Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull more rdma updates from Doug Ledford:
 "Round two of 4.6 merge window patches.

  This is a monster pull request.  I held off on the hfi1 driver updates
  (the hfi1 driver is intimately tied to the qib driver and the new
  rdmavt software library that was created to help both of them) in my
  first pull request.  The hfi1/qib/rdmavt update is probably 90% of
  this pull request.  The hfi1 driver is being left in staging so that
  it can be fixed up in regards to the API that Al and yourself didn't
  like.  Intel has agreed to do the work, but in the meantime, this
  clears out 300+ patches in the backlog queue and brings my tree and
  their tree closer to sync.

  This also includes about 10 patches to the core and a few to mlx5 to
  create an infrastructure for configuring SRIOV ports on IB devices.
  That series includes one patch to the net core that we sent to netdev@
  and Dave Miller with each of the three revisions to the series.  We
  didn't get any response to the patch, so we took that as implicit
  approval.

  Finally, this series includes Intel's new iWARP driver for their x722
  cards.  It's not nearly the beast as the hfi1 driver.  It also has a
  linux-next merge issue, but that has been resolved and it now passes
  just fine.

  Summary:

   - A few minor core fixups needed for the next patch series

   - The IB SRIOV series.  This has bounced around for several versions.
     Of note is the fact that the first patch in this series effects the
     net core.  It was directed to netdev and DaveM for each iteration
     of the series (three versions total).  Dave did not object, but did
     not respond either.  I've taken this as permission to move forward
     with the series.

   - The new Intel X722 iWARP driver

   - A huge set of updates to the Intel hfi1 driver.  Of particular
     interest here is that we have left the driver in staging since it
     still has an API that people object to.  Intel is working on a fix,
     but getting these patches in now helps keep me sane as the upstream
     and Intel's trees were over 300 patches apart"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (362 commits)
  IB/ipoib: Allow mcast packets from other VFs
  IB/mlx5: Implement callbacks for manipulating VFs
  net/mlx5_core: Implement modify HCA vport command
  net/mlx5_core: Add VF param when querying vport counter
  IB/ipoib: Add ndo operations for configuring VFs
  IB/core: Add interfaces to control VF attributes
  IB/core: Support accessing SA in virtualized environment
  IB/core: Add subnet prefix to port info
  IB/mlx5: Fix decision on using MAD_IFC
  net/core: Add support for configuring VF GUIDs
  IB/{core, ulp} Support above 32 possible device capability flags
  IB/core: Replace setting the zero values in ib_uverbs_ex_query_device
  net/mlx5_core: Introduce offload arithmetic hardware capabilities
  net/mlx5_core: Refactor device capability function
  net/mlx5_core: Fix caching ATOMIC endian mode capability
  ib_srpt: fix a WARN_ON() message
  i40iw: Replace the obsolete crypto hash interface with shash
  IB/hfi1: Add SDMA cache eviction algorithm
  IB/hfi1: Switch to using the pin query function
  IB/hfi1: Specify mm when releasing pages
  ...

175 files changed, 47756 insertions, 15197 deletions

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index 8a8440c0eed1..6425c0e5d18a 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -68,6 +68,7 @@ source "drivers/infiniband/hw/mthca/Kconfig"
 source "drivers/infiniband/hw/qib/Kconfig"
 source "drivers/infiniband/hw/cxgb3/Kconfig"
 source "drivers/infiniband/hw/cxgb4/Kconfig"
+source "drivers/infiniband/hw/i40iw/Kconfig"
 source "drivers/infiniband/hw/mlx4/Kconfig"
 source "drivers/infiniband/hw/mlx5/Kconfig"
 source "drivers/infiniband/hw/nes/Kconfig"
@@ -82,4 +83,6 @@ source "drivers/infiniband/ulp/srpt/Kconfig"
 source "drivers/infiniband/ulp/iser/Kconfig"
 source "drivers/infiniband/ulp/isert/Kconfig"
 
+source "drivers/infiniband/sw/rdmavt/Kconfig"
+
 endif # INFINIBAND
diff --git a/drivers/infiniband/Makefile b/drivers/infiniband/Makefile
index dc21836b5a8d..fad0b44c356f 100644
--- a/drivers/infiniband/Makefile
+++ b/drivers/infiniband/Makefile
@@ -1,3 +1,4 @@
 obj-$(CONFIG_INFINIBAND)		+= core/
 obj-$(CONFIG_INFINIBAND)		+= hw/
 obj-$(CONFIG_INFINIBAND)		+= ulp/
+obj-$(CONFIG_INFINIBAND)		+= sw/
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 270c7ff6cba7..10979844026a 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -650,10 +650,23 @@ int ib_query_port(struct ib_device *device,
 		  u8 port_num,
 		  struct ib_port_attr *port_attr)
 {
+	union ib_gid gid;
+	int err;
+
 	if (port_num < rdma_start_port(device) || port_num > rdma_end_port(device))
 		return -EINVAL;
 
-	return device->query_port(device, port_num, port_attr);
+	memset(port_attr, 0, sizeof(*port_attr));
+	err = device->query_port(device, port_num, port_attr);
+	if (err || port_attr->subnet_prefix)
+		return err;
+
+	err = ib_query_gid(device, port_num, 0, &gid, NULL);
+	if (err)
+		return err;
+
+	port_attr->subnet_prefix = be64_to_cpu(gid.global.subnet_prefix);
+	return 0;
 }
 EXPORT_SYMBOL(ib_query_port);
 
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index b5656a2298ee..8a09c0fb268d 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -885,6 +885,11 @@ static void update_sm_ah(struct work_struct *work)
 	ah_attr.dlid     = port_attr.sm_lid;
 	ah_attr.sl       = port_attr.sm_sl;
 	ah_attr.port_num = port->port_num;
+	if (port_attr.grh_required) {
+		ah_attr.ah_flags = IB_AH_GRH;
+		ah_attr.grh.dgid.global.subnet_prefix = cpu_to_be64(port_attr.subnet_prefix);
+		ah_attr.grh.dgid.global.interface_id = cpu_to_be64(IB_SA_WELL_KNOWN_GUID);
+	}
 
 	new_ah->ah = ib_create_ah(port->agent->qp->pd, &ah_attr);
 	if (IS_ERR(new_ah->ah)) {
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 3638c787cb7c..6fdc7ecdaca0 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -402,7 +402,7 @@ static void copy_query_dev_fields(struct ib_uverbs_file *file,
 	resp->hw_ver		= attr->hw_ver;
 	resp->max_qp		= attr->max_qp;
 	resp->max_qp_wr		= attr->max_qp_wr;
-	resp->device_cap_flags	= attr->device_cap_flags;
+	resp->device_cap_flags	= lower_32_bits(attr->device_cap_flags);
 	resp->max_sge		= attr->max_sge;
 	resp->max_sge_rd	= attr->max_sge_rd;
 	resp->max_cq		= attr->max_cq;
@@ -3600,9 +3600,9 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 			      struct ib_udata *ucore,
 			      struct ib_udata *uhw)
 {
-	struct ib_uverbs_ex_query_device_resp resp;
+	struct ib_uverbs_ex_query_device_resp resp = { {0} };
 	struct ib_uverbs_ex_query_device  cmd;
-	struct ib_device_attr attr;
+	struct ib_device_attr attr = {0};
 	int err;
 
 	if (ucore->inlen < sizeof(cmd))
@@ -3623,14 +3623,11 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 	if (ucore->outlen < resp.response_length)
 		return -ENOSPC;
 
-	memset(&attr, 0, sizeof(attr));
-
 	err = ib_dev->query_device(ib_dev, &attr, uhw);
 	if (err)
 		return err;
 
 	copy_query_dev_fields(file, ib_dev, &resp.base, &attr);
-	resp.comp_mask = 0;
 
 	if (ucore->outlen < resp.response_length + sizeof(resp.odp_caps))
 		goto end;
@@ -3643,9 +3640,6 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 		attr.odp_caps.per_transport_caps.uc_odp_caps;
 	resp.odp_caps.per_transport_caps.ud_odp_caps =
 		attr.odp_caps.per_transport_caps.ud_odp_caps;
-	resp.odp_caps.reserved = 0;
-#else
-	memset(&resp.odp_caps, 0, sizeof(resp.odp_caps));
 #endif
 	resp.response_length += sizeof(resp.odp_caps);
 
@@ -3663,8 +3657,5 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 
 end:
 	err = ib_copy_to_udata(ucore, &resp, resp.response_length);
-	if (err)
-		return err;
-
-	return 0;
+	return err;
 }
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index 5cd1e3987f2b..15b8adbf39c0 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -1551,6 +1551,46 @@ int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
 }
 EXPORT_SYMBOL(ib_check_mr_status);
 
+int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port,
+			 int state)
+{
+	if (!device->set_vf_link_state)
+		return -ENOSYS;
+
+	return device->set_vf_link_state(device, vf, port, state);
+}
+EXPORT_SYMBOL(ib_set_vf_link_state);
+
+int ib_get_vf_config(struct ib_device *device, int vf, u8 port,
+		     struct ifla_vf_info *info)
+{
+	if (!device->get_vf_config)
+		return -ENOSYS;
+
+	return device->get_vf_config(device, vf, port, info);
+}
+EXPORT_SYMBOL(ib_get_vf_config);
+
+int ib_get_vf_stats(struct ib_device *device, int vf, u8 port,
+		    struct ifla_vf_stats *stats)
+{
+	if (!device->get_vf_stats)
+		return -ENOSYS;
+
+	return device->get_vf_stats(device, vf, port, stats);
+}
+EXPORT_SYMBOL(ib_get_vf_stats);
+
+int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid,
+		   int type)
+{
+	if (!device->set_vf_guid)
+		return -ENOSYS;
+
+	return device->set_vf_guid(device, vf, port, guid, type);
+}
+EXPORT_SYMBOL(ib_set_vf_guid);
+
 /**
  * ib_map_mr_sg() - Map the largest prefix of a dma mapped SG list
  *     and set it the memory region.
diff --git a/drivers/infiniband/hw/Makefile b/drivers/infiniband/hw/Makefile
index aded2a5cc2d5..c7ad0a4c8b15 100644
--- a/drivers/infiniband/hw/Makefile
+++ b/drivers/infiniband/hw/Makefile
@@ -2,6 +2,7 @@ obj-$(CONFIG_INFINIBAND_MTHCA)		+= mthca/
 obj-$(CONFIG_INFINIBAND_QIB)		+= qib/
 obj-$(CONFIG_INFINIBAND_CXGB3)		+= cxgb3/
 obj-$(CONFIG_INFINIBAND_CXGB4)		+= cxgb4/
+obj-$(CONFIG_INFINIBAND_I40IW)		+= i40iw/
 obj-$(CONFIG_MLX4_INFINIBAND)		+= mlx4/
 obj-$(CONFIG_MLX5_INFINIBAND)		+= mlx5/
 obj-$(CONFIG_INFINIBAND_NES)		+= nes/
diff --git a/drivers/infiniband/hw/i40iw/Kconfig b/drivers/infiniband/hw/i40iw/Kconfig
new file mode 100644
index 000000000000..6e7d27a14061
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/Kconfig
@@ -0,0 +1,7 @@
+config INFINIBAND_I40IW
+	tristate "Intel(R) Ethernet X722 iWARP Driver"
+	depends on INET && I40E
+	select GENERIC_ALLOCATOR
+	---help---
+	Intel(R) Ethernet X722 iWARP Driver
+	INET && I40IW && INFINIBAND && I40E
diff --git a/drivers/infiniband/hw/i40iw/Makefile b/drivers/infiniband/hw/i40iw/Makefile
new file mode 100644
index 000000000000..90068c03d217
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/Makefile
@@ -0,0 +1,9 @@
+ccflags-y :=  -Idrivers/net/ethernet/intel/i40e
+
+obj-$(CONFIG_INFINIBAND_I40IW) += i40iw.o
+
+i40iw-objs :=\
+               i40iw_cm.o i40iw_ctrl.o \
+               i40iw_hmc.o i40iw_hw.o i40iw_main.o  \
+               i40iw_pble.o i40iw_puda.o i40iw_uk.o i40iw_utils.o \
+               i40iw_verbs.o i40iw_virtchnl.o i40iw_vf.o
diff --git a/drivers/infiniband/hw/i40iw/i40iw.h b/drivers/infiniband/hw/i40iw/i40iw.h
new file mode 100644
index 000000000000..819767681445
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw.h
@@ -0,0 +1,570 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_IW_H
+#define I40IW_IW_H
+#include <linux/netdevice.h>
+#include <linux/inetdevice.h>
+#include <linux/spinlock.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/workqueue.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+#include <linux/crc32c.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_pack.h>
+#include <rdma/rdma_cm.h>
+#include <rdma/iw_cm.h>
+#include <rdma/iw_portmap.h>
+#include <rdma/rdma_netlink.h>
+#include <crypto/hash.h>
+
+#include "i40iw_status.h"
+#include "i40iw_osdep.h"
+#include "i40iw_d.h"
+#include "i40iw_hmc.h"
+
+#include <i40e_client.h>
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_ucontext.h"
+#include "i40iw_pble.h"
+#include "i40iw_verbs.h"
+#include "i40iw_cm.h"
+#include "i40iw_user.h"
+#include "i40iw_puda.h"
+
+#define I40IW_FW_VERSION  2
+#define I40IW_HW_VERSION  2
+
+#define I40IW_ARP_ADD     1
+#define I40IW_ARP_DELETE  2
+#define I40IW_ARP_RESOLVE 3
+
+#define I40IW_MACIP_ADD     1
+#define I40IW_MACIP_DELETE  2
+
+#define IW_CCQ_SIZE         (I40IW_CQP_SW_SQSIZE_2048 + 1)
+#define IW_CEQ_SIZE         2048
+#define IW_AEQ_SIZE         2048
+
+#define RX_BUF_SIZE            (1536 + 8)
+#define IW_REG0_SIZE           (4 * 1024)
+#define IW_TX_TIMEOUT          (6 * HZ)
+#define IW_FIRST_QPN           1
+#define IW_SW_CONTEXT_ALIGN    1024
+
+#define MAX_DPC_ITERATIONS		128
+
+#define I40IW_EVENT_TIMEOUT		100000
+#define I40IW_VCHNL_EVENT_TIMEOUT	100000
+
+#define	I40IW_NO_VLAN			0xffff
+#define	I40IW_NO_QSET			0xffff
+
+/* access to mcast filter list */
+#define IW_ADD_MCAST false
+#define IW_DEL_MCAST true
+
+#define I40IW_DRV_OPT_ENABLE_MPA_VER_0     0x00000001
+#define I40IW_DRV_OPT_DISABLE_MPA_CRC      0x00000002
+#define I40IW_DRV_OPT_DISABLE_FIRST_WRITE  0x00000004
+#define I40IW_DRV_OPT_DISABLE_INTF         0x00000008
+#define I40IW_DRV_OPT_ENABLE_MSI           0x00000010
+#define I40IW_DRV_OPT_DUAL_LOGICAL_PORT    0x00000020
+#define I40IW_DRV_OPT_NO_INLINE_DATA       0x00000080
+#define I40IW_DRV_OPT_DISABLE_INT_MOD      0x00000100
+#define I40IW_DRV_OPT_DISABLE_VIRT_WQ      0x00000200
+#define I40IW_DRV_OPT_ENABLE_PAU           0x00000400
+#define I40IW_DRV_OPT_MCAST_LOGPORT_MAP    0x00000800
+
+#define IW_HMC_OBJ_TYPE_NUM ARRAY_SIZE(iw_hmc_obj_types)
+#define IW_CFG_FPM_QP_COUNT		32768
+
+#define I40IW_MTU_TO_MSS		40
+#define I40IW_DEFAULT_MSS		1460
+
+struct i40iw_cqp_compl_info {
+	u32 op_ret_val;
+	u16 maj_err_code;
+	u16 min_err_code;
+	bool error;
+	u8 op_code;
+};
+
+#define i40iw_pr_err(fmt, args ...) pr_err("%s: "fmt, __func__, ## args)
+
+#define i40iw_pr_info(fmt, args ...) pr_info("%s: " fmt, __func__, ## args)
+
+#define i40iw_pr_warn(fmt, args ...) pr_warn("%s: " fmt, __func__, ## args)
+
+struct i40iw_cqp_request {
+	struct cqp_commands_info info;
+	wait_queue_head_t waitq;
+	struct list_head list;
+	atomic_t refcount;
+	void (*callback_fcn)(struct i40iw_cqp_request*, u32);
+	void *param;
+	struct i40iw_cqp_compl_info compl_info;
+	bool waiting;
+	bool request_done;
+	bool dynamic;
+};
+
+struct i40iw_cqp {
+	struct i40iw_sc_cqp sc_cqp;
+	spinlock_t req_lock; /*cqp request list */
+	wait_queue_head_t waitq;
+	struct i40iw_dma_mem sq;
+	struct i40iw_dma_mem host_ctx;
+	u64 *scratch_array;
+	struct i40iw_cqp_request *cqp_requests;
+	struct list_head cqp_avail_reqs;
+	struct list_head cqp_pending_reqs;
+};
+
+struct i40iw_device;
+
+struct i40iw_ccq {
+	struct i40iw_sc_cq sc_cq;
+	spinlock_t lock; /* ccq control */
+	wait_queue_head_t waitq;
+	struct i40iw_dma_mem mem_cq;
+	struct i40iw_dma_mem shadow_area;
+};
+
+struct i40iw_ceq {
+	struct i40iw_sc_ceq sc_ceq;
+	struct i40iw_dma_mem mem;
+	u32 irq;
+	u32 msix_idx;
+	struct i40iw_device *iwdev;
+	struct tasklet_struct dpc_tasklet;
+};
+
+struct i40iw_aeq {
+	struct i40iw_sc_aeq sc_aeq;
+	struct i40iw_dma_mem mem;
+};
+
+struct i40iw_arp_entry {
+	u32 ip_addr[4];
+	u8 mac_addr[ETH_ALEN];
+};
+
+enum init_completion_state {
+	INVALID_STATE = 0,
+	INITIAL_STATE,
+	CQP_CREATED,
+	HMC_OBJS_CREATED,
+	PBLE_CHUNK_MEM,
+	CCQ_CREATED,
+	AEQ_CREATED,
+	CEQ_CREATED,
+	ILQ_CREATED,
+	IEQ_CREATED,
+	INET_NOTIFIER,
+	IP_ADDR_REGISTERED,
+	RDMA_DEV_REGISTERED
+};
+
+struct i40iw_msix_vector {
+	u32 idx;
+	u32 irq;
+	u32 cpu_affinity;
+	u32 ceq_id;
+};
+
+#define I40IW_MSIX_TABLE_SIZE   65
+
+struct virtchnl_work {
+	struct work_struct work;
+	union {
+		struct i40iw_cqp_request *cqp_request;
+		struct i40iw_virtchnl_work_info work_info;
+	};
+};
+
+struct i40e_qvlist_info;
+
+struct i40iw_device {
+	struct i40iw_ib_device *iwibdev;
+	struct net_device *netdev;
+	wait_queue_head_t vchnl_waitq;
+	struct i40iw_sc_dev sc_dev;
+	struct i40iw_handler *hdl;
+	struct i40e_info *ldev;
+	struct i40e_client *client;
+	struct i40iw_hw hw;
+	struct i40iw_cm_core cm_core;
+	unsigned long *mem_resources;
+	unsigned long *allocated_qps;
+	unsigned long *allocated_cqs;
+	unsigned long *allocated_mrs;
+	unsigned long *allocated_pds;
+	unsigned long *allocated_arps;
+	struct i40iw_qp **qp_table;
+	bool msix_shared;
+	u32 msix_count;
+	struct i40iw_msix_vector *iw_msixtbl;
+	struct i40e_qvlist_info *iw_qvlist;
+
+	struct i40iw_hmc_pble_rsrc *pble_rsrc;
+	struct i40iw_arp_entry *arp_table;
+	struct i40iw_cqp cqp;
+	struct i40iw_ccq ccq;
+	u32 ceqs_count;
+	struct i40iw_ceq *ceqlist;
+	struct i40iw_aeq aeq;
+	u32 arp_table_size;
+	u32 next_arp_index;
+	spinlock_t resource_lock; /* hw resource access */
+	u32 vendor_id;
+	u32 vendor_part_id;
+	u32 of_device_registered;
+
+	u32 device_cap_flags;
+	unsigned long db_start;
+	u8 resource_profile;
+	u8 max_rdma_vfs;
+	u8 max_enabled_vfs;
+	u8 max_sge;
+	u8 iw_status;
+	u8 send_term_ok;
+	bool push_mode;		/* Initialized from parameter passed to driver */
+
+	/* x710 specific */
+	struct mutex pbl_mutex;
+	struct tasklet_struct dpc_tasklet;
+	struct workqueue_struct *virtchnl_wq;
+	struct virtchnl_work virtchnl_w[I40IW_MAX_PE_ENABLED_VF_COUNT];
+	struct i40iw_dma_mem obj_mem;
+	struct i40iw_dma_mem obj_next;
+	u8 *hmc_info_mem;
+	u32 sd_type;
+	struct workqueue_struct *param_wq;
+	atomic_t params_busy;
+	u32 mss;
+	enum init_completion_state init_state;
+	u16 mac_ip_table_idx;
+	atomic_t vchnl_msgs;
+	u32 max_mr;
+	u32 max_qp;
+	u32 max_cq;
+	u32 max_pd;
+	u32 next_qp;
+	u32 next_cq;
+	u32 next_pd;
+	u32 max_mr_size;
+	u32 max_qp_wr;
+	u32 max_cqe;
+	u32 mr_stagmask;
+	u32 mpa_version;
+	bool dcb;
+};
+
+struct i40iw_ib_device {
+	struct ib_device ibdev;
+	struct i40iw_device *iwdev;
+};
+
+struct i40iw_handler {
+	struct list_head list;
+	struct i40e_client *client;
+	struct i40iw_device device;
+	struct i40e_info ldev;
+};
+
+/**
+ * to_iwdev - get device
+ * @ibdev: ib device
+ **/
+static inline struct i40iw_device *to_iwdev(struct ib_device *ibdev)
+{
+	return container_of(ibdev, struct i40iw_ib_device, ibdev)->iwdev;
+}
+
+/**
+ * to_ucontext - get user context
+ * @ibucontext: ib user context
+ **/
+static inline struct i40iw_ucontext *to_ucontext(struct ib_ucontext *ibucontext)
+{
+	return container_of(ibucontext, struct i40iw_ucontext, ibucontext);
+}
+
+/**
+ * to_iwpd - get protection domain
+ * @ibpd: ib pd
+ **/
+static inline struct i40iw_pd *to_iwpd(struct ib_pd *ibpd)
+{
+	return container_of(ibpd, struct i40iw_pd, ibpd);
+}
+
+/**
+ * to_iwmr - get device memory region
+ * @ibdev: ib memory region
+ **/
+static inline struct i40iw_mr *to_iwmr(struct ib_mr *ibmr)
+{
+	return container_of(ibmr, struct i40iw_mr, ibmr);
+}
+
+/**
+ * to_iwmr_from_ibfmr - get device memory region
+ * @ibfmr: ib fmr
+ **/
+static inline struct i40iw_mr *to_iwmr_from_ibfmr(struct ib_fmr *ibfmr)
+{
+	return container_of(ibfmr, struct i40iw_mr, ibfmr);
+}
+
+/**
+ * to_iwmw - get device memory window
+ * @ibmw: ib memory window
+ **/
+static inline struct i40iw_mr *to_iwmw(struct ib_mw *ibmw)
+{
+	return container_of(ibmw, struct i40iw_mr, ibmw);
+}
+
+/**
+ * to_iwcq - get completion queue
+ * @ibcq: ib cqdevice
+ **/
+static inline struct i40iw_cq *to_iwcq(struct ib_cq *ibcq)
+{
+	return container_of(ibcq, struct i40iw_cq, ibcq);
+}
+
+/**
+ * to_iwqp - get device qp
+ * @ibqp: ib qp
+ **/
+static inline struct i40iw_qp *to_iwqp(struct ib_qp *ibqp)
+{
+	return container_of(ibqp, struct i40iw_qp, ibqp);
+}
+
+/* i40iw.c */
+void i40iw_add_ref(struct ib_qp *);
+void i40iw_rem_ref(struct ib_qp *);
+struct ib_qp *i40iw_get_qp(struct ib_device *, int);
+
+void i40iw_flush_wqes(struct i40iw_device *iwdev,
+		      struct i40iw_qp *qp);
+
+void i40iw_manage_arp_cache(struct i40iw_device *iwdev,
+			    unsigned char *mac_addr,
+			    __be32 *ip_addr,
+			    bool ipv4,
+			    u32 action);
+
+int i40iw_manage_apbvt(struct i40iw_device *iwdev,
+		       u16 accel_local_port,
+		       bool add_port);
+
+struct i40iw_cqp_request *i40iw_get_cqp_request(struct i40iw_cqp *cqp, bool wait);
+void i40iw_free_cqp_request(struct i40iw_cqp *cqp, struct i40iw_cqp_request *cqp_request);
+void i40iw_put_cqp_request(struct i40iw_cqp *cqp, struct i40iw_cqp_request *cqp_request);
+
+/**
+ * i40iw_alloc_resource - allocate a resource
+ * @iwdev: device pointer
+ * @resource_array: resource bit array:
+ * @max_resources: maximum resource number
+ * @req_resources_num: Allocated resource number
+ * @next: next free id
+ **/
+static inline int i40iw_alloc_resource(struct i40iw_device *iwdev,
+				       unsigned long *resource_array,
+				       u32 max_resources,
+				       u32 *req_resource_num,
+				       u32 *next)
+{
+	u32 resource_num;
+	unsigned long flags;
+
+	spin_lock_irqsave(&iwdev->resource_lock, flags);
+	resource_num = find_next_zero_bit(resource_array, max_resources, *next);
+	if (resource_num >= max_resources) {
+		resource_num = find_first_zero_bit(resource_array, max_resources);
+		if (resource_num >= max_resources) {
+			spin_unlock_irqrestore(&iwdev->resource_lock, flags);
+			return -EOVERFLOW;
+		}
+	}
+	set_bit(resource_num, resource_array);
+	*next = resource_num + 1;
+	if (*next == max_resources)
+		*next = 0;
+	spin_unlock_irqrestore(&iwdev->resource_lock, flags);
+	*req_resource_num = resource_num;
+
+	return 0;
+}
+
+/**
+ * i40iw_is_resource_allocated - detrmine if resource is
+ * allocated
+ * @iwdev: device pointer
+ * @resource_array: resource array for the resource_num
+ * @resource_num: resource number to check
+ **/
+static inline bool i40iw_is_resource_allocated(struct i40iw_device *iwdev,
+					       unsigned long *resource_array,
+					       u32 resource_num)
+{
+	bool bit_is_set;
+	unsigned long flags;
+
+	spin_lock_irqsave(&iwdev->resource_lock, flags);
+
+	bit_is_set = test_bit(resource_num, resource_array);
+	spin_unlock_irqrestore(&iwdev->resource_lock, flags);
+
+	return bit_is_set;
+}
+
+/**
+ * i40iw_free_resource - free a resource
+ * @iwdev: device pointer
+ * @resource_array: resource array for the resource_num
+ * @resource_num: resource number to free
+ **/
+static inline void i40iw_free_resource(struct i40iw_device *iwdev,
+				       unsigned long *resource_array,
+				       u32 resource_num)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&iwdev->resource_lock, flags);
+	clear_bit(resource_num, resource_array);
+	spin_unlock_irqrestore(&iwdev->resource_lock, flags);
+}
+
+/**
+ * to_iwhdl - Get the handler from the device pointer
+ * @iwdev: device pointer
+ **/
+static inline struct i40iw_handler *to_iwhdl(struct i40iw_device *iw_dev)
+{
+	return container_of(iw_dev, struct i40iw_handler, device);
+}
+
+struct i40iw_handler *i40iw_find_netdev(struct net_device *netdev);
+
+/**
+ * iw_init_resources -
+ */
+u32 i40iw_initialize_hw_resources(struct i40iw_device *iwdev);
+
+int i40iw_register_rdma_device(struct i40iw_device *iwdev);
+void i40iw_port_ibevent(struct i40iw_device *iwdev);
+int i40iw_cm_disconn(struct i40iw_qp *);
+void i40iw_cm_disconn_worker(void *);
+int mini_cm_recv_pkt(struct i40iw_cm_core *, struct i40iw_device *,
+		     struct sk_buff *);
+
+enum i40iw_status_code i40iw_handle_cqp_op(struct i40iw_device *iwdev,
+					   struct i40iw_cqp_request *cqp_request);
+enum i40iw_status_code i40iw_add_mac_addr(struct i40iw_device *iwdev,
+					  u8 *mac_addr, u8 *mac_index);
+int i40iw_modify_qp(struct ib_qp *, struct ib_qp_attr *, int, struct ib_udata *);
+
+void i40iw_rem_pdusecount(struct i40iw_pd *iwpd, struct i40iw_device *iwdev);
+void i40iw_add_pdusecount(struct i40iw_pd *iwpd);
+void i40iw_hw_modify_qp(struct i40iw_device *iwdev, struct i40iw_qp *iwqp,
+			struct i40iw_modify_qp_info *info, bool wait);
+
+enum i40iw_status_code i40iw_manage_qhash(struct i40iw_device *iwdev,
+					  struct i40iw_cm_info *cminfo,
+					  enum i40iw_quad_entry_type etype,
+					  enum i40iw_quad_hash_manage_type mtype,
+					  void *cmnode,
+					  bool wait);
+void i40iw_receive_ilq(struct i40iw_sc_dev *dev, struct i40iw_puda_buf *rbuf);
+void i40iw_free_sqbuf(struct i40iw_sc_dev *dev, void *bufp);
+void i40iw_free_qp_resources(struct i40iw_device *iwdev,
+			     struct i40iw_qp *iwqp,
+			     u32 qp_num);
+enum i40iw_status_code i40iw_obj_aligned_mem(struct i40iw_device *iwdev,
+					     struct i40iw_dma_mem *memptr,
+					     u32 size, u32 mask);
+
+void i40iw_request_reset(struct i40iw_device *iwdev);
+void i40iw_destroy_rdma_device(struct i40iw_ib_device *iwibdev);
+void i40iw_setup_cm_core(struct i40iw_device *iwdev);
+void i40iw_cleanup_cm_core(struct i40iw_cm_core *cm_core);
+void i40iw_process_ceq(struct i40iw_device *, struct i40iw_ceq *iwceq);
+void i40iw_process_aeq(struct i40iw_device *);
+void i40iw_next_iw_state(struct i40iw_qp *iwqp,
+			 u8 state, u8 del_hash,
+			 u8 term, u8 term_len);
+int i40iw_send_syn(struct i40iw_cm_node *cm_node, u32 sendack);
+struct i40iw_cm_node *i40iw_find_node(struct i40iw_cm_core *cm_core,
+				      u16 rem_port,
+				      u32 *rem_addr,
+				      u16 loc_port,
+				      u32 *loc_addr,
+				      bool add_refcnt);
+
+enum i40iw_status_code i40iw_hw_flush_wqes(struct i40iw_device *iwdev,
+					   struct i40iw_sc_qp *qp,
+					   struct i40iw_qp_flush_info *info,
+					   bool wait);
+
+void i40iw_copy_ip_ntohl(u32 *dst, u32 *src);
+struct ib_mr *i40iw_reg_phys_mr(struct ib_pd *ib_pd,
+				u64 addr,
+				u64 size,
+				int acc,
+				u64 *iova_start);
+
+int i40iw_inetaddr_event(struct notifier_block *notifier,
+			 unsigned long event,
+			 void *ptr);
+int i40iw_inet6addr_event(struct notifier_block *notifier,
+			  unsigned long event,
+			  void *ptr);
+int i40iw_net_event(struct notifier_block *notifier,
+		    unsigned long event,
+		    void *ptr);
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_cm.c b/drivers/infiniband/hw/i40iw/i40iw_cm.c
new file mode 100644
index 000000000000..92745d755272
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_cm.c
@@ -0,0 +1,4141 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include <linux/atomic.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/init.h>
+#include <linux/if_arp.h>
+#include <linux/if_vlan.h>
+#include <linux/notifier.h>
+#include <linux/net.h>
+#include <linux/types.h>
+#include <linux/timer.h>
+#include <linux/time.h>
+#include <linux/delay.h>
+#include <linux/etherdevice.h>
+#include <linux/netdevice.h>
+#include <linux/random.h>
+#include <linux/list.h>
+#include <linux/threads.h>
+#include <linux/highmem.h>
+#include <net/arp.h>
+#include <net/ndisc.h>
+#include <net/neighbour.h>
+#include <net/route.h>
+#include <net/addrconf.h>
+#include <net/ip6_route.h>
+#include <net/ip_fib.h>
+#include <net/tcp.h>
+#include <asm/checksum.h>
+
+#include "i40iw.h"
+
+static void i40iw_rem_ref_cm_node(struct i40iw_cm_node *);
+static void i40iw_cm_post_event(struct i40iw_cm_event *event);
+static void i40iw_disconnect_worker(struct work_struct *work);
+
+/**
+ * i40iw_free_sqbuf - put back puda buffer if refcount = 0
+ * @dev: FPK device
+ * @buf: puda buffer to free
+ */
+void i40iw_free_sqbuf(struct i40iw_sc_dev *dev, void *bufp)
+{
+	struct i40iw_puda_buf *buf = (struct i40iw_puda_buf *)bufp;
+	struct i40iw_puda_rsrc *ilq = dev->ilq;
+
+	if (!atomic_dec_return(&buf->refcount))
+		i40iw_puda_ret_bufpool(ilq, buf);
+}
+
+/**
+ * i40iw_derive_hw_ird_setting - Calculate IRD
+ *
+ * @cm_ird: IRD of connection's node
+ *
+ * The ird from the connection is rounded to a supported HW
+ * setting (2,8,32,64) and then encoded for ird_size field of
+ * qp_ctx
+ */
+static u8 i40iw_derive_hw_ird_setting(u16 cm_ird)
+{
+	u8 encoded_ird_size;
+	u8 pof2_cm_ird = 1;
+
+	/* round-off to next powerof2 */
+	while (pof2_cm_ird < cm_ird)
+		pof2_cm_ird *= 2;
+
+	/* ird_size field is encoded in qp_ctx */
+	switch (pof2_cm_ird) {
+	case I40IW_HW_IRD_SETTING_64:
+		encoded_ird_size = 3;
+		break;
+	case I40IW_HW_IRD_SETTING_32:
+	case I40IW_HW_IRD_SETTING_16:
+		encoded_ird_size = 2;
+		break;
+	case I40IW_HW_IRD_SETTING_8:
+	case I40IW_HW_IRD_SETTING_4:
+		encoded_ird_size = 1;
+		break;
+	case I40IW_HW_IRD_SETTING_2:
+	default:
+		encoded_ird_size = 0;
+		break;
+	}
+	return encoded_ird_size;
+}
+
+/**
+ * i40iw_record_ird_ord - Record IRD/ORD passed in
+ * @cm_node: connection's node
+ * @conn_ird: connection IRD
+ * @conn_ord: connection ORD
+ */
+static void i40iw_record_ird_ord(struct i40iw_cm_node *cm_node, u16 conn_ird, u16 conn_ord)
+{
+	if (conn_ird > I40IW_MAX_IRD_SIZE)
+		conn_ird = I40IW_MAX_IRD_SIZE;
+
+	if (conn_ord > I40IW_MAX_ORD_SIZE)
+		conn_ord = I40IW_MAX_ORD_SIZE;
+
+	cm_node->ird_size = conn_ird;
+	cm_node->ord_size = conn_ord;
+}
+
+/**
+ * i40iw_copy_ip_ntohl - change network to host ip
+ * @dst: host ip
+ * @src: big endian
+ */
+void i40iw_copy_ip_ntohl(u32 *dst, __be32 *src)
+{
+	*dst++ = ntohl(*src++);
+	*dst++ = ntohl(*src++);
+	*dst++ = ntohl(*src++);
+	*dst = ntohl(*src);
+}
+
+/**
+ * i40iw_copy_ip_htonl - change host addr to network ip
+ * @dst: host ip
+ * @src: little endian
+ */
+static inline void i40iw_copy_ip_htonl(__be32 *dst, u32 *src)
+{
+	*dst++ = htonl(*src++);
+	*dst++ = htonl(*src++);
+	*dst++ = htonl(*src++);
+	*dst = htonl(*src);
+}
+
+/**
+ * i40iw_fill_sockaddr4 - get addr info for passive connection
+ * @cm_node: connection's node
+ * @event: upper layer's cm event
+ */
+static inline void i40iw_fill_sockaddr4(struct i40iw_cm_node *cm_node,
+					struct iw_cm_event *event)
+{
+	struct sockaddr_in *laddr = (struct sockaddr_in *)&event->local_addr;
+	struct sockaddr_in *raddr = (struct sockaddr_in *)&event->remote_addr;
+
+	laddr->sin_family = AF_INET;
+	raddr->sin_family = AF_INET;
+
+	laddr->sin_port = htons(cm_node->loc_port);
+	raddr->sin_port = htons(cm_node->rem_port);
+
+	laddr->sin_addr.s_addr = htonl(cm_node->loc_addr[0]);
+	raddr->sin_addr.s_addr = htonl(cm_node->rem_addr[0]);
+}
+
+/**
+ * i40iw_fill_sockaddr6 - get ipv6 addr info for passive side
+ * @cm_node: connection's node
+ * @event: upper layer's cm event
+ */
+static inline void i40iw_fill_sockaddr6(struct i40iw_cm_node *cm_node,
+					struct iw_cm_event *event)
+{
+	struct sockaddr_in6 *laddr6 = (struct sockaddr_in6 *)&event->local_addr;
+	struct sockaddr_in6 *raddr6 = (struct sockaddr_in6 *)&event->remote_addr;
+
+	laddr6->sin6_family = AF_INET6;
+	raddr6->sin6_family = AF_INET6;
+
+	laddr6->sin6_port = htons(cm_node->loc_port);
+	raddr6->sin6_port = htons(cm_node->rem_port);
+
+	i40iw_copy_ip_htonl(laddr6->sin6_addr.in6_u.u6_addr32,
+			    cm_node->loc_addr);
+	i40iw_copy_ip_htonl(raddr6->sin6_addr.in6_u.u6_addr32,
+			    cm_node->rem_addr);
+}
+
+/**
+ * i40iw_get_addr_info
+ * @cm_node: contains ip/tcp info
+ * @cm_info: to get a copy of the cm_node ip/tcp info
+*/
+static void i40iw_get_addr_info(struct i40iw_cm_node *cm_node,
+				struct i40iw_cm_info *cm_info)
+{
+	cm_info->ipv4 = cm_node->ipv4;
+	cm_info->vlan_id = cm_node->vlan_id;
+	memcpy(cm_info->loc_addr, cm_node->loc_addr, sizeof(cm_info->loc_addr));
+	memcpy(cm_info->rem_addr, cm_node->rem_addr, sizeof(cm_info->rem_addr));
+	cm_info->loc_port = cm_node->loc_port;
+	cm_info->rem_port = cm_node->rem_port;
+}
+
+/**
+ * i40iw_get_cmevent_info - for cm event upcall
+ * @cm_node: connection's node
+ * @cm_id: upper layers cm struct for the event
+ * @event: upper layer's cm event
+ */
+static inline void i40iw_get_cmevent_info(struct i40iw_cm_node *cm_node,
+					  struct iw_cm_id *cm_id,
+					  struct iw_cm_event *event)
+{
+	memcpy(&event->local_addr, &cm_id->m_local_addr,
+	       sizeof(event->local_addr));
+	memcpy(&event->remote_addr, &cm_id->m_remote_addr,
+	       sizeof(event->remote_addr));
+	if (cm_node) {
+		event->private_data = (void *)cm_node->pdata_buf;
+		event->private_data_len = (u8)cm_node->pdata.size;
+		event->ird = cm_node->ird_size;
+		event->ord = cm_node->ord_size;
+	}
+}
+
+/**
+ * i40iw_send_cm_event - upcall cm's event handler
+ * @cm_node: connection's node
+ * @cm_id: upper layer's cm info struct
+ * @type: Event type to indicate
+ * @status: status for the event type
+ */
+static int i40iw_send_cm_event(struct i40iw_cm_node *cm_node,
+			       struct iw_cm_id *cm_id,
+			       enum iw_cm_event_type type,
+			       int status)
+{
+	struct iw_cm_event event;
+
+	memset(&event, 0, sizeof(event));
+	event.event = type;
+	event.status = status;
+	switch (type) {
+	case IW_CM_EVENT_CONNECT_REQUEST:
+		if (cm_node->ipv4)
+			i40iw_fill_sockaddr4(cm_node, &event);
+		else
+			i40iw_fill_sockaddr6(cm_node, &event);
+		event.provider_data = (void *)cm_node;
+		event.private_data = (void *)cm_node->pdata_buf;
+		event.private_data_len = (u8)cm_node->pdata.size;
+		break;
+	case IW_CM_EVENT_CONNECT_REPLY:
+		i40iw_get_cmevent_info(cm_node, cm_id, &event);
+		break;
+	case IW_CM_EVENT_ESTABLISHED:
+		event.ird = cm_node->ird_size;
+		event.ord = cm_node->ord_size;
+		break;
+	case IW_CM_EVENT_DISCONNECT:
+		break;
+	case IW_CM_EVENT_CLOSE:
+		break;
+	default:
+		i40iw_pr_err("event type received type = %d\n", type);
+		return -1;
+	}
+	return cm_id->event_handler(cm_id, &event);
+}
+
+/**
+ * i40iw_create_event - create cm event
+ * @cm_node: connection's node
+ * @type: Event type to generate
+ */
+static struct i40iw_cm_event *i40iw_create_event(struct i40iw_cm_node *cm_node,
+						 enum i40iw_cm_event_type type)
+{
+	struct i40iw_cm_event *event;
+
+	if (!cm_node->cm_id)
+		return NULL;
+
+	event = kzalloc(sizeof(*event), GFP_ATOMIC);
+
+	if (!event)
+		return NULL;
+
+	event->type = type;
+	event->cm_node = cm_node;
+	memcpy(event->cm_info.rem_addr, cm_node->rem_addr, sizeof(event->cm_info.rem_addr));
+	memcpy(event->cm_info.loc_addr, cm_node->loc_addr, sizeof(event->cm_info.loc_addr));
+	event->cm_info.rem_port = cm_node->rem_port;
+	event->cm_info.loc_port = cm_node->loc_port;
+	event->cm_info.cm_id = cm_node->cm_id;
+
+	i40iw_debug(cm_node->dev,
+		    I40IW_DEBUG_CM,
+		    "node=%p event=%p type=%u dst=%pI4 src=%pI4\n",
+		    cm_node,
+		    event,
+		    type,
+		    event->cm_info.loc_addr,
+		    event->cm_info.rem_addr);
+
+	i40iw_cm_post_event(event);
+	return event;
+}
+
+/**
+ * i40iw_free_retrans_entry - free send entry
+ * @cm_node: connection's node
+ */
+static void i40iw_free_retrans_entry(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_sc_dev *dev = cm_node->dev;
+	struct i40iw_timer_entry *send_entry;
+
+	send_entry = cm_node->send_entry;
+	if (send_entry) {
+		cm_node->send_entry = NULL;
+		i40iw_free_sqbuf(dev, (void *)send_entry->sqbuf);
+		kfree(send_entry);
+		atomic_dec(&cm_node->ref_count);
+	}
+}
+
+/**
+ * i40iw_cleanup_retrans_entry - free send entry with lock
+ * @cm_node: connection's node
+ */
+static void i40iw_cleanup_retrans_entry(struct i40iw_cm_node *cm_node)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cm_node->retrans_list_lock, flags);
+	i40iw_free_retrans_entry(cm_node);
+	spin_unlock_irqrestore(&cm_node->retrans_list_lock, flags);
+}
+
+static bool is_remote_ne020_or_chelsio(struct i40iw_cm_node *cm_node)
+{
+	if ((cm_node->rem_mac[0] == 0x0) &&
+	    (((cm_node->rem_mac[1] == 0x12) && (cm_node->rem_mac[2] == 0x55)) ||
+	     ((cm_node->rem_mac[1] == 0x07 && (cm_node->rem_mac[2] == 0x43)))))
+		return true;
+	return false;
+}
+
+/**
+ * i40iw_form_cm_frame - get a free packet and build frame
+ * @cm_node: connection's node ionfo to use in frame
+ * @options: pointer to options info
+ * @hdr: pointer mpa header
+ * @pdata: pointer to private data
+ * @flags:  indicates FIN or ACK
+ */
+static struct i40iw_puda_buf *i40iw_form_cm_frame(struct i40iw_cm_node *cm_node,
+						  struct i40iw_kmem_info *options,
+						  struct i40iw_kmem_info *hdr,
+						  struct i40iw_kmem_info *pdata,
+						  u8 flags)
+{
+	struct i40iw_puda_buf *sqbuf;
+	struct i40iw_sc_dev *dev = cm_node->dev;
+	u8 *buf;
+
+	struct tcphdr *tcph;
+	struct iphdr *iph;
+	struct ipv6hdr *ip6h;
+	struct ethhdr *ethh;
+	u16 packetsize;
+	u16 eth_hlen = ETH_HLEN;
+	u32 opts_len = 0;
+	u32 pd_len = 0;
+	u32 hdr_len = 0;
+
+	sqbuf = i40iw_puda_get_bufpool(dev->ilq);
+	if (!sqbuf)
+		return NULL;
+	buf = sqbuf->mem.va;
+
+	if (options)
+		opts_len = (u32)options->size;
+
+	if (hdr)
+		hdr_len = hdr->size;
+
+	if (pdata) {
+		pd_len = pdata->size;
+		if (!is_remote_ne020_or_chelsio(cm_node))
+			pd_len += MPA_ZERO_PAD_LEN;
+	}
+
+	if (cm_node->vlan_id < VLAN_TAG_PRESENT)
+		eth_hlen += 4;
+
+	if (cm_node->ipv4)
+		packetsize = sizeof(*iph) + sizeof(*tcph);
+	else
+		packetsize = sizeof(*ip6h) + sizeof(*tcph);
+	packetsize += opts_len + hdr_len + pd_len;
+
+	memset(buf, 0x00, eth_hlen + packetsize);
+
+	sqbuf->totallen = packetsize + eth_hlen;
+	sqbuf->maclen = eth_hlen;
+	sqbuf->tcphlen = sizeof(*tcph) + opts_len;
+	sqbuf->scratch = (void *)cm_node;
+
+	ethh = (struct ethhdr *)buf;
+	buf += eth_hlen;
+
+	if (cm_node->ipv4) {
+		sqbuf->ipv4 = true;
+
+		iph = (struct iphdr *)buf;
+		buf += sizeof(*iph);
+		tcph = (struct tcphdr *)buf;
+		buf += sizeof(*tcph);
+
+		ether_addr_copy(ethh->h_dest, cm_node->rem_mac);
+		ether_addr_copy(ethh->h_source, cm_node->loc_mac);
+		if (cm_node->vlan_id < VLAN_TAG_PRESENT) {
+			((struct vlan_ethhdr *)ethh)->h_vlan_proto = htons(ETH_P_8021Q);
+			((struct vlan_ethhdr *)ethh)->h_vlan_TCI = htons(cm_node->vlan_id);
+
+			((struct vlan_ethhdr *)ethh)->h_vlan_encapsulated_proto = htons(ETH_P_IP);
+		} else {
+			ethh->h_proto = htons(ETH_P_IP);
+		}
+
+		iph->version = IPVERSION;
+		iph->ihl = 5;	/* 5 * 4Byte words, IP headr len */
+		iph->tos = 0;
+		iph->tot_len = htons(packetsize);
+		iph->id = htons(++cm_node->tcp_cntxt.loc_id);
+
+		iph->frag_off = htons(0x4000);
+		iph->ttl = 0x40;
+		iph->protocol = IPPROTO_TCP;
+		iph->saddr = htonl(cm_node->loc_addr[0]);
+		iph->daddr = htonl(cm_node->rem_addr[0]);
+	} else {
+		sqbuf->ipv4 = false;
+		ip6h = (struct ipv6hdr *)buf;
+		buf += sizeof(*ip6h);
+		tcph = (struct tcphdr *)buf;
+		buf += sizeof(*tcph);
+
+		ether_addr_copy(ethh->h_dest, cm_node->rem_mac);
+		ether_addr_copy(ethh->h_source, cm_node->loc_mac);
+		if (cm_node->vlan_id < VLAN_TAG_PRESENT) {
+			((struct vlan_ethhdr *)ethh)->h_vlan_proto = htons(ETH_P_8021Q);
+			((struct vlan_ethhdr *)ethh)->h_vlan_TCI = htons(cm_node->vlan_id);
+			((struct vlan_ethhdr *)ethh)->h_vlan_encapsulated_proto = htons(ETH_P_IPV6);
+		} else {
+			ethh->h_proto = htons(ETH_P_IPV6);
+		}
+		ip6h->version = 6;
+		ip6h->flow_lbl[0] = 0;
+		ip6h->flow_lbl[1] = 0;
+		ip6h->flow_lbl[2] = 0;
+		ip6h->payload_len = htons(packetsize - sizeof(*ip6h));
+		ip6h->nexthdr = 6;
+		ip6h->hop_limit = 128;
+		i40iw_copy_ip_htonl(ip6h->saddr.in6_u.u6_addr32,
+				    cm_node->loc_addr);
+		i40iw_copy_ip_htonl(ip6h->daddr.in6_u.u6_addr32,
+				    cm_node->rem_addr);
+	}
+
+	tcph->source = htons(cm_node->loc_port);
+	tcph->dest = htons(cm_node->rem_port);
+
+	tcph->seq = htonl(cm_node->tcp_cntxt.loc_seq_num);
+
+	if (flags & SET_ACK) {
+		cm_node->tcp_cntxt.loc_ack_num = cm_node->tcp_cntxt.rcv_nxt;
+		tcph->ack_seq = htonl(cm_node->tcp_cntxt.loc_ack_num);
+		tcph->ack = 1;
+	} else {
+		tcph->ack_seq = 0;
+	}
+
+	if (flags & SET_SYN) {
+		cm_node->tcp_cntxt.loc_seq_num++;
+		tcph->syn = 1;
+	} else {
+		cm_node->tcp_cntxt.loc_seq_num += hdr_len + pd_len;
+	}
+
+	if (flags & SET_FIN) {
+		cm_node->tcp_cntxt.loc_seq_num++;
+		tcph->fin = 1;
+	}
+
+	if (flags & SET_RST)
+		tcph->rst = 1;
+
+	tcph->doff = (u16)((sizeof(*tcph) + opts_len + 3) >> 2);
+	sqbuf->tcphlen = tcph->doff << 2;
+	tcph->window = htons(cm_node->tcp_cntxt.rcv_wnd);
+	tcph->urg_ptr = 0;
+
+	if (opts_len) {
+		memcpy(buf, options->addr, opts_len);
+		buf += opts_len;
+	}
+
+	if (hdr_len) {
+		memcpy(buf, hdr->addr, hdr_len);
+		buf += hdr_len;
+	}
+
+	if (pd_len)
+		memcpy(buf, pdata->addr, pd_len);
+
+	atomic_set(&sqbuf->refcount, 1);
+
+	return sqbuf;
+}
+
+/**
+ * i40iw_send_reset - Send RST packet
+ * @cm_node: connection's node
+ */
+static int i40iw_send_reset(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_puda_buf *sqbuf;
+	int flags = SET_RST | SET_ACK;
+
+	sqbuf = i40iw_form_cm_frame(cm_node, NULL, NULL, NULL, flags);
+	if (!sqbuf) {
+		i40iw_pr_err("no sqbuf\n");
+		return -1;
+	}
+
+	return i40iw_schedule_cm_timer(cm_node, sqbuf, I40IW_TIMER_TYPE_SEND, 0, 1);
+}
+
+/**
+ * i40iw_active_open_err - send event for active side cm error
+ * @cm_node: connection's node
+ * @reset: Flag to send reset or not
+ */
+static void i40iw_active_open_err(struct i40iw_cm_node *cm_node, bool reset)
+{
+	i40iw_cleanup_retrans_entry(cm_node);
+	cm_node->cm_core->stats_connect_errs++;
+	if (reset) {
+		i40iw_debug(cm_node->dev,
+			    I40IW_DEBUG_CM,
+			    "%s cm_node=%p state=%d\n",
+			    __func__,
+			    cm_node,
+			    cm_node->state);
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+	}
+
+	cm_node->state = I40IW_CM_STATE_CLOSED;
+	i40iw_create_event(cm_node, I40IW_CM_EVENT_ABORTED);
+}
+
+/**
+ * i40iw_passive_open_err - handle passive side cm error
+ * @cm_node: connection's node
+ * @reset: send reset or just free cm_node
+ */
+static void i40iw_passive_open_err(struct i40iw_cm_node *cm_node, bool reset)
+{
+	i40iw_cleanup_retrans_entry(cm_node);
+	cm_node->cm_core->stats_passive_errs++;
+	cm_node->state = I40IW_CM_STATE_CLOSED;
+	i40iw_debug(cm_node->dev,
+		    I40IW_DEBUG_CM,
+		    "%s cm_node=%p state =%d\n",
+		    __func__,
+		    cm_node,
+		    cm_node->state);
+	if (reset)
+		i40iw_send_reset(cm_node);
+	else
+		i40iw_rem_ref_cm_node(cm_node);
+}
+
+/**
+ * i40iw_event_connect_error - to create connect error event
+ * @event: cm information for connect event
+ */
+static void i40iw_event_connect_error(struct i40iw_cm_event *event)
+{
+	struct i40iw_qp *iwqp;
+	struct iw_cm_id *cm_id;
+
+	cm_id = event->cm_node->cm_id;
+	if (!cm_id)
+		return;
+
+	iwqp = cm_id->provider_data;
+
+	if (!iwqp || !iwqp->iwdev)
+		return;
+
+	iwqp->cm_id = NULL;
+	cm_id->provider_data = NULL;
+	i40iw_send_cm_event(event->cm_node, cm_id,
+			    IW_CM_EVENT_CONNECT_REPLY,
+			    -ECONNRESET);
+	cm_id->rem_ref(cm_id);
+	i40iw_rem_ref_cm_node(event->cm_node);
+}
+
+/**
+ * i40iw_process_options
+ * @cm_node: connection's node
+ * @optionsloc: point to start of options
+ * @optionsize: size of all options
+ * @syn_packet: flag if syn packet
+ */
+static int i40iw_process_options(struct i40iw_cm_node *cm_node,
+				 u8 *optionsloc,
+				 u32 optionsize,
+				 u32 syn_packet)
+{
+	u32 tmp;
+	u32 offset = 0;
+	union all_known_options *all_options;
+	char got_mss_option = 0;
+
+	while (offset < optionsize) {
+		all_options = (union all_known_options *)(optionsloc + offset);
+		switch (all_options->as_base.optionnum) {
+		case OPTION_NUMBER_END:
+			offset = optionsize;
+			break;
+		case OPTION_NUMBER_NONE:
+			offset += 1;
+			continue;
+		case OPTION_NUMBER_MSS:
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "%s: MSS Length: %d Offset: %d Size: %d\n",
+				    __func__,
+				    all_options->as_mss.length,
+				    offset,
+				    optionsize);
+			got_mss_option = 1;
+			if (all_options->as_mss.length != 4)
+				return -1;
+			tmp = ntohs(all_options->as_mss.mss);
+			if (tmp > 0 && tmp < cm_node->tcp_cntxt.mss)
+				cm_node->tcp_cntxt.mss = tmp;
+			break;
+		case OPTION_NUMBER_WINDOW_SCALE:
+			cm_node->tcp_cntxt.snd_wscale =
+			    all_options->as_windowscale.shiftcount;
+			break;
+		default:
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "TCP Option not understood: %x\n",
+				    all_options->as_base.optionnum);
+			break;
+		}
+		offset += all_options->as_base.length;
+	}
+	if (!got_mss_option && syn_packet)
+		cm_node->tcp_cntxt.mss = I40IW_CM_DEFAULT_MSS;
+	return 0;
+}
+
+/**
+ * i40iw_handle_tcp_options -
+ * @cm_node: connection's node
+ * @tcph: pointer tcp header
+ * @optionsize: size of options rcvd
+ * @passive: active or passive flag
+ */
+static int i40iw_handle_tcp_options(struct i40iw_cm_node *cm_node,
+				    struct tcphdr *tcph,
+				    int optionsize,
+				    int passive)
+{
+	u8 *optionsloc = (u8 *)&tcph[1];
+
+	if (optionsize) {
+		if (i40iw_process_options(cm_node,
+					  optionsloc,
+					  optionsize,
+					  (u32)tcph->syn)) {
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "%s: Node %p, Sending RESET\n",
+				    __func__,
+				    cm_node);
+			if (passive)
+				i40iw_passive_open_err(cm_node, true);
+			else
+				i40iw_active_open_err(cm_node, true);
+			return -1;
+		}
+	}
+
+	cm_node->tcp_cntxt.snd_wnd = ntohs(tcph->window) <<
+	    cm_node->tcp_cntxt.snd_wscale;
+
+	if (cm_node->tcp_cntxt.snd_wnd > cm_node->tcp_cntxt.max_snd_wnd)
+		cm_node->tcp_cntxt.max_snd_wnd = cm_node->tcp_cntxt.snd_wnd;
+	return 0;
+}
+
+/**
+ * i40iw_build_mpa_v1 - build a MPA V1 frame
+ * @cm_node: connection's node
+ * @mpa_key: to do read0 or write0
+ */
+static void i40iw_build_mpa_v1(struct i40iw_cm_node *cm_node,
+			       void *start_addr,
+			       u8 mpa_key)
+{
+	struct ietf_mpa_v1 *mpa_frame = (struct ietf_mpa_v1 *)start_addr;
+
+	switch (mpa_key) {
+	case MPA_KEY_REQUEST:
+		memcpy(mpa_frame->key, IEFT_MPA_KEY_REQ, IETF_MPA_KEY_SIZE);
+		break;
+	case MPA_KEY_REPLY:
+		memcpy(mpa_frame->key, IEFT_MPA_KEY_REP, IETF_MPA_KEY_SIZE);
+		break;
+	default:
+		break;
+	}
+	mpa_frame->flags = IETF_MPA_FLAGS_CRC;
+	mpa_frame->rev = cm_node->mpa_frame_rev;
+	mpa_frame->priv_data_len = htons(cm_node->pdata.size);
+}
+
+/**
+ * i40iw_build_mpa_v2 - build a MPA V2 frame
+ * @cm_node: connection's node
+ * @start_addr: buffer start address
+ * @mpa_key: to do read0 or write0
+ */
+static void i40iw_build_mpa_v2(struct i40iw_cm_node *cm_node,
+			       void *start_addr,
+			       u8 mpa_key)
+{
+	struct ietf_mpa_v2 *mpa_frame = (struct ietf_mpa_v2 *)start_addr;
+	struct ietf_rtr_msg *rtr_msg = &mpa_frame->rtr_msg;
+
+	/* initialize the upper 5 bytes of the frame */
+	i40iw_build_mpa_v1(cm_node, start_addr, mpa_key);
+	mpa_frame->flags |= IETF_MPA_V2_FLAG;
+	mpa_frame->priv_data_len += htons(IETF_RTR_MSG_SIZE);
+
+	/* initialize RTR msg */
+	if (cm_node->mpav2_ird_ord == IETF_NO_IRD_ORD) {
+		rtr_msg->ctrl_ird = IETF_NO_IRD_ORD;
+		rtr_msg->ctrl_ord = IETF_NO_IRD_ORD;
+	} else {
+		rtr_msg->ctrl_ird = (cm_node->ird_size > IETF_NO_IRD_ORD) ?
+			IETF_NO_IRD_ORD : cm_node->ird_size;
+		rtr_msg->ctrl_ord = (cm_node->ord_size > IETF_NO_IRD_ORD) ?
+			IETF_NO_IRD_ORD : cm_node->ord_size;
+	}
+
+	rtr_msg->ctrl_ird |= IETF_PEER_TO_PEER;
+	rtr_msg->ctrl_ird |= IETF_FLPDU_ZERO_LEN;
+
+	switch (mpa_key) {
+	case MPA_KEY_REQUEST:
+		rtr_msg->ctrl_ord |= IETF_RDMA0_WRITE;
+		rtr_msg->ctrl_ord |= IETF_RDMA0_READ;
+		break;
+	case MPA_KEY_REPLY:
+		switch (cm_node->send_rdma0_op) {
+		case SEND_RDMA_WRITE_ZERO:
+			rtr_msg->ctrl_ord |= IETF_RDMA0_WRITE;
+			break;
+		case SEND_RDMA_READ_ZERO:
+			rtr_msg->ctrl_ord |= IETF_RDMA0_READ;
+			break;
+		}
+		break;
+	default:
+		break;
+	}
+	rtr_msg->ctrl_ird = htons(rtr_msg->ctrl_ird);
+	rtr_msg->ctrl_ord = htons(rtr_msg->ctrl_ord);
+}
+
+/**
+ * i40iw_cm_build_mpa_frame - build mpa frame for mpa version 1 or version 2
+ * @cm_node: connection's node
+ * @mpa: mpa: data buffer
+ * @mpa_key: to do read0 or write0
+ */
+static int i40iw_cm_build_mpa_frame(struct i40iw_cm_node *cm_node,
+				    struct i40iw_kmem_info *mpa,
+				    u8 mpa_key)
+{
+	int hdr_len = 0;
+
+	switch (cm_node->mpa_frame_rev) {
+	case IETF_MPA_V1:
+		hdr_len = sizeof(struct ietf_mpa_v1);
+		i40iw_build_mpa_v1(cm_node, mpa->addr, mpa_key);
+		break;
+	case IETF_MPA_V2:
+		hdr_len = sizeof(struct ietf_mpa_v2);
+		i40iw_build_mpa_v2(cm_node, mpa->addr, mpa_key);
+		break;
+	default:
+		break;
+	}
+
+	return hdr_len;
+}
+
+/**
+ * i40iw_send_mpa_request - active node send mpa request to passive node
+ * @cm_node: connection's node
+ */
+static int i40iw_send_mpa_request(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_puda_buf *sqbuf;
+
+	if (!cm_node) {
+		i40iw_pr_err("cm_node == NULL\n");
+		return -1;
+	}
+
+	cm_node->mpa_hdr.addr = &cm_node->mpa_frame;
+	cm_node->mpa_hdr.size = i40iw_cm_build_mpa_frame(cm_node,
+							 &cm_node->mpa_hdr,
+							 MPA_KEY_REQUEST);
+	if (!cm_node->mpa_hdr.size) {
+		i40iw_pr_err("mpa size = %d\n", cm_node->mpa_hdr.size);
+		return -1;
+	}
+
+	sqbuf = i40iw_form_cm_frame(cm_node,
+				    NULL,
+				    &cm_node->mpa_hdr,
+				    &cm_node->pdata,
+				    SET_ACK);
+	if (!sqbuf) {
+		i40iw_pr_err("sq_buf == NULL\n");
+		return -1;
+	}
+	return i40iw_schedule_cm_timer(cm_node, sqbuf, I40IW_TIMER_TYPE_SEND, 1, 0);
+}
+
+/**
+ * i40iw_send_mpa_reject -
+ * @cm_node: connection's node
+ * @pdata: reject data for connection
+ * @plen: length of reject data
+ */
+static int i40iw_send_mpa_reject(struct i40iw_cm_node *cm_node,
+				 const void *pdata,
+				 u8 plen)
+{
+	struct i40iw_puda_buf *sqbuf;
+	struct i40iw_kmem_info priv_info;
+
+	cm_node->mpa_hdr.addr = &cm_node->mpa_frame;
+	cm_node->mpa_hdr.size = i40iw_cm_build_mpa_frame(cm_node,
+							 &cm_node->mpa_hdr,
+							 MPA_KEY_REPLY);
+
+	cm_node->mpa_frame.flags |= IETF_MPA_FLAGS_REJECT;
+	priv_info.addr = (void *)pdata;
+	priv_info.size = plen;
+
+	sqbuf = i40iw_form_cm_frame(cm_node,
+				    NULL,
+				    &cm_node->mpa_hdr,
+				    &priv_info,
+				    SET_ACK | SET_FIN);
+	if (!sqbuf) {
+		i40iw_pr_err("no sqbuf\n");
+		return -ENOMEM;
+	}
+	cm_node->state = I40IW_CM_STATE_FIN_WAIT1;
+	return i40iw_schedule_cm_timer(cm_node, sqbuf, I40IW_TIMER_TYPE_SEND, 1, 0);
+}
+
+/**
+ * recv_mpa - process an IETF MPA frame
+ * @cm_node: connection's node
+ * @buffer: Data pointer
+ * @type: to return accept or reject
+ * @len: Len of mpa buffer
+ */
+static int i40iw_parse_mpa(struct i40iw_cm_node *cm_node, u8 *buffer, u32 *type, u32 len)
+{
+	struct ietf_mpa_v1 *mpa_frame;
+	struct ietf_mpa_v2 *mpa_v2_frame;
+	struct ietf_rtr_msg *rtr_msg;
+	int mpa_hdr_len;
+	int priv_data_len;
+
+	*type = I40IW_MPA_REQUEST_ACCEPT;
+
+	if (len < sizeof(struct ietf_mpa_v1)) {
+		i40iw_pr_err("ietf buffer small (%x)\n", len);
+		return -1;
+	}
+
+	mpa_frame = (struct ietf_mpa_v1 *)buffer;
+	mpa_hdr_len = sizeof(struct ietf_mpa_v1);
+	priv_data_len = ntohs(mpa_frame->priv_data_len);
+
+	if (priv_data_len > IETF_MAX_PRIV_DATA_LEN) {
+		i40iw_pr_err("large pri_data %d\n", priv_data_len);
+		return -1;
+	}
+	if (mpa_frame->rev != IETF_MPA_V1 && mpa_frame->rev != IETF_MPA_V2) {
+		i40iw_pr_err("unsupported mpa rev = %d\n", mpa_frame->rev);
+		return -1;
+	}
+	if (mpa_frame->rev > cm_node->mpa_frame_rev) {
+		i40iw_pr_err("rev %d\n", mpa_frame->rev);
+		return -1;
+	}
+	cm_node->mpa_frame_rev = mpa_frame->rev;
+
+	if (cm_node->state != I40IW_CM_STATE_MPAREQ_SENT) {
+		if (memcmp(mpa_frame->key, IEFT_MPA_KEY_REQ, IETF_MPA_KEY_SIZE)) {
+			i40iw_pr_err("Unexpected MPA Key received\n");
+			return -1;
+		}
+	} else {
+		if (memcmp(mpa_frame->key, IEFT_MPA_KEY_REP, IETF_MPA_KEY_SIZE)) {
+			i40iw_pr_err("Unexpected MPA Key received\n");
+			return -1;
+		}
+	}
+
+	if (priv_data_len + mpa_hdr_len > len) {
+		i40iw_pr_err("ietf buffer len(%x + %x != %x)\n",
+			     priv_data_len, mpa_hdr_len, len);
+		return -1;
+	}
+	if (len > MAX_CM_BUFFER) {
+		i40iw_pr_err("ietf buffer large len = %d\n", len);
+		return -1;
+	}
+
+	switch (mpa_frame->rev) {
+	case IETF_MPA_V2:{
+			u16 ird_size;
+			u16 ord_size;
+			u16 ctrl_ord;
+			u16 ctrl_ird;
+
+			mpa_v2_frame = (struct ietf_mpa_v2 *)buffer;
+			mpa_hdr_len += IETF_RTR_MSG_SIZE;
+			rtr_msg = &mpa_v2_frame->rtr_msg;
+
+			/* parse rtr message */
+			ctrl_ord = ntohs(rtr_msg->ctrl_ord);
+			ctrl_ird = ntohs(rtr_msg->ctrl_ird);
+			ird_size = ctrl_ird & IETF_NO_IRD_ORD;
+			ord_size = ctrl_ord & IETF_NO_IRD_ORD;
+
+			if (!(ctrl_ird & IETF_PEER_TO_PEER))
+				return -1;
+
+			if (ird_size == IETF_NO_IRD_ORD || ord_size == IETF_NO_IRD_ORD) {
+				cm_node->mpav2_ird_ord = IETF_NO_IRD_ORD;
+				goto negotiate_done;
+			}
+
+			if (cm_node->state != I40IW_CM_STATE_MPAREQ_SENT) {
+				/* responder */
+				if (!ord_size && (ctrl_ord & IETF_RDMA0_READ))
+					cm_node->ird_size = 1;
+				if (cm_node->ord_size > ird_size)
+					cm_node->ord_size = ird_size;
+			} else {
+				/* initiator */
+				if (!ird_size && (ctrl_ord & IETF_RDMA0_READ))
+					return -1;
+				if (cm_node->ord_size > ird_size)
+					cm_node->ord_size = ird_size;
+
+				if (cm_node->ird_size < ord_size)
+					/* no resources available */
+					return -1;
+			}
+
+negotiate_done:
+			if (ctrl_ord & IETF_RDMA0_READ)
+				cm_node->send_rdma0_op = SEND_RDMA_READ_ZERO;
+			else if (ctrl_ord & IETF_RDMA0_WRITE)
+				cm_node->send_rdma0_op = SEND_RDMA_WRITE_ZERO;
+			else	/* Not supported RDMA0 operation */
+				return -1;
+			i40iw_debug(cm_node->dev, I40IW_DEBUG_CM,
+				    "MPAV2: Negotiated ORD: %d, IRD: %d\n",
+				    cm_node->ord_size, cm_node->ird_size);
+			break;
+		}
+		break;
+	case IETF_MPA_V1:
+	default:
+		break;
+	}
+
+	memcpy(cm_node->pdata_buf, buffer + mpa_hdr_len, priv_data_len);
+	cm_node->pdata.size = priv_data_len;
+
+	if (mpa_frame->flags & IETF_MPA_FLAGS_REJECT)
+		*type = I40IW_MPA_REQUEST_REJECT;
+
+	if (mpa_frame->flags & IETF_MPA_FLAGS_MARKERS)
+		cm_node->snd_mark_en = true;
+
+	return 0;
+}
+
+/**
+ * i40iw_schedule_cm_timer
+ * @@cm_node: connection's node
+ * @sqbuf: buffer to send
+ * @type: if it es send ot close
+ * @send_retrans: if rexmits to be done
+ * @close_when_complete: is cm_node to be removed
+ *
+ * note - cm_node needs to be protected before calling this. Encase in:
+ *		i40iw_rem_ref_cm_node(cm_core, cm_node);
+ *		i40iw_schedule_cm_timer(...)
+ *		atomic_inc(&cm_node->ref_count);
+ */
+int i40iw_schedule_cm_timer(struct i40iw_cm_node *cm_node,
+			    struct i40iw_puda_buf *sqbuf,
+			    enum i40iw_timer_type type,
+			    int send_retrans,
+			    int close_when_complete)
+{
+	struct i40iw_sc_dev *dev = cm_node->dev;
+	struct i40iw_cm_core *cm_core = cm_node->cm_core;
+	struct i40iw_timer_entry *new_send;
+	int ret = 0;
+	u32 was_timer_set;
+	unsigned long flags;
+
+	new_send = kzalloc(sizeof(*new_send), GFP_ATOMIC);
+	if (!new_send) {
+		i40iw_free_sqbuf(cm_node->dev, (void *)sqbuf);
+		return -ENOMEM;
+	}
+	new_send->retrycount = I40IW_DEFAULT_RETRYS;
+	new_send->retranscount = I40IW_DEFAULT_RETRANS;
+	new_send->sqbuf = sqbuf;
+	new_send->timetosend = jiffies;
+	new_send->type = type;
+	new_send->send_retrans = send_retrans;
+	new_send->close_when_complete = close_when_complete;
+
+	if (type == I40IW_TIMER_TYPE_CLOSE) {
+		new_send->timetosend += (HZ / 10);
+		if (cm_node->close_entry) {
+			kfree(new_send);
+			i40iw_free_sqbuf(cm_node->dev, (void *)sqbuf);
+			i40iw_pr_err("already close entry\n");
+			return -EINVAL;
+		}
+		cm_node->close_entry = new_send;
+	}
+
+	if (type == I40IW_TIMER_TYPE_SEND) {
+		spin_lock_irqsave(&cm_node->retrans_list_lock, flags);
+		cm_node->send_entry = new_send;
+		atomic_inc(&cm_node->ref_count);
+		spin_unlock_irqrestore(&cm_node->retrans_list_lock, flags);
+		new_send->timetosend = jiffies + I40IW_RETRY_TIMEOUT;
+
+		atomic_inc(&sqbuf->refcount);
+		i40iw_puda_send_buf(dev->ilq, sqbuf);
+		if (!send_retrans) {
+			i40iw_cleanup_retrans_entry(cm_node);
+			if (close_when_complete)
+				i40iw_rem_ref_cm_node(cm_node);
+			return ret;
+		}
+	}
+
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	was_timer_set = timer_pending(&cm_core->tcp_timer);
+
+	if (!was_timer_set) {
+		cm_core->tcp_timer.expires = new_send->timetosend;
+		add_timer(&cm_core->tcp_timer);
+	}
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+
+	return ret;
+}
+
+/**
+ * i40iw_retrans_expired - Could not rexmit the packet
+ * @cm_node: connection's node
+ */
+static void i40iw_retrans_expired(struct i40iw_cm_node *cm_node)
+{
+	struct iw_cm_id *cm_id = cm_node->cm_id;
+	enum i40iw_cm_node_state state = cm_node->state;
+
+	cm_node->state = I40IW_CM_STATE_CLOSED;
+	switch (state) {
+	case I40IW_CM_STATE_SYN_RCVD:
+	case I40IW_CM_STATE_CLOSING:
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	case I40IW_CM_STATE_FIN_WAIT1:
+	case I40IW_CM_STATE_LAST_ACK:
+		if (cm_node->cm_id)
+			cm_id->rem_ref(cm_id);
+		i40iw_send_reset(cm_node);
+		break;
+	default:
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+		i40iw_create_event(cm_node, I40IW_CM_EVENT_ABORTED);
+		break;
+	}
+}
+
+/**
+ * i40iw_handle_close_entry - for handling retry/timeouts
+ * @cm_node: connection's node
+ * @rem_node: flag for remove cm_node
+ */
+static void i40iw_handle_close_entry(struct i40iw_cm_node *cm_node, u32 rem_node)
+{
+	struct i40iw_timer_entry *close_entry = cm_node->close_entry;
+	struct iw_cm_id *cm_id = cm_node->cm_id;
+	struct i40iw_qp *iwqp;
+	unsigned long flags;
+
+	if (!close_entry)
+		return;
+	iwqp = (struct i40iw_qp *)close_entry->sqbuf;
+	if (iwqp) {
+		spin_lock_irqsave(&iwqp->lock, flags);
+		if (iwqp->cm_id) {
+			iwqp->hw_tcp_state = I40IW_TCP_STATE_CLOSED;
+			iwqp->hw_iwarp_state = I40IW_QP_STATE_ERROR;
+			iwqp->last_aeq = I40IW_AE_RESET_SENT;
+			iwqp->ibqp_state = IB_QPS_ERR;
+			spin_unlock_irqrestore(&iwqp->lock, flags);
+			i40iw_cm_disconn(iwqp);
+		} else {
+			spin_unlock_irqrestore(&iwqp->lock, flags);
+		}
+	} else if (rem_node) {
+		/* TIME_WAIT state */
+		i40iw_rem_ref_cm_node(cm_node);
+	}
+	if (cm_id)
+		cm_id->rem_ref(cm_id);
+	kfree(close_entry);
+	cm_node->close_entry = NULL;
+}
+
+/**
+ * i40iw_cm_timer_tick - system's timer expired callback
+ * @pass: Pointing to cm_core
+ */
+static void i40iw_cm_timer_tick(unsigned long pass)
+{
+	unsigned long nexttimeout = jiffies + I40IW_LONG_TIME;
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_timer_entry *send_entry, *close_entry;
+	struct list_head *list_core_temp;
+	struct list_head *list_node;
+	struct i40iw_cm_core *cm_core = (struct i40iw_cm_core *)pass;
+	u32 settimer = 0;
+	unsigned long timetosend;
+	struct i40iw_sc_dev *dev;
+	unsigned long flags;
+
+	struct list_head timer_list;
+
+	INIT_LIST_HEAD(&timer_list);
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+
+	list_for_each_safe(list_node, list_core_temp, &cm_core->connected_nodes) {
+		cm_node = container_of(list_node, struct i40iw_cm_node, list);
+		if (cm_node->close_entry || cm_node->send_entry) {
+			atomic_inc(&cm_node->ref_count);
+			list_add(&cm_node->timer_entry, &timer_list);
+		}
+	}
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+
+	list_for_each_safe(list_node, list_core_temp, &timer_list) {
+		cm_node = container_of(list_node,
+				       struct i40iw_cm_node,
+				       timer_entry);
+		close_entry = cm_node->close_entry;
+
+		if (close_entry) {
+			if (time_after(close_entry->timetosend, jiffies)) {
+				if (nexttimeout > close_entry->timetosend ||
+				    !settimer) {
+					nexttimeout = close_entry->timetosend;
+					settimer = 1;
+				}
+			} else {
+				i40iw_handle_close_entry(cm_node, 1);
+			}
+		}
+
+		spin_lock_irqsave(&cm_node->retrans_list_lock, flags);
+
+		send_entry = cm_node->send_entry;
+		if (!send_entry)
+			goto done;
+		if (time_after(send_entry->timetosend, jiffies)) {
+			if (cm_node->state != I40IW_CM_STATE_OFFLOADED) {
+				if ((nexttimeout > send_entry->timetosend) ||
+				    !settimer) {
+					nexttimeout = send_entry->timetosend;
+					settimer = 1;
+				}
+			} else {
+				i40iw_free_retrans_entry(cm_node);
+			}
+			goto done;
+		}
+
+		if ((cm_node->state == I40IW_CM_STATE_OFFLOADED) ||
+		    (cm_node->state == I40IW_CM_STATE_CLOSED)) {
+			i40iw_free_retrans_entry(cm_node);
+			goto done;
+		}
+
+		if (!send_entry->retranscount || !send_entry->retrycount) {
+			i40iw_free_retrans_entry(cm_node);
+
+			spin_unlock_irqrestore(&cm_node->retrans_list_lock, flags);
+			i40iw_retrans_expired(cm_node);
+			cm_node->state = I40IW_CM_STATE_CLOSED;
+			spin_lock_irqsave(&cm_node->retrans_list_lock, flags);
+			goto done;
+		}
+		cm_node->cm_core->stats_pkt_retrans++;
+		spin_unlock_irqrestore(&cm_node->retrans_list_lock, flags);
+
+		dev = cm_node->dev;
+		atomic_inc(&send_entry->sqbuf->refcount);
+		i40iw_puda_send_buf(dev->ilq, send_entry->sqbuf);
+		spin_lock_irqsave(&cm_node->retrans_list_lock, flags);
+		if (send_entry->send_retrans) {
+			send_entry->retranscount--;
+			timetosend = (I40IW_RETRY_TIMEOUT <<
+				      (I40IW_DEFAULT_RETRANS -
+				       send_entry->retranscount));
+
+			send_entry->timetosend = jiffies +
+			    min(timetosend, I40IW_MAX_TIMEOUT);
+			if (nexttimeout > send_entry->timetosend || !settimer) {
+				nexttimeout = send_entry->timetosend;
+				settimer = 1;
+			}
+		} else {
+			int close_when_complete;
+
+			close_when_complete = send_entry->close_when_complete;
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "cm_node=%p state=%d\n",
+				    cm_node,
+				    cm_node->state);
+			i40iw_free_retrans_entry(cm_node);
+			if (close_when_complete)
+				i40iw_rem_ref_cm_node(cm_node);
+		}
+done:
+		spin_unlock_irqrestore(&cm_node->retrans_list_lock, flags);
+		i40iw_rem_ref_cm_node(cm_node);
+	}
+
+	if (settimer) {
+		spin_lock_irqsave(&cm_core->ht_lock, flags);
+		if (!timer_pending(&cm_core->tcp_timer)) {
+			cm_core->tcp_timer.expires = nexttimeout;
+			add_timer(&cm_core->tcp_timer);
+		}
+		spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+	}
+}
+
+/**
+ * i40iw_send_syn - send SYN packet
+ * @cm_node: connection's node
+ * @sendack: flag to set ACK bit or not
+ */
+int i40iw_send_syn(struct i40iw_cm_node *cm_node, u32 sendack)
+{
+	struct i40iw_puda_buf *sqbuf;
+	int flags = SET_SYN;
+	char optionsbuffer[sizeof(struct option_mss) +
+			   sizeof(struct option_windowscale) +
+			   sizeof(struct option_base) + TCP_OPTIONS_PADDING];
+	struct i40iw_kmem_info opts;
+
+	int optionssize = 0;
+	/* Sending MSS option */
+	union all_known_options *options;
+
+	opts.addr = optionsbuffer;
+	if (!cm_node) {
+		i40iw_pr_err("no cm_node\n");
+		return -EINVAL;
+	}
+
+	options = (union all_known_options *)&optionsbuffer[optionssize];
+	options->as_mss.optionnum = OPTION_NUMBER_MSS;
+	options->as_mss.length = sizeof(struct option_mss);
+	options->as_mss.mss = htons(cm_node->tcp_cntxt.mss);
+	optionssize += sizeof(struct option_mss);
+
+	options = (union all_known_options *)&optionsbuffer[optionssize];
+	options->as_windowscale.optionnum = OPTION_NUMBER_WINDOW_SCALE;
+	options->as_windowscale.length = sizeof(struct option_windowscale);
+	options->as_windowscale.shiftcount = cm_node->tcp_cntxt.rcv_wscale;
+	optionssize += sizeof(struct option_windowscale);
+	options = (union all_known_options *)&optionsbuffer[optionssize];
+	options->as_end = OPTION_NUMBER_END;
+	optionssize += 1;
+
+	if (sendack)
+		flags |= SET_ACK;
+
+	opts.size = optionssize;
+
+	sqbuf = i40iw_form_cm_frame(cm_node, &opts, NULL, NULL, flags);
+	if (!sqbuf) {
+		i40iw_pr_err("no sqbuf\n");
+		return -1;
+	}
+	return i40iw_schedule_cm_timer(cm_node, sqbuf, I40IW_TIMER_TYPE_SEND, 1, 0);
+}
+
+/**
+ * i40iw_send_ack - Send ACK packet
+ * @cm_node: connection's node
+ */
+static void i40iw_send_ack(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_puda_buf *sqbuf;
+
+	sqbuf = i40iw_form_cm_frame(cm_node, NULL, NULL, NULL, SET_ACK);
+	if (sqbuf)
+		i40iw_puda_send_buf(cm_node->dev->ilq, sqbuf);
+	else
+		i40iw_pr_err("no sqbuf\n");
+}
+
+/**
+ * i40iw_send_fin - Send FIN pkt
+ * @cm_node: connection's node
+ */
+static int i40iw_send_fin(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_puda_buf *sqbuf;
+
+	sqbuf = i40iw_form_cm_frame(cm_node, NULL, NULL, NULL, SET_ACK | SET_FIN);
+	if (!sqbuf) {
+		i40iw_pr_err("no sqbuf\n");
+		return -1;
+	}
+	return i40iw_schedule_cm_timer(cm_node, sqbuf, I40IW_TIMER_TYPE_SEND, 1, 0);
+}
+
+/**
+ * i40iw_find_node - find a cm node that matches the reference cm node
+ * @cm_core: cm's core
+ * @rem_port: remote tcp port num
+ * @rem_addr: remote ip addr
+ * @loc_port: local tcp port num
+ * @loc_addr: loc ip addr
+ * @add_refcnt: flag to increment refcount of cm_node
+ */
+struct i40iw_cm_node *i40iw_find_node(struct i40iw_cm_core *cm_core,
+				      u16 rem_port,
+				      u32 *rem_addr,
+				      u16 loc_port,
+				      u32 *loc_addr,
+				      bool add_refcnt)
+{
+	struct list_head *hte;
+	struct i40iw_cm_node *cm_node;
+	unsigned long flags;
+
+	hte = &cm_core->connected_nodes;
+
+	/* walk list and find cm_node associated with this session ID */
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	list_for_each_entry(cm_node, hte, list) {
+		if (!memcmp(cm_node->loc_addr, loc_addr, sizeof(cm_node->loc_addr)) &&
+		    (cm_node->loc_port == loc_port) &&
+		    !memcmp(cm_node->rem_addr, rem_addr, sizeof(cm_node->rem_addr)) &&
+		    (cm_node->rem_port == rem_port)) {
+			if (add_refcnt)
+				atomic_inc(&cm_node->ref_count);
+			spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+			return cm_node;
+		}
+	}
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+
+	/* no owner node */
+	return NULL;
+}
+
+/**
+ * i40iw_find_listener - find a cm node listening on this addr-port pair
+ * @cm_core: cm's core
+ * @dst_port: listener tcp port num
+ * @dst_addr: listener ip addr
+ * @listener_state: state to match with listen node's
+ */
+static struct i40iw_cm_listener *i40iw_find_listener(
+						     struct i40iw_cm_core *cm_core,
+						     u32 *dst_addr,
+						     u16 dst_port,
+						     u16 vlan_id,
+						     enum i40iw_cm_listener_state
+						     listener_state)
+{
+	struct i40iw_cm_listener *listen_node;
+	static const u32 ip_zero[4] = { 0, 0, 0, 0 };
+	u32 listen_addr[4];
+	u16 listen_port;
+	unsigned long flags;
+
+	/* walk list and find cm_node associated with this session ID */
+	spin_lock_irqsave(&cm_core->listen_list_lock, flags);
+	list_for_each_entry(listen_node, &cm_core->listen_nodes, list) {
+		memcpy(listen_addr, listen_node->loc_addr, sizeof(listen_addr));
+		listen_port = listen_node->loc_port;
+		/* compare node pair, return node handle if a match */
+		if ((!memcmp(listen_addr, dst_addr, sizeof(listen_addr)) ||
+		     !memcmp(listen_addr, ip_zero, sizeof(listen_addr))) &&
+		     (listen_port == dst_port) &&
+		     (listener_state & listen_node->listener_state)) {
+			atomic_inc(&listen_node->ref_count);
+			spin_unlock_irqrestore(&cm_core->listen_list_lock, flags);
+			return listen_node;
+		}
+	}
+	spin_unlock_irqrestore(&cm_core->listen_list_lock, flags);
+	return NULL;
+}
+
+/**
+ * i40iw_add_hte_node - add a cm node to the hash table
+ * @cm_core: cm's core
+ * @cm_node: connection's node
+ */
+static void i40iw_add_hte_node(struct i40iw_cm_core *cm_core,
+			       struct i40iw_cm_node *cm_node)
+{
+	struct list_head *hte;
+	unsigned long flags;
+
+	if (!cm_node || !cm_core) {
+		i40iw_pr_err("cm_node or cm_core == NULL\n");
+		return;
+	}
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+
+	/* get a handle on the hash table element (list head for this slot) */
+	hte = &cm_core->connected_nodes;
+	list_add_tail(&cm_node->list, hte);
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+}
+
+/**
+ * listen_port_in_use - determine if port is in use
+ * @port: Listen port number
+ */
+static bool i40iw_listen_port_in_use(struct i40iw_cm_core *cm_core, u16 port)
+{
+	struct i40iw_cm_listener *listen_node;
+	unsigned long flags;
+	bool ret = false;
+
+	spin_lock_irqsave(&cm_core->listen_list_lock, flags);
+	list_for_each_entry(listen_node, &cm_core->listen_nodes, list) {
+		if (listen_node->loc_port == port) {
+			ret = true;
+			break;
+		}
+	}
+	spin_unlock_irqrestore(&cm_core->listen_list_lock, flags);
+	return ret;
+}
+
+/**
+ * i40iw_del_multiple_qhash - Remove qhash and child listens
+ * @iwdev: iWarp device
+ * @cm_info: CM info for parent listen node
+ * @cm_parent_listen_node: The parent listen node
+ */
+static enum i40iw_status_code i40iw_del_multiple_qhash(
+						       struct i40iw_device *iwdev,
+						       struct i40iw_cm_info *cm_info,
+						       struct i40iw_cm_listener *cm_parent_listen_node)
+{
+	struct i40iw_cm_listener *child_listen_node;
+	enum i40iw_status_code ret = I40IW_ERR_CONFIG;
+	struct list_head *pos, *tpos;
+	unsigned long flags;
+
+	spin_lock_irqsave(&iwdev->cm_core.listen_list_lock, flags);
+	list_for_each_safe(pos, tpos, &cm_parent_listen_node->child_listen_list) {
+		child_listen_node = list_entry(pos, struct i40iw_cm_listener, child_listen_list);
+		if (child_listen_node->ipv4)
+			i40iw_debug(&iwdev->sc_dev,
+				    I40IW_DEBUG_CM,
+				    "removing child listen for IP=%pI4, port=%d, vlan=%d\n",
+				    child_listen_node->loc_addr,
+				    child_listen_node->loc_port,
+				    child_listen_node->vlan_id);
+		else
+			i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_CM,
+				    "removing child listen for IP=%pI6, port=%d, vlan=%d\n",
+				    child_listen_node->loc_addr,
+				    child_listen_node->loc_port,
+				    child_listen_node->vlan_id);
+		list_del(pos);
+		memcpy(cm_info->loc_addr, child_listen_node->loc_addr,
+		       sizeof(cm_info->loc_addr));
+		cm_info->vlan_id = child_listen_node->vlan_id;
+		ret = i40iw_manage_qhash(iwdev, cm_info,
+					 I40IW_QHASH_TYPE_TCP_SYN,
+					 I40IW_QHASH_MANAGE_TYPE_DELETE, NULL, false);
+		kfree(child_listen_node);
+		cm_parent_listen_node->cm_core->stats_listen_nodes_destroyed++;
+		i40iw_debug(&iwdev->sc_dev,
+			    I40IW_DEBUG_CM,
+			    "freed pointer = %p\n",
+			    child_listen_node);
+	}
+	spin_unlock_irqrestore(&iwdev->cm_core.listen_list_lock, flags);
+
+	return ret;
+}
+
+/**
+ * i40iw_netdev_vlan_ipv6 - Gets the netdev and mac
+ * @addr: local IPv6 address
+ * @vlan_id: vlan id for the given IPv6 address
+ * @mac: mac address for the given IPv6 address
+ *
+ * Returns the net_device of the IPv6 address and also sets the
+ * vlan id and mac for that address.
+ */
+static struct net_device *i40iw_netdev_vlan_ipv6(u32 *addr, u16 *vlan_id, u8 *mac)
+{
+	struct net_device *ip_dev = NULL;
+#if IS_ENABLED(CONFIG_IPV6)
+	struct in6_addr laddr6;
+
+	i40iw_copy_ip_htonl(laddr6.in6_u.u6_addr32, addr);
+	if (vlan_id)
+		*vlan_id = I40IW_NO_VLAN;
+	if (mac)
+		eth_zero_addr(mac);
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, ip_dev) {
+		if (ipv6_chk_addr(&init_net, &laddr6, ip_dev, 1)) {
+			if (vlan_id)
+				*vlan_id = rdma_vlan_dev_vlan_id(ip_dev);
+			if (ip_dev->dev_addr && mac)
+				ether_addr_copy(mac, ip_dev->dev_addr);
+			break;
+		}
+	}
+	rcu_read_unlock();
+#endif
+	return ip_dev;
+}
+
+/**
+ * i40iw_get_vlan_ipv4 - Returns the vlan_id for IPv4 address
+ * @addr: local IPv4 address
+ */
+static u16 i40iw_get_vlan_ipv4(u32 *addr)
+{
+	struct net_device *netdev;
+	u16 vlan_id = I40IW_NO_VLAN;
+
+	netdev = ip_dev_find(&init_net, htonl(addr[0]));
+	if (netdev) {
+		vlan_id = rdma_vlan_dev_vlan_id(netdev);
+		dev_put(netdev);
+	}
+	return vlan_id;
+}
+
+/**
+ * i40iw_add_mqh_6 - Adds multiple qhashes for IPv6
+ * @iwdev: iWarp device
+ * @cm_info: CM info for parent listen node
+ * @cm_parent_listen_node: The parent listen node
+ *
+ * Adds a qhash and a child listen node for every IPv6 address
+ * on the adapter and adds the associated qhash filter
+ */
+static enum i40iw_status_code i40iw_add_mqh_6(struct i40iw_device *iwdev,
+					      struct i40iw_cm_info *cm_info,
+					      struct i40iw_cm_listener *cm_parent_listen_node)
+{
+	struct net_device *ip_dev;
+	struct inet6_dev *idev;
+	struct inet6_ifaddr *ifp;
+	enum i40iw_status_code ret = 0;
+	struct i40iw_cm_listener *child_listen_node;
+	unsigned long flags;
+
+	rtnl_lock();
+	for_each_netdev_rcu(&init_net, ip_dev) {
+		if ((((rdma_vlan_dev_vlan_id(ip_dev) < I40IW_NO_VLAN) &&
+		      (rdma_vlan_dev_real_dev(ip_dev) == iwdev->netdev)) ||
+		     (ip_dev == iwdev->netdev)) && (ip_dev->flags & IFF_UP)) {
+			idev = __in6_dev_get(ip_dev);
+			if (!idev) {
+				i40iw_pr_err("idev == NULL\n");
+				break;
+			}
+			list_for_each_entry(ifp, &idev->addr_list, if_list) {
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "IP=%pI6, vlan_id=%d, MAC=%pM\n",
+					    &ifp->addr,
+					    rdma_vlan_dev_vlan_id(ip_dev),
+					    ip_dev->dev_addr);
+				child_listen_node =
+					kzalloc(sizeof(*child_listen_node), GFP_ATOMIC);
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "Allocating child listener %p\n",
+					    child_listen_node);
+				if (!child_listen_node) {
+					i40iw_pr_err("listener memory allocation\n");
+					ret = I40IW_ERR_NO_MEMORY;
+					goto exit;
+				}
+				cm_info->vlan_id = rdma_vlan_dev_vlan_id(ip_dev);
+				cm_parent_listen_node->vlan_id = cm_info->vlan_id;
+
+				memcpy(child_listen_node, cm_parent_listen_node,
+				       sizeof(*child_listen_node));
+
+				i40iw_copy_ip_ntohl(child_listen_node->loc_addr,
+						    ifp->addr.in6_u.u6_addr32);
+				memcpy(cm_info->loc_addr, child_listen_node->loc_addr,
+				       sizeof(cm_info->loc_addr));
+
+				ret = i40iw_manage_qhash(iwdev, cm_info,
+							 I40IW_QHASH_TYPE_TCP_SYN,
+							 I40IW_QHASH_MANAGE_TYPE_ADD,
+							 NULL, true);
+				if (!ret) {
+					spin_lock_irqsave(&iwdev->cm_core.listen_list_lock, flags);
+					list_add(&child_listen_node->child_listen_list,
+						 &cm_parent_listen_node->child_listen_list);
+					spin_unlock_irqrestore(&iwdev->cm_core.listen_list_lock, flags);
+					cm_parent_listen_node->cm_core->stats_listen_nodes_created++;
+				} else {
+					kfree(child_listen_node);
+				}
+			}
+		}
+	}
+exit:
+	rtnl_unlock();
+	return ret;
+}
+
+/**
+ * i40iw_add_mqh_4 - Adds multiple qhashes for IPv4
+ * @iwdev: iWarp device
+ * @cm_info: CM info for parent listen node
+ * @cm_parent_listen_node: The parent listen node
+ *
+ * Adds a qhash and a child listen node for every IPv4 address
+ * on the adapter and adds the associated qhash filter
+ */
+static enum i40iw_status_code i40iw_add_mqh_4(
+				struct i40iw_device *iwdev,
+				struct i40iw_cm_info *cm_info,
+				struct i40iw_cm_listener *cm_parent_listen_node)
+{
+	struct net_device *dev;
+	struct in_device *idev;
+	struct i40iw_cm_listener *child_listen_node;
+	enum i40iw_status_code ret = 0;
+	unsigned long flags;
+
+	rtnl_lock();
+	for_each_netdev(&init_net, dev) {
+		if ((((rdma_vlan_dev_vlan_id(dev) < I40IW_NO_VLAN) &&
+		      (rdma_vlan_dev_real_dev(dev) == iwdev->netdev)) ||
+		    (dev == iwdev->netdev)) && (dev->flags & IFF_UP)) {
+			idev = in_dev_get(dev);
+			for_ifa(idev) {
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "Allocating child CM Listener forIP=%pI4, vlan_id=%d, MAC=%pM\n",
+					    &ifa->ifa_address,
+					    rdma_vlan_dev_vlan_id(dev),
+					    dev->dev_addr);
+				child_listen_node = kzalloc(sizeof(*child_listen_node), GFP_ATOMIC);
+				cm_parent_listen_node->cm_core->stats_listen_nodes_created++;
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "Allocating child listener %p\n",
+					    child_listen_node);
+				if (!child_listen_node) {
+					i40iw_pr_err("listener memory allocation\n");
+					in_dev_put(idev);
+					ret = I40IW_ERR_NO_MEMORY;
+					goto exit;
+				}
+				cm_info->vlan_id = rdma_vlan_dev_vlan_id(dev);
+				cm_parent_listen_node->vlan_id = cm_info->vlan_id;
+				memcpy(child_listen_node,
+				       cm_parent_listen_node,
+				       sizeof(*child_listen_node));
+
+				child_listen_node->loc_addr[0] = ntohl(ifa->ifa_address);
+				memcpy(cm_info->loc_addr, child_listen_node->loc_addr,
+				       sizeof(cm_info->loc_addr));
+
+				ret = i40iw_manage_qhash(iwdev,
+							 cm_info,
+							 I40IW_QHASH_TYPE_TCP_SYN,
+							 I40IW_QHASH_MANAGE_TYPE_ADD,
+							 NULL,
+							 true);
+				if (!ret) {
+					spin_lock_irqsave(&iwdev->cm_core.listen_list_lock, flags);
+					list_add(&child_listen_node->child_listen_list,
+						 &cm_parent_listen_node->child_listen_list);
+					spin_unlock_irqrestore(&iwdev->cm_core.listen_list_lock, flags);
+				} else {
+					kfree(child_listen_node);
+					cm_parent_listen_node->cm_core->stats_listen_nodes_created--;
+				}
+			}
+			endfor_ifa(idev);
+			in_dev_put(idev);
+		}
+	}
+exit:
+	rtnl_unlock();
+	return ret;
+}
+
+/**
+ * i40iw_dec_refcnt_listen - delete listener and associated cm nodes
+ * @cm_core: cm's core
+ * @free_hanging_nodes: to free associated cm_nodes
+ * @apbvt_del: flag to delete the apbvt
+ */
+static int i40iw_dec_refcnt_listen(struct i40iw_cm_core *cm_core,
+				   struct i40iw_cm_listener *listener,
+				   int free_hanging_nodes, bool apbvt_del)
+{
+	int ret = -EINVAL;
+	int err = 0;
+	struct list_head *list_pos;
+	struct list_head *list_temp;
+	struct i40iw_cm_node *cm_node;
+	struct list_head reset_list;
+	struct i40iw_cm_info nfo;
+	struct i40iw_cm_node *loopback;
+	enum i40iw_cm_node_state old_state;
+	unsigned long flags;
+
+	/* free non-accelerated child nodes for this listener */
+	INIT_LIST_HEAD(&reset_list);
+	if (free_hanging_nodes) {
+		spin_lock_irqsave(&cm_core->ht_lock, flags);
+		list_for_each_safe(list_pos, list_temp, &cm_core->connected_nodes) {
+			cm_node = container_of(list_pos, struct i40iw_cm_node, list);
+			if ((cm_node->listener == listener) && !cm_node->accelerated) {
+				atomic_inc(&cm_node->ref_count);
+				list_add(&cm_node->reset_entry, &reset_list);
+			}
+		}
+		spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+	}
+
+	list_for_each_safe(list_pos, list_temp, &reset_list) {
+		cm_node = container_of(list_pos, struct i40iw_cm_node, reset_entry);
+		loopback = cm_node->loopbackpartner;
+		if (cm_node->state >= I40IW_CM_STATE_FIN_WAIT1) {
+			i40iw_rem_ref_cm_node(cm_node);
+		} else {
+			if (!loopback) {
+				i40iw_cleanup_retrans_entry(cm_node);
+				err = i40iw_send_reset(cm_node);
+				if (err) {
+					cm_node->state = I40IW_CM_STATE_CLOSED;
+					i40iw_pr_err("send reset\n");
+				} else {
+					old_state = cm_node->state;
+					cm_node->state = I40IW_CM_STATE_LISTENER_DESTROYED;
+					if (old_state != I40IW_CM_STATE_MPAREQ_RCVD)
+						i40iw_rem_ref_cm_node(cm_node);
+				}
+			} else {
+				struct i40iw_cm_event event;
+
+				event.cm_node = loopback;
+				memcpy(event.cm_info.rem_addr,
+				       loopback->rem_addr, sizeof(event.cm_info.rem_addr));
+				memcpy(event.cm_info.loc_addr,
+				       loopback->loc_addr, sizeof(event.cm_info.loc_addr));
+				event.cm_info.rem_port = loopback->rem_port;
+				event.cm_info.loc_port = loopback->loc_port;
+				event.cm_info.cm_id = loopback->cm_id;
+				event.cm_info.ipv4 = loopback->ipv4;
+				atomic_inc(&loopback->ref_count);
+				loopback->state = I40IW_CM_STATE_CLOSED;
+				i40iw_event_connect_error(&event);
+				cm_node->state = I40IW_CM_STATE_LISTENER_DESTROYED;
+				i40iw_rem_ref_cm_node(cm_node);
+			}
+		}
+	}
+
+	if (!atomic_dec_return(&listener->ref_count)) {
+		spin_lock_irqsave(&cm_core->listen_list_lock, flags);
+		list_del(&listener->list);
+		spin_unlock_irqrestore(&cm_core->listen_list_lock, flags);
+
+		if (listener->iwdev) {
+			if (apbvt_del && !i40iw_listen_port_in_use(cm_core, listener->loc_port))
+				i40iw_manage_apbvt(listener->iwdev,
+						   listener->loc_port,
+						   I40IW_MANAGE_APBVT_DEL);
+
+			memcpy(nfo.loc_addr, listener->loc_addr, sizeof(nfo.loc_addr));
+			nfo.loc_port = listener->loc_port;
+			nfo.ipv4 = listener->ipv4;
+			nfo.vlan_id = listener->vlan_id;
+
+			if (!list_empty(&listener->child_listen_list)) {
+				i40iw_del_multiple_qhash(listener->iwdev, &nfo, listener);
+			} else {
+				if (listener->qhash_set)
+					i40iw_manage_qhash(listener->iwdev,
+							   &nfo,
+							   I40IW_QHASH_TYPE_TCP_SYN,
+							   I40IW_QHASH_MANAGE_TYPE_DELETE,
+							   NULL,
+							   false);
+			}
+		}
+
+		cm_core->stats_listen_destroyed++;
+		kfree(listener);
+		cm_core->stats_listen_nodes_destroyed++;
+		listener = NULL;
+		ret = 0;
+	}
+
+	if (listener) {
+		if (atomic_read(&listener->pend_accepts_cnt) > 0)
+			i40iw_debug(cm_core->dev,
+				    I40IW_DEBUG_CM,
+				    "%s: listener (%p) pending accepts=%u\n",
+				    __func__,
+				    listener,
+				    atomic_read(&listener->pend_accepts_cnt));
+	}
+
+	return ret;
+}
+
+/**
+ * i40iw_cm_del_listen - delete a linstener
+ * @cm_core: cm's core
+  * @listener: passive connection's listener
+ * @apbvt_del: flag to delete apbvt
+ */
+static int i40iw_cm_del_listen(struct i40iw_cm_core *cm_core,
+			       struct i40iw_cm_listener *listener,
+			       bool apbvt_del)
+{
+	listener->listener_state = I40IW_CM_LISTENER_PASSIVE_STATE;
+	listener->cm_id = NULL;	/* going to be destroyed pretty soon */
+	return i40iw_dec_refcnt_listen(cm_core, listener, 1, apbvt_del);
+}
+
+/**
+ * i40iw_addr_resolve_neigh - resolve neighbor address
+ * @iwdev: iwarp device structure
+ * @src_ip: local ip address
+ * @dst_ip: remote ip address
+ * @arpindex: if there is an arp entry
+ */
+static int i40iw_addr_resolve_neigh(struct i40iw_device *iwdev,
+				    u32 src_ip,
+				    u32 dst_ip,
+				    int arpindex)
+{
+	struct rtable *rt;
+	struct neighbour *neigh;
+	int rc = arpindex;
+	struct net_device *netdev = iwdev->netdev;
+	__be32 dst_ipaddr = htonl(dst_ip);
+	__be32 src_ipaddr = htonl(src_ip);
+
+	rt = ip_route_output(&init_net, dst_ipaddr, src_ipaddr, 0, 0);
+	if (IS_ERR(rt)) {
+		i40iw_pr_err("ip_route_output\n");
+		return rc;
+	}
+
+	if (netif_is_bond_slave(netdev))
+		netdev = netdev_master_upper_dev_get(netdev);
+
+	neigh = dst_neigh_lookup(&rt->dst, &dst_ipaddr);
+
+	rcu_read_lock();
+	if (neigh) {
+		if (neigh->nud_state & NUD_VALID) {
+			if (arpindex >= 0) {
+				if (ether_addr_equal(iwdev->arp_table[arpindex].mac_addr,
+						     neigh->ha))
+					/* Mac address same as arp table */
+					goto resolve_neigh_exit;
+				i40iw_manage_arp_cache(iwdev,
+						       iwdev->arp_table[arpindex].mac_addr,
+						       &dst_ip,
+						       true,
+						       I40IW_ARP_DELETE);
+			}
+
+			i40iw_manage_arp_cache(iwdev, neigh->ha, &dst_ip, true, I40IW_ARP_ADD);
+			rc = i40iw_arp_table(iwdev, &dst_ip, true, NULL, I40IW_ARP_RESOLVE);
+		} else {
+			neigh_event_send(neigh, NULL);
+		}
+	}
+ resolve_neigh_exit:
+
+	rcu_read_unlock();
+	if (neigh)
+		neigh_release(neigh);
+
+	ip_rt_put(rt);
+	return rc;
+}
+
+/**
+ * i40iw_get_dst_ipv6
+ */
+#if IS_ENABLED(CONFIG_IPV6)
+static struct dst_entry *i40iw_get_dst_ipv6(struct sockaddr_in6 *src_addr,
+					    struct sockaddr_in6 *dst_addr)
+{
+	struct dst_entry *dst;
+	struct flowi6 fl6;
+
+	memset(&fl6, 0, sizeof(fl6));
+	fl6.daddr = dst_addr->sin6_addr;
+	fl6.saddr = src_addr->sin6_addr;
+	if (ipv6_addr_type(&fl6.daddr) & IPV6_ADDR_LINKLOCAL)
+		fl6.flowi6_oif = dst_addr->sin6_scope_id;
+
+	dst = ip6_route_output(&init_net, NULL, &fl6);
+	return dst;
+}
+#endif
+
+/**
+ * i40iw_addr_resolve_neigh_ipv6 - resolve neighbor ipv6 address
+ * @iwdev: iwarp device structure
+ * @dst_ip: remote ip address
+ * @arpindex: if there is an arp entry
+ */
+#if IS_ENABLED(CONFIG_IPV6)
+static int i40iw_addr_resolve_neigh_ipv6(struct i40iw_device *iwdev,
+					 u32 *src,
+					 u32 *dest,
+					 int arpindex)
+{
+	struct neighbour *neigh;
+	int rc = arpindex;
+	struct net_device *netdev = iwdev->netdev;
+	struct dst_entry *dst;
+	struct sockaddr_in6 dst_addr;
+	struct sockaddr_in6 src_addr;
+
+	memset(&dst_addr, 0, sizeof(dst_addr));
+	dst_addr.sin6_family = AF_INET6;
+	i40iw_copy_ip_htonl(dst_addr.sin6_addr.in6_u.u6_addr32, dest);
+	memset(&src_addr, 0, sizeof(src_addr));
+	src_addr.sin6_family = AF_INET6;
+	i40iw_copy_ip_htonl(src_addr.sin6_addr.in6_u.u6_addr32, src);
+	dst = i40iw_get_dst_ipv6(&src_addr, &dst_addr);
+	if (!dst || dst->error) {
+		if (dst) {
+			dst_release(dst);
+			i40iw_pr_err("ip6_route_output returned dst->error = %d\n",
+				     dst->error);
+		}
+		return rc;
+	}
+
+	if (netif_is_bond_slave(netdev))
+		netdev = netdev_master_upper_dev_get(netdev);
+
+	neigh = dst_neigh_lookup(dst, &dst_addr);
+
+	rcu_read_lock();
+	if (neigh) {
+		i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_CM, "dst_neigh_lookup MAC=%pM\n", neigh->ha);
+		if (neigh->nud_state & NUD_VALID) {
+			if (arpindex >= 0) {
+				if (ether_addr_equal
+				    (iwdev->arp_table[arpindex].mac_addr,
+				     neigh->ha)) {
+					/* Mac address same as in arp table */
+					goto resolve_neigh_exit6;
+				}
+				i40iw_manage_arp_cache(iwdev,
+						       iwdev->arp_table[arpindex].mac_addr,
+						       dest,
+						       false,
+						       I40IW_ARP_DELETE);
+			}
+			i40iw_manage_arp_cache(iwdev,
+					       neigh->ha,
+					       dest,
+					       false,
+					       I40IW_ARP_ADD);
+			rc = i40iw_arp_table(iwdev,
+					     dest,
+					     false,
+					     NULL,
+					     I40IW_ARP_RESOLVE);
+		} else {
+			neigh_event_send(neigh, NULL);
+		}
+	}
+
+ resolve_neigh_exit6:
+	rcu_read_unlock();
+	if (neigh)
+		neigh_release(neigh);
+	dst_release(dst);
+	return rc;
+}
+#endif
+
+/**
+ * i40iw_ipv4_is_loopback - check if loopback
+ * @loc_addr: local addr to compare
+ * @rem_addr: remote address
+ */
+static bool i40iw_ipv4_is_loopback(u32 loc_addr, u32 rem_addr)
+{
+	return ipv4_is_loopback(htonl(rem_addr)) || (loc_addr == rem_addr);
+}
+
+/**
+ * i40iw_ipv6_is_loopback - check if loopback
+ * @loc_addr: local addr to compare
+ * @rem_addr: remote address
+ */
+static bool i40iw_ipv6_is_loopback(u32 *loc_addr, u32 *rem_addr)
+{
+	struct in6_addr raddr6;
+
+	i40iw_copy_ip_htonl(raddr6.in6_u.u6_addr32, rem_addr);
+	return (!memcmp(loc_addr, rem_addr, 16) || ipv6_addr_loopback(&raddr6));
+}
+
+/**
+ * i40iw_make_cm_node - create a new instance of a cm node
+ * @cm_core: cm's core
+ * @iwdev: iwarp device structure
+ * @cm_info: quad info for connection
+ * @listener: passive connection's listener
+ */
+static struct i40iw_cm_node *i40iw_make_cm_node(
+				   struct i40iw_cm_core *cm_core,
+				   struct i40iw_device *iwdev,
+				   struct i40iw_cm_info *cm_info,
+				   struct i40iw_cm_listener *listener)
+{
+	struct i40iw_cm_node *cm_node;
+	struct timespec ts;
+	int oldarpindex;
+	int arpindex;
+	struct net_device *netdev = iwdev->netdev;
+
+	/* create an hte and cm_node for this instance */
+	cm_node = kzalloc(sizeof(*cm_node), GFP_ATOMIC);
+	if (!cm_node)
+		return NULL;
+
+	/* set our node specific transport info */
+	cm_node->ipv4 = cm_info->ipv4;
+	cm_node->vlan_id = cm_info->vlan_id;
+	memcpy(cm_node->loc_addr, cm_info->loc_addr, sizeof(cm_node->loc_addr));
+	memcpy(cm_node->rem_addr, cm_info->rem_addr, sizeof(cm_node->rem_addr));
+	cm_node->loc_port = cm_info->loc_port;
+	cm_node->rem_port = cm_info->rem_port;
+
+	cm_node->mpa_frame_rev = iwdev->mpa_version;
+	cm_node->send_rdma0_op = SEND_RDMA_READ_ZERO;
+	cm_node->ird_size = I40IW_MAX_IRD_SIZE;
+	cm_node->ord_size = I40IW_MAX_ORD_SIZE;
+
+	cm_node->listener = listener;
+	cm_node->cm_id = cm_info->cm_id;
+	ether_addr_copy(cm_node->loc_mac, netdev->dev_addr);
+	spin_lock_init(&cm_node->retrans_list_lock);
+
+	atomic_set(&cm_node->ref_count, 1);
+	/* associate our parent CM core */
+	cm_node->cm_core = cm_core;
+	cm_node->tcp_cntxt.loc_id = I40IW_CM_DEF_LOCAL_ID;
+	cm_node->tcp_cntxt.rcv_wscale = I40IW_CM_DEFAULT_RCV_WND_SCALE;
+	cm_node->tcp_cntxt.rcv_wnd =
+			I40IW_CM_DEFAULT_RCV_WND_SCALED >> I40IW_CM_DEFAULT_RCV_WND_SCALE;
+	ts = current_kernel_time();
+	cm_node->tcp_cntxt.loc_seq_num = htonl(ts.tv_nsec);
+	cm_node->tcp_cntxt.mss = iwdev->mss;
+
+	cm_node->iwdev = iwdev;
+	cm_node->dev = &iwdev->sc_dev;
+
+	if ((cm_node->ipv4 &&
+	     i40iw_ipv4_is_loopback(cm_node->loc_addr[0], cm_node->rem_addr[0])) ||
+	     (!cm_node->ipv4 && i40iw_ipv6_is_loopback(cm_node->loc_addr,
+						       cm_node->rem_addr))) {
+		arpindex = i40iw_arp_table(iwdev,
+					   cm_node->rem_addr,
+					   false,
+					   NULL,
+					   I40IW_ARP_RESOLVE);
+	} else {
+		oldarpindex = i40iw_arp_table(iwdev,
+					      cm_node->rem_addr,
+					      false,
+					      NULL,
+					      I40IW_ARP_RESOLVE);
+		if (cm_node->ipv4)
+			arpindex = i40iw_addr_resolve_neigh(iwdev,
+							    cm_info->loc_addr[0],
+							    cm_info->rem_addr[0],
+							    oldarpindex);
+#if IS_ENABLED(CONFIG_IPV6)
+		else
+			arpindex = i40iw_addr_resolve_neigh_ipv6(iwdev,
+								 cm_info->loc_addr,
+								 cm_info->rem_addr,
+								 oldarpindex);
+#endif
+	}
+	if (arpindex < 0) {
+		i40iw_pr_err("cm_node arpindex\n");
+		kfree(cm_node);
+		return NULL;
+	}
+	ether_addr_copy(cm_node->rem_mac, iwdev->arp_table[arpindex].mac_addr);
+	i40iw_add_hte_node(cm_core, cm_node);
+	cm_core->stats_nodes_created++;
+	return cm_node;
+}
+
+/**
+ * i40iw_rem_ref_cm_node - destroy an instance of a cm node
+ * @cm_node: connection's node
+ */
+static void i40iw_rem_ref_cm_node(struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_cm_core *cm_core = cm_node->cm_core;
+	struct i40iw_qp *iwqp;
+	struct i40iw_cm_info nfo;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cm_node->cm_core->ht_lock, flags);
+	if (atomic_dec_return(&cm_node->ref_count)) {
+		spin_unlock_irqrestore(&cm_node->cm_core->ht_lock, flags);
+		return;
+	}
+	list_del(&cm_node->list);
+	spin_unlock_irqrestore(&cm_node->cm_core->ht_lock, flags);
+
+	/* if the node is destroyed before connection was accelerated */
+	if (!cm_node->accelerated && cm_node->accept_pend) {
+		pr_err("node destroyed before established\n");
+		atomic_dec(&cm_node->listener->pend_accepts_cnt);
+	}
+	if (cm_node->close_entry)
+		i40iw_handle_close_entry(cm_node, 0);
+	if (cm_node->listener) {
+		i40iw_dec_refcnt_listen(cm_core, cm_node->listener, 0, true);
+	} else {
+		if (!i40iw_listen_port_in_use(cm_core, htons(cm_node->loc_port)) &&
+		    cm_node->apbvt_set && cm_node->iwdev) {
+			i40iw_manage_apbvt(cm_node->iwdev,
+					   cm_node->loc_port,
+					   I40IW_MANAGE_APBVT_DEL);
+			i40iw_get_addr_info(cm_node, &nfo);
+			if (cm_node->qhash_set) {
+				i40iw_manage_qhash(cm_node->iwdev,
+						   &nfo,
+						   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+						   I40IW_QHASH_MANAGE_TYPE_DELETE,
+						   NULL,
+						   false);
+				cm_node->qhash_set = 0;
+			}
+		}
+	}
+
+	iwqp = cm_node->iwqp;
+	if (iwqp) {
+		iwqp->cm_node = NULL;
+		i40iw_rem_ref(&iwqp->ibqp);
+		cm_node->iwqp = NULL;
+	} else if (cm_node->qhash_set) {
+		i40iw_get_addr_info(cm_node, &nfo);
+		i40iw_manage_qhash(cm_node->iwdev,
+				   &nfo,
+				   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+				   I40IW_QHASH_MANAGE_TYPE_DELETE,
+				   NULL,
+				   false);
+		cm_node->qhash_set = 0;
+	}
+
+	cm_node->cm_core->stats_nodes_destroyed++;
+	kfree(cm_node);
+}
+
+/**
+ * i40iw_handle_fin_pkt - FIN packet received
+ * @cm_node: connection's node
+ */
+static void i40iw_handle_fin_pkt(struct i40iw_cm_node *cm_node)
+{
+	u32 ret;
+
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_RCVD:
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_MPAREJ_RCVD:
+		cm_node->tcp_cntxt.rcv_nxt++;
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_LAST_ACK;
+		i40iw_send_fin(cm_node);
+		break;
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		i40iw_create_event(cm_node, I40IW_CM_EVENT_ABORTED);
+		cm_node->tcp_cntxt.rcv_nxt++;
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_FIN_WAIT1:
+		cm_node->tcp_cntxt.rcv_nxt++;
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSING;
+		i40iw_send_ack(cm_node);
+		/*
+		 * Wait for ACK as this is simultaneous close.
+		 * After we receive ACK, do not send anything.
+		 * Just rm the node.
+		 */
+		break;
+	case I40IW_CM_STATE_FIN_WAIT2:
+		cm_node->tcp_cntxt.rcv_nxt++;
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_TIME_WAIT;
+		i40iw_send_ack(cm_node);
+		ret =
+		    i40iw_schedule_cm_timer(cm_node, NULL, I40IW_TIMER_TYPE_CLOSE, 1, 0);
+		if (ret)
+			i40iw_pr_err("node %p state = %d\n", cm_node, cm_node->state);
+		break;
+	case I40IW_CM_STATE_TIME_WAIT:
+		cm_node->tcp_cntxt.rcv_nxt++;
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	case I40IW_CM_STATE_OFFLOADED:
+	default:
+		i40iw_pr_err("bad state node %p state = %d\n", cm_node, cm_node->state);
+		break;
+	}
+}
+
+/**
+ * i40iw_handle_rst_pkt - process received RST packet
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static void i40iw_handle_rst_pkt(struct i40iw_cm_node *cm_node,
+				 struct i40iw_puda_buf *rbuf)
+{
+	i40iw_cleanup_retrans_entry(cm_node);
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		switch (cm_node->mpa_frame_rev) {
+		case IETF_MPA_V2:
+			cm_node->mpa_frame_rev = IETF_MPA_V1;
+			/* send a syn and goto syn sent state */
+			cm_node->state = I40IW_CM_STATE_SYN_SENT;
+			if (i40iw_send_syn(cm_node, 0))
+				i40iw_active_open_err(cm_node, false);
+			break;
+		case IETF_MPA_V1:
+		default:
+			i40iw_active_open_err(cm_node, false);
+			break;
+		}
+		break;
+	case I40IW_CM_STATE_MPAREQ_RCVD:
+		atomic_add_return(1, &cm_node->passive_state);
+		break;
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_SYN_RCVD:
+	case I40IW_CM_STATE_LISTENING:
+		i40iw_pr_err("Bad state state = %d\n", cm_node->state);
+		i40iw_passive_open_err(cm_node, false);
+		break;
+	case I40IW_CM_STATE_OFFLOADED:
+		i40iw_active_open_err(cm_node, false);
+		break;
+	case I40IW_CM_STATE_CLOSED:
+		break;
+	case I40IW_CM_STATE_FIN_WAIT2:
+	case I40IW_CM_STATE_FIN_WAIT1:
+	case I40IW_CM_STATE_LAST_ACK:
+		cm_node->cm_id->rem_ref(cm_node->cm_id);
+	case I40IW_CM_STATE_TIME_WAIT:
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	default:
+		break;
+	}
+}
+
+/**
+ * i40iw_handle_rcv_mpa - Process a recv'd mpa buffer
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static void i40iw_handle_rcv_mpa(struct i40iw_cm_node *cm_node,
+				 struct i40iw_puda_buf *rbuf)
+{
+	int ret;
+	int datasize = rbuf->datalen;
+	u8 *dataloc = rbuf->data;
+
+	enum i40iw_cm_event_type type = I40IW_CM_EVENT_UNKNOWN;
+	u32 res_type;
+
+	ret = i40iw_parse_mpa(cm_node, dataloc, &res_type, datasize);
+	if (ret) {
+		if (cm_node->state == I40IW_CM_STATE_MPAREQ_SENT)
+			i40iw_active_open_err(cm_node, true);
+		else
+			i40iw_passive_open_err(cm_node, true);
+		return;
+	}
+
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_ESTABLISHED:
+		if (res_type == I40IW_MPA_REQUEST_REJECT)
+			i40iw_pr_err("state for reject\n");
+		cm_node->state = I40IW_CM_STATE_MPAREQ_RCVD;
+		type = I40IW_CM_EVENT_MPA_REQ;
+		i40iw_send_ack(cm_node);	/* ACK received MPA request */
+		atomic_set(&cm_node->passive_state,
+			   I40IW_PASSIVE_STATE_INDICATED);
+		break;
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		i40iw_cleanup_retrans_entry(cm_node);
+		if (res_type == I40IW_MPA_REQUEST_REJECT) {
+			type = I40IW_CM_EVENT_MPA_REJECT;
+			cm_node->state = I40IW_CM_STATE_MPAREJ_RCVD;
+		} else {
+			type = I40IW_CM_EVENT_CONNECTED;
+			cm_node->state = I40IW_CM_STATE_OFFLOADED;
+			i40iw_send_ack(cm_node);
+		}
+		break;
+	default:
+		pr_err("%s wrong cm_node state =%d\n", __func__, cm_node->state);
+		break;
+	}
+	i40iw_create_event(cm_node, type);
+}
+
+/**
+ * i40iw_indicate_pkt_err - Send up err event to cm
+ * @cm_node: connection's node
+ */
+static void i40iw_indicate_pkt_err(struct i40iw_cm_node *cm_node)
+{
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		i40iw_active_open_err(cm_node, true);
+		break;
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_SYN_RCVD:
+		i40iw_passive_open_err(cm_node, true);
+		break;
+	case I40IW_CM_STATE_OFFLOADED:
+	default:
+		break;
+	}
+}
+
+/**
+ * i40iw_check_syn - Check for error on received syn ack
+ * @cm_node: connection's node
+ * @tcph: pointer tcp header
+ */
+static int i40iw_check_syn(struct i40iw_cm_node *cm_node, struct tcphdr *tcph)
+{
+	int err = 0;
+
+	if (ntohl(tcph->ack_seq) != cm_node->tcp_cntxt.loc_seq_num) {
+		err = 1;
+		i40iw_active_open_err(cm_node, true);
+	}
+	return err;
+}
+
+/**
+ * i40iw_check_seq - check seq numbers if OK
+ * @cm_node: connection's node
+ * @tcph: pointer tcp header
+ */
+static int i40iw_check_seq(struct i40iw_cm_node *cm_node, struct tcphdr *tcph)
+{
+	int err = 0;
+	u32 seq;
+	u32 ack_seq;
+	u32 loc_seq_num = cm_node->tcp_cntxt.loc_seq_num;
+	u32 rcv_nxt = cm_node->tcp_cntxt.rcv_nxt;
+	u32 rcv_wnd;
+
+	seq = ntohl(tcph->seq);
+	ack_seq = ntohl(tcph->ack_seq);
+	rcv_wnd = cm_node->tcp_cntxt.rcv_wnd;
+	if (ack_seq != loc_seq_num)
+		err = -1;
+	else if (!between(seq, rcv_nxt, (rcv_nxt + rcv_wnd)))
+		err = -1;
+	if (err) {
+		i40iw_pr_err("seq number\n");
+		i40iw_indicate_pkt_err(cm_node);
+	}
+	return err;
+}
+
+/**
+ * i40iw_handle_syn_pkt - is for Passive node
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static void i40iw_handle_syn_pkt(struct i40iw_cm_node *cm_node,
+				 struct i40iw_puda_buf *rbuf)
+{
+	struct tcphdr *tcph = (struct tcphdr *)rbuf->tcph;
+	int ret;
+	u32 inc_sequence;
+	int optionsize;
+	struct i40iw_cm_info nfo;
+
+	optionsize = (tcph->doff << 2) - sizeof(struct tcphdr);
+	inc_sequence = ntohl(tcph->seq);
+
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		/* Rcvd syn on active open connection */
+		i40iw_active_open_err(cm_node, 1);
+		break;
+	case I40IW_CM_STATE_LISTENING:
+		/* Passive OPEN */
+		if (atomic_read(&cm_node->listener->pend_accepts_cnt) >
+		    cm_node->listener->backlog) {
+			cm_node->cm_core->stats_backlog_drops++;
+			i40iw_passive_open_err(cm_node, false);
+			break;
+		}
+		ret = i40iw_handle_tcp_options(cm_node, tcph, optionsize, 1);
+		if (ret) {
+			i40iw_passive_open_err(cm_node, false);
+			/* drop pkt */
+			break;
+		}
+		cm_node->tcp_cntxt.rcv_nxt = inc_sequence + 1;
+		cm_node->accept_pend = 1;
+		atomic_inc(&cm_node->listener->pend_accepts_cnt);
+
+		cm_node->state = I40IW_CM_STATE_SYN_RCVD;
+		i40iw_get_addr_info(cm_node, &nfo);
+		ret = i40iw_manage_qhash(cm_node->iwdev,
+					 &nfo,
+					 I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+					 I40IW_QHASH_MANAGE_TYPE_ADD,
+					 (void *)cm_node,
+					 false);
+		cm_node->qhash_set = true;
+		break;
+	case I40IW_CM_STATE_CLOSED:
+		i40iw_cleanup_retrans_entry(cm_node);
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_OFFLOADED:
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_FIN_WAIT1:
+	case I40IW_CM_STATE_FIN_WAIT2:
+	case I40IW_CM_STATE_MPAREQ_RCVD:
+	case I40IW_CM_STATE_LAST_ACK:
+	case I40IW_CM_STATE_CLOSING:
+	case I40IW_CM_STATE_UNKNOWN:
+	default:
+		break;
+	}
+}
+
+/**
+ * i40iw_handle_synack_pkt - Process SYN+ACK packet (active side)
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static void i40iw_handle_synack_pkt(struct i40iw_cm_node *cm_node,
+				    struct i40iw_puda_buf *rbuf)
+{
+	struct tcphdr *tcph = (struct tcphdr *)rbuf->tcph;
+	int ret;
+	u32 inc_sequence;
+	int optionsize;
+
+	optionsize = (tcph->doff << 2) - sizeof(struct tcphdr);
+	inc_sequence = ntohl(tcph->seq);
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_SENT:
+		i40iw_cleanup_retrans_entry(cm_node);
+		/* active open */
+		if (i40iw_check_syn(cm_node, tcph)) {
+			i40iw_pr_err("check syn fail\n");
+			return;
+		}
+		cm_node->tcp_cntxt.rem_ack_num = ntohl(tcph->ack_seq);
+		/* setup options */
+		ret = i40iw_handle_tcp_options(cm_node, tcph, optionsize, 0);
+		if (ret) {
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "cm_node=%p tcp_options failed\n",
+				    cm_node);
+			break;
+		}
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->tcp_cntxt.rcv_nxt = inc_sequence + 1;
+		i40iw_send_ack(cm_node);	/* ACK  for the syn_ack */
+		ret = i40iw_send_mpa_request(cm_node);
+		if (ret) {
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "cm_node=%p i40iw_send_mpa_request failed\n",
+				    cm_node);
+			break;
+		}
+		cm_node->state = I40IW_CM_STATE_MPAREQ_SENT;
+		break;
+	case I40IW_CM_STATE_MPAREQ_RCVD:
+		i40iw_passive_open_err(cm_node, true);
+		break;
+	case I40IW_CM_STATE_LISTENING:
+		cm_node->tcp_cntxt.loc_seq_num = ntohl(tcph->ack_seq);
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_CLOSED:
+		cm_node->tcp_cntxt.loc_seq_num = ntohl(tcph->ack_seq);
+		i40iw_cleanup_retrans_entry(cm_node);
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_FIN_WAIT1:
+	case I40IW_CM_STATE_FIN_WAIT2:
+	case I40IW_CM_STATE_LAST_ACK:
+	case I40IW_CM_STATE_OFFLOADED:
+	case I40IW_CM_STATE_CLOSING:
+	case I40IW_CM_STATE_UNKNOWN:
+	case I40IW_CM_STATE_MPAREQ_SENT:
+	default:
+		break;
+	}
+}
+
+/**
+ * i40iw_handle_ack_pkt - process packet with ACK
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static int i40iw_handle_ack_pkt(struct i40iw_cm_node *cm_node,
+				struct i40iw_puda_buf *rbuf)
+{
+	struct tcphdr *tcph = (struct tcphdr *)rbuf->tcph;
+	u32 inc_sequence;
+	int ret = 0;
+	int optionsize;
+	u32 datasize = rbuf->datalen;
+
+	optionsize = (tcph->doff << 2) - sizeof(struct tcphdr);
+
+	if (i40iw_check_seq(cm_node, tcph))
+		return -EINVAL;
+
+	inc_sequence = ntohl(tcph->seq);
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_RCVD:
+		i40iw_cleanup_retrans_entry(cm_node);
+		ret = i40iw_handle_tcp_options(cm_node, tcph, optionsize, 1);
+		if (ret)
+			break;
+		cm_node->tcp_cntxt.rem_ack_num = ntohl(tcph->ack_seq);
+		cm_node->state = I40IW_CM_STATE_ESTABLISHED;
+		if (datasize) {
+			cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize;
+			i40iw_handle_rcv_mpa(cm_node, rbuf);
+		}
+		break;
+	case I40IW_CM_STATE_ESTABLISHED:
+		i40iw_cleanup_retrans_entry(cm_node);
+		if (datasize) {
+			cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize;
+			i40iw_handle_rcv_mpa(cm_node, rbuf);
+		}
+		break;
+	case I40IW_CM_STATE_MPAREQ_SENT:
+		cm_node->tcp_cntxt.rem_ack_num = ntohl(tcph->ack_seq);
+		if (datasize) {
+			cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize;
+			i40iw_handle_rcv_mpa(cm_node, rbuf);
+		}
+		break;
+	case I40IW_CM_STATE_LISTENING:
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_CLOSED:
+		i40iw_cleanup_retrans_entry(cm_node);
+		atomic_inc(&cm_node->ref_count);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_LAST_ACK:
+	case I40IW_CM_STATE_CLOSING:
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_CLOSED;
+		if (!cm_node->accept_pend)
+			cm_node->cm_id->rem_ref(cm_node->cm_id);
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	case I40IW_CM_STATE_FIN_WAIT1:
+		i40iw_cleanup_retrans_entry(cm_node);
+		cm_node->state = I40IW_CM_STATE_FIN_WAIT2;
+		break;
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_FIN_WAIT2:
+	case I40IW_CM_STATE_OFFLOADED:
+	case I40IW_CM_STATE_MPAREQ_RCVD:
+	case I40IW_CM_STATE_UNKNOWN:
+	default:
+		i40iw_cleanup_retrans_entry(cm_node);
+		break;
+	}
+	return ret;
+}
+
+/**
+ * i40iw_process_packet - process cm packet
+ * @cm_node: connection's node
+ * @rbuf: receive buffer
+ */
+static void i40iw_process_packet(struct i40iw_cm_node *cm_node,
+				 struct i40iw_puda_buf *rbuf)
+{
+	enum i40iw_tcpip_pkt_type pkt_type = I40IW_PKT_TYPE_UNKNOWN;
+	struct tcphdr *tcph = (struct tcphdr *)rbuf->tcph;
+	u32 fin_set = 0;
+	int ret;
+
+	if (tcph->rst) {
+		pkt_type = I40IW_PKT_TYPE_RST;
+	} else if (tcph->syn) {
+		pkt_type = I40IW_PKT_TYPE_SYN;
+		if (tcph->ack)
+			pkt_type = I40IW_PKT_TYPE_SYNACK;
+	} else if (tcph->ack) {
+		pkt_type = I40IW_PKT_TYPE_ACK;
+	}
+	if (tcph->fin)
+		fin_set = 1;
+
+	switch (pkt_type) {
+	case I40IW_PKT_TYPE_SYN:
+		i40iw_handle_syn_pkt(cm_node, rbuf);
+		break;
+	case I40IW_PKT_TYPE_SYNACK:
+		i40iw_handle_synack_pkt(cm_node, rbuf);
+		break;
+	case I40IW_PKT_TYPE_ACK:
+		ret = i40iw_handle_ack_pkt(cm_node, rbuf);
+		if (fin_set && !ret)
+			i40iw_handle_fin_pkt(cm_node);
+		break;
+	case I40IW_PKT_TYPE_RST:
+		i40iw_handle_rst_pkt(cm_node, rbuf);
+		break;
+	default:
+		if (fin_set &&
+		    (!i40iw_check_seq(cm_node, (struct tcphdr *)rbuf->tcph)))
+			i40iw_handle_fin_pkt(cm_node);
+		break;
+	}
+}
+
+/**
+ * i40iw_make_listen_node - create a listen node with params
+ * @cm_core: cm's core
+ * @iwdev: iwarp device structure
+ * @cm_info: quad info for connection
+ */
+static struct i40iw_cm_listener *i40iw_make_listen_node(
+					struct i40iw_cm_core *cm_core,
+					struct i40iw_device *iwdev,
+					struct i40iw_cm_info *cm_info)
+{
+	struct i40iw_cm_listener *listener;
+	unsigned long flags;
+
+	/* cannot have multiple matching listeners */
+	listener = i40iw_find_listener(cm_core, cm_info->loc_addr,
+				       cm_info->loc_port,
+				       cm_info->vlan_id,
+				       I40IW_CM_LISTENER_EITHER_STATE);
+	if (listener &&
+	    (listener->listener_state == I40IW_CM_LISTENER_ACTIVE_STATE)) {
+		atomic_dec(&listener->ref_count);
+		i40iw_debug(cm_core->dev,
+			    I40IW_DEBUG_CM,
+			    "Not creating listener since it already exists\n");
+		return NULL;
+	}
+
+	if (!listener) {
+		/* create a CM listen node (1/2 node to compare incoming traffic to) */
+		listener = kzalloc(sizeof(*listener), GFP_ATOMIC);
+		if (!listener)
+			return NULL;
+		cm_core->stats_listen_nodes_created++;
+		memcpy(listener->loc_addr, cm_info->loc_addr, sizeof(listener->loc_addr));
+		listener->loc_port = cm_info->loc_port;
+
+		INIT_LIST_HEAD(&listener->child_listen_list);
+
+		atomic_set(&listener->ref_count, 1);
+	} else {
+		listener->reused_node = 1;
+	}
+
+	listener->cm_id = cm_info->cm_id;
+	listener->ipv4 = cm_info->ipv4;
+	listener->vlan_id = cm_info->vlan_id;
+	atomic_set(&listener->pend_accepts_cnt, 0);
+	listener->cm_core = cm_core;
+	listener->iwdev = iwdev;
+
+	listener->backlog = cm_info->backlog;
+	listener->listener_state = I40IW_CM_LISTENER_ACTIVE_STATE;
+
+	if (!listener->reused_node) {
+		spin_lock_irqsave(&cm_core->listen_list_lock, flags);
+		list_add(&listener->list, &cm_core->listen_nodes);
+		spin_unlock_irqrestore(&cm_core->listen_list_lock, flags);
+	}
+
+	return listener;
+}
+
+/**
+ * i40iw_create_cm_node - make a connection node with params
+ * @cm_core: cm's core
+ * @iwdev: iwarp device structure
+ * @private_data_len: len to provate data for mpa request
+ * @private_data: pointer to private data for connection
+ * @cm_info: quad info for connection
+ */
+static struct i40iw_cm_node *i40iw_create_cm_node(
+					struct i40iw_cm_core *cm_core,
+					struct i40iw_device *iwdev,
+					u16 private_data_len,
+					void *private_data,
+					struct i40iw_cm_info *cm_info)
+{
+	int ret;
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_cm_listener *loopback_remotelistener;
+	struct i40iw_cm_node *loopback_remotenode;
+	struct i40iw_cm_info loopback_cm_info;
+
+	/* create a CM connection node */
+	cm_node = i40iw_make_cm_node(cm_core, iwdev, cm_info, NULL);
+	if (!cm_node)
+		return NULL;
+	/* set our node side to client (active) side */
+	cm_node->tcp_cntxt.client = 1;
+	cm_node->tcp_cntxt.rcv_wscale = I40IW_CM_DEFAULT_RCV_WND_SCALE;
+
+	if (!memcmp(cm_info->loc_addr, cm_info->rem_addr, sizeof(cm_info->loc_addr))) {
+		loopback_remotelistener = i40iw_find_listener(
+						cm_core,
+						cm_info->rem_addr,
+						cm_node->rem_port,
+						cm_node->vlan_id,
+						I40IW_CM_LISTENER_ACTIVE_STATE);
+		if (!loopback_remotelistener) {
+			i40iw_create_event(cm_node, I40IW_CM_EVENT_ABORTED);
+		} else {
+			loopback_cm_info = *cm_info;
+			loopback_cm_info.loc_port = cm_info->rem_port;
+			loopback_cm_info.rem_port = cm_info->loc_port;
+			loopback_cm_info.cm_id = loopback_remotelistener->cm_id;
+			loopback_cm_info.ipv4 = cm_info->ipv4;
+			loopback_remotenode = i40iw_make_cm_node(cm_core,
+								 iwdev,
+								 &loopback_cm_info,
+								 loopback_remotelistener);
+			if (!loopback_remotenode) {
+				i40iw_rem_ref_cm_node(cm_node);
+				return NULL;
+			}
+			cm_core->stats_loopbacks++;
+			loopback_remotenode->loopbackpartner = cm_node;
+			loopback_remotenode->tcp_cntxt.rcv_wscale =
+				I40IW_CM_DEFAULT_RCV_WND_SCALE;
+			cm_node->loopbackpartner = loopback_remotenode;
+			memcpy(loopback_remotenode->pdata_buf, private_data,
+			       private_data_len);
+			loopback_remotenode->pdata.size = private_data_len;
+
+			cm_node->state = I40IW_CM_STATE_OFFLOADED;
+			cm_node->tcp_cntxt.rcv_nxt =
+				loopback_remotenode->tcp_cntxt.loc_seq_num;
+			loopback_remotenode->tcp_cntxt.rcv_nxt =
+				cm_node->tcp_cntxt.loc_seq_num;
+			cm_node->tcp_cntxt.max_snd_wnd =
+				loopback_remotenode->tcp_cntxt.rcv_wnd;
+			loopback_remotenode->tcp_cntxt.max_snd_wnd = cm_node->tcp_cntxt.rcv_wnd;
+			cm_node->tcp_cntxt.snd_wnd = loopback_remotenode->tcp_cntxt.rcv_wnd;
+			loopback_remotenode->tcp_cntxt.snd_wnd = cm_node->tcp_cntxt.rcv_wnd;
+			cm_node->tcp_cntxt.snd_wscale = loopback_remotenode->tcp_cntxt.rcv_wscale;
+			loopback_remotenode->tcp_cntxt.snd_wscale = cm_node->tcp_cntxt.rcv_wscale;
+			loopback_remotenode->state = I40IW_CM_STATE_MPAREQ_RCVD;
+			i40iw_create_event(loopback_remotenode, I40IW_CM_EVENT_MPA_REQ);
+		}
+		return cm_node;
+	}
+
+	cm_node->pdata.size = private_data_len;
+	cm_node->pdata.addr = cm_node->pdata_buf;
+
+	memcpy(cm_node->pdata_buf, private_data, private_data_len);
+
+	cm_node->state = I40IW_CM_STATE_SYN_SENT;
+	ret = i40iw_send_syn(cm_node, 0);
+
+	if (ret) {
+		if (cm_node->ipv4)
+			i40iw_debug(cm_node->dev,
+				    I40IW_DEBUG_CM,
+				    "Api - connect() FAILED: dest addr=%pI4",
+				    cm_node->rem_addr);
+		else
+			i40iw_debug(cm_node->dev, I40IW_DEBUG_CM,
+				    "Api - connect() FAILED: dest addr=%pI6",
+				    cm_node->rem_addr);
+		i40iw_rem_ref_cm_node(cm_node);
+		cm_node = NULL;
+	}
+
+	if (cm_node)
+		i40iw_debug(cm_node->dev,
+			    I40IW_DEBUG_CM,
+			    "Api - connect(): port=0x%04x, cm_node=%p, cm_id = %p.\n",
+			    cm_node->rem_port,
+			    cm_node,
+			    cm_node->cm_id);
+
+	return cm_node;
+}
+
+/**
+ * i40iw_cm_reject - reject and teardown a connection
+ * @cm_node: connection's node
+ * @pdate: ptr to private data for reject
+ * @plen: size of private data
+ */
+static int i40iw_cm_reject(struct i40iw_cm_node *cm_node, const void *pdata, u8 plen)
+{
+	int ret = 0;
+	int err;
+	int passive_state;
+	struct iw_cm_id *cm_id = cm_node->cm_id;
+	struct i40iw_cm_node *loopback = cm_node->loopbackpartner;
+
+	if (cm_node->tcp_cntxt.client)
+		return ret;
+	i40iw_cleanup_retrans_entry(cm_node);
+
+	if (!loopback) {
+		passive_state = atomic_add_return(1, &cm_node->passive_state);
+		if (passive_state == I40IW_SEND_RESET_EVENT) {
+			cm_node->state = I40IW_CM_STATE_CLOSED;
+			i40iw_rem_ref_cm_node(cm_node);
+		} else {
+			if (cm_node->state == I40IW_CM_STATE_LISTENER_DESTROYED) {
+				i40iw_rem_ref_cm_node(cm_node);
+			} else {
+				ret = i40iw_send_mpa_reject(cm_node, pdata, plen);
+				if (ret) {
+					cm_node->state = I40IW_CM_STATE_CLOSED;
+					err = i40iw_send_reset(cm_node);
+					if (err)
+						i40iw_pr_err("send reset failed\n");
+				} else {
+					cm_id->add_ref(cm_id);
+				}
+			}
+		}
+	} else {
+		cm_node->cm_id = NULL;
+		if (cm_node->state == I40IW_CM_STATE_LISTENER_DESTROYED) {
+			i40iw_rem_ref_cm_node(cm_node);
+			i40iw_rem_ref_cm_node(loopback);
+		} else {
+			ret = i40iw_send_cm_event(loopback,
+						  loopback->cm_id,
+						  IW_CM_EVENT_CONNECT_REPLY,
+						  -ECONNREFUSED);
+			i40iw_rem_ref_cm_node(cm_node);
+			loopback->state = I40IW_CM_STATE_CLOSING;
+
+			cm_id = loopback->cm_id;
+			i40iw_rem_ref_cm_node(loopback);
+			cm_id->rem_ref(cm_id);
+		}
+	}
+
+	return ret;
+}
+
+/**
+ * i40iw_cm_close - close of cm connection
+ * @cm_node: connection's node
+ */
+static int i40iw_cm_close(struct i40iw_cm_node *cm_node)
+{
+	int ret = 0;
+
+	if (!cm_node)
+		return -EINVAL;
+
+	switch (cm_node->state) {
+	case I40IW_CM_STATE_SYN_RCVD:
+	case I40IW_CM_STATE_SYN_SENT:
+	case I40IW_CM_STATE_ONE_SIDE_ESTABLISHED:
+	case I40IW_CM_STATE_ESTABLISHED:
+	case I40IW_CM_STATE_ACCEPTING:
+	case I40IW_CM_STATE_MPAREQ_SENT:
+	case I40IW_CM_STATE_MPAREQ_RCVD:
+		i40iw_cleanup_retrans_entry(cm_node);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_CLOSE_WAIT:
+		cm_node->state = I40IW_CM_STATE_LAST_ACK;
+		i40iw_send_fin(cm_node);
+		break;
+	case I40IW_CM_STATE_FIN_WAIT1:
+	case I40IW_CM_STATE_FIN_WAIT2:
+	case I40IW_CM_STATE_LAST_ACK:
+	case I40IW_CM_STATE_TIME_WAIT:
+	case I40IW_CM_STATE_CLOSING:
+		ret = -1;
+		break;
+	case I40IW_CM_STATE_LISTENING:
+		i40iw_cleanup_retrans_entry(cm_node);
+		i40iw_send_reset(cm_node);
+		break;
+	case I40IW_CM_STATE_MPAREJ_RCVD:
+	case I40IW_CM_STATE_UNKNOWN:
+	case I40IW_CM_STATE_INITED:
+	case I40IW_CM_STATE_CLOSED:
+	case I40IW_CM_STATE_LISTENER_DESTROYED:
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	case I40IW_CM_STATE_OFFLOADED:
+		if (cm_node->send_entry)
+			i40iw_pr_err("send_entry\n");
+		i40iw_rem_ref_cm_node(cm_node);
+		break;
+	}
+	return ret;
+}
+
+/**
+ * i40iw_receive_ilq - recv an ETHERNET packet, and process it
+ * through CM
+ * @dev: FPK dev struct
+ * @rbuf: receive buffer
+ */
+void i40iw_receive_ilq(struct i40iw_sc_dev *dev, struct i40iw_puda_buf *rbuf)
+{
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_cm_listener *listener;
+	struct iphdr *iph;
+	struct ipv6hdr *ip6h;
+	struct tcphdr *tcph;
+	struct i40iw_cm_info cm_info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+	struct i40iw_cm_core *cm_core = &iwdev->cm_core;
+	struct vlan_ethhdr *ethh;
+
+	/* if vlan, then maclen = 18 else 14 */
+	iph = (struct iphdr *)rbuf->iph;
+	memset(&cm_info, 0, sizeof(cm_info));
+
+	i40iw_debug_buf(dev,
+			I40IW_DEBUG_ILQ,
+			"RECEIVE ILQ BUFFER",
+			rbuf->mem.va,
+			rbuf->totallen);
+	ethh = (struct vlan_ethhdr *)rbuf->mem.va;
+
+	if (ethh->h_vlan_proto == htons(ETH_P_8021Q)) {
+		cm_info.vlan_id = ntohs(ethh->h_vlan_TCI) & VLAN_VID_MASK;
+		i40iw_debug(cm_core->dev,
+			    I40IW_DEBUG_CM,
+			    "%s vlan_id=%d\n",
+			    __func__,
+			    cm_info.vlan_id);
+	} else {
+		cm_info.vlan_id = I40IW_NO_VLAN;
+	}
+	tcph = (struct tcphdr *)rbuf->tcph;
+
+	if (rbuf->ipv4) {
+		cm_info.loc_addr[0] = ntohl(iph->daddr);
+		cm_info.rem_addr[0] = ntohl(iph->saddr);
+		cm_info.ipv4 = true;
+	} else {
+		ip6h = (struct ipv6hdr *)rbuf->iph;
+		i40iw_copy_ip_ntohl(cm_info.loc_addr,
+				    ip6h->daddr.in6_u.u6_addr32);
+		i40iw_copy_ip_ntohl(cm_info.rem_addr,
+				    ip6h->saddr.in6_u.u6_addr32);
+		cm_info.ipv4 = false;
+	}
+	cm_info.loc_port = ntohs(tcph->dest);
+	cm_info.rem_port = ntohs(tcph->source);
+	cm_node = i40iw_find_node(cm_core,
+				  cm_info.rem_port,
+				  cm_info.rem_addr,
+				  cm_info.loc_port,
+				  cm_info.loc_addr,
+				  true);
+
+	if (!cm_node) {
+		/* Only type of packet accepted are for */
+		/* the PASSIVE open (syn only) */
+		if (!tcph->syn || tcph->ack)
+			return;
+		listener =
+		    i40iw_find_listener(cm_core,
+					cm_info.loc_addr,
+					cm_info.loc_port,
+					cm_info.vlan_id,
+					I40IW_CM_LISTENER_ACTIVE_STATE);
+		if (!listener) {
+			cm_info.cm_id = NULL;
+			i40iw_debug(cm_core->dev,
+				    I40IW_DEBUG_CM,
+				    "%s no listener found\n",
+				    __func__);
+			return;
+		}
+		cm_info.cm_id = listener->cm_id;
+		cm_node = i40iw_make_cm_node(cm_core, iwdev, &cm_info, listener);
+		if (!cm_node) {
+			i40iw_debug(cm_core->dev,
+				    I40IW_DEBUG_CM,
+				    "%s allocate node failed\n",
+				    __func__);
+			atomic_dec(&listener->ref_count);
+			return;
+		}
+		if (!tcph->rst && !tcph->fin) {
+			cm_node->state = I40IW_CM_STATE_LISTENING;
+		} else {
+			i40iw_rem_ref_cm_node(cm_node);
+			return;
+		}
+		atomic_inc(&cm_node->ref_count);
+	} else if (cm_node->state == I40IW_CM_STATE_OFFLOADED) {
+		i40iw_rem_ref_cm_node(cm_node);
+		return;
+	}
+	i40iw_process_packet(cm_node, rbuf);
+	i40iw_rem_ref_cm_node(cm_node);
+}
+
+/**
+ * i40iw_setup_cm_core - allocate a top level instance of a cm
+ * core
+ * @iwdev: iwarp device structure
+ */
+void i40iw_setup_cm_core(struct i40iw_device *iwdev)
+{
+	struct i40iw_cm_core *cm_core = &iwdev->cm_core;
+
+	cm_core->iwdev = iwdev;
+	cm_core->dev = &iwdev->sc_dev;
+
+	INIT_LIST_HEAD(&cm_core->connected_nodes);
+	INIT_LIST_HEAD(&cm_core->listen_nodes);
+
+	init_timer(&cm_core->tcp_timer);
+	cm_core->tcp_timer.function = i40iw_cm_timer_tick;
+	cm_core->tcp_timer.data = (unsigned long)cm_core;
+
+	spin_lock_init(&cm_core->ht_lock);
+	spin_lock_init(&cm_core->listen_list_lock);
+
+	cm_core->event_wq = create_singlethread_workqueue("iwewq");
+	cm_core->disconn_wq = create_singlethread_workqueue("iwdwq");
+}
+
+/**
+ * i40iw_cleanup_cm_core - deallocate a top level instance of a
+ * cm core
+ * @cm_core: cm's core
+ */
+void i40iw_cleanup_cm_core(struct i40iw_cm_core *cm_core)
+{
+	unsigned long flags;
+
+	if (!cm_core)
+		return;
+
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	if (timer_pending(&cm_core->tcp_timer))
+		del_timer_sync(&cm_core->tcp_timer);
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+
+	destroy_workqueue(cm_core->event_wq);
+	destroy_workqueue(cm_core->disconn_wq);
+}
+
+/**
+ * i40iw_init_tcp_ctx - setup qp context
+ * @cm_node: connection's node
+ * @tcp_info: offload info for tcp
+ * @iwqp: associate qp for the connection
+ */
+static void i40iw_init_tcp_ctx(struct i40iw_cm_node *cm_node,
+			       struct i40iw_tcp_offload_info *tcp_info,
+			       struct i40iw_qp *iwqp)
+{
+	tcp_info->ipv4 = cm_node->ipv4;
+	tcp_info->drop_ooo_seg = true;
+	tcp_info->wscale = true;
+	tcp_info->ignore_tcp_opt = true;
+	tcp_info->ignore_tcp_uns_opt = true;
+	tcp_info->no_nagle = false;
+
+	tcp_info->ttl = I40IW_DEFAULT_TTL;
+	tcp_info->rtt_var = cpu_to_le32(I40IW_DEFAULT_RTT_VAR);
+	tcp_info->ss_thresh = cpu_to_le32(I40IW_DEFAULT_SS_THRESH);
+	tcp_info->rexmit_thresh = I40IW_DEFAULT_REXMIT_THRESH;
+
+	tcp_info->tcp_state = I40IW_TCP_STATE_ESTABLISHED;
+	tcp_info->snd_wscale = cm_node->tcp_cntxt.snd_wscale;
+	tcp_info->rcv_wscale = cm_node->tcp_cntxt.rcv_wscale;
+
+	tcp_info->snd_nxt = cpu_to_le32(cm_node->tcp_cntxt.loc_seq_num);
+	tcp_info->snd_wnd = cpu_to_le32(cm_node->tcp_cntxt.snd_wnd);
+	tcp_info->rcv_nxt = cpu_to_le32(cm_node->tcp_cntxt.rcv_nxt);
+	tcp_info->snd_max = cpu_to_le32(cm_node->tcp_cntxt.loc_seq_num);
+
+	tcp_info->snd_una = cpu_to_le32(cm_node->tcp_cntxt.loc_seq_num);
+	tcp_info->cwnd = cpu_to_le32(2 * cm_node->tcp_cntxt.mss);
+	tcp_info->snd_wl1 = cpu_to_le32(cm_node->tcp_cntxt.rcv_nxt);
+	tcp_info->snd_wl2 = cpu_to_le32(cm_node->tcp_cntxt.loc_seq_num);
+	tcp_info->max_snd_window = cpu_to_le32(cm_node->tcp_cntxt.max_snd_wnd);
+	tcp_info->rcv_wnd = cpu_to_le32(cm_node->tcp_cntxt.rcv_wnd <<
+					cm_node->tcp_cntxt.rcv_wscale);
+
+	tcp_info->flow_label = 0;
+	tcp_info->snd_mss = cpu_to_le32(((u32)cm_node->tcp_cntxt.mss));
+	if (cm_node->vlan_id < VLAN_TAG_PRESENT) {
+		tcp_info->insert_vlan_tag = true;
+		tcp_info->vlan_tag = cpu_to_le16(cm_node->vlan_id);
+	}
+	if (cm_node->ipv4) {
+		tcp_info->src_port = cpu_to_le16(cm_node->loc_port);
+		tcp_info->dst_port = cpu_to_le16(cm_node->rem_port);
+
+		tcp_info->dest_ip_addr3 = cpu_to_le32(cm_node->rem_addr[0]);
+		tcp_info->local_ipaddr3 = cpu_to_le32(cm_node->loc_addr[0]);
+		tcp_info->arp_idx = cpu_to_le32(i40iw_arp_table(iwqp->iwdev,
+								&tcp_info->dest_ip_addr3,
+								true,
+								NULL,
+								I40IW_ARP_RESOLVE));
+	} else {
+		tcp_info->src_port = cpu_to_le16(cm_node->loc_port);
+		tcp_info->dst_port = cpu_to_le16(cm_node->rem_port);
+		tcp_info->dest_ip_addr0 = cpu_to_le32(cm_node->rem_addr[0]);
+		tcp_info->dest_ip_addr1 = cpu_to_le32(cm_node->rem_addr[1]);
+		tcp_info->dest_ip_addr2 = cpu_to_le32(cm_node->rem_addr[2]);
+		tcp_info->dest_ip_addr3 = cpu_to_le32(cm_node->rem_addr[3]);
+		tcp_info->local_ipaddr0 = cpu_to_le32(cm_node->loc_addr[0]);
+		tcp_info->local_ipaddr1 = cpu_to_le32(cm_node->loc_addr[1]);
+		tcp_info->local_ipaddr2 = cpu_to_le32(cm_node->loc_addr[2]);
+		tcp_info->local_ipaddr3 = cpu_to_le32(cm_node->loc_addr[3]);
+		tcp_info->arp_idx = cpu_to_le32(i40iw_arp_table(
+							iwqp->iwdev,
+							&tcp_info->dest_ip_addr0,
+							false,
+							NULL,
+							I40IW_ARP_RESOLVE));
+	}
+}
+
+/**
+ * i40iw_cm_init_tsa_conn - setup qp for RTS
+ * @iwqp: associate qp for the connection
+ * @cm_node: connection's node
+ */
+static void i40iw_cm_init_tsa_conn(struct i40iw_qp *iwqp,
+				   struct i40iw_cm_node *cm_node)
+{
+	struct i40iw_tcp_offload_info tcp_info;
+	struct i40iwarp_offload_info *iwarp_info;
+	struct i40iw_qp_host_ctx_info *ctx_info;
+	struct i40iw_device *iwdev = iwqp->iwdev;
+	struct i40iw_sc_dev *dev = &iwqp->iwdev->sc_dev;
+
+	memset(&tcp_info, 0x00, sizeof(struct i40iw_tcp_offload_info));
+	iwarp_info = &iwqp->iwarp_info;
+	ctx_info = &iwqp->ctx_info;
+
+	ctx_info->tcp_info = &tcp_info;
+	ctx_info->send_cq_num = iwqp->iwscq->sc_cq.cq_uk.cq_id;
+	ctx_info->rcv_cq_num = iwqp->iwrcq->sc_cq.cq_uk.cq_id;
+
+	iwarp_info->ord_size = cm_node->ord_size;
+	iwarp_info->ird_size = i40iw_derive_hw_ird_setting(cm_node->ird_size);
+
+	if (iwarp_info->ord_size == 1)
+		iwarp_info->ord_size = 2;
+
+	iwarp_info->rd_enable = true;
+	iwarp_info->rdmap_ver = 1;
+	iwarp_info->ddp_ver = 1;
+
+	iwarp_info->pd_id = iwqp->iwpd->sc_pd.pd_id;
+
+	ctx_info->tcp_info_valid = true;
+	ctx_info->iwarp_info_valid = true;
+
+	i40iw_init_tcp_ctx(cm_node, &tcp_info, iwqp);
+	if (cm_node->snd_mark_en) {
+		iwarp_info->snd_mark_en = true;
+		iwarp_info->snd_mark_offset = (tcp_info.snd_nxt &
+				SNDMARKER_SEQNMASK) + cm_node->lsmm_size;
+	}
+
+	cm_node->state = I40IW_CM_STATE_OFFLOADED;
+	tcp_info.tcp_state = I40IW_TCP_STATE_ESTABLISHED;
+	tcp_info.src_mac_addr_idx = iwdev->mac_ip_table_idx;
+
+	dev->iw_priv_qp_ops->qp_setctx(&iwqp->sc_qp, (u64 *)(iwqp->host_ctx.va), ctx_info);
+
+	/* once tcp_info is set, no need to do it again */
+	ctx_info->tcp_info_valid = false;
+	ctx_info->iwarp_info_valid = false;
+}
+
+/**
+ * i40iw_cm_disconn - when a connection is being closed
+ * @iwqp: associate qp for the connection
+ */
+int i40iw_cm_disconn(struct i40iw_qp *iwqp)
+{
+	struct disconn_work *work;
+	struct i40iw_device *iwdev = iwqp->iwdev;
+	struct i40iw_cm_core *cm_core = &iwdev->cm_core;
+
+	work = kzalloc(sizeof(*work), GFP_ATOMIC);
+	if (!work)
+		return -ENOMEM;	/* Timer will clean up */
+
+	i40iw_add_ref(&iwqp->ibqp);
+	work->iwqp = iwqp;
+	INIT_WORK(&work->work, i40iw_disconnect_worker);
+	queue_work(cm_core->disconn_wq, &work->work);
+	return 0;
+}
+
+/**
+ * i40iw_loopback_nop - Send a nop
+ * @qp: associated hw qp
+ */
+static void i40iw_loopback_nop(struct i40iw_sc_qp *qp)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = qp->qp_uk.sq_base->elem;
+	set_64bit_val(wqe, 0, 0);
+	set_64bit_val(wqe, 8, 0);
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64(I40IWQP_OP_NOP, I40IWQPSQ_OPCODE) |
+	    LS_64(0, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+	set_64bit_val(wqe, 24, header);
+}
+
+/**
+ * i40iw_qp_disconnect - free qp and close cm
+ * @iwqp: associate qp for the connection
+ */
+static void i40iw_qp_disconnect(struct i40iw_qp *iwqp)
+{
+	struct i40iw_device *iwdev;
+	struct i40iw_ib_device *iwibdev;
+
+	iwdev = to_iwdev(iwqp->ibqp.device);
+	if (!iwdev) {
+		i40iw_pr_err("iwdev == NULL\n");
+		return;
+	}
+
+	iwibdev = iwdev->iwibdev;
+
+	if (iwqp->active_conn) {
+		/* indicate this connection is NOT active */
+		iwqp->active_conn = 0;
+	} else {
+		/* Need to free the Last Streaming Mode Message */
+		if (iwqp->ietf_mem.va) {
+			if (iwqp->lsmm_mr)
+				iwibdev->ibdev.dereg_mr(iwqp->lsmm_mr);
+			i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwqp->ietf_mem);
+		}
+	}
+
+	/* close the CM node down if it is still active */
+	if (iwqp->cm_node) {
+		i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_CM, "%s Call close API\n", __func__);
+		i40iw_cm_close(iwqp->cm_node);
+	}
+}
+
+/**
+ * i40iw_cm_disconn_true - called by worker thread to disconnect qp
+ * @iwqp: associate qp for the connection
+ */
+static void i40iw_cm_disconn_true(struct i40iw_qp *iwqp)
+{
+	struct iw_cm_id *cm_id;
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_qp *qp = &iwqp->sc_qp;
+	u16 last_ae;
+	u8 original_hw_tcp_state;
+	u8 original_ibqp_state;
+	int disconn_status = 0;
+	int issue_disconn = 0;
+	int issue_close = 0;
+	int issue_flush = 0;
+	struct ib_event ibevent;
+	unsigned long flags;
+	int ret;
+
+	if (!iwqp) {
+		i40iw_pr_err("iwqp == NULL\n");
+		return;
+	}
+
+	spin_lock_irqsave(&iwqp->lock, flags);
+	cm_id = iwqp->cm_id;
+	/* make sure we havent already closed this connection */
+	if (!cm_id) {
+		spin_unlock_irqrestore(&iwqp->lock, flags);
+		return;
+	}
+
+	iwdev = to_iwdev(iwqp->ibqp.device);
+
+	original_hw_tcp_state = iwqp->hw_tcp_state;
+	original_ibqp_state = iwqp->ibqp_state;
+	last_ae = iwqp->last_aeq;
+
+	if (qp->term_flags) {
+		issue_disconn = 1;
+		issue_close = 1;
+		iwqp->cm_id = NULL;
+		/*When term timer expires after cm_timer, don't want
+		 *terminate-handler to issue cm_disconn which can re-free
+		 *a QP even after its refcnt=0.
+		 */
+		del_timer(&iwqp->terminate_timer);
+		if (!iwqp->flush_issued) {
+			iwqp->flush_issued = 1;
+			issue_flush = 1;
+		}
+	} else if ((original_hw_tcp_state == I40IW_TCP_STATE_CLOSE_WAIT) ||
+		   ((original_ibqp_state == IB_QPS_RTS) &&
+		    (last_ae == I40IW_AE_LLP_CONNECTION_RESET))) {
+		issue_disconn = 1;
+		if (last_ae == I40IW_AE_LLP_CONNECTION_RESET)
+			disconn_status = -ECONNRESET;
+	}
+
+	if (((original_hw_tcp_state == I40IW_TCP_STATE_CLOSED) ||
+	     (original_hw_tcp_state == I40IW_TCP_STATE_TIME_WAIT) ||
+	     (last_ae == I40IW_AE_RDMAP_ROE_BAD_LLP_CLOSE) ||
+	     (last_ae == I40IW_AE_LLP_CONNECTION_RESET))) {
+		issue_close = 1;
+		iwqp->cm_id = NULL;
+		if (!iwqp->flush_issued) {
+			iwqp->flush_issued = 1;
+			issue_flush = 1;
+		}
+	}
+
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+	if (issue_flush && !iwqp->destroyed) {
+		/* Flush the queues */
+		i40iw_flush_wqes(iwdev, iwqp);
+
+		if (qp->term_flags) {
+			ibevent.device = iwqp->ibqp.device;
+			ibevent.event = (qp->eventtype == TERM_EVENT_QP_FATAL) ?
+					IB_EVENT_QP_FATAL : IB_EVENT_QP_ACCESS_ERR;
+			ibevent.element.qp = &iwqp->ibqp;
+			iwqp->ibqp.event_handler(&ibevent, iwqp->ibqp.qp_context);
+		}
+	}
+
+	if (cm_id && cm_id->event_handler) {
+		if (issue_disconn) {
+			ret = i40iw_send_cm_event(NULL,
+						  cm_id,
+						  IW_CM_EVENT_DISCONNECT,
+						  disconn_status);
+
+			if (ret)
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "disconnect event failed %s: - cm_id = %p\n",
+					    __func__, cm_id);
+		}
+		if (issue_close) {
+			i40iw_qp_disconnect(iwqp);
+			cm_id->provider_data = iwqp;
+			ret = i40iw_send_cm_event(NULL, cm_id, IW_CM_EVENT_CLOSE, 0);
+			if (ret)
+				i40iw_debug(&iwdev->sc_dev,
+					    I40IW_DEBUG_CM,
+					    "close event failed %s: - cm_id = %p\n",
+					    __func__, cm_id);
+			cm_id->rem_ref(cm_id);
+		}
+	}
+}
+
+/**
+ * i40iw_disconnect_worker - worker for connection close
+ * @work: points or disconn structure
+ */
+static void i40iw_disconnect_worker(struct work_struct *work)
+{
+	struct disconn_work *dwork = container_of(work, struct disconn_work, work);
+	struct i40iw_qp *iwqp = dwork->iwqp;
+
+	kfree(dwork);
+	i40iw_cm_disconn_true(iwqp);
+	i40iw_rem_ref(&iwqp->ibqp);
+}
+
+/**
+ * i40iw_accept - registered call for connection to be accepted
+ * @cm_id: cm information for passive connection
+ * @conn_param: accpet parameters
+ */
+int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
+{
+	struct ib_qp *ibqp;
+	struct i40iw_qp *iwqp;
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_cm_node *cm_node;
+	struct ib_qp_attr attr;
+	int passive_state;
+	struct i40iw_ib_device *iwibdev;
+	struct ib_mr *ibmr;
+	struct i40iw_pd *iwpd;
+	u16 buf_len = 0;
+	struct i40iw_kmem_info accept;
+	enum i40iw_status_code status;
+	u64 tagged_offset;
+
+	memset(&attr, 0, sizeof(attr));
+	ibqp = i40iw_get_qp(cm_id->device, conn_param->qpn);
+	if (!ibqp)
+		return -EINVAL;
+
+	iwqp = to_iwqp(ibqp);
+	iwdev = iwqp->iwdev;
+	dev = &iwdev->sc_dev;
+	cm_node = (struct i40iw_cm_node *)cm_id->provider_data;
+
+	if (((struct sockaddr_in *)&cm_id->local_addr)->sin_family == AF_INET) {
+		cm_node->ipv4 = true;
+		cm_node->vlan_id = i40iw_get_vlan_ipv4(cm_node->loc_addr);
+	} else {
+		cm_node->ipv4 = false;
+		i40iw_netdev_vlan_ipv6(cm_node->loc_addr, &cm_node->vlan_id, NULL);
+	}
+	i40iw_debug(cm_node->dev,
+		    I40IW_DEBUG_CM,
+		    "Accept vlan_id=%d\n",
+		    cm_node->vlan_id);
+	if (cm_node->state == I40IW_CM_STATE_LISTENER_DESTROYED) {
+		if (cm_node->loopbackpartner)
+			i40iw_rem_ref_cm_node(cm_node->loopbackpartner);
+		i40iw_rem_ref_cm_node(cm_node);
+		return -EINVAL;
+	}
+
+	passive_state = atomic_add_return(1, &cm_node->passive_state);
+	if (passive_state == I40IW_SEND_RESET_EVENT) {
+		i40iw_rem_ref_cm_node(cm_node);
+		return -ECONNRESET;
+	}
+
+	cm_node->cm_core->stats_accepts++;
+	iwqp->cm_node = (void *)cm_node;
+	cm_node->iwqp = iwqp;
+
+	buf_len = conn_param->private_data_len + I40IW_MAX_IETF_SIZE + MPA_ZERO_PAD_LEN;
+
+	status = i40iw_allocate_dma_mem(dev->hw, &iwqp->ietf_mem, buf_len, 1);
+
+	if (status)
+		return -ENOMEM;
+	cm_node->pdata.size = conn_param->private_data_len;
+	accept.addr = iwqp->ietf_mem.va;
+	accept.size = i40iw_cm_build_mpa_frame(cm_node, &accept, MPA_KEY_REPLY);
+	memcpy(accept.addr + accept.size, conn_param->private_data,
+	       conn_param->private_data_len);
+
+	/* setup our first outgoing iWarp send WQE (the IETF frame response) */
+	if ((cm_node->ipv4 &&
+	     !i40iw_ipv4_is_loopback(cm_node->loc_addr[0], cm_node->rem_addr[0])) ||
+	    (!cm_node->ipv4 &&
+	     !i40iw_ipv6_is_loopback(cm_node->loc_addr, cm_node->rem_addr))) {
+		iwibdev = iwdev->iwibdev;
+		iwpd = iwqp->iwpd;
+		tagged_offset = (uintptr_t)iwqp->ietf_mem.va;
+		ibmr = i40iw_reg_phys_mr(&iwpd->ibpd,
+					 iwqp->ietf_mem.pa,
+					 buf_len,
+					 IB_ACCESS_LOCAL_WRITE,
+					 &tagged_offset);
+		if (IS_ERR(ibmr)) {
+			i40iw_free_dma_mem(dev->hw, &iwqp->ietf_mem);
+			return -ENOMEM;
+		}
+
+		ibmr->pd = &iwpd->ibpd;
+		ibmr->device = iwpd->ibpd.device;
+		iwqp->lsmm_mr = ibmr;
+		if (iwqp->page)
+			iwqp->sc_qp.qp_uk.sq_base = kmap(iwqp->page);
+		if (is_remote_ne020_or_chelsio(cm_node))
+			dev->iw_priv_qp_ops->qp_send_lsmm(
+							&iwqp->sc_qp,
+							iwqp->ietf_mem.va,
+							(accept.size + conn_param->private_data_len),
+							ibmr->lkey);
+		else
+			dev->iw_priv_qp_ops->qp_send_lsmm(
+							&iwqp->sc_qp,
+							iwqp->ietf_mem.va,
+							(accept.size + conn_param->private_data_len + MPA_ZERO_PAD_LEN),
+							ibmr->lkey);
+
+	} else {
+		if (iwqp->page)
+			iwqp->sc_qp.qp_uk.sq_base = kmap(iwqp->page);
+		i40iw_loopback_nop(&iwqp->sc_qp);
+	}
+
+	if (iwqp->page)
+		kunmap(iwqp->page);
+
+	iwqp->cm_id = cm_id;
+	cm_node->cm_id = cm_id;
+
+	cm_id->provider_data = (void *)iwqp;
+	iwqp->active_conn = 0;
+
+	cm_node->lsmm_size = accept.size + conn_param->private_data_len;
+	i40iw_cm_init_tsa_conn(iwqp, cm_node);
+	cm_id->add_ref(cm_id);
+	i40iw_add_ref(&iwqp->ibqp);
+
+	i40iw_send_cm_event(cm_node, cm_id, IW_CM_EVENT_ESTABLISHED, 0);
+
+	attr.qp_state = IB_QPS_RTS;
+	cm_node->qhash_set = false;
+	i40iw_modify_qp(&iwqp->ibqp, &attr, IB_QP_STATE, NULL);
+	if (cm_node->loopbackpartner) {
+		cm_node->loopbackpartner->pdata.size = conn_param->private_data_len;
+
+		/* copy entire MPA frame to our cm_node's frame */
+		memcpy(cm_node->loopbackpartner->pdata_buf,
+		       conn_param->private_data,
+		       conn_param->private_data_len);
+		i40iw_create_event(cm_node->loopbackpartner, I40IW_CM_EVENT_CONNECTED);
+	}
+
+	cm_node->accelerated = 1;
+	if (cm_node->accept_pend) {
+		if (!cm_node->listener)
+			i40iw_pr_err("cm_node->listener NULL for passive node\n");
+		atomic_dec(&cm_node->listener->pend_accepts_cnt);
+		cm_node->accept_pend = 0;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_reject - registered call for connection to be rejected
+ * @cm_id: cm information for passive connection
+ * @pdata: private data to be sent
+ * @pdata_len: private data length
+ */
+int i40iw_reject(struct iw_cm_id *cm_id, const void *pdata, u8 pdata_len)
+{
+	struct i40iw_device *iwdev;
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_cm_node *loopback;
+
+	cm_node = (struct i40iw_cm_node *)cm_id->provider_data;
+	loopback = cm_node->loopbackpartner;
+	cm_node->cm_id = cm_id;
+	cm_node->pdata.size = pdata_len;
+
+	iwdev = to_iwdev(cm_id->device);
+	if (!iwdev)
+		return -EINVAL;
+	cm_node->cm_core->stats_rejects++;
+
+	if (pdata_len + sizeof(struct ietf_mpa_v2) > MAX_CM_BUFFER)
+		return -EINVAL;
+
+	if (loopback) {
+		memcpy(&loopback->pdata_buf, pdata, pdata_len);
+		loopback->pdata.size = pdata_len;
+	}
+
+	return i40iw_cm_reject(cm_node, pdata, pdata_len);
+}
+
+/**
+ * i40iw_connect - registered call for connection to be established
+ * @cm_id: cm information for passive connection
+ * @conn_param: Information about the connection
+ */
+int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
+{
+	struct ib_qp *ibqp;
+	struct i40iw_qp *iwqp;
+	struct i40iw_device *iwdev;
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_cm_info cm_info;
+	struct sockaddr_in *laddr;
+	struct sockaddr_in *raddr;
+	struct sockaddr_in6 *laddr6;
+	struct sockaddr_in6 *raddr6;
+	int apbvt_set = 0;
+	enum i40iw_status_code status;
+
+	ibqp = i40iw_get_qp(cm_id->device, conn_param->qpn);
+	if (!ibqp)
+		return -EINVAL;
+	iwqp = to_iwqp(ibqp);
+	if (!iwqp)
+		return -EINVAL;
+	iwdev = to_iwdev(iwqp->ibqp.device);
+	if (!iwdev)
+		return -EINVAL;
+
+	laddr = (struct sockaddr_in *)&cm_id->m_local_addr;
+	raddr = (struct sockaddr_in *)&cm_id->m_remote_addr;
+	laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr;
+	raddr6 = (struct sockaddr_in6 *)&cm_id->m_remote_addr;
+
+	if (!(laddr->sin_port) || !(raddr->sin_port))
+		return -EINVAL;
+
+	iwqp->active_conn = 1;
+	iwqp->cm_id = NULL;
+	cm_id->provider_data = iwqp;
+
+	/* set up the connection params for the node */
+	if (cm_id->remote_addr.ss_family == AF_INET) {
+		cm_info.ipv4 = true;
+		memset(cm_info.loc_addr, 0, sizeof(cm_info.loc_addr));
+		memset(cm_info.rem_addr, 0, sizeof(cm_info.rem_addr));
+		cm_info.loc_addr[0] = ntohl(laddr->sin_addr.s_addr);
+		cm_info.rem_addr[0] = ntohl(raddr->sin_addr.s_addr);
+		cm_info.loc_port = ntohs(laddr->sin_port);
+		cm_info.rem_port = ntohs(raddr->sin_port);
+		cm_info.vlan_id = i40iw_get_vlan_ipv4(cm_info.loc_addr);
+	} else {
+		cm_info.ipv4 = false;
+		i40iw_copy_ip_ntohl(cm_info.loc_addr,
+				    laddr6->sin6_addr.in6_u.u6_addr32);
+		i40iw_copy_ip_ntohl(cm_info.rem_addr,
+				    raddr6->sin6_addr.in6_u.u6_addr32);
+		cm_info.loc_port = ntohs(laddr6->sin6_port);
+		cm_info.rem_port = ntohs(raddr6->sin6_port);
+		i40iw_netdev_vlan_ipv6(cm_info.loc_addr, &cm_info.vlan_id, NULL);
+	}
+	cm_info.cm_id = cm_id;
+	if ((cm_info.ipv4 && (laddr->sin_addr.s_addr != raddr->sin_addr.s_addr)) ||
+	    (!cm_info.ipv4 && memcmp(laddr6->sin6_addr.in6_u.u6_addr32,
+				     raddr6->sin6_addr.in6_u.u6_addr32,
+				     sizeof(laddr6->sin6_addr.in6_u.u6_addr32)))) {
+		status = i40iw_manage_qhash(iwdev,
+					    &cm_info,
+					    I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+					    I40IW_QHASH_MANAGE_TYPE_ADD,
+					    NULL,
+					    true);
+		if (status)
+			return -EINVAL;
+	}
+	status = i40iw_manage_apbvt(iwdev, cm_info.loc_port, I40IW_MANAGE_APBVT_ADD);
+	if (status) {
+		i40iw_manage_qhash(iwdev,
+				   &cm_info,
+				   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+				   I40IW_QHASH_MANAGE_TYPE_DELETE,
+				   NULL,
+				   false);
+		return -EINVAL;
+	}
+
+	apbvt_set = 1;
+	cm_id->add_ref(cm_id);
+	cm_node = i40iw_create_cm_node(&iwdev->cm_core, iwdev,
+				       conn_param->private_data_len,
+				       (void *)conn_param->private_data,
+				       &cm_info);
+	if (!cm_node) {
+		i40iw_manage_qhash(iwdev,
+				   &cm_info,
+				   I40IW_QHASH_TYPE_TCP_ESTABLISHED,
+				   I40IW_QHASH_MANAGE_TYPE_DELETE,
+				   NULL,
+				   false);
+
+		if (apbvt_set && !i40iw_listen_port_in_use(&iwdev->cm_core,
+							   cm_info.loc_port))
+			i40iw_manage_apbvt(iwdev,
+					   cm_info.loc_port,
+					   I40IW_MANAGE_APBVT_DEL);
+		cm_id->rem_ref(cm_id);
+		iwdev->cm_core.stats_connect_errs++;
+		return -ENOMEM;
+	}
+
+	i40iw_record_ird_ord(cm_node, (u16)conn_param->ird, (u16)conn_param->ord);
+	if (cm_node->send_rdma0_op == SEND_RDMA_READ_ZERO &&
+	    !cm_node->ord_size)
+		cm_node->ord_size = 1;
+
+	cm_node->apbvt_set = apbvt_set;
+	cm_node->qhash_set = true;
+	iwqp->cm_node = cm_node;
+	cm_node->iwqp = iwqp;
+	iwqp->cm_id = cm_id;
+	i40iw_add_ref(&iwqp->ibqp);
+	return 0;
+}
+
+/**
+ * i40iw_create_listen - registered call creating listener
+ * @cm_id: cm information for passive connection
+ * @backlog: to max accept pending count
+ */
+int i40iw_create_listen(struct iw_cm_id *cm_id, int backlog)
+{
+	struct i40iw_device *iwdev;
+	struct i40iw_cm_listener *cm_listen_node;
+	struct i40iw_cm_info cm_info;
+	enum i40iw_status_code ret;
+	struct sockaddr_in *laddr;
+	struct sockaddr_in6 *laddr6;
+	bool wildcard = false;
+
+	iwdev = to_iwdev(cm_id->device);
+	if (!iwdev)
+		return -EINVAL;
+
+	laddr = (struct sockaddr_in *)&cm_id->m_local_addr;
+	laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr;
+	memset(&cm_info, 0, sizeof(cm_info));
+	if (laddr->sin_family == AF_INET) {
+		cm_info.ipv4 = true;
+		cm_info.loc_addr[0] = ntohl(laddr->sin_addr.s_addr);
+		cm_info.loc_port = ntohs(laddr->sin_port);
+
+		if (laddr->sin_addr.s_addr != INADDR_ANY)
+			cm_info.vlan_id = i40iw_get_vlan_ipv4(cm_info.loc_addr);
+		else
+			wildcard = true;
+
+	} else {
+		cm_info.ipv4 = false;
+		i40iw_copy_ip_ntohl(cm_info.loc_addr,
+				    laddr6->sin6_addr.in6_u.u6_addr32);
+		cm_info.loc_port = ntohs(laddr6->sin6_port);
+		if (ipv6_addr_type(&laddr6->sin6_addr) != IPV6_ADDR_ANY)
+			i40iw_netdev_vlan_ipv6(cm_info.loc_addr,
+					       &cm_info.vlan_id,
+					       NULL);
+		else
+			wildcard = true;
+	}
+	cm_info.backlog = backlog;
+	cm_info.cm_id = cm_id;
+
+	cm_listen_node = i40iw_make_listen_node(&iwdev->cm_core, iwdev, &cm_info);
+	if (!cm_listen_node) {
+		i40iw_pr_err("cm_listen_node == NULL\n");
+		return -ENOMEM;
+	}
+
+	cm_id->provider_data = cm_listen_node;
+
+	if (!cm_listen_node->reused_node) {
+		if (wildcard) {
+			if (cm_info.ipv4)
+				ret = i40iw_add_mqh_4(iwdev,
+						      &cm_info,
+						      cm_listen_node);
+			else
+				ret = i40iw_add_mqh_6(iwdev,
+						      &cm_info,
+						      cm_listen_node);
+			if (ret)
+				goto error;
+
+			ret = i40iw_manage_apbvt(iwdev,
+						 cm_info.loc_port,
+						 I40IW_MANAGE_APBVT_ADD);
+
+			if (ret)
+				goto error;
+		} else {
+			ret = i40iw_manage_qhash(iwdev,
+						 &cm_info,
+						 I40IW_QHASH_TYPE_TCP_SYN,
+						 I40IW_QHASH_MANAGE_TYPE_ADD,
+						 NULL,
+						 true);
+			if (ret)
+				goto error;
+			cm_listen_node->qhash_set = true;
+			ret = i40iw_manage_apbvt(iwdev,
+						 cm_info.loc_port,
+						 I40IW_MANAGE_APBVT_ADD);
+			if (ret)
+				goto error;
+		}
+	}
+	cm_id->add_ref(cm_id);
+	cm_listen_node->cm_core->stats_listen_created++;
+	return 0;
+ error:
+	i40iw_cm_del_listen(&iwdev->cm_core, (void *)cm_listen_node, false);
+	return -EINVAL;
+}
+
+/**
+ * i40iw_destroy_listen - registered call to destroy listener
+ * @cm_id: cm information for passive connection
+ */
+int i40iw_destroy_listen(struct iw_cm_id *cm_id)
+{
+	struct i40iw_device *iwdev;
+
+	iwdev = to_iwdev(cm_id->device);
+	if (cm_id->provider_data)
+		i40iw_cm_del_listen(&iwdev->cm_core, cm_id->provider_data, true);
+	else
+		i40iw_pr_err("cm_id->provider_data was NULL\n");
+
+	cm_id->rem_ref(cm_id);
+
+	return 0;
+}
+
+/**
+ * i40iw_cm_event_connected - handle connected active node
+ * @event: the info for cm_node of connection
+ */
+static void i40iw_cm_event_connected(struct i40iw_cm_event *event)
+{
+	struct i40iw_qp *iwqp;
+	struct i40iw_device *iwdev;
+	struct i40iw_cm_node *cm_node;
+	struct i40iw_sc_dev *dev;
+	struct ib_qp_attr attr;
+	struct iw_cm_id *cm_id;
+	int status;
+	bool read0;
+
+	cm_node = event->cm_node;
+	cm_id = cm_node->cm_id;
+	iwqp = (struct i40iw_qp *)cm_id->provider_data;
+	iwdev = to_iwdev(iwqp->ibqp.device);
+	dev = &iwdev->sc_dev;
+
+	if (iwqp->destroyed) {
+		status = -ETIMEDOUT;
+		goto error;
+	}
+	i40iw_cm_init_tsa_conn(iwqp, cm_node);
+	read0 = (cm_node->send_rdma0_op == SEND_RDMA_READ_ZERO);
+	if (iwqp->page)
+		iwqp->sc_qp.qp_uk.sq_base = kmap(iwqp->page);
+	dev->iw_priv_qp_ops->qp_send_rtt(&iwqp->sc_qp, read0);
+	if (iwqp->page)
+		kunmap(iwqp->page);
+	status = i40iw_send_cm_event(cm_node, cm_id, IW_CM_EVENT_CONNECT_REPLY, 0);
+	if (status)
+		i40iw_pr_err("send cm event\n");
+
+	memset(&attr, 0, sizeof(attr));
+	attr.qp_state = IB_QPS_RTS;
+	cm_node->qhash_set = false;
+	i40iw_modify_qp(&iwqp->ibqp, &attr, IB_QP_STATE, NULL);
+
+	cm_node->accelerated = 1;
+	if (cm_node->accept_pend) {
+		if (!cm_node->listener)
+			i40iw_pr_err("listener is null for passive node\n");
+		atomic_dec(&cm_node->listener->pend_accepts_cnt);
+		cm_node->accept_pend = 0;
+	}
+	return;
+
+error:
+	iwqp->cm_id = NULL;
+	cm_id->provider_data = NULL;
+	i40iw_send_cm_event(event->cm_node,
+			    cm_id,
+			    IW_CM_EVENT_CONNECT_REPLY,
+			    status);
+	cm_id->rem_ref(cm_id);
+	i40iw_rem_ref_cm_node(event->cm_node);
+}
+
+/**
+ * i40iw_cm_event_reset - handle reset
+ * @event: the info for cm_node of connection
+ */
+static void i40iw_cm_event_reset(struct i40iw_cm_event *event)
+{
+	struct i40iw_cm_node *cm_node = event->cm_node;
+	struct iw_cm_id   *cm_id = cm_node->cm_id;
+	struct i40iw_qp *iwqp;
+
+	if (!cm_id)
+		return;
+
+	iwqp = cm_id->provider_data;
+	if (!iwqp)
+		return;
+
+	i40iw_debug(cm_node->dev,
+		    I40IW_DEBUG_CM,
+		    "reset event %p - cm_id = %p\n",
+		     event->cm_node, cm_id);
+	iwqp->cm_id = NULL;
+
+	i40iw_send_cm_event(cm_node, cm_node->cm_id, IW_CM_EVENT_DISCONNECT, -ECONNRESET);
+	i40iw_send_cm_event(cm_node, cm_node->cm_id, IW_CM_EVENT_CLOSE, 0);
+}
+
+/**
+ * i40iw_cm_event_handler - worker thread callback to send event to cm upper layer
+ * @work: pointer of cm event info.
+ */
+static void i40iw_cm_event_handler(struct work_struct *work)
+{
+	struct i40iw_cm_event *event = container_of(work,
+						    struct i40iw_cm_event,
+						    event_work);
+	struct i40iw_cm_node *cm_node;
+
+	if (!event || !event->cm_node || !event->cm_node->cm_core)
+		return;
+
+	cm_node = event->cm_node;
+
+	switch (event->type) {
+	case I40IW_CM_EVENT_MPA_REQ:
+		i40iw_send_cm_event(cm_node,
+				    cm_node->cm_id,
+				    IW_CM_EVENT_CONNECT_REQUEST,
+				    0);
+		break;
+	case I40IW_CM_EVENT_RESET:
+		i40iw_cm_event_reset(event);
+		break;
+	case I40IW_CM_EVENT_CONNECTED:
+		if (!event->cm_node->cm_id ||
+		    (event->cm_node->state != I40IW_CM_STATE_OFFLOADED))
+			break;
+		i40iw_cm_event_connected(event);
+		break;
+	case I40IW_CM_EVENT_MPA_REJECT:
+		if (!event->cm_node->cm_id ||
+		    (cm_node->state == I40IW_CM_STATE_OFFLOADED))
+			break;
+		i40iw_send_cm_event(cm_node,
+				    cm_node->cm_id,
+				    IW_CM_EVENT_CONNECT_REPLY,
+				    -ECONNREFUSED);
+		break;
+	case I40IW_CM_EVENT_ABORTED:
+		if (!event->cm_node->cm_id ||
+		    (event->cm_node->state == I40IW_CM_STATE_OFFLOADED))
+			break;
+		i40iw_event_connect_error(event);
+		break;
+	default:
+		i40iw_pr_err("event type = %d\n", event->type);
+		break;
+	}
+
+	event->cm_info.cm_id->rem_ref(event->cm_info.cm_id);
+	i40iw_rem_ref_cm_node(event->cm_node);
+	kfree(event);
+}
+
+/**
+ * i40iw_cm_post_event - queue event request for worker thread
+ * @event: cm node's info for up event call
+ */
+static void i40iw_cm_post_event(struct i40iw_cm_event *event)
+{
+	atomic_inc(&event->cm_node->ref_count);
+	event->cm_info.cm_id->add_ref(event->cm_info.cm_id);
+	INIT_WORK(&event->event_work, i40iw_cm_event_handler);
+
+	queue_work(event->cm_node->cm_core->event_wq, &event->event_work);
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_cm.h b/drivers/infiniband/hw/i40iw/i40iw_cm.h
new file mode 100644
index 000000000000..5f8ceb4a8e84
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_cm.h
@@ -0,0 +1,456 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_CM_H
+#define I40IW_CM_H
+
+#define QUEUE_EVENTS
+
+#define I40IW_MANAGE_APBVT_DEL 0
+#define I40IW_MANAGE_APBVT_ADD 1
+
+#define I40IW_MPA_REQUEST_ACCEPT  1
+#define I40IW_MPA_REQUEST_REJECT  2
+
+/* IETF MPA -- defines, enums, structs */
+#define IEFT_MPA_KEY_REQ  "MPA ID Req Frame"
+#define IEFT_MPA_KEY_REP  "MPA ID Rep Frame"
+#define IETF_MPA_KEY_SIZE 16
+#define IETF_MPA_VERSION  1
+#define IETF_MAX_PRIV_DATA_LEN 512
+#define IETF_MPA_FRAME_SIZE    20
+#define IETF_RTR_MSG_SIZE      4
+#define IETF_MPA_V2_FLAG       0x10
+#define SNDMARKER_SEQNMASK     0x000001FF
+
+#define I40IW_MAX_IETF_SIZE      32
+
+#define MPA_ZERO_PAD_LEN	4
+
+/* IETF RTR MSG Fields               */
+#define IETF_PEER_TO_PEER       0x8000
+#define IETF_FLPDU_ZERO_LEN     0x4000
+#define IETF_RDMA0_WRITE        0x8000
+#define IETF_RDMA0_READ         0x4000
+#define IETF_NO_IRD_ORD         0x3FFF
+
+/* HW-supported IRD sizes*/
+#define	I40IW_HW_IRD_SETTING_2	2
+#define	I40IW_HW_IRD_SETTING_4	4
+#define	I40IW_HW_IRD_SETTING_8	8
+#define	I40IW_HW_IRD_SETTING_16	16
+#define	I40IW_HW_IRD_SETTING_32	32
+#define	I40IW_HW_IRD_SETTING_64	64
+
+enum ietf_mpa_flags {
+	IETF_MPA_FLAGS_MARKERS = 0x80,	/* receive Markers */
+	IETF_MPA_FLAGS_CRC = 0x40,	/* receive Markers */
+	IETF_MPA_FLAGS_REJECT = 0x20,	/* Reject */
+};
+
+struct ietf_mpa_v1 {
+	u8 key[IETF_MPA_KEY_SIZE];
+	u8 flags;
+	u8 rev;
+	__be16 priv_data_len;
+	u8 priv_data[0];
+};
+
+#define ietf_mpa_req_resp_frame ietf_mpa_frame
+
+struct ietf_rtr_msg {
+	__be16 ctrl_ird;
+	__be16 ctrl_ord;
+};
+
+struct ietf_mpa_v2 {
+	u8 key[IETF_MPA_KEY_SIZE];
+	u8 flags;
+	u8 rev;
+	__be16 priv_data_len;
+	struct ietf_rtr_msg rtr_msg;
+	u8 priv_data[0];
+};
+
+struct i40iw_cm_node;
+enum i40iw_timer_type {
+	I40IW_TIMER_TYPE_SEND,
+	I40IW_TIMER_TYPE_RECV,
+	I40IW_TIMER_NODE_CLEANUP,
+	I40IW_TIMER_TYPE_CLOSE,
+};
+
+#define I40IW_PASSIVE_STATE_INDICATED    0
+#define I40IW_DO_NOT_SEND_RESET_EVENT    1
+#define I40IW_SEND_RESET_EVENT           2
+
+#define MAX_I40IW_IFS 4
+
+#define SET_ACK 0x1
+#define SET_SYN 0x2
+#define SET_FIN 0x4
+#define SET_RST 0x8
+
+#define TCP_OPTIONS_PADDING     3
+
+struct option_base {
+	u8 optionnum;
+	u8 length;
+};
+
+enum option_numbers {
+	OPTION_NUMBER_END,
+	OPTION_NUMBER_NONE,
+	OPTION_NUMBER_MSS,
+	OPTION_NUMBER_WINDOW_SCALE,
+	OPTION_NUMBER_SACK_PERM,
+	OPTION_NUMBER_SACK,
+	OPTION_NUMBER_WRITE0 = 0xbc
+};
+
+struct option_mss {
+	u8 optionnum;
+	u8 length;
+	__be16 mss;
+};
+
+struct option_windowscale {
+	u8 optionnum;
+	u8 length;
+	u8 shiftcount;
+};
+
+union all_known_options {
+	char as_end;
+	struct option_base as_base;
+	struct option_mss as_mss;
+	struct option_windowscale as_windowscale;
+};
+
+struct i40iw_timer_entry {
+	struct list_head list;
+	unsigned long timetosend;	/* jiffies */
+	struct i40iw_puda_buf *sqbuf;
+	u32 type;
+	u32 retrycount;
+	u32 retranscount;
+	u32 context;
+	u32 send_retrans;
+	int close_when_complete;
+};
+
+#define I40IW_DEFAULT_RETRYS	64
+#define I40IW_DEFAULT_RETRANS	8
+#define I40IW_DEFAULT_TTL	0x40
+#define I40IW_DEFAULT_RTT_VAR	0x6
+#define I40IW_DEFAULT_SS_THRESH 0x3FFFFFFF
+#define I40IW_DEFAULT_REXMIT_THRESH 8
+
+#define I40IW_RETRY_TIMEOUT   HZ
+#define I40IW_SHORT_TIME      10
+#define I40IW_LONG_TIME       (2 * HZ)
+#define I40IW_MAX_TIMEOUT     ((unsigned long)(12 * HZ))
+
+#define I40IW_CM_HASHTABLE_SIZE         1024
+#define I40IW_CM_TCP_TIMER_INTERVAL     3000
+#define I40IW_CM_DEFAULT_MTU            1540
+#define I40IW_CM_DEFAULT_FRAME_CNT      10
+#define I40IW_CM_THREAD_STACK_SIZE      256
+#define I40IW_CM_DEFAULT_RCV_WND        64240
+#define I40IW_CM_DEFAULT_RCV_WND_SCALED 0x3fffc
+#define I40IW_CM_DEFAULT_RCV_WND_SCALE  2
+#define I40IW_CM_DEFAULT_FREE_PKTS      0x000A
+#define I40IW_CM_FREE_PKT_LO_WATERMARK  2
+
+#define I40IW_CM_DEFAULT_MSS   536
+
+#define I40IW_CM_DEF_SEQ       0x159bf75f
+#define I40IW_CM_DEF_LOCAL_ID  0x3b47
+
+#define I40IW_CM_DEF_SEQ2      0x18ed5740
+#define I40IW_CM_DEF_LOCAL_ID2 0xb807
+#define MAX_CM_BUFFER   (I40IW_MAX_IETF_SIZE + IETF_MAX_PRIV_DATA_LEN)
+
+typedef u32 i40iw_addr_t;
+
+#define i40iw_cm_tsa_context i40iw_qp_context
+
+struct i40iw_qp;
+
+/* cm node transition states */
+enum i40iw_cm_node_state {
+	I40IW_CM_STATE_UNKNOWN,
+	I40IW_CM_STATE_INITED,
+	I40IW_CM_STATE_LISTENING,
+	I40IW_CM_STATE_SYN_RCVD,
+	I40IW_CM_STATE_SYN_SENT,
+	I40IW_CM_STATE_ONE_SIDE_ESTABLISHED,
+	I40IW_CM_STATE_ESTABLISHED,
+	I40IW_CM_STATE_ACCEPTING,
+	I40IW_CM_STATE_MPAREQ_SENT,
+	I40IW_CM_STATE_MPAREQ_RCVD,
+	I40IW_CM_STATE_MPAREJ_RCVD,
+	I40IW_CM_STATE_OFFLOADED,
+	I40IW_CM_STATE_FIN_WAIT1,
+	I40IW_CM_STATE_FIN_WAIT2,
+	I40IW_CM_STATE_CLOSE_WAIT,
+	I40IW_CM_STATE_TIME_WAIT,
+	I40IW_CM_STATE_LAST_ACK,
+	I40IW_CM_STATE_CLOSING,
+	I40IW_CM_STATE_LISTENER_DESTROYED,
+	I40IW_CM_STATE_CLOSED
+};
+
+enum mpa_frame_version {
+	IETF_MPA_V1 = 1,
+	IETF_MPA_V2 = 2
+};
+
+enum mpa_frame_key {
+	MPA_KEY_REQUEST,
+	MPA_KEY_REPLY
+};
+
+enum send_rdma0 {
+	SEND_RDMA_READ_ZERO = 1,
+	SEND_RDMA_WRITE_ZERO = 2
+};
+
+enum i40iw_tcpip_pkt_type {
+	I40IW_PKT_TYPE_UNKNOWN,
+	I40IW_PKT_TYPE_SYN,
+	I40IW_PKT_TYPE_SYNACK,
+	I40IW_PKT_TYPE_ACK,
+	I40IW_PKT_TYPE_FIN,
+	I40IW_PKT_TYPE_RST
+};
+
+/* CM context params */
+struct i40iw_cm_tcp_context {
+	u8 client;
+
+	u32 loc_seq_num;
+	u32 loc_ack_num;
+	u32 rem_ack_num;
+	u32 rcv_nxt;
+
+	u32 loc_id;
+	u32 rem_id;
+
+	u32 snd_wnd;
+	u32 max_snd_wnd;
+
+	u32 rcv_wnd;
+	u32 mss;
+	u8 snd_wscale;
+	u8 rcv_wscale;
+
+	struct timeval sent_ts;
+};
+
+enum i40iw_cm_listener_state {
+	I40IW_CM_LISTENER_PASSIVE_STATE = 1,
+	I40IW_CM_LISTENER_ACTIVE_STATE = 2,
+	I40IW_CM_LISTENER_EITHER_STATE = 3
+};
+
+struct i40iw_cm_listener {
+	struct list_head list;
+	struct i40iw_cm_core *cm_core;
+	u8 loc_mac[ETH_ALEN];
+	u32 loc_addr[4];
+	u16 loc_port;
+	u32 map_loc_addr[4];
+	u16 map_loc_port;
+	struct iw_cm_id *cm_id;
+	atomic_t ref_count;
+	struct i40iw_device *iwdev;
+	atomic_t pend_accepts_cnt;
+	int backlog;
+	enum i40iw_cm_listener_state listener_state;
+	u32 reused_node;
+	u8 user_pri;
+	u16 vlan_id;
+	bool qhash_set;
+	bool ipv4;
+	struct list_head child_listen_list;
+
+};
+
+struct i40iw_kmem_info {
+	void *addr;
+	u32 size;
+};
+
+/* per connection node and node state information */
+struct i40iw_cm_node {
+	u32 loc_addr[4], rem_addr[4];
+	u16 loc_port, rem_port;
+	u32 map_loc_addr[4], map_rem_addr[4];
+	u16 map_loc_port, map_rem_port;
+	u16 vlan_id;
+	enum i40iw_cm_node_state state;
+	u8 loc_mac[ETH_ALEN];
+	u8 rem_mac[ETH_ALEN];
+	atomic_t ref_count;
+	struct i40iw_qp *iwqp;
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_cm_tcp_context tcp_cntxt;
+	struct i40iw_cm_core *cm_core;
+	struct i40iw_cm_node *loopbackpartner;
+	struct i40iw_timer_entry *send_entry;
+	struct i40iw_timer_entry *close_entry;
+	spinlock_t retrans_list_lock; /* cm transmit packet */
+	enum send_rdma0 send_rdma0_op;
+	u16 ird_size;
+	u16 ord_size;
+	u16     mpav2_ird_ord;
+	struct iw_cm_id *cm_id;
+	struct list_head list;
+	int accelerated;
+	struct i40iw_cm_listener *listener;
+	int apbvt_set;
+	int accept_pend;
+	struct list_head timer_entry;
+	struct list_head reset_entry;
+	atomic_t passive_state;
+	bool qhash_set;
+	u8 user_pri;
+	bool ipv4;
+	bool snd_mark_en;
+	u16 lsmm_size;
+	enum mpa_frame_version mpa_frame_rev;
+	struct i40iw_kmem_info pdata;
+	union {
+		struct ietf_mpa_v1 mpa_frame;
+		struct ietf_mpa_v2 mpa_v2_frame;
+	};
+
+	u8 pdata_buf[IETF_MAX_PRIV_DATA_LEN];
+	struct i40iw_kmem_info mpa_hdr;
+};
+
+/* structure for client or CM to fill when making CM api calls. */
+/*	- only need to set relevant data, based on op. */
+struct i40iw_cm_info {
+	struct iw_cm_id *cm_id;
+	u16 loc_port;
+	u16 rem_port;
+	u32 loc_addr[4];
+	u32 rem_addr[4];
+	u16 map_loc_port;
+	u16 map_rem_port;
+	u32 map_loc_addr[4];
+	u32 map_rem_addr[4];
+	u16 vlan_id;
+	int backlog;
+	u16 user_pri;
+	bool ipv4;
+};
+
+/* CM event codes */
+enum i40iw_cm_event_type {
+	I40IW_CM_EVENT_UNKNOWN,
+	I40IW_CM_EVENT_ESTABLISHED,
+	I40IW_CM_EVENT_MPA_REQ,
+	I40IW_CM_EVENT_MPA_CONNECT,
+	I40IW_CM_EVENT_MPA_ACCEPT,
+	I40IW_CM_EVENT_MPA_REJECT,
+	I40IW_CM_EVENT_MPA_ESTABLISHED,
+	I40IW_CM_EVENT_CONNECTED,
+	I40IW_CM_EVENT_RESET,
+	I40IW_CM_EVENT_ABORTED
+};
+
+/* event to post to CM event handler */
+struct i40iw_cm_event {
+	enum i40iw_cm_event_type type;
+	struct i40iw_cm_info cm_info;
+	struct work_struct event_work;
+	struct i40iw_cm_node *cm_node;
+};
+
+struct i40iw_cm_core {
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_dev *dev;
+
+	struct list_head listen_nodes;
+	struct list_head connected_nodes;
+
+	struct timer_list tcp_timer;
+
+	struct workqueue_struct *event_wq;
+	struct workqueue_struct *disconn_wq;
+
+	spinlock_t ht_lock; /* manage hash table */
+	spinlock_t listen_list_lock; /* listen list */
+
+	u64	stats_nodes_created;
+	u64	stats_nodes_destroyed;
+	u64	stats_listen_created;
+	u64	stats_listen_destroyed;
+	u64	stats_listen_nodes_created;
+	u64	stats_listen_nodes_destroyed;
+	u64	stats_loopbacks;
+	u64	stats_accepts;
+	u64	stats_rejects;
+	u64	stats_connect_errs;
+	u64	stats_passive_errs;
+	u64	stats_pkt_retrans;
+	u64	stats_backlog_drops;
+};
+
+int i40iw_schedule_cm_timer(struct i40iw_cm_node *cm_node,
+			    struct i40iw_puda_buf *sqbuf,
+			    enum i40iw_timer_type type,
+			    int send_retrans,
+			    int close_when_complete);
+
+int i40iw_accept(struct iw_cm_id *, struct iw_cm_conn_param *);
+int i40iw_reject(struct iw_cm_id *, const void *, u8);
+int i40iw_connect(struct iw_cm_id *, struct iw_cm_conn_param *);
+int i40iw_create_listen(struct iw_cm_id *, int);
+int i40iw_destroy_listen(struct iw_cm_id *);
+
+int i40iw_cm_start(struct i40iw_device *);
+int i40iw_cm_stop(struct i40iw_device *);
+
+int i40iw_arp_table(struct i40iw_device *iwdev,
+		    u32 *ip_addr,
+		    bool ipv4,
+		    u8 *mac_addr,
+		    u32 action);
+
+#endif /* I40IW_CM_H */
diff --git a/drivers/infiniband/hw/i40iw/i40iw_ctrl.c b/drivers/infiniband/hw/i40iw/i40iw_ctrl.c
new file mode 100644
index 000000000000..f05802bf6ca0
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_ctrl.c
@@ -0,0 +1,4743 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_status.h"
+#include "i40iw_hmc.h"
+
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_vf.h"
+#include "i40iw_virtchnl.h"
+
+/**
+ * i40iw_insert_wqe_hdr - write wqe header
+ * @wqe: cqp wqe for header
+ * @header: header for the cqp wqe
+ */
+static inline void i40iw_insert_wqe_hdr(u64 *wqe, u64 header)
+{
+	wmb();            /* make sure WQE is populated before polarity is set */
+	set_64bit_val(wqe, 24, header);
+}
+
+/**
+ * i40iw_get_cqp_reg_info - get head and tail for cqp using registers
+ * @cqp: struct for cqp hw
+ * @val: cqp tail register value
+ * @tail:wqtail register value
+ * @error: cqp processing err
+ */
+static inline void i40iw_get_cqp_reg_info(struct i40iw_sc_cqp *cqp,
+					  u32 *val,
+					  u32 *tail,
+					  u32 *error)
+{
+	if (cqp->dev->is_pf) {
+		*val = i40iw_rd32(cqp->dev->hw, I40E_PFPE_CQPTAIL);
+		*tail = RS_32(*val, I40E_PFPE_CQPTAIL_WQTAIL);
+		*error = RS_32(*val, I40E_PFPE_CQPTAIL_CQP_OP_ERR);
+	} else {
+		*val = i40iw_rd32(cqp->dev->hw, I40E_VFPE_CQPTAIL1);
+		*tail = RS_32(*val, I40E_VFPE_CQPTAIL_WQTAIL);
+		*error = RS_32(*val, I40E_VFPE_CQPTAIL_CQP_OP_ERR);
+	}
+}
+
+/**
+ * i40iw_cqp_poll_registers - poll cqp registers
+ * @cqp: struct for cqp hw
+ * @tail:wqtail register value
+ * @count: how many times to try for completion
+ */
+static enum i40iw_status_code i40iw_cqp_poll_registers(
+						struct i40iw_sc_cqp *cqp,
+						u32 tail,
+						u32 count)
+{
+	u32 i = 0;
+	u32 newtail, error, val;
+
+	while (i < count) {
+		i++;
+		i40iw_get_cqp_reg_info(cqp, &val, &newtail, &error);
+		if (error) {
+			error = (cqp->dev->is_pf) ?
+				 i40iw_rd32(cqp->dev->hw, I40E_PFPE_CQPERRCODES) :
+				 i40iw_rd32(cqp->dev->hw, I40E_VFPE_CQPERRCODES1);
+			return I40IW_ERR_CQP_COMPL_ERROR;
+		}
+		if (newtail != tail) {
+			/* SUCCESS */
+			I40IW_RING_MOVE_TAIL(cqp->sq_ring);
+			return 0;
+		}
+		udelay(I40IW_SLEEP_COUNT);
+	}
+	return I40IW_ERR_TIMEOUT;
+}
+
+/**
+ * i40iw_sc_parse_fpm_commit_buf - parse fpm commit buffer
+ * @buf: ptr to fpm commit buffer
+ * @info: ptr to i40iw_hmc_obj_info struct
+ *
+ * parses fpm commit info and copy base value
+ * of hmc objects in hmc_info
+ */
+static enum i40iw_status_code i40iw_sc_parse_fpm_commit_buf(
+				u64 *buf,
+				struct i40iw_hmc_obj_info *info)
+{
+	u64 temp;
+	u32 i, j;
+	u32 low;
+
+	/* copy base values in obj_info */
+	for (i = I40IW_HMC_IW_QP, j = 0;
+			i <= I40IW_HMC_IW_PBLE; i++, j += 8) {
+		get_64bit_val(buf, j, &temp);
+		info[i].base = RS_64_1(temp, 32) * 512;
+		low = (u32)(temp);
+		if (low)
+			info[i].cnt = low;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_sc_parse_fpm_query_buf() - parses fpm query buffer
+ * @buf: ptr to fpm query buffer
+ * @info: ptr to i40iw_hmc_obj_info struct
+ * @hmc_fpm_misc: ptr to fpm data
+ *
+ * parses fpm query buffer and copy max_cnt and
+ * size value of hmc objects in hmc_info
+ */
+static enum i40iw_status_code i40iw_sc_parse_fpm_query_buf(
+				u64 *buf,
+				struct i40iw_hmc_info *hmc_info,
+				struct i40iw_hmc_fpm_misc *hmc_fpm_misc)
+{
+	u64 temp;
+	struct i40iw_hmc_obj_info *obj_info;
+	u32 i, j, size;
+	u16 max_pe_sds;
+
+	obj_info = hmc_info->hmc_obj;
+
+	get_64bit_val(buf, 0, &temp);
+	hmc_info->first_sd_index = (u16)RS_64(temp, I40IW_QUERY_FPM_FIRST_PE_SD_INDEX);
+	max_pe_sds = (u16)RS_64(temp, I40IW_QUERY_FPM_MAX_PE_SDS);
+
+	/* Reduce SD count for VFs by 1 to account for PBLE backing page rounding */
+	if (hmc_info->hmc_fn_id >= I40IW_FIRST_VF_FPM_ID)
+		max_pe_sds--;
+	hmc_fpm_misc->max_sds = max_pe_sds;
+	hmc_info->sd_table.sd_cnt = max_pe_sds + hmc_info->first_sd_index;
+
+	for (i = I40IW_HMC_IW_QP, j = 8;
+	     i <= I40IW_HMC_IW_ARP; i++, j += 8) {
+		get_64bit_val(buf, j, &temp);
+		if (i == I40IW_HMC_IW_QP)
+			obj_info[i].max_cnt = (u32)RS_64(temp, I40IW_QUERY_FPM_MAX_QPS);
+		else if (i == I40IW_HMC_IW_CQ)
+			obj_info[i].max_cnt = (u32)RS_64(temp, I40IW_QUERY_FPM_MAX_CQS);
+		else
+			obj_info[i].max_cnt = (u32)temp;
+
+		size = (u32)RS_64_1(temp, 32);
+		obj_info[i].size = ((u64)1 << size);
+	}
+	for (i = I40IW_HMC_IW_MR, j = 48;
+			i <= I40IW_HMC_IW_PBLE; i++, j += 8) {
+		get_64bit_val(buf, j, &temp);
+		obj_info[i].max_cnt = (u32)temp;
+		size = (u32)RS_64_1(temp, 32);
+		obj_info[i].size = LS_64_1(1, size);
+	}
+
+	get_64bit_val(buf, 120, &temp);
+	hmc_fpm_misc->max_ceqs = (u8)RS_64(temp, I40IW_QUERY_FPM_MAX_CEQS);
+	get_64bit_val(buf, 120, &temp);
+	hmc_fpm_misc->ht_multiplier = RS_64(temp, I40IW_QUERY_FPM_HTMULTIPLIER);
+	get_64bit_val(buf, 120, &temp);
+	hmc_fpm_misc->timer_bucket = RS_64(temp, I40IW_QUERY_FPM_TIMERBUCKET);
+	get_64bit_val(buf, 64, &temp);
+	hmc_fpm_misc->xf_block_size = RS_64(temp, I40IW_QUERY_FPM_XFBLOCKSIZE);
+	if (!hmc_fpm_misc->xf_block_size)
+		return I40IW_ERR_INVALID_SIZE;
+	get_64bit_val(buf, 80, &temp);
+	hmc_fpm_misc->q1_block_size = RS_64(temp, I40IW_QUERY_FPM_Q1BLOCKSIZE);
+	if (!hmc_fpm_misc->q1_block_size)
+		return I40IW_ERR_INVALID_SIZE;
+	return 0;
+}
+
+/**
+ * i40iw_sc_pd_init - initialize sc pd struct
+ * @dev: sc device struct
+ * @pd: sc pd ptr
+ * @pd_id: pd_id for allocated pd
+ */
+static void i40iw_sc_pd_init(struct i40iw_sc_dev *dev,
+			     struct i40iw_sc_pd *pd,
+			     u16 pd_id)
+{
+	pd->size = sizeof(*pd);
+	pd->pd_id = pd_id;
+	pd->dev = dev;
+}
+
+/**
+ * i40iw_get_encoded_wqe_size - given wq size, returns hardware encoded size
+ * @wqsize: size of the wq (sq, rq, srq) to encoded_size
+ * @cqpsq: encoded size for sq for cqp as its encoded size is 1+ other wq's
+ */
+u8 i40iw_get_encoded_wqe_size(u32 wqsize, bool cqpsq)
+{
+	u8 encoded_size = 0;
+
+	/* cqp sq's hw coded value starts from 1 for size of 4
+	 * while it starts from 0 for qp' wq's.
+	 */
+	if (cqpsq)
+		encoded_size = 1;
+	wqsize >>= 2;
+	while (wqsize >>= 1)
+		encoded_size++;
+	return encoded_size;
+}
+
+/**
+ * i40iw_sc_cqp_init - Initialize buffers for a control Queue Pair
+ * @cqp: IWARP control queue pair pointer
+ * @info: IWARP control queue pair init info pointer
+ *
+ * Initializes the object and context buffers for a control Queue Pair.
+ */
+static enum i40iw_status_code i40iw_sc_cqp_init(struct i40iw_sc_cqp *cqp,
+						struct i40iw_cqp_init_info *info)
+{
+	u8 hw_sq_size;
+
+	if ((info->sq_size > I40IW_CQP_SW_SQSIZE_2048) ||
+	    (info->sq_size < I40IW_CQP_SW_SQSIZE_4) ||
+	    ((info->sq_size & (info->sq_size - 1))))
+		return I40IW_ERR_INVALID_SIZE;
+
+	hw_sq_size = i40iw_get_encoded_wqe_size(info->sq_size, true);
+	cqp->size = sizeof(*cqp);
+	cqp->sq_size = info->sq_size;
+	cqp->hw_sq_size = hw_sq_size;
+	cqp->sq_base = info->sq;
+	cqp->host_ctx = info->host_ctx;
+	cqp->sq_pa = info->sq_pa;
+	cqp->host_ctx_pa = info->host_ctx_pa;
+	cqp->dev = info->dev;
+	cqp->struct_ver = info->struct_ver;
+	cqp->scratch_array = info->scratch_array;
+	cqp->polarity = 0;
+	cqp->en_datacenter_tcp = info->en_datacenter_tcp;
+	cqp->enabled_vf_count = info->enabled_vf_count;
+	cqp->hmc_profile = info->hmc_profile;
+	info->dev->cqp = cqp;
+
+	I40IW_RING_INIT(cqp->sq_ring, cqp->sq_size);
+	i40iw_debug(cqp->dev, I40IW_DEBUG_WQE,
+		    "%s: sq_size[%04d] hw_sq_size[%04d] sq_base[%p] sq_pa[%llxh] cqp[%p] polarity[x%04X]\n",
+		    __func__, cqp->sq_size, cqp->hw_sq_size,
+		    cqp->sq_base, cqp->sq_pa, cqp, cqp->polarity);
+	return 0;
+}
+
+/**
+ * i40iw_sc_cqp_create - create cqp during bringup
+ * @cqp: struct for cqp hw
+ * @disable_pfpdus: if pfpdu to be disabled
+ * @maj_err: If error, major err number
+ * @min_err: If error, minor err number
+ */
+static enum i40iw_status_code i40iw_sc_cqp_create(struct i40iw_sc_cqp *cqp,
+						  bool disable_pfpdus,
+						  u16 *maj_err,
+						  u16 *min_err)
+{
+	u64 temp;
+	u32 cnt = 0, p1, p2, val = 0, err_code;
+	enum i40iw_status_code ret_code;
+
+	ret_code = i40iw_allocate_dma_mem(cqp->dev->hw,
+					  &cqp->sdbuf,
+					  128,
+					  I40IW_SD_BUF_ALIGNMENT);
+
+	if (ret_code)
+		goto exit;
+
+	temp = LS_64(cqp->hw_sq_size, I40IW_CQPHC_SQSIZE) |
+	       LS_64(cqp->struct_ver, I40IW_CQPHC_SVER);
+
+	if (disable_pfpdus)
+		temp |= LS_64(1, I40IW_CQPHC_DISABLE_PFPDUS);
+
+	set_64bit_val(cqp->host_ctx, 0, temp);
+	set_64bit_val(cqp->host_ctx, 8, cqp->sq_pa);
+	temp = LS_64(cqp->enabled_vf_count, I40IW_CQPHC_ENABLED_VFS) |
+	       LS_64(cqp->hmc_profile, I40IW_CQPHC_HMC_PROFILE);
+	set_64bit_val(cqp->host_ctx, 16, temp);
+	set_64bit_val(cqp->host_ctx, 24, (uintptr_t)cqp);
+	set_64bit_val(cqp->host_ctx, 32, 0);
+	set_64bit_val(cqp->host_ctx, 40, 0);
+	set_64bit_val(cqp->host_ctx, 48, 0);
+	set_64bit_val(cqp->host_ctx, 56, 0);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CQP_HOST_CTX",
+			cqp->host_ctx, I40IW_CQP_CTX_SIZE * 8);
+
+	p1 = RS_32_1(cqp->host_ctx_pa, 32);
+	p2 = (u32)cqp->host_ctx_pa;
+
+	if (cqp->dev->is_pf) {
+		i40iw_wr32(cqp->dev->hw, I40E_PFPE_CCQPHIGH, p1);
+		i40iw_wr32(cqp->dev->hw, I40E_PFPE_CCQPLOW, p2);
+	} else {
+		i40iw_wr32(cqp->dev->hw, I40E_VFPE_CCQPHIGH1, p1);
+		i40iw_wr32(cqp->dev->hw, I40E_VFPE_CCQPLOW1, p2);
+	}
+	do {
+		if (cnt++ > I40IW_DONE_COUNT) {
+			i40iw_free_dma_mem(cqp->dev->hw, &cqp->sdbuf);
+			ret_code = I40IW_ERR_TIMEOUT;
+			/*
+			 * read PFPE_CQPERRORCODES register to get the minor
+			 * and major error code
+			 */
+			if (cqp->dev->is_pf)
+				err_code = i40iw_rd32(cqp->dev->hw, I40E_PFPE_CQPERRCODES);
+			else
+				err_code = i40iw_rd32(cqp->dev->hw, I40E_VFPE_CQPERRCODES1);
+			*min_err = RS_32(err_code, I40E_PFPE_CQPERRCODES_CQP_MINOR_CODE);
+			*maj_err = RS_32(err_code, I40E_PFPE_CQPERRCODES_CQP_MAJOR_CODE);
+			goto exit;
+		}
+		udelay(I40IW_SLEEP_COUNT);
+		if (cqp->dev->is_pf)
+			val = i40iw_rd32(cqp->dev->hw, I40E_PFPE_CCQPSTATUS);
+		else
+			val = i40iw_rd32(cqp->dev->hw, I40E_VFPE_CCQPSTATUS1);
+	} while (!val);
+
+exit:
+	if (!ret_code)
+		cqp->process_cqp_sds = i40iw_update_sds_noccq;
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_cqp_post_sq - post of cqp's sq
+ * @cqp: struct for cqp hw
+ */
+void i40iw_sc_cqp_post_sq(struct i40iw_sc_cqp *cqp)
+{
+	if (cqp->dev->is_pf)
+		i40iw_wr32(cqp->dev->hw, I40E_PFPE_CQPDB, I40IW_RING_GETCURRENT_HEAD(cqp->sq_ring));
+	else
+		i40iw_wr32(cqp->dev->hw, I40E_VFPE_CQPDB1, I40IW_RING_GETCURRENT_HEAD(cqp->sq_ring));
+
+	i40iw_debug(cqp->dev,
+		    I40IW_DEBUG_WQE,
+		    "%s: HEAD_TAIL[%04d,%04d,%04d]\n",
+		    __func__,
+		    cqp->sq_ring.head,
+		    cqp->sq_ring.tail,
+		    cqp->sq_ring.size);
+}
+
+/**
+ * i40iw_sc_cqp_get_next_send_wqe - get next wqe on cqp sq
+ * @cqp: struct for cqp hw
+ * @wqe_idx: we index of cqp ring
+ */
+u64 *i40iw_sc_cqp_get_next_send_wqe(struct i40iw_sc_cqp *cqp, u64 scratch)
+{
+	u64 *wqe = NULL;
+	u32	wqe_idx;
+	enum i40iw_status_code ret_code;
+
+	if (I40IW_RING_FULL_ERR(cqp->sq_ring)) {
+		i40iw_debug(cqp->dev,
+			    I40IW_DEBUG_WQE,
+			    "%s: ring is full head %x tail %x size %x\n",
+			    __func__,
+			    cqp->sq_ring.head,
+			    cqp->sq_ring.tail,
+			    cqp->sq_ring.size);
+		return NULL;
+	}
+	I40IW_ATOMIC_RING_MOVE_HEAD(cqp->sq_ring, wqe_idx, ret_code);
+	if (ret_code)
+		return NULL;
+	if (!wqe_idx)
+		cqp->polarity = !cqp->polarity;
+
+	wqe = cqp->sq_base[wqe_idx].elem;
+	cqp->scratch_array[wqe_idx] = scratch;
+	I40IW_CQP_INIT_WQE(wqe);
+
+	return wqe;
+}
+
+/**
+ * i40iw_sc_cqp_destroy - destroy cqp during close
+ * @cqp: struct for cqp hw
+ */
+static enum i40iw_status_code i40iw_sc_cqp_destroy(struct i40iw_sc_cqp *cqp)
+{
+	u32 cnt = 0, val = 1;
+	enum i40iw_status_code ret_code = 0;
+	u32 cqpstat_addr;
+
+	if (cqp->dev->is_pf) {
+		i40iw_wr32(cqp->dev->hw, I40E_PFPE_CCQPHIGH, 0);
+		i40iw_wr32(cqp->dev->hw, I40E_PFPE_CCQPLOW, 0);
+		cqpstat_addr = I40E_PFPE_CCQPSTATUS;
+	} else {
+		i40iw_wr32(cqp->dev->hw, I40E_VFPE_CCQPHIGH1, 0);
+		i40iw_wr32(cqp->dev->hw, I40E_VFPE_CCQPLOW1, 0);
+		cqpstat_addr = I40E_VFPE_CCQPSTATUS1;
+	}
+	do {
+		if (cnt++ > I40IW_DONE_COUNT) {
+			ret_code = I40IW_ERR_TIMEOUT;
+			break;
+		}
+		udelay(I40IW_SLEEP_COUNT);
+		val = i40iw_rd32(cqp->dev->hw, cqpstat_addr);
+	} while (val);
+
+	i40iw_free_dma_mem(cqp->dev->hw, &cqp->sdbuf);
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_ccq_arm - enable intr for control cq
+ * @ccq: ccq sc struct
+ */
+static void i40iw_sc_ccq_arm(struct i40iw_sc_cq *ccq)
+{
+	u64 temp_val;
+	u16 sw_cq_sel;
+	u8 arm_next_se;
+	u8 arm_seq_num;
+
+	/* write to cq doorbell shadow area */
+	/* arm next se should always be zero */
+	get_64bit_val(ccq->cq_uk.shadow_area, 32, &temp_val);
+
+	sw_cq_sel = (u16)RS_64(temp_val, I40IW_CQ_DBSA_SW_CQ_SELECT);
+	arm_next_se = (u8)RS_64(temp_val, I40IW_CQ_DBSA_ARM_NEXT_SE);
+
+	arm_seq_num = (u8)RS_64(temp_val, I40IW_CQ_DBSA_ARM_SEQ_NUM);
+	arm_seq_num++;
+
+	temp_val = LS_64(arm_seq_num, I40IW_CQ_DBSA_ARM_SEQ_NUM) |
+		   LS_64(sw_cq_sel, I40IW_CQ_DBSA_SW_CQ_SELECT) |
+		   LS_64(arm_next_se, I40IW_CQ_DBSA_ARM_NEXT_SE) |
+		   LS_64(1, I40IW_CQ_DBSA_ARM_NEXT);
+
+	set_64bit_val(ccq->cq_uk.shadow_area, 32, temp_val);
+
+	wmb();       /* make sure shadow area is updated before arming */
+
+	if (ccq->dev->is_pf)
+		i40iw_wr32(ccq->dev->hw, I40E_PFPE_CQARM, ccq->cq_uk.cq_id);
+	else
+		i40iw_wr32(ccq->dev->hw, I40E_VFPE_CQARM1, ccq->cq_uk.cq_id);
+}
+
+/**
+ * i40iw_sc_ccq_get_cqe_info - get ccq's cq entry
+ * @ccq: ccq sc struct
+ * @info: completion q entry to return
+ */
+static enum i40iw_status_code i40iw_sc_ccq_get_cqe_info(
+					struct i40iw_sc_cq *ccq,
+					struct i40iw_ccq_cqe_info *info)
+{
+	u64 qp_ctx, temp, temp1;
+	u64 *cqe;
+	struct i40iw_sc_cqp *cqp;
+	u32 wqe_idx;
+	u8 polarity;
+	enum i40iw_status_code ret_code = 0;
+
+	if (ccq->cq_uk.avoid_mem_cflct)
+		cqe = (u64 *)I40IW_GET_CURRENT_EXTENDED_CQ_ELEMENT(&ccq->cq_uk);
+	else
+		cqe = (u64 *)I40IW_GET_CURRENT_CQ_ELEMENT(&ccq->cq_uk);
+
+	get_64bit_val(cqe, 24, &temp);
+	polarity = (u8)RS_64(temp, I40IW_CQ_VALID);
+	if (polarity != ccq->cq_uk.polarity)
+		return I40IW_ERR_QUEUE_EMPTY;
+
+	get_64bit_val(cqe, 8, &qp_ctx);
+	cqp = (struct i40iw_sc_cqp *)(unsigned long)qp_ctx;
+	info->error = (bool)RS_64(temp, I40IW_CQ_ERROR);
+	info->min_err_code = (u16)RS_64(temp, I40IW_CQ_MINERR);
+	if (info->error) {
+		info->maj_err_code = (u16)RS_64(temp, I40IW_CQ_MAJERR);
+		info->min_err_code = (u16)RS_64(temp, I40IW_CQ_MINERR);
+	}
+	wqe_idx = (u32)RS_64(temp, I40IW_CQ_WQEIDX);
+	info->scratch = cqp->scratch_array[wqe_idx];
+
+	get_64bit_val(cqe, 16, &temp1);
+	info->op_ret_val = (u32)RS_64(temp1, I40IW_CCQ_OPRETVAL);
+	get_64bit_val(cqp->sq_base[wqe_idx].elem, 24, &temp1);
+	info->op_code = (u8)RS_64(temp1, I40IW_CQPSQ_OPCODE);
+	info->cqp = cqp;
+
+	/*  move the head for cq */
+	I40IW_RING_MOVE_HEAD(ccq->cq_uk.cq_ring, ret_code);
+	if (I40IW_RING_GETCURRENT_HEAD(ccq->cq_uk.cq_ring) == 0)
+		ccq->cq_uk.polarity ^= 1;
+
+	/* update cq tail in cq shadow memory also */
+	I40IW_RING_MOVE_TAIL(ccq->cq_uk.cq_ring);
+	set_64bit_val(ccq->cq_uk.shadow_area,
+		      0,
+		      I40IW_RING_GETCURRENT_HEAD(ccq->cq_uk.cq_ring));
+	wmb(); /* write shadow area before tail */
+	I40IW_RING_MOVE_TAIL(cqp->sq_ring);
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_poll_for_cqp_op_done - Waits for last write to complete in CQP SQ
+ * @cqp: struct for cqp hw
+ * @op_code: cqp opcode for completion
+ * @info: completion q entry to return
+ */
+static enum i40iw_status_code i40iw_sc_poll_for_cqp_op_done(
+					struct i40iw_sc_cqp *cqp,
+					u8 op_code,
+					struct i40iw_ccq_cqe_info *compl_info)
+{
+	struct i40iw_ccq_cqe_info info;
+	struct i40iw_sc_cq *ccq;
+	enum i40iw_status_code ret_code = 0;
+	u32 cnt = 0;
+
+	memset(&info, 0, sizeof(info));
+	ccq = cqp->dev->ccq;
+	while (1) {
+		if (cnt++ > I40IW_DONE_COUNT)
+			return I40IW_ERR_TIMEOUT;
+
+		if (i40iw_sc_ccq_get_cqe_info(ccq, &info)) {
+			udelay(I40IW_SLEEP_COUNT);
+			continue;
+		}
+
+		if (info.error) {
+			ret_code = I40IW_ERR_CQP_COMPL_ERROR;
+			break;
+		}
+		/* check if opcode is cq create */
+		if (op_code != info.op_code) {
+			i40iw_debug(cqp->dev, I40IW_DEBUG_WQE,
+				    "%s: opcode mismatch for my op code 0x%x, returned opcode %x\n",
+				    __func__, op_code, info.op_code);
+		}
+		/* success, exit out of the loop */
+		if (op_code == info.op_code)
+			break;
+	}
+
+	if (compl_info)
+		memcpy(compl_info, &info, sizeof(*compl_info));
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_manage_push_page - Handle push page
+ * @cqp: struct for cqp hw
+ * @info: push page info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_manage_push_page(
+				struct i40iw_sc_cqp *cqp,
+				struct i40iw_cqp_manage_push_page_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	if (info->push_idx >= I40IW_MAX_PUSH_PAGE_COUNT)
+		return I40IW_ERR_INVALID_PUSH_PAGE_INDEX;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, info->qs_handle);
+
+	header = LS_64(info->push_idx, I40IW_CQPSQ_MPP_PPIDX) |
+		 LS_64(I40IW_CQP_OP_MANAGE_PUSH_PAGES, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID) |
+		 LS_64(info->free_page, I40IW_CQPSQ_MPP_FREE_PAGE);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE_PUSH_PAGES WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_manage_hmc_pm_func_table - manage of function table
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @vf_index: vf index for cqp
+ * @free_pm_fcn: function number
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_manage_hmc_pm_func_table(
+				struct i40iw_sc_cqp *cqp,
+				u64 scratch,
+				u8 vf_index,
+				bool free_pm_fcn,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	if (vf_index >= I40IW_MAX_VF_PER_PF)
+		return I40IW_ERR_INVALID_VF_ID;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	header = LS_64(vf_index, I40IW_CQPSQ_MHMC_VFIDX) |
+		 LS_64(I40IW_CQP_OP_MANAGE_HMC_PM_FUNC_TABLE, I40IW_CQPSQ_OPCODE) |
+		 LS_64(free_pm_fcn, I40IW_CQPSQ_MHMC_FREEPMFN) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE_HMC_PM_FUNC_TABLE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_set_hmc_resource_profile - cqp wqe for hmc profile
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @hmc_profile_type: type of profile to set
+ * @vf_num: vf number for profile
+ * @post_sq: flag for cqp db to ring
+ * @poll_registers: flag to poll register for cqp completion
+ */
+static enum i40iw_status_code i40iw_sc_set_hmc_resource_profile(
+				struct i40iw_sc_cqp *cqp,
+				u64 scratch,
+				u8 hmc_profile_type,
+				u8 vf_num, bool post_sq,
+				bool poll_registers)
+{
+	u64 *wqe;
+	u64 header;
+	u32 val, tail, error;
+	enum i40iw_status_code ret_code = 0;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16,
+		      (LS_64(hmc_profile_type, I40IW_CQPSQ_SHMCRP_HMC_PROFILE) |
+				LS_64(vf_num, I40IW_CQPSQ_SHMCRP_VFNUM)));
+
+	header = LS_64(I40IW_CQP_OP_SET_HMC_RESOURCE_PROFILE, I40IW_CQPSQ_OPCODE) |
+		       LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE_HMC_PM_FUNC_TABLE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+	if (error)
+		return I40IW_ERR_CQP_COMPL_ERROR;
+
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+		if (poll_registers)
+			ret_code = i40iw_cqp_poll_registers(cqp, tail, 1000000);
+		else
+			ret_code = i40iw_sc_poll_for_cqp_op_done(cqp,
+								 I40IW_CQP_OP_SHMC_PAGES_ALLOCATED,
+								 NULL);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_manage_hmc_pm_func_table_done - wait for cqp wqe completion for function table
+ * @cqp: struct for cqp hw
+ */
+static enum i40iw_status_code i40iw_sc_manage_hmc_pm_func_table_done(struct i40iw_sc_cqp *cqp)
+{
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_MANAGE_HMC_PM_FUNC_TABLE, NULL);
+}
+
+/**
+ * i40iw_sc_commit_fpm_values_done - wait for cqp eqe completion for fpm commit
+ * @cqp: struct for cqp hw
+ */
+static enum i40iw_status_code i40iw_sc_commit_fpm_values_done(struct i40iw_sc_cqp *cqp)
+{
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_COMMIT_FPM_VALUES, NULL);
+}
+
+/**
+ * i40iw_sc_commit_fpm_values - cqp wqe for commit fpm values
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @hmc_fn_id: hmc function id
+ * @commit_fpm_mem; Memory for fpm values
+ * @post_sq: flag for cqp db to ring
+ * @wait_type: poll ccq or cqp registers for cqp completion
+ */
+static enum i40iw_status_code i40iw_sc_commit_fpm_values(
+					struct i40iw_sc_cqp *cqp,
+					u64 scratch,
+					u8 hmc_fn_id,
+					struct i40iw_dma_mem *commit_fpm_mem,
+					bool post_sq,
+					u8 wait_type)
+{
+	u64 *wqe;
+	u64 header;
+	u32 tail, val, error;
+	enum i40iw_status_code ret_code = 0;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, hmc_fn_id);
+	set_64bit_val(wqe, 32, commit_fpm_mem->pa);
+
+	header = LS_64(I40IW_CQP_OP_COMMIT_FPM_VALUES, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "COMMIT_FPM_VALUES WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+	if (error)
+		return I40IW_ERR_CQP_COMPL_ERROR;
+
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+
+		if (wait_type == I40IW_CQP_WAIT_POLL_REGS)
+			ret_code = i40iw_cqp_poll_registers(cqp, tail, I40IW_DONE_COUNT);
+		else if (wait_type == I40IW_CQP_WAIT_POLL_CQ)
+			ret_code = i40iw_sc_commit_fpm_values_done(cqp);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_query_fpm_values_done - poll for cqp wqe completion for query fpm
+ * @cqp: struct for cqp hw
+ */
+static enum i40iw_status_code i40iw_sc_query_fpm_values_done(struct i40iw_sc_cqp *cqp)
+{
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_QUERY_FPM_VALUES, NULL);
+}
+
+/**
+ * i40iw_sc_query_fpm_values - cqp wqe query fpm values
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @hmc_fn_id: hmc function id
+ * @query_fpm_mem: memory for return fpm values
+ * @post_sq: flag for cqp db to ring
+ * @wait_type: poll ccq or cqp registers for cqp completion
+ */
+static enum i40iw_status_code i40iw_sc_query_fpm_values(
+					struct i40iw_sc_cqp *cqp,
+					u64 scratch,
+					u8 hmc_fn_id,
+					struct i40iw_dma_mem *query_fpm_mem,
+					bool post_sq,
+					u8 wait_type)
+{
+	u64 *wqe;
+	u64 header;
+	u32 tail, val, error;
+	enum i40iw_status_code ret_code = 0;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, hmc_fn_id);
+	set_64bit_val(wqe, 32, query_fpm_mem->pa);
+
+	header = LS_64(I40IW_CQP_OP_QUERY_FPM_VALUES, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QUERY_FPM WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	/* read the tail from CQP_TAIL register */
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+
+	if (error)
+		return I40IW_ERR_CQP_COMPL_ERROR;
+
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+		if (wait_type == I40IW_CQP_WAIT_POLL_REGS)
+			ret_code = i40iw_cqp_poll_registers(cqp, tail, I40IW_DONE_COUNT);
+		else if (wait_type == I40IW_CQP_WAIT_POLL_CQ)
+			ret_code = i40iw_sc_query_fpm_values_done(cqp);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_add_arp_cache_entry - cqp wqe add arp cache entry
+ * @cqp: struct for cqp hw
+ * @info: arp entry information
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_add_arp_cache_entry(
+				struct i40iw_sc_cqp *cqp,
+				struct i40iw_add_arp_cache_entry_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 temp, header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 8, info->reach_max);
+
+	temp = info->mac_addr[5] |
+	       LS_64_1(info->mac_addr[4], 8) |
+	       LS_64_1(info->mac_addr[3], 16) |
+	       LS_64_1(info->mac_addr[2], 24) |
+	       LS_64_1(info->mac_addr[1], 32) |
+	       LS_64_1(info->mac_addr[0], 40);
+
+	set_64bit_val(wqe, 16, temp);
+
+	header = info->arp_index |
+		 LS_64(I40IW_CQP_OP_MANAGE_ARP, I40IW_CQPSQ_OPCODE) |
+		 LS_64((info->permanent ? 1 : 0), I40IW_CQPSQ_MAT_PERMANENT) |
+		 LS_64(1, I40IW_CQPSQ_MAT_ENTRYVALID) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "ARP_CACHE_ENTRY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_del_arp_cache_entry - dele arp cache entry
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @arp_index: arp index to delete arp entry
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_del_arp_cache_entry(
+					struct i40iw_sc_cqp *cqp,
+					u64 scratch,
+					u16 arp_index,
+					bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	header = arp_index |
+		 LS_64(I40IW_CQP_OP_MANAGE_ARP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "ARP_CACHE_DEL_ENTRY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_query_arp_cache_entry - cqp wqe to query arp and arp index
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @arp_index: arp index to delete arp entry
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_query_arp_cache_entry(
+				struct i40iw_sc_cqp *cqp,
+				u64 scratch,
+				u16 arp_index,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	header = arp_index |
+		 LS_64(I40IW_CQP_OP_MANAGE_ARP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_MAT_QUERY) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QUERY_ARP_CACHE_ENTRY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_manage_apbvt_entry - for adding and deleting apbvt entries
+ * @cqp: struct for cqp hw
+ * @info: info for apbvt entry to add or delete
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_manage_apbvt_entry(
+				struct i40iw_sc_cqp *cqp,
+				struct i40iw_apbvt_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, info->port);
+
+	header = LS_64(I40IW_CQP_OP_MANAGE_APBVT, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->add, I40IW_CQPSQ_MAPT_ADDPORT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE_APBVT WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_manage_qhash_table_entry - manage quad hash entries
+ * @cqp: struct for cqp hw
+ * @info: info for quad hash to manage
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ *
+ * This is called before connection establishment is started. For passive connections, when
+ * listener is created, it will call with entry type of  I40IW_QHASH_TYPE_TCP_SYN with local
+ * ip address and tcp port. When SYN is received (passive connections) or
+ * sent (active connections), this routine is called with entry type of
+ * I40IW_QHASH_TYPE_TCP_ESTABLISHED and quad is passed in info.
+ *
+ * When iwarp connection is done and its state moves to RTS, the quad hash entry in
+ * the hardware will point to iwarp's qp number and requires no calls from the driver.
+ */
+static enum i40iw_status_code i40iw_sc_manage_qhash_table_entry(
+					struct i40iw_sc_cqp *cqp,
+					struct i40iw_qhash_table_info *info,
+					u64 scratch,
+					bool post_sq)
+{
+	u64 *wqe;
+	u64 qw1 = 0;
+	u64 qw2 = 0;
+	u64 temp;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	temp = info->mac_addr[5] |
+		LS_64_1(info->mac_addr[4], 8) |
+		LS_64_1(info->mac_addr[3], 16) |
+		LS_64_1(info->mac_addr[2], 24) |
+		LS_64_1(info->mac_addr[1], 32) |
+		LS_64_1(info->mac_addr[0], 40);
+
+	set_64bit_val(wqe, 0, temp);
+
+	qw1 = LS_64(info->qp_num, I40IW_CQPSQ_QHASH_QPN) |
+	      LS_64(info->dest_port, I40IW_CQPSQ_QHASH_DEST_PORT);
+	if (info->ipv4_valid) {
+		set_64bit_val(wqe,
+			      48,
+			      LS_64(info->dest_ip[0], I40IW_CQPSQ_QHASH_ADDR3));
+	} else {
+		set_64bit_val(wqe,
+			      56,
+			      LS_64(info->dest_ip[0], I40IW_CQPSQ_QHASH_ADDR0) |
+			      LS_64(info->dest_ip[1], I40IW_CQPSQ_QHASH_ADDR1));
+
+		set_64bit_val(wqe,
+			      48,
+			      LS_64(info->dest_ip[2], I40IW_CQPSQ_QHASH_ADDR2) |
+			      LS_64(info->dest_ip[3], I40IW_CQPSQ_QHASH_ADDR3));
+	}
+	qw2 = LS_64(cqp->dev->qs_handle, I40IW_CQPSQ_QHASH_QS_HANDLE);
+	if (info->vlan_valid)
+		qw2 |= LS_64(info->vlan_id, I40IW_CQPSQ_QHASH_VLANID);
+	set_64bit_val(wqe, 16, qw2);
+	if (info->entry_type == I40IW_QHASH_TYPE_TCP_ESTABLISHED) {
+		qw1 |= LS_64(info->src_port, I40IW_CQPSQ_QHASH_SRC_PORT);
+		if (!info->ipv4_valid) {
+			set_64bit_val(wqe,
+				      40,
+				      LS_64(info->src_ip[0], I40IW_CQPSQ_QHASH_ADDR0) |
+				      LS_64(info->src_ip[1], I40IW_CQPSQ_QHASH_ADDR1));
+			set_64bit_val(wqe,
+				      32,
+				      LS_64(info->src_ip[2], I40IW_CQPSQ_QHASH_ADDR2) |
+				      LS_64(info->src_ip[3], I40IW_CQPSQ_QHASH_ADDR3));
+		} else {
+			set_64bit_val(wqe,
+				      32,
+				      LS_64(info->src_ip[0], I40IW_CQPSQ_QHASH_ADDR3));
+		}
+	}
+
+	set_64bit_val(wqe, 8, qw1);
+	temp = LS_64(cqp->polarity, I40IW_CQPSQ_QHASH_WQEVALID) |
+	       LS_64(I40IW_CQP_OP_MANAGE_QUAD_HASH_TABLE_ENTRY, I40IW_CQPSQ_QHASH_OPCODE) |
+	       LS_64(info->manage, I40IW_CQPSQ_QHASH_MANAGE) |
+	       LS_64(info->ipv4_valid, I40IW_CQPSQ_QHASH_IPV4VALID) |
+	       LS_64(info->vlan_valid, I40IW_CQPSQ_QHASH_VLANVALID) |
+	       LS_64(info->entry_type, I40IW_CQPSQ_QHASH_ENTRYTYPE);
+
+	i40iw_insert_wqe_hdr(wqe, temp);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE_QHASH WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_alloc_local_mac_ipaddr_entry - cqp wqe for loc mac entry
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_alloc_local_mac_ipaddr_entry(
+					struct i40iw_sc_cqp *cqp,
+					u64 scratch,
+					bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	header = LS_64(I40IW_CQP_OP_ALLOCATE_LOC_MAC_IP_TABLE_ENTRY, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "ALLOCATE_LOCAL_MAC_IPADDR WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_add_local_mac_ipaddr_entry - add mac enry
+ * @cqp: struct for cqp hw
+ * @info:mac addr info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_add_local_mac_ipaddr_entry(
+				struct i40iw_sc_cqp *cqp,
+				struct i40iw_local_mac_ipaddr_entry_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 temp, header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	temp = info->mac_addr[5] |
+		LS_64_1(info->mac_addr[4], 8) |
+		LS_64_1(info->mac_addr[3], 16) |
+		LS_64_1(info->mac_addr[2], 24) |
+		LS_64_1(info->mac_addr[1], 32) |
+		LS_64_1(info->mac_addr[0], 40);
+
+	set_64bit_val(wqe, 32, temp);
+
+	header = LS_64(info->entry_idx, I40IW_CQPSQ_MLIPA_IPTABLEIDX) |
+		 LS_64(I40IW_CQP_OP_MANAGE_LOC_MAC_IP_TABLE, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "ADD_LOCAL_MAC_IPADDR WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_del_local_mac_ipaddr_entry - cqp wqe to dele local mac
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @entry_idx: index of mac entry
+ * @ ignore_ref_count: to force mac adde delete
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_del_local_mac_ipaddr_entry(
+				struct i40iw_sc_cqp *cqp,
+				u64 scratch,
+				u8 entry_idx,
+				u8 ignore_ref_count,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	header = LS_64(entry_idx, I40IW_CQPSQ_MLIPA_IPTABLEIDX) |
+		 LS_64(I40IW_CQP_OP_MANAGE_LOC_MAC_IP_TABLE, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_MLIPA_FREEENTRY) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID) |
+		 LS_64(ignore_ref_count, I40IW_CQPSQ_MLIPA_IGNORE_REF_CNT);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "DEL_LOCAL_MAC_IPADDR WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_cqp_nop - send a nop wqe
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_cqp_nop(struct i40iw_sc_cqp *cqp,
+					       u64 scratch,
+					       bool post_sq)
+{
+	u64 *wqe;
+	u64 header;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	header = LS_64(I40IW_CQP_OP_NOP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "NOP WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_ceq_init - initialize ceq
+ * @ceq: ceq sc structure
+ * @info: ceq initialization info
+ */
+static enum i40iw_status_code i40iw_sc_ceq_init(struct i40iw_sc_ceq *ceq,
+						struct i40iw_ceq_init_info *info)
+{
+	u32 pble_obj_cnt;
+
+	if ((info->elem_cnt < I40IW_MIN_CEQ_ENTRIES) ||
+	    (info->elem_cnt > I40IW_MAX_CEQ_ENTRIES))
+		return I40IW_ERR_INVALID_SIZE;
+
+	if (info->ceq_id >= I40IW_MAX_CEQID)
+		return I40IW_ERR_INVALID_CEQ_ID;
+
+	pble_obj_cnt = info->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->virtual_map && (info->first_pm_pbl_idx >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	ceq->size = sizeof(*ceq);
+	ceq->ceqe_base = (struct i40iw_ceqe *)info->ceqe_base;
+	ceq->ceq_id = info->ceq_id;
+	ceq->dev = info->dev;
+	ceq->elem_cnt = info->elem_cnt;
+	ceq->ceq_elem_pa = info->ceqe_pa;
+	ceq->virtual_map = info->virtual_map;
+
+	ceq->pbl_chunk_size = (ceq->virtual_map ? info->pbl_chunk_size : 0);
+	ceq->first_pm_pbl_idx = (ceq->virtual_map ? info->first_pm_pbl_idx : 0);
+	ceq->pbl_list = (ceq->virtual_map ? info->pbl_list : NULL);
+
+	ceq->tph_en = info->tph_en;
+	ceq->tph_val = info->tph_val;
+	ceq->polarity = 1;
+	I40IW_RING_INIT(ceq->ceq_ring, ceq->elem_cnt);
+	ceq->dev->ceq[info->ceq_id] = ceq;
+
+	return 0;
+}
+
+/**
+ * i40iw_sc_ceq_create - create ceq wqe
+ * @ceq: ceq sc structure
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_ceq_create(struct i40iw_sc_ceq *ceq,
+						  u64 scratch,
+						  bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+
+	cqp = ceq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, ceq->elem_cnt);
+	set_64bit_val(wqe, 32, (ceq->virtual_map ? 0 : ceq->ceq_elem_pa));
+	set_64bit_val(wqe, 48, (ceq->virtual_map ? ceq->first_pm_pbl_idx : 0));
+	set_64bit_val(wqe, 56, LS_64(ceq->tph_val, I40IW_CQPSQ_TPHVAL));
+
+	header = ceq->ceq_id |
+		 LS_64(I40IW_CQP_OP_CREATE_CEQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(ceq->pbl_chunk_size, I40IW_CQPSQ_CEQ_LPBLSIZE) |
+		 LS_64(ceq->virtual_map, I40IW_CQPSQ_CEQ_VMAP) |
+		 LS_64(ceq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CEQ_CREATE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_cceq_create_done - poll for control ceq wqe to complete
+ * @ceq: ceq sc structure
+ */
+static enum i40iw_status_code i40iw_sc_cceq_create_done(struct i40iw_sc_ceq *ceq)
+{
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = ceq->dev->cqp;
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_CREATE_CEQ, NULL);
+}
+
+/**
+ * i40iw_sc_cceq_destroy_done - poll for destroy cceq to complete
+ * @ceq: ceq sc structure
+ */
+static enum i40iw_status_code i40iw_sc_cceq_destroy_done(struct i40iw_sc_ceq *ceq)
+{
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = ceq->dev->cqp;
+	cqp->process_cqp_sds = i40iw_update_sds_noccq;
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_DESTROY_CEQ, NULL);
+}
+
+/**
+ * i40iw_sc_cceq_create - create cceq
+ * @ceq: ceq sc structure
+ * @scratch: u64 saved to be used during cqp completion
+ */
+static enum i40iw_status_code i40iw_sc_cceq_create(struct i40iw_sc_ceq *ceq, u64 scratch)
+{
+	enum i40iw_status_code ret_code;
+
+	ret_code = i40iw_sc_ceq_create(ceq, scratch, true);
+	if (!ret_code)
+		ret_code = i40iw_sc_cceq_create_done(ceq);
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_ceq_destroy - destroy ceq
+ * @ceq: ceq sc structure
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_ceq_destroy(struct i40iw_sc_ceq *ceq,
+						   u64 scratch,
+						   bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+
+	cqp = ceq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, ceq->elem_cnt);
+	set_64bit_val(wqe, 48, ceq->first_pm_pbl_idx);
+	header = ceq->ceq_id |
+		 LS_64(I40IW_CQP_OP_DESTROY_CEQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(ceq->pbl_chunk_size, I40IW_CQPSQ_CEQ_LPBLSIZE) |
+		 LS_64(ceq->virtual_map, I40IW_CQPSQ_CEQ_VMAP) |
+		 LS_64(ceq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CEQ_DESTROY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_process_ceq - process ceq
+ * @dev: sc device struct
+ * @ceq: ceq sc structure
+ */
+static void *i40iw_sc_process_ceq(struct i40iw_sc_dev *dev, struct i40iw_sc_ceq *ceq)
+{
+	u64 temp;
+	u64 *ceqe;
+	struct i40iw_sc_cq *cq = NULL;
+	u8 polarity;
+
+	ceqe = (u64 *)I40IW_GET_CURRENT_CEQ_ELEMENT(ceq);
+	get_64bit_val(ceqe, 0, &temp);
+	polarity = (u8)RS_64(temp, I40IW_CEQE_VALID);
+	if (polarity != ceq->polarity)
+		return cq;
+
+	cq = (struct i40iw_sc_cq *)(unsigned long)LS_64_1(temp, 1);
+
+	I40IW_RING_MOVE_TAIL(ceq->ceq_ring);
+	if (I40IW_RING_GETCURRENT_TAIL(ceq->ceq_ring) == 0)
+		ceq->polarity ^= 1;
+
+	if (dev->is_pf)
+		i40iw_wr32(dev->hw, I40E_PFPE_CQACK, cq->cq_uk.cq_id);
+	else
+		i40iw_wr32(dev->hw, I40E_VFPE_CQACK1, cq->cq_uk.cq_id);
+
+	return cq;
+}
+
+/**
+ * i40iw_sc_aeq_init - initialize aeq
+ * @aeq: aeq structure ptr
+ * @info: aeq initialization info
+ */
+static enum i40iw_status_code i40iw_sc_aeq_init(struct i40iw_sc_aeq *aeq,
+						struct i40iw_aeq_init_info *info)
+{
+	u32 pble_obj_cnt;
+
+	if ((info->elem_cnt < I40IW_MIN_AEQ_ENTRIES) ||
+	    (info->elem_cnt > I40IW_MAX_AEQ_ENTRIES))
+		return I40IW_ERR_INVALID_SIZE;
+	pble_obj_cnt = info->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->virtual_map && (info->first_pm_pbl_idx >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	aeq->size = sizeof(*aeq);
+	aeq->polarity = 1;
+	aeq->aeqe_base = (struct i40iw_sc_aeqe *)info->aeqe_base;
+	aeq->dev = info->dev;
+	aeq->elem_cnt = info->elem_cnt;
+
+	aeq->aeq_elem_pa = info->aeq_elem_pa;
+	I40IW_RING_INIT(aeq->aeq_ring, aeq->elem_cnt);
+	info->dev->aeq = aeq;
+
+	aeq->virtual_map = info->virtual_map;
+	aeq->pbl_list = (aeq->virtual_map ? info->pbl_list : NULL);
+	aeq->pbl_chunk_size = (aeq->virtual_map ? info->pbl_chunk_size : 0);
+	aeq->first_pm_pbl_idx = (aeq->virtual_map ? info->first_pm_pbl_idx : 0);
+	info->dev->aeq = aeq;
+	return 0;
+}
+
+/**
+ * i40iw_sc_aeq_create - create aeq
+ * @aeq: aeq structure ptr
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_aeq_create(struct i40iw_sc_aeq *aeq,
+						  u64 scratch,
+						  bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = aeq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, aeq->elem_cnt);
+	set_64bit_val(wqe, 32,
+		      (aeq->virtual_map ? 0 : aeq->aeq_elem_pa));
+	set_64bit_val(wqe, 48,
+		      (aeq->virtual_map ? aeq->first_pm_pbl_idx : 0));
+
+	header = LS_64(I40IW_CQP_OP_CREATE_AEQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(aeq->pbl_chunk_size, I40IW_CQPSQ_AEQ_LPBLSIZE) |
+		 LS_64(aeq->virtual_map, I40IW_CQPSQ_AEQ_VMAP) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "AEQ_CREATE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_aeq_destroy - destroy aeq during close
+ * @aeq: aeq structure ptr
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_aeq_destroy(struct i40iw_sc_aeq *aeq,
+						   u64 scratch,
+						   bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = aeq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, aeq->elem_cnt);
+	set_64bit_val(wqe, 48, aeq->first_pm_pbl_idx);
+	header = LS_64(I40IW_CQP_OP_DESTROY_AEQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(aeq->pbl_chunk_size, I40IW_CQPSQ_AEQ_LPBLSIZE) |
+		 LS_64(aeq->virtual_map, I40IW_CQPSQ_AEQ_VMAP) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "AEQ_DESTROY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_get_next_aeqe - get next aeq entry
+ * @aeq: aeq structure ptr
+ * @info: aeqe info to be returned
+ */
+static enum i40iw_status_code i40iw_sc_get_next_aeqe(struct i40iw_sc_aeq *aeq,
+						     struct i40iw_aeqe_info *info)
+{
+	u64 temp, compl_ctx;
+	u64 *aeqe;
+	u16 wqe_idx;
+	u8 ae_src;
+	u8 polarity;
+
+	aeqe = (u64 *)I40IW_GET_CURRENT_AEQ_ELEMENT(aeq);
+	get_64bit_val(aeqe, 0, &compl_ctx);
+	get_64bit_val(aeqe, 8, &temp);
+	polarity = (u8)RS_64(temp, I40IW_AEQE_VALID);
+
+	if (aeq->polarity != polarity)
+		return I40IW_ERR_QUEUE_EMPTY;
+
+	i40iw_debug_buf(aeq->dev, I40IW_DEBUG_WQE, "AEQ_ENTRY", aeqe, 16);
+
+	ae_src = (u8)RS_64(temp, I40IW_AEQE_AESRC);
+	wqe_idx = (u16)RS_64(temp, I40IW_AEQE_WQDESCIDX);
+	info->qp_cq_id = (u32)RS_64(temp, I40IW_AEQE_QPCQID);
+	info->ae_id = (u16)RS_64(temp, I40IW_AEQE_AECODE);
+	info->tcp_state = (u8)RS_64(temp, I40IW_AEQE_TCPSTATE);
+	info->iwarp_state = (u8)RS_64(temp, I40IW_AEQE_IWSTATE);
+	info->q2_data_written = (u8)RS_64(temp, I40IW_AEQE_Q2DATA);
+	info->aeqe_overflow = (bool)RS_64(temp, I40IW_AEQE_OVERFLOW);
+	switch (ae_src) {
+	case I40IW_AE_SOURCE_RQ:
+	case I40IW_AE_SOURCE_RQ_0011:
+		info->qp = true;
+		info->wqe_idx = wqe_idx;
+		info->compl_ctx = compl_ctx;
+		break;
+	case I40IW_AE_SOURCE_CQ:
+	case I40IW_AE_SOURCE_CQ_0110:
+	case I40IW_AE_SOURCE_CQ_1010:
+	case I40IW_AE_SOURCE_CQ_1110:
+		info->cq = true;
+		info->compl_ctx = LS_64_1(compl_ctx, 1);
+		break;
+	case I40IW_AE_SOURCE_SQ:
+	case I40IW_AE_SOURCE_SQ_0111:
+		info->qp = true;
+		info->sq = true;
+		info->wqe_idx = wqe_idx;
+		info->compl_ctx = compl_ctx;
+		break;
+	case I40IW_AE_SOURCE_IN_RR_WR:
+	case I40IW_AE_SOURCE_IN_RR_WR_1011:
+		info->qp = true;
+		info->compl_ctx = compl_ctx;
+		info->in_rdrsp_wr = true;
+		break;
+	case I40IW_AE_SOURCE_OUT_RR:
+	case I40IW_AE_SOURCE_OUT_RR_1111:
+		info->qp = true;
+		info->compl_ctx = compl_ctx;
+		info->out_rdrsp = true;
+		break;
+	default:
+		break;
+	}
+	I40IW_RING_MOVE_TAIL(aeq->aeq_ring);
+	if (I40IW_RING_GETCURRENT_TAIL(aeq->aeq_ring) == 0)
+		aeq->polarity ^= 1;
+	return 0;
+}
+
+/**
+ * i40iw_sc_repost_aeq_entries - repost completed aeq entries
+ * @dev: sc device struct
+ * @count: allocate count
+ */
+static enum i40iw_status_code i40iw_sc_repost_aeq_entries(struct i40iw_sc_dev *dev,
+							  u32 count)
+{
+	if (count > I40IW_MAX_AEQ_ALLOCATE_COUNT)
+		return I40IW_ERR_INVALID_SIZE;
+
+	if (dev->is_pf)
+		i40iw_wr32(dev->hw, I40E_PFPE_AEQALLOC, count);
+	else
+		i40iw_wr32(dev->hw, I40E_VFPE_AEQALLOC1, count);
+
+	return 0;
+}
+
+/**
+ * i40iw_sc_aeq_create_done - create aeq
+ * @aeq: aeq structure ptr
+ */
+static enum i40iw_status_code i40iw_sc_aeq_create_done(struct i40iw_sc_aeq *aeq)
+{
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = aeq->dev->cqp;
+	return i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_CREATE_AEQ, NULL);
+}
+
+/**
+ * i40iw_sc_aeq_destroy_done - destroy of aeq during close
+ * @aeq: aeq structure ptr
+ */
+static enum i40iw_status_code i40iw_sc_aeq_destroy_done(struct i40iw_sc_aeq *aeq)
+{
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = aeq->dev->cqp;
+	return  i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_DESTROY_AEQ, NULL);
+}
+
+/**
+ * i40iw_sc_ccq_init - initialize control cq
+ * @cq: sc's cq ctruct
+ * @info: info for control cq initialization
+ */
+static enum i40iw_status_code i40iw_sc_ccq_init(struct i40iw_sc_cq *cq,
+						struct i40iw_ccq_init_info *info)
+{
+	u32 pble_obj_cnt;
+
+	if (info->num_elem < I40IW_MIN_CQ_SIZE || info->num_elem > I40IW_MAX_CQ_SIZE)
+		return I40IW_ERR_INVALID_SIZE;
+
+	if (info->ceq_id > I40IW_MAX_CEQID)
+		return I40IW_ERR_INVALID_CEQ_ID;
+
+	pble_obj_cnt = info->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->virtual_map && (info->first_pm_pbl_idx >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	cq->cq_pa = info->cq_pa;
+	cq->cq_uk.cq_base = info->cq_base;
+	cq->shadow_area_pa = info->shadow_area_pa;
+	cq->cq_uk.shadow_area = info->shadow_area;
+	cq->shadow_read_threshold = info->shadow_read_threshold;
+	cq->dev = info->dev;
+	cq->ceq_id = info->ceq_id;
+	cq->cq_uk.cq_size = info->num_elem;
+	cq->cq_type = I40IW_CQ_TYPE_CQP;
+	cq->ceqe_mask = info->ceqe_mask;
+	I40IW_RING_INIT(cq->cq_uk.cq_ring, info->num_elem);
+
+	cq->cq_uk.cq_id = 0;    /* control cq is id 0 always */
+	cq->ceq_id_valid = info->ceq_id_valid;
+	cq->tph_en = info->tph_en;
+	cq->tph_val = info->tph_val;
+	cq->cq_uk.avoid_mem_cflct = info->avoid_mem_cflct;
+
+	cq->pbl_list = info->pbl_list;
+	cq->virtual_map = info->virtual_map;
+	cq->pbl_chunk_size = info->pbl_chunk_size;
+	cq->first_pm_pbl_idx = info->first_pm_pbl_idx;
+	cq->cq_uk.polarity = true;
+
+	/* following are only for iw cqs so initialize them to zero */
+	cq->cq_uk.cqe_alloc_reg = NULL;
+	info->dev->ccq = cq;
+	return 0;
+}
+
+/**
+ * i40iw_sc_ccq_create_done - poll cqp for ccq create
+ * @ccq: ccq sc struct
+ */
+static enum i40iw_status_code i40iw_sc_ccq_create_done(struct i40iw_sc_cq *ccq)
+{
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = ccq->dev->cqp;
+	return	i40iw_sc_poll_for_cqp_op_done(cqp, I40IW_CQP_OP_CREATE_CQ, NULL);
+}
+
+/**
+ * i40iw_sc_ccq_create - create control cq
+ * @ccq: ccq sc struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @check_overflow: overlow flag for ccq
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_ccq_create(struct i40iw_sc_cq *ccq,
+						  u64 scratch,
+						  bool check_overflow,
+						  bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+	enum i40iw_status_code ret_code;
+
+	cqp = ccq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 0, ccq->cq_uk.cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(ccq, 1));
+	set_64bit_val(wqe, 16,
+		      LS_64(ccq->shadow_read_threshold, I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD));
+	set_64bit_val(wqe, 32, (ccq->virtual_map ? 0 : ccq->cq_pa));
+	set_64bit_val(wqe, 40, ccq->shadow_area_pa);
+	set_64bit_val(wqe, 48,
+		      (ccq->virtual_map ? ccq->first_pm_pbl_idx : 0));
+	set_64bit_val(wqe, 56,
+		      LS_64(ccq->tph_val, I40IW_CQPSQ_TPHVAL));
+
+	header = ccq->cq_uk.cq_id |
+		 LS_64((ccq->ceq_id_valid ? ccq->ceq_id : 0), I40IW_CQPSQ_CQ_CEQID) |
+		 LS_64(I40IW_CQP_OP_CREATE_CQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(ccq->pbl_chunk_size, I40IW_CQPSQ_CQ_LPBLSIZE) |
+		 LS_64(check_overflow, I40IW_CQPSQ_CQ_CHKOVERFLOW) |
+		 LS_64(ccq->virtual_map, I40IW_CQPSQ_CQ_VIRTMAP) |
+		 LS_64(ccq->ceqe_mask, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+		 LS_64(ccq->ceq_id_valid, I40IW_CQPSQ_CQ_CEQIDVALID) |
+		 LS_64(ccq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(ccq->cq_uk.avoid_mem_cflct, I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CCQ_CREATE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+		ret_code = i40iw_sc_ccq_create_done(ccq);
+		if (ret_code)
+			return ret_code;
+	}
+	cqp->process_cqp_sds = i40iw_cqp_sds_cmd;
+
+	return 0;
+}
+
+/**
+ * i40iw_sc_ccq_destroy - destroy ccq during close
+ * @ccq: ccq sc struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_ccq_destroy(struct i40iw_sc_cq *ccq,
+						   u64 scratch,
+						   bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+	enum i40iw_status_code ret_code = 0;
+	u32 tail, val, error;
+
+	cqp = ccq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 0, ccq->cq_uk.cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(ccq, 1));
+	set_64bit_val(wqe, 40, ccq->shadow_area_pa);
+
+	header = ccq->cq_uk.cq_id |
+		 LS_64((ccq->ceq_id_valid ? ccq->ceq_id : 0), I40IW_CQPSQ_CQ_CEQID) |
+		 LS_64(I40IW_CQP_OP_DESTROY_CQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(ccq->ceqe_mask, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+		 LS_64(ccq->ceq_id_valid, I40IW_CQPSQ_CQ_CEQIDVALID) |
+		 LS_64(ccq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(ccq->cq_uk.avoid_mem_cflct, I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CCQ_DESTROY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+	if (error)
+		return I40IW_ERR_CQP_COMPL_ERROR;
+
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+		ret_code = i40iw_cqp_poll_registers(cqp, tail, 1000);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_cq_init - initialize completion q
+ * @cq: cq struct
+ * @info: cq initialization info
+ */
+static enum i40iw_status_code i40iw_sc_cq_init(struct i40iw_sc_cq *cq,
+					       struct i40iw_cq_init_info *info)
+{
+	u32 __iomem *cqe_alloc_reg = NULL;
+	enum i40iw_status_code ret_code;
+	u32 pble_obj_cnt;
+	u32 arm_offset;
+
+	pble_obj_cnt = info->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->virtual_map && (info->first_pm_pbl_idx >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	cq->cq_pa = info->cq_base_pa;
+	cq->dev = info->dev;
+	cq->ceq_id = info->ceq_id;
+	arm_offset = (info->dev->is_pf) ? I40E_PFPE_CQARM : I40E_VFPE_CQARM1;
+	if (i40iw_get_hw_addr(cq->dev))
+		cqe_alloc_reg = (u32 __iomem *)(i40iw_get_hw_addr(cq->dev) +
+					      arm_offset);
+	info->cq_uk_init_info.cqe_alloc_reg = cqe_alloc_reg;
+	ret_code = i40iw_cq_uk_init(&cq->cq_uk, &info->cq_uk_init_info);
+	if (ret_code)
+		return ret_code;
+	cq->virtual_map = info->virtual_map;
+	cq->pbl_chunk_size = info->pbl_chunk_size;
+	cq->ceqe_mask = info->ceqe_mask;
+	cq->cq_type = (info->type) ? info->type : I40IW_CQ_TYPE_IWARP;
+
+	cq->shadow_area_pa = info->shadow_area_pa;
+	cq->shadow_read_threshold = info->shadow_read_threshold;
+
+	cq->ceq_id_valid = info->ceq_id_valid;
+	cq->tph_en = info->tph_en;
+	cq->tph_val = info->tph_val;
+
+	cq->first_pm_pbl_idx = info->first_pm_pbl_idx;
+
+	return 0;
+}
+
+/**
+ * i40iw_sc_cq_create - create completion q
+ * @cq: cq struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @check_overflow: flag for overflow check
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_cq_create(struct i40iw_sc_cq *cq,
+						 u64 scratch,
+						 bool check_overflow,
+						 bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	if (cq->cq_uk.cq_id > I40IW_MAX_CQID)
+		return I40IW_ERR_INVALID_CQ_ID;
+
+	if (cq->ceq_id > I40IW_MAX_CEQID)
+		return I40IW_ERR_INVALID_CEQ_ID;
+
+	cqp = cq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 0, cq->cq_uk.cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(cq, 1));
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(cq->shadow_read_threshold, I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD));
+
+	set_64bit_val(wqe, 32, (cq->virtual_map ? 0 : cq->cq_pa));
+
+	set_64bit_val(wqe, 40, cq->shadow_area_pa);
+	set_64bit_val(wqe, 48, (cq->virtual_map ? cq->first_pm_pbl_idx : 0));
+	set_64bit_val(wqe, 56, LS_64(cq->tph_val, I40IW_CQPSQ_TPHVAL));
+
+	header = cq->cq_uk.cq_id |
+		 LS_64((cq->ceq_id_valid ? cq->ceq_id : 0), I40IW_CQPSQ_CQ_CEQID) |
+		 LS_64(I40IW_CQP_OP_CREATE_CQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cq->pbl_chunk_size, I40IW_CQPSQ_CQ_LPBLSIZE) |
+		 LS_64(check_overflow, I40IW_CQPSQ_CQ_CHKOVERFLOW) |
+		 LS_64(cq->virtual_map, I40IW_CQPSQ_CQ_VIRTMAP) |
+		 LS_64(cq->ceqe_mask, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+		 LS_64(cq->ceq_id_valid, I40IW_CQPSQ_CQ_CEQIDVALID) |
+		 LS_64(cq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(cq->cq_uk.avoid_mem_cflct, I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CQ_CREATE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_cq_destroy - destroy completion q
+ * @cq: cq struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_cq_destroy(struct i40iw_sc_cq *cq,
+						  u64 scratch,
+						  bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+
+	cqp = cq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 0, cq->cq_uk.cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(cq, 1));
+	set_64bit_val(wqe, 40, cq->shadow_area_pa);
+	set_64bit_val(wqe, 48, (cq->virtual_map ? cq->first_pm_pbl_idx : 0));
+
+	header = cq->cq_uk.cq_id |
+		 LS_64((cq->ceq_id_valid ? cq->ceq_id : 0), I40IW_CQPSQ_CQ_CEQID) |
+		 LS_64(I40IW_CQP_OP_DESTROY_CQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cq->pbl_chunk_size, I40IW_CQPSQ_CQ_LPBLSIZE) |
+		 LS_64(cq->virtual_map, I40IW_CQPSQ_CQ_VIRTMAP) |
+		 LS_64(cq->ceqe_mask, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+		 LS_64(cq->ceq_id_valid, I40IW_CQPSQ_CQ_CEQIDVALID) |
+		 LS_64(cq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(cq->cq_uk.avoid_mem_cflct, I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CQ_DESTROY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_cq_modify - modify a Completion Queue
+ * @cq: cq struct
+ * @info: modification info struct
+ * @scratch:
+ * @post_sq: flag to post to sq
+ */
+static enum i40iw_status_code i40iw_sc_cq_modify(struct i40iw_sc_cq *cq,
+						 struct i40iw_modify_cq_info *info,
+						 u64 scratch,
+						 bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+	u32 cq_size, ceq_id, first_pm_pbl_idx;
+	u8 pbl_chunk_size;
+	bool virtual_map, ceq_id_valid, check_overflow;
+	u32 pble_obj_cnt;
+
+	if (info->ceq_valid && (info->ceq_id > I40IW_MAX_CEQID))
+		return I40IW_ERR_INVALID_CEQ_ID;
+
+	pble_obj_cnt = cq->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->cq_resize && info->virtual_map &&
+	    (info->first_pm_pbl_idx >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	cqp = cq->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	cq->pbl_list = info->pbl_list;
+	cq->cq_pa = info->cq_pa;
+	cq->first_pm_pbl_idx = info->first_pm_pbl_idx;
+
+	cq_size = info->cq_resize ? info->cq_size : cq->cq_uk.cq_size;
+	if (info->ceq_change) {
+		ceq_id_valid = true;
+		ceq_id = info->ceq_id;
+	} else {
+		ceq_id_valid = cq->ceq_id_valid;
+		ceq_id = ceq_id_valid ? cq->ceq_id : 0;
+	}
+	virtual_map = info->cq_resize ? info->virtual_map : cq->virtual_map;
+	first_pm_pbl_idx = (info->cq_resize ?
+			    (info->virtual_map ? info->first_pm_pbl_idx : 0) :
+			    (cq->virtual_map ? cq->first_pm_pbl_idx : 0));
+	pbl_chunk_size = (info->cq_resize ?
+			  (info->virtual_map ? info->pbl_chunk_size : 0) :
+			  (cq->virtual_map ? cq->pbl_chunk_size : 0));
+	check_overflow = info->check_overflow_change ? info->check_overflow :
+			 cq->check_overflow;
+	cq->cq_uk.cq_size = cq_size;
+	cq->ceq_id_valid = ceq_id_valid;
+	cq->ceq_id = ceq_id;
+	cq->virtual_map = virtual_map;
+	cq->first_pm_pbl_idx = first_pm_pbl_idx;
+	cq->pbl_chunk_size = pbl_chunk_size;
+	cq->check_overflow = check_overflow;
+
+	set_64bit_val(wqe, 0, cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(cq, 1));
+	set_64bit_val(wqe, 16,
+		      LS_64(info->shadow_read_threshold, I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD));
+	set_64bit_val(wqe, 32, (cq->virtual_map ? 0 : cq->cq_pa));
+	set_64bit_val(wqe, 40, cq->shadow_area_pa);
+	set_64bit_val(wqe, 48, (cq->virtual_map ? first_pm_pbl_idx : 0));
+	set_64bit_val(wqe, 56, LS_64(cq->tph_val, I40IW_CQPSQ_TPHVAL));
+
+	header = cq->cq_uk.cq_id |
+		 LS_64(ceq_id, I40IW_CQPSQ_CQ_CEQID) |
+		 LS_64(I40IW_CQP_OP_MODIFY_CQ, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->cq_resize, I40IW_CQPSQ_CQ_CQRESIZE) |
+		 LS_64(pbl_chunk_size, I40IW_CQPSQ_CQ_LPBLSIZE) |
+		 LS_64(check_overflow, I40IW_CQPSQ_CQ_CHKOVERFLOW) |
+		 LS_64(virtual_map, I40IW_CQPSQ_CQ_VIRTMAP) |
+		 LS_64(cq->ceqe_mask, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+		 LS_64(ceq_id_valid, I40IW_CQPSQ_CQ_CEQIDVALID) |
+		 LS_64(cq->tph_en, I40IW_CQPSQ_TPHEN) |
+		 LS_64(cq->cq_uk.avoid_mem_cflct, I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "CQ_MODIFY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_init - initialize qp
+ * @qp: sc qp
+ * @info: initialization qp info
+ */
+static enum i40iw_status_code i40iw_sc_qp_init(struct i40iw_sc_qp *qp,
+					       struct i40iw_qp_init_info *info)
+{
+	u32 __iomem *wqe_alloc_reg = NULL;
+	enum i40iw_status_code ret_code;
+	u32 pble_obj_cnt;
+	u8 wqe_size;
+	u32 offset;
+
+	qp->dev = info->pd->dev;
+	qp->sq_pa = info->sq_pa;
+	qp->rq_pa = info->rq_pa;
+	qp->hw_host_ctx_pa = info->host_ctx_pa;
+	qp->q2_pa = info->q2_pa;
+	qp->shadow_area_pa = info->shadow_area_pa;
+
+	qp->q2_buf = info->q2;
+	qp->pd = info->pd;
+	qp->hw_host_ctx = info->host_ctx;
+	offset = (qp->pd->dev->is_pf) ? I40E_PFPE_WQEALLOC : I40E_VFPE_WQEALLOC1;
+	if (i40iw_get_hw_addr(qp->pd->dev))
+		wqe_alloc_reg = (u32 __iomem *)(i40iw_get_hw_addr(qp->pd->dev) +
+					      offset);
+
+	info->qp_uk_init_info.wqe_alloc_reg = wqe_alloc_reg;
+	ret_code = i40iw_qp_uk_init(&qp->qp_uk, &info->qp_uk_init_info);
+	if (ret_code)
+		return ret_code;
+	qp->virtual_map = info->virtual_map;
+
+	pble_obj_cnt = info->pd->dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if ((info->virtual_map && (info->sq_pa >= pble_obj_cnt)) ||
+	    (info->virtual_map && (info->rq_pa >= pble_obj_cnt)))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	qp->llp_stream_handle = (void *)(-1);
+	qp->qp_type = (info->type) ? info->type : I40IW_QP_TYPE_IWARP;
+
+	qp->hw_sq_size = i40iw_get_encoded_wqe_size(qp->qp_uk.sq_ring.size,
+						    false);
+	i40iw_debug(qp->dev, I40IW_DEBUG_WQE, "%s: hw_sq_size[%04d] sq_ring.size[%04d]\n",
+		    __func__, qp->hw_sq_size, qp->qp_uk.sq_ring.size);
+	ret_code = i40iw_fragcnt_to_wqesize_rq(qp->qp_uk.max_rq_frag_cnt,
+					       &wqe_size);
+	if (ret_code)
+		return ret_code;
+	qp->hw_rq_size = i40iw_get_encoded_wqe_size(qp->qp_uk.rq_size *
+				(wqe_size / I40IW_QP_WQE_MIN_SIZE), false);
+	i40iw_debug(qp->dev, I40IW_DEBUG_WQE,
+		    "%s: hw_rq_size[%04d] qp_uk.rq_size[%04d] wqe_size[%04d]\n",
+		    __func__, qp->hw_rq_size, qp->qp_uk.rq_size, wqe_size);
+	qp->sq_tph_val = info->sq_tph_val;
+	qp->rq_tph_val = info->rq_tph_val;
+	qp->sq_tph_en = info->sq_tph_en;
+	qp->rq_tph_en = info->rq_tph_en;
+	qp->rcv_tph_en = info->rcv_tph_en;
+	qp->xmit_tph_en = info->xmit_tph_en;
+	qp->qs_handle = qp->pd->dev->qs_handle;
+	qp->exception_lan_queue = qp->pd->dev->exception_lan_queue;
+
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_create - create qp
+ * @qp: sc qp
+ * @info: qp create info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_qp_create(
+				struct i40iw_sc_qp *qp,
+				struct i40iw_create_qp_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+
+	if ((qp->qp_uk.qp_id < I40IW_MIN_IW_QP_ID) ||
+	    (qp->qp_uk.qp_id > I40IW_MAX_IW_QP_ID))
+		return I40IW_ERR_INVALID_QP_ID;
+
+	cqp = qp->pd->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, qp->hw_host_ctx_pa);
+
+	set_64bit_val(wqe, 40, qp->shadow_area_pa);
+
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_CREATE_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64((info->ord_valid ? 1 : 0), I40IW_CQPSQ_QP_ORDVALID) |
+		 LS_64(info->tcp_ctx_valid, I40IW_CQPSQ_QP_TOECTXVALID) |
+		 LS_64(qp->qp_type, I40IW_CQPSQ_QP_QPTYPE) |
+		 LS_64(qp->virtual_map, I40IW_CQPSQ_QP_VQ) |
+		 LS_64(info->cq_num_valid, I40IW_CQPSQ_QP_CQNUMVALID) |
+		 LS_64(info->static_rsrc, I40IW_CQPSQ_QP_STATRSRC) |
+		 LS_64(info->arp_cache_idx_valid, I40IW_CQPSQ_QP_ARPTABIDXVALID) |
+		 LS_64(info->next_iwarp_state, I40IW_CQPSQ_QP_NEXTIWSTATE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QP_CREATE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_modify - modify qp cqp wqe
+ * @qp: sc qp
+ * @info: modify qp info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_qp_modify(
+				struct i40iw_sc_qp *qp,
+				struct i40iw_modify_qp_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+	u8 term_actions = 0;
+	u8 term_len = 0;
+
+	cqp = qp->pd->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	if (info->next_iwarp_state == I40IW_QP_STATE_TERMINATE) {
+		if (info->dont_send_fin)
+			term_actions += I40IWQP_TERM_SEND_TERM_ONLY;
+		if (info->dont_send_term)
+			term_actions += I40IWQP_TERM_SEND_FIN_ONLY;
+		if ((term_actions == I40IWQP_TERM_SEND_TERM_AND_FIN) ||
+		    (term_actions == I40IWQP_TERM_SEND_TERM_ONLY))
+			term_len = info->termlen;
+	}
+
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(info->new_mss, I40IW_CQPSQ_QP_NEWMSS) |
+		      LS_64(term_len, I40IW_CQPSQ_QP_TERMLEN));
+
+	set_64bit_val(wqe, 16, qp->hw_host_ctx_pa);
+	set_64bit_val(wqe, 40, qp->shadow_area_pa);
+
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_MODIFY_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->ord_valid, I40IW_CQPSQ_QP_ORDVALID) |
+		 LS_64(info->tcp_ctx_valid, I40IW_CQPSQ_QP_TOECTXVALID) |
+		 LS_64(info->cached_var_valid, I40IW_CQPSQ_QP_CACHEDVARVALID) |
+		 LS_64(qp->virtual_map, I40IW_CQPSQ_QP_VQ) |
+		 LS_64(info->cq_num_valid, I40IW_CQPSQ_QP_CQNUMVALID) |
+		 LS_64(info->force_loopback, I40IW_CQPSQ_QP_FORCELOOPBACK) |
+		 LS_64(qp->qp_type, I40IW_CQPSQ_QP_QPTYPE) |
+		 LS_64(info->mss_change, I40IW_CQPSQ_QP_MSSCHANGE) |
+		 LS_64(info->static_rsrc, I40IW_CQPSQ_QP_STATRSRC) |
+		 LS_64(info->remove_hash_idx, I40IW_CQPSQ_QP_REMOVEHASHENTRY) |
+		 LS_64(term_actions, I40IW_CQPSQ_QP_TERMACT) |
+		 LS_64(info->reset_tcp_conn, I40IW_CQPSQ_QP_RESETCON) |
+		 LS_64(info->arp_cache_idx_valid, I40IW_CQPSQ_QP_ARPTABIDXVALID) |
+		 LS_64(info->next_iwarp_state, I40IW_CQPSQ_QP_NEXTIWSTATE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QP_MODIFY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_destroy - cqp destroy qp
+ * @qp: sc qp
+ * @scratch: u64 saved to be used during cqp completion
+ * @remove_hash_idx: flag if to remove hash idx
+ * @ignore_mw_bnd: memory window bind flag
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_qp_destroy(
+					struct i40iw_sc_qp *qp,
+					u64 scratch,
+					bool remove_hash_idx,
+					bool ignore_mw_bnd,
+					bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = qp->pd->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, qp->hw_host_ctx_pa);
+	set_64bit_val(wqe, 40, qp->shadow_area_pa);
+
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_DESTROY_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(qp->qp_type, I40IW_CQPSQ_QP_QPTYPE) |
+		 LS_64(ignore_mw_bnd, I40IW_CQPSQ_QP_IGNOREMWBOUND) |
+		 LS_64(remove_hash_idx, I40IW_CQPSQ_QP_REMOVEHASHENTRY) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QP_DESTROY WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_flush_wqes - flush qp's wqe
+ * @qp: sc qp
+ * @info: dlush information
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_qp_flush_wqes(
+				struct i40iw_sc_qp *qp,
+				struct i40iw_qp_flush_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 temp = 0;
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+	bool flush_sq = false, flush_rq = false;
+
+	if (info->rq && !qp->flush_rq)
+		flush_rq = true;
+
+	if (info->sq && !qp->flush_sq)
+		flush_sq = true;
+
+	qp->flush_sq |= flush_sq;
+	qp->flush_rq |= flush_rq;
+	if (!flush_sq && !flush_rq) {
+		if (info->ae_code != I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR)
+			return 0;
+	}
+
+	cqp = qp->pd->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	if (info->userflushcode) {
+		if (flush_rq) {
+			temp |= LS_64(info->rq_minor_code, I40IW_CQPSQ_FWQE_RQMNERR) |
+				LS_64(info->rq_major_code, I40IW_CQPSQ_FWQE_RQMJERR);
+		}
+		if (flush_sq) {
+			temp |= LS_64(info->sq_minor_code, I40IW_CQPSQ_FWQE_SQMNERR) |
+				LS_64(info->sq_major_code, I40IW_CQPSQ_FWQE_SQMJERR);
+		}
+	}
+	set_64bit_val(wqe, 16, temp);
+
+	temp = (info->generate_ae) ?
+		info->ae_code | LS_64(info->ae_source, I40IW_CQPSQ_FWQE_AESOURCE) : 0;
+
+	set_64bit_val(wqe, 8, temp);
+
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_FLUSH_WQES, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->generate_ae, I40IW_CQPSQ_FWQE_GENERATE_AE) |
+		 LS_64(info->userflushcode, I40IW_CQPSQ_FWQE_USERFLCODE) |
+		 LS_64(flush_sq, I40IW_CQPSQ_FWQE_FLUSHSQ) |
+		 LS_64(flush_rq, I40IW_CQPSQ_FWQE_FLUSHRQ) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "QP_FLUSH WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_upload_context - upload qp's context
+ * @dev: sc device struct
+ * @info: upload context info ptr for return
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_qp_upload_context(
+					struct i40iw_sc_dev *dev,
+					struct i40iw_upload_context_info *info,
+					u64 scratch,
+					bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 16, info->buf_pa);
+
+	header = LS_64(info->qp_id, I40IW_CQPSQ_UCTX_QPID) |
+		 LS_64(I40IW_CQP_OP_UPLOAD_CONTEXT, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->qp_type, I40IW_CQPSQ_UCTX_QPTYPE) |
+		 LS_64(info->raw_format, I40IW_CQPSQ_UCTX_RAWFORMAT) |
+		 LS_64(info->freeze_qp, I40IW_CQPSQ_UCTX_FREEZEQP) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "QP_UPLOAD_CTX WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_qp_setctx - set qp's context
+ * @qp: sc qp
+ * @qp_ctx: context ptr
+ * @info: ctx info
+ */
+static enum i40iw_status_code i40iw_sc_qp_setctx(
+				struct i40iw_sc_qp *qp,
+				u64 *qp_ctx,
+				struct i40iw_qp_host_ctx_info *info)
+{
+	struct i40iwarp_offload_info *iw;
+	struct i40iw_tcp_offload_info *tcp;
+	u64 qw0, qw3, qw7 = 0;
+
+	iw = info->iwarp_info;
+	tcp = info->tcp_info;
+	qw0 = LS_64(qp->qp_uk.rq_wqe_size, I40IWQPC_RQWQESIZE) |
+	      LS_64(info->err_rq_idx_valid, I40IWQPC_ERR_RQ_IDX_VALID) |
+	      LS_64(qp->rcv_tph_en, I40IWQPC_RCVTPHEN) |
+	      LS_64(qp->xmit_tph_en, I40IWQPC_XMITTPHEN) |
+	      LS_64(qp->rq_tph_en, I40IWQPC_RQTPHEN) |
+	      LS_64(qp->sq_tph_en, I40IWQPC_SQTPHEN) |
+	      LS_64(info->push_idx, I40IWQPC_PPIDX) |
+	      LS_64(info->push_mode_en, I40IWQPC_PMENA);
+
+	set_64bit_val(qp_ctx, 8, qp->sq_pa);
+	set_64bit_val(qp_ctx, 16, qp->rq_pa);
+
+	qw3 = LS_64(qp->src_mac_addr_idx, I40IWQPC_SRCMACADDRIDX) |
+	      LS_64(qp->hw_rq_size, I40IWQPC_RQSIZE) |
+	      LS_64(qp->hw_sq_size, I40IWQPC_SQSIZE);
+
+	set_64bit_val(qp_ctx,
+		      128,
+		      LS_64(info->err_rq_idx, I40IWQPC_ERR_RQ_IDX));
+
+	set_64bit_val(qp_ctx,
+		      136,
+		      LS_64(info->send_cq_num, I40IWQPC_TXCQNUM) |
+		      LS_64(info->rcv_cq_num, I40IWQPC_RXCQNUM));
+
+	set_64bit_val(qp_ctx,
+		      168,
+		      LS_64(info->qp_compl_ctx, I40IWQPC_QPCOMPCTX));
+	set_64bit_val(qp_ctx,
+		      176,
+		      LS_64(qp->sq_tph_val, I40IWQPC_SQTPHVAL) |
+		      LS_64(qp->rq_tph_val, I40IWQPC_RQTPHVAL) |
+		      LS_64(qp->qs_handle, I40IWQPC_QSHANDLE) |
+		      LS_64(qp->exception_lan_queue, I40IWQPC_EXCEPTION_LAN_QUEUE));
+
+	if (info->iwarp_info_valid) {
+		qw0 |= LS_64(iw->ddp_ver, I40IWQPC_DDP_VER) |
+		       LS_64(iw->rdmap_ver, I40IWQPC_RDMAP_VER);
+
+		qw7 |= LS_64(iw->pd_id, I40IWQPC_PDIDX);
+		set_64bit_val(qp_ctx, 144, qp->q2_pa);
+		set_64bit_val(qp_ctx,
+			      152,
+			      LS_64(iw->last_byte_sent, I40IWQPC_LASTBYTESENT));
+
+		/*
+		* Hard-code IRD_SIZE to hw-limit, 128, in qpctx, i.e matching an
+		*advertisable IRD of 64
+		*/
+		iw->ird_size = I40IW_QPCTX_ENCD_MAXIRD;
+		set_64bit_val(qp_ctx,
+			      160,
+			      LS_64(iw->ord_size, I40IWQPC_ORDSIZE) |
+			      LS_64(iw->ird_size, I40IWQPC_IRDSIZE) |
+			      LS_64(iw->wr_rdresp_en, I40IWQPC_WRRDRSPOK) |
+			      LS_64(iw->rd_enable, I40IWQPC_RDOK) |
+			      LS_64(iw->snd_mark_en, I40IWQPC_SNDMARKERS) |
+			      LS_64(iw->bind_en, I40IWQPC_BINDEN) |
+			      LS_64(iw->fast_reg_en, I40IWQPC_FASTREGEN) |
+			      LS_64(iw->priv_mode_en, I40IWQPC_PRIVEN) |
+			      LS_64(1, I40IWQPC_IWARPMODE) |
+			      LS_64(iw->rcv_mark_en, I40IWQPC_RCVMARKERS) |
+			      LS_64(iw->align_hdrs, I40IWQPC_ALIGNHDRS) |
+			      LS_64(iw->rcv_no_mpa_crc, I40IWQPC_RCVNOMPACRC) |
+			      LS_64(iw->rcv_mark_offset, I40IWQPC_RCVMARKOFFSET) |
+			      LS_64(iw->snd_mark_offset, I40IWQPC_SNDMARKOFFSET));
+	}
+	if (info->tcp_info_valid) {
+		qw0 |= LS_64(tcp->ipv4, I40IWQPC_IPV4) |
+		       LS_64(tcp->no_nagle, I40IWQPC_NONAGLE) |
+		       LS_64(tcp->insert_vlan_tag, I40IWQPC_INSERTVLANTAG) |
+		       LS_64(tcp->time_stamp, I40IWQPC_TIMESTAMP) |
+		       LS_64(tcp->cwnd_inc_limit, I40IWQPC_LIMIT) |
+		       LS_64(tcp->drop_ooo_seg, I40IWQPC_DROPOOOSEG) |
+		       LS_64(tcp->dup_ack_thresh, I40IWQPC_DUPACK_THRESH);
+
+		qw3 |= LS_64(tcp->ttl, I40IWQPC_TTL) |
+		       LS_64(tcp->src_mac_addr_idx, I40IWQPC_SRCMACADDRIDX) |
+		       LS_64(tcp->avoid_stretch_ack, I40IWQPC_AVOIDSTRETCHACK) |
+		       LS_64(tcp->tos, I40IWQPC_TOS) |
+		       LS_64(tcp->src_port, I40IWQPC_SRCPORTNUM) |
+		       LS_64(tcp->dst_port, I40IWQPC_DESTPORTNUM);
+
+		qp->src_mac_addr_idx = tcp->src_mac_addr_idx;
+		set_64bit_val(qp_ctx,
+			      32,
+			      LS_64(tcp->dest_ip_addr2, I40IWQPC_DESTIPADDR2) |
+			      LS_64(tcp->dest_ip_addr3, I40IWQPC_DESTIPADDR3));
+
+		set_64bit_val(qp_ctx,
+			      40,
+			      LS_64(tcp->dest_ip_addr0, I40IWQPC_DESTIPADDR0) |
+			      LS_64(tcp->dest_ip_addr1, I40IWQPC_DESTIPADDR1));
+
+		set_64bit_val(qp_ctx,
+			      48,
+			      LS_64(tcp->snd_mss, I40IWQPC_SNDMSS) |
+				LS_64(tcp->vlan_tag, I40IWQPC_VLANTAG) |
+				LS_64(tcp->arp_idx, I40IWQPC_ARPIDX));
+
+		qw7 |= LS_64(tcp->flow_label, I40IWQPC_FLOWLABEL) |
+		       LS_64(tcp->wscale, I40IWQPC_WSCALE) |
+		       LS_64(tcp->ignore_tcp_opt, I40IWQPC_IGNORE_TCP_OPT) |
+		       LS_64(tcp->ignore_tcp_uns_opt, I40IWQPC_IGNORE_TCP_UNS_OPT) |
+		       LS_64(tcp->tcp_state, I40IWQPC_TCPSTATE) |
+		       LS_64(tcp->rcv_wscale, I40IWQPC_RCVSCALE) |
+		       LS_64(tcp->snd_wscale, I40IWQPC_SNDSCALE);
+
+		set_64bit_val(qp_ctx,
+			      72,
+			      LS_64(tcp->time_stamp_recent, I40IWQPC_TIMESTAMP_RECENT) |
+			      LS_64(tcp->time_stamp_age, I40IWQPC_TIMESTAMP_AGE));
+		set_64bit_val(qp_ctx,
+			      80,
+			      LS_64(tcp->snd_nxt, I40IWQPC_SNDNXT) |
+			      LS_64(tcp->snd_wnd, I40IWQPC_SNDWND));
+
+		set_64bit_val(qp_ctx,
+			      88,
+			      LS_64(tcp->rcv_nxt, I40IWQPC_RCVNXT) |
+			      LS_64(tcp->rcv_wnd, I40IWQPC_RCVWND));
+		set_64bit_val(qp_ctx,
+			      96,
+			      LS_64(tcp->snd_max, I40IWQPC_SNDMAX) |
+			      LS_64(tcp->snd_una, I40IWQPC_SNDUNA));
+		set_64bit_val(qp_ctx,
+			      104,
+			      LS_64(tcp->srtt, I40IWQPC_SRTT) |
+			      LS_64(tcp->rtt_var, I40IWQPC_RTTVAR));
+		set_64bit_val(qp_ctx,
+			      112,
+			      LS_64(tcp->ss_thresh, I40IWQPC_SSTHRESH) |
+			      LS_64(tcp->cwnd, I40IWQPC_CWND));
+		set_64bit_val(qp_ctx,
+			      120,
+			      LS_64(tcp->snd_wl1, I40IWQPC_SNDWL1) |
+			      LS_64(tcp->snd_wl2, I40IWQPC_SNDWL2));
+		set_64bit_val(qp_ctx,
+			      128,
+			      LS_64(tcp->max_snd_window, I40IWQPC_MAXSNDWND) |
+			      LS_64(tcp->rexmit_thresh, I40IWQPC_REXMIT_THRESH));
+		set_64bit_val(qp_ctx,
+			      184,
+			      LS_64(tcp->local_ipaddr3, I40IWQPC_LOCAL_IPADDR3) |
+			      LS_64(tcp->local_ipaddr2, I40IWQPC_LOCAL_IPADDR2));
+		set_64bit_val(qp_ctx,
+			      192,
+			      LS_64(tcp->local_ipaddr1, I40IWQPC_LOCAL_IPADDR1) |
+			      LS_64(tcp->local_ipaddr0, I40IWQPC_LOCAL_IPADDR0));
+	}
+
+	set_64bit_val(qp_ctx, 0, qw0);
+	set_64bit_val(qp_ctx, 24, qw3);
+	set_64bit_val(qp_ctx, 56, qw7);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_WQE, "QP_HOST)CTX WQE",
+			qp_ctx, I40IW_QP_CTX_SIZE);
+	return 0;
+}
+
+/**
+ * i40iw_sc_alloc_stag - mr stag alloc
+ * @dev: sc device struct
+ * @info: stag info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_alloc_stag(
+				struct i40iw_sc_dev *dev,
+				struct i40iw_allocate_stag_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(info->pd_id, I40IW_CQPSQ_STAG_PDID) |
+		      LS_64(info->total_len, I40IW_CQPSQ_STAG_STAGLEN));
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(info->stag_idx, I40IW_CQPSQ_STAG_IDX));
+	set_64bit_val(wqe,
+		      40,
+		      LS_64(info->hmc_fcn_index, I40IW_CQPSQ_STAG_HMCFNIDX));
+
+	header = LS_64(I40IW_CQP_OP_ALLOC_STAG, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_STAG_MR) |
+		 LS_64(info->access_rights, I40IW_CQPSQ_STAG_ARIGHTS) |
+		 LS_64(info->chunk_size, I40IW_CQPSQ_STAG_LPBLSIZE) |
+		 LS_64(info->page_size, I40IW_CQPSQ_STAG_HPAGESIZE) |
+		 LS_64(info->remote_access, I40IW_CQPSQ_STAG_REMACCENABLED) |
+		 LS_64(info->use_hmc_fcn_index, I40IW_CQPSQ_STAG_USEHMCFNIDX) |
+		 LS_64(info->use_pf_rid, I40IW_CQPSQ_STAG_USEPFRID) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "ALLOC_STAG WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_mr_reg_non_shared - non-shared mr registration
+ * @dev: sc device struct
+ * @info: mr info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_mr_reg_non_shared(
+				struct i40iw_sc_dev *dev,
+				struct i40iw_reg_ns_stag_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 *wqe;
+	u64 temp;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+	u32 pble_obj_cnt;
+	bool remote_access;
+	u8 addr_type;
+
+	if (info->access_rights & (I40IW_ACCESS_FLAGS_REMOTEREAD_ONLY |
+				   I40IW_ACCESS_FLAGS_REMOTEWRITE_ONLY))
+		remote_access = true;
+	else
+		remote_access = false;
+
+	pble_obj_cnt = dev->hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt;
+
+	if (info->chunk_size && (info->first_pm_pbl_index >= pble_obj_cnt))
+		return I40IW_ERR_INVALID_PBLE_INDEX;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	temp = (info->addr_type == I40IW_ADDR_TYPE_VA_BASED) ? (uintptr_t)info->va : info->fbo;
+	set_64bit_val(wqe, 0, temp);
+
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(info->total_len, I40IW_CQPSQ_STAG_STAGLEN) |
+		      LS_64(info->pd_id, I40IW_CQPSQ_STAG_PDID));
+
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(info->stag_key, I40IW_CQPSQ_STAG_KEY) |
+		      LS_64(info->stag_idx, I40IW_CQPSQ_STAG_IDX));
+	if (!info->chunk_size) {
+		set_64bit_val(wqe, 32, info->reg_addr_pa);
+		set_64bit_val(wqe, 48, 0);
+	} else {
+		set_64bit_val(wqe, 32, 0);
+		set_64bit_val(wqe, 48, info->first_pm_pbl_index);
+	}
+	set_64bit_val(wqe, 40, info->hmc_fcn_index);
+	set_64bit_val(wqe, 56, 0);
+
+	addr_type = (info->addr_type == I40IW_ADDR_TYPE_VA_BASED) ? 1 : 0;
+	header = LS_64(I40IW_CQP_OP_REG_MR, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_STAG_MR) |
+		 LS_64(info->chunk_size, I40IW_CQPSQ_STAG_LPBLSIZE) |
+		 LS_64(info->page_size, I40IW_CQPSQ_STAG_HPAGESIZE) |
+		 LS_64(info->access_rights, I40IW_CQPSQ_STAG_ARIGHTS) |
+		 LS_64(remote_access, I40IW_CQPSQ_STAG_REMACCENABLED) |
+		 LS_64(addr_type, I40IW_CQPSQ_STAG_VABASEDTO) |
+		 LS_64(info->use_hmc_fcn_index, I40IW_CQPSQ_STAG_USEHMCFNIDX) |
+		 LS_64(info->use_pf_rid, I40IW_CQPSQ_STAG_USEPFRID) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "MR_REG_NS WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_mr_reg_shared - registered shared memory region
+ * @dev: sc device struct
+ * @info: info for shared memory registeration
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_mr_reg_shared(
+					struct i40iw_sc_dev *dev,
+					struct i40iw_register_shared_stag *info,
+					u64 scratch,
+					bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 temp, va64, fbo, header;
+	u32 va32;
+	bool remote_access;
+	u8 addr_type;
+
+	if (info->access_rights & (I40IW_ACCESS_FLAGS_REMOTEREAD_ONLY |
+				   I40IW_ACCESS_FLAGS_REMOTEWRITE_ONLY))
+		remote_access = true;
+	else
+		remote_access = false;
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	va64 = (uintptr_t)(info->va);
+	va32 = (u32)(va64 & 0x00000000FFFFFFFF);
+	fbo = (u64)(va32 & (4096 - 1));
+
+	set_64bit_val(wqe,
+		      0,
+		      (info->addr_type == I40IW_ADDR_TYPE_VA_BASED ? (uintptr_t)info->va : fbo));
+
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(info->pd_id, I40IW_CQPSQ_STAG_PDID));
+	temp = LS_64(info->new_stag_key, I40IW_CQPSQ_STAG_KEY) |
+	       LS_64(info->new_stag_idx, I40IW_CQPSQ_STAG_IDX) |
+	       LS_64(info->parent_stag_idx, I40IW_CQPSQ_STAG_PARENTSTAGIDX);
+	set_64bit_val(wqe, 16, temp);
+
+	addr_type = (info->addr_type == I40IW_ADDR_TYPE_VA_BASED) ? 1 : 0;
+	header = LS_64(I40IW_CQP_OP_REG_SMR, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_STAG_MR) |
+		 LS_64(info->access_rights, I40IW_CQPSQ_STAG_ARIGHTS) |
+		 LS_64(remote_access, I40IW_CQPSQ_STAG_REMACCENABLED) |
+		 LS_64(addr_type, I40IW_CQPSQ_STAG_VABASEDTO) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "MR_REG_SHARED WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_dealloc_stag - deallocate stag
+ * @dev: sc device struct
+ * @info: dealloc stag info
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_dealloc_stag(
+					struct i40iw_sc_dev *dev,
+					struct i40iw_dealloc_stag_info *info,
+					u64 scratch,
+					bool post_sq)
+{
+	u64 header;
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe,
+		      8,
+		      LS_64(info->pd_id, I40IW_CQPSQ_STAG_PDID));
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(info->stag_idx, I40IW_CQPSQ_STAG_IDX));
+
+	header = LS_64(I40IW_CQP_OP_DEALLOC_STAG, I40IW_CQPSQ_OPCODE) |
+		 LS_64(info->mr, I40IW_CQPSQ_STAG_MR) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "DEALLOC_STAG WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_query_stag - query hardware for stag
+ * @dev: sc device struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @stag_index: stag index for query
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_query_stag(struct i40iw_sc_dev *dev,
+						  u64 scratch,
+						  u32 stag_index,
+						  bool post_sq)
+{
+	u64 header;
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(stag_index, I40IW_CQPSQ_QUERYSTAG_IDX));
+
+	header = LS_64(I40IW_CQP_OP_QUERY_STAG, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "QUERY_STAG WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_mw_alloc - mw allocate
+ * @dev: sc device struct
+ * @scratch: u64 saved to be used during cqp completion
+ * @mw_stag_index:stag index
+ * @pd_id: pd is for this mw
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_mw_alloc(
+					struct i40iw_sc_dev *dev,
+					u64 scratch,
+					u32 mw_stag_index,
+					u16 pd_id,
+					bool post_sq)
+{
+	u64 header;
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe, 8, LS_64(pd_id, I40IW_CQPSQ_STAG_PDID));
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(mw_stag_index, I40IW_CQPSQ_STAG_IDX));
+
+	header = LS_64(I40IW_CQP_OP_ALLOC_STAG, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_WQE, "MW_ALLOC WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_send_lsmm - send last streaming mode message
+ * @qp: sc qp struct
+ * @lsmm_buf: buffer with lsmm message
+ * @size: size of lsmm buffer
+ * @stag: stag of lsmm buffer
+ */
+static void i40iw_sc_send_lsmm(struct i40iw_sc_qp *qp,
+			       void *lsmm_buf,
+			       u32 size,
+			       i40iw_stag stag)
+{
+	u64 *wqe;
+	u64 header;
+	struct i40iw_qp_uk *qp_uk;
+
+	qp_uk = &qp->qp_uk;
+	wqe = qp_uk->sq_base->elem;
+
+	set_64bit_val(wqe, 0, (uintptr_t)lsmm_buf);
+
+	set_64bit_val(wqe, 8, (size | LS_64(stag, I40IWQPSQ_FRAG_STAG)));
+
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64(I40IWQP_OP_RDMA_SEND, I40IWQPSQ_OPCODE) |
+		 LS_64(1, I40IWQPSQ_STREAMMODE) |
+		 LS_64(1, I40IWQPSQ_WAITFORRCVPDU) |
+		 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_QP, "SEND_LSMM WQE",
+			wqe, I40IW_QP_WQE_MIN_SIZE);
+}
+
+/**
+ * i40iw_sc_send_lsmm_nostag - for privilege qp
+ * @qp: sc qp struct
+ * @lsmm_buf: buffer with lsmm message
+ * @size: size of lsmm buffer
+ */
+static void i40iw_sc_send_lsmm_nostag(struct i40iw_sc_qp *qp,
+				      void *lsmm_buf,
+				      u32 size)
+{
+	u64 *wqe;
+	u64 header;
+	struct i40iw_qp_uk *qp_uk;
+
+	qp_uk = &qp->qp_uk;
+	wqe = qp_uk->sq_base->elem;
+
+	set_64bit_val(wqe, 0, (uintptr_t)lsmm_buf);
+
+	set_64bit_val(wqe, 8, size);
+
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64(I40IWQP_OP_RDMA_SEND, I40IWQPSQ_OPCODE) |
+		 LS_64(1, I40IWQPSQ_STREAMMODE) |
+		 LS_64(1, I40IWQPSQ_WAITFORRCVPDU) |
+		 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_WQE, "SEND_LSMM_NOSTAG WQE",
+			wqe, I40IW_QP_WQE_MIN_SIZE);
+}
+
+/**
+ * i40iw_sc_send_rtt - send last read0 or write0
+ * @qp: sc qp struct
+ * @read: Do read0 or write0
+ */
+static void i40iw_sc_send_rtt(struct i40iw_sc_qp *qp, bool read)
+{
+	u64 *wqe;
+	u64 header;
+	struct i40iw_qp_uk *qp_uk;
+
+	qp_uk = &qp->qp_uk;
+	wqe = qp_uk->sq_base->elem;
+
+	set_64bit_val(wqe, 0, 0);
+	set_64bit_val(wqe, 8, 0);
+	set_64bit_val(wqe, 16, 0);
+	if (read) {
+		header = LS_64(0x1234, I40IWQPSQ_REMSTAG) |
+			 LS_64(I40IWQP_OP_RDMA_READ, I40IWQPSQ_OPCODE) |
+			 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+		set_64bit_val(wqe, 8, ((u64)0xabcd << 32));
+	} else {
+		header = LS_64(I40IWQP_OP_RDMA_WRITE, I40IWQPSQ_OPCODE) |
+			 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+	}
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_WQE, "RTR WQE",
+			wqe, I40IW_QP_WQE_MIN_SIZE);
+}
+
+/**
+ * i40iw_sc_post_wqe0 - send wqe with opcode
+ * @qp: sc qp struct
+ * @opcode: opcode to use for wqe0
+ */
+static enum i40iw_status_code i40iw_sc_post_wqe0(struct i40iw_sc_qp *qp, u8 opcode)
+{
+	u64 *wqe;
+	u64 header;
+	struct i40iw_qp_uk *qp_uk;
+
+	qp_uk = &qp->qp_uk;
+	wqe = qp_uk->sq_base->elem;
+
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+	switch (opcode) {
+	case I40IWQP_OP_NOP:
+		set_64bit_val(wqe, 0, 0);
+		set_64bit_val(wqe, 8, 0);
+		set_64bit_val(wqe, 16, 0);
+		header = LS_64(I40IWQP_OP_NOP, I40IWQPSQ_OPCODE) |
+			 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
+
+		i40iw_insert_wqe_hdr(wqe, header);
+		break;
+	case I40IWQP_OP_RDMA_SEND:
+		set_64bit_val(wqe, 0, 0);
+		set_64bit_val(wqe, 8, 0);
+		set_64bit_val(wqe, 16, 0);
+		header = LS_64(I40IWQP_OP_RDMA_SEND, I40IWQPSQ_OPCODE) |
+			 LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID) |
+			 LS_64(1, I40IWQPSQ_STREAMMODE) |
+			 LS_64(1, I40IWQPSQ_WAITFORRCVPDU);
+
+		i40iw_insert_wqe_hdr(wqe, header);
+		break;
+	default:
+		i40iw_debug(qp->dev, I40IW_DEBUG_QP, "%s: Invalid WQE zero opcode\n",
+			    __func__);
+		break;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_sc_init_iw_hmc() - queries fpm values using cqp and populates hmc_info
+ * @dev : ptr to i40iw_dev struct
+ * @hmc_fn_id: hmc function id
+ */
+enum i40iw_status_code i40iw_sc_init_iw_hmc(struct i40iw_sc_dev *dev, u8 hmc_fn_id)
+{
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_dma_mem query_fpm_mem;
+	struct i40iw_virt_mem virt_mem;
+	struct i40iw_vfdev *vf_dev = NULL;
+	u32 mem_size;
+	enum i40iw_status_code ret_code = 0;
+	bool poll_registers = true;
+	u16 iw_vf_idx;
+	u8 wait_type;
+
+	if (hmc_fn_id >= I40IW_MAX_VF_FPM_ID ||
+	    (dev->hmc_fn_id != hmc_fn_id && hmc_fn_id < I40IW_FIRST_VF_FPM_ID))
+		return I40IW_ERR_INVALID_HMCFN_ID;
+
+	i40iw_debug(dev, I40IW_DEBUG_HMC, "hmc_fn_id %u, dev->hmc_fn_id %u\n", hmc_fn_id,
+		    dev->hmc_fn_id);
+	if (hmc_fn_id == dev->hmc_fn_id) {
+		hmc_info = dev->hmc_info;
+		query_fpm_mem.pa = dev->fpm_query_buf_pa;
+		query_fpm_mem.va = dev->fpm_query_buf;
+	} else {
+		vf_dev = i40iw_vfdev_from_fpm(dev, hmc_fn_id);
+		if (!vf_dev)
+			return I40IW_ERR_INVALID_VF_ID;
+
+		hmc_info = &vf_dev->hmc_info;
+		iw_vf_idx = vf_dev->iw_vf_idx;
+		i40iw_debug(dev, I40IW_DEBUG_HMC, "vf_dev %p, hmc_info %p, hmc_obj %p\n", vf_dev,
+			    hmc_info, hmc_info->hmc_obj);
+		if (!vf_dev->fpm_query_buf) {
+			if (!dev->vf_fpm_query_buf[iw_vf_idx].va) {
+				ret_code = i40iw_alloc_query_fpm_buf(dev,
+								     &dev->vf_fpm_query_buf[iw_vf_idx]);
+				if (ret_code)
+					return ret_code;
+			}
+			vf_dev->fpm_query_buf = dev->vf_fpm_query_buf[iw_vf_idx].va;
+			vf_dev->fpm_query_buf_pa = dev->vf_fpm_query_buf[iw_vf_idx].pa;
+		}
+		query_fpm_mem.pa = vf_dev->fpm_query_buf_pa;
+		query_fpm_mem.va = vf_dev->fpm_query_buf;
+		/**
+		 * It is HARDWARE specific:
+		 * this call is done by PF for VF and
+		 * i40iw_sc_query_fpm_values needs ccq poll
+		 * because PF ccq is already created.
+		 */
+		poll_registers = false;
+	}
+
+	hmc_info->hmc_fn_id = hmc_fn_id;
+
+	if (hmc_fn_id != dev->hmc_fn_id) {
+		ret_code =
+			i40iw_cqp_query_fpm_values_cmd(dev, &query_fpm_mem, hmc_fn_id);
+	} else {
+		wait_type = poll_registers ? (u8)I40IW_CQP_WAIT_POLL_REGS :
+			    (u8)I40IW_CQP_WAIT_POLL_CQ;
+
+		ret_code = i40iw_sc_query_fpm_values(
+					dev->cqp,
+					0,
+					hmc_info->hmc_fn_id,
+					&query_fpm_mem,
+					true,
+					wait_type);
+	}
+	if (ret_code)
+		return ret_code;
+
+	/* parse the fpm_query_buf and fill hmc obj info */
+	ret_code =
+		i40iw_sc_parse_fpm_query_buf((u64 *)query_fpm_mem.va,
+					     hmc_info,
+					     &dev->hmc_fpm_misc);
+	if (ret_code)
+		return ret_code;
+	i40iw_debug_buf(dev, I40IW_DEBUG_HMC, "QUERY FPM BUFFER",
+			query_fpm_mem.va, I40IW_QUERY_FPM_BUF_SIZE);
+
+	if (hmc_fn_id != dev->hmc_fn_id) {
+		i40iw_cqp_commit_fpm_values_cmd(dev, &query_fpm_mem, hmc_fn_id);
+
+		/* parse the fpm_commit_buf and fill hmc obj info */
+		i40iw_sc_parse_fpm_commit_buf((u64 *)query_fpm_mem.va, hmc_info->hmc_obj);
+		mem_size = sizeof(struct i40iw_hmc_sd_entry) *
+			   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index);
+		ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);
+		if (ret_code)
+			return ret_code;
+		hmc_info->sd_table.sd_entry = virt_mem.va;
+	}
+
+	/* fill size of objects which are fixed */
+	hmc_info->hmc_obj[I40IW_HMC_IW_XFFL].size = 4;
+	hmc_info->hmc_obj[I40IW_HMC_IW_Q1FL].size = 4;
+	hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].size = 8;
+	hmc_info->hmc_obj[I40IW_HMC_IW_APBVT_ENTRY].size = 8192;
+	hmc_info->hmc_obj[I40IW_HMC_IW_APBVT_ENTRY].max_cnt = 1;
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_configure_iw_fpm() - commits hmc obj cnt values using cqp command and
+ * populates fpm base address in hmc_info
+ * @dev : ptr to i40iw_dev struct
+ * @hmc_fn_id: hmc function id
+ */
+static enum i40iw_status_code i40iw_sc_configure_iw_fpm(struct i40iw_sc_dev *dev,
+							u8 hmc_fn_id)
+{
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_hmc_obj_info *obj_info;
+	u64 *buf;
+	struct i40iw_dma_mem commit_fpm_mem;
+	u32 i, j;
+	enum i40iw_status_code ret_code = 0;
+	bool poll_registers = true;
+	u8 wait_type;
+
+	if (hmc_fn_id >= I40IW_MAX_VF_FPM_ID ||
+	    (dev->hmc_fn_id != hmc_fn_id && hmc_fn_id < I40IW_FIRST_VF_FPM_ID))
+		return I40IW_ERR_INVALID_HMCFN_ID;
+
+	if (hmc_fn_id == dev->hmc_fn_id) {
+		hmc_info = dev->hmc_info;
+	} else {
+		hmc_info = i40iw_vf_hmcinfo_from_fpm(dev, hmc_fn_id);
+		poll_registers = false;
+	}
+	if (!hmc_info)
+		return I40IW_ERR_BAD_PTR;
+
+	obj_info = hmc_info->hmc_obj;
+	buf = dev->fpm_commit_buf;
+
+	/* copy cnt values in commit buf */
+	for (i = I40IW_HMC_IW_QP, j = 0; i <= I40IW_HMC_IW_PBLE;
+	     i++, j += 8)
+		set_64bit_val(buf, j, (u64)obj_info[i].cnt);
+
+	set_64bit_val(buf, 40, 0);   /* APBVT rsvd */
+
+	commit_fpm_mem.pa = dev->fpm_commit_buf_pa;
+	commit_fpm_mem.va = dev->fpm_commit_buf;
+	wait_type = poll_registers ? (u8)I40IW_CQP_WAIT_POLL_REGS :
+			(u8)I40IW_CQP_WAIT_POLL_CQ;
+	ret_code = i40iw_sc_commit_fpm_values(
+					dev->cqp,
+					0,
+					hmc_info->hmc_fn_id,
+					&commit_fpm_mem,
+					true,
+					wait_type);
+
+	/* parse the fpm_commit_buf and fill hmc obj info */
+	if (!ret_code)
+		ret_code = i40iw_sc_parse_fpm_commit_buf(dev->fpm_commit_buf, hmc_info->hmc_obj);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_HMC, "COMMIT FPM BUFFER",
+			commit_fpm_mem.va, I40IW_COMMIT_FPM_BUF_SIZE);
+
+	return ret_code;
+}
+
+/**
+ * cqp_sds_wqe_fill - fill cqp wqe doe sd
+ * @cqp: struct for cqp hw
+ * @info; sd info for wqe
+ * @scratch: u64 saved to be used during cqp completion
+ */
+static enum i40iw_status_code cqp_sds_wqe_fill(struct i40iw_sc_cqp *cqp,
+					       struct i40iw_update_sds_info *info,
+					       u64 scratch)
+{
+	u64 data;
+	u64 header;
+	u64 *wqe;
+	int mem_entries, wqe_entries;
+	struct i40iw_dma_mem *sdbuf = &cqp->sdbuf;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	I40IW_CQP_INIT_WQE(wqe);
+	wqe_entries = (info->cnt > 3) ? 3 : info->cnt;
+	mem_entries = info->cnt - wqe_entries;
+
+	header = LS_64(I40IW_CQP_OP_UPDATE_PE_SDS, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID) |
+		 LS_64(mem_entries, I40IW_CQPSQ_UPESD_ENTRY_COUNT);
+
+	if (mem_entries) {
+		memcpy(sdbuf->va, &info->entry[3], (mem_entries << 4));
+		data = sdbuf->pa;
+	} else {
+		data = 0;
+	}
+	data |= LS_64(info->hmc_fn_id, I40IW_CQPSQ_UPESD_HMCFNID);
+
+	set_64bit_val(wqe, 16, data);
+
+	switch (wqe_entries) {
+	case 3:
+		set_64bit_val(wqe, 48,
+			      (LS_64(info->entry[2].cmd, I40IW_CQPSQ_UPESD_SDCMD) |
+					LS_64(1, I40IW_CQPSQ_UPESD_ENTRY_VALID)));
+
+		set_64bit_val(wqe, 56, info->entry[2].data);
+		/* fallthrough */
+	case 2:
+		set_64bit_val(wqe, 32,
+			      (LS_64(info->entry[1].cmd, I40IW_CQPSQ_UPESD_SDCMD) |
+					LS_64(1, I40IW_CQPSQ_UPESD_ENTRY_VALID)));
+
+		set_64bit_val(wqe, 40, info->entry[1].data);
+		/* fallthrough */
+	case 1:
+		set_64bit_val(wqe, 0,
+			      LS_64(info->entry[0].cmd, I40IW_CQPSQ_UPESD_SDCMD));
+
+		set_64bit_val(wqe, 8, info->entry[0].data);
+		break;
+	default:
+		break;
+	}
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "UPDATE_PE_SDS WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	return 0;
+}
+
+/**
+ * i40iw_update_pe_sds - cqp wqe for sd
+ * @dev: ptr to i40iw_dev struct
+ * @info: sd info for sd's
+ * @scratch: u64 saved to be used during cqp completion
+ */
+static enum i40iw_status_code i40iw_update_pe_sds(struct i40iw_sc_dev *dev,
+						  struct i40iw_update_sds_info *info,
+						  u64 scratch)
+{
+	struct i40iw_sc_cqp *cqp = dev->cqp;
+	enum i40iw_status_code ret_code;
+
+	ret_code = cqp_sds_wqe_fill(cqp, info, scratch);
+	if (!ret_code)
+		i40iw_sc_cqp_post_sq(cqp);
+
+	return ret_code;
+}
+
+/**
+ * i40iw_update_sds_noccq - update sd before ccq created
+ * @dev: sc device struct
+ * @info: sd info for sd's
+ */
+enum i40iw_status_code i40iw_update_sds_noccq(struct i40iw_sc_dev *dev,
+					      struct i40iw_update_sds_info *info)
+{
+	u32 error, val, tail;
+	struct i40iw_sc_cqp *cqp = dev->cqp;
+	enum i40iw_status_code ret_code;
+
+	ret_code = cqp_sds_wqe_fill(cqp, info, 0);
+	if (ret_code)
+		return ret_code;
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+	if (error)
+		return I40IW_ERR_CQP_COMPL_ERROR;
+
+	i40iw_sc_cqp_post_sq(cqp);
+	ret_code = i40iw_cqp_poll_registers(cqp, tail, I40IW_DONE_COUNT);
+
+	return ret_code;
+}
+
+/**
+ * i40iw_sc_suspend_qp - suspend qp for param change
+ * @cqp: struct for cqp hw
+ * @qp: sc qp struct
+ * @scratch: u64 saved to be used during cqp completion
+ */
+enum i40iw_status_code i40iw_sc_suspend_qp(struct i40iw_sc_cqp *cqp,
+					   struct i40iw_sc_qp *qp,
+					   u64 scratch)
+{
+	u64 header;
+	u64 *wqe;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	header = LS_64(qp->qp_uk.qp_id, I40IW_CQPSQ_SUSPENDQP_QPID) |
+		 LS_64(I40IW_CQP_OP_SUSPEND_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "SUSPEND_QP WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_resume_qp - resume qp after suspend
+ * @cqp: struct for cqp hw
+ * @qp: sc qp struct
+ * @scratch: u64 saved to be used during cqp completion
+ */
+enum i40iw_status_code i40iw_sc_resume_qp(struct i40iw_sc_cqp *cqp,
+					  struct i40iw_sc_qp *qp,
+					  u64 scratch)
+{
+	u64 header;
+	u64 *wqe;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe,
+		      16,
+			LS_64(qp->qs_handle, I40IW_CQPSQ_RESUMEQP_QSHANDLE));
+
+	header = LS_64(qp->qp_uk.qp_id, I40IW_CQPSQ_RESUMEQP_QPID) |
+		 LS_64(I40IW_CQP_OP_RESUME_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "RESUME_QP WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+/**
+ * i40iw_sc_static_hmc_pages_allocated - cqp wqe to allocate hmc pages
+ * @cqp: struct for cqp hw
+ * @scratch: u64 saved to be used during cqp completion
+ * @hmc_fn_id: hmc function id
+ * @post_sq: flag for cqp db to ring
+ * @poll_registers: flag to poll register for cqp completion
+ */
+enum i40iw_status_code i40iw_sc_static_hmc_pages_allocated(
+					struct i40iw_sc_cqp *cqp,
+					u64 scratch,
+					u8 hmc_fn_id,
+					bool post_sq,
+					bool poll_registers)
+{
+	u64 header;
+	u64 *wqe;
+	u32 tail, val, error;
+	enum i40iw_status_code ret_code = 0;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+	set_64bit_val(wqe,
+		      16,
+		      LS_64(hmc_fn_id, I40IW_SHMC_PAGE_ALLOCATED_HMC_FN_ID));
+
+	header = LS_64(I40IW_CQP_OP_SHMC_PAGES_ALLOCATED, I40IW_CQPSQ_OPCODE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "SHMC_PAGES_ALLOCATED WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+	i40iw_get_cqp_reg_info(cqp, &val, &tail, &error);
+	if (error) {
+		ret_code = I40IW_ERR_CQP_COMPL_ERROR;
+		return ret_code;
+	}
+	if (post_sq) {
+		i40iw_sc_cqp_post_sq(cqp);
+		if (poll_registers)
+			/* check for cqp sq tail update */
+			ret_code = i40iw_cqp_poll_registers(cqp, tail, 1000);
+		else
+			ret_code = i40iw_sc_poll_for_cqp_op_done(cqp,
+								 I40IW_CQP_OP_SHMC_PAGES_ALLOCATED,
+								 NULL);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_ring_full - check if cqp ring is full
+ * @cqp: struct for cqp hw
+ */
+static bool i40iw_ring_full(struct i40iw_sc_cqp *cqp)
+{
+	return I40IW_RING_FULL_ERR(cqp->sq_ring);
+}
+
+/**
+ * i40iw_config_fpm_values - configure HMC objects
+ * @dev: sc device struct
+ * @qp_count: desired qp count
+ */
+enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_count)
+{
+	struct i40iw_virt_mem virt_mem;
+	u32 i, mem_size;
+	u32 qpwantedoriginal, qpwanted, mrwanted, pblewanted;
+	u32 powerof2;
+	u64 sd_needed, bytes_needed;
+	u32 loop_count = 0;
+
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_hmc_fpm_misc *hmc_fpm_misc;
+	enum i40iw_status_code ret_code = 0;
+
+	hmc_info = dev->hmc_info;
+	hmc_fpm_misc = &dev->hmc_fpm_misc;
+
+	ret_code = i40iw_sc_init_iw_hmc(dev, dev->hmc_fn_id);
+	if (ret_code) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "i40iw_sc_init_iw_hmc returned error_code = %d\n",
+			    ret_code);
+		return ret_code;
+	}
+
+	bytes_needed = 0;
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
+		hmc_info->hmc_obj[i].cnt = hmc_info->hmc_obj[i].max_cnt;
+		bytes_needed +=
+		    (hmc_info->hmc_obj[i].max_cnt) * (hmc_info->hmc_obj[i].size);
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s i[%04d] max_cnt[0x%04X] size[0x%04llx]\n",
+			    __func__, i, hmc_info->hmc_obj[i].max_cnt,
+			    hmc_info->hmc_obj[i].size);
+	}
+	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1; /* round up */
+	i40iw_debug(dev, I40IW_DEBUG_HMC,
+		    "%s: FW initial max sd_count[%08lld] first_sd_index[%04d]\n",
+		    __func__, sd_needed, hmc_info->first_sd_index);
+	i40iw_debug(dev, I40IW_DEBUG_HMC,
+		    "%s: bytes_needed=0x%llx sd count %d where max sd is %d\n",
+		    __func__, bytes_needed, hmc_info->sd_table.sd_cnt,
+		    hmc_fpm_misc->max_sds);
+
+	qpwanted = min(qp_count, hmc_info->hmc_obj[I40IW_HMC_IW_QP].max_cnt);
+	qpwantedoriginal = qpwanted;
+	mrwanted = hmc_info->hmc_obj[I40IW_HMC_IW_MR].max_cnt;
+	pblewanted = hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].max_cnt;
+
+	i40iw_debug(dev, I40IW_DEBUG_HMC,
+		    "req_qp=%d max_sd=%d, max_qp = %d, max_cq=%d, max_mr=%d, max_pble=%d\n",
+		    qp_count, hmc_fpm_misc->max_sds,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_QP].max_cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_CQ].max_cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_MR].max_cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].max_cnt);
+
+	do {
+		++loop_count;
+		hmc_info->hmc_obj[I40IW_HMC_IW_QP].cnt = qpwanted;
+		hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt =
+			min(2 * qpwanted, hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt);
+		hmc_info->hmc_obj[I40IW_HMC_IW_SRQ].cnt = 0x00; /* Reserved */
+		hmc_info->hmc_obj[I40IW_HMC_IW_HTE].cnt =
+					qpwanted * hmc_fpm_misc->ht_multiplier;
+		hmc_info->hmc_obj[I40IW_HMC_IW_ARP].cnt =
+			hmc_info->hmc_obj[I40IW_HMC_IW_ARP].max_cnt;
+		hmc_info->hmc_obj[I40IW_HMC_IW_APBVT_ENTRY].cnt = 1;
+		hmc_info->hmc_obj[I40IW_HMC_IW_MR].cnt = mrwanted;
+
+		hmc_info->hmc_obj[I40IW_HMC_IW_XF].cnt = I40IW_MAX_WQ_ENTRIES * qpwanted;
+		hmc_info->hmc_obj[I40IW_HMC_IW_Q1].cnt = 4 * I40IW_MAX_IRD_SIZE * qpwanted;
+		hmc_info->hmc_obj[I40IW_HMC_IW_XFFL].cnt =
+			hmc_info->hmc_obj[I40IW_HMC_IW_XF].cnt / hmc_fpm_misc->xf_block_size;
+		hmc_info->hmc_obj[I40IW_HMC_IW_Q1FL].cnt =
+			hmc_info->hmc_obj[I40IW_HMC_IW_Q1].cnt / hmc_fpm_misc->q1_block_size;
+		hmc_info->hmc_obj[I40IW_HMC_IW_TIMER].cnt =
+			((qpwanted) / 512 + 1) * hmc_fpm_misc->timer_bucket;
+		hmc_info->hmc_obj[I40IW_HMC_IW_FSIMC].cnt = 0x00;
+		hmc_info->hmc_obj[I40IW_HMC_IW_FSIAV].cnt = 0x00;
+		hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt = pblewanted;
+
+		/* How much memory is needed for all the objects. */
+		bytes_needed = 0;
+		for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
+			bytes_needed +=
+			    (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
+		sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;
+		if ((loop_count > 1000) ||
+		    ((!(loop_count % 10)) &&
+		    (qpwanted > qpwantedoriginal * 2 / 3))) {
+			if (qpwanted > FPM_MULTIPLIER) {
+				qpwanted -= FPM_MULTIPLIER;
+				powerof2 = 1;
+				while (powerof2 < qpwanted)
+					powerof2 *= 2;
+				powerof2 /= 2;
+				qpwanted = powerof2;
+			} else {
+				qpwanted /= 2;
+			}
+		}
+		if (mrwanted > FPM_MULTIPLIER * 10)
+			mrwanted -= FPM_MULTIPLIER * 10;
+		if (pblewanted > FPM_MULTIPLIER * 1000)
+			pblewanted -= FPM_MULTIPLIER * 1000;
+	} while (sd_needed > hmc_fpm_misc->max_sds && loop_count < 2000);
+
+	bytes_needed = 0;
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
+		bytes_needed += (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s i[%04d] cnt[0x%04x] size[0x%04llx]\n",
+			    __func__, i, hmc_info->hmc_obj[i].cnt,
+			    hmc_info->hmc_obj[i].size);
+	}
+	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;    /* round up not truncate. */
+
+	i40iw_debug(dev, I40IW_DEBUG_HMC,
+		    "loop_cnt=%d, sd_needed=%lld, qpcnt = %d, cqcnt=%d, mrcnt=%d, pblecnt=%d\n",
+		    loop_count, sd_needed,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_QP].cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_MR].cnt,
+		    hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt);
+
+	ret_code = i40iw_sc_configure_iw_fpm(dev, dev->hmc_fn_id);
+	if (ret_code) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "configure_iw_fpm returned error_code[x%08X]\n",
+			    i40iw_rd32(dev->hw, dev->is_pf ? I40E_PFPE_CQPERRCODES : I40E_VFPE_CQPERRCODES1));
+		return ret_code;
+	}
+
+	hmc_info->sd_table.sd_cnt = (u32)sd_needed;
+
+	mem_size = sizeof(struct i40iw_hmc_sd_entry) *
+		   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index + 1);
+	ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);
+	if (ret_code) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s: failed to allocate memory for sd_entry buffer\n",
+			    __func__);
+		return ret_code;
+	}
+	hmc_info->sd_table.sd_entry = virt_mem.va;
+
+	return ret_code;
+}
+
+/**
+ * i40iw_exec_cqp_cmd - execute cqp cmd when wqe are available
+ * @dev: rdma device
+ * @pcmdinfo: cqp command info
+ */
+static enum i40iw_status_code i40iw_exec_cqp_cmd(struct i40iw_sc_dev *dev,
+						 struct cqp_commands_info *pcmdinfo)
+{
+	enum i40iw_status_code status;
+	struct i40iw_dma_mem values_mem;
+
+	dev->cqp_cmd_stats[pcmdinfo->cqp_cmd]++;
+	switch (pcmdinfo->cqp_cmd) {
+	case OP_DELETE_LOCAL_MAC_IPADDR_ENTRY:
+		status = i40iw_sc_del_local_mac_ipaddr_entry(
+				pcmdinfo->in.u.del_local_mac_ipaddr_entry.cqp,
+				pcmdinfo->in.u.del_local_mac_ipaddr_entry.scratch,
+				pcmdinfo->in.u.del_local_mac_ipaddr_entry.entry_idx,
+				pcmdinfo->in.u.del_local_mac_ipaddr_entry.ignore_ref_count,
+				pcmdinfo->post_sq);
+		break;
+	case OP_CEQ_DESTROY:
+		status = i40iw_sc_ceq_destroy(pcmdinfo->in.u.ceq_destroy.ceq,
+					      pcmdinfo->in.u.ceq_destroy.scratch,
+					      pcmdinfo->post_sq);
+		break;
+	case OP_AEQ_DESTROY:
+		status = i40iw_sc_aeq_destroy(pcmdinfo->in.u.aeq_destroy.aeq,
+					      pcmdinfo->in.u.aeq_destroy.scratch,
+					      pcmdinfo->post_sq);
+
+		break;
+	case OP_DELETE_ARP_CACHE_ENTRY:
+		status = i40iw_sc_del_arp_cache_entry(
+				pcmdinfo->in.u.del_arp_cache_entry.cqp,
+				pcmdinfo->in.u.del_arp_cache_entry.scratch,
+				pcmdinfo->in.u.del_arp_cache_entry.arp_index,
+				pcmdinfo->post_sq);
+		break;
+	case OP_MANAGE_APBVT_ENTRY:
+		status = i40iw_sc_manage_apbvt_entry(
+				pcmdinfo->in.u.manage_apbvt_entry.cqp,
+				&pcmdinfo->in.u.manage_apbvt_entry.info,
+				pcmdinfo->in.u.manage_apbvt_entry.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_CEQ_CREATE:
+		status = i40iw_sc_ceq_create(pcmdinfo->in.u.ceq_create.ceq,
+					     pcmdinfo->in.u.ceq_create.scratch,
+					     pcmdinfo->post_sq);
+		break;
+	case OP_AEQ_CREATE:
+		status = i40iw_sc_aeq_create(pcmdinfo->in.u.aeq_create.aeq,
+					     pcmdinfo->in.u.aeq_create.scratch,
+					     pcmdinfo->post_sq);
+		break;
+	case OP_ALLOC_LOCAL_MAC_IPADDR_ENTRY:
+		status = i40iw_sc_alloc_local_mac_ipaddr_entry(
+				pcmdinfo->in.u.alloc_local_mac_ipaddr_entry.cqp,
+				pcmdinfo->in.u.alloc_local_mac_ipaddr_entry.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_ADD_LOCAL_MAC_IPADDR_ENTRY:
+		status = i40iw_sc_add_local_mac_ipaddr_entry(
+				pcmdinfo->in.u.add_local_mac_ipaddr_entry.cqp,
+				&pcmdinfo->in.u.add_local_mac_ipaddr_entry.info,
+				pcmdinfo->in.u.add_local_mac_ipaddr_entry.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_MANAGE_QHASH_TABLE_ENTRY:
+		status = i40iw_sc_manage_qhash_table_entry(
+				pcmdinfo->in.u.manage_qhash_table_entry.cqp,
+				&pcmdinfo->in.u.manage_qhash_table_entry.info,
+				pcmdinfo->in.u.manage_qhash_table_entry.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_QP_MODIFY:
+		status = i40iw_sc_qp_modify(
+				pcmdinfo->in.u.qp_modify.qp,
+				&pcmdinfo->in.u.qp_modify.info,
+				pcmdinfo->in.u.qp_modify.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_QP_UPLOAD_CONTEXT:
+		status = i40iw_sc_qp_upload_context(
+				pcmdinfo->in.u.qp_upload_context.dev,
+				&pcmdinfo->in.u.qp_upload_context.info,
+				pcmdinfo->in.u.qp_upload_context.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_CQ_CREATE:
+		status = i40iw_sc_cq_create(
+				pcmdinfo->in.u.cq_create.cq,
+				pcmdinfo->in.u.cq_create.scratch,
+				pcmdinfo->in.u.cq_create.check_overflow,
+				pcmdinfo->post_sq);
+		break;
+	case OP_CQ_DESTROY:
+		status = i40iw_sc_cq_destroy(
+				pcmdinfo->in.u.cq_destroy.cq,
+				pcmdinfo->in.u.cq_destroy.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_QP_CREATE:
+		status = i40iw_sc_qp_create(
+				pcmdinfo->in.u.qp_create.qp,
+				&pcmdinfo->in.u.qp_create.info,
+				pcmdinfo->in.u.qp_create.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_QP_DESTROY:
+		status = i40iw_sc_qp_destroy(
+				pcmdinfo->in.u.qp_destroy.qp,
+				pcmdinfo->in.u.qp_destroy.scratch,
+				pcmdinfo->in.u.qp_destroy.remove_hash_idx,
+				pcmdinfo->in.u.qp_destroy.
+				ignore_mw_bnd,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_ALLOC_STAG:
+		status = i40iw_sc_alloc_stag(
+				pcmdinfo->in.u.alloc_stag.dev,
+				&pcmdinfo->in.u.alloc_stag.info,
+				pcmdinfo->in.u.alloc_stag.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_MR_REG_NON_SHARED:
+		status = i40iw_sc_mr_reg_non_shared(
+				pcmdinfo->in.u.mr_reg_non_shared.dev,
+				&pcmdinfo->in.u.mr_reg_non_shared.info,
+				pcmdinfo->in.u.mr_reg_non_shared.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_DEALLOC_STAG:
+		status = i40iw_sc_dealloc_stag(
+				pcmdinfo->in.u.dealloc_stag.dev,
+				&pcmdinfo->in.u.dealloc_stag.info,
+				pcmdinfo->in.u.dealloc_stag.scratch,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_MW_ALLOC:
+		status = i40iw_sc_mw_alloc(
+				pcmdinfo->in.u.mw_alloc.dev,
+				pcmdinfo->in.u.mw_alloc.scratch,
+				pcmdinfo->in.u.mw_alloc.mw_stag_index,
+				pcmdinfo->in.u.mw_alloc.pd_id,
+				pcmdinfo->post_sq);
+
+		break;
+	case OP_QP_FLUSH_WQES:
+		status = i40iw_sc_qp_flush_wqes(
+				pcmdinfo->in.u.qp_flush_wqes.qp,
+				&pcmdinfo->in.u.qp_flush_wqes.info,
+				pcmdinfo->in.u.qp_flush_wqes.
+				scratch, pcmdinfo->post_sq);
+		break;
+	case OP_ADD_ARP_CACHE_ENTRY:
+		status = i40iw_sc_add_arp_cache_entry(
+				pcmdinfo->in.u.add_arp_cache_entry.cqp,
+				&pcmdinfo->in.u.add_arp_cache_entry.info,
+				pcmdinfo->in.u.add_arp_cache_entry.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_MANAGE_PUSH_PAGE:
+		status = i40iw_sc_manage_push_page(
+				pcmdinfo->in.u.manage_push_page.cqp,
+				&pcmdinfo->in.u.manage_push_page.info,
+				pcmdinfo->in.u.manage_push_page.scratch,
+				pcmdinfo->post_sq);
+		break;
+	case OP_UPDATE_PE_SDS:
+		/* case I40IW_CQP_OP_UPDATE_PE_SDS */
+		status = i40iw_update_pe_sds(
+				pcmdinfo->in.u.update_pe_sds.dev,
+				&pcmdinfo->in.u.update_pe_sds.info,
+				pcmdinfo->in.u.update_pe_sds.
+				scratch);
+
+		break;
+	case OP_MANAGE_HMC_PM_FUNC_TABLE:
+		status = i40iw_sc_manage_hmc_pm_func_table(
+				pcmdinfo->in.u.manage_hmc_pm.dev->cqp,
+				pcmdinfo->in.u.manage_hmc_pm.scratch,
+				(u8)pcmdinfo->in.u.manage_hmc_pm.info.vf_id,
+				pcmdinfo->in.u.manage_hmc_pm.info.free_fcn,
+				true);
+		break;
+	case OP_SUSPEND:
+		status = i40iw_sc_suspend_qp(
+				pcmdinfo->in.u.suspend_resume.cqp,
+				pcmdinfo->in.u.suspend_resume.qp,
+				pcmdinfo->in.u.suspend_resume.scratch);
+		break;
+	case OP_RESUME:
+		status = i40iw_sc_resume_qp(
+				pcmdinfo->in.u.suspend_resume.cqp,
+				pcmdinfo->in.u.suspend_resume.qp,
+				pcmdinfo->in.u.suspend_resume.scratch);
+		break;
+	case OP_MANAGE_VF_PBLE_BP:
+		status = i40iw_manage_vf_pble_bp(
+				pcmdinfo->in.u.manage_vf_pble_bp.cqp,
+				&pcmdinfo->in.u.manage_vf_pble_bp.info,
+				pcmdinfo->in.u.manage_vf_pble_bp.scratch, true);
+		break;
+	case OP_QUERY_FPM_VALUES:
+		values_mem.pa = pcmdinfo->in.u.query_fpm_values.fpm_values_pa;
+		values_mem.va = pcmdinfo->in.u.query_fpm_values.fpm_values_va;
+		status = i40iw_sc_query_fpm_values(
+				pcmdinfo->in.u.query_fpm_values.cqp,
+				pcmdinfo->in.u.query_fpm_values.scratch,
+				pcmdinfo->in.u.query_fpm_values.hmc_fn_id,
+				&values_mem, true, I40IW_CQP_WAIT_EVENT);
+		break;
+	case OP_COMMIT_FPM_VALUES:
+		values_mem.pa = pcmdinfo->in.u.commit_fpm_values.fpm_values_pa;
+		values_mem.va = pcmdinfo->in.u.commit_fpm_values.fpm_values_va;
+		status = i40iw_sc_commit_fpm_values(
+				pcmdinfo->in.u.commit_fpm_values.cqp,
+				pcmdinfo->in.u.commit_fpm_values.scratch,
+				pcmdinfo->in.u.commit_fpm_values.hmc_fn_id,
+				&values_mem,
+				true,
+				I40IW_CQP_WAIT_EVENT);
+		break;
+	default:
+		status = I40IW_NOT_SUPPORTED;
+		break;
+	}
+
+	return status;
+}
+
+/**
+ * i40iw_process_cqp_cmd - process all cqp commands
+ * @dev: sc device struct
+ * @pcmdinfo: cqp command info
+ */
+enum i40iw_status_code i40iw_process_cqp_cmd(struct i40iw_sc_dev *dev,
+					     struct cqp_commands_info *pcmdinfo)
+{
+	enum i40iw_status_code status = 0;
+	unsigned long	flags;
+
+	spin_lock_irqsave(&dev->cqp_lock, flags);
+	if (list_empty(&dev->cqp_cmd_head) && !i40iw_ring_full(dev->cqp))
+		status = i40iw_exec_cqp_cmd(dev, pcmdinfo);
+	else
+		list_add_tail(&pcmdinfo->cqp_cmd_entry, &dev->cqp_cmd_head);
+	spin_unlock_irqrestore(&dev->cqp_lock, flags);
+	return status;
+}
+
+/**
+ * i40iw_process_bh - called from tasklet for cqp list
+ * @dev: sc device struct
+ */
+enum i40iw_status_code i40iw_process_bh(struct i40iw_sc_dev *dev)
+{
+	enum i40iw_status_code status = 0;
+	struct cqp_commands_info *pcmdinfo;
+	unsigned long	flags;
+
+	spin_lock_irqsave(&dev->cqp_lock, flags);
+	while (!list_empty(&dev->cqp_cmd_head) && !i40iw_ring_full(dev->cqp)) {
+		pcmdinfo = (struct cqp_commands_info *)i40iw_remove_head(&dev->cqp_cmd_head);
+
+		status = i40iw_exec_cqp_cmd(dev, pcmdinfo);
+		if (status)
+			break;
+	}
+	spin_unlock_irqrestore(&dev->cqp_lock, flags);
+	return status;
+}
+
+/**
+ * i40iw_iwarp_opcode - determine if incoming is rdma layer
+ * @info: aeq info for the packet
+ * @pkt: packet for error
+ */
+static u32 i40iw_iwarp_opcode(struct i40iw_aeqe_info *info, u8 *pkt)
+{
+	u16 *mpa;
+	u32 opcode = 0xffffffff;
+
+	if (info->q2_data_written) {
+		mpa = (u16 *)pkt;
+		opcode = ntohs(mpa[1]) & 0xf;
+	}
+	return opcode;
+}
+
+/**
+ * i40iw_locate_mpa - return pointer to mpa in the pkt
+ * @pkt: packet with data
+ */
+static u8 *i40iw_locate_mpa(u8 *pkt)
+{
+	/* skip over ethernet header */
+	pkt += I40IW_MAC_HLEN;
+
+	/* Skip over IP and TCP headers */
+	pkt += 4 * (pkt[0] & 0x0f);
+	pkt += 4 * ((pkt[12] >> 4) & 0x0f);
+	return pkt;
+}
+
+/**
+ * i40iw_setup_termhdr - termhdr for terminate pkt
+ * @qp: sc qp ptr for pkt
+ * @hdr: term hdr
+ * @opcode: flush opcode for termhdr
+ * @layer_etype: error layer + error type
+ * @err: error cod ein the header
+ */
+static void i40iw_setup_termhdr(struct i40iw_sc_qp *qp,
+				struct i40iw_terminate_hdr *hdr,
+				enum i40iw_flush_opcode opcode,
+				u8 layer_etype,
+				u8 err)
+{
+	qp->flush_code = opcode;
+	hdr->layer_etype = layer_etype;
+	hdr->error_code = err;
+}
+
+/**
+ * i40iw_bld_terminate_hdr - build terminate message header
+ * @qp: qp associated with received terminate AE
+ * @info: the struct contiaing AE information
+ */
+static int i40iw_bld_terminate_hdr(struct i40iw_sc_qp *qp,
+				   struct i40iw_aeqe_info *info)
+{
+	u8 *pkt = qp->q2_buf + Q2_BAD_FRAME_OFFSET;
+	u16 ddp_seg_len;
+	int copy_len = 0;
+	u8 is_tagged = 0;
+	enum i40iw_flush_opcode flush_code = FLUSH_INVALID;
+	u32 opcode;
+	struct i40iw_terminate_hdr *termhdr;
+
+	termhdr = (struct i40iw_terminate_hdr *)qp->q2_buf;
+	memset(termhdr, 0, Q2_BAD_FRAME_OFFSET);
+
+	if (info->q2_data_written) {
+		/* Use data from offending packet to fill in ddp & rdma hdrs */
+		pkt = i40iw_locate_mpa(pkt);
+		ddp_seg_len = ntohs(*(u16 *)pkt);
+		if (ddp_seg_len) {
+			copy_len = 2;
+			termhdr->hdrct = DDP_LEN_FLAG;
+			if (pkt[2] & 0x80) {
+				is_tagged = 1;
+				if (ddp_seg_len >= TERM_DDP_LEN_TAGGED) {
+					copy_len += TERM_DDP_LEN_TAGGED;
+					termhdr->hdrct |= DDP_HDR_FLAG;
+				}
+			} else {
+				if (ddp_seg_len >= TERM_DDP_LEN_UNTAGGED) {
+					copy_len += TERM_DDP_LEN_UNTAGGED;
+					termhdr->hdrct |= DDP_HDR_FLAG;
+				}
+
+				if (ddp_seg_len >= (TERM_DDP_LEN_UNTAGGED + TERM_RDMA_LEN)) {
+					if ((pkt[3] & RDMA_OPCODE_MASK) == RDMA_READ_REQ_OPCODE) {
+						copy_len += TERM_RDMA_LEN;
+						termhdr->hdrct |= RDMA_HDR_FLAG;
+					}
+				}
+			}
+		}
+	}
+
+	opcode = i40iw_iwarp_opcode(info, pkt);
+
+	switch (info->ae_id) {
+	case I40IW_AE_AMP_UNALLOCATED_STAG:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		if (opcode == I40IW_OP_TYPE_RDMA_WRITE)
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_PROT_ERR,
+					    (LAYER_DDP << 4) | DDP_TAGGED_BUFFER, DDP_TAGGED_INV_STAG);
+		else
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_INV_STAG);
+		break;
+	case I40IW_AE_AMP_BOUNDS_VIOLATION:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		if (info->q2_data_written)
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_PROT_ERR,
+					    (LAYER_DDP << 4) | DDP_TAGGED_BUFFER, DDP_TAGGED_BOUNDS);
+		else
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_INV_BOUNDS);
+		break;
+	case I40IW_AE_AMP_BAD_PD:
+		switch (opcode) {
+		case I40IW_OP_TYPE_RDMA_WRITE:
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_PROT_ERR,
+					    (LAYER_DDP << 4) | DDP_TAGGED_BUFFER, DDP_TAGGED_UNASSOC_STAG);
+			break;
+		case I40IW_OP_TYPE_SEND_INV:
+		case I40IW_OP_TYPE_SEND_SOL_INV:
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_CANT_INV_STAG);
+			break;
+		default:
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_UNASSOC_STAG);
+		}
+		break;
+	case I40IW_AE_AMP_INVALID_STAG:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_INV_STAG);
+		break;
+	case I40IW_AE_AMP_BAD_QP:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_LOC_QP_OP_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_QN);
+		break;
+	case I40IW_AE_AMP_BAD_STAG_KEY:
+	case I40IW_AE_AMP_BAD_STAG_INDEX:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		switch (opcode) {
+		case I40IW_OP_TYPE_SEND_INV:
+		case I40IW_OP_TYPE_SEND_SOL_INV:
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_OP_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_OP, RDMAP_CANT_INV_STAG);
+			break;
+		default:
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+					    (LAYER_RDMA << 4) | RDMAP_REMOTE_OP, RDMAP_INV_STAG);
+		}
+		break;
+	case I40IW_AE_AMP_RIGHTS_VIOLATION:
+	case I40IW_AE_AMP_INVALIDATE_NO_REMOTE_ACCESS_RIGHTS:
+	case I40IW_AE_PRIV_OPERATION_DENIED:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_ACCESS);
+		break;
+	case I40IW_AE_AMP_TO_WRAP:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_ACCESS_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT, RDMAP_TO_WRAP);
+		break;
+	case I40IW_AE_LLP_RECEIVED_MARKER_AND_LENGTH_FIELDS_DONT_MATCH:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_LOC_LEN_ERR,
+				    (LAYER_MPA << 4) | DDP_LLP, MPA_MARKER);
+		break;
+	case I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+				    (LAYER_MPA << 4) | DDP_LLP, MPA_CRC);
+		break;
+	case I40IW_AE_LLP_SEGMENT_TOO_LARGE:
+	case I40IW_AE_LLP_SEGMENT_TOO_SMALL:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_LOC_LEN_ERR,
+				    (LAYER_DDP << 4) | DDP_CATASTROPHIC, DDP_CATASTROPHIC_LOCAL);
+		break;
+	case I40IW_AE_LCE_QP_CATASTROPHIC:
+	case I40IW_AE_DDP_NO_L_BIT:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_FATAL_ERR,
+				    (LAYER_DDP << 4) | DDP_CATASTROPHIC, DDP_CATASTROPHIC_LOCAL);
+		break;
+	case I40IW_AE_DDP_INVALID_MSN_GAP_IN_MSN:
+	case I40IW_AE_DDP_INVALID_MSN_RANGE_IS_NOT_VALID:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_MSN_RANGE);
+		break;
+	case I40IW_AE_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER:
+		qp->eventtype = TERM_EVENT_QP_ACCESS_ERR;
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_LOC_LEN_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_TOO_LONG);
+		break;
+	case I40IW_AE_DDP_UBE_INVALID_DDP_VERSION:
+		if (is_tagged)
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+					    (LAYER_DDP << 4) | DDP_TAGGED_BUFFER, DDP_TAGGED_INV_DDP_VER);
+		else
+			i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+					    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_DDP_VER);
+		break;
+	case I40IW_AE_DDP_UBE_INVALID_MO:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_MO);
+		break;
+	case I40IW_AE_DDP_UBE_INVALID_MSN_NO_BUFFER_AVAILABLE:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_REM_OP_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_MSN_NO_BUF);
+		break;
+	case I40IW_AE_DDP_UBE_INVALID_QN:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+				    (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER, DDP_UNTAGGED_INV_QN);
+		break;
+	case I40IW_AE_RDMAP_ROE_INVALID_RDMAP_VERSION:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_GENERAL_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_OP, RDMAP_INV_RDMAP_VER);
+		break;
+	case I40IW_AE_RDMAP_ROE_UNEXPECTED_OPCODE:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_LOC_QP_OP_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_OP, RDMAP_UNEXPECTED_OP);
+		break;
+	default:
+		i40iw_setup_termhdr(qp, termhdr, FLUSH_FATAL_ERR,
+				    (LAYER_RDMA << 4) | RDMAP_REMOTE_OP, RDMAP_UNSPECIFIED);
+		break;
+	}
+
+	if (copy_len)
+		memcpy(termhdr + 1, pkt, copy_len);
+
+	if (flush_code && !info->in_rdrsp_wr)
+		qp->sq_flush = (info->sq) ? true : false;
+
+	return sizeof(struct i40iw_terminate_hdr) + copy_len;
+}
+
+/**
+ * i40iw_terminate_send_fin() - Send fin for terminate message
+ * @qp: qp associated with received terminate AE
+ */
+void i40iw_terminate_send_fin(struct i40iw_sc_qp *qp)
+{
+	/* Send the fin only */
+	i40iw_term_modify_qp(qp,
+			     I40IW_QP_STATE_TERMINATE,
+			     I40IWQP_TERM_SEND_FIN_ONLY,
+			     0);
+}
+
+/**
+ * i40iw_terminate_connection() - Bad AE and send terminate to remote QP
+ * @qp: qp associated with received terminate AE
+ * @info: the struct contiaing AE information
+ */
+void i40iw_terminate_connection(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *info)
+{
+	u8 termlen = 0;
+
+	if (qp->term_flags & I40IW_TERM_SENT)
+		return;         /* Sanity check */
+
+	/* Eventtype can change from bld_terminate_hdr */
+	qp->eventtype = TERM_EVENT_QP_FATAL;
+	termlen = i40iw_bld_terminate_hdr(qp, info);
+	i40iw_terminate_start_timer(qp);
+	qp->term_flags |= I40IW_TERM_SENT;
+	i40iw_term_modify_qp(qp, I40IW_QP_STATE_TERMINATE,
+			     I40IWQP_TERM_SEND_TERM_ONLY, termlen);
+}
+
+/**
+ * i40iw_terminate_received - handle terminate received AE
+ * @qp: qp associated with received terminate AE
+ * @info: the struct contiaing AE information
+ */
+void i40iw_terminate_received(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *info)
+{
+	u8 *pkt = qp->q2_buf + Q2_BAD_FRAME_OFFSET;
+	u32 *mpa;
+	u8 ddp_ctl;
+	u8 rdma_ctl;
+	u16 aeq_id = 0;
+	struct i40iw_terminate_hdr *termhdr;
+
+	mpa = (u32 *)i40iw_locate_mpa(pkt);
+	if (info->q2_data_written) {
+		/* did not validate the frame - do it now */
+		ddp_ctl = (ntohl(mpa[0]) >> 8) & 0xff;
+		rdma_ctl = ntohl(mpa[0]) & 0xff;
+		if ((ddp_ctl & 0xc0) != 0x40)
+			aeq_id = I40IW_AE_LCE_QP_CATASTROPHIC;
+		else if ((ddp_ctl & 0x03) != 1)
+			aeq_id = I40IW_AE_DDP_UBE_INVALID_DDP_VERSION;
+		else if (ntohl(mpa[2]) != 2)
+			aeq_id = I40IW_AE_DDP_UBE_INVALID_QN;
+		else if (ntohl(mpa[3]) != 1)
+			aeq_id = I40IW_AE_DDP_INVALID_MSN_GAP_IN_MSN;
+		else if (ntohl(mpa[4]) != 0)
+			aeq_id = I40IW_AE_DDP_UBE_INVALID_MO;
+		else if ((rdma_ctl & 0xc0) != 0x40)
+			aeq_id = I40IW_AE_RDMAP_ROE_INVALID_RDMAP_VERSION;
+
+		info->ae_id = aeq_id;
+		if (info->ae_id) {
+			/* Bad terminate recvd - send back a terminate */
+			i40iw_terminate_connection(qp, info);
+			return;
+		}
+	}
+
+	qp->term_flags |= I40IW_TERM_RCVD;
+	qp->eventtype = TERM_EVENT_QP_FATAL;
+	termhdr = (struct i40iw_terminate_hdr *)&mpa[5];
+	if (termhdr->layer_etype == RDMAP_REMOTE_PROT ||
+	    termhdr->layer_etype == RDMAP_REMOTE_OP) {
+		i40iw_terminate_done(qp, 0);
+	} else {
+		i40iw_terminate_start_timer(qp);
+		i40iw_terminate_send_fin(qp);
+	}
+}
+
+/**
+ * i40iw_hw_stat_init - Initiliaze HW stats table
+ * @devstat: pestat struct
+ * @fcn_idx: PCI fn id
+ * @hw: PF i40iw_hw structure.
+ * @is_pf: Is it a PF?
+ *
+ * Populate the HW stat table with register offset addr for each
+ * stat. And start the perioidic stats timer.
+ */
+static void i40iw_hw_stat_init(struct i40iw_dev_pestat *devstat,
+			       u8 fcn_idx,
+			       struct i40iw_hw *hw, bool is_pf)
+{
+	u32 stat_reg_offset;
+	u32 stat_index;
+	struct i40iw_dev_hw_stat_offsets *stat_table =
+		&devstat->hw_stat_offsets;
+	struct i40iw_dev_hw_stats *last_rd_stats = &devstat->last_read_hw_stats;
+
+	devstat->hw = hw;
+
+	if (is_pf) {
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4RXDISCARD] =
+				I40E_GLPES_PFIP4RXDISCARD(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4RXTRUNC] =
+				I40E_GLPES_PFIP4RXTRUNC(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4TXNOROUTE] =
+				I40E_GLPES_PFIP4TXNOROUTE(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6RXDISCARD] =
+				I40E_GLPES_PFIP6RXDISCARD(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6RXTRUNC] =
+				I40E_GLPES_PFIP6RXTRUNC(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6TXNOROUTE] =
+				I40E_GLPES_PFIP6TXNOROUTE(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRTXSEG] =
+				I40E_GLPES_PFTCPRTXSEG(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRXOPTERR] =
+				I40E_GLPES_PFTCPRXOPTERR(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRXPROTOERR] =
+				I40E_GLPES_PFTCPRXPROTOERR(fcn_idx);
+
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXOCTS] =
+				I40E_GLPES_PFIP4RXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXPKTS] =
+				I40E_GLPES_PFIP4RXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXFRAGS] =
+				I40E_GLPES_PFIP4RXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXMCPKTS] =
+				I40E_GLPES_PFIP4RXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXOCTS] =
+				I40E_GLPES_PFIP4TXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXPKTS] =
+				I40E_GLPES_PFIP4TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXFRAGS] =
+				I40E_GLPES_PFIP4TXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXMCPKTS] =
+				I40E_GLPES_PFIP4TXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXOCTS] =
+				I40E_GLPES_PFIP6RXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXPKTS] =
+				I40E_GLPES_PFIP6RXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXFRAGS] =
+				I40E_GLPES_PFIP6RXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXMCPKTS] =
+				I40E_GLPES_PFIP6RXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXOCTS] =
+				I40E_GLPES_PFIP6TXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXPKTS] =
+				I40E_GLPES_PFIP6TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXPKTS] =
+				I40E_GLPES_PFIP6TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXFRAGS] =
+				I40E_GLPES_PFIP6TXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_TCPRXSEGS] =
+				I40E_GLPES_PFTCPRXSEGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_TCPTXSEG] =
+				I40E_GLPES_PFTCPTXSEGLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXRDS] =
+				I40E_GLPES_PFRDMARXRDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXSNDS] =
+				I40E_GLPES_PFRDMARXSNDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXWRS] =
+				I40E_GLPES_PFRDMARXWRSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXRDS] =
+				I40E_GLPES_PFRDMATXRDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXSNDS] =
+				I40E_GLPES_PFRDMATXSNDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXWRS] =
+				I40E_GLPES_PFRDMATXWRSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMAVBND] =
+				I40E_GLPES_PFRDMAVBNDLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMAVINV] =
+				I40E_GLPES_PFRDMAVINVLO(fcn_idx);
+	} else {
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4RXDISCARD] =
+				I40E_GLPES_VFIP4RXDISCARD(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4RXTRUNC] =
+				I40E_GLPES_VFIP4RXTRUNC(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP4TXNOROUTE] =
+				I40E_GLPES_VFIP4TXNOROUTE(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6RXDISCARD] =
+				I40E_GLPES_VFIP6RXDISCARD(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6RXTRUNC] =
+				I40E_GLPES_VFIP6RXTRUNC(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_IP6TXNOROUTE] =
+				I40E_GLPES_VFIP6TXNOROUTE(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRTXSEG] =
+				I40E_GLPES_VFTCPRTXSEG(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRXOPTERR] =
+				I40E_GLPES_VFTCPRXOPTERR(fcn_idx);
+		stat_table->stat_offset_32[I40IW_HW_STAT_INDEX_TCPRXPROTOERR] =
+				I40E_GLPES_VFTCPRXPROTOERR(fcn_idx);
+
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXOCTS] =
+				I40E_GLPES_VFIP4RXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXPKTS] =
+				I40E_GLPES_VFIP4RXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXFRAGS] =
+				I40E_GLPES_VFIP4RXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4RXMCPKTS] =
+				I40E_GLPES_VFIP4RXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXOCTS] =
+				I40E_GLPES_VFIP4TXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXPKTS] =
+				I40E_GLPES_VFIP4TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXFRAGS] =
+				I40E_GLPES_VFIP4TXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP4TXMCPKTS] =
+				I40E_GLPES_VFIP4TXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXOCTS] =
+				I40E_GLPES_VFIP6RXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXPKTS] =
+				I40E_GLPES_VFIP6RXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXFRAGS] =
+				I40E_GLPES_VFIP6RXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6RXMCPKTS] =
+				I40E_GLPES_VFIP6RXMCPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXOCTS] =
+				I40E_GLPES_VFIP6TXOCTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXPKTS] =
+				I40E_GLPES_VFIP6TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXPKTS] =
+				I40E_GLPES_VFIP6TXPKTSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_IP6TXFRAGS] =
+				I40E_GLPES_VFIP6TXFRAGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_TCPRXSEGS] =
+				I40E_GLPES_VFTCPRXSEGSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_TCPTXSEG] =
+				I40E_GLPES_VFTCPTXSEGLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXRDS] =
+				I40E_GLPES_VFRDMARXRDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXSNDS] =
+				I40E_GLPES_VFRDMARXSNDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMARXWRS] =
+				I40E_GLPES_VFRDMARXWRSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXRDS] =
+				I40E_GLPES_VFRDMATXRDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXSNDS] =
+				I40E_GLPES_VFRDMATXSNDSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMATXWRS] =
+				I40E_GLPES_VFRDMATXWRSLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMAVBND] =
+				I40E_GLPES_VFRDMAVBNDLO(fcn_idx);
+		stat_table->stat_offset_64[I40IW_HW_STAT_INDEX_RDMAVINV] =
+				I40E_GLPES_VFRDMAVINVLO(fcn_idx);
+	}
+
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_64;
+	     stat_index++) {
+		stat_reg_offset = stat_table->stat_offset_64[stat_index];
+		last_rd_stats->stat_value_64[stat_index] =
+			readq(devstat->hw->hw_addr + stat_reg_offset);
+	}
+
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_32;
+	     stat_index++) {
+		stat_reg_offset = stat_table->stat_offset_32[stat_index];
+		last_rd_stats->stat_value_32[stat_index] =
+			i40iw_rd32(devstat->hw, stat_reg_offset);
+	}
+}
+
+/**
+ * i40iw_hw_stat_read_32 - Read 32-bit HW stat counters and accommodates for roll-overs.
+ * @devstat: pestat struct
+ * @index: index in HW stat table which contains offset reg-addr
+ * @value: hw stat value
+ */
+static void i40iw_hw_stat_read_32(struct i40iw_dev_pestat *devstat,
+				  enum i40iw_hw_stat_index_32b index,
+				  u64 *value)
+{
+	struct i40iw_dev_hw_stat_offsets *stat_table =
+		&devstat->hw_stat_offsets;
+	struct i40iw_dev_hw_stats *last_rd_stats = &devstat->last_read_hw_stats;
+	struct i40iw_dev_hw_stats *hw_stats = &devstat->hw_stats;
+	u64 new_stat_value = 0;
+	u32 stat_reg_offset = stat_table->stat_offset_32[index];
+
+	new_stat_value = i40iw_rd32(devstat->hw, stat_reg_offset);
+	/*roll-over case */
+	if (new_stat_value < last_rd_stats->stat_value_32[index])
+		hw_stats->stat_value_32[index] += new_stat_value;
+	else
+		hw_stats->stat_value_32[index] +=
+			new_stat_value - last_rd_stats->stat_value_32[index];
+	last_rd_stats->stat_value_32[index] = new_stat_value;
+	*value = hw_stats->stat_value_32[index];
+}
+
+/**
+ * i40iw_hw_stat_read_64 - Read HW stat counters (greater than 32-bit) and accommodates for roll-overs.
+ * @devstat: pestat struct
+ * @index: index in HW stat table which contains offset reg-addr
+ * @value: hw stat value
+ */
+static void i40iw_hw_stat_read_64(struct i40iw_dev_pestat *devstat,
+				  enum i40iw_hw_stat_index_64b index,
+				  u64 *value)
+{
+	struct i40iw_dev_hw_stat_offsets *stat_table =
+		&devstat->hw_stat_offsets;
+	struct i40iw_dev_hw_stats *last_rd_stats = &devstat->last_read_hw_stats;
+	struct i40iw_dev_hw_stats *hw_stats = &devstat->hw_stats;
+	u64 new_stat_value = 0;
+	u32 stat_reg_offset = stat_table->stat_offset_64[index];
+
+	new_stat_value = readq(devstat->hw->hw_addr + stat_reg_offset);
+	/*roll-over case */
+	if (new_stat_value < last_rd_stats->stat_value_64[index])
+		hw_stats->stat_value_64[index] += new_stat_value;
+	else
+		hw_stats->stat_value_64[index] +=
+			new_stat_value - last_rd_stats->stat_value_64[index];
+	last_rd_stats->stat_value_64[index] = new_stat_value;
+	*value = hw_stats->stat_value_64[index];
+}
+
+/**
+ * i40iw_hw_stat_read_all - read all HW stat counters
+ * @devstat: pestat struct
+ * @stat_values: hw stats structure
+ *
+ * Read all the HW stat counters and populates hw_stats structure
+ * of passed-in dev's pestat as well as copy created in stat_values.
+ */
+static void i40iw_hw_stat_read_all(struct i40iw_dev_pestat *devstat,
+				   struct i40iw_dev_hw_stats *stat_values)
+{
+	u32 stat_index;
+
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_32;
+	     stat_index++)
+		i40iw_hw_stat_read_32(devstat, stat_index,
+				      &stat_values->stat_value_32[stat_index]);
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_64;
+	     stat_index++)
+		i40iw_hw_stat_read_64(devstat, stat_index,
+				      &stat_values->stat_value_64[stat_index]);
+}
+
+/**
+ * i40iw_hw_stat_refresh_all - Update all HW stat structs
+ * @devstat: pestat struct
+ * @stat_values: hw stats structure
+ *
+ * Read all the HW stat counters to refresh values in hw_stats structure
+ * of passed-in dev's pestat
+ */
+static void i40iw_hw_stat_refresh_all(struct i40iw_dev_pestat *devstat)
+{
+	u64 stat_value;
+	u32 stat_index;
+
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_32;
+	     stat_index++)
+		i40iw_hw_stat_read_32(devstat, stat_index, &stat_value);
+	for (stat_index = 0; stat_index < I40IW_HW_STAT_INDEX_MAX_64;
+	     stat_index++)
+		i40iw_hw_stat_read_64(devstat, stat_index, &stat_value);
+}
+
+static struct i40iw_cqp_ops iw_cqp_ops = {
+	i40iw_sc_cqp_init,
+	i40iw_sc_cqp_create,
+	i40iw_sc_cqp_post_sq,
+	i40iw_sc_cqp_get_next_send_wqe,
+	i40iw_sc_cqp_destroy,
+	i40iw_sc_poll_for_cqp_op_done
+};
+
+static struct i40iw_ccq_ops iw_ccq_ops = {
+	i40iw_sc_ccq_init,
+	i40iw_sc_ccq_create,
+	i40iw_sc_ccq_destroy,
+	i40iw_sc_ccq_create_done,
+	i40iw_sc_ccq_get_cqe_info,
+	i40iw_sc_ccq_arm
+};
+
+static struct i40iw_ceq_ops iw_ceq_ops = {
+	i40iw_sc_ceq_init,
+	i40iw_sc_ceq_create,
+	i40iw_sc_cceq_create_done,
+	i40iw_sc_cceq_destroy_done,
+	i40iw_sc_cceq_create,
+	i40iw_sc_ceq_destroy,
+	i40iw_sc_process_ceq
+};
+
+static struct i40iw_aeq_ops iw_aeq_ops = {
+	i40iw_sc_aeq_init,
+	i40iw_sc_aeq_create,
+	i40iw_sc_aeq_destroy,
+	i40iw_sc_get_next_aeqe,
+	i40iw_sc_repost_aeq_entries,
+	i40iw_sc_aeq_create_done,
+	i40iw_sc_aeq_destroy_done
+};
+
+/* iwarp pd ops */
+static struct i40iw_pd_ops iw_pd_ops = {
+	i40iw_sc_pd_init,
+};
+
+static struct i40iw_priv_qp_ops iw_priv_qp_ops = {
+	i40iw_sc_qp_init,
+	i40iw_sc_qp_create,
+	i40iw_sc_qp_modify,
+	i40iw_sc_qp_destroy,
+	i40iw_sc_qp_flush_wqes,
+	i40iw_sc_qp_upload_context,
+	i40iw_sc_qp_setctx,
+	i40iw_sc_send_lsmm,
+	i40iw_sc_send_lsmm_nostag,
+	i40iw_sc_send_rtt,
+	i40iw_sc_post_wqe0,
+};
+
+static struct i40iw_priv_cq_ops iw_priv_cq_ops = {
+	i40iw_sc_cq_init,
+	i40iw_sc_cq_create,
+	i40iw_sc_cq_destroy,
+	i40iw_sc_cq_modify,
+};
+
+static struct i40iw_mr_ops iw_mr_ops = {
+	i40iw_sc_alloc_stag,
+	i40iw_sc_mr_reg_non_shared,
+	i40iw_sc_mr_reg_shared,
+	i40iw_sc_dealloc_stag,
+	i40iw_sc_query_stag,
+	i40iw_sc_mw_alloc
+};
+
+static struct i40iw_cqp_misc_ops iw_cqp_misc_ops = {
+	i40iw_sc_manage_push_page,
+	i40iw_sc_manage_hmc_pm_func_table,
+	i40iw_sc_set_hmc_resource_profile,
+	i40iw_sc_commit_fpm_values,
+	i40iw_sc_query_fpm_values,
+	i40iw_sc_static_hmc_pages_allocated,
+	i40iw_sc_add_arp_cache_entry,
+	i40iw_sc_del_arp_cache_entry,
+	i40iw_sc_query_arp_cache_entry,
+	i40iw_sc_manage_apbvt_entry,
+	i40iw_sc_manage_qhash_table_entry,
+	i40iw_sc_alloc_local_mac_ipaddr_entry,
+	i40iw_sc_add_local_mac_ipaddr_entry,
+	i40iw_sc_del_local_mac_ipaddr_entry,
+	i40iw_sc_cqp_nop,
+	i40iw_sc_commit_fpm_values_done,
+	i40iw_sc_query_fpm_values_done,
+	i40iw_sc_manage_hmc_pm_func_table_done,
+	i40iw_sc_suspend_qp,
+	i40iw_sc_resume_qp
+};
+
+static struct i40iw_hmc_ops iw_hmc_ops = {
+	i40iw_sc_init_iw_hmc,
+	i40iw_sc_parse_fpm_query_buf,
+	i40iw_sc_configure_iw_fpm,
+	i40iw_sc_parse_fpm_commit_buf,
+	i40iw_sc_create_hmc_obj,
+	i40iw_sc_del_hmc_obj,
+	NULL,
+	NULL
+};
+
+static const struct i40iw_device_pestat_ops iw_device_pestat_ops = {
+	i40iw_hw_stat_init,
+	i40iw_hw_stat_read_32,
+	i40iw_hw_stat_read_64,
+	i40iw_hw_stat_read_all,
+	i40iw_hw_stat_refresh_all
+};
+
+/**
+ * i40iw_device_init_pestat - Initialize the pestat structure
+ * @dev: pestat struct
+ */
+enum i40iw_status_code i40iw_device_init_pestat(struct i40iw_dev_pestat *devstat)
+{
+	devstat->ops = iw_device_pestat_ops;
+	return 0;
+}
+
+/**
+ * i40iw_device_init - Initialize IWARP device
+ * @dev: IWARP device pointer
+ * @info: IWARP init info
+ */
+enum i40iw_status_code i40iw_device_init(struct i40iw_sc_dev *dev,
+					 struct i40iw_device_init_info *info)
+{
+	u32 val;
+	u32 vchnl_ver = 0;
+	u16 hmc_fcn = 0;
+	enum i40iw_status_code ret_code = 0;
+	u8 db_size;
+
+	spin_lock_init(&dev->cqp_lock);
+	INIT_LIST_HEAD(&dev->cqp_cmd_head);             /* for the cqp commands backlog. */
+
+	i40iw_device_init_uk(&dev->dev_uk);
+
+	dev->debug_mask = info->debug_mask;
+
+	ret_code = i40iw_device_init_pestat(&dev->dev_pestat);
+	if (ret_code) {
+		i40iw_debug(dev, I40IW_DEBUG_DEV,
+			    "%s: i40iw_device_init_pestat failed\n", __func__);
+		return ret_code;
+	}
+	dev->hmc_fn_id = info->hmc_fn_id;
+	dev->qs_handle = info->qs_handle;
+	dev->exception_lan_queue = info->exception_lan_queue;
+	dev->is_pf = info->is_pf;
+
+	dev->fpm_query_buf_pa = info->fpm_query_buf_pa;
+	dev->fpm_query_buf = info->fpm_query_buf;
+
+	dev->fpm_commit_buf_pa = info->fpm_commit_buf_pa;
+	dev->fpm_commit_buf = info->fpm_commit_buf;
+
+	dev->hw = info->hw;
+	dev->hw->hw_addr = info->bar0;
+
+	val = i40iw_rd32(dev->hw, I40E_GLPCI_DREVID);
+	dev->hw_rev = (u8)RS_32(val, I40E_GLPCI_DREVID_DEFAULT_REVID);
+
+	if (dev->is_pf) {
+		dev->dev_pestat.ops.iw_hw_stat_init(&dev->dev_pestat,
+			dev->hmc_fn_id, dev->hw, true);
+		spin_lock_init(&dev->dev_pestat.stats_lock);
+		/*start the periodic stats_timer */
+		i40iw_hw_stats_start_timer(dev);
+		val = i40iw_rd32(dev->hw, I40E_GLPCI_LBARCTRL);
+		db_size = (u8)RS_32(val, I40E_GLPCI_LBARCTRL_PE_DB_SIZE);
+		if ((db_size != I40IW_PE_DB_SIZE_4M) &&
+		    (db_size != I40IW_PE_DB_SIZE_8M)) {
+			i40iw_debug(dev, I40IW_DEBUG_DEV,
+				    "%s: PE doorbell is not enabled in CSR val 0x%x\n",
+				    __func__, val);
+			ret_code = I40IW_ERR_PE_DOORBELL_NOT_ENABLED;
+			return ret_code;
+		}
+		dev->db_addr = dev->hw->hw_addr + I40IW_DB_ADDR_OFFSET;
+		dev->vchnl_if.vchnl_recv = i40iw_vchnl_recv_pf;
+	} else {
+		dev->db_addr = dev->hw->hw_addr + I40IW_VF_DB_ADDR_OFFSET;
+	}
+
+	dev->cqp_ops = &iw_cqp_ops;
+	dev->ccq_ops = &iw_ccq_ops;
+	dev->ceq_ops = &iw_ceq_ops;
+	dev->aeq_ops = &iw_aeq_ops;
+	dev->cqp_misc_ops = &iw_cqp_misc_ops;
+	dev->iw_pd_ops = &iw_pd_ops;
+	dev->iw_priv_qp_ops = &iw_priv_qp_ops;
+	dev->iw_priv_cq_ops = &iw_priv_cq_ops;
+	dev->mr_ops = &iw_mr_ops;
+	dev->hmc_ops = &iw_hmc_ops;
+	dev->vchnl_if.vchnl_send = info->vchnl_send;
+	if (dev->vchnl_if.vchnl_send)
+		dev->vchnl_up = true;
+	else
+		dev->vchnl_up = false;
+	if (!dev->is_pf) {
+		dev->vchnl_if.vchnl_recv = i40iw_vchnl_recv_vf;
+		ret_code = i40iw_vchnl_vf_get_ver(dev, &vchnl_ver);
+		if (!ret_code) {
+			i40iw_debug(dev, I40IW_DEBUG_DEV,
+				    "%s: Get Channel version rc = 0x%0x, version is %u\n",
+				__func__, ret_code, vchnl_ver);
+			ret_code = i40iw_vchnl_vf_get_hmc_fcn(dev, &hmc_fcn);
+			if (!ret_code) {
+				i40iw_debug(dev, I40IW_DEBUG_DEV,
+					    "%s Get HMC function rc = 0x%0x, hmc fcn is %u\n",
+					    __func__, ret_code, hmc_fcn);
+				dev->hmc_fn_id = (u8)hmc_fcn;
+			}
+		}
+	}
+	dev->iw_vf_cqp_ops = &iw_vf_cqp_ops;
+
+	return ret_code;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_d.h b/drivers/infiniband/hw/i40iw/i40iw_d.h
new file mode 100644
index 000000000000..aab88d65f805
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_d.h
@@ -0,0 +1,1713 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_D_H
+#define I40IW_D_H
+
+#define I40IW_DB_ADDR_OFFSET    (4 * 1024 * 1024 - 64 * 1024)
+#define I40IW_VF_DB_ADDR_OFFSET (64 * 1024)
+
+#define I40IW_PUSH_OFFSET       (4 * 1024 * 1024)
+#define I40IW_PF_FIRST_PUSH_PAGE_INDEX 16
+#define I40IW_VF_PUSH_OFFSET    ((8 + 64) * 1024)
+#define I40IW_VF_FIRST_PUSH_PAGE_INDEX 2
+
+#define I40IW_PE_DB_SIZE_4M     1
+#define I40IW_PE_DB_SIZE_8M     2
+
+#define I40IW_DDP_VER 1
+#define I40IW_RDMAP_VER 1
+
+#define I40IW_RDMA_MODE_RDMAC 0
+#define I40IW_RDMA_MODE_IETF  1
+
+#define I40IW_QP_STATE_INVALID 0
+#define I40IW_QP_STATE_IDLE 1
+#define I40IW_QP_STATE_RTS 2
+#define I40IW_QP_STATE_CLOSING 3
+#define I40IW_QP_STATE_RESERVED 4
+#define I40IW_QP_STATE_TERMINATE 5
+#define I40IW_QP_STATE_ERROR 6
+
+#define I40IW_STAG_STATE_INVALID 0
+#define I40IW_STAG_STATE_VALID 1
+
+#define I40IW_STAG_TYPE_SHARED 0
+#define I40IW_STAG_TYPE_NONSHARED 1
+
+#define I40IW_MAX_USER_PRIORITY 8
+
+#define LS_64_1(val, bits)      ((u64)(uintptr_t)val << bits)
+#define RS_64_1(val, bits)      ((u64)(uintptr_t)val >> bits)
+#define LS_32_1(val, bits)      (u32)(val << bits)
+#define RS_32_1(val, bits)      (u32)(val >> bits)
+#define I40E_HI_DWORD(x)        ((u32)((((x) >> 16) >> 16) & 0xFFFFFFFF))
+
+#define LS_64(val, field) (((u64)val << field ## _SHIFT) & (field ## _MASK))
+
+#define RS_64(val, field) ((u64)(val & field ## _MASK) >> field ## _SHIFT)
+#define LS_32(val, field) ((val << field ## _SHIFT) & (field ## _MASK))
+#define RS_32(val, field) ((val & field ## _MASK) >> field ## _SHIFT)
+
+#define TERM_DDP_LEN_TAGGED     14
+#define TERM_DDP_LEN_UNTAGGED   18
+#define TERM_RDMA_LEN           28
+#define RDMA_OPCODE_MASK        0x0f
+#define RDMA_READ_REQ_OPCODE    1
+#define Q2_BAD_FRAME_OFFSET     72
+#define CQE_MAJOR_DRV           0x8000
+
+#define I40IW_TERM_SENT 0x01
+#define I40IW_TERM_RCVD 0x02
+#define I40IW_TERM_DONE 0x04
+#define I40IW_MAC_HLEN  14
+
+#define I40IW_INVALID_WQE_INDEX 0xffffffff
+
+#define I40IW_CQP_WAIT_POLL_REGS 1
+#define I40IW_CQP_WAIT_POLL_CQ 2
+#define I40IW_CQP_WAIT_EVENT 3
+
+#define I40IW_CQP_INIT_WQE(wqe) memset(wqe, 0, 64)
+
+#define I40IW_GET_CURRENT_CQ_ELEMENT(_cq) \
+	( \
+		&((_cq)->cq_base[I40IW_RING_GETCURRENT_HEAD((_cq)->cq_ring)])  \
+	)
+#define I40IW_GET_CURRENT_EXTENDED_CQ_ELEMENT(_cq) \
+	( \
+		&(((struct i40iw_extended_cqe *)        \
+		   ((_cq)->cq_base))[I40IW_RING_GETCURRENT_HEAD((_cq)->cq_ring)]) \
+	)
+
+#define I40IW_GET_CURRENT_AEQ_ELEMENT(_aeq) \
+	( \
+		&_aeq->aeqe_base[I40IW_RING_GETCURRENT_TAIL(_aeq->aeq_ring)]   \
+	)
+
+#define I40IW_GET_CURRENT_CEQ_ELEMENT(_ceq) \
+	( \
+		&_ceq->ceqe_base[I40IW_RING_GETCURRENT_TAIL(_ceq->ceq_ring)]   \
+	)
+
+#define I40IW_AE_SOURCE_RQ              0x1
+#define I40IW_AE_SOURCE_RQ_0011         0x3
+
+#define I40IW_AE_SOURCE_CQ              0x2
+#define I40IW_AE_SOURCE_CQ_0110         0x6
+#define I40IW_AE_SOURCE_CQ_1010         0xA
+#define I40IW_AE_SOURCE_CQ_1110         0xE
+
+#define I40IW_AE_SOURCE_SQ              0x5
+#define I40IW_AE_SOURCE_SQ_0111         0x7
+
+#define I40IW_AE_SOURCE_IN_RR_WR        0x9
+#define I40IW_AE_SOURCE_IN_RR_WR_1011   0xB
+#define I40IW_AE_SOURCE_OUT_RR          0xD
+#define I40IW_AE_SOURCE_OUT_RR_1111     0xF
+
+#define I40IW_TCP_STATE_NON_EXISTENT 0
+#define I40IW_TCP_STATE_CLOSED 1
+#define I40IW_TCP_STATE_LISTEN 2
+#define I40IW_STATE_SYN_SEND 3
+#define I40IW_TCP_STATE_SYN_RECEIVED 4
+#define I40IW_TCP_STATE_ESTABLISHED 5
+#define I40IW_TCP_STATE_CLOSE_WAIT 6
+#define I40IW_TCP_STATE_FIN_WAIT_1 7
+#define I40IW_TCP_STATE_CLOSING  8
+#define I40IW_TCP_STATE_LAST_ACK 9
+#define I40IW_TCP_STATE_FIN_WAIT_2 10
+#define I40IW_TCP_STATE_TIME_WAIT 11
+#define I40IW_TCP_STATE_RESERVED_1 12
+#define I40IW_TCP_STATE_RESERVED_2 13
+#define I40IW_TCP_STATE_RESERVED_3 14
+#define I40IW_TCP_STATE_RESERVED_4 15
+
+/* ILQ CQP hash table fields */
+#define I40IW_CQPSQ_QHASH_VLANID_SHIFT 32
+#define I40IW_CQPSQ_QHASH_VLANID_MASK \
+	((u64)0xfff << I40IW_CQPSQ_QHASH_VLANID_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_QPN_SHIFT 32
+#define I40IW_CQPSQ_QHASH_QPN_MASK \
+	((u64)0x3ffff << I40IW_CQPSQ_QHASH_QPN_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_QS_HANDLE_SHIFT 0
+#define I40IW_CQPSQ_QHASH_QS_HANDLE_MASK ((u64)0x3ff << I40IW_CQPSQ_QHASH_QS_HANDLE_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_SRC_PORT_SHIFT 16
+#define I40IW_CQPSQ_QHASH_SRC_PORT_MASK \
+	((u64)0xffff << I40IW_CQPSQ_QHASH_SRC_PORT_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_DEST_PORT_SHIFT 0
+#define I40IW_CQPSQ_QHASH_DEST_PORT_MASK \
+	((u64)0xffff << I40IW_CQPSQ_QHASH_DEST_PORT_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_ADDR0_SHIFT 32
+#define I40IW_CQPSQ_QHASH_ADDR0_MASK \
+	((u64)0xffffffff << I40IW_CQPSQ_QHASH_ADDR0_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_ADDR1_SHIFT 0
+#define I40IW_CQPSQ_QHASH_ADDR1_MASK \
+	((u64)0xffffffff << I40IW_CQPSQ_QHASH_ADDR1_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_ADDR2_SHIFT 32
+#define I40IW_CQPSQ_QHASH_ADDR2_MASK \
+	((u64)0xffffffff << I40IW_CQPSQ_QHASH_ADDR2_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_ADDR3_SHIFT 0
+#define I40IW_CQPSQ_QHASH_ADDR3_MASK \
+	((u64)0xffffffff << I40IW_CQPSQ_QHASH_ADDR3_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_WQEVALID_SHIFT 63
+#define I40IW_CQPSQ_QHASH_WQEVALID_MASK \
+	((u64)0x1 << I40IW_CQPSQ_QHASH_WQEVALID_SHIFT)
+#define I40IW_CQPSQ_QHASH_OPCODE_SHIFT 32
+#define I40IW_CQPSQ_QHASH_OPCODE_MASK \
+	((u64)0x3f << I40IW_CQPSQ_QHASH_OPCODE_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_MANAGE_SHIFT 61
+#define I40IW_CQPSQ_QHASH_MANAGE_MASK \
+	((u64)0x3 << I40IW_CQPSQ_QHASH_MANAGE_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_IPV4VALID_SHIFT 60
+#define I40IW_CQPSQ_QHASH_IPV4VALID_MASK \
+	((u64)0x1 << I40IW_CQPSQ_QHASH_IPV4VALID_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_VLANVALID_SHIFT 59
+#define I40IW_CQPSQ_QHASH_VLANVALID_MASK \
+	((u64)0x1 << I40IW_CQPSQ_QHASH_VLANVALID_SHIFT)
+
+#define I40IW_CQPSQ_QHASH_ENTRYTYPE_SHIFT 42
+#define I40IW_CQPSQ_QHASH_ENTRYTYPE_MASK \
+	((u64)0x7 << I40IW_CQPSQ_QHASH_ENTRYTYPE_SHIFT)
+/* CQP Host Context */
+#define I40IW_CQPHC_EN_DC_TCP_SHIFT 0
+#define I40IW_CQPHC_EN_DC_TCP_MASK (1UL << I40IW_CQPHC_EN_DC_TCP_SHIFT)
+
+#define I40IW_CQPHC_SQSIZE_SHIFT 8
+#define I40IW_CQPHC_SQSIZE_MASK (0xfUL << I40IW_CQPHC_SQSIZE_SHIFT)
+
+#define I40IW_CQPHC_DISABLE_PFPDUS_SHIFT 1
+#define I40IW_CQPHC_DISABLE_PFPDUS_MASK (0x1UL << I40IW_CQPHC_DISABLE_PFPDUS_SHIFT)
+
+#define I40IW_CQPHC_ENABLED_VFS_SHIFT 32
+#define I40IW_CQPHC_ENABLED_VFS_MASK (0x3fULL << I40IW_CQPHC_ENABLED_VFS_SHIFT)
+
+#define I40IW_CQPHC_HMC_PROFILE_SHIFT 0
+#define I40IW_CQPHC_HMC_PROFILE_MASK (0x7ULL << I40IW_CQPHC_HMC_PROFILE_SHIFT)
+
+#define I40IW_CQPHC_SVER_SHIFT 24
+#define I40IW_CQPHC_SVER_MASK (0xffUL << I40IW_CQPHC_SVER_SHIFT)
+
+#define I40IW_CQPHC_SQBASE_SHIFT 9
+#define I40IW_CQPHC_SQBASE_MASK \
+	(0xfffffffffffffeULL << I40IW_CQPHC_SQBASE_SHIFT)
+
+#define I40IW_CQPHC_QPCTX_SHIFT 0
+#define I40IW_CQPHC_QPCTX_MASK  \
+	(0xffffffffffffffffULL << I40IW_CQPHC_QPCTX_SHIFT)
+#define I40IW_CQPHC_SVER        1
+
+#define I40IW_CQP_SW_SQSIZE_4 4
+#define I40IW_CQP_SW_SQSIZE_2048 2048
+
+/* iWARP QP Doorbell shadow area */
+#define I40IW_QP_DBSA_HW_SQ_TAIL_SHIFT 0
+#define I40IW_QP_DBSA_HW_SQ_TAIL_MASK \
+	(0x3fffUL << I40IW_QP_DBSA_HW_SQ_TAIL_SHIFT)
+
+/* Completion Queue Doorbell shadow area */
+#define I40IW_CQ_DBSA_CQEIDX_SHIFT 0
+#define I40IW_CQ_DBSA_CQEIDX_MASK (0xfffffUL << I40IW_CQ_DBSA_CQEIDX_SHIFT)
+
+#define I40IW_CQ_DBSA_SW_CQ_SELECT_SHIFT 0
+#define I40IW_CQ_DBSA_SW_CQ_SELECT_MASK \
+	(0x3fffUL << I40IW_CQ_DBSA_SW_CQ_SELECT_SHIFT)
+
+#define I40IW_CQ_DBSA_ARM_NEXT_SHIFT 14
+#define I40IW_CQ_DBSA_ARM_NEXT_MASK (1UL << I40IW_CQ_DBSA_ARM_NEXT_SHIFT)
+
+#define I40IW_CQ_DBSA_ARM_NEXT_SE_SHIFT 15
+#define I40IW_CQ_DBSA_ARM_NEXT_SE_MASK (1UL << I40IW_CQ_DBSA_ARM_NEXT_SE_SHIFT)
+
+#define I40IW_CQ_DBSA_ARM_SEQ_NUM_SHIFT 16
+#define I40IW_CQ_DBSA_ARM_SEQ_NUM_MASK \
+	(0x3UL << I40IW_CQ_DBSA_ARM_SEQ_NUM_SHIFT)
+
+/* CQP and iWARP Completion Queue */
+#define I40IW_CQ_QPCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQ_QPCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CCQ_OPRETVAL_SHIFT 0
+#define I40IW_CCQ_OPRETVAL_MASK (0xffffffffUL << I40IW_CCQ_OPRETVAL_SHIFT)
+
+#define I40IW_CQ_MINERR_SHIFT 0
+#define I40IW_CQ_MINERR_MASK (0xffffUL << I40IW_CQ_MINERR_SHIFT)
+
+#define I40IW_CQ_MAJERR_SHIFT 16
+#define I40IW_CQ_MAJERR_MASK (0xffffUL << I40IW_CQ_MAJERR_SHIFT)
+
+#define I40IW_CQ_WQEIDX_SHIFT 32
+#define I40IW_CQ_WQEIDX_MASK (0x3fffULL << I40IW_CQ_WQEIDX_SHIFT)
+
+#define I40IW_CQ_ERROR_SHIFT 55
+#define I40IW_CQ_ERROR_MASK (1ULL << I40IW_CQ_ERROR_SHIFT)
+
+#define I40IW_CQ_SQ_SHIFT 62
+#define I40IW_CQ_SQ_MASK (1ULL << I40IW_CQ_SQ_SHIFT)
+
+#define I40IW_CQ_VALID_SHIFT 63
+#define I40IW_CQ_VALID_MASK (1ULL << I40IW_CQ_VALID_SHIFT)
+
+#define I40IWCQ_PAYLDLEN_SHIFT 0
+#define I40IWCQ_PAYLDLEN_MASK (0xffffffffUL << I40IWCQ_PAYLDLEN_SHIFT)
+
+#define I40IWCQ_TCPSEQNUM_SHIFT 32
+#define I40IWCQ_TCPSEQNUM_MASK (0xffffffffULL << I40IWCQ_TCPSEQNUM_SHIFT)
+
+#define I40IWCQ_INVSTAG_SHIFT 0
+#define I40IWCQ_INVSTAG_MASK (0xffffffffUL << I40IWCQ_INVSTAG_SHIFT)
+
+#define I40IWCQ_QPID_SHIFT 32
+#define I40IWCQ_QPID_MASK (0x3ffffULL << I40IWCQ_QPID_SHIFT)
+
+#define I40IWCQ_PSHDROP_SHIFT 51
+#define I40IWCQ_PSHDROP_MASK (1ULL << I40IWCQ_PSHDROP_SHIFT)
+
+#define I40IWCQ_SRQ_SHIFT 52
+#define I40IWCQ_SRQ_MASK (1ULL << I40IWCQ_SRQ_SHIFT)
+
+#define I40IWCQ_STAG_SHIFT 53
+#define I40IWCQ_STAG_MASK (1ULL << I40IWCQ_STAG_SHIFT)
+
+#define I40IWCQ_SOEVENT_SHIFT 54
+#define I40IWCQ_SOEVENT_MASK (1ULL << I40IWCQ_SOEVENT_SHIFT)
+
+#define I40IWCQ_OP_SHIFT 56
+#define I40IWCQ_OP_MASK (0x3fULL << I40IWCQ_OP_SHIFT)
+
+/* CEQE format */
+#define I40IW_CEQE_CQCTX_SHIFT 0
+#define I40IW_CEQE_CQCTX_MASK   \
+	(0x7fffffffffffffffULL << I40IW_CEQE_CQCTX_SHIFT)
+
+#define I40IW_CEQE_VALID_SHIFT 63
+#define I40IW_CEQE_VALID_MASK (1ULL << I40IW_CEQE_VALID_SHIFT)
+
+/* AEQE format */
+#define I40IW_AEQE_COMPCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_AEQE_COMPCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_AEQE_QPCQID_SHIFT 0
+#define I40IW_AEQE_QPCQID_MASK (0x3ffffUL << I40IW_AEQE_QPCQID_SHIFT)
+
+#define I40IW_AEQE_WQDESCIDX_SHIFT 18
+#define I40IW_AEQE_WQDESCIDX_MASK (0x3fffULL << I40IW_AEQE_WQDESCIDX_SHIFT)
+
+#define I40IW_AEQE_OVERFLOW_SHIFT 33
+#define I40IW_AEQE_OVERFLOW_MASK (1ULL << I40IW_AEQE_OVERFLOW_SHIFT)
+
+#define I40IW_AEQE_AECODE_SHIFT 34
+#define I40IW_AEQE_AECODE_MASK (0xffffULL << I40IW_AEQE_AECODE_SHIFT)
+
+#define I40IW_AEQE_AESRC_SHIFT 50
+#define I40IW_AEQE_AESRC_MASK (0xfULL << I40IW_AEQE_AESRC_SHIFT)
+
+#define I40IW_AEQE_IWSTATE_SHIFT 54
+#define I40IW_AEQE_IWSTATE_MASK (0x7ULL << I40IW_AEQE_IWSTATE_SHIFT)
+
+#define I40IW_AEQE_TCPSTATE_SHIFT 57
+#define I40IW_AEQE_TCPSTATE_MASK (0xfULL << I40IW_AEQE_TCPSTATE_SHIFT)
+
+#define I40IW_AEQE_Q2DATA_SHIFT 61
+#define I40IW_AEQE_Q2DATA_MASK (0x3ULL << I40IW_AEQE_Q2DATA_SHIFT)
+
+#define I40IW_AEQE_VALID_SHIFT 63
+#define I40IW_AEQE_VALID_MASK (1ULL << I40IW_AEQE_VALID_SHIFT)
+
+/* CQP SQ WQES */
+#define I40IW_QP_TYPE_IWARP     1
+#define I40IW_QP_TYPE_UDA       2
+#define I40IW_QP_TYPE_CQP       4
+
+#define I40IW_CQ_TYPE_IWARP     1
+#define I40IW_CQ_TYPE_ILQ       2
+#define I40IW_CQ_TYPE_IEQ       3
+#define I40IW_CQ_TYPE_CQP       4
+
+#define I40IWQP_TERM_SEND_TERM_AND_FIN          0
+#define I40IWQP_TERM_SEND_TERM_ONLY             1
+#define I40IWQP_TERM_SEND_FIN_ONLY              2
+#define I40IWQP_TERM_DONOT_SEND_TERM_OR_FIN     3
+
+#define I40IW_CQP_OP_CREATE_QP                  0
+#define I40IW_CQP_OP_MODIFY_QP                  0x1
+#define I40IW_CQP_OP_DESTROY_QP                 0x02
+#define I40IW_CQP_OP_CREATE_CQ                  0x03
+#define I40IW_CQP_OP_MODIFY_CQ                  0x04
+#define I40IW_CQP_OP_DESTROY_CQ                 0x05
+#define I40IW_CQP_OP_CREATE_SRQ                 0x06
+#define I40IW_CQP_OP_MODIFY_SRQ                 0x07
+#define I40IW_CQP_OP_DESTROY_SRQ                0x08
+#define I40IW_CQP_OP_ALLOC_STAG                 0x09
+#define I40IW_CQP_OP_REG_MR                     0x0a
+#define I40IW_CQP_OP_QUERY_STAG                 0x0b
+#define I40IW_CQP_OP_REG_SMR                    0x0c
+#define I40IW_CQP_OP_DEALLOC_STAG               0x0d
+#define I40IW_CQP_OP_MANAGE_LOC_MAC_IP_TABLE    0x0e
+#define I40IW_CQP_OP_MANAGE_ARP                 0x0f
+#define I40IW_CQP_OP_MANAGE_VF_PBLE_BP          0x10
+#define I40IW_CQP_OP_MANAGE_PUSH_PAGES          0x11
+#define I40IW_CQP_OP_MANAGE_PE_TEAM             0x12
+#define I40IW_CQP_OP_UPLOAD_CONTEXT             0x13
+#define I40IW_CQP_OP_ALLOCATE_LOC_MAC_IP_TABLE_ENTRY 0x14
+#define I40IW_CQP_OP_MANAGE_HMC_PM_FUNC_TABLE   0x15
+#define I40IW_CQP_OP_CREATE_CEQ                 0x16
+#define I40IW_CQP_OP_DESTROY_CEQ                0x18
+#define I40IW_CQP_OP_CREATE_AEQ                 0x19
+#define I40IW_CQP_OP_DESTROY_AEQ                0x1b
+#define I40IW_CQP_OP_CREATE_ADDR_VECT           0x1c
+#define I40IW_CQP_OP_MODIFY_ADDR_VECT           0x1d
+#define I40IW_CQP_OP_DESTROY_ADDR_VECT          0x1e
+#define I40IW_CQP_OP_UPDATE_PE_SDS              0x1f
+#define I40IW_CQP_OP_QUERY_FPM_VALUES           0x20
+#define I40IW_CQP_OP_COMMIT_FPM_VALUES          0x21
+#define I40IW_CQP_OP_FLUSH_WQES                 0x22
+#define I40IW_CQP_OP_MANAGE_APBVT               0x23
+#define I40IW_CQP_OP_NOP                        0x24
+#define I40IW_CQP_OP_MANAGE_QUAD_HASH_TABLE_ENTRY 0x25
+#define I40IW_CQP_OP_CREATE_UDA_MCAST_GROUP     0x26
+#define I40IW_CQP_OP_MODIFY_UDA_MCAST_GROUP     0x27
+#define I40IW_CQP_OP_DESTROY_UDA_MCAST_GROUP    0x28
+#define I40IW_CQP_OP_SUSPEND_QP                 0x29
+#define I40IW_CQP_OP_RESUME_QP                  0x2a
+#define I40IW_CQP_OP_SHMC_PAGES_ALLOCATED       0x2b
+#define I40IW_CQP_OP_SET_HMC_RESOURCE_PROFILE   0x2d
+
+#define I40IW_UDA_QPSQ_NEXT_HEADER_SHIFT 16
+#define I40IW_UDA_QPSQ_NEXT_HEADER_MASK ((u64)0xff << I40IW_UDA_QPSQ_NEXT_HEADER_SHIFT)
+
+#define I40IW_UDA_QPSQ_OPCODE_SHIFT 32
+#define I40IW_UDA_QPSQ_OPCODE_MASK ((u64)0x3f << I40IW_UDA_QPSQ_OPCODE_SHIFT)
+
+#define I40IW_UDA_QPSQ_MACLEN_SHIFT 56
+#define I40IW_UDA_QPSQ_MACLEN_MASK \
+	((u64)0x7f << I40IW_UDA_QPSQ_MACLEN_SHIFT)
+
+#define I40IW_UDA_QPSQ_IPLEN_SHIFT 48
+#define I40IW_UDA_QPSQ_IPLEN_MASK \
+	((u64)0x7f << I40IW_UDA_QPSQ_IPLEN_SHIFT)
+
+#define I40IW_UDA_QPSQ_L4T_SHIFT 30
+#define I40IW_UDA_QPSQ_L4T_MASK \
+	((u64)0x3 << I40IW_UDA_QPSQ_L4T_SHIFT)
+
+#define I40IW_UDA_QPSQ_IIPT_SHIFT 28
+#define I40IW_UDA_QPSQ_IIPT_MASK \
+	((u64)0x3 << I40IW_UDA_QPSQ_IIPT_SHIFT)
+
+#define I40IW_UDA_QPSQ_L4LEN_SHIFT 24
+#define I40IW_UDA_QPSQ_L4LEN_MASK ((u64)0xf << I40IW_UDA_QPSQ_L4LEN_SHIFT)
+
+#define I40IW_UDA_QPSQ_AVIDX_SHIFT 0
+#define I40IW_UDA_QPSQ_AVIDX_MASK ((u64)0xffff << I40IW_UDA_QPSQ_AVIDX_SHIFT)
+
+#define I40IW_UDA_QPSQ_VALID_SHIFT 63
+#define I40IW_UDA_QPSQ_VALID_MASK \
+	((u64)0x1 << I40IW_UDA_QPSQ_VALID_SHIFT)
+
+#define I40IW_UDA_QPSQ_SIGCOMPL_SHIFT 62
+#define I40IW_UDA_QPSQ_SIGCOMPL_MASK ((u64)0x1 << I40IW_UDA_QPSQ_SIGCOMPL_SHIFT)
+
+#define I40IW_UDA_PAYLOADLEN_SHIFT 0
+#define I40IW_UDA_PAYLOADLEN_MASK ((u64)0x3fff << I40IW_UDA_PAYLOADLEN_SHIFT)
+
+#define I40IW_UDA_HDRLEN_SHIFT 16
+#define I40IW_UDA_HDRLEN_MASK ((u64)0x1ff << I40IW_UDA_HDRLEN_SHIFT)
+
+#define I40IW_VLAN_TAG_VALID_SHIFT 50
+#define I40IW_VLAN_TAG_VALID_MASK ((u64)0x1 << I40IW_VLAN_TAG_VALID_SHIFT)
+
+#define I40IW_UDA_L3PROTO_SHIFT 0
+#define I40IW_UDA_L3PROTO_MASK ((u64)0x3 << I40IW_UDA_L3PROTO_SHIFT)
+
+#define I40IW_UDA_L4PROTO_SHIFT 16
+#define I40IW_UDA_L4PROTO_MASK ((u64)0x3 << I40IW_UDA_L4PROTO_SHIFT)
+
+#define I40IW_UDA_QPSQ_DOLOOPBACK_SHIFT 44
+#define I40IW_UDA_QPSQ_DOLOOPBACK_MASK \
+	((u64)0x1 << I40IW_UDA_QPSQ_DOLOOPBACK_SHIFT)
+
+/* CQP SQ WQE common fields */
+#define I40IW_CQPSQ_OPCODE_SHIFT 32
+#define I40IW_CQPSQ_OPCODE_MASK (0x3fULL << I40IW_CQPSQ_OPCODE_SHIFT)
+
+#define I40IW_CQPSQ_WQEVALID_SHIFT 63
+#define I40IW_CQPSQ_WQEVALID_MASK (1ULL << I40IW_CQPSQ_WQEVALID_SHIFT)
+
+#define I40IW_CQPSQ_TPHVAL_SHIFT 0
+#define I40IW_CQPSQ_TPHVAL_MASK (0xffUL << I40IW_CQPSQ_TPHVAL_SHIFT)
+
+#define I40IW_CQPSQ_TPHEN_SHIFT 60
+#define I40IW_CQPSQ_TPHEN_MASK (1ULL << I40IW_CQPSQ_TPHEN_SHIFT)
+
+#define I40IW_CQPSQ_PBUFADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_PBUFADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+/* Create/Modify/Destroy QP */
+
+#define I40IW_CQPSQ_QP_NEWMSS_SHIFT 32
+#define I40IW_CQPSQ_QP_NEWMSS_MASK (0x3fffULL << I40IW_CQPSQ_QP_NEWMSS_SHIFT)
+
+#define I40IW_CQPSQ_QP_TERMLEN_SHIFT 48
+#define I40IW_CQPSQ_QP_TERMLEN_MASK (0xfULL << I40IW_CQPSQ_QP_TERMLEN_SHIFT)
+
+#define I40IW_CQPSQ_QP_QPCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_QP_QPCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_QP_QPID_SHIFT 0
+#define I40IW_CQPSQ_QP_QPID_MASK (0x3FFFFUL)
+/* I40IWCQ_QPID_MASK */
+
+#define I40IW_CQPSQ_QP_OP_SHIFT 32
+#define I40IW_CQPSQ_QP_OP_MASK I40IWCQ_OP_MASK
+
+#define I40IW_CQPSQ_QP_ORDVALID_SHIFT 42
+#define I40IW_CQPSQ_QP_ORDVALID_MASK (1ULL << I40IW_CQPSQ_QP_ORDVALID_SHIFT)
+
+#define I40IW_CQPSQ_QP_TOECTXVALID_SHIFT 43
+#define I40IW_CQPSQ_QP_TOECTXVALID_MASK \
+	(1ULL << I40IW_CQPSQ_QP_TOECTXVALID_SHIFT)
+
+#define I40IW_CQPSQ_QP_CACHEDVARVALID_SHIFT 44
+#define I40IW_CQPSQ_QP_CACHEDVARVALID_MASK      \
+	(1ULL << I40IW_CQPSQ_QP_CACHEDVARVALID_SHIFT)
+
+#define I40IW_CQPSQ_QP_VQ_SHIFT 45
+#define I40IW_CQPSQ_QP_VQ_MASK (1ULL << I40IW_CQPSQ_QP_VQ_SHIFT)
+
+#define I40IW_CQPSQ_QP_FORCELOOPBACK_SHIFT 46
+#define I40IW_CQPSQ_QP_FORCELOOPBACK_MASK       \
+	(1ULL << I40IW_CQPSQ_QP_FORCELOOPBACK_SHIFT)
+
+#define I40IW_CQPSQ_QP_CQNUMVALID_SHIFT 47
+#define I40IW_CQPSQ_QP_CQNUMVALID_MASK  \
+	(1ULL << I40IW_CQPSQ_QP_CQNUMVALID_SHIFT)
+
+#define I40IW_CQPSQ_QP_QPTYPE_SHIFT 48
+#define I40IW_CQPSQ_QP_QPTYPE_MASK (0x3ULL << I40IW_CQPSQ_QP_QPTYPE_SHIFT)
+
+#define I40IW_CQPSQ_QP_MSSCHANGE_SHIFT 52
+#define I40IW_CQPSQ_QP_MSSCHANGE_MASK (1ULL << I40IW_CQPSQ_QP_MSSCHANGE_SHIFT)
+
+#define I40IW_CQPSQ_QP_STATRSRC_SHIFT 53
+#define I40IW_CQPSQ_QP_STATRSRC_MASK (1ULL << I40IW_CQPSQ_QP_STATRSRC_SHIFT)
+
+#define I40IW_CQPSQ_QP_IGNOREMWBOUND_SHIFT 54
+#define I40IW_CQPSQ_QP_IGNOREMWBOUND_MASK       \
+	(1ULL << I40IW_CQPSQ_QP_IGNOREMWBOUND_SHIFT)
+
+#define I40IW_CQPSQ_QP_REMOVEHASHENTRY_SHIFT 55
+#define I40IW_CQPSQ_QP_REMOVEHASHENTRY_MASK     \
+	(1ULL << I40IW_CQPSQ_QP_REMOVEHASHENTRY_SHIFT)
+
+#define I40IW_CQPSQ_QP_TERMACT_SHIFT 56
+#define I40IW_CQPSQ_QP_TERMACT_MASK (0x3ULL << I40IW_CQPSQ_QP_TERMACT_SHIFT)
+
+#define I40IW_CQPSQ_QP_RESETCON_SHIFT 58
+#define I40IW_CQPSQ_QP_RESETCON_MASK (1ULL << I40IW_CQPSQ_QP_RESETCON_SHIFT)
+
+#define I40IW_CQPSQ_QP_ARPTABIDXVALID_SHIFT 59
+#define I40IW_CQPSQ_QP_ARPTABIDXVALID_MASK      \
+	(1ULL << I40IW_CQPSQ_QP_ARPTABIDXVALID_SHIFT)
+
+#define I40IW_CQPSQ_QP_NEXTIWSTATE_SHIFT 60
+#define I40IW_CQPSQ_QP_NEXTIWSTATE_MASK \
+	(0x7ULL << I40IW_CQPSQ_QP_NEXTIWSTATE_SHIFT)
+
+#define I40IW_CQPSQ_QP_DBSHADOWADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_QP_DBSHADOWADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+/* Create/Modify/Destroy CQ */
+#define I40IW_CQPSQ_CQ_CQSIZE_SHIFT 0
+#define I40IW_CQPSQ_CQ_CQSIZE_MASK (0x3ffffUL << I40IW_CQPSQ_CQ_CQSIZE_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CQCTX_SHIFT 0
+#define I40IW_CQPSQ_CQ_CQCTX_MASK       \
+	(0x7fffffffffffffffULL << I40IW_CQPSQ_CQ_CQCTX_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CQCTX_SHIFT 0
+#define I40IW_CQPSQ_CQ_CQCTX_MASK       \
+	(0x7fffffffffffffffULL << I40IW_CQPSQ_CQ_CQCTX_SHIFT)
+
+#define I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD_SHIFT 0
+#define I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD_MASK       \
+	(0x3ffff << I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CEQID_SHIFT 24
+#define I40IW_CQPSQ_CQ_CEQID_MASK (0x7fUL << I40IW_CQPSQ_CQ_CEQID_SHIFT)
+
+#define I40IW_CQPSQ_CQ_OP_SHIFT 32
+#define I40IW_CQPSQ_CQ_OP_MASK (0x3fULL << I40IW_CQPSQ_CQ_OP_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CQRESIZE_SHIFT 43
+#define I40IW_CQPSQ_CQ_CQRESIZE_MASK (1ULL << I40IW_CQPSQ_CQ_CQRESIZE_SHIFT)
+
+#define I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT 44
+#define I40IW_CQPSQ_CQ_LPBLSIZE_MASK (3ULL << I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CHKOVERFLOW_SHIFT 46
+#define I40IW_CQPSQ_CQ_CHKOVERFLOW_MASK         \
+	(1ULL << I40IW_CQPSQ_CQ_CHKOVERFLOW_SHIFT)
+
+#define I40IW_CQPSQ_CQ_VIRTMAP_SHIFT 47
+#define I40IW_CQPSQ_CQ_VIRTMAP_MASK (1ULL << I40IW_CQPSQ_CQ_VIRTMAP_SHIFT)
+
+#define I40IW_CQPSQ_CQ_ENCEQEMASK_SHIFT 48
+#define I40IW_CQPSQ_CQ_ENCEQEMASK_MASK  \
+	(1ULL << I40IW_CQPSQ_CQ_ENCEQEMASK_SHIFT)
+
+#define I40IW_CQPSQ_CQ_CEQIDVALID_SHIFT 49
+#define I40IW_CQPSQ_CQ_CEQIDVALID_MASK  \
+	(1ULL << I40IW_CQPSQ_CQ_CEQIDVALID_SHIFT)
+
+#define I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT_SHIFT 61
+#define I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT_MASK      \
+	(1ULL << I40IW_CQPSQ_CQ_AVOIDMEMCNFLCT_SHIFT)
+
+/* Create/Modify/Destroy Shared Receive Queue */
+
+#define I40IW_CQPSQ_SRQ_RQSIZE_SHIFT 0
+#define I40IW_CQPSQ_SRQ_RQSIZE_MASK (0xfUL << I40IW_CQPSQ_SRQ_RQSIZE_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_RQWQESIZE_SHIFT 4
+#define I40IW_CQPSQ_SRQ_RQWQESIZE_MASK \
+	(0x7UL << I40IW_CQPSQ_SRQ_RQWQESIZE_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_SRQLIMIT_SHIFT 32
+#define I40IW_CQPSQ_SRQ_SRQLIMIT_MASK   \
+	(0xfffULL << I40IW_CQPSQ_SRQ_SRQLIMIT_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_SRQCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_SRQ_SRQCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_SRQ_PDID_SHIFT 16
+#define I40IW_CQPSQ_SRQ_PDID_MASK       \
+	(0x7fffULL << I40IW_CQPSQ_SRQ_PDID_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_SRQID_SHIFT 0
+#define I40IW_CQPSQ_SRQ_SRQID_MASK (0x7fffUL << I40IW_CQPSQ_SRQ_SRQID_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_LPBLSIZE_SHIFT I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT
+#define I40IW_CQPSQ_SRQ_LPBLSIZE_MASK I40IW_CQPSQ_CQ_LPBLSIZE_MASK
+
+#define I40IW_CQPSQ_SRQ_VIRTMAP_SHIFT I40IW_CQPSQ_CQ_VIRTMAP_SHIFT
+#define I40IW_CQPSQ_SRQ_VIRTMAP_MASK I40IW_CQPSQ_CQ_VIRTMAP_MASK
+
+#define I40IW_CQPSQ_SRQ_TPHEN_SHIFT I40IW_CQPSQ_TPHEN_SHIFT
+#define I40IW_CQPSQ_SRQ_TPHEN_MASK I40IW_CQPSQ_TPHEN_MASK
+
+#define I40IW_CQPSQ_SRQ_ARMLIMITEVENT_SHIFT 61
+#define I40IW_CQPSQ_SRQ_ARMLIMITEVENT_MASK      \
+	(1ULL << I40IW_CQPSQ_SRQ_ARMLIMITEVENT_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_DBSHADOWAREA_SHIFT 6
+#define I40IW_CQPSQ_SRQ_DBSHADOWAREA_MASK       \
+	(0x3ffffffffffffffULL << I40IW_CQPSQ_SRQ_DBSHADOWAREA_SHIFT)
+
+#define I40IW_CQPSQ_SRQ_FIRSTPMPBLIDX_SHIFT 0
+#define I40IW_CQPSQ_SRQ_FIRSTPMPBLIDX_MASK      \
+	(0xfffffffUL << I40IW_CQPSQ_SRQ_FIRSTPMPBLIDX_SHIFT)
+
+/* Allocate/Register/Register Shared/Deallocate Stag */
+#define I40IW_CQPSQ_STAG_VA_FBO_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_STAG_VA_FBO_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_STAG_STAGLEN_SHIFT 0
+#define I40IW_CQPSQ_STAG_STAGLEN_MASK   \
+	(0x3fffffffffffULL << I40IW_CQPSQ_STAG_STAGLEN_SHIFT)
+
+#define I40IW_CQPSQ_STAG_PDID_SHIFT 48
+#define I40IW_CQPSQ_STAG_PDID_MASK (0x7fffULL << I40IW_CQPSQ_STAG_PDID_SHIFT)
+
+#define I40IW_CQPSQ_STAG_KEY_SHIFT 0
+#define I40IW_CQPSQ_STAG_KEY_MASK (0xffUL << I40IW_CQPSQ_STAG_KEY_SHIFT)
+
+#define I40IW_CQPSQ_STAG_IDX_SHIFT 8
+#define I40IW_CQPSQ_STAG_IDX_MASK (0xffffffUL << I40IW_CQPSQ_STAG_IDX_SHIFT)
+
+#define I40IW_CQPSQ_STAG_PARENTSTAGIDX_SHIFT 32
+#define I40IW_CQPSQ_STAG_PARENTSTAGIDX_MASK     \
+	(0xffffffULL << I40IW_CQPSQ_STAG_PARENTSTAGIDX_SHIFT)
+
+#define I40IW_CQPSQ_STAG_MR_SHIFT 43
+#define I40IW_CQPSQ_STAG_MR_MASK (1ULL << I40IW_CQPSQ_STAG_MR_SHIFT)
+
+#define I40IW_CQPSQ_STAG_LPBLSIZE_SHIFT I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT
+#define I40IW_CQPSQ_STAG_LPBLSIZE_MASK I40IW_CQPSQ_CQ_LPBLSIZE_MASK
+
+#define I40IW_CQPSQ_STAG_HPAGESIZE_SHIFT 46
+#define I40IW_CQPSQ_STAG_HPAGESIZE_MASK \
+	(1ULL << I40IW_CQPSQ_STAG_HPAGESIZE_SHIFT)
+
+#define I40IW_CQPSQ_STAG_ARIGHTS_SHIFT 48
+#define I40IW_CQPSQ_STAG_ARIGHTS_MASK   \
+	(0x1fULL << I40IW_CQPSQ_STAG_ARIGHTS_SHIFT)
+
+#define I40IW_CQPSQ_STAG_REMACCENABLED_SHIFT 53
+#define I40IW_CQPSQ_STAG_REMACCENABLED_MASK     \
+	(1ULL << I40IW_CQPSQ_STAG_REMACCENABLED_SHIFT)
+
+#define I40IW_CQPSQ_STAG_VABASEDTO_SHIFT 59
+#define I40IW_CQPSQ_STAG_VABASEDTO_MASK \
+	(1ULL << I40IW_CQPSQ_STAG_VABASEDTO_SHIFT)
+
+#define I40IW_CQPSQ_STAG_USEHMCFNIDX_SHIFT 60
+#define I40IW_CQPSQ_STAG_USEHMCFNIDX_MASK       \
+	(1ULL << I40IW_CQPSQ_STAG_USEHMCFNIDX_SHIFT)
+
+#define I40IW_CQPSQ_STAG_USEPFRID_SHIFT 61
+#define I40IW_CQPSQ_STAG_USEPFRID_MASK  \
+	(1ULL << I40IW_CQPSQ_STAG_USEPFRID_SHIFT)
+
+#define I40IW_CQPSQ_STAG_PBA_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_STAG_PBA_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_STAG_HMCFNIDX_SHIFT 0
+#define I40IW_CQPSQ_STAG_HMCFNIDX_MASK \
+	(0x3fUL << I40IW_CQPSQ_STAG_HMCFNIDX_SHIFT)
+
+#define I40IW_CQPSQ_STAG_FIRSTPMPBLIDX_SHIFT 0
+#define I40IW_CQPSQ_STAG_FIRSTPMPBLIDX_MASK     \
+	(0xfffffffUL << I40IW_CQPSQ_STAG_FIRSTPMPBLIDX_SHIFT)
+
+/* Query stag */
+#define I40IW_CQPSQ_QUERYSTAG_IDX_SHIFT I40IW_CQPSQ_STAG_IDX_SHIFT
+#define I40IW_CQPSQ_QUERYSTAG_IDX_MASK I40IW_CQPSQ_STAG_IDX_MASK
+
+/* Allocate Local IP Address Entry */
+
+/* Manage Local IP Address Table - MLIPA */
+#define I40IW_CQPSQ_MLIPA_IPV6LO_SHIFT  I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_MLIPA_IPV6LO_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_MLIPA_IPV6HI_SHIFT  I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_MLIPA_IPV6HI_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_MLIPA_IPV4_SHIFT 0
+#define I40IW_CQPSQ_MLIPA_IPV4_MASK \
+	(0xffffffffUL << I40IW_CQPSQ_MLIPA_IPV4_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_IPTABLEIDX_SHIFT 0
+#define I40IW_CQPSQ_MLIPA_IPTABLEIDX_MASK       \
+	(0x3fUL << I40IW_CQPSQ_MLIPA_IPTABLEIDX_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_IPV4VALID_SHIFT 42
+#define I40IW_CQPSQ_MLIPA_IPV4VALID_MASK        \
+	(1ULL << I40IW_CQPSQ_MLIPA_IPV4VALID_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_IPV6VALID_SHIFT 43
+#define I40IW_CQPSQ_MLIPA_IPV6VALID_MASK        \
+	(1ULL << I40IW_CQPSQ_MLIPA_IPV6VALID_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_FREEENTRY_SHIFT 62
+#define I40IW_CQPSQ_MLIPA_FREEENTRY_MASK        \
+	(1ULL << I40IW_CQPSQ_MLIPA_FREEENTRY_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_IGNORE_REF_CNT_SHIFT 61
+#define I40IW_CQPSQ_MLIPA_IGNORE_REF_CNT_MASK   \
+	(1ULL << I40IW_CQPSQ_MLIPA_IGNORE_REF_CNT_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC0_SHIFT 0
+#define I40IW_CQPSQ_MLIPA_MAC0_MASK (0xffUL << I40IW_CQPSQ_MLIPA_MAC0_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC1_SHIFT 8
+#define I40IW_CQPSQ_MLIPA_MAC1_MASK (0xffUL << I40IW_CQPSQ_MLIPA_MAC1_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC2_SHIFT 16
+#define I40IW_CQPSQ_MLIPA_MAC2_MASK (0xffUL << I40IW_CQPSQ_MLIPA_MAC2_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC3_SHIFT 24
+#define I40IW_CQPSQ_MLIPA_MAC3_MASK (0xffUL << I40IW_CQPSQ_MLIPA_MAC3_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC4_SHIFT 32
+#define I40IW_CQPSQ_MLIPA_MAC4_MASK (0xffULL << I40IW_CQPSQ_MLIPA_MAC4_SHIFT)
+
+#define I40IW_CQPSQ_MLIPA_MAC5_SHIFT 40
+#define I40IW_CQPSQ_MLIPA_MAC5_MASK (0xffULL << I40IW_CQPSQ_MLIPA_MAC5_SHIFT)
+
+/* Manage ARP Table  - MAT */
+#define I40IW_CQPSQ_MAT_REACHMAX_SHIFT 0
+#define I40IW_CQPSQ_MAT_REACHMAX_MASK   \
+	(0xffffffffUL << I40IW_CQPSQ_MAT_REACHMAX_SHIFT)
+
+#define I40IW_CQPSQ_MAT_MACADDR_SHIFT 0
+#define I40IW_CQPSQ_MAT_MACADDR_MASK    \
+	(0xffffffffffffULL << I40IW_CQPSQ_MAT_MACADDR_SHIFT)
+
+#define I40IW_CQPSQ_MAT_ARPENTRYIDX_SHIFT 0
+#define I40IW_CQPSQ_MAT_ARPENTRYIDX_MASK        \
+	(0xfffUL << I40IW_CQPSQ_MAT_ARPENTRYIDX_SHIFT)
+
+#define I40IW_CQPSQ_MAT_ENTRYVALID_SHIFT 42
+#define I40IW_CQPSQ_MAT_ENTRYVALID_MASK \
+	(1ULL << I40IW_CQPSQ_MAT_ENTRYVALID_SHIFT)
+
+#define I40IW_CQPSQ_MAT_PERMANENT_SHIFT 43
+#define I40IW_CQPSQ_MAT_PERMANENT_MASK  \
+	(1ULL << I40IW_CQPSQ_MAT_PERMANENT_SHIFT)
+
+#define I40IW_CQPSQ_MAT_QUERY_SHIFT 44
+#define I40IW_CQPSQ_MAT_QUERY_MASK (1ULL << I40IW_CQPSQ_MAT_QUERY_SHIFT)
+
+/* Manage VF PBLE Backing Pages - MVPBP*/
+#define I40IW_CQPSQ_MVPBP_PD_ENTRY_CNT_SHIFT 0
+#define I40IW_CQPSQ_MVPBP_PD_ENTRY_CNT_MASK \
+	(0x3ffULL << I40IW_CQPSQ_MVPBP_PD_ENTRY_CNT_SHIFT)
+
+#define I40IW_CQPSQ_MVPBP_FIRST_PD_INX_SHIFT 16
+#define I40IW_CQPSQ_MVPBP_FIRST_PD_INX_MASK \
+	(0x1ffULL << I40IW_CQPSQ_MVPBP_FIRST_PD_INX_SHIFT)
+
+#define I40IW_CQPSQ_MVPBP_SD_INX_SHIFT 32
+#define I40IW_CQPSQ_MVPBP_SD_INX_MASK \
+	(0xfffULL << I40IW_CQPSQ_MVPBP_SD_INX_SHIFT)
+
+#define I40IW_CQPSQ_MVPBP_INV_PD_ENT_SHIFT 62
+#define I40IW_CQPSQ_MVPBP_INV_PD_ENT_MASK \
+	(0x1ULL << I40IW_CQPSQ_MVPBP_INV_PD_ENT_SHIFT)
+
+#define I40IW_CQPSQ_MVPBP_PD_PLPBA_SHIFT 3
+#define I40IW_CQPSQ_MVPBP_PD_PLPBA_MASK \
+	(0x1fffffffffffffffULL << I40IW_CQPSQ_MVPBP_PD_PLPBA_SHIFT)
+
+/* Manage Push Page - MPP */
+#define I40IW_INVALID_PUSH_PAGE_INDEX 0xffff
+
+#define I40IW_CQPSQ_MPP_QS_HANDLE_SHIFT 0
+#define I40IW_CQPSQ_MPP_QS_HANDLE_MASK (0xffffUL << \
+					I40IW_CQPSQ_MPP_QS_HANDLE_SHIFT)
+
+#define I40IW_CQPSQ_MPP_PPIDX_SHIFT 0
+#define I40IW_CQPSQ_MPP_PPIDX_MASK (0x3ffUL << I40IW_CQPSQ_MPP_PPIDX_SHIFT)
+
+#define I40IW_CQPSQ_MPP_FREE_PAGE_SHIFT 62
+#define I40IW_CQPSQ_MPP_FREE_PAGE_MASK (1ULL << I40IW_CQPSQ_MPP_FREE_PAGE_SHIFT)
+
+/* Upload Context - UCTX */
+#define I40IW_CQPSQ_UCTX_QPCTXADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IW_CQPSQ_UCTX_QPCTXADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IW_CQPSQ_UCTX_QPID_SHIFT 0
+#define I40IW_CQPSQ_UCTX_QPID_MASK (0x3ffffUL << I40IW_CQPSQ_UCTX_QPID_SHIFT)
+
+#define I40IW_CQPSQ_UCTX_QPTYPE_SHIFT 48
+#define I40IW_CQPSQ_UCTX_QPTYPE_MASK (0xfULL << I40IW_CQPSQ_UCTX_QPTYPE_SHIFT)
+
+#define I40IW_CQPSQ_UCTX_RAWFORMAT_SHIFT 61
+#define I40IW_CQPSQ_UCTX_RAWFORMAT_MASK \
+	(1ULL << I40IW_CQPSQ_UCTX_RAWFORMAT_SHIFT)
+
+#define I40IW_CQPSQ_UCTX_FREEZEQP_SHIFT 62
+#define I40IW_CQPSQ_UCTX_FREEZEQP_MASK  \
+	(1ULL << I40IW_CQPSQ_UCTX_FREEZEQP_SHIFT)
+
+/* Manage HMC PM Function Table - MHMC */
+#define I40IW_CQPSQ_MHMC_VFIDX_SHIFT 0
+#define I40IW_CQPSQ_MHMC_VFIDX_MASK (0x7fUL << I40IW_CQPSQ_MHMC_VFIDX_SHIFT)
+
+#define I40IW_CQPSQ_MHMC_FREEPMFN_SHIFT 62
+#define I40IW_CQPSQ_MHMC_FREEPMFN_MASK  \
+	(1ULL << I40IW_CQPSQ_MHMC_FREEPMFN_SHIFT)
+
+/* Set HMC Resource Profile - SHMCRP */
+#define I40IW_CQPSQ_SHMCRP_HMC_PROFILE_SHIFT 0
+#define I40IW_CQPSQ_SHMCRP_HMC_PROFILE_MASK \
+	(0x7ULL << I40IW_CQPSQ_SHMCRP_HMC_PROFILE_SHIFT)
+#define I40IW_CQPSQ_SHMCRP_VFNUM_SHIFT 32
+#define I40IW_CQPSQ_SHMCRP_VFNUM_MASK (0x3fULL << I40IW_CQPSQ_SHMCRP_VFNUM_SHIFT)
+
+/* Create/Destroy CEQ */
+#define I40IW_CQPSQ_CEQ_CEQSIZE_SHIFT 0
+#define I40IW_CQPSQ_CEQ_CEQSIZE_MASK \
+	(0x1ffffUL << I40IW_CQPSQ_CEQ_CEQSIZE_SHIFT)
+
+#define I40IW_CQPSQ_CEQ_CEQID_SHIFT 0
+#define I40IW_CQPSQ_CEQ_CEQID_MASK (0x7fUL << I40IW_CQPSQ_CEQ_CEQID_SHIFT)
+
+#define I40IW_CQPSQ_CEQ_LPBLSIZE_SHIFT I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT
+#define I40IW_CQPSQ_CEQ_LPBLSIZE_MASK I40IW_CQPSQ_CQ_LPBLSIZE_MASK
+
+#define I40IW_CQPSQ_CEQ_VMAP_SHIFT 47
+#define I40IW_CQPSQ_CEQ_VMAP_MASK (1ULL << I40IW_CQPSQ_CEQ_VMAP_SHIFT)
+
+#define I40IW_CQPSQ_CEQ_FIRSTPMPBLIDX_SHIFT 0
+#define I40IW_CQPSQ_CEQ_FIRSTPMPBLIDX_MASK      \
+	(0xfffffffUL << I40IW_CQPSQ_CEQ_FIRSTPMPBLIDX_SHIFT)
+
+/* Create/Destroy AEQ */
+#define I40IW_CQPSQ_AEQ_AEQECNT_SHIFT 0
+#define I40IW_CQPSQ_AEQ_AEQECNT_MASK \
+	(0x7ffffUL << I40IW_CQPSQ_AEQ_AEQECNT_SHIFT)
+
+#define I40IW_CQPSQ_AEQ_LPBLSIZE_SHIFT I40IW_CQPSQ_CQ_LPBLSIZE_SHIFT
+#define I40IW_CQPSQ_AEQ_LPBLSIZE_MASK I40IW_CQPSQ_CQ_LPBLSIZE_MASK
+
+#define I40IW_CQPSQ_AEQ_VMAP_SHIFT 47
+#define I40IW_CQPSQ_AEQ_VMAP_MASK (1ULL << I40IW_CQPSQ_AEQ_VMAP_SHIFT)
+
+#define I40IW_CQPSQ_AEQ_FIRSTPMPBLIDX_SHIFT 0
+#define I40IW_CQPSQ_AEQ_FIRSTPMPBLIDX_MASK      \
+	(0xfffffffUL << I40IW_CQPSQ_AEQ_FIRSTPMPBLIDX_SHIFT)
+
+/* Commit FPM Values - CFPM */
+#define I40IW_CQPSQ_CFPM_HMCFNID_SHIFT 0
+#define I40IW_CQPSQ_CFPM_HMCFNID_MASK (0x3fUL << I40IW_CQPSQ_CFPM_HMCFNID_SHIFT)
+
+/* Flush WQEs - FWQE */
+#define I40IW_CQPSQ_FWQE_AECODE_SHIFT 0
+#define I40IW_CQPSQ_FWQE_AECODE_MASK (0xffffUL << I40IW_CQPSQ_FWQE_AECODE_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_AESOURCE_SHIFT 16
+#define I40IW_CQPSQ_FWQE_AESOURCE_MASK \
+	(0xfUL << I40IW_CQPSQ_FWQE_AESOURCE_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_RQMNERR_SHIFT 0
+#define I40IW_CQPSQ_FWQE_RQMNERR_MASK \
+	(0xffffUL << I40IW_CQPSQ_FWQE_RQMNERR_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_RQMJERR_SHIFT 16
+#define I40IW_CQPSQ_FWQE_RQMJERR_MASK \
+	(0xffffUL << I40IW_CQPSQ_FWQE_RQMJERR_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_SQMNERR_SHIFT 32
+#define I40IW_CQPSQ_FWQE_SQMNERR_MASK   \
+	(0xffffULL << I40IW_CQPSQ_FWQE_SQMNERR_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_SQMJERR_SHIFT 48
+#define I40IW_CQPSQ_FWQE_SQMJERR_MASK   \
+	(0xffffULL << I40IW_CQPSQ_FWQE_SQMJERR_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_QPID_SHIFT 0
+#define I40IW_CQPSQ_FWQE_QPID_MASK (0x3ffffULL << I40IW_CQPSQ_FWQE_QPID_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_GENERATE_AE_SHIFT 59
+#define I40IW_CQPSQ_FWQE_GENERATE_AE_MASK (1ULL <<      \
+					   I40IW_CQPSQ_FWQE_GENERATE_AE_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_USERFLCODE_SHIFT 60
+#define I40IW_CQPSQ_FWQE_USERFLCODE_MASK        \
+	(1ULL << I40IW_CQPSQ_FWQE_USERFLCODE_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_FLUSHSQ_SHIFT 61
+#define I40IW_CQPSQ_FWQE_FLUSHSQ_MASK (1ULL << I40IW_CQPSQ_FWQE_FLUSHSQ_SHIFT)
+
+#define I40IW_CQPSQ_FWQE_FLUSHRQ_SHIFT 62
+#define I40IW_CQPSQ_FWQE_FLUSHRQ_MASK (1ULL << I40IW_CQPSQ_FWQE_FLUSHRQ_SHIFT)
+
+/* Manage Accelerated Port Table - MAPT */
+#define I40IW_CQPSQ_MAPT_PORT_SHIFT 0
+#define I40IW_CQPSQ_MAPT_PORT_MASK (0xffffUL << I40IW_CQPSQ_MAPT_PORT_SHIFT)
+
+#define I40IW_CQPSQ_MAPT_ADDPORT_SHIFT 62
+#define I40IW_CQPSQ_MAPT_ADDPORT_MASK (1ULL << I40IW_CQPSQ_MAPT_ADDPORT_SHIFT)
+
+/* Update Protocol Engine SDs */
+#define I40IW_CQPSQ_UPESD_SDCMD_SHIFT 0
+#define I40IW_CQPSQ_UPESD_SDCMD_MASK (0xffffffffUL << I40IW_CQPSQ_UPESD_SDCMD_SHIFT)
+
+#define I40IW_CQPSQ_UPESD_SDDATALOW_SHIFT 0
+#define I40IW_CQPSQ_UPESD_SDDATALOW_MASK        \
+	(0xffffffffUL << I40IW_CQPSQ_UPESD_SDDATALOW_SHIFT)
+
+#define I40IW_CQPSQ_UPESD_SDDATAHI_SHIFT 32
+#define I40IW_CQPSQ_UPESD_SDDATAHI_MASK \
+	(0xffffffffULL << I40IW_CQPSQ_UPESD_SDDATAHI_SHIFT)
+#define I40IW_CQPSQ_UPESD_HMCFNID_SHIFT 0
+#define I40IW_CQPSQ_UPESD_HMCFNID_MASK  \
+	(0x3fUL << I40IW_CQPSQ_UPESD_HMCFNID_SHIFT)
+
+#define I40IW_CQPSQ_UPESD_ENTRY_VALID_SHIFT 63
+#define I40IW_CQPSQ_UPESD_ENTRY_VALID_MASK      \
+	((u64)1 << I40IW_CQPSQ_UPESD_ENTRY_VALID_SHIFT)
+
+#define I40IW_CQPSQ_UPESD_ENTRY_COUNT_SHIFT 0
+#define I40IW_CQPSQ_UPESD_ENTRY_COUNT_MASK      \
+	(0xfUL << I40IW_CQPSQ_UPESD_ENTRY_COUNT_SHIFT)
+
+#define I40IW_CQPSQ_UPESD_SKIP_ENTRY_SHIFT 7
+#define I40IW_CQPSQ_UPESD_SKIP_ENTRY_MASK       \
+	(0x1UL << I40IW_CQPSQ_UPESD_SKIP_ENTRY_SHIFT)
+
+/* Suspend QP */
+#define I40IW_CQPSQ_SUSPENDQP_QPID_SHIFT 0
+#define I40IW_CQPSQ_SUSPENDQP_QPID_MASK (0x3FFFFUL)
+/* I40IWCQ_QPID_MASK */
+
+/* Resume QP */
+#define I40IW_CQPSQ_RESUMEQP_QSHANDLE_SHIFT 0
+#define I40IW_CQPSQ_RESUMEQP_QSHANDLE_MASK      \
+	(0xffffffffUL << I40IW_CQPSQ_RESUMEQP_QSHANDLE_SHIFT)
+
+#define I40IW_CQPSQ_RESUMEQP_QPID_SHIFT 0
+#define I40IW_CQPSQ_RESUMEQP_QPID_MASK (0x3FFFFUL)
+/* I40IWCQ_QPID_MASK */
+
+/* IW QP Context */
+#define I40IWQPC_DDP_VER_SHIFT 0
+#define I40IWQPC_DDP_VER_MASK (3UL << I40IWQPC_DDP_VER_SHIFT)
+
+#define I40IWQPC_SNAP_SHIFT 2
+#define I40IWQPC_SNAP_MASK (1UL << I40IWQPC_SNAP_SHIFT)
+
+#define I40IWQPC_IPV4_SHIFT 3
+#define I40IWQPC_IPV4_MASK (1UL << I40IWQPC_IPV4_SHIFT)
+
+#define I40IWQPC_NONAGLE_SHIFT 4
+#define I40IWQPC_NONAGLE_MASK (1UL << I40IWQPC_NONAGLE_SHIFT)
+
+#define I40IWQPC_INSERTVLANTAG_SHIFT 5
+#define I40IWQPC_INSERTVLANTAG_MASK (1 << I40IWQPC_INSERTVLANTAG_SHIFT)
+
+#define I40IWQPC_USESRQ_SHIFT 6
+#define I40IWQPC_USESRQ_MASK (1UL << I40IWQPC_USESRQ_SHIFT)
+
+#define I40IWQPC_TIMESTAMP_SHIFT 7
+#define I40IWQPC_TIMESTAMP_MASK (1UL << I40IWQPC_TIMESTAMP_SHIFT)
+
+#define I40IWQPC_RQWQESIZE_SHIFT 8
+#define I40IWQPC_RQWQESIZE_MASK (3UL << I40IWQPC_RQWQESIZE_SHIFT)
+
+#define I40IWQPC_INSERTL2TAG2_SHIFT 11
+#define I40IWQPC_INSERTL2TAG2_MASK (1UL << I40IWQPC_INSERTL2TAG2_SHIFT)
+
+#define I40IWQPC_LIMIT_SHIFT 12
+#define I40IWQPC_LIMIT_MASK (3UL << I40IWQPC_LIMIT_SHIFT)
+
+#define I40IWQPC_DROPOOOSEG_SHIFT 15
+#define I40IWQPC_DROPOOOSEG_MASK (1UL << I40IWQPC_DROPOOOSEG_SHIFT)
+
+#define I40IWQPC_DUPACK_THRESH_SHIFT 16
+#define I40IWQPC_DUPACK_THRESH_MASK (7UL << I40IWQPC_DUPACK_THRESH_SHIFT)
+
+#define I40IWQPC_ERR_RQ_IDX_VALID_SHIFT 19
+#define I40IWQPC_ERR_RQ_IDX_VALID_MASK  (1UL << I40IWQPC_ERR_RQ_IDX_VALID_SHIFT)
+
+#define I40IWQPC_DIS_VLAN_CHECKS_SHIFT 19
+#define I40IWQPC_DIS_VLAN_CHECKS_MASK (7UL << I40IWQPC_DIS_VLAN_CHECKS_SHIFT)
+
+#define I40IWQPC_RCVTPHEN_SHIFT 28
+#define I40IWQPC_RCVTPHEN_MASK (1UL << I40IWQPC_RCVTPHEN_SHIFT)
+
+#define I40IWQPC_XMITTPHEN_SHIFT 29
+#define I40IWQPC_XMITTPHEN_MASK (1ULL << I40IWQPC_XMITTPHEN_SHIFT)
+
+#define I40IWQPC_RQTPHEN_SHIFT 30
+#define I40IWQPC_RQTPHEN_MASK (1UL << I40IWQPC_RQTPHEN_SHIFT)
+
+#define I40IWQPC_SQTPHEN_SHIFT 31
+#define I40IWQPC_SQTPHEN_MASK (1ULL << I40IWQPC_SQTPHEN_SHIFT)
+
+#define I40IWQPC_PPIDX_SHIFT 32
+#define I40IWQPC_PPIDX_MASK (0x3ffULL << I40IWQPC_PPIDX_SHIFT)
+
+#define I40IWQPC_PMENA_SHIFT 47
+#define I40IWQPC_PMENA_MASK (1ULL << I40IWQPC_PMENA_SHIFT)
+
+#define I40IWQPC_RDMAP_VER_SHIFT 62
+#define I40IWQPC_RDMAP_VER_MASK (3ULL << I40IWQPC_RDMAP_VER_SHIFT)
+
+#define I40IWQPC_SQADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPC_SQADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPC_RQADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPC_RQADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPC_TTL_SHIFT 0
+#define I40IWQPC_TTL_MASK (0xffUL << I40IWQPC_TTL_SHIFT)
+
+#define I40IWQPC_RQSIZE_SHIFT 8
+#define I40IWQPC_RQSIZE_MASK (0xfUL << I40IWQPC_RQSIZE_SHIFT)
+
+#define I40IWQPC_SQSIZE_SHIFT 12
+#define I40IWQPC_SQSIZE_MASK (0xfUL << I40IWQPC_SQSIZE_SHIFT)
+
+#define I40IWQPC_SRCMACADDRIDX_SHIFT 16
+#define I40IWQPC_SRCMACADDRIDX_MASK (0x3fUL << I40IWQPC_SRCMACADDRIDX_SHIFT)
+
+#define I40IWQPC_AVOIDSTRETCHACK_SHIFT 23
+#define I40IWQPC_AVOIDSTRETCHACK_MASK (1UL << I40IWQPC_AVOIDSTRETCHACK_SHIFT)
+
+#define I40IWQPC_TOS_SHIFT 24
+#define I40IWQPC_TOS_MASK (0xffUL << I40IWQPC_TOS_SHIFT)
+
+#define I40IWQPC_SRCPORTNUM_SHIFT 32
+#define I40IWQPC_SRCPORTNUM_MASK (0xffffULL << I40IWQPC_SRCPORTNUM_SHIFT)
+
+#define I40IWQPC_DESTPORTNUM_SHIFT 48
+#define I40IWQPC_DESTPORTNUM_MASK (0xffffULL << I40IWQPC_DESTPORTNUM_SHIFT)
+
+#define I40IWQPC_DESTIPADDR0_SHIFT 32
+#define I40IWQPC_DESTIPADDR0_MASK       \
+	(0xffffffffULL << I40IWQPC_DESTIPADDR0_SHIFT)
+
+#define I40IWQPC_DESTIPADDR1_SHIFT 0
+#define I40IWQPC_DESTIPADDR1_MASK       \
+	(0xffffffffULL << I40IWQPC_DESTIPADDR1_SHIFT)
+
+#define I40IWQPC_DESTIPADDR2_SHIFT 32
+#define I40IWQPC_DESTIPADDR2_MASK       \
+	(0xffffffffULL << I40IWQPC_DESTIPADDR2_SHIFT)
+
+#define I40IWQPC_DESTIPADDR3_SHIFT 0
+#define I40IWQPC_DESTIPADDR3_MASK       \
+	(0xffffffffULL << I40IWQPC_DESTIPADDR3_SHIFT)
+
+#define I40IWQPC_SNDMSS_SHIFT 16
+#define I40IWQPC_SNDMSS_MASK (0x3fffUL << I40IWQPC_SNDMSS_SHIFT)
+
+#define I40IWQPC_VLANTAG_SHIFT 32
+#define I40IWQPC_VLANTAG_MASK (0xffffULL << I40IWQPC_VLANTAG_SHIFT)
+
+#define I40IWQPC_ARPIDX_SHIFT 48
+#define I40IWQPC_ARPIDX_MASK (0xfffULL << I40IWQPC_ARPIDX_SHIFT)
+
+#define I40IWQPC_FLOWLABEL_SHIFT 0
+#define I40IWQPC_FLOWLABEL_MASK (0xfffffUL << I40IWQPC_FLOWLABEL_SHIFT)
+
+#define I40IWQPC_WSCALE_SHIFT 20
+#define I40IWQPC_WSCALE_MASK (1UL << I40IWQPC_WSCALE_SHIFT)
+
+#define I40IWQPC_KEEPALIVE_SHIFT 21
+#define I40IWQPC_KEEPALIVE_MASK (1UL << I40IWQPC_KEEPALIVE_SHIFT)
+
+#define I40IWQPC_IGNORE_TCP_OPT_SHIFT 22
+#define I40IWQPC_IGNORE_TCP_OPT_MASK (1UL << I40IWQPC_IGNORE_TCP_OPT_SHIFT)
+
+#define I40IWQPC_IGNORE_TCP_UNS_OPT_SHIFT 23
+#define I40IWQPC_IGNORE_TCP_UNS_OPT_MASK        \
+	(1UL << I40IWQPC_IGNORE_TCP_UNS_OPT_SHIFT)
+
+#define I40IWQPC_TCPSTATE_SHIFT 28
+#define I40IWQPC_TCPSTATE_MASK (0xfUL << I40IWQPC_TCPSTATE_SHIFT)
+
+#define I40IWQPC_RCVSCALE_SHIFT 32
+#define I40IWQPC_RCVSCALE_MASK (0xfULL << I40IWQPC_RCVSCALE_SHIFT)
+
+#define I40IWQPC_SNDSCALE_SHIFT 40
+#define I40IWQPC_SNDSCALE_MASK (0xfULL << I40IWQPC_SNDSCALE_SHIFT)
+
+#define I40IWQPC_PDIDX_SHIFT 48
+#define I40IWQPC_PDIDX_MASK (0x7fffULL << I40IWQPC_PDIDX_SHIFT)
+
+#define I40IWQPC_KALIVE_TIMER_MAX_PROBES_SHIFT 16
+#define I40IWQPC_KALIVE_TIMER_MAX_PROBES_MASK   \
+	(0xffUL << I40IWQPC_KALIVE_TIMER_MAX_PROBES_SHIFT)
+
+#define I40IWQPC_KEEPALIVE_INTERVAL_SHIFT 24
+#define I40IWQPC_KEEPALIVE_INTERVAL_MASK        \
+	(0xffUL << I40IWQPC_KEEPALIVE_INTERVAL_SHIFT)
+
+#define I40IWQPC_TIMESTAMP_RECENT_SHIFT 0
+#define I40IWQPC_TIMESTAMP_RECENT_MASK  \
+	(0xffffffffUL << I40IWQPC_TIMESTAMP_RECENT_SHIFT)
+
+#define I40IWQPC_TIMESTAMP_AGE_SHIFT 32
+#define I40IWQPC_TIMESTAMP_AGE_MASK     \
+	(0xffffffffULL << I40IWQPC_TIMESTAMP_AGE_SHIFT)
+
+#define I40IWQPC_SNDNXT_SHIFT 0
+#define I40IWQPC_SNDNXT_MASK (0xffffffffUL << I40IWQPC_SNDNXT_SHIFT)
+
+#define I40IWQPC_SNDWND_SHIFT 32
+#define I40IWQPC_SNDWND_MASK (0xffffffffULL << I40IWQPC_SNDWND_SHIFT)
+
+#define I40IWQPC_RCVNXT_SHIFT 0
+#define I40IWQPC_RCVNXT_MASK (0xffffffffUL << I40IWQPC_RCVNXT_SHIFT)
+
+#define I40IWQPC_RCVWND_SHIFT 32
+#define I40IWQPC_RCVWND_MASK (0xffffffffULL << I40IWQPC_RCVWND_SHIFT)
+
+#define I40IWQPC_SNDMAX_SHIFT 0
+#define I40IWQPC_SNDMAX_MASK (0xffffffffUL << I40IWQPC_SNDMAX_SHIFT)
+
+#define I40IWQPC_SNDUNA_SHIFT 32
+#define I40IWQPC_SNDUNA_MASK (0xffffffffULL << I40IWQPC_SNDUNA_SHIFT)
+
+#define I40IWQPC_SRTT_SHIFT 0
+#define I40IWQPC_SRTT_MASK (0xffffffffUL << I40IWQPC_SRTT_SHIFT)
+
+#define I40IWQPC_RTTVAR_SHIFT 32
+#define I40IWQPC_RTTVAR_MASK (0xffffffffULL << I40IWQPC_RTTVAR_SHIFT)
+
+#define I40IWQPC_SSTHRESH_SHIFT 0
+#define I40IWQPC_SSTHRESH_MASK (0xffffffffUL << I40IWQPC_SSTHRESH_SHIFT)
+
+#define I40IWQPC_CWND_SHIFT 32
+#define I40IWQPC_CWND_MASK (0xffffffffULL << I40IWQPC_CWND_SHIFT)
+
+#define I40IWQPC_SNDWL1_SHIFT 0
+#define I40IWQPC_SNDWL1_MASK (0xffffffffUL << I40IWQPC_SNDWL1_SHIFT)
+
+#define I40IWQPC_SNDWL2_SHIFT 32
+#define I40IWQPC_SNDWL2_MASK (0xffffffffULL << I40IWQPC_SNDWL2_SHIFT)
+
+#define I40IWQPC_ERR_RQ_IDX_SHIFT 32
+#define I40IWQPC_ERR_RQ_IDX_MASK  (0x3fffULL << I40IWQPC_ERR_RQ_IDX_SHIFT)
+
+#define I40IWQPC_MAXSNDWND_SHIFT 0
+#define I40IWQPC_MAXSNDWND_MASK (0xffffffffUL << I40IWQPC_MAXSNDWND_SHIFT)
+
+#define I40IWQPC_REXMIT_THRESH_SHIFT 48
+#define I40IWQPC_REXMIT_THRESH_MASK (0x3fULL << I40IWQPC_REXMIT_THRESH_SHIFT)
+
+#define I40IWQPC_TXCQNUM_SHIFT 0
+#define I40IWQPC_TXCQNUM_MASK (0x1ffffUL << I40IWQPC_TXCQNUM_SHIFT)
+
+#define I40IWQPC_RXCQNUM_SHIFT 32
+#define I40IWQPC_RXCQNUM_MASK (0x1ffffULL << I40IWQPC_RXCQNUM_SHIFT)
+
+#define I40IWQPC_Q2ADDR_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPC_Q2ADDR_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPC_LASTBYTESENT_SHIFT 0
+#define I40IWQPC_LASTBYTESENT_MASK (0xffUL << I40IWQPC_LASTBYTESENT_SHIFT)
+
+#define I40IWQPC_SRQID_SHIFT 32
+#define I40IWQPC_SRQID_MASK (0xffULL << I40IWQPC_SRQID_SHIFT)
+
+#define I40IWQPC_ORDSIZE_SHIFT 0
+#define I40IWQPC_ORDSIZE_MASK (0x7fUL << I40IWQPC_ORDSIZE_SHIFT)
+
+#define I40IWQPC_IRDSIZE_SHIFT 16
+#define I40IWQPC_IRDSIZE_MASK (0x3UL << I40IWQPC_IRDSIZE_SHIFT)
+
+#define I40IWQPC_WRRDRSPOK_SHIFT 20
+#define I40IWQPC_WRRDRSPOK_MASK (1UL << I40IWQPC_WRRDRSPOK_SHIFT)
+
+#define I40IWQPC_RDOK_SHIFT 21
+#define I40IWQPC_RDOK_MASK (1UL << I40IWQPC_RDOK_SHIFT)
+
+#define I40IWQPC_SNDMARKERS_SHIFT 22
+#define I40IWQPC_SNDMARKERS_MASK (1UL << I40IWQPC_SNDMARKERS_SHIFT)
+
+#define I40IWQPC_BINDEN_SHIFT 23
+#define I40IWQPC_BINDEN_MASK (1UL << I40IWQPC_BINDEN_SHIFT)
+
+#define I40IWQPC_FASTREGEN_SHIFT 24
+#define I40IWQPC_FASTREGEN_MASK (1UL << I40IWQPC_FASTREGEN_SHIFT)
+
+#define I40IWQPC_PRIVEN_SHIFT 25
+#define I40IWQPC_PRIVEN_MASK (1UL << I40IWQPC_PRIVEN_SHIFT)
+
+#define I40IWQPC_LSMMPRESENT_SHIFT 26
+#define I40IWQPC_LSMMPRESENT_MASK (1UL << I40IWQPC_LSMMPRESENT_SHIFT)
+
+#define I40IWQPC_ADJUSTFORLSMM_SHIFT 27
+#define I40IWQPC_ADJUSTFORLSMM_MASK (1UL << I40IWQPC_ADJUSTFORLSMM_SHIFT)
+
+#define I40IWQPC_IWARPMODE_SHIFT 28
+#define I40IWQPC_IWARPMODE_MASK (1UL << I40IWQPC_IWARPMODE_SHIFT)
+
+#define I40IWQPC_RCVMARKERS_SHIFT 29
+#define I40IWQPC_RCVMARKERS_MASK (1UL << I40IWQPC_RCVMARKERS_SHIFT)
+
+#define I40IWQPC_ALIGNHDRS_SHIFT 30
+#define I40IWQPC_ALIGNHDRS_MASK (1UL << I40IWQPC_ALIGNHDRS_SHIFT)
+
+#define I40IWQPC_RCVNOMPACRC_SHIFT 31
+#define I40IWQPC_RCVNOMPACRC_MASK (1UL << I40IWQPC_RCVNOMPACRC_SHIFT)
+
+#define I40IWQPC_RCVMARKOFFSET_SHIFT 33
+#define I40IWQPC_RCVMARKOFFSET_MASK (0x1ffULL << I40IWQPC_RCVMARKOFFSET_SHIFT)
+
+#define I40IWQPC_SNDMARKOFFSET_SHIFT 48
+#define I40IWQPC_SNDMARKOFFSET_MASK (0x1ffULL << I40IWQPC_SNDMARKOFFSET_SHIFT)
+
+#define I40IWQPC_QPCOMPCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPC_QPCOMPCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPC_SQTPHVAL_SHIFT 0
+#define I40IWQPC_SQTPHVAL_MASK (0xffUL << I40IWQPC_SQTPHVAL_SHIFT)
+
+#define I40IWQPC_RQTPHVAL_SHIFT 8
+#define I40IWQPC_RQTPHVAL_MASK (0xffUL << I40IWQPC_RQTPHVAL_SHIFT)
+
+#define I40IWQPC_QSHANDLE_SHIFT 16
+#define I40IWQPC_QSHANDLE_MASK (0x3ffUL << I40IWQPC_QSHANDLE_SHIFT)
+
+#define I40IWQPC_EXCEPTION_LAN_QUEUE_SHIFT 32
+#define I40IWQPC_EXCEPTION_LAN_QUEUE_MASK (0xfffULL <<  \
+					   I40IWQPC_EXCEPTION_LAN_QUEUE_SHIFT)
+
+#define I40IWQPC_LOCAL_IPADDR3_SHIFT 0
+#define I40IWQPC_LOCAL_IPADDR3_MASK \
+	(0xffffffffUL << I40IWQPC_LOCAL_IPADDR3_SHIFT)
+
+#define I40IWQPC_LOCAL_IPADDR2_SHIFT 32
+#define I40IWQPC_LOCAL_IPADDR2_MASK     \
+	(0xffffffffULL << I40IWQPC_LOCAL_IPADDR2_SHIFT)
+
+#define I40IWQPC_LOCAL_IPADDR1_SHIFT 0
+#define I40IWQPC_LOCAL_IPADDR1_MASK     \
+	(0xffffffffUL << I40IWQPC_LOCAL_IPADDR1_SHIFT)
+
+#define I40IWQPC_LOCAL_IPADDR0_SHIFT 32
+#define I40IWQPC_LOCAL_IPADDR0_MASK     \
+	(0xffffffffULL << I40IWQPC_LOCAL_IPADDR0_SHIFT)
+
+/* wqe size considering 32 bytes per wqe*/
+#define I40IWQP_SW_MIN_WQSIZE 4		/* 128 bytes */
+#define I40IWQP_SW_MAX_WQSIZE 16384	/* 524288 bytes */
+
+#define I40IWQP_OP_RDMA_WRITE 0
+#define I40IWQP_OP_RDMA_READ 1
+#define I40IWQP_OP_RDMA_SEND 3
+#define I40IWQP_OP_RDMA_SEND_INV 4
+#define I40IWQP_OP_RDMA_SEND_SOL_EVENT 5
+#define I40IWQP_OP_RDMA_SEND_SOL_EVENT_INV 6
+#define I40IWQP_OP_BIND_MW 8
+#define I40IWQP_OP_FAST_REGISTER 9
+#define I40IWQP_OP_LOCAL_INVALIDATE 10
+#define I40IWQP_OP_RDMA_READ_LOC_INV 11
+#define I40IWQP_OP_NOP 12
+
+#define I40IW_RSVD_SHIFT        41
+#define I40IW_RSVD_MASK (0x7fffULL << I40IW_RSVD_SHIFT)
+
+/* iwarp QP SQ WQE common fields */
+#define I40IWQPSQ_OPCODE_SHIFT 32
+#define I40IWQPSQ_OPCODE_MASK (0x3fULL << I40IWQPSQ_OPCODE_SHIFT)
+
+#define I40IWQPSQ_ADDFRAGCNT_SHIFT 38
+#define I40IWQPSQ_ADDFRAGCNT_MASK (0x7ULL << I40IWQPSQ_ADDFRAGCNT_SHIFT)
+
+#define I40IWQPSQ_PUSHWQE_SHIFT 56
+#define I40IWQPSQ_PUSHWQE_MASK (1ULL << I40IWQPSQ_PUSHWQE_SHIFT)
+
+#define I40IWQPSQ_STREAMMODE_SHIFT 58
+#define I40IWQPSQ_STREAMMODE_MASK (1ULL << I40IWQPSQ_STREAMMODE_SHIFT)
+
+#define I40IWQPSQ_WAITFORRCVPDU_SHIFT 59
+#define I40IWQPSQ_WAITFORRCVPDU_MASK (1ULL << I40IWQPSQ_WAITFORRCVPDU_SHIFT)
+
+#define I40IWQPSQ_READFENCE_SHIFT 60
+#define I40IWQPSQ_READFENCE_MASK (1ULL << I40IWQPSQ_READFENCE_SHIFT)
+
+#define I40IWQPSQ_LOCALFENCE_SHIFT 61
+#define I40IWQPSQ_LOCALFENCE_MASK (1ULL << I40IWQPSQ_LOCALFENCE_SHIFT)
+
+#define I40IWQPSQ_SIGCOMPL_SHIFT 62
+#define I40IWQPSQ_SIGCOMPL_MASK (1ULL << I40IWQPSQ_SIGCOMPL_SHIFT)
+
+#define I40IWQPSQ_VALID_SHIFT 63
+#define I40IWQPSQ_VALID_MASK (1ULL << I40IWQPSQ_VALID_SHIFT)
+
+#define I40IWQPSQ_FRAG_TO_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPSQ_FRAG_TO_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPSQ_FRAG_LEN_SHIFT 0
+#define I40IWQPSQ_FRAG_LEN_MASK (0xffffffffUL << I40IWQPSQ_FRAG_LEN_SHIFT)
+
+#define I40IWQPSQ_FRAG_STAG_SHIFT 32
+#define I40IWQPSQ_FRAG_STAG_MASK (0xffffffffULL << I40IWQPSQ_FRAG_STAG_SHIFT)
+
+#define I40IWQPSQ_REMSTAGINV_SHIFT 0
+#define I40IWQPSQ_REMSTAGINV_MASK (0xffffffffUL << I40IWQPSQ_REMSTAGINV_SHIFT)
+
+#define I40IWQPSQ_INLINEDATAFLAG_SHIFT 57
+#define I40IWQPSQ_INLINEDATAFLAG_MASK (1ULL << I40IWQPSQ_INLINEDATAFLAG_SHIFT)
+
+#define I40IWQPSQ_INLINEDATALEN_SHIFT 48
+#define I40IWQPSQ_INLINEDATALEN_MASK    \
+	(0x7fULL << I40IWQPSQ_INLINEDATALEN_SHIFT)
+
+/* iwarp send with push mode */
+#define I40IWQPSQ_WQDESCIDX_SHIFT 0
+#define I40IWQPSQ_WQDESCIDX_MASK (0x3fffUL << I40IWQPSQ_WQDESCIDX_SHIFT)
+
+/* rdma write */
+#define I40IWQPSQ_REMSTAG_SHIFT 0
+#define I40IWQPSQ_REMSTAG_MASK (0xffffffffUL << I40IWQPSQ_REMSTAG_SHIFT)
+
+#define I40IWQPSQ_REMTO_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPSQ_REMTO_MASK I40IW_CQPHC_QPCTX_MASK
+
+/* memory window */
+#define I40IWQPSQ_STAGRIGHTS_SHIFT 48
+#define I40IWQPSQ_STAGRIGHTS_MASK (0x1fULL << I40IWQPSQ_STAGRIGHTS_SHIFT)
+
+#define I40IWQPSQ_VABASEDTO_SHIFT 53
+#define I40IWQPSQ_VABASEDTO_MASK (1ULL << I40IWQPSQ_VABASEDTO_SHIFT)
+
+#define I40IWQPSQ_MWLEN_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPSQ_MWLEN_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPSQ_PARENTMRSTAG_SHIFT 0
+#define I40IWQPSQ_PARENTMRSTAG_MASK \
+	(0xffffffffUL << I40IWQPSQ_PARENTMRSTAG_SHIFT)
+
+#define I40IWQPSQ_MWSTAG_SHIFT 32
+#define I40IWQPSQ_MWSTAG_MASK (0xffffffffULL << I40IWQPSQ_MWSTAG_SHIFT)
+
+#define I40IWQPSQ_BASEVA_TO_FBO_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPSQ_BASEVA_TO_FBO_MASK I40IW_CQPHC_QPCTX_MASK
+
+/* Local Invalidate */
+#define I40IWQPSQ_LOCSTAG_SHIFT 32
+#define I40IWQPSQ_LOCSTAG_MASK (0xffffffffULL << I40IWQPSQ_LOCSTAG_SHIFT)
+
+/* Fast Register */
+#define I40IWQPSQ_STAGKEY_SHIFT 0
+#define I40IWQPSQ_STAGKEY_MASK (0xffUL << I40IWQPSQ_STAGKEY_SHIFT)
+
+#define I40IWQPSQ_STAGINDEX_SHIFT 8
+#define I40IWQPSQ_STAGINDEX_MASK (0xffffffUL << I40IWQPSQ_STAGINDEX_SHIFT)
+
+#define I40IWQPSQ_COPYHOSTPBLS_SHIFT 43
+#define I40IWQPSQ_COPYHOSTPBLS_MASK (1ULL << I40IWQPSQ_COPYHOSTPBLS_SHIFT)
+
+#define I40IWQPSQ_LPBLSIZE_SHIFT 44
+#define I40IWQPSQ_LPBLSIZE_MASK (3ULL << I40IWQPSQ_LPBLSIZE_SHIFT)
+
+#define I40IWQPSQ_HPAGESIZE_SHIFT 46
+#define I40IWQPSQ_HPAGESIZE_MASK (3ULL << I40IWQPSQ_HPAGESIZE_SHIFT)
+
+#define I40IWQPSQ_STAGLEN_SHIFT 0
+#define I40IWQPSQ_STAGLEN_MASK (0x1ffffffffffULL << I40IWQPSQ_STAGLEN_SHIFT)
+
+#define I40IWQPSQ_FIRSTPMPBLIDXLO_SHIFT 48
+#define I40IWQPSQ_FIRSTPMPBLIDXLO_MASK  \
+	(0xffffULL << I40IWQPSQ_FIRSTPMPBLIDXLO_SHIFT)
+
+#define I40IWQPSQ_FIRSTPMPBLIDXHI_SHIFT 0
+#define I40IWQPSQ_FIRSTPMPBLIDXHI_MASK  \
+	(0xfffUL << I40IWQPSQ_FIRSTPMPBLIDXHI_SHIFT)
+
+#define I40IWQPSQ_PBLADDR_SHIFT 12
+#define I40IWQPSQ_PBLADDR_MASK (0xfffffffffffffULL << I40IWQPSQ_PBLADDR_SHIFT)
+
+/*  iwarp QP RQ WQE common fields */
+#define I40IWQPRQ_ADDFRAGCNT_SHIFT I40IWQPSQ_ADDFRAGCNT_SHIFT
+#define I40IWQPRQ_ADDFRAGCNT_MASK I40IWQPSQ_ADDFRAGCNT_MASK
+
+#define I40IWQPRQ_VALID_SHIFT I40IWQPSQ_VALID_SHIFT
+#define I40IWQPRQ_VALID_MASK I40IWQPSQ_VALID_MASK
+
+#define I40IWQPRQ_COMPLCTX_SHIFT I40IW_CQPHC_QPCTX_SHIFT
+#define I40IWQPRQ_COMPLCTX_MASK I40IW_CQPHC_QPCTX_MASK
+
+#define I40IWQPRQ_FRAG_LEN_SHIFT I40IWQPSQ_FRAG_LEN_SHIFT
+#define I40IWQPRQ_FRAG_LEN_MASK I40IWQPSQ_FRAG_LEN_MASK
+
+#define I40IWQPRQ_STAG_SHIFT I40IWQPSQ_FRAG_STAG_SHIFT
+#define I40IWQPRQ_STAG_MASK I40IWQPSQ_FRAG_STAG_MASK
+
+#define I40IWQPRQ_TO_SHIFT I40IWQPSQ_FRAG_TO_SHIFT
+#define I40IWQPRQ_TO_MASK I40IWQPSQ_FRAG_TO_MASK
+
+/* Query FPM CQP buf */
+#define I40IW_QUERY_FPM_MAX_QPS_SHIFT 0
+#define I40IW_QUERY_FPM_MAX_QPS_MASK               \
+	(0x7ffffUL << I40IW_QUERY_FPM_MAX_QPS_SHIFT)
+
+#define I40IW_QUERY_FPM_MAX_CQS_SHIFT 0
+#define I40IW_QUERY_FPM_MAX_CQS_MASK               \
+	(0x3ffffUL << I40IW_QUERY_FPM_MAX_CQS_SHIFT)
+
+#define I40IW_QUERY_FPM_FIRST_PE_SD_INDEX_SHIFT 0
+#define I40IW_QUERY_FPM_FIRST_PE_SD_INDEX_MASK  \
+	(0x3fffUL << I40IW_QUERY_FPM_FIRST_PE_SD_INDEX_SHIFT)
+
+#define I40IW_QUERY_FPM_MAX_PE_SDS_SHIFT 32
+#define I40IW_QUERY_FPM_MAX_PE_SDS_MASK \
+	(0x3fffULL << I40IW_QUERY_FPM_MAX_PE_SDS_SHIFT)
+
+#define I40IW_QUERY_FPM_MAX_QPS_SHIFT 0
+#define I40IW_QUERY_FPM_MAX_QPS_MASK    \
+	(0x7ffffUL << I40IW_QUERY_FPM_MAX_QPS_SHIFT)
+
+#define I40IW_QUERY_FPM_MAX_CQS_SHIFT 0
+#define I40IW_QUERY_FPM_MAX_CQS_MASK    \
+	(0x3ffffUL << I40IW_QUERY_FPM_MAX_CQS_SHIFT)
+
+#define I40IW_QUERY_FPM_MAX_CEQS_SHIFT 0
+#define I40IW_QUERY_FPM_MAX_CEQS_MASK   \
+	(0xffUL << I40IW_QUERY_FPM_MAX_CEQS_SHIFT)
+
+#define I40IW_QUERY_FPM_XFBLOCKSIZE_SHIFT 32
+#define I40IW_QUERY_FPM_XFBLOCKSIZE_MASK        \
+	(0xffffffffULL << I40IW_QUERY_FPM_XFBLOCKSIZE_SHIFT)
+
+#define I40IW_QUERY_FPM_Q1BLOCKSIZE_SHIFT 32
+#define I40IW_QUERY_FPM_Q1BLOCKSIZE_MASK        \
+	(0xffffffffULL << I40IW_QUERY_FPM_Q1BLOCKSIZE_SHIFT)
+
+#define I40IW_QUERY_FPM_HTMULTIPLIER_SHIFT 16
+#define I40IW_QUERY_FPM_HTMULTIPLIER_MASK       \
+	(0xfUL << I40IW_QUERY_FPM_HTMULTIPLIER_SHIFT)
+
+#define I40IW_QUERY_FPM_TIMERBUCKET_SHIFT 32
+#define I40IW_QUERY_FPM_TIMERBUCKET_MASK        \
+	(0xffFFULL << I40IW_QUERY_FPM_TIMERBUCKET_SHIFT)
+
+/* Static HMC pages allocated buf */
+#define I40IW_SHMC_PAGE_ALLOCATED_HMC_FN_ID_SHIFT 0
+#define I40IW_SHMC_PAGE_ALLOCATED_HMC_FN_ID_MASK        \
+	(0x3fUL << I40IW_SHMC_PAGE_ALLOCATED_HMC_FN_ID_SHIFT)
+
+#define I40IW_HW_PAGE_SIZE	4096
+#define I40IW_DONE_COUNT	1000
+#define I40IW_SLEEP_COUNT	10
+
+enum {
+	I40IW_QUEUES_ALIGNMENT_MASK =		(128 - 1),
+	I40IW_AEQ_ALIGNMENT_MASK =		(256 - 1),
+	I40IW_Q2_ALIGNMENT_MASK =		(256 - 1),
+	I40IW_CEQ_ALIGNMENT_MASK =		(256 - 1),
+	I40IW_CQ0_ALIGNMENT_MASK =		(256 - 1),
+	I40IW_HOST_CTX_ALIGNMENT_MASK =		(4 - 1),
+	I40IW_SHADOWAREA_MASK =			(128 - 1),
+	I40IW_FPM_QUERY_BUF_ALIGNMENT_MASK =	0,
+	I40IW_FPM_COMMIT_BUF_ALIGNMENT_MASK =	0
+};
+
+enum i40iw_alignment {
+	I40IW_CQP_ALIGNMENT =		0x200,
+	I40IW_AEQ_ALIGNMENT =		0x100,
+	I40IW_CEQ_ALIGNMENT =		0x100,
+	I40IW_CQ0_ALIGNMENT =		0x100,
+	I40IW_SD_BUF_ALIGNMENT =	0x100
+};
+
+#define I40IW_QP_WQE_MIN_SIZE	32
+#define I40IW_QP_WQE_MAX_SIZE	128
+
+#define I40IW_CQE_QTYPE_RQ 0
+#define I40IW_CQE_QTYPE_SQ 1
+
+#define I40IW_RING_INIT(_ring, _size) \
+	{ \
+		(_ring).head = 0; \
+		(_ring).tail = 0; \
+		(_ring).size = (_size); \
+	}
+#define I40IW_RING_GETSIZE(_ring) ((_ring).size)
+#define I40IW_RING_GETCURRENT_HEAD(_ring) ((_ring).head)
+#define I40IW_RING_GETCURRENT_TAIL(_ring) ((_ring).tail)
+
+#define I40IW_RING_MOVE_HEAD(_ring, _retcode) \
+	{ \
+		register u32 size; \
+		size = (_ring).size;  \
+		if (!I40IW_RING_FULL_ERR(_ring)) { \
+			(_ring).head = ((_ring).head + 1) % size; \
+			(_retcode) = 0; \
+		} else { \
+			(_retcode) = I40IW_ERR_RING_FULL; \
+		} \
+	}
+
+#define I40IW_RING_MOVE_HEAD_BY_COUNT(_ring, _count, _retcode) \
+	{ \
+		register u32 size; \
+		size = (_ring).size; \
+		if ((I40IW_RING_WORK_AVAILABLE(_ring) + (_count)) < size) { \
+			(_ring).head = ((_ring).head + (_count)) % size; \
+			(_retcode) = 0; \
+		} else { \
+			(_retcode) = I40IW_ERR_RING_FULL; \
+		} \
+	}
+
+#define I40IW_RING_MOVE_TAIL(_ring) \
+	(_ring).tail = ((_ring).tail + 1) % (_ring).size
+
+#define I40IW_RING_MOVE_TAIL_BY_COUNT(_ring, _count) \
+	(_ring).tail = ((_ring).tail + (_count)) % (_ring).size
+
+#define I40IW_RING_SET_TAIL(_ring, _pos) \
+	(_ring).tail = (_pos) % (_ring).size
+
+#define I40IW_RING_FULL_ERR(_ring) \
+	( \
+		(I40IW_RING_WORK_AVAILABLE(_ring) == ((_ring).size - 1))  \
+	)
+
+#define I40IW_ERR_RING_FULL2(_ring) \
+	( \
+		(I40IW_RING_WORK_AVAILABLE(_ring) == ((_ring).size - 2))  \
+	)
+
+#define I40IW_ERR_RING_FULL3(_ring) \
+	( \
+		(I40IW_RING_WORK_AVAILABLE(_ring) == ((_ring).size - 3))  \
+	)
+
+#define I40IW_RING_MORE_WORK(_ring) \
+	( \
+		(I40IW_RING_WORK_AVAILABLE(_ring) != 0) \
+	)
+
+#define I40IW_RING_WORK_AVAILABLE(_ring) \
+	( \
+		(((_ring).head + (_ring).size - (_ring).tail) % (_ring).size) \
+	)
+
+#define I40IW_RING_GET_WQES_AVAILABLE(_ring) \
+	( \
+		((_ring).size - I40IW_RING_WORK_AVAILABLE(_ring) - 1) \
+	)
+
+#define I40IW_ATOMIC_RING_MOVE_HEAD(_ring, index, _retcode) \
+	{ \
+		index = I40IW_RING_GETCURRENT_HEAD(_ring); \
+		I40IW_RING_MOVE_HEAD(_ring, _retcode); \
+	}
+
+/* Async Events codes */
+#define I40IW_AE_AMP_UNALLOCATED_STAG                                   0x0102
+#define I40IW_AE_AMP_INVALID_STAG                                       0x0103
+#define I40IW_AE_AMP_BAD_QP                                             0x0104
+#define I40IW_AE_AMP_BAD_PD                                             0x0105
+#define I40IW_AE_AMP_BAD_STAG_KEY                                       0x0106
+#define I40IW_AE_AMP_BAD_STAG_INDEX                                     0x0107
+#define I40IW_AE_AMP_BOUNDS_VIOLATION                                   0x0108
+#define I40IW_AE_AMP_RIGHTS_VIOLATION                                   0x0109
+#define I40IW_AE_AMP_TO_WRAP                                            0x010a
+#define I40IW_AE_AMP_FASTREG_SHARED                                     0x010b
+#define I40IW_AE_AMP_FASTREG_VALID_STAG                                 0x010c
+#define I40IW_AE_AMP_FASTREG_MW_STAG                                    0x010d
+#define I40IW_AE_AMP_FASTREG_INVALID_RIGHTS                             0x010e
+#define I40IW_AE_AMP_FASTREG_PBL_TABLE_OVERFLOW                         0x010f
+#define I40IW_AE_AMP_FASTREG_INVALID_LENGTH                             0x0110
+#define I40IW_AE_AMP_INVALIDATE_SHARED                                  0x0111
+#define I40IW_AE_AMP_INVALIDATE_NO_REMOTE_ACCESS_RIGHTS                 0x0112
+#define I40IW_AE_AMP_INVALIDATE_MR_WITH_BOUND_WINDOWS                   0x0113
+#define I40IW_AE_AMP_MWBIND_VALID_STAG                                  0x0114
+#define I40IW_AE_AMP_MWBIND_OF_MR_STAG                                  0x0115
+#define I40IW_AE_AMP_MWBIND_TO_ZERO_BASED_STAG                          0x0116
+#define I40IW_AE_AMP_MWBIND_TO_MW_STAG                                  0x0117
+#define I40IW_AE_AMP_MWBIND_INVALID_RIGHTS                              0x0118
+#define I40IW_AE_AMP_MWBIND_INVALID_BOUNDS                              0x0119
+#define I40IW_AE_AMP_MWBIND_TO_INVALID_PARENT                           0x011a
+#define I40IW_AE_AMP_MWBIND_BIND_DISABLED                               0x011b
+#define I40IW_AE_AMP_WQE_INVALID_PARAMETER                              0x0130
+#define I40IW_AE_BAD_CLOSE                                              0x0201
+#define I40IW_AE_RDMAP_ROE_BAD_LLP_CLOSE                                0x0202
+#define I40IW_AE_CQ_OPERATION_ERROR                                     0x0203
+#define I40IW_AE_PRIV_OPERATION_DENIED                                  0x011c
+#define I40IW_AE_RDMA_READ_WHILE_ORD_ZERO                               0x0205
+#define I40IW_AE_STAG_ZERO_INVALID                                      0x0206
+#define I40IW_AE_IB_RREQ_AND_Q1_FULL                                    0x0207
+#define I40IW_AE_SRQ_LIMIT                                              0x0209
+#define I40IW_AE_WQE_UNEXPECTED_OPCODE                                  0x020a
+#define I40IW_AE_WQE_INVALID_PARAMETER                                  0x020b
+#define I40IW_AE_WQE_LSMM_TOO_LONG                                      0x0220
+#define I40IW_AE_DDP_INVALID_MSN_GAP_IN_MSN                             0x0301
+#define I40IW_AE_DDP_INVALID_MSN_RANGE_IS_NOT_VALID                     0x0302
+#define I40IW_AE_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER      0x0303
+#define I40IW_AE_DDP_UBE_INVALID_DDP_VERSION                            0x0304
+#define I40IW_AE_DDP_UBE_INVALID_MO                                     0x0305
+#define I40IW_AE_DDP_UBE_INVALID_MSN_NO_BUFFER_AVAILABLE                0x0306
+#define I40IW_AE_DDP_UBE_INVALID_QN                                     0x0307
+#define I40IW_AE_DDP_NO_L_BIT                                           0x0308
+#define I40IW_AE_RDMAP_ROE_INVALID_RDMAP_VERSION                        0x0311
+#define I40IW_AE_RDMAP_ROE_UNEXPECTED_OPCODE                            0x0312
+#define I40IW_AE_ROE_INVALID_RDMA_READ_REQUEST                          0x0313
+#define I40IW_AE_ROE_INVALID_RDMA_WRITE_OR_READ_RESP                    0x0314
+#define I40IW_AE_INVALID_ARP_ENTRY                                      0x0401
+#define I40IW_AE_INVALID_TCP_OPTION_RCVD                                0x0402
+#define I40IW_AE_STALE_ARP_ENTRY                                        0x0403
+#define I40IW_AE_INVALID_WQE_LENGTH                                     0x0404
+#define I40IW_AE_INVALID_MAC_ENTRY                                      0x0405
+#define I40IW_AE_LLP_CLOSE_COMPLETE                                     0x0501
+#define I40IW_AE_LLP_CONNECTION_RESET                                   0x0502
+#define I40IW_AE_LLP_FIN_RECEIVED                                       0x0503
+#define I40IW_AE_LLP_RECEIVED_MARKER_AND_LENGTH_FIELDS_DONT_MATCH       0x0504
+#define I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR                             0x0505
+#define I40IW_AE_LLP_SEGMENT_TOO_LARGE                                  0x0506
+#define I40IW_AE_LLP_SEGMENT_TOO_SMALL                                  0x0507
+#define I40IW_AE_LLP_SYN_RECEIVED                                       0x0508
+#define I40IW_AE_LLP_TERMINATE_RECEIVED                                 0x0509
+#define I40IW_AE_LLP_TOO_MANY_RETRIES                                   0x050a
+#define I40IW_AE_LLP_TOO_MANY_KEEPALIVE_RETRIES                         0x050b
+#define I40IW_AE_LLP_DOUBT_REACHABILITY                                 0x050c
+#define I40IW_AE_LLP_RX_VLAN_MISMATCH                                   0x050d
+#define I40IW_AE_RESOURCE_EXHAUSTION                                    0x0520
+#define I40IW_AE_RESET_SENT                                             0x0601
+#define I40IW_AE_TERMINATE_SENT                                         0x0602
+#define I40IW_AE_RESET_NOT_SENT                                         0x0603
+#define I40IW_AE_LCE_QP_CATASTROPHIC                                    0x0700
+#define I40IW_AE_LCE_FUNCTION_CATASTROPHIC                              0x0701
+#define I40IW_AE_LCE_CQ_CATASTROPHIC                                    0x0702
+#define I40IW_AE_UDA_XMIT_FRAG_SEQ                                      0x0800
+#define I40IW_AE_UDA_XMIT_DGRAM_TOO_LONG                                0x0801
+#define I40IW_AE_UDA_XMIT_IPADDR_MISMATCH                               0x0802
+#define I40IW_AE_QP_SUSPEND_COMPLETE                                    0x0900
+
+#define OP_DELETE_LOCAL_MAC_IPADDR_ENTRY        1
+#define OP_CEQ_DESTROY                          2
+#define OP_AEQ_DESTROY                          3
+#define OP_DELETE_ARP_CACHE_ENTRY               4
+#define OP_MANAGE_APBVT_ENTRY                   5
+#define OP_CEQ_CREATE                           6
+#define OP_AEQ_CREATE                           7
+#define OP_ALLOC_LOCAL_MAC_IPADDR_ENTRY         8
+#define OP_ADD_LOCAL_MAC_IPADDR_ENTRY           9
+#define OP_MANAGE_QHASH_TABLE_ENTRY             10
+#define OP_QP_MODIFY                            11
+#define OP_QP_UPLOAD_CONTEXT                    12
+#define OP_CQ_CREATE                            13
+#define OP_CQ_DESTROY                           14
+#define OP_QP_CREATE                            15
+#define OP_QP_DESTROY                           16
+#define OP_ALLOC_STAG                           17
+#define OP_MR_REG_NON_SHARED                    18
+#define OP_DEALLOC_STAG                         19
+#define OP_MW_ALLOC                             20
+#define OP_QP_FLUSH_WQES                        21
+#define OP_ADD_ARP_CACHE_ENTRY                  22
+#define OP_MANAGE_PUSH_PAGE                     23
+#define OP_UPDATE_PE_SDS                        24
+#define OP_MANAGE_HMC_PM_FUNC_TABLE             25
+#define OP_SUSPEND                              26
+#define OP_RESUME                               27
+#define OP_MANAGE_VF_PBLE_BP                    28
+#define OP_QUERY_FPM_VALUES                     29
+#define OP_COMMIT_FPM_VALUES                    30
+#define OP_SIZE_CQP_STAT_ARRAY                  31
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_hmc.c b/drivers/infiniband/hw/i40iw/i40iw_hmc.c
new file mode 100644
index 000000000000..5484cbf55f0f
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_hmc.c
@@ -0,0 +1,821 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_status.h"
+#include "i40iw_hmc.h"
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_vf.h"
+#include "i40iw_virtchnl.h"
+
+/**
+ * i40iw_find_sd_index_limit - finds segment descriptor index limit
+ * @hmc_info: pointer to the HMC configuration information structure
+ * @type: type of HMC resources we're searching
+ * @index: starting index for the object
+ * @cnt: number of objects we're trying to create
+ * @sd_idx: pointer to return index of the segment descriptor in question
+ * @sd_limit: pointer to return the maximum number of segment descriptors
+ *
+ * This function calculates the segment descriptor index and index limit
+ * for the resource defined by i40iw_hmc_rsrc_type.
+ */
+
+static inline void i40iw_find_sd_index_limit(struct i40iw_hmc_info *hmc_info,
+					     u32 type,
+					     u32 idx,
+					     u32 cnt,
+					     u32 *sd_idx,
+					     u32 *sd_limit)
+{
+	u64 fpm_addr, fpm_limit;
+
+	fpm_addr = hmc_info->hmc_obj[(type)].base +
+			hmc_info->hmc_obj[type].size * idx;
+	fpm_limit = fpm_addr + hmc_info->hmc_obj[type].size * cnt;
+	*sd_idx = (u32)(fpm_addr / I40IW_HMC_DIRECT_BP_SIZE);
+	*sd_limit = (u32)((fpm_limit - 1) / I40IW_HMC_DIRECT_BP_SIZE);
+	*sd_limit += 1;
+}
+
+/**
+ * i40iw_find_pd_index_limit - finds page descriptor index limit
+ * @hmc_info: pointer to the HMC configuration information struct
+ * @type: HMC resource type we're examining
+ * @idx: starting index for the object
+ * @cnt: number of objects we're trying to create
+ * @pd_index: pointer to return page descriptor index
+ * @pd_limit: pointer to return page descriptor index limit
+ *
+ * Calculates the page descriptor index and index limit for the resource
+ * defined by i40iw_hmc_rsrc_type.
+ */
+
+static inline void i40iw_find_pd_index_limit(struct i40iw_hmc_info *hmc_info,
+					     u32 type,
+					     u32 idx,
+					     u32 cnt,
+					     u32 *pd_idx,
+					     u32 *pd_limit)
+{
+	u64 fpm_adr, fpm_limit;
+
+	fpm_adr = hmc_info->hmc_obj[type].base +
+			hmc_info->hmc_obj[type].size * idx;
+	fpm_limit = fpm_adr + (hmc_info)->hmc_obj[(type)].size * (cnt);
+	*(pd_idx) = (u32)(fpm_adr / I40IW_HMC_PAGED_BP_SIZE);
+	*(pd_limit) = (u32)((fpm_limit - 1) / I40IW_HMC_PAGED_BP_SIZE);
+	*(pd_limit) += 1;
+}
+
+/**
+ * i40iw_set_sd_entry - setup entry for sd programming
+ * @pa: physical addr
+ * @idx: sd index
+ * @type: paged or direct sd
+ * @entry: sd entry ptr
+ */
+static inline void i40iw_set_sd_entry(u64 pa,
+				      u32 idx,
+				      enum i40iw_sd_entry_type type,
+				      struct update_sd_entry *entry)
+{
+	entry->data = pa | (I40IW_HMC_MAX_BP_COUNT << I40E_PFHMC_SDDATALOW_PMSDBPCOUNT_SHIFT) |
+			(((type == I40IW_SD_TYPE_PAGED) ? 0 : 1) <<
+				I40E_PFHMC_SDDATALOW_PMSDTYPE_SHIFT) |
+			(1 << I40E_PFHMC_SDDATALOW_PMSDVALID_SHIFT);
+	entry->cmd = (idx | (1 << I40E_PFHMC_SDCMD_PMSDWR_SHIFT) | (1 << 15));
+}
+
+/**
+ * i40iw_clr_sd_entry - setup entry for sd clear
+ * @idx: sd index
+ * @type: paged or direct sd
+ * @entry: sd entry ptr
+ */
+static inline void i40iw_clr_sd_entry(u32 idx, enum i40iw_sd_entry_type type,
+				      struct update_sd_entry *entry)
+{
+	entry->data = (I40IW_HMC_MAX_BP_COUNT <<
+			I40E_PFHMC_SDDATALOW_PMSDBPCOUNT_SHIFT) |
+			(((type == I40IW_SD_TYPE_PAGED) ? 0 : 1) <<
+				I40E_PFHMC_SDDATALOW_PMSDTYPE_SHIFT);
+	entry->cmd = (idx | (1 << I40E_PFHMC_SDCMD_PMSDWR_SHIFT) | (1 << 15));
+}
+
+/**
+ * i40iw_hmc_sd_one - setup 1 sd entry for cqp
+ * @dev: pointer to the device structure
+ * @hmc_fn_id: hmc's function id
+ * @pa: physical addr
+ * @sd_idx: sd index
+ * @type: paged or direct sd
+ * @setsd: flag to set or clear sd
+ */
+enum i40iw_status_code i40iw_hmc_sd_one(struct i40iw_sc_dev *dev,
+					u8 hmc_fn_id,
+					u64 pa, u32 sd_idx,
+					enum i40iw_sd_entry_type type,
+					bool setsd)
+{
+	struct i40iw_update_sds_info sdinfo;
+
+	sdinfo.cnt = 1;
+	sdinfo.hmc_fn_id = hmc_fn_id;
+	if (setsd)
+		i40iw_set_sd_entry(pa, sd_idx, type, sdinfo.entry);
+	else
+		i40iw_clr_sd_entry(sd_idx, type, sdinfo.entry);
+
+	return dev->cqp->process_cqp_sds(dev, &sdinfo);
+}
+
+/**
+ * i40iw_hmc_sd_grp - setup group od sd entries for cqp
+ * @dev: pointer to the device structure
+ * @hmc_info: pointer to the HMC configuration information struct
+ * @sd_index: sd index
+ * @sd_cnt: number of sd entries
+ * @setsd: flag to set or clear sd
+ */
+static enum i40iw_status_code i40iw_hmc_sd_grp(struct i40iw_sc_dev *dev,
+					       struct i40iw_hmc_info *hmc_info,
+					       u32 sd_index,
+					       u32 sd_cnt,
+					       bool setsd)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+	struct i40iw_update_sds_info sdinfo;
+	u64 pa;
+	u32 i;
+	enum i40iw_status_code ret_code = 0;
+
+	memset(&sdinfo, 0, sizeof(sdinfo));
+	sdinfo.hmc_fn_id = hmc_info->hmc_fn_id;
+	for (i = sd_index; i < sd_index + sd_cnt; i++) {
+		sd_entry = &hmc_info->sd_table.sd_entry[i];
+		if (!sd_entry ||
+		    (!sd_entry->valid && setsd) ||
+		    (sd_entry->valid && !setsd))
+			continue;
+		if (setsd) {
+			pa = (sd_entry->entry_type == I40IW_SD_TYPE_PAGED) ?
+			    sd_entry->u.pd_table.pd_page_addr.pa :
+			    sd_entry->u.bp.addr.pa;
+			i40iw_set_sd_entry(pa, i, sd_entry->entry_type,
+					   &sdinfo.entry[sdinfo.cnt]);
+		} else {
+			i40iw_clr_sd_entry(i, sd_entry->entry_type,
+					   &sdinfo.entry[sdinfo.cnt]);
+		}
+		sdinfo.cnt++;
+		if (sdinfo.cnt == I40IW_MAX_SD_ENTRIES) {
+			ret_code = dev->cqp->process_cqp_sds(dev, &sdinfo);
+			if (ret_code) {
+				i40iw_debug(dev, I40IW_DEBUG_HMC,
+					    "i40iw_hmc_sd_grp: sd_programming failed err=%d\n",
+					    ret_code);
+				return ret_code;
+			}
+			sdinfo.cnt = 0;
+		}
+	}
+	if (sdinfo.cnt)
+		ret_code = dev->cqp->process_cqp_sds(dev, &sdinfo);
+
+	return ret_code;
+}
+
+/**
+ * i40iw_vfdev_from_fpm - return vf dev ptr for hmc function id
+ * @dev: pointer to the device structure
+ * @hmc_fn_id: hmc's function id
+ */
+struct i40iw_vfdev *i40iw_vfdev_from_fpm(struct i40iw_sc_dev *dev, u8 hmc_fn_id)
+{
+	struct i40iw_vfdev *vf_dev = NULL;
+	u16 idx;
+
+	for (idx = 0; idx < I40IW_MAX_PE_ENABLED_VF_COUNT; idx++) {
+		if (dev->vf_dev[idx] &&
+		    ((u8)dev->vf_dev[idx]->pmf_index == hmc_fn_id)) {
+			vf_dev = dev->vf_dev[idx];
+			break;
+		}
+	}
+	return vf_dev;
+}
+
+/**
+ * i40iw_vf_hmcinfo_from_fpm - get ptr to hmc for func_id
+ * @dev: pointer to the device structure
+ * @hmc_fn_id: hmc's function id
+ */
+struct i40iw_hmc_info *i40iw_vf_hmcinfo_from_fpm(struct i40iw_sc_dev *dev,
+						 u8 hmc_fn_id)
+{
+	struct i40iw_hmc_info *hmc_info = NULL;
+	u16 idx;
+
+	for (idx = 0; idx < I40IW_MAX_PE_ENABLED_VF_COUNT; idx++) {
+		if (dev->vf_dev[idx] &&
+		    ((u8)dev->vf_dev[idx]->pmf_index == hmc_fn_id)) {
+			hmc_info = &dev->vf_dev[idx]->hmc_info;
+			break;
+		}
+	}
+	return hmc_info;
+}
+
+/**
+ * i40iw_hmc_finish_add_sd_reg - program sd entries for objects
+ * @dev: pointer to the device structure
+ * @info: create obj info
+ */
+static enum i40iw_status_code i40iw_hmc_finish_add_sd_reg(struct i40iw_sc_dev *dev,
+							  struct i40iw_hmc_create_obj_info *info)
+{
+	if (info->start_idx >= info->hmc_info->hmc_obj[info->rsrc_type].cnt)
+		return I40IW_ERR_INVALID_HMC_OBJ_INDEX;
+
+	if ((info->start_idx + info->count) >
+			info->hmc_info->hmc_obj[info->rsrc_type].cnt)
+		return I40IW_ERR_INVALID_HMC_OBJ_COUNT;
+
+	if (!info->add_sd_cnt)
+		return 0;
+
+	return i40iw_hmc_sd_grp(dev, info->hmc_info,
+				info->hmc_info->sd_indexes[0],
+				info->add_sd_cnt, true);
+}
+
+/**
+ * i40iw_create_iw_hmc_obj - allocate backing store for hmc objects
+ * @dev: pointer to the device structure
+ * @info: pointer to i40iw_hmc_iw_create_obj_info struct
+ *
+ * This will allocate memory for PDs and backing pages and populate
+ * the sd and pd entries.
+ */
+enum i40iw_status_code i40iw_sc_create_hmc_obj(struct i40iw_sc_dev *dev,
+					       struct i40iw_hmc_create_obj_info *info)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+	u32 sd_idx, sd_lmt;
+	u32 pd_idx = 0, pd_lmt = 0;
+	u32 pd_idx1 = 0, pd_lmt1 = 0;
+	u32 i, j;
+	bool pd_error = false;
+	enum i40iw_status_code ret_code = 0;
+
+	if (info->start_idx >= info->hmc_info->hmc_obj[info->rsrc_type].cnt)
+		return I40IW_ERR_INVALID_HMC_OBJ_INDEX;
+
+	if ((info->start_idx + info->count) >
+	    info->hmc_info->hmc_obj[info->rsrc_type].cnt) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s: error type %u, start = %u, req cnt %u, cnt = %u\n",
+			    __func__, info->rsrc_type, info->start_idx, info->count,
+			    info->hmc_info->hmc_obj[info->rsrc_type].cnt);
+		return I40IW_ERR_INVALID_HMC_OBJ_COUNT;
+	}
+
+	if (!dev->is_pf)
+		return i40iw_vchnl_vf_add_hmc_objs(dev, info->rsrc_type, 0, info->count);
+
+	i40iw_find_sd_index_limit(info->hmc_info, info->rsrc_type,
+				  info->start_idx, info->count,
+				  &sd_idx, &sd_lmt);
+	if (sd_idx >= info->hmc_info->sd_table.sd_cnt ||
+	    sd_lmt > info->hmc_info->sd_table.sd_cnt) {
+		return I40IW_ERR_INVALID_SD_INDEX;
+	}
+	i40iw_find_pd_index_limit(info->hmc_info, info->rsrc_type,
+				  info->start_idx, info->count, &pd_idx, &pd_lmt);
+
+	for (j = sd_idx; j < sd_lmt; j++) {
+		ret_code = i40iw_add_sd_table_entry(dev->hw, info->hmc_info,
+						    j,
+						    info->entry_type,
+						    I40IW_HMC_DIRECT_BP_SIZE);
+		if (ret_code)
+			goto exit_sd_error;
+		sd_entry = &info->hmc_info->sd_table.sd_entry[j];
+
+		if ((sd_entry->entry_type == I40IW_SD_TYPE_PAGED) &&
+		    ((dev->hmc_info == info->hmc_info) &&
+		     (info->rsrc_type != I40IW_HMC_IW_PBLE))) {
+			pd_idx1 = max(pd_idx, (j * I40IW_HMC_MAX_BP_COUNT));
+			pd_lmt1 = min(pd_lmt,
+				      (j + 1) * I40IW_HMC_MAX_BP_COUNT);
+			for (i = pd_idx1; i < pd_lmt1; i++) {
+				/* update the pd table entry */
+				ret_code = i40iw_add_pd_table_entry(dev->hw, info->hmc_info,
+								    i, NULL);
+				if (ret_code) {
+					pd_error = true;
+					break;
+				}
+			}
+			if (pd_error) {
+				while (i && (i > pd_idx1)) {
+					i40iw_remove_pd_bp(dev->hw, info->hmc_info, (i - 1),
+							   info->is_pf);
+					i--;
+				}
+			}
+		}
+		if (sd_entry->valid)
+			continue;
+
+		info->hmc_info->sd_indexes[info->add_sd_cnt] = (u16)j;
+		info->add_sd_cnt++;
+		sd_entry->valid = true;
+	}
+	return i40iw_hmc_finish_add_sd_reg(dev, info);
+
+exit_sd_error:
+	while (j && (j > sd_idx)) {
+		sd_entry = &info->hmc_info->sd_table.sd_entry[j - 1];
+		switch (sd_entry->entry_type) {
+		case I40IW_SD_TYPE_PAGED:
+			pd_idx1 = max(pd_idx,
+				      (j - 1) * I40IW_HMC_MAX_BP_COUNT);
+			pd_lmt1 = min(pd_lmt, (j * I40IW_HMC_MAX_BP_COUNT));
+			for (i = pd_idx1; i < pd_lmt1; i++)
+				i40iw_prep_remove_pd_page(info->hmc_info, i);
+			break;
+		case I40IW_SD_TYPE_DIRECT:
+			i40iw_prep_remove_pd_page(info->hmc_info, (j - 1));
+			break;
+		default:
+			ret_code = I40IW_ERR_INVALID_SD_TYPE;
+			break;
+		}
+		j--;
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_finish_del_sd_reg - delete sd entries for objects
+ * @dev: pointer to the device structure
+ * @info: dele obj info
+ * @reset: true if called before reset
+ */
+static enum i40iw_status_code i40iw_finish_del_sd_reg(struct i40iw_sc_dev *dev,
+						      struct i40iw_hmc_del_obj_info *info,
+						      bool reset)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+	enum i40iw_status_code ret_code = 0;
+	u32 i, sd_idx;
+	struct i40iw_dma_mem *mem;
+
+	if (dev->is_pf && !reset)
+		ret_code = i40iw_hmc_sd_grp(dev, info->hmc_info,
+					    info->hmc_info->sd_indexes[0],
+					    info->del_sd_cnt, false);
+
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_HMC, "%s: error cqp sd sd_grp\n", __func__);
+
+	for (i = 0; i < info->del_sd_cnt; i++) {
+		sd_idx = info->hmc_info->sd_indexes[i];
+		sd_entry = &info->hmc_info->sd_table.sd_entry[sd_idx];
+		if (!sd_entry)
+			continue;
+		mem = (sd_entry->entry_type == I40IW_SD_TYPE_PAGED) ?
+			&sd_entry->u.pd_table.pd_page_addr :
+			&sd_entry->u.bp.addr;
+
+		if (!mem || !mem->va)
+			i40iw_debug(dev, I40IW_DEBUG_HMC, "%s: error cqp sd mem\n", __func__);
+		else
+			i40iw_free_dma_mem(dev->hw, mem);
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_del_iw_hmc_obj - remove pe hmc objects
+ * @dev: pointer to the device structure
+ * @info: pointer to i40iw_hmc_del_obj_info struct
+ * @reset: true if called before reset
+ *
+ * This will de-populate the SDs and PDs.  It frees
+ * the memory for PDS and backing storage.  After this function is returned,
+ * caller should deallocate memory allocated previously for
+ * book-keeping information about PDs and backing storage.
+ */
+enum i40iw_status_code i40iw_sc_del_hmc_obj(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_del_obj_info *info,
+					    bool reset)
+{
+	struct i40iw_hmc_pd_table *pd_table;
+	u32 sd_idx, sd_lmt;
+	u32 pd_idx, pd_lmt, rel_pd_idx;
+	u32 i, j;
+	enum i40iw_status_code ret_code = 0;
+
+	if (info->start_idx >= info->hmc_info->hmc_obj[info->rsrc_type].cnt) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s: error start_idx[%04d]  >= [type %04d].cnt[%04d]\n",
+			    __func__, info->start_idx, info->rsrc_type,
+			    info->hmc_info->hmc_obj[info->rsrc_type].cnt);
+		return I40IW_ERR_INVALID_HMC_OBJ_INDEX;
+	}
+
+	if ((info->start_idx + info->count) >
+	    info->hmc_info->hmc_obj[info->rsrc_type].cnt) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC,
+			    "%s: error start_idx[%04d] + count %04d  >= [type %04d].cnt[%04d]\n",
+			    __func__, info->start_idx, info->count,
+			    info->rsrc_type,
+			    info->hmc_info->hmc_obj[info->rsrc_type].cnt);
+		return I40IW_ERR_INVALID_HMC_OBJ_COUNT;
+	}
+	if (!dev->is_pf) {
+		ret_code = i40iw_vchnl_vf_del_hmc_obj(dev, info->rsrc_type, 0,
+						      info->count);
+		if (info->rsrc_type != I40IW_HMC_IW_PBLE)
+			return ret_code;
+	}
+
+	i40iw_find_pd_index_limit(info->hmc_info, info->rsrc_type,
+				  info->start_idx, info->count, &pd_idx, &pd_lmt);
+
+	for (j = pd_idx; j < pd_lmt; j++) {
+		sd_idx = j / I40IW_HMC_PD_CNT_IN_SD;
+
+		if (info->hmc_info->sd_table.sd_entry[sd_idx].entry_type !=
+		    I40IW_SD_TYPE_PAGED)
+			continue;
+
+		rel_pd_idx = j % I40IW_HMC_PD_CNT_IN_SD;
+		pd_table = &info->hmc_info->sd_table.sd_entry[sd_idx].u.pd_table;
+		if (pd_table->pd_entry[rel_pd_idx].valid) {
+			ret_code = i40iw_remove_pd_bp(dev->hw, info->hmc_info, j,
+						      info->is_pf);
+			if (ret_code) {
+				i40iw_debug(dev, I40IW_DEBUG_HMC, "%s: error\n", __func__);
+				return ret_code;
+			}
+		}
+	}
+
+	i40iw_find_sd_index_limit(info->hmc_info, info->rsrc_type,
+				  info->start_idx, info->count, &sd_idx, &sd_lmt);
+	if (sd_idx >= info->hmc_info->sd_table.sd_cnt ||
+	    sd_lmt > info->hmc_info->sd_table.sd_cnt) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC, "%s: error invalid sd_idx\n", __func__);
+		return I40IW_ERR_INVALID_SD_INDEX;
+	}
+
+	for (i = sd_idx; i < sd_lmt; i++) {
+		if (!info->hmc_info->sd_table.sd_entry[i].valid)
+			continue;
+		switch (info->hmc_info->sd_table.sd_entry[i].entry_type) {
+		case I40IW_SD_TYPE_DIRECT:
+			ret_code = i40iw_prep_remove_sd_bp(info->hmc_info, i);
+			if (!ret_code) {
+				info->hmc_info->sd_indexes[info->del_sd_cnt] = (u16)i;
+				info->del_sd_cnt++;
+			}
+			break;
+		case I40IW_SD_TYPE_PAGED:
+			ret_code = i40iw_prep_remove_pd_page(info->hmc_info, i);
+			if (!ret_code) {
+				info->hmc_info->sd_indexes[info->del_sd_cnt] = (u16)i;
+				info->del_sd_cnt++;
+			}
+			break;
+		default:
+			break;
+		}
+	}
+	return i40iw_finish_del_sd_reg(dev, info, reset);
+}
+
+/**
+ * i40iw_add_sd_table_entry - Adds a segment descriptor to the table
+ * @hw: pointer to our hw struct
+ * @hmc_info: pointer to the HMC configuration information struct
+ * @sd_index: segment descriptor index to manipulate
+ * @type: what type of segment descriptor we're manipulating
+ * @direct_mode_sz: size to alloc in direct mode
+ */
+enum i40iw_status_code i40iw_add_sd_table_entry(struct i40iw_hw *hw,
+						struct i40iw_hmc_info *hmc_info,
+						u32 sd_index,
+						enum i40iw_sd_entry_type type,
+						u64 direct_mode_sz)
+{
+	enum i40iw_status_code ret_code = 0;
+	struct i40iw_hmc_sd_entry *sd_entry;
+	bool dma_mem_alloc_done = false;
+	struct i40iw_dma_mem mem;
+	u64 alloc_len;
+
+	sd_entry = &hmc_info->sd_table.sd_entry[sd_index];
+	if (!sd_entry->valid) {
+		if (type == I40IW_SD_TYPE_PAGED)
+			alloc_len = I40IW_HMC_PAGED_BP_SIZE;
+		else
+			alloc_len = direct_mode_sz;
+
+		/* allocate a 4K pd page or 2M backing page */
+		ret_code = i40iw_allocate_dma_mem(hw, &mem, alloc_len,
+						  I40IW_HMC_PD_BP_BUF_ALIGNMENT);
+		if (ret_code)
+			goto exit;
+		dma_mem_alloc_done = true;
+		if (type == I40IW_SD_TYPE_PAGED) {
+			ret_code = i40iw_allocate_virt_mem(hw,
+							   &sd_entry->u.pd_table.pd_entry_virt_mem,
+							   sizeof(struct i40iw_hmc_pd_entry) * 512);
+			if (ret_code)
+				goto exit;
+			sd_entry->u.pd_table.pd_entry = (struct i40iw_hmc_pd_entry *)
+							 sd_entry->u.pd_table.pd_entry_virt_mem.va;
+
+			memcpy(&sd_entry->u.pd_table.pd_page_addr, &mem, sizeof(struct i40iw_dma_mem));
+		} else {
+			memcpy(&sd_entry->u.bp.addr, &mem, sizeof(struct i40iw_dma_mem));
+			sd_entry->u.bp.sd_pd_index = sd_index;
+		}
+
+		hmc_info->sd_table.sd_entry[sd_index].entry_type = type;
+
+		I40IW_INC_SD_REFCNT(&hmc_info->sd_table);
+	}
+	if (sd_entry->entry_type == I40IW_SD_TYPE_DIRECT)
+		I40IW_INC_BP_REFCNT(&sd_entry->u.bp);
+exit:
+	if (ret_code)
+		if (dma_mem_alloc_done)
+			i40iw_free_dma_mem(hw, &mem);
+
+	return ret_code;
+}
+
+/**
+ * i40iw_add_pd_table_entry - Adds page descriptor to the specified table
+ * @hw: pointer to our HW structure
+ * @hmc_info: pointer to the HMC configuration information structure
+ * @pd_index: which page descriptor index to manipulate
+ * @rsrc_pg: if not NULL, use preallocated page instead of allocating new one.
+ *
+ * This function:
+ *	1. Initializes the pd entry
+ *	2. Adds pd_entry in the pd_table
+ *	3. Mark the entry valid in i40iw_hmc_pd_entry structure
+ *	4. Initializes the pd_entry's ref count to 1
+ * assumptions:
+ *	1. The memory for pd should be pinned down, physically contiguous and
+ *	   aligned on 4K boundary and zeroed memory.
+ *	2. It should be 4K in size.
+ */
+enum i40iw_status_code i40iw_add_pd_table_entry(struct i40iw_hw *hw,
+						struct i40iw_hmc_info *hmc_info,
+						u32 pd_index,
+						struct i40iw_dma_mem *rsrc_pg)
+{
+	enum i40iw_status_code ret_code = 0;
+	struct i40iw_hmc_pd_table *pd_table;
+	struct i40iw_hmc_pd_entry *pd_entry;
+	struct i40iw_dma_mem mem;
+	struct i40iw_dma_mem *page = &mem;
+	u32 sd_idx, rel_pd_idx;
+	u64 *pd_addr;
+	u64 page_desc;
+
+	if (pd_index / I40IW_HMC_PD_CNT_IN_SD >= hmc_info->sd_table.sd_cnt)
+		return I40IW_ERR_INVALID_PAGE_DESC_INDEX;
+
+	sd_idx = (pd_index / I40IW_HMC_PD_CNT_IN_SD);
+	if (hmc_info->sd_table.sd_entry[sd_idx].entry_type != I40IW_SD_TYPE_PAGED)
+		return 0;
+
+	rel_pd_idx = (pd_index % I40IW_HMC_PD_CNT_IN_SD);
+	pd_table = &hmc_info->sd_table.sd_entry[sd_idx].u.pd_table;
+	pd_entry = &pd_table->pd_entry[rel_pd_idx];
+	if (!pd_entry->valid) {
+		if (rsrc_pg) {
+			pd_entry->rsrc_pg = true;
+			page = rsrc_pg;
+		} else {
+			ret_code = i40iw_allocate_dma_mem(hw, page,
+							  I40IW_HMC_PAGED_BP_SIZE,
+							  I40IW_HMC_PD_BP_BUF_ALIGNMENT);
+			if (ret_code)
+				return ret_code;
+			pd_entry->rsrc_pg = false;
+		}
+
+		memcpy(&pd_entry->bp.addr, page, sizeof(struct i40iw_dma_mem));
+		pd_entry->bp.sd_pd_index = pd_index;
+		pd_entry->bp.entry_type = I40IW_SD_TYPE_PAGED;
+		page_desc = page->pa | 0x1;
+
+		pd_addr = (u64 *)pd_table->pd_page_addr.va;
+		pd_addr += rel_pd_idx;
+
+		memcpy(pd_addr, &page_desc, sizeof(*pd_addr));
+
+		pd_entry->sd_index = sd_idx;
+		pd_entry->valid = true;
+		I40IW_INC_PD_REFCNT(pd_table);
+		if (hmc_info->hmc_fn_id < I40IW_FIRST_VF_FPM_ID)
+			I40IW_INVALIDATE_PF_HMC_PD(hw, sd_idx, rel_pd_idx);
+		else if (hw->hmc.hmc_fn_id != hmc_info->hmc_fn_id)
+			I40IW_INVALIDATE_VF_HMC_PD(hw, sd_idx, rel_pd_idx,
+						   hmc_info->hmc_fn_id);
+	}
+	I40IW_INC_BP_REFCNT(&pd_entry->bp);
+
+	return 0;
+}
+
+/**
+ * i40iw_remove_pd_bp - remove a backing page from a page descriptor
+ * @hw: pointer to our HW structure
+ * @hmc_info: pointer to the HMC configuration information structure
+ * @idx: the page index
+ * @is_pf: distinguishes a VF from a PF
+ *
+ * This function:
+ *	1. Marks the entry in pd table (for paged address mode) or in sd table
+ *	   (for direct address mode) invalid.
+ *	2. Write to register PMPDINV to invalidate the backing page in FV cache
+ *	3. Decrement the ref count for the pd _entry
+ * assumptions:
+ *	1. Caller can deallocate the memory used by backing storage after this
+ *	   function returns.
+ */
+enum i40iw_status_code i40iw_remove_pd_bp(struct i40iw_hw *hw,
+					  struct i40iw_hmc_info *hmc_info,
+					  u32 idx,
+					  bool is_pf)
+{
+	struct i40iw_hmc_pd_entry *pd_entry;
+	struct i40iw_hmc_pd_table *pd_table;
+	struct i40iw_hmc_sd_entry *sd_entry;
+	u32 sd_idx, rel_pd_idx;
+	struct i40iw_dma_mem *mem;
+	u64 *pd_addr;
+
+	sd_idx = idx / I40IW_HMC_PD_CNT_IN_SD;
+	rel_pd_idx = idx % I40IW_HMC_PD_CNT_IN_SD;
+	if (sd_idx >= hmc_info->sd_table.sd_cnt)
+		return I40IW_ERR_INVALID_PAGE_DESC_INDEX;
+
+	sd_entry = &hmc_info->sd_table.sd_entry[sd_idx];
+	if (sd_entry->entry_type != I40IW_SD_TYPE_PAGED)
+		return I40IW_ERR_INVALID_SD_TYPE;
+
+	pd_table = &hmc_info->sd_table.sd_entry[sd_idx].u.pd_table;
+	pd_entry = &pd_table->pd_entry[rel_pd_idx];
+	I40IW_DEC_BP_REFCNT(&pd_entry->bp);
+	if (pd_entry->bp.ref_cnt)
+		return 0;
+
+	pd_entry->valid = false;
+	I40IW_DEC_PD_REFCNT(pd_table);
+	pd_addr = (u64 *)pd_table->pd_page_addr.va;
+	pd_addr += rel_pd_idx;
+	memset(pd_addr, 0, sizeof(u64));
+	if (is_pf)
+		I40IW_INVALIDATE_PF_HMC_PD(hw, sd_idx, idx);
+	else
+		I40IW_INVALIDATE_VF_HMC_PD(hw, sd_idx, idx,
+					   hmc_info->hmc_fn_id);
+
+	if (!pd_entry->rsrc_pg) {
+		mem = &pd_entry->bp.addr;
+		if (!mem || !mem->va)
+			return I40IW_ERR_PARAM;
+		i40iw_free_dma_mem(hw, mem);
+	}
+	if (!pd_table->ref_cnt)
+		i40iw_free_virt_mem(hw, &pd_table->pd_entry_virt_mem);
+
+	return 0;
+}
+
+/**
+ * i40iw_prep_remove_sd_bp - Prepares to remove a backing page from a sd entry
+ * @hmc_info: pointer to the HMC configuration information structure
+ * @idx: the page index
+ */
+enum i40iw_status_code i40iw_prep_remove_sd_bp(struct i40iw_hmc_info *hmc_info, u32 idx)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+
+	sd_entry = &hmc_info->sd_table.sd_entry[idx];
+	I40IW_DEC_BP_REFCNT(&sd_entry->u.bp);
+	if (sd_entry->u.bp.ref_cnt)
+		return I40IW_ERR_NOT_READY;
+
+	I40IW_DEC_SD_REFCNT(&hmc_info->sd_table);
+	sd_entry->valid = false;
+
+	return 0;
+}
+
+/**
+ * i40iw_prep_remove_pd_page - Prepares to remove a PD page from sd entry.
+ * @hmc_info: pointer to the HMC configuration information structure
+ * @idx: segment descriptor index to find the relevant page descriptor
+ */
+enum i40iw_status_code i40iw_prep_remove_pd_page(struct i40iw_hmc_info *hmc_info,
+						 u32 idx)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+
+	sd_entry = &hmc_info->sd_table.sd_entry[idx];
+
+	if (sd_entry->u.pd_table.ref_cnt)
+		return I40IW_ERR_NOT_READY;
+
+	sd_entry->valid = false;
+	I40IW_DEC_SD_REFCNT(&hmc_info->sd_table);
+
+	return 0;
+}
+
+/**
+ * i40iw_pf_init_vfhmc -
+ * @vf_cnt_array: array of cnt values of iwarp hmc objects
+ * @vf_hmc_fn_id: hmc function id ofr vf driver
+ * @dev: pointer to i40iw_dev struct
+ *
+ * Called by pf driver to initialize hmc_info for vf driver instance.
+ */
+enum i40iw_status_code i40iw_pf_init_vfhmc(struct i40iw_sc_dev *dev,
+					   u8 vf_hmc_fn_id,
+					   u32 *vf_cnt_array)
+{
+	struct i40iw_hmc_info *hmc_info;
+	enum i40iw_status_code ret_code = 0;
+	u32 i;
+
+	if ((vf_hmc_fn_id < I40IW_FIRST_VF_FPM_ID) ||
+	    (vf_hmc_fn_id >= I40IW_FIRST_VF_FPM_ID +
+	     I40IW_MAX_PE_ENABLED_VF_COUNT)) {
+		i40iw_debug(dev, I40IW_DEBUG_HMC, "%s: invalid vf_hmc_fn_id  0x%x\n",
+			    __func__, vf_hmc_fn_id);
+		return I40IW_ERR_INVALID_HMCFN_ID;
+	}
+
+	ret_code = i40iw_sc_init_iw_hmc(dev, vf_hmc_fn_id);
+	if (ret_code)
+		return ret_code;
+
+	hmc_info = i40iw_vf_hmcinfo_from_fpm(dev, vf_hmc_fn_id);
+
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
+		if (vf_cnt_array)
+			hmc_info->hmc_obj[i].cnt =
+			    vf_cnt_array[i - I40IW_HMC_IW_QP];
+		else
+			hmc_info->hmc_obj[i].cnt = hmc_info->hmc_obj[i].max_cnt;
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_hmc.h b/drivers/infiniband/hw/i40iw/i40iw_hmc.h
new file mode 100644
index 000000000000..4c3fdd875621
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_hmc.h
@@ -0,0 +1,241 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_HMC_H
+#define I40IW_HMC_H
+
+#include "i40iw_d.h"
+
+struct i40iw_hw;
+enum i40iw_status_code;
+
+#define I40IW_HMC_MAX_BP_COUNT 512
+#define I40IW_MAX_SD_ENTRIES 11
+#define I40IW_HW_DBG_HMC_INVALID_BP_MARK     0xCA
+
+#define I40IW_HMC_INFO_SIGNATURE	0x484D5347
+#define I40IW_HMC_PD_CNT_IN_SD		512
+#define I40IW_HMC_DIRECT_BP_SIZE	0x200000
+#define I40IW_HMC_MAX_SD_COUNT		4096
+#define I40IW_HMC_PAGED_BP_SIZE		4096
+#define I40IW_HMC_PD_BP_BUF_ALIGNMENT	4096
+#define I40IW_FIRST_VF_FPM_ID		16
+#define FPM_MULTIPLIER			1024
+
+#define I40IW_INC_SD_REFCNT(sd_table)   ((sd_table)->ref_cnt++)
+#define I40IW_INC_PD_REFCNT(pd_table)   ((pd_table)->ref_cnt++)
+#define I40IW_INC_BP_REFCNT(bp)         ((bp)->ref_cnt++)
+
+#define I40IW_DEC_SD_REFCNT(sd_table)   ((sd_table)->ref_cnt--)
+#define I40IW_DEC_PD_REFCNT(pd_table)   ((pd_table)->ref_cnt--)
+#define I40IW_DEC_BP_REFCNT(bp)         ((bp)->ref_cnt--)
+
+/**
+ * I40IW_INVALIDATE_PF_HMC_PD - Invalidates the pd cache in the hardware
+ * @hw: pointer to our hw struct
+ * @sd_idx: segment descriptor index
+ * @pd_idx: page descriptor index
+ */
+#define I40IW_INVALIDATE_PF_HMC_PD(hw, sd_idx, pd_idx)                  \
+	i40iw_wr32((hw), I40E_PFHMC_PDINV,                                    \
+		(((sd_idx) << I40E_PFHMC_PDINV_PMSDIDX_SHIFT) |             \
+		(0x1 << I40E_PFHMC_PDINV_PMSDPARTSEL_SHIFT) | \
+		((pd_idx) << I40E_PFHMC_PDINV_PMPDIDX_SHIFT)))
+
+/**
+ * I40IW_INVALIDATE_VF_HMC_PD - Invalidates the pd cache in the hardware
+ * @hw: pointer to our hw struct
+ * @sd_idx: segment descriptor index
+ * @pd_idx: page descriptor index
+ * @hmc_fn_id: VF's function id
+ */
+#define I40IW_INVALIDATE_VF_HMC_PD(hw, sd_idx, pd_idx, hmc_fn_id)        \
+	i40iw_wr32(hw, I40E_GLHMC_VFPDINV(hmc_fn_id - I40IW_FIRST_VF_FPM_ID),  \
+	     ((sd_idx << I40E_PFHMC_PDINV_PMSDIDX_SHIFT) |              \
+	      (pd_idx << I40E_PFHMC_PDINV_PMPDIDX_SHIFT)))
+
+struct i40iw_hmc_obj_info {
+	u64 base;
+	u32 max_cnt;
+	u32 cnt;
+	u64 size;
+};
+
+enum i40iw_sd_entry_type {
+	I40IW_SD_TYPE_INVALID = 0,
+	I40IW_SD_TYPE_PAGED = 1,
+	I40IW_SD_TYPE_DIRECT = 2
+};
+
+struct i40iw_hmc_bp {
+	enum i40iw_sd_entry_type entry_type;
+	struct i40iw_dma_mem addr;
+	u32 sd_pd_index;
+	u32 ref_cnt;
+};
+
+struct i40iw_hmc_pd_entry {
+	struct i40iw_hmc_bp bp;
+	u32 sd_index;
+	bool rsrc_pg;
+	bool valid;
+};
+
+struct i40iw_hmc_pd_table {
+	struct i40iw_dma_mem pd_page_addr;
+	struct i40iw_hmc_pd_entry *pd_entry;
+	struct i40iw_virt_mem pd_entry_virt_mem;
+	u32 ref_cnt;
+	u32 sd_index;
+};
+
+struct i40iw_hmc_sd_entry {
+	enum i40iw_sd_entry_type entry_type;
+	bool valid;
+
+	union {
+		struct i40iw_hmc_pd_table pd_table;
+		struct i40iw_hmc_bp bp;
+	} u;
+};
+
+struct i40iw_hmc_sd_table {
+	struct i40iw_virt_mem addr;
+	u32 sd_cnt;
+	u32 ref_cnt;
+	struct i40iw_hmc_sd_entry *sd_entry;
+};
+
+struct i40iw_hmc_info {
+	u32 signature;
+	u8 hmc_fn_id;
+	u16 first_sd_index;
+
+	struct i40iw_hmc_obj_info *hmc_obj;
+	struct i40iw_virt_mem hmc_obj_virt_mem;
+	struct i40iw_hmc_sd_table sd_table;
+	u16 sd_indexes[I40IW_HMC_MAX_SD_COUNT];
+};
+
+struct update_sd_entry {
+	u64 cmd;
+	u64 data;
+};
+
+struct i40iw_update_sds_info {
+	u32 cnt;
+	u8 hmc_fn_id;
+	struct update_sd_entry entry[I40IW_MAX_SD_ENTRIES];
+};
+
+struct i40iw_ccq_cqe_info;
+struct i40iw_hmc_fcn_info {
+	void (*callback_fcn)(struct i40iw_sc_dev *, void *,
+			     struct i40iw_ccq_cqe_info *);
+	void *cqp_callback_param;
+	u32 vf_id;
+	u16 iw_vf_idx;
+	bool free_fcn;
+};
+
+enum i40iw_hmc_rsrc_type {
+	I40IW_HMC_IW_QP = 0,
+	I40IW_HMC_IW_CQ = 1,
+	I40IW_HMC_IW_SRQ = 2,
+	I40IW_HMC_IW_HTE = 3,
+	I40IW_HMC_IW_ARP = 4,
+	I40IW_HMC_IW_APBVT_ENTRY = 5,
+	I40IW_HMC_IW_MR = 6,
+	I40IW_HMC_IW_XF = 7,
+	I40IW_HMC_IW_XFFL = 8,
+	I40IW_HMC_IW_Q1 = 9,
+	I40IW_HMC_IW_Q1FL = 10,
+	I40IW_HMC_IW_TIMER = 11,
+	I40IW_HMC_IW_FSIMC = 12,
+	I40IW_HMC_IW_FSIAV = 13,
+	I40IW_HMC_IW_PBLE = 14,
+	I40IW_HMC_IW_MAX = 15,
+};
+
+struct i40iw_hmc_create_obj_info {
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_virt_mem add_sd_virt_mem;
+	u32 rsrc_type;
+	u32 start_idx;
+	u32 count;
+	u32 add_sd_cnt;
+	enum i40iw_sd_entry_type entry_type;
+	bool is_pf;
+};
+
+struct i40iw_hmc_del_obj_info {
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_virt_mem del_sd_virt_mem;
+	u32 rsrc_type;
+	u32 start_idx;
+	u32 count;
+	u32 del_sd_cnt;
+	bool is_pf;
+};
+
+enum i40iw_status_code i40iw_copy_dma_mem(struct i40iw_hw *hw, void *dest_buf,
+					  struct i40iw_dma_mem *src_mem, u64 src_offset, u64 size);
+enum i40iw_status_code i40iw_sc_create_hmc_obj(struct i40iw_sc_dev *dev,
+					       struct i40iw_hmc_create_obj_info *info);
+enum i40iw_status_code i40iw_sc_del_hmc_obj(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_del_obj_info *info,
+					    bool reset);
+enum i40iw_status_code i40iw_hmc_sd_one(struct i40iw_sc_dev *dev, u8 hmc_fn_id,
+					u64 pa, u32 sd_idx, enum i40iw_sd_entry_type type,
+					bool setsd);
+enum i40iw_status_code i40iw_update_sds_noccq(struct i40iw_sc_dev *dev,
+					      struct i40iw_update_sds_info *info);
+struct i40iw_vfdev *i40iw_vfdev_from_fpm(struct i40iw_sc_dev *dev, u8 hmc_fn_id);
+struct i40iw_hmc_info *i40iw_vf_hmcinfo_from_fpm(struct i40iw_sc_dev *dev,
+						 u8 hmc_fn_id);
+enum i40iw_status_code i40iw_add_sd_table_entry(struct i40iw_hw *hw,
+						struct i40iw_hmc_info *hmc_info, u32 sd_index,
+						enum i40iw_sd_entry_type type, u64 direct_mode_sz);
+enum i40iw_status_code i40iw_add_pd_table_entry(struct i40iw_hw *hw,
+						struct i40iw_hmc_info *hmc_info, u32 pd_index,
+						struct i40iw_dma_mem *rsrc_pg);
+enum i40iw_status_code i40iw_remove_pd_bp(struct i40iw_hw *hw,
+					  struct i40iw_hmc_info *hmc_info, u32 idx, bool is_pf);
+enum i40iw_status_code i40iw_prep_remove_sd_bp(struct i40iw_hmc_info *hmc_info, u32 idx);
+enum i40iw_status_code i40iw_prep_remove_pd_page(struct i40iw_hmc_info *hmc_info, u32 idx);
+
+#define     ENTER_SHARED_FUNCTION()
+#define     EXIT_SHARED_FUNCTION()
+
+#endif				/* I40IW_HMC_H */
diff --git a/drivers/infiniband/hw/i40iw/i40iw_hw.c b/drivers/infiniband/hw/i40iw/i40iw_hw.c
new file mode 100644
index 000000000000..9fd302425563
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_hw.c
@@ -0,0 +1,730 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/if_vlan.h>
+
+#include "i40iw.h"
+
+/**
+ * i40iw_initialize_hw_resources - initialize hw resource during open
+ * @iwdev: iwarp device
+ */
+u32 i40iw_initialize_hw_resources(struct i40iw_device *iwdev)
+{
+	unsigned long num_pds;
+	u32 resources_size;
+	u32 max_mr;
+	u32 max_qp;
+	u32 max_cq;
+	u32 arp_table_size;
+	u32 mrdrvbits;
+	void *resource_ptr;
+
+	max_qp = iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_QP].cnt;
+	max_cq = iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt;
+	max_mr = iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_MR].cnt;
+	arp_table_size = iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_ARP].cnt;
+	iwdev->max_cqe = 0xFFFFF;
+	num_pds = max_qp * 4;
+	resources_size = sizeof(struct i40iw_arp_entry) * arp_table_size;
+	resources_size += sizeof(unsigned long) * BITS_TO_LONGS(max_qp);
+	resources_size += sizeof(unsigned long) * BITS_TO_LONGS(max_mr);
+	resources_size += sizeof(unsigned long) * BITS_TO_LONGS(max_cq);
+	resources_size += sizeof(unsigned long) * BITS_TO_LONGS(num_pds);
+	resources_size += sizeof(unsigned long) * BITS_TO_LONGS(arp_table_size);
+	resources_size += sizeof(struct i40iw_qp **) * max_qp;
+	iwdev->mem_resources = kzalloc(resources_size, GFP_KERNEL);
+
+	if (!iwdev->mem_resources)
+		return -ENOMEM;
+
+	iwdev->max_qp = max_qp;
+	iwdev->max_mr = max_mr;
+	iwdev->max_cq = max_cq;
+	iwdev->max_pd = num_pds;
+	iwdev->arp_table_size = arp_table_size;
+	iwdev->arp_table = (struct i40iw_arp_entry *)iwdev->mem_resources;
+	resource_ptr = iwdev->mem_resources + (sizeof(struct i40iw_arp_entry) * arp_table_size);
+
+	iwdev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY |
+	    IB_DEVICE_MEM_WINDOW | IB_DEVICE_MEM_MGT_EXTENSIONS;
+
+	iwdev->allocated_qps = resource_ptr;
+	iwdev->allocated_cqs = &iwdev->allocated_qps[BITS_TO_LONGS(max_qp)];
+	iwdev->allocated_mrs = &iwdev->allocated_cqs[BITS_TO_LONGS(max_cq)];
+	iwdev->allocated_pds = &iwdev->allocated_mrs[BITS_TO_LONGS(max_mr)];
+	iwdev->allocated_arps = &iwdev->allocated_pds[BITS_TO_LONGS(num_pds)];
+	iwdev->qp_table = (struct i40iw_qp **)(&iwdev->allocated_arps[BITS_TO_LONGS(arp_table_size)]);
+	set_bit(0, iwdev->allocated_mrs);
+	set_bit(0, iwdev->allocated_qps);
+	set_bit(0, iwdev->allocated_cqs);
+	set_bit(0, iwdev->allocated_pds);
+	set_bit(0, iwdev->allocated_arps);
+
+	/* Following for ILQ/IEQ */
+	set_bit(1, iwdev->allocated_qps);
+	set_bit(1, iwdev->allocated_cqs);
+	set_bit(1, iwdev->allocated_pds);
+	set_bit(2, iwdev->allocated_cqs);
+	set_bit(2, iwdev->allocated_pds);
+
+	spin_lock_init(&iwdev->resource_lock);
+	mrdrvbits = 24 - get_count_order(iwdev->max_mr);
+	iwdev->mr_stagmask = ~(((1 << mrdrvbits) - 1) << (32 - mrdrvbits));
+	return 0;
+}
+
+/**
+ * i40iw_cqp_ce_handler - handle cqp completions
+ * @iwdev: iwarp device
+ * @arm: flag to arm after completions
+ * @cq: cq for cqp completions
+ */
+static void i40iw_cqp_ce_handler(struct i40iw_device *iwdev, struct i40iw_sc_cq *cq, bool arm)
+{
+	struct i40iw_cqp_request *cqp_request;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	u32 cqe_count = 0;
+	struct i40iw_ccq_cqe_info info;
+	int ret;
+
+	do {
+		memset(&info, 0, sizeof(info));
+		ret = dev->ccq_ops->ccq_get_cqe_info(cq, &info);
+		if (ret)
+			break;
+		cqp_request = (struct i40iw_cqp_request *)(unsigned long)info.scratch;
+		if (info.error)
+			i40iw_pr_err("opcode = 0x%x maj_err_code = 0x%x min_err_code = 0x%x\n",
+				     info.op_code, info.maj_err_code, info.min_err_code);
+		if (cqp_request) {
+			cqp_request->compl_info.maj_err_code = info.maj_err_code;
+			cqp_request->compl_info.min_err_code = info.min_err_code;
+			cqp_request->compl_info.op_ret_val = info.op_ret_val;
+			cqp_request->compl_info.error = info.error;
+
+			if (cqp_request->waiting) {
+				cqp_request->request_done = true;
+				wake_up(&cqp_request->waitq);
+				i40iw_put_cqp_request(&iwdev->cqp, cqp_request);
+			} else {
+				if (cqp_request->callback_fcn)
+					cqp_request->callback_fcn(cqp_request, 1);
+				i40iw_put_cqp_request(&iwdev->cqp, cqp_request);
+			}
+		}
+
+		cqe_count++;
+	} while (1);
+
+	if (arm && cqe_count) {
+		i40iw_process_bh(dev);
+		dev->ccq_ops->ccq_arm(cq);
+	}
+}
+
+/**
+ * i40iw_iwarp_ce_handler - handle iwarp completions
+ * @iwdev: iwarp device
+ * @iwcp: iwarp cq receiving event
+ */
+static void i40iw_iwarp_ce_handler(struct i40iw_device *iwdev,
+				   struct i40iw_sc_cq *iwcq)
+{
+	struct i40iw_cq *i40iwcq = iwcq->back_cq;
+
+	if (i40iwcq->ibcq.comp_handler)
+		i40iwcq->ibcq.comp_handler(&i40iwcq->ibcq,
+					   i40iwcq->ibcq.cq_context);
+}
+
+/**
+ * i40iw_puda_ce_handler - handle puda completion events
+ * @iwdev: iwarp device
+ * @cq: puda completion q for event
+ */
+static void i40iw_puda_ce_handler(struct i40iw_device *iwdev,
+				  struct i40iw_sc_cq *cq)
+{
+	struct i40iw_sc_dev *dev = (struct i40iw_sc_dev *)&iwdev->sc_dev;
+	enum i40iw_status_code status;
+	u32 compl_error;
+
+	do {
+		status = i40iw_puda_poll_completion(dev, cq, &compl_error);
+		if (status == I40IW_ERR_QUEUE_EMPTY)
+			break;
+		if (status) {
+			i40iw_pr_err("puda  status = %d\n", status);
+			break;
+		}
+		if (compl_error) {
+			i40iw_pr_err("puda compl_err  =0x%x\n", compl_error);
+			break;
+		}
+	} while (1);
+
+	dev->ccq_ops->ccq_arm(cq);
+}
+
+/**
+ * i40iw_process_ceq - handle ceq for completions
+ * @iwdev: iwarp device
+ * @ceq: ceq having cq for completion
+ */
+void i40iw_process_ceq(struct i40iw_device *iwdev, struct i40iw_ceq *ceq)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_sc_ceq *sc_ceq;
+	struct i40iw_sc_cq *cq;
+	bool arm = true;
+
+	sc_ceq = &ceq->sc_ceq;
+	do {
+		cq = dev->ceq_ops->process_ceq(dev, sc_ceq);
+		if (!cq)
+			break;
+
+		if (cq->cq_type == I40IW_CQ_TYPE_CQP)
+			i40iw_cqp_ce_handler(iwdev, cq, arm);
+		else if (cq->cq_type == I40IW_CQ_TYPE_IWARP)
+			i40iw_iwarp_ce_handler(iwdev, cq);
+		else if ((cq->cq_type == I40IW_CQ_TYPE_ILQ) ||
+			 (cq->cq_type == I40IW_CQ_TYPE_IEQ))
+			i40iw_puda_ce_handler(iwdev, cq);
+	} while (1);
+}
+
+/**
+ * i40iw_next_iw_state - modify qp state
+ * @iwqp: iwarp qp to modify
+ * @state: next state for qp
+ * @del_hash: del hash
+ * @term: term message
+ * @termlen: length of term message
+ */
+void i40iw_next_iw_state(struct i40iw_qp *iwqp,
+			 u8 state,
+			 u8 del_hash,
+			 u8 term,
+			 u8 termlen)
+{
+	struct i40iw_modify_qp_info info;
+
+	memset(&info, 0, sizeof(info));
+	info.next_iwarp_state = state;
+	info.remove_hash_idx = del_hash;
+	info.cq_num_valid = true;
+	info.arp_cache_idx_valid = true;
+	info.dont_send_term = true;
+	info.dont_send_fin = true;
+	info.termlen = termlen;
+
+	if (term & I40IWQP_TERM_SEND_TERM_ONLY)
+		info.dont_send_term = false;
+	if (term & I40IWQP_TERM_SEND_FIN_ONLY)
+		info.dont_send_fin = false;
+	if (iwqp->sc_qp.term_flags && (state == I40IW_QP_STATE_ERROR))
+		info.reset_tcp_conn = true;
+	i40iw_hw_modify_qp(iwqp->iwdev, iwqp, &info, 0);
+}
+
+/**
+ * i40iw_process_aeq - handle aeq events
+ * @iwdev: iwarp device
+ */
+void i40iw_process_aeq(struct i40iw_device *iwdev)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_aeq *aeq = &iwdev->aeq;
+	struct i40iw_sc_aeq *sc_aeq = &aeq->sc_aeq;
+	struct i40iw_aeqe_info aeinfo;
+	struct i40iw_aeqe_info *info = &aeinfo;
+	int ret;
+	struct i40iw_qp *iwqp = NULL;
+	struct i40iw_sc_cq *cq = NULL;
+	struct i40iw_cq *iwcq = NULL;
+	struct i40iw_sc_qp *qp = NULL;
+	struct i40iw_qp_host_ctx_info *ctx_info = NULL;
+	unsigned long flags;
+
+	u32 aeqcnt = 0;
+
+	if (!sc_aeq->size)
+		return;
+
+	do {
+		memset(info, 0, sizeof(*info));
+		ret = dev->aeq_ops->get_next_aeqe(sc_aeq, info);
+		if (ret)
+			break;
+
+		aeqcnt++;
+		i40iw_debug(dev, I40IW_DEBUG_AEQ,
+			    "%s ae_id = 0x%x bool qp=%d qp_id = %d\n",
+			    __func__, info->ae_id, info->qp, info->qp_cq_id);
+		if (info->qp) {
+			iwqp = iwdev->qp_table[info->qp_cq_id];
+			if (!iwqp) {
+				i40iw_pr_err("qp_id %d is already freed\n", info->qp_cq_id);
+				continue;
+			}
+			qp = &iwqp->sc_qp;
+			spin_lock_irqsave(&iwqp->lock, flags);
+			iwqp->hw_tcp_state = info->tcp_state;
+			iwqp->hw_iwarp_state = info->iwarp_state;
+			iwqp->last_aeq = info->ae_id;
+			spin_unlock_irqrestore(&iwqp->lock, flags);
+			ctx_info = &iwqp->ctx_info;
+			ctx_info->err_rq_idx_valid = true;
+		} else {
+			if (info->ae_id != I40IW_AE_CQ_OPERATION_ERROR)
+				continue;
+		}
+
+		switch (info->ae_id) {
+		case I40IW_AE_LLP_FIN_RECEIVED:
+			if (qp->term_flags)
+				continue;
+			if (atomic_inc_return(&iwqp->close_timer_started) == 1) {
+				iwqp->hw_tcp_state = I40IW_TCP_STATE_CLOSE_WAIT;
+				if ((iwqp->hw_tcp_state == I40IW_TCP_STATE_CLOSE_WAIT) &&
+				    (iwqp->ibqp_state == IB_QPS_RTS)) {
+					i40iw_next_iw_state(iwqp,
+							    I40IW_QP_STATE_CLOSING, 0, 0, 0);
+					i40iw_cm_disconn(iwqp);
+				}
+				iwqp->cm_id->add_ref(iwqp->cm_id);
+				i40iw_schedule_cm_timer(iwqp->cm_node,
+							(struct i40iw_puda_buf *)iwqp,
+							I40IW_TIMER_TYPE_CLOSE, 1, 0);
+			}
+			break;
+		case I40IW_AE_LLP_CLOSE_COMPLETE:
+			if (qp->term_flags)
+				i40iw_terminate_done(qp, 0);
+			else
+				i40iw_cm_disconn(iwqp);
+			break;
+		case I40IW_AE_RESET_SENT:
+			i40iw_next_iw_state(iwqp, I40IW_QP_STATE_ERROR, 1, 0, 0);
+			i40iw_cm_disconn(iwqp);
+			break;
+		case I40IW_AE_LLP_CONNECTION_RESET:
+			if (atomic_read(&iwqp->close_timer_started))
+				continue;
+			i40iw_cm_disconn(iwqp);
+			break;
+		case I40IW_AE_TERMINATE_SENT:
+			i40iw_terminate_send_fin(qp);
+			break;
+		case I40IW_AE_LLP_TERMINATE_RECEIVED:
+			i40iw_terminate_received(qp, info);
+			break;
+		case I40IW_AE_CQ_OPERATION_ERROR:
+			i40iw_pr_err("Processing an iWARP related AE for CQ misc = 0x%04X\n",
+				     info->ae_id);
+			cq = (struct i40iw_sc_cq *)(unsigned long)info->compl_ctx;
+			iwcq = (struct i40iw_cq *)cq->back_cq;
+
+			if (iwcq->ibcq.event_handler) {
+				struct ib_event ibevent;
+
+				ibevent.device = iwcq->ibcq.device;
+				ibevent.event = IB_EVENT_CQ_ERR;
+				ibevent.element.cq = &iwcq->ibcq;
+				iwcq->ibcq.event_handler(&ibevent, iwcq->ibcq.cq_context);
+			}
+			break;
+		case I40IW_AE_PRIV_OPERATION_DENIED:
+		case I40IW_AE_STAG_ZERO_INVALID:
+		case I40IW_AE_IB_RREQ_AND_Q1_FULL:
+		case I40IW_AE_DDP_UBE_INVALID_DDP_VERSION:
+		case I40IW_AE_DDP_UBE_INVALID_MO:
+		case I40IW_AE_DDP_UBE_INVALID_QN:
+		case I40IW_AE_DDP_NO_L_BIT:
+		case I40IW_AE_RDMAP_ROE_INVALID_RDMAP_VERSION:
+		case I40IW_AE_RDMAP_ROE_UNEXPECTED_OPCODE:
+		case I40IW_AE_ROE_INVALID_RDMA_READ_REQUEST:
+		case I40IW_AE_ROE_INVALID_RDMA_WRITE_OR_READ_RESP:
+		case I40IW_AE_INVALID_ARP_ENTRY:
+		case I40IW_AE_INVALID_TCP_OPTION_RCVD:
+		case I40IW_AE_STALE_ARP_ENTRY:
+		case I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR:
+		case I40IW_AE_LLP_SEGMENT_TOO_SMALL:
+		case I40IW_AE_LLP_SYN_RECEIVED:
+		case I40IW_AE_LLP_TOO_MANY_RETRIES:
+		case I40IW_AE_LLP_DOUBT_REACHABILITY:
+		case I40IW_AE_LCE_QP_CATASTROPHIC:
+		case I40IW_AE_LCE_FUNCTION_CATASTROPHIC:
+		case I40IW_AE_LCE_CQ_CATASTROPHIC:
+		case I40IW_AE_UDA_XMIT_DGRAM_TOO_LONG:
+		case I40IW_AE_UDA_XMIT_IPADDR_MISMATCH:
+		case I40IW_AE_QP_SUSPEND_COMPLETE:
+			ctx_info->err_rq_idx_valid = false;
+		default:
+				if (!info->sq && ctx_info->err_rq_idx_valid) {
+					ctx_info->err_rq_idx = info->wqe_idx;
+					ctx_info->tcp_info_valid = false;
+					ctx_info->iwarp_info_valid = false;
+					ret = dev->iw_priv_qp_ops->qp_setctx(&iwqp->sc_qp,
+									     iwqp->host_ctx.va,
+									     ctx_info);
+				}
+				i40iw_terminate_connection(qp, info);
+				break;
+		}
+	} while (1);
+
+	if (aeqcnt)
+		dev->aeq_ops->repost_aeq_entries(dev, aeqcnt);
+}
+
+/**
+ * i40iw_manage_apbvt - add or delete tcp port
+ * @iwdev: iwarp device
+ * @accel_local_port: port for apbvt
+ * @add_port: add or delete port
+ */
+int i40iw_manage_apbvt(struct i40iw_device *iwdev, u16 accel_local_port, bool add_port)
+{
+	struct i40iw_apbvt_info *info;
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, add_port);
+	if (!cqp_request)
+		return -ENOMEM;
+
+	cqp_info = &cqp_request->info;
+	info = &cqp_info->in.u.manage_apbvt_entry.info;
+
+	memset(info, 0, sizeof(*info));
+	info->add = add_port;
+	info->port = cpu_to_le16(accel_local_port);
+
+	cqp_info->cqp_cmd = OP_MANAGE_APBVT_ENTRY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.manage_apbvt_entry.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.manage_apbvt_entry.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Manage APBVT entry fail");
+	return status;
+}
+
+/**
+ * i40iw_manage_arp_cache - manage hw arp cache
+ * @iwdev: iwarp device
+ * @mac_addr: mac address ptr
+ * @ip_addr: ip addr for arp cache
+ * @action: add, delete or modify
+ */
+void i40iw_manage_arp_cache(struct i40iw_device *iwdev,
+			    unsigned char *mac_addr,
+			    __be32 *ip_addr,
+			    bool ipv4,
+			    u32 action)
+{
+	struct i40iw_add_arp_cache_entry_info *info;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	int arp_index;
+
+	arp_index = i40iw_arp_table(iwdev, ip_addr, ipv4, mac_addr, action);
+	if (arp_index == -1)
+		return;
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
+	if (!cqp_request)
+		return;
+
+	cqp_info = &cqp_request->info;
+	if (action == I40IW_ARP_ADD) {
+		cqp_info->cqp_cmd = OP_ADD_ARP_CACHE_ENTRY;
+		info = &cqp_info->in.u.add_arp_cache_entry.info;
+		memset(info, 0, sizeof(*info));
+		info->arp_index = cpu_to_le32(arp_index);
+		info->permanent = true;
+		ether_addr_copy(info->mac_addr, mac_addr);
+		cqp_info->in.u.add_arp_cache_entry.scratch = (uintptr_t)cqp_request;
+		cqp_info->in.u.add_arp_cache_entry.cqp = &iwdev->cqp.sc_cqp;
+	} else {
+		cqp_info->cqp_cmd = OP_DELETE_ARP_CACHE_ENTRY;
+		cqp_info->in.u.del_arp_cache_entry.scratch = (uintptr_t)cqp_request;
+		cqp_info->in.u.del_arp_cache_entry.cqp = &iwdev->cqp.sc_cqp;
+		cqp_info->in.u.del_arp_cache_entry.arp_index = arp_index;
+	}
+
+	cqp_info->in.u.add_arp_cache_entry.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.add_arp_cache_entry.scratch = (uintptr_t)cqp_request;
+	cqp_info->post_sq = 1;
+	if (i40iw_handle_cqp_op(iwdev, cqp_request))
+		i40iw_pr_err("CQP-OP Add/Del Arp Cache entry fail");
+}
+
+/**
+ * i40iw_send_syn_cqp_callback - do syn/ack after qhash
+ * @cqp_request: qhash cqp completion
+ * @send_ack: flag send ack
+ */
+static void i40iw_send_syn_cqp_callback(struct i40iw_cqp_request *cqp_request, u32 send_ack)
+{
+	i40iw_send_syn(cqp_request->param, send_ack);
+}
+
+/**
+ * i40iw_manage_qhash - add or modify qhash
+ * @iwdev: iwarp device
+ * @cminfo: cm info for qhash
+ * @etype: type (syn or quad)
+ * @mtype: type of qhash
+ * @cmnode: cmnode associated with connection
+ * @wait: wait for completion
+ * @user_pri:user pri of the connection
+ */
+enum i40iw_status_code i40iw_manage_qhash(struct i40iw_device *iwdev,
+					  struct i40iw_cm_info *cminfo,
+					  enum i40iw_quad_entry_type etype,
+					  enum i40iw_quad_hash_manage_type mtype,
+					  void *cmnode,
+					  bool wait)
+{
+	struct i40iw_qhash_table_info *info;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(iwcqp, wait);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+	cqp_info = &cqp_request->info;
+	info = &cqp_info->in.u.manage_qhash_table_entry.info;
+	memset(info, 0, sizeof(*info));
+
+	info->manage = mtype;
+	info->entry_type = etype;
+	if (cminfo->vlan_id != 0xFFFF) {
+		info->vlan_valid = true;
+		info->vlan_id = cpu_to_le16(cminfo->vlan_id);
+	} else {
+		info->vlan_valid = false;
+	}
+
+	info->ipv4_valid = cminfo->ipv4;
+	ether_addr_copy(info->mac_addr, iwdev->netdev->dev_addr);
+	info->qp_num = cpu_to_le32(dev->ilq->qp_id);
+	info->dest_port = cpu_to_le16(cminfo->loc_port);
+	info->dest_ip[0] = cpu_to_le32(cminfo->loc_addr[0]);
+	info->dest_ip[1] = cpu_to_le32(cminfo->loc_addr[1]);
+	info->dest_ip[2] = cpu_to_le32(cminfo->loc_addr[2]);
+	info->dest_ip[3] = cpu_to_le32(cminfo->loc_addr[3]);
+	if (etype == I40IW_QHASH_TYPE_TCP_ESTABLISHED) {
+		info->src_port = cpu_to_le16(cminfo->rem_port);
+		info->src_ip[0] = cpu_to_le32(cminfo->rem_addr[0]);
+		info->src_ip[1] = cpu_to_le32(cminfo->rem_addr[1]);
+		info->src_ip[2] = cpu_to_le32(cminfo->rem_addr[2]);
+		info->src_ip[3] = cpu_to_le32(cminfo->rem_addr[3]);
+	}
+	if (cmnode) {
+		cqp_request->callback_fcn = i40iw_send_syn_cqp_callback;
+		cqp_request->param = (void *)cmnode;
+	}
+
+	if (info->ipv4_valid)
+		i40iw_debug(dev, I40IW_DEBUG_CM,
+			    "%s:%s IP=%pI4, port=%d, mac=%pM, vlan_id=%d\n",
+			    __func__, (!mtype) ? "DELETE" : "ADD",
+			    info->dest_ip,
+			    info->dest_port, info->mac_addr, cminfo->vlan_id);
+	else
+		i40iw_debug(dev, I40IW_DEBUG_CM,
+			    "%s:%s IP=%pI6, port=%d, mac=%pM, vlan_id=%d\n",
+			    __func__, (!mtype) ? "DELETE" : "ADD",
+			    info->dest_ip,
+			    info->dest_port, info->mac_addr, cminfo->vlan_id);
+	cqp_info->in.u.manage_qhash_table_entry.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.manage_qhash_table_entry.scratch = (uintptr_t)cqp_request;
+	cqp_info->cqp_cmd = OP_MANAGE_QHASH_TABLE_ENTRY;
+	cqp_info->post_sq = 1;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Manage Qhash Entry fail");
+	return status;
+}
+
+/**
+ * i40iw_hw_flush_wqes - flush qp's wqe
+ * @iwdev: iwarp device
+ * @qp: hardware control qp
+ * @info: info for flush
+ * @wait: flag wait for completion
+ */
+enum i40iw_status_code i40iw_hw_flush_wqes(struct i40iw_device *iwdev,
+					   struct i40iw_sc_qp *qp,
+					   struct i40iw_qp_flush_info *info,
+					   bool wait)
+{
+	enum i40iw_status_code status;
+	struct i40iw_qp_flush_info *hw_info;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, wait);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+
+	cqp_info = &cqp_request->info;
+	hw_info = &cqp_request->info.in.u.qp_flush_wqes.info;
+	memcpy(hw_info, info, sizeof(*hw_info));
+
+	cqp_info->cqp_cmd = OP_QP_FLUSH_WQES;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.qp_flush_wqes.qp = qp;
+	cqp_info->in.u.qp_flush_wqes.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Flush WQE's fail");
+	return status;
+}
+
+/**
+ * i40iw_hw_manage_vf_pble_bp - manage vf pbles
+ * @iwdev: iwarp device
+ * @info: info for managing pble
+ * @wait: flag wait for completion
+ */
+enum i40iw_status_code i40iw_hw_manage_vf_pble_bp(struct i40iw_device *iwdev,
+						  struct i40iw_manage_vf_pble_info *info,
+						  bool wait)
+{
+	enum i40iw_status_code status;
+	struct i40iw_manage_vf_pble_info *hw_info;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	if ((iwdev->init_state < CCQ_CREATED) && wait)
+		wait = false;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, wait);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+
+	cqp_info = &cqp_request->info;
+	hw_info = &cqp_request->info.in.u.manage_vf_pble_bp.info;
+	memcpy(hw_info, info, sizeof(*hw_info));
+
+	cqp_info->cqp_cmd = OP_MANAGE_VF_PBLE_BP;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.manage_vf_pble_bp.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.manage_vf_pble_bp.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Manage VF pble_bp fail");
+	return status;
+}
+
+/**
+ * i40iw_get_ib_wc - return change flush code to IB's
+ * @opcode: iwarp flush code
+ */
+static enum ib_wc_status i40iw_get_ib_wc(enum i40iw_flush_opcode opcode)
+{
+	switch (opcode) {
+	case FLUSH_PROT_ERR:
+		return IB_WC_LOC_PROT_ERR;
+	case FLUSH_REM_ACCESS_ERR:
+		return IB_WC_REM_ACCESS_ERR;
+	case FLUSH_LOC_QP_OP_ERR:
+		return IB_WC_LOC_QP_OP_ERR;
+	case FLUSH_REM_OP_ERR:
+		return IB_WC_REM_OP_ERR;
+	case FLUSH_LOC_LEN_ERR:
+		return IB_WC_LOC_LEN_ERR;
+	case FLUSH_GENERAL_ERR:
+		return IB_WC_GENERAL_ERR;
+	case FLUSH_FATAL_ERR:
+	default:
+		return IB_WC_FATAL_ERR;
+	}
+}
+
+/**
+ * i40iw_set_flush_info - set flush info
+ * @pinfo: set flush info
+ * @min: minor err
+ * @maj: major err
+ * @opcode: flush error code
+ */
+static void i40iw_set_flush_info(struct i40iw_qp_flush_info *pinfo,
+				 u16 *min,
+				 u16 *maj,
+				 enum i40iw_flush_opcode opcode)
+{
+	*min = (u16)i40iw_get_ib_wc(opcode);
+	*maj = CQE_MAJOR_DRV;
+	pinfo->userflushcode = true;
+}
+
+/**
+ * i40iw_flush_wqes - flush wqe for qp
+ * @iwdev: iwarp device
+ * @iwqp: qp to flush wqes
+ */
+void i40iw_flush_wqes(struct i40iw_device *iwdev, struct i40iw_qp *iwqp)
+{
+	struct i40iw_qp_flush_info info;
+	struct i40iw_qp_flush_info *pinfo = &info;
+
+	struct i40iw_sc_qp *qp = &iwqp->sc_qp;
+
+	memset(pinfo, 0, sizeof(*pinfo));
+	info.sq = true;
+	info.rq = true;
+	if (qp->term_flags) {
+		i40iw_set_flush_info(pinfo, &pinfo->sq_minor_code,
+				     &pinfo->sq_major_code, qp->flush_code);
+		i40iw_set_flush_info(pinfo, &pinfo->rq_minor_code,
+				     &pinfo->rq_major_code, qp->flush_code);
+	}
+	(void)i40iw_hw_flush_wqes(iwdev, &iwqp->sc_qp, &info, true);
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_main.c b/drivers/infiniband/hw/i40iw/i40iw_main.c
new file mode 100644
index 000000000000..90e5af21737e
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_main.c
@@ -0,0 +1,1910 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/if_vlan.h>
+#include <net/addrconf.h>
+
+#include "i40iw.h"
+#include "i40iw_register.h"
+#include <net/netevent.h>
+#define CLIENT_IW_INTERFACE_VERSION_MAJOR 0
+#define CLIENT_IW_INTERFACE_VERSION_MINOR 01
+#define CLIENT_IW_INTERFACE_VERSION_BUILD 00
+
+#define DRV_VERSION_MAJOR 0
+#define DRV_VERSION_MINOR 5
+#define DRV_VERSION_BUILD 123
+#define DRV_VERSION	__stringify(DRV_VERSION_MAJOR) "."		\
+	__stringify(DRV_VERSION_MINOR) "." __stringify(DRV_VERSION_BUILD)
+
+static int push_mode;
+module_param(push_mode, int, 0644);
+MODULE_PARM_DESC(push_mode, "Low latency mode: 0=disabled (default), 1=enabled)");
+
+static int debug;
+module_param(debug, int, 0644);
+MODULE_PARM_DESC(debug, "debug flags: 0=disabled (default), 0x7fffffff=all");
+
+static int resource_profile;
+module_param(resource_profile, int, 0644);
+MODULE_PARM_DESC(resource_profile,
+		 "Resource Profile: 0=no VF RDMA support (default), 1=Weighted VF, 2=Even Distribution");
+
+static int max_rdma_vfs = 32;
+module_param(max_rdma_vfs, int, 0644);
+MODULE_PARM_DESC(max_rdma_vfs, "Maximum VF count: 0-32 32=default");
+static int mpa_version = 2;
+module_param(mpa_version, int, 0644);
+MODULE_PARM_DESC(mpa_version, "MPA version to be used in MPA Req/Resp 1 or 2");
+
+MODULE_AUTHOR("Intel Corporation, <e1000-rdma@lists.sourceforge.net>");
+MODULE_DESCRIPTION("Intel(R) Ethernet Connection X722 iWARP RDMA Driver");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_VERSION(DRV_VERSION);
+
+static struct i40e_client i40iw_client;
+static char i40iw_client_name[I40E_CLIENT_STR_LENGTH] = "i40iw";
+
+static LIST_HEAD(i40iw_handlers);
+static spinlock_t i40iw_handler_lock;
+
+static enum i40iw_status_code i40iw_virtchnl_send(struct i40iw_sc_dev *dev,
+						  u32 vf_id, u8 *msg, u16 len);
+
+static struct notifier_block i40iw_inetaddr_notifier = {
+	.notifier_call = i40iw_inetaddr_event
+};
+
+static struct notifier_block i40iw_inetaddr6_notifier = {
+	.notifier_call = i40iw_inet6addr_event
+};
+
+static struct notifier_block i40iw_net_notifier = {
+	.notifier_call = i40iw_net_event
+};
+
+static int i40iw_notifiers_registered;
+
+/**
+ * i40iw_find_i40e_handler - find a handler given a client info
+ * @ldev: pointer to a client info
+ */
+static struct i40iw_handler *i40iw_find_i40e_handler(struct i40e_info *ldev)
+{
+	struct i40iw_handler *hdl;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i40iw_handler_lock, flags);
+	list_for_each_entry(hdl, &i40iw_handlers, list) {
+		if (hdl->ldev.netdev == ldev->netdev) {
+			spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+			return hdl;
+		}
+	}
+	spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+	return NULL;
+}
+
+/**
+ * i40iw_find_netdev - find a handler given a netdev
+ * @netdev: pointer to net_device
+ */
+struct i40iw_handler *i40iw_find_netdev(struct net_device *netdev)
+{
+	struct i40iw_handler *hdl;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i40iw_handler_lock, flags);
+	list_for_each_entry(hdl, &i40iw_handlers, list) {
+		if (hdl->ldev.netdev == netdev) {
+			spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+			return hdl;
+		}
+	}
+	spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+	return NULL;
+}
+
+/**
+ * i40iw_add_handler - add a handler to the list
+ * @hdl: handler to be added to the handler list
+ */
+static void i40iw_add_handler(struct i40iw_handler *hdl)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&i40iw_handler_lock, flags);
+	list_add(&hdl->list, &i40iw_handlers);
+	spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+}
+
+/**
+ * i40iw_del_handler - delete a handler from the list
+ * @hdl: handler to be deleted from the handler list
+ */
+static int i40iw_del_handler(struct i40iw_handler *hdl)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&i40iw_handler_lock, flags);
+	list_del(&hdl->list);
+	spin_unlock_irqrestore(&i40iw_handler_lock, flags);
+	return 0;
+}
+
+/**
+ * i40iw_enable_intr - set up device interrupts
+ * @dev: hardware control device structure
+ * @msix_id: id of the interrupt to be enabled
+ */
+static void i40iw_enable_intr(struct i40iw_sc_dev *dev, u32 msix_id)
+{
+	u32 val;
+
+	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
+		I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
+		(3 << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
+	if (dev->is_pf)
+		i40iw_wr32(dev->hw, I40E_PFINT_DYN_CTLN(msix_id - 1), val);
+	else
+		i40iw_wr32(dev->hw, I40E_VFINT_DYN_CTLN1(msix_id - 1), val);
+}
+
+/**
+ * i40iw_dpc - tasklet for aeq and ceq 0
+ * @data: iwarp device
+ */
+static void i40iw_dpc(unsigned long data)
+{
+	struct i40iw_device *iwdev = (struct i40iw_device *)data;
+
+	if (iwdev->msix_shared)
+		i40iw_process_ceq(iwdev, iwdev->ceqlist);
+	i40iw_process_aeq(iwdev);
+	i40iw_enable_intr(&iwdev->sc_dev, iwdev->iw_msixtbl[0].idx);
+}
+
+/**
+ * i40iw_ceq_dpc - dpc handler for CEQ
+ * @data: data points to CEQ
+ */
+static void i40iw_ceq_dpc(unsigned long data)
+{
+	struct i40iw_ceq *iwceq = (struct i40iw_ceq *)data;
+	struct i40iw_device *iwdev = iwceq->iwdev;
+
+	i40iw_process_ceq(iwdev, iwceq);
+	i40iw_enable_intr(&iwdev->sc_dev, iwceq->msix_idx);
+}
+
+/**
+ * i40iw_irq_handler - interrupt handler for aeq and ceq0
+ * @irq: Interrupt request number
+ * @data: iwarp device
+ */
+static irqreturn_t i40iw_irq_handler(int irq, void *data)
+{
+	struct i40iw_device *iwdev = (struct i40iw_device *)data;
+
+	tasklet_schedule(&iwdev->dpc_tasklet);
+	return IRQ_HANDLED;
+}
+
+/**
+ * i40iw_destroy_cqp  - destroy control qp
+ * @iwdev: iwarp device
+ * @create_done: 1 if cqp create poll was success
+ *
+ * Issue destroy cqp request and
+ * free the resources associated with the cqp
+ */
+static void i40iw_destroy_cqp(struct i40iw_device *iwdev, bool free_hwcqp)
+{
+	enum i40iw_status_code status = 0;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_cqp *cqp = &iwdev->cqp;
+
+	if (free_hwcqp && dev->cqp_ops->cqp_destroy)
+		status = dev->cqp_ops->cqp_destroy(dev->cqp);
+	if (status)
+		i40iw_pr_err("destroy cqp failed");
+
+	i40iw_free_dma_mem(dev->hw, &cqp->sq);
+	kfree(cqp->scratch_array);
+	iwdev->cqp.scratch_array = NULL;
+
+	kfree(cqp->cqp_requests);
+	cqp->cqp_requests = NULL;
+}
+
+/**
+ * i40iw_disable_irqs - disable device interrupts
+ * @dev: hardware control device structure
+ * @msic_vec: msix vector to disable irq
+ * @dev_id: parameter to pass to free_irq (used during irq setup)
+ *
+ * The function is called when destroying aeq/ceq
+ */
+static void i40iw_disable_irq(struct i40iw_sc_dev *dev,
+			      struct i40iw_msix_vector *msix_vec,
+			      void *dev_id)
+{
+	if (dev->is_pf)
+		i40iw_wr32(dev->hw, I40E_PFINT_DYN_CTLN(msix_vec->idx - 1), 0);
+	else
+		i40iw_wr32(dev->hw, I40E_VFINT_DYN_CTLN1(msix_vec->idx - 1), 0);
+	synchronize_irq(msix_vec->irq);
+	free_irq(msix_vec->irq, dev_id);
+}
+
+/**
+ * i40iw_destroy_aeq - destroy aeq
+ * @iwdev: iwarp device
+ * @reset: true if called before reset
+ *
+ * Issue a destroy aeq request and
+ * free the resources associated with the aeq
+ * The function is called during driver unload
+ */
+static void i40iw_destroy_aeq(struct i40iw_device *iwdev, bool reset)
+{
+	enum i40iw_status_code status = I40IW_ERR_NOT_READY;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_aeq *aeq = &iwdev->aeq;
+
+	if (!iwdev->msix_shared)
+		i40iw_disable_irq(dev, iwdev->iw_msixtbl, (void *)iwdev);
+	if (reset)
+		goto exit;
+
+	if (!dev->aeq_ops->aeq_destroy(&aeq->sc_aeq, 0, 1))
+		status = dev->aeq_ops->aeq_destroy_done(&aeq->sc_aeq);
+	if (status)
+		i40iw_pr_err("destroy aeq failed %d\n", status);
+
+exit:
+	i40iw_free_dma_mem(dev->hw, &aeq->mem);
+}
+
+/**
+ * i40iw_destroy_ceq - destroy ceq
+ * @iwdev: iwarp device
+ * @iwceq: ceq to be destroyed
+ * @reset: true if called before reset
+ *
+ * Issue a destroy ceq request and
+ * free the resources associated with the ceq
+ */
+static void i40iw_destroy_ceq(struct i40iw_device *iwdev,
+			      struct i40iw_ceq *iwceq,
+			      bool reset)
+{
+	enum i40iw_status_code status;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+
+	if (reset)
+		goto exit;
+
+	status = dev->ceq_ops->ceq_destroy(&iwceq->sc_ceq, 0, 1);
+	if (status) {
+		i40iw_pr_err("ceq destroy command failed %d\n", status);
+		goto exit;
+	}
+
+	status = dev->ceq_ops->cceq_destroy_done(&iwceq->sc_ceq);
+	if (status)
+		i40iw_pr_err("ceq destroy completion failed %d\n", status);
+exit:
+	i40iw_free_dma_mem(dev->hw, &iwceq->mem);
+}
+
+/**
+ * i40iw_dele_ceqs - destroy all ceq's
+ * @iwdev: iwarp device
+ * @reset: true if called before reset
+ *
+ * Go through all of the device ceq's and for each ceq
+ * disable the ceq interrupt and destroy the ceq
+ */
+static void i40iw_dele_ceqs(struct i40iw_device *iwdev, bool reset)
+{
+	u32 i = 0;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_ceq *iwceq = iwdev->ceqlist;
+	struct i40iw_msix_vector *msix_vec = iwdev->iw_msixtbl;
+
+	if (iwdev->msix_shared) {
+		i40iw_disable_irq(dev, msix_vec, (void *)iwdev);
+		i40iw_destroy_ceq(iwdev, iwceq, reset);
+		iwceq++;
+		i++;
+	}
+
+	for (msix_vec++; i < iwdev->ceqs_count; i++, msix_vec++, iwceq++) {
+		i40iw_disable_irq(dev, msix_vec, (void *)iwceq);
+		i40iw_destroy_ceq(iwdev, iwceq, reset);
+	}
+}
+
+/**
+ * i40iw_destroy_ccq - destroy control cq
+ * @iwdev: iwarp device
+ * @reset: true if called before reset
+ *
+ * Issue destroy ccq request and
+ * free the resources associated with the ccq
+ */
+static void i40iw_destroy_ccq(struct i40iw_device *iwdev, bool reset)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_ccq *ccq = &iwdev->ccq;
+	enum i40iw_status_code status = 0;
+
+	if (!reset)
+		status = dev->ccq_ops->ccq_destroy(dev->ccq, 0, true);
+	if (status)
+		i40iw_pr_err("ccq destroy failed %d\n", status);
+	i40iw_free_dma_mem(dev->hw, &ccq->mem_cq);
+}
+
+/* types of hmc objects */
+static enum i40iw_hmc_rsrc_type iw_hmc_obj_types[] = {
+	I40IW_HMC_IW_QP,
+	I40IW_HMC_IW_CQ,
+	I40IW_HMC_IW_HTE,
+	I40IW_HMC_IW_ARP,
+	I40IW_HMC_IW_APBVT_ENTRY,
+	I40IW_HMC_IW_MR,
+	I40IW_HMC_IW_XF,
+	I40IW_HMC_IW_XFFL,
+	I40IW_HMC_IW_Q1,
+	I40IW_HMC_IW_Q1FL,
+	I40IW_HMC_IW_TIMER,
+};
+
+/**
+ * i40iw_close_hmc_objects_type - delete hmc objects of a given type
+ * @iwdev: iwarp device
+ * @obj_type: the hmc object type to be deleted
+ * @is_pf: true if the function is PF otherwise false
+ * @reset: true if called before reset
+ */
+static void i40iw_close_hmc_objects_type(struct i40iw_sc_dev *dev,
+					 enum i40iw_hmc_rsrc_type obj_type,
+					 struct i40iw_hmc_info *hmc_info,
+					 bool is_pf,
+					 bool reset)
+{
+	struct i40iw_hmc_del_obj_info info;
+
+	memset(&info, 0, sizeof(info));
+	info.hmc_info = hmc_info;
+	info.rsrc_type = obj_type;
+	info.count = hmc_info->hmc_obj[obj_type].cnt;
+	info.is_pf = is_pf;
+	if (dev->hmc_ops->del_hmc_object(dev, &info, reset))
+		i40iw_pr_err("del obj of type %d failed\n", obj_type);
+}
+
+/**
+ * i40iw_del_hmc_objects - remove all device hmc objects
+ * @dev: iwarp device
+ * @hmc_info: hmc_info to free
+ * @is_pf: true if hmc_info belongs to PF, not vf nor allocated
+ *	   by PF on behalf of VF
+ * @reset: true if called before reset
+ */
+static void i40iw_del_hmc_objects(struct i40iw_sc_dev *dev,
+				  struct i40iw_hmc_info *hmc_info,
+				  bool is_pf,
+				  bool reset)
+{
+	unsigned int i;
+
+	for (i = 0; i < IW_HMC_OBJ_TYPE_NUM; i++)
+		i40iw_close_hmc_objects_type(dev, iw_hmc_obj_types[i], hmc_info, is_pf, reset);
+}
+
+/**
+ * i40iw_ceq_handler - interrupt handler for ceq
+ * @data: ceq pointer
+ */
+static irqreturn_t i40iw_ceq_handler(int irq, void *data)
+{
+	struct i40iw_ceq *iwceq = (struct i40iw_ceq *)data;
+
+	if (iwceq->irq != irq)
+		i40iw_pr_err("expected irq = %d received irq = %d\n", iwceq->irq, irq);
+	tasklet_schedule(&iwceq->dpc_tasklet);
+	return IRQ_HANDLED;
+}
+
+/**
+ * i40iw_create_hmc_obj_type - create hmc object of a given type
+ * @dev: hardware control device structure
+ * @info: information for the hmc object to create
+ */
+static enum i40iw_status_code i40iw_create_hmc_obj_type(struct i40iw_sc_dev *dev,
+							struct i40iw_hmc_create_obj_info *info)
+{
+	return dev->hmc_ops->create_hmc_object(dev, info);
+}
+
+/**
+ * i40iw_create_hmc_objs - create all hmc objects for the device
+ * @iwdev: iwarp device
+ * @is_pf: true if the function is PF otherwise false
+ *
+ * Create the device hmc objects and allocate hmc pages
+ * Return 0 if successful, otherwise clean up and return error
+ */
+static enum i40iw_status_code i40iw_create_hmc_objs(struct i40iw_device *iwdev,
+						    bool is_pf)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_hmc_create_obj_info info;
+	enum i40iw_status_code status;
+	int i;
+
+	memset(&info, 0, sizeof(info));
+	info.hmc_info = dev->hmc_info;
+	info.is_pf = is_pf;
+	info.entry_type = iwdev->sd_type;
+	for (i = 0; i < IW_HMC_OBJ_TYPE_NUM; i++) {
+		info.rsrc_type = iw_hmc_obj_types[i];
+		info.count = dev->hmc_info->hmc_obj[info.rsrc_type].cnt;
+		status = i40iw_create_hmc_obj_type(dev, &info);
+		if (status) {
+			i40iw_pr_err("create obj type %d status = %d\n",
+				     iw_hmc_obj_types[i], status);
+			break;
+		}
+	}
+	if (!status)
+		return (dev->cqp_misc_ops->static_hmc_pages_allocated(dev->cqp, 0,
+								      dev->hmc_fn_id,
+								      true, true));
+
+	while (i) {
+		i--;
+		/* destroy the hmc objects of a given type */
+		i40iw_close_hmc_objects_type(dev,
+					     iw_hmc_obj_types[i],
+					     dev->hmc_info,
+					     is_pf,
+					     false);
+	}
+	return status;
+}
+
+/**
+ * i40iw_obj_aligned_mem - get aligned memory from device allocated memory
+ * @iwdev: iwarp device
+ * @memptr: points to the memory addresses
+ * @size: size of memory needed
+ * @mask: mask for the aligned memory
+ *
+ * Get aligned memory of the requested size and
+ * update the memptr to point to the new aligned memory
+ * Return 0 if successful, otherwise return no memory error
+ */
+enum i40iw_status_code i40iw_obj_aligned_mem(struct i40iw_device *iwdev,
+					     struct i40iw_dma_mem *memptr,
+					     u32 size,
+					     u32 mask)
+{
+	unsigned long va, newva;
+	unsigned long extra;
+
+	va = (unsigned long)iwdev->obj_next.va;
+	newva = va;
+	if (mask)
+		newva = ALIGN(va, (mask + 1));
+	extra = newva - va;
+	memptr->va = (u8 *)va + extra;
+	memptr->pa = iwdev->obj_next.pa + extra;
+	memptr->size = size;
+	if ((memptr->va + size) > (iwdev->obj_mem.va + iwdev->obj_mem.size))
+		return I40IW_ERR_NO_MEMORY;
+
+	iwdev->obj_next.va = memptr->va + size;
+	iwdev->obj_next.pa = memptr->pa + size;
+	return 0;
+}
+
+/**
+ * i40iw_create_cqp - create control qp
+ * @iwdev: iwarp device
+ *
+ * Return 0, if the cqp and all the resources associated with it
+ * are successfully created, otherwise return error
+ */
+static enum i40iw_status_code i40iw_create_cqp(struct i40iw_device *iwdev)
+{
+	enum i40iw_status_code status;
+	u32 sqsize = I40IW_CQP_SW_SQSIZE_2048;
+	struct i40iw_dma_mem mem;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_cqp_init_info cqp_init_info;
+	struct i40iw_cqp *cqp = &iwdev->cqp;
+	u16 maj_err, min_err;
+	int i;
+
+	cqp->cqp_requests = kcalloc(sqsize, sizeof(*cqp->cqp_requests), GFP_KERNEL);
+	if (!cqp->cqp_requests)
+		return I40IW_ERR_NO_MEMORY;
+	cqp->scratch_array = kcalloc(sqsize, sizeof(*cqp->scratch_array), GFP_KERNEL);
+	if (!cqp->scratch_array) {
+		kfree(cqp->cqp_requests);
+		return I40IW_ERR_NO_MEMORY;
+	}
+	dev->cqp = &cqp->sc_cqp;
+	dev->cqp->dev = dev;
+	memset(&cqp_init_info, 0, sizeof(cqp_init_info));
+	status = i40iw_allocate_dma_mem(dev->hw, &cqp->sq,
+					(sizeof(struct i40iw_cqp_sq_wqe) * sqsize),
+					I40IW_CQP_ALIGNMENT);
+	if (status)
+		goto exit;
+	status = i40iw_obj_aligned_mem(iwdev, &mem, sizeof(struct i40iw_cqp_ctx),
+				       I40IW_HOST_CTX_ALIGNMENT_MASK);
+	if (status)
+		goto exit;
+	dev->cqp->host_ctx_pa = mem.pa;
+	dev->cqp->host_ctx = mem.va;
+	/* populate the cqp init info */
+	cqp_init_info.dev = dev;
+	cqp_init_info.sq_size = sqsize;
+	cqp_init_info.sq = cqp->sq.va;
+	cqp_init_info.sq_pa = cqp->sq.pa;
+	cqp_init_info.host_ctx_pa = mem.pa;
+	cqp_init_info.host_ctx = mem.va;
+	cqp_init_info.hmc_profile = iwdev->resource_profile;
+	cqp_init_info.enabled_vf_count = iwdev->max_rdma_vfs;
+	cqp_init_info.scratch_array = cqp->scratch_array;
+	status = dev->cqp_ops->cqp_init(dev->cqp, &cqp_init_info);
+	if (status) {
+		i40iw_pr_err("cqp init status %d maj_err %d min_err %d\n",
+			     status, maj_err, min_err);
+		goto exit;
+	}
+	status = dev->cqp_ops->cqp_create(dev->cqp, true, &maj_err, &min_err);
+	if (status) {
+		i40iw_pr_err("cqp create status %d maj_err %d min_err %d\n",
+			     status, maj_err, min_err);
+		goto exit;
+	}
+	spin_lock_init(&cqp->req_lock);
+	INIT_LIST_HEAD(&cqp->cqp_avail_reqs);
+	INIT_LIST_HEAD(&cqp->cqp_pending_reqs);
+	/* init the waitq of the cqp_requests and add them to the list */
+	for (i = 0; i < I40IW_CQP_SW_SQSIZE_2048; i++) {
+		init_waitqueue_head(&cqp->cqp_requests[i].waitq);
+		list_add_tail(&cqp->cqp_requests[i].list, &cqp->cqp_avail_reqs);
+	}
+	return 0;
+exit:
+	/* clean up the created resources */
+	i40iw_destroy_cqp(iwdev, false);
+	return status;
+}
+
+/**
+ * i40iw_create_ccq - create control cq
+ * @iwdev: iwarp device
+ *
+ * Return 0, if the ccq and the resources associated with it
+ * are successfully created, otherwise return error
+ */
+static enum i40iw_status_code i40iw_create_ccq(struct i40iw_device *iwdev)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_dma_mem mem;
+	enum i40iw_status_code status;
+	struct i40iw_ccq_init_info info;
+	struct i40iw_ccq *ccq = &iwdev->ccq;
+
+	memset(&info, 0, sizeof(info));
+	dev->ccq = &ccq->sc_cq;
+	dev->ccq->dev = dev;
+	info.dev = dev;
+	ccq->shadow_area.size = sizeof(struct i40iw_cq_shadow_area);
+	ccq->mem_cq.size = sizeof(struct i40iw_cqe) * IW_CCQ_SIZE;
+	status = i40iw_allocate_dma_mem(dev->hw, &ccq->mem_cq,
+					ccq->mem_cq.size, I40IW_CQ0_ALIGNMENT);
+	if (status)
+		goto exit;
+	status = i40iw_obj_aligned_mem(iwdev, &mem, ccq->shadow_area.size,
+				       I40IW_SHADOWAREA_MASK);
+	if (status)
+		goto exit;
+	ccq->sc_cq.back_cq = (void *)ccq;
+	/* populate the ccq init info */
+	info.cq_base = ccq->mem_cq.va;
+	info.cq_pa = ccq->mem_cq.pa;
+	info.num_elem = IW_CCQ_SIZE;
+	info.shadow_area = mem.va;
+	info.shadow_area_pa = mem.pa;
+	info.ceqe_mask = false;
+	info.ceq_id_valid = true;
+	info.shadow_read_threshold = 16;
+	status = dev->ccq_ops->ccq_init(dev->ccq, &info);
+	if (!status)
+		status = dev->ccq_ops->ccq_create(dev->ccq, 0, true, true);
+exit:
+	if (status)
+		i40iw_free_dma_mem(dev->hw, &ccq->mem_cq);
+	return status;
+}
+
+/**
+ * i40iw_configure_ceq_vector - set up the msix interrupt vector for ceq
+ * @iwdev: iwarp device
+ * @msix_vec: interrupt vector information
+ * @iwceq: ceq associated with the vector
+ * @ceq_id: the id number of the iwceq
+ *
+ * Allocate interrupt resources and enable irq handling
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_configure_ceq_vector(struct i40iw_device *iwdev,
+							 struct i40iw_ceq *iwceq,
+							 u32 ceq_id,
+							 struct i40iw_msix_vector *msix_vec)
+{
+	enum i40iw_status_code status;
+
+	if (iwdev->msix_shared && !ceq_id) {
+		tasklet_init(&iwdev->dpc_tasklet, i40iw_dpc, (unsigned long)iwdev);
+		status = request_irq(msix_vec->irq, i40iw_irq_handler, 0, "AEQCEQ", iwdev);
+	} else {
+		tasklet_init(&iwceq->dpc_tasklet, i40iw_ceq_dpc, (unsigned long)iwceq);
+		status = request_irq(msix_vec->irq, i40iw_ceq_handler, 0, "CEQ", iwceq);
+	}
+
+	if (status) {
+		i40iw_pr_err("ceq irq config fail\n");
+		return I40IW_ERR_CONFIG;
+	}
+	msix_vec->ceq_id = ceq_id;
+	msix_vec->cpu_affinity = 0;
+
+	return 0;
+}
+
+/**
+ * i40iw_create_ceq - create completion event queue
+ * @iwdev: iwarp device
+ * @iwceq: pointer to the ceq resources to be created
+ * @ceq_id: the id number of the iwceq
+ *
+ * Return 0, if the ceq and the resources associated with it
+ * are successfully created, otherwise return error
+ */
+static enum i40iw_status_code i40iw_create_ceq(struct i40iw_device *iwdev,
+					       struct i40iw_ceq *iwceq,
+					       u32 ceq_id)
+{
+	enum i40iw_status_code status;
+	struct i40iw_ceq_init_info info;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	u64 scratch;
+
+	memset(&info, 0, sizeof(info));
+	info.ceq_id = ceq_id;
+	iwceq->iwdev = iwdev;
+	iwceq->mem.size = sizeof(struct i40iw_ceqe) *
+		iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt;
+	status = i40iw_allocate_dma_mem(dev->hw, &iwceq->mem, iwceq->mem.size,
+					I40IW_CEQ_ALIGNMENT);
+	if (status)
+		goto exit;
+	info.ceq_id = ceq_id;
+	info.ceqe_base = iwceq->mem.va;
+	info.ceqe_pa = iwceq->mem.pa;
+
+	info.elem_cnt = iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt;
+	iwceq->sc_ceq.ceq_id = ceq_id;
+	info.dev = dev;
+	scratch = (uintptr_t)&iwdev->cqp.sc_cqp;
+	status = dev->ceq_ops->ceq_init(&iwceq->sc_ceq, &info);
+	if (!status)
+		status = dev->ceq_ops->cceq_create(&iwceq->sc_ceq, scratch);
+
+exit:
+	if (status)
+		i40iw_free_dma_mem(dev->hw, &iwceq->mem);
+	return status;
+}
+
+void i40iw_request_reset(struct i40iw_device *iwdev)
+{
+	struct i40e_info *ldev = iwdev->ldev;
+
+	ldev->ops->request_reset(ldev, iwdev->client, 1);
+}
+
+/**
+ * i40iw_setup_ceqs - manage the device ceq's and their interrupt resources
+ * @iwdev: iwarp device
+ * @ldev: i40e lan device
+ *
+ * Allocate a list for all device completion event queues
+ * Create the ceq's and configure their msix interrupt vectors
+ * Return 0, if at least one ceq is successfully set up, otherwise return error
+ */
+static enum i40iw_status_code i40iw_setup_ceqs(struct i40iw_device *iwdev,
+					       struct i40e_info *ldev)
+{
+	u32 i;
+	u32 ceq_id;
+	struct i40iw_ceq *iwceq;
+	struct i40iw_msix_vector *msix_vec;
+	enum i40iw_status_code status = 0;
+	u32 num_ceqs;
+
+	if (ldev && ldev->ops && ldev->ops->setup_qvlist) {
+		status = ldev->ops->setup_qvlist(ldev, &i40iw_client,
+						 iwdev->iw_qvlist);
+		if (status)
+			goto exit;
+	} else {
+		status = I40IW_ERR_BAD_PTR;
+		goto exit;
+	}
+
+	num_ceqs = min(iwdev->msix_count, iwdev->sc_dev.hmc_fpm_misc.max_ceqs);
+	iwdev->ceqlist = kcalloc(num_ceqs, sizeof(*iwdev->ceqlist), GFP_KERNEL);
+	if (!iwdev->ceqlist) {
+		status = I40IW_ERR_NO_MEMORY;
+		goto exit;
+	}
+	i = (iwdev->msix_shared) ? 0 : 1;
+	for (ceq_id = 0; i < num_ceqs; i++, ceq_id++) {
+		iwceq = &iwdev->ceqlist[ceq_id];
+		status = i40iw_create_ceq(iwdev, iwceq, ceq_id);
+		if (status) {
+			i40iw_pr_err("create ceq status = %d\n", status);
+			break;
+		}
+
+		msix_vec = &iwdev->iw_msixtbl[i];
+		iwceq->irq = msix_vec->irq;
+		iwceq->msix_idx = msix_vec->idx;
+		status = i40iw_configure_ceq_vector(iwdev, iwceq, ceq_id, msix_vec);
+		if (status) {
+			i40iw_destroy_ceq(iwdev, iwceq, false);
+			break;
+		}
+		i40iw_enable_intr(&iwdev->sc_dev, msix_vec->idx);
+		iwdev->ceqs_count++;
+	}
+
+exit:
+	if (status) {
+		if (!iwdev->ceqs_count) {
+			kfree(iwdev->ceqlist);
+			iwdev->ceqlist = NULL;
+		} else {
+			status = 0;
+		}
+	}
+	return status;
+}
+
+/**
+ * i40iw_configure_aeq_vector - set up the msix vector for aeq
+ * @iwdev: iwarp device
+ *
+ * Allocate interrupt resources and enable irq handling
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_configure_aeq_vector(struct i40iw_device *iwdev)
+{
+	struct i40iw_msix_vector *msix_vec = iwdev->iw_msixtbl;
+	u32 ret = 0;
+
+	if (!iwdev->msix_shared) {
+		tasklet_init(&iwdev->dpc_tasklet, i40iw_dpc, (unsigned long)iwdev);
+		ret = request_irq(msix_vec->irq, i40iw_irq_handler, 0, "i40iw", iwdev);
+	}
+	if (ret) {
+		i40iw_pr_err("aeq irq config fail\n");
+		return I40IW_ERR_CONFIG;
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_create_aeq - create async event queue
+ * @iwdev: iwarp device
+ *
+ * Return 0, if the aeq and the resources associated with it
+ * are successfully created, otherwise return error
+ */
+static enum i40iw_status_code i40iw_create_aeq(struct i40iw_device *iwdev)
+{
+	enum i40iw_status_code status;
+	struct i40iw_aeq_init_info info;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_aeq *aeq = &iwdev->aeq;
+	u64 scratch = 0;
+	u32 aeq_size;
+
+	aeq_size = 2 * iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_QP].cnt +
+		iwdev->sc_dev.hmc_info->hmc_obj[I40IW_HMC_IW_CQ].cnt;
+	memset(&info, 0, sizeof(info));
+	aeq->mem.size = sizeof(struct i40iw_sc_aeqe) * aeq_size;
+	status = i40iw_allocate_dma_mem(dev->hw, &aeq->mem, aeq->mem.size,
+					I40IW_AEQ_ALIGNMENT);
+	if (status)
+		goto exit;
+
+	info.aeqe_base = aeq->mem.va;
+	info.aeq_elem_pa = aeq->mem.pa;
+	info.elem_cnt = aeq_size;
+	info.dev = dev;
+	status = dev->aeq_ops->aeq_init(&aeq->sc_aeq, &info);
+	if (status)
+		goto exit;
+	status = dev->aeq_ops->aeq_create(&aeq->sc_aeq, scratch, 1);
+	if (!status)
+		status = dev->aeq_ops->aeq_create_done(&aeq->sc_aeq);
+exit:
+	if (status)
+		i40iw_free_dma_mem(dev->hw, &aeq->mem);
+	return status;
+}
+
+/**
+ * i40iw_setup_aeq - set up the device aeq
+ * @iwdev: iwarp device
+ *
+ * Create the aeq and configure its msix interrupt vector
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_setup_aeq(struct i40iw_device *iwdev)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+
+	status = i40iw_create_aeq(iwdev);
+	if (status)
+		return status;
+
+	status = i40iw_configure_aeq_vector(iwdev);
+	if (status) {
+		i40iw_destroy_aeq(iwdev, false);
+		return status;
+	}
+
+	if (!iwdev->msix_shared)
+		i40iw_enable_intr(dev, iwdev->iw_msixtbl[0].idx);
+	return 0;
+}
+
+/**
+ * i40iw_initialize_ilq - create iwarp local queue for cm
+ * @iwdev: iwarp device
+ *
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_initialize_ilq(struct i40iw_device *iwdev)
+{
+	struct i40iw_puda_rsrc_info info;
+	enum i40iw_status_code status;
+
+	info.type = I40IW_PUDA_RSRC_TYPE_ILQ;
+	info.cq_id = 1;
+	info.qp_id = 0;
+	info.count = 1;
+	info.pd_id = 1;
+	info.sq_size = 8192;
+	info.rq_size = 8192;
+	info.buf_size = 1024;
+	info.tx_buf_cnt = 16384;
+	info.mss = iwdev->mss;
+	info.receive = i40iw_receive_ilq;
+	info.xmit_complete = i40iw_free_sqbuf;
+	status = i40iw_puda_create_rsrc(&iwdev->sc_dev, &info);
+	if (status)
+		i40iw_pr_err("ilq create fail\n");
+	return status;
+}
+
+/**
+ * i40iw_initialize_ieq - create iwarp exception queue
+ * @iwdev: iwarp device
+ *
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_initialize_ieq(struct i40iw_device *iwdev)
+{
+	struct i40iw_puda_rsrc_info info;
+	enum i40iw_status_code status;
+
+	info.type = I40IW_PUDA_RSRC_TYPE_IEQ;
+	info.cq_id = 2;
+	info.qp_id = iwdev->sc_dev.exception_lan_queue;
+	info.count = 1;
+	info.pd_id = 2;
+	info.sq_size = 8192;
+	info.rq_size = 8192;
+	info.buf_size = 2048;
+	info.mss = iwdev->mss;
+	info.tx_buf_cnt = 16384;
+	status = i40iw_puda_create_rsrc(&iwdev->sc_dev, &info);
+	if (status)
+		i40iw_pr_err("ieq create fail\n");
+	return status;
+}
+
+/**
+ * i40iw_hmc_setup - create hmc objects for the device
+ * @iwdev: iwarp device
+ *
+ * Set up the device private memory space for the number and size of
+ * the hmc objects and create the objects
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_hmc_setup(struct i40iw_device *iwdev)
+{
+	enum i40iw_status_code status;
+
+	iwdev->sd_type = I40IW_SD_TYPE_DIRECT;
+	status = i40iw_config_fpm_values(&iwdev->sc_dev, IW_CFG_FPM_QP_COUNT);
+	if (status)
+		goto exit;
+	status = i40iw_create_hmc_objs(iwdev, true);
+	if (status)
+		goto exit;
+	iwdev->init_state = HMC_OBJS_CREATED;
+exit:
+	return status;
+}
+
+/**
+ * i40iw_del_init_mem - deallocate memory resources
+ * @iwdev: iwarp device
+ */
+static void i40iw_del_init_mem(struct i40iw_device *iwdev)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+
+	i40iw_free_dma_mem(&iwdev->hw, &iwdev->obj_mem);
+	kfree(dev->hmc_info->sd_table.sd_entry);
+	dev->hmc_info->sd_table.sd_entry = NULL;
+	kfree(iwdev->mem_resources);
+	iwdev->mem_resources = NULL;
+	kfree(iwdev->ceqlist);
+	iwdev->ceqlist = NULL;
+	kfree(iwdev->iw_msixtbl);
+	iwdev->iw_msixtbl = NULL;
+	kfree(iwdev->hmc_info_mem);
+	iwdev->hmc_info_mem = NULL;
+}
+
+/**
+ * i40iw_del_macip_entry - remove a mac ip address entry from the hw table
+ * @iwdev: iwarp device
+ * @idx: the index of the mac ip address to delete
+ */
+static void i40iw_del_macip_entry(struct i40iw_device *iwdev, u8 idx)
+{
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	enum i40iw_status_code status = 0;
+
+	cqp_request = i40iw_get_cqp_request(iwcqp, true);
+	if (!cqp_request) {
+		i40iw_pr_err("cqp_request memory failed\n");
+		return;
+	}
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_DELETE_LOCAL_MAC_IPADDR_ENTRY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.del_local_mac_ipaddr_entry.cqp = &iwcqp->sc_cqp;
+	cqp_info->in.u.del_local_mac_ipaddr_entry.scratch = (uintptr_t)cqp_request;
+	cqp_info->in.u.del_local_mac_ipaddr_entry.entry_idx = idx;
+	cqp_info->in.u.del_local_mac_ipaddr_entry.ignore_ref_count = 0;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Del MAC Ip entry fail");
+}
+
+/**
+ * i40iw_add_mac_ipaddr_entry - add a mac ip address entry to the hw table
+ * @iwdev: iwarp device
+ * @mac_addr: pointer to mac address
+ * @idx: the index of the mac ip address to add
+ */
+static enum i40iw_status_code i40iw_add_mac_ipaddr_entry(struct i40iw_device *iwdev,
+							 u8 *mac_addr,
+							 u8 idx)
+{
+	struct i40iw_local_mac_ipaddr_entry_info *info;
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	enum i40iw_status_code status = 0;
+
+	cqp_request = i40iw_get_cqp_request(iwcqp, true);
+	if (!cqp_request) {
+		i40iw_pr_err("cqp_request memory failed\n");
+		return I40IW_ERR_NO_MEMORY;
+	}
+
+	cqp_info = &cqp_request->info;
+
+	cqp_info->post_sq = 1;
+	info = &cqp_info->in.u.add_local_mac_ipaddr_entry.info;
+	ether_addr_copy(info->mac_addr, mac_addr);
+	info->entry_idx = idx;
+	cqp_info->in.u.add_local_mac_ipaddr_entry.scratch = (uintptr_t)cqp_request;
+	cqp_info->cqp_cmd = OP_ADD_LOCAL_MAC_IPADDR_ENTRY;
+	cqp_info->in.u.add_local_mac_ipaddr_entry.cqp = &iwcqp->sc_cqp;
+	cqp_info->in.u.add_local_mac_ipaddr_entry.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Add MAC Ip entry fail");
+	return status;
+}
+
+/**
+ * i40iw_alloc_local_mac_ipaddr_entry - allocate a mac ip address entry
+ * @iwdev: iwarp device
+ * @mac_ip_tbl_idx: the index of the new mac ip address
+ *
+ * Allocate a mac ip address entry and update the mac_ip_tbl_idx
+ * to hold the index of the newly created mac ip address
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_alloc_local_mac_ipaddr_entry(struct i40iw_device *iwdev,
+								 u16 *mac_ip_tbl_idx)
+{
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	enum i40iw_status_code status = 0;
+
+	cqp_request = i40iw_get_cqp_request(iwcqp, true);
+	if (!cqp_request) {
+		i40iw_pr_err("cqp_request memory failed\n");
+		return I40IW_ERR_NO_MEMORY;
+	}
+
+	/* increment refcount, because we need the cqp request ret value */
+	atomic_inc(&cqp_request->refcount);
+
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_ALLOC_LOCAL_MAC_IPADDR_ENTRY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.alloc_local_mac_ipaddr_entry.cqp = &iwcqp->sc_cqp;
+	cqp_info->in.u.alloc_local_mac_ipaddr_entry.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (!status)
+		*mac_ip_tbl_idx = cqp_request->compl_info.op_ret_val;
+	else
+		i40iw_pr_err("CQP-OP Alloc MAC Ip entry fail");
+	/* decrement refcount and free the cqp request, if no longer used */
+	i40iw_put_cqp_request(iwcqp, cqp_request);
+	return status;
+}
+
+/**
+ * i40iw_alloc_set_mac_ipaddr - set up a mac ip address table entry
+ * @iwdev: iwarp device
+ * @macaddr: pointer to mac address
+ *
+ * Allocate a mac ip address entry and add it to the hw table
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_alloc_set_mac_ipaddr(struct i40iw_device *iwdev,
+							 u8 *macaddr)
+{
+	enum i40iw_status_code status;
+
+	status = i40iw_alloc_local_mac_ipaddr_entry(iwdev, &iwdev->mac_ip_table_idx);
+	if (!status) {
+		status = i40iw_add_mac_ipaddr_entry(iwdev, macaddr,
+						    (u8)iwdev->mac_ip_table_idx);
+		if (!status)
+			status = i40iw_add_mac_ipaddr_entry(iwdev, macaddr,
+							    (u8)iwdev->mac_ip_table_idx);
+		else
+			i40iw_del_macip_entry(iwdev, (u8)iwdev->mac_ip_table_idx);
+	}
+	return status;
+}
+
+/**
+ * i40iw_add_ipv6_addr - add ipv6 address to the hw arp table
+ * @iwdev: iwarp device
+ */
+static void i40iw_add_ipv6_addr(struct i40iw_device *iwdev)
+{
+	struct net_device *ip_dev;
+	struct inet6_dev *idev;
+	struct inet6_ifaddr *ifp;
+	__be32 local_ipaddr6[4];
+
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, ip_dev) {
+		if ((((rdma_vlan_dev_vlan_id(ip_dev) < 0xFFFF) &&
+		      (rdma_vlan_dev_real_dev(ip_dev) == iwdev->netdev)) ||
+		     (ip_dev == iwdev->netdev)) && (ip_dev->flags & IFF_UP)) {
+			idev = __in6_dev_get(ip_dev);
+			if (!idev) {
+				i40iw_pr_err("ipv6 inet device not found\n");
+				break;
+			}
+			list_for_each_entry(ifp, &idev->addr_list, if_list) {
+				i40iw_pr_info("IP=%pI6, vlan_id=%d, MAC=%pM\n", &ifp->addr,
+					      rdma_vlan_dev_vlan_id(ip_dev), ip_dev->dev_addr);
+				i40iw_copy_ip_ntohl(local_ipaddr6,
+						    ifp->addr.in6_u.u6_addr32);
+				i40iw_manage_arp_cache(iwdev,
+						       ip_dev->dev_addr,
+						       local_ipaddr6,
+						       false,
+						       I40IW_ARP_ADD);
+			}
+		}
+	}
+	rcu_read_unlock();
+}
+
+/**
+ * i40iw_add_ipv4_addr - add ipv4 address to the hw arp table
+ * @iwdev: iwarp device
+ */
+static void i40iw_add_ipv4_addr(struct i40iw_device *iwdev)
+{
+	struct net_device *dev;
+	struct in_device *idev;
+	bool got_lock = true;
+	u32 ip_addr;
+
+	if (!rtnl_trylock())
+		got_lock = false;
+
+	for_each_netdev(&init_net, dev) {
+		if ((((rdma_vlan_dev_vlan_id(dev) < 0xFFFF) &&
+		      (rdma_vlan_dev_real_dev(dev) == iwdev->netdev)) ||
+		    (dev == iwdev->netdev)) && (dev->flags & IFF_UP)) {
+			idev = in_dev_get(dev);
+			for_ifa(idev) {
+				i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_CM,
+					    "IP=%pI4, vlan_id=%d, MAC=%pM\n", &ifa->ifa_address,
+					     rdma_vlan_dev_vlan_id(dev), dev->dev_addr);
+
+				ip_addr = ntohl(ifa->ifa_address);
+				i40iw_manage_arp_cache(iwdev,
+						       dev->dev_addr,
+						       &ip_addr,
+						       true,
+						       I40IW_ARP_ADD);
+			}
+			endfor_ifa(idev);
+			in_dev_put(idev);
+		}
+	}
+	if (got_lock)
+		rtnl_unlock();
+}
+
+/**
+ * i40iw_add_mac_ip - add mac and ip addresses
+ * @iwdev: iwarp device
+ *
+ * Create and add a mac ip address entry to the hw table and
+ * ipv4/ipv6 addresses to the arp cache
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_add_mac_ip(struct i40iw_device *iwdev)
+{
+	struct net_device *netdev = iwdev->netdev;
+	enum i40iw_status_code status;
+
+	status = i40iw_alloc_set_mac_ipaddr(iwdev, (u8 *)netdev->dev_addr);
+	if (status)
+		return status;
+	i40iw_add_ipv4_addr(iwdev);
+	i40iw_add_ipv6_addr(iwdev);
+	return 0;
+}
+
+/**
+ * i40iw_wait_pe_ready - Check if firmware is ready
+ * @hw: provides access to registers
+ */
+static void i40iw_wait_pe_ready(struct i40iw_hw *hw)
+{
+	u32 statusfw;
+	u32 statuscpu0;
+	u32 statuscpu1;
+	u32 statuscpu2;
+	u32 retrycount = 0;
+
+	do {
+		statusfw = i40iw_rd32(hw, I40E_GLPE_FWLDSTATUS);
+		i40iw_pr_info("[%04d] fm load status[x%04X]\n", __LINE__, statusfw);
+		statuscpu0 = i40iw_rd32(hw, I40E_GLPE_CPUSTATUS0);
+		i40iw_pr_info("[%04d] CSR_CQP status[x%04X]\n", __LINE__, statuscpu0);
+		statuscpu1 = i40iw_rd32(hw, I40E_GLPE_CPUSTATUS1);
+		i40iw_pr_info("[%04d] I40E_GLPE_CPUSTATUS1 status[x%04X]\n",
+			      __LINE__, statuscpu1);
+		statuscpu2 = i40iw_rd32(hw, I40E_GLPE_CPUSTATUS2);
+		i40iw_pr_info("[%04d] I40E_GLPE_CPUSTATUS2 status[x%04X]\n",
+			      __LINE__, statuscpu2);
+		if ((statuscpu0 == 0x80) && (statuscpu1 == 0x80) && (statuscpu2 == 0x80))
+			break;	/* SUCCESS */
+		mdelay(1000);
+		retrycount++;
+	} while (retrycount < 14);
+	i40iw_wr32(hw, 0xb4040, 0x4C104C5);
+}
+
+/**
+ * i40iw_initialize_dev - initialize device
+ * @iwdev: iwarp device
+ * @ldev: lan device information
+ *
+ * Allocate memory for the hmc objects and initialize iwdev
+ * Return 0 if successful, otherwise clean up the resources
+ * and return error
+ */
+static enum i40iw_status_code i40iw_initialize_dev(struct i40iw_device *iwdev,
+						   struct i40e_info *ldev)
+{
+	enum i40iw_status_code status;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_device_init_info info;
+	struct i40iw_dma_mem mem;
+	u32 size;
+
+	memset(&info, 0, sizeof(info));
+	size = sizeof(struct i40iw_hmc_pble_rsrc) + sizeof(struct i40iw_hmc_info) +
+				(sizeof(struct i40iw_hmc_obj_info) * I40IW_HMC_IW_MAX);
+	iwdev->hmc_info_mem = kzalloc(size, GFP_KERNEL);
+	if (!iwdev->hmc_info_mem) {
+		i40iw_pr_err("memory alloc fail\n");
+		return I40IW_ERR_NO_MEMORY;
+	}
+	iwdev->pble_rsrc = (struct i40iw_hmc_pble_rsrc *)iwdev->hmc_info_mem;
+	dev->hmc_info = &iwdev->hw.hmc;
+	dev->hmc_info->hmc_obj = (struct i40iw_hmc_obj_info *)(iwdev->pble_rsrc + 1);
+	status = i40iw_obj_aligned_mem(iwdev, &mem, I40IW_QUERY_FPM_BUF_SIZE,
+				       I40IW_FPM_QUERY_BUF_ALIGNMENT_MASK);
+	if (status)
+		goto exit;
+	info.fpm_query_buf_pa = mem.pa;
+	info.fpm_query_buf = mem.va;
+	status = i40iw_obj_aligned_mem(iwdev, &mem, I40IW_COMMIT_FPM_BUF_SIZE,
+				       I40IW_FPM_COMMIT_BUF_ALIGNMENT_MASK);
+	if (status)
+		goto exit;
+	info.fpm_commit_buf_pa = mem.pa;
+	info.fpm_commit_buf = mem.va;
+	info.hmc_fn_id = ldev->fid;
+	info.is_pf = (ldev->ftype) ? false : true;
+	info.bar0 = ldev->hw_addr;
+	info.hw = &iwdev->hw;
+	info.debug_mask = debug;
+	info.qs_handle = ldev->params.qos.prio_qos[0].qs_handle;
+	info.exception_lan_queue = 1;
+	info.vchnl_send = i40iw_virtchnl_send;
+	status = i40iw_device_init(&iwdev->sc_dev, &info);
+exit:
+	if (status) {
+		kfree(iwdev->hmc_info_mem);
+		iwdev->hmc_info_mem = NULL;
+	}
+	return status;
+}
+
+/**
+ * i40iw_register_notifiers - register tcp ip notifiers
+ */
+static void i40iw_register_notifiers(void)
+{
+	if (!i40iw_notifiers_registered) {
+		register_inetaddr_notifier(&i40iw_inetaddr_notifier);
+		register_inet6addr_notifier(&i40iw_inetaddr6_notifier);
+		register_netevent_notifier(&i40iw_net_notifier);
+	}
+	i40iw_notifiers_registered++;
+}
+
+/**
+ * i40iw_save_msix_info - copy msix vector information to iwarp device
+ * @iwdev: iwarp device
+ * @ldev: lan device information
+ *
+ * Allocate iwdev msix table and copy the ldev msix info to the table
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_save_msix_info(struct i40iw_device *iwdev,
+						   struct i40e_info *ldev)
+{
+	struct i40e_qvlist_info *iw_qvlist;
+	struct i40e_qv_info *iw_qvinfo;
+	u32 ceq_idx;
+	u32 i;
+	u32 size;
+
+	iwdev->msix_count = ldev->msix_count;
+
+	size = sizeof(struct i40iw_msix_vector) * iwdev->msix_count;
+	size += sizeof(struct i40e_qvlist_info);
+	size +=  sizeof(struct i40e_qv_info) * iwdev->msix_count - 1;
+	iwdev->iw_msixtbl = kzalloc(size, GFP_KERNEL);
+
+	if (!iwdev->iw_msixtbl)
+		return I40IW_ERR_NO_MEMORY;
+	iwdev->iw_qvlist = (struct i40e_qvlist_info *)(&iwdev->iw_msixtbl[iwdev->msix_count]);
+	iw_qvlist = iwdev->iw_qvlist;
+	iw_qvinfo = iw_qvlist->qv_info;
+	iw_qvlist->num_vectors = iwdev->msix_count;
+	if (iwdev->msix_count <= num_online_cpus())
+		iwdev->msix_shared = true;
+	for (i = 0, ceq_idx = 0; i < iwdev->msix_count; i++, iw_qvinfo++) {
+		iwdev->iw_msixtbl[i].idx = ldev->msix_entries[i].entry;
+		iwdev->iw_msixtbl[i].irq = ldev->msix_entries[i].vector;
+		if (i == 0) {
+			iw_qvinfo->aeq_idx = 0;
+			if (iwdev->msix_shared)
+				iw_qvinfo->ceq_idx = ceq_idx++;
+			else
+				iw_qvinfo->ceq_idx = I40E_QUEUE_INVALID_IDX;
+		} else {
+			iw_qvinfo->aeq_idx = I40E_QUEUE_INVALID_IDX;
+			iw_qvinfo->ceq_idx = ceq_idx++;
+		}
+		iw_qvinfo->itr_idx = 3;
+		iw_qvinfo->v_idx = iwdev->iw_msixtbl[i].idx;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_deinit_device - clean up the device resources
+ * @iwdev: iwarp device
+ * @reset: true if called before reset
+ * @del_hdl: true if delete hdl entry
+ *
+ * Destroy the ib device interface, remove the mac ip entry and ipv4/ipv6 addresses,
+ * destroy the device queues and free the pble and the hmc objects
+ */
+static void i40iw_deinit_device(struct i40iw_device *iwdev, bool reset, bool del_hdl)
+{
+	struct i40e_info *ldev = iwdev->ldev;
+
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+
+	i40iw_pr_info("state = %d\n", iwdev->init_state);
+
+	switch (iwdev->init_state) {
+	case RDMA_DEV_REGISTERED:
+		iwdev->iw_status = 0;
+		i40iw_port_ibevent(iwdev);
+		i40iw_destroy_rdma_device(iwdev->iwibdev);
+		/* fallthrough */
+	case IP_ADDR_REGISTERED:
+		if (!reset)
+			i40iw_del_macip_entry(iwdev, (u8)iwdev->mac_ip_table_idx);
+		/* fallthrough */
+	case INET_NOTIFIER:
+		if (i40iw_notifiers_registered > 0) {
+			i40iw_notifiers_registered--;
+			unregister_netevent_notifier(&i40iw_net_notifier);
+			unregister_inetaddr_notifier(&i40iw_inetaddr_notifier);
+			unregister_inet6addr_notifier(&i40iw_inetaddr6_notifier);
+		}
+		/* fallthrough */
+	case CEQ_CREATED:
+		i40iw_dele_ceqs(iwdev, reset);
+		/* fallthrough */
+	case AEQ_CREATED:
+		i40iw_destroy_aeq(iwdev, reset);
+		/* fallthrough */
+	case IEQ_CREATED:
+		i40iw_puda_dele_resources(dev, I40IW_PUDA_RSRC_TYPE_IEQ, reset);
+		/* fallthrough */
+	case ILQ_CREATED:
+		i40iw_puda_dele_resources(dev, I40IW_PUDA_RSRC_TYPE_ILQ, reset);
+		/* fallthrough */
+	case CCQ_CREATED:
+		i40iw_destroy_ccq(iwdev, reset);
+		/* fallthrough */
+	case PBLE_CHUNK_MEM:
+		i40iw_destroy_pble_pool(dev, iwdev->pble_rsrc);
+		/* fallthrough */
+	case HMC_OBJS_CREATED:
+		i40iw_del_hmc_objects(dev, dev->hmc_info, true, reset);
+		/* fallthrough */
+	case CQP_CREATED:
+		i40iw_destroy_cqp(iwdev, !reset);
+		/* fallthrough */
+	case INITIAL_STATE:
+		i40iw_cleanup_cm_core(&iwdev->cm_core);
+		if (dev->is_pf)
+			i40iw_hw_stats_del_timer(dev);
+
+		i40iw_del_init_mem(iwdev);
+		break;
+	case INVALID_STATE:
+		/* fallthrough */
+	default:
+		i40iw_pr_err("bad init_state = %d\n", iwdev->init_state);
+		break;
+	}
+
+	if (del_hdl)
+		i40iw_del_handler(i40iw_find_i40e_handler(ldev));
+	kfree(iwdev->hdl);
+}
+
+/**
+ * i40iw_setup_init_state - set up the initial device struct
+ * @hdl: handler for iwarp device - one per instance
+ * @ldev: lan device information
+ * @client: iwarp client information, provided during registration
+ *
+ * Initialize the iwarp device and its hdl information
+ * using the ldev and client information
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
+						     struct i40e_info *ldev,
+						     struct i40e_client *client)
+{
+	struct i40iw_device *iwdev = &hdl->device;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+
+	memcpy(&hdl->ldev, ldev, sizeof(*ldev));
+	if (resource_profile == 1)
+		resource_profile = 2;
+
+	iwdev->mpa_version = mpa_version;
+	iwdev->resource_profile = (resource_profile < I40IW_HMC_PROFILE_EQUAL) ?
+	    (u8)resource_profile + I40IW_HMC_PROFILE_DEFAULT :
+	    I40IW_HMC_PROFILE_DEFAULT;
+	iwdev->max_rdma_vfs =
+		(iwdev->resource_profile != I40IW_HMC_PROFILE_DEFAULT) ?  max_rdma_vfs : 0;
+	iwdev->netdev = ldev->netdev;
+	hdl->client = client;
+	iwdev->mss = (!ldev->params.mtu) ? I40IW_DEFAULT_MSS : ldev->params.mtu - I40IW_MTU_TO_MSS;
+	if (!ldev->ftype)
+		iwdev->db_start = pci_resource_start(ldev->pcidev, 0) + I40IW_DB_ADDR_OFFSET;
+	else
+		iwdev->db_start = pci_resource_start(ldev->pcidev, 0) + I40IW_VF_DB_ADDR_OFFSET;
+
+	status = i40iw_save_msix_info(iwdev, ldev);
+	if (status)
+		goto exit;
+	iwdev->hw.dev_context = (void *)ldev->pcidev;
+	iwdev->hw.hw_addr = ldev->hw_addr;
+	status = i40iw_allocate_dma_mem(&iwdev->hw,
+					&iwdev->obj_mem, 8192, 4096);
+	if (status)
+		goto exit;
+	iwdev->obj_next = iwdev->obj_mem;
+	iwdev->push_mode = push_mode;
+	init_waitqueue_head(&iwdev->vchnl_waitq);
+	status = i40iw_initialize_dev(iwdev, ldev);
+exit:
+	if (status) {
+		kfree(iwdev->iw_msixtbl);
+		i40iw_free_dma_mem(dev->hw, &iwdev->obj_mem);
+		iwdev->iw_msixtbl = NULL;
+	}
+	return status;
+}
+
+/**
+ * i40iw_open - client interface operation open for iwarp/uda device
+ * @ldev: lan device information
+ * @client: iwarp client information, provided during registration
+ *
+ * Called by the lan driver during the processing of client register
+ * Create device resources, set up queues, pble and hmc objects and
+ * register the device with the ib verbs interface
+ * Return 0 if successful, otherwise return error
+ */
+static int i40iw_open(struct i40e_info *ldev, struct i40e_client *client)
+{
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_dev *dev;
+	enum i40iw_status_code status;
+	struct i40iw_handler *hdl;
+
+	hdl = kzalloc(sizeof(*hdl), GFP_KERNEL);
+	if (!hdl)
+		return -ENOMEM;
+	iwdev = &hdl->device;
+	iwdev->hdl = hdl;
+	dev = &iwdev->sc_dev;
+	i40iw_setup_cm_core(iwdev);
+
+	dev->back_dev = (void *)iwdev;
+	iwdev->ldev = &hdl->ldev;
+	iwdev->client = client;
+	mutex_init(&iwdev->pbl_mutex);
+	i40iw_add_handler(hdl);
+
+	do {
+		status = i40iw_setup_init_state(hdl, ldev, client);
+		if (status)
+			break;
+		iwdev->init_state = INITIAL_STATE;
+		if (dev->is_pf)
+			i40iw_wait_pe_ready(dev->hw);
+		status = i40iw_create_cqp(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = CQP_CREATED;
+		status = i40iw_hmc_setup(iwdev);
+		if (status)
+			break;
+		status = i40iw_create_ccq(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = CCQ_CREATED;
+		status = i40iw_initialize_ilq(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = ILQ_CREATED;
+		status = i40iw_initialize_ieq(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = IEQ_CREATED;
+		status = i40iw_setup_aeq(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = AEQ_CREATED;
+		status = i40iw_setup_ceqs(iwdev, ldev);
+		if (status)
+			break;
+		iwdev->init_state = CEQ_CREATED;
+		status = i40iw_initialize_hw_resources(iwdev);
+		if (status)
+			break;
+		dev->ccq_ops->ccq_arm(dev->ccq);
+		status = i40iw_hmc_init_pble(&iwdev->sc_dev, iwdev->pble_rsrc);
+		if (status)
+			break;
+		iwdev->virtchnl_wq = create_singlethread_workqueue("iwvch");
+		i40iw_register_notifiers();
+		iwdev->init_state = INET_NOTIFIER;
+		status = i40iw_add_mac_ip(iwdev);
+		if (status)
+			break;
+		iwdev->init_state = IP_ADDR_REGISTERED;
+		if (i40iw_register_rdma_device(iwdev)) {
+			i40iw_pr_err("register rdma device fail\n");
+			break;
+		};
+
+		iwdev->init_state = RDMA_DEV_REGISTERED;
+		iwdev->iw_status = 1;
+		i40iw_port_ibevent(iwdev);
+		i40iw_pr_info("i40iw_open completed\n");
+		return 0;
+	} while (0);
+
+	i40iw_pr_err("status = %d last completion = %d\n", status, iwdev->init_state);
+	i40iw_deinit_device(iwdev, false, false);
+	return -ERESTART;
+}
+
+/**
+ * i40iw_l2param_change : handle qs handles for qos and mss change
+ * @ldev: lan device information
+ * @client: client for paramater change
+ * @params: new parameters from L2
+ */
+static void i40iw_l2param_change(struct i40e_info *ldev,
+				 struct i40e_client *client,
+				 struct i40e_params *params)
+{
+	struct i40iw_handler *hdl;
+	struct i40iw_device *iwdev;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return;
+
+	iwdev = &hdl->device;
+	if (params->mtu)
+		iwdev->mss = params->mtu - I40IW_MTU_TO_MSS;
+}
+
+/**
+ * i40iw_close - client interface operation close for iwarp/uda device
+ * @ldev: lan device information
+ * @client: client to close
+ *
+ * Called by the lan driver during the processing of client unregister
+ * Destroy and clean up the driver resources
+ */
+static void i40iw_close(struct i40e_info *ldev, struct i40e_client *client, bool reset)
+{
+	struct i40iw_device *iwdev;
+	struct i40iw_handler *hdl;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return;
+
+	iwdev = &hdl->device;
+	destroy_workqueue(iwdev->virtchnl_wq);
+	i40iw_deinit_device(iwdev, reset, true);
+}
+
+/**
+ * i40iw_vf_reset - process VF reset
+ * @ldev: lan device information
+ * @client: client interface instance
+ * @vf_id: virtual function id
+ *
+ * Called when a VF is reset by the PF
+ * Destroy and clean up the VF resources
+ */
+static void i40iw_vf_reset(struct i40e_info *ldev, struct i40e_client *client, u32 vf_id)
+{
+	struct i40iw_handler *hdl;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_hmc_fcn_info hmc_fcn_info;
+	struct i40iw_virt_mem vf_dev_mem;
+	struct i40iw_vfdev *tmp_vfdev;
+	unsigned int i;
+	unsigned long flags;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return;
+
+	dev = &hdl->device.sc_dev;
+
+	for (i = 0; i < I40IW_MAX_PE_ENABLED_VF_COUNT; i++) {
+		if (!dev->vf_dev[i] || (dev->vf_dev[i]->vf_id != vf_id))
+			continue;
+
+		/* free all resources allocated on behalf of vf */
+		tmp_vfdev = dev->vf_dev[i];
+		spin_lock_irqsave(&dev->dev_pestat.stats_lock, flags);
+		dev->vf_dev[i] = NULL;
+		spin_unlock_irqrestore(&dev->dev_pestat.stats_lock, flags);
+		i40iw_del_hmc_objects(dev, &tmp_vfdev->hmc_info, false, false);
+		/* remove vf hmc function */
+		memset(&hmc_fcn_info, 0, sizeof(hmc_fcn_info));
+		hmc_fcn_info.vf_id = vf_id;
+		hmc_fcn_info.iw_vf_idx = tmp_vfdev->iw_vf_idx;
+		hmc_fcn_info.free_fcn = true;
+		i40iw_cqp_manage_hmc_fcn_cmd(dev, &hmc_fcn_info);
+		/* free vf_dev */
+		vf_dev_mem.va = tmp_vfdev;
+		vf_dev_mem.size = sizeof(struct i40iw_vfdev) +
+					sizeof(struct i40iw_hmc_obj_info) * I40IW_HMC_IW_MAX;
+		i40iw_free_virt_mem(dev->hw, &vf_dev_mem);
+		break;
+	}
+}
+
+/**
+ * i40iw_vf_enable - enable a number of VFs
+ * @ldev: lan device information
+ * @client: client interface instance
+ * @num_vfs: number of VFs for the PF
+ *
+ * Called when the number of VFs changes
+ */
+static void i40iw_vf_enable(struct i40e_info *ldev,
+			    struct i40e_client *client,
+			    u32 num_vfs)
+{
+	struct i40iw_handler *hdl;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return;
+
+	if (num_vfs > I40IW_MAX_PE_ENABLED_VF_COUNT)
+		hdl->device.max_enabled_vfs = I40IW_MAX_PE_ENABLED_VF_COUNT;
+	else
+		hdl->device.max_enabled_vfs = num_vfs;
+}
+
+/**
+ * i40iw_vf_capable - check if VF capable
+ * @ldev: lan device information
+ * @client: client interface instance
+ * @vf_id: virtual function id
+ *
+ * Return 1 if a VF slot is available or if VF is already RDMA enabled
+ * Return 0 otherwise
+ */
+static int i40iw_vf_capable(struct i40e_info *ldev,
+			    struct i40e_client *client,
+			    u32 vf_id)
+{
+	struct i40iw_handler *hdl;
+	struct i40iw_sc_dev *dev;
+	unsigned int i;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return 0;
+
+	dev = &hdl->device.sc_dev;
+
+	for (i = 0; i < hdl->device.max_enabled_vfs; i++) {
+		if (!dev->vf_dev[i] || (dev->vf_dev[i]->vf_id == vf_id))
+			return 1;
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_virtchnl_receive - receive a message through the virtual channel
+ * @ldev: lan device information
+ * @client: client interface instance
+ * @vf_id: virtual function id associated with the message
+ * @msg: message buffer pointer
+ * @len: length of the message
+ *
+ * Invoke virtual channel receive operation for the given msg
+ * Return 0 if successful, otherwise return error
+ */
+static int i40iw_virtchnl_receive(struct i40e_info *ldev,
+				  struct i40e_client *client,
+				  u32 vf_id,
+				  u8 *msg,
+				  u16 len)
+{
+	struct i40iw_handler *hdl;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_device *iwdev;
+	int ret_code = I40IW_NOT_SUPPORTED;
+
+	if (!len || !msg)
+		return I40IW_ERR_PARAM;
+
+	hdl = i40iw_find_i40e_handler(ldev);
+	if (!hdl)
+		return I40IW_ERR_PARAM;
+
+	dev = &hdl->device.sc_dev;
+	iwdev = dev->back_dev;
+
+	i40iw_debug(dev, I40IW_DEBUG_VIRT, "msg %p, message length %u\n", msg, len);
+
+	if (dev->vchnl_if.vchnl_recv) {
+		ret_code = dev->vchnl_if.vchnl_recv(dev, vf_id, msg, len);
+		if (!dev->is_pf) {
+			atomic_dec(&iwdev->vchnl_msgs);
+			wake_up(&iwdev->vchnl_waitq);
+		}
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_virtchnl_send - send a message through the virtual channel
+ * @dev: iwarp device
+ * @vf_id: virtual function id associated with the message
+ * @msg: virtual channel message buffer pointer
+ * @len: length of the message
+ *
+ * Invoke virtual channel send operation for the given msg
+ * Return 0 if successful, otherwise return error
+ */
+static enum i40iw_status_code i40iw_virtchnl_send(struct i40iw_sc_dev *dev,
+						  u32 vf_id,
+						  u8 *msg,
+						  u16 len)
+{
+	struct i40iw_device *iwdev;
+	struct i40e_info *ldev;
+	enum i40iw_status_code ret_code = I40IW_ERR_BAD_PTR;
+
+	if (!dev || !dev->back_dev)
+		return ret_code;
+
+	iwdev = dev->back_dev;
+	ldev = iwdev->ldev;
+
+	if (ldev && ldev->ops && ldev->ops->virtchnl_send)
+		ret_code = ldev->ops->virtchnl_send(ldev, &i40iw_client, vf_id, msg, len);
+
+	return ret_code;
+}
+
+/* client interface functions */
+static struct i40e_client_ops i40e_ops = {
+	.open = i40iw_open,
+	.close = i40iw_close,
+	.l2_param_change = i40iw_l2param_change,
+	.virtchnl_receive = i40iw_virtchnl_receive,
+	.vf_reset = i40iw_vf_reset,
+	.vf_enable = i40iw_vf_enable,
+	.vf_capable = i40iw_vf_capable
+};
+
+/**
+ * i40iw_init_module - driver initialization function
+ *
+ * First function to call when the driver is loaded
+ * Register the driver as i40e client and port mapper client
+ */
+static int __init i40iw_init_module(void)
+{
+	int ret;
+
+	memset(&i40iw_client, 0, sizeof(i40iw_client));
+	i40iw_client.version.major = CLIENT_IW_INTERFACE_VERSION_MAJOR;
+	i40iw_client.version.minor = CLIENT_IW_INTERFACE_VERSION_MINOR;
+	i40iw_client.version.build = CLIENT_IW_INTERFACE_VERSION_BUILD;
+	i40iw_client.ops = &i40e_ops;
+	memcpy(i40iw_client.name, i40iw_client_name, I40E_CLIENT_STR_LENGTH);
+	i40iw_client.type = I40E_CLIENT_IWARP;
+	spin_lock_init(&i40iw_handler_lock);
+	ret = i40e_register_client(&i40iw_client);
+	return ret;
+}
+
+/**
+ * i40iw_exit_module - driver exit clean up function
+ *
+ * The function is called just before the driver is unloaded
+ * Unregister the driver as i40e client and port mapper client
+ */
+static void __exit i40iw_exit_module(void)
+{
+	i40e_unregister_client(&i40iw_client);
+}
+
+module_init(i40iw_init_module);
+module_exit(i40iw_exit_module);
diff --git a/drivers/infiniband/hw/i40iw/i40iw_osdep.h b/drivers/infiniband/hw/i40iw/i40iw_osdep.h
new file mode 100644
index 000000000000..7e20493510e8
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_osdep.h
@@ -0,0 +1,215 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_OSDEP_H
+#define I40IW_OSDEP_H
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/bitops.h>
+#include <net/tcp.h>
+#include <crypto/hash.h>
+/* get readq/writeq support for 32 bit kernels, use the low-first version */
+#include <linux/io-64-nonatomic-lo-hi.h>
+
+#define STATS_TIMER_DELAY 1000
+
+static inline void set_64bit_val(u64 *wqe_words, u32 byte_index, u64 value)
+{
+	wqe_words[byte_index >> 3] = value;
+}
+
+/**
+ * set_32bit_val - set 32 value to hw wqe
+ * @wqe_words: wqe addr to write
+ * @byte_index: index in wqe
+ * @value: value to write
+ **/
+static inline void set_32bit_val(u32 *wqe_words, u32 byte_index, u32 value)
+{
+	wqe_words[byte_index >> 2] = value;
+}
+
+/**
+ * get_64bit_val - read 64 bit value from wqe
+ * @wqe_words: wqe addr
+ * @byte_index: index to read from
+ * @value: read value
+ **/
+static inline void get_64bit_val(u64 *wqe_words, u32 byte_index, u64 *value)
+{
+	*value = wqe_words[byte_index >> 3];
+}
+
+/**
+ * get_32bit_val - read 32 bit value from wqe
+ * @wqe_words: wqe addr
+ * @byte_index: index to reaad from
+ * @value: return 32 bit value
+ **/
+static inline void get_32bit_val(u32 *wqe_words, u32 byte_index, u32 *value)
+{
+	*value = wqe_words[byte_index >> 2];
+}
+
+struct i40iw_dma_mem {
+	void *va;
+	dma_addr_t pa;
+	u32 size;
+} __packed;
+
+struct i40iw_virt_mem {
+	void *va;
+	u32 size;
+} __packed;
+
+#define i40iw_debug(h, m, s, ...)                               \
+do {                                                            \
+	if (((m) & (h)->debug_mask))                            \
+		pr_info("i40iw " s, ##__VA_ARGS__);             \
+} while (0)
+
+#define i40iw_flush(a)          readl((a)->hw_addr + I40E_GLGEN_STAT)
+
+#define I40E_GLHMC_VFSDCMD(_i)  (0x000C8000 + ((_i) * 4)) \
+				/* _i=0...31 */
+#define I40E_GLHMC_VFSDCMD_MAX_INDEX    31
+#define I40E_GLHMC_VFSDCMD_PMSDIDX_SHIFT  0
+#define I40E_GLHMC_VFSDCMD_PMSDIDX_MASK  (0xFFF \
+					  << I40E_GLHMC_VFSDCMD_PMSDIDX_SHIFT)
+#define I40E_GLHMC_VFSDCMD_PF_SHIFT       16
+#define I40E_GLHMC_VFSDCMD_PF_MASK        (0xF << I40E_GLHMC_VFSDCMD_PF_SHIFT)
+#define I40E_GLHMC_VFSDCMD_VF_SHIFT       20
+#define I40E_GLHMC_VFSDCMD_VF_MASK        (0x1FF << I40E_GLHMC_VFSDCMD_VF_SHIFT)
+#define I40E_GLHMC_VFSDCMD_PMF_TYPE_SHIFT 29
+#define I40E_GLHMC_VFSDCMD_PMF_TYPE_MASK  (0x3 \
+					   << I40E_GLHMC_VFSDCMD_PMF_TYPE_SHIFT)
+#define I40E_GLHMC_VFSDCMD_PMSDWR_SHIFT   31
+#define I40E_GLHMC_VFSDCMD_PMSDWR_MASK  (0x1 << I40E_GLHMC_VFSDCMD_PMSDWR_SHIFT)
+
+#define I40E_GLHMC_VFSDDATAHIGH(_i)     (0x000C8200 + ((_i) * 4)) \
+				/* _i=0...31 */
+#define I40E_GLHMC_VFSDDATAHIGH_MAX_INDEX       31
+#define I40E_GLHMC_VFSDDATAHIGH_PMSDDATAHIGH_SHIFT 0
+#define I40E_GLHMC_VFSDDATAHIGH_PMSDDATAHIGH_MASK  (0xFFFFFFFF \
+			<< I40E_GLHMC_VFSDDATAHIGH_PMSDDATAHIGH_SHIFT)
+
+#define I40E_GLHMC_VFSDDATALOW(_i)      (0x000C8100 + ((_i) * 4)) \
+				/* _i=0...31 */
+#define I40E_GLHMC_VFSDDATALOW_MAX_INDEX        31
+#define I40E_GLHMC_VFSDDATALOW_PMSDVALID_SHIFT   0
+#define I40E_GLHMC_VFSDDATALOW_PMSDVALID_MASK  (0x1 \
+			<< I40E_GLHMC_VFSDDATALOW_PMSDVALID_SHIFT)
+#define I40E_GLHMC_VFSDDATALOW_PMSDTYPE_SHIFT    1
+#define I40E_GLHMC_VFSDDATALOW_PMSDTYPE_MASK  (0x1 \
+			<< I40E_GLHMC_VFSDDATALOW_PMSDTYPE_SHIFT)
+#define I40E_GLHMC_VFSDDATALOW_PMSDBPCOUNT_SHIFT 2
+#define I40E_GLHMC_VFSDDATALOW_PMSDBPCOUNT_MASK  (0x3FF \
+			<< I40E_GLHMC_VFSDDATALOW_PMSDBPCOUNT_SHIFT)
+#define I40E_GLHMC_VFSDDATALOW_PMSDDATALOW_SHIFT 12
+#define I40E_GLHMC_VFSDDATALOW_PMSDDATALOW_MASK  (0xFFFFF \
+			<< I40E_GLHMC_VFSDDATALOW_PMSDDATALOW_SHIFT)
+
+#define I40E_GLPE_FWLDSTATUS                     0x0000D200
+#define I40E_GLPE_FWLDSTATUS_LOAD_REQUESTED_SHIFT 0
+#define I40E_GLPE_FWLDSTATUS_LOAD_REQUESTED_MASK  (0x1 \
+			<< I40E_GLPE_FWLDSTATUS_LOAD_REQUESTED_SHIFT)
+#define I40E_GLPE_FWLDSTATUS_DONE_SHIFT           1
+#define I40E_GLPE_FWLDSTATUS_DONE_MASK  (0x1 << I40E_GLPE_FWLDSTATUS_DONE_SHIFT)
+#define I40E_GLPE_FWLDSTATUS_CQP_FAIL_SHIFT       2
+#define I40E_GLPE_FWLDSTATUS_CQP_FAIL_MASK  (0x1 \
+			 << I40E_GLPE_FWLDSTATUS_CQP_FAIL_SHIFT)
+#define I40E_GLPE_FWLDSTATUS_TEP_FAIL_SHIFT       3
+#define I40E_GLPE_FWLDSTATUS_TEP_FAIL_MASK  (0x1 \
+			 << I40E_GLPE_FWLDSTATUS_TEP_FAIL_SHIFT)
+#define I40E_GLPE_FWLDSTATUS_OOP_FAIL_SHIFT       4
+#define I40E_GLPE_FWLDSTATUS_OOP_FAIL_MASK  (0x1 \
+			 << I40E_GLPE_FWLDSTATUS_OOP_FAIL_SHIFT)
+
+struct i40iw_sc_dev;
+struct i40iw_sc_qp;
+struct i40iw_puda_buf;
+struct i40iw_puda_completion_info;
+struct i40iw_update_sds_info;
+struct i40iw_hmc_fcn_info;
+struct i40iw_virtchnl_work_info;
+struct i40iw_manage_vf_pble_info;
+struct i40iw_device;
+struct i40iw_hmc_info;
+struct i40iw_hw;
+
+u8 __iomem *i40iw_get_hw_addr(void *dev);
+void i40iw_ieq_mpa_crc_ae(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp);
+enum i40iw_status_code i40iw_vf_wait_vchnl_resp(struct i40iw_sc_dev *dev);
+enum i40iw_status_code i40iw_ieq_check_mpacrc(struct shash_desc *desc, void *addr,
+					      u32 length, u32 value);
+struct i40iw_sc_qp *i40iw_ieq_get_qp(struct i40iw_sc_dev *dev, struct i40iw_puda_buf *buf);
+void i40iw_ieq_update_tcpip_info(struct i40iw_puda_buf *buf, u16 length, u32 seqnum);
+void i40iw_free_hash_desc(struct shash_desc *);
+enum i40iw_status_code i40iw_init_hash_desc(struct shash_desc **);
+enum i40iw_status_code i40iw_puda_get_tcpip_info(struct i40iw_puda_completion_info *info,
+						 struct i40iw_puda_buf *buf);
+enum i40iw_status_code i40iw_cqp_sds_cmd(struct i40iw_sc_dev *dev,
+					 struct i40iw_update_sds_info *info);
+enum i40iw_status_code i40iw_cqp_manage_hmc_fcn_cmd(struct i40iw_sc_dev *dev,
+						    struct i40iw_hmc_fcn_info *hmcfcninfo);
+enum i40iw_status_code i40iw_cqp_query_fpm_values_cmd(struct i40iw_sc_dev *dev,
+						      struct i40iw_dma_mem *values_mem,
+						      u8 hmc_fn_id);
+enum i40iw_status_code i40iw_cqp_commit_fpm_values_cmd(struct i40iw_sc_dev *dev,
+						       struct i40iw_dma_mem *values_mem,
+						       u8 hmc_fn_id);
+enum i40iw_status_code i40iw_alloc_query_fpm_buf(struct i40iw_sc_dev *dev,
+						 struct i40iw_dma_mem *mem);
+enum i40iw_status_code i40iw_cqp_manage_vf_pble_bp(struct i40iw_sc_dev *dev,
+						   struct i40iw_manage_vf_pble_info *info);
+void i40iw_cqp_spawn_worker(struct i40iw_sc_dev *dev,
+			    struct i40iw_virtchnl_work_info *work_info, u32 iw_vf_idx);
+void *i40iw_remove_head(struct list_head *list);
+
+void i40iw_term_modify_qp(struct i40iw_sc_qp *qp, u8 next_state, u8 term, u8 term_len);
+void i40iw_terminate_done(struct i40iw_sc_qp *qp, int timeout_occurred);
+void i40iw_terminate_start_timer(struct i40iw_sc_qp *qp);
+void i40iw_terminate_del_timer(struct i40iw_sc_qp *qp);
+
+enum i40iw_status_code i40iw_hw_manage_vf_pble_bp(struct i40iw_device *iwdev,
+						  struct i40iw_manage_vf_pble_info *info,
+						  bool wait);
+struct i40iw_dev_pestat;
+void i40iw_hw_stats_start_timer(struct i40iw_sc_dev *);
+void i40iw_hw_stats_del_timer(struct i40iw_sc_dev *);
+#define i40iw_mmiowb() mmiowb()
+void i40iw_wr32(struct i40iw_hw *hw, u32 reg, u32 value);
+u32  i40iw_rd32(struct i40iw_hw *hw, u32 reg);
+#endif				/* _I40IW_OSDEP_H_ */
diff --git a/drivers/infiniband/hw/i40iw/i40iw_p.h b/drivers/infiniband/hw/i40iw/i40iw_p.h
new file mode 100644
index 000000000000..a0b8ca10d67e
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_p.h
@@ -0,0 +1,106 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_P_H
+#define I40IW_P_H
+
+#define PAUSE_TIMER_VALUE  0xFFFF
+#define REFRESH_THRESHOLD  0x7FFF
+#define HIGH_THRESHOLD     0x800
+#define LOW_THRESHOLD      0x200
+#define ALL_TC2PFC         0xFF
+
+void i40iw_debug_buf(struct i40iw_sc_dev *dev, enum i40iw_debug_flag mask,
+		     char *desc, u64 *buf, u32 size);
+/* init operations */
+enum i40iw_status_code i40iw_device_init(struct i40iw_sc_dev *dev,
+					 struct i40iw_device_init_info *info);
+
+enum i40iw_status_code i40iw_device_init_pestat(struct i40iw_dev_pestat *);
+
+void i40iw_sc_cqp_post_sq(struct i40iw_sc_cqp *cqp);
+
+u64 *i40iw_sc_cqp_get_next_send_wqe(struct i40iw_sc_cqp *cqp, u64 scratch);
+
+enum i40iw_status_code i40iw_sc_mr_fast_register(struct i40iw_sc_qp *qp,
+						 struct i40iw_fast_reg_stag_info *info,
+						 bool post_sq);
+
+/* HMC/FPM functions */
+enum i40iw_status_code i40iw_sc_init_iw_hmc(struct i40iw_sc_dev *dev,
+					    u8 hmc_fn_id);
+
+enum i40iw_status_code i40iw_pf_init_vfhmc(struct i40iw_sc_dev *dev, u8 vf_hmc_fn_id,
+					   u32 *vf_cnt_array);
+
+/* cqp misc functions */
+
+void i40iw_terminate_send_fin(struct i40iw_sc_qp *qp);
+
+void i40iw_terminate_connection(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *info);
+
+void i40iw_terminate_received(struct i40iw_sc_qp *qp, struct i40iw_aeqe_info *info);
+
+enum i40iw_status_code i40iw_sc_suspend_qp(struct i40iw_sc_cqp *cqp,
+					   struct i40iw_sc_qp *qp, u64 scratch);
+
+enum i40iw_status_code i40iw_sc_resume_qp(struct i40iw_sc_cqp *cqp,
+					  struct i40iw_sc_qp *qp, u64 scratch);
+
+enum i40iw_status_code i40iw_sc_static_hmc_pages_allocated(struct i40iw_sc_cqp *cqp,
+							   u64 scratch, u8 hmc_fn_id,
+							   bool post_sq,
+							   bool poll_registers);
+
+enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_count);
+
+void free_sd_mem(struct i40iw_sc_dev *dev);
+
+enum i40iw_status_code i40iw_process_cqp_cmd(struct i40iw_sc_dev *dev,
+					     struct cqp_commands_info *pcmdinfo);
+
+enum i40iw_status_code i40iw_process_bh(struct i40iw_sc_dev *dev);
+
+/* prototype for functions used for dynamic memory allocation */
+enum i40iw_status_code i40iw_allocate_dma_mem(struct i40iw_hw *hw,
+					      struct i40iw_dma_mem *mem, u64 size,
+					      u32 alignment);
+void i40iw_free_dma_mem(struct i40iw_hw *hw, struct i40iw_dma_mem *mem);
+enum i40iw_status_code i40iw_allocate_virt_mem(struct i40iw_hw *hw,
+					       struct i40iw_virt_mem *mem, u32 size);
+enum i40iw_status_code i40iw_free_virt_mem(struct i40iw_hw *hw,
+					   struct i40iw_virt_mem *mem);
+u8 i40iw_get_encoded_wqe_size(u32 wqsize, bool cqpsq);
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_pble.c b/drivers/infiniband/hw/i40iw/i40iw_pble.c
new file mode 100644
index 000000000000..ded853d2fad8
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_pble.c
@@ -0,0 +1,618 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_status.h"
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_hmc.h"
+
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+
+#include <linux/pci.h>
+#include <linux/genalloc.h>
+#include <linux/vmalloc.h>
+#include "i40iw_pble.h"
+#include "i40iw.h"
+
+struct i40iw_device;
+static enum i40iw_status_code add_pble_pool(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_pble_rsrc *pble_rsrc);
+static void i40iw_free_vmalloc_mem(struct i40iw_hw *hw, struct i40iw_chunk *chunk);
+
+/**
+ * i40iw_destroy_pble_pool - destroy pool during module unload
+ * @pble_rsrc:	pble resources
+ */
+void i40iw_destroy_pble_pool(struct i40iw_sc_dev *dev, struct i40iw_hmc_pble_rsrc *pble_rsrc)
+{
+	struct list_head *clist;
+	struct list_head *tlist;
+	struct i40iw_chunk *chunk;
+	struct i40iw_pble_pool *pinfo = &pble_rsrc->pinfo;
+
+	if (pinfo->pool) {
+		list_for_each_safe(clist, tlist, &pinfo->clist) {
+			chunk = list_entry(clist, struct i40iw_chunk, list);
+			if (chunk->type == I40IW_VMALLOC)
+				i40iw_free_vmalloc_mem(dev->hw, chunk);
+			kfree(chunk);
+		}
+		gen_pool_destroy(pinfo->pool);
+	}
+}
+
+/**
+ * i40iw_hmc_init_pble - Initialize pble resources during module load
+ * @dev: i40iw_sc_dev struct
+ * @pble_rsrc:	pble resources
+ */
+enum i40iw_status_code i40iw_hmc_init_pble(struct i40iw_sc_dev *dev,
+					   struct i40iw_hmc_pble_rsrc *pble_rsrc)
+{
+	struct i40iw_hmc_info *hmc_info;
+	u32 fpm_idx = 0;
+
+	hmc_info = dev->hmc_info;
+	pble_rsrc->fpm_base_addr = hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].base;
+	/* Now start the pble' on 4k boundary */
+	if (pble_rsrc->fpm_base_addr & 0xfff)
+		fpm_idx = (PAGE_SIZE - (pble_rsrc->fpm_base_addr & 0xfff)) >> 3;
+
+	pble_rsrc->unallocated_pble =
+	    hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt - fpm_idx;
+	pble_rsrc->next_fpm_addr = pble_rsrc->fpm_base_addr + (fpm_idx << 3);
+
+	pble_rsrc->pinfo.pool_shift = POOL_SHIFT;
+	pble_rsrc->pinfo.pool = gen_pool_create(pble_rsrc->pinfo.pool_shift, -1);
+	INIT_LIST_HEAD(&pble_rsrc->pinfo.clist);
+	if (!pble_rsrc->pinfo.pool)
+		goto error;
+
+	if (add_pble_pool(dev, pble_rsrc))
+		goto error;
+
+	return 0;
+
+ error:i40iw_destroy_pble_pool(dev, pble_rsrc);
+	return I40IW_ERR_NO_MEMORY;
+}
+
+/**
+ * get_sd_pd_idx -  Returns sd index, pd index and rel_pd_idx from fpm address
+ * @ pble_rsrc:	structure containing fpm address
+ * @ idx: where to return indexes
+ */
+static inline void get_sd_pd_idx(struct i40iw_hmc_pble_rsrc *pble_rsrc,
+				 struct sd_pd_idx *idx)
+{
+	idx->sd_idx = (u32)(pble_rsrc->next_fpm_addr) / I40IW_HMC_DIRECT_BP_SIZE;
+	idx->pd_idx = (u32)(pble_rsrc->next_fpm_addr) / I40IW_HMC_PAGED_BP_SIZE;
+	idx->rel_pd_idx = (idx->pd_idx % I40IW_HMC_PD_CNT_IN_SD);
+}
+
+/**
+ * add_sd_direct - add sd direct for pble
+ * @dev: hardware control device structure
+ * @pble_rsrc: pble resource ptr
+ * @info: page info for sd
+ */
+static enum i40iw_status_code add_sd_direct(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_pble_rsrc *pble_rsrc,
+					    struct i40iw_add_page_info *info)
+{
+	enum i40iw_status_code ret_code = 0;
+	struct sd_pd_idx *idx = &info->idx;
+	struct i40iw_chunk *chunk = info->chunk;
+	struct i40iw_hmc_info *hmc_info = info->hmc_info;
+	struct i40iw_hmc_sd_entry *sd_entry = info->sd_entry;
+	u32 offset = 0;
+
+	if (!sd_entry->valid) {
+		if (dev->is_pf) {
+			ret_code = i40iw_add_sd_table_entry(dev->hw, hmc_info,
+							    info->idx.sd_idx,
+							    I40IW_SD_TYPE_DIRECT,
+							    I40IW_HMC_DIRECT_BP_SIZE);
+			if (ret_code)
+				return ret_code;
+			chunk->type = I40IW_DMA_COHERENT;
+		}
+	}
+	offset = idx->rel_pd_idx << I40IW_HMC_PAGED_BP_SHIFT;
+	chunk->size = info->pages << I40IW_HMC_PAGED_BP_SHIFT;
+	chunk->vaddr = ((u8 *)sd_entry->u.bp.addr.va + offset);
+	chunk->fpm_addr = pble_rsrc->next_fpm_addr;
+	i40iw_debug(dev, I40IW_DEBUG_PBLE, "chunk_size[%d] = 0x%x vaddr=%p fpm_addr = %llx\n",
+		    chunk->size, chunk->size, chunk->vaddr, chunk->fpm_addr);
+	return 0;
+}
+
+/**
+ * i40iw_free_vmalloc_mem - free vmalloc during close
+ * @hw: hw struct
+ * @chunk: chunk information for vmalloc
+ */
+static void i40iw_free_vmalloc_mem(struct i40iw_hw *hw, struct i40iw_chunk *chunk)
+{
+	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	int i;
+
+	if (!chunk->pg_cnt)
+		goto done;
+	for (i = 0; i < chunk->pg_cnt; i++)
+		dma_unmap_page(&pcidev->dev, chunk->dmaaddrs[i], PAGE_SIZE, DMA_BIDIRECTIONAL);
+
+ done:
+	kfree(chunk->dmaaddrs);
+	chunk->dmaaddrs = NULL;
+	vfree(chunk->vaddr);
+	chunk->vaddr = NULL;
+	chunk->type = 0;
+}
+
+/**
+ * i40iw_get_vmalloc_mem - get 2M page for sd
+ * @hw: hardware address
+ * @chunk: chunk to adf
+ * @pg_cnt: #of 4 K pages
+ */
+static enum i40iw_status_code i40iw_get_vmalloc_mem(struct i40iw_hw *hw,
+						    struct i40iw_chunk *chunk,
+						    int pg_cnt)
+{
+	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	struct page *page;
+	u8 *addr;
+	u32 size;
+	int i;
+
+	chunk->dmaaddrs = kzalloc(pg_cnt << 3, GFP_KERNEL);
+	if (!chunk->dmaaddrs)
+		return I40IW_ERR_NO_MEMORY;
+	size = PAGE_SIZE * pg_cnt;
+	chunk->vaddr = vmalloc(size);
+	if (!chunk->vaddr) {
+		kfree(chunk->dmaaddrs);
+		chunk->dmaaddrs = NULL;
+		return I40IW_ERR_NO_MEMORY;
+	}
+	chunk->size = size;
+	addr = (u8 *)chunk->vaddr;
+	for (i = 0; i < pg_cnt; i++) {
+		page = vmalloc_to_page((void *)addr);
+		if (!page)
+			break;
+		chunk->dmaaddrs[i] = dma_map_page(&pcidev->dev, page, 0,
+						  PAGE_SIZE, DMA_BIDIRECTIONAL);
+		if (dma_mapping_error(&pcidev->dev, chunk->dmaaddrs[i]))
+			break;
+		addr += PAGE_SIZE;
+	}
+
+	chunk->pg_cnt = i;
+	chunk->type = I40IW_VMALLOC;
+	if (i == pg_cnt)
+		return 0;
+
+	i40iw_free_vmalloc_mem(hw, chunk);
+	return I40IW_ERR_NO_MEMORY;
+}
+
+/**
+ * fpm_to_idx - given fpm address, get pble index
+ * @pble_rsrc: pble resource management
+ * @addr: fpm address for index
+ */
+static inline u32 fpm_to_idx(struct i40iw_hmc_pble_rsrc *pble_rsrc, u64 addr)
+{
+	return (addr - (pble_rsrc->fpm_base_addr)) >> 3;
+}
+
+/**
+ * add_bp_pages - add backing pages for sd
+ * @dev: hardware control device structure
+ * @pble_rsrc: pble resource management
+ * @info: page info for sd
+ */
+static enum i40iw_status_code add_bp_pages(struct i40iw_sc_dev *dev,
+					   struct i40iw_hmc_pble_rsrc *pble_rsrc,
+					   struct i40iw_add_page_info *info)
+{
+	u8 *addr;
+	struct i40iw_dma_mem mem;
+	struct i40iw_hmc_pd_entry *pd_entry;
+	struct i40iw_hmc_sd_entry *sd_entry = info->sd_entry;
+	struct i40iw_hmc_info *hmc_info = info->hmc_info;
+	struct i40iw_chunk *chunk = info->chunk;
+	struct i40iw_manage_vf_pble_info vf_pble_info;
+	enum i40iw_status_code status = 0;
+	u32 rel_pd_idx = info->idx.rel_pd_idx;
+	u32 pd_idx = info->idx.pd_idx;
+	u32 i;
+
+	status = i40iw_get_vmalloc_mem(dev->hw, chunk, info->pages);
+	if (status)
+		return I40IW_ERR_NO_MEMORY;
+	status = i40iw_add_sd_table_entry(dev->hw, hmc_info,
+					  info->idx.sd_idx, I40IW_SD_TYPE_PAGED,
+					  I40IW_HMC_DIRECT_BP_SIZE);
+	if (status) {
+		i40iw_free_vmalloc_mem(dev->hw, chunk);
+		return status;
+	}
+	if (!dev->is_pf) {
+		status = i40iw_vchnl_vf_add_hmc_objs(dev, I40IW_HMC_IW_PBLE,
+						     fpm_to_idx(pble_rsrc,
+								pble_rsrc->next_fpm_addr),
+						     (info->pages << PBLE_512_SHIFT));
+		if (status) {
+			i40iw_pr_err("allocate PBLEs in the PF.  Error %i\n", status);
+			i40iw_free_vmalloc_mem(dev->hw, chunk);
+			return status;
+		}
+	}
+	addr = chunk->vaddr;
+	for (i = 0; i < info->pages; i++) {
+		mem.pa = chunk->dmaaddrs[i];
+		mem.size = PAGE_SIZE;
+		mem.va = (void *)(addr);
+		pd_entry = &sd_entry->u.pd_table.pd_entry[rel_pd_idx++];
+		if (!pd_entry->valid) {
+			status = i40iw_add_pd_table_entry(dev->hw, hmc_info, pd_idx++, &mem);
+			if (status)
+				goto error;
+			addr += PAGE_SIZE;
+		} else {
+			i40iw_pr_err("pd entry is valid expecting to be invalid\n");
+		}
+	}
+	if (!dev->is_pf) {
+		vf_pble_info.first_pd_index = info->idx.rel_pd_idx;
+		vf_pble_info.inv_pd_ent = false;
+		vf_pble_info.pd_entry_cnt = PBLE_PER_PAGE;
+		vf_pble_info.pd_pl_pba = sd_entry->u.pd_table.pd_page_addr.pa;
+		vf_pble_info.sd_index = info->idx.sd_idx;
+		status = i40iw_hw_manage_vf_pble_bp(dev->back_dev,
+						    &vf_pble_info, true);
+		if (status) {
+			i40iw_pr_err("CQP manage VF PBLE BP failed.  %i\n", status);
+			goto error;
+		}
+	}
+	chunk->fpm_addr = pble_rsrc->next_fpm_addr;
+	return 0;
+error:
+	i40iw_free_vmalloc_mem(dev->hw, chunk);
+	return status;
+}
+
+/**
+ * add_pble_pool - add a sd entry for pble resoure
+ * @dev: hardware control device structure
+ * @pble_rsrc: pble resource management
+ */
+static enum i40iw_status_code add_pble_pool(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_pble_rsrc *pble_rsrc)
+{
+	struct i40iw_hmc_sd_entry *sd_entry;
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_chunk *chunk;
+	struct i40iw_add_page_info info;
+	struct sd_pd_idx *idx = &info.idx;
+	enum i40iw_status_code ret_code = 0;
+	enum i40iw_sd_entry_type sd_entry_type;
+	u64 sd_reg_val = 0;
+	u32 pages;
+
+	if (pble_rsrc->unallocated_pble < PBLE_PER_PAGE)
+		return I40IW_ERR_NO_MEMORY;
+	if (pble_rsrc->next_fpm_addr & 0xfff) {
+		i40iw_pr_err("next fpm_addr %llx\n", pble_rsrc->next_fpm_addr);
+		return I40IW_ERR_INVALID_PAGE_DESC_INDEX;
+	}
+	chunk = kzalloc(sizeof(*chunk), GFP_KERNEL);
+	if (!chunk)
+		return I40IW_ERR_NO_MEMORY;
+	hmc_info = dev->hmc_info;
+	chunk->fpm_addr = pble_rsrc->next_fpm_addr;
+	get_sd_pd_idx(pble_rsrc, idx);
+	sd_entry = &hmc_info->sd_table.sd_entry[idx->sd_idx];
+	pages = (idx->rel_pd_idx) ? (I40IW_HMC_PD_CNT_IN_SD -
+			idx->rel_pd_idx) : I40IW_HMC_PD_CNT_IN_SD;
+	pages = min(pages, pble_rsrc->unallocated_pble >> PBLE_512_SHIFT);
+	if (!pages) {
+		ret_code = I40IW_ERR_NO_PBLCHUNKS_AVAILABLE;
+		goto error;
+	}
+	info.chunk = chunk;
+	info.hmc_info = hmc_info;
+	info.pages = pages;
+	info.sd_entry = sd_entry;
+	if (!sd_entry->valid) {
+		sd_entry_type = (!idx->rel_pd_idx &&
+				 (pages == I40IW_HMC_PD_CNT_IN_SD) &&
+				 dev->is_pf) ? I40IW_SD_TYPE_DIRECT : I40IW_SD_TYPE_PAGED;
+	} else {
+		sd_entry_type = sd_entry->entry_type;
+	}
+	i40iw_debug(dev, I40IW_DEBUG_PBLE,
+		    "pages = %d, unallocated_pble[%u] current_fpm_addr = %llx\n",
+		    pages, pble_rsrc->unallocated_pble, pble_rsrc->next_fpm_addr);
+	i40iw_debug(dev, I40IW_DEBUG_PBLE, "sd_entry_type = %d sd_entry valid = %d\n",
+		    sd_entry_type, sd_entry->valid);
+
+	if (sd_entry_type == I40IW_SD_TYPE_DIRECT)
+		ret_code = add_sd_direct(dev, pble_rsrc, &info);
+	if (ret_code)
+		sd_entry_type = I40IW_SD_TYPE_PAGED;
+	else
+		pble_rsrc->stats_direct_sds++;
+
+	if (sd_entry_type == I40IW_SD_TYPE_PAGED) {
+		ret_code = add_bp_pages(dev, pble_rsrc, &info);
+		if (ret_code)
+			goto error;
+		else
+			pble_rsrc->stats_paged_sds++;
+	}
+
+	if (gen_pool_add_virt(pble_rsrc->pinfo.pool, (unsigned long)chunk->vaddr,
+			      (phys_addr_t)chunk->fpm_addr, chunk->size, -1)) {
+		i40iw_pr_err("could not allocate memory by gen_pool_addr_virt()\n");
+		ret_code = I40IW_ERR_NO_MEMORY;
+		goto error;
+	}
+	pble_rsrc->next_fpm_addr += chunk->size;
+	i40iw_debug(dev, I40IW_DEBUG_PBLE, "next_fpm_addr = %llx chunk_size[%u] = 0x%x\n",
+		    pble_rsrc->next_fpm_addr, chunk->size, chunk->size);
+	pble_rsrc->unallocated_pble -= (chunk->size >> 3);
+	list_add(&chunk->list, &pble_rsrc->pinfo.clist);
+	sd_reg_val = (sd_entry_type == I40IW_SD_TYPE_PAGED) ?
+			sd_entry->u.pd_table.pd_page_addr.pa : sd_entry->u.bp.addr.pa;
+	if (sd_entry->valid)
+		return 0;
+	if (dev->is_pf)
+		ret_code = i40iw_hmc_sd_one(dev, hmc_info->hmc_fn_id,
+					    sd_reg_val, idx->sd_idx,
+					    sd_entry->entry_type, true);
+	if (ret_code) {
+		i40iw_pr_err("cqp cmd failed for sd (pbles)\n");
+		goto error;
+	}
+
+	sd_entry->valid = true;
+	return 0;
+ error:
+	kfree(chunk);
+	return ret_code;
+}
+
+/**
+ * free_lvl2 - fee level 2 pble
+ * @pble_rsrc: pble resource management
+ * @palloc: level 2 pble allocation
+ */
+static void free_lvl2(struct i40iw_hmc_pble_rsrc *pble_rsrc,
+		      struct i40iw_pble_alloc *palloc)
+{
+	u32 i;
+	struct gen_pool *pool;
+	struct i40iw_pble_level2 *lvl2 = &palloc->level2;
+	struct i40iw_pble_info *root = &lvl2->root;
+	struct i40iw_pble_info *leaf = lvl2->leaf;
+
+	pool = pble_rsrc->pinfo.pool;
+
+	for (i = 0; i < lvl2->leaf_cnt; i++, leaf++) {
+		if (leaf->addr)
+			gen_pool_free(pool, leaf->addr, (leaf->cnt << 3));
+		else
+			break;
+	}
+
+	if (root->addr)
+		gen_pool_free(pool, root->addr, (root->cnt << 3));
+
+	kfree(lvl2->leaf);
+	lvl2->leaf = NULL;
+}
+
+/**
+ * get_lvl2_pble - get level 2 pble resource
+ * @pble_rsrc: pble resource management
+ * @palloc: level 2 pble allocation
+ * @pool: pool pointer
+ */
+static enum i40iw_status_code get_lvl2_pble(struct i40iw_hmc_pble_rsrc *pble_rsrc,
+					    struct i40iw_pble_alloc *palloc,
+					    struct gen_pool *pool)
+{
+	u32 lf4k, lflast, total, i;
+	u32 pblcnt = PBLE_PER_PAGE;
+	u64 *addr;
+	struct i40iw_pble_level2 *lvl2 = &palloc->level2;
+	struct i40iw_pble_info *root = &lvl2->root;
+	struct i40iw_pble_info *leaf;
+
+	/* number of full 512 (4K) leafs) */
+	lf4k = palloc->total_cnt >> 9;
+	lflast = palloc->total_cnt % PBLE_PER_PAGE;
+	total = (lflast == 0) ? lf4k : lf4k + 1;
+	lvl2->leaf_cnt = total;
+
+	leaf = kzalloc((sizeof(*leaf) * total), GFP_ATOMIC);
+	if (!leaf)
+		return I40IW_ERR_NO_MEMORY;
+	lvl2->leaf = leaf;
+	/* allocate pbles for the root */
+	root->addr = gen_pool_alloc(pool, (total << 3));
+	if (!root->addr) {
+		kfree(lvl2->leaf);
+		lvl2->leaf = NULL;
+		return I40IW_ERR_NO_MEMORY;
+	}
+	root->idx = fpm_to_idx(pble_rsrc,
+			       (u64)gen_pool_virt_to_phys(pool, root->addr));
+	root->cnt = total;
+	addr = (u64 *)root->addr;
+	for (i = 0; i < total; i++, leaf++) {
+		pblcnt = (lflast && ((i + 1) == total)) ? lflast : PBLE_PER_PAGE;
+		leaf->addr = gen_pool_alloc(pool, (pblcnt << 3));
+		if (!leaf->addr)
+			goto error;
+		leaf->idx = fpm_to_idx(pble_rsrc, (u64)gen_pool_virt_to_phys(pool, leaf->addr));
+
+		leaf->cnt = pblcnt;
+		*addr = (u64)leaf->idx;
+		addr++;
+	}
+	palloc->level = I40IW_LEVEL_2;
+	pble_rsrc->stats_lvl2++;
+	return 0;
+ error:
+	free_lvl2(pble_rsrc, palloc);
+	return I40IW_ERR_NO_MEMORY;
+}
+
+/**
+ * get_lvl1_pble - get level 1 pble resource
+ * @dev: hardware control device structure
+ * @pble_rsrc: pble resource management
+ * @palloc: level 1 pble allocation
+ */
+static enum i40iw_status_code get_lvl1_pble(struct i40iw_sc_dev *dev,
+					    struct i40iw_hmc_pble_rsrc *pble_rsrc,
+					    struct i40iw_pble_alloc *palloc)
+{
+	u64 *addr;
+	struct gen_pool *pool;
+	struct i40iw_pble_info *lvl1 = &palloc->level1;
+
+	pool = pble_rsrc->pinfo.pool;
+	addr = (u64 *)gen_pool_alloc(pool, (palloc->total_cnt << 3));
+
+	if (!addr)
+		return I40IW_ERR_NO_MEMORY;
+
+	palloc->level = I40IW_LEVEL_1;
+	lvl1->addr = (unsigned long)addr;
+	lvl1->idx = fpm_to_idx(pble_rsrc, (u64)gen_pool_virt_to_phys(pool,
+			       (unsigned long)addr));
+	lvl1->cnt = palloc->total_cnt;
+	pble_rsrc->stats_lvl1++;
+	return 0;
+}
+
+/**
+ * get_lvl1_lvl2_pble - calls get_lvl1 and get_lvl2 pble routine
+ * @dev: i40iw_sc_dev struct
+ * @pble_rsrc:	pble resources
+ * @palloc: contains all inforamtion regarding pble (idx + pble addr)
+ * @pool: pointer to general purpose special memory pool descriptor
+ */
+static inline enum i40iw_status_code get_lvl1_lvl2_pble(struct i40iw_sc_dev *dev,
+							struct i40iw_hmc_pble_rsrc *pble_rsrc,
+							struct i40iw_pble_alloc *palloc,
+							struct gen_pool *pool)
+{
+	enum i40iw_status_code status = 0;
+
+	status = get_lvl1_pble(dev, pble_rsrc, palloc);
+	if (status && (palloc->total_cnt > PBLE_PER_PAGE))
+		status = get_lvl2_pble(pble_rsrc, palloc, pool);
+	return status;
+}
+
+/**
+ * i40iw_get_pble - allocate pbles from the pool
+ * @dev: i40iw_sc_dev struct
+ * @pble_rsrc:	pble resources
+ * @palloc: contains all inforamtion regarding pble (idx + pble addr)
+ * @pble_cnt: #of pbles requested
+ */
+enum i40iw_status_code i40iw_get_pble(struct i40iw_sc_dev *dev,
+				      struct i40iw_hmc_pble_rsrc *pble_rsrc,
+				      struct i40iw_pble_alloc *palloc,
+				      u32 pble_cnt)
+{
+	struct gen_pool *pool;
+	enum i40iw_status_code status = 0;
+	u32 max_sds = 0;
+	int i;
+
+	pool = pble_rsrc->pinfo.pool;
+	palloc->total_cnt = pble_cnt;
+	palloc->level = I40IW_LEVEL_0;
+	/*check first to see if we can get pble's without acquiring additional sd's */
+	status = get_lvl1_lvl2_pble(dev, pble_rsrc, palloc, pool);
+	if (!status)
+		goto exit;
+	max_sds = (palloc->total_cnt >> 18) + 1;
+	for (i = 0; i < max_sds; i++) {
+		status = add_pble_pool(dev, pble_rsrc);
+		if (status)
+			break;
+		status = get_lvl1_lvl2_pble(dev, pble_rsrc, palloc, pool);
+		if (!status)
+			break;
+	}
+exit:
+	if (!status)
+		pble_rsrc->stats_alloc_ok++;
+	else
+		pble_rsrc->stats_alloc_fail++;
+
+	return status;
+}
+
+/**
+ * i40iw_free_pble - put pbles back into pool
+ * @pble_rsrc:	pble resources
+ * @palloc: contains all inforamtion regarding pble resource being freed
+ */
+void i40iw_free_pble(struct i40iw_hmc_pble_rsrc *pble_rsrc,
+		     struct i40iw_pble_alloc *palloc)
+{
+	struct gen_pool *pool;
+
+	pool = pble_rsrc->pinfo.pool;
+	if (palloc->level == I40IW_LEVEL_2)
+		free_lvl2(pble_rsrc, palloc);
+	else
+		gen_pool_free(pool, palloc->level1.addr,
+			      (palloc->level1.cnt << 3));
+	pble_rsrc->stats_alloc_freed++;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_pble.h b/drivers/infiniband/hw/i40iw/i40iw_pble.h
new file mode 100644
index 000000000000..7b1851d21cc0
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_pble.h
@@ -0,0 +1,131 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_PBLE_H
+#define I40IW_PBLE_H
+
+#define POOL_SHIFT      6
+#define PBLE_PER_PAGE   512
+#define I40IW_HMC_PAGED_BP_SHIFT 12
+#define PBLE_512_SHIFT  9
+
+enum i40iw_pble_level {
+	I40IW_LEVEL_0 = 0,
+	I40IW_LEVEL_1 = 1,
+	I40IW_LEVEL_2 = 2
+};
+
+enum i40iw_alloc_type {
+	I40IW_NO_ALLOC = 0,
+	I40IW_DMA_COHERENT = 1,
+	I40IW_VMALLOC = 2
+};
+
+struct i40iw_pble_info {
+	unsigned long addr;
+	u32 idx;
+	u32 cnt;
+};
+
+struct i40iw_pble_level2 {
+	struct i40iw_pble_info root;
+	struct i40iw_pble_info *leaf;
+	u32 leaf_cnt;
+};
+
+struct i40iw_pble_alloc {
+	u32 total_cnt;
+	enum i40iw_pble_level level;
+	union {
+		struct i40iw_pble_info level1;
+		struct i40iw_pble_level2 level2;
+	};
+};
+
+struct sd_pd_idx {
+	u32 sd_idx;
+	u32 pd_idx;
+	u32 rel_pd_idx;
+};
+
+struct i40iw_add_page_info {
+	struct i40iw_chunk *chunk;
+	struct i40iw_hmc_sd_entry *sd_entry;
+	struct i40iw_hmc_info *hmc_info;
+	struct sd_pd_idx idx;
+	u32 pages;
+};
+
+struct i40iw_chunk {
+	struct list_head list;
+	u32 size;
+	void *vaddr;
+	u64 fpm_addr;
+	u32 pg_cnt;
+	dma_addr_t *dmaaddrs;
+	enum i40iw_alloc_type type;
+};
+
+struct i40iw_pble_pool {
+	struct gen_pool *pool;
+	struct list_head clist;
+	u32 total_pble_alloc;
+	u32 free_pble_cnt;
+	u32 pool_shift;
+};
+
+struct i40iw_hmc_pble_rsrc {
+	u32 unallocated_pble;
+	u64 fpm_base_addr;
+	u64 next_fpm_addr;
+	struct i40iw_pble_pool pinfo;
+
+	u32 stats_direct_sds;
+	u32 stats_paged_sds;
+	u64 stats_alloc_ok;
+	u64 stats_alloc_fail;
+	u64 stats_alloc_freed;
+	u64 stats_lvl1;
+	u64 stats_lvl2;
+};
+
+void i40iw_destroy_pble_pool(struct i40iw_sc_dev *dev, struct i40iw_hmc_pble_rsrc *pble_rsrc);
+enum i40iw_status_code i40iw_hmc_init_pble(struct i40iw_sc_dev *dev,
+					   struct i40iw_hmc_pble_rsrc *pble_rsrc);
+void i40iw_free_pble(struct i40iw_hmc_pble_rsrc *pble_rsrc, struct i40iw_pble_alloc *palloc);
+enum i40iw_status_code i40iw_get_pble(struct i40iw_sc_dev *dev,
+				      struct i40iw_hmc_pble_rsrc *pble_rsrc,
+				      struct i40iw_pble_alloc *palloc,
+				      u32 pble_cnt);
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_puda.c b/drivers/infiniband/hw/i40iw/i40iw_puda.c
new file mode 100644
index 000000000000..8eb400d8a7a0
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_puda.c
@@ -0,0 +1,1436 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_status.h"
+#include "i40iw_hmc.h"
+
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_puda.h"
+
+static void i40iw_ieq_receive(struct i40iw_sc_dev *dev,
+			      struct i40iw_puda_buf *buf);
+static void i40iw_ieq_tx_compl(struct i40iw_sc_dev *dev, void *sqwrid);
+static void i40iw_ilq_putback_rcvbuf(struct i40iw_sc_qp *qp, u32 wqe_idx);
+static enum i40iw_status_code i40iw_puda_replenish_rq(struct i40iw_puda_rsrc
+						      *rsrc, bool initial);
+/**
+ * i40iw_puda_get_listbuf - get buffer from puda list
+ * @list: list to use for buffers (ILQ or IEQ)
+ */
+static struct i40iw_puda_buf *i40iw_puda_get_listbuf(struct list_head *list)
+{
+	struct i40iw_puda_buf *buf = NULL;
+
+	if (!list_empty(list)) {
+		buf = (struct i40iw_puda_buf *)list->next;
+		list_del((struct list_head *)&buf->list);
+	}
+	return buf;
+}
+
+/**
+ * i40iw_puda_get_bufpool - return buffer from resource
+ * @rsrc: resource to use for buffer
+ */
+struct i40iw_puda_buf *i40iw_puda_get_bufpool(struct i40iw_puda_rsrc *rsrc)
+{
+	struct i40iw_puda_buf *buf = NULL;
+	struct list_head *list = &rsrc->bufpool;
+	unsigned long	flags;
+
+	spin_lock_irqsave(&rsrc->bufpool_lock, flags);
+	buf = i40iw_puda_get_listbuf(list);
+	if (buf)
+		rsrc->avail_buf_count--;
+	else
+		rsrc->stats_buf_alloc_fail++;
+	spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
+	return buf;
+}
+
+/**
+ * i40iw_puda_ret_bufpool - return buffer to rsrc list
+ * @rsrc: resource to use for buffer
+ * @buf: buffe to return to resouce
+ */
+void i40iw_puda_ret_bufpool(struct i40iw_puda_rsrc *rsrc,
+			    struct i40iw_puda_buf *buf)
+{
+	unsigned long	flags;
+
+	spin_lock_irqsave(&rsrc->bufpool_lock, flags);
+	list_add(&buf->list, &rsrc->bufpool);
+	spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
+	rsrc->avail_buf_count++;
+}
+
+/**
+ * i40iw_puda_post_recvbuf - set wqe for rcv buffer
+ * @rsrc: resource ptr
+ * @wqe_idx: wqe index to use
+ * @buf: puda buffer for rcv q
+ * @initial: flag if during init time
+ */
+static void i40iw_puda_post_recvbuf(struct i40iw_puda_rsrc *rsrc, u32 wqe_idx,
+				    struct i40iw_puda_buf *buf, bool initial)
+{
+	u64 *wqe;
+	struct i40iw_sc_qp *qp = &rsrc->qp;
+	u64 offset24 = 0;
+
+	qp->qp_uk.rq_wrid_array[wqe_idx] = (uintptr_t)buf;
+	wqe = qp->qp_uk.rq_base[wqe_idx].elem;
+	i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+		    "%s: wqe_idx= %d buf = %p wqe = %p\n", __func__,
+		    wqe_idx, buf, wqe);
+	if (!initial)
+		get_64bit_val(wqe, 24, &offset24);
+
+	offset24 = (offset24) ? 0 : LS_64(1, I40IWQPSQ_VALID);
+	set_64bit_val(wqe, 24, offset24);
+
+	set_64bit_val(wqe, 0, buf->mem.pa);
+	set_64bit_val(wqe, 8,
+		      LS_64(buf->mem.size, I40IWQPSQ_FRAG_LEN));
+	set_64bit_val(wqe, 24, offset24);
+}
+
+/**
+ * i40iw_puda_replenish_rq - post rcv buffers
+ * @rsrc: resource to use for buffer
+ * @initial: flag if during init time
+ */
+static enum i40iw_status_code i40iw_puda_replenish_rq(struct i40iw_puda_rsrc *rsrc,
+						      bool initial)
+{
+	u32 i;
+	u32 invalid_cnt = rsrc->rxq_invalid_cnt;
+	struct i40iw_puda_buf *buf = NULL;
+
+	for (i = 0; i < invalid_cnt; i++) {
+		buf = i40iw_puda_get_bufpool(rsrc);
+		if (!buf)
+			return I40IW_ERR_list_empty;
+		i40iw_puda_post_recvbuf(rsrc, rsrc->rx_wqe_idx, buf,
+					initial);
+		rsrc->rx_wqe_idx =
+		    ((rsrc->rx_wqe_idx + 1) % rsrc->rq_size);
+		rsrc->rxq_invalid_cnt--;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_puda_alloc_buf - allocate mem for buffer
+ * @dev: iwarp device
+ * @length: length of buffer
+ */
+static struct i40iw_puda_buf *i40iw_puda_alloc_buf(struct i40iw_sc_dev *dev,
+						   u32 length)
+{
+	struct i40iw_puda_buf *buf = NULL;
+	struct i40iw_virt_mem buf_mem;
+	enum i40iw_status_code ret;
+
+	ret = i40iw_allocate_virt_mem(dev->hw, &buf_mem,
+				      sizeof(struct i40iw_puda_buf));
+	if (ret) {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA,
+			    "%s: error mem for buf\n", __func__);
+		return NULL;
+	}
+	buf = (struct i40iw_puda_buf *)buf_mem.va;
+	ret = i40iw_allocate_dma_mem(dev->hw, &buf->mem, length, 1);
+	if (ret) {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA,
+			    "%s: error dma mem for buf\n", __func__);
+		i40iw_free_virt_mem(dev->hw, &buf_mem);
+		return NULL;
+	}
+	buf->buf_mem.va = buf_mem.va;
+	buf->buf_mem.size = buf_mem.size;
+	return buf;
+}
+
+/**
+ * i40iw_puda_dele_buf - delete buffer back to system
+ * @dev: iwarp device
+ * @buf: buffer to free
+ */
+static void i40iw_puda_dele_buf(struct i40iw_sc_dev *dev,
+				struct i40iw_puda_buf *buf)
+{
+	i40iw_free_dma_mem(dev->hw, &buf->mem);
+	i40iw_free_virt_mem(dev->hw, &buf->buf_mem);
+}
+
+/**
+ * i40iw_puda_get_next_send_wqe - return next wqe for processing
+ * @qp: puda qp for wqe
+ * @wqe_idx: wqe index for caller
+ */
+static u64 *i40iw_puda_get_next_send_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx)
+{
+	u64 *wqe = NULL;
+	enum i40iw_status_code ret_code = 0;
+
+	*wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+	if (!*wqe_idx)
+		qp->swqe_polarity = !qp->swqe_polarity;
+	I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
+	if (ret_code)
+		return wqe;
+	wqe = qp->sq_base[*wqe_idx].elem;
+
+	return wqe;
+}
+
+/**
+ * i40iw_puda_poll_info - poll cq for completion
+ * @cq: cq for poll
+ * @info: info return for successful completion
+ */
+static enum i40iw_status_code i40iw_puda_poll_info(struct i40iw_sc_cq *cq,
+						   struct i40iw_puda_completion_info *info)
+{
+	u64 qword0, qword2, qword3;
+	u64 *cqe;
+	u64 comp_ctx;
+	bool valid_bit;
+	u32 major_err, minor_err;
+	bool error;
+
+	cqe = (u64 *)I40IW_GET_CURRENT_CQ_ELEMENT(&cq->cq_uk);
+	get_64bit_val(cqe, 24, &qword3);
+	valid_bit = (bool)RS_64(qword3, I40IW_CQ_VALID);
+
+	if (valid_bit != cq->cq_uk.polarity)
+		return I40IW_ERR_QUEUE_EMPTY;
+
+	i40iw_debug_buf(cq->dev, I40IW_DEBUG_PUDA, "PUDA CQE", cqe, 32);
+	error = (bool)RS_64(qword3, I40IW_CQ_ERROR);
+	if (error) {
+		i40iw_debug(cq->dev, I40IW_DEBUG_PUDA, "%s receive error\n", __func__);
+		major_err = (u32)(RS_64(qword3, I40IW_CQ_MAJERR));
+		minor_err = (u32)(RS_64(qword3, I40IW_CQ_MINERR));
+		info->compl_error = major_err << 16 | minor_err;
+		return I40IW_ERR_CQ_COMPL_ERROR;
+	}
+
+	get_64bit_val(cqe, 0, &qword0);
+	get_64bit_val(cqe, 16, &qword2);
+
+	info->q_type = (u8)RS_64(qword3, I40IW_CQ_SQ);
+	info->qp_id = (u32)RS_64(qword2, I40IWCQ_QPID);
+
+	get_64bit_val(cqe, 8, &comp_ctx);
+	info->qp = (struct i40iw_qp_uk *)(unsigned long)comp_ctx;
+	info->wqe_idx = (u32)RS_64(qword3, I40IW_CQ_WQEIDX);
+
+	if (info->q_type == I40IW_CQE_QTYPE_RQ) {
+		info->vlan_valid = (bool)RS_64(qword3, I40IW_VLAN_TAG_VALID);
+		info->l4proto = (u8)RS_64(qword2, I40IW_UDA_L4PROTO);
+		info->l3proto = (u8)RS_64(qword2, I40IW_UDA_L3PROTO);
+		info->payload_len = (u16)RS_64(qword0, I40IW_UDA_PAYLOADLEN);
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_puda_poll_completion - processes completion for cq
+ * @dev: iwarp device
+ * @cq: cq getting interrupt
+ * @compl_err: return any completion err
+ */
+enum i40iw_status_code i40iw_puda_poll_completion(struct i40iw_sc_dev *dev,
+						  struct i40iw_sc_cq *cq, u32 *compl_err)
+{
+	struct i40iw_qp_uk *qp;
+	struct i40iw_cq_uk *cq_uk = &cq->cq_uk;
+	struct i40iw_puda_completion_info info;
+	enum i40iw_status_code ret = 0;
+	struct i40iw_puda_buf *buf;
+	struct i40iw_puda_rsrc *rsrc;
+	void *sqwrid;
+	u8 cq_type = cq->cq_type;
+	unsigned long	flags;
+
+	if ((cq_type == I40IW_CQ_TYPE_ILQ) || (cq_type == I40IW_CQ_TYPE_IEQ)) {
+		rsrc = (cq_type == I40IW_CQ_TYPE_ILQ) ? dev->ilq : dev->ieq;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "%s qp_type error\n", __func__);
+		return I40IW_ERR_BAD_PTR;
+	}
+	memset(&info, 0, sizeof(info));
+	ret = i40iw_puda_poll_info(cq, &info);
+	*compl_err = info.compl_error;
+	if (ret == I40IW_ERR_QUEUE_EMPTY)
+		return ret;
+	if (ret)
+		goto done;
+
+	qp = info.qp;
+	if (!qp || !rsrc) {
+		ret = I40IW_ERR_BAD_PTR;
+		goto done;
+	}
+
+	if (qp->qp_id != rsrc->qp_id) {
+		ret = I40IW_ERR_BAD_PTR;
+		goto done;
+	}
+
+	if (info.q_type == I40IW_CQE_QTYPE_RQ) {
+		buf = (struct i40iw_puda_buf *)(uintptr_t)qp->rq_wrid_array[info.wqe_idx];
+		/* Get all the tcpip information in the buf header */
+		ret = i40iw_puda_get_tcpip_info(&info, buf);
+		if (ret) {
+			rsrc->stats_rcvd_pkt_err++;
+			if (cq_type == I40IW_CQ_TYPE_ILQ) {
+				i40iw_ilq_putback_rcvbuf(&rsrc->qp,
+							 info.wqe_idx);
+			} else {
+				i40iw_puda_ret_bufpool(rsrc, buf);
+				i40iw_puda_replenish_rq(rsrc, false);
+			}
+			goto done;
+		}
+
+		rsrc->stats_pkt_rcvd++;
+		rsrc->compl_rxwqe_idx = info.wqe_idx;
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "%s RQ completion\n", __func__);
+		rsrc->receive(rsrc->dev, buf);
+		if (cq_type == I40IW_CQ_TYPE_ILQ)
+			i40iw_ilq_putback_rcvbuf(&rsrc->qp, info.wqe_idx);
+		else
+			i40iw_puda_replenish_rq(rsrc, false);
+
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "%s SQ completion\n", __func__);
+		sqwrid = (void *)(uintptr_t)qp->sq_wrtrk_array[info.wqe_idx].wrid;
+		I40IW_RING_SET_TAIL(qp->sq_ring, info.wqe_idx);
+		rsrc->xmit_complete(rsrc->dev, sqwrid);
+		spin_lock_irqsave(&rsrc->bufpool_lock, flags);
+		rsrc->tx_wqe_avail_cnt++;
+		spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
+		if (!list_empty(&dev->ilq->txpend))
+			i40iw_puda_send_buf(dev->ilq, NULL);
+	}
+
+done:
+	I40IW_RING_MOVE_HEAD(cq_uk->cq_ring, ret);
+	if (I40IW_RING_GETCURRENT_HEAD(cq_uk->cq_ring) == 0)
+		cq_uk->polarity = !cq_uk->polarity;
+	/* update cq tail in cq shadow memory also */
+	I40IW_RING_MOVE_TAIL(cq_uk->cq_ring);
+	set_64bit_val(cq_uk->shadow_area, 0,
+		      I40IW_RING_GETCURRENT_HEAD(cq_uk->cq_ring));
+	return 0;
+}
+
+/**
+ * i40iw_puda_send - complete send wqe for transmit
+ * @qp: puda qp for send
+ * @info: buffer information for transmit
+ */
+enum i40iw_status_code i40iw_puda_send(struct i40iw_sc_qp *qp,
+				       struct i40iw_puda_send_info *info)
+{
+	u64 *wqe;
+	u32 iplen, l4len;
+	u64 header[2];
+	u32 wqe_idx;
+	u8 iipt;
+
+	/* number of 32 bits DWORDS in header */
+	l4len = info->tcplen >> 2;
+	if (info->ipv4) {
+		iipt = 3;
+		iplen = 5;
+	} else {
+		iipt = 1;
+		iplen = 10;
+	}
+
+	wqe = i40iw_puda_get_next_send_wqe(&qp->qp_uk, &wqe_idx);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+	qp->qp_uk.sq_wrtrk_array[wqe_idx].wrid = (uintptr_t)info->scratch;
+	/* Third line of WQE descriptor */
+	/* maclen is in words */
+	header[0] = LS_64((info->maclen >> 1), I40IW_UDA_QPSQ_MACLEN) |
+		    LS_64(iplen, I40IW_UDA_QPSQ_IPLEN) | LS_64(1, I40IW_UDA_QPSQ_L4T) |
+		    LS_64(iipt, I40IW_UDA_QPSQ_IIPT) |
+		    LS_64(l4len, I40IW_UDA_QPSQ_L4LEN);
+	/* Forth line of WQE descriptor */
+	header[1] = LS_64(I40IW_OP_TYPE_SEND, I40IW_UDA_QPSQ_OPCODE) |
+		    LS_64(1, I40IW_UDA_QPSQ_SIGCOMPL) |
+		    LS_64(info->doloopback, I40IW_UDA_QPSQ_DOLOOPBACK) |
+		    LS_64(qp->qp_uk.swqe_polarity, I40IW_UDA_QPSQ_VALID);
+
+	set_64bit_val(wqe, 0, info->paddr);
+	set_64bit_val(wqe, 8, LS_64(info->len, I40IWQPSQ_FRAG_LEN));
+	set_64bit_val(wqe, 16, header[0]);
+	set_64bit_val(wqe, 24, header[1]);
+
+	i40iw_debug_buf(qp->dev, I40IW_DEBUG_PUDA, "PUDA SEND WQE", wqe, 32);
+	i40iw_qp_post_wr(&qp->qp_uk);
+	return 0;
+}
+
+/**
+ * i40iw_puda_send_buf - transmit puda buffer
+ * @rsrc: resource to use for buffer
+ * @buf: puda buffer to transmit
+ */
+void i40iw_puda_send_buf(struct i40iw_puda_rsrc *rsrc, struct i40iw_puda_buf *buf)
+{
+	struct i40iw_puda_send_info info;
+	enum i40iw_status_code ret = 0;
+	unsigned long	flags;
+
+	spin_lock_irqsave(&rsrc->bufpool_lock, flags);
+	/* if no wqe available or not from a completion and we have
+	 * pending buffers, we must queue new buffer
+	 */
+	if (!rsrc->tx_wqe_avail_cnt || (buf && !list_empty(&rsrc->txpend))) {
+		list_add_tail(&buf->list, &rsrc->txpend);
+		spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
+		rsrc->stats_sent_pkt_q++;
+		if (rsrc->type == I40IW_PUDA_RSRC_TYPE_ILQ)
+			i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+				    "%s: adding to txpend\n", __func__);
+		return;
+	}
+	rsrc->tx_wqe_avail_cnt--;
+	/* if we are coming from a completion and have pending buffers
+	 * then Get one from pending list
+	 */
+	if (!buf) {
+		buf = i40iw_puda_get_listbuf(&rsrc->txpend);
+		if (!buf)
+			goto done;
+	}
+
+	info.scratch = (void *)buf;
+	info.paddr = buf->mem.pa;
+	info.len = buf->totallen;
+	info.tcplen = buf->tcphlen;
+	info.maclen = buf->maclen;
+	info.ipv4 = buf->ipv4;
+	info.doloopback = (rsrc->type == I40IW_PUDA_RSRC_TYPE_IEQ);
+
+	ret = i40iw_puda_send(&rsrc->qp, &info);
+	if (ret) {
+		rsrc->tx_wqe_avail_cnt++;
+		rsrc->stats_sent_pkt_q++;
+		list_add(&buf->list, &rsrc->txpend);
+		if (rsrc->type == I40IW_PUDA_RSRC_TYPE_ILQ)
+			i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+				    "%s: adding to puda_send\n", __func__);
+	} else {
+		rsrc->stats_pkt_sent++;
+	}
+done:
+	spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
+}
+
+/**
+ * i40iw_puda_qp_setctx - during init, set qp's context
+ * @rsrc: qp's resource
+ */
+static void i40iw_puda_qp_setctx(struct i40iw_puda_rsrc *rsrc)
+{
+	struct i40iw_sc_qp *qp = &rsrc->qp;
+	u64 *qp_ctx = qp->hw_host_ctx;
+
+	set_64bit_val(qp_ctx, 8, qp->sq_pa);
+	set_64bit_val(qp_ctx, 16, qp->rq_pa);
+
+	set_64bit_val(qp_ctx, 24,
+		      LS_64(qp->hw_rq_size, I40IWQPC_RQSIZE) |
+		      LS_64(qp->hw_sq_size, I40IWQPC_SQSIZE));
+
+	set_64bit_val(qp_ctx, 48, LS_64(1514, I40IWQPC_SNDMSS));
+	set_64bit_val(qp_ctx, 56, 0);
+	set_64bit_val(qp_ctx, 64, 1);
+
+	set_64bit_val(qp_ctx, 136,
+		      LS_64(rsrc->cq_id, I40IWQPC_TXCQNUM) |
+		      LS_64(rsrc->cq_id, I40IWQPC_RXCQNUM));
+
+	set_64bit_val(qp_ctx, 160, LS_64(1, I40IWQPC_PRIVEN));
+
+	set_64bit_val(qp_ctx, 168,
+		      LS_64((uintptr_t)qp, I40IWQPC_QPCOMPCTX));
+
+	set_64bit_val(qp_ctx, 176,
+		      LS_64(qp->sq_tph_val, I40IWQPC_SQTPHVAL) |
+		      LS_64(qp->rq_tph_val, I40IWQPC_RQTPHVAL) |
+		      LS_64(qp->qs_handle, I40IWQPC_QSHANDLE));
+
+	i40iw_debug_buf(rsrc->dev, I40IW_DEBUG_PUDA, "PUDA QP CONTEXT",
+			qp_ctx, I40IW_QP_CTX_SIZE);
+}
+
+/**
+ * i40iw_puda_qp_wqe - setup wqe for qp create
+ * @rsrc: resource for qp
+ */
+static enum i40iw_status_code i40iw_puda_qp_wqe(struct i40iw_puda_rsrc *rsrc)
+{
+	struct i40iw_sc_qp *qp = &rsrc->qp;
+	struct i40iw_sc_dev *dev = rsrc->dev;
+	struct i40iw_sc_cqp *cqp;
+	u64 *wqe;
+	u64 header;
+	struct i40iw_ccq_cqe_info compl_info;
+	enum i40iw_status_code status = 0;
+
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, 0);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	set_64bit_val(wqe, 16, qp->hw_host_ctx_pa);
+	set_64bit_val(wqe, 40, qp->shadow_area_pa);
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_CREATE_QP, I40IW_CQPSQ_OPCODE) |
+		 LS_64(I40IW_QP_TYPE_UDA, I40IW_CQPSQ_QP_QPTYPE) |
+		 LS_64(1, I40IW_CQPSQ_QP_CQNUMVALID) |
+		 LS_64(2, I40IW_CQPSQ_QP_NEXTIWSTATE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	set_64bit_val(wqe, 24, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_PUDA, "PUDA CQE", wqe, 32);
+	i40iw_sc_cqp_post_sq(cqp);
+	status = dev->cqp_ops->poll_for_cqp_op_done(dev->cqp,
+						    I40IW_CQP_OP_CREATE_QP,
+						    &compl_info);
+	return status;
+}
+
+/**
+ * i40iw_puda_qp_create - create qp for resource
+ * @rsrc: resource to use for buffer
+ */
+static enum i40iw_status_code i40iw_puda_qp_create(struct i40iw_puda_rsrc *rsrc)
+{
+	struct i40iw_sc_qp *qp = &rsrc->qp;
+	struct i40iw_qp_uk *ukqp = &qp->qp_uk;
+	enum i40iw_status_code ret = 0;
+	u32 sq_size, rq_size, t_size;
+	struct i40iw_dma_mem *mem;
+
+	sq_size = rsrc->sq_size * I40IW_QP_WQE_MIN_SIZE;
+	rq_size = rsrc->rq_size * I40IW_QP_WQE_MIN_SIZE;
+	t_size = (sq_size + rq_size + (I40IW_SHADOW_AREA_SIZE << 3) +
+		  I40IW_QP_CTX_SIZE);
+	/* Get page aligned memory */
+	ret =
+	    i40iw_allocate_dma_mem(rsrc->dev->hw, &rsrc->qpmem, t_size,
+				   I40IW_HW_PAGE_SIZE);
+	if (ret) {
+		i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA, "%s: error dma mem\n", __func__);
+		return ret;
+	}
+
+	mem = &rsrc->qpmem;
+	memset(mem->va, 0, t_size);
+	qp->hw_sq_size = i40iw_get_encoded_wqe_size(rsrc->sq_size, false);
+	qp->hw_rq_size = i40iw_get_encoded_wqe_size(rsrc->rq_size, false);
+	qp->pd = &rsrc->sc_pd;
+	qp->qp_type = I40IW_QP_TYPE_UDA;
+	qp->dev = rsrc->dev;
+	qp->back_qp = (void *)rsrc;
+	qp->sq_pa = mem->pa;
+	qp->rq_pa = qp->sq_pa + sq_size;
+	ukqp->sq_base = mem->va;
+	ukqp->rq_base = &ukqp->sq_base[rsrc->sq_size];
+	ukqp->shadow_area = ukqp->rq_base[rsrc->rq_size].elem;
+	qp->shadow_area_pa = qp->rq_pa + rq_size;
+	qp->hw_host_ctx = ukqp->shadow_area + I40IW_SHADOW_AREA_SIZE;
+	qp->hw_host_ctx_pa =
+		qp->shadow_area_pa + (I40IW_SHADOW_AREA_SIZE << 3);
+	ukqp->qp_id = rsrc->qp_id;
+	ukqp->sq_wrtrk_array = rsrc->sq_wrtrk_array;
+	ukqp->rq_wrid_array = rsrc->rq_wrid_array;
+
+	ukqp->qp_id = rsrc->qp_id;
+	ukqp->sq_size = rsrc->sq_size;
+	ukqp->rq_size = rsrc->rq_size;
+
+	I40IW_RING_INIT(ukqp->sq_ring, ukqp->sq_size);
+	I40IW_RING_INIT(ukqp->initial_ring, ukqp->sq_size);
+	I40IW_RING_INIT(ukqp->rq_ring, ukqp->rq_size);
+
+	if (qp->pd->dev->is_pf)
+		ukqp->wqe_alloc_reg = (u32 __iomem *)(i40iw_get_hw_addr(qp->pd->dev) +
+						    I40E_PFPE_WQEALLOC);
+	else
+		ukqp->wqe_alloc_reg = (u32 __iomem *)(i40iw_get_hw_addr(qp->pd->dev) +
+						    I40E_VFPE_WQEALLOC1);
+
+	qp->qs_handle = qp->dev->qs_handle;
+	i40iw_puda_qp_setctx(rsrc);
+	ret = i40iw_puda_qp_wqe(rsrc);
+	if (ret)
+		i40iw_free_dma_mem(rsrc->dev->hw, &rsrc->qpmem);
+	return ret;
+}
+
+/**
+ * i40iw_puda_cq_create - create cq for resource
+ * @rsrc: resource for which cq to create
+ */
+static enum i40iw_status_code i40iw_puda_cq_create(struct i40iw_puda_rsrc *rsrc)
+{
+	struct i40iw_sc_dev *dev = rsrc->dev;
+	struct i40iw_sc_cq *cq = &rsrc->cq;
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+	enum i40iw_status_code ret = 0;
+	u32 tsize, cqsize;
+	u32 shadow_read_threshold = 128;
+	struct i40iw_dma_mem *mem;
+	struct i40iw_ccq_cqe_info compl_info;
+	struct i40iw_cq_init_info info;
+	struct i40iw_cq_uk_init_info *init_info = &info.cq_uk_init_info;
+
+	cq->back_cq = (void *)rsrc;
+	cqsize = rsrc->cq_size * (sizeof(struct i40iw_cqe));
+	tsize = cqsize + sizeof(struct i40iw_cq_shadow_area);
+	ret = i40iw_allocate_dma_mem(dev->hw, &rsrc->cqmem, tsize,
+				     I40IW_CQ0_ALIGNMENT_MASK);
+	if (ret)
+		return ret;
+
+	mem = &rsrc->cqmem;
+	memset(&info, 0, sizeof(info));
+	info.dev = dev;
+	info.type = (rsrc->type == I40IW_PUDA_RSRC_TYPE_ILQ) ?
+			 I40IW_CQ_TYPE_ILQ : I40IW_CQ_TYPE_IEQ;
+	info.shadow_read_threshold = rsrc->cq_size >> 2;
+	info.ceq_id_valid = true;
+	info.cq_base_pa = mem->pa;
+	info.shadow_area_pa = mem->pa + cqsize;
+	init_info->cq_base = mem->va;
+	init_info->shadow_area = (u64 *)((u8 *)mem->va + cqsize);
+	init_info->cq_size = rsrc->cq_size;
+	init_info->cq_id = rsrc->cq_id;
+	ret = dev->iw_priv_cq_ops->cq_init(cq, &info);
+	if (ret)
+		goto error;
+	cqp = dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, 0);
+	if (!wqe) {
+		ret = I40IW_ERR_RING_FULL;
+		goto error;
+	}
+
+	set_64bit_val(wqe, 0, rsrc->cq_size);
+	set_64bit_val(wqe, 8, RS_64_1(cq, 1));
+	set_64bit_val(wqe, 16, LS_64(shadow_read_threshold, I40IW_CQPSQ_CQ_SHADOW_READ_THRESHOLD));
+	set_64bit_val(wqe, 32, cq->cq_pa);
+
+	set_64bit_val(wqe, 40, cq->shadow_area_pa);
+
+	header = rsrc->cq_id |
+	    LS_64(I40IW_CQP_OP_CREATE_CQ, I40IW_CQPSQ_OPCODE) |
+	    LS_64(1, I40IW_CQPSQ_CQ_CHKOVERFLOW) |
+	    LS_64(1, I40IW_CQPSQ_CQ_ENCEQEMASK) |
+	    LS_64(1, I40IW_CQPSQ_CQ_CEQIDVALID) |
+	    LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	set_64bit_val(wqe, 24, header);
+
+	i40iw_debug_buf(dev, I40IW_DEBUG_PUDA, "PUDA CQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	i40iw_sc_cqp_post_sq(dev->cqp);
+	ret = dev->cqp_ops->poll_for_cqp_op_done(dev->cqp,
+						 I40IW_CQP_OP_CREATE_CQ,
+						 &compl_info);
+
+error:
+	if (ret)
+		i40iw_free_dma_mem(dev->hw, &rsrc->cqmem);
+	return ret;
+}
+
+/**
+ * i40iw_puda_dele_resources - delete all resources during close
+ * @dev: iwarp device
+ * @type: type of resource to dele
+ * @reset: true if reset chip
+ */
+void i40iw_puda_dele_resources(struct i40iw_sc_dev *dev,
+			       enum puda_resource_type type,
+			       bool reset)
+{
+	struct i40iw_ccq_cqe_info compl_info;
+	struct i40iw_puda_rsrc *rsrc;
+	struct i40iw_puda_buf *buf = NULL;
+	struct i40iw_puda_buf *nextbuf = NULL;
+	struct i40iw_virt_mem *vmem;
+	enum i40iw_status_code ret;
+
+	switch (type) {
+	case I40IW_PUDA_RSRC_TYPE_ILQ:
+		rsrc = dev->ilq;
+		vmem = &dev->ilq_mem;
+		break;
+	case I40IW_PUDA_RSRC_TYPE_IEQ:
+		rsrc = dev->ieq;
+		vmem = &dev->ieq_mem;
+		break;
+	default:
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "%s: error resource type = 0x%x\n",
+			    __func__, type);
+		return;
+	}
+
+	switch (rsrc->completion) {
+	case PUDA_HASH_CRC_COMPLETE:
+		i40iw_free_hash_desc(rsrc->hash_desc);
+	case PUDA_QP_CREATED:
+		do {
+			if (reset)
+				break;
+			ret = dev->iw_priv_qp_ops->qp_destroy(&rsrc->qp,
+							      0, false, true, true);
+			if (ret)
+				i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+					    "%s error ieq qp destroy\n",
+					    __func__);
+
+			ret = dev->cqp_ops->poll_for_cqp_op_done(dev->cqp,
+								 I40IW_CQP_OP_DESTROY_QP,
+								 &compl_info);
+			if (ret)
+				i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+					    "%s error ieq qp destroy done\n",
+					    __func__);
+		} while (0);
+
+		i40iw_free_dma_mem(dev->hw, &rsrc->qpmem);
+		/* fallthrough */
+	case PUDA_CQ_CREATED:
+		do {
+			if (reset)
+				break;
+			ret = dev->iw_priv_cq_ops->cq_destroy(&rsrc->cq, 0, true);
+			if (ret)
+				i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+					    "%s error ieq cq destroy\n",
+					    __func__);
+
+			ret = dev->cqp_ops->poll_for_cqp_op_done(dev->cqp,
+								 I40IW_CQP_OP_DESTROY_CQ,
+								 &compl_info);
+			if (ret)
+				i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA,
+					    "%s error ieq qp destroy done\n",
+					    __func__);
+		} while (0);
+
+		i40iw_free_dma_mem(dev->hw, &rsrc->cqmem);
+		break;
+	default:
+		i40iw_debug(rsrc->dev, I40IW_DEBUG_PUDA, "%s error no resources\n", __func__);
+		break;
+	}
+	/* Free all allocated puda buffers for both tx and rx */
+	buf = rsrc->alloclist;
+	while (buf) {
+		nextbuf = buf->next;
+		i40iw_puda_dele_buf(dev, buf);
+		buf = nextbuf;
+		rsrc->alloc_buf_count--;
+	}
+	i40iw_free_virt_mem(dev->hw, vmem);
+}
+
+/**
+ * i40iw_puda_allocbufs - allocate buffers for resource
+ * @rsrc: resource for buffer allocation
+ * @count: number of buffers to create
+ */
+static enum i40iw_status_code i40iw_puda_allocbufs(struct i40iw_puda_rsrc *rsrc,
+						   u32 count)
+{
+	u32 i;
+	struct i40iw_puda_buf *buf;
+	struct i40iw_puda_buf *nextbuf;
+
+	for (i = 0; i < count; i++) {
+		buf = i40iw_puda_alloc_buf(rsrc->dev, rsrc->buf_size);
+		if (!buf) {
+			rsrc->stats_buf_alloc_fail++;
+			return I40IW_ERR_NO_MEMORY;
+		}
+		i40iw_puda_ret_bufpool(rsrc, buf);
+		rsrc->alloc_buf_count++;
+		if (!rsrc->alloclist) {
+			rsrc->alloclist = buf;
+		} else {
+			nextbuf = rsrc->alloclist;
+			rsrc->alloclist = buf;
+			buf->next = nextbuf;
+		}
+	}
+	rsrc->avail_buf_count = rsrc->alloc_buf_count;
+	return 0;
+}
+
+/**
+ * i40iw_puda_create_rsrc - create resouce (ilq or ieq)
+ * @dev: iwarp device
+ * @info: resource information
+ */
+enum i40iw_status_code i40iw_puda_create_rsrc(struct i40iw_sc_dev *dev,
+					      struct i40iw_puda_rsrc_info *info)
+{
+	enum i40iw_status_code ret = 0;
+	struct i40iw_puda_rsrc *rsrc;
+	u32 pudasize;
+	u32 sqwridsize, rqwridsize;
+	struct i40iw_virt_mem *vmem;
+
+	info->count = 1;
+	pudasize = sizeof(struct i40iw_puda_rsrc);
+	sqwridsize = info->sq_size * sizeof(struct i40iw_sq_uk_wr_trk_info);
+	rqwridsize = info->rq_size * 8;
+	switch (info->type) {
+	case I40IW_PUDA_RSRC_TYPE_ILQ:
+		vmem = &dev->ilq_mem;
+		break;
+	case I40IW_PUDA_RSRC_TYPE_IEQ:
+		vmem = &dev->ieq_mem;
+		break;
+	default:
+		return I40IW_NOT_SUPPORTED;
+	}
+	ret =
+	    i40iw_allocate_virt_mem(dev->hw, vmem,
+				    pudasize + sqwridsize + rqwridsize);
+	if (ret)
+		return ret;
+	rsrc = (struct i40iw_puda_rsrc *)vmem->va;
+	spin_lock_init(&rsrc->bufpool_lock);
+	if (info->type == I40IW_PUDA_RSRC_TYPE_ILQ) {
+		dev->ilq = (struct i40iw_puda_rsrc *)vmem->va;
+		dev->ilq_count = info->count;
+		rsrc->receive = info->receive;
+		rsrc->xmit_complete = info->xmit_complete;
+	} else {
+		vmem = &dev->ieq_mem;
+		dev->ieq_count = info->count;
+		dev->ieq = (struct i40iw_puda_rsrc *)vmem->va;
+		rsrc->receive = i40iw_ieq_receive;
+		rsrc->xmit_complete = i40iw_ieq_tx_compl;
+	}
+
+	rsrc->type = info->type;
+	rsrc->sq_wrtrk_array = (struct i40iw_sq_uk_wr_trk_info *)((u8 *)vmem->va + pudasize);
+	rsrc->rq_wrid_array = (u64 *)((u8 *)vmem->va + pudasize + sqwridsize);
+	rsrc->mss = info->mss;
+	/* Initialize all ieq lists */
+	INIT_LIST_HEAD(&rsrc->bufpool);
+	INIT_LIST_HEAD(&rsrc->txpend);
+
+	rsrc->tx_wqe_avail_cnt = info->sq_size - 1;
+	dev->iw_pd_ops->pd_init(dev, &rsrc->sc_pd, info->pd_id);
+	rsrc->qp_id = info->qp_id;
+	rsrc->cq_id = info->cq_id;
+	rsrc->sq_size = info->sq_size;
+	rsrc->rq_size = info->rq_size;
+	rsrc->cq_size = info->rq_size + info->sq_size;
+	rsrc->buf_size = info->buf_size;
+	rsrc->dev = dev;
+
+	ret = i40iw_puda_cq_create(rsrc);
+	if (!ret) {
+		rsrc->completion = PUDA_CQ_CREATED;
+		ret = i40iw_puda_qp_create(rsrc);
+	}
+	if (ret) {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "[%s] error qp_create\n", __func__);
+		goto error;
+	}
+	rsrc->completion = PUDA_QP_CREATED;
+
+	ret = i40iw_puda_allocbufs(rsrc, info->tx_buf_cnt + info->rq_size);
+	if (ret) {
+		i40iw_debug(dev, I40IW_DEBUG_PUDA, "[%s] error allloc_buf\n", __func__);
+		goto error;
+	}
+
+	rsrc->rxq_invalid_cnt = info->rq_size;
+	ret = i40iw_puda_replenish_rq(rsrc, true);
+	if (ret)
+		goto error;
+
+	if (info->type == I40IW_PUDA_RSRC_TYPE_IEQ) {
+		if (!i40iw_init_hash_desc(&rsrc->hash_desc)) {
+			rsrc->check_crc = true;
+			rsrc->completion = PUDA_HASH_CRC_COMPLETE;
+			ret = 0;
+		}
+	}
+
+	dev->ccq_ops->ccq_arm(&rsrc->cq);
+	return ret;
+ error:
+	i40iw_puda_dele_resources(dev, info->type, false);
+
+	return ret;
+}
+
+/**
+ * i40iw_ilq_putback_rcvbuf - ilq buffer to put back on rq
+ * @qp: ilq's qp resource
+ * @wqe_idx:  wqe index of completed rcvbuf
+ */
+static void i40iw_ilq_putback_rcvbuf(struct i40iw_sc_qp *qp, u32 wqe_idx)
+{
+	u64 *wqe;
+	u64 offset24;
+
+	wqe = qp->qp_uk.rq_base[wqe_idx].elem;
+	get_64bit_val(wqe, 24, &offset24);
+	offset24 = (offset24) ? 0 : LS_64(1, I40IWQPSQ_VALID);
+	set_64bit_val(wqe, 24, offset24);
+}
+
+/**
+ * i40iw_ieq_get_fpdu - given length return fpdu length
+ * @length: length if fpdu
+ */
+static u16 i40iw_ieq_get_fpdu_length(u16 length)
+{
+	u16 fpdu_len;
+
+	fpdu_len = length + I40IW_IEQ_MPA_FRAMING;
+	fpdu_len = (fpdu_len + 3) & 0xfffffffc;
+	return fpdu_len;
+}
+
+/**
+ * i40iw_ieq_copy_to_txbuf - copydata from rcv buf to tx buf
+ * @buf: rcv buffer with partial
+ * @txbuf: tx buffer for sendign back
+ * @buf_offset: rcv buffer offset to copy from
+ * @txbuf_offset: at offset in tx buf to copy
+ * @length: length of data to copy
+ */
+static void i40iw_ieq_copy_to_txbuf(struct i40iw_puda_buf *buf,
+				    struct i40iw_puda_buf *txbuf,
+				    u16 buf_offset, u32 txbuf_offset,
+				    u32 length)
+{
+	void *mem1 = (u8 *)buf->mem.va + buf_offset;
+	void *mem2 = (u8 *)txbuf->mem.va + txbuf_offset;
+
+	memcpy(mem2, mem1, length);
+}
+
+/**
+ * i40iw_ieq_setup_tx_buf - setup tx buffer for partial handling
+ * @buf: reeive buffer with partial
+ * @txbuf: buffer to prepare
+ */
+static void i40iw_ieq_setup_tx_buf(struct i40iw_puda_buf *buf,
+				   struct i40iw_puda_buf *txbuf)
+{
+	txbuf->maclen = buf->maclen;
+	txbuf->tcphlen = buf->tcphlen;
+	txbuf->ipv4 = buf->ipv4;
+	txbuf->hdrlen = buf->hdrlen;
+	i40iw_ieq_copy_to_txbuf(buf, txbuf, 0, 0, buf->hdrlen);
+}
+
+/**
+ * i40iw_ieq_check_first_buf - check if rcv buffer's seq is in range
+ * @buf: receive exception buffer
+ * @fps: first partial sequence number
+ */
+static void i40iw_ieq_check_first_buf(struct i40iw_puda_buf *buf, u32 fps)
+{
+	u32 offset;
+
+	if (buf->seqnum < fps) {
+		offset = fps - buf->seqnum;
+		if (offset > buf->datalen)
+			return;
+		buf->data += offset;
+		buf->datalen -= (u16)offset;
+		buf->seqnum = fps;
+	}
+}
+
+/**
+ * i40iw_ieq_compl_pfpdu - write txbuf with full fpdu
+ * @ieq: ieq resource
+ * @rxlist: ieq's received buffer list
+ * @pbufl: temporary list for buffers for fpddu
+ * @txbuf: tx buffer for fpdu
+ * @fpdu_len: total length of fpdu
+ */
+static void  i40iw_ieq_compl_pfpdu(struct i40iw_puda_rsrc *ieq,
+				   struct list_head *rxlist,
+				   struct list_head *pbufl,
+				   struct i40iw_puda_buf *txbuf,
+				   u16 fpdu_len)
+{
+	struct i40iw_puda_buf *buf;
+	u32 nextseqnum;
+	u16 txoffset, bufoffset;
+
+	buf = i40iw_puda_get_listbuf(pbufl);
+	nextseqnum = buf->seqnum + fpdu_len;
+	txbuf->totallen = buf->hdrlen + fpdu_len;
+	txbuf->data = (u8 *)txbuf->mem.va + buf->hdrlen;
+	i40iw_ieq_setup_tx_buf(buf, txbuf);
+
+	txoffset = buf->hdrlen;
+	bufoffset = (u16)(buf->data - (u8 *)buf->mem.va);
+
+	do {
+		if (buf->datalen >= fpdu_len) {
+			/* copied full fpdu */
+			i40iw_ieq_copy_to_txbuf(buf, txbuf, bufoffset, txoffset, fpdu_len);
+			buf->datalen -= fpdu_len;
+			buf->data += fpdu_len;
+			buf->seqnum = nextseqnum;
+			break;
+		}
+		/* copy partial fpdu */
+		i40iw_ieq_copy_to_txbuf(buf, txbuf, bufoffset, txoffset, buf->datalen);
+		txoffset += buf->datalen;
+		fpdu_len -= buf->datalen;
+		i40iw_puda_ret_bufpool(ieq, buf);
+		buf = i40iw_puda_get_listbuf(pbufl);
+		bufoffset = (u16)(buf->data - (u8 *)buf->mem.va);
+	} while (1);
+
+	/* last buffer on the list*/
+	if (buf->datalen)
+		list_add(&buf->list, rxlist);
+	else
+		i40iw_puda_ret_bufpool(ieq, buf);
+}
+
+/**
+ * i40iw_ieq_create_pbufl - create buffer list for single fpdu
+ * @rxlist: resource list for receive ieq buffes
+ * @pbufl: temp. list for buffers for fpddu
+ * @buf: first receive buffer
+ * @fpdu_len: total length of fpdu
+ */
+static enum i40iw_status_code i40iw_ieq_create_pbufl(
+						     struct i40iw_pfpdu *pfpdu,
+						     struct list_head *rxlist,
+						     struct list_head *pbufl,
+						     struct i40iw_puda_buf *buf,
+						     u16 fpdu_len)
+{
+	enum i40iw_status_code status = 0;
+	struct i40iw_puda_buf *nextbuf;
+	u32	nextseqnum;
+	u16 plen = fpdu_len - buf->datalen;
+	bool done = false;
+
+	nextseqnum = buf->seqnum + buf->datalen;
+	do {
+		nextbuf = i40iw_puda_get_listbuf(rxlist);
+		if (!nextbuf) {
+			status = I40IW_ERR_list_empty;
+			break;
+		}
+		list_add_tail(&nextbuf->list, pbufl);
+		if (nextbuf->seqnum != nextseqnum) {
+			pfpdu->bad_seq_num++;
+			status = I40IW_ERR_SEQ_NUM;
+			break;
+		}
+		if (nextbuf->datalen >= plen) {
+			done = true;
+		} else {
+			plen -= nextbuf->datalen;
+			nextseqnum = nextbuf->seqnum + nextbuf->datalen;
+		}
+
+	} while (!done);
+
+	return status;
+}
+
+/**
+ * i40iw_ieq_handle_partial - process partial fpdu buffer
+ * @ieq: ieq resource
+ * @pfpdu: partial management per user qp
+ * @buf: receive buffer
+ * @fpdu_len: fpdu len in the buffer
+ */
+static enum i40iw_status_code i40iw_ieq_handle_partial(struct i40iw_puda_rsrc *ieq,
+						       struct i40iw_pfpdu *pfpdu,
+						       struct i40iw_puda_buf *buf,
+						       u16 fpdu_len)
+{
+	enum i40iw_status_code status = 0;
+	u8 *crcptr;
+	u32 mpacrc;
+	u32 seqnum = buf->seqnum;
+	struct list_head pbufl;	/* partial buffer list */
+	struct i40iw_puda_buf *txbuf = NULL;
+	struct list_head *rxlist = &pfpdu->rxlist;
+
+	INIT_LIST_HEAD(&pbufl);
+	list_add(&buf->list, &pbufl);
+
+	status = i40iw_ieq_create_pbufl(pfpdu, rxlist, &pbufl, buf, fpdu_len);
+	if (!status)
+		goto error;
+
+	txbuf = i40iw_puda_get_bufpool(ieq);
+	if (!txbuf) {
+		pfpdu->no_tx_bufs++;
+		status = I40IW_ERR_NO_TXBUFS;
+		goto error;
+	}
+
+	i40iw_ieq_compl_pfpdu(ieq, rxlist, &pbufl, txbuf, fpdu_len);
+	i40iw_ieq_update_tcpip_info(txbuf, fpdu_len, seqnum);
+	crcptr = txbuf->data + fpdu_len - 4;
+	mpacrc = *(u32 *)crcptr;
+	if (ieq->check_crc) {
+		status = i40iw_ieq_check_mpacrc(ieq->hash_desc, txbuf->data,
+						(fpdu_len - 4), mpacrc);
+		if (status) {
+			i40iw_debug(ieq->dev, I40IW_DEBUG_IEQ,
+				    "%s: error bad crc\n", __func__);
+			goto error;
+		}
+	}
+
+	i40iw_debug_buf(ieq->dev, I40IW_DEBUG_IEQ, "IEQ TX BUFFER",
+			txbuf->mem.va, txbuf->totallen);
+	i40iw_puda_send_buf(ieq, txbuf);
+	pfpdu->rcv_nxt = seqnum + fpdu_len;
+	return status;
+ error:
+	while (!list_empty(&pbufl)) {
+		buf = (struct i40iw_puda_buf *)(pbufl.prev);
+		list_del(&buf->list);
+		list_add(&buf->list, rxlist);
+	}
+	if (txbuf)
+		i40iw_puda_ret_bufpool(ieq, txbuf);
+	return status;
+}
+
+/**
+ * i40iw_ieq_process_buf - process buffer rcvd for ieq
+ * @ieq: ieq resource
+ * @pfpdu: partial management per user qp
+ * @buf: receive buffer
+ */
+static enum i40iw_status_code i40iw_ieq_process_buf(struct i40iw_puda_rsrc *ieq,
+						    struct i40iw_pfpdu *pfpdu,
+						    struct i40iw_puda_buf *buf)
+{
+	u16 fpdu_len = 0;
+	u16 datalen = buf->datalen;
+	u8 *datap = buf->data;
+	u8 *crcptr;
+	u16 ioffset = 0;
+	u32 mpacrc;
+	u32 seqnum = buf->seqnum;
+	u16 length = 0;
+	u16 full = 0;
+	bool partial = false;
+	struct i40iw_puda_buf *txbuf;
+	struct list_head *rxlist = &pfpdu->rxlist;
+	enum i40iw_status_code ret = 0;
+	enum i40iw_status_code status = 0;
+
+	ioffset = (u16)(buf->data - (u8 *)buf->mem.va);
+	while (datalen) {
+		fpdu_len = i40iw_ieq_get_fpdu_length(ntohs(*(u16 *)datap));
+		if (fpdu_len > pfpdu->max_fpdu_data) {
+			i40iw_debug(ieq->dev, I40IW_DEBUG_IEQ,
+				    "%s: error bad fpdu_len\n", __func__);
+			status = I40IW_ERR_MPA_CRC;
+			list_add(&buf->list, rxlist);
+			return status;
+		}
+
+		if (datalen < fpdu_len) {
+			partial = true;
+			break;
+		}
+		crcptr = datap + fpdu_len - 4;
+		mpacrc = *(u32 *)crcptr;
+		if (ieq->check_crc)
+			ret = i40iw_ieq_check_mpacrc(ieq->hash_desc,
+						     datap, fpdu_len - 4, mpacrc);
+		if (ret) {
+			status = I40IW_ERR_MPA_CRC;
+			list_add(&buf->list, rxlist);
+			return status;
+		}
+		full++;
+		pfpdu->fpdu_processed++;
+		datap += fpdu_len;
+		length += fpdu_len;
+		datalen -= fpdu_len;
+	}
+	if (full) {
+		/* copy full pdu's in the txbuf and send them out */
+		txbuf = i40iw_puda_get_bufpool(ieq);
+		if (!txbuf) {
+			pfpdu->no_tx_bufs++;
+			status = I40IW_ERR_NO_TXBUFS;
+			list_add(&buf->list, rxlist);
+			return status;
+		}
+		/* modify txbuf's buffer header */
+		i40iw_ieq_setup_tx_buf(buf, txbuf);
+		/* copy full fpdu's to new buffer */
+		i40iw_ieq_copy_to_txbuf(buf, txbuf, ioffset, buf->hdrlen,
+					length);
+		txbuf->totallen = buf->hdrlen + length;
+
+		i40iw_ieq_update_tcpip_info(txbuf, length, buf->seqnum);
+		i40iw_puda_send_buf(ieq, txbuf);
+
+		if (!datalen) {
+			pfpdu->rcv_nxt = buf->seqnum + length;
+			i40iw_puda_ret_bufpool(ieq, buf);
+			return status;
+		}
+		buf->data = datap;
+		buf->seqnum = seqnum + length;
+		buf->datalen = datalen;
+		pfpdu->rcv_nxt = buf->seqnum;
+	}
+	if (partial)
+		status = i40iw_ieq_handle_partial(ieq, pfpdu, buf, fpdu_len);
+
+	return status;
+}
+
+/**
+ * i40iw_ieq_process_fpdus - process fpdu's buffers on its list
+ * @qp: qp for which partial fpdus
+ * @ieq: ieq resource
+ */
+static void i40iw_ieq_process_fpdus(struct i40iw_sc_qp *qp,
+				    struct i40iw_puda_rsrc *ieq)
+{
+	struct i40iw_pfpdu *pfpdu = &qp->pfpdu;
+	struct list_head *rxlist = &pfpdu->rxlist;
+	struct i40iw_puda_buf *buf;
+	enum i40iw_status_code status;
+
+	do {
+		if (list_empty(rxlist))
+			break;
+		buf = i40iw_puda_get_listbuf(rxlist);
+		if (!buf) {
+			i40iw_debug(ieq->dev, I40IW_DEBUG_IEQ,
+				    "%s: error no buf\n", __func__);
+			break;
+		}
+		if (buf->seqnum != pfpdu->rcv_nxt) {
+			/* This could be out of order or missing packet */
+			pfpdu->out_of_order++;
+			list_add(&buf->list, rxlist);
+			break;
+		}
+		/* keep processing buffers from the head of the list */
+		status = i40iw_ieq_process_buf(ieq, pfpdu, buf);
+		if (status == I40IW_ERR_MPA_CRC) {
+			pfpdu->mpa_crc_err = true;
+			while (!list_empty(rxlist)) {
+				buf = i40iw_puda_get_listbuf(rxlist);
+				i40iw_puda_ret_bufpool(ieq, buf);
+				pfpdu->crc_err++;
+			}
+			/* create CQP for AE */
+			i40iw_ieq_mpa_crc_ae(ieq->dev, qp);
+		}
+	} while (!status);
+}
+
+/**
+ * i40iw_ieq_handle_exception - handle qp's exception
+ * @ieq: ieq resource
+ * @qp: qp receiving excpetion
+ * @buf: receive buffer
+ */
+static void i40iw_ieq_handle_exception(struct i40iw_puda_rsrc *ieq,
+				       struct i40iw_sc_qp *qp,
+				       struct i40iw_puda_buf *buf)
+{
+	struct i40iw_puda_buf *tmpbuf = NULL;
+	struct i40iw_pfpdu *pfpdu = &qp->pfpdu;
+	u32 *hw_host_ctx = (u32 *)qp->hw_host_ctx;
+	u32 rcv_wnd = hw_host_ctx[23];
+	/* first partial seq # in q2 */
+	u32 fps = qp->q2_buf[16];
+	struct list_head *rxlist = &pfpdu->rxlist;
+	struct list_head *plist;
+
+	pfpdu->total_ieq_bufs++;
+
+	if (pfpdu->mpa_crc_err) {
+		pfpdu->crc_err++;
+		goto error;
+	}
+	if (pfpdu->mode && (fps != pfpdu->fps)) {
+		/* clean up qp as it is new partial sequence */
+		i40iw_ieq_cleanup_qp(ieq->dev, qp);
+		i40iw_debug(ieq->dev, I40IW_DEBUG_IEQ,
+			    "%s: restarting new partial\n", __func__);
+		pfpdu->mode = false;
+	}
+
+	if (!pfpdu->mode) {
+		i40iw_debug_buf(ieq->dev, I40IW_DEBUG_IEQ, "Q2 BUFFER", (u64 *)qp->q2_buf, 128);
+		/* First_Partial_Sequence_Number check */
+		pfpdu->rcv_nxt = fps;
+		pfpdu->fps = fps;
+		pfpdu->mode = true;
+		pfpdu->max_fpdu_data = ieq->mss;
+		pfpdu->pmode_count++;
+		INIT_LIST_HEAD(rxlist);
+		i40iw_ieq_check_first_buf(buf, fps);
+	}
+
+	if (!(rcv_wnd >= (buf->seqnum - pfpdu->rcv_nxt))) {
+		pfpdu->bad_seq_num++;
+		goto error;
+	}
+
+	if (!list_empty(rxlist)) {
+		tmpbuf = (struct i40iw_puda_buf *)rxlist->next;
+		plist = &tmpbuf->list;
+		while ((struct list_head *)tmpbuf != rxlist) {
+			if ((int)(buf->seqnum - tmpbuf->seqnum) < 0)
+				break;
+			tmpbuf = (struct i40iw_puda_buf *)plist->next;
+		}
+		/* Insert buf before tmpbuf */
+		list_add_tail(&buf->list, &tmpbuf->list);
+	} else {
+		list_add_tail(&buf->list, rxlist);
+	}
+	i40iw_ieq_process_fpdus(qp, ieq);
+	return;
+ error:
+	i40iw_puda_ret_bufpool(ieq, buf);
+}
+
+/**
+ * i40iw_ieq_receive - received exception buffer
+ * @dev: iwarp device
+ * @buf: exception buffer received
+ */
+static void i40iw_ieq_receive(struct i40iw_sc_dev *dev,
+			      struct i40iw_puda_buf *buf)
+{
+	struct i40iw_puda_rsrc *ieq = dev->ieq;
+	struct i40iw_sc_qp *qp = NULL;
+	u32 wqe_idx = ieq->compl_rxwqe_idx;
+
+	qp = i40iw_ieq_get_qp(dev, buf);
+	if (!qp) {
+		ieq->stats_bad_qp_id++;
+		i40iw_puda_ret_bufpool(ieq, buf);
+	} else {
+		i40iw_ieq_handle_exception(ieq, qp, buf);
+	}
+	/*
+	 * ieq->rx_wqe_idx is used by i40iw_puda_replenish_rq()
+	 * on which wqe_idx to start replenish rq
+	 */
+	if (!ieq->rxq_invalid_cnt)
+		ieq->rx_wqe_idx = wqe_idx;
+	ieq->rxq_invalid_cnt++;
+}
+
+/**
+ * i40iw_ieq_tx_compl - put back after sending completed exception buffer
+ * @dev: iwarp device
+ * @sqwrid: pointer to puda buffer
+ */
+static void i40iw_ieq_tx_compl(struct i40iw_sc_dev *dev, void *sqwrid)
+{
+	struct i40iw_puda_rsrc *ieq = dev->ieq;
+	struct i40iw_puda_buf *buf = (struct i40iw_puda_buf *)sqwrid;
+
+	i40iw_puda_ret_bufpool(ieq, buf);
+	if (!list_empty(&ieq->txpend)) {
+		buf = i40iw_puda_get_listbuf(&ieq->txpend);
+		i40iw_puda_send_buf(ieq, buf);
+	}
+}
+
+/**
+ * i40iw_ieq_cleanup_qp - qp is being destroyed
+ * @dev: iwarp device
+ * @qp: all pending fpdu buffers
+ */
+void i40iw_ieq_cleanup_qp(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp)
+{
+	struct i40iw_puda_buf *buf;
+	struct i40iw_pfpdu *pfpdu = &qp->pfpdu;
+	struct list_head *rxlist = &pfpdu->rxlist;
+	struct i40iw_puda_rsrc *ieq = dev->ieq;
+
+	if (!pfpdu->mode)
+		return;
+	while (!list_empty(rxlist)) {
+		buf = i40iw_puda_get_listbuf(rxlist);
+		i40iw_puda_ret_bufpool(ieq, buf);
+	}
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_puda.h b/drivers/infiniband/hw/i40iw/i40iw_puda.h
new file mode 100644
index 000000000000..52bf7826ce4e
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_puda.h
@@ -0,0 +1,183 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_PUDA_H
+#define I40IW_PUDA_H
+
+#define I40IW_IEQ_MPA_FRAMING 6
+
+struct i40iw_sc_dev;
+struct i40iw_sc_qp;
+struct i40iw_sc_cq;
+
+enum puda_resource_type {
+	I40IW_PUDA_RSRC_TYPE_ILQ = 1,
+	I40IW_PUDA_RSRC_TYPE_IEQ
+};
+
+enum puda_rsrc_complete {
+	PUDA_CQ_CREATED = 1,
+	PUDA_QP_CREATED,
+	PUDA_TX_COMPLETE,
+	PUDA_RX_COMPLETE,
+	PUDA_HASH_CRC_COMPLETE
+};
+
+struct i40iw_puda_completion_info {
+	struct i40iw_qp_uk *qp;
+	u8 q_type;
+	u8 vlan_valid;
+	u8 l3proto;
+	u8 l4proto;
+	u16 payload_len;
+	u32 compl_error;	/* No_err=0, else major and minor err code */
+	u32 qp_id;
+	u32 wqe_idx;
+};
+
+struct i40iw_puda_send_info {
+	u64 paddr;		/* Physical address */
+	u32 len;
+	u8 tcplen;
+	u8 maclen;
+	bool ipv4;
+	bool doloopback;
+	void *scratch;
+};
+
+struct i40iw_puda_buf {
+	struct list_head list;	/* MUST be first entry */
+	struct i40iw_dma_mem mem;	/* DMA memory for the buffer */
+	struct i40iw_puda_buf *next;	/* for alloclist in rsrc struct */
+	struct i40iw_virt_mem buf_mem;	/* Buffer memory for this buffer */
+	void *scratch;
+	u8 *iph;
+	u8 *tcph;
+	u8 *data;
+	u16 datalen;
+	u16 vlan_id;
+	u8 tcphlen;		/* tcp length in bytes */
+	u8 maclen;		/* mac length in bytes */
+	u32 totallen;		/* machlen+iphlen+tcphlen+datalen */
+	atomic_t refcount;
+	u8 hdrlen;
+	bool ipv4;
+	u32 seqnum;
+};
+
+struct i40iw_puda_rsrc_info {
+	enum puda_resource_type type;	/* ILQ or IEQ */
+	u32 count;
+	u16 pd_id;
+	u32 cq_id;
+	u32 qp_id;
+	u32 sq_size;
+	u32 rq_size;
+	u16 buf_size;
+	u16 mss;
+	u32 tx_buf_cnt;		/* total bufs allocated will be rq_size + tx_buf_cnt */
+	void (*receive)(struct i40iw_sc_dev *, struct i40iw_puda_buf *);
+	void (*xmit_complete)(struct i40iw_sc_dev *, void *);
+};
+
+struct i40iw_puda_rsrc {
+	struct i40iw_sc_cq cq;
+	struct i40iw_sc_qp qp;
+	struct i40iw_sc_pd sc_pd;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_dma_mem cqmem;
+	struct i40iw_dma_mem qpmem;
+	struct i40iw_virt_mem ilq_mem;
+	enum puda_rsrc_complete completion;
+	enum puda_resource_type type;
+	u16 buf_size;		/*buffer must be max datalen + tcpip hdr + mac */
+	u16 mss;
+	u32 cq_id;
+	u32 qp_id;
+	u32 sq_size;
+	u32 rq_size;
+	u32 cq_size;
+	struct i40iw_sq_uk_wr_trk_info *sq_wrtrk_array;
+	u64 *rq_wrid_array;
+	u32 compl_rxwqe_idx;
+	u32 rx_wqe_idx;
+	u32 rxq_invalid_cnt;
+	u32 tx_wqe_avail_cnt;
+	bool check_crc;
+	struct shash_desc *hash_desc;
+	struct list_head txpend;
+	struct list_head bufpool;	/* free buffers pool list for recv and xmit */
+	u32 alloc_buf_count;
+	u32 avail_buf_count;		/* snapshot of currently available buffers */
+	spinlock_t bufpool_lock;
+	struct i40iw_puda_buf *alloclist;
+	void (*receive)(struct i40iw_sc_dev *, struct i40iw_puda_buf *);
+	void (*xmit_complete)(struct i40iw_sc_dev *, void *);
+	/* puda stats */
+	u64 stats_buf_alloc_fail;
+	u64 stats_pkt_rcvd;
+	u64 stats_pkt_sent;
+	u64 stats_rcvd_pkt_err;
+	u64 stats_sent_pkt_q;
+	u64 stats_bad_qp_id;
+};
+
+struct i40iw_puda_buf *i40iw_puda_get_bufpool(struct i40iw_puda_rsrc *rsrc);
+void i40iw_puda_ret_bufpool(struct i40iw_puda_rsrc *rsrc,
+			    struct i40iw_puda_buf *buf);
+void i40iw_puda_send_buf(struct i40iw_puda_rsrc *rsrc,
+			 struct i40iw_puda_buf *buf);
+enum i40iw_status_code i40iw_puda_send(struct i40iw_sc_qp *qp,
+				       struct i40iw_puda_send_info *info);
+enum i40iw_status_code i40iw_puda_create_rsrc(struct i40iw_sc_dev *dev,
+					      struct i40iw_puda_rsrc_info *info);
+void i40iw_puda_dele_resources(struct i40iw_sc_dev *dev,
+			       enum puda_resource_type type,
+			       bool reset);
+enum i40iw_status_code i40iw_puda_poll_completion(struct i40iw_sc_dev *dev,
+						  struct i40iw_sc_cq *cq, u32 *compl_err);
+void i40iw_ieq_cleanup_qp(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp);
+
+struct i40iw_sc_qp *i40iw_ieq_get_qp(struct i40iw_sc_dev *dev,
+				     struct i40iw_puda_buf *buf);
+enum i40iw_status_code i40iw_puda_get_tcpip_info(struct i40iw_puda_completion_info *info,
+						 struct i40iw_puda_buf *buf);
+enum i40iw_status_code i40iw_ieq_check_mpacrc(struct shash_desc *desc,
+					      void *addr, u32 length, u32 value);
+enum i40iw_status_code i40iw_init_hash_desc(struct shash_desc **desc);
+void i40iw_ieq_mpa_crc_ae(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp);
+void i40iw_free_hash_desc(struct shash_desc *desc);
+void i40iw_ieq_update_tcpip_info(struct i40iw_puda_buf *buf, u16 length,
+				 u32 seqnum);
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_register.h b/drivers/infiniband/hw/i40iw/i40iw_register.h
new file mode 100644
index 000000000000..57768184e251
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_register.h
@@ -0,0 +1,1030 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_REGISTER_H
+#define I40IW_REGISTER_H
+
+#define I40E_GLGEN_STAT               0x000B612C /* Reset: POR */
+
+#define I40E_PFHMC_PDINV               0x000C0300 /* Reset: PFR */
+#define I40E_PFHMC_PDINV_PMSDIDX_SHIFT 0
+#define I40E_PFHMC_PDINV_PMSDIDX_MASK  (0xFFF <<  I40E_PFHMC_PDINV_PMSDIDX_SHIFT)
+#define I40E_PFHMC_PDINV_PMPDIDX_SHIFT 16
+#define I40E_PFHMC_PDINV_PMPDIDX_MASK  (0x1FF <<  I40E_PFHMC_PDINV_PMPDIDX_SHIFT)
+#define I40E_PFHMC_SDCMD_PMSDWR_SHIFT  31
+#define I40E_PFHMC_SDCMD_PMSDWR_MASK   (0x1 <<  I40E_PFHMC_SDCMD_PMSDWR_SHIFT)
+#define I40E_PFHMC_SDDATALOW_PMSDVALID_SHIFT   0
+#define I40E_PFHMC_SDDATALOW_PMSDVALID_MASK    (0x1 <<  I40E_PFHMC_SDDATALOW_PMSDVALID_SHIFT)
+#define I40E_PFHMC_SDDATALOW_PMSDTYPE_SHIFT    1
+#define I40E_PFHMC_SDDATALOW_PMSDTYPE_MASK     (0x1 <<  I40E_PFHMC_SDDATALOW_PMSDTYPE_SHIFT)
+#define I40E_PFHMC_SDDATALOW_PMSDBPCOUNT_SHIFT 2
+#define I40E_PFHMC_SDDATALOW_PMSDBPCOUNT_MASK  (0x3FF <<  I40E_PFHMC_SDDATALOW_PMSDBPCOUNT_SHIFT)
+
+#define I40E_PFINT_DYN_CTLN(_INTPF) (0x00034800 + ((_INTPF) * 4)) /* _i=0...511 */	/* Reset: PFR */
+#define I40E_PFINT_DYN_CTLN_INTENA_SHIFT          0
+#define I40E_PFINT_DYN_CTLN_INTENA_MASK           (0x1 <<  I40E_PFINT_DYN_CTLN_INTENA_SHIFT)
+#define I40E_PFINT_DYN_CTLN_CLEARPBA_SHIFT        1
+#define I40E_PFINT_DYN_CTLN_CLEARPBA_MASK         (0x1 <<  I40E_PFINT_DYN_CTLN_CLEARPBA_SHIFT)
+#define I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT        3
+#define I40E_PFINT_DYN_CTLN_ITR_INDX_MASK         (0x3 <<  I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT)
+
+#define I40E_VFINT_DYN_CTLN1(_INTVF)               (0x00003800 + ((_INTVF) * 4)) /* _i=0...15 */ /* Reset: VFR */
+#define I40E_GLHMC_VFPDINV(_i)               (0x000C8300 + ((_i) * 4)) /* _i=0...31 */ /* Reset: CORER */
+
+#define I40E_PFHMC_PDINV_PMSDPARTSEL_SHIFT 15
+#define I40E_PFHMC_PDINV_PMSDPARTSEL_MASK  (0x1 <<  I40E_PFHMC_PDINV_PMSDPARTSEL_SHIFT)
+#define I40E_GLPCI_LBARCTRL                    0x000BE484 /* Reset: POR */
+#define I40E_GLPCI_LBARCTRL_PE_DB_SIZE_SHIFT    4
+#define I40E_GLPCI_LBARCTRL_PE_DB_SIZE_MASK     (0x3 <<  I40E_GLPCI_LBARCTRL_PE_DB_SIZE_SHIFT)
+#define I40E_GLPCI_DREVID			0x0009C480 /* Reset: PCIR */
+#define I40E_GLPCI_DREVID_DEFAULT_REVID_SHIFT 0
+#define I40E_GLPCI_DREVID_DEFAULT_REVID_MASK 0xFF
+
+#define I40E_PFPE_AEQALLOC               0x00131180 /* Reset: PFR */
+#define I40E_PFPE_AEQALLOC_AECOUNT_SHIFT 0
+#define I40E_PFPE_AEQALLOC_AECOUNT_MASK  (0xFFFFFFFF <<  I40E_PFPE_AEQALLOC_AECOUNT_SHIFT)
+#define I40E_PFPE_CCQPHIGH                  0x00008200 /* Reset: PFR */
+#define I40E_PFPE_CCQPHIGH_PECCQPHIGH_SHIFT 0
+#define I40E_PFPE_CCQPHIGH_PECCQPHIGH_MASK  (0xFFFFFFFF <<  I40E_PFPE_CCQPHIGH_PECCQPHIGH_SHIFT)
+#define I40E_PFPE_CCQPLOW                 0x00008180 /* Reset: PFR */
+#define I40E_PFPE_CCQPLOW_PECCQPLOW_SHIFT 0
+#define I40E_PFPE_CCQPLOW_PECCQPLOW_MASK  (0xFFFFFFFF <<  I40E_PFPE_CCQPLOW_PECCQPLOW_SHIFT)
+#define I40E_PFPE_CCQPSTATUS                   0x00008100 /* Reset: PFR */
+#define I40E_PFPE_CCQPSTATUS_CCQP_DONE_SHIFT   0
+#define I40E_PFPE_CCQPSTATUS_CCQP_DONE_MASK    (0x1 <<  I40E_PFPE_CCQPSTATUS_CCQP_DONE_SHIFT)
+#define I40E_PFPE_CCQPSTATUS_HMC_PROFILE_SHIFT 4
+#define I40E_PFPE_CCQPSTATUS_HMC_PROFILE_MASK  (0x7 <<  I40E_PFPE_CCQPSTATUS_HMC_PROFILE_SHIFT)
+#define I40E_PFPE_CCQPSTATUS_RDMA_EN_VFS_SHIFT 16
+#define I40E_PFPE_CCQPSTATUS_RDMA_EN_VFS_MASK  (0x3F <<  I40E_PFPE_CCQPSTATUS_RDMA_EN_VFS_SHIFT)
+#define I40E_PFPE_CCQPSTATUS_CCQP_ERR_SHIFT    31
+#define I40E_PFPE_CCQPSTATUS_CCQP_ERR_MASK     (0x1 <<  I40E_PFPE_CCQPSTATUS_CCQP_ERR_SHIFT)
+#define I40E_PFPE_CQACK              0x00131100 /* Reset: PFR */
+#define I40E_PFPE_CQACK_PECQID_SHIFT 0
+#define I40E_PFPE_CQACK_PECQID_MASK  (0x1FFFF <<  I40E_PFPE_CQACK_PECQID_SHIFT)
+#define I40E_PFPE_CQARM              0x00131080 /* Reset: PFR */
+#define I40E_PFPE_CQARM_PECQID_SHIFT 0
+#define I40E_PFPE_CQARM_PECQID_MASK  (0x1FFFF <<  I40E_PFPE_CQARM_PECQID_SHIFT)
+#define I40E_PFPE_CQPDB              0x00008000 /* Reset: PFR */
+#define I40E_PFPE_CQPDB_WQHEAD_SHIFT 0
+#define I40E_PFPE_CQPDB_WQHEAD_MASK  (0x7FF <<  I40E_PFPE_CQPDB_WQHEAD_SHIFT)
+#define I40E_PFPE_CQPERRCODES                      0x00008880 /* Reset: PFR */
+#define I40E_PFPE_CQPERRCODES_CQP_MINOR_CODE_SHIFT 0
+#define I40E_PFPE_CQPERRCODES_CQP_MINOR_CODE_MASK  (0xFFFF <<  I40E_PFPE_CQPERRCODES_CQP_MINOR_CODE_SHIFT)
+#define I40E_PFPE_CQPERRCODES_CQP_MAJOR_CODE_SHIFT 16
+#define I40E_PFPE_CQPERRCODES_CQP_MAJOR_CODE_MASK  (0xFFFF <<  I40E_PFPE_CQPERRCODES_CQP_MAJOR_CODE_SHIFT)
+#define I40E_PFPE_CQPTAIL                  0x00008080 /* Reset: PFR */
+#define I40E_PFPE_CQPTAIL_WQTAIL_SHIFT     0
+#define I40E_PFPE_CQPTAIL_WQTAIL_MASK      (0x7FF <<  I40E_PFPE_CQPTAIL_WQTAIL_SHIFT)
+#define I40E_PFPE_CQPTAIL_CQP_OP_ERR_SHIFT 31
+#define I40E_PFPE_CQPTAIL_CQP_OP_ERR_MASK  (0x1 <<  I40E_PFPE_CQPTAIL_CQP_OP_ERR_SHIFT)
+#define I40E_PFPE_FLMQ1ALLOCERR                   0x00008980 /* Reset: PFR */
+#define I40E_PFPE_FLMQ1ALLOCERR_ERROR_COUNT_SHIFT 0
+#define I40E_PFPE_FLMQ1ALLOCERR_ERROR_COUNT_MASK  (0xFFFF <<  I40E_PFPE_FLMQ1ALLOCERR_ERROR_COUNT_SHIFT)
+#define I40E_PFPE_FLMXMITALLOCERR                   0x00008900 /* Reset: PFR */
+#define I40E_PFPE_FLMXMITALLOCERR_ERROR_COUNT_SHIFT 0
+#define I40E_PFPE_FLMXMITALLOCERR_ERROR_COUNT_MASK  (0xFFFF <<  I40E_PFPE_FLMXMITALLOCERR_ERROR_COUNT_SHIFT)
+#define I40E_PFPE_IPCONFIG0                        0x00008280 /* Reset: PFR */
+#define I40E_PFPE_IPCONFIG0_PEIPID_SHIFT           0
+#define I40E_PFPE_IPCONFIG0_PEIPID_MASK            (0xFFFF <<  I40E_PFPE_IPCONFIG0_PEIPID_SHIFT)
+#define I40E_PFPE_IPCONFIG0_USEENTIREIDRANGE_SHIFT 16
+#define I40E_PFPE_IPCONFIG0_USEENTIREIDRANGE_MASK  (0x1 <<  I40E_PFPE_IPCONFIG0_USEENTIREIDRANGE_SHIFT)
+#define I40E_PFPE_MRTEIDXMASK                       0x00008600 /* Reset: PFR */
+#define I40E_PFPE_MRTEIDXMASK_MRTEIDXMASKBITS_SHIFT 0
+#define I40E_PFPE_MRTEIDXMASK_MRTEIDXMASKBITS_MASK  (0x1F <<  I40E_PFPE_MRTEIDXMASK_MRTEIDXMASKBITS_SHIFT)
+#define I40E_PFPE_RCVUNEXPECTEDERROR                        0x00008680 /* Reset: PFR */
+#define I40E_PFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_SHIFT 0
+#define I40E_PFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_MASK  (0xFFFFFF <<  I40E_PFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_SHIFT)
+#define I40E_PFPE_TCPNOWTIMER               0x00008580 /* Reset: PFR */
+#define I40E_PFPE_TCPNOWTIMER_TCP_NOW_SHIFT 0
+#define I40E_PFPE_TCPNOWTIMER_TCP_NOW_MASK  (0xFFFFFFFF <<  I40E_PFPE_TCPNOWTIMER_TCP_NOW_SHIFT)
+
+#define I40E_PFPE_WQEALLOC                      0x00138C00 /* Reset: PFR */
+#define I40E_PFPE_WQEALLOC_PEQPID_SHIFT         0
+#define I40E_PFPE_WQEALLOC_PEQPID_MASK          (0x3FFFF <<  I40E_PFPE_WQEALLOC_PEQPID_SHIFT)
+#define I40E_PFPE_WQEALLOC_WQE_DESC_INDEX_SHIFT 20
+#define I40E_PFPE_WQEALLOC_WQE_DESC_INDEX_MASK  (0xFFF <<  I40E_PFPE_WQEALLOC_WQE_DESC_INDEX_SHIFT)
+
+#define I40E_VFPE_AEQALLOC(_VF)          (0x00130C00 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_AEQALLOC_MAX_INDEX     127
+#define I40E_VFPE_AEQALLOC_AECOUNT_SHIFT 0
+#define I40E_VFPE_AEQALLOC_AECOUNT_MASK  (0xFFFFFFFF <<  I40E_VFPE_AEQALLOC_AECOUNT_SHIFT)
+#define I40E_VFPE_CCQPHIGH(_VF)             (0x00001000 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CCQPHIGH_MAX_INDEX        127
+#define I40E_VFPE_CCQPHIGH_PECCQPHIGH_SHIFT 0
+#define I40E_VFPE_CCQPHIGH_PECCQPHIGH_MASK  (0xFFFFFFFF <<  I40E_VFPE_CCQPHIGH_PECCQPHIGH_SHIFT)
+#define I40E_VFPE_CCQPLOW(_VF)            (0x00000C00 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CCQPLOW_MAX_INDEX       127
+#define I40E_VFPE_CCQPLOW_PECCQPLOW_SHIFT 0
+#define I40E_VFPE_CCQPLOW_PECCQPLOW_MASK  (0xFFFFFFFF <<  I40E_VFPE_CCQPLOW_PECCQPLOW_SHIFT)
+#define I40E_VFPE_CCQPSTATUS(_VF)              (0x00000800 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CCQPSTATUS_MAX_INDEX         127
+#define I40E_VFPE_CCQPSTATUS_CCQP_DONE_SHIFT   0
+#define I40E_VFPE_CCQPSTATUS_CCQP_DONE_MASK    (0x1 <<  I40E_VFPE_CCQPSTATUS_CCQP_DONE_SHIFT)
+#define I40E_VFPE_CCQPSTATUS_HMC_PROFILE_SHIFT 4
+#define I40E_VFPE_CCQPSTATUS_HMC_PROFILE_MASK  (0x7 <<  I40E_VFPE_CCQPSTATUS_HMC_PROFILE_SHIFT)
+#define I40E_VFPE_CCQPSTATUS_RDMA_EN_VFS_SHIFT 16
+#define I40E_VFPE_CCQPSTATUS_RDMA_EN_VFS_MASK  (0x3F <<  I40E_VFPE_CCQPSTATUS_RDMA_EN_VFS_SHIFT)
+#define I40E_VFPE_CCQPSTATUS_CCQP_ERR_SHIFT    31
+#define I40E_VFPE_CCQPSTATUS_CCQP_ERR_MASK     (0x1 <<  I40E_VFPE_CCQPSTATUS_CCQP_ERR_SHIFT)
+#define I40E_VFPE_CQACK(_VF)         (0x00130800 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CQACK_MAX_INDEX    127
+#define I40E_VFPE_CQACK_PECQID_SHIFT 0
+#define I40E_VFPE_CQACK_PECQID_MASK  (0x1FFFF <<  I40E_VFPE_CQACK_PECQID_SHIFT)
+#define I40E_VFPE_CQARM(_VF)         (0x00130400 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CQARM_MAX_INDEX    127
+#define I40E_VFPE_CQARM_PECQID_SHIFT 0
+#define I40E_VFPE_CQARM_PECQID_MASK  (0x1FFFF <<  I40E_VFPE_CQARM_PECQID_SHIFT)
+#define I40E_VFPE_CQPDB(_VF)         (0x00000000 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CQPDB_MAX_INDEX    127
+#define I40E_VFPE_CQPDB_WQHEAD_SHIFT 0
+#define I40E_VFPE_CQPDB_WQHEAD_MASK  (0x7FF <<  I40E_VFPE_CQPDB_WQHEAD_SHIFT)
+#define I40E_VFPE_CQPERRCODES(_VF)                 (0x00001800 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CQPERRCODES_MAX_INDEX            127
+#define I40E_VFPE_CQPERRCODES_CQP_MINOR_CODE_SHIFT 0
+#define I40E_VFPE_CQPERRCODES_CQP_MINOR_CODE_MASK  (0xFFFF <<  I40E_VFPE_CQPERRCODES_CQP_MINOR_CODE_SHIFT)
+#define I40E_VFPE_CQPERRCODES_CQP_MAJOR_CODE_SHIFT 16
+#define I40E_VFPE_CQPERRCODES_CQP_MAJOR_CODE_MASK  (0xFFFF <<  I40E_VFPE_CQPERRCODES_CQP_MAJOR_CODE_SHIFT)
+#define I40E_VFPE_CQPTAIL(_VF)             (0x00000400 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_CQPTAIL_MAX_INDEX        127
+#define I40E_VFPE_CQPTAIL_WQTAIL_SHIFT     0
+#define I40E_VFPE_CQPTAIL_WQTAIL_MASK      (0x7FF <<  I40E_VFPE_CQPTAIL_WQTAIL_SHIFT)
+#define I40E_VFPE_CQPTAIL_CQP_OP_ERR_SHIFT 31
+#define I40E_VFPE_CQPTAIL_CQP_OP_ERR_MASK  (0x1 <<  I40E_VFPE_CQPTAIL_CQP_OP_ERR_SHIFT)
+#define I40E_VFPE_IPCONFIG0(_VF)                   (0x00001400 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_IPCONFIG0_MAX_INDEX              127
+#define I40E_VFPE_IPCONFIG0_PEIPID_SHIFT           0
+#define I40E_VFPE_IPCONFIG0_PEIPID_MASK            (0xFFFF <<  I40E_VFPE_IPCONFIG0_PEIPID_SHIFT)
+#define I40E_VFPE_IPCONFIG0_USEENTIREIDRANGE_SHIFT 16
+#define I40E_VFPE_IPCONFIG0_USEENTIREIDRANGE_MASK  (0x1 <<  I40E_VFPE_IPCONFIG0_USEENTIREIDRANGE_SHIFT)
+#define I40E_VFPE_MRTEIDXMASK(_VF)                  (0x00003000 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_MRTEIDXMASK_MAX_INDEX             127
+#define I40E_VFPE_MRTEIDXMASK_MRTEIDXMASKBITS_SHIFT 0
+#define I40E_VFPE_MRTEIDXMASK_MRTEIDXMASKBITS_MASK  (0x1F <<  I40E_VFPE_MRTEIDXMASK_MRTEIDXMASKBITS_SHIFT)
+#define I40E_VFPE_RCVUNEXPECTEDERROR(_VF)                   (0x00003400 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_RCVUNEXPECTEDERROR_MAX_INDEX              127
+#define I40E_VFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_SHIFT 0
+#define I40E_VFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_MASK  (0xFFFFFF <<  I40E_VFPE_RCVUNEXPECTEDERROR_TCP_RX_UNEXP_ERR_SHIFT)
+#define I40E_VFPE_TCPNOWTIMER(_VF)          (0x00002C00 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_TCPNOWTIMER_MAX_INDEX     127
+#define I40E_VFPE_TCPNOWTIMER_TCP_NOW_SHIFT 0
+#define I40E_VFPE_TCPNOWTIMER_TCP_NOW_MASK  (0xFFFFFFFF <<  I40E_VFPE_TCPNOWTIMER_TCP_NOW_SHIFT)
+#define I40E_VFPE_WQEALLOC(_VF)                 (0x00138000 + ((_VF) * 4)) /* _i=0...127 */ /* Reset: VFR */
+#define I40E_VFPE_WQEALLOC_MAX_INDEX            127
+#define I40E_VFPE_WQEALLOC_PEQPID_SHIFT         0
+#define I40E_VFPE_WQEALLOC_PEQPID_MASK          (0x3FFFF <<  I40E_VFPE_WQEALLOC_PEQPID_SHIFT)
+#define I40E_VFPE_WQEALLOC_WQE_DESC_INDEX_SHIFT 20
+#define I40E_VFPE_WQEALLOC_WQE_DESC_INDEX_MASK  (0xFFF <<  I40E_VFPE_WQEALLOC_WQE_DESC_INDEX_SHIFT)
+
+#define I40E_GLPE_CPUSTATUS0                    0x0000D040 /* Reset: PE_CORER */
+#define I40E_GLPE_CPUSTATUS0_PECPUSTATUS0_SHIFT 0
+#define I40E_GLPE_CPUSTATUS0_PECPUSTATUS0_MASK  (0xFFFFFFFF <<  I40E_GLPE_CPUSTATUS0_PECPUSTATUS0_SHIFT)
+#define I40E_GLPE_CPUSTATUS1                    0x0000D044 /* Reset: PE_CORER */
+#define I40E_GLPE_CPUSTATUS1_PECPUSTATUS1_SHIFT 0
+#define I40E_GLPE_CPUSTATUS1_PECPUSTATUS1_MASK  (0xFFFFFFFF <<  I40E_GLPE_CPUSTATUS1_PECPUSTATUS1_SHIFT)
+#define I40E_GLPE_CPUSTATUS2                    0x0000D048 /* Reset: PE_CORER */
+#define I40E_GLPE_CPUSTATUS2_PECPUSTATUS2_SHIFT 0
+#define I40E_GLPE_CPUSTATUS2_PECPUSTATUS2_MASK  (0xFFFFFFFF <<  I40E_GLPE_CPUSTATUS2_PECPUSTATUS2_SHIFT)
+#define I40E_GLPE_CPUTRIG0                   0x0000D060 /* Reset: PE_CORER */
+#define I40E_GLPE_CPUTRIG0_PECPUTRIG0_SHIFT  0
+#define I40E_GLPE_CPUTRIG0_PECPUTRIG0_MASK   (0xFFFF <<  I40E_GLPE_CPUTRIG0_PECPUTRIG0_SHIFT)
+#define I40E_GLPE_CPUTRIG0_TEPREQUEST0_SHIFT 17
+#define I40E_GLPE_CPUTRIG0_TEPREQUEST0_MASK  (0x1 <<  I40E_GLPE_CPUTRIG0_TEPREQUEST0_SHIFT)
+#define I40E_GLPE_CPUTRIG0_OOPREQUEST0_SHIFT 18
+#define I40E_GLPE_CPUTRIG0_OOPREQUEST0_MASK  (0x1 <<  I40E_GLPE_CPUTRIG0_OOPREQUEST0_SHIFT)
+#define I40E_GLPE_DUAL40_RUPM                     0x0000DA04 /* Reset: PE_CORER */
+#define I40E_GLPE_DUAL40_RUPM_DUAL_40G_MODE_SHIFT 0
+#define I40E_GLPE_DUAL40_RUPM_DUAL_40G_MODE_MASK  (0x1 <<  I40E_GLPE_DUAL40_RUPM_DUAL_40G_MODE_SHIFT)
+#define I40E_GLPE_PFAEQEDROPCNT(_i)               (0x00131440 + ((_i) * 4)) /* _i=0...15 */ /* Reset: CORER */
+#define I40E_GLPE_PFAEQEDROPCNT_MAX_INDEX         15
+#define I40E_GLPE_PFAEQEDROPCNT_AEQEDROPCNT_SHIFT 0
+#define I40E_GLPE_PFAEQEDROPCNT_AEQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_PFAEQEDROPCNT_AEQEDROPCNT_SHIFT)
+#define I40E_GLPE_PFCEQEDROPCNT(_i)               (0x001313C0 + ((_i) * 4)) /* _i=0...15 */ /* Reset: CORER */
+#define I40E_GLPE_PFCEQEDROPCNT_MAX_INDEX         15
+#define I40E_GLPE_PFCEQEDROPCNT_CEQEDROPCNT_SHIFT 0
+#define I40E_GLPE_PFCEQEDROPCNT_CEQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_PFCEQEDROPCNT_CEQEDROPCNT_SHIFT)
+#define I40E_GLPE_PFCQEDROPCNT(_i)              (0x00131340 + ((_i) * 4)) /* _i=0...15 */ /* Reset: CORER */
+#define I40E_GLPE_PFCQEDROPCNT_MAX_INDEX        15
+#define I40E_GLPE_PFCQEDROPCNT_CQEDROPCNT_SHIFT 0
+#define I40E_GLPE_PFCQEDROPCNT_CQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_PFCQEDROPCNT_CQEDROPCNT_SHIFT)
+#define I40E_GLPE_RUPM_CQPPOOL                0x0000DACC /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_CQPPOOL_CQPSPADS_SHIFT 0
+#define I40E_GLPE_RUPM_CQPPOOL_CQPSPADS_MASK  (0xFF <<  I40E_GLPE_RUPM_CQPPOOL_CQPSPADS_SHIFT)
+#define I40E_GLPE_RUPM_FLRPOOL                0x0000DAC4 /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_FLRPOOL_FLRSPADS_SHIFT 0
+#define I40E_GLPE_RUPM_FLRPOOL_FLRSPADS_MASK  (0xFF <<  I40E_GLPE_RUPM_FLRPOOL_FLRSPADS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL                   0x0000DA00 /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_GCTL_ALLOFFTH_SHIFT    0
+#define I40E_GLPE_RUPM_GCTL_ALLOFFTH_MASK     (0xFF <<  I40E_GLPE_RUPM_GCTL_ALLOFFTH_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_RUPM_P0_DIS_SHIFT 26
+#define I40E_GLPE_RUPM_GCTL_RUPM_P0_DIS_MASK  (0x1 <<  I40E_GLPE_RUPM_GCTL_RUPM_P0_DIS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_RUPM_P1_DIS_SHIFT 27
+#define I40E_GLPE_RUPM_GCTL_RUPM_P1_DIS_MASK  (0x1 <<  I40E_GLPE_RUPM_GCTL_RUPM_P1_DIS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_RUPM_P2_DIS_SHIFT 28
+#define I40E_GLPE_RUPM_GCTL_RUPM_P2_DIS_MASK  (0x1 <<  I40E_GLPE_RUPM_GCTL_RUPM_P2_DIS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_RUPM_P3_DIS_SHIFT 29
+#define I40E_GLPE_RUPM_GCTL_RUPM_P3_DIS_MASK  (0x1 <<  I40E_GLPE_RUPM_GCTL_RUPM_P3_DIS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_RUPM_DIS_SHIFT    30
+#define I40E_GLPE_RUPM_GCTL_RUPM_DIS_MASK     (0x1 <<  I40E_GLPE_RUPM_GCTL_RUPM_DIS_SHIFT)
+#define I40E_GLPE_RUPM_GCTL_SWLB_MODE_SHIFT   31
+#define I40E_GLPE_RUPM_GCTL_SWLB_MODE_MASK    (0x1 <<  I40E_GLPE_RUPM_GCTL_SWLB_MODE_SHIFT)
+#define I40E_GLPE_RUPM_PTXPOOL                0x0000DAC8 /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_PTXPOOL_PTXSPADS_SHIFT 0
+#define I40E_GLPE_RUPM_PTXPOOL_PTXSPADS_MASK  (0xFF <<  I40E_GLPE_RUPM_PTXPOOL_PTXSPADS_SHIFT)
+#define I40E_GLPE_RUPM_PUSHPOOL                 0x0000DAC0 /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_PUSHPOOL_PUSHSPADS_SHIFT 0
+#define I40E_GLPE_RUPM_PUSHPOOL_PUSHSPADS_MASK  (0xFF <<  I40E_GLPE_RUPM_PUSHPOOL_PUSHSPADS_SHIFT)
+#define I40E_GLPE_RUPM_TXHOST_EN                 0x0000DA08 /* Reset: PE_CORER */
+#define I40E_GLPE_RUPM_TXHOST_EN_TXHOST_EN_SHIFT 0
+#define I40E_GLPE_RUPM_TXHOST_EN_TXHOST_EN_MASK  (0x1 <<  I40E_GLPE_RUPM_TXHOST_EN_TXHOST_EN_SHIFT)
+#define I40E_GLPE_VFAEQEDROPCNT(_i)               (0x00132540 + ((_i) * 4)) /* _i=0...31 */ /* Reset: CORER */
+#define I40E_GLPE_VFAEQEDROPCNT_MAX_INDEX         31
+#define I40E_GLPE_VFAEQEDROPCNT_AEQEDROPCNT_SHIFT 0
+#define I40E_GLPE_VFAEQEDROPCNT_AEQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_VFAEQEDROPCNT_AEQEDROPCNT_SHIFT)
+#define I40E_GLPE_VFCEQEDROPCNT(_i)               (0x00132440 + ((_i) * 4)) /* _i=0...31 */ /* Reset: CORER */
+#define I40E_GLPE_VFCEQEDROPCNT_MAX_INDEX         31
+#define I40E_GLPE_VFCEQEDROPCNT_CEQEDROPCNT_SHIFT 0
+#define I40E_GLPE_VFCEQEDROPCNT_CEQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_VFCEQEDROPCNT_CEQEDROPCNT_SHIFT)
+#define I40E_GLPE_VFCQEDROPCNT(_i)              (0x00132340 + ((_i) * 4)) /* _i=0...31 */ /* Reset: CORER */
+#define I40E_GLPE_VFCQEDROPCNT_MAX_INDEX        31
+#define I40E_GLPE_VFCQEDROPCNT_CQEDROPCNT_SHIFT 0
+#define I40E_GLPE_VFCQEDROPCNT_CQEDROPCNT_MASK  (0xFFFF <<  I40E_GLPE_VFCQEDROPCNT_CQEDROPCNT_SHIFT)
+#define I40E_GLPE_VFFLMOBJCTRL(_i)                  (0x0000D400 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPE_VFFLMOBJCTRL_MAX_INDEX            31
+#define I40E_GLPE_VFFLMOBJCTRL_XMIT_BLOCKSIZE_SHIFT 0
+#define I40E_GLPE_VFFLMOBJCTRL_XMIT_BLOCKSIZE_MASK  (0x7 <<  I40E_GLPE_VFFLMOBJCTRL_XMIT_BLOCKSIZE_SHIFT)
+#define I40E_GLPE_VFFLMOBJCTRL_Q1_BLOCKSIZE_SHIFT   8
+#define I40E_GLPE_VFFLMOBJCTRL_Q1_BLOCKSIZE_MASK    (0x7 <<  I40E_GLPE_VFFLMOBJCTRL_Q1_BLOCKSIZE_SHIFT)
+#define I40E_GLPE_VFFLMQ1ALLOCERR(_i)               (0x0000C700 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPE_VFFLMQ1ALLOCERR_MAX_INDEX         31
+#define I40E_GLPE_VFFLMQ1ALLOCERR_ERROR_COUNT_SHIFT 0
+#define I40E_GLPE_VFFLMQ1ALLOCERR_ERROR_COUNT_MASK  (0xFFFF <<  I40E_GLPE_VFFLMQ1ALLOCERR_ERROR_COUNT_SHIFT)
+#define I40E_GLPE_VFFLMXMITALLOCERR(_i)               (0x0000C600 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPE_VFFLMXMITALLOCERR_MAX_INDEX         31
+#define I40E_GLPE_VFFLMXMITALLOCERR_ERROR_COUNT_SHIFT 0
+#define I40E_GLPE_VFFLMXMITALLOCERR_ERROR_COUNT_MASK  (0xFFFF <<  I40E_GLPE_VFFLMXMITALLOCERR_ERROR_COUNT_SHIFT)
+#define I40E_GLPE_VFUDACTRL(_i)                    (0x0000C000 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPE_VFUDACTRL_MAX_INDEX              31
+#define I40E_GLPE_VFUDACTRL_IPV4MCFRAGRESBP_SHIFT  0
+#define I40E_GLPE_VFUDACTRL_IPV4MCFRAGRESBP_MASK   (0x1 <<  I40E_GLPE_VFUDACTRL_IPV4MCFRAGRESBP_SHIFT)
+#define I40E_GLPE_VFUDACTRL_IPV4UCFRAGRESBP_SHIFT  1
+#define I40E_GLPE_VFUDACTRL_IPV4UCFRAGRESBP_MASK   (0x1 <<  I40E_GLPE_VFUDACTRL_IPV4UCFRAGRESBP_SHIFT)
+#define I40E_GLPE_VFUDACTRL_IPV6MCFRAGRESBP_SHIFT  2
+#define I40E_GLPE_VFUDACTRL_IPV6MCFRAGRESBP_MASK   (0x1 <<  I40E_GLPE_VFUDACTRL_IPV6MCFRAGRESBP_SHIFT)
+#define I40E_GLPE_VFUDACTRL_IPV6UCFRAGRESBP_SHIFT  3
+#define I40E_GLPE_VFUDACTRL_IPV6UCFRAGRESBP_MASK   (0x1 <<  I40E_GLPE_VFUDACTRL_IPV6UCFRAGRESBP_SHIFT)
+#define I40E_GLPE_VFUDACTRL_UDPMCFRAGRESFAIL_SHIFT 4
+#define I40E_GLPE_VFUDACTRL_UDPMCFRAGRESFAIL_MASK  (0x1 <<  I40E_GLPE_VFUDACTRL_UDPMCFRAGRESFAIL_SHIFT)
+#define I40E_GLPE_VFUDAUCFBQPN(_i)         (0x0000C100 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPE_VFUDAUCFBQPN_MAX_INDEX   31
+#define I40E_GLPE_VFUDAUCFBQPN_QPN_SHIFT   0
+#define I40E_GLPE_VFUDAUCFBQPN_QPN_MASK    (0x3FFFF <<  I40E_GLPE_VFUDAUCFBQPN_QPN_SHIFT)
+#define I40E_GLPE_VFUDAUCFBQPN_VALID_SHIFT 31
+#define I40E_GLPE_VFUDAUCFBQPN_VALID_MASK  (0x1 <<  I40E_GLPE_VFUDAUCFBQPN_VALID_SHIFT)
+
+#define I40E_GLPES_PFIP4RXDISCARD(_i)                (0x00010600 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXDISCARD_MAX_INDEX          15
+#define I40E_GLPES_PFIP4RXDISCARD_IP4RXDISCARD_SHIFT 0
+#define I40E_GLPES_PFIP4RXDISCARD_IP4RXDISCARD_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXDISCARD_IP4RXDISCARD_SHIFT)
+#define I40E_GLPES_PFIP4RXFRAGSHI(_i)                (0x00010804 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXFRAGSHI_MAX_INDEX          15
+#define I40E_GLPES_PFIP4RXFRAGSHI_IP4RXFRAGSHI_SHIFT 0
+#define I40E_GLPES_PFIP4RXFRAGSHI_IP4RXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4RXFRAGSHI_IP4RXFRAGSHI_SHIFT)
+#define I40E_GLPES_PFIP4RXFRAGSLO(_i)                (0x00010800 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXFRAGSLO_MAX_INDEX          15
+#define I40E_GLPES_PFIP4RXFRAGSLO_IP4RXFRAGSLO_SHIFT 0
+#define I40E_GLPES_PFIP4RXFRAGSLO_IP4RXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXFRAGSLO_IP4RXFRAGSLO_SHIFT)
+#define I40E_GLPES_PFIP4RXMCOCTSHI(_i)                 (0x00010A04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXMCOCTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP4RXMCOCTSHI_IP4RXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4RXMCOCTSHI_IP4RXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4RXMCOCTSHI_IP4RXMCOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP4RXMCOCTSLO(_i)                 (0x00010A00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXMCOCTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP4RXMCOCTSLO_IP4RXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4RXMCOCTSLO_IP4RXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXMCOCTSLO_IP4RXMCOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP4RXMCPKTSHI(_i)                 (0x00010C04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXMCPKTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP4RXMCPKTSHI_IP4RXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4RXMCPKTSHI_IP4RXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4RXMCPKTSHI_IP4RXMCPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP4RXMCPKTSLO(_i)                 (0x00010C00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXMCPKTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP4RXMCPKTSLO_IP4RXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4RXMCPKTSLO_IP4RXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXMCPKTSLO_IP4RXMCPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP4RXOCTSHI(_i)               (0x00010204 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXOCTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP4RXOCTSHI_IP4RXOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4RXOCTSHI_IP4RXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4RXOCTSHI_IP4RXOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP4RXOCTSLO(_i)               (0x00010200 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXOCTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP4RXOCTSLO_IP4RXOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4RXOCTSLO_IP4RXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXOCTSLO_IP4RXOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP4RXPKTSHI(_i)               (0x00010404 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP4RXPKTSHI_IP4RXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4RXPKTSHI_IP4RXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4RXPKTSHI_IP4RXPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP4RXPKTSLO(_i)               (0x00010400 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP4RXPKTSLO_IP4RXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4RXPKTSLO_IP4RXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXPKTSLO_IP4RXPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP4RXTRUNC(_i)              (0x00010700 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4RXTRUNC_MAX_INDEX        15
+#define I40E_GLPES_PFIP4RXTRUNC_IP4RXTRUNC_SHIFT 0
+#define I40E_GLPES_PFIP4RXTRUNC_IP4RXTRUNC_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4RXTRUNC_IP4RXTRUNC_SHIFT)
+#define I40E_GLPES_PFIP4TXFRAGSHI(_i)                (0x00011E04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXFRAGSHI_MAX_INDEX          15
+#define I40E_GLPES_PFIP4TXFRAGSHI_IP4TXFRAGSHI_SHIFT 0
+#define I40E_GLPES_PFIP4TXFRAGSHI_IP4TXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4TXFRAGSHI_IP4TXFRAGSHI_SHIFT)
+#define I40E_GLPES_PFIP4TXFRAGSLO(_i)                (0x00011E00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXFRAGSLO_MAX_INDEX          15
+#define I40E_GLPES_PFIP4TXFRAGSLO_IP4TXFRAGSLO_SHIFT 0
+#define I40E_GLPES_PFIP4TXFRAGSLO_IP4TXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4TXFRAGSLO_IP4TXFRAGSLO_SHIFT)
+#define I40E_GLPES_PFIP4TXMCOCTSHI(_i)                 (0x00012004 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXMCOCTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP4TXMCOCTSHI_IP4TXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4TXMCOCTSHI_IP4TXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4TXMCOCTSHI_IP4TXMCOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP4TXMCOCTSLO(_i)                 (0x00012000 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXMCOCTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP4TXMCOCTSLO_IP4TXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4TXMCOCTSLO_IP4TXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4TXMCOCTSLO_IP4TXMCOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP4TXMCPKTSHI(_i)                 (0x00012204 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXMCPKTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP4TXMCPKTSHI_IP4TXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4TXMCPKTSHI_IP4TXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4TXMCPKTSHI_IP4TXMCPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP4TXMCPKTSLO(_i)                 (0x00012200 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXMCPKTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP4TXMCPKTSLO_IP4TXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4TXMCPKTSLO_IP4TXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4TXMCPKTSLO_IP4TXMCPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP4TXNOROUTE(_i)                (0x00012E00 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXNOROUTE_MAX_INDEX          15
+#define I40E_GLPES_PFIP4TXNOROUTE_IP4TXNOROUTE_SHIFT 0
+#define I40E_GLPES_PFIP4TXNOROUTE_IP4TXNOROUTE_MASK  (0xFFFFFF <<  I40E_GLPES_PFIP4TXNOROUTE_IP4TXNOROUTE_SHIFT)
+#define I40E_GLPES_PFIP4TXOCTSHI(_i)               (0x00011A04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXOCTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP4TXOCTSHI_IP4TXOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4TXOCTSHI_IP4TXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4TXOCTSHI_IP4TXOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP4TXOCTSLO(_i)               (0x00011A00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXOCTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP4TXOCTSLO_IP4TXOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4TXOCTSLO_IP4TXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4TXOCTSLO_IP4TXOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP4TXPKTSHI(_i)               (0x00011C04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP4TXPKTSHI_IP4TXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP4TXPKTSHI_IP4TXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP4TXPKTSHI_IP4TXPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP4TXPKTSLO(_i)               (0x00011C00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP4TXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP4TXPKTSLO_IP4TXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP4TXPKTSLO_IP4TXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP4TXPKTSLO_IP4TXPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXDISCARD(_i)                (0x00011200 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXDISCARD_MAX_INDEX          15
+#define I40E_GLPES_PFIP6RXDISCARD_IP6RXDISCARD_SHIFT 0
+#define I40E_GLPES_PFIP6RXDISCARD_IP6RXDISCARD_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXDISCARD_IP6RXDISCARD_SHIFT)
+#define I40E_GLPES_PFIP6RXFRAGSHI(_i)                (0x00011404 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXFRAGSHI_MAX_INDEX          15
+#define I40E_GLPES_PFIP6RXFRAGSHI_IP6RXFRAGSHI_SHIFT 0
+#define I40E_GLPES_PFIP6RXFRAGSHI_IP6RXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6RXFRAGSHI_IP6RXFRAGSHI_SHIFT)
+#define I40E_GLPES_PFIP6RXFRAGSLO(_i)                (0x00011400 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXFRAGSLO_MAX_INDEX          15
+#define I40E_GLPES_PFIP6RXFRAGSLO_IP6RXFRAGSLO_SHIFT 0
+#define I40E_GLPES_PFIP6RXFRAGSLO_IP6RXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXFRAGSLO_IP6RXFRAGSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXMCOCTSHI(_i)                 (0x00011604 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXMCOCTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP6RXMCOCTSHI_IP6RXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6RXMCOCTSHI_IP6RXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6RXMCOCTSHI_IP6RXMCOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP6RXMCOCTSLO(_i)                 (0x00011600 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXMCOCTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP6RXMCOCTSLO_IP6RXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6RXMCOCTSLO_IP6RXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXMCOCTSLO_IP6RXMCOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXMCPKTSHI(_i)                 (0x00011804 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXMCPKTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP6RXMCPKTSHI_IP6RXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6RXMCPKTSHI_IP6RXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6RXMCPKTSHI_IP6RXMCPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP6RXMCPKTSLO(_i)                 (0x00011800 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXMCPKTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP6RXMCPKTSLO_IP6RXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6RXMCPKTSLO_IP6RXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXMCPKTSLO_IP6RXMCPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXOCTSHI(_i)               (0x00010E04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXOCTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP6RXOCTSHI_IP6RXOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6RXOCTSHI_IP6RXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6RXOCTSHI_IP6RXOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP6RXOCTSLO(_i)               (0x00010E00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXOCTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP6RXOCTSLO_IP6RXOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6RXOCTSLO_IP6RXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXOCTSLO_IP6RXOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXPKTSHI(_i)               (0x00011004 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP6RXPKTSHI_IP6RXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6RXPKTSHI_IP6RXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6RXPKTSHI_IP6RXPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP6RXPKTSLO(_i)               (0x00011000 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP6RXPKTSLO_IP6RXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6RXPKTSLO_IP6RXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXPKTSLO_IP6RXPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP6RXTRUNC(_i)              (0x00011300 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6RXTRUNC_MAX_INDEX        15
+#define I40E_GLPES_PFIP6RXTRUNC_IP6RXTRUNC_SHIFT 0
+#define I40E_GLPES_PFIP6RXTRUNC_IP6RXTRUNC_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6RXTRUNC_IP6RXTRUNC_SHIFT)
+#define I40E_GLPES_PFIP6TXFRAGSHI(_i)                (0x00012804 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXFRAGSHI_MAX_INDEX          15
+#define I40E_GLPES_PFIP6TXFRAGSHI_IP6TXFRAGSHI_SHIFT 0
+#define I40E_GLPES_PFIP6TXFRAGSHI_IP6TXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6TXFRAGSHI_IP6TXFRAGSHI_SHIFT)
+#define I40E_GLPES_PFIP6TXFRAGSLO(_i)                (0x00012800 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXFRAGSLO_MAX_INDEX          15
+#define I40E_GLPES_PFIP6TXFRAGSLO_IP6TXFRAGSLO_SHIFT 0
+#define I40E_GLPES_PFIP6TXFRAGSLO_IP6TXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6TXFRAGSLO_IP6TXFRAGSLO_SHIFT)
+#define I40E_GLPES_PFIP6TXMCOCTSHI(_i)                 (0x00012A04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXMCOCTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP6TXMCOCTSHI_IP6TXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6TXMCOCTSHI_IP6TXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6TXMCOCTSHI_IP6TXMCOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP6TXMCOCTSLO(_i)                 (0x00012A00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXMCOCTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP6TXMCOCTSLO_IP6TXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6TXMCOCTSLO_IP6TXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6TXMCOCTSLO_IP6TXMCOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP6TXMCPKTSHI(_i)                 (0x00012C04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXMCPKTSHI_MAX_INDEX           15
+#define I40E_GLPES_PFIP6TXMCPKTSHI_IP6TXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6TXMCPKTSHI_IP6TXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6TXMCPKTSHI_IP6TXMCPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP6TXMCPKTSLO(_i)                 (0x00012C00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXMCPKTSLO_MAX_INDEX           15
+#define I40E_GLPES_PFIP6TXMCPKTSLO_IP6TXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6TXMCPKTSLO_IP6TXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6TXMCPKTSLO_IP6TXMCPKTSLO_SHIFT)
+#define I40E_GLPES_PFIP6TXNOROUTE(_i)                (0x00012F00 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXNOROUTE_MAX_INDEX          15
+#define I40E_GLPES_PFIP6TXNOROUTE_IP6TXNOROUTE_SHIFT 0
+#define I40E_GLPES_PFIP6TXNOROUTE_IP6TXNOROUTE_MASK  (0xFFFFFF <<  I40E_GLPES_PFIP6TXNOROUTE_IP6TXNOROUTE_SHIFT)
+#define I40E_GLPES_PFIP6TXOCTSHI(_i)               (0x00012404 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXOCTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP6TXOCTSHI_IP6TXOCTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6TXOCTSHI_IP6TXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6TXOCTSHI_IP6TXOCTSHI_SHIFT)
+#define I40E_GLPES_PFIP6TXOCTSLO(_i)               (0x00012400 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXOCTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP6TXOCTSLO_IP6TXOCTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6TXOCTSLO_IP6TXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6TXOCTSLO_IP6TXOCTSLO_SHIFT)
+#define I40E_GLPES_PFIP6TXPKTSHI(_i)               (0x00012604 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFIP6TXPKTSHI_IP6TXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFIP6TXPKTSHI_IP6TXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFIP6TXPKTSHI_IP6TXPKTSHI_SHIFT)
+#define I40E_GLPES_PFIP6TXPKTSLO(_i)               (0x00012600 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFIP6TXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFIP6TXPKTSLO_IP6TXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFIP6TXPKTSLO_IP6TXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFIP6TXPKTSLO_IP6TXPKTSLO_SHIFT)
+#define I40E_GLPES_PFRDMARXRDSHI(_i)               (0x00013E04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXRDSHI_MAX_INDEX         15
+#define I40E_GLPES_PFRDMARXRDSHI_RDMARXRDSHI_SHIFT 0
+#define I40E_GLPES_PFRDMARXRDSHI_RDMARXRDSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMARXRDSHI_RDMARXRDSHI_SHIFT)
+#define I40E_GLPES_PFRDMARXRDSLO(_i)               (0x00013E00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXRDSLO_MAX_INDEX         15
+#define I40E_GLPES_PFRDMARXRDSLO_RDMARXRDSLO_SHIFT 0
+#define I40E_GLPES_PFRDMARXRDSLO_RDMARXRDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMARXRDSLO_RDMARXRDSLO_SHIFT)
+#define I40E_GLPES_PFRDMARXSNDSHI(_i)                (0x00014004 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXSNDSHI_MAX_INDEX          15
+#define I40E_GLPES_PFRDMARXSNDSHI_RDMARXSNDSHI_SHIFT 0
+#define I40E_GLPES_PFRDMARXSNDSHI_RDMARXSNDSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMARXSNDSHI_RDMARXSNDSHI_SHIFT)
+#define I40E_GLPES_PFRDMARXSNDSLO(_i)                (0x00014000 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXSNDSLO_MAX_INDEX          15
+#define I40E_GLPES_PFRDMARXSNDSLO_RDMARXSNDSLO_SHIFT 0
+#define I40E_GLPES_PFRDMARXSNDSLO_RDMARXSNDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMARXSNDSLO_RDMARXSNDSLO_SHIFT)
+#define I40E_GLPES_PFRDMARXWRSHI(_i)               (0x00013C04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXWRSHI_MAX_INDEX         15
+#define I40E_GLPES_PFRDMARXWRSHI_RDMARXWRSHI_SHIFT 0
+#define I40E_GLPES_PFRDMARXWRSHI_RDMARXWRSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMARXWRSHI_RDMARXWRSHI_SHIFT)
+#define I40E_GLPES_PFRDMARXWRSLO(_i)               (0x00013C00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMARXWRSLO_MAX_INDEX         15
+#define I40E_GLPES_PFRDMARXWRSLO_RDMARXWRSLO_SHIFT 0
+#define I40E_GLPES_PFRDMARXWRSLO_RDMARXWRSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMARXWRSLO_RDMARXWRSLO_SHIFT)
+#define I40E_GLPES_PFRDMATXRDSHI(_i)               (0x00014404 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXRDSHI_MAX_INDEX         15
+#define I40E_GLPES_PFRDMATXRDSHI_RDMARXRDSHI_SHIFT 0
+#define I40E_GLPES_PFRDMATXRDSHI_RDMARXRDSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMATXRDSHI_RDMARXRDSHI_SHIFT)
+#define I40E_GLPES_PFRDMATXRDSLO(_i)               (0x00014400 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXRDSLO_MAX_INDEX         15
+#define I40E_GLPES_PFRDMATXRDSLO_RDMARXRDSLO_SHIFT 0
+#define I40E_GLPES_PFRDMATXRDSLO_RDMARXRDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMATXRDSLO_RDMARXRDSLO_SHIFT)
+#define I40E_GLPES_PFRDMATXSNDSHI(_i)                (0x00014604 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXSNDSHI_MAX_INDEX          15
+#define I40E_GLPES_PFRDMATXSNDSHI_RDMARXSNDSHI_SHIFT 0
+#define I40E_GLPES_PFRDMATXSNDSHI_RDMARXSNDSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMATXSNDSHI_RDMARXSNDSHI_SHIFT)
+#define I40E_GLPES_PFRDMATXSNDSLO(_i)                (0x00014600 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXSNDSLO_MAX_INDEX          15
+#define I40E_GLPES_PFRDMATXSNDSLO_RDMARXSNDSLO_SHIFT 0
+#define I40E_GLPES_PFRDMATXSNDSLO_RDMARXSNDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMATXSNDSLO_RDMARXSNDSLO_SHIFT)
+#define I40E_GLPES_PFRDMATXWRSHI(_i)               (0x00014204 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXWRSHI_MAX_INDEX         15
+#define I40E_GLPES_PFRDMATXWRSHI_RDMARXWRSHI_SHIFT 0
+#define I40E_GLPES_PFRDMATXWRSHI_RDMARXWRSHI_MASK  (0xFFFF <<  I40E_GLPES_PFRDMATXWRSHI_RDMARXWRSHI_SHIFT)
+#define I40E_GLPES_PFRDMATXWRSLO(_i)               (0x00014200 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMATXWRSLO_MAX_INDEX         15
+#define I40E_GLPES_PFRDMATXWRSLO_RDMARXWRSLO_SHIFT 0
+#define I40E_GLPES_PFRDMATXWRSLO_RDMARXWRSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMATXWRSLO_RDMARXWRSLO_SHIFT)
+#define I40E_GLPES_PFRDMAVBNDHI(_i)              (0x00014804 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMAVBNDHI_MAX_INDEX        15
+#define I40E_GLPES_PFRDMAVBNDHI_RDMAVBNDHI_SHIFT 0
+#define I40E_GLPES_PFRDMAVBNDHI_RDMAVBNDHI_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMAVBNDHI_RDMAVBNDHI_SHIFT)
+#define I40E_GLPES_PFRDMAVBNDLO(_i)              (0x00014800 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMAVBNDLO_MAX_INDEX        15
+#define I40E_GLPES_PFRDMAVBNDLO_RDMAVBNDLO_SHIFT 0
+#define I40E_GLPES_PFRDMAVBNDLO_RDMAVBNDLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMAVBNDLO_RDMAVBNDLO_SHIFT)
+#define I40E_GLPES_PFRDMAVINVHI(_i)              (0x00014A04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMAVINVHI_MAX_INDEX        15
+#define I40E_GLPES_PFRDMAVINVHI_RDMAVINVHI_SHIFT 0
+#define I40E_GLPES_PFRDMAVINVHI_RDMAVINVHI_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMAVINVHI_RDMAVINVHI_SHIFT)
+#define I40E_GLPES_PFRDMAVINVLO(_i)              (0x00014A00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRDMAVINVLO_MAX_INDEX        15
+#define I40E_GLPES_PFRDMAVINVLO_RDMAVINVLO_SHIFT 0
+#define I40E_GLPES_PFRDMAVINVLO_RDMAVINVLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFRDMAVINVLO_RDMAVINVLO_SHIFT)
+#define I40E_GLPES_PFRXVLANERR(_i)             (0x00010000 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFRXVLANERR_MAX_INDEX       15
+#define I40E_GLPES_PFRXVLANERR_RXVLANERR_SHIFT 0
+#define I40E_GLPES_PFRXVLANERR_RXVLANERR_MASK  (0xFFFFFF <<  I40E_GLPES_PFRXVLANERR_RXVLANERR_SHIFT)
+#define I40E_GLPES_PFTCPRTXSEG(_i)             (0x00013600 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPRTXSEG_MAX_INDEX       15
+#define I40E_GLPES_PFTCPRTXSEG_TCPRTXSEG_SHIFT 0
+#define I40E_GLPES_PFTCPRTXSEG_TCPRTXSEG_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFTCPRTXSEG_TCPRTXSEG_SHIFT)
+#define I40E_GLPES_PFTCPRXOPTERR(_i)               (0x00013200 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPRXOPTERR_MAX_INDEX         15
+#define I40E_GLPES_PFTCPRXOPTERR_TCPRXOPTERR_SHIFT 0
+#define I40E_GLPES_PFTCPRXOPTERR_TCPRXOPTERR_MASK  (0xFFFFFF <<  I40E_GLPES_PFTCPRXOPTERR_TCPRXOPTERR_SHIFT)
+#define I40E_GLPES_PFTCPRXPROTOERR(_i)                 (0x00013300 + ((_i) * 4)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPRXPROTOERR_MAX_INDEX           15
+#define I40E_GLPES_PFTCPRXPROTOERR_TCPRXPROTOERR_SHIFT 0
+#define I40E_GLPES_PFTCPRXPROTOERR_TCPRXPROTOERR_MASK  (0xFFFFFF <<  I40E_GLPES_PFTCPRXPROTOERR_TCPRXPROTOERR_SHIFT)
+#define I40E_GLPES_PFTCPRXSEGSHI(_i)               (0x00013004 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPRXSEGSHI_MAX_INDEX         15
+#define I40E_GLPES_PFTCPRXSEGSHI_TCPRXSEGSHI_SHIFT 0
+#define I40E_GLPES_PFTCPRXSEGSHI_TCPRXSEGSHI_MASK  (0xFFFF <<  I40E_GLPES_PFTCPRXSEGSHI_TCPRXSEGSHI_SHIFT)
+#define I40E_GLPES_PFTCPRXSEGSLO(_i)               (0x00013000 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPRXSEGSLO_MAX_INDEX         15
+#define I40E_GLPES_PFTCPRXSEGSLO_TCPRXSEGSLO_SHIFT 0
+#define I40E_GLPES_PFTCPRXSEGSLO_TCPRXSEGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFTCPRXSEGSLO_TCPRXSEGSLO_SHIFT)
+#define I40E_GLPES_PFTCPTXSEGHI(_i)              (0x00013404 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPTXSEGHI_MAX_INDEX        15
+#define I40E_GLPES_PFTCPTXSEGHI_TCPTXSEGHI_SHIFT 0
+#define I40E_GLPES_PFTCPTXSEGHI_TCPTXSEGHI_MASK  (0xFFFF <<  I40E_GLPES_PFTCPTXSEGHI_TCPTXSEGHI_SHIFT)
+#define I40E_GLPES_PFTCPTXSEGLO(_i)              (0x00013400 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFTCPTXSEGLO_MAX_INDEX        15
+#define I40E_GLPES_PFTCPTXSEGLO_TCPTXSEGLO_SHIFT 0
+#define I40E_GLPES_PFTCPTXSEGLO_TCPTXSEGLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFTCPTXSEGLO_TCPTXSEGLO_SHIFT)
+#define I40E_GLPES_PFUDPRXPKTSHI(_i)               (0x00013804 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFUDPRXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFUDPRXPKTSHI_UDPRXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFUDPRXPKTSHI_UDPRXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFUDPRXPKTSHI_UDPRXPKTSHI_SHIFT)
+#define I40E_GLPES_PFUDPRXPKTSLO(_i)               (0x00013800 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFUDPRXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFUDPRXPKTSLO_UDPRXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFUDPRXPKTSLO_UDPRXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFUDPRXPKTSLO_UDPRXPKTSLO_SHIFT)
+#define I40E_GLPES_PFUDPTXPKTSHI(_i)               (0x00013A04 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFUDPTXPKTSHI_MAX_INDEX         15
+#define I40E_GLPES_PFUDPTXPKTSHI_UDPTXPKTSHI_SHIFT 0
+#define I40E_GLPES_PFUDPTXPKTSHI_UDPTXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_PFUDPTXPKTSHI_UDPTXPKTSHI_SHIFT)
+#define I40E_GLPES_PFUDPTXPKTSLO(_i)               (0x00013A00 + ((_i) * 8)) /* _i=0...15 */ /* Reset: PE_CORER */
+#define I40E_GLPES_PFUDPTXPKTSLO_MAX_INDEX         15
+#define I40E_GLPES_PFUDPTXPKTSLO_UDPTXPKTSLO_SHIFT 0
+#define I40E_GLPES_PFUDPTXPKTSLO_UDPTXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_PFUDPTXPKTSLO_UDPTXPKTSLO_SHIFT)
+#define I40E_GLPES_RDMARXMULTFPDUSHI                         0x0001E014 /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXMULTFPDUSHI_RDMARXMULTFPDUSHI_SHIFT 0
+#define I40E_GLPES_RDMARXMULTFPDUSHI_RDMARXMULTFPDUSHI_MASK  (0xFFFFFF <<  I40E_GLPES_RDMARXMULTFPDUSHI_RDMARXMULTFPDUSHI_SHIFT)
+#define I40E_GLPES_RDMARXMULTFPDUSLO                         0x0001E010 /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXMULTFPDUSLO_RDMARXMULTFPDUSLO_SHIFT 0
+#define I40E_GLPES_RDMARXMULTFPDUSLO_RDMARXMULTFPDUSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_RDMARXMULTFPDUSLO_RDMARXMULTFPDUSLO_SHIFT)
+#define I40E_GLPES_RDMARXOOODDPHI                      0x0001E01C /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXOOODDPHI_RDMARXOOODDPHI_SHIFT 0
+#define I40E_GLPES_RDMARXOOODDPHI_RDMARXOOODDPHI_MASK  (0xFFFFFF <<  I40E_GLPES_RDMARXOOODDPHI_RDMARXOOODDPHI_SHIFT)
+#define I40E_GLPES_RDMARXOOODDPLO                      0x0001E018 /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXOOODDPLO_RDMARXOOODDPLO_SHIFT 0
+#define I40E_GLPES_RDMARXOOODDPLO_RDMARXOOODDPLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_RDMARXOOODDPLO_RDMARXOOODDPLO_SHIFT)
+#define I40E_GLPES_RDMARXOOONOMARK                     0x0001E004 /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXOOONOMARK_RDMAOOONOMARK_SHIFT 0
+#define I40E_GLPES_RDMARXOOONOMARK_RDMAOOONOMARK_MASK  (0xFFFFFFFF <<  I40E_GLPES_RDMARXOOONOMARK_RDMAOOONOMARK_SHIFT)
+#define I40E_GLPES_RDMARXUNALIGN                     0x0001E000 /* Reset: PE_CORER */
+#define I40E_GLPES_RDMARXUNALIGN_RDMRXAUNALIGN_SHIFT 0
+#define I40E_GLPES_RDMARXUNALIGN_RDMRXAUNALIGN_MASK  (0xFFFFFFFF <<  I40E_GLPES_RDMARXUNALIGN_RDMRXAUNALIGN_SHIFT)
+#define I40E_GLPES_TCPRXFOURHOLEHI                       0x0001E044 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXFOURHOLEHI_TCPRXFOURHOLEHI_SHIFT 0
+#define I40E_GLPES_TCPRXFOURHOLEHI_TCPRXFOURHOLEHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPRXFOURHOLEHI_TCPRXFOURHOLEHI_SHIFT)
+#define I40E_GLPES_TCPRXFOURHOLELO                       0x0001E040 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXFOURHOLELO_TCPRXFOURHOLELO_SHIFT 0
+#define I40E_GLPES_TCPRXFOURHOLELO_TCPRXFOURHOLELO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPRXFOURHOLELO_TCPRXFOURHOLELO_SHIFT)
+#define I40E_GLPES_TCPRXONEHOLEHI                      0x0001E02C /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXONEHOLEHI_TCPRXONEHOLEHI_SHIFT 0
+#define I40E_GLPES_TCPRXONEHOLEHI_TCPRXONEHOLEHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPRXONEHOLEHI_TCPRXONEHOLEHI_SHIFT)
+#define I40E_GLPES_TCPRXONEHOLELO                      0x0001E028 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXONEHOLELO_TCPRXONEHOLELO_SHIFT 0
+#define I40E_GLPES_TCPRXONEHOLELO_TCPRXONEHOLELO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPRXONEHOLELO_TCPRXONEHOLELO_SHIFT)
+#define I40E_GLPES_TCPRXPUREACKHI                       0x0001E024 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXPUREACKHI_TCPRXPUREACKSHI_SHIFT 0
+#define I40E_GLPES_TCPRXPUREACKHI_TCPRXPUREACKSHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPRXPUREACKHI_TCPRXPUREACKSHI_SHIFT)
+#define I40E_GLPES_TCPRXPUREACKSLO                      0x0001E020 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXPUREACKSLO_TCPRXPUREACKLO_SHIFT 0
+#define I40E_GLPES_TCPRXPUREACKSLO_TCPRXPUREACKLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPRXPUREACKSLO_TCPRXPUREACKLO_SHIFT)
+#define I40E_GLPES_TCPRXTHREEHOLEHI                        0x0001E03C /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXTHREEHOLEHI_TCPRXTHREEHOLEHI_SHIFT 0
+#define I40E_GLPES_TCPRXTHREEHOLEHI_TCPRXTHREEHOLEHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPRXTHREEHOLEHI_TCPRXTHREEHOLEHI_SHIFT)
+#define I40E_GLPES_TCPRXTHREEHOLELO                        0x0001E038 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXTHREEHOLELO_TCPRXTHREEHOLELO_SHIFT 0
+#define I40E_GLPES_TCPRXTHREEHOLELO_TCPRXTHREEHOLELO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPRXTHREEHOLELO_TCPRXTHREEHOLELO_SHIFT)
+#define I40E_GLPES_TCPRXTWOHOLEHI                      0x0001E034 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXTWOHOLEHI_TCPRXTWOHOLEHI_SHIFT 0
+#define I40E_GLPES_TCPRXTWOHOLEHI_TCPRXTWOHOLEHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPRXTWOHOLEHI_TCPRXTWOHOLEHI_SHIFT)
+#define I40E_GLPES_TCPRXTWOHOLELO                      0x0001E030 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPRXTWOHOLELO_TCPRXTWOHOLELO_SHIFT 0
+#define I40E_GLPES_TCPRXTWOHOLELO_TCPRXTWOHOLELO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPRXTWOHOLELO_TCPRXTWOHOLELO_SHIFT)
+#define I40E_GLPES_TCPTXRETRANSFASTHI                          0x0001E04C /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXRETRANSFASTHI_TCPTXRETRANSFASTHI_SHIFT 0
+#define I40E_GLPES_TCPTXRETRANSFASTHI_TCPTXRETRANSFASTHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPTXRETRANSFASTHI_TCPTXRETRANSFASTHI_SHIFT)
+#define I40E_GLPES_TCPTXRETRANSFASTLO                          0x0001E048 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXRETRANSFASTLO_TCPTXRETRANSFASTLO_SHIFT 0
+#define I40E_GLPES_TCPTXRETRANSFASTLO_TCPTXRETRANSFASTLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPTXRETRANSFASTLO_TCPTXRETRANSFASTLO_SHIFT)
+#define I40E_GLPES_TCPTXTOUTSFASTHI                        0x0001E054 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXTOUTSFASTHI_TCPTXTOUTSFASTHI_SHIFT 0
+#define I40E_GLPES_TCPTXTOUTSFASTHI_TCPTXTOUTSFASTHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPTXTOUTSFASTHI_TCPTXTOUTSFASTHI_SHIFT)
+#define I40E_GLPES_TCPTXTOUTSFASTLO                        0x0001E050 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXTOUTSFASTLO_TCPTXTOUTSFASTLO_SHIFT 0
+#define I40E_GLPES_TCPTXTOUTSFASTLO_TCPTXTOUTSFASTLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPTXTOUTSFASTLO_TCPTXTOUTSFASTLO_SHIFT)
+#define I40E_GLPES_TCPTXTOUTSHI                    0x0001E05C /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXTOUTSHI_TCPTXTOUTSHI_SHIFT 0
+#define I40E_GLPES_TCPTXTOUTSHI_TCPTXTOUTSHI_MASK  (0xFFFFFF <<  I40E_GLPES_TCPTXTOUTSHI_TCPTXTOUTSHI_SHIFT)
+#define I40E_GLPES_TCPTXTOUTSLO                    0x0001E058 /* Reset: PE_CORER */
+#define I40E_GLPES_TCPTXTOUTSLO_TCPTXTOUTSLO_SHIFT 0
+#define I40E_GLPES_TCPTXTOUTSLO_TCPTXTOUTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_TCPTXTOUTSLO_TCPTXTOUTSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXDISCARD(_i)                (0x00018600 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXDISCARD_MAX_INDEX          31
+#define I40E_GLPES_VFIP4RXDISCARD_IP4RXDISCARD_SHIFT 0
+#define I40E_GLPES_VFIP4RXDISCARD_IP4RXDISCARD_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXDISCARD_IP4RXDISCARD_SHIFT)
+#define I40E_GLPES_VFIP4RXFRAGSHI(_i)                (0x00018804 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXFRAGSHI_MAX_INDEX          31
+#define I40E_GLPES_VFIP4RXFRAGSHI_IP4RXFRAGSHI_SHIFT 0
+#define I40E_GLPES_VFIP4RXFRAGSHI_IP4RXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4RXFRAGSHI_IP4RXFRAGSHI_SHIFT)
+#define I40E_GLPES_VFIP4RXFRAGSLO(_i)                (0x00018800 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXFRAGSLO_MAX_INDEX          31
+#define I40E_GLPES_VFIP4RXFRAGSLO_IP4RXFRAGSLO_SHIFT 0
+#define I40E_GLPES_VFIP4RXFRAGSLO_IP4RXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXFRAGSLO_IP4RXFRAGSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXMCOCTSHI(_i)                 (0x00018A04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXMCOCTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP4RXMCOCTSHI_IP4RXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4RXMCOCTSHI_IP4RXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4RXMCOCTSHI_IP4RXMCOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP4RXMCOCTSLO(_i)                 (0x00018A00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXMCOCTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP4RXMCOCTSLO_IP4RXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4RXMCOCTSLO_IP4RXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXMCOCTSLO_IP4RXMCOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXMCPKTSHI(_i)                 (0x00018C04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXMCPKTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP4RXMCPKTSHI_IP4RXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4RXMCPKTSHI_IP4RXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4RXMCPKTSHI_IP4RXMCPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP4RXMCPKTSLO(_i)                 (0x00018C00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXMCPKTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP4RXMCPKTSLO_IP4RXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4RXMCPKTSLO_IP4RXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXMCPKTSLO_IP4RXMCPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXOCTSHI(_i)               (0x00018204 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXOCTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP4RXOCTSHI_IP4RXOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4RXOCTSHI_IP4RXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4RXOCTSHI_IP4RXOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP4RXOCTSLO(_i)               (0x00018200 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXOCTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP4RXOCTSLO_IP4RXOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4RXOCTSLO_IP4RXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXOCTSLO_IP4RXOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXPKTSHI(_i)               (0x00018404 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP4RXPKTSHI_IP4RXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4RXPKTSHI_IP4RXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4RXPKTSHI_IP4RXPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP4RXPKTSLO(_i)               (0x00018400 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP4RXPKTSLO_IP4RXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4RXPKTSLO_IP4RXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXPKTSLO_IP4RXPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP4RXTRUNC(_i)              (0x00018700 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4RXTRUNC_MAX_INDEX        31
+#define I40E_GLPES_VFIP4RXTRUNC_IP4RXTRUNC_SHIFT 0
+#define I40E_GLPES_VFIP4RXTRUNC_IP4RXTRUNC_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4RXTRUNC_IP4RXTRUNC_SHIFT)
+#define I40E_GLPES_VFIP4TXFRAGSHI(_i)                (0x00019E04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXFRAGSHI_MAX_INDEX          31
+#define I40E_GLPES_VFIP4TXFRAGSHI_IP4TXFRAGSHI_SHIFT 0
+#define I40E_GLPES_VFIP4TXFRAGSHI_IP4TXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4TXFRAGSHI_IP4TXFRAGSHI_SHIFT)
+#define I40E_GLPES_VFIP4TXFRAGSLO(_i)                (0x00019E00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXFRAGSLO_MAX_INDEX          31
+#define I40E_GLPES_VFIP4TXFRAGSLO_IP4TXFRAGSLO_SHIFT 0
+#define I40E_GLPES_VFIP4TXFRAGSLO_IP4TXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4TXFRAGSLO_IP4TXFRAGSLO_SHIFT)
+#define I40E_GLPES_VFIP4TXMCOCTSHI(_i)                 (0x0001A004 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXMCOCTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP4TXMCOCTSHI_IP4TXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4TXMCOCTSHI_IP4TXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4TXMCOCTSHI_IP4TXMCOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP4TXMCOCTSLO(_i)                 (0x0001A000 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXMCOCTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP4TXMCOCTSLO_IP4TXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4TXMCOCTSLO_IP4TXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4TXMCOCTSLO_IP4TXMCOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP4TXMCPKTSHI(_i)                 (0x0001A204 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXMCPKTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP4TXMCPKTSHI_IP4TXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4TXMCPKTSHI_IP4TXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4TXMCPKTSHI_IP4TXMCPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP4TXMCPKTSLO(_i)                 (0x0001A200 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXMCPKTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP4TXMCPKTSLO_IP4TXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4TXMCPKTSLO_IP4TXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4TXMCPKTSLO_IP4TXMCPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP4TXNOROUTE(_i)                (0x0001AE00 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXNOROUTE_MAX_INDEX          31
+#define I40E_GLPES_VFIP4TXNOROUTE_IP4TXNOROUTE_SHIFT 0
+#define I40E_GLPES_VFIP4TXNOROUTE_IP4TXNOROUTE_MASK  (0xFFFFFF <<  I40E_GLPES_VFIP4TXNOROUTE_IP4TXNOROUTE_SHIFT)
+#define I40E_GLPES_VFIP4TXOCTSHI(_i)               (0x00019A04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXOCTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP4TXOCTSHI_IP4TXOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4TXOCTSHI_IP4TXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4TXOCTSHI_IP4TXOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP4TXOCTSLO(_i)               (0x00019A00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXOCTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP4TXOCTSLO_IP4TXOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4TXOCTSLO_IP4TXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4TXOCTSLO_IP4TXOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP4TXPKTSHI(_i)               (0x00019C04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP4TXPKTSHI_IP4TXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP4TXPKTSHI_IP4TXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP4TXPKTSHI_IP4TXPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP4TXPKTSLO(_i)               (0x00019C00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP4TXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP4TXPKTSLO_IP4TXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP4TXPKTSLO_IP4TXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP4TXPKTSLO_IP4TXPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXDISCARD(_i)                (0x00019200 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXDISCARD_MAX_INDEX          31
+#define I40E_GLPES_VFIP6RXDISCARD_IP6RXDISCARD_SHIFT 0
+#define I40E_GLPES_VFIP6RXDISCARD_IP6RXDISCARD_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXDISCARD_IP6RXDISCARD_SHIFT)
+#define I40E_GLPES_VFIP6RXFRAGSHI(_i)                (0x00019404 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXFRAGSHI_MAX_INDEX          31
+#define I40E_GLPES_VFIP6RXFRAGSHI_IP6RXFRAGSHI_SHIFT 0
+#define I40E_GLPES_VFIP6RXFRAGSHI_IP6RXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6RXFRAGSHI_IP6RXFRAGSHI_SHIFT)
+#define I40E_GLPES_VFIP6RXFRAGSLO(_i)                (0x00019400 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXFRAGSLO_MAX_INDEX          31
+#define I40E_GLPES_VFIP6RXFRAGSLO_IP6RXFRAGSLO_SHIFT 0
+#define I40E_GLPES_VFIP6RXFRAGSLO_IP6RXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXFRAGSLO_IP6RXFRAGSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXMCOCTSHI(_i)                 (0x00019604 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXMCOCTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP6RXMCOCTSHI_IP6RXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6RXMCOCTSHI_IP6RXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6RXMCOCTSHI_IP6RXMCOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP6RXMCOCTSLO(_i)                 (0x00019600 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXMCOCTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP6RXMCOCTSLO_IP6RXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6RXMCOCTSLO_IP6RXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXMCOCTSLO_IP6RXMCOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXMCPKTSHI(_i)                 (0x00019804 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXMCPKTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP6RXMCPKTSHI_IP6RXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6RXMCPKTSHI_IP6RXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6RXMCPKTSHI_IP6RXMCPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP6RXMCPKTSLO(_i)                 (0x00019800 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXMCPKTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP6RXMCPKTSLO_IP6RXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6RXMCPKTSLO_IP6RXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXMCPKTSLO_IP6RXMCPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXOCTSHI(_i)               (0x00018E04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXOCTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP6RXOCTSHI_IP6RXOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6RXOCTSHI_IP6RXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6RXOCTSHI_IP6RXOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP6RXOCTSLO(_i)               (0x00018E00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXOCTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP6RXOCTSLO_IP6RXOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6RXOCTSLO_IP6RXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXOCTSLO_IP6RXOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXPKTSHI(_i)               (0x00019004 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP6RXPKTSHI_IP6RXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6RXPKTSHI_IP6RXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6RXPKTSHI_IP6RXPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP6RXPKTSLO(_i)               (0x00019000 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP6RXPKTSLO_IP6RXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6RXPKTSLO_IP6RXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXPKTSLO_IP6RXPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP6RXTRUNC(_i)              (0x00019300 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6RXTRUNC_MAX_INDEX        31
+#define I40E_GLPES_VFIP6RXTRUNC_IP6RXTRUNC_SHIFT 0
+#define I40E_GLPES_VFIP6RXTRUNC_IP6RXTRUNC_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6RXTRUNC_IP6RXTRUNC_SHIFT)
+#define I40E_GLPES_VFIP6TXFRAGSHI(_i)                (0x0001A804 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXFRAGSHI_MAX_INDEX          31
+#define I40E_GLPES_VFIP6TXFRAGSHI_IP6TXFRAGSHI_SHIFT 0
+#define I40E_GLPES_VFIP6TXFRAGSHI_IP6TXFRAGSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6TXFRAGSHI_IP6TXFRAGSHI_SHIFT)
+#define I40E_GLPES_VFIP6TXFRAGSLO(_i)                (0x0001A800 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXFRAGSLO_MAX_INDEX          31
+#define I40E_GLPES_VFIP6TXFRAGSLO_IP6TXFRAGSLO_SHIFT 0
+#define I40E_GLPES_VFIP6TXFRAGSLO_IP6TXFRAGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6TXFRAGSLO_IP6TXFRAGSLO_SHIFT)
+#define I40E_GLPES_VFIP6TXMCOCTSHI(_i)                 (0x0001AA04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXMCOCTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP6TXMCOCTSHI_IP6TXMCOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6TXMCOCTSHI_IP6TXMCOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6TXMCOCTSHI_IP6TXMCOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP6TXMCOCTSLO(_i)                 (0x0001AA00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXMCOCTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP6TXMCOCTSLO_IP6TXMCOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6TXMCOCTSLO_IP6TXMCOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6TXMCOCTSLO_IP6TXMCOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP6TXMCPKTSHI(_i)                 (0x0001AC04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXMCPKTSHI_MAX_INDEX           31
+#define I40E_GLPES_VFIP6TXMCPKTSHI_IP6TXMCPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6TXMCPKTSHI_IP6TXMCPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6TXMCPKTSHI_IP6TXMCPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP6TXMCPKTSLO(_i)                 (0x0001AC00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXMCPKTSLO_MAX_INDEX           31
+#define I40E_GLPES_VFIP6TXMCPKTSLO_IP6TXMCPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6TXMCPKTSLO_IP6TXMCPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6TXMCPKTSLO_IP6TXMCPKTSLO_SHIFT)
+#define I40E_GLPES_VFIP6TXNOROUTE(_i)                (0x0001AF00 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXNOROUTE_MAX_INDEX          31
+#define I40E_GLPES_VFIP6TXNOROUTE_IP6TXNOROUTE_SHIFT 0
+#define I40E_GLPES_VFIP6TXNOROUTE_IP6TXNOROUTE_MASK  (0xFFFFFF <<  I40E_GLPES_VFIP6TXNOROUTE_IP6TXNOROUTE_SHIFT)
+#define I40E_GLPES_VFIP6TXOCTSHI(_i)               (0x0001A404 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXOCTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP6TXOCTSHI_IP6TXOCTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6TXOCTSHI_IP6TXOCTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6TXOCTSHI_IP6TXOCTSHI_SHIFT)
+#define I40E_GLPES_VFIP6TXOCTSLO(_i)               (0x0001A400 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXOCTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP6TXOCTSLO_IP6TXOCTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6TXOCTSLO_IP6TXOCTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6TXOCTSLO_IP6TXOCTSLO_SHIFT)
+#define I40E_GLPES_VFIP6TXPKTSHI(_i)               (0x0001A604 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFIP6TXPKTSHI_IP6TXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFIP6TXPKTSHI_IP6TXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFIP6TXPKTSHI_IP6TXPKTSHI_SHIFT)
+#define I40E_GLPES_VFIP6TXPKTSLO(_i)               (0x0001A600 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFIP6TXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFIP6TXPKTSLO_IP6TXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFIP6TXPKTSLO_IP6TXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFIP6TXPKTSLO_IP6TXPKTSLO_SHIFT)
+#define I40E_GLPES_VFRDMARXRDSHI(_i)               (0x0001BE04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXRDSHI_MAX_INDEX         31
+#define I40E_GLPES_VFRDMARXRDSHI_RDMARXRDSHI_SHIFT 0
+#define I40E_GLPES_VFRDMARXRDSHI_RDMARXRDSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMARXRDSHI_RDMARXRDSHI_SHIFT)
+#define I40E_GLPES_VFRDMARXRDSLO(_i)               (0x0001BE00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXRDSLO_MAX_INDEX         31
+#define I40E_GLPES_VFRDMARXRDSLO_RDMARXRDSLO_SHIFT 0
+#define I40E_GLPES_VFRDMARXRDSLO_RDMARXRDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMARXRDSLO_RDMARXRDSLO_SHIFT)
+#define I40E_GLPES_VFRDMARXSNDSHI(_i)                (0x0001C004 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXSNDSHI_MAX_INDEX          31
+#define I40E_GLPES_VFRDMARXSNDSHI_RDMARXSNDSHI_SHIFT 0
+#define I40E_GLPES_VFRDMARXSNDSHI_RDMARXSNDSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMARXSNDSHI_RDMARXSNDSHI_SHIFT)
+#define I40E_GLPES_VFRDMARXSNDSLO(_i)                (0x0001C000 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXSNDSLO_MAX_INDEX          31
+#define I40E_GLPES_VFRDMARXSNDSLO_RDMARXSNDSLO_SHIFT 0
+#define I40E_GLPES_VFRDMARXSNDSLO_RDMARXSNDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMARXSNDSLO_RDMARXSNDSLO_SHIFT)
+#define I40E_GLPES_VFRDMARXWRSHI(_i)               (0x0001BC04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXWRSHI_MAX_INDEX         31
+#define I40E_GLPES_VFRDMARXWRSHI_RDMARXWRSHI_SHIFT 0
+#define I40E_GLPES_VFRDMARXWRSHI_RDMARXWRSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMARXWRSHI_RDMARXWRSHI_SHIFT)
+#define I40E_GLPES_VFRDMARXWRSLO(_i)               (0x0001BC00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMARXWRSLO_MAX_INDEX         31
+#define I40E_GLPES_VFRDMARXWRSLO_RDMARXWRSLO_SHIFT 0
+#define I40E_GLPES_VFRDMARXWRSLO_RDMARXWRSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMARXWRSLO_RDMARXWRSLO_SHIFT)
+#define I40E_GLPES_VFRDMATXRDSHI(_i)               (0x0001C404 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXRDSHI_MAX_INDEX         31
+#define I40E_GLPES_VFRDMATXRDSHI_RDMARXRDSHI_SHIFT 0
+#define I40E_GLPES_VFRDMATXRDSHI_RDMARXRDSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMATXRDSHI_RDMARXRDSHI_SHIFT)
+#define I40E_GLPES_VFRDMATXRDSLO(_i)               (0x0001C400 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXRDSLO_MAX_INDEX         31
+#define I40E_GLPES_VFRDMATXRDSLO_RDMARXRDSLO_SHIFT 0
+#define I40E_GLPES_VFRDMATXRDSLO_RDMARXRDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMATXRDSLO_RDMARXRDSLO_SHIFT)
+#define I40E_GLPES_VFRDMATXSNDSHI(_i)                (0x0001C604 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXSNDSHI_MAX_INDEX          31
+#define I40E_GLPES_VFRDMATXSNDSHI_RDMARXSNDSHI_SHIFT 0
+#define I40E_GLPES_VFRDMATXSNDSHI_RDMARXSNDSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMATXSNDSHI_RDMARXSNDSHI_SHIFT)
+#define I40E_GLPES_VFRDMATXSNDSLO(_i)                (0x0001C600 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXSNDSLO_MAX_INDEX          31
+#define I40E_GLPES_VFRDMATXSNDSLO_RDMARXSNDSLO_SHIFT 0
+#define I40E_GLPES_VFRDMATXSNDSLO_RDMARXSNDSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMATXSNDSLO_RDMARXSNDSLO_SHIFT)
+#define I40E_GLPES_VFRDMATXWRSHI(_i)               (0x0001C204 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXWRSHI_MAX_INDEX         31
+#define I40E_GLPES_VFRDMATXWRSHI_RDMARXWRSHI_SHIFT 0
+#define I40E_GLPES_VFRDMATXWRSHI_RDMARXWRSHI_MASK  (0xFFFF <<  I40E_GLPES_VFRDMATXWRSHI_RDMARXWRSHI_SHIFT)
+#define I40E_GLPES_VFRDMATXWRSLO(_i)               (0x0001C200 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMATXWRSLO_MAX_INDEX         31
+#define I40E_GLPES_VFRDMATXWRSLO_RDMARXWRSLO_SHIFT 0
+#define I40E_GLPES_VFRDMATXWRSLO_RDMARXWRSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMATXWRSLO_RDMARXWRSLO_SHIFT)
+#define I40E_GLPES_VFRDMAVBNDHI(_i)              (0x0001C804 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMAVBNDHI_MAX_INDEX        31
+#define I40E_GLPES_VFRDMAVBNDHI_RDMAVBNDHI_SHIFT 0
+#define I40E_GLPES_VFRDMAVBNDHI_RDMAVBNDHI_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMAVBNDHI_RDMAVBNDHI_SHIFT)
+#define I40E_GLPES_VFRDMAVBNDLO(_i)              (0x0001C800 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMAVBNDLO_MAX_INDEX        31
+#define I40E_GLPES_VFRDMAVBNDLO_RDMAVBNDLO_SHIFT 0
+#define I40E_GLPES_VFRDMAVBNDLO_RDMAVBNDLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMAVBNDLO_RDMAVBNDLO_SHIFT)
+#define I40E_GLPES_VFRDMAVINVHI(_i)              (0x0001CA04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMAVINVHI_MAX_INDEX        31
+#define I40E_GLPES_VFRDMAVINVHI_RDMAVINVHI_SHIFT 0
+#define I40E_GLPES_VFRDMAVINVHI_RDMAVINVHI_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMAVINVHI_RDMAVINVHI_SHIFT)
+#define I40E_GLPES_VFRDMAVINVLO(_i)              (0x0001CA00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRDMAVINVLO_MAX_INDEX        31
+#define I40E_GLPES_VFRDMAVINVLO_RDMAVINVLO_SHIFT 0
+#define I40E_GLPES_VFRDMAVINVLO_RDMAVINVLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFRDMAVINVLO_RDMAVINVLO_SHIFT)
+#define I40E_GLPES_VFRXVLANERR(_i)             (0x00018000 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFRXVLANERR_MAX_INDEX       31
+#define I40E_GLPES_VFRXVLANERR_RXVLANERR_SHIFT 0
+#define I40E_GLPES_VFRXVLANERR_RXVLANERR_MASK  (0xFFFFFF <<  I40E_GLPES_VFRXVLANERR_RXVLANERR_SHIFT)
+#define I40E_GLPES_VFTCPRTXSEG(_i)             (0x0001B600 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPRTXSEG_MAX_INDEX       31
+#define I40E_GLPES_VFTCPRTXSEG_TCPRTXSEG_SHIFT 0
+#define I40E_GLPES_VFTCPRTXSEG_TCPRTXSEG_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFTCPRTXSEG_TCPRTXSEG_SHIFT)
+#define I40E_GLPES_VFTCPRXOPTERR(_i)               (0x0001B200 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPRXOPTERR_MAX_INDEX         31
+#define I40E_GLPES_VFTCPRXOPTERR_TCPRXOPTERR_SHIFT 0
+#define I40E_GLPES_VFTCPRXOPTERR_TCPRXOPTERR_MASK  (0xFFFFFF <<  I40E_GLPES_VFTCPRXOPTERR_TCPRXOPTERR_SHIFT)
+#define I40E_GLPES_VFTCPRXPROTOERR(_i)                 (0x0001B300 + ((_i) * 4)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPRXPROTOERR_MAX_INDEX           31
+#define I40E_GLPES_VFTCPRXPROTOERR_TCPRXPROTOERR_SHIFT 0
+#define I40E_GLPES_VFTCPRXPROTOERR_TCPRXPROTOERR_MASK  (0xFFFFFF <<  I40E_GLPES_VFTCPRXPROTOERR_TCPRXPROTOERR_SHIFT)
+#define I40E_GLPES_VFTCPRXSEGSHI(_i)               (0x0001B004 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPRXSEGSHI_MAX_INDEX         31
+#define I40E_GLPES_VFTCPRXSEGSHI_TCPRXSEGSHI_SHIFT 0
+#define I40E_GLPES_VFTCPRXSEGSHI_TCPRXSEGSHI_MASK  (0xFFFF <<  I40E_GLPES_VFTCPRXSEGSHI_TCPRXSEGSHI_SHIFT)
+#define I40E_GLPES_VFTCPRXSEGSLO(_i)               (0x0001B000 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPRXSEGSLO_MAX_INDEX         31
+#define I40E_GLPES_VFTCPRXSEGSLO_TCPRXSEGSLO_SHIFT 0
+#define I40E_GLPES_VFTCPRXSEGSLO_TCPRXSEGSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFTCPRXSEGSLO_TCPRXSEGSLO_SHIFT)
+#define I40E_GLPES_VFTCPTXSEGHI(_i)              (0x0001B404 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPTXSEGHI_MAX_INDEX        31
+#define I40E_GLPES_VFTCPTXSEGHI_TCPTXSEGHI_SHIFT 0
+#define I40E_GLPES_VFTCPTXSEGHI_TCPTXSEGHI_MASK  (0xFFFF <<  I40E_GLPES_VFTCPTXSEGHI_TCPTXSEGHI_SHIFT)
+#define I40E_GLPES_VFTCPTXSEGLO(_i)              (0x0001B400 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFTCPTXSEGLO_MAX_INDEX        31
+#define I40E_GLPES_VFTCPTXSEGLO_TCPTXSEGLO_SHIFT 0
+#define I40E_GLPES_VFTCPTXSEGLO_TCPTXSEGLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFTCPTXSEGLO_TCPTXSEGLO_SHIFT)
+#define I40E_GLPES_VFUDPRXPKTSHI(_i)               (0x0001B804 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFUDPRXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFUDPRXPKTSHI_UDPRXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFUDPRXPKTSHI_UDPRXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFUDPRXPKTSHI_UDPRXPKTSHI_SHIFT)
+#define I40E_GLPES_VFUDPRXPKTSLO(_i)               (0x0001B800 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFUDPRXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFUDPRXPKTSLO_UDPRXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFUDPRXPKTSLO_UDPRXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFUDPRXPKTSLO_UDPRXPKTSLO_SHIFT)
+#define I40E_GLPES_VFUDPTXPKTSHI(_i)               (0x0001BA04 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFUDPTXPKTSHI_MAX_INDEX         31
+#define I40E_GLPES_VFUDPTXPKTSHI_UDPTXPKTSHI_SHIFT 0
+#define I40E_GLPES_VFUDPTXPKTSHI_UDPTXPKTSHI_MASK  (0xFFFF <<  I40E_GLPES_VFUDPTXPKTSHI_UDPTXPKTSHI_SHIFT)
+#define I40E_GLPES_VFUDPTXPKTSLO(_i)               (0x0001BA00 + ((_i) * 8)) /* _i=0...31 */ /* Reset: PE_CORER */
+#define I40E_GLPES_VFUDPTXPKTSLO_MAX_INDEX         31
+#define I40E_GLPES_VFUDPTXPKTSLO_UDPTXPKTSLO_SHIFT 0
+#define I40E_GLPES_VFUDPTXPKTSLO_UDPTXPKTSLO_MASK  (0xFFFFFFFF <<  I40E_GLPES_VFUDPTXPKTSLO_UDPTXPKTSLO_SHIFT)
+
+#define I40E_VFPE_AEQALLOC1               0x0000A400 /* Reset: VFR */
+#define I40E_VFPE_AEQALLOC1_AECOUNT_SHIFT 0
+#define I40E_VFPE_AEQALLOC1_AECOUNT_MASK  (0xFFFFFFFF <<  I40E_VFPE_AEQALLOC1_AECOUNT_SHIFT)
+#define I40E_VFPE_CCQPHIGH1                  0x00009800 /* Reset: VFR */
+#define I40E_VFPE_CCQPHIGH1_PECCQPHIGH_SHIFT 0
+#define I40E_VFPE_CCQPHIGH1_PECCQPHIGH_MASK  (0xFFFFFFFF <<  I40E_VFPE_CCQPHIGH1_PECCQPHIGH_SHIFT)
+#define I40E_VFPE_CCQPLOW1                 0x0000AC00 /* Reset: VFR */
+#define I40E_VFPE_CCQPLOW1_PECCQPLOW_SHIFT 0
+#define I40E_VFPE_CCQPLOW1_PECCQPLOW_MASK  (0xFFFFFFFF <<  I40E_VFPE_CCQPLOW1_PECCQPLOW_SHIFT)
+#define I40E_VFPE_CCQPSTATUS1                   0x0000B800 /* Reset: VFR */
+#define I40E_VFPE_CCQPSTATUS1_CCQP_DONE_SHIFT   0
+#define I40E_VFPE_CCQPSTATUS1_CCQP_DONE_MASK    (0x1 <<  I40E_VFPE_CCQPSTATUS1_CCQP_DONE_SHIFT)
+#define I40E_VFPE_CCQPSTATUS1_HMC_PROFILE_SHIFT 4
+#define I40E_VFPE_CCQPSTATUS1_HMC_PROFILE_MASK  (0x7 <<  I40E_VFPE_CCQPSTATUS1_HMC_PROFILE_SHIFT)
+#define I40E_VFPE_CCQPSTATUS1_RDMA_EN_VFS_SHIFT 16
+#define I40E_VFPE_CCQPSTATUS1_RDMA_EN_VFS_MASK  (0x3F <<  I40E_VFPE_CCQPSTATUS1_RDMA_EN_VFS_SHIFT)
+#define I40E_VFPE_CCQPSTATUS1_CCQP_ERR_SHIFT    31
+#define I40E_VFPE_CCQPSTATUS1_CCQP_ERR_MASK     (0x1 <<  I40E_VFPE_CCQPSTATUS1_CCQP_ERR_SHIFT)
+#define I40E_VFPE_CQACK1              0x0000B000 /* Reset: VFR */
+#define I40E_VFPE_CQACK1_PECQID_SHIFT 0
+#define I40E_VFPE_CQACK1_PECQID_MASK  (0x1FFFF <<  I40E_VFPE_CQACK1_PECQID_SHIFT)
+#define I40E_VFPE_CQARM1              0x0000B400 /* Reset: VFR */
+#define I40E_VFPE_CQARM1_PECQID_SHIFT 0
+#define I40E_VFPE_CQARM1_PECQID_MASK  (0x1FFFF <<  I40E_VFPE_CQARM1_PECQID_SHIFT)
+#define I40E_VFPE_CQPDB1              0x0000BC00 /* Reset: VFR */
+#define I40E_VFPE_CQPDB1_WQHEAD_SHIFT 0
+#define I40E_VFPE_CQPDB1_WQHEAD_MASK  (0x7FF <<  I40E_VFPE_CQPDB1_WQHEAD_SHIFT)
+#define I40E_VFPE_CQPERRCODES1                      0x00009C00 /* Reset: VFR */
+#define I40E_VFPE_CQPERRCODES1_CQP_MINOR_CODE_SHIFT 0
+#define I40E_VFPE_CQPERRCODES1_CQP_MINOR_CODE_MASK  (0xFFFF <<  I40E_VFPE_CQPERRCODES1_CQP_MINOR_CODE_SHIFT)
+#define I40E_VFPE_CQPERRCODES1_CQP_MAJOR_CODE_SHIFT 16
+#define I40E_VFPE_CQPERRCODES1_CQP_MAJOR_CODE_MASK  (0xFFFF <<  I40E_VFPE_CQPERRCODES1_CQP_MAJOR_CODE_SHIFT)
+#define I40E_VFPE_CQPTAIL1                  0x0000A000 /* Reset: VFR */
+#define I40E_VFPE_CQPTAIL1_WQTAIL_SHIFT     0
+#define I40E_VFPE_CQPTAIL1_WQTAIL_MASK      (0x7FF <<  I40E_VFPE_CQPTAIL1_WQTAIL_SHIFT)
+#define I40E_VFPE_CQPTAIL1_CQP_OP_ERR_SHIFT 31
+#define I40E_VFPE_CQPTAIL1_CQP_OP_ERR_MASK  (0x1 <<  I40E_VFPE_CQPTAIL1_CQP_OP_ERR_SHIFT)
+#define I40E_VFPE_IPCONFIG01                        0x00008C00 /* Reset: VFR */
+#define I40E_VFPE_IPCONFIG01_PEIPID_SHIFT           0
+#define I40E_VFPE_IPCONFIG01_PEIPID_MASK            (0xFFFF <<  I40E_VFPE_IPCONFIG01_PEIPID_SHIFT)
+#define I40E_VFPE_IPCONFIG01_USEENTIREIDRANGE_SHIFT 16
+#define I40E_VFPE_IPCONFIG01_USEENTIREIDRANGE_MASK  (0x1 <<  I40E_VFPE_IPCONFIG01_USEENTIREIDRANGE_SHIFT)
+#define I40E_VFPE_MRTEIDXMASK1                       0x00009000 /* Reset: VFR */
+#define I40E_VFPE_MRTEIDXMASK1_MRTEIDXMASKBITS_SHIFT 0
+#define I40E_VFPE_MRTEIDXMASK1_MRTEIDXMASKBITS_MASK  (0x1F <<  I40E_VFPE_MRTEIDXMASK1_MRTEIDXMASKBITS_SHIFT)
+#define I40E_VFPE_RCVUNEXPECTEDERROR1                        0x00009400 /* Reset: VFR */
+#define I40E_VFPE_RCVUNEXPECTEDERROR1_TCP_RX_UNEXP_ERR_SHIFT 0
+#define I40E_VFPE_RCVUNEXPECTEDERROR1_TCP_RX_UNEXP_ERR_MASK  (0xFFFFFF <<  I40E_VFPE_RCVUNEXPECTEDERROR1_TCP_RX_UNEXP_ERR_SHIFT)
+#define I40E_VFPE_TCPNOWTIMER1               0x0000A800 /* Reset: VFR */
+#define I40E_VFPE_TCPNOWTIMER1_TCP_NOW_SHIFT 0
+#define I40E_VFPE_TCPNOWTIMER1_TCP_NOW_MASK  (0xFFFFFFFF <<  I40E_VFPE_TCPNOWTIMER1_TCP_NOW_SHIFT)
+#define I40E_VFPE_WQEALLOC1                      0x0000C000 /* Reset: VFR */
+#define I40E_VFPE_WQEALLOC1_PEQPID_SHIFT         0
+#define I40E_VFPE_WQEALLOC1_PEQPID_MASK          (0x3FFFF <<  I40E_VFPE_WQEALLOC1_PEQPID_SHIFT)
+#define I40E_VFPE_WQEALLOC1_WQE_DESC_INDEX_SHIFT 20
+#define I40E_VFPE_WQEALLOC1_WQE_DESC_INDEX_MASK  (0xFFF <<  I40E_VFPE_WQEALLOC1_WQE_DESC_INDEX_SHIFT)
+#endif /* I40IW_REGISTER_H */
diff --git a/drivers/infiniband/hw/i40iw/i40iw_status.h b/drivers/infiniband/hw/i40iw/i40iw_status.h
new file mode 100644
index 000000000000..b0110c15e044
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_status.h
@@ -0,0 +1,100 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_STATUS_H
+#define I40IW_STATUS_H
+
+/* Error Codes */
+enum i40iw_status_code {
+	I40IW_SUCCESS = 0,
+	I40IW_ERR_NVM = -1,
+	I40IW_ERR_NVM_CHECKSUM = -2,
+	I40IW_ERR_CONFIG = -4,
+	I40IW_ERR_PARAM = -5,
+	I40IW_ERR_DEVICE_NOT_SUPPORTED = -6,
+	I40IW_ERR_RESET_FAILED = -7,
+	I40IW_ERR_SWFW_SYNC = -8,
+	I40IW_ERR_NO_MEMORY = -9,
+	I40IW_ERR_BAD_PTR = -10,
+	I40IW_ERR_INVALID_PD_ID = -11,
+	I40IW_ERR_INVALID_QP_ID = -12,
+	I40IW_ERR_INVALID_CQ_ID = -13,
+	I40IW_ERR_INVALID_CEQ_ID = -14,
+	I40IW_ERR_INVALID_AEQ_ID = -15,
+	I40IW_ERR_INVALID_SIZE = -16,
+	I40IW_ERR_INVALID_ARP_INDEX = -17,
+	I40IW_ERR_INVALID_FPM_FUNC_ID = -18,
+	I40IW_ERR_QP_INVALID_MSG_SIZE = -19,
+	I40IW_ERR_QP_TOOMANY_WRS_POSTED = -20,
+	I40IW_ERR_INVALID_FRAG_COUNT = -21,
+	I40IW_ERR_QUEUE_EMPTY = -22,
+	I40IW_ERR_INVALID_ALIGNMENT = -23,
+	I40IW_ERR_FLUSHED_QUEUE = -24,
+	I40IW_ERR_INVALID_PUSH_PAGE_INDEX = -25,
+	I40IW_ERR_INVALID_IMM_DATA_SIZE = -26,
+	I40IW_ERR_TIMEOUT = -27,
+	I40IW_ERR_OPCODE_MISMATCH = -28,
+	I40IW_ERR_CQP_COMPL_ERROR = -29,
+	I40IW_ERR_INVALID_VF_ID = -30,
+	I40IW_ERR_INVALID_HMCFN_ID = -31,
+	I40IW_ERR_BACKING_PAGE_ERROR = -32,
+	I40IW_ERR_NO_PBLCHUNKS_AVAILABLE = -33,
+	I40IW_ERR_INVALID_PBLE_INDEX = -34,
+	I40IW_ERR_INVALID_SD_INDEX = -35,
+	I40IW_ERR_INVALID_PAGE_DESC_INDEX = -36,
+	I40IW_ERR_INVALID_SD_TYPE = -37,
+	I40IW_ERR_MEMCPY_FAILED = -38,
+	I40IW_ERR_INVALID_HMC_OBJ_INDEX = -39,
+	I40IW_ERR_INVALID_HMC_OBJ_COUNT = -40,
+	I40IW_ERR_INVALID_SRQ_ARM_LIMIT = -41,
+	I40IW_ERR_SRQ_ENABLED = -42,
+	I40IW_ERR_BUF_TOO_SHORT = -43,
+	I40IW_ERR_BAD_IWARP_CQE = -44,
+	I40IW_ERR_NVM_BLANK_MODE = -45,
+	I40IW_ERR_NOT_IMPLEMENTED = -46,
+	I40IW_ERR_PE_DOORBELL_NOT_ENABLED = -47,
+	I40IW_ERR_NOT_READY = -48,
+	I40IW_NOT_SUPPORTED = -49,
+	I40IW_ERR_FIRMWARE_API_VERSION = -50,
+	I40IW_ERR_RING_FULL = -51,
+	I40IW_ERR_MPA_CRC = -61,
+	I40IW_ERR_NO_TXBUFS = -62,
+	I40IW_ERR_SEQ_NUM = -63,
+	I40IW_ERR_list_empty = -64,
+	I40IW_ERR_INVALID_MAC_ADDR = -65,
+	I40IW_ERR_BAD_STAG      = -66,
+	I40IW_ERR_CQ_COMPL_ERROR = -67,
+
+};
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_type.h b/drivers/infiniband/hw/i40iw/i40iw_type.h
new file mode 100644
index 000000000000..edb3a8c8267a
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_type.h
@@ -0,0 +1,1312 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_TYPE_H
+#define I40IW_TYPE_H
+#include "i40iw_user.h"
+#include "i40iw_hmc.h"
+#include "i40iw_vf.h"
+#include "i40iw_virtchnl.h"
+
+struct i40iw_cqp_sq_wqe {
+	u64 buf[I40IW_CQP_WQE_SIZE];
+};
+
+struct i40iw_sc_aeqe {
+	u64 buf[I40IW_AEQE_SIZE];
+};
+
+struct i40iw_ceqe {
+	u64 buf[I40IW_CEQE_SIZE];
+};
+
+struct i40iw_cqp_ctx {
+	u64 buf[I40IW_CQP_CTX_SIZE];
+};
+
+struct i40iw_cq_shadow_area {
+	u64 buf[I40IW_SHADOW_AREA_SIZE];
+};
+
+struct i40iw_sc_dev;
+struct i40iw_hmc_info;
+struct i40iw_dev_pestat;
+
+struct i40iw_cqp_ops;
+struct i40iw_ccq_ops;
+struct i40iw_ceq_ops;
+struct i40iw_aeq_ops;
+struct i40iw_mr_ops;
+struct i40iw_cqp_misc_ops;
+struct i40iw_pd_ops;
+struct i40iw_priv_qp_ops;
+struct i40iw_priv_cq_ops;
+struct i40iw_hmc_ops;
+
+enum i40iw_resource_indicator_type {
+	I40IW_RSRC_INDICATOR_TYPE_ADAPTER = 0,
+	I40IW_RSRC_INDICATOR_TYPE_CQ,
+	I40IW_RSRC_INDICATOR_TYPE_QP,
+	I40IW_RSRC_INDICATOR_TYPE_SRQ
+};
+
+enum i40iw_hdrct_flags {
+	DDP_LEN_FLAG = 0x80,
+	DDP_HDR_FLAG = 0x40,
+	RDMA_HDR_FLAG = 0x20
+};
+
+enum i40iw_term_layers {
+	LAYER_RDMA = 0,
+	LAYER_DDP = 1,
+	LAYER_MPA = 2
+};
+
+enum i40iw_term_error_types {
+	RDMAP_REMOTE_PROT = 1,
+	RDMAP_REMOTE_OP = 2,
+	DDP_CATASTROPHIC = 0,
+	DDP_TAGGED_BUFFER = 1,
+	DDP_UNTAGGED_BUFFER = 2,
+	DDP_LLP = 3
+};
+
+enum i40iw_term_rdma_errors {
+	RDMAP_INV_STAG = 0x00,
+	RDMAP_INV_BOUNDS = 0x01,
+	RDMAP_ACCESS = 0x02,
+	RDMAP_UNASSOC_STAG = 0x03,
+	RDMAP_TO_WRAP = 0x04,
+	RDMAP_INV_RDMAP_VER = 0x05,
+	RDMAP_UNEXPECTED_OP = 0x06,
+	RDMAP_CATASTROPHIC_LOCAL = 0x07,
+	RDMAP_CATASTROPHIC_GLOBAL = 0x08,
+	RDMAP_CANT_INV_STAG = 0x09,
+	RDMAP_UNSPECIFIED = 0xff
+};
+
+enum i40iw_term_ddp_errors {
+	DDP_CATASTROPHIC_LOCAL = 0x00,
+	DDP_TAGGED_INV_STAG = 0x00,
+	DDP_TAGGED_BOUNDS = 0x01,
+	DDP_TAGGED_UNASSOC_STAG = 0x02,
+	DDP_TAGGED_TO_WRAP = 0x03,
+	DDP_TAGGED_INV_DDP_VER = 0x04,
+	DDP_UNTAGGED_INV_QN = 0x01,
+	DDP_UNTAGGED_INV_MSN_NO_BUF = 0x02,
+	DDP_UNTAGGED_INV_MSN_RANGE = 0x03,
+	DDP_UNTAGGED_INV_MO = 0x04,
+	DDP_UNTAGGED_INV_TOO_LONG = 0x05,
+	DDP_UNTAGGED_INV_DDP_VER = 0x06
+};
+
+enum i40iw_term_mpa_errors {
+	MPA_CLOSED = 0x01,
+	MPA_CRC = 0x02,
+	MPA_MARKER = 0x03,
+	MPA_REQ_RSP = 0x04,
+};
+
+enum i40iw_flush_opcode {
+	FLUSH_INVALID = 0,
+	FLUSH_PROT_ERR,
+	FLUSH_REM_ACCESS_ERR,
+	FLUSH_LOC_QP_OP_ERR,
+	FLUSH_REM_OP_ERR,
+	FLUSH_LOC_LEN_ERR,
+	FLUSH_GENERAL_ERR,
+	FLUSH_FATAL_ERR
+};
+
+enum i40iw_term_eventtypes {
+	TERM_EVENT_QP_FATAL,
+	TERM_EVENT_QP_ACCESS_ERR
+};
+
+struct i40iw_terminate_hdr {
+	u8 layer_etype;
+	u8 error_code;
+	u8 hdrct;
+	u8 rsvd;
+};
+
+enum i40iw_debug_flag {
+	I40IW_DEBUG_NONE	= 0x00000000,
+	I40IW_DEBUG_ERR		= 0x00000001,
+	I40IW_DEBUG_INIT	= 0x00000002,
+	I40IW_DEBUG_DEV		= 0x00000004,
+	I40IW_DEBUG_CM		= 0x00000008,
+	I40IW_DEBUG_VERBS	= 0x00000010,
+	I40IW_DEBUG_PUDA	= 0x00000020,
+	I40IW_DEBUG_ILQ		= 0x00000040,
+	I40IW_DEBUG_IEQ		= 0x00000080,
+	I40IW_DEBUG_QP		= 0x00000100,
+	I40IW_DEBUG_CQ		= 0x00000200,
+	I40IW_DEBUG_MR		= 0x00000400,
+	I40IW_DEBUG_PBLE	= 0x00000800,
+	I40IW_DEBUG_WQE		= 0x00001000,
+	I40IW_DEBUG_AEQ		= 0x00002000,
+	I40IW_DEBUG_CQP		= 0x00004000,
+	I40IW_DEBUG_HMC		= 0x00008000,
+	I40IW_DEBUG_USER	= 0x00010000,
+	I40IW_DEBUG_VIRT	= 0x00020000,
+	I40IW_DEBUG_DCB		= 0x00040000,
+	I40IW_DEBUG_CQE		= 0x00800000,
+	I40IW_DEBUG_ALL		= 0xFFFFFFFF
+};
+
+enum i40iw_hw_stat_index_32b {
+	I40IW_HW_STAT_INDEX_IP4RXDISCARD = 0,
+	I40IW_HW_STAT_INDEX_IP4RXTRUNC,
+	I40IW_HW_STAT_INDEX_IP4TXNOROUTE,
+	I40IW_HW_STAT_INDEX_IP6RXDISCARD,
+	I40IW_HW_STAT_INDEX_IP6RXTRUNC,
+	I40IW_HW_STAT_INDEX_IP6TXNOROUTE,
+	I40IW_HW_STAT_INDEX_TCPRTXSEG,
+	I40IW_HW_STAT_INDEX_TCPRXOPTERR,
+	I40IW_HW_STAT_INDEX_TCPRXPROTOERR,
+	I40IW_HW_STAT_INDEX_MAX_32
+};
+
+enum i40iw_hw_stat_index_64b {
+	I40IW_HW_STAT_INDEX_IP4RXOCTS = 0,
+	I40IW_HW_STAT_INDEX_IP4RXPKTS,
+	I40IW_HW_STAT_INDEX_IP4RXFRAGS,
+	I40IW_HW_STAT_INDEX_IP4RXMCPKTS,
+	I40IW_HW_STAT_INDEX_IP4TXOCTS,
+	I40IW_HW_STAT_INDEX_IP4TXPKTS,
+	I40IW_HW_STAT_INDEX_IP4TXFRAGS,
+	I40IW_HW_STAT_INDEX_IP4TXMCPKTS,
+	I40IW_HW_STAT_INDEX_IP6RXOCTS,
+	I40IW_HW_STAT_INDEX_IP6RXPKTS,
+	I40IW_HW_STAT_INDEX_IP6RXFRAGS,
+	I40IW_HW_STAT_INDEX_IP6RXMCPKTS,
+	I40IW_HW_STAT_INDEX_IP6TXOCTS,
+	I40IW_HW_STAT_INDEX_IP6TXPKTS,
+	I40IW_HW_STAT_INDEX_IP6TXFRAGS,
+	I40IW_HW_STAT_INDEX_IP6TXMCPKTS,
+	I40IW_HW_STAT_INDEX_TCPRXSEGS,
+	I40IW_HW_STAT_INDEX_TCPTXSEG,
+	I40IW_HW_STAT_INDEX_RDMARXRDS,
+	I40IW_HW_STAT_INDEX_RDMARXSNDS,
+	I40IW_HW_STAT_INDEX_RDMARXWRS,
+	I40IW_HW_STAT_INDEX_RDMATXRDS,
+	I40IW_HW_STAT_INDEX_RDMATXSNDS,
+	I40IW_HW_STAT_INDEX_RDMATXWRS,
+	I40IW_HW_STAT_INDEX_RDMAVBND,
+	I40IW_HW_STAT_INDEX_RDMAVINV,
+	I40IW_HW_STAT_INDEX_MAX_64
+};
+
+struct i40iw_dev_hw_stat_offsets {
+	u32 stat_offset_32[I40IW_HW_STAT_INDEX_MAX_32];
+	u32 stat_offset_64[I40IW_HW_STAT_INDEX_MAX_64];
+};
+
+struct i40iw_dev_hw_stats {
+	u64 stat_value_32[I40IW_HW_STAT_INDEX_MAX_32];
+	u64 stat_value_64[I40IW_HW_STAT_INDEX_MAX_64];
+};
+
+struct i40iw_device_pestat_ops {
+	void (*iw_hw_stat_init)(struct i40iw_dev_pestat *, u8, struct i40iw_hw *, bool);
+	void (*iw_hw_stat_read_32)(struct i40iw_dev_pestat *, enum i40iw_hw_stat_index_32b, u64 *);
+	void (*iw_hw_stat_read_64)(struct i40iw_dev_pestat *, enum i40iw_hw_stat_index_64b, u64 *);
+	void (*iw_hw_stat_read_all)(struct i40iw_dev_pestat *, struct i40iw_dev_hw_stats *);
+	void (*iw_hw_stat_refresh_all)(struct i40iw_dev_pestat *);
+};
+
+struct i40iw_dev_pestat {
+	struct i40iw_hw *hw;
+	struct i40iw_device_pestat_ops ops;
+	struct i40iw_dev_hw_stats hw_stats;
+	struct i40iw_dev_hw_stats last_read_hw_stats;
+	struct i40iw_dev_hw_stat_offsets hw_stat_offsets;
+	struct timer_list stats_timer;
+	spinlock_t stats_lock; /* rdma stats lock */
+};
+
+struct i40iw_hw {
+	u8 __iomem *hw_addr;
+	void *dev_context;
+	struct i40iw_hmc_info hmc;
+};
+
+struct i40iw_pfpdu {
+	struct list_head rxlist;
+	u32 rcv_nxt;
+	u32 fps;
+	u32 max_fpdu_data;
+	bool mode;
+	bool mpa_crc_err;
+	u64 total_ieq_bufs;
+	u64 fpdu_processed;
+	u64 bad_seq_num;
+	u64 crc_err;
+	u64 no_tx_bufs;
+	u64 tx_err;
+	u64 out_of_order;
+	u64 pmode_count;
+};
+
+struct i40iw_sc_pd {
+	u32 size;
+	struct i40iw_sc_dev *dev;
+	u16 pd_id;
+};
+
+struct i40iw_cqp_quanta {
+	u64 elem[I40IW_CQP_WQE_SIZE];
+};
+
+struct i40iw_sc_cqp {
+	u32 size;
+	u64 sq_pa;
+	u64 host_ctx_pa;
+	void *back_cqp;
+	struct i40iw_sc_dev *dev;
+	enum i40iw_status_code (*process_cqp_sds)(struct i40iw_sc_dev *,
+						  struct i40iw_update_sds_info *);
+	struct i40iw_dma_mem sdbuf;
+	struct i40iw_ring sq_ring;
+	struct i40iw_cqp_quanta *sq_base;
+	u64 *host_ctx;
+	u64 *scratch_array;
+	u32 cqp_id;
+	u32 sq_size;
+	u32 hw_sq_size;
+	u8 struct_ver;
+	u8 polarity;
+	bool en_datacenter_tcp;
+	u8 hmc_profile;
+	u8 enabled_vf_count;
+	u8 timeout_count;
+};
+
+struct i40iw_sc_aeq {
+	u32 size;
+	u64 aeq_elem_pa;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_sc_aeqe *aeqe_base;
+	void *pbl_list;
+	u32 elem_cnt;
+	struct i40iw_ring aeq_ring;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	u32 first_pm_pbl_idx;
+	u8 polarity;
+};
+
+struct i40iw_sc_ceq {
+	u32 size;
+	u64 ceq_elem_pa;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_ceqe *ceqe_base;
+	void *pbl_list;
+	u32 ceq_id;
+	u32 elem_cnt;
+	struct i40iw_ring ceq_ring;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	bool tph_en;
+	u8 tph_val;
+	u32 first_pm_pbl_idx;
+	u8 polarity;
+};
+
+struct i40iw_sc_cq {
+	struct i40iw_cq_uk cq_uk;
+	u64 cq_pa;
+	u64 shadow_area_pa;
+	struct i40iw_sc_dev *dev;
+	void *pbl_list;
+	void *back_cq;
+	u32 ceq_id;
+	u32 shadow_read_threshold;
+	bool ceqe_mask;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	u8 cq_type;
+	bool ceq_id_valid;
+	bool tph_en;
+	u8 tph_val;
+	u32 first_pm_pbl_idx;
+	bool check_overflow;
+};
+
+struct i40iw_sc_qp {
+	struct i40iw_qp_uk qp_uk;
+	u64 sq_pa;
+	u64 rq_pa;
+	u64 hw_host_ctx_pa;
+	u64 shadow_area_pa;
+	u64 q2_pa;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_sc_pd *pd;
+	u64 *hw_host_ctx;
+	void *llp_stream_handle;
+	void *back_qp;
+	struct i40iw_pfpdu pfpdu;
+	u8 *q2_buf;
+	u64 qp_compl_ctx;
+	u16 qs_handle;
+	u16 exception_lan_queue;
+	u16 push_idx;
+	u8 sq_tph_val;
+	u8 rq_tph_val;
+	u8 qp_state;
+	u8 qp_type;
+	u8 hw_sq_size;
+	u8 hw_rq_size;
+	u8 src_mac_addr_idx;
+	bool sq_tph_en;
+	bool rq_tph_en;
+	bool rcv_tph_en;
+	bool xmit_tph_en;
+	bool virtual_map;
+	bool flush_sq;
+	bool flush_rq;
+	bool sq_flush;
+	enum i40iw_flush_opcode flush_code;
+	enum i40iw_term_eventtypes eventtype;
+	u8 term_flags;
+};
+
+struct i40iw_hmc_fpm_misc {
+	u32 max_ceqs;
+	u32 max_sds;
+	u32 xf_block_size;
+	u32 q1_block_size;
+	u32 ht_multiplier;
+	u32 timer_bucket;
+};
+
+struct i40iw_vchnl_if {
+	enum i40iw_status_code (*vchnl_recv)(struct i40iw_sc_dev *, u32, u8 *, u16);
+	enum i40iw_status_code (*vchnl_send)(struct i40iw_sc_dev *dev, u32, u8 *, u16);
+};
+
+#define I40IW_VCHNL_MAX_VF_MSG_SIZE 512
+
+struct i40iw_vchnl_vf_msg_buffer {
+	struct i40iw_virtchnl_op_buf vchnl_msg;
+	char parm_buffer[I40IW_VCHNL_MAX_VF_MSG_SIZE - 1];
+};
+
+struct i40iw_vfdev {
+	struct i40iw_sc_dev *pf_dev;
+	u8 *hmc_info_mem;
+	struct i40iw_dev_pestat dev_pestat;
+	struct i40iw_hmc_pble_info *pble_info;
+	struct i40iw_hmc_info hmc_info;
+	struct i40iw_vchnl_vf_msg_buffer vf_msg_buffer;
+	u64 fpm_query_buf_pa;
+	u64 *fpm_query_buf;
+	u32 vf_id;
+	u32 msg_count;
+	bool pf_hmc_initialized;
+	u16 pmf_index;
+	u16 iw_vf_idx;		/* VF Device table index */
+	bool stats_initialized;
+};
+
+struct i40iw_sc_dev {
+	struct list_head cqp_cmd_head;	/* head of the CQP command list */
+	spinlock_t cqp_lock; /* cqp list sync */
+	struct i40iw_dev_uk dev_uk;
+	struct i40iw_dev_pestat dev_pestat;
+	struct i40iw_dma_mem vf_fpm_query_buf[I40IW_MAX_PE_ENABLED_VF_COUNT];
+	u64 fpm_query_buf_pa;
+	u64 fpm_commit_buf_pa;
+	u64 *fpm_query_buf;
+	u64 *fpm_commit_buf;
+	void *back_dev;
+	struct i40iw_hw *hw;
+	u8 __iomem *db_addr;
+	struct i40iw_hmc_info *hmc_info;
+	struct i40iw_hmc_pble_info *pble_info;
+	struct i40iw_vfdev *vf_dev[I40IW_MAX_PE_ENABLED_VF_COUNT];
+	struct i40iw_sc_cqp *cqp;
+	struct i40iw_sc_aeq *aeq;
+	struct i40iw_sc_ceq *ceq[I40IW_CEQ_MAX_COUNT];
+	struct i40iw_sc_cq *ccq;
+	struct i40iw_cqp_ops *cqp_ops;
+	struct i40iw_ccq_ops *ccq_ops;
+	struct i40iw_ceq_ops *ceq_ops;
+	struct i40iw_aeq_ops *aeq_ops;
+	struct i40iw_pd_ops *iw_pd_ops;
+	struct i40iw_priv_qp_ops *iw_priv_qp_ops;
+	struct i40iw_priv_cq_ops *iw_priv_cq_ops;
+	struct i40iw_mr_ops *mr_ops;
+	struct i40iw_cqp_misc_ops *cqp_misc_ops;
+	struct i40iw_hmc_ops *hmc_ops;
+	struct i40iw_vchnl_if vchnl_if;
+	u32 ilq_count;
+	struct i40iw_virt_mem ilq_mem;
+	struct i40iw_puda_rsrc *ilq;
+	u32 ieq_count;
+	struct i40iw_virt_mem ieq_mem;
+	struct i40iw_puda_rsrc *ieq;
+
+	struct i40iw_vf_cqp_ops *iw_vf_cqp_ops;
+
+	struct i40iw_hmc_fpm_misc hmc_fpm_misc;
+	u16 qs_handle;
+	u32	debug_mask;
+	u16 exception_lan_queue;
+	u8 hmc_fn_id;
+	bool is_pf;
+	bool vchnl_up;
+	u8 vf_id;
+	u64 cqp_cmd_stats[OP_SIZE_CQP_STAT_ARRAY];
+	struct i40iw_vchnl_vf_msg_buffer vchnl_vf_msg_buf;
+	u8 hw_rev;
+};
+
+struct i40iw_modify_cq_info {
+	u64 cq_pa;
+	struct i40iw_cqe *cq_base;
+	void *pbl_list;
+	u32 ceq_id;
+	u32 cq_size;
+	u32 shadow_read_threshold;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	bool check_overflow;
+	bool cq_resize;
+	bool ceq_change;
+	bool check_overflow_change;
+	u32 first_pm_pbl_idx;
+	bool ceq_valid;
+};
+
+struct i40iw_create_qp_info {
+	u8 next_iwarp_state;
+	bool ord_valid;
+	bool tcp_ctx_valid;
+	bool cq_num_valid;
+	bool static_rsrc;
+	bool arp_cache_idx_valid;
+};
+
+struct i40iw_modify_qp_info {
+	u64 rx_win0;
+	u64 rx_win1;
+	u16 new_mss;
+	u8 next_iwarp_state;
+	u8 termlen;
+	bool ord_valid;
+	bool tcp_ctx_valid;
+	bool cq_num_valid;
+	bool static_rsrc;
+	bool arp_cache_idx_valid;
+	bool reset_tcp_conn;
+	bool remove_hash_idx;
+	bool dont_send_term;
+	bool dont_send_fin;
+	bool cached_var_valid;
+	bool mss_change;
+	bool force_loopback;
+};
+
+struct i40iw_ccq_cqe_info {
+	struct i40iw_sc_cqp *cqp;
+	u64 scratch;
+	u32 op_ret_val;
+	u16 maj_err_code;
+	u16 min_err_code;
+	u8 op_code;
+	bool error;
+};
+
+struct i40iw_l2params {
+	u16 qs_handle_list[I40IW_MAX_USER_PRIORITY];
+	u16 mss;
+};
+
+struct i40iw_device_init_info {
+	u64 fpm_query_buf_pa;
+	u64 fpm_commit_buf_pa;
+	u64 *fpm_query_buf;
+	u64 *fpm_commit_buf;
+	struct i40iw_hw *hw;
+	void __iomem *bar0;
+	enum i40iw_status_code (*vchnl_send)(struct i40iw_sc_dev *, u32, u8 *, u16);
+	u16 qs_handle;
+	u16 exception_lan_queue;
+	u8 hmc_fn_id;
+	bool is_pf;
+	u32 debug_mask;
+};
+
+enum i40iw_cqp_hmc_profile {
+	I40IW_HMC_PROFILE_DEFAULT = 1,
+	I40IW_HMC_PROFILE_FAVOR_VF = 2,
+	I40IW_HMC_PROFILE_EQUAL = 3,
+};
+
+struct i40iw_cqp_init_info {
+	u64 cqp_compl_ctx;
+	u64 host_ctx_pa;
+	u64 sq_pa;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_cqp_quanta *sq;
+	u64 *host_ctx;
+	u64 *scratch_array;
+	u32 sq_size;
+	u8 struct_ver;
+	bool en_datacenter_tcp;
+	u8 hmc_profile;
+	u8 enabled_vf_count;
+};
+
+struct i40iw_ceq_init_info {
+	u64 ceqe_pa;
+	struct i40iw_sc_dev *dev;
+	u64 *ceqe_base;
+	void *pbl_list;
+	u32 elem_cnt;
+	u32 ceq_id;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	bool tph_en;
+	u8 tph_val;
+	u32 first_pm_pbl_idx;
+};
+
+struct i40iw_aeq_init_info {
+	u64 aeq_elem_pa;
+	struct i40iw_sc_dev *dev;
+	u32 *aeqe_base;
+	void *pbl_list;
+	u32 elem_cnt;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	u32 first_pm_pbl_idx;
+};
+
+struct i40iw_ccq_init_info {
+	u64 cq_pa;
+	u64 shadow_area_pa;
+	struct i40iw_sc_dev *dev;
+	struct i40iw_cqe *cq_base;
+	u64 *shadow_area;
+	void *pbl_list;
+	u32 num_elem;
+	u32 ceq_id;
+	u32 shadow_read_threshold;
+	bool ceqe_mask;
+	bool ceq_id_valid;
+	bool tph_en;
+	u8 tph_val;
+	bool avoid_mem_cflct;
+	bool virtual_map;
+	u8 pbl_chunk_size;
+	u32 first_pm_pbl_idx;
+};
+
+struct i40iwarp_offload_info {
+	u16 rcv_mark_offset;
+	u16 snd_mark_offset;
+	u16 pd_id;
+	u8 ddp_ver;
+	u8 rdmap_ver;
+	u8 ord_size;
+	u8 ird_size;
+	bool wr_rdresp_en;
+	bool rd_enable;
+	bool snd_mark_en;
+	bool rcv_mark_en;
+	bool bind_en;
+	bool fast_reg_en;
+	bool priv_mode_en;
+	bool lsmm_present;
+	u8 iwarp_mode;
+	bool align_hdrs;
+	bool rcv_no_mpa_crc;
+
+	u8 last_byte_sent;
+};
+
+struct i40iw_tcp_offload_info {
+	bool ipv4;
+	bool no_nagle;
+	bool insert_vlan_tag;
+	bool time_stamp;
+	u8 cwnd_inc_limit;
+	bool drop_ooo_seg;
+	bool dup_ack_thresh;
+	u8 ttl;
+	u8 src_mac_addr_idx;
+	bool avoid_stretch_ack;
+	u8 tos;
+	u16 src_port;
+	u16 dst_port;
+	u32 dest_ip_addr0;
+	u32 dest_ip_addr1;
+	u32 dest_ip_addr2;
+	u32 dest_ip_addr3;
+	u32 snd_mss;
+	u16 vlan_tag;
+	u16 arp_idx;
+	u32 flow_label;
+	bool wscale;
+	u8 tcp_state;
+	u8 snd_wscale;
+	u8 rcv_wscale;
+	u32 time_stamp_recent;
+	u32 time_stamp_age;
+	u32 snd_nxt;
+	u32 snd_wnd;
+	u32 rcv_nxt;
+	u32 rcv_wnd;
+	u32 snd_max;
+	u32 snd_una;
+	u32 srtt;
+	u32 rtt_var;
+	u32 ss_thresh;
+	u32 cwnd;
+	u32 snd_wl1;
+	u32 snd_wl2;
+	u32 max_snd_window;
+	u8 rexmit_thresh;
+	u32 local_ipaddr0;
+	u32 local_ipaddr1;
+	u32 local_ipaddr2;
+	u32 local_ipaddr3;
+	bool ignore_tcp_opt;
+	bool ignore_tcp_uns_opt;
+};
+
+struct i40iw_qp_host_ctx_info {
+	u64 qp_compl_ctx;
+	struct i40iw_tcp_offload_info *tcp_info;
+	struct i40iwarp_offload_info *iwarp_info;
+	u32 send_cq_num;
+	u32 rcv_cq_num;
+	u16 push_idx;
+	bool push_mode_en;
+	bool tcp_info_valid;
+	bool iwarp_info_valid;
+	bool err_rq_idx_valid;
+	u16 err_rq_idx;
+};
+
+struct i40iw_aeqe_info {
+	u64 compl_ctx;
+	u32 qp_cq_id;
+	u16 ae_id;
+	u16 wqe_idx;
+	u8 tcp_state;
+	u8 iwarp_state;
+	bool qp;
+	bool cq;
+	bool sq;
+	bool in_rdrsp_wr;
+	bool out_rdrsp;
+	u8 q2_data_written;
+	bool aeqe_overflow;
+};
+
+struct i40iw_allocate_stag_info {
+	u64 total_len;
+	u32 chunk_size;
+	u32 stag_idx;
+	u32 page_size;
+	u16 pd_id;
+	u16 access_rights;
+	bool remote_access;
+	bool use_hmc_fcn_index;
+	u8 hmc_fcn_index;
+	bool use_pf_rid;
+};
+
+struct i40iw_reg_ns_stag_info {
+	u64 reg_addr_pa;
+	u64 fbo;
+	void *va;
+	u64 total_len;
+	u32 page_size;
+	u32 chunk_size;
+	u32 first_pm_pbl_index;
+	enum i40iw_addressing_type addr_type;
+	i40iw_stag_index stag_idx;
+	u16 access_rights;
+	u16 pd_id;
+	i40iw_stag_key stag_key;
+	bool use_hmc_fcn_index;
+	u8 hmc_fcn_index;
+	bool use_pf_rid;
+};
+
+struct i40iw_fast_reg_stag_info {
+	u64 wr_id;
+	u64 reg_addr_pa;
+	u64 fbo;
+	void *va;
+	u64 total_len;
+	u32 page_size;
+	u32 chunk_size;
+	u32 first_pm_pbl_index;
+	enum i40iw_addressing_type addr_type;
+	i40iw_stag_index stag_idx;
+	u16 access_rights;
+	u16 pd_id;
+	i40iw_stag_key stag_key;
+	bool local_fence;
+	bool read_fence;
+	bool signaled;
+	bool use_hmc_fcn_index;
+	u8 hmc_fcn_index;
+	bool use_pf_rid;
+	bool defer_flag;
+};
+
+struct i40iw_dealloc_stag_info {
+	u32 stag_idx;
+	u16 pd_id;
+	bool mr;
+	bool dealloc_pbl;
+};
+
+struct i40iw_register_shared_stag {
+	void *va;
+	enum i40iw_addressing_type addr_type;
+	i40iw_stag_index new_stag_idx;
+	i40iw_stag_index parent_stag_idx;
+	u32 access_rights;
+	u16 pd_id;
+	i40iw_stag_key new_stag_key;
+};
+
+struct i40iw_qp_init_info {
+	struct i40iw_qp_uk_init_info qp_uk_init_info;
+	struct i40iw_sc_pd *pd;
+	u64 *host_ctx;
+	u8 *q2;
+	u64 sq_pa;
+	u64 rq_pa;
+	u64 host_ctx_pa;
+	u64 q2_pa;
+	u64 shadow_area_pa;
+	u8 sq_tph_val;
+	u8 rq_tph_val;
+	u8 type;
+	bool sq_tph_en;
+	bool rq_tph_en;
+	bool rcv_tph_en;
+	bool xmit_tph_en;
+	bool virtual_map;
+};
+
+struct i40iw_cq_init_info {
+	struct i40iw_sc_dev *dev;
+	u64 cq_base_pa;
+	u64 shadow_area_pa;
+	u32 ceq_id;
+	u32 shadow_read_threshold;
+	bool virtual_map;
+	bool ceqe_mask;
+	u8 pbl_chunk_size;
+	u32 first_pm_pbl_idx;
+	bool ceq_id_valid;
+	bool tph_en;
+	u8 tph_val;
+	u8 type;
+	struct i40iw_cq_uk_init_info cq_uk_init_info;
+};
+
+struct i40iw_upload_context_info {
+	u64 buf_pa;
+	bool freeze_qp;
+	bool raw_format;
+	u32 qp_id;
+	u8 qp_type;
+};
+
+struct i40iw_add_arp_cache_entry_info {
+	u8 mac_addr[6];
+	u32 reach_max;
+	u16 arp_index;
+	bool permanent;
+};
+
+struct i40iw_apbvt_info {
+	u16 port;
+	bool add;
+};
+
+enum i40iw_quad_entry_type {
+	I40IW_QHASH_TYPE_TCP_ESTABLISHED = 1,
+	I40IW_QHASH_TYPE_TCP_SYN,
+};
+
+enum i40iw_quad_hash_manage_type {
+	I40IW_QHASH_MANAGE_TYPE_DELETE = 0,
+	I40IW_QHASH_MANAGE_TYPE_ADD,
+	I40IW_QHASH_MANAGE_TYPE_MODIFY
+};
+
+struct i40iw_qhash_table_info {
+	enum i40iw_quad_hash_manage_type manage;
+	enum i40iw_quad_entry_type entry_type;
+	bool vlan_valid;
+	bool ipv4_valid;
+	u8 mac_addr[6];
+	u16 vlan_id;
+	u16 qs_handle;
+	u32 qp_num;
+	u32 dest_ip[4];
+	u32 src_ip[4];
+	u32 dest_port;
+	u32 src_port;
+};
+
+struct i40iw_local_mac_ipaddr_entry_info {
+	u8 mac_addr[6];
+	u8 entry_idx;
+};
+
+struct i40iw_cqp_manage_push_page_info {
+	u32 push_idx;
+	u16 qs_handle;
+	u8 free_page;
+};
+
+struct i40iw_qp_flush_info {
+	u16 sq_minor_code;
+	u16 sq_major_code;
+	u16 rq_minor_code;
+	u16 rq_major_code;
+	u16 ae_code;
+	u8 ae_source;
+	bool sq;
+	bool rq;
+	bool userflushcode;
+	bool generate_ae;
+};
+
+struct i40iw_cqp_commit_fpm_values {
+	u64 qp_base;
+	u64 cq_base;
+	u32 hte_base;
+	u32 arp_base;
+	u32 apbvt_inuse_base;
+	u32 mr_base;
+	u32 xf_base;
+	u32 xffl_base;
+	u32 q1_base;
+	u32 q1fl_base;
+	u32 fsimc_base;
+	u32 fsiav_base;
+	u32 pbl_base;
+
+	u32 qp_cnt;
+	u32 cq_cnt;
+	u32 hte_cnt;
+	u32 arp_cnt;
+	u32 mr_cnt;
+	u32 xf_cnt;
+	u32 xffl_cnt;
+	u32 q1_cnt;
+	u32 q1fl_cnt;
+	u32 fsimc_cnt;
+	u32 fsiav_cnt;
+	u32 pbl_cnt;
+};
+
+struct i40iw_cqp_query_fpm_values {
+	u16 first_pe_sd_index;
+	u32 qp_objsize;
+	u32 cq_objsize;
+	u32 hte_objsize;
+	u32 arp_objsize;
+	u32 mr_objsize;
+	u32 xf_objsize;
+	u32 q1_objsize;
+	u32 fsimc_objsize;
+	u32 fsiav_objsize;
+
+	u32 qp_max;
+	u32 cq_max;
+	u32 hte_max;
+	u32 arp_max;
+	u32 mr_max;
+	u32 xf_max;
+	u32 xffl_max;
+	u32 q1_max;
+	u32 q1fl_max;
+	u32 fsimc_max;
+	u32 fsiav_max;
+	u32 pbl_max;
+};
+
+struct i40iw_cqp_ops {
+	enum i40iw_status_code (*cqp_init)(struct i40iw_sc_cqp *,
+					   struct i40iw_cqp_init_info *);
+	enum i40iw_status_code (*cqp_create)(struct i40iw_sc_cqp *, bool, u16 *, u16 *);
+	void (*cqp_post_sq)(struct i40iw_sc_cqp *);
+	u64 *(*cqp_get_next_send_wqe)(struct i40iw_sc_cqp *, u64 scratch);
+	enum i40iw_status_code (*cqp_destroy)(struct i40iw_sc_cqp *);
+	enum i40iw_status_code (*poll_for_cqp_op_done)(struct i40iw_sc_cqp *, u8,
+						       struct i40iw_ccq_cqe_info *);
+};
+
+struct i40iw_ccq_ops {
+	enum i40iw_status_code (*ccq_init)(struct i40iw_sc_cq *,
+					   struct i40iw_ccq_init_info *);
+	enum i40iw_status_code (*ccq_create)(struct i40iw_sc_cq *, u64, bool, bool);
+	enum i40iw_status_code (*ccq_destroy)(struct i40iw_sc_cq *, u64, bool);
+	enum i40iw_status_code (*ccq_create_done)(struct i40iw_sc_cq *);
+	enum i40iw_status_code (*ccq_get_cqe_info)(struct i40iw_sc_cq *,
+						   struct i40iw_ccq_cqe_info *);
+	void (*ccq_arm)(struct i40iw_sc_cq *);
+};
+
+struct i40iw_ceq_ops {
+	enum i40iw_status_code (*ceq_init)(struct i40iw_sc_ceq *,
+					   struct i40iw_ceq_init_info *);
+	enum i40iw_status_code (*ceq_create)(struct i40iw_sc_ceq *, u64, bool);
+	enum i40iw_status_code (*cceq_create_done)(struct i40iw_sc_ceq *);
+	enum i40iw_status_code (*cceq_destroy_done)(struct i40iw_sc_ceq *);
+	enum i40iw_status_code (*cceq_create)(struct i40iw_sc_ceq *, u64);
+	enum i40iw_status_code (*ceq_destroy)(struct i40iw_sc_ceq *, u64, bool);
+	void *(*process_ceq)(struct i40iw_sc_dev *, struct i40iw_sc_ceq *);
+};
+
+struct i40iw_aeq_ops {
+	enum i40iw_status_code (*aeq_init)(struct i40iw_sc_aeq *,
+					   struct i40iw_aeq_init_info *);
+	enum i40iw_status_code (*aeq_create)(struct i40iw_sc_aeq *, u64, bool);
+	enum i40iw_status_code (*aeq_destroy)(struct i40iw_sc_aeq *, u64, bool);
+	enum i40iw_status_code (*get_next_aeqe)(struct i40iw_sc_aeq *,
+						struct i40iw_aeqe_info *);
+	enum i40iw_status_code (*repost_aeq_entries)(struct i40iw_sc_dev *, u32);
+	enum i40iw_status_code (*aeq_create_done)(struct i40iw_sc_aeq *);
+	enum i40iw_status_code (*aeq_destroy_done)(struct i40iw_sc_aeq *);
+};
+
+struct i40iw_pd_ops {
+	void (*pd_init)(struct i40iw_sc_dev *, struct i40iw_sc_pd *, u16);
+};
+
+struct i40iw_priv_qp_ops {
+	enum i40iw_status_code (*qp_init)(struct i40iw_sc_qp *, struct i40iw_qp_init_info *);
+	enum i40iw_status_code (*qp_create)(struct i40iw_sc_qp *,
+					    struct i40iw_create_qp_info *, u64, bool);
+	enum i40iw_status_code (*qp_modify)(struct i40iw_sc_qp *,
+					    struct i40iw_modify_qp_info *, u64, bool);
+	enum i40iw_status_code (*qp_destroy)(struct i40iw_sc_qp *, u64, bool, bool, bool);
+	enum i40iw_status_code (*qp_flush_wqes)(struct i40iw_sc_qp *,
+						struct i40iw_qp_flush_info *, u64, bool);
+	enum i40iw_status_code (*qp_upload_context)(struct i40iw_sc_dev *,
+						    struct i40iw_upload_context_info *,
+						    u64, bool);
+	enum i40iw_status_code (*qp_setctx)(struct i40iw_sc_qp *, u64 *,
+					    struct i40iw_qp_host_ctx_info *);
+
+	void (*qp_send_lsmm)(struct i40iw_sc_qp *, void *, u32, i40iw_stag);
+	void (*qp_send_lsmm_nostag)(struct i40iw_sc_qp *, void *, u32);
+	void (*qp_send_rtt)(struct i40iw_sc_qp *, bool);
+	enum i40iw_status_code (*qp_post_wqe0)(struct i40iw_sc_qp *, u8);
+};
+
+struct i40iw_priv_cq_ops {
+	enum i40iw_status_code (*cq_init)(struct i40iw_sc_cq *, struct i40iw_cq_init_info *);
+	enum i40iw_status_code (*cq_create)(struct i40iw_sc_cq *, u64, bool, bool);
+	enum i40iw_status_code (*cq_destroy)(struct i40iw_sc_cq *, u64, bool);
+	enum i40iw_status_code (*cq_modify)(struct i40iw_sc_cq *,
+					    struct i40iw_modify_cq_info *, u64, bool);
+};
+
+struct i40iw_mr_ops {
+	enum i40iw_status_code (*alloc_stag)(struct i40iw_sc_dev *,
+					     struct i40iw_allocate_stag_info *, u64, bool);
+	enum i40iw_status_code (*mr_reg_non_shared)(struct i40iw_sc_dev *,
+						    struct i40iw_reg_ns_stag_info *,
+						    u64, bool);
+	enum i40iw_status_code (*mr_reg_shared)(struct i40iw_sc_dev *,
+						struct i40iw_register_shared_stag *,
+						u64, bool);
+	enum i40iw_status_code (*dealloc_stag)(struct i40iw_sc_dev *,
+					       struct i40iw_dealloc_stag_info *,
+					       u64, bool);
+	enum i40iw_status_code (*query_stag)(struct i40iw_sc_dev *, u64, u32, bool);
+	enum i40iw_status_code (*mw_alloc)(struct i40iw_sc_dev *, u64, u32, u16, bool);
+};
+
+struct i40iw_cqp_misc_ops {
+	enum i40iw_status_code (*manage_push_page)(struct i40iw_sc_cqp *,
+						   struct i40iw_cqp_manage_push_page_info *,
+						   u64, bool);
+	enum i40iw_status_code (*manage_hmc_pm_func_table)(struct i40iw_sc_cqp *,
+							   u64, u8, bool, bool);
+	enum i40iw_status_code (*set_hmc_resource_profile)(struct i40iw_sc_cqp *,
+							   u64, u8, u8, bool, bool);
+	enum i40iw_status_code (*commit_fpm_values)(struct i40iw_sc_cqp *, u64, u8,
+						    struct i40iw_dma_mem *, bool, u8);
+	enum i40iw_status_code (*query_fpm_values)(struct i40iw_sc_cqp *, u64, u8,
+						   struct i40iw_dma_mem *, bool, u8);
+	enum i40iw_status_code (*static_hmc_pages_allocated)(struct i40iw_sc_cqp *,
+							     u64, u8, bool, bool);
+	enum i40iw_status_code (*add_arp_cache_entry)(struct i40iw_sc_cqp *,
+						      struct i40iw_add_arp_cache_entry_info *,
+						      u64, bool);
+	enum i40iw_status_code (*del_arp_cache_entry)(struct i40iw_sc_cqp *, u64, u16, bool);
+	enum i40iw_status_code (*query_arp_cache_entry)(struct i40iw_sc_cqp *, u64, u16, bool);
+	enum i40iw_status_code (*manage_apbvt_entry)(struct i40iw_sc_cqp *,
+						     struct i40iw_apbvt_info *, u64, bool);
+	enum i40iw_status_code (*manage_qhash_table_entry)(struct i40iw_sc_cqp *,
+							   struct i40iw_qhash_table_info *, u64, bool);
+	enum i40iw_status_code (*alloc_local_mac_ipaddr_table_entry)(struct i40iw_sc_cqp *, u64, bool);
+	enum i40iw_status_code (*add_local_mac_ipaddr_entry)(struct i40iw_sc_cqp *,
+							     struct i40iw_local_mac_ipaddr_entry_info *,
+							     u64, bool);
+	enum i40iw_status_code (*del_local_mac_ipaddr_entry)(struct i40iw_sc_cqp *, u64, u8, u8, bool);
+	enum i40iw_status_code (*cqp_nop)(struct i40iw_sc_cqp *, u64, bool);
+	enum i40iw_status_code (*commit_fpm_values_done)(struct i40iw_sc_cqp
+							  *);
+	enum i40iw_status_code (*query_fpm_values_done)(struct i40iw_sc_cqp *);
+	enum i40iw_status_code (*manage_hmc_pm_func_table_done)(struct i40iw_sc_cqp *);
+	enum i40iw_status_code (*update_suspend_qp)(struct i40iw_sc_cqp *, struct i40iw_sc_qp *, u64);
+	enum i40iw_status_code (*update_resume_qp)(struct i40iw_sc_cqp *, struct i40iw_sc_qp *, u64);
+};
+
+struct i40iw_hmc_ops {
+	enum i40iw_status_code (*init_iw_hmc)(struct i40iw_sc_dev *, u8);
+	enum i40iw_status_code (*parse_fpm_query_buf)(u64 *, struct i40iw_hmc_info *,
+						      struct i40iw_hmc_fpm_misc *);
+	enum i40iw_status_code (*configure_iw_fpm)(struct i40iw_sc_dev *, u8);
+	enum i40iw_status_code (*parse_fpm_commit_buf)(u64 *, struct i40iw_hmc_obj_info *);
+	enum i40iw_status_code (*create_hmc_object)(struct i40iw_sc_dev *dev,
+						    struct i40iw_hmc_create_obj_info *);
+	enum i40iw_status_code (*del_hmc_object)(struct i40iw_sc_dev *dev,
+						 struct i40iw_hmc_del_obj_info *,
+						 bool reset);
+	enum i40iw_status_code (*pf_init_vfhmc)(struct i40iw_sc_dev *, u8, u32 *);
+	enum i40iw_status_code (*vf_configure_vffpm)(struct i40iw_sc_dev *, u32 *);
+};
+
+struct cqp_info {
+	union {
+		struct {
+			struct i40iw_sc_qp *qp;
+			struct i40iw_create_qp_info info;
+			u64 scratch;
+		} qp_create;
+
+		struct {
+			struct i40iw_sc_qp *qp;
+			struct i40iw_modify_qp_info info;
+			u64 scratch;
+		} qp_modify;
+
+		struct {
+			struct i40iw_sc_qp *qp;
+			u64 scratch;
+			bool remove_hash_idx;
+			bool ignore_mw_bnd;
+		} qp_destroy;
+
+		struct {
+			struct i40iw_sc_cq *cq;
+			u64 scratch;
+			bool check_overflow;
+		} cq_create;
+
+		struct {
+			struct i40iw_sc_cq *cq;
+			u64 scratch;
+		} cq_destroy;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_allocate_stag_info info;
+			u64 scratch;
+		} alloc_stag;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			u64 scratch;
+			u32 mw_stag_index;
+			u16 pd_id;
+		} mw_alloc;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_reg_ns_stag_info info;
+			u64 scratch;
+		} mr_reg_non_shared;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_dealloc_stag_info info;
+			u64 scratch;
+		} dealloc_stag;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_local_mac_ipaddr_entry_info info;
+			u64 scratch;
+		} add_local_mac_ipaddr_entry;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_add_arp_cache_entry_info info;
+			u64 scratch;
+		} add_arp_cache_entry;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			u64 scratch;
+			u8 entry_idx;
+			u8 ignore_ref_count;
+		} del_local_mac_ipaddr_entry;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			u64 scratch;
+			u16 arp_index;
+		} del_arp_cache_entry;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_manage_vf_pble_info info;
+			u64 scratch;
+		} manage_vf_pble_bp;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_cqp_manage_push_page_info info;
+			u64 scratch;
+		} manage_push_page;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_upload_context_info info;
+			u64 scratch;
+		} qp_upload_context;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			u64 scratch;
+		} alloc_local_mac_ipaddr_entry;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_hmc_fcn_info info;
+			u64 scratch;
+		} manage_hmc_pm;
+
+		struct {
+			struct i40iw_sc_ceq *ceq;
+			u64 scratch;
+		} ceq_create;
+
+		struct {
+			struct i40iw_sc_ceq *ceq;
+			u64 scratch;
+		} ceq_destroy;
+
+		struct {
+			struct i40iw_sc_aeq *aeq;
+			u64 scratch;
+		} aeq_create;
+
+		struct {
+			struct i40iw_sc_aeq *aeq;
+			u64 scratch;
+		} aeq_destroy;
+
+		struct {
+			struct i40iw_sc_qp *qp;
+			struct i40iw_qp_flush_info info;
+			u64 scratch;
+		} qp_flush_wqes;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			void *fpm_values_va;
+			u64 fpm_values_pa;
+			u8 hmc_fn_id;
+			u64 scratch;
+		} query_fpm_values;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			void *fpm_values_va;
+			u64 fpm_values_pa;
+			u8 hmc_fn_id;
+			u64 scratch;
+		} commit_fpm_values;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_apbvt_info info;
+			u64 scratch;
+		} manage_apbvt_entry;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_qhash_table_info info;
+			u64 scratch;
+		} manage_qhash_table_entry;
+
+		struct {
+			struct i40iw_sc_dev *dev;
+			struct i40iw_update_sds_info info;
+			u64 scratch;
+		} update_pe_sds;
+
+		struct {
+			struct i40iw_sc_cqp *cqp;
+			struct i40iw_sc_qp *qp;
+			u64 scratch;
+		} suspend_resume;
+	} u;
+};
+
+struct cqp_commands_info {
+	struct list_head cqp_cmd_entry;
+	u8 cqp_cmd;
+	u8 post_sq;
+	struct cqp_info in;
+};
+
+struct i40iw_virtchnl_work_info {
+	void (*callback_fcn)(void *vf_dev);
+	void *worker_vf_dev;
+};
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_ucontext.h b/drivers/infiniband/hw/i40iw/i40iw_ucontext.h
new file mode 100644
index 000000000000..12acd688def4
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_ucontext.h
@@ -0,0 +1,107 @@
+/*
+ * Copyright (c) 2006 - 2016 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2005 Topspin Communications.  All rights reserved.
+ * Copyright (c) 2005 Cisco Systems.  All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef I40IW_USER_CONTEXT_H
+#define I40IW_USER_CONTEXT_H
+
+#include <linux/types.h>
+
+#define I40IW_ABI_USERSPACE_VER 4
+#define I40IW_ABI_KERNEL_VER    4
+struct i40iw_alloc_ucontext_req {
+	__u32 reserved32;
+	__u8 userspace_ver;
+	__u8 reserved8[3];
+};
+
+struct i40iw_alloc_ucontext_resp {
+	__u32 max_pds;		/* maximum pds allowed for this user process */
+	__u32 max_qps;		/* maximum qps allowed for this user process */
+	__u32 wq_size;		/* size of the WQs (sq+rq) allocated to the mmaped area */
+	__u8 kernel_ver;
+	__u8 reserved[3];
+};
+
+struct i40iw_alloc_pd_resp {
+	__u32 pd_id;
+	__u8 reserved[4];
+};
+
+struct i40iw_create_cq_req {
+	__u64 user_cq_buffer;
+	__u64 user_shadow_area;
+};
+
+struct i40iw_create_qp_req {
+	__u64 user_wqe_buffers;
+	__u64 user_compl_ctx;
+
+	/* UDA QP PHB */
+	__u64 user_sq_phb;	/* place for VA of the sq phb buff */
+	__u64 user_rq_phb;	/* place for VA of the rq phb buff */
+};
+
+enum i40iw_memreg_type {
+	IW_MEMREG_TYPE_MEM = 0x0000,
+	IW_MEMREG_TYPE_QP = 0x0001,
+	IW_MEMREG_TYPE_CQ = 0x0002,
+};
+
+struct i40iw_mem_reg_req {
+	__u16 reg_type;		/* Memory, QP or CQ */
+	__u16 cq_pages;
+	__u16 rq_pages;
+	__u16 sq_pages;
+};
+
+struct i40iw_create_cq_resp {
+	__u32 cq_id;
+	__u32 cq_size;
+	__u32 mmap_db_index;
+	__u32 reserved;
+};
+
+struct i40iw_create_qp_resp {
+	__u32 qp_id;
+	__u32 actual_sq_size;
+	__u32 actual_rq_size;
+	__u32 i40iw_drv_opt;
+	__u16 push_idx;
+	__u8  lsmm;
+	__u8  rsvd2;
+};
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_uk.c b/drivers/infiniband/hw/i40iw/i40iw_uk.c
new file mode 100644
index 000000000000..f78c3dc8bdb2
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_uk.c
@@ -0,0 +1,1204 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_status.h"
+#include "i40iw_d.h"
+#include "i40iw_user.h"
+#include "i40iw_register.h"
+
+static u32 nop_signature = 0x55550000;
+
+/**
+ * i40iw_nop_1 - insert a nop wqe and move head. no post work
+ * @qp: hw qp ptr
+ */
+static enum i40iw_status_code i40iw_nop_1(struct i40iw_qp_uk *qp)
+{
+	u64 header, *wqe;
+	u64 *wqe_0 = NULL;
+	u32 wqe_idx, peek_head;
+	bool signaled = false;
+
+	if (!qp->sq_ring.head)
+		return I40IW_ERR_PARAM;
+
+	wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+	wqe = qp->sq_base[wqe_idx].elem;
+	peek_head = (qp->sq_ring.head + 1) % qp->sq_ring.size;
+	wqe_0 = qp->sq_base[peek_head].elem;
+	if (peek_head)
+		wqe_0[3] = LS_64(!qp->swqe_polarity, I40IWQPSQ_VALID);
+	else
+		wqe_0[3] = LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	set_64bit_val(wqe, 0, 0);
+	set_64bit_val(wqe, 8, 0);
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64(I40IWQP_OP_NOP, I40IWQPSQ_OPCODE) |
+	    LS_64(signaled, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->swqe_polarity, I40IWQPSQ_VALID) | nop_signature++;
+
+	wmb();	/* Memory barrier to ensure data is written before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+	return 0;
+}
+
+/**
+ * i40iw_qp_post_wr - post wr to hrdware
+ * @qp: hw qp ptr
+ */
+void i40iw_qp_post_wr(struct i40iw_qp_uk *qp)
+{
+	u64 temp;
+	u32 hw_sq_tail;
+	u32 sw_sq_head;
+
+	mb(); /* valid bit is written and loads completed before reading shadow */
+
+	/* read the doorbell shadow area */
+	get_64bit_val(qp->shadow_area, 0, &temp);
+
+	hw_sq_tail = (u32)RS_64(temp, I40IW_QP_DBSA_HW_SQ_TAIL);
+	sw_sq_head = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+	if (sw_sq_head != hw_sq_tail) {
+		if (sw_sq_head > qp->initial_ring.head) {
+			if ((hw_sq_tail >= qp->initial_ring.head) &&
+			    (hw_sq_tail < sw_sq_head)) {
+				writel(qp->qp_id, qp->wqe_alloc_reg);
+			}
+		} else if (sw_sq_head != qp->initial_ring.head) {
+			if ((hw_sq_tail >= qp->initial_ring.head) ||
+			    (hw_sq_tail < sw_sq_head)) {
+				writel(qp->qp_id, qp->wqe_alloc_reg);
+			}
+		}
+	}
+
+	qp->initial_ring.head = qp->sq_ring.head;
+}
+
+/**
+ * i40iw_qp_ring_push_db -  ring qp doorbell
+ * @qp: hw qp ptr
+ * @wqe_idx: wqe index
+ */
+static void i40iw_qp_ring_push_db(struct i40iw_qp_uk *qp, u32 wqe_idx)
+{
+	set_32bit_val(qp->push_db, 0, LS_32((wqe_idx >> 2), I40E_PFPE_WQEALLOC_WQE_DESC_INDEX) | qp->qp_id);
+	qp->initial_ring.head = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+}
+
+/**
+ * i40iw_qp_get_next_send_wqe - return next wqe ptr
+ * @qp: hw qp ptr
+ * @wqe_idx: return wqe index
+ * @wqe_size: size of sq wqe
+ */
+u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp,
+				u32 *wqe_idx,
+				u8 wqe_size)
+{
+	u64 *wqe = NULL;
+	u64 wqe_ptr;
+	u32 peek_head = 0;
+	u16 offset;
+	enum i40iw_status_code ret_code = 0;
+	u8 nop_wqe_cnt = 0, i;
+	u64 *wqe_0 = NULL;
+
+	*wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+
+	if (!*wqe_idx)
+		qp->swqe_polarity = !qp->swqe_polarity;
+	wqe_ptr = (uintptr_t)qp->sq_base[*wqe_idx].elem;
+	offset = (u16)(wqe_ptr) & 0x7F;
+	if ((offset + wqe_size) > I40IW_QP_WQE_MAX_SIZE) {
+		nop_wqe_cnt = (u8)(I40IW_QP_WQE_MAX_SIZE - offset) / I40IW_QP_WQE_MIN_SIZE;
+		for (i = 0; i < nop_wqe_cnt; i++) {
+			i40iw_nop_1(qp);
+			I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
+			if (ret_code)
+				return NULL;
+		}
+
+		*wqe_idx = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+		if (!*wqe_idx)
+			qp->swqe_polarity = !qp->swqe_polarity;
+	}
+	for (i = 0; i < wqe_size / I40IW_QP_WQE_MIN_SIZE; i++) {
+		I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
+		if (ret_code)
+			return NULL;
+	}
+
+	wqe = qp->sq_base[*wqe_idx].elem;
+
+	peek_head = I40IW_RING_GETCURRENT_HEAD(qp->sq_ring);
+	wqe_0 = qp->sq_base[peek_head].elem;
+	if (peek_head & 0x3)
+		wqe_0[3] = LS_64(!qp->swqe_polarity, I40IWQPSQ_VALID);
+	return wqe;
+}
+
+/**
+ * i40iw_set_fragment - set fragment in wqe
+ * @wqe: wqe for setting fragment
+ * @offset: offset value
+ * @sge: sge length and stag
+ */
+static void i40iw_set_fragment(u64 *wqe, u32 offset, struct i40iw_sge *sge)
+{
+	if (sge) {
+		set_64bit_val(wqe, offset, LS_64(sge->tag_off, I40IWQPSQ_FRAG_TO));
+		set_64bit_val(wqe, (offset + 8),
+			      (LS_64(sge->len, I40IWQPSQ_FRAG_LEN) |
+			       LS_64(sge->stag, I40IWQPSQ_FRAG_STAG)));
+	}
+}
+
+/**
+ * i40iw_qp_get_next_recv_wqe - get next qp's rcv wqe
+ * @qp: hw qp ptr
+ * @wqe_idx: return wqe index
+ */
+u64 *i40iw_qp_get_next_recv_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx)
+{
+	u64 *wqe = NULL;
+	enum i40iw_status_code ret_code;
+
+	if (I40IW_RING_FULL_ERR(qp->rq_ring))
+		return NULL;
+
+	I40IW_ATOMIC_RING_MOVE_HEAD(qp->rq_ring, *wqe_idx, ret_code);
+	if (ret_code)
+		return NULL;
+	if (!*wqe_idx)
+		qp->rwqe_polarity = !qp->rwqe_polarity;
+	/* rq_wqe_size_multiplier is no of qwords in one rq wqe */
+	wqe = qp->rq_base[*wqe_idx * (qp->rq_wqe_size_multiplier >> 2)].elem;
+
+	return wqe;
+}
+
+/**
+ * i40iw_rdma_write - rdma write operation
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_rdma_write(struct i40iw_qp_uk *qp,
+					       struct i40iw_post_sq_info *info,
+					       bool post_sq)
+{
+	u64 header;
+	u64 *wqe;
+	struct i40iw_rdma_write *op_info;
+	u32 i, wqe_idx;
+	u32 total_size = 0, byte_off;
+	enum i40iw_status_code ret_code;
+	bool read_fence = false;
+	u8 wqe_size;
+
+	op_info = &info->op.rdma_write;
+	if (op_info->num_lo_sges > qp->max_sq_frag_cnt)
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+
+	for (i = 0; i < op_info->num_lo_sges; i++)
+		total_size += op_info->lo_sg_list[i].len;
+
+	if (total_size > I40IW_MAX_OUTBOUND_MESSAGE_SIZE)
+		return I40IW_ERR_QP_INVALID_MSG_SIZE;
+
+	read_fence |= info->read_fence;
+
+	ret_code = i40iw_fragcnt_to_wqesize_sq(op_info->num_lo_sges, &wqe_size);
+	if (ret_code)
+		return ret_code;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = total_size;
+	set_64bit_val(wqe, 16,
+		      LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
+	if (!op_info->rem_addr.stag)
+		return I40IW_ERR_BAD_STAG;
+
+	header = LS_64(op_info->rem_addr.stag, I40IWQPSQ_REMSTAG) |
+		 LS_64(I40IWQP_OP_RDMA_WRITE, I40IWQPSQ_OPCODE) |
+		 LS_64((op_info->num_lo_sges > 1 ?  (op_info->num_lo_sges - 1) : 0), I40IWQPSQ_ADDFRAGCNT) |
+		 LS_64(read_fence, I40IWQPSQ_READFENCE) |
+		 LS_64(info->local_fence, I40IWQPSQ_LOCALFENCE) |
+		 LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+		 LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_set_fragment(wqe, 0, op_info->lo_sg_list);
+
+	for (i = 1; i < op_info->num_lo_sges; i++) {
+		byte_off = 32 + (i - 1) * 16;
+		i40iw_set_fragment(wqe, byte_off, &op_info->lo_sg_list[i]);
+	}
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_rdma_read - rdma read command
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @inv_stag: flag for inv_stag
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_rdma_read(struct i40iw_qp_uk *qp,
+					      struct i40iw_post_sq_info *info,
+					      bool inv_stag,
+					      bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_rdma_read *op_info;
+	u64 header;
+	u32 wqe_idx;
+	enum i40iw_status_code ret_code;
+	u8 wqe_size;
+	bool local_fence = false;
+
+	op_info = &info->op.rdma_read;
+	ret_code = i40iw_fragcnt_to_wqesize_sq(1, &wqe_size);
+	if (ret_code)
+		return ret_code;
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->lo_addr.len;
+	local_fence |= info->local_fence;
+
+	set_64bit_val(wqe, 16, LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
+	header = LS_64(op_info->rem_addr.stag, I40IWQPSQ_REMSTAG) |
+		 LS_64((inv_stag ? I40IWQP_OP_RDMA_READ_LOC_INV : I40IWQP_OP_RDMA_READ), I40IWQPSQ_OPCODE) |
+		 LS_64(info->read_fence, I40IWQPSQ_READFENCE) |
+		 LS_64(local_fence, I40IWQPSQ_LOCALFENCE) |
+		 LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+		 LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_set_fragment(wqe, 0, &op_info->lo_addr);
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_send - rdma send command
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @stag_to_inv: stag_to_inv value
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_send(struct i40iw_qp_uk *qp,
+					 struct i40iw_post_sq_info *info,
+					 u32 stag_to_inv,
+					 bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_post_send *op_info;
+	u64 header;
+	u32 i, wqe_idx, total_size = 0, byte_off;
+	enum i40iw_status_code ret_code;
+	bool read_fence = false;
+	u8 wqe_size;
+
+	op_info = &info->op.send;
+	if (qp->max_sq_frag_cnt < op_info->num_sges)
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+
+	for (i = 0; i < op_info->num_sges; i++)
+		total_size += op_info->sg_list[i].len;
+	ret_code = i40iw_fragcnt_to_wqesize_sq(op_info->num_sges, &wqe_size);
+	if (ret_code)
+		return ret_code;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	read_fence |= info->read_fence;
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = total_size;
+	set_64bit_val(wqe, 16, 0);
+	header = LS_64(stag_to_inv, I40IWQPSQ_REMSTAG) |
+		 LS_64(info->op_type, I40IWQPSQ_OPCODE) |
+		 LS_64((op_info->num_sges > 1 ? (op_info->num_sges - 1) : 0),
+		       I40IWQPSQ_ADDFRAGCNT) |
+		 LS_64(read_fence, I40IWQPSQ_READFENCE) |
+		 LS_64(info->local_fence, I40IWQPSQ_LOCALFENCE) |
+		 LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+		 LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_set_fragment(wqe, 0, op_info->sg_list);
+
+	for (i = 1; i < op_info->num_sges; i++) {
+		byte_off = 32 + (i - 1) * 16;
+		i40iw_set_fragment(wqe, byte_off, &op_info->sg_list[i]);
+	}
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_inline_rdma_write - inline rdma write operation
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_inline_rdma_write(struct i40iw_qp_uk *qp,
+						      struct i40iw_post_sq_info *info,
+						      bool post_sq)
+{
+	u64 *wqe;
+	u8 *dest, *src;
+	struct i40iw_inline_rdma_write *op_info;
+	u64 *push;
+	u64 header = 0;
+	u32 i, wqe_idx;
+	enum i40iw_status_code ret_code;
+	bool read_fence = false;
+	u8 wqe_size;
+
+	op_info = &info->op.inline_rdma_write;
+	if (op_info->len > I40IW_MAX_INLINE_DATA_SIZE)
+		return I40IW_ERR_INVALID_IMM_DATA_SIZE;
+
+	ret_code = i40iw_inline_data_size_to_wqesize(op_info->len, &wqe_size);
+	if (ret_code)
+		return ret_code;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	read_fence |= info->read_fence;
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->len;
+	set_64bit_val(wqe, 16,
+		      LS_64(op_info->rem_addr.tag_off, I40IWQPSQ_FRAG_TO));
+
+	header = LS_64(op_info->rem_addr.stag, I40IWQPSQ_REMSTAG) |
+		 LS_64(I40IWQP_OP_RDMA_WRITE, I40IWQPSQ_OPCODE) |
+		 LS_64(op_info->len, I40IWQPSQ_INLINEDATALEN) |
+		 LS_64(1, I40IWQPSQ_INLINEDATAFLAG) |
+		 LS_64((qp->push_db ? 1 : 0), I40IWQPSQ_PUSHWQE) |
+		 LS_64(read_fence, I40IWQPSQ_READFENCE) |
+		 LS_64(info->local_fence, I40IWQPSQ_LOCALFENCE) |
+		 LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+		 LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	dest = (u8 *)wqe;
+	src = (u8 *)(op_info->data);
+
+	if (op_info->len <= 16) {
+		for (i = 0; i < op_info->len; i++, src++, dest++)
+			*dest = *src;
+	} else {
+		for (i = 0; i < 16; i++, src++, dest++)
+			*dest = *src;
+		dest = (u8 *)wqe + 32;
+		for (; i < op_info->len; i++, src++, dest++)
+			*dest = *src;
+	}
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	if (qp->push_db) {
+		push = (u64 *)((uintptr_t)qp->push_wqe + (wqe_idx & 0x3) * 0x20);
+		memcpy(push, wqe, (op_info->len > 16) ? op_info->len + 16 : 32);
+		i40iw_qp_ring_push_db(qp, wqe_idx);
+	} else {
+		if (post_sq)
+			i40iw_qp_post_wr(qp);
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_inline_send - inline send operation
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @stag_to_inv: remote stag
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_inline_send(struct i40iw_qp_uk *qp,
+						struct i40iw_post_sq_info *info,
+						u32 stag_to_inv,
+						bool post_sq)
+{
+	u64 *wqe;
+	u8 *dest, *src;
+	struct i40iw_post_inline_send *op_info;
+	u64 header;
+	u32 wqe_idx, i;
+	enum i40iw_status_code ret_code;
+	bool read_fence = false;
+	u8 wqe_size;
+	u64 *push;
+
+	op_info = &info->op.inline_send;
+	if (op_info->len > I40IW_MAX_INLINE_DATA_SIZE)
+		return I40IW_ERR_INVALID_IMM_DATA_SIZE;
+
+	ret_code = i40iw_inline_data_size_to_wqesize(op_info->len, &wqe_size);
+	if (ret_code)
+		return ret_code;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, wqe_size);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	read_fence |= info->read_fence;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = op_info->len;
+	header = LS_64(stag_to_inv, I40IWQPSQ_REMSTAG) |
+	    LS_64(info->op_type, I40IWQPSQ_OPCODE) |
+	    LS_64(op_info->len, I40IWQPSQ_INLINEDATALEN) |
+	    LS_64(1, I40IWQPSQ_INLINEDATAFLAG) |
+	    LS_64((qp->push_db ? 1 : 0), I40IWQPSQ_PUSHWQE) |
+	    LS_64(read_fence, I40IWQPSQ_READFENCE) |
+	    LS_64(info->local_fence, I40IWQPSQ_LOCALFENCE) |
+	    LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	dest = (u8 *)wqe;
+	src = (u8 *)(op_info->data);
+
+	if (op_info->len <= 16) {
+		for (i = 0; i < op_info->len; i++, src++, dest++)
+			*dest = *src;
+	} else {
+		for (i = 0; i < 16; i++, src++, dest++)
+			*dest = *src;
+		dest = (u8 *)wqe + 32;
+		for (; i < op_info->len; i++, src++, dest++)
+			*dest = *src;
+	}
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	if (qp->push_db) {
+		push = (u64 *)((uintptr_t)qp->push_wqe + (wqe_idx & 0x3) * 0x20);
+		memcpy(push, wqe, (op_info->len > 16) ? op_info->len + 16 : 32);
+		i40iw_qp_ring_push_db(qp, wqe_idx);
+	} else {
+		if (post_sq)
+			i40iw_qp_post_wr(qp);
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_stag_local_invalidate - stag invalidate operation
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_stag_local_invalidate(struct i40iw_qp_uk *qp,
+							  struct i40iw_post_sq_info *info,
+							  bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_inv_local_stag *op_info;
+	u64 header;
+	u32 wqe_idx;
+	bool local_fence = false;
+
+	op_info = &info->op.inv_local_stag;
+	local_fence = info->local_fence;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
+	set_64bit_val(wqe, 0, 0);
+	set_64bit_val(wqe, 8,
+		      LS_64(op_info->target_stag, I40IWQPSQ_LOCSTAG));
+	set_64bit_val(wqe, 16, 0);
+	header = LS_64(I40IW_OP_TYPE_INV_STAG, I40IWQPSQ_OPCODE) |
+	    LS_64(info->read_fence, I40IWQPSQ_READFENCE) |
+	    LS_64(local_fence, I40IWQPSQ_LOCALFENCE) |
+	    LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_mw_bind - Memory Window bind operation
+ * @qp: hw qp ptr
+ * @info: post sq information
+ * @post_sq: flag to post sq
+ */
+static enum i40iw_status_code i40iw_mw_bind(struct i40iw_qp_uk *qp,
+					    struct i40iw_post_sq_info *info,
+					    bool post_sq)
+{
+	u64 *wqe;
+	struct i40iw_bind_window *op_info;
+	u64 header;
+	u32 wqe_idx;
+	bool local_fence = false;
+
+	op_info = &info->op.bind_window;
+
+	local_fence |= info->local_fence;
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = info->wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
+	set_64bit_val(wqe, 0, (uintptr_t)op_info->va);
+	set_64bit_val(wqe, 8,
+		      LS_64(op_info->mr_stag, I40IWQPSQ_PARENTMRSTAG) |
+		      LS_64(op_info->mw_stag, I40IWQPSQ_MWSTAG));
+	set_64bit_val(wqe, 16, op_info->bind_length);
+	header = LS_64(I40IW_OP_TYPE_BIND_MW, I40IWQPSQ_OPCODE) |
+	    LS_64(((op_info->enable_reads << 2) |
+		   (op_info->enable_writes << 3)),
+		  I40IWQPSQ_STAGRIGHTS) |
+	    LS_64((op_info->addressing_type == I40IW_ADDR_TYPE_VA_BASED ?  1 : 0),
+		  I40IWQPSQ_VABASEDTO) |
+	    LS_64(info->read_fence, I40IWQPSQ_READFENCE) |
+	    LS_64(local_fence, I40IWQPSQ_LOCALFENCE) |
+	    LS_64(info->signaled, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_post_receive - post receive wqe
+ * @qp: hw qp ptr
+ * @info: post rq information
+ */
+static enum i40iw_status_code i40iw_post_receive(struct i40iw_qp_uk *qp,
+						 struct i40iw_post_rq_info *info)
+{
+	u64 *wqe;
+	u64 header;
+	u32 total_size = 0, wqe_idx, i, byte_off;
+
+	if (qp->max_rq_frag_cnt < info->num_sges)
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+	for (i = 0; i < info->num_sges; i++)
+		total_size += info->sg_list[i].len;
+	wqe = i40iw_qp_get_next_recv_wqe(qp, &wqe_idx);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->rq_wrid_array[wqe_idx] = info->wr_id;
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64((info->num_sges > 1 ? (info->num_sges - 1) : 0),
+		       I40IWQPSQ_ADDFRAGCNT) |
+	    LS_64(qp->rwqe_polarity, I40IWQPSQ_VALID);
+
+	i40iw_set_fragment(wqe, 0, info->sg_list);
+
+	for (i = 1; i < info->num_sges; i++) {
+		byte_off = 32 + (i - 1) * 16;
+		i40iw_set_fragment(wqe, byte_off, &info->sg_list[i]);
+	}
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+
+	return 0;
+}
+
+/**
+ * i40iw_cq_request_notification - cq notification request (door bell)
+ * @cq: hw cq
+ * @cq_notify: notification type
+ */
+static void i40iw_cq_request_notification(struct i40iw_cq_uk *cq,
+					  enum i40iw_completion_notify cq_notify)
+{
+	u64 temp_val;
+	u16 sw_cq_sel;
+	u8 arm_next_se = 0;
+	u8 arm_next = 0;
+	u8 arm_seq_num;
+
+	get_64bit_val(cq->shadow_area, 32, &temp_val);
+	arm_seq_num = (u8)RS_64(temp_val, I40IW_CQ_DBSA_ARM_SEQ_NUM);
+	arm_seq_num++;
+
+	sw_cq_sel = (u16)RS_64(temp_val, I40IW_CQ_DBSA_SW_CQ_SELECT);
+	arm_next_se = (u8)RS_64(temp_val, I40IW_CQ_DBSA_ARM_NEXT_SE);
+	arm_next_se |= 1;
+	if (cq_notify == IW_CQ_COMPL_EVENT)
+		arm_next = 1;
+	temp_val = LS_64(arm_seq_num, I40IW_CQ_DBSA_ARM_SEQ_NUM) |
+	    LS_64(sw_cq_sel, I40IW_CQ_DBSA_SW_CQ_SELECT) |
+	    LS_64(arm_next_se, I40IW_CQ_DBSA_ARM_NEXT_SE) |
+	    LS_64(arm_next, I40IW_CQ_DBSA_ARM_NEXT);
+
+	set_64bit_val(cq->shadow_area, 32, temp_val);
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	writel(cq->cq_id, cq->cqe_alloc_reg);
+}
+
+/**
+ * i40iw_cq_post_entries - update tail in shadow memory
+ * @cq: hw cq
+ * @count: # of entries processed
+ */
+static enum i40iw_status_code i40iw_cq_post_entries(struct i40iw_cq_uk *cq,
+						    u8 count)
+{
+	I40IW_RING_MOVE_TAIL_BY_COUNT(cq->cq_ring, count);
+	set_64bit_val(cq->shadow_area, 0,
+		      I40IW_RING_GETCURRENT_HEAD(cq->cq_ring));
+	return 0;
+}
+
+/**
+ * i40iw_cq_poll_completion - get cq completion info
+ * @cq: hw cq
+ * @info: cq poll information returned
+ * @post_cq: update cq tail
+ */
+static enum i40iw_status_code i40iw_cq_poll_completion(struct i40iw_cq_uk *cq,
+						       struct i40iw_cq_poll_info *info,
+						       bool post_cq)
+{
+	u64 comp_ctx, qword0, qword2, qword3, wqe_qword;
+	u64 *cqe, *sw_wqe;
+	struct i40iw_qp_uk *qp;
+	struct i40iw_ring *pring = NULL;
+	u32 wqe_idx, q_type, array_idx = 0;
+	enum i40iw_status_code ret_code = 0;
+	enum i40iw_status_code ret_code2 = 0;
+	bool move_cq_head = true;
+	u8 polarity;
+	u8 addl_frag_cnt, addl_wqes = 0;
+
+	if (cq->avoid_mem_cflct)
+		cqe = (u64 *)I40IW_GET_CURRENT_EXTENDED_CQ_ELEMENT(cq);
+	else
+		cqe = (u64 *)I40IW_GET_CURRENT_CQ_ELEMENT(cq);
+
+	get_64bit_val(cqe, 24, &qword3);
+	polarity = (u8)RS_64(qword3, I40IW_CQ_VALID);
+
+	if (polarity != cq->polarity)
+		return I40IW_ERR_QUEUE_EMPTY;
+
+	q_type = (u8)RS_64(qword3, I40IW_CQ_SQ);
+	info->error = (bool)RS_64(qword3, I40IW_CQ_ERROR);
+	info->push_dropped = (bool)RS_64(qword3, I40IWCQ_PSHDROP);
+	if (info->error) {
+		info->comp_status = I40IW_COMPL_STATUS_FLUSHED;
+		info->major_err = (bool)RS_64(qword3, I40IW_CQ_MAJERR);
+		info->minor_err = (bool)RS_64(qword3, I40IW_CQ_MINERR);
+	} else {
+		info->comp_status = I40IW_COMPL_STATUS_SUCCESS;
+	}
+
+	get_64bit_val(cqe, 0, &qword0);
+	get_64bit_val(cqe, 16, &qword2);
+
+	info->tcp_seq_num = (u8)RS_64(qword0, I40IWCQ_TCPSEQNUM);
+
+	info->qp_id = (u32)RS_64(qword2, I40IWCQ_QPID);
+
+	get_64bit_val(cqe, 8, &comp_ctx);
+
+	info->solicited_event = (bool)RS_64(qword3, I40IWCQ_SOEVENT);
+	info->is_srq = (bool)RS_64(qword3, I40IWCQ_SRQ);
+
+	qp = (struct i40iw_qp_uk *)(unsigned long)comp_ctx;
+	wqe_idx = (u32)RS_64(qword3, I40IW_CQ_WQEIDX);
+	info->qp_handle = (i40iw_qp_handle)(unsigned long)qp;
+
+	if (q_type == I40IW_CQE_QTYPE_RQ) {
+		array_idx = (wqe_idx * 4) / qp->rq_wqe_size_multiplier;
+		if (info->comp_status == I40IW_COMPL_STATUS_FLUSHED) {
+			info->wr_id = qp->rq_wrid_array[qp->rq_ring.tail];
+			array_idx = qp->rq_ring.tail;
+		} else {
+			info->wr_id = qp->rq_wrid_array[array_idx];
+		}
+
+		info->op_type = I40IW_OP_TYPE_REC;
+		if (qword3 & I40IWCQ_STAG_MASK) {
+			info->stag_invalid_set = true;
+			info->inv_stag = (u32)RS_64(qword2, I40IWCQ_INVSTAG);
+		} else {
+			info->stag_invalid_set = false;
+		}
+		info->bytes_xfered = (u32)RS_64(qword0, I40IWCQ_PAYLDLEN);
+		I40IW_RING_SET_TAIL(qp->rq_ring, array_idx + 1);
+		pring = &qp->rq_ring;
+	} else {
+		if (info->comp_status != I40IW_COMPL_STATUS_FLUSHED) {
+			info->wr_id = qp->sq_wrtrk_array[wqe_idx].wrid;
+			info->bytes_xfered = qp->sq_wrtrk_array[wqe_idx].wr_len;
+
+			info->op_type = (u8)RS_64(qword3, I40IWCQ_OP);
+			sw_wqe = qp->sq_base[wqe_idx].elem;
+			get_64bit_val(sw_wqe, 24, &wqe_qword);
+			addl_frag_cnt =
+			    (u8)RS_64(wqe_qword, I40IWQPSQ_ADDFRAGCNT);
+			i40iw_fragcnt_to_wqesize_sq(addl_frag_cnt + 1, &addl_wqes);
+
+			addl_wqes = (addl_wqes / I40IW_QP_WQE_MIN_SIZE);
+			I40IW_RING_SET_TAIL(qp->sq_ring, (wqe_idx + addl_wqes));
+		} else {
+			do {
+				u8 op_type;
+				u32 tail;
+
+				tail = qp->sq_ring.tail;
+				sw_wqe = qp->sq_base[tail].elem;
+				get_64bit_val(sw_wqe, 24, &wqe_qword);
+				op_type = (u8)RS_64(wqe_qword, I40IWQPSQ_OPCODE);
+				info->op_type = op_type;
+				addl_frag_cnt = (u8)RS_64(wqe_qword, I40IWQPSQ_ADDFRAGCNT);
+				i40iw_fragcnt_to_wqesize_sq(addl_frag_cnt + 1, &addl_wqes);
+				addl_wqes = (addl_wqes / I40IW_QP_WQE_MIN_SIZE);
+				I40IW_RING_SET_TAIL(qp->sq_ring, (tail + addl_wqes));
+				if (op_type != I40IWQP_OP_NOP) {
+					info->wr_id = qp->sq_wrtrk_array[tail].wrid;
+					info->bytes_xfered = qp->sq_wrtrk_array[tail].wr_len;
+					break;
+				}
+			} while (1);
+		}
+		pring = &qp->sq_ring;
+	}
+
+	ret_code = 0;
+
+	if (!ret_code &&
+	    (info->comp_status == I40IW_COMPL_STATUS_FLUSHED))
+		if (pring && (I40IW_RING_MORE_WORK(*pring)))
+			move_cq_head = false;
+
+	if (move_cq_head) {
+		I40IW_RING_MOVE_HEAD(cq->cq_ring, ret_code2);
+
+		if (ret_code2 && !ret_code)
+			ret_code = ret_code2;
+
+		if (I40IW_RING_GETCURRENT_HEAD(cq->cq_ring) == 0)
+			cq->polarity ^= 1;
+
+		if (post_cq) {
+			I40IW_RING_MOVE_TAIL(cq->cq_ring);
+			set_64bit_val(cq->shadow_area, 0,
+				      I40IW_RING_GETCURRENT_HEAD(cq->cq_ring));
+		}
+	} else {
+		if (info->is_srq)
+			return ret_code;
+		qword3 &= ~I40IW_CQ_WQEIDX_MASK;
+		qword3 |= LS_64(pring->tail, I40IW_CQ_WQEIDX);
+		set_64bit_val(cqe, 24, qword3);
+	}
+
+	return ret_code;
+}
+
+/**
+ * i40iw_get_wqe_shift - get shift count for maximum wqe size
+ * @wqdepth: depth of wq required.
+ * @sge: Maximum Scatter Gather Elements wqe
+ * @shift: Returns the shift needed based on sge
+ *
+ * Shift can be used to left shift the wqe size based on sge.
+ * If sge, == 1, shift =0 (wqe_size of 32 bytes), for sge=2 and 3, shift =1
+ * (64 bytes wqes) and 2 otherwise (128 bytes wqe).
+ */
+enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u8 sge, u8 *shift)
+{
+	u32 size;
+
+	*shift = 0;
+	if (sge > 1)
+		*shift = (sge < 4) ? 1 : 2;
+
+	/* check if wqdepth is multiple of 2 or not */
+
+	if ((wqdepth < I40IWQP_SW_MIN_WQSIZE) || (wqdepth & (wqdepth - 1)))
+		return I40IW_ERR_INVALID_SIZE;
+
+	size = wqdepth << *shift;	/* multiple of 32 bytes count */
+	if (size > I40IWQP_SW_MAX_WQSIZE)
+		return I40IW_ERR_INVALID_SIZE;
+	return 0;
+}
+
+static struct i40iw_qp_uk_ops iw_qp_uk_ops = {
+	i40iw_qp_post_wr,
+	i40iw_qp_ring_push_db,
+	i40iw_rdma_write,
+	i40iw_rdma_read,
+	i40iw_send,
+	i40iw_inline_rdma_write,
+	i40iw_inline_send,
+	i40iw_stag_local_invalidate,
+	i40iw_mw_bind,
+	i40iw_post_receive,
+	i40iw_nop
+};
+
+static struct i40iw_cq_ops iw_cq_ops = {
+	i40iw_cq_request_notification,
+	i40iw_cq_poll_completion,
+	i40iw_cq_post_entries,
+	i40iw_clean_cq
+};
+
+static struct i40iw_device_uk_ops iw_device_uk_ops = {
+	i40iw_cq_uk_init,
+	i40iw_qp_uk_init,
+};
+
+/**
+ * i40iw_qp_uk_init - initialize shared qp
+ * @qp: hw qp (user and kernel)
+ * @info: qp initialization info
+ *
+ * initializes the vars used in both user and kernel mode.
+ * size of the wqe depends on numbers of max. fragements
+ * allowed. Then size of wqe * the number of wqes should be the
+ * amount of memory allocated for sq and rq. If srq is used,
+ * then rq_base will point to one rq wqe only (not the whole
+ * array of wqes)
+ */
+enum i40iw_status_code i40iw_qp_uk_init(struct i40iw_qp_uk *qp,
+					struct i40iw_qp_uk_init_info *info)
+{
+	enum i40iw_status_code ret_code = 0;
+	u32 sq_ring_size;
+	u8 sqshift, rqshift;
+
+	if (info->max_sq_frag_cnt > I40IW_MAX_WQ_FRAGMENT_COUNT)
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+
+	if (info->max_rq_frag_cnt > I40IW_MAX_WQ_FRAGMENT_COUNT)
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+	ret_code = i40iw_get_wqe_shift(info->sq_size, info->max_sq_frag_cnt, &sqshift);
+	if (ret_code)
+		return ret_code;
+
+	ret_code = i40iw_get_wqe_shift(info->rq_size, info->max_rq_frag_cnt, &rqshift);
+	if (ret_code)
+		return ret_code;
+
+	qp->sq_base = info->sq;
+	qp->rq_base = info->rq;
+	qp->shadow_area = info->shadow_area;
+	qp->sq_wrtrk_array = info->sq_wrtrk_array;
+	qp->rq_wrid_array = info->rq_wrid_array;
+
+	qp->wqe_alloc_reg = info->wqe_alloc_reg;
+	qp->qp_id = info->qp_id;
+
+	qp->sq_size = info->sq_size;
+	qp->push_db = info->push_db;
+	qp->push_wqe = info->push_wqe;
+
+	qp->max_sq_frag_cnt = info->max_sq_frag_cnt;
+	sq_ring_size = qp->sq_size << sqshift;
+
+	I40IW_RING_INIT(qp->sq_ring, sq_ring_size);
+	I40IW_RING_INIT(qp->initial_ring, sq_ring_size);
+	I40IW_RING_MOVE_HEAD(qp->sq_ring, ret_code);
+	I40IW_RING_MOVE_TAIL(qp->sq_ring);
+	I40IW_RING_MOVE_HEAD(qp->initial_ring, ret_code);
+	qp->swqe_polarity = 1;
+	qp->swqe_polarity_deferred = 1;
+	qp->rwqe_polarity = 0;
+
+	if (!qp->use_srq) {
+		qp->rq_size = info->rq_size;
+		qp->max_rq_frag_cnt = info->max_rq_frag_cnt;
+		qp->rq_wqe_size = rqshift;
+		I40IW_RING_INIT(qp->rq_ring, qp->rq_size);
+		qp->rq_wqe_size_multiplier = 4 << rqshift;
+	}
+	qp->ops = iw_qp_uk_ops;
+
+	return ret_code;
+}
+
+/**
+ * i40iw_cq_uk_init - initialize shared cq (user and kernel)
+ * @cq: hw cq
+ * @info: hw cq initialization info
+ */
+enum i40iw_status_code i40iw_cq_uk_init(struct i40iw_cq_uk *cq,
+					struct i40iw_cq_uk_init_info *info)
+{
+	if ((info->cq_size < I40IW_MIN_CQ_SIZE) ||
+	    (info->cq_size > I40IW_MAX_CQ_SIZE))
+		return I40IW_ERR_INVALID_SIZE;
+	cq->cq_base = (struct i40iw_cqe *)info->cq_base;
+	cq->cq_id = info->cq_id;
+	cq->cq_size = info->cq_size;
+	cq->cqe_alloc_reg = info->cqe_alloc_reg;
+	cq->shadow_area = info->shadow_area;
+	cq->avoid_mem_cflct = info->avoid_mem_cflct;
+
+	I40IW_RING_INIT(cq->cq_ring, cq->cq_size);
+	cq->polarity = 1;
+	cq->ops = iw_cq_ops;
+
+	return 0;
+}
+
+/**
+ * i40iw_device_init_uk - setup routines for iwarp shared device
+ * @dev: iwarp shared (user and kernel)
+ */
+void i40iw_device_init_uk(struct i40iw_dev_uk *dev)
+{
+	dev->ops_uk = iw_device_uk_ops;
+}
+
+/**
+ * i40iw_clean_cq - clean cq entries
+ * @ queue completion context
+ * @cq: cq to clean
+ */
+void i40iw_clean_cq(void *queue, struct i40iw_cq_uk *cq)
+{
+	u64 *cqe;
+	u64 qword3, comp_ctx;
+	u32 cq_head;
+	u8 polarity, temp;
+
+	cq_head = cq->cq_ring.head;
+	temp = cq->polarity;
+	do {
+		if (cq->avoid_mem_cflct)
+			cqe = (u64 *)&(((struct i40iw_extended_cqe *)cq->cq_base)[cq_head]);
+		else
+			cqe = (u64 *)&cq->cq_base[cq_head];
+		get_64bit_val(cqe, 24, &qword3);
+		polarity = (u8)RS_64(qword3, I40IW_CQ_VALID);
+
+		if (polarity != temp)
+			break;
+
+		get_64bit_val(cqe, 8, &comp_ctx);
+		if ((void *)(unsigned long)comp_ctx == queue)
+			set_64bit_val(cqe, 8, 0);
+
+		cq_head = (cq_head + 1) % cq->cq_ring.size;
+		if (!cq_head)
+			temp ^= 1;
+	} while (true);
+}
+
+/**
+ * i40iw_nop - send a nop
+ * @qp: hw qp ptr
+ * @wr_id: work request id
+ * @signaled: flag if signaled for completion
+ * @post_sq: flag to post sq
+ */
+enum i40iw_status_code i40iw_nop(struct i40iw_qp_uk *qp,
+				 u64 wr_id,
+				 bool signaled,
+				 bool post_sq)
+{
+	u64 header, *wqe;
+	u32 wqe_idx;
+
+	wqe = i40iw_qp_get_next_send_wqe(qp, &wqe_idx, I40IW_QP_WQE_MIN_SIZE);
+	if (!wqe)
+		return I40IW_ERR_QP_TOOMANY_WRS_POSTED;
+
+	qp->sq_wrtrk_array[wqe_idx].wrid = wr_id;
+	qp->sq_wrtrk_array[wqe_idx].wr_len = 0;
+	set_64bit_val(wqe, 0, 0);
+	set_64bit_val(wqe, 8, 0);
+	set_64bit_val(wqe, 16, 0);
+
+	header = LS_64(I40IWQP_OP_NOP, I40IWQPSQ_OPCODE) |
+	    LS_64(signaled, I40IWQPSQ_SIGCOMPL) |
+	    LS_64(qp->swqe_polarity, I40IWQPSQ_VALID);
+
+	wmb(); /* make sure WQE is populated before valid bit is set */
+
+	set_64bit_val(wqe, 24, header);
+	if (post_sq)
+		i40iw_qp_post_wr(qp);
+
+	return 0;
+}
+
+/**
+ * i40iw_fragcnt_to_wqesize_sq - calculate wqe size based on fragment count for SQ
+ * @frag_cnt: number of fragments
+ * @wqe_size: size of sq wqe returned
+ */
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u8 frag_cnt, u8 *wqe_size)
+{
+	switch (frag_cnt) {
+	case 0:
+	case 1:
+		*wqe_size = I40IW_QP_WQE_MIN_SIZE;
+		break;
+	case 2:
+	case 3:
+		*wqe_size = 64;
+		break;
+	case 4:
+	case 5:
+		*wqe_size = 96;
+		break;
+	case 6:
+	case 7:
+		*wqe_size = 128;
+		break;
+	default:
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_fragcnt_to_wqesize_rq - calculate wqe size based on fragment count for RQ
+ * @frag_cnt: number of fragments
+ * @wqe_size: size of rq wqe returned
+ */
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u8 frag_cnt, u8 *wqe_size)
+{
+	switch (frag_cnt) {
+	case 0:
+	case 1:
+		*wqe_size = 32;
+		break;
+	case 2:
+	case 3:
+		*wqe_size = 64;
+		break;
+	case 4:
+	case 5:
+	case 6:
+	case 7:
+		*wqe_size = 128;
+		break;
+	default:
+		return I40IW_ERR_INVALID_FRAG_COUNT;
+	}
+
+	return 0;
+}
+
+/**
+ * i40iw_inline_data_size_to_wqesize - based on inline data, wqe size
+ * @data_size: data size for inline
+ * @wqe_size: size of sq wqe returned
+ */
+enum i40iw_status_code i40iw_inline_data_size_to_wqesize(u32 data_size,
+							 u8 *wqe_size)
+{
+	if (data_size > I40IW_MAX_INLINE_DATA_SIZE)
+		return I40IW_ERR_INVALID_IMM_DATA_SIZE;
+
+	if (data_size <= 16)
+		*wqe_size = I40IW_QP_WQE_MIN_SIZE;
+	else if (data_size <= 48)
+		*wqe_size = 64;
+	else if (data_size <= 80)
+		*wqe_size = 96;
+	else
+		*wqe_size = 128;
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_user.h b/drivers/infiniband/hw/i40iw/i40iw_user.h
new file mode 100644
index 000000000000..5cd971bb8cc7
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_user.h
@@ -0,0 +1,442 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_USER_H
+#define I40IW_USER_H
+
+enum i40iw_device_capabilities_const {
+	I40IW_WQE_SIZE =			4,
+	I40IW_CQP_WQE_SIZE =			8,
+	I40IW_CQE_SIZE =			4,
+	I40IW_EXTENDED_CQE_SIZE =		8,
+	I40IW_AEQE_SIZE =			2,
+	I40IW_CEQE_SIZE =			1,
+	I40IW_CQP_CTX_SIZE =			8,
+	I40IW_SHADOW_AREA_SIZE =		8,
+	I40IW_CEQ_MAX_COUNT =			256,
+	I40IW_QUERY_FPM_BUF_SIZE =		128,
+	I40IW_COMMIT_FPM_BUF_SIZE =		128,
+	I40IW_MIN_IW_QP_ID =			1,
+	I40IW_MAX_IW_QP_ID =			262143,
+	I40IW_MIN_CEQID =			0,
+	I40IW_MAX_CEQID =			256,
+	I40IW_MIN_CQID =			0,
+	I40IW_MAX_CQID =			131071,
+	I40IW_MIN_AEQ_ENTRIES =			1,
+	I40IW_MAX_AEQ_ENTRIES =			524287,
+	I40IW_MIN_CEQ_ENTRIES =			1,
+	I40IW_MAX_CEQ_ENTRIES =			131071,
+	I40IW_MIN_CQ_SIZE =			1,
+	I40IW_MAX_CQ_SIZE =			1048575,
+	I40IW_MAX_AEQ_ALLOCATE_COUNT =		255,
+	I40IW_DB_ID_ZERO =			0,
+	I40IW_MAX_WQ_FRAGMENT_COUNT =		6,
+	I40IW_MAX_SGE_RD =			1,
+	I40IW_MAX_OUTBOUND_MESSAGE_SIZE =	2147483647,
+	I40IW_MAX_INBOUND_MESSAGE_SIZE =	2147483647,
+	I40IW_MAX_PUSH_PAGE_COUNT =		4096,
+	I40IW_MAX_PE_ENABLED_VF_COUNT =		32,
+	I40IW_MAX_VF_FPM_ID =			47,
+	I40IW_MAX_VF_PER_PF =			127,
+	I40IW_MAX_SQ_PAYLOAD_SIZE =		2145386496,
+	I40IW_MAX_INLINE_DATA_SIZE =		112,
+	I40IW_MAX_PUSHMODE_INLINE_DATA_SIZE =	112,
+	I40IW_MAX_IRD_SIZE =			32,
+	I40IW_QPCTX_ENCD_MAXIRD =		3,
+	I40IW_MAX_WQ_ENTRIES =			2048,
+	I40IW_MAX_ORD_SIZE =			32,
+	I40IW_Q2_BUFFER_SIZE =			(248 + 100),
+	I40IW_QP_CTX_SIZE =			248
+};
+
+#define i40iw_handle void *
+#define i40iw_adapter_handle i40iw_handle
+#define i40iw_qp_handle i40iw_handle
+#define i40iw_cq_handle i40iw_handle
+#define i40iw_srq_handle i40iw_handle
+#define i40iw_pd_id i40iw_handle
+#define i40iw_stag_handle i40iw_handle
+#define i40iw_stag_index u32
+#define i40iw_stag u32
+#define i40iw_stag_key u8
+
+#define i40iw_tagged_offset u64
+#define i40iw_access_privileges u32
+#define i40iw_physical_fragment u64
+#define i40iw_address_list u64 *
+
+#define I40IW_CREATE_STAG(index, key)       (((index) << 8) + (key))
+
+#define I40IW_STAG_KEY_FROM_STAG(stag)      ((stag) && 0x000000FF)
+
+#define I40IW_STAG_INDEX_FROM_STAG(stag)    (((stag) && 0xFFFFFF00) >> 8)
+
+struct i40iw_qp_uk;
+struct i40iw_cq_uk;
+struct i40iw_srq_uk;
+struct i40iw_qp_uk_init_info;
+struct i40iw_cq_uk_init_info;
+struct i40iw_srq_uk_init_info;
+
+struct i40iw_sge {
+	i40iw_tagged_offset tag_off;
+	u32 len;
+	i40iw_stag stag;
+};
+
+#define i40iw_sgl struct i40iw_sge *
+
+struct i40iw_ring {
+	u32 head;
+	u32 tail;
+	u32 size;
+};
+
+struct i40iw_cqe {
+	u64 buf[I40IW_CQE_SIZE];
+};
+
+struct i40iw_extended_cqe {
+	u64 buf[I40IW_EXTENDED_CQE_SIZE];
+};
+
+struct i40iw_wqe {
+	u64 buf[I40IW_WQE_SIZE];
+};
+
+struct i40iw_qp_uk_ops;
+
+enum i40iw_addressing_type {
+	I40IW_ADDR_TYPE_ZERO_BASED = 0,
+	I40IW_ADDR_TYPE_VA_BASED = 1,
+};
+
+#define I40IW_ACCESS_FLAGS_LOCALREAD		0x01
+#define I40IW_ACCESS_FLAGS_LOCALWRITE		0x02
+#define I40IW_ACCESS_FLAGS_REMOTEREAD_ONLY	0x04
+#define I40IW_ACCESS_FLAGS_REMOTEREAD		0x05
+#define I40IW_ACCESS_FLAGS_REMOTEWRITE_ONLY	0x08
+#define I40IW_ACCESS_FLAGS_REMOTEWRITE		0x0a
+#define I40IW_ACCESS_FLAGS_BIND_WINDOW		0x10
+#define I40IW_ACCESS_FLAGS_ALL			0x1F
+
+#define I40IW_OP_TYPE_RDMA_WRITE	0
+#define I40IW_OP_TYPE_RDMA_READ		1
+#define I40IW_OP_TYPE_SEND		3
+#define I40IW_OP_TYPE_SEND_INV		4
+#define I40IW_OP_TYPE_SEND_SOL		5
+#define I40IW_OP_TYPE_SEND_SOL_INV	6
+#define I40IW_OP_TYPE_REC		7
+#define I40IW_OP_TYPE_BIND_MW		8
+#define I40IW_OP_TYPE_FAST_REG_NSMR	9
+#define I40IW_OP_TYPE_INV_STAG		10
+#define I40IW_OP_TYPE_RDMA_READ_INV_STAG 11
+#define I40IW_OP_TYPE_NOP		12
+
+enum i40iw_completion_status {
+	I40IW_COMPL_STATUS_SUCCESS = 0,
+	I40IW_COMPL_STATUS_FLUSHED,
+	I40IW_COMPL_STATUS_INVALID_WQE,
+	I40IW_COMPL_STATUS_QP_CATASTROPHIC,
+	I40IW_COMPL_STATUS_REMOTE_TERMINATION,
+	I40IW_COMPL_STATUS_INVALID_STAG,
+	I40IW_COMPL_STATUS_BASE_BOUND_VIOLATION,
+	I40IW_COMPL_STATUS_ACCESS_VIOLATION,
+	I40IW_COMPL_STATUS_INVALID_PD_ID,
+	I40IW_COMPL_STATUS_WRAP_ERROR,
+	I40IW_COMPL_STATUS_STAG_INVALID_PDID,
+	I40IW_COMPL_STATUS_RDMA_READ_ZERO_ORD,
+	I40IW_COMPL_STATUS_QP_NOT_PRIVLEDGED,
+	I40IW_COMPL_STATUS_STAG_NOT_INVALID,
+	I40IW_COMPL_STATUS_INVALID_PHYS_BUFFER_SIZE,
+	I40IW_COMPL_STATUS_INVALID_PHYS_BUFFER_ENTRY,
+	I40IW_COMPL_STATUS_INVALID_FBO,
+	I40IW_COMPL_STATUS_INVALID_LENGTH,
+	I40IW_COMPL_STATUS_INVALID_ACCESS,
+	I40IW_COMPL_STATUS_PHYS_BUFFER_LIST_TOO_LONG,
+	I40IW_COMPL_STATUS_INVALID_VIRT_ADDRESS,
+	I40IW_COMPL_STATUS_INVALID_REGION,
+	I40IW_COMPL_STATUS_INVALID_WINDOW,
+	I40IW_COMPL_STATUS_INVALID_TOTAL_LENGTH
+};
+
+enum i40iw_completion_notify {
+	IW_CQ_COMPL_EVENT = 0,
+	IW_CQ_COMPL_SOLICITED = 1
+};
+
+struct i40iw_post_send {
+	i40iw_sgl sg_list;
+	u8 num_sges;
+};
+
+struct i40iw_post_inline_send {
+	void *data;
+	u32 len;
+};
+
+struct i40iw_post_send_w_inv {
+	i40iw_sgl sg_list;
+	u32 num_sges;
+	i40iw_stag remote_stag_to_inv;
+};
+
+struct i40iw_post_inline_send_w_inv {
+	void *data;
+	u32 len;
+	i40iw_stag remote_stag_to_inv;
+};
+
+struct i40iw_rdma_write {
+	i40iw_sgl lo_sg_list;
+	u8 num_lo_sges;
+	struct i40iw_sge rem_addr;
+};
+
+struct i40iw_inline_rdma_write {
+	void *data;
+	u32 len;
+	struct i40iw_sge rem_addr;
+};
+
+struct i40iw_rdma_read {
+	struct i40iw_sge lo_addr;
+	struct i40iw_sge rem_addr;
+};
+
+struct i40iw_bind_window {
+	i40iw_stag mr_stag;
+	u64 bind_length;
+	void *va;
+	enum i40iw_addressing_type addressing_type;
+	bool enable_reads;
+	bool enable_writes;
+	i40iw_stag mw_stag;
+};
+
+struct i40iw_inv_local_stag {
+	i40iw_stag target_stag;
+};
+
+struct i40iw_post_sq_info {
+	u64 wr_id;
+	u8 op_type;
+	bool signaled;
+	bool read_fence;
+	bool local_fence;
+	bool inline_data;
+	bool defer_flag;
+	union {
+		struct i40iw_post_send send;
+		struct i40iw_post_send send_w_sol;
+		struct i40iw_post_send_w_inv send_w_inv;
+		struct i40iw_post_send_w_inv send_w_sol_inv;
+		struct i40iw_rdma_write rdma_write;
+		struct i40iw_rdma_read rdma_read;
+		struct i40iw_rdma_read rdma_read_inv;
+		struct i40iw_bind_window bind_window;
+		struct i40iw_inv_local_stag inv_local_stag;
+		struct i40iw_inline_rdma_write inline_rdma_write;
+		struct i40iw_post_inline_send inline_send;
+		struct i40iw_post_inline_send inline_send_w_sol;
+		struct i40iw_post_inline_send_w_inv inline_send_w_inv;
+		struct i40iw_post_inline_send_w_inv inline_send_w_sol_inv;
+	} op;
+};
+
+struct i40iw_post_rq_info {
+	u64 wr_id;
+	i40iw_sgl sg_list;
+	u32 num_sges;
+};
+
+struct i40iw_cq_poll_info {
+	u64 wr_id;
+	i40iw_qp_handle qp_handle;
+	u32 bytes_xfered;
+	u32 tcp_seq_num;
+	u32 qp_id;
+	i40iw_stag inv_stag;
+	enum i40iw_completion_status comp_status;
+	u16 major_err;
+	u16 minor_err;
+	u8 op_type;
+	bool stag_invalid_set;
+	bool push_dropped;
+	bool error;
+	bool is_srq;
+	bool solicited_event;
+};
+
+struct i40iw_qp_uk_ops {
+	void (*iw_qp_post_wr)(struct i40iw_qp_uk *);
+	void (*iw_qp_ring_push_db)(struct i40iw_qp_uk *, u32);
+	enum i40iw_status_code (*iw_rdma_write)(struct i40iw_qp_uk *,
+						struct i40iw_post_sq_info *, bool);
+	enum i40iw_status_code (*iw_rdma_read)(struct i40iw_qp_uk *,
+					       struct i40iw_post_sq_info *, bool, bool);
+	enum i40iw_status_code (*iw_send)(struct i40iw_qp_uk *,
+					  struct i40iw_post_sq_info *, u32, bool);
+	enum i40iw_status_code (*iw_inline_rdma_write)(struct i40iw_qp_uk *,
+						       struct i40iw_post_sq_info *, bool);
+	enum i40iw_status_code (*iw_inline_send)(struct i40iw_qp_uk *,
+						 struct i40iw_post_sq_info *, u32, bool);
+	enum i40iw_status_code (*iw_stag_local_invalidate)(struct i40iw_qp_uk *,
+							   struct i40iw_post_sq_info *, bool);
+	enum i40iw_status_code (*iw_mw_bind)(struct i40iw_qp_uk *,
+					     struct i40iw_post_sq_info *, bool);
+	enum i40iw_status_code (*iw_post_receive)(struct i40iw_qp_uk *,
+						  struct i40iw_post_rq_info *);
+	enum i40iw_status_code (*iw_post_nop)(struct i40iw_qp_uk *, u64, bool, bool);
+};
+
+struct i40iw_cq_ops {
+	void (*iw_cq_request_notification)(struct i40iw_cq_uk *,
+					   enum i40iw_completion_notify);
+	enum i40iw_status_code (*iw_cq_poll_completion)(struct i40iw_cq_uk *,
+							struct i40iw_cq_poll_info *, bool);
+	enum i40iw_status_code (*iw_cq_post_entries)(struct i40iw_cq_uk *, u8 count);
+	void (*iw_cq_clean)(void *, struct i40iw_cq_uk *);
+};
+
+struct i40iw_dev_uk;
+
+struct i40iw_device_uk_ops {
+	enum i40iw_status_code (*iwarp_cq_uk_init)(struct i40iw_cq_uk *,
+						   struct i40iw_cq_uk_init_info *);
+	enum i40iw_status_code (*iwarp_qp_uk_init)(struct i40iw_qp_uk *,
+						   struct i40iw_qp_uk_init_info *);
+};
+
+struct i40iw_dev_uk {
+	struct i40iw_device_uk_ops ops_uk;
+};
+
+struct i40iw_sq_uk_wr_trk_info {
+	u64 wrid;
+	u64 wr_len;
+};
+
+struct i40iw_qp_quanta {
+	u64 elem[I40IW_WQE_SIZE];
+};
+
+struct i40iw_qp_uk {
+	struct i40iw_qp_quanta *sq_base;
+	struct i40iw_qp_quanta *rq_base;
+	u32 __iomem *wqe_alloc_reg;
+	struct i40iw_sq_uk_wr_trk_info *sq_wrtrk_array;
+	u64 *rq_wrid_array;
+	u64 *shadow_area;
+	u32 *push_db;
+	u64 *push_wqe;
+	struct i40iw_ring sq_ring;
+	struct i40iw_ring rq_ring;
+	struct i40iw_ring initial_ring;
+	u32 qp_id;
+	u32 sq_size;
+	u32 rq_size;
+	struct i40iw_qp_uk_ops ops;
+	bool use_srq;
+	u8 swqe_polarity;
+	u8 swqe_polarity_deferred;
+	u8 rwqe_polarity;
+	u8 rq_wqe_size;
+	u8 rq_wqe_size_multiplier;
+	u8 max_sq_frag_cnt;
+	u8 max_rq_frag_cnt;
+	bool deferred_flag;
+};
+
+struct i40iw_cq_uk {
+	struct i40iw_cqe *cq_base;
+	u32 __iomem *cqe_alloc_reg;
+	u64 *shadow_area;
+	u32 cq_id;
+	u32 cq_size;
+	struct i40iw_ring cq_ring;
+	u8 polarity;
+	bool avoid_mem_cflct;
+
+	struct i40iw_cq_ops ops;
+};
+
+struct i40iw_qp_uk_init_info {
+	struct i40iw_qp_quanta *sq;
+	struct i40iw_qp_quanta *rq;
+	u32 __iomem *wqe_alloc_reg;
+	u64 *shadow_area;
+	struct i40iw_sq_uk_wr_trk_info *sq_wrtrk_array;
+	u64 *rq_wrid_array;
+	u32 *push_db;
+	u64 *push_wqe;
+	u32 qp_id;
+	u32 sq_size;
+	u32 rq_size;
+	u8 max_sq_frag_cnt;
+	u8 max_rq_frag_cnt;
+
+};
+
+struct i40iw_cq_uk_init_info {
+	u32 __iomem *cqe_alloc_reg;
+	struct i40iw_cqe *cq_base;
+	u64 *shadow_area;
+	u32 cq_size;
+	u32 cq_id;
+	bool avoid_mem_cflct;
+};
+
+void i40iw_device_init_uk(struct i40iw_dev_uk *dev);
+
+void i40iw_qp_post_wr(struct i40iw_qp_uk *qp);
+u64 *i40iw_qp_get_next_send_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx,
+				u8 wqe_size);
+u64 *i40iw_qp_get_next_recv_wqe(struct i40iw_qp_uk *qp, u32 *wqe_idx);
+u64 *i40iw_qp_get_next_srq_wqe(struct i40iw_srq_uk *srq, u32 *wqe_idx);
+
+enum i40iw_status_code i40iw_cq_uk_init(struct i40iw_cq_uk *cq,
+					struct i40iw_cq_uk_init_info *info);
+enum i40iw_status_code i40iw_qp_uk_init(struct i40iw_qp_uk *qp,
+					struct i40iw_qp_uk_init_info *info);
+
+void i40iw_clean_cq(void *queue, struct i40iw_cq_uk *cq);
+enum i40iw_status_code i40iw_nop(struct i40iw_qp_uk *qp, u64 wr_id,
+				 bool signaled, bool post_sq);
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_sq(u8 frag_cnt, u8 *wqe_size);
+enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u8 frag_cnt, u8 *wqe_size);
+enum i40iw_status_code i40iw_inline_data_size_to_wqesize(u32 data_size,
+							 u8 *wqe_size);
+enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u8 sge, u8 *shift);
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_utils.c b/drivers/infiniband/hw/i40iw/i40iw_utils.c
new file mode 100644
index 000000000000..1ceec81bd8eb
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_utils.c
@@ -0,0 +1,1270 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/if_vlan.h>
+#include <linux/crc32.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <asm/irq.h>
+#include <asm/byteorder.h>
+#include <net/netevent.h>
+#include <net/neighbour.h>
+#include "i40iw.h"
+
+/**
+ * i40iw_arp_table - manage arp table
+ * @iwdev: iwarp device
+ * @ip_addr: ip address for device
+ * @mac_addr: mac address ptr
+ * @action: modify, delete or add
+ */
+int i40iw_arp_table(struct i40iw_device *iwdev,
+		    __be32 *ip_addr,
+		    bool ipv4,
+		    u8 *mac_addr,
+		    u32 action)
+{
+	int arp_index;
+	int err;
+	u32 ip[4];
+
+	if (ipv4) {
+		memset(ip, 0, sizeof(ip));
+		ip[0] = *ip_addr;
+	} else {
+		memcpy(ip, ip_addr, sizeof(ip));
+	}
+
+	for (arp_index = 0; (u32)arp_index < iwdev->arp_table_size; arp_index++)
+		if (memcmp(iwdev->arp_table[arp_index].ip_addr, ip, sizeof(ip)) == 0)
+			break;
+	switch (action) {
+	case I40IW_ARP_ADD:
+		if (arp_index != iwdev->arp_table_size)
+			return -1;
+
+		arp_index = 0;
+		err = i40iw_alloc_resource(iwdev, iwdev->allocated_arps,
+					   iwdev->arp_table_size,
+					   (u32 *)&arp_index,
+					   &iwdev->next_arp_index);
+
+		if (err)
+			return err;
+
+		memcpy(iwdev->arp_table[arp_index].ip_addr, ip, sizeof(ip));
+		ether_addr_copy(iwdev->arp_table[arp_index].mac_addr, mac_addr);
+		break;
+	case I40IW_ARP_RESOLVE:
+		if (arp_index == iwdev->arp_table_size)
+			return -1;
+		break;
+	case I40IW_ARP_DELETE:
+		if (arp_index == iwdev->arp_table_size)
+			return -1;
+		memset(iwdev->arp_table[arp_index].ip_addr, 0,
+		       sizeof(iwdev->arp_table[arp_index].ip_addr));
+		eth_zero_addr(iwdev->arp_table[arp_index].mac_addr);
+		i40iw_free_resource(iwdev, iwdev->allocated_arps, arp_index);
+		break;
+	default:
+		return -1;
+	}
+	return arp_index;
+}
+
+/**
+ * i40iw_wr32 - write 32 bits to hw register
+ * @hw: hardware information including registers
+ * @reg: register offset
+ * @value: vvalue to write to register
+ */
+inline void i40iw_wr32(struct i40iw_hw *hw, u32 reg, u32 value)
+{
+	writel(value, hw->hw_addr + reg);
+}
+
+/**
+ * i40iw_rd32 - read a 32 bit hw register
+ * @hw: hardware information including registers
+ * @reg: register offset
+ *
+ * Return value of register content
+ */
+inline u32 i40iw_rd32(struct i40iw_hw *hw, u32 reg)
+{
+	return readl(hw->hw_addr + reg);
+}
+
+/**
+ * i40iw_inetaddr_event - system notifier for netdev events
+ * @notfier: not used
+ * @event: event for notifier
+ * @ptr: if address
+ */
+int i40iw_inetaddr_event(struct notifier_block *notifier,
+			 unsigned long event,
+			 void *ptr)
+{
+	struct in_ifaddr *ifa = ptr;
+	struct net_device *event_netdev = ifa->ifa_dev->dev;
+	struct net_device *netdev;
+	struct net_device *upper_dev;
+	struct i40iw_device *iwdev;
+	struct i40iw_handler *hdl;
+	__be32 local_ipaddr;
+
+	hdl = i40iw_find_netdev(event_netdev);
+	if (!hdl)
+		return NOTIFY_DONE;
+
+	iwdev = &hdl->device;
+	netdev = iwdev->ldev->netdev;
+	upper_dev = netdev_master_upper_dev_get(netdev);
+	if (netdev != event_netdev)
+		return NOTIFY_DONE;
+
+	switch (event) {
+	case NETDEV_DOWN:
+		if (upper_dev)
+			local_ipaddr =
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+		else
+			local_ipaddr = ifa->ifa_address;
+		local_ipaddr = ntohl(local_ipaddr);
+		i40iw_manage_arp_cache(iwdev,
+				       netdev->dev_addr,
+				       &local_ipaddr,
+				       true,
+				       I40IW_ARP_DELETE);
+		return NOTIFY_OK;
+	case NETDEV_UP:
+		if (upper_dev)
+			local_ipaddr =
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+		else
+			local_ipaddr = ifa->ifa_address;
+		local_ipaddr = ntohl(local_ipaddr);
+		i40iw_manage_arp_cache(iwdev,
+				       netdev->dev_addr,
+				       &local_ipaddr,
+				       true,
+				       I40IW_ARP_ADD);
+		break;
+	case NETDEV_CHANGEADDR:
+		/* Add the address to the IP table */
+		if (upper_dev)
+			local_ipaddr =
+				((struct in_device *)upper_dev->ip_ptr)->ifa_list->ifa_address;
+		else
+			local_ipaddr = ifa->ifa_address;
+
+		local_ipaddr = ntohl(local_ipaddr);
+		i40iw_manage_arp_cache(iwdev,
+				       netdev->dev_addr,
+				       &local_ipaddr,
+				       true,
+				       I40IW_ARP_ADD);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_DONE;
+}
+
+/**
+ * i40iw_inet6addr_event - system notifier for ipv6 netdev events
+ * @notfier: not used
+ * @event: event for notifier
+ * @ptr: if address
+ */
+int i40iw_inet6addr_event(struct notifier_block *notifier,
+			  unsigned long event,
+			  void *ptr)
+{
+	struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr;
+	struct net_device *event_netdev = ifa->idev->dev;
+	struct net_device *netdev;
+	struct i40iw_device *iwdev;
+	struct i40iw_handler *hdl;
+	__be32 local_ipaddr6[4];
+
+	hdl = i40iw_find_netdev(event_netdev);
+	if (!hdl)
+		return NOTIFY_DONE;
+
+	iwdev = &hdl->device;
+	netdev = iwdev->ldev->netdev;
+	if (netdev != event_netdev)
+		return NOTIFY_DONE;
+
+	switch (event) {
+	case NETDEV_DOWN:
+		i40iw_copy_ip_ntohl(local_ipaddr6, ifa->addr.in6_u.u6_addr32);
+		i40iw_manage_arp_cache(iwdev,
+				       netdev->dev_addr,
+				       local_ipaddr6,
+				       false,
+				       I40IW_ARP_DELETE);
+		return NOTIFY_OK;
+	case NETDEV_UP:
+		/* Fall through */
+	case NETDEV_CHANGEADDR:
+		i40iw_copy_ip_ntohl(local_ipaddr6, ifa->addr.in6_u.u6_addr32);
+		i40iw_manage_arp_cache(iwdev,
+				       netdev->dev_addr,
+				       local_ipaddr6,
+				       false,
+				       I40IW_ARP_ADD);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_DONE;
+}
+
+/**
+ * i40iw_net_event - system notifier for net events
+ * @notfier: not used
+ * @event: event for notifier
+ * @ptr: neighbor
+ */
+int i40iw_net_event(struct notifier_block *notifier, unsigned long event, void *ptr)
+{
+	struct neighbour *neigh = ptr;
+	struct i40iw_device *iwdev;
+	struct i40iw_handler *iwhdl;
+	__be32 *p;
+	u32 local_ipaddr[4];
+
+	switch (event) {
+	case NETEVENT_NEIGH_UPDATE:
+		iwhdl = i40iw_find_netdev((struct net_device *)neigh->dev);
+		if (!iwhdl)
+			return NOTIFY_DONE;
+		iwdev = &iwhdl->device;
+		p = (__be32 *)neigh->primary_key;
+		i40iw_copy_ip_ntohl(local_ipaddr, p);
+		if (neigh->nud_state & NUD_VALID) {
+			i40iw_manage_arp_cache(iwdev,
+					       neigh->ha,
+					       local_ipaddr,
+					       false,
+					       I40IW_ARP_ADD);
+
+		} else {
+			i40iw_manage_arp_cache(iwdev,
+					       neigh->ha,
+					       local_ipaddr,
+					       false,
+					       I40IW_ARP_DELETE);
+		}
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_DONE;
+}
+
+/**
+ * i40iw_get_cqp_request - get cqp struct
+ * @cqp: device cqp ptr
+ * @wait: cqp to be used in wait mode
+ */
+struct i40iw_cqp_request *i40iw_get_cqp_request(struct i40iw_cqp *cqp, bool wait)
+{
+	struct i40iw_cqp_request *cqp_request = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cqp->req_lock, flags);
+	if (!list_empty(&cqp->cqp_avail_reqs)) {
+		cqp_request = list_entry(cqp->cqp_avail_reqs.next,
+					 struct i40iw_cqp_request, list);
+		list_del_init(&cqp_request->list);
+	}
+	spin_unlock_irqrestore(&cqp->req_lock, flags);
+	if (!cqp_request) {
+		cqp_request = kzalloc(sizeof(*cqp_request), GFP_ATOMIC);
+		if (cqp_request) {
+			cqp_request->dynamic = true;
+			INIT_LIST_HEAD(&cqp_request->list);
+			init_waitqueue_head(&cqp_request->waitq);
+		}
+	}
+	if (!cqp_request) {
+		i40iw_pr_err("CQP Request Fail: No Memory");
+		return NULL;
+	}
+
+	if (wait) {
+		atomic_set(&cqp_request->refcount, 2);
+		cqp_request->waiting = true;
+	} else {
+		atomic_set(&cqp_request->refcount, 1);
+	}
+	return cqp_request;
+}
+
+/**
+ * i40iw_free_cqp_request - free cqp request
+ * @cqp: cqp ptr
+ * @cqp_request: to be put back in cqp list
+ */
+void i40iw_free_cqp_request(struct i40iw_cqp *cqp, struct i40iw_cqp_request *cqp_request)
+{
+	unsigned long flags;
+
+	if (cqp_request->dynamic) {
+		kfree(cqp_request);
+	} else {
+		cqp_request->request_done = false;
+		cqp_request->callback_fcn = NULL;
+		cqp_request->waiting = false;
+
+		spin_lock_irqsave(&cqp->req_lock, flags);
+		list_add_tail(&cqp_request->list, &cqp->cqp_avail_reqs);
+		spin_unlock_irqrestore(&cqp->req_lock, flags);
+	}
+}
+
+/**
+ * i40iw_put_cqp_request - dec ref count and free if 0
+ * @cqp: cqp ptr
+ * @cqp_request: to be put back in cqp list
+ */
+void i40iw_put_cqp_request(struct i40iw_cqp *cqp,
+			   struct i40iw_cqp_request *cqp_request)
+{
+	if (atomic_dec_and_test(&cqp_request->refcount))
+		i40iw_free_cqp_request(cqp, cqp_request);
+}
+
+/**
+ * i40iw_free_qp - callback after destroy cqp completes
+ * @cqp_request: cqp request for destroy qp
+ * @num: not used
+ */
+static void i40iw_free_qp(struct i40iw_cqp_request *cqp_request, u32 num)
+{
+	struct i40iw_sc_qp *qp = (struct i40iw_sc_qp *)cqp_request->param;
+	struct i40iw_qp *iwqp = (struct i40iw_qp *)qp->back_qp;
+	struct i40iw_device *iwdev;
+	u32 qp_num = iwqp->ibqp.qp_num;
+
+	iwdev = iwqp->iwdev;
+
+	i40iw_rem_pdusecount(iwqp->iwpd, iwdev);
+	i40iw_free_qp_resources(iwdev, iwqp, qp_num);
+}
+
+/**
+ * i40iw_wait_event - wait for completion
+ * @iwdev: iwarp device
+ * @cqp_request: cqp request to wait
+ */
+static int i40iw_wait_event(struct i40iw_device *iwdev,
+			    struct i40iw_cqp_request *cqp_request)
+{
+	struct cqp_commands_info *info = &cqp_request->info;
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	bool cqp_error = false;
+	int err_code = 0;
+	int timeout_ret = 0;
+
+	timeout_ret = wait_event_timeout(cqp_request->waitq,
+					 cqp_request->request_done,
+					 I40IW_EVENT_TIMEOUT);
+	if (!timeout_ret) {
+		i40iw_pr_err("error cqp command 0x%x timed out ret = %d\n",
+			     info->cqp_cmd, timeout_ret);
+		err_code = -ETIME;
+		i40iw_request_reset(iwdev);
+		goto done;
+	}
+	cqp_error = cqp_request->compl_info.error;
+	if (cqp_error) {
+		i40iw_pr_err("error cqp command 0x%x completion maj = 0x%x min=0x%x\n",
+			     info->cqp_cmd, cqp_request->compl_info.maj_err_code,
+			     cqp_request->compl_info.min_err_code);
+		err_code = -EPROTO;
+		goto done;
+	}
+done:
+	i40iw_put_cqp_request(iwcqp, cqp_request);
+	return err_code;
+}
+
+/**
+ * i40iw_handle_cqp_op - process cqp command
+ * @iwdev: iwarp device
+ * @cqp_request: cqp request to process
+ */
+enum i40iw_status_code i40iw_handle_cqp_op(struct i40iw_device *iwdev,
+					   struct i40iw_cqp_request
+					   *cqp_request)
+{
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+	struct cqp_commands_info *info = &cqp_request->info;
+	int err_code = 0;
+
+	status = i40iw_process_cqp_cmd(dev, info);
+	if (status) {
+		i40iw_pr_err("error cqp command 0x%x failed\n", info->cqp_cmd);
+		i40iw_free_cqp_request(&iwdev->cqp, cqp_request);
+		return status;
+	}
+	if (cqp_request->waiting)
+		err_code = i40iw_wait_event(iwdev, cqp_request);
+	if (err_code)
+		status = I40IW_ERR_CQP_COMPL_ERROR;
+	return status;
+}
+
+/**
+ * i40iw_add_pdusecount - add pd refcount
+ * @iwpd: pd for refcount
+ */
+void i40iw_add_pdusecount(struct i40iw_pd *iwpd)
+{
+	atomic_inc(&iwpd->usecount);
+}
+
+/**
+ * i40iw_rem_pdusecount - decrement refcount for pd and free if 0
+ * @iwpd: pd for refcount
+ * @iwdev: iwarp device
+ */
+void i40iw_rem_pdusecount(struct i40iw_pd *iwpd, struct i40iw_device *iwdev)
+{
+	if (!atomic_dec_and_test(&iwpd->usecount))
+		return;
+	i40iw_free_resource(iwdev, iwdev->allocated_pds, iwpd->sc_pd.pd_id);
+	kfree(iwpd);
+}
+
+/**
+ * i40iw_add_ref - add refcount for qp
+ * @ibqp: iqarp qp
+ */
+void i40iw_add_ref(struct ib_qp *ibqp)
+{
+	struct i40iw_qp *iwqp = (struct i40iw_qp *)ibqp;
+
+	atomic_inc(&iwqp->refcount);
+}
+
+/**
+ * i40iw_rem_ref - rem refcount for qp and free if 0
+ * @ibqp: iqarp qp
+ */
+void i40iw_rem_ref(struct ib_qp *ibqp)
+{
+	struct i40iw_qp *iwqp;
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_device *iwdev;
+	u32 qp_num;
+
+	iwqp = to_iwqp(ibqp);
+	if (!atomic_dec_and_test(&iwqp->refcount))
+		return;
+
+	iwdev = iwqp->iwdev;
+	qp_num = iwqp->ibqp.qp_num;
+	iwdev->qp_table[qp_num] = NULL;
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
+	if (!cqp_request)
+		return;
+
+	cqp_request->callback_fcn = i40iw_free_qp;
+	cqp_request->param = (void *)&iwqp->sc_qp;
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_QP_DESTROY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.qp_destroy.qp = &iwqp->sc_qp;
+	cqp_info->in.u.qp_destroy.scratch = (uintptr_t)cqp_request;
+	cqp_info->in.u.qp_destroy.remove_hash_idx = true;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Destroy QP fail");
+}
+
+/**
+ * i40iw_get_qp - get qp address
+ * @device: iwarp device
+ * @qpn: qp number
+ */
+struct ib_qp *i40iw_get_qp(struct ib_device *device, int qpn)
+{
+	struct i40iw_device *iwdev = to_iwdev(device);
+
+	if ((qpn < IW_FIRST_QPN) || (qpn >= iwdev->max_qp))
+		return NULL;
+
+	return &iwdev->qp_table[qpn]->ibqp;
+}
+
+/**
+ * i40iw_debug_buf - print debug msg and buffer is mask set
+ * @dev: hardware control device structure
+ * @mask: mask to compare if to print debug buffer
+ * @buf: points buffer addr
+ * @size: saize of buffer to print
+ */
+void i40iw_debug_buf(struct i40iw_sc_dev *dev,
+		     enum i40iw_debug_flag mask,
+		     char *desc,
+		     u64 *buf,
+		     u32 size)
+{
+	u32 i;
+
+	if (!(dev->debug_mask & mask))
+		return;
+	i40iw_debug(dev, mask, "%s\n", desc);
+	i40iw_debug(dev, mask, "starting address virt=%p phy=%llxh\n", buf,
+		    (unsigned long long)virt_to_phys(buf));
+
+	for (i = 0; i < size; i += 8)
+		i40iw_debug(dev, mask, "index %03d val: %016llx\n", i, buf[i / 8]);
+}
+
+/**
+ * i40iw_get_hw_addr - return hw addr
+ * @par: points to shared dev
+ */
+u8 __iomem *i40iw_get_hw_addr(void *par)
+{
+	struct i40iw_sc_dev *dev = (struct i40iw_sc_dev *)par;
+
+	return dev->hw->hw_addr;
+}
+
+/**
+ * i40iw_remove_head - return head entry and remove from list
+ * @list: list for entry
+ */
+void *i40iw_remove_head(struct list_head *list)
+{
+	struct list_head *entry;
+
+	if (list_empty(list))
+		return NULL;
+
+	entry = (void *)list->next;
+	list_del(entry);
+	return (void *)entry;
+}
+
+/**
+ * i40iw_allocate_dma_mem - Memory alloc helper fn
+ * @hw:   pointer to the HW structure
+ * @mem:  ptr to mem struct to fill out
+ * @size: size of memory requested
+ * @alignment: what to align the allocation to
+ */
+enum i40iw_status_code i40iw_allocate_dma_mem(struct i40iw_hw *hw,
+					      struct i40iw_dma_mem *mem,
+					      u64 size,
+					      u32 alignment)
+{
+	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+
+	if (!mem)
+		return I40IW_ERR_PARAM;
+	mem->size = ALIGN(size, alignment);
+	mem->va = dma_zalloc_coherent(&pcidev->dev, mem->size,
+				      (dma_addr_t *)&mem->pa, GFP_KERNEL);
+	if (!mem->va)
+		return I40IW_ERR_NO_MEMORY;
+	return 0;
+}
+
+/**
+ * i40iw_free_dma_mem - Memory free helper fn
+ * @hw:   pointer to the HW structure
+ * @mem:  ptr to mem struct to free
+ */
+void i40iw_free_dma_mem(struct i40iw_hw *hw, struct i40iw_dma_mem *mem)
+{
+	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+
+	if (!mem || !mem->va)
+		return;
+
+	dma_free_coherent(&pcidev->dev, mem->size,
+			  mem->va, (dma_addr_t)mem->pa);
+	mem->va = NULL;
+}
+
+/**
+ * i40iw_allocate_virt_mem - virtual memory alloc helper fn
+ * @hw:   pointer to the HW structure
+ * @mem:  ptr to mem struct to fill out
+ * @size: size of memory requested
+ */
+enum i40iw_status_code i40iw_allocate_virt_mem(struct i40iw_hw *hw,
+					       struct i40iw_virt_mem *mem,
+					       u32 size)
+{
+	if (!mem)
+		return I40IW_ERR_PARAM;
+
+	mem->size = size;
+	mem->va = kzalloc(size, GFP_KERNEL);
+
+	if (mem->va)
+		return 0;
+	else
+		return I40IW_ERR_NO_MEMORY;
+}
+
+/**
+ * i40iw_free_virt_mem - virtual memory free helper fn
+ * @hw:   pointer to the HW structure
+ * @mem:  ptr to mem struct to free
+ */
+enum i40iw_status_code i40iw_free_virt_mem(struct i40iw_hw *hw,
+					   struct i40iw_virt_mem *mem)
+{
+	if (!mem)
+		return I40IW_ERR_PARAM;
+	kfree(mem->va);
+	mem->va = NULL;
+	return 0;
+}
+
+/**
+ * i40iw_cqp_sds_cmd - create cqp command for sd
+ * @dev: hardware control device structure
+ * @sd_info: information  for sd cqp
+ *
+ */
+enum i40iw_status_code i40iw_cqp_sds_cmd(struct i40iw_sc_dev *dev,
+					 struct i40iw_update_sds_info *sdinfo)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+	cqp_info = &cqp_request->info;
+	memcpy(&cqp_info->in.u.update_pe_sds.info, sdinfo,
+	       sizeof(cqp_info->in.u.update_pe_sds.info));
+	cqp_info->cqp_cmd = OP_UPDATE_PE_SDS;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.update_pe_sds.dev = dev;
+	cqp_info->in.u.update_pe_sds.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Update SD's fail");
+	return status;
+}
+
+/**
+ * i40iw_term_modify_qp - modify qp for term message
+ * @qp: hardware control qp
+ * @next_state: qp's next state
+ * @term: terminate code
+ * @term_len: length
+ */
+void i40iw_term_modify_qp(struct i40iw_sc_qp *qp, u8 next_state, u8 term, u8 term_len)
+{
+	struct i40iw_qp *iwqp;
+
+	iwqp = (struct i40iw_qp *)qp->back_qp;
+	i40iw_next_iw_state(iwqp, next_state, 0, term, term_len);
+};
+
+/**
+ * i40iw_terminate_done - after terminate is completed
+ * @qp: hardware control qp
+ * @timeout_occurred: indicates if terminate timer expired
+ */
+void i40iw_terminate_done(struct i40iw_sc_qp *qp, int timeout_occurred)
+{
+	struct i40iw_qp *iwqp;
+	u32 next_iwarp_state = I40IW_QP_STATE_ERROR;
+	u8 hte = 0;
+	bool first_time;
+	unsigned long flags;
+
+	iwqp = (struct i40iw_qp *)qp->back_qp;
+	spin_lock_irqsave(&iwqp->lock, flags);
+	if (iwqp->hte_added) {
+		iwqp->hte_added = 0;
+		hte = 1;
+	}
+	first_time = !(qp->term_flags & I40IW_TERM_DONE);
+	qp->term_flags |= I40IW_TERM_DONE;
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+	if (first_time) {
+		if (!timeout_occurred)
+			i40iw_terminate_del_timer(qp);
+		else
+			next_iwarp_state = I40IW_QP_STATE_CLOSING;
+
+		i40iw_next_iw_state(iwqp, next_iwarp_state, hte, 0, 0);
+		i40iw_cm_disconn(iwqp);
+	}
+}
+
+/**
+ * i40iw_terminate_imeout - timeout happened
+ * @context: points to iwarp qp
+ */
+static void i40iw_terminate_timeout(unsigned long context)
+{
+	struct i40iw_qp *iwqp = (struct i40iw_qp *)context;
+	struct i40iw_sc_qp *qp = (struct i40iw_sc_qp *)&iwqp->sc_qp;
+
+	i40iw_terminate_done(qp, 1);
+}
+
+/**
+ * i40iw_terminate_start_timer - start terminate timeout
+ * @qp: hardware control qp
+ */
+void i40iw_terminate_start_timer(struct i40iw_sc_qp *qp)
+{
+	struct i40iw_qp *iwqp;
+
+	iwqp = (struct i40iw_qp *)qp->back_qp;
+	init_timer(&iwqp->terminate_timer);
+	iwqp->terminate_timer.function = i40iw_terminate_timeout;
+	iwqp->terminate_timer.expires = jiffies + HZ;
+	iwqp->terminate_timer.data = (unsigned long)iwqp;
+	add_timer(&iwqp->terminate_timer);
+}
+
+/**
+ * i40iw_terminate_del_timer - delete terminate timeout
+ * @qp: hardware control qp
+ */
+void i40iw_terminate_del_timer(struct i40iw_sc_qp *qp)
+{
+	struct i40iw_qp *iwqp;
+
+	iwqp = (struct i40iw_qp *)qp->back_qp;
+	del_timer(&iwqp->terminate_timer);
+}
+
+/**
+ * i40iw_cqp_generic_worker - generic worker for cqp
+ * @work: work pointer
+ */
+static void i40iw_cqp_generic_worker(struct work_struct *work)
+{
+	struct i40iw_virtchnl_work_info *work_info =
+	    &((struct virtchnl_work *)work)->work_info;
+
+	if (work_info->worker_vf_dev)
+		work_info->callback_fcn(work_info->worker_vf_dev);
+}
+
+/**
+ * i40iw_cqp_spawn_worker - spawn worket thread
+ * @iwdev: device struct pointer
+ * @work_info: work request info
+ * @iw_vf_idx: virtual function index
+ */
+void i40iw_cqp_spawn_worker(struct i40iw_sc_dev *dev,
+			    struct i40iw_virtchnl_work_info *work_info,
+			    u32 iw_vf_idx)
+{
+	struct virtchnl_work *work;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	work = &iwdev->virtchnl_w[iw_vf_idx];
+	memcpy(&work->work_info, work_info, sizeof(*work_info));
+	INIT_WORK(&work->work, i40iw_cqp_generic_worker);
+	queue_work(iwdev->virtchnl_wq, &work->work);
+}
+
+/**
+ * i40iw_cqp_manage_hmc_fcn_worker -
+ * @work: work pointer for hmc info
+ */
+static void i40iw_cqp_manage_hmc_fcn_worker(struct work_struct *work)
+{
+	struct i40iw_cqp_request *cqp_request =
+	    ((struct virtchnl_work *)work)->cqp_request;
+	struct i40iw_ccq_cqe_info ccq_cqe_info;
+	struct i40iw_hmc_fcn_info *hmcfcninfo =
+			&cqp_request->info.in.u.manage_hmc_pm.info;
+	struct i40iw_device *iwdev =
+	    (struct i40iw_device *)cqp_request->info.in.u.manage_hmc_pm.dev->back_dev;
+
+	ccq_cqe_info.cqp = NULL;
+	ccq_cqe_info.maj_err_code = cqp_request->compl_info.maj_err_code;
+	ccq_cqe_info.min_err_code = cqp_request->compl_info.min_err_code;
+	ccq_cqe_info.op_code = cqp_request->compl_info.op_code;
+	ccq_cqe_info.op_ret_val = cqp_request->compl_info.op_ret_val;
+	ccq_cqe_info.scratch = 0;
+	ccq_cqe_info.error = cqp_request->compl_info.error;
+	hmcfcninfo->callback_fcn(cqp_request->info.in.u.manage_hmc_pm.dev,
+				 hmcfcninfo->cqp_callback_param, &ccq_cqe_info);
+	i40iw_put_cqp_request(&iwdev->cqp, cqp_request);
+}
+
+/**
+ * i40iw_cqp_manage_hmc_fcn_callback - called function after cqp completion
+ * @cqp_request: cqp request info struct for hmc fun
+ * @unused: unused param of callback
+ */
+static void i40iw_cqp_manage_hmc_fcn_callback(struct i40iw_cqp_request *cqp_request,
+					      u32 unused)
+{
+	struct virtchnl_work *work;
+	struct i40iw_hmc_fcn_info *hmcfcninfo =
+	    &cqp_request->info.in.u.manage_hmc_pm.info;
+	struct i40iw_device *iwdev =
+	    (struct i40iw_device *)cqp_request->info.in.u.manage_hmc_pm.dev->
+	    back_dev;
+
+	if (hmcfcninfo && hmcfcninfo->callback_fcn) {
+		i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_HMC, "%s1\n", __func__);
+		atomic_inc(&cqp_request->refcount);
+		work = &iwdev->virtchnl_w[hmcfcninfo->iw_vf_idx];
+		work->cqp_request = cqp_request;
+		INIT_WORK(&work->work, i40iw_cqp_manage_hmc_fcn_worker);
+		queue_work(iwdev->virtchnl_wq, &work->work);
+		i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_HMC, "%s2\n", __func__);
+	} else {
+		i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_HMC, "%s: Something wrong\n", __func__);
+	}
+}
+
+/**
+ * i40iw_cqp_manage_hmc_fcn_cmd - issue cqp command to manage hmc
+ * @dev: hardware control device structure
+ * @hmcfcninfo: info for hmc
+ */
+enum i40iw_status_code i40iw_cqp_manage_hmc_fcn_cmd(struct i40iw_sc_dev *dev,
+						    struct i40iw_hmc_fcn_info *hmcfcninfo)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	i40iw_debug(&iwdev->sc_dev, I40IW_DEBUG_HMC, "%s\n", __func__);
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+	cqp_info = &cqp_request->info;
+	cqp_request->callback_fcn = i40iw_cqp_manage_hmc_fcn_callback;
+	cqp_request->param = hmcfcninfo;
+	memcpy(&cqp_info->in.u.manage_hmc_pm.info, hmcfcninfo,
+	       sizeof(*hmcfcninfo));
+	cqp_info->in.u.manage_hmc_pm.dev = dev;
+	cqp_info->cqp_cmd = OP_MANAGE_HMC_PM_FUNC_TABLE;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.manage_hmc_pm.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Manage HMC fail");
+	return status;
+}
+
+/**
+ * i40iw_cqp_query_fpm_values_cmd - send cqp command for fpm
+ * @iwdev: function device struct
+ * @values_mem: buffer for fpm
+ * @hmc_fn_id: function id for fpm
+ */
+enum i40iw_status_code i40iw_cqp_query_fpm_values_cmd(struct i40iw_sc_dev *dev,
+						      struct i40iw_dma_mem *values_mem,
+						      u8 hmc_fn_id)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+	cqp_info = &cqp_request->info;
+	cqp_request->param = NULL;
+	cqp_info->in.u.query_fpm_values.cqp = dev->cqp;
+	cqp_info->in.u.query_fpm_values.fpm_values_pa = values_mem->pa;
+	cqp_info->in.u.query_fpm_values.fpm_values_va = values_mem->va;
+	cqp_info->in.u.query_fpm_values.hmc_fn_id = hmc_fn_id;
+	cqp_info->cqp_cmd = OP_QUERY_FPM_VALUES;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.query_fpm_values.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Query FPM fail");
+	return status;
+}
+
+/**
+ * i40iw_cqp_commit_fpm_values_cmd - commit fpm values in hw
+ * @dev: hardware control device structure
+ * @values_mem: buffer with fpm values
+ * @hmc_fn_id: function id for fpm
+ */
+enum i40iw_status_code i40iw_cqp_commit_fpm_values_cmd(struct i40iw_sc_dev *dev,
+						       struct i40iw_dma_mem *values_mem,
+						       u8 hmc_fn_id)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return I40IW_ERR_NO_MEMORY;
+	cqp_info = &cqp_request->info;
+	cqp_request->param = NULL;
+	cqp_info->in.u.commit_fpm_values.cqp = dev->cqp;
+	cqp_info->in.u.commit_fpm_values.fpm_values_pa = values_mem->pa;
+	cqp_info->in.u.commit_fpm_values.fpm_values_va = values_mem->va;
+	cqp_info->in.u.commit_fpm_values.hmc_fn_id = hmc_fn_id;
+	cqp_info->cqp_cmd = OP_COMMIT_FPM_VALUES;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.commit_fpm_values.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Commit FPM fail");
+	return status;
+}
+
+/**
+ * i40iw_vf_wait_vchnl_resp - wait for channel msg
+ * @iwdev: function's device struct
+ */
+enum i40iw_status_code i40iw_vf_wait_vchnl_resp(struct i40iw_sc_dev *dev)
+{
+	struct i40iw_device *iwdev = dev->back_dev;
+	enum i40iw_status_code err_code = 0;
+	int timeout_ret;
+
+	i40iw_debug(dev, I40IW_DEBUG_VIRT, "%s[%u] dev %p, iwdev %p\n",
+		    __func__, __LINE__, dev, iwdev);
+	atomic_add(2, &iwdev->vchnl_msgs);
+	timeout_ret = wait_event_timeout(iwdev->vchnl_waitq,
+					 (atomic_read(&iwdev->vchnl_msgs) == 1),
+					 I40IW_VCHNL_EVENT_TIMEOUT);
+	atomic_dec(&iwdev->vchnl_msgs);
+	if (!timeout_ret) {
+		i40iw_pr_err("virt channel completion timeout = 0x%x\n", timeout_ret);
+		err_code = I40IW_ERR_TIMEOUT;
+	}
+	return err_code;
+}
+
+/**
+ * i40iw_ieq_mpa_crc_ae - generate AE for crc error
+ * @dev: hardware control device structure
+ * @qp: hardware control qp
+ */
+void i40iw_ieq_mpa_crc_ae(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp)
+{
+	struct i40iw_qp_flush_info info;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	i40iw_debug(dev, I40IW_DEBUG_AEQ, "%s entered\n", __func__);
+	memset(&info, 0, sizeof(info));
+	info.ae_code = I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR;
+	info.generate_ae = true;
+	info.ae_source = 0x3;
+	(void)i40iw_hw_flush_wqes(iwdev, qp, &info, false);
+}
+
+/**
+ * i40iw_init_hash_desc - initialize hash for crc calculation
+ * @desc: cryption type
+ */
+enum i40iw_status_code i40iw_init_hash_desc(struct shash_desc **desc)
+{
+	struct crypto_shash *tfm;
+	struct shash_desc *tdesc;
+
+	tfm = crypto_alloc_shash("crc32c", 0, 0);
+	if (IS_ERR(tfm))
+		return I40IW_ERR_MPA_CRC;
+
+	tdesc = kzalloc(sizeof(*tdesc) + crypto_shash_descsize(tfm),
+			GFP_KERNEL);
+	if (!tdesc) {
+		crypto_free_shash(tfm);
+		return I40IW_ERR_MPA_CRC;
+	}
+	tdesc->tfm = tfm;
+	*desc = tdesc;
+
+	return 0;
+}
+
+/**
+ * i40iw_free_hash_desc - free hash desc
+ * @desc: to be freed
+ */
+void i40iw_free_hash_desc(struct shash_desc *desc)
+{
+	if (desc) {
+		crypto_free_shash(desc->tfm);
+		kfree(desc);
+	}
+}
+
+/**
+ * i40iw_alloc_query_fpm_buf - allocate buffer for fpm
+ * @dev: hardware control device structure
+ * @mem: buffer ptr for fpm to be allocated
+ * @return: memory allocation status
+ */
+enum i40iw_status_code i40iw_alloc_query_fpm_buf(struct i40iw_sc_dev *dev,
+						 struct i40iw_dma_mem *mem)
+{
+	enum i40iw_status_code status;
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+
+	status = i40iw_obj_aligned_mem(iwdev, mem, I40IW_QUERY_FPM_BUF_SIZE,
+				       I40IW_FPM_QUERY_BUF_ALIGNMENT_MASK);
+	return status;
+}
+
+/**
+ * i40iw_ieq_check_mpacrc - check if mpa crc is OK
+ * @desc: desc for hash
+ * @addr: address of buffer for crc
+ * @length: length of buffer
+ * @value: value to be compared
+ */
+enum i40iw_status_code i40iw_ieq_check_mpacrc(struct shash_desc *desc,
+					      void *addr,
+					      u32 length,
+					      u32 value)
+{
+	u32 crc = 0;
+	int ret;
+	enum i40iw_status_code ret_code = 0;
+
+	crypto_shash_init(desc);
+	ret = crypto_shash_update(desc, addr, length);
+	if (!ret)
+		crypto_shash_final(desc, (u8 *)&crc);
+	if (crc != value) {
+		i40iw_pr_err("mpa crc check fail\n");
+		ret_code = I40IW_ERR_MPA_CRC;
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_ieq_get_qp - get qp based on quad in puda buffer
+ * @dev: hardware control device structure
+ * @buf: receive puda buffer on exception q
+ */
+struct i40iw_sc_qp *i40iw_ieq_get_qp(struct i40iw_sc_dev *dev,
+				     struct i40iw_puda_buf *buf)
+{
+	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
+	struct i40iw_qp *iwqp;
+	struct i40iw_cm_node *cm_node;
+	u32 loc_addr[4], rem_addr[4];
+	u16 loc_port, rem_port;
+	struct ipv6hdr *ip6h;
+	struct iphdr *iph = (struct iphdr *)buf->iph;
+	struct tcphdr *tcph = (struct tcphdr *)buf->tcph;
+
+	if (iph->version == 4) {
+		memset(loc_addr, 0, sizeof(loc_addr));
+		loc_addr[0] = ntohl(iph->daddr);
+		memset(rem_addr, 0, sizeof(rem_addr));
+		rem_addr[0] = ntohl(iph->saddr);
+	} else {
+		ip6h = (struct ipv6hdr *)buf->iph;
+		i40iw_copy_ip_ntohl(loc_addr, ip6h->daddr.in6_u.u6_addr32);
+		i40iw_copy_ip_ntohl(rem_addr, ip6h->saddr.in6_u.u6_addr32);
+	}
+	loc_port = ntohs(tcph->dest);
+	rem_port = ntohs(tcph->source);
+
+	cm_node = i40iw_find_node(&iwdev->cm_core, rem_port, rem_addr, loc_port,
+				  loc_addr, false);
+	if (!cm_node)
+		return NULL;
+	iwqp = cm_node->iwqp;
+	return &iwqp->sc_qp;
+}
+
+/**
+ * i40iw_ieq_update_tcpip_info - update tcpip in the buffer
+ * @buf: puda to update
+ * @length: length of buffer
+ * @seqnum: seq number for tcp
+ */
+void i40iw_ieq_update_tcpip_info(struct i40iw_puda_buf *buf, u16 length, u32 seqnum)
+{
+	struct tcphdr *tcph;
+	struct iphdr *iph;
+	u16 iphlen;
+	u16 packetsize;
+	u8 *addr = (u8 *)buf->mem.va;
+
+	iphlen = (buf->ipv4) ? 20 : 40;
+	iph = (struct iphdr *)(addr + buf->maclen);
+	tcph = (struct tcphdr *)(addr + buf->maclen + iphlen);
+	packetsize = length + buf->tcphlen + iphlen;
+
+	iph->tot_len = htons(packetsize);
+	tcph->seq = htonl(seqnum);
+}
+
+/**
+ * i40iw_puda_get_tcpip_info - get tcpip info from puda buffer
+ * @info: to get information
+ * @buf: puda buffer
+ */
+enum i40iw_status_code i40iw_puda_get_tcpip_info(struct i40iw_puda_completion_info *info,
+						 struct i40iw_puda_buf *buf)
+{
+	struct iphdr *iph;
+	struct ipv6hdr *ip6h;
+	struct tcphdr *tcph;
+	u16 iphlen;
+	u16 pkt_len;
+	u8 *mem = (u8 *)buf->mem.va;
+	struct ethhdr *ethh = (struct ethhdr *)buf->mem.va;
+
+	if (ethh->h_proto == htons(0x8100)) {
+		info->vlan_valid = true;
+		buf->vlan_id = ntohs(((struct vlan_ethhdr *)ethh)->h_vlan_TCI) & VLAN_VID_MASK;
+	}
+	buf->maclen = (info->vlan_valid) ? 18 : 14;
+	iphlen = (info->l3proto) ? 40 : 20;
+	buf->ipv4 = (info->l3proto) ? false : true;
+	buf->iph = mem + buf->maclen;
+	iph = (struct iphdr *)buf->iph;
+
+	buf->tcph = buf->iph + iphlen;
+	tcph = (struct tcphdr *)buf->tcph;
+
+	if (buf->ipv4) {
+		pkt_len = ntohs(iph->tot_len);
+	} else {
+		ip6h = (struct ipv6hdr *)buf->iph;
+		pkt_len = ntohs(ip6h->payload_len) + iphlen;
+	}
+
+	buf->totallen = pkt_len + buf->maclen;
+
+	if (info->payload_len < buf->totallen - 4) {
+		i40iw_pr_err("payload_len = 0x%x totallen expected0x%x\n",
+			     info->payload_len, buf->totallen);
+		return I40IW_ERR_INVALID_SIZE;
+	}
+
+	buf->tcphlen = (tcph->doff) << 2;
+	buf->datalen = pkt_len - iphlen - buf->tcphlen;
+	buf->data = (buf->datalen) ? buf->tcph + buf->tcphlen : NULL;
+	buf->hdrlen = buf->maclen + iphlen + buf->tcphlen;
+	buf->seqnum = ntohl(tcph->seq);
+	return 0;
+}
+
+/**
+ * i40iw_hw_stats_timeout - Stats timer-handler which updates all HW stats
+ * @dev: hardware control device structure
+ */
+static void i40iw_hw_stats_timeout(unsigned long dev)
+{
+	struct i40iw_sc_dev *pf_dev = (struct i40iw_sc_dev *)dev;
+	struct i40iw_dev_pestat *pf_devstat = &pf_dev->dev_pestat;
+	struct i40iw_dev_pestat *vf_devstat = NULL;
+	u16 iw_vf_idx;
+	unsigned long flags;
+
+	/*PF*/
+	pf_devstat->ops.iw_hw_stat_read_all(pf_devstat, &pf_devstat->hw_stats);
+	for (iw_vf_idx = 0; iw_vf_idx < I40IW_MAX_PE_ENABLED_VF_COUNT; iw_vf_idx++) {
+		spin_lock_irqsave(&pf_devstat->stats_lock, flags);
+		if (pf_dev->vf_dev[iw_vf_idx]) {
+			if (pf_dev->vf_dev[iw_vf_idx]->stats_initialized) {
+				vf_devstat = &pf_dev->vf_dev[iw_vf_idx]->dev_pestat;
+				vf_devstat->ops.iw_hw_stat_read_all(vf_devstat, &vf_devstat->hw_stats);
+			}
+		}
+		spin_unlock_irqrestore(&pf_devstat->stats_lock, flags);
+	}
+
+	mod_timer(&pf_devstat->stats_timer,
+		  jiffies + msecs_to_jiffies(STATS_TIMER_DELAY));
+}
+
+/**
+ * i40iw_hw_stats_start_timer - Start periodic stats timer
+ * @dev: hardware control device structure
+ */
+void i40iw_hw_stats_start_timer(struct i40iw_sc_dev *dev)
+{
+	struct i40iw_dev_pestat *devstat = &dev->dev_pestat;
+
+	init_timer(&devstat->stats_timer);
+	devstat->stats_timer.function = i40iw_hw_stats_timeout;
+	devstat->stats_timer.data = (unsigned long)dev;
+	mod_timer(&devstat->stats_timer,
+		  jiffies + msecs_to_jiffies(STATS_TIMER_DELAY));
+}
+
+/**
+ * i40iw_hw_stats_del_timer - Delete periodic stats timer
+ * @dev: hardware control device structure
+ */
+void i40iw_hw_stats_del_timer(struct i40iw_sc_dev *dev)
+{
+	struct i40iw_dev_pestat *devstat = &dev->dev_pestat;
+
+	del_timer_sync(&devstat->stats_timer);
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
new file mode 100644
index 000000000000..1fe3b84a06e4
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@@ -0,0 +1,2437 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/random.h>
+#include <linux/highmem.h>
+#include <linux/time.h>
+#include <asm/byteorder.h>
+#include <net/ip.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/iw_cm.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_umem.h>
+#include "i40iw.h"
+
+/**
+ * i40iw_query_device - get device attributes
+ * @ibdev: device pointer from stack
+ * @props: returning device attributes
+ * @udata: user data
+ */
+static int i40iw_query_device(struct ib_device *ibdev,
+			      struct ib_device_attr *props,
+			      struct ib_udata *udata)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+
+	if (udata->inlen || udata->outlen)
+		return -EINVAL;
+	memset(props, 0, sizeof(*props));
+	ether_addr_copy((u8 *)&props->sys_image_guid, iwdev->netdev->dev_addr);
+	props->fw_ver = I40IW_FW_VERSION;
+	props->device_cap_flags = iwdev->device_cap_flags;
+	props->vendor_id = iwdev->vendor_id;
+	props->vendor_part_id = iwdev->vendor_part_id;
+	props->hw_ver = (u32)iwdev->sc_dev.hw_rev;
+	props->max_mr_size = I40IW_MAX_OUTBOUND_MESSAGE_SIZE;
+	props->max_qp = iwdev->max_qp;
+	props->max_qp_wr = (I40IW_MAX_WQ_ENTRIES >> 2) - 1;
+	props->max_sge = I40IW_MAX_WQ_FRAGMENT_COUNT;
+	props->max_cq = iwdev->max_cq;
+	props->max_cqe = iwdev->max_cqe;
+	props->max_mr = iwdev->max_mr;
+	props->max_pd = iwdev->max_pd;
+	props->max_sge_rd = 1;
+	props->max_qp_rd_atom = I40IW_MAX_IRD_SIZE;
+	props->max_qp_init_rd_atom = props->max_qp_rd_atom;
+	props->atomic_cap = IB_ATOMIC_NONE;
+	props->max_map_per_fmr = 1;
+	return 0;
+}
+
+/**
+ * i40iw_query_port - get port attrubutes
+ * @ibdev: device pointer from stack
+ * @port: port number for query
+ * @props: returning device attributes
+ */
+static int i40iw_query_port(struct ib_device *ibdev,
+			    u8 port,
+			    struct ib_port_attr *props)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct net_device *netdev = iwdev->netdev;
+
+	memset(props, 0, sizeof(*props));
+
+	props->max_mtu = IB_MTU_4096;
+	if (netdev->mtu >= 4096)
+		props->active_mtu = IB_MTU_4096;
+	else if (netdev->mtu >= 2048)
+		props->active_mtu = IB_MTU_2048;
+	else if (netdev->mtu >= 1024)
+		props->active_mtu = IB_MTU_1024;
+	else if (netdev->mtu >= 512)
+		props->active_mtu = IB_MTU_512;
+	else
+		props->active_mtu = IB_MTU_256;
+
+	props->lid = 1;
+	if (netif_carrier_ok(iwdev->netdev))
+		props->state = IB_PORT_ACTIVE;
+	else
+		props->state = IB_PORT_DOWN;
+	props->port_cap_flags = IB_PORT_CM_SUP | IB_PORT_REINIT_SUP |
+		IB_PORT_VENDOR_CLASS_SUP | IB_PORT_BOOT_MGMT_SUP;
+	props->gid_tbl_len = 1;
+	props->pkey_tbl_len = 1;
+	props->active_width = IB_WIDTH_4X;
+	props->active_speed = 1;
+	props->max_msg_sz = 0x80000000;
+	return 0;
+}
+
+/**
+ * i40iw_alloc_ucontext - Allocate the user context data structure
+ * @ibdev: device pointer from stack
+ * @udata: user data
+ *
+ * This keeps track of all objects associated with a particular
+ * user-mode client.
+ */
+static struct ib_ucontext *i40iw_alloc_ucontext(struct ib_device *ibdev,
+						struct ib_udata *udata)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct i40iw_alloc_ucontext_req req;
+	struct i40iw_alloc_ucontext_resp uresp;
+	struct i40iw_ucontext *ucontext;
+
+	if (ib_copy_from_udata(&req, udata, sizeof(req)))
+		return ERR_PTR(-EINVAL);
+
+	if (req.userspace_ver != I40IW_ABI_USERSPACE_VER) {
+		i40iw_pr_err("Invalid userspace driver version detected. Detected version %d, should be %d\n",
+			     req.userspace_ver, I40IW_ABI_USERSPACE_VER);
+		return ERR_PTR(-EINVAL);
+	}
+
+	memset(&uresp, 0, sizeof(uresp));
+	uresp.max_qps = iwdev->max_qp;
+	uresp.max_pds = iwdev->max_pd;
+	uresp.wq_size = iwdev->max_qp_wr * 2;
+	uresp.kernel_ver = I40IW_ABI_KERNEL_VER;
+
+	ucontext = kzalloc(sizeof(*ucontext), GFP_KERNEL);
+	if (!ucontext)
+		return ERR_PTR(-ENOMEM);
+
+	ucontext->iwdev = iwdev;
+
+	if (ib_copy_to_udata(udata, &uresp, sizeof(uresp))) {
+		kfree(ucontext);
+		return ERR_PTR(-EFAULT);
+	}
+
+	INIT_LIST_HEAD(&ucontext->cq_reg_mem_list);
+	spin_lock_init(&ucontext->cq_reg_mem_list_lock);
+	INIT_LIST_HEAD(&ucontext->qp_reg_mem_list);
+	spin_lock_init(&ucontext->qp_reg_mem_list_lock);
+
+	return &ucontext->ibucontext;
+}
+
+/**
+ * i40iw_dealloc_ucontext - deallocate the user context data structure
+ * @context: user context created during alloc
+ */
+static int i40iw_dealloc_ucontext(struct ib_ucontext *context)
+{
+	struct i40iw_ucontext *ucontext = to_ucontext(context);
+	unsigned long flags;
+
+	spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
+	if (!list_empty(&ucontext->cq_reg_mem_list)) {
+		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
+		return -EBUSY;
+	}
+	spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
+	spin_lock_irqsave(&ucontext->qp_reg_mem_list_lock, flags);
+	if (!list_empty(&ucontext->qp_reg_mem_list)) {
+		spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
+		return -EBUSY;
+	}
+	spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
+
+	kfree(ucontext);
+	return 0;
+}
+
+/**
+ * i40iw_mmap - user memory map
+ * @context: context created during alloc
+ * @vma: kernel info for user memory map
+ */
+static int i40iw_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct i40iw_ucontext *ucontext;
+	u64 db_addr_offset;
+	u64 push_offset;
+
+	ucontext = to_ucontext(context);
+	if (ucontext->iwdev->sc_dev.is_pf) {
+		db_addr_offset = I40IW_DB_ADDR_OFFSET;
+		push_offset = I40IW_PUSH_OFFSET;
+		if (vma->vm_pgoff)
+			vma->vm_pgoff += I40IW_PF_FIRST_PUSH_PAGE_INDEX - 1;
+	} else {
+		db_addr_offset = I40IW_VF_DB_ADDR_OFFSET;
+		push_offset = I40IW_VF_PUSH_OFFSET;
+		if (vma->vm_pgoff)
+			vma->vm_pgoff += I40IW_VF_FIRST_PUSH_PAGE_INDEX - 1;
+	}
+
+	vma->vm_pgoff += db_addr_offset >> PAGE_SHIFT;
+
+	if (vma->vm_pgoff == (db_addr_offset >> PAGE_SHIFT)) {
+		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+		vma->vm_private_data = ucontext;
+	} else {
+		if ((vma->vm_pgoff - (push_offset >> PAGE_SHIFT)) % 2)
+			vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+		else
+			vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+	}
+
+	if (io_remap_pfn_range(vma, vma->vm_start,
+			       vma->vm_pgoff + (pci_resource_start(ucontext->iwdev->ldev->pcidev, 0) >> PAGE_SHIFT),
+			       PAGE_SIZE, vma->vm_page_prot))
+		return -EAGAIN;
+
+	return 0;
+}
+
+/**
+ * i40iw_alloc_push_page - allocate a push page for qp
+ * @iwdev: iwarp device
+ * @qp: hardware control qp
+ */
+static void i40iw_alloc_push_page(struct i40iw_device *iwdev, struct i40iw_sc_qp *qp)
+{
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+
+	if (qp->push_idx != I40IW_INVALID_PUSH_PAGE_INDEX)
+		return;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return;
+
+	atomic_inc(&cqp_request->refcount);
+
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_MANAGE_PUSH_PAGE;
+	cqp_info->post_sq = 1;
+
+	cqp_info->in.u.manage_push_page.info.qs_handle = dev->qs_handle;
+	cqp_info->in.u.manage_push_page.info.free_page = 0;
+	cqp_info->in.u.manage_push_page.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.manage_push_page.scratch = (uintptr_t)cqp_request;
+
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (!status)
+		qp->push_idx = cqp_request->compl_info.op_ret_val;
+	else
+		i40iw_pr_err("CQP-OP Push page fail");
+	i40iw_put_cqp_request(&iwdev->cqp, cqp_request);
+}
+
+/**
+ * i40iw_dealloc_push_page - free a push page for qp
+ * @iwdev: iwarp device
+ * @qp: hardware control qp
+ */
+static void i40iw_dealloc_push_page(struct i40iw_device *iwdev, struct i40iw_sc_qp *qp)
+{
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	enum i40iw_status_code status;
+
+	if (qp->push_idx == I40IW_INVALID_PUSH_PAGE_INDEX)
+		return;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
+	if (!cqp_request)
+		return;
+
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_MANAGE_PUSH_PAGE;
+	cqp_info->post_sq = 1;
+
+	cqp_info->in.u.manage_push_page.info.push_idx = qp->push_idx;
+	cqp_info->in.u.manage_push_page.info.qs_handle = dev->qs_handle;
+	cqp_info->in.u.manage_push_page.info.free_page = 1;
+	cqp_info->in.u.manage_push_page.cqp = &iwdev->cqp.sc_cqp;
+	cqp_info->in.u.manage_push_page.scratch = (uintptr_t)cqp_request;
+
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (!status)
+		qp->push_idx = I40IW_INVALID_PUSH_PAGE_INDEX;
+	else
+		i40iw_pr_err("CQP-OP Push page fail");
+}
+
+/**
+ * i40iw_alloc_pd - allocate protection domain
+ * @ibdev: device pointer from stack
+ * @context: user context created during alloc
+ * @udata: user data
+ */
+static struct ib_pd *i40iw_alloc_pd(struct ib_device *ibdev,
+				    struct ib_ucontext *context,
+				    struct ib_udata *udata)
+{
+	struct i40iw_pd *iwpd;
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_alloc_pd_resp uresp;
+	struct i40iw_sc_pd *sc_pd;
+	u32 pd_id = 0;
+	int err;
+
+	err = i40iw_alloc_resource(iwdev, iwdev->allocated_pds,
+				   iwdev->max_pd, &pd_id, &iwdev->next_pd);
+	if (err) {
+		i40iw_pr_err("alloc resource failed\n");
+		return ERR_PTR(err);
+	}
+
+	iwpd = kzalloc(sizeof(*iwpd), GFP_KERNEL);
+	if (!iwpd) {
+		err = -ENOMEM;
+		goto free_res;
+	}
+
+	sc_pd = &iwpd->sc_pd;
+	dev->iw_pd_ops->pd_init(dev, sc_pd, pd_id);
+
+	if (context) {
+		memset(&uresp, 0, sizeof(uresp));
+		uresp.pd_id = pd_id;
+		if (ib_copy_to_udata(udata, &uresp, sizeof(uresp))) {
+			err = -EFAULT;
+			goto error;
+		}
+	}
+
+	i40iw_add_pdusecount(iwpd);
+	return &iwpd->ibpd;
+error:
+	kfree(iwpd);
+free_res:
+	i40iw_free_resource(iwdev, iwdev->allocated_pds, pd_id);
+	return ERR_PTR(err);
+}
+
+/**
+ * i40iw_dealloc_pd - deallocate pd
+ * @ibpd: ptr of pd to be deallocated
+ */
+static int i40iw_dealloc_pd(struct ib_pd *ibpd)
+{
+	struct i40iw_pd *iwpd = to_iwpd(ibpd);
+	struct i40iw_device *iwdev = to_iwdev(ibpd->device);
+
+	i40iw_rem_pdusecount(iwpd, iwdev);
+	return 0;
+}
+
+/**
+ * i40iw_qp_roundup - return round up qp ring size
+ * @wr_ring_size: ring size to round up
+ */
+static int i40iw_qp_roundup(u32 wr_ring_size)
+{
+	int scount = 1;
+
+	if (wr_ring_size < I40IWQP_SW_MIN_WQSIZE)
+		wr_ring_size = I40IWQP_SW_MIN_WQSIZE;
+
+	for (wr_ring_size--; scount <= 16; scount *= 2)
+		wr_ring_size |= wr_ring_size >> scount;
+	return ++wr_ring_size;
+}
+
+/**
+ * i40iw_get_pbl - Retrieve pbl from a list given a virtual
+ * address
+ * @va: user virtual address
+ * @pbl_list: pbl list to search in (QP's or CQ's)
+ */
+static struct i40iw_pbl *i40iw_get_pbl(unsigned long va,
+				       struct list_head *pbl_list)
+{
+	struct i40iw_pbl *iwpbl;
+
+	list_for_each_entry(iwpbl, pbl_list, list) {
+		if (iwpbl->user_base == va) {
+			list_del(&iwpbl->list);
+			return iwpbl;
+		}
+	}
+	return NULL;
+}
+
+/**
+ * i40iw_free_qp_resources - free up memory resources for qp
+ * @iwdev: iwarp device
+ * @iwqp: qp ptr (user or kernel)
+ * @qp_num: qp number assigned
+ */
+void i40iw_free_qp_resources(struct i40iw_device *iwdev,
+			     struct i40iw_qp *iwqp,
+			     u32 qp_num)
+{
+	i40iw_dealloc_push_page(iwdev, &iwqp->sc_qp);
+	if (qp_num)
+		i40iw_free_resource(iwdev, iwdev->allocated_qps, qp_num);
+	i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwqp->q2_ctx_mem);
+	i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwqp->kqp.dma_mem);
+	kfree(iwqp->kqp.wrid_mem);
+	iwqp->kqp.wrid_mem = NULL;
+	kfree(iwqp->allocated_buffer);
+	iwqp->allocated_buffer = NULL;
+}
+
+/**
+ * i40iw_clean_cqes - clean cq entries for qp
+ * @iwqp: qp ptr (user or kernel)
+ * @iwcq: cq ptr
+ */
+static void i40iw_clean_cqes(struct i40iw_qp *iwqp, struct i40iw_cq *iwcq)
+{
+	struct i40iw_cq_uk *ukcq = &iwcq->sc_cq.cq_uk;
+
+	ukcq->ops.iw_cq_clean(&iwqp->sc_qp.qp_uk, ukcq);
+}
+
+/**
+ * i40iw_destroy_qp - destroy qp
+ * @ibqp: qp's ib pointer also to get to device's qp address
+ */
+static int i40iw_destroy_qp(struct ib_qp *ibqp)
+{
+	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+
+	iwqp->destroyed = 1;
+
+	if (iwqp->ibqp_state >= IB_QPS_INIT && iwqp->ibqp_state < IB_QPS_RTS)
+		i40iw_next_iw_state(iwqp, I40IW_QP_STATE_ERROR, 0, 0, 0);
+
+	if (!iwqp->user_mode) {
+		if (iwqp->iwscq) {
+			i40iw_clean_cqes(iwqp, iwqp->iwscq);
+			if (iwqp->iwrcq != iwqp->iwscq)
+				i40iw_clean_cqes(iwqp, iwqp->iwrcq);
+		}
+	}
+
+	i40iw_rem_ref(&iwqp->ibqp);
+	return 0;
+}
+
+/**
+ * i40iw_setup_virt_qp - setup for allocation of virtual qp
+ * @dev: iwarp device
+ * @qp: qp ptr
+ * @init_info: initialize info to return
+ */
+static int i40iw_setup_virt_qp(struct i40iw_device *iwdev,
+			       struct i40iw_qp *iwqp,
+			       struct i40iw_qp_init_info *init_info)
+{
+	struct i40iw_pbl *iwpbl = iwqp->iwpbl;
+	struct i40iw_qp_mr *qpmr = &iwpbl->qp_mr;
+
+	iwqp->page = qpmr->sq_page;
+	init_info->shadow_area_pa = cpu_to_le64(qpmr->shadow);
+	if (iwpbl->pbl_allocated) {
+		init_info->virtual_map = true;
+		init_info->sq_pa = qpmr->sq_pbl.idx;
+		init_info->rq_pa = qpmr->rq_pbl.idx;
+	} else {
+		init_info->sq_pa = qpmr->sq_pbl.addr;
+		init_info->rq_pa = qpmr->rq_pbl.addr;
+	}
+	return 0;
+}
+
+/**
+ * i40iw_setup_kmode_qp - setup initialization for kernel mode qp
+ * @iwdev: iwarp device
+ * @iwqp: qp ptr (user or kernel)
+ * @info: initialize info to return
+ */
+static int i40iw_setup_kmode_qp(struct i40iw_device *iwdev,
+				struct i40iw_qp *iwqp,
+				struct i40iw_qp_init_info *info)
+{
+	struct i40iw_dma_mem *mem = &iwqp->kqp.dma_mem;
+	u32 sqdepth, rqdepth;
+	u32 sq_size, rq_size;
+	u8 sqshift, rqshift;
+	u32 size;
+	enum i40iw_status_code status;
+	struct i40iw_qp_uk_init_info *ukinfo = &info->qp_uk_init_info;
+
+	ukinfo->max_sq_frag_cnt = I40IW_MAX_WQ_FRAGMENT_COUNT;
+
+	sq_size = i40iw_qp_roundup(ukinfo->sq_size + 1);
+	rq_size = i40iw_qp_roundup(ukinfo->rq_size + 1);
+
+	status = i40iw_get_wqe_shift(sq_size, ukinfo->max_sq_frag_cnt, &sqshift);
+	if (!status)
+		status = i40iw_get_wqe_shift(rq_size, ukinfo->max_rq_frag_cnt, &rqshift);
+
+	if (status)
+		return -ENOSYS;
+
+	sqdepth = sq_size << sqshift;
+	rqdepth = rq_size << rqshift;
+
+	size = sqdepth * sizeof(struct i40iw_sq_uk_wr_trk_info) + (rqdepth << 3);
+	iwqp->kqp.wrid_mem = kzalloc(size, GFP_KERNEL);
+
+	ukinfo->sq_wrtrk_array = (struct i40iw_sq_uk_wr_trk_info *)iwqp->kqp.wrid_mem;
+	if (!ukinfo->sq_wrtrk_array)
+		return -ENOMEM;
+
+	ukinfo->rq_wrid_array = (u64 *)&ukinfo->sq_wrtrk_array[sqdepth];
+
+	size = (sqdepth + rqdepth) * I40IW_QP_WQE_MIN_SIZE;
+	size += (I40IW_SHADOW_AREA_SIZE << 3);
+
+	status = i40iw_allocate_dma_mem(iwdev->sc_dev.hw, mem, size, 256);
+	if (status) {
+		kfree(ukinfo->sq_wrtrk_array);
+		ukinfo->sq_wrtrk_array = NULL;
+		return -ENOMEM;
+	}
+
+	ukinfo->sq = mem->va;
+	info->sq_pa = mem->pa;
+
+	ukinfo->rq = &ukinfo->sq[sqdepth];
+	info->rq_pa = info->sq_pa + (sqdepth * I40IW_QP_WQE_MIN_SIZE);
+
+	ukinfo->shadow_area = ukinfo->rq[rqdepth].elem;
+	info->shadow_area_pa = info->rq_pa + (rqdepth * I40IW_QP_WQE_MIN_SIZE);
+
+	ukinfo->sq_size = sq_size;
+	ukinfo->rq_size = rq_size;
+	ukinfo->qp_id = iwqp->ibqp.qp_num;
+	return 0;
+}
+
+/**
+ * i40iw_create_qp - create qp
+ * @ibpd: ptr of pd
+ * @init_attr: attributes for qp
+ * @udata: user data for create qp
+ */
+static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
+				     struct ib_qp_init_attr *init_attr,
+				     struct ib_udata *udata)
+{
+	struct i40iw_pd *iwpd = to_iwpd(ibpd);
+	struct i40iw_device *iwdev = to_iwdev(ibpd->device);
+	struct i40iw_cqp *iwcqp = &iwdev->cqp;
+	struct i40iw_qp *iwqp;
+	struct i40iw_ucontext *ucontext;
+	struct i40iw_create_qp_req req;
+	struct i40iw_create_qp_resp uresp;
+	u32 qp_num = 0;
+	void *mem;
+	enum i40iw_status_code ret;
+	int err_code;
+	int sq_size;
+	int rq_size;
+	struct i40iw_sc_qp *qp;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_qp_init_info init_info;
+	struct i40iw_create_qp_info *qp_info;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	struct i40iw_qp_host_ctx_info *ctx_info;
+	struct i40iwarp_offload_info *iwarp_info;
+	unsigned long flags;
+
+	if (init_attr->create_flags)
+		return ERR_PTR(-EINVAL);
+	if (init_attr->cap.max_inline_data > I40IW_MAX_INLINE_DATA_SIZE)
+		init_attr->cap.max_inline_data = I40IW_MAX_INLINE_DATA_SIZE;
+
+	memset(&init_info, 0, sizeof(init_info));
+
+	sq_size = init_attr->cap.max_send_wr;
+	rq_size = init_attr->cap.max_recv_wr;
+
+	init_info.qp_uk_init_info.sq_size = sq_size;
+	init_info.qp_uk_init_info.rq_size = rq_size;
+	init_info.qp_uk_init_info.max_sq_frag_cnt = init_attr->cap.max_send_sge;
+	init_info.qp_uk_init_info.max_rq_frag_cnt = init_attr->cap.max_recv_sge;
+
+	mem = kzalloc(sizeof(*iwqp), GFP_KERNEL);
+	if (!mem)
+		return ERR_PTR(-ENOMEM);
+
+	iwqp = (struct i40iw_qp *)mem;
+	qp = &iwqp->sc_qp;
+	qp->back_qp = (void *)iwqp;
+	qp->push_idx = I40IW_INVALID_PUSH_PAGE_INDEX;
+
+	iwqp->ctx_info.iwarp_info = &iwqp->iwarp_info;
+
+	if (i40iw_allocate_dma_mem(dev->hw,
+				   &iwqp->q2_ctx_mem,
+				   I40IW_Q2_BUFFER_SIZE + I40IW_QP_CTX_SIZE,
+				   256)) {
+		i40iw_pr_err("dma_mem failed\n");
+		err_code = -ENOMEM;
+		goto error;
+	}
+
+	init_info.q2 = iwqp->q2_ctx_mem.va;
+	init_info.q2_pa = iwqp->q2_ctx_mem.pa;
+
+	init_info.host_ctx = (void *)init_info.q2 + I40IW_Q2_BUFFER_SIZE;
+	init_info.host_ctx_pa = init_info.q2_pa + I40IW_Q2_BUFFER_SIZE;
+
+	err_code = i40iw_alloc_resource(iwdev, iwdev->allocated_qps, iwdev->max_qp,
+					&qp_num, &iwdev->next_qp);
+	if (err_code) {
+		i40iw_pr_err("qp resource\n");
+		goto error;
+	}
+
+	iwqp->allocated_buffer = mem;
+	iwqp->iwdev = iwdev;
+	iwqp->iwpd = iwpd;
+	iwqp->ibqp.qp_num = qp_num;
+	qp = &iwqp->sc_qp;
+	iwqp->iwscq = to_iwcq(init_attr->send_cq);
+	iwqp->iwrcq = to_iwcq(init_attr->recv_cq);
+
+	iwqp->host_ctx.va = init_info.host_ctx;
+	iwqp->host_ctx.pa = init_info.host_ctx_pa;
+	iwqp->host_ctx.size = I40IW_QP_CTX_SIZE;
+
+	init_info.pd = &iwpd->sc_pd;
+	init_info.qp_uk_init_info.qp_id = iwqp->ibqp.qp_num;
+	iwqp->ctx_info.qp_compl_ctx = (uintptr_t)qp;
+
+	if (init_attr->qp_type != IB_QPT_RC) {
+		err_code = -ENOSYS;
+		goto error;
+	}
+	if (iwdev->push_mode)
+		i40iw_alloc_push_page(iwdev, qp);
+	if (udata) {
+		err_code = ib_copy_from_udata(&req, udata, sizeof(req));
+		if (err_code) {
+			i40iw_pr_err("ib_copy_from_data\n");
+			goto error;
+		}
+		iwqp->ctx_info.qp_compl_ctx = req.user_compl_ctx;
+		if (ibpd->uobject && ibpd->uobject->context) {
+			iwqp->user_mode = 1;
+			ucontext = to_ucontext(ibpd->uobject->context);
+
+			if (req.user_wqe_buffers) {
+				spin_lock_irqsave(
+				    &ucontext->qp_reg_mem_list_lock, flags);
+				iwqp->iwpbl = i40iw_get_pbl(
+				    (unsigned long)req.user_wqe_buffers,
+				    &ucontext->qp_reg_mem_list);
+				spin_unlock_irqrestore(
+				    &ucontext->qp_reg_mem_list_lock, flags);
+
+				if (!iwqp->iwpbl) {
+					err_code = -ENODATA;
+					i40iw_pr_err("no pbl info\n");
+					goto error;
+				}
+			}
+		}
+		err_code = i40iw_setup_virt_qp(iwdev, iwqp, &init_info);
+	} else {
+		err_code = i40iw_setup_kmode_qp(iwdev, iwqp, &init_info);
+	}
+
+	if (err_code) {
+		i40iw_pr_err("setup qp failed\n");
+		goto error;
+	}
+
+	init_info.type = I40IW_QP_TYPE_IWARP;
+	ret = dev->iw_priv_qp_ops->qp_init(qp, &init_info);
+	if (ret) {
+		err_code = -EPROTO;
+		i40iw_pr_err("qp_init fail\n");
+		goto error;
+	}
+	ctx_info = &iwqp->ctx_info;
+	iwarp_info = &iwqp->iwarp_info;
+	iwarp_info->rd_enable = true;
+	iwarp_info->wr_rdresp_en = true;
+	if (!iwqp->user_mode)
+		iwarp_info->priv_mode_en = true;
+	iwarp_info->ddp_ver = 1;
+	iwarp_info->rdmap_ver = 1;
+
+	ctx_info->iwarp_info_valid = true;
+	ctx_info->send_cq_num = iwqp->iwscq->sc_cq.cq_uk.cq_id;
+	ctx_info->rcv_cq_num = iwqp->iwrcq->sc_cq.cq_uk.cq_id;
+	if (qp->push_idx == I40IW_INVALID_PUSH_PAGE_INDEX) {
+		ctx_info->push_mode_en = false;
+	} else {
+		ctx_info->push_mode_en = true;
+		ctx_info->push_idx = qp->push_idx;
+	}
+
+	ret = dev->iw_priv_qp_ops->qp_setctx(&iwqp->sc_qp,
+					     (u64 *)iwqp->host_ctx.va,
+					     ctx_info);
+	ctx_info->iwarp_info_valid = false;
+	cqp_request = i40iw_get_cqp_request(iwcqp, true);
+	if (!cqp_request) {
+		err_code = -ENOMEM;
+		goto error;
+	}
+	cqp_info = &cqp_request->info;
+	qp_info = &cqp_request->info.in.u.qp_create.info;
+
+	memset(qp_info, 0, sizeof(*qp_info));
+
+	qp_info->cq_num_valid = true;
+	qp_info->next_iwarp_state = I40IW_QP_STATE_IDLE;
+
+	cqp_info->cqp_cmd = OP_QP_CREATE;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.qp_create.qp = qp;
+	cqp_info->in.u.qp_create.scratch = (uintptr_t)cqp_request;
+	ret = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (ret) {
+		i40iw_pr_err("CQP-OP QP create fail");
+		err_code = -EACCES;
+		goto error;
+	}
+
+	i40iw_add_ref(&iwqp->ibqp);
+	spin_lock_init(&iwqp->lock);
+	iwqp->sig_all = (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) ? 1 : 0;
+	iwdev->qp_table[qp_num] = iwqp;
+	i40iw_add_pdusecount(iwqp->iwpd);
+	if (ibpd->uobject && udata) {
+		memset(&uresp, 0, sizeof(uresp));
+		uresp.actual_sq_size = sq_size;
+		uresp.actual_rq_size = rq_size;
+		uresp.qp_id = qp_num;
+		uresp.push_idx = qp->push_idx;
+		err_code = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
+		if (err_code) {
+			i40iw_pr_err("copy_to_udata failed\n");
+			i40iw_destroy_qp(&iwqp->ibqp);
+			   /* let the completion of the qp destroy free the qp */
+			return ERR_PTR(err_code);
+		}
+	}
+
+	return &iwqp->ibqp;
+error:
+	i40iw_free_qp_resources(iwdev, iwqp, qp_num);
+	kfree(mem);
+	return ERR_PTR(err_code);
+}
+
+/**
+ * i40iw_query - query qp attributes
+ * @ibqp: qp pointer
+ * @attr: attributes pointer
+ * @attr_mask: Not used
+ * @init_attr: qp attributes to return
+ */
+static int i40iw_query_qp(struct ib_qp *ibqp,
+			  struct ib_qp_attr *attr,
+			  int attr_mask,
+			  struct ib_qp_init_attr *init_attr)
+{
+	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+	struct i40iw_sc_qp *qp = &iwqp->sc_qp;
+
+	attr->qp_access_flags = 0;
+	attr->cap.max_send_wr = qp->qp_uk.sq_size;
+	attr->cap.max_recv_wr = qp->qp_uk.rq_size;
+	attr->cap.max_recv_sge = 1;
+	attr->cap.max_inline_data = I40IW_MAX_INLINE_DATA_SIZE;
+	init_attr->event_handler = iwqp->ibqp.event_handler;
+	init_attr->qp_context = iwqp->ibqp.qp_context;
+	init_attr->send_cq = iwqp->ibqp.send_cq;
+	init_attr->recv_cq = iwqp->ibqp.recv_cq;
+	init_attr->srq = iwqp->ibqp.srq;
+	init_attr->cap = attr->cap;
+	return 0;
+}
+
+/**
+ * i40iw_hw_modify_qp - setup cqp for modify qp
+ * @iwdev: iwarp device
+ * @iwqp: qp ptr (user or kernel)
+ * @info: info for modify qp
+ * @wait: flag to wait or not for modify qp completion
+ */
+void i40iw_hw_modify_qp(struct i40iw_device *iwdev, struct i40iw_qp *iwqp,
+			struct i40iw_modify_qp_info *info, bool wait)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_modify_qp_info *m_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, wait);
+	if (!cqp_request)
+		return;
+
+	cqp_info = &cqp_request->info;
+	m_info = &cqp_info->in.u.qp_modify.info;
+	memcpy(m_info, info, sizeof(*m_info));
+	cqp_info->cqp_cmd = OP_QP_MODIFY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.qp_modify.qp = &iwqp->sc_qp;
+	cqp_info->in.u.qp_modify.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Modify QP fail");
+}
+
+/**
+ * i40iw_modify_qp - modify qp request
+ * @ibqp: qp's pointer for modify
+ * @attr: access attributes
+ * @attr_mask: state mask
+ * @udata: user data
+ */
+int i40iw_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		    int attr_mask, struct ib_udata *udata)
+{
+	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+	struct i40iw_device *iwdev = iwqp->iwdev;
+	struct i40iw_qp_host_ctx_info *ctx_info;
+	struct i40iwarp_offload_info *iwarp_info;
+	struct i40iw_modify_qp_info info;
+	u8 issue_modify_qp = 0;
+	u8 dont_wait = 0;
+	u32 err;
+	unsigned long flags;
+
+	memset(&info, 0, sizeof(info));
+	ctx_info = &iwqp->ctx_info;
+	iwarp_info = &iwqp->iwarp_info;
+
+	spin_lock_irqsave(&iwqp->lock, flags);
+
+	if (attr_mask & IB_QP_STATE) {
+		switch (attr->qp_state) {
+		case IB_QPS_INIT:
+		case IB_QPS_RTR:
+			if (iwqp->iwarp_state > (u32)I40IW_QP_STATE_IDLE) {
+				err = -EINVAL;
+				goto exit;
+			}
+			if (iwqp->iwarp_state == I40IW_QP_STATE_INVALID) {
+				info.next_iwarp_state = I40IW_QP_STATE_IDLE;
+				issue_modify_qp = 1;
+			}
+			break;
+		case IB_QPS_RTS:
+			if ((iwqp->iwarp_state > (u32)I40IW_QP_STATE_RTS) ||
+			    (!iwqp->cm_id)) {
+				err = -EINVAL;
+				goto exit;
+			}
+
+			issue_modify_qp = 1;
+			iwqp->hw_tcp_state = I40IW_TCP_STATE_ESTABLISHED;
+			iwqp->hte_added = 1;
+			info.next_iwarp_state = I40IW_QP_STATE_RTS;
+			info.tcp_ctx_valid = true;
+			info.ord_valid = true;
+			info.arp_cache_idx_valid = true;
+			info.cq_num_valid = true;
+			break;
+		case IB_QPS_SQD:
+			if (iwqp->hw_iwarp_state > (u32)I40IW_QP_STATE_RTS) {
+				err = 0;
+				goto exit;
+			}
+			if ((iwqp->iwarp_state == (u32)I40IW_QP_STATE_CLOSING) ||
+			    (iwqp->iwarp_state < (u32)I40IW_QP_STATE_RTS)) {
+				err = 0;
+				goto exit;
+			}
+			if (iwqp->iwarp_state > (u32)I40IW_QP_STATE_CLOSING) {
+				err = -EINVAL;
+				goto exit;
+			}
+			info.next_iwarp_state = I40IW_QP_STATE_CLOSING;
+			issue_modify_qp = 1;
+			break;
+		case IB_QPS_SQE:
+			if (iwqp->iwarp_state >= (u32)I40IW_QP_STATE_TERMINATE) {
+				err = -EINVAL;
+				goto exit;
+			}
+			info.next_iwarp_state = I40IW_QP_STATE_TERMINATE;
+			issue_modify_qp = 1;
+			break;
+		case IB_QPS_ERR:
+		case IB_QPS_RESET:
+			if (iwqp->iwarp_state == (u32)I40IW_QP_STATE_ERROR) {
+				err = -EINVAL;
+				goto exit;
+			}
+			if (iwqp->sc_qp.term_flags)
+				del_timer(&iwqp->terminate_timer);
+			info.next_iwarp_state = I40IW_QP_STATE_ERROR;
+			if ((iwqp->hw_tcp_state > I40IW_TCP_STATE_CLOSED) &&
+			    iwdev->iw_status &&
+			    (iwqp->hw_tcp_state != I40IW_TCP_STATE_TIME_WAIT))
+				info.reset_tcp_conn = true;
+			else
+				dont_wait = 1;
+			issue_modify_qp = 1;
+			info.next_iwarp_state = I40IW_QP_STATE_ERROR;
+			break;
+		default:
+			err = -EINVAL;
+			goto exit;
+		}
+
+		iwqp->ibqp_state = attr->qp_state;
+
+		if (issue_modify_qp)
+			iwqp->iwarp_state = info.next_iwarp_state;
+		else
+			info.next_iwarp_state = iwqp->iwarp_state;
+	}
+	if (attr_mask & IB_QP_ACCESS_FLAGS) {
+		ctx_info->iwarp_info_valid = true;
+		if (attr->qp_access_flags & IB_ACCESS_LOCAL_WRITE)
+			iwarp_info->wr_rdresp_en = true;
+		if (attr->qp_access_flags & IB_ACCESS_REMOTE_WRITE)
+			iwarp_info->wr_rdresp_en = true;
+		if (attr->qp_access_flags & IB_ACCESS_REMOTE_READ)
+			iwarp_info->rd_enable = true;
+		if (attr->qp_access_flags & IB_ACCESS_MW_BIND)
+			iwarp_info->bind_en = true;
+
+		if (iwqp->user_mode) {
+			iwarp_info->rd_enable = true;
+			iwarp_info->wr_rdresp_en = true;
+			iwarp_info->priv_mode_en = false;
+		}
+	}
+
+	if (ctx_info->iwarp_info_valid) {
+		struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+		int ret;
+
+		ctx_info->send_cq_num = iwqp->iwscq->sc_cq.cq_uk.cq_id;
+		ctx_info->rcv_cq_num = iwqp->iwrcq->sc_cq.cq_uk.cq_id;
+		ret = dev->iw_priv_qp_ops->qp_setctx(&iwqp->sc_qp,
+						     (u64 *)iwqp->host_ctx.va,
+						     ctx_info);
+		if (ret) {
+			i40iw_pr_err("setting QP context\n");
+			err = -EINVAL;
+			goto exit;
+		}
+	}
+
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+
+	if (issue_modify_qp)
+		i40iw_hw_modify_qp(iwdev, iwqp, &info, true);
+
+	if (issue_modify_qp && (iwqp->ibqp_state > IB_QPS_RTS)) {
+		if (dont_wait) {
+			if (iwqp->cm_id && iwqp->hw_tcp_state) {
+				spin_lock_irqsave(&iwqp->lock, flags);
+				iwqp->hw_tcp_state = I40IW_TCP_STATE_CLOSED;
+				iwqp->last_aeq = I40IW_AE_RESET_SENT;
+				spin_unlock_irqrestore(&iwqp->lock, flags);
+			}
+		}
+	}
+	return 0;
+exit:
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+	return err;
+}
+
+/**
+ * cq_free_resources - free up recources for cq
+ * @iwdev: iwarp device
+ * @iwcq: cq ptr
+ */
+static void cq_free_resources(struct i40iw_device *iwdev, struct i40iw_cq *iwcq)
+{
+	struct i40iw_sc_cq *cq = &iwcq->sc_cq;
+
+	if (!iwcq->user_mode)
+		i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwcq->kmem);
+	i40iw_free_resource(iwdev, iwdev->allocated_cqs, cq->cq_uk.cq_id);
+}
+
+/**
+ * cq_wq_destroy - send cq destroy cqp
+ * @iwdev: iwarp device
+ * @cq: hardware control cq
+ */
+static void cq_wq_destroy(struct i40iw_device *iwdev, struct i40iw_sc_cq *cq)
+{
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return;
+
+	cqp_info = &cqp_request->info;
+
+	cqp_info->cqp_cmd = OP_CQ_DESTROY;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.cq_destroy.cq = cq;
+	cqp_info->in.u.cq_destroy.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP Destroy QP fail");
+}
+
+/**
+ * i40iw_destroy_cq - destroy cq
+ * @ib_cq: cq pointer
+ */
+static int i40iw_destroy_cq(struct ib_cq *ib_cq)
+{
+	struct i40iw_cq *iwcq;
+	struct i40iw_device *iwdev;
+	struct i40iw_sc_cq *cq;
+
+	if (!ib_cq) {
+		i40iw_pr_err("ib_cq == NULL\n");
+		return 0;
+	}
+
+	iwcq = to_iwcq(ib_cq);
+	iwdev = to_iwdev(ib_cq->device);
+	cq = &iwcq->sc_cq;
+	cq_wq_destroy(iwdev, cq);
+	cq_free_resources(iwdev, iwcq);
+	kfree(iwcq);
+	return 0;
+}
+
+/**
+ * i40iw_create_cq - create cq
+ * @ibdev: device pointer from stack
+ * @attr: attributes for cq
+ * @context: user context created during alloc
+ * @udata: user data
+ */
+static struct ib_cq *i40iw_create_cq(struct ib_device *ibdev,
+				     const struct ib_cq_init_attr *attr,
+				     struct ib_ucontext *context,
+				     struct ib_udata *udata)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct i40iw_cq *iwcq;
+	struct i40iw_pbl *iwpbl;
+	u32 cq_num = 0;
+	struct i40iw_sc_cq *cq;
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_cq_init_info info;
+	enum i40iw_status_code status;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	struct i40iw_cq_uk_init_info *ukinfo = &info.cq_uk_init_info;
+	unsigned long flags;
+	int err_code;
+	int entries = attr->cqe;
+
+	if (entries > iwdev->max_cqe)
+		return ERR_PTR(-EINVAL);
+
+	iwcq = kzalloc(sizeof(*iwcq), GFP_KERNEL);
+	if (!iwcq)
+		return ERR_PTR(-ENOMEM);
+
+	memset(&info, 0, sizeof(info));
+
+	err_code = i40iw_alloc_resource(iwdev, iwdev->allocated_cqs,
+					iwdev->max_cq, &cq_num,
+					&iwdev->next_cq);
+	if (err_code)
+		goto error;
+
+	cq = &iwcq->sc_cq;
+	cq->back_cq = (void *)iwcq;
+	spin_lock_init(&iwcq->lock);
+
+	info.dev = dev;
+	ukinfo->cq_size = max(entries, 4);
+	ukinfo->cq_id = cq_num;
+	iwcq->ibcq.cqe = info.cq_uk_init_info.cq_size;
+	info.ceqe_mask = 0;
+	info.ceq_id = 0;
+	info.ceq_id_valid = true;
+	info.ceqe_mask = 1;
+	info.type = I40IW_CQ_TYPE_IWARP;
+	if (context) {
+		struct i40iw_ucontext *ucontext;
+		struct i40iw_create_cq_req req;
+		struct i40iw_cq_mr *cqmr;
+
+		memset(&req, 0, sizeof(req));
+		iwcq->user_mode = true;
+		ucontext = to_ucontext(context);
+		if (ib_copy_from_udata(&req, udata, sizeof(struct i40iw_create_cq_req)))
+			goto cq_free_resources;
+
+		spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
+		iwpbl = i40iw_get_pbl((unsigned long)req.user_cq_buffer,
+				      &ucontext->cq_reg_mem_list);
+		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
+		if (!iwpbl) {
+			err_code = -EPROTO;
+			goto cq_free_resources;
+		}
+
+		iwcq->iwpbl = iwpbl;
+		iwcq->cq_mem_size = 0;
+		cqmr = &iwpbl->cq_mr;
+		info.shadow_area_pa = cpu_to_le64(cqmr->shadow);
+		if (iwpbl->pbl_allocated) {
+			info.virtual_map = true;
+			info.pbl_chunk_size = 1;
+			info.first_pm_pbl_idx = cqmr->cq_pbl.idx;
+		} else {
+			info.cq_base_pa = cqmr->cq_pbl.addr;
+		}
+	} else {
+		/* Kmode allocations */
+		int rsize;
+		int shadow;
+
+		rsize = info.cq_uk_init_info.cq_size * sizeof(struct i40iw_cqe);
+		rsize = round_up(rsize, 256);
+		shadow = I40IW_SHADOW_AREA_SIZE << 3;
+		status = i40iw_allocate_dma_mem(dev->hw, &iwcq->kmem,
+						rsize + shadow, 256);
+		if (status) {
+			err_code = -ENOMEM;
+			goto cq_free_resources;
+		}
+		ukinfo->cq_base = iwcq->kmem.va;
+		info.cq_base_pa = iwcq->kmem.pa;
+		info.shadow_area_pa = info.cq_base_pa + rsize;
+		ukinfo->shadow_area = iwcq->kmem.va + rsize;
+	}
+
+	if (dev->iw_priv_cq_ops->cq_init(cq, &info)) {
+		i40iw_pr_err("init cq fail\n");
+		err_code = -EPROTO;
+		goto cq_free_resources;
+	}
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request) {
+		err_code = -ENOMEM;
+		goto cq_free_resources;
+	}
+
+	cqp_info = &cqp_request->info;
+	cqp_info->cqp_cmd = OP_CQ_CREATE;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.cq_create.cq = cq;
+	cqp_info->in.u.cq_create.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status) {
+		i40iw_pr_err("CQP-OP Create QP fail");
+		err_code = -EPROTO;
+		goto cq_free_resources;
+	}
+
+	if (context) {
+		struct i40iw_create_cq_resp resp;
+
+		memset(&resp, 0, sizeof(resp));
+		resp.cq_id = info.cq_uk_init_info.cq_id;
+		resp.cq_size = info.cq_uk_init_info.cq_size;
+		if (ib_copy_to_udata(udata, &resp, sizeof(resp))) {
+			i40iw_pr_err("copy to user data\n");
+			err_code = -EPROTO;
+			goto cq_destroy;
+		}
+	}
+
+	return (struct ib_cq *)iwcq;
+
+cq_destroy:
+	cq_wq_destroy(iwdev, cq);
+cq_free_resources:
+	cq_free_resources(iwdev, iwcq);
+error:
+	kfree(iwcq);
+	return ERR_PTR(err_code);
+}
+
+/**
+ * i40iw_get_user_access - get hw access from IB access
+ * @acc: IB access to return hw access
+ */
+static inline u16 i40iw_get_user_access(int acc)
+{
+	u16 access = 0;
+
+	access |= (acc & IB_ACCESS_LOCAL_WRITE) ? I40IW_ACCESS_FLAGS_LOCALWRITE : 0;
+	access |= (acc & IB_ACCESS_REMOTE_WRITE) ? I40IW_ACCESS_FLAGS_REMOTEWRITE : 0;
+	access |= (acc & IB_ACCESS_REMOTE_READ) ? I40IW_ACCESS_FLAGS_REMOTEREAD : 0;
+	access |= (acc & IB_ACCESS_MW_BIND) ? I40IW_ACCESS_FLAGS_BIND_WINDOW : 0;
+	return access;
+}
+
+/**
+ * i40iw_free_stag - free stag resource
+ * @iwdev: iwarp device
+ * @stag: stag to free
+ */
+static void i40iw_free_stag(struct i40iw_device *iwdev, u32 stag)
+{
+	u32 stag_idx;
+
+	stag_idx = (stag & iwdev->mr_stagmask) >> I40IW_CQPSQ_STAG_IDX_SHIFT;
+	i40iw_free_resource(iwdev, iwdev->allocated_mrs, stag_idx);
+}
+
+/**
+ * i40iw_create_stag - create random stag
+ * @iwdev: iwarp device
+ */
+static u32 i40iw_create_stag(struct i40iw_device *iwdev)
+{
+	u32 stag = 0;
+	u32 stag_index = 0;
+	u32 next_stag_index;
+	u32 driver_key;
+	u32 random;
+	u8 consumer_key;
+	int ret;
+
+	get_random_bytes(&random, sizeof(random));
+	consumer_key = (u8)random;
+
+	driver_key = random & ~iwdev->mr_stagmask;
+	next_stag_index = (random & iwdev->mr_stagmask) >> 8;
+	next_stag_index %= iwdev->max_mr;
+
+	ret = i40iw_alloc_resource(iwdev,
+				   iwdev->allocated_mrs, iwdev->max_mr,
+				   &stag_index, &next_stag_index);
+	if (!ret) {
+		stag = stag_index << I40IW_CQPSQ_STAG_IDX_SHIFT;
+		stag |= driver_key;
+		stag += (u32)consumer_key;
+	}
+	return stag;
+}
+
+/**
+ * i40iw_next_pbl_addr - Get next pbl address
+ * @palloc: Poiner to allocated pbles
+ * @pbl: pointer to a pble
+ * @pinfo: info pointer
+ * @idx: index
+ */
+static inline u64 *i40iw_next_pbl_addr(struct i40iw_pble_alloc *palloc,
+				       u64 *pbl,
+				       struct i40iw_pble_info **pinfo,
+				       u32 *idx)
+{
+	*idx += 1;
+	if ((!(*pinfo)) || (*idx != (*pinfo)->cnt))
+		return ++pbl;
+	*idx = 0;
+	(*pinfo)++;
+	return (u64 *)(*pinfo)->addr;
+}
+
+/**
+ * i40iw_copy_user_pgaddrs - copy user page address to pble's os locally
+ * @iwmr: iwmr for IB's user page addresses
+ * @pbl: ple pointer to save 1 level or 0 level pble
+ * @level: indicated level 0, 1 or 2
+ */
+static void i40iw_copy_user_pgaddrs(struct i40iw_mr *iwmr,
+				    u64 *pbl,
+				    enum i40iw_pble_level level)
+{
+	struct ib_umem *region = iwmr->region;
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	int chunk_pages, entry, pg_shift, i;
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	struct i40iw_pble_info *pinfo;
+	struct scatterlist *sg;
+	u32 idx = 0;
+
+	pinfo = (level == I40IW_LEVEL_1) ? NULL : palloc->level2.leaf;
+	pg_shift = ffs(region->page_size) - 1;
+	for_each_sg(region->sg_head.sgl, sg, region->nmap, entry) {
+		chunk_pages = sg_dma_len(sg) >> pg_shift;
+		if ((iwmr->type == IW_MEMREG_TYPE_QP) &&
+		    !iwpbl->qp_mr.sq_page)
+			iwpbl->qp_mr.sq_page = sg_page(sg);
+		for (i = 0; i < chunk_pages; i++) {
+			*pbl = cpu_to_le64(sg_dma_address(sg) + region->page_size * i);
+			pbl = i40iw_next_pbl_addr(palloc, pbl, &pinfo, &idx);
+		}
+	}
+}
+
+/**
+ * i40iw_setup_pbles - copy user pg address to pble's
+ * @iwdev: iwarp device
+ * @iwmr: mr pointer for this memory registration
+ * @use_pbles: flag if to use pble's or memory (level 0)
+ */
+static int i40iw_setup_pbles(struct i40iw_device *iwdev,
+			     struct i40iw_mr *iwmr,
+			     bool use_pbles)
+{
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	struct i40iw_pble_info *pinfo;
+	u64 *pbl;
+	enum i40iw_status_code status;
+	enum i40iw_pble_level level = I40IW_LEVEL_1;
+
+	if (!use_pbles && (iwmr->page_cnt > MAX_SAVE_PAGE_ADDRS))
+		return -ENOMEM;
+
+	if (use_pbles) {
+		mutex_lock(&iwdev->pbl_mutex);
+		status = i40iw_get_pble(&iwdev->sc_dev, iwdev->pble_rsrc, palloc, iwmr->page_cnt);
+		mutex_unlock(&iwdev->pbl_mutex);
+		if (status)
+			return -ENOMEM;
+
+		iwpbl->pbl_allocated = true;
+		level = palloc->level;
+		pinfo = (level == I40IW_LEVEL_1) ? &palloc->level1 : palloc->level2.leaf;
+		pbl = (u64 *)pinfo->addr;
+	} else {
+		pbl = iwmr->pgaddrmem;
+	}
+
+	i40iw_copy_user_pgaddrs(iwmr, pbl, level);
+	return 0;
+}
+
+/**
+ * i40iw_handle_q_mem - handle memory for qp and cq
+ * @iwdev: iwarp device
+ * @req: information for q memory management
+ * @iwpbl: pble struct
+ * @use_pbles: flag to use pble
+ */
+static int i40iw_handle_q_mem(struct i40iw_device *iwdev,
+			      struct i40iw_mem_reg_req *req,
+			      struct i40iw_pbl *iwpbl,
+			      bool use_pbles)
+{
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	struct i40iw_mr *iwmr = iwpbl->iwmr;
+	struct i40iw_qp_mr *qpmr = &iwpbl->qp_mr;
+	struct i40iw_cq_mr *cqmr = &iwpbl->cq_mr;
+	struct i40iw_hmc_pble *hmc_p;
+	u64 *arr = iwmr->pgaddrmem;
+	int err;
+	int total;
+
+	total = req->sq_pages + req->rq_pages + req->cq_pages;
+
+	err = i40iw_setup_pbles(iwdev, iwmr, use_pbles);
+	if (err)
+		return err;
+	if (use_pbles && (palloc->level != I40IW_LEVEL_1)) {
+		i40iw_free_pble(iwdev->pble_rsrc, palloc);
+		iwpbl->pbl_allocated = false;
+		return -ENOMEM;
+	}
+
+	if (use_pbles)
+		arr = (u64 *)palloc->level1.addr;
+	if (req->reg_type == IW_MEMREG_TYPE_QP) {
+		hmc_p = &qpmr->sq_pbl;
+		qpmr->shadow = (dma_addr_t)arr[total];
+		if (use_pbles) {
+			hmc_p->idx = palloc->level1.idx;
+			hmc_p = &qpmr->rq_pbl;
+			hmc_p->idx = palloc->level1.idx + req->sq_pages;
+		} else {
+			hmc_p->addr = arr[0];
+			hmc_p = &qpmr->rq_pbl;
+			hmc_p->addr = arr[1];
+		}
+	} else {		/* CQ */
+		hmc_p = &cqmr->cq_pbl;
+		cqmr->shadow = (dma_addr_t)arr[total];
+		if (use_pbles)
+			hmc_p->idx = palloc->level1.idx;
+		else
+			hmc_p->addr = arr[0];
+	}
+	return err;
+}
+
+/**
+ * i40iw_hwreg_mr - send cqp command for memory registration
+ * @iwdev: iwarp device
+ * @iwmr: iwarp mr pointer
+ * @access: access for MR
+ */
+static int i40iw_hwreg_mr(struct i40iw_device *iwdev,
+			  struct i40iw_mr *iwmr,
+			  u16 access)
+{
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	struct i40iw_reg_ns_stag_info *stag_info;
+	struct i40iw_pd *iwpd = to_iwpd(iwmr->ibmr.pd);
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	enum i40iw_status_code status;
+	int err = 0;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return -ENOMEM;
+
+	cqp_info = &cqp_request->info;
+	stag_info = &cqp_info->in.u.mr_reg_non_shared.info;
+	memset(stag_info, 0, sizeof(*stag_info));
+	stag_info->va = (void *)(unsigned long)iwpbl->user_base;
+	stag_info->stag_idx = iwmr->stag >> I40IW_CQPSQ_STAG_IDX_SHIFT;
+	stag_info->stag_key = (u8)iwmr->stag;
+	stag_info->total_len = iwmr->length;
+	stag_info->access_rights = access;
+	stag_info->pd_id = iwpd->sc_pd.pd_id;
+	stag_info->addr_type = I40IW_ADDR_TYPE_VA_BASED;
+
+	if (iwmr->page_cnt > 1) {
+		if (palloc->level == I40IW_LEVEL_1) {
+			stag_info->first_pm_pbl_index = palloc->level1.idx;
+			stag_info->chunk_size = 1;
+		} else {
+			stag_info->first_pm_pbl_index = palloc->level2.root.idx;
+			stag_info->chunk_size = 3;
+		}
+	} else {
+		stag_info->reg_addr_pa = iwmr->pgaddrmem[0];
+	}
+
+	cqp_info->cqp_cmd = OP_MR_REG_NON_SHARED;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.mr_reg_non_shared.dev = &iwdev->sc_dev;
+	cqp_info->in.u.mr_reg_non_shared.scratch = (uintptr_t)cqp_request;
+
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status) {
+		err = -ENOMEM;
+		i40iw_pr_err("CQP-OP MR Reg fail");
+	}
+	return err;
+}
+
+/**
+ * i40iw_reg_user_mr - Register a user memory region
+ * @pd: ptr of pd
+ * @start: virtual start address
+ * @length: length of mr
+ * @virt: virtual address
+ * @acc: access of mr
+ * @udata: user data
+ */
+static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
+				       u64 start,
+				       u64 length,
+				       u64 virt,
+				       int acc,
+				       struct ib_udata *udata)
+{
+	struct i40iw_pd *iwpd = to_iwpd(pd);
+	struct i40iw_device *iwdev = to_iwdev(pd->device);
+	struct i40iw_ucontext *ucontext;
+	struct i40iw_pble_alloc *palloc;
+	struct i40iw_pbl *iwpbl;
+	struct i40iw_mr *iwmr;
+	struct ib_umem *region;
+	struct i40iw_mem_reg_req req;
+	u32 pbl_depth = 0;
+	u32 stag = 0;
+	u16 access;
+	u32 region_length;
+	bool use_pbles = false;
+	unsigned long flags;
+	int err = -ENOSYS;
+
+	region = ib_umem_get(pd->uobject->context, start, length, acc, 0);
+	if (IS_ERR(region))
+		return (struct ib_mr *)region;
+
+	if (ib_copy_from_udata(&req, udata, sizeof(req))) {
+		ib_umem_release(region);
+		return ERR_PTR(-EFAULT);
+	}
+
+	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
+	if (!iwmr) {
+		ib_umem_release(region);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	iwpbl = &iwmr->iwpbl;
+	iwpbl->iwmr = iwmr;
+	iwmr->region = region;
+	iwmr->ibmr.pd = pd;
+	iwmr->ibmr.device = pd->device;
+	ucontext = to_ucontext(pd->uobject->context);
+	region_length = region->length + (start & 0xfff);
+	pbl_depth = region_length >> 12;
+	pbl_depth += (region_length & (4096 - 1)) ? 1 : 0;
+	iwmr->length = region->length;
+
+	iwpbl->user_base = virt;
+	palloc = &iwpbl->pble_alloc;
+
+	iwmr->type = req.reg_type;
+	iwmr->page_cnt = pbl_depth;
+
+	switch (req.reg_type) {
+	case IW_MEMREG_TYPE_QP:
+		use_pbles = ((req.sq_pages + req.rq_pages) > 2);
+		err = i40iw_handle_q_mem(iwdev, &req, iwpbl, use_pbles);
+		if (err)
+			goto error;
+		spin_lock_irqsave(&ucontext->qp_reg_mem_list_lock, flags);
+		list_add_tail(&iwpbl->list, &ucontext->qp_reg_mem_list);
+		spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
+		break;
+	case IW_MEMREG_TYPE_CQ:
+		use_pbles = (req.cq_pages > 1);
+		err = i40iw_handle_q_mem(iwdev, &req, iwpbl, use_pbles);
+		if (err)
+			goto error;
+
+		spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
+		list_add_tail(&iwpbl->list, &ucontext->cq_reg_mem_list);
+		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
+		break;
+	case IW_MEMREG_TYPE_MEM:
+		access = I40IW_ACCESS_FLAGS_LOCALREAD;
+
+		use_pbles = (iwmr->page_cnt != 1);
+		err = i40iw_setup_pbles(iwdev, iwmr, use_pbles);
+		if (err)
+			goto error;
+
+		access |= i40iw_get_user_access(acc);
+		stag = i40iw_create_stag(iwdev);
+		if (!stag) {
+			err = -ENOMEM;
+			goto error;
+		}
+
+		iwmr->stag = stag;
+		iwmr->ibmr.rkey = stag;
+		iwmr->ibmr.lkey = stag;
+
+		err = i40iw_hwreg_mr(iwdev, iwmr, access);
+		if (err) {
+			i40iw_free_stag(iwdev, stag);
+			goto error;
+		}
+		break;
+	default:
+		goto error;
+	}
+
+	iwmr->type = req.reg_type;
+	if (req.reg_type == IW_MEMREG_TYPE_MEM)
+		i40iw_add_pdusecount(iwpd);
+	return &iwmr->ibmr;
+
+error:
+	if (palloc->level != I40IW_LEVEL_0)
+		i40iw_free_pble(iwdev->pble_rsrc, palloc);
+	ib_umem_release(region);
+	kfree(iwmr);
+	return ERR_PTR(err);
+}
+
+/**
+ * i40iw_reg_phys_mr - register kernel physical memory
+ * @pd: ibpd pointer
+ * @addr: physical address of memory to register
+ * @size: size of memory to register
+ * @acc: Access rights
+ * @iova_start: start of virtual address for physical buffers
+ */
+struct ib_mr *i40iw_reg_phys_mr(struct ib_pd *pd,
+				u64 addr,
+				u64 size,
+				int acc,
+				u64 *iova_start)
+{
+	struct i40iw_pd *iwpd = to_iwpd(pd);
+	struct i40iw_device *iwdev = to_iwdev(pd->device);
+	struct i40iw_pbl *iwpbl;
+	struct i40iw_mr *iwmr;
+	enum i40iw_status_code status;
+	u32 stag;
+	u16 access = I40IW_ACCESS_FLAGS_LOCALREAD;
+	int ret;
+
+	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
+	if (!iwmr)
+		return ERR_PTR(-ENOMEM);
+	iwmr->ibmr.pd = pd;
+	iwmr->ibmr.device = pd->device;
+	iwpbl = &iwmr->iwpbl;
+	iwpbl->iwmr = iwmr;
+	iwmr->type = IW_MEMREG_TYPE_MEM;
+	iwpbl->user_base = *iova_start;
+	stag = i40iw_create_stag(iwdev);
+	if (!stag) {
+		ret = -EOVERFLOW;
+		goto err;
+	}
+	access |= i40iw_get_user_access(acc);
+	iwmr->stag = stag;
+	iwmr->ibmr.rkey = stag;
+	iwmr->ibmr.lkey = stag;
+	iwmr->page_cnt = 1;
+	iwmr->pgaddrmem[0]  = addr;
+	status = i40iw_hwreg_mr(iwdev, iwmr, access);
+	if (status) {
+		i40iw_free_stag(iwdev, stag);
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	i40iw_add_pdusecount(iwpd);
+	return &iwmr->ibmr;
+ err:
+	kfree(iwmr);
+	return ERR_PTR(ret);
+}
+
+/**
+ * i40iw_get_dma_mr - register physical mem
+ * @pd: ptr of pd
+ * @acc: access for memory
+ */
+static struct ib_mr *i40iw_get_dma_mr(struct ib_pd *pd, int acc)
+{
+	u64 kva = 0;
+
+	return i40iw_reg_phys_mr(pd, 0, 0xffffffffffULL, acc, &kva);
+}
+
+/**
+ * i40iw_del_mem_list - Deleting pbl list entries for CQ/QP
+ * @iwmr: iwmr for IB's user page addresses
+ * @ucontext: ptr to user context
+ */
+static void i40iw_del_memlist(struct i40iw_mr *iwmr,
+			      struct i40iw_ucontext *ucontext)
+{
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	unsigned long flags;
+
+	switch (iwmr->type) {
+	case IW_MEMREG_TYPE_CQ:
+		spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
+		if (!list_empty(&ucontext->cq_reg_mem_list))
+			list_del(&iwpbl->list);
+		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
+		break;
+	case IW_MEMREG_TYPE_QP:
+		spin_lock_irqsave(&ucontext->qp_reg_mem_list_lock, flags);
+		if (!list_empty(&ucontext->qp_reg_mem_list))
+			list_del(&iwpbl->list);
+		spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
+		break;
+	default:
+		break;
+	}
+}
+
+/**
+ * i40iw_dereg_mr - deregister mr
+ * @ib_mr: mr ptr for dereg
+ */
+static int i40iw_dereg_mr(struct ib_mr *ib_mr)
+{
+	struct ib_pd *ibpd = ib_mr->pd;
+	struct i40iw_pd *iwpd = to_iwpd(ibpd);
+	struct i40iw_mr *iwmr = to_iwmr(ib_mr);
+	struct i40iw_device *iwdev = to_iwdev(ib_mr->device);
+	enum i40iw_status_code status;
+	struct i40iw_dealloc_stag_info *info;
+	struct i40iw_pbl *iwpbl = &iwmr->iwpbl;
+	struct i40iw_pble_alloc *palloc = &iwpbl->pble_alloc;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+	u32 stag_idx;
+
+	if (iwmr->region)
+		ib_umem_release(iwmr->region);
+
+	if (iwmr->type != IW_MEMREG_TYPE_MEM) {
+		if (ibpd->uobject) {
+			struct i40iw_ucontext *ucontext;
+
+			ucontext = to_ucontext(ibpd->uobject->context);
+			i40iw_del_memlist(iwmr, ucontext);
+		}
+		if (iwpbl->pbl_allocated)
+			i40iw_free_pble(iwdev->pble_rsrc, palloc);
+		kfree(iwpbl->iwmr);
+		iwpbl->iwmr = NULL;
+		return 0;
+	}
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, true);
+	if (!cqp_request)
+		return -ENOMEM;
+
+	cqp_info = &cqp_request->info;
+	info = &cqp_info->in.u.dealloc_stag.info;
+	memset(info, 0, sizeof(*info));
+
+	info->pd_id = cpu_to_le32(iwpd->sc_pd.pd_id & 0x00007fff);
+	info->stag_idx = RS_64_1(ib_mr->rkey, I40IW_CQPSQ_STAG_IDX_SHIFT);
+	stag_idx = info->stag_idx;
+	info->mr = true;
+	if (iwpbl->pbl_allocated)
+		info->dealloc_pbl = true;
+
+	cqp_info->cqp_cmd = OP_DEALLOC_STAG;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.dealloc_stag.dev = &iwdev->sc_dev;
+	cqp_info->in.u.dealloc_stag.scratch = (uintptr_t)cqp_request;
+	status = i40iw_handle_cqp_op(iwdev, cqp_request);
+	if (status)
+		i40iw_pr_err("CQP-OP dealloc failed for stag_idx = 0x%x\n", stag_idx);
+	i40iw_rem_pdusecount(iwpd, iwdev);
+	i40iw_free_stag(iwdev, iwmr->stag);
+	if (iwpbl->pbl_allocated)
+		i40iw_free_pble(iwdev->pble_rsrc, palloc);
+	kfree(iwmr);
+	return 0;
+}
+
+/**
+ * i40iw_show_rev
+ */
+static ssize_t i40iw_show_rev(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct i40iw_ib_device *iwibdev = container_of(dev,
+						       struct i40iw_ib_device,
+						       ibdev.dev);
+	u32 hw_rev = iwibdev->iwdev->sc_dev.hw_rev;
+
+	return sprintf(buf, "%x\n", hw_rev);
+}
+
+/**
+ * i40iw_show_fw_ver
+ */
+static ssize_t i40iw_show_fw_ver(struct device *dev,
+				 struct device_attribute *attr, char *buf)
+{
+	u32 firmware_version = I40IW_FW_VERSION;
+
+	return sprintf(buf, "%u.%u\n", firmware_version,
+		       (firmware_version & 0x000000ff));
+}
+
+/**
+ * i40iw_show_hca
+ */
+static ssize_t i40iw_show_hca(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	return sprintf(buf, "I40IW\n");
+}
+
+/**
+ * i40iw_show_board
+ */
+static ssize_t i40iw_show_board(struct device *dev,
+				struct device_attribute *attr,
+				char *buf)
+{
+	return sprintf(buf, "%.*s\n", 32, "I40IW Board ID");
+}
+
+static DEVICE_ATTR(hw_rev, S_IRUGO, i40iw_show_rev, NULL);
+static DEVICE_ATTR(fw_ver, S_IRUGO, i40iw_show_fw_ver, NULL);
+static DEVICE_ATTR(hca_type, S_IRUGO, i40iw_show_hca, NULL);
+static DEVICE_ATTR(board_id, S_IRUGO, i40iw_show_board, NULL);
+
+static struct device_attribute *i40iw_dev_attributes[] = {
+	&dev_attr_hw_rev,
+	&dev_attr_fw_ver,
+	&dev_attr_hca_type,
+	&dev_attr_board_id
+};
+
+/**
+ * i40iw_copy_sg_list - copy sg list for qp
+ * @sg_list: copied into sg_list
+ * @sgl: copy from sgl
+ * @num_sges: count of sg entries
+ */
+static void i40iw_copy_sg_list(struct i40iw_sge *sg_list, struct ib_sge *sgl, int num_sges)
+{
+	unsigned int i;
+
+	for (i = 0; (i < num_sges) && (i < I40IW_MAX_WQ_FRAGMENT_COUNT); i++) {
+		sg_list[i].tag_off = sgl[i].addr;
+		sg_list[i].len = sgl[i].length;
+		sg_list[i].stag = sgl[i].lkey;
+	}
+}
+
+/**
+ * i40iw_post_send -  kernel application wr
+ * @ibqp: qp ptr for wr
+ * @ib_wr: work request ptr
+ * @bad_wr: return of bad wr if err
+ */
+static int i40iw_post_send(struct ib_qp *ibqp,
+			   struct ib_send_wr *ib_wr,
+			   struct ib_send_wr **bad_wr)
+{
+	struct i40iw_qp *iwqp;
+	struct i40iw_qp_uk *ukqp;
+	struct i40iw_post_sq_info info;
+	enum i40iw_status_code ret;
+	int err = 0;
+	unsigned long flags;
+
+	iwqp = (struct i40iw_qp *)ibqp;
+	ukqp = &iwqp->sc_qp.qp_uk;
+
+	spin_lock_irqsave(&iwqp->lock, flags);
+	while (ib_wr) {
+		memset(&info, 0, sizeof(info));
+		info.wr_id = (u64)(ib_wr->wr_id);
+		if ((ib_wr->send_flags & IB_SEND_SIGNALED) || iwqp->sig_all)
+			info.signaled = true;
+		if (ib_wr->send_flags & IB_SEND_FENCE)
+			info.read_fence = true;
+
+		switch (ib_wr->opcode) {
+		case IB_WR_SEND:
+			if (ib_wr->send_flags & IB_SEND_SOLICITED)
+				info.op_type = I40IW_OP_TYPE_SEND_SOL;
+			else
+				info.op_type = I40IW_OP_TYPE_SEND;
+
+			if (ib_wr->send_flags & IB_SEND_INLINE) {
+				info.op.inline_send.data = (void *)(unsigned long)ib_wr->sg_list[0].addr;
+				info.op.inline_send.len = ib_wr->sg_list[0].length;
+				ret = ukqp->ops.iw_inline_send(ukqp, &info, rdma_wr(ib_wr)->rkey, false);
+			} else {
+				info.op.send.num_sges = ib_wr->num_sge;
+				info.op.send.sg_list = (struct i40iw_sge *)ib_wr->sg_list;
+				ret = ukqp->ops.iw_send(ukqp, &info, rdma_wr(ib_wr)->rkey, false);
+			}
+
+			if (ret)
+				err = -EIO;
+			break;
+		case IB_WR_RDMA_WRITE:
+			info.op_type = I40IW_OP_TYPE_RDMA_WRITE;
+
+			if (ib_wr->send_flags & IB_SEND_INLINE) {
+				info.op.inline_rdma_write.data = (void *)(unsigned long)ib_wr->sg_list[0].addr;
+				info.op.inline_rdma_write.len = ib_wr->sg_list[0].length;
+				info.op.inline_rdma_write.rem_addr.tag_off = rdma_wr(ib_wr)->remote_addr;
+				info.op.inline_rdma_write.rem_addr.stag = rdma_wr(ib_wr)->rkey;
+				info.op.inline_rdma_write.rem_addr.len = ib_wr->sg_list->length;
+				ret = ukqp->ops.iw_inline_rdma_write(ukqp, &info, false);
+			} else {
+				info.op.rdma_write.lo_sg_list = (void *)ib_wr->sg_list;
+				info.op.rdma_write.num_lo_sges = ib_wr->num_sge;
+				info.op.rdma_write.rem_addr.tag_off = rdma_wr(ib_wr)->remote_addr;
+				info.op.rdma_write.rem_addr.stag = rdma_wr(ib_wr)->rkey;
+				info.op.rdma_write.rem_addr.len = ib_wr->sg_list->length;
+				ret = ukqp->ops.iw_rdma_write(ukqp, &info, false);
+			}
+
+			if (ret)
+				err = -EIO;
+			break;
+		case IB_WR_RDMA_READ:
+			info.op_type = I40IW_OP_TYPE_RDMA_READ;
+			info.op.rdma_read.rem_addr.tag_off = rdma_wr(ib_wr)->remote_addr;
+			info.op.rdma_read.rem_addr.stag = rdma_wr(ib_wr)->rkey;
+			info.op.rdma_read.rem_addr.len = ib_wr->sg_list->length;
+			info.op.rdma_read.lo_addr.tag_off = ib_wr->sg_list->addr;
+			info.op.rdma_read.lo_addr.stag = ib_wr->sg_list->lkey;
+			info.op.rdma_read.lo_addr.len = ib_wr->sg_list->length;
+			ret = ukqp->ops.iw_rdma_read(ukqp, &info, false, false);
+			if (ret)
+				err = -EIO;
+			break;
+		default:
+			err = -EINVAL;
+			i40iw_pr_err(" upost_send bad opcode = 0x%x\n",
+				     ib_wr->opcode);
+			break;
+		}
+
+		if (err)
+			break;
+		ib_wr = ib_wr->next;
+	}
+
+	if (err)
+		*bad_wr = ib_wr;
+	else
+		ukqp->ops.iw_qp_post_wr(ukqp);
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+
+	return err;
+}
+
+/**
+ * i40iw_post_recv - post receive wr for kernel application
+ * @ibqp: ib qp pointer
+ * @ib_wr: work request for receive
+ * @bad_wr: bad wr caused an error
+ */
+static int i40iw_post_recv(struct ib_qp *ibqp,
+			   struct ib_recv_wr *ib_wr,
+			   struct ib_recv_wr **bad_wr)
+{
+	struct i40iw_qp *iwqp;
+	struct i40iw_qp_uk *ukqp;
+	struct i40iw_post_rq_info post_recv;
+	struct i40iw_sge sg_list[I40IW_MAX_WQ_FRAGMENT_COUNT];
+	enum i40iw_status_code ret = 0;
+	unsigned long flags;
+
+	iwqp = (struct i40iw_qp *)ibqp;
+	ukqp = &iwqp->sc_qp.qp_uk;
+
+	memset(&post_recv, 0, sizeof(post_recv));
+	spin_lock_irqsave(&iwqp->lock, flags);
+	while (ib_wr) {
+		post_recv.num_sges = ib_wr->num_sge;
+		post_recv.wr_id = ib_wr->wr_id;
+		i40iw_copy_sg_list(sg_list, ib_wr->sg_list, ib_wr->num_sge);
+		post_recv.sg_list = sg_list;
+		ret = ukqp->ops.iw_post_receive(ukqp, &post_recv);
+		if (ret) {
+			i40iw_pr_err(" post_recv err %d\n", ret);
+			*bad_wr = ib_wr;
+			goto out;
+		}
+		ib_wr = ib_wr->next;
+	}
+ out:
+	spin_unlock_irqrestore(&iwqp->lock, flags);
+	if (ret)
+		return -ENOSYS;
+	return 0;
+}
+
+/**
+ * i40iw_poll_cq - poll cq for completion (kernel apps)
+ * @ibcq: cq to poll
+ * @num_entries: number of entries to poll
+ * @entry: wr of entry completed
+ */
+static int i40iw_poll_cq(struct ib_cq *ibcq,
+			 int num_entries,
+			 struct ib_wc *entry)
+{
+	struct i40iw_cq *iwcq;
+	int cqe_count = 0;
+	struct i40iw_cq_poll_info cq_poll_info;
+	enum i40iw_status_code ret;
+	struct i40iw_cq_uk *ukcq;
+	struct i40iw_sc_qp *qp;
+	unsigned long flags;
+
+	iwcq = (struct i40iw_cq *)ibcq;
+	ukcq = &iwcq->sc_cq.cq_uk;
+
+	spin_lock_irqsave(&iwcq->lock, flags);
+	while (cqe_count < num_entries) {
+		ret = ukcq->ops.iw_cq_poll_completion(ukcq, &cq_poll_info, true);
+		if (ret == I40IW_ERR_QUEUE_EMPTY) {
+			break;
+		} else if (ret) {
+			if (!cqe_count)
+				cqe_count = -1;
+			break;
+		}
+		entry->wc_flags = 0;
+		entry->wr_id = cq_poll_info.wr_id;
+		if (!cq_poll_info.error)
+			entry->status = IB_WC_SUCCESS;
+		else
+			entry->status = IB_WC_WR_FLUSH_ERR;
+
+		switch (cq_poll_info.op_type) {
+		case I40IW_OP_TYPE_RDMA_WRITE:
+			entry->opcode = IB_WC_RDMA_WRITE;
+			break;
+		case I40IW_OP_TYPE_RDMA_READ_INV_STAG:
+		case I40IW_OP_TYPE_RDMA_READ:
+			entry->opcode = IB_WC_RDMA_READ;
+			break;
+		case I40IW_OP_TYPE_SEND_SOL:
+		case I40IW_OP_TYPE_SEND_SOL_INV:
+		case I40IW_OP_TYPE_SEND_INV:
+		case I40IW_OP_TYPE_SEND:
+			entry->opcode = IB_WC_SEND;
+			break;
+		case I40IW_OP_TYPE_REC:
+			entry->opcode = IB_WC_RECV;
+			break;
+		default:
+			entry->opcode = IB_WC_RECV;
+			break;
+		}
+
+		entry->vendor_err =
+		    cq_poll_info.major_err << 16 | cq_poll_info.minor_err;
+		entry->ex.imm_data = 0;
+		qp = (struct i40iw_sc_qp *)cq_poll_info.qp_handle;
+		entry->qp = (struct ib_qp *)qp->back_qp;
+		entry->src_qp = cq_poll_info.qp_id;
+		entry->byte_len = cq_poll_info.bytes_xfered;
+		entry++;
+		cqe_count++;
+	}
+	spin_unlock_irqrestore(&iwcq->lock, flags);
+	return cqe_count;
+}
+
+/**
+ * i40iw_req_notify_cq - arm cq kernel application
+ * @ibcq: cq to arm
+ * @notify_flags: notofication flags
+ */
+static int i40iw_req_notify_cq(struct ib_cq *ibcq,
+			       enum ib_cq_notify_flags notify_flags)
+{
+	struct i40iw_cq *iwcq;
+	struct i40iw_cq_uk *ukcq;
+	enum i40iw_completion_notify cq_notify = IW_CQ_COMPL_SOLICITED;
+
+	iwcq = (struct i40iw_cq *)ibcq;
+	ukcq = &iwcq->sc_cq.cq_uk;
+	if (notify_flags == IB_CQ_NEXT_COMP)
+		cq_notify = IW_CQ_COMPL_EVENT;
+	ukcq->ops.iw_cq_request_notification(ukcq, cq_notify);
+	return 0;
+}
+
+/**
+ * i40iw_port_immutable - return port's immutable data
+ * @ibdev: ib dev struct
+ * @port_num: port number
+ * @immutable: immutable data for the port return
+ */
+static int i40iw_port_immutable(struct ib_device *ibdev, u8 port_num,
+				struct ib_port_immutable *immutable)
+{
+	struct ib_port_attr attr;
+	int err;
+
+	err = i40iw_query_port(ibdev, port_num, &attr);
+
+	if (err)
+		return err;
+
+	immutable->pkey_tbl_len = attr.pkey_tbl_len;
+	immutable->gid_tbl_len = attr.gid_tbl_len;
+	immutable->core_cap_flags = RDMA_CORE_PORT_IWARP;
+
+	return 0;
+}
+
+/**
+ * i40iw_get_protocol_stats - Populates the rdma_stats structure
+ * @ibdev: ib dev struct
+ * @stats: iw protocol stats struct
+ */
+static int i40iw_get_protocol_stats(struct ib_device *ibdev,
+				    union rdma_protocol_stats *stats)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct i40iw_sc_dev *dev = &iwdev->sc_dev;
+	struct i40iw_dev_pestat *devstat = &dev->dev_pestat;
+	struct i40iw_dev_hw_stats *hw_stats = &devstat->hw_stats;
+	struct timespec curr_time;
+	static struct timespec last_rd_time = {0, 0};
+	enum i40iw_status_code status = 0;
+	unsigned long flags;
+
+	curr_time = current_kernel_time();
+	memset(stats, 0, sizeof(*stats));
+
+	if (dev->is_pf) {
+		spin_lock_irqsave(&devstat->stats_lock, flags);
+		devstat->ops.iw_hw_stat_read_all(devstat,
+			&devstat->hw_stats);
+		spin_unlock_irqrestore(&devstat->stats_lock, flags);
+	} else {
+		if (((u64)curr_time.tv_sec - (u64)last_rd_time.tv_sec) > 1)
+			status = i40iw_vchnl_vf_get_pe_stats(dev,
+							     &devstat->hw_stats);
+
+		if (status)
+			return -ENOSYS;
+	}
+
+	stats->iw.ipInReceives = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4RXPKTS] +
+				 hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP6RXPKTS];
+	stats->iw.ipInTruncatedPkts = hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP4RXTRUNC] +
+				      hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP6RXTRUNC];
+	stats->iw.ipInDiscards = hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP4RXDISCARD] +
+				 hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP6RXDISCARD];
+	stats->iw.ipOutNoRoutes = hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP4TXNOROUTE] +
+				  hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_IP6TXNOROUTE];
+	stats->iw.ipReasmReqds = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4RXFRAGS] +
+				 hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP6RXFRAGS];
+	stats->iw.ipFragCreates = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4TXFRAGS] +
+				  hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP6TXFRAGS];
+	stats->iw.ipInMcastPkts = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4RXMCPKTS] +
+				  hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP6RXMCPKTS];
+	stats->iw.ipOutMcastPkts = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP4TXMCPKTS] +
+				   hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_IP6TXMCPKTS];
+	stats->iw.tcpOutSegs = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_TCPTXSEG];
+	stats->iw.tcpInSegs = hw_stats->stat_value_64[I40IW_HW_STAT_INDEX_TCPRXSEGS];
+	stats->iw.tcpRetransSegs = hw_stats->stat_value_32[I40IW_HW_STAT_INDEX_TCPRTXSEG];
+
+	last_rd_time = curr_time;
+	return 0;
+}
+
+/**
+ * i40iw_query_gid - Query port GID
+ * @ibdev: device pointer from stack
+ * @port: port number
+ * @index: Entry index
+ * @gid: Global ID
+ */
+static int i40iw_query_gid(struct ib_device *ibdev,
+			   u8 port,
+			   int index,
+			   union ib_gid *gid)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+
+	memset(gid->raw, 0, sizeof(gid->raw));
+	ether_addr_copy(gid->raw, iwdev->netdev->dev_addr);
+	return 0;
+}
+
+/**
+ * i40iw_modify_port  Modify port properties
+ * @ibdev: device pointer from stack
+ * @port: port number
+ * @port_modify_mask: mask for port modifications
+ * @props: port properties
+ */
+static int i40iw_modify_port(struct ib_device *ibdev,
+			     u8 port,
+			     int port_modify_mask,
+			     struct ib_port_modify *props)
+{
+	return 0;
+}
+
+/**
+ * i40iw_query_pkey - Query partition key
+ * @ibdev: device pointer from stack
+ * @port: port number
+ * @index: index of pkey
+ * @pkey: pointer to store the pkey
+ */
+static int i40iw_query_pkey(struct ib_device *ibdev,
+			    u8 port,
+			    u16 index,
+			    u16 *pkey)
+{
+	*pkey = 0;
+	return 0;
+}
+
+/**
+ * i40iw_create_ah - create address handle
+ * @ibpd: ptr of pd
+ * @ah_attr: address handle attributes
+ */
+static struct ib_ah *i40iw_create_ah(struct ib_pd *ibpd,
+				     struct ib_ah_attr *attr)
+{
+	return ERR_PTR(-ENOSYS);
+}
+
+/**
+ * i40iw_destroy_ah - Destroy address handle
+ * @ah: pointer to address handle
+ */
+static int i40iw_destroy_ah(struct ib_ah *ah)
+{
+	return -ENOSYS;
+}
+
+/**
+ * i40iw_init_rdma_device - initialization of iwarp device
+ * @iwdev: iwarp device
+ */
+static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev)
+{
+	struct i40iw_ib_device *iwibdev;
+	struct net_device *netdev = iwdev->netdev;
+	struct pci_dev *pcidev = (struct pci_dev *)iwdev->hw.dev_context;
+
+	iwibdev = (struct i40iw_ib_device *)ib_alloc_device(sizeof(*iwibdev));
+	if (!iwibdev) {
+		i40iw_pr_err("iwdev == NULL\n");
+		return NULL;
+	}
+	strlcpy(iwibdev->ibdev.name, "i40iw%d", IB_DEVICE_NAME_MAX);
+	iwibdev->ibdev.owner = THIS_MODULE;
+	iwdev->iwibdev = iwibdev;
+	iwibdev->iwdev = iwdev;
+
+	iwibdev->ibdev.node_type = RDMA_NODE_RNIC;
+	ether_addr_copy((u8 *)&iwibdev->ibdev.node_guid, netdev->dev_addr);
+
+	iwibdev->ibdev.uverbs_cmd_mask =
+	    (1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
+	    (1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) |
+	    (1ull << IB_USER_VERBS_CMD_QUERY_PORT) |
+	    (1ull << IB_USER_VERBS_CMD_ALLOC_PD) |
+	    (1ull << IB_USER_VERBS_CMD_DEALLOC_PD) |
+	    (1ull << IB_USER_VERBS_CMD_REG_MR) |
+	    (1ull << IB_USER_VERBS_CMD_DEREG_MR) |
+	    (1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
+	    (1ull << IB_USER_VERBS_CMD_CREATE_CQ) |
+	    (1ull << IB_USER_VERBS_CMD_DESTROY_CQ) |
+	    (1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ) |
+	    (1ull << IB_USER_VERBS_CMD_CREATE_QP) |
+	    (1ull << IB_USER_VERBS_CMD_MODIFY_QP) |
+	    (1ull << IB_USER_VERBS_CMD_QUERY_QP) |
+	    (1ull << IB_USER_VERBS_CMD_POLL_CQ) |
+	    (1ull << IB_USER_VERBS_CMD_CREATE_AH) |
+	    (1ull << IB_USER_VERBS_CMD_DESTROY_AH) |
+	    (1ull << IB_USER_VERBS_CMD_DESTROY_QP) |
+	    (1ull << IB_USER_VERBS_CMD_POST_RECV) |
+	    (1ull << IB_USER_VERBS_CMD_POST_SEND);
+	iwibdev->ibdev.phys_port_cnt = 1;
+	iwibdev->ibdev.num_comp_vectors = 1;
+	iwibdev->ibdev.dma_device = &pcidev->dev;
+	iwibdev->ibdev.dev.parent = &pcidev->dev;
+	iwibdev->ibdev.query_port = i40iw_query_port;
+	iwibdev->ibdev.modify_port = i40iw_modify_port;
+	iwibdev->ibdev.query_pkey = i40iw_query_pkey;
+	iwibdev->ibdev.query_gid = i40iw_query_gid;
+	iwibdev->ibdev.alloc_ucontext = i40iw_alloc_ucontext;
+	iwibdev->ibdev.dealloc_ucontext = i40iw_dealloc_ucontext;
+	iwibdev->ibdev.mmap = i40iw_mmap;
+	iwibdev->ibdev.alloc_pd = i40iw_alloc_pd;
+	iwibdev->ibdev.dealloc_pd = i40iw_dealloc_pd;
+	iwibdev->ibdev.create_qp = i40iw_create_qp;
+	iwibdev->ibdev.modify_qp = i40iw_modify_qp;
+	iwibdev->ibdev.query_qp = i40iw_query_qp;
+	iwibdev->ibdev.destroy_qp = i40iw_destroy_qp;
+	iwibdev->ibdev.create_cq = i40iw_create_cq;
+	iwibdev->ibdev.destroy_cq = i40iw_destroy_cq;
+	iwibdev->ibdev.get_dma_mr = i40iw_get_dma_mr;
+	iwibdev->ibdev.reg_user_mr = i40iw_reg_user_mr;
+	iwibdev->ibdev.dereg_mr = i40iw_dereg_mr;
+	iwibdev->ibdev.get_protocol_stats = i40iw_get_protocol_stats;
+	iwibdev->ibdev.query_device = i40iw_query_device;
+	iwibdev->ibdev.create_ah = i40iw_create_ah;
+	iwibdev->ibdev.destroy_ah = i40iw_destroy_ah;
+	iwibdev->ibdev.iwcm = kzalloc(sizeof(*iwibdev->ibdev.iwcm), GFP_KERNEL);
+	if (!iwibdev->ibdev.iwcm) {
+		ib_dealloc_device(&iwibdev->ibdev);
+		i40iw_pr_err("iwcm == NULL\n");
+		return NULL;
+	}
+
+	iwibdev->ibdev.iwcm->add_ref = i40iw_add_ref;
+	iwibdev->ibdev.iwcm->rem_ref = i40iw_rem_ref;
+	iwibdev->ibdev.iwcm->get_qp = i40iw_get_qp;
+	iwibdev->ibdev.iwcm->connect = i40iw_connect;
+	iwibdev->ibdev.iwcm->accept = i40iw_accept;
+	iwibdev->ibdev.iwcm->reject = i40iw_reject;
+	iwibdev->ibdev.iwcm->create_listen = i40iw_create_listen;
+	iwibdev->ibdev.iwcm->destroy_listen = i40iw_destroy_listen;
+	memcpy(iwibdev->ibdev.iwcm->ifname, netdev->name,
+	       sizeof(iwibdev->ibdev.iwcm->ifname));
+	iwibdev->ibdev.get_port_immutable   = i40iw_port_immutable;
+	iwibdev->ibdev.poll_cq = i40iw_poll_cq;
+	iwibdev->ibdev.req_notify_cq = i40iw_req_notify_cq;
+	iwibdev->ibdev.post_send = i40iw_post_send;
+	iwibdev->ibdev.post_recv = i40iw_post_recv;
+
+	return iwibdev;
+}
+
+/**
+ * i40iw_port_ibevent - indicate port event
+ * @iwdev: iwarp device
+ */
+void i40iw_port_ibevent(struct i40iw_device *iwdev)
+{
+	struct i40iw_ib_device *iwibdev = iwdev->iwibdev;
+	struct ib_event event;
+
+	event.device = &iwibdev->ibdev;
+	event.element.port_num = 1;
+	event.event = iwdev->iw_status ? IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR;
+	ib_dispatch_event(&event);
+}
+
+/**
+ * i40iw_unregister_rdma_device - unregister of iwarp from IB
+ * @iwibdev: rdma device ptr
+ */
+static void i40iw_unregister_rdma_device(struct i40iw_ib_device *iwibdev)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(i40iw_dev_attributes); ++i)
+		device_remove_file(&iwibdev->ibdev.dev,
+				   i40iw_dev_attributes[i]);
+	ib_unregister_device(&iwibdev->ibdev);
+}
+
+/**
+ * i40iw_destroy_rdma_device - destroy rdma device and free resources
+ * @iwibdev: IB device ptr
+ */
+void i40iw_destroy_rdma_device(struct i40iw_ib_device *iwibdev)
+{
+	if (!iwibdev)
+		return;
+
+	i40iw_unregister_rdma_device(iwibdev);
+	kfree(iwibdev->ibdev.iwcm);
+	iwibdev->ibdev.iwcm = NULL;
+	ib_dealloc_device(&iwibdev->ibdev);
+}
+
+/**
+ * i40iw_register_rdma_device - register iwarp device to IB
+ * @iwdev: iwarp device
+ */
+int i40iw_register_rdma_device(struct i40iw_device *iwdev)
+{
+	int i, ret;
+	struct i40iw_ib_device *iwibdev;
+
+	iwdev->iwibdev = i40iw_init_rdma_device(iwdev);
+	if (!iwdev->iwibdev)
+		return -ENOSYS;
+	iwibdev = iwdev->iwibdev;
+
+	ret = ib_register_device(&iwibdev->ibdev, NULL);
+	if (ret)
+		goto error;
+
+	for (i = 0; i < ARRAY_SIZE(i40iw_dev_attributes); ++i) {
+		ret =
+		    device_create_file(&iwibdev->ibdev.dev,
+				       i40iw_dev_attributes[i]);
+		if (ret) {
+			while (i > 0) {
+				i--;
+				device_remove_file(&iwibdev->ibdev.dev, i40iw_dev_attributes[i]);
+			}
+			ib_unregister_device(&iwibdev->ibdev);
+			goto error;
+		}
+	}
+	return 0;
+error:
+	kfree(iwdev->iwibdev->ibdev.iwcm);
+	iwdev->iwibdev->ibdev.iwcm = NULL;
+	ib_dealloc_device(&iwdev->iwibdev->ibdev);
+	return -ENOSYS;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.h b/drivers/infiniband/hw/i40iw/i40iw_verbs.h
new file mode 100644
index 000000000000..1101f77080e6
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.h
@@ -0,0 +1,173 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_VERBS_H
+#define I40IW_VERBS_H
+
+struct i40iw_ucontext {
+	struct ib_ucontext ibucontext;
+	struct i40iw_device *iwdev;
+	struct list_head cq_reg_mem_list;
+	spinlock_t cq_reg_mem_list_lock; /* memory list for cq's */
+	struct list_head qp_reg_mem_list;
+	spinlock_t qp_reg_mem_list_lock; /* memory list for qp's */
+};
+
+struct i40iw_pd {
+	struct ib_pd ibpd;
+	struct i40iw_sc_pd sc_pd;
+	atomic_t usecount;
+};
+
+struct i40iw_hmc_pble {
+	union {
+		u32 idx;
+		dma_addr_t addr;
+	};
+};
+
+struct i40iw_cq_mr {
+	struct i40iw_hmc_pble cq_pbl;
+	dma_addr_t shadow;
+};
+
+struct i40iw_qp_mr {
+	struct i40iw_hmc_pble sq_pbl;
+	struct i40iw_hmc_pble rq_pbl;
+	dma_addr_t shadow;
+	struct page *sq_page;
+};
+
+struct i40iw_pbl {
+	struct list_head list;
+	union {
+		struct i40iw_qp_mr qp_mr;
+		struct i40iw_cq_mr cq_mr;
+	};
+
+	bool pbl_allocated;
+	u64 user_base;
+	struct i40iw_pble_alloc pble_alloc;
+	struct i40iw_mr *iwmr;
+};
+
+#define MAX_SAVE_PAGE_ADDRS     4
+struct i40iw_mr {
+	union {
+		struct ib_mr ibmr;
+		struct ib_mw ibmw;
+		struct ib_fmr ibfmr;
+	};
+	struct ib_umem *region;
+	u16 type;
+	u32 page_cnt;
+	u32 stag;
+	u64 length;
+	u64 pgaddrmem[MAX_SAVE_PAGE_ADDRS];
+	struct i40iw_pbl iwpbl;
+};
+
+struct i40iw_cq {
+	struct ib_cq ibcq;
+	struct i40iw_sc_cq sc_cq;
+	u16 cq_head;
+	u16 cq_size;
+	u16 cq_number;
+	bool user_mode;
+	u32 polled_completions;
+	u32 cq_mem_size;
+	struct i40iw_dma_mem kmem;
+	spinlock_t lock; /* for poll cq */
+	struct i40iw_pbl *iwpbl;
+};
+
+struct disconn_work {
+	struct work_struct work;
+	struct i40iw_qp *iwqp;
+};
+
+struct iw_cm_id;
+struct ietf_mpa_frame;
+struct i40iw_ud_file;
+
+struct i40iw_qp_kmode {
+	struct i40iw_dma_mem dma_mem;
+	u64 *wrid_mem;
+};
+
+struct i40iw_qp {
+	struct ib_qp ibqp;
+	struct i40iw_sc_qp sc_qp;
+	struct i40iw_device *iwdev;
+	struct i40iw_cq *iwscq;
+	struct i40iw_cq *iwrcq;
+	struct i40iw_pd *iwpd;
+	struct i40iw_qp_host_ctx_info ctx_info;
+	struct i40iwarp_offload_info iwarp_info;
+	void *allocated_buffer;
+	atomic_t refcount;
+	struct iw_cm_id *cm_id;
+	void *cm_node;
+	struct ib_mr *lsmm_mr;
+	struct work_struct work;
+	enum ib_qp_state ibqp_state;
+	u32 iwarp_state;
+	u32 qp_mem_size;
+	u32 last_aeq;
+	atomic_t close_timer_started;
+	spinlock_t lock; /* for post work requests */
+	struct i40iw_qp_context *iwqp_context;
+	void *pbl_vbase;
+	dma_addr_t pbl_pbase;
+	struct page *page;
+	u8 active_conn:1;
+	u8 user_mode:1;
+	u8 hte_added:1;
+	u8 flush_issued:1;
+	u8 destroyed:1;
+	u8 sig_all:1;
+	u8 pau_mode:1;
+	u8 rsvd:1;
+	u16 term_sq_flush_code;
+	u16 term_rq_flush_code;
+	u8 hw_iwarp_state;
+	u8 hw_tcp_state;
+	struct i40iw_qp_kmode kqp;
+	struct i40iw_dma_mem host_ctx;
+	struct timer_list terminate_timer;
+	struct i40iw_pbl *iwpbl;
+	struct i40iw_dma_mem q2_ctx_mem;
+	struct i40iw_dma_mem ietf_mem;
+};
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_vf.c b/drivers/infiniband/hw/i40iw/i40iw_vf.c
new file mode 100644
index 000000000000..cb0f18340e14
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_vf.c
@@ -0,0 +1,85 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_status.h"
+#include "i40iw_hmc.h"
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_vf.h"
+
+/**
+ * i40iw_manage_vf_pble_bp - manage vf pble
+ * @cqp: cqp for cqp' sq wqe
+ * @info: pble info
+ * @scratch: pointer for completion
+ * @post_sq: to post and ring
+ */
+enum i40iw_status_code i40iw_manage_vf_pble_bp(struct i40iw_sc_cqp *cqp,
+					       struct i40iw_manage_vf_pble_info *info,
+					       u64 scratch,
+					       bool post_sq)
+{
+	u64 *wqe;
+	u64 temp, header, pd_pl_pba = 0;
+
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	temp = LS_64(info->pd_entry_cnt, I40IW_CQPSQ_MVPBP_PD_ENTRY_CNT) |
+	    LS_64(info->first_pd_index, I40IW_CQPSQ_MVPBP_FIRST_PD_INX) |
+	    LS_64(info->sd_index, I40IW_CQPSQ_MVPBP_SD_INX);
+	set_64bit_val(wqe, 16, temp);
+
+	header = LS_64((info->inv_pd_ent ? 1 : 0), I40IW_CQPSQ_MVPBP_INV_PD_ENT) |
+	    LS_64(I40IW_CQP_OP_MANAGE_VF_PBLE_BP, I40IW_CQPSQ_OPCODE) |
+	    LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+	set_64bit_val(wqe, 24, header);
+
+	pd_pl_pba = LS_64(info->pd_pl_pba >> 3, I40IW_CQPSQ_MVPBP_PD_PLPBA);
+	set_64bit_val(wqe, 32, pd_pl_pba);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "MANAGE VF_PBLE_BP WQE", wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
+struct i40iw_vf_cqp_ops iw_vf_cqp_ops = {
+	i40iw_manage_vf_pble_bp
+};
diff --git a/drivers/infiniband/hw/i40iw/i40iw_vf.h b/drivers/infiniband/hw/i40iw/i40iw_vf.h
new file mode 100644
index 000000000000..f649f3a62e13
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_vf.h
@@ -0,0 +1,62 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_VF_H
+#define I40IW_VF_H
+
+struct i40iw_sc_cqp;
+
+struct i40iw_manage_vf_pble_info {
+	u32 sd_index;
+	u16 first_pd_index;
+	u16 pd_entry_cnt;
+	u8 inv_pd_ent;
+	u64 pd_pl_pba;
+};
+
+struct i40iw_vf_cqp_ops {
+	enum i40iw_status_code (*manage_vf_pble_bp)(struct i40iw_sc_cqp *,
+						    struct i40iw_manage_vf_pble_info *,
+						    u64,
+						    bool);
+};
+
+enum i40iw_status_code i40iw_manage_vf_pble_bp(struct i40iw_sc_cqp *cqp,
+					       struct i40iw_manage_vf_pble_info *info,
+					       u64 scratch,
+					       bool post_sq);
+
+extern struct i40iw_vf_cqp_ops iw_vf_cqp_ops;
+
+#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_virtchnl.c b/drivers/infiniband/hw/i40iw/i40iw_virtchnl.c
new file mode 100644
index 000000000000..6b68f7890b76
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_virtchnl.c
@@ -0,0 +1,748 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#include "i40iw_osdep.h"
+#include "i40iw_register.h"
+#include "i40iw_status.h"
+#include "i40iw_hmc.h"
+#include "i40iw_d.h"
+#include "i40iw_type.h"
+#include "i40iw_p.h"
+#include "i40iw_virtchnl.h"
+
+/**
+ * vchnl_vf_send_get_ver_req - Request Channel version
+ * @dev: IWARP device pointer
+ * @vchnl_req: Virtual channel message request pointer
+ */
+static enum i40iw_status_code vchnl_vf_send_get_ver_req(struct i40iw_sc_dev *dev,
+							struct i40iw_virtchnl_req *vchnl_req)
+{
+	enum i40iw_status_code ret_code = I40IW_ERR_NOT_READY;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = vchnl_req->vchnl_msg;
+
+	if (!dev->vchnl_up)
+		return ret_code;
+
+	memset(vchnl_msg, 0, sizeof(*vchnl_msg));
+	vchnl_msg->iw_chnl_op_ctx = (uintptr_t)vchnl_req;
+	vchnl_msg->iw_chnl_buf_len = sizeof(*vchnl_msg);
+	vchnl_msg->iw_op_code = I40IW_VCHNL_OP_GET_VER;
+	vchnl_msg->iw_op_ver = I40IW_VCHNL_OP_GET_VER_V0;
+	ret_code = dev->vchnl_if.vchnl_send(dev, 0, (u8 *)vchnl_msg, vchnl_msg->iw_chnl_buf_len);
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+	return ret_code;
+}
+
+/**
+ * vchnl_vf_send_get_hmc_fcn_req - Request HMC Function from VF
+ * @dev: IWARP device pointer
+ * @vchnl_req: Virtual channel message request pointer
+ */
+static enum i40iw_status_code vchnl_vf_send_get_hmc_fcn_req(struct i40iw_sc_dev *dev,
+							    struct i40iw_virtchnl_req *vchnl_req)
+{
+	enum i40iw_status_code ret_code = I40IW_ERR_NOT_READY;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = vchnl_req->vchnl_msg;
+
+	if (!dev->vchnl_up)
+		return ret_code;
+
+	memset(vchnl_msg, 0, sizeof(*vchnl_msg));
+	vchnl_msg->iw_chnl_op_ctx = (uintptr_t)vchnl_req;
+	vchnl_msg->iw_chnl_buf_len = sizeof(*vchnl_msg);
+	vchnl_msg->iw_op_code = I40IW_VCHNL_OP_GET_HMC_FCN;
+	vchnl_msg->iw_op_ver = I40IW_VCHNL_OP_GET_HMC_FCN_V0;
+	ret_code = dev->vchnl_if.vchnl_send(dev, 0, (u8 *)vchnl_msg, vchnl_msg->iw_chnl_buf_len);
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+	return ret_code;
+}
+
+/**
+ * vchnl_vf_send_get_pe_stats_req - Request PE stats from VF
+ * @dev: IWARP device pointer
+ * @vchnl_req: Virtual channel message request pointer
+ */
+static enum i40iw_status_code vchnl_vf_send_get_pe_stats_req(struct i40iw_sc_dev *dev,
+							     struct i40iw_virtchnl_req  *vchnl_req)
+{
+	enum i40iw_status_code ret_code = I40IW_ERR_NOT_READY;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = vchnl_req->vchnl_msg;
+
+	if (!dev->vchnl_up)
+		return ret_code;
+
+	memset(vchnl_msg, 0, sizeof(*vchnl_msg));
+	vchnl_msg->iw_chnl_op_ctx = (uintptr_t)vchnl_req;
+	vchnl_msg->iw_chnl_buf_len = sizeof(*vchnl_msg) + sizeof(struct i40iw_dev_hw_stats) - 1;
+	vchnl_msg->iw_op_code = I40IW_VCHNL_OP_GET_STATS;
+	vchnl_msg->iw_op_ver = I40IW_VCHNL_OP_GET_STATS_V0;
+	ret_code = dev->vchnl_if.vchnl_send(dev, 0, (u8 *)vchnl_msg, vchnl_msg->iw_chnl_buf_len);
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+	return ret_code;
+}
+
+/**
+ * vchnl_vf_send_add_hmc_objs_req - Add HMC objects
+ * @dev: IWARP device pointer
+ * @vchnl_req: Virtual channel message request pointer
+ */
+static enum i40iw_status_code vchnl_vf_send_add_hmc_objs_req(struct i40iw_sc_dev *dev,
+							     struct i40iw_virtchnl_req *vchnl_req,
+							     enum i40iw_hmc_rsrc_type rsrc_type,
+							     u32 start_index,
+							     u32 rsrc_count)
+{
+	enum i40iw_status_code ret_code = I40IW_ERR_NOT_READY;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = vchnl_req->vchnl_msg;
+	struct i40iw_virtchnl_hmc_obj_range *add_hmc_obj;
+
+	if (!dev->vchnl_up)
+		return ret_code;
+
+	add_hmc_obj = (struct i40iw_virtchnl_hmc_obj_range *)vchnl_msg->iw_chnl_buf;
+	memset(vchnl_msg, 0, sizeof(*vchnl_msg));
+	memset(add_hmc_obj, 0, sizeof(*add_hmc_obj));
+	vchnl_msg->iw_chnl_op_ctx = (uintptr_t)vchnl_req;
+	vchnl_msg->iw_chnl_buf_len = sizeof(*vchnl_msg) + sizeof(struct i40iw_virtchnl_hmc_obj_range) - 1;
+	vchnl_msg->iw_op_code = I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE;
+	vchnl_msg->iw_op_ver = I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE_V0;
+	add_hmc_obj->obj_type = (u16)rsrc_type;
+	add_hmc_obj->start_index = start_index;
+	add_hmc_obj->obj_count = rsrc_count;
+	ret_code = dev->vchnl_if.vchnl_send(dev, 0, (u8 *)vchnl_msg, vchnl_msg->iw_chnl_buf_len);
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+	return ret_code;
+}
+
+/**
+ * vchnl_vf_send_del_hmc_objs_req - del HMC objects
+ * @dev: IWARP device pointer
+ * @vchnl_req: Virtual channel message request pointer
+ * @ rsrc_type - resource type to delete
+ * @ start_index - starting index for resource
+ * @ rsrc_count - number of resource type to delete
+ */
+static enum i40iw_status_code vchnl_vf_send_del_hmc_objs_req(struct i40iw_sc_dev *dev,
+							     struct i40iw_virtchnl_req *vchnl_req,
+							     enum i40iw_hmc_rsrc_type rsrc_type,
+							     u32 start_index,
+							     u32 rsrc_count)
+{
+	enum i40iw_status_code ret_code = I40IW_ERR_NOT_READY;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = vchnl_req->vchnl_msg;
+	struct i40iw_virtchnl_hmc_obj_range *add_hmc_obj;
+
+	if (!dev->vchnl_up)
+		return ret_code;
+
+	add_hmc_obj = (struct i40iw_virtchnl_hmc_obj_range *)vchnl_msg->iw_chnl_buf;
+	memset(vchnl_msg, 0, sizeof(*vchnl_msg));
+	memset(add_hmc_obj, 0, sizeof(*add_hmc_obj));
+	vchnl_msg->iw_chnl_op_ctx = (uintptr_t)vchnl_req;
+	vchnl_msg->iw_chnl_buf_len = sizeof(*vchnl_msg) + sizeof(struct i40iw_virtchnl_hmc_obj_range) - 1;
+	vchnl_msg->iw_op_code = I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE;
+	vchnl_msg->iw_op_ver = I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE_V0;
+	add_hmc_obj->obj_type = (u16)rsrc_type;
+	add_hmc_obj->start_index = start_index;
+	add_hmc_obj->obj_count = rsrc_count;
+	ret_code = dev->vchnl_if.vchnl_send(dev, 0, (u8 *)vchnl_msg, vchnl_msg->iw_chnl_buf_len);
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+	return ret_code;
+}
+
+/**
+ * vchnl_pf_send_get_ver_resp - Send channel version to VF
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @vchnl_msg: Virtual channel message buffer pointer
+ */
+static void vchnl_pf_send_get_ver_resp(struct i40iw_sc_dev *dev,
+				       u32 vf_id,
+				       struct i40iw_virtchnl_op_buf *vchnl_msg)
+{
+	enum i40iw_status_code ret_code;
+	u8 resp_buffer[sizeof(struct i40iw_virtchnl_resp_buf) + sizeof(u32) - 1];
+	struct i40iw_virtchnl_resp_buf *vchnl_msg_resp = (struct i40iw_virtchnl_resp_buf *)resp_buffer;
+
+	memset(resp_buffer, 0, sizeof(*resp_buffer));
+	vchnl_msg_resp->iw_chnl_op_ctx = vchnl_msg->iw_chnl_op_ctx;
+	vchnl_msg_resp->iw_chnl_buf_len = sizeof(resp_buffer);
+	vchnl_msg_resp->iw_op_ret_code = I40IW_SUCCESS;
+	*((u32 *)vchnl_msg_resp->iw_chnl_buf) = I40IW_VCHNL_CHNL_VER_V0;
+	ret_code = dev->vchnl_if.vchnl_send(dev, vf_id, resp_buffer, sizeof(resp_buffer));
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+}
+
+/**
+ * vchnl_pf_send_get_hmc_fcn_resp - Send HMC Function to VF
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @vchnl_msg: Virtual channel message buffer pointer
+ */
+static void vchnl_pf_send_get_hmc_fcn_resp(struct i40iw_sc_dev *dev,
+					   u32 vf_id,
+					   struct i40iw_virtchnl_op_buf *vchnl_msg,
+					   u16 hmc_fcn)
+{
+	enum i40iw_status_code ret_code;
+	u8 resp_buffer[sizeof(struct i40iw_virtchnl_resp_buf) + sizeof(u16) - 1];
+	struct i40iw_virtchnl_resp_buf *vchnl_msg_resp = (struct i40iw_virtchnl_resp_buf *)resp_buffer;
+
+	memset(resp_buffer, 0, sizeof(*resp_buffer));
+	vchnl_msg_resp->iw_chnl_op_ctx = vchnl_msg->iw_chnl_op_ctx;
+	vchnl_msg_resp->iw_chnl_buf_len = sizeof(resp_buffer);
+	vchnl_msg_resp->iw_op_ret_code = I40IW_SUCCESS;
+	*((u16 *)vchnl_msg_resp->iw_chnl_buf) = hmc_fcn;
+	ret_code = dev->vchnl_if.vchnl_send(dev, vf_id, resp_buffer, sizeof(resp_buffer));
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+}
+
+/**
+ * vchnl_pf_send_get_pe_stats_resp - Send PE Stats to VF
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @vchnl_msg: Virtual channel message buffer pointer
+ * @hw_stats: HW Stats struct
+ */
+
+static void vchnl_pf_send_get_pe_stats_resp(struct i40iw_sc_dev *dev,
+					    u32 vf_id,
+					    struct i40iw_virtchnl_op_buf *vchnl_msg,
+					    struct i40iw_dev_hw_stats hw_stats)
+{
+	enum i40iw_status_code ret_code;
+	u8 resp_buffer[sizeof(struct i40iw_virtchnl_resp_buf) + sizeof(struct i40iw_dev_hw_stats) - 1];
+	struct i40iw_virtchnl_resp_buf *vchnl_msg_resp = (struct i40iw_virtchnl_resp_buf *)resp_buffer;
+
+	memset(resp_buffer, 0, sizeof(*resp_buffer));
+	vchnl_msg_resp->iw_chnl_op_ctx = vchnl_msg->iw_chnl_op_ctx;
+	vchnl_msg_resp->iw_chnl_buf_len = sizeof(resp_buffer);
+	vchnl_msg_resp->iw_op_ret_code = I40IW_SUCCESS;
+	*((struct i40iw_dev_hw_stats *)vchnl_msg_resp->iw_chnl_buf) = hw_stats;
+	ret_code = dev->vchnl_if.vchnl_send(dev, vf_id, resp_buffer, sizeof(resp_buffer));
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+}
+
+/**
+ * vchnl_pf_send_error_resp - Send an error response to VF
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @vchnl_msg: Virtual channel message buffer pointer
+ */
+static void vchnl_pf_send_error_resp(struct i40iw_sc_dev *dev, u32 vf_id,
+				     struct i40iw_virtchnl_op_buf *vchnl_msg,
+				     u16 op_ret_code)
+{
+	enum i40iw_status_code ret_code;
+	u8 resp_buffer[sizeof(struct i40iw_virtchnl_resp_buf)];
+	struct i40iw_virtchnl_resp_buf *vchnl_msg_resp = (struct i40iw_virtchnl_resp_buf *)resp_buffer;
+
+	memset(resp_buffer, 0, sizeof(resp_buffer));
+	vchnl_msg_resp->iw_chnl_op_ctx = vchnl_msg->iw_chnl_op_ctx;
+	vchnl_msg_resp->iw_chnl_buf_len = sizeof(resp_buffer);
+	vchnl_msg_resp->iw_op_ret_code = (u16)op_ret_code;
+	ret_code = dev->vchnl_if.vchnl_send(dev, vf_id, resp_buffer, sizeof(resp_buffer));
+	if (ret_code)
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: virt channel send failed 0x%x\n", __func__, ret_code);
+}
+
+/**
+ * pf_cqp_get_hmc_fcn_callback - Callback for Get HMC Fcn
+ * @cqp_req_param: CQP Request param value
+ * @not_used: unused CQP callback parameter
+ */
+static void pf_cqp_get_hmc_fcn_callback(struct i40iw_sc_dev *dev, void *callback_param,
+					struct i40iw_ccq_cqe_info *cqe_info)
+{
+	struct i40iw_vfdev *vf_dev = callback_param;
+	struct i40iw_virt_mem vf_dev_mem;
+
+	if (cqe_info->error) {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "CQP Completion Error on Get HMC Function.  Maj = 0x%04x, Minor = 0x%04x\n",
+			    cqe_info->maj_err_code, cqe_info->min_err_code);
+		dev->vf_dev[vf_dev->iw_vf_idx] = NULL;
+		vchnl_pf_send_error_resp(dev, vf_dev->vf_id, &vf_dev->vf_msg_buffer.vchnl_msg,
+					 (u16)I40IW_ERR_CQP_COMPL_ERROR);
+		vf_dev_mem.va = vf_dev;
+		vf_dev_mem.size = sizeof(*vf_dev);
+		i40iw_free_virt_mem(dev->hw, &vf_dev_mem);
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "CQP Completion Operation Return information = 0x%08x\n",
+			    cqe_info->op_ret_val);
+		vf_dev->pmf_index = (u16)cqe_info->op_ret_val;
+		vf_dev->msg_count--;
+		vchnl_pf_send_get_hmc_fcn_resp(dev,
+					       vf_dev->vf_id,
+					       &vf_dev->vf_msg_buffer.vchnl_msg,
+					       vf_dev->pmf_index);
+	}
+}
+
+/**
+ * pf_add_hmc_obj - Callback for Add HMC Object
+ * @vf_dev: pointer to the VF Device
+ */
+static void pf_add_hmc_obj_callback(void *work_vf_dev)
+{
+	struct i40iw_vfdev *vf_dev = (struct i40iw_vfdev *)work_vf_dev;
+	struct i40iw_hmc_info *hmc_info = &vf_dev->hmc_info;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = &vf_dev->vf_msg_buffer.vchnl_msg;
+	struct i40iw_hmc_create_obj_info info;
+	struct i40iw_virtchnl_hmc_obj_range *add_hmc_obj;
+	enum i40iw_status_code ret_code;
+
+	if (!vf_dev->pf_hmc_initialized) {
+		ret_code = i40iw_pf_init_vfhmc(vf_dev->pf_dev, (u8)vf_dev->pmf_index, NULL);
+		if (ret_code)
+			goto add_out;
+		vf_dev->pf_hmc_initialized = true;
+	}
+
+	add_hmc_obj = (struct i40iw_virtchnl_hmc_obj_range *)vchnl_msg->iw_chnl_buf;
+
+	memset(&info, 0, sizeof(info));
+	info.hmc_info = hmc_info;
+	info.is_pf = false;
+	info.rsrc_type = (u32)add_hmc_obj->obj_type;
+	info.entry_type = (info.rsrc_type == I40IW_HMC_IW_PBLE) ? I40IW_SD_TYPE_PAGED : I40IW_SD_TYPE_DIRECT;
+	info.start_idx = add_hmc_obj->start_index;
+	info.count = add_hmc_obj->obj_count;
+	i40iw_debug(vf_dev->pf_dev, I40IW_DEBUG_VIRT,
+		    "I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE.  Add %u type %u objects\n",
+		    info.count, info.rsrc_type);
+	ret_code = i40iw_sc_create_hmc_obj(vf_dev->pf_dev, &info);
+	if (!ret_code)
+		vf_dev->hmc_info.hmc_obj[add_hmc_obj->obj_type].cnt = add_hmc_obj->obj_count;
+add_out:
+	vf_dev->msg_count--;
+	vchnl_pf_send_error_resp(vf_dev->pf_dev, vf_dev->vf_id, vchnl_msg, (u16)ret_code);
+}
+
+/**
+ * pf_del_hmc_obj_callback - Callback for delete HMC Object
+ * @work_vf_dev: pointer to the VF Device
+ */
+static void pf_del_hmc_obj_callback(void *work_vf_dev)
+{
+	struct i40iw_vfdev *vf_dev = (struct i40iw_vfdev *)work_vf_dev;
+	struct i40iw_hmc_info *hmc_info = &vf_dev->hmc_info;
+	struct i40iw_virtchnl_op_buf *vchnl_msg = &vf_dev->vf_msg_buffer.vchnl_msg;
+	struct i40iw_hmc_del_obj_info info;
+	struct i40iw_virtchnl_hmc_obj_range *del_hmc_obj;
+	enum i40iw_status_code ret_code = I40IW_SUCCESS;
+
+	if (!vf_dev->pf_hmc_initialized)
+		goto del_out;
+
+	del_hmc_obj = (struct i40iw_virtchnl_hmc_obj_range *)vchnl_msg->iw_chnl_buf;
+
+	memset(&info, 0, sizeof(info));
+	info.hmc_info = hmc_info;
+	info.is_pf = false;
+	info.rsrc_type = (u32)del_hmc_obj->obj_type;
+	info.start_idx = del_hmc_obj->start_index;
+	info.count = del_hmc_obj->obj_count;
+	i40iw_debug(vf_dev->pf_dev, I40IW_DEBUG_VIRT,
+		    "I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE.  Delete %u type %u objects\n",
+		    info.count, info.rsrc_type);
+	ret_code = i40iw_sc_del_hmc_obj(vf_dev->pf_dev, &info, false);
+del_out:
+	vf_dev->msg_count--;
+	vchnl_pf_send_error_resp(vf_dev->pf_dev, vf_dev->vf_id, vchnl_msg, (u16)ret_code);
+}
+
+/**
+ * i40iw_vchnl_recv_pf - Receive PF virtual channel messages
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @msg: Virtual channel message buffer pointer
+ * @len: Length of the virtual channels message
+ */
+enum i40iw_status_code i40iw_vchnl_recv_pf(struct i40iw_sc_dev *dev,
+					   u32 vf_id,
+					   u8 *msg,
+					   u16 len)
+{
+	struct i40iw_virtchnl_op_buf *vchnl_msg = (struct i40iw_virtchnl_op_buf *)msg;
+	struct i40iw_vfdev *vf_dev = NULL;
+	struct i40iw_hmc_fcn_info hmc_fcn_info;
+	u16 iw_vf_idx;
+	u16 first_avail_iw_vf = I40IW_MAX_PE_ENABLED_VF_COUNT;
+	struct i40iw_virt_mem vf_dev_mem;
+	struct i40iw_virtchnl_work_info work_info;
+	struct i40iw_dev_pestat *devstat;
+	enum i40iw_status_code ret_code;
+	unsigned long flags;
+
+	if (!dev || !msg || !len)
+		return I40IW_ERR_PARAM;
+
+	if (!dev->vchnl_up)
+		return I40IW_ERR_NOT_READY;
+	if (vchnl_msg->iw_op_code == I40IW_VCHNL_OP_GET_VER) {
+		if (vchnl_msg->iw_op_ver != I40IW_VCHNL_OP_GET_VER_V0)
+			vchnl_pf_send_get_ver_resp(dev, vf_id, vchnl_msg);
+		else
+			vchnl_pf_send_get_ver_resp(dev, vf_id, vchnl_msg);
+		return I40IW_SUCCESS;
+	}
+	for (iw_vf_idx = 0; iw_vf_idx < I40IW_MAX_PE_ENABLED_VF_COUNT;
+	     iw_vf_idx++) {
+		if (!dev->vf_dev[iw_vf_idx]) {
+			if (first_avail_iw_vf ==
+			    I40IW_MAX_PE_ENABLED_VF_COUNT)
+				first_avail_iw_vf = iw_vf_idx;
+			continue;
+		}
+		if (dev->vf_dev[iw_vf_idx]->vf_id == vf_id) {
+			vf_dev = dev->vf_dev[iw_vf_idx];
+			break;
+		}
+	}
+	if (vf_dev) {
+		if (!vf_dev->msg_count) {
+			vf_dev->msg_count++;
+		} else {
+			i40iw_debug(dev, I40IW_DEBUG_VIRT,
+				    "VF%u already has a channel message in progress.\n",
+				    vf_id);
+			return I40IW_SUCCESS;
+		}
+	}
+	switch (vchnl_msg->iw_op_code) {
+	case I40IW_VCHNL_OP_GET_HMC_FCN:
+		if (!vf_dev &&
+		    (first_avail_iw_vf != I40IW_MAX_PE_ENABLED_VF_COUNT)) {
+			ret_code = i40iw_allocate_virt_mem(dev->hw, &vf_dev_mem, sizeof(struct i40iw_vfdev) +
+							   (sizeof(struct i40iw_hmc_obj_info) * I40IW_HMC_IW_MAX));
+			if (!ret_code) {
+				vf_dev = vf_dev_mem.va;
+				vf_dev->stats_initialized = false;
+				vf_dev->pf_dev = dev;
+				vf_dev->msg_count = 1;
+				vf_dev->vf_id = vf_id;
+				vf_dev->iw_vf_idx = first_avail_iw_vf;
+				vf_dev->pf_hmc_initialized = false;
+				vf_dev->hmc_info.hmc_obj = (struct i40iw_hmc_obj_info *)(&vf_dev[1]);
+				i40iw_debug(dev, I40IW_DEBUG_VIRT,
+					    "vf_dev %p, hmc_info %p, hmc_obj %p\n",
+					    vf_dev, &vf_dev->hmc_info, vf_dev->hmc_info.hmc_obj);
+				dev->vf_dev[first_avail_iw_vf] = vf_dev;
+				iw_vf_idx = first_avail_iw_vf;
+			} else {
+				i40iw_debug(dev, I40IW_DEBUG_VIRT,
+					    "VF%u Unable to allocate a VF device structure.\n",
+					    vf_id);
+				vchnl_pf_send_error_resp(dev, vf_id, vchnl_msg, (u16)I40IW_ERR_NO_MEMORY);
+				return I40IW_SUCCESS;
+			}
+			memcpy(&vf_dev->vf_msg_buffer.vchnl_msg, vchnl_msg, len);
+			hmc_fcn_info.callback_fcn = pf_cqp_get_hmc_fcn_callback;
+			hmc_fcn_info.vf_id = vf_id;
+			hmc_fcn_info.iw_vf_idx = vf_dev->iw_vf_idx;
+			hmc_fcn_info.cqp_callback_param = vf_dev;
+			hmc_fcn_info.free_fcn = false;
+			ret_code = i40iw_cqp_manage_hmc_fcn_cmd(dev, &hmc_fcn_info);
+			if (ret_code)
+				i40iw_debug(dev, I40IW_DEBUG_VIRT,
+					    "VF%u error CQP HMC Function operation.\n",
+					    vf_id);
+			ret_code = i40iw_device_init_pestat(&vf_dev->dev_pestat);
+			if (ret_code)
+				i40iw_debug(dev, I40IW_DEBUG_VIRT,
+					    "VF%u - i40iw_device_init_pestat failed\n",
+					    vf_id);
+			vf_dev->dev_pestat.ops.iw_hw_stat_init(&vf_dev->dev_pestat,
+							      (u8)vf_dev->pmf_index,
+							      dev->hw, false);
+			vf_dev->stats_initialized = true;
+		} else {
+			if (vf_dev) {
+				vf_dev->msg_count--;
+				vchnl_pf_send_get_hmc_fcn_resp(dev, vf_id, vchnl_msg, vf_dev->pmf_index);
+			} else {
+				vchnl_pf_send_error_resp(dev, vf_id, vchnl_msg,
+							 (u16)I40IW_ERR_NO_MEMORY);
+			}
+		}
+		break;
+	case I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE:
+		if (!vf_dev)
+			return I40IW_ERR_BAD_PTR;
+		work_info.worker_vf_dev = vf_dev;
+		work_info.callback_fcn = pf_add_hmc_obj_callback;
+		memcpy(&vf_dev->vf_msg_buffer.vchnl_msg, vchnl_msg, len);
+		i40iw_cqp_spawn_worker(dev, &work_info, vf_dev->iw_vf_idx);
+		break;
+	case I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE:
+		if (!vf_dev)
+			return I40IW_ERR_BAD_PTR;
+		work_info.worker_vf_dev = vf_dev;
+		work_info.callback_fcn = pf_del_hmc_obj_callback;
+		memcpy(&vf_dev->vf_msg_buffer.vchnl_msg, vchnl_msg, len);
+		i40iw_cqp_spawn_worker(dev, &work_info, vf_dev->iw_vf_idx);
+		break;
+	case I40IW_VCHNL_OP_GET_STATS:
+		if (!vf_dev)
+			return I40IW_ERR_BAD_PTR;
+		devstat = &vf_dev->dev_pestat;
+		spin_lock_irqsave(&dev->dev_pestat.stats_lock, flags);
+		devstat->ops.iw_hw_stat_read_all(devstat, &devstat->hw_stats);
+		spin_unlock_irqrestore(&dev->dev_pestat.stats_lock, flags);
+		vf_dev->msg_count--;
+		vchnl_pf_send_get_pe_stats_resp(dev, vf_id, vchnl_msg, devstat->hw_stats);
+		break;
+	default:
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "40iw_vchnl_recv_pf: Invalid OpCode 0x%x\n",
+			    vchnl_msg->iw_op_code);
+		vchnl_pf_send_error_resp(dev, vf_id,
+					 vchnl_msg, (u16)I40IW_ERR_NOT_IMPLEMENTED);
+	}
+	return I40IW_SUCCESS;
+}
+
+/**
+ * i40iw_vchnl_recv_vf - Receive VF virtual channel messages
+ * @dev: IWARP device pointer
+ * @vf_id: Virtual function ID associated with the message
+ * @msg: Virtual channel message buffer pointer
+ * @len: Length of the virtual channels message
+ */
+enum i40iw_status_code i40iw_vchnl_recv_vf(struct i40iw_sc_dev *dev,
+					   u32 vf_id,
+					   u8 *msg,
+					   u16 len)
+{
+	struct i40iw_virtchnl_resp_buf *vchnl_msg_resp = (struct i40iw_virtchnl_resp_buf *)msg;
+	struct i40iw_virtchnl_req *vchnl_req;
+
+	vchnl_req = (struct i40iw_virtchnl_req *)(uintptr_t)vchnl_msg_resp->iw_chnl_op_ctx;
+	vchnl_req->ret_code = (enum i40iw_status_code)vchnl_msg_resp->iw_op_ret_code;
+	if (len == (sizeof(*vchnl_msg_resp) + vchnl_req->parm_len - 1)) {
+		if (vchnl_req->parm_len && vchnl_req->parm)
+			memcpy(vchnl_req->parm, vchnl_msg_resp->iw_chnl_buf, vchnl_req->parm_len);
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: Got response, data size %u\n", __func__,
+			    vchnl_req->parm_len);
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s: error length on response, Got %u, expected %u\n", __func__,
+			    len, (u32)(sizeof(*vchnl_msg_resp) + vchnl_req->parm_len - 1));
+	}
+
+	return I40IW_SUCCESS;
+}
+
+/**
+ * i40iw_vchnl_vf_get_ver - Request Channel version
+ * @dev: IWARP device pointer
+ * @vchnl_ver: Virtual channel message version pointer
+ */
+enum i40iw_status_code i40iw_vchnl_vf_get_ver(struct i40iw_sc_dev *dev,
+					      u32 *vchnl_ver)
+{
+	struct i40iw_virtchnl_req vchnl_req;
+	enum i40iw_status_code ret_code;
+
+	memset(&vchnl_req, 0, sizeof(vchnl_req));
+	vchnl_req.dev = dev;
+	vchnl_req.parm = vchnl_ver;
+	vchnl_req.parm_len = sizeof(*vchnl_ver);
+	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+	ret_code = vchnl_vf_send_get_ver_req(dev, &vchnl_req);
+	if (!ret_code) {
+		ret_code = i40iw_vf_wait_vchnl_resp(dev);
+		if (!ret_code)
+			ret_code = vchnl_req.ret_code;
+		else
+			dev->vchnl_up = false;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_vchnl_vf_get_hmc_fcn - Request HMC Function
+ * @dev: IWARP device pointer
+ * @hmc_fcn: HMC function index pointer
+ */
+enum i40iw_status_code i40iw_vchnl_vf_get_hmc_fcn(struct i40iw_sc_dev *dev,
+						  u16 *hmc_fcn)
+{
+	struct i40iw_virtchnl_req vchnl_req;
+	enum i40iw_status_code ret_code;
+
+	memset(&vchnl_req, 0, sizeof(vchnl_req));
+	vchnl_req.dev = dev;
+	vchnl_req.parm = hmc_fcn;
+	vchnl_req.parm_len = sizeof(*hmc_fcn);
+	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+	ret_code = vchnl_vf_send_get_hmc_fcn_req(dev, &vchnl_req);
+	if (!ret_code) {
+		ret_code = i40iw_vf_wait_vchnl_resp(dev);
+		if (!ret_code)
+			ret_code = vchnl_req.ret_code;
+		else
+			dev->vchnl_up = false;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_vchnl_vf_add_hmc_objs - Add HMC Object
+ * @dev: IWARP device pointer
+ * @rsrc_type: HMC Resource type
+ * @start_index: Starting index of the objects to be added
+ * @rsrc_count: Number of resources to be added
+ */
+enum i40iw_status_code i40iw_vchnl_vf_add_hmc_objs(struct i40iw_sc_dev *dev,
+						   enum i40iw_hmc_rsrc_type rsrc_type,
+						   u32 start_index,
+						   u32 rsrc_count)
+{
+	struct i40iw_virtchnl_req vchnl_req;
+	enum i40iw_status_code ret_code;
+
+	memset(&vchnl_req, 0, sizeof(vchnl_req));
+	vchnl_req.dev = dev;
+	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+	ret_code = vchnl_vf_send_add_hmc_objs_req(dev,
+						  &vchnl_req,
+						  rsrc_type,
+						  start_index,
+						  rsrc_count);
+	if (!ret_code) {
+		ret_code = i40iw_vf_wait_vchnl_resp(dev);
+		if (!ret_code)
+			ret_code = vchnl_req.ret_code;
+		else
+			dev->vchnl_up = false;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_vchnl_vf_del_hmc_obj - del HMC obj
+ * @dev: IWARP device pointer
+ * @rsrc_type: HMC Resource type
+ * @start_index: Starting index of the object to delete
+ * @rsrc_count: Number of resources to be delete
+ */
+enum i40iw_status_code i40iw_vchnl_vf_del_hmc_obj(struct i40iw_sc_dev *dev,
+						  enum i40iw_hmc_rsrc_type rsrc_type,
+						  u32 start_index,
+						  u32 rsrc_count)
+{
+	struct i40iw_virtchnl_req vchnl_req;
+	enum i40iw_status_code ret_code;
+
+	memset(&vchnl_req, 0, sizeof(vchnl_req));
+	vchnl_req.dev = dev;
+	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+	ret_code = vchnl_vf_send_del_hmc_objs_req(dev,
+						  &vchnl_req,
+						  rsrc_type,
+						  start_index,
+						  rsrc_count);
+	if (!ret_code) {
+		ret_code = i40iw_vf_wait_vchnl_resp(dev);
+		if (!ret_code)
+			ret_code = vchnl_req.ret_code;
+		else
+			dev->vchnl_up = false;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+	}
+	return ret_code;
+}
+
+/**
+ * i40iw_vchnl_vf_get_pe_stats - Get PE stats
+ * @dev: IWARP device pointer
+ * @hw_stats: HW stats struct
+ */
+enum i40iw_status_code i40iw_vchnl_vf_get_pe_stats(struct i40iw_sc_dev *dev,
+						   struct i40iw_dev_hw_stats *hw_stats)
+{
+	struct i40iw_virtchnl_req  vchnl_req;
+	enum i40iw_status_code ret_code;
+
+	memset(&vchnl_req, 0, sizeof(vchnl_req));
+	vchnl_req.dev = dev;
+	vchnl_req.parm = hw_stats;
+	vchnl_req.parm_len = sizeof(*hw_stats);
+	vchnl_req.vchnl_msg = &dev->vchnl_vf_msg_buf.vchnl_msg;
+	ret_code = vchnl_vf_send_get_pe_stats_req(dev, &vchnl_req);
+	if (!ret_code) {
+		ret_code = i40iw_vf_wait_vchnl_resp(dev);
+		if (!ret_code)
+			ret_code = vchnl_req.ret_code;
+		else
+			dev->vchnl_up = false;
+	} else {
+		i40iw_debug(dev, I40IW_DEBUG_VIRT,
+			    "%s Send message failed 0x%0x\n", __func__, ret_code);
+	}
+	return ret_code;
+}
diff --git a/drivers/infiniband/hw/i40iw/i40iw_virtchnl.h b/drivers/infiniband/hw/i40iw/i40iw_virtchnl.h
new file mode 100644
index 000000000000..24886ef08293
--- /dev/null
+++ b/drivers/infiniband/hw/i40iw/i40iw_virtchnl.h
@@ -0,0 +1,124 @@
+/*******************************************************************************
+*
+* Copyright (c) 2015-2016 Intel Corporation.  All rights reserved.
+*
+* This software is available to you under a choice of one of two
+* licenses.  You may choose to be licensed under the terms of the GNU
+* General Public License (GPL) Version 2, available from the file
+* COPYING in the main directory of this source tree, or the
+* OpenFabrics.org BSD license below:
+*
+*   Redistribution and use in source and binary forms, with or
+*   without modification, are permitted provided that the following
+*   conditions are met:
+*
+*    - Redistributions of source code must retain the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer.
+*
+*    - Redistributions in binary form must reproduce the above
+*	copyright notice, this list of conditions and the following
+*	disclaimer in the documentation and/or other materials
+*	provided with the distribution.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+* SOFTWARE.
+*
+*******************************************************************************/
+
+#ifndef I40IW_VIRTCHNL_H
+#define I40IW_VIRTCHNL_H
+
+#include "i40iw_hmc.h"
+
+#pragma pack(push, 1)
+
+struct i40iw_virtchnl_op_buf {
+	u16 iw_op_code;
+	u16 iw_op_ver;
+	u16 iw_chnl_buf_len;
+	u16 rsvd;
+	u64 iw_chnl_op_ctx;
+	/* Member alignment MUST be maintained above this location */
+	u8 iw_chnl_buf[1];
+};
+
+struct i40iw_virtchnl_resp_buf {
+	u64 iw_chnl_op_ctx;
+	u16 iw_chnl_buf_len;
+	s16 iw_op_ret_code;
+	/* Member alignment MUST be maintained above this location */
+	u16 rsvd[2];
+	u8 iw_chnl_buf[1];
+};
+
+enum i40iw_virtchnl_ops {
+	I40IW_VCHNL_OP_GET_VER = 0,
+	I40IW_VCHNL_OP_GET_HMC_FCN,
+	I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE,
+	I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE,
+	I40IW_VCHNL_OP_GET_STATS
+};
+
+#define I40IW_VCHNL_OP_GET_VER_V0 0
+#define I40IW_VCHNL_OP_GET_HMC_FCN_V0 0
+#define I40IW_VCHNL_OP_ADD_HMC_OBJ_RANGE_V0 0
+#define I40IW_VCHNL_OP_DEL_HMC_OBJ_RANGE_V0 0
+#define I40IW_VCHNL_OP_GET_STATS_V0 0
+#define I40IW_VCHNL_CHNL_VER_V0 0
+
+struct i40iw_dev_hw_stats;
+
+struct i40iw_virtchnl_hmc_obj_range {
+	u16 obj_type;
+	u16 rsvd;
+	u32 start_index;
+	u32 obj_count;
+};
+
+enum i40iw_status_code i40iw_vchnl_recv_pf(struct i40iw_sc_dev *dev,
+					   u32 vf_id,
+					   u8 *msg,
+					   u16 len);
+
+enum i40iw_status_code i40iw_vchnl_recv_vf(struct i40iw_sc_dev *dev,
+					   u32 vf_id,
+					   u8 *msg,
+					   u16 len);
+
+struct i40iw_virtchnl_req {
+	struct i40iw_sc_dev *dev;
+	struct i40iw_virtchnl_op_buf *vchnl_msg;
+	void *parm;
+	u32 vf_id;
+	u16 parm_len;
+	s16 ret_code;
+};
+
+#pragma pack(pop)
+
+enum i40iw_status_code i40iw_vchnl_vf_get_ver(struct i40iw_sc_dev *dev,
+					      u32 *vchnl_ver);
+
+enum i40iw_status_code i40iw_vchnl_vf_get_hmc_fcn(struct i40iw_sc_dev *dev,
+						  u16 *hmc_fcn);
+
+enum i40iw_status_code i40iw_vchnl_vf_add_hmc_objs(struct i40iw_sc_dev *dev,
+						   enum i40iw_hmc_rsrc_type rsrc_type,
+						   u32 start_index,
+						   u32 rsrc_count);
+
+enum i40iw_status_code i40iw_vchnl_vf_del_hmc_obj(struct i40iw_sc_dev *dev,
+						  enum i40iw_hmc_rsrc_type rsrc_type,
+						  u32 start_index,
+						  u32 rsrc_count);
+
+enum i40iw_status_code i40iw_vchnl_vf_get_pe_stats(struct i40iw_sc_dev *dev,
+						   struct i40iw_dev_hw_stats *hw_stats);
+#endif
diff --git a/drivers/infiniband/hw/mlx5/Makefile b/drivers/infiniband/hw/mlx5/Makefile
index 4e851889355a..7493a83acd28 100644
--- a/drivers/infiniband/hw/mlx5/Makefile
+++ b/drivers/infiniband/hw/mlx5/Makefile
@@ -1,4 +1,4 @@
 obj-$(CONFIG_MLX5_INFINIBAND)	+= mlx5_ib.o
 
-mlx5_ib-y :=	main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o
+mlx5_ib-y :=	main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o ib_virt.o
 mlx5_ib-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += odp.o
diff --git a/drivers/infiniband/hw/mlx5/ib_virt.c b/drivers/infiniband/hw/mlx5/ib_virt.c
new file mode 100644
index 000000000000..c1b9de800fe5
--- /dev/null
+++ b/drivers/infiniband/hw/mlx5/ib_virt.c
@@ -0,0 +1,194 @@
+/*
+ * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/mlx5/vport.h>
+#include "mlx5_ib.h"
+
+static inline u32 mlx_to_net_policy(enum port_state_policy mlx_policy)
+{
+	switch (mlx_policy) {
+	case MLX5_POLICY_DOWN:
+		return IFLA_VF_LINK_STATE_DISABLE;
+	case MLX5_POLICY_UP:
+		return IFLA_VF_LINK_STATE_ENABLE;
+	case MLX5_POLICY_FOLLOW:
+		return IFLA_VF_LINK_STATE_AUTO;
+	default:
+		return __IFLA_VF_LINK_STATE_MAX;
+	}
+}
+
+int mlx5_ib_get_vf_config(struct ib_device *device, int vf, u8 port,
+			  struct ifla_vf_info *info)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *rep;
+	int err;
+
+	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
+	if (!rep)
+		return -ENOMEM;
+
+	err = mlx5_query_hca_vport_context(mdev, 1, 1,  vf + 1, rep);
+	if (err) {
+		mlx5_ib_warn(dev, "failed to query port policy for vf %d (%d)\n",
+			     vf, err);
+		goto free;
+	}
+	memset(info, 0, sizeof(*info));
+	info->linkstate = mlx_to_net_policy(rep->policy);
+	if (info->linkstate == __IFLA_VF_LINK_STATE_MAX)
+		err = -EINVAL;
+
+free:
+	kfree(rep);
+	return err;
+}
+
+static inline enum port_state_policy net_to_mlx_policy(int policy)
+{
+	switch (policy) {
+	case IFLA_VF_LINK_STATE_DISABLE:
+		return MLX5_POLICY_DOWN;
+	case IFLA_VF_LINK_STATE_ENABLE:
+		return MLX5_POLICY_UP;
+	case IFLA_VF_LINK_STATE_AUTO:
+		return MLX5_POLICY_FOLLOW;
+	default:
+		return MLX5_POLICY_INVALID;
+	}
+}
+
+int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf,
+			      u8 port, int state)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->policy = net_to_mlx_policy(state);
+	if (in->policy == MLX5_POLICY_INVALID) {
+		err = -EINVAL;
+		goto out;
+	}
+	in->field_select = MLX5_HCA_VPORT_SEL_STATE_POLICY;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+
+out:
+	kfree(in);
+	return err;
+}
+
+int mlx5_ib_get_vf_stats(struct ib_device *device, int vf,
+			 u8 port, struct ifla_vf_stats *stats)
+{
+	int out_sz = MLX5_ST_SZ_BYTES(query_vport_counter_out);
+	struct mlx5_core_dev *mdev;
+	struct mlx5_ib_dev *dev;
+	void *out;
+	int err;
+
+	dev = to_mdev(device);
+	mdev = dev->mdev;
+
+	out = kzalloc(out_sz, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	err = mlx5_core_query_vport_counter(mdev, true, vf, port, out, out_sz);
+	if (err)
+		goto ex;
+
+	stats->rx_packets = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_unicast.packets);
+	stats->tx_packets = MLX5_GET64_PR(query_vport_counter_out, out, transmitted_ib_unicast.packets);
+	stats->rx_bytes = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_unicast.octets);
+	stats->tx_bytes = MLX5_GET64_PR(query_vport_counter_out, out, transmitted_ib_unicast.octets);
+	stats->multicast = MLX5_GET64_PR(query_vport_counter_out, out, received_ib_multicast.packets);
+
+ex:
+	kfree(out);
+	return err;
+}
+
+static int set_vf_node_guid(struct ib_device *device, int vf, u8 port, u64 guid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->field_select = MLX5_HCA_VPORT_SEL_NODE_GUID;
+	in->node_guid = guid;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+	kfree(in);
+	return err;
+}
+
+static int set_vf_port_guid(struct ib_device *device, int vf, u8 port, u64 guid)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_hca_vport_context *in;
+	int err;
+
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	in->field_select = MLX5_HCA_VPORT_SEL_PORT_GUID;
+	in->port_guid = guid;
+	err = mlx5_core_modify_hca_vport_context(mdev, 1, 1, vf + 1, in);
+	kfree(in);
+	return err;
+}
+
+int mlx5_ib_set_vf_guid(struct ib_device *device, int vf, u8 port,
+			u64 guid, int type)
+{
+	if (type == IFLA_VF_IB_NODE_GUID)
+		return set_vf_node_guid(device, vf, port, guid);
+	else if (type == IFLA_VF_IB_PORT_GUID)
+		return set_vf_port_guid(device, vf, port, guid);
+
+	return -EINVAL;
+}
diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c
index 41d8a0036465..1534af113058 100644
--- a/drivers/infiniband/hw/mlx5/mad.c
+++ b/drivers/infiniband/hw/mlx5/mad.c
@@ -208,7 +208,7 @@ static int process_pma_cmd(struct ib_device *ibdev, u8 port_num,
 		if (!out_cnt)
 			return IB_MAD_RESULT_FAILURE;
 
-		err = mlx5_core_query_vport_counter(dev->mdev, 0,
+		err = mlx5_core_query_vport_counter(dev->mdev, 0, 0,
 						    port_num, out_cnt, sz);
 		if (!err)
 			pma_cnt_ext_assign(pma_cnt_ext, out_cnt);
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index edd8b8741846..5acf346e048e 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -284,7 +284,7 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
 
 static int mlx5_use_mad_ifc(struct mlx5_ib_dev *dev)
 {
-	return !dev->mdev->issi;
+	return !MLX5_CAP_GEN(dev->mdev, ib_virt);
 }
 
 enum {
@@ -563,6 +563,9 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 	if (MLX5_CAP_GEN(mdev, cd))
 		props->device_cap_flags |= IB_DEVICE_CROSS_CHANNEL;
 
+	if (!mlx5_core_is_pf(mdev))
+		props->device_cap_flags |= IB_DEVICE_VIRTUAL_FUNCTION;
+
 	return 0;
 }
 
@@ -700,6 +703,7 @@ static int mlx5_query_hca_port(struct ib_device *ibdev, u8 port,
 	props->qkey_viol_cntr	= rep->qkey_violation_counter;
 	props->subnet_timeout	= rep->subnet_timeout;
 	props->init_type_reply	= rep->init_type_reply;
+	props->grh_required	= rep->grh_required;
 
 	err = mlx5_query_port_link_width_oper(mdev, &ib_link_width_oper, port);
 	if (err)
@@ -2350,6 +2354,12 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 	dev->ib_dev.map_mr_sg		= mlx5_ib_map_mr_sg;
 	dev->ib_dev.check_mr_status	= mlx5_ib_check_mr_status;
 	dev->ib_dev.get_port_immutable  = mlx5_port_immutable;
+	if (mlx5_core_is_pf(mdev)) {
+		dev->ib_dev.get_vf_config	= mlx5_ib_get_vf_config;
+		dev->ib_dev.set_vf_link_state	= mlx5_ib_set_vf_link_state;
+		dev->ib_dev.get_vf_stats	= mlx5_ib_get_vf_stats;
+		dev->ib_dev.set_vf_guid		= mlx5_ib_set_vf_guid;
+	}
 
 	mlx5_ib_internal_fill_odp_caps(dev);
 
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 76b2b42e0535..f16c818ad2e6 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -776,6 +776,14 @@ void mlx5_ib_qp_disable_pagefaults(struct mlx5_ib_qp *qp);
 void mlx5_ib_qp_enable_pagefaults(struct mlx5_ib_qp *qp);
 void mlx5_ib_invalidate_range(struct ib_umem *umem, unsigned long start,
 			      unsigned long end);
+int mlx5_ib_get_vf_config(struct ib_device *device, int vf,
+			  u8 port, struct ifla_vf_info *info);
+int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf,
+			      u8 port, int state);
+int mlx5_ib_get_vf_stats(struct ib_device *device, int vf,
+			 u8 port, struct ifla_vf_stats *stats);
+int mlx5_ib_set_vf_guid(struct ib_device *device, int vf, u8 port,
+			u64 guid, int type);
 
 #else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
 static inline void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev)
diff --git a/drivers/infiniband/hw/qib/Kconfig b/drivers/infiniband/hw/qib/Kconfig
index 495be09781b1..e0fdb9201423 100644
--- a/drivers/infiniband/hw/qib/Kconfig
+++ b/drivers/infiniband/hw/qib/Kconfig
@@ -1,6 +1,6 @@
 config INFINIBAND_QIB
 	tristate "Intel PCIe HCA support"
-	depends on 64BIT
+	depends on 64BIT && INFINIBAND_RDMAVT
 	---help---
 	This is a low-level driver for Intel PCIe QLE InfiniBand host
 	channel adapters.  This driver does not support the Intel
diff --git a/drivers/infiniband/hw/qib/Makefile b/drivers/infiniband/hw/qib/Makefile
index 57f8103e51f8..79ebd79e8405 100644
--- a/drivers/infiniband/hw/qib/Makefile
+++ b/drivers/infiniband/hw/qib/Makefile
@@ -1,11 +1,11 @@
 obj-$(CONFIG_INFINIBAND_QIB) += ib_qib.o
 
-ib_qib-y := qib_cq.o qib_diag.o qib_dma.o qib_driver.o qib_eeprom.o \
-	qib_file_ops.o qib_fs.o qib_init.o qib_intr.o qib_keys.o \
-	qib_mad.o qib_mmap.o qib_mr.o qib_pcie.o qib_pio_copy.o \
-	qib_qp.o qib_qsfp.o qib_rc.o qib_ruc.o qib_sdma.o qib_srq.o \
+ib_qib-y := qib_diag.o qib_driver.o qib_eeprom.o \
+	qib_file_ops.o qib_fs.o qib_init.o qib_intr.o \
+	qib_mad.o qib_pcie.o qib_pio_copy.o \
+	qib_qp.o qib_qsfp.o qib_rc.o qib_ruc.o qib_sdma.o \
 	qib_sysfs.o qib_twsi.o qib_tx.o qib_uc.o qib_ud.o \
-	qib_user_pages.o qib_user_sdma.o qib_verbs_mcast.o qib_iba7220.o \
+	qib_user_pages.o qib_user_sdma.o qib_iba7220.o \
 	qib_sd7220.o qib_iba7322.o qib_verbs.o
 
 # 6120 has no fallback if no MSI interrupts, others can do INTx
diff --git a/drivers/infiniband/hw/qib/qib.h b/drivers/infiniband/hw/qib/qib.h
index 7df16f74bb45..bbf0a163aeab 100644
--- a/drivers/infiniband/hw/qib/qib.h
+++ b/drivers/infiniband/hw/qib/qib.h
@@ -52,6 +52,7 @@
 #include <linux/kref.h>
 #include <linux/sched.h>
 #include <linux/kthread.h>
+#include <rdma/rdma_vt.h>
 
 #include "qib_common.h"
 #include "qib_verbs.h"
@@ -229,9 +230,6 @@ struct qib_ctxtdata {
 	u8 redirect_seq_cnt;
 	/* ctxt rcvhdrq head offset */
 	u32 head;
-	/* lookaside fields */
-	struct qib_qp *lookaside_qp;
-	u32 lookaside_qpn;
 	/* QPs waiting for context processing */
 	struct list_head qp_wait_list;
 #ifdef CONFIG_DEBUG_FS
@@ -240,7 +238,7 @@ struct qib_ctxtdata {
 #endif
 };
 
-struct qib_sge_state;
+struct rvt_sge_state;
 
 struct qib_sdma_txreq {
 	int                 flags;
@@ -258,14 +256,14 @@ struct qib_sdma_desc {
 
 struct qib_verbs_txreq {
 	struct qib_sdma_txreq   txreq;
-	struct qib_qp           *qp;
-	struct qib_swqe         *wqe;
+	struct rvt_qp           *qp;
+	struct rvt_swqe         *wqe;
 	u32                     dwords;
 	u16                     hdr_dwords;
 	u16                     hdr_inx;
 	struct qib_pio_header	*align_buf;
-	struct qib_mregion	*mr;
-	struct qib_sge_state    *ss;
+	struct rvt_mregion	*mr;
+	struct rvt_sge_state    *ss;
 };
 
 #define QIB_SDMA_TXREQ_F_USELARGEBUF  0x1
@@ -1096,8 +1094,6 @@ struct qib_devdata {
 	u16 psxmitwait_check_rate;
 	/* high volume overflow errors defered to tasklet */
 	struct tasklet_struct error_tasklet;
-	/* per device cq worker */
-	struct kthread_worker *worker;
 
 	int assigned_node_id; /* NUMA node closest to HCA */
 };
@@ -1135,8 +1131,9 @@ extern spinlock_t qib_devs_lock;
 extern struct qib_devdata *qib_lookup(int unit);
 extern u32 qib_cpulist_count;
 extern unsigned long *qib_cpulist;
-
+extern u16 qpt_mask;
 extern unsigned qib_cc_table_size;
+
 int qib_init(struct qib_devdata *, int);
 int init_chip_wc_pat(struct qib_devdata *dd, u32);
 int qib_enable_wc(struct qib_devdata *dd);
@@ -1323,7 +1320,7 @@ void __qib_sdma_intr(struct qib_pportdata *);
 void qib_sdma_intr(struct qib_pportdata *);
 void qib_user_sdma_send_desc(struct qib_pportdata *dd,
 			struct list_head *pktlist);
-int qib_sdma_verbs_send(struct qib_pportdata *, struct qib_sge_state *,
+int qib_sdma_verbs_send(struct qib_pportdata *, struct rvt_sge_state *,
 			u32, struct qib_verbs_txreq *);
 /* ppd->sdma_lock should be locked before calling this. */
 int qib_sdma_make_progress(struct qib_pportdata *dd);
@@ -1454,6 +1451,8 @@ u64 qib_sps_ints(void);
 dma_addr_t qib_map_page(struct pci_dev *, struct page *, unsigned long,
 			  size_t, int);
 const char *qib_get_unit_name(int unit);
+const char *qib_get_card_name(struct rvt_dev_info *rdi);
+struct pci_dev *qib_get_pci_dev(struct rvt_dev_info *rdi);
 
 /*
  * Flush write combining store buffers (if present) and perform a write
@@ -1540,4 +1539,14 @@ struct qib_hwerror_msgs {
 void qib_format_hwerrors(u64 hwerrs,
 			 const struct qib_hwerror_msgs *hwerrmsgs,
 			 size_t nhwerrmsgs, char *msg, size_t lmsg);
+
+void qib_stop_send_queue(struct rvt_qp *qp);
+void qib_quiesce_qp(struct rvt_qp *qp);
+void qib_flush_qp_waiters(struct rvt_qp *qp);
+int qib_mtu_to_path_mtu(u32 mtu);
+u32 qib_mtu_from_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, u32 pmtu);
+void qib_notify_error_qp(struct rvt_qp *qp);
+int qib_get_pmtu_from_attr(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+			   struct ib_qp_attr *attr);
+
 #endif                          /* _QIB_KERNEL_H */
diff --git a/drivers/infiniband/hw/qib/qib_common.h b/drivers/infiniband/hw/qib/qib_common.h
index 4fb78abd8ba1..1d6e63eb1146 100644
--- a/drivers/infiniband/hw/qib/qib_common.h
+++ b/drivers/infiniband/hw/qib/qib_common.h
@@ -742,14 +742,11 @@ struct qib_tid_session_member {
 #define SIZE_OF_CRC 1
 
 #define QIB_DEFAULT_P_KEY 0xFFFF
-#define QIB_PERMISSIVE_LID 0xFFFF
 #define QIB_AETH_CREDIT_SHIFT 24
 #define QIB_AETH_CREDIT_MASK 0x1F
 #define QIB_AETH_CREDIT_INVAL 0x1F
 #define QIB_PSN_MASK 0xFFFFFF
 #define QIB_MSN_MASK 0xFFFFFF
-#define QIB_QPN_MASK 0xFFFFFF
-#define QIB_MULTICAST_LID_BASE 0xC000
 #define QIB_EAGER_TID_ID QLOGIC_IB_I_TID_MASK
 #define QIB_MULTICAST_QPN 0xFFFFFF
 
diff --git a/drivers/infiniband/hw/qib/qib_cq.c b/drivers/infiniband/hw/qib/qib_cq.c
deleted file mode 100644
index 2b45d0b02300..000000000000
--- a/drivers/infiniband/hw/qib/qib_cq.c
+++ /dev/null
@@ -1,545 +0,0 @@
-/*
- * Copyright (c) 2013 Intel Corporation.  All rights reserved.
- * Copyright (c) 2006, 2007, 2008, 2010 QLogic Corporation. All rights reserved.
- * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <linux/err.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-#include <linux/kthread.h>
-
-#include "qib_verbs.h"
-#include "qib.h"
-
-/**
- * qib_cq_enter - add a new entry to the completion queue
- * @cq: completion queue
- * @entry: work completion entry to add
- * @sig: true if @entry is a solicitated entry
- *
- * This may be called with qp->s_lock held.
- */
-void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int solicited)
-{
-	struct qib_cq_wc *wc;
-	unsigned long flags;
-	u32 head;
-	u32 next;
-
-	spin_lock_irqsave(&cq->lock, flags);
-
-	/*
-	 * Note that the head pointer might be writable by user processes.
-	 * Take care to verify it is a sane value.
-	 */
-	wc = cq->queue;
-	head = wc->head;
-	if (head >= (unsigned) cq->ibcq.cqe) {
-		head = cq->ibcq.cqe;
-		next = 0;
-	} else
-		next = head + 1;
-	if (unlikely(next == wc->tail)) {
-		spin_unlock_irqrestore(&cq->lock, flags);
-		if (cq->ibcq.event_handler) {
-			struct ib_event ev;
-
-			ev.device = cq->ibcq.device;
-			ev.element.cq = &cq->ibcq;
-			ev.event = IB_EVENT_CQ_ERR;
-			cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
-		}
-		return;
-	}
-	if (cq->ip) {
-		wc->uqueue[head].wr_id = entry->wr_id;
-		wc->uqueue[head].status = entry->status;
-		wc->uqueue[head].opcode = entry->opcode;
-		wc->uqueue[head].vendor_err = entry->vendor_err;
-		wc->uqueue[head].byte_len = entry->byte_len;
-		wc->uqueue[head].ex.imm_data =
-			(__u32 __force)entry->ex.imm_data;
-		wc->uqueue[head].qp_num = entry->qp->qp_num;
-		wc->uqueue[head].src_qp = entry->src_qp;
-		wc->uqueue[head].wc_flags = entry->wc_flags;
-		wc->uqueue[head].pkey_index = entry->pkey_index;
-		wc->uqueue[head].slid = entry->slid;
-		wc->uqueue[head].sl = entry->sl;
-		wc->uqueue[head].dlid_path_bits = entry->dlid_path_bits;
-		wc->uqueue[head].port_num = entry->port_num;
-		/* Make sure entry is written before the head index. */
-		smp_wmb();
-	} else
-		wc->kqueue[head] = *entry;
-	wc->head = next;
-
-	if (cq->notify == IB_CQ_NEXT_COMP ||
-	    (cq->notify == IB_CQ_SOLICITED &&
-	     (solicited || entry->status != IB_WC_SUCCESS))) {
-		struct kthread_worker *worker;
-		/*
-		 * This will cause send_complete() to be called in
-		 * another thread.
-		 */
-		smp_rmb();
-		worker = cq->dd->worker;
-		if (likely(worker)) {
-			cq->notify = IB_CQ_NONE;
-			cq->triggered++;
-			queue_kthread_work(worker, &cq->comptask);
-		}
-	}
-
-	spin_unlock_irqrestore(&cq->lock, flags);
-}
-
-/**
- * qib_poll_cq - poll for work completion entries
- * @ibcq: the completion queue to poll
- * @num_entries: the maximum number of entries to return
- * @entry: pointer to array where work completions are placed
- *
- * Returns the number of completion entries polled.
- *
- * This may be called from interrupt context.  Also called by ib_poll_cq()
- * in the generic verbs code.
- */
-int qib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
-{
-	struct qib_cq *cq = to_icq(ibcq);
-	struct qib_cq_wc *wc;
-	unsigned long flags;
-	int npolled;
-	u32 tail;
-
-	/* The kernel can only poll a kernel completion queue */
-	if (cq->ip) {
-		npolled = -EINVAL;
-		goto bail;
-	}
-
-	spin_lock_irqsave(&cq->lock, flags);
-
-	wc = cq->queue;
-	tail = wc->tail;
-	if (tail > (u32) cq->ibcq.cqe)
-		tail = (u32) cq->ibcq.cqe;
-	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
-		if (tail == wc->head)
-			break;
-		/* The kernel doesn't need a RMB since it has the lock. */
-		*entry = wc->kqueue[tail];
-		if (tail >= cq->ibcq.cqe)
-			tail = 0;
-		else
-			tail++;
-	}
-	wc->tail = tail;
-
-	spin_unlock_irqrestore(&cq->lock, flags);
-
-bail:
-	return npolled;
-}
-
-static void send_complete(struct kthread_work *work)
-{
-	struct qib_cq *cq = container_of(work, struct qib_cq, comptask);
-
-	/*
-	 * The completion handler will most likely rearm the notification
-	 * and poll for all pending entries.  If a new completion entry
-	 * is added while we are in this routine, queue_work()
-	 * won't call us again until we return so we check triggered to
-	 * see if we need to call the handler again.
-	 */
-	for (;;) {
-		u8 triggered = cq->triggered;
-
-		/*
-		 * IPoIB connected mode assumes the callback is from a
-		 * soft IRQ. We simulate this by blocking "bottom halves".
-		 * See the implementation for ipoib_cm_handle_tx_wc(),
-		 * netif_tx_lock_bh() and netif_tx_lock().
-		 */
-		local_bh_disable();
-		cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
-		local_bh_enable();
-
-		if (cq->triggered == triggered)
-			return;
-	}
-}
-
-/**
- * qib_create_cq - create a completion queue
- * @ibdev: the device this completion queue is attached to
- * @attr: creation attributes
- * @context: unused by the QLogic_IB driver
- * @udata: user data for libibverbs.so
- *
- * Returns a pointer to the completion queue or negative errno values
- * for failure.
- *
- * Called by ib_create_cq() in the generic verbs code.
- */
-struct ib_cq *qib_create_cq(struct ib_device *ibdev,
-			    const struct ib_cq_init_attr *attr,
-			    struct ib_ucontext *context,
-			    struct ib_udata *udata)
-{
-	int entries = attr->cqe;
-	struct qib_ibdev *dev = to_idev(ibdev);
-	struct qib_cq *cq;
-	struct qib_cq_wc *wc;
-	struct ib_cq *ret;
-	u32 sz;
-
-	if (attr->flags)
-		return ERR_PTR(-EINVAL);
-
-	if (entries < 1 || entries > ib_qib_max_cqes) {
-		ret = ERR_PTR(-EINVAL);
-		goto done;
-	}
-
-	/* Allocate the completion queue structure. */
-	cq = kmalloc(sizeof(*cq), GFP_KERNEL);
-	if (!cq) {
-		ret = ERR_PTR(-ENOMEM);
-		goto done;
-	}
-
-	/*
-	 * Allocate the completion queue entries and head/tail pointers.
-	 * This is allocated separately so that it can be resized and
-	 * also mapped into user space.
-	 * We need to use vmalloc() in order to support mmap and large
-	 * numbers of entries.
-	 */
-	sz = sizeof(*wc);
-	if (udata && udata->outlen >= sizeof(__u64))
-		sz += sizeof(struct ib_uverbs_wc) * (entries + 1);
-	else
-		sz += sizeof(struct ib_wc) * (entries + 1);
-	wc = vmalloc_user(sz);
-	if (!wc) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_cq;
-	}
-
-	/*
-	 * Return the address of the WC as the offset to mmap.
-	 * See qib_mmap() for details.
-	 */
-	if (udata && udata->outlen >= sizeof(__u64)) {
-		int err;
-
-		cq->ip = qib_create_mmap_info(dev, sz, context, wc);
-		if (!cq->ip) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_wc;
-		}
-
-		err = ib_copy_to_udata(udata, &cq->ip->offset,
-				       sizeof(cq->ip->offset));
-		if (err) {
-			ret = ERR_PTR(err);
-			goto bail_ip;
-		}
-	} else
-		cq->ip = NULL;
-
-	spin_lock(&dev->n_cqs_lock);
-	if (dev->n_cqs_allocated == ib_qib_max_cqs) {
-		spin_unlock(&dev->n_cqs_lock);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_ip;
-	}
-
-	dev->n_cqs_allocated++;
-	spin_unlock(&dev->n_cqs_lock);
-
-	if (cq->ip) {
-		spin_lock_irq(&dev->pending_lock);
-		list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-	}
-
-	/*
-	 * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
-	 * The number of entries should be >= the number requested or return
-	 * an error.
-	 */
-	cq->dd = dd_from_dev(dev);
-	cq->ibcq.cqe = entries;
-	cq->notify = IB_CQ_NONE;
-	cq->triggered = 0;
-	spin_lock_init(&cq->lock);
-	init_kthread_work(&cq->comptask, send_complete);
-	wc->head = 0;
-	wc->tail = 0;
-	cq->queue = wc;
-
-	ret = &cq->ibcq;
-
-	goto done;
-
-bail_ip:
-	kfree(cq->ip);
-bail_wc:
-	vfree(wc);
-bail_cq:
-	kfree(cq);
-done:
-	return ret;
-}
-
-/**
- * qib_destroy_cq - destroy a completion queue
- * @ibcq: the completion queue to destroy.
- *
- * Returns 0 for success.
- *
- * Called by ib_destroy_cq() in the generic verbs code.
- */
-int qib_destroy_cq(struct ib_cq *ibcq)
-{
-	struct qib_ibdev *dev = to_idev(ibcq->device);
-	struct qib_cq *cq = to_icq(ibcq);
-
-	flush_kthread_work(&cq->comptask);
-	spin_lock(&dev->n_cqs_lock);
-	dev->n_cqs_allocated--;
-	spin_unlock(&dev->n_cqs_lock);
-	if (cq->ip)
-		kref_put(&cq->ip->ref, qib_release_mmap_info);
-	else
-		vfree(cq->queue);
-	kfree(cq);
-
-	return 0;
-}
-
-/**
- * qib_req_notify_cq - change the notification type for a completion queue
- * @ibcq: the completion queue
- * @notify_flags: the type of notification to request
- *
- * Returns 0 for success.
- *
- * This may be called from interrupt context.  Also called by
- * ib_req_notify_cq() in the generic verbs code.
- */
-int qib_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
-{
-	struct qib_cq *cq = to_icq(ibcq);
-	unsigned long flags;
-	int ret = 0;
-
-	spin_lock_irqsave(&cq->lock, flags);
-	/*
-	 * Don't change IB_CQ_NEXT_COMP to IB_CQ_SOLICITED but allow
-	 * any other transitions (see C11-31 and C11-32 in ch. 11.4.2.2).
-	 */
-	if (cq->notify != IB_CQ_NEXT_COMP)
-		cq->notify = notify_flags & IB_CQ_SOLICITED_MASK;
-
-	if ((notify_flags & IB_CQ_REPORT_MISSED_EVENTS) &&
-	    cq->queue->head != cq->queue->tail)
-		ret = 1;
-
-	spin_unlock_irqrestore(&cq->lock, flags);
-
-	return ret;
-}
-
-/**
- * qib_resize_cq - change the size of the CQ
- * @ibcq: the completion queue
- *
- * Returns 0 for success.
- */
-int qib_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
-{
-	struct qib_cq *cq = to_icq(ibcq);
-	struct qib_cq_wc *old_wc;
-	struct qib_cq_wc *wc;
-	u32 head, tail, n;
-	int ret;
-	u32 sz;
-
-	if (cqe < 1 || cqe > ib_qib_max_cqes) {
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	/*
-	 * Need to use vmalloc() if we want to support large #s of entries.
-	 */
-	sz = sizeof(*wc);
-	if (udata && udata->outlen >= sizeof(__u64))
-		sz += sizeof(struct ib_uverbs_wc) * (cqe + 1);
-	else
-		sz += sizeof(struct ib_wc) * (cqe + 1);
-	wc = vmalloc_user(sz);
-	if (!wc) {
-		ret = -ENOMEM;
-		goto bail;
-	}
-
-	/* Check that we can write the offset to mmap. */
-	if (udata && udata->outlen >= sizeof(__u64)) {
-		__u64 offset = 0;
-
-		ret = ib_copy_to_udata(udata, &offset, sizeof(offset));
-		if (ret)
-			goto bail_free;
-	}
-
-	spin_lock_irq(&cq->lock);
-	/*
-	 * Make sure head and tail are sane since they
-	 * might be user writable.
-	 */
-	old_wc = cq->queue;
-	head = old_wc->head;
-	if (head > (u32) cq->ibcq.cqe)
-		head = (u32) cq->ibcq.cqe;
-	tail = old_wc->tail;
-	if (tail > (u32) cq->ibcq.cqe)
-		tail = (u32) cq->ibcq.cqe;
-	if (head < tail)
-		n = cq->ibcq.cqe + 1 + head - tail;
-	else
-		n = head - tail;
-	if (unlikely((u32)cqe < n)) {
-		ret = -EINVAL;
-		goto bail_unlock;
-	}
-	for (n = 0; tail != head; n++) {
-		if (cq->ip)
-			wc->uqueue[n] = old_wc->uqueue[tail];
-		else
-			wc->kqueue[n] = old_wc->kqueue[tail];
-		if (tail == (u32) cq->ibcq.cqe)
-			tail = 0;
-		else
-			tail++;
-	}
-	cq->ibcq.cqe = cqe;
-	wc->head = n;
-	wc->tail = 0;
-	cq->queue = wc;
-	spin_unlock_irq(&cq->lock);
-
-	vfree(old_wc);
-
-	if (cq->ip) {
-		struct qib_ibdev *dev = to_idev(ibcq->device);
-		struct qib_mmap_info *ip = cq->ip;
-
-		qib_update_mmap_info(dev, ip, sz, wc);
-
-		/*
-		 * Return the offset to mmap.
-		 * See qib_mmap() for details.
-		 */
-		if (udata && udata->outlen >= sizeof(__u64)) {
-			ret = ib_copy_to_udata(udata, &ip->offset,
-					       sizeof(ip->offset));
-			if (ret)
-				goto bail;
-		}
-
-		spin_lock_irq(&dev->pending_lock);
-		if (list_empty(&ip->pending_mmaps))
-			list_add(&ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-	}
-
-	ret = 0;
-	goto bail;
-
-bail_unlock:
-	spin_unlock_irq(&cq->lock);
-bail_free:
-	vfree(wc);
-bail:
-	return ret;
-}
-
-int qib_cq_init(struct qib_devdata *dd)
-{
-	int ret = 0;
-	int cpu;
-	struct task_struct *task;
-
-	if (dd->worker)
-		return 0;
-	dd->worker = kzalloc(sizeof(*dd->worker), GFP_KERNEL);
-	if (!dd->worker)
-		return -ENOMEM;
-	init_kthread_worker(dd->worker);
-	task = kthread_create_on_node(
-		kthread_worker_fn,
-		dd->worker,
-		dd->assigned_node_id,
-		"qib_cq%d", dd->unit);
-	if (IS_ERR(task))
-		goto task_fail;
-	cpu = cpumask_first(cpumask_of_node(dd->assigned_node_id));
-	kthread_bind(task, cpu);
-	wake_up_process(task);
-out:
-	return ret;
-task_fail:
-	ret = PTR_ERR(task);
-	kfree(dd->worker);
-	dd->worker = NULL;
-	goto out;
-}
-
-void qib_cq_exit(struct qib_devdata *dd)
-{
-	struct kthread_worker *worker;
-
-	worker = dd->worker;
-	if (!worker)
-		return;
-	/* blocks future queuing from send_complete() */
-	dd->worker = NULL;
-	smp_wmb();
-	flush_kthread_worker(worker);
-	kthread_stop(worker->task);
-	kfree(worker);
-}
diff --git a/drivers/infiniband/hw/qib/qib_driver.c b/drivers/infiniband/hw/qib/qib_driver.c
index f58fdc3d25a2..67ee6438cf59 100644
--- a/drivers/infiniband/hw/qib/qib_driver.c
+++ b/drivers/infiniband/hw/qib/qib_driver.c
@@ -90,6 +90,22 @@ const char *qib_get_unit_name(int unit)
 	return iname;
 }
 
+const char *qib_get_card_name(struct rvt_dev_info *rdi)
+{
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(ibdev,
+					      struct qib_devdata, verbs_dev);
+	return qib_get_unit_name(dd->unit);
+}
+
+struct pci_dev *qib_get_pci_dev(struct rvt_dev_info *rdi)
+{
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(ibdev,
+					      struct qib_devdata, verbs_dev);
+	return dd->pcidev;
+}
+
 /*
  * Return count of units with at least one port ACTIVE.
  */
@@ -306,7 +322,9 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 		struct qib_ib_header *hdr = (struct qib_ib_header *) rhdr;
 		struct qib_other_headers *ohdr = NULL;
 		struct qib_ibport *ibp = &ppd->ibport_data;
-		struct qib_qp *qp = NULL;
+		struct qib_devdata *dd = ppd->dd;
+		struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
+		struct rvt_qp *qp = NULL;
 		u32 tlen = qib_hdrget_length_in_bytes(rhf_addr);
 		u16 lid  = be16_to_cpu(hdr->lrh[1]);
 		int lnh = be16_to_cpu(hdr->lrh[0]) & 3;
@@ -319,7 +337,7 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 		if (tlen < 24)
 			goto drop;
 
-		if (lid < QIB_MULTICAST_LID_BASE) {
+		if (lid < be16_to_cpu(IB_MULTICAST_LID_BASE)) {
 			lid &= ~((1 << ppd->lmc) - 1);
 			if (unlikely(lid != ppd->lid))
 				goto drop;
@@ -346,13 +364,16 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 		psn = be32_to_cpu(ohdr->bth[2]);
 
 		/* Get the destination QP number. */
-		qp_num = be32_to_cpu(ohdr->bth[1]) & QIB_QPN_MASK;
+		qp_num = be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
 		if (qp_num != QIB_MULTICAST_QPN) {
 			int ruc_res;
 
-			qp = qib_lookup_qpn(ibp, qp_num);
-			if (!qp)
+			rcu_read_lock();
+			qp = rvt_lookup_qpn(rdi, &ibp->rvp, qp_num);
+			if (!qp) {
+				rcu_read_unlock();
 				goto drop;
+			}
 
 			/*
 			 * Handle only RC QPs - for other QP types drop error
@@ -361,9 +382,9 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 			spin_lock(&qp->r_lock);
 
 			/* Check for valid receive state. */
-			if (!(ib_qib_state_ops[qp->state] &
-			      QIB_PROCESS_RECV_OK)) {
-				ibp->n_pkt_drops++;
+			if (!(ib_rvt_state_ops[qp->state] &
+			      RVT_PROCESS_RECV_OK)) {
+				ibp->rvp.n_pkt_drops++;
 				goto unlock;
 			}
 
@@ -383,7 +404,7 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 				    IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST) {
 					diff = qib_cmp24(psn, qp->r_psn);
 					if (!qp->r_nak_state && diff >= 0) {
-						ibp->n_rc_seqnak++;
+						ibp->rvp.n_rc_seqnak++;
 						qp->r_nak_state =
 							IB_NAK_PSN_ERROR;
 						/* Use the expected PSN. */
@@ -398,7 +419,7 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 						 */
 						if (list_empty(&qp->rspwait)) {
 							qp->r_flags |=
-								QIB_R_RSP_NAK;
+								RVT_R_RSP_NAK;
 							atomic_inc(
 								&qp->refcount);
 							list_add_tail(
@@ -419,12 +440,7 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd,
 
 unlock:
 			spin_unlock(&qp->r_lock);
-			/*
-			 * Notify qib_destroy_qp() if it is waiting
-			 * for us to finish.
-			 */
-			if (atomic_dec_and_test(&qp->refcount))
-				wake_up(&qp->wait);
+			rcu_read_unlock();
 		} /* Unicast QP */
 	} /* Valid packet with TIDErr */
 
@@ -456,7 +472,7 @@ u32 qib_kreceive(struct qib_ctxtdata *rcd, u32 *llic, u32 *npkts)
 	u32 eflags, etype, tlen, i = 0, updegr = 0, crcs = 0;
 	int last;
 	u64 lval;
-	struct qib_qp *qp, *nqp;
+	struct rvt_qp *qp, *nqp;
 
 	l = rcd->head;
 	rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset;
@@ -549,15 +565,6 @@ move_along:
 			updegr = 0;
 		}
 	}
-	/*
-	 * Notify qib_destroy_qp() if it is waiting
-	 * for lookaside_qp to finish.
-	 */
-	if (rcd->lookaside_qp) {
-		if (atomic_dec_and_test(&rcd->lookaside_qp->refcount))
-			wake_up(&rcd->lookaside_qp->wait);
-		rcd->lookaside_qp = NULL;
-	}
 
 	rcd->head = l;
 
@@ -567,17 +574,17 @@ move_along:
 	 */
 	list_for_each_entry_safe(qp, nqp, &rcd->qp_wait_list, rspwait) {
 		list_del_init(&qp->rspwait);
-		if (qp->r_flags & QIB_R_RSP_NAK) {
-			qp->r_flags &= ~QIB_R_RSP_NAK;
+		if (qp->r_flags & RVT_R_RSP_NAK) {
+			qp->r_flags &= ~RVT_R_RSP_NAK;
 			qib_send_rc_ack(qp);
 		}
-		if (qp->r_flags & QIB_R_RSP_SEND) {
+		if (qp->r_flags & RVT_R_RSP_SEND) {
 			unsigned long flags;
 
-			qp->r_flags &= ~QIB_R_RSP_SEND;
+			qp->r_flags &= ~RVT_R_RSP_SEND;
 			spin_lock_irqsave(&qp->s_lock, flags);
-			if (ib_qib_state_ops[qp->state] &
-					QIB_PROCESS_OR_FLUSH_SEND)
+			if (ib_rvt_state_ops[qp->state] &
+					RVT_PROCESS_OR_FLUSH_SEND)
 				qib_schedule_send(qp);
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 		}
diff --git a/drivers/infiniband/hw/qib/qib_iba6120.c b/drivers/infiniband/hw/qib/qib_iba6120.c
index 4b927809d1a1..a3733f25280f 100644
--- a/drivers/infiniband/hw/qib/qib_iba6120.c
+++ b/drivers/infiniband/hw/qib/qib_iba6120.c
@@ -2956,13 +2956,13 @@ static void pma_6120_timer(unsigned long data)
 	struct qib_ibport *ibp = &ppd->ibport_data;
 	unsigned long flags;
 
-	spin_lock_irqsave(&ibp->lock, flags);
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
 	if (cs->pma_sample_status == IB_PMA_SAMPLE_STATUS_STARTED) {
 		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_RUNNING;
 		qib_snapshot_counters(ppd, &cs->sword, &cs->rword,
 				      &cs->spkts, &cs->rpkts, &cs->xmit_wait);
 		mod_timer(&cs->pma_timer,
-			  jiffies + usecs_to_jiffies(ibp->pma_sample_interval));
+		      jiffies + usecs_to_jiffies(ibp->rvp.pma_sample_interval));
 	} else if (cs->pma_sample_status == IB_PMA_SAMPLE_STATUS_RUNNING) {
 		u64 ta, tb, tc, td, te;
 
@@ -2975,11 +2975,11 @@ static void pma_6120_timer(unsigned long data)
 		cs->rpkts = td - cs->rpkts;
 		cs->xmit_wait = te - cs->xmit_wait;
 	}
-	spin_unlock_irqrestore(&ibp->lock, flags);
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 }
 
 /*
- * Note that the caller has the ibp->lock held.
+ * Note that the caller has the ibp->rvp.lock held.
  */
 static void qib_set_cntr_6120_sample(struct qib_pportdata *ppd, u32 intv,
 				     u32 start)
diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c
index 6c8ff10101c0..82d7c4bf5970 100644
--- a/drivers/infiniband/hw/qib/qib_iba7322.c
+++ b/drivers/infiniband/hw/qib/qib_iba7322.c
@@ -2910,8 +2910,6 @@ static void qib_setup_7322_cleanup(struct qib_devdata *dd)
 			spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
 			qib_qsfp_deinit(&dd->pport[i].cpspec->qsfp_data);
 		}
-		if (dd->pport[i].ibport_data.smi_ah)
-			ib_destroy_ah(&dd->pport[i].ibport_data.smi_ah->ibah);
 	}
 }
 
@@ -5497,7 +5495,7 @@ static void try_7322_ipg(struct qib_pportdata *ppd)
 	unsigned delay;
 	int ret;
 
-	agent = ibp->send_agent;
+	agent = ibp->rvp.send_agent;
 	if (!agent)
 		goto retry;
 
@@ -5515,7 +5513,7 @@ static void try_7322_ipg(struct qib_pportdata *ppd)
 			ret = PTR_ERR(ah);
 		else {
 			send_buf->ah = ah;
-			ibp->smi_ah = to_iah(ah);
+			ibp->smi_ah = ibah_to_rvtah(ah);
 			ret = 0;
 		}
 	} else {
diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c
index 4ff340fe904f..3f062f0dd9d8 100644
--- a/drivers/infiniband/hw/qib/qib_init.c
+++ b/drivers/infiniband/hw/qib/qib_init.c
@@ -42,6 +42,7 @@
 #ifdef CONFIG_INFINIBAND_QIB_DCA
 #include <linux/dca.h>
 #endif
+#include <rdma/rdma_vt.h>
 
 #include "qib.h"
 #include "qib_common.h"
@@ -244,6 +245,13 @@ int qib_init_pportdata(struct qib_pportdata *ppd, struct qib_devdata *dd,
 		alloc_percpu(struct qib_pma_counters);
 	if (!ppd->ibport_data.pmastats)
 		return -ENOMEM;
+	ppd->ibport_data.rvp.rc_acks = alloc_percpu(u64);
+	ppd->ibport_data.rvp.rc_qacks = alloc_percpu(u64);
+	ppd->ibport_data.rvp.rc_delayed_comp = alloc_percpu(u64);
+	if (!(ppd->ibport_data.rvp.rc_acks) ||
+	    !(ppd->ibport_data.rvp.rc_qacks) ||
+	    !(ppd->ibport_data.rvp.rc_delayed_comp))
+		return -ENOMEM;
 
 	if (qib_cc_table_size < IB_CCT_MIN_ENTRIES)
 		goto bail;
@@ -449,8 +457,6 @@ static int loadtime_init(struct qib_devdata *dd)
 	init_timer(&dd->intrchk_timer);
 	dd->intrchk_timer.function = verify_interrupt;
 	dd->intrchk_timer.data = (unsigned long) dd;
-
-	ret = qib_cq_init(dd);
 done:
 	return ret;
 }
@@ -631,6 +637,9 @@ wq_error:
 static void qib_free_pportdata(struct qib_pportdata *ppd)
 {
 	free_percpu(ppd->ibport_data.pmastats);
+	free_percpu(ppd->ibport_data.rvp.rc_acks);
+	free_percpu(ppd->ibport_data.rvp.rc_qacks);
+	free_percpu(ppd->ibport_data.rvp.rc_delayed_comp);
 	ppd->ibport_data.pmastats = NULL;
 }
 
@@ -1081,7 +1090,7 @@ void qib_free_devdata(struct qib_devdata *dd)
 	qib_dbg_ibdev_exit(&dd->verbs_dev);
 #endif
 	free_percpu(dd->int_counter);
-	ib_dealloc_device(&dd->verbs_dev.ibdev);
+	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
 }
 
 u64 qib_int_counter(struct qib_devdata *dd)
@@ -1120,9 +1129,12 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
 {
 	unsigned long flags;
 	struct qib_devdata *dd;
-	int ret;
+	int ret, nports;
 
-	dd = (struct qib_devdata *) ib_alloc_device(sizeof(*dd) + extra);
+	/* extra is * number of ports */
+	nports = extra / sizeof(struct qib_pportdata);
+	dd = (struct qib_devdata *)rvt_alloc_device(sizeof(*dd) + extra,
+						    nports);
 	if (!dd)
 		return ERR_PTR(-ENOMEM);
 
@@ -1171,7 +1183,7 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
 bail:
 	if (!list_empty(&dd->list))
 		list_del_init(&dd->list);
-	ib_dealloc_device(&dd->verbs_dev.ibdev);
+	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
 	return ERR_PTR(ret);
 }
 
@@ -1421,7 +1433,6 @@ static void cleanup_device_data(struct qib_devdata *dd)
 	}
 	kfree(tmp);
 	kfree(dd->boardname);
-	qib_cq_exit(dd);
 }
 
 /*
diff --git a/drivers/infiniband/hw/qib/qib_intr.c b/drivers/infiniband/hw/qib/qib_intr.c
index 086616d071b9..a014fd4cd076 100644
--- a/drivers/infiniband/hw/qib/qib_intr.c
+++ b/drivers/infiniband/hw/qib/qib_intr.c
@@ -74,7 +74,7 @@ static void signal_ib_event(struct qib_pportdata *ppd, enum ib_event_type ev)
 	struct ib_event event;
 	struct qib_devdata *dd = ppd->dd;
 
-	event.device = &dd->verbs_dev.ibdev;
+	event.device = &dd->verbs_dev.rdi.ibdev;
 	event.element.port_num = ppd->port;
 	event.event = ev;
 	ib_dispatch_event(&event);
diff --git a/drivers/infiniband/hw/qib/qib_keys.c b/drivers/infiniband/hw/qib/qib_keys.c
index d725c565518d..2c3c93572c17 100644
--- a/drivers/infiniband/hw/qib/qib_keys.c
+++ b/drivers/infiniband/hw/qib/qib_keys.c
@@ -46,20 +46,20 @@
  *
  */
 
-int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
+int qib_alloc_lkey(struct rvt_mregion *mr, int dma_region)
 {
 	unsigned long flags;
 	u32 r;
 	u32 n;
 	int ret = 0;
 	struct qib_ibdev *dev = to_idev(mr->pd->device);
-	struct qib_lkey_table *rkt = &dev->lk_table;
+	struct rvt_lkey_table *rkt = &dev->lk_table;
 
 	spin_lock_irqsave(&rkt->lock, flags);
 
 	/* special case for dma_mr lkey == 0 */
 	if (dma_region) {
-		struct qib_mregion *tmr;
+		struct rvt_mregion *tmr;
 
 		tmr = rcu_access_pointer(dev->dma_mr);
 		if (!tmr) {
@@ -90,8 +90,8 @@ int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
 	 * bits are capped in qib_verbs.c to insure enough bits
 	 * for generation number
 	 */
-	mr->lkey = (r << (32 - ib_qib_lkey_table_size)) |
-		((((1 << (24 - ib_qib_lkey_table_size)) - 1) & rkt->gen)
+	mr->lkey = (r << (32 - ib_rvt_lkey_table_size)) |
+		((((1 << (24 - ib_rvt_lkey_table_size)) - 1) & rkt->gen)
 		 << 8);
 	if (mr->lkey == 0) {
 		mr->lkey |= 1 << 8;
@@ -114,13 +114,13 @@ bail:
  * qib_free_lkey - free an lkey
  * @mr: mr to free from tables
  */
-void qib_free_lkey(struct qib_mregion *mr)
+void qib_free_lkey(struct rvt_mregion *mr)
 {
 	unsigned long flags;
 	u32 lkey = mr->lkey;
 	u32 r;
 	struct qib_ibdev *dev = to_idev(mr->pd->device);
-	struct qib_lkey_table *rkt = &dev->lk_table;
+	struct rvt_lkey_table *rkt = &dev->lk_table;
 
 	spin_lock_irqsave(&rkt->lock, flags);
 	if (!mr->lkey_published)
@@ -128,7 +128,7 @@ void qib_free_lkey(struct qib_mregion *mr)
 	if (lkey == 0)
 		RCU_INIT_POINTER(dev->dma_mr, NULL);
 	else {
-		r = lkey >> (32 - ib_qib_lkey_table_size);
+		r = lkey >> (32 - ib_rvt_lkey_table_size);
 		RCU_INIT_POINTER(rkt->table[r], NULL);
 	}
 	qib_put_mr(mr);
@@ -138,105 +138,6 @@ out:
 }
 
 /**
- * qib_lkey_ok - check IB SGE for validity and initialize
- * @rkt: table containing lkey to check SGE against
- * @pd: protection domain
- * @isge: outgoing internal SGE
- * @sge: SGE to check
- * @acc: access flags
- *
- * Return 1 if valid and successful, otherwise returns 0.
- *
- * increments the reference count upon success
- *
- * Check the IB SGE for validity and initialize our internal version
- * of it.
- */
-int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
-		struct qib_sge *isge, struct ib_sge *sge, int acc)
-{
-	struct qib_mregion *mr;
-	unsigned n, m;
-	size_t off;
-
-	/*
-	 * We use LKEY == zero for kernel virtual addresses
-	 * (see qib_get_dma_mr and qib_dma.c).
-	 */
-	rcu_read_lock();
-	if (sge->lkey == 0) {
-		struct qib_ibdev *dev = to_idev(pd->ibpd.device);
-
-		if (pd->user)
-			goto bail;
-		mr = rcu_dereference(dev->dma_mr);
-		if (!mr)
-			goto bail;
-		if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
-			goto bail;
-		rcu_read_unlock();
-
-		isge->mr = mr;
-		isge->vaddr = (void *) sge->addr;
-		isge->length = sge->length;
-		isge->sge_length = sge->length;
-		isge->m = 0;
-		isge->n = 0;
-		goto ok;
-	}
-	mr = rcu_dereference(
-		rkt->table[(sge->lkey >> (32 - ib_qib_lkey_table_size))]);
-	if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
-		goto bail;
-
-	off = sge->addr - mr->user_base;
-	if (unlikely(sge->addr < mr->user_base ||
-		     off + sge->length > mr->length ||
-		     (mr->access_flags & acc) != acc))
-		goto bail;
-	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
-		goto bail;
-	rcu_read_unlock();
-
-	off += mr->offset;
-	if (mr->page_shift) {
-		/*
-		page sizes are uniform power of 2 so no loop is necessary
-		entries_spanned_by_off is the number of times the loop below
-		would have executed.
-		*/
-		size_t entries_spanned_by_off;
-
-		entries_spanned_by_off = off >> mr->page_shift;
-		off -= (entries_spanned_by_off << mr->page_shift);
-		m = entries_spanned_by_off/QIB_SEGSZ;
-		n = entries_spanned_by_off%QIB_SEGSZ;
-	} else {
-		m = 0;
-		n = 0;
-		while (off >= mr->map[m]->segs[n].length) {
-			off -= mr->map[m]->segs[n].length;
-			n++;
-			if (n >= QIB_SEGSZ) {
-				m++;
-				n = 0;
-			}
-		}
-	}
-	isge->mr = mr;
-	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
-	isge->length = mr->map[m]->segs[n].length - off;
-	isge->sge_length = sge->length;
-	isge->m = m;
-	isge->n = n;
-ok:
-	return 1;
-bail:
-	rcu_read_unlock();
-	return 0;
-}
-
-/**
  * qib_rkey_ok - check the IB virtual address, length, and RKEY
  * @qp: qp for validation
  * @sge: SGE state
@@ -249,11 +150,11 @@ bail:
  *
  * increments the reference count upon success
  */
-int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
+int qib_rkey_ok(struct rvt_qp *qp, struct rvt_sge *sge,
 		u32 len, u64 vaddr, u32 rkey, int acc)
 {
-	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
-	struct qib_mregion *mr;
+	struct rvt_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
+	struct rvt_mregion *mr;
 	unsigned n, m;
 	size_t off;
 
@@ -263,7 +164,7 @@ int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
 	 */
 	rcu_read_lock();
 	if (rkey == 0) {
-		struct qib_pd *pd = to_ipd(qp->ibqp.pd);
+		struct rvt_pd *pd = ibpd_to_rvtpd(qp->ibqp.pd);
 		struct qib_ibdev *dev = to_idev(pd->ibpd.device);
 
 		if (pd->user)
@@ -285,7 +186,7 @@ int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
 	}
 
 	mr = rcu_dereference(
-		rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))]);
+		rkt->table[(rkey >> (32 - ib_rvt_lkey_table_size))]);
 	if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
 		goto bail;
 
@@ -308,15 +209,15 @@ int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
 
 		entries_spanned_by_off = off >> mr->page_shift;
 		off -= (entries_spanned_by_off << mr->page_shift);
-		m = entries_spanned_by_off/QIB_SEGSZ;
-		n = entries_spanned_by_off%QIB_SEGSZ;
+		m = entries_spanned_by_off / RVT_SEGSZ;
+		n = entries_spanned_by_off % RVT_SEGSZ;
 	} else {
 		m = 0;
 		n = 0;
 		while (off >= mr->map[m]->segs[n].length) {
 			off -= mr->map[m]->segs[n].length;
 			n++;
-			if (n >= QIB_SEGSZ) {
+			if (n >= RVT_SEGSZ) {
 				m++;
 				n = 0;
 			}
@@ -335,58 +236,3 @@ bail:
 	return 0;
 }
 
-/*
- * Initialize the memory region specified by the work request.
- */
-int qib_reg_mr(struct qib_qp *qp, struct ib_reg_wr *wr)
-{
-	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
-	struct qib_pd *pd = to_ipd(qp->ibqp.pd);
-	struct qib_mr *mr = to_imr(wr->mr);
-	struct qib_mregion *mrg;
-	u32 key = wr->key;
-	unsigned i, n, m;
-	int ret = -EINVAL;
-	unsigned long flags;
-	u64 *page_list;
-	size_t ps;
-
-	spin_lock_irqsave(&rkt->lock, flags);
-	if (pd->user || key == 0)
-		goto bail;
-
-	mrg = rcu_dereference_protected(
-		rkt->table[(key >> (32 - ib_qib_lkey_table_size))],
-		lockdep_is_held(&rkt->lock));
-	if (unlikely(mrg == NULL || qp->ibqp.pd != mrg->pd))
-		goto bail;
-
-	if (mr->npages > mrg->max_segs)
-		goto bail;
-
-	ps = mr->ibmr.page_size;
-	if (mr->ibmr.length > ps * mr->npages)
-		goto bail;
-
-	mrg->user_base = mr->ibmr.iova;
-	mrg->iova = mr->ibmr.iova;
-	mrg->lkey = key;
-	mrg->length = mr->ibmr.length;
-	mrg->access_flags = wr->access;
-	page_list = mr->pages;
-	m = 0;
-	n = 0;
-	for (i = 0; i < mr->npages; i++) {
-		mrg->map[m]->segs[n].vaddr = (void *) page_list[i];
-		mrg->map[m]->segs[n].length = ps;
-		if (++n == QIB_SEGSZ) {
-			m++;
-			n = 0;
-		}
-	}
-
-	ret = 0;
-bail:
-	spin_unlock_irqrestore(&rkt->lock, flags);
-	return ret;
-}
diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c
index 9625e7c438e5..0bd18375d7df 100644
--- a/drivers/infiniband/hw/qib/qib_mad.c
+++ b/drivers/infiniband/hw/qib/qib_mad.c
@@ -70,7 +70,7 @@ static void qib_send_trap(struct qib_ibport *ibp, void *data, unsigned len)
 	unsigned long flags;
 	unsigned long timeout;
 
-	agent = ibp->send_agent;
+	agent = ibp->rvp.send_agent;
 	if (!agent)
 		return;
 
@@ -79,7 +79,8 @@ static void qib_send_trap(struct qib_ibport *ibp, void *data, unsigned len)
 		return;
 
 	/* o14-2 */
-	if (ibp->trap_timeout && time_before(jiffies, ibp->trap_timeout))
+	if (ibp->rvp.trap_timeout &&
+	    time_before(jiffies, ibp->rvp.trap_timeout))
 		return;
 
 	send_buf = ib_create_send_mad(agent, 0, 0, 0, IB_MGMT_MAD_HDR,
@@ -93,42 +94,42 @@ static void qib_send_trap(struct qib_ibport *ibp, void *data, unsigned len)
 	smp->mgmt_class = IB_MGMT_CLASS_SUBN_LID_ROUTED;
 	smp->class_version = 1;
 	smp->method = IB_MGMT_METHOD_TRAP;
-	ibp->tid++;
-	smp->tid = cpu_to_be64(ibp->tid);
+	ibp->rvp.tid++;
+	smp->tid = cpu_to_be64(ibp->rvp.tid);
 	smp->attr_id = IB_SMP_ATTR_NOTICE;
 	/* o14-1: smp->mkey = 0; */
 	memcpy(smp->data, data, len);
 
-	spin_lock_irqsave(&ibp->lock, flags);
-	if (!ibp->sm_ah) {
-		if (ibp->sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
+	if (!ibp->rvp.sm_ah) {
+		if (ibp->rvp.sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
 			struct ib_ah *ah;
 
-			ah = qib_create_qp0_ah(ibp, ibp->sm_lid);
+			ah = qib_create_qp0_ah(ibp, ibp->rvp.sm_lid);
 			if (IS_ERR(ah))
 				ret = PTR_ERR(ah);
 			else {
 				send_buf->ah = ah;
-				ibp->sm_ah = to_iah(ah);
+				ibp->rvp.sm_ah = ibah_to_rvtah(ah);
 				ret = 0;
 			}
 		} else
 			ret = -EINVAL;
 	} else {
-		send_buf->ah = &ibp->sm_ah->ibah;
+		send_buf->ah = &ibp->rvp.sm_ah->ibah;
 		ret = 0;
 	}
-	spin_unlock_irqrestore(&ibp->lock, flags);
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 	if (!ret)
 		ret = ib_post_send_mad(send_buf, NULL);
 	if (!ret) {
 		/* 4.096 usec. */
-		timeout = (4096 * (1UL << ibp->subnet_timeout)) / 1000;
-		ibp->trap_timeout = jiffies + usecs_to_jiffies(timeout);
+		timeout = (4096 * (1UL << ibp->rvp.subnet_timeout)) / 1000;
+		ibp->rvp.trap_timeout = jiffies + usecs_to_jiffies(timeout);
 	} else {
 		ib_free_send_mad(send_buf);
-		ibp->trap_timeout = 0;
+		ibp->rvp.trap_timeout = 0;
 	}
 }
 
@@ -141,10 +142,10 @@ void qib_bad_pqkey(struct qib_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
 	struct ib_mad_notice_attr data;
 
 	if (trap_num == IB_NOTICE_TRAP_BAD_PKEY)
-		ibp->pkey_violations++;
+		ibp->rvp.pkey_violations++;
 	else
-		ibp->qkey_violations++;
-	ibp->n_pkt_drops++;
+		ibp->rvp.qkey_violations++;
+	ibp->rvp.n_pkt_drops++;
 
 	/* Send violation trap */
 	data.generic_type = IB_NOTICE_TYPE_SECURITY;
@@ -205,8 +206,11 @@ static void qib_bad_mkey(struct qib_ibport *ibp, struct ib_smp *smp)
 /*
  * Send a Port Capability Mask Changed trap (ch. 14.3.11).
  */
-void qib_cap_mask_chg(struct qib_ibport *ibp)
+void qib_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num)
 {
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = dd_from_dev(ibdev);
+	struct qib_ibport *ibp = &dd->pport[port_num - 1].ibport_data;
 	struct ib_mad_notice_attr data;
 
 	data.generic_type = IB_NOTICE_TYPE_INFO;
@@ -217,8 +221,8 @@ void qib_cap_mask_chg(struct qib_ibport *ibp)
 	data.toggle_count = 0;
 	memset(&data.details, 0, sizeof(data.details));
 	data.details.ntc_144.lid = data.issuer_lid;
-	data.details.ntc_144.new_cap_mask = cpu_to_be32(ibp->port_cap_flags);
-
+	data.details.ntc_144.new_cap_mask =
+					cpu_to_be32(ibp->rvp.port_cap_flags);
 	qib_send_trap(ibp, &data, sizeof(data));
 }
 
@@ -409,37 +413,38 @@ static int check_mkey(struct qib_ibport *ibp, struct ib_smp *smp, int mad_flags)
 	int ret = 0;
 
 	/* Is the mkey in the process of expiring? */
-	if (ibp->mkey_lease_timeout &&
-	    time_after_eq(jiffies, ibp->mkey_lease_timeout)) {
+	if (ibp->rvp.mkey_lease_timeout &&
+	    time_after_eq(jiffies, ibp->rvp.mkey_lease_timeout)) {
 		/* Clear timeout and mkey protection field. */
-		ibp->mkey_lease_timeout = 0;
-		ibp->mkeyprot = 0;
+		ibp->rvp.mkey_lease_timeout = 0;
+		ibp->rvp.mkeyprot = 0;
 	}
 
-	if ((mad_flags & IB_MAD_IGNORE_MKEY) ||  ibp->mkey == 0 ||
-	    ibp->mkey == smp->mkey)
+	if ((mad_flags & IB_MAD_IGNORE_MKEY) ||  ibp->rvp.mkey == 0 ||
+	    ibp->rvp.mkey == smp->mkey)
 		valid_mkey = 1;
 
 	/* Unset lease timeout on any valid Get/Set/TrapRepress */
-	if (valid_mkey && ibp->mkey_lease_timeout &&
+	if (valid_mkey && ibp->rvp.mkey_lease_timeout &&
 	    (smp->method == IB_MGMT_METHOD_GET ||
 	     smp->method == IB_MGMT_METHOD_SET ||
 	     smp->method == IB_MGMT_METHOD_TRAP_REPRESS))
-		ibp->mkey_lease_timeout = 0;
+		ibp->rvp.mkey_lease_timeout = 0;
 
 	if (!valid_mkey) {
 		switch (smp->method) {
 		case IB_MGMT_METHOD_GET:
 			/* Bad mkey not a violation below level 2 */
-			if (ibp->mkeyprot < 2)
+			if (ibp->rvp.mkeyprot < 2)
 				break;
 		case IB_MGMT_METHOD_SET:
 		case IB_MGMT_METHOD_TRAP_REPRESS:
-			if (ibp->mkey_violations != 0xFFFF)
-				++ibp->mkey_violations;
-			if (!ibp->mkey_lease_timeout && ibp->mkey_lease_period)
-				ibp->mkey_lease_timeout = jiffies +
-					ibp->mkey_lease_period * HZ;
+			if (ibp->rvp.mkey_violations != 0xFFFF)
+				++ibp->rvp.mkey_violations;
+			if (!ibp->rvp.mkey_lease_timeout &&
+			    ibp->rvp.mkey_lease_period)
+				ibp->rvp.mkey_lease_timeout = jiffies +
+					ibp->rvp.mkey_lease_period * HZ;
 			/* Generate a trap notice. */
 			qib_bad_mkey(ibp, smp);
 			ret = 1;
@@ -489,15 +494,15 @@ static int subn_get_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 
 	/* Only return the mkey if the protection field allows it. */
 	if (!(smp->method == IB_MGMT_METHOD_GET &&
-	      ibp->mkey != smp->mkey &&
-	      ibp->mkeyprot == 1))
-		pip->mkey = ibp->mkey;
-	pip->gid_prefix = ibp->gid_prefix;
+	      ibp->rvp.mkey != smp->mkey &&
+	      ibp->rvp.mkeyprot == 1))
+		pip->mkey = ibp->rvp.mkey;
+	pip->gid_prefix = ibp->rvp.gid_prefix;
 	pip->lid = cpu_to_be16(ppd->lid);
-	pip->sm_lid = cpu_to_be16(ibp->sm_lid);
-	pip->cap_mask = cpu_to_be32(ibp->port_cap_flags);
+	pip->sm_lid = cpu_to_be16(ibp->rvp.sm_lid);
+	pip->cap_mask = cpu_to_be32(ibp->rvp.port_cap_flags);
 	/* pip->diag_code; */
-	pip->mkey_lease_period = cpu_to_be16(ibp->mkey_lease_period);
+	pip->mkey_lease_period = cpu_to_be16(ibp->rvp.mkey_lease_period);
 	pip->local_port_num = port;
 	pip->link_width_enabled = ppd->link_width_enabled;
 	pip->link_width_supported = ppd->link_width_supported;
@@ -508,7 +513,7 @@ static int subn_get_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	pip->portphysstate_linkdown =
 		(dd->f_ibphys_portstate(ppd->lastibcstat) << 4) |
 		(get_linkdowndefaultstate(ppd) ? 1 : 2);
-	pip->mkeyprot_resv_lmc = (ibp->mkeyprot << 6) | ppd->lmc;
+	pip->mkeyprot_resv_lmc = (ibp->rvp.mkeyprot << 6) | ppd->lmc;
 	pip->linkspeedactive_enabled = (ppd->link_speed_active << 4) |
 		ppd->link_speed_enabled;
 	switch (ppd->ibmtu) {
@@ -529,9 +534,9 @@ static int subn_get_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 		mtu = IB_MTU_256;
 		break;
 	}
-	pip->neighbormtu_mastersmsl = (mtu << 4) | ibp->sm_sl;
+	pip->neighbormtu_mastersmsl = (mtu << 4) | ibp->rvp.sm_sl;
 	pip->vlcap_inittype = ppd->vls_supported << 4;  /* InitType = 0 */
-	pip->vl_high_limit = ibp->vl_high_limit;
+	pip->vl_high_limit = ibp->rvp.vl_high_limit;
 	pip->vl_arb_high_cap =
 		dd->f_get_ib_cfg(ppd, QIB_IB_CFG_VL_HIGH_CAP);
 	pip->vl_arb_low_cap =
@@ -542,20 +547,20 @@ static int subn_get_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	/* pip->vlstallcnt_hoqlife; */
 	pip->operationalvl_pei_peo_fpi_fpo =
 		dd->f_get_ib_cfg(ppd, QIB_IB_CFG_OP_VLS) << 4;
-	pip->mkey_violations = cpu_to_be16(ibp->mkey_violations);
+	pip->mkey_violations = cpu_to_be16(ibp->rvp.mkey_violations);
 	/* P_KeyViolations are counted by hardware. */
-	pip->pkey_violations = cpu_to_be16(ibp->pkey_violations);
-	pip->qkey_violations = cpu_to_be16(ibp->qkey_violations);
+	pip->pkey_violations = cpu_to_be16(ibp->rvp.pkey_violations);
+	pip->qkey_violations = cpu_to_be16(ibp->rvp.qkey_violations);
 	/* Only the hardware GUID is supported for now */
 	pip->guid_cap = QIB_GUIDS_PER_PORT;
-	pip->clientrereg_resv_subnetto = ibp->subnet_timeout;
+	pip->clientrereg_resv_subnetto = ibp->rvp.subnet_timeout;
 	/* 32.768 usec. response time (guessing) */
 	pip->resv_resptimevalue = 3;
 	pip->localphyerrors_overrunerrors =
 		(get_phyerrthreshold(ppd) << 4) |
 		get_overrunthreshold(ppd);
 	/* pip->max_credit_hint; */
-	if (ibp->port_cap_flags & IB_PORT_LINK_LATENCY_SUP) {
+	if (ibp->rvp.port_cap_flags & IB_PORT_LINK_LATENCY_SUP) {
 		u32 v;
 
 		v = dd->f_get_ib_cfg(ppd, QIB_IB_CFG_LINKLATENCY);
@@ -685,13 +690,13 @@ static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	event.device = ibdev;
 	event.element.port_num = port;
 
-	ibp->mkey = pip->mkey;
-	ibp->gid_prefix = pip->gid_prefix;
-	ibp->mkey_lease_period = be16_to_cpu(pip->mkey_lease_period);
+	ibp->rvp.mkey = pip->mkey;
+	ibp->rvp.gid_prefix = pip->gid_prefix;
+	ibp->rvp.mkey_lease_period = be16_to_cpu(pip->mkey_lease_period);
 
 	lid = be16_to_cpu(pip->lid);
 	/* Must be a valid unicast LID address. */
-	if (lid == 0 || lid >= QIB_MULTICAST_LID_BASE)
+	if (lid == 0 || lid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
 		smp->status |= IB_SMP_INVALID_FIELD;
 	else if (ppd->lid != lid || ppd->lmc != (pip->mkeyprot_resv_lmc & 7)) {
 		if (ppd->lid != lid)
@@ -706,21 +711,21 @@ static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	smlid = be16_to_cpu(pip->sm_lid);
 	msl = pip->neighbormtu_mastersmsl & 0xF;
 	/* Must be a valid unicast LID address. */
-	if (smlid == 0 || smlid >= QIB_MULTICAST_LID_BASE)
+	if (smlid == 0 || smlid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
 		smp->status |= IB_SMP_INVALID_FIELD;
-	else if (smlid != ibp->sm_lid || msl != ibp->sm_sl) {
-		spin_lock_irqsave(&ibp->lock, flags);
-		if (ibp->sm_ah) {
-			if (smlid != ibp->sm_lid)
-				ibp->sm_ah->attr.dlid = smlid;
-			if (msl != ibp->sm_sl)
-				ibp->sm_ah->attr.sl = msl;
+	else if (smlid != ibp->rvp.sm_lid || msl != ibp->rvp.sm_sl) {
+		spin_lock_irqsave(&ibp->rvp.lock, flags);
+		if (ibp->rvp.sm_ah) {
+			if (smlid != ibp->rvp.sm_lid)
+				ibp->rvp.sm_ah->attr.dlid = smlid;
+			if (msl != ibp->rvp.sm_sl)
+				ibp->rvp.sm_ah->attr.sl = msl;
 		}
-		spin_unlock_irqrestore(&ibp->lock, flags);
-		if (smlid != ibp->sm_lid)
-			ibp->sm_lid = smlid;
-		if (msl != ibp->sm_sl)
-			ibp->sm_sl = msl;
+		spin_unlock_irqrestore(&ibp->rvp.lock, flags);
+		if (smlid != ibp->rvp.sm_lid)
+			ibp->rvp.sm_lid = smlid;
+		if (msl != ibp->rvp.sm_sl)
+			ibp->rvp.sm_sl = msl;
 		event.event = IB_EVENT_SM_CHANGE;
 		ib_dispatch_event(&event);
 	}
@@ -768,10 +773,10 @@ static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 		smp->status |= IB_SMP_INVALID_FIELD;
 	}
 
-	ibp->mkeyprot = pip->mkeyprot_resv_lmc >> 6;
-	ibp->vl_high_limit = pip->vl_high_limit;
+	ibp->rvp.mkeyprot = pip->mkeyprot_resv_lmc >> 6;
+	ibp->rvp.vl_high_limit = pip->vl_high_limit;
 	(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_VL_HIGH_LIMIT,
-				    ibp->vl_high_limit);
+				    ibp->rvp.vl_high_limit);
 
 	mtu = ib_mtu_enum_to_int((pip->neighbormtu_mastersmsl >> 4) & 0xF);
 	if (mtu == -1)
@@ -789,13 +794,13 @@ static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	}
 
 	if (pip->mkey_violations == 0)
-		ibp->mkey_violations = 0;
+		ibp->rvp.mkey_violations = 0;
 
 	if (pip->pkey_violations == 0)
-		ibp->pkey_violations = 0;
+		ibp->rvp.pkey_violations = 0;
 
 	if (pip->qkey_violations == 0)
-		ibp->qkey_violations = 0;
+		ibp->rvp.qkey_violations = 0;
 
 	ore = pip->localphyerrors_overrunerrors;
 	if (set_phyerrthreshold(ppd, (ore >> 4) & 0xF))
@@ -804,7 +809,7 @@ static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
 	if (set_overrunthreshold(ppd, (ore & 0xF)))
 		smp->status |= IB_SMP_INVALID_FIELD;
 
-	ibp->subnet_timeout = pip->clientrereg_resv_subnetto & 0x1F;
+	ibp->rvp.subnet_timeout = pip->clientrereg_resv_subnetto & 0x1F;
 
 	/*
 	 * Do the port state change now that the other link parameters
@@ -1028,7 +1033,7 @@ static int set_pkeys(struct qib_devdata *dd, u8 port, u16 *pkeys)
 		(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PKEYS, 0);
 
 		event.event = IB_EVENT_PKEY_CHANGE;
-		event.device = &dd->verbs_dev.ibdev;
+		event.device = &dd->verbs_dev.rdi.ibdev;
 		event.element.port_num = port;
 		ib_dispatch_event(&event);
 	}
@@ -1062,7 +1067,7 @@ static int subn_get_sl_to_vl(struct ib_smp *smp, struct ib_device *ibdev,
 
 	memset(smp->data, 0, sizeof(smp->data));
 
-	if (!(ibp->port_cap_flags & IB_PORT_SL_MAP_SUP))
+	if (!(ibp->rvp.port_cap_flags & IB_PORT_SL_MAP_SUP))
 		smp->status |= IB_SMP_UNSUP_METHOD;
 	else
 		for (i = 0; i < ARRAY_SIZE(ibp->sl_to_vl); i += 2)
@@ -1078,7 +1083,7 @@ static int subn_set_sl_to_vl(struct ib_smp *smp, struct ib_device *ibdev,
 	u8 *p = (u8 *) smp->data;
 	unsigned i;
 
-	if (!(ibp->port_cap_flags & IB_PORT_SL_MAP_SUP)) {
+	if (!(ibp->rvp.port_cap_flags & IB_PORT_SL_MAP_SUP)) {
 		smp->status |= IB_SMP_UNSUP_METHOD;
 		return reply(smp);
 	}
@@ -1195,20 +1200,20 @@ static int pma_get_portsamplescontrol(struct ib_pma_mad *pmp,
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		goto bail;
 	}
-	spin_lock_irqsave(&ibp->lock, flags);
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
 	p->tick = dd->f_get_ib_cfg(ppd, QIB_IB_CFG_PMA_TICKS);
 	p->sample_status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
 	p->counter_width = 4;   /* 32 bit counters */
 	p->counter_mask0_9 = COUNTER_MASK0_9;
-	p->sample_start = cpu_to_be32(ibp->pma_sample_start);
-	p->sample_interval = cpu_to_be32(ibp->pma_sample_interval);
-	p->tag = cpu_to_be16(ibp->pma_tag);
-	p->counter_select[0] = ibp->pma_counter_select[0];
-	p->counter_select[1] = ibp->pma_counter_select[1];
-	p->counter_select[2] = ibp->pma_counter_select[2];
-	p->counter_select[3] = ibp->pma_counter_select[3];
-	p->counter_select[4] = ibp->pma_counter_select[4];
-	spin_unlock_irqrestore(&ibp->lock, flags);
+	p->sample_start = cpu_to_be32(ibp->rvp.pma_sample_start);
+	p->sample_interval = cpu_to_be32(ibp->rvp.pma_sample_interval);
+	p->tag = cpu_to_be16(ibp->rvp.pma_tag);
+	p->counter_select[0] = ibp->rvp.pma_counter_select[0];
+	p->counter_select[1] = ibp->rvp.pma_counter_select[1];
+	p->counter_select[2] = ibp->rvp.pma_counter_select[2];
+	p->counter_select[3] = ibp->rvp.pma_counter_select[3];
+	p->counter_select[4] = ibp->rvp.pma_counter_select[4];
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 bail:
 	return reply((struct ib_smp *) pmp);
@@ -1233,7 +1238,7 @@ static int pma_set_portsamplescontrol(struct ib_pma_mad *pmp,
 		goto bail;
 	}
 
-	spin_lock_irqsave(&ibp->lock, flags);
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
 
 	/* Port Sampling code owns the PS* HW counters */
 	xmit_flags = ppd->cong_stats.flags;
@@ -1242,18 +1247,18 @@ static int pma_set_portsamplescontrol(struct ib_pma_mad *pmp,
 	if (status == IB_PMA_SAMPLE_STATUS_DONE ||
 	    (status == IB_PMA_SAMPLE_STATUS_RUNNING &&
 	     xmit_flags == IB_PMA_CONG_HW_CONTROL_TIMER)) {
-		ibp->pma_sample_start = be32_to_cpu(p->sample_start);
-		ibp->pma_sample_interval = be32_to_cpu(p->sample_interval);
-		ibp->pma_tag = be16_to_cpu(p->tag);
-		ibp->pma_counter_select[0] = p->counter_select[0];
-		ibp->pma_counter_select[1] = p->counter_select[1];
-		ibp->pma_counter_select[2] = p->counter_select[2];
-		ibp->pma_counter_select[3] = p->counter_select[3];
-		ibp->pma_counter_select[4] = p->counter_select[4];
-		dd->f_set_cntr_sample(ppd, ibp->pma_sample_interval,
-				      ibp->pma_sample_start);
+		ibp->rvp.pma_sample_start = be32_to_cpu(p->sample_start);
+		ibp->rvp.pma_sample_interval = be32_to_cpu(p->sample_interval);
+		ibp->rvp.pma_tag = be16_to_cpu(p->tag);
+		ibp->rvp.pma_counter_select[0] = p->counter_select[0];
+		ibp->rvp.pma_counter_select[1] = p->counter_select[1];
+		ibp->rvp.pma_counter_select[2] = p->counter_select[2];
+		ibp->rvp.pma_counter_select[3] = p->counter_select[3];
+		ibp->rvp.pma_counter_select[4] = p->counter_select[4];
+		dd->f_set_cntr_sample(ppd, ibp->rvp.pma_sample_interval,
+				      ibp->rvp.pma_sample_start);
 	}
-	spin_unlock_irqrestore(&ibp->lock, flags);
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 	ret = pma_get_portsamplescontrol(pmp, ibdev, port);
 
@@ -1357,8 +1362,8 @@ static int pma_get_portsamplesresult(struct ib_pma_mad *pmp,
 	int i;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
-	spin_lock_irqsave(&ibp->lock, flags);
-	p->tag = cpu_to_be16(ibp->pma_tag);
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
+	p->tag = cpu_to_be16(ibp->rvp.pma_tag);
 	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_TIMER)
 		p->sample_status = IB_PMA_SAMPLE_STATUS_DONE;
 	else {
@@ -1373,11 +1378,11 @@ static int pma_get_portsamplesresult(struct ib_pma_mad *pmp,
 			ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_TIMER;
 		}
 	}
-	for (i = 0; i < ARRAY_SIZE(ibp->pma_counter_select); i++)
+	for (i = 0; i < ARRAY_SIZE(ibp->rvp.pma_counter_select); i++)
 		p->counter[i] = cpu_to_be32(
 			get_cache_hw_sample_counters(
-				ppd, ibp->pma_counter_select[i]));
-	spin_unlock_irqrestore(&ibp->lock, flags);
+				ppd, ibp->rvp.pma_counter_select[i]));
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 	return reply((struct ib_smp *) pmp);
 }
@@ -1397,8 +1402,8 @@ static int pma_get_portsamplesresult_ext(struct ib_pma_mad *pmp,
 
 	/* Port Sampling code owns the PS* HW counters */
 	memset(pmp->data, 0, sizeof(pmp->data));
-	spin_lock_irqsave(&ibp->lock, flags);
-	p->tag = cpu_to_be16(ibp->pma_tag);
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
+	p->tag = cpu_to_be16(ibp->rvp.pma_tag);
 	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_TIMER)
 		p->sample_status = IB_PMA_SAMPLE_STATUS_DONE;
 	else {
@@ -1415,11 +1420,11 @@ static int pma_get_portsamplesresult_ext(struct ib_pma_mad *pmp,
 			ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_TIMER;
 		}
 	}
-	for (i = 0; i < ARRAY_SIZE(ibp->pma_counter_select); i++)
+	for (i = 0; i < ARRAY_SIZE(ibp->rvp.pma_counter_select); i++)
 		p->counter[i] = cpu_to_be64(
 			get_cache_hw_sample_counters(
-				ppd, ibp->pma_counter_select[i]));
-	spin_unlock_irqrestore(&ibp->lock, flags);
+				ppd, ibp->rvp.pma_counter_select[i]));
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 	return reply((struct ib_smp *) pmp);
 }
@@ -1453,7 +1458,7 @@ static int pma_get_portcounters(struct ib_pma_mad *pmp,
 	cntrs.excessive_buffer_overrun_errors -=
 		ibp->z_excessive_buffer_overrun_errors;
 	cntrs.vl15_dropped -= ibp->z_vl15_dropped;
-	cntrs.vl15_dropped += ibp->n_vl15_dropped;
+	cntrs.vl15_dropped += ibp->rvp.n_vl15_dropped;
 
 	memset(pmp->data, 0, sizeof(pmp->data));
 
@@ -1546,9 +1551,9 @@ static int pma_get_portcounters_cong(struct ib_pma_mad *pmp,
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 
 	qib_get_counters(ppd, &cntrs);
-	spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+	spin_lock_irqsave(&ppd->ibport_data.rvp.lock, flags);
 	xmit_wait_counter = xmit_wait_get_value_delta(ppd);
-	spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+	spin_unlock_irqrestore(&ppd->ibport_data.rvp.lock, flags);
 
 	/* Adjust counters for any resets done. */
 	cntrs.symbol_error_counter -= ibp->z_symbol_error_counter;
@@ -1564,7 +1569,7 @@ static int pma_get_portcounters_cong(struct ib_pma_mad *pmp,
 	cntrs.excessive_buffer_overrun_errors -=
 		ibp->z_excessive_buffer_overrun_errors;
 	cntrs.vl15_dropped -= ibp->z_vl15_dropped;
-	cntrs.vl15_dropped += ibp->n_vl15_dropped;
+	cntrs.vl15_dropped += ibp->rvp.n_vl15_dropped;
 	cntrs.port_xmit_data -= ibp->z_port_xmit_data;
 	cntrs.port_rcv_data -= ibp->z_port_rcv_data;
 	cntrs.port_xmit_packets -= ibp->z_port_xmit_packets;
@@ -1743,7 +1748,7 @@ static int pma_set_portcounters(struct ib_pma_mad *pmp,
 			cntrs.excessive_buffer_overrun_errors;
 
 	if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED) {
-		ibp->n_vl15_dropped = 0;
+		ibp->rvp.n_vl15_dropped = 0;
 		ibp->z_vl15_dropped = cntrs.vl15_dropped;
 	}
 
@@ -1778,11 +1783,11 @@ static int pma_set_portcounters_cong(struct ib_pma_mad *pmp,
 	ret = pma_get_portcounters_cong(pmp, ibdev, port);
 
 	if (counter_select & IB_PMA_SEL_CONG_XMIT) {
-		spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+		spin_lock_irqsave(&ppd->ibport_data.rvp.lock, flags);
 		ppd->cong_stats.counter = 0;
 		dd->f_set_cntr_sample(ppd, QIB_CONG_TIMER_PSINTERVAL,
 				      0x0);
-		spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+		spin_unlock_irqrestore(&ppd->ibport_data.rvp.lock, flags);
 	}
 	if (counter_select & IB_PMA_SEL_CONG_PORT_DATA) {
 		ibp->z_port_xmit_data = cntrs.port_xmit_data;
@@ -1806,7 +1811,7 @@ static int pma_set_portcounters_cong(struct ib_pma_mad *pmp,
 			cntrs.local_link_integrity_errors;
 		ibp->z_excessive_buffer_overrun_errors =
 			cntrs.excessive_buffer_overrun_errors;
-		ibp->n_vl15_dropped = 0;
+		ibp->rvp.n_vl15_dropped = 0;
 		ibp->z_vl15_dropped = cntrs.vl15_dropped;
 	}
 
@@ -1916,12 +1921,12 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 			ret = subn_get_vl_arb(smp, ibdev, port);
 			goto bail;
 		case IB_SMP_ATTR_SM_INFO:
-			if (ibp->port_cap_flags & IB_PORT_SM_DISABLED) {
+			if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED) {
 				ret = IB_MAD_RESULT_SUCCESS |
 					IB_MAD_RESULT_CONSUMED;
 				goto bail;
 			}
-			if (ibp->port_cap_flags & IB_PORT_SM) {
+			if (ibp->rvp.port_cap_flags & IB_PORT_SM) {
 				ret = IB_MAD_RESULT_SUCCESS;
 				goto bail;
 			}
@@ -1950,12 +1955,12 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 			ret = subn_set_vl_arb(smp, ibdev, port);
 			goto bail;
 		case IB_SMP_ATTR_SM_INFO:
-			if (ibp->port_cap_flags & IB_PORT_SM_DISABLED) {
+			if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED) {
 				ret = IB_MAD_RESULT_SUCCESS |
 					IB_MAD_RESULT_CONSUMED;
 				goto bail;
 			}
-			if (ibp->port_cap_flags & IB_PORT_SM) {
+			if (ibp->rvp.port_cap_flags & IB_PORT_SM) {
 				ret = IB_MAD_RESULT_SUCCESS;
 				goto bail;
 			}
@@ -2443,12 +2448,6 @@ bail:
 	return ret;
 }
 
-static void send_handler(struct ib_mad_agent *agent,
-			 struct ib_mad_send_wc *mad_send_wc)
-{
-	ib_free_send_mad(mad_send_wc->send_buf);
-}
-
 static void xmit_wait_timer_func(unsigned long opaque)
 {
 	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
@@ -2456,7 +2455,7 @@ static void xmit_wait_timer_func(unsigned long opaque)
 	unsigned long flags;
 	u8 status;
 
-	spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+	spin_lock_irqsave(&ppd->ibport_data.rvp.lock, flags);
 	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_SAMPLE) {
 		status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
 		if (status == IB_PMA_SAMPLE_STATUS_DONE) {
@@ -2469,74 +2468,35 @@ static void xmit_wait_timer_func(unsigned long opaque)
 	ppd->cong_stats.counter = xmit_wait_get_value_delta(ppd);
 	dd->f_set_cntr_sample(ppd, QIB_CONG_TIMER_PSINTERVAL, 0x0);
 done:
-	spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+	spin_unlock_irqrestore(&ppd->ibport_data.rvp.lock, flags);
 	mod_timer(&ppd->cong_stats.timer, jiffies + HZ);
 }
 
-int qib_create_agents(struct qib_ibdev *dev)
+void qib_notify_create_mad_agent(struct rvt_dev_info *rdi, int port_idx)
 {
-	struct qib_devdata *dd = dd_from_dev(dev);
-	struct ib_mad_agent *agent;
-	struct qib_ibport *ibp;
-	int p;
-	int ret;
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(ibdev,
+					      struct qib_devdata, verbs_dev);
 
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI,
-					      NULL, 0, send_handler,
-					      NULL, NULL, 0);
-		if (IS_ERR(agent)) {
-			ret = PTR_ERR(agent);
-			goto err;
-		}
-
-		/* Initialize xmit_wait structure */
-		dd->pport[p].cong_stats.counter = 0;
-		init_timer(&dd->pport[p].cong_stats.timer);
-		dd->pport[p].cong_stats.timer.function = xmit_wait_timer_func;
-		dd->pport[p].cong_stats.timer.data =
-			(unsigned long)(&dd->pport[p]);
-		dd->pport[p].cong_stats.timer.expires = 0;
-		add_timer(&dd->pport[p].cong_stats.timer);
-
-		ibp->send_agent = agent;
-	}
-
-	return 0;
-
-err:
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		if (ibp->send_agent) {
-			agent = ibp->send_agent;
-			ibp->send_agent = NULL;
-			ib_unregister_mad_agent(agent);
-		}
-	}
-
-	return ret;
+	/* Initialize xmit_wait structure */
+	dd->pport[port_idx].cong_stats.counter = 0;
+	init_timer(&dd->pport[port_idx].cong_stats.timer);
+	dd->pport[port_idx].cong_stats.timer.function = xmit_wait_timer_func;
+	dd->pport[port_idx].cong_stats.timer.data =
+		(unsigned long)(&dd->pport[port_idx]);
+	dd->pport[port_idx].cong_stats.timer.expires = 0;
+	add_timer(&dd->pport[port_idx].cong_stats.timer);
 }
 
-void qib_free_agents(struct qib_ibdev *dev)
+void qib_notify_free_mad_agent(struct rvt_dev_info *rdi, int port_idx)
 {
-	struct qib_devdata *dd = dd_from_dev(dev);
-	struct ib_mad_agent *agent;
-	struct qib_ibport *ibp;
-	int p;
-
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		if (ibp->send_agent) {
-			agent = ibp->send_agent;
-			ibp->send_agent = NULL;
-			ib_unregister_mad_agent(agent);
-		}
-		if (ibp->sm_ah) {
-			ib_destroy_ah(&ibp->sm_ah->ibah);
-			ibp->sm_ah = NULL;
-		}
-		if (dd->pport[p].cong_stats.timer.data)
-			del_timer_sync(&dd->pport[p].cong_stats.timer);
-	}
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(ibdev,
+					      struct qib_devdata, verbs_dev);
+
+	if (dd->pport[port_idx].cong_stats.timer.data)
+		del_timer_sync(&dd->pport[port_idx].cong_stats.timer);
+
+	if (dd->pport[port_idx].ibport_data.smi_ah)
+		ib_destroy_ah(&dd->pport[port_idx].ibport_data.smi_ah->ibah);
 }
diff --git a/drivers/infiniband/hw/qib/qib_mmap.c b/drivers/infiniband/hw/qib/qib_mmap.c
deleted file mode 100644
index 34927b700b0e..000000000000
--- a/drivers/infiniband/hw/qib/qib_mmap.c
+++ /dev/null
@@ -1,174 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <linux/module.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-#include <linux/mm.h>
-#include <linux/errno.h>
-#include <asm/pgtable.h>
-
-#include "qib_verbs.h"
-
-/**
- * qib_release_mmap_info - free mmap info structure
- * @ref: a pointer to the kref within struct qib_mmap_info
- */
-void qib_release_mmap_info(struct kref *ref)
-{
-	struct qib_mmap_info *ip =
-		container_of(ref, struct qib_mmap_info, ref);
-	struct qib_ibdev *dev = to_idev(ip->context->device);
-
-	spin_lock_irq(&dev->pending_lock);
-	list_del(&ip->pending_mmaps);
-	spin_unlock_irq(&dev->pending_lock);
-
-	vfree(ip->obj);
-	kfree(ip);
-}
-
-/*
- * open and close keep track of how many times the CQ is mapped,
- * to avoid releasing it.
- */
-static void qib_vma_open(struct vm_area_struct *vma)
-{
-	struct qib_mmap_info *ip = vma->vm_private_data;
-
-	kref_get(&ip->ref);
-}
-
-static void qib_vma_close(struct vm_area_struct *vma)
-{
-	struct qib_mmap_info *ip = vma->vm_private_data;
-
-	kref_put(&ip->ref, qib_release_mmap_info);
-}
-
-static const struct vm_operations_struct qib_vm_ops = {
-	.open =     qib_vma_open,
-	.close =    qib_vma_close,
-};
-
-/**
- * qib_mmap - create a new mmap region
- * @context: the IB user context of the process making the mmap() call
- * @vma: the VMA to be initialized
- * Return zero if the mmap is OK. Otherwise, return an errno.
- */
-int qib_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
-{
-	struct qib_ibdev *dev = to_idev(context->device);
-	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
-	unsigned long size = vma->vm_end - vma->vm_start;
-	struct qib_mmap_info *ip, *pp;
-	int ret = -EINVAL;
-
-	/*
-	 * Search the device's list of objects waiting for a mmap call.
-	 * Normally, this list is very short since a call to create a
-	 * CQ, QP, or SRQ is soon followed by a call to mmap().
-	 */
-	spin_lock_irq(&dev->pending_lock);
-	list_for_each_entry_safe(ip, pp, &dev->pending_mmaps,
-				 pending_mmaps) {
-		/* Only the creator is allowed to mmap the object */
-		if (context != ip->context || (__u64) offset != ip->offset)
-			continue;
-		/* Don't allow a mmap larger than the object. */
-		if (size > ip->size)
-			break;
-
-		list_del_init(&ip->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-
-		ret = remap_vmalloc_range(vma, ip->obj, 0);
-		if (ret)
-			goto done;
-		vma->vm_ops = &qib_vm_ops;
-		vma->vm_private_data = ip;
-		qib_vma_open(vma);
-		goto done;
-	}
-	spin_unlock_irq(&dev->pending_lock);
-done:
-	return ret;
-}
-
-/*
- * Allocate information for qib_mmap
- */
-struct qib_mmap_info *qib_create_mmap_info(struct qib_ibdev *dev,
-					   u32 size,
-					   struct ib_ucontext *context,
-					   void *obj) {
-	struct qib_mmap_info *ip;
-
-	ip = kmalloc(sizeof(*ip), GFP_KERNEL);
-	if (!ip)
-		goto bail;
-
-	size = PAGE_ALIGN(size);
-
-	spin_lock_irq(&dev->mmap_offset_lock);
-	if (dev->mmap_offset == 0)
-		dev->mmap_offset = PAGE_SIZE;
-	ip->offset = dev->mmap_offset;
-	dev->mmap_offset += size;
-	spin_unlock_irq(&dev->mmap_offset_lock);
-
-	INIT_LIST_HEAD(&ip->pending_mmaps);
-	ip->size = size;
-	ip->context = context;
-	ip->obj = obj;
-	kref_init(&ip->ref);
-
-bail:
-	return ip;
-}
-
-void qib_update_mmap_info(struct qib_ibdev *dev, struct qib_mmap_info *ip,
-			  u32 size, void *obj)
-{
-	size = PAGE_ALIGN(size);
-
-	spin_lock_irq(&dev->mmap_offset_lock);
-	if (dev->mmap_offset == 0)
-		dev->mmap_offset = PAGE_SIZE;
-	ip->offset = dev->mmap_offset;
-	dev->mmap_offset += size;
-	spin_unlock_irq(&dev->mmap_offset_lock);
-
-	ip->size = size;
-	ip->obj = obj;
-}
diff --git a/drivers/infiniband/hw/qib/qib_mr.c b/drivers/infiniband/hw/qib/qib_mr.c
deleted file mode 100644
index 5f53304e8a9b..000000000000
--- a/drivers/infiniband/hw/qib/qib_mr.c
+++ /dev/null
@@ -1,490 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
- * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <rdma/ib_umem.h>
-#include <rdma/ib_smi.h>
-
-#include "qib.h"
-
-/* Fast memory region */
-struct qib_fmr {
-	struct ib_fmr ibfmr;
-	struct qib_mregion mr;        /* must be last */
-};
-
-static inline struct qib_fmr *to_ifmr(struct ib_fmr *ibfmr)
-{
-	return container_of(ibfmr, struct qib_fmr, ibfmr);
-}
-
-static int init_qib_mregion(struct qib_mregion *mr, struct ib_pd *pd,
-	int count)
-{
-	int m, i = 0;
-	int rval = 0;
-
-	m = (count + QIB_SEGSZ - 1) / QIB_SEGSZ;
-	for (; i < m; i++) {
-		mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL);
-		if (!mr->map[i])
-			goto bail;
-	}
-	mr->mapsz = m;
-	init_completion(&mr->comp);
-	/* count returning the ptr to user */
-	atomic_set(&mr->refcount, 1);
-	mr->pd = pd;
-	mr->max_segs = count;
-out:
-	return rval;
-bail:
-	while (i)
-		kfree(mr->map[--i]);
-	rval = -ENOMEM;
-	goto out;
-}
-
-static void deinit_qib_mregion(struct qib_mregion *mr)
-{
-	int i = mr->mapsz;
-
-	mr->mapsz = 0;
-	while (i)
-		kfree(mr->map[--i]);
-}
-
-
-/**
- * qib_get_dma_mr - get a DMA memory region
- * @pd: protection domain for this memory region
- * @acc: access flags
- *
- * Returns the memory region on success, otherwise returns an errno.
- * Note that all DMA addresses should be created via the
- * struct ib_dma_mapping_ops functions (see qib_dma.c).
- */
-struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc)
-{
-	struct qib_mr *mr = NULL;
-	struct ib_mr *ret;
-	int rval;
-
-	if (to_ipd(pd)->user) {
-		ret = ERR_PTR(-EPERM);
-		goto bail;
-	}
-
-	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
-	if (!mr) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	rval = init_qib_mregion(&mr->mr, pd, 0);
-	if (rval) {
-		ret = ERR_PTR(rval);
-		goto bail;
-	}
-
-
-	rval = qib_alloc_lkey(&mr->mr, 1);
-	if (rval) {
-		ret = ERR_PTR(rval);
-		goto bail_mregion;
-	}
-
-	mr->mr.access_flags = acc;
-	ret = &mr->ibmr;
-done:
-	return ret;
-
-bail_mregion:
-	deinit_qib_mregion(&mr->mr);
-bail:
-	kfree(mr);
-	goto done;
-}
-
-static struct qib_mr *alloc_mr(int count, struct ib_pd *pd)
-{
-	struct qib_mr *mr;
-	int rval = -ENOMEM;
-	int m;
-
-	/* Allocate struct plus pointers to first level page tables. */
-	m = (count + QIB_SEGSZ - 1) / QIB_SEGSZ;
-	mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL);
-	if (!mr)
-		goto bail;
-
-	rval = init_qib_mregion(&mr->mr, pd, count);
-	if (rval)
-		goto bail;
-
-	rval = qib_alloc_lkey(&mr->mr, 0);
-	if (rval)
-		goto bail_mregion;
-	mr->ibmr.lkey = mr->mr.lkey;
-	mr->ibmr.rkey = mr->mr.lkey;
-done:
-	return mr;
-
-bail_mregion:
-	deinit_qib_mregion(&mr->mr);
-bail:
-	kfree(mr);
-	mr = ERR_PTR(rval);
-	goto done;
-}
-
-/**
- * qib_reg_user_mr - register a userspace memory region
- * @pd: protection domain for this memory region
- * @start: starting userspace address
- * @length: length of region to register
- * @mr_access_flags: access flags for this memory region
- * @udata: unused by the QLogic_IB driver
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_mr *qib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
-			      u64 virt_addr, int mr_access_flags,
-			      struct ib_udata *udata)
-{
-	struct qib_mr *mr;
-	struct ib_umem *umem;
-	struct scatterlist *sg;
-	int n, m, entry;
-	struct ib_mr *ret;
-
-	if (length == 0) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
-
-	umem = ib_umem_get(pd->uobject->context, start, length,
-			   mr_access_flags, 0);
-	if (IS_ERR(umem))
-		return (void *) umem;
-
-	n = umem->nmap;
-
-	mr = alloc_mr(n, pd);
-	if (IS_ERR(mr)) {
-		ret = (struct ib_mr *)mr;
-		ib_umem_release(umem);
-		goto bail;
-	}
-
-	mr->mr.user_base = start;
-	mr->mr.iova = virt_addr;
-	mr->mr.length = length;
-	mr->mr.offset = ib_umem_offset(umem);
-	mr->mr.access_flags = mr_access_flags;
-	mr->umem = umem;
-
-	if (is_power_of_2(umem->page_size))
-		mr->mr.page_shift = ilog2(umem->page_size);
-	m = 0;
-	n = 0;
-	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
-			void *vaddr;
-
-			vaddr = page_address(sg_page(sg));
-			if (!vaddr) {
-				ret = ERR_PTR(-EINVAL);
-				goto bail;
-			}
-			mr->mr.map[m]->segs[n].vaddr = vaddr;
-			mr->mr.map[m]->segs[n].length = umem->page_size;
-			n++;
-			if (n == QIB_SEGSZ) {
-				m++;
-				n = 0;
-			}
-	}
-	ret = &mr->ibmr;
-
-bail:
-	return ret;
-}
-
-/**
- * qib_dereg_mr - unregister and free a memory region
- * @ibmr: the memory region to free
- *
- * Returns 0 on success.
- *
- * Note that this is called to free MRs created by qib_get_dma_mr()
- * or qib_reg_user_mr().
- */
-int qib_dereg_mr(struct ib_mr *ibmr)
-{
-	struct qib_mr *mr = to_imr(ibmr);
-	int ret = 0;
-	unsigned long timeout;
-
-	kfree(mr->pages);
-	qib_free_lkey(&mr->mr);
-
-	qib_put_mr(&mr->mr); /* will set completion if last */
-	timeout = wait_for_completion_timeout(&mr->mr.comp,
-		5 * HZ);
-	if (!timeout) {
-		qib_get_mr(&mr->mr);
-		ret = -EBUSY;
-		goto out;
-	}
-	deinit_qib_mregion(&mr->mr);
-	if (mr->umem)
-		ib_umem_release(mr->umem);
-	kfree(mr);
-out:
-	return ret;
-}
-
-/*
- * Allocate a memory region usable with the
- * IB_WR_REG_MR send work request.
- *
- * Return the memory region on success, otherwise return an errno.
- */
-struct ib_mr *qib_alloc_mr(struct ib_pd *pd,
-			   enum ib_mr_type mr_type,
-			   u32 max_num_sg)
-{
-	struct qib_mr *mr;
-
-	if (mr_type != IB_MR_TYPE_MEM_REG)
-		return ERR_PTR(-EINVAL);
-
-	mr = alloc_mr(max_num_sg, pd);
-	if (IS_ERR(mr))
-		return (struct ib_mr *)mr;
-
-	mr->pages = kcalloc(max_num_sg, sizeof(u64), GFP_KERNEL);
-	if (!mr->pages)
-		goto err;
-
-	return &mr->ibmr;
-
-err:
-	qib_dereg_mr(&mr->ibmr);
-	return ERR_PTR(-ENOMEM);
-}
-
-static int qib_set_page(struct ib_mr *ibmr, u64 addr)
-{
-	struct qib_mr *mr = to_imr(ibmr);
-
-	if (unlikely(mr->npages == mr->mr.max_segs))
-		return -ENOMEM;
-
-	mr->pages[mr->npages++] = addr;
-
-	return 0;
-}
-
-int qib_map_mr_sg(struct ib_mr *ibmr,
-		  struct scatterlist *sg,
-		  int sg_nents)
-{
-	struct qib_mr *mr = to_imr(ibmr);
-
-	mr->npages = 0;
-
-	return ib_sg_to_pages(ibmr, sg, sg_nents, qib_set_page);
-}
-
-/**
- * qib_alloc_fmr - allocate a fast memory region
- * @pd: the protection domain for this memory region
- * @mr_access_flags: access flags for this memory region
- * @fmr_attr: fast memory region attributes
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_fmr *qib_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
-			     struct ib_fmr_attr *fmr_attr)
-{
-	struct qib_fmr *fmr;
-	int m;
-	struct ib_fmr *ret;
-	int rval = -ENOMEM;
-
-	/* Allocate struct plus pointers to first level page tables. */
-	m = (fmr_attr->max_pages + QIB_SEGSZ - 1) / QIB_SEGSZ;
-	fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL);
-	if (!fmr)
-		goto bail;
-
-	rval = init_qib_mregion(&fmr->mr, pd, fmr_attr->max_pages);
-	if (rval)
-		goto bail;
-
-	/*
-	 * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
-	 * rkey.
-	 */
-	rval = qib_alloc_lkey(&fmr->mr, 0);
-	if (rval)
-		goto bail_mregion;
-	fmr->ibfmr.rkey = fmr->mr.lkey;
-	fmr->ibfmr.lkey = fmr->mr.lkey;
-	/*
-	 * Resources are allocated but no valid mapping (RKEY can't be
-	 * used).
-	 */
-	fmr->mr.access_flags = mr_access_flags;
-	fmr->mr.max_segs = fmr_attr->max_pages;
-	fmr->mr.page_shift = fmr_attr->page_shift;
-
-	ret = &fmr->ibfmr;
-done:
-	return ret;
-
-bail_mregion:
-	deinit_qib_mregion(&fmr->mr);
-bail:
-	kfree(fmr);
-	ret = ERR_PTR(rval);
-	goto done;
-}
-
-/**
- * qib_map_phys_fmr - set up a fast memory region
- * @ibmfr: the fast memory region to set up
- * @page_list: the list of pages to associate with the fast memory region
- * @list_len: the number of pages to associate with the fast memory region
- * @iova: the virtual address of the start of the fast memory region
- *
- * This may be called from interrupt context.
- */
-
-int qib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
-		     int list_len, u64 iova)
-{
-	struct qib_fmr *fmr = to_ifmr(ibfmr);
-	struct qib_lkey_table *rkt;
-	unsigned long flags;
-	int m, n, i;
-	u32 ps;
-	int ret;
-
-	i = atomic_read(&fmr->mr.refcount);
-	if (i > 2)
-		return -EBUSY;
-
-	if (list_len > fmr->mr.max_segs) {
-		ret = -EINVAL;
-		goto bail;
-	}
-	rkt = &to_idev(ibfmr->device)->lk_table;
-	spin_lock_irqsave(&rkt->lock, flags);
-	fmr->mr.user_base = iova;
-	fmr->mr.iova = iova;
-	ps = 1 << fmr->mr.page_shift;
-	fmr->mr.length = list_len * ps;
-	m = 0;
-	n = 0;
-	for (i = 0; i < list_len; i++) {
-		fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
-		fmr->mr.map[m]->segs[n].length = ps;
-		if (++n == QIB_SEGSZ) {
-			m++;
-			n = 0;
-		}
-	}
-	spin_unlock_irqrestore(&rkt->lock, flags);
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * qib_unmap_fmr - unmap fast memory regions
- * @fmr_list: the list of fast memory regions to unmap
- *
- * Returns 0 on success.
- */
-int qib_unmap_fmr(struct list_head *fmr_list)
-{
-	struct qib_fmr *fmr;
-	struct qib_lkey_table *rkt;
-	unsigned long flags;
-
-	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
-		rkt = &to_idev(fmr->ibfmr.device)->lk_table;
-		spin_lock_irqsave(&rkt->lock, flags);
-		fmr->mr.user_base = 0;
-		fmr->mr.iova = 0;
-		fmr->mr.length = 0;
-		spin_unlock_irqrestore(&rkt->lock, flags);
-	}
-	return 0;
-}
-
-/**
- * qib_dealloc_fmr - deallocate a fast memory region
- * @ibfmr: the fast memory region to deallocate
- *
- * Returns 0 on success.
- */
-int qib_dealloc_fmr(struct ib_fmr *ibfmr)
-{
-	struct qib_fmr *fmr = to_ifmr(ibfmr);
-	int ret = 0;
-	unsigned long timeout;
-
-	qib_free_lkey(&fmr->mr);
-	qib_put_mr(&fmr->mr); /* will set completion if last */
-	timeout = wait_for_completion_timeout(&fmr->mr.comp,
-		5 * HZ);
-	if (!timeout) {
-		qib_get_mr(&fmr->mr);
-		ret = -EBUSY;
-		goto out;
-	}
-	deinit_qib_mregion(&fmr->mr);
-	kfree(fmr);
-out:
-	return ret;
-}
-
-void mr_rcu_callback(struct rcu_head *list)
-{
-	struct qib_mregion *mr = container_of(list, struct qib_mregion, list);
-
-	complete(&mr->comp);
-}
diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c
index 3eff35c2d453..575b737d9ef3 100644
--- a/drivers/infiniband/hw/qib/qib_qp.c
+++ b/drivers/infiniband/hw/qib/qib_qp.c
@@ -34,32 +34,38 @@
 
 #include <linux/err.h>
 #include <linux/vmalloc.h>
-#include <linux/jhash.h>
+#include <rdma/rdma_vt.h>
 #ifdef CONFIG_DEBUG_FS
 #include <linux/seq_file.h>
 #endif
 
 #include "qib.h"
 
-#define BITS_PER_PAGE           (PAGE_SIZE*BITS_PER_BYTE)
-#define BITS_PER_PAGE_MASK      (BITS_PER_PAGE-1)
+/*
+ * mask field which was present in now deleted qib_qpn_table
+ * is not present in rvt_qpn_table. Defining the same field
+ * as qpt_mask here instead of adding the mask field to
+ * rvt_qpn_table.
+ */
+u16 qpt_mask;
 
-static inline unsigned mk_qpn(struct qib_qpn_table *qpt,
-			      struct qpn_map *map, unsigned off)
+static inline unsigned mk_qpn(struct rvt_qpn_table *qpt,
+			      struct rvt_qpn_map *map, unsigned off)
 {
-	return (map - qpt->map) * BITS_PER_PAGE + off;
+	return (map - qpt->map) * RVT_BITS_PER_PAGE + off;
 }
 
-static inline unsigned find_next_offset(struct qib_qpn_table *qpt,
-					struct qpn_map *map, unsigned off,
+static inline unsigned find_next_offset(struct rvt_qpn_table *qpt,
+					struct rvt_qpn_map *map, unsigned off,
 					unsigned n)
 {
-	if (qpt->mask) {
+	if (qpt_mask) {
 		off++;
-		if (((off & qpt->mask) >> 1) >= n)
-			off = (off | qpt->mask) + 2;
-	} else
-		off = find_next_zero_bit(map->page, BITS_PER_PAGE, off);
+		if (((off & qpt_mask) >> 1) >= n)
+			off = (off | qpt_mask) + 2;
+	} else {
+		off = find_next_zero_bit(map->page, RVT_BITS_PER_PAGE, off);
+	}
 	return off;
 }
 
@@ -100,7 +106,7 @@ static u32 credit_table[31] = {
 	32768                   /* 1E */
 };
 
-static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map,
+static void get_map_page(struct rvt_qpn_table *qpt, struct rvt_qpn_map *map,
 			 gfp_t gfp)
 {
 	unsigned long page = get_zeroed_page(gfp);
@@ -121,12 +127,15 @@ static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map,
  * Allocate the next available QPN or
  * zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
  */
-static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
-		     enum ib_qp_type type, u8 port, gfp_t gfp)
+int qib_alloc_qpn(struct rvt_dev_info *rdi, struct rvt_qpn_table *qpt,
+		  enum ib_qp_type type, u8 port, gfp_t gfp)
 {
 	u32 i, offset, max_scan, qpn;
-	struct qpn_map *map;
+	struct rvt_qpn_map *map;
 	u32 ret;
+	struct qib_ibdev *verbs_dev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(verbs_dev, struct qib_devdata,
+					      verbs_dev);
 
 	if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
 		unsigned n;
@@ -143,12 +152,12 @@ static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
 	}
 
 	qpn = qpt->last + 2;
-	if (qpn >= QPN_MAX)
+	if (qpn >= RVT_QPN_MAX)
 		qpn = 2;
-	if (qpt->mask && ((qpn & qpt->mask) >> 1) >= dd->n_krcv_queues)
-		qpn = (qpn | qpt->mask) + 2;
-	offset = qpn & BITS_PER_PAGE_MASK;
-	map = &qpt->map[qpn / BITS_PER_PAGE];
+	if (qpt_mask && ((qpn & qpt_mask) >> 1) >= dd->n_krcv_queues)
+		qpn = (qpn | qpt_mask) + 2;
+	offset = qpn & RVT_BITS_PER_PAGE_MASK;
+	map = &qpt->map[qpn / RVT_BITS_PER_PAGE];
 	max_scan = qpt->nmaps - !offset;
 	for (i = 0;;) {
 		if (unlikely(!map->page)) {
@@ -173,14 +182,14 @@ static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
 			 * We just need to be sure we don't loop
 			 * forever.
 			 */
-		} while (offset < BITS_PER_PAGE && qpn < QPN_MAX);
+		} while (offset < RVT_BITS_PER_PAGE && qpn < RVT_QPN_MAX);
 		/*
 		 * In order to keep the number of pages allocated to a
 		 * minimum, we scan the all existing pages before increasing
 		 * the size of the bitmap table.
 		 */
 		if (++i > max_scan) {
-			if (qpt->nmaps == QPNMAP_ENTRIES)
+			if (qpt->nmaps == RVT_QPNMAP_ENTRIES)
 				break;
 			map = &qpt->map[qpt->nmaps++];
 			offset = 0;
@@ -200,706 +209,113 @@ bail:
 	return ret;
 }
 
-static void free_qpn(struct qib_qpn_table *qpt, u32 qpn)
-{
-	struct qpn_map *map;
-
-	map = qpt->map + qpn / BITS_PER_PAGE;
-	if (map->page)
-		clear_bit(qpn & BITS_PER_PAGE_MASK, map->page);
-}
-
-static inline unsigned qpn_hash(struct qib_ibdev *dev, u32 qpn)
-{
-	return jhash_1word(qpn, dev->qp_rnd) &
-		(dev->qp_table_size - 1);
-}
-
-
-/*
- * Put the QP into the hash table.
- * The hash table holds a reference to the QP.
- */
-static void insert_qp(struct qib_ibdev *dev, struct qib_qp *qp)
-{
-	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	unsigned long flags;
-	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
-
-	atomic_inc(&qp->refcount);
-	spin_lock_irqsave(&dev->qpt_lock, flags);
-
-	if (qp->ibqp.qp_num == 0)
-		rcu_assign_pointer(ibp->qp0, qp);
-	else if (qp->ibqp.qp_num == 1)
-		rcu_assign_pointer(ibp->qp1, qp);
-	else {
-		qp->next = dev->qp_table[n];
-		rcu_assign_pointer(dev->qp_table[n], qp);
-	}
-
-	spin_unlock_irqrestore(&dev->qpt_lock, flags);
-}
-
-/*
- * Remove the QP from the table so it can't be found asynchronously by
- * the receive interrupt routine.
- */
-static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp)
-{
-	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
-	unsigned long flags;
-	int removed = 1;
-
-	spin_lock_irqsave(&dev->qpt_lock, flags);
-
-	if (rcu_dereference_protected(ibp->qp0,
-			lockdep_is_held(&dev->qpt_lock)) == qp) {
-		RCU_INIT_POINTER(ibp->qp0, NULL);
-	} else if (rcu_dereference_protected(ibp->qp1,
-			lockdep_is_held(&dev->qpt_lock)) == qp) {
-		RCU_INIT_POINTER(ibp->qp1, NULL);
-	} else {
-		struct qib_qp *q;
-		struct qib_qp __rcu **qpp;
-
-		removed = 0;
-		qpp = &dev->qp_table[n];
-		for (; (q = rcu_dereference_protected(*qpp,
-				lockdep_is_held(&dev->qpt_lock))) != NULL;
-				qpp = &q->next)
-			if (q == qp) {
-				RCU_INIT_POINTER(*qpp,
-					rcu_dereference_protected(qp->next,
-					 lockdep_is_held(&dev->qpt_lock)));
-				removed = 1;
-				break;
-			}
-	}
-
-	spin_unlock_irqrestore(&dev->qpt_lock, flags);
-	if (removed) {
-		synchronize_rcu();
-		atomic_dec(&qp->refcount);
-	}
-}
-
 /**
  * qib_free_all_qps - check for QPs still in use
- * @qpt: the QP table to empty
- *
- * There should not be any QPs still in use.
- * Free memory for table.
  */
-unsigned qib_free_all_qps(struct qib_devdata *dd)
+unsigned qib_free_all_qps(struct rvt_dev_info *rdi)
 {
-	struct qib_ibdev *dev = &dd->verbs_dev;
-	unsigned long flags;
-	struct qib_qp *qp;
+	struct qib_ibdev *verbs_dev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(verbs_dev, struct qib_devdata,
+					      verbs_dev);
 	unsigned n, qp_inuse = 0;
 
 	for (n = 0; n < dd->num_pports; n++) {
 		struct qib_ibport *ibp = &dd->pport[n].ibport_data;
 
-		if (!qib_mcast_tree_empty(ibp))
-			qp_inuse++;
 		rcu_read_lock();
-		if (rcu_dereference(ibp->qp0))
+		if (rcu_dereference(ibp->rvp.qp[0]))
 			qp_inuse++;
-		if (rcu_dereference(ibp->qp1))
+		if (rcu_dereference(ibp->rvp.qp[1]))
 			qp_inuse++;
 		rcu_read_unlock();
 	}
-
-	spin_lock_irqsave(&dev->qpt_lock, flags);
-	for (n = 0; n < dev->qp_table_size; n++) {
-		qp = rcu_dereference_protected(dev->qp_table[n],
-			lockdep_is_held(&dev->qpt_lock));
-		RCU_INIT_POINTER(dev->qp_table[n], NULL);
-
-		for (; qp; qp = rcu_dereference_protected(qp->next,
-					lockdep_is_held(&dev->qpt_lock)))
-			qp_inuse++;
-	}
-	spin_unlock_irqrestore(&dev->qpt_lock, flags);
-	synchronize_rcu();
-
 	return qp_inuse;
 }
 
-/**
- * qib_lookup_qpn - return the QP with the given QPN
- * @qpt: the QP table
- * @qpn: the QP number to look up
- *
- * The caller is responsible for decrementing the QP reference count
- * when done.
- */
-struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn)
+void qib_notify_qp_reset(struct rvt_qp *qp)
 {
-	struct qib_qp *qp = NULL;
-
-	rcu_read_lock();
-	if (unlikely(qpn <= 1)) {
-		if (qpn == 0)
-			qp = rcu_dereference(ibp->qp0);
-		else
-			qp = rcu_dereference(ibp->qp1);
-		if (qp)
-			atomic_inc(&qp->refcount);
-	} else {
-		struct qib_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
-		unsigned n = qpn_hash(dev, qpn);
-
-		for (qp = rcu_dereference(dev->qp_table[n]); qp;
-			qp = rcu_dereference(qp->next))
-			if (qp->ibqp.qp_num == qpn) {
-				atomic_inc(&qp->refcount);
-				break;
-			}
-	}
-	rcu_read_unlock();
-	return qp;
-}
-
-/**
- * qib_reset_qp - initialize the QP state to the reset state
- * @qp: the QP to reset
- * @type: the QP type
- */
-static void qib_reset_qp(struct qib_qp *qp, enum ib_qp_type type)
-{
-	qp->remote_qpn = 0;
-	qp->qkey = 0;
-	qp->qp_access_flags = 0;
-	atomic_set(&qp->s_dma_busy, 0);
-	qp->s_flags &= QIB_S_SIGNAL_REQ_WR;
-	qp->s_hdrwords = 0;
-	qp->s_wqe = NULL;
-	qp->s_draining = 0;
-	qp->s_next_psn = 0;
-	qp->s_last_psn = 0;
-	qp->s_sending_psn = 0;
-	qp->s_sending_hpsn = 0;
-	qp->s_psn = 0;
-	qp->r_psn = 0;
-	qp->r_msn = 0;
-	if (type == IB_QPT_RC) {
-		qp->s_state = IB_OPCODE_RC_SEND_LAST;
-		qp->r_state = IB_OPCODE_RC_SEND_LAST;
-	} else {
-		qp->s_state = IB_OPCODE_UC_SEND_LAST;
-		qp->r_state = IB_OPCODE_UC_SEND_LAST;
-	}
-	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
-	qp->r_nak_state = 0;
-	qp->r_aflags = 0;
-	qp->r_flags = 0;
-	qp->s_head = 0;
-	qp->s_tail = 0;
-	qp->s_cur = 0;
-	qp->s_acked = 0;
-	qp->s_last = 0;
-	qp->s_ssn = 1;
-	qp->s_lsn = 0;
-	qp->s_mig_state = IB_MIG_MIGRATED;
-	memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
-	qp->r_head_ack_queue = 0;
-	qp->s_tail_ack_queue = 0;
-	qp->s_num_rd_atomic = 0;
-	if (qp->r_rq.wq) {
-		qp->r_rq.wq->head = 0;
-		qp->r_rq.wq->tail = 0;
-	}
-	qp->r_sge.num_sge = 0;
-}
-
-static void clear_mr_refs(struct qib_qp *qp, int clr_sends)
-{
-	unsigned n;
-
-	if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
-		qib_put_ss(&qp->s_rdma_read_sge);
-
-	qib_put_ss(&qp->r_sge);
-
-	if (clr_sends) {
-		while (qp->s_last != qp->s_head) {
-			struct qib_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
-			unsigned i;
-
-			for (i = 0; i < wqe->wr.num_sge; i++) {
-				struct qib_sge *sge = &wqe->sg_list[i];
-
-				qib_put_mr(sge->mr);
-			}
-			if (qp->ibqp.qp_type == IB_QPT_UD ||
-			    qp->ibqp.qp_type == IB_QPT_SMI ||
-			    qp->ibqp.qp_type == IB_QPT_GSI)
-				atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
-			if (++qp->s_last >= qp->s_size)
-				qp->s_last = 0;
-		}
-		if (qp->s_rdma_mr) {
-			qib_put_mr(qp->s_rdma_mr);
-			qp->s_rdma_mr = NULL;
-		}
-	}
-
-	if (qp->ibqp.qp_type != IB_QPT_RC)
-		return;
+	struct qib_qp_priv *priv = qp->priv;
 
-	for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
-		struct qib_ack_entry *e = &qp->s_ack_queue[n];
-
-		if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
-		    e->rdma_sge.mr) {
-			qib_put_mr(e->rdma_sge.mr);
-			e->rdma_sge.mr = NULL;
-		}
-	}
+	atomic_set(&priv->s_dma_busy, 0);
 }
 
-/**
- * qib_error_qp - put a QP into the error state
- * @qp: the QP to put into the error state
- * @err: the receive completion error to signal if a RWQE is active
- *
- * Flushes both send and receive work queues.
- * Returns true if last WQE event should be generated.
- * The QP r_lock and s_lock should be held and interrupts disabled.
- * If we are already in error state, just return.
- */
-int qib_error_qp(struct qib_qp *qp, enum ib_wc_status err)
+void qib_notify_error_qp(struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
-	struct ib_wc wc;
-	int ret = 0;
-
-	if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
-		goto bail;
-
-	qp->state = IB_QPS_ERR;
-
-	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
-		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
-		del_timer(&qp->s_timer);
-	}
-
-	if (qp->s_flags & QIB_S_ANY_WAIT_SEND)
-		qp->s_flags &= ~QIB_S_ANY_WAIT_SEND;
 
-	spin_lock(&dev->pending_lock);
-	if (!list_empty(&qp->iowait) && !(qp->s_flags & QIB_S_BUSY)) {
-		qp->s_flags &= ~QIB_S_ANY_WAIT_IO;
-		list_del_init(&qp->iowait);
+	spin_lock(&dev->rdi.pending_lock);
+	if (!list_empty(&priv->iowait) && !(qp->s_flags & RVT_S_BUSY)) {
+		qp->s_flags &= ~RVT_S_ANY_WAIT_IO;
+		list_del_init(&priv->iowait);
 	}
-	spin_unlock(&dev->pending_lock);
+	spin_unlock(&dev->rdi.pending_lock);
 
-	if (!(qp->s_flags & QIB_S_BUSY)) {
+	if (!(qp->s_flags & RVT_S_BUSY)) {
 		qp->s_hdrwords = 0;
 		if (qp->s_rdma_mr) {
-			qib_put_mr(qp->s_rdma_mr);
+			rvt_put_mr(qp->s_rdma_mr);
 			qp->s_rdma_mr = NULL;
 		}
-		if (qp->s_tx) {
-			qib_put_txreq(qp->s_tx);
-			qp->s_tx = NULL;
+		if (priv->s_tx) {
+			qib_put_txreq(priv->s_tx);
+			priv->s_tx = NULL;
 		}
 	}
-
-	/* Schedule the sending tasklet to drain the send work queue. */
-	if (qp->s_last != qp->s_head)
-		qib_schedule_send(qp);
-
-	clear_mr_refs(qp, 0);
-
-	memset(&wc, 0, sizeof(wc));
-	wc.qp = &qp->ibqp;
-	wc.opcode = IB_WC_RECV;
-
-	if (test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags)) {
-		wc.wr_id = qp->r_wr_id;
-		wc.status = err;
-		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
-	}
-	wc.status = IB_WC_WR_FLUSH_ERR;
-
-	if (qp->r_rq.wq) {
-		struct qib_rwq *wq;
-		u32 head;
-		u32 tail;
-
-		spin_lock(&qp->r_rq.lock);
-
-		/* sanity check pointers before trusting them */
-		wq = qp->r_rq.wq;
-		head = wq->head;
-		if (head >= qp->r_rq.size)
-			head = 0;
-		tail = wq->tail;
-		if (tail >= qp->r_rq.size)
-			tail = 0;
-		while (tail != head) {
-			wc.wr_id = get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
-			if (++tail >= qp->r_rq.size)
-				tail = 0;
-			qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
-		}
-		wq->tail = tail;
-
-		spin_unlock(&qp->r_rq.lock);
-	} else if (qp->ibqp.event_handler)
-		ret = 1;
-
-bail:
-	return ret;
 }
 
-/**
- * qib_modify_qp - modify the attributes of a queue pair
- * @ibqp: the queue pair who's attributes we're modifying
- * @attr: the new attributes
- * @attr_mask: the mask of attributes to modify
- * @udata: user data for libibverbs.so
- *
- * Returns 0 on success, otherwise returns an errno.
- */
-int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		  int attr_mask, struct ib_udata *udata)
+static int mtu_to_enum(u32 mtu)
 {
-	struct qib_ibdev *dev = to_idev(ibqp->device);
-	struct qib_qp *qp = to_iqp(ibqp);
-	enum ib_qp_state cur_state, new_state;
-	struct ib_event ev;
-	int lastwqe = 0;
-	int mig = 0;
-	int ret;
-	u32 pmtu = 0; /* for gcc warning only */
-
-	spin_lock_irq(&qp->r_lock);
-	spin_lock(&qp->s_lock);
-
-	cur_state = attr_mask & IB_QP_CUR_STATE ?
-		attr->cur_qp_state : qp->state;
-	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
-
-	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
-				attr_mask, IB_LINK_LAYER_UNSPECIFIED))
-		goto inval;
-
-	if (attr_mask & IB_QP_AV) {
-		if (attr->ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
-			goto inval;
-		if (qib_check_ah(qp->ibqp.device, &attr->ah_attr))
-			goto inval;
-	}
-
-	if (attr_mask & IB_QP_ALT_PATH) {
-		if (attr->alt_ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
-			goto inval;
-		if (qib_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
-			goto inval;
-		if (attr->alt_pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
-			goto inval;
-	}
-
-	if (attr_mask & IB_QP_PKEY_INDEX)
-		if (attr->pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
-			goto inval;
-
-	if (attr_mask & IB_QP_MIN_RNR_TIMER)
-		if (attr->min_rnr_timer > 31)
-			goto inval;
-
-	if (attr_mask & IB_QP_PORT)
-		if (qp->ibqp.qp_type == IB_QPT_SMI ||
-		    qp->ibqp.qp_type == IB_QPT_GSI ||
-		    attr->port_num == 0 ||
-		    attr->port_num > ibqp->device->phys_port_cnt)
-			goto inval;
-
-	if (attr_mask & IB_QP_DEST_QPN)
-		if (attr->dest_qp_num > QIB_QPN_MASK)
-			goto inval;
-
-	if (attr_mask & IB_QP_RETRY_CNT)
-		if (attr->retry_cnt > 7)
-			goto inval;
-
-	if (attr_mask & IB_QP_RNR_RETRY)
-		if (attr->rnr_retry > 7)
-			goto inval;
-
-	/*
-	 * Don't allow invalid path_mtu values.  OK to set greater
-	 * than the active mtu (or even the max_cap, if we have tuned
-	 * that to a small mtu.  We'll set qp->path_mtu
-	 * to the lesser of requested attribute mtu and active,
-	 * for packetizing messages.
-	 * Note that the QP port has to be set in INIT and MTU in RTR.
-	 */
-	if (attr_mask & IB_QP_PATH_MTU) {
-		struct qib_devdata *dd = dd_from_dev(dev);
-		int mtu, pidx = qp->port_num - 1;
-
-		mtu = ib_mtu_enum_to_int(attr->path_mtu);
-		if (mtu == -1)
-			goto inval;
-		if (mtu > dd->pport[pidx].ibmtu) {
-			switch (dd->pport[pidx].ibmtu) {
-			case 4096:
-				pmtu = IB_MTU_4096;
-				break;
-			case 2048:
-				pmtu = IB_MTU_2048;
-				break;
-			case 1024:
-				pmtu = IB_MTU_1024;
-				break;
-			case 512:
-				pmtu = IB_MTU_512;
-				break;
-			case 256:
-				pmtu = IB_MTU_256;
-				break;
-			default:
-				pmtu = IB_MTU_2048;
-			}
-		} else
-			pmtu = attr->path_mtu;
-	}
-
-	if (attr_mask & IB_QP_PATH_MIG_STATE) {
-		if (attr->path_mig_state == IB_MIG_REARM) {
-			if (qp->s_mig_state == IB_MIG_ARMED)
-				goto inval;
-			if (new_state != IB_QPS_RTS)
-				goto inval;
-		} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
-			if (qp->s_mig_state == IB_MIG_REARM)
-				goto inval;
-			if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
-				goto inval;
-			if (qp->s_mig_state == IB_MIG_ARMED)
-				mig = 1;
-		} else
-			goto inval;
-	}
-
-	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
-		if (attr->max_dest_rd_atomic > QIB_MAX_RDMA_ATOMIC)
-			goto inval;
+	int enum_mtu;
 
-	switch (new_state) {
-	case IB_QPS_RESET:
-		if (qp->state != IB_QPS_RESET) {
-			qp->state = IB_QPS_RESET;
-			spin_lock(&dev->pending_lock);
-			if (!list_empty(&qp->iowait))
-				list_del_init(&qp->iowait);
-			spin_unlock(&dev->pending_lock);
-			qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
-			spin_unlock(&qp->s_lock);
-			spin_unlock_irq(&qp->r_lock);
-			/* Stop the sending work queue and retry timer */
-			cancel_work_sync(&qp->s_work);
-			del_timer_sync(&qp->s_timer);
-			wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
-			if (qp->s_tx) {
-				qib_put_txreq(qp->s_tx);
-				qp->s_tx = NULL;
-			}
-			remove_qp(dev, qp);
-			wait_event(qp->wait, !atomic_read(&qp->refcount));
-			spin_lock_irq(&qp->r_lock);
-			spin_lock(&qp->s_lock);
-			clear_mr_refs(qp, 1);
-			qib_reset_qp(qp, ibqp->qp_type);
-		}
+	switch (mtu) {
+	case 4096:
+		enum_mtu = IB_MTU_4096;
 		break;
-
-	case IB_QPS_RTR:
-		/* Allow event to retrigger if QP set to RTR more than once */
-		qp->r_flags &= ~QIB_R_COMM_EST;
-		qp->state = new_state;
+	case 2048:
+		enum_mtu = IB_MTU_2048;
 		break;
-
-	case IB_QPS_SQD:
-		qp->s_draining = qp->s_last != qp->s_cur;
-		qp->state = new_state;
+	case 1024:
+		enum_mtu = IB_MTU_1024;
 		break;
-
-	case IB_QPS_SQE:
-		if (qp->ibqp.qp_type == IB_QPT_RC)
-			goto inval;
-		qp->state = new_state;
+	case 512:
+		enum_mtu = IB_MTU_512;
 		break;
-
-	case IB_QPS_ERR:
-		lastwqe = qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+	case 256:
+		enum_mtu = IB_MTU_256;
 		break;
-
 	default:
-		qp->state = new_state;
-		break;
-	}
-
-	if (attr_mask & IB_QP_PKEY_INDEX)
-		qp->s_pkey_index = attr->pkey_index;
-
-	if (attr_mask & IB_QP_PORT)
-		qp->port_num = attr->port_num;
-
-	if (attr_mask & IB_QP_DEST_QPN)
-		qp->remote_qpn = attr->dest_qp_num;
-
-	if (attr_mask & IB_QP_SQ_PSN) {
-		qp->s_next_psn = attr->sq_psn & QIB_PSN_MASK;
-		qp->s_psn = qp->s_next_psn;
-		qp->s_sending_psn = qp->s_next_psn;
-		qp->s_last_psn = qp->s_next_psn - 1;
-		qp->s_sending_hpsn = qp->s_last_psn;
-	}
-
-	if (attr_mask & IB_QP_RQ_PSN)
-		qp->r_psn = attr->rq_psn & QIB_PSN_MASK;
-
-	if (attr_mask & IB_QP_ACCESS_FLAGS)
-		qp->qp_access_flags = attr->qp_access_flags;
-
-	if (attr_mask & IB_QP_AV) {
-		qp->remote_ah_attr = attr->ah_attr;
-		qp->s_srate = attr->ah_attr.static_rate;
-	}
-
-	if (attr_mask & IB_QP_ALT_PATH) {
-		qp->alt_ah_attr = attr->alt_ah_attr;
-		qp->s_alt_pkey_index = attr->alt_pkey_index;
-	}
-
-	if (attr_mask & IB_QP_PATH_MIG_STATE) {
-		qp->s_mig_state = attr->path_mig_state;
-		if (mig) {
-			qp->remote_ah_attr = qp->alt_ah_attr;
-			qp->port_num = qp->alt_ah_attr.port_num;
-			qp->s_pkey_index = qp->s_alt_pkey_index;
-		}
-	}
-
-	if (attr_mask & IB_QP_PATH_MTU) {
-		qp->path_mtu = pmtu;
-		qp->pmtu = ib_mtu_enum_to_int(pmtu);
-	}
-
-	if (attr_mask & IB_QP_RETRY_CNT) {
-		qp->s_retry_cnt = attr->retry_cnt;
-		qp->s_retry = attr->retry_cnt;
-	}
-
-	if (attr_mask & IB_QP_RNR_RETRY) {
-		qp->s_rnr_retry_cnt = attr->rnr_retry;
-		qp->s_rnr_retry = attr->rnr_retry;
+		enum_mtu = IB_MTU_2048;
 	}
-
-	if (attr_mask & IB_QP_MIN_RNR_TIMER)
-		qp->r_min_rnr_timer = attr->min_rnr_timer;
-
-	if (attr_mask & IB_QP_TIMEOUT) {
-		qp->timeout = attr->timeout;
-		qp->timeout_jiffies =
-			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
-				1000UL);
-	}
-
-	if (attr_mask & IB_QP_QKEY)
-		qp->qkey = attr->qkey;
-
-	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
-		qp->r_max_rd_atomic = attr->max_dest_rd_atomic;
-
-	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
-		qp->s_max_rd_atomic = attr->max_rd_atomic;
-
-	spin_unlock(&qp->s_lock);
-	spin_unlock_irq(&qp->r_lock);
-
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
-		insert_qp(dev, qp);
-
-	if (lastwqe) {
-		ev.device = qp->ibqp.device;
-		ev.element.qp = &qp->ibqp;
-		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
-		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
-	}
-	if (mig) {
-		ev.device = qp->ibqp.device;
-		ev.element.qp = &qp->ibqp;
-		ev.event = IB_EVENT_PATH_MIG;
-		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
-	}
-	ret = 0;
-	goto bail;
-
-inval:
-	spin_unlock(&qp->s_lock);
-	spin_unlock_irq(&qp->r_lock);
-	ret = -EINVAL;
-
-bail:
-	return ret;
+	return enum_mtu;
 }
 
-int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		 int attr_mask, struct ib_qp_init_attr *init_attr)
+int qib_get_pmtu_from_attr(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+			   struct ib_qp_attr *attr)
 {
-	struct qib_qp *qp = to_iqp(ibqp);
+	int mtu, pmtu, pidx = qp->port_num - 1;
+	struct qib_ibdev *verbs_dev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = container_of(verbs_dev, struct qib_devdata,
+					      verbs_dev);
+	mtu = ib_mtu_enum_to_int(attr->path_mtu);
+	if (mtu == -1)
+		return -EINVAL;
+
+	if (mtu > dd->pport[pidx].ibmtu)
+		pmtu = mtu_to_enum(dd->pport[pidx].ibmtu);
+	else
+		pmtu = attr->path_mtu;
+	return pmtu;
+}
 
-	attr->qp_state = qp->state;
-	attr->cur_qp_state = attr->qp_state;
-	attr->path_mtu = qp->path_mtu;
-	attr->path_mig_state = qp->s_mig_state;
-	attr->qkey = qp->qkey;
-	attr->rq_psn = qp->r_psn & QIB_PSN_MASK;
-	attr->sq_psn = qp->s_next_psn & QIB_PSN_MASK;
-	attr->dest_qp_num = qp->remote_qpn;
-	attr->qp_access_flags = qp->qp_access_flags;
-	attr->cap.max_send_wr = qp->s_size - 1;
-	attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
-	attr->cap.max_send_sge = qp->s_max_sge;
-	attr->cap.max_recv_sge = qp->r_rq.max_sge;
-	attr->cap.max_inline_data = 0;
-	attr->ah_attr = qp->remote_ah_attr;
-	attr->alt_ah_attr = qp->alt_ah_attr;
-	attr->pkey_index = qp->s_pkey_index;
-	attr->alt_pkey_index = qp->s_alt_pkey_index;
-	attr->en_sqd_async_notify = 0;
-	attr->sq_draining = qp->s_draining;
-	attr->max_rd_atomic = qp->s_max_rd_atomic;
-	attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
-	attr->min_rnr_timer = qp->r_min_rnr_timer;
-	attr->port_num = qp->port_num;
-	attr->timeout = qp->timeout;
-	attr->retry_cnt = qp->s_retry_cnt;
-	attr->rnr_retry = qp->s_rnr_retry_cnt;
-	attr->alt_port_num = qp->alt_ah_attr.port_num;
-	attr->alt_timeout = qp->alt_timeout;
+int qib_mtu_to_path_mtu(u32 mtu)
+{
+	return mtu_to_enum(mtu);
+}
 
-	init_attr->event_handler = qp->ibqp.event_handler;
-	init_attr->qp_context = qp->ibqp.qp_context;
-	init_attr->send_cq = qp->ibqp.send_cq;
-	init_attr->recv_cq = qp->ibqp.recv_cq;
-	init_attr->srq = qp->ibqp.srq;
-	init_attr->cap = attr->cap;
-	if (qp->s_flags & QIB_S_SIGNAL_REQ_WR)
-		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
-	else
-		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
-	init_attr->qp_type = qp->ibqp.qp_type;
-	init_attr->port_num = qp->port_num;
-	return 0;
+u32 qib_mtu_from_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, u32 pmtu)
+{
+	return ib_mtu_enum_to_int(pmtu);
 }
 
 /**
@@ -908,7 +324,7 @@ int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
  *
  * Returns the AETH.
  */
-__be32 qib_compute_aeth(struct qib_qp *qp)
+__be32 qib_compute_aeth(struct rvt_qp *qp)
 {
 	u32 aeth = qp->r_msn & QIB_MSN_MASK;
 
@@ -921,7 +337,7 @@ __be32 qib_compute_aeth(struct qib_qp *qp)
 	} else {
 		u32 min, max, x;
 		u32 credits;
-		struct qib_rwq *wq = qp->r_rq.wq;
+		struct rvt_rwq *wq = qp->r_rq.wq;
 		u32 head;
 		u32 tail;
 
@@ -962,315 +378,63 @@ __be32 qib_compute_aeth(struct qib_qp *qp)
 	return cpu_to_be32(aeth);
 }
 
-/**
- * qib_create_qp - create a queue pair for a device
- * @ibpd: the protection domain who's device we create the queue pair for
- * @init_attr: the attributes of the queue pair
- * @udata: user data for libibverbs.so
- *
- * Returns the queue pair on success, otherwise returns an errno.
- *
- * Called by the ib_create_qp() core verbs function.
- */
-struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
-			    struct ib_qp_init_attr *init_attr,
-			    struct ib_udata *udata)
+void *qib_qp_priv_alloc(struct rvt_dev_info *rdi, struct rvt_qp *qp, gfp_t gfp)
 {
-	struct qib_qp *qp;
-	int err;
-	struct qib_swqe *swq = NULL;
-	struct qib_ibdev *dev;
-	struct qib_devdata *dd;
-	size_t sz;
-	size_t sg_list_sz;
-	struct ib_qp *ret;
-	gfp_t gfp;
+	struct qib_qp_priv *priv;
 
+	priv = kzalloc(sizeof(*priv), gfp);
+	if (!priv)
+		return ERR_PTR(-ENOMEM);
+	priv->owner = qp;
 
-	if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
-	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs ||
-	    init_attr->create_flags & ~(IB_QP_CREATE_USE_GFP_NOIO))
-		return ERR_PTR(-EINVAL);
-
-	/* GFP_NOIO is applicable in RC QPs only */
-	if (init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO &&
-	    init_attr->qp_type != IB_QPT_RC)
-		return ERR_PTR(-EINVAL);
-
-	gfp = init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO ?
-			GFP_NOIO : GFP_KERNEL;
-
-	/* Check receive queue parameters if no SRQ is specified. */
-	if (!init_attr->srq) {
-		if (init_attr->cap.max_recv_sge > ib_qib_max_sges ||
-		    init_attr->cap.max_recv_wr > ib_qib_max_qp_wrs) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-		if (init_attr->cap.max_send_sge +
-		    init_attr->cap.max_send_wr +
-		    init_attr->cap.max_recv_sge +
-		    init_attr->cap.max_recv_wr == 0) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
+	priv->s_hdr = kzalloc(sizeof(*priv->s_hdr), gfp);
+	if (!priv->s_hdr) {
+		kfree(priv);
+		return ERR_PTR(-ENOMEM);
 	}
+	init_waitqueue_head(&priv->wait_dma);
+	INIT_WORK(&priv->s_work, _qib_do_send);
+	INIT_LIST_HEAD(&priv->iowait);
 
-	switch (init_attr->qp_type) {
-	case IB_QPT_SMI:
-	case IB_QPT_GSI:
-		if (init_attr->port_num == 0 ||
-		    init_attr->port_num > ibpd->device->phys_port_cnt) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-	case IB_QPT_UC:
-	case IB_QPT_RC:
-	case IB_QPT_UD:
-		sz = sizeof(struct qib_sge) *
-			init_attr->cap.max_send_sge +
-			sizeof(struct qib_swqe);
-		swq = __vmalloc((init_attr->cap.max_send_wr + 1) * sz,
-				gfp, PAGE_KERNEL);
-		if (swq == NULL) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail;
-		}
-		sz = sizeof(*qp);
-		sg_list_sz = 0;
-		if (init_attr->srq) {
-			struct qib_srq *srq = to_isrq(init_attr->srq);
-
-			if (srq->rq.max_sge > 1)
-				sg_list_sz = sizeof(*qp->r_sg_list) *
-					(srq->rq.max_sge - 1);
-		} else if (init_attr->cap.max_recv_sge > 1)
-			sg_list_sz = sizeof(*qp->r_sg_list) *
-				(init_attr->cap.max_recv_sge - 1);
-		qp = kzalloc(sz + sg_list_sz, gfp);
-		if (!qp) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_swq;
-		}
-		RCU_INIT_POINTER(qp->next, NULL);
-		qp->s_hdr = kzalloc(sizeof(*qp->s_hdr), gfp);
-		if (!qp->s_hdr) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_qp;
-		}
-		qp->timeout_jiffies =
-			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
-				1000UL);
-		if (init_attr->srq)
-			sz = 0;
-		else {
-			qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
-			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
-			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
-				sizeof(struct qib_rwqe);
-			if (gfp != GFP_NOIO)
-				qp->r_rq.wq = vmalloc_user(
-						sizeof(struct qib_rwq) +
-						qp->r_rq.size * sz);
-			else
-				qp->r_rq.wq = __vmalloc(
-						sizeof(struct qib_rwq) +
-						qp->r_rq.size * sz,
-						gfp, PAGE_KERNEL);
-
-			if (!qp->r_rq.wq) {
-				ret = ERR_PTR(-ENOMEM);
-				goto bail_qp;
-			}
-		}
-
-		/*
-		 * ib_create_qp() will initialize qp->ibqp
-		 * except for qp->ibqp.qp_num.
-		 */
-		spin_lock_init(&qp->r_lock);
-		spin_lock_init(&qp->s_lock);
-		spin_lock_init(&qp->r_rq.lock);
-		atomic_set(&qp->refcount, 0);
-		init_waitqueue_head(&qp->wait);
-		init_waitqueue_head(&qp->wait_dma);
-		init_timer(&qp->s_timer);
-		qp->s_timer.data = (unsigned long)qp;
-		INIT_WORK(&qp->s_work, qib_do_send);
-		INIT_LIST_HEAD(&qp->iowait);
-		INIT_LIST_HEAD(&qp->rspwait);
-		qp->state = IB_QPS_RESET;
-		qp->s_wq = swq;
-		qp->s_size = init_attr->cap.max_send_wr + 1;
-		qp->s_max_sge = init_attr->cap.max_send_sge;
-		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
-			qp->s_flags = QIB_S_SIGNAL_REQ_WR;
-		dev = to_idev(ibpd->device);
-		dd = dd_from_dev(dev);
-		err = alloc_qpn(dd, &dev->qpn_table, init_attr->qp_type,
-				init_attr->port_num, gfp);
-		if (err < 0) {
-			ret = ERR_PTR(err);
-			vfree(qp->r_rq.wq);
-			goto bail_qp;
-		}
-		qp->ibqp.qp_num = err;
-		qp->port_num = init_attr->port_num;
-		qib_reset_qp(qp, init_attr->qp_type);
-		break;
-
-	default:
-		/* Don't support raw QPs */
-		ret = ERR_PTR(-ENOSYS);
-		goto bail;
-	}
-
-	init_attr->cap.max_inline_data = 0;
-
-	/*
-	 * Return the address of the RWQ as the offset to mmap.
-	 * See qib_mmap() for details.
-	 */
-	if (udata && udata->outlen >= sizeof(__u64)) {
-		if (!qp->r_rq.wq) {
-			__u64 offset = 0;
-
-			err = ib_copy_to_udata(udata, &offset,
-					       sizeof(offset));
-			if (err) {
-				ret = ERR_PTR(err);
-				goto bail_ip;
-			}
-		} else {
-			u32 s = sizeof(struct qib_rwq) + qp->r_rq.size * sz;
-
-			qp->ip = qib_create_mmap_info(dev, s,
-						      ibpd->uobject->context,
-						      qp->r_rq.wq);
-			if (!qp->ip) {
-				ret = ERR_PTR(-ENOMEM);
-				goto bail_ip;
-			}
-
-			err = ib_copy_to_udata(udata, &(qp->ip->offset),
-					       sizeof(qp->ip->offset));
-			if (err) {
-				ret = ERR_PTR(err);
-				goto bail_ip;
-			}
-		}
-	}
-
-	spin_lock(&dev->n_qps_lock);
-	if (dev->n_qps_allocated == ib_qib_max_qps) {
-		spin_unlock(&dev->n_qps_lock);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_ip;
-	}
-
-	dev->n_qps_allocated++;
-	spin_unlock(&dev->n_qps_lock);
-
-	if (qp->ip) {
-		spin_lock_irq(&dev->pending_lock);
-		list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-	}
-
-	ret = &qp->ibqp;
-	goto bail;
-
-bail_ip:
-	if (qp->ip)
-		kref_put(&qp->ip->ref, qib_release_mmap_info);
-	else
-		vfree(qp->r_rq.wq);
-	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
-bail_qp:
-	kfree(qp->s_hdr);
-	kfree(qp);
-bail_swq:
-	vfree(swq);
-bail:
-	return ret;
+	return priv;
 }
 
-/**
- * qib_destroy_qp - destroy a queue pair
- * @ibqp: the queue pair to destroy
- *
- * Returns 0 on success.
- *
- * Note that this can be called while the QP is actively sending or
- * receiving!
- */
-int qib_destroy_qp(struct ib_qp *ibqp)
+void qib_qp_priv_free(struct rvt_dev_info *rdi, struct rvt_qp *qp)
 {
-	struct qib_qp *qp = to_iqp(ibqp);
-	struct qib_ibdev *dev = to_idev(ibqp->device);
+	struct qib_qp_priv *priv = qp->priv;
 
-	/* Make sure HW and driver activity is stopped. */
-	spin_lock_irq(&qp->s_lock);
-	if (qp->state != IB_QPS_RESET) {
-		qp->state = IB_QPS_RESET;
-		spin_lock(&dev->pending_lock);
-		if (!list_empty(&qp->iowait))
-			list_del_init(&qp->iowait);
-		spin_unlock(&dev->pending_lock);
-		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
-		spin_unlock_irq(&qp->s_lock);
-		cancel_work_sync(&qp->s_work);
-		del_timer_sync(&qp->s_timer);
-		wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
-		if (qp->s_tx) {
-			qib_put_txreq(qp->s_tx);
-			qp->s_tx = NULL;
-		}
-		remove_qp(dev, qp);
-		wait_event(qp->wait, !atomic_read(&qp->refcount));
-		clear_mr_refs(qp, 1);
-	} else
-		spin_unlock_irq(&qp->s_lock);
+	kfree(priv->s_hdr);
+	kfree(priv);
+}
 
-	/* all user's cleaned up, mark it available */
-	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
-	spin_lock(&dev->n_qps_lock);
-	dev->n_qps_allocated--;
-	spin_unlock(&dev->n_qps_lock);
+void qib_stop_send_queue(struct rvt_qp *qp)
+{
+	struct qib_qp_priv *priv = qp->priv;
 
-	if (qp->ip)
-		kref_put(&qp->ip->ref, qib_release_mmap_info);
-	else
-		vfree(qp->r_rq.wq);
-	vfree(qp->s_wq);
-	kfree(qp->s_hdr);
-	kfree(qp);
-	return 0;
+	cancel_work_sync(&priv->s_work);
+	del_timer_sync(&qp->s_timer);
 }
 
-/**
- * qib_init_qpn_table - initialize the QP number table for a device
- * @qpt: the QPN table
- */
-void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt)
+void qib_quiesce_qp(struct rvt_qp *qp)
 {
-	spin_lock_init(&qpt->lock);
-	qpt->last = 1;          /* start with QPN 2 */
-	qpt->nmaps = 1;
-	qpt->mask = dd->qpn_mask;
+	struct qib_qp_priv *priv = qp->priv;
+
+	wait_event(priv->wait_dma, !atomic_read(&priv->s_dma_busy));
+	if (priv->s_tx) {
+		qib_put_txreq(priv->s_tx);
+		priv->s_tx = NULL;
+	}
 }
 
-/**
- * qib_free_qpn_table - free the QP number table for a device
- * @qpt: the QPN table
- */
-void qib_free_qpn_table(struct qib_qpn_table *qpt)
+void qib_flush_qp_waiters(struct rvt_qp *qp)
 {
-	int i;
+	struct qib_qp_priv *priv = qp->priv;
+	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
 
-	for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
-		if (qpt->map[i].page)
-			free_page((unsigned long) qpt->map[i].page);
+	spin_lock(&dev->rdi.pending_lock);
+	if (!list_empty(&priv->iowait))
+		list_del_init(&priv->iowait);
+	spin_unlock(&dev->rdi.pending_lock);
 }
 
 /**
@@ -1280,7 +444,7 @@ void qib_free_qpn_table(struct qib_qpn_table *qpt)
  *
  * The QP s_lock should be held.
  */
-void qib_get_credit(struct qib_qp *qp, u32 aeth)
+void qib_get_credit(struct rvt_qp *qp, u32 aeth)
 {
 	u32 credit = (aeth >> QIB_AETH_CREDIT_SHIFT) & QIB_AETH_CREDIT_MASK;
 
@@ -1290,31 +454,70 @@ void qib_get_credit(struct qib_qp *qp, u32 aeth)
 	 * honor the credit field.
 	 */
 	if (credit == QIB_AETH_CREDIT_INVAL) {
-		if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
-			qp->s_flags |= QIB_S_UNLIMITED_CREDIT;
-			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
-				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
+		if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
+			qp->s_flags |= RVT_S_UNLIMITED_CREDIT;
+			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
 				qib_schedule_send(qp);
 			}
 		}
-	} else if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
+	} else if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
 		/* Compute new LSN (i.e., MSN + credit) */
 		credit = (aeth + credit_table[credit]) & QIB_MSN_MASK;
 		if (qib_cmp24(credit, qp->s_lsn) > 0) {
 			qp->s_lsn = credit;
-			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
-				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
+			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
 				qib_schedule_send(qp);
 			}
 		}
 	}
 }
 
+/**
+ * qib_check_send_wqe - validate wr/wqe
+ * @qp - The qp
+ * @wqe - The built wqe
+ *
+ * validate wr/wqe.  This is called
+ * prior to inserting the wqe into
+ * the ring but after the wqe has been
+ * setup.
+ *
+ * Returns 1 to force direct progress, 0 otherwise, -EINVAL on failure
+ */
+int qib_check_send_wqe(struct rvt_qp *qp,
+		       struct rvt_swqe *wqe)
+{
+	struct rvt_ah *ah;
+	int ret = 0;
+
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+		if (wqe->length > 0x80000000U)
+			return -EINVAL;
+		break;
+	case IB_QPT_SMI:
+	case IB_QPT_GSI:
+	case IB_QPT_UD:
+		ah = ibah_to_rvtah(wqe->ud_wr.ah);
+		if (wqe->length > (1 << ah->log_pmtu))
+			return -EINVAL;
+		/* progress hint */
+		ret = 1;
+		break;
+	default:
+		break;
+	}
+	return ret;
+}
+
 #ifdef CONFIG_DEBUG_FS
 
 struct qib_qp_iter {
 	struct qib_ibdev *dev;
-	struct qib_qp *qp;
+	struct rvt_qp *qp;
 	int n;
 };
 
@@ -1340,14 +543,14 @@ int qib_qp_iter_next(struct qib_qp_iter *iter)
 	struct qib_ibdev *dev = iter->dev;
 	int n = iter->n;
 	int ret = 1;
-	struct qib_qp *pqp = iter->qp;
-	struct qib_qp *qp;
+	struct rvt_qp *pqp = iter->qp;
+	struct rvt_qp *qp;
 
-	for (; n < dev->qp_table_size; n++) {
+	for (; n < dev->rdi.qp_dev->qp_table_size; n++) {
 		if (pqp)
 			qp = rcu_dereference(pqp->next);
 		else
-			qp = rcu_dereference(dev->qp_table[n]);
+			qp = rcu_dereference(dev->rdi.qp_dev->qp_table[n]);
 		pqp = qp;
 		if (qp) {
 			iter->qp = qp;
@@ -1364,10 +567,11 @@ static const char * const qp_type_str[] = {
 
 void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter)
 {
-	struct qib_swqe *wqe;
-	struct qib_qp *qp = iter->qp;
+	struct rvt_swqe *wqe;
+	struct rvt_qp *qp = iter->qp;
+	struct qib_qp_priv *priv = qp->priv;
 
-	wqe = get_swqe_ptr(qp, qp->s_last);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 	seq_printf(s,
 		   "N %d QP%u %s %u %u %u f=%x %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u) QP%u LID %x\n",
 		   iter->n,
@@ -1377,8 +581,8 @@ void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter)
 		   wqe->wr.opcode,
 		   qp->s_hdrwords,
 		   qp->s_flags,
-		   atomic_read(&qp->s_dma_busy),
-		   !list_empty(&qp->iowait),
+		   atomic_read(&priv->s_dma_busy),
+		   !list_empty(&priv->iowait),
 		   qp->timeout,
 		   wqe->ssn,
 		   qp->s_lsn,
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index e6b7556d5221..9088e26d3ac8 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -40,7 +40,7 @@
 
 static void rc_timeout(unsigned long arg);
 
-static u32 restart_sge(struct qib_sge_state *ss, struct qib_swqe *wqe,
+static u32 restart_sge(struct rvt_sge_state *ss, struct rvt_swqe *wqe,
 		       u32 psn, u32 pmtu)
 {
 	u32 len;
@@ -54,9 +54,9 @@ static u32 restart_sge(struct qib_sge_state *ss, struct qib_swqe *wqe,
 	return wqe->length - len;
 }
 
-static void start_timer(struct qib_qp *qp)
+static void start_timer(struct rvt_qp *qp)
 {
-	qp->s_flags |= QIB_S_TIMER;
+	qp->s_flags |= RVT_S_TIMER;
 	qp->s_timer.function = rc_timeout;
 	/* 4.096 usec. * (1 << qp->timeout) */
 	qp->s_timer.expires = jiffies + qp->timeout_jiffies;
@@ -74,17 +74,17 @@ static void start_timer(struct qib_qp *qp)
  * Note that we are in the responder's side of the QP context.
  * Note the QP s_lock must be held.
  */
-static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
+static int qib_make_rc_ack(struct qib_ibdev *dev, struct rvt_qp *qp,
 			   struct qib_other_headers *ohdr, u32 pmtu)
 {
-	struct qib_ack_entry *e;
+	struct rvt_ack_entry *e;
 	u32 hwords;
 	u32 len;
 	u32 bth0;
 	u32 bth2;
 
 	/* Don't send an ACK if we aren't supposed to. */
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
 		goto bail;
 
 	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
@@ -95,7 +95,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
 	case OP(RDMA_READ_RESPONSE_ONLY):
 		e = &qp->s_ack_queue[qp->s_tail_ack_queue];
 		if (e->rdma_sge.mr) {
-			qib_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		/* FALLTHROUGH */
@@ -112,7 +112,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
 	case OP(ACKNOWLEDGE):
 		/* Check for no next entry in the queue. */
 		if (qp->r_head_ack_queue == qp->s_tail_ack_queue) {
-			if (qp->s_flags & QIB_S_ACK_PENDING)
+			if (qp->s_flags & RVT_S_ACK_PENDING)
 				goto normal;
 			goto bail;
 		}
@@ -133,7 +133,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
 			/* Copy SGE state in case we need to resend */
 			qp->s_rdma_mr = e->rdma_sge.mr;
 			if (qp->s_rdma_mr)
-				qib_get_mr(qp->s_rdma_mr);
+				rvt_get_mr(qp->s_rdma_mr);
 			qp->s_ack_rdma_sge.sge = e->rdma_sge;
 			qp->s_ack_rdma_sge.num_sge = 1;
 			qp->s_cur_sge = &qp->s_ack_rdma_sge;
@@ -172,7 +172,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
 		qp->s_cur_sge = &qp->s_ack_rdma_sge;
 		qp->s_rdma_mr = qp->s_ack_rdma_sge.sge.mr;
 		if (qp->s_rdma_mr)
-			qib_get_mr(qp->s_rdma_mr);
+			rvt_get_mr(qp->s_rdma_mr);
 		len = qp->s_ack_rdma_sge.sge.sge_length;
 		if (len > pmtu)
 			len = pmtu;
@@ -196,7 +196,7 @@ normal:
 		 * (see above).
 		 */
 		qp->s_ack_state = OP(SEND_ONLY);
-		qp->s_flags &= ~QIB_S_ACK_PENDING;
+		qp->s_flags &= ~RVT_S_ACK_PENDING;
 		qp->s_cur_sge = NULL;
 		if (qp->s_nak_state)
 			ohdr->u.aeth =
@@ -218,7 +218,7 @@ normal:
 
 bail:
 	qp->s_ack_state = OP(ACKNOWLEDGE);
-	qp->s_flags &= ~(QIB_S_RESP_PENDING | QIB_S_ACK_PENDING);
+	qp->s_flags &= ~(RVT_S_RESP_PENDING | RVT_S_ACK_PENDING);
 	return 0;
 }
 
@@ -226,63 +226,60 @@ bail:
  * qib_make_rc_req - construct a request packet (SEND, RDMA r/w, ATOMIC)
  * @qp: a pointer to the QP
  *
+ * Assumes the s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_rc_req(struct qib_qp *qp)
+int qib_make_rc_req(struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
 	struct qib_other_headers *ohdr;
-	struct qib_sge_state *ss;
-	struct qib_swqe *wqe;
+	struct rvt_sge_state *ss;
+	struct rvt_swqe *wqe;
 	u32 hwords;
 	u32 len;
 	u32 bth0;
 	u32 bth2;
 	u32 pmtu = qp->pmtu;
 	char newreq;
-	unsigned long flags;
 	int ret = 0;
 	int delta;
 
-	ohdr = &qp->s_hdr->u.oth;
+	ohdr = &priv->s_hdr->u.oth;
 	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
-		ohdr = &qp->s_hdr->u.l.oth;
-
-	/*
-	 * The lock is needed to synchronize between the sending tasklet,
-	 * the receive interrupt handler, and timeout resends.
-	 */
-	spin_lock_irqsave(&qp->s_lock, flags);
+		ohdr = &priv->s_hdr->u.l.oth;
 
 	/* Sending responses has higher priority over sending requests. */
-	if ((qp->s_flags & QIB_S_RESP_PENDING) &&
+	if ((qp->s_flags & RVT_S_RESP_PENDING) &&
 	    qib_make_rc_ack(dev, qp, ohdr, pmtu))
 		goto done;
 
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_SEND_OK)) {
-		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_dma_busy)) {
-			qp->s_flags |= QIB_S_WAIT_DMA;
+		if (atomic_read(&priv->s_dma_busy)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		qib_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
 			IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR);
 		/* will get called again */
 		goto done;
 	}
 
-	if (qp->s_flags & (QIB_S_WAIT_RNR | QIB_S_WAIT_ACK))
+	if (qp->s_flags & (RVT_S_WAIT_RNR | RVT_S_WAIT_ACK))
 		goto bail;
 
 	if (qib_cmp24(qp->s_psn, qp->s_sending_hpsn) <= 0) {
 		if (qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0) {
-			qp->s_flags |= QIB_S_WAIT_PSN;
+			qp->s_flags |= RVT_S_WAIT_PSN;
 			goto bail;
 		}
 		qp->s_sending_psn = qp->s_psn;
@@ -294,10 +291,10 @@ int qib_make_rc_req(struct qib_qp *qp)
 	bth0 = 0;
 
 	/* Send a request. */
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	switch (qp->s_state) {
 	default:
-		if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_NEXT_SEND_OK))
+		if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_NEXT_SEND_OK))
 			goto bail;
 		/*
 		 * Resend an old request or start a new one.
@@ -317,11 +314,11 @@ int qib_make_rc_req(struct qib_qp *qp)
 			 */
 			if ((wqe->wr.send_flags & IB_SEND_FENCE) &&
 			    qp->s_num_rd_atomic) {
-				qp->s_flags |= QIB_S_WAIT_FENCE;
+				qp->s_flags |= RVT_S_WAIT_FENCE;
 				goto bail;
 			}
-			wqe->psn = qp->s_next_psn;
 			newreq = 1;
+			qp->s_psn = wqe->psn;
 		}
 		/*
 		 * Note that we have to be careful not to modify the
@@ -335,14 +332,12 @@ int qib_make_rc_req(struct qib_qp *qp)
 		case IB_WR_SEND:
 		case IB_WR_SEND_WITH_IMM:
 			/* If no credit, return. */
-			if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT) &&
+			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
 			    qib_cmp24(wqe->ssn, qp->s_lsn + 1) > 0) {
-				qp->s_flags |= QIB_S_WAIT_SSN_CREDIT;
+				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
-			wqe->lpsn = wqe->psn;
 			if (len > pmtu) {
-				wqe->lpsn += (len - 1) / pmtu;
 				qp->s_state = OP(SEND_FIRST);
 				len = pmtu;
 				break;
@@ -363,14 +358,14 @@ int qib_make_rc_req(struct qib_qp *qp)
 			break;
 
 		case IB_WR_RDMA_WRITE:
-			if (newreq && !(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+			if (newreq && !(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 				qp->s_lsn++;
 			/* FALLTHROUGH */
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			/* If no credit, return. */
-			if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT) &&
+			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
 			    qib_cmp24(wqe->ssn, qp->s_lsn + 1) > 0) {
-				qp->s_flags |= QIB_S_WAIT_SSN_CREDIT;
+				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
 
@@ -380,9 +375,7 @@ int qib_make_rc_req(struct qib_qp *qp)
 				cpu_to_be32(wqe->rdma_wr.rkey);
 			ohdr->u.rc.reth.length = cpu_to_be32(len);
 			hwords += sizeof(struct ib_reth) / sizeof(u32);
-			wqe->lpsn = wqe->psn;
 			if (len > pmtu) {
-				wqe->lpsn += (len - 1) / pmtu;
 				qp->s_state = OP(RDMA_WRITE_FIRST);
 				len = pmtu;
 				break;
@@ -411,19 +404,12 @@ int qib_make_rc_req(struct qib_qp *qp)
 			if (newreq) {
 				if (qp->s_num_rd_atomic >=
 				    qp->s_max_rd_atomic) {
-					qp->s_flags |= QIB_S_WAIT_RDMAR;
+					qp->s_flags |= RVT_S_WAIT_RDMAR;
 					goto bail;
 				}
 				qp->s_num_rd_atomic++;
-				if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+				if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 					qp->s_lsn++;
-				/*
-				 * Adjust s_next_psn to count the
-				 * expected number of responses.
-				 */
-				if (len > pmtu)
-					qp->s_next_psn += (len - 1) / pmtu;
-				wqe->lpsn = qp->s_next_psn++;
 			}
 
 			ohdr->u.rc.reth.vaddr =
@@ -449,13 +435,12 @@ int qib_make_rc_req(struct qib_qp *qp)
 			if (newreq) {
 				if (qp->s_num_rd_atomic >=
 				    qp->s_max_rd_atomic) {
-					qp->s_flags |= QIB_S_WAIT_RDMAR;
+					qp->s_flags |= RVT_S_WAIT_RDMAR;
 					goto bail;
 				}
 				qp->s_num_rd_atomic++;
-				if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+				if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 					qp->s_lsn++;
-				wqe->lpsn = wqe->psn;
 			}
 			if (wqe->atomic_wr.wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP) {
 				qp->s_state = OP(COMPARE_SWAP);
@@ -498,11 +483,8 @@ int qib_make_rc_req(struct qib_qp *qp)
 		}
 		if (wqe->wr.opcode == IB_WR_RDMA_READ)
 			qp->s_psn = wqe->lpsn + 1;
-		else {
+		else
 			qp->s_psn++;
-			if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
-				qp->s_next_psn = qp->s_psn;
-		}
 		break;
 
 	case OP(RDMA_READ_RESPONSE_FIRST):
@@ -522,8 +504,6 @@ int qib_make_rc_req(struct qib_qp *qp)
 		/* FALLTHROUGH */
 	case OP(SEND_MIDDLE):
 		bth2 = qp->s_psn++ & QIB_PSN_MASK;
-		if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
-			qp->s_next_psn = qp->s_psn;
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
@@ -563,8 +543,6 @@ int qib_make_rc_req(struct qib_qp *qp)
 		/* FALLTHROUGH */
 	case OP(RDMA_WRITE_MIDDLE):
 		bth2 = qp->s_psn++ & QIB_PSN_MASK;
-		if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
-			qp->s_next_psn = qp->s_psn;
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
@@ -618,9 +596,9 @@ int qib_make_rc_req(struct qib_qp *qp)
 	delta = (((int) bth2 - (int) wqe->psn) << 8) >> 8;
 	if (delta && delta % QIB_PSN_CREDIT == 0)
 		bth2 |= IB_BTH_REQ_ACK;
-	if (qp->s_flags & QIB_S_SEND_ONE) {
-		qp->s_flags &= ~QIB_S_SEND_ONE;
-		qp->s_flags |= QIB_S_WAIT_ACK;
+	if (qp->s_flags & RVT_S_SEND_ONE) {
+		qp->s_flags &= ~RVT_S_SEND_ONE;
+		qp->s_flags |= RVT_S_WAIT_ACK;
 		bth2 |= IB_BTH_REQ_ACK;
 	}
 	qp->s_len -= len;
@@ -629,13 +607,9 @@ int qib_make_rc_req(struct qib_qp *qp)
 	qp->s_cur_size = len;
 	qib_make_ruc_header(qp, ohdr, bth0 | (qp->s_state << 24), bth2);
 done:
-	ret = 1;
-	goto unlock;
-
+	return 1;
 bail:
-	qp->s_flags &= ~QIB_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	qp->s_flags &= ~RVT_S_BUSY;
 	return ret;
 }
 
@@ -647,7 +621,7 @@ unlock:
  * Note that RDMA reads and atomics are handled in the
  * send side QP state and tasklet.
  */
-void qib_send_rc_ack(struct qib_qp *qp)
+void qib_send_rc_ack(struct rvt_qp *qp)
 {
 	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
@@ -665,11 +639,11 @@ void qib_send_rc_ack(struct qib_qp *qp)
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
 		goto unlock;
 
 	/* Don't send ACK or NAK if a RDMA read or atomic is pending. */
-	if ((qp->s_flags & QIB_S_RESP_PENDING) || qp->s_rdma_ack_cnt)
+	if ((qp->s_flags & RVT_S_RESP_PENDING) || qp->s_rdma_ack_cnt)
 		goto queue_ack;
 
 	/* Construct the header with s_lock held so APM doesn't change it. */
@@ -758,9 +732,9 @@ void qib_send_rc_ack(struct qib_qp *qp)
 	goto done;
 
 queue_ack:
-	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
-		ibp->n_rc_qacks++;
-		qp->s_flags |= QIB_S_ACK_PENDING | QIB_S_RESP_PENDING;
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
+		this_cpu_inc(*ibp->rvp.rc_qacks);
+		qp->s_flags |= RVT_S_ACK_PENDING | RVT_S_RESP_PENDING;
 		qp->s_nak_state = qp->r_nak_state;
 		qp->s_ack_psn = qp->r_ack_psn;
 
@@ -782,10 +756,10 @@ done:
  * for the given QP.
  * Called at interrupt level with the QP s_lock held.
  */
-static void reset_psn(struct qib_qp *qp, u32 psn)
+static void reset_psn(struct rvt_qp *qp, u32 psn)
 {
 	u32 n = qp->s_acked;
-	struct qib_swqe *wqe = get_swqe_ptr(qp, n);
+	struct rvt_swqe *wqe = rvt_get_swqe_ptr(qp, n);
 	u32 opcode;
 
 	qp->s_cur = n;
@@ -808,7 +782,7 @@ static void reset_psn(struct qib_qp *qp, u32 psn)
 			n = 0;
 		if (n == qp->s_tail)
 			break;
-		wqe = get_swqe_ptr(qp, n);
+		wqe = rvt_get_swqe_ptr(qp, n);
 		diff = qib_cmp24(psn, wqe->psn);
 		if (diff < 0)
 			break;
@@ -854,22 +828,22 @@ static void reset_psn(struct qib_qp *qp, u32 psn)
 done:
 	qp->s_psn = psn;
 	/*
-	 * Set QIB_S_WAIT_PSN as qib_rc_complete() may start the timer
+	 * Set RVT_S_WAIT_PSN as qib_rc_complete() may start the timer
 	 * asynchronously before the send tasklet can get scheduled.
 	 * Doing it in qib_make_rc_req() is too late.
 	 */
 	if ((qib_cmp24(qp->s_psn, qp->s_sending_hpsn) <= 0) &&
 	    (qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0))
-		qp->s_flags |= QIB_S_WAIT_PSN;
+		qp->s_flags |= RVT_S_WAIT_PSN;
 }
 
 /*
  * Back up requester to resend the last un-ACKed request.
  * The QP r_lock and s_lock should be held and interrupts disabled.
  */
-static void qib_restart_rc(struct qib_qp *qp, u32 psn, int wait)
+static void qib_restart_rc(struct rvt_qp *qp, u32 psn, int wait)
 {
-	struct qib_swqe *wqe = get_swqe_ptr(qp, qp->s_acked);
+	struct rvt_swqe *wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	struct qib_ibport *ibp;
 
 	if (qp->s_retry == 0) {
@@ -878,7 +852,7 @@ static void qib_restart_rc(struct qib_qp *qp, u32 psn, int wait)
 			qp->s_retry = qp->s_retry_cnt;
 		} else if (qp->s_last == qp->s_acked) {
 			qib_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
-			qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+			rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			return;
 		} else /* XXX need to handle delayed completion */
 			return;
@@ -887,15 +861,15 @@ static void qib_restart_rc(struct qib_qp *qp, u32 psn, int wait)
 
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	if (wqe->wr.opcode == IB_WR_RDMA_READ)
-		ibp->n_rc_resends++;
+		ibp->rvp.n_rc_resends++;
 	else
-		ibp->n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
+		ibp->rvp.n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
 
-	qp->s_flags &= ~(QIB_S_WAIT_FENCE | QIB_S_WAIT_RDMAR |
-			 QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_PSN |
-			 QIB_S_WAIT_ACK);
+	qp->s_flags &= ~(RVT_S_WAIT_FENCE | RVT_S_WAIT_RDMAR |
+			 RVT_S_WAIT_SSN_CREDIT | RVT_S_WAIT_PSN |
+			 RVT_S_WAIT_ACK);
 	if (wait)
-		qp->s_flags |= QIB_S_SEND_ONE;
+		qp->s_flags |= RVT_S_SEND_ONE;
 	reset_psn(qp, psn);
 }
 
@@ -904,16 +878,16 @@ static void qib_restart_rc(struct qib_qp *qp, u32 psn, int wait)
  */
 static void rc_timeout(unsigned long arg)
 {
-	struct qib_qp *qp = (struct qib_qp *)arg;
+	struct rvt_qp *qp = (struct rvt_qp *)arg;
 	struct qib_ibport *ibp;
 	unsigned long flags;
 
 	spin_lock_irqsave(&qp->r_lock, flags);
 	spin_lock(&qp->s_lock);
-	if (qp->s_flags & QIB_S_TIMER) {
+	if (qp->s_flags & RVT_S_TIMER) {
 		ibp = to_iport(qp->ibqp.device, qp->port_num);
-		ibp->n_rc_timeouts++;
-		qp->s_flags &= ~QIB_S_TIMER;
+		ibp->rvp.n_rc_timeouts++;
+		qp->s_flags &= ~RVT_S_TIMER;
 		del_timer(&qp->s_timer);
 		qib_restart_rc(qp, qp->s_last_psn + 1, 1);
 		qib_schedule_send(qp);
@@ -927,12 +901,12 @@ static void rc_timeout(unsigned long arg)
  */
 void qib_rc_rnr_retry(unsigned long arg)
 {
-	struct qib_qp *qp = (struct qib_qp *)arg;
+	struct rvt_qp *qp = (struct rvt_qp *)arg;
 	unsigned long flags;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (qp->s_flags & QIB_S_WAIT_RNR) {
-		qp->s_flags &= ~QIB_S_WAIT_RNR;
+	if (qp->s_flags & RVT_S_WAIT_RNR) {
+		qp->s_flags &= ~RVT_S_WAIT_RNR;
 		del_timer(&qp->s_timer);
 		qib_schedule_send(qp);
 	}
@@ -943,14 +917,14 @@ void qib_rc_rnr_retry(unsigned long arg)
  * Set qp->s_sending_psn to the next PSN after the given one.
  * This would be psn+1 except when RDMA reads are present.
  */
-static void reset_sending_psn(struct qib_qp *qp, u32 psn)
+static void reset_sending_psn(struct rvt_qp *qp, u32 psn)
 {
-	struct qib_swqe *wqe;
+	struct rvt_swqe *wqe;
 	u32 n = qp->s_last;
 
 	/* Find the work request corresponding to the given PSN. */
 	for (;;) {
-		wqe = get_swqe_ptr(qp, n);
+		wqe = rvt_get_swqe_ptr(qp, n);
 		if (qib_cmp24(psn, wqe->lpsn) <= 0) {
 			if (wqe->wr.opcode == IB_WR_RDMA_READ)
 				qp->s_sending_psn = wqe->lpsn + 1;
@@ -968,16 +942,16 @@ static void reset_sending_psn(struct qib_qp *qp, u32 psn)
 /*
  * This should be called with the QP s_lock held and interrupts disabled.
  */
-void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr)
+void qib_rc_send_complete(struct rvt_qp *qp, struct qib_ib_header *hdr)
 {
 	struct qib_other_headers *ohdr;
-	struct qib_swqe *wqe;
+	struct rvt_swqe *wqe;
 	struct ib_wc wc;
 	unsigned i;
 	u32 opcode;
 	u32 psn;
 
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		return;
 
 	/* Find out where the BTH is */
@@ -1002,22 +976,30 @@ void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr)
 	 * there are still requests that haven't been acked.
 	 */
 	if ((psn & IB_BTH_REQ_ACK) && qp->s_acked != qp->s_tail &&
-	    !(qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR | QIB_S_WAIT_PSN)) &&
-	    (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+	    !(qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR | RVT_S_WAIT_PSN)) &&
+	    (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
 		start_timer(qp);
 
 	while (qp->s_last != qp->s_acked) {
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		u32 s_last;
+
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		if (qib_cmp24(wqe->lpsn, qp->s_sending_psn) >= 0 &&
 		    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0)
 			break;
+		s_last = qp->s_last;
+		if (++s_last >= qp->s_size)
+			s_last = 0;
+		qp->s_last = s_last;
+		/* see post_send() */
+		barrier();
 		for (i = 0; i < wqe->wr.num_sge; i++) {
-			struct qib_sge *sge = &wqe->sg_list[i];
+			struct rvt_sge *sge = &wqe->sg_list[i];
 
-			qib_put_mr(sge->mr);
+			rvt_put_mr(sge->mr);
 		}
 		/* Post a send completion queue entry if requested. */
-		if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+		if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
 			memset(&wc, 0, sizeof(wc));
 			wc.wr_id = wqe->wr.wr_id;
@@ -1025,25 +1007,23 @@ void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr)
 			wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode];
 			wc.byte_len = wqe->length;
 			wc.qp = &qp->ibqp;
-			qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc, 0);
 		}
-		if (++qp->s_last >= qp->s_size)
-			qp->s_last = 0;
 	}
 	/*
 	 * If we were waiting for sends to complete before resending,
 	 * and they are now complete, restart sending.
 	 */
-	if (qp->s_flags & QIB_S_WAIT_PSN &&
+	if (qp->s_flags & RVT_S_WAIT_PSN &&
 	    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
-		qp->s_flags &= ~QIB_S_WAIT_PSN;
+		qp->s_flags &= ~RVT_S_WAIT_PSN;
 		qp->s_sending_psn = qp->s_psn;
 		qp->s_sending_hpsn = qp->s_psn - 1;
 		qib_schedule_send(qp);
 	}
 }
 
-static inline void update_last_psn(struct qib_qp *qp, u32 psn)
+static inline void update_last_psn(struct rvt_qp *qp, u32 psn)
 {
 	qp->s_last_psn = psn;
 }
@@ -1053,8 +1033,8 @@ static inline void update_last_psn(struct qib_qp *qp, u32 psn)
  * This is similar to qib_send_complete but has to check to be sure
  * that the SGEs are not being referenced if the SWQE is being resent.
  */
-static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
-					 struct qib_swqe *wqe,
+static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp,
+					 struct rvt_swqe *wqe,
 					 struct qib_ibport *ibp)
 {
 	struct ib_wc wc;
@@ -1067,13 +1047,21 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
 	 */
 	if (qib_cmp24(wqe->lpsn, qp->s_sending_psn) < 0 ||
 	    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
+		u32 s_last;
+
 		for (i = 0; i < wqe->wr.num_sge; i++) {
-			struct qib_sge *sge = &wqe->sg_list[i];
+			struct rvt_sge *sge = &wqe->sg_list[i];
 
-			qib_put_mr(sge->mr);
+			rvt_put_mr(sge->mr);
 		}
+		s_last = qp->s_last;
+		if (++s_last >= qp->s_size)
+			s_last = 0;
+		qp->s_last = s_last;
+		/* see post_send() */
+		barrier();
 		/* Post a send completion queue entry if requested. */
-		if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+		if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
 			memset(&wc, 0, sizeof(wc));
 			wc.wr_id = wqe->wr.wr_id;
@@ -1081,12 +1069,10 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
 			wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode];
 			wc.byte_len = wqe->length;
 			wc.qp = &qp->ibqp;
-			qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc, 0);
 		}
-		if (++qp->s_last >= qp->s_size)
-			qp->s_last = 0;
 	} else
-		ibp->n_rc_delayed_comp++;
+		this_cpu_inc(*ibp->rvp.rc_delayed_comp);
 
 	qp->s_retry = qp->s_retry_cnt;
 	update_last_psn(qp, wqe->lpsn);
@@ -1100,7 +1086,7 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
 		if (++qp->s_cur >= qp->s_size)
 			qp->s_cur = 0;
 		qp->s_acked = qp->s_cur;
-		wqe = get_swqe_ptr(qp, qp->s_cur);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 		if (qp->s_acked != qp->s_tail) {
 			qp->s_state = OP(SEND_LAST);
 			qp->s_psn = wqe->psn;
@@ -1110,7 +1096,7 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
 			qp->s_acked = 0;
 		if (qp->state == IB_QPS_SQD && qp->s_acked == qp->s_cur)
 			qp->s_draining = 0;
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	}
 	return wqe;
 }
@@ -1126,19 +1112,19 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
  * Called at interrupt level with the QP s_lock held.
  * Returns 1 if OK, 0 if current operation should be aborted (NAK).
  */
-static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
+static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
 		     u64 val, struct qib_ctxtdata *rcd)
 {
 	struct qib_ibport *ibp;
 	enum ib_wc_status status;
-	struct qib_swqe *wqe;
+	struct rvt_swqe *wqe;
 	int ret = 0;
 	u32 ack_psn;
 	int diff;
 
 	/* Remove QP from retry timer */
-	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
-		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
+	if (qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR)) {
+		qp->s_flags &= ~(RVT_S_TIMER | RVT_S_WAIT_RNR);
 		del_timer(&qp->s_timer);
 	}
 
@@ -1151,7 +1137,7 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 	ack_psn = psn;
 	if (aeth >> 29)
 		ack_psn--;
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 
 	/*
@@ -1186,11 +1172,11 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 		      wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) &&
 		     (opcode != OP(ATOMIC_ACKNOWLEDGE) || diff != 0))) {
 			/* Retry this request. */
-			if (!(qp->r_flags & QIB_R_RDMAR_SEQ)) {
-				qp->r_flags |= QIB_R_RDMAR_SEQ;
+			if (!(qp->r_flags & RVT_R_RDMAR_SEQ)) {
+				qp->r_flags |= RVT_R_RDMAR_SEQ;
 				qib_restart_rc(qp, qp->s_last_psn + 1, 0);
 				if (list_empty(&qp->rspwait)) {
-					qp->r_flags |= QIB_R_RSP_SEND;
+					qp->r_flags |= RVT_R_RSP_SEND;
 					atomic_inc(&qp->refcount);
 					list_add_tail(&qp->rspwait,
 						      &rcd->qp_wait_list);
@@ -1213,14 +1199,14 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 		     wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD)) {
 			qp->s_num_rd_atomic--;
 			/* Restart sending task if fence is complete */
-			if ((qp->s_flags & QIB_S_WAIT_FENCE) &&
+			if ((qp->s_flags & RVT_S_WAIT_FENCE) &&
 			    !qp->s_num_rd_atomic) {
-				qp->s_flags &= ~(QIB_S_WAIT_FENCE |
-						 QIB_S_WAIT_ACK);
+				qp->s_flags &= ~(RVT_S_WAIT_FENCE |
+						 RVT_S_WAIT_ACK);
 				qib_schedule_send(qp);
-			} else if (qp->s_flags & QIB_S_WAIT_RDMAR) {
-				qp->s_flags &= ~(QIB_S_WAIT_RDMAR |
-						 QIB_S_WAIT_ACK);
+			} else if (qp->s_flags & RVT_S_WAIT_RDMAR) {
+				qp->s_flags &= ~(RVT_S_WAIT_RDMAR |
+						 RVT_S_WAIT_ACK);
 				qib_schedule_send(qp);
 			}
 		}
@@ -1231,7 +1217,7 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 
 	switch (aeth >> 29) {
 	case 0:         /* ACK */
-		ibp->n_rc_acks++;
+		this_cpu_inc(*ibp->rvp.rc_acks);
 		if (qp->s_acked != qp->s_tail) {
 			/*
 			 * We are expecting more ACKs so
@@ -1248,8 +1234,8 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 			qp->s_state = OP(SEND_LAST);
 			qp->s_psn = psn + 1;
 		}
-		if (qp->s_flags & QIB_S_WAIT_ACK) {
-			qp->s_flags &= ~QIB_S_WAIT_ACK;
+		if (qp->s_flags & RVT_S_WAIT_ACK) {
+			qp->s_flags &= ~RVT_S_WAIT_ACK;
 			qib_schedule_send(qp);
 		}
 		qib_get_credit(qp, aeth);
@@ -1260,10 +1246,10 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 		goto bail;
 
 	case 1:         /* RNR NAK */
-		ibp->n_rnr_naks++;
+		ibp->rvp.n_rnr_naks++;
 		if (qp->s_acked == qp->s_tail)
 			goto bail;
-		if (qp->s_flags & QIB_S_WAIT_RNR)
+		if (qp->s_flags & RVT_S_WAIT_RNR)
 			goto bail;
 		if (qp->s_rnr_retry == 0) {
 			status = IB_WC_RNR_RETRY_EXC_ERR;
@@ -1275,12 +1261,12 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 		/* The last valid PSN is the previous PSN. */
 		update_last_psn(qp, psn - 1);
 
-		ibp->n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
+		ibp->rvp.n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
 
 		reset_psn(qp, psn);
 
-		qp->s_flags &= ~(QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_ACK);
-		qp->s_flags |= QIB_S_WAIT_RNR;
+		qp->s_flags &= ~(RVT_S_WAIT_SSN_CREDIT | RVT_S_WAIT_ACK);
+		qp->s_flags |= RVT_S_WAIT_RNR;
 		qp->s_timer.function = qib_rc_rnr_retry;
 		qp->s_timer.expires = jiffies + usecs_to_jiffies(
 			ib_qib_rnr_table[(aeth >> QIB_AETH_CREDIT_SHIFT) &
@@ -1296,7 +1282,7 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 		switch ((aeth >> QIB_AETH_CREDIT_SHIFT) &
 			QIB_AETH_CREDIT_MASK) {
 		case 0: /* PSN sequence error */
-			ibp->n_seq_naks++;
+			ibp->rvp.n_seq_naks++;
 			/*
 			 * Back up to the responder's expected PSN.
 			 * Note that we might get a NAK in the middle of an
@@ -1309,21 +1295,21 @@ static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
 
 		case 1: /* Invalid Request */
 			status = IB_WC_REM_INV_REQ_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 			goto class_b;
 
 		case 2: /* Remote Access Error */
 			status = IB_WC_REM_ACCESS_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 			goto class_b;
 
 		case 3: /* Remote Operation Error */
 			status = IB_WC_REM_OP_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 class_b:
 			if (qp->s_last == qp->s_acked) {
 				qib_send_complete(qp, wqe, status);
-				qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+				rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			}
 			break;
 
@@ -1349,18 +1335,18 @@ bail:
  * We have seen an out of sequence RDMA read middle or last packet.
  * This ACKs SENDs and RDMA writes up to the first RDMA read or atomic SWQE.
  */
-static void rdma_seq_err(struct qib_qp *qp, struct qib_ibport *ibp, u32 psn,
+static void rdma_seq_err(struct rvt_qp *qp, struct qib_ibport *ibp, u32 psn,
 			 struct qib_ctxtdata *rcd)
 {
-	struct qib_swqe *wqe;
+	struct rvt_swqe *wqe;
 
 	/* Remove QP from retry timer */
-	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
-		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
+	if (qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR)) {
+		qp->s_flags &= ~(RVT_S_TIMER | RVT_S_WAIT_RNR);
 		del_timer(&qp->s_timer);
 	}
 
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 
 	while (qib_cmp24(psn, wqe->lpsn) > 0) {
 		if (wqe->wr.opcode == IB_WR_RDMA_READ ||
@@ -1370,11 +1356,11 @@ static void rdma_seq_err(struct qib_qp *qp, struct qib_ibport *ibp, u32 psn,
 		wqe = do_rc_completion(qp, wqe, ibp);
 	}
 
-	ibp->n_rdma_seq++;
-	qp->r_flags |= QIB_R_RDMAR_SEQ;
+	ibp->rvp.n_rdma_seq++;
+	qp->r_flags |= RVT_R_RDMAR_SEQ;
 	qib_restart_rc(qp, qp->s_last_psn + 1, 0);
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= QIB_R_RSP_SEND;
+		qp->r_flags |= RVT_R_RSP_SEND;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
@@ -1399,12 +1385,12 @@ static void rdma_seq_err(struct qib_qp *qp, struct qib_ibport *ibp, u32 psn,
 static void qib_rc_rcv_resp(struct qib_ibport *ibp,
 			    struct qib_other_headers *ohdr,
 			    void *data, u32 tlen,
-			    struct qib_qp *qp,
+			    struct rvt_qp *qp,
 			    u32 opcode,
 			    u32 psn, u32 hdrsize, u32 pmtu,
 			    struct qib_ctxtdata *rcd)
 {
-	struct qib_swqe *wqe;
+	struct rvt_swqe *wqe;
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
 	enum ib_wc_status status;
 	unsigned long flags;
@@ -1425,7 +1411,7 @@ static void qib_rc_rcv_resp(struct qib_ibport *ibp,
 			 * If send tasklet not running attempt to progress
 			 * SDMA queue.
 			 */
-			if (!(qp->s_flags & QIB_S_BUSY)) {
+			if (!(qp->s_flags & RVT_S_BUSY)) {
 				/* Acquire SDMA Lock */
 				spin_lock_irqsave(&ppd->sdma_lock, flags);
 				/* Invoke sdma make progress */
@@ -1437,11 +1423,12 @@ static void qib_rc_rcv_resp(struct qib_ibport *ibp,
 	}
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
 		goto ack_done;
 
 	/* Ignore invalid responses. */
-	if (qib_cmp24(psn, qp->s_next_psn) >= 0)
+	smp_read_barrier_depends(); /* see post_one_send */
+	if (qib_cmp24(psn, ACCESS_ONCE(qp->s_next_psn)) >= 0)
 		goto ack_done;
 
 	/* Ignore duplicate responses. */
@@ -1460,15 +1447,15 @@ static void qib_rc_rcv_resp(struct qib_ibport *ibp,
 	 * Skip everything other than the PSN we expect, if we are waiting
 	 * for a reply to a restarted RDMA read or atomic op.
 	 */
-	if (qp->r_flags & QIB_R_RDMAR_SEQ) {
+	if (qp->r_flags & RVT_R_RDMAR_SEQ) {
 		if (qib_cmp24(psn, qp->s_last_psn + 1) != 0)
 			goto ack_done;
-		qp->r_flags &= ~QIB_R_RDMAR_SEQ;
+		qp->r_flags &= ~RVT_R_RDMAR_SEQ;
 	}
 
 	if (unlikely(qp->s_acked == qp->s_tail))
 		goto ack_done;
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	status = IB_WC_SUCCESS;
 
 	switch (opcode) {
@@ -1487,7 +1474,7 @@ static void qib_rc_rcv_resp(struct qib_ibport *ibp,
 		    opcode != OP(RDMA_READ_RESPONSE_FIRST))
 			goto ack_done;
 		hdrsize += 4;
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 		if (unlikely(wqe->wr.opcode != IB_WR_RDMA_READ))
 			goto ack_op_err;
 		/*
@@ -1515,10 +1502,10 @@ read_middle:
 		 * We got a response so update the timeout.
 		 * 4.096 usec. * (1 << qp->timeout)
 		 */
-		qp->s_flags |= QIB_S_TIMER;
+		qp->s_flags |= RVT_S_TIMER;
 		mod_timer(&qp->s_timer, jiffies + qp->timeout_jiffies);
-		if (qp->s_flags & QIB_S_WAIT_ACK) {
-			qp->s_flags &= ~QIB_S_WAIT_ACK;
+		if (qp->s_flags & RVT_S_WAIT_ACK) {
+			qp->s_flags &= ~RVT_S_WAIT_ACK;
 			qib_schedule_send(qp);
 		}
 
@@ -1553,7 +1540,7 @@ read_middle:
 		 * have to be careful to copy the data to the right
 		 * location.
 		 */
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 		qp->s_rdma_read_len = restart_sge(&qp->s_rdma_read_sge,
 						  wqe, psn, pmtu);
 		goto read_last;
@@ -1598,7 +1585,7 @@ ack_len_err:
 ack_err:
 	if (qp->s_last == qp->s_acked) {
 		qib_send_complete(qp, wqe, status);
-		qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+		rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 	}
 ack_done:
 	spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -1623,14 +1610,14 @@ bail:
  */
 static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 			    void *data,
-			    struct qib_qp *qp,
+			    struct rvt_qp *qp,
 			    u32 opcode,
 			    u32 psn,
 			    int diff,
 			    struct qib_ctxtdata *rcd)
 {
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	struct qib_ack_entry *e;
+	struct rvt_ack_entry *e;
 	unsigned long flags;
 	u8 i, prev;
 	int old_req;
@@ -1642,7 +1629,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 		 * Don't queue the NAK if we already sent one.
 		 */
 		if (!qp->r_nak_state) {
-			ibp->n_rc_seqnak++;
+			ibp->rvp.n_rc_seqnak++;
 			qp->r_nak_state = IB_NAK_PSN_ERROR;
 			/* Use the expected PSN. */
 			qp->r_ack_psn = qp->r_psn;
@@ -1652,7 +1639,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 			 * Otherwise, we end up propagating congestion.
 			 */
 			if (list_empty(&qp->rspwait)) {
-				qp->r_flags |= QIB_R_RSP_NAK;
+				qp->r_flags |= RVT_R_RSP_NAK;
 				atomic_inc(&qp->refcount);
 				list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 			}
@@ -1678,7 +1665,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 	 */
 	e = NULL;
 	old_req = 1;
-	ibp->n_rc_dupreq++;
+	ibp->rvp.n_rc_dupreq++;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 
@@ -1732,7 +1719,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 		if (unlikely(offset + len != e->rdma_sge.sge_length))
 			goto unlock_done;
 		if (e->rdma_sge.mr) {
-			qib_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		if (len != 0) {
@@ -1740,7 +1727,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 			u64 vaddr = be64_to_cpu(reth->vaddr);
 			int ok;
 
-			ok = qib_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
+			ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
 					 IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok))
 				goto unlock_done;
@@ -1791,7 +1778,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 		 * which doesn't accept a RDMA read response or atomic
 		 * response as an ACK for earlier SENDs or RDMA writes.
 		 */
-		if (!(qp->s_flags & QIB_S_RESP_PENDING)) {
+		if (!(qp->s_flags & RVT_S_RESP_PENDING)) {
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 			qp->r_nak_state = 0;
 			qp->r_ack_psn = qp->s_ack_queue[i].psn - 1;
@@ -1805,7 +1792,7 @@ static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
 		break;
 	}
 	qp->s_ack_state = OP(ACKNOWLEDGE);
-	qp->s_flags |= QIB_S_RESP_PENDING;
+	qp->s_flags |= RVT_S_RESP_PENDING;
 	qp->r_nak_state = 0;
 	qib_schedule_send(qp);
 
@@ -1818,13 +1805,13 @@ send_ack:
 	return 0;
 }
 
-void qib_rc_error(struct qib_qp *qp, enum ib_wc_status err)
+void qib_rc_error(struct rvt_qp *qp, enum ib_wc_status err)
 {
 	unsigned long flags;
 	int lastwqe;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	lastwqe = qib_error_qp(qp, err);
+	lastwqe = rvt_error_qp(qp, err);
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 
 	if (lastwqe) {
@@ -1837,7 +1824,7 @@ void qib_rc_error(struct qib_qp *qp, enum ib_wc_status err)
 	}
 }
 
-static inline void qib_update_ack_queue(struct qib_qp *qp, unsigned n)
+static inline void qib_update_ack_queue(struct rvt_qp *qp, unsigned n)
 {
 	unsigned next;
 
@@ -1862,7 +1849,7 @@ static inline void qib_update_ack_queue(struct qib_qp *qp, unsigned n)
  * Called at interrupt level.
  */
 void qib_rc_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp)
 {
 	struct qib_ibport *ibp = &rcd->ppd->ibport_data;
 	struct qib_other_headers *ohdr;
@@ -1948,8 +1935,8 @@ void qib_rc_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
 		break;
 	}
 
-	if (qp->state == IB_QPS_RTR && !(qp->r_flags & QIB_R_COMM_EST)) {
-		qp->r_flags |= QIB_R_COMM_EST;
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & RVT_R_COMM_EST)) {
+		qp->r_flags |= RVT_R_COMM_EST;
 		if (qp->ibqp.event_handler) {
 			struct ib_event ev;
 
@@ -2026,9 +2013,9 @@ send_last:
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto nack_inv;
 		qib_copy_sge(&qp->r_sge, data, tlen, 1);
-		qib_put_ss(&qp->r_sge);
+		rvt_put_ss(&qp->r_sge);
 		qp->r_msn++;
-		if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+		if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 			break;
 		wc.wr_id = qp->r_wr_id;
 		wc.status = IB_WC_SUCCESS;
@@ -2047,7 +2034,7 @@ send_last:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
 			     (ohdr->bth[0] &
 			      cpu_to_be32(IB_BTH_SOLICITED)) != 0);
 		break;
@@ -2069,7 +2056,7 @@ send_last:
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = qib_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
+			ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
 					 rkey, IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok))
 				goto nack_acc;
@@ -2096,7 +2083,7 @@ send_last:
 		goto send_last;
 
 	case OP(RDMA_READ_REQUEST): {
-		struct qib_ack_entry *e;
+		struct rvt_ack_entry *e;
 		u32 len;
 		u8 next;
 
@@ -2114,7 +2101,7 @@ send_last:
 		}
 		e = &qp->s_ack_queue[qp->r_head_ack_queue];
 		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
-			qib_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		reth = &ohdr->u.rc.reth;
@@ -2125,7 +2112,7 @@ send_last:
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = qib_rkey_ok(qp, &e->rdma_sge, len, vaddr,
+			ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr,
 					 rkey, IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok))
 				goto nack_acc_unlck;
@@ -2157,7 +2144,7 @@ send_last:
 		qp->r_head_ack_queue = next;
 
 		/* Schedule the send tasklet. */
-		qp->s_flags |= QIB_S_RESP_PENDING;
+		qp->s_flags |= RVT_S_RESP_PENDING;
 		qib_schedule_send(qp);
 
 		goto sunlock;
@@ -2166,7 +2153,7 @@ send_last:
 	case OP(COMPARE_SWAP):
 	case OP(FETCH_ADD): {
 		struct ib_atomic_eth *ateth;
-		struct qib_ack_entry *e;
+		struct rvt_ack_entry *e;
 		u64 vaddr;
 		atomic64_t *maddr;
 		u64 sdata;
@@ -2186,7 +2173,7 @@ send_last:
 		}
 		e = &qp->s_ack_queue[qp->r_head_ack_queue];
 		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
-			qib_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		ateth = &ohdr->u.atomic_eth;
@@ -2196,7 +2183,7 @@ send_last:
 			goto nack_inv_unlck;
 		rkey = be32_to_cpu(ateth->rkey);
 		/* Check rkey & NAK */
-		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
 					  vaddr, rkey,
 					  IB_ACCESS_REMOTE_ATOMIC)))
 			goto nack_acc_unlck;
@@ -2208,7 +2195,7 @@ send_last:
 			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
 				      be64_to_cpu(ateth->compare_data),
 				      sdata);
-		qib_put_mr(qp->r_sge.sge.mr);
+		rvt_put_mr(qp->r_sge.sge.mr);
 		qp->r_sge.num_sge = 0;
 		e->opcode = opcode;
 		e->sent = 0;
@@ -2221,7 +2208,7 @@ send_last:
 		qp->r_head_ack_queue = next;
 
 		/* Schedule the send tasklet. */
-		qp->s_flags |= QIB_S_RESP_PENDING;
+		qp->s_flags |= RVT_S_RESP_PENDING;
 		qib_schedule_send(qp);
 
 		goto sunlock;
@@ -2245,7 +2232,7 @@ rnr_nak:
 	qp->r_ack_psn = qp->r_psn;
 	/* Queue RNR NAK for later */
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= QIB_R_RSP_NAK;
+		qp->r_flags |= RVT_R_RSP_NAK;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
@@ -2257,7 +2244,7 @@ nack_op_err:
 	qp->r_ack_psn = qp->r_psn;
 	/* Queue NAK for later */
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= QIB_R_RSP_NAK;
+		qp->r_flags |= RVT_R_RSP_NAK;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
@@ -2271,7 +2258,7 @@ nack_inv:
 	qp->r_ack_psn = qp->r_psn;
 	/* Queue NAK for later */
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= QIB_R_RSP_NAK;
+		qp->r_flags |= RVT_R_RSP_NAK;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index b1aa21bdd484..a5f07a64b228 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -79,16 +79,16 @@ const u32 ib_qib_rnr_table[32] = {
  * Validate a RWQE and fill in the SGE state.
  * Return 1 if OK.
  */
-static int qib_init_sge(struct qib_qp *qp, struct qib_rwqe *wqe)
+static int qib_init_sge(struct rvt_qp *qp, struct rvt_rwqe *wqe)
 {
 	int i, j, ret;
 	struct ib_wc wc;
-	struct qib_lkey_table *rkt;
-	struct qib_pd *pd;
-	struct qib_sge_state *ss;
+	struct rvt_lkey_table *rkt;
+	struct rvt_pd *pd;
+	struct rvt_sge_state *ss;
 
-	rkt = &to_idev(qp->ibqp.device)->lk_table;
-	pd = to_ipd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
+	rkt = &to_idev(qp->ibqp.device)->rdi.lkey_table;
+	pd = ibpd_to_rvtpd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
 	ss = &qp->r_sge;
 	ss->sg_list = qp->r_sg_list;
 	qp->r_len = 0;
@@ -96,7 +96,7 @@ static int qib_init_sge(struct qib_qp *qp, struct qib_rwqe *wqe)
 		if (wqe->sg_list[i].length == 0)
 			continue;
 		/* Check LKEY */
-		if (!qib_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
+		if (!rvt_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
 				 &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
 			goto bad_lkey;
 		qp->r_len += wqe->sg_list[i].length;
@@ -109,9 +109,9 @@ static int qib_init_sge(struct qib_qp *qp, struct qib_rwqe *wqe)
 
 bad_lkey:
 	while (j) {
-		struct qib_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
+		struct rvt_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
 
-		qib_put_mr(sge->mr);
+		rvt_put_mr(sge->mr);
 	}
 	ss->num_sge = 0;
 	memset(&wc, 0, sizeof(wc));
@@ -120,7 +120,7 @@ bad_lkey:
 	wc.opcode = IB_WC_RECV;
 	wc.qp = &qp->ibqp;
 	/* Signal solicited completion event. */
-	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
 	ret = 0;
 bail:
 	return ret;
@@ -136,19 +136,19 @@ bail:
  *
  * Can be called from interrupt level.
  */
-int qib_get_rwqe(struct qib_qp *qp, int wr_id_only)
+int qib_get_rwqe(struct rvt_qp *qp, int wr_id_only)
 {
 	unsigned long flags;
-	struct qib_rq *rq;
-	struct qib_rwq *wq;
-	struct qib_srq *srq;
-	struct qib_rwqe *wqe;
+	struct rvt_rq *rq;
+	struct rvt_rwq *wq;
+	struct rvt_srq *srq;
+	struct rvt_rwqe *wqe;
 	void (*handler)(struct ib_event *, void *);
 	u32 tail;
 	int ret;
 
 	if (qp->ibqp.srq) {
-		srq = to_isrq(qp->ibqp.srq);
+		srq = ibsrq_to_rvtsrq(qp->ibqp.srq);
 		handler = srq->ibsrq.event_handler;
 		rq = &srq->rq;
 	} else {
@@ -158,7 +158,7 @@ int qib_get_rwqe(struct qib_qp *qp, int wr_id_only)
 	}
 
 	spin_lock_irqsave(&rq->lock, flags);
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
 		ret = 0;
 		goto unlock;
 	}
@@ -174,7 +174,7 @@ int qib_get_rwqe(struct qib_qp *qp, int wr_id_only)
 	}
 	/* Make sure entry is read after head index is read. */
 	smp_rmb();
-	wqe = get_rwqe_ptr(rq, tail);
+	wqe = rvt_get_rwqe_ptr(rq, tail);
 	/*
 	 * Even though we update the tail index in memory, the verbs
 	 * consumer is not supposed to post more entries until a
@@ -190,7 +190,7 @@ int qib_get_rwqe(struct qib_qp *qp, int wr_id_only)
 	qp->r_wr_id = wqe->wr_id;
 
 	ret = 1;
-	set_bit(QIB_R_WRID_VALID, &qp->r_aflags);
+	set_bit(RVT_R_WRID_VALID, &qp->r_aflags);
 	if (handler) {
 		u32 n;
 
@@ -227,7 +227,7 @@ bail:
  * Switch to alternate path.
  * The QP s_lock should be held and interrupts disabled.
  */
-void qib_migrate_qp(struct qib_qp *qp)
+void qib_migrate_qp(struct rvt_qp *qp)
 {
 	struct ib_event ev;
 
@@ -266,7 +266,7 @@ static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id)
  * The s_lock will be acquired around the qib_migrate_qp() call.
  */
 int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		      int has_grh, struct qib_qp *qp, u32 bth0)
+		      int has_grh, struct rvt_qp *qp, u32 bth0)
 {
 	__be64 guid;
 	unsigned long flags;
@@ -279,7 +279,8 @@ int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 			if (!(qp->alt_ah_attr.ah_flags & IB_AH_GRH))
 				goto err;
 			guid = get_sguid(ibp, qp->alt_ah_attr.grh.sgid_index);
-			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+			if (!gid_ok(&hdr->u.l.grh.dgid,
+				    ibp->rvp.gid_prefix, guid))
 				goto err;
 			if (!gid_ok(&hdr->u.l.grh.sgid,
 			    qp->alt_ah_attr.grh.dgid.global.subnet_prefix,
@@ -311,7 +312,8 @@ int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 				goto err;
 			guid = get_sguid(ibp,
 					 qp->remote_ah_attr.grh.sgid_index);
-			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+			if (!gid_ok(&hdr->u.l.grh.dgid,
+				    ibp->rvp.gid_prefix, guid))
 				goto err;
 			if (!gid_ok(&hdr->u.l.grh.sgid,
 			    qp->remote_ah_attr.grh.dgid.global.subnet_prefix,
@@ -353,12 +355,15 @@ err:
  * receive interrupts since this is a connected protocol and all packets
  * will pass through here.
  */
-static void qib_ruc_loopback(struct qib_qp *sqp)
+static void qib_ruc_loopback(struct rvt_qp *sqp)
 {
 	struct qib_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
-	struct qib_qp *qp;
-	struct qib_swqe *wqe;
-	struct qib_sge *sge;
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = ppd->dd;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
+	struct rvt_qp *qp;
+	struct rvt_swqe *wqe;
+	struct rvt_sge *sge;
 	unsigned long flags;
 	struct ib_wc wc;
 	u64 sdata;
@@ -367,29 +372,33 @@ static void qib_ruc_loopback(struct qib_qp *sqp)
 	int release;
 	int ret;
 
+	rcu_read_lock();
 	/*
 	 * Note that we check the responder QP state after
 	 * checking the requester's state.
 	 */
-	qp = qib_lookup_qpn(ibp, sqp->remote_qpn);
+	qp = rvt_lookup_qpn(rdi, &ibp->rvp, sqp->remote_qpn);
+	if (!qp)
+		goto done;
 
 	spin_lock_irqsave(&sqp->s_lock, flags);
 
 	/* Return if we are already busy processing a work request. */
-	if ((sqp->s_flags & (QIB_S_BUSY | QIB_S_ANY_WAIT)) ||
-	    !(ib_qib_state_ops[sqp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+	if ((sqp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT)) ||
+	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		goto unlock;
 
-	sqp->s_flags |= QIB_S_BUSY;
+	sqp->s_flags |= RVT_S_BUSY;
 
 again:
-	if (sqp->s_last == sqp->s_head)
+	smp_read_barrier_depends(); /* see post_one_send() */
+	if (sqp->s_last == ACCESS_ONCE(sqp->s_head))
 		goto clr_busy;
-	wqe = get_swqe_ptr(sqp, sqp->s_last);
+	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
 
 	/* Return if it is not OK to start a new work reqeust. */
-	if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_qib_state_ops[sqp->state] & QIB_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
 			goto clr_busy;
 		/* We are in the error state, flush the work request. */
 		send_status = IB_WC_WR_FLUSH_ERR;
@@ -407,9 +416,9 @@ again:
 	}
 	spin_unlock_irqrestore(&sqp->s_lock, flags);
 
-	if (!qp || !(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) ||
+	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
 	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
-		ibp->n_pkt_drops++;
+		ibp->rvp.n_pkt_drops++;
 		/*
 		 * For RC, the requester would timeout and retry so
 		 * shortcut the timeouts and just signal too many retries.
@@ -458,7 +467,7 @@ again:
 			goto inv_err;
 		if (wqe->length == 0)
 			break;
-		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
 					  wqe->rdma_wr.remote_addr,
 					  wqe->rdma_wr.rkey,
 					  IB_ACCESS_REMOTE_WRITE)))
@@ -471,7 +480,7 @@ again:
 	case IB_WR_RDMA_READ:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
 			goto inv_err;
-		if (unlikely(!qib_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
+		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
 					  wqe->rdma_wr.remote_addr,
 					  wqe->rdma_wr.rkey,
 					  IB_ACCESS_REMOTE_READ)))
@@ -489,7 +498,7 @@ again:
 	case IB_WR_ATOMIC_FETCH_AND_ADD:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
 			goto inv_err;
-		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
 					  wqe->atomic_wr.remote_addr,
 					  wqe->atomic_wr.rkey,
 					  IB_ACCESS_REMOTE_ATOMIC)))
@@ -502,7 +511,7 @@ again:
 			(u64) atomic64_add_return(sdata, maddr) - sdata :
 			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
 				      sdata, wqe->atomic_wr.swap);
-		qib_put_mr(qp->r_sge.sge.mr);
+		rvt_put_mr(qp->r_sge.sge.mr);
 		qp->r_sge.num_sge = 0;
 		goto send_comp;
 
@@ -526,11 +535,11 @@ again:
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (!release)
-				qib_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--sqp->s_sge.num_sge)
 				*sge = *sqp->s_sge.sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -543,9 +552,9 @@ again:
 		sqp->s_len -= len;
 	}
 	if (release)
-		qib_put_ss(&qp->r_sge);
+		rvt_put_ss(&qp->r_sge);
 
-	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		goto send_comp;
 
 	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
@@ -561,12 +570,12 @@ again:
 	wc.sl = qp->remote_ah_attr.sl;
 	wc.port_num = 1;
 	/* Signal completion event if the solicited bit is set. */
-	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-		       wqe->wr.send_flags & IB_SEND_SOLICITED);
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+		     wqe->wr.send_flags & IB_SEND_SOLICITED);
 
 send_comp:
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	ibp->n_loop_pkts++;
+	ibp->rvp.n_loop_pkts++;
 flush_send:
 	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
 	qib_send_complete(sqp, wqe, send_status);
@@ -576,7 +585,7 @@ rnr_nak:
 	/* Handle RNR NAK */
 	if (qp->ibqp.qp_type == IB_QPT_UC)
 		goto send_comp;
-	ibp->n_rnr_naks++;
+	ibp->rvp.n_rnr_naks++;
 	/*
 	 * Note: we don't need the s_lock held since the BUSY flag
 	 * makes this single threaded.
@@ -588,9 +597,9 @@ rnr_nak:
 	if (sqp->s_rnr_retry_cnt < 7)
 		sqp->s_rnr_retry--;
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
 		goto clr_busy;
-	sqp->s_flags |= QIB_S_WAIT_RNR;
+	sqp->s_flags |= RVT_S_WAIT_RNR;
 	sqp->s_timer.function = qib_rc_rnr_retry;
 	sqp->s_timer.expires = jiffies +
 		usecs_to_jiffies(ib_qib_rnr_table[qp->r_min_rnr_timer]);
@@ -618,9 +627,9 @@ serr:
 	spin_lock_irqsave(&sqp->s_lock, flags);
 	qib_send_complete(sqp, wqe, send_status);
 	if (sqp->ibqp.qp_type == IB_QPT_RC) {
-		int lastwqe = qib_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
 
-		sqp->s_flags &= ~QIB_S_BUSY;
+		sqp->s_flags &= ~RVT_S_BUSY;
 		spin_unlock_irqrestore(&sqp->s_lock, flags);
 		if (lastwqe) {
 			struct ib_event ev;
@@ -633,12 +642,11 @@ serr:
 		goto done;
 	}
 clr_busy:
-	sqp->s_flags &= ~QIB_S_BUSY;
+	sqp->s_flags &= ~RVT_S_BUSY;
 unlock:
 	spin_unlock_irqrestore(&sqp->s_lock, flags);
 done:
-	if (qp && atomic_dec_and_test(&qp->refcount))
-		wake_up(&qp->wait);
+	rcu_read_unlock();
 }
 
 /**
@@ -663,7 +671,7 @@ u32 qib_make_grh(struct qib_ibport *ibp, struct ib_grh *hdr,
 	hdr->next_hdr = IB_GRH_NEXT_HDR;
 	hdr->hop_limit = grh->hop_limit;
 	/* The SGID is 32-bit aligned. */
-	hdr->sgid.global.subnet_prefix = ibp->gid_prefix;
+	hdr->sgid.global.subnet_prefix = ibp->rvp.gid_prefix;
 	hdr->sgid.global.interface_id = grh->sgid_index ?
 		ibp->guids[grh->sgid_index - 1] : ppd_from_ibp(ibp)->guid;
 	hdr->dgid = grh->dgid;
@@ -672,9 +680,10 @@ u32 qib_make_grh(struct qib_ibport *ibp, struct ib_grh *hdr,
 	return sizeof(struct ib_grh) / sizeof(u32);
 }
 
-void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
+void qib_make_ruc_header(struct rvt_qp *qp, struct qib_other_headers *ohdr,
 			 u32 bth0, u32 bth2)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	u16 lrh0;
 	u32 nwords;
@@ -685,17 +694,18 @@ void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
 	nwords = (qp->s_cur_size + extra_bytes) >> 2;
 	lrh0 = QIB_LRH_BTH;
 	if (unlikely(qp->remote_ah_attr.ah_flags & IB_AH_GRH)) {
-		qp->s_hdrwords += qib_make_grh(ibp, &qp->s_hdr->u.l.grh,
+		qp->s_hdrwords += qib_make_grh(ibp, &priv->s_hdr->u.l.grh,
 					       &qp->remote_ah_attr.grh,
 					       qp->s_hdrwords, nwords);
 		lrh0 = QIB_LRH_GRH;
 	}
 	lrh0 |= ibp->sl_to_vl[qp->remote_ah_attr.sl] << 12 |
 		qp->remote_ah_attr.sl << 4;
-	qp->s_hdr->lrh[0] = cpu_to_be16(lrh0);
-	qp->s_hdr->lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
-	qp->s_hdr->lrh[2] = cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
-	qp->s_hdr->lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
+	priv->s_hdr->lrh[0] = cpu_to_be16(lrh0);
+	priv->s_hdr->lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
+	priv->s_hdr->lrh[2] =
+			cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
+	priv->s_hdr->lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
 				       qp->remote_ah_attr.src_path_bits);
 	bth0 |= qib_get_pkey(ibp, qp->s_pkey_index);
 	bth0 |= extra_bytes << 20;
@@ -707,20 +717,29 @@ void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
 	this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 }
 
+void _qib_do_send(struct work_struct *work)
+{
+	struct qib_qp_priv *priv = container_of(work, struct qib_qp_priv,
+						s_work);
+	struct rvt_qp *qp = priv->owner;
+
+	qib_do_send(qp);
+}
+
 /**
  * qib_do_send - perform a send on a QP
- * @work: contains a pointer to the QP
+ * @qp: pointer to the QP
  *
  * Process entries in the send work queue until credit or queue is
  * exhausted.  Only allow one CPU to send a packet per QP (tasklet).
  * Otherwise, two threads could send packets out of order.
  */
-void qib_do_send(struct work_struct *work)
+void qib_do_send(struct rvt_qp *qp)
 {
-	struct qib_qp *qp = container_of(work, struct qib_qp, s_work);
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-	int (*make_req)(struct qib_qp *qp);
+	int (*make_req)(struct rvt_qp *qp);
 	unsigned long flags;
 
 	if ((qp->ibqp.qp_type == IB_QPT_RC ||
@@ -745,50 +764,59 @@ void qib_do_send(struct work_struct *work)
 		return;
 	}
 
-	qp->s_flags |= QIB_S_BUSY;
-
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	qp->s_flags |= RVT_S_BUSY;
 
 	do {
 		/* Check for a constructed packet to be sent. */
 		if (qp->s_hdrwords != 0) {
+			spin_unlock_irqrestore(&qp->s_lock, flags);
 			/*
 			 * If the packet cannot be sent now, return and
 			 * the send tasklet will be woken up later.
 			 */
-			if (qib_verbs_send(qp, qp->s_hdr, qp->s_hdrwords,
+			if (qib_verbs_send(qp, priv->s_hdr, qp->s_hdrwords,
 					   qp->s_cur_sge, qp->s_cur_size))
-				break;
+				return;
 			/* Record that s_hdr is empty. */
 			qp->s_hdrwords = 0;
+			spin_lock_irqsave(&qp->s_lock, flags);
 		}
 	} while (make_req(qp));
+
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
 
 /*
  * This should be called with s_lock held.
  */
-void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe,
+void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 		       enum ib_wc_status status)
 {
 	u32 old_last, last;
 	unsigned i;
 
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		return;
 
+	last = qp->s_last;
+	old_last = last;
+	if (++last >= qp->s_size)
+		last = 0;
+	qp->s_last = last;
+	/* See post_send() */
+	barrier();
 	for (i = 0; i < wqe->wr.num_sge; i++) {
-		struct qib_sge *sge = &wqe->sg_list[i];
+		struct rvt_sge *sge = &wqe->sg_list[i];
 
-		qib_put_mr(sge->mr);
+		rvt_put_mr(sge->mr);
 	}
 	if (qp->ibqp.qp_type == IB_QPT_UD ||
 	    qp->ibqp.qp_type == IB_QPT_SMI ||
 	    qp->ibqp.qp_type == IB_QPT_GSI)
-		atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
+		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
 
 	/* See ch. 11.2.4.1 and 10.7.3.1 */
-	if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+	if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 	    (wqe->wr.send_flags & IB_SEND_SIGNALED) ||
 	    status != IB_WC_SUCCESS) {
 		struct ib_wc wc;
@@ -800,15 +828,10 @@ void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe,
 		wc.qp = &qp->ibqp;
 		if (status == IB_WC_SUCCESS)
 			wc.byte_len = wqe->length;
-		qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc,
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc,
 			     status != IB_WC_SUCCESS);
 	}
 
-	last = qp->s_last;
-	old_last = last;
-	if (++last >= qp->s_size)
-		last = 0;
-	qp->s_last = last;
 	if (qp->s_acked == old_last)
 		qp->s_acked = last;
 	if (qp->s_cur == old_last)
diff --git a/drivers/infiniband/hw/qib/qib_sdma.c b/drivers/infiniband/hw/qib/qib_sdma.c
index c6d6a54d2e19..891873b38a1e 100644
--- a/drivers/infiniband/hw/qib/qib_sdma.c
+++ b/drivers/infiniband/hw/qib/qib_sdma.c
@@ -513,7 +513,9 @@ int qib_sdma_running(struct qib_pportdata *ppd)
 static void complete_sdma_err_req(struct qib_pportdata *ppd,
 				  struct qib_verbs_txreq *tx)
 {
-	atomic_inc(&tx->qp->s_dma_busy);
+	struct qib_qp_priv *priv = tx->qp->priv;
+
+	atomic_inc(&priv->s_dma_busy);
 	/* no sdma descriptors, so no unmap_desc */
 	tx->txreq.start_idx = 0;
 	tx->txreq.next_descq_idx = 0;
@@ -531,18 +533,19 @@ static void complete_sdma_err_req(struct qib_pportdata *ppd,
  * 3) The SGE addresses are suitable for passing to dma_map_single().
  */
 int qib_sdma_verbs_send(struct qib_pportdata *ppd,
-			struct qib_sge_state *ss, u32 dwords,
+			struct rvt_sge_state *ss, u32 dwords,
 			struct qib_verbs_txreq *tx)
 {
 	unsigned long flags;
-	struct qib_sge *sge;
-	struct qib_qp *qp;
+	struct rvt_sge *sge;
+	struct rvt_qp *qp;
 	int ret = 0;
 	u16 tail;
 	__le64 *descqp;
 	u64 sdmadesc[2];
 	u32 dwoffset;
 	dma_addr_t addr;
+	struct qib_qp_priv *priv;
 
 	spin_lock_irqsave(&ppd->sdma_lock, flags);
 
@@ -621,7 +624,7 @@ retry:
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -644,8 +647,8 @@ retry:
 		descqp[0] |= cpu_to_le64(SDMA_DESC_DMA_HEAD);
 	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_INTREQ)
 		descqp[0] |= cpu_to_le64(SDMA_DESC_INTR);
-
-	atomic_inc(&tx->qp->s_dma_busy);
+	priv = tx->qp->priv;
+	atomic_inc(&priv->s_dma_busy);
 	tx->txreq.next_descq_idx = tail;
 	ppd->dd->f_sdma_update_tail(ppd, tail);
 	ppd->sdma_descq_added += tx->txreq.sg_count;
@@ -663,13 +666,14 @@ unmap:
 		unmap_desc(ppd, tail);
 	}
 	qp = tx->qp;
+	priv = qp->priv;
 	qib_put_txreq(tx);
 	spin_lock(&qp->r_lock);
 	spin_lock(&qp->s_lock);
 	if (qp->ibqp.qp_type == IB_QPT_RC) {
 		/* XXX what about error sending RDMA read responses? */
-		if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)
-			qib_error_qp(qp, IB_WC_GENERAL_ERR);
+		if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)
+			rvt_error_qp(qp, IB_WC_GENERAL_ERR);
 	} else if (qp->s_wqe)
 		qib_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
 	spin_unlock(&qp->s_lock);
@@ -679,8 +683,9 @@ unmap:
 
 busy:
 	qp = tx->qp;
+	priv = qp->priv;
 	spin_lock(&qp->s_lock);
-	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
 		struct qib_ibdev *dev;
 
 		/*
@@ -690,19 +695,19 @@ busy:
 		 */
 		tx->ss = ss;
 		tx->dwords = dwords;
-		qp->s_tx = tx;
+		priv->s_tx = tx;
 		dev = &ppd->dd->verbs_dev;
-		spin_lock(&dev->pending_lock);
-		if (list_empty(&qp->iowait)) {
+		spin_lock(&dev->rdi.pending_lock);
+		if (list_empty(&priv->iowait)) {
 			struct qib_ibport *ibp;
 
 			ibp = &ppd->ibport_data;
-			ibp->n_dmawait++;
-			qp->s_flags |= QIB_S_WAIT_DMA_DESC;
-			list_add_tail(&qp->iowait, &dev->dmawait);
+			ibp->rvp.n_dmawait++;
+			qp->s_flags |= RVT_S_WAIT_DMA_DESC;
+			list_add_tail(&priv->iowait, &dev->dmawait);
 		}
-		spin_unlock(&dev->pending_lock);
-		qp->s_flags &= ~QIB_S_BUSY;
+		spin_unlock(&dev->rdi.pending_lock);
+		qp->s_flags &= ~RVT_S_BUSY;
 		spin_unlock(&qp->s_lock);
 		ret = -EBUSY;
 	} else {
diff --git a/drivers/infiniband/hw/qib/qib_srq.c b/drivers/infiniband/hw/qib/qib_srq.c
deleted file mode 100644
index d6235931a1ba..000000000000
--- a/drivers/infiniband/hw/qib/qib_srq.c
+++ /dev/null
@@ -1,380 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
- * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <linux/err.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-
-#include "qib_verbs.h"
-
-/**
- * qib_post_srq_receive - post a receive on a shared receive queue
- * @ibsrq: the SRQ to post the receive on
- * @wr: the list of work requests to post
- * @bad_wr: A pointer to the first WR to cause a problem is put here
- *
- * This may be called from interrupt context.
- */
-int qib_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
-			 struct ib_recv_wr **bad_wr)
-{
-	struct qib_srq *srq = to_isrq(ibsrq);
-	struct qib_rwq *wq;
-	unsigned long flags;
-	int ret;
-
-	for (; wr; wr = wr->next) {
-		struct qib_rwqe *wqe;
-		u32 next;
-		int i;
-
-		if ((unsigned) wr->num_sge > srq->rq.max_sge) {
-			*bad_wr = wr;
-			ret = -EINVAL;
-			goto bail;
-		}
-
-		spin_lock_irqsave(&srq->rq.lock, flags);
-		wq = srq->rq.wq;
-		next = wq->head + 1;
-		if (next >= srq->rq.size)
-			next = 0;
-		if (next == wq->tail) {
-			spin_unlock_irqrestore(&srq->rq.lock, flags);
-			*bad_wr = wr;
-			ret = -ENOMEM;
-			goto bail;
-		}
-
-		wqe = get_rwqe_ptr(&srq->rq, wq->head);
-		wqe->wr_id = wr->wr_id;
-		wqe->num_sge = wr->num_sge;
-		for (i = 0; i < wr->num_sge; i++)
-			wqe->sg_list[i] = wr->sg_list[i];
-		/* Make sure queue entry is written before the head index. */
-		smp_wmb();
-		wq->head = next;
-		spin_unlock_irqrestore(&srq->rq.lock, flags);
-	}
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * qib_create_srq - create a shared receive queue
- * @ibpd: the protection domain of the SRQ to create
- * @srq_init_attr: the attributes of the SRQ
- * @udata: data from libibverbs when creating a user SRQ
- */
-struct ib_srq *qib_create_srq(struct ib_pd *ibpd,
-			      struct ib_srq_init_attr *srq_init_attr,
-			      struct ib_udata *udata)
-{
-	struct qib_ibdev *dev = to_idev(ibpd->device);
-	struct qib_srq *srq;
-	u32 sz;
-	struct ib_srq *ret;
-
-	if (srq_init_attr->srq_type != IB_SRQT_BASIC) {
-		ret = ERR_PTR(-ENOSYS);
-		goto done;
-	}
-
-	if (srq_init_attr->attr.max_sge == 0 ||
-	    srq_init_attr->attr.max_sge > ib_qib_max_srq_sges ||
-	    srq_init_attr->attr.max_wr == 0 ||
-	    srq_init_attr->attr.max_wr > ib_qib_max_srq_wrs) {
-		ret = ERR_PTR(-EINVAL);
-		goto done;
-	}
-
-	srq = kmalloc(sizeof(*srq), GFP_KERNEL);
-	if (!srq) {
-		ret = ERR_PTR(-ENOMEM);
-		goto done;
-	}
-
-	/*
-	 * Need to use vmalloc() if we want to support large #s of entries.
-	 */
-	srq->rq.size = srq_init_attr->attr.max_wr + 1;
-	srq->rq.max_sge = srq_init_attr->attr.max_sge;
-	sz = sizeof(struct ib_sge) * srq->rq.max_sge +
-		sizeof(struct qib_rwqe);
-	srq->rq.wq = vmalloc_user(sizeof(struct qib_rwq) + srq->rq.size * sz);
-	if (!srq->rq.wq) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_srq;
-	}
-
-	/*
-	 * Return the address of the RWQ as the offset to mmap.
-	 * See qib_mmap() for details.
-	 */
-	if (udata && udata->outlen >= sizeof(__u64)) {
-		int err;
-		u32 s = sizeof(struct qib_rwq) + srq->rq.size * sz;
-
-		srq->ip =
-		    qib_create_mmap_info(dev, s, ibpd->uobject->context,
-					 srq->rq.wq);
-		if (!srq->ip) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_wq;
-		}
-
-		err = ib_copy_to_udata(udata, &srq->ip->offset,
-				       sizeof(srq->ip->offset));
-		if (err) {
-			ret = ERR_PTR(err);
-			goto bail_ip;
-		}
-	} else
-		srq->ip = NULL;
-
-	/*
-	 * ib_create_srq() will initialize srq->ibsrq.
-	 */
-	spin_lock_init(&srq->rq.lock);
-	srq->rq.wq->head = 0;
-	srq->rq.wq->tail = 0;
-	srq->limit = srq_init_attr->attr.srq_limit;
-
-	spin_lock(&dev->n_srqs_lock);
-	if (dev->n_srqs_allocated == ib_qib_max_srqs) {
-		spin_unlock(&dev->n_srqs_lock);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_ip;
-	}
-
-	dev->n_srqs_allocated++;
-	spin_unlock(&dev->n_srqs_lock);
-
-	if (srq->ip) {
-		spin_lock_irq(&dev->pending_lock);
-		list_add(&srq->ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-	}
-
-	ret = &srq->ibsrq;
-	goto done;
-
-bail_ip:
-	kfree(srq->ip);
-bail_wq:
-	vfree(srq->rq.wq);
-bail_srq:
-	kfree(srq);
-done:
-	return ret;
-}
-
-/**
- * qib_modify_srq - modify a shared receive queue
- * @ibsrq: the SRQ to modify
- * @attr: the new attributes of the SRQ
- * @attr_mask: indicates which attributes to modify
- * @udata: user data for libibverbs.so
- */
-int qib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
-		   enum ib_srq_attr_mask attr_mask,
-		   struct ib_udata *udata)
-{
-	struct qib_srq *srq = to_isrq(ibsrq);
-	struct qib_rwq *wq;
-	int ret = 0;
-
-	if (attr_mask & IB_SRQ_MAX_WR) {
-		struct qib_rwq *owq;
-		struct qib_rwqe *p;
-		u32 sz, size, n, head, tail;
-
-		/* Check that the requested sizes are below the limits. */
-		if ((attr->max_wr > ib_qib_max_srq_wrs) ||
-		    ((attr_mask & IB_SRQ_LIMIT) ?
-		     attr->srq_limit : srq->limit) > attr->max_wr) {
-			ret = -EINVAL;
-			goto bail;
-		}
-
-		sz = sizeof(struct qib_rwqe) +
-			srq->rq.max_sge * sizeof(struct ib_sge);
-		size = attr->max_wr + 1;
-		wq = vmalloc_user(sizeof(struct qib_rwq) + size * sz);
-		if (!wq) {
-			ret = -ENOMEM;
-			goto bail;
-		}
-
-		/* Check that we can write the offset to mmap. */
-		if (udata && udata->inlen >= sizeof(__u64)) {
-			__u64 offset_addr;
-			__u64 offset = 0;
-
-			ret = ib_copy_from_udata(&offset_addr, udata,
-						 sizeof(offset_addr));
-			if (ret)
-				goto bail_free;
-			udata->outbuf =
-				(void __user *) (unsigned long) offset_addr;
-			ret = ib_copy_to_udata(udata, &offset,
-					       sizeof(offset));
-			if (ret)
-				goto bail_free;
-		}
-
-		spin_lock_irq(&srq->rq.lock);
-		/*
-		 * validate head and tail pointer values and compute
-		 * the number of remaining WQEs.
-		 */
-		owq = srq->rq.wq;
-		head = owq->head;
-		tail = owq->tail;
-		if (head >= srq->rq.size || tail >= srq->rq.size) {
-			ret = -EINVAL;
-			goto bail_unlock;
-		}
-		n = head;
-		if (n < tail)
-			n += srq->rq.size - tail;
-		else
-			n -= tail;
-		if (size <= n) {
-			ret = -EINVAL;
-			goto bail_unlock;
-		}
-		n = 0;
-		p = wq->wq;
-		while (tail != head) {
-			struct qib_rwqe *wqe;
-			int i;
-
-			wqe = get_rwqe_ptr(&srq->rq, tail);
-			p->wr_id = wqe->wr_id;
-			p->num_sge = wqe->num_sge;
-			for (i = 0; i < wqe->num_sge; i++)
-				p->sg_list[i] = wqe->sg_list[i];
-			n++;
-			p = (struct qib_rwqe *)((char *) p + sz);
-			if (++tail >= srq->rq.size)
-				tail = 0;
-		}
-		srq->rq.wq = wq;
-		srq->rq.size = size;
-		wq->head = n;
-		wq->tail = 0;
-		if (attr_mask & IB_SRQ_LIMIT)
-			srq->limit = attr->srq_limit;
-		spin_unlock_irq(&srq->rq.lock);
-
-		vfree(owq);
-
-		if (srq->ip) {
-			struct qib_mmap_info *ip = srq->ip;
-			struct qib_ibdev *dev = to_idev(srq->ibsrq.device);
-			u32 s = sizeof(struct qib_rwq) + size * sz;
-
-			qib_update_mmap_info(dev, ip, s, wq);
-
-			/*
-			 * Return the offset to mmap.
-			 * See qib_mmap() for details.
-			 */
-			if (udata && udata->inlen >= sizeof(__u64)) {
-				ret = ib_copy_to_udata(udata, &ip->offset,
-						       sizeof(ip->offset));
-				if (ret)
-					goto bail;
-			}
-
-			/*
-			 * Put user mapping info onto the pending list
-			 * unless it already is on the list.
-			 */
-			spin_lock_irq(&dev->pending_lock);
-			if (list_empty(&ip->pending_mmaps))
-				list_add(&ip->pending_mmaps,
-					 &dev->pending_mmaps);
-			spin_unlock_irq(&dev->pending_lock);
-		}
-	} else if (attr_mask & IB_SRQ_LIMIT) {
-		spin_lock_irq(&srq->rq.lock);
-		if (attr->srq_limit >= srq->rq.size)
-			ret = -EINVAL;
-		else
-			srq->limit = attr->srq_limit;
-		spin_unlock_irq(&srq->rq.lock);
-	}
-	goto bail;
-
-bail_unlock:
-	spin_unlock_irq(&srq->rq.lock);
-bail_free:
-	vfree(wq);
-bail:
-	return ret;
-}
-
-int qib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
-{
-	struct qib_srq *srq = to_isrq(ibsrq);
-
-	attr->max_wr = srq->rq.size - 1;
-	attr->max_sge = srq->rq.max_sge;
-	attr->srq_limit = srq->limit;
-	return 0;
-}
-
-/**
- * qib_destroy_srq - destroy a shared receive queue
- * @ibsrq: the SRQ to destroy
- */
-int qib_destroy_srq(struct ib_srq *ibsrq)
-{
-	struct qib_srq *srq = to_isrq(ibsrq);
-	struct qib_ibdev *dev = to_idev(ibsrq->device);
-
-	spin_lock(&dev->n_srqs_lock);
-	dev->n_srqs_allocated--;
-	spin_unlock(&dev->n_srqs_lock);
-	if (srq->ip)
-		kref_put(&srq->ip->ref, qib_release_mmap_info);
-	else
-		vfree(srq->rq.wq);
-	kfree(srq);
-
-	return 0;
-}
diff --git a/drivers/infiniband/hw/qib/qib_sysfs.c b/drivers/infiniband/hw/qib/qib_sysfs.c
index 81f56cdff2bc..fe4cf5e4acec 100644
--- a/drivers/infiniband/hw/qib/qib_sysfs.c
+++ b/drivers/infiniband/hw/qib/qib_sysfs.c
@@ -406,7 +406,13 @@ static struct kobj_type qib_sl2vl_ktype = {
 #define QIB_DIAGC_ATTR(N) \
 	static struct qib_diagc_attr qib_diagc_attr_##N = { \
 		.attr = { .name = __stringify(N), .mode = 0664 }, \
-		.counter = offsetof(struct qib_ibport, n_##N) \
+		.counter = offsetof(struct qib_ibport, rvp.n_##N) \
+	}
+
+#define QIB_DIAGC_ATTR_PER_CPU(N) \
+	static struct qib_diagc_attr qib_diagc_attr_##N = { \
+		.attr = { .name = __stringify(N), .mode = 0664 }, \
+		.counter = offsetof(struct qib_ibport, rvp.z_##N) \
 	}
 
 struct qib_diagc_attr {
@@ -414,10 +420,11 @@ struct qib_diagc_attr {
 	size_t counter;
 };
 
+QIB_DIAGC_ATTR_PER_CPU(rc_acks);
+QIB_DIAGC_ATTR_PER_CPU(rc_qacks);
+QIB_DIAGC_ATTR_PER_CPU(rc_delayed_comp);
+
 QIB_DIAGC_ATTR(rc_resends);
-QIB_DIAGC_ATTR(rc_acks);
-QIB_DIAGC_ATTR(rc_qacks);
-QIB_DIAGC_ATTR(rc_delayed_comp);
 QIB_DIAGC_ATTR(seq_naks);
 QIB_DIAGC_ATTR(rdma_seq);
 QIB_DIAGC_ATTR(rnr_naks);
@@ -449,6 +456,35 @@ static struct attribute *diagc_default_attributes[] = {
 	NULL
 };
 
+static u64 get_all_cpu_total(u64 __percpu *cntr)
+{
+	int cpu;
+	u64 counter = 0;
+
+	for_each_possible_cpu(cpu)
+		counter += *per_cpu_ptr(cntr, cpu);
+	return counter;
+}
+
+#define def_write_per_cpu(cntr) \
+static void write_per_cpu_##cntr(struct qib_pportdata *ppd, u32 data)	\
+{									\
+	struct qib_devdata *dd = ppd->dd;				\
+	struct qib_ibport *qibp = &ppd->ibport_data;			\
+	/*  A write can only zero the counter */			\
+	if (data == 0)							\
+		qibp->rvp.z_##cntr = get_all_cpu_total(qibp->rvp.cntr); \
+	else								\
+		qib_dev_err(dd, "Per CPU cntrs can only be zeroed");	\
+}
+
+def_write_per_cpu(rc_acks)
+def_write_per_cpu(rc_qacks)
+def_write_per_cpu(rc_delayed_comp)
+
+#define READ_PER_CPU_CNTR(cntr) (get_all_cpu_total(qibp->rvp.cntr) - \
+							qibp->rvp.z_##cntr)
+
 static ssize_t diagc_attr_show(struct kobject *kobj, struct attribute *attr,
 			       char *buf)
 {
@@ -458,7 +494,16 @@ static ssize_t diagc_attr_show(struct kobject *kobj, struct attribute *attr,
 		container_of(kobj, struct qib_pportdata, diagc_kobj);
 	struct qib_ibport *qibp = &ppd->ibport_data;
 
-	return sprintf(buf, "%u\n", *(u32 *)((char *)qibp + dattr->counter));
+	if (!strncmp(dattr->attr.name, "rc_acks", 7))
+		return sprintf(buf, "%llu\n", READ_PER_CPU_CNTR(rc_acks));
+	else if (!strncmp(dattr->attr.name, "rc_qacks", 8))
+		return sprintf(buf, "%llu\n", READ_PER_CPU_CNTR(rc_qacks));
+	else if (!strncmp(dattr->attr.name, "rc_delayed_comp", 15))
+		return sprintf(buf, "%llu\n",
+					READ_PER_CPU_CNTR(rc_delayed_comp));
+	else
+		return sprintf(buf, "%u\n",
+				*(u32 *)((char *)qibp + dattr->counter));
 }
 
 static ssize_t diagc_attr_store(struct kobject *kobj, struct attribute *attr,
@@ -475,7 +520,15 @@ static ssize_t diagc_attr_store(struct kobject *kobj, struct attribute *attr,
 	ret = kstrtou32(buf, 0, &val);
 	if (ret)
 		return ret;
-	*(u32 *)((char *) qibp + dattr->counter) = val;
+
+	if (!strncmp(dattr->attr.name, "rc_acks", 7))
+		write_per_cpu_rc_acks(ppd, val);
+	else if (!strncmp(dattr->attr.name, "rc_qacks", 8))
+		write_per_cpu_rc_qacks(ppd, val);
+	else if (!strncmp(dattr->attr.name, "rc_delayed_comp", 15))
+		write_per_cpu_rc_delayed_comp(ppd, val);
+	else
+		*(u32 *)((char *)qibp + dattr->counter) = val;
 	return size;
 }
 
@@ -502,7 +555,7 @@ static ssize_t show_rev(struct device *device, struct device_attribute *attr,
 			char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 
 	return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
 }
@@ -511,7 +564,7 @@ static ssize_t show_hca(struct device *device, struct device_attribute *attr,
 			char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 	int ret;
 
@@ -533,7 +586,7 @@ static ssize_t show_boardversion(struct device *device,
 				 struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 
 	/* The string printed here is already newline-terminated. */
@@ -545,7 +598,7 @@ static ssize_t show_localbus_info(struct device *device,
 				  struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 
 	/* The string printed here is already newline-terminated. */
@@ -557,7 +610,7 @@ static ssize_t show_nctxts(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 
 	/* Return the number of user ports (contexts) available. */
@@ -572,7 +625,7 @@ static ssize_t show_nfreectxts(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 
 	/* Return the number of free user ports (contexts) available. */
@@ -583,7 +636,7 @@ static ssize_t show_serial(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 
 	buf[sizeof(dd->serial)] = '\0';
@@ -597,7 +650,7 @@ static ssize_t store_chip_reset(struct device *device,
 				size_t count)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 	int ret;
 
@@ -618,7 +671,7 @@ static ssize_t show_tempsense(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
-		container_of(device, struct qib_ibdev, ibdev.dev);
+		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 	struct qib_devdata *dd = dd_from_dev(dev);
 	int ret;
 	int idx;
@@ -778,7 +831,7 @@ bail:
  */
 int qib_verbs_register_sysfs(struct qib_devdata *dd)
 {
-	struct ib_device *dev = &dd->verbs_dev.ibdev;
+	struct ib_device *dev = &dd->verbs_dev.rdi.ibdev;
 	int i, ret;
 
 	for (i = 0; i < ARRAY_SIZE(qib_attributes); ++i) {
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
index 06a564589c35..7bdbc79ceaa3 100644
--- a/drivers/infiniband/hw/qib/qib_uc.c
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -41,61 +41,62 @@
  * qib_make_uc_req - construct a request packet (SEND, RDMA write)
  * @qp: a pointer to the QP
  *
+ * Assumes the s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_uc_req(struct qib_qp *qp)
+int qib_make_uc_req(struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;
-	struct qib_swqe *wqe;
-	unsigned long flags;
+	struct rvt_swqe *wqe;
 	u32 hwords;
 	u32 bth0;
 	u32 len;
 	u32 pmtu = qp->pmtu;
 	int ret = 0;
 
-	spin_lock_irqsave(&qp->s_lock, flags);
-
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_SEND_OK)) {
-		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_dma_busy)) {
-			qp->s_flags |= QIB_S_WAIT_DMA;
+		if (atomic_read(&priv->s_dma_busy)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done;
 	}
 
-	ohdr = &qp->s_hdr->u.oth;
+	ohdr = &priv->s_hdr->u.oth;
 	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
-		ohdr = &qp->s_hdr->u.l.oth;
+		ohdr = &priv->s_hdr->u.l.oth;
 
 	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
 	hwords = 5;
 	bth0 = 0;
 
 	/* Get the next send request. */
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	qp->s_wqe = NULL;
 	switch (qp->s_state) {
 	default:
-		if (!(ib_qib_state_ops[qp->state] &
-		    QIB_PROCESS_NEXT_SEND_OK))
+		if (!(ib_rvt_state_ops[qp->state] &
+		    RVT_PROCESS_NEXT_SEND_OK))
 			goto bail;
 		/* Check if send work queue is empty. */
-		if (qp->s_cur == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_cur == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/*
 		 * Start a new request.
 		 */
-		wqe->psn = qp->s_next_psn;
-		qp->s_psn = qp->s_next_psn;
+		qp->s_psn = wqe->psn;
 		qp->s_sge.sge = wqe->sg_list[0];
 		qp->s_sge.sg_list = wqe->sg_list + 1;
 		qp->s_sge.num_sge = wqe->wr.num_sge;
@@ -214,15 +215,11 @@ int qib_make_uc_req(struct qib_qp *qp)
 	qp->s_cur_sge = &qp->s_sge;
 	qp->s_cur_size = len;
 	qib_make_ruc_header(qp, ohdr, bth0 | (qp->s_state << 24),
-			    qp->s_next_psn++ & QIB_PSN_MASK);
+			    qp->s_psn++ & QIB_PSN_MASK);
 done:
-	ret = 1;
-	goto unlock;
-
+	return 1;
 bail:
-	qp->s_flags &= ~QIB_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	qp->s_flags &= ~RVT_S_BUSY;
 	return ret;
 }
 
@@ -240,7 +237,7 @@ unlock:
  * Called at interrupt level.
  */
 void qib_uc_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp)
 {
 	struct qib_other_headers *ohdr;
 	u32 opcode;
@@ -278,10 +275,10 @@ void qib_uc_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 inv:
 		if (qp->r_state == OP(SEND_FIRST) ||
 		    qp->r_state == OP(SEND_MIDDLE)) {
-			set_bit(QIB_R_REWIND_SGE, &qp->r_aflags);
+			set_bit(RVT_R_REWIND_SGE, &qp->r_aflags);
 			qp->r_sge.num_sge = 0;
 		} else
-			qib_put_ss(&qp->r_sge);
+			rvt_put_ss(&qp->r_sge);
 		qp->r_state = OP(SEND_LAST);
 		switch (opcode) {
 		case OP(SEND_FIRST):
@@ -328,8 +325,8 @@ inv:
 		goto inv;
 	}
 
-	if (qp->state == IB_QPS_RTR && !(qp->r_flags & QIB_R_COMM_EST)) {
-		qp->r_flags |= QIB_R_COMM_EST;
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & RVT_R_COMM_EST)) {
+		qp->r_flags |= RVT_R_COMM_EST;
 		if (qp->ibqp.event_handler) {
 			struct ib_event ev;
 
@@ -346,7 +343,7 @@ inv:
 	case OP(SEND_ONLY):
 	case OP(SEND_ONLY_WITH_IMMEDIATE):
 send_first:
-		if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
+		if (test_and_clear_bit(RVT_R_REWIND_SGE, &qp->r_aflags))
 			qp->r_sge = qp->s_rdma_read_sge;
 		else {
 			ret = qib_get_rwqe(qp, 0);
@@ -400,7 +397,7 @@ send_last:
 			goto rewind;
 		wc.opcode = IB_WC_RECV;
 		qib_copy_sge(&qp->r_sge, data, tlen, 0);
-		qib_put_ss(&qp->s_rdma_read_sge);
+		rvt_put_ss(&qp->s_rdma_read_sge);
 last_imm:
 		wc.wr_id = qp->r_wr_id;
 		wc.status = IB_WC_SUCCESS;
@@ -414,7 +411,7 @@ last_imm:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
 			     (ohdr->bth[0] &
 				cpu_to_be32(IB_BTH_SOLICITED)) != 0);
 		break;
@@ -438,7 +435,7 @@ rdma_first:
 			int ok;
 
 			/* Check rkey */
-			ok = qib_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
+			ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
 					 vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok))
 				goto drop;
@@ -483,8 +480,8 @@ rdma_last_imm:
 		tlen -= (hdrsize + pad + 4);
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
-		if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
-			qib_put_ss(&qp->s_rdma_read_sge);
+		if (test_and_clear_bit(RVT_R_REWIND_SGE, &qp->r_aflags))
+			rvt_put_ss(&qp->s_rdma_read_sge);
 		else {
 			ret = qib_get_rwqe(qp, 1);
 			if (ret < 0)
@@ -495,7 +492,7 @@ rdma_last_imm:
 		wc.byte_len = qp->r_len;
 		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
 		qib_copy_sge(&qp->r_sge, data, tlen, 1);
-		qib_put_ss(&qp->r_sge);
+		rvt_put_ss(&qp->r_sge);
 		goto last_imm;
 
 	case OP(RDMA_WRITE_LAST):
@@ -511,7 +508,7 @@ rdma_last:
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
 		qib_copy_sge(&qp->r_sge, data, tlen, 1);
-		qib_put_ss(&qp->r_sge);
+		rvt_put_ss(&qp->r_sge);
 		break;
 
 	default:
@@ -523,10 +520,10 @@ rdma_last:
 	return;
 
 rewind:
-	set_bit(QIB_R_REWIND_SGE, &qp->r_aflags);
+	set_bit(RVT_R_REWIND_SGE, &qp->r_aflags);
 	qp->r_sge.num_sge = 0;
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 	return;
 
 op_err:
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index 59193f67ea78..d9502137de62 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -32,6 +32,7 @@
  */
 
 #include <rdma/ib_smi.h>
+#include <rdma/ib_verbs.h>
 
 #include "qib.h"
 #include "qib_mad.h"
@@ -46,22 +47,26 @@
  * Note that the receive interrupt handler may be calling qib_ud_rcv()
  * while this is being called.
  */
-static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
+static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 {
 	struct qib_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
-	struct qib_pportdata *ppd;
-	struct qib_qp *qp;
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = ppd->dd;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
+	struct rvt_qp *qp;
 	struct ib_ah_attr *ah_attr;
 	unsigned long flags;
-	struct qib_sge_state ssge;
-	struct qib_sge *sge;
+	struct rvt_sge_state ssge;
+	struct rvt_sge *sge;
 	struct ib_wc wc;
 	u32 length;
 	enum ib_qp_type sqptype, dqptype;
 
-	qp = qib_lookup_qpn(ibp, swqe->ud_wr.remote_qpn);
+	rcu_read_lock();
+	qp = rvt_lookup_qpn(rdi, &ibp->rvp, swqe->ud_wr.remote_qpn);
 	if (!qp) {
-		ibp->n_pkt_drops++;
+		ibp->rvp.n_pkt_drops++;
+		rcu_read_unlock();
 		return;
 	}
 
@@ -71,12 +76,12 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 			IB_QPT_UD : qp->ibqp.qp_type;
 
 	if (dqptype != sqptype ||
-	    !(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
-		ibp->n_pkt_drops++;
+	    !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
+		ibp->rvp.n_pkt_drops++;
 		goto drop;
 	}
 
-	ah_attr = &to_iah(swqe->ud_wr.ah)->attr;
+	ah_attr = &ibah_to_rvtah(swqe->ud_wr.ah)->attr;
 	ppd = ppd_from_ibp(ibp);
 
 	if (qp->ibqp.qp_num > 1) {
@@ -140,8 +145,8 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 	/*
 	 * Get the next work request entry to find where to put the data.
 	 */
-	if (qp->r_flags & QIB_R_REUSE_SGE)
-		qp->r_flags &= ~QIB_R_REUSE_SGE;
+	if (qp->r_flags & RVT_R_REUSE_SGE)
+		qp->r_flags &= ~RVT_R_REUSE_SGE;
 	else {
 		int ret;
 
@@ -152,14 +157,14 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 		}
 		if (!ret) {
 			if (qp->ibqp.qp_num == 0)
-				ibp->n_vl15_dropped++;
+				ibp->rvp.n_vl15_dropped++;
 			goto bail_unlock;
 		}
 	}
 	/* Silently drop packets which are too big. */
 	if (unlikely(wc.byte_len > qp->r_len)) {
-		qp->r_flags |= QIB_R_REUSE_SGE;
-		ibp->n_pkt_drops++;
+		qp->r_flags |= RVT_R_REUSE_SGE;
+		ibp->rvp.n_pkt_drops++;
 		goto bail_unlock;
 	}
 
@@ -189,7 +194,7 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 			if (--ssge.num_sge)
 				*sge = *ssge.sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -201,8 +206,8 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 		}
 		length -= len;
 	}
-	qib_put_ss(&qp->r_sge);
-	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+	rvt_put_ss(&qp->r_sge);
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		goto bail_unlock;
 	wc.wr_id = qp->r_wr_id;
 	wc.status = IB_WC_SUCCESS;
@@ -216,30 +221,31 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
 	wc.dlid_path_bits = ah_attr->dlid & ((1 << ppd->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
 		     swqe->wr.send_flags & IB_SEND_SOLICITED);
-	ibp->n_loop_pkts++;
+	ibp->rvp.n_loop_pkts++;
 bail_unlock:
 	spin_unlock_irqrestore(&qp->r_lock, flags);
 drop:
-	if (atomic_dec_and_test(&qp->refcount))
-		wake_up(&qp->wait);
+	rcu_read_unlock();
 }
 
 /**
  * qib_make_ud_req - construct a UD request packet
  * @qp: the QP
  *
+ * Assumes the s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_ud_req(struct qib_qp *qp)
+int qib_make_ud_req(struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;
 	struct ib_ah_attr *ah_attr;
 	struct qib_pportdata *ppd;
 	struct qib_ibport *ibp;
-	struct qib_swqe *wqe;
-	unsigned long flags;
+	struct rvt_swqe *wqe;
 	u32 nwords;
 	u32 extra_bytes;
 	u32 bth0;
@@ -248,28 +254,29 @@ int qib_make_ud_req(struct qib_qp *qp)
 	int ret = 0;
 	int next_cur;
 
-	spin_lock_irqsave(&qp->s_lock, flags);
-
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_dma_busy)) {
-			qp->s_flags |= QIB_S_WAIT_DMA;
+		if (atomic_read(&priv->s_dma_busy)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done;
 	}
 
-	if (qp->s_cur == qp->s_head)
+	/* see post_one_send() */
+	smp_read_barrier_depends();
+	if (qp->s_cur == ACCESS_ONCE(qp->s_head))
 		goto bail;
 
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	next_cur = qp->s_cur + 1;
 	if (next_cur >= qp->s_size)
 		next_cur = 0;
@@ -277,9 +284,9 @@ int qib_make_ud_req(struct qib_qp *qp)
 	/* Construct the header. */
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &to_iah(wqe->ud_wr.ah)->attr;
-	if (ah_attr->dlid >= QIB_MULTICAST_LID_BASE) {
-		if (ah_attr->dlid != QIB_PERMISSIVE_LID)
+	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	if (ah_attr->dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
+		if (ah_attr->dlid != be16_to_cpu(IB_LID_PERMISSIVE))
 			this_cpu_inc(ibp->pmastats->n_multicast_xmit);
 		else
 			this_cpu_inc(ibp->pmastats->n_unicast_xmit);
@@ -287,6 +294,7 @@ int qib_make_ud_req(struct qib_qp *qp)
 		this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
 		if (unlikely(lid == ppd->lid)) {
+			unsigned long flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -294,11 +302,12 @@ int qib_make_ud_req(struct qib_qp *qp)
 			 * XXX Instead of waiting, we could queue a
 			 * zero length descriptor so we get a callback.
 			 */
-			if (atomic_read(&qp->s_dma_busy)) {
-				qp->s_flags |= QIB_S_WAIT_DMA;
+			if (atomic_read(&priv->s_dma_busy)) {
+				qp->s_flags |= RVT_S_WAIT_DMA;
 				goto bail;
 			}
 			qp->s_cur = next_cur;
+			local_irq_save(flags);
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 			qib_ud_loopback(qp, wqe);
 			spin_lock_irqsave(&qp->s_lock, flags);
@@ -324,11 +333,11 @@ int qib_make_ud_req(struct qib_qp *qp)
 
 	if (ah_attr->ah_flags & IB_AH_GRH) {
 		/* Header size in 32-bit words. */
-		qp->s_hdrwords += qib_make_grh(ibp, &qp->s_hdr->u.l.grh,
+		qp->s_hdrwords += qib_make_grh(ibp, &priv->s_hdr->u.l.grh,
 					       &ah_attr->grh,
 					       qp->s_hdrwords, nwords);
 		lrh0 = QIB_LRH_GRH;
-		ohdr = &qp->s_hdr->u.l.oth;
+		ohdr = &priv->s_hdr->u.l.oth;
 		/*
 		 * Don't worry about sending to locally attached multicast
 		 * QPs.  It is unspecified by the spec. what happens.
@@ -336,7 +345,7 @@ int qib_make_ud_req(struct qib_qp *qp)
 	} else {
 		/* Header size in 32-bit words. */
 		lrh0 = QIB_LRH_BTH;
-		ohdr = &qp->s_hdr->u.oth;
+		ohdr = &priv->s_hdr->u.oth;
 	}
 	if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
 		qp->s_hdrwords++;
@@ -349,15 +358,16 @@ int qib_make_ud_req(struct qib_qp *qp)
 		lrh0 |= 0xF000; /* Set VL (see ch. 13.5.3.1) */
 	else
 		lrh0 |= ibp->sl_to_vl[ah_attr->sl] << 12;
-	qp->s_hdr->lrh[0] = cpu_to_be16(lrh0);
-	qp->s_hdr->lrh[1] = cpu_to_be16(ah_attr->dlid);  /* DEST LID */
-	qp->s_hdr->lrh[2] = cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
+	priv->s_hdr->lrh[0] = cpu_to_be16(lrh0);
+	priv->s_hdr->lrh[1] = cpu_to_be16(ah_attr->dlid);  /* DEST LID */
+	priv->s_hdr->lrh[2] =
+			cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
 	lid = ppd->lid;
 	if (lid) {
 		lid |= ah_attr->src_path_bits & ((1 << ppd->lmc) - 1);
-		qp->s_hdr->lrh[3] = cpu_to_be16(lid);
+		priv->s_hdr->lrh[3] = cpu_to_be16(lid);
 	} else
-		qp->s_hdr->lrh[3] = IB_LID_PERMISSIVE;
+		priv->s_hdr->lrh[3] = IB_LID_PERMISSIVE;
 	if (wqe->wr.send_flags & IB_SEND_SOLICITED)
 		bth0 |= IB_BTH_SOLICITED;
 	bth0 |= extra_bytes << 20;
@@ -368,11 +378,11 @@ int qib_make_ud_req(struct qib_qp *qp)
 	/*
 	 * Use the multicast QP if the destination LID is a multicast LID.
 	 */
-	ohdr->bth[1] = ah_attr->dlid >= QIB_MULTICAST_LID_BASE &&
-		ah_attr->dlid != QIB_PERMISSIVE_LID ?
+	ohdr->bth[1] = ah_attr->dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE) &&
+		ah_attr->dlid != be16_to_cpu(IB_LID_PERMISSIVE) ?
 		cpu_to_be32(QIB_MULTICAST_QPN) :
 		cpu_to_be32(wqe->ud_wr.remote_qpn);
-	ohdr->bth[2] = cpu_to_be32(qp->s_next_psn++ & QIB_PSN_MASK);
+	ohdr->bth[2] = cpu_to_be32(wqe->psn & QIB_PSN_MASK);
 	/*
 	 * Qkeys with the high order bit set mean use the
 	 * qkey from the QP context instead of the WR (see 10.2.5).
@@ -382,13 +392,9 @@ int qib_make_ud_req(struct qib_qp *qp)
 	ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
 
 done:
-	ret = 1;
-	goto unlock;
-
+	return 1;
 bail:
-	qp->s_flags &= ~QIB_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	qp->s_flags &= ~RVT_S_BUSY;
 	return ret;
 }
 
@@ -426,7 +432,7 @@ static unsigned qib_lookup_pkey(struct qib_ibport *ibp, u16 pkey)
  * Called at interrupt level.
  */
 void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp)
 {
 	struct qib_other_headers *ohdr;
 	int opcode;
@@ -446,7 +452,7 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 		hdrsize = 8 + 40 + 12 + 8; /* LRH + GRH + BTH + DETH */
 	}
 	qkey = be32_to_cpu(ohdr->u.ud.deth[0]);
-	src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & QIB_QPN_MASK;
+	src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & RVT_QPN_MASK;
 
 	/*
 	 * Get the number of bytes the message was padded by
@@ -531,8 +537,8 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 	/*
 	 * Get the next work request entry to find where to put the data.
 	 */
-	if (qp->r_flags & QIB_R_REUSE_SGE)
-		qp->r_flags &= ~QIB_R_REUSE_SGE;
+	if (qp->r_flags & RVT_R_REUSE_SGE)
+		qp->r_flags &= ~RVT_R_REUSE_SGE;
 	else {
 		int ret;
 
@@ -543,13 +549,13 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 		}
 		if (!ret) {
 			if (qp->ibqp.qp_num == 0)
-				ibp->n_vl15_dropped++;
+				ibp->rvp.n_vl15_dropped++;
 			return;
 		}
 	}
 	/* Silently drop packets which are too big. */
 	if (unlikely(wc.byte_len > qp->r_len)) {
-		qp->r_flags |= QIB_R_REUSE_SGE;
+		qp->r_flags |= RVT_R_REUSE_SGE;
 		goto drop;
 	}
 	if (has_grh) {
@@ -559,8 +565,8 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 	} else
 		qib_skip_sge(&qp->r_sge, sizeof(struct ib_grh), 1);
 	qib_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh), 1);
-	qib_put_ss(&qp->r_sge);
-	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+	rvt_put_ss(&qp->r_sge);
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		return;
 	wc.wr_id = qp->r_wr_id;
 	wc.status = IB_WC_SUCCESS;
@@ -576,15 +582,15 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
 	/*
 	 * Save the LMC lower bits if the destination LID is a unicast LID.
 	 */
-	wc.dlid_path_bits = dlid >= QIB_MULTICAST_LID_BASE ? 0 :
+	wc.dlid_path_bits = dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE) ? 0 :
 		dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
 		     (ohdr->bth[0] &
 			cpu_to_be32(IB_BTH_SOLICITED)) != 0);
 	return;
 
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 }
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index baf1e42b6896..cbf6200e6afc 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -41,6 +41,7 @@
 #include <linux/mm.h>
 #include <linux/random.h>
 #include <linux/vmalloc.h>
+#include <rdma/rdma_vt.h>
 
 #include "qib.h"
 #include "qib_common.h"
@@ -49,8 +50,8 @@ static unsigned int ib_qib_qp_table_size = 256;
 module_param_named(qp_table_size, ib_qib_qp_table_size, uint, S_IRUGO);
 MODULE_PARM_DESC(qp_table_size, "QP table size");
 
-unsigned int ib_qib_lkey_table_size = 16;
-module_param_named(lkey_table_size, ib_qib_lkey_table_size, uint,
+static unsigned int qib_lkey_table_size = 16;
+module_param_named(lkey_table_size, qib_lkey_table_size, uint,
 		   S_IRUGO);
 MODULE_PARM_DESC(lkey_table_size,
 		 "LKEY table size in bits (2^n, 1 <= n <= 23)");
@@ -113,36 +114,6 @@ module_param_named(disable_sma, ib_qib_disable_sma, uint, S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(disable_sma, "Disable the SMA");
 
 /*
- * Note that it is OK to post send work requests in the SQE and ERR
- * states; qib_do_send() will process them and generate error
- * completions as per IB 1.2 C10-96.
- */
-const int ib_qib_state_ops[IB_QPS_ERR + 1] = {
-	[IB_QPS_RESET] = 0,
-	[IB_QPS_INIT] = QIB_POST_RECV_OK,
-	[IB_QPS_RTR] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK,
-	[IB_QPS_RTS] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
-	    QIB_POST_SEND_OK | QIB_PROCESS_SEND_OK |
-	    QIB_PROCESS_NEXT_SEND_OK,
-	[IB_QPS_SQD] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
-	    QIB_POST_SEND_OK | QIB_PROCESS_SEND_OK,
-	[IB_QPS_SQE] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
-	    QIB_POST_SEND_OK | QIB_FLUSH_SEND,
-	[IB_QPS_ERR] = QIB_POST_RECV_OK | QIB_FLUSH_RECV |
-	    QIB_POST_SEND_OK | QIB_FLUSH_SEND,
-};
-
-struct qib_ucontext {
-	struct ib_ucontext ibucontext;
-};
-
-static inline struct qib_ucontext *to_iucontext(struct ib_ucontext
-						  *ibucontext)
-{
-	return container_of(ibucontext, struct qib_ucontext, ibucontext);
-}
-
-/*
  * Translate ib_wr_opcode into ib_wc_opcode.
  */
 const enum ib_wc_opcode ib_qib_wc_opcode[] = {
@@ -166,9 +137,9 @@ __be64 ib_qib_sys_image_guid;
  * @data: the data to copy
  * @length: the length of the data
  */
-void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length, int release)
+void qib_copy_sge(struct rvt_sge_state *ss, void *data, u32 length, int release)
 {
-	struct qib_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	while (length) {
 		u32 len = sge->length;
@@ -184,11 +155,11 @@ void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length, int release)
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (release)
-				qib_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -208,9 +179,9 @@ void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length, int release)
  * @ss: the SGE state
  * @length: the number of bytes to skip
  */
-void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release)
+void qib_skip_sge(struct rvt_sge_state *ss, u32 length, int release)
 {
-	struct qib_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	while (length) {
 		u32 len = sge->length;
@@ -225,11 +196,11 @@ void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release)
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (release)
-				qib_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -248,10 +219,10 @@ void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release)
  * Don't modify the qib_sge_state to get the count.
  * Return zero if any of the segments is not aligned.
  */
-static u32 qib_count_sge(struct qib_sge_state *ss, u32 length)
+static u32 qib_count_sge(struct rvt_sge_state *ss, u32 length)
 {
-	struct qib_sge *sg_list = ss->sg_list;
-	struct qib_sge sge = ss->sge;
+	struct rvt_sge *sg_list = ss->sg_list;
+	struct rvt_sge sge = ss->sge;
 	u8 num_sge = ss->num_sge;
 	u32 ndesc = 1;  /* count the header */
 
@@ -276,7 +247,7 @@ static u32 qib_count_sge(struct qib_sge_state *ss, u32 length)
 			if (--num_sge)
 				sge = *sg_list++;
 		} else if (sge.length == 0 && sge.mr->lkey) {
-			if (++sge.n >= QIB_SEGSZ) {
+			if (++sge.n >= RVT_SEGSZ) {
 				if (++sge.m >= sge.mr->mapsz)
 					break;
 				sge.n = 0;
@@ -294,9 +265,9 @@ static u32 qib_count_sge(struct qib_sge_state *ss, u32 length)
 /*
  * Copy from the SGEs to the data buffer.
  */
-static void qib_copy_from_sge(void *data, struct qib_sge_state *ss, u32 length)
+static void qib_copy_from_sge(void *data, struct rvt_sge_state *ss, u32 length)
 {
-	struct qib_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	while (length) {
 		u32 len = sge->length;
@@ -314,7 +285,7 @@ static void qib_copy_from_sge(void *data, struct qib_sge_state *ss, u32 length)
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -330,242 +301,6 @@ static void qib_copy_from_sge(void *data, struct qib_sge_state *ss, u32 length)
 }
 
 /**
- * qib_post_one_send - post one RC, UC, or UD send work request
- * @qp: the QP to post on
- * @wr: the work request to send
- */
-static int qib_post_one_send(struct qib_qp *qp, struct ib_send_wr *wr,
-	int *scheduled)
-{
-	struct qib_swqe *wqe;
-	u32 next;
-	int i;
-	int j;
-	int acc;
-	int ret;
-	unsigned long flags;
-	struct qib_lkey_table *rkt;
-	struct qib_pd *pd;
-	int avoid_schedule = 0;
-
-	spin_lock_irqsave(&qp->s_lock, flags);
-
-	/* Check that state is OK to post send. */
-	if (unlikely(!(ib_qib_state_ops[qp->state] & QIB_POST_SEND_OK)))
-		goto bail_inval;
-
-	/* IB spec says that num_sge == 0 is OK. */
-	if (wr->num_sge > qp->s_max_sge)
-		goto bail_inval;
-
-	/*
-	 * Don't allow RDMA reads or atomic operations on UC or
-	 * undefined operations.
-	 * Make sure buffer is large enough to hold the result for atomics.
-	 */
-	if (wr->opcode == IB_WR_REG_MR) {
-		if (qib_reg_mr(qp, reg_wr(wr)))
-			goto bail_inval;
-	} else if (qp->ibqp.qp_type == IB_QPT_UC) {
-		if ((unsigned) wr->opcode >= IB_WR_RDMA_READ)
-			goto bail_inval;
-	} else if (qp->ibqp.qp_type != IB_QPT_RC) {
-		/* Check IB_QPT_SMI, IB_QPT_GSI, IB_QPT_UD opcode */
-		if (wr->opcode != IB_WR_SEND &&
-		    wr->opcode != IB_WR_SEND_WITH_IMM)
-			goto bail_inval;
-		/* Check UD destination address PD */
-		if (qp->ibqp.pd != ud_wr(wr)->ah->pd)
-			goto bail_inval;
-	} else if ((unsigned) wr->opcode > IB_WR_ATOMIC_FETCH_AND_ADD)
-		goto bail_inval;
-	else if (wr->opcode >= IB_WR_ATOMIC_CMP_AND_SWP &&
-		   (wr->num_sge == 0 ||
-		    wr->sg_list[0].length < sizeof(u64) ||
-		    wr->sg_list[0].addr & (sizeof(u64) - 1)))
-		goto bail_inval;
-	else if (wr->opcode >= IB_WR_RDMA_READ && !qp->s_max_rd_atomic)
-		goto bail_inval;
-
-	next = qp->s_head + 1;
-	if (next >= qp->s_size)
-		next = 0;
-	if (next == qp->s_last) {
-		ret = -ENOMEM;
-		goto bail;
-	}
-
-	rkt = &to_idev(qp->ibqp.device)->lk_table;
-	pd = to_ipd(qp->ibqp.pd);
-	wqe = get_swqe_ptr(qp, qp->s_head);
-
-	if (qp->ibqp.qp_type != IB_QPT_UC &&
-	    qp->ibqp.qp_type != IB_QPT_RC)
-		memcpy(&wqe->ud_wr, ud_wr(wr), sizeof(wqe->ud_wr));
-	else if (wr->opcode == IB_WR_REG_MR)
-		memcpy(&wqe->reg_wr, reg_wr(wr),
-			sizeof(wqe->reg_wr));
-	else if (wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM ||
-		 wr->opcode == IB_WR_RDMA_WRITE ||
-		 wr->opcode == IB_WR_RDMA_READ)
-		memcpy(&wqe->rdma_wr, rdma_wr(wr), sizeof(wqe->rdma_wr));
-	else if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
-		 wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
-		memcpy(&wqe->atomic_wr, atomic_wr(wr), sizeof(wqe->atomic_wr));
-	else
-		memcpy(&wqe->wr, wr, sizeof(wqe->wr));
-
-	wqe->length = 0;
-	j = 0;
-	if (wr->num_sge) {
-		acc = wr->opcode >= IB_WR_RDMA_READ ?
-			IB_ACCESS_LOCAL_WRITE : 0;
-		for (i = 0; i < wr->num_sge; i++) {
-			u32 length = wr->sg_list[i].length;
-			int ok;
-
-			if (length == 0)
-				continue;
-			ok = qib_lkey_ok(rkt, pd, &wqe->sg_list[j],
-					 &wr->sg_list[i], acc);
-			if (!ok)
-				goto bail_inval_free;
-			wqe->length += length;
-			j++;
-		}
-		wqe->wr.num_sge = j;
-	}
-	if (qp->ibqp.qp_type == IB_QPT_UC ||
-	    qp->ibqp.qp_type == IB_QPT_RC) {
-		if (wqe->length > 0x80000000U)
-			goto bail_inval_free;
-		if (wqe->length <= qp->pmtu)
-			avoid_schedule = 1;
-	} else if (wqe->length > (dd_from_ibdev(qp->ibqp.device)->pport +
-				  qp->port_num - 1)->ibmtu) {
-		goto bail_inval_free;
-	} else {
-		atomic_inc(&to_iah(ud_wr(wr)->ah)->refcount);
-		avoid_schedule = 1;
-	}
-	wqe->ssn = qp->s_ssn++;
-	qp->s_head = next;
-
-	ret = 0;
-	goto bail;
-
-bail_inval_free:
-	while (j) {
-		struct qib_sge *sge = &wqe->sg_list[--j];
-
-		qib_put_mr(sge->mr);
-	}
-bail_inval:
-	ret = -EINVAL;
-bail:
-	if (!ret && !wr->next && !avoid_schedule &&
-	 !qib_sdma_empty(
-	   dd_from_ibdev(qp->ibqp.device)->pport + qp->port_num - 1)) {
-		qib_schedule_send(qp);
-		*scheduled = 1;
-	}
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-	return ret;
-}
-
-/**
- * qib_post_send - post a send on a QP
- * @ibqp: the QP to post the send on
- * @wr: the list of work requests to post
- * @bad_wr: the first bad WR is put here
- *
- * This may be called from interrupt context.
- */
-static int qib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
-			 struct ib_send_wr **bad_wr)
-{
-	struct qib_qp *qp = to_iqp(ibqp);
-	int err = 0;
-	int scheduled = 0;
-
-	for (; wr; wr = wr->next) {
-		err = qib_post_one_send(qp, wr, &scheduled);
-		if (err) {
-			*bad_wr = wr;
-			goto bail;
-		}
-	}
-
-	/* Try to do the send work in the caller's context. */
-	if (!scheduled)
-		qib_do_send(&qp->s_work);
-
-bail:
-	return err;
-}
-
-/**
- * qib_post_receive - post a receive on a QP
- * @ibqp: the QP to post the receive on
- * @wr: the WR to post
- * @bad_wr: the first bad WR is put here
- *
- * This may be called from interrupt context.
- */
-static int qib_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
-			    struct ib_recv_wr **bad_wr)
-{
-	struct qib_qp *qp = to_iqp(ibqp);
-	struct qib_rwq *wq = qp->r_rq.wq;
-	unsigned long flags;
-	int ret;
-
-	/* Check that state is OK to post receive. */
-	if (!(ib_qib_state_ops[qp->state] & QIB_POST_RECV_OK) || !wq) {
-		*bad_wr = wr;
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	for (; wr; wr = wr->next) {
-		struct qib_rwqe *wqe;
-		u32 next;
-		int i;
-
-		if ((unsigned) wr->num_sge > qp->r_rq.max_sge) {
-			*bad_wr = wr;
-			ret = -EINVAL;
-			goto bail;
-		}
-
-		spin_lock_irqsave(&qp->r_rq.lock, flags);
-		next = wq->head + 1;
-		if (next >= qp->r_rq.size)
-			next = 0;
-		if (next == wq->tail) {
-			spin_unlock_irqrestore(&qp->r_rq.lock, flags);
-			*bad_wr = wr;
-			ret = -ENOMEM;
-			goto bail;
-		}
-
-		wqe = get_rwqe_ptr(&qp->r_rq, wq->head);
-		wqe->wr_id = wr->wr_id;
-		wqe->num_sge = wr->num_sge;
-		for (i = 0; i < wr->num_sge; i++)
-			wqe->sg_list[i] = wr->sg_list[i];
-		/* Make sure queue entry is written before the head index. */
-		smp_wmb();
-		wq->head = next;
-		spin_unlock_irqrestore(&qp->r_rq.lock, flags);
-	}
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
  * qib_qp_rcv - processing an incoming packet on a QP
  * @rcd: the context pointer
  * @hdr: the packet header
@@ -579,15 +314,15 @@ bail:
  * Called at interrupt level.
  */
 static void qib_qp_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
-		       int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+		       int has_grh, void *data, u32 tlen, struct rvt_qp *qp)
 {
 	struct qib_ibport *ibp = &rcd->ppd->ibport_data;
 
 	spin_lock(&qp->r_lock);
 
 	/* Check for valid receive state. */
-	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
-		ibp->n_pkt_drops++;
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
+		ibp->rvp.n_pkt_drops++;
 		goto unlock;
 	}
 
@@ -632,8 +367,10 @@ void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
 	struct qib_pportdata *ppd = rcd->ppd;
 	struct qib_ibport *ibp = &ppd->ibport_data;
 	struct qib_ib_header *hdr = rhdr;
+	struct qib_devdata *dd = ppd->dd;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
 	struct qib_other_headers *ohdr;
-	struct qib_qp *qp;
+	struct rvt_qp *qp;
 	u32 qp_num;
 	int lnh;
 	u8 opcode;
@@ -645,7 +382,7 @@ void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
 
 	/* Check for a valid destination LID (see ch. 7.11.1). */
 	lid = be16_to_cpu(hdr->lrh[1]);
-	if (lid < QIB_MULTICAST_LID_BASE) {
+	if (lid < be16_to_cpu(IB_MULTICAST_LID_BASE)) {
 		lid &= ~((1 << ppd->lmc) - 1);
 		if (unlikely(lid != ppd->lid))
 			goto drop;
@@ -674,50 +411,40 @@ void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
 #endif
 
 	/* Get the destination QP number. */
-	qp_num = be32_to_cpu(ohdr->bth[1]) & QIB_QPN_MASK;
+	qp_num = be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
 	if (qp_num == QIB_MULTICAST_QPN) {
-		struct qib_mcast *mcast;
-		struct qib_mcast_qp *p;
+		struct rvt_mcast *mcast;
+		struct rvt_mcast_qp *p;
 
 		if (lnh != QIB_LRH_GRH)
 			goto drop;
-		mcast = qib_mcast_find(ibp, &hdr->u.l.grh.dgid);
+		mcast = rvt_mcast_find(&ibp->rvp, &hdr->u.l.grh.dgid);
 		if (mcast == NULL)
 			goto drop;
 		this_cpu_inc(ibp->pmastats->n_multicast_rcv);
 		list_for_each_entry_rcu(p, &mcast->qp_list, list)
 			qib_qp_rcv(rcd, hdr, 1, data, tlen, p->qp);
 		/*
-		 * Notify qib_multicast_detach() if it is waiting for us
+		 * Notify rvt_multicast_detach() if it is waiting for us
 		 * to finish.
 		 */
 		if (atomic_dec_return(&mcast->refcount) <= 1)
 			wake_up(&mcast->wait);
 	} else {
-		if (rcd->lookaside_qp) {
-			if (rcd->lookaside_qpn != qp_num) {
-				if (atomic_dec_and_test(
-					&rcd->lookaside_qp->refcount))
-					wake_up(
-					 &rcd->lookaside_qp->wait);
-				rcd->lookaside_qp = NULL;
-			}
+		rcu_read_lock();
+		qp = rvt_lookup_qpn(rdi, &ibp->rvp, qp_num);
+		if (!qp) {
+			rcu_read_unlock();
+			goto drop;
 		}
-		if (!rcd->lookaside_qp) {
-			qp = qib_lookup_qpn(ibp, qp_num);
-			if (!qp)
-				goto drop;
-			rcd->lookaside_qp = qp;
-			rcd->lookaside_qpn = qp_num;
-		} else
-			qp = rcd->lookaside_qp;
 		this_cpu_inc(ibp->pmastats->n_unicast_rcv);
 		qib_qp_rcv(rcd, hdr, lnh == QIB_LRH_GRH, data, tlen, qp);
+		rcu_read_unlock();
 	}
 	return;
 
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 }
 
 /*
@@ -728,23 +455,25 @@ static void mem_timer(unsigned long data)
 {
 	struct qib_ibdev *dev = (struct qib_ibdev *) data;
 	struct list_head *list = &dev->memwait;
-	struct qib_qp *qp = NULL;
+	struct rvt_qp *qp = NULL;
+	struct qib_qp_priv *priv = NULL;
 	unsigned long flags;
 
-	spin_lock_irqsave(&dev->pending_lock, flags);
+	spin_lock_irqsave(&dev->rdi.pending_lock, flags);
 	if (!list_empty(list)) {
-		qp = list_entry(list->next, struct qib_qp, iowait);
-		list_del_init(&qp->iowait);
+		priv = list_entry(list->next, struct qib_qp_priv, iowait);
+		qp = priv->owner;
+		list_del_init(&priv->iowait);
 		atomic_inc(&qp->refcount);
 		if (!list_empty(list))
 			mod_timer(&dev->mem_timer, jiffies + 1);
 	}
-	spin_unlock_irqrestore(&dev->pending_lock, flags);
+	spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 
 	if (qp) {
 		spin_lock_irqsave(&qp->s_lock, flags);
-		if (qp->s_flags & QIB_S_WAIT_KMEM) {
-			qp->s_flags &= ~QIB_S_WAIT_KMEM;
+		if (qp->s_flags & RVT_S_WAIT_KMEM) {
+			qp->s_flags &= ~RVT_S_WAIT_KMEM;
 			qib_schedule_send(qp);
 		}
 		spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -753,9 +482,9 @@ static void mem_timer(unsigned long data)
 	}
 }
 
-static void update_sge(struct qib_sge_state *ss, u32 length)
+static void update_sge(struct rvt_sge_state *ss, u32 length)
 {
-	struct qib_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	sge->vaddr += length;
 	sge->length -= length;
@@ -764,7 +493,7 @@ static void update_sge(struct qib_sge_state *ss, u32 length)
 		if (--ss->num_sge)
 			*sge = *ss->sg_list++;
 	} else if (sge->length == 0 && sge->mr->lkey) {
-		if (++sge->n >= QIB_SEGSZ) {
+		if (++sge->n >= RVT_SEGSZ) {
 			if (++sge->m >= sge->mr->mapsz)
 				return;
 			sge->n = 0;
@@ -810,7 +539,7 @@ static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
 }
 #endif
 
-static void copy_io(u32 __iomem *piobuf, struct qib_sge_state *ss,
+static void copy_io(u32 __iomem *piobuf, struct rvt_sge_state *ss,
 		    u32 length, unsigned flush_wc)
 {
 	u32 extra = 0;
@@ -947,30 +676,31 @@ static void copy_io(u32 __iomem *piobuf, struct qib_sge_state *ss,
 }
 
 static noinline struct qib_verbs_txreq *__get_txreq(struct qib_ibdev *dev,
-					   struct qib_qp *qp)
+					   struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_verbs_txreq *tx;
 	unsigned long flags;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	spin_lock(&dev->pending_lock);
+	spin_lock(&dev->rdi.pending_lock);
 
 	if (!list_empty(&dev->txreq_free)) {
 		struct list_head *l = dev->txreq_free.next;
 
 		list_del(l);
-		spin_unlock(&dev->pending_lock);
+		spin_unlock(&dev->rdi.pending_lock);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
 	} else {
-		if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK &&
-		    list_empty(&qp->iowait)) {
+		if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK &&
+		    list_empty(&priv->iowait)) {
 			dev->n_txwait++;
-			qp->s_flags |= QIB_S_WAIT_TX;
-			list_add_tail(&qp->iowait, &dev->txwait);
+			qp->s_flags |= RVT_S_WAIT_TX;
+			list_add_tail(&priv->iowait, &dev->txwait);
 		}
-		qp->s_flags &= ~QIB_S_BUSY;
-		spin_unlock(&dev->pending_lock);
+		qp->s_flags &= ~RVT_S_BUSY;
+		spin_unlock(&dev->rdi.pending_lock);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 		tx = ERR_PTR(-EBUSY);
 	}
@@ -978,22 +708,22 @@ static noinline struct qib_verbs_txreq *__get_txreq(struct qib_ibdev *dev,
 }
 
 static inline struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
-					 struct qib_qp *qp)
+					 struct rvt_qp *qp)
 {
 	struct qib_verbs_txreq *tx;
 	unsigned long flags;
 
-	spin_lock_irqsave(&dev->pending_lock, flags);
+	spin_lock_irqsave(&dev->rdi.pending_lock, flags);
 	/* assume the list non empty */
 	if (likely(!list_empty(&dev->txreq_free))) {
 		struct list_head *l = dev->txreq_free.next;
 
 		list_del(l);
-		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
 	} else {
 		/* call slow path to get the extra lock */
-		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 		tx =  __get_txreq(dev, qp);
 	}
 	return tx;
@@ -1002,16 +732,15 @@ static inline struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
 void qib_put_txreq(struct qib_verbs_txreq *tx)
 {
 	struct qib_ibdev *dev;
-	struct qib_qp *qp;
+	struct rvt_qp *qp;
+	struct qib_qp_priv *priv;
 	unsigned long flags;
 
 	qp = tx->qp;
 	dev = to_idev(qp->ibqp.device);
 
-	if (atomic_dec_and_test(&qp->refcount))
-		wake_up(&qp->wait);
 	if (tx->mr) {
-		qib_put_mr(tx->mr);
+		rvt_put_mr(tx->mr);
 		tx->mr = NULL;
 	}
 	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_FREEBUF) {
@@ -1022,21 +751,23 @@ void qib_put_txreq(struct qib_verbs_txreq *tx)
 		kfree(tx->align_buf);
 	}
 
-	spin_lock_irqsave(&dev->pending_lock, flags);
+	spin_lock_irqsave(&dev->rdi.pending_lock, flags);
 
 	/* Put struct back on free list */
 	list_add(&tx->txreq.list, &dev->txreq_free);
 
 	if (!list_empty(&dev->txwait)) {
 		/* Wake up first QP wanting a free struct */
-		qp = list_entry(dev->txwait.next, struct qib_qp, iowait);
-		list_del_init(&qp->iowait);
+		priv = list_entry(dev->txwait.next, struct qib_qp_priv,
+				  iowait);
+		qp = priv->owner;
+		list_del_init(&priv->iowait);
 		atomic_inc(&qp->refcount);
-		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 
 		spin_lock_irqsave(&qp->s_lock, flags);
-		if (qp->s_flags & QIB_S_WAIT_TX) {
-			qp->s_flags &= ~QIB_S_WAIT_TX;
+		if (qp->s_flags & RVT_S_WAIT_TX) {
+			qp->s_flags &= ~RVT_S_WAIT_TX;
 			qib_schedule_send(qp);
 		}
 		spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -1044,7 +775,7 @@ void qib_put_txreq(struct qib_verbs_txreq *tx)
 		if (atomic_dec_and_test(&qp->refcount))
 			wake_up(&qp->wait);
 	} else
-		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 }
 
 /*
@@ -1055,36 +786,39 @@ void qib_put_txreq(struct qib_verbs_txreq *tx)
  */
 void qib_verbs_sdma_desc_avail(struct qib_pportdata *ppd, unsigned avail)
 {
-	struct qib_qp *qp, *nqp;
-	struct qib_qp *qps[20];
+	struct rvt_qp *qp, *nqp;
+	struct qib_qp_priv *qpp, *nqpp;
+	struct rvt_qp *qps[20];
 	struct qib_ibdev *dev;
 	unsigned i, n;
 
 	n = 0;
 	dev = &ppd->dd->verbs_dev;
-	spin_lock(&dev->pending_lock);
+	spin_lock(&dev->rdi.pending_lock);
 
 	/* Search wait list for first QP wanting DMA descriptors. */
-	list_for_each_entry_safe(qp, nqp, &dev->dmawait, iowait) {
+	list_for_each_entry_safe(qpp, nqpp, &dev->dmawait, iowait) {
+		qp = qpp->owner;
+		nqp = nqpp->owner;
 		if (qp->port_num != ppd->port)
 			continue;
 		if (n == ARRAY_SIZE(qps))
 			break;
-		if (qp->s_tx->txreq.sg_count > avail)
+		if (qpp->s_tx->txreq.sg_count > avail)
 			break;
-		avail -= qp->s_tx->txreq.sg_count;
-		list_del_init(&qp->iowait);
+		avail -= qpp->s_tx->txreq.sg_count;
+		list_del_init(&qpp->iowait);
 		atomic_inc(&qp->refcount);
 		qps[n++] = qp;
 	}
 
-	spin_unlock(&dev->pending_lock);
+	spin_unlock(&dev->rdi.pending_lock);
 
 	for (i = 0; i < n; i++) {
 		qp = qps[i];
 		spin_lock(&qp->s_lock);
-		if (qp->s_flags & QIB_S_WAIT_DMA_DESC) {
-			qp->s_flags &= ~QIB_S_WAIT_DMA_DESC;
+		if (qp->s_flags & RVT_S_WAIT_DMA_DESC) {
+			qp->s_flags &= ~RVT_S_WAIT_DMA_DESC;
 			qib_schedule_send(qp);
 		}
 		spin_unlock(&qp->s_lock);
@@ -1100,7 +834,8 @@ static void sdma_complete(struct qib_sdma_txreq *cookie, int status)
 {
 	struct qib_verbs_txreq *tx =
 		container_of(cookie, struct qib_verbs_txreq, txreq);
-	struct qib_qp *qp = tx->qp;
+	struct rvt_qp *qp = tx->qp;
+	struct qib_qp_priv *priv = qp->priv;
 
 	spin_lock(&qp->s_lock);
 	if (tx->wqe)
@@ -1117,11 +852,11 @@ static void sdma_complete(struct qib_sdma_txreq *cookie, int status)
 		}
 		qib_rc_send_complete(qp, hdr);
 	}
-	if (atomic_dec_and_test(&qp->s_dma_busy)) {
+	if (atomic_dec_and_test(&priv->s_dma_busy)) {
 		if (qp->state == IB_QPS_RESET)
-			wake_up(&qp->wait_dma);
-		else if (qp->s_flags & QIB_S_WAIT_DMA) {
-			qp->s_flags &= ~QIB_S_WAIT_DMA;
+			wake_up(&priv->wait_dma);
+		else if (qp->s_flags & RVT_S_WAIT_DMA) {
+			qp->s_flags &= ~RVT_S_WAIT_DMA;
 			qib_schedule_send(qp);
 		}
 	}
@@ -1130,22 +865,23 @@ static void sdma_complete(struct qib_sdma_txreq *cookie, int status)
 	qib_put_txreq(tx);
 }
 
-static int wait_kmem(struct qib_ibdev *dev, struct qib_qp *qp)
+static int wait_kmem(struct qib_ibdev *dev, struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	unsigned long flags;
 	int ret = 0;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
-		spin_lock(&dev->pending_lock);
-		if (list_empty(&qp->iowait)) {
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
+		spin_lock(&dev->rdi.pending_lock);
+		if (list_empty(&priv->iowait)) {
 			if (list_empty(&dev->memwait))
 				mod_timer(&dev->mem_timer, jiffies + 1);
-			qp->s_flags |= QIB_S_WAIT_KMEM;
-			list_add_tail(&qp->iowait, &dev->memwait);
+			qp->s_flags |= RVT_S_WAIT_KMEM;
+			list_add_tail(&priv->iowait, &dev->memwait);
 		}
-		spin_unlock(&dev->pending_lock);
-		qp->s_flags &= ~QIB_S_BUSY;
+		spin_unlock(&dev->rdi.pending_lock);
+		qp->s_flags &= ~RVT_S_BUSY;
 		ret = -EBUSY;
 	}
 	spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -1153,10 +889,11 @@ static int wait_kmem(struct qib_ibdev *dev, struct qib_qp *qp)
 	return ret;
 }
 
-static int qib_verbs_send_dma(struct qib_qp *qp, struct qib_ib_header *hdr,
-			      u32 hdrwords, struct qib_sge_state *ss, u32 len,
+static int qib_verbs_send_dma(struct rvt_qp *qp, struct qib_ib_header *hdr,
+			      u32 hdrwords, struct rvt_sge_state *ss, u32 len,
 			      u32 plen, u32 dwords)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
 	struct qib_devdata *dd = dd_from_dev(dev);
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
@@ -1167,9 +904,9 @@ static int qib_verbs_send_dma(struct qib_qp *qp, struct qib_ib_header *hdr,
 	u32 ndesc;
 	int ret;
 
-	tx = qp->s_tx;
+	tx = priv->s_tx;
 	if (tx) {
-		qp->s_tx = NULL;
+		priv->s_tx = NULL;
 		/* resend previously constructed packet */
 		ret = qib_sdma_verbs_send(ppd, tx->ss, tx->dwords, tx);
 		goto bail;
@@ -1182,7 +919,6 @@ static int qib_verbs_send_dma(struct qib_qp *qp, struct qib_ib_header *hdr,
 	control = dd->f_setpbc_control(ppd, plen, qp->s_srate,
 				       be16_to_cpu(hdr->lrh[0]) >> 12);
 	tx->qp = qp;
-	atomic_inc(&qp->refcount);
 	tx->wqe = qp->s_wqe;
 	tx->mr = qp->s_rdma_mr;
 	if (qp->s_rdma_mr)
@@ -1245,7 +981,7 @@ err_tx:
 	qib_put_txreq(tx);
 	ret = wait_kmem(dev, qp);
 unaligned:
-	ibp->n_unaligned++;
+	ibp->rvp.n_unaligned++;
 bail:
 	return ret;
 bail_tx:
@@ -1257,8 +993,9 @@ bail_tx:
  * If we are now in the error state, return zero to flush the
  * send work request.
  */
-static int no_bufs_available(struct qib_qp *qp)
+static int no_bufs_available(struct rvt_qp *qp)
 {
+	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
 	struct qib_devdata *dd;
 	unsigned long flags;
@@ -1271,25 +1008,25 @@ static int no_bufs_available(struct qib_qp *qp)
 	 * enabling the PIO avail interrupt.
 	 */
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
-		spin_lock(&dev->pending_lock);
-		if (list_empty(&qp->iowait)) {
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
+		spin_lock(&dev->rdi.pending_lock);
+		if (list_empty(&priv->iowait)) {
 			dev->n_piowait++;
-			qp->s_flags |= QIB_S_WAIT_PIO;
-			list_add_tail(&qp->iowait, &dev->piowait);
+			qp->s_flags |= RVT_S_WAIT_PIO;
+			list_add_tail(&priv->iowait, &dev->piowait);
 			dd = dd_from_dev(dev);
 			dd->f_wantpiobuf_intr(dd, 1);
 		}
-		spin_unlock(&dev->pending_lock);
-		qp->s_flags &= ~QIB_S_BUSY;
+		spin_unlock(&dev->rdi.pending_lock);
+		qp->s_flags &= ~RVT_S_BUSY;
 		ret = -EBUSY;
 	}
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 	return ret;
 }
 
-static int qib_verbs_send_pio(struct qib_qp *qp, struct qib_ib_header *ibhdr,
-			      u32 hdrwords, struct qib_sge_state *ss, u32 len,
+static int qib_verbs_send_pio(struct rvt_qp *qp, struct qib_ib_header *ibhdr,
+			      u32 hdrwords, struct rvt_sge_state *ss, u32 len,
 			      u32 plen, u32 dwords)
 {
 	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
@@ -1370,7 +1107,7 @@ done:
 	}
 	qib_sendbuf_done(dd, pbufn);
 	if (qp->s_rdma_mr) {
-		qib_put_mr(qp->s_rdma_mr);
+		rvt_put_mr(qp->s_rdma_mr);
 		qp->s_rdma_mr = NULL;
 	}
 	if (qp->s_wqe) {
@@ -1394,10 +1131,10 @@ done:
  * @len: the length of the packet in bytes
  *
  * Return zero if packet is sent or queued OK.
- * Return non-zero and clear qp->s_flags QIB_S_BUSY otherwise.
+ * Return non-zero and clear qp->s_flags RVT_S_BUSY otherwise.
  */
-int qib_verbs_send(struct qib_qp *qp, struct qib_ib_header *hdr,
-		   u32 hdrwords, struct qib_sge_state *ss, u32 len)
+int qib_verbs_send(struct rvt_qp *qp, struct qib_ib_header *hdr,
+		   u32 hdrwords, struct rvt_sge_state *ss, u32 len)
 {
 	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
 	u32 plen;
@@ -1529,10 +1266,11 @@ void qib_ib_piobufavail(struct qib_devdata *dd)
 {
 	struct qib_ibdev *dev = &dd->verbs_dev;
 	struct list_head *list;
-	struct qib_qp *qps[5];
-	struct qib_qp *qp;
+	struct rvt_qp *qps[5];
+	struct rvt_qp *qp;
 	unsigned long flags;
 	unsigned i, n;
+	struct qib_qp_priv *priv;
 
 	list = &dev->piowait;
 	n = 0;
@@ -1543,25 +1281,26 @@ void qib_ib_piobufavail(struct qib_devdata *dd)
 	 * could end up with QPs on the wait list with the interrupt
 	 * disabled.
 	 */
-	spin_lock_irqsave(&dev->pending_lock, flags);
+	spin_lock_irqsave(&dev->rdi.pending_lock, flags);
 	while (!list_empty(list)) {
 		if (n == ARRAY_SIZE(qps))
 			goto full;
-		qp = list_entry(list->next, struct qib_qp, iowait);
-		list_del_init(&qp->iowait);
+		priv = list_entry(list->next, struct qib_qp_priv, iowait);
+		qp = priv->owner;
+		list_del_init(&priv->iowait);
 		atomic_inc(&qp->refcount);
 		qps[n++] = qp;
 	}
 	dd->f_wantpiobuf_intr(dd, 0);
 full:
-	spin_unlock_irqrestore(&dev->pending_lock, flags);
+	spin_unlock_irqrestore(&dev->rdi.pending_lock, flags);
 
 	for (i = 0; i < n; i++) {
 		qp = qps[i];
 
 		spin_lock_irqsave(&qp->s_lock, flags);
-		if (qp->s_flags & QIB_S_WAIT_PIO) {
-			qp->s_flags &= ~QIB_S_WAIT_PIO;
+		if (qp->s_flags & RVT_S_WAIT_PIO) {
+			qp->s_flags &= ~RVT_S_WAIT_PIO;
 			qib_schedule_send(qp);
 		}
 		spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -1572,82 +1311,24 @@ full:
 	}
 }
 
-static int qib_query_device(struct ib_device *ibdev, struct ib_device_attr *props,
-			    struct ib_udata *uhw)
-{
-	struct qib_devdata *dd = dd_from_ibdev(ibdev);
-	struct qib_ibdev *dev = to_idev(ibdev);
-
-	if (uhw->inlen || uhw->outlen)
-		return -EINVAL;
-	memset(props, 0, sizeof(*props));
-
-	props->device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
-		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
-		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
-		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
-	props->page_size_cap = PAGE_SIZE;
-	props->vendor_id =
-		QIB_SRC_OUI_1 << 16 | QIB_SRC_OUI_2 << 8 | QIB_SRC_OUI_3;
-	props->vendor_part_id = dd->deviceid;
-	props->hw_ver = dd->minrev;
-	props->sys_image_guid = ib_qib_sys_image_guid;
-	props->max_mr_size = ~0ULL;
-	props->max_qp = ib_qib_max_qps;
-	props->max_qp_wr = ib_qib_max_qp_wrs;
-	props->max_sge = ib_qib_max_sges;
-	props->max_sge_rd = ib_qib_max_sges;
-	props->max_cq = ib_qib_max_cqs;
-	props->max_ah = ib_qib_max_ahs;
-	props->max_cqe = ib_qib_max_cqes;
-	props->max_mr = dev->lk_table.max;
-	props->max_fmr = dev->lk_table.max;
-	props->max_map_per_fmr = 32767;
-	props->max_pd = ib_qib_max_pds;
-	props->max_qp_rd_atom = QIB_MAX_RDMA_ATOMIC;
-	props->max_qp_init_rd_atom = 255;
-	/* props->max_res_rd_atom */
-	props->max_srq = ib_qib_max_srqs;
-	props->max_srq_wr = ib_qib_max_srq_wrs;
-	props->max_srq_sge = ib_qib_max_srq_sges;
-	/* props->local_ca_ack_delay */
-	props->atomic_cap = IB_ATOMIC_GLOB;
-	props->max_pkeys = qib_get_npkeys(dd);
-	props->max_mcast_grp = ib_qib_max_mcast_grps;
-	props->max_mcast_qp_attach = ib_qib_max_mcast_qp_attached;
-	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
-		props->max_mcast_grp;
-
-	return 0;
-}
-
-static int qib_query_port(struct ib_device *ibdev, u8 port,
+static int qib_query_port(struct rvt_dev_info *rdi, u8 port_num,
 			  struct ib_port_attr *props)
 {
-	struct qib_devdata *dd = dd_from_ibdev(ibdev);
-	struct qib_ibport *ibp = to_iport(ibdev, port);
-	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = dd_from_dev(ibdev);
+	struct qib_pportdata *ppd = &dd->pport[port_num - 1];
 	enum ib_mtu mtu;
 	u16 lid = ppd->lid;
 
-	memset(props, 0, sizeof(*props));
 	props->lid = lid ? lid : be16_to_cpu(IB_LID_PERMISSIVE);
 	props->lmc = ppd->lmc;
-	props->sm_lid = ibp->sm_lid;
-	props->sm_sl = ibp->sm_sl;
 	props->state = dd->f_iblink_state(ppd->lastibcstat);
 	props->phys_state = dd->f_ibphys_portstate(ppd->lastibcstat);
-	props->port_cap_flags = ibp->port_cap_flags;
 	props->gid_tbl_len = QIB_GUIDS_PER_PORT;
-	props->max_msg_sz = 0x80000000;
-	props->pkey_tbl_len = qib_get_npkeys(dd);
-	props->bad_pkey_cntr = ibp->pkey_violations;
-	props->qkey_viol_cntr = ibp->qkey_violations;
 	props->active_width = ppd->link_width_active;
 	/* See rate_show() */
 	props->active_speed = ppd->link_speed_active;
 	props->max_vl_num = qib_num_vls(ppd->vls_supported);
-	props->init_type_reply = 0;
 
 	props->max_mtu = qib_ibmtu ? qib_ibmtu : IB_MTU_4096;
 	switch (ppd->ibmtu) {
@@ -1670,7 +1351,6 @@ static int qib_query_port(struct ib_device *ibdev, u8 port,
 		mtu = IB_MTU_2048;
 	}
 	props->active_mtu = mtu;
-	props->subnet_timeout = ibp->subnet_timeout;
 
 	return 0;
 }
@@ -1714,185 +1394,70 @@ bail:
 	return ret;
 }
 
-static int qib_modify_port(struct ib_device *ibdev, u8 port,
-			   int port_modify_mask, struct ib_port_modify *props)
+static int qib_shut_down_port(struct rvt_dev_info *rdi, u8 port_num)
 {
-	struct qib_ibport *ibp = to_iport(ibdev, port);
-	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-
-	ibp->port_cap_flags |= props->set_port_cap_mask;
-	ibp->port_cap_flags &= ~props->clr_port_cap_mask;
-	if (props->set_port_cap_mask || props->clr_port_cap_mask)
-		qib_cap_mask_chg(ibp);
-	if (port_modify_mask & IB_PORT_SHUTDOWN)
-		qib_set_linkstate(ppd, QIB_IB_LINKDOWN);
-	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
-		ibp->qkey_violations = 0;
-	return 0;
-}
-
-static int qib_query_gid(struct ib_device *ibdev, u8 port,
-			 int index, union ib_gid *gid)
-{
-	struct qib_devdata *dd = dd_from_ibdev(ibdev);
-	int ret = 0;
-
-	if (!port || port > dd->num_pports)
-		ret = -EINVAL;
-	else {
-		struct qib_ibport *ibp = to_iport(ibdev, port);
-		struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-
-		gid->global.subnet_prefix = ibp->gid_prefix;
-		if (index == 0)
-			gid->global.interface_id = ppd->guid;
-		else if (index < QIB_GUIDS_PER_PORT)
-			gid->global.interface_id = ibp->guids[index - 1];
-		else
-			ret = -EINVAL;
-	}
-
-	return ret;
-}
+	struct qib_ibdev *ibdev = container_of(rdi, struct qib_ibdev, rdi);
+	struct qib_devdata *dd = dd_from_dev(ibdev);
+	struct qib_pportdata *ppd = &dd->pport[port_num - 1];
 
-static struct ib_pd *qib_alloc_pd(struct ib_device *ibdev,
-				  struct ib_ucontext *context,
-				  struct ib_udata *udata)
-{
-	struct qib_ibdev *dev = to_idev(ibdev);
-	struct qib_pd *pd;
-	struct ib_pd *ret;
+	qib_set_linkstate(ppd, QIB_IB_LINKDOWN);
 
-	/*
-	 * This is actually totally arbitrary.  Some correctness tests
-	 * assume there's a maximum number of PDs that can be allocated.
-	 * We don't actually have this limit, but we fail the test if
-	 * we allow allocations of more than we report for this value.
-	 */
-
-	pd = kmalloc(sizeof(*pd), GFP_KERNEL);
-	if (!pd) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	spin_lock(&dev->n_pds_lock);
-	if (dev->n_pds_allocated == ib_qib_max_pds) {
-		spin_unlock(&dev->n_pds_lock);
-		kfree(pd);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	dev->n_pds_allocated++;
-	spin_unlock(&dev->n_pds_lock);
-
-	/* ib_alloc_pd() will initialize pd->ibpd. */
-	pd->user = udata != NULL;
-
-	ret = &pd->ibpd;
-
-bail:
-	return ret;
+	return 0;
 }
 
-static int qib_dealloc_pd(struct ib_pd *ibpd)
+static int qib_get_guid_be(struct rvt_dev_info *rdi, struct rvt_ibport *rvp,
+			   int guid_index, __be64 *guid)
 {
-	struct qib_pd *pd = to_ipd(ibpd);
-	struct qib_ibdev *dev = to_idev(ibpd->device);
-
-	spin_lock(&dev->n_pds_lock);
-	dev->n_pds_allocated--;
-	spin_unlock(&dev->n_pds_lock);
+	struct qib_ibport *ibp = container_of(rvp, struct qib_ibport, rvp);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
 
-	kfree(pd);
+	if (guid_index == 0)
+		*guid = ppd->guid;
+	else if (guid_index < QIB_GUIDS_PER_PORT)
+		*guid = ibp->guids[guid_index - 1];
+	else
+		return -EINVAL;
 
 	return 0;
 }
 
 int qib_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
 {
-	/* A multicast address requires a GRH (see ch. 8.4.1). */
-	if (ah_attr->dlid >= QIB_MULTICAST_LID_BASE &&
-	    ah_attr->dlid != QIB_PERMISSIVE_LID &&
-	    !(ah_attr->ah_flags & IB_AH_GRH))
-		goto bail;
-	if ((ah_attr->ah_flags & IB_AH_GRH) &&
-	    ah_attr->grh.sgid_index >= QIB_GUIDS_PER_PORT)
-		goto bail;
-	if (ah_attr->dlid == 0)
-		goto bail;
-	if (ah_attr->port_num < 1 ||
-	    ah_attr->port_num > ibdev->phys_port_cnt)
-		goto bail;
-	if (ah_attr->static_rate != IB_RATE_PORT_CURRENT &&
-	    ib_rate_to_mult(ah_attr->static_rate) < 0)
-		goto bail;
 	if (ah_attr->sl > 15)
-		goto bail;
+		return -EINVAL;
+
 	return 0;
-bail:
-	return -EINVAL;
 }
 
-/**
- * qib_create_ah - create an address handle
- * @pd: the protection domain
- * @ah_attr: the attributes of the AH
- *
- * This may be called from interrupt context.
- */
-static struct ib_ah *qib_create_ah(struct ib_pd *pd,
-				   struct ib_ah_attr *ah_attr)
+static void qib_notify_new_ah(struct ib_device *ibdev,
+			      struct ib_ah_attr *ah_attr,
+			      struct rvt_ah *ah)
 {
-	struct qib_ah *ah;
-	struct ib_ah *ret;
-	struct qib_ibdev *dev = to_idev(pd->device);
-	unsigned long flags;
+	struct qib_ibport *ibp;
+	struct qib_pportdata *ppd;
 
-	if (qib_check_ah(pd->device, ah_attr)) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
-
-	ah = kmalloc(sizeof(*ah), GFP_ATOMIC);
-	if (!ah) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	spin_lock_irqsave(&dev->n_ahs_lock, flags);
-	if (dev->n_ahs_allocated == ib_qib_max_ahs) {
-		spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-		kfree(ah);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	dev->n_ahs_allocated++;
-	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
-	/* ib_create_ah() will initialize ah->ibah. */
-	ah->attr = *ah_attr;
-	atomic_set(&ah->refcount, 0);
-
-	ret = &ah->ibah;
+	/*
+	 * Do not trust reading anything from rvt_ah at this point as it is not
+	 * done being setup. We can however modify things which we need to set.
+	 */
 
-bail:
-	return ret;
+	ibp = to_iport(ibdev, ah_attr->port_num);
+	ppd = ppd_from_ibp(ibp);
+	ah->vl = ibp->sl_to_vl[ah->attr.sl];
+	ah->log_pmtu = ilog2(ppd->ibmtu);
 }
 
 struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid)
 {
 	struct ib_ah_attr attr;
 	struct ib_ah *ah = ERR_PTR(-EINVAL);
-	struct qib_qp *qp0;
+	struct rvt_qp *qp0;
 
 	memset(&attr, 0, sizeof(attr));
 	attr.dlid = dlid;
 	attr.port_num = ppd_from_ibp(ibp)->port;
 	rcu_read_lock();
-	qp0 = rcu_dereference(ibp->qp0);
+	qp0 = rcu_dereference(ibp->rvp.qp[0]);
 	if (qp0)
 		ah = ib_create_ah(qp0->ibqp.pd, &attr);
 	rcu_read_unlock();
@@ -1900,51 +1465,6 @@ struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid)
 }
 
 /**
- * qib_destroy_ah - destroy an address handle
- * @ibah: the AH to destroy
- *
- * This may be called from interrupt context.
- */
-static int qib_destroy_ah(struct ib_ah *ibah)
-{
-	struct qib_ibdev *dev = to_idev(ibah->device);
-	struct qib_ah *ah = to_iah(ibah);
-	unsigned long flags;
-
-	if (atomic_read(&ah->refcount) != 0)
-		return -EBUSY;
-
-	spin_lock_irqsave(&dev->n_ahs_lock, flags);
-	dev->n_ahs_allocated--;
-	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
-	kfree(ah);
-
-	return 0;
-}
-
-static int qib_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
-	struct qib_ah *ah = to_iah(ibah);
-
-	if (qib_check_ah(ibah->device, ah_attr))
-		return -EINVAL;
-
-	ah->attr = *ah_attr;
-
-	return 0;
-}
-
-static int qib_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
-	struct qib_ah *ah = to_iah(ibah);
-
-	*ah_attr = ah->attr;
-
-	return 0;
-}
-
-/**
  * qib_get_npkeys - return the size of the PKEY table for context 0
  * @dd: the qlogic_ib device
  */
@@ -1973,75 +1493,27 @@ unsigned qib_get_pkey(struct qib_ibport *ibp, unsigned index)
 	return ret;
 }
 
-static int qib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
-			  u16 *pkey)
-{
-	struct qib_devdata *dd = dd_from_ibdev(ibdev);
-	int ret;
-
-	if (index >= qib_get_npkeys(dd)) {
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	*pkey = qib_get_pkey(to_iport(ibdev, port), index);
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * qib_alloc_ucontext - allocate a ucontest
- * @ibdev: the infiniband device
- * @udata: not used by the QLogic_IB driver
- */
-
-static struct ib_ucontext *qib_alloc_ucontext(struct ib_device *ibdev,
-					      struct ib_udata *udata)
-{
-	struct qib_ucontext *context;
-	struct ib_ucontext *ret;
-
-	context = kmalloc(sizeof(*context), GFP_KERNEL);
-	if (!context) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	ret = &context->ibucontext;
-
-bail:
-	return ret;
-}
-
-static int qib_dealloc_ucontext(struct ib_ucontext *context)
-{
-	kfree(to_iucontext(context));
-	return 0;
-}
-
 static void init_ibport(struct qib_pportdata *ppd)
 {
 	struct qib_verbs_counters cntrs;
 	struct qib_ibport *ibp = &ppd->ibport_data;
 
-	spin_lock_init(&ibp->lock);
+	spin_lock_init(&ibp->rvp.lock);
 	/* Set the prefix to the default value (see ch. 4.1.1) */
-	ibp->gid_prefix = IB_DEFAULT_GID_PREFIX;
-	ibp->sm_lid = be16_to_cpu(IB_LID_PERMISSIVE);
-	ibp->port_cap_flags = IB_PORT_SYS_IMAGE_GUID_SUP |
+	ibp->rvp.gid_prefix = IB_DEFAULT_GID_PREFIX;
+	ibp->rvp.sm_lid = be16_to_cpu(IB_LID_PERMISSIVE);
+	ibp->rvp.port_cap_flags = IB_PORT_SYS_IMAGE_GUID_SUP |
 		IB_PORT_CLIENT_REG_SUP | IB_PORT_SL_MAP_SUP |
 		IB_PORT_TRAP_SUP | IB_PORT_AUTO_MIGR_SUP |
 		IB_PORT_DR_NOTICE_SUP | IB_PORT_CAP_MASK_NOTICE_SUP |
 		IB_PORT_OTHER_LOCAL_CHANGES_SUP;
 	if (ppd->dd->flags & QIB_HAS_LINK_LATENCY)
-		ibp->port_cap_flags |= IB_PORT_LINK_LATENCY_SUP;
-	ibp->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
-	ibp->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
-	ibp->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
-	ibp->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
-	ibp->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
+		ibp->rvp.port_cap_flags |= IB_PORT_LINK_LATENCY_SUP;
+	ibp->rvp.pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
+	ibp->rvp.pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
+	ibp->rvp.pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
+	ibp->rvp.pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
+	ibp->rvp.pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
 
 	/* Snapshot current HW counters to "clear" them. */
 	qib_get_counters(ppd, &cntrs);
@@ -2061,26 +1533,55 @@ static void init_ibport(struct qib_pportdata *ppd)
 	ibp->z_excessive_buffer_overrun_errors =
 		cntrs.excessive_buffer_overrun_errors;
 	ibp->z_vl15_dropped = cntrs.vl15_dropped;
-	RCU_INIT_POINTER(ibp->qp0, NULL);
-	RCU_INIT_POINTER(ibp->qp1, NULL);
+	RCU_INIT_POINTER(ibp->rvp.qp[0], NULL);
+	RCU_INIT_POINTER(ibp->rvp.qp[1], NULL);
 }
 
-static int qib_port_immutable(struct ib_device *ibdev, u8 port_num,
-			      struct ib_port_immutable *immutable)
+/**
+ * qib_fill_device_attr - Fill in rvt dev info device attributes.
+ * @dd: the device data structure
+ */
+static void qib_fill_device_attr(struct qib_devdata *dd)
 {
-	struct ib_port_attr attr;
-	int err;
-
-	err = qib_query_port(ibdev, port_num, &attr);
-	if (err)
-		return err;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
 
-	immutable->pkey_tbl_len = attr.pkey_tbl_len;
-	immutable->gid_tbl_len = attr.gid_tbl_len;
-	immutable->core_cap_flags = RDMA_CORE_PORT_IBA_IB;
-	immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+	memset(&rdi->dparms.props, 0, sizeof(rdi->dparms.props));
 
-	return 0;
+	rdi->dparms.props.max_pd = ib_qib_max_pds;
+	rdi->dparms.props.max_ah = ib_qib_max_ahs;
+	rdi->dparms.props.device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
+		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
+		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
+		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
+	rdi->dparms.props.page_size_cap = PAGE_SIZE;
+	rdi->dparms.props.vendor_id =
+		QIB_SRC_OUI_1 << 16 | QIB_SRC_OUI_2 << 8 | QIB_SRC_OUI_3;
+	rdi->dparms.props.vendor_part_id = dd->deviceid;
+	rdi->dparms.props.hw_ver = dd->minrev;
+	rdi->dparms.props.sys_image_guid = ib_qib_sys_image_guid;
+	rdi->dparms.props.max_mr_size = ~0ULL;
+	rdi->dparms.props.max_qp = ib_qib_max_qps;
+	rdi->dparms.props.max_qp_wr = ib_qib_max_qp_wrs;
+	rdi->dparms.props.max_sge = ib_qib_max_sges;
+	rdi->dparms.props.max_sge_rd = ib_qib_max_sges;
+	rdi->dparms.props.max_cq = ib_qib_max_cqs;
+	rdi->dparms.props.max_cqe = ib_qib_max_cqes;
+	rdi->dparms.props.max_ah = ib_qib_max_ahs;
+	rdi->dparms.props.max_mr = rdi->lkey_table.max;
+	rdi->dparms.props.max_fmr = rdi->lkey_table.max;
+	rdi->dparms.props.max_map_per_fmr = 32767;
+	rdi->dparms.props.max_qp_rd_atom = QIB_MAX_RDMA_ATOMIC;
+	rdi->dparms.props.max_qp_init_rd_atom = 255;
+	rdi->dparms.props.max_srq = ib_qib_max_srqs;
+	rdi->dparms.props.max_srq_wr = ib_qib_max_srq_wrs;
+	rdi->dparms.props.max_srq_sge = ib_qib_max_srq_sges;
+	rdi->dparms.props.atomic_cap = IB_ATOMIC_GLOB;
+	rdi->dparms.props.max_pkeys = qib_get_npkeys(dd);
+	rdi->dparms.props.max_mcast_grp = ib_qib_max_mcast_grps;
+	rdi->dparms.props.max_mcast_qp_attach = ib_qib_max_mcast_qp_attached;
+	rdi->dparms.props.max_total_mcast_qp_attach =
+					rdi->dparms.props.max_mcast_qp_attach *
+					rdi->dparms.props.max_mcast_grp;
 }
 
 /**
@@ -2091,68 +1592,20 @@ static int qib_port_immutable(struct ib_device *ibdev, u8 port_num,
 int qib_register_ib_device(struct qib_devdata *dd)
 {
 	struct qib_ibdev *dev = &dd->verbs_dev;
-	struct ib_device *ibdev = &dev->ibdev;
+	struct ib_device *ibdev = &dev->rdi.ibdev;
 	struct qib_pportdata *ppd = dd->pport;
-	unsigned i, lk_tab_size;
+	unsigned i, ctxt;
 	int ret;
 
-	dev->qp_table_size = ib_qib_qp_table_size;
 	get_random_bytes(&dev->qp_rnd, sizeof(dev->qp_rnd));
-	dev->qp_table = kmalloc_array(
-				dev->qp_table_size,
-				sizeof(*dev->qp_table),
-				GFP_KERNEL);
-	if (!dev->qp_table) {
-		ret = -ENOMEM;
-		goto err_qpt;
-	}
-	for (i = 0; i < dev->qp_table_size; i++)
-		RCU_INIT_POINTER(dev->qp_table[i], NULL);
-
 	for (i = 0; i < dd->num_pports; i++)
 		init_ibport(ppd + i);
 
 	/* Only need to initialize non-zero fields. */
-	spin_lock_init(&dev->qpt_lock);
-	spin_lock_init(&dev->n_pds_lock);
-	spin_lock_init(&dev->n_ahs_lock);
-	spin_lock_init(&dev->n_cqs_lock);
-	spin_lock_init(&dev->n_qps_lock);
-	spin_lock_init(&dev->n_srqs_lock);
-	spin_lock_init(&dev->n_mcast_grps_lock);
-	init_timer(&dev->mem_timer);
-	dev->mem_timer.function = mem_timer;
-	dev->mem_timer.data = (unsigned long) dev;
-
-	qib_init_qpn_table(dd, &dev->qpn_table);
+	setup_timer(&dev->mem_timer, mem_timer, (unsigned long)dev);
+
+	qpt_mask = dd->qpn_mask;
 
-	/*
-	 * The top ib_qib_lkey_table_size bits are used to index the
-	 * table.  The lower 8 bits can be owned by the user (copied from
-	 * the LKEY).  The remaining bits act as a generation number or tag.
-	 */
-	spin_lock_init(&dev->lk_table.lock);
-	/* insure generation is at least 4 bits see keys.c */
-	if (ib_qib_lkey_table_size > MAX_LKEY_TABLE_BITS) {
-		qib_dev_warn(dd, "lkey bits %u too large, reduced to %u\n",
-			ib_qib_lkey_table_size, MAX_LKEY_TABLE_BITS);
-		ib_qib_lkey_table_size = MAX_LKEY_TABLE_BITS;
-	}
-	dev->lk_table.max = 1 << ib_qib_lkey_table_size;
-	lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
-	dev->lk_table.table = (struct qib_mregion __rcu **)
-		vmalloc(lk_tab_size);
-	if (dev->lk_table.table == NULL) {
-		ret = -ENOMEM;
-		goto err_lk;
-	}
-	RCU_INIT_POINTER(dev->dma_mr, NULL);
-	for (i = 0; i < dev->lk_table.max; i++)
-		RCU_INIT_POINTER(dev->lk_table.table[i], NULL);
-	INIT_LIST_HEAD(&dev->pending_mmaps);
-	spin_lock_init(&dev->pending_lock);
-	dev->mmap_offset = PAGE_SIZE;
-	spin_lock_init(&dev->mmap_offset_lock);
 	INIT_LIST_HEAD(&dev->piowait);
 	INIT_LIST_HEAD(&dev->dmawait);
 	INIT_LIST_HEAD(&dev->txwait);
@@ -2194,110 +1647,91 @@ int qib_register_ib_device(struct qib_devdata *dd)
 	strlcpy(ibdev->name, "qib%d", IB_DEVICE_NAME_MAX);
 	ibdev->owner = THIS_MODULE;
 	ibdev->node_guid = ppd->guid;
-	ibdev->uverbs_abi_ver = QIB_UVERBS_ABI_VERSION;
-	ibdev->uverbs_cmd_mask =
-		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)         |
-		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)        |
-		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)          |
-		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)            |
-		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)          |
-		(1ull << IB_USER_VERBS_CMD_CREATE_AH)           |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_AH)           |
-		(1ull << IB_USER_VERBS_CMD_QUERY_AH)            |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)          |
-		(1ull << IB_USER_VERBS_CMD_REG_MR)              |
-		(1ull << IB_USER_VERBS_CMD_DEREG_MR)            |
-		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
-		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)           |
-		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)          |
-		(1ull << IB_USER_VERBS_CMD_POLL_CQ)             |
-		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)       |
-		(1ull << IB_USER_VERBS_CMD_CREATE_QP)           |
-		(1ull << IB_USER_VERBS_CMD_QUERY_QP)            |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)          |
-		(1ull << IB_USER_VERBS_CMD_POST_SEND)           |
-		(1ull << IB_USER_VERBS_CMD_POST_RECV)           |
-		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)        |
-		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)        |
-		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)          |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)          |
-		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)         |
-		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
-	ibdev->node_type = RDMA_NODE_IB_CA;
 	ibdev->phys_port_cnt = dd->num_pports;
-	ibdev->num_comp_vectors = 1;
 	ibdev->dma_device = &dd->pcidev->dev;
-	ibdev->query_device = qib_query_device;
 	ibdev->modify_device = qib_modify_device;
-	ibdev->query_port = qib_query_port;
-	ibdev->modify_port = qib_modify_port;
-	ibdev->query_pkey = qib_query_pkey;
-	ibdev->query_gid = qib_query_gid;
-	ibdev->alloc_ucontext = qib_alloc_ucontext;
-	ibdev->dealloc_ucontext = qib_dealloc_ucontext;
-	ibdev->alloc_pd = qib_alloc_pd;
-	ibdev->dealloc_pd = qib_dealloc_pd;
-	ibdev->create_ah = qib_create_ah;
-	ibdev->destroy_ah = qib_destroy_ah;
-	ibdev->modify_ah = qib_modify_ah;
-	ibdev->query_ah = qib_query_ah;
-	ibdev->create_srq = qib_create_srq;
-	ibdev->modify_srq = qib_modify_srq;
-	ibdev->query_srq = qib_query_srq;
-	ibdev->destroy_srq = qib_destroy_srq;
-	ibdev->create_qp = qib_create_qp;
-	ibdev->modify_qp = qib_modify_qp;
-	ibdev->query_qp = qib_query_qp;
-	ibdev->destroy_qp = qib_destroy_qp;
-	ibdev->post_send = qib_post_send;
-	ibdev->post_recv = qib_post_receive;
-	ibdev->post_srq_recv = qib_post_srq_receive;
-	ibdev->create_cq = qib_create_cq;
-	ibdev->destroy_cq = qib_destroy_cq;
-	ibdev->resize_cq = qib_resize_cq;
-	ibdev->poll_cq = qib_poll_cq;
-	ibdev->req_notify_cq = qib_req_notify_cq;
-	ibdev->get_dma_mr = qib_get_dma_mr;
-	ibdev->reg_user_mr = qib_reg_user_mr;
-	ibdev->dereg_mr = qib_dereg_mr;
-	ibdev->alloc_mr = qib_alloc_mr;
-	ibdev->map_mr_sg = qib_map_mr_sg;
-	ibdev->alloc_fmr = qib_alloc_fmr;
-	ibdev->map_phys_fmr = qib_map_phys_fmr;
-	ibdev->unmap_fmr = qib_unmap_fmr;
-	ibdev->dealloc_fmr = qib_dealloc_fmr;
-	ibdev->attach_mcast = qib_multicast_attach;
-	ibdev->detach_mcast = qib_multicast_detach;
 	ibdev->process_mad = qib_process_mad;
-	ibdev->mmap = qib_mmap;
-	ibdev->dma_ops = &qib_dma_mapping_ops;
-	ibdev->get_port_immutable = qib_port_immutable;
 
 	snprintf(ibdev->node_desc, sizeof(ibdev->node_desc),
 		 "Intel Infiniband HCA %s", init_utsname()->nodename);
 
-	ret = ib_register_device(ibdev, qib_create_port_files);
-	if (ret)
-		goto err_reg;
+	/*
+	 * Fill in rvt info object.
+	 */
+	dd->verbs_dev.rdi.driver_f.port_callback = qib_create_port_files;
+	dd->verbs_dev.rdi.driver_f.get_card_name = qib_get_card_name;
+	dd->verbs_dev.rdi.driver_f.get_pci_dev = qib_get_pci_dev;
+	dd->verbs_dev.rdi.driver_f.check_ah = qib_check_ah;
+	dd->verbs_dev.rdi.driver_f.check_send_wqe = qib_check_send_wqe;
+	dd->verbs_dev.rdi.driver_f.notify_new_ah = qib_notify_new_ah;
+	dd->verbs_dev.rdi.driver_f.alloc_qpn = qib_alloc_qpn;
+	dd->verbs_dev.rdi.driver_f.qp_priv_alloc = qib_qp_priv_alloc;
+	dd->verbs_dev.rdi.driver_f.qp_priv_free = qib_qp_priv_free;
+	dd->verbs_dev.rdi.driver_f.free_all_qps = qib_free_all_qps;
+	dd->verbs_dev.rdi.driver_f.notify_qp_reset = qib_notify_qp_reset;
+	dd->verbs_dev.rdi.driver_f.do_send = qib_do_send;
+	dd->verbs_dev.rdi.driver_f.schedule_send = qib_schedule_send;
+	dd->verbs_dev.rdi.driver_f.quiesce_qp = qib_quiesce_qp;
+	dd->verbs_dev.rdi.driver_f.stop_send_queue = qib_stop_send_queue;
+	dd->verbs_dev.rdi.driver_f.flush_qp_waiters = qib_flush_qp_waiters;
+	dd->verbs_dev.rdi.driver_f.notify_error_qp = qib_notify_error_qp;
+	dd->verbs_dev.rdi.driver_f.mtu_to_path_mtu = qib_mtu_to_path_mtu;
+	dd->verbs_dev.rdi.driver_f.mtu_from_qp = qib_mtu_from_qp;
+	dd->verbs_dev.rdi.driver_f.get_pmtu_from_attr = qib_get_pmtu_from_attr;
+	dd->verbs_dev.rdi.driver_f.schedule_send_no_lock = _qib_schedule_send;
+	dd->verbs_dev.rdi.driver_f.query_port_state = qib_query_port;
+	dd->verbs_dev.rdi.driver_f.shut_down_port = qib_shut_down_port;
+	dd->verbs_dev.rdi.driver_f.cap_mask_chg = qib_cap_mask_chg;
+	dd->verbs_dev.rdi.driver_f.notify_create_mad_agent =
+						qib_notify_create_mad_agent;
+	dd->verbs_dev.rdi.driver_f.notify_free_mad_agent =
+						qib_notify_free_mad_agent;
+
+	dd->verbs_dev.rdi.dparms.max_rdma_atomic = QIB_MAX_RDMA_ATOMIC;
+	dd->verbs_dev.rdi.driver_f.get_guid_be = qib_get_guid_be;
+	dd->verbs_dev.rdi.dparms.lkey_table_size = qib_lkey_table_size;
+	dd->verbs_dev.rdi.dparms.qp_table_size = ib_qib_qp_table_size;
+	dd->verbs_dev.rdi.dparms.qpn_start = 1;
+	dd->verbs_dev.rdi.dparms.qpn_res_start = QIB_KD_QP;
+	dd->verbs_dev.rdi.dparms.qpn_res_end = QIB_KD_QP; /* Reserve one QP */
+	dd->verbs_dev.rdi.dparms.qpn_inc = 1;
+	dd->verbs_dev.rdi.dparms.qos_shift = 1;
+	dd->verbs_dev.rdi.dparms.psn_mask = QIB_PSN_MASK;
+	dd->verbs_dev.rdi.dparms.psn_shift = QIB_PSN_SHIFT;
+	dd->verbs_dev.rdi.dparms.psn_modify_mask = QIB_PSN_MASK;
+	dd->verbs_dev.rdi.dparms.nports = dd->num_pports;
+	dd->verbs_dev.rdi.dparms.npkeys = qib_get_npkeys(dd);
+	dd->verbs_dev.rdi.dparms.node = dd->assigned_node_id;
+	dd->verbs_dev.rdi.dparms.core_cap_flags = RDMA_CORE_PORT_IBA_IB;
+	dd->verbs_dev.rdi.dparms.max_mad_size = IB_MGMT_MAD_SIZE;
+
+	snprintf(dd->verbs_dev.rdi.dparms.cq_name,
+		 sizeof(dd->verbs_dev.rdi.dparms.cq_name),
+		 "qib_cq%d", dd->unit);
+
+	qib_fill_device_attr(dd);
+
+	ppd = dd->pport;
+	for (i = 0; i < dd->num_pports; i++, ppd++) {
+		ctxt = ppd->hw_pidx;
+		rvt_init_port(&dd->verbs_dev.rdi,
+			      &ppd->ibport_data.rvp,
+			      i,
+			      dd->rcd[ctxt]->pkeys);
+	}
 
-	ret = qib_create_agents(dev);
+	ret = rvt_register_device(&dd->verbs_dev.rdi);
 	if (ret)
-		goto err_agents;
+		goto err_tx;
 
 	ret = qib_verbs_register_sysfs(dd);
 	if (ret)
 		goto err_class;
 
-	goto bail;
+	return ret;
 
 err_class:
-	qib_free_agents(dev);
-err_agents:
-	ib_unregister_device(ibdev);
-err_reg:
+	rvt_unregister_device(&dd->verbs_dev.rdi);
 err_tx:
 	while (!list_empty(&dev->txreq_free)) {
 		struct list_head *l = dev->txreq_free.next;
@@ -2313,27 +1747,17 @@ err_tx:
 					sizeof(struct qib_pio_header),
 				  dev->pio_hdrs, dev->pio_hdrs_phys);
 err_hdrs:
-	vfree(dev->lk_table.table);
-err_lk:
-	kfree(dev->qp_table);
-err_qpt:
 	qib_dev_err(dd, "cannot register verbs: %d!\n", -ret);
-bail:
 	return ret;
 }
 
 void qib_unregister_ib_device(struct qib_devdata *dd)
 {
 	struct qib_ibdev *dev = &dd->verbs_dev;
-	struct ib_device *ibdev = &dev->ibdev;
-	u32 qps_inuse;
-	unsigned lk_tab_size;
 
 	qib_verbs_unregister_sysfs(dd);
 
-	qib_free_agents(dev);
-
-	ib_unregister_device(ibdev);
+	rvt_unregister_device(&dd->verbs_dev.rdi);
 
 	if (!list_empty(&dev->piowait))
 		qib_dev_err(dd, "piowait list not empty!\n");
@@ -2343,16 +1767,8 @@ void qib_unregister_ib_device(struct qib_devdata *dd)
 		qib_dev_err(dd, "txwait list not empty!\n");
 	if (!list_empty(&dev->memwait))
 		qib_dev_err(dd, "memwait list not empty!\n");
-	if (dev->dma_mr)
-		qib_dev_err(dd, "DMA MR not NULL!\n");
-
-	qps_inuse = qib_free_all_qps(dd);
-	if (qps_inuse)
-		qib_dev_err(dd, "QP memory leak! %u still in use\n",
-			    qps_inuse);
 
 	del_timer_sync(&dev->mem_timer);
-	qib_free_qpn_table(&dev->qpn_table);
 	while (!list_empty(&dev->txreq_free)) {
 		struct list_head *l = dev->txreq_free.next;
 		struct qib_verbs_txreq *tx;
@@ -2366,21 +1782,36 @@ void qib_unregister_ib_device(struct qib_devdata *dd)
 				  dd->pport->sdma_descq_cnt *
 					sizeof(struct qib_pio_header),
 				  dev->pio_hdrs, dev->pio_hdrs_phys);
-	lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
-	vfree(dev->lk_table.table);
-	kfree(dev->qp_table);
 }
 
-/*
- * This must be called with s_lock held.
+/**
+ * _qib_schedule_send - schedule progress
+ * @qp - the qp
+ *
+ * This schedules progress w/o regard to the s_flags.
+ *
+ * It is only used in post send, which doesn't hold
+ * the s_lock.
  */
-void qib_schedule_send(struct qib_qp *qp)
+void _qib_schedule_send(struct rvt_qp *qp)
 {
-	if (qib_send_ok(qp)) {
-		struct qib_ibport *ibp =
-			to_iport(qp->ibqp.device, qp->port_num);
-		struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_ibport *ibp =
+		to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_qp_priv *priv = qp->priv;
 
-		queue_work(ppd->qib_wq, &qp->s_work);
-	}
+	queue_work(ppd->qib_wq, &priv->s_work);
+}
+
+/**
+ * qib_schedule_send - schedule progress
+ * @qp - the qp
+ *
+ * This schedules qp progress.  The s_lock
+ * should be held.
+ */
+void qib_schedule_send(struct rvt_qp *qp)
+{
+	if (qib_send_ok(qp))
+		_qib_schedule_send(qp);
 }
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index 6c5e77753d85..4b76a8d59337 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -45,6 +45,8 @@
 #include <linux/completion.h>
 #include <rdma/ib_pack.h>
 #include <rdma/ib_user_verbs.h>
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_cq.h>
 
 struct qib_ctxtdata;
 struct qib_pportdata;
@@ -53,9 +55,7 @@ struct qib_verbs_txreq;
 
 #define QIB_MAX_RDMA_ATOMIC     16
 #define QIB_GUIDS_PER_PORT	5
-
-#define QPN_MAX                 (1 << 24)
-#define QPNMAP_ENTRIES          (QPN_MAX / PAGE_SIZE / BITS_PER_BYTE)
+#define QIB_PSN_SHIFT		8
 
 /*
  * Increment this value if any changes that break userspace ABI
@@ -63,12 +63,6 @@ struct qib_verbs_txreq;
  */
 #define QIB_UVERBS_ABI_VERSION       2
 
-/*
- * Define an ib_cq_notify value that is not valid so we know when CQ
- * notifications are armed.
- */
-#define IB_CQ_NONE      (IB_CQ_NEXT_COMP + 1)
-
 #define IB_SEQ_NAK	(3 << 29)
 
 /* AETH NAK opcode values */
@@ -79,17 +73,6 @@ struct qib_verbs_txreq;
 #define IB_NAK_REMOTE_OPERATIONAL_ERROR 0x63
 #define IB_NAK_INVALID_RD_REQUEST       0x64
 
-/* Flags for checking QP state (see ib_qib_state_ops[]) */
-#define QIB_POST_SEND_OK                0x01
-#define QIB_POST_RECV_OK                0x02
-#define QIB_PROCESS_RECV_OK             0x04
-#define QIB_PROCESS_SEND_OK             0x08
-#define QIB_PROCESS_NEXT_SEND_OK        0x10
-#define QIB_FLUSH_SEND			0x20
-#define QIB_FLUSH_RECV			0x40
-#define QIB_PROCESS_OR_FLUSH_SEND \
-	(QIB_PROCESS_SEND_OK | QIB_FLUSH_SEND)
-
 /* IB Performance Manager status values */
 #define IB_PMA_SAMPLE_STATUS_DONE       0x00
 #define IB_PMA_SAMPLE_STATUS_STARTED    0x01
@@ -203,468 +186,21 @@ struct qib_pio_header {
 } __packed;
 
 /*
- * There is one struct qib_mcast for each multicast GID.
- * All attached QPs are then stored as a list of
- * struct qib_mcast_qp.
+ * qib specific data structure that will be hidden from rvt after the queue pair
+ * is made common.
  */
-struct qib_mcast_qp {
-	struct list_head list;
-	struct qib_qp *qp;
-};
-
-struct qib_mcast {
-	struct rb_node rb_node;
-	union ib_gid mgid;
-	struct list_head qp_list;
-	wait_queue_head_t wait;
-	atomic_t refcount;
-	int n_attached;
-};
-
-/* Protection domain */
-struct qib_pd {
-	struct ib_pd ibpd;
-	int user;               /* non-zero if created from user space */
-};
-
-/* Address Handle */
-struct qib_ah {
-	struct ib_ah ibah;
-	struct ib_ah_attr attr;
-	atomic_t refcount;
-};
-
-/*
- * This structure is used by qib_mmap() to validate an offset
- * when an mmap() request is made.  The vm_area_struct then uses
- * this as its vm_private_data.
- */
-struct qib_mmap_info {
-	struct list_head pending_mmaps;
-	struct ib_ucontext *context;
-	void *obj;
-	__u64 offset;
-	struct kref ref;
-	unsigned size;
-};
-
-/*
- * This structure is used to contain the head pointer, tail pointer,
- * and completion queue entries as a single memory allocation so
- * it can be mmap'ed into user space.
- */
-struct qib_cq_wc {
-	u32 head;               /* index of next entry to fill */
-	u32 tail;               /* index of next ib_poll_cq() entry */
-	union {
-		/* these are actually size ibcq.cqe + 1 */
-		struct ib_uverbs_wc uqueue[0];
-		struct ib_wc kqueue[0];
-	};
-};
-
-/*
- * The completion queue structure.
- */
-struct qib_cq {
-	struct ib_cq ibcq;
-	struct kthread_work comptask;
-	struct qib_devdata *dd;
-	spinlock_t lock; /* protect changes in this struct */
-	u8 notify;
-	u8 triggered;
-	struct qib_cq_wc *queue;
-	struct qib_mmap_info *ip;
-};
-
-/*
- * A segment is a linear region of low physical memory.
- * XXX Maybe we should use phys addr here and kmap()/kunmap().
- * Used by the verbs layer.
- */
-struct qib_seg {
-	void *vaddr;
-	size_t length;
-};
-
-/* The number of qib_segs that fit in a page. */
-#define QIB_SEGSZ     (PAGE_SIZE / sizeof(struct qib_seg))
-
-struct qib_segarray {
-	struct qib_seg segs[QIB_SEGSZ];
-};
-
-struct qib_mregion {
-	struct ib_pd *pd;       /* shares refcnt of ibmr.pd */
-	u64 user_base;          /* User's address for this region */
-	u64 iova;               /* IB start address of this region */
-	size_t length;
-	u32 lkey;
-	u32 offset;             /* offset (bytes) to start of region */
-	int access_flags;
-	u32 max_segs;           /* number of qib_segs in all the arrays */
-	u32 mapsz;              /* size of the map array */
-	u8  page_shift;         /* 0 - non unform/non powerof2 sizes */
-	u8  lkey_published;     /* in global table */
-	struct completion comp; /* complete when refcount goes to zero */
-	struct rcu_head list;
-	atomic_t refcount;
-	struct qib_segarray *map[0];    /* the segments */
-};
-
-/*
- * These keep track of the copy progress within a memory region.
- * Used by the verbs layer.
- */
-struct qib_sge {
-	struct qib_mregion *mr;
-	void *vaddr;            /* kernel virtual address of segment */
-	u32 sge_length;         /* length of the SGE */
-	u32 length;             /* remaining length of the segment */
-	u16 m;                  /* current index: mr->map[m] */
-	u16 n;                  /* current index: mr->map[m]->segs[n] */
-};
-
-/* Memory region */
-struct qib_mr {
-	struct ib_mr ibmr;
-	struct ib_umem *umem;
-	u64 *pages;
-	u32 npages;
-	struct qib_mregion mr;  /* must be last */
-};
-
-/*
- * Send work request queue entry.
- * The size of the sg_list is determined when the QP is created and stored
- * in qp->s_max_sge.
- */
-struct qib_swqe {
-	union {
-		struct ib_send_wr wr;   /* don't use wr.sg_list */
-		struct ib_ud_wr ud_wr;
-		struct ib_reg_wr reg_wr;
-		struct ib_rdma_wr rdma_wr;
-		struct ib_atomic_wr atomic_wr;
-	};
-	u32 psn;                /* first packet sequence number */
-	u32 lpsn;               /* last packet sequence number */
-	u32 ssn;                /* send sequence number */
-	u32 length;             /* total length of data in sg_list */
-	struct qib_sge sg_list[0];
-};
-
-/*
- * Receive work request queue entry.
- * The size of the sg_list is determined when the QP (or SRQ) is created
- * and stored in qp->r_rq.max_sge (or srq->rq.max_sge).
- */
-struct qib_rwqe {
-	u64 wr_id;
-	u8 num_sge;
-	struct ib_sge sg_list[0];
-};
-
-/*
- * This structure is used to contain the head pointer, tail pointer,
- * and receive work queue entries as a single memory allocation so
- * it can be mmap'ed into user space.
- * Note that the wq array elements are variable size so you can't
- * just index into the array to get the N'th element;
- * use get_rwqe_ptr() instead.
- */
-struct qib_rwq {
-	u32 head;               /* new work requests posted to the head */
-	u32 tail;               /* receives pull requests from here. */
-	struct qib_rwqe wq[0];
-};
-
-struct qib_rq {
-	struct qib_rwq *wq;
-	u32 size;               /* size of RWQE array */
-	u8 max_sge;
-	spinlock_t lock /* protect changes in this struct */
-		____cacheline_aligned_in_smp;
-};
-
-struct qib_srq {
-	struct ib_srq ibsrq;
-	struct qib_rq rq;
-	struct qib_mmap_info *ip;
-	/* send signal when number of RWQEs < limit */
-	u32 limit;
-};
-
-struct qib_sge_state {
-	struct qib_sge *sg_list;      /* next SGE to be used if any */
-	struct qib_sge sge;   /* progress state for the current SGE */
-	u32 total_len;
-	u8 num_sge;
-};
-
-/*
- * This structure holds the information that the send tasklet needs
- * to send a RDMA read response or atomic operation.
- */
-struct qib_ack_entry {
-	u8 opcode;
-	u8 sent;
-	u32 psn;
-	u32 lpsn;
-	union {
-		struct qib_sge rdma_sge;
-		u64 atomic_data;
-	};
-};
-
-/*
- * Variables prefixed with s_ are for the requester (sender).
- * Variables prefixed with r_ are for the responder (receiver).
- * Variables prefixed with ack_ are for responder replies.
- *
- * Common variables are protected by both r_rq.lock and s_lock in that order
- * which only happens in modify_qp() or changing the QP 'state'.
- */
-struct qib_qp {
-	struct ib_qp ibqp;
-	/* read mostly fields above and below */
-	struct ib_ah_attr remote_ah_attr;
-	struct ib_ah_attr alt_ah_attr;
-	struct qib_qp __rcu *next;            /* link list for QPN hash table */
-	struct qib_swqe *s_wq;  /* send work queue */
-	struct qib_mmap_info *ip;
-	struct qib_ib_header *s_hdr;     /* next packet header to send */
-	unsigned long timeout_jiffies;  /* computed from timeout */
-
-	enum ib_mtu path_mtu;
-	u32 remote_qpn;
-	u32 pmtu;		/* decoded from path_mtu */
-	u32 qkey;               /* QKEY for this QP (for UD or RD) */
-	u32 s_size;             /* send work queue size */
-	u32 s_rnr_timeout;      /* number of milliseconds for RNR timeout */
-
-	u8 state;               /* QP state */
-	u8 qp_access_flags;
-	u8 alt_timeout;         /* Alternate path timeout for this QP */
-	u8 timeout;             /* Timeout for this QP */
-	u8 s_srate;
-	u8 s_mig_state;
-	u8 port_num;
-	u8 s_pkey_index;        /* PKEY index to use */
-	u8 s_alt_pkey_index;    /* Alternate path PKEY index to use */
-	u8 r_max_rd_atomic;     /* max number of RDMA read/atomic to receive */
-	u8 s_max_rd_atomic;     /* max number of RDMA read/atomic to send */
-	u8 s_retry_cnt;         /* number of times to retry */
-	u8 s_rnr_retry_cnt;
-	u8 r_min_rnr_timer;     /* retry timeout value for RNR NAKs */
-	u8 s_max_sge;           /* size of s_wq->sg_list */
-	u8 s_draining;
-
-	/* start of read/write fields */
-
-	atomic_t refcount ____cacheline_aligned_in_smp;
-	wait_queue_head_t wait;
-
-
-	struct qib_ack_entry s_ack_queue[QIB_MAX_RDMA_ATOMIC + 1]
-		____cacheline_aligned_in_smp;
-	struct qib_sge_state s_rdma_read_sge;
-
-	spinlock_t r_lock ____cacheline_aligned_in_smp;      /* used for APM */
-	unsigned long r_aflags;
-	u64 r_wr_id;            /* ID for current receive WQE */
-	u32 r_ack_psn;          /* PSN for next ACK or atomic ACK */
-	u32 r_len;              /* total length of r_sge */
-	u32 r_rcv_len;          /* receive data len processed */
-	u32 r_psn;              /* expected rcv packet sequence number */
-	u32 r_msn;              /* message sequence number */
-
-	u8 r_state;             /* opcode of last packet received */
-	u8 r_flags;
-	u8 r_head_ack_queue;    /* index into s_ack_queue[] */
-
-	struct list_head rspwait;       /* link for waititing to respond */
-
-	struct qib_sge_state r_sge;     /* current receive data */
-	struct qib_rq r_rq;             /* receive work queue */
-
-	spinlock_t s_lock ____cacheline_aligned_in_smp;
-	struct qib_sge_state *s_cur_sge;
-	u32 s_flags;
-	struct qib_verbs_txreq *s_tx;
-	struct qib_swqe *s_wqe;
-	struct qib_sge_state s_sge;     /* current send request data */
-	struct qib_mregion *s_rdma_mr;
-	atomic_t s_dma_busy;
-	u32 s_cur_size;         /* size of send packet in bytes */
-	u32 s_len;              /* total length of s_sge */
-	u32 s_rdma_read_len;    /* total length of s_rdma_read_sge */
-	u32 s_next_psn;         /* PSN for next request */
-	u32 s_last_psn;         /* last response PSN processed */
-	u32 s_sending_psn;      /* lowest PSN that is being sent */
-	u32 s_sending_hpsn;     /* highest PSN that is being sent */
-	u32 s_psn;              /* current packet sequence number */
-	u32 s_ack_rdma_psn;     /* PSN for sending RDMA read responses */
-	u32 s_ack_psn;          /* PSN for acking sends and RDMA writes */
-	u32 s_head;             /* new entries added here */
-	u32 s_tail;             /* next entry to process */
-	u32 s_cur;              /* current work queue entry */
-	u32 s_acked;            /* last un-ACK'ed entry */
-	u32 s_last;             /* last completed entry */
-	u32 s_ssn;              /* SSN of tail entry */
-	u32 s_lsn;              /* limit sequence number (credit) */
-	u16 s_hdrwords;         /* size of s_hdr in 32 bit words */
-	u16 s_rdma_ack_cnt;
-	u8 s_state;             /* opcode of last packet sent */
-	u8 s_ack_state;         /* opcode of packet to ACK */
-	u8 s_nak_state;         /* non-zero if NAK is pending */
-	u8 r_nak_state;         /* non-zero if NAK is pending */
-	u8 s_retry;             /* requester retry counter */
-	u8 s_rnr_retry;         /* requester RNR retry counter */
-	u8 s_num_rd_atomic;     /* number of RDMA read/atomic pending */
-	u8 s_tail_ack_queue;    /* index into s_ack_queue[] */
-
-	struct qib_sge_state s_ack_rdma_sge;
-	struct timer_list s_timer;
+struct qib_qp_priv {
+	struct qib_ib_header *s_hdr;    /* next packet header to send */
 	struct list_head iowait;        /* link for wait PIO buf */
-
+	atomic_t s_dma_busy;
+	struct qib_verbs_txreq *s_tx;
 	struct work_struct s_work;
-
 	wait_queue_head_t wait_dma;
-
-	struct qib_sge r_sg_list[0] /* verified SGEs */
-		____cacheline_aligned_in_smp;
+	struct rvt_qp *owner;
 };
 
-/*
- * Atomic bit definitions for r_aflags.
- */
-#define QIB_R_WRID_VALID        0
-#define QIB_R_REWIND_SGE        1
-
-/*
- * Bit definitions for r_flags.
- */
-#define QIB_R_REUSE_SGE 0x01
-#define QIB_R_RDMAR_SEQ 0x02
-#define QIB_R_RSP_NAK   0x04
-#define QIB_R_RSP_SEND  0x08
-#define QIB_R_COMM_EST  0x10
-
-/*
- * Bit definitions for s_flags.
- *
- * QIB_S_SIGNAL_REQ_WR - set if QP send WRs contain completion signaled
- * QIB_S_BUSY - send tasklet is processing the QP
- * QIB_S_TIMER - the RC retry timer is active
- * QIB_S_ACK_PENDING - an ACK is waiting to be sent after RDMA read/atomics
- * QIB_S_WAIT_FENCE - waiting for all prior RDMA read or atomic SWQEs
- *                         before processing the next SWQE
- * QIB_S_WAIT_RDMAR - waiting for a RDMA read or atomic SWQE to complete
- *                         before processing the next SWQE
- * QIB_S_WAIT_RNR - waiting for RNR timeout
- * QIB_S_WAIT_SSN_CREDIT - waiting for RC credits to process next SWQE
- * QIB_S_WAIT_DMA - waiting for send DMA queue to drain before generating
- *                  next send completion entry not via send DMA
- * QIB_S_WAIT_PIO - waiting for a send buffer to be available
- * QIB_S_WAIT_TX - waiting for a struct qib_verbs_txreq to be available
- * QIB_S_WAIT_DMA_DESC - waiting for DMA descriptors to be available
- * QIB_S_WAIT_KMEM - waiting for kernel memory to be available
- * QIB_S_WAIT_PSN - waiting for a packet to exit the send DMA queue
- * QIB_S_WAIT_ACK - waiting for an ACK packet before sending more requests
- * QIB_S_SEND_ONE - send one packet, request ACK, then wait for ACK
- */
-#define QIB_S_SIGNAL_REQ_WR	0x0001
-#define QIB_S_BUSY		0x0002
-#define QIB_S_TIMER		0x0004
-#define QIB_S_RESP_PENDING	0x0008
-#define QIB_S_ACK_PENDING	0x0010
-#define QIB_S_WAIT_FENCE	0x0020
-#define QIB_S_WAIT_RDMAR	0x0040
-#define QIB_S_WAIT_RNR		0x0080
-#define QIB_S_WAIT_SSN_CREDIT	0x0100
-#define QIB_S_WAIT_DMA		0x0200
-#define QIB_S_WAIT_PIO		0x0400
-#define QIB_S_WAIT_TX		0x0800
-#define QIB_S_WAIT_DMA_DESC	0x1000
-#define QIB_S_WAIT_KMEM		0x2000
-#define QIB_S_WAIT_PSN		0x4000
-#define QIB_S_WAIT_ACK		0x8000
-#define QIB_S_SEND_ONE		0x10000
-#define QIB_S_UNLIMITED_CREDIT	0x20000
-
-/*
- * Wait flags that would prevent any packet type from being sent.
- */
-#define QIB_S_ANY_WAIT_IO (QIB_S_WAIT_PIO | QIB_S_WAIT_TX | \
-	QIB_S_WAIT_DMA_DESC | QIB_S_WAIT_KMEM)
-
-/*
- * Wait flags that would prevent send work requests from making progress.
- */
-#define QIB_S_ANY_WAIT_SEND (QIB_S_WAIT_FENCE | QIB_S_WAIT_RDMAR | \
-	QIB_S_WAIT_RNR | QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_DMA | \
-	QIB_S_WAIT_PSN | QIB_S_WAIT_ACK)
-
-#define QIB_S_ANY_WAIT (QIB_S_ANY_WAIT_IO | QIB_S_ANY_WAIT_SEND)
-
 #define QIB_PSN_CREDIT  16
 
-/*
- * Since struct qib_swqe is not a fixed size, we can't simply index into
- * struct qib_qp.s_wq.  This function does the array index computation.
- */
-static inline struct qib_swqe *get_swqe_ptr(struct qib_qp *qp,
-					      unsigned n)
-{
-	return (struct qib_swqe *)((char *)qp->s_wq +
-				     (sizeof(struct qib_swqe) +
-				      qp->s_max_sge *
-				      sizeof(struct qib_sge)) * n);
-}
-
-/*
- * Since struct qib_rwqe is not a fixed size, we can't simply index into
- * struct qib_rwq.wq.  This function does the array index computation.
- */
-static inline struct qib_rwqe *get_rwqe_ptr(struct qib_rq *rq, unsigned n)
-{
-	return (struct qib_rwqe *)
-		((char *) rq->wq->wq +
-		 (sizeof(struct qib_rwqe) +
-		  rq->max_sge * sizeof(struct ib_sge)) * n);
-}
-
-/*
- * QPN-map pages start out as NULL, they get allocated upon
- * first use and are never deallocated. This way,
- * large bitmaps are not allocated unless large numbers of QPs are used.
- */
-struct qpn_map {
-	void *page;
-};
-
-struct qib_qpn_table {
-	spinlock_t lock; /* protect changes in this struct */
-	unsigned flags;         /* flags for QP0/1 allocated for each port */
-	u32 last;               /* last QP number allocated */
-	u32 nmaps;              /* size of the map table */
-	u16 limit;
-	u16 mask;
-	/* bit map of free QP numbers other than 0/1 */
-	struct qpn_map map[QPNMAP_ENTRIES];
-};
-
-#define MAX_LKEY_TABLE_BITS 23
-
-struct qib_lkey_table {
-	spinlock_t lock; /* protect changes in this struct */
-	u32 next;               /* next unused index (speeds search) */
-	u32 gen;                /* generation count */
-	u32 max;                /* size of the table */
-	struct qib_mregion __rcu **table;
-};
-
 struct qib_opcode_stats {
 	u64 n_packets;          /* number of packets */
 	u64 n_bytes;            /* total number of bytes */
@@ -682,21 +218,9 @@ struct qib_pma_counters {
 };
 
 struct qib_ibport {
-	struct qib_qp __rcu *qp0;
-	struct qib_qp __rcu *qp1;
-	struct ib_mad_agent *send_agent;	/* agent for SMI (traps) */
-	struct qib_ah *sm_ah;
-	struct qib_ah *smi_ah;
-	struct rb_root mcast_tree;
-	spinlock_t lock;		/* protect changes in this struct */
-
-	/* non-zero when timer is set */
-	unsigned long mkey_lease_timeout;
-	unsigned long trap_timeout;
-	__be64 gid_prefix;      /* in network order */
-	__be64 mkey;
+	struct rvt_ibport rvp;
+	struct rvt_ah *smi_ah;
 	__be64 guids[QIB_GUIDS_PER_PORT	- 1];	/* writable GUIDs */
-	u64 tid;		/* TID for traps */
 	struct qib_pma_counters __percpu *pmastats;
 	u64 z_unicast_xmit;     /* starting count for PMA */
 	u64 z_unicast_rcv;      /* starting count for PMA */
@@ -715,82 +239,25 @@ struct qib_ibport {
 	u32 z_local_link_integrity_errors;      /* starting count for PMA */
 	u32 z_excessive_buffer_overrun_errors;  /* starting count for PMA */
 	u32 z_vl15_dropped;                     /* starting count for PMA */
-	u32 n_rc_resends;
-	u32 n_rc_acks;
-	u32 n_rc_qacks;
-	u32 n_rc_delayed_comp;
-	u32 n_seq_naks;
-	u32 n_rdma_seq;
-	u32 n_rnr_naks;
-	u32 n_other_naks;
-	u32 n_loop_pkts;
-	u32 n_pkt_drops;
-	u32 n_vl15_dropped;
-	u32 n_rc_timeouts;
-	u32 n_dmawait;
-	u32 n_unaligned;
-	u32 n_rc_dupreq;
-	u32 n_rc_seqnak;
-	u32 port_cap_flags;
-	u32 pma_sample_start;
-	u32 pma_sample_interval;
-	__be16 pma_counter_select[5];
-	u16 pma_tag;
-	u16 pkey_violations;
-	u16 qkey_violations;
-	u16 mkey_violations;
-	u16 mkey_lease_period;
-	u16 sm_lid;
-	u16 repress_traps;
-	u8 sm_sl;
-	u8 mkeyprot;
-	u8 subnet_timeout;
-	u8 vl_high_limit;
 	u8 sl_to_vl[16];
-
 };
 
-
 struct qib_ibdev {
-	struct ib_device ibdev;
-	struct list_head pending_mmaps;
-	spinlock_t mmap_offset_lock; /* protect mmap_offset */
-	u32 mmap_offset;
-	struct qib_mregion __rcu *dma_mr;
-
-	/* QP numbers are shared by all IB ports */
-	struct qib_qpn_table qpn_table;
-	struct qib_lkey_table lk_table;
+	struct rvt_dev_info rdi;
+
 	struct list_head piowait;       /* list for wait PIO buf */
 	struct list_head dmawait;	/* list for wait DMA */
 	struct list_head txwait;        /* list for wait qib_verbs_txreq */
 	struct list_head memwait;       /* list for wait kernel memory */
 	struct list_head txreq_free;
 	struct timer_list mem_timer;
-	struct qib_qp __rcu **qp_table;
 	struct qib_pio_header *pio_hdrs;
 	dma_addr_t pio_hdrs_phys;
-	/* list of QPs waiting for RNR timer */
-	spinlock_t pending_lock; /* protect wait lists, PMA counters, etc. */
-	u32 qp_table_size; /* size of the hash table */
 	u32 qp_rnd; /* random bytes for hash */
-	spinlock_t qpt_lock;
 
 	u32 n_piowait;
 	u32 n_txwait;
 
-	u32 n_pds_allocated;    /* number of PDs allocated for device */
-	spinlock_t n_pds_lock;
-	u32 n_ahs_allocated;    /* number of AHs allocated for device */
-	spinlock_t n_ahs_lock;
-	u32 n_cqs_allocated;    /* number of CQs allocated for device */
-	spinlock_t n_cqs_lock;
-	u32 n_qps_allocated;    /* number of QPs allocated for device */
-	spinlock_t n_qps_lock;
-	u32 n_srqs_allocated;   /* number of SRQs allocated for device */
-	spinlock_t n_srqs_lock;
-	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
-	spinlock_t n_mcast_grps_lock;
 #ifdef CONFIG_DEBUG_FS
 	/* per HCA debugfs */
 	struct dentry *qib_ibdev_dbg;
@@ -813,56 +280,27 @@ struct qib_verbs_counters {
 	u32 vl15_dropped;
 };
 
-static inline struct qib_mr *to_imr(struct ib_mr *ibmr)
-{
-	return container_of(ibmr, struct qib_mr, ibmr);
-}
-
-static inline struct qib_pd *to_ipd(struct ib_pd *ibpd)
-{
-	return container_of(ibpd, struct qib_pd, ibpd);
-}
-
-static inline struct qib_ah *to_iah(struct ib_ah *ibah)
-{
-	return container_of(ibah, struct qib_ah, ibah);
-}
-
-static inline struct qib_cq *to_icq(struct ib_cq *ibcq)
-{
-	return container_of(ibcq, struct qib_cq, ibcq);
-}
-
-static inline struct qib_srq *to_isrq(struct ib_srq *ibsrq)
-{
-	return container_of(ibsrq, struct qib_srq, ibsrq);
-}
-
-static inline struct qib_qp *to_iqp(struct ib_qp *ibqp)
-{
-	return container_of(ibqp, struct qib_qp, ibqp);
-}
-
 static inline struct qib_ibdev *to_idev(struct ib_device *ibdev)
 {
-	return container_of(ibdev, struct qib_ibdev, ibdev);
+	struct rvt_dev_info *rdi;
+
+	rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
+	return container_of(rdi, struct qib_ibdev, rdi);
 }
 
 /*
  * Send if not busy or waiting for I/O and either
  * a RC response is pending or we can process send work requests.
  */
-static inline int qib_send_ok(struct qib_qp *qp)
+static inline int qib_send_ok(struct rvt_qp *qp)
 {
-	return !(qp->s_flags & (QIB_S_BUSY | QIB_S_ANY_WAIT_IO)) &&
-		(qp->s_hdrwords || (qp->s_flags & QIB_S_RESP_PENDING) ||
-		 !(qp->s_flags & QIB_S_ANY_WAIT_SEND));
+	return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) &&
+		(qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) ||
+		 !(qp->s_flags & RVT_S_ANY_WAIT_SEND));
 }
 
-/*
- * This must be called with s_lock held.
- */
-void qib_schedule_send(struct qib_qp *qp);
+void _qib_schedule_send(struct rvt_qp *qp);
+void qib_schedule_send(struct rvt_qp *qp);
 
 static inline int qib_pkey_ok(u16 pkey1, u16 pkey2)
 {
@@ -878,7 +316,7 @@ static inline int qib_pkey_ok(u16 pkey1, u16 pkey2)
 
 void qib_bad_pqkey(struct qib_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
 		   u32 qp1, u32 qp2, __be16 lid1, __be16 lid2);
-void qib_cap_mask_chg(struct qib_ibport *ibp);
+void qib_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num);
 void qib_sys_guid_chg(struct qib_ibport *ibp);
 void qib_node_desc_chg(struct qib_ibport *ibp);
 int qib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
@@ -886,8 +324,8 @@ int qib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
 		    const struct ib_mad_hdr *in, size_t in_mad_size,
 		    struct ib_mad_hdr *out, size_t *out_mad_size,
 		    u16 *out_mad_pkey_index);
-int qib_create_agents(struct qib_ibdev *dev);
-void qib_free_agents(struct qib_ibdev *dev);
+void qib_notify_create_mad_agent(struct rvt_dev_info *rdi, int port_idx);
+void qib_notify_free_mad_agent(struct rvt_dev_info *rdi, int port_idx);
 
 /*
  * Compare the lower 24 bits of the two values.
@@ -898,8 +336,6 @@ static inline int qib_cmp24(u32 a, u32 b)
 	return (((int) a) - ((int) b)) << 8;
 }
 
-struct qib_mcast *qib_mcast_find(struct qib_ibport *ibp, union ib_gid *mgid);
-
 int qib_snapshot_counters(struct qib_pportdata *ppd, u64 *swords,
 			  u64 *rwords, u64 *spkts, u64 *rpkts,
 			  u64 *xmit_wait);
@@ -907,35 +343,17 @@ int qib_snapshot_counters(struct qib_pportdata *ppd, u64 *swords,
 int qib_get_counters(struct qib_pportdata *ppd,
 		     struct qib_verbs_counters *cntrs);
 
-int qib_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
-
-int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
-
-int qib_mcast_tree_empty(struct qib_ibport *ibp);
-
-__be32 qib_compute_aeth(struct qib_qp *qp);
-
-struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn);
-
-struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
-			    struct ib_qp_init_attr *init_attr,
-			    struct ib_udata *udata);
-
-int qib_destroy_qp(struct ib_qp *ibqp);
-
-int qib_error_qp(struct qib_qp *qp, enum ib_wc_status err);
-
-int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		  int attr_mask, struct ib_udata *udata);
-
-int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		 int attr_mask, struct ib_qp_init_attr *init_attr);
-
-unsigned qib_free_all_qps(struct qib_devdata *dd);
+__be32 qib_compute_aeth(struct rvt_qp *qp);
 
-void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt);
-
-void qib_free_qpn_table(struct qib_qpn_table *qpt);
+/*
+ * Functions provided by qib driver for rdmavt to use
+ */
+unsigned qib_free_all_qps(struct rvt_dev_info *rdi);
+void *qib_qp_priv_alloc(struct rvt_dev_info *rdi, struct rvt_qp *qp, gfp_t gfp);
+void qib_qp_priv_free(struct rvt_dev_info *rdi, struct rvt_qp *qp);
+void qib_notify_qp_reset(struct rvt_qp *qp);
+int qib_alloc_qpn(struct rvt_dev_info *rdi, struct rvt_qpn_table *qpt,
+		  enum ib_qp_type type, u8 port, gfp_t gfp);
 
 #ifdef CONFIG_DEBUG_FS
 
@@ -949,7 +367,7 @@ void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter);
 
 #endif
 
-void qib_get_credit(struct qib_qp *qp, u32 aeth);
+void qib_get_credit(struct rvt_qp *qp, u32 aeth);
 
 unsigned qib_pkt_delay(u32 plen, u8 snd_mult, u8 rcv_mult);
 
@@ -957,166 +375,66 @@ void qib_verbs_sdma_desc_avail(struct qib_pportdata *ppd, unsigned avail);
 
 void qib_put_txreq(struct qib_verbs_txreq *tx);
 
-int qib_verbs_send(struct qib_qp *qp, struct qib_ib_header *hdr,
-		   u32 hdrwords, struct qib_sge_state *ss, u32 len);
+int qib_verbs_send(struct rvt_qp *qp, struct qib_ib_header *hdr,
+		   u32 hdrwords, struct rvt_sge_state *ss, u32 len);
 
-void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length,
+void qib_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
 		  int release);
 
-void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release);
+void qib_skip_sge(struct rvt_sge_state *ss, u32 length, int release);
 
 void qib_uc_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp);
 
 void qib_rc_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp);
 
 int qib_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
 
+int qib_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
+
 struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid);
 
 void qib_rc_rnr_retry(unsigned long arg);
 
-void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr);
+void qib_rc_send_complete(struct rvt_qp *qp, struct qib_ib_header *hdr);
 
-void qib_rc_error(struct qib_qp *qp, enum ib_wc_status err);
+void qib_rc_error(struct rvt_qp *qp, enum ib_wc_status err);
 
-int qib_post_ud_send(struct qib_qp *qp, struct ib_send_wr *wr);
+int qib_post_ud_send(struct rvt_qp *qp, struct ib_send_wr *wr);
 
 void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
-
-int qib_alloc_lkey(struct qib_mregion *mr, int dma_region);
-
-void qib_free_lkey(struct qib_mregion *mr);
-
-int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
-		struct qib_sge *isge, struct ib_sge *sge, int acc);
-
-int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
-		u32 len, u64 vaddr, u32 rkey, int acc);
-
-int qib_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
-			 struct ib_recv_wr **bad_wr);
-
-struct ib_srq *qib_create_srq(struct ib_pd *ibpd,
-			      struct ib_srq_init_attr *srq_init_attr,
-			      struct ib_udata *udata);
-
-int qib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
-		   enum ib_srq_attr_mask attr_mask,
-		   struct ib_udata *udata);
-
-int qib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr);
-
-int qib_destroy_srq(struct ib_srq *ibsrq);
-
-int qib_cq_init(struct qib_devdata *dd);
-
-void qib_cq_exit(struct qib_devdata *dd);
-
-void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int sig);
-
-int qib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
-
-struct ib_cq *qib_create_cq(struct ib_device *ibdev,
-			    const struct ib_cq_init_attr *attr,
-			    struct ib_ucontext *context,
-			    struct ib_udata *udata);
-
-int qib_destroy_cq(struct ib_cq *ibcq);
-
-int qib_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags);
-
-int qib_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
-
-struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc);
-
-struct ib_mr *qib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
-			      u64 virt_addr, int mr_access_flags,
-			      struct ib_udata *udata);
-
-int qib_dereg_mr(struct ib_mr *ibmr);
-
-struct ib_mr *qib_alloc_mr(struct ib_pd *pd,
-			   enum ib_mr_type mr_type,
-			   u32 max_entries);
-
-int qib_map_mr_sg(struct ib_mr *ibmr,
-		  struct scatterlist *sg,
-		  int sg_nents);
-
-int qib_reg_mr(struct qib_qp *qp, struct ib_reg_wr *wr);
-
-struct ib_fmr *qib_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
-			     struct ib_fmr_attr *fmr_attr);
-
-int qib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
-		     int list_len, u64 iova);
-
-int qib_unmap_fmr(struct list_head *fmr_list);
-
-int qib_dealloc_fmr(struct ib_fmr *ibfmr);
-
-static inline void qib_get_mr(struct qib_mregion *mr)
-{
-	atomic_inc(&mr->refcount);
-}
+		int has_grh, void *data, u32 tlen, struct rvt_qp *qp);
 
 void mr_rcu_callback(struct rcu_head *list);
 
-static inline void qib_put_mr(struct qib_mregion *mr)
-{
-	if (unlikely(atomic_dec_and_test(&mr->refcount)))
-		call_rcu(&mr->list, mr_rcu_callback);
-}
-
-static inline void qib_put_ss(struct qib_sge_state *ss)
-{
-	while (ss->num_sge) {
-		qib_put_mr(ss->sge.mr);
-		if (--ss->num_sge)
-			ss->sge = *ss->sg_list++;
-	}
-}
-
-
-void qib_release_mmap_info(struct kref *ref);
+int qib_get_rwqe(struct rvt_qp *qp, int wr_id_only);
 
-struct qib_mmap_info *qib_create_mmap_info(struct qib_ibdev *dev, u32 size,
-					   struct ib_ucontext *context,
-					   void *obj);
-
-void qib_update_mmap_info(struct qib_ibdev *dev, struct qib_mmap_info *ip,
-			  u32 size, void *obj);
-
-int qib_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
-
-int qib_get_rwqe(struct qib_qp *qp, int wr_id_only);
-
-void qib_migrate_qp(struct qib_qp *qp);
+void qib_migrate_qp(struct rvt_qp *qp);
 
 int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
-		      int has_grh, struct qib_qp *qp, u32 bth0);
+		      int has_grh, struct rvt_qp *qp, u32 bth0);
 
 u32 qib_make_grh(struct qib_ibport *ibp, struct ib_grh *hdr,
 		 struct ib_global_route *grh, u32 hwords, u32 nwords);
 
-void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
+void qib_make_ruc_header(struct rvt_qp *qp, struct qib_other_headers *ohdr,
 			 u32 bth0, u32 bth2);
 
-void qib_do_send(struct work_struct *work);
+void _qib_do_send(struct work_struct *work);
+
+void qib_do_send(struct rvt_qp *qp);
 
-void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe,
+void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 		       enum ib_wc_status status);
 
-void qib_send_rc_ack(struct qib_qp *qp);
+void qib_send_rc_ack(struct rvt_qp *qp);
 
-int qib_make_rc_req(struct qib_qp *qp);
+int qib_make_rc_req(struct rvt_qp *qp);
 
-int qib_make_uc_req(struct qib_qp *qp);
+int qib_make_uc_req(struct rvt_qp *qp);
 
-int qib_make_ud_req(struct qib_qp *qp);
+int qib_make_ud_req(struct rvt_qp *qp);
 
 int qib_register_ib_device(struct qib_devdata *);
 
@@ -1150,11 +468,11 @@ extern const enum ib_wc_opcode ib_qib_wc_opcode[];
 #define IB_PHYSPORTSTATE_CFG_ENH 0x10
 #define IB_PHYSPORTSTATE_CFG_WAIT_ENH 0x13
 
-extern const int ib_qib_state_ops[];
+extern const int ib_rvt_state_ops[];
 
 extern __be64 ib_qib_sys_image_guid;    /* in network order */
 
-extern unsigned int ib_qib_lkey_table_size;
+extern unsigned int ib_rvt_lkey_table_size;
 
 extern unsigned int ib_qib_max_cqes;
 
@@ -1178,6 +496,4 @@ extern unsigned int ib_qib_max_srq_wrs;
 
 extern const u32 ib_qib_rnr_table[];
 
-extern struct ib_dma_mapping_ops qib_dma_mapping_ops;
-
 #endif                          /* QIB_VERBS_H */
diff --git a/drivers/infiniband/hw/qib/qib_verbs_mcast.c b/drivers/infiniband/hw/qib/qib_verbs_mcast.c
deleted file mode 100644
index b2fb5286dbd9..000000000000
--- a/drivers/infiniband/hw/qib/qib_verbs_mcast.c
+++ /dev/null
@@ -1,363 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
- * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include <linux/rculist.h>
-
-#include "qib.h"
-
-/**
- * qib_mcast_qp_alloc - alloc a struct to link a QP to mcast GID struct
- * @qp: the QP to link
- */
-static struct qib_mcast_qp *qib_mcast_qp_alloc(struct qib_qp *qp)
-{
-	struct qib_mcast_qp *mqp;
-
-	mqp = kmalloc(sizeof(*mqp), GFP_KERNEL);
-	if (!mqp)
-		goto bail;
-
-	mqp->qp = qp;
-	atomic_inc(&qp->refcount);
-
-bail:
-	return mqp;
-}
-
-static void qib_mcast_qp_free(struct qib_mcast_qp *mqp)
-{
-	struct qib_qp *qp = mqp->qp;
-
-	/* Notify qib_destroy_qp() if it is waiting. */
-	if (atomic_dec_and_test(&qp->refcount))
-		wake_up(&qp->wait);
-
-	kfree(mqp);
-}
-
-/**
- * qib_mcast_alloc - allocate the multicast GID structure
- * @mgid: the multicast GID
- *
- * A list of QPs will be attached to this structure.
- */
-static struct qib_mcast *qib_mcast_alloc(union ib_gid *mgid)
-{
-	struct qib_mcast *mcast;
-
-	mcast = kmalloc(sizeof(*mcast), GFP_KERNEL);
-	if (!mcast)
-		goto bail;
-
-	mcast->mgid = *mgid;
-	INIT_LIST_HEAD(&mcast->qp_list);
-	init_waitqueue_head(&mcast->wait);
-	atomic_set(&mcast->refcount, 0);
-	mcast->n_attached = 0;
-
-bail:
-	return mcast;
-}
-
-static void qib_mcast_free(struct qib_mcast *mcast)
-{
-	struct qib_mcast_qp *p, *tmp;
-
-	list_for_each_entry_safe(p, tmp, &mcast->qp_list, list)
-		qib_mcast_qp_free(p);
-
-	kfree(mcast);
-}
-
-/**
- * qib_mcast_find - search the global table for the given multicast GID
- * @ibp: the IB port structure
- * @mgid: the multicast GID to search for
- *
- * Returns NULL if not found.
- *
- * The caller is responsible for decrementing the reference count if found.
- */
-struct qib_mcast *qib_mcast_find(struct qib_ibport *ibp, union ib_gid *mgid)
-{
-	struct rb_node *n;
-	unsigned long flags;
-	struct qib_mcast *mcast;
-
-	spin_lock_irqsave(&ibp->lock, flags);
-	n = ibp->mcast_tree.rb_node;
-	while (n) {
-		int ret;
-
-		mcast = rb_entry(n, struct qib_mcast, rb_node);
-
-		ret = memcmp(mgid->raw, mcast->mgid.raw,
-			     sizeof(union ib_gid));
-		if (ret < 0)
-			n = n->rb_left;
-		else if (ret > 0)
-			n = n->rb_right;
-		else {
-			atomic_inc(&mcast->refcount);
-			spin_unlock_irqrestore(&ibp->lock, flags);
-			goto bail;
-		}
-	}
-	spin_unlock_irqrestore(&ibp->lock, flags);
-
-	mcast = NULL;
-
-bail:
-	return mcast;
-}
-
-/**
- * qib_mcast_add - insert mcast GID into table and attach QP struct
- * @mcast: the mcast GID table
- * @mqp: the QP to attach
- *
- * Return zero if both were added.  Return EEXIST if the GID was already in
- * the table but the QP was added.  Return ESRCH if the QP was already
- * attached and neither structure was added.
- */
-static int qib_mcast_add(struct qib_ibdev *dev, struct qib_ibport *ibp,
-			 struct qib_mcast *mcast, struct qib_mcast_qp *mqp)
-{
-	struct rb_node **n = &ibp->mcast_tree.rb_node;
-	struct rb_node *pn = NULL;
-	int ret;
-
-	spin_lock_irq(&ibp->lock);
-
-	while (*n) {
-		struct qib_mcast *tmcast;
-		struct qib_mcast_qp *p;
-
-		pn = *n;
-		tmcast = rb_entry(pn, struct qib_mcast, rb_node);
-
-		ret = memcmp(mcast->mgid.raw, tmcast->mgid.raw,
-			     sizeof(union ib_gid));
-		if (ret < 0) {
-			n = &pn->rb_left;
-			continue;
-		}
-		if (ret > 0) {
-			n = &pn->rb_right;
-			continue;
-		}
-
-		/* Search the QP list to see if this is already there. */
-		list_for_each_entry_rcu(p, &tmcast->qp_list, list) {
-			if (p->qp == mqp->qp) {
-				ret = ESRCH;
-				goto bail;
-			}
-		}
-		if (tmcast->n_attached == ib_qib_max_mcast_qp_attached) {
-			ret = ENOMEM;
-			goto bail;
-		}
-
-		tmcast->n_attached++;
-
-		list_add_tail_rcu(&mqp->list, &tmcast->qp_list);
-		ret = EEXIST;
-		goto bail;
-	}
-
-	spin_lock(&dev->n_mcast_grps_lock);
-	if (dev->n_mcast_grps_allocated == ib_qib_max_mcast_grps) {
-		spin_unlock(&dev->n_mcast_grps_lock);
-		ret = ENOMEM;
-		goto bail;
-	}
-
-	dev->n_mcast_grps_allocated++;
-	spin_unlock(&dev->n_mcast_grps_lock);
-
-	mcast->n_attached++;
-
-	list_add_tail_rcu(&mqp->list, &mcast->qp_list);
-
-	atomic_inc(&mcast->refcount);
-	rb_link_node(&mcast->rb_node, pn, n);
-	rb_insert_color(&mcast->rb_node, &ibp->mcast_tree);
-
-	ret = 0;
-
-bail:
-	spin_unlock_irq(&ibp->lock);
-
-	return ret;
-}
-
-int qib_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
-{
-	struct qib_qp *qp = to_iqp(ibqp);
-	struct qib_ibdev *dev = to_idev(ibqp->device);
-	struct qib_ibport *ibp;
-	struct qib_mcast *mcast;
-	struct qib_mcast_qp *mqp;
-	int ret;
-
-	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	/*
-	 * Allocate data structures since its better to do this outside of
-	 * spin locks and it will most likely be needed.
-	 */
-	mcast = qib_mcast_alloc(gid);
-	if (mcast == NULL) {
-		ret = -ENOMEM;
-		goto bail;
-	}
-	mqp = qib_mcast_qp_alloc(qp);
-	if (mqp == NULL) {
-		qib_mcast_free(mcast);
-		ret = -ENOMEM;
-		goto bail;
-	}
-	ibp = to_iport(ibqp->device, qp->port_num);
-	switch (qib_mcast_add(dev, ibp, mcast, mqp)) {
-	case ESRCH:
-		/* Neither was used: OK to attach the same QP twice. */
-		qib_mcast_qp_free(mqp);
-		qib_mcast_free(mcast);
-		break;
-
-	case EEXIST:            /* The mcast wasn't used */
-		qib_mcast_free(mcast);
-		break;
-
-	case ENOMEM:
-		/* Exceeded the maximum number of mcast groups. */
-		qib_mcast_qp_free(mqp);
-		qib_mcast_free(mcast);
-		ret = -ENOMEM;
-		goto bail;
-
-	default:
-		break;
-	}
-
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
-{
-	struct qib_qp *qp = to_iqp(ibqp);
-	struct qib_ibdev *dev = to_idev(ibqp->device);
-	struct qib_ibport *ibp = to_iport(ibqp->device, qp->port_num);
-	struct qib_mcast *mcast = NULL;
-	struct qib_mcast_qp *p, *tmp, *delp = NULL;
-	struct rb_node *n;
-	int last = 0;
-	int ret;
-
-	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET)
-		return -EINVAL;
-
-	spin_lock_irq(&ibp->lock);
-
-	/* Find the GID in the mcast table. */
-	n = ibp->mcast_tree.rb_node;
-	while (1) {
-		if (n == NULL) {
-			spin_unlock_irq(&ibp->lock);
-			return -EINVAL;
-		}
-
-		mcast = rb_entry(n, struct qib_mcast, rb_node);
-		ret = memcmp(gid->raw, mcast->mgid.raw,
-			     sizeof(union ib_gid));
-		if (ret < 0)
-			n = n->rb_left;
-		else if (ret > 0)
-			n = n->rb_right;
-		else
-			break;
-	}
-
-	/* Search the QP list. */
-	list_for_each_entry_safe(p, tmp, &mcast->qp_list, list) {
-		if (p->qp != qp)
-			continue;
-		/*
-		 * We found it, so remove it, but don't poison the forward
-		 * link until we are sure there are no list walkers.
-		 */
-		list_del_rcu(&p->list);
-		mcast->n_attached--;
-		delp = p;
-
-		/* If this was the last attached QP, remove the GID too. */
-		if (list_empty(&mcast->qp_list)) {
-			rb_erase(&mcast->rb_node, &ibp->mcast_tree);
-			last = 1;
-		}
-		break;
-	}
-
-	spin_unlock_irq(&ibp->lock);
-	/* QP not attached */
-	if (!delp)
-		return -EINVAL;
-	/*
-	 * Wait for any list walkers to finish before freeing the
-	 * list element.
-	 */
-	wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
-	qib_mcast_qp_free(delp);
-
-	if (last) {
-		atomic_dec(&mcast->refcount);
-		wait_event(mcast->wait, !atomic_read(&mcast->refcount));
-		qib_mcast_free(mcast);
-		spin_lock_irq(&dev->n_mcast_grps_lock);
-		dev->n_mcast_grps_allocated--;
-		spin_unlock_irq(&dev->n_mcast_grps_lock);
-	}
-	return 0;
-}
-
-int qib_mcast_tree_empty(struct qib_ibport *ibp)
-{
-	return ibp->mcast_tree.rb_node == NULL;
-}
diff --git a/drivers/infiniband/sw/Makefile b/drivers/infiniband/sw/Makefile
new file mode 100644
index 000000000000..988b6a0101a4
--- /dev/null
+++ b/drivers/infiniband/sw/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_INFINIBAND_RDMAVT)		+= rdmavt/
diff --git a/drivers/infiniband/sw/rdmavt/Kconfig b/drivers/infiniband/sw/rdmavt/Kconfig
new file mode 100644
index 000000000000..11aa6a34bd71
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/Kconfig
@@ -0,0 +1,6 @@
+config INFINIBAND_RDMAVT
+	tristate "RDMA verbs transport library"
+	depends on 64BIT
+	default m
+	---help---
+	This is a common software verbs provider for RDMA networks.
diff --git a/drivers/infiniband/sw/rdmavt/Makefile b/drivers/infiniband/sw/rdmavt/Makefile
new file mode 100644
index 000000000000..ccaa7992ac97
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/Makefile
@@ -0,0 +1,13 @@
+#
+# rdmavt driver
+#
+#
+#
+# Called from the kernel module build system.
+#
+obj-$(CONFIG_INFINIBAND_RDMAVT) += rdmavt.o
+
+rdmavt-y := vt.o ah.o cq.o dma.o mad.o mcast.o mmap.o mr.o pd.o qp.o srq.o \
+	trace.o
+
+CFLAGS_trace.o = -I$(src)
diff --git a/drivers/infiniband/sw/rdmavt/ah.c b/drivers/infiniband/sw/rdmavt/ah.c
new file mode 100644
index 000000000000..16c446142c2a
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/ah.c
@@ -0,0 +1,196 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/slab.h>
+#include "ah.h"
+#include "vt.h" /* for prints */
+
+/**
+ * rvt_check_ah - validate the attributes of AH
+ * @ibdev: the ib device
+ * @ah_attr: the attributes of the AH
+ *
+ * If driver supports a more detailed check_ah function call back to it
+ * otherwise just check the basics.
+ *
+ * Return: 0 on success
+ */
+int rvt_check_ah(struct ib_device *ibdev,
+		 struct ib_ah_attr *ah_attr)
+{
+	int err;
+	struct ib_port_attr port_attr;
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	enum rdma_link_layer link = rdma_port_get_link_layer(ibdev,
+							     ah_attr->port_num);
+
+	err = ib_query_port(ibdev, ah_attr->port_num, &port_attr);
+	if (err)
+		return -EINVAL;
+	if (ah_attr->port_num < 1 ||
+	    ah_attr->port_num > ibdev->phys_port_cnt)
+		return -EINVAL;
+	if (ah_attr->static_rate != IB_RATE_PORT_CURRENT &&
+	    ib_rate_to_mbps(ah_attr->static_rate) < 0)
+		return -EINVAL;
+	if ((ah_attr->ah_flags & IB_AH_GRH) &&
+	    ah_attr->grh.sgid_index >= port_attr.gid_tbl_len)
+		return -EINVAL;
+	if (link != IB_LINK_LAYER_ETHERNET) {
+		if (ah_attr->dlid == 0)
+			return -EINVAL;
+		if (ah_attr->dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE) &&
+		    ah_attr->dlid != be16_to_cpu(IB_LID_PERMISSIVE) &&
+		    !(ah_attr->ah_flags & IB_AH_GRH))
+			return -EINVAL;
+	}
+	if (rdi->driver_f.check_ah)
+		return rdi->driver_f.check_ah(ibdev, ah_attr);
+	return 0;
+}
+EXPORT_SYMBOL(rvt_check_ah);
+
+/**
+ * rvt_create_ah - create an address handle
+ * @pd: the protection domain
+ * @ah_attr: the attributes of the AH
+ *
+ * This may be called from interrupt context.
+ *
+ * Return: newly allocated ah
+ */
+struct ib_ah *rvt_create_ah(struct ib_pd *pd,
+			    struct ib_ah_attr *ah_attr)
+{
+	struct rvt_ah *ah;
+	struct rvt_dev_info *dev = ib_to_rvt(pd->device);
+	unsigned long flags;
+
+	if (rvt_check_ah(pd->device, ah_attr))
+		return ERR_PTR(-EINVAL);
+
+	ah = kmalloc(sizeof(*ah), GFP_ATOMIC);
+	if (!ah)
+		return ERR_PTR(-ENOMEM);
+
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
+	if (dev->n_ahs_allocated == dev->dparms.props.max_ah) {
+		spin_unlock(&dev->n_ahs_lock);
+		kfree(ah);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	dev->n_ahs_allocated++;
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
+
+	ah->attr = *ah_attr;
+	atomic_set(&ah->refcount, 0);
+
+	if (dev->driver_f.notify_new_ah)
+		dev->driver_f.notify_new_ah(pd->device, ah_attr, ah);
+
+	return &ah->ibah;
+}
+
+/**
+ * rvt_destory_ah - Destory an address handle
+ * @ibah: address handle
+ *
+ * Return: 0 on success
+ */
+int rvt_destroy_ah(struct ib_ah *ibah)
+{
+	struct rvt_dev_info *dev = ib_to_rvt(ibah->device);
+	struct rvt_ah *ah = ibah_to_rvtah(ibah);
+	unsigned long flags;
+
+	if (atomic_read(&ah->refcount) != 0)
+		return -EBUSY;
+
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
+	dev->n_ahs_allocated--;
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
+
+	kfree(ah);
+
+	return 0;
+}
+
+/**
+ * rvt_modify_ah - modify an ah with given attrs
+ * @ibah: address handle to modify
+ * @ah_attr: attrs to apply
+ *
+ * Return: 0 on success
+ */
+int rvt_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
+{
+	struct rvt_ah *ah = ibah_to_rvtah(ibah);
+
+	if (rvt_check_ah(ibah->device, ah_attr))
+		return -EINVAL;
+
+	ah->attr = *ah_attr;
+
+	return 0;
+}
+
+/**
+ * rvt_query_ah - return attrs for ah
+ * @ibah: address handle to query
+ * @ah_attr: return info in this
+ *
+ * Return: always 0
+ */
+int rvt_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
+{
+	struct rvt_ah *ah = ibah_to_rvtah(ibah);
+
+	*ah_attr = ah->attr;
+
+	return 0;
+}
diff --git a/drivers/infiniband/sw/rdmavt/ah.h b/drivers/infiniband/sw/rdmavt/ah.h
new file mode 100644
index 000000000000..e9c36be87d79
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/ah.h
@@ -0,0 +1,59 @@
+#ifndef DEF_RVTAH_H
+#define DEF_RVTAH_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+struct ib_ah *rvt_create_ah(struct ib_pd *pd,
+			    struct ib_ah_attr *ah_attr);
+int rvt_destroy_ah(struct ib_ah *ibah);
+int rvt_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr);
+int rvt_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr);
+
+#endif          /* DEF_RVTAH_H */
diff --git a/drivers/staging/rdma/hfi1/cq.c b/drivers/infiniband/sw/rdmavt/cq.c
index 4f046ffe7e60..b1ffc8b4a6c0 100644
--- a/drivers/staging/rdma/hfi1/cq.c
+++ b/drivers/infiniband/sw/rdmavt/cq.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -48,25 +45,23 @@
  *
  */
 
-#include <linux/err.h>
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
 #include <linux/kthread.h>
-
-#include "verbs.h"
-#include "hfi.h"
+#include "cq.h"
+#include "vt.h"
 
 /**
- * hfi1_cq_enter - add a new entry to the completion queue
+ * rvt_cq_enter - add a new entry to the completion queue
  * @cq: completion queue
  * @entry: work completion entry to add
- * @sig: true if @entry is a solicited entry
+ * @sig: true if @entry is solicited
  *
  * This may be called with qp->s_lock held.
  */
-void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
+void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
 {
-	struct hfi1_cq_wc *wc;
+	struct rvt_cq_wc *wc;
 	unsigned long flags;
 	u32 head;
 	u32 next;
@@ -79,11 +74,13 @@ void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
 	 */
 	wc = cq->queue;
 	head = wc->head;
-	if (head >= (unsigned) cq->ibcq.cqe) {
+	if (head >= (unsigned)cq->ibcq.cqe) {
 		head = cq->ibcq.cqe;
 		next = 0;
-	} else
+	} else {
 		next = head + 1;
+	}
+
 	if (unlikely(next == wc->tail)) {
 		spin_unlock_irqrestore(&cq->lock, flags);
 		if (cq->ibcq.event_handler) {
@@ -114,8 +111,9 @@ void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
 		wc->uqueue[head].port_num = entry->port_num;
 		/* Make sure entry is written before the head index. */
 		smp_wmb();
-	} else
+	} else {
 		wc->kqueue[head] = *entry;
+	}
 	wc->head = next;
 
 	if (cq->notify == IB_CQ_NEXT_COMP ||
@@ -126,10 +124,10 @@ void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
 		 * This will cause send_complete() to be called in
 		 * another thread.
 		 */
-		smp_read_barrier_depends(); /* see hfi1_cq_exit */
-		worker = cq->dd->worker;
+		smp_read_barrier_depends(); /* see rvt_cq_exit */
+		worker = cq->rdi->worker;
 		if (likely(worker)) {
-			cq->notify = IB_CQ_NONE;
+			cq->notify = RVT_CQ_NONE;
 			cq->triggered++;
 			queue_kthread_work(worker, &cq->comptask);
 		}
@@ -137,59 +135,11 @@ void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int solicited)
 
 	spin_unlock_irqrestore(&cq->lock, flags);
 }
-
-/**
- * hfi1_poll_cq - poll for work completion entries
- * @ibcq: the completion queue to poll
- * @num_entries: the maximum number of entries to return
- * @entry: pointer to array where work completions are placed
- *
- * Returns the number of completion entries polled.
- *
- * This may be called from interrupt context.  Also called by ib_poll_cq()
- * in the generic verbs code.
- */
-int hfi1_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
-{
-	struct hfi1_cq *cq = to_icq(ibcq);
-	struct hfi1_cq_wc *wc;
-	unsigned long flags;
-	int npolled;
-	u32 tail;
-
-	/* The kernel can only poll a kernel completion queue */
-	if (cq->ip) {
-		npolled = -EINVAL;
-		goto bail;
-	}
-
-	spin_lock_irqsave(&cq->lock, flags);
-
-	wc = cq->queue;
-	tail = wc->tail;
-	if (tail > (u32) cq->ibcq.cqe)
-		tail = (u32) cq->ibcq.cqe;
-	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
-		if (tail == wc->head)
-			break;
-		/* The kernel doesn't need a RMB since it has the lock. */
-		*entry = wc->kqueue[tail];
-		if (tail >= cq->ibcq.cqe)
-			tail = 0;
-		else
-			tail++;
-	}
-	wc->tail = tail;
-
-	spin_unlock_irqrestore(&cq->lock, flags);
-
-bail:
-	return npolled;
-}
+EXPORT_SYMBOL(rvt_cq_enter);
 
 static void send_complete(struct kthread_work *work)
 {
-	struct hfi1_cq *cq = container_of(work, struct hfi1_cq, comptask);
+	struct rvt_cq *cq = container_of(work, struct rvt_cq, comptask);
 
 	/*
 	 * The completion handler will most likely rearm the notification
@@ -217,26 +167,25 @@ static void send_complete(struct kthread_work *work)
 }
 
 /**
- * hfi1_create_cq - create a completion queue
+ * rvt_create_cq - create a completion queue
  * @ibdev: the device this completion queue is attached to
  * @attr: creation attributes
- * @context: unused by the driver
+ * @context: unused by the QLogic_IB driver
  * @udata: user data for libibverbs.so
  *
- * Returns a pointer to the completion queue or negative errno values
- * for failure.
- *
  * Called by ib_create_cq() in the generic verbs code.
+ *
+ * Return: pointer to the completion queue or negative errno values
+ * for failure.
  */
-struct ib_cq *hfi1_create_cq(
-	struct ib_device *ibdev,
-	const struct ib_cq_init_attr *attr,
-	struct ib_ucontext *context,
-	struct ib_udata *udata)
+struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
+			    const struct ib_cq_init_attr *attr,
+			    struct ib_ucontext *context,
+			    struct ib_udata *udata)
 {
-	struct hfi1_ibdev *dev = to_idev(ibdev);
-	struct hfi1_cq *cq;
-	struct hfi1_cq_wc *wc;
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	struct rvt_cq *cq;
+	struct rvt_cq_wc *wc;
 	struct ib_cq *ret;
 	u32 sz;
 	unsigned int entries = attr->cqe;
@@ -244,11 +193,11 @@ struct ib_cq *hfi1_create_cq(
 	if (attr->flags)
 		return ERR_PTR(-EINVAL);
 
-	if (entries < 1 || entries > hfi1_max_cqes)
+	if (entries < 1 || entries > rdi->dparms.props.max_cqe)
 		return ERR_PTR(-EINVAL);
 
 	/* Allocate the completion queue structure. */
-	cq = kmalloc(sizeof(*cq), GFP_KERNEL);
+	cq = kzalloc(sizeof(*cq), GFP_KERNEL);
 	if (!cq)
 		return ERR_PTR(-ENOMEM);
 
@@ -272,12 +221,12 @@ struct ib_cq *hfi1_create_cq(
 
 	/*
 	 * Return the address of the WC as the offset to mmap.
-	 * See hfi1_mmap() for details.
+	 * See rvt_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
 		int err;
 
-		cq->ip = hfi1_create_mmap_info(dev, sz, context, wc);
+		cq->ip = rvt_create_mmap_info(rdi, sz, context, wc);
 		if (!cq->ip) {
 			ret = ERR_PTR(-ENOMEM);
 			goto bail_wc;
@@ -289,23 +238,22 @@ struct ib_cq *hfi1_create_cq(
 			ret = ERR_PTR(err);
 			goto bail_ip;
 		}
-	} else
-		cq->ip = NULL;
+	}
 
-	spin_lock(&dev->n_cqs_lock);
-	if (dev->n_cqs_allocated == hfi1_max_cqs) {
-		spin_unlock(&dev->n_cqs_lock);
+	spin_lock(&rdi->n_cqs_lock);
+	if (rdi->n_cqs_allocated == rdi->dparms.props.max_cq) {
+		spin_unlock(&rdi->n_cqs_lock);
 		ret = ERR_PTR(-ENOMEM);
 		goto bail_ip;
 	}
 
-	dev->n_cqs_allocated++;
-	spin_unlock(&dev->n_cqs_lock);
+	rdi->n_cqs_allocated++;
+	spin_unlock(&rdi->n_cqs_lock);
 
 	if (cq->ip) {
-		spin_lock_irq(&dev->pending_lock);
-		list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
+		spin_lock_irq(&rdi->pending_lock);
+		list_add(&cq->ip->pending_mmaps, &rdi->pending_mmaps);
+		spin_unlock_irq(&rdi->pending_lock);
 	}
 
 	/*
@@ -313,14 +261,11 @@ struct ib_cq *hfi1_create_cq(
 	 * The number of entries should be >= the number requested or return
 	 * an error.
 	 */
-	cq->dd = dd_from_dev(dev);
+	cq->rdi = rdi;
 	cq->ibcq.cqe = entries;
-	cq->notify = IB_CQ_NONE;
-	cq->triggered = 0;
+	cq->notify = RVT_CQ_NONE;
 	spin_lock_init(&cq->lock);
 	init_kthread_work(&cq->comptask, send_complete);
-	wc->head = 0;
-	wc->tail = 0;
 	cq->queue = wc;
 
 	ret = &cq->ibcq;
@@ -338,24 +283,24 @@ done:
 }
 
 /**
- * hfi1_destroy_cq - destroy a completion queue
+ * rvt_destroy_cq - destroy a completion queue
  * @ibcq: the completion queue to destroy.
  *
- * Returns 0 for success.
- *
  * Called by ib_destroy_cq() in the generic verbs code.
+ *
+ * Return: always 0
  */
-int hfi1_destroy_cq(struct ib_cq *ibcq)
+int rvt_destroy_cq(struct ib_cq *ibcq)
 {
-	struct hfi1_ibdev *dev = to_idev(ibcq->device);
-	struct hfi1_cq *cq = to_icq(ibcq);
+	struct rvt_cq *cq = ibcq_to_rvtcq(ibcq);
+	struct rvt_dev_info *rdi = cq->rdi;
 
 	flush_kthread_work(&cq->comptask);
-	spin_lock(&dev->n_cqs_lock);
-	dev->n_cqs_allocated--;
-	spin_unlock(&dev->n_cqs_lock);
+	spin_lock(&rdi->n_cqs_lock);
+	rdi->n_cqs_allocated--;
+	spin_unlock(&rdi->n_cqs_lock);
 	if (cq->ip)
-		kref_put(&cq->ip->ref, hfi1_release_mmap_info);
+		kref_put(&cq->ip->ref, rvt_release_mmap_info);
 	else
 		vfree(cq->queue);
 	kfree(cq);
@@ -364,18 +309,18 @@ int hfi1_destroy_cq(struct ib_cq *ibcq)
 }
 
 /**
- * hfi1_req_notify_cq - change the notification type for a completion queue
+ * rvt_req_notify_cq - change the notification type for a completion queue
  * @ibcq: the completion queue
  * @notify_flags: the type of notification to request
  *
- * Returns 0 for success.
- *
  * This may be called from interrupt context.  Also called by
  * ib_req_notify_cq() in the generic verbs code.
+ *
+ * Return: 0 for success.
  */
-int hfi1_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
+int rvt_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
 {
-	struct hfi1_cq *cq = to_icq(ibcq);
+	struct rvt_cq *cq = ibcq_to_rvtcq(ibcq);
 	unsigned long flags;
 	int ret = 0;
 
@@ -397,24 +342,23 @@ int hfi1_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
 }
 
 /**
- * hfi1_resize_cq - change the size of the CQ
+ * rvt_resize_cq - change the size of the CQ
  * @ibcq: the completion queue
  *
- * Returns 0 for success.
+ * Return: 0 for success.
  */
-int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
+int rvt_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 {
-	struct hfi1_cq *cq = to_icq(ibcq);
-	struct hfi1_cq_wc *old_wc;
-	struct hfi1_cq_wc *wc;
+	struct rvt_cq *cq = ibcq_to_rvtcq(ibcq);
+	struct rvt_cq_wc *old_wc;
+	struct rvt_cq_wc *wc;
 	u32 head, tail, n;
 	int ret;
 	u32 sz;
+	struct rvt_dev_info *rdi = cq->rdi;
 
-	if (cqe < 1 || cqe > hfi1_max_cqes) {
-		ret = -EINVAL;
-		goto bail;
-	}
+	if (cqe < 1 || cqe > rdi->dparms.props.max_cqe)
+		return -EINVAL;
 
 	/*
 	 * Need to use vmalloc() if we want to support large #s of entries.
@@ -425,10 +369,8 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	else
 		sz += sizeof(struct ib_wc) * (cqe + 1);
 	wc = vmalloc_user(sz);
-	if (!wc) {
-		ret = -ENOMEM;
-		goto bail;
-	}
+	if (!wc)
+		return -ENOMEM;
 
 	/* Check that we can write the offset to mmap. */
 	if (udata && udata->outlen >= sizeof(__u64)) {
@@ -446,11 +388,11 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	 */
 	old_wc = cq->queue;
 	head = old_wc->head;
-	if (head > (u32) cq->ibcq.cqe)
-		head = (u32) cq->ibcq.cqe;
+	if (head > (u32)cq->ibcq.cqe)
+		head = (u32)cq->ibcq.cqe;
 	tail = old_wc->tail;
-	if (tail > (u32) cq->ibcq.cqe)
-		tail = (u32) cq->ibcq.cqe;
+	if (tail > (u32)cq->ibcq.cqe)
+		tail = (u32)cq->ibcq.cqe;
 	if (head < tail)
 		n = cq->ibcq.cqe + 1 + head - tail;
 	else
@@ -464,7 +406,7 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 			wc->uqueue[n] = old_wc->uqueue[tail];
 		else
 			wc->kqueue[n] = old_wc->kqueue[tail];
-		if (tail == (u32) cq->ibcq.cqe)
+		if (tail == (u32)cq->ibcq.cqe)
 			tail = 0;
 		else
 			tail++;
@@ -478,80 +420,131 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	vfree(old_wc);
 
 	if (cq->ip) {
-		struct hfi1_ibdev *dev = to_idev(ibcq->device);
-		struct hfi1_mmap_info *ip = cq->ip;
+		struct rvt_mmap_info *ip = cq->ip;
 
-		hfi1_update_mmap_info(dev, ip, sz, wc);
+		rvt_update_mmap_info(rdi, ip, sz, wc);
 
 		/*
 		 * Return the offset to mmap.
-		 * See hfi1_mmap() for details.
+		 * See rvt_mmap() for details.
 		 */
 		if (udata && udata->outlen >= sizeof(__u64)) {
 			ret = ib_copy_to_udata(udata, &ip->offset,
 					       sizeof(ip->offset));
 			if (ret)
-				goto bail;
+				return ret;
 		}
 
-		spin_lock_irq(&dev->pending_lock);
+		spin_lock_irq(&rdi->pending_lock);
 		if (list_empty(&ip->pending_mmaps))
-			list_add(&ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
+			list_add(&ip->pending_mmaps, &rdi->pending_mmaps);
+		spin_unlock_irq(&rdi->pending_lock);
 	}
 
-	ret = 0;
-	goto bail;
+	return 0;
 
 bail_unlock:
 	spin_unlock_irq(&cq->lock);
 bail_free:
 	vfree(wc);
-bail:
 	return ret;
 }
 
-int hfi1_cq_init(struct hfi1_devdata *dd)
+/**
+ * rvt_poll_cq - poll for work completion entries
+ * @ibcq: the completion queue to poll
+ * @num_entries: the maximum number of entries to return
+ * @entry: pointer to array where work completions are placed
+ *
+ * This may be called from interrupt context.  Also called by ib_poll_cq()
+ * in the generic verbs code.
+ *
+ * Return: the number of completion entries polled.
+ */
+int rvt_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
+{
+	struct rvt_cq *cq = ibcq_to_rvtcq(ibcq);
+	struct rvt_cq_wc *wc;
+	unsigned long flags;
+	int npolled;
+	u32 tail;
+
+	/* The kernel can only poll a kernel completion queue */
+	if (cq->ip)
+		return -EINVAL;
+
+	spin_lock_irqsave(&cq->lock, flags);
+
+	wc = cq->queue;
+	tail = wc->tail;
+	if (tail > (u32)cq->ibcq.cqe)
+		tail = (u32)cq->ibcq.cqe;
+	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
+		if (tail == wc->head)
+			break;
+		/* The kernel doesn't need a RMB since it has the lock. */
+		*entry = wc->kqueue[tail];
+		if (tail >= cq->ibcq.cqe)
+			tail = 0;
+		else
+			tail++;
+	}
+	wc->tail = tail;
+
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+	return npolled;
+}
+
+/**
+ * rvt_driver_cq_init - Init cq resources on behalf of driver
+ * @rdi: rvt dev structure
+ *
+ * Return: 0 on success
+ */
+int rvt_driver_cq_init(struct rvt_dev_info *rdi)
 {
 	int ret = 0;
 	int cpu;
 	struct task_struct *task;
 
-	if (dd->worker)
+	if (rdi->worker)
 		return 0;
-	dd->worker = kzalloc(sizeof(*dd->worker), GFP_KERNEL);
-	if (!dd->worker)
+	rdi->worker = kzalloc(sizeof(*rdi->worker), GFP_KERNEL);
+	if (!rdi->worker)
 		return -ENOMEM;
-	init_kthread_worker(dd->worker);
+	init_kthread_worker(rdi->worker);
 	task = kthread_create_on_node(
 		kthread_worker_fn,
-		dd->worker,
-		dd->assigned_node_id,
-		"hfi1_cq%d", dd->unit);
-	if (IS_ERR(task))
-		goto task_fail;
-	cpu = cpumask_first(cpumask_of_node(dd->assigned_node_id));
+		rdi->worker,
+		rdi->dparms.node,
+		"%s", rdi->dparms.cq_name);
+	if (IS_ERR(task)) {
+		kfree(rdi->worker);
+		rdi->worker = NULL;
+		return PTR_ERR(task);
+	}
+
+	cpu = cpumask_first(cpumask_of_node(rdi->dparms.node));
 	kthread_bind(task, cpu);
 	wake_up_process(task);
-out:
 	return ret;
-task_fail:
-	ret = PTR_ERR(task);
-	kfree(dd->worker);
-	dd->worker = NULL;
-	goto out;
 }
 
-void hfi1_cq_exit(struct hfi1_devdata *dd)
+/**
+ * rvt_cq_exit - tear down cq reources
+ * @rdi: rvt dev structure
+ */
+void rvt_cq_exit(struct rvt_dev_info *rdi)
 {
 	struct kthread_worker *worker;
 
-	worker = dd->worker;
+	worker = rdi->worker;
 	if (!worker)
 		return;
 	/* blocks future queuing from send_complete() */
-	dd->worker = NULL;
-	smp_wmb(); /* See hfi1_cq_enter */
+	rdi->worker = NULL;
+	smp_wmb(); /* See rdi_cq_enter */
 	flush_kthread_worker(worker);
 	kthread_stop(worker->task);
 	kfree(worker);
diff --git a/drivers/infiniband/sw/rdmavt/cq.h b/drivers/infiniband/sw/rdmavt/cq.h
new file mode 100644
index 000000000000..6182c29eff66
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/cq.h
@@ -0,0 +1,64 @@
+#ifndef DEF_RVTCQ_H
+#define DEF_RVTCQ_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_cq.h>
+
+struct ib_cq *rvt_create_cq(struct ib_device *ibdev,
+			    const struct ib_cq_init_attr *attr,
+			    struct ib_ucontext *context,
+			    struct ib_udata *udata);
+int rvt_destroy_cq(struct ib_cq *ibcq);
+int rvt_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags);
+int rvt_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
+int rvt_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
+int rvt_driver_cq_init(struct rvt_dev_info *rdi);
+void rvt_cq_exit(struct rvt_dev_info *rdi);
+#endif          /* DEF_RVTCQ_H */
diff --git a/drivers/infiniband/sw/rdmavt/dma.c b/drivers/infiniband/sw/rdmavt/dma.c
new file mode 100644
index 000000000000..33076a5eee2f
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/dma.c
@@ -0,0 +1,184 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <linux/types.h>
+#include <linux/scatterlist.h>
+#include <rdma/ib_verbs.h>
+
+#include "dma.h"
+
+#define BAD_DMA_ADDRESS ((u64)0)
+
+/*
+ * The following functions implement driver specific replacements
+ * for the ib_dma_*() functions.
+ *
+ * These functions return kernel virtual addresses instead of
+ * device bus addresses since the driver uses the CPU to copy
+ * data instead of using hardware DMA.
+ */
+
+static int rvt_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dma_addr == BAD_DMA_ADDRESS;
+}
+
+static u64 rvt_dma_map_single(struct ib_device *dev, void *cpu_addr,
+			      size_t size, enum dma_data_direction direction)
+{
+	if (WARN_ON(!valid_dma_direction(direction)))
+		return BAD_DMA_ADDRESS;
+
+	return (u64)cpu_addr;
+}
+
+static void rvt_dma_unmap_single(struct ib_device *dev, u64 addr, size_t size,
+				 enum dma_data_direction direction)
+{
+	/* This is a stub, nothing to be done here */
+}
+
+static u64 rvt_dma_map_page(struct ib_device *dev, struct page *page,
+			    unsigned long offset, size_t size,
+			    enum dma_data_direction direction)
+{
+	u64 addr;
+
+	if (WARN_ON(!valid_dma_direction(direction)))
+		return BAD_DMA_ADDRESS;
+
+	if (offset + size > PAGE_SIZE)
+		return BAD_DMA_ADDRESS;
+
+	addr = (u64)page_address(page);
+	if (addr)
+		addr += offset;
+
+	return addr;
+}
+
+static void rvt_dma_unmap_page(struct ib_device *dev, u64 addr, size_t size,
+			       enum dma_data_direction direction)
+{
+	/* This is a stub, nothing to be done here */
+}
+
+static int rvt_map_sg(struct ib_device *dev, struct scatterlist *sgl,
+		      int nents, enum dma_data_direction direction)
+{
+	struct scatterlist *sg;
+	u64 addr;
+	int i;
+	int ret = nents;
+
+	if (WARN_ON(!valid_dma_direction(direction)))
+		return 0;
+
+	for_each_sg(sgl, sg, nents, i) {
+		addr = (u64)page_address(sg_page(sg));
+		if (!addr) {
+			ret = 0;
+			break;
+		}
+		sg->dma_address = addr + sg->offset;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+		sg->dma_length = sg->length;
+#endif
+	}
+	return ret;
+}
+
+static void rvt_unmap_sg(struct ib_device *dev,
+			 struct scatterlist *sg, int nents,
+			 enum dma_data_direction direction)
+{
+	/* This is a stub, nothing to be done here */
+}
+
+static void rvt_sync_single_for_cpu(struct ib_device *dev, u64 addr,
+				    size_t size, enum dma_data_direction dir)
+{
+}
+
+static void rvt_sync_single_for_device(struct ib_device *dev, u64 addr,
+				       size_t size,
+				       enum dma_data_direction dir)
+{
+}
+
+static void *rvt_dma_alloc_coherent(struct ib_device *dev, size_t size,
+				    u64 *dma_handle, gfp_t flag)
+{
+	struct page *p;
+	void *addr = NULL;
+
+	p = alloc_pages(flag, get_order(size));
+	if (p)
+		addr = page_address(p);
+	if (dma_handle)
+		*dma_handle = (u64)addr;
+	return addr;
+}
+
+static void rvt_dma_free_coherent(struct ib_device *dev, size_t size,
+				  void *cpu_addr, u64 dma_handle)
+{
+	free_pages((unsigned long)cpu_addr, get_order(size));
+}
+
+struct ib_dma_mapping_ops rvt_default_dma_mapping_ops = {
+	.mapping_error = rvt_mapping_error,
+	.map_single = rvt_dma_map_single,
+	.unmap_single = rvt_dma_unmap_single,
+	.map_page = rvt_dma_map_page,
+	.unmap_page = rvt_dma_unmap_page,
+	.map_sg = rvt_map_sg,
+	.unmap_sg = rvt_unmap_sg,
+	.sync_single_for_cpu = rvt_sync_single_for_cpu,
+	.sync_single_for_device = rvt_sync_single_for_device,
+	.alloc_coherent = rvt_dma_alloc_coherent,
+	.free_coherent = rvt_dma_free_coherent
+};
diff --git a/drivers/infiniband/sw/rdmavt/dma.h b/drivers/infiniband/sw/rdmavt/dma.h
new file mode 100644
index 000000000000..979f07e09195
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/dma.h
@@ -0,0 +1,53 @@
+#ifndef DEF_RDMAVTDMA_H
+#define DEF_RDMAVTDMA_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+extern struct ib_dma_mapping_ops rvt_default_dma_mapping_ops;
+
+#endif          /* DEF_RDMAVTDMA_H */
diff --git a/drivers/infiniband/sw/rdmavt/mad.c b/drivers/infiniband/sw/rdmavt/mad.c
new file mode 100644
index 000000000000..f6e99778d7ca
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mad.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/ib_mad.h>
+#include "mad.h"
+#include "vt.h"
+
+/**
+ * rvt_process_mad - process an incoming MAD packet
+ * @ibdev: the infiniband device this packet came in on
+ * @mad_flags: MAD flags
+ * @port_num: the port number this packet came in on, 1 based from ib core
+ * @in_wc: the work completion entry for this packet
+ * @in_grh: the global route header for this packet
+ * @in_mad: the incoming MAD
+ * @out_mad: any outgoing MAD reply
+ *
+ * Note that the verbs framework has already done the MAD sanity checks,
+ * and hop count/pointer updating for IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE
+ * MADs.
+ *
+ * This is called by the ib_mad module.
+ *
+ * Return: IB_MAD_RESULT_SUCCESS or error
+ */
+int rvt_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+		    const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+		    const struct ib_mad_hdr *in, size_t in_mad_size,
+		    struct ib_mad_hdr *out, size_t *out_mad_size,
+		    u16 *out_mad_pkey_index)
+{
+	/*
+	 * MAD processing is quite different between hfi1 and qib. Therfore this
+	 * is expected to be provided by the driver. Other drivers in the future
+	 * may chose to implement this but it should not be made into a
+	 * requirement.
+	 */
+	if (ibport_num_to_idx(ibdev, port_num) < 0)
+		return -EINVAL;
+
+	return IB_MAD_RESULT_FAILURE;
+}
+
+static void rvt_send_mad_handler(struct ib_mad_agent *agent,
+				 struct ib_mad_send_wc *mad_send_wc)
+{
+	ib_free_send_mad(mad_send_wc->send_buf);
+}
+
+/**
+ * rvt_create_mad_agents - create mad agents
+ * @rdi: rvt dev struct
+ *
+ * If driver needs to be notified of mad agent creation then call back
+ *
+ * Return 0 on success
+ */
+int rvt_create_mad_agents(struct rvt_dev_info *rdi)
+{
+	struct ib_mad_agent *agent;
+	struct rvt_ibport *rvp;
+	int p;
+	int ret;
+
+	for (p = 0; p < rdi->dparms.nports; p++) {
+		rvp = rdi->ports[p];
+		agent = ib_register_mad_agent(&rdi->ibdev, p + 1,
+					      IB_QPT_SMI,
+					      NULL, 0, rvt_send_mad_handler,
+					      NULL, NULL, 0);
+		if (IS_ERR(agent)) {
+			ret = PTR_ERR(agent);
+			goto err;
+		}
+
+		rvp->send_agent = agent;
+
+		if (rdi->driver_f.notify_create_mad_agent)
+			rdi->driver_f.notify_create_mad_agent(rdi, p);
+	}
+
+	return 0;
+
+err:
+	for (p = 0; p < rdi->dparms.nports; p++) {
+		rvp = rdi->ports[p];
+		if (rvp->send_agent) {
+			agent = rvp->send_agent;
+			rvp->send_agent = NULL;
+			ib_unregister_mad_agent(agent);
+			if (rdi->driver_f.notify_free_mad_agent)
+				rdi->driver_f.notify_free_mad_agent(rdi, p);
+		}
+	}
+
+	return ret;
+}
+
+/**
+ * rvt_free_mad_agents - free up mad agents
+ * @rdi: rvt dev struct
+ *
+ * If driver needs notification of mad agent removal make the call back
+ */
+void rvt_free_mad_agents(struct rvt_dev_info *rdi)
+{
+	struct ib_mad_agent *agent;
+	struct rvt_ibport *rvp;
+	int p;
+
+	for (p = 0; p < rdi->dparms.nports; p++) {
+		rvp = rdi->ports[p];
+		if (rvp->send_agent) {
+			agent = rvp->send_agent;
+			rvp->send_agent = NULL;
+			ib_unregister_mad_agent(agent);
+		}
+		if (rvp->sm_ah) {
+			ib_destroy_ah(&rvp->sm_ah->ibah);
+			rvp->sm_ah = NULL;
+		}
+
+		if (rdi->driver_f.notify_free_mad_agent)
+			rdi->driver_f.notify_free_mad_agent(rdi, p);
+	}
+}
+
diff --git a/drivers/infiniband/sw/rdmavt/mad.h b/drivers/infiniband/sw/rdmavt/mad.h
new file mode 100644
index 000000000000..a9d6eecc3723
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mad.h
@@ -0,0 +1,60 @@
+#ifndef DEF_RVTMAD_H
+#define DEF_RVTMAD_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+int rvt_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+		    const struct ib_wc *in_wc, const struct ib_grh *in_grh,
+		    const struct ib_mad_hdr *in, size_t in_mad_size,
+		    struct ib_mad_hdr *out, size_t *out_mad_size,
+		    u16 *out_mad_pkey_index);
+int rvt_create_mad_agents(struct rvt_dev_info *rdi);
+void rvt_free_mad_agents(struct rvt_dev_info *rdi);
+#endif          /* DEF_RVTMAD_H */
diff --git a/drivers/staging/rdma/hfi1/verbs_mcast.c b/drivers/infiniband/sw/rdmavt/mcast.c
index afc6b4c61a1d..983d319ac976 100644
--- a/drivers/staging/rdma/hfi1/verbs_mcast.c
+++ b/drivers/infiniband/sw/rdmavt/mcast.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -48,17 +45,36 @@
  *
  */
 
+#include <linux/slab.h>
+#include <linux/sched.h>
 #include <linux/rculist.h>
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_qp.h>
 
-#include "hfi.h"
+#include "mcast.h"
+
+/**
+ * rvt_driver_mcast - init resources for multicast
+ * @rdi: rvt dev struct
+ *
+ * This is per device that registers with rdmavt
+ */
+void rvt_driver_mcast_init(struct rvt_dev_info *rdi)
+{
+	/*
+	 * Anything that needs setup for multicast on a per driver or per rdi
+	 * basis should be done in here.
+	 */
+	spin_lock_init(&rdi->n_mcast_grps_lock);
+}
 
 /**
  * mcast_qp_alloc - alloc a struct to link a QP to mcast GID struct
  * @qp: the QP to link
  */
-static struct hfi1_mcast_qp *mcast_qp_alloc(struct hfi1_qp *qp)
+static struct rvt_mcast_qp *rvt_mcast_qp_alloc(struct rvt_qp *qp)
 {
-	struct hfi1_mcast_qp *mqp;
+	struct rvt_mcast_qp *mqp;
 
 	mqp = kmalloc(sizeof(*mqp), GFP_KERNEL);
 	if (!mqp)
@@ -71,9 +87,9 @@ bail:
 	return mqp;
 }
 
-static void mcast_qp_free(struct hfi1_mcast_qp *mqp)
+static void rvt_mcast_qp_free(struct rvt_mcast_qp *mqp)
 {
-	struct hfi1_qp *qp = mqp->qp;
+	struct rvt_qp *qp = mqp->qp;
 
 	/* Notify hfi1_destroy_qp() if it is waiting. */
 	if (atomic_dec_and_test(&qp->refcount))
@@ -88,11 +104,11 @@ static void mcast_qp_free(struct hfi1_mcast_qp *mqp)
  *
  * A list of QPs will be attached to this structure.
  */
-static struct hfi1_mcast *mcast_alloc(union ib_gid *mgid)
+static struct rvt_mcast *rvt_mcast_alloc(union ib_gid *mgid)
 {
-	struct hfi1_mcast *mcast;
+	struct rvt_mcast *mcast;
 
-	mcast = kmalloc(sizeof(*mcast), GFP_KERNEL);
+	mcast = kzalloc(sizeof(*mcast), GFP_KERNEL);
 	if (!mcast)
 		goto bail;
 
@@ -100,75 +116,72 @@ static struct hfi1_mcast *mcast_alloc(union ib_gid *mgid)
 	INIT_LIST_HEAD(&mcast->qp_list);
 	init_waitqueue_head(&mcast->wait);
 	atomic_set(&mcast->refcount, 0);
-	mcast->n_attached = 0;
 
 bail:
 	return mcast;
 }
 
-static void mcast_free(struct hfi1_mcast *mcast)
+static void rvt_mcast_free(struct rvt_mcast *mcast)
 {
-	struct hfi1_mcast_qp *p, *tmp;
+	struct rvt_mcast_qp *p, *tmp;
 
 	list_for_each_entry_safe(p, tmp, &mcast->qp_list, list)
-		mcast_qp_free(p);
+		rvt_mcast_qp_free(p);
 
 	kfree(mcast);
 }
 
 /**
- * hfi1_mcast_find - search the global table for the given multicast GID
+ * rvt_mcast_find - search the global table for the given multicast GID
  * @ibp: the IB port structure
  * @mgid: the multicast GID to search for
  *
- * Returns NULL if not found.
- *
  * The caller is responsible for decrementing the reference count if found.
+ *
+ * Return: NULL if not found.
  */
-struct hfi1_mcast *hfi1_mcast_find(struct hfi1_ibport *ibp, union ib_gid *mgid)
+struct rvt_mcast *rvt_mcast_find(struct rvt_ibport *ibp, union ib_gid *mgid)
 {
 	struct rb_node *n;
 	unsigned long flags;
-	struct hfi1_mcast *mcast;
+	struct rvt_mcast *found = NULL;
 
 	spin_lock_irqsave(&ibp->lock, flags);
 	n = ibp->mcast_tree.rb_node;
 	while (n) {
 		int ret;
+		struct rvt_mcast *mcast;
 
-		mcast = rb_entry(n, struct hfi1_mcast, rb_node);
+		mcast = rb_entry(n, struct rvt_mcast, rb_node);
 
 		ret = memcmp(mgid->raw, mcast->mgid.raw,
 			     sizeof(union ib_gid));
-		if (ret < 0)
+		if (ret < 0) {
 			n = n->rb_left;
-		else if (ret > 0)
+		} else if (ret > 0) {
 			n = n->rb_right;
-		else {
+		} else {
 			atomic_inc(&mcast->refcount);
-			spin_unlock_irqrestore(&ibp->lock, flags);
-			goto bail;
+			found = mcast;
+			break;
 		}
 	}
 	spin_unlock_irqrestore(&ibp->lock, flags);
-
-	mcast = NULL;
-
-bail:
-	return mcast;
+	return found;
 }
+EXPORT_SYMBOL(rvt_mcast_find);
 
 /**
  * mcast_add - insert mcast GID into table and attach QP struct
  * @mcast: the mcast GID table
  * @mqp: the QP to attach
  *
- * Return zero if both were added.  Return EEXIST if the GID was already in
+ * Return: zero if both were added.  Return EEXIST if the GID was already in
  * the table but the QP was added.  Return ESRCH if the QP was already
  * attached and neither structure was added.
  */
-static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
-		     struct hfi1_mcast *mcast, struct hfi1_mcast_qp *mqp)
+static int rvt_mcast_add(struct rvt_dev_info *rdi, struct rvt_ibport *ibp,
+			 struct rvt_mcast *mcast, struct rvt_mcast_qp *mqp)
 {
 	struct rb_node **n = &ibp->mcast_tree.rb_node;
 	struct rb_node *pn = NULL;
@@ -177,11 +190,11 @@ static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
 	spin_lock_irq(&ibp->lock);
 
 	while (*n) {
-		struct hfi1_mcast *tmcast;
-		struct hfi1_mcast_qp *p;
+		struct rvt_mcast *tmcast;
+		struct rvt_mcast_qp *p;
 
 		pn = *n;
-		tmcast = rb_entry(pn, struct hfi1_mcast, rb_node);
+		tmcast = rb_entry(pn, struct rvt_mcast, rb_node);
 
 		ret = memcmp(mcast->mgid.raw, tmcast->mgid.raw,
 			     sizeof(union ib_gid));
@@ -201,7 +214,8 @@ static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
 				goto bail;
 			}
 		}
-		if (tmcast->n_attached == hfi1_max_mcast_qp_attached) {
+		if (tmcast->n_attached ==
+		    rdi->dparms.props.max_mcast_qp_attach) {
 			ret = ENOMEM;
 			goto bail;
 		}
@@ -213,15 +227,15 @@ static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
 		goto bail;
 	}
 
-	spin_lock(&dev->n_mcast_grps_lock);
-	if (dev->n_mcast_grps_allocated == hfi1_max_mcast_grps) {
-		spin_unlock(&dev->n_mcast_grps_lock);
+	spin_lock(&rdi->n_mcast_grps_lock);
+	if (rdi->n_mcast_grps_allocated == rdi->dparms.props.max_mcast_grp) {
+		spin_unlock(&rdi->n_mcast_grps_lock);
 		ret = ENOMEM;
 		goto bail;
 	}
 
-	dev->n_mcast_grps_allocated++;
-	spin_unlock(&dev->n_mcast_grps_lock);
+	rdi->n_mcast_grps_allocated++;
+	spin_unlock(&rdi->n_mcast_grps_lock);
 
 	mcast->n_attached++;
 
@@ -239,92 +253,98 @@ bail:
 	return ret;
 }
 
-int hfi1_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+/**
+ * rvt_attach_mcast - attach a qp to a multicast group
+ * @ibqp: Infiniband qp
+ * @igd: multicast guid
+ * @lid: multicast lid
+ *
+ * Return: 0 on success
+ */
+int rvt_attach_mcast(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	struct hfi1_ibdev *dev = to_idev(ibqp->device);
-	struct hfi1_ibport *ibp;
-	struct hfi1_mcast *mcast;
-	struct hfi1_mcast_qp *mqp;
-	int ret;
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+	struct rvt_ibport *ibp = rdi->ports[qp->port_num - 1];
+	struct rvt_mcast *mcast;
+	struct rvt_mcast_qp *mqp;
+	int ret = -ENOMEM;
 
-	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
-		ret = -EINVAL;
-		goto bail;
-	}
+	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET)
+		return -EINVAL;
 
 	/*
 	 * Allocate data structures since its better to do this outside of
 	 * spin locks and it will most likely be needed.
 	 */
-	mcast = mcast_alloc(gid);
-	if (mcast == NULL) {
-		ret = -ENOMEM;
-		goto bail;
-	}
-	mqp = mcast_qp_alloc(qp);
-	if (mqp == NULL) {
-		mcast_free(mcast);
-		ret = -ENOMEM;
-		goto bail;
-	}
-	ibp = to_iport(ibqp->device, qp->port_num);
-	switch (mcast_add(dev, ibp, mcast, mqp)) {
-	case ESRCH:
-		/* Neither was used: OK to attach the same QP twice. */
-		mcast_qp_free(mqp);
-		mcast_free(mcast);
-		break;
+	mcast = rvt_mcast_alloc(gid);
+	if (!mcast)
+		return -ENOMEM;
 
-	case EEXIST:            /* The mcast wasn't used */
-		mcast_free(mcast);
-		break;
+	mqp = rvt_mcast_qp_alloc(qp);
+	if (!mqp)
+		goto bail_mcast;
 
+	switch (rvt_mcast_add(rdi, ibp, mcast, mqp)) {
+	case ESRCH:
+		/* Neither was used: OK to attach the same QP twice. */
+		ret = 0;
+		goto bail_mqp;
+	case EEXIST: /* The mcast wasn't used */
+		ret = 0;
+		goto bail_mcast;
 	case ENOMEM:
 		/* Exceeded the maximum number of mcast groups. */
-		mcast_qp_free(mqp);
-		mcast_free(mcast);
 		ret = -ENOMEM;
-		goto bail;
-
+		goto bail_mqp;
 	default:
 		break;
 	}
 
-	ret = 0;
+	return 0;
+
+bail_mqp:
+	rvt_mcast_qp_free(mqp);
+
+bail_mcast:
+	rvt_mcast_free(mcast);
 
-bail:
 	return ret;
 }
 
-int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+/**
+ * rvt_detach_mcast - remove a qp from a multicast group
+ * @ibqp: Infiniband qp
+ * @igd: multicast guid
+ * @lid: multicast lid
+ *
+ * Return: 0 on success
+ */
+int rvt_detach_mcast(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	struct hfi1_ibdev *dev = to_idev(ibqp->device);
-	struct hfi1_ibport *ibp = to_iport(ibqp->device, qp->port_num);
-	struct hfi1_mcast *mcast = NULL;
-	struct hfi1_mcast_qp *p, *tmp;
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+	struct rvt_ibport *ibp = rdi->ports[qp->port_num - 1];
+	struct rvt_mcast *mcast = NULL;
+	struct rvt_mcast_qp *p, *tmp, *delp = NULL;
 	struct rb_node *n;
 	int last = 0;
-	int ret;
+	int ret = 0;
 
-	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
-		ret = -EINVAL;
-		goto bail;
-	}
+	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET)
+		return -EINVAL;
 
 	spin_lock_irq(&ibp->lock);
 
 	/* Find the GID in the mcast table. */
 	n = ibp->mcast_tree.rb_node;
 	while (1) {
-		if (n == NULL) {
+		if (!n) {
 			spin_unlock_irq(&ibp->lock);
-			ret = -EINVAL;
-			goto bail;
+			return -EINVAL;
 		}
 
-		mcast = rb_entry(n, struct hfi1_mcast, rb_node);
+		mcast = rb_entry(n, struct rvt_mcast, rb_node);
 		ret = memcmp(gid->raw, mcast->mgid.raw,
 			     sizeof(union ib_gid));
 		if (ret < 0)
@@ -345,6 +365,7 @@ int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 		 */
 		list_del_rcu(&p->list);
 		mcast->n_attached--;
+		delp = p;
 
 		/* If this was the last attached QP, remove the GID too. */
 		if (list_empty(&mcast->qp_list)) {
@@ -355,31 +376,42 @@ int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	}
 
 	spin_unlock_irq(&ibp->lock);
+	/* QP not attached */
+	if (!delp)
+		return -EINVAL;
+
+	/*
+	 * Wait for any list walkers to finish before freeing the
+	 * list element.
+	 */
+	wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
+	rvt_mcast_qp_free(delp);
 
-	if (p) {
-		/*
-		 * Wait for any list walkers to finish before freeing the
-		 * list element.
-		 */
-		wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
-		mcast_qp_free(p);
-	}
 	if (last) {
 		atomic_dec(&mcast->refcount);
 		wait_event(mcast->wait, !atomic_read(&mcast->refcount));
-		mcast_free(mcast);
-		spin_lock_irq(&dev->n_mcast_grps_lock);
-		dev->n_mcast_grps_allocated--;
-		spin_unlock_irq(&dev->n_mcast_grps_lock);
+		rvt_mcast_free(mcast);
+		spin_lock_irq(&rdi->n_mcast_grps_lock);
+		rdi->n_mcast_grps_allocated--;
+		spin_unlock_irq(&rdi->n_mcast_grps_lock);
 	}
 
-	ret = 0;
-
-bail:
-	return ret;
+	return 0;
 }
 
-int hfi1_mcast_tree_empty(struct hfi1_ibport *ibp)
+/**
+ *rvt_mast_tree_empty - determine if any qps are attached to any mcast group
+ *@rdi: rvt dev struct
+ *
+ * Return: in use count
+ */
+int rvt_mcast_tree_empty(struct rvt_dev_info *rdi)
 {
-	return ibp->mcast_tree.rb_node == NULL;
+	int i;
+	int in_use = 0;
+
+	for (i = 0; i < rdi->dparms.nports; i++)
+		if (rdi->ports[i]->mcast_tree.rb_node)
+			in_use++;
+	return in_use;
 }
diff --git a/drivers/infiniband/sw/rdmavt/mcast.h b/drivers/infiniband/sw/rdmavt/mcast.h
new file mode 100644
index 000000000000..29f579267608
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mcast.h
@@ -0,0 +1,58 @@
+#ifndef DEF_RVTMCAST_H
+#define DEF_RVTMCAST_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+void rvt_driver_mcast_init(struct rvt_dev_info *rdi);
+int rvt_attach_mcast(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
+int rvt_detach_mcast(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
+int rvt_mcast_tree_empty(struct rvt_dev_info *rdi);
+
+#endif          /* DEF_RVTMCAST_H */
diff --git a/drivers/staging/rdma/hfi1/mmap.c b/drivers/infiniband/sw/rdmavt/mmap.c
index 5173b1c60b3d..e202b8142759 100644
--- a/drivers/staging/rdma/hfi1/mmap.c
+++ b/drivers/infiniband/sw/rdmavt/mmap.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -48,68 +45,74 @@
  *
  */
 
-#include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
 #include <linux/mm.h>
-#include <linux/errno.h>
 #include <asm/pgtable.h>
+#include "mmap.h"
 
-#include "verbs.h"
+/**
+ * rvt_mmap_init - init link list and lock for mem map
+ * @rdi: rvt dev struct
+ */
+void rvt_mmap_init(struct rvt_dev_info *rdi)
+{
+	INIT_LIST_HEAD(&rdi->pending_mmaps);
+	spin_lock_init(&rdi->pending_lock);
+	rdi->mmap_offset = PAGE_SIZE;
+	spin_lock_init(&rdi->mmap_offset_lock);
+}
 
 /**
- * hfi1_release_mmap_info - free mmap info structure
- * @ref: a pointer to the kref within struct hfi1_mmap_info
+ * rvt_release_mmap_info - free mmap info structure
+ * @ref: a pointer to the kref within struct rvt_mmap_info
  */
-void hfi1_release_mmap_info(struct kref *ref)
+void rvt_release_mmap_info(struct kref *ref)
 {
-	struct hfi1_mmap_info *ip =
-		container_of(ref, struct hfi1_mmap_info, ref);
-	struct hfi1_ibdev *dev = to_idev(ip->context->device);
+	struct rvt_mmap_info *ip =
+		container_of(ref, struct rvt_mmap_info, ref);
+	struct rvt_dev_info *rdi = ib_to_rvt(ip->context->device);
 
-	spin_lock_irq(&dev->pending_lock);
+	spin_lock_irq(&rdi->pending_lock);
 	list_del(&ip->pending_mmaps);
-	spin_unlock_irq(&dev->pending_lock);
+	spin_unlock_irq(&rdi->pending_lock);
 
 	vfree(ip->obj);
 	kfree(ip);
 }
 
-/*
- * open and close keep track of how many times the CQ is mapped,
- * to avoid releasing it.
- */
-static void hfi1_vma_open(struct vm_area_struct *vma)
+static void rvt_vma_open(struct vm_area_struct *vma)
 {
-	struct hfi1_mmap_info *ip = vma->vm_private_data;
+	struct rvt_mmap_info *ip = vma->vm_private_data;
 
 	kref_get(&ip->ref);
 }
 
-static void hfi1_vma_close(struct vm_area_struct *vma)
+static void rvt_vma_close(struct vm_area_struct *vma)
 {
-	struct hfi1_mmap_info *ip = vma->vm_private_data;
+	struct rvt_mmap_info *ip = vma->vm_private_data;
 
-	kref_put(&ip->ref, hfi1_release_mmap_info);
+	kref_put(&ip->ref, rvt_release_mmap_info);
 }
 
-static struct vm_operations_struct hfi1_vm_ops = {
-	.open =     hfi1_vma_open,
-	.close =    hfi1_vma_close,
+static const struct vm_operations_struct rvt_vm_ops = {
+	.open = rvt_vma_open,
+	.close = rvt_vma_close,
 };
 
 /**
- * hfi1_mmap - create a new mmap region
+ * rvt_mmap - create a new mmap region
  * @context: the IB user context of the process making the mmap() call
  * @vma: the VMA to be initialized
- * Return zero if the mmap is OK. Otherwise, return an errno.
+ *
+ * Return: zero if the mmap is OK. Otherwise, return an errno.
  */
-int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+int rvt_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 {
-	struct hfi1_ibdev *dev = to_idev(context->device);
+	struct rvt_dev_info *rdi = ib_to_rvt(context->device);
 	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
 	unsigned long size = vma->vm_end - vma->vm_start;
-	struct hfi1_mmap_info *ip, *pp;
+	struct rvt_mmap_info *ip, *pp;
 	int ret = -EINVAL;
 
 	/*
@@ -117,53 +120,60 @@ int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 	 * Normally, this list is very short since a call to create a
 	 * CQ, QP, or SRQ is soon followed by a call to mmap().
 	 */
-	spin_lock_irq(&dev->pending_lock);
-	list_for_each_entry_safe(ip, pp, &dev->pending_mmaps,
+	spin_lock_irq(&rdi->pending_lock);
+	list_for_each_entry_safe(ip, pp, &rdi->pending_mmaps,
 				 pending_mmaps) {
 		/* Only the creator is allowed to mmap the object */
-		if (context != ip->context || (__u64) offset != ip->offset)
+		if (context != ip->context || (__u64)offset != ip->offset)
 			continue;
 		/* Don't allow a mmap larger than the object. */
 		if (size > ip->size)
 			break;
 
 		list_del_init(&ip->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
+		spin_unlock_irq(&rdi->pending_lock);
 
 		ret = remap_vmalloc_range(vma, ip->obj, 0);
 		if (ret)
 			goto done;
-		vma->vm_ops = &hfi1_vm_ops;
+		vma->vm_ops = &rvt_vm_ops;
 		vma->vm_private_data = ip;
-		hfi1_vma_open(vma);
+		rvt_vma_open(vma);
 		goto done;
 	}
-	spin_unlock_irq(&dev->pending_lock);
+	spin_unlock_irq(&rdi->pending_lock);
 done:
 	return ret;
 }
 
-/*
- * Allocate information for hfi1_mmap
+/**
+ * rvt_create_mmap_info - allocate information for hfi1_mmap
+ * @rdi: rvt dev struct
+ * @size: size in bytes to map
+ * @context: user context
+ * @obj: opaque pointer to a cq, wq etc
+ *
+ * Return: rvt_mmap struct on success
  */
-struct hfi1_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev,
-					     u32 size,
-					     struct ib_ucontext *context,
-					     void *obj) {
-	struct hfi1_mmap_info *ip;
+struct rvt_mmap_info *rvt_create_mmap_info(struct rvt_dev_info *rdi,
+					   u32 size,
+					   struct ib_ucontext *context,
+					   void *obj)
+{
+	struct rvt_mmap_info *ip;
 
-	ip = kmalloc(sizeof(*ip), GFP_KERNEL);
+	ip = kmalloc_node(sizeof(*ip), GFP_KERNEL, rdi->dparms.node);
 	if (!ip)
-		goto bail;
+		return ip;
 
 	size = PAGE_ALIGN(size);
 
-	spin_lock_irq(&dev->mmap_offset_lock);
-	if (dev->mmap_offset == 0)
-		dev->mmap_offset = PAGE_SIZE;
-	ip->offset = dev->mmap_offset;
-	dev->mmap_offset += size;
-	spin_unlock_irq(&dev->mmap_offset_lock);
+	spin_lock_irq(&rdi->mmap_offset_lock);
+	if (rdi->mmap_offset == 0)
+		rdi->mmap_offset = PAGE_SIZE;
+	ip->offset = rdi->mmap_offset;
+	rdi->mmap_offset += size;
+	spin_unlock_irq(&rdi->mmap_offset_lock);
 
 	INIT_LIST_HEAD(&ip->pending_mmaps);
 	ip->size = size;
@@ -171,21 +181,27 @@ struct hfi1_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev,
 	ip->obj = obj;
 	kref_init(&ip->ref);
 
-bail:
 	return ip;
 }
 
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct hfi1_mmap_info *ip,
-			   u32 size, void *obj)
+/**
+ * rvt_update_mmap_info - update a mem map
+ * @rdi: rvt dev struct
+ * @ip: mmap info pointer
+ * @size: size to grow by
+ * @obj: opaque pointer to cq, wq, etc.
+ */
+void rvt_update_mmap_info(struct rvt_dev_info *rdi, struct rvt_mmap_info *ip,
+			  u32 size, void *obj)
 {
 	size = PAGE_ALIGN(size);
 
-	spin_lock_irq(&dev->mmap_offset_lock);
-	if (dev->mmap_offset == 0)
-		dev->mmap_offset = PAGE_SIZE;
-	ip->offset = dev->mmap_offset;
-	dev->mmap_offset += size;
-	spin_unlock_irq(&dev->mmap_offset_lock);
+	spin_lock_irq(&rdi->mmap_offset_lock);
+	if (rdi->mmap_offset == 0)
+		rdi->mmap_offset = PAGE_SIZE;
+	ip->offset = rdi->mmap_offset;
+	rdi->mmap_offset += size;
+	spin_unlock_irq(&rdi->mmap_offset_lock);
 
 	ip->size = size;
 	ip->obj = obj;
diff --git a/drivers/infiniband/sw/rdmavt/mmap.h b/drivers/infiniband/sw/rdmavt/mmap.h
new file mode 100644
index 000000000000..fab0e7b1daf9
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mmap.h
@@ -0,0 +1,63 @@
+#ifndef DEF_RDMAVTMMAP_H
+#define DEF_RDMAVTMMAP_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+void rvt_mmap_init(struct rvt_dev_info *rdi);
+void rvt_release_mmap_info(struct kref *ref);
+int rvt_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
+struct rvt_mmap_info *rvt_create_mmap_info(struct rvt_dev_info *rdi,
+					   u32 size,
+					   struct ib_ucontext *context,
+					   void *obj);
+void rvt_update_mmap_info(struct rvt_dev_info *rdi, struct rvt_mmap_info *ip,
+			  u32 size, void *obj);
+
+#endif          /* DEF_RDMAVTMMAP_H */
diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c
new file mode 100644
index 000000000000..0ff765bfd619
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mr.c
@@ -0,0 +1,830 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <rdma/ib_umem.h>
+#include <rdma/rdma_vt.h>
+#include "vt.h"
+#include "mr.h"
+
+/**
+ * rvt_driver_mr_init - Init MR resources per driver
+ * @rdi: rvt dev struct
+ *
+ * Do any intilization needed when a driver registers with rdmavt.
+ *
+ * Return: 0 on success or errno on failure
+ */
+int rvt_driver_mr_init(struct rvt_dev_info *rdi)
+{
+	unsigned int lkey_table_size = rdi->dparms.lkey_table_size;
+	unsigned lk_tab_size;
+	int i;
+
+	/*
+	 * The top hfi1_lkey_table_size bits are used to index the
+	 * table.  The lower 8 bits can be owned by the user (copied from
+	 * the LKEY).  The remaining bits act as a generation number or tag.
+	 */
+	if (!lkey_table_size)
+		return -EINVAL;
+
+	spin_lock_init(&rdi->lkey_table.lock);
+
+	/* ensure generation is at least 4 bits */
+	if (lkey_table_size > RVT_MAX_LKEY_TABLE_BITS) {
+		rvt_pr_warn(rdi, "lkey bits %u too large, reduced to %u\n",
+			    lkey_table_size, RVT_MAX_LKEY_TABLE_BITS);
+		rdi->dparms.lkey_table_size = RVT_MAX_LKEY_TABLE_BITS;
+		lkey_table_size = rdi->dparms.lkey_table_size;
+	}
+	rdi->lkey_table.max = 1 << lkey_table_size;
+	lk_tab_size = rdi->lkey_table.max * sizeof(*rdi->lkey_table.table);
+	rdi->lkey_table.table = (struct rvt_mregion __rcu **)
+			       vmalloc_node(lk_tab_size, rdi->dparms.node);
+	if (!rdi->lkey_table.table)
+		return -ENOMEM;
+
+	RCU_INIT_POINTER(rdi->dma_mr, NULL);
+	for (i = 0; i < rdi->lkey_table.max; i++)
+		RCU_INIT_POINTER(rdi->lkey_table.table[i], NULL);
+
+	return 0;
+}
+
+/**
+ *rvt_mr_exit: clean up MR
+ *@rdi: rvt dev structure
+ *
+ * called when drivers have unregistered or perhaps failed to register with us
+ */
+void rvt_mr_exit(struct rvt_dev_info *rdi)
+{
+	if (rdi->dma_mr)
+		rvt_pr_err(rdi, "DMA MR not null!\n");
+
+	vfree(rdi->lkey_table.table);
+}
+
+static void rvt_deinit_mregion(struct rvt_mregion *mr)
+{
+	int i = mr->mapsz;
+
+	mr->mapsz = 0;
+	while (i)
+		kfree(mr->map[--i]);
+}
+
+static int rvt_init_mregion(struct rvt_mregion *mr, struct ib_pd *pd,
+			    int count)
+{
+	int m, i = 0;
+
+	mr->mapsz = 0;
+	m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
+	for (; i < m; i++) {
+		mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL);
+		if (!mr->map[i]) {
+			rvt_deinit_mregion(mr);
+			return -ENOMEM;
+		}
+		mr->mapsz++;
+	}
+	init_completion(&mr->comp);
+	/* count returning the ptr to user */
+	atomic_set(&mr->refcount, 1);
+	mr->pd = pd;
+	mr->max_segs = count;
+	return 0;
+}
+
+/**
+ * rvt_alloc_lkey - allocate an lkey
+ * @mr: memory region that this lkey protects
+ * @dma_region: 0->normal key, 1->restricted DMA key
+ *
+ * Returns 0 if successful, otherwise returns -errno.
+ *
+ * Increments mr reference count as required.
+ *
+ * Sets the lkey field mr for non-dma regions.
+ *
+ */
+static int rvt_alloc_lkey(struct rvt_mregion *mr, int dma_region)
+{
+	unsigned long flags;
+	u32 r;
+	u32 n;
+	int ret = 0;
+	struct rvt_dev_info *dev = ib_to_rvt(mr->pd->device);
+	struct rvt_lkey_table *rkt = &dev->lkey_table;
+
+	rvt_get_mr(mr);
+	spin_lock_irqsave(&rkt->lock, flags);
+
+	/* special case for dma_mr lkey == 0 */
+	if (dma_region) {
+		struct rvt_mregion *tmr;
+
+		tmr = rcu_access_pointer(dev->dma_mr);
+		if (!tmr) {
+			rcu_assign_pointer(dev->dma_mr, mr);
+			mr->lkey_published = 1;
+		} else {
+			rvt_put_mr(mr);
+		}
+		goto success;
+	}
+
+	/* Find the next available LKEY */
+	r = rkt->next;
+	n = r;
+	for (;;) {
+		if (!rcu_access_pointer(rkt->table[r]))
+			break;
+		r = (r + 1) & (rkt->max - 1);
+		if (r == n)
+			goto bail;
+	}
+	rkt->next = (r + 1) & (rkt->max - 1);
+	/*
+	 * Make sure lkey is never zero which is reserved to indicate an
+	 * unrestricted LKEY.
+	 */
+	rkt->gen++;
+	/*
+	 * bits are capped to ensure enough bits for generation number
+	 */
+	mr->lkey = (r << (32 - dev->dparms.lkey_table_size)) |
+		((((1 << (24 - dev->dparms.lkey_table_size)) - 1) & rkt->gen)
+		 << 8);
+	if (mr->lkey == 0) {
+		mr->lkey |= 1 << 8;
+		rkt->gen++;
+	}
+	rcu_assign_pointer(rkt->table[r], mr);
+	mr->lkey_published = 1;
+success:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+out:
+	return ret;
+bail:
+	rvt_put_mr(mr);
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	ret = -ENOMEM;
+	goto out;
+}
+
+/**
+ * rvt_free_lkey - free an lkey
+ * @mr: mr to free from tables
+ */
+static void rvt_free_lkey(struct rvt_mregion *mr)
+{
+	unsigned long flags;
+	u32 lkey = mr->lkey;
+	u32 r;
+	struct rvt_dev_info *dev = ib_to_rvt(mr->pd->device);
+	struct rvt_lkey_table *rkt = &dev->lkey_table;
+	int freed = 0;
+
+	spin_lock_irqsave(&rkt->lock, flags);
+	if (!mr->lkey_published)
+		goto out;
+	if (lkey == 0) {
+		RCU_INIT_POINTER(dev->dma_mr, NULL);
+	} else {
+		r = lkey >> (32 - dev->dparms.lkey_table_size);
+		RCU_INIT_POINTER(rkt->table[r], NULL);
+	}
+	mr->lkey_published = 0;
+	freed++;
+out:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	if (freed) {
+		synchronize_rcu();
+		rvt_put_mr(mr);
+	}
+}
+
+static struct rvt_mr *__rvt_alloc_mr(int count, struct ib_pd *pd)
+{
+	struct rvt_mr *mr;
+	int rval = -ENOMEM;
+	int m;
+
+	/* Allocate struct plus pointers to first level page tables. */
+	m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
+	mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL);
+	if (!mr)
+		goto bail;
+
+	rval = rvt_init_mregion(&mr->mr, pd, count);
+	if (rval)
+		goto bail;
+	/*
+	 * ib_reg_phys_mr() will initialize mr->ibmr except for
+	 * lkey and rkey.
+	 */
+	rval = rvt_alloc_lkey(&mr->mr, 0);
+	if (rval)
+		goto bail_mregion;
+	mr->ibmr.lkey = mr->mr.lkey;
+	mr->ibmr.rkey = mr->mr.lkey;
+done:
+	return mr;
+
+bail_mregion:
+	rvt_deinit_mregion(&mr->mr);
+bail:
+	kfree(mr);
+	mr = ERR_PTR(rval);
+	goto done;
+}
+
+static void __rvt_free_mr(struct rvt_mr *mr)
+{
+	rvt_deinit_mregion(&mr->mr);
+	rvt_free_lkey(&mr->mr);
+	vfree(mr);
+}
+
+/**
+ * rvt_get_dma_mr - get a DMA memory region
+ * @pd: protection domain for this memory region
+ * @acc: access flags
+ *
+ * Return: the memory region on success, otherwise returns an errno.
+ * Note that all DMA addresses should be created via the
+ * struct ib_dma_mapping_ops functions (see dma.c).
+ */
+struct ib_mr *rvt_get_dma_mr(struct ib_pd *pd, int acc)
+{
+	struct rvt_mr *mr;
+	struct ib_mr *ret;
+	int rval;
+
+	if (ibpd_to_rvtpd(pd)->user)
+		return ERR_PTR(-EPERM);
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	rval = rvt_init_mregion(&mr->mr, pd, 0);
+	if (rval) {
+		ret = ERR_PTR(rval);
+		goto bail;
+	}
+
+	rval = rvt_alloc_lkey(&mr->mr, 1);
+	if (rval) {
+		ret = ERR_PTR(rval);
+		goto bail_mregion;
+	}
+
+	mr->mr.access_flags = acc;
+	ret = &mr->ibmr;
+done:
+	return ret;
+
+bail_mregion:
+	rvt_deinit_mregion(&mr->mr);
+bail:
+	kfree(mr);
+	goto done;
+}
+
+/**
+ * rvt_reg_user_mr - register a userspace memory region
+ * @pd: protection domain for this memory region
+ * @start: starting userspace address
+ * @length: length of region to register
+ * @mr_access_flags: access flags for this memory region
+ * @udata: unused by the driver
+ *
+ * Return: the memory region on success, otherwise returns an errno.
+ */
+struct ib_mr *rvt_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+			      u64 virt_addr, int mr_access_flags,
+			      struct ib_udata *udata)
+{
+	struct rvt_mr *mr;
+	struct ib_umem *umem;
+	struct scatterlist *sg;
+	int n, m, entry;
+	struct ib_mr *ret;
+
+	if (length == 0)
+		return ERR_PTR(-EINVAL);
+
+	umem = ib_umem_get(pd->uobject->context, start, length,
+			   mr_access_flags, 0);
+	if (IS_ERR(umem))
+		return (void *)umem;
+
+	n = umem->nmap;
+
+	mr = __rvt_alloc_mr(n, pd);
+	if (IS_ERR(mr)) {
+		ret = (struct ib_mr *)mr;
+		goto bail_umem;
+	}
+
+	mr->mr.user_base = start;
+	mr->mr.iova = virt_addr;
+	mr->mr.length = length;
+	mr->mr.offset = ib_umem_offset(umem);
+	mr->mr.access_flags = mr_access_flags;
+	mr->umem = umem;
+
+	if (is_power_of_2(umem->page_size))
+		mr->mr.page_shift = ilog2(umem->page_size);
+	m = 0;
+	n = 0;
+	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+		void *vaddr;
+
+		vaddr = page_address(sg_page(sg));
+		if (!vaddr) {
+			ret = ERR_PTR(-EINVAL);
+			goto bail_inval;
+		}
+		mr->mr.map[m]->segs[n].vaddr = vaddr;
+		mr->mr.map[m]->segs[n].length = umem->page_size;
+		n++;
+		if (n == RVT_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+	return &mr->ibmr;
+
+bail_inval:
+	__rvt_free_mr(mr);
+
+bail_umem:
+	ib_umem_release(umem);
+
+	return ret;
+}
+
+/**
+ * rvt_dereg_mr - unregister and free a memory region
+ * @ibmr: the memory region to free
+ *
+ *
+ * Note that this is called to free MRs created by rvt_get_dma_mr()
+ * or rvt_reg_user_mr().
+ *
+ * Returns 0 on success.
+ */
+int rvt_dereg_mr(struct ib_mr *ibmr)
+{
+	struct rvt_mr *mr = to_imr(ibmr);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibmr->pd->device);
+	int ret = 0;
+	unsigned long timeout;
+
+	rvt_free_lkey(&mr->mr);
+
+	rvt_put_mr(&mr->mr); /* will set completion if last */
+	timeout = wait_for_completion_timeout(&mr->mr.comp, 5 * HZ);
+	if (!timeout) {
+		rvt_pr_err(rdi,
+			   "rvt_dereg_mr timeout mr %p pd %p refcount %u\n",
+			   mr, mr->mr.pd, atomic_read(&mr->mr.refcount));
+		rvt_get_mr(&mr->mr);
+		ret = -EBUSY;
+		goto out;
+	}
+	rvt_deinit_mregion(&mr->mr);
+	if (mr->umem)
+		ib_umem_release(mr->umem);
+	kfree(mr);
+out:
+	return ret;
+}
+
+/**
+ * rvt_alloc_mr - Allocate a memory region usable with the
+ * @pd: protection domain for this memory region
+ * @mr_type: mem region type
+ * @max_num_sg: Max number of segments allowed
+ *
+ * Return: the memory region on success, otherwise return an errno.
+ */
+struct ib_mr *rvt_alloc_mr(struct ib_pd *pd,
+			   enum ib_mr_type mr_type,
+			   u32 max_num_sg)
+{
+	struct rvt_mr *mr;
+
+	if (mr_type != IB_MR_TYPE_MEM_REG)
+		return ERR_PTR(-EINVAL);
+
+	mr = __rvt_alloc_mr(max_num_sg, pd);
+	if (IS_ERR(mr))
+		return (struct ib_mr *)mr;
+
+	return &mr->ibmr;
+}
+
+/**
+ * rvt_alloc_fmr - allocate a fast memory region
+ * @pd: the protection domain for this memory region
+ * @mr_access_flags: access flags for this memory region
+ * @fmr_attr: fast memory region attributes
+ *
+ * Return: the memory region on success, otherwise returns an errno.
+ */
+struct ib_fmr *rvt_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
+			     struct ib_fmr_attr *fmr_attr)
+{
+	struct rvt_fmr *fmr;
+	int m;
+	struct ib_fmr *ret;
+	int rval = -ENOMEM;
+
+	/* Allocate struct plus pointers to first level page tables. */
+	m = (fmr_attr->max_pages + RVT_SEGSZ - 1) / RVT_SEGSZ;
+	fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL);
+	if (!fmr)
+		goto bail;
+
+	rval = rvt_init_mregion(&fmr->mr, pd, fmr_attr->max_pages);
+	if (rval)
+		goto bail;
+
+	/*
+	 * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
+	 * rkey.
+	 */
+	rval = rvt_alloc_lkey(&fmr->mr, 0);
+	if (rval)
+		goto bail_mregion;
+	fmr->ibfmr.rkey = fmr->mr.lkey;
+	fmr->ibfmr.lkey = fmr->mr.lkey;
+	/*
+	 * Resources are allocated but no valid mapping (RKEY can't be
+	 * used).
+	 */
+	fmr->mr.access_flags = mr_access_flags;
+	fmr->mr.max_segs = fmr_attr->max_pages;
+	fmr->mr.page_shift = fmr_attr->page_shift;
+
+	ret = &fmr->ibfmr;
+done:
+	return ret;
+
+bail_mregion:
+	rvt_deinit_mregion(&fmr->mr);
+bail:
+	kfree(fmr);
+	ret = ERR_PTR(rval);
+	goto done;
+}
+
+/**
+ * rvt_map_phys_fmr - set up a fast memory region
+ * @ibmfr: the fast memory region to set up
+ * @page_list: the list of pages to associate with the fast memory region
+ * @list_len: the number of pages to associate with the fast memory region
+ * @iova: the virtual address of the start of the fast memory region
+ *
+ * This may be called from interrupt context.
+ *
+ * Return: 0 on success
+ */
+
+int rvt_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
+		     int list_len, u64 iova)
+{
+	struct rvt_fmr *fmr = to_ifmr(ibfmr);
+	struct rvt_lkey_table *rkt;
+	unsigned long flags;
+	int m, n, i;
+	u32 ps;
+	struct rvt_dev_info *rdi = ib_to_rvt(ibfmr->device);
+
+	i = atomic_read(&fmr->mr.refcount);
+	if (i > 2)
+		return -EBUSY;
+
+	if (list_len > fmr->mr.max_segs)
+		return -EINVAL;
+
+	rkt = &rdi->lkey_table;
+	spin_lock_irqsave(&rkt->lock, flags);
+	fmr->mr.user_base = iova;
+	fmr->mr.iova = iova;
+	ps = 1 << fmr->mr.page_shift;
+	fmr->mr.length = list_len * ps;
+	m = 0;
+	n = 0;
+	for (i = 0; i < list_len; i++) {
+		fmr->mr.map[m]->segs[n].vaddr = (void *)page_list[i];
+		fmr->mr.map[m]->segs[n].length = ps;
+		if (++n == RVT_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	return 0;
+}
+
+/**
+ * rvt_unmap_fmr - unmap fast memory regions
+ * @fmr_list: the list of fast memory regions to unmap
+ *
+ * Return: 0 on success.
+ */
+int rvt_unmap_fmr(struct list_head *fmr_list)
+{
+	struct rvt_fmr *fmr;
+	struct rvt_lkey_table *rkt;
+	unsigned long flags;
+	struct rvt_dev_info *rdi;
+
+	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
+		rdi = ib_to_rvt(fmr->ibfmr.device);
+		rkt = &rdi->lkey_table;
+		spin_lock_irqsave(&rkt->lock, flags);
+		fmr->mr.user_base = 0;
+		fmr->mr.iova = 0;
+		fmr->mr.length = 0;
+		spin_unlock_irqrestore(&rkt->lock, flags);
+	}
+	return 0;
+}
+
+/**
+ * rvt_dealloc_fmr - deallocate a fast memory region
+ * @ibfmr: the fast memory region to deallocate
+ *
+ * Return: 0 on success.
+ */
+int rvt_dealloc_fmr(struct ib_fmr *ibfmr)
+{
+	struct rvt_fmr *fmr = to_ifmr(ibfmr);
+	int ret = 0;
+	unsigned long timeout;
+
+	rvt_free_lkey(&fmr->mr);
+	rvt_put_mr(&fmr->mr); /* will set completion if last */
+	timeout = wait_for_completion_timeout(&fmr->mr.comp, 5 * HZ);
+	if (!timeout) {
+		rvt_get_mr(&fmr->mr);
+		ret = -EBUSY;
+		goto out;
+	}
+	rvt_deinit_mregion(&fmr->mr);
+	kfree(fmr);
+out:
+	return ret;
+}
+
+/**
+ * rvt_lkey_ok - check IB SGE for validity and initialize
+ * @rkt: table containing lkey to check SGE against
+ * @pd: protection domain
+ * @isge: outgoing internal SGE
+ * @sge: SGE to check
+ * @acc: access flags
+ *
+ * Check the IB SGE for validity and initialize our internal version
+ * of it.
+ *
+ * Return: 1 if valid and successful, otherwise returns 0.
+ *
+ * increments the reference count upon success
+ *
+ */
+int rvt_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
+		struct rvt_sge *isge, struct ib_sge *sge, int acc)
+{
+	struct rvt_mregion *mr;
+	unsigned n, m;
+	size_t off;
+	struct rvt_dev_info *dev = ib_to_rvt(pd->ibpd.device);
+
+	/*
+	 * We use LKEY == zero for kernel virtual addresses
+	 * (see rvt_get_dma_mr and dma.c).
+	 */
+	rcu_read_lock();
+	if (sge->lkey == 0) {
+		if (pd->user)
+			goto bail;
+		mr = rcu_dereference(dev->dma_mr);
+		if (!mr)
+			goto bail;
+		atomic_inc(&mr->refcount);
+		rcu_read_unlock();
+
+		isge->mr = mr;
+		isge->vaddr = (void *)sge->addr;
+		isge->length = sge->length;
+		isge->sge_length = sge->length;
+		isge->m = 0;
+		isge->n = 0;
+		goto ok;
+	}
+	mr = rcu_dereference(
+		rkt->table[(sge->lkey >> (32 - dev->dparms.lkey_table_size))]);
+	if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
+		goto bail;
+
+	off = sge->addr - mr->user_base;
+	if (unlikely(sge->addr < mr->user_base ||
+		     off + sge->length > mr->length ||
+		     (mr->access_flags & acc) != acc))
+		goto bail;
+	atomic_inc(&mr->refcount);
+	rcu_read_unlock();
+
+	off += mr->offset;
+	if (mr->page_shift) {
+		/*
+		 * page sizes are uniform power of 2 so no loop is necessary
+		 * entries_spanned_by_off is the number of times the loop below
+		 * would have executed.
+		*/
+		size_t entries_spanned_by_off;
+
+		entries_spanned_by_off = off >> mr->page_shift;
+		off -= (entries_spanned_by_off << mr->page_shift);
+		m = entries_spanned_by_off / RVT_SEGSZ;
+		n = entries_spanned_by_off % RVT_SEGSZ;
+	} else {
+		m = 0;
+		n = 0;
+		while (off >= mr->map[m]->segs[n].length) {
+			off -= mr->map[m]->segs[n].length;
+			n++;
+			if (n >= RVT_SEGSZ) {
+				m++;
+				n = 0;
+			}
+		}
+	}
+	isge->mr = mr;
+	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
+	isge->length = mr->map[m]->segs[n].length - off;
+	isge->sge_length = sge->length;
+	isge->m = m;
+	isge->n = n;
+ok:
+	return 1;
+bail:
+	rcu_read_unlock();
+	return 0;
+}
+EXPORT_SYMBOL(rvt_lkey_ok);
+
+/**
+ * rvt_rkey_ok - check the IB virtual address, length, and RKEY
+ * @qp: qp for validation
+ * @sge: SGE state
+ * @len: length of data
+ * @vaddr: virtual address to place data
+ * @rkey: rkey to check
+ * @acc: access flags
+ *
+ * Return: 1 if successful, otherwise 0.
+ *
+ * increments the reference count upon success
+ */
+int rvt_rkey_ok(struct rvt_qp *qp, struct rvt_sge *sge,
+		u32 len, u64 vaddr, u32 rkey, int acc)
+{
+	struct rvt_dev_info *dev = ib_to_rvt(qp->ibqp.device);
+	struct rvt_lkey_table *rkt = &dev->lkey_table;
+	struct rvt_mregion *mr;
+	unsigned n, m;
+	size_t off;
+
+	/*
+	 * We use RKEY == zero for kernel virtual addresses
+	 * (see rvt_get_dma_mr and dma.c).
+	 */
+	rcu_read_lock();
+	if (rkey == 0) {
+		struct rvt_pd *pd = ibpd_to_rvtpd(qp->ibqp.pd);
+		struct rvt_dev_info *rdi = ib_to_rvt(pd->ibpd.device);
+
+		if (pd->user)
+			goto bail;
+		mr = rcu_dereference(rdi->dma_mr);
+		if (!mr)
+			goto bail;
+		atomic_inc(&mr->refcount);
+		rcu_read_unlock();
+
+		sge->mr = mr;
+		sge->vaddr = (void *)vaddr;
+		sge->length = len;
+		sge->sge_length = len;
+		sge->m = 0;
+		sge->n = 0;
+		goto ok;
+	}
+
+	mr = rcu_dereference(
+		rkt->table[(rkey >> (32 - dev->dparms.lkey_table_size))]);
+	if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
+		goto bail;
+
+	off = vaddr - mr->iova;
+	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
+		     (mr->access_flags & acc) == 0))
+		goto bail;
+	atomic_inc(&mr->refcount);
+	rcu_read_unlock();
+
+	off += mr->offset;
+	if (mr->page_shift) {
+		/*
+		 * page sizes are uniform power of 2 so no loop is necessary
+		 * entries_spanned_by_off is the number of times the loop below
+		 * would have executed.
+		*/
+		size_t entries_spanned_by_off;
+
+		entries_spanned_by_off = off >> mr->page_shift;
+		off -= (entries_spanned_by_off << mr->page_shift);
+		m = entries_spanned_by_off / RVT_SEGSZ;
+		n = entries_spanned_by_off % RVT_SEGSZ;
+	} else {
+		m = 0;
+		n = 0;
+		while (off >= mr->map[m]->segs[n].length) {
+			off -= mr->map[m]->segs[n].length;
+			n++;
+			if (n >= RVT_SEGSZ) {
+				m++;
+				n = 0;
+			}
+		}
+	}
+	sge->mr = mr;
+	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
+	sge->length = mr->map[m]->segs[n].length - off;
+	sge->sge_length = len;
+	sge->m = m;
+	sge->n = n;
+ok:
+	return 1;
+bail:
+	rcu_read_unlock();
+	return 0;
+}
+EXPORT_SYMBOL(rvt_rkey_ok);
diff --git a/drivers/infiniband/sw/rdmavt/mr.h b/drivers/infiniband/sw/rdmavt/mr.h
new file mode 100644
index 000000000000..69380512c6d1
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/mr.h
@@ -0,0 +1,92 @@
+#ifndef DEF_RVTMR_H
+#define DEF_RVTMR_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+struct rvt_fmr {
+	struct ib_fmr ibfmr;
+	struct rvt_mregion mr;        /* must be last */
+};
+
+struct rvt_mr {
+	struct ib_mr ibmr;
+	struct ib_umem *umem;
+	struct rvt_mregion mr;  /* must be last */
+};
+
+static inline struct rvt_fmr *to_ifmr(struct ib_fmr *ibfmr)
+{
+	return container_of(ibfmr, struct rvt_fmr, ibfmr);
+}
+
+static inline struct rvt_mr *to_imr(struct ib_mr *ibmr)
+{
+	return container_of(ibmr, struct rvt_mr, ibmr);
+}
+
+int rvt_driver_mr_init(struct rvt_dev_info *rdi);
+void rvt_mr_exit(struct rvt_dev_info *rdi);
+
+/* Mem Regions */
+struct ib_mr *rvt_get_dma_mr(struct ib_pd *pd, int acc);
+struct ib_mr *rvt_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+			      u64 virt_addr, int mr_access_flags,
+			      struct ib_udata *udata);
+int rvt_dereg_mr(struct ib_mr *ibmr);
+struct ib_mr *rvt_alloc_mr(struct ib_pd *pd,
+			   enum ib_mr_type mr_type,
+			   u32 max_num_sg);
+struct ib_fmr *rvt_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
+			     struct ib_fmr_attr *fmr_attr);
+int rvt_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
+		     int list_len, u64 iova);
+int rvt_unmap_fmr(struct list_head *fmr_list);
+int rvt_dealloc_fmr(struct ib_fmr *ibfmr);
+
+#endif          /* DEF_RVTMR_H */
diff --git a/drivers/infiniband/sw/rdmavt/pd.c b/drivers/infiniband/sw/rdmavt/pd.c
new file mode 100644
index 000000000000..d1292f324c67
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/pd.c
@@ -0,0 +1,119 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/slab.h>
+#include "pd.h"
+
+/**
+ * rvt_alloc_pd - allocate a protection domain
+ * @ibdev: ib device
+ * @context: optional user context
+ * @udata: optional user data
+ *
+ * Allocate and keep track of a PD.
+ *
+ * Return: 0 on success
+ */
+struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev,
+			   struct ib_ucontext *context,
+			   struct ib_udata *udata)
+{
+	struct rvt_dev_info *dev = ib_to_rvt(ibdev);
+	struct rvt_pd *pd;
+	struct ib_pd *ret;
+
+	pd = kmalloc(sizeof(*pd), GFP_KERNEL);
+	if (!pd) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+	/*
+	 * While we could continue allocating protecetion domains, being
+	 * constrained only by system resources. The IBTA spec defines that
+	 * there is a max_pd limit that can be set and we need to check for
+	 * that.
+	 */
+
+	spin_lock(&dev->n_pds_lock);
+	if (dev->n_pds_allocated == dev->dparms.props.max_pd) {
+		spin_unlock(&dev->n_pds_lock);
+		kfree(pd);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	dev->n_pds_allocated++;
+	spin_unlock(&dev->n_pds_lock);
+
+	/* ib_alloc_pd() will initialize pd->ibpd. */
+	pd->user = udata ? 1 : 0;
+
+	ret = &pd->ibpd;
+
+bail:
+	return ret;
+}
+
+/**
+ * rvt_dealloc_pd - Free PD
+ * @ibpd: Free up PD
+ *
+ * Return: always 0
+ */
+int rvt_dealloc_pd(struct ib_pd *ibpd)
+{
+	struct rvt_pd *pd = ibpd_to_rvtpd(ibpd);
+	struct rvt_dev_info *dev = ib_to_rvt(ibpd->device);
+
+	spin_lock(&dev->n_pds_lock);
+	dev->n_pds_allocated--;
+	spin_unlock(&dev->n_pds_lock);
+
+	kfree(pd);
+
+	return 0;
+}
diff --git a/drivers/infiniband/sw/rdmavt/pd.h b/drivers/infiniband/sw/rdmavt/pd.h
new file mode 100644
index 000000000000..1892ca4a9746
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/pd.h
@@ -0,0 +1,58 @@
+#ifndef DEF_RDMAVTPD_H
+#define DEF_RDMAVTPD_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev,
+			   struct ib_ucontext *context,
+			   struct ib_udata *udata);
+int rvt_dealloc_pd(struct ib_pd *ibpd);
+
+#endif          /* DEF_RDMAVTPD_H */
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
new file mode 100644
index 000000000000..bd82a6948dc8
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -0,0 +1,1696 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/hash.h>
+#include <linux/bitops.h>
+#include <linux/lockdep.h>
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+#include <rdma/ib_verbs.h>
+#include "qp.h"
+#include "vt.h"
+#include "trace.h"
+
+/*
+ * Note that it is OK to post send work requests in the SQE and ERR
+ * states; rvt_do_send() will process them and generate error
+ * completions as per IB 1.2 C10-96.
+ */
+const int ib_rvt_state_ops[IB_QPS_ERR + 1] = {
+	[IB_QPS_RESET] = 0,
+	[IB_QPS_INIT] = RVT_POST_RECV_OK,
+	[IB_QPS_RTR] = RVT_POST_RECV_OK | RVT_PROCESS_RECV_OK,
+	[IB_QPS_RTS] = RVT_POST_RECV_OK | RVT_PROCESS_RECV_OK |
+	    RVT_POST_SEND_OK | RVT_PROCESS_SEND_OK |
+	    RVT_PROCESS_NEXT_SEND_OK,
+	[IB_QPS_SQD] = RVT_POST_RECV_OK | RVT_PROCESS_RECV_OK |
+	    RVT_POST_SEND_OK | RVT_PROCESS_SEND_OK,
+	[IB_QPS_SQE] = RVT_POST_RECV_OK | RVT_PROCESS_RECV_OK |
+	    RVT_POST_SEND_OK | RVT_FLUSH_SEND,
+	[IB_QPS_ERR] = RVT_POST_RECV_OK | RVT_FLUSH_RECV |
+	    RVT_POST_SEND_OK | RVT_FLUSH_SEND,
+};
+EXPORT_SYMBOL(ib_rvt_state_ops);
+
+static void get_map_page(struct rvt_qpn_table *qpt,
+			 struct rvt_qpn_map *map,
+			 gfp_t gfp)
+{
+	unsigned long page = get_zeroed_page(gfp);
+
+	/*
+	 * Free the page if someone raced with us installing it.
+	 */
+
+	spin_lock(&qpt->lock);
+	if (map->page)
+		free_page(page);
+	else
+		map->page = (void *)page;
+	spin_unlock(&qpt->lock);
+}
+
+/**
+ * init_qpn_table - initialize the QP number table for a device
+ * @qpt: the QPN table
+ */
+static int init_qpn_table(struct rvt_dev_info *rdi, struct rvt_qpn_table *qpt)
+{
+	u32 offset, i;
+	struct rvt_qpn_map *map;
+	int ret = 0;
+
+	if (!(rdi->dparms.qpn_res_end >= rdi->dparms.qpn_res_start))
+		return -EINVAL;
+
+	spin_lock_init(&qpt->lock);
+
+	qpt->last = rdi->dparms.qpn_start;
+	qpt->incr = rdi->dparms.qpn_inc << rdi->dparms.qos_shift;
+
+	/*
+	 * Drivers may want some QPs beyond what we need for verbs let them use
+	 * our qpn table. No need for two. Lets go ahead and mark the bitmaps
+	 * for those. The reserved range must be *after* the range which verbs
+	 * will pick from.
+	 */
+
+	/* Figure out number of bit maps needed before reserved range */
+	qpt->nmaps = rdi->dparms.qpn_res_start / RVT_BITS_PER_PAGE;
+
+	/* This should always be zero */
+	offset = rdi->dparms.qpn_res_start & RVT_BITS_PER_PAGE_MASK;
+
+	/* Starting with the first reserved bit map */
+	map = &qpt->map[qpt->nmaps];
+
+	rvt_pr_info(rdi, "Reserving QPNs from 0x%x to 0x%x for non-verbs use\n",
+		    rdi->dparms.qpn_res_start, rdi->dparms.qpn_res_end);
+	for (i = rdi->dparms.qpn_res_start; i <= rdi->dparms.qpn_res_end; i++) {
+		if (!map->page) {
+			get_map_page(qpt, map, GFP_KERNEL);
+			if (!map->page) {
+				ret = -ENOMEM;
+				break;
+			}
+		}
+		set_bit(offset, map->page);
+		offset++;
+		if (offset == RVT_BITS_PER_PAGE) {
+			/* next page */
+			qpt->nmaps++;
+			map++;
+			offset = 0;
+		}
+	}
+	return ret;
+}
+
+/**
+ * free_qpn_table - free the QP number table for a device
+ * @qpt: the QPN table
+ */
+static void free_qpn_table(struct rvt_qpn_table *qpt)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
+		free_page((unsigned long)qpt->map[i].page);
+}
+
+/**
+ * rvt_driver_qp_init - Init driver qp resources
+ * @rdi: rvt dev strucutre
+ *
+ * Return: 0 on success
+ */
+int rvt_driver_qp_init(struct rvt_dev_info *rdi)
+{
+	int i;
+	int ret = -ENOMEM;
+
+	if (!rdi->dparms.qp_table_size)
+		return -EINVAL;
+
+	/*
+	 * If driver is not doing any QP allocation then make sure it is
+	 * providing the necessary QP functions.
+	 */
+	if (!rdi->driver_f.free_all_qps ||
+	    !rdi->driver_f.qp_priv_alloc ||
+	    !rdi->driver_f.qp_priv_free ||
+	    !rdi->driver_f.notify_qp_reset)
+		return -EINVAL;
+
+	/* allocate parent object */
+	rdi->qp_dev = kzalloc_node(sizeof(*rdi->qp_dev), GFP_KERNEL,
+				   rdi->dparms.node);
+	if (!rdi->qp_dev)
+		return -ENOMEM;
+
+	/* allocate hash table */
+	rdi->qp_dev->qp_table_size = rdi->dparms.qp_table_size;
+	rdi->qp_dev->qp_table_bits = ilog2(rdi->dparms.qp_table_size);
+	rdi->qp_dev->qp_table =
+		kmalloc_node(rdi->qp_dev->qp_table_size *
+			     sizeof(*rdi->qp_dev->qp_table),
+			     GFP_KERNEL, rdi->dparms.node);
+	if (!rdi->qp_dev->qp_table)
+		goto no_qp_table;
+
+	for (i = 0; i < rdi->qp_dev->qp_table_size; i++)
+		RCU_INIT_POINTER(rdi->qp_dev->qp_table[i], NULL);
+
+	spin_lock_init(&rdi->qp_dev->qpt_lock);
+
+	/* initialize qpn map */
+	if (init_qpn_table(rdi, &rdi->qp_dev->qpn_table))
+		goto fail_table;
+
+	spin_lock_init(&rdi->n_qps_lock);
+
+	return 0;
+
+fail_table:
+	kfree(rdi->qp_dev->qp_table);
+	free_qpn_table(&rdi->qp_dev->qpn_table);
+
+no_qp_table:
+	kfree(rdi->qp_dev);
+
+	return ret;
+}
+
+/**
+ * free_all_qps - check for QPs still in use
+ * @qpt: the QP table to empty
+ *
+ * There should not be any QPs still in use.
+ * Free memory for table.
+ */
+static unsigned rvt_free_all_qps(struct rvt_dev_info *rdi)
+{
+	unsigned long flags;
+	struct rvt_qp *qp;
+	unsigned n, qp_inuse = 0;
+	spinlock_t *ql; /* work around too long line below */
+
+	if (rdi->driver_f.free_all_qps)
+		qp_inuse = rdi->driver_f.free_all_qps(rdi);
+
+	qp_inuse += rvt_mcast_tree_empty(rdi);
+
+	if (!rdi->qp_dev)
+		return qp_inuse;
+
+	ql = &rdi->qp_dev->qpt_lock;
+	spin_lock_irqsave(ql, flags);
+	for (n = 0; n < rdi->qp_dev->qp_table_size; n++) {
+		qp = rcu_dereference_protected(rdi->qp_dev->qp_table[n],
+					       lockdep_is_held(ql));
+		RCU_INIT_POINTER(rdi->qp_dev->qp_table[n], NULL);
+
+		for (; qp; qp = rcu_dereference_protected(qp->next,
+							  lockdep_is_held(ql)))
+			qp_inuse++;
+	}
+	spin_unlock_irqrestore(ql, flags);
+	synchronize_rcu();
+	return qp_inuse;
+}
+
+/**
+ * rvt_qp_exit - clean up qps on device exit
+ * @rdi: rvt dev structure
+ *
+ * Check for qp leaks and free resources.
+ */
+void rvt_qp_exit(struct rvt_dev_info *rdi)
+{
+	u32 qps_inuse = rvt_free_all_qps(rdi);
+
+	if (qps_inuse)
+		rvt_pr_err(rdi, "QP memory leak! %u still in use\n",
+			   qps_inuse);
+	if (!rdi->qp_dev)
+		return;
+
+	kfree(rdi->qp_dev->qp_table);
+	free_qpn_table(&rdi->qp_dev->qpn_table);
+	kfree(rdi->qp_dev);
+}
+
+static inline unsigned mk_qpn(struct rvt_qpn_table *qpt,
+			      struct rvt_qpn_map *map, unsigned off)
+{
+	return (map - qpt->map) * RVT_BITS_PER_PAGE + off;
+}
+
+/**
+ * alloc_qpn - Allocate the next available qpn or zero/one for QP type
+ *	       IB_QPT_SMI/IB_QPT_GSI
+ *@rdi:	rvt device info structure
+ *@qpt: queue pair number table pointer
+ *@port_num: IB port number, 1 based, comes from core
+ *
+ * Return: The queue pair number
+ */
+static int alloc_qpn(struct rvt_dev_info *rdi, struct rvt_qpn_table *qpt,
+		     enum ib_qp_type type, u8 port_num, gfp_t gfp)
+{
+	u32 i, offset, max_scan, qpn;
+	struct rvt_qpn_map *map;
+	u32 ret;
+
+	if (rdi->driver_f.alloc_qpn)
+		return rdi->driver_f.alloc_qpn(rdi, qpt, type, port_num, gfp);
+
+	if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
+		unsigned n;
+
+		ret = type == IB_QPT_GSI;
+		n = 1 << (ret + 2 * (port_num - 1));
+		spin_lock(&qpt->lock);
+		if (qpt->flags & n)
+			ret = -EINVAL;
+		else
+			qpt->flags |= n;
+		spin_unlock(&qpt->lock);
+		goto bail;
+	}
+
+	qpn = qpt->last + qpt->incr;
+	if (qpn >= RVT_QPN_MAX)
+		qpn = qpt->incr | ((qpt->last & 1) ^ 1);
+	/* offset carries bit 0 */
+	offset = qpn & RVT_BITS_PER_PAGE_MASK;
+	map = &qpt->map[qpn / RVT_BITS_PER_PAGE];
+	max_scan = qpt->nmaps - !offset;
+	for (i = 0;;) {
+		if (unlikely(!map->page)) {
+			get_map_page(qpt, map, gfp);
+			if (unlikely(!map->page))
+				break;
+		}
+		do {
+			if (!test_and_set_bit(offset, map->page)) {
+				qpt->last = qpn;
+				ret = qpn;
+				goto bail;
+			}
+			offset += qpt->incr;
+			/*
+			 * This qpn might be bogus if offset >= BITS_PER_PAGE.
+			 * That is OK.   It gets re-assigned below
+			 */
+			qpn = mk_qpn(qpt, map, offset);
+		} while (offset < RVT_BITS_PER_PAGE && qpn < RVT_QPN_MAX);
+		/*
+		 * In order to keep the number of pages allocated to a
+		 * minimum, we scan the all existing pages before increasing
+		 * the size of the bitmap table.
+		 */
+		if (++i > max_scan) {
+			if (qpt->nmaps == RVT_QPNMAP_ENTRIES)
+				break;
+			map = &qpt->map[qpt->nmaps++];
+			/* start at incr with current bit 0 */
+			offset = qpt->incr | (offset & 1);
+		} else if (map < &qpt->map[qpt->nmaps]) {
+			++map;
+			/* start at incr with current bit 0 */
+			offset = qpt->incr | (offset & 1);
+		} else {
+			map = &qpt->map[0];
+			/* wrap to first map page, invert bit 0 */
+			offset = qpt->incr | ((offset & 1) ^ 1);
+		}
+		/* there can be no bits at shift and below */
+		WARN_ON(offset & (rdi->dparms.qos_shift - 1));
+		qpn = mk_qpn(qpt, map, offset);
+	}
+
+	ret = -ENOMEM;
+
+bail:
+	return ret;
+}
+
+static void free_qpn(struct rvt_qpn_table *qpt, u32 qpn)
+{
+	struct rvt_qpn_map *map;
+
+	map = qpt->map + qpn / RVT_BITS_PER_PAGE;
+	if (map->page)
+		clear_bit(qpn & RVT_BITS_PER_PAGE_MASK, map->page);
+}
+
+/**
+ * rvt_clear_mr_refs - Drop help mr refs
+ * @qp: rvt qp data structure
+ * @clr_sends: If shoudl clear send side or not
+ */
+static void rvt_clear_mr_refs(struct rvt_qp *qp, int clr_sends)
+{
+	unsigned n;
+
+	if (test_and_clear_bit(RVT_R_REWIND_SGE, &qp->r_aflags))
+		rvt_put_ss(&qp->s_rdma_read_sge);
+
+	rvt_put_ss(&qp->r_sge);
+
+	if (clr_sends) {
+		while (qp->s_last != qp->s_head) {
+			struct rvt_swqe *wqe = rvt_get_swqe_ptr(qp, qp->s_last);
+			unsigned i;
+
+			for (i = 0; i < wqe->wr.num_sge; i++) {
+				struct rvt_sge *sge = &wqe->sg_list[i];
+
+				rvt_put_mr(sge->mr);
+			}
+			if (qp->ibqp.qp_type == IB_QPT_UD ||
+			    qp->ibqp.qp_type == IB_QPT_SMI ||
+			    qp->ibqp.qp_type == IB_QPT_GSI)
+				atomic_dec(&ibah_to_rvtah(
+						wqe->ud_wr.ah)->refcount);
+			if (++qp->s_last >= qp->s_size)
+				qp->s_last = 0;
+			smp_wmb(); /* see qp_set_savail */
+		}
+		if (qp->s_rdma_mr) {
+			rvt_put_mr(qp->s_rdma_mr);
+			qp->s_rdma_mr = NULL;
+		}
+	}
+
+	if (qp->ibqp.qp_type != IB_QPT_RC)
+		return;
+
+	for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
+		struct rvt_ack_entry *e = &qp->s_ack_queue[n];
+
+		if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
+		    e->rdma_sge.mr) {
+			rvt_put_mr(e->rdma_sge.mr);
+			e->rdma_sge.mr = NULL;
+		}
+	}
+}
+
+/**
+ * rvt_remove_qp - remove qp form table
+ * @rdi: rvt dev struct
+ * @qp: qp to remove
+ *
+ * Remove the QP from the table so it can't be found asynchronously by
+ * the receive routine.
+ */
+static void rvt_remove_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp)
+{
+	struct rvt_ibport *rvp = rdi->ports[qp->port_num - 1];
+	u32 n = hash_32(qp->ibqp.qp_num, rdi->qp_dev->qp_table_bits);
+	unsigned long flags;
+	int removed = 1;
+
+	spin_lock_irqsave(&rdi->qp_dev->qpt_lock, flags);
+
+	if (rcu_dereference_protected(rvp->qp[0],
+			lockdep_is_held(&rdi->qp_dev->qpt_lock)) == qp) {
+		RCU_INIT_POINTER(rvp->qp[0], NULL);
+	} else if (rcu_dereference_protected(rvp->qp[1],
+			lockdep_is_held(&rdi->qp_dev->qpt_lock)) == qp) {
+		RCU_INIT_POINTER(rvp->qp[1], NULL);
+	} else {
+		struct rvt_qp *q;
+		struct rvt_qp __rcu **qpp;
+
+		removed = 0;
+		qpp = &rdi->qp_dev->qp_table[n];
+		for (; (q = rcu_dereference_protected(*qpp,
+			lockdep_is_held(&rdi->qp_dev->qpt_lock))) != NULL;
+			qpp = &q->next) {
+			if (q == qp) {
+				RCU_INIT_POINTER(*qpp,
+				     rcu_dereference_protected(qp->next,
+				     lockdep_is_held(&rdi->qp_dev->qpt_lock)));
+				removed = 1;
+				trace_rvt_qpremove(qp, n);
+				break;
+			}
+		}
+	}
+
+	spin_unlock_irqrestore(&rdi->qp_dev->qpt_lock, flags);
+	if (removed) {
+		synchronize_rcu();
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+}
+
+/**
+ * reset_qp - initialize the QP state to the reset state
+ * @qp: the QP to reset
+ * @type: the QP type
+ * r and s lock are required to be held by the caller
+ */
+static void rvt_reset_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+		  enum ib_qp_type type)
+{
+	if (qp->state != IB_QPS_RESET) {
+		qp->state = IB_QPS_RESET;
+
+		/* Let drivers flush their waitlist */
+		rdi->driver_f.flush_qp_waiters(qp);
+		qp->s_flags &= ~(RVT_S_TIMER | RVT_S_ANY_WAIT);
+		spin_unlock(&qp->s_lock);
+		spin_unlock(&qp->s_hlock);
+		spin_unlock_irq(&qp->r_lock);
+
+		/* Stop the send queue and the retry timer */
+		rdi->driver_f.stop_send_queue(qp);
+
+		/* Wait for things to stop */
+		rdi->driver_f.quiesce_qp(qp);
+
+		/* take qp out the hash and wait for it to be unused */
+		rvt_remove_qp(rdi, qp);
+		wait_event(qp->wait, !atomic_read(&qp->refcount));
+
+		/* grab the lock b/c it was locked at call time */
+		spin_lock_irq(&qp->r_lock);
+		spin_lock(&qp->s_hlock);
+		spin_lock(&qp->s_lock);
+
+		rvt_clear_mr_refs(qp, 1);
+	}
+
+	/*
+	 * Let the driver do any tear down it needs to for a qp
+	 * that has been reset
+	 */
+	rdi->driver_f.notify_qp_reset(qp);
+
+	qp->remote_qpn = 0;
+	qp->qkey = 0;
+	qp->qp_access_flags = 0;
+	qp->s_flags &= RVT_S_SIGNAL_REQ_WR;
+	qp->s_hdrwords = 0;
+	qp->s_wqe = NULL;
+	qp->s_draining = 0;
+	qp->s_next_psn = 0;
+	qp->s_last_psn = 0;
+	qp->s_sending_psn = 0;
+	qp->s_sending_hpsn = 0;
+	qp->s_psn = 0;
+	qp->r_psn = 0;
+	qp->r_msn = 0;
+	if (type == IB_QPT_RC) {
+		qp->s_state = IB_OPCODE_RC_SEND_LAST;
+		qp->r_state = IB_OPCODE_RC_SEND_LAST;
+	} else {
+		qp->s_state = IB_OPCODE_UC_SEND_LAST;
+		qp->r_state = IB_OPCODE_UC_SEND_LAST;
+	}
+	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
+	qp->r_nak_state = 0;
+	qp->r_aflags = 0;
+	qp->r_flags = 0;
+	qp->s_head = 0;
+	qp->s_tail = 0;
+	qp->s_cur = 0;
+	qp->s_acked = 0;
+	qp->s_last = 0;
+	qp->s_ssn = 1;
+	qp->s_lsn = 0;
+	qp->s_mig_state = IB_MIG_MIGRATED;
+	memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
+	qp->r_head_ack_queue = 0;
+	qp->s_tail_ack_queue = 0;
+	qp->s_num_rd_atomic = 0;
+	if (qp->r_rq.wq) {
+		qp->r_rq.wq->head = 0;
+		qp->r_rq.wq->tail = 0;
+	}
+	qp->r_sge.num_sge = 0;
+}
+
+/**
+ * rvt_create_qp - create a queue pair for a device
+ * @ibpd: the protection domain who's device we create the queue pair for
+ * @init_attr: the attributes of the queue pair
+ * @udata: user data for libibverbs.so
+ *
+ * Queue pair creation is mostly an rvt issue. However, drivers have their own
+ * unique idea of what queue pair numbers mean. For instance there is a reserved
+ * range for PSM.
+ *
+ * Return: the queue pair on success, otherwise returns an errno.
+ *
+ * Called by the ib_create_qp() core verbs function.
+ */
+struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
+			    struct ib_qp_init_attr *init_attr,
+			    struct ib_udata *udata)
+{
+	struct rvt_qp *qp;
+	int err;
+	struct rvt_swqe *swq = NULL;
+	size_t sz;
+	size_t sg_list_sz;
+	struct ib_qp *ret = ERR_PTR(-ENOMEM);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibpd->device);
+	void *priv = NULL;
+	gfp_t gfp;
+
+	if (!rdi)
+		return ERR_PTR(-EINVAL);
+
+	if (init_attr->cap.max_send_sge > rdi->dparms.props.max_sge ||
+	    init_attr->cap.max_send_wr > rdi->dparms.props.max_qp_wr ||
+	    init_attr->create_flags & ~(IB_QP_CREATE_USE_GFP_NOIO))
+		return ERR_PTR(-EINVAL);
+
+	/* GFP_NOIO is applicable to RC QP's only */
+
+	if (init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO &&
+	    init_attr->qp_type != IB_QPT_RC)
+		return ERR_PTR(-EINVAL);
+
+	gfp = init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO ?
+						GFP_NOIO : GFP_KERNEL;
+
+	/* Check receive queue parameters if no SRQ is specified. */
+	if (!init_attr->srq) {
+		if (init_attr->cap.max_recv_sge > rdi->dparms.props.max_sge ||
+		    init_attr->cap.max_recv_wr > rdi->dparms.props.max_qp_wr)
+			return ERR_PTR(-EINVAL);
+
+		if (init_attr->cap.max_send_sge +
+		    init_attr->cap.max_send_wr +
+		    init_attr->cap.max_recv_sge +
+		    init_attr->cap.max_recv_wr == 0)
+			return ERR_PTR(-EINVAL);
+	}
+
+	switch (init_attr->qp_type) {
+	case IB_QPT_SMI:
+	case IB_QPT_GSI:
+		if (init_attr->port_num == 0 ||
+		    init_attr->port_num > ibpd->device->phys_port_cnt)
+			return ERR_PTR(-EINVAL);
+	case IB_QPT_UC:
+	case IB_QPT_RC:
+	case IB_QPT_UD:
+		sz = sizeof(struct rvt_sge) *
+			init_attr->cap.max_send_sge +
+			sizeof(struct rvt_swqe);
+		if (gfp == GFP_NOIO)
+			swq = __vmalloc(
+				(init_attr->cap.max_send_wr + 1) * sz,
+				gfp, PAGE_KERNEL);
+		else
+			swq = vmalloc_node(
+				(init_attr->cap.max_send_wr + 1) * sz,
+				rdi->dparms.node);
+		if (!swq)
+			return ERR_PTR(-ENOMEM);
+
+		sz = sizeof(*qp);
+		sg_list_sz = 0;
+		if (init_attr->srq) {
+			struct rvt_srq *srq = ibsrq_to_rvtsrq(init_attr->srq);
+
+			if (srq->rq.max_sge > 1)
+				sg_list_sz = sizeof(*qp->r_sg_list) *
+					(srq->rq.max_sge - 1);
+		} else if (init_attr->cap.max_recv_sge > 1)
+			sg_list_sz = sizeof(*qp->r_sg_list) *
+				(init_attr->cap.max_recv_sge - 1);
+		qp = kzalloc_node(sz + sg_list_sz, gfp, rdi->dparms.node);
+		if (!qp)
+			goto bail_swq;
+
+		RCU_INIT_POINTER(qp->next, NULL);
+
+		/*
+		 * Driver needs to set up it's private QP structure and do any
+		 * initialization that is needed.
+		 */
+		priv = rdi->driver_f.qp_priv_alloc(rdi, qp, gfp);
+		if (!priv)
+			goto bail_qp;
+		qp->priv = priv;
+		qp->timeout_jiffies =
+			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
+				1000UL);
+		if (init_attr->srq) {
+			sz = 0;
+		} else {
+			qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
+			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
+			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
+				sizeof(struct rvt_rwqe);
+			if (udata)
+				qp->r_rq.wq = vmalloc_user(
+						sizeof(struct rvt_rwq) +
+						qp->r_rq.size * sz);
+			else if (gfp == GFP_NOIO)
+				qp->r_rq.wq = __vmalloc(
+						sizeof(struct rvt_rwq) +
+						qp->r_rq.size * sz,
+						gfp, PAGE_KERNEL);
+			else
+				qp->r_rq.wq = vmalloc_node(
+						sizeof(struct rvt_rwq) +
+						qp->r_rq.size * sz,
+						rdi->dparms.node);
+			if (!qp->r_rq.wq)
+				goto bail_driver_priv;
+		}
+
+		/*
+		 * ib_create_qp() will initialize qp->ibqp
+		 * except for qp->ibqp.qp_num.
+		 */
+		spin_lock_init(&qp->r_lock);
+		spin_lock_init(&qp->s_hlock);
+		spin_lock_init(&qp->s_lock);
+		spin_lock_init(&qp->r_rq.lock);
+		atomic_set(&qp->refcount, 0);
+		init_waitqueue_head(&qp->wait);
+		init_timer(&qp->s_timer);
+		qp->s_timer.data = (unsigned long)qp;
+		INIT_LIST_HEAD(&qp->rspwait);
+		qp->state = IB_QPS_RESET;
+		qp->s_wq = swq;
+		qp->s_size = init_attr->cap.max_send_wr + 1;
+		qp->s_avail = init_attr->cap.max_send_wr;
+		qp->s_max_sge = init_attr->cap.max_send_sge;
+		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
+			qp->s_flags = RVT_S_SIGNAL_REQ_WR;
+
+		err = alloc_qpn(rdi, &rdi->qp_dev->qpn_table,
+				init_attr->qp_type,
+				init_attr->port_num, gfp);
+		if (err < 0) {
+			ret = ERR_PTR(err);
+			goto bail_rq_wq;
+		}
+		qp->ibqp.qp_num = err;
+		qp->port_num = init_attr->port_num;
+		rvt_reset_qp(rdi, qp, init_attr->qp_type);
+		break;
+
+	default:
+		/* Don't support raw QPs */
+		return ERR_PTR(-EINVAL);
+	}
+
+	init_attr->cap.max_inline_data = 0;
+
+	/*
+	 * Return the address of the RWQ as the offset to mmap.
+	 * See rvt_mmap() for details.
+	 */
+	if (udata && udata->outlen >= sizeof(__u64)) {
+		if (!qp->r_rq.wq) {
+			__u64 offset = 0;
+
+			err = ib_copy_to_udata(udata, &offset,
+					       sizeof(offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_qpn;
+			}
+		} else {
+			u32 s = sizeof(struct rvt_rwq) + qp->r_rq.size * sz;
+
+			qp->ip = rvt_create_mmap_info(rdi, s,
+						      ibpd->uobject->context,
+						      qp->r_rq.wq);
+			if (!qp->ip) {
+				ret = ERR_PTR(-ENOMEM);
+				goto bail_qpn;
+			}
+
+			err = ib_copy_to_udata(udata, &qp->ip->offset,
+					       sizeof(qp->ip->offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_ip;
+			}
+		}
+		qp->pid = current->pid;
+	}
+
+	spin_lock(&rdi->n_qps_lock);
+	if (rdi->n_qps_allocated == rdi->dparms.props.max_qp) {
+		spin_unlock(&rdi->n_qps_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_ip;
+	}
+
+	rdi->n_qps_allocated++;
+	/*
+	 * Maintain a busy_jiffies variable that will be added to the timeout
+	 * period in mod_retry_timer and add_retry_timer. This busy jiffies
+	 * is scaled by the number of rc qps created for the device to reduce
+	 * the number of timeouts occurring when there is a large number of
+	 * qps. busy_jiffies is incremented every rc qp scaling interval.
+	 * The scaling interval is selected based on extensive performance
+	 * evaluation of targeted workloads.
+	 */
+	if (init_attr->qp_type == IB_QPT_RC) {
+		rdi->n_rc_qps++;
+		rdi->busy_jiffies = rdi->n_rc_qps / RC_QP_SCALING_INTERVAL;
+	}
+	spin_unlock(&rdi->n_qps_lock);
+
+	if (qp->ip) {
+		spin_lock_irq(&rdi->pending_lock);
+		list_add(&qp->ip->pending_mmaps, &rdi->pending_mmaps);
+		spin_unlock_irq(&rdi->pending_lock);
+	}
+
+	ret = &qp->ibqp;
+
+	/*
+	 * We have our QP and its good, now keep track of what types of opcodes
+	 * can be processed on this QP. We do this by keeping track of what the
+	 * 3 high order bits of the opcode are.
+	 */
+	switch (init_attr->qp_type) {
+	case IB_QPT_SMI:
+	case IB_QPT_GSI:
+	case IB_QPT_UD:
+		qp->allowed_ops = IB_OPCODE_UD_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		break;
+	case IB_QPT_RC:
+		qp->allowed_ops = IB_OPCODE_RC_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		break;
+	case IB_QPT_UC:
+		qp->allowed_ops = IB_OPCODE_UC_SEND_ONLY & RVT_OPCODE_QP_MASK;
+		break;
+	default:
+		ret = ERR_PTR(-EINVAL);
+		goto bail_ip;
+	}
+
+	return ret;
+
+bail_ip:
+	kref_put(&qp->ip->ref, rvt_release_mmap_info);
+
+bail_qpn:
+	free_qpn(&rdi->qp_dev->qpn_table, qp->ibqp.qp_num);
+
+bail_rq_wq:
+	vfree(qp->r_rq.wq);
+
+bail_driver_priv:
+	rdi->driver_f.qp_priv_free(rdi, qp);
+
+bail_qp:
+	kfree(qp);
+
+bail_swq:
+	vfree(swq);
+
+	return ret;
+}
+
+/**
+ * rvt_error_qp - put a QP into the error state
+ * @qp: the QP to put into the error state
+ * @err: the receive completion error to signal if a RWQE is active
+ *
+ * Flushes both send and receive work queues.
+ *
+ * Return: true if last WQE event should be generated.
+ * The QP r_lock and s_lock should be held and interrupts disabled.
+ * If we are already in error state, just return.
+ */
+int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err)
+{
+	struct ib_wc wc;
+	int ret = 0;
+	struct rvt_dev_info *rdi = ib_to_rvt(qp->ibqp.device);
+
+	if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
+		goto bail;
+
+	qp->state = IB_QPS_ERR;
+
+	if (qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR)) {
+		qp->s_flags &= ~(RVT_S_TIMER | RVT_S_WAIT_RNR);
+		del_timer(&qp->s_timer);
+	}
+
+	if (qp->s_flags & RVT_S_ANY_WAIT_SEND)
+		qp->s_flags &= ~RVT_S_ANY_WAIT_SEND;
+
+	rdi->driver_f.notify_error_qp(qp);
+
+	/* Schedule the sending tasklet to drain the send work queue. */
+	if (ACCESS_ONCE(qp->s_last) != qp->s_head)
+		rdi->driver_f.schedule_send(qp);
+
+	rvt_clear_mr_refs(qp, 0);
+
+	memset(&wc, 0, sizeof(wc));
+	wc.qp = &qp->ibqp;
+	wc.opcode = IB_WC_RECV;
+
+	if (test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags)) {
+		wc.wr_id = qp->r_wr_id;
+		wc.status = err;
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
+	}
+	wc.status = IB_WC_WR_FLUSH_ERR;
+
+	if (qp->r_rq.wq) {
+		struct rvt_rwq *wq;
+		u32 head;
+		u32 tail;
+
+		spin_lock(&qp->r_rq.lock);
+
+		/* sanity check pointers before trusting them */
+		wq = qp->r_rq.wq;
+		head = wq->head;
+		if (head >= qp->r_rq.size)
+			head = 0;
+		tail = wq->tail;
+		if (tail >= qp->r_rq.size)
+			tail = 0;
+		while (tail != head) {
+			wc.wr_id = rvt_get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
+			if (++tail >= qp->r_rq.size)
+				tail = 0;
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
+		}
+		wq->tail = tail;
+
+		spin_unlock(&qp->r_rq.lock);
+	} else if (qp->ibqp.event_handler) {
+		ret = 1;
+	}
+
+bail:
+	return ret;
+}
+EXPORT_SYMBOL(rvt_error_qp);
+
+/*
+ * Put the QP into the hash table.
+ * The hash table holds a reference to the QP.
+ */
+static void rvt_insert_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp)
+{
+	struct rvt_ibport *rvp = rdi->ports[qp->port_num - 1];
+	unsigned long flags;
+
+	atomic_inc(&qp->refcount);
+	spin_lock_irqsave(&rdi->qp_dev->qpt_lock, flags);
+
+	if (qp->ibqp.qp_num <= 1) {
+		rcu_assign_pointer(rvp->qp[qp->ibqp.qp_num], qp);
+	} else {
+		u32 n = hash_32(qp->ibqp.qp_num, rdi->qp_dev->qp_table_bits);
+
+		qp->next = rdi->qp_dev->qp_table[n];
+		rcu_assign_pointer(rdi->qp_dev->qp_table[n], qp);
+		trace_rvt_qpinsert(qp, n);
+	}
+
+	spin_unlock_irqrestore(&rdi->qp_dev->qpt_lock, flags);
+}
+
+/**
+ * qib_modify_qp - modify the attributes of a queue pair
+ * @ibqp: the queue pair who's attributes we're modifying
+ * @attr: the new attributes
+ * @attr_mask: the mask of attributes to modify
+ * @udata: user data for libibverbs.so
+ *
+ * Return: 0 on success, otherwise returns an errno.
+ */
+int rvt_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		  int attr_mask, struct ib_udata *udata)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	enum ib_qp_state cur_state, new_state;
+	struct ib_event ev;
+	int lastwqe = 0;
+	int mig = 0;
+	int pmtu = 0; /* for gcc warning only */
+	enum rdma_link_layer link;
+
+	link = rdma_port_get_link_layer(ibqp->device, qp->port_num);
+
+	spin_lock_irq(&qp->r_lock);
+	spin_lock(&qp->s_hlock);
+	spin_lock(&qp->s_lock);
+
+	cur_state = attr_mask & IB_QP_CUR_STATE ?
+		attr->cur_qp_state : qp->state;
+	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
+
+	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
+				attr_mask, link))
+		goto inval;
+
+	if (rdi->driver_f.check_modify_qp &&
+	    rdi->driver_f.check_modify_qp(qp, attr, attr_mask, udata))
+		goto inval;
+
+	if (attr_mask & IB_QP_AV) {
+		if (attr->ah_attr.dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
+			goto inval;
+		if (rvt_check_ah(qp->ibqp.device, &attr->ah_attr))
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_ALT_PATH) {
+		if (attr->alt_ah_attr.dlid >=
+		    be16_to_cpu(IB_MULTICAST_LID_BASE))
+			goto inval;
+		if (rvt_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
+			goto inval;
+		if (attr->alt_pkey_index >= rvt_get_npkeys(rdi))
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		if (attr->pkey_index >= rvt_get_npkeys(rdi))
+			goto inval;
+
+	if (attr_mask & IB_QP_MIN_RNR_TIMER)
+		if (attr->min_rnr_timer > 31)
+			goto inval;
+
+	if (attr_mask & IB_QP_PORT)
+		if (qp->ibqp.qp_type == IB_QPT_SMI ||
+		    qp->ibqp.qp_type == IB_QPT_GSI ||
+		    attr->port_num == 0 ||
+		    attr->port_num > ibqp->device->phys_port_cnt)
+			goto inval;
+
+	if (attr_mask & IB_QP_DEST_QPN)
+		if (attr->dest_qp_num > RVT_QPN_MASK)
+			goto inval;
+
+	if (attr_mask & IB_QP_RETRY_CNT)
+		if (attr->retry_cnt > 7)
+			goto inval;
+
+	if (attr_mask & IB_QP_RNR_RETRY)
+		if (attr->rnr_retry > 7)
+			goto inval;
+
+	/*
+	 * Don't allow invalid path_mtu values.  OK to set greater
+	 * than the active mtu (or even the max_cap, if we have tuned
+	 * that to a small mtu.  We'll set qp->path_mtu
+	 * to the lesser of requested attribute mtu and active,
+	 * for packetizing messages.
+	 * Note that the QP port has to be set in INIT and MTU in RTR.
+	 */
+	if (attr_mask & IB_QP_PATH_MTU) {
+		pmtu = rdi->driver_f.get_pmtu_from_attr(rdi, qp, attr);
+		if (pmtu < 0)
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_PATH_MIG_STATE) {
+		if (attr->path_mig_state == IB_MIG_REARM) {
+			if (qp->s_mig_state == IB_MIG_ARMED)
+				goto inval;
+			if (new_state != IB_QPS_RTS)
+				goto inval;
+		} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
+			if (qp->s_mig_state == IB_MIG_REARM)
+				goto inval;
+			if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
+				goto inval;
+			if (qp->s_mig_state == IB_MIG_ARMED)
+				mig = 1;
+		} else {
+			goto inval;
+		}
+	}
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		if (attr->max_dest_rd_atomic > rdi->dparms.max_rdma_atomic)
+			goto inval;
+
+	switch (new_state) {
+	case IB_QPS_RESET:
+		if (qp->state != IB_QPS_RESET)
+			rvt_reset_qp(rdi, qp, ibqp->qp_type);
+		break;
+
+	case IB_QPS_RTR:
+		/* Allow event to re-trigger if QP set to RTR more than once */
+		qp->r_flags &= ~RVT_R_COMM_EST;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_SQD:
+		qp->s_draining = qp->s_last != qp->s_cur;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_SQE:
+		if (qp->ibqp.qp_type == IB_QPT_RC)
+			goto inval;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_ERR:
+		lastwqe = rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+		break;
+
+	default:
+		qp->state = new_state;
+		break;
+	}
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		qp->s_pkey_index = attr->pkey_index;
+
+	if (attr_mask & IB_QP_PORT)
+		qp->port_num = attr->port_num;
+
+	if (attr_mask & IB_QP_DEST_QPN)
+		qp->remote_qpn = attr->dest_qp_num;
+
+	if (attr_mask & IB_QP_SQ_PSN) {
+		qp->s_next_psn = attr->sq_psn & rdi->dparms.psn_modify_mask;
+		qp->s_psn = qp->s_next_psn;
+		qp->s_sending_psn = qp->s_next_psn;
+		qp->s_last_psn = qp->s_next_psn - 1;
+		qp->s_sending_hpsn = qp->s_last_psn;
+	}
+
+	if (attr_mask & IB_QP_RQ_PSN)
+		qp->r_psn = attr->rq_psn & rdi->dparms.psn_modify_mask;
+
+	if (attr_mask & IB_QP_ACCESS_FLAGS)
+		qp->qp_access_flags = attr->qp_access_flags;
+
+	if (attr_mask & IB_QP_AV) {
+		qp->remote_ah_attr = attr->ah_attr;
+		qp->s_srate = attr->ah_attr.static_rate;
+		qp->srate_mbps = ib_rate_to_mbps(qp->s_srate);
+	}
+
+	if (attr_mask & IB_QP_ALT_PATH) {
+		qp->alt_ah_attr = attr->alt_ah_attr;
+		qp->s_alt_pkey_index = attr->alt_pkey_index;
+	}
+
+	if (attr_mask & IB_QP_PATH_MIG_STATE) {
+		qp->s_mig_state = attr->path_mig_state;
+		if (mig) {
+			qp->remote_ah_attr = qp->alt_ah_attr;
+			qp->port_num = qp->alt_ah_attr.port_num;
+			qp->s_pkey_index = qp->s_alt_pkey_index;
+		}
+	}
+
+	if (attr_mask & IB_QP_PATH_MTU) {
+		qp->pmtu = rdi->driver_f.mtu_from_qp(rdi, qp, pmtu);
+		qp->path_mtu = rdi->driver_f.mtu_to_path_mtu(qp->pmtu);
+		qp->log_pmtu = ilog2(qp->pmtu);
+	}
+
+	if (attr_mask & IB_QP_RETRY_CNT) {
+		qp->s_retry_cnt = attr->retry_cnt;
+		qp->s_retry = attr->retry_cnt;
+	}
+
+	if (attr_mask & IB_QP_RNR_RETRY) {
+		qp->s_rnr_retry_cnt = attr->rnr_retry;
+		qp->s_rnr_retry = attr->rnr_retry;
+	}
+
+	if (attr_mask & IB_QP_MIN_RNR_TIMER)
+		qp->r_min_rnr_timer = attr->min_rnr_timer;
+
+	if (attr_mask & IB_QP_TIMEOUT) {
+		qp->timeout = attr->timeout;
+		qp->timeout_jiffies =
+			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
+				1000UL);
+	}
+
+	if (attr_mask & IB_QP_QKEY)
+		qp->qkey = attr->qkey;
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		qp->r_max_rd_atomic = attr->max_dest_rd_atomic;
+
+	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
+		qp->s_max_rd_atomic = attr->max_rd_atomic;
+
+	if (rdi->driver_f.modify_qp)
+		rdi->driver_f.modify_qp(qp, attr, attr_mask, udata);
+
+	spin_unlock(&qp->s_lock);
+	spin_unlock(&qp->s_hlock);
+	spin_unlock_irq(&qp->r_lock);
+
+	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+		rvt_insert_qp(rdi, qp);
+
+	if (lastwqe) {
+		ev.device = qp->ibqp.device;
+		ev.element.qp = &qp->ibqp;
+		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+	}
+	if (mig) {
+		ev.device = qp->ibqp.device;
+		ev.element.qp = &qp->ibqp;
+		ev.event = IB_EVENT_PATH_MIG;
+		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+	}
+	return 0;
+
+inval:
+	spin_unlock(&qp->s_lock);
+	spin_unlock(&qp->s_hlock);
+	spin_unlock_irq(&qp->r_lock);
+	return -EINVAL;
+}
+
+/** rvt_free_qpn - Free a qpn from the bit map
+ * @qpt: QP table
+ * @qpn: queue pair number to free
+ */
+static void rvt_free_qpn(struct rvt_qpn_table *qpt, u32 qpn)
+{
+	struct rvt_qpn_map *map;
+
+	map = qpt->map + qpn / RVT_BITS_PER_PAGE;
+	if (map->page)
+		clear_bit(qpn & RVT_BITS_PER_PAGE_MASK, map->page);
+}
+
+/**
+ * rvt_destroy_qp - destroy a queue pair
+ * @ibqp: the queue pair to destroy
+ *
+ * Note that this can be called while the QP is actively sending or
+ * receiving!
+ *
+ * Return: 0 on success.
+ */
+int rvt_destroy_qp(struct ib_qp *ibqp)
+{
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+
+	spin_lock_irq(&qp->r_lock);
+	spin_lock(&qp->s_hlock);
+	spin_lock(&qp->s_lock);
+	rvt_reset_qp(rdi, qp, ibqp->qp_type);
+	spin_unlock(&qp->s_lock);
+	spin_unlock(&qp->s_hlock);
+	spin_unlock_irq(&qp->r_lock);
+
+	/* qpn is now available for use again */
+	rvt_free_qpn(&rdi->qp_dev->qpn_table, qp->ibqp.qp_num);
+
+	spin_lock(&rdi->n_qps_lock);
+	rdi->n_qps_allocated--;
+	if (qp->ibqp.qp_type == IB_QPT_RC) {
+		rdi->n_rc_qps--;
+		rdi->busy_jiffies = rdi->n_rc_qps / RC_QP_SCALING_INTERVAL;
+	}
+	spin_unlock(&rdi->n_qps_lock);
+
+	if (qp->ip)
+		kref_put(&qp->ip->ref, rvt_release_mmap_info);
+	else
+		vfree(qp->r_rq.wq);
+	vfree(qp->s_wq);
+	rdi->driver_f.qp_priv_free(rdi, qp);
+	kfree(qp);
+	return 0;
+}
+
+/**
+ * rvt_query_qp - query an ipbq
+ * @ibqp: IB qp to query
+ * @attr: attr struct to fill in
+ * @attr_mask: attr mask ignored
+ * @init_attr: struct to fill in
+ *
+ * Return: always 0
+ */
+int rvt_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		 int attr_mask, struct ib_qp_init_attr *init_attr)
+{
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+
+	attr->qp_state = qp->state;
+	attr->cur_qp_state = attr->qp_state;
+	attr->path_mtu = qp->path_mtu;
+	attr->path_mig_state = qp->s_mig_state;
+	attr->qkey = qp->qkey;
+	attr->rq_psn = qp->r_psn & rdi->dparms.psn_mask;
+	attr->sq_psn = qp->s_next_psn & rdi->dparms.psn_mask;
+	attr->dest_qp_num = qp->remote_qpn;
+	attr->qp_access_flags = qp->qp_access_flags;
+	attr->cap.max_send_wr = qp->s_size - 1;
+	attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
+	attr->cap.max_send_sge = qp->s_max_sge;
+	attr->cap.max_recv_sge = qp->r_rq.max_sge;
+	attr->cap.max_inline_data = 0;
+	attr->ah_attr = qp->remote_ah_attr;
+	attr->alt_ah_attr = qp->alt_ah_attr;
+	attr->pkey_index = qp->s_pkey_index;
+	attr->alt_pkey_index = qp->s_alt_pkey_index;
+	attr->en_sqd_async_notify = 0;
+	attr->sq_draining = qp->s_draining;
+	attr->max_rd_atomic = qp->s_max_rd_atomic;
+	attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
+	attr->min_rnr_timer = qp->r_min_rnr_timer;
+	attr->port_num = qp->port_num;
+	attr->timeout = qp->timeout;
+	attr->retry_cnt = qp->s_retry_cnt;
+	attr->rnr_retry = qp->s_rnr_retry_cnt;
+	attr->alt_port_num = qp->alt_ah_attr.port_num;
+	attr->alt_timeout = qp->alt_timeout;
+
+	init_attr->event_handler = qp->ibqp.event_handler;
+	init_attr->qp_context = qp->ibqp.qp_context;
+	init_attr->send_cq = qp->ibqp.send_cq;
+	init_attr->recv_cq = qp->ibqp.recv_cq;
+	init_attr->srq = qp->ibqp.srq;
+	init_attr->cap = attr->cap;
+	if (qp->s_flags & RVT_S_SIGNAL_REQ_WR)
+		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
+	else
+		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
+	init_attr->qp_type = qp->ibqp.qp_type;
+	init_attr->port_num = qp->port_num;
+	return 0;
+}
+
+/**
+ * rvt_post_receive - post a receive on a QP
+ * @ibqp: the QP to post the receive on
+ * @wr: the WR to post
+ * @bad_wr: the first bad WR is put here
+ *
+ * This may be called from interrupt context.
+ *
+ * Return: 0 on success otherwise errno
+ */
+int rvt_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+		  struct ib_recv_wr **bad_wr)
+{
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_rwq *wq = qp->r_rq.wq;
+	unsigned long flags;
+	int qp_err_flush = (ib_rvt_state_ops[qp->state] & RVT_FLUSH_RECV) &&
+				!qp->ibqp.srq;
+
+	/* Check that state is OK to post receive. */
+	if (!(ib_rvt_state_ops[qp->state] & RVT_POST_RECV_OK) || !wq) {
+		*bad_wr = wr;
+		return -EINVAL;
+	}
+
+	for (; wr; wr = wr->next) {
+		struct rvt_rwqe *wqe;
+		u32 next;
+		int i;
+
+		if ((unsigned)wr->num_sge > qp->r_rq.max_sge) {
+			*bad_wr = wr;
+			return -EINVAL;
+		}
+
+		spin_lock_irqsave(&qp->r_rq.lock, flags);
+		next = wq->head + 1;
+		if (next >= qp->r_rq.size)
+			next = 0;
+		if (next == wq->tail) {
+			spin_unlock_irqrestore(&qp->r_rq.lock, flags);
+			*bad_wr = wr;
+			return -ENOMEM;
+		}
+		if (unlikely(qp_err_flush)) {
+			struct ib_wc wc;
+
+			memset(&wc, 0, sizeof(wc));
+			wc.qp = &qp->ibqp;
+			wc.opcode = IB_WC_RECV;
+			wc.wr_id = wr->wr_id;
+			wc.status = IB_WC_WR_FLUSH_ERR;
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
+		} else {
+			wqe = rvt_get_rwqe_ptr(&qp->r_rq, wq->head);
+			wqe->wr_id = wr->wr_id;
+			wqe->num_sge = wr->num_sge;
+			for (i = 0; i < wr->num_sge; i++)
+				wqe->sg_list[i] = wr->sg_list[i];
+			/*
+			 * Make sure queue entry is written
+			 * before the head index.
+			 */
+			smp_wmb();
+			wq->head = next;
+		}
+		spin_unlock_irqrestore(&qp->r_rq.lock, flags);
+	}
+	return 0;
+}
+
+/**
+ * qp_get_savail - return number of avail send entries
+ *
+ * @qp - the qp
+ *
+ * This assumes the s_hlock is held but the s_last
+ * qp variable is uncontrolled.
+ */
+static inline u32 qp_get_savail(struct rvt_qp *qp)
+{
+	u32 slast;
+	u32 ret;
+
+	smp_read_barrier_depends(); /* see rc.c */
+	slast = ACCESS_ONCE(qp->s_last);
+	if (qp->s_head >= slast)
+		ret = qp->s_size - (qp->s_head - slast);
+	else
+		ret = slast - qp->s_head;
+	return ret - 1;
+}
+
+/**
+ * rvt_post_one_wr - post one RC, UC, or UD send work request
+ * @qp: the QP to post on
+ * @wr: the work request to send
+ */
+static int rvt_post_one_wr(struct rvt_qp *qp,
+			   struct ib_send_wr *wr,
+			   int *call_send)
+{
+	struct rvt_swqe *wqe;
+	u32 next;
+	int i;
+	int j;
+	int acc;
+	struct rvt_lkey_table *rkt;
+	struct rvt_pd *pd;
+	struct rvt_dev_info *rdi = ib_to_rvt(qp->ibqp.device);
+	u8 log_pmtu;
+	int ret;
+
+	/* IB spec says that num_sge == 0 is OK. */
+	if (unlikely(wr->num_sge > qp->s_max_sge))
+		return -EINVAL;
+
+	/*
+	 * Don't allow RDMA reads or atomic operations on UC or
+	 * undefined operations.
+	 * Make sure buffer is large enough to hold the result for atomics.
+	 */
+	if (qp->ibqp.qp_type == IB_QPT_UC) {
+		if ((unsigned)wr->opcode >= IB_WR_RDMA_READ)
+			return -EINVAL;
+	} else if (qp->ibqp.qp_type != IB_QPT_RC) {
+		/* Check IB_QPT_SMI, IB_QPT_GSI, IB_QPT_UD opcode */
+		if (wr->opcode != IB_WR_SEND &&
+		    wr->opcode != IB_WR_SEND_WITH_IMM)
+			return -EINVAL;
+		/* Check UD destination address PD */
+		if (qp->ibqp.pd != ud_wr(wr)->ah->pd)
+			return -EINVAL;
+	} else if ((unsigned)wr->opcode > IB_WR_ATOMIC_FETCH_AND_ADD) {
+		return -EINVAL;
+	} else if (wr->opcode >= IB_WR_ATOMIC_CMP_AND_SWP &&
+		   (wr->num_sge == 0 ||
+		    wr->sg_list[0].length < sizeof(u64) ||
+		    wr->sg_list[0].addr & (sizeof(u64) - 1))) {
+		return -EINVAL;
+	} else if (wr->opcode >= IB_WR_RDMA_READ && !qp->s_max_rd_atomic) {
+		return -EINVAL;
+	}
+	/* check for avail */
+	if (unlikely(!qp->s_avail)) {
+		qp->s_avail = qp_get_savail(qp);
+		if (WARN_ON(qp->s_avail > (qp->s_size - 1)))
+			rvt_pr_err(rdi,
+				   "More avail entries than QP RB size.\nQP: %u, size: %u, avail: %u\nhead: %u, tail: %u, cur: %u, acked: %u, last: %u",
+				   qp->ibqp.qp_num, qp->s_size, qp->s_avail,
+				   qp->s_head, qp->s_tail, qp->s_cur,
+				   qp->s_acked, qp->s_last);
+		if (!qp->s_avail)
+			return -ENOMEM;
+	}
+	next = qp->s_head + 1;
+	if (next >= qp->s_size)
+		next = 0;
+
+	rkt = &rdi->lkey_table;
+	pd = ibpd_to_rvtpd(qp->ibqp.pd);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_head);
+
+	if (qp->ibqp.qp_type != IB_QPT_UC &&
+	    qp->ibqp.qp_type != IB_QPT_RC)
+		memcpy(&wqe->ud_wr, ud_wr(wr), sizeof(wqe->ud_wr));
+	else if (wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM ||
+		 wr->opcode == IB_WR_RDMA_WRITE ||
+		 wr->opcode == IB_WR_RDMA_READ)
+		memcpy(&wqe->rdma_wr, rdma_wr(wr), sizeof(wqe->rdma_wr));
+	else if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
+		 wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
+		memcpy(&wqe->atomic_wr, atomic_wr(wr), sizeof(wqe->atomic_wr));
+	else
+		memcpy(&wqe->wr, wr, sizeof(wqe->wr));
+
+	wqe->length = 0;
+	j = 0;
+	if (wr->num_sge) {
+		acc = wr->opcode >= IB_WR_RDMA_READ ?
+			IB_ACCESS_LOCAL_WRITE : 0;
+		for (i = 0; i < wr->num_sge; i++) {
+			u32 length = wr->sg_list[i].length;
+			int ok;
+
+			if (length == 0)
+				continue;
+			ok = rvt_lkey_ok(rkt, pd, &wqe->sg_list[j],
+					 &wr->sg_list[i], acc);
+			if (!ok) {
+				ret = -EINVAL;
+				goto bail_inval_free;
+			}
+			wqe->length += length;
+			j++;
+		}
+		wqe->wr.num_sge = j;
+	}
+
+	/* general part of wqe valid - allow for driver checks */
+	if (rdi->driver_f.check_send_wqe) {
+		ret = rdi->driver_f.check_send_wqe(qp, wqe);
+		if (ret < 0)
+			goto bail_inval_free;
+		if (ret)
+			*call_send = ret;
+	}
+
+	log_pmtu = qp->log_pmtu;
+	if (qp->ibqp.qp_type != IB_QPT_UC &&
+	    qp->ibqp.qp_type != IB_QPT_RC) {
+		struct rvt_ah *ah = ibah_to_rvtah(wqe->ud_wr.ah);
+
+		log_pmtu = ah->log_pmtu;
+		atomic_inc(&ibah_to_rvtah(ud_wr(wr)->ah)->refcount);
+	}
+
+	wqe->ssn = qp->s_ssn++;
+	wqe->psn = qp->s_next_psn;
+	wqe->lpsn = wqe->psn +
+			(wqe->length ? ((wqe->length - 1) >> log_pmtu) : 0);
+	qp->s_next_psn = wqe->lpsn + 1;
+	trace_rvt_post_one_wr(qp, wqe);
+	smp_wmb(); /* see request builders */
+	qp->s_avail--;
+	qp->s_head = next;
+
+	return 0;
+
+bail_inval_free:
+	/* release mr holds */
+	while (j) {
+		struct rvt_sge *sge = &wqe->sg_list[--j];
+
+		rvt_put_mr(sge->mr);
+	}
+	return ret;
+}
+
+/**
+ * rvt_post_send - post a send on a QP
+ * @ibqp: the QP to post the send on
+ * @wr: the list of work requests to post
+ * @bad_wr: the first bad WR is put here
+ *
+ * This may be called from interrupt context.
+ *
+ * Return: 0 on success else errno
+ */
+int rvt_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+		  struct ib_send_wr **bad_wr)
+{
+	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+	unsigned long flags = 0;
+	int call_send;
+	unsigned nreq = 0;
+	int err = 0;
+
+	spin_lock_irqsave(&qp->s_hlock, flags);
+
+	/*
+	 * Ensure QP state is such that we can send. If not bail out early,
+	 * there is no need to do this every time we post a send.
+	 */
+	if (unlikely(!(ib_rvt_state_ops[qp->state] & RVT_POST_SEND_OK))) {
+		spin_unlock_irqrestore(&qp->s_hlock, flags);
+		return -EINVAL;
+	}
+
+	/*
+	 * If the send queue is empty, and we only have a single WR then just go
+	 * ahead and kick the send engine into gear. Otherwise we will always
+	 * just schedule the send to happen later.
+	 */
+	call_send = qp->s_head == ACCESS_ONCE(qp->s_last) && !wr->next;
+
+	for (; wr; wr = wr->next) {
+		err = rvt_post_one_wr(qp, wr, &call_send);
+		if (unlikely(err)) {
+			*bad_wr = wr;
+			goto bail;
+		}
+		nreq++;
+	}
+bail:
+	spin_unlock_irqrestore(&qp->s_hlock, flags);
+	if (nreq) {
+		if (call_send)
+			rdi->driver_f.schedule_send_no_lock(qp);
+		else
+			rdi->driver_f.do_send(qp);
+	}
+	return err;
+}
+
+/**
+ * rvt_post_srq_receive - post a receive on a shared receive queue
+ * @ibsrq: the SRQ to post the receive on
+ * @wr: the list of work requests to post
+ * @bad_wr: A pointer to the first WR to cause a problem is put here
+ *
+ * This may be called from interrupt context.
+ *
+ * Return: 0 on success else errno
+ */
+int rvt_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
+		      struct ib_recv_wr **bad_wr)
+{
+	struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
+	struct rvt_rwq *wq;
+	unsigned long flags;
+
+	for (; wr; wr = wr->next) {
+		struct rvt_rwqe *wqe;
+		u32 next;
+		int i;
+
+		if ((unsigned)wr->num_sge > srq->rq.max_sge) {
+			*bad_wr = wr;
+			return -EINVAL;
+		}
+
+		spin_lock_irqsave(&srq->rq.lock, flags);
+		wq = srq->rq.wq;
+		next = wq->head + 1;
+		if (next >= srq->rq.size)
+			next = 0;
+		if (next == wq->tail) {
+			spin_unlock_irqrestore(&srq->rq.lock, flags);
+			*bad_wr = wr;
+			return -ENOMEM;
+		}
+
+		wqe = rvt_get_rwqe_ptr(&srq->rq, wq->head);
+		wqe->wr_id = wr->wr_id;
+		wqe->num_sge = wr->num_sge;
+		for (i = 0; i < wr->num_sge; i++)
+			wqe->sg_list[i] = wr->sg_list[i];
+		/* Make sure queue entry is written before the head index. */
+		smp_wmb();
+		wq->head = next;
+		spin_unlock_irqrestore(&srq->rq.lock, flags);
+	}
+	return 0;
+}
diff --git a/drivers/infiniband/sw/rdmavt/qp.h b/drivers/infiniband/sw/rdmavt/qp.h
new file mode 100644
index 000000000000..8409f80d5f25
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/qp.h
@@ -0,0 +1,69 @@
+#ifndef DEF_RVTQP_H
+#define DEF_RVTQP_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+
+int rvt_driver_qp_init(struct rvt_dev_info *rdi);
+void rvt_qp_exit(struct rvt_dev_info *rdi);
+struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
+			    struct ib_qp_init_attr *init_attr,
+			    struct ib_udata *udata);
+int rvt_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		  int attr_mask, struct ib_udata *udata);
+int rvt_destroy_qp(struct ib_qp *ibqp);
+int rvt_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		 int attr_mask, struct ib_qp_init_attr *init_attr);
+int rvt_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+		  struct ib_recv_wr **bad_wr);
+int rvt_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+		  struct ib_send_wr **bad_wr);
+int rvt_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
+		      struct ib_recv_wr **bad_wr);
+#endif          /* DEF_RVTQP_H */
diff --git a/drivers/staging/rdma/hfi1/srq.c b/drivers/infiniband/sw/rdmavt/srq.c
index 67786d417493..f7c48e9023de 100644
--- a/drivers/staging/rdma/hfi1/srq.c
+++ b/drivers/infiniband/sw/rdmavt/srq.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -52,96 +49,50 @@
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
 
-#include "verbs.h"
+#include "srq.h"
+#include "vt.h"
 
 /**
- * hfi1_post_srq_receive - post a receive on a shared receive queue
- * @ibsrq: the SRQ to post the receive on
- * @wr: the list of work requests to post
- * @bad_wr: A pointer to the first WR to cause a problem is put here
+ * rvt_driver_srq_init - init srq resources on a per driver basis
+ * @rdi: rvt dev structure
  *
- * This may be called from interrupt context.
+ * Do any initialization needed when a driver registers with rdmavt.
  */
-int hfi1_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
-			  struct ib_recv_wr **bad_wr)
+void rvt_driver_srq_init(struct rvt_dev_info *rdi)
 {
-	struct hfi1_srq *srq = to_isrq(ibsrq);
-	struct hfi1_rwq *wq;
-	unsigned long flags;
-	int ret;
-
-	for (; wr; wr = wr->next) {
-		struct hfi1_rwqe *wqe;
-		u32 next;
-		int i;
-
-		if ((unsigned) wr->num_sge > srq->rq.max_sge) {
-			*bad_wr = wr;
-			ret = -EINVAL;
-			goto bail;
-		}
-
-		spin_lock_irqsave(&srq->rq.lock, flags);
-		wq = srq->rq.wq;
-		next = wq->head + 1;
-		if (next >= srq->rq.size)
-			next = 0;
-		if (next == wq->tail) {
-			spin_unlock_irqrestore(&srq->rq.lock, flags);
-			*bad_wr = wr;
-			ret = -ENOMEM;
-			goto bail;
-		}
-
-		wqe = get_rwqe_ptr(&srq->rq, wq->head);
-		wqe->wr_id = wr->wr_id;
-		wqe->num_sge = wr->num_sge;
-		for (i = 0; i < wr->num_sge; i++)
-			wqe->sg_list[i] = wr->sg_list[i];
-		/* Make sure queue entry is written before the head index. */
-		smp_wmb();
-		wq->head = next;
-		spin_unlock_irqrestore(&srq->rq.lock, flags);
-	}
-	ret = 0;
-
-bail:
-	return ret;
+	spin_lock_init(&rdi->n_srqs_lock);
+	rdi->n_srqs_allocated = 0;
 }
 
 /**
- * hfi1_create_srq - create a shared receive queue
+ * rvt_create_srq - create a shared receive queue
  * @ibpd: the protection domain of the SRQ to create
  * @srq_init_attr: the attributes of the SRQ
  * @udata: data from libibverbs when creating a user SRQ
+ *
+ * Return: Allocated srq object
  */
-struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
-			       struct ib_srq_init_attr *srq_init_attr,
-			       struct ib_udata *udata)
+struct ib_srq *rvt_create_srq(struct ib_pd *ibpd,
+			      struct ib_srq_init_attr *srq_init_attr,
+			      struct ib_udata *udata)
 {
-	struct hfi1_ibdev *dev = to_idev(ibpd->device);
-	struct hfi1_srq *srq;
+	struct rvt_dev_info *dev = ib_to_rvt(ibpd->device);
+	struct rvt_srq *srq;
 	u32 sz;
 	struct ib_srq *ret;
 
-	if (srq_init_attr->srq_type != IB_SRQT_BASIC) {
-		ret = ERR_PTR(-ENOSYS);
-		goto done;
-	}
+	if (srq_init_attr->srq_type != IB_SRQT_BASIC)
+		return ERR_PTR(-ENOSYS);
 
 	if (srq_init_attr->attr.max_sge == 0 ||
-	    srq_init_attr->attr.max_sge > hfi1_max_srq_sges ||
+	    srq_init_attr->attr.max_sge > dev->dparms.props.max_srq_sge ||
 	    srq_init_attr->attr.max_wr == 0 ||
-	    srq_init_attr->attr.max_wr > hfi1_max_srq_wrs) {
-		ret = ERR_PTR(-EINVAL);
-		goto done;
-	}
+	    srq_init_attr->attr.max_wr > dev->dparms.props.max_srq_wr)
+		return ERR_PTR(-EINVAL);
 
 	srq = kmalloc(sizeof(*srq), GFP_KERNEL);
-	if (!srq) {
-		ret = ERR_PTR(-ENOMEM);
-		goto done;
-	}
+	if (!srq)
+		return ERR_PTR(-ENOMEM);
 
 	/*
 	 * Need to use vmalloc() if we want to support large #s of entries.
@@ -149,8 +100,8 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
 	srq->rq.size = srq_init_attr->attr.max_wr + 1;
 	srq->rq.max_sge = srq_init_attr->attr.max_sge;
 	sz = sizeof(struct ib_sge) * srq->rq.max_sge +
-		sizeof(struct hfi1_rwqe);
-	srq->rq.wq = vmalloc_user(sizeof(struct hfi1_rwq) + srq->rq.size * sz);
+		sizeof(struct rvt_rwqe);
+	srq->rq.wq = vmalloc_user(sizeof(struct rvt_rwq) + srq->rq.size * sz);
 	if (!srq->rq.wq) {
 		ret = ERR_PTR(-ENOMEM);
 		goto bail_srq;
@@ -158,15 +109,15 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
 
 	/*
 	 * Return the address of the RWQ as the offset to mmap.
-	 * See hfi1_mmap() for details.
+	 * See rvt_mmap() for details.
 	 */
 	if (udata && udata->outlen >= sizeof(__u64)) {
 		int err;
-		u32 s = sizeof(struct hfi1_rwq) + srq->rq.size * sz;
+		u32 s = sizeof(struct rvt_rwq) + srq->rq.size * sz;
 
 		srq->ip =
-		    hfi1_create_mmap_info(dev, s, ibpd->uobject->context,
-					  srq->rq.wq);
+		    rvt_create_mmap_info(dev, s, ibpd->uobject->context,
+					 srq->rq.wq);
 		if (!srq->ip) {
 			ret = ERR_PTR(-ENOMEM);
 			goto bail_wq;
@@ -178,8 +129,9 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
 			ret = ERR_PTR(err);
 			goto bail_ip;
 		}
-	} else
+	} else {
 		srq->ip = NULL;
+	}
 
 	/*
 	 * ib_create_srq() will initialize srq->ibsrq.
@@ -190,7 +142,7 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
 	srq->limit = srq_init_attr->attr.srq_limit;
 
 	spin_lock(&dev->n_srqs_lock);
-	if (dev->n_srqs_allocated == hfi1_max_srqs) {
+	if (dev->n_srqs_allocated == dev->dparms.props.max_srq) {
 		spin_unlock(&dev->n_srqs_lock);
 		ret = ERR_PTR(-ENOMEM);
 		goto bail_ip;
@@ -205,8 +157,7 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
 		spin_unlock_irq(&dev->pending_lock);
 	}
 
-	ret = &srq->ibsrq;
-	goto done;
+	return &srq->ibsrq;
 
 bail_ip:
 	kfree(srq->ip);
@@ -214,46 +165,44 @@ bail_wq:
 	vfree(srq->rq.wq);
 bail_srq:
 	kfree(srq);
-done:
 	return ret;
 }
 
 /**
- * hfi1_modify_srq - modify a shared receive queue
+ * rvt_modify_srq - modify a shared receive queue
  * @ibsrq: the SRQ to modify
  * @attr: the new attributes of the SRQ
  * @attr_mask: indicates which attributes to modify
  * @udata: user data for libibverbs.so
+ *
+ * Return: 0 on success
  */
-int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
-		    enum ib_srq_attr_mask attr_mask,
-		    struct ib_udata *udata)
+int rvt_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		   enum ib_srq_attr_mask attr_mask,
+		   struct ib_udata *udata)
 {
-	struct hfi1_srq *srq = to_isrq(ibsrq);
-	struct hfi1_rwq *wq;
+	struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
+	struct rvt_dev_info *dev = ib_to_rvt(ibsrq->device);
+	struct rvt_rwq *wq;
 	int ret = 0;
 
 	if (attr_mask & IB_SRQ_MAX_WR) {
-		struct hfi1_rwq *owq;
-		struct hfi1_rwqe *p;
+		struct rvt_rwq *owq;
+		struct rvt_rwqe *p;
 		u32 sz, size, n, head, tail;
 
 		/* Check that the requested sizes are below the limits. */
-		if ((attr->max_wr > hfi1_max_srq_wrs) ||
+		if ((attr->max_wr > dev->dparms.props.max_srq_wr) ||
 		    ((attr_mask & IB_SRQ_LIMIT) ?
-		     attr->srq_limit : srq->limit) > attr->max_wr) {
-			ret = -EINVAL;
-			goto bail;
-		}
+		     attr->srq_limit : srq->limit) > attr->max_wr)
+			return -EINVAL;
 
-		sz = sizeof(struct hfi1_rwqe) +
+		sz = sizeof(struct rvt_rwqe) +
 			srq->rq.max_sge * sizeof(struct ib_sge);
 		size = attr->max_wr + 1;
-		wq = vmalloc_user(sizeof(struct hfi1_rwq) + size * sz);
-		if (!wq) {
-			ret = -ENOMEM;
-			goto bail;
-		}
+		wq = vmalloc_user(sizeof(struct rvt_rwq) + size * sz);
+		if (!wq)
+			return -ENOMEM;
 
 		/* Check that we can write the offset to mmap. */
 		if (udata && udata->inlen >= sizeof(__u64)) {
@@ -264,8 +213,8 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 						 sizeof(offset_addr));
 			if (ret)
 				goto bail_free;
-			udata->outbuf =
-				(void __user *) (unsigned long) offset_addr;
+			udata->outbuf = (void __user *)
+					(unsigned long)offset_addr;
 			ret = ib_copy_to_udata(udata, &offset,
 					       sizeof(offset));
 			if (ret)
@@ -296,16 +245,16 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		n = 0;
 		p = wq->wq;
 		while (tail != head) {
-			struct hfi1_rwqe *wqe;
+			struct rvt_rwqe *wqe;
 			int i;
 
-			wqe = get_rwqe_ptr(&srq->rq, tail);
+			wqe = rvt_get_rwqe_ptr(&srq->rq, tail);
 			p->wr_id = wqe->wr_id;
 			p->num_sge = wqe->num_sge;
 			for (i = 0; i < wqe->num_sge; i++)
 				p->sg_list[i] = wqe->sg_list[i];
 			n++;
-			p = (struct hfi1_rwqe *)((char *)p + sz);
+			p = (struct rvt_rwqe *)((char *)p + sz);
 			if (++tail >= srq->rq.size)
 				tail = 0;
 		}
@@ -320,21 +269,21 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		vfree(owq);
 
 		if (srq->ip) {
-			struct hfi1_mmap_info *ip = srq->ip;
-			struct hfi1_ibdev *dev = to_idev(srq->ibsrq.device);
-			u32 s = sizeof(struct hfi1_rwq) + size * sz;
+			struct rvt_mmap_info *ip = srq->ip;
+			struct rvt_dev_info *dev = ib_to_rvt(srq->ibsrq.device);
+			u32 s = sizeof(struct rvt_rwq) + size * sz;
 
-			hfi1_update_mmap_info(dev, ip, s, wq);
+			rvt_update_mmap_info(dev, ip, s, wq);
 
 			/*
 			 * Return the offset to mmap.
-			 * See hfi1_mmap() for details.
+			 * See rvt_mmap() for details.
 			 */
 			if (udata && udata->inlen >= sizeof(__u64)) {
 				ret = ib_copy_to_udata(udata, &ip->offset,
 						       sizeof(ip->offset));
 				if (ret)
-					goto bail;
+					return ret;
 			}
 
 			/*
@@ -355,19 +304,24 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 			srq->limit = attr->srq_limit;
 		spin_unlock_irq(&srq->rq.lock);
 	}
-	goto bail;
+	return ret;
 
 bail_unlock:
 	spin_unlock_irq(&srq->rq.lock);
 bail_free:
 	vfree(wq);
-bail:
 	return ret;
 }
 
-int hfi1_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
+/** rvt_query_srq - query srq data
+ * @ibsrq: srq to query
+ * @attr: return info in attr
+ *
+ * Return: always 0
+ */
+int rvt_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
 {
-	struct hfi1_srq *srq = to_isrq(ibsrq);
+	struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
 
 	attr->max_wr = srq->rq.size - 1;
 	attr->max_sge = srq->rq.max_sge;
@@ -376,19 +330,21 @@ int hfi1_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
 }
 
 /**
- * hfi1_destroy_srq - destroy a shared receive queue
- * @ibsrq: the SRQ to destroy
+ * rvt_destroy_srq - destory an srq
+ * @ibsrq: srq object to destroy
+ *
+ * Return always 0
  */
-int hfi1_destroy_srq(struct ib_srq *ibsrq)
+int rvt_destroy_srq(struct ib_srq *ibsrq)
 {
-	struct hfi1_srq *srq = to_isrq(ibsrq);
-	struct hfi1_ibdev *dev = to_idev(ibsrq->device);
+	struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
+	struct rvt_dev_info *dev = ib_to_rvt(ibsrq->device);
 
 	spin_lock(&dev->n_srqs_lock);
 	dev->n_srqs_allocated--;
 	spin_unlock(&dev->n_srqs_lock);
 	if (srq->ip)
-		kref_put(&srq->ip->ref, hfi1_release_mmap_info);
+		kref_put(&srq->ip->ref, rvt_release_mmap_info);
 	else
 		vfree(srq->rq.wq);
 	kfree(srq);
diff --git a/drivers/infiniband/sw/rdmavt/srq.h b/drivers/infiniband/sw/rdmavt/srq.h
new file mode 100644
index 000000000000..bf0eaaf56465
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/srq.h
@@ -0,0 +1,62 @@
+#ifndef DEF_RVTSRQ_H
+#define DEF_RVTSRQ_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+void rvt_driver_srq_init(struct rvt_dev_info *rdi);
+struct ib_srq *rvt_create_srq(struct ib_pd *ibpd,
+			      struct ib_srq_init_attr *srq_init_attr,
+			      struct ib_udata *udata);
+int rvt_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		   enum ib_srq_attr_mask attr_mask,
+		   struct ib_udata *udata);
+int rvt_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr);
+int rvt_destroy_srq(struct ib_srq *ibsrq);
+
+#endif          /* DEF_RVTSRQ_H */
diff --git a/drivers/infiniband/sw/rdmavt/trace.c b/drivers/infiniband/sw/rdmavt/trace.c
new file mode 100644
index 000000000000..d593285a349c
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/trace.c
@@ -0,0 +1,49 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#define CREATE_TRACE_POINTS
+#include "trace.h"
diff --git a/drivers/infiniband/sw/rdmavt/trace.h b/drivers/infiniband/sw/rdmavt/trace.h
new file mode 100644
index 000000000000..6c0457db5499
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/trace.h
@@ -0,0 +1,187 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#undef TRACE_SYSTEM_VAR
+#define TRACE_SYSTEM_VAR rdmavt
+
+#if !defined(__RDMAVT_TRACE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define __RDMAVT_TRACE_H
+
+#include <linux/tracepoint.h>
+#include <linux/trace_seq.h>
+
+#include <rdma/ib_verbs.h>
+#include <rdma/rdma_vt.h>
+
+#define RDI_DEV_ENTRY(rdi)   __string(dev, rdi->driver_f.get_card_name(rdi))
+#define RDI_DEV_ASSIGN(rdi)  __assign_str(dev, rdi->driver_f.get_card_name(rdi))
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM rdmavt
+
+TRACE_EVENT(rvt_dbg,
+	TP_PROTO(struct rvt_dev_info *rdi,
+		 const char *msg),
+	TP_ARGS(rdi, msg),
+	TP_STRUCT__entry(
+		RDI_DEV_ENTRY(rdi)
+		__string(msg, msg)
+	),
+	TP_fast_assign(
+		RDI_DEV_ASSIGN(rdi);
+		__assign_str(msg, msg);
+	),
+	TP_printk("[%s]: %s", __get_str(dev), __get_str(msg))
+);
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM rvt_qphash
+DECLARE_EVENT_CLASS(rvt_qphash_template,
+	TP_PROTO(struct rvt_qp *qp, u32 bucket),
+	TP_ARGS(qp, bucket),
+	TP_STRUCT__entry(
+		RDI_DEV_ENTRY(ib_to_rvt(qp->ibqp.device))
+		__field(u32, qpn)
+		__field(u32, bucket)
+	),
+	TP_fast_assign(
+		RDI_DEV_ASSIGN(ib_to_rvt(qp->ibqp.device))
+		__entry->qpn = qp->ibqp.qp_num;
+		__entry->bucket = bucket;
+	),
+	TP_printk(
+		"[%s] qpn 0x%x bucket %u",
+		__get_str(dev),
+		__entry->qpn,
+		__entry->bucket
+	)
+);
+
+DEFINE_EVENT(rvt_qphash_template, rvt_qpinsert,
+	TP_PROTO(struct rvt_qp *qp, u32 bucket),
+	TP_ARGS(qp, bucket));
+
+DEFINE_EVENT(rvt_qphash_template, rvt_qpremove,
+	TP_PROTO(struct rvt_qp *qp, u32 bucket),
+	TP_ARGS(qp, bucket));
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM rvt_tx
+
+#define wr_opcode_name(opcode) { IB_WR_##opcode, #opcode  }
+#define show_wr_opcode(opcode)                             \
+__print_symbolic(opcode,                                   \
+	wr_opcode_name(RDMA_WRITE),                        \
+	wr_opcode_name(RDMA_WRITE_WITH_IMM),               \
+	wr_opcode_name(SEND),                              \
+	wr_opcode_name(SEND_WITH_IMM),                     \
+	wr_opcode_name(RDMA_READ),                         \
+	wr_opcode_name(ATOMIC_CMP_AND_SWP),                \
+	wr_opcode_name(ATOMIC_FETCH_AND_ADD),              \
+	wr_opcode_name(LSO),                               \
+	wr_opcode_name(SEND_WITH_INV),                     \
+	wr_opcode_name(RDMA_READ_WITH_INV),                \
+	wr_opcode_name(LOCAL_INV),                         \
+	wr_opcode_name(MASKED_ATOMIC_CMP_AND_SWP),         \
+	wr_opcode_name(MASKED_ATOMIC_FETCH_AND_ADD))
+
+#define POS_PRN \
+"[%s] wr_id %llx qpn %x psn 0x%x lpsn 0x%x length %u opcode 0x%.2x,%s size %u avail %u head %u last %u"
+
+TRACE_EVENT(
+	rvt_post_one_wr,
+	TP_PROTO(struct rvt_qp *qp, struct rvt_swqe *wqe),
+	TP_ARGS(qp, wqe),
+	TP_STRUCT__entry(
+		RDI_DEV_ENTRY(ib_to_rvt(qp->ibqp.device))
+		__field(u64, wr_id)
+		__field(u32, qpn)
+		__field(u32, psn)
+		__field(u32, lpsn)
+		__field(u32, length)
+		__field(u32, opcode)
+		__field(u32, size)
+		__field(u32, avail)
+		__field(u32, head)
+		__field(u32, last)
+	),
+	TP_fast_assign(
+		RDI_DEV_ASSIGN(ib_to_rvt(qp->ibqp.device))
+		__entry->wr_id = wqe->wr.wr_id;
+		__entry->qpn = qp->ibqp.qp_num;
+		__entry->psn = wqe->psn;
+		__entry->lpsn = wqe->lpsn;
+		__entry->length = wqe->length;
+		__entry->opcode = wqe->wr.opcode;
+		__entry->size = qp->s_size;
+		__entry->avail = qp->s_avail;
+		__entry->head = qp->s_head;
+		__entry->last = qp->s_last;
+	),
+	TP_printk(
+		POS_PRN,
+		__get_str(dev),
+		__entry->wr_id,
+		__entry->qpn,
+		__entry->psn,
+		__entry->lpsn,
+		__entry->length,
+		__entry->opcode, show_wr_opcode(__entry->opcode),
+		__entry->size,
+		__entry->avail,
+		__entry->head,
+		__entry->last
+	)
+);
+
+#endif /* __RDMAVT_TRACE_H */
+
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_FILE trace
+#include <trace/define_trace.h>
diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c
new file mode 100644
index 000000000000..6caf5272ba1f
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/vt.c
@@ -0,0 +1,873 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include "vt.h"
+#include "trace.h"
+
+#define RVT_UVERBS_ABI_VERSION 2
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_DESCRIPTION("RDMA Verbs Transport Library");
+
+static int rvt_init(void)
+{
+	/*
+	 * rdmavt does not need to do anything special when it starts up. All it
+	 * needs to do is sit and wait until a driver attempts registration.
+	 */
+	return 0;
+}
+module_init(rvt_init);
+
+static void rvt_cleanup(void)
+{
+	/*
+	 * Nothing to do at exit time either. The module won't be able to be
+	 * removed until all drivers are gone which means all the dev structs
+	 * are gone so there is really nothing to do.
+	 */
+}
+module_exit(rvt_cleanup);
+
+/**
+ * rvt_alloc_device - allocate rdi
+ * @size: how big of a structure to allocate
+ * @nports: number of ports to allocate array slots for
+ *
+ * Use IB core device alloc to allocate space for the rdi which is assumed to be
+ * inside of the ib_device. Any extra space that drivers require should be
+ * included in size.
+ *
+ * We also allocate a port array based on the number of ports.
+ *
+ * Return: pointer to allocated rdi
+ */
+struct rvt_dev_info *rvt_alloc_device(size_t size, int nports)
+{
+	struct rvt_dev_info *rdi = ERR_PTR(-ENOMEM);
+
+	rdi = (struct rvt_dev_info *)ib_alloc_device(size);
+	if (!rdi)
+		return rdi;
+
+	rdi->ports = kcalloc(nports,
+			     sizeof(struct rvt_ibport **),
+			     GFP_KERNEL);
+	if (!rdi->ports)
+		ib_dealloc_device(&rdi->ibdev);
+
+	return rdi;
+}
+EXPORT_SYMBOL(rvt_alloc_device);
+
+static int rvt_query_device(struct ib_device *ibdev,
+			    struct ib_device_attr *props,
+			    struct ib_udata *uhw)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+
+	if (uhw->inlen || uhw->outlen)
+		return -EINVAL;
+	/*
+	 * Return rvt_dev_info.dparms.props contents
+	 */
+	*props = rdi->dparms.props;
+	return 0;
+}
+
+static int rvt_modify_device(struct ib_device *device,
+			     int device_modify_mask,
+			     struct ib_device_modify *device_modify)
+{
+	/*
+	 * There is currently no need to supply this based on qib and hfi1.
+	 * Future drivers may need to implement this though.
+	 */
+
+	return -EOPNOTSUPP;
+}
+
+/**
+ * rvt_query_port: Passes the query port call to the driver
+ * @ibdev: Verbs IB dev
+ * @port_num: port number, 1 based from ib core
+ * @props: structure to hold returned properties
+ *
+ * Return: 0 on success
+ */
+static int rvt_query_port(struct ib_device *ibdev, u8 port_num,
+			  struct ib_port_attr *props)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	struct rvt_ibport *rvp;
+	int port_index = ibport_num_to_idx(ibdev, port_num);
+
+	if (port_index < 0)
+		return -EINVAL;
+
+	rvp = rdi->ports[port_index];
+	memset(props, 0, sizeof(*props));
+	props->sm_lid = rvp->sm_lid;
+	props->sm_sl = rvp->sm_sl;
+	props->port_cap_flags = rvp->port_cap_flags;
+	props->max_msg_sz = 0x80000000;
+	props->pkey_tbl_len = rvt_get_npkeys(rdi);
+	props->bad_pkey_cntr = rvp->pkey_violations;
+	props->qkey_viol_cntr = rvp->qkey_violations;
+	props->subnet_timeout = rvp->subnet_timeout;
+	props->init_type_reply = 0;
+
+	/* Populate the remaining ib_port_attr elements */
+	return rdi->driver_f.query_port_state(rdi, port_num, props);
+}
+
+/**
+ * rvt_modify_port
+ * @ibdev: Verbs IB dev
+ * @port_num: Port number, 1 based from ib core
+ * @port_modify_mask: How to change the port
+ * @props: Structure to fill in
+ *
+ * Return: 0 on success
+ */
+static int rvt_modify_port(struct ib_device *ibdev, u8 port_num,
+			   int port_modify_mask, struct ib_port_modify *props)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	struct rvt_ibport *rvp;
+	int ret = 0;
+	int port_index = ibport_num_to_idx(ibdev, port_num);
+
+	if (port_index < 0)
+		return -EINVAL;
+
+	rvp = rdi->ports[port_index];
+	rvp->port_cap_flags |= props->set_port_cap_mask;
+	rvp->port_cap_flags &= ~props->clr_port_cap_mask;
+
+	if (props->set_port_cap_mask || props->clr_port_cap_mask)
+		rdi->driver_f.cap_mask_chg(rdi, port_num);
+	if (port_modify_mask & IB_PORT_SHUTDOWN)
+		ret = rdi->driver_f.shut_down_port(rdi, port_num);
+	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
+		rvp->qkey_violations = 0;
+
+	return ret;
+}
+
+/**
+ * rvt_query_pkey - Return a pkey from the table at a given index
+ * @ibdev: Verbs IB dev
+ * @port_num: Port number, 1 based from ib core
+ * @intex: Index into pkey table
+ *
+ * Return: 0 on failure pkey otherwise
+ */
+static int rvt_query_pkey(struct ib_device *ibdev, u8 port_num, u16 index,
+			  u16 *pkey)
+{
+	/*
+	 * Driver will be responsible for keeping rvt_dev_info.pkey_table up to
+	 * date. This function will just return that value. There is no need to
+	 * lock, if a stale value is read and sent to the user so be it there is
+	 * no way to protect against that anyway.
+	 */
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	int port_index;
+
+	port_index = ibport_num_to_idx(ibdev, port_num);
+	if (port_index < 0)
+		return -EINVAL;
+
+	if (index >= rvt_get_npkeys(rdi))
+		return -EINVAL;
+
+	*pkey = rvt_get_pkey(rdi, port_index, index);
+	return 0;
+}
+
+/**
+ * rvt_query_gid - Return a gid from the table
+ * @ibdev: Verbs IB dev
+ * @port_num: Port number, 1 based from ib core
+ * @index: = Index in table
+ * @gid: Gid to return
+ *
+ * Return: 0 on success
+ */
+static int rvt_query_gid(struct ib_device *ibdev, u8 port_num,
+			 int guid_index, union ib_gid *gid)
+{
+	struct rvt_dev_info *rdi;
+	struct rvt_ibport *rvp;
+	int port_index;
+
+	/*
+	 * Driver is responsible for updating the guid table. Which will be used
+	 * to craft the return value. This will work similar to how query_pkey()
+	 * is being done.
+	 */
+	port_index = ibport_num_to_idx(ibdev, port_num);
+	if (port_index < 0)
+		return -EINVAL;
+
+	rdi = ib_to_rvt(ibdev);
+	rvp = rdi->ports[port_index];
+
+	gid->global.subnet_prefix = rvp->gid_prefix;
+
+	return rdi->driver_f.get_guid_be(rdi, rvp, guid_index,
+					 &gid->global.interface_id);
+}
+
+struct rvt_ucontext {
+	struct ib_ucontext ibucontext;
+};
+
+static inline struct rvt_ucontext *to_iucontext(struct ib_ucontext
+						*ibucontext)
+{
+	return container_of(ibucontext, struct rvt_ucontext, ibucontext);
+}
+
+/**
+ * rvt_alloc_ucontext - Allocate a user context
+ * @ibdev: Vers IB dev
+ * @data: User data allocated
+ */
+static struct ib_ucontext *rvt_alloc_ucontext(struct ib_device *ibdev,
+					      struct ib_udata *udata)
+{
+	struct rvt_ucontext *context;
+
+	context = kmalloc(sizeof(*context), GFP_KERNEL);
+	if (!context)
+		return ERR_PTR(-ENOMEM);
+	return &context->ibucontext;
+}
+
+/**
+ *rvt_dealloc_ucontext - Free a user context
+ *@context - Free this
+ */
+static int rvt_dealloc_ucontext(struct ib_ucontext *context)
+{
+	kfree(to_iucontext(context));
+	return 0;
+}
+
+static int rvt_get_port_immutable(struct ib_device *ibdev, u8 port_num,
+				  struct ib_port_immutable *immutable)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	struct ib_port_attr attr;
+	int err, port_index;
+
+	port_index = ibport_num_to_idx(ibdev, port_num);
+	if (port_index < 0)
+		return -EINVAL;
+
+	err = rvt_query_port(ibdev, port_num, &attr);
+	if (err)
+		return err;
+
+	immutable->pkey_tbl_len = attr.pkey_tbl_len;
+	immutable->gid_tbl_len = attr.gid_tbl_len;
+	immutable->core_cap_flags = rdi->dparms.core_cap_flags;
+	immutable->max_mad_size = rdi->dparms.max_mad_size;
+
+	return 0;
+}
+
+enum {
+	MISC,
+	QUERY_DEVICE,
+	MODIFY_DEVICE,
+	QUERY_PORT,
+	MODIFY_PORT,
+	QUERY_PKEY,
+	QUERY_GID,
+	ALLOC_UCONTEXT,
+	DEALLOC_UCONTEXT,
+	GET_PORT_IMMUTABLE,
+	CREATE_QP,
+	MODIFY_QP,
+	DESTROY_QP,
+	QUERY_QP,
+	POST_SEND,
+	POST_RECV,
+	POST_SRQ_RECV,
+	CREATE_AH,
+	DESTROY_AH,
+	MODIFY_AH,
+	QUERY_AH,
+	CREATE_SRQ,
+	MODIFY_SRQ,
+	DESTROY_SRQ,
+	QUERY_SRQ,
+	ATTACH_MCAST,
+	DETACH_MCAST,
+	GET_DMA_MR,
+	REG_USER_MR,
+	DEREG_MR,
+	ALLOC_MR,
+	ALLOC_FMR,
+	MAP_PHYS_FMR,
+	UNMAP_FMR,
+	DEALLOC_FMR,
+	MMAP,
+	CREATE_CQ,
+	DESTROY_CQ,
+	POLL_CQ,
+	REQ_NOTFIY_CQ,
+	RESIZE_CQ,
+	ALLOC_PD,
+	DEALLOC_PD,
+	_VERB_IDX_MAX /* Must always be last! */
+};
+
+static inline int check_driver_override(struct rvt_dev_info *rdi,
+					size_t offset, void *func)
+{
+	if (!*(void **)((void *)&rdi->ibdev + offset)) {
+		*(void **)((void *)&rdi->ibdev + offset) = func;
+		return 0;
+	}
+
+	return 1;
+}
+
+static noinline int check_support(struct rvt_dev_info *rdi, int verb)
+{
+	switch (verb) {
+	case MISC:
+		/*
+		 * These functions are not part of verbs specifically but are
+		 * required for rdmavt to function.
+		 */
+		if ((!rdi->driver_f.port_callback) ||
+		    (!rdi->driver_f.get_card_name) ||
+		    (!rdi->driver_f.get_pci_dev))
+			return -EINVAL;
+		break;
+
+	case QUERY_DEVICE:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    query_device),
+						    rvt_query_device);
+		break;
+
+	case MODIFY_DEVICE:
+		/*
+		 * rdmavt does not support modify device currently drivers must
+		 * provide.
+		 */
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 modify_device),
+					   rvt_modify_device))
+			return -EOPNOTSUPP;
+		break;
+
+	case QUERY_PORT:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 query_port),
+					   rvt_query_port))
+			if (!rdi->driver_f.query_port_state)
+				return -EINVAL;
+		break;
+
+	case MODIFY_PORT:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 modify_port),
+					   rvt_modify_port))
+			if (!rdi->driver_f.cap_mask_chg ||
+			    !rdi->driver_f.shut_down_port)
+				return -EINVAL;
+		break;
+
+	case QUERY_PKEY:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    query_pkey),
+				      rvt_query_pkey);
+		break;
+
+	case QUERY_GID:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 query_gid),
+					   rvt_query_gid))
+			if (!rdi->driver_f.get_guid_be)
+				return -EINVAL;
+		break;
+
+	case ALLOC_UCONTEXT:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    alloc_ucontext),
+				      rvt_alloc_ucontext);
+		break;
+
+	case DEALLOC_UCONTEXT:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    dealloc_ucontext),
+				      rvt_dealloc_ucontext);
+		break;
+
+	case GET_PORT_IMMUTABLE:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    get_port_immutable),
+				      rvt_get_port_immutable);
+		break;
+
+	case CREATE_QP:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 create_qp),
+					   rvt_create_qp))
+			if (!rdi->driver_f.qp_priv_alloc ||
+			    !rdi->driver_f.qp_priv_free ||
+			    !rdi->driver_f.notify_qp_reset ||
+			    !rdi->driver_f.flush_qp_waiters ||
+			    !rdi->driver_f.stop_send_queue ||
+			    !rdi->driver_f.quiesce_qp)
+				return -EINVAL;
+		break;
+
+	case MODIFY_QP:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 modify_qp),
+					   rvt_modify_qp))
+			if (!rdi->driver_f.notify_qp_reset ||
+			    !rdi->driver_f.schedule_send ||
+			    !rdi->driver_f.get_pmtu_from_attr ||
+			    !rdi->driver_f.flush_qp_waiters ||
+			    !rdi->driver_f.stop_send_queue ||
+			    !rdi->driver_f.quiesce_qp ||
+			    !rdi->driver_f.notify_error_qp ||
+			    !rdi->driver_f.mtu_from_qp ||
+			    !rdi->driver_f.mtu_to_path_mtu ||
+			    !rdi->driver_f.shut_down_port ||
+			    !rdi->driver_f.cap_mask_chg)
+				return -EINVAL;
+		break;
+
+	case DESTROY_QP:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 destroy_qp),
+					   rvt_destroy_qp))
+			if (!rdi->driver_f.qp_priv_free ||
+			    !rdi->driver_f.notify_qp_reset ||
+			    !rdi->driver_f.flush_qp_waiters ||
+			    !rdi->driver_f.stop_send_queue ||
+			    !rdi->driver_f.quiesce_qp)
+				return -EINVAL;
+		break;
+
+	case QUERY_QP:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    query_qp),
+						    rvt_query_qp);
+		break;
+
+	case POST_SEND:
+		if (!check_driver_override(rdi, offsetof(struct ib_device,
+							 post_send),
+					   rvt_post_send))
+			if (!rdi->driver_f.schedule_send ||
+			    !rdi->driver_f.do_send)
+				return -EINVAL;
+		break;
+
+	case POST_RECV:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    post_recv),
+				      rvt_post_recv);
+		break;
+	case POST_SRQ_RECV:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    post_srq_recv),
+				      rvt_post_srq_recv);
+		break;
+
+	case CREATE_AH:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    create_ah),
+				      rvt_create_ah);
+		break;
+
+	case DESTROY_AH:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    destroy_ah),
+				      rvt_destroy_ah);
+		break;
+
+	case MODIFY_AH:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    modify_ah),
+				      rvt_modify_ah);
+		break;
+
+	case QUERY_AH:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    query_ah),
+				      rvt_query_ah);
+		break;
+
+	case CREATE_SRQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    create_srq),
+				      rvt_create_srq);
+		break;
+
+	case MODIFY_SRQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    modify_srq),
+				      rvt_modify_srq);
+		break;
+
+	case DESTROY_SRQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    destroy_srq),
+				      rvt_destroy_srq);
+		break;
+
+	case QUERY_SRQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    query_srq),
+				      rvt_query_srq);
+		break;
+
+	case ATTACH_MCAST:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    attach_mcast),
+				      rvt_attach_mcast);
+		break;
+
+	case DETACH_MCAST:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    detach_mcast),
+				      rvt_detach_mcast);
+		break;
+
+	case GET_DMA_MR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    get_dma_mr),
+				      rvt_get_dma_mr);
+		break;
+
+	case REG_USER_MR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    reg_user_mr),
+				      rvt_reg_user_mr);
+		break;
+
+	case DEREG_MR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    dereg_mr),
+				      rvt_dereg_mr);
+		break;
+
+	case ALLOC_FMR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    alloc_fmr),
+				      rvt_alloc_fmr);
+		break;
+
+	case ALLOC_MR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    alloc_mr),
+				      rvt_alloc_mr);
+		break;
+
+	case MAP_PHYS_FMR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    map_phys_fmr),
+				      rvt_map_phys_fmr);
+		break;
+
+	case UNMAP_FMR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    unmap_fmr),
+				      rvt_unmap_fmr);
+		break;
+
+	case DEALLOC_FMR:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    dealloc_fmr),
+				      rvt_dealloc_fmr);
+		break;
+
+	case MMAP:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    mmap),
+				      rvt_mmap);
+		break;
+
+	case CREATE_CQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    create_cq),
+				      rvt_create_cq);
+		break;
+
+	case DESTROY_CQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    destroy_cq),
+				      rvt_destroy_cq);
+		break;
+
+	case POLL_CQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    poll_cq),
+				      rvt_poll_cq);
+		break;
+
+	case REQ_NOTFIY_CQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    req_notify_cq),
+				      rvt_req_notify_cq);
+		break;
+
+	case RESIZE_CQ:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    resize_cq),
+				      rvt_resize_cq);
+		break;
+
+	case ALLOC_PD:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    alloc_pd),
+				      rvt_alloc_pd);
+		break;
+
+	case DEALLOC_PD:
+		check_driver_override(rdi, offsetof(struct ib_device,
+						    dealloc_pd),
+				      rvt_dealloc_pd);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/**
+ * rvt_register_device - register a driver
+ * @rdi: main dev structure for all of rdmavt operations
+ *
+ * It is up to drivers to allocate the rdi and fill in the appropriate
+ * information.
+ *
+ * Return: 0 on success otherwise an errno.
+ */
+int rvt_register_device(struct rvt_dev_info *rdi)
+{
+	int ret = 0, i;
+
+	if (!rdi)
+		return -EINVAL;
+
+	/*
+	 * Check to ensure drivers have setup the required helpers for the verbs
+	 * they want rdmavt to handle
+	 */
+	for (i = 0; i < _VERB_IDX_MAX; i++)
+		if (check_support(rdi, i)) {
+			pr_err("Driver support req not met at %d\n", i);
+			return -EINVAL;
+		}
+
+
+	/* Once we get past here we can use rvt_pr macros and tracepoints */
+	trace_rvt_dbg(rdi, "Driver attempting registration");
+	rvt_mmap_init(rdi);
+
+	/* Queue Pairs */
+	ret = rvt_driver_qp_init(rdi);
+	if (ret) {
+		pr_err("Error in driver QP init.\n");
+		return -EINVAL;
+	}
+
+	/* Address Handle */
+	spin_lock_init(&rdi->n_ahs_lock);
+	rdi->n_ahs_allocated = 0;
+
+	/* Shared Receive Queue */
+	rvt_driver_srq_init(rdi);
+
+	/* Multicast */
+	rvt_driver_mcast_init(rdi);
+
+	/* Mem Region */
+	ret = rvt_driver_mr_init(rdi);
+	if (ret) {
+		pr_err("Error in driver MR init.\n");
+		goto bail_no_mr;
+	}
+
+	/* Completion queues */
+	ret = rvt_driver_cq_init(rdi);
+	if (ret) {
+		pr_err("Error in driver CQ init.\n");
+		goto bail_mr;
+	}
+
+	/* DMA Operations */
+	rdi->ibdev.dma_ops =
+		rdi->ibdev.dma_ops ? : &rvt_default_dma_mapping_ops;
+
+	/* Protection Domain */
+	spin_lock_init(&rdi->n_pds_lock);
+	rdi->n_pds_allocated = 0;
+
+	/*
+	 * There are some things which could be set by underlying drivers but
+	 * really should be up to rdmavt to set. For instance drivers can't know
+	 * exactly which functions rdmavt supports, nor do they know the ABI
+	 * version, so we do all of this sort of stuff here.
+	 */
+	rdi->ibdev.uverbs_abi_ver = RVT_UVERBS_ABI_VERSION;
+	rdi->ibdev.uverbs_cmd_mask =
+		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)         |
+		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)        |
+		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)          |
+		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)            |
+		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)          |
+		(1ull << IB_USER_VERBS_CMD_CREATE_AH)           |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_AH)           |
+		(1ull << IB_USER_VERBS_CMD_QUERY_AH)            |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)          |
+		(1ull << IB_USER_VERBS_CMD_REG_MR)              |
+		(1ull << IB_USER_VERBS_CMD_DEREG_MR)            |
+		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
+		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)           |
+		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)          |
+		(1ull << IB_USER_VERBS_CMD_POLL_CQ)             |
+		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)       |
+		(1ull << IB_USER_VERBS_CMD_CREATE_QP)           |
+		(1ull << IB_USER_VERBS_CMD_QUERY_QP)            |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)          |
+		(1ull << IB_USER_VERBS_CMD_POST_SEND)           |
+		(1ull << IB_USER_VERBS_CMD_POST_RECV)           |
+		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)        |
+		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)        |
+		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)          |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)          |
+		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)         |
+		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
+	rdi->ibdev.node_type = RDMA_NODE_IB_CA;
+	rdi->ibdev.num_comp_vectors = 1;
+
+	/* We are now good to announce we exist */
+	ret =  ib_register_device(&rdi->ibdev, rdi->driver_f.port_callback);
+	if (ret) {
+		rvt_pr_err(rdi, "Failed to register driver with ib core.\n");
+		goto bail_cq;
+	}
+
+	rvt_create_mad_agents(rdi);
+
+	rvt_pr_info(rdi, "Registration with rdmavt done.\n");
+	return ret;
+
+bail_cq:
+	rvt_cq_exit(rdi);
+
+bail_mr:
+	rvt_mr_exit(rdi);
+
+bail_no_mr:
+	rvt_qp_exit(rdi);
+
+	return ret;
+}
+EXPORT_SYMBOL(rvt_register_device);
+
+/**
+ * rvt_unregister_device - remove a driver
+ * @rdi: rvt dev struct
+ */
+void rvt_unregister_device(struct rvt_dev_info *rdi)
+{
+	trace_rvt_dbg(rdi, "Driver is unregistering.");
+	if (!rdi)
+		return;
+
+	rvt_free_mad_agents(rdi);
+
+	ib_unregister_device(&rdi->ibdev);
+	rvt_cq_exit(rdi);
+	rvt_mr_exit(rdi);
+	rvt_qp_exit(rdi);
+}
+EXPORT_SYMBOL(rvt_unregister_device);
+
+/**
+ * rvt_init_port - init internal data for driver port
+ * @rdi: rvt dev strut
+ * @port: rvt port
+ * @port_index: 0 based index of ports, different from IB core port num
+ *
+ * Keep track of a list of ports. No need to have a detach port.
+ * They persist until the driver goes away.
+ *
+ * Return: always 0
+ */
+int rvt_init_port(struct rvt_dev_info *rdi, struct rvt_ibport *port,
+		  int port_index, u16 *pkey_table)
+{
+
+	rdi->ports[port_index] = port;
+	rdi->ports[port_index]->pkey_table = pkey_table;
+
+	return 0;
+}
+EXPORT_SYMBOL(rvt_init_port);
diff --git a/drivers/infiniband/sw/rdmavt/vt.h b/drivers/infiniband/sw/rdmavt/vt.h
new file mode 100644
index 000000000000..6b01eaa4461b
--- /dev/null
+++ b/drivers/infiniband/sw/rdmavt/vt.h
@@ -0,0 +1,104 @@
+#ifndef DEF_RDMAVT_H
+#define DEF_RDMAVT_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <rdma/rdma_vt.h>
+#include <linux/pci.h>
+#include "dma.h"
+#include "pd.h"
+#include "qp.h"
+#include "ah.h"
+#include "mr.h"
+#include "srq.h"
+#include "mcast.h"
+#include "mmap.h"
+#include "cq.h"
+#include "mad.h"
+#include "mmap.h"
+
+#define rvt_pr_info(rdi, fmt, ...) \
+	__rvt_pr_info(rdi->driver_f.get_pci_dev(rdi), \
+		      rdi->driver_f.get_card_name(rdi), \
+		      fmt, \
+		      ##__VA_ARGS__)
+
+#define rvt_pr_warn(rdi, fmt, ...) \
+	__rvt_pr_warn(rdi->driver_f.get_pci_dev(rdi), \
+		      rdi->driver_f.get_card_name(rdi), \
+		      fmt, \
+		      ##__VA_ARGS__)
+
+#define rvt_pr_err(rdi, fmt, ...) \
+	__rvt_pr_err(rdi->driver_f.get_pci_dev(rdi), \
+		     rdi->driver_f.get_card_name(rdi), \
+		     fmt, \
+		     ##__VA_ARGS__)
+
+#define __rvt_pr_info(pdev, name, fmt, ...) \
+	dev_info(&pdev->dev, "%s: " fmt, name, ##__VA_ARGS__)
+
+#define __rvt_pr_warn(pdev, name, fmt, ...) \
+	dev_warn(&pdev->dev, "%s: " fmt, name, ##__VA_ARGS__)
+
+#define __rvt_pr_err(pdev, name, fmt, ...) \
+	dev_err(&pdev->dev, "%s: " fmt, name, ##__VA_ARGS__)
+
+static inline int ibport_num_to_idx(struct ib_device *ibdev, u8 port_num)
+{
+	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
+	int port_index;
+
+	port_index = port_num - 1; /* IB ports start at 1 our arrays at 0 */
+	if ((port_index < 0) || (port_index >= rdi->dparms.nports))
+		return -EINVAL;
+
+	return port_index;
+}
+
+#endif          /* DEF_RDMAVT_H */
diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h
index 85be0de3ab26..caec8e9c4666 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib.h
+++ b/drivers/infiniband/ulp/ipoib/ipoib.h
@@ -388,7 +388,7 @@ struct ipoib_dev_priv {
 	struct dentry *mcg_dentry;
 	struct dentry *path_dentry;
 #endif
-	int	hca_caps;
+	u64	hca_caps;
 	struct ipoib_ethtool_st ethtool;
 	struct timer_list poll_timer;
 	unsigned max_send_sge;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index 899e6b7fb8a5..f0e55e47eb54 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -180,6 +180,7 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	struct sk_buff *skb;
 	u64 mapping[IPOIB_UD_RX_SG];
 	union ib_gid *dgid;
+	union ib_gid *sgid;
 
 	ipoib_dbg_data(priv, "recv completion: id %d, status: %d\n",
 		       wr_id, wc->status);
@@ -203,13 +204,6 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 		return;
 	}
 
-	/*
-	 * Drop packets that this interface sent, ie multicast packets
-	 * that the HCA has replicated.
-	 */
-	if (wc->slid == priv->local_lid && wc->src_qp == priv->qp->qp_num)
-		goto repost;
-
 	memcpy(mapping, priv->rx_ring[wr_id].mapping,
 	       IPOIB_UD_RX_SG * sizeof *mapping);
 
@@ -239,6 +233,25 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	else
 		skb->pkt_type = PACKET_MULTICAST;
 
+	sgid = &((struct ib_grh *)skb->data)->sgid;
+
+	/*
+	 * Drop packets that this interface sent, ie multicast packets
+	 * that the HCA has replicated.
+	 */
+	if (wc->slid == priv->local_lid && wc->src_qp == priv->qp->qp_num) {
+		int need_repost = 1;
+
+		if ((wc->wc_flags & IB_WC_GRH) &&
+		    sgid->global.interface_id != priv->local_gid.global.interface_id)
+			need_repost = 0;
+
+		if (need_repost) {
+			dev_kfree_skb_any(skb);
+			goto repost;
+		}
+	}
+
 	skb_pull(skb, IB_GRH_BYTES);
 
 	skb->protocol = ((struct ipoib_header *) skb->data)->proto;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 25509bbd4a05..80807d6e5c4c 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -51,6 +51,7 @@
 #include <net/addrconf.h>
 #include <linux/inetdevice.h>
 #include <rdma/ib_cache.h>
+#include <linux/pci.h>
 
 #define DRV_VERSION "1.0.0"
 
@@ -1590,11 +1591,67 @@ void ipoib_dev_cleanup(struct net_device *dev)
 	priv->tx_ring = NULL;
 }
 
+static int ipoib_set_vf_link_state(struct net_device *dev, int vf, int link_state)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+
+	return ib_set_vf_link_state(priv->ca, vf, priv->port, link_state);
+}
+
+static int ipoib_get_vf_config(struct net_device *dev, int vf,
+			       struct ifla_vf_info *ivf)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	int err;
+
+	err = ib_get_vf_config(priv->ca, vf, priv->port, ivf);
+	if (err)
+		return err;
+
+	ivf->vf = vf;
+
+	return 0;
+}
+
+static int ipoib_set_vf_guid(struct net_device *dev, int vf, u64 guid, int type)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+
+	if (type != IFLA_VF_IB_NODE_GUID && type != IFLA_VF_IB_PORT_GUID)
+		return -EINVAL;
+
+	return ib_set_vf_guid(priv->ca, vf, priv->port, guid, type);
+}
+
+static int ipoib_get_vf_stats(struct net_device *dev, int vf,
+			      struct ifla_vf_stats *vf_stats)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+
+	return ib_get_vf_stats(priv->ca, vf, priv->port, vf_stats);
+}
+
 static const struct header_ops ipoib_header_ops = {
 	.create	= ipoib_hard_header,
 };
 
-static const struct net_device_ops ipoib_netdev_ops = {
+static const struct net_device_ops ipoib_netdev_ops_pf = {
+	.ndo_uninit		 = ipoib_uninit,
+	.ndo_open		 = ipoib_open,
+	.ndo_stop		 = ipoib_stop,
+	.ndo_change_mtu		 = ipoib_change_mtu,
+	.ndo_fix_features	 = ipoib_fix_features,
+	.ndo_start_xmit		 = ipoib_start_xmit,
+	.ndo_tx_timeout		 = ipoib_timeout,
+	.ndo_set_rx_mode	 = ipoib_set_mcast_list,
+	.ndo_get_iflink		 = ipoib_get_iflink,
+	.ndo_set_vf_link_state	 = ipoib_set_vf_link_state,
+	.ndo_get_vf_config	 = ipoib_get_vf_config,
+	.ndo_get_vf_stats	 = ipoib_get_vf_stats,
+	.ndo_set_vf_guid	 = ipoib_set_vf_guid,
+};
+
+static const struct net_device_ops ipoib_netdev_ops_vf = {
 	.ndo_uninit		 = ipoib_uninit,
 	.ndo_open		 = ipoib_open,
 	.ndo_stop		 = ipoib_stop,
@@ -1610,7 +1667,11 @@ void ipoib_setup(struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 
-	dev->netdev_ops		 = &ipoib_netdev_ops;
+	if (priv->hca_caps & IB_DEVICE_VIRTUAL_FUNCTION)
+		dev->netdev_ops	= &ipoib_netdev_ops_vf;
+	else
+		dev->netdev_ops	= &ipoib_netdev_ops_pf;
+
 	dev->header_ops		 = &ipoib_header_ops;
 
 	ipoib_set_ethtool_ops(dev);
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index 1d1309091aba..0bd3cb2f3c67 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -839,7 +839,7 @@ static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc)
 		if (srpt_set_ch_state(ch, CH_DISCONNECTED))
 			schedule_work(&ch->release_work);
 		else
-			WARN_ONCE("%s-%d\n", ch->sess_name, ch->qp->qp_num);
+			WARN_ONCE(1, "%s-%d\n", ch->sess_name, ch->qp->qp_num);
 	}
 }
 
diff --git a/drivers/net/ethernet/intel/i40e/Makefile b/drivers/net/ethernet/intel/i40e/Makefile
index b4729ba57c9c..3b3c63e54ed6 100644
--- a/drivers/net/ethernet/intel/i40e/Makefile
+++ b/drivers/net/ethernet/intel/i40e/Makefile
@@ -41,6 +41,7 @@ i40e-objs := i40e_main.o \
 	i40e_diag.o	\
 	i40e_txrx.o	\
 	i40e_ptp.o	\
+	i40e_client.o   \
 	i40e_virtchnl_pf.o
 
 i40e-$(CONFIG_I40E_DCB) += i40e_dcb.o i40e_dcb_nl.o
diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index 2f6210ae8ba0..1ce6e9c0427d 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -58,6 +58,7 @@
 #ifdef I40E_FCOE
 #include "i40e_fcoe.h"
 #endif
+#include "i40e_client.h"
 #include "i40e_virtchnl.h"
 #include "i40e_virtchnl_pf.h"
 #include "i40e_txrx.h"
@@ -190,6 +191,7 @@ struct i40e_lump_tracking {
 	u16 search_hint;
 	u16 list[0];
 #define I40E_PILE_VALID_BIT  0x8000
+#define I40E_IWARP_IRQ_PILE_ID  (I40E_PILE_VALID_BIT - 2)
 };
 
 #define I40E_DEFAULT_ATR_SAMPLE_RATE	20
@@ -282,6 +284,8 @@ struct i40e_pf {
 #endif /* I40E_FCOE */
 	u16 num_lan_qps;           /* num lan queues this PF has set up */
 	u16 num_lan_msix;          /* num queue vectors for the base PF vsi */
+	u16 num_iwarp_msix;        /* num of iwarp vectors for this PF */
+	int iwarp_base_vector;
 	int queues_left;           /* queues left unclaimed */
 	u16 alloc_rss_size;        /* allocated RSS queues */
 	u16 rss_size_max;          /* HW defined max RSS queues */
@@ -329,6 +333,7 @@ struct i40e_pf {
 #define I40E_FLAG_16BYTE_RX_DESC_ENABLED	BIT_ULL(13)
 #define I40E_FLAG_CLEAN_ADMINQ			BIT_ULL(14)
 #define I40E_FLAG_FILTER_SYNC			BIT_ULL(15)
+#define I40E_FLAG_SERVICE_CLIENT_REQUESTED	BIT_ULL(16)
 #define I40E_FLAG_PROCESS_MDD_EVENT		BIT_ULL(17)
 #define I40E_FLAG_PROCESS_VFLR_EVENT		BIT_ULL(18)
 #define I40E_FLAG_SRIOV_ENABLED			BIT_ULL(19)
@@ -571,6 +576,8 @@ struct i40e_vsi {
 	struct kobject *kobj;  /* sysfs object */
 	bool current_isup;     /* Sync 'link up' logging */
 
+	void *priv;	/* client driver data reference. */
+
 	/* VSI specific handlers */
 	irqreturn_t (*irq_handler)(int irq, void *data);
 
@@ -728,6 +735,10 @@ void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
 			      struct i40e_vsi_context *ctxt,
 			      u8 enabled_tc, bool is_add);
 #endif
+void i40e_service_event_schedule(struct i40e_pf *pf);
+void i40e_notify_client_of_vf_msg(struct i40e_vsi *vsi, u32 vf_id,
+				  u8 *msg, u16 len);
+
 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool enable);
 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count);
 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags, u16 uplink_seid,
@@ -750,6 +761,17 @@ static inline void i40e_dbg_pf_exit(struct i40e_pf *pf) {}
 static inline void i40e_dbg_init(void) {}
 static inline void i40e_dbg_exit(void) {}
 #endif /* CONFIG_DEBUG_FS*/
+/* needed by client drivers */
+int i40e_lan_add_device(struct i40e_pf *pf);
+int i40e_lan_del_device(struct i40e_pf *pf);
+void i40e_client_subtask(struct i40e_pf *pf);
+void i40e_notify_client_of_l2_param_changes(struct i40e_vsi *vsi);
+void i40e_notify_client_of_netdev_open(struct i40e_vsi *vsi);
+void i40e_notify_client_of_netdev_close(struct i40e_vsi *vsi, bool reset);
+void i40e_notify_client_of_vf_enable(struct i40e_pf *pf, u32 num_vfs);
+void i40e_notify_client_of_vf_reset(struct i40e_pf *pf, u32 vf_id);
+int i40e_vf_client_capable(struct i40e_pf *pf, u32 vf_id,
+			   enum i40e_client_type type);
 /**
  * i40e_irq_dynamic_enable - Enable default interrupt generation settings
  * @vsi: pointer to a vsi
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c
new file mode 100644
index 000000000000..0e6ac841321c
--- /dev/null
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.c
@@ -0,0 +1,1012 @@
+/*******************************************************************************
+ *
+ * Intel Ethernet Controller XL710 Family Linux Driver
+ * Copyright(c) 2013 - 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * The full GNU General Public License is included in this distribution in
+ * the file called "COPYING".
+ *
+ * Contact Information:
+ * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
+ * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
+ *
+ ******************************************************************************/
+
+#include <linux/list.h>
+#include <linux/errno.h>
+
+#include "i40e.h"
+#include "i40e_prototype.h"
+#include "i40e_client.h"
+
+static const char i40e_client_interface_version_str[] = I40E_CLIENT_VERSION_STR;
+
+static LIST_HEAD(i40e_devices);
+static DEFINE_MUTEX(i40e_device_mutex);
+
+static LIST_HEAD(i40e_clients);
+static DEFINE_MUTEX(i40e_client_mutex);
+
+static LIST_HEAD(i40e_client_instances);
+static DEFINE_MUTEX(i40e_client_instance_mutex);
+
+static int i40e_client_virtchnl_send(struct i40e_info *ldev,
+				     struct i40e_client *client,
+				     u32 vf_id, u8 *msg, u16 len);
+
+static int i40e_client_setup_qvlist(struct i40e_info *ldev,
+				    struct i40e_client *client,
+				    struct i40e_qvlist_info *qvlist_info);
+
+static void i40e_client_request_reset(struct i40e_info *ldev,
+				      struct i40e_client *client,
+				      u32 reset_level);
+
+static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
+				       struct i40e_client *client,
+				       bool is_vf, u32 vf_id,
+				       u32 flag, u32 valid_flag);
+
+static struct i40e_ops i40e_lan_ops = {
+	.virtchnl_send = i40e_client_virtchnl_send,
+	.setup_qvlist = i40e_client_setup_qvlist,
+	.request_reset = i40e_client_request_reset,
+	.update_vsi_ctxt = i40e_client_update_vsi_ctxt,
+};
+
+/**
+ * i40e_client_type_to_vsi_type - convert client type to vsi type
+ * @client_type: the i40e_client type
+ *
+ * returns the related vsi type value
+ **/
+static
+enum i40e_vsi_type i40e_client_type_to_vsi_type(enum i40e_client_type type)
+{
+	switch (type) {
+	case I40E_CLIENT_IWARP:
+		return I40E_VSI_IWARP;
+
+	case I40E_CLIENT_VMDQ2:
+		return I40E_VSI_VMDQ2;
+
+	default:
+		pr_err("i40e: Client type unknown\n");
+		return I40E_VSI_TYPE_UNKNOWN;
+	}
+}
+
+/**
+ * i40e_client_get_params - Get the params that can change at runtime
+ * @vsi: the VSI with the message
+ * @param: clinet param struct
+ *
+ **/
+static
+int i40e_client_get_params(struct i40e_vsi *vsi, struct i40e_params *params)
+{
+	struct i40e_dcbx_config *dcb_cfg = &vsi->back->hw.local_dcbx_config;
+	int i = 0;
+
+	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
+		u8 tc = dcb_cfg->etscfg.prioritytable[i];
+		u16 qs_handle;
+
+		/* If TC is not enabled for VSI use TC0 for UP */
+		if (!(vsi->tc_config.enabled_tc & BIT(tc)))
+			tc = 0;
+
+		qs_handle = le16_to_cpu(vsi->info.qs_handle[tc]);
+		params->qos.prio_qos[i].tc = tc;
+		params->qos.prio_qos[i].qs_handle = qs_handle;
+		if (qs_handle == I40E_AQ_VSI_QS_HANDLE_INVALID) {
+			dev_err(&vsi->back->pdev->dev, "Invalid queue set handle for TC = %d, vsi id = %d\n",
+				tc, vsi->id);
+			return -EINVAL;
+		}
+	}
+
+	params->mtu = vsi->netdev->mtu;
+	return 0;
+}
+
+/**
+ * i40e_notify_client_of_vf_msg - call the client vf message callback
+ * @vsi: the VSI with the message
+ * @vf_id: the absolute VF id that sent the message
+ * @msg: message buffer
+ * @len: length of the message
+ *
+ * If there is a client to this VSI, call the client
+ **/
+void
+i40e_notify_client_of_vf_msg(struct i40e_vsi *vsi, u32 vf_id, u8 *msg, u16 len)
+{
+	struct i40e_client_instance *cdev;
+
+	if (!vsi)
+		return;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.pf == vsi->back) {
+			if (!cdev->client ||
+			    !cdev->client->ops ||
+			    !cdev->client->ops->virtchnl_receive) {
+				dev_dbg(&vsi->back->pdev->dev,
+					"Cannot locate client instance virtual channel receive routine\n");
+				continue;
+			}
+			cdev->client->ops->virtchnl_receive(&cdev->lan_info,
+							    cdev->client,
+							    vf_id, msg, len);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_notify_client_of_l2_param_changes - call the client notify callback
+ * @vsi: the VSI with l2 param changes
+ *
+ * If there is a client to this VSI, call the client
+ **/
+void i40e_notify_client_of_l2_param_changes(struct i40e_vsi *vsi)
+{
+	struct i40e_client_instance *cdev;
+	struct i40e_params params;
+
+	if (!vsi)
+		return;
+	memset(&params, 0, sizeof(params));
+	i40e_client_get_params(vsi, &params);
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.pf == vsi->back) {
+			if (!cdev->client ||
+			    !cdev->client->ops ||
+			    !cdev->client->ops->l2_param_change) {
+				dev_dbg(&vsi->back->pdev->dev,
+					"Cannot locate client instance l2_param_change routine\n");
+				continue;
+			}
+			cdev->lan_info.params = params;
+			cdev->client->ops->l2_param_change(&cdev->lan_info,
+							   cdev->client,
+							   &params);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_notify_client_of_netdev_open - call the client open callback
+ * @vsi: the VSI with netdev opened
+ *
+ * If there is a client to this netdev, call the client with open
+ **/
+void i40e_notify_client_of_netdev_open(struct i40e_vsi *vsi)
+{
+	struct i40e_client_instance *cdev;
+
+	if (!vsi)
+		return;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.netdev == vsi->netdev) {
+			if (!cdev->client ||
+			    !cdev->client->ops || !cdev->client->ops->open) {
+				dev_dbg(&vsi->back->pdev->dev,
+					"Cannot locate client instance open routine\n");
+				continue;
+			}
+			cdev->client->ops->open(&cdev->lan_info, cdev->client);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_client_release_qvlist
+ * @ldev: pointer to L2 context.
+ *
+ **/
+static void i40e_client_release_qvlist(struct i40e_info *ldev)
+{
+	struct i40e_qvlist_info *qvlist_info = ldev->qvlist_info;
+	u32 i;
+
+	if (!ldev->qvlist_info)
+		return;
+
+	for (i = 0; i < qvlist_info->num_vectors; i++) {
+		struct i40e_pf *pf = ldev->pf;
+		struct i40e_qv_info *qv_info;
+		u32 reg_idx;
+
+		qv_info = &qvlist_info->qv_info[i];
+		if (!qv_info)
+			continue;
+		reg_idx = I40E_PFINT_LNKLSTN(qv_info->v_idx - 1);
+		wr32(&pf->hw, reg_idx, I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK);
+	}
+	kfree(ldev->qvlist_info);
+	ldev->qvlist_info = NULL;
+}
+
+/**
+ * i40e_notify_client_of_netdev_close - call the client close callback
+ * @vsi: the VSI with netdev closed
+ * @reset: true when close called due to a reset pending
+ *
+ * If there is a client to this netdev, call the client with close
+ **/
+void i40e_notify_client_of_netdev_close(struct i40e_vsi *vsi, bool reset)
+{
+	struct i40e_client_instance *cdev;
+
+	if (!vsi)
+		return;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.netdev == vsi->netdev) {
+			if (!cdev->client ||
+			    !cdev->client->ops || !cdev->client->ops->close) {
+				dev_dbg(&vsi->back->pdev->dev,
+					"Cannot locate client instance close routine\n");
+				continue;
+			}
+			cdev->client->ops->close(&cdev->lan_info, cdev->client,
+						 reset);
+			i40e_client_release_qvlist(&cdev->lan_info);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_notify_client_of_vf_reset - call the client vf reset callback
+ * @pf: PF device pointer
+ * @vf_id: asolute id of VF being reset
+ *
+ * If there is a client attached to this PF, notify when a VF is reset
+ **/
+void i40e_notify_client_of_vf_reset(struct i40e_pf *pf, u32 vf_id)
+{
+	struct i40e_client_instance *cdev;
+
+	if (!pf)
+		return;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.pf == pf) {
+			if (!cdev->client ||
+			    !cdev->client->ops ||
+			    !cdev->client->ops->vf_reset) {
+				dev_dbg(&pf->pdev->dev,
+					"Cannot locate client instance VF reset routine\n");
+				continue;
+			}
+			cdev->client->ops->vf_reset(&cdev->lan_info,
+						    cdev->client, vf_id);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_notify_client_of_vf_enable - call the client vf notification callback
+ * @pf: PF device pointer
+ * @num_vfs: the number of VFs currently enabled, 0 for disable
+ *
+ * If there is a client attached to this PF, call its VF notification routine
+ **/
+void i40e_notify_client_of_vf_enable(struct i40e_pf *pf, u32 num_vfs)
+{
+	struct i40e_client_instance *cdev;
+
+	if (!pf)
+		return;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.pf == pf) {
+			if (!cdev->client ||
+			    !cdev->client->ops ||
+			    !cdev->client->ops->vf_enable) {
+				dev_dbg(&pf->pdev->dev,
+					"Cannot locate client instance VF enable routine\n");
+				continue;
+			}
+			cdev->client->ops->vf_enable(&cdev->lan_info,
+						     cdev->client, num_vfs);
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+}
+
+/**
+ * i40e_vf_client_capable - ask the client if it likes the specified VF
+ * @pf: PF device pointer
+ * @vf_id: the VF in question
+ *
+ * If there is a client of the specified type attached to this PF, call
+ * its vf_capable routine
+ **/
+int i40e_vf_client_capable(struct i40e_pf *pf, u32 vf_id,
+			   enum i40e_client_type type)
+{
+	struct i40e_client_instance *cdev;
+	int capable = false;
+
+	if (!pf)
+		return false;
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if (cdev->lan_info.pf == pf) {
+			if (!cdev->client ||
+			    !cdev->client->ops ||
+			    !cdev->client->ops->vf_capable ||
+			    !(cdev->client->type == type)) {
+				dev_dbg(&pf->pdev->dev,
+					"Cannot locate client instance VF capability routine\n");
+				continue;
+			}
+			capable = cdev->client->ops->vf_capable(&cdev->lan_info,
+								cdev->client,
+								vf_id);
+			break;
+		}
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+	return capable;
+}
+
+/**
+ * i40e_vsi_lookup - finds a matching VSI from the PF list starting at start_vsi
+ * @pf: board private structure
+ * @type: vsi type
+ * @start_vsi: a VSI pointer from where to start the search
+ *
+ * Returns non NULL on success or NULL for failure
+ **/
+struct i40e_vsi *i40e_vsi_lookup(struct i40e_pf *pf,
+				 enum i40e_vsi_type type,
+				 struct i40e_vsi *start_vsi)
+{
+	struct i40e_vsi *vsi;
+	int i = 0;
+
+	if (start_vsi) {
+		for (i = 0; i < pf->num_alloc_vsi; i++) {
+			vsi = pf->vsi[i];
+			if (vsi == start_vsi)
+				break;
+		}
+	}
+	for (; i < pf->num_alloc_vsi; i++) {
+		vsi = pf->vsi[i];
+		if (vsi && vsi->type == type)
+			return vsi;
+	}
+
+	return NULL;
+}
+
+/**
+ * i40e_client_add_instance - add a client instance struct to the instance list
+ * @pf: pointer to the board struct
+ * @client: pointer to a client struct in the client list.
+ *
+ * Returns cdev ptr on success, NULL on failure
+ **/
+static
+struct i40e_client_instance *i40e_client_add_instance(struct i40e_pf *pf,
+						      struct i40e_client *client)
+{
+	struct i40e_client_instance *cdev;
+	struct netdev_hw_addr *mac = NULL;
+	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
+
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry(cdev, &i40e_client_instances, list) {
+		if ((cdev->lan_info.pf == pf) && (cdev->client == client)) {
+			cdev = NULL;
+			goto out;
+		}
+	}
+	cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
+	if (!cdev)
+		goto out;
+
+	cdev->lan_info.pf = (void *)pf;
+	cdev->lan_info.netdev = vsi->netdev;
+	cdev->lan_info.pcidev = pf->pdev;
+	cdev->lan_info.fid = pf->hw.pf_id;
+	cdev->lan_info.ftype = I40E_CLIENT_FTYPE_PF;
+	cdev->lan_info.hw_addr = pf->hw.hw_addr;
+	cdev->lan_info.ops = &i40e_lan_ops;
+	cdev->lan_info.version.major = I40E_CLIENT_VERSION_MAJOR;
+	cdev->lan_info.version.minor = I40E_CLIENT_VERSION_MINOR;
+	cdev->lan_info.version.build = I40E_CLIENT_VERSION_BUILD;
+	cdev->lan_info.fw_maj_ver = pf->hw.aq.fw_maj_ver;
+	cdev->lan_info.fw_min_ver = pf->hw.aq.fw_min_ver;
+	cdev->lan_info.fw_build = pf->hw.aq.fw_build;
+	set_bit(__I40E_CLIENT_INSTANCE_NONE, &cdev->state);
+
+	if (i40e_client_get_params(vsi, &cdev->lan_info.params)) {
+		kfree(cdev);
+		cdev = NULL;
+		goto out;
+	}
+
+	cdev->lan_info.msix_count = pf->num_iwarp_msix;
+	cdev->lan_info.msix_entries = &pf->msix_entries[pf->iwarp_base_vector];
+
+	mac = list_first_entry(&cdev->lan_info.netdev->dev_addrs.list,
+			       struct netdev_hw_addr, list);
+	if (mac)
+		ether_addr_copy(cdev->lan_info.lanmac, mac->addr);
+	else
+		dev_err(&pf->pdev->dev, "MAC address list is empty!\n");
+
+	cdev->client = client;
+	INIT_LIST_HEAD(&cdev->list);
+	list_add(&cdev->list, &i40e_client_instances);
+out:
+	mutex_unlock(&i40e_client_instance_mutex);
+	return cdev;
+}
+
+/**
+ * i40e_client_del_instance - removes a client instance from the list
+ * @pf: pointer to the board struct
+ *
+ * Returns 0 on success or non-0 on error
+ **/
+static
+int i40e_client_del_instance(struct i40e_pf *pf, struct i40e_client *client)
+{
+	struct i40e_client_instance *cdev, *tmp;
+	int ret = -ENODEV;
+
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry_safe(cdev, tmp, &i40e_client_instances, list) {
+		if ((cdev->lan_info.pf != pf) || (cdev->client != client))
+			continue;
+
+		dev_info(&pf->pdev->dev, "Deleted instance of Client %s, of dev %d bus=0x%02x func=0x%02x)\n",
+			 client->name, pf->hw.pf_id,
+			 pf->hw.bus.device, pf->hw.bus.func);
+		list_del(&cdev->list);
+		kfree(cdev);
+		ret = 0;
+		break;
+	}
+	mutex_unlock(&i40e_client_instance_mutex);
+	return ret;
+}
+
+/**
+ * i40e_client_subtask - client maintenance work
+ * @pf: board private structure
+ **/
+void i40e_client_subtask(struct i40e_pf *pf)
+{
+	struct i40e_client_instance *cdev;
+	struct i40e_client *client;
+	int ret = 0;
+
+	if (!(pf->flags & I40E_FLAG_SERVICE_CLIENT_REQUESTED))
+		return;
+	pf->flags &= ~I40E_FLAG_SERVICE_CLIENT_REQUESTED;
+
+	/* If we're down or resetting, just bail */
+	if (test_bit(__I40E_DOWN, &pf->state) ||
+	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
+		return;
+
+	/* Check client state and instantiate client if client registered */
+	mutex_lock(&i40e_client_mutex);
+	list_for_each_entry(client, &i40e_clients, list) {
+		/* first check client is registered */
+		if (!test_bit(__I40E_CLIENT_REGISTERED, &client->state))
+			continue;
+
+		/* Do we also need the LAN VSI to be up, to create instance */
+		if (!(client->flags & I40E_CLIENT_FLAGS_LAUNCH_ON_PROBE)) {
+			/* check if L2 VSI is up, if not we are not ready */
+			if (test_bit(__I40E_DOWN, &pf->vsi[pf->lan_vsi]->state))
+				continue;
+		}
+
+		/* Add the client instance to the instance list */
+		cdev = i40e_client_add_instance(pf, client);
+		if (!cdev)
+			continue;
+
+		/* Also up the ref_cnt of no. of instances of this client */
+		atomic_inc(&client->ref_cnt);
+		dev_info(&pf->pdev->dev, "Added instance of Client %s to PF%d bus=0x%02x func=0x%02x\n",
+			 client->name, pf->hw.pf_id,
+			 pf->hw.bus.device, pf->hw.bus.func);
+
+		/* Send an Open request to the client */
+		atomic_inc(&cdev->ref_cnt);
+		if (client->ops && client->ops->open)
+			ret = client->ops->open(&cdev->lan_info, client);
+		atomic_dec(&cdev->ref_cnt);
+		if (!ret) {
+			set_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state);
+		} else {
+			/* remove client instance */
+			i40e_client_del_instance(pf, client);
+			atomic_dec(&client->ref_cnt);
+			continue;
+		}
+	}
+	mutex_unlock(&i40e_client_mutex);
+}
+
+/**
+ * i40e_lan_add_device - add a lan device struct to the list of lan devices
+ * @pf: pointer to the board struct
+ *
+ * Returns 0 on success or none 0 on error
+ **/
+int i40e_lan_add_device(struct i40e_pf *pf)
+{
+	struct i40e_device *ldev;
+	int ret = 0;
+
+	mutex_lock(&i40e_device_mutex);
+	list_for_each_entry(ldev, &i40e_devices, list) {
+		if (ldev->pf == pf) {
+			ret = -EEXIST;
+			goto out;
+		}
+	}
+	ldev = kzalloc(sizeof(*ldev), GFP_KERNEL);
+	if (!ldev) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	ldev->pf = pf;
+	INIT_LIST_HEAD(&ldev->list);
+	list_add(&ldev->list, &i40e_devices);
+	dev_info(&pf->pdev->dev, "Added LAN device PF%d bus=0x%02x func=0x%02x\n",
+		 pf->hw.pf_id, pf->hw.bus.device, pf->hw.bus.func);
+
+	/* Since in some cases register may have happened before a device gets
+	 * added, we can schedule a subtask to go initiate the clients.
+	 */
+	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
+	i40e_service_event_schedule(pf);
+
+out:
+	mutex_unlock(&i40e_device_mutex);
+	return ret;
+}
+
+/**
+ * i40e_lan_del_device - removes a lan device from the device list
+ * @pf: pointer to the board struct
+ *
+ * Returns 0 on success or non-0 on error
+ **/
+int i40e_lan_del_device(struct i40e_pf *pf)
+{
+	struct i40e_device *ldev, *tmp;
+	int ret = -ENODEV;
+
+	mutex_lock(&i40e_device_mutex);
+	list_for_each_entry_safe(ldev, tmp, &i40e_devices, list) {
+		if (ldev->pf == pf) {
+			dev_info(&pf->pdev->dev, "Deleted LAN device PF%d bus=0x%02x func=0x%02x\n",
+				 pf->hw.pf_id, pf->hw.bus.device,
+				 pf->hw.bus.func);
+			list_del(&ldev->list);
+			kfree(ldev);
+			ret = 0;
+			break;
+		}
+	}
+
+	mutex_unlock(&i40e_device_mutex);
+	return ret;
+}
+
+/**
+ * i40e_client_release - release client specific resources
+ * @client: pointer to the registered client
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_client_release(struct i40e_client *client)
+{
+	struct i40e_client_instance *cdev, *tmp;
+	struct i40e_pf *pf = NULL;
+	int ret = 0;
+
+	LIST_HEAD(cdevs_tmp);
+
+	mutex_lock(&i40e_client_instance_mutex);
+	list_for_each_entry_safe(cdev, tmp, &i40e_client_instances, list) {
+		if (strncmp(cdev->client->name, client->name,
+			    I40E_CLIENT_STR_LENGTH))
+			continue;
+		if (test_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state)) {
+			if (atomic_read(&cdev->ref_cnt) > 0) {
+				ret = I40E_ERR_NOT_READY;
+				goto out;
+			}
+			pf = (struct i40e_pf *)cdev->lan_info.pf;
+			if (client->ops && client->ops->close)
+				client->ops->close(&cdev->lan_info, client,
+						   false);
+			i40e_client_release_qvlist(&cdev->lan_info);
+			clear_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state);
+
+			dev_warn(&pf->pdev->dev,
+				 "Client %s instance for PF id %d closed\n",
+				 client->name, pf->hw.pf_id);
+		}
+		/* delete the client instance from the list */
+		list_del(&cdev->list);
+		list_add(&cdev->list, &cdevs_tmp);
+		atomic_dec(&client->ref_cnt);
+		dev_info(&pf->pdev->dev, "Deleted client instance of Client %s\n",
+			 client->name);
+	}
+out:
+	mutex_unlock(&i40e_client_instance_mutex);
+
+	/* free the client device and release its vsi */
+	list_for_each_entry_safe(cdev, tmp, &cdevs_tmp, list) {
+		kfree(cdev);
+	}
+	return ret;
+}
+
+/**
+ * i40e_client_prepare - prepare client specific resources
+ * @client: pointer to the registered client
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_client_prepare(struct i40e_client *client)
+{
+	struct i40e_device *ldev;
+	struct i40e_pf *pf;
+	int ret = 0;
+
+	mutex_lock(&i40e_device_mutex);
+	list_for_each_entry(ldev, &i40e_devices, list) {
+		pf = ldev->pf;
+		/* Start the client subtask */
+		pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
+		i40e_service_event_schedule(pf);
+	}
+	mutex_unlock(&i40e_device_mutex);
+	return ret;
+}
+
+/**
+ * i40e_client_virtchnl_send - TBD
+ * @ldev: pointer to L2 context
+ * @client: Client pointer
+ * @vf_id: absolute VF identifier
+ * @msg: message buffer
+ * @len: length of message buffer
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_client_virtchnl_send(struct i40e_info *ldev,
+				     struct i40e_client *client,
+				     u32 vf_id, u8 *msg, u16 len)
+{
+	struct i40e_pf *pf = ldev->pf;
+	struct i40e_hw *hw = &pf->hw;
+	i40e_status err;
+
+	err = i40e_aq_send_msg_to_vf(hw, vf_id, I40E_VIRTCHNL_OP_IWARP,
+				     0, msg, len, NULL);
+	if (err)
+		dev_err(&pf->pdev->dev, "Unable to send iWarp message to VF, error %d, aq status %d\n",
+			err, hw->aq.asq_last_status);
+
+	return err;
+}
+
+/**
+ * i40e_client_setup_qvlist
+ * @ldev: pointer to L2 context.
+ * @client: Client pointer.
+ * @qv_info: queue and vector list
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_client_setup_qvlist(struct i40e_info *ldev,
+				    struct i40e_client *client,
+				    struct i40e_qvlist_info *qvlist_info)
+{
+	struct i40e_pf *pf = ldev->pf;
+	struct i40e_hw *hw = &pf->hw;
+	struct i40e_qv_info *qv_info;
+	u32 v_idx, i, reg_idx, reg;
+	u32 size;
+
+	size = sizeof(struct i40e_qvlist_info) +
+	       (sizeof(struct i40e_qv_info) * (qvlist_info->num_vectors - 1));
+	ldev->qvlist_info = kzalloc(size, GFP_KERNEL);
+	ldev->qvlist_info->num_vectors = qvlist_info->num_vectors;
+
+	for (i = 0; i < qvlist_info->num_vectors; i++) {
+		qv_info = &qvlist_info->qv_info[i];
+		if (!qv_info)
+			continue;
+		v_idx = qv_info->v_idx;
+
+		/* Validate vector id belongs to this client */
+		if ((v_idx >= (pf->iwarp_base_vector + pf->num_iwarp_msix)) ||
+		    (v_idx < pf->iwarp_base_vector))
+			goto err;
+
+		ldev->qvlist_info->qv_info[i] = *qv_info;
+		reg_idx = I40E_PFINT_LNKLSTN(v_idx - 1);
+
+		if (qv_info->ceq_idx == I40E_QUEUE_INVALID_IDX) {
+			/* Special case - No CEQ mapped on this vector */
+			wr32(hw, reg_idx, I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK);
+		} else {
+			reg = (qv_info->ceq_idx &
+			       I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK) |
+			       (I40E_QUEUE_TYPE_PE_CEQ <<
+			       I40E_PFINT_LNKLSTN_FIRSTQ_TYPE_SHIFT);
+			wr32(hw, reg_idx, reg);
+
+			reg = (I40E_PFINT_CEQCTL_CAUSE_ENA_MASK |
+			       (v_idx << I40E_PFINT_CEQCTL_MSIX_INDX_SHIFT) |
+			       (qv_info->itr_idx <<
+				I40E_PFINT_CEQCTL_ITR_INDX_SHIFT) |
+			       (I40E_QUEUE_END_OF_LIST <<
+				I40E_PFINT_CEQCTL_NEXTQ_INDX_SHIFT));
+			wr32(hw, I40E_PFINT_CEQCTL(qv_info->ceq_idx), reg);
+		}
+		if (qv_info->aeq_idx != I40E_QUEUE_INVALID_IDX) {
+			reg = (I40E_PFINT_AEQCTL_CAUSE_ENA_MASK |
+			       (v_idx << I40E_PFINT_AEQCTL_MSIX_INDX_SHIFT) |
+			       (qv_info->itr_idx <<
+				I40E_PFINT_AEQCTL_ITR_INDX_SHIFT));
+
+			wr32(hw, I40E_PFINT_AEQCTL, reg);
+		}
+	}
+
+	return 0;
+err:
+	kfree(ldev->qvlist_info);
+	ldev->qvlist_info = NULL;
+	return -EINVAL;
+}
+
+/**
+ * i40e_client_request_reset
+ * @ldev: pointer to L2 context.
+ * @client: Client pointer.
+ * @level: reset level
+ **/
+static void i40e_client_request_reset(struct i40e_info *ldev,
+				      struct i40e_client *client,
+				      u32 reset_level)
+{
+	struct i40e_pf *pf = ldev->pf;
+
+	switch (reset_level) {
+	case I40E_CLIENT_RESET_LEVEL_PF:
+		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
+		break;
+	case I40E_CLIENT_RESET_LEVEL_CORE:
+		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
+		break;
+	default:
+		dev_warn(&pf->pdev->dev,
+			 "Client %s instance for PF id %d request an unsupported reset: %d.\n",
+			 client->name, pf->hw.pf_id, reset_level);
+		break;
+	}
+
+	i40e_service_event_schedule(pf);
+}
+
+/**
+ * i40e_client_update_vsi_ctxt
+ * @ldev: pointer to L2 context.
+ * @client: Client pointer.
+ * @is_vf: if this for the VF
+ * @vf_id: if is_vf true this carries the vf_id
+ * @flag: Any device level setting that needs to be done for PE
+ * @valid_flag: Bits in this match up and enable changing of flag bits
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
+				       struct i40e_client *client,
+				       bool is_vf, u32 vf_id,
+				       u32 flag, u32 valid_flag)
+{
+	struct i40e_pf *pf = ldev->pf;
+	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
+	struct i40e_vsi_context ctxt;
+	bool update = true;
+	i40e_status err;
+
+	/* TODO: for now do not allow setting VF's VSI setting */
+	if (is_vf)
+		return -EINVAL;
+
+	ctxt.seid = pf->main_vsi_seid;
+	ctxt.pf_num = pf->hw.pf_id;
+	err = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
+	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
+	if (err) {
+		dev_info(&pf->pdev->dev,
+			 "couldn't get PF vsi config, err %s aq_err %s\n",
+			 i40e_stat_str(&pf->hw, err),
+			 i40e_aq_str(&pf->hw,
+				     pf->hw.aq.asq_last_status));
+		return -ENOENT;
+	}
+
+	if ((valid_flag & I40E_CLIENT_VSI_FLAG_TCP_PACKET_ENABLE) &&
+	    (flag & I40E_CLIENT_VSI_FLAG_TCP_PACKET_ENABLE)) {
+		ctxt.info.valid_sections =
+			cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
+		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
+	} else if ((valid_flag & I40E_CLIENT_VSI_FLAG_TCP_PACKET_ENABLE) &&
+		  !(flag & I40E_CLIENT_VSI_FLAG_TCP_PACKET_ENABLE)) {
+		ctxt.info.valid_sections =
+			cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
+		ctxt.info.queueing_opt_flags &= ~I40E_AQ_VSI_QUE_OPT_TCP_ENA;
+	} else {
+		update = false;
+		dev_warn(&pf->pdev->dev,
+			 "Client %s instance for PF id %d request an unsupported Config: %x.\n",
+			 client->name, pf->hw.pf_id, flag);
+	}
+
+	if (update) {
+		err = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
+		if (err) {
+			dev_info(&pf->pdev->dev,
+				 "update VSI ctxt for PE failed, err %s aq_err %s\n",
+				 i40e_stat_str(&pf->hw, err),
+				 i40e_aq_str(&pf->hw,
+					     pf->hw.aq.asq_last_status));
+		}
+	}
+	return err;
+}
+
+/**
+ * i40e_register_client - Register a i40e client driver with the L2 driver
+ * @client: pointer to the i40e_client struct
+ *
+ * Returns 0 on success or non-0 on error
+ **/
+int i40e_register_client(struct i40e_client *client)
+{
+	int ret = 0;
+	enum i40e_vsi_type vsi_type;
+
+	if (!client) {
+		ret = -EIO;
+		goto out;
+	}
+
+	if (strlen(client->name) == 0) {
+		pr_info("i40e: Failed to register client with no name\n");
+		ret = -EIO;
+		goto out;
+	}
+
+	mutex_lock(&i40e_client_mutex);
+	if (i40e_client_is_registered(client)) {
+		pr_info("i40e: Client %s has already been registered!\n",
+			client->name);
+		mutex_unlock(&i40e_client_mutex);
+		ret = -EEXIST;
+		goto out;
+	}
+
+	if ((client->version.major != I40E_CLIENT_VERSION_MAJOR) ||
+	    (client->version.minor != I40E_CLIENT_VERSION_MINOR)) {
+		pr_info("i40e: Failed to register client %s due to mismatched client interface version\n",
+			client->name);
+		pr_info("Client is using version: %02d.%02d.%02d while LAN driver supports %s\n",
+			client->version.major, client->version.minor,
+			client->version.build,
+			i40e_client_interface_version_str);
+		mutex_unlock(&i40e_client_mutex);
+		ret = -EIO;
+		goto out;
+	}
+
+	vsi_type = i40e_client_type_to_vsi_type(client->type);
+	if (vsi_type == I40E_VSI_TYPE_UNKNOWN) {
+		pr_info("i40e: Failed to register client %s due to unknown client type %d\n",
+			client->name, client->type);
+		mutex_unlock(&i40e_client_mutex);
+		ret = -EIO;
+		goto out;
+	}
+	list_add(&client->list, &i40e_clients);
+	set_bit(__I40E_CLIENT_REGISTERED, &client->state);
+	mutex_unlock(&i40e_client_mutex);
+
+	if (i40e_client_prepare(client)) {
+		ret = -EIO;
+		goto out;
+	}
+
+	pr_info("i40e: Registered client %s with return code %d\n",
+		client->name, ret);
+out:
+	return ret;
+}
+EXPORT_SYMBOL(i40e_register_client);
+
+/**
+ * i40e_unregister_client - Unregister a i40e client driver with the L2 driver
+ * @client: pointer to the i40e_client struct
+ *
+ * Returns 0 on success or non-0 on error
+ **/
+int i40e_unregister_client(struct i40e_client *client)
+{
+	int ret = 0;
+
+	/* When a unregister request comes through we would have to send
+	 * a close for each of the client instances that were opened.
+	 * client_release function is called to handle this.
+	 */
+	if (!client || i40e_client_release(client)) {
+		ret = -EIO;
+		goto out;
+	}
+
+	/* TODO: check if device is in reset, or if that matters? */
+	mutex_lock(&i40e_client_mutex);
+	if (!i40e_client_is_registered(client)) {
+		pr_info("i40e: Client %s has not been registered\n",
+			client->name);
+		mutex_unlock(&i40e_client_mutex);
+		ret = -ENODEV;
+		goto out;
+	}
+	if (atomic_read(&client->ref_cnt) == 0) {
+		clear_bit(__I40E_CLIENT_REGISTERED, &client->state);
+		list_del(&client->list);
+		pr_info("i40e: Unregistered client %s with return code %d\n",
+			client->name, ret);
+	} else {
+		ret = I40E_ERR_NOT_READY;
+		pr_err("i40e: Client %s failed unregister - client has open instances\n",
+		       client->name);
+	}
+
+	mutex_unlock(&i40e_client_mutex);
+out:
+	return ret;
+}
+EXPORT_SYMBOL(i40e_unregister_client);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.h b/drivers/net/ethernet/intel/i40e/i40e_client.h
new file mode 100644
index 000000000000..bf6b453d93a1
--- /dev/null
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.h
@@ -0,0 +1,232 @@
+/*******************************************************************************
+ *
+ * Intel Ethernet Controller XL710 Family Linux Driver
+ * Copyright(c) 2013 - 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * The full GNU General Public License is included in this distribution in
+ * the file called "COPYING".
+ *
+ * Contact Information:
+ * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
+ * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
+ *
+ ******************************************************************************/
+
+#ifndef _I40E_CLIENT_H_
+#define _I40E_CLIENT_H_
+
+#define I40E_CLIENT_STR_LENGTH 10
+
+/* Client interface version should be updated anytime there is a change in the
+ * existing APIs or data structures.
+ */
+#define I40E_CLIENT_VERSION_MAJOR 0
+#define I40E_CLIENT_VERSION_MINOR 01
+#define I40E_CLIENT_VERSION_BUILD 00
+#define I40E_CLIENT_VERSION_STR     \
+	XSTRINGIFY(I40E_CLIENT_VERSION_MAJOR) "." \
+	XSTRINGIFY(I40E_CLIENT_VERSION_MINOR) "." \
+	XSTRINGIFY(I40E_CLIENT_VERSION_BUILD)
+
+struct i40e_client_version {
+	u8 major;
+	u8 minor;
+	u8 build;
+	u8 rsvd;
+};
+
+enum i40e_client_state {
+	__I40E_CLIENT_NULL,
+	__I40E_CLIENT_REGISTERED
+};
+
+enum i40e_client_instance_state {
+	__I40E_CLIENT_INSTANCE_NONE,
+	__I40E_CLIENT_INSTANCE_OPENED,
+};
+
+enum i40e_client_type {
+	I40E_CLIENT_IWARP,
+	I40E_CLIENT_VMDQ2
+};
+
+struct i40e_ops;
+struct i40e_client;
+
+/* HW does not define a type value for AEQ; only for RX/TX and CEQ.
+ * In order for us to keep the interface simple, SW will define a
+ * unique type value for AEQ.
+ */
+#define I40E_QUEUE_TYPE_PE_AEQ  0x80
+#define I40E_QUEUE_INVALID_IDX	0xFFFF
+
+struct i40e_qv_info {
+	u32 v_idx; /* msix_vector */
+	u16 ceq_idx;
+	u16 aeq_idx;
+	u8 itr_idx;
+};
+
+struct i40e_qvlist_info {
+	u32 num_vectors;
+	struct i40e_qv_info qv_info[1];
+};
+
+#define I40E_CLIENT_MSIX_ALL 0xFFFFFFFF
+
+/* set of LAN parameters useful for clients managed by LAN */
+
+/* Struct to hold per priority info */
+struct i40e_prio_qos_params {
+	u16 qs_handle; /* qs handle for prio */
+	u8 tc; /* TC mapped to prio */
+	u8 reserved;
+};
+
+#define I40E_CLIENT_MAX_USER_PRIORITY        8
+/* Struct to hold Client QoS */
+struct i40e_qos_params {
+	struct i40e_prio_qos_params prio_qos[I40E_CLIENT_MAX_USER_PRIORITY];
+};
+
+struct i40e_params {
+	struct i40e_qos_params qos;
+	u16 mtu;
+};
+
+/* Structure to hold Lan device info for a client device */
+struct i40e_info {
+	struct i40e_client_version version;
+	u8 lanmac[6];
+	struct net_device *netdev;
+	struct pci_dev *pcidev;
+	u8 __iomem *hw_addr;
+	u8 fid;	/* function id, PF id or VF id */
+#define I40E_CLIENT_FTYPE_PF 0
+#define I40E_CLIENT_FTYPE_VF 1
+	u8 ftype; /* function type, PF or VF */
+	void *pf;
+
+	/* All L2 params that could change during the life span of the PF
+	 * and needs to be communicated to the client when they change
+	 */
+	struct i40e_qvlist_info *qvlist_info;
+	struct i40e_params params;
+	struct i40e_ops *ops;
+
+	u16 msix_count;	 /* number of msix vectors*/
+	/* Array down below will be dynamically allocated based on msix_count */
+	struct msix_entry *msix_entries;
+	u16 itr_index; /* Which ITR index the PE driver is suppose to use */
+	u16 fw_maj_ver;                 /* firmware major version */
+	u16 fw_min_ver;                 /* firmware minor version */
+	u32 fw_build;                   /* firmware build number */
+};
+
+#define I40E_CLIENT_RESET_LEVEL_PF   1
+#define I40E_CLIENT_RESET_LEVEL_CORE 2
+#define I40E_CLIENT_VSI_FLAG_TCP_PACKET_ENABLE  BIT(1)
+
+struct i40e_ops {
+	/* setup_q_vector_list enables queues with a particular vector */
+	int (*setup_qvlist)(struct i40e_info *ldev, struct i40e_client *client,
+			    struct i40e_qvlist_info *qv_info);
+
+	int (*virtchnl_send)(struct i40e_info *ldev, struct i40e_client *client,
+			     u32 vf_id, u8 *msg, u16 len);
+
+	/* If the PE Engine is unresponsive, RDMA driver can request a reset.
+	 * The level helps determine the level of reset being requested.
+	 */
+	void (*request_reset)(struct i40e_info *ldev,
+			      struct i40e_client *client, u32 level);
+
+	/* API for the RDMA driver to set certain VSI flags that control
+	 * PE Engine.
+	 */
+	int (*update_vsi_ctxt)(struct i40e_info *ldev,
+			       struct i40e_client *client,
+			       bool is_vf, u32 vf_id,
+			       u32 flag, u32 valid_flag);
+};
+
+struct i40e_client_ops {
+	/* Should be called from register_client() or whenever PF is ready
+	 * to create a specific client instance.
+	 */
+	int (*open)(struct i40e_info *ldev, struct i40e_client *client);
+
+	/* Should be called when netdev is unavailable or when unregister
+	 * call comes in. If the close is happenening due to a reset being
+	 * triggered set the reset bit to true.
+	 */
+	void (*close)(struct i40e_info *ldev, struct i40e_client *client,
+		      bool reset);
+
+	/* called when some l2 managed parameters changes - mtu */
+	void (*l2_param_change)(struct i40e_info *ldev,
+				struct i40e_client *client,
+				struct i40e_params *params);
+
+	int (*virtchnl_receive)(struct i40e_info *ldev,
+				struct i40e_client *client, u32 vf_id,
+				u8 *msg, u16 len);
+
+	/* called when a VF is reset by the PF */
+	void (*vf_reset)(struct i40e_info *ldev,
+			 struct i40e_client *client, u32 vf_id);
+
+	/* called when the number of VFs changes */
+	void (*vf_enable)(struct i40e_info *ldev,
+			  struct i40e_client *client, u32 num_vfs);
+
+	/* returns true if VF is capable of specified offload */
+	int (*vf_capable)(struct i40e_info *ldev,
+			  struct i40e_client *client, u32 vf_id);
+};
+
+/* Client device */
+struct i40e_client_instance {
+	struct list_head list;
+	struct i40e_info lan_info;
+	struct i40e_client *client;
+	unsigned long  state;
+	/* A count of all the in-progress calls to the client */
+	atomic_t ref_cnt;
+};
+
+struct i40e_client {
+	struct list_head list;		/* list of registered clients */
+	char name[I40E_CLIENT_STR_LENGTH];
+	struct i40e_client_version version;
+	unsigned long state;		/* client state */
+	atomic_t ref_cnt;  /* Count of all the client devices of this kind */
+	u32 flags;
+#define I40E_CLIENT_FLAGS_LAUNCH_ON_PROBE	BIT(0)
+#define I40E_TX_FLAGS_NOTIFY_OTHER_EVENTS	BIT(2)
+	enum i40e_client_type type;
+	struct i40e_client_ops *ops;	/* client ops provided by the client */
+};
+
+static inline bool i40e_client_is_registered(struct i40e_client *client)
+{
+	return test_bit(__I40E_CLIENT_REGISTERED, &client->state);
+}
+
+/* used by clients */
+int i40e_register_client(struct i40e_client *client);
+int i40e_unregister_client(struct i40e_client *client);
+
+#endif /* _I40E_CLIENT_H_ */
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 70d9605a0d9e..67006431726a 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -289,7 +289,7 @@ struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
  *
  * If not already scheduled, this puts the task into the work queue
  **/
-static void i40e_service_event_schedule(struct i40e_pf *pf)
+void i40e_service_event_schedule(struct i40e_pf *pf)
 {
 	if (!test_bit(__I40E_DOWN, &pf->state) &&
 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
@@ -2230,7 +2230,7 @@ static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
 	netdev->mtu = new_mtu;
 	if (netif_running(netdev))
 		i40e_vsi_reinit_locked(vsi);
-
+	i40e_notify_client_of_l2_param_changes(vsi);
 	return 0;
 }
 
@@ -4169,6 +4169,9 @@ static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
 		free_irq(pf->msix_entries[0].vector, pf);
 	}
 
+	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
+		      I40E_IWARP_IRQ_PILE_ID);
+
 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
 	for (i = 0; i < pf->num_alloc_vsi; i++)
 		if (pf->vsi[i])
@@ -4212,12 +4215,17 @@ static void i40e_napi_disable_all(struct i40e_vsi *vsi)
  **/
 static void i40e_vsi_close(struct i40e_vsi *vsi)
 {
+	bool reset = false;
+
 	if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
 		i40e_down(vsi);
 	i40e_vsi_free_irq(vsi);
 	i40e_vsi_free_tx_resources(vsi);
 	i40e_vsi_free_rx_resources(vsi);
 	vsi->current_netdev_flags = 0;
+	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
+		reset = true;
+	i40e_notify_client_of_netdev_close(vsi, reset);
 }
 
 /**
@@ -4850,6 +4858,12 @@ static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
 	ctxt.info = vsi->info;
 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
 
+	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
+		ctxt.info.valid_sections |=
+				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
+		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
+	}
+
 	/* Update the VSI after updating the VSI queue-mapping information */
 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
 	if (ret) {
@@ -4993,6 +5007,7 @@ static void i40e_dcb_reconfigure(struct i40e_pf *pf)
 			if (pf->vsi[v]->netdev)
 				i40e_dcbnl_set_all(pf->vsi[v]);
 		}
+		i40e_notify_client_of_l2_param_changes(pf->vsi[v]);
 	}
 }
 
@@ -5191,6 +5206,11 @@ static int i40e_up_complete(struct i40e_vsi *vsi)
 		}
 		i40e_fdir_filter_restore(vsi);
 	}
+
+	/* On the next run of the service_task, notify any clients of the new
+	 * opened netdev
+	 */
+	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
 	i40e_service_event_schedule(pf);
 
 	return 0;
@@ -5379,6 +5399,8 @@ int i40e_open(struct net_device *netdev)
 		geneve_get_rx_port(netdev);
 #endif
 
+	i40e_notify_client_of_netdev_open(vsi);
+
 	return 0;
 }
 
@@ -6043,6 +6065,7 @@ static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
 	case I40E_VSI_SRIOV:
 	case I40E_VSI_VMDQ2:
 	case I40E_VSI_CTRL:
+	case I40E_VSI_IWARP:
 	case I40E_VSI_MIRROR:
 	default:
 		/* there is no notification for other VSIs */
@@ -7148,6 +7171,7 @@ static void i40e_service_task(struct work_struct *work)
 	i40e_vc_process_vflr_event(pf);
 	i40e_watchdog_subtask(pf);
 	i40e_fdir_reinit_subtask(pf);
+	i40e_client_subtask(pf);
 	i40e_sync_filters_subtask(pf);
 	i40e_sync_udp_filters_subtask(pf);
 	i40e_clean_adminq_subtask(pf);
@@ -7550,6 +7574,7 @@ static int i40e_init_msix(struct i40e_pf *pf)
 	int vectors_left;
 	int v_budget, i;
 	int v_actual;
+	int iwarp_requested = 0;
 
 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
 		return -ENODEV;
@@ -7563,6 +7588,7 @@ static int i40e_init_msix(struct i40e_pf *pf)
 	 *		is governed by number of cpus in the system.
 	 *	- assumes symmetric Tx/Rx pairing
 	 *   - The number of VMDq pairs
+	 *   - The CPU count within the NUMA node if iWARP is enabled
 #ifdef I40E_FCOE
 	 *   - The number of FCOE qps.
 #endif
@@ -7609,6 +7635,16 @@ static int i40e_init_msix(struct i40e_pf *pf)
 	}
 
 #endif
+	/* can we reserve enough for iWARP? */
+	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
+		if (!vectors_left)
+			pf->num_iwarp_msix = 0;
+		else if (vectors_left < pf->num_iwarp_msix)
+			pf->num_iwarp_msix = 1;
+		v_budget += pf->num_iwarp_msix;
+		vectors_left -= pf->num_iwarp_msix;
+	}
+
 	/* any vectors left over go for VMDq support */
 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
@@ -7643,6 +7679,8 @@ static int i40e_init_msix(struct i40e_pf *pf)
 		 * of these features based on the policy and at the end disable
 		 * the features that did not get any vectors.
 		 */
+		iwarp_requested = pf->num_iwarp_msix;
+		pf->num_iwarp_msix = 0;
 #ifdef I40E_FCOE
 		pf->num_fcoe_qps = 0;
 		pf->num_fcoe_msix = 0;
@@ -7681,17 +7719,33 @@ static int i40e_init_msix(struct i40e_pf *pf)
 			pf->num_lan_msix = 1;
 			break;
 		case 3:
+			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
+				pf->num_lan_msix = 1;
+				pf->num_iwarp_msix = 1;
+			} else {
+				pf->num_lan_msix = 2;
+			}
 #ifdef I40E_FCOE
 			/* give one vector to FCoE */
 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
 				pf->num_lan_msix = 1;
 				pf->num_fcoe_msix = 1;
 			}
-#else
-			pf->num_lan_msix = 2;
 #endif
 			break;
 		default:
+			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
+				pf->num_iwarp_msix = min_t(int, (vec / 3),
+						 iwarp_requested);
+				pf->num_vmdq_vsis = min_t(int, (vec / 3),
+						  I40E_DEFAULT_NUM_VMDQ_VSI);
+			} else {
+				pf->num_vmdq_vsis = min_t(int, (vec / 2),
+						  I40E_DEFAULT_NUM_VMDQ_VSI);
+			}
+			pf->num_lan_msix = min_t(int,
+			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
+							      pf->num_lan_msix);
 #ifdef I40E_FCOE
 			/* give one vector to FCoE */
 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
@@ -7699,8 +7753,6 @@ static int i40e_init_msix(struct i40e_pf *pf)
 				vec--;
 			}
 #endif
-			/* give the rest to the PF */
-			pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
 			break;
 		}
 	}
@@ -7710,6 +7762,12 @@ static int i40e_init_msix(struct i40e_pf *pf)
 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
 	}
+
+	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
+	    (pf->num_iwarp_msix == 0)) {
+		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
+		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
+	}
 #ifdef I40E_FCOE
 
 	if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
@@ -7801,6 +7859,7 @@ static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
 		vectors = i40e_init_msix(pf);
 		if (vectors < 0) {
 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
+				       I40E_FLAG_IWARP_ENABLED	|
 #ifdef I40E_FCOE
 				       I40E_FLAG_FCOE_ENABLED	|
 #endif
@@ -8474,6 +8533,12 @@ static int i40e_sw_init(struct i40e_pf *pf)
 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
 	}
 
+	if (pf->hw.func_caps.iwarp) {
+		pf->flags |= I40E_FLAG_IWARP_ENABLED;
+		/* IWARP needs one extra vector for CQP just like MISC.*/
+		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
+	}
+
 #ifdef I40E_FCOE
 	i40e_init_pf_fcoe(pf);
 
@@ -9328,6 +9393,13 @@ static int i40e_add_vsi(struct i40e_vsi *vsi)
 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
 		}
 
+		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
+			ctxt.info.valid_sections |=
+				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
+			ctxt.info.queueing_opt_flags |=
+						I40E_AQ_VSI_QUE_OPT_TCP_ENA;
+		}
+
 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
 		if (pf->vf[vsi->vf_id].spoofchk) {
@@ -9351,6 +9423,10 @@ static int i40e_add_vsi(struct i40e_vsi *vsi)
 		break;
 
 #endif /* I40E_FCOE */
+	case I40E_VSI_IWARP:
+		/* send down message to iWARP */
+		break;
+
 	default:
 		return -ENODEV;
 	}
@@ -10467,6 +10543,7 @@ static void i40e_determine_queue_usage(struct i40e_pf *pf)
 
 		/* make sure all the fancies are disabled */
 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
+			       I40E_FLAG_IWARP_ENABLED	|
 #ifdef I40E_FCOE
 			       I40E_FLAG_FCOE_ENABLED	|
 #endif
@@ -10484,6 +10561,7 @@ static void i40e_determine_queue_usage(struct i40e_pf *pf)
 		queues_left -= pf->num_lan_qps;
 
 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
+			       I40E_FLAG_IWARP_ENABLED	|
 #ifdef I40E_FCOE
 			       I40E_FLAG_FCOE_ENABLED	|
 #endif
@@ -11059,7 +11137,17 @@ static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	}
 #endif /* CONFIG_PCI_IOV */
 
-	pfs_found++;
+	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
+		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
+						      pf->num_iwarp_msix,
+						      I40E_IWARP_IRQ_PILE_ID);
+		if (pf->iwarp_base_vector < 0) {
+			dev_info(&pdev->dev,
+				 "failed to get tracking for %d vectors for IWARP err=%d\n",
+				 pf->num_iwarp_msix, pf->iwarp_base_vector);
+			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
+		}
+	}
 
 	i40e_dbg_pf_init(pf);
 
@@ -11070,6 +11158,12 @@ static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	mod_timer(&pf->service_timer,
 		  round_jiffies(jiffies + pf->service_timer_period));
 
+	/* add this PF to client device list and launch a client service task */
+	err = i40e_lan_add_device(pf);
+	if (err)
+		dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
+			 err);
+
 #ifdef I40E_FCOE
 	/* create FCoE interface */
 	i40e_fcoe_vsi_setup(pf);
@@ -11245,6 +11339,13 @@ static void i40e_remove(struct pci_dev *pdev)
 	if (pf->vsi[pf->lan_vsi])
 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
 
+	/* remove attached clients */
+	ret_code = i40e_lan_del_device(pf);
+	if (ret_code) {
+		dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
+			 ret_code);
+	}
+
 	/* shutdown and destroy the HMC */
 	if (hw->hmc.hmc_obj) {
 		ret_code = i40e_shutdown_lan_hmc(hw);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_type.h b/drivers/net/ethernet/intel/i40e/i40e_type.h
index 0a0baf71041b..3335f9d13374 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_type.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_type.h
@@ -78,7 +78,7 @@ enum i40e_debug_mask {
 	I40E_DEBUG_DCB			= 0x00000400,
 	I40E_DEBUG_DIAG			= 0x00000800,
 	I40E_DEBUG_FD			= 0x00001000,
-
+	I40E_DEBUG_IWARP		= 0x00F00000,
 	I40E_DEBUG_AQ_MESSAGE		= 0x01000000,
 	I40E_DEBUG_AQ_DESCRIPTOR	= 0x02000000,
 	I40E_DEBUG_AQ_DESC_BUFFER	= 0x04000000,
@@ -160,6 +160,7 @@ enum i40e_vsi_type {
 	I40E_VSI_MIRROR	= 5,
 	I40E_VSI_SRIOV	= 6,
 	I40E_VSI_FDIR	= 7,
+	I40E_VSI_IWARP	= 8,
 	I40E_VSI_TYPE_UNKNOWN
 };
 
diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl.h b/drivers/net/ethernet/intel/i40e/i40e_virtchnl.h
index 3226946bf3d4..ab866cf3dc18 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl.h
@@ -81,6 +81,9 @@ enum i40e_virtchnl_ops {
 	I40E_VIRTCHNL_OP_GET_STATS = 15,
 	I40E_VIRTCHNL_OP_FCOE = 16,
 	I40E_VIRTCHNL_OP_EVENT = 17,
+	I40E_VIRTCHNL_OP_IWARP = 20,
+	I40E_VIRTCHNL_OP_CONFIG_IWARP_IRQ_MAP = 21,
+	I40E_VIRTCHNL_OP_RELEASE_IWARP_IRQ_MAP = 22,
 };
 
 /* Virtual channel message descriptor. This overlays the admin queue
@@ -348,6 +351,37 @@ struct i40e_virtchnl_pf_event {
 	int severity;
 };
 
+/* I40E_VIRTCHNL_OP_CONFIG_IWARP_IRQ_MAP
+ * VF uses this message to request PF to map IWARP vectors to IWARP queues.
+ * The request for this originates from the VF IWARP driver through
+ * a client interface between VF LAN and VF IWARP driver.
+ * A vector could have an AEQ and CEQ attached to it although
+ * there is a single AEQ per VF IWARP instance in which case
+ * most vectors will have an INVALID_IDX for aeq and valid idx for ceq.
+ * There will never be a case where there will be multiple CEQs attached
+ * to a single vector.
+ * PF configures interrupt mapping and returns status.
+ */
+
+/* HW does not define a type value for AEQ; only for RX/TX and CEQ.
+ * In order for us to keep the interface simple, SW will define a
+ * unique type value for AEQ.
+*/
+#define I40E_QUEUE_TYPE_PE_AEQ  0x80
+#define I40E_QUEUE_INVALID_IDX  0xFFFF
+
+struct i40e_virtchnl_iwarp_qv_info {
+	u32 v_idx; /* msix_vector */
+	u16 ceq_idx;
+	u16 aeq_idx;
+	u8 itr_idx;
+};
+
+struct i40e_virtchnl_iwarp_qvlist_info {
+	u32 num_vectors;
+	struct i40e_virtchnl_iwarp_qv_info qv_info[1];
+};
+
 /* VF reset states - these are written into the RSTAT register:
  * I40E_VFGEN_RSTAT1 on the PF
  * I40E_VFGEN_RSTAT on the VF
diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
index acd2693a4e97..816c6bbf7093 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
@@ -352,6 +352,136 @@ irq_list_done:
 }
 
 /**
+ * i40e_release_iwarp_qvlist
+ * @vf: pointer to the VF.
+ *
+ **/
+static void i40e_release_iwarp_qvlist(struct i40e_vf *vf)
+{
+	struct i40e_pf *pf = vf->pf;
+	struct i40e_virtchnl_iwarp_qvlist_info *qvlist_info = vf->qvlist_info;
+	u32 msix_vf;
+	u32 i;
+
+	if (!vf->qvlist_info)
+		return;
+
+	msix_vf = pf->hw.func_caps.num_msix_vectors_vf;
+	for (i = 0; i < qvlist_info->num_vectors; i++) {
+		struct i40e_virtchnl_iwarp_qv_info *qv_info;
+		u32 next_q_index, next_q_type;
+		struct i40e_hw *hw = &pf->hw;
+		u32 v_idx, reg_idx, reg;
+
+		qv_info = &qvlist_info->qv_info[i];
+		if (!qv_info)
+			continue;
+		v_idx = qv_info->v_idx;
+		if (qv_info->ceq_idx != I40E_QUEUE_INVALID_IDX) {
+			/* Figure out the queue after CEQ and make that the
+			 * first queue.
+			 */
+			reg_idx = (msix_vf - 1) * vf->vf_id + qv_info->ceq_idx;
+			reg = rd32(hw, I40E_VPINT_CEQCTL(reg_idx));
+			next_q_index = (reg & I40E_VPINT_CEQCTL_NEXTQ_INDX_MASK)
+					>> I40E_VPINT_CEQCTL_NEXTQ_INDX_SHIFT;
+			next_q_type = (reg & I40E_VPINT_CEQCTL_NEXTQ_TYPE_MASK)
+					>> I40E_VPINT_CEQCTL_NEXTQ_TYPE_SHIFT;
+
+			reg_idx = ((msix_vf - 1) * vf->vf_id) + (v_idx - 1);
+			reg = (next_q_index &
+			       I40E_VPINT_LNKLSTN_FIRSTQ_INDX_MASK) |
+			       (next_q_type <<
+			       I40E_VPINT_LNKLSTN_FIRSTQ_TYPE_SHIFT);
+
+			wr32(hw, I40E_VPINT_LNKLSTN(reg_idx), reg);
+		}
+	}
+	kfree(vf->qvlist_info);
+	vf->qvlist_info = NULL;
+}
+
+/**
+ * i40e_config_iwarp_qvlist
+ * @vf: pointer to the VF info
+ * @qvlist_info: queue and vector list
+ *
+ * Return 0 on success or < 0 on error
+ **/
+static int i40e_config_iwarp_qvlist(struct i40e_vf *vf,
+				    struct i40e_virtchnl_iwarp_qvlist_info *qvlist_info)
+{
+	struct i40e_pf *pf = vf->pf;
+	struct i40e_hw *hw = &pf->hw;
+	struct i40e_virtchnl_iwarp_qv_info *qv_info;
+	u32 v_idx, i, reg_idx, reg;
+	u32 next_q_idx, next_q_type;
+	u32 msix_vf, size;
+
+	size = sizeof(struct i40e_virtchnl_iwarp_qvlist_info) +
+	       (sizeof(struct i40e_virtchnl_iwarp_qv_info) *
+						(qvlist_info->num_vectors - 1));
+	vf->qvlist_info = kzalloc(size, GFP_KERNEL);
+	vf->qvlist_info->num_vectors = qvlist_info->num_vectors;
+
+	msix_vf = pf->hw.func_caps.num_msix_vectors_vf;
+	for (i = 0; i < qvlist_info->num_vectors; i++) {
+		qv_info = &qvlist_info->qv_info[i];
+		if (!qv_info)
+			continue;
+		v_idx = qv_info->v_idx;
+
+		/* Validate vector id belongs to this vf */
+		if (!i40e_vc_isvalid_vector_id(vf, v_idx))
+			goto err;
+
+		vf->qvlist_info->qv_info[i] = *qv_info;
+
+		reg_idx = ((msix_vf - 1) * vf->vf_id) + (v_idx - 1);
+		/* We might be sharing the interrupt, so get the first queue
+		 * index and type, push it down the list by adding the new
+		 * queue on top. Also link it with the new queue in CEQCTL.
+		 */
+		reg = rd32(hw, I40E_VPINT_LNKLSTN(reg_idx));
+		next_q_idx = ((reg & I40E_VPINT_LNKLSTN_FIRSTQ_INDX_MASK) >>
+				I40E_VPINT_LNKLSTN_FIRSTQ_INDX_SHIFT);
+		next_q_type = ((reg & I40E_VPINT_LNKLSTN_FIRSTQ_TYPE_MASK) >>
+				I40E_VPINT_LNKLSTN_FIRSTQ_TYPE_SHIFT);
+
+		if (qv_info->ceq_idx != I40E_QUEUE_INVALID_IDX) {
+			reg_idx = (msix_vf - 1) * vf->vf_id + qv_info->ceq_idx;
+			reg = (I40E_VPINT_CEQCTL_CAUSE_ENA_MASK |
+			(v_idx << I40E_VPINT_CEQCTL_MSIX_INDX_SHIFT) |
+			(qv_info->itr_idx << I40E_VPINT_CEQCTL_ITR_INDX_SHIFT) |
+			(next_q_type << I40E_VPINT_CEQCTL_NEXTQ_TYPE_SHIFT) |
+			(next_q_idx << I40E_VPINT_CEQCTL_NEXTQ_INDX_SHIFT));
+			wr32(hw, I40E_VPINT_CEQCTL(reg_idx), reg);
+
+			reg_idx = ((msix_vf - 1) * vf->vf_id) + (v_idx - 1);
+			reg = (qv_info->ceq_idx &
+			       I40E_VPINT_LNKLSTN_FIRSTQ_INDX_MASK) |
+			       (I40E_QUEUE_TYPE_PE_CEQ <<
+			       I40E_VPINT_LNKLSTN_FIRSTQ_TYPE_SHIFT);
+			wr32(hw, I40E_VPINT_LNKLSTN(reg_idx), reg);
+		}
+
+		if (qv_info->aeq_idx != I40E_QUEUE_INVALID_IDX) {
+			reg = (I40E_VPINT_AEQCTL_CAUSE_ENA_MASK |
+			(v_idx << I40E_VPINT_AEQCTL_MSIX_INDX_SHIFT) |
+			(qv_info->itr_idx << I40E_VPINT_AEQCTL_ITR_INDX_SHIFT));
+
+			wr32(hw, I40E_VPINT_AEQCTL(vf->vf_id), reg);
+		}
+	}
+
+	return 0;
+err:
+	kfree(vf->qvlist_info);
+	vf->qvlist_info = NULL;
+	return -EINVAL;
+}
+
+/**
  * i40e_config_vsi_tx_queue
  * @vf: pointer to the VF info
  * @vsi_id: id of VSI as provided by the FW
@@ -850,9 +980,11 @@ complete_reset:
 	/* reallocate VF resources to reset the VSI state */
 	i40e_free_vf_res(vf);
 	if (!i40e_alloc_vf_res(vf)) {
+		int abs_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
 		i40e_enable_vf_mappings(vf);
 		set_bit(I40E_VF_STAT_ACTIVE, &vf->vf_states);
 		clear_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
+		i40e_notify_client_of_vf_reset(pf, abs_vf_id);
 	}
 	/* tell the VF the reset is done */
 	wr32(hw, I40E_VFGEN_RSTAT1(vf->vf_id), I40E_VFR_VFACTIVE);
@@ -877,11 +1009,7 @@ void i40e_free_vfs(struct i40e_pf *pf)
 	while (test_and_set_bit(__I40E_VF_DISABLE, &pf->state))
 		usleep_range(1000, 2000);
 
-	for (i = 0; i < pf->num_alloc_vfs; i++)
-		if (test_bit(I40E_VF_STAT_INIT, &pf->vf[i].vf_states))
-			i40e_vsi_control_rings(pf->vsi[pf->vf[i].lan_vsi_idx],
-					       false);
-
+	i40e_notify_client_of_vf_enable(pf, 0);
 	for (i = 0; i < pf->num_alloc_vfs; i++)
 		if (test_bit(I40E_VF_STAT_INIT, &pf->vf[i].vf_states))
 			i40e_vsi_control_rings(pf->vsi[pf->vf[i].lan_vsi_idx],
@@ -953,6 +1081,7 @@ int i40e_alloc_vfs(struct i40e_pf *pf, u16 num_alloc_vfs)
 			goto err_iov;
 		}
 	}
+	i40e_notify_client_of_vf_enable(pf, num_alloc_vfs);
 	/* allocate memory */
 	vfs = kcalloc(num_alloc_vfs, sizeof(struct i40e_vf), GFP_KERNEL);
 	if (!vfs) {
@@ -1206,6 +1335,13 @@ static int i40e_vc_get_vf_resources_msg(struct i40e_vf *vf, u8 *msg)
 	vsi = pf->vsi[vf->lan_vsi_idx];
 	if (!vsi->info.pvid)
 		vfres->vf_offload_flags |= I40E_VIRTCHNL_VF_OFFLOAD_VLAN;
+
+	if (i40e_vf_client_capable(pf, vf->vf_id, I40E_CLIENT_IWARP) &&
+	    (vf->driver_caps & I40E_VIRTCHNL_VF_OFFLOAD_IWARP)) {
+		vfres->vf_offload_flags |= I40E_VIRTCHNL_VF_OFFLOAD_IWARP;
+		set_bit(I40E_VF_STAT_IWARPENA, &vf->vf_states);
+	}
+
 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) {
 		if (vf->driver_caps & I40E_VIRTCHNL_VF_OFFLOAD_RSS_AQ)
 			vfres->vf_offload_flags |=
@@ -1827,6 +1963,72 @@ error_param:
 }
 
 /**
+ * i40e_vc_iwarp_msg
+ * @vf: pointer to the VF info
+ * @msg: pointer to the msg buffer
+ * @msglen: msg length
+ *
+ * called from the VF for the iwarp msgs
+ **/
+static int i40e_vc_iwarp_msg(struct i40e_vf *vf, u8 *msg, u16 msglen)
+{
+	struct i40e_pf *pf = vf->pf;
+	int abs_vf_id = vf->vf_id + pf->hw.func_caps.vf_base_id;
+	i40e_status aq_ret = 0;
+
+	if (!test_bit(I40E_VF_STAT_ACTIVE, &vf->vf_states) ||
+	    !test_bit(I40E_VF_STAT_IWARPENA, &vf->vf_states)) {
+		aq_ret = I40E_ERR_PARAM;
+		goto error_param;
+	}
+
+	i40e_notify_client_of_vf_msg(pf->vsi[pf->lan_vsi], abs_vf_id,
+				     msg, msglen);
+
+error_param:
+	/* send the response to the VF */
+	return i40e_vc_send_resp_to_vf(vf, I40E_VIRTCHNL_OP_IWARP,
+				       aq_ret);
+}
+
+/**
+ * i40e_vc_iwarp_qvmap_msg
+ * @vf: pointer to the VF info
+ * @msg: pointer to the msg buffer
+ * @msglen: msg length
+ * @config: config qvmap or release it
+ *
+ * called from the VF for the iwarp msgs
+ **/
+static int i40e_vc_iwarp_qvmap_msg(struct i40e_vf *vf, u8 *msg, u16 msglen,
+				   bool config)
+{
+	struct i40e_virtchnl_iwarp_qvlist_info *qvlist_info =
+				(struct i40e_virtchnl_iwarp_qvlist_info *)msg;
+	i40e_status aq_ret = 0;
+
+	if (!test_bit(I40E_VF_STAT_ACTIVE, &vf->vf_states) ||
+	    !test_bit(I40E_VF_STAT_IWARPENA, &vf->vf_states)) {
+		aq_ret = I40E_ERR_PARAM;
+		goto error_param;
+	}
+
+	if (config) {
+		if (i40e_config_iwarp_qvlist(vf, qvlist_info))
+			aq_ret = I40E_ERR_PARAM;
+	} else {
+		i40e_release_iwarp_qvlist(vf);
+	}
+
+error_param:
+	/* send the response to the VF */
+	return i40e_vc_send_resp_to_vf(vf,
+			       config ? I40E_VIRTCHNL_OP_RELEASE_IWARP_IRQ_MAP :
+			       I40E_VIRTCHNL_OP_CONFIG_IWARP_IRQ_MAP,
+			       aq_ret);
+}
+
+/**
  * i40e_vc_validate_vf_msg
  * @vf: pointer to the VF info
  * @msg: pointer to the msg buffer
@@ -1921,6 +2123,32 @@ static int i40e_vc_validate_vf_msg(struct i40e_vf *vf, u32 v_opcode,
 	case I40E_VIRTCHNL_OP_GET_STATS:
 		valid_len = sizeof(struct i40e_virtchnl_queue_select);
 		break;
+	case I40E_VIRTCHNL_OP_IWARP:
+		/* These messages are opaque to us and will be validated in
+		 * the RDMA client code. We just need to check for nonzero
+		 * length. The firmware will enforce max length restrictions.
+		 */
+		if (msglen)
+			valid_len = msglen;
+		else
+			err_msg_format = true;
+		break;
+	case I40E_VIRTCHNL_OP_RELEASE_IWARP_IRQ_MAP:
+		valid_len = 0;
+		break;
+	case I40E_VIRTCHNL_OP_CONFIG_IWARP_IRQ_MAP:
+		valid_len = sizeof(struct i40e_virtchnl_iwarp_qvlist_info);
+		if (msglen >= valid_len) {
+			struct i40e_virtchnl_iwarp_qvlist_info *qv =
+				(struct i40e_virtchnl_iwarp_qvlist_info *)msg;
+			if (qv->num_vectors == 0) {
+				err_msg_format = true;
+				break;
+			}
+			valid_len += ((qv->num_vectors - 1) *
+				sizeof(struct i40e_virtchnl_iwarp_qv_info));
+		}
+		break;
 	/* These are always errors coming from the VF. */
 	case I40E_VIRTCHNL_OP_EVENT:
 	case I40E_VIRTCHNL_OP_UNKNOWN:
@@ -2010,6 +2238,15 @@ int i40e_vc_process_vf_msg(struct i40e_pf *pf, u16 vf_id, u32 v_opcode,
 	case I40E_VIRTCHNL_OP_GET_STATS:
 		ret = i40e_vc_get_stats_msg(vf, msg, msglen);
 		break;
+	case I40E_VIRTCHNL_OP_IWARP:
+		ret = i40e_vc_iwarp_msg(vf, msg, msglen);
+		break;
+	case I40E_VIRTCHNL_OP_CONFIG_IWARP_IRQ_MAP:
+		ret = i40e_vc_iwarp_qvmap_msg(vf, msg, msglen, true);
+		break;
+	case I40E_VIRTCHNL_OP_RELEASE_IWARP_IRQ_MAP:
+		ret = i40e_vc_iwarp_qvmap_msg(vf, msg, msglen, false);
+		break;
 	case I40E_VIRTCHNL_OP_UNKNOWN:
 	default:
 		dev_err(&pf->pdev->dev, "Unsupported opcode %d from VF %d\n",
diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
index e74642a0c42e..e7b2fba0309e 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
@@ -58,6 +58,7 @@ enum i40e_queue_ctrl {
 enum i40e_vf_states {
 	I40E_VF_STAT_INIT = 0,
 	I40E_VF_STAT_ACTIVE,
+	I40E_VF_STAT_IWARPENA,
 	I40E_VF_STAT_FCOEENA,
 	I40E_VF_STAT_DISABLED,
 };
@@ -66,6 +67,7 @@ enum i40e_vf_states {
 enum i40e_vf_capabilities {
 	I40E_VIRTCHNL_VF_CAP_PRIVILEGE = 0,
 	I40E_VIRTCHNL_VF_CAP_L2,
+	I40E_VIRTCHNL_VF_CAP_IWARP,
 };
 
 /* VF information structure */
@@ -106,6 +108,8 @@ struct i40e_vf {
 	bool link_forced;
 	bool link_up;		/* only valid if VF link is forced */
 	bool spoofchk;
+	/* RDMA Client */
+	struct i40e_virtchnl_iwarp_qvlist_info *qvlist_info;
 };
 
 void i40e_free_vfs(struct i40e_pf *pf);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
index 97f5114fc113..eb926e1ee71c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
@@ -407,6 +407,12 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 const char *mlx5_command_str(int command)
 {
 	switch (command) {
+	case MLX5_CMD_OP_QUERY_HCA_VPORT_CONTEXT:
+		return "QUERY_HCA_VPORT_CONTEXT";
+
+	case MLX5_CMD_OP_MODIFY_HCA_VPORT_CONTEXT:
+		return "MODIFY_HCA_VPORT_CONTEXT";
+
 	case MLX5_CMD_OP_QUERY_HCA_CAP:
 		return "QUERY_HCA_CAP";
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw.c b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
index aa1ab4702385..75c7ae6a5cc4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
@@ -98,88 +98,55 @@ int mlx5_query_hca_caps(struct mlx5_core_dev *dev)
 {
 	int err;
 
-	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL, HCA_CAP_OPMOD_GET_CUR);
-	if (err)
-		return err;
-
-	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL, HCA_CAP_OPMOD_GET_MAX);
+	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL);
 	if (err)
 		return err;
 
 	if (MLX5_CAP_GEN(dev, eth_net_offloads)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ETHERNET_OFFLOADS,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ETHERNET_OFFLOADS,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ETHERNET_OFFLOADS);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, pg)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ODP,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ODP,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ODP);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, atomic)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ATOMIC,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ATOMIC,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ATOMIC);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, roce)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ROCE,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ROCE,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ROCE);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, nic_flow_table)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_FLOW_TABLE,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_FLOW_TABLE,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_FLOW_TABLE);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, vport_group_manager) &&
 	    MLX5_CAP_GEN(dev, eswitch_flow_table)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH_FLOW_TABLE,
-					 HCA_CAP_OPMOD_GET_CUR);
-		if (err)
-			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH_FLOW_TABLE,
-					 HCA_CAP_OPMOD_GET_MAX);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH_FLOW_TABLE);
 		if (err)
 			return err;
 	}
 
 	if (MLX5_CAP_GEN(dev, eswitch_flow_table)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH,
-					 HCA_CAP_OPMOD_GET_CUR);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH);
 		if (err)
 			return err;
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ESWITCH,
-					 HCA_CAP_OPMOD_GET_MAX);
+	}
+
+	if (MLX5_CAP_GEN(dev, vector_calc)) {
+		err = mlx5_core_get_caps(dev, MLX5_CAP_VECTOR_CALC);
 		if (err)
 			return err;
 	}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 72a94e72ee25..3f3b2fae4991 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -341,8 +341,9 @@ static u16 to_fw_pkey_sz(u32 size)
 	}
 }
 
-int mlx5_core_get_caps(struct mlx5_core_dev *dev, enum mlx5_cap_type cap_type,
-		       enum mlx5_cap_mode cap_mode)
+static int mlx5_core_get_caps_mode(struct mlx5_core_dev *dev,
+				   enum mlx5_cap_type cap_type,
+				   enum mlx5_cap_mode cap_mode)
 {
 	u8 in[MLX5_ST_SZ_BYTES(query_hca_cap_in)];
 	int out_sz = MLX5_ST_SZ_BYTES(query_hca_cap_out);
@@ -392,6 +393,16 @@ query_ex:
 	return err;
 }
 
+int mlx5_core_get_caps(struct mlx5_core_dev *dev, enum mlx5_cap_type cap_type)
+{
+	int ret;
+
+	ret = mlx5_core_get_caps_mode(dev, cap_type, HCA_CAP_OPMOD_GET_CUR);
+	if (ret)
+		return ret;
+	return mlx5_core_get_caps_mode(dev, cap_type, HCA_CAP_OPMOD_GET_MAX);
+}
+
 static int set_caps(struct mlx5_core_dev *dev, void *in, int in_sz, int opmod)
 {
 	u32 out[MLX5_ST_SZ_DW(set_hca_cap_out)];
@@ -419,8 +430,7 @@ static int handle_hca_cap_atomic(struct mlx5_core_dev *dev)
 	int err;
 
 	if (MLX5_CAP_GEN(dev, atomic)) {
-		err = mlx5_core_get_caps(dev, MLX5_CAP_ATOMIC,
-					 HCA_CAP_OPMOD_GET_CUR);
+		err = mlx5_core_get_caps(dev, MLX5_CAP_ATOMIC);
 		if (err)
 			return err;
 	} else {
@@ -462,11 +472,7 @@ static int handle_hca_cap(struct mlx5_core_dev *dev)
 	if (!set_ctx)
 		goto query_ex;
 
-	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL, HCA_CAP_OPMOD_GET_MAX);
-	if (err)
-		goto query_ex;
-
-	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL, HCA_CAP_OPMOD_GET_CUR);
+	err = mlx5_core_get_caps(dev, MLX5_CAP_GENERAL);
 	if (err)
 		goto query_ex;
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
index 90ab09e375b8..bd518405859e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
@@ -852,7 +852,8 @@ int mlx5_nic_vport_disable_roce(struct mlx5_core_dev *mdev)
 EXPORT_SYMBOL_GPL(mlx5_nic_vport_disable_roce);
 
 int mlx5_core_query_vport_counter(struct mlx5_core_dev *dev, u8 other_vport,
-				  u8 port_num, void *out, size_t out_sz)
+				  int vf, u8 port_num, void *out,
+				  size_t out_sz)
 {
 	int	in_sz = MLX5_ST_SZ_BYTES(query_vport_counter_in);
 	int	is_group_manager;
@@ -871,7 +872,7 @@ int mlx5_core_query_vport_counter(struct mlx5_core_dev *dev, u8 other_vport,
 	if (other_vport) {
 		if (is_group_manager) {
 			MLX5_SET(query_vport_counter_in, in, other_vport, 1);
-			MLX5_SET(query_vport_counter_in, in, vport_number, 0);
+			MLX5_SET(query_vport_counter_in, in, vport_number, vf + 1);
 		} else {
 			err = -EPERM;
 			goto free;
@@ -890,3 +891,70 @@ free:
 	return err;
 }
 EXPORT_SYMBOL_GPL(mlx5_core_query_vport_counter);
+
+int mlx5_core_modify_hca_vport_context(struct mlx5_core_dev *dev,
+				       u8 other_vport, u8 port_num,
+				       int vf,
+				       struct mlx5_hca_vport_context *req)
+{
+	int in_sz = MLX5_ST_SZ_BYTES(modify_hca_vport_context_in);
+	u8 out[MLX5_ST_SZ_BYTES(modify_hca_vport_context_out)];
+	int is_group_manager;
+	void *in;
+	int err;
+	void *ctx;
+
+	mlx5_core_dbg(dev, "vf %d\n", vf);
+	is_group_manager = MLX5_CAP_GEN(dev, vport_group_manager);
+	in = kzalloc(in_sz, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	memset(out, 0, sizeof(out));
+	MLX5_SET(modify_hca_vport_context_in, in, opcode, MLX5_CMD_OP_MODIFY_HCA_VPORT_CONTEXT);
+	if (other_vport) {
+		if (is_group_manager) {
+			MLX5_SET(modify_hca_vport_context_in, in, other_vport, 1);
+			MLX5_SET(modify_hca_vport_context_in, in, vport_number, vf);
+		} else {
+			err = -EPERM;
+			goto ex;
+		}
+	}
+
+	if (MLX5_CAP_GEN(dev, num_ports) > 1)
+		MLX5_SET(modify_hca_vport_context_in, in, port_num, port_num);
+
+	ctx = MLX5_ADDR_OF(modify_hca_vport_context_in, in, hca_vport_context);
+	MLX5_SET(hca_vport_context, ctx, field_select, req->field_select);
+	MLX5_SET(hca_vport_context, ctx, sm_virt_aware, req->sm_virt_aware);
+	MLX5_SET(hca_vport_context, ctx, has_smi, req->has_smi);
+	MLX5_SET(hca_vport_context, ctx, has_raw, req->has_raw);
+	MLX5_SET(hca_vport_context, ctx, vport_state_policy, req->policy);
+	MLX5_SET(hca_vport_context, ctx, port_physical_state, req->phys_state);
+	MLX5_SET(hca_vport_context, ctx, vport_state, req->vport_state);
+	MLX5_SET64(hca_vport_context, ctx, port_guid, req->port_guid);
+	MLX5_SET64(hca_vport_context, ctx, node_guid, req->node_guid);
+	MLX5_SET(hca_vport_context, ctx, cap_mask1, req->cap_mask1);
+	MLX5_SET(hca_vport_context, ctx, cap_mask1_field_select, req->cap_mask1_perm);
+	MLX5_SET(hca_vport_context, ctx, cap_mask2, req->cap_mask2);
+	MLX5_SET(hca_vport_context, ctx, cap_mask2_field_select, req->cap_mask2_perm);
+	MLX5_SET(hca_vport_context, ctx, lid, req->lid);
+	MLX5_SET(hca_vport_context, ctx, init_type_reply, req->init_type_reply);
+	MLX5_SET(hca_vport_context, ctx, lmc, req->lmc);
+	MLX5_SET(hca_vport_context, ctx, subnet_timeout, req->subnet_timeout);
+	MLX5_SET(hca_vport_context, ctx, sm_lid, req->sm_lid);
+	MLX5_SET(hca_vport_context, ctx, sm_sl, req->sm_sl);
+	MLX5_SET(hca_vport_context, ctx, qkey_violation_counter, req->qkey_violation_counter);
+	MLX5_SET(hca_vport_context, ctx, pkey_violation_counter, req->pkey_violation_counter);
+	err = mlx5_cmd_exec(dev, in, in_sz, out, sizeof(out));
+	if (err)
+		goto ex;
+
+	err = mlx5_cmd_status_to_err_v2(out);
+
+ex:
+	kfree(in);
+	return err;
+}
+EXPORT_SYMBOL_GPL(mlx5_core_modify_hca_vport_context);
diff --git a/drivers/staging/rdma/hfi1/Kconfig b/drivers/staging/rdma/hfi1/Kconfig
index fd25078ee923..3e668d852f03 100644
--- a/drivers/staging/rdma/hfi1/Kconfig
+++ b/drivers/staging/rdma/hfi1/Kconfig
@@ -1,6 +1,7 @@
 config INFINIBAND_HFI1
 	tristate "Intel OPA Gen1 support"
-	depends on X86_64
+	depends on X86_64 && INFINIBAND_RDMAVT
+	select MMU_NOTIFIER
 	default m
 	---help---
 	This is a low-level driver for Intel OPA Gen1 adapter.
@@ -25,13 +26,3 @@ config SDMA_VERBOSITY
 	---help---
 	This is a configuration flag to enable verbose
 	SDMA debug
-config PRESCAN_RXQ
-	bool "Enable prescanning of the RX queue for ECNs"
-	depends on INFINIBAND_HFI1
-	default n
-	---help---
-	This option toggles the prescanning of the receive queue for
-	Explicit Congestion Notifications. If an ECN is detected, it
-	is processed as quickly as possible, the ECN is toggled off.
-	After the prescanning step, the receive queue is processed as
-	usual.
diff --git a/drivers/staging/rdma/hfi1/Makefile b/drivers/staging/rdma/hfi1/Makefile
index 68c5a315e557..8dc59382ee96 100644
--- a/drivers/staging/rdma/hfi1/Makefile
+++ b/drivers/staging/rdma/hfi1/Makefile
@@ -7,10 +7,12 @@
 #
 obj-$(CONFIG_INFINIBAND_HFI1) += hfi1.o
 
-hfi1-y := chip.o cq.o device.o diag.o dma.o driver.o efivar.o eprom.o file_ops.o firmware.o \
-	init.o intr.o keys.o mad.o mmap.o mr.o pcie.o pio.o pio_copy.o \
-	qp.o qsfp.o rc.o ruc.o sdma.o srq.o sysfs.o trace.o twsi.o \
-	uc.o ud.o user_pages.o user_sdma.o verbs_mcast.o verbs.o
+hfi1-y := affinity.o chip.o device.o diag.o driver.o efivar.o \
+	eprom.o file_ops.o firmware.o \
+	init.o intr.o mad.o mmu_rb.o pcie.o pio.o pio_copy.o platform.o \
+	qp.o qsfp.o rc.o ruc.o sdma.o sysfs.o trace.o twsi.o \
+	uc.o ud.o user_exp_rcv.o user_pages.o user_sdma.o verbs.o \
+	verbs_txreq.o
 hfi1-$(CONFIG_DEBUG_FS) += debugfs.o
 
 CFLAGS_trace.o = -I$(src)
diff --git a/drivers/staging/rdma/hfi1/affinity.c b/drivers/staging/rdma/hfi1/affinity.c
new file mode 100644
index 000000000000..2cb8ca77f876
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/affinity.c
@@ -0,0 +1,430 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <linux/topology.h>
+#include <linux/cpumask.h>
+#include <linux/module.h>
+
+#include "hfi.h"
+#include "affinity.h"
+#include "sdma.h"
+#include "trace.h"
+
+struct cpu_mask_set {
+	struct cpumask mask;
+	struct cpumask used;
+	uint gen;
+};
+
+struct hfi1_affinity {
+	struct cpu_mask_set def_intr;
+	struct cpu_mask_set rcv_intr;
+	struct cpu_mask_set proc;
+	/* spin lock to protect affinity struct */
+	spinlock_t lock;
+};
+
+/* Name of IRQ types, indexed by enum irq_type */
+static const char * const irq_type_names[] = {
+	"SDMA",
+	"RCVCTXT",
+	"GENERAL",
+	"OTHER",
+};
+
+static inline void init_cpu_mask_set(struct cpu_mask_set *set)
+{
+	cpumask_clear(&set->mask);
+	cpumask_clear(&set->used);
+	set->gen = 0;
+}
+
+/*
+ * Interrupt affinity.
+ *
+ * non-rcv avail gets a default mask that
+ * starts as possible cpus with threads reset
+ * and each rcv avail reset.
+ *
+ * rcv avail gets node relative 1 wrapping back
+ * to the node relative 1 as necessary.
+ *
+ */
+int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
+{
+	int node = pcibus_to_node(dd->pcidev->bus);
+	struct hfi1_affinity *info;
+	const struct cpumask *local_mask;
+	int curr_cpu, possible, i, ht;
+
+	if (node < 0)
+		node = numa_node_id();
+	dd->node = node;
+
+	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	if (!info)
+		return -ENOMEM;
+	spin_lock_init(&info->lock);
+
+	init_cpu_mask_set(&info->def_intr);
+	init_cpu_mask_set(&info->rcv_intr);
+	init_cpu_mask_set(&info->proc);
+
+	local_mask = cpumask_of_node(dd->node);
+	if (cpumask_first(local_mask) >= nr_cpu_ids)
+		local_mask = topology_core_cpumask(0);
+	/* use local mask as default */
+	cpumask_copy(&info->def_intr.mask, local_mask);
+	/*
+	 * Remove HT cores from the default mask.  Do this in two steps below.
+	 */
+	possible = cpumask_weight(&info->def_intr.mask);
+	ht = cpumask_weight(topology_sibling_cpumask(
+					cpumask_first(&info->def_intr.mask)));
+	/*
+	 * Step 1.  Skip over the first N HT siblings and use them as the
+	 * "real" cores.  Assumes that HT cores are not enumerated in
+	 * succession (except in the single core case).
+	 */
+	curr_cpu = cpumask_first(&info->def_intr.mask);
+	for (i = 0; i < possible / ht; i++)
+		curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
+	/*
+	 * Step 2.  Remove the remaining HT siblings.  Use cpumask_next() to
+	 * skip any gaps.
+	 */
+	for (; i < possible; i++) {
+		cpumask_clear_cpu(curr_cpu, &info->def_intr.mask);
+		curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
+	}
+
+	/*  fill in the receive list */
+	possible = cpumask_weight(&info->def_intr.mask);
+	curr_cpu = cpumask_first(&info->def_intr.mask);
+	if (possible == 1) {
+		/*  only one CPU, everyone will use it */
+		cpumask_set_cpu(curr_cpu, &info->rcv_intr.mask);
+	} else {
+		/*
+		 * Retain the first CPU in the default list for the control
+		 * context.
+		 */
+		curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
+		/*
+		 * Remove the remaining kernel receive queues from
+		 * the default list and add them to the receive list.
+		 */
+		for (i = 0; i < dd->n_krcv_queues - 1; i++) {
+			cpumask_clear_cpu(curr_cpu, &info->def_intr.mask);
+			cpumask_set_cpu(curr_cpu, &info->rcv_intr.mask);
+			curr_cpu = cpumask_next(curr_cpu, &info->def_intr.mask);
+			if (curr_cpu >= nr_cpu_ids)
+				break;
+		}
+	}
+
+	cpumask_copy(&info->proc.mask, cpu_online_mask);
+	dd->affinity = info;
+	return 0;
+}
+
+void hfi1_dev_affinity_free(struct hfi1_devdata *dd)
+{
+	kfree(dd->affinity);
+}
+
+int hfi1_get_irq_affinity(struct hfi1_devdata *dd, struct hfi1_msix_entry *msix)
+{
+	int ret;
+	cpumask_var_t diff;
+	struct cpu_mask_set *set;
+	struct sdma_engine *sde = NULL;
+	struct hfi1_ctxtdata *rcd = NULL;
+	char extra[64];
+	int cpu = -1;
+
+	extra[0] = '\0';
+	cpumask_clear(&msix->mask);
+
+	ret = zalloc_cpumask_var(&diff, GFP_KERNEL);
+	if (!ret)
+		return -ENOMEM;
+
+	switch (msix->type) {
+	case IRQ_SDMA:
+		sde = (struct sdma_engine *)msix->arg;
+		scnprintf(extra, 64, "engine %u", sde->this_idx);
+		/* fall through */
+	case IRQ_GENERAL:
+		set = &dd->affinity->def_intr;
+		break;
+	case IRQ_RCVCTXT:
+		rcd = (struct hfi1_ctxtdata *)msix->arg;
+		if (rcd->ctxt == HFI1_CTRL_CTXT) {
+			set = &dd->affinity->def_intr;
+			cpu = cpumask_first(&set->mask);
+		} else {
+			set = &dd->affinity->rcv_intr;
+		}
+		scnprintf(extra, 64, "ctxt %u", rcd->ctxt);
+		break;
+	default:
+		dd_dev_err(dd, "Invalid IRQ type %d\n", msix->type);
+		return -EINVAL;
+	}
+
+	/*
+	 * The control receive context is placed on a particular CPU, which
+	 * is set above.  Skip accounting for it.  Everything else finds its
+	 * CPU here.
+	 */
+	if (cpu == -1) {
+		spin_lock(&dd->affinity->lock);
+		if (cpumask_equal(&set->mask, &set->used)) {
+			/*
+			 * We've used up all the CPUs, bump up the generation
+			 * and reset the 'used' map
+			 */
+			set->gen++;
+			cpumask_clear(&set->used);
+		}
+		cpumask_andnot(diff, &set->mask, &set->used);
+		cpu = cpumask_first(diff);
+		cpumask_set_cpu(cpu, &set->used);
+		spin_unlock(&dd->affinity->lock);
+	}
+
+	switch (msix->type) {
+	case IRQ_SDMA:
+		sde->cpu = cpu;
+		break;
+	case IRQ_GENERAL:
+	case IRQ_RCVCTXT:
+	case IRQ_OTHER:
+		break;
+	}
+
+	cpumask_set_cpu(cpu, &msix->mask);
+	dd_dev_info(dd, "IRQ vector: %u, type %s %s -> cpu: %d\n",
+		    msix->msix.vector, irq_type_names[msix->type],
+		    extra, cpu);
+	irq_set_affinity_hint(msix->msix.vector, &msix->mask);
+
+	free_cpumask_var(diff);
+	return 0;
+}
+
+void hfi1_put_irq_affinity(struct hfi1_devdata *dd,
+			   struct hfi1_msix_entry *msix)
+{
+	struct cpu_mask_set *set = NULL;
+	struct hfi1_ctxtdata *rcd;
+
+	switch (msix->type) {
+	case IRQ_SDMA:
+	case IRQ_GENERAL:
+		set = &dd->affinity->def_intr;
+		break;
+	case IRQ_RCVCTXT:
+		rcd = (struct hfi1_ctxtdata *)msix->arg;
+		/* only do accounting for non control contexts */
+		if (rcd->ctxt != HFI1_CTRL_CTXT)
+			set = &dd->affinity->rcv_intr;
+		break;
+	default:
+		return;
+	}
+
+	if (set) {
+		spin_lock(&dd->affinity->lock);
+		cpumask_andnot(&set->used, &set->used, &msix->mask);
+		if (cpumask_empty(&set->used) && set->gen) {
+			set->gen--;
+			cpumask_copy(&set->used, &set->mask);
+		}
+		spin_unlock(&dd->affinity->lock);
+	}
+
+	irq_set_affinity_hint(msix->msix.vector, NULL);
+	cpumask_clear(&msix->mask);
+}
+
+int hfi1_get_proc_affinity(struct hfi1_devdata *dd, int node)
+{
+	int cpu = -1, ret;
+	cpumask_var_t diff, mask, intrs;
+	const struct cpumask *node_mask,
+		*proc_mask = tsk_cpus_allowed(current);
+	struct cpu_mask_set *set = &dd->affinity->proc;
+	char buf[1024];
+
+	/*
+	 * check whether process/context affinity has already
+	 * been set
+	 */
+	if (cpumask_weight(proc_mask) == 1) {
+		scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(proc_mask));
+		hfi1_cdbg(PROC, "PID %u %s affinity set to CPU %s",
+			  current->pid, current->comm, buf);
+		/*
+		 * Mark the pre-set CPU as used. This is atomic so we don't
+		 * need the lock
+		 */
+		cpu = cpumask_first(proc_mask);
+		cpumask_set_cpu(cpu, &set->used);
+		goto done;
+	} else if (cpumask_weight(proc_mask) < cpumask_weight(&set->mask)) {
+		scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(proc_mask));
+		hfi1_cdbg(PROC, "PID %u %s affinity set to CPU set(s) %s",
+			  current->pid, current->comm, buf);
+		goto done;
+	}
+
+	/*
+	 * The process does not have a preset CPU affinity so find one to
+	 * recommend. We prefer CPUs on the same NUMA as the device.
+	 */
+
+	ret = zalloc_cpumask_var(&diff, GFP_KERNEL);
+	if (!ret)
+		goto done;
+	ret = zalloc_cpumask_var(&mask, GFP_KERNEL);
+	if (!ret)
+		goto free_diff;
+	ret = zalloc_cpumask_var(&intrs, GFP_KERNEL);
+	if (!ret)
+		goto free_mask;
+
+	spin_lock(&dd->affinity->lock);
+	/*
+	 * If we've used all available CPUs, clear the mask and start
+	 * overloading.
+	 */
+	if (cpumask_equal(&set->mask, &set->used)) {
+		set->gen++;
+		cpumask_clear(&set->used);
+	}
+
+	/* CPUs used by interrupt handlers */
+	cpumask_copy(intrs, (dd->affinity->def_intr.gen ?
+			     &dd->affinity->def_intr.mask :
+			     &dd->affinity->def_intr.used));
+	cpumask_or(intrs, intrs, (dd->affinity->rcv_intr.gen ?
+				  &dd->affinity->rcv_intr.mask :
+				  &dd->affinity->rcv_intr.used));
+	scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(intrs));
+	hfi1_cdbg(PROC, "CPUs used by interrupts: %s", buf);
+
+	/*
+	 * If we don't have a NUMA node requested, preference is towards
+	 * device NUMA node
+	 */
+	if (node == -1)
+		node = dd->node;
+	node_mask = cpumask_of_node(node);
+	scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(node_mask));
+	hfi1_cdbg(PROC, "device on NUMA %u, CPUs %s", node, buf);
+
+	/* diff will hold all unused cpus */
+	cpumask_andnot(diff, &set->mask, &set->used);
+	scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(diff));
+	hfi1_cdbg(PROC, "unused CPUs (all) %s", buf);
+
+	/* get cpumask of available CPUs on preferred NUMA */
+	cpumask_and(mask, diff, node_mask);
+	scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(mask));
+	hfi1_cdbg(PROC, "available cpus on NUMA %s", buf);
+
+	/*
+	 * At first, we don't want to place processes on the same
+	 * CPUs as interrupt handlers.
+	 */
+	cpumask_andnot(diff, mask, intrs);
+	if (!cpumask_empty(diff))
+		cpumask_copy(mask, diff);
+
+	/*
+	 * if we don't have a cpu on the preferred NUMA, get
+	 * the list of the remaining available CPUs
+	 */
+	if (cpumask_empty(mask)) {
+		cpumask_andnot(diff, &set->mask, &set->used);
+		cpumask_andnot(mask, diff, node_mask);
+	}
+	scnprintf(buf, 1024, "%*pbl", cpumask_pr_args(mask));
+	hfi1_cdbg(PROC, "possible CPUs for process %s", buf);
+
+	cpu = cpumask_first(mask);
+	if (cpu >= nr_cpu_ids) /* empty */
+		cpu = -1;
+	else
+		cpumask_set_cpu(cpu, &set->used);
+	spin_unlock(&dd->affinity->lock);
+
+	free_cpumask_var(intrs);
+free_mask:
+	free_cpumask_var(mask);
+free_diff:
+	free_cpumask_var(diff);
+done:
+	return cpu;
+}
+
+void hfi1_put_proc_affinity(struct hfi1_devdata *dd, int cpu)
+{
+	struct cpu_mask_set *set = &dd->affinity->proc;
+
+	if (cpu < 0)
+		return;
+	spin_lock(&dd->affinity->lock);
+	cpumask_clear_cpu(cpu, &set->used);
+	if (cpumask_empty(&set->used) && set->gen) {
+		set->gen--;
+		cpumask_copy(&set->used, &set->mask);
+	}
+	spin_unlock(&dd->affinity->lock);
+}
+
diff --git a/drivers/staging/rdma/hfi1/affinity.h b/drivers/staging/rdma/hfi1/affinity.h
new file mode 100644
index 000000000000..b287e4963024
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/affinity.h
@@ -0,0 +1,91 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#ifndef _HFI1_AFFINITY_H
+#define _HFI1_AFFINITY_H
+
+#include "hfi.h"
+
+enum irq_type {
+	IRQ_SDMA,
+	IRQ_RCVCTXT,
+	IRQ_GENERAL,
+	IRQ_OTHER
+};
+
+/* Can be used for both memory and cpu */
+enum affinity_flags {
+	AFF_AUTO,
+	AFF_NUMA_LOCAL,
+	AFF_DEV_LOCAL,
+	AFF_IRQ_LOCAL
+};
+
+struct hfi1_msix_entry;
+
+/* Initialize driver affinity data */
+int hfi1_dev_affinity_init(struct hfi1_devdata *);
+/* Free driver affinity data */
+void hfi1_dev_affinity_free(struct hfi1_devdata *);
+/*
+ * Set IRQ affinity to a CPU. The function will determine the
+ * CPU and set the affinity to it.
+ */
+int hfi1_get_irq_affinity(struct hfi1_devdata *, struct hfi1_msix_entry *);
+/*
+ * Remove the IRQ's CPU affinity. This function also updates
+ * any internal CPU tracking data
+ */
+void hfi1_put_irq_affinity(struct hfi1_devdata *, struct hfi1_msix_entry *);
+/*
+ * Determine a CPU affinity for a user process, if the process does not
+ * have an affinity set yet.
+ */
+int hfi1_get_proc_affinity(struct hfi1_devdata *, int);
+/* Release a CPU used by a user process. */
+void hfi1_put_proc_affinity(struct hfi1_devdata *, int);
+
+#endif /* _HFI1_AFFINITY_H */
diff --git a/drivers/staging/rdma/hfi1/aspm.h b/drivers/staging/rdma/hfi1/aspm.h
new file mode 100644
index 000000000000..0d58fe3b49b5
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/aspm.h
@@ -0,0 +1,309 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#ifndef _ASPM_H
+#define _ASPM_H
+
+#include "hfi.h"
+
+extern uint aspm_mode;
+
+enum aspm_mode {
+	ASPM_MODE_DISABLED = 0,	/* ASPM always disabled, performance mode */
+	ASPM_MODE_ENABLED = 1,	/* ASPM always enabled, power saving mode */
+	ASPM_MODE_DYNAMIC = 2,	/* ASPM enabled/disabled dynamically */
+};
+
+/* Time after which the timer interrupt will re-enable ASPM */
+#define ASPM_TIMER_MS 1000
+/* Time for which interrupts are ignored after a timer has been scheduled */
+#define ASPM_RESCHED_TIMER_MS (ASPM_TIMER_MS / 2)
+/* Two interrupts within this time trigger ASPM disable */
+#define ASPM_TRIGGER_MS 1
+#define ASPM_TRIGGER_NS (ASPM_TRIGGER_MS * 1000 * 1000ull)
+#define ASPM_L1_SUPPORTED(reg) \
+	(((reg & PCI_EXP_LNKCAP_ASPMS) >> 10) & 0x2)
+
+static inline bool aspm_hw_l1_supported(struct hfi1_devdata *dd)
+{
+	struct pci_dev *parent = dd->pcidev->bus->self;
+	u32 up, dn;
+
+	/*
+	 * If the driver does not have access to the upstream component,
+	 * it cannot support ASPM L1 at all.
+	 */
+	if (!parent)
+		return false;
+
+	pcie_capability_read_dword(dd->pcidev, PCI_EXP_LNKCAP, &dn);
+	dn = ASPM_L1_SUPPORTED(dn);
+
+	pcie_capability_read_dword(parent, PCI_EXP_LNKCAP, &up);
+	up = ASPM_L1_SUPPORTED(up);
+
+	/* ASPM works on A-step but is reported as not supported */
+	return (!!dn || is_ax(dd)) && !!up;
+}
+
+/* Set L1 entrance latency for slower entry to L1 */
+static inline void aspm_hw_set_l1_ent_latency(struct hfi1_devdata *dd)
+{
+	u32 l1_ent_lat = 0x4u;
+	u32 reg32;
+
+	pci_read_config_dword(dd->pcidev, PCIE_CFG_REG_PL3, &reg32);
+	reg32 &= ~PCIE_CFG_REG_PL3_L1_ENT_LATENCY_SMASK;
+	reg32 |= l1_ent_lat << PCIE_CFG_REG_PL3_L1_ENT_LATENCY_SHIFT;
+	pci_write_config_dword(dd->pcidev, PCIE_CFG_REG_PL3, reg32);
+}
+
+static inline void aspm_hw_enable_l1(struct hfi1_devdata *dd)
+{
+	struct pci_dev *parent = dd->pcidev->bus->self;
+
+	/*
+	 * If the driver does not have access to the upstream component,
+	 * it cannot support ASPM L1 at all.
+	 */
+	if (!parent)
+		return;
+
+	/* Enable ASPM L1 first in upstream component and then downstream */
+	pcie_capability_clear_and_set_word(parent, PCI_EXP_LNKCTL,
+					   PCI_EXP_LNKCTL_ASPMC,
+					   PCI_EXP_LNKCTL_ASPM_L1);
+	pcie_capability_clear_and_set_word(dd->pcidev, PCI_EXP_LNKCTL,
+					   PCI_EXP_LNKCTL_ASPMC,
+					   PCI_EXP_LNKCTL_ASPM_L1);
+}
+
+static inline void aspm_hw_disable_l1(struct hfi1_devdata *dd)
+{
+	struct pci_dev *parent = dd->pcidev->bus->self;
+
+	/* Disable ASPM L1 first in downstream component and then upstream */
+	pcie_capability_clear_and_set_word(dd->pcidev, PCI_EXP_LNKCTL,
+					   PCI_EXP_LNKCTL_ASPMC, 0x0);
+	if (parent)
+		pcie_capability_clear_and_set_word(parent, PCI_EXP_LNKCTL,
+						   PCI_EXP_LNKCTL_ASPMC, 0x0);
+}
+
+static inline void aspm_enable(struct hfi1_devdata *dd)
+{
+	if (dd->aspm_enabled || aspm_mode == ASPM_MODE_DISABLED ||
+	    !dd->aspm_supported)
+		return;
+
+	aspm_hw_enable_l1(dd);
+	dd->aspm_enabled = true;
+}
+
+static inline void aspm_disable(struct hfi1_devdata *dd)
+{
+	if (!dd->aspm_enabled || aspm_mode == ASPM_MODE_ENABLED)
+		return;
+
+	aspm_hw_disable_l1(dd);
+	dd->aspm_enabled = false;
+}
+
+static inline void aspm_disable_inc(struct hfi1_devdata *dd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->aspm_lock, flags);
+	aspm_disable(dd);
+	atomic_inc(&dd->aspm_disabled_cnt);
+	spin_unlock_irqrestore(&dd->aspm_lock, flags);
+}
+
+static inline void aspm_enable_dec(struct hfi1_devdata *dd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->aspm_lock, flags);
+	if (atomic_dec_and_test(&dd->aspm_disabled_cnt))
+		aspm_enable(dd);
+	spin_unlock_irqrestore(&dd->aspm_lock, flags);
+}
+
+/* ASPM processing for each receive context interrupt */
+static inline void aspm_ctx_disable(struct hfi1_ctxtdata *rcd)
+{
+	bool restart_timer;
+	bool close_interrupts;
+	unsigned long flags;
+	ktime_t now, prev;
+
+	/* Quickest exit for minimum impact */
+	if (!rcd->aspm_intr_supported)
+		return;
+
+	spin_lock_irqsave(&rcd->aspm_lock, flags);
+	/* PSM contexts are open */
+	if (!rcd->aspm_intr_enable)
+		goto unlock;
+
+	prev = rcd->aspm_ts_last_intr;
+	now = ktime_get();
+	rcd->aspm_ts_last_intr = now;
+
+	/* An interrupt pair close together in time */
+	close_interrupts = ktime_to_ns(ktime_sub(now, prev)) < ASPM_TRIGGER_NS;
+
+	/* Don't push out our timer till this much time has elapsed */
+	restart_timer = ktime_to_ns(ktime_sub(now, rcd->aspm_ts_timer_sched)) >
+				    ASPM_RESCHED_TIMER_MS * NSEC_PER_MSEC;
+	restart_timer = restart_timer && close_interrupts;
+
+	/* Disable ASPM and schedule timer */
+	if (rcd->aspm_enabled && close_interrupts) {
+		aspm_disable_inc(rcd->dd);
+		rcd->aspm_enabled = false;
+		restart_timer = true;
+	}
+
+	if (restart_timer) {
+		mod_timer(&rcd->aspm_timer,
+			  jiffies + msecs_to_jiffies(ASPM_TIMER_MS));
+		rcd->aspm_ts_timer_sched = now;
+	}
+unlock:
+	spin_unlock_irqrestore(&rcd->aspm_lock, flags);
+}
+
+/* Timer function for re-enabling ASPM in the absence of interrupt activity */
+static inline void aspm_ctx_timer_function(unsigned long data)
+{
+	struct hfi1_ctxtdata *rcd = (struct hfi1_ctxtdata *)data;
+	unsigned long flags;
+
+	spin_lock_irqsave(&rcd->aspm_lock, flags);
+	aspm_enable_dec(rcd->dd);
+	rcd->aspm_enabled = true;
+	spin_unlock_irqrestore(&rcd->aspm_lock, flags);
+}
+
+/* Disable interrupt processing for verbs contexts when PSM contexts are open */
+static inline void aspm_disable_all(struct hfi1_devdata *dd)
+{
+	struct hfi1_ctxtdata *rcd;
+	unsigned long flags;
+	unsigned i;
+
+	for (i = 0; i < dd->first_user_ctxt; i++) {
+		rcd = dd->rcd[i];
+		del_timer_sync(&rcd->aspm_timer);
+		spin_lock_irqsave(&rcd->aspm_lock, flags);
+		rcd->aspm_intr_enable = false;
+		spin_unlock_irqrestore(&rcd->aspm_lock, flags);
+	}
+
+	aspm_disable(dd);
+	atomic_set(&dd->aspm_disabled_cnt, 0);
+}
+
+/* Re-enable interrupt processing for verbs contexts */
+static inline void aspm_enable_all(struct hfi1_devdata *dd)
+{
+	struct hfi1_ctxtdata *rcd;
+	unsigned long flags;
+	unsigned i;
+
+	aspm_enable(dd);
+
+	if (aspm_mode != ASPM_MODE_DYNAMIC)
+		return;
+
+	for (i = 0; i < dd->first_user_ctxt; i++) {
+		rcd = dd->rcd[i];
+		spin_lock_irqsave(&rcd->aspm_lock, flags);
+		rcd->aspm_intr_enable = true;
+		rcd->aspm_enabled = true;
+		spin_unlock_irqrestore(&rcd->aspm_lock, flags);
+	}
+}
+
+static inline void aspm_ctx_init(struct hfi1_ctxtdata *rcd)
+{
+	spin_lock_init(&rcd->aspm_lock);
+	setup_timer(&rcd->aspm_timer, aspm_ctx_timer_function,
+		    (unsigned long)rcd);
+	rcd->aspm_intr_supported = rcd->dd->aspm_supported &&
+		aspm_mode == ASPM_MODE_DYNAMIC &&
+		rcd->ctxt < rcd->dd->first_user_ctxt;
+}
+
+static inline void aspm_init(struct hfi1_devdata *dd)
+{
+	unsigned i;
+
+	spin_lock_init(&dd->aspm_lock);
+	dd->aspm_supported = aspm_hw_l1_supported(dd);
+
+	for (i = 0; i < dd->first_user_ctxt; i++)
+		aspm_ctx_init(dd->rcd[i]);
+
+	/* Start with ASPM disabled */
+	aspm_hw_set_l1_ent_latency(dd);
+	dd->aspm_enabled = false;
+	aspm_hw_disable_l1(dd);
+
+	/* Now turn on ASPM if configured */
+	aspm_enable_all(dd);
+}
+
+static inline void aspm_exit(struct hfi1_devdata *dd)
+{
+	aspm_disable_all(dd);
+
+	/* Turn on ASPM on exit to conserve power */
+	aspm_enable(dd);
+}
+
+#endif /* _ASPM_H */
diff --git a/drivers/staging/rdma/hfi1/chip.c b/drivers/staging/rdma/hfi1/chip.c
index 46a1830b509b..16eb653903e0 100644
--- a/drivers/staging/rdma/hfi1/chip.c
+++ b/drivers/staging/rdma/hfi1/chip.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -64,6 +61,8 @@
 #include "sdma.h"
 #include "eprom.h"
 #include "efivar.h"
+#include "platform.h"
+#include "aspm.h"
 
 #define NUM_IB_PORTS 1
 
@@ -420,10 +419,10 @@ static struct flag_table pio_err_status_flags[] = {
 	SEC_SPC_FREEZE,
 	SEND_PIO_ERR_STATUS_PIO_STATE_MACHINE_ERR_SMASK),
 /*23*/	FLAG_ENTRY("PioWriteQwValidParity",
-	SEC_WRITE_DROPPED|SEC_SPC_FREEZE,
+	SEC_WRITE_DROPPED | SEC_SPC_FREEZE,
 	SEND_PIO_ERR_STATUS_PIO_WRITE_QW_VALID_PARITY_ERR_SMASK),
 /*24*/	FLAG_ENTRY("PioBlockQwCountParity",
-	SEC_WRITE_DROPPED|SEC_SPC_FREEZE,
+	SEC_WRITE_DROPPED | SEC_SPC_FREEZE,
 	SEND_PIO_ERR_STATUS_PIO_BLOCK_QW_COUNT_PARITY_ERR_SMASK),
 /*25*/	FLAG_ENTRY("PioVlfVlLenParity",
 	SEC_SPC_FREEZE,
@@ -509,6 +508,12 @@ static struct flag_table sdma_err_status_flags[] = {
 		| SEND_DMA_ERR_STATUS_SDMA_CSR_PARITY_ERR_SMASK \
 		| SEND_DMA_ERR_STATUS_SDMA_PCIE_REQ_TRACKING_UNC_ERR_SMASK)
 
+/* SendEgressErrInfo bits that correspond to a PortXmitDiscard counter */
+#define PORT_DISCARD_EGRESS_ERRS \
+	(SEND_EGRESS_ERR_INFO_TOO_LONG_IB_PACKET_ERR_SMASK \
+	| SEND_EGRESS_ERR_INFO_VL_MAPPING_ERR_SMASK \
+	| SEND_EGRESS_ERR_INFO_VL_ERR_SMASK)
+
 /*
  * TXE Egress Error flags
  */
@@ -936,7 +941,7 @@ static struct flag_table dc8051_err_flags[] = {
 	FLAG_ENTRY0("IRAM_MBE", D8E(IRAM_MBE)),
 	FLAG_ENTRY0("IRAM_SBE", D8E(IRAM_SBE)),
 	FLAG_ENTRY0("UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES",
-		D8E(UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES)),
+		    D8E(UNMATCHED_SECURE_MSG_ACROSS_BCC_LANES)),
 	FLAG_ENTRY0("INVALID_CSR_ADDR", D8E(INVALID_CSR_ADDR)),
 };
 
@@ -950,7 +955,7 @@ static struct flag_table dc8051_info_err_flags[] = {
 	FLAG_ENTRY0("Unknown frame received",  UNKNOWN_FRAME),
 	FLAG_ENTRY0("Target BER not met",      TARGET_BER_NOT_MET),
 	FLAG_ENTRY0("Serdes internal loopback failure",
-					FAILED_SERDES_INTERNAL_LOOPBACK),
+		    FAILED_SERDES_INTERNAL_LOOPBACK),
 	FLAG_ENTRY0("Failed SerDes init",      FAILED_SERDES_INIT),
 	FLAG_ENTRY0("Failed LNI(Polling)",     FAILED_LNI_POLLING),
 	FLAG_ENTRY0("Failed LNI(Debounce)",    FAILED_LNI_DEBOUNCE),
@@ -958,7 +963,8 @@ static struct flag_table dc8051_info_err_flags[] = {
 	FLAG_ENTRY0("Failed LNI(OptEq)",       FAILED_LNI_OPTEQ),
 	FLAG_ENTRY0("Failed LNI(VerifyCap_1)", FAILED_LNI_VERIFY_CAP1),
 	FLAG_ENTRY0("Failed LNI(VerifyCap_2)", FAILED_LNI_VERIFY_CAP2),
-	FLAG_ENTRY0("Failed LNI(ConfigLT)",    FAILED_LNI_CONFIGLT)
+	FLAG_ENTRY0("Failed LNI(ConfigLT)",    FAILED_LNI_CONFIGLT),
+	FLAG_ENTRY0("Host Handshake Timeout",  HOST_HANDSHAKE_TIMEOUT)
 };
 
 /*
@@ -978,7 +984,6 @@ static struct flag_table dc8051_info_host_msg_flags[] = {
 	FLAG_ENTRY0("Link going down", 0x0100),
 };
 
-
 static u32 encoded_size(u32 size);
 static u32 chip_to_opa_lstate(struct hfi1_devdata *dd, u32 chip_lstate);
 static int set_physical_link_state(struct hfi1_devdata *dd, u64 state);
@@ -1140,11 +1145,8 @@ struct cntr_entry {
 	/*
 	 * accessor for stat element, context either dd or ppd
 	 */
-	u64 (*rw_cntr)(const struct cntr_entry *,
-			       void *context,
-			       int vl,
-			       int mode,
-			       u64 data);
+	u64 (*rw_cntr)(const struct cntr_entry *, void *context, int vl,
+		       int mode, u64 data);
 };
 
 #define C_RCV_HDR_OVF_FIRST C_RCV_HDR_OVF_0
@@ -1188,7 +1190,7 @@ CNTR_ELEM(#name, \
 #define OVR_LBL(ctx) C_RCV_HDR_OVF_ ## ctx
 #define OVR_ELM(ctx) \
 CNTR_ELEM("RcvHdrOvr" #ctx, \
-	  (RCV_HDR_OVFL_CNT + ctx*0x100), \
+	  (RCV_HDR_OVFL_CNT + ctx * 0x100), \
 	  0, CNTR_NORMAL, port_access_u64_csr)
 
 /* 32bit TXE */
@@ -1274,7 +1276,6 @@ static inline u64 read_write_csr(const struct hfi1_devdata *dd, u32 csr,
 {
 	u64 ret;
 
-
 	if (mode == CNTR_MODE_R) {
 		ret = read_csr(dd, csr);
 	} else if (mode == CNTR_MODE_W) {
@@ -1291,17 +1292,65 @@ static inline u64 read_write_csr(const struct hfi1_devdata *dd, u32 csr,
 
 /* Dev Access */
 static u64 dev_access_u32_csr(const struct cntr_entry *entry,
-			    void *context, int vl, int mode, u64 data)
+			      void *context, int vl, int mode, u64 data)
 {
 	struct hfi1_devdata *dd = context;
+	u64 csr = entry->csr;
 
-	if (vl != CNTR_INVALID_VL)
-		return 0;
-	return read_write_csr(dd, entry->csr, mode, data);
+	if (entry->flags & CNTR_SDMA) {
+		if (vl == CNTR_INVALID_VL)
+			return 0;
+		csr += 0x100 * vl;
+	} else {
+		if (vl != CNTR_INVALID_VL)
+			return 0;
+	}
+	return read_write_csr(dd, csr, mode, data);
+}
+
+static u64 access_sde_err_cnt(const struct cntr_entry *entry,
+			      void *context, int idx, int mode, u64 data)
+{
+	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+	if (dd->per_sdma && idx < dd->num_sdma)
+		return dd->per_sdma[idx].err_cnt;
+	return 0;
+}
+
+static u64 access_sde_int_cnt(const struct cntr_entry *entry,
+			      void *context, int idx, int mode, u64 data)
+{
+	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+	if (dd->per_sdma && idx < dd->num_sdma)
+		return dd->per_sdma[idx].sdma_int_cnt;
+	return 0;
+}
+
+static u64 access_sde_idle_int_cnt(const struct cntr_entry *entry,
+				   void *context, int idx, int mode, u64 data)
+{
+	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+	if (dd->per_sdma && idx < dd->num_sdma)
+		return dd->per_sdma[idx].idle_int_cnt;
+	return 0;
+}
+
+static u64 access_sde_progress_int_cnt(const struct cntr_entry *entry,
+				       void *context, int idx, int mode,
+				       u64 data)
+{
+	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+	if (dd->per_sdma && idx < dd->num_sdma)
+		return dd->per_sdma[idx].progress_int_cnt;
+	return 0;
 }
 
 static u64 dev_access_u64_csr(const struct cntr_entry *entry, void *context,
-			    int vl, int mode, u64 data)
+			      int vl, int mode, u64 data)
 {
 	struct hfi1_devdata *dd = context;
 
@@ -1322,7 +1371,7 @@ static u64 dev_access_u64_csr(const struct cntr_entry *entry, void *context,
 }
 
 static u64 dc_access_lcb_cntr(const struct cntr_entry *entry, void *context,
-			    int vl, int mode, u64 data)
+			      int vl, int mode, u64 data)
 {
 	struct hfi1_devdata *dd = context;
 	u32 csr = entry->csr;
@@ -1346,7 +1395,7 @@ static u64 dc_access_lcb_cntr(const struct cntr_entry *entry, void *context,
 
 /* Port Access */
 static u64 port_access_u32_csr(const struct cntr_entry *entry, void *context,
-			     int vl, int mode, u64 data)
+			       int vl, int mode, u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 
@@ -1356,7 +1405,7 @@ static u64 port_access_u32_csr(const struct cntr_entry *entry, void *context,
 }
 
 static u64 port_access_u64_csr(const struct cntr_entry *entry,
-			     void *context, int vl, int mode, u64 data)
+			       void *context, int vl, int mode, u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 	u64 val;
@@ -1396,7 +1445,7 @@ static inline u64 read_write_sw(struct hfi1_devdata *dd, u64 *cntr, int mode,
 }
 
 static u64 access_sw_link_dn_cnt(const struct cntr_entry *entry, void *context,
-			       int vl, int mode, u64 data)
+				 int vl, int mode, u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 
@@ -1406,7 +1455,7 @@ static u64 access_sw_link_dn_cnt(const struct cntr_entry *entry, void *context,
 }
 
 static u64 access_sw_link_up_cnt(const struct cntr_entry *entry, void *context,
-			       int vl, int mode, u64 data)
+				 int vl, int mode, u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 
@@ -1427,18 +1476,25 @@ static u64 access_sw_unknown_frame_cnt(const struct cntr_entry *entry,
 }
 
 static u64 access_sw_xmit_discards(const struct cntr_entry *entry,
-				    void *context, int vl, int mode, u64 data)
+				   void *context, int vl, int mode, u64 data)
 {
-	struct hfi1_pportdata *ppd = context;
+	struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;
+	u64 zero = 0;
+	u64 *counter;
 
-	if (vl != CNTR_INVALID_VL)
-		return 0;
+	if (vl == CNTR_INVALID_VL)
+		counter = &ppd->port_xmit_discards;
+	else if (vl >= 0 && vl < C_VL_COUNT)
+		counter = &ppd->port_xmit_discards_vl[vl];
+	else
+		counter = &zero;
 
-	return read_write_sw(ppd->dd, &ppd->port_xmit_discards, mode, data);
+	return read_write_sw(ppd->dd, counter, mode, data);
 }
 
 static u64 access_xmit_constraint_errs(const struct cntr_entry *entry,
-				     void *context, int vl, int mode, u64 data)
+				       void *context, int vl, int mode,
+				       u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 
@@ -1450,7 +1506,7 @@ static u64 access_xmit_constraint_errs(const struct cntr_entry *entry,
 }
 
 static u64 access_rcv_constraint_errs(const struct cntr_entry *entry,
-				     void *context, int vl, int mode, u64 data)
+				      void *context, int vl, int mode, u64 data)
 {
 	struct hfi1_pportdata *ppd = context;
 
@@ -1475,7 +1531,6 @@ static u64 read_write_cpu(struct hfi1_devdata *dd, u64 *z_val,
 			  u64 __percpu *cntr,
 			  int vl, int mode, u64 data)
 {
-
 	u64 ret = 0;
 
 	if (vl != CNTR_INVALID_VL)
@@ -1507,7 +1562,7 @@ static u64 access_sw_cpu_intr(const struct cntr_entry *entry,
 }
 
 static u64 access_sw_cpu_rcv_limit(const struct cntr_entry *entry,
-			      void *context, int vl, int mode, u64 data)
+				   void *context, int vl, int mode, u64 data)
 {
 	struct hfi1_devdata *dd = context;
 
@@ -1523,6 +1578,14 @@ static u64 access_sw_pio_wait(const struct cntr_entry *entry,
 	return dd->verbs_dev.n_piowait;
 }
 
+static u64 access_sw_pio_drain(const struct cntr_entry *entry,
+			       void *context, int vl, int mode, u64 data)
+{
+	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
+
+	return dd->verbs_dev.n_piodrain;
+}
+
 static u64 access_sw_vtx_wait(const struct cntr_entry *entry,
 			      void *context, int vl, int mode, u64 data)
 {
@@ -1540,11 +1603,12 @@ static u64 access_sw_kmem_wait(const struct cntr_entry *entry,
 }
 
 static u64 access_sw_send_schedule(const struct cntr_entry *entry,
-			       void *context, int vl, int mode, u64 data)
+				   void *context, int vl, int mode, u64 data)
 {
 	struct hfi1_devdata *dd = (struct hfi1_devdata *)context;
 
-	return dd->verbs_dev.n_send_schedule;
+	return read_write_cpu(dd, &dd->z_send_schedule, dd->send_schedule, vl,
+			      mode, data);
 }
 
 /* Software counters for the error status bits within MISC_ERR_STATUS */
@@ -3882,8 +3946,8 @@ static u64 access_sw_cpu_##cntr(const struct cntr_entry *entry,		      \
 			      void *context, int vl, int mode, u64 data)      \
 {									      \
 	struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context;	      \
-	return read_write_cpu(ppd->dd, &ppd->ibport_data.z_ ##cntr,	      \
-			      ppd->ibport_data.cntr, vl,		      \
+	return read_write_cpu(ppd->dd, &ppd->ibport_data.rvp.z_ ##cntr,	      \
+			      ppd->ibport_data.rvp.cntr, vl,		      \
 			      mode, data);				      \
 }
 
@@ -3900,7 +3964,7 @@ static u64 access_ibp_##cntr(const struct cntr_entry *entry,		      \
 	if (vl != CNTR_INVALID_VL)					      \
 		return 0;						      \
 									      \
-	return read_write_sw(ppd->dd, &ppd->ibport_data.n_ ##cntr,	      \
+	return read_write_sw(ppd->dd, &ppd->ibport_data.rvp.n_ ##cntr,	      \
 			     mode, data);				      \
 }
 
@@ -4063,10 +4127,28 @@ static struct cntr_entry dev_cntrs[DEV_CNTR_LAST] = {
 			    access_sw_vtx_wait),
 [C_SW_PIO_WAIT] = CNTR_ELEM("PioWait", 0, 0, CNTR_NORMAL,
 			    access_sw_pio_wait),
+[C_SW_PIO_DRAIN] = CNTR_ELEM("PioDrain", 0, 0, CNTR_NORMAL,
+			    access_sw_pio_drain),
 [C_SW_KMEM_WAIT] = CNTR_ELEM("KmemWait", 0, 0, CNTR_NORMAL,
 			    access_sw_kmem_wait),
 [C_SW_SEND_SCHED] = CNTR_ELEM("SendSched", 0, 0, CNTR_NORMAL,
 			    access_sw_send_schedule),
+[C_SDMA_DESC_FETCHED_CNT] = CNTR_ELEM("SDEDscFdCn",
+				      SEND_DMA_DESC_FETCHED_CNT, 0,
+				      CNTR_NORMAL | CNTR_32BIT | CNTR_SDMA,
+				      dev_access_u32_csr),
+[C_SDMA_INT_CNT] = CNTR_ELEM("SDMAInt", 0, 0,
+			     CNTR_NORMAL | CNTR_32BIT | CNTR_SDMA,
+			     access_sde_int_cnt),
+[C_SDMA_ERR_CNT] = CNTR_ELEM("SDMAErrCt", 0, 0,
+			     CNTR_NORMAL | CNTR_32BIT | CNTR_SDMA,
+			     access_sde_err_cnt),
+[C_SDMA_IDLE_INT_CNT] = CNTR_ELEM("SDMAIdInt", 0, 0,
+				  CNTR_NORMAL | CNTR_32BIT | CNTR_SDMA,
+				  access_sde_idle_int_cnt),
+[C_SDMA_PROGRESS_INT_CNT] = CNTR_ELEM("SDMAPrIntCn", 0, 0,
+				      CNTR_NORMAL | CNTR_32BIT | CNTR_SDMA,
+				      access_sde_progress_int_cnt),
 /* MISC_ERR_STATUS */
 [C_MISC_PLL_LOCK_FAIL_ERR] = CNTR_ELEM("MISC_PLL_LOCK_FAIL_ERR", 0, 0,
 				CNTR_NORMAL,
@@ -4876,28 +4958,28 @@ static struct cntr_entry port_cntrs[PORT_CNTR_LAST] = {
 [C_TX_WORDS] = TXE64_PORT_CNTR_ELEM(TxWords, SEND_DWORD_CNT, CNTR_NORMAL),
 [C_TX_WAIT] = TXE64_PORT_CNTR_ELEM(TxWait, SEND_WAIT_CNT, CNTR_SYNTH),
 [C_TX_FLIT_VL] = TXE64_PORT_CNTR_ELEM(TxFlitVL, SEND_DATA_VL0_CNT,
-			CNTR_SYNTH | CNTR_VL),
+				      CNTR_SYNTH | CNTR_VL),
 [C_TX_PKT_VL] = TXE64_PORT_CNTR_ELEM(TxPktVL, SEND_DATA_PKT_VL0_CNT,
-			CNTR_SYNTH | CNTR_VL),
+				     CNTR_SYNTH | CNTR_VL),
 [C_TX_WAIT_VL] = TXE64_PORT_CNTR_ELEM(TxWaitVL, SEND_WAIT_VL0_CNT,
-			CNTR_SYNTH | CNTR_VL),
+				      CNTR_SYNTH | CNTR_VL),
 [C_RX_PKT] = RXE64_PORT_CNTR_ELEM(RxPkt, RCV_DATA_PKT_CNT, CNTR_NORMAL),
 [C_RX_WORDS] = RXE64_PORT_CNTR_ELEM(RxWords, RCV_DWORD_CNT, CNTR_NORMAL),
 [C_SW_LINK_DOWN] = CNTR_ELEM("SwLinkDown", 0, 0, CNTR_SYNTH | CNTR_32BIT,
-			access_sw_link_dn_cnt),
+			     access_sw_link_dn_cnt),
 [C_SW_LINK_UP] = CNTR_ELEM("SwLinkUp", 0, 0, CNTR_SYNTH | CNTR_32BIT,
-			access_sw_link_up_cnt),
+			   access_sw_link_up_cnt),
 [C_SW_UNKNOWN_FRAME] = CNTR_ELEM("UnknownFrame", 0, 0, CNTR_NORMAL,
 				 access_sw_unknown_frame_cnt),
 [C_SW_XMIT_DSCD] = CNTR_ELEM("XmitDscd", 0, 0, CNTR_SYNTH | CNTR_32BIT,
-			access_sw_xmit_discards),
+			     access_sw_xmit_discards),
 [C_SW_XMIT_DSCD_VL] = CNTR_ELEM("XmitDscdVl", 0, 0,
-			CNTR_SYNTH | CNTR_32BIT | CNTR_VL,
-			access_sw_xmit_discards),
+				CNTR_SYNTH | CNTR_32BIT | CNTR_VL,
+				access_sw_xmit_discards),
 [C_SW_XMIT_CSTR_ERR] = CNTR_ELEM("XmitCstrErr", 0, 0, CNTR_SYNTH,
-			access_xmit_constraint_errs),
+				 access_xmit_constraint_errs),
 [C_SW_RCV_CSTR_ERR] = CNTR_ELEM("RcvCstrErr", 0, 0, CNTR_SYNTH,
-			access_rcv_constraint_errs),
+				access_rcv_constraint_errs),
 [C_SW_IBP_LOOP_PKTS] = SW_IBP_CNTR(LoopPkts, loop_pkts),
 [C_SW_IBP_RC_RESENDS] = SW_IBP_CNTR(RcResend, rc_resends),
 [C_SW_IBP_RNR_NAKS] = SW_IBP_CNTR(RnrNak, rnr_naks),
@@ -4913,9 +4995,9 @@ static struct cntr_entry port_cntrs[PORT_CNTR_LAST] = {
 [C_SW_CPU_RC_ACKS] = CNTR_ELEM("RcAcks", 0, 0, CNTR_NORMAL,
 			       access_sw_cpu_rc_acks),
 [C_SW_CPU_RC_QACKS] = CNTR_ELEM("RcQacks", 0, 0, CNTR_NORMAL,
-			       access_sw_cpu_rc_qacks),
+				access_sw_cpu_rc_qacks),
 [C_SW_CPU_RC_DELAYED_COMP] = CNTR_ELEM("RcDelayComp", 0, 0, CNTR_NORMAL,
-			       access_sw_cpu_rc_delayed_comp),
+				       access_sw_cpu_rc_delayed_comp),
 [OVR_LBL(0)] = OVR_ELM(0), [OVR_LBL(1)] = OVR_ELM(1),
 [OVR_LBL(2)] = OVR_ELM(2), [OVR_LBL(3)] = OVR_ELM(3),
 [OVR_LBL(4)] = OVR_ELM(4), [OVR_LBL(5)] = OVR_ELM(5),
@@ -5064,7 +5146,7 @@ done:
  * the buffer.  End in '*' if the buffer is too short.
  */
 static char *flag_string(char *buf, int buf_len, u64 flags,
-				struct flag_table *table, int table_size)
+			 struct flag_table *table, int table_size)
 {
 	char extra[32];
 	char *p = buf;
@@ -5125,10 +5207,8 @@ static char *is_misc_err_name(char *buf, size_t bsize, unsigned int source)
 	if (source < ARRAY_SIZE(cce_misc_names))
 		strncpy(buf, cce_misc_names[source], bsize);
 	else
-		snprintf(buf,
-			bsize,
-			"Reserved%u",
-			source + IS_GENERAL_ERR_START);
+		snprintf(buf, bsize, "Reserved%u",
+			 source + IS_GENERAL_ERR_START);
 
 	return buf;
 }
@@ -5167,7 +5247,7 @@ static char *is_various_name(char *buf, size_t bsize, unsigned int source)
 	if (source < ARRAY_SIZE(various_names))
 		strncpy(buf, various_names[source], bsize);
 	else
-		snprintf(buf, bsize, "Reserved%u", source+IS_VARIOUS_START);
+		snprintf(buf, bsize, "Reserved%u", source + IS_VARIOUS_START);
 	return buf;
 }
 
@@ -5252,51 +5332,56 @@ static char *is_reserved_name(char *buf, size_t bsize, unsigned int source)
 static char *cce_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			cce_err_status_flags, ARRAY_SIZE(cce_err_status_flags));
+			   cce_err_status_flags,
+			   ARRAY_SIZE(cce_err_status_flags));
 }
 
 static char *rxe_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			rxe_err_status_flags, ARRAY_SIZE(rxe_err_status_flags));
+			   rxe_err_status_flags,
+			   ARRAY_SIZE(rxe_err_status_flags));
 }
 
 static char *misc_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags, misc_err_status_flags,
-			ARRAY_SIZE(misc_err_status_flags));
+			   ARRAY_SIZE(misc_err_status_flags));
 }
 
 static char *pio_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			pio_err_status_flags, ARRAY_SIZE(pio_err_status_flags));
+			   pio_err_status_flags,
+			   ARRAY_SIZE(pio_err_status_flags));
 }
 
 static char *sdma_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			sdma_err_status_flags,
-			ARRAY_SIZE(sdma_err_status_flags));
+			   sdma_err_status_flags,
+			   ARRAY_SIZE(sdma_err_status_flags));
 }
 
 static char *egress_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-		egress_err_status_flags, ARRAY_SIZE(egress_err_status_flags));
+			   egress_err_status_flags,
+			   ARRAY_SIZE(egress_err_status_flags));
 }
 
 static char *egress_err_info_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-		egress_err_info_flags, ARRAY_SIZE(egress_err_info_flags));
+			   egress_err_info_flags,
+			   ARRAY_SIZE(egress_err_info_flags));
 }
 
 static char *send_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			send_err_status_flags,
-			ARRAY_SIZE(send_err_status_flags));
+			   send_err_status_flags,
+			   ARRAY_SIZE(send_err_status_flags));
 }
 
 static void handle_cce_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
@@ -5309,7 +5394,7 @@ static void handle_cce_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	 * report or record it.
 	 */
 	dd_dev_info(dd, "CCE Error: %s\n",
-		cce_err_status_string(buf, sizeof(buf), reg));
+		    cce_err_status_string(buf, sizeof(buf), reg));
 
 	if ((reg & CCE_ERR_STATUS_CCE_CLI2_ASYNC_FIFO_PARITY_ERR_SMASK) &&
 	    is_ax(dd) && (dd->icode != ICODE_FUNCTIONAL_SIMULATOR)) {
@@ -5339,14 +5424,14 @@ static void update_rcverr_timer(unsigned long opaque)
 	u32 cur_ovfl_cnt = read_dev_cntr(dd, C_RCV_OVF, CNTR_INVALID_VL);
 
 	if (dd->rcv_ovfl_cnt < cur_ovfl_cnt &&
-		ppd->port_error_action & OPA_PI_MASK_EX_BUFFER_OVERRUN) {
+	    ppd->port_error_action & OPA_PI_MASK_EX_BUFFER_OVERRUN) {
 		dd_dev_info(dd, "%s: PortErrorAction bounce\n", __func__);
-		set_link_down_reason(ppd,
-		  OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN, 0,
-			OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN);
+		set_link_down_reason(
+		ppd, OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN, 0,
+		OPA_LINKDOWN_REASON_EXCESSIVE_BUFFER_OVERRUN);
 		queue_work(ppd->hfi1_wq, &ppd->link_bounce_work);
 	}
-	dd->rcv_ovfl_cnt = (u32) cur_ovfl_cnt;
+	dd->rcv_ovfl_cnt = (u32)cur_ovfl_cnt;
 
 	mod_timer(&dd->rcverr_timer, jiffies + HZ * RCVERR_CHECK_TIME);
 }
@@ -5372,7 +5457,7 @@ static void handle_rxe_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	int i = 0;
 
 	dd_dev_info(dd, "Receive Error: %s\n",
-		rxe_err_status_string(buf, sizeof(buf), reg));
+		    rxe_err_status_string(buf, sizeof(buf), reg));
 
 	if (reg & ALL_RXE_FREEZE_ERR) {
 		int flags = 0;
@@ -5399,7 +5484,7 @@ static void handle_misc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	int i = 0;
 
 	dd_dev_info(dd, "Misc Error: %s",
-		misc_err_status_string(buf, sizeof(buf), reg));
+		    misc_err_status_string(buf, sizeof(buf), reg));
 	for (i = 0; i < NUM_MISC_ERR_STATUS_COUNTERS; i++) {
 		if (reg & (1ull << i))
 			incr_cntr64(&dd->misc_err_status_cnt[i]);
@@ -5412,7 +5497,7 @@ static void handle_pio_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	int i = 0;
 
 	dd_dev_info(dd, "PIO Error: %s\n",
-		pio_err_status_string(buf, sizeof(buf), reg));
+		    pio_err_status_string(buf, sizeof(buf), reg));
 
 	if (reg & ALL_PIO_FREEZE_ERR)
 		start_freeze_handling(dd->pport, 0);
@@ -5429,7 +5514,7 @@ static void handle_sdma_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	int i = 0;
 
 	dd_dev_info(dd, "SDMA Error: %s\n",
-		sdma_err_status_string(buf, sizeof(buf), reg));
+		    sdma_err_status_string(buf, sizeof(buf), reg));
 
 	if (reg & ALL_SDMA_FREEZE_ERR)
 		start_freeze_handling(dd->pport, 0);
@@ -5440,12 +5525,14 @@ static void handle_sdma_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	}
 }
 
-static void count_port_inactive(struct hfi1_devdata *dd)
+static inline void __count_port_discards(struct hfi1_pportdata *ppd)
 {
-	struct hfi1_pportdata *ppd = dd->pport;
+	incr_cntr64(&ppd->port_xmit_discards);
+}
 
-	if (ppd->port_xmit_discards < ~(u64)0)
-		ppd->port_xmit_discards++;
+static void count_port_inactive(struct hfi1_devdata *dd)
+{
+	__count_port_discards(dd->pport);
 }
 
 /*
@@ -5457,7 +5544,8 @@ static void count_port_inactive(struct hfi1_devdata *dd)
  * egress error if more than one packet fails the same integrity check
  * since we cleared the corresponding bit in SEND_EGRESS_ERR_INFO.
  */
-static void handle_send_egress_err_info(struct hfi1_devdata *dd)
+static void handle_send_egress_err_info(struct hfi1_devdata *dd,
+					int vl)
 {
 	struct hfi1_pportdata *ppd = dd->pport;
 	u64 src = read_csr(dd, SEND_EGRESS_ERR_SOURCE); /* read first */
@@ -5468,14 +5556,44 @@ static void handle_send_egress_err_info(struct hfi1_devdata *dd)
 	write_csr(dd, SEND_EGRESS_ERR_INFO, info);
 
 	dd_dev_info(dd,
-		"Egress Error Info: 0x%llx, %s Egress Error Src 0x%llx\n",
-		info, egress_err_info_string(buf, sizeof(buf), info), src);
+		    "Egress Error Info: 0x%llx, %s Egress Error Src 0x%llx\n",
+		    info, egress_err_info_string(buf, sizeof(buf), info), src);
 
 	/* Eventually add other counters for each bit */
+	if (info & PORT_DISCARD_EGRESS_ERRS) {
+		int weight, i;
 
-	if (info & SEND_EGRESS_ERR_INFO_TOO_LONG_IB_PACKET_ERR_SMASK) {
-		if (ppd->port_xmit_discards < ~(u64)0)
-			ppd->port_xmit_discards++;
+		/*
+		 * Count all applicable bits as individual errors and
+		 * attribute them to the packet that triggered this handler.
+		 * This may not be completely accurate due to limitations
+		 * on the available hardware error information.  There is
+		 * a single information register and any number of error
+		 * packets may have occurred and contributed to it before
+		 * this routine is called.  This means that:
+		 * a) If multiple packets with the same error occur before
+		 *    this routine is called, earlier packets are missed.
+		 *    There is only a single bit for each error type.
+		 * b) Errors may not be attributed to the correct VL.
+		 *    The driver is attributing all bits in the info register
+		 *    to the packet that triggered this call, but bits
+		 *    could be an accumulation of different packets with
+		 *    different VLs.
+		 * c) A single error packet may have multiple counts attached
+		 *    to it.  There is no way for the driver to know if
+		 *    multiple bits set in the info register are due to a
+		 *    single packet or multiple packets.  The driver assumes
+		 *    multiple packets.
+		 */
+		weight = hweight64(info & PORT_DISCARD_EGRESS_ERRS);
+		for (i = 0; i < weight; i++) {
+			__count_port_discards(ppd);
+			if (vl >= 0 && vl < TXE_NUM_DATA_VL)
+				incr_cntr64(&ppd->port_xmit_discards_vl[vl]);
+			else if (vl == 15)
+				incr_cntr64(&ppd->port_xmit_discards_vl
+					    [C_VL_15]);
+		}
 	}
 }
 
@@ -5493,12 +5611,71 @@ static inline int port_inactive_err(u64 posn)
  * Input value is a bit position within the SEND_EGRESS_ERR_STATUS
  * register. Does it represent a 'disallowed packet' error?
  */
-static inline int disallowed_pkt_err(u64 posn)
+static inline int disallowed_pkt_err(int posn)
 {
 	return (posn >= SEES(TX_SDMA0_DISALLOWED_PACKET) &&
 		posn <= SEES(TX_SDMA15_DISALLOWED_PACKET));
 }
 
+/*
+ * Input value is a bit position of one of the SDMA engine disallowed
+ * packet errors.  Return which engine.  Use of this must be guarded by
+ * disallowed_pkt_err().
+ */
+static inline int disallowed_pkt_engine(int posn)
+{
+	return posn - SEES(TX_SDMA0_DISALLOWED_PACKET);
+}
+
+/*
+ * Translate an SDMA engine to a VL.  Return -1 if the tranlation cannot
+ * be done.
+ */
+static int engine_to_vl(struct hfi1_devdata *dd, int engine)
+{
+	struct sdma_vl_map *m;
+	int vl;
+
+	/* range check */
+	if (engine < 0 || engine >= TXE_NUM_SDMA_ENGINES)
+		return -1;
+
+	rcu_read_lock();
+	m = rcu_dereference(dd->sdma_map);
+	vl = m->engine_to_vl[engine];
+	rcu_read_unlock();
+
+	return vl;
+}
+
+/*
+ * Translate the send context (sofware index) into a VL.  Return -1 if the
+ * translation cannot be done.
+ */
+static int sc_to_vl(struct hfi1_devdata *dd, int sw_index)
+{
+	struct send_context_info *sci;
+	struct send_context *sc;
+	int i;
+
+	sci = &dd->send_contexts[sw_index];
+
+	/* there is no information for user (PSM) and ack contexts */
+	if (sci->type != SC_KERNEL)
+		return -1;
+
+	sc = sci->sc;
+	if (!sc)
+		return -1;
+	if (dd->vld[15].sc == sc)
+		return 15;
+	for (i = 0; i < num_vls; i++)
+		if (dd->vld[i].sc == sc)
+			return i;
+
+	return -1;
+}
+
 static void handle_egress_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 {
 	u64 reg_copy = reg, handled = 0;
@@ -5507,34 +5684,34 @@ static void handle_egress_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 
 	if (reg & ALL_TXE_EGRESS_FREEZE_ERR)
 		start_freeze_handling(dd->pport, 0);
-	if (is_ax(dd) && (reg &
-		    SEND_EGRESS_ERR_STATUS_TX_CREDIT_RETURN_VL_ERR_SMASK)
-		    && (dd->icode != ICODE_FUNCTIONAL_SIMULATOR))
+	else if (is_ax(dd) &&
+		 (reg & SEND_EGRESS_ERR_STATUS_TX_CREDIT_RETURN_VL_ERR_SMASK) &&
+		 (dd->icode != ICODE_FUNCTIONAL_SIMULATOR))
 		start_freeze_handling(dd->pport, 0);
 
 	while (reg_copy) {
 		int posn = fls64(reg_copy);
-		/*
-		 * fls64() returns a 1-based offset, but we generally
-		 * want 0-based offsets.
-		 */
+		/* fls64() returns a 1-based offset, we want it zero based */
 		int shift = posn - 1;
+		u64 mask = 1ULL << shift;
 
 		if (port_inactive_err(shift)) {
 			count_port_inactive(dd);
-			handled |= (1ULL << shift);
+			handled |= mask;
 		} else if (disallowed_pkt_err(shift)) {
-			handle_send_egress_err_info(dd);
-			handled |= (1ULL << shift);
+			int vl = engine_to_vl(dd, disallowed_pkt_engine(shift));
+
+			handle_send_egress_err_info(dd, vl);
+			handled |= mask;
 		}
-		clear_bit(shift, (unsigned long *)&reg_copy);
+		reg_copy &= ~mask;
 	}
 
 	reg &= ~handled;
 
 	if (reg)
 		dd_dev_info(dd, "Egress Error: %s\n",
-			egress_err_status_string(buf, sizeof(buf), reg));
+			    egress_err_status_string(buf, sizeof(buf), reg));
 
 	for (i = 0; i < NUM_SEND_EGRESS_ERR_STATUS_COUNTERS; i++) {
 		if (reg & (1ull << i))
@@ -5548,7 +5725,7 @@ static void handle_txe_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	int i = 0;
 
 	dd_dev_info(dd, "Send Error: %s\n",
-		send_err_status_string(buf, sizeof(buf), reg));
+		    send_err_status_string(buf, sizeof(buf), reg));
 
 	for (i = 0; i < NUM_SEND_ERR_STATUS_COUNTERS; i++) {
 		if (reg & (1ull << i))
@@ -5594,7 +5771,7 @@ static void interrupt_clear_down(struct hfi1_devdata *dd,
 			u64 mask;
 
 			dd_dev_err(dd, "Repeating %s bits 0x%llx - masking\n",
-				eri->desc, reg);
+				   eri->desc, reg);
 			/*
 			 * Read-modify-write so any other masked bits
 			 * remain masked.
@@ -5618,14 +5795,15 @@ static void is_misc_err_int(struct hfi1_devdata *dd, unsigned int source)
 		interrupt_clear_down(dd, 0, eri);
 	} else {
 		dd_dev_err(dd, "Unexpected misc interrupt (%u) - reserved\n",
-			source);
+			   source);
 	}
 }
 
 static char *send_context_err_status_string(char *buf, int buf_len, u64 flags)
 {
 	return flag_string(buf, buf_len, flags,
-			sc_err_status_flags, ARRAY_SIZE(sc_err_status_flags));
+			   sc_err_status_flags,
+			   ARRAY_SIZE(sc_err_status_flags));
 }
 
 /*
@@ -5650,15 +5828,15 @@ static void is_sendctxt_err_int(struct hfi1_devdata *dd,
 	sw_index = dd->hw_to_sw[hw_context];
 	if (sw_index >= dd->num_send_contexts) {
 		dd_dev_err(dd,
-			"out of range sw index %u for send context %u\n",
-			sw_index, hw_context);
+			   "out of range sw index %u for send context %u\n",
+			   sw_index, hw_context);
 		return;
 	}
 	sci = &dd->send_contexts[sw_index];
 	sc = sci->sc;
 	if (!sc) {
 		dd_dev_err(dd, "%s: context %u(%u): no sc?\n", __func__,
-			sw_index, hw_context);
+			   sw_index, hw_context);
 		return;
 	}
 
@@ -5668,10 +5846,11 @@ static void is_sendctxt_err_int(struct hfi1_devdata *dd,
 	status = read_kctxt_csr(dd, hw_context, SEND_CTXT_ERR_STATUS);
 
 	dd_dev_info(dd, "Send Context %u(%u) Error: %s\n", sw_index, hw_context,
-		send_context_err_status_string(flags, sizeof(flags), status));
+		    send_context_err_status_string(flags, sizeof(flags),
+						   status));
 
 	if (status & SEND_CTXT_ERR_STATUS_PIO_DISALLOWED_PACKET_ERR_SMASK)
-		handle_send_egress_err_info(dd);
+		handle_send_egress_err_info(dd, sc_to_vl(dd, sw_index));
 
 	/*
 	 * Automatically restart halted kernel contexts out of interrupt
@@ -5704,6 +5883,7 @@ static void handle_sdma_eng_err(struct hfi1_devdata *dd,
 	dd_dev_err(sde->dd, "CONFIG SDMA(%u) source: %u status 0x%llx\n",
 		   sde->this_idx, source, (unsigned long long)status);
 #endif
+	sde->err_cnt++;
 	sdma_engine_error(sde, status);
 
 	/*
@@ -5752,23 +5932,22 @@ static void is_various_int(struct hfi1_devdata *dd, unsigned int source)
 		interrupt_clear_down(dd, 0, eri);
 	else
 		dd_dev_info(dd,
-			"%s: Unimplemented/reserved interrupt %d\n",
-			__func__, source);
+			    "%s: Unimplemented/reserved interrupt %d\n",
+			    __func__, source);
 }
 
 static void handle_qsfp_int(struct hfi1_devdata *dd, u32 src_ctx, u64 reg)
 {
-	/* source is always zero */
+	/* src_ctx is always zero */
 	struct hfi1_pportdata *ppd = dd->pport;
 	unsigned long flags;
 	u64 qsfp_int_mgmt = (u64)(QSFP_HFI0_INT_N | QSFP_HFI0_MODPRST_N);
 
 	if (reg & QSFP_HFI0_MODPRST_N) {
-
-		dd_dev_info(dd, "%s: ModPresent triggered QSFP interrupt\n",
-				__func__);
-
 		if (!qsfp_mod_present(ppd)) {
+			dd_dev_info(dd, "%s: QSFP module removed\n",
+				    __func__);
+
 			ppd->driver_link_ready = 0;
 			/*
 			 * Cable removed, reset all our information about the
@@ -5781,14 +5960,23 @@ static void handle_qsfp_int(struct hfi1_devdata *dd, u32 src_ctx, u64 reg)
 			 * an interrupt when a cable is inserted
 			 */
 			ppd->qsfp_info.cache_valid = 0;
-			ppd->qsfp_info.qsfp_interrupt_functional = 0;
+			ppd->qsfp_info.reset_needed = 0;
+			ppd->qsfp_info.limiting_active = 0;
 			spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
-						flags);
-			write_csr(dd,
-					dd->hfi1_id ?
-						ASIC_QSFP2_INVERT :
-						ASIC_QSFP1_INVERT,
-				qsfp_int_mgmt);
+					       flags);
+			/* Invert the ModPresent pin now to detect plug-in */
+			write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_INVERT :
+				  ASIC_QSFP1_INVERT, qsfp_int_mgmt);
+
+			if ((ppd->offline_disabled_reason >
+			  HFI1_ODR_MASK(
+			  OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED)) ||
+			  (ppd->offline_disabled_reason ==
+			  HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE)))
+				ppd->offline_disabled_reason =
+				HFI1_ODR_MASK(
+				OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED);
+
 			if (ppd->host_link_state == HLS_DN_POLL) {
 				/*
 				 * The link is still in POLL. This means
@@ -5799,28 +5987,33 @@ static void handle_qsfp_int(struct hfi1_devdata *dd, u32 src_ctx, u64 reg)
 				queue_work(ppd->hfi1_wq, &ppd->link_down_work);
 			}
 		} else {
+			dd_dev_info(dd, "%s: QSFP module inserted\n",
+				    __func__);
+
 			spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
 			ppd->qsfp_info.cache_valid = 0;
 			ppd->qsfp_info.cache_refresh_required = 1;
 			spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
-						flags);
+					       flags);
 
+			/*
+			 * Stop inversion of ModPresent pin to detect
+			 * removal of the cable
+			 */
 			qsfp_int_mgmt &= ~(u64)QSFP_HFI0_MODPRST_N;
-			write_csr(dd,
-					dd->hfi1_id ?
-						ASIC_QSFP2_INVERT :
-						ASIC_QSFP1_INVERT,
-				qsfp_int_mgmt);
+			write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_INVERT :
+				  ASIC_QSFP1_INVERT, qsfp_int_mgmt);
+
+			ppd->offline_disabled_reason =
+				HFI1_ODR_MASK(OPA_LINKDOWN_REASON_TRANSIENT);
 		}
 	}
 
 	if (reg & QSFP_HFI0_INT_N) {
-
-		dd_dev_info(dd, "%s: IntN triggered QSFP interrupt\n",
-				__func__);
+		dd_dev_info(dd, "%s: Interrupt received from QSFP module\n",
+			    __func__);
 		spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
 		ppd->qsfp_info.check_interrupt_flags = 1;
-		ppd->qsfp_info.qsfp_interrupt_functional = 1;
 		spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock, flags);
 	}
 
@@ -5834,11 +6027,11 @@ static int request_host_lcb_access(struct hfi1_devdata *dd)
 	int ret;
 
 	ret = do_8051_command(dd, HCMD_MISC,
-		(u64)HCMD_MISC_REQUEST_LCB_ACCESS << LOAD_DATA_FIELD_ID_SHIFT,
-		NULL);
+			      (u64)HCMD_MISC_REQUEST_LCB_ACCESS <<
+			      LOAD_DATA_FIELD_ID_SHIFT, NULL);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd, "%s: command failed with error %d\n",
-			__func__, ret);
+			   __func__, ret);
 	}
 	return ret == HCMD_SUCCESS ? 0 : -EBUSY;
 }
@@ -5848,11 +6041,11 @@ static int request_8051_lcb_access(struct hfi1_devdata *dd)
 	int ret;
 
 	ret = do_8051_command(dd, HCMD_MISC,
-		(u64)HCMD_MISC_GRANT_LCB_ACCESS << LOAD_DATA_FIELD_ID_SHIFT,
-		NULL);
+			      (u64)HCMD_MISC_GRANT_LCB_ACCESS <<
+			      LOAD_DATA_FIELD_ID_SHIFT, NULL);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd, "%s: command failed with error %d\n",
-			__func__, ret);
+			   __func__, ret);
 	}
 	return ret == HCMD_SUCCESS ? 0 : -EBUSY;
 }
@@ -5864,8 +6057,8 @@ static int request_8051_lcb_access(struct hfi1_devdata *dd)
 static inline void set_host_lcb_access(struct hfi1_devdata *dd)
 {
 	write_csr(dd, DC_DC8051_CFG_CSR_ACCESS_SEL,
-				DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK
-				| DC_DC8051_CFG_CSR_ACCESS_SEL_LCB_SMASK);
+		  DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK |
+		  DC_DC8051_CFG_CSR_ACCESS_SEL_LCB_SMASK);
 }
 
 /*
@@ -5875,7 +6068,7 @@ static inline void set_host_lcb_access(struct hfi1_devdata *dd)
 static inline void set_8051_lcb_access(struct hfi1_devdata *dd)
 {
 	write_csr(dd, DC_DC8051_CFG_CSR_ACCESS_SEL,
-				DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK);
+		  DC_DC8051_CFG_CSR_ACCESS_SEL_DCC_SMASK);
 }
 
 /*
@@ -5909,7 +6102,7 @@ int acquire_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
 	/* this access is valid only when the link is up */
 	if ((ppd->host_link_state & HLS_UP) == 0) {
 		dd_dev_info(dd, "%s: link state %s not up\n",
-			__func__, link_state_name(ppd->host_link_state));
+			    __func__, link_state_name(ppd->host_link_state));
 		ret = -EBUSY;
 		goto done;
 	}
@@ -5918,8 +6111,8 @@ int acquire_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
 		ret = request_host_lcb_access(dd);
 		if (ret) {
 			dd_dev_err(dd,
-				"%s: unable to acquire LCB access, err %d\n",
-				__func__, ret);
+				   "%s: unable to acquire LCB access, err %d\n",
+				   __func__, ret);
 			goto done;
 		}
 		set_host_lcb_access(dd);
@@ -5956,7 +6149,7 @@ int release_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
 
 	if (dd->lcb_access_count == 0) {
 		dd_dev_err(dd, "%s: LCB access count is zero.  Skipping.\n",
-			__func__);
+			   __func__);
 		goto done;
 	}
 
@@ -5965,8 +6158,8 @@ int release_lcb_access(struct hfi1_devdata *dd, int sleep_ok)
 		ret = request_8051_lcb_access(dd);
 		if (ret) {
 			dd_dev_err(dd,
-				"%s: unable to release LCB access, err %d\n",
-				__func__, ret);
+				   "%s: unable to release LCB access, err %d\n",
+				   __func__, ret);
 			/* restore host access if the grant didn't work */
 			set_host_lcb_access(dd);
 			goto done;
@@ -5998,19 +6191,26 @@ static void init_lcb_access(struct hfi1_devdata *dd)
 static void hreq_response(struct hfi1_devdata *dd, u8 return_code, u16 rsp_data)
 {
 	write_csr(dd, DC_DC8051_CFG_EXT_DEV_0,
-		DC_DC8051_CFG_EXT_DEV_0_COMPLETED_SMASK
-		| (u64)return_code << DC_DC8051_CFG_EXT_DEV_0_RETURN_CODE_SHIFT
-		| (u64)rsp_data << DC_DC8051_CFG_EXT_DEV_0_RSP_DATA_SHIFT);
+		  DC_DC8051_CFG_EXT_DEV_0_COMPLETED_SMASK |
+		  (u64)return_code <<
+		  DC_DC8051_CFG_EXT_DEV_0_RETURN_CODE_SHIFT |
+		  (u64)rsp_data << DC_DC8051_CFG_EXT_DEV_0_RSP_DATA_SHIFT);
 }
 
 /*
- * Handle requests from the 8051.
+ * Handle host requests from the 8051.
+ *
+ * This is a work-queue function outside of the interrupt.
  */
-static void handle_8051_request(struct hfi1_devdata *dd)
+void handle_8051_request(struct work_struct *work)
 {
+	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+							dc_host_req_work);
+	struct hfi1_devdata *dd = ppd->dd;
 	u64 reg;
-	u16 data;
-	u8 type;
+	u16 data = 0;
+	u8 type, i, lanes, *cache = ppd->qsfp_info.cache;
+	u8 cdr_ctrl_byte = cache[QSFP_CDR_CTRL_BYTE_OFFS];
 
 	reg = read_csr(dd, DC_DC8051_CFG_EXT_DEV_1);
 	if ((reg & DC_DC8051_CFG_EXT_DEV_1_REQ_NEW_SMASK) == 0)
@@ -6031,12 +6231,46 @@ static void handle_8051_request(struct hfi1_devdata *dd)
 	case HREQ_READ_CONFIG:
 	case HREQ_SET_TX_EQ_ABS:
 	case HREQ_SET_TX_EQ_REL:
-	case HREQ_ENABLE:
 		dd_dev_info(dd, "8051 request: request 0x%x not supported\n",
-			type);
+			    type);
 		hreq_response(dd, HREQ_NOT_SUPPORTED, 0);
 		break;
 
+	case HREQ_ENABLE:
+		lanes = data & 0xF;
+		for (i = 0; lanes; lanes >>= 1, i++) {
+			if (!(lanes & 1))
+				continue;
+			if (data & 0x200) {
+				/* enable TX CDR */
+				if (cache[QSFP_MOD_PWR_OFFS] & 0x8 &&
+				    cache[QSFP_CDR_INFO_OFFS] & 0x80)
+					cdr_ctrl_byte |= (1 << (i + 4));
+			} else {
+				/* disable TX CDR */
+				if (cache[QSFP_MOD_PWR_OFFS] & 0x8 &&
+				    cache[QSFP_CDR_INFO_OFFS] & 0x80)
+					cdr_ctrl_byte &= ~(1 << (i + 4));
+			}
+
+			if (data & 0x800) {
+				/* enable RX CDR */
+				if (cache[QSFP_MOD_PWR_OFFS] & 0x4 &&
+				    cache[QSFP_CDR_INFO_OFFS] & 0x40)
+					cdr_ctrl_byte |= (1 << i);
+			} else {
+				/* disable RX CDR */
+				if (cache[QSFP_MOD_PWR_OFFS] & 0x4 &&
+				    cache[QSFP_CDR_INFO_OFFS] & 0x40)
+					cdr_ctrl_byte &= ~(1 << i);
+			}
+		}
+		one_qsfp_write(ppd, dd->hfi1_id, QSFP_CDR_CTRL_BYTE_OFFS,
+			       &cdr_ctrl_byte, 1);
+		hreq_response(dd, HREQ_SUCCESS, data);
+		refresh_qsfp_cache(ppd, &ppd->qsfp_info);
+		break;
+
 	case HREQ_CONFIG_DONE:
 		hreq_response(dd, HREQ_SUCCESS, 0);
 		break;
@@ -6056,11 +6290,11 @@ static void write_global_credit(struct hfi1_devdata *dd,
 				u8 vau, u16 total, u16 shared)
 {
 	write_csr(dd, SEND_CM_GLOBAL_CREDIT,
-		((u64)total
-			<< SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT)
-		| ((u64)shared
-			<< SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT)
-		| ((u64)vau << SEND_CM_GLOBAL_CREDIT_AU_SHIFT));
+		  ((u64)total <<
+		   SEND_CM_GLOBAL_CREDIT_TOTAL_CREDIT_LIMIT_SHIFT) |
+		  ((u64)shared <<
+		   SEND_CM_GLOBAL_CREDIT_SHARED_LIMIT_SHIFT) |
+		  ((u64)vau << SEND_CM_GLOBAL_CREDIT_AU_SHIFT));
 }
 
 /*
@@ -6097,7 +6331,7 @@ void reset_link_credits(struct hfi1_devdata *dd)
 
 	/* remove all previous VL credit limits */
 	for (i = 0; i < TXE_NUM_DATA_VL; i++)
-		write_csr(dd, SEND_CM_CREDIT_VL + (8*i), 0);
+		write_csr(dd, SEND_CM_CREDIT_VL + (8 * i), 0);
 	write_csr(dd, SEND_CM_CREDIT_VL15, 0);
 	write_global_credit(dd, 0, 0, 0);
 	/* reset the CM block */
@@ -6139,15 +6373,14 @@ static void lcb_shutdown(struct hfi1_devdata *dd, int abort)
 	write_csr(dd, DC_LCB_CFG_RUN, 0);
 	/* set tx fifo reset: LCB_CFG_TX_FIFOS_RESET.VAL = 1 */
 	write_csr(dd, DC_LCB_CFG_TX_FIFOS_RESET,
-		1ull << DC_LCB_CFG_TX_FIFOS_RESET_VAL_SHIFT);
+		  1ull << DC_LCB_CFG_TX_FIFOS_RESET_VAL_SHIFT);
 	/* set dcc reset csr: DCC_CFG_RESET.{reset_lcb,reset_rx_fpe} = 1 */
 	dd->lcb_err_en = read_csr(dd, DC_LCB_ERR_EN);
 	reg = read_csr(dd, DCC_CFG_RESET);
-	write_csr(dd, DCC_CFG_RESET,
-		reg
-		| (1ull << DCC_CFG_RESET_RESET_LCB_SHIFT)
-		| (1ull << DCC_CFG_RESET_RESET_RX_FPE_SHIFT));
-	(void) read_csr(dd, DCC_CFG_RESET); /* make sure the write completed */
+	write_csr(dd, DCC_CFG_RESET, reg |
+		  (1ull << DCC_CFG_RESET_RESET_LCB_SHIFT) |
+		  (1ull << DCC_CFG_RESET_RESET_RX_FPE_SHIFT));
+	(void)read_csr(dd, DCC_CFG_RESET); /* make sure the write completed */
 	if (!abort) {
 		udelay(1);    /* must hold for the longer of 16cclks or 20ns */
 		write_csr(dd, DCC_CFG_RESET, reg);
@@ -6176,14 +6409,18 @@ static void dc_shutdown(struct hfi1_devdata *dd)
 	spin_unlock_irqrestore(&dd->dc8051_lock, flags);
 	/* Shutdown the LCB */
 	lcb_shutdown(dd, 1);
-	/* Going to OFFLINE would have causes the 8051 to put the
+	/*
+	 * Going to OFFLINE would have causes the 8051 to put the
 	 * SerDes into reset already. Just need to shut down the 8051,
-	 * itself. */
+	 * itself.
+	 */
 	write_csr(dd, DC_DC8051_CFG_RST, 0x1);
 }
 
-/* Calling this after the DC has been brought out of reset should not
- * do any damage. */
+/*
+ * Calling this after the DC has been brought out of reset should not
+ * do any damage.
+ */
 static void dc_start(struct hfi1_devdata *dd)
 {
 	unsigned long flags;
@@ -6199,7 +6436,7 @@ static void dc_start(struct hfi1_devdata *dd)
 	ret = wait_fm_ready(dd, TIMEOUT_8051_START);
 	if (ret) {
 		dd_dev_err(dd, "%s: timeout starting 8051 firmware\n",
-			__func__);
+			   __func__);
 	}
 	/* Take away reset for LCB and RX FPE (set in lcb_shutdown). */
 	write_csr(dd, DCC_CFG_RESET, 0x10);
@@ -6292,7 +6529,7 @@ static void adjust_lcb_for_fpga_serdes(struct hfi1_devdata *dd)
 	write_csr(dd, DC_LCB_CFG_RX_FIFOS_RADR, rx_radr);
 	/* LCB_CFG_IGNORE_LOST_RCLK.EN = 1 */
 	write_csr(dd, DC_LCB_CFG_IGNORE_LOST_RCLK,
-		DC_LCB_CFG_IGNORE_LOST_RCLK_EN_SMASK);
+		  DC_LCB_CFG_IGNORE_LOST_RCLK_EN_SMASK);
 	write_csr(dd, DC_LCB_CFG_TX_FIFOS_RADR, tx_radr);
 }
 
@@ -6309,8 +6546,10 @@ void handle_sma_message(struct work_struct *work)
 	u64 msg;
 	int ret;
 
-	/* msg is bytes 1-4 of the 40-bit idle message - the command code
-	   is stripped off */
+	/*
+	 * msg is bytes 1-4 of the 40-bit idle message - the command code
+	 * is stripped off
+	 */
 	ret = read_idle_sma(dd, &msg);
 	if (ret)
 		return;
@@ -6336,8 +6575,8 @@ void handle_sma_message(struct work_struct *work)
 		 *
 		 * Can activate the node.  Discard otherwise.
 		 */
-		if (ppd->host_link_state == HLS_UP_ARMED
-					&& ppd->is_active_optimize_enabled) {
+		if (ppd->host_link_state == HLS_UP_ARMED &&
+		    ppd->is_active_optimize_enabled) {
 			ppd->neighbor_normal = 1;
 			ret = set_link_state(ppd, HLS_UP_ACTIVE);
 			if (ret)
@@ -6349,8 +6588,8 @@ void handle_sma_message(struct work_struct *work)
 		break;
 	default:
 		dd_dev_err(dd,
-			"%s: received unexpected SMA idle message 0x%llx\n",
-			__func__, msg);
+			   "%s: received unexpected SMA idle message 0x%llx\n",
+			   __func__, msg);
 		break;
 	}
 }
@@ -6442,10 +6681,9 @@ static void wait_for_freeze_status(struct hfi1_devdata *dd, int freeze)
 
 		if (time_after(jiffies, timeout)) {
 			dd_dev_err(dd,
-				"Time out waiting for SPC %sfreeze, bits 0x%llx, expecting 0x%llx, continuing",
-				freeze ? "" : "un",
-				reg & ALL_FROZE,
-				freeze ? ALL_FROZE : 0ull);
+				   "Time out waiting for SPC %sfreeze, bits 0x%llx, expecting 0x%llx, continuing",
+				   freeze ? "" : "un", reg & ALL_FROZE,
+				   freeze ? ALL_FROZE : 0ull);
 			return;
 		}
 		usleep_range(80, 120);
@@ -6475,11 +6713,17 @@ static void rxe_freeze(struct hfi1_devdata *dd)
  */
 static void rxe_kernel_unfreeze(struct hfi1_devdata *dd)
 {
+	u32 rcvmask;
 	int i;
 
 	/* enable all kernel contexts */
-	for (i = 0; i < dd->n_krcv_queues; i++)
-		hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB, i);
+	for (i = 0; i < dd->n_krcv_queues; i++) {
+		rcvmask = HFI1_RCVCTRL_CTXT_ENB;
+		/* HFI1_RCVCTRL_TAILUPD_[ENB|DIS] needs to be set explicitly */
+		rcvmask |= HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, DMA_RTAIL) ?
+			HFI1_RCVCTRL_TAILUPD_ENB : HFI1_RCVCTRL_TAILUPD_DIS;
+		hfi1_rcvctrl(dd, rcvmask, i);
+	}
 
 	/* enable port */
 	add_rcvctrl(dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
@@ -6564,7 +6808,7 @@ void handle_freeze(struct work_struct *work)
 void handle_link_up(struct work_struct *work)
 {
 	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
-								link_up_work);
+						  link_up_work);
 	set_link_state(ppd, HLS_UP_INIT);
 
 	/* cache the read of DC_LCB_STS_ROUND_TRIP_LTP_CNT */
@@ -6583,17 +6827,20 @@ void handle_link_up(struct work_struct *work)
 	if ((ppd->link_speed_active & ppd->link_speed_enabled) == 0) {
 		/* oops - current speed is not enabled, bounce */
 		dd_dev_err(ppd->dd,
-			"Link speed active 0x%x is outside enabled 0x%x, downing link\n",
-			ppd->link_speed_active, ppd->link_speed_enabled);
+			   "Link speed active 0x%x is outside enabled 0x%x, downing link\n",
+			   ppd->link_speed_active, ppd->link_speed_enabled);
 		set_link_down_reason(ppd, OPA_LINKDOWN_REASON_SPEED_POLICY, 0,
-			OPA_LINKDOWN_REASON_SPEED_POLICY);
+				     OPA_LINKDOWN_REASON_SPEED_POLICY);
 		set_link_state(ppd, HLS_DN_OFFLINE);
+		tune_serdes(ppd);
 		start_link(ppd);
 	}
 }
 
-/* Several pieces of LNI information were cached for SMA in ppd.
- * Reset these on link down */
+/*
+ * Several pieces of LNI information were cached for SMA in ppd.
+ * Reset these on link down
+ */
 static void reset_neighbor_info(struct hfi1_pportdata *ppd)
 {
 	ppd->neighbor_guid = 0;
@@ -6613,7 +6860,13 @@ void handle_link_down(struct work_struct *work)
 	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
 								link_down_work);
 
-	/* go offline first, then deal with reasons */
+	if ((ppd->host_link_state &
+	     (HLS_DN_POLL | HLS_VERIFY_CAP | HLS_GOING_UP)) &&
+	     ppd->port_type == PORT_TYPE_FIXED)
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NOT_INSTALLED);
+
+	/* Go offline first, then deal with reading/writing through 8051 */
 	set_link_state(ppd, HLS_DN_OFFLINE);
 
 	lcl_reason = 0;
@@ -6633,12 +6886,16 @@ void handle_link_down(struct work_struct *work)
 	/* disable the port */
 	clear_rcvctrl(ppd->dd, RCV_CTRL_RCV_PORT_ENABLE_SMASK);
 
-	/* If there is no cable attached, turn the DC off. Otherwise,
-	 * start the link bring up. */
-	if (!qsfp_mod_present(ppd))
+	/*
+	 * If there is no cable attached, turn the DC off. Otherwise,
+	 * start the link bring up.
+	 */
+	if (!qsfp_mod_present(ppd)) {
 		dc_shutdown(ppd->dd);
-	else
+	} else {
+		tune_serdes(ppd);
 		start_link(ppd);
+	}
 }
 
 void handle_link_bounce(struct work_struct *work)
@@ -6651,10 +6908,11 @@ void handle_link_bounce(struct work_struct *work)
 	 */
 	if (ppd->host_link_state & HLS_UP) {
 		set_link_state(ppd, HLS_DN_OFFLINE);
+		tune_serdes(ppd);
 		start_link(ppd);
 	} else {
 		dd_dev_info(ppd->dd, "%s: link not up (%s), nothing to do\n",
-			__func__, link_state_name(ppd->host_link_state));
+			    __func__, link_state_name(ppd->host_link_state));
 	}
 }
 
@@ -6751,7 +7009,7 @@ static u16 link_width_to_bits(struct hfi1_devdata *dd, u16 width)
 	case 3: return OPA_LINK_WIDTH_3X;
 	default:
 		dd_dev_info(dd, "%s: invalid width %d, using 4\n",
-			__func__, width);
+			    __func__, width);
 		/* fall through */
 	case 4: return OPA_LINK_WIDTH_4X;
 	}
@@ -6763,6 +7021,7 @@ static u16 link_width_to_bits(struct hfi1_devdata *dd, u16 width)
 static const u8 bit_counts[16] = {
 	0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4
 };
+
 static inline u8 nibble_to_count(u8 nibble)
 {
 	return bit_counts[nibble & 0xf];
@@ -6788,7 +7047,7 @@ static void get_link_widths(struct hfi1_devdata *dd, u16 *tx_width,
 
 	/* read the active lanes */
 	read_tx_settings(dd, &enable_lane_tx, &tx_polarity_inversion,
-				&rx_polarity_inversion, &max_rate);
+			 &rx_polarity_inversion, &max_rate);
 	read_local_lni(dd, &enable_lane_rx);
 
 	/* convert to counts */
@@ -6800,8 +7059,8 @@ static void get_link_widths(struct hfi1_devdata *dd, u16 *tx_width,
 	 * handle_verify_cap().  The ASIC 8051 firmware does not correctly
 	 * set the max_rate field in handle_verify_cap until v0.19.
 	 */
-	if ((dd->icode == ICODE_RTL_SILICON)
-				&& (dd->dc8051_ver < dc8051_ver(0, 19))) {
+	if ((dd->icode == ICODE_RTL_SILICON) &&
+	    (dd->dc8051_ver < dc8051_ver(0, 19))) {
 		/* max_rate: 0 = 12.5G, 1 = 25G */
 		switch (max_rate) {
 		case 0:
@@ -6809,8 +7068,8 @@ static void get_link_widths(struct hfi1_devdata *dd, u16 *tx_width,
 			break;
 		default:
 			dd_dev_err(dd,
-				"%s: unexpected max rate %d, using 25Gb\n",
-				__func__, (int)max_rate);
+				   "%s: unexpected max rate %d, using 25Gb\n",
+				   __func__, (int)max_rate);
 			/* fall through */
 		case 1:
 			dd->pport[0].link_speed_active = OPA_LINK_SPEED_25G;
@@ -6819,8 +7078,8 @@ static void get_link_widths(struct hfi1_devdata *dd, u16 *tx_width,
 	}
 
 	dd_dev_info(dd,
-		"Fabric active lanes (width): tx 0x%x (%d), rx 0x%x (%d)\n",
-		enable_lane_tx, tx, enable_lane_rx, rx);
+		    "Fabric active lanes (width): tx 0x%x (%d), rx 0x%x (%d)\n",
+		    enable_lane_tx, tx, enable_lane_rx, rx);
 	*tx_width = link_width_to_bits(dd, tx);
 	*rx_width = link_width_to_bits(dd, rx);
 }
@@ -6923,13 +7182,8 @@ void handle_verify_cap(struct work_struct *work)
 	 */
 
 	read_vc_remote_phy(dd, &power_management, &continious);
-	read_vc_remote_fabric(
-		dd,
-		&vau,
-		&z,
-		&vcu,
-		&vl15buf,
-		&partner_supported_crc);
+	read_vc_remote_fabric(dd, &vau, &z, &vcu, &vl15buf,
+			      &partner_supported_crc);
 	read_vc_remote_link_width(dd, &remote_tx_rate, &link_widths);
 	read_remote_device_id(dd, &device_id, &device_rev);
 	/*
@@ -6940,19 +7194,16 @@ void handle_verify_cap(struct work_struct *work)
 	/* print the active widths */
 	get_link_widths(dd, &active_tx, &active_rx);
 	dd_dev_info(dd,
-		"Peer PHY: power management 0x%x, continuous updates 0x%x\n",
-		(int)power_management, (int)continious);
+		    "Peer PHY: power management 0x%x, continuous updates 0x%x\n",
+		    (int)power_management, (int)continious);
 	dd_dev_info(dd,
-		"Peer Fabric: vAU %d, Z %d, vCU %d, vl15 credits 0x%x, CRC sizes 0x%x\n",
-		(int)vau,
-		(int)z,
-		(int)vcu,
-		(int)vl15buf,
-		(int)partner_supported_crc);
+		    "Peer Fabric: vAU %d, Z %d, vCU %d, vl15 credits 0x%x, CRC sizes 0x%x\n",
+		    (int)vau, (int)z, (int)vcu, (int)vl15buf,
+		    (int)partner_supported_crc);
 	dd_dev_info(dd, "Peer Link Width: tx rate 0x%x, widths 0x%x\n",
-		(u32)remote_tx_rate, (u32)link_widths);
+		    (u32)remote_tx_rate, (u32)link_widths);
 	dd_dev_info(dd, "Peer Device ID: 0x%04x, Revision 0x%02x\n",
-		(u32)device_id, (u32)device_rev);
+		    (u32)device_id, (u32)device_rev);
 	/*
 	 * The peer vAU value just read is the peer receiver value.  HFI does
 	 * not support a transmit vAU of 0 (AU == 8).  We advertised that
@@ -6987,10 +7238,10 @@ void handle_verify_cap(struct work_struct *work)
 	reg = read_csr(dd, SEND_CM_CTRL);
 	if (crc_val == LCB_CRC_14B && crc_14b_sideband) {
 		write_csr(dd, SEND_CM_CTRL,
-			reg | SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
+			  reg | SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
 	} else {
 		write_csr(dd, SEND_CM_CTRL,
-			reg & ~SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
+			  reg & ~SEND_CM_CTRL_FORCE_CREDIT_MODE_SMASK);
 	}
 
 	ppd->link_speed_active = 0;	/* invalid value */
@@ -7015,7 +7266,7 @@ void handle_verify_cap(struct work_struct *work)
 	}
 	if (ppd->link_speed_active == 0) {
 		dd_dev_err(dd, "%s: unexpected remote tx rate %d, using 25Gb\n",
-			__func__, (int)remote_tx_rate);
+			   __func__, (int)remote_tx_rate);
 		ppd->link_speed_active = OPA_LINK_SPEED_25G;
 	}
 
@@ -7071,9 +7322,9 @@ void handle_verify_cap(struct work_struct *work)
 		read_csr(dd, DC_DC8051_STS_REMOTE_FM_SECURITY) &
 		DC_DC8051_STS_LOCAL_FM_SECURITY_DISABLED_MASK;
 	dd_dev_info(dd,
-		"Neighbor Guid: %llx Neighbor type %d MgmtAllowed %d FM security bypass %d\n",
-		ppd->neighbor_guid, ppd->neighbor_type,
-		ppd->mgmt_allowed, ppd->neighbor_fm_security);
+		    "Neighbor Guid: %llx Neighbor type %d MgmtAllowed %d FM security bypass %d\n",
+		    ppd->neighbor_guid, ppd->neighbor_type,
+		    ppd->mgmt_allowed, ppd->neighbor_fm_security);
 	if (ppd->mgmt_allowed)
 		add_full_mgmt_pkey(ppd);
 
@@ -7127,28 +7378,27 @@ retry:
 
 		/* bounce if not at starting active width */
 		if ((ppd->link_width_active !=
-					ppd->link_width_downgrade_tx_active)
-				|| (ppd->link_width_active !=
-					ppd->link_width_downgrade_rx_active)) {
+		     ppd->link_width_downgrade_tx_active) ||
+		    (ppd->link_width_active !=
+		     ppd->link_width_downgrade_rx_active)) {
 			dd_dev_err(ppd->dd,
-				"Link downgrade is disabled and link has downgraded, downing link\n");
+				   "Link downgrade is disabled and link has downgraded, downing link\n");
 			dd_dev_err(ppd->dd,
-				"  original 0x%x, tx active 0x%x, rx active 0x%x\n",
-				ppd->link_width_active,
-				ppd->link_width_downgrade_tx_active,
-				ppd->link_width_downgrade_rx_active);
+				   "  original 0x%x, tx active 0x%x, rx active 0x%x\n",
+				   ppd->link_width_active,
+				   ppd->link_width_downgrade_tx_active,
+				   ppd->link_width_downgrade_rx_active);
 			do_bounce = 1;
 		}
-	} else if ((lwde & ppd->link_width_downgrade_tx_active) == 0
-		|| (lwde & ppd->link_width_downgrade_rx_active) == 0) {
+	} else if ((lwde & ppd->link_width_downgrade_tx_active) == 0 ||
+		   (lwde & ppd->link_width_downgrade_rx_active) == 0) {
 		/* Tx or Rx is outside the enabled policy */
 		dd_dev_err(ppd->dd,
-			"Link is outside of downgrade allowed, downing link\n");
+			   "Link is outside of downgrade allowed, downing link\n");
 		dd_dev_err(ppd->dd,
-			"  enabled 0x%x, tx active 0x%x, rx active 0x%x\n",
-			lwde,
-			ppd->link_width_downgrade_tx_active,
-			ppd->link_width_downgrade_rx_active);
+			   "  enabled 0x%x, tx active 0x%x, rx active 0x%x\n",
+			   lwde, ppd->link_width_downgrade_tx_active,
+			   ppd->link_width_downgrade_rx_active);
 		do_bounce = 1;
 	}
 
@@ -7157,8 +7407,9 @@ done:
 
 	if (do_bounce) {
 		set_link_down_reason(ppd, OPA_LINKDOWN_REASON_WIDTH_POLICY, 0,
-		  OPA_LINKDOWN_REASON_WIDTH_POLICY);
+				     OPA_LINKDOWN_REASON_WIDTH_POLICY);
 		set_link_state(ppd, HLS_DN_OFFLINE);
+		tune_serdes(ppd);
 		start_link(ppd);
 	}
 }
@@ -7239,9 +7490,10 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 			    & (HLS_DN_POLL | HLS_VERIFY_CAP | HLS_GOING_UP)) {
 				queue_link_down = 1;
 				dd_dev_info(dd, "Link error: %s\n",
-					dc8051_info_err_string(buf,
-						sizeof(buf),
-						err & FAILED_LNI));
+					    dc8051_info_err_string(buf,
+								   sizeof(buf),
+								   err &
+								   FAILED_LNI));
 			}
 			err &= ~(u64)FAILED_LNI;
 		}
@@ -7253,7 +7505,8 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 		if (err) {
 			/* report remaining errors, but do not do anything */
 			dd_dev_err(dd, "8051 info error: %s\n",
-				dc8051_info_err_string(buf, sizeof(buf), err));
+				   dc8051_info_err_string(buf, sizeof(buf),
+							  err));
 		}
 
 		/*
@@ -7281,7 +7534,7 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 			host_msg &= ~(u64)LINKUP_ACHIEVED;
 		}
 		if (host_msg & EXT_DEVICE_CFG_REQ) {
-			handle_8051_request(dd);
+			queue_work(ppd->hfi1_wq, &ppd->dc_host_req_work);
 			host_msg &= ~(u64)EXT_DEVICE_CFG_REQ;
 		}
 		if (host_msg & VERIFY_CAP_FRAME) {
@@ -7306,8 +7559,9 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 		if (host_msg) {
 			/* report remaining messages, but do not do anything */
 			dd_dev_info(dd, "8051 info host message: %s\n",
-				dc8051_info_host_msg_string(buf, sizeof(buf),
-					host_msg));
+				    dc8051_info_host_msg_string(buf,
+								sizeof(buf),
+								host_msg));
 		}
 
 		reg &= ~DC_DC8051_ERR_FLG_SET_BY_8051_SMASK;
@@ -7320,25 +7574,27 @@ static void handle_8051_interrupt(struct hfi1_devdata *dd, u32 unused, u64 reg)
 		 */
 		dd_dev_err(dd, "Lost 8051 heartbeat\n");
 		write_csr(dd, DC_DC8051_ERR_EN,
-			read_csr(dd, DC_DC8051_ERR_EN)
-			  & ~DC_DC8051_ERR_EN_LOST_8051_HEART_BEAT_SMASK);
+			  read_csr(dd, DC_DC8051_ERR_EN) &
+			  ~DC_DC8051_ERR_EN_LOST_8051_HEART_BEAT_SMASK);
 
 		reg &= ~DC_DC8051_ERR_FLG_LOST_8051_HEART_BEAT_SMASK;
 	}
 	if (reg) {
 		/* report the error, but do not do anything */
 		dd_dev_err(dd, "8051 error: %s\n",
-			dc8051_err_string(buf, sizeof(buf), reg));
+			   dc8051_err_string(buf, sizeof(buf), reg));
 	}
 
 	if (queue_link_down) {
-		/* if the link is already going down or disabled, do not
-		 * queue another */
-		if ((ppd->host_link_state
-				    & (HLS_GOING_OFFLINE|HLS_LINK_COOLDOWN))
-				|| ppd->link_enabled == 0) {
+		/*
+		 * if the link is already going down or disabled, do not
+		 * queue another
+		 */
+		if ((ppd->host_link_state &
+		    (HLS_GOING_OFFLINE | HLS_LINK_COOLDOWN)) ||
+		    ppd->link_enabled == 0) {
 			dd_dev_info(dd, "%s: not queuing link down\n",
-				__func__);
+				    __func__);
 		} else {
 			queue_work(ppd->hfi1_wq, &ppd->link_down_work);
 		}
@@ -7480,8 +7736,10 @@ static void handle_dcc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 			/* set status bit */
 			dd->err_info_rcvport.status_and_code |=
 				OPA_EI_STATUS_SMASK;
-			/* save first 2 flits in the packet that caused
-			 * the error */
+			/*
+			 * save first 2 flits in the packet that caused
+			 * the error
+			 */
 			 dd->err_info_rcvport.packet_flit1 = hdr0;
 			 dd->err_info_rcvport.packet_flit2 = hdr1;
 		}
@@ -7514,7 +7772,7 @@ static void handle_dcc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 		/* just report this */
 		dd_dev_info(dd, "DCC Error: PortRcv error: %s\n", extra);
 		dd_dev_info(dd, "           hdr0 0x%llx, hdr1 0x%llx\n",
-			hdr0, hdr1);
+			    hdr0, hdr1);
 
 		reg &= ~DCC_ERR_FLG_RCVPORT_ERR_SMASK;
 	}
@@ -7533,7 +7791,7 @@ static void handle_dcc_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	/* report any remaining errors */
 	if (reg)
 		dd_dev_info(dd, "DCC Error: %s\n",
-			dcc_err_string(buf, sizeof(buf), reg));
+			    dcc_err_string(buf, sizeof(buf), reg));
 
 	if (lcl_reason == 0)
 		lcl_reason = OPA_LINKDOWN_REASON_UNKNOWN;
@@ -7550,7 +7808,7 @@ static void handle_lcb_err(struct hfi1_devdata *dd, u32 unused, u64 reg)
 	char buf[96];
 
 	dd_dev_info(dd, "LCB Error: %s\n",
-		lcb_err_string(buf, sizeof(buf), reg));
+		    lcb_err_string(buf, sizeof(buf), reg));
 }
 
 /*
@@ -7640,7 +7898,7 @@ static void is_rcv_avail_int(struct hfi1_devdata *dd, unsigned int source)
 		err_detail = "out of range";
 	}
 	dd_dev_err(dd, "unexpected %s receive available context interrupt %u\n",
-		err_detail, source);
+		   err_detail, source);
 }
 
 /*
@@ -7666,7 +7924,7 @@ static void is_rcv_urgent_int(struct hfi1_devdata *dd, unsigned int source)
 		err_detail = "out of range";
 	}
 	dd_dev_err(dd, "unexpected %s receive urgent context interrupt %u\n",
-		err_detail, source);
+		   err_detail, source);
 }
 
 /*
@@ -7677,12 +7935,14 @@ static void is_reserved_int(struct hfi1_devdata *dd, unsigned int source)
 	char name[64];
 
 	dd_dev_err(dd, "unexpected %s interrupt\n",
-				is_reserved_name(name, sizeof(name), source));
+		   is_reserved_name(name, sizeof(name), source));
 }
 
 static const struct is_table is_table[] = {
-/* start		     end
-				name func		interrupt func */
+/*
+ * start		 end
+ *				name func		interrupt func
+ */
 { IS_GENERAL_ERR_START,  IS_GENERAL_ERR_END,
 				is_misc_err_name,	is_misc_err_int },
 { IS_SDMAENG_ERR_START,  IS_SDMAENG_ERR_END,
@@ -7753,7 +8013,7 @@ static irqreturn_t general_interrupt(int irq, void *data)
 
 	/* phase 2: call the appropriate handler */
 	for_each_set_bit(bit, (unsigned long *)&regs[0],
-						CCE_NUM_INT_CSRS*64) {
+			 CCE_NUM_INT_CSRS * 64) {
 		is_interrupt(dd, bit);
 	}
 
@@ -7776,27 +8036,27 @@ static irqreturn_t sdma_interrupt(int irq, void *data)
 
 	/* This read_csr is really bad in the hot path */
 	status = read_csr(dd,
-			CCE_INT_STATUS + (8*(IS_SDMA_START/64)))
-			& sde->imask;
+			  CCE_INT_STATUS + (8 * (IS_SDMA_START / 64)))
+			  & sde->imask;
 	if (likely(status)) {
 		/* clear the interrupt(s) */
 		write_csr(dd,
-			CCE_INT_CLEAR + (8*(IS_SDMA_START/64)),
-			status);
+			  CCE_INT_CLEAR + (8 * (IS_SDMA_START / 64)),
+			  status);
 
 		/* handle the interrupt(s) */
 		sdma_engine_interrupt(sde, status);
 	} else
 		dd_dev_err(dd, "SDMA engine %u interrupt, but no status bits set\n",
-			sde->this_idx);
+			   sde->this_idx);
 
 	return IRQ_HANDLED;
 }
 
 /*
- * Clear the receive interrupt, forcing the write and making sure
- * we have data from the chip, pushing everything in front of it
- * back to the host.
+ * Clear the receive interrupt.  Use a read of the interrupt clear CSR
+ * to insure that the write completed.  This does NOT guarantee that
+ * queued DMA writes to memory from the chip are pushed.
  */
 static inline void clear_recv_intr(struct hfi1_ctxtdata *rcd)
 {
@@ -7810,27 +8070,45 @@ static inline void clear_recv_intr(struct hfi1_ctxtdata *rcd)
 }
 
 /* force the receive interrupt */
-static inline void force_recv_intr(struct hfi1_ctxtdata *rcd)
+void force_recv_intr(struct hfi1_ctxtdata *rcd)
 {
 	write_csr(rcd->dd, CCE_INT_FORCE + (8 * rcd->ireg), rcd->imask);
 }
 
-/* return non-zero if a packet is present */
+/*
+ * Return non-zero if a packet is present.
+ *
+ * This routine is called when rechecking for packets after the RcvAvail
+ * interrupt has been cleared down.  First, do a quick check of memory for
+ * a packet present.  If not found, use an expensive CSR read of the context
+ * tail to determine the actual tail.  The CSR read is necessary because there
+ * is no method to push pending DMAs to memory other than an interrupt and we
+ * are trying to determine if we need to force an interrupt.
+ */
 static inline int check_packet_present(struct hfi1_ctxtdata *rcd)
 {
+	u32 tail;
+	int present;
+
 	if (!HFI1_CAP_IS_KSET(DMA_RTAIL))
-		return (rcd->seq_cnt ==
+		present = (rcd->seq_cnt ==
 				rhf_rcv_seq(rhf_to_cpu(get_rhf_addr(rcd))));
+	else /* is RDMA rtail */
+		present = (rcd->head != get_rcvhdrtail(rcd));
+
+	if (present)
+		return 1;
 
-	/* else is RDMA rtail */
-	return (rcd->head != get_rcvhdrtail(rcd));
+	/* fall back to a CSR read, correct indpendent of DMA_RTAIL */
+	tail = (u32)read_uctxt_csr(rcd->dd, rcd->ctxt, RCV_HDR_TAIL);
+	return rcd->head != tail;
 }
 
 /*
  * Receive packet IRQ handler.  This routine expects to be on its own IRQ.
  * This routine will try to handle packets immediately (latency), but if
  * it finds too many, it will invoke the thread handler (bandwitdh).  The
- * chip receive interupt is *not* cleared down until this or the thread (if
+ * chip receive interrupt is *not* cleared down until this or the thread (if
  * invoked) is finished.  The intent is to avoid extra interrupts while we
  * are processing packets anyway.
  */
@@ -7843,6 +8121,7 @@ static irqreturn_t receive_context_interrupt(int irq, void *data)
 
 	trace_hfi1_receive_interrupt(dd, rcd->ctxt);
 	this_cpu_inc(*dd->int_counter);
+	aspm_ctx_disable(rcd);
 
 	/* receive interrupt remains blocked while processing packets */
 	disposition = rcd->do_interrupt(rcd, 0);
@@ -7909,7 +8188,7 @@ u32 read_physical_state(struct hfi1_devdata *dd)
 				& DC_DC8051_STS_CUR_STATE_PORT_MASK;
 }
 
-static u32 read_logical_state(struct hfi1_devdata *dd)
+u32 read_logical_state(struct hfi1_devdata *dd)
 {
 	u64 reg;
 
@@ -8157,8 +8436,8 @@ static int set_physical_link_state(struct hfi1_devdata *dd, u64 state)
 	return do_8051_command(dd, HCMD_CHANGE_PHY_STATE, state, NULL);
 }
 
-static int load_8051_config(struct hfi1_devdata *dd, u8 field_id,
-			    u8 lane_id, u32 config_data)
+int load_8051_config(struct hfi1_devdata *dd, u8 field_id,
+		     u8 lane_id, u32 config_data)
 {
 	u64 data;
 	int ret;
@@ -8169,8 +8448,8 @@ static int load_8051_config(struct hfi1_devdata *dd, u8 field_id,
 	ret = do_8051_command(dd, HCMD_LOAD_CONFIG_DATA, data, NULL);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd,
-			"load 8051 config: field id %d, lane %d, err %d\n",
-			(int)field_id, (int)lane_id, ret);
+			   "load 8051 config: field id %d, lane %d, err %d\n",
+			   (int)field_id, (int)lane_id, ret);
 	}
 	return ret;
 }
@@ -8180,8 +8459,8 @@ static int load_8051_config(struct hfi1_devdata *dd, u8 field_id,
  * set the result, even on error.
  * Return 0 on success, -errno on failure
  */
-static int read_8051_config(struct hfi1_devdata *dd, u8 field_id, u8 lane_id,
-			    u32 *result)
+int read_8051_config(struct hfi1_devdata *dd, u8 field_id, u8 lane_id,
+		     u32 *result)
 {
 	u64 big_data;
 	u32 addr;
@@ -8207,7 +8486,7 @@ static int read_8051_config(struct hfi1_devdata *dd, u8 field_id, u8 lane_id,
 	} else {
 		*result = 0;
 		dd_dev_err(dd, "%s: direct read failed, lane %d, field %d!\n",
-			__func__, lane_id, field_id);
+			   __func__, lane_id, field_id);
 	}
 
 	return ret;
@@ -8244,7 +8523,7 @@ static void read_vc_local_link_width(struct hfi1_devdata *dd, u8 *misc_bits,
 	u32 frame;
 
 	read_8051_config(dd, VERIFY_CAP_LOCAL_LINK_WIDTH, GENERAL_CONFIG,
-				&frame);
+			 &frame);
 	*misc_bits = (frame >> MISC_CONFIG_BITS_SHIFT) & MISC_CONFIG_BITS_MASK;
 	*flag_bits = (frame >> LOCAL_FLAG_BITS_SHIFT) & LOCAL_FLAG_BITS_MASK;
 	*link_widths = (frame >> LINK_WIDTH_SHIFT) & LINK_WIDTH_MASK;
@@ -8326,7 +8605,7 @@ static void read_vc_remote_link_width(struct hfi1_devdata *dd,
 	u32 frame;
 
 	read_8051_config(dd, VERIFY_CAP_REMOTE_LINK_WIDTH, GENERAL_CONFIG,
-				&frame);
+			 &frame);
 	*remote_tx_rate = (frame >> REMOTE_TX_RATE_SHIFT)
 				& REMOTE_TX_RATE_MASK;
 	*link_widths = (frame >> LINK_WIDTH_SHIFT) & LINK_WIDTH_MASK;
@@ -8366,7 +8645,7 @@ void hfi1_read_link_quality(struct hfi1_devdata *dd, u8 *link_quality)
 	*link_quality = 0;
 	if (dd->pport->host_link_state & HLS_UP) {
 		ret = read_8051_config(dd, LINK_QUALITY_INFO, GENERAL_CONFIG,
-					&frame);
+				       &frame);
 		if (ret == 0)
 			*link_quality = (frame >> LINK_QUALITY_SHIFT)
 						& LINK_QUALITY_MASK;
@@ -8426,10 +8705,9 @@ static void check_fabric_firmware_versions(struct hfi1_devdata *dd)
 	for (lane = 0; lane < 4; lane++) {
 		ret = read_8051_config(dd, SPICO_FW_VERSION, lane, &frame);
 		if (ret) {
-			dd_dev_err(
-				dd,
-				"Unable to read lane %d firmware details\n",
-				lane);
+			dd_dev_err(dd,
+				   "Unable to read lane %d firmware details\n",
+				   lane);
 			continue;
 		}
 		version = (frame >> SPICO_ROM_VERSION_SHIFT)
@@ -8437,8 +8715,8 @@ static void check_fabric_firmware_versions(struct hfi1_devdata *dd)
 		prod_id = (frame >> SPICO_ROM_PROD_ID_SHIFT)
 					& SPICO_ROM_PROD_ID_MASK;
 		dd_dev_info(dd,
-			"Lane %d firmware: version 0x%04x, prod_id 0x%04x\n",
-			lane, version, prod_id);
+			    "Lane %d firmware: version 0x%04x, prod_id 0x%04x\n",
+			    lane, version, prod_id);
 	}
 }
 
@@ -8451,11 +8729,10 @@ static int read_idle_message(struct hfi1_devdata *dd, u64 type, u64 *data_out)
 {
 	int ret;
 
-	ret = do_8051_command(dd, HCMD_READ_LCB_IDLE_MSG,
-		type, data_out);
+	ret = do_8051_command(dd, HCMD_READ_LCB_IDLE_MSG, type, data_out);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd, "read idle message: type %d, err %d\n",
-			(u32)type, ret);
+			   (u32)type, ret);
 		return -EINVAL;
 	}
 	dd_dev_info(dd, "%s: read idle message 0x%llx\n", __func__, *data_out);
@@ -8472,8 +8749,8 @@ static int read_idle_message(struct hfi1_devdata *dd, u64 type, u64 *data_out)
  */
 static int read_idle_sma(struct hfi1_devdata *dd, u64 *data)
 {
-	return read_idle_message(dd,
-			(u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT, data);
+	return read_idle_message(dd, (u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT,
+				 data);
 }
 
 /*
@@ -8489,7 +8766,7 @@ static int send_idle_message(struct hfi1_devdata *dd, u64 data)
 	ret = do_8051_command(dd, HCMD_SEND_LCB_IDLE_MSG, data, NULL);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd, "send idle message: data 0x%llx, err %d\n",
-			data, ret);
+			   data, ret);
 		return -EINVAL;
 	}
 	return 0;
@@ -8504,8 +8781,8 @@ int send_idle_sma(struct hfi1_devdata *dd, u64 message)
 {
 	u64 data;
 
-	data = ((message & IDLE_PAYLOAD_MASK) << IDLE_PAYLOAD_SHIFT)
-		| ((u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT);
+	data = ((message & IDLE_PAYLOAD_MASK) << IDLE_PAYLOAD_SHIFT) |
+		((u64)IDLE_SMA << IDLE_MSG_TYPE_SHIFT);
 	return send_idle_message(dd, data);
 }
 
@@ -8527,7 +8804,7 @@ static int do_quick_linkup(struct hfi1_devdata *dd)
 		/* LCB_CFG_LOOPBACK.VAL = 2 */
 		/* LCB_CFG_LANE_WIDTH.VAL = 0 */
 		write_csr(dd, DC_LCB_CFG_LOOPBACK,
-			IB_PACKET_TYPE << DC_LCB_CFG_LOOPBACK_VAL_SHIFT);
+			  IB_PACKET_TYPE << DC_LCB_CFG_LOOPBACK_VAL_SHIFT);
 		write_csr(dd, DC_LCB_CFG_LANE_WIDTH, 0);
 	}
 
@@ -8539,25 +8816,24 @@ static int do_quick_linkup(struct hfi1_devdata *dd)
 	if (loopback && dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
 		/* LCB_CFG_RUN.EN = 1 */
 		write_csr(dd, DC_LCB_CFG_RUN,
-			1ull << DC_LCB_CFG_RUN_EN_SHIFT);
+			  1ull << DC_LCB_CFG_RUN_EN_SHIFT);
 
 		/* watch LCB_STS_LINK_TRANSFER_ACTIVE */
 		timeout = jiffies + msecs_to_jiffies(10);
 		while (1) {
-			reg = read_csr(dd,
-				DC_LCB_STS_LINK_TRANSFER_ACTIVE);
+			reg = read_csr(dd, DC_LCB_STS_LINK_TRANSFER_ACTIVE);
 			if (reg)
 				break;
 			if (time_after(jiffies, timeout)) {
 				dd_dev_err(dd,
-					"timeout waiting for LINK_TRANSFER_ACTIVE\n");
+					   "timeout waiting for LINK_TRANSFER_ACTIVE\n");
 				return -ETIMEDOUT;
 			}
 			udelay(2);
 		}
 
 		write_csr(dd, DC_LCB_CFG_ALLOW_LINK_UP,
-			1ull << DC_LCB_CFG_ALLOW_LINK_UP_VAL_SHIFT);
+			  1ull << DC_LCB_CFG_ALLOW_LINK_UP_VAL_SHIFT);
 	}
 
 	if (!loopback) {
@@ -8569,10 +8845,9 @@ static int do_quick_linkup(struct hfi1_devdata *dd)
 		 * done with LCB set up before resuming.
 		 */
 		dd_dev_err(dd,
-			"Pausing for peer to be finished with LCB set up\n");
+			   "Pausing for peer to be finished with LCB set up\n");
 		msleep(5000);
-		dd_dev_err(dd,
-			"Continuing with quick linkup\n");
+		dd_dev_err(dd, "Continuing with quick linkup\n");
 	}
 
 	write_csr(dd, DC_LCB_ERR_EN, 0); /* mask LCB errors */
@@ -8586,8 +8861,8 @@ static int do_quick_linkup(struct hfi1_devdata *dd)
 	ret = set_physical_link_state(dd, PLS_QUICK_LINKUP);
 	if (ret != HCMD_SUCCESS) {
 		dd_dev_err(dd,
-			"%s: set physical link state to quick LinkUp failed with return %d\n",
-			__func__, ret);
+			   "%s: set physical link state to quick LinkUp failed with return %d\n",
+			   __func__, ret);
 
 		set_host_lcb_access(dd);
 		write_csr(dd, DC_LCB_ERR_EN, ~0ull); /* watch LCB errors */
@@ -8612,8 +8887,8 @@ static int set_serdes_loopback_mode(struct hfi1_devdata *dd)
 	if (ret == HCMD_SUCCESS)
 		return 0;
 	dd_dev_err(dd,
-		"Set physical link state to SerDes Loopback failed with return %d\n",
-		ret);
+		   "Set physical link state to SerDes Loopback failed with return %d\n",
+		   ret);
 	if (ret >= 0)
 		ret = -EINVAL;
 	return ret;
@@ -8628,7 +8903,7 @@ static int init_loopback(struct hfi1_devdata *dd)
 
 	/* all loopbacks should disable self GUID check */
 	write_csr(dd, DC_DC8051_CFG_MODE,
-		(read_csr(dd, DC_DC8051_CFG_MODE) | DISABLE_SELF_GUID_CHECK));
+		  (read_csr(dd, DC_DC8051_CFG_MODE) | DISABLE_SELF_GUID_CHECK));
 
 	/*
 	 * The simulator has only one loopback option - LCB.  Switch
@@ -8636,10 +8911,9 @@ static int init_loopback(struct hfi1_devdata *dd)
 	 *
 	 * Accept all valid loopback values.
 	 */
-	if ((dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
-		&& (loopback == LOOPBACK_SERDES
-			|| loopback == LOOPBACK_LCB
-			|| loopback == LOOPBACK_CABLE)) {
+	if ((dd->icode == ICODE_FUNCTIONAL_SIMULATOR) &&
+	    (loopback == LOOPBACK_SERDES || loopback == LOOPBACK_LCB ||
+	     loopback == LOOPBACK_CABLE)) {
 		loopback = LOOPBACK_LCB;
 		quick_linkup = 1;
 		return 0;
@@ -8660,7 +8934,7 @@ static int init_loopback(struct hfi1_devdata *dd)
 		/* not supported in emulation due to emulation RTL changes */
 		if (dd->icode == ICODE_FPGA_EMULATION) {
 			dd_dev_err(dd,
-				"LCB loopback not supported in emulation\n");
+				   "LCB loopback not supported in emulation\n");
 			return -EINVAL;
 		}
 		return 0;
@@ -8687,10 +8961,10 @@ static u16 opa_to_vc_link_widths(u16 opa_widths)
 		u16 from;
 		u16 to;
 	} opa_link_xlate[] = {
-		{ OPA_LINK_WIDTH_1X, 1 << (1-1)  },
-		{ OPA_LINK_WIDTH_2X, 1 << (2-1)  },
-		{ OPA_LINK_WIDTH_3X, 1 << (3-1)  },
-		{ OPA_LINK_WIDTH_4X, 1 << (4-1)  },
+		{ OPA_LINK_WIDTH_1X, 1 << (1 - 1)  },
+		{ OPA_LINK_WIDTH_2X, 1 << (2 - 1)  },
+		{ OPA_LINK_WIDTH_3X, 1 << (3 - 1)  },
+		{ OPA_LINK_WIDTH_4X, 1 << (4 - 1)  },
 	};
 
 	for (i = 0; i < ARRAY_SIZE(opa_link_xlate); i++) {
@@ -8716,7 +8990,7 @@ static int set_local_link_attributes(struct hfi1_pportdata *ppd)
 
 	/* set the local tx rate - need to read-modify-write */
 	ret = read_tx_settings(dd, &enable_lane_tx, &tx_polarity_inversion,
-		&rx_polarity_inversion, &ppd->local_tx_rate);
+			       &rx_polarity_inversion, &ppd->local_tx_rate);
 	if (ret)
 		goto set_local_link_attributes_fail;
 
@@ -8737,15 +9011,16 @@ static int set_local_link_attributes(struct hfi1_pportdata *ppd)
 
 	enable_lane_tx = 0xF; /* enable all four lanes */
 	ret = write_tx_settings(dd, enable_lane_tx, tx_polarity_inversion,
-		     rx_polarity_inversion, ppd->local_tx_rate);
+				rx_polarity_inversion, ppd->local_tx_rate);
 	if (ret != HCMD_SUCCESS)
 		goto set_local_link_attributes_fail;
 
 	/*
 	 * DC supports continuous updates.
 	 */
-	ret = write_vc_local_phy(dd, 0 /* no power management */,
-				     1 /* continuous updates */);
+	ret = write_vc_local_phy(dd,
+				 0 /* no power management */,
+				 1 /* continuous updates */);
 	if (ret != HCMD_SUCCESS)
 		goto set_local_link_attributes_fail;
 
@@ -8756,7 +9031,8 @@ static int set_local_link_attributes(struct hfi1_pportdata *ppd)
 		goto set_local_link_attributes_fail;
 
 	ret = write_vc_local_link_width(dd, 0, 0,
-		     opa_to_vc_link_widths(ppd->link_width_enabled));
+					opa_to_vc_link_widths(
+						ppd->link_width_enabled));
 	if (ret != HCMD_SUCCESS)
 		goto set_local_link_attributes_fail;
 
@@ -8767,8 +9043,8 @@ static int set_local_link_attributes(struct hfi1_pportdata *ppd)
 
 set_local_link_attributes_fail:
 	dd_dev_err(dd,
-		"Failed to set local link attributes, return 0x%x\n",
-		ret);
+		   "Failed to set local link attributes, return 0x%x\n",
+		   ret);
 	return ret;
 }
 
@@ -8781,54 +9057,101 @@ int start_link(struct hfi1_pportdata *ppd)
 {
 	if (!ppd->link_enabled) {
 		dd_dev_info(ppd->dd,
-			"%s: stopping link start because link is disabled\n",
-			__func__);
+			    "%s: stopping link start because link is disabled\n",
+			    __func__);
 		return 0;
 	}
 	if (!ppd->driver_link_ready) {
 		dd_dev_info(ppd->dd,
-			"%s: stopping link start because driver is not ready\n",
-			__func__);
+			    "%s: stopping link start because driver is not ready\n",
+			    __func__);
 		return 0;
 	}
 
 	if (qsfp_mod_present(ppd) || loopback == LOOPBACK_SERDES ||
-			loopback == LOOPBACK_LCB ||
-			ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
+	    loopback == LOOPBACK_LCB ||
+	    ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
 		return set_link_state(ppd, HLS_DN_POLL);
 
 	dd_dev_info(ppd->dd,
-		"%s: stopping link start because no cable is present\n",
-		__func__);
+		    "%s: stopping link start because no cable is present\n",
+		    __func__);
 	return -EAGAIN;
 }
 
-static void reset_qsfp(struct hfi1_pportdata *ppd)
+static void wait_for_qsfp_init(struct hfi1_pportdata *ppd)
+{
+	struct hfi1_devdata *dd = ppd->dd;
+	u64 mask;
+	unsigned long timeout;
+
+	/*
+	 * Check for QSFP interrupt for t_init (SFF 8679)
+	 */
+	timeout = jiffies + msecs_to_jiffies(2000);
+	while (1) {
+		mask = read_csr(dd, dd->hfi1_id ?
+				ASIC_QSFP2_IN : ASIC_QSFP1_IN);
+		if (!(mask & QSFP_HFI0_INT_N)) {
+			write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_CLEAR :
+				  ASIC_QSFP1_CLEAR, QSFP_HFI0_INT_N);
+			break;
+		}
+		if (time_after(jiffies, timeout)) {
+			dd_dev_info(dd, "%s: No IntN detected, reset complete\n",
+				    __func__);
+			break;
+		}
+		udelay(2);
+	}
+}
+
+static void set_qsfp_int_n(struct hfi1_pportdata *ppd, u8 enable)
+{
+	struct hfi1_devdata *dd = ppd->dd;
+	u64 mask;
+
+	mask = read_csr(dd, dd->hfi1_id ? ASIC_QSFP2_MASK : ASIC_QSFP1_MASK);
+	if (enable)
+		mask |= (u64)QSFP_HFI0_INT_N;
+	else
+		mask &= ~(u64)QSFP_HFI0_INT_N;
+	write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_MASK : ASIC_QSFP1_MASK, mask);
+}
+
+void reset_qsfp(struct hfi1_pportdata *ppd)
 {
 	struct hfi1_devdata *dd = ppd->dd;
 	u64 mask, qsfp_mask;
 
+	/* Disable INT_N from triggering QSFP interrupts */
+	set_qsfp_int_n(ppd, 0);
+
+	/* Reset the QSFP */
 	mask = (u64)QSFP_HFI0_RESET_N;
-	qsfp_mask = read_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE);
+	qsfp_mask = read_csr(dd, dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE);
 	qsfp_mask |= mask;
-	write_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE,
-		qsfp_mask);
+	write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_OE : ASIC_QSFP1_OE, qsfp_mask);
 
 	qsfp_mask = read_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT);
+			     dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT);
 	qsfp_mask &= ~mask;
 	write_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT,
-		qsfp_mask);
+		  dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT, qsfp_mask);
 
 	udelay(10);
 
 	qsfp_mask |= mask;
 	write_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT,
-		qsfp_mask);
+		  dd->hfi1_id ? ASIC_QSFP2_OUT : ASIC_QSFP1_OUT, qsfp_mask);
+
+	wait_for_qsfp_init(ppd);
+
+	/*
+	 * Allow INT_N to trigger the QSFP interrupt to watch
+	 * for alarms and warnings
+	 */
+	set_qsfp_int_n(ppd, 1);
 }
 
 static int handle_qsfp_error_conditions(struct hfi1_pportdata *ppd,
@@ -8837,102 +9160,86 @@ static int handle_qsfp_error_conditions(struct hfi1_pportdata *ppd,
 	struct hfi1_devdata *dd = ppd->dd;
 
 	if ((qsfp_interrupt_status[0] & QSFP_HIGH_TEMP_ALARM) ||
-		(qsfp_interrupt_status[0] & QSFP_HIGH_TEMP_WARNING))
-		dd_dev_info(dd,
-			"%s: QSFP cable on fire\n",
-			__func__);
+	    (qsfp_interrupt_status[0] & QSFP_HIGH_TEMP_WARNING))
+		dd_dev_info(dd, "%s: QSFP cable on fire\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[0] & QSFP_LOW_TEMP_ALARM) ||
-		(qsfp_interrupt_status[0] & QSFP_LOW_TEMP_WARNING))
-		dd_dev_info(dd,
-			"%s: QSFP cable temperature too low\n",
-			__func__);
+	    (qsfp_interrupt_status[0] & QSFP_LOW_TEMP_WARNING))
+		dd_dev_info(dd, "%s: QSFP cable temperature too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[1] & QSFP_HIGH_VCC_ALARM) ||
-		(qsfp_interrupt_status[1] & QSFP_HIGH_VCC_WARNING))
-		dd_dev_info(dd,
-			"%s: QSFP supply voltage too high\n",
-			__func__);
+	    (qsfp_interrupt_status[1] & QSFP_HIGH_VCC_WARNING))
+		dd_dev_info(dd, "%s: QSFP supply voltage too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[1] & QSFP_LOW_VCC_ALARM) ||
-		(qsfp_interrupt_status[1] & QSFP_LOW_VCC_WARNING))
-		dd_dev_info(dd,
-			"%s: QSFP supply voltage too low\n",
-			__func__);
+	    (qsfp_interrupt_status[1] & QSFP_LOW_VCC_WARNING))
+		dd_dev_info(dd, "%s: QSFP supply voltage too low\n",
+			    __func__);
 
 	/* Byte 2 is vendor specific */
 
 	if ((qsfp_interrupt_status[3] & QSFP_HIGH_POWER_ALARM) ||
-		(qsfp_interrupt_status[3] & QSFP_HIGH_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable RX channel 1/2 power too high\n",
-			__func__);
+	    (qsfp_interrupt_status[3] & QSFP_HIGH_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable RX channel 1/2 power too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[3] & QSFP_LOW_POWER_ALARM) ||
-		(qsfp_interrupt_status[3] & QSFP_LOW_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable RX channel 1/2 power too low\n",
-			__func__);
+	    (qsfp_interrupt_status[3] & QSFP_LOW_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable RX channel 1/2 power too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[4] & QSFP_HIGH_POWER_ALARM) ||
-		(qsfp_interrupt_status[4] & QSFP_HIGH_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable RX channel 3/4 power too high\n",
-			__func__);
+	    (qsfp_interrupt_status[4] & QSFP_HIGH_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable RX channel 3/4 power too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[4] & QSFP_LOW_POWER_ALARM) ||
-		(qsfp_interrupt_status[4] & QSFP_LOW_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable RX channel 3/4 power too low\n",
-			__func__);
+	    (qsfp_interrupt_status[4] & QSFP_LOW_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable RX channel 3/4 power too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[5] & QSFP_HIGH_BIAS_ALARM) ||
-		(qsfp_interrupt_status[5] & QSFP_HIGH_BIAS_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 1/2 bias too high\n",
-			__func__);
+	    (qsfp_interrupt_status[5] & QSFP_HIGH_BIAS_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 1/2 bias too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[5] & QSFP_LOW_BIAS_ALARM) ||
-		(qsfp_interrupt_status[5] & QSFP_LOW_BIAS_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 1/2 bias too low\n",
-			__func__);
+	    (qsfp_interrupt_status[5] & QSFP_LOW_BIAS_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 1/2 bias too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[6] & QSFP_HIGH_BIAS_ALARM) ||
-		(qsfp_interrupt_status[6] & QSFP_HIGH_BIAS_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 3/4 bias too high\n",
-			__func__);
+	    (qsfp_interrupt_status[6] & QSFP_HIGH_BIAS_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 3/4 bias too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[6] & QSFP_LOW_BIAS_ALARM) ||
-		(qsfp_interrupt_status[6] & QSFP_LOW_BIAS_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 3/4 bias too low\n",
-			__func__);
+	    (qsfp_interrupt_status[6] & QSFP_LOW_BIAS_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 3/4 bias too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[7] & QSFP_HIGH_POWER_ALARM) ||
-		(qsfp_interrupt_status[7] & QSFP_HIGH_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 1/2 power too high\n",
-			__func__);
+	    (qsfp_interrupt_status[7] & QSFP_HIGH_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 1/2 power too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[7] & QSFP_LOW_POWER_ALARM) ||
-		(qsfp_interrupt_status[7] & QSFP_LOW_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 1/2 power too low\n",
-			__func__);
+	    (qsfp_interrupt_status[7] & QSFP_LOW_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 1/2 power too low\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[8] & QSFP_HIGH_POWER_ALARM) ||
-		(qsfp_interrupt_status[8] & QSFP_HIGH_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 3/4 power too high\n",
-			__func__);
+	    (qsfp_interrupt_status[8] & QSFP_HIGH_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 3/4 power too high\n",
+			    __func__);
 
 	if ((qsfp_interrupt_status[8] & QSFP_LOW_POWER_ALARM) ||
-		(qsfp_interrupt_status[8] & QSFP_LOW_POWER_WARNING))
-		dd_dev_info(dd,
-			"%s: Cable TX channel 3/4 power too low\n",
-			__func__);
+	    (qsfp_interrupt_status[8] & QSFP_LOW_POWER_WARNING))
+		dd_dev_info(dd, "%s: Cable TX channel 3/4 power too low\n",
+			    __func__);
 
 	/* Bytes 9-10 and 11-12 are reserved */
 	/* Bytes 13-15 are vendor specific */
@@ -8940,35 +9247,8 @@ static int handle_qsfp_error_conditions(struct hfi1_pportdata *ppd,
 	return 0;
 }
 
-static int do_pre_lni_host_behaviors(struct hfi1_pportdata *ppd)
-{
-	refresh_qsfp_cache(ppd, &ppd->qsfp_info);
-
-	return 0;
-}
-
-static int do_qsfp_intr_fallback(struct hfi1_pportdata *ppd)
-{
-	struct hfi1_devdata *dd = ppd->dd;
-	u8 qsfp_interrupt_status = 0;
-
-	if (qsfp_read(ppd, dd->hfi1_id, 2, &qsfp_interrupt_status, 1)
-		!= 1) {
-		dd_dev_info(dd,
-			"%s: Failed to read status of QSFP module\n",
-			__func__);
-		return -EIO;
-	}
-
-	/* We don't care about alarms & warnings with a non-functional INT_N */
-	if (!(qsfp_interrupt_status & QSFP_DATA_NOT_READY))
-		do_pre_lni_host_behaviors(ppd);
-
-	return 0;
-}
-
 /* This routine will only be scheduled if the QSFP module is present */
-static void qsfp_event(struct work_struct *work)
+void qsfp_event(struct work_struct *work)
 {
 	struct qsfp_data *qd;
 	struct hfi1_pportdata *ppd;
@@ -8990,76 +9270,75 @@ static void qsfp_event(struct work_struct *work)
 	dc_start(dd);
 
 	if (qd->cache_refresh_required) {
-		msleep(3000);
-		reset_qsfp(ppd);
+		set_qsfp_int_n(ppd, 0);
 
-		/* Check for QSFP interrupt after t_init (SFF 8679)
-		 * + extra
+		wait_for_qsfp_init(ppd);
+
+		/*
+		 * Allow INT_N to trigger the QSFP interrupt to watch
+		 * for alarms and warnings
 		 */
-		msleep(3000);
-		if (!qd->qsfp_interrupt_functional) {
-			if (do_qsfp_intr_fallback(ppd) < 0)
-				dd_dev_info(dd, "%s: QSFP fallback failed\n",
-					__func__);
-			ppd->driver_link_ready = 1;
-			start_link(ppd);
-		}
+		set_qsfp_int_n(ppd, 1);
+
+		tune_serdes(ppd);
+
+		start_link(ppd);
 	}
 
 	if (qd->check_interrupt_flags) {
 		u8 qsfp_interrupt_status[16] = {0,};
 
-		if (qsfp_read(ppd, dd->hfi1_id, 6,
-			      &qsfp_interrupt_status[0], 16) != 16) {
+		if (one_qsfp_read(ppd, dd->hfi1_id, 6,
+				  &qsfp_interrupt_status[0], 16) != 16) {
 			dd_dev_info(dd,
-				"%s: Failed to read status of QSFP module\n",
-				__func__);
+				    "%s: Failed to read status of QSFP module\n",
+				    __func__);
 		} else {
 			unsigned long flags;
-			u8 data_status;
 
+			handle_qsfp_error_conditions(
+					ppd, qsfp_interrupt_status);
 			spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
 			ppd->qsfp_info.check_interrupt_flags = 0;
 			spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock,
-								flags);
-
-			if (qsfp_read(ppd, dd->hfi1_id, 2, &data_status, 1)
-				 != 1) {
-				dd_dev_info(dd,
-				"%s: Failed to read status of QSFP module\n",
-					__func__);
-			}
-			if (!(data_status & QSFP_DATA_NOT_READY)) {
-				do_pre_lni_host_behaviors(ppd);
-				start_link(ppd);
-			} else
-				handle_qsfp_error_conditions(ppd,
-						qsfp_interrupt_status);
+					       flags);
 		}
 	}
 }
 
-void init_qsfp(struct hfi1_pportdata *ppd)
+static void init_qsfp_int(struct hfi1_devdata *dd)
 {
-	struct hfi1_devdata *dd = ppd->dd;
-	u64 qsfp_mask;
+	struct hfi1_pportdata *ppd = dd->pport;
+	u64 qsfp_mask, cce_int_mask;
+	const int qsfp1_int_smask = QSFP1_INT % 64;
+	const int qsfp2_int_smask = QSFP2_INT % 64;
 
-	if (loopback == LOOPBACK_SERDES || loopback == LOOPBACK_LCB ||
-			ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
-		ppd->driver_link_ready = 1;
-		return;
+	/*
+	 * disable QSFP1 interrupts for HFI1, QSFP2 interrupts for HFI0
+	 * Qsfp1Int and Qsfp2Int are adjacent bits in the same CSR,
+	 * therefore just one of QSFP1_INT/QSFP2_INT can be used to find
+	 * the index of the appropriate CSR in the CCEIntMask CSR array
+	 */
+	cce_int_mask = read_csr(dd, CCE_INT_MASK +
+				(8 * (QSFP1_INT / 64)));
+	if (dd->hfi1_id) {
+		cce_int_mask &= ~((u64)1 << qsfp1_int_smask);
+		write_csr(dd, CCE_INT_MASK + (8 * (QSFP1_INT / 64)),
+			  cce_int_mask);
+	} else {
+		cce_int_mask &= ~((u64)1 << qsfp2_int_smask);
+		write_csr(dd, CCE_INT_MASK + (8 * (QSFP2_INT / 64)),
+			  cce_int_mask);
 	}
 
-	ppd->qsfp_info.ppd = ppd;
-	INIT_WORK(&ppd->qsfp_info.qsfp_work, qsfp_event);
-
 	qsfp_mask = (u64)(QSFP_HFI0_INT_N | QSFP_HFI0_MODPRST_N);
 	/* Clear current status to avoid spurious interrupts */
-	write_csr(dd,
-			dd->hfi1_id ?
-				ASIC_QSFP2_CLEAR :
-				ASIC_QSFP1_CLEAR,
-		qsfp_mask);
+	write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_CLEAR : ASIC_QSFP1_CLEAR,
+		  qsfp_mask);
+	write_csr(dd, dd->hfi1_id ? ASIC_QSFP2_MASK : ASIC_QSFP1_MASK,
+		  qsfp_mask);
+
+	set_qsfp_int_n(ppd, 0);
 
 	/* Handle active low nature of INT_N and MODPRST_N pins */
 	if (qsfp_mod_present(ppd))
@@ -9067,29 +9346,6 @@ void init_qsfp(struct hfi1_pportdata *ppd)
 	write_csr(dd,
 		  dd->hfi1_id ? ASIC_QSFP2_INVERT : ASIC_QSFP1_INVERT,
 		  qsfp_mask);
-
-	/* Allow only INT_N and MODPRST_N to trigger QSFP interrupts */
-	qsfp_mask |= (u64)QSFP_HFI0_MODPRST_N;
-	write_csr(dd,
-		dd->hfi1_id ? ASIC_QSFP2_MASK : ASIC_QSFP1_MASK,
-		qsfp_mask);
-
-	if (qsfp_mod_present(ppd)) {
-		msleep(3000);
-		reset_qsfp(ppd);
-
-		/* Check for QSFP interrupt after t_init (SFF 8679)
-		 * + extra
-		 */
-		msleep(3000);
-		if (!ppd->qsfp_info.qsfp_interrupt_functional) {
-			if (do_qsfp_intr_fallback(ppd) < 0)
-				dd_dev_info(dd,
-					"%s: QSFP fallback failed\n",
-					__func__);
-			ppd->driver_link_ready = 1;
-		}
-	}
 }
 
 /*
@@ -9097,6 +9353,10 @@ void init_qsfp(struct hfi1_pportdata *ppd)
  */
 static void init_lcb(struct hfi1_devdata *dd)
 {
+	/* simulator does not correctly handle LCB cclk loopback, skip */
+	if (dd->icode == ICODE_FUNCTIONAL_SIMULATOR)
+		return;
+
 	/* the DC has been reset earlier in the driver load */
 
 	/* set LCB for cclk loopback on the port */
@@ -9125,8 +9385,6 @@ int bringup_serdes(struct hfi1_pportdata *ppd)
 		ppd->guid = guid;
 	}
 
-	/* the link defaults to enabled */
-	ppd->link_enabled = 1;
 	/* Set linkinit_reason on power up per OPA spec */
 	ppd->linkinit_reason = OPA_LINKINIT_REASON_LINKUP;
 
@@ -9139,6 +9397,12 @@ int bringup_serdes(struct hfi1_pportdata *ppd)
 			return ret;
 	}
 
+	/* tune the SERDES to a ballpark setting for
+	 * optimal signal and bit error rate
+	 * Needs to be done before starting the link
+	 */
+	tune_serdes(ppd);
+
 	return start_link(ppd);
 }
 
@@ -9156,8 +9420,10 @@ void hfi1_quiet_serdes(struct hfi1_pportdata *ppd)
 	ppd->driver_link_ready = 0;
 	ppd->link_enabled = 0;
 
+	ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_SMA_DISABLED);
 	set_link_down_reason(ppd, OPA_LINKDOWN_REASON_SMA_DISABLED, 0,
-	  OPA_LINKDOWN_REASON_SMA_DISABLED);
+			     OPA_LINKDOWN_REASON_SMA_DISABLED);
 	set_link_state(ppd, HLS_DN_OFFLINE);
 
 	/* disable the port */
@@ -9171,14 +9437,14 @@ static inline int init_cpu_counters(struct hfi1_devdata *dd)
 
 	ppd = (struct hfi1_pportdata *)(dd + 1);
 	for (i = 0; i < dd->num_pports; i++, ppd++) {
-		ppd->ibport_data.rc_acks = NULL;
-		ppd->ibport_data.rc_qacks = NULL;
-		ppd->ibport_data.rc_acks = alloc_percpu(u64);
-		ppd->ibport_data.rc_qacks = alloc_percpu(u64);
-		ppd->ibport_data.rc_delayed_comp = alloc_percpu(u64);
-		if ((ppd->ibport_data.rc_acks == NULL) ||
-		    (ppd->ibport_data.rc_delayed_comp == NULL) ||
-		    (ppd->ibport_data.rc_qacks == NULL))
+		ppd->ibport_data.rvp.rc_acks = NULL;
+		ppd->ibport_data.rvp.rc_qacks = NULL;
+		ppd->ibport_data.rvp.rc_acks = alloc_percpu(u64);
+		ppd->ibport_data.rvp.rc_qacks = alloc_percpu(u64);
+		ppd->ibport_data.rvp.rc_delayed_comp = alloc_percpu(u64);
+		if (!ppd->ibport_data.rvp.rc_acks ||
+		    !ppd->ibport_data.rvp.rc_delayed_comp ||
+		    !ppd->ibport_data.rvp.rc_qacks)
 			return -ENOMEM;
 	}
 
@@ -9213,8 +9479,8 @@ void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
 		pa = 0;
 	} else if (type > PT_INVALID) {
 		dd_dev_err(dd,
-			"unexpected receive array type %u for index %u, not handled\n",
-			type, index);
+			   "unexpected receive array type %u for index %u, not handled\n",
+			   type, index);
 		goto done;
 	}
 
@@ -9429,12 +9695,15 @@ static void set_send_length(struct hfi1_pportdata *ppd)
 	/* all kernel receive contexts have the same hdrqentsize */
 	for (i = 0; i < ppd->vls_supported; i++) {
 		sc_set_cr_threshold(dd->vld[i].sc,
-			sc_mtu_to_threshold(dd->vld[i].sc, dd->vld[i].mtu,
-				dd->rcd[0]->rcvhdrqentsize));
+				    sc_mtu_to_threshold(dd->vld[i].sc,
+							dd->vld[i].mtu,
+							dd->rcd[0]->
+							rcvhdrqentsize));
 	}
 	sc_set_cr_threshold(dd->vld[15].sc,
-		sc_mtu_to_threshold(dd->vld[15].sc, dd->vld[15].mtu,
-			dd->rcd[0]->rcvhdrqentsize));
+			    sc_mtu_to_threshold(dd->vld[15].sc,
+						dd->vld[15].mtu,
+						dd->rcd[0]->rcvhdrqentsize));
 
 	/* Adjust maximum MTU for the port in DC */
 	dcmtu = maxvlmtu == 10240 ? DCC_CFG_PORT_MTU_CAP_10240 :
@@ -9460,7 +9729,7 @@ static void set_lidlmc(struct hfi1_pportdata *ppd)
 	c1 &= ~(DCC_CFG_PORT_CONFIG1_TARGET_DLID_SMASK
 		| DCC_CFG_PORT_CONFIG1_DLID_MASK_SMASK);
 	c1 |= ((ppd->lid & DCC_CFG_PORT_CONFIG1_TARGET_DLID_MASK)
-			<< DCC_CFG_PORT_CONFIG1_TARGET_DLID_SHIFT)|
+			<< DCC_CFG_PORT_CONFIG1_TARGET_DLID_SHIFT) |
 	      ((mask & DCC_CFG_PORT_CONFIG1_DLID_MASK_MASK)
 			<< DCC_CFG_PORT_CONFIG1_DLID_MASK_SHIFT);
 	write_csr(ppd->dd, DCC_CFG_PORT_CONFIG1, c1);
@@ -9495,8 +9764,8 @@ static int wait_phy_linkstate(struct hfi1_devdata *dd, u32 state, u32 msecs)
 			break;
 		if (time_after(jiffies, timeout)) {
 			dd_dev_err(dd,
-				"timeout waiting for phy link state 0x%x, current state is 0x%x\n",
-				state, curr_state);
+				   "timeout waiting for phy link state 0x%x, current state is 0x%x\n",
+				   state, curr_state);
 			return -ETIMEDOUT;
 		}
 		usleep_range(1950, 2050); /* sleep 2ms-ish */
@@ -9539,17 +9808,18 @@ static int goto_offline(struct hfi1_pportdata *ppd, u8 rem_reason)
 
 	if (do_transition) {
 		ret = set_physical_link_state(dd,
-			PLS_OFFLINE | (rem_reason << 8));
+					      (rem_reason << 8) | PLS_OFFLINE);
 
 		if (ret != HCMD_SUCCESS) {
 			dd_dev_err(dd,
-				"Failed to transition to Offline link state, return %d\n",
-				ret);
+				   "Failed to transition to Offline link state, return %d\n",
+				   ret);
 			return -EINVAL;
 		}
-		if (ppd->offline_disabled_reason == OPA_LINKDOWN_REASON_NONE)
+		if (ppd->offline_disabled_reason ==
+				HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE))
 			ppd->offline_disabled_reason =
-			OPA_LINKDOWN_REASON_TRANSIENT;
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_TRANSIENT);
 	}
 
 	if (do_wait) {
@@ -9570,6 +9840,22 @@ static int goto_offline(struct hfi1_pportdata *ppd, u8 rem_reason)
 	write_csr(dd, DC_LCB_ERR_EN, ~0ull); /* watch LCB errors */
 	ppd->host_link_state = HLS_LINK_COOLDOWN; /* LCB access allowed */
 
+	if (ppd->port_type == PORT_TYPE_QSFP &&
+	    ppd->qsfp_info.limiting_active &&
+	    qsfp_mod_present(ppd)) {
+		int ret;
+
+		ret = acquire_chip_resource(dd, qsfp_resource(dd), QSFP_WAIT);
+		if (ret == 0) {
+			set_qsfp_tx(ppd, 0);
+			release_chip_resource(dd, qsfp_resource(dd));
+		} else {
+			/* not fatal, but should warn */
+			dd_dev_err(dd,
+				   "Unable to acquire lock to turn off QSFP TX\n");
+		}
+	}
+
 	/*
 	 * The LNI has a mandatory wait time after the physical state
 	 * moves to Offline.Quiet.  The wait time may be different
@@ -9582,7 +9868,7 @@ static int goto_offline(struct hfi1_pportdata *ppd, u8 rem_reason)
 	ret = wait_fm_ready(dd, 7000);
 	if (ret) {
 		dd_dev_err(dd,
-			"After going offline, timed out waiting for the 8051 to become ready to accept host requests\n");
+			   "After going offline, timed out waiting for the 8051 to become ready to accept host requests\n");
 		/* state is really offline, so make it so */
 		ppd->host_link_state = HLS_DN_OFFLINE;
 		return ret;
@@ -9605,8 +9891,8 @@ static int goto_offline(struct hfi1_pportdata *ppd, u8 rem_reason)
 		read_last_local_state(dd, &last_local_state);
 		read_last_remote_state(dd, &last_remote_state);
 		dd_dev_err(dd,
-			"LNI failure last states: local 0x%08x, remote 0x%08x\n",
-			last_local_state, last_remote_state);
+			   "LNI failure last states: local 0x%08x, remote 0x%08x\n",
+			   last_local_state, last_remote_state);
 	}
 
 	/* the active link width (downgrade) is 0 on link down */
@@ -9754,14 +10040,14 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 		state = dd->link_default;
 
 	/* interpret poll -> poll as a link bounce */
-	poll_bounce = ppd->host_link_state == HLS_DN_POLL
-				&& state == HLS_DN_POLL;
+	poll_bounce = ppd->host_link_state == HLS_DN_POLL &&
+		      state == HLS_DN_POLL;
 
 	dd_dev_info(dd, "%s: current %s, new %s %s%s\n", __func__,
-		link_state_name(ppd->host_link_state),
-		link_state_name(orig_new_state),
-		poll_bounce ? "(bounce) " : "",
-		link_state_reason_name(ppd, state));
+		    link_state_name(ppd->host_link_state),
+		    link_state_name(orig_new_state),
+		    poll_bounce ? "(bounce) " : "",
+		    link_state_reason_name(ppd, state));
 
 	was_up = !!(ppd->host_link_state & HLS_UP);
 
@@ -9782,8 +10068,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 
 	switch (state) {
 	case HLS_UP_INIT:
-		if (ppd->host_link_state == HLS_DN_POLL && (quick_linkup
-			    || dd->icode == ICODE_FUNCTIONAL_SIMULATOR)) {
+		if (ppd->host_link_state == HLS_DN_POLL &&
+		    (quick_linkup || dd->icode == ICODE_FUNCTIONAL_SIMULATOR)) {
 			/*
 			 * Quick link up jumps from polling to here.
 			 *
@@ -9791,7 +10077,7 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 			 * simulator jumps from polling to link up.
 			 * Accept that here.
 			 */
-			/* OK */;
+			/* OK */
 		} else if (ppd->host_link_state != HLS_GOING_UP) {
 			goto unexpected;
 		}
@@ -9802,8 +10088,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 			/* logical state didn't change, stay at going_up */
 			ppd->host_link_state = HLS_GOING_UP;
 			dd_dev_err(dd,
-				"%s: logical state did not change to INIT\n",
-				__func__);
+				   "%s: logical state did not change to INIT\n",
+				   __func__);
 		} else {
 			/* clear old transient LINKINIT_REASON code */
 			if (ppd->linkinit_reason >= OPA_LINKINIT_REASON_CLEAR)
@@ -9827,8 +10113,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 			/* logical state didn't change, stay at init */
 			ppd->host_link_state = HLS_UP_INIT;
 			dd_dev_err(dd,
-				"%s: logical state did not change to ARMED\n",
-				__func__);
+				   "%s: logical state did not change to ARMED\n",
+				   __func__);
 		}
 		/*
 		 * The simulator does not currently implement SMA messages,
@@ -9849,15 +10135,14 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 			/* logical state didn't change, stay at armed */
 			ppd->host_link_state = HLS_UP_ARMED;
 			dd_dev_err(dd,
-				"%s: logical state did not change to ACTIVE\n",
-				__func__);
+				   "%s: logical state did not change to ACTIVE\n",
+				   __func__);
 		} else {
-
 			/* tell all engines to go running */
 			sdma_all_running(dd);
 
 			/* Signal the IB layer that the port has went active */
-			event.device = &dd->verbs_dev.ibdev;
+			event.device = &dd->verbs_dev.rdi.ibdev;
 			event.element.port_num = ppd->port;
 			event.event = IB_EVENT_PORT_ACTIVE;
 		}
@@ -9884,6 +10169,7 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 				ppd->link_enabled = 1;
 		}
 
+		set_all_slowpath(ppd->dd);
 		ret = set_local_link_attributes(ppd);
 		if (ret)
 			break;
@@ -9898,12 +10184,13 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 			ret1 = set_physical_link_state(dd, PLS_POLLING);
 			if (ret1 != HCMD_SUCCESS) {
 				dd_dev_err(dd,
-					"Failed to transition to Polling link state, return 0x%x\n",
-					ret1);
+					   "Failed to transition to Polling link state, return 0x%x\n",
+					   ret1);
 				ret = -EINVAL;
 			}
 		}
-		ppd->offline_disabled_reason = OPA_LINKDOWN_REASON_NONE;
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE);
 		/*
 		 * If an error occurred above, go back to offline.  The
 		 * caller may reschedule another attempt.
@@ -9928,8 +10215,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 		ret1 = set_physical_link_state(dd, PLS_DISABLED);
 		if (ret1 != HCMD_SUCCESS) {
 			dd_dev_err(dd,
-				"Failed to transition to Disabled link state, return 0x%x\n",
-				ret1);
+				   "Failed to transition to Disabled link state, return 0x%x\n",
+				   ret1);
 			ret = -EINVAL;
 			break;
 		}
@@ -9957,8 +10244,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 		ret1 = set_physical_link_state(dd, PLS_LINKUP);
 		if (ret1 != HCMD_SUCCESS) {
 			dd_dev_err(dd,
-				"Failed to transition to link up state, return 0x%x\n",
-				ret1);
+				   "Failed to transition to link up state, return 0x%x\n",
+				   ret1);
 			ret = -EINVAL;
 			break;
 		}
@@ -9969,7 +10256,7 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 	case HLS_LINK_COOLDOWN:		/* transient within goto_offline() */
 	default:
 		dd_dev_info(dd, "%s: state 0x%x: not supported\n",
-			__func__, state);
+			    __func__, state);
 		ret = -EINVAL;
 		break;
 	}
@@ -9989,8 +10276,8 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
 
 unexpected:
 	dd_dev_err(dd, "%s: unexpected state transition from %s to %s\n",
-		__func__, link_state_name(ppd->host_link_state),
-		link_state_name(state));
+		   __func__, link_state_name(ppd->host_link_state),
+		   link_state_name(state));
 	ret = -EINVAL;
 
 done:
@@ -10016,7 +10303,7 @@ int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val)
 		 * The VL Arbitrator high limit is sent in units of 4k
 		 * bytes, while HFI stores it in units of 64 bytes.
 		 */
-		val *= 4096/64;
+		val *= 4096 / 64;
 		reg = ((u64)val & SEND_HIGH_PRIORITY_LIMIT_LIMIT_MASK)
 			<< SEND_HIGH_PRIORITY_LIMIT_LIMIT_SHIFT;
 		write_csr(ppd->dd, SEND_HIGH_PRIORITY_LIMIT, reg);
@@ -10031,12 +10318,6 @@ int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val)
 			ppd->vls_operational = val;
 			if (!ppd->port)
 				ret = -EINVAL;
-			else
-				ret = sdma_map_init(
-					ppd->dd,
-					ppd->port - 1,
-					val,
-					NULL);
 		}
 		break;
 	/*
@@ -10084,8 +10365,8 @@ int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val)
 	default:
 		if (HFI1_CAP_IS_KSET(PRINT_UNIMPL))
 			dd_dev_info(ppd->dd,
-			  "%s: which %s, val 0x%x: not implemented\n",
-			  __func__, ib_cfg_name(which), val);
+				    "%s: which %s, val 0x%x: not implemented\n",
+				    __func__, ib_cfg_name(which), val);
 		break;
 	}
 	return ret;
@@ -10152,6 +10433,7 @@ static int vl_arb_match_cache(struct vl_arb_cache *cache,
 {
 	return !memcmp(cache->table, vl, VL_ARB_TABLE_SIZE * sizeof(*vl));
 }
+
 /* end functions related to vl arbitration table caching */
 
 static int set_vl_weights(struct hfi1_pportdata *ppd, u32 target,
@@ -10239,7 +10521,7 @@ static int get_buffer_control(struct hfi1_devdata *dd,
 
 	/* OPA and HFI have a 1-1 mapping */
 	for (i = 0; i < TXE_NUM_DATA_VL; i++)
-		read_one_cm_vl(dd, SEND_CM_CREDIT_VL + (8*i), &bc->vl[i]);
+		read_one_cm_vl(dd, SEND_CM_CREDIT_VL + (8 * i), &bc->vl[i]);
 
 	/* NOTE: assumes that VL* and VL15 CSRs are bit-wise identical */
 	read_one_cm_vl(dd, SEND_CM_CREDIT_VL15, &bc->vl[15]);
@@ -10293,41 +10575,41 @@ static void get_vlarb_preempt(struct hfi1_devdata *dd, u32 nelems,
 static void set_sc2vlnt(struct hfi1_devdata *dd, struct sc2vlnt *dp)
 {
 	write_csr(dd, DCC_CFG_SC_VL_TABLE_15_0,
-		DC_SC_VL_VAL(15_0,
-		0, dp->vlnt[0] & 0xf,
-		1, dp->vlnt[1] & 0xf,
-		2, dp->vlnt[2] & 0xf,
-		3, dp->vlnt[3] & 0xf,
-		4, dp->vlnt[4] & 0xf,
-		5, dp->vlnt[5] & 0xf,
-		6, dp->vlnt[6] & 0xf,
-		7, dp->vlnt[7] & 0xf,
-		8, dp->vlnt[8] & 0xf,
-		9, dp->vlnt[9] & 0xf,
-		10, dp->vlnt[10] & 0xf,
-		11, dp->vlnt[11] & 0xf,
-		12, dp->vlnt[12] & 0xf,
-		13, dp->vlnt[13] & 0xf,
-		14, dp->vlnt[14] & 0xf,
-		15, dp->vlnt[15] & 0xf));
+		  DC_SC_VL_VAL(15_0,
+			       0, dp->vlnt[0] & 0xf,
+			       1, dp->vlnt[1] & 0xf,
+			       2, dp->vlnt[2] & 0xf,
+			       3, dp->vlnt[3] & 0xf,
+			       4, dp->vlnt[4] & 0xf,
+			       5, dp->vlnt[5] & 0xf,
+			       6, dp->vlnt[6] & 0xf,
+			       7, dp->vlnt[7] & 0xf,
+			       8, dp->vlnt[8] & 0xf,
+			       9, dp->vlnt[9] & 0xf,
+			       10, dp->vlnt[10] & 0xf,
+			       11, dp->vlnt[11] & 0xf,
+			       12, dp->vlnt[12] & 0xf,
+			       13, dp->vlnt[13] & 0xf,
+			       14, dp->vlnt[14] & 0xf,
+			       15, dp->vlnt[15] & 0xf));
 	write_csr(dd, DCC_CFG_SC_VL_TABLE_31_16,
-		DC_SC_VL_VAL(31_16,
-		16, dp->vlnt[16] & 0xf,
-		17, dp->vlnt[17] & 0xf,
-		18, dp->vlnt[18] & 0xf,
-		19, dp->vlnt[19] & 0xf,
-		20, dp->vlnt[20] & 0xf,
-		21, dp->vlnt[21] & 0xf,
-		22, dp->vlnt[22] & 0xf,
-		23, dp->vlnt[23] & 0xf,
-		24, dp->vlnt[24] & 0xf,
-		25, dp->vlnt[25] & 0xf,
-		26, dp->vlnt[26] & 0xf,
-		27, dp->vlnt[27] & 0xf,
-		28, dp->vlnt[28] & 0xf,
-		29, dp->vlnt[29] & 0xf,
-		30, dp->vlnt[30] & 0xf,
-		31, dp->vlnt[31] & 0xf));
+		  DC_SC_VL_VAL(31_16,
+			       16, dp->vlnt[16] & 0xf,
+			       17, dp->vlnt[17] & 0xf,
+			       18, dp->vlnt[18] & 0xf,
+			       19, dp->vlnt[19] & 0xf,
+			       20, dp->vlnt[20] & 0xf,
+			       21, dp->vlnt[21] & 0xf,
+			       22, dp->vlnt[22] & 0xf,
+			       23, dp->vlnt[23] & 0xf,
+			       24, dp->vlnt[24] & 0xf,
+			       25, dp->vlnt[25] & 0xf,
+			       26, dp->vlnt[26] & 0xf,
+			       27, dp->vlnt[27] & 0xf,
+			       28, dp->vlnt[28] & 0xf,
+			       29, dp->vlnt[29] & 0xf,
+			       30, dp->vlnt[30] & 0xf,
+			       31, dp->vlnt[31] & 0xf));
 }
 
 static void nonzero_msg(struct hfi1_devdata *dd, int idx, const char *what,
@@ -10335,7 +10617,7 @@ static void nonzero_msg(struct hfi1_devdata *dd, int idx, const char *what,
 {
 	if (limit != 0)
 		dd_dev_info(dd, "Invalid %s limit %d on VL %d, ignoring\n",
-			what, (int)limit, idx);
+			    what, (int)limit, idx);
 }
 
 /* change only the shared limit portion of SendCmGLobalCredit */
@@ -10413,14 +10695,14 @@ static void wait_for_vl_status_clear(struct hfi1_devdata *dd, u64 mask,
 	}
 
 	dd_dev_err(dd,
-		"%s credit change status not clearing after %dms, mask 0x%llx, not clear 0x%llx\n",
-		which, VL_STATUS_CLEAR_TIMEOUT, mask, reg);
+		   "%s credit change status not clearing after %dms, mask 0x%llx, not clear 0x%llx\n",
+		   which, VL_STATUS_CLEAR_TIMEOUT, mask, reg);
 	/*
 	 * If this occurs, it is likely there was a credit loss on the link.
 	 * The only recovery from that is a link bounce.
 	 */
 	dd_dev_err(dd,
-		"Continuing anyway.  A credit loss may occur.  Suggest a link bounce\n");
+		   "Continuing anyway.  A credit loss may occur.  Suggest a link bounce\n");
 }
 
 /*
@@ -10447,13 +10729,15 @@ static void wait_for_vl_status_clear(struct hfi1_devdata *dd, u64 mask,
  * raise = if the new limit is higher than the current value (may be changed
  *	earlier in the algorithm), set the new limit to the new value
  */
-static int set_buffer_control(struct hfi1_devdata *dd,
-			      struct buffer_control *new_bc)
+int set_buffer_control(struct hfi1_pportdata *ppd,
+		       struct buffer_control *new_bc)
 {
+	struct hfi1_devdata *dd = ppd->dd;
 	u64 changing_mask, ld_mask, stat_mask;
 	int change_count;
 	int i, use_all_mask;
 	int this_shared_changing;
+	int vl_count = 0, ret;
 	/*
 	 * A0: add the variable any_shared_limit_changing below and in the
 	 * algorithm above.  If removing A0 support, it can be removed.
@@ -10478,7 +10762,6 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 #define valid_vl(idx) ((idx) < TXE_NUM_DATA_VL || (idx) == 15)
 #define NUM_USABLE_VLS 16	/* look at VL15 and less */
 
-
 	/* find the new total credits, do sanity check on unused VLs */
 	for (i = 0; i < OPA_MAX_VLS; i++) {
 		if (valid_vl(i)) {
@@ -10486,9 +10769,9 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 			continue;
 		}
 		nonzero_msg(dd, i, "dedicated",
-			be16_to_cpu(new_bc->vl[i].dedicated));
+			    be16_to_cpu(new_bc->vl[i].dedicated));
 		nonzero_msg(dd, i, "shared",
-			be16_to_cpu(new_bc->vl[i].shared));
+			    be16_to_cpu(new_bc->vl[i].shared));
 		new_bc->vl[i].dedicated = 0;
 		new_bc->vl[i].shared = 0;
 	}
@@ -10502,8 +10785,10 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 	 */
 	memset(changing, 0, sizeof(changing));
 	memset(lowering_dedicated, 0, sizeof(lowering_dedicated));
-	/* NOTE: Assumes that the individual VL bits are adjacent and in
-	   increasing order */
+	/*
+	 * NOTE: Assumes that the individual VL bits are adjacent and in
+	 * increasing order
+	 */
 	stat_mask =
 		SEND_CM_CREDIT_USED_STATUS_VL0_RETURN_CREDIT_STATUS_SMASK;
 	changing_mask = 0;
@@ -10517,8 +10802,8 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 						!= cur_bc.vl[i].shared;
 		if (this_shared_changing)
 			any_shared_limit_changing = 1;
-		if (new_bc->vl[i].dedicated != cur_bc.vl[i].dedicated
-				|| this_shared_changing) {
+		if (new_bc->vl[i].dedicated != cur_bc.vl[i].dedicated ||
+		    this_shared_changing) {
 			changing[i] = 1;
 			changing_mask |= stat_mask;
 			change_count++;
@@ -10557,7 +10842,7 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 	}
 
 	wait_for_vl_status_clear(dd, use_all_mask ? all_mask : changing_mask,
-		"shared");
+				 "shared");
 
 	if (change_count > 0) {
 		for (i = 0; i < NUM_USABLE_VLS; i++) {
@@ -10566,7 +10851,8 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 
 			if (lowering_dedicated[i]) {
 				set_vl_dedicated(dd, i,
-					be16_to_cpu(new_bc->vl[i].dedicated));
+						 be16_to_cpu(new_bc->
+							     vl[i].dedicated));
 				cur_bc.vl[i].dedicated =
 						new_bc->vl[i].dedicated;
 			}
@@ -10582,7 +10868,8 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 			if (be16_to_cpu(new_bc->vl[i].dedicated) >
 					be16_to_cpu(cur_bc.vl[i].dedicated))
 				set_vl_dedicated(dd, i,
-					be16_to_cpu(new_bc->vl[i].dedicated));
+						 be16_to_cpu(new_bc->
+							     vl[i].dedicated));
 		}
 	}
 
@@ -10598,13 +10885,35 @@ static int set_buffer_control(struct hfi1_devdata *dd,
 
 	/* finally raise the global shared */
 	if (be16_to_cpu(new_bc->overall_shared_limit) >
-			be16_to_cpu(cur_bc.overall_shared_limit))
+	    be16_to_cpu(cur_bc.overall_shared_limit))
 		set_global_shared(dd,
-			be16_to_cpu(new_bc->overall_shared_limit));
+				  be16_to_cpu(new_bc->overall_shared_limit));
 
 	/* bracket the credit change with a total adjustment */
 	if (new_total < cur_total)
 		set_global_limit(dd, new_total);
+
+	/*
+	 * Determine the actual number of operational VLS using the number of
+	 * dedicated and shared credits for each VL.
+	 */
+	if (change_count > 0) {
+		for (i = 0; i < TXE_NUM_DATA_VL; i++)
+			if (be16_to_cpu(new_bc->vl[i].dedicated) > 0 ||
+			    be16_to_cpu(new_bc->vl[i].shared) > 0)
+				vl_count++;
+		ppd->actual_vls_operational = vl_count;
+		ret = sdma_map_init(dd, ppd->port - 1, vl_count ?
+				    ppd->actual_vls_operational :
+				    ppd->vls_operational,
+				    NULL);
+		if (ret == 0)
+			ret = pio_map_init(dd, ppd->port - 1, vl_count ?
+					   ppd->actual_vls_operational :
+					   ppd->vls_operational, NULL);
+		if (ret)
+			return ret;
+	}
 	return 0;
 }
 
@@ -10696,7 +11005,7 @@ int fm_set_table(struct hfi1_pportdata *ppd, int which, void *t)
 				     VL_ARB_LOW_PRIO_TABLE_SIZE, t);
 		break;
 	case FM_TBL_BUFFER_CONTROL:
-		ret = set_buffer_control(ppd->dd, t);
+		ret = set_buffer_control(ppd, t);
 		break;
 	case FM_TBL_SC2VLNT:
 		set_sc2vlnt(ppd->dd, t);
@@ -10846,10 +11155,13 @@ static void adjust_rcv_timeout(struct hfi1_ctxtdata *rcd, u32 npkts)
 	}
 
 	rcd->rcvavail_timeout = timeout;
-	/* timeout cannot be larger than rcv_intr_timeout_csr which has already
-	   been verified to be in range */
+	/*
+	 * timeout cannot be larger than rcv_intr_timeout_csr which has already
+	 * been verified to be in range
+	 */
 	write_kctxt_csr(dd, rcd->ctxt, RCV_AVAIL_TIME_OUT,
-		(u64)timeout << RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT);
+			(u64)timeout <<
+			RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT);
 }
 
 void update_usrhead(struct hfi1_ctxtdata *rcd, u32 hd, u32 updegr, u32 egrhd,
@@ -10915,16 +11227,16 @@ u32 hdrqempty(struct hfi1_ctxtdata *rcd)
 static u32 encoded_size(u32 size)
 {
 	switch (size) {
-	case   4*1024: return 0x1;
-	case   8*1024: return 0x2;
-	case  16*1024: return 0x3;
-	case  32*1024: return 0x4;
-	case  64*1024: return 0x5;
-	case 128*1024: return 0x6;
-	case 256*1024: return 0x7;
-	case 512*1024: return 0x8;
-	case   1*1024*1024: return 0x9;
-	case   2*1024*1024: return 0xa;
+	case   4 * 1024: return 0x1;
+	case   8 * 1024: return 0x2;
+	case  16 * 1024: return 0x3;
+	case  32 * 1024: return 0x4;
+	case  64 * 1024: return 0x5;
+	case 128 * 1024: return 0x6;
+	case 256 * 1024: return 0x7;
+	case 512 * 1024: return 0x8;
+	case   1 * 1024 * 1024: return 0x9;
+	case   2 * 1024 * 1024: return 0xa;
 	}
 	return 0x1;	/* if invalid, go with the minimum size */
 }
@@ -10943,8 +11255,8 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 
 	rcvctrl = read_kctxt_csr(dd, ctxt, RCV_CTXT_CTRL);
 	/* if the context already enabled, don't do the extra steps */
-	if ((op & HFI1_RCVCTRL_CTXT_ENB)
-			&& !(rcvctrl & RCV_CTXT_CTRL_ENABLE_SMASK)) {
+	if ((op & HFI1_RCVCTRL_CTXT_ENB) &&
+	    !(rcvctrl & RCV_CTXT_CTRL_ENABLE_SMASK)) {
 		/* reset the tail and hdr addresses, and sequence count */
 		write_kctxt_csr(dd, ctxt, RCV_HDR_ADDR,
 				rcd->rcvhdrq_phys);
@@ -11018,6 +11330,7 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 		if (dd->rcvhdrtail_dummy_physaddr) {
 			write_kctxt_csr(dd, ctxt, RCV_HDR_TAIL_ADDR,
 					dd->rcvhdrtail_dummy_physaddr);
+			/* Enabling RcvCtxtCtrl.TailUpd is intentional. */
 			rcvctrl |= RCV_CTXT_CTRL_TAIL_UPD_SMASK;
 		}
 
@@ -11029,15 +11342,20 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 		rcvctrl &= ~RCV_CTXT_CTRL_INTR_AVAIL_SMASK;
 	if (op & HFI1_RCVCTRL_TAILUPD_ENB && rcd->rcvhdrqtailaddr_phys)
 		rcvctrl |= RCV_CTXT_CTRL_TAIL_UPD_SMASK;
-	if (op & HFI1_RCVCTRL_TAILUPD_DIS)
-		rcvctrl &= ~RCV_CTXT_CTRL_TAIL_UPD_SMASK;
+	if (op & HFI1_RCVCTRL_TAILUPD_DIS) {
+		/* See comment on RcvCtxtCtrl.TailUpd above */
+		if (!(op & HFI1_RCVCTRL_CTXT_DIS))
+			rcvctrl &= ~RCV_CTXT_CTRL_TAIL_UPD_SMASK;
+	}
 	if (op & HFI1_RCVCTRL_TIDFLOW_ENB)
 		rcvctrl |= RCV_CTXT_CTRL_TID_FLOW_ENABLE_SMASK;
 	if (op & HFI1_RCVCTRL_TIDFLOW_DIS)
 		rcvctrl &= ~RCV_CTXT_CTRL_TID_FLOW_ENABLE_SMASK;
 	if (op & HFI1_RCVCTRL_ONE_PKT_EGR_ENB) {
-		/* In one-packet-per-eager mode, the size comes from
-		   the RcvArray entry. */
+		/*
+		 * In one-packet-per-eager mode, the size comes from
+		 * the RcvArray entry.
+		 */
 		rcvctrl &= ~RCV_CTXT_CTRL_EGR_BUF_SIZE_SMASK;
 		rcvctrl |= RCV_CTXT_CTRL_ONE_PACKET_PER_EGR_BUFFER_SMASK;
 	}
@@ -11056,19 +11374,19 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 	write_kctxt_csr(dd, ctxt, RCV_CTXT_CTRL, rcd->rcvctrl);
 
 	/* work around sticky RcvCtxtStatus.BlockedRHQFull */
-	if (did_enable
-	    && (rcvctrl & RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK)) {
+	if (did_enable &&
+	    (rcvctrl & RCV_CTXT_CTRL_DONT_DROP_RHQ_FULL_SMASK)) {
 		reg = read_kctxt_csr(dd, ctxt, RCV_CTXT_STATUS);
 		if (reg != 0) {
 			dd_dev_info(dd, "ctxt %d status %lld (blocked)\n",
-				ctxt, reg);
+				    ctxt, reg);
 			read_uctxt_csr(dd, ctxt, RCV_HDR_HEAD);
 			write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, 0x10);
 			write_uctxt_csr(dd, ctxt, RCV_HDR_HEAD, 0x00);
 			read_uctxt_csr(dd, ctxt, RCV_HDR_HEAD);
 			reg = read_kctxt_csr(dd, ctxt, RCV_CTXT_STATUS);
 			dd_dev_info(dd, "ctxt %d status %lld (%s blocked)\n",
-				ctxt, reg, reg == 0 ? "not" : "still");
+				    ctxt, reg, reg == 0 ? "not" : "still");
 		}
 	}
 
@@ -11079,7 +11397,7 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 		 */
 		/* set interrupt timeout */
 		write_kctxt_csr(dd, ctxt, RCV_AVAIL_TIME_OUT,
-			(u64)rcd->rcvavail_timeout <<
+				(u64)rcd->rcvavail_timeout <<
 				RCV_AVAIL_TIME_OUT_TIME_OUT_RELOAD_SHIFT);
 
 		/* set RcvHdrHead.Counter, zero RcvHdrHead.Head (again) */
@@ -11097,28 +11415,19 @@ void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt)
 				dd->rcvhdrtail_dummy_physaddr);
 }
 
-u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
-		    u64 **cntrp)
+u32 hfi1_read_cntrs(struct hfi1_devdata *dd, char **namep, u64 **cntrp)
 {
 	int ret;
 	u64 val = 0;
 
 	if (namep) {
 		ret = dd->cntrnameslen;
-		if (pos != 0) {
-			dd_dev_err(dd, "read_cntrs does not support indexing");
-			return 0;
-		}
 		*namep = dd->cntrnames;
 	} else {
 		const struct cntr_entry *entry;
 		int i, j;
 
 		ret = (dd->ndevcntrs) * sizeof(u64);
-		if (pos != 0) {
-			dd_dev_err(dd, "read_cntrs does not support indexing");
-			return 0;
-		}
 
 		/* Get the start of the block of counters */
 		*cntrp = dd->cntrs;
@@ -11147,6 +11456,20 @@ u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
 						dd->cntrs[entry->offset + j] =
 									    val;
 					}
+				} else if (entry->flags & CNTR_SDMA) {
+					hfi1_cdbg(CNTR,
+						  "\t Per SDMA Engine\n");
+					for (j = 0; j < dd->chip_sdma_engines;
+					     j++) {
+						val =
+						entry->rw_cntr(entry, dd, j,
+							       CNTR_MODE_R, 0);
+						hfi1_cdbg(CNTR,
+							  "\t\tRead 0x%llx for %d\n",
+							  val, j);
+						dd->cntrs[entry->offset + j] =
+									val;
+					}
 				} else {
 					val = entry->rw_cntr(entry, dd,
 							CNTR_INVALID_VL,
@@ -11163,30 +11486,19 @@ u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
 /*
  * Used by sysfs to create files for hfi stats to read
  */
-u32 hfi1_read_portcntrs(struct hfi1_devdata *dd, loff_t pos, u32 port,
-			char **namep, u64 **cntrp)
+u32 hfi1_read_portcntrs(struct hfi1_pportdata *ppd, char **namep, u64 **cntrp)
 {
 	int ret;
 	u64 val = 0;
 
 	if (namep) {
-		ret = dd->portcntrnameslen;
-		if (pos != 0) {
-			dd_dev_err(dd, "index not supported");
-			return 0;
-		}
-		*namep = dd->portcntrnames;
+		ret = ppd->dd->portcntrnameslen;
+		*namep = ppd->dd->portcntrnames;
 	} else {
 		const struct cntr_entry *entry;
-		struct hfi1_pportdata *ppd;
 		int i, j;
 
-		ret = (dd->nportcntrs) * sizeof(u64);
-		if (pos != 0) {
-			dd_dev_err(dd, "indexing not supported");
-			return 0;
-		}
-		ppd = (struct hfi1_pportdata *)(dd + 1 + port);
+		ret = ppd->dd->nportcntrs * sizeof(u64);
 		*cntrp = ppd->cntrs;
 
 		for (i = 0; i < PORT_CNTR_LAST; i++) {
@@ -11235,14 +11547,14 @@ static void free_cntrs(struct hfi1_devdata *dd)
 	for (i = 0; i < dd->num_pports; i++, ppd++) {
 		kfree(ppd->cntrs);
 		kfree(ppd->scntrs);
-		free_percpu(ppd->ibport_data.rc_acks);
-		free_percpu(ppd->ibport_data.rc_qacks);
-		free_percpu(ppd->ibport_data.rc_delayed_comp);
+		free_percpu(ppd->ibport_data.rvp.rc_acks);
+		free_percpu(ppd->ibport_data.rvp.rc_qacks);
+		free_percpu(ppd->ibport_data.rvp.rc_delayed_comp);
 		ppd->cntrs = NULL;
 		ppd->scntrs = NULL;
-		ppd->ibport_data.rc_acks = NULL;
-		ppd->ibport_data.rc_qacks = NULL;
-		ppd->ibport_data.rc_delayed_comp = NULL;
+		ppd->ibport_data.rvp.rc_acks = NULL;
+		ppd->ibport_data.rvp.rc_qacks = NULL;
+		ppd->ibport_data.rvp.rc_delayed_comp = NULL;
 	}
 	kfree(dd->portcntrnames);
 	dd->portcntrnames = NULL;
@@ -11510,11 +11822,13 @@ mod_timer(&dd->synth_stats_timer, jiffies + HZ * SYNTH_CNT_TIME);
 #define C_MAX_NAME 13 /* 12 chars + one for /0 */
 static int init_cntrs(struct hfi1_devdata *dd)
 {
-	int i, rcv_ctxts, index, j;
+	int i, rcv_ctxts, j;
 	size_t sz;
 	char *p;
 	char name[C_MAX_NAME];
 	struct hfi1_pportdata *ppd;
+	const char *bit_type_32 = ",32";
+	const int bit_type_32_sz = strlen(bit_type_32);
 
 	/* set up the stats timer; the add_timer is done at the end */
 	setup_timer(&dd->synth_stats_timer, update_synth_timer,
@@ -11527,49 +11841,57 @@ static int init_cntrs(struct hfi1_devdata *dd)
 	/* size names and determine how many we have*/
 	dd->ndevcntrs = 0;
 	sz = 0;
-	index = 0;
 
 	for (i = 0; i < DEV_CNTR_LAST; i++) {
-		hfi1_dbg_early("Init cntr %s\n", dev_cntrs[i].name);
 		if (dev_cntrs[i].flags & CNTR_DISABLED) {
 			hfi1_dbg_early("\tSkipping %s\n", dev_cntrs[i].name);
 			continue;
 		}
 
 		if (dev_cntrs[i].flags & CNTR_VL) {
-			hfi1_dbg_early("\tProcessing VL cntr\n");
-			dev_cntrs[i].offset = index;
+			dev_cntrs[i].offset = dd->ndevcntrs;
 			for (j = 0; j < C_VL_COUNT; j++) {
-				memset(name, '\0', C_MAX_NAME);
 				snprintf(name, C_MAX_NAME, "%s%d",
-					dev_cntrs[i].name,
-					vl_from_idx(j));
+					 dev_cntrs[i].name, vl_from_idx(j));
+				sz += strlen(name);
+				/* Add ",32" for 32-bit counters */
+				if (dev_cntrs[i].flags & CNTR_32BIT)
+					sz += bit_type_32_sz;
+				sz++;
+				dd->ndevcntrs++;
+			}
+		} else if (dev_cntrs[i].flags & CNTR_SDMA) {
+			dev_cntrs[i].offset = dd->ndevcntrs;
+			for (j = 0; j < dd->chip_sdma_engines; j++) {
+				snprintf(name, C_MAX_NAME, "%s%d",
+					 dev_cntrs[i].name, j);
 				sz += strlen(name);
+				/* Add ",32" for 32-bit counters */
+				if (dev_cntrs[i].flags & CNTR_32BIT)
+					sz += bit_type_32_sz;
 				sz++;
-				hfi1_dbg_early("\t\t%s\n", name);
 				dd->ndevcntrs++;
-				index++;
 			}
 		} else {
-			/* +1 for newline  */
+			/* +1 for newline. */
 			sz += strlen(dev_cntrs[i].name) + 1;
+			/* Add ",32" for 32-bit counters */
+			if (dev_cntrs[i].flags & CNTR_32BIT)
+				sz += bit_type_32_sz;
+			dev_cntrs[i].offset = dd->ndevcntrs;
 			dd->ndevcntrs++;
-			dev_cntrs[i].offset = index;
-			index++;
-			hfi1_dbg_early("\tAdding %s\n", dev_cntrs[i].name);
 		}
 	}
 
 	/* allocate space for the counter values */
-	dd->cntrs = kcalloc(index, sizeof(u64), GFP_KERNEL);
+	dd->cntrs = kcalloc(dd->ndevcntrs, sizeof(u64), GFP_KERNEL);
 	if (!dd->cntrs)
 		goto bail;
 
-	dd->scntrs = kcalloc(index, sizeof(u64), GFP_KERNEL);
+	dd->scntrs = kcalloc(dd->ndevcntrs, sizeof(u64), GFP_KERNEL);
 	if (!dd->scntrs)
 		goto bail;
 
-
 	/* allocate space for the counter names */
 	dd->cntrnameslen = sz;
 	dd->cntrnames = kmalloc(sz, GFP_KERNEL);
@@ -11577,27 +11899,51 @@ static int init_cntrs(struct hfi1_devdata *dd)
 		goto bail;
 
 	/* fill in the names */
-	for (p = dd->cntrnames, i = 0, index = 0; i < DEV_CNTR_LAST; i++) {
+	for (p = dd->cntrnames, i = 0; i < DEV_CNTR_LAST; i++) {
 		if (dev_cntrs[i].flags & CNTR_DISABLED) {
 			/* Nothing */
-		} else {
-			if (dev_cntrs[i].flags & CNTR_VL) {
-				for (j = 0; j < C_VL_COUNT; j++) {
-					memset(name, '\0', C_MAX_NAME);
-					snprintf(name, C_MAX_NAME, "%s%d",
-						dev_cntrs[i].name,
-						vl_from_idx(j));
-					memcpy(p, name, strlen(name));
-					p += strlen(name);
-					*p++ = '\n';
+		} else if (dev_cntrs[i].flags & CNTR_VL) {
+			for (j = 0; j < C_VL_COUNT; j++) {
+				snprintf(name, C_MAX_NAME, "%s%d",
+					 dev_cntrs[i].name,
+					 vl_from_idx(j));
+				memcpy(p, name, strlen(name));
+				p += strlen(name);
+
+				/* Counter is 32 bits */
+				if (dev_cntrs[i].flags & CNTR_32BIT) {
+					memcpy(p, bit_type_32, bit_type_32_sz);
+					p += bit_type_32_sz;
 				}
-			} else {
-				memcpy(p, dev_cntrs[i].name,
-				       strlen(dev_cntrs[i].name));
-				p += strlen(dev_cntrs[i].name);
+
+				*p++ = '\n';
+			}
+		} else if (dev_cntrs[i].flags & CNTR_SDMA) {
+			for (j = 0; j < dd->chip_sdma_engines; j++) {
+				snprintf(name, C_MAX_NAME, "%s%d",
+					 dev_cntrs[i].name, j);
+				memcpy(p, name, strlen(name));
+				p += strlen(name);
+
+				/* Counter is 32 bits */
+				if (dev_cntrs[i].flags & CNTR_32BIT) {
+					memcpy(p, bit_type_32, bit_type_32_sz);
+					p += bit_type_32_sz;
+				}
+
 				*p++ = '\n';
 			}
-			index++;
+		} else {
+			memcpy(p, dev_cntrs[i].name, strlen(dev_cntrs[i].name));
+			p += strlen(dev_cntrs[i].name);
+
+			/* Counter is 32 bits */
+			if (dev_cntrs[i].flags & CNTR_32BIT) {
+				memcpy(p, bit_type_32, bit_type_32_sz);
+				p += bit_type_32_sz;
+			}
+
+			*p++ = '\n';
 		}
 	}
 
@@ -11620,31 +11966,31 @@ static int init_cntrs(struct hfi1_devdata *dd)
 	sz = 0;
 	dd->nportcntrs = 0;
 	for (i = 0; i < PORT_CNTR_LAST; i++) {
-		hfi1_dbg_early("Init pcntr %s\n", port_cntrs[i].name);
 		if (port_cntrs[i].flags & CNTR_DISABLED) {
 			hfi1_dbg_early("\tSkipping %s\n", port_cntrs[i].name);
 			continue;
 		}
 
 		if (port_cntrs[i].flags & CNTR_VL) {
-			hfi1_dbg_early("\tProcessing VL cntr\n");
 			port_cntrs[i].offset = dd->nportcntrs;
 			for (j = 0; j < C_VL_COUNT; j++) {
-				memset(name, '\0', C_MAX_NAME);
 				snprintf(name, C_MAX_NAME, "%s%d",
-					port_cntrs[i].name,
-					vl_from_idx(j));
+					 port_cntrs[i].name, vl_from_idx(j));
 				sz += strlen(name);
+				/* Add ",32" for 32-bit counters */
+				if (port_cntrs[i].flags & CNTR_32BIT)
+					sz += bit_type_32_sz;
 				sz++;
-				hfi1_dbg_early("\t\t%s\n", name);
 				dd->nportcntrs++;
 			}
 		} else {
-			/* +1 for newline  */
+			/* +1 for newline */
 			sz += strlen(port_cntrs[i].name) + 1;
+			/* Add ",32" for 32-bit counters */
+			if (port_cntrs[i].flags & CNTR_32BIT)
+				sz += bit_type_32_sz;
 			port_cntrs[i].offset = dd->nportcntrs;
 			dd->nportcntrs++;
-			hfi1_dbg_early("\tAdding %s\n", port_cntrs[i].name);
 		}
 	}
 
@@ -11661,18 +12007,30 @@ static int init_cntrs(struct hfi1_devdata *dd)
 
 		if (port_cntrs[i].flags & CNTR_VL) {
 			for (j = 0; j < C_VL_COUNT; j++) {
-				memset(name, '\0', C_MAX_NAME);
 				snprintf(name, C_MAX_NAME, "%s%d",
-					port_cntrs[i].name,
-					vl_from_idx(j));
+					 port_cntrs[i].name, vl_from_idx(j));
 				memcpy(p, name, strlen(name));
 				p += strlen(name);
+
+				/* Counter is 32 bits */
+				if (port_cntrs[i].flags & CNTR_32BIT) {
+					memcpy(p, bit_type_32, bit_type_32_sz);
+					p += bit_type_32_sz;
+				}
+
 				*p++ = '\n';
 			}
 		} else {
 			memcpy(p, port_cntrs[i].name,
 			       strlen(port_cntrs[i].name));
 			p += strlen(port_cntrs[i].name);
+
+			/* Counter is 32 bits */
+			if (port_cntrs[i].flags & CNTR_32BIT) {
+				memcpy(p, bit_type_32, bit_type_32_sz);
+				p += bit_type_32_sz;
+			}
+
 			*p++ = '\n';
 		}
 	}
@@ -11700,14 +12058,13 @@ bail:
 	return -ENOMEM;
 }
 
-
 static u32 chip_to_opa_lstate(struct hfi1_devdata *dd, u32 chip_lstate)
 {
 	switch (chip_lstate) {
 	default:
 		dd_dev_err(dd,
-			 "Unknown logical state 0x%x, reporting IB_PORT_DOWN\n",
-			 chip_lstate);
+			   "Unknown logical state 0x%x, reporting IB_PORT_DOWN\n",
+			   chip_lstate);
 		/* fall through */
 	case LSTATE_DOWN:
 		return IB_PORT_DOWN;
@@ -11726,7 +12083,7 @@ u32 chip_to_opa_pstate(struct hfi1_devdata *dd, u32 chip_pstate)
 	switch (chip_pstate & 0xf0) {
 	default:
 		dd_dev_err(dd, "Unexpected chip physical state of 0x%x\n",
-			chip_pstate);
+			   chip_pstate);
 		/* fall through */
 	case PLS_DISABLED:
 		return IB_PORTPHYSSTATE_DISABLED;
@@ -11792,7 +12149,7 @@ u32 get_logical_state(struct hfi1_pportdata *ppd)
 	new_state = chip_to_opa_lstate(ppd->dd, read_logical_state(ppd->dd));
 	if (new_state != ppd->lstate) {
 		dd_dev_info(ppd->dd, "logical state changed to %s (0x%x)\n",
-			opa_lstate_name(new_state), new_state);
+			    opa_lstate_name(new_state), new_state);
 		ppd->lstate = new_state;
 	}
 	/*
@@ -11851,18 +12208,17 @@ static int wait_logical_linkstate(struct hfi1_pportdata *ppd, u32 state,
 
 u8 hfi1_ibphys_portstate(struct hfi1_pportdata *ppd)
 {
-	static u32 remembered_state = 0xff;
 	u32 pstate;
 	u32 ib_pstate;
 
 	pstate = read_physical_state(ppd->dd);
 	ib_pstate = chip_to_opa_pstate(ppd->dd, pstate);
-	if (remembered_state != ib_pstate) {
+	if (ppd->last_pstate != ib_pstate) {
 		dd_dev_info(ppd->dd,
-			"%s: physical state changed to %s (0x%x), phy 0x%x\n",
-			__func__, opa_pstate_name(ib_pstate), ib_pstate,
-			pstate);
-		remembered_state = ib_pstate;
+			    "%s: physical state changed to %s (0x%x), phy 0x%x\n",
+			    __func__, opa_pstate_name(ib_pstate), ib_pstate,
+			    pstate);
+		ppd->last_pstate = ib_pstate;
 	}
 	return ib_pstate;
 }
@@ -11906,7 +12262,7 @@ u64 hfi1_gpio_mod(struct hfi1_devdata *dd, u32 target, u32 data, u32 dir,
 
 int hfi1_init_ctxt(struct send_context *sc)
 {
-	if (sc != NULL) {
+	if (sc) {
 		struct hfi1_devdata *dd = sc->dd;
 		u64 reg;
 		u8 set = (sc->type == SC_USER ?
@@ -11963,34 +12319,14 @@ void set_intr_state(struct hfi1_devdata *dd, u32 enable)
 	 * In HFI, the mask needs to be 1 to allow interrupts.
 	 */
 	if (enable) {
-		u64 cce_int_mask;
-		const int qsfp1_int_smask = QSFP1_INT % 64;
-		const int qsfp2_int_smask = QSFP2_INT % 64;
-
 		/* enable all interrupts */
 		for (i = 0; i < CCE_NUM_INT_CSRS; i++)
-			write_csr(dd, CCE_INT_MASK + (8*i), ~(u64)0);
+			write_csr(dd, CCE_INT_MASK + (8 * i), ~(u64)0);
 
-		/*
-		 * disable QSFP1 interrupts for HFI1, QSFP2 interrupts for HFI0
-		 * Qsfp1Int and Qsfp2Int are adjacent bits in the same CSR,
-		 * therefore just one of QSFP1_INT/QSFP2_INT can be used to find
-		 * the index of the appropriate CSR in the CCEIntMask CSR array
-		 */
-		cce_int_mask = read_csr(dd, CCE_INT_MASK +
-						(8*(QSFP1_INT/64)));
-		if (dd->hfi1_id) {
-			cce_int_mask &= ~((u64)1 << qsfp1_int_smask);
-			write_csr(dd, CCE_INT_MASK + (8*(QSFP1_INT/64)),
-					cce_int_mask);
-		} else {
-			cce_int_mask &= ~((u64)1 << qsfp2_int_smask);
-			write_csr(dd, CCE_INT_MASK + (8*(QSFP2_INT/64)),
-					cce_int_mask);
-		}
+		init_qsfp_int(dd);
 	} else {
 		for (i = 0; i < CCE_NUM_INT_CSRS; i++)
-			write_csr(dd, CCE_INT_MASK + (8*i), 0ull);
+			write_csr(dd, CCE_INT_MASK + (8 * i), 0ull);
 	}
 }
 
@@ -12002,7 +12338,7 @@ static void clear_all_interrupts(struct hfi1_devdata *dd)
 	int i;
 
 	for (i = 0; i < CCE_NUM_INT_CSRS; i++)
-		write_csr(dd, CCE_INT_CLEAR + (8*i), ~(u64)0);
+		write_csr(dd, CCE_INT_CLEAR + (8 * i), ~(u64)0);
 
 	write_csr(dd, CCE_ERR_CLEAR, ~(u64)0);
 	write_csr(dd, MISC_ERR_CLEAR, ~(u64)0);
@@ -12037,10 +12373,9 @@ static void clean_up_interrupts(struct hfi1_devdata *dd)
 		struct hfi1_msix_entry *me = dd->msix_entries;
 
 		for (i = 0; i < dd->num_msix_entries; i++, me++) {
-			if (me->arg == NULL) /* => no irq, no affinity */
-				break;
-			irq_set_affinity_hint(dd->msix_entries[i].msix.vector,
-					NULL);
+			if (!me->arg) /* => no irq, no affinity */
+				continue;
+			hfi1_put_irq_affinity(dd, &dd->msix_entries[i]);
 			free_irq(me->msix.vector, me->arg);
 		}
 	} else {
@@ -12061,8 +12396,6 @@ static void clean_up_interrupts(struct hfi1_devdata *dd)
 	}
 
 	/* clean structures */
-	for (i = 0; i < dd->num_msix_entries; i++)
-		free_cpumask_var(dd->msix_entries[i].mask);
 	kfree(dd->msix_entries);
 	dd->msix_entries = NULL;
 	dd->num_msix_entries = 0;
@@ -12085,10 +12418,10 @@ static void remap_intr(struct hfi1_devdata *dd, int isrc, int msix_intr)
 	/* direct the chip source to the given MSI-X interrupt */
 	m = isrc / 8;
 	n = isrc % 8;
-	reg = read_csr(dd, CCE_INT_MAP + (8*m));
-	reg &= ~((u64)0xff << (8*n));
-	reg |= ((u64)msix_intr & 0xff) << (8*n);
-	write_csr(dd, CCE_INT_MAP + (8*m), reg);
+	reg = read_csr(dd, CCE_INT_MAP + (8 * m));
+	reg &= ~((u64)0xff << (8 * n));
+	reg |= ((u64)msix_intr & 0xff) << (8 * n);
+	write_csr(dd, CCE_INT_MAP + (8 * m), reg);
 }
 
 static void remap_sdma_interrupts(struct hfi1_devdata *dd,
@@ -12101,12 +12434,12 @@ static void remap_sdma_interrupts(struct hfi1_devdata *dd,
 	 *	SDMAProgress
 	 *	SDMAIdle
 	 */
-	remap_intr(dd, IS_SDMA_START + 0*TXE_NUM_SDMA_ENGINES + engine,
-		msix_intr);
-	remap_intr(dd, IS_SDMA_START + 1*TXE_NUM_SDMA_ENGINES + engine,
-		msix_intr);
-	remap_intr(dd, IS_SDMA_START + 2*TXE_NUM_SDMA_ENGINES + engine,
-		msix_intr);
+	remap_intr(dd, IS_SDMA_START + 0 * TXE_NUM_SDMA_ENGINES + engine,
+		   msix_intr);
+	remap_intr(dd, IS_SDMA_START + 1 * TXE_NUM_SDMA_ENGINES + engine,
+		   msix_intr);
+	remap_intr(dd, IS_SDMA_START + 2 * TXE_NUM_SDMA_ENGINES + engine,
+		   msix_intr);
 }
 
 static int request_intx_irq(struct hfi1_devdata *dd)
@@ -12116,10 +12449,10 @@ static int request_intx_irq(struct hfi1_devdata *dd)
 	snprintf(dd->intx_name, sizeof(dd->intx_name), DRIVER_NAME "_%d",
 		 dd->unit);
 	ret = request_irq(dd->pcidev->irq, general_interrupt,
-				  IRQF_SHARED, dd->intx_name, dd);
+			  IRQF_SHARED, dd->intx_name, dd);
 	if (ret)
 		dd_dev_err(dd, "unable to request INTx interrupt, err %d\n",
-				ret);
+			   ret);
 	else
 		dd->requested_intx_irq = 1;
 	return ret;
@@ -12127,70 +12460,20 @@ static int request_intx_irq(struct hfi1_devdata *dd)
 
 static int request_msix_irqs(struct hfi1_devdata *dd)
 {
-	const struct cpumask *local_mask;
-	cpumask_var_t def, rcv;
-	bool def_ret, rcv_ret;
 	int first_general, last_general;
 	int first_sdma, last_sdma;
 	int first_rx, last_rx;
-	int first_cpu, curr_cpu;
-	int rcv_cpu, sdma_cpu;
-	int i, ret = 0, possible;
-	int ht;
+	int i, ret = 0;
 
 	/* calculate the ranges we are going to use */
 	first_general = 0;
-	first_sdma = last_general = first_general + 1;
-	first_rx = last_sdma = first_sdma + dd->num_sdma;
+	last_general = first_general + 1;
+	first_sdma = last_general;
+	last_sdma = first_sdma + dd->num_sdma;
+	first_rx = last_sdma;
 	last_rx = first_rx + dd->n_krcv_queues;
 
 	/*
-	 * Interrupt affinity.
-	 *
-	 * non-rcv avail gets a default mask that
-	 * starts as possible cpus with threads reset
-	 * and each rcv avail reset.
-	 *
-	 * rcv avail gets node relative 1 wrapping back
-	 * to the node relative 1 as necessary.
-	 *
-	 */
-	local_mask = cpumask_of_pcibus(dd->pcidev->bus);
-	/* if first cpu is invalid, use NUMA 0 */
-	if (cpumask_first(local_mask) >= nr_cpu_ids)
-		local_mask = topology_core_cpumask(0);
-
-	def_ret = zalloc_cpumask_var(&def, GFP_KERNEL);
-	rcv_ret = zalloc_cpumask_var(&rcv, GFP_KERNEL);
-	if (!def_ret || !rcv_ret)
-		goto bail;
-	/* use local mask as default */
-	cpumask_copy(def, local_mask);
-	possible = cpumask_weight(def);
-	/* disarm threads from default */
-	ht = cpumask_weight(
-			topology_sibling_cpumask(cpumask_first(local_mask)));
-	for (i = possible/ht; i < possible; i++)
-		cpumask_clear_cpu(i, def);
-	/* def now has full cores on chosen node*/
-	first_cpu = cpumask_first(def);
-	if (nr_cpu_ids >= first_cpu)
-		first_cpu++;
-	curr_cpu = first_cpu;
-
-	/*  One context is reserved as control context */
-	for (i = first_cpu; i < dd->n_krcv_queues + first_cpu - 1; i++) {
-		cpumask_clear_cpu(curr_cpu, def);
-		cpumask_set_cpu(curr_cpu, rcv);
-		curr_cpu = cpumask_next(curr_cpu, def);
-		if (curr_cpu >= nr_cpu_ids)
-			break;
-	}
-	/* def mask has non-rcv, rcv has recv mask */
-	rcv_cpu = cpumask_first(rcv);
-	sdma_cpu = cpumask_first(def);
-
-	/*
 	 * Sanity check - the code expects all SDMA chip source
 	 * interrupts to be in the same CSR, starting at bit 0.  Verify
 	 * that this is true by checking the bit location of the start.
@@ -12215,6 +12498,7 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 			snprintf(me->name, sizeof(me->name),
 				 DRIVER_NAME "_%d", dd->unit);
 			err_info = "general";
+			me->type = IRQ_GENERAL;
 		} else if (first_sdma <= i && i < last_sdma) {
 			idx = i - first_sdma;
 			sde = &dd->per_sdma[idx];
@@ -12224,6 +12508,7 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 				 DRIVER_NAME "_%d sdma%d", dd->unit, idx);
 			err_info = "sdma";
 			remap_sdma_interrupts(dd, idx, i);
+			me->type = IRQ_SDMA;
 		} else if (first_rx <= i && i < last_rx) {
 			idx = i - first_rx;
 			rcd = dd->rcd[idx];
@@ -12234,9 +12519,9 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 			 * Set the interrupt register and mask for this
 			 * context's interrupt.
 			 */
-			rcd->ireg = (IS_RCVAVAIL_START+idx) / 64;
+			rcd->ireg = (IS_RCVAVAIL_START + idx) / 64;
 			rcd->imask = ((u64)1) <<
-					((IS_RCVAVAIL_START+idx) % 64);
+					((IS_RCVAVAIL_START + idx) % 64);
 			handler = receive_context_interrupt;
 			thread = receive_context_thread;
 			arg = rcd;
@@ -12244,25 +12529,27 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 				 DRIVER_NAME "_%d kctxt%d", dd->unit, idx);
 			err_info = "receive context";
 			remap_intr(dd, IS_RCVAVAIL_START + idx, i);
+			me->type = IRQ_RCVCTXT;
 		} else {
 			/* not in our expected range - complain, then
-			   ignore it */
+			 * ignore it
+			 */
 			dd_dev_err(dd,
-				"Unexpected extra MSI-X interrupt %d\n", i);
+				   "Unexpected extra MSI-X interrupt %d\n", i);
 			continue;
 		}
 		/* no argument, no interrupt */
-		if (arg == NULL)
+		if (!arg)
 			continue;
 		/* make sure the name is terminated */
-		me->name[sizeof(me->name)-1] = 0;
+		me->name[sizeof(me->name) - 1] = 0;
 
 		ret = request_threaded_irq(me->msix.vector, handler, thread, 0,
-						me->name, arg);
+					   me->name, arg);
 		if (ret) {
 			dd_dev_err(dd,
-				"unable to allocate %s interrupt, vector %d, index %d, err %d\n",
-				 err_info, me->msix.vector, idx, ret);
+				   "unable to allocate %s interrupt, vector %d, index %d, err %d\n",
+				   err_info, me->msix.vector, idx, ret);
 			return ret;
 		}
 		/*
@@ -12271,52 +12558,13 @@ static int request_msix_irqs(struct hfi1_devdata *dd)
 		 */
 		me->arg = arg;
 
-		if (!zalloc_cpumask_var(
-			&dd->msix_entries[i].mask,
-			GFP_KERNEL))
-			goto bail;
-		if (handler == sdma_interrupt) {
-			dd_dev_info(dd, "sdma engine %d cpu %d\n",
-				sde->this_idx, sdma_cpu);
-			sde->cpu = sdma_cpu;
-			cpumask_set_cpu(sdma_cpu, dd->msix_entries[i].mask);
-			sdma_cpu = cpumask_next(sdma_cpu, def);
-			if (sdma_cpu >= nr_cpu_ids)
-				sdma_cpu = cpumask_first(def);
-		} else if (handler == receive_context_interrupt) {
-			dd_dev_info(dd, "rcv ctxt %d cpu %d\n", rcd->ctxt,
-				    (rcd->ctxt == HFI1_CTRL_CTXT) ?
-					    cpumask_first(def) : rcv_cpu);
-			if (rcd->ctxt == HFI1_CTRL_CTXT) {
-				/* map to first default */
-				cpumask_set_cpu(cpumask_first(def),
-						dd->msix_entries[i].mask);
-			} else {
-				cpumask_set_cpu(rcv_cpu,
-						dd->msix_entries[i].mask);
-				rcv_cpu = cpumask_next(rcv_cpu, rcv);
-				if (rcv_cpu >= nr_cpu_ids)
-					rcv_cpu = cpumask_first(rcv);
-			}
-		} else {
-			/* otherwise first def */
-			dd_dev_info(dd, "%s cpu %d\n",
-				err_info, cpumask_first(def));
-			cpumask_set_cpu(
-				cpumask_first(def), dd->msix_entries[i].mask);
-		}
-		irq_set_affinity_hint(
-			dd->msix_entries[i].msix.vector,
-			dd->msix_entries[i].mask);
+		ret = hfi1_get_irq_affinity(dd, me);
+		if (ret)
+			dd_dev_err(dd,
+				   "unable to pin IRQ %d\n", ret);
 	}
 
-out:
-	free_cpumask_var(def);
-	free_cpumask_var(rcv);
 	return ret;
-bail:
-	ret = -ENOMEM;
-	goto  out;
 }
 
 /*
@@ -12333,7 +12581,7 @@ static void reset_interrupts(struct hfi1_devdata *dd)
 
 	/* all chip interrupts map to MSI-X 0 */
 	for (i = 0; i < CCE_NUM_INT_MAP_CSRS; i++)
-		write_csr(dd, CCE_INT_MAP + (8*i), 0);
+		write_csr(dd, CCE_INT_MAP + (8 * i), 0);
 }
 
 static int set_up_interrupts(struct hfi1_devdata *dd)
@@ -12442,7 +12690,7 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 		 */
 		num_kernel_contexts = n_krcvqs + MIN_KERNEL_KCTXTS - 1;
 	else
-		num_kernel_contexts = num_online_nodes();
+		num_kernel_contexts = num_online_nodes() + 1;
 	num_kernel_contexts =
 		max_t(int, MIN_KERNEL_KCTXTS, num_kernel_contexts);
 	/*
@@ -12483,13 +12731,14 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 	dd->num_rcv_contexts = total_contexts;
 	dd->n_krcv_queues = num_kernel_contexts;
 	dd->first_user_ctxt = num_kernel_contexts;
+	dd->num_user_contexts = num_user_contexts;
 	dd->freectxts = num_user_contexts;
 	dd_dev_info(dd,
-		"rcv contexts: chip %d, used %d (kernel %d, user %d)\n",
-		(int)dd->chip_rcv_contexts,
-		(int)dd->num_rcv_contexts,
-		(int)dd->n_krcv_queues,
-		(int)dd->num_rcv_contexts - dd->n_krcv_queues);
+		    "rcv contexts: chip %d, used %d (kernel %d, user %d)\n",
+		    (int)dd->chip_rcv_contexts,
+		    (int)dd->num_rcv_contexts,
+		    (int)dd->n_krcv_queues,
+		    (int)dd->num_rcv_contexts - dd->n_krcv_queues);
 
 	/*
 	 * Receive array allocation:
@@ -12515,8 +12764,8 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 		dd->rcv_entries.ngroups = (MAX_EAGER_ENTRIES * 2) /
 			dd->rcv_entries.group_size;
 		dd_dev_info(dd,
-		   "RcvArray group count too high, change to %u\n",
-		   dd->rcv_entries.ngroups);
+			    "RcvArray group count too high, change to %u\n",
+			    dd->rcv_entries.ngroups);
 		dd->rcv_entries.nctxt_extra = 0;
 	}
 	/*
@@ -12582,7 +12831,7 @@ static void write_uninitialized_csrs_and_memories(struct hfi1_devdata *dd)
 
 	/* CceIntMap */
 	for (i = 0; i < CCE_NUM_INT_MAP_CSRS; i++)
-		write_csr(dd, CCE_INT_MAP+(8*i), 0);
+		write_csr(dd, CCE_INT_MAP + (8 * i), 0);
 
 	/* SendCtxtCreditReturnAddr */
 	for (i = 0; i < dd->chip_send_contexts; i++)
@@ -12590,8 +12839,10 @@ static void write_uninitialized_csrs_and_memories(struct hfi1_devdata *dd)
 
 	/* PIO Send buffers */
 	/* SDMA Send buffers */
-	/* These are not normally read, and (presently) have no method
-	   to be read, so are not pre-initialized */
+	/*
+	 * These are not normally read, and (presently) have no method
+	 * to be read, so are not pre-initialized
+	 */
 
 	/* RcvHdrAddr */
 	/* RcvHdrTailAddr */
@@ -12600,13 +12851,13 @@ static void write_uninitialized_csrs_and_memories(struct hfi1_devdata *dd)
 		write_kctxt_csr(dd, i, RCV_HDR_ADDR, 0);
 		write_kctxt_csr(dd, i, RCV_HDR_TAIL_ADDR, 0);
 		for (j = 0; j < RXE_NUM_TID_FLOWS; j++)
-			write_uctxt_csr(dd, i, RCV_TID_FLOW_TABLE+(8*j), 0);
+			write_uctxt_csr(dd, i, RCV_TID_FLOW_TABLE + (8 * j), 0);
 	}
 
 	/* RcvArray */
 	for (i = 0; i < dd->chip_rcv_array_count; i++)
-		write_csr(dd, RCV_ARRAY + (8*i),
-					RCV_ARRAY_RT_WRITE_ENABLE_SMASK);
+		write_csr(dd, RCV_ARRAY + (8 * i),
+			  RCV_ARRAY_RT_WRITE_ENABLE_SMASK);
 
 	/* RcvQPMapTable */
 	for (i = 0; i < 32; i++)
@@ -12638,8 +12889,8 @@ static void clear_cce_status(struct hfi1_devdata *dd, u64 status_bits,
 			return;
 		if (time_after(jiffies, timeout)) {
 			dd_dev_err(dd,
-				"Timeout waiting for CceStatus to clear bits 0x%llx, remaining 0x%llx\n",
-				status_bits, reg & status_bits);
+				   "Timeout waiting for CceStatus to clear bits 0x%llx, remaining 0x%llx\n",
+				   status_bits, reg & status_bits);
 			return;
 		}
 		udelay(1);
@@ -12671,7 +12922,7 @@ static void reset_cce_csrs(struct hfi1_devdata *dd)
 	for (i = 0; i < CCE_NUM_MSIX_VECTORS; i++) {
 		write_csr(dd, CCE_MSIX_TABLE_LOWER + (8 * i), 0);
 		write_csr(dd, CCE_MSIX_TABLE_UPPER + (8 * i),
-					CCE_MSIX_TABLE_UPPER_RESETCSR);
+			  CCE_MSIX_TABLE_UPPER_RESETCSR);
 	}
 	for (i = 0; i < CCE_NUM_MSIX_PBAS; i++) {
 		/* CCE_MSIX_PBA read-only */
@@ -12691,91 +12942,6 @@ static void reset_cce_csrs(struct hfi1_devdata *dd)
 		write_csr(dd, CCE_INT_COUNTER_ARRAY32 + (8 * i), 0);
 }
 
-/* set ASIC CSRs to chip reset defaults */
-static void reset_asic_csrs(struct hfi1_devdata *dd)
-{
-	int i;
-
-	/*
-	 * If the HFIs are shared between separate nodes or VMs,
-	 * then more will need to be done here.  One idea is a module
-	 * parameter that returns early, letting the first power-on or
-	 * a known first load do the reset and blocking all others.
-	 */
-
-	if (!(dd->flags & HFI1_DO_INIT_ASIC))
-		return;
-
-	if (dd->icode != ICODE_FPGA_EMULATION) {
-		/* emulation does not have an SBus - leave these alone */
-		/*
-		 * All writes to ASIC_CFG_SBUS_REQUEST do something.
-		 * Notes:
-		 * o The reset is not zero if aimed at the core.  See the
-		 *   SBus documentation for details.
-		 * o If the SBus firmware has been updated (e.g. by the BIOS),
-		 *   will the reset revert that?
-		 */
-		/* ASIC_CFG_SBUS_REQUEST leave alone */
-		write_csr(dd, ASIC_CFG_SBUS_EXECUTE, 0);
-	}
-	/* ASIC_SBUS_RESULT read-only */
-	write_csr(dd, ASIC_STS_SBUS_COUNTERS, 0);
-	for (i = 0; i < ASIC_NUM_SCRATCH; i++)
-		write_csr(dd, ASIC_CFG_SCRATCH + (8 * i), 0);
-	write_csr(dd, ASIC_CFG_MUTEX, 0);	/* this will clear it */
-
-	/* We might want to retain this state across FLR if we ever use it */
-	write_csr(dd, ASIC_CFG_DRV_STR, 0);
-
-	/* ASIC_CFG_THERM_POLL_EN leave alone */
-	/* ASIC_STS_THERM read-only */
-	/* ASIC_CFG_RESET leave alone */
-
-	write_csr(dd, ASIC_PCIE_SD_HOST_CMD, 0);
-	/* ASIC_PCIE_SD_HOST_STATUS read-only */
-	write_csr(dd, ASIC_PCIE_SD_INTRPT_DATA_CODE, 0);
-	write_csr(dd, ASIC_PCIE_SD_INTRPT_ENABLE, 0);
-	/* ASIC_PCIE_SD_INTRPT_PROGRESS read-only */
-	write_csr(dd, ASIC_PCIE_SD_INTRPT_STATUS, ~0ull); /* clear */
-	/* ASIC_HFI0_PCIE_SD_INTRPT_RSPD_DATA read-only */
-	/* ASIC_HFI1_PCIE_SD_INTRPT_RSPD_DATA read-only */
-	for (i = 0; i < 16; i++)
-		write_csr(dd, ASIC_PCIE_SD_INTRPT_LIST + (8 * i), 0);
-
-	/* ASIC_GPIO_IN read-only */
-	write_csr(dd, ASIC_GPIO_OE, 0);
-	write_csr(dd, ASIC_GPIO_INVERT, 0);
-	write_csr(dd, ASIC_GPIO_OUT, 0);
-	write_csr(dd, ASIC_GPIO_MASK, 0);
-	/* ASIC_GPIO_STATUS read-only */
-	write_csr(dd, ASIC_GPIO_CLEAR, ~0ull);
-	/* ASIC_GPIO_FORCE leave alone */
-
-	/* ASIC_QSFP1_IN read-only */
-	write_csr(dd, ASIC_QSFP1_OE, 0);
-	write_csr(dd, ASIC_QSFP1_INVERT, 0);
-	write_csr(dd, ASIC_QSFP1_OUT, 0);
-	write_csr(dd, ASIC_QSFP1_MASK, 0);
-	/* ASIC_QSFP1_STATUS read-only */
-	write_csr(dd, ASIC_QSFP1_CLEAR, ~0ull);
-	/* ASIC_QSFP1_FORCE leave alone */
-
-	/* ASIC_QSFP2_IN read-only */
-	write_csr(dd, ASIC_QSFP2_OE, 0);
-	write_csr(dd, ASIC_QSFP2_INVERT, 0);
-	write_csr(dd, ASIC_QSFP2_OUT, 0);
-	write_csr(dd, ASIC_QSFP2_MASK, 0);
-	/* ASIC_QSFP2_STATUS read-only */
-	write_csr(dd, ASIC_QSFP2_CLEAR, ~0ull);
-	/* ASIC_QSFP2_FORCE leave alone */
-
-	write_csr(dd, ASIC_EEP_CTL_STAT, ASIC_EEP_CTL_STAT_RESETCSR);
-	/* this also writes a NOP command, clearing paging mode */
-	write_csr(dd, ASIC_EEP_ADDR_CMD, 0);
-	write_csr(dd, ASIC_EEP_DATA, 0);
-}
-
 /* set MISC CSRs to chip reset defaults */
 static void reset_misc_csrs(struct hfi1_devdata *dd)
 {
@@ -12786,8 +12952,10 @@ static void reset_misc_csrs(struct hfi1_devdata *dd)
 		write_csr(dd, MISC_CFG_RSA_SIGNATURE + (8 * i), 0);
 		write_csr(dd, MISC_CFG_RSA_MODULUS + (8 * i), 0);
 	}
-	/* MISC_CFG_SHA_PRELOAD leave alone - always reads 0 and can
-	   only be written 128-byte chunks */
+	/*
+	 * MISC_CFG_SHA_PRELOAD leave alone - always reads 0 and can
+	 * only be written 128-byte chunks
+	 */
 	/* init RSA engine to clear lingering errors */
 	write_csr(dd, MISC_CFG_RSA_CMD, 1);
 	write_csr(dd, MISC_CFG_RSA_MU, 0);
@@ -12843,18 +13011,17 @@ static void reset_txe_csrs(struct hfi1_devdata *dd)
 	write_csr(dd, SEND_ERR_CLEAR, ~0ull);
 	/* SEND_ERR_FORCE read-only */
 	for (i = 0; i < VL_ARB_LOW_PRIO_TABLE_SIZE; i++)
-		write_csr(dd, SEND_LOW_PRIORITY_LIST + (8*i), 0);
+		write_csr(dd, SEND_LOW_PRIORITY_LIST + (8 * i), 0);
 	for (i = 0; i < VL_ARB_HIGH_PRIO_TABLE_SIZE; i++)
-		write_csr(dd, SEND_HIGH_PRIORITY_LIST + (8*i), 0);
-	for (i = 0; i < dd->chip_send_contexts/NUM_CONTEXTS_PER_SET; i++)
-		write_csr(dd, SEND_CONTEXT_SET_CTRL + (8*i), 0);
+		write_csr(dd, SEND_HIGH_PRIORITY_LIST + (8 * i), 0);
+	for (i = 0; i < dd->chip_send_contexts / NUM_CONTEXTS_PER_SET; i++)
+		write_csr(dd, SEND_CONTEXT_SET_CTRL + (8 * i), 0);
 	for (i = 0; i < TXE_NUM_32_BIT_COUNTER; i++)
-		write_csr(dd, SEND_COUNTER_ARRAY32 + (8*i), 0);
+		write_csr(dd, SEND_COUNTER_ARRAY32 + (8 * i), 0);
 	for (i = 0; i < TXE_NUM_64_BIT_COUNTER; i++)
-		write_csr(dd, SEND_COUNTER_ARRAY64 + (8*i), 0);
+		write_csr(dd, SEND_COUNTER_ARRAY64 + (8 * i), 0);
 	write_csr(dd, SEND_CM_CTRL, SEND_CM_CTRL_RESETCSR);
-	write_csr(dd, SEND_CM_GLOBAL_CREDIT,
-					SEND_CM_GLOBAL_CREDIT_RESETCSR);
+	write_csr(dd, SEND_CM_GLOBAL_CREDIT, SEND_CM_GLOBAL_CREDIT_RESETCSR);
 	/* SEND_CM_CREDIT_USED_STATUS read-only */
 	write_csr(dd, SEND_CM_TIMER_CTRL, 0);
 	write_csr(dd, SEND_CM_LOCAL_AU_TABLE0_TO3, 0);
@@ -12862,7 +13029,7 @@ static void reset_txe_csrs(struct hfi1_devdata *dd)
 	write_csr(dd, SEND_CM_REMOTE_AU_TABLE0_TO3, 0);
 	write_csr(dd, SEND_CM_REMOTE_AU_TABLE4_TO7, 0);
 	for (i = 0; i < TXE_NUM_DATA_VL; i++)
-		write_csr(dd, SEND_CM_CREDIT_VL + (8*i), 0);
+		write_csr(dd, SEND_CM_CREDIT_VL + (8 * i), 0);
 	write_csr(dd, SEND_CM_CREDIT_VL15, 0);
 	/* SEND_CM_CREDIT_USED_VL read-only */
 	/* SEND_CM_CREDIT_USED_VL15 read-only */
@@ -12948,8 +13115,8 @@ static void init_rbufs(struct hfi1_devdata *dd)
 		 */
 		if (count++ > 500) {
 			dd_dev_err(dd,
-				"%s: in-progress DMA not clearing: RcvStatus 0x%llx, continuing\n",
-				__func__, reg);
+				   "%s: in-progress DMA not clearing: RcvStatus 0x%llx, continuing\n",
+				   __func__, reg);
 			break;
 		}
 		udelay(2); /* do not busy-wait the CSR */
@@ -12978,8 +13145,8 @@ static void init_rbufs(struct hfi1_devdata *dd)
 		/* give up after 100us - slowest possible at 33MHz is 73us */
 		if (count++ > 50) {
 			dd_dev_err(dd,
-				"%s: RcvStatus.RxRbufInit not set, continuing\n",
-				__func__);
+				   "%s: RcvStatus.RxRbufInit not set, continuing\n",
+				   __func__);
 			break;
 		}
 	}
@@ -13005,7 +13172,7 @@ static void reset_rxe_csrs(struct hfi1_devdata *dd)
 	write_csr(dd, RCV_VL15, 0);
 	/* this is a clear-down */
 	write_csr(dd, RCV_ERR_INFO,
-			RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK);
+		  RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK);
 	/* RCV_ERR_STATUS read-only */
 	write_csr(dd, RCV_ERR_MASK, 0);
 	write_csr(dd, RCV_ERR_CLEAR, ~0ull);
@@ -13051,8 +13218,8 @@ static void reset_rxe_csrs(struct hfi1_devdata *dd)
 		write_uctxt_csr(dd, i, RCV_EGR_INDEX_HEAD, 0);
 		/* RCV_EGR_OFFSET_TAIL read-only */
 		for (j = 0; j < RXE_NUM_TID_FLOWS; j++) {
-			write_uctxt_csr(dd, i, RCV_TID_FLOW_TABLE + (8 * j),
-				0);
+			write_uctxt_csr(dd, i,
+					RCV_TID_FLOW_TABLE + (8 * j), 0);
 		}
 	}
 }
@@ -13154,7 +13321,7 @@ static void init_chip(struct hfi1_devdata *dd)
 		write_csr(dd, RCV_CTXT_CTRL, 0);
 	/* mask all interrupt sources */
 	for (i = 0; i < CCE_NUM_INT_CSRS; i++)
-		write_csr(dd, CCE_INT_MASK + (8*i), 0ull);
+		write_csr(dd, CCE_INT_MASK + (8 * i), 0ull);
 
 	/*
 	 * DC Reset: do a full DC reset before the register clear.
@@ -13163,7 +13330,7 @@ static void init_chip(struct hfi1_devdata *dd)
 	 * across the clear.
 	 */
 	write_csr(dd, CCE_DC_CTRL, CCE_DC_CTRL_DC_RESET_SMASK);
-	(void) read_csr(dd, CCE_DC_CTRL);
+	(void)read_csr(dd, CCE_DC_CTRL);
 
 	if (use_flr) {
 		/*
@@ -13184,22 +13351,19 @@ static void init_chip(struct hfi1_devdata *dd)
 			hfi1_pcie_flr(dd);
 			restore_pci_variables(dd);
 		}
-
-		reset_asic_csrs(dd);
 	} else {
 		dd_dev_info(dd, "Resetting CSRs with writes\n");
 		reset_cce_csrs(dd);
 		reset_txe_csrs(dd);
 		reset_rxe_csrs(dd);
-		reset_asic_csrs(dd);
 		reset_misc_csrs(dd);
 	}
 	/* clear the DC reset */
 	write_csr(dd, CCE_DC_CTRL, 0);
 
 	/* Set the LED off */
-	if (is_ax(dd))
-		setextled(dd, 0);
+	setextled(dd, 0);
+
 	/*
 	 * Clear the QSFP reset.
 	 * An FLR enforces a 0 on all out pins. The driver does not touch
@@ -13212,6 +13376,7 @@ static void init_chip(struct hfi1_devdata *dd)
 	 */
 	write_csr(dd, ASIC_QSFP1_OUT, 0x1f);
 	write_csr(dd, ASIC_QSFP2_OUT, 0x1f);
+	init_chip_resources(dd);
 }
 
 static void init_early_variables(struct hfi1_devdata *dd)
@@ -13252,12 +13417,12 @@ static void init_kdeth_qp(struct hfi1_devdata *dd)
 		kdeth_qp = DEFAULT_KDETH_QP;
 
 	write_csr(dd, SEND_BTH_QP,
-			(kdeth_qp & SEND_BTH_QP_KDETH_QP_MASK)
-				<< SEND_BTH_QP_KDETH_QP_SHIFT);
+		  (kdeth_qp & SEND_BTH_QP_KDETH_QP_MASK) <<
+		  SEND_BTH_QP_KDETH_QP_SHIFT);
 
 	write_csr(dd, RCV_BTH_QP,
-			(kdeth_qp & RCV_BTH_QP_KDETH_QP_MASK)
-				<< RCV_BTH_QP_KDETH_QP_SHIFT);
+		  (kdeth_qp & RCV_BTH_QP_KDETH_QP_MASK) <<
+		  RCV_BTH_QP_KDETH_QP_SHIFT);
 }
 
 /**
@@ -13382,22 +13547,21 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 		write_csr(dd, RCV_RSM_MAP_TABLE + (8 * i), rsmmap[i]);
 	/* add rule0 */
 	write_csr(dd, RCV_RSM_CFG /* + (8 * 0) */,
-		RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK
-			<< RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
-		2ull << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
+		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK <<
+		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
+		  2ull << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
 	write_csr(dd, RCV_RSM_SELECT /* + (8 * 0) */,
-		LRH_BTH_MATCH_OFFSET
-			<< RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
-		LRH_SC_MATCH_OFFSET << RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT |
-		LRH_SC_SELECT_OFFSET << RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT |
-		((u64)n) << RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT |
-		QPN_SELECT_OFFSET << RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT |
-		((u64)m + (u64)n) << RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT);
+		  LRH_BTH_MATCH_OFFSET << RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
+		  LRH_SC_MATCH_OFFSET << RCV_RSM_SELECT_FIELD2_OFFSET_SHIFT |
+		  LRH_SC_SELECT_OFFSET << RCV_RSM_SELECT_INDEX1_OFFSET_SHIFT |
+		  ((u64)n) << RCV_RSM_SELECT_INDEX1_WIDTH_SHIFT |
+		  QPN_SELECT_OFFSET << RCV_RSM_SELECT_INDEX2_OFFSET_SHIFT |
+		  ((u64)m + (u64)n) << RCV_RSM_SELECT_INDEX2_WIDTH_SHIFT);
 	write_csr(dd, RCV_RSM_MATCH /* + (8 * 0) */,
-		LRH_BTH_MASK << RCV_RSM_MATCH_MASK1_SHIFT |
-		LRH_BTH_VALUE << RCV_RSM_MATCH_VALUE1_SHIFT |
-		LRH_SC_MASK << RCV_RSM_MATCH_MASK2_SHIFT |
-		LRH_SC_VALUE << RCV_RSM_MATCH_VALUE2_SHIFT);
+		  LRH_BTH_MASK << RCV_RSM_MATCH_MASK1_SHIFT |
+		  LRH_BTH_VALUE << RCV_RSM_MATCH_VALUE1_SHIFT |
+		  LRH_SC_MASK << RCV_RSM_MATCH_MASK2_SHIFT |
+		  LRH_SC_VALUE << RCV_RSM_MATCH_VALUE2_SHIFT);
 	/* Enable RSM */
 	add_rcvctrl(dd, RCV_CTRL_RCV_RSM_ENABLE_SMASK);
 	kfree(rsmmap);
@@ -13415,9 +13579,8 @@ static void init_rxe(struct hfi1_devdata *dd)
 	/* enable all receive errors */
 	write_csr(dd, RCV_ERR_MASK, ~0ull);
 	/* setup QPN map table - start where VL15 context leaves off */
-	init_qos(
-		dd,
-		dd->n_krcv_queues > MIN_KERNEL_KCTXTS ? MIN_KERNEL_KCTXTS : 0);
+	init_qos(dd, dd->n_krcv_queues > MIN_KERNEL_KCTXTS ?
+		 MIN_KERNEL_KCTXTS : 0);
 	/*
 	 * make sure RcvCtrl.RcvWcb <= PCIe Device Control
 	 * Register Max_Payload_Size (PCI_EXP_DEVCTL in Linux PCIe config
@@ -13454,36 +13617,33 @@ static void assign_cm_au_table(struct hfi1_devdata *dd, u32 cu,
 			       u32 csr0to3, u32 csr4to7)
 {
 	write_csr(dd, csr0to3,
-		   0ull <<
-			SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE0_SHIFT
-		|  1ull <<
-			SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE1_SHIFT
-		|  2ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE2_SHIFT
-		|  4ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE3_SHIFT);
+		  0ull << SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE0_SHIFT |
+		  1ull << SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE1_SHIFT |
+		  2ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE2_SHIFT |
+		  4ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE0_TO3_LOCAL_AU_TABLE3_SHIFT);
 	write_csr(dd, csr4to7,
-		   8ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE4_SHIFT
-		| 16ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE5_SHIFT
-		| 32ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE6_SHIFT
-		| 64ull * cu <<
-			SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE7_SHIFT);
-
+		  8ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE4_SHIFT |
+		  16ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE5_SHIFT |
+		  32ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE6_SHIFT |
+		  64ull * cu <<
+		  SEND_CM_LOCAL_AU_TABLE4_TO7_LOCAL_AU_TABLE7_SHIFT);
 }
 
 static void assign_local_cm_au_table(struct hfi1_devdata *dd, u8 vcu)
 {
 	assign_cm_au_table(dd, vcu_to_cu(vcu), SEND_CM_LOCAL_AU_TABLE0_TO3,
-					SEND_CM_LOCAL_AU_TABLE4_TO7);
+			   SEND_CM_LOCAL_AU_TABLE4_TO7);
 }
 
 void assign_remote_cm_au_table(struct hfi1_devdata *dd, u8 vcu)
 {
 	assign_cm_au_table(dd, vcu_to_cu(vcu), SEND_CM_REMOTE_AU_TABLE0_TO3,
-					SEND_CM_REMOTE_AU_TABLE4_TO7);
+			   SEND_CM_REMOTE_AU_TABLE4_TO7);
 }
 
 static void init_txe(struct hfi1_devdata *dd)
@@ -13586,9 +13746,9 @@ int hfi1_set_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt, u16 pkey)
 	int ret = 0;
 	u64 reg;
 
-	if (ctxt < dd->num_rcv_contexts)
+	if (ctxt < dd->num_rcv_contexts) {
 		rcd = dd->rcd[ctxt];
-	else {
+	} else {
 		ret = -EINVAL;
 		goto done;
 	}
@@ -13614,9 +13774,9 @@ int hfi1_clear_ctxt_pkey(struct hfi1_devdata *dd, unsigned ctxt)
 	int ret = 0;
 	u64 reg;
 
-	if (ctxt < dd->num_rcv_contexts)
+	if (ctxt < dd->num_rcv_contexts) {
 		rcd = dd->rcd[ctxt];
-	else {
+	} else {
 		ret = -EINVAL;
 		goto done;
 	}
@@ -13639,24 +13799,26 @@ done:
  */
 void hfi1_start_cleanup(struct hfi1_devdata *dd)
 {
+	aspm_exit(dd);
 	free_cntrs(dd);
 	free_rcverr(dd);
 	clean_up_interrupts(dd);
+	finish_chip_resources(dd);
 }
 
 #define HFI_BASE_GUID(dev) \
 	((dev)->base_guid & ~(1ULL << GUID_HFI_INDEX_SHIFT))
 
 /*
- * Certain chip functions need to be initialized only once per asic
- * instead of per-device. This function finds the peer device and
- * checks whether that chip initialization needs to be done by this
- * device.
+ * Information can be shared between the two HFIs on the same ASIC
+ * in the same OS.  This function finds the peer device and sets
+ * up a shared structure.
  */
-static void asic_should_init(struct hfi1_devdata *dd)
+static int init_asic_data(struct hfi1_devdata *dd)
 {
 	unsigned long flags;
 	struct hfi1_devdata *tmp, *peer = NULL;
+	int ret = 0;
 
 	spin_lock_irqsave(&hfi1_devs_lock, flags);
 	/* Find our peer device */
@@ -13668,13 +13830,21 @@ static void asic_should_init(struct hfi1_devdata *dd)
 		}
 	}
 
-	/*
-	 * "Claim" the ASIC for initialization if it hasn't been
-	 " "claimed" yet.
-	 */
-	if (!peer || !(peer->flags & HFI1_DO_INIT_ASIC))
-		dd->flags |= HFI1_DO_INIT_ASIC;
+	if (peer) {
+		dd->asic_data = peer->asic_data;
+	} else {
+		dd->asic_data = kzalloc(sizeof(*dd->asic_data), GFP_KERNEL);
+		if (!dd->asic_data) {
+			ret = -ENOMEM;
+			goto done;
+		}
+		mutex_init(&dd->asic_data->asic_resource_mutex);
+	}
+	dd->asic_data->dds[dd->hfi1_id] = dd; /* self back-pointer */
+
+done:
 	spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+	return ret;
 }
 
 /*
@@ -13694,7 +13864,7 @@ static int obtain_boardname(struct hfi1_devdata *dd)
 	ret = read_hfi1_efi_var(dd, "description", &size,
 				(void **)&dd->boardname);
 	if (ret) {
-		dd_dev_err(dd, "Board description not found\n");
+		dd_dev_info(dd, "Board description not found\n");
 		/* use generic description */
 		dd->boardname = kstrdup(generic, GFP_KERNEL);
 		if (!dd->boardname)
@@ -13703,6 +13873,50 @@ static int obtain_boardname(struct hfi1_devdata *dd)
 	return 0;
 }
 
+/*
+ * Check the interrupt registers to make sure that they are mapped correctly.
+ * It is intended to help user identify any mismapping by VMM when the driver
+ * is running in a VM. This function should only be called before interrupt
+ * is set up properly.
+ *
+ * Return 0 on success, -EINVAL on failure.
+ */
+static int check_int_registers(struct hfi1_devdata *dd)
+{
+	u64 reg;
+	u64 all_bits = ~(u64)0;
+	u64 mask;
+
+	/* Clear CceIntMask[0] to avoid raising any interrupts */
+	mask = read_csr(dd, CCE_INT_MASK);
+	write_csr(dd, CCE_INT_MASK, 0ull);
+	reg = read_csr(dd, CCE_INT_MASK);
+	if (reg)
+		goto err_exit;
+
+	/* Clear all interrupt status bits */
+	write_csr(dd, CCE_INT_CLEAR, all_bits);
+	reg = read_csr(dd, CCE_INT_STATUS);
+	if (reg)
+		goto err_exit;
+
+	/* Set all interrupt status bits */
+	write_csr(dd, CCE_INT_FORCE, all_bits);
+	reg = read_csr(dd, CCE_INT_STATUS);
+	if (reg != all_bits)
+		goto err_exit;
+
+	/* Restore the interrupt mask */
+	write_csr(dd, CCE_INT_CLEAR, all_bits);
+	write_csr(dd, CCE_INT_MASK, mask);
+
+	return 0;
+err_exit:
+	write_csr(dd, CCE_INT_MASK, mask);
+	dd_dev_err(dd, "Interrupt registers not properly mapped by VMM\n");
+	return -EINVAL;
+}
+
 /**
  * Allocate and initialize the device structure for the hfi.
  * @dev: the pci_dev for hfi1_ib device
@@ -13727,9 +13941,10 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		"RTL FPGA emulation",
 		"Functional simulator"
 	};
+	struct pci_dev *parent = pdev->bus->self;
 
-	dd = hfi1_alloc_devdata(pdev,
-		NUM_IB_PORTS * sizeof(struct hfi1_pportdata));
+	dd = hfi1_alloc_devdata(pdev, NUM_IB_PORTS *
+				sizeof(struct hfi1_pportdata));
 	if (IS_ERR(dd))
 		goto bail;
 	ppd = dd->pport;
@@ -13750,8 +13965,8 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		/* link width active is 0 when link is down */
 		/* link width downgrade active is 0 when link is down */
 
-		if (num_vls < HFI1_MIN_VLS_SUPPORTED
-			|| num_vls > HFI1_MAX_VLS_SUPPORTED) {
+		if (num_vls < HFI1_MIN_VLS_SUPPORTED ||
+		    num_vls > HFI1_MAX_VLS_SUPPORTED) {
 			hfi1_early_err(&pdev->dev,
 				       "Invalid num_vls %u, using %u VLs\n",
 				    num_vls, HFI1_MAX_VLS_SUPPORTED);
@@ -13759,6 +13974,7 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		}
 		ppd->vls_supported = num_vls;
 		ppd->vls_operational = ppd->vls_supported;
+		ppd->actual_vls_operational = ppd->vls_supported;
 		/* Set the default MTU. */
 		for (vl = 0; vl < num_vls; vl++)
 			dd->vld[vl].mtu = hfi1_max_mtu;
@@ -13778,6 +13994,7 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 		/* start in offline */
 		ppd->host_link_state = HLS_DN_OFFLINE;
 		init_vl_arb_caches(ppd);
+		ppd->last_pstate = 0xff; /* invalid value */
 	}
 
 	dd->link_default = HLS_DN_POLL;
@@ -13803,8 +14020,21 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	dd->minrev = (dd->revision >> CCE_REVISION_CHIP_REV_MINOR_SHIFT)
 			& CCE_REVISION_CHIP_REV_MINOR_MASK;
 
-	/* obtain the hardware ID - NOT related to unit, which is a
-	   software enumeration */
+	/*
+	 * Check interrupt registers mapping if the driver has no access to
+	 * the upstream component. In this case, it is likely that the driver
+	 * is running in a VM.
+	 */
+	if (!parent) {
+		ret = check_int_registers(dd);
+		if (ret)
+			goto bail_cleanup;
+	}
+
+	/*
+	 * obtain the hardware ID - NOT related to unit, which is a
+	 * software enumeration
+	 */
 	reg = read_csr(dd, CCE_REVISION2);
 	dd->hfi1_id = (reg >> CCE_REVISION2_HFI_ID_SHIFT)
 					& CCE_REVISION2_HFI_ID_MASK;
@@ -13812,8 +14042,8 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	dd->icode = reg >> CCE_REVISION2_IMPL_CODE_SHIFT;
 	dd->irev = reg >> CCE_REVISION2_IMPL_REVISION_SHIFT;
 	dd_dev_info(dd, "Implementation: %s, revision 0x%x\n",
-		dd->icode < ARRAY_SIZE(inames) ? inames[dd->icode] : "unknown",
-		(int)dd->irev);
+		    dd->icode < ARRAY_SIZE(inames) ?
+		    inames[dd->icode] : "unknown", (int)dd->irev);
 
 	/* speeds the hardware can support */
 	dd->pport->link_speed_supported = OPA_LINK_SPEED_25G;
@@ -13842,6 +14072,7 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 			   num_vls, dd->chip_sdma_engines);
 		num_vls = dd->chip_sdma_engines;
 		ppd->vls_supported = dd->chip_sdma_engines;
+		ppd->vls_operational = ppd->vls_supported;
 	}
 
 	/*
@@ -13863,8 +14094,10 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	/* needs to be done before we look for the peer device */
 	read_guid(dd);
 
-	/* should this device init the ASIC block? */
-	asic_should_init(dd);
+	/* set up shared ASIC data with peer device */
+	ret = init_asic_data(dd);
+	if (ret)
+		goto bail_cleanup;
 
 	/* obtain chip sizes, reset chip CSRs */
 	init_chip(dd);
@@ -13874,6 +14107,9 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	if (ret)
 		goto bail_cleanup;
 
+	/* Needs to be called before hfi1_firmware_init */
+	get_platform_config(dd);
+
 	/* read in firmware */
 	ret = hfi1_firmware_init(dd);
 	if (ret)
@@ -13925,6 +14161,10 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	/* set up KDETH QP prefix in both RX and TX CSRs */
 	init_kdeth_qp(dd);
 
+	ret = hfi1_dev_affinity_init(dd);
+	if (ret)
+		goto bail_cleanup;
+
 	/* send contexts must be set up before receive contexts */
 	ret = init_send_contexts(dd);
 	if (ret)
@@ -14022,7 +14262,6 @@ static u16 delay_cycles(struct hfi1_pportdata *ppd, u32 desired_egress_rate,
 	return (u16)delta_cycles;
 }
 
-
 /**
  * create_pbc - build a pbc for transmission
  * @flags: special case flags or-ed in built pbc
@@ -14078,10 +14317,15 @@ static int thermal_init(struct hfi1_devdata *dd)
 	int ret = 0;
 
 	if (dd->icode != ICODE_RTL_SILICON ||
-	    !(dd->flags & HFI1_DO_INIT_ASIC))
+	    check_chip_resource(dd, CR_THERM_INIT, NULL))
 		return ret;
 
-	acquire_hw_mutex(dd);
+	ret = acquire_chip_resource(dd, CR_SBUS, SBUS_TIMEOUT);
+	if (ret) {
+		THERM_FAILURE(dd, ret, "Acquire SBus");
+		return ret;
+	}
+
 	dd_dev_info(dd, "Initializing thermal sensor\n");
 	/* Disable polling of thermal readings */
 	write_csr(dd, ASIC_CFG_THERM_POLL_EN, 0x0);
@@ -14128,8 +14372,14 @@ static int thermal_init(struct hfi1_devdata *dd)
 
 	/* Enable polling of thermal readings */
 	write_csr(dd, ASIC_CFG_THERM_POLL_EN, 0x1);
+
+	/* Set initialized flag */
+	ret = acquire_chip_resource(dd, CR_THERM_INIT, 0);
+	if (ret)
+		THERM_FAILURE(dd, ret, "Unable to set thermal init flag");
+
 done:
-	release_hw_mutex(dd);
+	release_chip_resource(dd, CR_SBUS);
 	return ret;
 }
 
@@ -14144,7 +14394,7 @@ static void handle_temp_err(struct hfi1_devdata *dd)
 	dd_dev_emerg(dd,
 		     "Critical temperature reached! Forcing device into freeze mode!\n");
 	dd->flags |= HFI1_FORCED_FREEZE;
-	start_freeze_handling(ppd, FREEZE_SELF|FREEZE_ABORT);
+	start_freeze_handling(ppd, FREEZE_SELF | FREEZE_ABORT);
 	/*
 	 * Shut DC down as much and as quickly as possible.
 	 *
@@ -14158,8 +14408,8 @@ static void handle_temp_err(struct hfi1_devdata *dd)
 	 */
 	ppd->driver_link_ready = 0;
 	ppd->link_enabled = 0;
-	set_physical_link_state(dd, PLS_OFFLINE |
-				(OPA_LINKDOWN_REASON_SMA_DISABLED << 8));
+	set_physical_link_state(dd, (OPA_LINKDOWN_REASON_SMA_DISABLED << 8) |
+				PLS_OFFLINE);
 	/*
 	 * Step 2: Shutdown LCB and 8051
 	 *         After shutdown, do not restore DC_CFG_RESET value.
diff --git a/drivers/staging/rdma/hfi1/chip.h b/drivers/staging/rdma/hfi1/chip.h
index 5b375ddc345d..4f3b878e43eb 100644
--- a/drivers/staging/rdma/hfi1/chip.h
+++ b/drivers/staging/rdma/hfi1/chip.h
@@ -1,14 +1,13 @@
 #ifndef _CHIP_H
 #define _CHIP_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -79,8 +76,10 @@
 #define PIO_CMASK 0x7ff	/* counter mask for free and fill counters */
 #define MAX_EAGER_ENTRIES    2048	/* max receive eager entries */
 #define MAX_TID_PAIR_ENTRIES 1024	/* max receive expected pairs */
-/* Virtual? Allocation Unit, defined as AU = 8*2^vAU, 64 bytes, AU is fixed
-   at 64 bytes for all generation one devices */
+/*
+ * Virtual? Allocation Unit, defined as AU = 8*2^vAU, 64 bytes, AU is fixed
+ * at 64 bytes for all generation one devices
+ */
 #define CM_VAU 3
 /* HFI link credit count, AKA receive buffer depth (RBUF_DEPTH) */
 #define CM_GLOBAL_CREDITS 0x940
@@ -93,15 +92,15 @@
 #define TXE_PIO_SEND (TXE + TXE_PIO_SEND_OFFSET)
 
 /* PBC flags */
-#define PBC_INTR		(1ull << 31)
+#define PBC_INTR		BIT_ULL(31)
 #define PBC_DC_INFO_SHIFT	(30)
-#define PBC_DC_INFO		(1ull << PBC_DC_INFO_SHIFT)
-#define PBC_TEST_EBP	(1ull << 29)
-#define PBC_PACKET_BYPASS	(1ull << 28)
-#define PBC_CREDIT_RETURN	(1ull << 25)
-#define PBC_INSERT_BYPASS_ICRC (1ull << 24)
-#define PBC_TEST_BAD_ICRC	(1ull << 23)
-#define PBC_FECN		(1ull << 22)
+#define PBC_DC_INFO		BIT_ULL(PBC_DC_INFO_SHIFT)
+#define PBC_TEST_EBP		BIT_ULL(29)
+#define PBC_PACKET_BYPASS	BIT_ULL(28)
+#define PBC_CREDIT_RETURN	BIT_ULL(25)
+#define PBC_INSERT_BYPASS_ICRC	BIT_ULL(24)
+#define PBC_TEST_BAD_ICRC	BIT_ULL(23)
+#define PBC_FECN		BIT_ULL(22)
 
 /* PbcInsertHcrc field settings */
 #define PBC_IHCRC_LKDETH 0x0	/* insert @ local KDETH offset */
@@ -212,7 +211,7 @@
 #define PLS_CONFIGPHY_DEBOUCE		   0x40
 #define PLS_CONFIGPHY_ESTCOMM		   0x41
 #define PLS_CONFIGPHY_ESTCOMM_TXRX_HUNT	   0x42
-#define PLS_CONFIGPHY_ESTcOMM_LOCAL_COMPLETE   0x43
+#define PLS_CONFIGPHY_ESTCOMM_LOCAL_COMPLETE   0x43
 #define PLS_CONFIGPHY_OPTEQ			   0x44
 #define PLS_CONFIGPHY_OPTEQ_OPTIMIZING	   0x44
 #define PLS_CONFIGPHY_OPTEQ_LOCAL_COMPLETE	   0x45
@@ -242,36 +241,37 @@
 #define HCMD_SUCCESS 2
 
 /* DC_DC8051_DBG_ERR_INFO_SET_BY_8051.ERROR - error flags */
-#define SPICO_ROM_FAILED		    (1 <<  0)
-#define UNKNOWN_FRAME		    (1 <<  1)
-#define TARGET_BER_NOT_MET		    (1 <<  2)
-#define FAILED_SERDES_INTERNAL_LOOPBACK (1 <<  3)
-#define FAILED_SERDES_INIT		    (1 <<  4)
-#define FAILED_LNI_POLLING		    (1 <<  5)
-#define FAILED_LNI_DEBOUNCE		    (1 <<  6)
-#define FAILED_LNI_ESTBCOMM		    (1 <<  7)
-#define FAILED_LNI_OPTEQ		    (1 <<  8)
-#define FAILED_LNI_VERIFY_CAP1	    (1 <<  9)
-#define FAILED_LNI_VERIFY_CAP2	    (1 << 10)
-#define FAILED_LNI_CONFIGLT		    (1 << 11)
+#define SPICO_ROM_FAILED		BIT(0)
+#define UNKNOWN_FRAME			BIT(1)
+#define TARGET_BER_NOT_MET		BIT(2)
+#define FAILED_SERDES_INTERNAL_LOOPBACK	BIT(3)
+#define FAILED_SERDES_INIT		BIT(4)
+#define FAILED_LNI_POLLING		BIT(5)
+#define FAILED_LNI_DEBOUNCE		BIT(6)
+#define FAILED_LNI_ESTBCOMM		BIT(7)
+#define FAILED_LNI_OPTEQ		BIT(8)
+#define FAILED_LNI_VERIFY_CAP1		BIT(9)
+#define FAILED_LNI_VERIFY_CAP2		BIT(10)
+#define FAILED_LNI_CONFIGLT		BIT(11)
+#define HOST_HANDSHAKE_TIMEOUT		BIT(12)
 
 #define FAILED_LNI (FAILED_LNI_POLLING | FAILED_LNI_DEBOUNCE \
 			| FAILED_LNI_ESTBCOMM | FAILED_LNI_OPTEQ \
 			| FAILED_LNI_VERIFY_CAP1 \
 			| FAILED_LNI_VERIFY_CAP2 \
-			| FAILED_LNI_CONFIGLT)
+			| FAILED_LNI_CONFIGLT | HOST_HANDSHAKE_TIMEOUT)
 
 /* DC_DC8051_DBG_ERR_INFO_SET_BY_8051.HOST_MSG - host message flags */
-#define HOST_REQ_DONE	   (1 << 0)
-#define BC_PWR_MGM_MSG	   (1 << 1)
-#define BC_SMA_MSG		   (1 << 2)
-#define BC_BCC_UNKOWN_MSG	   (1 << 3)
-#define BC_IDLE_UNKNOWN_MSG	   (1 << 4)
-#define EXT_DEVICE_CFG_REQ	   (1 << 5)
-#define VERIFY_CAP_FRAME	   (1 << 6)
-#define LINKUP_ACHIEVED	   (1 << 7)
-#define LINK_GOING_DOWN	   (1 << 8)
-#define LINK_WIDTH_DOWNGRADED  (1 << 9)
+#define HOST_REQ_DONE		BIT(0)
+#define BC_PWR_MGM_MSG		BIT(1)
+#define BC_SMA_MSG		BIT(2)
+#define BC_BCC_UNKNOWN_MSG	BIT(3)
+#define BC_IDLE_UNKNOWN_MSG	BIT(4)
+#define EXT_DEVICE_CFG_REQ	BIT(5)
+#define VERIFY_CAP_FRAME	BIT(6)
+#define LINKUP_ACHIEVED		BIT(7)
+#define LINK_GOING_DOWN		BIT(8)
+#define LINK_WIDTH_DOWNGRADED	BIT(9)
 
 /* DC_DC8051_CFG_EXT_DEV_1.REQ_TYPE - 8051 host requests */
 #define HREQ_LOAD_CONFIG	0x01
@@ -335,14 +335,14 @@
  * the CSR fields hold multiples of this value.
  */
 #define RCV_SHIFT 3
-#define RCV_INCREMENT (1 << RCV_SHIFT)
+#define RCV_INCREMENT BIT(RCV_SHIFT)
 
 /*
  * Receive header queue entry increment - the CSR holds multiples of
  * this value.
  */
 #define HDRQ_SIZE_SHIFT 5
-#define HDRQ_INCREMENT (1 << HDRQ_SIZE_SHIFT)
+#define HDRQ_INCREMENT BIT(HDRQ_SIZE_SHIFT)
 
 /*
  * Freeze handling flags
@@ -371,6 +371,9 @@
 #define NUM_LANE_FIELDS    0x8
 
 /* 8051 general register Field IDs */
+#define LINK_OPTIMIZATION_SETTINGS   0x00
+#define LINK_TUNING_PARAMETERS	     0x02
+#define DC_HOST_COMM_SETTINGS	     0x03
 #define TX_SETTINGS		     0x06
 #define VERIFY_CAP_LOCAL_PHY	     0x07
 #define VERIFY_CAP_LOCAL_FABRIC	     0x08
@@ -387,6 +390,10 @@
 #define LINK_QUALITY_INFO            0x14
 #define REMOTE_DEVICE_ID	     0x15
 
+/* 8051 lane specific register field IDs */
+#define TX_EQ_SETTINGS		0x00
+#define CHANNEL_LOSS_SETTINGS	0x05
+
 /* Lane ID for general configuration registers */
 #define GENERAL_CONFIG 4
 
@@ -511,8 +518,10 @@ enum {
 #define LCB_CRC_48B			0x2	/* 48b CRC */
 #define LCB_CRC_12B_16B_PER_LANE	0x3	/* 12b-16b per lane CRC */
 
-/* the following enum is (almost) a copy/paste of the definition
- * in the OPA spec, section 20.2.2.6.8 (PortInfo) */
+/*
+ * the following enum is (almost) a copy/paste of the definition
+ * in the OPA spec, section 20.2.2.6.8 (PortInfo)
+ */
 enum {
 	PORT_LTP_CRC_MODE_NONE = 0,
 	PORT_LTP_CRC_MODE_14 = 1, /* 14-bit LTP CRC mode (optional) */
@@ -614,6 +623,8 @@ u64 create_pbc(struct hfi1_pportdata *ppd, u64, int, u32, u32);
 #define NUM_PCIE_SERDES 16	/* number of PCIe serdes on the SBus */
 extern const u8 pcie_serdes_broadcast[];
 extern const u8 pcie_pcs_addrs[2][NUM_PCIE_SERDES];
+extern uint platform_config_load;
+
 /* SBus commands */
 #define RESET_SBUS_RECEIVER 0x20
 #define WRITE_SBUS_RECEIVER 0x21
@@ -629,6 +640,42 @@ int load_firmware(struct hfi1_devdata *dd);
 void dispose_firmware(void);
 int acquire_hw_mutex(struct hfi1_devdata *dd);
 void release_hw_mutex(struct hfi1_devdata *dd);
+
+/*
+ * Bitmask of dynamic access for ASIC block chip resources.  Each HFI has its
+ * own range of bits for the resource so it can clear its own bits on
+ * starting and exiting.  If either HFI has the resource bit set, the
+ * resource is in use.  The separate bit ranges are:
+ *	HFI0 bits  7:0
+ *	HFI1 bits 15:8
+ */
+#define CR_SBUS  0x01	/* SBUS, THERM, and PCIE registers */
+#define CR_EPROM 0x02	/* EEP, GPIO registers */
+#define CR_I2C1  0x04	/* QSFP1_OE register */
+#define CR_I2C2  0x08	/* QSFP2_OE register */
+#define CR_DYN_SHIFT 8	/* dynamic flag shift */
+#define CR_DYN_MASK  ((1ull << CR_DYN_SHIFT) - 1)
+
+/*
+ * Bitmask of static ASIC states these are outside of the dynamic ASIC
+ * block chip resources above.  These are to be set once and never cleared.
+ * Must be holding the SBus dynamic flag when setting.
+ */
+#define CR_THERM_INIT	0x010000
+
+int acquire_chip_resource(struct hfi1_devdata *dd, u32 resource, u32 mswait);
+void release_chip_resource(struct hfi1_devdata *dd, u32 resource);
+bool check_chip_resource(struct hfi1_devdata *dd, u32 resource,
+			 const char *func);
+void init_chip_resources(struct hfi1_devdata *dd);
+void finish_chip_resources(struct hfi1_devdata *dd);
+
+/* ms wait time for access to an SBus resoure */
+#define SBUS_TIMEOUT 4000 /* long enough for a FW download and SBR */
+
+/* ms wait time for a qsfp (i2c) chain to become available */
+#define QSFP_WAIT 20000 /* long enough for FW update to the F4 uc */
+
 void fabric_serdes_reset(struct hfi1_devdata *dd);
 int read_8051_data(struct hfi1_devdata *dd, u32 addr, u32 len, u64 *result);
 
@@ -644,13 +691,17 @@ void handle_verify_cap(struct work_struct *work);
 void handle_freeze(struct work_struct *work);
 void handle_link_up(struct work_struct *work);
 void handle_link_down(struct work_struct *work);
+void handle_8051_request(struct work_struct *work);
 void handle_link_downgrade(struct work_struct *work);
 void handle_link_bounce(struct work_struct *work);
 void handle_sma_message(struct work_struct *work);
+void reset_qsfp(struct hfi1_pportdata *ppd);
+void qsfp_event(struct work_struct *work);
 void start_freeze_handling(struct hfi1_pportdata *ppd, int flags);
 int send_idle_sma(struct hfi1_devdata *dd, u64 message);
+int load_8051_config(struct hfi1_devdata *, u8, u8, u32);
+int read_8051_config(struct hfi1_devdata *, u8, u8, u32 *);
 int start_link(struct hfi1_pportdata *ppd);
-void init_qsfp(struct hfi1_pportdata *ppd);
 int bringup_serdes(struct hfi1_pportdata *ppd);
 void set_intr_state(struct hfi1_devdata *dd, u32 enable);
 void apply_link_downgrade_policy(struct hfi1_pportdata *ppd,
@@ -690,6 +741,8 @@ u64 read_dev_cntr(struct hfi1_devdata *dd, int index, int vl);
 u64 write_dev_cntr(struct hfi1_devdata *dd, int index, int vl, u64 data);
 u64 read_port_cntr(struct hfi1_pportdata *ppd, int index, int vl);
 u64 write_port_cntr(struct hfi1_pportdata *ppd, int index, int vl, u64 data);
+u32 read_logical_state(struct hfi1_devdata *dd);
+void force_recv_intr(struct hfi1_ctxtdata *rcd);
 
 /* Per VL indexes */
 enum {
@@ -785,8 +838,14 @@ enum {
 	C_SW_CPU_RCV_LIM,
 	C_SW_VTX_WAIT,
 	C_SW_PIO_WAIT,
+	C_SW_PIO_DRAIN,
 	C_SW_KMEM_WAIT,
 	C_SW_SEND_SCHED,
+	C_SDMA_DESC_FETCHED_CNT,
+	C_SDMA_INT_CNT,
+	C_SDMA_ERR_CNT,
+	C_SDMA_IDLE_INT_CNT,
+	C_SDMA_PROGRESS_INT_CNT,
 /* MISC_ERR_STATUS */
 	C_MISC_PLL_LOCK_FAIL_ERR,
 	C_MISC_MBIST_FAIL_ERR,
@@ -1275,10 +1334,8 @@ void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
 		  u32 type, unsigned long pa, u16 order);
 void hfi1_quiet_serdes(struct hfi1_pportdata *ppd);
 void hfi1_rcvctrl(struct hfi1_devdata *dd, unsigned int op, int ctxt);
-u32 hfi1_read_cntrs(struct hfi1_devdata *dd, loff_t pos, char **namep,
-		    u64 **cntrp);
-u32 hfi1_read_portcntrs(struct hfi1_devdata *dd, loff_t pos, u32 port,
-			char **namep, u64 **cntrp);
+u32 hfi1_read_cntrs(struct hfi1_devdata *dd, char **namep, u64 **cntrp);
+u32 hfi1_read_portcntrs(struct hfi1_pportdata *ppd, char **namep, u64 **cntrp);
 u8 hfi1_ibphys_portstate(struct hfi1_pportdata *ppd);
 int hfi1_get_ib_cfg(struct hfi1_pportdata *ppd, int which);
 int hfi1_set_ib_cfg(struct hfi1_pportdata *ppd, int which, u32 val);
diff --git a/drivers/staging/rdma/hfi1/chip_registers.h b/drivers/staging/rdma/hfi1/chip_registers.h
index 014d7a609ea0..770f05c9b8de 100644
--- a/drivers/staging/rdma/hfi1/chip_registers.h
+++ b/drivers/staging/rdma/hfi1/chip_registers.h
@@ -2,14 +2,13 @@
 #define DEF_CHIP_REG
 
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -21,8 +20,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -1281,6 +1278,9 @@
 #define SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SHIFT 0
 #define SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SMASK 0xFFFFull
 #define PCIE_CFG_REG_PL2 (PCIE + 0x000000000708)
+#define PCIE_CFG_REG_PL3 (PCIE + 0x00000000070C)
+#define PCIE_CFG_REG_PL3_L1_ENT_LATENCY_SHIFT 27
+#define PCIE_CFG_REG_PL3_L1_ENT_LATENCY_SMASK 0x38000000
 #define PCIE_CFG_REG_PL102 (PCIE + 0x000000000898)
 #define PCIE_CFG_REG_PL102_GEN3_EQ_POST_CURSOR_PSET_SHIFT 12
 #define PCIE_CFG_REG_PL102_GEN3_EQ_CURSOR_PSET_SHIFT 6
@@ -1301,5 +1301,6 @@
 #define CCE_INT_BLOCKED (CCE + 0x000000110C00)
 #define SEND_DMA_IDLE_CNT (TXE + 0x000000200040)
 #define SEND_DMA_DESC_FETCHED_CNT (TXE + 0x000000200058)
+#define CCE_MSIX_PBA_OFFSET 0X0110000
 
 #endif          /* DEF_CHIP_REG */
diff --git a/drivers/staging/rdma/hfi1/common.h b/drivers/staging/rdma/hfi1/common.h
index 5dd92720faae..e9b6bb322025 100644
--- a/drivers/staging/rdma/hfi1/common.h
+++ b/drivers/staging/rdma/hfi1/common.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -341,19 +338,16 @@ struct hfi1_message_header {
 #define FULL_MGMT_P_KEY      0xFFFF
 
 #define DEFAULT_P_KEY LIM_MGMT_P_KEY
-#define HFI1_PERMISSIVE_LID 0xFFFF
 #define HFI1_AETH_CREDIT_SHIFT 24
 #define HFI1_AETH_CREDIT_MASK 0x1F
 #define HFI1_AETH_CREDIT_INVAL 0x1F
 #define HFI1_MSN_MASK 0xFFFFFF
-#define HFI1_QPN_MASK 0xFFFFFF
 #define HFI1_FECN_SHIFT 31
 #define HFI1_FECN_MASK 1
-#define HFI1_FECN_SMASK (1 << HFI1_FECN_SHIFT)
+#define HFI1_FECN_SMASK BIT(HFI1_FECN_SHIFT)
 #define HFI1_BECN_SHIFT 30
 #define HFI1_BECN_MASK 1
-#define HFI1_BECN_SMASK (1 << HFI1_BECN_SHIFT)
-#define HFI1_MULTICAST_LID_BASE 0xC000
+#define HFI1_BECN_SMASK BIT(HFI1_BECN_SHIFT)
 
 static inline __u64 rhf_to_cpu(const __le32 *rbuf)
 {
diff --git a/drivers/staging/rdma/hfi1/debugfs.c b/drivers/staging/rdma/hfi1/debugfs.c
index acd2269e9f14..dbab9d9cc288 100644
--- a/drivers/staging/rdma/hfi1/debugfs.c
+++ b/drivers/staging/rdma/hfi1/debugfs.c
@@ -1,13 +1,12 @@
 #ifdef CONFIG_DEBUG_FS
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -19,8 +18,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -52,6 +49,7 @@
 #include <linux/seq_file.h>
 #include <linux/kernel.h>
 #include <linux/export.h>
+#include <linux/module.h>
 
 #include "hfi.h"
 #include "debugfs.h"
@@ -71,6 +69,7 @@ static const struct seq_operations _##name##_seq_ops = { \
 	.stop  = _##name##_seq_stop, \
 	.show  = _##name##_seq_show \
 }
+
 #define DEBUGFS_SEQ_FILE_OPEN(name) \
 static int _##name##_open(struct inode *inode, struct file *s) \
 { \
@@ -102,7 +101,6 @@ do { \
 		pr_warn("create of %s failed\n", name); \
 } while (0)
 
-
 #define DEBUGFS_SEQ_FILE_CREATE(name, parent, data) \
 	DEBUGFS_FILE_CREATE(#name, parent, data, &_##name##_file_ops, S_IRUGO)
 
@@ -127,7 +125,6 @@ static void *_opcode_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
 	return pos;
 }
 
-
 static void _opcode_stats_seq_stop(struct seq_file *s, void *v)
 __releases(RCU)
 {
@@ -151,8 +148,8 @@ static int _opcode_stats_seq_show(struct seq_file *s, void *v)
 	if (!n_packets && !n_bytes)
 		return SEQ_SKIP;
 	seq_printf(s, "%02llx %llu/%llu\n", i,
-		(unsigned long long) n_packets,
-		(unsigned long long) n_bytes);
+		   (unsigned long long)n_packets,
+		   (unsigned long long)n_bytes);
 
 	return 0;
 }
@@ -247,7 +244,7 @@ __acquires(RCU)
 }
 
 static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
-				   loff_t *pos)
+				loff_t *pos)
 {
 	struct qp_iter *iter = iter_ptr;
 
@@ -308,7 +305,6 @@ static void *_sdes_seq_next(struct seq_file *s, void *v, loff_t *pos)
 	return pos;
 }
 
-
 static void _sdes_seq_stop(struct seq_file *s, void *v)
 __releases(RCU)
 {
@@ -341,7 +337,7 @@ static ssize_t dev_counters_read(struct file *file, char __user *buf,
 
 	rcu_read_lock();
 	dd = private2dd(file);
-	avail = hfi1_read_cntrs(dd, *ppos, NULL, &counters);
+	avail = hfi1_read_cntrs(dd, NULL, &counters);
 	rval =  simple_read_from_buffer(buf, count, ppos, counters, avail);
 	rcu_read_unlock();
 	return rval;
@@ -358,7 +354,7 @@ static ssize_t dev_names_read(struct file *file, char __user *buf,
 
 	rcu_read_lock();
 	dd = private2dd(file);
-	avail = hfi1_read_cntrs(dd, *ppos, &names, NULL);
+	avail = hfi1_read_cntrs(dd, &names, NULL);
 	rval =  simple_read_from_buffer(buf, count, ppos, names, avail);
 	rcu_read_unlock();
 	return rval;
@@ -385,8 +381,7 @@ static ssize_t portnames_read(struct file *file, char __user *buf,
 
 	rcu_read_lock();
 	dd = private2dd(file);
-	/* port number n/a here since names are constant */
-	avail = hfi1_read_portcntrs(dd, *ppos, 0, &names, NULL);
+	avail = hfi1_read_portcntrs(dd->pport, &names, NULL);
 	rval = simple_read_from_buffer(buf, count, ppos, names, avail);
 	rcu_read_unlock();
 	return rval;
@@ -394,28 +389,150 @@ static ssize_t portnames_read(struct file *file, char __user *buf,
 
 /* read the per-port counters */
 static ssize_t portcntrs_debugfs_read(struct file *file, char __user *buf,
-				size_t count, loff_t *ppos)
+				      size_t count, loff_t *ppos)
 {
 	u64 *counters;
 	size_t avail;
-	struct hfi1_devdata *dd;
 	struct hfi1_pportdata *ppd;
 	ssize_t rval;
 
 	rcu_read_lock();
 	ppd = private2ppd(file);
-	dd = ppd->dd;
-	avail = hfi1_read_portcntrs(dd, *ppos, ppd->port - 1, NULL, &counters);
+	avail = hfi1_read_portcntrs(ppd, NULL, &counters);
 	rval = simple_read_from_buffer(buf, count, ppos, counters, avail);
 	rcu_read_unlock();
 	return rval;
 }
 
+static void check_dyn_flag(u64 scratch0, char *p, int size, int *used,
+			   int this_hfi, int hfi, u32 flag, const char *what)
+{
+	u32 mask;
+
+	mask = flag << (hfi ? CR_DYN_SHIFT : 0);
+	if (scratch0 & mask) {
+		*used += scnprintf(p + *used, size - *used,
+				   "  0x%08x - HFI%d %s in use, %s device\n",
+				   mask, hfi, what,
+				   this_hfi == hfi ? "this" : "other");
+	}
+}
+
+static ssize_t asic_flags_read(struct file *file, char __user *buf,
+			       size_t count, loff_t *ppos)
+{
+	struct hfi1_pportdata *ppd;
+	struct hfi1_devdata *dd;
+	u64 scratch0;
+	char *tmp;
+	int ret = 0;
+	int size;
+	int used;
+	int i;
+
+	rcu_read_lock();
+	ppd = private2ppd(file);
+	dd = ppd->dd;
+	size = PAGE_SIZE;
+	used = 0;
+	tmp = kmalloc(size, GFP_KERNEL);
+	if (!tmp) {
+		rcu_read_unlock();
+		return -ENOMEM;
+	}
+
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	used += scnprintf(tmp + used, size - used,
+			  "Resource flags: 0x%016llx\n", scratch0);
+
+	/* check permanent flag */
+	if (scratch0 & CR_THERM_INIT) {
+		used += scnprintf(tmp + used, size - used,
+				  "  0x%08x - thermal monitoring initialized\n",
+				  (u32)CR_THERM_INIT);
+	}
+
+	/* check each dynamic flag on each HFI */
+	for (i = 0; i < 2; i++) {
+		check_dyn_flag(scratch0, tmp, size, &used, dd->hfi1_id, i,
+			       CR_SBUS, "SBus");
+		check_dyn_flag(scratch0, tmp, size, &used, dd->hfi1_id, i,
+			       CR_EPROM, "EPROM");
+		check_dyn_flag(scratch0, tmp, size, &used, dd->hfi1_id, i,
+			       CR_I2C1, "i2c chain 1");
+		check_dyn_flag(scratch0, tmp, size, &used, dd->hfi1_id, i,
+			       CR_I2C2, "i2c chain 2");
+	}
+	used += scnprintf(tmp + used, size - used, "Write bits to clear\n");
+
+	ret = simple_read_from_buffer(buf, count, ppos, tmp, used);
+	rcu_read_unlock();
+	kfree(tmp);
+	return ret;
+}
+
+static ssize_t asic_flags_write(struct file *file, const char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	struct hfi1_pportdata *ppd;
+	struct hfi1_devdata *dd;
+	char *buff;
+	int ret;
+	unsigned long long value;
+	u64 scratch0;
+	u64 clear;
+
+	rcu_read_lock();
+	ppd = private2ppd(file);
+	dd = ppd->dd;
+
+	buff = kmalloc(count + 1, GFP_KERNEL);
+	if (!buff) {
+		ret = -ENOMEM;
+		goto do_return;
+	}
+
+	ret = copy_from_user(buff, buf, count);
+	if (ret > 0) {
+		ret = -EFAULT;
+		goto do_free;
+	}
+
+	/* zero terminate and read the expected integer */
+	buff[count] = 0;
+	ret = kstrtoull(buff, 0, &value);
+	if (ret)
+		goto do_free;
+	clear = value;
+
+	/* obtain exclusive access */
+	mutex_lock(&dd->asic_data->asic_resource_mutex);
+	acquire_hw_mutex(dd);
+
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	scratch0 &= ~clear;
+	write_csr(dd, ASIC_CFG_SCRATCH, scratch0);
+	/* force write to be visible to other HFI on another OS */
+	(void)read_csr(dd, ASIC_CFG_SCRATCH);
+
+	release_hw_mutex(dd);
+	mutex_unlock(&dd->asic_data->asic_resource_mutex);
+
+	/* return the number of bytes written */
+	ret = count;
+
+ do_free:
+	kfree(buff);
+ do_return:
+	rcu_read_unlock();
+	return ret;
+}
+
 /*
  * read the per-port QSFP data for ppd
  */
 static ssize_t qsfp_debugfs_dump(struct file *file, char __user *buf,
-			   size_t count, loff_t *ppos)
+				 size_t count, loff_t *ppos)
 {
 	struct hfi1_pportdata *ppd;
 	char *tmp;
@@ -439,7 +556,7 @@ static ssize_t qsfp_debugfs_dump(struct file *file, char __user *buf,
 
 /* Do an i2c write operation on the chain for the given HFI. */
 static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos, u32 target)
+				   size_t count, loff_t *ppos, u32 target)
 {
 	struct hfi1_pportdata *ppd;
 	char *buff;
@@ -451,6 +568,16 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 	rcu_read_lock();
 	ppd = private2ppd(file);
 
+	/* byte offset format: [offsetSize][i2cAddr][offsetHigh][offsetLow] */
+	i2c_addr = (*ppos >> 16) & 0xffff;
+	offset = *ppos & 0xffff;
+
+	/* explicitly reject invalid address 0 to catch cp and cat */
+	if (i2c_addr == 0) {
+		ret = -EINVAL;
+		goto _return;
+	}
+
 	buff = kmalloc(count, GFP_KERNEL);
 	if (!buff) {
 		ret = -ENOMEM;
@@ -463,9 +590,6 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 		goto _free;
 	}
 
-	i2c_addr = (*ppos >> 16) & 0xff;
-	offset = *ppos & 0xffff;
-
 	total_written = i2c_write(ppd, target, i2c_addr, offset, buff, count);
 	if (total_written < 0) {
 		ret = total_written;
@@ -485,21 +609,21 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 
 /* Do an i2c write operation on chain for HFI 0. */
 static ssize_t i2c1_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos)
+				  size_t count, loff_t *ppos)
 {
 	return __i2c_debugfs_write(file, buf, count, ppos, 0);
 }
 
 /* Do an i2c write operation on chain for HFI 1. */
 static ssize_t i2c2_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos)
+				  size_t count, loff_t *ppos)
 {
 	return __i2c_debugfs_write(file, buf, count, ppos, 1);
 }
 
 /* Do an i2c read operation on the chain for the given HFI. */
 static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos, u32 target)
+				  size_t count, loff_t *ppos, u32 target)
 {
 	struct hfi1_pportdata *ppd;
 	char *buff;
@@ -511,15 +635,22 @@ static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
 	rcu_read_lock();
 	ppd = private2ppd(file);
 
+	/* byte offset format: [offsetSize][i2cAddr][offsetHigh][offsetLow] */
+	i2c_addr = (*ppos >> 16) & 0xffff;
+	offset = *ppos & 0xffff;
+
+	/* explicitly reject invalid address 0 to catch cp and cat */
+	if (i2c_addr == 0) {
+		ret = -EINVAL;
+		goto _return;
+	}
+
 	buff = kmalloc(count, GFP_KERNEL);
 	if (!buff) {
 		ret = -ENOMEM;
 		goto _return;
 	}
 
-	i2c_addr = (*ppos >> 16) & 0xff;
-	offset = *ppos & 0xffff;
-
 	total_read = i2c_read(ppd, target, i2c_addr, offset, buff, count);
 	if (total_read < 0) {
 		ret = total_read;
@@ -545,21 +676,21 @@ static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
 
 /* Do an i2c read operation on chain for HFI 0. */
 static ssize_t i2c1_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos)
+				 size_t count, loff_t *ppos)
 {
 	return __i2c_debugfs_read(file, buf, count, ppos, 0);
 }
 
 /* Do an i2c read operation on chain for HFI 1. */
 static ssize_t i2c2_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos)
+				 size_t count, loff_t *ppos)
 {
 	return __i2c_debugfs_read(file, buf, count, ppos, 1);
 }
 
 /* Do a QSFP write operation on the i2c chain for the given HFI. */
 static ssize_t __qsfp_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos, u32 target)
+				    size_t count, loff_t *ppos, u32 target)
 {
 	struct hfi1_pportdata *ppd;
 	char *buff;
@@ -605,21 +736,21 @@ static ssize_t __qsfp_debugfs_write(struct file *file, const char __user *buf,
 
 /* Do a QSFP write operation on i2c chain for HFI 0. */
 static ssize_t qsfp1_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos)
+				   size_t count, loff_t *ppos)
 {
 	return __qsfp_debugfs_write(file, buf, count, ppos, 0);
 }
 
 /* Do a QSFP write operation on i2c chain for HFI 1. */
 static ssize_t qsfp2_debugfs_write(struct file *file, const char __user *buf,
-			   size_t count, loff_t *ppos)
+				   size_t count, loff_t *ppos)
 {
 	return __qsfp_debugfs_write(file, buf, count, ppos, 1);
 }
 
 /* Do a QSFP read operation on the i2c chain for the given HFI. */
 static ssize_t __qsfp_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos, u32 target)
+				   size_t count, loff_t *ppos, u32 target)
 {
 	struct hfi1_pportdata *ppd;
 	char *buff;
@@ -665,18 +796,116 @@ static ssize_t __qsfp_debugfs_read(struct file *file, char __user *buf,
 
 /* Do a QSFP read operation on i2c chain for HFI 0. */
 static ssize_t qsfp1_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos)
+				  size_t count, loff_t *ppos)
 {
 	return __qsfp_debugfs_read(file, buf, count, ppos, 0);
 }
 
 /* Do a QSFP read operation on i2c chain for HFI 1. */
 static ssize_t qsfp2_debugfs_read(struct file *file, char __user *buf,
-			size_t count, loff_t *ppos)
+				  size_t count, loff_t *ppos)
 {
 	return __qsfp_debugfs_read(file, buf, count, ppos, 1);
 }
 
+static int __i2c_debugfs_open(struct inode *in, struct file *fp, u32 target)
+{
+	struct hfi1_pportdata *ppd;
+	int ret;
+
+	if (!try_module_get(THIS_MODULE))
+		return -ENODEV;
+
+	ppd = private2ppd(fp);
+
+	ret = acquire_chip_resource(ppd->dd, i2c_target(target), 0);
+	if (ret) /* failed - release the module */
+		module_put(THIS_MODULE);
+
+	return ret;
+}
+
+static int i2c1_debugfs_open(struct inode *in, struct file *fp)
+{
+	return __i2c_debugfs_open(in, fp, 0);
+}
+
+static int i2c2_debugfs_open(struct inode *in, struct file *fp)
+{
+	return __i2c_debugfs_open(in, fp, 1);
+}
+
+static int __i2c_debugfs_release(struct inode *in, struct file *fp, u32 target)
+{
+	struct hfi1_pportdata *ppd;
+
+	ppd = private2ppd(fp);
+
+	release_chip_resource(ppd->dd, i2c_target(target));
+	module_put(THIS_MODULE);
+
+	return 0;
+}
+
+static int i2c1_debugfs_release(struct inode *in, struct file *fp)
+{
+	return __i2c_debugfs_release(in, fp, 0);
+}
+
+static int i2c2_debugfs_release(struct inode *in, struct file *fp)
+{
+	return __i2c_debugfs_release(in, fp, 1);
+}
+
+static int __qsfp_debugfs_open(struct inode *in, struct file *fp, u32 target)
+{
+	struct hfi1_pportdata *ppd;
+	int ret;
+
+	if (!try_module_get(THIS_MODULE))
+		return -ENODEV;
+
+	ppd = private2ppd(fp);
+
+	ret = acquire_chip_resource(ppd->dd, i2c_target(target), 0);
+	if (ret) /* failed - release the module */
+		module_put(THIS_MODULE);
+
+	return ret;
+}
+
+static int qsfp1_debugfs_open(struct inode *in, struct file *fp)
+{
+	return __qsfp_debugfs_open(in, fp, 0);
+}
+
+static int qsfp2_debugfs_open(struct inode *in, struct file *fp)
+{
+	return __qsfp_debugfs_open(in, fp, 1);
+}
+
+static int __qsfp_debugfs_release(struct inode *in, struct file *fp, u32 target)
+{
+	struct hfi1_pportdata *ppd;
+
+	ppd = private2ppd(fp);
+
+	release_chip_resource(ppd->dd, i2c_target(target));
+	module_put(THIS_MODULE);
+
+	return 0;
+}
+
+static int qsfp1_debugfs_release(struct inode *in, struct file *fp)
+{
+	return __qsfp_debugfs_release(in, fp, 0);
+}
+
+static int qsfp2_debugfs_release(struct inode *in, struct file *fp)
+{
+	return __qsfp_debugfs_release(in, fp, 1);
+}
+
 #define DEBUGFS_OPS(nm, readroutine, writeroutine)	\
 { \
 	.name = nm, \
@@ -687,6 +916,18 @@ static ssize_t qsfp2_debugfs_read(struct file *file, char __user *buf,
 	}, \
 }
 
+#define DEBUGFS_XOPS(nm, readf, writef, openf, releasef) \
+{ \
+	.name = nm, \
+	.ops = { \
+		.read = readf, \
+		.write = writef, \
+		.llseek = generic_file_llseek, \
+		.open = openf, \
+		.release = releasef \
+	}, \
+}
+
 static const struct counter_info cntr_ops[] = {
 	DEBUGFS_OPS("counter_names", dev_names_read, NULL),
 	DEBUGFS_OPS("counters", dev_counters_read, NULL),
@@ -695,11 +936,16 @@ static const struct counter_info cntr_ops[] = {
 
 static const struct counter_info port_cntr_ops[] = {
 	DEBUGFS_OPS("port%dcounters", portcntrs_debugfs_read, NULL),
-	DEBUGFS_OPS("i2c1", i2c1_debugfs_read, i2c1_debugfs_write),
-	DEBUGFS_OPS("i2c2", i2c2_debugfs_read, i2c2_debugfs_write),
+	DEBUGFS_XOPS("i2c1", i2c1_debugfs_read, i2c1_debugfs_write,
+		     i2c1_debugfs_open, i2c1_debugfs_release),
+	DEBUGFS_XOPS("i2c2", i2c2_debugfs_read, i2c2_debugfs_write,
+		     i2c2_debugfs_open, i2c2_debugfs_release),
 	DEBUGFS_OPS("qsfp_dump%d", qsfp_debugfs_dump, NULL),
-	DEBUGFS_OPS("qsfp1", qsfp1_debugfs_read, qsfp1_debugfs_write),
-	DEBUGFS_OPS("qsfp2", qsfp2_debugfs_read, qsfp2_debugfs_write),
+	DEBUGFS_XOPS("qsfp1", qsfp1_debugfs_read, qsfp1_debugfs_write,
+		     qsfp1_debugfs_open, qsfp1_debugfs_release),
+	DEBUGFS_XOPS("qsfp2", qsfp2_debugfs_read, qsfp2_debugfs_write,
+		     qsfp2_debugfs_open, qsfp2_debugfs_release),
+	DEBUGFS_OPS("asic_flags", asic_flags_read, asic_flags_write),
 };
 
 void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd)
@@ -747,8 +993,8 @@ void hfi1_dbg_ibdev_init(struct hfi1_ibdev *ibd)
 					    ibd->hfi1_ibdev_dbg,
 					    ppd,
 					    &port_cntr_ops[i].ops,
-					    port_cntr_ops[i].ops.write == NULL ?
-					    S_IRUGO : S_IRUGO|S_IWUSR);
+					    !port_cntr_ops[i].ops.write ?
+					    S_IRUGO : S_IRUGO | S_IWUSR);
 		}
 }
 
diff --git a/drivers/staging/rdma/hfi1/debugfs.h b/drivers/staging/rdma/hfi1/debugfs.h
index 92d6fe146714..b6fb6814f1b8 100644
--- a/drivers/staging/rdma/hfi1/debugfs.h
+++ b/drivers/staging/rdma/hfi1/debugfs.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_DEBUGFS_H
 #define _HFI1_DEBUGFS_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
diff --git a/drivers/staging/rdma/hfi1/device.c b/drivers/staging/rdma/hfi1/device.c
index 58472e5ac4e5..c05c39da83b1 100644
--- a/drivers/staging/rdma/hfi1/device.c
+++ b/drivers/staging/rdma/hfi1/device.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
diff --git a/drivers/staging/rdma/hfi1/device.h b/drivers/staging/rdma/hfi1/device.h
index 2850ff739d81..5bb3e83cf2da 100644
--- a/drivers/staging/rdma/hfi1/device.h
+++ b/drivers/staging/rdma/hfi1/device.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_DEVICE_H
 #define _HFI1_DEVICE_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
diff --git a/drivers/staging/rdma/hfi1/diag.c b/drivers/staging/rdma/hfi1/diag.c
index e41159fe6889..c5b520bf610e 100644
--- a/drivers/staging/rdma/hfi1/diag.c
+++ b/drivers/staging/rdma/hfi1/diag.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -70,6 +67,7 @@
 #include "hfi.h"
 #include "device.h"
 #include "common.h"
+#include "verbs_txreq.h"
 #include "trace.h"
 
 #undef pr_fmt
@@ -80,15 +78,15 @@
 /* Snoop option mask */
 #define SNOOP_DROP_SEND		BIT(0)
 #define SNOOP_USE_METADATA	BIT(1)
+#define SNOOP_SET_VL0TOVL15     BIT(2)
 
 static u8 snoop_flags;
 
 /*
  * Extract packet length from LRH header.
- * Why & 0x7FF? Because len is only 11 bits in case it wasn't 0'd we throw the
- * bogus bits away. This is in Dwords so multiply by 4 to get size in bytes
+ * This is in Dwords so multiply by 4 to get size in bytes
  */
-#define HFI1_GET_PKT_LEN(x)      (((be16_to_cpu((x)->lrh[2]) & 0x7FF)) << 2)
+#define HFI1_GET_PKT_LEN(x)      (((be16_to_cpu((x)->lrh[2]) & 0xFFF)) << 2)
 
 enum hfi1_filter_status {
 	HFI1_FILTER_HIT,
@@ -860,7 +858,7 @@ static ssize_t hfi1_snoop_write(struct file *fp, const char __user *data,
 			vl = sc4;
 		} else {
 			sl = (byte_two >> 4) & 0xf;
-			ibp = to_iport(&dd->verbs_dev.ibdev, 1);
+			ibp = to_iport(&dd->verbs_dev.rdi.ibdev, 1);
 			sc5 = ibp->sl_to_sc[sl];
 			vl = sc_to_vlt(dd, sc5);
 			if (vl != sc4) {
@@ -966,6 +964,65 @@ static ssize_t hfi1_snoop_read(struct file *fp, char __user *data,
 	return ret;
 }
 
+/**
+ * hfi1_assign_snoop_link_credits -- Set up credits for VL15 and others
+ * @ppd : ptr to hfi1 port data
+ * @value : options from user space
+ *
+ * Assumes the rest of the CM credit registers are zero from a
+ * previous global or credit reset.
+ * Leave shared count at zero for both global and all vls.
+ * In snoop mode ideally we don't use shared credits
+ * Reserve 8.5k for VL15
+ * If total credits less than 8.5kbytes return error.
+ * Divide the rest of the credits across VL0 to VL7 and if
+ * each of these levels has less than 34 credits (at least 2048 + 128 bytes)
+ * return with an error.
+ * The credit registers will be reset to zero on link negotiation or link up
+ * so this function should be activated from user space only if the port has
+ * gone past link negotiation and link up.
+ *
+ * Return -- 0 if successful else error condition
+ *
+ */
+static long hfi1_assign_snoop_link_credits(struct hfi1_pportdata *ppd,
+					   int value)
+{
+#define  OPA_MIN_PER_VL_CREDITS  34  /* 2048 + 128 bytes */
+	struct buffer_control t;
+	int i;
+	struct hfi1_devdata *dd = ppd->dd;
+	u16  total_credits = (value >> 16) & 0xffff;
+	u16  vl15_credits = dd->vl15_init / 2;
+	u16  per_vl_credits;
+	__be16 be_per_vl_credits;
+
+	if (!(ppd->host_link_state & HLS_UP))
+		goto err_exit;
+	if (total_credits  <  vl15_credits)
+		goto err_exit;
+
+	per_vl_credits = (total_credits - vl15_credits) / TXE_NUM_DATA_VL;
+
+	if (per_vl_credits < OPA_MIN_PER_VL_CREDITS)
+		goto err_exit;
+
+	memset(&t, 0, sizeof(t));
+	be_per_vl_credits = cpu_to_be16(per_vl_credits);
+
+	for (i = 0; i < TXE_NUM_DATA_VL; i++)
+		t.vl[i].dedicated = be_per_vl_credits;
+
+	t.vl[15].dedicated  = cpu_to_be16(vl15_credits);
+	return set_buffer_control(ppd, &t);
+
+err_exit:
+	snoop_dbg("port_state = 0x%x, total_credits = %d, vl15_credits = %d",
+		  ppd->host_link_state, total_credits, vl15_credits);
+
+	return -EINVAL;
+}
+
 static long hfi1_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
 {
 	struct hfi1_devdata *dd;
@@ -1192,6 +1249,10 @@ static long hfi1_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
 			snoop_flags |= SNOOP_DROP_SEND;
 		if (value & SNOOP_USE_METADATA)
 			snoop_flags |= SNOOP_USE_METADATA;
+		if (value & (SNOOP_SET_VL0TOVL15)) {
+			ppd = &dd->pport[0];  /* first port will do */
+			ret = hfi1_assign_snoop_link_credits(ppd, value);
+		}
 		break;
 	default:
 		return -ENOTTY;
@@ -1603,7 +1664,7 @@ int snoop_recv_handler(struct hfi1_packet *packet)
 /*
  * Handle snooping and capturing packets when sdma is being used.
  */
-int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_dma_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			   u64 pbc)
 {
 	pr_alert("Snooping/Capture of Send DMA Packets Is Not Supported!\n");
@@ -1616,20 +1677,19 @@ int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
  * bypass packets. The only way to send a bypass packet currently is to use the
  * diagpkt interface. When that interface is enable snoop/capture is not.
  */
-int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_pio_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			   u64 pbc)
 {
-	struct ahg_ib_header *ahdr = qp->s_hdr;
 	u32 hdrwords = qp->s_hdrwords;
-	struct hfi1_sge_state *ss = qp->s_cur_sge;
+	struct rvt_sge_state *ss = qp->s_cur_sge;
 	u32 len = qp->s_cur_size;
 	u32 dwords = (len + 3) >> 2;
 	u32 plen = hdrwords + dwords + 2; /* includes pbc */
 	struct hfi1_pportdata *ppd = ps->ppd;
 	struct snoop_packet *s_packet = NULL;
-	u32 *hdr = (u32 *)&ahdr->ibh;
+	u32 *hdr = (u32 *)&ps->s_txreq->phdr.hdr;
 	u32 length = 0;
-	struct hfi1_sge_state temp_ss;
+	struct rvt_sge_state temp_ss;
 	void *data = NULL;
 	void *data_start = NULL;
 	int ret;
@@ -1638,7 +1698,7 @@ int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 	struct capture_md md;
 	u32 vl;
 	u32 hdr_len = hdrwords << 2;
-	u32 tlen = HFI1_GET_PKT_LEN(&ahdr->ibh);
+	u32 tlen = HFI1_GET_PKT_LEN(&ps->s_txreq->phdr.hdr);
 
 	md.u.pbc = 0;
 
@@ -1665,7 +1725,7 @@ int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 		md.port = 1;
 		md.dir = PKT_DIR_EGRESS;
 		if (likely(pbc == 0)) {
-			vl = be16_to_cpu(ahdr->ibh.lrh[0]) >> 12;
+			vl = be16_to_cpu(ps->s_txreq->phdr.hdr.lrh[0]) >> 12;
 			md.u.pbc = create_pbc(ppd, 0, qp->s_srate, vl, plen);
 		} else {
 			md.u.pbc = 0;
@@ -1727,7 +1787,7 @@ int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 		ret = HFI1_FILTER_HIT;
 	} else {
 		ret = ppd->dd->hfi1_snoop.filter_callback(
-					&ahdr->ibh,
+					&ps->s_txreq->phdr.hdr,
 					NULL,
 					ppd->dd->hfi1_snoop.filter_value);
 	}
@@ -1759,9 +1819,16 @@ int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 				spin_unlock_irqrestore(&qp->s_lock, flags);
 			} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 				spin_lock_irqsave(&qp->s_lock, flags);
-				hfi1_rc_send_complete(qp, &ahdr->ibh);
+				hfi1_rc_send_complete(qp,
+						      &ps->s_txreq->phdr.hdr);
 				spin_unlock_irqrestore(&qp->s_lock, flags);
 			}
+
+			/*
+			 * If snoop is dropping the packet we need to put the
+			 * txreq back because no one else will.
+			 */
+			hfi1_put_txreq(ps->s_txreq);
 			return 0;
 		}
 		break;
diff --git a/drivers/staging/rdma/hfi1/dma.c b/drivers/staging/rdma/hfi1/dma.c
index e03bd735173c..7e8dab892848 100644
--- a/drivers/staging/rdma/hfi1/dma.c
+++ b/drivers/staging/rdma/hfi1/dma.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -52,7 +49,7 @@
 
 #include "verbs.h"
 
-#define BAD_DMA_ADDRESS ((u64) 0)
+#define BAD_DMA_ADDRESS ((u64)0)
 
 /*
  * The following functions implement driver specific replacements
@@ -74,7 +71,7 @@ static u64 hfi1_dma_map_single(struct ib_device *dev, void *cpu_addr,
 	if (WARN_ON(!valid_dma_direction(direction)))
 		return BAD_DMA_ADDRESS;
 
-	return (u64) cpu_addr;
+	return (u64)cpu_addr;
 }
 
 static void hfi1_dma_unmap_single(struct ib_device *dev, u64 addr, size_t size,
@@ -95,7 +92,7 @@ static u64 hfi1_dma_map_page(struct ib_device *dev, struct page *page,
 	if (offset + size > PAGE_SIZE)
 		return BAD_DMA_ADDRESS;
 
-	addr = (u64) page_address(page);
+	addr = (u64)page_address(page);
 	if (addr)
 		addr += offset;
 
@@ -120,7 +117,7 @@ static int hfi1_map_sg(struct ib_device *dev, struct scatterlist *sgl,
 		return BAD_DMA_ADDRESS;
 
 	for_each_sg(sgl, sg, nents, i) {
-		addr = (u64) page_address(sg_page(sg));
+		addr = (u64)page_address(sg_page(sg));
 		if (!addr) {
 			ret = 0;
 			break;
@@ -161,14 +158,14 @@ static void *hfi1_dma_alloc_coherent(struct ib_device *dev, size_t size,
 	if (p)
 		addr = page_address(p);
 	if (dma_handle)
-		*dma_handle = (u64) addr;
+		*dma_handle = (u64)addr;
 	return addr;
 }
 
 static void hfi1_dma_free_coherent(struct ib_device *dev, size_t size,
 				   void *cpu_addr, u64 dma_handle)
 {
-	free_pages((unsigned long) cpu_addr, get_order(size));
+	free_pages((unsigned long)cpu_addr, get_order(size));
 }
 
 struct ib_dma_mapping_ops hfi1_dma_mapping_ops = {
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
index ee50bbf64d39..34511e5df1d5 100644
--- a/drivers/staging/rdma/hfi1/driver.c
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -56,6 +53,7 @@
 #include <linux/vmalloc.h>
 #include <linux/module.h>
 #include <linux/prefetch.h>
+#include <rdma/ib_verbs.h>
 
 #include "hfi.h"
 #include "trace.h"
@@ -162,6 +160,22 @@ const char *get_unit_name(int unit)
 	return iname;
 }
 
+const char *get_card_name(struct rvt_dev_info *rdi)
+{
+	struct hfi1_ibdev *ibdev = container_of(rdi, struct hfi1_ibdev, rdi);
+	struct hfi1_devdata *dd = container_of(ibdev,
+					       struct hfi1_devdata, verbs_dev);
+	return get_unit_name(dd->unit);
+}
+
+struct pci_dev *get_pci_dev(struct rvt_dev_info *rdi)
+{
+	struct hfi1_ibdev *ibdev = container_of(rdi, struct hfi1_ibdev, rdi);
+	struct hfi1_devdata *dd = container_of(ibdev,
+					       struct hfi1_devdata, verbs_dev);
+	return dd->pcidev;
+}
+
 /*
  * Return count of units with at least one port ACTIVE.
  */
@@ -265,6 +279,8 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 	u32 rte = rhf_rcv_type_err(packet->rhf);
 	int lnh = be16_to_cpu(rhdr->lrh[0]) & 3;
 	struct hfi1_ibport *ibp = &ppd->ibport_data;
+	struct hfi1_devdata *dd = ppd->dd;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
 
 	if (packet->rhf & (RHF_VCRC_ERR | RHF_ICRC_ERR))
 		return;
@@ -283,9 +299,9 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 			goto drop;
 
 		/* Check for GRH */
-		if (lnh == HFI1_LRH_BTH)
+		if (lnh == HFI1_LRH_BTH) {
 			ohdr = &hdr->u.oth;
-		else if (lnh == HFI1_LRH_GRH) {
+		} else if (lnh == HFI1_LRH_GRH) {
 			u32 vtf;
 
 			ohdr = &hdr->u.l.oth;
@@ -295,17 +311,17 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 			if ((vtf >> IB_GRH_VERSION_SHIFT) != IB_GRH_VERSION)
 				goto drop;
 			rcv_flags |= HFI1_HAS_GRH;
-		} else
+		} else {
 			goto drop;
-
+		}
 		/* Get the destination QP number. */
-		qp_num = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
-		if (lid < HFI1_MULTICAST_LID_BASE) {
-			struct hfi1_qp *qp;
+		qp_num = be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
+		if (lid < be16_to_cpu(IB_MULTICAST_LID_BASE)) {
+			struct rvt_qp *qp;
 			unsigned long flags;
 
 			rcu_read_lock();
-			qp = hfi1_lookup_qpn(ibp, qp_num);
+			qp = rvt_lookup_qpn(rdi, &ibp->rvp, qp_num);
 			if (!qp) {
 				rcu_read_unlock();
 				goto drop;
@@ -318,9 +334,9 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 			spin_lock_irqsave(&qp->r_lock, flags);
 
 			/* Check for valid receive state. */
-			if (!(ib_hfi1_state_ops[qp->state] &
-			      HFI1_PROCESS_RECV_OK)) {
-				ibp->n_pkt_drops++;
+			if (!(ib_rvt_state_ops[qp->state] &
+			      RVT_PROCESS_RECV_OK)) {
+				ibp->rvp.n_pkt_drops++;
 			}
 
 			switch (qp->ibqp.qp_type) {
@@ -352,7 +368,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 		if (rhf_use_egr_bfr(packet->rhf))
 			ebuf = packet->ebuf;
 
-		if (ebuf == NULL)
+		if (!ebuf)
 			goto drop; /* this should never happen */
 
 		if (lnh == HFI1_LRH_BTH)
@@ -370,7 +386,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 			 * Only in pre-B0 h/w is the CNP_OPCODE handled
 			 * via this code path.
 			 */
-			struct hfi1_qp *qp = NULL;
+			struct rvt_qp *qp = NULL;
 			u32 lqpn, rqpn;
 			u16 rlid;
 			u8 svc_type, sl, sc5;
@@ -380,10 +396,10 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 				sc5 |= 0x10;
 			sl = ibp->sc_to_sl[sc5];
 
-			lqpn = be32_to_cpu(bth[1]) & HFI1_QPN_MASK;
+			lqpn = be32_to_cpu(bth[1]) & RVT_QPN_MASK;
 			rcu_read_lock();
-			qp = hfi1_lookup_qpn(ibp, lqpn);
-			if (qp == NULL) {
+			qp = rvt_lookup_qpn(rdi, &ibp->rvp, lqpn);
+			if (!qp) {
 				rcu_read_unlock();
 				goto drop;
 			}
@@ -419,9 +435,8 @@ drop:
 }
 
 static inline void init_packet(struct hfi1_ctxtdata *rcd,
-			      struct hfi1_packet *packet)
+			       struct hfi1_packet *packet)
 {
-
 	packet->rsize = rcd->rcvhdrqentsize; /* words */
 	packet->maxcnt = rcd->rcvhdrq_cnt * packet->rsize; /* words */
 	packet->rcd = rcd;
@@ -434,12 +449,7 @@ static inline void init_packet(struct hfi1_ctxtdata *rcd,
 	packet->rcv_flags = 0;
 }
 
-#ifndef CONFIG_PRESCAN_RXQ
-static void prescan_rxq(struct hfi1_packet *packet) {}
-#else /* !CONFIG_PRESCAN_RXQ */
-static int prescan_receive_queue;
-
-static void process_ecn(struct hfi1_qp *qp, struct hfi1_ib_header *hdr,
+static void process_ecn(struct rvt_qp *qp, struct hfi1_ib_header *hdr,
 			struct hfi1_other_headers *ohdr,
 			u64 rhf, u32 bth1, struct ib_grh *grh)
 {
@@ -453,7 +463,7 @@ static void process_ecn(struct hfi1_qp *qp, struct hfi1_ib_header *hdr,
 	case IB_QPT_GSI:
 	case IB_QPT_UD:
 		rlid = be16_to_cpu(hdr->lrh[3]);
-		rqpn = be32_to_cpu(ohdr->u.ud.deth[1]) & HFI1_QPN_MASK;
+		rqpn = be32_to_cpu(ohdr->u.ud.deth[1]) & RVT_QPN_MASK;
 		svc_type = IB_CC_SVCTYPE_UD;
 		break;
 	case IB_QPT_UC:
@@ -483,7 +493,7 @@ static void process_ecn(struct hfi1_qp *qp, struct hfi1_ib_header *hdr,
 
 	if (bth1 & HFI1_BECN_SMASK) {
 		struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
-		u32 lqpn = bth1 & HFI1_QPN_MASK;
+		u32 lqpn = bth1 & RVT_QPN_MASK;
 		u8 sl = ibp->sc_to_sl[sc5];
 
 		process_becn(ppd, sl, rlid, lqpn, rqpn, svc_type);
@@ -562,26 +572,31 @@ static inline void update_ps_mdata(struct ps_mdata *mdata,
  * containing Excplicit Congestion Notifications (FECNs, or BECNs).
  * When an ECN is found, process the Congestion Notification, and toggle
  * it off.
+ * This is declared as a macro to allow quick checking of the port to avoid
+ * the overhead of a function call if not enabled.
  */
-static void prescan_rxq(struct hfi1_packet *packet)
+#define prescan_rxq(rcd, packet) \
+	do { \
+		if (rcd->ppd->cc_prescan) \
+			__prescan_rxq(packet); \
+	} while (0)
+static void __prescan_rxq(struct hfi1_packet *packet)
 {
 	struct hfi1_ctxtdata *rcd = packet->rcd;
 	struct ps_mdata mdata;
 
-	if (!prescan_receive_queue)
-		return;
-
 	init_ps_mdata(&mdata, packet);
 
 	while (1) {
 		struct hfi1_devdata *dd = rcd->dd;
 		struct hfi1_ibport *ibp = &rcd->ppd->ibport_data;
-		__le32 *rhf_addr = (__le32 *) rcd->rcvhdrq + mdata.ps_head +
+		__le32 *rhf_addr = (__le32 *)rcd->rcvhdrq + mdata.ps_head +
 					 dd->rhf_offset;
-		struct hfi1_qp *qp;
+		struct rvt_qp *qp;
 		struct hfi1_ib_header *hdr;
 		struct hfi1_other_headers *ohdr;
 		struct ib_grh *grh = NULL;
+		struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
 		u64 rhf = rhf_to_cpu(rhf_addr);
 		u32 etype = rhf_rcv_type(rhf), qpn, bth1;
 		int is_ecn = 0;
@@ -600,25 +615,25 @@ static void prescan_rxq(struct hfi1_packet *packet)
 			hfi1_get_msgheader(dd, rhf_addr);
 		lnh = be16_to_cpu(hdr->lrh[0]) & 3;
 
-		if (lnh == HFI1_LRH_BTH)
+		if (lnh == HFI1_LRH_BTH) {
 			ohdr = &hdr->u.oth;
-		else if (lnh == HFI1_LRH_GRH) {
+		} else if (lnh == HFI1_LRH_GRH) {
 			ohdr = &hdr->u.l.oth;
 			grh = &hdr->u.l.grh;
-		} else
+		} else {
 			goto next; /* just in case */
-
+		}
 		bth1 = be32_to_cpu(ohdr->bth[1]);
 		is_ecn = !!(bth1 & (HFI1_FECN_SMASK | HFI1_BECN_SMASK));
 
 		if (!is_ecn)
 			goto next;
 
-		qpn = bth1 & HFI1_QPN_MASK;
+		qpn = bth1 & RVT_QPN_MASK;
 		rcu_read_lock();
-		qp = hfi1_lookup_qpn(ibp, qpn);
+		qp = rvt_lookup_qpn(rdi, &ibp->rvp, qpn);
 
-		if (qp == NULL) {
+		if (!qp) {
 			rcu_read_unlock();
 			goto next;
 		}
@@ -633,7 +648,6 @@ next:
 		update_ps_mdata(&mdata, rcd);
 	}
 }
-#endif /* CONFIG_PRESCAN_RXQ */
 
 static inline int skip_rcv_packet(struct hfi1_packet *packet, int thread)
 {
@@ -683,8 +697,9 @@ static inline int process_rcv_packet(struct hfi1_packet *packet, int thread)
 		 * The +2 is the size of the RHF.
 		 */
 		prefetch_range(packet->ebuf,
-			packet->tlen - ((packet->rcd->rcvhdrqentsize -
-				  (rhf_hdrq_offset(packet->rhf)+2)) * 4));
+			       packet->tlen - ((packet->rcd->rcvhdrqentsize -
+					       (rhf_hdrq_offset(packet->rhf)
+						+ 2)) * 4));
 	}
 
 	/*
@@ -712,7 +727,7 @@ static inline int process_rcv_packet(struct hfi1_packet *packet, int thread)
 		}
 	}
 
-	packet->rhf_addr = (__le32 *) packet->rcd->rcvhdrq + packet->rhqoff +
+	packet->rhf_addr = (__le32 *)packet->rcd->rcvhdrq + packet->rhqoff +
 				      packet->rcd->dd->rhf_offset;
 	packet->rhf = rhf_to_cpu(packet->rhf_addr);
 
@@ -737,7 +752,6 @@ static inline void process_rcv_update(int last, struct hfi1_packet *packet)
 
 static inline void finish_packet(struct hfi1_packet *packet)
 {
-
 	/*
 	 * Nothing we need to free for the packet.
 	 *
@@ -746,14 +760,12 @@ static inline void finish_packet(struct hfi1_packet *packet)
 	 */
 	update_usrhead(packet->rcd, packet->rcd->head, packet->updegr,
 		       packet->etail, rcv_intr_dynamic, packet->numpkt);
-
 }
 
 static inline void process_rcv_qp_work(struct hfi1_packet *packet)
 {
-
 	struct hfi1_ctxtdata *rcd;
-	struct hfi1_qp *qp, *nqp;
+	struct rvt_qp *qp, *nqp;
 
 	rcd = packet->rcd;
 	rcd->head = packet->rhqoff;
@@ -764,17 +776,17 @@ static inline void process_rcv_qp_work(struct hfi1_packet *packet)
 	 */
 	list_for_each_entry_safe(qp, nqp, &rcd->qp_wait_list, rspwait) {
 		list_del_init(&qp->rspwait);
-		if (qp->r_flags & HFI1_R_RSP_DEFERED_ACK) {
-			qp->r_flags &= ~HFI1_R_RSP_DEFERED_ACK;
+		if (qp->r_flags & RVT_R_RSP_NAK) {
+			qp->r_flags &= ~RVT_R_RSP_NAK;
 			hfi1_send_rc_ack(rcd, qp, 0);
 		}
-		if (qp->r_flags & HFI1_R_RSP_SEND) {
+		if (qp->r_flags & RVT_R_RSP_SEND) {
 			unsigned long flags;
 
-			qp->r_flags &= ~HFI1_R_RSP_SEND;
+			qp->r_flags &= ~RVT_R_RSP_SEND;
 			spin_lock_irqsave(&qp->s_lock, flags);
-			if (ib_hfi1_state_ops[qp->state] &
-					HFI1_PROCESS_OR_FLUSH_SEND)
+			if (ib_rvt_state_ops[qp->state] &
+					RVT_PROCESS_OR_FLUSH_SEND)
 				hfi1_schedule_send(qp);
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 		}
@@ -799,7 +811,7 @@ int handle_receive_interrupt_nodma_rtail(struct hfi1_ctxtdata *rcd, int thread)
 		goto bail;
 	}
 
-	prescan_rxq(&packet);
+	prescan_rxq(rcd, &packet);
 
 	while (last == RCV_PKT_OK) {
 		last = process_rcv_packet(&packet, thread);
@@ -830,7 +842,7 @@ int handle_receive_interrupt_dma_rtail(struct hfi1_ctxtdata *rcd, int thread)
 	}
 	smp_rmb();  /* prevent speculative reads of dma'ed hdrq */
 
-	prescan_rxq(&packet);
+	prescan_rxq(rcd, &packet);
 
 	while (last == RCV_PKT_OK) {
 		last = process_rcv_packet(&packet, thread);
@@ -862,6 +874,37 @@ static inline void set_all_dma_rtail(struct hfi1_devdata *dd)
 			&handle_receive_interrupt_dma_rtail;
 }
 
+void set_all_slowpath(struct hfi1_devdata *dd)
+{
+	int i;
+
+	/* HFI1_CTRL_CTXT must always use the slow path interrupt handler */
+	for (i = HFI1_CTRL_CTXT + 1; i < dd->first_user_ctxt; i++)
+		dd->rcd[i]->do_interrupt = &handle_receive_interrupt;
+}
+
+static inline int set_armed_to_active(struct hfi1_ctxtdata *rcd,
+				      struct hfi1_packet packet,
+				      struct hfi1_devdata *dd)
+{
+	struct work_struct *lsaw = &rcd->ppd->linkstate_active_work;
+	struct hfi1_message_header *hdr = hfi1_get_msgheader(packet.rcd->dd,
+							     packet.rhf_addr);
+
+	if (hdr2sc(hdr, packet.rhf) != 0xf) {
+		int hwstate = read_logical_state(dd);
+
+		if (hwstate != LSTATE_ACTIVE) {
+			dd_dev_info(dd, "Unexpected link state %d\n", hwstate);
+			return 0;
+		}
+
+		queue_work(rcd->ppd->hfi1_wq, lsaw);
+		return 1;
+	}
+	return 0;
+}
+
 /*
  * handle_receive_interrupt - receive a packet
  * @rcd: the context
@@ -910,17 +953,17 @@ int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread)
 		}
 	}
 
-	prescan_rxq(&packet);
+	prescan_rxq(rcd, &packet);
 
 	while (last == RCV_PKT_OK) {
-
-		if (unlikely(dd->do_drop && atomic_xchg(&dd->drop_packet,
-			DROP_PACKET_OFF) == DROP_PACKET_ON)) {
+		if (unlikely(dd->do_drop &&
+			     atomic_xchg(&dd->drop_packet, DROP_PACKET_OFF) ==
+			     DROP_PACKET_ON)) {
 			dd->do_drop = 0;
 
 			/* On to the next packet */
 			packet.rhqoff += packet.rsize;
-			packet.rhf_addr = (__le32 *) rcd->rcvhdrq +
+			packet.rhf_addr = (__le32 *)rcd->rcvhdrq +
 					  packet.rhqoff +
 					  dd->rhf_offset;
 			packet.rhf = rhf_to_cpu(packet.rhf_addr);
@@ -929,6 +972,11 @@ int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread)
 			last = skip_rcv_packet(&packet, thread);
 			skip_pkt = 0;
 		} else {
+			/* Auto activate link on non-SC15 packet receive */
+			if (unlikely(rcd->ppd->host_link_state ==
+				     HLS_UP_ARMED) &&
+			    set_armed_to_active(rcd, packet, dd))
+				goto bail;
 			last = process_rcv_packet(&packet, thread);
 		}
 
@@ -940,8 +988,7 @@ int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread)
 			if (seq != rcd->seq_cnt)
 				last = RCV_PKT_DONE;
 			if (needset) {
-				dd_dev_info(dd,
-					"Switching to NO_DMA_RTAIL\n");
+				dd_dev_info(dd, "Switching to NO_DMA_RTAIL\n");
 				set_all_nodma_rtail(dd);
 				needset = 0;
 			}
@@ -984,6 +1031,42 @@ bail:
 }
 
 /*
+ * We may discover in the interrupt that the hardware link state has
+ * changed from ARMED to ACTIVE (due to the arrival of a non-SC15 packet),
+ * and we need to update the driver's notion of the link state.  We cannot
+ * run set_link_state from interrupt context, so we queue this function on
+ * a workqueue.
+ *
+ * We delay the regular interrupt processing until after the state changes
+ * so that the link will be in the correct state by the time any application
+ * we wake up attempts to send a reply to any message it received.
+ * (Subsequent receive interrupts may possibly force the wakeup before we
+ * update the link state.)
+ *
+ * The rcd is freed in hfi1_free_ctxtdata after hfi1_postinit_cleanup invokes
+ * dd->f_cleanup(dd) to disable the interrupt handler and flush workqueues,
+ * so we're safe from use-after-free of the rcd.
+ */
+void receive_interrupt_work(struct work_struct *work)
+{
+	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+						  linkstate_active_work);
+	struct hfi1_devdata *dd = ppd->dd;
+	int i;
+
+	/* Received non-SC15 packet implies neighbor_normal */
+	ppd->neighbor_normal = 1;
+	set_link_state(ppd, HLS_UP_ACTIVE);
+
+	/*
+	 * Interrupt all kernel contexts that could have had an
+	 * interrupt during auto activation.
+	 */
+	for (i = HFI1_CTRL_CTXT; i < dd->first_user_ctxt; i++)
+		force_recv_intr(dd->rcd[i]);
+}
+
+/*
  * Convert a given MTU size to the on-wire MAD packet enumeration.
  * Return -1 if the size is invalid.
  */
@@ -1037,9 +1120,9 @@ int set_mtu(struct hfi1_pportdata *ppd)
 	ppd->ibmaxlen = ppd->ibmtu + lrh_max_header_bytes(ppd->dd);
 
 	mutex_lock(&ppd->hls_lock);
-	if (ppd->host_link_state == HLS_UP_INIT
-			|| ppd->host_link_state == HLS_UP_ARMED
-			|| ppd->host_link_state == HLS_UP_ACTIVE)
+	if (ppd->host_link_state == HLS_UP_INIT ||
+	    ppd->host_link_state == HLS_UP_ARMED ||
+	    ppd->host_link_state == HLS_UP_ACTIVE)
 		is_up = 1;
 
 	drain = !is_ax(dd) && is_up;
@@ -1082,79 +1165,80 @@ int hfi1_set_lid(struct hfi1_pportdata *ppd, u32 lid, u8 lmc)
 	return 0;
 }
 
-/*
- * Following deal with the "obviously simple" task of overriding the state
- * of the LEDs, which normally indicate link physical and logical status.
- * The complications arise in dealing with different hardware mappings
- * and the board-dependent routine being called from interrupts.
- * and then there's the requirement to _flash_ them.
- */
-#define LED_OVER_FREQ_SHIFT 8
-#define LED_OVER_FREQ_MASK (0xFF<<LED_OVER_FREQ_SHIFT)
-/* Below is "non-zero" to force override, but both actual LEDs are off */
-#define LED_OVER_BOTH_OFF (8)
+void shutdown_led_override(struct hfi1_pportdata *ppd)
+{
+	struct hfi1_devdata *dd = ppd->dd;
+
+	/*
+	 * This pairs with the memory barrier in hfi1_start_led_override to
+	 * ensure that we read the correct state of LED beaconing represented
+	 * by led_override_timer_active
+	 */
+	smp_rmb();
+	if (atomic_read(&ppd->led_override_timer_active)) {
+		del_timer_sync(&ppd->led_override_timer);
+		atomic_set(&ppd->led_override_timer_active, 0);
+		/* Ensure the atomic_set is visible to all CPUs */
+		smp_wmb();
+	}
+
+	/* Hand control of the LED to the DC for normal operation */
+	write_csr(dd, DCC_CFG_LED_CNTRL, 0);
+}
 
 static void run_led_override(unsigned long opaque)
 {
 	struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)opaque;
 	struct hfi1_devdata *dd = ppd->dd;
-	int timeoff;
-	int ph_idx;
+	unsigned long timeout;
+	int phase_idx;
 
 	if (!(dd->flags & HFI1_INITTED))
 		return;
 
-	ph_idx = ppd->led_override_phase++ & 1;
-	ppd->led_override = ppd->led_override_vals[ph_idx];
-	timeoff = ppd->led_override_timeoff;
+	phase_idx = ppd->led_override_phase & 1;
 
-	/*
-	 * don't re-fire the timer if user asked for it to be off; we let
-	 * it fire one more time after they turn it off to simplify
-	 */
-	if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
-		mod_timer(&ppd->led_override_timer, jiffies + timeoff);
+	setextled(dd, phase_idx);
+
+	timeout = ppd->led_override_vals[phase_idx];
+
+	/* Set up for next phase */
+	ppd->led_override_phase = !ppd->led_override_phase;
+
+	mod_timer(&ppd->led_override_timer, jiffies + timeout);
 }
 
-void hfi1_set_led_override(struct hfi1_pportdata *ppd, unsigned int val)
+/*
+ * To have the LED blink in a particular pattern, provide timeon and timeoff
+ * in milliseconds.
+ * To turn off custom blinking and return to normal operation, use
+ * shutdown_led_override()
+ */
+void hfi1_start_led_override(struct hfi1_pportdata *ppd, unsigned int timeon,
+			     unsigned int timeoff)
 {
-	struct hfi1_devdata *dd = ppd->dd;
-	int timeoff, freq;
-
-	if (!(dd->flags & HFI1_INITTED))
+	if (!(ppd->dd->flags & HFI1_INITTED))
 		return;
 
-	/* First check if we are blinking. If not, use 1HZ polling */
-	timeoff = HZ;
-	freq = (val & LED_OVER_FREQ_MASK) >> LED_OVER_FREQ_SHIFT;
+	/* Convert to jiffies for direct use in timer */
+	ppd->led_override_vals[0] = msecs_to_jiffies(timeoff);
+	ppd->led_override_vals[1] = msecs_to_jiffies(timeon);
 
-	if (freq) {
-		/* For blink, set each phase from one nybble of val */
-		ppd->led_override_vals[0] = val & 0xF;
-		ppd->led_override_vals[1] = (val >> 4) & 0xF;
-		timeoff = (HZ << 4)/freq;
-	} else {
-		/* Non-blink set both phases the same. */
-		ppd->led_override_vals[0] = val & 0xF;
-		ppd->led_override_vals[1] = val & 0xF;
-	}
-	ppd->led_override_timeoff = timeoff;
+	/* Arbitrarily start from LED on phase */
+	ppd->led_override_phase = 1;
 
 	/*
 	 * If the timer has not already been started, do so. Use a "quick"
-	 * timeout so the function will be called soon, to look at our request.
+	 * timeout so the handler will be called soon to look at our request.
 	 */
-	if (atomic_inc_return(&ppd->led_override_timer_active) == 1) {
-		/* Need to start timer */
+	if (!timer_pending(&ppd->led_override_timer)) {
 		setup_timer(&ppd->led_override_timer, run_led_override,
-				(unsigned long)ppd);
-
+			    (unsigned long)ppd);
 		ppd->led_override_timer.expires = jiffies + 1;
 		add_timer(&ppd->led_override_timer);
-	} else {
-		if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
-			mod_timer(&ppd->led_override_timer, jiffies + 1);
-		atomic_dec(&ppd->led_override_timer_active);
+		atomic_set(&ppd->led_override_timer_active, 1);
+		/* Ensure the atomic_set is visible to all CPUs */
+		smp_wmb();
 	}
 }
 
@@ -1184,8 +1268,8 @@ int hfi1_reset_device(int unit)
 
 	if (!dd->kregbase || !(dd->flags & HFI1_PRESENT)) {
 		dd_dev_info(dd,
-			"Invalid unit number %u or not initialized or not present\n",
-			unit);
+			    "Invalid unit number %u or not initialized or not present\n",
+			    unit);
 		ret = -ENXIO;
 		goto bail;
 	}
@@ -1203,14 +1287,8 @@ int hfi1_reset_device(int unit)
 
 	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
 		ppd = dd->pport + pidx;
-		if (atomic_read(&ppd->led_override_timer_active)) {
-			/* Need to stop LED timer, _then_ shut off LEDs */
-			del_timer_sync(&ppd->led_override_timer);
-			atomic_set(&ppd->led_override_timer_active, 0);
-		}
 
-		/* Shut off LEDs after we are sure timer is not running */
-		ppd->led_override = LED_OVER_BOTH_OFF;
+		shutdown_led_override(ppd);
 	}
 	if (dd->flags & HFI1_HAS_SEND_DMA)
 		sdma_exit(dd);
@@ -1221,11 +1299,11 @@ int hfi1_reset_device(int unit)
 
 	if (ret)
 		dd_dev_err(dd,
-			"Reinitialize unit %u after reset failed with %d\n",
-			unit, ret);
+			   "Reinitialize unit %u after reset failed with %d\n",
+			   unit, ret);
 	else
 		dd_dev_info(dd, "Reinitialized unit %u after resetting\n",
-			unit);
+			    unit);
 
 bail:
 	return ret;
@@ -1282,7 +1360,7 @@ int process_receive_bypass(struct hfi1_packet *packet)
 		handle_eflags(packet);
 
 	dd_dev_err(packet->rcd->dd,
-	   "Bypass packets are not supported in normal operation. Dropping\n");
+		   "Bypass packets are not supported in normal operation. Dropping\n");
 	return RHF_RCV_CONTINUE;
 }
 
@@ -1320,6 +1398,6 @@ int kdeth_process_eager(struct hfi1_packet *packet)
 int process_receive_invalid(struct hfi1_packet *packet)
 {
 	dd_dev_err(packet->rcd->dd, "Invalid packet type %d. Dropping\n",
-		rhf_rcv_type(packet->rhf));
+		   rhf_rcv_type(packet->rhf));
 	return RHF_RCV_CONTINUE;
 }
diff --git a/drivers/staging/rdma/hfi1/efivar.c b/drivers/staging/rdma/hfi1/efivar.c
index 47dfe2584760..106349fc1fb9 100644
--- a/drivers/staging/rdma/hfi1/efivar.c
+++ b/drivers/staging/rdma/hfi1/efivar.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
diff --git a/drivers/staging/rdma/hfi1/efivar.h b/drivers/staging/rdma/hfi1/efivar.h
index 070706225c51..94e9e70de568 100644
--- a/drivers/staging/rdma/hfi1/efivar.h
+++ b/drivers/staging/rdma/hfi1/efivar.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
diff --git a/drivers/staging/rdma/hfi1/eprom.c b/drivers/staging/rdma/hfi1/eprom.c
index fb620c97f592..bd8771570f81 100644
--- a/drivers/staging/rdma/hfi1/eprom.c
+++ b/drivers/staging/rdma/hfi1/eprom.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -99,17 +96,17 @@
 
 /* sleep length while waiting for controller */
 #define WAIT_SLEEP_US 100	/* must be larger than 5 (see usage) */
-#define COUNT_DELAY_SEC(n) ((n) * (1000000/WAIT_SLEEP_US))
+#define COUNT_DELAY_SEC(n) ((n) * (1000000 / WAIT_SLEEP_US))
 
 /* GPIO pins */
-#define EPROM_WP_N (1ull << 14)	/* EPROM write line */
+#define EPROM_WP_N BIT_ULL(14)	/* EPROM write line */
 
 /*
- * Use the EP mutex to guard against other callers from within the driver.
- * Also covers usage of eprom_available.
+ * How long to wait for the EPROM to become available, in ms.
+ * The spec 32 Mb EPROM takes around 40s to erase then write.
+ * Double it for safety.
  */
-static DEFINE_MUTEX(eprom_mutex);
-static int eprom_available;	/* default: not available */
+#define EPROM_TIMEOUT 80000 /* ms */
 
 /*
  * Turn on external enable line that allows writing on the flash.
@@ -117,11 +114,9 @@ static int eprom_available;	/* default: not available */
 static void write_enable(struct hfi1_devdata *dd)
 {
 	/* raise signal */
-	write_csr(dd, ASIC_GPIO_OUT,
-		read_csr(dd, ASIC_GPIO_OUT) | EPROM_WP_N);
+	write_csr(dd, ASIC_GPIO_OUT, read_csr(dd, ASIC_GPIO_OUT) | EPROM_WP_N);
 	/* raise enable */
-	write_csr(dd, ASIC_GPIO_OE,
-		read_csr(dd, ASIC_GPIO_OE) | EPROM_WP_N);
+	write_csr(dd, ASIC_GPIO_OE, read_csr(dd, ASIC_GPIO_OE) | EPROM_WP_N);
 }
 
 /*
@@ -130,11 +125,9 @@ static void write_enable(struct hfi1_devdata *dd)
 static void write_disable(struct hfi1_devdata *dd)
 {
 	/* lower signal */
-	write_csr(dd, ASIC_GPIO_OUT,
-		read_csr(dd, ASIC_GPIO_OUT) & ~EPROM_WP_N);
+	write_csr(dd, ASIC_GPIO_OUT, read_csr(dd, ASIC_GPIO_OUT) & ~EPROM_WP_N);
 	/* lower enable */
-	write_csr(dd, ASIC_GPIO_OE,
-		read_csr(dd, ASIC_GPIO_OE) & ~EPROM_WP_N);
+	write_csr(dd, ASIC_GPIO_OE, read_csr(dd, ASIC_GPIO_OE) & ~EPROM_WP_N);
 }
 
 /*
@@ -212,8 +205,8 @@ static int erase_range(struct hfi1_devdata *dd, u32 start, u32 len)
 	/* check the end points for the minimum erase */
 	if ((start & MASK_4KB) || (end & MASK_4KB)) {
 		dd_dev_err(dd,
-			"%s: non-aligned range (0x%x,0x%x) for a 4KB erase\n",
-			__func__, start, end);
+			   "%s: non-aligned range (0x%x,0x%x) for a 4KB erase\n",
+			   __func__, start, end);
 		return -EINVAL;
 	}
 
@@ -256,7 +249,7 @@ static void read_page(struct hfi1_devdata *dd, u32 offset, u32 *result)
 	int i;
 
 	write_csr(dd, ASIC_EEP_ADDR_CMD, CMD_READ_DATA(offset));
-	for (i = 0; i < EP_PAGE_SIZE/sizeof(u32); i++)
+	for (i = 0; i < EP_PAGE_SIZE / sizeof(u32); i++)
 		result[i] = (u32)read_csr(dd, ASIC_EEP_DATA);
 	write_csr(dd, ASIC_EEP_ADDR_CMD, CMD_NOP); /* close open page */
 }
@@ -267,7 +260,7 @@ static void read_page(struct hfi1_devdata *dd, u32 offset, u32 *result)
 static int read_length(struct hfi1_devdata *dd, u32 start, u32 len, u64 addr)
 {
 	u32 offset;
-	u32 buffer[EP_PAGE_SIZE/sizeof(u32)];
+	u32 buffer[EP_PAGE_SIZE / sizeof(u32)];
 	int ret = 0;
 
 	/* reject anything not on an EPROM page boundary */
@@ -277,7 +270,7 @@ static int read_length(struct hfi1_devdata *dd, u32 start, u32 len, u64 addr)
 	for (offset = 0; offset < len; offset += EP_PAGE_SIZE) {
 		read_page(dd, start + offset, buffer);
 		if (copy_to_user((void __user *)(addr + offset),
-						buffer, EP_PAGE_SIZE)) {
+				 buffer, EP_PAGE_SIZE)) {
 			ret = -EFAULT;
 			goto done;
 		}
@@ -298,7 +291,7 @@ static int write_page(struct hfi1_devdata *dd, u32 offset, u32 *data)
 	write_csr(dd, ASIC_EEP_ADDR_CMD, CMD_WRITE_ENABLE);
 	write_csr(dd, ASIC_EEP_DATA, data[0]);
 	write_csr(dd, ASIC_EEP_ADDR_CMD, CMD_PAGE_PROGRAM(offset));
-	for (i = 1; i < EP_PAGE_SIZE/sizeof(u32); i++)
+	for (i = 1; i < EP_PAGE_SIZE / sizeof(u32); i++)
 		write_csr(dd, ASIC_EEP_DATA, data[i]);
 	/* will close the open page */
 	return wait_for_not_busy(dd);
@@ -310,7 +303,7 @@ static int write_page(struct hfi1_devdata *dd, u32 offset, u32 *data)
 static int write_length(struct hfi1_devdata *dd, u32 start, u32 len, u64 addr)
 {
 	u32 offset;
-	u32 buffer[EP_PAGE_SIZE/sizeof(u32)];
+	u32 buffer[EP_PAGE_SIZE / sizeof(u32)];
 	int ret = 0;
 
 	/* reject anything not on an EPROM page boundary */
@@ -321,7 +314,7 @@ static int write_length(struct hfi1_devdata *dd, u32 start, u32 len, u64 addr)
 
 	for (offset = 0; offset < len; offset += EP_PAGE_SIZE) {
 		if (copy_from_user(buffer, (void __user *)(addr + offset),
-						EP_PAGE_SIZE)) {
+				   EP_PAGE_SIZE)) {
 			ret = -EFAULT;
 			goto done;
 		}
@@ -353,44 +346,42 @@ static inline u32 extract_rstart(u32 composite)
  *
  * Return 0 on success, -ERRNO on error
  */
-int handle_eprom_command(const struct hfi1_cmd *cmd)
+int handle_eprom_command(struct file *fp, const struct hfi1_cmd *cmd)
 {
 	struct hfi1_devdata *dd;
 	u32 dev_id;
 	u32 rlen;	/* range length */
 	u32 rstart;	/* range start */
+	int i_minor;
 	int ret = 0;
 
 	/*
-	 * The EPROM is per-device, so use unit 0 as that will always
-	 * exist.
+	 * Map the device file to device data using the relative minor.
+	 * The device file minor number is the unit number + 1.  0 is
+	 * the generic device file - reject it.
 	 */
-	dd = hfi1_lookup(0);
+	i_minor = iminor(file_inode(fp)) - HFI1_USER_MINOR_BASE;
+	if (i_minor <= 0)
+		return -EINVAL;
+	dd = hfi1_lookup(i_minor - 1);
 	if (!dd) {
-		pr_err("%s: cannot find unit 0!\n", __func__);
+		pr_err("%s: cannot find unit %d!\n", __func__, i_minor);
 		return -EINVAL;
 	}
 
-	/* lock against other callers touching the ASIC block */
-	mutex_lock(&eprom_mutex);
-
-	/* some platforms do not have an EPROM */
-	if (!eprom_available) {
-		ret = -ENOSYS;
-		goto done_asic;
-	}
+	/* some devices do not have an EPROM */
+	if (!dd->eprom_available)
+		return -EOPNOTSUPP;
 
-	/* lock against the other HFI on another OS */
-	ret = acquire_hw_mutex(dd);
+	ret = acquire_chip_resource(dd, CR_EPROM, EPROM_TIMEOUT);
 	if (ret) {
-		dd_dev_err(dd,
-			"%s: unable to acquire hw mutex, no EPROM support\n",
-			__func__);
+		dd_dev_err(dd, "%s: unable to acquire EPROM resource\n",
+			   __func__);
 		goto done_asic;
 	}
 
 	dd_dev_info(dd, "%s: cmd: type %d, len 0x%x, addr 0x%016llx\n",
-		__func__, cmd->type, cmd->len, cmd->addr);
+		    __func__, cmd->type, cmd->len, cmd->addr);
 
 	switch (cmd->type) {
 	case HFI1_CMD_EP_INFO:
@@ -401,7 +392,7 @@ int handle_eprom_command(const struct hfi1_cmd *cmd)
 		dev_id = read_device_id(dd);
 		/* addr points to a u32 user buffer */
 		if (copy_to_user((void __user *)cmd->addr, &dev_id,
-								sizeof(u32)))
+				 sizeof(u32)))
 			ret = -EFAULT;
 		break;
 
@@ -429,14 +420,13 @@ int handle_eprom_command(const struct hfi1_cmd *cmd)
 
 	default:
 		dd_dev_err(dd, "%s: unexpected command %d\n",
-			__func__, cmd->type);
+			   __func__, cmd->type);
 		ret = -EINVAL;
 		break;
 	}
 
-	release_hw_mutex(dd);
+	release_chip_resource(dd, CR_EPROM);
 done_asic:
-	mutex_unlock(&eprom_mutex);
 	return ret;
 }
 
@@ -447,44 +437,35 @@ int eprom_init(struct hfi1_devdata *dd)
 {
 	int ret = 0;
 
-	/* only the discrete chip has an EPROM, nothing to do */
+	/* only the discrete chip has an EPROM */
 	if (dd->pcidev->device != PCI_DEVICE_ID_INTEL0)
 		return 0;
 
-	/* lock against other callers */
-	mutex_lock(&eprom_mutex);
-	if (eprom_available)	/* already initialized */
-		goto done_asic;
-
 	/*
-	 * Lock against the other HFI on another OS - the mutex above
-	 * would have caught anything in this driver.  It is OK if
-	 * both OSes reset the EPROM - as long as they don't do it at
-	 * the same time.
+	 * It is OK if both HFIs reset the EPROM as long as they don't
+	 * do it at the same time.
 	 */
-	ret = acquire_hw_mutex(dd);
+	ret = acquire_chip_resource(dd, CR_EPROM, EPROM_TIMEOUT);
 	if (ret) {
 		dd_dev_err(dd,
-			"%s: unable to acquire hw mutex, no EPROM support\n",
-			__func__);
+			   "%s: unable to acquire EPROM resource, no EPROM support\n",
+			   __func__);
 		goto done_asic;
 	}
 
 	/* reset EPROM to be sure it is in a good state */
 
 	/* set reset */
-	write_csr(dd, ASIC_EEP_CTL_STAT,
-					ASIC_EEP_CTL_STAT_EP_RESET_SMASK);
+	write_csr(dd, ASIC_EEP_CTL_STAT, ASIC_EEP_CTL_STAT_EP_RESET_SMASK);
 	/* clear reset, set speed */
 	write_csr(dd, ASIC_EEP_CTL_STAT,
-			EP_SPEED_FULL << ASIC_EEP_CTL_STAT_RATE_SPI_SHIFT);
+		  EP_SPEED_FULL << ASIC_EEP_CTL_STAT_RATE_SPI_SHIFT);
 
 	/* wake the device with command "release powerdown NoID" */
 	write_csr(dd, ASIC_EEP_ADDR_CMD, CMD_RELEASE_POWERDOWN_NOID);
 
-	eprom_available = 1;
-	release_hw_mutex(dd);
+	dd->eprom_available = true;
+	release_chip_resource(dd, CR_EPROM);
 done_asic:
-	mutex_unlock(&eprom_mutex);
 	return ret;
 }
diff --git a/drivers/staging/rdma/hfi1/eprom.h b/drivers/staging/rdma/hfi1/eprom.h
index 64a64276be81..d41f0b1afb15 100644
--- a/drivers/staging/rdma/hfi1/eprom.h
+++ b/drivers/staging/rdma/hfi1/eprom.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -52,4 +49,4 @@ struct hfi1_cmd;
 struct hfi1_devdata;
 
 int eprom_init(struct hfi1_devdata *dd);
-int handle_eprom_command(const struct hfi1_cmd *cmd);
+int handle_eprom_command(struct file *fp, const struct hfi1_cmd *cmd);
diff --git a/drivers/staging/rdma/hfi1/file_ops.c b/drivers/staging/rdma/hfi1/file_ops.c
index 8b911e8bf0df..8396dc5fb6c1 100644
--- a/drivers/staging/rdma/hfi1/file_ops.c
+++ b/drivers/staging/rdma/hfi1/file_ops.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -60,6 +57,8 @@
 #include "user_sdma.h"
 #include "user_exp_rcv.h"
 #include "eprom.h"
+#include "aspm.h"
+#include "mmu_rb.h"
 
 #undef pr_fmt
 #define pr_fmt(fmt) DRIVER_NAME ": " fmt
@@ -96,9 +95,6 @@ static int user_event_ack(struct hfi1_ctxtdata *, int, unsigned long);
 static int set_ctxt_pkey(struct hfi1_ctxtdata *, unsigned, u16);
 static int manage_rcvq(struct hfi1_ctxtdata *, unsigned, int);
 static int vma_fault(struct vm_area_struct *, struct vm_fault *);
-static int exp_tid_setup(struct file *, struct hfi1_tid_info *);
-static int exp_tid_free(struct file *, struct hfi1_tid_info *);
-static void unlock_exp_tids(struct hfi1_ctxtdata *);
 
 static const struct file_operations hfi1_file_ops = {
 	.owner = THIS_MODULE,
@@ -164,7 +160,6 @@ enum mmap_types {
 #define dbg(fmt, ...)				\
 	pr_info(fmt, ##__VA_ARGS__)
 
-
 static inline int is_valid_mmap(u64 token)
 {
 	return (HFI1_MMAP_TOKEN_GET(MAGIC, token) == HFI1_MMAP_MAGIC);
@@ -188,6 +183,7 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 	struct hfi1_cmd cmd;
 	struct hfi1_user_info uinfo;
 	struct hfi1_tid_info tinfo;
+	unsigned long addr;
 	ssize_t consumed = 0, copy = 0, ret = 0;
 	void *dest = NULL;
 	__u64 user_val = 0;
@@ -219,6 +215,7 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 		break;
 	case HFI1_CMD_TID_UPDATE:
 	case HFI1_CMD_TID_FREE:
+	case HFI1_CMD_TID_INVAL_READ:
 		copy = sizeof(tinfo);
 		dest = &tinfo;
 		break;
@@ -294,9 +291,8 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 			sc_return_credits(uctxt->sc);
 		break;
 	case HFI1_CMD_TID_UPDATE:
-		ret = exp_tid_setup(fp, &tinfo);
+		ret = hfi1_user_exp_rcv_setup(fp, &tinfo);
 		if (!ret) {
-			unsigned long addr;
 			/*
 			 * Copy the number of tidlist entries we used
 			 * and the length of the buffer we registered.
@@ -311,8 +307,25 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 				ret = -EFAULT;
 		}
 		break;
+	case HFI1_CMD_TID_INVAL_READ:
+		ret = hfi1_user_exp_rcv_invalid(fp, &tinfo);
+		if (ret)
+			break;
+		addr = (unsigned long)cmd.addr +
+			offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			ret = -EFAULT;
+		break;
 	case HFI1_CMD_TID_FREE:
-		ret = exp_tid_free(fp, &tinfo);
+		ret = hfi1_user_exp_rcv_clear(fp, &tinfo);
+		if (ret)
+			break;
+		addr = (unsigned long)cmd.addr +
+			offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			ret = -EFAULT;
 		break;
 	case HFI1_CMD_RECV_CTRL:
 		ret = manage_rcvq(uctxt, fd->subctxt, (int)user_val);
@@ -373,8 +386,10 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 				break;
 			}
 			if (dd->flags & HFI1_FORCED_FREEZE) {
-				/* Don't allow context reset if we are into
-				 * forced freeze */
+				/*
+				 * Don't allow context reset if we are into
+				 * forced freeze
+				 */
 				ret = -ENODEV;
 				break;
 			}
@@ -382,8 +397,9 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 			ret = sc_enable(sc);
 			hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB,
 				     uctxt->ctxt);
-		} else
+		} else {
 			ret = sc_restart(sc);
+		}
 		if (!ret)
 			sc_return_credits(sc);
 		break;
@@ -393,7 +409,7 @@ static ssize_t hfi1_file_write(struct file *fp, const char __user *data,
 	case HFI1_CMD_EP_ERASE_RANGE:
 	case HFI1_CMD_EP_READ_RANGE:
 	case HFI1_CMD_EP_WRITE_RANGE:
-		ret = handle_eprom_command(&cmd);
+		ret = handle_eprom_command(fp, &cmd);
 		break;
 	}
 
@@ -732,6 +748,9 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 	/* drain user sdma queue */
 	hfi1_user_sdma_free_queues(fdata);
 
+	/* release the cpu */
+	hfi1_put_proc_affinity(dd, fdata->rec_cpu_num);
+
 	/*
 	 * Clear any left over, unhandled events so the next process that
 	 * gets this context doesn't get confused.
@@ -755,6 +774,7 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 	hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_DIS |
 		     HFI1_RCVCTRL_TIDFLOW_DIS |
 		     HFI1_RCVCTRL_INTRAVAIL_DIS |
+		     HFI1_RCVCTRL_TAILUPD_DIS |
 		     HFI1_RCVCTRL_ONE_PKT_EGR_DIS |
 		     HFI1_RCVCTRL_NO_RHQ_DROP_DIS |
 		     HFI1_RCVCTRL_NO_EGR_DROP_DIS, uctxt->ctxt);
@@ -777,14 +797,12 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 	uctxt->pionowait = 0;
 	uctxt->event_flags = 0;
 
-	hfi1_clear_tids(uctxt);
+	hfi1_user_exp_rcv_free(fdata);
 	hfi1_clear_ctxt_pkey(dd, uctxt->ctxt);
 
-	if (uctxt->tid_pg_list)
-		unlock_exp_tids(uctxt);
-
 	hfi1_stats.sps_ctxts--;
-	dd->freectxts++;
+	if (++dd->freectxts == dd->num_user_contexts)
+		aspm_enable_all(dd);
 	mutex_unlock(&hfi1_mutex);
 	hfi1_free_ctxtdata(dd, uctxt);
 done:
@@ -826,8 +844,16 @@ static int assign_ctxt(struct file *fp, struct hfi1_user_info *uinfo)
 
 	mutex_lock(&hfi1_mutex);
 	/* First, lets check if we need to setup a shared context? */
-	if (uinfo->subctxt_cnt)
+	if (uinfo->subctxt_cnt) {
+		struct hfi1_filedata *fd = fp->private_data;
+
 		ret = find_shared_ctxt(fp, uinfo);
+		if (ret < 0)
+			goto done_unlock;
+		if (ret)
+			fd->rec_cpu_num = hfi1_get_proc_affinity(
+				fd->uctxt->dd, fd->uctxt->numa_id);
+	}
 
 	/*
 	 * We execute the following block if we couldn't find a
@@ -837,6 +863,7 @@ static int assign_ctxt(struct file *fp, struct hfi1_user_info *uinfo)
 		i_minor = iminor(file_inode(fp)) - HFI1_USER_MINOR_BASE;
 		ret = get_user_context(fp, uinfo, i_minor - 1, alg);
 	}
+done_unlock:
 	mutex_unlock(&hfi1_mutex);
 done:
 	return ret;
@@ -962,7 +989,7 @@ static int allocate_ctxt(struct file *fp, struct hfi1_devdata *dd,
 	struct hfi1_filedata *fd = fp->private_data;
 	struct hfi1_ctxtdata *uctxt;
 	unsigned ctxt;
-	int ret;
+	int ret, numa;
 
 	if (dd->flags & HFI1_FROZEN) {
 		/*
@@ -982,17 +1009,26 @@ static int allocate_ctxt(struct file *fp, struct hfi1_devdata *dd,
 	if (ctxt == dd->num_rcv_contexts)
 		return -EBUSY;
 
-	uctxt = hfi1_create_ctxtdata(dd->pport, ctxt);
+	fd->rec_cpu_num = hfi1_get_proc_affinity(dd, -1);
+	if (fd->rec_cpu_num != -1)
+		numa = cpu_to_node(fd->rec_cpu_num);
+	else
+		numa = numa_node_id();
+	uctxt = hfi1_create_ctxtdata(dd->pport, ctxt, numa);
 	if (!uctxt) {
 		dd_dev_err(dd,
 			   "Unable to allocate ctxtdata memory, failing open\n");
 		return -ENOMEM;
 	}
+	hfi1_cdbg(PROC, "[%u:%u] pid %u assigned to CPU %d (NUMA %u)",
+		  uctxt->ctxt, fd->subctxt, current->pid, fd->rec_cpu_num,
+		  uctxt->numa_id);
+
 	/*
 	 * Allocate and enable a PIO send context.
 	 */
 	uctxt->sc = sc_alloc(dd, SC_USER, uctxt->rcvhdrqentsize,
-			     uctxt->numa_id);
+			     uctxt->dd->node);
 	if (!uctxt->sc)
 		return -ENOMEM;
 
@@ -1026,7 +1062,12 @@ static int allocate_ctxt(struct file *fp, struct hfi1_devdata *dd,
 	INIT_LIST_HEAD(&uctxt->sdma_queues);
 	spin_lock_init(&uctxt->sdma_qlock);
 	hfi1_stats.sps_ctxts++;
-	dd->freectxts--;
+	/*
+	 * Disable ASPM when there are open user/PSM contexts to avoid
+	 * issues with ASPM L1 exit latency
+	 */
+	if (dd->freectxts-- == dd->num_user_contexts)
+		aspm_disable_all(dd);
 	fd->uctxt = uctxt;
 
 	return 0;
@@ -1035,22 +1076,19 @@ static int allocate_ctxt(struct file *fp, struct hfi1_devdata *dd,
 static int init_subctxts(struct hfi1_ctxtdata *uctxt,
 			 const struct hfi1_user_info *uinfo)
 {
-	int ret = 0;
 	unsigned num_subctxts;
 
 	num_subctxts = uinfo->subctxt_cnt;
-	if (num_subctxts > HFI1_MAX_SHARED_CTXTS) {
-		ret = -EINVAL;
-		goto bail;
-	}
+	if (num_subctxts > HFI1_MAX_SHARED_CTXTS)
+		return -EINVAL;
 
 	uctxt->subctxt_cnt = uinfo->subctxt_cnt;
 	uctxt->subctxt_id = uinfo->subctxt_id;
 	uctxt->active_slaves = 1;
 	uctxt->redirect_seq_cnt = 1;
 	set_bit(HFI1_CTXT_MASTER_UNINIT, &uctxt->event_flags);
-bail:
-	return ret;
+
+	return 0;
 }
 
 static int setup_subctxt(struct hfi1_ctxtdata *uctxt)
@@ -1105,10 +1143,10 @@ static int user_init(struct file *fp)
 	 * has done it.
 	 */
 	if (fd->subctxt) {
-		ret = wait_event_interruptible(uctxt->wait,
-			!test_bit(HFI1_CTXT_MASTER_UNINIT,
-			&uctxt->event_flags));
-		goto done;
+		ret = wait_event_interruptible(uctxt->wait, !test_bit(
+					       HFI1_CTXT_MASTER_UNINIT,
+					       &uctxt->event_flags));
+		goto expected;
 	}
 
 	/* initialize poll variables... */
@@ -1146,8 +1184,16 @@ static int user_init(struct file *fp)
 		rcvctrl_ops |= HFI1_RCVCTRL_NO_EGR_DROP_ENB;
 	if (HFI1_CAP_KGET_MASK(uctxt->flags, NODROP_RHQ_FULL))
 		rcvctrl_ops |= HFI1_RCVCTRL_NO_RHQ_DROP_ENB;
+	/*
+	 * The RcvCtxtCtrl.TailUpd bit has to be explicitly written.
+	 * We can't rely on the correct value to be set from prior
+	 * uses of the chip or ctxt. Therefore, add the rcvctrl op
+	 * for both cases.
+	 */
 	if (HFI1_CAP_KGET_MASK(uctxt->flags, DMA_RTAIL))
 		rcvctrl_ops |= HFI1_RCVCTRL_TAILUPD_ENB;
+	else
+		rcvctrl_ops |= HFI1_RCVCTRL_TAILUPD_DIS;
 	hfi1_rcvctrl(uctxt->dd, rcvctrl_ops, uctxt->ctxt);
 
 	/* Notify any waiting slaves */
@@ -1155,8 +1201,18 @@ static int user_init(struct file *fp)
 		clear_bit(HFI1_CTXT_MASTER_UNINIT, &uctxt->event_flags);
 		wake_up(&uctxt->wait);
 	}
-	ret = 0;
 
+expected:
+	/*
+	 * Expected receive has to be setup for all processes (including
+	 * shared contexts). However, it has to be done after the master
+	 * context has been fully configured as it depends on the
+	 * eager/expected split of the RcvArray entries.
+	 * Setting it up here ensures that the subcontexts will be waiting
+	 * (due to the above wait_event_interruptible() until the master
+	 * is setup.
+	 */
+	ret = hfi1_user_exp_rcv_init(fp);
 done:
 	return ret;
 }
@@ -1226,46 +1282,6 @@ static int setup_ctxt(struct file *fp)
 			if (ret)
 				goto done;
 		}
-		/* Setup Expected Rcv memories */
-		uctxt->tid_pg_list = vzalloc(uctxt->expected_count *
-					     sizeof(struct page **));
-		if (!uctxt->tid_pg_list) {
-			ret = -ENOMEM;
-			goto done;
-		}
-		uctxt->physshadow = vzalloc(uctxt->expected_count *
-					    sizeof(*uctxt->physshadow));
-		if (!uctxt->physshadow) {
-			ret = -ENOMEM;
-			goto done;
-		}
-		/* allocate expected TID map and initialize the cursor */
-		atomic_set(&uctxt->tidcursor, 0);
-		uctxt->numtidgroups = uctxt->expected_count /
-			dd->rcv_entries.group_size;
-		uctxt->tidmapcnt = uctxt->numtidgroups / BITS_PER_LONG +
-			!!(uctxt->numtidgroups % BITS_PER_LONG);
-		uctxt->tidusemap = kzalloc_node(uctxt->tidmapcnt *
-						sizeof(*uctxt->tidusemap),
-						GFP_KERNEL, uctxt->numa_id);
-		if (!uctxt->tidusemap) {
-			ret = -ENOMEM;
-			goto done;
-		}
-		/*
-		 * In case that the number of groups is not a multiple of
-		 * 64 (the number of groups in a tidusemap element), mark
-		 * the extra ones as used. This will effectively make them
-		 * permanently used and should never be assigned. Otherwise,
-		 * the code which checks how many free groups we have will
-		 * get completely confused about the state of the bits.
-		 */
-		if (uctxt->numtidgroups % BITS_PER_LONG)
-			uctxt->tidusemap[uctxt->tidmapcnt - 1] =
-				~((1ULL << (uctxt->numtidgroups %
-					    BITS_PER_LONG)) - 1);
-		trace_hfi1_exp_tid_map(uctxt->ctxt, fd->subctxt, 0,
-				       uctxt->tidusemap, uctxt->tidmapcnt);
 	}
 	ret = hfi1_user_sdma_alloc_queues(uctxt, fp);
 	if (ret)
@@ -1391,8 +1407,9 @@ static unsigned int poll_next(struct file *fp,
 		set_bit(HFI1_CTXT_WAITING_RCV, &uctxt->event_flags);
 		hfi1_rcvctrl(dd, HFI1_RCVCTRL_INTRAVAIL_ENB, uctxt->ctxt);
 		pollflag = 0;
-	} else
+	} else {
 		pollflag = POLLIN | POLLRDNORM;
+	}
 	spin_unlock_irq(&dd->uctxt_lock);
 
 	return pollflag;
@@ -1470,8 +1487,9 @@ static int manage_rcvq(struct hfi1_ctxtdata *uctxt, unsigned subctxt,
 		if (uctxt->rcvhdrtail_kvaddr)
 			clear_rcvhdrtail(uctxt);
 		rcvctrl_op = HFI1_RCVCTRL_CTXT_ENB;
-	} else
+	} else {
 		rcvctrl_op = HFI1_RCVCTRL_CTXT_DIS;
+	}
 	hfi1_rcvctrl(dd, rcvctrl_op, uctxt->ctxt);
 	/* always; new head should be equal to new tail; see above */
 bail:
@@ -1504,367 +1522,6 @@ static int user_event_ack(struct hfi1_ctxtdata *uctxt, int subctxt,
 	return 0;
 }
 
-#define num_user_pages(vaddr, len)					\
-	(1 + (((((unsigned long)(vaddr) +				\
-		 (unsigned long)(len) - 1) & PAGE_MASK) -		\
-	       ((unsigned long)vaddr & PAGE_MASK)) >> PAGE_SHIFT))
-
-/**
- * tzcnt - count the number of trailing zeros in a 64bit value
- * @value: the value to be examined
- *
- * Returns the number of trailing least significant zeros in the
- * the input value. If the value is zero, return the number of
- * bits of the value.
- */
-static inline u8 tzcnt(u64 value)
-{
-	return value ? __builtin_ctzl(value) : sizeof(value) * 8;
-}
-
-static inline unsigned num_free_groups(unsigned long map, u16 *start)
-{
-	unsigned free;
-	u16 bitidx = *start;
-
-	if (bitidx >= BITS_PER_LONG)
-		return 0;
-	/* "Turn off" any bits set before our bit index */
-	map &= ~((1ULL << bitidx) - 1);
-	free = tzcnt(map) - bitidx;
-	while (!free && bitidx < BITS_PER_LONG) {
-		/* Zero out the last set bit so we look at the rest */
-		map &= ~(1ULL << bitidx);
-		/*
-		 * Account for the previously checked bits and advance
-		 * the bit index. We don't have to check for bitidx
-		 * getting bigger than BITS_PER_LONG here as it would
-		 * mean extra instructions that we don't need. If it
-		 * did happen, it would push free to a negative value
-		 * which will break the loop.
-		 */
-		free = tzcnt(map) - ++bitidx;
-	}
-	*start = bitidx;
-	return free;
-}
-
-static int exp_tid_setup(struct file *fp, struct hfi1_tid_info *tinfo)
-{
-	int ret = 0;
-	struct hfi1_filedata *fd = fp->private_data;
-	struct hfi1_ctxtdata *uctxt = fd->uctxt;
-	struct hfi1_devdata *dd = uctxt->dd;
-	unsigned tid, mapped = 0, npages, ngroups, exp_groups,
-		tidpairs = uctxt->expected_count / 2;
-	struct page **pages;
-	unsigned long vaddr, tidmap[uctxt->tidmapcnt];
-	dma_addr_t *phys;
-	u32 tidlist[tidpairs], pairidx = 0, tidcursor;
-	u16 useidx, idx, bitidx, tidcnt = 0;
-
-	vaddr = tinfo->vaddr;
-
-	if (offset_in_page(vaddr)) {
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	npages = num_user_pages(vaddr, tinfo->length);
-	if (!npages) {
-		ret = -EINVAL;
-		goto bail;
-	}
-	if (!access_ok(VERIFY_WRITE, (void __user *)vaddr,
-		       npages * PAGE_SIZE)) {
-		dd_dev_err(dd, "Fail vaddr %p, %u pages, !access_ok\n",
-			   (void *)vaddr, npages);
-		ret = -EFAULT;
-		goto bail;
-	}
-
-	memset(tidmap, 0, sizeof(tidmap[0]) * uctxt->tidmapcnt);
-	memset(tidlist, 0, sizeof(tidlist[0]) * tidpairs);
-
-	exp_groups = uctxt->expected_count / dd->rcv_entries.group_size;
-	/* which group set do we look at first? */
-	tidcursor = atomic_read(&uctxt->tidcursor);
-	useidx = (tidcursor >> 16) & 0xffff;
-	bitidx = tidcursor & 0xffff;
-
-	/*
-	 * Keep going until we've mapped all pages or we've exhausted all
-	 * RcvArray entries.
-	 * This iterates over the number of tidmaps + 1
-	 * (idx <= uctxt->tidmapcnt) so we check the bitmap which we
-	 * started from one more time for any free bits before the
-	 * starting point bit.
-	 */
-	for (mapped = 0, idx = 0;
-	     mapped < npages && idx <= uctxt->tidmapcnt;) {
-		u64 i, offset = 0;
-		unsigned free, pinned, pmapped = 0, bits_used;
-		u16 grp;
-
-		/*
-		 * "Reserve" the needed group bits under lock so other
-		 * processes can't step in the middle of it. Once
-		 * reserved, we don't need the lock anymore since we
-		 * are guaranteed the groups.
-		 */
-		spin_lock(&uctxt->exp_lock);
-		if (uctxt->tidusemap[useidx] == -1ULL ||
-		    bitidx >= BITS_PER_LONG) {
-			/* no free groups in the set, use the next */
-			useidx = (useidx + 1) % uctxt->tidmapcnt;
-			idx++;
-			bitidx = 0;
-			spin_unlock(&uctxt->exp_lock);
-			continue;
-		}
-		ngroups = ((npages - mapped) / dd->rcv_entries.group_size) +
-			!!((npages - mapped) % dd->rcv_entries.group_size);
-
-		/*
-		 * If we've gotten here, the current set of groups does have
-		 * one or more free groups.
-		 */
-		free = num_free_groups(uctxt->tidusemap[useidx], &bitidx);
-		if (!free) {
-			/*
-			 * Despite the check above, free could still come back
-			 * as 0 because we don't check the entire bitmap but
-			 * we start from bitidx.
-			 */
-			spin_unlock(&uctxt->exp_lock);
-			continue;
-		}
-		bits_used = min(free, ngroups);
-		tidmap[useidx] |= ((1ULL << bits_used) - 1) << bitidx;
-		uctxt->tidusemap[useidx] |= tidmap[useidx];
-		spin_unlock(&uctxt->exp_lock);
-
-		/*
-		 * At this point, we know where in the map we have free bits.
-		 * properly offset into the various "shadow" arrays and compute
-		 * the RcvArray entry index.
-		 */
-		offset = ((useidx * BITS_PER_LONG) + bitidx) *
-			dd->rcv_entries.group_size;
-		pages = uctxt->tid_pg_list + offset;
-		phys = uctxt->physshadow + offset;
-		tid = uctxt->expected_base + offset;
-
-		/* Calculate how many pages we can pin based on free bits */
-		pinned = min((bits_used * dd->rcv_entries.group_size),
-			     (npages - mapped));
-		/*
-		 * Now that we know how many free RcvArray entries we have,
-		 * we can pin that many user pages.
-		 */
-		ret = hfi1_acquire_user_pages(vaddr + (mapped * PAGE_SIZE),
-					      pinned, true, pages);
-		if (ret) {
-			/*
-			 * We can't continue because the pages array won't be
-			 * initialized. This should never happen,
-			 * unless perhaps the user has mpin'ed the pages
-			 * themselves.
-			 */
-			dd_dev_info(dd,
-				    "Failed to lock addr %p, %u pages: errno %d\n",
-				    (void *) vaddr, pinned, -ret);
-			/*
-			 * Let go of the bits that we reserved since we are not
-			 * going to use them.
-			 */
-			spin_lock(&uctxt->exp_lock);
-			uctxt->tidusemap[useidx] &=
-				~(((1ULL << bits_used) - 1) << bitidx);
-			spin_unlock(&uctxt->exp_lock);
-			goto done;
-		}
-		/*
-		 * How many groups do we need based on how many pages we have
-		 * pinned?
-		 */
-		ngroups = (pinned / dd->rcv_entries.group_size) +
-			!!(pinned % dd->rcv_entries.group_size);
-		/*
-		 * Keep programming RcvArray entries for all the <ngroups> free
-		 * groups.
-		 */
-		for (i = 0, grp = 0; grp < ngroups; i++, grp++) {
-			unsigned j;
-			u32 pair_size = 0, tidsize;
-			/*
-			 * This inner loop will program an entire group or the
-			 * array of pinned pages (which ever limit is hit
-			 * first).
-			 */
-			for (j = 0; j < dd->rcv_entries.group_size &&
-				     pmapped < pinned; j++, pmapped++, tid++) {
-				tidsize = PAGE_SIZE;
-				phys[pmapped] = hfi1_map_page(dd->pcidev,
-						   pages[pmapped], 0,
-						   tidsize, PCI_DMA_FROMDEVICE);
-				trace_hfi1_exp_rcv_set(uctxt->ctxt,
-						       fd->subctxt,
-						       tid, vaddr,
-						       phys[pmapped],
-						       pages[pmapped]);
-				/*
-				 * Each RcvArray entry is programmed with one
-				 * page * worth of memory. This will handle
-				 * the 8K MTU as well as anything smaller
-				 * due to the fact that both entries in the
-				 * RcvTidPair are programmed with a page.
-				 * PSM currently does not handle anything
-				 * bigger than 8K MTU, so should we even worry
-				 * about 10K here?
-				 */
-				hfi1_put_tid(dd, tid, PT_EXPECTED,
-					     phys[pmapped],
-					     ilog2(tidsize >> PAGE_SHIFT) + 1);
-				pair_size += tidsize >> PAGE_SHIFT;
-				EXP_TID_RESET(tidlist[pairidx], LEN, pair_size);
-				if (!(tid % 2)) {
-					tidlist[pairidx] |=
-					   EXP_TID_SET(IDX,
-						(tid - uctxt->expected_base)
-						       / 2);
-					tidlist[pairidx] |=
-						EXP_TID_SET(CTRL, 1);
-					tidcnt++;
-				} else {
-					tidlist[pairidx] |=
-						EXP_TID_SET(CTRL, 2);
-					pair_size = 0;
-					pairidx++;
-				}
-			}
-			/*
-			 * We've programmed the entire group (or as much of the
-			 * group as we'll use. Now, it's time to push it out...
-			 */
-			flush_wc();
-		}
-		mapped += pinned;
-		atomic_set(&uctxt->tidcursor,
-			   (((useidx & 0xffffff) << 16) |
-			    ((bitidx + bits_used) & 0xffffff)));
-	}
-	trace_hfi1_exp_tid_map(uctxt->ctxt, fd->subctxt, 0, uctxt->tidusemap,
-			       uctxt->tidmapcnt);
-
-done:
-	/* If we've mapped anything, copy relevant info to user */
-	if (mapped) {
-		if (copy_to_user((void __user *)(unsigned long)tinfo->tidlist,
-				 tidlist, sizeof(tidlist[0]) * tidcnt)) {
-			ret = -EFAULT;
-			goto done;
-		}
-		/* copy TID info to user */
-		if (copy_to_user((void __user *)(unsigned long)tinfo->tidmap,
-				 tidmap, sizeof(tidmap[0]) * uctxt->tidmapcnt))
-			ret = -EFAULT;
-	}
-bail:
-	/*
-	 * Calculate mapped length. New Exp TID protocol does not "unwind" and
-	 * report an error if it can't map the entire buffer. It just reports
-	 * the length that was mapped.
-	 */
-	tinfo->length = mapped * PAGE_SIZE;
-	tinfo->tidcnt = tidcnt;
-	return ret;
-}
-
-static int exp_tid_free(struct file *fp, struct hfi1_tid_info *tinfo)
-{
-	struct hfi1_filedata *fd = fp->private_data;
-	struct hfi1_ctxtdata *uctxt = fd->uctxt;
-	struct hfi1_devdata *dd = uctxt->dd;
-	unsigned long tidmap[uctxt->tidmapcnt];
-	struct page **pages;
-	dma_addr_t *phys;
-	u16 idx, bitidx, tid;
-	int ret = 0;
-
-	if (copy_from_user(&tidmap, (void __user *)(unsigned long)
-			   tinfo->tidmap,
-			   sizeof(tidmap[0]) * uctxt->tidmapcnt)) {
-		ret = -EFAULT;
-		goto done;
-	}
-	for (idx = 0; idx < uctxt->tidmapcnt; idx++) {
-		unsigned long map;
-
-		bitidx = 0;
-		if (!tidmap[idx])
-			continue;
-		map = tidmap[idx];
-		while ((bitidx = tzcnt(map)) < BITS_PER_LONG) {
-			int i, pcount = 0;
-			struct page *pshadow[dd->rcv_entries.group_size];
-			unsigned offset = ((idx * BITS_PER_LONG) + bitidx) *
-				dd->rcv_entries.group_size;
-
-			pages = uctxt->tid_pg_list + offset;
-			phys = uctxt->physshadow + offset;
-			tid = uctxt->expected_base + offset;
-			for (i = 0; i < dd->rcv_entries.group_size;
-			     i++, tid++) {
-				if (pages[i]) {
-					hfi1_put_tid(dd, tid, PT_INVALID,
-						      0, 0);
-					trace_hfi1_exp_rcv_free(uctxt->ctxt,
-								fd->subctxt,
-								tid, phys[i],
-								pages[i]);
-					pci_unmap_page(dd->pcidev, phys[i],
-					      PAGE_SIZE, PCI_DMA_FROMDEVICE);
-					pshadow[pcount] = pages[i];
-					pages[i] = NULL;
-					pcount++;
-					phys[i] = 0;
-				}
-			}
-			flush_wc();
-			hfi1_release_user_pages(pshadow, pcount, true);
-			clear_bit(bitidx, &uctxt->tidusemap[idx]);
-			map &= ~(1ULL<<bitidx);
-		}
-	}
-	trace_hfi1_exp_tid_map(uctxt->ctxt, fd->subctxt, 1, uctxt->tidusemap,
-			       uctxt->tidmapcnt);
-done:
-	return ret;
-}
-
-static void unlock_exp_tids(struct hfi1_ctxtdata *uctxt)
-{
-	struct hfi1_devdata *dd = uctxt->dd;
-	unsigned tid;
-
-	dd_dev_info(dd, "ctxt %u unlocking any locked expTID pages\n",
-		    uctxt->ctxt);
-	for (tid = 0; tid < uctxt->expected_count; tid++) {
-		struct page *p = uctxt->tid_pg_list[tid];
-		dma_addr_t phys;
-
-		if (!p)
-			continue;
-
-		phys = uctxt->physshadow[tid];
-		uctxt->physshadow[tid] = 0;
-		uctxt->tid_pg_list[tid] = NULL;
-		pci_unmap_page(dd->pcidev, phys, PAGE_SIZE, PCI_DMA_FROMDEVICE);
-		hfi1_release_user_pages(&p, 1, true);
-	}
-}
-
 static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, unsigned subctxt,
 			 u16 pkey)
 {
@@ -1933,10 +1590,9 @@ static loff_t ui_lseek(struct file *filp, loff_t offset, int whence)
 	return filp->f_pos;
 }
 
-
 /* NOTE: assumes unsigned long is 8 bytes */
 static ssize_t ui_read(struct file *filp, char __user *buf, size_t count,
-			loff_t *f_pos)
+		       loff_t *f_pos)
 {
 	struct hfi1_devdata *dd = filp->private_data;
 	void __iomem *base = dd->kregbase;
@@ -1972,12 +1628,12 @@ static ssize_t ui_read(struct file *filp, char __user *buf, size_t count,
 		 * them.  These registers are defined as having a read value
 		 * of 0.
 		 */
-		else if (csr_off == ASIC_GPIO_CLEAR
-				|| csr_off == ASIC_GPIO_FORCE
-				|| csr_off == ASIC_QSFP1_CLEAR
-				|| csr_off == ASIC_QSFP1_FORCE
-				|| csr_off == ASIC_QSFP2_CLEAR
-				|| csr_off == ASIC_QSFP2_FORCE)
+		else if (csr_off == ASIC_GPIO_CLEAR ||
+			 csr_off == ASIC_GPIO_FORCE ||
+			 csr_off == ASIC_QSFP1_CLEAR ||
+			 csr_off == ASIC_QSFP1_FORCE ||
+			 csr_off == ASIC_QSFP2_CLEAR ||
+			 csr_off == ASIC_QSFP2_FORCE)
 			data = 0;
 		else if (csr_off >= barlen) {
 			/*
diff --git a/drivers/staging/rdma/hfi1/firmware.c b/drivers/staging/rdma/hfi1/firmware.c
index 28ae42faa018..3040162cb326 100644
--- a/drivers/staging/rdma/hfi1/firmware.c
+++ b/drivers/staging/rdma/hfi1/firmware.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -77,7 +74,13 @@ static uint fw_8051_load = 1;
 static uint fw_fabric_serdes_load = 1;
 static uint fw_pcie_serdes_load = 1;
 static uint fw_sbus_load = 1;
-static uint platform_config_load = 1;
+
+/*
+ * Access required in platform.c
+ * Maintains state of whether the platform config was fetched via the
+ * fallback option
+ */
+uint platform_config_load;
 
 /* Firmware file names get set in hfi1_firmware_init() based on the above */
 static char *fw_8051_name;
@@ -107,6 +110,7 @@ struct css_header {
 	u32 exponent_size;	/* in DWORDs */
 	u32 reserved[22];
 };
+
 /* expected field values */
 #define CSS_MODULE_TYPE	   0x00000006
 #define CSS_HEADER_LEN	   0x000000a1
@@ -166,6 +170,7 @@ enum fw_state {
 	FW_FINAL,
 	FW_ERR
 };
+
 static enum fw_state fw_state = FW_EMPTY;
 static int fw_err;
 static struct firmware_details fw_8051;
@@ -193,7 +198,7 @@ static const struct firmware *platform_config;
 #define RSA_ENGINE_TIMEOUT 100 /* ms */
 
 /* hardware mutex timeout, in ms */
-#define HM_TIMEOUT 4000 /* 4 s */
+#define HM_TIMEOUT 10 /* ms */
 
 /* 8051 memory access timeout, in us */
 #define DC8051_ACCESS_TIMEOUT 100 /* us */
@@ -233,6 +238,8 @@ static const u8 all_pcie_serdes_broadcast = 0xe0;
 
 /* forwards */
 static void dispose_one_firmware(struct firmware_details *fdet);
+static int load_fabric_serdes_firmware(struct hfi1_devdata *dd,
+				       struct firmware_details *fdet);
 
 /*
  * Read a single 64-bit value from 8051 data memory.
@@ -372,8 +379,8 @@ static int invalid_header(struct hfi1_devdata *dd, const char *what,
 		return 0;
 
 	dd_dev_err(dd,
-		"invalid firmware header field %s: expected 0x%x, actual 0x%x\n",
-		what, expected, actual);
+		   "invalid firmware header field %s: expected 0x%x, actual 0x%x\n",
+		   what, expected, actual);
 	return 1;
 }
 
@@ -383,19 +390,19 @@ static int invalid_header(struct hfi1_devdata *dd, const char *what,
 static int verify_css_header(struct hfi1_devdata *dd, struct css_header *css)
 {
 	/* verify CSS header fields (most sizes are in DW, so add /4) */
-	if (invalid_header(dd, "module_type", css->module_type, CSS_MODULE_TYPE)
-			|| invalid_header(dd, "header_len", css->header_len,
-					(sizeof(struct firmware_file)/4))
-			|| invalid_header(dd, "header_version",
-					css->header_version, CSS_HEADER_VERSION)
-			|| invalid_header(dd, "module_vendor",
-					css->module_vendor, CSS_MODULE_VENDOR)
-			|| invalid_header(dd, "key_size",
-					css->key_size, KEY_SIZE/4)
-			|| invalid_header(dd, "modulus_size",
-					css->modulus_size, KEY_SIZE/4)
-			|| invalid_header(dd, "exponent_size",
-					css->exponent_size, EXPONENT_SIZE/4)) {
+	if (invalid_header(dd, "module_type", css->module_type,
+			   CSS_MODULE_TYPE) ||
+	    invalid_header(dd, "header_len", css->header_len,
+			   (sizeof(struct firmware_file) / 4)) ||
+	    invalid_header(dd, "header_version", css->header_version,
+			   CSS_HEADER_VERSION) ||
+	    invalid_header(dd, "module_vendor", css->module_vendor,
+			   CSS_MODULE_VENDOR) ||
+	    invalid_header(dd, "key_size", css->key_size, KEY_SIZE / 4) ||
+	    invalid_header(dd, "modulus_size", css->modulus_size,
+			   KEY_SIZE / 4) ||
+	    invalid_header(dd, "exponent_size", css->exponent_size,
+			   EXPONENT_SIZE / 4)) {
 		return -EINVAL;
 	}
 	return 0;
@@ -410,8 +417,8 @@ static int payload_check(struct hfi1_devdata *dd, const char *name,
 	/* make sure we have some payload */
 	if (prefix_size >= file_size) {
 		dd_dev_err(dd,
-			"firmware \"%s\", size %ld, must be larger than %ld bytes\n",
-			name, file_size, prefix_size);
+			   "firmware \"%s\", size %ld, must be larger than %ld bytes\n",
+			   name, file_size, prefix_size);
 		return -EINVAL;
 	}
 
@@ -433,8 +440,8 @@ static int obtain_one_firmware(struct hfi1_devdata *dd, const char *name,
 
 	ret = request_firmware(&fdet->fw, name, &dd->pcidev->dev);
 	if (ret) {
-		dd_dev_err(dd, "cannot find firmware \"%s\", err %d\n",
-			   name, ret);
+		dd_dev_warn(dd, "cannot find firmware \"%s\", err %d\n",
+			    name, ret);
 		return ret;
 	}
 
@@ -480,14 +487,14 @@ static int obtain_one_firmware(struct hfi1_devdata *dd, const char *name,
 	ret = verify_css_header(dd, css);
 	if (ret) {
 		dd_dev_info(dd, "Invalid CSS header for \"%s\"\n", name);
-	} else if ((css->size*4) == fdet->fw->size) {
+	} else if ((css->size * 4) == fdet->fw->size) {
 		/* non-augmented firmware file */
 		struct firmware_file *ff = (struct firmware_file *)
 							fdet->fw->data;
 
 		/* make sure there are bytes in the payload */
 		ret = payload_check(dd, name, fdet->fw->size,
-						sizeof(struct firmware_file));
+				    sizeof(struct firmware_file));
 		if (ret == 0) {
 			fdet->css_header = css;
 			fdet->modulus = ff->modulus;
@@ -505,14 +512,14 @@ static int obtain_one_firmware(struct hfi1_devdata *dd, const char *name,
 			dd_dev_err(dd, "driver is unable to validate firmware without r2 and mu (not in firmware file)\n");
 			ret = -EINVAL;
 		}
-	} else if ((css->size*4) + AUGMENT_SIZE == fdet->fw->size) {
+	} else if ((css->size * 4) + AUGMENT_SIZE == fdet->fw->size) {
 		/* augmented firmware file */
 		struct augmented_firmware_file *aff =
 			(struct augmented_firmware_file *)fdet->fw->data;
 
 		/* make sure there are bytes in the payload */
 		ret = payload_check(dd, name, fdet->fw->size,
-					sizeof(struct augmented_firmware_file));
+				    sizeof(struct augmented_firmware_file));
 		if (ret == 0) {
 			fdet->css_header = css;
 			fdet->modulus = aff->modulus;
@@ -527,9 +534,10 @@ static int obtain_one_firmware(struct hfi1_devdata *dd, const char *name,
 	} else {
 		/* css->size check failed */
 		dd_dev_err(dd,
-			"invalid firmware header field size: expected 0x%lx or 0x%lx, actual 0x%x\n",
-			fdet->fw->size/4, (fdet->fw->size - AUGMENT_SIZE)/4,
-			css->size);
+			   "invalid firmware header field size: expected 0x%lx or 0x%lx, actual 0x%x\n",
+			   fdet->fw->size / 4,
+			   (fdet->fw->size - AUGMENT_SIZE) / 4,
+			   css->size);
 
 		ret = -EINVAL;
 	}
@@ -572,7 +580,7 @@ retry:
 		 * We tried the original and it failed.  Move to the
 		 * alternate.
 		 */
-		dd_dev_info(dd, "using alternate firmware names\n");
+		dd_dev_warn(dd, "using alternate firmware names\n");
 		/*
 		 * Let others run.  Some systems, when missing firmware, does
 		 * something that holds for 30 seconds.  If we do that twice
@@ -593,27 +601,27 @@ retry:
 		fw_pcie_serdes_name = ALT_FW_PCIE_NAME;
 	}
 
-	if (fw_8051_load) {
-		err = obtain_one_firmware(dd, fw_8051_name, &fw_8051);
+	if (fw_sbus_load) {
+		err = obtain_one_firmware(dd, fw_sbus_name, &fw_sbus);
 		if (err)
 			goto done;
 	}
 
-	if (fw_fabric_serdes_load) {
-		err = obtain_one_firmware(dd, fw_fabric_serdes_name,
-			&fw_fabric);
+	if (fw_pcie_serdes_load) {
+		err = obtain_one_firmware(dd, fw_pcie_serdes_name, &fw_pcie);
 		if (err)
 			goto done;
 	}
 
-	if (fw_sbus_load) {
-		err = obtain_one_firmware(dd, fw_sbus_name, &fw_sbus);
+	if (fw_fabric_serdes_load) {
+		err = obtain_one_firmware(dd, fw_fabric_serdes_name,
+					  &fw_fabric);
 		if (err)
 			goto done;
 	}
 
-	if (fw_pcie_serdes_load) {
-		err = obtain_one_firmware(dd, fw_pcie_serdes_name, &fw_pcie);
+	if (fw_8051_load) {
+		err = obtain_one_firmware(dd, fw_8051_name, &fw_8051);
 		if (err)
 			goto done;
 	}
@@ -621,16 +629,18 @@ retry:
 done:
 	if (err) {
 		/* oops, had problems obtaining a firmware */
-		if (fw_state == FW_EMPTY) {
-			/* retry with alternate */
+		if (fw_state == FW_EMPTY && dd->icode == ICODE_RTL_SILICON) {
+			/* retry with alternate (RTL only) */
 			fw_state = FW_TRY;
 			goto retry;
 		}
+		dd_dev_err(dd, "unable to obtain working firmware\n");
 		fw_state = FW_ERR;
 		fw_err = -ENOENT;
 	} else {
 		/* success */
-		if (fw_state == FW_EMPTY)
+		if (fw_state == FW_EMPTY &&
+		    dd->icode != ICODE_FUNCTIONAL_SIMULATOR)
 			fw_state = FW_TRY;	/* may retry later */
 		else
 			fw_state = FW_FINAL;	/* cannot try again */
@@ -673,10 +683,15 @@ static int obtain_firmware(struct hfi1_devdata *dd)
 	}
 	/* not in FW_TRY state */
 
-	if (fw_state == FW_FINAL)
+	if (fw_state == FW_FINAL) {
+		if (platform_config) {
+			dd->platform_config.data = platform_config->data;
+			dd->platform_config.size = platform_config->size;
+		}
 		goto done;	/* already acquired */
-	else if (fw_state == FW_ERR)
+	} else if (fw_state == FW_ERR) {
 		goto done;	/* already tried and failed */
+	}
 	/* fw_state is FW_EMPTY */
 
 	/* set fw_state to FW_TRY, FW_FINAL, or FW_ERR, and fw_err */
@@ -685,9 +700,13 @@ static int obtain_firmware(struct hfi1_devdata *dd)
 	if (platform_config_load) {
 		platform_config = NULL;
 		err = request_firmware(&platform_config, platform_config_name,
-						&dd->pcidev->dev);
-		if (err)
+				       &dd->pcidev->dev);
+		if (err) {
 			platform_config = NULL;
+			goto done;
+		}
+		dd->platform_config.data = platform_config->data;
+		dd->platform_config.size = platform_config->size;
 	}
 
 done:
@@ -761,7 +780,7 @@ static int retry_firmware(struct hfi1_devdata *dd, int load_result)
 static void write_rsa_data(struct hfi1_devdata *dd, int what,
 			   const u8 *data, int nbytes)
 {
-	int qw_size = nbytes/8;
+	int qw_size = nbytes / 8;
 	int i;
 
 	if (((unsigned long)data & 0x7) == 0) {
@@ -769,14 +788,14 @@ static void write_rsa_data(struct hfi1_devdata *dd, int what,
 		u64 *ptr = (u64 *)data;
 
 		for (i = 0; i < qw_size; i++, ptr++)
-			write_csr(dd, what + (8*i), *ptr);
+			write_csr(dd, what + (8 * i), *ptr);
 	} else {
 		/* not aligned */
 		for (i = 0; i < qw_size; i++, data += 8) {
 			u64 value;
 
 			memcpy(&value, data, 8);
-			write_csr(dd, what + (8*i), value);
+			write_csr(dd, what + (8 * i), value);
 		}
 	}
 }
@@ -789,7 +808,7 @@ static void write_streamed_rsa_data(struct hfi1_devdata *dd, int what,
 				    const u8 *data, int nbytes)
 {
 	u64 *ptr = (u64 *)data;
-	int qw_size = nbytes/8;
+	int qw_size = nbytes / 8;
 
 	for (; qw_size > 0; qw_size--, ptr++)
 		write_csr(dd, what, *ptr);
@@ -822,7 +841,7 @@ static int run_rsa(struct hfi1_devdata *dd, const char *who,
 			     >> MISC_CFG_FW_CTRL_RSA_STATUS_SHIFT;
 	if (status != RSA_STATUS_IDLE) {
 		dd_dev_err(dd, "%s security engine not idle - giving up\n",
-			who);
+			   who);
 		return -EBUSY;
 	}
 
@@ -859,7 +878,7 @@ static int run_rsa(struct hfi1_devdata *dd, const char *who,
 		if (status == RSA_STATUS_IDLE) {
 			/* should not happen */
 			dd_dev_err(dd, "%s firmware security bad idle state\n",
-				who);
+				   who);
 			ret = -EINVAL;
 			break;
 		} else if (status == RSA_STATUS_DONE) {
@@ -893,19 +912,20 @@ static int run_rsa(struct hfi1_devdata *dd, const char *who,
 	 * is not keeping the error high.
 	 */
 	write_csr(dd, MISC_ERR_CLEAR,
-			MISC_ERR_STATUS_MISC_FW_AUTH_FAILED_ERR_SMASK
-			| MISC_ERR_STATUS_MISC_KEY_MISMATCH_ERR_SMASK);
+		  MISC_ERR_STATUS_MISC_FW_AUTH_FAILED_ERR_SMASK |
+		  MISC_ERR_STATUS_MISC_KEY_MISMATCH_ERR_SMASK);
 	/*
-	 * All that is left are the current errors.  Print failure details,
-	 * if any.
+	 * All that is left are the current errors.  Print warnings on
+	 * authorization failure details, if any.  Firmware authorization
+	 * can be retried, so these are only warnings.
 	 */
 	reg = read_csr(dd, MISC_ERR_STATUS);
 	if (ret) {
 		if (reg & MISC_ERR_STATUS_MISC_FW_AUTH_FAILED_ERR_SMASK)
-			dd_dev_err(dd, "%s firmware authorization failed\n",
-				who);
+			dd_dev_warn(dd, "%s firmware authorization failed\n",
+				    who);
 		if (reg & MISC_ERR_STATUS_MISC_KEY_MISMATCH_ERR_SMASK)
-			dd_dev_err(dd, "%s firmware key mismatch\n", who);
+			dd_dev_warn(dd, "%s firmware key mismatch\n", who);
 	}
 
 	return ret;
@@ -922,7 +942,8 @@ static void load_security_variables(struct hfi1_devdata *dd,
 	write_rsa_data(dd, MISC_CFG_RSA_MU, fdet->mu, MU_SIZE);
 	/* Security variables d.  Write the header */
 	write_streamed_rsa_data(dd, MISC_CFG_SHA_PRELOAD,
-			(u8 *)fdet->css_header, sizeof(struct css_header));
+				(u8 *)fdet->css_header,
+				sizeof(struct css_header));
 }
 
 /* return the 8051 firmware state */
@@ -1002,7 +1023,7 @@ static int load_8051_firmware(struct hfi1_devdata *dd,
 
 	/* Firmware load steps 3-5 */
 	ret = write_8051(dd, 1/*code*/, 0, fdet->firmware_ptr,
-							fdet->firmware_len);
+			 fdet->firmware_len);
 	if (ret)
 		return ret;
 
@@ -1029,13 +1050,13 @@ static int load_8051_firmware(struct hfi1_devdata *dd,
 	ret = wait_fm_ready(dd, TIMEOUT_8051_START);
 	if (ret) { /* timed out */
 		dd_dev_err(dd, "8051 start timeout, current state 0x%x\n",
-			get_firmware_state(dd));
+			   get_firmware_state(dd));
 		return -ETIMEDOUT;
 	}
 
 	read_misc_status(dd, &ver_a, &ver_b);
 	dd_dev_info(dd, "8051 firmware version %d.%d\n",
-		(int)ver_b, (int)ver_a);
+		    (int)ver_b, (int)ver_a);
 	dd->dc8051_ver = dc8051_ver(ver_b, ver_a);
 
 	return 0;
@@ -1050,11 +1071,11 @@ void sbus_request(struct hfi1_devdata *dd,
 		  u8 receiver_addr, u8 data_addr, u8 command, u32 data_in)
 {
 	write_csr(dd, ASIC_CFG_SBUS_REQUEST,
-		((u64)data_in << ASIC_CFG_SBUS_REQUEST_DATA_IN_SHIFT)
-		| ((u64)command << ASIC_CFG_SBUS_REQUEST_COMMAND_SHIFT)
-		| ((u64)data_addr << ASIC_CFG_SBUS_REQUEST_DATA_ADDR_SHIFT)
-		| ((u64)receiver_addr
-			<< ASIC_CFG_SBUS_REQUEST_RECEIVER_ADDR_SHIFT));
+		  ((u64)data_in << ASIC_CFG_SBUS_REQUEST_DATA_IN_SHIFT) |
+		  ((u64)command << ASIC_CFG_SBUS_REQUEST_COMMAND_SHIFT) |
+		  ((u64)data_addr << ASIC_CFG_SBUS_REQUEST_DATA_ADDR_SHIFT) |
+		  ((u64)receiver_addr <<
+		   ASIC_CFG_SBUS_REQUEST_RECEIVER_ADDR_SHIFT));
 }
 
 /*
@@ -1072,14 +1093,14 @@ static void turn_off_spicos(struct hfi1_devdata *dd, int flags)
 		return;
 
 	dd_dev_info(dd, "Turning off spicos:%s%s\n",
-		flags & SPICO_SBUS ? " SBus" : "",
-		flags & SPICO_FABRIC ? " fabric" : "");
+		    flags & SPICO_SBUS ? " SBus" : "",
+		    flags & SPICO_FABRIC ? " fabric" : "");
 
 	write_csr(dd, MISC_CFG_FW_CTRL, ENABLE_SPICO_SMASK);
 	/* disable SBus spico */
 	if (flags & SPICO_SBUS)
 		sbus_request(dd, SBUS_MASTER_BROADCAST, 0x01,
-			WRITE_SBUS_RECEIVER, 0x00000040);
+			     WRITE_SBUS_RECEIVER, 0x00000040);
 
 	/* disable the fabric serdes spicos */
 	if (flags & SPICO_FABRIC)
@@ -1089,29 +1110,60 @@ static void turn_off_spicos(struct hfi1_devdata *dd, int flags)
 }
 
 /*
- *  Reset all of the fabric serdes for our HFI.
+ * Reset all of the fabric serdes for this HFI in preparation to take the
+ * link to Polling.
+ *
+ * To do a reset, we need to write to to the serdes registers.  Unfortunately,
+ * the fabric serdes download to the other HFI on the ASIC will have turned
+ * off the firmware validation on this HFI.  This means we can't write to the
+ * registers to reset the serdes.  Work around this by performing a complete
+ * re-download and validation of the fabric serdes firmware.  This, as a
+ * by-product, will reset the serdes.  NOTE: the re-download requires that
+ * the 8051 be in the Offline state.  I.e. not actively trying to use the
+ * serdes.  This routine is called at the point where the link is Offline and
+ * is getting ready to go to Polling.
  */
 void fabric_serdes_reset(struct hfi1_devdata *dd)
 {
-	u8 ra;
+	int ret;
 
-	if (dd->icode != ICODE_RTL_SILICON) /* only for RTL */
+	if (!fw_fabric_serdes_load)
 		return;
 
-	ra = fabric_serdes_broadcast[dd->hfi1_id];
-
-	acquire_hw_mutex(dd);
+	ret = acquire_chip_resource(dd, CR_SBUS, SBUS_TIMEOUT);
+	if (ret) {
+		dd_dev_err(dd,
+			   "Cannot acquire SBus resource to reset fabric SerDes - perhaps you should reboot\n");
+		return;
+	}
 	set_sbus_fast_mode(dd);
-	/* place SerDes in reset and disable SPICO */
-	sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000011);
-	/* wait 100 refclk cycles @ 156.25MHz => 640ns */
-	udelay(1);
-	/* remove SerDes reset */
-	sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000010);
-	/* turn SPICO enable on */
-	sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000002);
+
+	if (is_ax(dd)) {
+		/* A0 serdes do not work with a re-download */
+		u8 ra = fabric_serdes_broadcast[dd->hfi1_id];
+
+		/* place SerDes in reset and disable SPICO */
+		sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000011);
+		/* wait 100 refclk cycles @ 156.25MHz => 640ns */
+		udelay(1);
+		/* remove SerDes reset */
+		sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000010);
+		/* turn SPICO enable on */
+		sbus_request(dd, ra, 0x07, WRITE_SBUS_RECEIVER, 0x00000002);
+	} else {
+		turn_off_spicos(dd, SPICO_FABRIC);
+		/*
+		 * No need for firmware retry - what to download has already
+		 * been decided.
+		 * No need to pay attention to the load return - the only
+		 * failure is a validation failure, which has already been
+		 * checked by the initial download.
+		 */
+		(void)load_fabric_serdes_firmware(dd, &fw_fabric);
+	}
+
 	clear_sbus_fast_mode(dd);
-	release_hw_mutex(dd);
+	release_chip_resource(dd, CR_SBUS);
 }
 
 /* Access to the SBus in this routine should probably be serialized */
@@ -1120,6 +1172,9 @@ int sbus_request_slow(struct hfi1_devdata *dd,
 {
 	u64 reg, count = 0;
 
+	/* make sure fast mode is clear */
+	clear_sbus_fast_mode(dd);
+
 	sbus_request(dd, receiver_addr, data_addr, command, data_in);
 	write_csr(dd, ASIC_CFG_SBUS_EXECUTE,
 		  ASIC_CFG_SBUS_EXECUTE_EXECUTE_SMASK);
@@ -1177,7 +1232,7 @@ static int load_fabric_serdes_firmware(struct hfi1_devdata *dd,
 	/* step 5: download SerDes machine code */
 	for (i = 0; i < fdet->firmware_len; i += 4) {
 		sbus_request(dd, ra, 0x0a, WRITE_SBUS_RECEIVER,
-					*(u32 *)&fdet->firmware_ptr[i]);
+			     *(u32 *)&fdet->firmware_ptr[i]);
 	}
 	/* step 6: IMEM override off */
 	sbus_request(dd, ra, 0x00, WRITE_SBUS_RECEIVER, 0x00000000);
@@ -1216,7 +1271,7 @@ static int load_sbus_firmware(struct hfi1_devdata *dd,
 	/* step 5: download the SBus Master machine code */
 	for (i = 0; i < fdet->firmware_len; i += 4) {
 		sbus_request(dd, ra, 0x14, WRITE_SBUS_RECEIVER,
-					*(u32 *)&fdet->firmware_ptr[i]);
+			     *(u32 *)&fdet->firmware_ptr[i]);
 	}
 	/* step 6: set IMEM_CNTL_EN off */
 	sbus_request(dd, ra, 0x01, WRITE_SBUS_RECEIVER, 0x00000040);
@@ -1249,19 +1304,23 @@ static int load_pcie_serdes_firmware(struct hfi1_devdata *dd,
 	/* step 3: enable XDMEM access */
 	sbus_request(dd, ra, 0x01, WRITE_SBUS_RECEIVER, 0x00000d40);
 	/* step 4: load firmware into SBus Master XDMEM */
-	/* NOTE: the dmem address, write_en, and wdata are all pre-packed,
-	   we only need to pick up the bytes and write them */
+	/*
+	 * NOTE: the dmem address, write_en, and wdata are all pre-packed,
+	 * we only need to pick up the bytes and write them
+	 */
 	for (i = 0; i < fdet->firmware_len; i += 4) {
 		sbus_request(dd, ra, 0x04, WRITE_SBUS_RECEIVER,
-					*(u32 *)&fdet->firmware_ptr[i]);
+			     *(u32 *)&fdet->firmware_ptr[i]);
 	}
 	/* step 5: disable XDMEM access */
 	sbus_request(dd, ra, 0x01, WRITE_SBUS_RECEIVER, 0x00000140);
 	/* step 6: allow SBus Spico to run */
 	sbus_request(dd, ra, 0x05, WRITE_SBUS_RECEIVER, 0x00000000);
 
-	/* steps 7-11: run RSA, if it succeeds, firmware is available to
-	   be swapped */
+	/*
+	 * steps 7-11: run RSA, if it succeeds, firmware is available to
+	 * be swapped
+	 */
 	return run_rsa(dd, "PCIe serdes", fdet->signature);
 }
 
@@ -1285,7 +1344,7 @@ static void set_serdes_broadcast(struct hfi1_devdata *dd, u8 bg1, u8 bg2,
 		 *	23:16	BROADCAST_GROUP_2 (default 0xff)
 		 */
 		sbus_request(dd, addrs[count], 0xfd, WRITE_SBUS_RECEIVER,
-				(u32)bg1 << 4 | (u32)bg2 << 16);
+			     (u32)bg1 << 4 | (u32)bg2 << 16);
 	}
 }
 
@@ -1310,8 +1369,8 @@ retry:
 
 	/* timed out */
 	dd_dev_err(dd,
-		"Unable to acquire hardware mutex, mutex mask %u, my mask %u (%s)\n",
-		(u32)user, (u32)mask, (try == 0) ? "retrying" : "giving up");
+		   "Unable to acquire hardware mutex, mutex mask %u, my mask %u (%s)\n",
+		   (u32)user, (u32)mask, (try == 0) ? "retrying" : "giving up");
 
 	if (try == 0) {
 		/* break mutex and retry */
@@ -1328,10 +1387,197 @@ void release_hw_mutex(struct hfi1_devdata *dd)
 	write_csr(dd, ASIC_CFG_MUTEX, 0);
 }
 
+/* return the given resource bit(s) as a mask for the given HFI */
+static inline u64 resource_mask(u32 hfi1_id, u32 resource)
+{
+	return ((u64)resource) << (hfi1_id ? CR_DYN_SHIFT : 0);
+}
+
+static void fail_mutex_acquire_message(struct hfi1_devdata *dd,
+				       const char *func)
+{
+	dd_dev_err(dd,
+		   "%s: hardware mutex stuck - suggest rebooting the machine\n",
+		   func);
+}
+
+/*
+ * Acquire access to a chip resource.
+ *
+ * Return 0 on success, -EBUSY if resource busy, -EIO if mutex acquire failed.
+ */
+static int __acquire_chip_resource(struct hfi1_devdata *dd, u32 resource)
+{
+	u64 scratch0, all_bits, my_bit;
+	int ret;
+
+	if (resource & CR_DYN_MASK) {
+		/* a dynamic resource is in use if either HFI has set the bit */
+		all_bits = resource_mask(0, resource) |
+						resource_mask(1, resource);
+		my_bit = resource_mask(dd->hfi1_id, resource);
+	} else {
+		/* non-dynamic resources are not split between HFIs */
+		all_bits = resource;
+		my_bit = resource;
+	}
+
+	/* lock against other callers within the driver wanting a resource */
+	mutex_lock(&dd->asic_data->asic_resource_mutex);
+
+	ret = acquire_hw_mutex(dd);
+	if (ret) {
+		fail_mutex_acquire_message(dd, __func__);
+		ret = -EIO;
+		goto done;
+	}
+
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	if (scratch0 & all_bits) {
+		ret = -EBUSY;
+	} else {
+		write_csr(dd, ASIC_CFG_SCRATCH, scratch0 | my_bit);
+		/* force write to be visible to other HFI on another OS */
+		(void)read_csr(dd, ASIC_CFG_SCRATCH);
+	}
+
+	release_hw_mutex(dd);
+
+done:
+	mutex_unlock(&dd->asic_data->asic_resource_mutex);
+	return ret;
+}
+
+/*
+ * Acquire access to a chip resource, wait up to mswait milliseconds for
+ * the resource to become available.
+ *
+ * Return 0 on success, -EBUSY if busy (even after wait), -EIO if mutex
+ * acquire failed.
+ */
+int acquire_chip_resource(struct hfi1_devdata *dd, u32 resource, u32 mswait)
+{
+	unsigned long timeout;
+	int ret;
+
+	timeout = jiffies + msecs_to_jiffies(mswait);
+	while (1) {
+		ret = __acquire_chip_resource(dd, resource);
+		if (ret != -EBUSY)
+			return ret;
+		/* resource is busy, check our timeout */
+		if (time_after_eq(jiffies, timeout))
+			return -EBUSY;
+		usleep_range(80, 120);	/* arbitrary delay */
+	}
+}
+
+/*
+ * Release access to a chip resource
+ */
+void release_chip_resource(struct hfi1_devdata *dd, u32 resource)
+{
+	u64 scratch0, bit;
+
+	/* only dynamic resources should ever be cleared */
+	if (!(resource & CR_DYN_MASK)) {
+		dd_dev_err(dd, "%s: invalid resource 0x%x\n", __func__,
+			   resource);
+		return;
+	}
+	bit = resource_mask(dd->hfi1_id, resource);
+
+	/* lock against other callers within the driver wanting a resource */
+	mutex_lock(&dd->asic_data->asic_resource_mutex);
+
+	if (acquire_hw_mutex(dd)) {
+		fail_mutex_acquire_message(dd, __func__);
+		goto done;
+	}
+
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	if ((scratch0 & bit) != 0) {
+		scratch0 &= ~bit;
+		write_csr(dd, ASIC_CFG_SCRATCH, scratch0);
+		/* force write to be visible to other HFI on another OS */
+		(void)read_csr(dd, ASIC_CFG_SCRATCH);
+	} else {
+		dd_dev_warn(dd, "%s: id %d, resource 0x%x: bit not set\n",
+			    __func__, dd->hfi1_id, resource);
+	}
+
+	release_hw_mutex(dd);
+
+done:
+	mutex_unlock(&dd->asic_data->asic_resource_mutex);
+}
+
+/*
+ * Return true if resource is set, false otherwise.  Print a warning
+ * if not set and a function is supplied.
+ */
+bool check_chip_resource(struct hfi1_devdata *dd, u32 resource,
+			 const char *func)
+{
+	u64 scratch0, bit;
+
+	if (resource & CR_DYN_MASK)
+		bit = resource_mask(dd->hfi1_id, resource);
+	else
+		bit = resource;
+
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	if ((scratch0 & bit) == 0) {
+		if (func)
+			dd_dev_warn(dd,
+				    "%s: id %d, resource 0x%x, not acquired!\n",
+				    func, dd->hfi1_id, resource);
+		return false;
+	}
+	return true;
+}
+
+static void clear_chip_resources(struct hfi1_devdata *dd, const char *func)
+{
+	u64 scratch0;
+
+	/* lock against other callers within the driver wanting a resource */
+	mutex_lock(&dd->asic_data->asic_resource_mutex);
+
+	if (acquire_hw_mutex(dd)) {
+		fail_mutex_acquire_message(dd, func);
+		goto done;
+	}
+
+	/* clear all dynamic access bits for this HFI */
+	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
+	scratch0 &= ~resource_mask(dd->hfi1_id, CR_DYN_MASK);
+	write_csr(dd, ASIC_CFG_SCRATCH, scratch0);
+	/* force write to be visible to other HFI on another OS */
+	(void)read_csr(dd, ASIC_CFG_SCRATCH);
+
+	release_hw_mutex(dd);
+
+done:
+	mutex_unlock(&dd->asic_data->asic_resource_mutex);
+}
+
+void init_chip_resources(struct hfi1_devdata *dd)
+{
+	/* clear any holds left by us */
+	clear_chip_resources(dd, __func__);
+}
+
+void finish_chip_resources(struct hfi1_devdata *dd)
+{
+	/* clear any holds left by us */
+	clear_chip_resources(dd, __func__);
+}
+
 void set_sbus_fast_mode(struct hfi1_devdata *dd)
 {
 	write_csr(dd, ASIC_CFG_SBUS_EXECUTE,
-				ASIC_CFG_SBUS_EXECUTE_FAST_MODE_SMASK);
+		  ASIC_CFG_SBUS_EXECUTE_FAST_MODE_SMASK);
 }
 
 void clear_sbus_fast_mode(struct hfi1_devdata *dd)
@@ -1354,23 +1600,23 @@ int load_firmware(struct hfi1_devdata *dd)
 	int ret;
 
 	if (fw_fabric_serdes_load) {
-		ret = acquire_hw_mutex(dd);
+		ret = acquire_chip_resource(dd, CR_SBUS, SBUS_TIMEOUT);
 		if (ret)
 			return ret;
 
 		set_sbus_fast_mode(dd);
 
 		set_serdes_broadcast(dd, all_fabric_serdes_broadcast,
-				fabric_serdes_broadcast[dd->hfi1_id],
-				fabric_serdes_addrs[dd->hfi1_id],
-				NUM_FABRIC_SERDES);
+				     fabric_serdes_broadcast[dd->hfi1_id],
+				     fabric_serdes_addrs[dd->hfi1_id],
+				     NUM_FABRIC_SERDES);
 		turn_off_spicos(dd, SPICO_FABRIC);
 		do {
 			ret = load_fabric_serdes_firmware(dd, &fw_fabric);
 		} while (retry_firmware(dd, ret));
 
 		clear_sbus_fast_mode(dd);
-		release_hw_mutex(dd);
+		release_chip_resource(dd, CR_SBUS);
 		if (ret)
 			return ret;
 	}
@@ -1419,18 +1665,57 @@ int hfi1_firmware_init(struct hfi1_devdata *dd)
 	return obtain_firmware(dd);
 }
 
+/*
+ * This function is a helper function for parse_platform_config(...) and
+ * does not check for validity of the platform configuration cache
+ * (because we know it is invalid as we are building up the cache).
+ * As such, this should not be called from anywhere other than
+ * parse_platform_config
+ */
+static int check_meta_version(struct hfi1_devdata *dd, u32 *system_table)
+{
+	u32 meta_ver, meta_ver_meta, ver_start, ver_len, mask;
+	struct platform_config_cache *pcfgcache = &dd->pcfg_cache;
+
+	if (!system_table)
+		return -EINVAL;
+
+	meta_ver_meta =
+	*(pcfgcache->config_tables[PLATFORM_CONFIG_SYSTEM_TABLE].table_metadata
+	+ SYSTEM_TABLE_META_VERSION);
+
+	mask = ((1 << METADATA_TABLE_FIELD_START_LEN_BITS) - 1);
+	ver_start = meta_ver_meta & mask;
+
+	meta_ver_meta >>= METADATA_TABLE_FIELD_LEN_SHIFT;
+
+	mask = ((1 << METADATA_TABLE_FIELD_LEN_LEN_BITS) - 1);
+	ver_len = meta_ver_meta & mask;
+
+	ver_start /= 8;
+	meta_ver = *((u8 *)system_table + ver_start) & ((1 << ver_len) - 1);
+
+	if (meta_ver < 5) {
+		dd_dev_info(
+			dd, "%s:Please update platform config\n", __func__);
+		return -EINVAL;
+	}
+	return 0;
+}
+
 int parse_platform_config(struct hfi1_devdata *dd)
 {
 	struct platform_config_cache *pcfgcache = &dd->pcfg_cache;
 	u32 *ptr = NULL;
-	u32 header1 = 0, header2 = 0, magic_num = 0, crc = 0;
+	u32 header1 = 0, header2 = 0, magic_num = 0, crc = 0, file_length = 0;
 	u32 record_idx = 0, table_type = 0, table_length_dwords = 0;
+	int ret = -EINVAL; /* assume failure */
 
-	if (platform_config == NULL) {
+	if (!dd->platform_config.data) {
 		dd_dev_info(dd, "%s: Missing config file\n", __func__);
 		goto bail;
 	}
-	ptr = (u32 *)platform_config->data;
+	ptr = (u32 *)dd->platform_config.data;
 
 	magic_num = *ptr;
 	ptr++;
@@ -1439,12 +1724,32 @@ int parse_platform_config(struct hfi1_devdata *dd)
 		goto bail;
 	}
 
-	while (ptr < (u32 *)(platform_config->data + platform_config->size)) {
+	/* Field is file size in DWORDs */
+	file_length = (*ptr) * 4;
+	ptr++;
+
+	if (file_length > dd->platform_config.size) {
+		dd_dev_info(dd, "%s:File claims to be larger than read size\n",
+			    __func__);
+		goto bail;
+	} else if (file_length < dd->platform_config.size) {
+		dd_dev_info(dd,
+			    "%s:File claims to be smaller than read size, continuing\n",
+			    __func__);
+	}
+	/* exactly equal, perfection */
+
+	/*
+	 * In both cases where we proceed, using the self-reported file length
+	 * is the safer option
+	 */
+	while (ptr < (u32 *)(dd->platform_config.data + file_length)) {
 		header1 = *ptr;
 		header2 = *(ptr + 1);
 		if (header1 != ~header2) {
 			dd_dev_info(dd, "%s: Failed validation at offset %ld\n",
-				__func__, (ptr - (u32 *)platform_config->data));
+				    __func__, (ptr - (u32 *)
+					       dd->platform_config.data));
 			goto bail;
 		}
 
@@ -1467,6 +1772,9 @@ int parse_platform_config(struct hfi1_devdata *dd)
 			case PLATFORM_CONFIG_SYSTEM_TABLE:
 				pcfgcache->config_tables[table_type].num_table =
 									1;
+				ret = check_meta_version(dd, ptr);
+				if (ret)
+					goto bail;
 				break;
 			case PLATFORM_CONFIG_PORT_TABLE:
 				pcfgcache->config_tables[table_type].num_table =
@@ -1484,9 +1792,10 @@ int parse_platform_config(struct hfi1_devdata *dd)
 				break;
 			default:
 				dd_dev_info(dd,
-				      "%s: Unknown data table %d, offset %ld\n",
-					__func__, table_type,
-				       (ptr - (u32 *)platform_config->data));
+					    "%s: Unknown data table %d, offset %ld\n",
+					    __func__, table_type,
+					    (ptr - (u32 *)
+					     dd->platform_config.data));
 				goto bail; /* We don't trust this file now */
 			}
 			pcfgcache->config_tables[table_type].table = ptr;
@@ -1507,9 +1816,10 @@ int parse_platform_config(struct hfi1_devdata *dd)
 				break;
 			default:
 				dd_dev_info(dd,
-				  "%s: Unknown metadata table %d, offset %ld\n",
-				  __func__, table_type,
-				  (ptr - (u32 *)platform_config->data));
+					    "%s: Unknown meta table %d, offset %ld\n",
+					    __func__, table_type,
+					    (ptr -
+					     (u32 *)dd->platform_config.data));
 				goto bail; /* We don't trust this file now */
 			}
 			pcfgcache->config_tables[table_type].table_metadata =
@@ -1518,14 +1828,16 @@ int parse_platform_config(struct hfi1_devdata *dd)
 
 		/* Calculate and check table crc */
 		crc = crc32_le(~(u32)0, (unsigned char const *)ptr,
-				(table_length_dwords * 4));
+			       (table_length_dwords * 4));
 		crc ^= ~(u32)0;
 
 		/* Jump the table */
 		ptr += table_length_dwords;
 		if (crc != *ptr) {
 			dd_dev_info(dd, "%s: Failed CRC check at offset %ld\n",
-				__func__, (ptr - (u32 *)platform_config->data));
+				    __func__, (ptr -
+					       (u32 *)
+					       dd->platform_config.data));
 			goto bail;
 		}
 		/* Jump the CRC DWORD */
@@ -1536,11 +1848,12 @@ int parse_platform_config(struct hfi1_devdata *dd)
 	return 0;
 bail:
 	memset(pcfgcache, 0, sizeof(struct platform_config_cache));
-	return -EINVAL;
+	return ret;
 }
 
 static int get_platform_fw_field_metadata(struct hfi1_devdata *dd, int table,
-		int field, u32 *field_len_bits, u32 *field_start_bits)
+					  int field, u32 *field_len_bits,
+					  u32 *field_start_bits)
 {
 	struct platform_config_cache *pcfgcache = &dd->pcfg_cache;
 	u32 *src_ptr = NULL;
@@ -1600,8 +1913,9 @@ static int get_platform_fw_field_metadata(struct hfi1_devdata *dd, int table,
  * @len: length of memory pointed by @data in bytes.
  */
 int get_platform_config_field(struct hfi1_devdata *dd,
-			enum platform_config_table_type_encoding table_type,
-			int table_index, int field_index, u32 *data, u32 len)
+			      enum platform_config_table_type_encoding
+			      table_type, int table_index, int field_index,
+			      u32 *data, u32 len)
 {
 	int ret = 0, wlen = 0, seek = 0;
 	u32 field_len_bits = 0, field_start_bits = 0, *src_ptr = NULL;
@@ -1613,7 +1927,8 @@ int get_platform_config_field(struct hfi1_devdata *dd,
 		return -EINVAL;
 
 	ret = get_platform_fw_field_metadata(dd, table_type, field_index,
-					&field_len_bits, &field_start_bits);
+					     &field_len_bits,
+					     &field_start_bits);
 	if (ret)
 		return -EINVAL;
 
@@ -1629,19 +1944,21 @@ int get_platform_config_field(struct hfi1_devdata *dd,
 			if (len < field_len_bits)
 				return -EINVAL;
 
-			seek = field_start_bits/8;
-			wlen = field_len_bits/8;
+			seek = field_start_bits / 8;
+			wlen = field_len_bits / 8;
 
 			src_ptr = (u32 *)((u8 *)src_ptr + seek);
 
-			/* We expect the field to be byte aligned and whole byte
-			 * lengths if we are here */
+			/*
+			 * We expect the field to be byte aligned and whole byte
+			 * lengths if we are here
+			 */
 			memcpy(data, src_ptr, wlen);
 			return 0;
 		}
 		break;
 	case PLATFORM_CONFIG_PORT_TABLE:
-		/* Port table is 4 DWORDS in META_VERSION 0 */
+		/* Port table is 4 DWORDS */
 		src_ptr = dd->hfi1_id ?
 			pcfgcache->config_tables[table_type].table + 4 :
 			pcfgcache->config_tables[table_type].table;
@@ -1669,7 +1986,7 @@ int get_platform_config_field(struct hfi1_devdata *dd,
 	if (!src_ptr || len < field_len_bits)
 		return -EINVAL;
 
-	src_ptr += (field_start_bits/32);
+	src_ptr += (field_start_bits / 32);
 	*data = (*src_ptr >> (field_start_bits % 32)) &
 			((1 << field_len_bits) - 1);
 
@@ -1680,7 +1997,7 @@ int get_platform_config_field(struct hfi1_devdata *dd,
  * Download the firmware needed for the Gen3 PCIe SerDes.  An update
  * to the SBus firmware is needed before updating the PCIe firmware.
  *
- * Note: caller must be holding the HW mutex.
+ * Note: caller must be holding the SBus resource.
  */
 int load_pcie_firmware(struct hfi1_devdata *dd)
 {
@@ -1701,9 +2018,9 @@ int load_pcie_firmware(struct hfi1_devdata *dd)
 	if (fw_pcie_serdes_load) {
 		dd_dev_info(dd, "Setting PCIe SerDes broadcast\n");
 		set_serdes_broadcast(dd, all_pcie_serdes_broadcast,
-					pcie_serdes_broadcast[dd->hfi1_id],
-					pcie_serdes_addrs[dd->hfi1_id],
-					NUM_PCIE_SERDES);
+				     pcie_serdes_broadcast[dd->hfi1_id],
+				     pcie_serdes_addrs[dd->hfi1_id],
+				     NUM_PCIE_SERDES);
 		do {
 			ret = load_pcie_serdes_firmware(dd, &fw_pcie);
 		} while (retry_firmware(dd, ret));
@@ -1724,9 +2041,9 @@ void read_guid(struct hfi1_devdata *dd)
 {
 	/* Take the DC out of reset to get a valid GUID value */
 	write_csr(dd, CCE_DC_CTRL, 0);
-	(void) read_csr(dd, CCE_DC_CTRL);
+	(void)read_csr(dd, CCE_DC_CTRL);
 
 	dd->base_guid = read_csr(dd, DC_DC8051_CFG_LOCAL_GUID);
 	dd_dev_info(dd, "GUID %llx",
-		(unsigned long long)dd->base_guid);
+		    (unsigned long long)dd->base_guid);
 }
diff --git a/drivers/staging/rdma/hfi1/hfi.h b/drivers/staging/rdma/hfi1/hfi.h
index d4826a9ab8d3..16cbdc4073e0 100644
--- a/drivers/staging/rdma/hfi1/hfi.h
+++ b/drivers/staging/rdma/hfi1/hfi.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_KERNEL_H
 #define _HFI1_KERNEL_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -65,6 +62,7 @@
 #include <linux/cdev.h>
 #include <linux/delay.h>
 #include <linux/kthread.h>
+#include <rdma/rdma_vt.h>
 
 #include "chip_registers.h"
 #include "common.h"
@@ -73,7 +71,8 @@
 #include "chip.h"
 #include "mad.h"
 #include "qsfp.h"
-#include "platform_config.h"
+#include "platform.h"
+#include "affinity.h"
 
 /* bumped 1 from s/w major version of TrueScale */
 #define HFI1_CHIP_VERS_MAJ 3U
@@ -98,6 +97,8 @@ extern unsigned long hfi1_cap_mask;
 #define HFI1_CAP_IS_USET(cap) (!!HFI1_CAP_UGET(cap))
 #define HFI1_MISC_GET() ((hfi1_cap_mask >> HFI1_CAP_MISC_SHIFT) & \
 			HFI1_CAP_MISC_MASK)
+/* Offline Disabled Reason is 4-bits */
+#define HFI1_ODR_MASK(rsn) ((rsn) & OPA_PI_MASK_OFFLINE_REASON)
 
 /*
  * Control context is always 0 and handles the error packets.
@@ -177,6 +178,11 @@ struct ctxt_eager_bufs {
 	} *rcvtids;
 };
 
+struct exp_tid_set {
+	struct list_head list;
+	u32 count;
+};
+
 struct hfi1_ctxtdata {
 	/* shadow the ctxt's RcvCtrl register */
 	u64 rcvctrl;
@@ -233,20 +239,13 @@ struct hfi1_ctxtdata {
 	u32 expected_count;
 	/* index of first expected TID entry. */
 	u32 expected_base;
-	/* cursor into the exp group sets */
-	atomic_t tidcursor;
-	/* number of exp TID groups assigned to the ctxt */
-	u16 numtidgroups;
-	/* size of exp TID group fields in tidusemap */
-	u16 tidmapcnt;
-	/* exp TID group usage bitfield array */
-	unsigned long *tidusemap;
-	/* pinned pages for exp sends, allocated at open */
-	struct page **tid_pg_list;
-	/* dma handles for exp tid pages */
-	dma_addr_t *physshadow;
+
+	struct exp_tid_set tid_group_list;
+	struct exp_tid_set tid_used_list;
+	struct exp_tid_set tid_full_list;
+
 	/* lock protecting all Expected TID data */
-	spinlock_t exp_lock;
+	struct mutex exp_lock;
 	/* number of pio bufs for this ctxt (all procs, if shared) */
 	u32 piocnt;
 	/* first pio buffer for this ctxt */
@@ -311,8 +310,24 @@ struct hfi1_ctxtdata {
 	 */
 	struct task_struct *progress;
 	struct list_head sdma_queues;
+	/* protect sdma queues */
 	spinlock_t sdma_qlock;
 
+	/* Is ASPM interrupt supported for this context */
+	bool aspm_intr_supported;
+	/* ASPM state (enabled/disabled) for this context */
+	bool aspm_enabled;
+	/* Timer for re-enabling ASPM if interrupt activity quietens down */
+	struct timer_list aspm_timer;
+	/* Lock to serialize between intr, timer intr and user threads */
+	spinlock_t aspm_lock;
+	/* Is ASPM processing enabled for this context (in intr context) */
+	bool aspm_intr_enable;
+	/* Last interrupt timestamp */
+	ktime_t aspm_ts_last_intr;
+	/* Last timestamp at which we scheduled a timer for this context */
+	ktime_t aspm_ts_timer_sched;
+
 	/*
 	 * The interrupt handler for a particular receive context can vary
 	 * throughout it's lifetime. This is not a lock protected data member so
@@ -335,7 +350,7 @@ struct hfi1_packet {
 	void *hdr;
 	struct hfi1_ctxtdata *rcd;
 	__le32 *rhf_addr;
-	struct hfi1_qp *qp;
+	struct rvt_qp *qp;
 	struct hfi1_other_headers *ohdr;
 	u64 rhf;
 	u32 maxcnt;
@@ -363,6 +378,7 @@ struct hfi1_snoop_data {
 	int mode_flag;
 	struct cdev cdev;
 	struct device *class_dev;
+	/* protect snoop data */
 	spinlock_t snoop_lock;
 	struct list_head queue;
 	wait_queue_head_t waitq;
@@ -375,7 +391,7 @@ struct hfi1_snoop_data {
 #define HFI1_PORT_SNOOP_MODE     1U
 #define HFI1_PORT_CAPTURE_MODE   2U
 
-struct hfi1_sge_state;
+struct rvt_sge_state;
 
 /*
  * Get/Set IB link-level config parameters for f_get/set_ib_cfg()
@@ -424,17 +440,17 @@ struct hfi1_sge_state;
 #define __HLS_GOING_OFFLINE_BP  9
 #define __HLS_LINK_COOLDOWN_BP 10
 
-#define HLS_UP_INIT	  (1 << __HLS_UP_INIT_BP)
-#define HLS_UP_ARMED	  (1 << __HLS_UP_ARMED_BP)
-#define HLS_UP_ACTIVE	  (1 << __HLS_UP_ACTIVE_BP)
-#define HLS_DN_DOWNDEF	  (1 << __HLS_DN_DOWNDEF_BP) /* link down default */
-#define HLS_DN_POLL	  (1 << __HLS_DN_POLL_BP)
-#define HLS_DN_DISABLE	  (1 << __HLS_DN_DISABLE_BP)
-#define HLS_DN_OFFLINE	  (1 << __HLS_DN_OFFLINE_BP)
-#define HLS_VERIFY_CAP	  (1 << __HLS_VERIFY_CAP_BP)
-#define HLS_GOING_UP	  (1 << __HLS_GOING_UP_BP)
-#define HLS_GOING_OFFLINE (1 << __HLS_GOING_OFFLINE_BP)
-#define HLS_LINK_COOLDOWN (1 << __HLS_LINK_COOLDOWN_BP)
+#define HLS_UP_INIT	  BIT(__HLS_UP_INIT_BP)
+#define HLS_UP_ARMED	  BIT(__HLS_UP_ARMED_BP)
+#define HLS_UP_ACTIVE	  BIT(__HLS_UP_ACTIVE_BP)
+#define HLS_DN_DOWNDEF	  BIT(__HLS_DN_DOWNDEF_BP) /* link down default */
+#define HLS_DN_POLL	  BIT(__HLS_DN_POLL_BP)
+#define HLS_DN_DISABLE	  BIT(__HLS_DN_DISABLE_BP)
+#define HLS_DN_OFFLINE	  BIT(__HLS_DN_OFFLINE_BP)
+#define HLS_VERIFY_CAP	  BIT(__HLS_VERIFY_CAP_BP)
+#define HLS_GOING_UP	  BIT(__HLS_GOING_UP_BP)
+#define HLS_GOING_OFFLINE BIT(__HLS_GOING_OFFLINE_BP)
+#define HLS_LINK_COOLDOWN BIT(__HLS_LINK_COOLDOWN_BP)
 
 #define HLS_UP (HLS_UP_INIT | HLS_UP_ARMED | HLS_UP_ACTIVE)
 
@@ -490,6 +506,7 @@ struct hfi1_sge_state;
 #define CNTR_DISABLED		0x2 /* Disable this counter */
 #define CNTR_32BIT		0x4 /* Simulate 64 bits for this counter */
 #define CNTR_VL			0x8 /* Per VL counter */
+#define CNTR_SDMA              0x10
 #define CNTR_INVALID_VL		-1  /* Specifies invalid VL */
 #define CNTR_MODE_W		0x0
 #define CNTR_MODE_R		0x1
@@ -512,10 +529,11 @@ static inline void incr_cntr32(u32 *cntr)
 
 #define MAX_NAME_SIZE 64
 struct hfi1_msix_entry {
+	enum irq_type type;
 	struct msix_entry msix;
 	void *arg;
 	char name[MAX_NAME_SIZE];
-	cpumask_var_t mask;
+	cpumask_t mask;
 };
 
 /* per-SL CCA information */
@@ -542,6 +560,7 @@ enum {
 };
 
 struct vl_arb_cache {
+	/* protect vl arb cache */
 	spinlock_t lock;
 	struct ib_vl_weight_elem table[VL_ARB_TABLE_SIZE];
 };
@@ -561,7 +580,8 @@ struct hfi1_pportdata {
 	struct kobject sl2sc_kobj;
 	struct kobject vl2mtu_kobj;
 
-	/* QSFP support */
+	/* PHY support */
+	u32 port_type;
 	struct qsfp_data qsfp_info;
 
 	/* GUID for this interface, in host order */
@@ -586,6 +606,7 @@ struct hfi1_pportdata {
 	struct work_struct link_vc_work;
 	struct work_struct link_up_work;
 	struct work_struct link_down_work;
+	struct work_struct dc_host_req_work;
 	struct work_struct sma_message_work;
 	struct work_struct freeze_work;
 	struct work_struct link_downgrade_work;
@@ -623,6 +644,7 @@ struct hfi1_pportdata {
 	u16 link_speed_active;
 	u8 vls_supported;
 	u8 vls_operational;
+	u8 actual_vls_operational;
 	/* LID mask control */
 	u8 lmc;
 	/* Rx Polarity inversion (compensate for ~tx on partner) */
@@ -642,19 +664,23 @@ struct hfi1_pportdata {
 	u8 link_enabled;	/* link enabled? */
 	u8 linkinit_reason;
 	u8 local_tx_rate;	/* rate given to 8051 firmware */
+	u8 last_pstate;		/* info only */
 
 	/* placeholders for IB MAD packet settings */
 	u8 overrun_threshold;
 	u8 phy_error_threshold;
 
-	/* used to override LED behavior */
-	u8 led_override;  /* Substituted for normal value, if non-zero */
-	u16 led_override_timeoff; /* delta to next timer event */
-	u8 led_override_vals[2]; /* Alternates per blink-frame */
-	u8 led_override_phase; /* Just counts, LSB picks from vals[] */
+	/* Used to override LED behavior for things like maintenance beaconing*/
+	/*
+	 * Alternates per phase of blink
+	 * [0] holds LED off duration, [1] holds LED on duration
+	 */
+	unsigned long led_override_vals[2];
+	u8 led_override_phase; /* LSB picks from vals[] */
 	atomic_t led_override_timer_active;
 	/* Used to flash LEDs in override mode */
 	struct timer_list led_override_timer;
+
 	u32 sm_trap_qp;
 	u32 sa_qp;
 
@@ -689,10 +715,12 @@ struct hfi1_pportdata {
 	/* CA's max number of 64 entry units in the congestion control table */
 	u8 cc_max_table_entries;
 
-	/* begin congestion log related entries
-	 * cc_log_lock protects all congestion log related data */
+	/*
+	 * begin congestion log related entries
+	 * cc_log_lock protects all congestion log related data
+	 */
 	spinlock_t cc_log_lock ____cacheline_aligned_in_smp;
-	u8 threshold_cong_event_map[OPA_MAX_SLS/8];
+	u8 threshold_cong_event_map[OPA_MAX_SLS / 8];
 	u16 threshold_event_counter;
 	struct opa_hfi1_cong_log_event_internal cc_events[OPA_CONG_LOG_ELEMS];
 	int cc_log_idx; /* index for logging events */
@@ -705,8 +733,9 @@ struct hfi1_pportdata {
 	u64 *cntrs;
 	/* port relative synthetic counter buffer */
 	u64 *scntrs;
-	/* we synthesize port_xmit_discards from several egress errors */
+	/* port_xmit_discards are synthesized from different egress errors */
 	u64 port_xmit_discards;
+	u64 port_xmit_discards_vl[C_VL_COUNT];
 	u64 port_xmit_constraint_errors;
 	u64 port_rcv_constraint_errors;
 	/* count of 'link_err' interrupts from DC */
@@ -728,6 +757,9 @@ struct hfi1_pportdata {
 	u8 remote_link_down_reason;
 	/* Error events that will cause a port bounce. */
 	u32 port_error_action;
+	struct work_struct linkstate_active_work;
+	/* Does this port need to prescan for FECNs */
+	bool cc_prescan;
 };
 
 typedef int (*rhf_rcv_function_ptr)(struct hfi1_packet *packet);
@@ -773,6 +805,12 @@ struct hfi1_temp {
 	u8 triggers;      /* temperature triggers */
 };
 
+/* common data between shared ASIC HFIs */
+struct hfi1_asic_data {
+	struct hfi1_devdata *dds[2];	/* back pointers */
+	struct mutex asic_resource_mutex;
+};
+
 /* device data struct now contains only "general per-device" info.
  * fields related to a physical IB port are in a hfi1_pportdata struct.
  */
@@ -782,6 +820,7 @@ struct sdma_vl_map;
 #define BOARD_VERS_MAX 96 /* how long the version string can be */
 #define SERIAL_MAX 16 /* length of the serial number */
 
+typedef int (*send_routine)(struct rvt_qp *, struct hfi1_pkt_state *, u64);
 struct hfi1_devdata {
 	struct hfi1_ibdev verbs_dev;     /* must be first */
 	struct list_head list;
@@ -811,6 +850,12 @@ struct hfi1_devdata {
 	spinlock_t sc_lock;
 	/* Per VL data. Enough for all VLs but not all elements are set/used. */
 	struct per_vl_data vld[PER_VL_SEND_CONTEXTS];
+	/* lock for pio_map */
+	spinlock_t pio_map_lock;
+	/* array of kernel send contexts */
+	struct send_context **kernel_send_context;
+	/* array of vl maps */
+	struct pio_vl_map __rcu *pio_map;
 	/* seqlock for sc2vl */
 	seqlock_t sc2vl_lock;
 	u64 sc2vl[4];
@@ -841,6 +886,8 @@ struct hfi1_devdata {
 	wait_queue_head_t		  sdma_unfreeze_wq;
 	atomic_t			  sdma_unfreeze_count;
 
+	/* common data between shared ASIC HFIs in this OS */
+	struct hfi1_asic_data *asic_data;
 
 	/* hfi1_pportdata, points to array of (physical) port-specific
 	 * data structs, indexed by pidx (0..n-1)
@@ -873,10 +920,11 @@ struct hfi1_devdata {
 	/* reset value */
 	u64 z_int_counter;
 	u64 z_rcv_limit;
+	u64 z_send_schedule;
 	/* percpu int_counter */
 	u64 __percpu *int_counter;
 	u64 __percpu *rcv_limit;
-
+	u64 __percpu *send_schedule;
 	/* number of receive contexts in use by the driver */
 	u32 num_rcv_contexts;
 	/* number of pio send contexts in use by the driver */
@@ -885,6 +933,8 @@ struct hfi1_devdata {
 	 * number of ctxts available for PSM open
 	 */
 	u32 freectxts;
+	/* total number of available user/PSM contexts */
+	u32 num_user_contexts;
 	/* base receive interrupt timeout, in CSR units */
 	u32 rcv_intr_timeout_csr;
 
@@ -996,9 +1046,8 @@ struct hfi1_devdata {
 	u16 irev;	/* implementation revision */
 	u16 dc8051_ver; /* 8051 firmware version */
 
+	struct platform_config platform_config;
 	struct platform_config_cache pcfg_cache;
-	/* control high-level access to qsfp */
-	struct mutex qsfp_i2c_mutex;
 
 	struct diag_client *diag_client;
 	spinlock_t hfi1_diag_trans_lock; /* protect diag observer ops */
@@ -1008,8 +1057,6 @@ struct hfi1_devdata {
 	u16 psxmitwait_check_rate;
 	/* high volume overflow errors deferred to tasklet */
 	struct tasklet_struct error_tasklet;
-	/* per device cq worker */
-	struct kthread_worker *worker;
 
 	/* MSI-X information */
 	struct hfi1_msix_entry *msix_entries;
@@ -1090,10 +1137,8 @@ struct hfi1_devdata {
 	 * Handlers for outgoing data so that snoop/capture does not
 	 * have to have its hooks in the send path
 	 */
-	int (*process_pio_send)(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
-				u64 pbc);
-	int (*process_dma_send)(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
-				u64 pbc);
+	send_routine process_pio_send;
+	send_routine process_dma_send;
 	void (*pio_inline_send)(struct hfi1_devdata *dd, struct pio_buf *pbuf,
 				u64 pbc, const void *from, size_t count);
 
@@ -1105,7 +1150,6 @@ struct hfi1_devdata {
 	struct timer_list rcverr_timer;
 	u32 rcv_ovfl_cnt;
 
-	int assigned_node_id;
 	wait_queue_head_t event_queue;
 
 	/* Save the enabled LCB error bits */
@@ -1115,6 +1159,16 @@ struct hfi1_devdata {
 	/* receive context tail dummy address */
 	__le64 *rcvhdrtail_dummy_kvaddr;
 	dma_addr_t rcvhdrtail_dummy_physaddr;
+
+	bool eprom_available;	/* true if EPROM is available for this device */
+	bool aspm_supported;	/* Does HW support ASPM */
+	bool aspm_enabled;	/* ASPM state: enabled/disabled */
+	/* Serialize ASPM enable/disable between multiple verbs contexts */
+	spinlock_t aspm_lock;
+	/* Number of verbs contexts which have disabled ASPM */
+	atomic_t aspm_disabled_cnt;
+
+	struct hfi1_affinity *affinity;
 };
 
 /* 8051 firmware version helper */
@@ -1125,6 +1179,9 @@ struct hfi1_devdata {
 #define PT_EAGER    1
 #define PT_INVALID  2
 
+struct tid_rb_node;
+struct mmu_rb_node;
+
 /* Private data for file operations */
 struct hfi1_filedata {
 	struct hfi1_ctxtdata *uctxt;
@@ -1133,6 +1190,16 @@ struct hfi1_filedata {
 	struct hfi1_user_sdma_pkt_q *pq;
 	/* for cpu affinity; -1 if none */
 	int rec_cpu_num;
+	u32 tid_n_pinned;
+	struct rb_root tid_rb_root;
+	struct tid_rb_node **entry_to_rb;
+	spinlock_t tid_lock; /* protect tid_[limit,used] counters */
+	u32 tid_limit;
+	u32 tid_used;
+	u32 *invalid_tids;
+	u32 invalid_tid_idx;
+	/* protect invalid_tids array and invalid_tid_idx */
+	spinlock_t invalid_lock;
 };
 
 extern struct list_head hfi1_dev_list;
@@ -1156,7 +1223,7 @@ void handle_user_interrupt(struct hfi1_ctxtdata *rcd);
 int hfi1_create_rcvhdrq(struct hfi1_devdata *, struct hfi1_ctxtdata *);
 int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *);
 int hfi1_create_ctxts(struct hfi1_devdata *dd);
-struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *, u32);
+struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *, u32, int);
 void hfi1_init_pportdata(struct pci_dev *, struct hfi1_pportdata *,
 			 struct hfi1_devdata *, u8, u8);
 void hfi1_free_ctxtdata(struct hfi1_devdata *, struct hfi1_ctxtdata *);
@@ -1164,6 +1231,7 @@ void hfi1_free_ctxtdata(struct hfi1_devdata *, struct hfi1_ctxtdata *);
 int handle_receive_interrupt(struct hfi1_ctxtdata *, int);
 int handle_receive_interrupt_nodma_rtail(struct hfi1_ctxtdata *, int);
 int handle_receive_interrupt_dma_rtail(struct hfi1_ctxtdata *, int);
+void set_all_slowpath(struct hfi1_devdata *dd);
 
 /* receive packet handler dispositions */
 #define RCV_PKT_OK      0x0 /* keep going */
@@ -1184,6 +1252,15 @@ static inline u32 driver_lstate(struct hfi1_pportdata *ppd)
 	return ppd->lstate; /* use the cached value */
 }
 
+void receive_interrupt_work(struct work_struct *work);
+
+/* extract service channel from header and rhf */
+static inline int hdr2sc(struct hfi1_message_header *hdr, u64 rhf)
+{
+	return ((be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf) |
+	       ((!!(rhf & RHF_DC_INFO_MASK)) << 4);
+}
+
 static inline u16 generate_jkey(kuid_t uid)
 {
 	return from_kuid(current_user_ns(), uid) & 0xffff;
@@ -1253,7 +1330,7 @@ static inline u32 egress_cycles(u32 len, u32 rate)
 void set_link_ipg(struct hfi1_pportdata *ppd);
 void process_becn(struct hfi1_pportdata *ppd, u8 sl,  u16 rlid, u32 lqpn,
 		  u32 rqpn, u8 svc_type);
-void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
+void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn,
 		u32 pkey, u32 slid, u32 dlid, u8 sc5,
 		const struct ib_grh *old_grh);
 
@@ -1424,6 +1501,7 @@ static inline int valid_ib_mtu(unsigned int mtu)
 		mtu == 1024 || mtu == 2048 ||
 		mtu == 4096;
 }
+
 static inline int valid_opa_max_mtu(unsigned int mtu)
 {
 	return mtu >= 2048 &&
@@ -1445,12 +1523,13 @@ void reset_link_credits(struct hfi1_devdata *dd);
 void assign_remote_cm_au_table(struct hfi1_devdata *dd, u8 vcu);
 
 int snoop_recv_handler(struct hfi1_packet *packet);
-int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_dma_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			   u64 pbc);
-int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_pio_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			   u64 pbc);
 void snoop_inline_pio_send(struct hfi1_devdata *dd, struct pio_buf *pbuf,
 			   u64 pbc, const void *from, size_t count);
+int set_buffer_control(struct hfi1_pportdata *ppd, struct buffer_control *bc);
 
 static inline struct hfi1_devdata *dd_from_ppd(struct hfi1_pportdata *ppd)
 {
@@ -1472,6 +1551,11 @@ static inline struct hfi1_pportdata *ppd_from_ibp(struct hfi1_ibport *ibp)
 	return container_of(ibp, struct hfi1_pportdata, ibport_data);
 }
 
+static inline struct hfi1_ibdev *dev_from_rdi(struct rvt_dev_info *rdi)
+{
+	return container_of(rdi, struct hfi1_ibdev, rdi);
+}
+
 static inline struct hfi1_ibport *to_iport(struct ib_device *ibdev, u8 port)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
@@ -1515,12 +1599,10 @@ static inline struct cc_state *get_cc_state(struct hfi1_pportdata *ppd)
 #define HFI1_HAS_SDMA_TIMEOUT  0x8
 #define HFI1_HAS_SEND_DMA      0x10   /* Supports Send DMA */
 #define HFI1_FORCED_FREEZE     0x80   /* driver forced freeze mode */
-#define HFI1_DO_INIT_ASIC      0x100  /* This device will init the ASIC */
 
 /* IB dword length mask in PBC (lower 11 bits); same for all chips */
 #define HFI1_PBC_LENGTH_MASK                     ((1 << 11) - 1)
 
-
 /* ctxt_flag bit offsets */
 		/* context has been setup */
 #define HFI1_CTXT_SETUP_DONE 1
@@ -1538,14 +1620,10 @@ void hfi1_free_devdata(struct hfi1_devdata *);
 void cc_state_reclaim(struct rcu_head *rcu);
 struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra);
 
-/*
- * Set LED override, only the two LSBs have "public" meaning, but
- * any non-zero value substitutes them for the Link and LinkTrain
- * LED states.
- */
-#define HFI1_LED_PHYS 1 /* Physical (linktraining) GREEN LED */
-#define HFI1_LED_LOG 2  /* Logical (link) YELLOW LED */
-void hfi1_set_led_override(struct hfi1_pportdata *ppd, unsigned int val);
+/* LED beaconing functions */
+void hfi1_start_led_override(struct hfi1_pportdata *ppd, unsigned int timeon,
+			     unsigned int timeoff);
+void shutdown_led_override(struct hfi1_pportdata *ppd);
 
 #define HFI1_CREDIT_RETURN_RATE (100)
 
@@ -1587,12 +1665,13 @@ void hfi1_set_led_override(struct hfi1_pportdata *ppd, unsigned int val);
  */
 #define DEFAULT_RCVHDR_ENTSIZE 32
 
+bool hfi1_can_pin_pages(struct hfi1_devdata *, u32, u32);
 int hfi1_acquire_user_pages(unsigned long, size_t, bool, struct page **);
-void hfi1_release_user_pages(struct page **, size_t, bool);
+void hfi1_release_user_pages(struct mm_struct *, struct page **, size_t, bool);
 
 static inline void clear_rcvhdrtail(const struct hfi1_ctxtdata *rcd)
 {
-	*((u64 *) rcd->rcvhdrtail_kvaddr) = 0ULL;
+	*((u64 *)rcd->rcvhdrtail_kvaddr) = 0ULL;
 }
 
 static inline u32 get_rcvhdrtail(const struct hfi1_ctxtdata *rcd)
@@ -1601,7 +1680,7 @@ static inline u32 get_rcvhdrtail(const struct hfi1_ctxtdata *rcd)
 	 * volatile because it's a DMA target from the chip, routine is
 	 * inlined, and don't want register caching or reordering.
 	 */
-	return (u32) le64_to_cpu(*rcd->rcvhdrtail_kvaddr);
+	return (u32)le64_to_cpu(*rcd->rcvhdrtail_kvaddr);
 }
 
 /*
@@ -1633,12 +1712,13 @@ void restore_pci_variables(struct hfi1_devdata *dd);
 int do_pcie_gen3_transition(struct hfi1_devdata *dd);
 int parse_platform_config(struct hfi1_devdata *dd);
 int get_platform_config_field(struct hfi1_devdata *dd,
-			enum platform_config_table_type_encoding table_type,
-			int table_index, int field_index, u32 *data, u32 len);
+			      enum platform_config_table_type_encoding
+			      table_type, int table_index, int field_index,
+			      u32 *data, u32 len);
 
-dma_addr_t hfi1_map_page(struct pci_dev *, struct page *, unsigned long,
-			 size_t, int);
 const char *get_unit_name(int unit);
+const char *get_card_name(struct rvt_dev_info *rdi);
+struct pci_dev *get_pci_dev(struct rvt_dev_info *rdi);
 
 /*
  * Flush write combining store buffers (if present) and perform a write
@@ -1659,7 +1739,7 @@ int process_receive_invalid(struct hfi1_packet *packet);
 
 extern rhf_rcv_function_ptr snoop_rhf_rcv_functions[8];
 
-void update_sge(struct hfi1_sge_state *ss, u32 length);
+void update_sge(struct rvt_sge_state *ss, u32 length);
 
 /* global module parameter variables */
 extern unsigned int hfi1_max_mtu;
@@ -1667,7 +1747,7 @@ extern unsigned int hfi1_cu;
 extern unsigned int user_credit_return_threshold;
 extern int num_user_contexts;
 extern unsigned n_krcvqs;
-extern u8 krcvqs[];
+extern uint krcvqs[];
 extern int krcvqsset;
 extern uint kdeth_qp;
 extern uint loopback;
@@ -1829,13 +1909,14 @@ static inline void hfi1_reset_cpu_counters(struct hfi1_devdata *dd)
 
 	dd->z_int_counter = get_all_cpu_total(dd->int_counter);
 	dd->z_rcv_limit = get_all_cpu_total(dd->rcv_limit);
+	dd->z_send_schedule = get_all_cpu_total(dd->send_schedule);
 
 	ppd = (struct hfi1_pportdata *)(dd + 1);
 	for (i = 0; i < dd->num_pports; i++, ppd++) {
-		ppd->ibport_data.z_rc_acks =
-			get_all_cpu_total(ppd->ibport_data.rc_acks);
-		ppd->ibport_data.z_rc_qacks =
-			get_all_cpu_total(ppd->ibport_data.rc_qacks);
+		ppd->ibport_data.rvp.z_rc_acks =
+			get_all_cpu_total(ppd->ibport_data.rvp.rc_acks);
+		ppd->ibport_data.rvp.z_rc_qacks =
+			get_all_cpu_total(ppd->ibport_data.rvp.rc_qacks);
 	}
 }
 
@@ -1848,6 +1929,18 @@ static inline void setextled(struct hfi1_devdata *dd, u32 on)
 		write_csr(dd, DCC_CFG_LED_CNTRL, 0x10);
 }
 
+/* return the i2c resource given the target */
+static inline u32 i2c_target(u32 target)
+{
+	return target ? CR_I2C2 : CR_I2C1;
+}
+
+/* return the i2c chain chip resource that this HFI uses for QSFP */
+static inline u32 qsfp_resource(struct hfi1_devdata *dd)
+{
+	return i2c_target(dd->hfi1_id);
+}
+
 int hfi1_tempsense_rd(struct hfi1_devdata *dd, struct hfi1_temp *temp);
 
 #endif                          /* _HFI1_KERNEL_H */
diff --git a/drivers/staging/rdma/hfi1/init.c b/drivers/staging/rdma/hfi1/init.c
index 02df291eb172..cfcdc16b41c3 100644
--- a/drivers/staging/rdma/hfi1/init.c
+++ b/drivers/staging/rdma/hfi1/init.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -56,6 +53,7 @@
 #include <linux/module.h>
 #include <linux/printk.h>
 #include <linux/hrtimer.h>
+#include <rdma/rdma_vt.h>
 
 #include "hfi.h"
 #include "device.h"
@@ -65,6 +63,7 @@
 #include "sdma.h"
 #include "debugfs.h"
 #include "verbs.h"
+#include "aspm.h"
 
 #undef pr_fmt
 #define pr_fmt(fmt) DRIVER_NAME ": " fmt
@@ -75,6 +74,7 @@
 #define HFI1_MIN_USER_CTXT_BUFCNT 7
 
 #define HFI1_MIN_HDRQ_EGRBUF_CNT 2
+#define HFI1_MAX_HDRQ_EGRBUF_CNT 16352
 #define HFI1_MIN_EAGER_BUFFER_SIZE (4 * 1024) /* 4KB */
 #define HFI1_MAX_EAGER_BUFFER_SIZE (256 * 1024) /* 256KB */
 
@@ -87,9 +87,9 @@ module_param_named(num_user_contexts, num_user_contexts, uint, S_IRUGO);
 MODULE_PARM_DESC(
 	num_user_contexts, "Set max number of user contexts to use");
 
-u8 krcvqs[RXE_NUM_DATA_VL];
+uint krcvqs[RXE_NUM_DATA_VL];
 int krcvqsset;
-module_param_array(krcvqs, byte, &krcvqsset, S_IRUGO);
+module_param_array(krcvqs, uint, &krcvqsset, S_IRUGO);
 MODULE_PARM_DESC(krcvqs, "Array of the number of non-control kernel receive queues by VL");
 
 /* computed based on above array */
@@ -128,16 +128,12 @@ int hfi1_create_ctxts(struct hfi1_devdata *dd)
 {
 	unsigned i;
 	int ret;
-	int local_node_id = pcibus_to_node(dd->pcidev->bus);
 
 	/* Control context has to be always 0 */
 	BUILD_BUG_ON(HFI1_CTRL_CTXT != 0);
 
-	if (local_node_id < 0)
-		local_node_id = numa_node_id();
-	dd->assigned_node_id = local_node_id;
-
-	dd->rcd = kcalloc(dd->num_rcv_contexts, sizeof(*dd->rcd), GFP_KERNEL);
+	dd->rcd = kzalloc_node(dd->num_rcv_contexts * sizeof(*dd->rcd),
+			       GFP_KERNEL, dd->node);
 	if (!dd->rcd)
 		goto nomem;
 
@@ -147,10 +143,10 @@ int hfi1_create_ctxts(struct hfi1_devdata *dd)
 		struct hfi1_ctxtdata *rcd;
 
 		ppd = dd->pport + (i % dd->num_pports);
-		rcd = hfi1_create_ctxtdata(ppd, i);
+		rcd = hfi1_create_ctxtdata(ppd, i, dd->node);
 		if (!rcd) {
 			dd_dev_err(dd,
-				"Unable to allocate kernel receive context, failing\n");
+				   "Unable to allocate kernel receive context, failing\n");
 			goto nomem;
 		}
 		/*
@@ -171,7 +167,7 @@ int hfi1_create_ctxts(struct hfi1_devdata *dd)
 		rcd->sc = sc_alloc(dd, SC_ACK, rcd->rcvhdrqentsize, dd->node);
 		if (!rcd->sc) {
 			dd_dev_err(dd,
-				"Unable to allocate kernel send context, failing\n");
+				   "Unable to allocate kernel send context, failing\n");
 			dd->rcd[rcd->ctxt] = NULL;
 			hfi1_free_ctxtdata(dd, rcd);
 			goto nomem;
@@ -189,6 +185,12 @@ int hfi1_create_ctxts(struct hfi1_devdata *dd)
 		}
 	}
 
+	/*
+	 * Initialize aspm, to be done after gen3 transition and setting up
+	 * contexts and before enabling interrupts
+	 */
+	aspm_init(dd);
+
 	return 0;
 nomem:
 	ret = -ENOMEM;
@@ -201,7 +203,8 @@ bail:
 /*
  * Common code for user and kernel context setup.
  */
-struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt)
+struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt,
+					   int numa)
 {
 	struct hfi1_devdata *dd = ppd->dd;
 	struct hfi1_ctxtdata *rcd;
@@ -224,10 +227,10 @@ struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt)
 		rcd->cnt = 1;
 		rcd->ctxt = ctxt;
 		dd->rcd[ctxt] = rcd;
-		rcd->numa_id = numa_node_id();
+		rcd->numa_id = numa;
 		rcd->rcv_array_groups = dd->rcv_entries.ngroups;
 
-		spin_lock_init(&rcd->exp_lock);
+		mutex_init(&rcd->exp_lock);
 
 		/*
 		 * Calculate the context's RcvArray entry starting point.
@@ -260,7 +263,7 @@ struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt)
 		/* Validate and initialize Rcv Hdr Q variables */
 		if (rcvhdrcnt % HDRQ_INCREMENT) {
 			dd_dev_err(dd,
-				   "ctxt%u: header queue count %d must be divisible by %d\n",
+				   "ctxt%u: header queue count %d must be divisible by %lu\n",
 				   rcd->ctxt, rcvhdrcnt, HDRQ_INCREMENT);
 			goto bail;
 		}
@@ -379,7 +382,7 @@ void set_link_ipg(struct hfi1_pportdata *ppd)
 
 	cc_state = get_cc_state(ppd);
 
-	if (cc_state == NULL)
+	if (!cc_state)
 		/*
 		 * This should _never_ happen - rcu_read_lock() is held,
 		 * and set_link_ipg() should not be called if cc_state
@@ -431,7 +434,7 @@ static enum hrtimer_restart cca_timer_fn(struct hrtimer *t)
 
 	cc_state = get_cc_state(ppd);
 
-	if (cc_state == NULL) {
+	if (!cc_state) {
 		rcu_read_unlock();
 		return HRTIMER_NORESTART;
 	}
@@ -493,14 +496,19 @@ void hfi1_init_pportdata(struct pci_dev *pdev, struct hfi1_pportdata *ppd,
 	INIT_WORK(&ppd->link_vc_work, handle_verify_cap);
 	INIT_WORK(&ppd->link_up_work, handle_link_up);
 	INIT_WORK(&ppd->link_down_work, handle_link_down);
+	INIT_WORK(&ppd->dc_host_req_work, handle_8051_request);
 	INIT_WORK(&ppd->freeze_work, handle_freeze);
 	INIT_WORK(&ppd->link_downgrade_work, handle_link_downgrade);
 	INIT_WORK(&ppd->sma_message_work, handle_sma_message);
 	INIT_WORK(&ppd->link_bounce_work, handle_link_bounce);
+	INIT_WORK(&ppd->linkstate_active_work, receive_interrupt_work);
+	INIT_WORK(&ppd->qsfp_info.qsfp_work, qsfp_event);
+
 	mutex_init(&ppd->hls_lock);
 	spin_lock_init(&ppd->sdma_alllock);
 	spin_lock_init(&ppd->qsfp_info.qsfp_lock);
 
+	ppd->qsfp_info.ppd = ppd;
 	ppd->sm_trap_qp = 0x0;
 	ppd->sa_qp = 0x1;
 
@@ -582,8 +590,8 @@ static void enable_chip(struct hfi1_devdata *dd)
 	 * Enable kernel ctxts' receive and receive interrupt.
 	 * Other ctxts done as user opens and initializes them.
 	 */
-	rcvmask = HFI1_RCVCTRL_CTXT_ENB | HFI1_RCVCTRL_INTRAVAIL_ENB;
 	for (i = 0; i < dd->first_user_ctxt; ++i) {
+		rcvmask = HFI1_RCVCTRL_CTXT_ENB | HFI1_RCVCTRL_INTRAVAIL_ENB;
 		rcvmask |= HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, DMA_RTAIL) ?
 			HFI1_RCVCTRL_TAILUPD_ENB : HFI1_RCVCTRL_TAILUPD_DIS;
 		if (!HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, MULTI_PKT_EGR))
@@ -729,7 +737,7 @@ int hfi1_init(struct hfi1_devdata *dd, int reinit)
 			lastfail = hfi1_setup_eagerbufs(rcd);
 		if (lastfail)
 			dd_dev_err(dd,
-				"failed to allocate kernel ctxt's rcvhdrq and/or egr bufs\n");
+				   "failed to allocate kernel ctxt's rcvhdrq and/or egr bufs\n");
 	}
 	if (lastfail)
 		ret = lastfail;
@@ -762,7 +770,6 @@ int hfi1_init(struct hfi1_devdata *dd, int reinit)
 	/* enable chip even if we have an error, so we can debug cause */
 	enable_chip(dd);
 
-	ret = hfi1_cq_init(dd);
 done:
 	/*
 	 * Set status even if port serdes is not initialized
@@ -779,20 +786,15 @@ done:
 		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
 			ppd = dd->pport + pidx;
 
-			/* initialize the qsfp if it exists
-			 * Requires interrupts to be enabled so we are notified
-			 * when the QSFP completes reset, and has
-			 * to be done before bringing up the SERDES
+			/*
+			 * start the serdes - must be after interrupts are
+			 * enabled so we are notified when the link goes up
 			 */
-			init_qsfp(ppd);
-
-			/* start the serdes - must be after interrupts are
-			   enabled so we are notified when the link goes up */
 			lastfail = bringup_serdes(ppd);
 			if (lastfail)
 				dd_dev_info(dd,
-					"Failed to bring up port %u\n",
-					ppd->port);
+					    "Failed to bring up port %u\n",
+					    ppd->port);
 
 			/*
 			 * Set status even if port serdes is not initialized
@@ -904,6 +906,8 @@ static void shutdown_device(struct hfi1_devdata *dd)
 		/* disable the send device */
 		pio_send_control(dd, PSC_GLOBAL_DISABLE);
 
+		shutdown_led_override(ppd);
+
 		/*
 		 * Clear SerdesEnable.
 		 * We can't count on interrupts since we are stopping.
@@ -961,17 +965,33 @@ void hfi1_free_ctxtdata(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
 	kfree(rcd->egrbufs.buffers);
 
 	sc_free(rcd->sc);
-	vfree(rcd->physshadow);
-	vfree(rcd->tid_pg_list);
 	vfree(rcd->user_event_mask);
 	vfree(rcd->subctxt_uregbase);
 	vfree(rcd->subctxt_rcvegrbuf);
 	vfree(rcd->subctxt_rcvhdr_base);
-	kfree(rcd->tidusemap);
 	kfree(rcd->opstats);
 	kfree(rcd);
 }
 
+/*
+ * Release our hold on the shared asic data.  If we are the last one,
+ * free the structure.  Must be holding hfi1_devs_lock.
+ */
+static void release_asic_data(struct hfi1_devdata *dd)
+{
+	int other;
+
+	if (!dd->asic_data)
+		return;
+	dd->asic_data->dds[dd->hfi1_id] = NULL;
+	other = dd->hfi1_id ? 0 : 1;
+	if (!dd->asic_data->dds[other]) {
+		/* we are the last holder, free it */
+		kfree(dd->asic_data);
+	}
+	dd->asic_data = NULL;
+}
+
 void hfi1_free_devdata(struct hfi1_devdata *dd)
 {
 	unsigned long flags;
@@ -979,12 +999,15 @@ void hfi1_free_devdata(struct hfi1_devdata *dd)
 	spin_lock_irqsave(&hfi1_devs_lock, flags);
 	idr_remove(&hfi1_unit_table, dd->unit);
 	list_del(&dd->list);
+	release_asic_data(dd);
 	spin_unlock_irqrestore(&hfi1_devs_lock, flags);
-	hfi1_dbg_ibdev_exit(&dd->verbs_dev);
+	free_platform_config(dd);
 	rcu_barrier(); /* wait for rcu callbacks to complete */
 	free_percpu(dd->int_counter);
 	free_percpu(dd->rcv_limit);
-	ib_dealloc_device(&dd->verbs_dev.ibdev);
+	hfi1_dev_affinity_free(dd);
+	free_percpu(dd->send_schedule);
+	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
 }
 
 /*
@@ -999,19 +1022,19 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 {
 	unsigned long flags;
 	struct hfi1_devdata *dd;
-	int ret;
+	int ret, nports;
 
-	dd = (struct hfi1_devdata *)ib_alloc_device(sizeof(*dd) + extra);
+	/* extra is * number of ports */
+	nports = extra / sizeof(struct hfi1_pportdata);
+
+	dd = (struct hfi1_devdata *)rvt_alloc_device(sizeof(*dd) + extra,
+						     nports);
 	if (!dd)
 		return ERR_PTR(-ENOMEM);
-	/* extra is * number of ports */
-	dd->num_pports = extra / sizeof(struct hfi1_pportdata);
+	dd->num_pports = nports;
 	dd->pport = (struct hfi1_pportdata *)(dd + 1);
 
 	INIT_LIST_HEAD(&dd->list);
-	dd->node = dev_to_node(&pdev->dev);
-	if (dd->node < 0)
-		dd->node = 0;
 	idr_preload(GFP_KERNEL);
 	spin_lock_irqsave(&hfi1_devs_lock, flags);
 
@@ -1041,9 +1064,9 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 	spin_lock_init(&dd->sc_init_lock);
 	spin_lock_init(&dd->dc8051_lock);
 	spin_lock_init(&dd->dc8051_memlock);
-	mutex_init(&dd->qsfp_i2c_mutex);
 	seqlock_init(&dd->sc2vl_lock);
 	spin_lock_init(&dd->sde_map_lock);
+	spin_lock_init(&dd->pio_map_lock);
 	init_waitqueue_head(&dd->event_queue);
 
 	dd->int_counter = alloc_percpu(u64);
@@ -1062,6 +1085,14 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 		goto bail;
 	}
 
+	dd->send_schedule = alloc_percpu(u64);
+	if (!dd->send_schedule) {
+		ret = -ENOMEM;
+		hfi1_early_err(&pdev->dev,
+			       "Could not allocate per-cpu int_counter\n");
+		goto bail;
+	}
+
 	if (!hfi1_cpulist_count) {
 		u32 count = num_online_cpus();
 
@@ -1074,13 +1105,12 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 			&pdev->dev,
 			"Could not alloc cpulist info, cpu affinity might be wrong\n");
 	}
-	hfi1_dbg_ibdev_init(&dd->verbs_dev);
 	return dd;
 
 bail:
 	if (!list_empty(&dd->list))
 		list_del_init(&dd->list);
-	ib_dealloc_device(&dd->verbs_dev.ibdev);
+	ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
 	return ERR_PTR(ret);
 }
 
@@ -1173,8 +1203,10 @@ static int __init hfi1_mod_init(void)
 		user_credit_return_threshold = 100;
 
 	compute_krcvqs();
-	/* sanitize receive interrupt count, time must wait until after
-	   the hardware type is known */
+	/*
+	 * sanitize receive interrupt count, time must wait until after
+	 * the hardware type is known
+	 */
 	if (rcv_intr_count > RCV_HDR_HEAD_COUNTER_MASK)
 		rcv_intr_count = RCV_HDR_HEAD_COUNTER_MASK;
 	/* reject invalid combinations */
@@ -1209,6 +1241,9 @@ static int __init hfi1_mod_init(void)
 	idr_init(&hfi1_unit_table);
 
 	hfi1_dbg_init();
+	ret = hfi1_wss_init();
+	if (ret < 0)
+		goto bail_wss;
 	ret = pci_register_driver(&hfi1_pci_driver);
 	if (ret < 0) {
 		pr_err("Unable to register driver: error %d\n", -ret);
@@ -1217,6 +1252,8 @@ static int __init hfi1_mod_init(void)
 	goto bail; /* all OK */
 
 bail_dev:
+	hfi1_wss_exit();
+bail_wss:
 	hfi1_dbg_exit();
 	idr_destroy(&hfi1_unit_table);
 	dev_cleanup();
@@ -1232,6 +1269,7 @@ module_init(hfi1_mod_init);
 static void __exit hfi1_mod_cleanup(void)
 {
 	pci_unregister_driver(&hfi1_pci_driver);
+	hfi1_wss_exit();
 	hfi1_dbg_exit();
 	hfi1_cpulist_count = 0;
 	kfree(hfi1_cpulist);
@@ -1303,16 +1341,18 @@ static void cleanup_device_data(struct hfi1_devdata *dd)
 		}
 	}
 	kfree(tmp);
+	free_pio_map(dd);
 	/* must follow rcv context free - need to remove rcv's hooks */
 	for (ctxt = 0; ctxt < dd->num_send_contexts; ctxt++)
 		sc_free(dd->send_contexts[ctxt].sc);
 	dd->num_send_contexts = 0;
 	kfree(dd->send_contexts);
 	dd->send_contexts = NULL;
+	kfree(dd->hw_to_sw);
+	dd->hw_to_sw = NULL;
 	kfree(dd->boardname);
 	vfree(dd->events);
 	vfree(dd->status);
-	hfi1_cq_exit(dd);
 }
 
 /*
@@ -1346,6 +1386,13 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		ret = -EINVAL;
 		goto bail;
 	}
+	if (rcvhdrcnt > HFI1_MAX_HDRQ_EGRBUF_CNT) {
+		hfi1_early_err(&pdev->dev,
+			       "Receive header queue count cannot be greater than %u\n",
+			       HFI1_MAX_HDRQ_EGRBUF_CNT);
+		ret = -EINVAL;
+		goto bail;
+	}
 	/* use the encoding function as a sanitization check */
 	if (!encode_rcv_header_entry_size(hfi1_hdrq_entsize)) {
 		hfi1_early_err(&pdev->dev, "Invalid HdrQ Entry size %u\n",
@@ -1422,8 +1469,11 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	 * we still create devices, so diags, etc. can be used
 	 * to determine cause of problem.
 	 */
-	if (!initfail && !ret)
+	if (!initfail && !ret) {
 		dd->flags |= HFI1_INITTED;
+		/* create debufs files after init and ib register */
+		hfi1_dbg_ibdev_init(&dd->verbs_dev);
+	}
 
 	j = hfi1_device_create(dd);
 	if (j)
@@ -1464,6 +1514,8 @@ static void remove_one(struct pci_dev *pdev)
 {
 	struct hfi1_devdata *dd = pci_get_drvdata(pdev);
 
+	/* close debugfs files before ib unregister */
+	hfi1_dbg_ibdev_exit(&dd->verbs_dev);
 	/* unregister from IB core */
 	hfi1_unregister_ib_device(dd);
 
@@ -1516,18 +1568,11 @@ int hfi1_create_rcvhdrq(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
 
 		if (!rcd->rcvhdrq) {
 			dd_dev_err(dd,
-				"attempt to allocate %d bytes for ctxt %u rcvhdrq failed\n",
-				amt, rcd->ctxt);
+				   "attempt to allocate %d bytes for ctxt %u rcvhdrq failed\n",
+				   amt, rcd->ctxt);
 			goto bail;
 		}
 
-		/* Event mask is per device now and is in hfi1_devdata */
-		/*if (rcd->ctxt >= dd->first_user_ctxt) {
-			rcd->user_event_mask = vmalloc_user(PAGE_SIZE);
-			if (!rcd->user_event_mask)
-				goto bail_free_hdrq;
-				}*/
-
 		if (HFI1_CAP_KGET_MASK(rcd->flags, DMA_RTAIL)) {
 			rcd->rcvhdrtail_kvaddr = dma_zalloc_coherent(
 				&dd->pcidev->dev, PAGE_SIZE, &phys_hdrqtail,
@@ -1568,8 +1613,8 @@ int hfi1_create_rcvhdrq(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
 
 bail_free:
 	dd_dev_err(dd,
-		"attempt to allocate 1 page for ctxt %u rcvhdrqtailaddr failed\n",
-		rcd->ctxt);
+		   "attempt to allocate 1 page for ctxt %u rcvhdrqtailaddr failed\n",
+		   rcd->ctxt);
 	vfree(rcd->user_event_mask);
 	rcd->user_event_mask = NULL;
 	dma_free_coherent(&dd->pcidev->dev, amt, rcd->rcvhdrq,
@@ -1659,7 +1704,7 @@ int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
 			if (rcd->egrbufs.rcvtid_size == round_mtu ||
 			    !HFI1_CAP_KGET_MASK(rcd->flags, MULTI_PKT_EGR)) {
 				dd_dev_err(dd, "ctxt%u: Failed to allocate eager buffers\n",
-					rcd->ctxt);
+					   rcd->ctxt);
 				goto bail_rcvegrbuf_phys;
 			}
 
@@ -1694,8 +1739,9 @@ int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
 				     rcd->egrbufs.buffers[j].len)) {
 					j++;
 					offset = 0;
-				} else
+				} else {
 					offset += new_size;
+				}
 			}
 			rcd->egrbufs.rcvtid_size = new_size;
 		}
@@ -1708,7 +1754,6 @@ int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
 		  rcd->ctxt, rcd->egrbufs.alloced, rcd->egrbufs.rcvtid_size,
 		  rcd->egrbufs.size);
 
-
 	/*
 	 * Set the contexts rcv array head update threshold to the closest
 	 * power of 2 (so we can use a mask instead of modulo) below half
@@ -1742,14 +1787,14 @@ int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
 
 	for (idx = 0; idx < rcd->egrbufs.alloced; idx++) {
 		hfi1_put_tid(dd, rcd->eager_base + idx, PT_EAGER,
-			      rcd->egrbufs.rcvtids[idx].phys, order);
+			     rcd->egrbufs.rcvtids[idx].phys, order);
 		cond_resched();
 	}
 	goto bail;
 
 bail_rcvegrbuf_phys:
 	for (idx = 0; idx < rcd->egrbufs.alloced &&
-		     rcd->egrbufs.buffers[idx].addr;
+	     rcd->egrbufs.buffers[idx].addr;
 	     idx++) {
 		dma_free_coherent(&dd->pcidev->dev,
 				  rcd->egrbufs.buffers[idx].len,
diff --git a/drivers/staging/rdma/hfi1/intr.c b/drivers/staging/rdma/hfi1/intr.c
index 426582b9ab65..65348d16ab2f 100644
--- a/drivers/staging/rdma/hfi1/intr.c
+++ b/drivers/staging/rdma/hfi1/intr.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -98,7 +95,7 @@ static void signal_ib_event(struct hfi1_pportdata *ppd, enum ib_event_type ev)
 	 */
 	if (!(dd->flags & HFI1_INITTED))
 		return;
-	event.device = &dd->verbs_dev.ibdev;
+	event.device = &dd->verbs_dev.rdi.ibdev;
 	event.element.port_num = ppd->port;
 	event.event = ev;
 	ib_dispatch_event(&event);
@@ -131,28 +128,26 @@ void handle_linkup_change(struct hfi1_devdata *dd, u32 linkup)
 		 * NOTE: This uses this device's vAU, vCU, and vl15_init for
 		 * the remote values.  Both sides must be using the values.
 		 */
-		if (quick_linkup
-			    || dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
+		if (quick_linkup || dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
 			set_up_vl15(dd, dd->vau, dd->vl15_init);
 			assign_remote_cm_au_table(dd, dd->vcu);
 			ppd->neighbor_guid =
-				read_csr(dd,
-					DC_DC8051_STS_REMOTE_GUID);
+				read_csr(dd, DC_DC8051_STS_REMOTE_GUID);
 			ppd->neighbor_type =
 				read_csr(dd, DC_DC8051_STS_REMOTE_NODE_TYPE) &
 					DC_DC8051_STS_REMOTE_NODE_TYPE_VAL_MASK;
 			ppd->neighbor_port_number =
 				read_csr(dd, DC_DC8051_STS_REMOTE_PORT_NO) &
-					DC_DC8051_STS_REMOTE_PORT_NO_VAL_SMASK;
-			dd_dev_info(dd,
-				"Neighbor GUID: %llx Neighbor type %d\n",
-				ppd->neighbor_guid,
-				ppd->neighbor_type);
+					 DC_DC8051_STS_REMOTE_PORT_NO_VAL_SMASK;
+			dd_dev_info(dd, "Neighbor GUID: %llx Neighbor type %d\n",
+				    ppd->neighbor_guid,
+				    ppd->neighbor_type);
 		}
 
 		/* physical link went up */
 		ppd->linkup = 1;
-		ppd->offline_disabled_reason = OPA_LINKDOWN_REASON_NONE;
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE);
 
 		/* link widths are not available until the link is fully up */
 		get_linkup_link_widths(ppd);
@@ -165,7 +160,7 @@ void handle_linkup_change(struct hfi1_devdata *dd, u32 linkup)
 		reset_link_credits(dd);
 
 		/* freeze after a link down to guarantee a clean egress */
-		start_freeze_handling(ppd, FREEZE_SELF|FREEZE_LINK_DOWN);
+		start_freeze_handling(ppd, FREEZE_SELF | FREEZE_LINK_DOWN);
 
 		ev = IB_EVENT_PORT_ERR;
 
@@ -177,8 +172,6 @@ void handle_linkup_change(struct hfi1_devdata *dd, u32 linkup)
 		/* notify IB of the link change */
 		signal_ib_event(ppd, ev);
 	}
-
-
 }
 
 /*
diff --git a/drivers/staging/rdma/hfi1/iowait.h b/drivers/staging/rdma/hfi1/iowait.h
index e8ba5606d08d..2ec6ef38d389 100644
--- a/drivers/staging/rdma/hfi1/iowait.h
+++ b/drivers/staging/rdma/hfi1/iowait.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_IOWAIT_H
 #define _HFI1_IOWAIT_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -54,6 +51,8 @@
 #include <linux/workqueue.h>
 #include <linux/sched.h>
 
+#include "sdma_txreq.h"
+
 /*
  * typedef (*restart_t)() - restart callback
  * @work: pointer to work structure
@@ -67,9 +66,11 @@ struct sdma_engine;
  * @list: used to add/insert into QP/PQ wait lists
  * @tx_head: overflow list of sdma_txreq's
  * @sleep: no space callback
- * @wakeup: space callback
+ * @wakeup: space callback wakeup
+ * @sdma_drained: sdma count drained
  * @iowork: workqueue overhead
  * @wait_dma: wait for sdma_busy == 0
+ * @wait_pio: wait for pio_busy == 0
  * @sdma_busy: # of packets in flight
  * @count: total number of descriptors in tx_head'ed list
  * @tx_limit: limit for overflow queuing
@@ -101,9 +102,12 @@ struct iowait {
 		struct sdma_txreq *tx,
 		unsigned seq);
 	void (*wakeup)(struct iowait *wait, int reason);
+	void (*sdma_drained)(struct iowait *wait);
 	struct work_struct iowork;
 	wait_queue_head_t wait_dma;
+	wait_queue_head_t wait_pio;
 	atomic_t sdma_busy;
+	atomic_t pio_busy;
 	u32 count;
 	u32 tx_limit;
 	u32 tx_count;
@@ -117,7 +121,7 @@ struct iowait {
  * @tx_limit: limit for overflow queuing
  * @func: restart function for workqueue
  * @sleep: sleep function for no space
- * @wakeup: wakeup function for no space
+ * @resume: wakeup function for no space
  *
  * This function initializes the iowait
  * structure embedded in the QP or PQ.
@@ -133,17 +137,21 @@ static inline void iowait_init(
 		struct iowait *wait,
 		struct sdma_txreq *tx,
 		unsigned seq),
-	void (*wakeup)(struct iowait *wait, int reason))
+	void (*wakeup)(struct iowait *wait, int reason),
+	void (*sdma_drained)(struct iowait *wait))
 {
 	wait->count = 0;
 	INIT_LIST_HEAD(&wait->list);
 	INIT_LIST_HEAD(&wait->tx_head);
 	INIT_WORK(&wait->iowork, func);
 	init_waitqueue_head(&wait->wait_dma);
+	init_waitqueue_head(&wait->wait_pio);
 	atomic_set(&wait->sdma_busy, 0);
+	atomic_set(&wait->pio_busy, 0);
 	wait->tx_limit = tx_limit;
 	wait->sleep = sleep;
 	wait->wakeup = wakeup;
+	wait->sdma_drained = sdma_drained;
 }
 
 /**
@@ -174,6 +182,88 @@ static inline void iowait_sdma_drain(struct iowait *wait)
 }
 
 /**
+ * iowait_sdma_pending() - return sdma pending count
+ *
+ * @wait: iowait structure
+ *
+ */
+static inline int iowait_sdma_pending(struct iowait *wait)
+{
+	return atomic_read(&wait->sdma_busy);
+}
+
+/**
+ * iowait_sdma_inc - note sdma io pending
+ * @wait: iowait structure
+ */
+static inline void iowait_sdma_inc(struct iowait *wait)
+{
+	atomic_inc(&wait->sdma_busy);
+}
+
+/**
+ * iowait_sdma_add - add count to pending
+ * @wait: iowait structure
+ */
+static inline void iowait_sdma_add(struct iowait *wait, int count)
+{
+	atomic_add(count, &wait->sdma_busy);
+}
+
+/**
+ * iowait_sdma_dec - note sdma complete
+ * @wait: iowait structure
+ */
+static inline int iowait_sdma_dec(struct iowait *wait)
+{
+	return atomic_dec_and_test(&wait->sdma_busy);
+}
+
+/**
+ * iowait_pio_drain() - wait for pios to drain
+ *
+ * @wait: iowait structure
+ *
+ * This will delay until the iowait pios have
+ * completed.
+ */
+static inline void iowait_pio_drain(struct iowait *wait)
+{
+	wait_event_timeout(wait->wait_pio,
+			   !atomic_read(&wait->pio_busy),
+			   HZ);
+}
+
+/**
+ * iowait_pio_pending() - return pio pending count
+ *
+ * @wait: iowait structure
+ *
+ */
+static inline int iowait_pio_pending(struct iowait *wait)
+{
+	return atomic_read(&wait->pio_busy);
+}
+
+/**
+ * iowait_pio_inc - note pio pending
+ * @wait: iowait structure
+ */
+static inline void iowait_pio_inc(struct iowait *wait)
+{
+	atomic_inc(&wait->pio_busy);
+}
+
+/**
+ * iowait_sdma_dec - note pio complete
+ * @wait: iowait structure
+ */
+static inline int iowait_pio_dec(struct iowait *wait)
+{
+	return atomic_dec_and_test(&wait->pio_busy);
+}
+
+/**
  * iowait_drain_wakeup() - trigger iowait_drain() waiter
  *
  * @wait: iowait structure
@@ -183,6 +273,28 @@ static inline void iowait_sdma_drain(struct iowait *wait)
 static inline void iowait_drain_wakeup(struct iowait *wait)
 {
 	wake_up(&wait->wait_dma);
+	wake_up(&wait->wait_pio);
+	if (wait->sdma_drained)
+		wait->sdma_drained(wait);
+}
+
+/**
+ * iowait_get_txhead() - get packet off of iowait list
+ *
+ * @wait wait struture
+ */
+static inline struct sdma_txreq *iowait_get_txhead(struct iowait *wait)
+{
+	struct sdma_txreq *tx = NULL;
+
+	if (!list_empty(&wait->tx_head)) {
+		tx = list_first_entry(
+			&wait->tx_head,
+			struct sdma_txreq,
+			list);
+		list_del_init(&tx->list);
+	}
+	return tx;
 }
 
 #endif
diff --git a/drivers/staging/rdma/hfi1/keys.c b/drivers/staging/rdma/hfi1/keys.c
deleted file mode 100644
index e34f093a6b55..000000000000
--- a/drivers/staging/rdma/hfi1/keys.c
+++ /dev/null
@@ -1,356 +0,0 @@
-/*
- *
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *  - Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  - Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *  - Neither the name of Intel Corporation nor the names of its
- *    contributors may be used to endorse or promote products derived
- *    from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- */
-
-#include "hfi.h"
-
-/**
- * hfi1_alloc_lkey - allocate an lkey
- * @mr: memory region that this lkey protects
- * @dma_region: 0->normal key, 1->restricted DMA key
- *
- * Returns 0 if successful, otherwise returns -errno.
- *
- * Increments mr reference count as required.
- *
- * Sets the lkey field mr for non-dma regions.
- *
- */
-
-int hfi1_alloc_lkey(struct hfi1_mregion *mr, int dma_region)
-{
-	unsigned long flags;
-	u32 r;
-	u32 n;
-	int ret = 0;
-	struct hfi1_ibdev *dev = to_idev(mr->pd->device);
-	struct hfi1_lkey_table *rkt = &dev->lk_table;
-
-	hfi1_get_mr(mr);
-	spin_lock_irqsave(&rkt->lock, flags);
-
-	/* special case for dma_mr lkey == 0 */
-	if (dma_region) {
-		struct hfi1_mregion *tmr;
-
-		tmr = rcu_access_pointer(dev->dma_mr);
-		if (!tmr) {
-			rcu_assign_pointer(dev->dma_mr, mr);
-			mr->lkey_published = 1;
-		} else {
-			hfi1_put_mr(mr);
-		}
-		goto success;
-	}
-
-	/* Find the next available LKEY */
-	r = rkt->next;
-	n = r;
-	for (;;) {
-		if (!rcu_access_pointer(rkt->table[r]))
-			break;
-		r = (r + 1) & (rkt->max - 1);
-		if (r == n)
-			goto bail;
-	}
-	rkt->next = (r + 1) & (rkt->max - 1);
-	/*
-	 * Make sure lkey is never zero which is reserved to indicate an
-	 * unrestricted LKEY.
-	 */
-	rkt->gen++;
-	/*
-	 * bits are capped in verbs.c to ensure enough bits for
-	 * generation number
-	 */
-	mr->lkey = (r << (32 - hfi1_lkey_table_size)) |
-		((((1 << (24 - hfi1_lkey_table_size)) - 1) & rkt->gen)
-		 << 8);
-	if (mr->lkey == 0) {
-		mr->lkey = 1 << 8;
-		rkt->gen++;
-	}
-	rcu_assign_pointer(rkt->table[r], mr);
-	mr->lkey_published = 1;
-success:
-	spin_unlock_irqrestore(&rkt->lock, flags);
-out:
-	return ret;
-bail:
-	hfi1_put_mr(mr);
-	spin_unlock_irqrestore(&rkt->lock, flags);
-	ret = -ENOMEM;
-	goto out;
-}
-
-/**
- * hfi1_free_lkey - free an lkey
- * @mr: mr to free from tables
- */
-void hfi1_free_lkey(struct hfi1_mregion *mr)
-{
-	unsigned long flags;
-	u32 lkey = mr->lkey;
-	u32 r;
-	struct hfi1_ibdev *dev = to_idev(mr->pd->device);
-	struct hfi1_lkey_table *rkt = &dev->lk_table;
-	int freed = 0;
-
-	spin_lock_irqsave(&rkt->lock, flags);
-	if (!mr->lkey_published)
-		goto out;
-	if (lkey == 0)
-		RCU_INIT_POINTER(dev->dma_mr, NULL);
-	else {
-		r = lkey >> (32 - hfi1_lkey_table_size);
-		RCU_INIT_POINTER(rkt->table[r], NULL);
-	}
-	mr->lkey_published = 0;
-	freed++;
-out:
-	spin_unlock_irqrestore(&rkt->lock, flags);
-	if (freed) {
-		synchronize_rcu();
-		hfi1_put_mr(mr);
-	}
-}
-
-/**
- * hfi1_lkey_ok - check IB SGE for validity and initialize
- * @rkt: table containing lkey to check SGE against
- * @pd: protection domain
- * @isge: outgoing internal SGE
- * @sge: SGE to check
- * @acc: access flags
- *
- * Return 1 if valid and successful, otherwise returns 0.
- *
- * increments the reference count upon success
- *
- * Check the IB SGE for validity and initialize our internal version
- * of it.
- */
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct hfi1_pd *pd,
-		 struct hfi1_sge *isge, struct ib_sge *sge, int acc)
-{
-	struct hfi1_mregion *mr;
-	unsigned n, m;
-	size_t off;
-
-	/*
-	 * We use LKEY == zero for kernel virtual addresses
-	 * (see hfi1_get_dma_mr and dma.c).
-	 */
-	rcu_read_lock();
-	if (sge->lkey == 0) {
-		struct hfi1_ibdev *dev = to_idev(pd->ibpd.device);
-
-		if (pd->user)
-			goto bail;
-		mr = rcu_dereference(dev->dma_mr);
-		if (!mr)
-			goto bail;
-		atomic_inc(&mr->refcount);
-		rcu_read_unlock();
-
-		isge->mr = mr;
-		isge->vaddr = (void *) sge->addr;
-		isge->length = sge->length;
-		isge->sge_length = sge->length;
-		isge->m = 0;
-		isge->n = 0;
-		goto ok;
-	}
-	mr = rcu_dereference(
-		rkt->table[(sge->lkey >> (32 - hfi1_lkey_table_size))]);
-	if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
-		goto bail;
-
-	off = sge->addr - mr->user_base;
-	if (unlikely(sge->addr < mr->user_base ||
-		     off + sge->length > mr->length ||
-		     (mr->access_flags & acc) != acc))
-		goto bail;
-	atomic_inc(&mr->refcount);
-	rcu_read_unlock();
-
-	off += mr->offset;
-	if (mr->page_shift) {
-		/*
-		page sizes are uniform power of 2 so no loop is necessary
-		entries_spanned_by_off is the number of times the loop below
-		would have executed.
-		*/
-		size_t entries_spanned_by_off;
-
-		entries_spanned_by_off = off >> mr->page_shift;
-		off -= (entries_spanned_by_off << mr->page_shift);
-		m = entries_spanned_by_off / HFI1_SEGSZ;
-		n = entries_spanned_by_off % HFI1_SEGSZ;
-	} else {
-		m = 0;
-		n = 0;
-		while (off >= mr->map[m]->segs[n].length) {
-			off -= mr->map[m]->segs[n].length;
-			n++;
-			if (n >= HFI1_SEGSZ) {
-				m++;
-				n = 0;
-			}
-		}
-	}
-	isge->mr = mr;
-	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
-	isge->length = mr->map[m]->segs[n].length - off;
-	isge->sge_length = sge->length;
-	isge->m = m;
-	isge->n = n;
-ok:
-	return 1;
-bail:
-	rcu_read_unlock();
-	return 0;
-}
-
-/**
- * hfi1_rkey_ok - check the IB virtual address, length, and RKEY
- * @qp: qp for validation
- * @sge: SGE state
- * @len: length of data
- * @vaddr: virtual address to place data
- * @rkey: rkey to check
- * @acc: access flags
- *
- * Return 1 if successful, otherwise 0.
- *
- * increments the reference count upon success
- */
-int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
-		 u32 len, u64 vaddr, u32 rkey, int acc)
-{
-	struct hfi1_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
-	struct hfi1_mregion *mr;
-	unsigned n, m;
-	size_t off;
-
-	/*
-	 * We use RKEY == zero for kernel virtual addresses
-	 * (see hfi1_get_dma_mr and dma.c).
-	 */
-	rcu_read_lock();
-	if (rkey == 0) {
-		struct hfi1_pd *pd = to_ipd(qp->ibqp.pd);
-		struct hfi1_ibdev *dev = to_idev(pd->ibpd.device);
-
-		if (pd->user)
-			goto bail;
-		mr = rcu_dereference(dev->dma_mr);
-		if (!mr)
-			goto bail;
-		atomic_inc(&mr->refcount);
-		rcu_read_unlock();
-
-		sge->mr = mr;
-		sge->vaddr = (void *) vaddr;
-		sge->length = len;
-		sge->sge_length = len;
-		sge->m = 0;
-		sge->n = 0;
-		goto ok;
-	}
-
-	mr = rcu_dereference(
-		rkt->table[(rkey >> (32 - hfi1_lkey_table_size))]);
-	if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
-		goto bail;
-
-	off = vaddr - mr->iova;
-	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
-		     (mr->access_flags & acc) == 0))
-		goto bail;
-	atomic_inc(&mr->refcount);
-	rcu_read_unlock();
-
-	off += mr->offset;
-	if (mr->page_shift) {
-		/*
-		page sizes are uniform power of 2 so no loop is necessary
-		entries_spanned_by_off is the number of times the loop below
-		would have executed.
-		*/
-		size_t entries_spanned_by_off;
-
-		entries_spanned_by_off = off >> mr->page_shift;
-		off -= (entries_spanned_by_off << mr->page_shift);
-		m = entries_spanned_by_off / HFI1_SEGSZ;
-		n = entries_spanned_by_off % HFI1_SEGSZ;
-	} else {
-		m = 0;
-		n = 0;
-		while (off >= mr->map[m]->segs[n].length) {
-			off -= mr->map[m]->segs[n].length;
-			n++;
-			if (n >= HFI1_SEGSZ) {
-				m++;
-				n = 0;
-			}
-		}
-	}
-	sge->mr = mr;
-	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
-	sge->length = mr->map[m]->segs[n].length - off;
-	sge->sge_length = len;
-	sge->m = m;
-	sge->n = n;
-ok:
-	return 1;
-bail:
-	rcu_read_unlock();
-	return 0;
-}
diff --git a/drivers/staging/rdma/hfi1/mad.c b/drivers/staging/rdma/hfi1/mad.c
index 77700b818e3d..d1e7f4d7cf6f 100644
--- a/drivers/staging/rdma/hfi1/mad.c
+++ b/drivers/staging/rdma/hfi1/mad.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -55,6 +52,7 @@
 #include "hfi.h"
 #include "mad.h"
 #include "trace.h"
+#include "qp.h"
 
 /* the reset value from the FM is supposed to be 0xffff, handle both */
 #define OPA_LINK_WIDTH_RESET_OLD 0x0fff
@@ -91,7 +89,7 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
 	int pkey_idx;
 	u32 qpn = ppd_from_ibp(ibp)->sm_trap_qp;
 
-	agent = ibp->send_agent;
+	agent = ibp->rvp.send_agent;
 	if (!agent)
 		return;
 
@@ -100,7 +98,8 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
 		return;
 
 	/* o14-2 */
-	if (ibp->trap_timeout && time_before(jiffies, ibp->trap_timeout))
+	if (ibp->rvp.trap_timeout && time_before(jiffies,
+						 ibp->rvp.trap_timeout))
 		return;
 
 	pkey_idx = hfi1_lookup_pkey_idx(ibp, LIM_MGMT_P_KEY);
@@ -121,42 +120,43 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
 	smp->mgmt_class = IB_MGMT_CLASS_SUBN_LID_ROUTED;
 	smp->class_version = OPA_SMI_CLASS_VERSION;
 	smp->method = IB_MGMT_METHOD_TRAP;
-	ibp->tid++;
-	smp->tid = cpu_to_be64(ibp->tid);
+	ibp->rvp.tid++;
+	smp->tid = cpu_to_be64(ibp->rvp.tid);
 	smp->attr_id = IB_SMP_ATTR_NOTICE;
 	/* o14-1: smp->mkey = 0; */
 	memcpy(smp->route.lid.data, data, len);
 
-	spin_lock_irqsave(&ibp->lock, flags);
-	if (!ibp->sm_ah) {
-		if (ibp->sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
+	spin_lock_irqsave(&ibp->rvp.lock, flags);
+	if (!ibp->rvp.sm_ah) {
+		if (ibp->rvp.sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
 			struct ib_ah *ah;
 
-			ah = hfi1_create_qp0_ah(ibp, ibp->sm_lid);
-			if (IS_ERR(ah))
+			ah = hfi1_create_qp0_ah(ibp, ibp->rvp.sm_lid);
+			if (IS_ERR(ah)) {
 				ret = PTR_ERR(ah);
-			else {
+			} else {
 				send_buf->ah = ah;
-				ibp->sm_ah = to_iah(ah);
+				ibp->rvp.sm_ah = ibah_to_rvtah(ah);
 				ret = 0;
 			}
-		} else
+		} else {
 			ret = -EINVAL;
+		}
 	} else {
-		send_buf->ah = &ibp->sm_ah->ibah;
+		send_buf->ah = &ibp->rvp.sm_ah->ibah;
 		ret = 0;
 	}
-	spin_unlock_irqrestore(&ibp->lock, flags);
+	spin_unlock_irqrestore(&ibp->rvp.lock, flags);
 
 	if (!ret)
 		ret = ib_post_send_mad(send_buf, NULL);
 	if (!ret) {
 		/* 4.096 usec. */
-		timeout = (4096 * (1UL << ibp->subnet_timeout)) / 1000;
-		ibp->trap_timeout = jiffies + usecs_to_jiffies(timeout);
+		timeout = (4096 * (1UL << ibp->rvp.subnet_timeout)) / 1000;
+		ibp->rvp.trap_timeout = jiffies + usecs_to_jiffies(timeout);
 	} else {
 		ib_free_send_mad(send_buf);
-		ibp->trap_timeout = 0;
+		ibp->rvp.trap_timeout = 0;
 	}
 }
 
@@ -174,10 +174,10 @@ void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
 	memset(&data, 0, sizeof(data));
 
 	if (trap_num == OPA_TRAP_BAD_P_KEY)
-		ibp->pkey_violations++;
+		ibp->rvp.pkey_violations++;
 	else
-		ibp->qkey_violations++;
-	ibp->n_pkt_drops++;
+		ibp->rvp.qkey_violations++;
+	ibp->rvp.n_pkt_drops++;
 
 	/* Send violation trap */
 	data.generic_type = IB_NOTICE_TYPE_SECURITY;
@@ -233,9 +233,12 @@ static void bad_mkey(struct hfi1_ibport *ibp, struct ib_mad_hdr *mad,
 /*
  * Send a Port Capability Mask Changed trap (ch. 14.3.11).
  */
-void hfi1_cap_mask_chg(struct hfi1_ibport *ibp)
+void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num)
 {
 	struct opa_mad_notice_attr data;
+	struct hfi1_ibdev *verbs_dev = dev_from_rdi(rdi);
+	struct hfi1_devdata *dd = dd_from_dev(verbs_dev);
+	struct hfi1_ibport *ibp = &dd->pport[port_num - 1].ibport_data;
 	u32 lid = ppd_from_ibp(ibp)->lid;
 
 	memset(&data, 0, sizeof(data));
@@ -245,7 +248,7 @@ void hfi1_cap_mask_chg(struct hfi1_ibport *ibp)
 	data.trap_num = OPA_TRAP_CHANGE_CAPABILITY;
 	data.issuer_lid = cpu_to_be32(lid);
 	data.ntc_144.lid = data.issuer_lid;
-	data.ntc_144.new_cap_mask = cpu_to_be32(ibp->port_cap_flags);
+	data.ntc_144.new_cap_mask = cpu_to_be32(ibp->rvp.port_cap_flags);
 
 	send_trap(ibp, &data, sizeof(data));
 }
@@ -407,37 +410,38 @@ static int check_mkey(struct hfi1_ibport *ibp, struct ib_mad_hdr *mad,
 	int ret = 0;
 
 	/* Is the mkey in the process of expiring? */
-	if (ibp->mkey_lease_timeout &&
-	    time_after_eq(jiffies, ibp->mkey_lease_timeout)) {
+	if (ibp->rvp.mkey_lease_timeout &&
+	    time_after_eq(jiffies, ibp->rvp.mkey_lease_timeout)) {
 		/* Clear timeout and mkey protection field. */
-		ibp->mkey_lease_timeout = 0;
-		ibp->mkeyprot = 0;
+		ibp->rvp.mkey_lease_timeout = 0;
+		ibp->rvp.mkeyprot = 0;
 	}
 
-	if ((mad_flags & IB_MAD_IGNORE_MKEY) ||  ibp->mkey == 0 ||
-	    ibp->mkey == mkey)
+	if ((mad_flags & IB_MAD_IGNORE_MKEY) ||  ibp->rvp.mkey == 0 ||
+	    ibp->rvp.mkey == mkey)
 		valid_mkey = 1;
 
 	/* Unset lease timeout on any valid Get/Set/TrapRepress */
-	if (valid_mkey && ibp->mkey_lease_timeout &&
+	if (valid_mkey && ibp->rvp.mkey_lease_timeout &&
 	    (mad->method == IB_MGMT_METHOD_GET ||
 	     mad->method == IB_MGMT_METHOD_SET ||
 	     mad->method == IB_MGMT_METHOD_TRAP_REPRESS))
-		ibp->mkey_lease_timeout = 0;
+		ibp->rvp.mkey_lease_timeout = 0;
 
 	if (!valid_mkey) {
 		switch (mad->method) {
 		case IB_MGMT_METHOD_GET:
 			/* Bad mkey not a violation below level 2 */
-			if (ibp->mkeyprot < 2)
+			if (ibp->rvp.mkeyprot < 2)
 				break;
 		case IB_MGMT_METHOD_SET:
 		case IB_MGMT_METHOD_TRAP_REPRESS:
-			if (ibp->mkey_violations != 0xFFFF)
-				++ibp->mkey_violations;
-			if (!ibp->mkey_lease_timeout && ibp->mkey_lease_period)
-				ibp->mkey_lease_timeout = jiffies +
-					ibp->mkey_lease_period * HZ;
+			if (ibp->rvp.mkey_violations != 0xFFFF)
+				++ibp->rvp.mkey_violations;
+			if (!ibp->rvp.mkey_lease_timeout &&
+			    ibp->rvp.mkey_lease_period)
+				ibp->rvp.mkey_lease_timeout = jiffies +
+					ibp->rvp.mkey_lease_period * HZ;
 			/* Generate a trap notice. */
 			bad_mkey(ibp, mad, mkey, dr_slid, return_path,
 				 hop_cnt);
@@ -501,16 +505,6 @@ void read_ltp_rtt(struct hfi1_devdata *dd)
 		write_lcb_cache(DC_LCB_STS_ROUND_TRIP_LTP_CNT, reg);
 }
 
-static u8 __opa_porttype(struct hfi1_pportdata *ppd)
-{
-	if (qsfp_mod_present(ppd)) {
-		if (ppd->qsfp_info.cache_valid)
-			return OPA_PORT_TYPE_STANDARD;
-		return OPA_PORT_TYPE_DISCONNECTED;
-	}
-	return OPA_PORT_TYPE_UNKNOWN;
-}
-
 static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 				   struct ib_device *ibdev, u8 port,
 				   u32 *resp_len)
@@ -522,6 +516,7 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	struct opa_port_info *pi = (struct opa_port_info *)data;
 	u8 mtu;
 	u8 credit_rate;
+	u8 is_beaconing_active;
 	u32 state;
 	u32 num_ports = OPA_AM_NPORT(am);
 	u32 start_of_sm_config = OPA_AM_START_SM_CFG(am);
@@ -538,8 +533,8 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	ppd = dd->pport + (port - 1);
 	ibp = &ppd->ibport_data;
 
-	if (ppd->vls_supported/2 > ARRAY_SIZE(pi->neigh_mtu.pvlx_to_mtu) ||
-		ppd->vls_supported > ARRAY_SIZE(dd->vld)) {
+	if (ppd->vls_supported / 2 > ARRAY_SIZE(pi->neigh_mtu.pvlx_to_mtu) ||
+	    ppd->vls_supported > ARRAY_SIZE(dd->vld)) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)smp);
 	}
@@ -548,14 +543,14 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 
 	/* Only return the mkey if the protection field allows it. */
 	if (!(smp->method == IB_MGMT_METHOD_GET &&
-	      ibp->mkey != smp->mkey &&
-	      ibp->mkeyprot == 1))
-		pi->mkey = ibp->mkey;
-
-	pi->subnet_prefix = ibp->gid_prefix;
-	pi->sm_lid = cpu_to_be32(ibp->sm_lid);
-	pi->ib_cap_mask = cpu_to_be32(ibp->port_cap_flags);
-	pi->mkey_lease_period = cpu_to_be16(ibp->mkey_lease_period);
+	      ibp->rvp.mkey != smp->mkey &&
+	      ibp->rvp.mkeyprot == 1))
+		pi->mkey = ibp->rvp.mkey;
+
+	pi->subnet_prefix = ibp->rvp.gid_prefix;
+	pi->sm_lid = cpu_to_be32(ibp->rvp.sm_lid);
+	pi->ib_cap_mask = cpu_to_be32(ibp->rvp.port_cap_flags);
+	pi->mkey_lease_period = cpu_to_be16(ibp->rvp.mkey_lease_period);
 	pi->sm_trap_qp = cpu_to_be32(ppd->sm_trap_qp);
 	pi->sa_qp = cpu_to_be32(ppd->sa_qp);
 
@@ -581,38 +576,45 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	if (start_of_sm_config && (state == IB_PORT_INIT))
 		ppd->is_sm_config_started = 1;
 
-	pi->port_phys_conf = __opa_porttype(ppd) & 0xf;
+	pi->port_phys_conf = (ppd->port_type & 0xf);
 
 #if PI_LED_ENABLE_SUP
 	pi->port_states.ledenable_offlinereason = ppd->neighbor_normal << 4;
 	pi->port_states.ledenable_offlinereason |=
 		ppd->is_sm_config_started << 5;
+	/*
+	 * This pairs with the memory barrier in hfi1_start_led_override to
+	 * ensure that we read the correct state of LED beaconing represented
+	 * by led_override_timer_active
+	 */
+	smp_rmb();
+	is_beaconing_active = !!atomic_read(&ppd->led_override_timer_active);
+	pi->port_states.ledenable_offlinereason |= is_beaconing_active << 6;
 	pi->port_states.ledenable_offlinereason |=
-		ppd->offline_disabled_reason & OPA_PI_MASK_OFFLINE_REASON;
+		ppd->offline_disabled_reason;
 #else
 	pi->port_states.offline_reason = ppd->neighbor_normal << 4;
 	pi->port_states.offline_reason |= ppd->is_sm_config_started << 5;
-	pi->port_states.offline_reason |= ppd->offline_disabled_reason &
-						OPA_PI_MASK_OFFLINE_REASON;
+	pi->port_states.offline_reason |= ppd->offline_disabled_reason;
 #endif /* PI_LED_ENABLE_SUP */
 
 	pi->port_states.portphysstate_portstate =
 		(hfi1_ibphys_portstate(ppd) << 4) | state;
 
-	pi->mkeyprotect_lmc = (ibp->mkeyprot << 6) | ppd->lmc;
+	pi->mkeyprotect_lmc = (ibp->rvp.mkeyprot << 6) | ppd->lmc;
 
 	memset(pi->neigh_mtu.pvlx_to_mtu, 0, sizeof(pi->neigh_mtu.pvlx_to_mtu));
 	for (i = 0; i < ppd->vls_supported; i++) {
 		mtu = mtu_to_enum(dd->vld[i].mtu, HFI1_DEFAULT_ACTIVE_MTU);
 		if ((i % 2) == 0)
-			pi->neigh_mtu.pvlx_to_mtu[i/2] |= (mtu << 4);
+			pi->neigh_mtu.pvlx_to_mtu[i / 2] |= (mtu << 4);
 		else
-			pi->neigh_mtu.pvlx_to_mtu[i/2] |= mtu;
+			pi->neigh_mtu.pvlx_to_mtu[i / 2] |= mtu;
 	}
 	/* don't forget VL 15 */
 	mtu = mtu_to_enum(dd->vld[15].mtu, 2048);
-	pi->neigh_mtu.pvlx_to_mtu[15/2] |= mtu;
-	pi->smsl = ibp->sm_sl & OPA_PI_MASK_SMSL;
+	pi->neigh_mtu.pvlx_to_mtu[15 / 2] |= mtu;
+	pi->smsl = ibp->rvp.sm_sl & OPA_PI_MASK_SMSL;
 	pi->operational_vls = hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_OP_VLS);
 	pi->partenforce_filterraw |=
 		(ppd->linkinit_reason & OPA_PI_MASK_LINKINIT_REASON);
@@ -620,17 +622,17 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 		pi->partenforce_filterraw |= OPA_PI_MASK_PARTITION_ENFORCE_IN;
 	if (ppd->part_enforce & HFI1_PART_ENFORCE_OUT)
 		pi->partenforce_filterraw |= OPA_PI_MASK_PARTITION_ENFORCE_OUT;
-	pi->mkey_violations = cpu_to_be16(ibp->mkey_violations);
+	pi->mkey_violations = cpu_to_be16(ibp->rvp.mkey_violations);
 	/* P_KeyViolations are counted by hardware. */
-	pi->pkey_violations = cpu_to_be16(ibp->pkey_violations);
-	pi->qkey_violations = cpu_to_be16(ibp->qkey_violations);
+	pi->pkey_violations = cpu_to_be16(ibp->rvp.pkey_violations);
+	pi->qkey_violations = cpu_to_be16(ibp->rvp.qkey_violations);
 
 	pi->vl.cap = ppd->vls_supported;
-	pi->vl.high_limit = cpu_to_be16(ibp->vl_high_limit);
+	pi->vl.high_limit = cpu_to_be16(ibp->rvp.vl_high_limit);
 	pi->vl.arb_high_cap = (u8)hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_VL_HIGH_CAP);
 	pi->vl.arb_low_cap = (u8)hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_VL_LOW_CAP);
 
-	pi->clientrereg_subnettimeout = ibp->subnet_timeout;
+	pi->clientrereg_subnettimeout = ibp->rvp.subnet_timeout;
 
 	pi->port_link_mode  = cpu_to_be16(OPA_PORT_LINK_MODE_OPA << 10 |
 					  OPA_PORT_LINK_MODE_OPA << 5 |
@@ -701,8 +703,10 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	/* read the cached value of DC_LCB_STS_ROUND_TRIP_LTP_CNT */
 	read_lcb_cache(DC_LCB_STS_ROUND_TRIP_LTP_CNT, &tmp);
 
-	/* this counter is 16 bits wide, but the replay_depth.wire
-	 * variable is only 8 bits */
+	/*
+	 * this counter is 16 bits wide, but the replay_depth.wire
+	 * variable is only 8 bits
+	 */
 	if (tmp > 0xff)
 		tmp = 0xff;
 	pi->replay_depth.wire = tmp;
@@ -749,7 +753,7 @@ static int __subn_get_opa_pkeytable(struct opa_smp *smp, u32 am, u8 *data,
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	n_blocks_avail = (u16) (npkeys/OPA_PARTITION_TABLE_BLK_SIZE) + 1;
+	n_blocks_avail = (u16)(npkeys / OPA_PARTITION_TABLE_BLK_SIZE) + 1;
 
 	size = (n_blocks_req * OPA_PARTITION_TABLE_BLK_SIZE) * sizeof(u16);
 
@@ -763,7 +767,7 @@ static int __subn_get_opa_pkeytable(struct opa_smp *smp, u32 am, u8 *data,
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	p = (__be16 *) data;
+	p = (__be16 *)data;
 	q = (u16 *)data;
 	/* get the real pkeys if we are requesting the first block */
 	if (start_block == 0) {
@@ -772,9 +776,9 @@ static int __subn_get_opa_pkeytable(struct opa_smp *smp, u32 am, u8 *data,
 			p[i] = cpu_to_be16(q[i]);
 		if (resp_len)
 			*resp_len += size;
-	} else
+	} else {
 		smp->status |= IB_SMP_INVALID_FIELD;
-
+	}
 	return reply((struct ib_mad_hdr *)smp);
 }
 
@@ -901,8 +905,8 @@ static int port_states_transition_allowed(struct hfi1_pportdata *ppd,
 	u32 logical_old = driver_logical_state(ppd);
 	int ret, logical_allowed, physical_allowed;
 
-	logical_allowed = ret =
-		logical_transition_allowed(logical_old, logical_new);
+	ret = logical_transition_allowed(logical_old, logical_new);
+	logical_allowed = ret;
 
 	if (ret == HFI_TRANSITION_DISALLOWED ||
 	    ret == HFI_TRANSITION_UNDEFINED) {
@@ -912,8 +916,8 @@ static int port_states_transition_allowed(struct hfi1_pportdata *ppd,
 		return ret;
 	}
 
-	physical_allowed = ret =
-		physical_transition_allowed(physical_old, physical_new);
+	ret = physical_transition_allowed(physical_old, physical_new);
+	physical_allowed = ret;
 
 	if (ret == HFI_TRANSITION_DISALLOWED ||
 	    ret == HFI_TRANSITION_UNDEFINED) {
@@ -928,6 +932,14 @@ static int port_states_transition_allowed(struct hfi1_pportdata *ppd,
 		return HFI_TRANSITION_IGNORED;
 
 	/*
+	 * A change request of Physical Port State from
+	 * 'Offline' to 'Polling' should be ignored.
+	 */
+	if ((physical_old == OPA_PORTPHYSSTATE_OFFLINE) &&
+	    (physical_new == IB_PORTPHYSSTATE_POLLING))
+		return HFI_TRANSITION_IGNORED;
+
+	/*
 	 * Either physical_allowed or logical_allowed is
 	 * HFI_TRANSITION_ALLOWED.
 	 */
@@ -972,16 +984,15 @@ static int set_port_states(struct hfi1_pportdata *ppd, struct opa_smp *smp,
 			break;
 		/* FALLTHROUGH */
 	case IB_PORT_DOWN:
-		if (phys_state == IB_PORTPHYSSTATE_NOP)
+		if (phys_state == IB_PORTPHYSSTATE_NOP) {
 			link_state = HLS_DN_DOWNDEF;
-		else if (phys_state == IB_PORTPHYSSTATE_POLLING) {
+		} else if (phys_state == IB_PORTPHYSSTATE_POLLING) {
 			link_state = HLS_DN_POLL;
-			set_link_down_reason(ppd,
-			     OPA_LINKDOWN_REASON_FM_BOUNCE, 0,
-			     OPA_LINKDOWN_REASON_FM_BOUNCE);
-		} else if (phys_state == IB_PORTPHYSSTATE_DISABLED)
+			set_link_down_reason(ppd, OPA_LINKDOWN_REASON_FM_BOUNCE,
+					     0, OPA_LINKDOWN_REASON_FM_BOUNCE);
+		} else if (phys_state == IB_PORTPHYSSTATE_DISABLED) {
 			link_state = HLS_DN_DISABLE;
-		else {
+		} else {
 			pr_warn("SubnSet(OPA_PortInfo) invalid physical state 0x%x\n",
 				phys_state);
 			smp->status |= IB_SMP_INVALID_FIELD;
@@ -991,11 +1002,11 @@ static int set_port_states(struct hfi1_pportdata *ppd, struct opa_smp *smp,
 		set_link_state(ppd, link_state);
 		if (link_state == HLS_DN_DISABLE &&
 		    (ppd->offline_disabled_reason >
-		     OPA_LINKDOWN_REASON_SMA_DISABLED ||
+		     HFI1_ODR_MASK(OPA_LINKDOWN_REASON_SMA_DISABLED) ||
 		     ppd->offline_disabled_reason ==
-		     OPA_LINKDOWN_REASON_NONE))
+		     HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE)))
 			ppd->offline_disabled_reason =
-			OPA_LINKDOWN_REASON_SMA_DISABLED;
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_SMA_DISABLED);
 		/*
 		 * Don't send a reply if the response would be sent
 		 * through the disabled port.
@@ -1091,13 +1102,13 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 
 	ls_old = driver_lstate(ppd);
 
-	ibp->mkey = pi->mkey;
-	ibp->gid_prefix = pi->subnet_prefix;
-	ibp->mkey_lease_period = be16_to_cpu(pi->mkey_lease_period);
+	ibp->rvp.mkey = pi->mkey;
+	ibp->rvp.gid_prefix = pi->subnet_prefix;
+	ibp->rvp.mkey_lease_period = be16_to_cpu(pi->mkey_lease_period);
 
 	/* Must be a valid unicast LID address. */
 	if ((lid == 0 && ls_old > IB_PORT_INIT) ||
-	     lid >= HFI1_MULTICAST_LID_BASE) {
+	    lid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		pr_warn("SubnSet(OPA_PortInfo) lid invalid 0x%x\n",
 			lid);
@@ -1130,23 +1141,23 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 
 	/* Must be a valid unicast LID address. */
 	if ((smlid == 0 && ls_old > IB_PORT_INIT) ||
-	     smlid >= HFI1_MULTICAST_LID_BASE) {
+	    smlid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		pr_warn("SubnSet(OPA_PortInfo) smlid invalid 0x%x\n", smlid);
-	} else if (smlid != ibp->sm_lid || msl != ibp->sm_sl) {
+	} else if (smlid != ibp->rvp.sm_lid || msl != ibp->rvp.sm_sl) {
 		pr_warn("SubnSet(OPA_PortInfo) smlid 0x%x\n", smlid);
-		spin_lock_irqsave(&ibp->lock, flags);
-		if (ibp->sm_ah) {
-			if (smlid != ibp->sm_lid)
-				ibp->sm_ah->attr.dlid = smlid;
-			if (msl != ibp->sm_sl)
-				ibp->sm_ah->attr.sl = msl;
+		spin_lock_irqsave(&ibp->rvp.lock, flags);
+		if (ibp->rvp.sm_ah) {
+			if (smlid != ibp->rvp.sm_lid)
+				ibp->rvp.sm_ah->attr.dlid = smlid;
+			if (msl != ibp->rvp.sm_sl)
+				ibp->rvp.sm_ah->attr.sl = msl;
 		}
-		spin_unlock_irqrestore(&ibp->lock, flags);
-		if (smlid != ibp->sm_lid)
-			ibp->sm_lid = smlid;
-		if (msl != ibp->sm_sl)
-			ibp->sm_sl = msl;
+		spin_unlock_irqrestore(&ibp->rvp.lock, flags);
+		if (smlid != ibp->rvp.sm_lid)
+			ibp->rvp.sm_lid = smlid;
+		if (msl != ibp->rvp.sm_sl)
+			ibp->rvp.sm_sl = msl;
 		event.event = IB_EVENT_SM_CHANGE;
 		ib_dispatch_event(&event);
 	}
@@ -1167,8 +1178,8 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	ppd->port_error_action = be32_to_cpu(pi->port_error_action);
 	lwe = be16_to_cpu(pi->link_width.enabled);
 	if (lwe) {
-		if (lwe == OPA_LINK_WIDTH_RESET
-				|| lwe == OPA_LINK_WIDTH_RESET_OLD)
+		if (lwe == OPA_LINK_WIDTH_RESET ||
+		    lwe == OPA_LINK_WIDTH_RESET_OLD)
 			set_link_width_enabled(ppd, ppd->link_width_supported);
 		else if ((lwe & ~ppd->link_width_supported) == 0)
 			set_link_width_enabled(ppd, lwe);
@@ -1177,19 +1188,21 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	}
 	lwe = be16_to_cpu(pi->link_width_downgrade.enabled);
 	/* LWD.E is always applied - 0 means "disabled" */
-	if (lwe == OPA_LINK_WIDTH_RESET
-			|| lwe == OPA_LINK_WIDTH_RESET_OLD) {
+	if (lwe == OPA_LINK_WIDTH_RESET ||
+	    lwe == OPA_LINK_WIDTH_RESET_OLD) {
 		set_link_width_downgrade_enabled(ppd,
-				ppd->link_width_downgrade_supported);
+						 ppd->
+						 link_width_downgrade_supported
+						 );
 	} else if ((lwe & ~ppd->link_width_downgrade_supported) == 0) {
 		/* only set and apply if something changed */
 		if (lwe != ppd->link_width_downgrade_enabled) {
 			set_link_width_downgrade_enabled(ppd, lwe);
 			call_link_downgrade_policy = 1;
 		}
-	} else
+	} else {
 		smp->status |= IB_SMP_INVALID_FIELD;
-
+	}
 	lse = be16_to_cpu(pi->link_speed.enabled);
 	if (lse) {
 		if (lse & be16_to_cpu(pi->link_speed.supported))
@@ -1198,22 +1211,24 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 			smp->status |= IB_SMP_INVALID_FIELD;
 	}
 
-	ibp->mkeyprot = (pi->mkeyprotect_lmc & OPA_PI_MASK_MKEY_PROT_BIT) >> 6;
-	ibp->vl_high_limit = be16_to_cpu(pi->vl.high_limit) & 0xFF;
+	ibp->rvp.mkeyprot =
+		(pi->mkeyprotect_lmc & OPA_PI_MASK_MKEY_PROT_BIT) >> 6;
+	ibp->rvp.vl_high_limit = be16_to_cpu(pi->vl.high_limit) & 0xFF;
 	(void)hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_VL_HIGH_LIMIT,
-				    ibp->vl_high_limit);
+				    ibp->rvp.vl_high_limit);
 
-	if (ppd->vls_supported/2 > ARRAY_SIZE(pi->neigh_mtu.pvlx_to_mtu) ||
-		ppd->vls_supported > ARRAY_SIZE(dd->vld)) {
+	if (ppd->vls_supported / 2 > ARRAY_SIZE(pi->neigh_mtu.pvlx_to_mtu) ||
+	    ppd->vls_supported > ARRAY_SIZE(dd->vld)) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)smp);
 	}
 	for (i = 0; i < ppd->vls_supported; i++) {
 		if ((i % 2) == 0)
-			mtu = enum_to_mtu((pi->neigh_mtu.pvlx_to_mtu[i/2] >> 4)
-					  & 0xF);
+			mtu = enum_to_mtu((pi->neigh_mtu.pvlx_to_mtu[i / 2] >>
+					   4) & 0xF);
 		else
-			mtu = enum_to_mtu(pi->neigh_mtu.pvlx_to_mtu[i/2] & 0xF);
+			mtu = enum_to_mtu(pi->neigh_mtu.pvlx_to_mtu[i / 2] &
+					  0xF);
 		if (mtu == 0xffff) {
 			pr_warn("SubnSet(OPA_PortInfo) mtu invalid %d (0x%x)\n",
 				mtu,
@@ -1223,8 +1238,8 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 		}
 		if (dd->vld[i].mtu != mtu) {
 			dd_dev_info(dd,
-				"MTU change on vl %d from %d to %d\n",
-				i, dd->vld[i].mtu, mtu);
+				    "MTU change on vl %d from %d to %d\n",
+				    i, dd->vld[i].mtu, mtu);
 			dd->vld[i].mtu = mtu;
 			call_set_mtu++;
 		}
@@ -1232,13 +1247,13 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 	/* As per OPAV1 spec: VL15 must support and be configured
 	 * for operation with a 2048 or larger MTU.
 	 */
-	mtu = enum_to_mtu(pi->neigh_mtu.pvlx_to_mtu[15/2] & 0xF);
+	mtu = enum_to_mtu(pi->neigh_mtu.pvlx_to_mtu[15 / 2] & 0xF);
 	if (mtu < 2048 || mtu == 0xffff)
 		mtu = 2048;
 	if (dd->vld[15].mtu != mtu) {
 		dd_dev_info(dd,
-			"MTU change on vl 15 from %d to %d\n",
-			dd->vld[15].mtu, mtu);
+			    "MTU change on vl 15 from %d to %d\n",
+			    dd->vld[15].mtu, mtu);
 		dd->vld[15].mtu = mtu;
 		call_set_mtu++;
 	}
@@ -1254,21 +1269,21 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
 			smp->status |= IB_SMP_INVALID_FIELD;
 		} else {
 			if (hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_OP_VLS,
-						vls) == -EINVAL)
+					    vls) == -EINVAL)
 				smp->status |= IB_SMP_INVALID_FIELD;
 		}
 	}
 
 	if (pi->mkey_violations == 0)
-		ibp->mkey_violations = 0;
+		ibp->rvp.mkey_violations = 0;
 
 	if (pi->pkey_violations == 0)
-		ibp->pkey_violations = 0;
+		ibp->rvp.pkey_violations = 0;
 
 	if (pi->qkey_violations == 0)
-		ibp->qkey_violations = 0;
+		ibp->rvp.qkey_violations = 0;
 
-	ibp->subnet_timeout =
+	ibp->rvp.subnet_timeout =
 		pi->clientrereg_subnettimeout & OPA_PI_MASK_SUBNET_TIMEOUT;
 
 	crc_enabled = be16_to_cpu(pi->port_ltp_crc_mode);
@@ -1388,7 +1403,7 @@ static int set_pkeys(struct hfi1_devdata *dd, u8 port, u16 *pkeys)
 		(void)hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_PKEYS, 0);
 
 		event.event = IB_EVENT_PKEY_CHANGE;
-		event.device = &dd->verbs_dev.ibdev;
+		event.device = &dd->verbs_dev.rdi.ibdev;
 		event.element.port_num = port;
 		ib_dispatch_event(&event);
 	}
@@ -1402,7 +1417,7 @@ static int __subn_set_opa_pkeytable(struct opa_smp *smp, u32 am, u8 *data,
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	u32 n_blocks_sent = OPA_AM_NBLK(am);
 	u32 start_block = am & 0x7ff;
-	u16 *p = (u16 *) data;
+	u16 *p = (u16 *)data;
 	__be16 *q = (__be16 *)data;
 	int i;
 	u16 n_blocks_avail;
@@ -1415,7 +1430,7 @@ static int __subn_set_opa_pkeytable(struct opa_smp *smp, u32 am, u8 *data,
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	n_blocks_avail = (u16)(npkeys/OPA_PARTITION_TABLE_BLK_SIZE) + 1;
+	n_blocks_avail = (u16)(npkeys / OPA_PARTITION_TABLE_BLK_SIZE) + 1;
 
 	if (start_block + n_blocks_sent > n_blocks_avail ||
 	    n_blocks_sent > OPA_NUM_PKEY_BLOCKS_PER_SMP) {
@@ -1514,14 +1529,22 @@ static int __subn_set_opa_sl_to_sc(struct opa_smp *smp, u32 am, u8 *data,
 	struct hfi1_ibport *ibp = to_iport(ibdev, port);
 	u8 *p = data;
 	int i;
+	u8 sc;
 
 	if (am) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	for (i = 0; i <  ARRAY_SIZE(ibp->sl_to_sc); i++)
-		ibp->sl_to_sc[i] = *p++;
+	for (i = 0; i <  ARRAY_SIZE(ibp->sl_to_sc); i++) {
+		sc = *p++;
+		if (ibp->sl_to_sc[i] != sc) {
+			ibp->sl_to_sc[i] = sc;
+
+			/* Put all stale qps into error state */
+			hfi1_error_port_qps(ibp, i);
+		}
+	}
 
 	return __subn_get_opa_sl_to_sc(smp, am, data, ibdev, port, resp_len);
 }
@@ -1574,7 +1597,7 @@ static int __subn_get_opa_sc_to_vlt(struct opa_smp *smp, u32 am, u8 *data,
 {
 	u32 n_blocks = OPA_AM_NBLK(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	void *vp = (void *) data;
+	void *vp = (void *)data;
 	size_t size = 4 * sizeof(u64);
 
 	if (n_blocks != 1) {
@@ -1597,7 +1620,7 @@ static int __subn_set_opa_sc_to_vlt(struct opa_smp *smp, u32 am, u8 *data,
 	u32 n_blocks = OPA_AM_NBLK(am);
 	int async_update = OPA_AM_ASYNC(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	void *vp = (void *) data;
+	void *vp = (void *)data;
 	struct hfi1_pportdata *ppd;
 	int lstate;
 
@@ -1609,8 +1632,10 @@ static int __subn_set_opa_sc_to_vlt(struct opa_smp *smp, u32 am, u8 *data,
 	/* IB numbers ports from 1, hw from 0 */
 	ppd = dd->pport + (port - 1);
 	lstate = driver_lstate(ppd);
-	/* it's known that async_update is 0 by this point, but include
-	 * the explicit check for clarity */
+	/*
+	 * it's known that async_update is 0 by this point, but include
+	 * the explicit check for clarity
+	 */
 	if (!async_update &&
 	    (lstate == IB_PORT_ARMED || lstate == IB_PORT_ACTIVE)) {
 		smp->status |= IB_SMP_INVALID_FIELD;
@@ -1629,7 +1654,7 @@ static int __subn_get_opa_sc_to_vlnt(struct opa_smp *smp, u32 am, u8 *data,
 	u32 n_blocks = OPA_AM_NPORT(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	struct hfi1_pportdata *ppd;
-	void *vp = (void *) data;
+	void *vp = (void *)data;
 	int size;
 
 	if (n_blocks != 1) {
@@ -1654,7 +1679,7 @@ static int __subn_set_opa_sc_to_vlnt(struct opa_smp *smp, u32 am, u8 *data,
 	u32 n_blocks = OPA_AM_NPORT(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	struct hfi1_pportdata *ppd;
-	void *vp = (void *) data;
+	void *vp = (void *)data;
 	int lstate;
 
 	if (n_blocks != 1) {
@@ -1687,7 +1712,7 @@ static int __subn_get_opa_psi(struct opa_smp *smp, u32 am, u8 *data,
 	u32 lstate;
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
-	struct opa_port_state_info *psi = (struct opa_port_state_info *) data;
+	struct opa_port_state_info *psi = (struct opa_port_state_info *)data;
 
 	if (nports != 1) {
 		smp->status |= IB_SMP_INVALID_FIELD;
@@ -1707,12 +1732,11 @@ static int __subn_get_opa_psi(struct opa_smp *smp, u32 am, u8 *data,
 	psi->port_states.ledenable_offlinereason |=
 		ppd->is_sm_config_started << 5;
 	psi->port_states.ledenable_offlinereason |=
-		ppd->offline_disabled_reason & OPA_PI_MASK_OFFLINE_REASON;
+		ppd->offline_disabled_reason;
 #else
 	psi->port_states.offline_reason = ppd->neighbor_normal << 4;
 	psi->port_states.offline_reason |= ppd->is_sm_config_started << 5;
-	psi->port_states.offline_reason |= ppd->offline_disabled_reason &
-				OPA_PI_MASK_OFFLINE_REASON;
+	psi->port_states.offline_reason |= ppd->offline_disabled_reason;
 #endif /* PI_LED_ENABLE_SUP */
 
 	psi->port_states.portphysstate_portstate =
@@ -1737,7 +1761,7 @@ static int __subn_set_opa_psi(struct opa_smp *smp, u32 am, u8 *data,
 	u8 ls_new, ps_new;
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
-	struct opa_port_state_info *psi = (struct opa_port_state_info *) data;
+	struct opa_port_state_info *psi = (struct opa_port_state_info *)data;
 	int ret, invalid = 0;
 
 	if (nports != 1) {
@@ -1782,14 +1806,16 @@ static int __subn_get_opa_cable_info(struct opa_smp *smp, u32 am, u8 *data,
 	u32 len = OPA_AM_CI_LEN(am) + 1;
 	int ret;
 
-#define __CI_PAGE_SIZE (1 << 7) /* 128 bytes */
+#define __CI_PAGE_SIZE BIT(7) /* 128 bytes */
 #define __CI_PAGE_MASK ~(__CI_PAGE_SIZE - 1)
 #define __CI_PAGE_NUM(a) ((a) & __CI_PAGE_MASK)
 
-	/* check that addr is within spec, and
-	 * addr and (addr + len - 1) are on the same "page" */
+	/*
+	 * check that addr is within spec, and
+	 * addr and (addr + len - 1) are on the same "page"
+	 */
 	if (addr >= 4096 ||
-		(__CI_PAGE_NUM(addr) != __CI_PAGE_NUM(addr + len - 1))) {
+	    (__CI_PAGE_NUM(addr) != __CI_PAGE_NUM(addr + len - 1))) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)smp);
 	}
@@ -1823,7 +1849,7 @@ static int __subn_get_opa_bct(struct opa_smp *smp, u32 am, u8 *data,
 	u32 num_ports = OPA_AM_NPORT(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	struct hfi1_pportdata *ppd;
-	struct buffer_control *p = (struct buffer_control *) data;
+	struct buffer_control *p = (struct buffer_control *)data;
 	int size;
 
 	if (num_ports != 1) {
@@ -1846,7 +1872,7 @@ static int __subn_set_opa_bct(struct opa_smp *smp, u32 am, u8 *data,
 	u32 num_ports = OPA_AM_NPORT(am);
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	struct hfi1_pportdata *ppd;
-	struct buffer_control *p = (struct buffer_control *) data;
+	struct buffer_control *p = (struct buffer_control *)data;
 
 	if (num_ports != 1) {
 		smp->status |= IB_SMP_INVALID_FIELD;
@@ -1919,13 +1945,15 @@ static int __subn_set_opa_vl_arb(struct opa_smp *smp, u32 am, u8 *data,
 
 	switch (section) {
 	case OPA_VLARB_LOW_ELEMENTS:
-		(void) fm_set_table(ppd, FM_TBL_VL_LOW_ARB, p);
+		(void)fm_set_table(ppd, FM_TBL_VL_LOW_ARB, p);
 		break;
 	case OPA_VLARB_HIGH_ELEMENTS:
-		(void) fm_set_table(ppd, FM_TBL_VL_HIGH_ARB, p);
+		(void)fm_set_table(ppd, FM_TBL_VL_HIGH_ARB, p);
 		break;
-	/* neither OPA_VLARB_PREEMPT_ELEMENTS, or OPA_VLARB_PREEMPT_MATRIX
-	 * can be changed from the default values */
+	/*
+	 * neither OPA_VLARB_PREEMPT_ELEMENTS, or OPA_VLARB_PREEMPT_MATRIX
+	 * can be changed from the default values
+	 */
 	case OPA_VLARB_PREEMPT_ELEMENTS:
 		/* FALLTHROUGH */
 	case OPA_VLARB_PREEMPT_MATRIX:
@@ -2137,8 +2165,10 @@ struct opa_port_data_counters_msg {
 };
 
 struct opa_port_error_counters64_msg {
-	/* Request contains first two fields, response contains the
-	 * whole magilla */
+	/*
+	 * Request contains first two fields, response contains the
+	 * whole magilla
+	 */
 	__be64 port_select_mask[4];
 	__be32 vl_select_mask;
 
@@ -2172,7 +2202,6 @@ struct opa_port_error_info_msg {
 	__be32 error_info_select_mask;
 	__be32 reserved1;
 	struct _port_ei {
-
 		u8 port_number;
 		u8 reserved2[7];
 
@@ -2251,7 +2280,7 @@ enum error_info_selects {
 };
 
 static int pma_get_opa_classportinfo(struct opa_pma_mad *pmp,
-				struct ib_device *ibdev, u32 *resp_len)
+				     struct ib_device *ibdev, u32 *resp_len)
 {
 	struct opa_class_port_info *p =
 		(struct opa_class_port_info *)pmp->data;
@@ -2299,9 +2328,9 @@ static void a0_portstatus(struct hfi1_pportdata *ppd,
 	}
 }
 
-
 static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				  struct ib_device *ibdev,
+				  u8 port, u32 *resp_len)
 {
 	struct opa_port_status_req *req =
 		(struct opa_port_status_req *)pmp->data;
@@ -2326,8 +2355,8 @@ static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
 		return reply((struct ib_mad_hdr *)pmp);
 	}
 
-	if (nports != 1 || (port_num && port_num != port)
-	    || num_vls > OPA_MAX_VLS || (vl_select_mask & ~VL_MASK_ALL)) {
+	if (nports != 1 || (port_num && port_num != port) ||
+	    num_vls > OPA_MAX_VLS || (vl_select_mask & ~VL_MASK_ALL)) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)pmp);
 	}
@@ -2357,7 +2386,7 @@ static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
 					 CNTR_INVALID_VL));
 	rsp->port_multicast_xmit_pkts =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_XMIT_PKTS,
-					CNTR_INVALID_VL));
+					  CNTR_INVALID_VL));
 	rsp->port_multicast_rcv_pkts =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_RCV_PKTS,
 					  CNTR_INVALID_VL));
@@ -2386,7 +2415,7 @@ static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
 	}
 	tmp = read_dev_cntr(dd, C_DC_SEQ_CRC_CNT, CNTR_INVALID_VL);
 	tmp2 = tmp + read_dev_cntr(dd, C_DC_REINIT_FROM_PEER_CNT,
-					CNTR_INVALID_VL);
+				   CNTR_INVALID_VL);
 	if (tmp2 > (u32)UINT_MAX || tmp2 < tmp) {
 		/* overflow/wrapped */
 		rsp->link_error_recovery = cpu_to_be32(~0);
@@ -2401,13 +2430,13 @@ static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
 		cpu_to_be64(read_dev_cntr(dd, C_DC_FM_CFG_ERR,
 					  CNTR_INVALID_VL));
 	rsp->link_downed = cpu_to_be32(read_port_cntr(ppd, C_SW_LINK_DOWN,
-					  CNTR_INVALID_VL));
+						      CNTR_INVALID_VL));
 
 	/* rsp->uncorrectable_errors is 8 bits wide, and it pegs at 0xff */
 	tmp = read_dev_cntr(dd, C_DC_UNC_ERR, CNTR_INVALID_VL);
 	rsp->uncorrectable_errors = tmp < 0x100 ? (tmp & 0xff) : 0xff;
 
-	vlinfo = &(rsp->vls[0]);
+	vlinfo = &rsp->vls[0];
 	vfi = 0;
 	/* The vl_select_mask has been checked above, and we know
 	 * that it contains only entries which represent valid VLs.
@@ -2423,27 +2452,27 @@ static int pma_get_opa_portstatus(struct opa_pma_mad *pmp,
 
 		rsp->vls[vfi].port_vl_rcv_pkts =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RX_PKT_VL,
-					idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_xmit_data =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_FLIT_VL,
-					idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_xmit_pkts =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_PKT_VL,
-					idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_xmit_wait =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_WAIT_VL,
-					idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_rcv_fecn =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_FCN_VL,
-					idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_rcv_becn =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_BCN_VL,
-					idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		vlinfo++;
 		vfi++;
@@ -2473,7 +2502,7 @@ static u64 get_error_counter_summary(struct ib_device *ibdev, u8 port,
 	error_counter_summary += read_port_cntr(ppd, C_SW_XMIT_CSTR_ERR,
 						CNTR_INVALID_VL);
 	error_counter_summary += read_dev_cntr(dd, C_DC_RMT_PHY_ERR,
-						CNTR_INVALID_VL);
+					       CNTR_INVALID_VL);
 	/* local link integrity must be right-shifted by the lli resolution */
 	tmp = read_dev_cntr(dd, C_DC_RX_REPLAY, CNTR_INVALID_VL);
 	tmp += read_dev_cntr(dd, C_DC_TX_REPLAY, CNTR_INVALID_VL);
@@ -2483,10 +2512,10 @@ static u64 get_error_counter_summary(struct ib_device *ibdev, u8 port,
 	tmp += read_dev_cntr(dd, C_DC_REINIT_FROM_PEER_CNT, CNTR_INVALID_VL);
 	error_counter_summary += (tmp >> res_ler);
 	error_counter_summary += read_dev_cntr(dd, C_DC_RCV_ERR,
-						CNTR_INVALID_VL);
+					       CNTR_INVALID_VL);
 	error_counter_summary += read_dev_cntr(dd, C_RCV_OVF, CNTR_INVALID_VL);
 	error_counter_summary += read_dev_cntr(dd, C_DC_FM_CFG_ERR,
-						CNTR_INVALID_VL);
+					       CNTR_INVALID_VL);
 	/* ppd->link_downed is a 32-bit value */
 	error_counter_summary += read_port_cntr(ppd, C_SW_LINK_DOWN,
 						CNTR_INVALID_VL);
@@ -2512,7 +2541,7 @@ static void a0_datacounters(struct hfi1_pportdata *ppd, struct _port_dctrs *rsp,
 						 idx_from_vl(vl));
 			if (tmp < sum_vl_xmit_wait) {
 				/* we wrapped */
-				sum_vl_xmit_wait = (u64) ~0;
+				sum_vl_xmit_wait = (u64)~0;
 				break;
 			}
 			sum_vl_xmit_wait = tmp;
@@ -2522,8 +2551,30 @@ static void a0_datacounters(struct hfi1_pportdata *ppd, struct _port_dctrs *rsp,
 	}
 }
 
+static void pma_get_opa_port_dctrs(struct ib_device *ibdev,
+				   struct _port_dctrs *rsp)
+{
+	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
+
+	rsp->port_xmit_data = cpu_to_be64(read_dev_cntr(dd, C_DC_XMIT_FLITS,
+						CNTR_INVALID_VL));
+	rsp->port_rcv_data = cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_FLITS,
+						CNTR_INVALID_VL));
+	rsp->port_xmit_pkts = cpu_to_be64(read_dev_cntr(dd, C_DC_XMIT_PKTS,
+						CNTR_INVALID_VL));
+	rsp->port_rcv_pkts = cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_PKTS,
+						CNTR_INVALID_VL));
+	rsp->port_multicast_xmit_pkts =
+		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_XMIT_PKTS,
+					  CNTR_INVALID_VL));
+	rsp->port_multicast_rcv_pkts =
+		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_RCV_PKTS,
+					  CNTR_INVALID_VL));
+}
+
 static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				    struct ib_device *ibdev,
+				    u8 port, u32 *resp_len)
 {
 	struct opa_port_data_counters_msg *req =
 		(struct opa_port_data_counters_msg *)pmp->data;
@@ -2590,39 +2641,19 @@ static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
 	 */
 	hfi1_read_link_quality(dd, &lq);
 	rsp->link_quality_indicator = cpu_to_be32((u32)lq);
+	pma_get_opa_port_dctrs(ibdev, rsp);
 
-	/* rsp->sw_port_congestion is 0 for HFIs */
-	/* rsp->port_xmit_time_cong is 0 for HFIs */
-	/* rsp->port_xmit_wasted_bw ??? */
-	/* rsp->port_xmit_wait_data ??? */
-	/* rsp->port_mark_fecn is 0 for HFIs */
-
-	rsp->port_xmit_data = cpu_to_be64(read_dev_cntr(dd, C_DC_XMIT_FLITS,
-						CNTR_INVALID_VL));
-	rsp->port_rcv_data = cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_FLITS,
-						CNTR_INVALID_VL));
-	rsp->port_xmit_pkts = cpu_to_be64(read_dev_cntr(dd, C_DC_XMIT_PKTS,
-						CNTR_INVALID_VL));
-	rsp->port_rcv_pkts = cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_PKTS,
-						CNTR_INVALID_VL));
-	rsp->port_multicast_xmit_pkts =
-		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_XMIT_PKTS,
-						CNTR_INVALID_VL));
-	rsp->port_multicast_rcv_pkts =
-		cpu_to_be64(read_dev_cntr(dd, C_DC_MC_RCV_PKTS,
-						CNTR_INVALID_VL));
 	rsp->port_xmit_wait =
 		cpu_to_be64(read_port_cntr(ppd, C_TX_WAIT, CNTR_INVALID_VL));
 	rsp->port_rcv_fecn =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_FCN, CNTR_INVALID_VL));
 	rsp->port_rcv_becn =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_BCN, CNTR_INVALID_VL));
-
 	rsp->port_error_counter_summary =
 		cpu_to_be64(get_error_counter_summary(ibdev, port,
 						      res_lli, res_ler));
 
-	vlinfo = &(rsp->vls[0]);
+	vlinfo = &rsp->vls[0];
 	vfi = 0;
 	/* The vl_select_mask has been checked above, and we know
 	 * that it contains only entries which represent valid VLs.
@@ -2630,44 +2661,45 @@ static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
 	 * any additional checks for vl.
 	 */
 	for_each_set_bit(vl, (unsigned long *)&(vl_select_mask),
-		 8 * sizeof(req->vl_select_mask)) {
+			 8 * sizeof(req->vl_select_mask)) {
 		memset(vlinfo, 0, sizeof(*vlinfo));
 
 		rsp->vls[vfi].port_vl_xmit_data =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_FLIT_VL,
-							idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_rcv_data =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RX_FLIT_VL,
-							idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_xmit_pkts =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_PKT_VL,
-							idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_rcv_pkts =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RX_PKT_VL,
-							idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_xmit_wait =
 			cpu_to_be64(read_port_cntr(ppd, C_TX_WAIT_VL,
-							idx_from_vl(vl)));
+						   idx_from_vl(vl)));
 
 		rsp->vls[vfi].port_vl_rcv_fecn =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_FCN_VL,
-							idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 		rsp->vls[vfi].port_vl_rcv_becn =
 			cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_BCN_VL,
-							idx_from_vl(vl)));
+						  idx_from_vl(vl)));
 
 		/* rsp->port_vl_xmit_time_cong is 0 for HFIs */
 		/* rsp->port_vl_xmit_wasted_bw ??? */
 		/* port_vl_xmit_wait_data - TXE (table 13-9 HFI spec) ???
-		 * does this differ from rsp->vls[vfi].port_vl_xmit_wait */
+		 * does this differ from rsp->vls[vfi].port_vl_xmit_wait
+		 */
 		/*rsp->vls[vfi].port_vl_mark_fecn =
-			cpu_to_be64(read_csr(dd, DCC_PRF_PORT_VL_MARK_FECN_CNT
-				+ offset));
-		*/
+		 *	cpu_to_be64(read_csr(dd, DCC_PRF_PORT_VL_MARK_FECN_CNT
+		 *		+ offset));
+		 */
 		vlinfo++;
 		vfi++;
 	}
@@ -2680,12 +2712,88 @@ static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
 	return reply((struct ib_mad_hdr *)pmp);
 }
 
+static int pma_get_ib_portcounters_ext(struct ib_pma_mad *pmp,
+				       struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters_ext *p = (struct ib_pma_portcounters_ext *)
+						pmp->data;
+	struct _port_dctrs rsp;
+
+	if (pmp->mad_hdr.attr_mod != 0 || p->port_select != port) {
+		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
+		goto bail;
+	}
+
+	memset(&rsp, 0, sizeof(rsp));
+	pma_get_opa_port_dctrs(ibdev, &rsp);
+
+	p->port_xmit_data = rsp.port_xmit_data;
+	p->port_rcv_data = rsp.port_rcv_data;
+	p->port_xmit_packets = rsp.port_xmit_pkts;
+	p->port_rcv_packets = rsp.port_rcv_pkts;
+	p->port_unicast_xmit_packets = 0;
+	p->port_unicast_rcv_packets =  0;
+	p->port_multicast_xmit_packets = rsp.port_multicast_xmit_pkts;
+	p->port_multicast_rcv_packets = rsp.port_multicast_rcv_pkts;
+
+bail:
+	return reply((struct ib_mad_hdr *)pmp);
+}
+
+static void pma_get_opa_port_ectrs(struct ib_device *ibdev,
+				   struct _port_ectrs *rsp, u8 port)
+{
+	u64 tmp, tmp2;
+	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
+	struct hfi1_ibport *ibp = to_iport(ibdev, port);
+	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+
+	tmp = read_dev_cntr(dd, C_DC_SEQ_CRC_CNT, CNTR_INVALID_VL);
+	tmp2 = tmp + read_dev_cntr(dd, C_DC_REINIT_FROM_PEER_CNT,
+					CNTR_INVALID_VL);
+	if (tmp2 > (u32)UINT_MAX || tmp2 < tmp) {
+		/* overflow/wrapped */
+		rsp->link_error_recovery = cpu_to_be32(~0);
+	} else {
+		rsp->link_error_recovery = cpu_to_be32(tmp2);
+	}
+
+	rsp->link_downed = cpu_to_be32(read_port_cntr(ppd, C_SW_LINK_DOWN,
+						CNTR_INVALID_VL));
+	rsp->port_rcv_errors =
+		cpu_to_be64(read_dev_cntr(dd, C_DC_RCV_ERR, CNTR_INVALID_VL));
+	rsp->port_rcv_remote_physical_errors =
+		cpu_to_be64(read_dev_cntr(dd, C_DC_RMT_PHY_ERR,
+					  CNTR_INVALID_VL));
+	rsp->port_rcv_switch_relay_errors = 0;
+	rsp->port_xmit_discards =
+		cpu_to_be64(read_port_cntr(ppd, C_SW_XMIT_DSCD,
+					   CNTR_INVALID_VL));
+	rsp->port_xmit_constraint_errors =
+		cpu_to_be64(read_port_cntr(ppd, C_SW_XMIT_CSTR_ERR,
+					   CNTR_INVALID_VL));
+	rsp->port_rcv_constraint_errors =
+		cpu_to_be64(read_port_cntr(ppd, C_SW_RCV_CSTR_ERR,
+					   CNTR_INVALID_VL));
+	tmp = read_dev_cntr(dd, C_DC_RX_REPLAY, CNTR_INVALID_VL);
+	tmp2 = tmp + read_dev_cntr(dd, C_DC_TX_REPLAY, CNTR_INVALID_VL);
+	if (tmp2 < tmp) {
+		/* overflow/wrapped */
+		rsp->local_link_integrity_errors = cpu_to_be64(~0);
+	} else {
+		rsp->local_link_integrity_errors = cpu_to_be64(tmp2);
+	}
+	rsp->excessive_buffer_overruns =
+		cpu_to_be64(read_dev_cntr(dd, C_RCV_OVF, CNTR_INVALID_VL));
+}
+
 static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				  struct ib_device *ibdev,
+				  u8 port, u32 *resp_len)
 {
 	size_t response_data_size;
 	struct _port_ectrs *rsp;
-	unsigned long port_num;
+	u8 port_num;
 	struct opa_port_error_counters64_msg *req;
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	u32 num_ports;
@@ -2695,7 +2803,7 @@ static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
 	struct hfi1_pportdata *ppd;
 	struct _vls_ectrs *vlinfo;
 	unsigned long vl;
-	u64 port_mask, tmp, tmp2;
+	u64 port_mask, tmp;
 	u32 vl_select_mask;
 	int vfi;
 
@@ -2724,9 +2832,9 @@ static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
 	 */
 	port_mask = be64_to_cpu(req->port_select_mask[3]);
 	port_num = find_first_bit((unsigned long *)&port_mask,
-					sizeof(port_mask));
+				  sizeof(port_mask));
 
-	if ((u8)port_num != port) {
+	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)pmp);
 	}
@@ -2737,46 +2845,18 @@ static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
 	ppd = ppd_from_ibp(ibp);
 
 	memset(rsp, 0, sizeof(*rsp));
-	rsp->port_number = (u8)port_num;
+	rsp->port_number = port_num;
+
+	pma_get_opa_port_ectrs(ibdev, rsp, port_num);
 
-	rsp->port_rcv_constraint_errors =
-		cpu_to_be64(read_port_cntr(ppd, C_SW_RCV_CSTR_ERR,
-					   CNTR_INVALID_VL));
-	/* port_rcv_switch_relay_errors is 0 for HFIs */
-	rsp->port_xmit_discards =
-		cpu_to_be64(read_port_cntr(ppd, C_SW_XMIT_DSCD,
-						CNTR_INVALID_VL));
 	rsp->port_rcv_remote_physical_errors =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_RMT_PHY_ERR,
-						CNTR_INVALID_VL));
-	tmp = read_dev_cntr(dd, C_DC_RX_REPLAY, CNTR_INVALID_VL);
-	tmp2 = tmp + read_dev_cntr(dd, C_DC_TX_REPLAY, CNTR_INVALID_VL);
-	if (tmp2 < tmp) {
-		/* overflow/wrapped */
-		rsp->local_link_integrity_errors = cpu_to_be64(~0);
-	} else {
-		rsp->local_link_integrity_errors = cpu_to_be64(tmp2);
-	}
-	tmp = read_dev_cntr(dd, C_DC_SEQ_CRC_CNT, CNTR_INVALID_VL);
-	tmp2 = tmp + read_dev_cntr(dd, C_DC_REINIT_FROM_PEER_CNT,
-					CNTR_INVALID_VL);
-	if (tmp2 > (u32)UINT_MAX || tmp2 < tmp) {
-		/* overflow/wrapped */
-		rsp->link_error_recovery = cpu_to_be32(~0);
-	} else {
-		rsp->link_error_recovery = cpu_to_be32(tmp2);
-	}
-	rsp->port_xmit_constraint_errors =
-		cpu_to_be64(read_port_cntr(ppd, C_SW_XMIT_CSTR_ERR,
-					   CNTR_INVALID_VL));
-	rsp->excessive_buffer_overruns =
-		cpu_to_be64(read_dev_cntr(dd, C_RCV_OVF, CNTR_INVALID_VL));
+					  CNTR_INVALID_VL));
 	rsp->fm_config_errors =
 		cpu_to_be64(read_dev_cntr(dd, C_DC_FM_CFG_ERR,
-						CNTR_INVALID_VL));
-	rsp->link_downed = cpu_to_be32(read_port_cntr(ppd, C_SW_LINK_DOWN,
-						CNTR_INVALID_VL));
+					  CNTR_INVALID_VL));
 	tmp = read_dev_cntr(dd, C_DC_UNC_ERR, CNTR_INVALID_VL);
+
 	rsp->uncorrectable_errors = tmp < 0x100 ? (tmp & 0xff) : 0xff;
 
 	vlinfo = &rsp->vls[0];
@@ -2796,8 +2876,94 @@ static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
 	return reply((struct ib_mad_hdr *)pmp);
 }
 
+static int pma_get_ib_portcounters(struct ib_pma_mad *pmp,
+				   struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
+		pmp->data;
+	struct _port_ectrs rsp;
+	u64 temp_link_overrun_errors;
+	u64 temp_64;
+	u32 temp_32;
+
+	memset(&rsp, 0, sizeof(rsp));
+	pma_get_opa_port_ectrs(ibdev, &rsp, port);
+
+	if (pmp->mad_hdr.attr_mod != 0 || p->port_select != port) {
+		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
+		goto bail;
+	}
+
+	p->symbol_error_counter = 0; /* N/A for OPA */
+
+	temp_32 = be32_to_cpu(rsp.link_error_recovery);
+	if (temp_32 > 0xFFUL)
+		p->link_error_recovery_counter = 0xFF;
+	else
+		p->link_error_recovery_counter = (u8)temp_32;
+
+	temp_32 = be32_to_cpu(rsp.link_downed);
+	if (temp_32 > 0xFFUL)
+		p->link_downed_counter = 0xFF;
+	else
+		p->link_downed_counter = (u8)temp_32;
+
+	temp_64 = be64_to_cpu(rsp.port_rcv_errors);
+	if (temp_64 > 0xFFFFUL)
+		p->port_rcv_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_errors = cpu_to_be16((u16)temp_64);
+
+	temp_64 = be64_to_cpu(rsp.port_rcv_remote_physical_errors);
+	if (temp_64 > 0xFFFFUL)
+		p->port_rcv_remphys_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_remphys_errors = cpu_to_be16((u16)temp_64);
+
+	temp_64 = be64_to_cpu(rsp.port_rcv_switch_relay_errors);
+	p->port_rcv_switch_relay_errors = cpu_to_be16((u16)temp_64);
+
+	temp_64 = be64_to_cpu(rsp.port_xmit_discards);
+	if (temp_64 > 0xFFFFUL)
+		p->port_xmit_discards = cpu_to_be16(0xFFFF);
+	else
+		p->port_xmit_discards = cpu_to_be16((u16)temp_64);
+
+	temp_64 = be64_to_cpu(rsp.port_xmit_constraint_errors);
+	if (temp_64 > 0xFFUL)
+		p->port_xmit_constraint_errors = 0xFF;
+	else
+		p->port_xmit_constraint_errors = (u8)temp_64;
+
+	temp_64 = be64_to_cpu(rsp.port_rcv_constraint_errors);
+	if (temp_64 > 0xFFUL)
+		p->port_rcv_constraint_errors = 0xFFUL;
+	else
+		p->port_rcv_constraint_errors = (u8)temp_64;
+
+	/* LocalLink: 7:4, BufferOverrun: 3:0 */
+	temp_64 = be64_to_cpu(rsp.local_link_integrity_errors);
+	if (temp_64 > 0xFUL)
+		temp_64 = 0xFUL;
+
+	temp_link_overrun_errors = temp_64 << 4;
+
+	temp_64 = be64_to_cpu(rsp.excessive_buffer_overruns);
+	if (temp_64 > 0xFUL)
+		temp_64 = 0xFUL;
+	temp_link_overrun_errors |= temp_64;
+
+	p->link_overrun_errors = (u8)temp_link_overrun_errors;
+
+	p->vl15_dropped = 0; /* N/A for OPA */
+
+bail:
+	return reply((struct ib_mad_hdr *)pmp);
+}
+
 static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				 struct ib_device *ibdev,
+				 u8 port, u32 *resp_len)
 {
 	size_t response_data_size;
 	struct _port_ei *rsp;
@@ -2805,7 +2971,7 @@ static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	u64 port_mask;
 	u32 num_ports;
-	unsigned long port_num;
+	u8 port_num;
 	u8 num_pslm;
 	u64 reg;
 
@@ -2838,7 +3004,7 @@ static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
 	port_num = find_first_bit((unsigned long *)&port_mask,
 				  sizeof(port_mask));
 
-	if ((u8)port_num != port) {
+	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)pmp);
 	}
@@ -2847,15 +3013,17 @@ static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
 	rsp->port_rcv_ei.status_and_code =
 		dd->err_info_rcvport.status_and_code;
 	memcpy(&rsp->port_rcv_ei.ei.ei1to12.packet_flit1,
-		&dd->err_info_rcvport.packet_flit1, sizeof(u64));
+	       &dd->err_info_rcvport.packet_flit1, sizeof(u64));
 	memcpy(&rsp->port_rcv_ei.ei.ei1to12.packet_flit2,
-		&dd->err_info_rcvport.packet_flit2, sizeof(u64));
+	       &dd->err_info_rcvport.packet_flit2, sizeof(u64));
 
 	/* ExcessiverBufferOverrunInfo */
 	reg = read_csr(dd, RCV_ERR_INFO);
 	if (reg & RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK) {
-		/* if the RcvExcessBufferOverrun bit is set, save SC of
-		 * first pkt that encountered an excess buffer overrun */
+		/*
+		 * if the RcvExcessBufferOverrun bit is set, save SC of
+		 * first pkt that encountered an excess buffer overrun
+		 */
 		u8 tmp = (u8)reg;
 
 		tmp &=  RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SC_SMASK;
@@ -2892,7 +3060,8 @@ static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
 }
 
 static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				  struct ib_device *ibdev,
+				  u8 port, u32 *resp_len)
 {
 	struct opa_clear_port_status *req =
 		(struct opa_clear_port_status *)pmp->data;
@@ -2951,8 +3120,9 @@ static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
 		write_dev_cntr(dd, C_DC_RCV_BBL, CNTR_INVALID_VL, 0);
 
 	/* Only applicable for switch */
-	/*if (counter_select & CS_PORT_MARK_FECN)
-		write_csr(dd, DCC_PRF_PORT_MARK_FECN_CNT, 0);*/
+	/* if (counter_select & CS_PORT_MARK_FECN)
+	 *	write_csr(dd, DCC_PRF_PORT_MARK_FECN_CNT, 0);
+	 */
 
 	if (counter_select & CS_PORT_RCV_CONSTRAINT_ERRORS)
 		write_port_cntr(ppd, C_SW_RCV_CSTR_ERR, CNTR_INVALID_VL, 0);
@@ -2975,7 +3145,7 @@ static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
 	if (counter_select & CS_LINK_ERROR_RECOVERY) {
 		write_dev_cntr(dd, C_DC_SEQ_CRC_CNT, CNTR_INVALID_VL, 0);
 		write_dev_cntr(dd, C_DC_REINIT_FROM_PEER_CNT,
-						CNTR_INVALID_VL, 0);
+			       CNTR_INVALID_VL, 0);
 	}
 
 	if (counter_select & CS_PORT_RCV_ERRORS)
@@ -2997,7 +3167,6 @@ static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
 
 	for_each_set_bit(vl, (unsigned long *)&(vl_select_mask),
 			 8 * sizeof(vl_select_mask)) {
-
 		if (counter_select & CS_PORT_XMIT_DATA)
 			write_port_cntr(ppd, C_TX_FLIT_VL, idx_from_vl(vl), 0);
 
@@ -3026,9 +3195,9 @@ static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
 		if (counter_select & CS_PORT_RCV_BUBBLE)
 			write_dev_cntr(dd, C_DC_RCV_BBL_VL, idx_from_vl(vl), 0);
 
-		/*if (counter_select & CS_PORT_MARK_FECN)
-		     write_csr(dd, DCC_PRF_PORT_VL_MARK_FECN_CNT + offset, 0);
-		*/
+		/* if (counter_select & CS_PORT_MARK_FECN)
+		 *     write_csr(dd, DCC_PRF_PORT_VL_MARK_FECN_CNT + offset, 0);
+		 */
 		/* port_vl_xmit_discards ??? */
 	}
 
@@ -3039,14 +3208,15 @@ static int pma_set_opa_portstatus(struct opa_pma_mad *pmp,
 }
 
 static int pma_set_opa_errorinfo(struct opa_pma_mad *pmp,
-			struct ib_device *ibdev, u8 port, u32 *resp_len)
+				 struct ib_device *ibdev,
+				 u8 port, u32 *resp_len)
 {
 	struct _port_ei *rsp;
 	struct opa_port_error_info_msg *req;
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
 	u64 port_mask;
 	u32 num_ports;
-	unsigned long port_num;
+	u8 port_num;
 	u8 num_pslm;
 	u32 error_info_select;
 
@@ -3071,7 +3241,7 @@ static int pma_set_opa_errorinfo(struct opa_pma_mad *pmp,
 	port_num = find_first_bit((unsigned long *)&port_mask,
 				  sizeof(port_mask));
 
-	if ((u8)port_num != port) {
+	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)pmp);
 	}
@@ -3085,8 +3255,10 @@ static int pma_set_opa_errorinfo(struct opa_pma_mad *pmp,
 
 	/* ExcessiverBufferOverrunInfo */
 	if (error_info_select & ES_EXCESSIVE_BUFFER_OVERRUN_INFO)
-		/* status bit is essentially kept in the h/w - bit 5 of
-		 * RCV_ERR_INFO */
+		/*
+		 * status bit is essentially kept in the h/w - bit 5 of
+		 * RCV_ERR_INFO
+		 */
 		write_csr(dd, RCV_ERR_INFO,
 			  RCV_ERR_INFO_RCV_EXCESS_BUFFER_OVERRUN_SMASK);
 
@@ -3138,13 +3310,12 @@ static int __subn_get_opa_cong_info(struct opa_smp *smp, u32 am, u8 *data,
 }
 
 static int __subn_get_opa_cong_setting(struct opa_smp *smp, u32 am,
-					     u8 *data,
-					     struct ib_device *ibdev,
-					     u8 port, u32 *resp_len)
+				       u8 *data, struct ib_device *ibdev,
+				       u8 port, u32 *resp_len)
 {
 	int i;
 	struct opa_congestion_setting_attr *p =
-		(struct opa_congestion_setting_attr *) data;
+		(struct opa_congestion_setting_attr *)data;
 	struct hfi1_ibport *ibp = to_iport(ibdev, port);
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	struct opa_congestion_setting_entry_shadow *entries;
@@ -3154,7 +3325,7 @@ static int __subn_get_opa_cong_setting(struct opa_smp *smp, u32 am,
 
 	cc_state = get_cc_state(ppd);
 
-	if (cc_state == NULL) {
+	if (!cc_state) {
 		rcu_read_unlock();
 		return reply((struct ib_mad_hdr *)smp);
 	}
@@ -3183,7 +3354,7 @@ static int __subn_set_opa_cong_setting(struct opa_smp *smp, u32 am, u8 *data,
 				       u32 *resp_len)
 {
 	struct opa_congestion_setting_attr *p =
-		(struct opa_congestion_setting_attr *) data;
+		(struct opa_congestion_setting_attr *)data;
 	struct hfi1_ibport *ibp = to_iport(ibdev, port);
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	struct opa_congestion_setting_entry_shadow *entries;
@@ -3245,7 +3416,7 @@ static int __subn_get_opa_hfi1_cong_log(struct opa_smp *smp, u32 am,
 			continue;
 		memcpy(cong_log->events[i].local_qp_cn_entry, &cce->lqpn, 3);
 		memcpy(cong_log->events[i].remote_qp_number_cn_entry,
-			&cce->rqpn, 3);
+		       &cce->rqpn, 3);
 		cong_log->events[i].sl_svc_type_cn_entry =
 			((cce->sl & 0x1f) << 3) | (cce->svc_type & 0x7);
 		cong_log->events[i].remote_lid_cn_entry =
@@ -3275,7 +3446,7 @@ static int __subn_get_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 				   u32 *resp_len)
 {
 	struct ib_cc_table_attr *cc_table_attr =
-		(struct ib_cc_table_attr *) data;
+		(struct ib_cc_table_attr *)data;
 	struct hfi1_ibport *ibp = to_iport(ibdev, port);
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	u32 start_block = OPA_AM_START_BLK(am);
@@ -3296,7 +3467,7 @@ static int __subn_get_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 
 	cc_state = get_cc_state(ppd);
 
-	if (cc_state == NULL) {
+	if (!cc_state) {
 		rcu_read_unlock();
 		return reply((struct ib_mad_hdr *)smp);
 	}
@@ -3316,7 +3487,7 @@ static int __subn_get_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 	rcu_read_unlock();
 
 	if (resp_len)
-		*resp_len += sizeof(u16)*(IB_CCT_ENTRIES * n_blocks + 1);
+		*resp_len += sizeof(u16) * (IB_CCT_ENTRIES * n_blocks + 1);
 
 	return reply((struct ib_mad_hdr *)smp);
 }
@@ -3332,7 +3503,7 @@ static int __subn_set_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 				   struct ib_device *ibdev, u8 port,
 				   u32 *resp_len)
 {
-	struct ib_cc_table_attr *p = (struct ib_cc_table_attr *) data;
+	struct ib_cc_table_attr *p = (struct ib_cc_table_attr *)data;
 	struct hfi1_ibport *ibp = to_iport(ibdev, port);
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	u32 start_block = OPA_AM_START_BLK(am);
@@ -3362,14 +3533,14 @@ static int __subn_set_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 	}
 
 	new_cc_state = kzalloc(sizeof(*new_cc_state), GFP_KERNEL);
-	if (new_cc_state == NULL)
+	if (!new_cc_state)
 		goto getit;
 
 	spin_lock(&ppd->cc_state_lock);
 
 	old_cc_state = get_cc_state(ppd);
 
-	if (old_cc_state == NULL) {
+	if (!old_cc_state) {
 		spin_unlock(&ppd->cc_state_lock);
 		kfree(new_cc_state);
 		return reply((struct ib_mad_hdr *)smp);
@@ -3409,26 +3580,31 @@ struct opa_led_info {
 };
 
 #define OPA_LED_SHIFT	31
-#define OPA_LED_MASK	(1 << OPA_LED_SHIFT)
+#define OPA_LED_MASK	BIT(OPA_LED_SHIFT)
 
 static int __subn_get_opa_led_info(struct opa_smp *smp, u32 am, u8 *data,
 				   struct ib_device *ibdev, u8 port,
 				   u32 *resp_len)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	struct opa_led_info *p = (struct opa_led_info *) data;
+	struct hfi1_pportdata *ppd = dd->pport;
+	struct opa_led_info *p = (struct opa_led_info *)data;
 	u32 nport = OPA_AM_NPORT(am);
-	u64 reg;
+	u32 is_beaconing_active;
 
 	if (nport != 1) {
 		smp->status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	reg = read_csr(dd, DCC_CFG_LED_CNTRL);
-	if ((reg & DCC_CFG_LED_CNTRL_LED_CNTRL_SMASK) &&
-		((reg & DCC_CFG_LED_CNTRL_LED_SW_BLINK_RATE_SMASK) == 0xf))
-			p->rsvd_led_mask = cpu_to_be32(OPA_LED_MASK);
+	/*
+	 * This pairs with the memory barrier in hfi1_start_led_override to
+	 * ensure that we read the correct state of LED beaconing represented
+	 * by led_override_timer_active
+	 */
+	smp_rmb();
+	is_beaconing_active = !!atomic_read(&ppd->led_override_timer_active);
+	p->rsvd_led_mask = cpu_to_be32(is_beaconing_active << OPA_LED_SHIFT);
 
 	if (resp_len)
 		*resp_len += sizeof(struct opa_led_info);
@@ -3441,7 +3617,7 @@ static int __subn_set_opa_led_info(struct opa_smp *smp, u32 am, u8 *data,
 				   u32 *resp_len)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	struct opa_led_info *p = (struct opa_led_info *) data;
+	struct opa_led_info *p = (struct opa_led_info *)data;
 	u32 nport = OPA_AM_NPORT(am);
 	int on = !!(be32_to_cpu(p->rsvd_led_mask) & OPA_LED_MASK);
 
@@ -3450,7 +3626,10 @@ static int __subn_set_opa_led_info(struct opa_smp *smp, u32 am, u8 *data,
 		return reply((struct ib_mad_hdr *)smp);
 	}
 
-	setextled(dd, on);
+	if (on)
+		hfi1_start_led_override(dd->pport, 2000, 1500);
+	else
+		shutdown_led_override(dd->pport);
 
 	return __subn_get_opa_led_info(smp, am, data, ibdev, port, resp_len);
 }
@@ -3493,7 +3672,7 @@ static int subn_get_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
 		break;
 	case OPA_ATTRIB_ID_SC_TO_VLNT_MAP:
 		ret = __subn_get_opa_sc_to_vlnt(smp, am, data, ibdev, port,
-					       resp_len);
+						resp_len);
 		break;
 	case OPA_ATTRIB_ID_PORT_STATE_INFO:
 		ret = __subn_get_opa_psi(smp, am, data, ibdev, port,
@@ -3532,9 +3711,9 @@ static int subn_get_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
 					      resp_len);
 		break;
 	case IB_SMP_ATTR_SM_INFO:
-		if (ibp->port_cap_flags & IB_PORT_SM_DISABLED)
+		if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED)
 			return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
-		if (ibp->port_cap_flags & IB_PORT_SM)
+		if (ibp->rvp.port_cap_flags & IB_PORT_SM)
 			return IB_MAD_RESULT_SUCCESS;
 		/* FALLTHROUGH */
 	default:
@@ -3575,7 +3754,7 @@ static int subn_set_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
 		break;
 	case OPA_ATTRIB_ID_SC_TO_VLNT_MAP:
 		ret = __subn_set_opa_sc_to_vlnt(smp, am, data, ibdev, port,
-					       resp_len);
+						resp_len);
 		break;
 	case OPA_ATTRIB_ID_PORT_STATE_INFO:
 		ret = __subn_set_opa_psi(smp, am, data, ibdev, port,
@@ -3602,9 +3781,9 @@ static int subn_set_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
 					      resp_len);
 		break;
 	case IB_SMP_ATTR_SM_INFO:
-		if (ibp->port_cap_flags & IB_PORT_SM_DISABLED)
+		if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED)
 			return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
-		if (ibp->port_cap_flags & IB_PORT_SM)
+		if (ibp->rvp.port_cap_flags & IB_PORT_SM)
 			return IB_MAD_RESULT_SUCCESS;
 		/* FALLTHROUGH */
 	default:
@@ -3654,14 +3833,13 @@ static int subn_get_opa_aggregate(struct opa_smp *smp,
 		/* zero the payload for this segment */
 		memset(next_smp + sizeof(*agg), 0, agg_data_len);
 
-		(void) subn_get_opa_sma(agg->attr_id, smp, am, agg->data,
+		(void)subn_get_opa_sma(agg->attr_id, smp, am, agg->data,
 					ibdev, port, NULL);
 		if (smp->status & ~IB_SMP_DIRECTION) {
 			set_aggr_error(agg);
 			return reply((struct ib_mad_hdr *)smp);
 		}
 		next_smp += agg_size;
-
 	}
 
 	return reply((struct ib_mad_hdr *)smp);
@@ -3698,14 +3876,13 @@ static int subn_set_opa_aggregate(struct opa_smp *smp,
 			return reply((struct ib_mad_hdr *)smp);
 		}
 
-		(void) subn_set_opa_sma(agg->attr_id, smp, am, agg->data,
+		(void)subn_set_opa_sma(agg->attr_id, smp, am, agg->data,
 					ibdev, port, NULL);
 		if (smp->status & ~IB_SMP_DIRECTION) {
 			set_aggr_error(agg);
 			return reply((struct ib_mad_hdr *)smp);
 		}
 		next_smp += agg_size;
-
 	}
 
 	return reply((struct ib_mad_hdr *)smp);
@@ -3823,7 +4000,7 @@ static int process_subn_opa(struct ib_device *ibdev, int mad_flags,
 	if (smp->class_version != OPA_SMI_CLASS_VERSION) {
 		smp->status |= IB_SMP_UNSUP_VERSION;
 		ret = reply((struct ib_mad_hdr *)smp);
-		goto bail;
+		return ret;
 	}
 	ret = check_mkey(ibp, (struct ib_mad_hdr *)smp, mad_flags, smp->mkey,
 			 smp->route.dr.dr_slid, smp->route.dr.return_path,
@@ -3843,13 +4020,13 @@ static int process_subn_opa(struct ib_device *ibdev, int mad_flags,
 		     smp->method == IB_MGMT_METHOD_SET) &&
 		    port_num && port_num <= ibdev->phys_port_cnt &&
 		    port != port_num)
-			(void) check_mkey(to_iport(ibdev, port_num),
+			(void)check_mkey(to_iport(ibdev, port_num),
 					  (struct ib_mad_hdr *)smp, 0,
 					  smp->mkey, smp->route.dr.dr_slid,
 					  smp->route.dr.return_path,
 					  smp->hop_cnt);
 		ret = IB_MAD_RESULT_FAILURE;
-		goto bail;
+		return ret;
 	}
 
 	*resp_len = opa_get_smp_header_size(smp);
@@ -3861,23 +4038,25 @@ static int process_subn_opa(struct ib_device *ibdev, int mad_flags,
 			clear_opa_smp_data(smp);
 			ret = subn_get_opa_sma(attr_id, smp, am, data,
 					       ibdev, port, resp_len);
-			goto bail;
+			break;
 		case OPA_ATTRIB_ID_AGGREGATE:
 			ret = subn_get_opa_aggregate(smp, ibdev, port,
 						     resp_len);
-			goto bail;
+			break;
 		}
+		break;
 	case IB_MGMT_METHOD_SET:
 		switch (attr_id) {
 		default:
 			ret = subn_set_opa_sma(attr_id, smp, am, data,
 					       ibdev, port, resp_len);
-			goto bail;
+			break;
 		case OPA_ATTRIB_ID_AGGREGATE:
 			ret = subn_set_opa_aggregate(smp, ibdev, port,
 						     resp_len);
-			goto bail;
+			break;
 		}
+		break;
 	case IB_MGMT_METHOD_TRAP:
 	case IB_MGMT_METHOD_REPORT:
 	case IB_MGMT_METHOD_REPORT_RESP:
@@ -3888,13 +4067,13 @@ static int process_subn_opa(struct ib_device *ibdev, int mad_flags,
 		 * Just tell the caller to process it normally.
 		 */
 		ret = IB_MAD_RESULT_SUCCESS;
-		goto bail;
+		break;
 	default:
 		smp->status |= IB_SMP_UNSUP_METHOD;
 		ret = reply((struct ib_mad_hdr *)smp);
+		break;
 	}
 
-bail:
 	return ret;
 }
 
@@ -3910,7 +4089,7 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 	if (smp->class_version != 1) {
 		smp->status |= IB_SMP_UNSUP_VERSION;
 		ret = reply((struct ib_mad_hdr *)smp);
-		goto bail;
+		return ret;
 	}
 
 	ret = check_mkey(ibp, (struct ib_mad_hdr *)smp, mad_flags,
@@ -3931,13 +4110,13 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 		     smp->method == IB_MGMT_METHOD_SET) &&
 		    port_num && port_num <= ibdev->phys_port_cnt &&
 		    port != port_num)
-			(void) check_mkey(to_iport(ibdev, port_num),
-					  (struct ib_mad_hdr *)smp, 0,
-					  smp->mkey,
-					  (__force __be32)smp->dr_slid,
-					  smp->return_path, smp->hop_cnt);
+			(void)check_mkey(to_iport(ibdev, port_num),
+					 (struct ib_mad_hdr *)smp, 0,
+					 smp->mkey,
+					 (__force __be32)smp->dr_slid,
+					 smp->return_path, smp->hop_cnt);
 		ret = IB_MAD_RESULT_FAILURE;
-		goto bail;
+		return ret;
 	}
 
 	switch (smp->method) {
@@ -3945,15 +4124,77 @@ static int process_subn(struct ib_device *ibdev, int mad_flags,
 		switch (smp->attr_id) {
 		case IB_SMP_ATTR_NODE_INFO:
 			ret = subn_get_nodeinfo(smp, ibdev, port);
-			goto bail;
+			break;
 		default:
 			smp->status |= IB_SMP_UNSUP_METH_ATTR;
 			ret = reply((struct ib_mad_hdr *)smp);
-			goto bail;
+			break;
 		}
+		break;
+	}
+
+	return ret;
+}
+
+static int process_perf(struct ib_device *ibdev, u8 port,
+			const struct ib_mad *in_mad,
+			struct ib_mad *out_mad)
+{
+	struct ib_pma_mad *pmp = (struct ib_pma_mad *)out_mad;
+	struct ib_class_port_info *cpi = (struct ib_class_port_info *)
+						&pmp->data;
+	int ret = IB_MAD_RESULT_FAILURE;
+
+	*out_mad = *in_mad;
+	if (pmp->mad_hdr.class_version != 1) {
+		pmp->mad_hdr.status |= IB_SMP_UNSUP_VERSION;
+		ret = reply((struct ib_mad_hdr *)pmp);
+		return ret;
+	}
+
+	switch (pmp->mad_hdr.method) {
+	case IB_MGMT_METHOD_GET:
+		switch (pmp->mad_hdr.attr_id) {
+		case IB_PMA_PORT_COUNTERS:
+			ret = pma_get_ib_portcounters(pmp, ibdev, port);
+			break;
+		case IB_PMA_PORT_COUNTERS_EXT:
+			ret = pma_get_ib_portcounters_ext(pmp, ibdev, port);
+			break;
+		case IB_PMA_CLASS_PORT_INFO:
+			cpi->capability_mask = IB_PMA_CLASS_CAP_EXT_WIDTH;
+			ret = reply((struct ib_mad_hdr *)pmp);
+			break;
+		default:
+			pmp->mad_hdr.status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply((struct ib_mad_hdr *)pmp);
+			break;
+		}
+		break;
+
+	case IB_MGMT_METHOD_SET:
+		if (pmp->mad_hdr.attr_id) {
+			pmp->mad_hdr.status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply((struct ib_mad_hdr *)pmp);
+		}
+		break;
+
+	case IB_MGMT_METHOD_TRAP:
+	case IB_MGMT_METHOD_GET_RESP:
+		/*
+		 * The ib_mad module will call us to process responses
+		 * before checking for other consumers.
+		 * Just tell the caller to process it normally.
+		 */
+		ret = IB_MAD_RESULT_SUCCESS;
+		break;
+
+	default:
+		pmp->mad_hdr.status |= IB_SMP_UNSUP_METHOD;
+		ret = reply((struct ib_mad_hdr *)pmp);
+		break;
 	}
 
-bail:
 	return ret;
 }
 
@@ -3978,44 +4219,46 @@ static int process_perf_opa(struct ib_device *ibdev, u8 port,
 		switch (pmp->mad_hdr.attr_id) {
 		case IB_PMA_CLASS_PORT_INFO:
 			ret = pma_get_opa_classportinfo(pmp, ibdev, resp_len);
-			goto bail;
+			break;
 		case OPA_PM_ATTRIB_ID_PORT_STATUS:
 			ret = pma_get_opa_portstatus(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						     resp_len);
+			break;
 		case OPA_PM_ATTRIB_ID_DATA_PORT_COUNTERS:
 			ret = pma_get_opa_datacounters(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						       resp_len);
+			break;
 		case OPA_PM_ATTRIB_ID_ERROR_PORT_COUNTERS:
 			ret = pma_get_opa_porterrors(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						     resp_len);
+			break;
 		case OPA_PM_ATTRIB_ID_ERROR_INFO:
 			ret = pma_get_opa_errorinfo(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						    resp_len);
+			break;
 		default:
 			pmp->mad_hdr.status |= IB_SMP_UNSUP_METH_ATTR;
 			ret = reply((struct ib_mad_hdr *)pmp);
-			goto bail;
+			break;
 		}
+		break;
 
 	case IB_MGMT_METHOD_SET:
 		switch (pmp->mad_hdr.attr_id) {
 		case OPA_PM_ATTRIB_ID_CLEAR_PORT_STATUS:
 			ret = pma_set_opa_portstatus(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						     resp_len);
+			break;
 		case OPA_PM_ATTRIB_ID_ERROR_INFO:
 			ret = pma_set_opa_errorinfo(pmp, ibdev, port,
-								resp_len);
-			goto bail;
+						    resp_len);
+			break;
 		default:
 			pmp->mad_hdr.status |= IB_SMP_UNSUP_METH_ATTR;
 			ret = reply((struct ib_mad_hdr *)pmp);
-			goto bail;
+			break;
 		}
+		break;
 
 	case IB_MGMT_METHOD_TRAP:
 	case IB_MGMT_METHOD_GET_RESP:
@@ -4025,14 +4268,14 @@ static int process_perf_opa(struct ib_device *ibdev, u8 port,
 		 * Just tell the caller to process it normally.
 		 */
 		ret = IB_MAD_RESULT_SUCCESS;
-		goto bail;
+		break;
 
 	default:
 		pmp->mad_hdr.status |= IB_SMP_UNSUP_METHOD;
 		ret = reply((struct ib_mad_hdr *)pmp);
+		break;
 	}
 
-bail:
 	return ret;
 }
 
@@ -4097,12 +4340,15 @@ static int hfi1_process_ib_mad(struct ib_device *ibdev, int mad_flags, u8 port,
 	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
 	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
 		ret = process_subn(ibdev, mad_flags, port, in_mad, out_mad);
-		goto bail;
+		break;
+	case IB_MGMT_CLASS_PERF_MGMT:
+		ret = process_perf(ibdev, port, in_mad, out_mad);
+		break;
 	default:
 		ret = IB_MAD_RESULT_SUCCESS;
+		break;
 	}
 
-bail:
 	return ret;
 }
 
@@ -4154,66 +4400,3 @@ int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
 
 	return IB_MAD_RESULT_FAILURE;
 }
-
-static void send_handler(struct ib_mad_agent *agent,
-			 struct ib_mad_send_wc *mad_send_wc)
-{
-	ib_free_send_mad(mad_send_wc->send_buf);
-}
-
-int hfi1_create_agents(struct hfi1_ibdev *dev)
-{
-	struct hfi1_devdata *dd = dd_from_dev(dev);
-	struct ib_mad_agent *agent;
-	struct hfi1_ibport *ibp;
-	int p;
-	int ret;
-
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI,
-					      NULL, 0, send_handler,
-					      NULL, NULL, 0);
-		if (IS_ERR(agent)) {
-			ret = PTR_ERR(agent);
-			goto err;
-		}
-
-		ibp->send_agent = agent;
-	}
-
-	return 0;
-
-err:
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		if (ibp->send_agent) {
-			agent = ibp->send_agent;
-			ibp->send_agent = NULL;
-			ib_unregister_mad_agent(agent);
-		}
-	}
-
-	return ret;
-}
-
-void hfi1_free_agents(struct hfi1_ibdev *dev)
-{
-	struct hfi1_devdata *dd = dd_from_dev(dev);
-	struct ib_mad_agent *agent;
-	struct hfi1_ibport *ibp;
-	int p;
-
-	for (p = 0; p < dd->num_pports; p++) {
-		ibp = &dd->pport[p].ibport_data;
-		if (ibp->send_agent) {
-			agent = ibp->send_agent;
-			ibp->send_agent = NULL;
-			ib_unregister_mad_agent(agent);
-		}
-		if (ibp->sm_ah) {
-			ib_destroy_ah(&ibp->sm_ah->ibah);
-			ibp->sm_ah = NULL;
-		}
-	}
-}
diff --git a/drivers/staging/rdma/hfi1/mad.h b/drivers/staging/rdma/hfi1/mad.h
index f0317750e2fc..55ee08675333 100644
--- a/drivers/staging/rdma/hfi1/mad.h
+++ b/drivers/staging/rdma/hfi1/mad.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -51,8 +48,10 @@
 #define _HFI1_MAD_H
 
 #include <rdma/ib_pma.h>
-#define USE_PI_LED_ENABLE	1 /* use led enabled bit in struct
-				   * opa_port_states, if available */
+#define USE_PI_LED_ENABLE	1 /*
+				   * use led enabled bit in struct
+				   * opa_port_states, if available
+				   */
 #include <rdma/opa_smi.h>
 #include <rdma/opa_port_info.h>
 #ifndef PI_LED_ENABLE_SUP
@@ -235,7 +234,6 @@ struct ib_pma_portcounters_cong {
 #define IB_CC_SVCTYPE_RD 0x2
 #define IB_CC_SVCTYPE_UD 0x3
 
-
 /*
  * There should be an equivalent IB #define for the following, but
  * I cannot find it.
@@ -267,7 +265,7 @@ struct opa_hfi1_cong_log {
 	u8 congestion_flags;
 	__be16 threshold_event_counter;
 	__be32 current_time_stamp;
-	u8 threshold_cong_event_map[OPA_MAX_SLS/8];
+	u8 threshold_cong_event_map[OPA_MAX_SLS / 8];
 	struct opa_hfi1_cong_log_event events[OPA_CONG_LOG_ELEMS];
 } __packed;
 
diff --git a/drivers/staging/rdma/hfi1/mmu_rb.c b/drivers/staging/rdma/hfi1/mmu_rb.c
new file mode 100644
index 000000000000..c7ad0164ea9a
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/mmu_rb.c
@@ -0,0 +1,292 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <linux/list.h>
+#include <linux/mmu_notifier.h>
+#include <linux/interval_tree_generic.h>
+
+#include "mmu_rb.h"
+#include "trace.h"
+
+struct mmu_rb_handler {
+	struct list_head list;
+	struct mmu_notifier mn;
+	struct rb_root *root;
+	spinlock_t lock;        /* protect the RB tree */
+	struct mmu_rb_ops *ops;
+};
+
+static LIST_HEAD(mmu_rb_handlers);
+static DEFINE_SPINLOCK(mmu_rb_lock); /* protect mmu_rb_handlers list */
+
+static unsigned long mmu_node_start(struct mmu_rb_node *);
+static unsigned long mmu_node_last(struct mmu_rb_node *);
+static struct mmu_rb_handler *find_mmu_handler(struct rb_root *);
+static inline void mmu_notifier_page(struct mmu_notifier *, struct mm_struct *,
+				     unsigned long);
+static inline void mmu_notifier_range_start(struct mmu_notifier *,
+					    struct mm_struct *,
+					    unsigned long, unsigned long);
+static void mmu_notifier_mem_invalidate(struct mmu_notifier *,
+					unsigned long, unsigned long);
+static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *,
+					   unsigned long, unsigned long);
+
+static struct mmu_notifier_ops mn_opts = {
+	.invalidate_page = mmu_notifier_page,
+	.invalidate_range_start = mmu_notifier_range_start,
+};
+
+INTERVAL_TREE_DEFINE(struct mmu_rb_node, node, unsigned long, __last,
+		     mmu_node_start, mmu_node_last, static, __mmu_int_rb);
+
+static unsigned long mmu_node_start(struct mmu_rb_node *node)
+{
+	return node->addr & PAGE_MASK;
+}
+
+static unsigned long mmu_node_last(struct mmu_rb_node *node)
+{
+	return PAGE_ALIGN((node->addr & PAGE_MASK) + node->len) - 1;
+}
+
+int hfi1_mmu_rb_register(struct rb_root *root, struct mmu_rb_ops *ops)
+{
+	struct mmu_rb_handler *handlr;
+	unsigned long flags;
+
+	if (!ops->invalidate)
+		return -EINVAL;
+
+	handlr = kmalloc(sizeof(*handlr), GFP_KERNEL);
+	if (!handlr)
+		return -ENOMEM;
+
+	handlr->root = root;
+	handlr->ops = ops;
+	INIT_HLIST_NODE(&handlr->mn.hlist);
+	spin_lock_init(&handlr->lock);
+	handlr->mn.ops = &mn_opts;
+	spin_lock_irqsave(&mmu_rb_lock, flags);
+	list_add_tail(&handlr->list, &mmu_rb_handlers);
+	spin_unlock_irqrestore(&mmu_rb_lock, flags);
+
+	return mmu_notifier_register(&handlr->mn, current->mm);
+}
+
+void hfi1_mmu_rb_unregister(struct rb_root *root)
+{
+	struct mmu_rb_handler *handler = find_mmu_handler(root);
+	unsigned long flags;
+
+	if (!handler)
+		return;
+
+	spin_lock_irqsave(&mmu_rb_lock, flags);
+	list_del(&handler->list);
+	spin_unlock_irqrestore(&mmu_rb_lock, flags);
+
+	if (!RB_EMPTY_ROOT(root)) {
+		struct rb_node *node;
+		struct mmu_rb_node *rbnode;
+
+		while ((node = rb_first(root))) {
+			rbnode = rb_entry(node, struct mmu_rb_node, node);
+			rb_erase(node, root);
+			if (handler->ops->remove)
+				handler->ops->remove(root, rbnode, false);
+		}
+	}
+
+	if (current->mm)
+		mmu_notifier_unregister(&handler->mn, current->mm);
+	kfree(handler);
+}
+
+int hfi1_mmu_rb_insert(struct rb_root *root, struct mmu_rb_node *mnode)
+{
+	struct mmu_rb_handler *handler = find_mmu_handler(root);
+	struct mmu_rb_node *node;
+	unsigned long flags;
+	int ret = 0;
+
+	if (!handler)
+		return -EINVAL;
+
+	spin_lock_irqsave(&handler->lock, flags);
+	hfi1_cdbg(MMU, "Inserting node addr 0x%llx, len %u", mnode->addr,
+		  mnode->len);
+	node = __mmu_rb_search(handler, mnode->addr, mnode->len);
+	if (node) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+	__mmu_int_rb_insert(mnode, root);
+
+	if (handler->ops->insert) {
+		ret = handler->ops->insert(root, mnode);
+		if (ret)
+			__mmu_int_rb_remove(mnode, root);
+	}
+unlock:
+	spin_unlock_irqrestore(&handler->lock, flags);
+	return ret;
+}
+
+/* Caller must host handler lock */
+static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
+					   unsigned long addr,
+					   unsigned long len)
+{
+	struct mmu_rb_node *node = NULL;
+
+	hfi1_cdbg(MMU, "Searching for addr 0x%llx, len %u", addr, len);
+	if (!handler->ops->filter) {
+		node = __mmu_int_rb_iter_first(handler->root, addr,
+					       (addr + len) - 1);
+	} else {
+		for (node = __mmu_int_rb_iter_first(handler->root, addr,
+						    (addr + len) - 1);
+		     node;
+		     node = __mmu_int_rb_iter_next(node, addr,
+						   (addr + len) - 1)) {
+			if (handler->ops->filter(node, addr, len))
+				return node;
+		}
+	}
+	return node;
+}
+
+static void __mmu_rb_remove(struct mmu_rb_handler *handler,
+			    struct mmu_rb_node *node, bool arg)
+{
+	/* Validity of handler and node pointers has been checked by caller. */
+	hfi1_cdbg(MMU, "Removing node addr 0x%llx, len %u", node->addr,
+		  node->len);
+	__mmu_int_rb_remove(node, handler->root);
+	if (handler->ops->remove)
+		handler->ops->remove(handler->root, node, arg);
+}
+
+struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *root, unsigned long addr,
+				       unsigned long len)
+{
+	struct mmu_rb_handler *handler = find_mmu_handler(root);
+	struct mmu_rb_node *node;
+	unsigned long flags;
+
+	if (!handler)
+		return ERR_PTR(-EINVAL);
+
+	spin_lock_irqsave(&handler->lock, flags);
+	node = __mmu_rb_search(handler, addr, len);
+	spin_unlock_irqrestore(&handler->lock, flags);
+
+	return node;
+}
+
+void hfi1_mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node)
+{
+	struct mmu_rb_handler *handler = find_mmu_handler(root);
+	unsigned long flags;
+
+	if (!handler || !node)
+		return;
+
+	spin_lock_irqsave(&handler->lock, flags);
+	__mmu_rb_remove(handler, node, false);
+	spin_unlock_irqrestore(&handler->lock, flags);
+}
+
+static struct mmu_rb_handler *find_mmu_handler(struct rb_root *root)
+{
+	struct mmu_rb_handler *handler;
+	unsigned long flags;
+
+	spin_lock_irqsave(&mmu_rb_lock, flags);
+	list_for_each_entry(handler, &mmu_rb_handlers, list) {
+		if (handler->root == root)
+			goto unlock;
+	}
+	handler = NULL;
+unlock:
+	spin_unlock_irqrestore(&mmu_rb_lock, flags);
+	return handler;
+}
+
+static inline void mmu_notifier_page(struct mmu_notifier *mn,
+				     struct mm_struct *mm, unsigned long addr)
+{
+	mmu_notifier_mem_invalidate(mn, addr, addr + PAGE_SIZE);
+}
+
+static inline void mmu_notifier_range_start(struct mmu_notifier *mn,
+					    struct mm_struct *mm,
+					    unsigned long start,
+					    unsigned long end)
+{
+	mmu_notifier_mem_invalidate(mn, start, end);
+}
+
+static void mmu_notifier_mem_invalidate(struct mmu_notifier *mn,
+					unsigned long start, unsigned long end)
+{
+	struct mmu_rb_handler *handler =
+		container_of(mn, struct mmu_rb_handler, mn);
+	struct rb_root *root = handler->root;
+	struct mmu_rb_node *node;
+	unsigned long flags;
+
+	spin_lock_irqsave(&handler->lock, flags);
+	for (node = __mmu_int_rb_iter_first(root, start, end - 1); node;
+	     node = __mmu_int_rb_iter_next(node, start, end - 1)) {
+		hfi1_cdbg(MMU, "Invalidating node addr 0x%llx, len %u",
+			  node->addr, node->len);
+		if (handler->ops->invalidate(root, node))
+			__mmu_rb_remove(handler, node, true);
+	}
+	spin_unlock_irqrestore(&handler->lock, flags);
+}
diff --git a/drivers/staging/rdma/hfi1/mmu_rb.h b/drivers/staging/rdma/hfi1/mmu_rb.h
new file mode 100644
index 000000000000..f8523fdb8a18
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/mmu_rb.h
@@ -0,0 +1,73 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#ifndef _HFI1_MMU_RB_H
+#define _HFI1_MMU_RB_H
+
+#include "hfi.h"
+
+struct mmu_rb_node {
+	unsigned long addr;
+	unsigned long len;
+	unsigned long __last;
+	struct rb_node node;
+};
+
+struct mmu_rb_ops {
+	bool (*filter)(struct mmu_rb_node *, unsigned long, unsigned long);
+	int (*insert)(struct rb_root *, struct mmu_rb_node *);
+	void (*remove)(struct rb_root *, struct mmu_rb_node *, bool);
+	int (*invalidate)(struct rb_root *, struct mmu_rb_node *);
+};
+
+int hfi1_mmu_rb_register(struct rb_root *root, struct mmu_rb_ops *ops);
+void hfi1_mmu_rb_unregister(struct rb_root *);
+int hfi1_mmu_rb_insert(struct rb_root *, struct mmu_rb_node *);
+void hfi1_mmu_rb_remove(struct rb_root *, struct mmu_rb_node *);
+struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *, unsigned long,
+				       unsigned long);
+
+#endif /* _HFI1_MMU_RB_H */
diff --git a/drivers/staging/rdma/hfi1/mr.c b/drivers/staging/rdma/hfi1/mr.c
deleted file mode 100644
index 38253212af7a..000000000000
--- a/drivers/staging/rdma/hfi1/mr.c
+++ /dev/null
@@ -1,473 +0,0 @@
-/*
- *
- * This file is provided under a dual BSD/GPLv2 license.  When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *  - Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  - Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *  - Neither the name of Intel Corporation nor the names of its
- *    contributors may be used to endorse or promote products derived
- *    from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- */
-
-#include <rdma/ib_umem.h>
-#include <rdma/ib_smi.h>
-
-#include "hfi.h"
-
-/* Fast memory region */
-struct hfi1_fmr {
-	struct ib_fmr ibfmr;
-	struct hfi1_mregion mr;        /* must be last */
-};
-
-static inline struct hfi1_fmr *to_ifmr(struct ib_fmr *ibfmr)
-{
-	return container_of(ibfmr, struct hfi1_fmr, ibfmr);
-}
-
-static int init_mregion(struct hfi1_mregion *mr, struct ib_pd *pd,
-			int count)
-{
-	int m, i = 0;
-	int rval = 0;
-
-	m = DIV_ROUND_UP(count, HFI1_SEGSZ);
-	for (; i < m; i++) {
-		mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL);
-		if (!mr->map[i])
-			goto bail;
-	}
-	mr->mapsz = m;
-	init_completion(&mr->comp);
-	/* count returning the ptr to user */
-	atomic_set(&mr->refcount, 1);
-	mr->pd = pd;
-	mr->max_segs = count;
-out:
-	return rval;
-bail:
-	while (i)
-		kfree(mr->map[--i]);
-	rval = -ENOMEM;
-	goto out;
-}
-
-static void deinit_mregion(struct hfi1_mregion *mr)
-{
-	int i = mr->mapsz;
-
-	mr->mapsz = 0;
-	while (i)
-		kfree(mr->map[--i]);
-}
-
-
-/**
- * hfi1_get_dma_mr - get a DMA memory region
- * @pd: protection domain for this memory region
- * @acc: access flags
- *
- * Returns the memory region on success, otherwise returns an errno.
- * Note that all DMA addresses should be created via the
- * struct ib_dma_mapping_ops functions (see dma.c).
- */
-struct ib_mr *hfi1_get_dma_mr(struct ib_pd *pd, int acc)
-{
-	struct hfi1_mr *mr = NULL;
-	struct ib_mr *ret;
-	int rval;
-
-	if (to_ipd(pd)->user) {
-		ret = ERR_PTR(-EPERM);
-		goto bail;
-	}
-
-	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
-	if (!mr) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	rval = init_mregion(&mr->mr, pd, 0);
-	if (rval) {
-		ret = ERR_PTR(rval);
-		goto bail;
-	}
-
-
-	rval = hfi1_alloc_lkey(&mr->mr, 1);
-	if (rval) {
-		ret = ERR_PTR(rval);
-		goto bail_mregion;
-	}
-
-	mr->mr.access_flags = acc;
-	ret = &mr->ibmr;
-done:
-	return ret;
-
-bail_mregion:
-	deinit_mregion(&mr->mr);
-bail:
-	kfree(mr);
-	goto done;
-}
-
-static struct hfi1_mr *alloc_mr(int count, struct ib_pd *pd)
-{
-	struct hfi1_mr *mr;
-	int rval = -ENOMEM;
-	int m;
-
-	/* Allocate struct plus pointers to first level page tables. */
-	m = DIV_ROUND_UP(count, HFI1_SEGSZ);
-	mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL);
-	if (!mr)
-		goto bail;
-
-	rval = init_mregion(&mr->mr, pd, count);
-	if (rval)
-		goto bail;
-
-	rval = hfi1_alloc_lkey(&mr->mr, 0);
-	if (rval)
-		goto bail_mregion;
-	mr->ibmr.lkey = mr->mr.lkey;
-	mr->ibmr.rkey = mr->mr.lkey;
-done:
-	return mr;
-
-bail_mregion:
-	deinit_mregion(&mr->mr);
-bail:
-	kfree(mr);
-	mr = ERR_PTR(rval);
-	goto done;
-}
-
-/**
- * hfi1_reg_user_mr - register a userspace memory region
- * @pd: protection domain for this memory region
- * @start: starting userspace address
- * @length: length of region to register
- * @mr_access_flags: access flags for this memory region
- * @udata: unused by the driver
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_mr *hfi1_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
-			       u64 virt_addr, int mr_access_flags,
-			       struct ib_udata *udata)
-{
-	struct hfi1_mr *mr;
-	struct ib_umem *umem;
-	struct scatterlist *sg;
-	int n, m, entry;
-	struct ib_mr *ret;
-
-	if (length == 0) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
-
-	umem = ib_umem_get(pd->uobject->context, start, length,
-			   mr_access_flags, 0);
-	if (IS_ERR(umem))
-		return (void *) umem;
-
-	n = umem->nmap;
-
-	mr = alloc_mr(n, pd);
-	if (IS_ERR(mr)) {
-		ret = (struct ib_mr *)mr;
-		ib_umem_release(umem);
-		goto bail;
-	}
-
-	mr->mr.user_base = start;
-	mr->mr.iova = virt_addr;
-	mr->mr.length = length;
-	mr->mr.offset = ib_umem_offset(umem);
-	mr->mr.access_flags = mr_access_flags;
-	mr->umem = umem;
-
-	if (is_power_of_2(umem->page_size))
-		mr->mr.page_shift = ilog2(umem->page_size);
-	m = 0;
-	n = 0;
-	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
-		void *vaddr;
-
-		vaddr = page_address(sg_page(sg));
-		if (!vaddr) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-		mr->mr.map[m]->segs[n].vaddr = vaddr;
-		mr->mr.map[m]->segs[n].length = umem->page_size;
-		n++;
-		if (n == HFI1_SEGSZ) {
-			m++;
-			n = 0;
-		}
-	}
-	ret = &mr->ibmr;
-
-bail:
-	return ret;
-}
-
-/**
- * hfi1_dereg_mr - unregister and free a memory region
- * @ibmr: the memory region to free
- *
- * Returns 0 on success.
- *
- * Note that this is called to free MRs created by hfi1_get_dma_mr()
- * or hfi1_reg_user_mr().
- */
-int hfi1_dereg_mr(struct ib_mr *ibmr)
-{
-	struct hfi1_mr *mr = to_imr(ibmr);
-	int ret = 0;
-	unsigned long timeout;
-
-	hfi1_free_lkey(&mr->mr);
-
-	hfi1_put_mr(&mr->mr); /* will set completion if last */
-	timeout = wait_for_completion_timeout(&mr->mr.comp,
-		5 * HZ);
-	if (!timeout) {
-		dd_dev_err(
-			dd_from_ibdev(mr->mr.pd->device),
-			"hfi1_dereg_mr timeout mr %p pd %p refcount %u\n",
-			mr, mr->mr.pd, atomic_read(&mr->mr.refcount));
-		hfi1_get_mr(&mr->mr);
-		ret = -EBUSY;
-		goto out;
-	}
-	deinit_mregion(&mr->mr);
-	if (mr->umem)
-		ib_umem_release(mr->umem);
-	kfree(mr);
-out:
-	return ret;
-}
-
-/*
- * Allocate a memory region usable with the
- * IB_WR_REG_MR send work request.
- *
- * Return the memory region on success, otherwise return an errno.
- * FIXME: IB_WR_REG_MR is not supported
- */
-struct ib_mr *hfi1_alloc_mr(struct ib_pd *pd,
-			    enum ib_mr_type mr_type,
-			    u32 max_num_sg)
-{
-	struct hfi1_mr *mr;
-
-	if (mr_type != IB_MR_TYPE_MEM_REG)
-		return ERR_PTR(-EINVAL);
-
-	mr = alloc_mr(max_num_sg, pd);
-	if (IS_ERR(mr))
-		return (struct ib_mr *)mr;
-
-	return &mr->ibmr;
-}
-
-/**
- * hfi1_alloc_fmr - allocate a fast memory region
- * @pd: the protection domain for this memory region
- * @mr_access_flags: access flags for this memory region
- * @fmr_attr: fast memory region attributes
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_fmr *hfi1_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
-			      struct ib_fmr_attr *fmr_attr)
-{
-	struct hfi1_fmr *fmr;
-	int m;
-	struct ib_fmr *ret;
-	int rval = -ENOMEM;
-
-	/* Allocate struct plus pointers to first level page tables. */
-	m = DIV_ROUND_UP(fmr_attr->max_pages, HFI1_SEGSZ);
-	fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL);
-	if (!fmr)
-		goto bail;
-
-	rval = init_mregion(&fmr->mr, pd, fmr_attr->max_pages);
-	if (rval)
-		goto bail;
-
-	/*
-	 * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
-	 * rkey.
-	 */
-	rval = hfi1_alloc_lkey(&fmr->mr, 0);
-	if (rval)
-		goto bail_mregion;
-	fmr->ibfmr.rkey = fmr->mr.lkey;
-	fmr->ibfmr.lkey = fmr->mr.lkey;
-	/*
-	 * Resources are allocated but no valid mapping (RKEY can't be
-	 * used).
-	 */
-	fmr->mr.access_flags = mr_access_flags;
-	fmr->mr.max_segs = fmr_attr->max_pages;
-	fmr->mr.page_shift = fmr_attr->page_shift;
-
-	ret = &fmr->ibfmr;
-done:
-	return ret;
-
-bail_mregion:
-	deinit_mregion(&fmr->mr);
-bail:
-	kfree(fmr);
-	ret = ERR_PTR(rval);
-	goto done;
-}
-
-/**
- * hfi1_map_phys_fmr - set up a fast memory region
- * @ibmfr: the fast memory region to set up
- * @page_list: the list of pages to associate with the fast memory region
- * @list_len: the number of pages to associate with the fast memory region
- * @iova: the virtual address of the start of the fast memory region
- *
- * This may be called from interrupt context.
- */
-
-int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
-		      int list_len, u64 iova)
-{
-	struct hfi1_fmr *fmr = to_ifmr(ibfmr);
-	struct hfi1_lkey_table *rkt;
-	unsigned long flags;
-	int m, n, i;
-	u32 ps;
-	int ret;
-
-	i = atomic_read(&fmr->mr.refcount);
-	if (i > 2)
-		return -EBUSY;
-
-	if (list_len > fmr->mr.max_segs) {
-		ret = -EINVAL;
-		goto bail;
-	}
-	rkt = &to_idev(ibfmr->device)->lk_table;
-	spin_lock_irqsave(&rkt->lock, flags);
-	fmr->mr.user_base = iova;
-	fmr->mr.iova = iova;
-	ps = 1 << fmr->mr.page_shift;
-	fmr->mr.length = list_len * ps;
-	m = 0;
-	n = 0;
-	for (i = 0; i < list_len; i++) {
-		fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
-		fmr->mr.map[m]->segs[n].length = ps;
-		if (++n == HFI1_SEGSZ) {
-			m++;
-			n = 0;
-		}
-	}
-	spin_unlock_irqrestore(&rkt->lock, flags);
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * hfi1_unmap_fmr - unmap fast memory regions
- * @fmr_list: the list of fast memory regions to unmap
- *
- * Returns 0 on success.
- */
-int hfi1_unmap_fmr(struct list_head *fmr_list)
-{
-	struct hfi1_fmr *fmr;
-	struct hfi1_lkey_table *rkt;
-	unsigned long flags;
-
-	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
-		rkt = &to_idev(fmr->ibfmr.device)->lk_table;
-		spin_lock_irqsave(&rkt->lock, flags);
-		fmr->mr.user_base = 0;
-		fmr->mr.iova = 0;
-		fmr->mr.length = 0;
-		spin_unlock_irqrestore(&rkt->lock, flags);
-	}
-	return 0;
-}
-
-/**
- * hfi1_dealloc_fmr - deallocate a fast memory region
- * @ibfmr: the fast memory region to deallocate
- *
- * Returns 0 on success.
- */
-int hfi1_dealloc_fmr(struct ib_fmr *ibfmr)
-{
-	struct hfi1_fmr *fmr = to_ifmr(ibfmr);
-	int ret = 0;
-	unsigned long timeout;
-
-	hfi1_free_lkey(&fmr->mr);
-	hfi1_put_mr(&fmr->mr); /* will set completion if last */
-	timeout = wait_for_completion_timeout(&fmr->mr.comp,
-		5 * HZ);
-	if (!timeout) {
-		hfi1_get_mr(&fmr->mr);
-		ret = -EBUSY;
-		goto out;
-	}
-	deinit_mregion(&fmr->mr);
-	kfree(fmr);
-out:
-	return ret;
-}
diff --git a/drivers/staging/rdma/hfi1/opa_compat.h b/drivers/staging/rdma/hfi1/opa_compat.h
index f64eec1c2951..6ef3c1cbdcd7 100644
--- a/drivers/staging/rdma/hfi1/opa_compat.h
+++ b/drivers/staging/rdma/hfi1/opa_compat.h
@@ -1,14 +1,13 @@
 #ifndef _LINUX_H
 #define _LINUX_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -111,19 +108,4 @@ enum opa_port_phys_state {
 	/* values 12-15 are reserved/ignored */
 };
 
-/* OPA_PORT_TYPE_* definitions - these belong in opa_port_info.h */
-#define OPA_PORT_TYPE_UNKNOWN          0
-#define OPA_PORT_TYPE_DISCONNECTED     1
-/* port is not currently usable, CableInfo not available */
-#define OPA_PORT_TYPE_FIXED            2
-/* A fixed backplane port in a director class switch. All OPA ASICS */
-#define OPA_PORT_TYPE_VARIABLE         3
-/* A backplane port in a blade system, possibly mixed configuration */
-#define OPA_PORT_TYPE_STANDARD         4
-/* implies a SFF-8636 defined format for CableInfo (QSFP) */
-#define OPA_PORT_TYPE_SI_PHOTONICS      5
-/* A silicon photonics module implies TBD defined format for CableInfo
- * as defined by Intel SFO group */
-/* 6 - 15 are reserved */
-
 #endif /* _LINUX_H */
diff --git a/drivers/staging/rdma/hfi1/pcie.c b/drivers/staging/rdma/hfi1/pcie.c
index 47ca6314e328..0bac21e6a658 100644
--- a/drivers/staging/rdma/hfi1/pcie.c
+++ b/drivers/staging/rdma/hfi1/pcie.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -57,6 +54,7 @@
 
 #include "hfi.h"
 #include "chip_registers.h"
+#include "aspm.h"
 
 /* link speed vector for Gen3 speed - not in Linux headers */
 #define GEN1_SPEED_VECTOR 0x1
@@ -122,8 +120,9 @@ int hfi1_pcie_init(struct pci_dev *pdev, const struct pci_device_id *ent)
 			goto bail;
 		}
 		ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
-	} else
+	} else {
 		ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+	}
 	if (ret) {
 		hfi1_early_err(&pdev->dev,
 			       "Unable to set DMA consistent mask: %d\n", ret);
@@ -131,13 +130,7 @@ int hfi1_pcie_init(struct pci_dev *pdev, const struct pci_device_id *ent)
 	}
 
 	pci_set_master(pdev);
-	ret = pci_enable_pcie_error_reporting(pdev);
-	if (ret) {
-		hfi1_early_err(&pdev->dev,
-			       "Unable to enable pcie error reporting: %d\n",
-			      ret);
-		ret = 0;
-	}
+	(void)pci_enable_pcie_error_reporting(pdev);
 	goto done;
 
 bail:
@@ -222,10 +215,9 @@ int hfi1_pcie_ddinit(struct hfi1_devdata *dd, struct pci_dev *pdev,
 	pcie_capability_read_word(dd->pcidev, PCI_EXP_DEVCTL, &dd->pcie_devctl);
 	pcie_capability_read_word(dd->pcidev, PCI_EXP_LNKCTL, &dd->pcie_lnkctl);
 	pcie_capability_read_word(dd->pcidev, PCI_EXP_DEVCTL2,
-							&dd->pcie_devctl2);
+				  &dd->pcie_devctl2);
 	pci_read_config_dword(dd->pcidev, PCI_CFG_MSIX0, &dd->pci_msix0);
-	pci_read_config_dword(dd->pcidev, PCIE_CFG_SPCIE1,
-							&dd->pci_lnkctl3);
+	pci_read_config_dword(dd->pcidev, PCIE_CFG_SPCIE1, &dd->pci_lnkctl3);
 	pci_read_config_dword(dd->pcidev, PCIE_CFG_TPH2, &dd->pci_tph2);
 
 	return 0;
@@ -238,7 +230,7 @@ int hfi1_pcie_ddinit(struct hfi1_devdata *dd, struct pci_dev *pdev,
  */
 void hfi1_pcie_ddcleanup(struct hfi1_devdata *dd)
 {
-	u64 __iomem *base = (void __iomem *) dd->kregbase;
+	u64 __iomem *base = (void __iomem *)dd->kregbase;
 
 	dd->flags &= ~HFI1_PRESENT;
 	dd->kregbase = NULL;
@@ -274,7 +266,7 @@ void hfi1_pcie_flr(struct hfi1_devdata *dd)
 
 clear:
 	pcie_capability_set_word(dd->pcidev, PCI_EXP_DEVCTL,
-						PCI_EXP_DEVCTL_BCR_FLR);
+				 PCI_EXP_DEVCTL_BCR_FLR);
 	/* PCIe spec requires the function to be back within 100ms */
 	msleep(100);
 }
@@ -287,9 +279,11 @@ static void msix_setup(struct hfi1_devdata *dd, int pos, u32 *msixcnt,
 	struct msix_entry *msix_entry;
 	int i;
 
-	/* We can't pass hfi1_msix_entry array to msix_setup
+	/*
+	 * We can't pass hfi1_msix_entry array to msix_setup
 	 * so use a dummy msix_entry array and copy the allocated
-	 * irq back to the hfi1_msix_entry array. */
+	 * irq back to the hfi1_msix_entry array.
+	 */
 	msix_entry = kmalloc_array(nvec, sizeof(*msix_entry), GFP_KERNEL);
 	if (!msix_entry) {
 		ret = -ENOMEM;
@@ -319,7 +313,6 @@ do_intx:
 		   nvec, ret);
 	*msixcnt = 0;
 	hfi1_enable_intx(dd->pcidev);
-
 }
 
 /* return the PCIe link speed from the given link status */
@@ -367,6 +360,7 @@ static void update_lbus_info(struct hfi1_devdata *dd)
 int pcie_speeds(struct hfi1_devdata *dd)
 {
 	u32 linkcap;
+	struct pci_dev *parent = dd->pcidev->bus->self;
 
 	if (!pci_is_pcie(dd->pcidev)) {
 		dd_dev_err(dd, "Can't find PCI Express capability!\n");
@@ -379,15 +373,15 @@ int pcie_speeds(struct hfi1_devdata *dd)
 	pcie_capability_read_dword(dd->pcidev, PCI_EXP_LNKCAP, &linkcap);
 	if ((linkcap & PCI_EXP_LNKCAP_SLS) != GEN3_SPEED_VECTOR) {
 		dd_dev_info(dd,
-			"This HFI is not Gen3 capable, max speed 0x%x, need 0x3\n",
-			linkcap & PCI_EXP_LNKCAP_SLS);
+			    "This HFI is not Gen3 capable, max speed 0x%x, need 0x3\n",
+			    linkcap & PCI_EXP_LNKCAP_SLS);
 		dd->link_gen3_capable = 0;
 	}
 
 	/*
 	 * bus->max_bus_speed is set from the bridge's linkcap Max Link Speed
 	 */
-	if (dd->pcidev->bus->max_bus_speed != PCIE_SPEED_8_0GT) {
+	if (parent && dd->pcidev->bus->max_bus_speed != PCIE_SPEED_8_0GT) {
 		dd_dev_info(dd, "Parent PCIe bridge does not support Gen3\n");
 		dd->link_gen3_capable = 0;
 	}
@@ -395,9 +389,7 @@ int pcie_speeds(struct hfi1_devdata *dd)
 	/* obtain the link width and current speed */
 	update_lbus_info(dd);
 
-	/* check against expected pcie width and complain if "wrong" */
-	if (dd->lbus_width < 16)
-		dd_dev_err(dd, "PCIe width %u (x16 HFI)\n", dd->lbus_width);
+	dd_dev_info(dd, "%s\n", dd->lbus_info);
 
 	return 0;
 }
@@ -436,23 +428,18 @@ void hfi1_enable_intx(struct pci_dev *pdev)
 void restore_pci_variables(struct hfi1_devdata *dd)
 {
 	pci_write_config_word(dd->pcidev, PCI_COMMAND, dd->pci_command);
-	pci_write_config_dword(dd->pcidev,
-				PCI_BASE_ADDRESS_0, dd->pcibar0);
-	pci_write_config_dword(dd->pcidev,
-				PCI_BASE_ADDRESS_1, dd->pcibar1);
-	pci_write_config_dword(dd->pcidev,
-				PCI_ROM_ADDRESS, dd->pci_rom);
+	pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_0, dd->pcibar0);
+	pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_1, dd->pcibar1);
+	pci_write_config_dword(dd->pcidev, PCI_ROM_ADDRESS, dd->pci_rom);
 	pcie_capability_write_word(dd->pcidev, PCI_EXP_DEVCTL, dd->pcie_devctl);
 	pcie_capability_write_word(dd->pcidev, PCI_EXP_LNKCTL, dd->pcie_lnkctl);
 	pcie_capability_write_word(dd->pcidev, PCI_EXP_DEVCTL2,
-							dd->pcie_devctl2);
+				   dd->pcie_devctl2);
 	pci_write_config_dword(dd->pcidev, PCI_CFG_MSIX0, dd->pci_msix0);
-	pci_write_config_dword(dd->pcidev, PCIE_CFG_SPCIE1,
-							dd->pci_lnkctl3);
+	pci_write_config_dword(dd->pcidev, PCIE_CFG_SPCIE1, dd->pci_lnkctl3);
 	pci_write_config_dword(dd->pcidev, PCIE_CFG_TPH2, dd->pci_tph2);
 }
 
-
 /*
  * BIOS may not set PCIe bus-utilization parameters for best performance.
  * Check and optionally adjust them to maximize our throughput.
@@ -461,6 +448,10 @@ static int hfi1_pcie_caps;
 module_param_named(pcie_caps, hfi1_pcie_caps, int, S_IRUGO);
 MODULE_PARM_DESC(pcie_caps, "Max PCIe tuning: Payload (0..3), ReadReq (4..7)");
 
+uint aspm_mode = ASPM_MODE_DISABLED;
+module_param_named(aspm, aspm_mode, uint, S_IRUGO);
+MODULE_PARM_DESC(aspm, "PCIe ASPM: 0: disable, 1: enable, 2: dynamic");
+
 static void tune_pcie_caps(struct hfi1_devdata *dd)
 {
 	struct pci_dev *parent;
@@ -479,6 +470,12 @@ static void tune_pcie_caps(struct hfi1_devdata *dd)
 	}
 	/* Find out supported and configured values for parent (root) */
 	parent = dd->pcidev->bus->self;
+	/*
+	 * The driver cannot perform the tuning if it does not have
+	 * access to the upstream component.
+	 */
+	if (!parent)
+		return;
 	if (!pci_is_root_bus(parent->bus)) {
 		dd_dev_info(dd, "Parent not root\n");
 		return;
@@ -532,6 +529,7 @@ static void tune_pcie_caps(struct hfi1_devdata *dd)
 		pcie_set_readrq(dd->pcidev, ep_mrrs);
 	}
 }
+
 /* End of PCIe capability tuning */
 
 /*
@@ -746,21 +744,22 @@ static int load_eq_table(struct hfi1_devdata *dd, const u8 eq[11][3], u8 fs,
 		c0 = fs - (eq[i][PREC] / div) - (eq[i][POST] / div);
 		c_plus1 = eq[i][POST] / div;
 		pci_write_config_dword(pdev, PCIE_CFG_REG_PL102,
-			eq_value(c_minus1, c0, c_plus1));
+				       eq_value(c_minus1, c0, c_plus1));
 		/* check if these coefficients violate EQ rules */
 		pci_read_config_dword(dd->pcidev, PCIE_CFG_REG_PL105,
-								&violation);
+				      &violation);
 		if (violation
 		    & PCIE_CFG_REG_PL105_GEN3_EQ_VIOLATE_COEF_RULES_SMASK){
 			if (hit_error == 0) {
 				dd_dev_err(dd,
-					"Gen3 EQ Table Coefficient rule violations\n");
+					   "Gen3 EQ Table Coefficient rule violations\n");
 				dd_dev_err(dd, "         prec   attn   post\n");
 			}
 			dd_dev_err(dd, "   p%02d:   %02x     %02x     %02x\n",
-				i, (u32)eq[i][0], (u32)eq[i][1], (u32)eq[i][2]);
+				   i, (u32)eq[i][0], (u32)eq[i][1],
+				   (u32)eq[i][2]);
 			dd_dev_err(dd, "            %02x     %02x     %02x\n",
-				(u32)c_minus1, (u32)c0, (u32)c_plus1);
+				   (u32)c_minus1, (u32)c0, (u32)c_plus1);
 			hit_error = 1;
 		}
 	}
@@ -772,7 +771,7 @@ static int load_eq_table(struct hfi1_devdata *dd, const u8 eq[11][3], u8 fs,
 /*
  * Steps to be done after the PCIe firmware is downloaded and
  * before the SBR for the Pcie Gen3.
- * The hardware mutex is already being held.
+ * The SBus resource is already being held.
  */
 static void pcie_post_steps(struct hfi1_devdata *dd)
 {
@@ -815,8 +814,8 @@ static int trigger_sbr(struct hfi1_devdata *dd)
 	list_for_each_entry(pdev, &dev->bus->devices, bus_list)
 		if (pdev != dev) {
 			dd_dev_err(dd,
-				"%s: another device is on the same bus\n",
-				__func__);
+				   "%s: another device is on the same bus\n",
+				   __func__);
 			return -ENOTTY;
 		}
 
@@ -840,8 +839,8 @@ static void write_gasket_interrupt(struct hfi1_devdata *dd, int index,
 				   u16 code, u16 data)
 {
 	write_csr(dd, ASIC_PCIE_SD_INTRPT_LIST + (index * 8),
-	    (((u64)code << ASIC_PCIE_SD_INTRPT_LIST_INTRPT_CODE_SHIFT)
-	    |((u64)data << ASIC_PCIE_SD_INTRPT_LIST_INTRPT_DATA_SHIFT)));
+		  (((u64)code << ASIC_PCIE_SD_INTRPT_LIST_INTRPT_CODE_SHIFT) |
+		   ((u64)data << ASIC_PCIE_SD_INTRPT_LIST_INTRPT_DATA_SHIFT)));
 }
 
 /*
@@ -851,14 +850,13 @@ static void arm_gasket_logic(struct hfi1_devdata *dd)
 {
 	u64 reg;
 
-	reg = (((u64)1 << dd->hfi1_id)
-			<< ASIC_PCIE_SD_HOST_CMD_INTRPT_CMD_SHIFT)
-		| ((u64)pcie_serdes_broadcast[dd->hfi1_id]
-			<< ASIC_PCIE_SD_HOST_CMD_SBUS_RCVR_ADDR_SHIFT
-		| ASIC_PCIE_SD_HOST_CMD_SBR_MODE_SMASK
-		| ((u64)SBR_DELAY_US & ASIC_PCIE_SD_HOST_CMD_TIMER_MASK)
-			<< ASIC_PCIE_SD_HOST_CMD_TIMER_SHIFT
-		);
+	reg = (((u64)1 << dd->hfi1_id) <<
+	       ASIC_PCIE_SD_HOST_CMD_INTRPT_CMD_SHIFT) |
+	      ((u64)pcie_serdes_broadcast[dd->hfi1_id] <<
+	       ASIC_PCIE_SD_HOST_CMD_SBUS_RCVR_ADDR_SHIFT |
+	       ASIC_PCIE_SD_HOST_CMD_SBR_MODE_SMASK |
+	       ((u64)SBR_DELAY_US & ASIC_PCIE_SD_HOST_CMD_TIMER_MASK) <<
+	       ASIC_PCIE_SD_HOST_CMD_TIMER_SHIFT);
 	write_csr(dd, ASIC_PCIE_SD_HOST_CMD, reg);
 	/* read back to push the write */
 	read_csr(dd, ASIC_PCIE_SD_HOST_CMD);
@@ -946,7 +944,7 @@ static void write_xmt_margin(struct hfi1_devdata *dd, const char *fname)
  */
 int do_pcie_gen3_transition(struct hfi1_devdata *dd)
 {
-	struct pci_dev *parent;
+	struct pci_dev *parent = dd->pcidev->bus->self;
 	u64 fw_ctrl;
 	u64 reg, therm;
 	u32 reg32, fs, lf;
@@ -955,8 +953,7 @@ int do_pcie_gen3_transition(struct hfi1_devdata *dd)
 	int do_retry, retry_count = 0;
 	uint default_pset;
 	u16 target_vector, target_speed;
-	u16 lnkctl, lnkctl2, vendor;
-	u8 nsbr = 1;
+	u16 lnkctl2, vendor;
 	u8 div;
 	const u8 (*eq)[3];
 	int return_error = 0;
@@ -983,17 +980,21 @@ int do_pcie_gen3_transition(struct hfi1_devdata *dd)
 	/* if already at target speed, done (unless forced) */
 	if (dd->lbus_speed == target_speed) {
 		dd_dev_info(dd, "%s: PCIe already at gen%d, %s\n", __func__,
-			pcie_target,
-			pcie_force ? "re-doing anyway" : "skipping");
+			    pcie_target,
+			    pcie_force ? "re-doing anyway" : "skipping");
 		if (!pcie_force)
 			return 0;
 	}
 
 	/*
-	 * A0 needs an additional SBR
+	 * The driver cannot do the transition if it has no access to the
+	 * upstream component
 	 */
-	if (is_ax(dd))
-		nsbr++;
+	if (!parent) {
+		dd_dev_info(dd, "%s: No upstream, Can't do gen3 transition\n",
+			    __func__);
+		return 0;
+	}
 
 	/*
 	 * Do the Gen3 transition.  Steps are those of the PCIe Gen3
@@ -1009,10 +1010,13 @@ int do_pcie_gen3_transition(struct hfi1_devdata *dd)
 		goto done_no_mutex;
 	}
 
-	/* hold the HW mutex across the firmware download and SBR */
-	ret = acquire_hw_mutex(dd);
-	if (ret)
+	/* hold the SBus resource across the firmware download and SBR */
+	ret = acquire_chip_resource(dd, CR_SBUS, SBUS_TIMEOUT);
+	if (ret) {
+		dd_dev_err(dd, "%s: unable to acquire SBus resource\n",
+			   __func__);
 		return ret;
+	}
 
 	/* make sure thermal polling is not causing interrupts */
 	therm = read_csr(dd, ASIC_CFG_THERM_POLL_EN);
@@ -1030,8 +1034,11 @@ retry:
 	/* step 4: download PCIe Gen3 SerDes firmware */
 	dd_dev_info(dd, "%s: downloading firmware\n", __func__);
 	ret = load_pcie_firmware(dd);
-	if (ret)
+	if (ret) {
+		/* do not proceed if the firmware cannot be downloaded */
+		return_error = 1;
 		goto done;
+	}
 
 	/* step 5: set up device parameter settings */
 	dd_dev_info(dd, "%s: setting PCIe registers\n", __func__);
@@ -1091,8 +1098,10 @@ retry:
 		default_pset = DEFAULT_MCP_PSET;
 	}
 	pci_write_config_dword(dd->pcidev, PCIE_CFG_REG_PL101,
-		(fs << PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_FS_SHIFT)
-		| (lf << PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_LF_SHIFT));
+			       (fs <<
+				PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_FS_SHIFT) |
+			       (lf <<
+				PCIE_CFG_REG_PL101_GEN3_EQ_LOCAL_LF_SHIFT));
 	ret = load_eq_table(dd, eq, fs, div);
 	if (ret)
 		goto done;
@@ -1106,15 +1115,15 @@ retry:
 		pcie_pset = default_pset;
 	if (pcie_pset > 10) {	/* valid range is 0-10, inclusive */
 		dd_dev_err(dd, "%s: Invalid Eq Pset %u, setting to %d\n",
-			__func__, pcie_pset, default_pset);
+			   __func__, pcie_pset, default_pset);
 		pcie_pset = default_pset;
 	}
 	dd_dev_info(dd, "%s: using EQ Pset %u\n", __func__, pcie_pset);
 	pci_write_config_dword(dd->pcidev, PCIE_CFG_REG_PL106,
-		((1 << pcie_pset)
-			<< PCIE_CFG_REG_PL106_GEN3_EQ_PSET_REQ_VEC_SHIFT)
-		| PCIE_CFG_REG_PL106_GEN3_EQ_EVAL2MS_DISABLE_SMASK
-		| PCIE_CFG_REG_PL106_GEN3_EQ_PHASE23_EXIT_MODE_SMASK);
+			       ((1 << pcie_pset) <<
+			PCIE_CFG_REG_PL106_GEN3_EQ_PSET_REQ_VEC_SHIFT) |
+			PCIE_CFG_REG_PL106_GEN3_EQ_EVAL2MS_DISABLE_SMASK |
+			PCIE_CFG_REG_PL106_GEN3_EQ_PHASE23_EXIT_MODE_SMASK);
 
 	/*
 	 * step 5b: Do post firmware download steps via SBus
@@ -1142,11 +1151,12 @@ retry:
 	 */
 	write_xmt_margin(dd, __func__);
 
-	/* step 5e: disable active state power management (ASPM) */
+	/*
+	 * step 5e: disable active state power management (ASPM). It
+	 * will be enabled if required later
+	 */
 	dd_dev_info(dd, "%s: clearing ASPM\n", __func__);
-	pcie_capability_read_word(dd->pcidev, PCI_EXP_LNKCTL, &lnkctl);
-	lnkctl &= ~PCI_EXP_LNKCTL_ASPMC;
-	pcie_capability_write_word(dd->pcidev, PCI_EXP_LNKCTL, lnkctl);
+	aspm_hw_disable_l1(dd);
 
 	/*
 	 * step 5f: clear DirectSpeedChange
@@ -1165,16 +1175,15 @@ retry:
 	 * that it is Gen3 capable earlier.
 	 */
 	dd_dev_info(dd, "%s: setting parent target link speed\n", __func__);
-	parent = dd->pcidev->bus->self;
 	pcie_capability_read_word(parent, PCI_EXP_LNKCTL2, &lnkctl2);
 	dd_dev_info(dd, "%s: ..old link control2: 0x%x\n", __func__,
-		(u32)lnkctl2);
+		    (u32)lnkctl2);
 	/* only write to parent if target is not as high as ours */
 	if ((lnkctl2 & LNKCTL2_TARGET_LINK_SPEED_MASK) < target_vector) {
 		lnkctl2 &= ~LNKCTL2_TARGET_LINK_SPEED_MASK;
 		lnkctl2 |= target_vector;
 		dd_dev_info(dd, "%s: ..new link control2: 0x%x\n", __func__,
-			(u32)lnkctl2);
+			    (u32)lnkctl2);
 		pcie_capability_write_word(parent, PCI_EXP_LNKCTL2, lnkctl2);
 	} else {
 		dd_dev_info(dd, "%s: ..target speed is OK\n", __func__);
@@ -1183,17 +1192,17 @@ retry:
 	dd_dev_info(dd, "%s: setting target link speed\n", __func__);
 	pcie_capability_read_word(dd->pcidev, PCI_EXP_LNKCTL2, &lnkctl2);
 	dd_dev_info(dd, "%s: ..old link control2: 0x%x\n", __func__,
-		(u32)lnkctl2);
+		    (u32)lnkctl2);
 	lnkctl2 &= ~LNKCTL2_TARGET_LINK_SPEED_MASK;
 	lnkctl2 |= target_vector;
 	dd_dev_info(dd, "%s: ..new link control2: 0x%x\n", __func__,
-		(u32)lnkctl2);
+		    (u32)lnkctl2);
 	pcie_capability_write_word(dd->pcidev, PCI_EXP_LNKCTL2, lnkctl2);
 
 	/* step 5h: arm gasket logic */
 	/* hold DC in reset across the SBR */
 	write_csr(dd, CCE_DC_CTRL, CCE_DC_CTRL_DC_RESET_SMASK);
-	(void) read_csr(dd, CCE_DC_CTRL); /* DC reset hold */
+	(void)read_csr(dd, CCE_DC_CTRL); /* DC reset hold */
 	/* save firmware control across the SBR */
 	fw_ctrl = read_csr(dd, MISC_CFG_FW_CTRL);
 
@@ -1224,8 +1233,8 @@ retry:
 	ret = pci_read_config_word(dd->pcidev, PCI_VENDOR_ID, &vendor);
 	if (ret) {
 		dd_dev_info(dd,
-			"%s: read of VendorID failed after SBR, err %d\n",
-			__func__, ret);
+			    "%s: read of VendorID failed after SBR, err %d\n",
+			    __func__, ret);
 		return_error = 1;
 		goto done;
 	}
@@ -1265,8 +1274,7 @@ retry:
 	write_csr(dd, CCE_DC_CTRL, 0);
 
 	/* Set the LED off */
-	if (is_ax(dd))
-		setextled(dd, 0);
+	setextled(dd, 0);
 
 	/* check for any per-lane errors */
 	pci_read_config_dword(dd->pcidev, PCIE_CFG_SPCIE2, &reg32);
@@ -1277,8 +1285,8 @@ retry:
 			& ASIC_PCIE_SD_HOST_STATUS_FW_DNLD_STS_MASK;
 	if ((status & (1 << dd->hfi1_id)) == 0) {
 		dd_dev_err(dd,
-			"%s: gasket status 0x%x, expecting 0x%x\n",
-			__func__, status, 1 << dd->hfi1_id);
+			   "%s: gasket status 0x%x, expecting 0x%x\n",
+			   __func__, status, 1 << dd->hfi1_id);
 		ret = -EIO;
 		goto done;
 	}
@@ -1295,13 +1303,13 @@ retry:
 	/* update our link information cache */
 	update_lbus_info(dd);
 	dd_dev_info(dd, "%s: new speed and width: %s\n", __func__,
-		dd->lbus_info);
+		    dd->lbus_info);
 
 	if (dd->lbus_speed != target_speed) { /* not target */
 		/* maybe retry */
 		do_retry = retry_count < pcie_retry;
 		dd_dev_err(dd, "PCIe link speed did not switch to Gen%d%s\n",
-			pcie_target, do_retry ? ", retrying" : "");
+			   pcie_target, do_retry ? ", retrying" : "");
 		retry_count++;
 		if (do_retry) {
 			msleep(100); /* allow time to settle */
@@ -1317,7 +1325,7 @@ done:
 		dd_dev_info(dd, "%s: Re-enable therm polling\n",
 			    __func__);
 	}
-	release_hw_mutex(dd);
+	release_chip_resource(dd, CR_SBUS);
 done_no_mutex:
 	/* return no error if it is OK to be at current speed */
 	if (ret && !return_error) {
diff --git a/drivers/staging/rdma/hfi1/pio.c b/drivers/staging/rdma/hfi1/pio.c
index b51a4416312b..c6849ce9e5eb 100644
--- a/drivers/staging/rdma/hfi1/pio.c
+++ b/drivers/staging/rdma/hfi1/pio.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -101,7 +98,7 @@ void pio_send_control(struct hfi1_devdata *dd, int op)
 	/* Fall through */
 	case PSC_DATA_VL_ENABLE:
 		/* Disallow sending on VLs not enabled */
-		mask = (((~0ull)<<num_vls) & SEND_CTRL_UNSUPPORTED_VL_MASK)<<
+		mask = (((~0ull) << num_vls) & SEND_CTRL_UNSUPPORTED_VL_MASK) <<
 				SEND_CTRL_UNSUPPORTED_VL_SHIFT;
 		reg = (reg & ~SEND_CTRL_UNSUPPORTED_VL_SMASK) | mask;
 		break;
@@ -130,7 +127,7 @@ void pio_send_control(struct hfi1_devdata *dd, int op)
 	if (write) {
 		write_csr(dd, SEND_CTRL, reg);
 		if (flush)
-			(void) read_csr(dd, SEND_CTRL); /* flush write */
+			(void)read_csr(dd, SEND_CTRL); /* flush write */
 	}
 
 	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
@@ -177,8 +174,10 @@ static struct mem_pool_config sc_mem_pool_config[NUM_SC_POOLS] = {
 
 /* memory pool information, used when calculating final sizes */
 struct mem_pool_info {
-	int centipercent;	/* 100th of 1% of memory to use, -1 if blocks
-				   already set */
+	int centipercent;	/*
+				 * 100th of 1% of memory to use, -1 if blocks
+				 * already set
+				 */
 	int count;		/* count of contexts in the pool */
 	int blocks;		/* block size of the pool */
 	int size;		/* context size, in blocks */
@@ -312,7 +311,7 @@ int init_sc_pools_and_sizes(struct hfi1_devdata *dd)
 		if (i == SC_ACK) {
 			count = dd->n_krcv_queues;
 		} else if (i == SC_KERNEL) {
-			count = num_vls + 1 /* VL15 */;
+			count = (INIT_SC_PER_VL * num_vls) + 1 /* VL15 */;
 		} else if (count == SCC_PER_CPU) {
 			count = dd->num_rcv_contexts - dd->n_krcv_queues;
 		} else if (count < 0) {
@@ -509,7 +508,7 @@ static void sc_hw_free(struct hfi1_devdata *dd, u32 sw_index, u32 hw_context)
 	sci = &dd->send_contexts[sw_index];
 	if (!sci->allocated) {
 		dd_dev_err(dd, "%s: sw_index %u not allocated? hw_context %u\n",
-			__func__, sw_index, hw_context);
+			   __func__, sw_index, hw_context);
 	}
 	sci->allocated = 0;
 	dd->hw_to_sw[hw_context] = INVALID_SCI;
@@ -625,7 +624,7 @@ void sc_set_cr_threshold(struct send_context *sc, u32 new_threshold)
 				& SC(CREDIT_CTRL_THRESHOLD_MASK))
 			   << SC(CREDIT_CTRL_THRESHOLD_SHIFT));
 		write_kctxt_csr(sc->dd, sc->hw_context,
-			SC(CREDIT_CTRL), sc->credit_ctrl);
+				SC(CREDIT_CTRL), sc->credit_ctrl);
 
 		/* force a credit return on change to avoid a possible stall */
 		force_return = 1;
@@ -700,7 +699,7 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 	if (dd->flags & HFI1_FROZEN)
 		return NULL;
 
-	sc = kzalloc_node(sizeof(struct send_context), GFP_KERNEL, numa);
+	sc = kzalloc_node(sizeof(*sc), GFP_KERNEL, numa);
 	if (!sc)
 		return NULL;
 
@@ -763,9 +762,9 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 
 	/* set the default partition key */
 	write_kctxt_csr(dd, hw_context, SC(CHECK_PARTITION_KEY),
-		(DEFAULT_PKEY &
-			SC(CHECK_PARTITION_KEY_VALUE_MASK))
-		    << SC(CHECK_PARTITION_KEY_VALUE_SHIFT));
+			(SC(CHECK_PARTITION_KEY_VALUE_MASK) &
+			 DEFAULT_PKEY) <<
+			SC(CHECK_PARTITION_KEY_VALUE_SHIFT));
 
 	/* per context type checks */
 	if (type == SC_USER) {
@@ -778,8 +777,8 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 
 	/* set the send context check opcode mask and value */
 	write_kctxt_csr(dd, hw_context, SC(CHECK_OPCODE),
-		((u64)opmask << SC(CHECK_OPCODE_MASK_SHIFT)) |
-		((u64)opval << SC(CHECK_OPCODE_VALUE_SHIFT)));
+			((u64)opmask << SC(CHECK_OPCODE_MASK_SHIFT)) |
+			((u64)opval << SC(CHECK_OPCODE_VALUE_SHIFT)));
 
 	/* set up credit return */
 	reg = pa & SC(CREDIT_RETURN_ADDR_ADDRESS_SMASK);
@@ -797,7 +796,7 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 		thresh = sc_percent_to_threshold(sc, 50);
 	} else if (type == SC_USER) {
 		thresh = sc_percent_to_threshold(sc,
-				user_credit_return_threshold);
+						 user_credit_return_threshold);
 	} else { /* kernel */
 		thresh = sc_mtu_to_threshold(sc, hfi1_max_mtu, hdrqentsize);
 	}
@@ -852,7 +851,6 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 		  sc->credit_ctrl,
 		  thresh);
 
-
 	return sc;
 }
 
@@ -971,11 +969,11 @@ static void sc_wait_for_packet_egress(struct send_context *sc, int pause)
 		if (loop > 500) {
 			/* timed out - bounce the link */
 			dd_dev_err(dd,
-				"%s: context %u(%u) timeout waiting for packets to egress, remaining count %u, bouncing link\n",
-				__func__, sc->sw_index,
-				sc->hw_context, (u32)reg);
+				   "%s: context %u(%u) timeout waiting for packets to egress, remaining count %u, bouncing link\n",
+				   __func__, sc->sw_index,
+				   sc->hw_context, (u32)reg);
 			queue_work(dd->pport->hfi1_wq,
-				&dd->pport->link_bounce_work);
+				   &dd->pport->link_bounce_work);
 			break;
 		}
 		loop++;
@@ -1021,7 +1019,7 @@ int sc_restart(struct send_context *sc)
 		return -EINVAL;
 
 	dd_dev_info(dd, "restarting send context %u(%u)\n", sc->sw_index,
-		sc->hw_context);
+		    sc->hw_context);
 
 	/*
 	 * Step 1: Wait for the context to actually halt.
@@ -1036,7 +1034,7 @@ int sc_restart(struct send_context *sc)
 			break;
 		if (loop > 100) {
 			dd_dev_err(dd, "%s: context %u(%u) not halting, skipping\n",
-				__func__, sc->sw_index, sc->hw_context);
+				   __func__, sc->sw_index, sc->hw_context);
 			return -ETIME;
 		}
 		loop++;
@@ -1062,9 +1060,9 @@ int sc_restart(struct send_context *sc)
 				break;
 			if (loop > 100) {
 				dd_dev_err(dd,
-					"%s: context %u(%u) timeout waiting for PIO buffers to zero, remaining %d\n",
-					__func__, sc->sw_index,
-					sc->hw_context, count);
+					   "%s: context %u(%u) timeout waiting for PIO buffers to zero, remaining %d\n",
+					   __func__, sc->sw_index,
+					   sc->hw_context, count);
 			}
 			loop++;
 			udelay(1);
@@ -1177,18 +1175,18 @@ void pio_reset_all(struct hfi1_devdata *dd)
 	if (ret == -EIO) {
 		/* clear the error */
 		write_csr(dd, SEND_PIO_ERR_CLEAR,
-			SEND_PIO_ERR_CLEAR_PIO_INIT_SM_IN_ERR_SMASK);
+			  SEND_PIO_ERR_CLEAR_PIO_INIT_SM_IN_ERR_SMASK);
 	}
 
 	/* reset init all */
 	write_csr(dd, SEND_PIO_INIT_CTXT,
-			SEND_PIO_INIT_CTXT_PIO_ALL_CTXT_INIT_SMASK);
+		  SEND_PIO_INIT_CTXT_PIO_ALL_CTXT_INIT_SMASK);
 	udelay(2);
 	ret = pio_init_wait_progress(dd);
 	if (ret < 0) {
 		dd_dev_err(dd,
-			"PIO send context init %s while initializing all PIO blocks\n",
-			ret == -ETIMEDOUT ? "is stuck" : "had an error");
+			   "PIO send context init %s while initializing all PIO blocks\n",
+			   ret == -ETIMEDOUT ? "is stuck" : "had an error");
 	}
 }
 
@@ -1236,8 +1234,7 @@ int sc_enable(struct send_context *sc)
 	 */
 	reg = read_kctxt_csr(dd, sc->hw_context, SC(ERR_STATUS));
 	if (reg)
-		write_kctxt_csr(dd, sc->hw_context, SC(ERR_CLEAR),
-			reg);
+		write_kctxt_csr(dd, sc->hw_context, SC(ERR_CLEAR), reg);
 
 	/*
 	 * The HW PIO initialization engine can handle only one init
@@ -1295,7 +1292,7 @@ void sc_return_credits(struct send_context *sc)
 
 	/* a 0->1 transition schedules a credit return */
 	write_kctxt_csr(sc->dd, sc->hw_context, SC(CREDIT_FORCE),
-		SC(CREDIT_FORCE_FORCE_RETURN_SMASK));
+			SC(CREDIT_FORCE_FORCE_RETURN_SMASK));
 	/*
 	 * Ensure that the write is flushed and the credit return is
 	 * scheduled. We care more about the 0 -> 1 transition.
@@ -1321,7 +1318,7 @@ void sc_drop(struct send_context *sc)
 		return;
 
 	dd_dev_info(sc->dd, "%s: context %u(%u) - not implemented\n",
-			__func__, sc->sw_index, sc->hw_context);
+		    __func__, sc->sw_index, sc->hw_context);
 }
 
 /*
@@ -1346,7 +1343,7 @@ void sc_stop(struct send_context *sc, int flag)
 	wake_up(&sc->halt_wait);
 }
 
-#define BLOCK_DWORDS (PIO_BLOCK_SIZE/sizeof(u32))
+#define BLOCK_DWORDS (PIO_BLOCK_SIZE / sizeof(u32))
 #define dwords_to_blocks(x) DIV_ROUND_UP(x, BLOCK_DWORDS)
 
 /*
@@ -1430,8 +1427,10 @@ retry:
 	next = head + 1;
 	if (next >= sc->sr_size)
 		next = 0;
-	/* update the head - must be last! - the releaser can look at fields
-	   in pbuf once we move the head */
+	/*
+	 * update the head - must be last! - the releaser can look at fields
+	 * in pbuf once we move the head
+	 */
 	smp_wmb();
 	sc->sr_head = next;
 	spin_unlock_irqrestore(&sc->alloc_lock, flags);
@@ -1469,7 +1468,7 @@ void sc_add_credit_return_intr(struct send_context *sc)
 	if (sc->credit_intr_count == 0) {
 		sc->credit_ctrl |= SC(CREDIT_CTRL_CREDIT_INTR_SMASK);
 		write_kctxt_csr(sc->dd, sc->hw_context,
-			SC(CREDIT_CTRL), sc->credit_ctrl);
+				SC(CREDIT_CTRL), sc->credit_ctrl);
 	}
 	sc->credit_intr_count++;
 	spin_unlock_irqrestore(&sc->credit_ctrl_lock, flags);
@@ -1491,7 +1490,7 @@ void sc_del_credit_return_intr(struct send_context *sc)
 	if (sc->credit_intr_count == 0) {
 		sc->credit_ctrl &= ~SC(CREDIT_CTRL_CREDIT_INTR_SMASK);
 		write_kctxt_csr(sc->dd, sc->hw_context,
-			SC(CREDIT_CTRL), sc->credit_ctrl);
+				SC(CREDIT_CTRL), sc->credit_ctrl);
 	}
 	spin_unlock_irqrestore(&sc->credit_ctrl_lock, flags);
 }
@@ -1526,8 +1525,9 @@ static void sc_piobufavail(struct send_context *sc)
 	struct hfi1_devdata *dd = sc->dd;
 	struct hfi1_ibdev *dev = &dd->verbs_dev;
 	struct list_head *list;
-	struct hfi1_qp *qps[PIO_WAIT_BATCH_SIZE];
-	struct hfi1_qp *qp;
+	struct rvt_qp *qps[PIO_WAIT_BATCH_SIZE];
+	struct rvt_qp *qp;
+	struct hfi1_qp_priv *priv;
 	unsigned long flags;
 	unsigned i, n = 0;
 
@@ -1545,24 +1545,28 @@ static void sc_piobufavail(struct send_context *sc)
 		struct iowait *wait;
 
 		if (n == ARRAY_SIZE(qps))
-			goto full;
+			break;
 		wait = list_first_entry(list, struct iowait, list);
-		qp = container_of(wait, struct hfi1_qp, s_iowait);
-		list_del_init(&qp->s_iowait.list);
+		qp = iowait_to_qp(wait);
+		priv = qp->priv;
+		list_del_init(&priv->s_iowait.list);
 		/* refcount held until actual wake up */
 		qps[n++] = qp;
 	}
 	/*
-	 * Counting: only call wantpiobuf_intr() if there were waiters and they
-	 * are now all gone.
+	 * If there had been waiters and there are more
+	 * insure that we redo the force to avoid a potential hang.
 	 */
-	if (n)
+	if (n) {
 		hfi1_sc_wantpiobuf_intr(sc, 0);
-full:
+		if (!list_empty(list))
+			hfi1_sc_wantpiobuf_intr(sc, 1);
+	}
 	write_sequnlock_irqrestore(&dev->iowait_lock, flags);
 
 	for (i = 0; i < n; i++)
-		hfi1_qp_wakeup(qps[i], HFI1_S_WAIT_PIO);
+		hfi1_qp_wakeup(qps[i],
+			       RVT_S_WAIT_PIO | RVT_S_WAIT_PIO_DRAIN);
 }
 
 /* translate a send credit update to a bit code of reasons */
@@ -1661,7 +1665,7 @@ void sc_group_release_update(struct hfi1_devdata *dd, u32 hw_context)
 	sw_index = dd->hw_to_sw[hw_context];
 	if (unlikely(sw_index >= dd->num_send_contexts)) {
 		dd_dev_err(dd, "%s: invalid hw (%u) to sw (%u) mapping\n",
-			__func__, hw_context, sw_index);
+			   __func__, hw_context, sw_index);
 		goto done;
 	}
 	sc = dd->send_contexts[sw_index].sc;
@@ -1674,8 +1678,8 @@ void sc_group_release_update(struct hfi1_devdata *dd, u32 hw_context)
 		sw_index = dd->hw_to_sw[gc];
 		if (unlikely(sw_index >= dd->num_send_contexts)) {
 			dd_dev_err(dd,
-				"%s: invalid hw (%u) to sw (%u) mapping\n",
-				__func__, hw_context, sw_index);
+				   "%s: invalid hw (%u) to sw (%u) mapping\n",
+				   __func__, hw_context, sw_index);
 			continue;
 		}
 		sc_release_update(dd->send_contexts[sw_index].sc);
@@ -1684,11 +1688,217 @@ done:
 	spin_unlock(&dd->sc_lock);
 }
 
+/*
+ * pio_select_send_context_vl() - select send context
+ * @dd: devdata
+ * @selector: a spreading factor
+ * @vl: this vl
+ *
+ * This function returns a send context based on the selector and a vl.
+ * The mapping fields are protected by RCU
+ */
+struct send_context *pio_select_send_context_vl(struct hfi1_devdata *dd,
+						u32 selector, u8 vl)
+{
+	struct pio_vl_map *m;
+	struct pio_map_elem *e;
+	struct send_context *rval;
+
+	/*
+	 * NOTE This should only happen if SC->VL changed after the initial
+	 * checks on the QP/AH
+	 * Default will return VL0's send context below
+	 */
+	if (unlikely(vl >= num_vls)) {
+		rval = NULL;
+		goto done;
+	}
+
+	rcu_read_lock();
+	m = rcu_dereference(dd->pio_map);
+	if (unlikely(!m)) {
+		rcu_read_unlock();
+		return dd->vld[0].sc;
+	}
+	e = m->map[vl & m->mask];
+	rval = e->ksc[selector & e->mask];
+	rcu_read_unlock();
+
+done:
+	rval = !rval ? dd->vld[0].sc : rval;
+	return rval;
+}
+
+/*
+ * pio_select_send_context_sc() - select send context
+ * @dd: devdata
+ * @selector: a spreading factor
+ * @sc5: the 5 bit sc
+ *
+ * This function returns an send context based on the selector and an sc
+ */
+struct send_context *pio_select_send_context_sc(struct hfi1_devdata *dd,
+						u32 selector, u8 sc5)
+{
+	u8 vl = sc_to_vlt(dd, sc5);
+
+	return pio_select_send_context_vl(dd, selector, vl);
+}
+
+/*
+ * Free the indicated map struct
+ */
+static void pio_map_free(struct pio_vl_map *m)
+{
+	int i;
+
+	for (i = 0; m && i < m->actual_vls; i++)
+		kfree(m->map[i]);
+	kfree(m);
+}
+
+/*
+ * Handle RCU callback
+ */
+static void pio_map_rcu_callback(struct rcu_head *list)
+{
+	struct pio_vl_map *m = container_of(list, struct pio_vl_map, list);
+
+	pio_map_free(m);
+}
+
+/*
+ * pio_map_init - called when #vls change
+ * @dd: hfi1_devdata
+ * @port: port number
+ * @num_vls: number of vls
+ * @vl_scontexts: per vl send context mapping (optional)
+ *
+ * This routine changes the mapping based on the number of vls.
+ *
+ * vl_scontexts is used to specify a non-uniform vl/send context
+ * loading. NULL implies auto computing the loading and giving each
+ * VL an uniform distribution of send contexts per VL.
+ *
+ * The auto algorithm computers the sc_per_vl and the number of extra
+ * send contexts. Any extra send contexts are added from the last VL
+ * on down
+ *
+ * rcu locking is used here to control access to the mapping fields.
+ *
+ * If either the num_vls or num_send_contexts are non-power of 2, the
+ * array sizes in the struct pio_vl_map and the struct pio_map_elem are
+ * rounded up to the next highest power of 2 and the first entry is
+ * reused in a round robin fashion.
+ *
+ * If an error occurs the map change is not done and the mapping is not
+ * chaged.
+ *
+ */
+int pio_map_init(struct hfi1_devdata *dd, u8 port, u8 num_vls, u8 *vl_scontexts)
+{
+	int i, j;
+	int extra, sc_per_vl;
+	int scontext = 1;
+	int num_kernel_send_contexts = 0;
+	u8 lvl_scontexts[OPA_MAX_VLS];
+	struct pio_vl_map *oldmap, *newmap;
+
+	if (!vl_scontexts) {
+		/* send context 0 reserved for VL15 */
+		for (i = 1; i < dd->num_send_contexts; i++)
+			if (dd->send_contexts[i].type == SC_KERNEL)
+				num_kernel_send_contexts++;
+		/* truncate divide */
+		sc_per_vl = num_kernel_send_contexts / num_vls;
+		/* extras */
+		extra = num_kernel_send_contexts % num_vls;
+		vl_scontexts = lvl_scontexts;
+		/* add extras from last vl down */
+		for (i = num_vls - 1; i >= 0; i--, extra--)
+			vl_scontexts[i] = sc_per_vl + (extra > 0 ? 1 : 0);
+	}
+	/* build new map */
+	newmap = kzalloc(sizeof(*newmap) +
+			 roundup_pow_of_two(num_vls) *
+			 sizeof(struct pio_map_elem *),
+			 GFP_KERNEL);
+	if (!newmap)
+		goto bail;
+	newmap->actual_vls = num_vls;
+	newmap->vls = roundup_pow_of_two(num_vls);
+	newmap->mask = (1 << ilog2(newmap->vls)) - 1;
+	for (i = 0; i < newmap->vls; i++) {
+		/* save for wrap around */
+		int first_scontext = scontext;
+
+		if (i < newmap->actual_vls) {
+			int sz = roundup_pow_of_two(vl_scontexts[i]);
+
+			/* only allocate once */
+			newmap->map[i] = kzalloc(sizeof(*newmap->map[i]) +
+						 sz * sizeof(struct
+							     send_context *),
+						 GFP_KERNEL);
+			if (!newmap->map[i])
+				goto bail;
+			newmap->map[i]->mask = (1 << ilog2(sz)) - 1;
+			/* assign send contexts */
+			for (j = 0; j < sz; j++) {
+				if (dd->kernel_send_context[scontext])
+					newmap->map[i]->ksc[j] =
+					dd->kernel_send_context[scontext];
+				if (++scontext >= first_scontext +
+						  vl_scontexts[i])
+					/* wrap back to first send context */
+					scontext = first_scontext;
+			}
+		} else {
+			/* just re-use entry without allocating */
+			newmap->map[i] = newmap->map[i % num_vls];
+		}
+		scontext = first_scontext + vl_scontexts[i];
+	}
+	/* newmap in hand, save old map */
+	spin_lock_irq(&dd->pio_map_lock);
+	oldmap = rcu_dereference_protected(dd->pio_map,
+					   lockdep_is_held(&dd->pio_map_lock));
+
+	/* publish newmap */
+	rcu_assign_pointer(dd->pio_map, newmap);
+
+	spin_unlock_irq(&dd->pio_map_lock);
+	/* success, free any old map after grace period */
+	if (oldmap)
+		call_rcu(&oldmap->list, pio_map_rcu_callback);
+	return 0;
+bail:
+	/* free any partial allocation */
+	pio_map_free(newmap);
+	return -ENOMEM;
+}
+
+void free_pio_map(struct hfi1_devdata *dd)
+{
+	/* Free PIO map if allocated */
+	if (rcu_access_pointer(dd->pio_map)) {
+		spin_lock_irq(&dd->pio_map_lock);
+		pio_map_free(rcu_access_pointer(dd->pio_map));
+		RCU_INIT_POINTER(dd->pio_map, NULL);
+		spin_unlock_irq(&dd->pio_map_lock);
+		synchronize_rcu();
+	}
+	kfree(dd->kernel_send_context);
+	dd->kernel_send_context = NULL;
+}
+
 int init_pervl_scs(struct hfi1_devdata *dd)
 {
 	int i;
-	u64 mask, all_vl_mask = (u64) 0x80ff; /* VLs 0-7, 15 */
+	u64 mask, all_vl_mask = (u64)0x80ff; /* VLs 0-7, 15 */
+	u64 data_vls_mask = (u64)0x00ff; /* VLs 0-7 */
 	u32 ctxt;
+	struct hfi1_pportdata *ppd = dd->pport;
 
 	dd->vld[15].sc = sc_alloc(dd, SC_KERNEL,
 				  dd->rcd[0]->rcvhdrqentsize, dd->node);
@@ -1696,6 +1906,12 @@ int init_pervl_scs(struct hfi1_devdata *dd)
 		goto nomem;
 	hfi1_init_ctxt(dd->vld[15].sc);
 	dd->vld[15].mtu = enum_to_mtu(OPA_MTU_2048);
+
+	dd->kernel_send_context = kmalloc_node(dd->num_send_contexts *
+					sizeof(struct send_context *),
+					GFP_KERNEL, dd->node);
+	dd->kernel_send_context[0] = dd->vld[15].sc;
+
 	for (i = 0; i < num_vls; i++) {
 		/*
 		 * Since this function does not deal with a specific
@@ -1708,12 +1924,19 @@ int init_pervl_scs(struct hfi1_devdata *dd)
 					 dd->rcd[0]->rcvhdrqentsize, dd->node);
 		if (!dd->vld[i].sc)
 			goto nomem;
-
+		dd->kernel_send_context[i + 1] = dd->vld[i].sc;
 		hfi1_init_ctxt(dd->vld[i].sc);
-
 		/* non VL15 start with the max MTU */
 		dd->vld[i].mtu = hfi1_max_mtu;
 	}
+	for (i = num_vls; i < INIT_SC_PER_VL * num_vls; i++) {
+		dd->kernel_send_context[i + 1] =
+		sc_alloc(dd, SC_KERNEL, dd->rcd[0]->rcvhdrqentsize, dd->node);
+		if (!dd->kernel_send_context[i + 1])
+			goto nomem;
+		hfi1_init_ctxt(dd->kernel_send_context[i + 1]);
+	}
+
 	sc_enable(dd->vld[15].sc);
 	ctxt = dd->vld[15].sc->hw_context;
 	mask = all_vl_mask & ~(1LL << 15);
@@ -1721,17 +1944,29 @@ int init_pervl_scs(struct hfi1_devdata *dd)
 	dd_dev_info(dd,
 		    "Using send context %u(%u) for VL15\n",
 		    dd->vld[15].sc->sw_index, ctxt);
+
 	for (i = 0; i < num_vls; i++) {
 		sc_enable(dd->vld[i].sc);
 		ctxt = dd->vld[i].sc->hw_context;
-		mask = all_vl_mask & ~(1LL << i);
+		mask = all_vl_mask & ~(data_vls_mask);
 		write_kctxt_csr(dd, ctxt, SC(CHECK_VL), mask);
 	}
+	for (i = num_vls; i < INIT_SC_PER_VL * num_vls; i++) {
+		sc_enable(dd->kernel_send_context[i + 1]);
+		ctxt = dd->kernel_send_context[i + 1]->hw_context;
+		mask = all_vl_mask & ~(data_vls_mask);
+		write_kctxt_csr(dd, ctxt, SC(CHECK_VL), mask);
+	}
+
+	if (pio_map_init(dd, ppd->port - 1, num_vls, NULL))
+		goto nomem;
 	return 0;
 nomem:
 	sc_free(dd->vld[15].sc);
 	for (i = 0; i < num_vls; i++)
 		sc_free(dd->vld[i].sc);
+	for (i = num_vls; i < INIT_SC_PER_VL * num_vls; i++)
+		sc_free(dd->kernel_send_context[i + 1]);
 	return -ENOMEM;
 }
 
@@ -1769,11 +2004,11 @@ int init_credit_return(struct hfi1_devdata *dd)
 					bytes,
 					&dd->cr_base[i].pa,
 					GFP_KERNEL);
-		if (dd->cr_base[i].va == NULL) {
+		if (!dd->cr_base[i].va) {
 			set_dev_node(&dd->pcidev->dev, dd->node);
 			dd_dev_err(dd,
-				"Unable to allocate credit return DMA range for NUMA %d\n",
-				i);
+				   "Unable to allocate credit return DMA range for NUMA %d\n",
+				   i);
 			ret = -ENOMEM;
 			goto done;
 		}
@@ -1797,10 +2032,10 @@ void free_credit_return(struct hfi1_devdata *dd)
 	for (i = 0; i < num_numa; i++) {
 		if (dd->cr_base[i].va) {
 			dma_free_coherent(&dd->pcidev->dev,
-				TXE_NUM_CONTEXTS
-					* sizeof(struct credit_return),
-				dd->cr_base[i].va,
-				dd->cr_base[i].pa);
+					  TXE_NUM_CONTEXTS *
+					  sizeof(struct credit_return),
+					  dd->cr_base[i].va,
+					  dd->cr_base[i].pa);
 		}
 	}
 	kfree(dd->cr_base);
diff --git a/drivers/staging/rdma/hfi1/pio.h b/drivers/staging/rdma/hfi1/pio.h
index 53d3e0a79375..0026976ce4f6 100644
--- a/drivers/staging/rdma/hfi1/pio.h
+++ b/drivers/staging/rdma/hfi1/pio.h
@@ -1,14 +1,13 @@
 #ifndef _PIO_H
 #define _PIO_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -50,7 +47,6 @@
  *
  */
 
-
 /* send context types */
 #define SC_KERNEL 0
 #define SC_ACK    1
@@ -106,6 +102,7 @@ struct send_context {
 	struct hfi1_devdata *dd;		/* device */
 	void __iomem *base_addr;	/* start of PIO memory */
 	union pio_shadow_ring *sr;	/* shadow ring */
+
 	volatile __le64 *hw_free;	/* HW free counter */
 	struct work_struct halt_work;	/* halted context work queue entry */
 	unsigned long flags;		/* flags */
@@ -165,6 +162,112 @@ struct sc_config_sizes {
 	short int count;
 };
 
+/*
+ * The diagram below details the relationship of the mapping structures
+ *
+ * Since the mapping now allows for non-uniform send contexts per vl, the
+ * number of send contexts for a vl is either the vl_scontexts[vl] or
+ * a computation based on num_kernel_send_contexts/num_vls:
+ *
+ * For example:
+ * nactual = vl_scontexts ? vl_scontexts[vl] : num_kernel_send_contexts/num_vls
+ *
+ * n = roundup to next highest power of 2 using nactual
+ *
+ * In the case where there are num_kernel_send_contexts/num_vls doesn't divide
+ * evenly, the extras are added from the last vl downward.
+ *
+ * For the case where n > nactual, the send contexts are assigned
+ * in a round robin fashion wrapping back to the first send context
+ * for a particular vl.
+ *
+ *               dd->pio_map
+ *                    |                                   pio_map_elem[0]
+ *                    |                                +--------------------+
+ *                    v                                |       mask         |
+ *               pio_vl_map                            |--------------------|
+ *      +--------------------------+                   | ksc[0] -> sc 1     |
+ *      |    list (RCU)            |                   |--------------------|
+ *      |--------------------------|                 ->| ksc[1] -> sc 2     |
+ *      |    mask                  |              --/  |--------------------|
+ *      |--------------------------|            -/     |        *           |
+ *      |    actual_vls (max 8)    |          -/       |--------------------|
+ *      |--------------------------|       --/         | ksc[n] -> sc n     |
+ *      |    vls (max 8)           |     -/            +--------------------+
+ *      |--------------------------|  --/
+ *      |    map[0]                |-/
+ *      |--------------------------|                   +--------------------+
+ *      |    map[1]                |---                |       mask         |
+ *      |--------------------------|   \----           |--------------------|
+ *      |           *              |        \--        | ksc[0] -> sc 1+n   |
+ *      |           *              |           \----   |--------------------|
+ *      |           *              |                \->| ksc[1] -> sc 2+n   |
+ *      |--------------------------|                   |--------------------|
+ *      |   map[vls - 1]           |-                  |         *          |
+ *      +--------------------------+ \-                |--------------------|
+ *                                     \-              | ksc[m] -> sc m+n   |
+ *                                       \             +--------------------+
+ *                                        \-
+ *                                          \
+ *                                           \-        +--------------------+
+ *                                             \-      |       mask         |
+ *                                               \     |--------------------|
+ *                                                \-   | ksc[0] -> sc 1+m+n |
+ *                                                  \- |--------------------|
+ *                                                    >| ksc[1] -> sc 2+m+n |
+ *                                                     |--------------------|
+ *                                                     |         *          |
+ *                                                     |--------------------|
+ *                                                     | ksc[o] -> sc o+m+n |
+ *                                                     +--------------------+
+ *
+ */
+
+/* Initial number of send contexts per VL */
+#define INIT_SC_PER_VL 2
+
+/*
+ * struct pio_map_elem - mapping for a vl
+ * @mask - selector mask
+ * @ksc - array of kernel send contexts for this vl
+ *
+ * The mask is used to "mod" the selector to
+ * produce index into the trailing array of
+ * kscs
+ */
+struct pio_map_elem {
+	u32 mask;
+	struct send_context *ksc[0];
+};
+
+/*
+ * struct pio_vl_map - mapping for a vl
+ * @list - rcu head for free callback
+ * @mask - vl mask to "mod" the vl to produce an index to map array
+ * @actual_vls - number of vls
+ * @vls - numbers of vls rounded to next power of 2
+ * @map - array of pio_map_elem entries
+ *
+ * This is the parent mapping structure. The trailing members of the
+ * struct point to pio_map_elem entries, which in turn point to an
+ * array of kscs for that vl.
+ */
+struct pio_vl_map {
+	struct rcu_head list;
+	u32 mask;
+	u8 actual_vls;
+	u8 vls;
+	struct pio_map_elem *map[0];
+};
+
+int pio_map_init(struct hfi1_devdata *dd, u8 port, u8 num_vls,
+		 u8 *vl_scontexts);
+void free_pio_map(struct hfi1_devdata *dd);
+struct send_context *pio_select_send_context_vl(struct hfi1_devdata *dd,
+						u32 selector, u8 vl);
+struct send_context *pio_select_send_context_sc(struct hfi1_devdata *dd,
+						u32 selector, u8 sc5);
+
 /* send context functions */
 int init_credit_return(struct hfi1_devdata *dd);
 void free_credit_return(struct hfi1_devdata *dd);
@@ -183,7 +286,7 @@ void sc_flush(struct send_context *sc);
 void sc_drop(struct send_context *sc);
 void sc_stop(struct send_context *sc, int bit);
 struct pio_buf *sc_buffer_alloc(struct send_context *sc, u32 dw_len,
-			pio_release_cb cb, void *arg);
+				pio_release_cb cb, void *arg);
 void sc_release_update(struct send_context *sc);
 void sc_return_credits(struct send_context *sc);
 void sc_group_release_update(struct hfi1_devdata *dd, u32 hw_context);
@@ -212,12 +315,11 @@ void pio_kernel_unfreeze(struct hfi1_devdata *dd);
 void __cm_reset(struct hfi1_devdata *dd, u64 sendctrl);
 void pio_send_control(struct hfi1_devdata *dd, int op);
 
-
 /* PIO copy routines */
 void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
 	      const void *from, size_t count);
 void seg_pio_copy_start(struct pio_buf *pbuf, u64 pbc,
-					const void *from, size_t nbytes);
+			const void *from, size_t nbytes);
 void seg_pio_copy_mid(struct pio_buf *pbuf, const void *from, size_t nbytes);
 void seg_pio_copy_end(struct pio_buf *pbuf);
 
diff --git a/drivers/staging/rdma/hfi1/pio_copy.c b/drivers/staging/rdma/hfi1/pio_copy.c
index 64bef6c26653..8c25e1b58849 100644
--- a/drivers/staging/rdma/hfi1/pio_copy.c
+++ b/drivers/staging/rdma/hfi1/pio_copy.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -52,9 +49,9 @@
 
 /* additive distance between non-SOP and SOP space */
 #define SOP_DISTANCE (TXE_PIO_SIZE / 2)
-#define PIO_BLOCK_MASK (PIO_BLOCK_SIZE-1)
+#define PIO_BLOCK_MASK (PIO_BLOCK_SIZE - 1)
 /* number of QUADWORDs in a block */
-#define PIO_BLOCK_QWS (PIO_BLOCK_SIZE/sizeof(u64))
+#define PIO_BLOCK_QWS (PIO_BLOCK_SIZE / sizeof(u64))
 
 /**
  * pio_copy - copy data block to MMIO space
@@ -83,11 +80,13 @@ void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
 	dest += sizeof(u64);
 
 	/* calculate where the QWORD data ends - in SOP=1 space */
-	dend = dest + ((count>>1) * sizeof(u64));
+	dend = dest + ((count >> 1) * sizeof(u64));
 
 	if (dend < send) {
-		/* all QWORD data is within the SOP block, does *not*
-		   reach the end of the SOP block */
+		/*
+		 * all QWORD data is within the SOP block, does *not*
+		 * reach the end of the SOP block
+		 */
 
 		while (dest < dend) {
 			writeq(*(u64 *)from, dest);
@@ -152,8 +151,10 @@ void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
 		writeq(val.val64, dest);
 		dest += sizeof(u64);
 	}
-	/* fill in rest of block, no need to check pbuf->end
-	   as we only wrap on a block boundary */
+	/*
+	 * fill in rest of block, no need to check pbuf->end
+	 * as we only wrap on a block boundary
+	 */
 	while (((unsigned long)dest & PIO_BLOCK_MASK) != 0) {
 		writeq(0, dest);
 		dest += sizeof(u64);
@@ -177,7 +178,7 @@ void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
  * "zero" shift - bit shift used to zero out upper bytes.  Input is
  * the count of LSB bytes to preserve.
  */
-#define zshift(x) (8 * (8-(x)))
+#define zshift(x) (8 * (8 - (x)))
 
 /*
  * "merge" shift - bit shift used to merge with carry bytes.  Input is
@@ -196,7 +197,7 @@ void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
  * o nbytes must not span a QW boundary
  */
 static inline void read_low_bytes(struct pio_buf *pbuf, const void *from,
-							unsigned int nbytes)
+				  unsigned int nbytes)
 {
 	unsigned long off;
 
@@ -223,7 +224,7 @@ static inline void read_low_bytes(struct pio_buf *pbuf, const void *from,
  * o nbytes may span a QW boundary
  */
 static inline void read_extra_bytes(struct pio_buf *pbuf,
-					const void *from, unsigned int nbytes)
+				    const void *from, unsigned int nbytes)
 {
 	unsigned long off = (unsigned long)from & 0x7;
 	unsigned int room, xbytes;
@@ -244,7 +245,7 @@ static inline void read_extra_bytes(struct pio_buf *pbuf,
 		pbuf->carry.val64 |= (((*(u64 *)from)
 					>> mshift(off))
 					<< zshift(xbytes))
-					>> zshift(xbytes+pbuf->carry_bytes);
+					>> zshift(xbytes + pbuf->carry_bytes);
 		off = 0;
 		pbuf->carry_bytes += xbytes;
 		nbytes -= xbytes;
@@ -362,7 +363,7 @@ static inline void jcopy(u8 *dest, const u8 *src, u32 n)
  * o from may _not_ be u64 aligned.
  */
 static inline void read_low_bytes(struct pio_buf *pbuf, const void *from,
-							unsigned int nbytes)
+				  unsigned int nbytes)
 {
 	jcopy(&pbuf->carry.val8[0], from, nbytes);
 	pbuf->carry_bytes = nbytes;
@@ -377,7 +378,7 @@ static inline void read_low_bytes(struct pio_buf *pbuf, const void *from,
  * o nbytes may span a QW boundary
  */
 static inline void read_extra_bytes(struct pio_buf *pbuf,
-					const void *from, unsigned int nbytes)
+				    const void *from, unsigned int nbytes)
 {
 	jcopy(&pbuf->carry.val8[pbuf->carry_bytes], from, nbytes);
 	pbuf->carry_bytes += nbytes;
@@ -411,7 +412,7 @@ static inline void merge_write8(
 
 	jcopy(&pbuf->carry.val8[pbuf->carry_bytes], src, remainder);
 	writeq(pbuf->carry.val64, dest);
-	jcopy(&pbuf->carry.val8[0], src+remainder, pbuf->carry_bytes);
+	jcopy(&pbuf->carry.val8[0], src + remainder, pbuf->carry_bytes);
 }
 
 /*
@@ -433,7 +434,7 @@ static inline int carry_write8(struct pio_buf *pbuf, void *dest)
 		u64 zero = 0;
 
 		jcopy(&pbuf->carry.val8[pbuf->carry_bytes], (u8 *)&zero,
-						8 - pbuf->carry_bytes);
+		      8 - pbuf->carry_bytes);
 		writeq(pbuf->carry.val64, dest);
 		return 1;
 	}
@@ -453,7 +454,7 @@ static inline int carry_write8(struct pio_buf *pbuf, void *dest)
  * @nbytes: bytes to copy
  */
 void seg_pio_copy_start(struct pio_buf *pbuf, u64 pbc,
-				const void *from, size_t nbytes)
+			const void *from, size_t nbytes)
 {
 	void __iomem *dest = pbuf->start + SOP_DISTANCE;
 	void __iomem *send = dest + PIO_BLOCK_SIZE;
@@ -463,11 +464,13 @@ void seg_pio_copy_start(struct pio_buf *pbuf, u64 pbc,
 	dest += sizeof(u64);
 
 	/* calculate where the QWORD data ends - in SOP=1 space */
-	dend = dest + ((nbytes>>3) * sizeof(u64));
+	dend = dest + ((nbytes >> 3) * sizeof(u64));
 
 	if (dend < send) {
-		/* all QWORD data is within the SOP block, does *not*
-		   reach the end of the SOP block */
+		/*
+		 * all QWORD data is within the SOP block, does *not*
+		 * reach the end of the SOP block
+		 */
 
 		while (dest < dend) {
 			writeq(*(u64 *)from, dest);
@@ -562,8 +565,10 @@ static void mid_copy_mix(struct pio_buf *pbuf, const void *from, size_t nbytes)
 		void __iomem *send;		/* SOP end */
 		void __iomem *xend;
 
-		/* calculate the end of data or end of block, whichever
-		   comes first */
+		/*
+		 * calculate the end of data or end of block, whichever
+		 * comes first
+		 */
 		send = pbuf->start + PIO_BLOCK_SIZE;
 		xend = min(send, dend);
 
@@ -639,13 +644,13 @@ static void mid_copy_mix(struct pio_buf *pbuf, const void *from, size_t nbytes)
  * Must handle nbytes < 8.
  */
 static void mid_copy_straight(struct pio_buf *pbuf,
-						const void *from, size_t nbytes)
+			      const void *from, size_t nbytes)
 {
 	void __iomem *dest = pbuf->start + (pbuf->qw_written * sizeof(u64));
 	void __iomem *dend;			/* 8-byte data end */
 
 	/* calculate 8-byte data end */
-	dend = dest + ((nbytes>>3) * sizeof(u64));
+	dend = dest + ((nbytes >> 3) * sizeof(u64));
 
 	if (pbuf->qw_written < PIO_BLOCK_QWS) {
 		/*
@@ -656,8 +661,10 @@ static void mid_copy_straight(struct pio_buf *pbuf,
 		void __iomem *send;		/* SOP end */
 		void __iomem *xend;
 
-		/* calculate the end of data or end of block, whichever
-		   comes first */
+		/*
+		 * calculate the end of data or end of block, whichever
+		 * comes first
+		 */
 		send = pbuf->start + PIO_BLOCK_SIZE;
 		xend = min(send, dend);
 
@@ -713,7 +720,7 @@ static void mid_copy_straight(struct pio_buf *pbuf,
 	/* we know carry_bytes was zero on entry to this routine */
 	read_low_bytes(pbuf, from, nbytes & 0x7);
 
-	pbuf->qw_written += nbytes>>3;
+	pbuf->qw_written += nbytes >> 3;
 }
 
 /*
diff --git a/drivers/staging/rdma/hfi1/platform.c b/drivers/staging/rdma/hfi1/platform.c
new file mode 100644
index 000000000000..0a1d074583e4
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/platform.c
@@ -0,0 +1,893 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include "hfi.h"
+#include "efivar.h"
+
+void get_platform_config(struct hfi1_devdata *dd)
+{
+	int ret = 0;
+	unsigned long size = 0;
+	u8 *temp_platform_config = NULL;
+
+	ret = read_hfi1_efi_var(dd, "configuration", &size,
+				(void **)&temp_platform_config);
+	if (ret) {
+		dd_dev_info(dd,
+			    "%s: Failed to get platform config from UEFI, falling back to request firmware\n",
+			    __func__);
+		/* fall back to request firmware */
+		platform_config_load = 1;
+		goto bail;
+	}
+
+	dd->platform_config.data = temp_platform_config;
+	dd->platform_config.size = size;
+
+bail:
+	/* exit */;
+}
+
+void free_platform_config(struct hfi1_devdata *dd)
+{
+	if (!platform_config_load) {
+		/*
+		 * was loaded from EFI, release memory
+		 * allocated by read_efi_var
+		 */
+		kfree(dd->platform_config.data);
+	}
+	/*
+	 * else do nothing, dispose_firmware will release
+	 * struct firmware platform_config on driver exit
+	 */
+}
+
+int set_qsfp_tx(struct hfi1_pportdata *ppd, int on)
+{
+	u8 tx_ctrl_byte = on ? 0x0 : 0xF;
+	int ret = 0;
+
+	ret = qsfp_write(ppd, ppd->dd->hfi1_id, QSFP_TX_CTRL_BYTE_OFFS,
+			 &tx_ctrl_byte, 1);
+	/* we expected 1, so consider 0 an error */
+	if (ret == 0)
+		ret = -EIO;
+	else if (ret == 1)
+		ret = 0;
+	return ret;
+}
+
+static int qual_power(struct hfi1_pportdata *ppd)
+{
+	u32 cable_power_class = 0, power_class_max = 0;
+	u8 *cache = ppd->qsfp_info.cache;
+	int ret = 0;
+
+	ret = get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_SYSTEM_TABLE, 0,
+		SYSTEM_TABLE_QSFP_POWER_CLASS_MAX, &power_class_max, 4);
+	if (ret)
+		return ret;
+
+	if (QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]) != 4)
+		cable_power_class = QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]);
+	else
+		cable_power_class = QSFP_PWR(cache[QSFP_MOD_PWR_OFFS]);
+
+	if (cable_power_class <= 3 && cable_power_class > (power_class_max - 1))
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY);
+	else if (cable_power_class > 4 && cable_power_class > (power_class_max))
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY);
+	/*
+	 * cable_power_class will never have value 4 as this simply
+	 * means the high power settings are unused
+	 */
+
+	if (ppd->offline_disabled_reason ==
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_POWER_POLICY)) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: Port disabled due to system power restrictions\n",
+			__func__);
+		ret = -EPERM;
+	}
+	return ret;
+}
+
+static int qual_bitrate(struct hfi1_pportdata *ppd)
+{
+	u16 lss = ppd->link_speed_supported, lse = ppd->link_speed_enabled;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	if ((lss & OPA_LINK_SPEED_25G) && (lse & OPA_LINK_SPEED_25G) &&
+	    cache[QSFP_NOM_BIT_RATE_250_OFFS] < 0x64)
+		ppd->offline_disabled_reason =
+			   HFI1_ODR_MASK(OPA_LINKDOWN_REASON_LINKSPEED_POLICY);
+
+	if ((lss & OPA_LINK_SPEED_12_5G) && (lse & OPA_LINK_SPEED_12_5G) &&
+	    cache[QSFP_NOM_BIT_RATE_100_OFFS] < 0x7D)
+		ppd->offline_disabled_reason =
+			   HFI1_ODR_MASK(OPA_LINKDOWN_REASON_LINKSPEED_POLICY);
+
+	if (ppd->offline_disabled_reason ==
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_LINKSPEED_POLICY)) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: Cable failed bitrate check, disabling port\n",
+			__func__);
+		return -EPERM;
+	}
+	return 0;
+}
+
+static int set_qsfp_high_power(struct hfi1_pportdata *ppd)
+{
+	u8 cable_power_class = 0, power_ctrl_byte = 0;
+	u8 *cache = ppd->qsfp_info.cache;
+	int ret;
+
+	if (QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]) != 4)
+		cable_power_class = QSFP_HIGH_PWR(cache[QSFP_MOD_PWR_OFFS]);
+	else
+		cable_power_class = QSFP_PWR(cache[QSFP_MOD_PWR_OFFS]);
+
+	if (cable_power_class) {
+		power_ctrl_byte = cache[QSFP_PWR_CTRL_BYTE_OFFS];
+
+		power_ctrl_byte |= 1;
+		power_ctrl_byte &= ~(0x2);
+
+		ret = qsfp_write(ppd, ppd->dd->hfi1_id,
+				 QSFP_PWR_CTRL_BYTE_OFFS,
+				 &power_ctrl_byte, 1);
+		if (ret != 1)
+			return -EIO;
+
+		if (cable_power_class > 3) {
+			/* > power class 4*/
+			power_ctrl_byte |= (1 << 2);
+			ret = qsfp_write(ppd, ppd->dd->hfi1_id,
+					 QSFP_PWR_CTRL_BYTE_OFFS,
+					 &power_ctrl_byte, 1);
+			if (ret != 1)
+				return -EIO;
+		}
+
+		/* SFF 8679 rev 1.7 LPMode Deassert time */
+		msleep(300);
+	}
+	return 0;
+}
+
+static void apply_rx_cdr(struct hfi1_pportdata *ppd,
+			 u32 rx_preset_index,
+			 u8 *cdr_ctrl_byte)
+{
+	u32 rx_preset;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	if (!((cache[QSFP_MOD_PWR_OFFS] & 0x4) &&
+	      (cache[QSFP_CDR_INFO_OFFS] & 0x40)))
+		return;
+
+	/* rx_preset preset to zero to catch error */
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_RX_PRESET_TABLE,
+		rx_preset_index, RX_PRESET_TABLE_QSFP_RX_CDR_APPLY,
+		&rx_preset, 4);
+
+	if (!rx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: RX_CDR_APPLY is set to disabled\n",
+			__func__);
+		return;
+	}
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_RX_PRESET_TABLE,
+		rx_preset_index, RX_PRESET_TABLE_QSFP_RX_CDR,
+		&rx_preset, 4);
+
+	/* Expand cdr setting to all 4 lanes */
+	rx_preset = (rx_preset | (rx_preset << 1) |
+			(rx_preset << 2) | (rx_preset << 3));
+
+	if (rx_preset) {
+		*cdr_ctrl_byte |= rx_preset;
+	} else {
+		*cdr_ctrl_byte &= rx_preset;
+		/* Preserve current TX CDR status */
+		*cdr_ctrl_byte |= (cache[QSFP_CDR_CTRL_BYTE_OFFS] & 0xF0);
+	}
+}
+
+static void apply_tx_cdr(struct hfi1_pportdata *ppd,
+			 u32 tx_preset_index,
+			 u8 *ctr_ctrl_byte)
+{
+	u32 tx_preset;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	if (!((cache[QSFP_MOD_PWR_OFFS] & 0x8) &&
+	      (cache[QSFP_CDR_INFO_OFFS] & 0x80)))
+		return;
+
+	get_platform_config_field(
+		ppd->dd,
+		PLATFORM_CONFIG_TX_PRESET_TABLE, tx_preset_index,
+		TX_PRESET_TABLE_QSFP_TX_CDR_APPLY, &tx_preset, 4);
+
+	if (!tx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: TX_CDR_APPLY is set to disabled\n",
+			__func__);
+		return;
+	}
+	get_platform_config_field(
+		ppd->dd,
+		PLATFORM_CONFIG_TX_PRESET_TABLE,
+		tx_preset_index,
+		TX_PRESET_TABLE_QSFP_TX_CDR, &tx_preset, 4);
+
+	/* Expand cdr setting to all 4 lanes */
+	tx_preset = (tx_preset | (tx_preset << 1) |
+			(tx_preset << 2) | (tx_preset << 3));
+
+	if (tx_preset)
+		*ctr_ctrl_byte |= (tx_preset << 4);
+	else
+		/* Preserve current/determined RX CDR status */
+		*ctr_ctrl_byte &= ((tx_preset << 4) | 0xF);
+}
+
+static void apply_cdr_settings(
+		struct hfi1_pportdata *ppd, u32 rx_preset_index,
+		u32 tx_preset_index)
+{
+	u8 *cache = ppd->qsfp_info.cache;
+	u8 cdr_ctrl_byte = cache[QSFP_CDR_CTRL_BYTE_OFFS];
+
+	apply_rx_cdr(ppd, rx_preset_index, &cdr_ctrl_byte);
+
+	apply_tx_cdr(ppd, tx_preset_index, &cdr_ctrl_byte);
+
+	qsfp_write(ppd, ppd->dd->hfi1_id, QSFP_CDR_CTRL_BYTE_OFFS,
+		   &cdr_ctrl_byte, 1);
+}
+
+static void apply_tx_eq_auto(struct hfi1_pportdata *ppd)
+{
+	u8 *cache = ppd->qsfp_info.cache;
+	u8 tx_eq;
+
+	if (!(cache[QSFP_EQ_INFO_OFFS] & 0x8))
+		return;
+	/* Disable adaptive TX EQ if present */
+	tx_eq = cache[(128 * 3) + 241];
+	tx_eq &= 0xF0;
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 241, &tx_eq, 1);
+}
+
+static void apply_tx_eq_prog(struct hfi1_pportdata *ppd, u32 tx_preset_index)
+{
+	u8 *cache = ppd->qsfp_info.cache;
+	u32 tx_preset;
+	u8 tx_eq;
+
+	if (!(cache[QSFP_EQ_INFO_OFFS] & 0x4))
+		return;
+
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_TX_PRESET_TABLE,
+		tx_preset_index, TX_PRESET_TABLE_QSFP_TX_EQ_APPLY,
+		&tx_preset, 4);
+	if (!tx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: TX_EQ_APPLY is set to disabled\n",
+			__func__);
+		return;
+	}
+	get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_TX_PRESET_TABLE,
+			tx_preset_index, TX_PRESET_TABLE_QSFP_TX_EQ,
+			&tx_preset, 4);
+
+	if (((cache[(128 * 3) + 224] & 0xF0) >> 4) < tx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: TX EQ %x unsupported\n",
+			__func__, tx_preset);
+
+		dd_dev_info(
+			ppd->dd,
+			"%s: Applying EQ %x\n",
+			__func__, cache[608] & 0xF0);
+
+		tx_preset = (cache[608] & 0xF0) >> 4;
+	}
+
+	tx_eq = tx_preset | (tx_preset << 4);
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 234, &tx_eq, 1);
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 235, &tx_eq, 1);
+}
+
+static void apply_rx_eq_emp(struct hfi1_pportdata *ppd, u32 rx_preset_index)
+{
+	u32 rx_preset;
+	u8 rx_eq, *cache = ppd->qsfp_info.cache;
+
+	if (!(cache[QSFP_EQ_INFO_OFFS] & 0x2))
+		return;
+	get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_RX_PRESET_TABLE,
+			rx_preset_index, RX_PRESET_TABLE_QSFP_RX_EMP_APPLY,
+			&rx_preset, 4);
+
+	if (!rx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: RX_EMP_APPLY is set to disabled\n",
+			__func__);
+		return;
+	}
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_RX_PRESET_TABLE,
+		rx_preset_index, RX_PRESET_TABLE_QSFP_RX_EMP,
+		&rx_preset, 4);
+
+	if ((cache[(128 * 3) + 224] & 0xF) < rx_preset) {
+		dd_dev_info(
+			ppd->dd,
+			"%s: Requested RX EMP %x\n",
+			__func__, rx_preset);
+
+		dd_dev_info(
+			ppd->dd,
+			"%s: Applying supported EMP %x\n",
+			__func__, cache[608] & 0xF);
+
+		rx_preset = cache[608] & 0xF;
+	}
+
+	rx_eq = rx_preset | (rx_preset << 4);
+
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 236, &rx_eq, 1);
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 237, &rx_eq, 1);
+}
+
+static void apply_eq_settings(struct hfi1_pportdata *ppd,
+			      u32 rx_preset_index, u32 tx_preset_index)
+{
+	u8 *cache = ppd->qsfp_info.cache;
+
+	/* no point going on w/o a page 3 */
+	if (cache[2] & 4) {
+		dd_dev_info(ppd->dd,
+			    "%s: Upper page 03 not present\n",
+			    __func__);
+		return;
+	}
+
+	apply_tx_eq_auto(ppd);
+
+	apply_tx_eq_prog(ppd, tx_preset_index);
+
+	apply_rx_eq_emp(ppd, rx_preset_index);
+}
+
+static void apply_rx_amplitude_settings(
+		struct hfi1_pportdata *ppd, u32 rx_preset_index,
+		u32 tx_preset_index)
+{
+	u32 rx_preset;
+	u8 rx_amp = 0, i = 0, preferred = 0, *cache = ppd->qsfp_info.cache;
+
+	/* no point going on w/o a page 3 */
+	if (cache[2] & 4) {
+		dd_dev_info(ppd->dd,
+			    "%s: Upper page 03 not present\n",
+			    __func__);
+		return;
+	}
+	if (!(cache[QSFP_EQ_INFO_OFFS] & 0x1)) {
+		dd_dev_info(ppd->dd,
+			    "%s: RX_AMP_APPLY is set to disabled\n",
+			    __func__);
+		return;
+	}
+
+	get_platform_config_field(ppd->dd,
+				  PLATFORM_CONFIG_RX_PRESET_TABLE,
+				  rx_preset_index,
+				  RX_PRESET_TABLE_QSFP_RX_AMP_APPLY,
+				  &rx_preset, 4);
+
+	if (!rx_preset) {
+		dd_dev_info(ppd->dd,
+			    "%s: RX_AMP_APPLY is set to disabled\n",
+			    __func__);
+		return;
+	}
+	get_platform_config_field(ppd->dd,
+				  PLATFORM_CONFIG_RX_PRESET_TABLE,
+				  rx_preset_index,
+				  RX_PRESET_TABLE_QSFP_RX_AMP,
+				  &rx_preset, 4);
+
+	dd_dev_info(ppd->dd,
+		    "%s: Requested RX AMP %x\n",
+		    __func__,
+		    rx_preset);
+
+	for (i = 0; i < 4; i++) {
+		if (cache[(128 * 3) + 225] & (1 << i)) {
+			preferred = i;
+			if (preferred == rx_preset)
+				break;
+		}
+	}
+
+	/*
+	 * Verify that preferred RX amplitude is not just a
+	 * fall through of the default
+	 */
+	if (!preferred && !(cache[(128 * 3) + 225] & 0x1)) {
+		dd_dev_info(ppd->dd, "No supported RX AMP, not applying\n");
+		return;
+	}
+
+	dd_dev_info(ppd->dd,
+		    "%s: Applying RX AMP %x\n", __func__, preferred);
+
+	rx_amp = preferred | (preferred << 4);
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 238, &rx_amp, 1);
+	qsfp_write(ppd, ppd->dd->hfi1_id, (256 * 3) + 239, &rx_amp, 1);
+}
+
+#define OPA_INVALID_INDEX 0xFFF
+
+static void apply_tx_lanes(struct hfi1_pportdata *ppd, u8 field_id,
+			   u32 config_data, const char *message)
+{
+	u8 i;
+	int ret = HCMD_SUCCESS;
+
+	for (i = 0; i < 4; i++) {
+		ret = load_8051_config(ppd->dd, field_id, i, config_data);
+		if (ret != HCMD_SUCCESS) {
+			dd_dev_err(
+				ppd->dd,
+				"%s: %s for lane %u failed\n",
+				message, __func__, i);
+		}
+	}
+}
+
+static void apply_tunings(
+		struct hfi1_pportdata *ppd, u32 tx_preset_index,
+		u8 tuning_method, u32 total_atten, u8 limiting_active)
+{
+	int ret = 0;
+	u32 config_data = 0, tx_preset = 0;
+	u8 precur = 0, attn = 0, postcur = 0, external_device_config = 0;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	/* Enable external device config if channel is limiting active */
+	read_8051_config(ppd->dd, LINK_OPTIMIZATION_SETTINGS,
+			 GENERAL_CONFIG, &config_data);
+	config_data |= limiting_active;
+	ret = load_8051_config(ppd->dd, LINK_OPTIMIZATION_SETTINGS,
+			       GENERAL_CONFIG, config_data);
+	if (ret != HCMD_SUCCESS)
+		dd_dev_err(
+			ppd->dd,
+			"%s: Failed to set enable external device config\n",
+			__func__);
+
+	config_data = 0; /* re-init  */
+	/* Pass tuning method to 8051 */
+	read_8051_config(ppd->dd, LINK_TUNING_PARAMETERS, GENERAL_CONFIG,
+			 &config_data);
+	config_data |= tuning_method;
+	ret = load_8051_config(ppd->dd, LINK_TUNING_PARAMETERS, GENERAL_CONFIG,
+			       config_data);
+	if (ret != HCMD_SUCCESS)
+		dd_dev_err(ppd->dd, "%s: Failed to set tuning method\n",
+			   __func__);
+
+	/* Set same channel loss for both TX and RX */
+	config_data = 0 | (total_atten << 16) | (total_atten << 24);
+	apply_tx_lanes(ppd, CHANNEL_LOSS_SETTINGS, config_data,
+		       "Setting channel loss");
+
+	/* Inform 8051 of cable capabilities */
+	if (ppd->qsfp_info.cache_valid) {
+		external_device_config =
+			((cache[QSFP_MOD_PWR_OFFS] & 0x4) << 3) |
+			((cache[QSFP_MOD_PWR_OFFS] & 0x8) << 2) |
+			((cache[QSFP_EQ_INFO_OFFS] & 0x2) << 1) |
+			(cache[QSFP_EQ_INFO_OFFS] & 0x4);
+		ret = read_8051_config(ppd->dd, DC_HOST_COMM_SETTINGS,
+				       GENERAL_CONFIG, &config_data);
+		/* Clear, then set the external device config field */
+		config_data &= ~(0xFF << 24);
+		config_data |= (external_device_config << 24);
+		ret = load_8051_config(ppd->dd, DC_HOST_COMM_SETTINGS,
+				       GENERAL_CONFIG, config_data);
+		if (ret != HCMD_SUCCESS)
+			dd_dev_info(ppd->dd,
+				    "%s: Failed set ext device config params\n",
+				    __func__);
+	}
+
+	if (tx_preset_index == OPA_INVALID_INDEX) {
+		if (ppd->port_type == PORT_TYPE_QSFP && limiting_active)
+			dd_dev_info(ppd->dd, "%s: Invalid Tx preset index\n",
+				    __func__);
+		return;
+	}
+
+	/* Following for limiting active channels only */
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_TX_PRESET_TABLE, tx_preset_index,
+		TX_PRESET_TABLE_PRECUR, &tx_preset, 4);
+	precur = tx_preset;
+
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_TX_PRESET_TABLE,
+		tx_preset_index, TX_PRESET_TABLE_ATTN, &tx_preset, 4);
+	attn = tx_preset;
+
+	get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_TX_PRESET_TABLE,
+		tx_preset_index, TX_PRESET_TABLE_POSTCUR, &tx_preset, 4);
+	postcur = tx_preset;
+
+	config_data = precur | (attn << 8) | (postcur << 16);
+
+	apply_tx_lanes(ppd, TX_EQ_SETTINGS, config_data,
+		       "Applying TX settings");
+}
+
+static int tune_active_qsfp(struct hfi1_pportdata *ppd, u32 *ptr_tx_preset,
+			    u32 *ptr_rx_preset, u32 *ptr_total_atten)
+{
+	int ret;
+	u16 lss = ppd->link_speed_supported, lse = ppd->link_speed_enabled;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	ret = acquire_chip_resource(ppd->dd, qsfp_resource(ppd->dd), QSFP_WAIT);
+	if (ret) {
+		dd_dev_err(ppd->dd, "%s: hfi%d: cannot lock i2c chain\n",
+			   __func__, (int)ppd->dd->hfi1_id);
+		return ret;
+	}
+
+	ppd->qsfp_info.limiting_active = 1;
+
+	ret = set_qsfp_tx(ppd, 0);
+	if (ret)
+		goto bail_unlock;
+
+	ret = qual_power(ppd);
+	if (ret)
+		goto bail_unlock;
+
+	ret = qual_bitrate(ppd);
+	if (ret)
+		goto bail_unlock;
+
+	if (ppd->qsfp_info.reset_needed) {
+		reset_qsfp(ppd);
+		ppd->qsfp_info.reset_needed = 0;
+		refresh_qsfp_cache(ppd, &ppd->qsfp_info);
+	} else {
+		ppd->qsfp_info.reset_needed = 1;
+	}
+
+	ret = set_qsfp_high_power(ppd);
+	if (ret)
+		goto bail_unlock;
+
+	if (cache[QSFP_EQ_INFO_OFFS] & 0x4) {
+		ret = get_platform_config_field(
+			ppd->dd,
+			PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_TX_PRESET_IDX_ACTIVE_EQ,
+			ptr_tx_preset, 4);
+		if (ret) {
+			*ptr_tx_preset = OPA_INVALID_INDEX;
+			goto bail_unlock;
+		}
+	} else {
+		ret = get_platform_config_field(
+			ppd->dd,
+			PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_TX_PRESET_IDX_ACTIVE_NO_EQ,
+			ptr_tx_preset, 4);
+		if (ret) {
+			*ptr_tx_preset = OPA_INVALID_INDEX;
+			goto bail_unlock;
+		}
+	}
+
+	ret = get_platform_config_field(
+		ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+		PORT_TABLE_RX_PRESET_IDX, ptr_rx_preset, 4);
+	if (ret) {
+		*ptr_rx_preset = OPA_INVALID_INDEX;
+		goto bail_unlock;
+	}
+
+	if ((lss & OPA_LINK_SPEED_25G) && (lse & OPA_LINK_SPEED_25G))
+		get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_LOCAL_ATTEN_25G, ptr_total_atten, 4);
+	else if ((lss & OPA_LINK_SPEED_12_5G) && (lse & OPA_LINK_SPEED_12_5G))
+		get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_LOCAL_ATTEN_12G, ptr_total_atten, 4);
+
+	apply_cdr_settings(ppd, *ptr_rx_preset, *ptr_tx_preset);
+
+	apply_eq_settings(ppd, *ptr_rx_preset, *ptr_tx_preset);
+
+	apply_rx_amplitude_settings(ppd, *ptr_rx_preset, *ptr_tx_preset);
+
+	ret = set_qsfp_tx(ppd, 1);
+
+bail_unlock:
+	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
+	return ret;
+}
+
+static int tune_qsfp(struct hfi1_pportdata *ppd,
+		     u32 *ptr_tx_preset, u32 *ptr_rx_preset,
+		     u8 *ptr_tuning_method, u32 *ptr_total_atten)
+{
+	u32 cable_atten = 0, remote_atten = 0, platform_atten = 0;
+	u16 lss = ppd->link_speed_supported, lse = ppd->link_speed_enabled;
+	int ret = 0;
+	u8 *cache = ppd->qsfp_info.cache;
+
+	switch ((cache[QSFP_MOD_TECH_OFFS] & 0xF0) >> 4) {
+	case 0xA ... 0xB:
+		ret = get_platform_config_field(
+			ppd->dd,
+			PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_LOCAL_ATTEN_25G,
+			&platform_atten, 4);
+		if (ret)
+			return ret;
+
+		if ((lss & OPA_LINK_SPEED_25G) && (lse & OPA_LINK_SPEED_25G))
+			cable_atten = cache[QSFP_CU_ATTEN_12G_OFFS];
+		else if ((lss & OPA_LINK_SPEED_12_5G) &&
+			 (lse & OPA_LINK_SPEED_12_5G))
+			cable_atten = cache[QSFP_CU_ATTEN_7G_OFFS];
+
+		/* Fallback to configured attenuation if cable memory is bad */
+		if (cable_atten == 0 || cable_atten > 36) {
+			ret = get_platform_config_field(
+				ppd->dd,
+				PLATFORM_CONFIG_SYSTEM_TABLE, 0,
+				SYSTEM_TABLE_QSFP_ATTENUATION_DEFAULT_25G,
+				&cable_atten, 4);
+			if (ret)
+				return ret;
+		}
+
+		ret = get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_REMOTE_ATTEN_25G, &remote_atten, 4);
+		if (ret)
+			return ret;
+
+		*ptr_total_atten = platform_atten + cable_atten + remote_atten;
+
+		*ptr_tuning_method = OPA_PASSIVE_TUNING;
+		break;
+	case 0x0 ... 0x9: /* fallthrough */
+	case 0xC: /* fallthrough */
+	case 0xE:
+		ret = tune_active_qsfp(ppd, ptr_tx_preset, ptr_rx_preset,
+				       ptr_total_atten);
+		if (ret)
+			return ret;
+
+		*ptr_tuning_method = OPA_ACTIVE_TUNING;
+		break;
+	case 0xD: /* fallthrough */
+	case 0xF:
+	default:
+		dd_dev_info(ppd->dd, "%s: Unknown/unsupported cable\n",
+			    __func__);
+		break;
+	}
+	return ret;
+}
+
+/*
+ * This function communicates its success or failure via ppd->driver_link_ready
+ * Thus, it depends on its association with start_link(...) which checks
+ * driver_link_ready before proceeding with the link negotiation and
+ * initialization process.
+ */
+void tune_serdes(struct hfi1_pportdata *ppd)
+{
+	int ret = 0;
+	u32 total_atten = 0;
+	u32 remote_atten = 0, platform_atten = 0;
+	u32 rx_preset_index, tx_preset_index;
+	u8 tuning_method = 0, limiting_active = 0;
+	struct hfi1_devdata *dd = ppd->dd;
+
+	rx_preset_index = OPA_INVALID_INDEX;
+	tx_preset_index = OPA_INVALID_INDEX;
+
+	/* the link defaults to enabled */
+	ppd->link_enabled = 1;
+	/* the driver link ready state defaults to not ready */
+	ppd->driver_link_ready = 0;
+	ppd->offline_disabled_reason = HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE);
+
+	/* Skip the tuning for testing (loopback != none) and simulations */
+	if (loopback != LOOPBACK_NONE ||
+	    ppd->dd->icode == ICODE_FUNCTIONAL_SIMULATOR) {
+		ppd->driver_link_ready = 1;
+		return;
+	}
+
+	ret = get_platform_config_field(ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+					PORT_TABLE_PORT_TYPE, &ppd->port_type,
+					4);
+	if (ret)
+		ppd->port_type = PORT_TYPE_UNKNOWN;
+
+	switch (ppd->port_type) {
+	case PORT_TYPE_DISCONNECTED:
+		ppd->offline_disabled_reason =
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_DISCONNECTED);
+		dd_dev_info(dd, "%s: Port disconnected, disabling port\n",
+			    __func__);
+		goto bail;
+	case PORT_TYPE_FIXED:
+		/* platform_atten, remote_atten pre-zeroed to catch error */
+		get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_LOCAL_ATTEN_25G, &platform_atten, 4);
+
+		get_platform_config_field(
+			ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+			PORT_TABLE_REMOTE_ATTEN_25G, &remote_atten, 4);
+
+		total_atten = platform_atten + remote_atten;
+
+		tuning_method = OPA_PASSIVE_TUNING;
+		break;
+	case PORT_TYPE_VARIABLE:
+		if (qsfp_mod_present(ppd)) {
+			/*
+			 * platform_atten, remote_atten pre-zeroed to
+			 * catch error
+			 */
+			get_platform_config_field(
+				ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+				PORT_TABLE_LOCAL_ATTEN_25G,
+				&platform_atten, 4);
+
+			get_platform_config_field(
+				ppd->dd, PLATFORM_CONFIG_PORT_TABLE, 0,
+				PORT_TABLE_REMOTE_ATTEN_25G,
+				&remote_atten, 4);
+
+			total_atten = platform_atten + remote_atten;
+
+			tuning_method = OPA_PASSIVE_TUNING;
+		} else
+			ppd->offline_disabled_reason =
+			     HFI1_ODR_MASK(OPA_LINKDOWN_REASON_CHASSIS_CONFIG);
+		break;
+	case PORT_TYPE_QSFP:
+		if (qsfp_mod_present(ppd)) {
+			refresh_qsfp_cache(ppd, &ppd->qsfp_info);
+
+			if (ppd->qsfp_info.cache_valid) {
+				ret = tune_qsfp(ppd,
+						&tx_preset_index,
+						&rx_preset_index,
+						&tuning_method,
+						&total_atten);
+
+				/*
+				 * We may have modified the QSFP memory, so
+				 * update the cache to reflect the changes
+				 */
+				refresh_qsfp_cache(ppd, &ppd->qsfp_info);
+				if (ret)
+					goto bail;
+
+				limiting_active =
+						ppd->qsfp_info.limiting_active;
+			} else {
+				dd_dev_err(dd,
+					   "%s: Reading QSFP memory failed\n",
+					   __func__);
+				goto bail;
+			}
+		} else
+			ppd->offline_disabled_reason =
+			   HFI1_ODR_MASK(
+				OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED);
+		break;
+	default:
+		dd_dev_info(ppd->dd, "%s: Unknown port type\n", __func__);
+		ppd->port_type = PORT_TYPE_UNKNOWN;
+		tuning_method = OPA_UNKNOWN_TUNING;
+		total_atten = 0;
+		limiting_active = 0;
+		tx_preset_index = OPA_INVALID_INDEX;
+		break;
+	}
+
+	if (ppd->offline_disabled_reason ==
+			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_NONE))
+		apply_tunings(ppd, tx_preset_index, tuning_method,
+			      total_atten, limiting_active);
+
+	if (!ret)
+		ppd->driver_link_ready = 1;
+
+	return;
+bail:
+	ppd->driver_link_ready = 0;
+}
diff --git a/drivers/staging/rdma/hfi1/platform_config.h b/drivers/staging/rdma/hfi1/platform.h
index 8a94a8342052..19620cf546d5 100644
--- a/drivers/staging/rdma/hfi1/platform_config.h
+++ b/drivers/staging/rdma/hfi1/platform.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -47,8 +44,8 @@
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  *
  */
-#ifndef __PLATFORM_CONFIG_H
-#define __PLATFORM_CONFIG_H
+#ifndef __PLATFORM_H
+#define __PLATFORM_H
 
 #define METADATA_TABLE_FIELD_START_SHIFT		0
 #define METADATA_TABLE_FIELD_START_LEN_BITS		15
@@ -94,17 +91,18 @@ enum platform_config_system_table_fields {
 enum platform_config_port_table_fields {
 	PORT_TABLE_RESERVED,
 	PORT_TABLE_PORT_TYPE,
-	PORT_TABLE_ATTENUATION_12G,
-	PORT_TABLE_ATTENUATION_25G,
+	PORT_TABLE_LOCAL_ATTEN_12G,
+	PORT_TABLE_LOCAL_ATTEN_25G,
 	PORT_TABLE_LINK_SPEED_SUPPORTED,
 	PORT_TABLE_LINK_WIDTH_SUPPORTED,
+	PORT_TABLE_AUTO_LANE_SHEDDING_ENABLED,
+	PORT_TABLE_EXTERNAL_LOOPBACK_ALLOWED,
 	PORT_TABLE_VL_CAP,
 	PORT_TABLE_MTU_CAP,
 	PORT_TABLE_TX_LANE_ENABLE_MASK,
 	PORT_TABLE_LOCAL_MAX_TIMEOUT,
-	PORT_TABLE_AUTO_LANE_SHEDDING_ENABLED,
-	PORT_TABLE_EXTERNAL_LOOPBACK_ALLOWED,
-	PORT_TABLE_TX_PRESET_IDX_PASSIVE_CU,
+	PORT_TABLE_REMOTE_ATTEN_12G,
+	PORT_TABLE_REMOTE_ATTEN_25G,
 	PORT_TABLE_TX_PRESET_IDX_ACTIVE_NO_EQ,
 	PORT_TABLE_TX_PRESET_IDX_ACTIVE_EQ,
 	PORT_TABLE_RX_PRESET_IDX,
@@ -115,10 +113,10 @@ enum platform_config_port_table_fields {
 enum platform_config_rx_preset_table_fields {
 	RX_PRESET_TABLE_RESERVED,
 	RX_PRESET_TABLE_QSFP_RX_CDR_APPLY,
-	RX_PRESET_TABLE_QSFP_RX_EQ_APPLY,
+	RX_PRESET_TABLE_QSFP_RX_EMP_APPLY,
 	RX_PRESET_TABLE_QSFP_RX_AMP_APPLY,
 	RX_PRESET_TABLE_QSFP_RX_CDR,
-	RX_PRESET_TABLE_QSFP_RX_EQ,
+	RX_PRESET_TABLE_QSFP_RX_EMP,
 	RX_PRESET_TABLE_QSFP_RX_AMP,
 	RX_PRESET_TABLE_MAX
 };
@@ -149,6 +147,11 @@ enum platform_config_variable_settings_table_fields {
 	VARIABLE_SETTINGS_TABLE_MAX
 };
 
+struct platform_config {
+	size_t size;
+	const u8 *data;
+};
+
 struct platform_config_data {
 	u32 *table;
 	u32 *table_metadata;
@@ -179,9 +182,11 @@ static const u32 platform_config_table_limits[PLATFORM_CONFIG_TABLE_MAX] = {
  * fields defined for each table above
  */
 
-/*=====================================================
+/*
+ * =====================================================
  *  System table encodings
- *====================================================*/
+ * =====================================================
+ */
 #define PLATFORM_CONFIG_MAGIC_NUM		0x3d4f5041
 #define PLATFORM_CONFIG_MAGIC_NUMBER_LEN	4
 
@@ -199,12 +204,13 @@ enum platform_config_qsfp_power_class_encoding {
 	QSFP_POWER_CLASS_7
 };
 
-
-/*=====================================================
+/*
+ * ====================================================
  *  Port table encodings
- *==================================================== */
+ * ====================================================
+ */
 enum platform_config_port_type_encoding {
-	PORT_TYPE_RESERVED,
+	PORT_TYPE_UNKNOWN,
 	PORT_TYPE_DISCONNECTED,
 	PORT_TYPE_FIXED,
 	PORT_TYPE_VARIABLE,
@@ -283,4 +289,16 @@ enum platform_config_local_max_timeout_encoding {
 	LOCAL_MAX_TIMEOUT_1000_S
 };
 
-#endif			/*__PLATFORM_CONFIG_H*/
+enum link_tuning_encoding {
+	OPA_PASSIVE_TUNING,
+	OPA_ACTIVE_TUNING,
+	OPA_UNKNOWN_TUNING
+};
+
+/* platform.c */
+void get_platform_config(struct hfi1_devdata *dd);
+void free_platform_config(struct hfi1_devdata *dd);
+int set_qsfp_tx(struct hfi1_pportdata *ppd, int on);
+void tune_serdes(struct hfi1_pportdata *ppd);
+
+#endif			/*__PLATFORM_H*/
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index ce036810d576..29a5ad28019b 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -54,31 +51,32 @@
 #include <linux/module.h>
 #include <linux/random.h>
 #include <linux/seq_file.h>
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_qp.h>
 
 #include "hfi.h"
 #include "qp.h"
 #include "trace.h"
-#include "sdma.h"
+#include "verbs_txreq.h"
 
-#define BITS_PER_PAGE           (PAGE_SIZE*BITS_PER_BYTE)
-#define BITS_PER_PAGE_MASK      (BITS_PER_PAGE-1)
-
-static unsigned int hfi1_qp_table_size = 256;
+unsigned int hfi1_qp_table_size = 256;
 module_param_named(qp_table_size, hfi1_qp_table_size, uint, S_IRUGO);
 MODULE_PARM_DESC(qp_table_size, "QP table size");
 
-static void flush_tx_list(struct hfi1_qp *qp);
+static void flush_tx_list(struct rvt_qp *qp);
 static int iowait_sleep(
 	struct sdma_engine *sde,
 	struct iowait *wait,
 	struct sdma_txreq *stx,
 	unsigned seq);
 static void iowait_wakeup(struct iowait *wait, int reason);
+static void iowait_sdma_drained(struct iowait *wait);
+static void qp_pio_drain(struct rvt_qp *qp);
 
-static inline unsigned mk_qpn(struct hfi1_qpn_table *qpt,
-			      struct qpn_map *map, unsigned off)
+static inline unsigned mk_qpn(struct rvt_qpn_table *qpt,
+			      struct rvt_qpn_map *map, unsigned off)
 {
-	return (map - qpt->map) * BITS_PER_PAGE + off;
+	return (map - qpt->map) * RVT_BITS_PER_PAGE + off;
 }
 
 /*
@@ -118,437 +116,15 @@ static const u16 credit_table[31] = {
 	32768                   /* 1E */
 };
 
-static void get_map_page(struct hfi1_qpn_table *qpt, struct qpn_map *map)
-{
-	unsigned long page = get_zeroed_page(GFP_KERNEL);
-
-	/*
-	 * Free the page if someone raced with us installing it.
-	 */
-
-	spin_lock(&qpt->lock);
-	if (map->page)
-		free_page(page);
-	else
-		map->page = (void *)page;
-	spin_unlock(&qpt->lock);
-}
-
-/*
- * Allocate the next available QPN or
- * zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
- */
-static int alloc_qpn(struct hfi1_devdata *dd, struct hfi1_qpn_table *qpt,
-		     enum ib_qp_type type, u8 port)
-{
-	u32 i, offset, max_scan, qpn;
-	struct qpn_map *map;
-	u32 ret;
-
-	if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
-		unsigned n;
-
-		ret = type == IB_QPT_GSI;
-		n = 1 << (ret + 2 * (port - 1));
-		spin_lock(&qpt->lock);
-		if (qpt->flags & n)
-			ret = -EINVAL;
-		else
-			qpt->flags |= n;
-		spin_unlock(&qpt->lock);
-		goto bail;
-	}
-
-	qpn = qpt->last + qpt->incr;
-	if (qpn >= QPN_MAX)
-		qpn = qpt->incr | ((qpt->last & 1) ^ 1);
-	/* offset carries bit 0 */
-	offset = qpn & BITS_PER_PAGE_MASK;
-	map = &qpt->map[qpn / BITS_PER_PAGE];
-	max_scan = qpt->nmaps - !offset;
-	for (i = 0;;) {
-		if (unlikely(!map->page)) {
-			get_map_page(qpt, map);
-			if (unlikely(!map->page))
-				break;
-		}
-		do {
-			if (!test_and_set_bit(offset, map->page)) {
-				qpt->last = qpn;
-				ret = qpn;
-				goto bail;
-			}
-			offset += qpt->incr;
-			/*
-			 * This qpn might be bogus if offset >= BITS_PER_PAGE.
-			 * That is OK.   It gets re-assigned below
-			 */
-			qpn = mk_qpn(qpt, map, offset);
-		} while (offset < BITS_PER_PAGE && qpn < QPN_MAX);
-		/*
-		 * In order to keep the number of pages allocated to a
-		 * minimum, we scan the all existing pages before increasing
-		 * the size of the bitmap table.
-		 */
-		if (++i > max_scan) {
-			if (qpt->nmaps == QPNMAP_ENTRIES)
-				break;
-			map = &qpt->map[qpt->nmaps++];
-			/* start at incr with current bit 0 */
-			offset = qpt->incr | (offset & 1);
-		} else if (map < &qpt->map[qpt->nmaps]) {
-			++map;
-			/* start at incr with current bit 0 */
-			offset = qpt->incr | (offset & 1);
-		} else {
-			map = &qpt->map[0];
-			/* wrap to first map page, invert bit 0 */
-			offset = qpt->incr | ((offset & 1) ^ 1);
-		}
-		/* there can be no bits at shift and below */
-		WARN_ON(offset & (dd->qos_shift - 1));
-		qpn = mk_qpn(qpt, map, offset);
-	}
-
-	ret = -ENOMEM;
-
-bail:
-	return ret;
-}
-
-static void free_qpn(struct hfi1_qpn_table *qpt, u32 qpn)
-{
-	struct qpn_map *map;
-
-	map = qpt->map + qpn / BITS_PER_PAGE;
-	if (map->page)
-		clear_bit(qpn & BITS_PER_PAGE_MASK, map->page);
-}
-
-/*
- * Put the QP into the hash table.
- * The hash table holds a reference to the QP.
- */
-static void insert_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
-{
-	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	unsigned long flags;
-
-	atomic_inc(&qp->refcount);
-	spin_lock_irqsave(&dev->qp_dev->qpt_lock, flags);
-
-	if (qp->ibqp.qp_num <= 1) {
-		rcu_assign_pointer(ibp->qp[qp->ibqp.qp_num], qp);
-	} else {
-		u32 n = qpn_hash(dev->qp_dev, qp->ibqp.qp_num);
-
-		qp->next = dev->qp_dev->qp_table[n];
-		rcu_assign_pointer(dev->qp_dev->qp_table[n], qp);
-		trace_hfi1_qpinsert(qp, n);
-	}
-
-	spin_unlock_irqrestore(&dev->qp_dev->qpt_lock, flags);
-}
-
-/*
- * Remove the QP from the table so it can't be found asynchronously by
- * the receive interrupt routine.
- */
-static void remove_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
-{
-	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	u32 n = qpn_hash(dev->qp_dev, qp->ibqp.qp_num);
-	unsigned long flags;
-	int removed = 1;
-
-	spin_lock_irqsave(&dev->qp_dev->qpt_lock, flags);
-
-	if (rcu_dereference_protected(ibp->qp[0],
-			lockdep_is_held(&dev->qp_dev->qpt_lock)) == qp) {
-		RCU_INIT_POINTER(ibp->qp[0], NULL);
-	} else if (rcu_dereference_protected(ibp->qp[1],
-			lockdep_is_held(&dev->qp_dev->qpt_lock)) == qp) {
-		RCU_INIT_POINTER(ibp->qp[1], NULL);
-	} else {
-		struct hfi1_qp *q;
-		struct hfi1_qp __rcu **qpp;
-
-		removed = 0;
-		qpp = &dev->qp_dev->qp_table[n];
-		for (; (q = rcu_dereference_protected(*qpp,
-				lockdep_is_held(&dev->qp_dev->qpt_lock)))
-					!= NULL;
-				qpp = &q->next)
-			if (q == qp) {
-				RCU_INIT_POINTER(*qpp,
-				 rcu_dereference_protected(qp->next,
-				 lockdep_is_held(&dev->qp_dev->qpt_lock)));
-				removed = 1;
-				trace_hfi1_qpremove(qp, n);
-				break;
-			}
-	}
-
-	spin_unlock_irqrestore(&dev->qp_dev->qpt_lock, flags);
-	if (removed) {
-		synchronize_rcu();
-		if (atomic_dec_and_test(&qp->refcount))
-			wake_up(&qp->wait);
-	}
-}
-
-/**
- * free_all_qps - check for QPs still in use
- * @qpt: the QP table to empty
- *
- * There should not be any QPs still in use.
- * Free memory for table.
- */
-static unsigned free_all_qps(struct hfi1_devdata *dd)
-{
-	struct hfi1_ibdev *dev = &dd->verbs_dev;
-	unsigned long flags;
-	struct hfi1_qp *qp;
-	unsigned n, qp_inuse = 0;
-
-	for (n = 0; n < dd->num_pports; n++) {
-		struct hfi1_ibport *ibp = &dd->pport[n].ibport_data;
-
-		if (!hfi1_mcast_tree_empty(ibp))
-			qp_inuse++;
-		rcu_read_lock();
-		if (rcu_dereference(ibp->qp[0]))
-			qp_inuse++;
-		if (rcu_dereference(ibp->qp[1]))
-			qp_inuse++;
-		rcu_read_unlock();
-	}
-
-	if (!dev->qp_dev)
-		goto bail;
-	spin_lock_irqsave(&dev->qp_dev->qpt_lock, flags);
-	for (n = 0; n < dev->qp_dev->qp_table_size; n++) {
-		qp = rcu_dereference_protected(dev->qp_dev->qp_table[n],
-			lockdep_is_held(&dev->qp_dev->qpt_lock));
-		RCU_INIT_POINTER(dev->qp_dev->qp_table[n], NULL);
-
-		for (; qp; qp = rcu_dereference_protected(qp->next,
-				lockdep_is_held(&dev->qp_dev->qpt_lock)))
-			qp_inuse++;
-	}
-	spin_unlock_irqrestore(&dev->qp_dev->qpt_lock, flags);
-	synchronize_rcu();
-bail:
-	return qp_inuse;
-}
-
-/**
- * reset_qp - initialize the QP state to the reset state
- * @qp: the QP to reset
- * @type: the QP type
- */
-static void reset_qp(struct hfi1_qp *qp, enum ib_qp_type type)
-{
-	qp->remote_qpn = 0;
-	qp->qkey = 0;
-	qp->qp_access_flags = 0;
-	iowait_init(
-		&qp->s_iowait,
-		1,
-		hfi1_do_send,
-		iowait_sleep,
-		iowait_wakeup);
-	qp->s_flags &= HFI1_S_SIGNAL_REQ_WR;
-	qp->s_hdrwords = 0;
-	qp->s_wqe = NULL;
-	qp->s_draining = 0;
-	qp->s_next_psn = 0;
-	qp->s_last_psn = 0;
-	qp->s_sending_psn = 0;
-	qp->s_sending_hpsn = 0;
-	qp->s_psn = 0;
-	qp->r_psn = 0;
-	qp->r_msn = 0;
-	if (type == IB_QPT_RC) {
-		qp->s_state = IB_OPCODE_RC_SEND_LAST;
-		qp->r_state = IB_OPCODE_RC_SEND_LAST;
-	} else {
-		qp->s_state = IB_OPCODE_UC_SEND_LAST;
-		qp->r_state = IB_OPCODE_UC_SEND_LAST;
-	}
-	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
-	qp->r_nak_state = 0;
-	qp->r_adefered = 0;
-	qp->r_aflags = 0;
-	qp->r_flags = 0;
-	qp->s_head = 0;
-	qp->s_tail = 0;
-	qp->s_cur = 0;
-	qp->s_acked = 0;
-	qp->s_last = 0;
-	qp->s_ssn = 1;
-	qp->s_lsn = 0;
-	clear_ahg(qp);
-	qp->s_mig_state = IB_MIG_MIGRATED;
-	memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
-	qp->r_head_ack_queue = 0;
-	qp->s_tail_ack_queue = 0;
-	qp->s_num_rd_atomic = 0;
-	if (qp->r_rq.wq) {
-		qp->r_rq.wq->head = 0;
-		qp->r_rq.wq->tail = 0;
-	}
-	qp->r_sge.num_sge = 0;
-}
-
-static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
-{
-	unsigned n;
-
-	if (test_and_clear_bit(HFI1_R_REWIND_SGE, &qp->r_aflags))
-		hfi1_put_ss(&qp->s_rdma_read_sge);
-
-	hfi1_put_ss(&qp->r_sge);
-
-	if (clr_sends) {
-		while (qp->s_last != qp->s_head) {
-			struct hfi1_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
-			unsigned i;
-
-			for (i = 0; i < wqe->wr.num_sge; i++) {
-				struct hfi1_sge *sge = &wqe->sg_list[i];
-
-				hfi1_put_mr(sge->mr);
-			}
-			if (qp->ibqp.qp_type == IB_QPT_UD ||
-			    qp->ibqp.qp_type == IB_QPT_SMI ||
-			    qp->ibqp.qp_type == IB_QPT_GSI)
-				atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
-			if (++qp->s_last >= qp->s_size)
-				qp->s_last = 0;
-		}
-		if (qp->s_rdma_mr) {
-			hfi1_put_mr(qp->s_rdma_mr);
-			qp->s_rdma_mr = NULL;
-		}
-	}
-
-	if (qp->ibqp.qp_type != IB_QPT_RC)
-		return;
-
-	for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
-		struct hfi1_ack_entry *e = &qp->s_ack_queue[n];
-
-		if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
-		    e->rdma_sge.mr) {
-			hfi1_put_mr(e->rdma_sge.mr);
-			e->rdma_sge.mr = NULL;
-		}
-	}
-}
-
-/**
- * hfi1_error_qp - put a QP into the error state
- * @qp: the QP to put into the error state
- * @err: the receive completion error to signal if a RWQE is active
- *
- * Flushes both send and receive work queues.
- * Returns true if last WQE event should be generated.
- * The QP r_lock and s_lock should be held and interrupts disabled.
- * If we are already in error state, just return.
- */
-int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
+static void flush_tx_list(struct rvt_qp *qp)
 {
-	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
-	struct ib_wc wc;
-	int ret = 0;
-
-	if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
-		goto bail;
-
-	qp->state = IB_QPS_ERR;
-
-	if (qp->s_flags & (HFI1_S_TIMER | HFI1_S_WAIT_RNR)) {
-		qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_WAIT_RNR);
-		del_timer(&qp->s_timer);
-	}
-
-	if (qp->s_flags & HFI1_S_ANY_WAIT_SEND)
-		qp->s_flags &= ~HFI1_S_ANY_WAIT_SEND;
-
-	write_seqlock(&dev->iowait_lock);
-	if (!list_empty(&qp->s_iowait.list) && !(qp->s_flags & HFI1_S_BUSY)) {
-		qp->s_flags &= ~HFI1_S_ANY_WAIT_IO;
-		list_del_init(&qp->s_iowait.list);
-		if (atomic_dec_and_test(&qp->refcount))
-			wake_up(&qp->wait);
-	}
-	write_sequnlock(&dev->iowait_lock);
-
-	if (!(qp->s_flags & HFI1_S_BUSY)) {
-		qp->s_hdrwords = 0;
-		if (qp->s_rdma_mr) {
-			hfi1_put_mr(qp->s_rdma_mr);
-			qp->s_rdma_mr = NULL;
-		}
-		flush_tx_list(qp);
-	}
-
-	/* Schedule the sending tasklet to drain the send work queue. */
-	if (qp->s_last != qp->s_head)
-		hfi1_schedule_send(qp);
-
-	clear_mr_refs(qp, 0);
-
-	memset(&wc, 0, sizeof(wc));
-	wc.qp = &qp->ibqp;
-	wc.opcode = IB_WC_RECV;
-
-	if (test_and_clear_bit(HFI1_R_WRID_VALID, &qp->r_aflags)) {
-		wc.wr_id = qp->r_wr_id;
-		wc.status = err;
-		hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
-	}
-	wc.status = IB_WC_WR_FLUSH_ERR;
+	struct hfi1_qp_priv *priv = qp->priv;
 
-	if (qp->r_rq.wq) {
-		struct hfi1_rwq *wq;
-		u32 head;
-		u32 tail;
-
-		spin_lock(&qp->r_rq.lock);
-
-		/* sanity check pointers before trusting them */
-		wq = qp->r_rq.wq;
-		head = wq->head;
-		if (head >= qp->r_rq.size)
-			head = 0;
-		tail = wq->tail;
-		if (tail >= qp->r_rq.size)
-			tail = 0;
-		while (tail != head) {
-			wc.wr_id = get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
-			if (++tail >= qp->r_rq.size)
-				tail = 0;
-			hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
-		}
-		wq->tail = tail;
-
-		spin_unlock(&qp->r_rq.lock);
-	} else if (qp->ibqp.event_handler)
-		ret = 1;
-
-bail:
-	return ret;
-}
-
-static void flush_tx_list(struct hfi1_qp *qp)
-{
-	while (!list_empty(&qp->s_iowait.tx_head)) {
+	while (!list_empty(&priv->s_iowait.tx_head)) {
 		struct sdma_txreq *tx;
 
 		tx = list_first_entry(
-			&qp->s_iowait.tx_head,
+			&priv->s_iowait.tx_head,
 			struct sdma_txreq,
 			list);
 		list_del_init(&tx->list);
@@ -557,14 +133,15 @@ static void flush_tx_list(struct hfi1_qp *qp)
 	}
 }
 
-static void flush_iowait(struct hfi1_qp *qp)
+static void flush_iowait(struct rvt_qp *qp)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
 	unsigned long flags;
 
 	write_seqlock_irqsave(&dev->iowait_lock, flags);
-	if (!list_empty(&qp->s_iowait.list)) {
-		list_del_init(&qp->s_iowait.list);
+	if (!list_empty(&priv->s_iowait.list)) {
+		list_del_init(&priv->s_iowait.list);
 		if (atomic_dec_and_test(&qp->refcount))
 			wake_up(&qp->wait);
 	}
@@ -597,362 +174,106 @@ static inline int verbs_mtu_enum_to_int(struct ib_device *dev, enum ib_mtu mtu)
 	return ib_mtu_enum_to_int(mtu);
 }
 
-
-/**
- * hfi1_modify_qp - modify the attributes of a queue pair
- * @ibqp: the queue pair who's attributes we're modifying
- * @attr: the new attributes
- * @attr_mask: the mask of attributes to modify
- * @udata: user data for libibverbs.so
- *
- * Returns 0 on success, otherwise returns an errno.
- */
-int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		   int attr_mask, struct ib_udata *udata)
+int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
+			 int attr_mask, struct ib_udata *udata)
 {
+	struct ib_qp *ibqp = &qp->ibqp;
 	struct hfi1_ibdev *dev = to_idev(ibqp->device);
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	enum ib_qp_state cur_state, new_state;
-	struct ib_event ev;
-	int lastwqe = 0;
-	int mig = 0;
-	int ret;
-	u32 pmtu = 0; /* for gcc warning only */
 	struct hfi1_devdata *dd = dd_from_dev(dev);
-
-	spin_lock_irq(&qp->r_lock);
-	spin_lock(&qp->s_lock);
-
-	cur_state = attr_mask & IB_QP_CUR_STATE ?
-		attr->cur_qp_state : qp->state;
-	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
-
-	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
-				attr_mask, IB_LINK_LAYER_UNSPECIFIED))
-		goto inval;
+	u8 sc;
 
 	if (attr_mask & IB_QP_AV) {
-		u8 sc;
-
-		if (attr->ah_attr.dlid >= HFI1_MULTICAST_LID_BASE)
-			goto inval;
-		if (hfi1_check_ah(qp->ibqp.device, &attr->ah_attr))
-			goto inval;
 		sc = ah_to_sc(ibqp->device, &attr->ah_attr);
+		if (sc == 0xf)
+			return -EINVAL;
+
 		if (!qp_to_sdma_engine(qp, sc) &&
 		    dd->flags & HFI1_HAS_SEND_DMA)
-			goto inval;
+			return -EINVAL;
+
+		if (!qp_to_send_context(qp, sc))
+			return -EINVAL;
 	}
 
 	if (attr_mask & IB_QP_ALT_PATH) {
-		u8 sc;
-
-		if (attr->alt_ah_attr.dlid >= HFI1_MULTICAST_LID_BASE)
-			goto inval;
-		if (hfi1_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
-			goto inval;
-		if (attr->alt_pkey_index >= hfi1_get_npkeys(dd))
-			goto inval;
 		sc = ah_to_sc(ibqp->device, &attr->alt_ah_attr);
+		if (sc == 0xf)
+			return -EINVAL;
+
 		if (!qp_to_sdma_engine(qp, sc) &&
 		    dd->flags & HFI1_HAS_SEND_DMA)
-			goto inval;
-	}
-
-	if (attr_mask & IB_QP_PKEY_INDEX)
-		if (attr->pkey_index >= hfi1_get_npkeys(dd))
-			goto inval;
-
-	if (attr_mask & IB_QP_MIN_RNR_TIMER)
-		if (attr->min_rnr_timer > 31)
-			goto inval;
+			return -EINVAL;
 
-	if (attr_mask & IB_QP_PORT)
-		if (qp->ibqp.qp_type == IB_QPT_SMI ||
-		    qp->ibqp.qp_type == IB_QPT_GSI ||
-		    attr->port_num == 0 ||
-		    attr->port_num > ibqp->device->phys_port_cnt)
-			goto inval;
-
-	if (attr_mask & IB_QP_DEST_QPN)
-		if (attr->dest_qp_num > HFI1_QPN_MASK)
-			goto inval;
+		if (!qp_to_send_context(qp, sc))
+			return -EINVAL;
+	}
 
-	if (attr_mask & IB_QP_RETRY_CNT)
-		if (attr->retry_cnt > 7)
-			goto inval;
+	return 0;
+}
 
-	if (attr_mask & IB_QP_RNR_RETRY)
-		if (attr->rnr_retry > 7)
-			goto inval;
+void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
+		    int attr_mask, struct ib_udata *udata)
+{
+	struct ib_qp *ibqp = &qp->ibqp;
+	struct hfi1_qp_priv *priv = qp->priv;
 
-	/*
-	 * Don't allow invalid path_mtu values.  OK to set greater
-	 * than the active mtu (or even the max_cap, if we have tuned
-	 * that to a small mtu.  We'll set qp->path_mtu
-	 * to the lesser of requested attribute mtu and active,
-	 * for packetizing messages.
-	 * Note that the QP port has to be set in INIT and MTU in RTR.
-	 */
-	if (attr_mask & IB_QP_PATH_MTU) {
-		int mtu, pidx = qp->port_num - 1;
-
-		dd = dd_from_dev(dev);
-		mtu = verbs_mtu_enum_to_int(ibqp->device, attr->path_mtu);
-		if (mtu == -1)
-			goto inval;
-
-		if (mtu > dd->pport[pidx].ibmtu)
-			pmtu = mtu_to_enum(dd->pport[pidx].ibmtu, IB_MTU_2048);
-		else
-			pmtu = attr->path_mtu;
+	if (attr_mask & IB_QP_AV) {
+		priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
+		priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
+		priv->s_sendcontext = qp_to_send_context(qp, priv->s_sc);
 	}
 
-	if (attr_mask & IB_QP_PATH_MIG_STATE) {
-		if (attr->path_mig_state == IB_MIG_REARM) {
-			if (qp->s_mig_state == IB_MIG_ARMED)
-				goto inval;
-			if (new_state != IB_QPS_RTS)
-				goto inval;
-		} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
-			if (qp->s_mig_state == IB_MIG_REARM)
-				goto inval;
-			if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
-				goto inval;
-			if (qp->s_mig_state == IB_MIG_ARMED)
-				mig = 1;
-		} else
-			goto inval;
+	if (attr_mask & IB_QP_PATH_MIG_STATE &&
+	    attr->path_mig_state == IB_MIG_MIGRATED &&
+	    qp->s_mig_state == IB_MIG_ARMED) {
+		qp->s_flags |= RVT_S_AHG_CLEAR;
+		priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
+		priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
+		priv->s_sendcontext = qp_to_send_context(qp, priv->s_sc);
 	}
+}
 
-	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
-		if (attr->max_dest_rd_atomic > HFI1_MAX_RDMA_ATOMIC)
-			goto inval;
-
-	switch (new_state) {
-	case IB_QPS_RESET:
-		if (qp->state != IB_QPS_RESET) {
-			qp->state = IB_QPS_RESET;
-			flush_iowait(qp);
-			qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_ANY_WAIT);
-			spin_unlock(&qp->s_lock);
-			spin_unlock_irq(&qp->r_lock);
-			/* Stop the sending work queue and retry timer */
-			cancel_work_sync(&qp->s_iowait.iowork);
-			del_timer_sync(&qp->s_timer);
-			iowait_sdma_drain(&qp->s_iowait);
-			flush_tx_list(qp);
-			remove_qp(dev, qp);
-			wait_event(qp->wait, !atomic_read(&qp->refcount));
-			spin_lock_irq(&qp->r_lock);
-			spin_lock(&qp->s_lock);
-			clear_mr_refs(qp, 1);
-			clear_ahg(qp);
-			reset_qp(qp, ibqp->qp_type);
-		}
-		break;
-
-	case IB_QPS_RTR:
-		/* Allow event to re-trigger if QP set to RTR more than once */
-		qp->r_flags &= ~HFI1_R_COMM_EST;
-		qp->state = new_state;
-		break;
-
-	case IB_QPS_SQD:
-		qp->s_draining = qp->s_last != qp->s_cur;
-		qp->state = new_state;
-		break;
+/**
+ * hfi1_check_send_wqe - validate wqe
+ * @qp - The qp
+ * @wqe - The built wqe
+ *
+ * validate wqe.  This is called
+ * prior to inserting the wqe into
+ * the ring but after the wqe has been
+ * setup.
+ *
+ * Returns 0 on success, -EINVAL on failure
+ *
+ */
+int hfi1_check_send_wqe(struct rvt_qp *qp,
+			struct rvt_swqe *wqe)
+{
+	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct rvt_ah *ah;
 
-	case IB_QPS_SQE:
-		if (qp->ibqp.qp_type == IB_QPT_RC)
-			goto inval;
-		qp->state = new_state;
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+		if (wqe->length > 0x80000000U)
+			return -EINVAL;
 		break;
-
-	case IB_QPS_ERR:
-		lastwqe = hfi1_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+	case IB_QPT_SMI:
+		ah = ibah_to_rvtah(wqe->ud_wr.ah);
+		if (wqe->length > (1 << ah->log_pmtu))
+			return -EINVAL;
 		break;
-
+	case IB_QPT_GSI:
+	case IB_QPT_UD:
+		ah = ibah_to_rvtah(wqe->ud_wr.ah);
+		if (wqe->length > (1 << ah->log_pmtu))
+			return -EINVAL;
+		if (ibp->sl_to_sc[ah->attr.sl] == 0xf)
+			return -EINVAL;
 	default:
-		qp->state = new_state;
 		break;
 	}
-
-	if (attr_mask & IB_QP_PKEY_INDEX)
-		qp->s_pkey_index = attr->pkey_index;
-
-	if (attr_mask & IB_QP_PORT)
-		qp->port_num = attr->port_num;
-
-	if (attr_mask & IB_QP_DEST_QPN)
-		qp->remote_qpn = attr->dest_qp_num;
-
-	if (attr_mask & IB_QP_SQ_PSN) {
-		qp->s_next_psn = attr->sq_psn & PSN_MODIFY_MASK;
-		qp->s_psn = qp->s_next_psn;
-		qp->s_sending_psn = qp->s_next_psn;
-		qp->s_last_psn = qp->s_next_psn - 1;
-		qp->s_sending_hpsn = qp->s_last_psn;
-	}
-
-	if (attr_mask & IB_QP_RQ_PSN)
-		qp->r_psn = attr->rq_psn & PSN_MODIFY_MASK;
-
-	if (attr_mask & IB_QP_ACCESS_FLAGS)
-		qp->qp_access_flags = attr->qp_access_flags;
-
-	if (attr_mask & IB_QP_AV) {
-		qp->remote_ah_attr = attr->ah_attr;
-		qp->s_srate = attr->ah_attr.static_rate;
-		qp->srate_mbps = ib_rate_to_mbps(qp->s_srate);
-		qp->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
-		qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
-	}
-
-	if (attr_mask & IB_QP_ALT_PATH) {
-		qp->alt_ah_attr = attr->alt_ah_attr;
-		qp->s_alt_pkey_index = attr->alt_pkey_index;
-	}
-
-	if (attr_mask & IB_QP_PATH_MIG_STATE) {
-		qp->s_mig_state = attr->path_mig_state;
-		if (mig) {
-			qp->remote_ah_attr = qp->alt_ah_attr;
-			qp->port_num = qp->alt_ah_attr.port_num;
-			qp->s_pkey_index = qp->s_alt_pkey_index;
-			qp->s_flags |= HFI1_S_AHG_CLEAR;
-			qp->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
-			qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
-		}
-	}
-
-	if (attr_mask & IB_QP_PATH_MTU) {
-		struct hfi1_ibport *ibp;
-		u8 sc, vl;
-		u32 mtu;
-
-		dd = dd_from_dev(dev);
-		ibp = &dd->pport[qp->port_num - 1].ibport_data;
-
-		sc = ibp->sl_to_sc[qp->remote_ah_attr.sl];
-		vl = sc_to_vlt(dd, sc);
-
-		mtu = verbs_mtu_enum_to_int(ibqp->device, pmtu);
-		if (vl < PER_VL_SEND_CONTEXTS)
-			mtu = min_t(u32, mtu, dd->vld[vl].mtu);
-		pmtu = mtu_to_enum(mtu, OPA_MTU_8192);
-
-		qp->path_mtu = pmtu;
-		qp->pmtu = mtu;
-	}
-
-	if (attr_mask & IB_QP_RETRY_CNT) {
-		qp->s_retry_cnt = attr->retry_cnt;
-		qp->s_retry = attr->retry_cnt;
-	}
-
-	if (attr_mask & IB_QP_RNR_RETRY) {
-		qp->s_rnr_retry_cnt = attr->rnr_retry;
-		qp->s_rnr_retry = attr->rnr_retry;
-	}
-
-	if (attr_mask & IB_QP_MIN_RNR_TIMER)
-		qp->r_min_rnr_timer = attr->min_rnr_timer;
-
-	if (attr_mask & IB_QP_TIMEOUT) {
-		qp->timeout = attr->timeout;
-		qp->timeout_jiffies =
-			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
-				1000UL);
-	}
-
-	if (attr_mask & IB_QP_QKEY)
-		qp->qkey = attr->qkey;
-
-	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
-		qp->r_max_rd_atomic = attr->max_dest_rd_atomic;
-
-	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
-		qp->s_max_rd_atomic = attr->max_rd_atomic;
-
-	spin_unlock(&qp->s_lock);
-	spin_unlock_irq(&qp->r_lock);
-
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
-		insert_qp(dev, qp);
-
-	if (lastwqe) {
-		ev.device = qp->ibqp.device;
-		ev.element.qp = &qp->ibqp;
-		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
-		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
-	}
-	if (mig) {
-		ev.device = qp->ibqp.device;
-		ev.element.qp = &qp->ibqp;
-		ev.event = IB_EVENT_PATH_MIG;
-		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
-	}
-	ret = 0;
-	goto bail;
-
-inval:
-	spin_unlock(&qp->s_lock);
-	spin_unlock_irq(&qp->r_lock);
-	ret = -EINVAL;
-
-bail:
-	return ret;
-}
-
-int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		  int attr_mask, struct ib_qp_init_attr *init_attr)
-{
-	struct hfi1_qp *qp = to_iqp(ibqp);
-
-	attr->qp_state = qp->state;
-	attr->cur_qp_state = attr->qp_state;
-	attr->path_mtu = qp->path_mtu;
-	attr->path_mig_state = qp->s_mig_state;
-	attr->qkey = qp->qkey;
-	attr->rq_psn = mask_psn(qp->r_psn);
-	attr->sq_psn = mask_psn(qp->s_next_psn);
-	attr->dest_qp_num = qp->remote_qpn;
-	attr->qp_access_flags = qp->qp_access_flags;
-	attr->cap.max_send_wr = qp->s_size - 1;
-	attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
-	attr->cap.max_send_sge = qp->s_max_sge;
-	attr->cap.max_recv_sge = qp->r_rq.max_sge;
-	attr->cap.max_inline_data = 0;
-	attr->ah_attr = qp->remote_ah_attr;
-	attr->alt_ah_attr = qp->alt_ah_attr;
-	attr->pkey_index = qp->s_pkey_index;
-	attr->alt_pkey_index = qp->s_alt_pkey_index;
-	attr->en_sqd_async_notify = 0;
-	attr->sq_draining = qp->s_draining;
-	attr->max_rd_atomic = qp->s_max_rd_atomic;
-	attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
-	attr->min_rnr_timer = qp->r_min_rnr_timer;
-	attr->port_num = qp->port_num;
-	attr->timeout = qp->timeout;
-	attr->retry_cnt = qp->s_retry_cnt;
-	attr->rnr_retry = qp->s_rnr_retry_cnt;
-	attr->alt_port_num = qp->alt_ah_attr.port_num;
-	attr->alt_timeout = qp->alt_timeout;
-
-	init_attr->event_handler = qp->ibqp.event_handler;
-	init_attr->qp_context = qp->ibqp.qp_context;
-	init_attr->send_cq = qp->ibqp.send_cq;
-	init_attr->recv_cq = qp->ibqp.recv_cq;
-	init_attr->srq = qp->ibqp.srq;
-	init_attr->cap = attr->cap;
-	if (qp->s_flags & HFI1_S_SIGNAL_REQ_WR)
-		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
-	else
-		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
-	init_attr->qp_type = qp->ibqp.qp_type;
-	init_attr->port_num = qp->port_num;
-	return 0;
+	return wqe->length <= piothreshold;
 }
 
 /**
@@ -961,7 +282,7 @@ int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
  *
  * Returns the AETH.
  */
-__be32 hfi1_compute_aeth(struct hfi1_qp *qp)
+__be32 hfi1_compute_aeth(struct rvt_qp *qp)
 {
 	u32 aeth = qp->r_msn & HFI1_MSN_MASK;
 
@@ -974,7 +295,7 @@ __be32 hfi1_compute_aeth(struct hfi1_qp *qp)
 	} else {
 		u32 min, max, x;
 		u32 credits;
-		struct hfi1_rwq *wq = qp->r_rq.wq;
+		struct rvt_rwq *wq = qp->r_rq.wq;
 		u32 head;
 		u32 tail;
 
@@ -1004,12 +325,13 @@ __be32 hfi1_compute_aeth(struct hfi1_qp *qp)
 			x = (min + max) / 2;
 			if (credit_table[x] == credits)
 				break;
-			if (credit_table[x] > credits)
+			if (credit_table[x] > credits) {
 				max = x;
-			else if (min == x)
-				break;
-			else
+			} else {
+				if (min == x)
+					break;
 				min = x;
+			}
 		}
 		aeth |= x << HFI1_AETH_CREDIT_SHIFT;
 	}
@@ -1017,348 +339,58 @@ __be32 hfi1_compute_aeth(struct hfi1_qp *qp)
 }
 
 /**
- * hfi1_create_qp - create a queue pair for a device
- * @ibpd: the protection domain who's device we create the queue pair for
- * @init_attr: the attributes of the queue pair
- * @udata: user data for libibverbs.so
- *
- * Returns the queue pair on success, otherwise returns an errno.
- *
- * Called by the ib_create_qp() core verbs function.
- */
-struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
-			     struct ib_qp_init_attr *init_attr,
-			     struct ib_udata *udata)
-{
-	struct hfi1_qp *qp;
-	int err;
-	struct hfi1_swqe *swq = NULL;
-	struct hfi1_ibdev *dev;
-	struct hfi1_devdata *dd;
-	size_t sz;
-	size_t sg_list_sz;
-	struct ib_qp *ret;
-
-	if (init_attr->cap.max_send_sge > hfi1_max_sges ||
-	    init_attr->cap.max_send_wr > hfi1_max_qp_wrs ||
-	    init_attr->create_flags) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
-
-	/* Check receive queue parameters if no SRQ is specified. */
-	if (!init_attr->srq) {
-		if (init_attr->cap.max_recv_sge > hfi1_max_sges ||
-		    init_attr->cap.max_recv_wr > hfi1_max_qp_wrs) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-		if (init_attr->cap.max_send_sge +
-		    init_attr->cap.max_send_wr +
-		    init_attr->cap.max_recv_sge +
-		    init_attr->cap.max_recv_wr == 0) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-	}
-
-	switch (init_attr->qp_type) {
-	case IB_QPT_SMI:
-	case IB_QPT_GSI:
-		if (init_attr->port_num == 0 ||
-		    init_attr->port_num > ibpd->device->phys_port_cnt) {
-			ret = ERR_PTR(-EINVAL);
-			goto bail;
-		}
-	case IB_QPT_UC:
-	case IB_QPT_RC:
-	case IB_QPT_UD:
-		sz = sizeof(struct hfi1_sge) *
-			init_attr->cap.max_send_sge +
-			sizeof(struct hfi1_swqe);
-		swq = vmalloc((init_attr->cap.max_send_wr + 1) * sz);
-		if (swq == NULL) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail;
-		}
-		sz = sizeof(*qp);
-		sg_list_sz = 0;
-		if (init_attr->srq) {
-			struct hfi1_srq *srq = to_isrq(init_attr->srq);
-
-			if (srq->rq.max_sge > 1)
-				sg_list_sz = sizeof(*qp->r_sg_list) *
-					(srq->rq.max_sge - 1);
-		} else if (init_attr->cap.max_recv_sge > 1)
-			sg_list_sz = sizeof(*qp->r_sg_list) *
-				(init_attr->cap.max_recv_sge - 1);
-		qp = kzalloc(sz + sg_list_sz, GFP_KERNEL);
-		if (!qp) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_swq;
-		}
-		RCU_INIT_POINTER(qp->next, NULL);
-		qp->s_hdr = kzalloc(sizeof(*qp->s_hdr), GFP_KERNEL);
-		if (!qp->s_hdr) {
-			ret = ERR_PTR(-ENOMEM);
-			goto bail_qp;
-		}
-		qp->timeout_jiffies =
-			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
-				1000UL);
-		if (init_attr->srq)
-			sz = 0;
-		else {
-			qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
-			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
-			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
-				sizeof(struct hfi1_rwqe);
-			qp->r_rq.wq = vmalloc_user(sizeof(struct hfi1_rwq) +
-						   qp->r_rq.size * sz);
-			if (!qp->r_rq.wq) {
-				ret = ERR_PTR(-ENOMEM);
-				goto bail_qp;
-			}
-		}
-
-		/*
-		 * ib_create_qp() will initialize qp->ibqp
-		 * except for qp->ibqp.qp_num.
-		 */
-		spin_lock_init(&qp->r_lock);
-		spin_lock_init(&qp->s_lock);
-		spin_lock_init(&qp->r_rq.lock);
-		atomic_set(&qp->refcount, 0);
-		init_waitqueue_head(&qp->wait);
-		init_timer(&qp->s_timer);
-		qp->s_timer.data = (unsigned long)qp;
-		INIT_LIST_HEAD(&qp->rspwait);
-		qp->state = IB_QPS_RESET;
-		qp->s_wq = swq;
-		qp->s_size = init_attr->cap.max_send_wr + 1;
-		qp->s_max_sge = init_attr->cap.max_send_sge;
-		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
-			qp->s_flags = HFI1_S_SIGNAL_REQ_WR;
-		dev = to_idev(ibpd->device);
-		dd = dd_from_dev(dev);
-		err = alloc_qpn(dd, &dev->qp_dev->qpn_table, init_attr->qp_type,
-				init_attr->port_num);
-		if (err < 0) {
-			ret = ERR_PTR(err);
-			vfree(qp->r_rq.wq);
-			goto bail_qp;
-		}
-		qp->ibqp.qp_num = err;
-		qp->port_num = init_attr->port_num;
-		reset_qp(qp, init_attr->qp_type);
-
-		break;
-
-	default:
-		/* Don't support raw QPs */
-		ret = ERR_PTR(-ENOSYS);
-		goto bail;
-	}
-
-	init_attr->cap.max_inline_data = 0;
-
-	/*
-	 * Return the address of the RWQ as the offset to mmap.
-	 * See hfi1_mmap() for details.
-	 */
-	if (udata && udata->outlen >= sizeof(__u64)) {
-		if (!qp->r_rq.wq) {
-			__u64 offset = 0;
-
-			err = ib_copy_to_udata(udata, &offset,
-					       sizeof(offset));
-			if (err) {
-				ret = ERR_PTR(err);
-				goto bail_ip;
-			}
-		} else {
-			u32 s = sizeof(struct hfi1_rwq) + qp->r_rq.size * sz;
-
-			qp->ip = hfi1_create_mmap_info(dev, s,
-						      ibpd->uobject->context,
-						      qp->r_rq.wq);
-			if (!qp->ip) {
-				ret = ERR_PTR(-ENOMEM);
-				goto bail_ip;
-			}
-
-			err = ib_copy_to_udata(udata, &(qp->ip->offset),
-					       sizeof(qp->ip->offset));
-			if (err) {
-				ret = ERR_PTR(err);
-				goto bail_ip;
-			}
-		}
-	}
-
-	spin_lock(&dev->n_qps_lock);
-	if (dev->n_qps_allocated == hfi1_max_qps) {
-		spin_unlock(&dev->n_qps_lock);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail_ip;
-	}
-
-	dev->n_qps_allocated++;
-	spin_unlock(&dev->n_qps_lock);
-
-	if (qp->ip) {
-		spin_lock_irq(&dev->pending_lock);
-		list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
-		spin_unlock_irq(&dev->pending_lock);
-	}
-
-	ret = &qp->ibqp;
-
-	/*
-	 * We have our QP and its good, now keep track of what types of opcodes
-	 * can be processed on this QP. We do this by keeping track of what the
-	 * 3 high order bits of the opcode are.
-	 */
-	switch (init_attr->qp_type) {
-	case IB_QPT_SMI:
-	case IB_QPT_GSI:
-	case IB_QPT_UD:
-		qp->allowed_ops = IB_OPCODE_UD_SEND_ONLY & OPCODE_QP_MASK;
-		break;
-	case IB_QPT_RC:
-		qp->allowed_ops = IB_OPCODE_RC_SEND_ONLY & OPCODE_QP_MASK;
-		break;
-	case IB_QPT_UC:
-		qp->allowed_ops = IB_OPCODE_UC_SEND_ONLY & OPCODE_QP_MASK;
-		break;
-	default:
-		ret = ERR_PTR(-EINVAL);
-		goto bail_ip;
-	}
-
-	goto bail;
-
-bail_ip:
-	if (qp->ip)
-		kref_put(&qp->ip->ref, hfi1_release_mmap_info);
-	else
-		vfree(qp->r_rq.wq);
-	free_qpn(&dev->qp_dev->qpn_table, qp->ibqp.qp_num);
-bail_qp:
-	kfree(qp->s_hdr);
-	kfree(qp);
-bail_swq:
-	vfree(swq);
-bail:
-	return ret;
-}
-
-/**
- * hfi1_destroy_qp - destroy a queue pair
- * @ibqp: the queue pair to destroy
+ * _hfi1_schedule_send - schedule progress
+ * @qp: the QP
  *
- * Returns 0 on success.
+ * This schedules qp progress w/o regard to the s_flags.
  *
- * Note that this can be called while the QP is actively sending or
- * receiving!
+ * It is only used in the post send, which doesn't hold
+ * the s_lock.
  */
-int hfi1_destroy_qp(struct ib_qp *ibqp)
+void _hfi1_schedule_send(struct rvt_qp *qp)
 {
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	struct hfi1_ibdev *dev = to_idev(ibqp->device);
-
-	/* Make sure HW and driver activity is stopped. */
-	spin_lock_irq(&qp->r_lock);
-	spin_lock(&qp->s_lock);
-	if (qp->state != IB_QPS_RESET) {
-		qp->state = IB_QPS_RESET;
-		flush_iowait(qp);
-		qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_ANY_WAIT);
-		spin_unlock(&qp->s_lock);
-		spin_unlock_irq(&qp->r_lock);
-		cancel_work_sync(&qp->s_iowait.iowork);
-		del_timer_sync(&qp->s_timer);
-		iowait_sdma_drain(&qp->s_iowait);
-		flush_tx_list(qp);
-		remove_qp(dev, qp);
-		wait_event(qp->wait, !atomic_read(&qp->refcount));
-		spin_lock_irq(&qp->r_lock);
-		spin_lock(&qp->s_lock);
-		clear_mr_refs(qp, 1);
-		clear_ahg(qp);
-	}
-	spin_unlock(&qp->s_lock);
-	spin_unlock_irq(&qp->r_lock);
-
-	/* all user's cleaned up, mark it available */
-	free_qpn(&dev->qp_dev->qpn_table, qp->ibqp.qp_num);
-	spin_lock(&dev->n_qps_lock);
-	dev->n_qps_allocated--;
-	spin_unlock(&dev->n_qps_lock);
+	struct hfi1_qp_priv *priv = qp->priv;
+	struct hfi1_ibport *ibp =
+		to_iport(qp->ibqp.device, qp->port_num);
+	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
 
-	if (qp->ip)
-		kref_put(&qp->ip->ref, hfi1_release_mmap_info);
-	else
-		vfree(qp->r_rq.wq);
-	vfree(qp->s_wq);
-	kfree(qp->s_hdr);
-	kfree(qp);
-	return 0;
+	iowait_schedule(&priv->s_iowait, ppd->hfi1_wq,
+			priv->s_sde ?
+			priv->s_sde->cpu :
+			cpumask_first(cpumask_of_node(dd->node)));
 }
 
-/**
- * init_qpn_table - initialize the QP number table for a device
- * @qpt: the QPN table
- */
-static int init_qpn_table(struct hfi1_devdata *dd, struct hfi1_qpn_table *qpt)
+static void qp_pio_drain(struct rvt_qp *qp)
 {
-	u32 offset, qpn, i;
-	struct qpn_map *map;
-	int ret = 0;
+	struct hfi1_ibdev *dev;
+	struct hfi1_qp_priv *priv = qp->priv;
 
-	spin_lock_init(&qpt->lock);
-
-	qpt->last = 0;
-	qpt->incr = 1 << dd->qos_shift;
-
-	/* insure we don't assign QPs from KDETH 64K window */
-	qpn = kdeth_qp << 16;
-	qpt->nmaps = qpn / BITS_PER_PAGE;
-	/* This should always be zero */
-	offset = qpn & BITS_PER_PAGE_MASK;
-	map = &qpt->map[qpt->nmaps];
-	dd_dev_info(dd, "Reserving QPNs for KDETH window from 0x%x to 0x%x\n",
-		qpn, qpn + 65535);
-	for (i = 0; i < 65536; i++) {
-		if (!map->page) {
-			get_map_page(qpt, map);
-			if (!map->page) {
-				ret = -ENOMEM;
-				break;
-			}
-		}
-		set_bit(offset, map->page);
-		offset++;
-		if (offset == BITS_PER_PAGE) {
-			/* next page */
-			qpt->nmaps++;
-			map++;
-			offset = 0;
-		}
+	if (!priv->s_sendcontext)
+		return;
+	dev = to_idev(qp->ibqp.device);
+	while (iowait_pio_pending(&priv->s_iowait)) {
+		write_seqlock_irq(&dev->iowait_lock);
+		hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 1);
+		write_sequnlock_irq(&dev->iowait_lock);
+		iowait_pio_drain(&priv->s_iowait);
+		write_seqlock_irq(&dev->iowait_lock);
+		hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 0);
+		write_sequnlock_irq(&dev->iowait_lock);
 	}
-	return ret;
 }
 
 /**
- * free_qpn_table - free the QP number table for a device
- * @qpt: the QPN table
+ * hfi1_schedule_send - schedule progress
+ * @qp: the QP
+ *
+ * This schedules qp progress and caller should hold
+ * the s_lock.
  */
-static void free_qpn_table(struct hfi1_qpn_table *qpt)
+void hfi1_schedule_send(struct rvt_qp *qp)
 {
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
-		free_page((unsigned long) qpt->map[i].page);
+	if (hfi1_send_ok(qp))
+		_hfi1_schedule_send(qp);
 }
 
 /**
@@ -1368,7 +400,7 @@ static void free_qpn_table(struct hfi1_qpn_table *qpt)
  *
  * The QP s_lock should be held.
  */
-void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth)
+void hfi1_get_credit(struct rvt_qp *qp, u32 aeth)
 {
 	u32 credit = (aeth >> HFI1_AETH_CREDIT_SHIFT) & HFI1_AETH_CREDIT_MASK;
 
@@ -1378,27 +410,27 @@ void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth)
 	 * honor the credit field.
 	 */
 	if (credit == HFI1_AETH_CREDIT_INVAL) {
-		if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT)) {
-			qp->s_flags |= HFI1_S_UNLIMITED_CREDIT;
-			if (qp->s_flags & HFI1_S_WAIT_SSN_CREDIT) {
-				qp->s_flags &= ~HFI1_S_WAIT_SSN_CREDIT;
+		if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
+			qp->s_flags |= RVT_S_UNLIMITED_CREDIT;
+			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
 				hfi1_schedule_send(qp);
 			}
 		}
-	} else if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT)) {
+	} else if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
 		/* Compute new LSN (i.e., MSN + credit) */
 		credit = (aeth + credit_table[credit]) & HFI1_MSN_MASK;
 		if (cmp_msn(credit, qp->s_lsn) > 0) {
 			qp->s_lsn = credit;
-			if (qp->s_flags & HFI1_S_WAIT_SSN_CREDIT) {
-				qp->s_flags &= ~HFI1_S_WAIT_SSN_CREDIT;
+			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
 				hfi1_schedule_send(qp);
 			}
 		}
 	}
 }
 
-void hfi1_qp_wakeup(struct hfi1_qp *qp, u32 flag)
+void hfi1_qp_wakeup(struct rvt_qp *qp, u32 flag)
 {
 	unsigned long flags;
 
@@ -1421,16 +453,17 @@ static int iowait_sleep(
 	unsigned seq)
 {
 	struct verbs_txreq *tx = container_of(stx, struct verbs_txreq, txreq);
-	struct hfi1_qp *qp;
+	struct rvt_qp *qp;
+	struct hfi1_qp_priv *priv;
 	unsigned long flags;
 	int ret = 0;
 	struct hfi1_ibdev *dev;
 
 	qp = tx->qp;
+	priv = qp->priv;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
-
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
 		/*
 		 * If we couldn't queue the DMA request, save the info
 		 * and try again later rather than destroying the
@@ -1442,18 +475,18 @@ static int iowait_sleep(
 		write_seqlock(&dev->iowait_lock);
 		if (sdma_progress(sde, seq, stx))
 			goto eagain;
-		if (list_empty(&qp->s_iowait.list)) {
+		if (list_empty(&priv->s_iowait.list)) {
 			struct hfi1_ibport *ibp =
 				to_iport(qp->ibqp.device, qp->port_num);
 
-			ibp->n_dmawait++;
-			qp->s_flags |= HFI1_S_WAIT_DMA_DESC;
-			list_add_tail(&qp->s_iowait.list, &sde->dmawait);
-			trace_hfi1_qpsleep(qp, HFI1_S_WAIT_DMA_DESC);
+			ibp->rvp.n_dmawait++;
+			qp->s_flags |= RVT_S_WAIT_DMA_DESC;
+			list_add_tail(&priv->s_iowait.list, &sde->dmawait);
+			trace_hfi1_qpsleep(qp, RVT_S_WAIT_DMA_DESC);
 			atomic_inc(&qp->refcount);
 		}
 		write_sequnlock(&dev->iowait_lock);
-		qp->s_flags &= ~HFI1_S_BUSY;
+		qp->s_flags &= ~RVT_S_BUSY;
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 		ret = -EBUSY;
 	} else {
@@ -1470,61 +503,25 @@ eagain:
 
 static void iowait_wakeup(struct iowait *wait, int reason)
 {
-	struct hfi1_qp *qp = container_of(wait, struct hfi1_qp, s_iowait);
+	struct rvt_qp *qp = iowait_to_qp(wait);
 
 	WARN_ON(reason != SDMA_AVAIL_REASON);
-	hfi1_qp_wakeup(qp, HFI1_S_WAIT_DMA_DESC);
+	hfi1_qp_wakeup(qp, RVT_S_WAIT_DMA_DESC);
 }
 
-int hfi1_qp_init(struct hfi1_ibdev *dev)
+static void iowait_sdma_drained(struct iowait *wait)
 {
-	struct hfi1_devdata *dd = dd_from_dev(dev);
-	int i;
-	int ret = -ENOMEM;
-
-	/* allocate parent object */
-	dev->qp_dev = kzalloc(sizeof(*dev->qp_dev), GFP_KERNEL);
-	if (!dev->qp_dev)
-		goto nomem;
-	/* allocate hash table */
-	dev->qp_dev->qp_table_size = hfi1_qp_table_size;
-	dev->qp_dev->qp_table_bits = ilog2(hfi1_qp_table_size);
-	dev->qp_dev->qp_table =
-		kmalloc(dev->qp_dev->qp_table_size *
-				sizeof(*dev->qp_dev->qp_table),
-			GFP_KERNEL);
-	if (!dev->qp_dev->qp_table)
-		goto nomem;
-	for (i = 0; i < dev->qp_dev->qp_table_size; i++)
-		RCU_INIT_POINTER(dev->qp_dev->qp_table[i], NULL);
-	spin_lock_init(&dev->qp_dev->qpt_lock);
-	/* initialize qpn map */
-	ret = init_qpn_table(dd, &dev->qp_dev->qpn_table);
-	if (ret)
-		goto nomem;
-	return ret;
-nomem:
-	if (dev->qp_dev) {
-		kfree(dev->qp_dev->qp_table);
-		free_qpn_table(&dev->qp_dev->qpn_table);
-		kfree(dev->qp_dev);
-	}
-	return ret;
-}
+	struct rvt_qp *qp = iowait_to_qp(wait);
 
-void hfi1_qp_exit(struct hfi1_ibdev *dev)
-{
-	struct hfi1_devdata *dd = dd_from_dev(dev);
-	u32 qps_inuse;
-
-	qps_inuse = free_all_qps(dd);
-	if (qps_inuse)
-		dd_dev_err(dd, "QP memory leak! %u still in use\n",
-			   qps_inuse);
-	if (dev->qp_dev) {
-		kfree(dev->qp_dev->qp_table);
-		free_qpn_table(&dev->qp_dev->qpn_table);
-		kfree(dev->qp_dev);
+	/*
+	 * This happens when the send engine notes
+	 * a QP in the error state and cannot
+	 * do the flush work until that QP's
+	 * sdma work has finished.
+	 */
+	if (qp->s_flags & RVT_S_WAIT_DMA) {
+		qp->s_flags &= ~RVT_S_WAIT_DMA;
+		hfi1_schedule_send(qp);
 	}
 }
 
@@ -1537,7 +534,7 @@ void hfi1_qp_exit(struct hfi1_ibdev *dev)
  * Return:
  * A send engine for the qp or NULL for SMI type qp.
  */
-struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5)
+struct sdma_engine *qp_to_sdma_engine(struct rvt_qp *qp, u8 sc5)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
 	struct sdma_engine *sde;
@@ -1554,9 +551,33 @@ struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5)
 	return sde;
 }
 
+/*
+ * qp_to_send_context - map a qp to a send context
+ * @qp: the QP
+ * @sc5: the 5 bit sc
+ *
+ * Return:
+ * A send context for the qp
+ */
+struct send_context *qp_to_send_context(struct rvt_qp *qp, u8 sc5)
+{
+	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_SMI:
+		/* SMA packets to VL15 */
+		return dd->vld[15].sc;
+	default:
+		break;
+	}
+
+	return pio_select_send_context_sc(dd, qp->ibqp.qp_num >> dd->qos_shift,
+					  sc5);
+}
+
 struct qp_iter {
 	struct hfi1_ibdev *dev;
-	struct hfi1_qp *qp;
+	struct rvt_qp *qp;
 	int specials;
 	int n;
 };
@@ -1570,7 +591,7 @@ struct qp_iter *qp_iter_init(struct hfi1_ibdev *dev)
 		return NULL;
 
 	iter->dev = dev;
-	iter->specials = dev->ibdev.phys_port_cnt * 2;
+	iter->specials = dev->rdi.ibdev.phys_port_cnt * 2;
 	if (qp_iter_next(iter)) {
 		kfree(iter);
 		return NULL;
@@ -1584,8 +605,8 @@ int qp_iter_next(struct qp_iter *iter)
 	struct hfi1_ibdev *dev = iter->dev;
 	int n = iter->n;
 	int ret = 1;
-	struct hfi1_qp *pqp = iter->qp;
-	struct hfi1_qp *qp;
+	struct rvt_qp *pqp = iter->qp;
+	struct rvt_qp *qp;
 
 	/*
 	 * The approach is to consider the special qps
@@ -1597,11 +618,11 @@ int qp_iter_next(struct qp_iter *iter)
 	 *
 	 * n = 0..iter->specials is the special qp indices
 	 *
-	 * n = iter->specials..dev->qp_dev->qp_table_size+iter->specials are
+	 * n = iter->specials..dev->rdi.qp_dev->qp_table_size+iter->specials are
 	 * the potential hash bucket entries
 	 *
 	 */
-	for (; n <  dev->qp_dev->qp_table_size + iter->specials; n++) {
+	for (; n <  dev->rdi.qp_dev->qp_table_size + iter->specials; n++) {
 		if (pqp) {
 			qp = rcu_dereference(pqp->next);
 		} else {
@@ -1610,17 +631,17 @@ int qp_iter_next(struct qp_iter *iter)
 				struct hfi1_ibport *ibp;
 				int pidx;
 
-				pidx = n % dev->ibdev.phys_port_cnt;
+				pidx = n % dev->rdi.ibdev.phys_port_cnt;
 				ppd = &dd_from_dev(dev)->pport[pidx];
 				ibp = &ppd->ibport_data;
 
 				if (!(n & 1))
-					qp = rcu_dereference(ibp->qp[0]);
+					qp = rcu_dereference(ibp->rvp.qp[0]);
 				else
-					qp = rcu_dereference(ibp->qp[1]);
+					qp = rcu_dereference(ibp->rvp.qp[1]);
 			} else {
 				qp = rcu_dereference(
-					dev->qp_dev->qp_table[
+					dev->rdi.qp_dev->qp_table[
 						(n - iter->specials)]);
 			}
 		}
@@ -1638,7 +659,7 @@ static const char * const qp_type_str[] = {
 	"SMI", "GSI", "RC", "UC", "UD",
 };
 
-static int qp_idle(struct hfi1_qp *qp)
+static int qp_idle(struct rvt_qp *qp)
 {
 	return
 		qp->s_last == qp->s_acked &&
@@ -1649,14 +670,17 @@ static int qp_idle(struct hfi1_qp *qp)
 
 void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
 {
-	struct hfi1_swqe *wqe;
-	struct hfi1_qp *qp = iter->qp;
+	struct rvt_swqe *wqe;
+	struct rvt_qp *qp = iter->qp;
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct sdma_engine *sde;
+	struct send_context *send_context;
 
-	sde = qp_to_sdma_engine(qp, qp->s_sc);
-	wqe = get_swqe_ptr(qp, qp->s_last);
+	sde = qp_to_sdma_engine(qp, priv->s_sc);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_last);
+	send_context = qp_to_send_context(qp, priv->s_sc);
 	seq_printf(s,
-		   "N %d %s QP%u R %u %s %u %u %u f=%x %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u) QP%u LID %x SL %u MTU %d %u %u %u SDE %p,%u\n",
+		   "N %d %s QP %x R %u %s %u %u %u f=%x %u %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u %u) RQP %x LID %x SL %u MTU %u %u %u %u SDE %p,%u SC %p,%u SCQ %u %u PID %d\n",
 		   iter->n,
 		   qp_idle(qp) ? "I" : "B",
 		   qp->ibqp.qp_num,
@@ -1666,8 +690,9 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
 		   wqe ? wqe->wr.opcode : 0,
 		   qp->s_hdrwords,
 		   qp->s_flags,
-		   atomic_read(&qp->s_iowait.sdma_busy),
-		   !list_empty(&qp->s_iowait.list),
+		   iowait_sdma_pending(&priv->s_iowait),
+		   iowait_pio_pending(&priv->s_iowait),
+		   !list_empty(&priv->s_iowait.list),
 		   qp->timeout,
 		   wqe ? wqe->ssn : 0,
 		   qp->s_lsn,
@@ -1676,20 +701,26 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
 		   qp->s_sending_psn, qp->s_sending_hpsn,
 		   qp->s_last, qp->s_acked, qp->s_cur,
 		   qp->s_tail, qp->s_head, qp->s_size,
+		   qp->s_avail,
 		   qp->remote_qpn,
 		   qp->remote_ah_attr.dlid,
 		   qp->remote_ah_attr.sl,
 		   qp->pmtu,
+		   qp->s_retry,
 		   qp->s_retry_cnt,
-		   qp->timeout,
 		   qp->s_rnr_retry_cnt,
 		   sde,
-		   sde ? sde->this_idx : 0);
+		   sde ? sde->this_idx : 0,
+		   send_context,
+		   send_context ? send_context->sw_index : 0,
+		   ibcq_to_rvtcq(qp->ibqp.send_cq)->queue->head,
+		   ibcq_to_rvtcq(qp->ibqp.send_cq)->queue->tail,
+		   qp->pid);
 }
 
-void qp_comm_est(struct hfi1_qp *qp)
+void qp_comm_est(struct rvt_qp *qp)
 {
-	qp->r_flags |= HFI1_R_COMM_EST;
+	qp->r_flags |= RVT_R_COMM_EST;
 	if (qp->ibqp.event_handler) {
 		struct ib_event ev;
 
@@ -1700,24 +731,241 @@ void qp_comm_est(struct hfi1_qp *qp)
 	}
 }
 
+void *qp_priv_alloc(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+		    gfp_t gfp)
+{
+	struct hfi1_qp_priv *priv;
+
+	priv = kzalloc_node(sizeof(*priv), gfp, rdi->dparms.node);
+	if (!priv)
+		return ERR_PTR(-ENOMEM);
+
+	priv->owner = qp;
+
+	priv->s_hdr = kzalloc_node(sizeof(*priv->s_hdr), gfp, rdi->dparms.node);
+	if (!priv->s_hdr) {
+		kfree(priv);
+		return ERR_PTR(-ENOMEM);
+	}
+	setup_timer(&priv->s_rnr_timer, hfi1_rc_rnr_retry, (unsigned long)qp);
+	qp->s_timer.function = hfi1_rc_timeout;
+	return priv;
+}
+
+void qp_priv_free(struct rvt_dev_info *rdi, struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	kfree(priv->s_hdr);
+	kfree(priv);
+}
+
+unsigned free_all_qps(struct rvt_dev_info *rdi)
+{
+	struct hfi1_ibdev *verbs_dev = container_of(rdi,
+						    struct hfi1_ibdev,
+						    rdi);
+	struct hfi1_devdata *dd = container_of(verbs_dev,
+					       struct hfi1_devdata,
+					       verbs_dev);
+	int n;
+	unsigned qp_inuse = 0;
+
+	for (n = 0; n < dd->num_pports; n++) {
+		struct hfi1_ibport *ibp = &dd->pport[n].ibport_data;
+
+		rcu_read_lock();
+		if (rcu_dereference(ibp->rvp.qp[0]))
+			qp_inuse++;
+		if (rcu_dereference(ibp->rvp.qp[1]))
+			qp_inuse++;
+		rcu_read_unlock();
+	}
+
+	return qp_inuse;
+}
+
+void flush_qp_waiters(struct rvt_qp *qp)
+{
+	flush_iowait(qp);
+	hfi1_stop_rc_timers(qp);
+}
+
+void stop_send_queue(struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	cancel_work_sync(&priv->s_iowait.iowork);
+	hfi1_del_timers_sync(qp);
+}
+
+void quiesce_qp(struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	iowait_sdma_drain(&priv->s_iowait);
+	qp_pio_drain(qp);
+	flush_tx_list(qp);
+}
+
+void notify_qp_reset(struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	iowait_init(
+		&priv->s_iowait,
+		1,
+		_hfi1_do_send,
+		iowait_sleep,
+		iowait_wakeup,
+		iowait_sdma_drained);
+	priv->r_adefered = 0;
+	clear_ahg(qp);
+}
+
 /*
  * Switch to alternate path.
  * The QP s_lock should be held and interrupts disabled.
  */
-void hfi1_migrate_qp(struct hfi1_qp *qp)
+void hfi1_migrate_qp(struct rvt_qp *qp)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct ib_event ev;
 
 	qp->s_mig_state = IB_MIG_MIGRATED;
 	qp->remote_ah_attr = qp->alt_ah_attr;
 	qp->port_num = qp->alt_ah_attr.port_num;
 	qp->s_pkey_index = qp->s_alt_pkey_index;
-	qp->s_flags |= HFI1_S_AHG_CLEAR;
-	qp->s_sc = ah_to_sc(qp->ibqp.device, &qp->remote_ah_attr);
-	qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
+	qp->s_flags |= RVT_S_AHG_CLEAR;
+	priv->s_sc = ah_to_sc(qp->ibqp.device, &qp->remote_ah_attr);
+	priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
 
 	ev.device = qp->ibqp.device;
 	ev.element.qp = &qp->ibqp;
 	ev.event = IB_EVENT_PATH_MIG;
 	qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
 }
+
+int mtu_to_path_mtu(u32 mtu)
+{
+	return mtu_to_enum(mtu, OPA_MTU_8192);
+}
+
+u32 mtu_from_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, u32 pmtu)
+{
+	u32 mtu;
+	struct hfi1_ibdev *verbs_dev = container_of(rdi,
+						    struct hfi1_ibdev,
+						    rdi);
+	struct hfi1_devdata *dd = container_of(verbs_dev,
+					       struct hfi1_devdata,
+					       verbs_dev);
+	struct hfi1_ibport *ibp;
+	u8 sc, vl;
+
+	ibp = &dd->pport[qp->port_num - 1].ibport_data;
+	sc = ibp->sl_to_sc[qp->remote_ah_attr.sl];
+	vl = sc_to_vlt(dd, sc);
+
+	mtu = verbs_mtu_enum_to_int(qp->ibqp.device, pmtu);
+	if (vl < PER_VL_SEND_CONTEXTS)
+		mtu = min_t(u32, mtu, dd->vld[vl].mtu);
+	return mtu;
+}
+
+int get_pmtu_from_attr(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+		       struct ib_qp_attr *attr)
+{
+	int mtu, pidx = qp->port_num - 1;
+	struct hfi1_ibdev *verbs_dev = container_of(rdi,
+						    struct hfi1_ibdev,
+						    rdi);
+	struct hfi1_devdata *dd = container_of(verbs_dev,
+					       struct hfi1_devdata,
+					       verbs_dev);
+	mtu = verbs_mtu_enum_to_int(qp->ibqp.device, attr->path_mtu);
+	if (mtu == -1)
+		return -1; /* values less than 0 are error */
+
+	if (mtu > dd->pport[pidx].ibmtu)
+		return mtu_to_enum(dd->pport[pidx].ibmtu, IB_MTU_2048);
+	else
+		return attr->path_mtu;
+}
+
+void notify_error_qp(struct rvt_qp *qp)
+{
+	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	write_seqlock(&dev->iowait_lock);
+	if (!list_empty(&priv->s_iowait.list) && !(qp->s_flags & RVT_S_BUSY)) {
+		qp->s_flags &= ~RVT_S_ANY_WAIT_IO;
+		list_del_init(&priv->s_iowait.list);
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+	write_sequnlock(&dev->iowait_lock);
+
+	if (!(qp->s_flags & RVT_S_BUSY)) {
+		qp->s_hdrwords = 0;
+		if (qp->s_rdma_mr) {
+			rvt_put_mr(qp->s_rdma_mr);
+			qp->s_rdma_mr = NULL;
+		}
+		flush_tx_list(qp);
+	}
+}
+
+/**
+ * hfi1_error_port_qps - put a port's RC/UC qps into error state
+ * @ibp: the ibport.
+ * @sl: the service level.
+ *
+ * This function places all RC/UC qps with a given service level into error
+ * state. It is generally called to force upper lay apps to abandon stale qps
+ * after an sl->sc mapping change.
+ */
+void hfi1_error_port_qps(struct hfi1_ibport *ibp, u8 sl)
+{
+	struct rvt_qp *qp = NULL;
+	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+	struct hfi1_ibdev *dev = &ppd->dd->verbs_dev;
+	int n;
+	int lastwqe;
+	struct ib_event ev;
+
+	rcu_read_lock();
+
+	/* Deal only with RC/UC qps that use the given SL. */
+	for (n = 0; n < dev->rdi.qp_dev->qp_table_size; n++) {
+		for (qp = rcu_dereference(dev->rdi.qp_dev->qp_table[n]); qp;
+			qp = rcu_dereference(qp->next)) {
+			if (qp->port_num == ppd->port &&
+			    (qp->ibqp.qp_type == IB_QPT_UC ||
+			     qp->ibqp.qp_type == IB_QPT_RC) &&
+			    qp->remote_ah_attr.sl == sl &&
+			    (ib_rvt_state_ops[qp->state] &
+			     RVT_POST_SEND_OK)) {
+				spin_lock_irq(&qp->r_lock);
+				spin_lock(&qp->s_hlock);
+				spin_lock(&qp->s_lock);
+				lastwqe = rvt_error_qp(qp,
+						       IB_WC_WR_FLUSH_ERR);
+				spin_unlock(&qp->s_lock);
+				spin_unlock(&qp->s_hlock);
+				spin_unlock_irq(&qp->r_lock);
+				if (lastwqe) {
+					ev.device = qp->ibqp.device;
+					ev.element.qp = &qp->ibqp;
+					ev.event =
+						IB_EVENT_QP_LAST_WQE_REACHED;
+					qp->ibqp.event_handler(&ev,
+						qp->ibqp.qp_context);
+				}
+			}
+		}
+	}
+
+	rcu_read_unlock();
+}
diff --git a/drivers/staging/rdma/hfi1/qp.h b/drivers/staging/rdma/hfi1/qp.h
index 62a94c5d7dca..e7bc8d6cf681 100644
--- a/drivers/staging/rdma/hfi1/qp.h
+++ b/drivers/staging/rdma/hfi1/qp.h
@@ -1,14 +1,13 @@
 #ifndef _QP_H
 #define _QP_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -51,119 +48,33 @@
  */
 
 #include <linux/hash.h>
+#include <rdma/rdmavt_qp.h>
 #include "verbs.h"
 #include "sdma.h"
 
-#define QPN_MAX                 (1 << 24)
-#define QPNMAP_ENTRIES          (QPN_MAX / PAGE_SIZE / BITS_PER_BYTE)
+extern unsigned int hfi1_qp_table_size;
 
 /*
- * QPN-map pages start out as NULL, they get allocated upon
- * first use and are never deallocated. This way,
- * large bitmaps are not allocated unless large numbers of QPs are used.
- */
-struct qpn_map {
-	void *page;
-};
-
-struct hfi1_qpn_table {
-	spinlock_t lock; /* protect changes in this struct */
-	unsigned flags;         /* flags for QP0/1 allocated for each port */
-	u32 last;               /* last QP number allocated */
-	u32 nmaps;              /* size of the map table */
-	u16 limit;
-	u8  incr;
-	/* bit map of free QP numbers other than 0/1 */
-	struct qpn_map map[QPNMAP_ENTRIES];
-};
-
-struct hfi1_qp_ibdev {
-	u32 qp_table_size;
-	u32 qp_table_bits;
-	struct hfi1_qp __rcu **qp_table;
-	spinlock_t qpt_lock;
-	struct hfi1_qpn_table qpn_table;
-};
-
-static inline u32 qpn_hash(struct hfi1_qp_ibdev *dev, u32 qpn)
-{
-	return hash_32(qpn, dev->qp_table_bits);
-}
-
-/**
- * hfi1_lookup_qpn - return the QP with the given QPN
- * @ibp: the ibport
- * @qpn: the QP number to look up
- *
- * The caller must hold the rcu_read_lock(), and keep the lock until
- * the returned qp is no longer in use.
+ * free_ahg - clear ahg from QP
  */
-static inline struct hfi1_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
-				u32 qpn) __must_hold(RCU)
+static inline void clear_ahg(struct rvt_qp *qp)
 {
-	struct hfi1_qp *qp = NULL;
-
-	if (unlikely(qpn <= 1)) {
-		qp = rcu_dereference(ibp->qp[qpn]);
-	} else {
-		struct hfi1_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
-		u32 n = qpn_hash(dev->qp_dev, qpn);
-
-		for (qp = rcu_dereference(dev->qp_dev->qp_table[n]); qp;
-			qp = rcu_dereference(qp->next))
-			if (qp->ibqp.qp_num == qpn)
-				break;
-	}
-	return qp;
-}
+	struct hfi1_qp_priv *priv = qp->priv;
 
-/**
- * clear_ahg - reset ahg status in qp
- * @qp - qp pointer
- */
-static inline void clear_ahg(struct hfi1_qp *qp)
-{
-	qp->s_hdr->ahgcount = 0;
-	qp->s_flags &= ~(HFI1_S_AHG_VALID | HFI1_S_AHG_CLEAR);
-	if (qp->s_sde && qp->s_ahgidx >= 0)
-		sdma_ahg_free(qp->s_sde, qp->s_ahgidx);
+	priv->s_hdr->ahgcount = 0;
+	qp->s_flags &= ~(RVT_S_AHG_VALID | RVT_S_AHG_CLEAR);
+	if (priv->s_sde && qp->s_ahgidx >= 0)
+		sdma_ahg_free(priv->s_sde, qp->s_ahgidx);
 	qp->s_ahgidx = -1;
 }
 
 /**
- * hfi1_error_qp - put a QP into the error state
- * @qp: the QP to put into the error state
- * @err: the receive completion error to signal if a RWQE is active
- *
- * Flushes both send and receive work queues.
- * Returns true if last WQE event should be generated.
- * The QP r_lock and s_lock should be held and interrupts disabled.
- * If we are already in error state, just return.
- */
-int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err);
-
-/**
- * hfi1_modify_qp - modify the attributes of a queue pair
- * @ibqp: the queue pair who's attributes we're modifying
- * @attr: the new attributes
- * @attr_mask: the mask of attributes to modify
- * @udata: user data for libibverbs.so
- *
- * Returns 0 on success, otherwise returns an errno.
- */
-int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		   int attr_mask, struct ib_udata *udata);
-
-int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
-		  int attr_mask, struct ib_qp_init_attr *init_attr);
-
-/**
  * hfi1_compute_aeth - compute the AETH (syndrome + MSN)
  * @qp: the queue pair to compute the AETH for
  *
  * Returns the AETH.
  */
-__be32 hfi1_compute_aeth(struct hfi1_qp *qp);
+__be32 hfi1_compute_aeth(struct rvt_qp *qp);
 
 /**
  * hfi1_create_qp - create a queue pair for a device
@@ -179,45 +90,23 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
 			     struct ib_qp_init_attr *init_attr,
 			     struct ib_udata *udata);
 /**
- * hfi1_destroy_qp - destroy a queue pair
- * @ibqp: the queue pair to destroy
- *
- * Returns 0 on success.
- *
- * Note that this can be called while the QP is actively sending or
- * receiving!
- */
-int hfi1_destroy_qp(struct ib_qp *ibqp);
-
-/**
  * hfi1_get_credit - flush the send work queue of a QP
  * @qp: the qp who's send work queue to flush
  * @aeth: the Acknowledge Extended Transport Header
  *
  * The QP s_lock should be held.
  */
-void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth);
-
-/**
- * hfi1_qp_init - allocate QP tables
- * @dev: a pointer to the hfi1_ibdev
- */
-int hfi1_qp_init(struct hfi1_ibdev *dev);
-
-/**
- * hfi1_qp_exit - free the QP related structures
- * @dev: a pointer to the hfi1_ibdev
- */
-void hfi1_qp_exit(struct hfi1_ibdev *dev);
+void hfi1_get_credit(struct rvt_qp *qp, u32 aeth);
 
 /**
  * hfi1_qp_wakeup - wake up on the indicated event
  * @qp: the QP
  * @flag: flag the qp on which the qp is stalled
  */
-void hfi1_qp_wakeup(struct hfi1_qp *qp, u32 flag);
+void hfi1_qp_wakeup(struct rvt_qp *qp, u32 flag);
 
-struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5);
+struct sdma_engine *qp_to_sdma_engine(struct rvt_qp *qp, u8 sc5);
+struct send_context *qp_to_send_context(struct rvt_qp *qp, u8 sc5);
 
 struct qp_iter;
 
@@ -244,43 +133,28 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter);
  * qp_comm_est - handle trap with QP established
  * @qp: the QP
  */
-void qp_comm_est(struct hfi1_qp *qp);
+void qp_comm_est(struct rvt_qp *qp);
 
-/**
- * _hfi1_schedule_send - schedule progress
- * @qp: the QP
- *
- * This schedules qp progress w/o regard to the s_flags.
- *
- * It is only used in the post send, which doesn't hold
- * the s_lock.
- */
-static inline void _hfi1_schedule_send(struct hfi1_qp *qp)
-{
-	struct hfi1_ibport *ibp =
-		to_iport(qp->ibqp.device, qp->port_num);
-	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
-	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+void _hfi1_schedule_send(struct rvt_qp *qp);
+void hfi1_schedule_send(struct rvt_qp *qp);
 
-	iowait_schedule(&qp->s_iowait, ppd->hfi1_wq,
-			qp->s_sde ?
-			qp->s_sde->cpu :
-			cpumask_first(cpumask_of_node(dd->assigned_node_id)));
-}
-
-/**
- * hfi1_schedule_send - schedule progress
- * @qp: the QP
- *
- * This schedules qp progress and caller should hold
- * the s_lock.
- */
-static inline void hfi1_schedule_send(struct hfi1_qp *qp)
-{
-	if (hfi1_send_ok(qp))
-		_hfi1_schedule_send(qp);
-}
-
-void hfi1_migrate_qp(struct hfi1_qp *qp);
+void hfi1_migrate_qp(struct rvt_qp *qp);
 
+/*
+ * Functions provided by hfi1 driver for rdmavt to use
+ */
+void *qp_priv_alloc(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+		    gfp_t gfp);
+void qp_priv_free(struct rvt_dev_info *rdi, struct rvt_qp *qp);
+unsigned free_all_qps(struct rvt_dev_info *rdi);
+void notify_qp_reset(struct rvt_qp *qp);
+int get_pmtu_from_attr(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+		       struct ib_qp_attr *attr);
+void flush_qp_waiters(struct rvt_qp *qp);
+void notify_error_qp(struct rvt_qp *qp);
+void stop_send_queue(struct rvt_qp *qp);
+void quiesce_qp(struct rvt_qp *qp);
+u32 mtu_from_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, u32 pmtu);
+int mtu_to_path_mtu(u32 mtu);
+void hfi1_error_port_qps(struct hfi1_ibport *ibp, u8 sl);
 #endif /* _QP_H */
diff --git a/drivers/staging/rdma/hfi1/qsfp.c b/drivers/staging/rdma/hfi1/qsfp.c
index 6326a915d7fd..9ed1963010fe 100644
--- a/drivers/staging/rdma/hfi1/qsfp.c
+++ b/drivers/staging/rdma/hfi1/qsfp.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -62,7 +59,7 @@
 #define I2C_MAX_RETRY 4
 
 /*
- * Unlocked i2c write.  Must hold dd->qsfp_i2c_mutex.
+ * Raw i2c write.  No set-up or lock checking.
  */
 static int __i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr,
 		       int offset, void *bp, int len)
@@ -71,14 +68,6 @@ static int __i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr,
 	int ret, cnt;
 	u8 *buff = bp;
 
-	/* Make sure TWSI bus is in sane state. */
-	ret = hfi1_twsi_reset(dd, target);
-	if (ret) {
-		hfi1_dev_porterr(dd, ppd->port,
-				 "I2C interface Reset for write failed\n");
-		return -EIO;
-	}
-
 	cnt = 0;
 	while (cnt < len) {
 		int wlen = len - cnt;
@@ -99,48 +88,45 @@ static int __i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr,
 	return cnt;
 }
 
+/*
+ * Caller must hold the i2c chain resource.
+ */
 int i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 	      void *bp, int len)
 {
-	struct hfi1_devdata *dd = ppd->dd;
 	int ret;
 
-	ret = mutex_lock_interruptible(&dd->qsfp_i2c_mutex);
-	if (!ret) {
-		ret = __i2c_write(ppd, target, i2c_addr, offset, bp, len);
-		mutex_unlock(&dd->qsfp_i2c_mutex);
+	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+		return -EACCES;
+
+	/* make sure the TWSI bus is in a sane state */
+	ret = hfi1_twsi_reset(ppd->dd, target);
+	if (ret) {
+		hfi1_dev_porterr(ppd->dd, ppd->port,
+				 "I2C chain %d write interface reset failed\n",
+				 target);
+		return ret;
 	}
 
-	return ret;
+	return __i2c_write(ppd, target, i2c_addr, offset, bp, len);
 }
 
 /*
- * Unlocked i2c read.  Must hold dd->qsfp_i2c_mutex.
+ * Raw i2c read.  No set-up or lock checking.
  */
 static int __i2c_read(struct hfi1_pportdata *ppd, u32 target, int i2c_addr,
 		      int offset, void *bp, int len)
 {
 	struct hfi1_devdata *dd = ppd->dd;
 	int ret, cnt, pass = 0;
-	int stuck = 0;
-	u8 *buff = bp;
-
-	/* Make sure TWSI bus is in sane state. */
-	ret = hfi1_twsi_reset(dd, target);
-	if (ret) {
-		hfi1_dev_porterr(dd, ppd->port,
-				 "I2C interface Reset for read failed\n");
-		ret = -EIO;
-		stuck = 1;
-		goto exit;
-	}
+	int orig_offset = offset;
 
 	cnt = 0;
 	while (cnt < len) {
 		int rlen = len - cnt;
 
 		ret = hfi1_twsi_blk_rd(dd, target, i2c_addr, offset,
-				       buff + cnt, rlen);
+				       bp + cnt, rlen);
 		/* Some QSFP's fail first try. Retry as experiment */
 		if (ret && cnt == 0 && ++pass < I2C_MAX_RETRY)
 			continue;
@@ -156,14 +142,11 @@ static int __i2c_read(struct hfi1_pportdata *ppd, u32 target, int i2c_addr,
 	ret = cnt;
 
 exit:
-	if (stuck)
-		dd_dev_err(dd, "I2C interface bus stuck non-idle\n");
-
-	if (pass >= I2C_MAX_RETRY && ret)
+	if (ret < 0) {
 		hfi1_dev_porterr(dd, ppd->port,
-				 "I2C failed even retrying\n");
-	else if (pass)
-		hfi1_dev_porterr(dd, ppd->port, "I2C retries: %d\n", pass);
+				 "I2C chain %d read failed, addr 0x%x, offset 0x%x, len %d\n",
+				 target, i2c_addr, orig_offset, len);
+	}
 
 	/* Must wait min 20us between qsfp i2c transactions */
 	udelay(20);
@@ -171,21 +154,35 @@ exit:
 	return ret;
 }
 
+/*
+ * Caller must hold the i2c chain resource.
+ */
 int i2c_read(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 	     void *bp, int len)
 {
-	struct hfi1_devdata *dd = ppd->dd;
 	int ret;
 
-	ret = mutex_lock_interruptible(&dd->qsfp_i2c_mutex);
-	if (!ret) {
-		ret = __i2c_read(ppd, target, i2c_addr, offset, bp, len);
-		mutex_unlock(&dd->qsfp_i2c_mutex);
+	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+		return -EACCES;
+
+	/* make sure the TWSI bus is in a sane state */
+	ret = hfi1_twsi_reset(ppd->dd, target);
+	if (ret) {
+		hfi1_dev_porterr(ppd->dd, ppd->port,
+				 "I2C chain %d read interface reset failed\n",
+				 target);
+		return ret;
 	}
 
-	return ret;
+	return __i2c_read(ppd, target, i2c_addr, offset, bp, len);
 }
 
+/*
+ * Write page n, offset m of QSFP memory as defined by SFF 8636
+ * by writing @addr = ((256 * n) + m)
+ *
+ * Caller must hold the i2c chain resource.
+ */
 int qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	       int len)
 {
@@ -195,50 +192,81 @@ int qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	ret = mutex_lock_interruptible(&ppd->dd->qsfp_i2c_mutex);
-	if (ret)
+	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+		return -EACCES;
+
+	/* make sure the TWSI bus is in a sane state */
+	ret = hfi1_twsi_reset(ppd->dd, target);
+	if (ret) {
+		hfi1_dev_porterr(ppd->dd, ppd->port,
+				 "QSFP chain %d write interface reset failed\n",
+				 target);
 		return ret;
+	}
 
 	while (count < len) {
 		/*
-		 * Set the qsfp page based on a zero-based addresss
+		 * Set the qsfp page based on a zero-based address
 		 * and a page size of QSFP_PAGESIZE bytes.
 		 */
 		page = (u8)(addr / QSFP_PAGESIZE);
 
-		ret = __i2c_write(ppd, target, QSFP_DEV,
-					QSFP_PAGE_SELECT_BYTE_OFFS, &page, 1);
+		ret = __i2c_write(ppd, target, QSFP_DEV | QSFP_OFFSET_SIZE,
+				  QSFP_PAGE_SELECT_BYTE_OFFS, &page, 1);
 		if (ret != 1) {
-			hfi1_dev_porterr(
-			ppd->dd,
-			ppd->port,
-			"can't write QSFP_PAGE_SELECT_BYTE: %d\n", ret);
+			hfi1_dev_porterr(ppd->dd, ppd->port,
+					 "QSFP chain %d can't write QSFP_PAGE_SELECT_BYTE: %d\n",
+					 target, ret);
 			ret = -EIO;
 			break;
 		}
 
-		/* truncate write to end of page if crossing page boundary */
 		offset = addr % QSFP_PAGESIZE;
 		nwrite = len - count;
-		if ((offset + nwrite) > QSFP_PAGESIZE)
-			nwrite = QSFP_PAGESIZE - offset;
+		/* truncate write to boundary if crossing boundary */
+		if (((addr % QSFP_RW_BOUNDARY) + nwrite) > QSFP_RW_BOUNDARY)
+			nwrite = QSFP_RW_BOUNDARY - (addr % QSFP_RW_BOUNDARY);
 
-		ret = __i2c_write(ppd, target, QSFP_DEV, offset, bp + count,
-					nwrite);
-		if (ret <= 0)	/* stop on error or nothing read */
+		ret = __i2c_write(ppd, target, QSFP_DEV | QSFP_OFFSET_SIZE,
+				  offset, bp + count, nwrite);
+		if (ret <= 0)	/* stop on error or nothing written */
 			break;
 
 		count += ret;
 		addr += ret;
 	}
 
-	mutex_unlock(&ppd->dd->qsfp_i2c_mutex);
-
 	if (ret < 0)
 		return ret;
 	return count;
 }
 
+/*
+ * Perform a stand-alone single QSFP write.  Acquire the resource, do the
+ * read, then release the resource.
+ */
+int one_qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
+		   int len)
+{
+	struct hfi1_devdata *dd = ppd->dd;
+	u32 resource = qsfp_resource(dd);
+	int ret;
+
+	ret = acquire_chip_resource(dd, resource, QSFP_WAIT);
+	if (ret)
+		return ret;
+	ret = qsfp_write(ppd, target, addr, bp, len);
+	release_chip_resource(dd, resource);
+
+	return ret;
+}
+
+/*
+ * Access page n, offset m of QSFP memory as defined by SFF 8636
+ * by reading @addr = ((256 * n) + m)
+ *
+ * Caller must hold the i2c chain resource.
+ */
 int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	      int len)
 {
@@ -248,9 +276,17 @@ int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	ret = mutex_lock_interruptible(&ppd->dd->qsfp_i2c_mutex);
-	if (ret)
+	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+		return -EACCES;
+
+	/* make sure the TWSI bus is in a sane state */
+	ret = hfi1_twsi_reset(ppd->dd, target);
+	if (ret) {
+		hfi1_dev_porterr(ppd->dd, ppd->port,
+				 "QSFP chain %d read interface reset failed\n",
+				 target);
 		return ret;
+	}
 
 	while (count < len) {
 		/*
@@ -258,25 +294,26 @@ int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 		 * and a page size of QSFP_PAGESIZE bytes.
 		 */
 		page = (u8)(addr / QSFP_PAGESIZE);
-		ret = __i2c_write(ppd, target, QSFP_DEV,
-					QSFP_PAGE_SELECT_BYTE_OFFS, &page, 1);
+		ret = __i2c_write(ppd, target, QSFP_DEV | QSFP_OFFSET_SIZE,
+				  QSFP_PAGE_SELECT_BYTE_OFFS, &page, 1);
 		if (ret != 1) {
-			hfi1_dev_porterr(
-			ppd->dd,
-			ppd->port,
-			"can't write QSFP_PAGE_SELECT_BYTE: %d\n", ret);
+			hfi1_dev_porterr(ppd->dd, ppd->port,
+					 "QSFP chain %d can't write QSFP_PAGE_SELECT_BYTE: %d\n",
+					 target, ret);
 			ret = -EIO;
 			break;
 		}
 
-		/* truncate read to end of page if crossing page boundary */
 		offset = addr % QSFP_PAGESIZE;
 		nread = len - count;
-		if ((offset + nread) > QSFP_PAGESIZE)
-			nread = QSFP_PAGESIZE - offset;
-
-		ret = __i2c_read(ppd, target, QSFP_DEV, offset, bp + count,
-					nread);
+		/* truncate read to boundary if crossing boundary */
+		if (((addr % QSFP_RW_BOUNDARY) + nread) > QSFP_RW_BOUNDARY)
+			nread = QSFP_RW_BOUNDARY - (addr % QSFP_RW_BOUNDARY);
+
+		/* QSFPs require a 5-10msec delay after write operations */
+		mdelay(5);
+		ret = __i2c_read(ppd, target, QSFP_DEV | QSFP_OFFSET_SIZE,
+				 offset, bp + count, nread);
 		if (ret <= 0)	/* stop on error or nothing read */
 			break;
 
@@ -284,17 +321,40 @@ int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 		addr += ret;
 	}
 
-	mutex_unlock(&ppd->dd->qsfp_i2c_mutex);
-
 	if (ret < 0)
 		return ret;
 	return count;
 }
 
 /*
+ * Perform a stand-alone single QSFP read.  Acquire the resource, do the
+ * read, then release the resource.
+ */
+int one_qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
+		  int len)
+{
+	struct hfi1_devdata *dd = ppd->dd;
+	u32 resource = qsfp_resource(dd);
+	int ret;
+
+	ret = acquire_chip_resource(dd, resource, QSFP_WAIT);
+	if (ret)
+		return ret;
+	ret = qsfp_read(ppd, target, addr, bp, len);
+	release_chip_resource(dd, resource);
+
+	return ret;
+}
+
+/*
  * This function caches the QSFP memory range in 128 byte chunks.
  * As an example, the next byte after address 255 is byte 128 from
  * upper page 01H (if existing) rather than byte 0 from lower page 00H.
+ * Access page n, offset m of QSFP memory as defined by SFF 8636
+ * in the cache by reading byte ((128 * n) + m)
+ * The calls to qsfp_{read,write} in this function correctly handle the
+ * address map difference between this mapping and the mapping implemented
+ * by those functions
  */
 int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 {
@@ -304,79 +364,84 @@ int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 	u8 *cache = &cp->cache[0];
 
 	/* ensure sane contents on invalid reads, for cable swaps */
-	memset(cache, 0, (QSFP_MAX_NUM_PAGES*128));
-	dd_dev_info(ppd->dd, "%s: called\n", __func__);
+	memset(cache, 0, (QSFP_MAX_NUM_PAGES * 128));
+	spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
+	ppd->qsfp_info.cache_valid = 0;
+	spin_unlock_irqrestore(&ppd->qsfp_info.qsfp_lock, flags);
+
 	if (!qsfp_mod_present(ppd)) {
 		ret = -ENODEV;
-		goto bail;
+		goto bail_no_release;
 	}
 
-	ret = qsfp_read(ppd, target, 0, cache, 256);
-	if (ret != 256) {
+	ret = acquire_chip_resource(ppd->dd, qsfp_resource(ppd->dd), QSFP_WAIT);
+	if (ret)
+		goto bail_no_release;
+
+	ret = qsfp_read(ppd, target, 0, cache, QSFP_PAGESIZE);
+	if (ret != QSFP_PAGESIZE) {
 		dd_dev_info(ppd->dd,
-			"%s: Read of pages 00H failed, expected 256, got %d\n",
-			__func__, ret);
+			    "%s: Page 0 read failed, expected %d, got %d\n",
+			    __func__, QSFP_PAGESIZE, ret);
 		goto bail;
 	}
 
-	if (cache[0] != 0x0C && cache[0] != 0x0D)
-		goto bail;
-
 	/* Is paging enabled? */
 	if (!(cache[2] & 4)) {
-
 		/* Paging enabled, page 03 required */
 		if ((cache[195] & 0xC0) == 0xC0) {
 			/* all */
 			ret = qsfp_read(ppd, target, 384, cache + 256, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 			ret = qsfp_read(ppd, target, 640, cache + 384, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 			ret = qsfp_read(ppd, target, 896, cache + 512, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 		} else if ((cache[195] & 0x80) == 0x80) {
 			/* only page 2 and 3 */
 			ret = qsfp_read(ppd, target, 640, cache + 384, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 			ret = qsfp_read(ppd, target, 896, cache + 512, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 		} else if ((cache[195] & 0x40) == 0x40) {
 			/* only page 1 and 3 */
 			ret = qsfp_read(ppd, target, 384, cache + 256, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 			ret = qsfp_read(ppd, target, 896, cache + 512, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 		} else {
 			/* only page 3 */
 			ret = qsfp_read(ppd, target, 896, cache + 512, 128);
 			if (ret <= 0 || ret != 128) {
-				dd_dev_info(ppd->dd, "%s: failed\n", __func__);
+				dd_dev_info(ppd->dd, "%s failed\n", __func__);
 				goto bail;
 			}
 		}
 	}
 
+	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
+
 	spin_lock_irqsave(&ppd->qsfp_info.qsfp_lock, flags);
 	ppd->qsfp_info.cache_valid = 1;
 	ppd->qsfp_info.cache_refresh_required = 0;
@@ -385,7 +450,9 @@ int refresh_qsfp_cache(struct hfi1_pportdata *ppd, struct qsfp_data *cp)
 	return 0;
 
 bail:
-	memset(cache, 0, (QSFP_MAX_NUM_PAGES*128));
+	release_chip_resource(ppd->dd, qsfp_resource(ppd->dd));
+bail_no_release:
+	memset(cache, 0, (QSFP_MAX_NUM_PAGES * 128));
 	return ret;
 }
 
@@ -434,7 +501,7 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
 
 	if (port_num > dd->num_pports || port_num < 1) {
 		dd_dev_info(dd, "%s: Invalid port number %d\n",
-				__func__, port_num);
+			    __func__, port_num);
 		ret = -EINVAL;
 		goto set_zeroes;
 	}
@@ -485,7 +552,6 @@ int qsfp_dump(struct hfi1_pportdata *ppd, char *buf, int len)
 	lenstr[1] = '\0';
 
 	if (ppd->qsfp_info.cache_valid) {
-
 		if (QSFP_IS_CU(cache[QSFP_MOD_TECH_OFFS]))
 			sprintf(lenstr, "%dM ", cache[QSFP_MOD_LEN_OFFS]);
 
@@ -529,7 +595,7 @@ int qsfp_dump(struct hfi1_pportdata *ppd, char *buf, int len)
 
 			memcpy(bin_buff, &cache[bidx], QSFP_DUMP_CHUNK);
 			for (iidx = 0; iidx < QSFP_DUMP_CHUNK; ++iidx) {
-				sofar += scnprintf(buf + sofar, len-sofar,
+				sofar += scnprintf(buf + sofar, len - sofar,
 					" %02X", bin_buff[iidx]);
 			}
 			sofar += scnprintf(buf + sofar, len - sofar, "\n");
diff --git a/drivers/staging/rdma/hfi1/qsfp.h b/drivers/staging/rdma/hfi1/qsfp.h
index d30c2a6baa0b..831fe4cf1345 100644
--- a/drivers/staging/rdma/hfi1/qsfp.h
+++ b/drivers/staging/rdma/hfi1/qsfp.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -59,23 +56,28 @@
  * Below are masks for QSFP pins.  Pins are the same for HFI0 and HFI1.
  * _N means asserted low
  */
-#define QSFP_HFI0_I2CCLK    (1 << 0)
-#define QSFP_HFI0_I2CDAT    (1 << 1)
-#define QSFP_HFI0_RESET_N   (1 << 2)
-#define QSFP_HFI0_INT_N	    (1 << 3)
-#define QSFP_HFI0_MODPRST_N (1 << 4)
+#define QSFP_HFI0_I2CCLK    BIT(0)
+#define QSFP_HFI0_I2CDAT    BIT(1)
+#define QSFP_HFI0_RESET_N   BIT(2)
+#define QSFP_HFI0_INT_N	    BIT(3)
+#define QSFP_HFI0_MODPRST_N BIT(4)
 
 /* QSFP is paged at 256 bytes */
 #define QSFP_PAGESIZE 256
+/* Reads/writes cannot cross 128 byte boundaries */
+#define QSFP_RW_BOUNDARY 128
+
+/* number of bytes in i2c offset for QSFP devices */
+#define __QSFP_OFFSET_SIZE 1                           /* num address bytes */
+#define QSFP_OFFSET_SIZE (__QSFP_OFFSET_SIZE << 8)     /* shifted value */
 
 /* Defined fields that Intel requires of qualified cables */
 /* Byte 0 is Identifier, not checked */
 /* Byte 1 is reserved "status MSB" */
-/* Byte 2 is "status LSB" We only care that D2 "Flat Mem" is set. */
-/*
- * Rest of first 128 not used, although 127 is reserved for page select
- * if module is not "Flat memory".
- */
+#define QSFP_TX_CTRL_BYTE_OFFS 86
+#define QSFP_PWR_CTRL_BYTE_OFFS 93
+#define QSFP_CDR_CTRL_BYTE_OFFS 98
+
 #define QSFP_PAGE_SELECT_BYTE_OFFS 127
 /* Byte 128 is Identifier: must be 0x0c for QSFP, or 0x0d for QSFP+ */
 #define QSFP_MOD_ID_OFFS 128
@@ -87,7 +89,8 @@
 /* Byte 130 is Connector type. Not Intel req'd */
 /* Bytes 131..138 are Transceiver types, bit maps for various tech, none IB */
 /* Byte 139 is encoding. code 0x01 is 8b10b. Not Intel req'd */
-/* byte 140 is nominal bit-rate, in units of 100Mbits/sec Not Intel req'd */
+/* byte 140 is nominal bit-rate, in units of 100Mbits/sec */
+#define QSFP_NOM_BIT_RATE_100_OFFS 140
 /* Byte 141 is Extended Rate Select. Not Intel req'd */
 /* Bytes 142..145 are lengths for various fiber types. Not Intel req'd */
 /* Byte 146 is length for Copper. Units of 1 meter */
@@ -135,11 +138,18 @@ extern const char *const hfi1_qsfp_devtech[16];
  */
 #define QSFP_ATTEN_OFFS 186
 #define QSFP_ATTEN_LEN 2
-/* Bytes 188,189 are Wavelength tolerance, not Intel req'd */
+/*
+ * Bytes 188,189 are Wavelength tolerance, if optical
+ * If copper, they are attenuation in dB:
+ * Byte 188 is at 12.5 Gb/s, Byte 189 at 25 Gb/s
+ */
+#define QSFP_CU_ATTEN_7G_OFFS 188
+#define QSFP_CU_ATTEN_12G_OFFS 189
 /* Byte 190 is Max Case Temp. Not Intel req'd */
 /* Byte 191 is LSB of sum of bytes 128..190. Not Intel req'd */
 #define QSFP_CC_OFFS 191
-/* Bytes 192..195 are Options implemented in qsfp. Not Intel req'd */
+#define QSFP_EQ_INFO_OFFS 193
+#define QSFP_CDR_INFO_OFFS 194
 /* Bytes 196..211 are Serial Number, String */
 #define QSFP_SN_OFFS 196
 #define QSFP_SN_LEN 16
@@ -150,6 +160,8 @@ extern const char *const hfi1_qsfp_devtech[16];
 #define QSFP_LOT_OFFS 218
 #define QSFP_LOT_LEN 2
 /* Bytes 220, 221 indicate monitoring options, Not Intel req'd */
+/* Byte 222 indicates nominal bitrate in units of 250Mbits/sec */
+#define QSFP_NOM_BIT_RATE_250_OFFS 222
 /* Byte 223 is LSB of sum of bytes 192..222 */
 #define QSFP_CC_EXT_OFFS 223
 
@@ -191,6 +203,7 @@ extern const char *const hfi1_qsfp_devtech[16];
  */
 
 #define QSFP_PWR(pbyte) (((pbyte) >> 6) & 3)
+#define QSFP_HIGH_PWR(pbyte) (((pbyte) & 3) | 4)
 #define QSFP_ATTEN_SDR(attenarray) (attenarray[0])
 #define QSFP_ATTEN_DDR(attenarray) (attenarray[1])
 
@@ -198,10 +211,12 @@ struct qsfp_data {
 	/* Helps to find our way */
 	struct hfi1_pportdata *ppd;
 	struct work_struct qsfp_work;
-	u8 cache[QSFP_MAX_NUM_PAGES*128];
+	u8 cache[QSFP_MAX_NUM_PAGES * 128];
+	/* protect qsfp data */
 	spinlock_t qsfp_lock;
 	u8 check_interrupt_flags;
-	u8 qsfp_interrupt_functional;
+	u8 reset_needed;
+	u8 limiting_active;
 	u8 cache_valid;
 	u8 cache_refresh_required;
 };
@@ -220,3 +235,7 @@ int qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	       int len);
 int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	      int len);
+int one_qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
+		   int len);
+int one_qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
+		  int len);
diff --git a/drivers/staging/rdma/hfi1/rc.c b/drivers/staging/rdma/hfi1/rc.c
index 6f4a155f7931..0d7e1017f3cb 100644
--- a/drivers/staging/rdma/hfi1/rc.c
+++ b/drivers/staging/rdma/hfi1/rc.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -49,18 +46,151 @@
  */
 
 #include <linux/io.h>
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_qp.h>
 
 #include "hfi.h"
 #include "qp.h"
-#include "sdma.h"
+#include "verbs_txreq.h"
 #include "trace.h"
 
 /* cut down ridiculously long IB macro names */
 #define OP(x) IB_OPCODE_RC_##x
 
-static void rc_timeout(unsigned long arg);
+/**
+ * hfi1_add_retry_timer - add/start a retry timer
+ * @qp - the QP
+ *
+ * add a retry timer on the QP
+ */
+static inline void hfi1_add_retry_timer(struct rvt_qp *qp)
+{
+	struct ib_qp *ibqp = &qp->ibqp;
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+
+	qp->s_flags |= RVT_S_TIMER;
+	/* 4.096 usec. * (1 << qp->timeout) */
+	qp->s_timer.expires = jiffies + qp->timeout_jiffies +
+			      rdi->busy_jiffies;
+	add_timer(&qp->s_timer);
+}
+
+/**
+ * hfi1_add_rnr_timer - add/start an rnr timer
+ * @qp - the QP
+ * @to - timeout in usecs
+ *
+ * add an rnr timer on the QP
+ */
+void hfi1_add_rnr_timer(struct rvt_qp *qp, u32 to)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	qp->s_flags |= RVT_S_WAIT_RNR;
+	qp->s_timer.expires = jiffies + usecs_to_jiffies(to);
+	add_timer(&priv->s_rnr_timer);
+}
+
+/**
+ * hfi1_mod_retry_timer - mod a retry timer
+ * @qp - the QP
+ *
+ * Modify a potentially already running retry
+ * timer
+ */
+static inline void hfi1_mod_retry_timer(struct rvt_qp *qp)
+{
+	struct ib_qp *ibqp = &qp->ibqp;
+	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
+
+	qp->s_flags |= RVT_S_TIMER;
+	/* 4.096 usec. * (1 << qp->timeout) */
+	mod_timer(&qp->s_timer, jiffies + qp->timeout_jiffies +
+		  rdi->busy_jiffies);
+}
+
+/**
+ * hfi1_stop_retry_timer - stop a retry timer
+ * @qp - the QP
+ *
+ * stop a retry timer and return if the timer
+ * had been pending.
+ */
+static inline int hfi1_stop_retry_timer(struct rvt_qp *qp)
+{
+	int rval = 0;
+
+	/* Remove QP from retry */
+	if (qp->s_flags & RVT_S_TIMER) {
+		qp->s_flags &= ~RVT_S_TIMER;
+		rval = del_timer(&qp->s_timer);
+	}
+	return rval;
+}
+
+/**
+ * hfi1_stop_rc_timers - stop all timers
+ * @qp - the QP
+ *
+ * stop any pending timers
+ */
+void hfi1_stop_rc_timers(struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	/* Remove QP from all timers */
+	if (qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR)) {
+		qp->s_flags &= ~(RVT_S_TIMER | RVT_S_WAIT_RNR);
+		del_timer(&qp->s_timer);
+		del_timer(&priv->s_rnr_timer);
+	}
+}
+
+/**
+ * hfi1_stop_rnr_timer - stop an rnr timer
+ * @qp - the QP
+ *
+ * stop an rnr timer and return if the timer
+ * had been pending.
+ */
+static inline int hfi1_stop_rnr_timer(struct rvt_qp *qp)
+{
+	int rval = 0;
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	/* Remove QP from rnr timer */
+	if (qp->s_flags & RVT_S_WAIT_RNR) {
+		qp->s_flags &= ~RVT_S_WAIT_RNR;
+		rval = del_timer(&priv->s_rnr_timer);
+	}
+	return rval;
+}
+
+/**
+ * hfi1_del_timers_sync - wait for any timeout routines to exit
+ * @qp - the QP
+ */
+void hfi1_del_timers_sync(struct rvt_qp *qp)
+{
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	del_timer_sync(&qp->s_timer);
+	del_timer_sync(&priv->s_rnr_timer);
+}
 
-static u32 restart_sge(struct hfi1_sge_state *ss, struct hfi1_swqe *wqe,
+/* only opcode mask for adaptive pio */
+const u32 rc_only_opcode =
+	BIT(OP(SEND_ONLY) & 0x1f) |
+	BIT(OP(SEND_ONLY_WITH_IMMEDIATE & 0x1f)) |
+	BIT(OP(RDMA_WRITE_ONLY & 0x1f)) |
+	BIT(OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE & 0x1f)) |
+	BIT(OP(RDMA_READ_REQUEST & 0x1f)) |
+	BIT(OP(ACKNOWLEDGE & 0x1f)) |
+	BIT(OP(ATOMIC_ACKNOWLEDGE & 0x1f)) |
+	BIT(OP(COMPARE_SWAP & 0x1f)) |
+	BIT(OP(FETCH_ADD & 0x1f));
+
+static u32 restart_sge(struct rvt_sge_state *ss, struct rvt_swqe *wqe,
 		       u32 psn, u32 pmtu)
 {
 	u32 len;
@@ -74,38 +204,32 @@ static u32 restart_sge(struct hfi1_sge_state *ss, struct hfi1_swqe *wqe,
 	return wqe->length - len;
 }
 
-static void start_timer(struct hfi1_qp *qp)
-{
-	qp->s_flags |= HFI1_S_TIMER;
-	qp->s_timer.function = rc_timeout;
-	/* 4.096 usec. * (1 << qp->timeout) */
-	qp->s_timer.expires = jiffies + qp->timeout_jiffies;
-	add_timer(&qp->s_timer);
-}
-
 /**
  * make_rc_ack - construct a response packet (ACK, NAK, or RDMA read)
  * @dev: the device for this QP
  * @qp: a pointer to the QP
  * @ohdr: a pointer to the IB header being constructed
- * @pmtu: the path MTU
+ * @ps: the xmit packet state
  *
  * Return 1 if constructed; otherwise, return 0.
  * Note that we are in the responder's side of the QP context.
  * Note the QP s_lock must be held.
  */
-static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
-		       struct hfi1_other_headers *ohdr, u32 pmtu)
+static int make_rc_ack(struct hfi1_ibdev *dev, struct rvt_qp *qp,
+		       struct hfi1_other_headers *ohdr,
+		       struct hfi1_pkt_state *ps)
 {
-	struct hfi1_ack_entry *e;
+	struct rvt_ack_entry *e;
 	u32 hwords;
 	u32 len;
 	u32 bth0;
 	u32 bth2;
 	int middle = 0;
+	u32 pmtu = qp->pmtu;
+	struct hfi1_qp_priv *priv = qp->priv;
 
 	/* Don't send an ACK if we aren't supposed to. */
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
 		goto bail;
 
 	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
@@ -116,7 +240,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
 	case OP(RDMA_READ_RESPONSE_ONLY):
 		e = &qp->s_ack_queue[qp->s_tail_ack_queue];
 		if (e->rdma_sge.mr) {
-			hfi1_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		/* FALLTHROUGH */
@@ -133,7 +257,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
 	case OP(ACKNOWLEDGE):
 		/* Check for no next entry in the queue. */
 		if (qp->r_head_ack_queue == qp->s_tail_ack_queue) {
-			if (qp->s_flags & HFI1_S_ACK_PENDING)
+			if (qp->s_flags & RVT_S_ACK_PENDING)
 				goto normal;
 			goto bail;
 		}
@@ -152,9 +276,9 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
 				goto bail;
 			}
 			/* Copy SGE state in case we need to resend */
-			qp->s_rdma_mr = e->rdma_sge.mr;
-			if (qp->s_rdma_mr)
-				hfi1_get_mr(qp->s_rdma_mr);
+			ps->s_txreq->mr = e->rdma_sge.mr;
+			if (ps->s_txreq->mr)
+				rvt_get_mr(ps->s_txreq->mr);
 			qp->s_ack_rdma_sge.sge = e->rdma_sge;
 			qp->s_ack_rdma_sge.num_sge = 1;
 			qp->s_cur_sge = &qp->s_ack_rdma_sge;
@@ -191,9 +315,9 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
 		/* FALLTHROUGH */
 	case OP(RDMA_READ_RESPONSE_MIDDLE):
 		qp->s_cur_sge = &qp->s_ack_rdma_sge;
-		qp->s_rdma_mr = qp->s_ack_rdma_sge.sge.mr;
-		if (qp->s_rdma_mr)
-			hfi1_get_mr(qp->s_rdma_mr);
+		ps->s_txreq->mr = qp->s_ack_rdma_sge.sge.mr;
+		if (ps->s_txreq->mr)
+			rvt_get_mr(ps->s_txreq->mr);
 		len = qp->s_ack_rdma_sge.sge.sge_length;
 		if (len > pmtu) {
 			len = pmtu;
@@ -218,7 +342,7 @@ normal:
 		 * (see above).
 		 */
 		qp->s_ack_state = OP(SEND_ONLY);
-		qp->s_flags &= ~HFI1_S_ACK_PENDING;
+		qp->s_flags &= ~RVT_S_ACK_PENDING;
 		qp->s_cur_sge = NULL;
 		if (qp->s_nak_state)
 			ohdr->u.aeth =
@@ -234,20 +358,23 @@ normal:
 	}
 	qp->s_rdma_ack_cnt++;
 	qp->s_hdrwords = hwords;
+	ps->s_txreq->sde = priv->s_sde;
 	qp->s_cur_size = len;
-	hfi1_make_ruc_header(qp, ohdr, bth0, bth2, middle);
+	hfi1_make_ruc_header(qp, ohdr, bth0, bth2, middle, ps);
+	/* pbc */
+	ps->s_txreq->hdr_dwords = qp->s_hdrwords + 2;
 	return 1;
 
 bail:
 	qp->s_ack_state = OP(ACKNOWLEDGE);
 	/*
 	 * Ensure s_rdma_ack_cnt changes are committed prior to resetting
-	 * HFI1_S_RESP_PENDING
+	 * RVT_S_RESP_PENDING
 	 */
 	smp_wmb();
-	qp->s_flags &= ~(HFI1_S_RESP_PENDING
-				| HFI1_S_ACK_PENDING
-				| HFI1_S_AHG_VALID);
+	qp->s_flags &= ~(RVT_S_RESP_PENDING
+				| RVT_S_ACK_PENDING
+				| RVT_S_AHG_VALID);
 	return 0;
 }
 
@@ -255,14 +382,17 @@ bail:
  * hfi1_make_rc_req - construct a request packet (SEND, RDMA r/w, ATOMIC)
  * @qp: a pointer to the QP
  *
+ * Assumes s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int hfi1_make_rc_req(struct hfi1_qp *qp)
+int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
 	struct hfi1_other_headers *ohdr;
-	struct hfi1_sge_state *ss;
-	struct hfi1_swqe *wqe;
+	struct rvt_sge_state *ss;
+	struct rvt_swqe *wqe;
 	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
 	u32 hwords = 5;
 	u32 len;
@@ -270,51 +400,48 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 	u32 bth2;
 	u32 pmtu = qp->pmtu;
 	char newreq;
-	unsigned long flags;
-	int ret = 0;
 	int middle = 0;
 	int delta;
 
-	ohdr = &qp->s_hdr->ibh.u.oth;
-	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
-		ohdr = &qp->s_hdr->ibh.u.l.oth;
+	ps->s_txreq = get_txreq(ps->dev, qp);
+	if (IS_ERR(ps->s_txreq))
+		goto bail_no_tx;
 
-	/*
-	 * The lock is needed to synchronize between the sending tasklet,
-	 * the receive interrupt handler, and timeout re-sends.
-	 */
-	spin_lock_irqsave(&qp->s_lock, flags);
+	ohdr = &ps->s_txreq->phdr.hdr.u.oth;
+	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
+		ohdr = &ps->s_txreq->phdr.hdr.u.l.oth;
 
 	/* Sending responses has higher priority over sending requests. */
-	if ((qp->s_flags & HFI1_S_RESP_PENDING) &&
-	    make_rc_ack(dev, qp, ohdr, pmtu))
-		goto done;
+	if ((qp->s_flags & RVT_S_RESP_PENDING) &&
+	    make_rc_ack(dev, qp, ohdr, ps))
+		return 1;
 
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_SEND_OK)) {
-		if (!(ib_hfi1_state_ops[qp->state] & HFI1_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_iowait.sdma_busy)) {
-			qp->s_flags |= HFI1_S_WAIT_DMA;
+		if (iowait_sdma_pending(&priv->s_iowait)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
 		clear_ahg(qp);
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		hfi1_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
 			IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR);
 		/* will get called again */
-		goto done;
+		goto done_free_tx;
 	}
 
-	if (qp->s_flags & (HFI1_S_WAIT_RNR | HFI1_S_WAIT_ACK))
+	if (qp->s_flags & (RVT_S_WAIT_RNR | RVT_S_WAIT_ACK))
 		goto bail;
 
 	if (cmp_psn(qp->s_psn, qp->s_sending_hpsn) <= 0) {
 		if (cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0) {
-			qp->s_flags |= HFI1_S_WAIT_PSN;
+			qp->s_flags |= RVT_S_WAIT_PSN;
 			goto bail;
 		}
 		qp->s_sending_psn = qp->s_psn;
@@ -322,10 +449,10 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 	}
 
 	/* Send a request. */
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	switch (qp->s_state) {
 	default:
-		if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_NEXT_SEND_OK))
+		if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_NEXT_SEND_OK))
 			goto bail;
 		/*
 		 * Resend an old request or start a new one.
@@ -347,11 +474,11 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			 */
 			if ((wqe->wr.send_flags & IB_SEND_FENCE) &&
 			    qp->s_num_rd_atomic) {
-				qp->s_flags |= HFI1_S_WAIT_FENCE;
+				qp->s_flags |= RVT_S_WAIT_FENCE;
 				goto bail;
 			}
-			wqe->psn = qp->s_next_psn;
 			newreq = 1;
+			qp->s_psn = wqe->psn;
 		}
 		/*
 		 * Note that we have to be careful not to modify the
@@ -365,21 +492,19 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 		case IB_WR_SEND:
 		case IB_WR_SEND_WITH_IMM:
 			/* If no credit, return. */
-			if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT) &&
+			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
 			    cmp_msn(wqe->ssn, qp->s_lsn + 1) > 0) {
-				qp->s_flags |= HFI1_S_WAIT_SSN_CREDIT;
+				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
-			wqe->lpsn = wqe->psn;
 			if (len > pmtu) {
-				wqe->lpsn += (len - 1) / pmtu;
 				qp->s_state = OP(SEND_FIRST);
 				len = pmtu;
 				break;
 			}
-			if (wqe->wr.opcode == IB_WR_SEND)
+			if (wqe->wr.opcode == IB_WR_SEND) {
 				qp->s_state = OP(SEND_ONLY);
-			else {
+			} else {
 				qp->s_state = OP(SEND_ONLY_WITH_IMMEDIATE);
 				/* Immediate data comes after the BTH */
 				ohdr->u.imm_data = wqe->wr.ex.imm_data;
@@ -393,14 +518,14 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			break;
 
 		case IB_WR_RDMA_WRITE:
-			if (newreq && !(qp->s_flags & HFI1_S_UNLIMITED_CREDIT))
+			if (newreq && !(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 				qp->s_lsn++;
 			/* FALLTHROUGH */
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			/* If no credit, return. */
-			if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT) &&
+			if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT) &&
 			    cmp_msn(wqe->ssn, qp->s_lsn + 1) > 0) {
-				qp->s_flags |= HFI1_S_WAIT_SSN_CREDIT;
+				qp->s_flags |= RVT_S_WAIT_SSN_CREDIT;
 				goto bail;
 			}
 			ohdr->u.rc.reth.vaddr =
@@ -409,16 +534,14 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 				cpu_to_be32(wqe->rdma_wr.rkey);
 			ohdr->u.rc.reth.length = cpu_to_be32(len);
 			hwords += sizeof(struct ib_reth) / sizeof(u32);
-			wqe->lpsn = wqe->psn;
 			if (len > pmtu) {
-				wqe->lpsn += (len - 1) / pmtu;
 				qp->s_state = OP(RDMA_WRITE_FIRST);
 				len = pmtu;
 				break;
 			}
-			if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+			if (wqe->wr.opcode == IB_WR_RDMA_WRITE) {
 				qp->s_state = OP(RDMA_WRITE_ONLY);
-			else {
+			} else {
 				qp->s_state =
 					OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE);
 				/* Immediate data comes after RETH */
@@ -440,19 +563,12 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			if (newreq) {
 				if (qp->s_num_rd_atomic >=
 				    qp->s_max_rd_atomic) {
-					qp->s_flags |= HFI1_S_WAIT_RDMAR;
+					qp->s_flags |= RVT_S_WAIT_RDMAR;
 					goto bail;
 				}
 				qp->s_num_rd_atomic++;
-				if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT))
+				if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 					qp->s_lsn++;
-				/*
-				 * Adjust s_next_psn to count the
-				 * expected number of responses.
-				 */
-				if (len > pmtu)
-					qp->s_next_psn += (len - 1) / pmtu;
-				wqe->lpsn = qp->s_next_psn++;
 			}
 			ohdr->u.rc.reth.vaddr =
 				cpu_to_be64(wqe->rdma_wr.remote_addr);
@@ -477,13 +593,12 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			if (newreq) {
 				if (qp->s_num_rd_atomic >=
 				    qp->s_max_rd_atomic) {
-					qp->s_flags |= HFI1_S_WAIT_RDMAR;
+					qp->s_flags |= RVT_S_WAIT_RDMAR;
 					goto bail;
 				}
 				qp->s_num_rd_atomic++;
-				if (!(qp->s_flags & HFI1_S_UNLIMITED_CREDIT))
+				if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT))
 					qp->s_lsn++;
-				wqe->lpsn = wqe->psn;
 			}
 			if (wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP) {
 				qp->s_state = OP(COMPARE_SWAP);
@@ -526,11 +641,8 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 		}
 		if (wqe->wr.opcode == IB_WR_RDMA_READ)
 			qp->s_psn = wqe->lpsn + 1;
-		else {
+		else
 			qp->s_psn++;
-			if (cmp_psn(qp->s_psn, qp->s_next_psn) > 0)
-				qp->s_next_psn = qp->s_psn;
-		}
 		break;
 
 	case OP(RDMA_READ_RESPONSE_FIRST):
@@ -550,8 +662,6 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 		/* FALLTHROUGH */
 	case OP(SEND_MIDDLE):
 		bth2 = mask_psn(qp->s_psn++);
-		if (cmp_psn(qp->s_psn, qp->s_next_psn) > 0)
-			qp->s_next_psn = qp->s_psn;
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
@@ -559,9 +669,9 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			middle = HFI1_CAP_IS_KSET(SDMA_AHG);
 			break;
 		}
-		if (wqe->wr.opcode == IB_WR_SEND)
+		if (wqe->wr.opcode == IB_WR_SEND) {
 			qp->s_state = OP(SEND_LAST);
-		else {
+		} else {
 			qp->s_state = OP(SEND_LAST_WITH_IMMEDIATE);
 			/* Immediate data comes after the BTH */
 			ohdr->u.imm_data = wqe->wr.ex.imm_data;
@@ -592,8 +702,6 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 		/* FALLTHROUGH */
 	case OP(RDMA_WRITE_MIDDLE):
 		bth2 = mask_psn(qp->s_psn++);
-		if (cmp_psn(qp->s_psn, qp->s_next_psn) > 0)
-			qp->s_next_psn = qp->s_psn;
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
@@ -601,9 +709,9 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 			middle = HFI1_CAP_IS_KSET(SDMA_AHG);
 			break;
 		}
-		if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE) {
 			qp->s_state = OP(RDMA_WRITE_LAST);
-		else {
+		} else {
 			qp->s_state = OP(RDMA_WRITE_LAST_WITH_IMMEDIATE);
 			/* Immediate data comes after the BTH */
 			ohdr->u.imm_data = wqe->wr.ex.imm_data;
@@ -648,13 +756,14 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 	delta = delta_psn(bth2, wqe->psn);
 	if (delta && delta % HFI1_PSN_CREDIT == 0)
 		bth2 |= IB_BTH_REQ_ACK;
-	if (qp->s_flags & HFI1_S_SEND_ONE) {
-		qp->s_flags &= ~HFI1_S_SEND_ONE;
-		qp->s_flags |= HFI1_S_WAIT_ACK;
+	if (qp->s_flags & RVT_S_SEND_ONE) {
+		qp->s_flags &= ~RVT_S_SEND_ONE;
+		qp->s_flags |= RVT_S_WAIT_ACK;
 		bth2 |= IB_BTH_REQ_ACK;
 	}
 	qp->s_len -= len;
 	qp->s_hdrwords = hwords;
+	ps->s_txreq->sde = priv->s_sde;
 	qp->s_cur_sge = ss;
 	qp->s_cur_size = len;
 	hfi1_make_ruc_header(
@@ -662,16 +771,25 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
 		ohdr,
 		bth0 | (qp->s_state << 24),
 		bth2,
-		middle);
-done:
-	ret = 1;
-	goto unlock;
+		middle,
+		ps);
+	/* pbc */
+	ps->s_txreq->hdr_dwords = qp->s_hdrwords + 2;
+	return 1;
+
+done_free_tx:
+	hfi1_put_txreq(ps->s_txreq);
+	ps->s_txreq = NULL;
+	return 1;
 
 bail:
-	qp->s_flags &= ~HFI1_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-	return ret;
+	hfi1_put_txreq(ps->s_txreq);
+
+bail_no_tx:
+	ps->s_txreq = NULL;
+	qp->s_flags &= ~RVT_S_BUSY;
+	qp->s_hdrwords = 0;
+	return 0;
 }
 
 /**
@@ -682,7 +800,7 @@ unlock:
  * Note that RDMA reads and atomics are handled in the
  * send side QP state and tasklet.
  */
-void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct hfi1_qp *qp,
+void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct rvt_qp *qp,
 		      int is_fecn)
 {
 	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
@@ -700,7 +818,7 @@ void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct hfi1_qp *qp,
 	unsigned long flags;
 
 	/* Don't send ACK or NAK if a RDMA read or atomic is pending. */
-	if (qp->s_flags & HFI1_S_RESP_PENDING)
+	if (qp->s_flags & RVT_S_RESP_PENDING)
 		goto queue_ack;
 
 	/* Ensure s_rdma_ack_cnt changes are committed */
@@ -763,7 +881,7 @@ void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct hfi1_qp *qp,
 		goto queue_ack;
 	}
 
-	trace_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &hdr);
+	trace_ack_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &hdr);
 
 	/* write the pbc and data */
 	ppd->dd->pio_inline_send(ppd->dd, pbuf, pbc, &hdr, hwords);
@@ -771,13 +889,13 @@ void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct hfi1_qp *qp,
 	return;
 
 queue_ack:
-	this_cpu_inc(*ibp->rc_qacks);
+	this_cpu_inc(*ibp->rvp.rc_qacks);
 	spin_lock_irqsave(&qp->s_lock, flags);
-	qp->s_flags |= HFI1_S_ACK_PENDING | HFI1_S_RESP_PENDING;
+	qp->s_flags |= RVT_S_ACK_PENDING | RVT_S_RESP_PENDING;
 	qp->s_nak_state = qp->r_nak_state;
 	qp->s_ack_psn = qp->r_ack_psn;
 	if (is_fecn)
-		qp->s_flags |= HFI1_S_ECN;
+		qp->s_flags |= RVT_S_ECN;
 
 	/* Schedule the send tasklet. */
 	hfi1_schedule_send(qp);
@@ -793,10 +911,10 @@ queue_ack:
  * for the given QP.
  * Called at interrupt level with the QP s_lock held.
  */
-static void reset_psn(struct hfi1_qp *qp, u32 psn)
+static void reset_psn(struct rvt_qp *qp, u32 psn)
 {
 	u32 n = qp->s_acked;
-	struct hfi1_swqe *wqe = get_swqe_ptr(qp, n);
+	struct rvt_swqe *wqe = rvt_get_swqe_ptr(qp, n);
 	u32 opcode;
 
 	qp->s_cur = n;
@@ -819,7 +937,7 @@ static void reset_psn(struct hfi1_qp *qp, u32 psn)
 			n = 0;
 		if (n == qp->s_tail)
 			break;
-		wqe = get_swqe_ptr(qp, n);
+		wqe = rvt_get_swqe_ptr(qp, n);
 		diff = cmp_psn(psn, wqe->psn);
 		if (diff < 0)
 			break;
@@ -865,23 +983,23 @@ static void reset_psn(struct hfi1_qp *qp, u32 psn)
 done:
 	qp->s_psn = psn;
 	/*
-	 * Set HFI1_S_WAIT_PSN as rc_complete() may start the timer
+	 * Set RVT_S_WAIT_PSN as rc_complete() may start the timer
 	 * asynchronously before the send tasklet can get scheduled.
 	 * Doing it in hfi1_make_rc_req() is too late.
 	 */
 	if ((cmp_psn(qp->s_psn, qp->s_sending_hpsn) <= 0) &&
 	    (cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0))
-		qp->s_flags |= HFI1_S_WAIT_PSN;
-	qp->s_flags &= ~HFI1_S_AHG_VALID;
+		qp->s_flags |= RVT_S_WAIT_PSN;
+	qp->s_flags &= ~RVT_S_AHG_VALID;
 }
 
 /*
  * Back up requester to resend the last un-ACKed request.
  * The QP r_lock and s_lock should be held and interrupts disabled.
  */
-static void restart_rc(struct hfi1_qp *qp, u32 psn, int wait)
+static void restart_rc(struct rvt_qp *qp, u32 psn, int wait)
 {
-	struct hfi1_swqe *wqe = get_swqe_ptr(qp, qp->s_acked);
+	struct rvt_swqe *wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	struct hfi1_ibport *ibp;
 
 	if (qp->s_retry == 0) {
@@ -890,42 +1008,44 @@ static void restart_rc(struct hfi1_qp *qp, u32 psn, int wait)
 			qp->s_retry = qp->s_retry_cnt;
 		} else if (qp->s_last == qp->s_acked) {
 			hfi1_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
-			hfi1_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+			rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			return;
-		} else /* need to handle delayed completion */
+		} else { /* need to handle delayed completion */
 			return;
-	} else
+		}
+	} else {
 		qp->s_retry--;
+	}
 
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	if (wqe->wr.opcode == IB_WR_RDMA_READ)
-		ibp->n_rc_resends++;
+		ibp->rvp.n_rc_resends++;
 	else
-		ibp->n_rc_resends += delta_psn(qp->s_psn, psn);
+		ibp->rvp.n_rc_resends += delta_psn(qp->s_psn, psn);
 
-	qp->s_flags &= ~(HFI1_S_WAIT_FENCE | HFI1_S_WAIT_RDMAR |
-			 HFI1_S_WAIT_SSN_CREDIT | HFI1_S_WAIT_PSN |
-			 HFI1_S_WAIT_ACK);
+	qp->s_flags &= ~(RVT_S_WAIT_FENCE | RVT_S_WAIT_RDMAR |
+			 RVT_S_WAIT_SSN_CREDIT | RVT_S_WAIT_PSN |
+			 RVT_S_WAIT_ACK);
 	if (wait)
-		qp->s_flags |= HFI1_S_SEND_ONE;
+		qp->s_flags |= RVT_S_SEND_ONE;
 	reset_psn(qp, psn);
 }
 
 /*
  * This is called from s_timer for missing responses.
  */
-static void rc_timeout(unsigned long arg)
+void hfi1_rc_timeout(unsigned long arg)
 {
-	struct hfi1_qp *qp = (struct hfi1_qp *)arg;
+	struct rvt_qp *qp = (struct rvt_qp *)arg;
 	struct hfi1_ibport *ibp;
 	unsigned long flags;
 
 	spin_lock_irqsave(&qp->r_lock, flags);
 	spin_lock(&qp->s_lock);
-	if (qp->s_flags & HFI1_S_TIMER) {
+	if (qp->s_flags & RVT_S_TIMER) {
 		ibp = to_iport(qp->ibqp.device, qp->port_num);
-		ibp->n_rc_timeouts++;
-		qp->s_flags &= ~HFI1_S_TIMER;
+		ibp->rvp.n_rc_timeouts++;
+		qp->s_flags &= ~RVT_S_TIMER;
 		del_timer(&qp->s_timer);
 		trace_hfi1_rc_timeout(qp, qp->s_last_psn + 1);
 		restart_rc(qp, qp->s_last_psn + 1, 1);
@@ -940,15 +1060,12 @@ static void rc_timeout(unsigned long arg)
  */
 void hfi1_rc_rnr_retry(unsigned long arg)
 {
-	struct hfi1_qp *qp = (struct hfi1_qp *)arg;
+	struct rvt_qp *qp = (struct rvt_qp *)arg;
 	unsigned long flags;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (qp->s_flags & HFI1_S_WAIT_RNR) {
-		qp->s_flags &= ~HFI1_S_WAIT_RNR;
-		del_timer(&qp->s_timer);
-		hfi1_schedule_send(qp);
-	}
+	hfi1_stop_rnr_timer(qp);
+	hfi1_schedule_send(qp);
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
 
@@ -956,14 +1073,14 @@ void hfi1_rc_rnr_retry(unsigned long arg)
  * Set qp->s_sending_psn to the next PSN after the given one.
  * This would be psn+1 except when RDMA reads are present.
  */
-static void reset_sending_psn(struct hfi1_qp *qp, u32 psn)
+static void reset_sending_psn(struct rvt_qp *qp, u32 psn)
 {
-	struct hfi1_swqe *wqe;
+	struct rvt_swqe *wqe;
 	u32 n = qp->s_last;
 
 	/* Find the work request corresponding to the given PSN. */
 	for (;;) {
-		wqe = get_swqe_ptr(qp, n);
+		wqe = rvt_get_swqe_ptr(qp, n);
 		if (cmp_psn(psn, wqe->lpsn) <= 0) {
 			if (wqe->wr.opcode == IB_WR_RDMA_READ)
 				qp->s_sending_psn = wqe->lpsn + 1;
@@ -981,16 +1098,16 @@ static void reset_sending_psn(struct hfi1_qp *qp, u32 psn)
 /*
  * This should be called with the QP s_lock held and interrupts disabled.
  */
-void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
+void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_ib_header *hdr)
 {
 	struct hfi1_other_headers *ohdr;
-	struct hfi1_swqe *wqe;
+	struct rvt_swqe *wqe;
 	struct ib_wc wc;
 	unsigned i;
 	u32 opcode;
 	u32 psn;
 
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_OR_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		return;
 
 	/* Find out where the BTH is */
@@ -1016,22 +1133,30 @@ void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
 	 */
 	if ((psn & IB_BTH_REQ_ACK) && qp->s_acked != qp->s_tail &&
 	    !(qp->s_flags &
-		(HFI1_S_TIMER | HFI1_S_WAIT_RNR | HFI1_S_WAIT_PSN)) &&
-		(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK))
-		start_timer(qp);
+		(RVT_S_TIMER | RVT_S_WAIT_RNR | RVT_S_WAIT_PSN)) &&
+		(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK))
+		hfi1_add_retry_timer(qp);
 
 	while (qp->s_last != qp->s_acked) {
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		u32 s_last;
+
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		if (cmp_psn(wqe->lpsn, qp->s_sending_psn) >= 0 &&
 		    cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0)
 			break;
+		s_last = qp->s_last;
+		if (++s_last >= qp->s_size)
+			s_last = 0;
+		qp->s_last = s_last;
+		/* see post_send() */
+		barrier();
 		for (i = 0; i < wqe->wr.num_sge; i++) {
-			struct hfi1_sge *sge = &wqe->sg_list[i];
+			struct rvt_sge *sge = &wqe->sg_list[i];
 
-			hfi1_put_mr(sge->mr);
+			rvt_put_mr(sge->mr);
 		}
 		/* Post a send completion queue entry if requested. */
-		if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
+		if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
 			memset(&wc, 0, sizeof(wc));
 			wc.wr_id = wqe->wr.wr_id;
@@ -1039,26 +1164,24 @@ void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
 			wc.opcode = ib_hfi1_wc_opcode[wqe->wr.opcode];
 			wc.byte_len = wqe->length;
 			wc.qp = &qp->ibqp;
-			hfi1_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc, 0);
 		}
-		if (++qp->s_last >= qp->s_size)
-			qp->s_last = 0;
 	}
 	/*
 	 * If we were waiting for sends to complete before re-sending,
 	 * and they are now complete, restart sending.
 	 */
 	trace_hfi1_rc_sendcomplete(qp, psn);
-	if (qp->s_flags & HFI1_S_WAIT_PSN &&
+	if (qp->s_flags & RVT_S_WAIT_PSN &&
 	    cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
-		qp->s_flags &= ~HFI1_S_WAIT_PSN;
+		qp->s_flags &= ~RVT_S_WAIT_PSN;
 		qp->s_sending_psn = qp->s_psn;
 		qp->s_sending_hpsn = qp->s_psn - 1;
 		hfi1_schedule_send(qp);
 	}
 }
 
-static inline void update_last_psn(struct hfi1_qp *qp, u32 psn)
+static inline void update_last_psn(struct rvt_qp *qp, u32 psn)
 {
 	qp->s_last_psn = psn;
 }
@@ -1068,9 +1191,9 @@ static inline void update_last_psn(struct hfi1_qp *qp, u32 psn)
  * This is similar to hfi1_send_complete but has to check to be sure
  * that the SGEs are not being referenced if the SWQE is being resent.
  */
-static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
-					  struct hfi1_swqe *wqe,
-					  struct hfi1_ibport *ibp)
+static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp,
+					 struct rvt_swqe *wqe,
+					 struct hfi1_ibport *ibp)
 {
 	struct ib_wc wc;
 	unsigned i;
@@ -1082,13 +1205,21 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
 	 */
 	if (cmp_psn(wqe->lpsn, qp->s_sending_psn) < 0 ||
 	    cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
+		u32 s_last;
+
 		for (i = 0; i < wqe->wr.num_sge; i++) {
-			struct hfi1_sge *sge = &wqe->sg_list[i];
+			struct rvt_sge *sge = &wqe->sg_list[i];
 
-			hfi1_put_mr(sge->mr);
+			rvt_put_mr(sge->mr);
 		}
+		s_last = qp->s_last;
+		if (++s_last >= qp->s_size)
+			s_last = 0;
+		qp->s_last = s_last;
+		/* see post_send() */
+		barrier();
 		/* Post a send completion queue entry if requested. */
-		if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
+		if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
 			memset(&wc, 0, sizeof(wc));
 			wc.wr_id = wqe->wr.wr_id;
@@ -1096,14 +1227,12 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
 			wc.opcode = ib_hfi1_wc_opcode[wqe->wr.opcode];
 			wc.byte_len = wqe->length;
 			wc.qp = &qp->ibqp;
-			hfi1_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+			rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc, 0);
 		}
-		if (++qp->s_last >= qp->s_size)
-			qp->s_last = 0;
 	} else {
 		struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 
-		this_cpu_inc(*ibp->rc_delayed_comp);
+		this_cpu_inc(*ibp->rvp.rc_delayed_comp);
 		/*
 		 * If send progress not running attempt to progress
 		 * SDMA queue.
@@ -1131,7 +1260,7 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
 		if (++qp->s_cur >= qp->s_size)
 			qp->s_cur = 0;
 		qp->s_acked = qp->s_cur;
-		wqe = get_swqe_ptr(qp, qp->s_cur);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 		if (qp->s_acked != qp->s_tail) {
 			qp->s_state = OP(SEND_LAST);
 			qp->s_psn = wqe->psn;
@@ -1141,7 +1270,7 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
 			qp->s_acked = 0;
 		if (qp->state == IB_QPS_SQD && qp->s_acked == qp->s_cur)
 			qp->s_draining = 0;
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	}
 	return wqe;
 }
@@ -1157,21 +1286,16 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
  * May be called at interrupt level, with the QP s_lock held.
  * Returns 1 if OK, 0 if current operation should be aborted (NAK).
  */
-static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
+static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
 		     u64 val, struct hfi1_ctxtdata *rcd)
 {
 	struct hfi1_ibport *ibp;
 	enum ib_wc_status status;
-	struct hfi1_swqe *wqe;
+	struct rvt_swqe *wqe;
 	int ret = 0;
 	u32 ack_psn;
 	int diff;
-
-	/* Remove QP from retry timer */
-	if (qp->s_flags & (HFI1_S_TIMER | HFI1_S_WAIT_RNR)) {
-		qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_WAIT_RNR);
-		del_timer(&qp->s_timer);
-	}
+	unsigned long to;
 
 	/*
 	 * Note that NAKs implicitly ACK outstanding SEND and RDMA write
@@ -1182,7 +1306,7 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 	ack_psn = psn;
 	if (aeth >> 29)
 		ack_psn--;
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 
 	/*
@@ -1200,7 +1324,7 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 		    opcode == OP(RDMA_READ_RESPONSE_ONLY) &&
 		    diff == 0) {
 			ret = 1;
-			goto bail;
+			goto bail_stop;
 		}
 		/*
 		 * If this request is a RDMA read or atomic, and the ACK is
@@ -1217,11 +1341,11 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 		      wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) &&
 		     (opcode != OP(ATOMIC_ACKNOWLEDGE) || diff != 0))) {
 			/* Retry this request. */
-			if (!(qp->r_flags & HFI1_R_RDMAR_SEQ)) {
-				qp->r_flags |= HFI1_R_RDMAR_SEQ;
+			if (!(qp->r_flags & RVT_R_RDMAR_SEQ)) {
+				qp->r_flags |= RVT_R_RDMAR_SEQ;
 				restart_rc(qp, qp->s_last_psn + 1, 0);
 				if (list_empty(&qp->rspwait)) {
-					qp->r_flags |= HFI1_R_RSP_SEND;
+					qp->r_flags |= RVT_R_RSP_SEND;
 					atomic_inc(&qp->refcount);
 					list_add_tail(&qp->rspwait,
 						      &rcd->qp_wait_list);
@@ -1231,7 +1355,7 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 			 * No need to process the ACK/NAK since we are
 			 * restarting an earlier request.
 			 */
-			goto bail;
+			goto bail_stop;
 		}
 		if (wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
 		    wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) {
@@ -1244,14 +1368,14 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 		     wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD)) {
 			qp->s_num_rd_atomic--;
 			/* Restart sending task if fence is complete */
-			if ((qp->s_flags & HFI1_S_WAIT_FENCE) &&
+			if ((qp->s_flags & RVT_S_WAIT_FENCE) &&
 			    !qp->s_num_rd_atomic) {
-				qp->s_flags &= ~(HFI1_S_WAIT_FENCE |
-						 HFI1_S_WAIT_ACK);
+				qp->s_flags &= ~(RVT_S_WAIT_FENCE |
+						 RVT_S_WAIT_ACK);
 				hfi1_schedule_send(qp);
-			} else if (qp->s_flags & HFI1_S_WAIT_RDMAR) {
-				qp->s_flags &= ~(HFI1_S_WAIT_RDMAR |
-						 HFI1_S_WAIT_ACK);
+			} else if (qp->s_flags & RVT_S_WAIT_RDMAR) {
+				qp->s_flags &= ~(RVT_S_WAIT_RDMAR |
+						 RVT_S_WAIT_ACK);
 				hfi1_schedule_send(qp);
 			}
 		}
@@ -1262,40 +1386,43 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 
 	switch (aeth >> 29) {
 	case 0:         /* ACK */
-		this_cpu_inc(*ibp->rc_acks);
+		this_cpu_inc(*ibp->rvp.rc_acks);
 		if (qp->s_acked != qp->s_tail) {
 			/*
 			 * We are expecting more ACKs so
-			 * reset the re-transmit timer.
+			 * mod the retry timer.
 			 */
-			start_timer(qp);
+			hfi1_mod_retry_timer(qp);
 			/*
 			 * We can stop re-sending the earlier packets and
 			 * continue with the next packet the receiver wants.
 			 */
 			if (cmp_psn(qp->s_psn, psn) <= 0)
 				reset_psn(qp, psn + 1);
-		} else if (cmp_psn(qp->s_psn, psn) <= 0) {
-			qp->s_state = OP(SEND_LAST);
-			qp->s_psn = psn + 1;
+		} else {
+			/* No more acks - kill all timers */
+			hfi1_stop_rc_timers(qp);
+			if (cmp_psn(qp->s_psn, psn) <= 0) {
+				qp->s_state = OP(SEND_LAST);
+				qp->s_psn = psn + 1;
+			}
 		}
-		if (qp->s_flags & HFI1_S_WAIT_ACK) {
-			qp->s_flags &= ~HFI1_S_WAIT_ACK;
+		if (qp->s_flags & RVT_S_WAIT_ACK) {
+			qp->s_flags &= ~RVT_S_WAIT_ACK;
 			hfi1_schedule_send(qp);
 		}
 		hfi1_get_credit(qp, aeth);
 		qp->s_rnr_retry = qp->s_rnr_retry_cnt;
 		qp->s_retry = qp->s_retry_cnt;
 		update_last_psn(qp, psn);
-		ret = 1;
-		goto bail;
+		return 1;
 
 	case 1:         /* RNR NAK */
-		ibp->n_rnr_naks++;
+		ibp->rvp.n_rnr_naks++;
 		if (qp->s_acked == qp->s_tail)
-			goto bail;
-		if (qp->s_flags & HFI1_S_WAIT_RNR)
-			goto bail;
+			goto bail_stop;
+		if (qp->s_flags & RVT_S_WAIT_RNR)
+			goto bail_stop;
 		if (qp->s_rnr_retry == 0) {
 			status = IB_WC_RNR_RETRY_EXC_ERR;
 			goto class_b;
@@ -1306,28 +1433,27 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 		/* The last valid PSN is the previous PSN. */
 		update_last_psn(qp, psn - 1);
 
-		ibp->n_rc_resends += delta_psn(qp->s_psn, psn);
+		ibp->rvp.n_rc_resends += delta_psn(qp->s_psn, psn);
 
 		reset_psn(qp, psn);
 
-		qp->s_flags &= ~(HFI1_S_WAIT_SSN_CREDIT | HFI1_S_WAIT_ACK);
-		qp->s_flags |= HFI1_S_WAIT_RNR;
-		qp->s_timer.function = hfi1_rc_rnr_retry;
-		qp->s_timer.expires = jiffies + usecs_to_jiffies(
+		qp->s_flags &= ~(RVT_S_WAIT_SSN_CREDIT | RVT_S_WAIT_ACK);
+		hfi1_stop_rc_timers(qp);
+		to =
 			ib_hfi1_rnr_table[(aeth >> HFI1_AETH_CREDIT_SHIFT) &
-					   HFI1_AETH_CREDIT_MASK]);
-		add_timer(&qp->s_timer);
-		goto bail;
+					   HFI1_AETH_CREDIT_MASK];
+		hfi1_add_rnr_timer(qp, to);
+		return 0;
 
 	case 3:         /* NAK */
 		if (qp->s_acked == qp->s_tail)
-			goto bail;
+			goto bail_stop;
 		/* The last valid PSN is the previous PSN. */
 		update_last_psn(qp, psn - 1);
 		switch ((aeth >> HFI1_AETH_CREDIT_SHIFT) &
 			HFI1_AETH_CREDIT_MASK) {
 		case 0: /* PSN sequence error */
-			ibp->n_seq_naks++;
+			ibp->rvp.n_seq_naks++;
 			/*
 			 * Back up to the responder's expected PSN.
 			 * Note that we might get a NAK in the middle of an
@@ -1340,21 +1466,21 @@ static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
 
 		case 1: /* Invalid Request */
 			status = IB_WC_REM_INV_REQ_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 			goto class_b;
 
 		case 2: /* Remote Access Error */
 			status = IB_WC_REM_ACCESS_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 			goto class_b;
 
 		case 3: /* Remote Operation Error */
 			status = IB_WC_REM_OP_ERR;
-			ibp->n_other_naks++;
+			ibp->rvp.n_other_naks++;
 class_b:
 			if (qp->s_last == qp->s_acked) {
 				hfi1_send_complete(qp, wqe, status);
-				hfi1_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+				rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			}
 			break;
 
@@ -1364,15 +1490,16 @@ class_b:
 		}
 		qp->s_retry = qp->s_retry_cnt;
 		qp->s_rnr_retry = qp->s_rnr_retry_cnt;
-		goto bail;
+		goto bail_stop;
 
 	default:                /* 2: reserved */
 reserved:
 		/* Ignore reserved NAK codes. */
-		goto bail;
+		goto bail_stop;
 	}
-
-bail:
+	return ret;
+bail_stop:
+	hfi1_stop_rc_timers(qp);
 	return ret;
 }
 
@@ -1380,18 +1507,15 @@ bail:
  * We have seen an out of sequence RDMA read middle or last packet.
  * This ACKs SENDs and RDMA writes up to the first RDMA read or atomic SWQE.
  */
-static void rdma_seq_err(struct hfi1_qp *qp, struct hfi1_ibport *ibp, u32 psn,
+static void rdma_seq_err(struct rvt_qp *qp, struct hfi1_ibport *ibp, u32 psn,
 			 struct hfi1_ctxtdata *rcd)
 {
-	struct hfi1_swqe *wqe;
+	struct rvt_swqe *wqe;
 
 	/* Remove QP from retry timer */
-	if (qp->s_flags & (HFI1_S_TIMER | HFI1_S_WAIT_RNR)) {
-		qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_WAIT_RNR);
-		del_timer(&qp->s_timer);
-	}
+	hfi1_stop_rc_timers(qp);
 
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 
 	while (cmp_psn(psn, wqe->lpsn) > 0) {
 		if (wqe->wr.opcode == IB_WR_RDMA_READ ||
@@ -1401,11 +1525,11 @@ static void rdma_seq_err(struct hfi1_qp *qp, struct hfi1_ibport *ibp, u32 psn,
 		wqe = do_rc_completion(qp, wqe, ibp);
 	}
 
-	ibp->n_rdma_seq++;
-	qp->r_flags |= HFI1_R_RDMAR_SEQ;
+	ibp->rvp.n_rdma_seq++;
+	qp->r_flags |= RVT_R_RDMAR_SEQ;
 	restart_rc(qp, qp->s_last_psn + 1, 0);
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= HFI1_R_RSP_SEND;
+		qp->r_flags |= RVT_R_RSP_SEND;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
@@ -1429,11 +1553,11 @@ static void rdma_seq_err(struct hfi1_qp *qp, struct hfi1_ibport *ibp, u32 psn,
  */
 static void rc_rcv_resp(struct hfi1_ibport *ibp,
 			struct hfi1_other_headers *ohdr,
-			void *data, u32 tlen, struct hfi1_qp *qp,
+			void *data, u32 tlen, struct rvt_qp *qp,
 			u32 opcode, u32 psn, u32 hdrsize, u32 pmtu,
 			struct hfi1_ctxtdata *rcd)
 {
-	struct hfi1_swqe *wqe;
+	struct rvt_swqe *wqe;
 	enum ib_wc_status status;
 	unsigned long flags;
 	int diff;
@@ -1446,7 +1570,8 @@ static void rc_rcv_resp(struct hfi1_ibport *ibp,
 	trace_hfi1_rc_ack(qp, psn);
 
 	/* Ignore invalid responses. */
-	if (cmp_psn(psn, qp->s_next_psn) >= 0)
+	smp_read_barrier_depends(); /* see post_one_send */
+	if (cmp_psn(psn, ACCESS_ONCE(qp->s_next_psn)) >= 0)
 		goto ack_done;
 
 	/* Ignore duplicate responses. */
@@ -1465,15 +1590,15 @@ static void rc_rcv_resp(struct hfi1_ibport *ibp,
 	 * Skip everything other than the PSN we expect, if we are waiting
 	 * for a reply to a restarted RDMA read or atomic op.
 	 */
-	if (qp->r_flags & HFI1_R_RDMAR_SEQ) {
+	if (qp->r_flags & RVT_R_RDMAR_SEQ) {
 		if (cmp_psn(psn, qp->s_last_psn + 1) != 0)
 			goto ack_done;
-		qp->r_flags &= ~HFI1_R_RDMAR_SEQ;
+		qp->r_flags &= ~RVT_R_RDMAR_SEQ;
 	}
 
 	if (unlikely(qp->s_acked == qp->s_tail))
 		goto ack_done;
-	wqe = get_swqe_ptr(qp, qp->s_acked);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 	status = IB_WC_SUCCESS;
 
 	switch (opcode) {
@@ -1484,14 +1609,15 @@ static void rc_rcv_resp(struct hfi1_ibport *ibp,
 		if (opcode == OP(ATOMIC_ACKNOWLEDGE)) {
 			__be32 *p = ohdr->u.at.atomic_ack_eth;
 
-			val = ((u64) be32_to_cpu(p[0]) << 32) |
+			val = ((u64)be32_to_cpu(p[0]) << 32) |
 				be32_to_cpu(p[1]);
-		} else
+		} else {
 			val = 0;
+		}
 		if (!do_rc_ack(qp, aeth, psn, opcode, val, rcd) ||
 		    opcode != OP(RDMA_READ_RESPONSE_FIRST))
 			goto ack_done;
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 		if (unlikely(wqe->wr.opcode != IB_WR_RDMA_READ))
 			goto ack_op_err;
 		/*
@@ -1519,10 +1645,10 @@ read_middle:
 		 * We got a response so update the timeout.
 		 * 4.096 usec. * (1 << qp->timeout)
 		 */
-		qp->s_flags |= HFI1_S_TIMER;
+		qp->s_flags |= RVT_S_TIMER;
 		mod_timer(&qp->s_timer, jiffies + qp->timeout_jiffies);
-		if (qp->s_flags & HFI1_S_WAIT_ACK) {
-			qp->s_flags &= ~HFI1_S_WAIT_ACK;
+		if (qp->s_flags & RVT_S_WAIT_ACK) {
+			qp->s_flags &= ~RVT_S_WAIT_ACK;
 			hfi1_schedule_send(qp);
 		}
 
@@ -1536,7 +1662,7 @@ read_middle:
 		qp->s_rdma_read_len -= pmtu;
 		update_last_psn(qp, psn);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
-		hfi1_copy_sge(&qp->s_rdma_read_sge, data, pmtu, 0);
+		hfi1_copy_sge(&qp->s_rdma_read_sge, data, pmtu, 0, 0);
 		goto bail;
 
 	case OP(RDMA_READ_RESPONSE_ONLY):
@@ -1556,7 +1682,7 @@ read_middle:
 		 * have to be careful to copy the data to the right
 		 * location.
 		 */
-		wqe = get_swqe_ptr(qp, qp->s_acked);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_acked);
 		qp->s_rdma_read_len = restart_sge(&qp->s_rdma_read_sge,
 						  wqe, psn, pmtu);
 		goto read_last;
@@ -1580,9 +1706,9 @@ read_last:
 		if (unlikely(tlen != qp->s_rdma_read_len))
 			goto ack_len_err;
 		aeth = be32_to_cpu(ohdr->u.aeth);
-		hfi1_copy_sge(&qp->s_rdma_read_sge, data, tlen, 0);
+		hfi1_copy_sge(&qp->s_rdma_read_sge, data, tlen, 0, 0);
 		WARN_ON(qp->s_rdma_read_sge.num_sge);
-		(void) do_rc_ack(qp, aeth, psn,
+		(void)do_rc_ack(qp, aeth, psn,
 				 OP(RDMA_READ_RESPONSE_LAST), 0, rcd);
 		goto ack_done;
 	}
@@ -1600,7 +1726,7 @@ ack_len_err:
 ack_err:
 	if (qp->s_last == qp->s_acked) {
 		hfi1_send_complete(qp, wqe, status);
-		hfi1_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+		rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 	}
 ack_done:
 	spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -1609,22 +1735,24 @@ bail:
 }
 
 static inline void rc_defered_ack(struct hfi1_ctxtdata *rcd,
-				  struct hfi1_qp *qp)
+				  struct rvt_qp *qp)
 {
 	if (list_empty(&qp->rspwait)) {
-		qp->r_flags |= HFI1_R_RSP_DEFERED_ACK;
+		qp->r_flags |= RVT_R_RSP_NAK;
 		atomic_inc(&qp->refcount);
 		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
 	}
 }
 
-static inline void rc_cancel_ack(struct hfi1_qp *qp)
+static inline void rc_cancel_ack(struct rvt_qp *qp)
 {
-	qp->r_adefered = 0;
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	priv->r_adefered = 0;
 	if (list_empty(&qp->rspwait))
 		return;
 	list_del_init(&qp->rspwait);
-	qp->r_flags &= ~HFI1_R_RSP_DEFERED_ACK;
+	qp->r_flags &= ~RVT_R_RSP_NAK;
 	if (atomic_dec_and_test(&qp->refcount))
 		wake_up(&qp->wait);
 }
@@ -1645,11 +1773,11 @@ static inline void rc_cancel_ack(struct hfi1_qp *qp)
  * schedule a response to be sent.
  */
 static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
-			struct hfi1_qp *qp, u32 opcode, u32 psn, int diff,
-			struct hfi1_ctxtdata *rcd)
+				 struct rvt_qp *qp, u32 opcode, u32 psn,
+				 int diff, struct hfi1_ctxtdata *rcd)
 {
 	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
-	struct hfi1_ack_entry *e;
+	struct rvt_ack_entry *e;
 	unsigned long flags;
 	u8 i, prev;
 	int old_req;
@@ -1662,7 +1790,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 		 * Don't queue the NAK if we already sent one.
 		 */
 		if (!qp->r_nak_state) {
-			ibp->n_rc_seqnak++;
+			ibp->rvp.n_rc_seqnak++;
 			qp->r_nak_state = IB_NAK_PSN_ERROR;
 			/* Use the expected PSN. */
 			qp->r_ack_psn = qp->r_psn;
@@ -1694,7 +1822,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 	 */
 	e = NULL;
 	old_req = 1;
-	ibp->n_rc_dupreq++;
+	ibp->rvp.n_rc_dupreq++;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 
@@ -1747,7 +1875,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 		if (unlikely(offset + len != e->rdma_sge.sge_length))
 			goto unlock_done;
 		if (e->rdma_sge.mr) {
-			hfi1_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		if (len != 0) {
@@ -1755,8 +1883,8 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 			u64 vaddr = be64_to_cpu(reth->vaddr);
 			int ok;
 
-			ok = hfi1_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
-					  IB_ACCESS_REMOTE_READ);
+			ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
+					 IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok))
 				goto unlock_done;
 		} else {
@@ -1778,7 +1906,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 		 * or the send tasklet is already backed up to send an
 		 * earlier entry, we can ignore this request.
 		 */
-		if (!e || e->opcode != (u8) opcode || old_req)
+		if (!e || e->opcode != (u8)opcode || old_req)
 			goto unlock_done;
 		qp->s_tail_ack_queue = prev;
 		break;
@@ -1810,7 +1938,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
 		break;
 	}
 	qp->s_ack_state = OP(ACKNOWLEDGE);
-	qp->s_flags |= HFI1_S_RESP_PENDING;
+	qp->s_flags |= RVT_S_RESP_PENDING;
 	qp->r_nak_state = 0;
 	hfi1_schedule_send(qp);
 
@@ -1823,13 +1951,13 @@ send_ack:
 	return 0;
 }
 
-void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err)
+void hfi1_rc_error(struct rvt_qp *qp, enum ib_wc_status err)
 {
 	unsigned long flags;
 	int lastwqe;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	lastwqe = hfi1_error_qp(qp, err);
+	lastwqe = rvt_error_qp(qp, err);
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 
 	if (lastwqe) {
@@ -1842,7 +1970,7 @@ void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err)
 	}
 }
 
-static inline void update_ack_queue(struct hfi1_qp *qp, unsigned n)
+static inline void update_ack_queue(struct rvt_qp *qp, unsigned n)
 {
 	unsigned next;
 
@@ -1864,14 +1992,14 @@ static void log_cca_event(struct hfi1_pportdata *ppd, u8 sl, u32 rlid,
 
 	spin_lock_irqsave(&ppd->cc_log_lock, flags);
 
-	ppd->threshold_cong_event_map[sl/8] |= 1 << (sl % 8);
+	ppd->threshold_cong_event_map[sl / 8] |= 1 << (sl % 8);
 	ppd->threshold_event_counter++;
 
 	cc_event = &ppd->cc_events[ppd->cc_log_idx++];
 	if (ppd->cc_log_idx == OPA_CONG_LOG_ELEMS)
 		ppd->cc_log_idx = 0;
-	cc_event->lqpn = lqpn & HFI1_QPN_MASK;
-	cc_event->rqpn = rqpn & HFI1_QPN_MASK;
+	cc_event->lqpn = lqpn & RVT_QPN_MASK;
+	cc_event->rqpn = rqpn & RVT_QPN_MASK;
 	cc_event->sl = sl;
 	cc_event->svc_type = svc_type;
 	cc_event->rlid = rlid;
@@ -1897,7 +2025,7 @@ void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
 
 	cc_state = get_cc_state(ppd);
 
-	if (cc_state == NULL)
+	if (!cc_state)
 		return;
 
 	/*
@@ -1957,7 +2085,7 @@ void hfi1_rc_rcv(struct hfi1_packet *packet)
 	u32 rcv_flags = packet->rcv_flags;
 	void *data = packet->ebuf;
 	u32 tlen = packet->tlen;
-	struct hfi1_qp *qp = packet->qp;
+	struct rvt_qp *qp = packet->qp;
 	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	struct hfi1_other_headers *ohdr = packet->ohdr;
@@ -1972,6 +2100,7 @@ void hfi1_rc_rcv(struct hfi1_packet *packet)
 	unsigned long flags;
 	u32 bth1;
 	int ret, is_fecn = 0;
+	int copy_last = 0;
 
 	bth0 = be32_to_cpu(ohdr->bth[0]);
 	if (hfi1_ruc_check_hdr(ibp, hdr, rcv_flags & HFI1_HAS_GRH, qp, bth0))
@@ -2054,13 +2183,13 @@ void hfi1_rc_rcv(struct hfi1_packet *packet)
 		break;
 	}
 
-	if (qp->state == IB_QPS_RTR && !(qp->r_flags & HFI1_R_COMM_EST))
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & RVT_R_COMM_EST))
 		qp_comm_est(qp);
 
 	/* OK, process the packet. */
 	switch (opcode) {
 	case OP(SEND_FIRST):
-		ret = hfi1_get_rwqe(qp, 0);
+		ret = hfi1_rvt_get_rwqe(qp, 0);
 		if (ret < 0)
 			goto nack_op_err;
 		if (!ret)
@@ -2076,12 +2205,12 @@ send_middle:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto nack_inv;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, 1);
+		hfi1_copy_sge(&qp->r_sge, data, pmtu, 1, 0);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
 		/* consume RWQE */
-		ret = hfi1_get_rwqe(qp, 1);
+		ret = hfi1_rvt_get_rwqe(qp, 1);
 		if (ret < 0)
 			goto nack_op_err;
 		if (!ret)
@@ -2090,7 +2219,7 @@ send_middle:
 
 	case OP(SEND_ONLY):
 	case OP(SEND_ONLY_WITH_IMMEDIATE):
-		ret = hfi1_get_rwqe(qp, 0);
+		ret = hfi1_rvt_get_rwqe(qp, 0);
 		if (ret < 0)
 			goto nack_op_err;
 		if (!ret)
@@ -2104,8 +2233,10 @@ send_last_imm:
 		wc.ex.imm_data = ohdr->u.imm_data;
 		wc.wc_flags = IB_WC_WITH_IMM;
 		goto send_last;
-	case OP(SEND_LAST):
 	case OP(RDMA_WRITE_LAST):
+		copy_last = ibpd_to_rvtpd(qp->ibqp.pd)->user;
+		/* fall through */
+	case OP(SEND_LAST):
 no_immediate_data:
 		wc.wc_flags = 0;
 		wc.ex.imm_data = 0;
@@ -2121,10 +2252,10 @@ send_last:
 		wc.byte_len = tlen + qp->r_rcv_len;
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto nack_inv;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, 1);
-		hfi1_put_ss(&qp->r_sge);
+		hfi1_copy_sge(&qp->r_sge, data, tlen, 1, copy_last);
+		rvt_put_ss(&qp->r_sge);
 		qp->r_msn++;
-		if (!test_and_clear_bit(HFI1_R_WRID_VALID, &qp->r_aflags))
+		if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 			break;
 		wc.wr_id = qp->r_wr_id;
 		wc.status = IB_WC_SUCCESS;
@@ -2154,12 +2285,14 @@ send_last:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-			      (bth0 & IB_BTH_SOLICITED) != 0);
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+			     (bth0 & IB_BTH_SOLICITED) != 0);
 		break;
 
-	case OP(RDMA_WRITE_FIRST):
 	case OP(RDMA_WRITE_ONLY):
+		copy_last = 1;
+		/* fall through */
+	case OP(RDMA_WRITE_FIRST):
 	case OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE):
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
 			goto nack_inv;
@@ -2174,8 +2307,8 @@ send_last:
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = hfi1_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
-					  rkey, IB_ACCESS_REMOTE_WRITE);
+			ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
+					 rkey, IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok))
 				goto nack_acc;
 			qp->r_sge.num_sge = 1;
@@ -2190,7 +2323,7 @@ send_last:
 			goto send_middle;
 		else if (opcode == OP(RDMA_WRITE_ONLY))
 			goto no_immediate_data;
-		ret = hfi1_get_rwqe(qp, 1);
+		ret = hfi1_rvt_get_rwqe(qp, 1);
 		if (ret < 0)
 			goto nack_op_err;
 		if (!ret)
@@ -2200,7 +2333,7 @@ send_last:
 		goto send_last;
 
 	case OP(RDMA_READ_REQUEST): {
-		struct hfi1_ack_entry *e;
+		struct rvt_ack_entry *e;
 		u32 len;
 		u8 next;
 
@@ -2218,7 +2351,7 @@ send_last:
 		}
 		e = &qp->s_ack_queue[qp->r_head_ack_queue];
 		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
-			hfi1_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		reth = &ohdr->u.rc.reth;
@@ -2229,8 +2362,8 @@ send_last:
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = hfi1_rkey_ok(qp, &e->rdma_sge, len, vaddr,
-					  rkey, IB_ACCESS_REMOTE_READ);
+			ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr,
+					 rkey, IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok))
 				goto nack_acc_unlck;
 			/*
@@ -2261,7 +2394,7 @@ send_last:
 		qp->r_head_ack_queue = next;
 
 		/* Schedule the send tasklet. */
-		qp->s_flags |= HFI1_S_RESP_PENDING;
+		qp->s_flags |= RVT_S_RESP_PENDING;
 		hfi1_schedule_send(qp);
 
 		spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -2273,7 +2406,7 @@ send_last:
 	case OP(COMPARE_SWAP):
 	case OP(FETCH_ADD): {
 		struct ib_atomic_eth *ateth;
-		struct hfi1_ack_entry *e;
+		struct rvt_ack_entry *e;
 		u64 vaddr;
 		atomic64_t *maddr;
 		u64 sdata;
@@ -2293,29 +2426,29 @@ send_last:
 		}
 		e = &qp->s_ack_queue[qp->r_head_ack_queue];
 		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
-			hfi1_put_mr(e->rdma_sge.mr);
+			rvt_put_mr(e->rdma_sge.mr);
 			e->rdma_sge.mr = NULL;
 		}
 		ateth = &ohdr->u.atomic_eth;
-		vaddr = ((u64) be32_to_cpu(ateth->vaddr[0]) << 32) |
+		vaddr = ((u64)be32_to_cpu(ateth->vaddr[0]) << 32) |
 			be32_to_cpu(ateth->vaddr[1]);
 		if (unlikely(vaddr & (sizeof(u64) - 1)))
 			goto nack_inv_unlck;
 		rkey = be32_to_cpu(ateth->rkey);
 		/* Check rkey & NAK */
-		if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
-					   vaddr, rkey,
-					   IB_ACCESS_REMOTE_ATOMIC)))
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+					  vaddr, rkey,
+					  IB_ACCESS_REMOTE_ATOMIC)))
 			goto nack_acc_unlck;
 		/* Perform atomic OP and save result. */
-		maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
+		maddr = (atomic64_t *)qp->r_sge.sge.vaddr;
 		sdata = be64_to_cpu(ateth->swap_data);
 		e->atomic_data = (opcode == OP(FETCH_ADD)) ?
-			(u64) atomic64_add_return(sdata, maddr) - sdata :
-			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
+			(u64)atomic64_add_return(sdata, maddr) - sdata :
+			(u64)cmpxchg((u64 *)qp->r_sge.sge.vaddr,
 				      be64_to_cpu(ateth->compare_data),
 				      sdata);
-		hfi1_put_mr(qp->r_sge.sge.mr);
+		rvt_put_mr(qp->r_sge.sge.mr);
 		qp->r_sge.num_sge = 0;
 		e->opcode = opcode;
 		e->sent = 0;
@@ -2328,7 +2461,7 @@ send_last:
 		qp->r_head_ack_queue = next;
 
 		/* Schedule the send tasklet. */
-		qp->s_flags |= HFI1_S_RESP_PENDING;
+		qp->s_flags |= RVT_S_RESP_PENDING;
 		hfi1_schedule_send(qp);
 
 		spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -2347,11 +2480,13 @@ send_last:
 	qp->r_nak_state = 0;
 	/* Send an ACK if requested or required. */
 	if (psn & IB_BTH_REQ_ACK) {
+		struct hfi1_qp_priv *priv = qp->priv;
+
 		if (packet->numpkt == 0) {
 			rc_cancel_ack(qp);
 			goto send_ack;
 		}
-		if (qp->r_adefered >= HFI1_PSN_CREDIT) {
+		if (priv->r_adefered >= HFI1_PSN_CREDIT) {
 			rc_cancel_ack(qp);
 			goto send_ack;
 		}
@@ -2359,13 +2494,13 @@ send_last:
 			rc_cancel_ack(qp);
 			goto send_ack;
 		}
-		qp->r_adefered++;
+		priv->r_adefered++;
 		rc_defered_ack(rcd, qp);
 	}
 	return;
 
 rnr_nak:
-	qp->r_nak_state = IB_RNR_NAK | qp->r_min_rnr_timer;
+	qp->r_nak_state = qp->r_min_rnr_timer | IB_RNR_NAK;
 	qp->r_ack_psn = qp->r_psn;
 	/* Queue RNR NAK for later */
 	rc_defered_ack(rcd, qp);
@@ -2403,7 +2538,7 @@ void hfi1_rc_hdrerr(
 	struct hfi1_ctxtdata *rcd,
 	struct hfi1_ib_header *hdr,
 	u32 rcv_flags,
-	struct hfi1_qp *qp)
+	struct rvt_qp *qp)
 {
 	int has_grh = rcv_flags & HFI1_HAS_GRH;
 	struct hfi1_other_headers *ohdr;
@@ -2428,7 +2563,7 @@ void hfi1_rc_hdrerr(
 	if (opcode < IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST) {
 		diff = delta_psn(psn, qp->r_psn);
 		if (!qp->r_nak_state && diff >= 0) {
-			ibp->n_rc_seqnak++;
+			ibp->rvp.n_rc_seqnak++;
 			qp->r_nak_state = IB_NAK_PSN_ERROR;
 			/* Use the expected PSN. */
 			qp->r_ack_psn = qp->r_psn;
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 4a91975b68d7..08813cdbd475 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -53,7 +50,8 @@
 #include "hfi.h"
 #include "mad.h"
 #include "qp.h"
-#include "sdma.h"
+#include "verbs_txreq.h"
+#include "trace.h"
 
 /*
  * Convert the AETH RNR timeout code into the number of microseconds.
@@ -97,16 +95,16 @@ const u32 ib_hfi1_rnr_table[32] = {
  * Validate a RWQE and fill in the SGE state.
  * Return 1 if OK.
  */
-static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
+static int init_sge(struct rvt_qp *qp, struct rvt_rwqe *wqe)
 {
 	int i, j, ret;
 	struct ib_wc wc;
-	struct hfi1_lkey_table *rkt;
-	struct hfi1_pd *pd;
-	struct hfi1_sge_state *ss;
+	struct rvt_lkey_table *rkt;
+	struct rvt_pd *pd;
+	struct rvt_sge_state *ss;
 
-	rkt = &to_idev(qp->ibqp.device)->lk_table;
-	pd = to_ipd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
+	rkt = &to_idev(qp->ibqp.device)->rdi.lkey_table;
+	pd = ibpd_to_rvtpd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
 	ss = &qp->r_sge;
 	ss->sg_list = qp->r_sg_list;
 	qp->r_len = 0;
@@ -114,8 +112,8 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
 		if (wqe->sg_list[i].length == 0)
 			continue;
 		/* Check LKEY */
-		if (!hfi1_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
-				  &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
+		if (!rvt_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
+				 &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
 			goto bad_lkey;
 		qp->r_len += wqe->sg_list[i].length;
 		j++;
@@ -127,9 +125,9 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
 
 bad_lkey:
 	while (j) {
-		struct hfi1_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
+		struct rvt_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
 
-		hfi1_put_mr(sge->mr);
+		rvt_put_mr(sge->mr);
 	}
 	ss->num_sge = 0;
 	memset(&wc, 0, sizeof(wc));
@@ -138,14 +136,14 @@ bad_lkey:
 	wc.opcode = IB_WC_RECV;
 	wc.qp = &qp->ibqp;
 	/* Signal solicited completion event. */
-	hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
 	ret = 0;
 bail:
 	return ret;
 }
 
 /**
- * hfi1_get_rwqe - copy the next RWQE into the QP's RWQE
+ * hfi1_rvt_get_rwqe - copy the next RWQE into the QP's RWQE
  * @qp: the QP
  * @wr_id_only: update qp->r_wr_id only, not qp->r_sge
  *
@@ -154,19 +152,19 @@ bail:
  *
  * Can be called from interrupt level.
  */
-int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only)
+int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only)
 {
 	unsigned long flags;
-	struct hfi1_rq *rq;
-	struct hfi1_rwq *wq;
-	struct hfi1_srq *srq;
-	struct hfi1_rwqe *wqe;
+	struct rvt_rq *rq;
+	struct rvt_rwq *wq;
+	struct rvt_srq *srq;
+	struct rvt_rwqe *wqe;
 	void (*handler)(struct ib_event *, void *);
 	u32 tail;
 	int ret;
 
 	if (qp->ibqp.srq) {
-		srq = to_isrq(qp->ibqp.srq);
+		srq = ibsrq_to_rvtsrq(qp->ibqp.srq);
 		handler = srq->ibsrq.event_handler;
 		rq = &srq->rq;
 	} else {
@@ -176,7 +174,7 @@ int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only)
 	}
 
 	spin_lock_irqsave(&rq->lock, flags);
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK)) {
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
 		ret = 0;
 		goto unlock;
 	}
@@ -192,7 +190,7 @@ int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only)
 	}
 	/* Make sure entry is read after head index is read. */
 	smp_rmb();
-	wqe = get_rwqe_ptr(rq, tail);
+	wqe = rvt_get_rwqe_ptr(rq, tail);
 	/*
 	 * Even though we update the tail index in memory, the verbs
 	 * consumer is not supposed to post more entries until a
@@ -208,7 +206,7 @@ int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only)
 	qp->r_wr_id = wqe->wr_id;
 
 	ret = 1;
-	set_bit(HFI1_R_WRID_VALID, &qp->r_aflags);
+	set_bit(RVT_R_WRID_VALID, &qp->r_aflags);
 	if (handler) {
 		u32 n;
 
@@ -265,7 +263,7 @@ static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id)
  * The s_lock will be acquired around the hfi1_migrate_qp() call.
  */
 int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
-		       int has_grh, struct hfi1_qp *qp, u32 bth0)
+		       int has_grh, struct rvt_qp *qp, u32 bth0)
 {
 	__be64 guid;
 	unsigned long flags;
@@ -279,11 +277,13 @@ int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
 			if (!(qp->alt_ah_attr.ah_flags & IB_AH_GRH))
 				goto err;
 			guid = get_sguid(ibp, qp->alt_ah_attr.grh.sgid_index);
-			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->rvp.gid_prefix,
+				    guid))
 				goto err;
-			if (!gid_ok(&hdr->u.l.grh.sgid,
-			    qp->alt_ah_attr.grh.dgid.global.subnet_prefix,
-			    qp->alt_ah_attr.grh.dgid.global.interface_id))
+			if (!gid_ok(
+				&hdr->u.l.grh.sgid,
+				qp->alt_ah_attr.grh.dgid.global.subnet_prefix,
+				qp->alt_ah_attr.grh.dgid.global.interface_id))
 				goto err;
 		}
 		if (unlikely(rcv_pkey_check(ppd_from_ibp(ibp), (u16)bth0,
@@ -312,11 +312,13 @@ int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
 				goto err;
 			guid = get_sguid(ibp,
 					 qp->remote_ah_attr.grh.sgid_index);
-			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->rvp.gid_prefix,
+				    guid))
 				goto err;
-			if (!gid_ok(&hdr->u.l.grh.sgid,
-			    qp->remote_ah_attr.grh.dgid.global.subnet_prefix,
-			    qp->remote_ah_attr.grh.dgid.global.interface_id))
+			if (!gid_ok(
+			     &hdr->u.l.grh.sgid,
+			     qp->remote_ah_attr.grh.dgid.global.subnet_prefix,
+			     qp->remote_ah_attr.grh.dgid.global.interface_id))
 				goto err;
 		}
 		if (unlikely(rcv_pkey_check(ppd_from_ibp(ibp), (u16)bth0,
@@ -355,12 +357,12 @@ err:
  * receive interrupts since this is a connected protocol and all packets
  * will pass through here.
  */
-static void ruc_loopback(struct hfi1_qp *sqp)
+static void ruc_loopback(struct rvt_qp *sqp)
 {
 	struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
-	struct hfi1_qp *qp;
-	struct hfi1_swqe *wqe;
-	struct hfi1_sge *sge;
+	struct rvt_qp *qp;
+	struct rvt_swqe *wqe;
+	struct rvt_sge *sge;
 	unsigned long flags;
 	struct ib_wc wc;
 	u64 sdata;
@@ -368,6 +370,8 @@ static void ruc_loopback(struct hfi1_qp *sqp)
 	enum ib_wc_status send_status;
 	int release;
 	int ret;
+	int copy_last = 0;
+	u32 to;
 
 	rcu_read_lock();
 
@@ -375,25 +379,27 @@ static void ruc_loopback(struct hfi1_qp *sqp)
 	 * Note that we check the responder QP state after
 	 * checking the requester's state.
 	 */
-	qp = hfi1_lookup_qpn(ibp, sqp->remote_qpn);
+	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), &ibp->rvp,
+			    sqp->remote_qpn);
 
 	spin_lock_irqsave(&sqp->s_lock, flags);
 
 	/* Return if we are already busy processing a work request. */
-	if ((sqp->s_flags & (HFI1_S_BUSY | HFI1_S_ANY_WAIT)) ||
-	    !(ib_hfi1_state_ops[sqp->state] & HFI1_PROCESS_OR_FLUSH_SEND))
+	if ((sqp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT)) ||
+	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		goto unlock;
 
-	sqp->s_flags |= HFI1_S_BUSY;
+	sqp->s_flags |= RVT_S_BUSY;
 
 again:
-	if (sqp->s_last == sqp->s_head)
+	smp_read_barrier_depends(); /* see post_one_send() */
+	if (sqp->s_last == ACCESS_ONCE(sqp->s_head))
 		goto clr_busy;
-	wqe = get_swqe_ptr(sqp, sqp->s_last);
+	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
 
 	/* Return if it is not OK to start a new work request. */
-	if (!(ib_hfi1_state_ops[sqp->state] & HFI1_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_hfi1_state_ops[sqp->state] & HFI1_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
 			goto clr_busy;
 		/* We are in the error state, flush the work request. */
 		send_status = IB_WC_WR_FLUSH_ERR;
@@ -411,9 +417,9 @@ again:
 	}
 	spin_unlock_irqrestore(&sqp->s_lock, flags);
 
-	if (!qp || !(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) ||
+	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
 	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
-		ibp->n_pkt_drops++;
+		ibp->rvp.n_pkt_drops++;
 		/*
 		 * For RC, the requester would timeout and retry so
 		 * shortcut the timeouts and just signal too many retries.
@@ -439,7 +445,7 @@ again:
 		wc.ex.imm_data = wqe->wr.ex.imm_data;
 		/* FALLTHROUGH */
 	case IB_WR_SEND:
-		ret = hfi1_get_rwqe(qp, 0);
+		ret = hfi1_rvt_get_rwqe(qp, 0);
 		if (ret < 0)
 			goto op_err;
 		if (!ret)
@@ -451,21 +457,24 @@ again:
 			goto inv_err;
 		wc.wc_flags = IB_WC_WITH_IMM;
 		wc.ex.imm_data = wqe->wr.ex.imm_data;
-		ret = hfi1_get_rwqe(qp, 1);
+		ret = hfi1_rvt_get_rwqe(qp, 1);
 		if (ret < 0)
 			goto op_err;
 		if (!ret)
 			goto rnr_nak;
-		/* FALLTHROUGH */
+		/* skip copy_last set and qp_access_flags recheck */
+		goto do_write;
 	case IB_WR_RDMA_WRITE:
+		copy_last = ibpd_to_rvtpd(qp->ibqp.pd)->user;
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
 			goto inv_err;
+do_write:
 		if (wqe->length == 0)
 			break;
-		if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
-					   wqe->rdma_wr.remote_addr,
-					   wqe->rdma_wr.rkey,
-					   IB_ACCESS_REMOTE_WRITE)))
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
+					  wqe->rdma_wr.remote_addr,
+					  wqe->rdma_wr.rkey,
+					  IB_ACCESS_REMOTE_WRITE)))
 			goto acc_err;
 		qp->r_sge.sg_list = NULL;
 		qp->r_sge.num_sge = 1;
@@ -475,10 +484,10 @@ again:
 	case IB_WR_RDMA_READ:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
 			goto inv_err;
-		if (unlikely(!hfi1_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
-					   wqe->rdma_wr.remote_addr,
-					   wqe->rdma_wr.rkey,
-					   IB_ACCESS_REMOTE_READ)))
+		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
+					  wqe->rdma_wr.remote_addr,
+					  wqe->rdma_wr.rkey,
+					  IB_ACCESS_REMOTE_READ)))
 			goto acc_err;
 		release = 0;
 		sqp->s_sge.sg_list = NULL;
@@ -493,20 +502,20 @@ again:
 	case IB_WR_ATOMIC_FETCH_AND_ADD:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
 			goto inv_err;
-		if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
-					   wqe->atomic_wr.remote_addr,
-					   wqe->atomic_wr.rkey,
-					   IB_ACCESS_REMOTE_ATOMIC)))
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+					  wqe->atomic_wr.remote_addr,
+					  wqe->atomic_wr.rkey,
+					  IB_ACCESS_REMOTE_ATOMIC)))
 			goto acc_err;
 		/* Perform atomic OP and save result. */
-		maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
+		maddr = (atomic64_t *)qp->r_sge.sge.vaddr;
 		sdata = wqe->atomic_wr.compare_add;
-		*(u64 *) sqp->s_sge.sge.vaddr =
+		*(u64 *)sqp->s_sge.sge.vaddr =
 			(wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
-			(u64) atomic64_add_return(sdata, maddr) - sdata :
-			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
+			(u64)atomic64_add_return(sdata, maddr) - sdata :
+			(u64)cmpxchg((u64 *)qp->r_sge.sge.vaddr,
 				      sdata, wqe->atomic_wr.swap);
-		hfi1_put_mr(qp->r_sge.sge.mr);
+		rvt_put_mr(qp->r_sge.sge.mr);
 		qp->r_sge.num_sge = 0;
 		goto send_comp;
 
@@ -524,17 +533,17 @@ again:
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		WARN_ON_ONCE(len == 0);
-		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, release);
+		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, release, copy_last);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (!release)
-				hfi1_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--sqp->s_sge.num_sge)
 				*sge = *sqp->s_sge.sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= HFI1_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -547,9 +556,9 @@ again:
 		sqp->s_len -= len;
 	}
 	if (release)
-		hfi1_put_ss(&qp->r_sge);
+		rvt_put_ss(&qp->r_sge);
 
-	if (!test_and_clear_bit(HFI1_R_WRID_VALID, &qp->r_aflags))
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		goto send_comp;
 
 	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
@@ -565,12 +574,12 @@ again:
 	wc.sl = qp->remote_ah_attr.sl;
 	wc.port_num = 1;
 	/* Signal completion event if the solicited bit is set. */
-	hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-		      wqe->wr.send_flags & IB_SEND_SOLICITED);
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+		     wqe->wr.send_flags & IB_SEND_SOLICITED);
 
 send_comp:
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	ibp->n_loop_pkts++;
+	ibp->rvp.n_loop_pkts++;
 flush_send:
 	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
 	hfi1_send_complete(sqp, wqe, send_status);
@@ -580,7 +589,7 @@ rnr_nak:
 	/* Handle RNR NAK */
 	if (qp->ibqp.qp_type == IB_QPT_UC)
 		goto send_comp;
-	ibp->n_rnr_naks++;
+	ibp->rvp.n_rnr_naks++;
 	/*
 	 * Note: we don't need the s_lock held since the BUSY flag
 	 * makes this single threaded.
@@ -592,13 +601,10 @@ rnr_nak:
 	if (sqp->s_rnr_retry_cnt < 7)
 		sqp->s_rnr_retry--;
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	if (!(ib_hfi1_state_ops[sqp->state] & HFI1_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
 		goto clr_busy;
-	sqp->s_flags |= HFI1_S_WAIT_RNR;
-	sqp->s_timer.function = hfi1_rc_rnr_retry;
-	sqp->s_timer.expires = jiffies +
-		usecs_to_jiffies(ib_hfi1_rnr_table[qp->r_min_rnr_timer]);
-	add_timer(&sqp->s_timer);
+	to = ib_hfi1_rnr_table[qp->r_min_rnr_timer];
+	hfi1_add_rnr_timer(sqp, to);
 	goto clr_busy;
 
 op_err:
@@ -622,9 +628,9 @@ serr:
 	spin_lock_irqsave(&sqp->s_lock, flags);
 	hfi1_send_complete(sqp, wqe, send_status);
 	if (sqp->ibqp.qp_type == IB_QPT_RC) {
-		int lastwqe = hfi1_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
 
-		sqp->s_flags &= ~HFI1_S_BUSY;
+		sqp->s_flags &= ~RVT_S_BUSY;
 		spin_unlock_irqrestore(&sqp->s_lock, flags);
 		if (lastwqe) {
 			struct ib_event ev;
@@ -637,7 +643,7 @@ serr:
 		goto done;
 	}
 clr_busy:
-	sqp->s_flags &= ~HFI1_S_BUSY;
+	sqp->s_flags &= ~RVT_S_BUSY;
 unlock:
 	spin_unlock_irqrestore(&sqp->s_lock, flags);
 done:
@@ -666,7 +672,7 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
 	hdr->next_hdr = IB_GRH_NEXT_HDR;
 	hdr->hop_limit = grh->hop_limit;
 	/* The SGID is 32-bit aligned. */
-	hdr->sgid.global.subnet_prefix = ibp->gid_prefix;
+	hdr->sgid.global.subnet_prefix = ibp->rvp.gid_prefix;
 	hdr->sgid.global.interface_id =
 		grh->sgid_index && grh->sgid_index < ARRAY_SIZE(ibp->guids) ?
 		ibp->guids[grh->sgid_index - 1] :
@@ -690,29 +696,31 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
  * Subsequent middles use the copied entry, editing the
  * PSN with 1 or 2 edits.
  */
-static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
+static inline void build_ahg(struct rvt_qp *qp, u32 npsn)
 {
-	if (unlikely(qp->s_flags & HFI1_S_AHG_CLEAR))
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	if (unlikely(qp->s_flags & RVT_S_AHG_CLEAR))
 		clear_ahg(qp);
-	if (!(qp->s_flags & HFI1_S_AHG_VALID)) {
+	if (!(qp->s_flags & RVT_S_AHG_VALID)) {
 		/* first middle that needs copy  */
 		if (qp->s_ahgidx < 0)
-			qp->s_ahgidx = sdma_ahg_alloc(qp->s_sde);
+			qp->s_ahgidx = sdma_ahg_alloc(priv->s_sde);
 		if (qp->s_ahgidx >= 0) {
 			qp->s_ahgpsn = npsn;
-			qp->s_hdr->tx_flags |= SDMA_TXREQ_F_AHG_COPY;
+			priv->s_hdr->tx_flags |= SDMA_TXREQ_F_AHG_COPY;
 			/* save to protect a change in another thread */
-			qp->s_hdr->sde = qp->s_sde;
-			qp->s_hdr->ahgidx = qp->s_ahgidx;
-			qp->s_flags |= HFI1_S_AHG_VALID;
+			priv->s_hdr->sde = priv->s_sde;
+			priv->s_hdr->ahgidx = qp->s_ahgidx;
+			qp->s_flags |= RVT_S_AHG_VALID;
 		}
 	} else {
 		/* subsequent middle after valid */
 		if (qp->s_ahgidx >= 0) {
-			qp->s_hdr->tx_flags |= SDMA_TXREQ_F_USE_AHG;
-			qp->s_hdr->ahgidx = qp->s_ahgidx;
-			qp->s_hdr->ahgcount++;
-			qp->s_hdr->ahgdesc[0] =
+			priv->s_hdr->tx_flags |= SDMA_TXREQ_F_USE_AHG;
+			priv->s_hdr->ahgidx = qp->s_ahgidx;
+			priv->s_hdr->ahgcount++;
+			priv->s_hdr->ahgdesc[0] =
 				sdma_build_ahg_descriptor(
 					(__force u16)cpu_to_be16((u16)npsn),
 					BTH2_OFFSET,
@@ -720,8 +728,8 @@ static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
 					16);
 			if ((npsn & 0xffff0000) !=
 					(qp->s_ahgpsn & 0xffff0000)) {
-				qp->s_hdr->ahgcount++;
-				qp->s_hdr->ahgdesc[1] =
+				priv->s_hdr->ahgcount++;
+				priv->s_hdr->ahgdesc[1] =
 					sdma_build_ahg_descriptor(
 						(__force u16)cpu_to_be16(
 							(u16)(npsn >> 16)),
@@ -733,10 +741,12 @@ static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
 	}
 }
 
-void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
-			  u32 bth0, u32 bth2, int middle)
+void hfi1_make_ruc_header(struct rvt_qp *qp, struct hfi1_other_headers *ohdr,
+			  u32 bth0, u32 bth2, int middle,
+			  struct hfi1_pkt_state *ps)
 {
-	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct hfi1_qp_priv *priv = qp->priv;
+	struct hfi1_ibport *ibp = ps->ibp;
 	u16 lrh0;
 	u32 nwords;
 	u32 extra_bytes;
@@ -747,13 +757,14 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
 	nwords = (qp->s_cur_size + extra_bytes) >> 2;
 	lrh0 = HFI1_LRH_BTH;
 	if (unlikely(qp->remote_ah_attr.ah_flags & IB_AH_GRH)) {
-		qp->s_hdrwords += hfi1_make_grh(ibp, &qp->s_hdr->ibh.u.l.grh,
-					       &qp->remote_ah_attr.grh,
-					       qp->s_hdrwords, nwords);
+		qp->s_hdrwords += hfi1_make_grh(ibp,
+						&ps->s_txreq->phdr.hdr.u.l.grh,
+						&qp->remote_ah_attr.grh,
+						qp->s_hdrwords, nwords);
 		lrh0 = HFI1_LRH_GRH;
 		middle = 0;
 	}
-	lrh0 |= (qp->s_sc & 0xf) << 12 | (qp->remote_ah_attr.sl & 0xf) << 4;
+	lrh0 |= (priv->s_sc & 0xf) << 12 | (qp->remote_ah_attr.sl & 0xf) << 4;
 	/*
 	 * reset s_hdr/AHG fields
 	 *
@@ -765,10 +776,10 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
 	 * build_ahg() will modify as appropriate
 	 * to use the AHG feature.
 	 */
-	qp->s_hdr->tx_flags = 0;
-	qp->s_hdr->ahgcount = 0;
-	qp->s_hdr->ahgidx = 0;
-	qp->s_hdr->sde = NULL;
+	priv->s_hdr->tx_flags = 0;
+	priv->s_hdr->ahgcount = 0;
+	priv->s_hdr->ahgidx = 0;
+	priv->s_hdr->sde = NULL;
 	if (qp->s_mig_state == IB_MIG_MIGRATED)
 		bth0 |= IB_BTH_MIG_REQ;
 	else
@@ -776,19 +787,19 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
 	if (middle)
 		build_ahg(qp, bth2);
 	else
-		qp->s_flags &= ~HFI1_S_AHG_VALID;
-	qp->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
-	qp->s_hdr->ibh.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
-	qp->s_hdr->ibh.lrh[2] =
+		qp->s_flags &= ~RVT_S_AHG_VALID;
+	ps->s_txreq->phdr.hdr.lrh[0] = cpu_to_be16(lrh0);
+	ps->s_txreq->phdr.hdr.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
+	ps->s_txreq->phdr.hdr.lrh[2] =
 		cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
-	qp->s_hdr->ibh.lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
+	ps->s_txreq->phdr.hdr.lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
 				       qp->remote_ah_attr.src_path_bits);
 	bth0 |= hfi1_get_pkey(ibp, qp->s_pkey_index);
 	bth0 |= extra_bytes << 20;
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	bth1 = qp->remote_qpn;
-	if (qp->s_flags & HFI1_S_ECN) {
-		qp->s_flags &= ~HFI1_S_ECN;
+	if (qp->s_flags & RVT_S_ECN) {
+		qp->s_flags &= ~RVT_S_ECN;
 		/* we recently received a FECN, so return a BECN */
 		bth1 |= (HFI1_BECN_MASK << HFI1_BECN_SHIFT);
 	}
@@ -799,6 +810,14 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
 /* when sending, force a reschedule every one of these periods */
 #define SEND_RESCHED_TIMEOUT (5 * HZ)  /* 5s in jiffies */
 
+void _hfi1_do_send(struct work_struct *work)
+{
+	struct iowait *wait = container_of(work, struct iowait, iowork);
+	struct rvt_qp *qp = iowait_to_qp(wait);
+
+	hfi1_do_send(qp);
+}
+
 /**
  * hfi1_do_send - perform a send on a QP
  * @work: contains a pointer to the QP
@@ -807,34 +826,45 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
  * exhausted.  Only allow one CPU to send a packet per QP (tasklet).
  * Otherwise, two threads could send packets out of order.
  */
-void hfi1_do_send(struct work_struct *work)
+void hfi1_do_send(struct rvt_qp *qp)
 {
-	struct iowait *wait = container_of(work, struct iowait, iowork);
-	struct hfi1_qp *qp = container_of(wait, struct hfi1_qp, s_iowait);
 	struct hfi1_pkt_state ps;
-	int (*make_req)(struct hfi1_qp *qp);
+	struct hfi1_qp_priv *priv = qp->priv;
+	int (*make_req)(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 	unsigned long flags;
 	unsigned long timeout;
+	unsigned long timeout_int;
+	int cpu;
 
 	ps.dev = to_idev(qp->ibqp.device);
 	ps.ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ps.ppd = ppd_from_ibp(ps.ibp);
 
-	if ((qp->ibqp.qp_type == IB_QPT_RC ||
-	     qp->ibqp.qp_type == IB_QPT_UC) &&
-	    !loopback &&
-	    (qp->remote_ah_attr.dlid & ~((1 << ps.ppd->lmc) - 1)) ==
-	    ps.ppd->lid) {
-		ruc_loopback(qp);
-		return;
-	}
-
-	if (qp->ibqp.qp_type == IB_QPT_RC)
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+		if (!loopback && ((qp->remote_ah_attr.dlid & ~((1 << ps.ppd->lmc
+								) - 1)) ==
+				 ps.ppd->lid)) {
+			ruc_loopback(qp);
+			return;
+		}
 		make_req = hfi1_make_rc_req;
-	else if (qp->ibqp.qp_type == IB_QPT_UC)
+		timeout_int = (qp->timeout_jiffies);
+		break;
+	case IB_QPT_UC:
+		if (!loopback && ((qp->remote_ah_attr.dlid & ~((1 << ps.ppd->lmc
+								) - 1)) ==
+				 ps.ppd->lid)) {
+			ruc_loopback(qp);
+			return;
+		}
 		make_req = hfi1_make_uc_req;
-	else
+		timeout_int = SEND_RESCHED_TIMEOUT;
+		break;
+	default:
 		make_req = hfi1_make_ud_req;
+		timeout_int = SEND_RESCHED_TIMEOUT;
+	}
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 
@@ -844,57 +874,83 @@ void hfi1_do_send(struct work_struct *work)
 		return;
 	}
 
-	qp->s_flags |= HFI1_S_BUSY;
+	qp->s_flags |= RVT_S_BUSY;
 
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-
-	timeout = jiffies + SEND_RESCHED_TIMEOUT;
+	timeout = jiffies + (timeout_int) / 8;
+	cpu = priv->s_sde ? priv->s_sde->cpu :
+			cpumask_first(cpumask_of_node(ps.ppd->dd->node));
+	/* insure a pre-built packet is handled  */
+	ps.s_txreq = get_waiting_verbs_txreq(qp);
 	do {
 		/* Check for a constructed packet to be sent. */
 		if (qp->s_hdrwords != 0) {
+			spin_unlock_irqrestore(&qp->s_lock, flags);
 			/*
 			 * If the packet cannot be sent now, return and
 			 * the send tasklet will be woken up later.
 			 */
 			if (hfi1_verbs_send(qp, &ps))
-				break;
+				return;
 			/* Record that s_hdr is empty. */
 			qp->s_hdrwords = 0;
+			/* allow other tasks to run */
+			if (unlikely(time_after(jiffies, timeout))) {
+				if (workqueue_congested(cpu,
+							ps.ppd->hfi1_wq)) {
+					spin_lock_irqsave(&qp->s_lock, flags);
+					qp->s_flags &= ~RVT_S_BUSY;
+					hfi1_schedule_send(qp);
+					spin_unlock_irqrestore(&qp->s_lock,
+							       flags);
+					this_cpu_inc(
+						*ps.ppd->dd->send_schedule);
+					return;
+				}
+				if (!irqs_disabled()) {
+					cond_resched();
+					this_cpu_inc(
+					   *ps.ppd->dd->send_schedule);
+				}
+				timeout = jiffies + (timeout_int) / 8;
+			}
+			spin_lock_irqsave(&qp->s_lock, flags);
 		}
+	} while (make_req(qp, &ps));
 
-		/* allow other tasks to run */
-		if (unlikely(time_after(jiffies, timeout))) {
-			cond_resched();
-			ps.ppd->dd->verbs_dev.n_send_schedule++;
-			timeout = jiffies + SEND_RESCHED_TIMEOUT;
-		}
-	} while (make_req(qp));
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
 
 /*
  * This should be called with s_lock held.
  */
-void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
+void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 			enum ib_wc_status status)
 {
 	u32 old_last, last;
 	unsigned i;
 
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_OR_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
 		return;
 
+	last = qp->s_last;
+	old_last = last;
+	if (++last >= qp->s_size)
+		last = 0;
+	qp->s_last = last;
+	/* See post_send() */
+	barrier();
 	for (i = 0; i < wqe->wr.num_sge; i++) {
-		struct hfi1_sge *sge = &wqe->sg_list[i];
+		struct rvt_sge *sge = &wqe->sg_list[i];
 
-		hfi1_put_mr(sge->mr);
+		rvt_put_mr(sge->mr);
 	}
 	if (qp->ibqp.qp_type == IB_QPT_UD ||
 	    qp->ibqp.qp_type == IB_QPT_SMI ||
 	    qp->ibqp.qp_type == IB_QPT_GSI)
-		atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
+		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
 
 	/* See ch. 11.2.4.1 and 10.7.3.1 */
-	if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
+	if (!(qp->s_flags & RVT_S_SIGNAL_REQ_WR) ||
 	    (wqe->wr.send_flags & IB_SEND_SIGNALED) ||
 	    status != IB_WC_SUCCESS) {
 		struct ib_wc wc;
@@ -906,15 +962,10 @@ void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
 		wc.qp = &qp->ibqp;
 		if (status == IB_WC_SUCCESS)
 			wc.byte_len = wqe->length;
-		hfi1_cq_enter(to_icq(qp->ibqp.send_cq), &wc,
-			      status != IB_WC_SUCCESS);
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &wc,
+			     status != IB_WC_SUCCESS);
 	}
 
-	last = qp->s_last;
-	old_last = last;
-	if (++last >= qp->s_size)
-		last = 0;
-	qp->s_last = last;
 	if (qp->s_acked == old_last)
 		qp->s_acked = last;
 	if (qp->s_cur == old_last)
diff --git a/drivers/staging/rdma/hfi1/sdma.c b/drivers/staging/rdma/hfi1/sdma.c
index 9a15f1f32b45..abb8ebc1fcac 100644
--- a/drivers/staging/rdma/hfi1/sdma.c
+++ b/drivers/staging/rdma/hfi1/sdma.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -112,10 +109,10 @@ MODULE_PARM_DESC(desct_intr, "Number of SDMA descriptor before interrupt");
 	| SD(ENG_ERR_STATUS_SDMA_HEADER_REQUEST_FIFO_UNC_ERR_SMASK))
 
 /* sdma_sendctrl operations */
-#define SDMA_SENDCTRL_OP_ENABLE    (1U << 0)
-#define SDMA_SENDCTRL_OP_INTENABLE (1U << 1)
-#define SDMA_SENDCTRL_OP_HALT      (1U << 2)
-#define SDMA_SENDCTRL_OP_CLEANUP   (1U << 3)
+#define SDMA_SENDCTRL_OP_ENABLE    BIT(0)
+#define SDMA_SENDCTRL_OP_INTENABLE BIT(1)
+#define SDMA_SENDCTRL_OP_HALT      BIT(2)
+#define SDMA_SENDCTRL_OP_CLEANUP   BIT(3)
 
 /* handle long defines */
 #define SDMA_EGRESS_PACKET_OCCUPANCY_SMASK \
@@ -325,9 +322,9 @@ static void sdma_wait_for_packet_egress(struct sdma_engine *sde,
 		if (lcnt++ > 500) {
 			/* timed out - bounce the link */
 			dd_dev_err(dd, "%s: engine %u timeout waiting for packets to egress, remaining count %u, bouncing link\n",
-				  __func__, sde->this_idx, (u32)reg);
+				   __func__, sde->this_idx, (u32)reg);
 			queue_work(dd->pport->hfi1_wq,
-				&dd->pport->link_bounce_work);
+				   &dd->pport->link_bounce_work);
 			break;
 		}
 		udelay(1);
@@ -361,6 +358,28 @@ static inline void sdma_set_desc_cnt(struct sdma_engine *sde, unsigned cnt)
 	write_sde_csr(sde, SD(DESC_CNT), reg);
 }
 
+static inline void complete_tx(struct sdma_engine *sde,
+			       struct sdma_txreq *tx,
+			       int res)
+{
+	/* protect against complete modifying */
+	struct iowait *wait = tx->wait;
+	callback_t complete = tx->complete;
+
+#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
+	trace_hfi1_sdma_out_sn(sde, tx->sn);
+	if (WARN_ON_ONCE(sde->head_sn != tx->sn))
+		dd_dev_err(sde->dd, "expected %llu got %llu\n",
+			   sde->head_sn, tx->sn);
+	sde->head_sn++;
+#endif
+	sdma_txclean(sde->dd, tx);
+	if (complete)
+		(*complete)(tx, res);
+	if (iowait_sdma_dec(wait) && wait)
+		iowait_drain_wakeup(wait);
+}
+
 /*
  * Complete all the sdma requests with a SDMA_TXREQ_S_ABORTED status
  *
@@ -395,27 +414,8 @@ static void sdma_flush(struct sdma_engine *sde)
 	}
 	spin_unlock_irqrestore(&sde->flushlist_lock, flags);
 	/* flush from flush list */
-	list_for_each_entry_safe(txp, txp_next, &flushlist, list) {
-		int drained = 0;
-		/* protect against complete modifying */
-		struct iowait *wait = txp->wait;
-
-		list_del_init(&txp->list);
-#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
-		trace_hfi1_sdma_out_sn(sde, txp->sn);
-		if (WARN_ON_ONCE(sde->head_sn != txp->sn))
-			dd_dev_err(sde->dd, "expected %llu got %llu\n",
-				sde->head_sn, txp->sn);
-		sde->head_sn++;
-#endif
-		sdma_txclean(sde->dd, txp);
-		if (wait)
-			drained = atomic_dec_and_test(&wait->sdma_busy);
-		if (txp->complete)
-			(*txp->complete)(txp, SDMA_TXREQ_S_ABORTED, drained);
-		if (wait && drained)
-			iowait_drain_wakeup(wait);
-	}
+	list_for_each_entry_safe(txp, txp_next, &flushlist, list)
+		complete_tx(sde, txp, SDMA_TXREQ_S_ABORTED);
 }
 
 /*
@@ -455,8 +455,8 @@ static void sdma_err_halt_wait(struct work_struct *work)
 			break;
 		if (time_after(jiffies, timeout)) {
 			dd_dev_err(sde->dd,
-				"SDMA engine %d - timeout waiting for engine to halt\n",
-				sde->this_idx);
+				   "SDMA engine %d - timeout waiting for engine to halt\n",
+				   sde->this_idx);
 			/*
 			 * Continue anyway.  This could happen if there was
 			 * an uncorrectable error in the wrong spot.
@@ -472,7 +472,6 @@ static void sdma_err_halt_wait(struct work_struct *work)
 static void sdma_err_progress_check_schedule(struct sdma_engine *sde)
 {
 	if (!is_bx(sde->dd) && HFI1_CAP_IS_KSET(SDMA_AHG)) {
-
 		unsigned index;
 		struct hfi1_devdata *dd = sde->dd;
 
@@ -531,7 +530,7 @@ static void sdma_err_progress_check(unsigned long data)
 
 static void sdma_hw_clean_up_task(unsigned long opaque)
 {
-	struct sdma_engine *sde = (struct sdma_engine *) opaque;
+	struct sdma_engine *sde = (struct sdma_engine *)opaque;
 	u64 statuscsr;
 
 	while (1) {
@@ -577,31 +576,10 @@ static void sdma_flush_descq(struct sdma_engine *sde)
 		head = ++sde->descq_head & sde->sdma_mask;
 		/* if now past this txp's descs, do the callback */
 		if (txp && txp->next_descq_idx == head) {
-			int drained = 0;
-			/* protect against complete modifying */
-			struct iowait *wait = txp->wait;
-
 			/* remove from list */
 			sde->tx_ring[sde->tx_head++ & sde->sdma_mask] = NULL;
-			if (wait)
-				drained = atomic_dec_and_test(&wait->sdma_busy);
-#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
-			trace_hfi1_sdma_out_sn(sde, txp->sn);
-			if (WARN_ON_ONCE(sde->head_sn != txp->sn))
-				dd_dev_err(sde->dd, "expected %llu got %llu\n",
-					sde->head_sn, txp->sn);
-			sde->head_sn++;
-#endif
-			sdma_txclean(sde->dd, txp);
+			complete_tx(sde, txp, SDMA_TXREQ_S_ABORTED);
 			trace_hfi1_sdma_progress(sde, head, tail, txp);
-			if (txp->complete)
-				(*txp->complete)(
-					txp,
-					SDMA_TXREQ_S_ABORTED,
-					drained);
-			if (wait && drained)
-				iowait_drain_wakeup(wait);
-			/* see if there is another txp */
 			txp = get_txhead(sde);
 		}
 		progress++;
@@ -612,7 +590,7 @@ static void sdma_flush_descq(struct sdma_engine *sde)
 
 static void sdma_sw_clean_up_task(unsigned long opaque)
 {
-	struct sdma_engine *sde = (struct sdma_engine *) opaque;
+	struct sdma_engine *sde = (struct sdma_engine *)opaque;
 	unsigned long flags;
 
 	spin_lock_irqsave(&sde->tail_lock, flags);
@@ -627,7 +605,6 @@ static void sdma_sw_clean_up_task(unsigned long opaque)
 	 *   descq are ours to play with.
 	 */
 
-
 	/*
 	 * In the error clean up sequence, software clean must be called
 	 * before the hardware clean so we can use the hardware head in
@@ -676,7 +653,7 @@ static void sdma_start_hw_clean_up(struct sdma_engine *sde)
 }
 
 static void sdma_set_state(struct sdma_engine *sde,
-	enum sdma_states next_state)
+			   enum sdma_states next_state)
 {
 	struct sdma_state *ss = &sde->state;
 	const struct sdma_set_state_action *action = sdma_action_table;
@@ -692,8 +669,8 @@ static void sdma_set_state(struct sdma_engine *sde,
 	ss->previous_op = ss->current_op;
 	ss->current_state = next_state;
 
-	if (ss->previous_state != sdma_state_s99_running
-		&& next_state == sdma_state_s99_running)
+	if (ss->previous_state != sdma_state_s99_running &&
+	    next_state == sdma_state_s99_running)
 		sdma_flush(sde);
 
 	if (action[next_state].op_enable)
@@ -890,6 +867,9 @@ int sdma_map_init(struct hfi1_devdata *dd, u8 port, u8 num_vls, u8 *vl_engines)
 	newmap->actual_vls = num_vls;
 	newmap->vls = roundup_pow_of_two(num_vls);
 	newmap->mask = (1 << ilog2(newmap->vls)) - 1;
+	/* initialize back-map */
+	for (i = 0; i < TXE_NUM_SDMA_ENGINES; i++)
+		newmap->engine_to_vl[i] = -1;
 	for (i = 0; i < newmap->vls; i++) {
 		/* save for wrap around */
 		int first_engine = engine;
@@ -913,6 +893,9 @@ int sdma_map_init(struct hfi1_devdata *dd, u8 port, u8 num_vls, u8 *vl_engines)
 					/* wrap back to first engine */
 					engine = first_engine;
 			}
+			/* assign back-map */
+			for (j = 0; j < vl_engines[i]; j++)
+				newmap->engine_to_vl[first_engine + j] = i;
 		} else {
 			/* just re-use entry without allocating */
 			newmap->map[i] = newmap->map[i % num_vls];
@@ -922,7 +905,7 @@ int sdma_map_init(struct hfi1_devdata *dd, u8 port, u8 num_vls, u8 *vl_engines)
 	/* newmap in hand, save old map */
 	spin_lock_irq(&dd->sde_map_lock);
 	oldmap = rcu_dereference_protected(dd->sdma_map,
-			lockdep_is_held(&dd->sde_map_lock));
+					   lockdep_is_held(&dd->sde_map_lock));
 
 	/* publish newmap */
 	rcu_assign_pointer(dd->sdma_map, newmap);
@@ -983,7 +966,7 @@ static void sdma_clean(struct hfi1_devdata *dd, size_t num_engines)
 		sde->tx_ring = NULL;
 	}
 	spin_lock_irq(&dd->sde_map_lock);
-	kfree(rcu_access_pointer(dd->sdma_map));
+	sdma_map_free(rcu_access_pointer(dd->sdma_map));
 	RCU_INIT_POINTER(dd->sdma_map, NULL);
 	spin_unlock_irq(&dd->sde_map_lock);
 	synchronize_rcu();
@@ -1020,19 +1003,19 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		return 0;
 	}
 	if (mod_num_sdma &&
-		/* can't exceed chip support */
-		mod_num_sdma <= dd->chip_sdma_engines &&
-		/* count must be >= vls */
-		mod_num_sdma >= num_vls)
+	    /* can't exceed chip support */
+	    mod_num_sdma <= dd->chip_sdma_engines &&
+	    /* count must be >= vls */
+	    mod_num_sdma >= num_vls)
 		num_engines = mod_num_sdma;
 
 	dd_dev_info(dd, "SDMA mod_num_sdma: %u\n", mod_num_sdma);
 	dd_dev_info(dd, "SDMA chip_sdma_engines: %u\n", dd->chip_sdma_engines);
 	dd_dev_info(dd, "SDMA chip_sdma_mem_size: %u\n",
-		dd->chip_sdma_mem_size);
+		    dd->chip_sdma_mem_size);
 
 	per_sdma_credits =
-		dd->chip_sdma_mem_size/(num_engines * SDMA_BLOCK_SIZE);
+		dd->chip_sdma_mem_size / (num_engines * SDMA_BLOCK_SIZE);
 
 	/* set up freeze waitqueue */
 	init_waitqueue_head(&dd->sdma_unfreeze_wq);
@@ -1040,7 +1023,7 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 
 	descq_cnt = sdma_get_descq_cnt();
 	dd_dev_info(dd, "SDMA engines %zu descq_cnt %u\n",
-		num_engines, descq_cnt);
+		    num_engines, descq_cnt);
 
 	/* alloc memory for array of send engines */
 	dd->per_sdma = kcalloc(num_engines, sizeof(*dd->per_sdma), GFP_KERNEL);
@@ -1061,18 +1044,18 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		sde->desc_avail = sdma_descq_freecnt(sde);
 		sde->sdma_shift = ilog2(descq_cnt);
 		sde->sdma_mask = (1 << sde->sdma_shift) - 1;
-		sde->descq_full_count = 0;
-
-		/* Create a mask for all 3 chip interrupt sources */
-		sde->imask = (u64)1 << (0*TXE_NUM_SDMA_ENGINES + this_idx)
-			| (u64)1 << (1*TXE_NUM_SDMA_ENGINES + this_idx)
-			| (u64)1 << (2*TXE_NUM_SDMA_ENGINES + this_idx);
-		/* Create a mask specifically for sdma_idle */
-		sde->idle_mask =
-			(u64)1 << (2*TXE_NUM_SDMA_ENGINES + this_idx);
-		/* Create a mask specifically for sdma_progress */
-		sde->progress_mask =
-			(u64)1 << (TXE_NUM_SDMA_ENGINES + this_idx);
+
+		/* Create a mask specifically for each interrupt source */
+		sde->int_mask = (u64)1 << (0 * TXE_NUM_SDMA_ENGINES +
+					   this_idx);
+		sde->progress_mask = (u64)1 << (1 * TXE_NUM_SDMA_ENGINES +
+						this_idx);
+		sde->idle_mask = (u64)1 << (2 * TXE_NUM_SDMA_ENGINES +
+					    this_idx);
+		/* Create a combined mask to cover all 3 interrupt sources */
+		sde->imask = sde->int_mask | sde->progress_mask |
+			     sde->idle_mask;
+
 		spin_lock_init(&sde->tail_lock);
 		seqlock_init(&sde->head_lock);
 		spin_lock_init(&sde->senddmactrl_lock);
@@ -1100,10 +1083,10 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 				SDMA_DESC1_INT_REQ_FLAG;
 
 		tasklet_init(&sde->sdma_hw_clean_up_task, sdma_hw_clean_up_task,
-			(unsigned long)sde);
+			     (unsigned long)sde);
 
 		tasklet_init(&sde->sdma_sw_clean_up_task, sdma_sw_clean_up_task,
-			(unsigned long)sde);
+			     (unsigned long)sde);
 		INIT_WORK(&sde->err_halt_worker, sdma_err_halt_wait);
 		INIT_WORK(&sde->flush_worker, sdma_field_flush);
 
@@ -1251,11 +1234,10 @@ void sdma_exit(struct hfi1_devdata *dd)
 
 	for (this_idx = 0; dd->per_sdma && this_idx < dd->num_sdma;
 			++this_idx) {
-
 		sde = &dd->per_sdma[this_idx];
 		if (!list_empty(&sde->dmawait))
 			dd_dev_err(dd, "sde %u: dmawait list not empty!\n",
-				sde->this_idx);
+				   sde->this_idx);
 		sdma_process_event(sde, sdma_event_e00_go_hw_down);
 
 		del_timer_sync(&sde->err_progress_check_timer);
@@ -1358,8 +1340,8 @@ retry:
 	use_dmahead = HFI1_CAP_IS_KSET(USE_SDMA_HEAD) && __sdma_running(sde) &&
 					(dd->flags & HFI1_HAS_SDMA_TIMEOUT);
 	hwhead = use_dmahead ?
-		(u16) le64_to_cpu(*sde->head_dma) :
-		(u16) read_sde_csr(sde, SD(HEAD));
+		(u16)le64_to_cpu(*sde->head_dma) :
+		(u16)read_sde_csr(sde, SD(HEAD));
 
 	if (unlikely(HFI1_CAP_IS_KSET(SDMA_HEAD_CHECK))) {
 		u16 cnt;
@@ -1385,9 +1367,9 @@ retry:
 
 		if (unlikely(!sane)) {
 			dd_dev_err(dd, "SDMA(%u) bad head (%s) hwhd=%hu swhd=%hu swtl=%hu cnt=%hu\n",
-				sde->this_idx,
-				use_dmahead ? "dma" : "kreg",
-				hwhead, swhead, swtail, cnt);
+				   sde->this_idx,
+				   use_dmahead ? "dma" : "kreg",
+				   hwhead, swhead, swtail, cnt);
 			if (use_dmahead) {
 				/* try one more time, using csr */
 				use_dmahead = 0;
@@ -1464,7 +1446,7 @@ static void sdma_make_progress(struct sdma_engine *sde, u64 status)
 {
 	struct sdma_txreq *txp = NULL;
 	int progress = 0;
-	u16 hwhead, swhead, swtail;
+	u16 hwhead, swhead;
 	int idle_check_done = 0;
 
 	hwhead = sdma_gethead(sde);
@@ -1485,29 +1467,9 @@ retry:
 
 		/* if now past this txp's descs, do the callback */
 		if (txp && txp->next_descq_idx == swhead) {
-			int drained = 0;
-			/* protect against complete modifying */
-			struct iowait *wait = txp->wait;
-
 			/* remove from list */
 			sde->tx_ring[sde->tx_head++ & sde->sdma_mask] = NULL;
-			if (wait)
-				drained = atomic_dec_and_test(&wait->sdma_busy);
-#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
-			trace_hfi1_sdma_out_sn(sde, txp->sn);
-			if (WARN_ON_ONCE(sde->head_sn != txp->sn))
-				dd_dev_err(sde->dd, "expected %llu got %llu\n",
-					sde->head_sn, txp->sn);
-			sde->head_sn++;
-#endif
-			sdma_txclean(sde->dd, txp);
-			if (txp->complete)
-				(*txp->complete)(
-					txp,
-					SDMA_TXREQ_S_OK,
-					drained);
-			if (wait && drained)
-				iowait_drain_wakeup(wait);
+			complete_tx(sde, txp, SDMA_TXREQ_S_OK);
 			/* see if there is another txp */
 			txp = get_txhead(sde);
 		}
@@ -1525,6 +1487,8 @@ retry:
 	 * of sdma_make_progress(..) which is ensured by idle_check_done flag
 	 */
 	if ((status & sde->idle_mask) && !idle_check_done) {
+		u16 swtail;
+
 		swtail = ACCESS_ONCE(sde->descq_tail) & sde->sdma_mask;
 		if (swtail != hwhead) {
 			hwhead = (u16)read_sde_csr(sde, SD(HEAD));
@@ -1552,6 +1516,12 @@ void sdma_engine_interrupt(struct sdma_engine *sde, u64 status)
 	trace_hfi1_sdma_engine_interrupt(sde, status);
 	write_seqlock(&sde->head_lock);
 	sdma_set_desc_cnt(sde, sdma_desct_intr);
+	if (status & sde->idle_mask)
+		sde->idle_int_cnt++;
+	else if (status & sde->progress_mask)
+		sde->progress_int_cnt++;
+	else if (status & sde->int_mask)
+		sde->sdma_int_cnt++;
 	sdma_make_progress(sde, status);
 	write_sequnlock(&sde->head_lock);
 }
@@ -1577,10 +1547,10 @@ void sdma_engine_error(struct sdma_engine *sde, u64 status)
 		__sdma_process_event(sde, sdma_event_e60_hw_halted);
 	if (status & ~SD(ENG_ERR_STATUS_SDMA_HALT_ERR_SMASK)) {
 		dd_dev_err(sde->dd,
-			"SDMA (%u) engine error: 0x%llx state %s\n",
-			sde->this_idx,
-			(unsigned long long)status,
-			sdma_state_names[sde->state.current_state]);
+			   "SDMA (%u) engine error: 0x%llx state %s\n",
+			   sde->this_idx,
+			   (unsigned long long)status,
+			   sdma_state_names[sde->state.current_state]);
 		dump_sdma_state(sde);
 	}
 	write_sequnlock(&sde->head_lock);
@@ -1624,8 +1594,8 @@ static void sdma_sendctrl(struct sdma_engine *sde, unsigned op)
 
 	if (op & SDMA_SENDCTRL_OP_CLEANUP)
 		write_sde_csr(sde, SD(CTRL),
-			sde->p_senddmactrl |
-			SD(CTRL_SDMA_CLEANUP_SMASK));
+			      sde->p_senddmactrl |
+			      SD(CTRL_SDMA_CLEANUP_SMASK));
 	else
 		write_sde_csr(sde, SD(CTRL), sde->p_senddmactrl);
 
@@ -1649,12 +1619,10 @@ static void sdma_setlengen(struct sdma_engine *sde)
 	 * generation counter.
 	 */
 	write_sde_csr(sde, SD(LEN_GEN),
-		(sde->descq_cnt/64) << SD(LEN_GEN_LENGTH_SHIFT)
-	);
+		      (sde->descq_cnt / 64) << SD(LEN_GEN_LENGTH_SHIFT));
 	write_sde_csr(sde, SD(LEN_GEN),
-		((sde->descq_cnt/64) << SD(LEN_GEN_LENGTH_SHIFT))
-		| (4ULL << SD(LEN_GEN_GENERATION_SHIFT))
-	);
+		      ((sde->descq_cnt / 64) << SD(LEN_GEN_LENGTH_SHIFT)) |
+		      (4ULL << SD(LEN_GEN_GENERATION_SHIFT)));
 }
 
 static inline void sdma_update_tail(struct sdma_engine *sde, u16 tail)
@@ -1714,7 +1682,6 @@ static void set_sdma_integrity(struct sdma_engine *sde)
 	write_sde_csr(sde, SD(CHECK_ENABLE), reg);
 }
 
-
 static void init_sdma_regs(
 	struct sdma_engine *sde,
 	u32 credits,
@@ -1735,17 +1702,16 @@ static void init_sdma_regs(
 	write_sde_csr(sde, SD(DESC_CNT), 0);
 	write_sde_csr(sde, SD(HEAD_ADDR), sde->head_phys);
 	write_sde_csr(sde, SD(MEMORY),
-		((u64)credits <<
-			SD(MEMORY_SDMA_MEMORY_CNT_SHIFT)) |
-		((u64)(credits * sde->this_idx) <<
-			SD(MEMORY_SDMA_MEMORY_INDEX_SHIFT)));
+		      ((u64)credits << SD(MEMORY_SDMA_MEMORY_CNT_SHIFT)) |
+		      ((u64)(credits * sde->this_idx) <<
+		       SD(MEMORY_SDMA_MEMORY_INDEX_SHIFT)));
 	write_sde_csr(sde, SD(ENG_ERR_MASK), ~0ull);
 	set_sdma_integrity(sde);
 	opmask = OPCODE_CHECK_MASK_DISABLED;
 	opval = OPCODE_CHECK_VAL_DISABLED;
 	write_sde_csr(sde, SD(CHECK_OPCODE),
-		(opmask << SEND_CTXT_CHECK_OPCODE_MASK_SHIFT) |
-		(opval << SEND_CTXT_CHECK_OPCODE_VALUE_SHIFT));
+		      (opmask << SEND_CTXT_CHECK_OPCODE_MASK_SHIFT) |
+		      (opval << SEND_CTXT_CHECK_OPCODE_VALUE_SHIFT));
 }
 
 #ifdef CONFIG_SDMA_VERBOSITY
@@ -1824,12 +1790,9 @@ static void dump_sdma_state(struct sdma_engine *sde)
 	descq = sde->descq;
 
 	dd_dev_err(sde->dd,
-		"SDMA (%u) descq_head: %u descq_tail: %u freecnt: %u FLE %d\n",
-		sde->this_idx,
-		head,
-		tail,
-		cnt,
-		!list_empty(&sde->flushlist));
+		   "SDMA (%u) descq_head: %u descq_tail: %u freecnt: %u FLE %d\n",
+		   sde->this_idx, head, tail, cnt,
+		   !list_empty(&sde->flushlist));
 
 	/* print info for each entry in the descriptor queue */
 	while (head != tail) {
@@ -1850,20 +1813,23 @@ static void dump_sdma_state(struct sdma_engine *sde)
 		len = (desc[0] >> SDMA_DESC0_BYTE_COUNT_SHIFT)
 			& SDMA_DESC0_BYTE_COUNT_MASK;
 		dd_dev_err(sde->dd,
-			"SDMA sdmadesc[%u]: flags:%s addr:0x%016llx gen:%u len:%u bytes\n",
-			 head, flags, addr, gen, len);
+			   "SDMA sdmadesc[%u]: flags:%s addr:0x%016llx gen:%u len:%u bytes\n",
+			   head, flags, addr, gen, len);
 		dd_dev_err(sde->dd,
-			"\tdesc0:0x%016llx desc1 0x%016llx\n",
-			 desc[0], desc[1]);
+			   "\tdesc0:0x%016llx desc1 0x%016llx\n",
+			   desc[0], desc[1]);
 		if (desc[0] & SDMA_DESC0_FIRST_DESC_FLAG)
 			dd_dev_err(sde->dd,
-				"\taidx: %u amode: %u alen: %u\n",
-				(u8)((desc[1] & SDMA_DESC1_HEADER_INDEX_SMASK)
-					>> SDMA_DESC1_HEADER_INDEX_SHIFT),
-				(u8)((desc[1] & SDMA_DESC1_HEADER_MODE_SMASK)
-					>> SDMA_DESC1_HEADER_MODE_SHIFT),
-				(u8)((desc[1] & SDMA_DESC1_HEADER_DWS_SMASK)
-					>> SDMA_DESC1_HEADER_DWS_SHIFT));
+				   "\taidx: %u amode: %u alen: %u\n",
+				   (u8)((desc[1] &
+					 SDMA_DESC1_HEADER_INDEX_SMASK) >>
+					SDMA_DESC1_HEADER_INDEX_SHIFT),
+				   (u8)((desc[1] &
+					 SDMA_DESC1_HEADER_MODE_SMASK) >>
+					SDMA_DESC1_HEADER_MODE_SHIFT),
+				   (u8)((desc[1] &
+					 SDMA_DESC1_HEADER_DWS_SMASK) >>
+					SDMA_DESC1_HEADER_DWS_SHIFT));
 		head++;
 		head &= sde->sdma_mask;
 	}
@@ -1890,29 +1856,26 @@ void sdma_seqfile_dump_sde(struct seq_file *s, struct sdma_engine *sde)
 	head = sde->descq_head & sde->sdma_mask;
 	tail = ACCESS_ONCE(sde->descq_tail) & sde->sdma_mask;
 	seq_printf(s, SDE_FMT, sde->this_idx,
-		sde->cpu,
-		sdma_state_name(sde->state.current_state),
-		(unsigned long long)read_sde_csr(sde, SD(CTRL)),
-		(unsigned long long)read_sde_csr(sde, SD(STATUS)),
-		(unsigned long long)read_sde_csr(sde,
-			SD(ENG_ERR_STATUS)),
-		(unsigned long long)read_sde_csr(sde, SD(TAIL)),
-		tail,
-		(unsigned long long)read_sde_csr(sde, SD(HEAD)),
-		head,
-		(unsigned long long)le64_to_cpu(*sde->head_dma),
-		(unsigned long long)read_sde_csr(sde, SD(MEMORY)),
-		(unsigned long long)read_sde_csr(sde, SD(LEN_GEN)),
-		(unsigned long long)read_sde_csr(sde, SD(RELOAD_CNT)),
-		(unsigned long long)sde->last_status,
-		(unsigned long long)sde->ahg_bits,
-		sde->tx_tail,
-		sde->tx_head,
-		sde->descq_tail,
-		sde->descq_head,
+		   sde->cpu,
+		   sdma_state_name(sde->state.current_state),
+		   (unsigned long long)read_sde_csr(sde, SD(CTRL)),
+		   (unsigned long long)read_sde_csr(sde, SD(STATUS)),
+		   (unsigned long long)read_sde_csr(sde, SD(ENG_ERR_STATUS)),
+		   (unsigned long long)read_sde_csr(sde, SD(TAIL)), tail,
+		   (unsigned long long)read_sde_csr(sde, SD(HEAD)), head,
+		   (unsigned long long)le64_to_cpu(*sde->head_dma),
+		   (unsigned long long)read_sde_csr(sde, SD(MEMORY)),
+		   (unsigned long long)read_sde_csr(sde, SD(LEN_GEN)),
+		   (unsigned long long)read_sde_csr(sde, SD(RELOAD_CNT)),
+		   (unsigned long long)sde->last_status,
+		   (unsigned long long)sde->ahg_bits,
+		   sde->tx_tail,
+		   sde->tx_head,
+		   sde->descq_tail,
+		   sde->descq_head,
 		   !list_empty(&sde->flushlist),
-		sde->descq_full_count,
-		(unsigned long long)read_sde_csr(sde, SEND_DMA_CHECK_SLID));
+		   sde->descq_full_count,
+		   (unsigned long long)read_sde_csr(sde, SEND_DMA_CHECK_SLID));
 
 	/* print info for each entry in the descriptor queue */
 	while (head != tail) {
@@ -1933,14 +1896,16 @@ void sdma_seqfile_dump_sde(struct seq_file *s, struct sdma_engine *sde)
 		len = (desc[0] >> SDMA_DESC0_BYTE_COUNT_SHIFT)
 			& SDMA_DESC0_BYTE_COUNT_MASK;
 		seq_printf(s,
-			"\tdesc[%u]: flags:%s addr:0x%016llx gen:%u len:%u bytes\n",
-			head, flags, addr, gen, len);
+			   "\tdesc[%u]: flags:%s addr:0x%016llx gen:%u len:%u bytes\n",
+			   head, flags, addr, gen, len);
 		if (desc[0] & SDMA_DESC0_FIRST_DESC_FLAG)
 			seq_printf(s, "\t\tahgidx: %u ahgmode: %u\n",
-				(u8)((desc[1] & SDMA_DESC1_HEADER_INDEX_SMASK)
-					>> SDMA_DESC1_HEADER_INDEX_SHIFT),
-				(u8)((desc[1] & SDMA_DESC1_HEADER_MODE_SMASK)
-					>> SDMA_DESC1_HEADER_MODE_SHIFT));
+				   (u8)((desc[1] &
+					 SDMA_DESC1_HEADER_INDEX_SMASK) >>
+					SDMA_DESC1_HEADER_INDEX_SHIFT),
+				   (u8)((desc[1] &
+					 SDMA_DESC1_HEADER_MODE_SMASK) >>
+					SDMA_DESC1_HEADER_MODE_SHIFT));
 		head = (head + 1) & sde->sdma_mask;
 	}
 }
@@ -2041,8 +2006,9 @@ static int sdma_check_progress(
 		ret = wait->sleep(sde, wait, tx, seq);
 		if (ret == -EAGAIN)
 			sde->desc_avail = sdma_descq_freecnt(sde);
-	} else
+	} else {
 		ret = -EBUSY;
+	}
 	return ret;
 }
 
@@ -2080,14 +2046,14 @@ retry:
 		goto nodesc;
 	tail = submit_tx(sde, tx);
 	if (wait)
-		atomic_inc(&wait->sdma_busy);
+		iowait_sdma_inc(wait);
 	sdma_update_tail(sde, tail);
 unlock:
 	spin_unlock_irqrestore(&sde->tail_lock, flags);
 	return ret;
 unlock_noconn:
 	if (wait)
-		atomic_inc(&wait->sdma_busy);
+		iowait_sdma_inc(wait);
 	tx->next_descq_idx = 0;
 #ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
 	tx->sn = sde->tail_sn++;
@@ -2132,13 +2098,12 @@ nodesc:
  * side locking.
  *
  * Return:
- * 0 - Success, -EINVAL - sdma_txreq incomplete, -EBUSY - no space in ring
- * (wait == NULL)
+ * > 0 - Success (value is number of sdma_txreq's submitted),
+ * -EINVAL - sdma_txreq incomplete, -EBUSY - no space in ring (wait == NULL)
  * -EIOCBQUEUED - tx queued to iowait, -ECOMM bad sdma state
  */
-int sdma_send_txlist(struct sdma_engine *sde,
-		    struct iowait *wait,
-		    struct list_head *tx_list)
+int sdma_send_txlist(struct sdma_engine *sde, struct iowait *wait,
+		     struct list_head *tx_list)
 {
 	struct sdma_txreq *tx, *tx_next;
 	int ret = 0;
@@ -2169,18 +2134,18 @@ retry:
 	}
 update_tail:
 	if (wait)
-		atomic_add(count, &wait->sdma_busy);
+		iowait_sdma_add(wait, count);
 	if (tail != INVALID_TAIL)
 		sdma_update_tail(sde, tail);
 	spin_unlock_irqrestore(&sde->tail_lock, flags);
-	return ret;
+	return ret == 0 ? count : ret;
 unlock_noconn:
 	spin_lock(&sde->flushlist_lock);
 	list_for_each_entry_safe(tx, tx_next, tx_list, list) {
 		tx->wait = wait;
 		list_del_init(&tx->list);
 		if (wait)
-			atomic_inc(&wait->sdma_busy);
+			iowait_sdma_inc(wait);
 		tx->next_descq_idx = 0;
 #ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
 		tx->sn = sde->tail_sn++;
@@ -2206,8 +2171,7 @@ nodesc:
 	goto update_tail;
 }
 
-static void sdma_process_event(struct sdma_engine *sde,
-	enum sdma_events event)
+static void sdma_process_event(struct sdma_engine *sde, enum sdma_events event)
 {
 	unsigned long flags;
 
@@ -2224,7 +2188,7 @@ static void sdma_process_event(struct sdma_engine *sde,
 }
 
 static void __sdma_process_event(struct sdma_engine *sde,
-	enum sdma_events event)
+				 enum sdma_events event)
 {
 	struct sdma_state *ss = &sde->state;
 	int need_progress = 0;
@@ -2247,14 +2211,15 @@ static void __sdma_process_event(struct sdma_engine *sde,
 			 * of link up, then we need to start up.
 			 * This can happen when hw down is requested while
 			 * bringing the link up with traffic active on
-			 * 7220, e.g. */
+			 * 7220, e.g.
+			 */
 			ss->go_s99_running = 1;
 			/* fall through and start dma engine */
 		case sdma_event_e10_go_hw_start:
 			/* This reference means the state machine is started */
 			sdma_get(&sde->state);
 			sdma_set_state(sde,
-				sdma_state_s10_hw_start_up_halt_wait);
+				       sdma_state_s10_hw_start_up_halt_wait);
 			break;
 		case sdma_event_e15_hw_halt_done:
 			break;
@@ -2292,7 +2257,7 @@ static void __sdma_process_event(struct sdma_engine *sde,
 			break;
 		case sdma_event_e15_hw_halt_done:
 			sdma_set_state(sde,
-				sdma_state_s15_hw_start_up_clean_wait);
+				       sdma_state_s15_hw_start_up_clean_wait);
 			sdma_start_hw_clean_up(sde);
 			break;
 		case sdma_event_e25_hw_clean_up_done:
@@ -2767,7 +2732,7 @@ enomem:
  * This function calls _extend_sdma_tx_descs to extend or allocate
  * coalesce buffer. If there is a allocated coalesce buffer, it will
  * copy the input packet data into the coalesce buffer. It also adds
- * coalesce buffer descriptor once whe whole packet is received.
+ * coalesce buffer descriptor once when whole packet is received.
  *
  * Return:
  * <0 - error
@@ -3030,7 +2995,8 @@ void sdma_freeze(struct hfi1_devdata *dd)
 	 * continuing.
 	 */
 	ret = wait_event_interruptible(dd->sdma_unfreeze_wq,
-				atomic_read(&dd->sdma_unfreeze_count) <= 0);
+				       atomic_read(&dd->sdma_unfreeze_count) <=
+				       0);
 	/* interrupted or count is negative, then unloading - just exit */
 	if (ret || atomic_read(&dd->sdma_unfreeze_count) < 0)
 		return;
@@ -3047,7 +3013,7 @@ void sdma_freeze(struct hfi1_devdata *dd)
 	 * software clean will read engine CSRs, so must be completed before
 	 * the next step, which will clear the engine CSRs.
 	 */
-	(void) wait_event_interruptible(dd->sdma_unfreeze_wq,
+	(void)wait_event_interruptible(dd->sdma_unfreeze_wq,
 				atomic_read(&dd->sdma_unfreeze_count) <= 0);
 	/* no need to check results - done no matter what */
 }
@@ -3067,7 +3033,7 @@ void sdma_unfreeze(struct hfi1_devdata *dd)
 	/* tell all engines start freeze clean up */
 	for (i = 0; i < dd->num_sdma; i++)
 		sdma_process_event(&dd->per_sdma[i],
-					sdma_event_e82_hw_unfreeze);
+				   sdma_event_e82_hw_unfreeze);
 }
 
 /**
@@ -3081,5 +3047,6 @@ void _sdma_engine_progress_schedule(
 	trace_hfi1_sdma_engine_progress(sde, sde->progress_mask);
 	/* assume we have selected a good cpu */
 	write_csr(sde->dd,
-		  CCE_INT_FORCE + (8*(IS_SDMA_START/64)), sde->progress_mask);
+		  CCE_INT_FORCE + (8 * (IS_SDMA_START / 64)),
+		  sde->progress_mask);
 }
diff --git a/drivers/staging/rdma/hfi1/sdma.h b/drivers/staging/rdma/hfi1/sdma.h
index da89e6458162..8f50c99fe711 100644
--- a/drivers/staging/rdma/hfi1/sdma.h
+++ b/drivers/staging/rdma/hfi1/sdma.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_SDMA_H
 #define _HFI1_SDMA_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -58,15 +55,13 @@
 
 #include "hfi.h"
 #include "verbs.h"
+#include "sdma_txreq.h"
 
-/* increased for AHG */
-#define NUM_DESC 6
 /* Hardware limit */
 #define MAX_DESC 64
 /* Hardware limit for SDMA packet size */
 #define MAX_SDMA_PKT_SIZE ((16 * 1024) - 1)
 
-
 #define SDMA_TXREQ_S_OK        0
 #define SDMA_TXREQ_S_SENDERROR 1
 #define SDMA_TXREQ_S_ABORTED   2
@@ -109,8 +104,8 @@
 /*
  * Bits defined in the send DMA descriptor.
  */
-#define SDMA_DESC0_FIRST_DESC_FLAG      (1ULL << 63)
-#define SDMA_DESC0_LAST_DESC_FLAG       (1ULL << 62)
+#define SDMA_DESC0_FIRST_DESC_FLAG      BIT_ULL(63)
+#define SDMA_DESC0_LAST_DESC_FLAG       BIT_ULL(62)
 #define SDMA_DESC0_BYTE_COUNT_SHIFT     48
 #define SDMA_DESC0_BYTE_COUNT_WIDTH     14
 #define SDMA_DESC0_BYTE_COUNT_MASK \
@@ -154,8 +149,8 @@
 	((1ULL << SDMA_DESC1_GENERATION_WIDTH) - 1)
 #define SDMA_DESC1_GENERATION_SMASK \
 	(SDMA_DESC1_GENERATION_MASK << SDMA_DESC1_GENERATION_SHIFT)
-#define SDMA_DESC1_INT_REQ_FLAG         (1ULL << 1)
-#define SDMA_DESC1_HEAD_TO_HOST_FLAG    (1ULL << 0)
+#define SDMA_DESC1_INT_REQ_FLAG         BIT_ULL(1)
+#define SDMA_DESC1_HEAD_TO_HOST_FLAG    BIT_ULL(0)
 
 enum sdma_states {
 	sdma_state_s00_hw_down,
@@ -311,83 +306,6 @@ struct hw_sdma_desc {
 	__le64 qw[2];
 };
 
-/*
- * struct sdma_desc - canonical fragment descriptor
- *
- * This is the descriptor carried in the tx request
- * corresponding to each fragment.
- *
- */
-struct sdma_desc {
-	/* private:  don't use directly */
-	u64 qw[2];
-};
-
-struct sdma_txreq;
-typedef void (*callback_t)(struct sdma_txreq *, int, int);
-
-/**
- * struct sdma_txreq - the sdma_txreq structure (one per packet)
- * @list: for use by user and by queuing for wait
- *
- * This is the representation of a packet which consists of some
- * number of fragments.   Storage is provided to within the structure.
- * for all fragments.
- *
- * The storage for the descriptors are automatically extended as needed
- * when the currently allocation is exceeded.
- *
- * The user (Verbs or PSM) may overload this structure with fields
- * specific to their use by putting this struct first in their struct.
- * The method of allocation of the overloaded structure is user dependent
- *
- * The list is the only public field in the structure.
- *
- */
-
-struct sdma_txreq {
-	struct list_head list;
-	/* private: */
-	struct sdma_desc *descp;
-	/* private: */
-	void *coalesce_buf;
-	/* private: */
-	u16 coalesce_idx;
-	/* private: */
-	struct iowait *wait;
-	/* private: */
-	callback_t                  complete;
-#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
-	u64 sn;
-#endif
-	/* private: - used in coalesce/pad processing */
-	u16                         packet_len;
-	/* private: - down-counted to trigger last */
-	u16                         tlen;
-	/* private: flags */
-	u16                         flags;
-	/* private: */
-	u16                         num_desc;
-	/* private: */
-	u16                         desc_limit;
-	/* private: */
-	u16                         next_descq_idx;
-	/* private: */
-	struct sdma_desc descs[NUM_DESC];
-};
-
-struct verbs_txreq {
-	struct hfi1_pio_header	phdr;
-	struct sdma_txreq       txreq;
-	struct hfi1_qp           *qp;
-	struct hfi1_swqe         *wqe;
-	struct hfi1_mregion	*mr;
-	struct hfi1_sge_state    *ss;
-	struct sdma_engine     *sde;
-	u16                     hdr_dwords;
-	u16                     hdr_inx;
-};
-
 /**
  * struct sdma_engine - Data pertaining to each SDMA engine.
  * @dd: a back-pointer to the device data
@@ -409,6 +327,7 @@ struct sdma_engine {
 	u64 imask;			/* clear interrupt mask */
 	u64 idle_mask;
 	u64 progress_mask;
+	u64 int_mask;
 	/* private: */
 	volatile __le64      *head_dma; /* DMA'ed by chip */
 	/* private: */
@@ -465,6 +384,12 @@ struct sdma_engine {
 	u16                   tx_head;
 	/* private: */
 	u64                   last_status;
+	/* private */
+	u64                     err_cnt;
+	/* private */
+	u64                     sdma_int_cnt;
+	u64                     idle_int_cnt;
+	u64                     progress_int_cnt;
 
 	/* private: */
 	struct list_head      dmawait;
@@ -484,12 +409,12 @@ struct sdma_engine {
 	u32                   progress_check_head;
 	/* private: */
 	struct work_struct flush_worker;
+	/* protect flush list */
 	spinlock_t flushlist_lock;
 	/* private: */
 	struct list_head flushlist;
 };
 
-
 int sdma_init(struct hfi1_devdata *dd, u8 port);
 void sdma_start(struct hfi1_devdata *dd);
 void sdma_exit(struct hfi1_devdata *dd);
@@ -535,7 +460,6 @@ static inline int __sdma_running(struct sdma_engine *engine)
 	return engine->state.current_state == sdma_state_s99_running;
 }
 
-
 /**
  * sdma_running() - state suitability test
  * @engine: sdma engine
@@ -565,7 +489,6 @@ void _sdma_txreq_ahgadd(
 	u32 *ahg,
 	u8 ahg_hlen);
 
-
 /**
  * sdma_txinit_ahg() - initialize an sdma_txreq struct with AHG
  * @tx: tx request to initialize
@@ -626,7 +549,7 @@ static inline int sdma_txinit_ahg(
 	u8 num_ahg,
 	u32 *ahg,
 	u8 ahg_hlen,
-	void (*cb)(struct sdma_txreq *, int, int))
+	void (*cb)(struct sdma_txreq *, int))
 {
 	if (tlen == 0)
 		return -ENODATA;
@@ -640,7 +563,8 @@ static inline int sdma_txinit_ahg(
 	tx->complete = cb;
 	tx->coalesce_buf = NULL;
 	tx->wait = NULL;
-	tx->tlen = tx->packet_len = tlen;
+	tx->packet_len = tlen;
+	tx->tlen = tx->packet_len;
 	tx->descs[0].qw[0] = SDMA_DESC0_FIRST_DESC_FLAG;
 	tx->descs[0].qw[1] = 0;
 	if (flags & SDMA_TXREQ_F_AHG_COPY)
@@ -689,7 +613,7 @@ static inline int sdma_txinit(
 	struct sdma_txreq *tx,
 	u16 flags,
 	u16 tlen,
-	void (*cb)(struct sdma_txreq *, int, int))
+	void (*cb)(struct sdma_txreq *, int))
 {
 	return sdma_txinit_ahg(tx, flags, tlen, 0, 0, NULL, 0, cb);
 }
@@ -753,7 +677,7 @@ static inline void _sdma_close_tx(struct hfi1_devdata *dd,
 		dd->default_desc1;
 	if (tx->flags & SDMA_TXREQ_F_URGENT)
 		tx->descp[tx->num_desc].qw[1] |=
-			(SDMA_DESC1_HEAD_TO_HOST_FLAG|
+			(SDMA_DESC1_HEAD_TO_HOST_FLAG |
 			 SDMA_DESC1_INT_REQ_FLAG);
 }
 
@@ -1080,6 +1004,7 @@ struct sdma_map_elem {
 
 /**
  * struct sdma_map_el - mapping for a vl
+ * @engine_to_vl - map of an engine to a vl
  * @list - rcu head for free callback
  * @mask - vl mask to "mod" the vl to produce an index to map array
  * @actual_vls - number of vls
@@ -1091,6 +1016,7 @@ struct sdma_map_elem {
  * in turn point to an array of sde's for that vl.
  */
 struct sdma_vl_map {
+	s8 engine_to_vl[TXE_NUM_SDMA_ENGINES];
 	struct rcu_head list;
 	u32 mask;
 	u8 actual_vls;
diff --git a/drivers/staging/rdma/hfi1/sdma_txreq.h b/drivers/staging/rdma/hfi1/sdma_txreq.h
new file mode 100644
index 000000000000..bf7d777d756e
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/sdma_txreq.h
@@ -0,0 +1,135 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef HFI1_SDMA_TXREQ_H
+#define HFI1_SDMA_TXREQ_H
+
+/* increased for AHG */
+#define NUM_DESC 6
+
+/*
+ * struct sdma_desc - canonical fragment descriptor
+ *
+ * This is the descriptor carried in the tx request
+ * corresponding to each fragment.
+ *
+ */
+struct sdma_desc {
+	/* private:  don't use directly */
+	u64 qw[2];
+};
+
+/**
+ * struct sdma_txreq - the sdma_txreq structure (one per packet)
+ * @list: for use by user and by queuing for wait
+ *
+ * This is the representation of a packet which consists of some
+ * number of fragments.   Storage is provided to within the structure.
+ * for all fragments.
+ *
+ * The storage for the descriptors are automatically extended as needed
+ * when the currently allocation is exceeded.
+ *
+ * The user (Verbs or PSM) may overload this structure with fields
+ * specific to their use by putting this struct first in their struct.
+ * The method of allocation of the overloaded structure is user dependent
+ *
+ * The list is the only public field in the structure.
+ *
+ */
+
+#define SDMA_TXREQ_S_OK        0
+#define SDMA_TXREQ_S_SENDERROR 1
+#define SDMA_TXREQ_S_ABORTED   2
+#define SDMA_TXREQ_S_SHUTDOWN  3
+
+/* flags bits */
+#define SDMA_TXREQ_F_URGENT       0x0001
+#define SDMA_TXREQ_F_AHG_COPY     0x0002
+#define SDMA_TXREQ_F_USE_AHG      0x0004
+
+struct sdma_txreq;
+typedef void (*callback_t)(struct sdma_txreq *, int);
+
+struct iowait;
+struct sdma_txreq {
+	struct list_head list;
+	/* private: */
+	struct sdma_desc *descp;
+	/* private: */
+	void *coalesce_buf;
+	/* private: */
+	struct iowait *wait;
+	/* private: */
+	callback_t                  complete;
+#ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
+	u64 sn;
+#endif
+	/* private: - used in coalesce/pad processing */
+	u16                         packet_len;
+	/* private: - down-counted to trigger last */
+	u16                         tlen;
+	/* private: */
+	u16                         num_desc;
+	/* private: */
+	u16                         desc_limit;
+	/* private: */
+	u16                         next_descq_idx;
+	/* private: */
+	u16 coalesce_idx;
+	/* private: flags */
+	u16                         flags;
+	/* private: */
+	struct sdma_desc descs[NUM_DESC];
+};
+
+static inline int sdma_txreq_built(struct sdma_txreq *tx)
+{
+	return tx->num_desc;
+}
+
+#endif                          /* HFI1_SDMA_TXREQ_H */
diff --git a/drivers/staging/rdma/hfi1/sysfs.c b/drivers/staging/rdma/hfi1/sysfs.c
index 1dd6727dd5ef..c7f1271190af 100644
--- a/drivers/staging/rdma/hfi1/sysfs.c
+++ b/drivers/staging/rdma/hfi1/sysfs.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -53,7 +50,6 @@
 #include "mad.h"
 #include "trace.h"
 
-
 /*
  * Start of per-port congestion control structures and support code
  */
@@ -62,8 +58,8 @@
  * Congestion control table size followed by table entries
  */
 static ssize_t read_cc_table_bin(struct file *filp, struct kobject *kobj,
-		struct bin_attribute *bin_attr,
-		char *buf, loff_t pos, size_t count)
+				 struct bin_attribute *bin_attr,
+				 char *buf, loff_t pos, size_t count)
 {
 	int ret;
 	struct hfi1_pportdata *ppd =
@@ -84,7 +80,7 @@ static ssize_t read_cc_table_bin(struct file *filp, struct kobject *kobj,
 
 	rcu_read_lock();
 	cc_state = get_cc_state(ppd);
-	if (cc_state == NULL) {
+	if (!cc_state) {
 		rcu_read_unlock();
 		return -EINVAL;
 	}
@@ -99,10 +95,6 @@ static void port_release(struct kobject *kobj)
 	/* nothing to do since memory is freed by hfi1_free_devdata() */
 }
 
-static struct kobj_type port_cc_ktype = {
-	.release = port_release,
-};
-
 static struct bin_attribute cc_table_bin_attr = {
 	.attr = {.name = "cc_table_bin", .mode = 0444},
 	.read = read_cc_table_bin,
@@ -115,8 +107,8 @@ static struct bin_attribute cc_table_bin_attr = {
  * trigger threshold and the minimum injection rate delay.
  */
 static ssize_t read_cc_setting_bin(struct file *filp, struct kobject *kobj,
-		struct bin_attribute *bin_attr,
-		char *buf, loff_t pos, size_t count)
+				   struct bin_attribute *bin_attr,
+				   char *buf, loff_t pos, size_t count)
 {
 	int ret;
 	struct hfi1_pportdata *ppd =
@@ -135,7 +127,7 @@ static ssize_t read_cc_setting_bin(struct file *filp, struct kobject *kobj,
 
 	rcu_read_lock();
 	cc_state = get_cc_state(ppd);
-	if (cc_state == NULL) {
+	if (!cc_state) {
 		rcu_read_unlock();
 		return -EINVAL;
 	}
@@ -151,6 +143,68 @@ static struct bin_attribute cc_setting_bin_attr = {
 	.size = PAGE_SIZE,
 };
 
+struct hfi1_port_attr {
+	struct attribute attr;
+	ssize_t	(*show)(struct hfi1_pportdata *, char *);
+	ssize_t	(*store)(struct hfi1_pportdata *, const char *, size_t);
+};
+
+static ssize_t cc_prescan_show(struct hfi1_pportdata *ppd, char *buf)
+{
+	return sprintf(buf, "%s\n", ppd->cc_prescan ? "on" : "off");
+}
+
+static ssize_t cc_prescan_store(struct hfi1_pportdata *ppd, const char *buf,
+				size_t count)
+{
+	if (!memcmp(buf, "on", 2))
+		ppd->cc_prescan = true;
+	else if (!memcmp(buf, "off", 3))
+		ppd->cc_prescan = false;
+
+	return count;
+}
+
+static struct hfi1_port_attr cc_prescan_attr =
+		__ATTR(cc_prescan, 0600, cc_prescan_show, cc_prescan_store);
+
+static ssize_t cc_attr_show(struct kobject *kobj, struct attribute *attr,
+			    char *buf)
+{
+	struct hfi1_port_attr *port_attr =
+		container_of(attr, struct hfi1_port_attr, attr);
+	struct hfi1_pportdata *ppd =
+		container_of(kobj, struct hfi1_pportdata, pport_cc_kobj);
+
+	return port_attr->show(ppd, buf);
+}
+
+static ssize_t cc_attr_store(struct kobject *kobj, struct attribute *attr,
+			     const char *buf, size_t count)
+{
+	struct hfi1_port_attr *port_attr =
+		container_of(attr, struct hfi1_port_attr, attr);
+	struct hfi1_pportdata *ppd =
+		container_of(kobj, struct hfi1_pportdata, pport_cc_kobj);
+
+	return port_attr->store(ppd, buf, count);
+}
+
+static const struct sysfs_ops port_cc_sysfs_ops = {
+	.show = cc_attr_show,
+	.store = cc_attr_store
+};
+
+static struct attribute *port_cc_default_attributes[] = {
+	&cc_prescan_attr.attr
+};
+
+static struct kobj_type port_cc_ktype = {
+	.release = port_release,
+	.sysfs_ops = &port_cc_sysfs_ops,
+	.default_attrs = port_cc_default_attributes
+};
+
 /* Start sc2vl */
 #define HFI1_SC2VL_ATTR(N)				    \
 	static struct hfi1_sc2vl_attr hfi1_sc2vl_attr_##N = { \
@@ -196,7 +250,6 @@ HFI1_SC2VL_ATTR(29);
 HFI1_SC2VL_ATTR(30);
 HFI1_SC2VL_ATTR(31);
 
-
 static struct attribute *sc2vl_default_attributes[] = {
 	&hfi1_sc2vl_attr_0.attr,
 	&hfi1_sc2vl_attr_1.attr,
@@ -302,7 +355,6 @@ HFI1_SL2SC_ATTR(29);
 HFI1_SL2SC_ATTR(30);
 HFI1_SL2SC_ATTR(31);
 
-
 static struct attribute *sl2sc_default_attributes[] = {
 	&hfi1_sl2sc_attr_0.attr,
 	&hfi1_sl2sc_attr_1.attr,
@@ -435,7 +487,6 @@ static struct kobj_type hfi1_vl2mtu_ktype = {
 	.default_attrs = vl2mtu_default_attributes
 };
 
-
 /* end of per-port file structures and support code */
 
 /*
@@ -446,7 +497,7 @@ static ssize_t show_rev(struct device *device, struct device_attribute *attr,
 			char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 
 	return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
 }
@@ -455,7 +506,7 @@ static ssize_t show_hfi(struct device *device, struct device_attribute *attr,
 			char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 	int ret;
 
@@ -470,19 +521,18 @@ static ssize_t show_boardversion(struct device *device,
 				 struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 
 	/* The string printed here is already newline-terminated. */
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->boardversion);
 }
 
-
 static ssize_t show_nctxts(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 
 	/*
@@ -497,10 +547,10 @@ static ssize_t show_nctxts(struct device *device,
 }
 
 static ssize_t show_nfreectxts(struct device *device,
-			   struct device_attribute *attr, char *buf)
+			       struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 
 	/* Return the number of free user ports (contexts) available. */
@@ -511,11 +561,10 @@ static ssize_t show_serial(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->serial);
-
 }
 
 static ssize_t store_chip_reset(struct device *device,
@@ -523,7 +572,7 @@ static ssize_t store_chip_reset(struct device *device,
 				size_t count)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 	int ret;
 
@@ -552,7 +601,7 @@ static ssize_t show_tempsense(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
-		container_of(device, struct hfi1_ibdev, ibdev.dev);
+		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 	struct hfi1_devdata *dd = dd_from_dev(dev);
 	struct hfi1_temp temp;
 	int ret;
@@ -608,8 +657,8 @@ int hfi1_create_port_files(struct ib_device *ibdev, u8 port_num,
 
 	if (!port_num || port_num > dd->num_pports) {
 		dd_dev_err(dd,
-			"Skipping infiniband class with invalid port %u\n",
-			port_num);
+			   "Skipping infiniband class with invalid port %u\n",
+			   port_num);
 		return -ENODEV;
 	}
 	ppd = &dd->pport[port_num - 1];
@@ -644,39 +693,36 @@ int hfi1_create_port_files(struct ib_device *ibdev, u8 port_num,
 	}
 	kobject_uevent(&ppd->vl2mtu_kobj, KOBJ_ADD);
 
-
 	ret = kobject_init_and_add(&ppd->pport_cc_kobj, &port_cc_ktype,
 				   kobj, "CCMgtA");
 	if (ret) {
 		dd_dev_err(dd,
-		 "Skipping Congestion Control sysfs info, (err %d) port %u\n",
-		 ret, port_num);
+			   "Skipping Congestion Control sysfs info, (err %d) port %u\n",
+			   ret, port_num);
 		goto bail_vl2mtu;
 	}
 
 	kobject_uevent(&ppd->pport_cc_kobj, KOBJ_ADD);
 
-	ret = sysfs_create_bin_file(&ppd->pport_cc_kobj,
-				&cc_setting_bin_attr);
+	ret = sysfs_create_bin_file(&ppd->pport_cc_kobj, &cc_setting_bin_attr);
 	if (ret) {
 		dd_dev_err(dd,
-		 "Skipping Congestion Control setting sysfs info, (err %d) port %u\n",
-		 ret, port_num);
+			   "Skipping Congestion Control setting sysfs info, (err %d) port %u\n",
+			   ret, port_num);
 		goto bail_cc;
 	}
 
-	ret = sysfs_create_bin_file(&ppd->pport_cc_kobj,
-				&cc_table_bin_attr);
+	ret = sysfs_create_bin_file(&ppd->pport_cc_kobj, &cc_table_bin_attr);
 	if (ret) {
 		dd_dev_err(dd,
-		 "Skipping Congestion Control table sysfs info, (err %d) port %u\n",
-		 ret, port_num);
+			   "Skipping Congestion Control table sysfs info, (err %d) port %u\n",
+			   ret, port_num);
 		goto bail_cc_entry_bin;
 	}
 
 	dd_dev_info(dd,
-		"IB%u: Congestion Control Agent enabled for port %d\n",
-		dd->unit, port_num);
+		    "IB%u: Congestion Control Agent enabled for port %d\n",
+		    dd->unit, port_num);
 
 	return 0;
 
@@ -700,7 +746,7 @@ bail:
  */
 int hfi1_verbs_register_sysfs(struct hfi1_devdata *dd)
 {
-	struct ib_device *dev = &dd->verbs_dev.ibdev;
+	struct ib_device *dev = &dd->verbs_dev.rdi.ibdev;
 	int i, ret;
 
 	for (i = 0; i < ARRAY_SIZE(hfi1_attributes); ++i) {
diff --git a/drivers/staging/rdma/hfi1/trace.c b/drivers/staging/rdma/hfi1/trace.c
index 10122e84cb2f..8b62fefcf903 100644
--- a/drivers/staging/rdma/hfi1/trace.c
+++ b/drivers/staging/rdma/hfi1/trace.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -109,17 +106,17 @@ const char *parse_everbs_hdrs(
 	case OP(RC, RDMA_WRITE_LAST_WITH_IMMEDIATE):
 	case OP(UC, RDMA_WRITE_LAST_WITH_IMMEDIATE):
 		trace_seq_printf(p, IMM_PRN,
-			be32_to_cpu(eh->imm_data));
+				 be32_to_cpu(eh->imm_data));
 		break;
 	/* reth + imm */
 	case OP(RC, RDMA_WRITE_ONLY_WITH_IMMEDIATE):
 	case OP(UC, RDMA_WRITE_ONLY_WITH_IMMEDIATE):
 		trace_seq_printf(p, RETH_PRN " " IMM_PRN,
-			(unsigned long long)ib_u64_get(
-				(__be32 *)&eh->rc.reth.vaddr),
-			be32_to_cpu(eh->rc.reth.rkey),
-			be32_to_cpu(eh->rc.reth.length),
-			be32_to_cpu(eh->rc.imm_data));
+				 (unsigned long long)ib_u64_get(
+				 (__be32 *)&eh->rc.reth.vaddr),
+				 be32_to_cpu(eh->rc.reth.rkey),
+				 be32_to_cpu(eh->rc.reth.length),
+				 be32_to_cpu(eh->rc.imm_data));
 		break;
 	/* reth */
 	case OP(RC, RDMA_READ_REQUEST):
@@ -128,10 +125,10 @@ const char *parse_everbs_hdrs(
 	case OP(RC, RDMA_WRITE_ONLY):
 	case OP(UC, RDMA_WRITE_ONLY):
 		trace_seq_printf(p, RETH_PRN,
-			(unsigned long long)ib_u64_get(
-				(__be32 *)&eh->rc.reth.vaddr),
-			be32_to_cpu(eh->rc.reth.rkey),
-			be32_to_cpu(eh->rc.reth.length));
+				 (unsigned long long)ib_u64_get(
+				 (__be32 *)&eh->rc.reth.vaddr),
+				 be32_to_cpu(eh->rc.reth.rkey),
+				 be32_to_cpu(eh->rc.reth.length));
 		break;
 	case OP(RC, RDMA_READ_RESPONSE_FIRST):
 	case OP(RC, RDMA_READ_RESPONSE_LAST):
@@ -154,19 +151,20 @@ const char *parse_everbs_hdrs(
 	case OP(RC, COMPARE_SWAP):
 	case OP(RC, FETCH_ADD):
 		trace_seq_printf(p, ATOMICETH_PRN,
-			(unsigned long long)ib_u64_get(eh->atomic_eth.vaddr),
-			eh->atomic_eth.rkey,
-			(unsigned long long)ib_u64_get(
-				(__be32 *)&eh->atomic_eth.swap_data),
-			(unsigned long long) ib_u64_get(
+				 (unsigned long long)ib_u64_get(
+				 eh->atomic_eth.vaddr),
+				 eh->atomic_eth.rkey,
+				 (unsigned long long)ib_u64_get(
+				 (__be32 *)&eh->atomic_eth.swap_data),
+				 (unsigned long long)ib_u64_get(
 				 (__be32 *)&eh->atomic_eth.compare_data));
 		break;
 	/* deth */
 	case OP(UD, SEND_ONLY):
 	case OP(UD, SEND_ONLY_WITH_IMMEDIATE):
 		trace_seq_printf(p, DETH_PRN,
-			be32_to_cpu(eh->ud.deth[0]),
-			be32_to_cpu(eh->ud.deth[1]) & HFI1_QPN_MASK);
+				 be32_to_cpu(eh->ud.deth[0]),
+				 be32_to_cpu(eh->ud.deth[1]) & RVT_QPN_MASK);
 		break;
 	}
 	trace_seq_putc(p, 0);
@@ -187,12 +185,12 @@ const char *parse_sdma_flags(
 	trace_seq_printf(p, "%s", flags);
 	if (desc0 & SDMA_DESC0_FIRST_DESC_FLAG)
 		trace_seq_printf(p, " amode:%u aidx:%u alen:%u",
-			(u8)((desc1 >> SDMA_DESC1_HEADER_MODE_SHIFT)
-				& SDMA_DESC1_HEADER_MODE_MASK),
-			(u8)((desc1 >> SDMA_DESC1_HEADER_INDEX_SHIFT)
-				& SDMA_DESC1_HEADER_INDEX_MASK),
-			(u8)((desc1 >> SDMA_DESC1_HEADER_DWS_SHIFT)
-				& SDMA_DESC1_HEADER_DWS_MASK));
+				 (u8)((desc1 >> SDMA_DESC1_HEADER_MODE_SHIFT) &
+				      SDMA_DESC1_HEADER_MODE_MASK),
+				 (u8)((desc1 >> SDMA_DESC1_HEADER_INDEX_SHIFT) &
+				      SDMA_DESC1_HEADER_INDEX_MASK),
+				 (u8)((desc1 >> SDMA_DESC1_HEADER_DWS_SHIFT) &
+				      SDMA_DESC1_HEADER_DWS_MASK));
 	return ret;
 }
 
@@ -234,3 +232,4 @@ __hfi1_trace_fn(DC8051);
 __hfi1_trace_fn(FIRMWARE);
 __hfi1_trace_fn(RCVCTRL);
 __hfi1_trace_fn(TID);
+__hfi1_trace_fn(MMU);
diff --git a/drivers/staging/rdma/hfi1/trace.h b/drivers/staging/rdma/hfi1/trace.h
index 86c12ebfd4f0..963dc948c38a 100644
--- a/drivers/staging/rdma/hfi1/trace.h
+++ b/drivers/staging/rdma/hfi1/trace.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -76,304 +73,295 @@ __print_symbolic(etype,                         \
 #define TRACE_SYSTEM hfi1_rx
 
 TRACE_EVENT(hfi1_rcvhdr,
-	TP_PROTO(struct hfi1_devdata *dd,
-		 u64 eflags,
-		 u32 ctxt,
-		 u32 etype,
-		 u32 hlen,
-		 u32 tlen,
-		 u32 updegr,
-		 u32 etail),
-	TP_ARGS(dd, ctxt, eflags, etype, hlen, tlen, updegr, etail),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		__field(u64, eflags)
-		__field(u32, ctxt)
-		__field(u32, etype)
-		__field(u32, hlen)
-		__field(u32, tlen)
-		__field(u32, updegr)
-		__field(u32, etail)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd);
-		__entry->eflags = eflags;
-		__entry->ctxt = ctxt;
-		__entry->etype = etype;
-		__entry->hlen = hlen;
-		__entry->tlen = tlen;
-		__entry->updegr = updegr;
-		__entry->etail = etail;
-	),
-	TP_printk(
-"[%s] ctxt %d eflags 0x%llx etype %d,%s hlen %d tlen %d updegr %d etail %d",
-		__get_str(dev),
-		__entry->ctxt,
-		__entry->eflags,
-		__entry->etype, show_packettype(__entry->etype),
-		__entry->hlen,
-		__entry->tlen,
-		__entry->updegr,
-		__entry->etail
-	)
+	    TP_PROTO(struct hfi1_devdata *dd,
+		     u64 eflags,
+		     u32 ctxt,
+		     u32 etype,
+		     u32 hlen,
+		     u32 tlen,
+		     u32 updegr,
+		     u32 etail
+		     ),
+	    TP_ARGS(dd, ctxt, eflags, etype, hlen, tlen, updegr, etail),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __field(u64, eflags)
+			     __field(u32, ctxt)
+			     __field(u32, etype)
+			     __field(u32, hlen)
+			     __field(u32, tlen)
+			     __field(u32, updegr)
+			     __field(u32, etail)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd);
+			   __entry->eflags = eflags;
+			   __entry->ctxt = ctxt;
+			   __entry->etype = etype;
+			   __entry->hlen = hlen;
+			   __entry->tlen = tlen;
+			   __entry->updegr = updegr;
+			   __entry->etail = etail;
+			   ),
+	    TP_printk(
+		      "[%s] ctxt %d eflags 0x%llx etype %d,%s hlen %d tlen %d updegr %d etail %d",
+		      __get_str(dev),
+		      __entry->ctxt,
+		      __entry->eflags,
+		      __entry->etype, show_packettype(__entry->etype),
+		      __entry->hlen,
+		      __entry->tlen,
+		      __entry->updegr,
+		      __entry->etail
+		      )
 );
 
 TRACE_EVENT(hfi1_receive_interrupt,
-	TP_PROTO(struct hfi1_devdata *dd, u32 ctxt),
-	TP_ARGS(dd, ctxt),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		__field(u32, ctxt)
-		__field(u8, slow_path)
-		__field(u8, dma_rtail)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd);
-		__entry->ctxt = ctxt;
-		if (dd->rcd[ctxt]->do_interrupt ==
-		    &handle_receive_interrupt) {
-			__entry->slow_path = 1;
-			__entry->dma_rtail = 0xFF;
-		} else if (dd->rcd[ctxt]->do_interrupt ==
-			&handle_receive_interrupt_dma_rtail){
-			__entry->dma_rtail = 1;
-			__entry->slow_path = 0;
-		} else if (dd->rcd[ctxt]->do_interrupt ==
-			 &handle_receive_interrupt_nodma_rtail) {
-			__entry->dma_rtail = 0;
-			__entry->slow_path = 0;
-		}
-	),
-	TP_printk(
-		"[%s] ctxt %d SlowPath: %d DmaRtail: %d",
-		__get_str(dev),
-		__entry->ctxt,
-		__entry->slow_path,
-		__entry->dma_rtail
-	)
+	    TP_PROTO(struct hfi1_devdata *dd, u32 ctxt),
+	    TP_ARGS(dd, ctxt),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __field(u32, ctxt)
+			     __field(u8, slow_path)
+			     __field(u8, dma_rtail)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd);
+			   __entry->ctxt = ctxt;
+			   if (dd->rcd[ctxt]->do_interrupt ==
+			       &handle_receive_interrupt) {
+				__entry->slow_path = 1;
+				__entry->dma_rtail = 0xFF;
+			   } else if (dd->rcd[ctxt]->do_interrupt ==
+				      &handle_receive_interrupt_dma_rtail){
+				__entry->dma_rtail = 1;
+				__entry->slow_path = 0;
+			   } else if (dd->rcd[ctxt]->do_interrupt ==
+				      &handle_receive_interrupt_nodma_rtail) {
+				__entry->dma_rtail = 0;
+				__entry->slow_path = 0;
+			   }
+			   ),
+	    TP_printk("[%s] ctxt %d SlowPath: %d DmaRtail: %d",
+		      __get_str(dev),
+		      __entry->ctxt,
+		      __entry->slow_path,
+		      __entry->dma_rtail
+		      )
 );
 
-const char *print_u64_array(struct trace_seq *, u64 *, int);
+TRACE_EVENT(hfi1_exp_tid_reg,
+	    TP_PROTO(unsigned ctxt, u16 subctxt, u32 rarr,
+		     u32 npages, unsigned long va, unsigned long pa,
+		     dma_addr_t dma),
+	    TP_ARGS(ctxt, subctxt, rarr, npages, va, pa, dma),
+	    TP_STRUCT__entry(
+		    __field(unsigned, ctxt)
+		    __field(u16, subctxt)
+		    __field(u32, rarr)
+		    __field(u32, npages)
+		    __field(unsigned long, va)
+		    __field(unsigned long, pa)
+		    __field(dma_addr_t, dma)
+		    ),
+	    TP_fast_assign(
+		    __entry->ctxt = ctxt;
+		    __entry->subctxt = subctxt;
+		    __entry->rarr = rarr;
+		    __entry->npages = npages;
+		    __entry->va = va;
+		    __entry->pa = pa;
+		    __entry->dma = dma;
+		    ),
+	    TP_printk("[%u:%u] entry:%u, %u pages @ 0x%lx, va:0x%lx dma:0x%llx",
+		      __entry->ctxt,
+		      __entry->subctxt,
+		      __entry->rarr,
+		      __entry->npages,
+		      __entry->pa,
+		      __entry->va,
+		      __entry->dma
+		    )
+	);
 
-TRACE_EVENT(hfi1_exp_tid_map,
-	    TP_PROTO(unsigned ctxt, u16 subctxt, int dir,
-		     unsigned long *maps, u16 count),
-	    TP_ARGS(ctxt, subctxt, dir, maps, count),
+TRACE_EVENT(hfi1_exp_tid_unreg,
+	    TP_PROTO(unsigned ctxt, u16 subctxt, u32 rarr, u32 npages,
+		     unsigned long va, unsigned long pa, dma_addr_t dma),
+	    TP_ARGS(ctxt, subctxt, rarr, npages, va, pa, dma),
 	    TP_STRUCT__entry(
 		    __field(unsigned, ctxt)
 		    __field(u16, subctxt)
-		    __field(int, dir)
-		    __field(u16, count)
-		    __dynamic_array(unsigned long, maps, sizeof(*maps) * count)
+		    __field(u32, rarr)
+		    __field(u32, npages)
+		    __field(unsigned long, va)
+		    __field(unsigned long, pa)
+		    __field(dma_addr_t, dma)
 		    ),
 	    TP_fast_assign(
 		    __entry->ctxt = ctxt;
 		    __entry->subctxt = subctxt;
-		    __entry->dir = dir;
-		    __entry->count = count;
-		    memcpy(__get_dynamic_array(maps), maps,
-			   sizeof(*maps) * count);
+		    __entry->rarr = rarr;
+		    __entry->npages = npages;
+		    __entry->va = va;
+		    __entry->pa = pa;
+		    __entry->dma = dma;
 		    ),
-	    TP_printk("[%3u:%02u] %s tidmaps %s",
+	    TP_printk("[%u:%u] entry:%u, %u pages @ 0x%lx, va:0x%lx dma:0x%llx",
 		      __entry->ctxt,
 		      __entry->subctxt,
-		      (__entry->dir ? ">" : "<"),
-		      print_u64_array(p, __get_dynamic_array(maps),
-				      __entry->count)
+		      __entry->rarr,
+		      __entry->npages,
+		      __entry->pa,
+		      __entry->va,
+		      __entry->dma
 		    )
 	);
 
-TRACE_EVENT(hfi1_exp_rcv_set,
-	    TP_PROTO(unsigned ctxt, u16 subctxt, u32 tid,
-		     unsigned long vaddr, u64 phys_addr, void *page),
-	    TP_ARGS(ctxt, subctxt, tid, vaddr, phys_addr, page),
+TRACE_EVENT(hfi1_exp_tid_inval,
+	    TP_PROTO(unsigned ctxt, u16 subctxt, unsigned long va, u32 rarr,
+		     u32 npages, dma_addr_t dma),
+	    TP_ARGS(ctxt, subctxt, va, rarr, npages, dma),
 	    TP_STRUCT__entry(
 		    __field(unsigned, ctxt)
 		    __field(u16, subctxt)
-		    __field(u32, tid)
-		    __field(unsigned long, vaddr)
-		    __field(u64, phys_addr)
-		    __field(void *, page)
+		    __field(unsigned long, va)
+		    __field(u32, rarr)
+		    __field(u32, npages)
+		    __field(dma_addr_t, dma)
 		    ),
 	    TP_fast_assign(
 		    __entry->ctxt = ctxt;
 		    __entry->subctxt = subctxt;
-		    __entry->tid = tid;
-		    __entry->vaddr = vaddr;
-		    __entry->phys_addr = phys_addr;
-		    __entry->page = page;
+		    __entry->va = va;
+		    __entry->rarr = rarr;
+		    __entry->npages = npages;
+		    __entry->dma = dma;
 		    ),
-	    TP_printk("[%u:%u] TID %u, vaddrs 0x%lx, physaddr 0x%llx, pgp %p",
+	    TP_printk("[%u:%u] entry:%u, %u pages @ 0x%lx dma: 0x%llx",
 		      __entry->ctxt,
 		      __entry->subctxt,
-		      __entry->tid,
-		      __entry->vaddr,
-		      __entry->phys_addr,
-		      __entry->page
+		      __entry->rarr,
+		      __entry->npages,
+		      __entry->va,
+		      __entry->dma
 		    )
 	);
 
-TRACE_EVENT(hfi1_exp_rcv_free,
-	    TP_PROTO(unsigned ctxt, u16 subctxt, u32 tid,
-		     unsigned long phys, void *page),
-	    TP_ARGS(ctxt, subctxt, tid, phys, page),
+TRACE_EVENT(hfi1_mmu_invalidate,
+	    TP_PROTO(unsigned ctxt, u16 subctxt, const char *type,
+		     unsigned long start, unsigned long end),
+	    TP_ARGS(ctxt, subctxt, type, start, end),
 	    TP_STRUCT__entry(
 		    __field(unsigned, ctxt)
 		    __field(u16, subctxt)
-		    __field(u32, tid)
-		    __field(unsigned long, phys)
-		    __field(void *, page)
+		    __string(type, type)
+		    __field(unsigned long, start)
+		    __field(unsigned long, end)
 		    ),
 	    TP_fast_assign(
 		    __entry->ctxt = ctxt;
 		    __entry->subctxt = subctxt;
-		    __entry->tid = tid;
-		    __entry->phys = phys;
-		    __entry->page = page;
+		    __assign_str(type, type);
+		    __entry->start = start;
+		    __entry->end = end;
 		    ),
-	    TP_printk("[%u:%u] freeing TID %u, 0x%lx, pgp %p",
+	    TP_printk("[%3u:%02u] MMU Invalidate (%s) 0x%lx - 0x%lx",
 		      __entry->ctxt,
 		      __entry->subctxt,
-		      __entry->tid,
-		      __entry->phys,
-		      __entry->page
+		      __get_str(type),
+		      __entry->start,
+		      __entry->end
 		    )
 	);
+
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_tx
 
 TRACE_EVENT(hfi1_piofree,
-	TP_PROTO(struct send_context *sc, int extra),
-	TP_ARGS(sc, extra),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sc->dd)
-		__field(u32, sw_index)
-		__field(u32, hw_context)
-		__field(int, extra)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sc->dd);
-		__entry->sw_index = sc->sw_index;
-		__entry->hw_context = sc->hw_context;
-		__entry->extra = extra;
-	),
-	TP_printk(
-		"[%s] ctxt %u(%u) extra %d",
-		__get_str(dev),
-		__entry->sw_index,
-		__entry->hw_context,
-		__entry->extra
-	)
+	    TP_PROTO(struct send_context *sc, int extra),
+	    TP_ARGS(sc, extra),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(sc->dd)
+			     __field(u32, sw_index)
+			     __field(u32, hw_context)
+			     __field(int, extra)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(sc->dd);
+			   __entry->sw_index = sc->sw_index;
+			   __entry->hw_context = sc->hw_context;
+			   __entry->extra = extra;
+			   ),
+	    TP_printk("[%s] ctxt %u(%u) extra %d",
+		      __get_str(dev),
+		      __entry->sw_index,
+		      __entry->hw_context,
+		      __entry->extra
+		      )
 );
 
 TRACE_EVENT(hfi1_wantpiointr,
-	TP_PROTO(struct send_context *sc, u32 needint, u64 credit_ctrl),
-	TP_ARGS(sc, needint, credit_ctrl),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sc->dd)
-		__field(u32, sw_index)
-		__field(u32, hw_context)
-		__field(u32, needint)
-		__field(u64, credit_ctrl)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sc->dd);
-		__entry->sw_index = sc->sw_index;
-		__entry->hw_context = sc->hw_context;
-		__entry->needint = needint;
-		__entry->credit_ctrl = credit_ctrl;
-	),
-	TP_printk(
-		"[%s] ctxt %u(%u) on %d credit_ctrl 0x%llx",
-		__get_str(dev),
-		__entry->sw_index,
-		__entry->hw_context,
-		__entry->needint,
-		(unsigned long long)__entry->credit_ctrl
-	)
+	    TP_PROTO(struct send_context *sc, u32 needint, u64 credit_ctrl),
+	    TP_ARGS(sc, needint, credit_ctrl),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(sc->dd)
+			     __field(u32, sw_index)
+			     __field(u32, hw_context)
+			     __field(u32, needint)
+			     __field(u64, credit_ctrl)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(sc->dd);
+			   __entry->sw_index = sc->sw_index;
+			   __entry->hw_context = sc->hw_context;
+			   __entry->needint = needint;
+			   __entry->credit_ctrl = credit_ctrl;
+			   ),
+	    TP_printk("[%s] ctxt %u(%u) on %d credit_ctrl 0x%llx",
+		      __get_str(dev),
+		      __entry->sw_index,
+		      __entry->hw_context,
+		      __entry->needint,
+		      (unsigned long long)__entry->credit_ctrl
+		       )
 );
 
 DECLARE_EVENT_CLASS(hfi1_qpsleepwakeup_template,
-	TP_PROTO(struct hfi1_qp *qp, u32 flags),
-	TP_ARGS(qp, flags),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
-		__field(u32, qpn)
-		__field(u32, flags)
-		__field(u32, s_flags)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd_from_ibdev(qp->ibqp.device))
-		__entry->flags = flags;
-		__entry->qpn = qp->ibqp.qp_num;
-		__entry->s_flags = qp->s_flags;
-	),
-	TP_printk(
-		"[%s] qpn 0x%x flags 0x%x s_flags 0x%x",
-		__get_str(dev),
-		__entry->qpn,
-		__entry->flags,
-		__entry->s_flags
-	)
+		    TP_PROTO(struct rvt_qp *qp, u32 flags),
+		    TP_ARGS(qp, flags),
+		    TP_STRUCT__entry(
+			    DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
+			    __field(u32, qpn)
+			    __field(u32, flags)
+			    __field(u32, s_flags)
+			    ),
+		    TP_fast_assign(
+			    DD_DEV_ASSIGN(dd_from_ibdev(qp->ibqp.device))
+			    __entry->flags = flags;
+			    __entry->qpn = qp->ibqp.qp_num;
+			    __entry->s_flags = qp->s_flags;
+			    ),
+		    TP_printk(
+			    "[%s] qpn 0x%x flags 0x%x s_flags 0x%x",
+			    __get_str(dev),
+			    __entry->qpn,
+			    __entry->flags,
+			    __entry->s_flags
+			    )
 );
 
 DEFINE_EVENT(hfi1_qpsleepwakeup_template, hfi1_qpwakeup,
-	     TP_PROTO(struct hfi1_qp *qp, u32 flags),
+	     TP_PROTO(struct rvt_qp *qp, u32 flags),
 	     TP_ARGS(qp, flags));
 
 DEFINE_EVENT(hfi1_qpsleepwakeup_template, hfi1_qpsleep,
-	     TP_PROTO(struct hfi1_qp *qp, u32 flags),
+	     TP_PROTO(struct rvt_qp *qp, u32 flags),
 	     TP_ARGS(qp, flags));
 
 #undef TRACE_SYSTEM
-#define TRACE_SYSTEM hfi1_qphash
-DECLARE_EVENT_CLASS(hfi1_qphash_template,
-	TP_PROTO(struct hfi1_qp *qp, u32 bucket),
-	TP_ARGS(qp, bucket),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
-		__field(u32, qpn)
-		__field(u32, bucket)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd_from_ibdev(qp->ibqp.device))
-		__entry->qpn = qp->ibqp.qp_num;
-		__entry->bucket = bucket;
-	),
-	TP_printk(
-		"[%s] qpn 0x%x bucket %u",
-		__get_str(dev),
-		__entry->qpn,
-		__entry->bucket
-	)
-);
-
-DEFINE_EVENT(hfi1_qphash_template, hfi1_qpinsert,
-	TP_PROTO(struct hfi1_qp *qp, u32 bucket),
-	TP_ARGS(qp, bucket));
-
-DEFINE_EVENT(hfi1_qphash_template, hfi1_qpremove,
-	TP_PROTO(struct hfi1_qp *qp, u32 bucket),
-	TP_ARGS(qp, bucket));
-
-#undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_ibhdrs
 
 u8 ibhdr_exhdr_len(struct hfi1_ib_header *hdr);
-const char *parse_everbs_hdrs(
-	struct trace_seq *p,
-	u8 opcode,
-	void *ehdrs);
+const char *parse_everbs_hdrs(struct trace_seq *p, u8 opcode, void *ehdrs);
 
 #define __parse_ib_ehdrs(op, ehdrs) parse_everbs_hdrs(p, op, ehdrs)
 
-const char *parse_sdma_flags(
-	struct trace_seq *p,
-	u64 desc0, u64 desc1);
+const char *parse_sdma_flags(struct trace_seq *p, u64 desc0, u64 desc1);
 
 #define __parse_sdma_flags(desc0, desc1) parse_sdma_flags(p, desc0, desc1)
 
-
 #define lrh_name(lrh) { HFI1_##lrh, #lrh }
 #define show_lnh(lrh)                    \
 __print_symbolic(lrh,                    \
@@ -420,7 +408,6 @@ __print_symbolic(opcode,                                   \
 	ib_opcode_name(UD_SEND_ONLY_WITH_IMMEDIATE),       \
 	ib_opcode_name(CNP))
 
-
 #define LRH_PRN "vl %d lver %d sl %d lnh %d,%s dlid %.4x len %d slid %.4x"
 #define BTH_PRN \
 	"op 0x%.2x,%s se %d m %d pad %d tver %d pkey 0x%.4x " \
@@ -428,124 +415,130 @@ __print_symbolic(opcode,                                   \
 #define EHDR_PRN "%s"
 
 DECLARE_EVENT_CLASS(hfi1_ibhdr_template,
-	TP_PROTO(struct hfi1_devdata *dd,
-		 struct hfi1_ib_header *hdr),
-	TP_ARGS(dd, hdr),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		/* LRH */
-		__field(u8, vl)
-		__field(u8, lver)
-		__field(u8, sl)
-		__field(u8, lnh)
-		__field(u16, dlid)
-		__field(u16, len)
-		__field(u16, slid)
-		/* BTH */
-		__field(u8, opcode)
-		__field(u8, se)
-		__field(u8, m)
-		__field(u8, pad)
-		__field(u8, tver)
-		__field(u16, pkey)
-		__field(u8, f)
-		__field(u8, b)
-		__field(u32, qpn)
-		__field(u8, a)
-		__field(u32, psn)
-		/* extended headers */
-		__dynamic_array(u8, ehdrs, ibhdr_exhdr_len(hdr))
-	),
-	TP_fast_assign(
-		struct hfi1_other_headers *ohdr;
-
-		DD_DEV_ASSIGN(dd);
-		/* LRH */
-		__entry->vl =
-			(u8)(be16_to_cpu(hdr->lrh[0]) >> 12);
-		__entry->lver =
-			(u8)(be16_to_cpu(hdr->lrh[0]) >> 8) & 0xf;
-		__entry->sl =
-			(u8)(be16_to_cpu(hdr->lrh[0]) >> 4) & 0xf;
-		__entry->lnh =
-			(u8)(be16_to_cpu(hdr->lrh[0]) & 3);
-		__entry->dlid =
-			be16_to_cpu(hdr->lrh[1]);
-		/* allow for larger len */
-		__entry->len =
-			be16_to_cpu(hdr->lrh[2]);
-		__entry->slid =
-			be16_to_cpu(hdr->lrh[3]);
-		/* BTH */
-		if (__entry->lnh == HFI1_LRH_BTH)
-			ohdr = &hdr->u.oth;
-		else
-			ohdr = &hdr->u.l.oth;
-		__entry->opcode =
-			(be32_to_cpu(ohdr->bth[0]) >> 24) & 0xff;
-		__entry->se =
-			(be32_to_cpu(ohdr->bth[0]) >> 23) & 1;
-		__entry->m =
-			 (be32_to_cpu(ohdr->bth[0]) >> 22) & 1;
-		__entry->pad =
-			(be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
-		__entry->tver =
-			(be32_to_cpu(ohdr->bth[0]) >> 16) & 0xf;
-		__entry->pkey =
-			be32_to_cpu(ohdr->bth[0]) & 0xffff;
-		__entry->f =
-			(be32_to_cpu(ohdr->bth[1]) >> HFI1_FECN_SHIFT)
-			& HFI1_FECN_MASK;
-		__entry->b =
-			(be32_to_cpu(ohdr->bth[1]) >> HFI1_BECN_SHIFT)
-			& HFI1_BECN_MASK;
-		__entry->qpn =
-			be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
-		__entry->a =
-			(be32_to_cpu(ohdr->bth[2]) >> 31) & 1;
-		/* allow for larger PSN */
-		__entry->psn =
-			be32_to_cpu(ohdr->bth[2]) & 0x7fffffff;
-		/* extended headers */
-		 memcpy(
-			__get_dynamic_array(ehdrs),
-			&ohdr->u,
-			ibhdr_exhdr_len(hdr));
-	),
-	TP_printk("[%s] " LRH_PRN " " BTH_PRN " " EHDR_PRN,
-		__get_str(dev),
-		/* LRH */
-		__entry->vl,
-		__entry->lver,
-		__entry->sl,
-		__entry->lnh, show_lnh(__entry->lnh),
-		__entry->dlid,
-		__entry->len,
-		__entry->slid,
-		/* BTH */
-		__entry->opcode, show_ib_opcode(__entry->opcode),
-		__entry->se,
-		__entry->m,
-		__entry->pad,
-		__entry->tver,
-		__entry->pkey,
-		__entry->f,
-		__entry->b,
-		__entry->qpn,
-		__entry->a,
-		__entry->psn,
-		/* extended headers */
-		__parse_ib_ehdrs(
-			__entry->opcode,
-			(void *)__get_dynamic_array(ehdrs))
-	)
+		    TP_PROTO(struct hfi1_devdata *dd,
+			     struct hfi1_ib_header *hdr),
+		    TP_ARGS(dd, hdr),
+		    TP_STRUCT__entry(
+			    DD_DEV_ENTRY(dd)
+			    /* LRH */
+			    __field(u8, vl)
+			    __field(u8, lver)
+			    __field(u8, sl)
+			    __field(u8, lnh)
+			    __field(u16, dlid)
+			    __field(u16, len)
+			    __field(u16, slid)
+			    /* BTH */
+			    __field(u8, opcode)
+			    __field(u8, se)
+			    __field(u8, m)
+			    __field(u8, pad)
+			    __field(u8, tver)
+			    __field(u16, pkey)
+			    __field(u8, f)
+			    __field(u8, b)
+			    __field(u32, qpn)
+			    __field(u8, a)
+			    __field(u32, psn)
+			    /* extended headers */
+			    __dynamic_array(u8, ehdrs, ibhdr_exhdr_len(hdr))
+			    ),
+		    TP_fast_assign(
+			   struct hfi1_other_headers *ohdr;
+
+			   DD_DEV_ASSIGN(dd);
+			   /* LRH */
+			   __entry->vl =
+			   (u8)(be16_to_cpu(hdr->lrh[0]) >> 12);
+			   __entry->lver =
+			   (u8)(be16_to_cpu(hdr->lrh[0]) >> 8) & 0xf;
+			   __entry->sl =
+			   (u8)(be16_to_cpu(hdr->lrh[0]) >> 4) & 0xf;
+			   __entry->lnh =
+			   (u8)(be16_to_cpu(hdr->lrh[0]) & 3);
+			   __entry->dlid =
+			   be16_to_cpu(hdr->lrh[1]);
+			   /* allow for larger len */
+			   __entry->len =
+			   be16_to_cpu(hdr->lrh[2]);
+			   __entry->slid =
+			   be16_to_cpu(hdr->lrh[3]);
+			   /* BTH */
+			   if (__entry->lnh == HFI1_LRH_BTH)
+				ohdr = &hdr->u.oth;
+			   else
+				ohdr = &hdr->u.l.oth;
+			  __entry->opcode =
+			  (be32_to_cpu(ohdr->bth[0]) >> 24) & 0xff;
+			  __entry->se =
+			  (be32_to_cpu(ohdr->bth[0]) >> 23) & 1;
+			  __entry->m =
+			  (be32_to_cpu(ohdr->bth[0]) >> 22) & 1;
+			  __entry->pad =
+			  (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+			  __entry->tver =
+			  (be32_to_cpu(ohdr->bth[0]) >> 16) & 0xf;
+			  __entry->pkey =
+			  be32_to_cpu(ohdr->bth[0]) & 0xffff;
+			  __entry->f =
+			  (be32_to_cpu(ohdr->bth[1]) >> HFI1_FECN_SHIFT) &
+			  HFI1_FECN_MASK;
+			  __entry->b =
+			  (be32_to_cpu(ohdr->bth[1]) >> HFI1_BECN_SHIFT) &
+			  HFI1_BECN_MASK;
+			  __entry->qpn =
+			  be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
+			  __entry->a =
+			  (be32_to_cpu(ohdr->bth[2]) >> 31) & 1;
+			  /* allow for larger PSN */
+			  __entry->psn =
+			  be32_to_cpu(ohdr->bth[2]) & 0x7fffffff;
+			  /* extended headers */
+			  memcpy(__get_dynamic_array(ehdrs), &ohdr->u,
+				 ibhdr_exhdr_len(hdr));
+			 ),
+		    TP_printk("[%s] " LRH_PRN " " BTH_PRN " " EHDR_PRN,
+			      __get_str(dev),
+			      /* LRH */
+			      __entry->vl,
+			      __entry->lver,
+			      __entry->sl,
+			      __entry->lnh, show_lnh(__entry->lnh),
+			      __entry->dlid,
+			      __entry->len,
+			      __entry->slid,
+			      /* BTH */
+			      __entry->opcode, show_ib_opcode(__entry->opcode),
+			      __entry->se,
+			      __entry->m,
+			      __entry->pad,
+			      __entry->tver,
+			      __entry->pkey,
+			      __entry->f,
+			      __entry->b,
+			      __entry->qpn,
+			      __entry->a,
+			      __entry->psn,
+			      /* extended headers */
+			      __parse_ib_ehdrs(
+					__entry->opcode,
+					(void *)__get_dynamic_array(ehdrs))
+			     )
 );
 
 DEFINE_EVENT(hfi1_ibhdr_template, input_ibhdr,
 	     TP_PROTO(struct hfi1_devdata *dd, struct hfi1_ib_header *hdr),
 	     TP_ARGS(dd, hdr));
 
-DEFINE_EVENT(hfi1_ibhdr_template, output_ibhdr,
+DEFINE_EVENT(hfi1_ibhdr_template, pio_output_ibhdr,
+	     TP_PROTO(struct hfi1_devdata *dd, struct hfi1_ib_header *hdr),
+	     TP_ARGS(dd, hdr));
+
+DEFINE_EVENT(hfi1_ibhdr_template, ack_output_ibhdr,
+	     TP_PROTO(struct hfi1_devdata *dd, struct hfi1_ib_header *hdr),
+	     TP_ARGS(dd, hdr));
+
+DEFINE_EVENT(hfi1_ibhdr_template, sdma_output_ibhdr,
 	     TP_PROTO(struct hfi1_devdata *dd, struct hfi1_ib_header *hdr),
 	     TP_ARGS(dd, hdr));
 
@@ -556,15 +549,14 @@ DEFINE_EVENT(hfi1_ibhdr_template, output_ibhdr,
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_snoop
 
-
 TRACE_EVENT(snoop_capture,
-	TP_PROTO(struct hfi1_devdata *dd,
-		 int hdr_len,
-		 struct hfi1_ib_header *hdr,
-		 int data_len,
-		 void *data),
-	TP_ARGS(dd, hdr_len, hdr, data_len, data),
-	TP_STRUCT__entry(
+	    TP_PROTO(struct hfi1_devdata *dd,
+		     int hdr_len,
+		     struct hfi1_ib_header *hdr,
+		     int data_len,
+		     void *data),
+	    TP_ARGS(dd, hdr_len, hdr, data_len, data),
+	    TP_STRUCT__entry(
 		DD_DEV_ENTRY(dd)
 		__field(u16, slid)
 		__field(u16, dlid)
@@ -577,8 +569,8 @@ TRACE_EVENT(snoop_capture,
 		__field(u8, lnh)
 		__dynamic_array(u8, raw_hdr, hdr_len)
 		__dynamic_array(u8, raw_pkt, data_len)
-	),
-	TP_fast_assign(
+		),
+	    TP_fast_assign(
 		struct hfi1_other_headers *ohdr;
 
 		__entry->lnh = (u8)(be16_to_cpu(hdr->lrh[0]) & 3);
@@ -589,7 +581,7 @@ TRACE_EVENT(snoop_capture,
 		DD_DEV_ASSIGN(dd);
 		__entry->slid = be16_to_cpu(hdr->lrh[3]);
 		__entry->dlid = be16_to_cpu(hdr->lrh[1]);
-		__entry->qpn = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
+		__entry->qpn = be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
 		__entry->opcode = (be32_to_cpu(ohdr->bth[0]) >> 24) & 0xff;
 		__entry->sl = (u8)(be16_to_cpu(hdr->lrh[0]) >> 4) & 0xf;
 		__entry->pkey =	be32_to_cpu(ohdr->bth[0]) & 0xffff;
@@ -597,8 +589,9 @@ TRACE_EVENT(snoop_capture,
 		__entry->data_len = data_len;
 		memcpy(__get_dynamic_array(raw_hdr), hdr, hdr_len);
 		memcpy(__get_dynamic_array(raw_pkt), data, data_len);
-	),
-	TP_printk("[%s] " SNOOP_PRN,
+		),
+	    TP_printk(
+		"[%s] " SNOOP_PRN,
 		__get_str(dev),
 		__entry->slid,
 		__entry->dlid,
@@ -609,7 +602,7 @@ TRACE_EVENT(snoop_capture,
 		__entry->pkey,
 		__entry->hdr_len,
 		__entry->data_len
-	)
+		)
 );
 
 #undef TRACE_SYSTEM
@@ -621,41 +614,39 @@ TRACE_EVENT(snoop_capture,
 TRACE_EVENT(hfi1_uctxtdata,
 	    TP_PROTO(struct hfi1_devdata *dd, struct hfi1_ctxtdata *uctxt),
 	    TP_ARGS(dd, uctxt),
-	    TP_STRUCT__entry(
-		    DD_DEV_ENTRY(dd)
-		    __field(unsigned, ctxt)
-		    __field(u32, credits)
-		    __field(u64, hw_free)
-		    __field(u64, piobase)
-		    __field(u16, rcvhdrq_cnt)
-		    __field(u64, rcvhdrq_phys)
-		    __field(u32, eager_cnt)
-		    __field(u64, rcvegr_phys)
-		    ),
-	    TP_fast_assign(
-		    DD_DEV_ASSIGN(dd);
-		    __entry->ctxt = uctxt->ctxt;
-		    __entry->credits = uctxt->sc->credits;
-		    __entry->hw_free = (u64)uctxt->sc->hw_free;
-		    __entry->piobase = (u64)uctxt->sc->base_addr;
-		    __entry->rcvhdrq_cnt = uctxt->rcvhdrq_cnt;
-		    __entry->rcvhdrq_phys = uctxt->rcvhdrq_phys;
-		    __entry->eager_cnt = uctxt->egrbufs.alloced;
-		    __entry->rcvegr_phys = uctxt->egrbufs.rcvtids[0].phys;
-		    ),
-	    TP_printk(
-		    "[%s] ctxt %u " UCTXT_FMT,
-		    __get_str(dev),
-		    __entry->ctxt,
-		    __entry->credits,
-		    __entry->hw_free,
-		    __entry->piobase,
-		    __entry->rcvhdrq_cnt,
-		    __entry->rcvhdrq_phys,
-		    __entry->eager_cnt,
-		    __entry->rcvegr_phys
-		    )
-	);
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __field(unsigned, ctxt)
+			     __field(u32, credits)
+			     __field(u64, hw_free)
+			     __field(u64, piobase)
+			     __field(u16, rcvhdrq_cnt)
+			     __field(u64, rcvhdrq_phys)
+			     __field(u32, eager_cnt)
+			     __field(u64, rcvegr_phys)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd);
+			   __entry->ctxt = uctxt->ctxt;
+			   __entry->credits = uctxt->sc->credits;
+			   __entry->hw_free = (u64)uctxt->sc->hw_free;
+			   __entry->piobase = (u64)uctxt->sc->base_addr;
+			   __entry->rcvhdrq_cnt = uctxt->rcvhdrq_cnt;
+			   __entry->rcvhdrq_phys = uctxt->rcvhdrq_phys;
+			   __entry->eager_cnt = uctxt->egrbufs.alloced;
+			   __entry->rcvegr_phys =
+			   uctxt->egrbufs.rcvtids[0].phys;
+			   ),
+	    TP_printk("[%s] ctxt %u " UCTXT_FMT,
+		      __get_str(dev),
+		      __entry->ctxt,
+		      __entry->credits,
+		      __entry->hw_free,
+		      __entry->piobase,
+		      __entry->rcvhdrq_cnt,
+		      __entry->rcvhdrq_phys,
+		      __entry->eager_cnt,
+		      __entry->rcvegr_phys
+		      )
+);
 
 #define CINFO_FMT \
 	"egrtids:%u, egr_size:%u, hdrq_cnt:%u, hdrq_size:%u, sdma_ring_size:%u"
@@ -663,38 +654,35 @@ TRACE_EVENT(hfi1_ctxt_info,
 	    TP_PROTO(struct hfi1_devdata *dd, unsigned ctxt, unsigned subctxt,
 		     struct hfi1_ctxt_info cinfo),
 	    TP_ARGS(dd, ctxt, subctxt, cinfo),
-	    TP_STRUCT__entry(
-		    DD_DEV_ENTRY(dd)
-		    __field(unsigned, ctxt)
-		    __field(unsigned, subctxt)
-		    __field(u16, egrtids)
-		    __field(u16, rcvhdrq_cnt)
-		    __field(u16, rcvhdrq_size)
-		    __field(u16, sdma_ring_size)
-		    __field(u32, rcvegr_size)
-		    ),
-	    TP_fast_assign(
-		    DD_DEV_ASSIGN(dd);
-		    __entry->ctxt = ctxt;
-		    __entry->subctxt = subctxt;
-		    __entry->egrtids = cinfo.egrtids;
-		    __entry->rcvhdrq_cnt = cinfo.rcvhdrq_cnt;
-		    __entry->rcvhdrq_size = cinfo.rcvhdrq_entsize;
-		    __entry->sdma_ring_size = cinfo.sdma_ring_size;
-		    __entry->rcvegr_size = cinfo.rcvegr_size;
-		    ),
-	    TP_printk(
-		    "[%s] ctxt %u:%u " CINFO_FMT,
-		    __get_str(dev),
-		    __entry->ctxt,
-		    __entry->subctxt,
-		    __entry->egrtids,
-		    __entry->rcvegr_size,
-		    __entry->rcvhdrq_cnt,
-		    __entry->rcvhdrq_size,
-		    __entry->sdma_ring_size
-		    )
-	);
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __field(unsigned, ctxt)
+			     __field(unsigned, subctxt)
+			     __field(u16, egrtids)
+			     __field(u16, rcvhdrq_cnt)
+			     __field(u16, rcvhdrq_size)
+			     __field(u16, sdma_ring_size)
+			     __field(u32, rcvegr_size)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd);
+			    __entry->ctxt = ctxt;
+			    __entry->subctxt = subctxt;
+			    __entry->egrtids = cinfo.egrtids;
+			    __entry->rcvhdrq_cnt = cinfo.rcvhdrq_cnt;
+			    __entry->rcvhdrq_size = cinfo.rcvhdrq_entsize;
+			    __entry->sdma_ring_size = cinfo.sdma_ring_size;
+			    __entry->rcvegr_size = cinfo.rcvegr_size;
+			    ),
+	    TP_printk("[%s] ctxt %u:%u " CINFO_FMT,
+		      __get_str(dev),
+		      __entry->ctxt,
+		      __entry->subctxt,
+		      __entry->egrtids,
+		      __entry->rcvegr_size,
+		      __entry->rcvhdrq_cnt,
+		      __entry->rcvhdrq_size,
+		      __entry->sdma_ring_size
+		      )
+);
 
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_sma
@@ -708,52 +696,48 @@ TRACE_EVENT(hfi1_ctxt_info,
 	)
 
 DECLARE_EVENT_CLASS(hfi1_bct_template,
-	TP_PROTO(struct hfi1_devdata *dd, struct buffer_control *bc),
-	TP_ARGS(dd, bc),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		__dynamic_array(u8, bct, sizeof(*bc))
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd);
-		memcpy(
-			__get_dynamic_array(bct),
-			bc,
-			sizeof(*bc));
-	),
-	TP_printk(BCT_FORMAT,
-		BCT(overall_shared_limit),
-
-		BCT(vl[0].dedicated),
-		BCT(vl[0].shared),
-
-		BCT(vl[1].dedicated),
-		BCT(vl[1].shared),
-
-		BCT(vl[2].dedicated),
-		BCT(vl[2].shared),
-
-		BCT(vl[3].dedicated),
-		BCT(vl[3].shared),
-
-		BCT(vl[4].dedicated),
-		BCT(vl[4].shared),
-
-		BCT(vl[5].dedicated),
-		BCT(vl[5].shared),
-
-		BCT(vl[6].dedicated),
-		BCT(vl[6].shared),
-
-		BCT(vl[7].dedicated),
-		BCT(vl[7].shared),
-
-		BCT(vl[15].dedicated),
-		BCT(vl[15].shared)
-	)
+		    TP_PROTO(struct hfi1_devdata *dd,
+			     struct buffer_control *bc),
+		    TP_ARGS(dd, bc),
+		    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+				     __dynamic_array(u8, bct, sizeof(*bc))
+				     ),
+		    TP_fast_assign(DD_DEV_ASSIGN(dd);
+				   memcpy(__get_dynamic_array(bct), bc,
+					  sizeof(*bc));
+				   ),
+		    TP_printk(BCT_FORMAT,
+			      BCT(overall_shared_limit),
+
+			      BCT(vl[0].dedicated),
+			      BCT(vl[0].shared),
+
+			      BCT(vl[1].dedicated),
+			      BCT(vl[1].shared),
+
+			      BCT(vl[2].dedicated),
+			      BCT(vl[2].shared),
+
+			      BCT(vl[3].dedicated),
+			      BCT(vl[3].shared),
+
+			      BCT(vl[4].dedicated),
+			      BCT(vl[4].shared),
+
+			      BCT(vl[5].dedicated),
+			      BCT(vl[5].shared),
+
+			      BCT(vl[6].dedicated),
+			      BCT(vl[6].shared),
+
+			      BCT(vl[7].dedicated),
+			      BCT(vl[7].shared),
+
+			      BCT(vl[15].dedicated),
+			      BCT(vl[15].shared)
+			      )
 );
 
-
 DEFINE_EVENT(hfi1_bct_template, bct_set,
 	     TP_PROTO(struct hfi1_devdata *dd, struct buffer_control *bc),
 	     TP_ARGS(dd, bc));
@@ -766,252 +750,209 @@ DEFINE_EVENT(hfi1_bct_template, bct_get,
 #define TRACE_SYSTEM hfi1_sdma
 
 TRACE_EVENT(hfi1_sdma_descriptor,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u64 desc0,
-		u64 desc1,
-		u16 e,
-		void *descp),
+	    TP_PROTO(struct sdma_engine *sde,
+		     u64 desc0,
+		     u64 desc1,
+		     u16 e,
+		     void *descp),
 	TP_ARGS(sde, desc0, desc1, e, descp),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(void *, descp)
-		__field(u64, desc0)
-		__field(u64, desc1)
-		__field(u16, e)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->desc0 = desc0;
-		__entry->desc1 = desc1;
-		__entry->idx = sde->this_idx;
-		__entry->descp = descp;
-		__entry->e = e;
-	),
+	TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+			 __field(void *, descp)
+			 __field(u64, desc0)
+			 __field(u64, desc1)
+			 __field(u16, e)
+			 __field(u8, idx)
+			 ),
+	TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+		       __entry->desc0 = desc0;
+		       __entry->desc1 = desc1;
+		       __entry->idx = sde->this_idx;
+		       __entry->descp = descp;
+		       __entry->e = e;
+		       ),
 	TP_printk(
-		"[%s] SDE(%u) flags:%s addr:0x%016llx gen:%u len:%u d0:%016llx d1:%016llx to %p,%u",
-		__get_str(dev),
-		__entry->idx,
-		__parse_sdma_flags(__entry->desc0, __entry->desc1),
-		(__entry->desc0 >> SDMA_DESC0_PHY_ADDR_SHIFT)
-			& SDMA_DESC0_PHY_ADDR_MASK,
-		(u8)((__entry->desc1 >> SDMA_DESC1_GENERATION_SHIFT)
-			& SDMA_DESC1_GENERATION_MASK),
-		(u16)((__entry->desc0 >> SDMA_DESC0_BYTE_COUNT_SHIFT)
-			& SDMA_DESC0_BYTE_COUNT_MASK),
-		__entry->desc0,
-		__entry->desc1,
-		__entry->descp,
-		__entry->e
-	)
+		  "[%s] SDE(%u) flags:%s addr:0x%016llx gen:%u len:%u d0:%016llx d1:%016llx to %p,%u",
+		  __get_str(dev),
+		  __entry->idx,
+		  __parse_sdma_flags(__entry->desc0, __entry->desc1),
+		  (__entry->desc0 >> SDMA_DESC0_PHY_ADDR_SHIFT) &
+		  SDMA_DESC0_PHY_ADDR_MASK,
+		  (u8)((__entry->desc1 >> SDMA_DESC1_GENERATION_SHIFT) &
+		       SDMA_DESC1_GENERATION_MASK),
+		  (u16)((__entry->desc0 >> SDMA_DESC0_BYTE_COUNT_SHIFT) &
+			SDMA_DESC0_BYTE_COUNT_MASK),
+		  __entry->desc0,
+		  __entry->desc1,
+		  __entry->descp,
+		  __entry->e
+		  )
 );
 
 TRACE_EVENT(hfi1_sdma_engine_select,
-	TP_PROTO(struct hfi1_devdata *dd, u32 sel, u8 vl, u8 idx),
-	TP_ARGS(dd, sel, vl, idx),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		__field(u32, sel)
-		__field(u8, vl)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd);
-		__entry->sel = sel;
-		__entry->vl = vl;
-		__entry->idx = idx;
-	),
-	TP_printk(
-		"[%s] selecting SDE %u sel 0x%x vl %u",
-		__get_str(dev),
-		__entry->idx,
-		__entry->sel,
-		__entry->vl
-	)
+	    TP_PROTO(struct hfi1_devdata *dd, u32 sel, u8 vl, u8 idx),
+	    TP_ARGS(dd, sel, vl, idx),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __field(u32, sel)
+			     __field(u8, vl)
+			     __field(u8, idx)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd);
+			   __entry->sel = sel;
+			   __entry->vl = vl;
+			   __entry->idx = idx;
+			   ),
+	    TP_printk("[%s] selecting SDE %u sel 0x%x vl %u",
+		      __get_str(dev),
+		      __entry->idx,
+		      __entry->sel,
+		      __entry->vl
+		      )
 );
 
 DECLARE_EVENT_CLASS(hfi1_sdma_engine_class,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u64 status
-	),
-	TP_ARGS(sde, status),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(u64, status)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->status = status;
-		__entry->idx = sde->this_idx;
-	),
-	TP_printk(
-		"[%s] SDE(%u) status %llx",
-		__get_str(dev),
-		__entry->idx,
-		(unsigned long long)__entry->status
-	)
+		    TP_PROTO(struct sdma_engine *sde, u64 status),
+		    TP_ARGS(sde, status),
+		    TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+				     __field(u64, status)
+				     __field(u8, idx)
+				     ),
+		    TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+				   __entry->status = status;
+				   __entry->idx = sde->this_idx;
+				   ),
+		    TP_printk("[%s] SDE(%u) status %llx",
+			      __get_str(dev),
+			      __entry->idx,
+			      (unsigned long long)__entry->status
+			      )
 );
 
 DEFINE_EVENT(hfi1_sdma_engine_class, hfi1_sdma_engine_interrupt,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u64 status
-	),
-	TP_ARGS(sde, status)
+	     TP_PROTO(struct sdma_engine *sde, u64 status),
+	     TP_ARGS(sde, status)
 );
 
 DEFINE_EVENT(hfi1_sdma_engine_class, hfi1_sdma_engine_progress,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u64 status
-	),
-	TP_ARGS(sde, status)
+	     TP_PROTO(struct sdma_engine *sde, u64 status),
+	     TP_ARGS(sde, status)
 );
 
 DECLARE_EVENT_CLASS(hfi1_sdma_ahg_ad,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		int aidx
-	),
-	TP_ARGS(sde, aidx),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(int, aidx)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->idx = sde->this_idx;
-		__entry->aidx = aidx;
-	),
-	TP_printk(
-		"[%s] SDE(%u) aidx %d",
-		__get_str(dev),
-		__entry->idx,
-		__entry->aidx
-	)
+		    TP_PROTO(struct sdma_engine *sde, int aidx),
+		    TP_ARGS(sde, aidx),
+		    TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+				     __field(int, aidx)
+				     __field(u8, idx)
+				     ),
+		    TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+				   __entry->idx = sde->this_idx;
+				   __entry->aidx = aidx;
+				   ),
+		    TP_printk("[%s] SDE(%u) aidx %d",
+			      __get_str(dev),
+			      __entry->idx,
+			      __entry->aidx
+			      )
 );
 
 DEFINE_EVENT(hfi1_sdma_ahg_ad, hfi1_ahg_allocate,
-	     TP_PROTO(
-		struct sdma_engine *sde,
-		int aidx
-	     ),
+	     TP_PROTO(struct sdma_engine *sde, int aidx),
 	     TP_ARGS(sde, aidx));
 
 DEFINE_EVENT(hfi1_sdma_ahg_ad, hfi1_ahg_deallocate,
-	     TP_PROTO(
-		struct sdma_engine *sde,
-		int aidx
-	     ),
+	     TP_PROTO(struct sdma_engine *sde, int aidx),
 	     TP_ARGS(sde, aidx));
 
 #ifdef CONFIG_HFI1_DEBUG_SDMA_ORDER
 TRACE_EVENT(hfi1_sdma_progress,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u16 hwhead,
-		u16 swhead,
-		struct sdma_txreq *txp
-	),
-	TP_ARGS(sde, hwhead, swhead, txp),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(u64, sn)
-		__field(u16, hwhead)
-		__field(u16, swhead)
-		__field(u16, txnext)
-		__field(u16, tx_tail)
-		__field(u16, tx_head)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->hwhead = hwhead;
-		__entry->swhead = swhead;
-		__entry->tx_tail = sde->tx_tail;
-		__entry->tx_head = sde->tx_head;
-		__entry->txnext = txp ? txp->next_descq_idx : ~0;
-		__entry->idx = sde->this_idx;
-		__entry->sn = txp ? txp->sn : ~0;
-	),
-	TP_printk(
-		"[%s] SDE(%u) sn %llu hwhead %u swhead %u next_descq_idx %u tx_head %u tx_tail %u",
-		__get_str(dev),
-		__entry->idx,
-		__entry->sn,
-		__entry->hwhead,
-		__entry->swhead,
-		__entry->txnext,
-		__entry->tx_head,
-		__entry->tx_tail
-	)
+	    TP_PROTO(struct sdma_engine *sde,
+		     u16 hwhead,
+		     u16 swhead,
+		     struct sdma_txreq *txp
+		     ),
+	    TP_ARGS(sde, hwhead, swhead, txp),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+			     __field(u64, sn)
+			     __field(u16, hwhead)
+			     __field(u16, swhead)
+			     __field(u16, txnext)
+			     __field(u16, tx_tail)
+			     __field(u16, tx_head)
+			     __field(u8, idx)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+			   __entry->hwhead = hwhead;
+			   __entry->swhead = swhead;
+			   __entry->tx_tail = sde->tx_tail;
+			   __entry->tx_head = sde->tx_head;
+			   __entry->txnext = txp ? txp->next_descq_idx : ~0;
+			   __entry->idx = sde->this_idx;
+			   __entry->sn = txp ? txp->sn : ~0;
+			   ),
+	    TP_printk(
+		      "[%s] SDE(%u) sn %llu hwhead %u swhead %u next_descq_idx %u tx_head %u tx_tail %u",
+		      __get_str(dev),
+		      __entry->idx,
+		      __entry->sn,
+		      __entry->hwhead,
+		      __entry->swhead,
+		      __entry->txnext,
+		      __entry->tx_head,
+		      __entry->tx_tail
+		      )
 );
 #else
 TRACE_EVENT(hfi1_sdma_progress,
-	    TP_PROTO(
-		struct sdma_engine *sde,
-		u16 hwhead,
-		u16 swhead,
-		struct sdma_txreq *txp
+	    TP_PROTO(struct sdma_engine *sde,
+		     u16 hwhead, u16 swhead,
+		     struct sdma_txreq *txp
 	    ),
 	TP_ARGS(sde, hwhead, swhead, txp),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(u16, hwhead)
-		__field(u16, swhead)
-		__field(u16, txnext)
-		__field(u16, tx_tail)
-		__field(u16, tx_head)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->hwhead = hwhead;
-		__entry->swhead = swhead;
-		__entry->tx_tail = sde->tx_tail;
-		__entry->tx_head = sde->tx_head;
-		__entry->txnext = txp ? txp->next_descq_idx : ~0;
-		__entry->idx = sde->this_idx;
-	),
+	TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+			 __field(u16, hwhead)
+			 __field(u16, swhead)
+			 __field(u16, txnext)
+			 __field(u16, tx_tail)
+			 __field(u16, tx_head)
+			 __field(u8, idx)
+			 ),
+	TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+		       __entry->hwhead = hwhead;
+		       __entry->swhead = swhead;
+		       __entry->tx_tail = sde->tx_tail;
+		       __entry->tx_head = sde->tx_head;
+		       __entry->txnext = txp ? txp->next_descq_idx : ~0;
+		       __entry->idx = sde->this_idx;
+		       ),
 	TP_printk(
-		"[%s] SDE(%u) hwhead %u swhead %u next_descq_idx %u tx_head %u tx_tail %u",
-		__get_str(dev),
-		__entry->idx,
-		__entry->hwhead,
-		__entry->swhead,
-		__entry->txnext,
-		__entry->tx_head,
-		__entry->tx_tail
-	)
+		  "[%s] SDE(%u) hwhead %u swhead %u next_descq_idx %u tx_head %u tx_tail %u",
+		  __get_str(dev),
+		  __entry->idx,
+		  __entry->hwhead,
+		  __entry->swhead,
+		  __entry->txnext,
+		  __entry->tx_head,
+		  __entry->tx_tail
+		  )
 );
 #endif
 
 DECLARE_EVENT_CLASS(hfi1_sdma_sn,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		u64 sn
-	),
-	TP_ARGS(sde, sn),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__field(u64, sn)
-		__field(u8, idx)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__entry->sn = sn;
-		__entry->idx = sde->this_idx;
-	),
-	TP_printk(
-		"[%s] SDE(%u) sn %llu",
-		__get_str(dev),
-		__entry->idx,
-		__entry->sn
-	)
+		    TP_PROTO(struct sdma_engine *sde, u64 sn),
+		    TP_ARGS(sde, sn),
+		    TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+				     __field(u64, sn)
+				     __field(u8, idx)
+				     ),
+		    TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+				   __entry->sn = sn;
+				   __entry->idx = sde->this_idx;
+				   ),
+		    TP_printk("[%s] SDE(%u) sn %llu",
+			      __get_str(dev),
+			      __entry->idx,
+			      __entry->sn
+			      )
 );
 
 DEFINE_EVENT(hfi1_sdma_sn, hfi1_sdma_out_sn,
@@ -1023,10 +964,7 @@ DEFINE_EVENT(hfi1_sdma_sn, hfi1_sdma_out_sn,
 );
 
 DEFINE_EVENT(hfi1_sdma_sn, hfi1_sdma_in_sn,
-	     TP_PROTO(
-		struct sdma_engine *sde,
-		u64 sn
-	     ),
+	     TP_PROTO(struct sdma_engine *sde, u64 sn),
 	     TP_ARGS(sde, sn)
 );
 
@@ -1227,88 +1165,85 @@ TRACE_EVENT(hfi1_sdma_user_header_ahg,
 	);
 
 TRACE_EVENT(hfi1_sdma_state,
-	TP_PROTO(
-		struct sdma_engine *sde,
-		const char *cstate,
-		const char *nstate
-	),
-	TP_ARGS(sde, cstate, nstate),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(sde->dd)
-		__string(curstate, cstate)
-		__string(newstate, nstate)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(sde->dd);
-		__assign_str(curstate, cstate);
-		__assign_str(newstate, nstate);
-	),
+	    TP_PROTO(struct sdma_engine *sde,
+		     const char *cstate,
+		     const char *nstate
+		     ),
+	    TP_ARGS(sde, cstate, nstate),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(sde->dd)
+			     __string(curstate, cstate)
+			     __string(newstate, nstate)
+			     ),
+	TP_fast_assign(DD_DEV_ASSIGN(sde->dd);
+		       __assign_str(curstate, cstate);
+		       __assign_str(newstate, nstate);
+		       ),
 	TP_printk("[%s] current state %s new state %s",
-		__get_str(dev),
-		__get_str(curstate),
-		__get_str(newstate)
-	)
+		  __get_str(dev),
+		  __get_str(curstate),
+		  __get_str(newstate)
+		  )
 );
 
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM hfi1_rc
 
 DECLARE_EVENT_CLASS(hfi1_rc_template,
-	TP_PROTO(struct hfi1_qp *qp, u32 psn),
-	TP_ARGS(qp, psn),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
-		__field(u32, qpn)
-		__field(u32, s_flags)
-		__field(u32, psn)
-		__field(u32, s_psn)
-		__field(u32, s_next_psn)
-		__field(u32, s_sending_psn)
-		__field(u32, s_sending_hpsn)
-		__field(u32, r_psn)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd_from_ibdev(qp->ibqp.device))
-		__entry->qpn = qp->ibqp.qp_num;
-		__entry->s_flags = qp->s_flags;
-		__entry->psn = psn;
-		__entry->s_psn = qp->s_psn;
-		__entry->s_next_psn = qp->s_next_psn;
-		__entry->s_sending_psn = qp->s_sending_psn;
-		__entry->s_sending_hpsn = qp->s_sending_hpsn;
-		__entry->r_psn = qp->r_psn;
-	),
-	TP_printk(
-		"[%s] qpn 0x%x s_flags 0x%x psn 0x%x s_psn 0x%x s_next_psn 0x%x s_sending_psn 0x%x sending_hpsn 0x%x r_psn 0x%x",
-		__get_str(dev),
-		__entry->qpn,
-		__entry->s_flags,
-		__entry->psn,
-		__entry->s_psn,
-		__entry->s_next_psn,
-		__entry->s_sending_psn,
-		__entry->s_sending_hpsn,
-		__entry->r_psn
-	)
+		    TP_PROTO(struct rvt_qp *qp, u32 psn),
+		    TP_ARGS(qp, psn),
+		    TP_STRUCT__entry(
+			DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
+			__field(u32, qpn)
+			__field(u32, s_flags)
+			__field(u32, psn)
+			__field(u32, s_psn)
+			__field(u32, s_next_psn)
+			__field(u32, s_sending_psn)
+			__field(u32, s_sending_hpsn)
+			__field(u32, r_psn)
+			),
+		    TP_fast_assign(
+			DD_DEV_ASSIGN(dd_from_ibdev(qp->ibqp.device))
+			__entry->qpn = qp->ibqp.qp_num;
+			__entry->s_flags = qp->s_flags;
+			__entry->psn = psn;
+			__entry->s_psn = qp->s_psn;
+			__entry->s_next_psn = qp->s_next_psn;
+			__entry->s_sending_psn = qp->s_sending_psn;
+			__entry->s_sending_hpsn = qp->s_sending_hpsn;
+			__entry->r_psn = qp->r_psn;
+			),
+		    TP_printk(
+			"[%s] qpn 0x%x s_flags 0x%x psn 0x%x s_psn 0x%x s_next_psn 0x%x s_sending_psn 0x%x sending_hpsn 0x%x r_psn 0x%x",
+			__get_str(dev),
+			__entry->qpn,
+			__entry->s_flags,
+			__entry->psn,
+			__entry->s_psn,
+			__entry->s_next_psn,
+			__entry->s_sending_psn,
+			__entry->s_sending_hpsn,
+			__entry->r_psn
+			)
 );
 
 DEFINE_EVENT(hfi1_rc_template, hfi1_rc_sendcomplete,
-	     TP_PROTO(struct hfi1_qp *qp, u32 psn),
+	     TP_PROTO(struct rvt_qp *qp, u32 psn),
 	     TP_ARGS(qp, psn)
 );
 
 DEFINE_EVENT(hfi1_rc_template, hfi1_rc_ack,
-	     TP_PROTO(struct hfi1_qp *qp, u32 psn),
+	     TP_PROTO(struct rvt_qp *qp, u32 psn),
 	     TP_ARGS(qp, psn)
 );
 
 DEFINE_EVENT(hfi1_rc_template, hfi1_rc_timeout,
-	     TP_PROTO(struct hfi1_qp *qp, u32 psn),
+	     TP_PROTO(struct rvt_qp *qp, u32 psn),
 	     TP_ARGS(qp, psn)
 );
 
 DEFINE_EVENT(hfi1_rc_template, hfi1_rc_rcv_error,
-	     TP_PROTO(struct hfi1_qp *qp, u32 psn),
+	     TP_PROTO(struct rvt_qp *qp, u32 psn),
 	     TP_ARGS(qp, psn)
 );
 
@@ -1316,21 +1251,20 @@ DEFINE_EVENT(hfi1_rc_template, hfi1_rc_rcv_error,
 #define TRACE_SYSTEM hfi1_misc
 
 TRACE_EVENT(hfi1_interrupt,
-	TP_PROTO(struct hfi1_devdata *dd, const struct is_table *is_entry,
-		 int src),
-	TP_ARGS(dd, is_entry, src),
-	TP_STRUCT__entry(
-		DD_DEV_ENTRY(dd)
-		__array(char, buf, 64)
-		__field(int, src)
-	),
-	TP_fast_assign(
-		DD_DEV_ASSIGN(dd)
-		is_entry->is_name(__entry->buf, 64, src - is_entry->start);
-		__entry->src = src;
-	),
-	TP_printk("[%s] source: %s [%d]", __get_str(dev), __entry->buf,
-		  __entry->src)
+	    TP_PROTO(struct hfi1_devdata *dd, const struct is_table *is_entry,
+		     int src),
+	    TP_ARGS(dd, is_entry, src),
+	    TP_STRUCT__entry(DD_DEV_ENTRY(dd)
+			     __array(char, buf, 64)
+			     __field(int, src)
+			     ),
+	    TP_fast_assign(DD_DEV_ASSIGN(dd)
+			   is_entry->is_name(__entry->buf, 64,
+					     src - is_entry->start);
+			   __entry->src = src;
+			   ),
+	    TP_printk("[%s] source: %s [%d]", __get_str(dev), __entry->buf,
+		      __entry->src)
 );
 
 /*
@@ -1345,21 +1279,21 @@ TRACE_EVENT(hfi1_interrupt,
 #define MAX_MSG_LEN 512
 
 DECLARE_EVENT_CLASS(hfi1_trace_template,
-	TP_PROTO(const char *function, struct va_format *vaf),
-	TP_ARGS(function, vaf),
-	TP_STRUCT__entry(
-		__string(function, function)
-		__dynamic_array(char, msg, MAX_MSG_LEN)
-	),
-	TP_fast_assign(
-		__assign_str(function, function);
-		WARN_ON_ONCE(vsnprintf(__get_dynamic_array(msg),
-		     MAX_MSG_LEN, vaf->fmt,
-		     *vaf->va) >= MAX_MSG_LEN);
-	),
-	TP_printk("(%s) %s",
-		  __get_str(function),
-		  __get_str(msg))
+		    TP_PROTO(const char *function, struct va_format *vaf),
+		    TP_ARGS(function, vaf),
+		    TP_STRUCT__entry(__string(function, function)
+				     __dynamic_array(char, msg, MAX_MSG_LEN)
+				     ),
+		    TP_fast_assign(__assign_str(function, function);
+				   WARN_ON_ONCE(vsnprintf
+						(__get_dynamic_array(msg),
+						 MAX_MSG_LEN, vaf->fmt,
+						 *vaf->va) >=
+						MAX_MSG_LEN);
+				   ),
+		    TP_printk("(%s) %s",
+			      __get_str(function),
+			      __get_str(msg))
 );
 
 /*
@@ -1406,6 +1340,7 @@ __hfi1_trace_def(DC8051);
 __hfi1_trace_def(FIRMWARE);
 __hfi1_trace_def(RCVCTRL);
 __hfi1_trace_def(TID);
+__hfi1_trace_def(MMU);
 
 #define hfi1_cdbg(which, fmt, ...) \
 	__hfi1_trace_##which(__func__, fmt, ##__VA_ARGS__)
diff --git a/drivers/staging/rdma/hfi1/twsi.c b/drivers/staging/rdma/hfi1/twsi.c
index ea54fd2700ad..e82e52a63d35 100644
--- a/drivers/staging/rdma/hfi1/twsi.c
+++ b/drivers/staging/rdma/hfi1/twsi.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -119,9 +116,9 @@ static void scl_out(struct hfi1_devdata *dd, u32 target, u8 bit)
 	 * Allow for slow slaves by simple
 	 * delay for falling edge, sampling on rise.
 	 */
-	if (!bit)
+	if (!bit) {
 		udelay(2);
-	else {
+	} else {
 		int rise_usec;
 
 		for (rise_usec = SCL_WAIT_USEC; rise_usec > 0; rise_usec -= 2) {
@@ -131,11 +128,24 @@ static void scl_out(struct hfi1_devdata *dd, u32 target, u8 bit)
 		}
 		if (rise_usec <= 0)
 			dd_dev_err(dd, "SCL interface stuck low > %d uSec\n",
-				    SCL_WAIT_USEC);
+				   SCL_WAIT_USEC);
 	}
 	i2c_wait_for_writes(dd, target);
 }
 
+static u8 scl_in(struct hfi1_devdata *dd, u32 target, int wait)
+{
+	u32 read_val, mask;
+
+	mask = QSFP_HFI0_I2CCLK;
+	/* SCL is meant to be bare-drain, so never set "OUT", just DIR */
+	hfi1_gpio_mod(dd, target, 0, 0, mask);
+	read_val = hfi1_gpio_mod(dd, target, 0, 0, 0);
+	if (wait)
+		i2c_wait_for_writes(dd, target);
+	return (read_val & mask) >> GPIO_SCL_NUM;
+}
+
 static void sda_out(struct hfi1_devdata *dd, u32 target, u8 bit)
 {
 	u32 mask;
@@ -274,13 +284,12 @@ static void stop_cmd(struct hfi1_devdata *dd, u32 target)
 /**
  * hfi1_twsi_reset - reset I2C communication
  * @dd: the hfi1_ib device
+ * returns 0 if ok, -EIO on error
  */
-
 int hfi1_twsi_reset(struct hfi1_devdata *dd, u32 target)
 {
 	int clock_cycles_left = 9;
-	int was_high = 0;
-	u32 pins, mask;
+	u32 mask;
 
 	/* Both SCL and SDA should be high. If not, there
 	 * is something wrong.
@@ -294,43 +303,23 @@ int hfi1_twsi_reset(struct hfi1_devdata *dd, u32 target)
 	 */
 	hfi1_gpio_mod(dd, target, 0, 0, mask);
 
-	/*
-	 * Clock nine times to get all listeners into a sane state.
-	 * If SDA does not go high at any point, we are wedged.
-	 * One vendor recommends then issuing START followed by STOP.
-	 * we cannot use our "normal" functions to do that, because
-	 * if SCL drops between them, another vendor's part will
-	 * wedge, dropping SDA and keeping it low forever, at the end of
-	 * the next transaction (even if it was not the device addressed).
-	 * So our START and STOP take place with SCL held high.
+	/* Check if SCL is low, if it is low then we have a slave device
+	 * misbehaving and there is not much we can do.
+	 */
+	if (!scl_in(dd, target, 0))
+		return -EIO;
+
+	/* Check if SDA is low, if it is low then we have to clock SDA
+	 * up to 9 times for the device to release the bus
 	 */
 	while (clock_cycles_left--) {
+		if (sda_in(dd, target, 0))
+			return 0;
 		scl_out(dd, target, 0);
 		scl_out(dd, target, 1);
-		/* Note if SDA is high, but keep clocking to sync slave */
-		was_high |= sda_in(dd, target, 0);
-	}
-
-	if (was_high) {
-		/*
-		 * We saw a high, which we hope means the slave is sync'd.
-		 * Issue START, STOP, pause for T_BUF.
-		 */
-
-		pins = hfi1_gpio_mod(dd, target, 0, 0, 0);
-		if ((pins & mask) != mask)
-			dd_dev_err(dd, "GPIO pins not at rest: %d\n",
-				    pins & mask);
-		/* Drop SDA to issue START */
-		udelay(1); /* Guarantee .6 uSec setup */
-		sda_out(dd, target, 0);
-		udelay(1); /* Guarantee .6 uSec hold */
-		/* At this point, SCL is high, SDA low. Raise SDA for STOP */
-		sda_out(dd, target, 1);
-		udelay(TWSI_BUF_WAIT_USEC);
 	}
 
-	return !was_high;
+	return -EIO;
 }
 
 #define HFI1_TWSI_START 0x100
@@ -365,17 +354,25 @@ static int twsi_wr(struct hfi1_devdata *dd, u32 target, int data, int flags)
  * HFI1_TWSI_NO_DEV and does the correct operation for the legacy part,
  * which responded to all TWSI device codes, interpreting them as
  * address within device. On all other devices found on board handled by
- * this driver, the device is followed by a one-byte "address" which selects
+ * this driver, the device is followed by a N-byte "address" which selects
  * the "register" or "offset" within the device from which data should
  * be read.
  */
 int hfi1_twsi_blk_rd(struct hfi1_devdata *dd, u32 target, int dev, int addr,
 		     void *buffer, int len)
 {
-	int ret;
 	u8 *bp = buffer;
+	int ret = 1;
+	int i;
+	int offset_size;
+
+	/* obtain the offset size, strip it from the device address */
+	offset_size = (dev >> 8) & 0xff;
+	dev &= 0xff;
 
-	ret = 1;
+	/* allow at most a 2 byte offset */
+	if (offset_size > 2)
+		goto bail;
 
 	if (dev == HFI1_TWSI_NO_DEV) {
 		/* legacy not-really-I2C */
@@ -383,34 +380,29 @@ int hfi1_twsi_blk_rd(struct hfi1_devdata *dd, u32 target, int dev, int addr,
 		ret = twsi_wr(dd, target, addr, HFI1_TWSI_START);
 	} else {
 		/* Actual I2C */
-		ret = twsi_wr(dd, target, dev | WRITE_CMD, HFI1_TWSI_START);
-		if (ret) {
-			stop_cmd(dd, target);
-			ret = 1;
-			goto bail;
-		}
-		/*
-		 * SFF spec claims we do _not_ stop after the addr
-		 * but simply issue a start with the "read" dev-addr.
-		 * Since we are implicitly waiting for ACK here,
-		 * we need t_buf (nominally 20uSec) before that start,
-		 * and cannot rely on the delay built in to the STOP
-		 */
-		ret = twsi_wr(dd, target, addr, 0);
-		udelay(TWSI_BUF_WAIT_USEC);
+		if (offset_size) {
+			ret = twsi_wr(dd, target,
+				      dev | WRITE_CMD, HFI1_TWSI_START);
+			if (ret) {
+				stop_cmd(dd, target);
+				goto bail;
+			}
 
-		if (ret) {
-			dd_dev_err(dd,
-				"Failed to write interface read addr %02X\n",
-				addr);
-			ret = 1;
-			goto bail;
+			for (i = 0; i < offset_size; i++) {
+				ret = twsi_wr(dd, target,
+					      (addr >> (i * 8)) & 0xff, 0);
+				udelay(TWSI_BUF_WAIT_USEC);
+				if (ret) {
+					dd_dev_err(dd, "Failed to write byte %d of offset 0x%04X\n",
+						   i, addr);
+					goto bail;
+				}
+			}
 		}
 		ret = twsi_wr(dd, target, dev | READ_CMD, HFI1_TWSI_START);
 	}
 	if (ret) {
 		stop_cmd(dd, target);
-		ret = 1;
 		goto bail;
 	}
 
@@ -442,76 +434,55 @@ bail:
  * HFI1_TWSI_NO_DEV and does the correct operation for the legacy part,
  * which responded to all TWSI device codes, interpreting them as
  * address within device. On all other devices found on board handled by
- * this driver, the device is followed by a one-byte "address" which selects
+ * this driver, the device is followed by a N-byte "address" which selects
  * the "register" or "offset" within the device to which data should
  * be written.
  */
 int hfi1_twsi_blk_wr(struct hfi1_devdata *dd, u32 target, int dev, int addr,
 		     const void *buffer, int len)
 {
-	int sub_len;
 	const u8 *bp = buffer;
-	int max_wait_time, i;
 	int ret = 1;
+	int i;
+	int offset_size;
 
-	while (len > 0) {
-		if (dev == HFI1_TWSI_NO_DEV) {
-			if (twsi_wr(dd, target, (addr << 1) | WRITE_CMD,
-				    HFI1_TWSI_START)) {
-				goto failed_write;
-			}
-		} else {
-			/* Real I2C */
-			if (twsi_wr(dd, target,
-				    dev | WRITE_CMD, HFI1_TWSI_START))
-				goto failed_write;
-			ret = twsi_wr(dd, target, addr, 0);
-			if (ret) {
-				dd_dev_err(dd,
-					"Failed to write interface write addr %02X\n",
-					addr);
-				goto failed_write;
-			}
-		}
-
-		sub_len = min(len, 4);
-		addr += sub_len;
-		len -= sub_len;
+	/* obtain the offset size, strip it from the device address */
+	offset_size = (dev >> 8) & 0xff;
+	dev &= 0xff;
 
-		for (i = 0; i < sub_len; i++)
-			if (twsi_wr(dd, target, *bp++, 0))
-				goto failed_write;
+	/* allow at most a 2 byte offset */
+	if (offset_size > 2)
+		goto bail;
 
-		stop_cmd(dd, target);
+	if (dev == HFI1_TWSI_NO_DEV) {
+		if (twsi_wr(dd, target, (addr << 1) | WRITE_CMD,
+			    HFI1_TWSI_START)) {
+			goto failed_write;
+		}
+	} else {
+		/* Real I2C */
+		if (twsi_wr(dd, target, dev | WRITE_CMD, HFI1_TWSI_START))
+			goto failed_write;
+	}
 
-		/*
-		 * Wait for write complete by waiting for a successful
-		 * read (the chip replies with a zero after the write
-		 * cmd completes, and before it writes to the eeprom.
-		 * The startcmd for the read will fail the ack until
-		 * the writes have completed.   We do this inline to avoid
-		 * the debug prints that are in the real read routine
-		 * if the startcmd fails.
-		 * We also use the proper device address, so it doesn't matter
-		 * whether we have real eeprom_dev. Legacy likes any address.
-		 */
-		max_wait_time = 100;
-		while (twsi_wr(dd, target,
-			       dev | READ_CMD, HFI1_TWSI_START)) {
-			stop_cmd(dd, target);
-			if (!--max_wait_time)
-				goto failed_write;
+	for (i = 0; i < offset_size; i++) {
+		ret = twsi_wr(dd, target, (addr >> (i * 8)) & 0xff, 0);
+		udelay(TWSI_BUF_WAIT_USEC);
+		if (ret) {
+			dd_dev_err(dd, "Failed to write byte %d of offset 0x%04X\n",
+				   i, addr);
+			goto bail;
 		}
-		/* now read (and ignore) the resulting byte */
-		rd_byte(dd, target, 1);
 	}
 
+	for (i = 0; i < len; i++)
+		if (twsi_wr(dd, target, *bp++, 0))
+			goto failed_write;
+
 	ret = 0;
-	goto bail;
 
 failed_write:
 	stop_cmd(dd, target);
-	ret = 1;
 
 bail:
 	return ret;
diff --git a/drivers/staging/rdma/hfi1/twsi.h b/drivers/staging/rdma/hfi1/twsi.h
index 5907e029613d..5b8a5b5e7eae 100644
--- a/drivers/staging/rdma/hfi1/twsi.h
+++ b/drivers/staging/rdma/hfi1/twsi.h
@@ -1,14 +1,13 @@
 #ifndef _TWSI_H
 #define _TWSI_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -54,8 +51,9 @@
 
 struct hfi1_devdata;
 
-/* Bit position of SDA pin in ASIC_QSFP* registers  */
+/* Bit position of SDA/SCL pins in ASIC_QSFP* registers  */
 #define  GPIO_SDA_NUM 1
+#define  GPIO_SCL_NUM 0
 
 /* these functions must be called with qsfp_lock held */
 int hfi1_twsi_reset(struct hfi1_devdata *dd, u32 target);
@@ -64,5 +62,4 @@ int hfi1_twsi_blk_rd(struct hfi1_devdata *dd, u32 target, int dev, int addr,
 int hfi1_twsi_blk_wr(struct hfi1_devdata *dd, u32 target, int dev, int addr,
 		     const void *buffer, int len);
 
-
 #endif /* _TWSI_H */
diff --git a/drivers/staging/rdma/hfi1/uc.c b/drivers/staging/rdma/hfi1/uc.c
index 4f2a7889a852..df773d433297 100644
--- a/drivers/staging/rdma/hfi1/uc.c
+++ b/drivers/staging/rdma/hfi1/uc.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -49,71 +46,82 @@
  */
 
 #include "hfi.h"
-#include "sdma.h"
+#include "verbs_txreq.h"
 #include "qp.h"
 
 /* cut down ridiculously long IB macro names */
 #define OP(x) IB_OPCODE_UC_##x
 
+/* only opcode mask for adaptive pio */
+const u32 uc_only_opcode =
+	BIT(OP(SEND_ONLY) & 0x1f) |
+	BIT(OP(SEND_ONLY_WITH_IMMEDIATE & 0x1f)) |
+	BIT(OP(RDMA_WRITE_ONLY & 0x1f)) |
+	BIT(OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE & 0x1f));
+
 /**
  * hfi1_make_uc_req - construct a request packet (SEND, RDMA write)
  * @qp: a pointer to the QP
  *
+ * Assume s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int hfi1_make_uc_req(struct hfi1_qp *qp)
+int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_other_headers *ohdr;
-	struct hfi1_swqe *wqe;
-	unsigned long flags;
+	struct rvt_swqe *wqe;
 	u32 hwords = 5;
 	u32 bth0 = 0;
 	u32 len;
 	u32 pmtu = qp->pmtu;
-	int ret = 0;
 	int middle = 0;
 
-	spin_lock_irqsave(&qp->s_lock, flags);
+	ps->s_txreq = get_txreq(ps->dev, qp);
+	if (IS_ERR(ps->s_txreq))
+		goto bail_no_tx;
 
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_SEND_OK)) {
-		if (!(ib_hfi1_state_ops[qp->state] & HFI1_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_iowait.sdma_busy)) {
-			qp->s_flags |= HFI1_S_WAIT_DMA;
+		if (iowait_sdma_pending(&priv->s_iowait)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
 		clear_ahg(qp);
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		hfi1_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
-		goto done;
+		goto done_free_tx;
 	}
 
-	ohdr = &qp->s_hdr->ibh.u.oth;
+	ohdr = &ps->s_txreq->phdr.hdr.u.oth;
 	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
-		ohdr = &qp->s_hdr->ibh.u.l.oth;
+		ohdr = &ps->s_txreq->phdr.hdr.u.l.oth;
 
 	/* Get the next send request. */
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	qp->s_wqe = NULL;
 	switch (qp->s_state) {
 	default:
-		if (!(ib_hfi1_state_ops[qp->state] &
-		    HFI1_PROCESS_NEXT_SEND_OK))
+		if (!(ib_rvt_state_ops[qp->state] &
+		    RVT_PROCESS_NEXT_SEND_OK))
 			goto bail;
 		/* Check if send work queue is empty. */
-		if (qp->s_cur == qp->s_head) {
+		smp_read_barrier_depends(); /* see post_one_send() */
+		if (qp->s_cur == ACCESS_ONCE(qp->s_head)) {
 			clear_ahg(qp);
 			goto bail;
 		}
 		/*
 		 * Start a new request.
 		 */
-		wqe->psn = qp->s_next_psn;
-		qp->s_psn = qp->s_next_psn;
+		qp->s_psn = wqe->psn;
 		qp->s_sge.sge = wqe->sg_list[0];
 		qp->s_sge.sg_list = wqe->sg_list + 1;
 		qp->s_sge.num_sge = wqe->wr.num_sge;
@@ -128,9 +136,9 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
 				len = pmtu;
 				break;
 			}
-			if (wqe->wr.opcode == IB_WR_SEND)
+			if (wqe->wr.opcode == IB_WR_SEND) {
 				qp->s_state = OP(SEND_ONLY);
-			else {
+			} else {
 				qp->s_state =
 					OP(SEND_ONLY_WITH_IMMEDIATE);
 				/* Immediate data comes after the BTH */
@@ -157,9 +165,9 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
 				len = pmtu;
 				break;
 			}
-			if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+			if (wqe->wr.opcode == IB_WR_RDMA_WRITE) {
 				qp->s_state = OP(RDMA_WRITE_ONLY);
-			else {
+			} else {
 				qp->s_state =
 					OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE);
 				/* Immediate data comes after the RETH */
@@ -188,9 +196,9 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
 			middle = HFI1_CAP_IS_KSET(SDMA_AHG);
 			break;
 		}
-		if (wqe->wr.opcode == IB_WR_SEND)
+		if (wqe->wr.opcode == IB_WR_SEND) {
 			qp->s_state = OP(SEND_LAST);
-		else {
+		} else {
 			qp->s_state = OP(SEND_LAST_WITH_IMMEDIATE);
 			/* Immediate data comes after the BTH */
 			ohdr->u.imm_data = wqe->wr.ex.imm_data;
@@ -213,9 +221,9 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
 			middle = HFI1_CAP_IS_KSET(SDMA_AHG);
 			break;
 		}
-		if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE) {
 			qp->s_state = OP(RDMA_WRITE_LAST);
-		else {
+		} else {
 			qp->s_state =
 				OP(RDMA_WRITE_LAST_WITH_IMMEDIATE);
 			/* Immediate data comes after the BTH */
@@ -231,19 +239,28 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
 	}
 	qp->s_len -= len;
 	qp->s_hdrwords = hwords;
+	ps->s_txreq->sde = priv->s_sde;
 	qp->s_cur_sge = &qp->s_sge;
 	qp->s_cur_size = len;
 	hfi1_make_ruc_header(qp, ohdr, bth0 | (qp->s_state << 24),
-			     mask_psn(qp->s_next_psn++), middle);
-done:
-	ret = 1;
-	goto unlock;
+			     mask_psn(qp->s_psn++), middle, ps);
+	/* pbc */
+	ps->s_txreq->hdr_dwords = qp->s_hdrwords + 2;
+	return 1;
+
+done_free_tx:
+	hfi1_put_txreq(ps->s_txreq);
+	ps->s_txreq = NULL;
+	return 1;
 
 bail:
-	qp->s_flags &= ~HFI1_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-	return ret;
+	hfi1_put_txreq(ps->s_txreq);
+
+bail_no_tx:
+	ps->s_txreq = NULL;
+	qp->s_flags &= ~RVT_S_BUSY;
+	qp->s_hdrwords = 0;
+	return 0;
 }
 
 /**
@@ -266,7 +283,7 @@ void hfi1_uc_rcv(struct hfi1_packet *packet)
 	u32 rcv_flags = packet->rcv_flags;
 	void *data = packet->ebuf;
 	u32 tlen = packet->tlen;
-	struct hfi1_qp *qp = packet->qp;
+	struct rvt_qp *qp = packet->qp;
 	struct hfi1_other_headers *ohdr = packet->ohdr;
 	u32 bth0, opcode;
 	u32 hdrsize = packet->hlen;
@@ -291,14 +308,14 @@ void hfi1_uc_rcv(struct hfi1_packet *packet)
 			u16 rlid = be16_to_cpu(hdr->lrh[3]);
 			u8 sl, sc5;
 
-			lqpn = bth1 & HFI1_QPN_MASK;
+			lqpn = bth1 & RVT_QPN_MASK;
 			rqpn = qp->remote_qpn;
 
 			sc5 = ibp->sl_to_sc[qp->remote_ah_attr.sl];
 			sl = ibp->sc_to_sl[sc5];
 
 			process_becn(ppd, sl, rlid, lqpn, rqpn,
-					IB_CC_SVCTYPE_UC);
+				     IB_CC_SVCTYPE_UC);
 		}
 
 		if (bth1 & HFI1_FECN_SMASK) {
@@ -331,10 +348,11 @@ void hfi1_uc_rcv(struct hfi1_packet *packet)
 inv:
 		if (qp->r_state == OP(SEND_FIRST) ||
 		    qp->r_state == OP(SEND_MIDDLE)) {
-			set_bit(HFI1_R_REWIND_SGE, &qp->r_aflags);
+			set_bit(RVT_R_REWIND_SGE, &qp->r_aflags);
 			qp->r_sge.num_sge = 0;
-		} else
-			hfi1_put_ss(&qp->r_sge);
+		} else {
+			rvt_put_ss(&qp->r_sge);
+		}
 		qp->r_state = OP(SEND_LAST);
 		switch (opcode) {
 		case OP(SEND_FIRST):
@@ -381,7 +399,7 @@ inv:
 		goto inv;
 	}
 
-	if (qp->state == IB_QPS_RTR && !(qp->r_flags & HFI1_R_COMM_EST))
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & RVT_R_COMM_EST))
 		qp_comm_est(qp);
 
 	/* OK, process the packet. */
@@ -390,10 +408,10 @@ inv:
 	case OP(SEND_ONLY):
 	case OP(SEND_ONLY_WITH_IMMEDIATE):
 send_first:
-		if (test_and_clear_bit(HFI1_R_REWIND_SGE, &qp->r_aflags))
+		if (test_and_clear_bit(RVT_R_REWIND_SGE, &qp->r_aflags)) {
 			qp->r_sge = qp->s_rdma_read_sge;
-		else {
-			ret = hfi1_get_rwqe(qp, 0);
+		} else {
+			ret = hfi1_rvt_get_rwqe(qp, 0);
 			if (ret < 0)
 				goto op_err;
 			if (!ret)
@@ -417,7 +435,7 @@ send_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto rewind;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, 0);
+		hfi1_copy_sge(&qp->r_sge, data, pmtu, 0, 0);
 		break;
 
 	case OP(SEND_LAST_WITH_IMMEDIATE):
@@ -442,8 +460,8 @@ send_last:
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto rewind;
 		wc.opcode = IB_WC_RECV;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, 0);
-		hfi1_put_ss(&qp->s_rdma_read_sge);
+		hfi1_copy_sge(&qp->r_sge, data, tlen, 0, 0);
+		rvt_put_ss(&qp->s_rdma_read_sge);
 last_imm:
 		wc.wr_id = qp->r_wr_id;
 		wc.status = IB_WC_SUCCESS;
@@ -468,9 +486,9 @@ last_imm:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-			      (ohdr->bth[0] &
-				cpu_to_be32(IB_BTH_SOLICITED)) != 0);
+		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+			     (ohdr->bth[0] &
+			      cpu_to_be32(IB_BTH_SOLICITED)) != 0);
 		break;
 
 	case OP(RDMA_WRITE_FIRST):
@@ -491,8 +509,8 @@ rdma_first:
 			int ok;
 
 			/* Check rkey */
-			ok = hfi1_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
-					  vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
+			ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
+					 vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok))
 				goto drop;
 			qp->r_sge.num_sge = 1;
@@ -503,9 +521,9 @@ rdma_first:
 			qp->r_sge.sge.length = 0;
 			qp->r_sge.sge.sge_length = 0;
 		}
-		if (opcode == OP(RDMA_WRITE_ONLY))
+		if (opcode == OP(RDMA_WRITE_ONLY)) {
 			goto rdma_last;
-		else if (opcode == OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE)) {
+		} else if (opcode == OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE)) {
 			wc.ex.imm_data = ohdr->u.rc.imm_data;
 			goto rdma_last_imm;
 		}
@@ -517,7 +535,7 @@ rdma_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto drop;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, 1);
+		hfi1_copy_sge(&qp->r_sge, data, pmtu, 1, 0);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
@@ -535,10 +553,10 @@ rdma_last_imm:
 		tlen -= (hdrsize + pad + 4);
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
-		if (test_and_clear_bit(HFI1_R_REWIND_SGE, &qp->r_aflags))
-			hfi1_put_ss(&qp->s_rdma_read_sge);
-		else {
-			ret = hfi1_get_rwqe(qp, 1);
+		if (test_and_clear_bit(RVT_R_REWIND_SGE, &qp->r_aflags)) {
+			rvt_put_ss(&qp->s_rdma_read_sge);
+		} else {
+			ret = hfi1_rvt_get_rwqe(qp, 1);
 			if (ret < 0)
 				goto op_err;
 			if (!ret)
@@ -546,8 +564,8 @@ rdma_last_imm:
 		}
 		wc.byte_len = qp->r_len;
 		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, 1);
-		hfi1_put_ss(&qp->r_sge);
+		hfi1_copy_sge(&qp->r_sge, data, tlen, 1, 0);
+		rvt_put_ss(&qp->r_sge);
 		goto last_imm;
 
 	case OP(RDMA_WRITE_LAST):
@@ -562,8 +580,8 @@ rdma_last:
 		tlen -= (hdrsize + pad + 4);
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, 1);
-		hfi1_put_ss(&qp->r_sge);
+		hfi1_copy_sge(&qp->r_sge, data, tlen, 1, 0);
+		rvt_put_ss(&qp->r_sge);
 		break;
 
 	default:
@@ -575,14 +593,12 @@ rdma_last:
 	return;
 
 rewind:
-	set_bit(HFI1_R_REWIND_SGE, &qp->r_aflags);
+	set_bit(RVT_R_REWIND_SGE, &qp->r_aflags);
 	qp->r_sge.num_sge = 0;
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 	return;
 
 op_err:
 	hfi1_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
-	return;
-
 }
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index 25e6053c38db..ae8a70f703eb 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -53,6 +50,7 @@
 
 #include "hfi.h"
 #include "mad.h"
+#include "verbs_txreq.h"
 #include "qp.h"
 
 /**
@@ -65,24 +63,25 @@
  * Note that the receive interrupt handler may be calling hfi1_ud_rcv()
  * while this is being called.
  */
-static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
+static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 {
 	struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
 	struct hfi1_pportdata *ppd;
-	struct hfi1_qp *qp;
+	struct rvt_qp *qp;
 	struct ib_ah_attr *ah_attr;
 	unsigned long flags;
-	struct hfi1_sge_state ssge;
-	struct hfi1_sge *sge;
+	struct rvt_sge_state ssge;
+	struct rvt_sge *sge;
 	struct ib_wc wc;
 	u32 length;
 	enum ib_qp_type sqptype, dqptype;
 
 	rcu_read_lock();
 
-	qp = hfi1_lookup_qpn(ibp, swqe->ud_wr.remote_qpn);
+	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), &ibp->rvp,
+			    swqe->ud_wr.remote_qpn);
 	if (!qp) {
-		ibp->n_pkt_drops++;
+		ibp->rvp.n_pkt_drops++;
 		rcu_read_unlock();
 		return;
 	}
@@ -93,12 +92,12 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 			IB_QPT_UD : qp->ibqp.qp_type;
 
 	if (dqptype != sqptype ||
-	    !(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK)) {
-		ibp->n_pkt_drops++;
+	    !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
+		ibp->rvp.n_pkt_drops++;
 		goto drop;
 	}
 
-	ah_attr = &to_iah(swqe->ud_wr.ah)->attr;
+	ah_attr = &ibah_to_rvtah(swqe->ud_wr.ah)->attr;
 	ppd = ppd_from_ibp(ibp);
 
 	if (qp->ibqp.qp_num > 1) {
@@ -161,35 +160,36 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 	/*
 	 * Get the next work request entry to find where to put the data.
 	 */
-	if (qp->r_flags & HFI1_R_REUSE_SGE)
-		qp->r_flags &= ~HFI1_R_REUSE_SGE;
-	else {
+	if (qp->r_flags & RVT_R_REUSE_SGE) {
+		qp->r_flags &= ~RVT_R_REUSE_SGE;
+	} else {
 		int ret;
 
-		ret = hfi1_get_rwqe(qp, 0);
+		ret = hfi1_rvt_get_rwqe(qp, 0);
 		if (ret < 0) {
 			hfi1_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
 			goto bail_unlock;
 		}
 		if (!ret) {
 			if (qp->ibqp.qp_num == 0)
-				ibp->n_vl15_dropped++;
+				ibp->rvp.n_vl15_dropped++;
 			goto bail_unlock;
 		}
 	}
 	/* Silently drop packets which are too big. */
 	if (unlikely(wc.byte_len > qp->r_len)) {
-		qp->r_flags |= HFI1_R_REUSE_SGE;
-		ibp->n_pkt_drops++;
+		qp->r_flags |= RVT_R_REUSE_SGE;
+		ibp->rvp.n_pkt_drops++;
 		goto bail_unlock;
 	}
 
 	if (ah_attr->ah_flags & IB_AH_GRH) {
 		hfi1_copy_sge(&qp->r_sge, &ah_attr->grh,
-			      sizeof(struct ib_grh), 1);
+			      sizeof(struct ib_grh), 1, 0);
 		wc.wc_flags |= IB_WC_GRH;
-	} else
+	} else {
 		hfi1_skip_sge(&qp->r_sge, sizeof(struct ib_grh), 1);
+	}
 	ssge.sg_list = swqe->sg_list + 1;
 	ssge.sge = *swqe->sg_list;
 	ssge.num_sge = swqe->wr.num_sge;
@@ -202,7 +202,7 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		WARN_ON_ONCE(len == 0);
-		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, 1);
+		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, 1, 0);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
@@ -210,7 +210,7 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 			if (--ssge.num_sge)
 				*sge = *ssge.sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= HFI1_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -222,8 +222,8 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 		}
 		length -= len;
 	}
-	hfi1_put_ss(&qp->r_sge);
-	if (!test_and_clear_bit(HFI1_R_WRID_VALID, &qp->r_aflags))
+	rvt_put_ss(&qp->r_sge);
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		goto bail_unlock;
 	wc.wr_id = qp->r_wr_id;
 	wc.status = IB_WC_SUCCESS;
@@ -242,14 +242,14 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
 	wc.slid = ppd->lid | (ah_attr->src_path_bits & ((1 << ppd->lmc) - 1));
 	/* Check for loopback when the port lid is not set */
 	if (wc.slid == 0 && sqp->ibqp.qp_type == IB_QPT_GSI)
-		wc.slid = HFI1_PERMISSIVE_LID;
+		wc.slid = be16_to_cpu(IB_LID_PERMISSIVE);
 	wc.sl = ah_attr->sl;
 	wc.dlid_path_bits = ah_attr->dlid & ((1 << ppd->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-		      swqe->wr.send_flags & IB_SEND_SOLICITED);
-	ibp->n_loop_pkts++;
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+		     swqe->wr.send_flags & IB_SEND_SOLICITED);
+	ibp->rvp.n_loop_pkts++;
 bail_unlock:
 	spin_unlock_irqrestore(&qp->r_lock, flags);
 drop:
@@ -260,47 +260,53 @@ drop:
  * hfi1_make_ud_req - construct a UD request packet
  * @qp: the QP
  *
+ * Assume s_lock is held.
+ *
  * Return 1 if constructed; otherwise, return 0.
  */
-int hfi1_make_ud_req(struct hfi1_qp *qp)
+int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_other_headers *ohdr;
 	struct ib_ah_attr *ah_attr;
 	struct hfi1_pportdata *ppd;
 	struct hfi1_ibport *ibp;
-	struct hfi1_swqe *wqe;
-	unsigned long flags;
+	struct rvt_swqe *wqe;
 	u32 nwords;
 	u32 extra_bytes;
 	u32 bth0;
 	u16 lrh0;
 	u16 lid;
-	int ret = 0;
 	int next_cur;
 	u8 sc5;
 
-	spin_lock_irqsave(&qp->s_lock, flags);
+	ps->s_txreq = get_txreq(ps->dev, qp);
+	if (IS_ERR(ps->s_txreq))
+		goto bail_no_tx;
 
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_hfi1_state_ops[qp->state] & HFI1_FLUSH_SEND))
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_rvt_state_ops[qp->state] & RVT_FLUSH_SEND))
 			goto bail;
 		/* We are in the error state, flush the work request. */
-		if (qp->s_last == qp->s_head)
+		smp_read_barrier_depends(); /* see post_one_send */
+		if (qp->s_last == ACCESS_ONCE(qp->s_head))
 			goto bail;
 		/* If DMAs are in progress, we can't flush immediately. */
-		if (atomic_read(&qp->s_iowait.sdma_busy)) {
-			qp->s_flags |= HFI1_S_WAIT_DMA;
+		if (iowait_sdma_pending(&priv->s_iowait)) {
+			qp->s_flags |= RVT_S_WAIT_DMA;
 			goto bail;
 		}
-		wqe = get_swqe_ptr(qp, qp->s_last);
+		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
 		hfi1_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
-		goto done;
+		goto done_free_tx;
 	}
 
-	if (qp->s_cur == qp->s_head)
+	/* see post_one_send() */
+	smp_read_barrier_depends();
+	if (qp->s_cur == ACCESS_ONCE(qp->s_head))
 		goto bail;
 
-	wqe = get_swqe_ptr(qp, qp->s_cur);
+	wqe = rvt_get_swqe_ptr(qp, qp->s_cur);
 	next_cur = qp->s_cur + 1;
 	if (next_cur >= qp->s_size)
 		next_cur = 0;
@@ -308,13 +314,15 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 	/* Construct the header. */
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &to_iah(wqe->ud_wr.ah)->attr;
-	if (ah_attr->dlid < HFI1_MULTICAST_LID_BASE ||
-	    ah_attr->dlid == HFI1_PERMISSIVE_LID) {
+	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	if (ah_attr->dlid < be16_to_cpu(IB_MULTICAST_LID_BASE) ||
+	    ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE)) {
 		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
-		if (unlikely(!loopback && (lid == ppd->lid ||
-		    (lid == HFI1_PERMISSIVE_LID &&
-		     qp->ibqp.qp_type == IB_QPT_GSI)))) {
+		if (unlikely(!loopback &&
+			     (lid == ppd->lid ||
+			      (lid == be16_to_cpu(IB_LID_PERMISSIVE) &&
+			      qp->ibqp.qp_type == IB_QPT_GSI)))) {
+			unsigned long flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -322,16 +330,17 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 			 * Instead of waiting, we could queue a
 			 * zero length descriptor so we get a callback.
 			 */
-			if (atomic_read(&qp->s_iowait.sdma_busy)) {
-				qp->s_flags |= HFI1_S_WAIT_DMA;
+			if (iowait_sdma_pending(&priv->s_iowait)) {
+				qp->s_flags |= RVT_S_WAIT_DMA;
 				goto bail;
 			}
 			qp->s_cur = next_cur;
+			local_irq_save(flags);
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 			ud_loopback(qp, wqe);
 			spin_lock_irqsave(&qp->s_lock, flags);
 			hfi1_send_complete(qp, wqe, IB_WC_SUCCESS);
-			goto done;
+			goto done_free_tx;
 		}
 	}
 
@@ -353,11 +362,12 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 
 	if (ah_attr->ah_flags & IB_AH_GRH) {
 		/* Header size in 32-bit words. */
-		qp->s_hdrwords += hfi1_make_grh(ibp, &qp->s_hdr->ibh.u.l.grh,
-					       &ah_attr->grh,
-					       qp->s_hdrwords, nwords);
+		qp->s_hdrwords += hfi1_make_grh(ibp,
+						&ps->s_txreq->phdr.hdr.u.l.grh,
+						&ah_attr->grh,
+						qp->s_hdrwords, nwords);
 		lrh0 = HFI1_LRH_GRH;
-		ohdr = &qp->s_hdr->ibh.u.l.oth;
+		ohdr = &ps->s_txreq->phdr.hdr.u.l.oth;
 		/*
 		 * Don't worry about sending to locally attached multicast
 		 * QPs.  It is unspecified by the spec. what happens.
@@ -365,37 +375,42 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 	} else {
 		/* Header size in 32-bit words. */
 		lrh0 = HFI1_LRH_BTH;
-		ohdr = &qp->s_hdr->ibh.u.oth;
+		ohdr = &ps->s_txreq->phdr.hdr.u.oth;
 	}
 	if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
 		qp->s_hdrwords++;
 		ohdr->u.ud.imm_data = wqe->wr.ex.imm_data;
 		bth0 = IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE << 24;
-	} else
+	} else {
 		bth0 = IB_OPCODE_UD_SEND_ONLY << 24;
+	}
 	sc5 = ibp->sl_to_sc[ah_attr->sl];
 	lrh0 |= (ah_attr->sl & 0xf) << 4;
 	if (qp->ibqp.qp_type == IB_QPT_SMI) {
 		lrh0 |= 0xF000; /* Set VL (see ch. 13.5.3.1) */
-		qp->s_sc = 0xf;
+		priv->s_sc = 0xf;
 	} else {
 		lrh0 |= (sc5 & 0xf) << 12;
-		qp->s_sc = sc5;
+		priv->s_sc = sc5;
 	}
-	qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
-	qp->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
-	qp->s_hdr->ibh.lrh[1] = cpu_to_be16(ah_attr->dlid);  /* DEST LID */
-	qp->s_hdr->ibh.lrh[2] =
+	priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
+	ps->s_txreq->sde = priv->s_sde;
+	priv->s_sendcontext = qp_to_send_context(qp, priv->s_sc);
+	ps->s_txreq->psc = priv->s_sendcontext;
+	ps->s_txreq->phdr.hdr.lrh[0] = cpu_to_be16(lrh0);
+	ps->s_txreq->phdr.hdr.lrh[1] = cpu_to_be16(ah_attr->dlid);
+	ps->s_txreq->phdr.hdr.lrh[2] =
 		cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
-	if (ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE))
-		qp->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
-	else {
+	if (ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE)) {
+		ps->s_txreq->phdr.hdr.lrh[3] = IB_LID_PERMISSIVE;
+	} else {
 		lid = ppd->lid;
 		if (lid) {
 			lid |= ah_attr->src_path_bits & ((1 << ppd->lmc) - 1);
-			qp->s_hdr->ibh.lrh[3] = cpu_to_be16(lid);
-		} else
-			qp->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
+			ps->s_txreq->phdr.hdr.lrh[3] = cpu_to_be16(lid);
+		} else {
+			ps->s_txreq->phdr.hdr.lrh[3] = IB_LID_PERMISSIVE;
+		}
 	}
 	if (wqe->wr.send_flags & IB_SEND_SOLICITED)
 		bth0 |= IB_BTH_SOLICITED;
@@ -406,7 +421,7 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 		bth0 |= hfi1_get_pkey(ibp, qp->s_pkey_index);
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	ohdr->bth[1] = cpu_to_be32(wqe->ud_wr.remote_qpn);
-	ohdr->bth[2] = cpu_to_be32(mask_psn(qp->s_next_psn++));
+	ohdr->bth[2] = cpu_to_be32(mask_psn(wqe->psn));
 	/*
 	 * Qkeys with the high order bit set mean use the
 	 * qkey from the QP context instead of the WR (see 10.2.5).
@@ -415,20 +430,28 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
 					 qp->qkey : wqe->ud_wr.remote_qkey);
 	ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
 	/* disarm any ahg */
-	qp->s_hdr->ahgcount = 0;
-	qp->s_hdr->ahgidx = 0;
-	qp->s_hdr->tx_flags = 0;
-	qp->s_hdr->sde = NULL;
+	priv->s_hdr->ahgcount = 0;
+	priv->s_hdr->ahgidx = 0;
+	priv->s_hdr->tx_flags = 0;
+	priv->s_hdr->sde = NULL;
+	/* pbc */
+	ps->s_txreq->hdr_dwords = qp->s_hdrwords + 2;
+
+	return 1;
 
-done:
-	ret = 1;
-	goto unlock;
+done_free_tx:
+	hfi1_put_txreq(ps->s_txreq);
+	ps->s_txreq = NULL;
+	return 1;
 
 bail:
-	qp->s_flags &= ~HFI1_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-	return ret;
+	hfi1_put_txreq(ps->s_txreq);
+
+bail_no_tx:
+	ps->s_txreq = NULL;
+	qp->s_flags &= ~RVT_S_BUSY;
+	qp->s_hdrwords = 0;
+	return 0;
 }
 
 /*
@@ -476,7 +499,7 @@ int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey)
 	return -1;
 }
 
-void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
+void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn,
 		u32 pkey, u32 slid, u32 dlid, u8 sc5,
 		const struct ib_grh *old_grh)
 {
@@ -550,7 +573,7 @@ void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
  * opa_smp_check() returns 0 if all checks succeed, 1 otherwise.
  */
 static int opa_smp_check(struct hfi1_ibport *ibp, u16 pkey, u8 sc5,
-			 struct hfi1_qp *qp, u16 slid, struct opa_smp *smp)
+			 struct rvt_qp *qp, u16 slid, struct opa_smp *smp)
 {
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 
@@ -607,7 +630,7 @@ static int opa_smp_check(struct hfi1_ibport *ibp, u16 pkey, u8 sc5,
 	case IB_MGMT_METHOD_TRAP:
 	case IB_MGMT_METHOD_GET_RESP:
 	case IB_MGMT_METHOD_REPORT_RESP:
-		if (ibp->port_cap_flags & IB_PORT_SM)
+		if (ibp->rvp.port_cap_flags & IB_PORT_SM)
 			return 0;
 		if (pkey == FULL_MGMT_P_KEY) {
 			smp->status |= IB_SMP_UNSUP_METHOD;
@@ -624,7 +647,6 @@ static int opa_smp_check(struct hfi1_ibport *ibp, u16 pkey, u8 sc5,
 	return 0;
 }
 
-
 /**
  * hfi1_ud_rcv - receive an incoming UD packet
  * @ibp: the port the packet came in on
@@ -654,7 +676,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	u32 rcv_flags = packet->rcv_flags;
 	void *data = packet->ebuf;
 	u32 tlen = packet->tlen;
-	struct hfi1_qp *qp = packet->qp;
+	struct rvt_qp *qp = packet->qp;
 	bool has_grh = rcv_flags & HFI1_HAS_GRH;
 	bool sc4_bit = has_sc4_bit(packet);
 	u8 sc;
@@ -663,10 +685,10 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	struct ib_grh *grh = NULL;
 
 	qkey = be32_to_cpu(ohdr->u.ud.deth[0]);
-	src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & HFI1_QPN_MASK;
+	src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & RVT_QPN_MASK;
 	dlid = be16_to_cpu(hdr->lrh[1]);
-	is_mcast = (dlid > HFI1_MULTICAST_LID_BASE) &&
-			(dlid != HFI1_PERMISSIVE_LID);
+	is_mcast = (dlid > be16_to_cpu(IB_MULTICAST_LID_BASE)) &&
+			(dlid != be16_to_cpu(IB_LID_PERMISSIVE));
 	bth1 = be32_to_cpu(ohdr->bth[1]);
 	if (unlikely(bth1 & HFI1_BECN_SMASK)) {
 		/*
@@ -674,7 +696,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 		 * error path.
 		 */
 		struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
-		u32 lqpn =  be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
+		u32 lqpn =  be32_to_cpu(ohdr->bth[1]) & RVT_QPN_MASK;
 		u8 sl, sc5;
 
 		sc5 = (be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf;
@@ -750,7 +772,6 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 			mgmt_pkey_idx = hfi1_lookup_pkey_idx(ibp, pkey);
 			if (mgmt_pkey_idx < 0)
 				goto drop;
-
 		}
 		if (unlikely(qkey != qp->qkey)) {
 			hfi1_bad_pqkey(ibp, OPA_TRAP_BAD_Q_KEY, qkey,
@@ -788,7 +809,6 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 		mgmt_pkey_idx = hfi1_lookup_pkey_idx(ibp, pkey);
 		if (mgmt_pkey_idx < 0)
 			goto drop;
-
 	}
 
 	if (qp->ibqp.qp_num > 1 &&
@@ -799,8 +819,9 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
 		wc.ex.imm_data = 0;
 		wc.wc_flags = 0;
-	} else
+	} else {
 		goto drop;
+	}
 
 	/*
 	 * A GRH is expected to precede the data even if not
@@ -811,36 +832,38 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	/*
 	 * Get the next work request entry to find where to put the data.
 	 */
-	if (qp->r_flags & HFI1_R_REUSE_SGE)
-		qp->r_flags &= ~HFI1_R_REUSE_SGE;
-	else {
+	if (qp->r_flags & RVT_R_REUSE_SGE) {
+		qp->r_flags &= ~RVT_R_REUSE_SGE;
+	} else {
 		int ret;
 
-		ret = hfi1_get_rwqe(qp, 0);
+		ret = hfi1_rvt_get_rwqe(qp, 0);
 		if (ret < 0) {
 			hfi1_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
 			return;
 		}
 		if (!ret) {
 			if (qp->ibqp.qp_num == 0)
-				ibp->n_vl15_dropped++;
+				ibp->rvp.n_vl15_dropped++;
 			return;
 		}
 	}
 	/* Silently drop packets which are too big. */
 	if (unlikely(wc.byte_len > qp->r_len)) {
-		qp->r_flags |= HFI1_R_REUSE_SGE;
+		qp->r_flags |= RVT_R_REUSE_SGE;
 		goto drop;
 	}
 	if (has_grh) {
 		hfi1_copy_sge(&qp->r_sge, &hdr->u.l.grh,
-			      sizeof(struct ib_grh), 1);
+			      sizeof(struct ib_grh), 1, 0);
 		wc.wc_flags |= IB_WC_GRH;
-	} else
+	} else {
 		hfi1_skip_sge(&qp->r_sge, sizeof(struct ib_grh), 1);
-	hfi1_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh), 1);
-	hfi1_put_ss(&qp->r_sge);
-	if (!test_and_clear_bit(HFI1_R_WRID_VALID, &qp->r_aflags))
+	}
+	hfi1_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh),
+		      1, 0);
+	rvt_put_ss(&qp->r_sge);
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		return;
 	wc.wr_id = qp->r_wr_id;
 	wc.status = IB_WC_SUCCESS;
@@ -862,8 +885,9 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 			}
 		}
 		wc.pkey_index = (unsigned)mgmt_pkey_idx;
-	} else
+	} else {
 		wc.pkey_index = 0;
+	}
 
 	wc.slid = be16_to_cpu(hdr->lrh[3]);
 	sc = (be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf;
@@ -873,15 +897,15 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	/*
 	 * Save the LMC lower bits if the destination LID is a unicast LID.
 	 */
-	wc.dlid_path_bits = dlid >= HFI1_MULTICAST_LID_BASE ? 0 :
+	wc.dlid_path_bits = dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE) ? 0 :
 		dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
-		      (ohdr->bth[0] &
-			cpu_to_be32(IB_BTH_SOLICITED)) != 0);
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+		     (ohdr->bth[0] &
+		      cpu_to_be32(IB_BTH_SOLICITED)) != 0);
 	return;
 
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 }
diff --git a/drivers/staging/rdma/hfi1/user_exp_rcv.c b/drivers/staging/rdma/hfi1/user_exp_rcv.c
new file mode 100644
index 000000000000..0861e095df8d
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/user_exp_rcv.c
@@ -0,0 +1,1044 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+#include <asm/page.h>
+
+#include "user_exp_rcv.h"
+#include "trace.h"
+#include "mmu_rb.h"
+
+struct tid_group {
+	struct list_head list;
+	unsigned base;
+	u8 size;
+	u8 used;
+	u8 map;
+};
+
+struct tid_rb_node {
+	struct mmu_rb_node mmu;
+	unsigned long phys;
+	struct tid_group *grp;
+	u32 rcventry;
+	dma_addr_t dma_addr;
+	bool freed;
+	unsigned npages;
+	struct page *pages[0];
+};
+
+struct tid_pageset {
+	u16 idx;
+	u16 count;
+};
+
+#define EXP_TID_SET_EMPTY(set) (set.count == 0 && list_empty(&set.list))
+
+#define num_user_pages(vaddr, len)				       \
+	(1 + (((((unsigned long)(vaddr) +			       \
+		 (unsigned long)(len) - 1) & PAGE_MASK) -	       \
+	       ((unsigned long)vaddr & PAGE_MASK)) >> PAGE_SHIFT))
+
+static void unlock_exp_tids(struct hfi1_ctxtdata *, struct exp_tid_set *,
+			    struct rb_root *);
+static u32 find_phys_blocks(struct page **, unsigned, struct tid_pageset *);
+static int set_rcvarray_entry(struct file *, unsigned long, u32,
+			      struct tid_group *, struct page **, unsigned);
+static int mmu_rb_insert(struct rb_root *, struct mmu_rb_node *);
+static void mmu_rb_remove(struct rb_root *, struct mmu_rb_node *, bool);
+static int mmu_rb_invalidate(struct rb_root *, struct mmu_rb_node *);
+static int program_rcvarray(struct file *, unsigned long, struct tid_group *,
+			    struct tid_pageset *, unsigned, u16, struct page **,
+			    u32 *, unsigned *, unsigned *);
+static int unprogram_rcvarray(struct file *, u32, struct tid_group **);
+static void clear_tid_node(struct hfi1_filedata *, u16, struct tid_rb_node *);
+
+static struct mmu_rb_ops tid_rb_ops = {
+	.insert = mmu_rb_insert,
+	.remove = mmu_rb_remove,
+	.invalidate = mmu_rb_invalidate
+};
+
+static inline u32 rcventry2tidinfo(u32 rcventry)
+{
+	u32 pair = rcventry & ~0x1;
+
+	return EXP_TID_SET(IDX, pair >> 1) |
+		EXP_TID_SET(CTRL, 1 << (rcventry - pair));
+}
+
+static inline void exp_tid_group_init(struct exp_tid_set *set)
+{
+	INIT_LIST_HEAD(&set->list);
+	set->count = 0;
+}
+
+static inline void tid_group_remove(struct tid_group *grp,
+				    struct exp_tid_set *set)
+{
+	list_del_init(&grp->list);
+	set->count--;
+}
+
+static inline void tid_group_add_tail(struct tid_group *grp,
+				      struct exp_tid_set *set)
+{
+	list_add_tail(&grp->list, &set->list);
+	set->count++;
+}
+
+static inline struct tid_group *tid_group_pop(struct exp_tid_set *set)
+{
+	struct tid_group *grp =
+		list_first_entry(&set->list, struct tid_group, list);
+	list_del_init(&grp->list);
+	set->count--;
+	return grp;
+}
+
+static inline void tid_group_move(struct tid_group *group,
+				  struct exp_tid_set *s1,
+				  struct exp_tid_set *s2)
+{
+	tid_group_remove(group, s1);
+	tid_group_add_tail(group, s2);
+}
+
+/*
+ * Initialize context and file private data needed for Expected
+ * receive caching. This needs to be done after the context has
+ * been configured with the eager/expected RcvEntry counts.
+ */
+int hfi1_user_exp_rcv_init(struct file *fp)
+{
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_devdata *dd = uctxt->dd;
+	unsigned tidbase;
+	int i, ret = 0;
+
+	spin_lock_init(&fd->tid_lock);
+	spin_lock_init(&fd->invalid_lock);
+	fd->tid_rb_root = RB_ROOT;
+
+	if (!uctxt->subctxt_cnt || !fd->subctxt) {
+		exp_tid_group_init(&uctxt->tid_group_list);
+		exp_tid_group_init(&uctxt->tid_used_list);
+		exp_tid_group_init(&uctxt->tid_full_list);
+
+		tidbase = uctxt->expected_base;
+		for (i = 0; i < uctxt->expected_count /
+			     dd->rcv_entries.group_size; i++) {
+			struct tid_group *grp;
+
+			grp = kzalloc(sizeof(*grp), GFP_KERNEL);
+			if (!grp) {
+				/*
+				 * If we fail here, the groups already
+				 * allocated will be freed by the close
+				 * call.
+				 */
+				ret = -ENOMEM;
+				goto done;
+			}
+			grp->size = dd->rcv_entries.group_size;
+			grp->base = tidbase;
+			tid_group_add_tail(grp, &uctxt->tid_group_list);
+			tidbase += dd->rcv_entries.group_size;
+		}
+	}
+
+	fd->entry_to_rb = kcalloc(uctxt->expected_count,
+				     sizeof(struct rb_node *),
+				     GFP_KERNEL);
+	if (!fd->entry_to_rb)
+		return -ENOMEM;
+
+	if (!HFI1_CAP_IS_USET(TID_UNMAP)) {
+		fd->invalid_tid_idx = 0;
+		fd->invalid_tids = kzalloc(uctxt->expected_count *
+					   sizeof(u32), GFP_KERNEL);
+		if (!fd->invalid_tids) {
+			ret = -ENOMEM;
+			goto done;
+		}
+
+		/*
+		 * Register MMU notifier callbacks. If the registration
+		 * fails, continue but turn off the TID caching for
+		 * all user contexts.
+		 */
+		ret = hfi1_mmu_rb_register(&fd->tid_rb_root, &tid_rb_ops);
+		if (ret) {
+			dd_dev_info(dd,
+				    "Failed MMU notifier registration %d\n",
+				    ret);
+			HFI1_CAP_USET(TID_UNMAP);
+			ret = 0;
+		}
+	}
+
+	/*
+	 * PSM does not have a good way to separate, count, and
+	 * effectively enforce a limit on RcvArray entries used by
+	 * subctxts (when context sharing is used) when TID caching
+	 * is enabled. To help with that, we calculate a per-process
+	 * RcvArray entry share and enforce that.
+	 * If TID caching is not in use, PSM deals with usage on its
+	 * own. In that case, we allow any subctxt to take all of the
+	 * entries.
+	 *
+	 * Make sure that we set the tid counts only after successful
+	 * init.
+	 */
+	spin_lock(&fd->tid_lock);
+	if (uctxt->subctxt_cnt && !HFI1_CAP_IS_USET(TID_UNMAP)) {
+		u16 remainder;
+
+		fd->tid_limit = uctxt->expected_count / uctxt->subctxt_cnt;
+		remainder = uctxt->expected_count % uctxt->subctxt_cnt;
+		if (remainder && fd->subctxt < remainder)
+			fd->tid_limit++;
+	} else {
+		fd->tid_limit = uctxt->expected_count;
+	}
+	spin_unlock(&fd->tid_lock);
+done:
+	return ret;
+}
+
+int hfi1_user_exp_rcv_free(struct hfi1_filedata *fd)
+{
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct tid_group *grp, *gptr;
+
+	/*
+	 * The notifier would have been removed when the process'es mm
+	 * was freed.
+	 */
+	if (!HFI1_CAP_IS_USET(TID_UNMAP))
+		hfi1_mmu_rb_unregister(&fd->tid_rb_root);
+
+	kfree(fd->invalid_tids);
+
+	if (!uctxt->cnt) {
+		if (!EXP_TID_SET_EMPTY(uctxt->tid_full_list))
+			unlock_exp_tids(uctxt, &uctxt->tid_full_list,
+					&fd->tid_rb_root);
+		if (!EXP_TID_SET_EMPTY(uctxt->tid_used_list))
+			unlock_exp_tids(uctxt, &uctxt->tid_used_list,
+					&fd->tid_rb_root);
+		list_for_each_entry_safe(grp, gptr, &uctxt->tid_group_list.list,
+					 list) {
+			list_del_init(&grp->list);
+			kfree(grp);
+		}
+		hfi1_clear_tids(uctxt);
+	}
+
+	kfree(fd->entry_to_rb);
+	return 0;
+}
+
+/*
+ * Write an "empty" RcvArray entry.
+ * This function exists so the TID registaration code can use it
+ * to write to unused/unneeded entries and still take advantage
+ * of the WC performance improvements. The HFI will ignore this
+ * write to the RcvArray entry.
+ */
+static inline void rcv_array_wc_fill(struct hfi1_devdata *dd, u32 index)
+{
+	/*
+	 * Doing the WC fill writes only makes sense if the device is
+	 * present and the RcvArray has been mapped as WC memory.
+	 */
+	if ((dd->flags & HFI1_PRESENT) && dd->rcvarray_wc)
+		writeq(0, dd->rcvarray_wc + (index * 8));
+}
+
+/*
+ * RcvArray entry allocation for Expected Receives is done by the
+ * following algorithm:
+ *
+ * The context keeps 3 lists of groups of RcvArray entries:
+ *   1. List of empty groups - tid_group_list
+ *      This list is created during user context creation and
+ *      contains elements which describe sets (of 8) of empty
+ *      RcvArray entries.
+ *   2. List of partially used groups - tid_used_list
+ *      This list contains sets of RcvArray entries which are
+ *      not completely used up. Another mapping request could
+ *      use some of all of the remaining entries.
+ *   3. List of full groups - tid_full_list
+ *      This is the list where sets that are completely used
+ *      up go.
+ *
+ * An attempt to optimize the usage of RcvArray entries is
+ * made by finding all sets of physically contiguous pages in a
+ * user's buffer.
+ * These physically contiguous sets are further split into
+ * sizes supported by the receive engine of the HFI. The
+ * resulting sets of pages are stored in struct tid_pageset,
+ * which describes the sets as:
+ *    * .count - number of pages in this set
+ *    * .idx - starting index into struct page ** array
+ *                    of this set
+ *
+ * From this point on, the algorithm deals with the page sets
+ * described above. The number of pagesets is divided by the
+ * RcvArray group size to produce the number of full groups
+ * needed.
+ *
+ * Groups from the 3 lists are manipulated using the following
+ * rules:
+ *   1. For each set of 8 pagesets, a complete group from
+ *      tid_group_list is taken, programmed, and moved to
+ *      the tid_full_list list.
+ *   2. For all remaining pagesets:
+ *      2.1 If the tid_used_list is empty and the tid_group_list
+ *          is empty, stop processing pageset and return only
+ *          what has been programmed up to this point.
+ *      2.2 If the tid_used_list is empty and the tid_group_list
+ *          is not empty, move a group from tid_group_list to
+ *          tid_used_list.
+ *      2.3 For each group is tid_used_group, program as much as
+ *          can fit into the group. If the group becomes fully
+ *          used, move it to tid_full_list.
+ */
+int hfi1_user_exp_rcv_setup(struct file *fp, struct hfi1_tid_info *tinfo)
+{
+	int ret = 0, need_group = 0, pinned;
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_devdata *dd = uctxt->dd;
+	unsigned npages, ngroups, pageidx = 0, pageset_count, npagesets,
+		tididx = 0, mapped, mapped_pages = 0;
+	unsigned long vaddr = tinfo->vaddr;
+	struct page **pages = NULL;
+	u32 *tidlist = NULL;
+	struct tid_pageset *pagesets = NULL;
+
+	/* Get the number of pages the user buffer spans */
+	npages = num_user_pages(vaddr, tinfo->length);
+	if (!npages)
+		return -EINVAL;
+
+	if (npages > uctxt->expected_count) {
+		dd_dev_err(dd, "Expected buffer too big\n");
+		return -EINVAL;
+	}
+
+	/* Verify that access is OK for the user buffer */
+	if (!access_ok(VERIFY_WRITE, (void __user *)vaddr,
+		       npages * PAGE_SIZE)) {
+		dd_dev_err(dd, "Fail vaddr %p, %u pages, !access_ok\n",
+			   (void *)vaddr, npages);
+		return -EFAULT;
+	}
+
+	pagesets = kcalloc(uctxt->expected_count, sizeof(*pagesets),
+			   GFP_KERNEL);
+	if (!pagesets)
+		return -ENOMEM;
+
+	/* Allocate the array of struct page pointers needed for pinning */
+	pages = kcalloc(npages, sizeof(*pages), GFP_KERNEL);
+	if (!pages) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	/*
+	 * Pin all the pages of the user buffer. If we can't pin all the
+	 * pages, accept the amount pinned so far and program only that.
+	 * User space knows how to deal with partially programmed buffers.
+	 */
+	if (!hfi1_can_pin_pages(dd, fd->tid_n_pinned, npages))
+		return -ENOMEM;
+	pinned = hfi1_acquire_user_pages(vaddr, npages, true, pages);
+	if (pinned <= 0) {
+		ret = pinned;
+		goto bail;
+	}
+	fd->tid_n_pinned += npages;
+
+	/* Find sets of physically contiguous pages */
+	npagesets = find_phys_blocks(pages, pinned, pagesets);
+
+	/*
+	 * We don't need to access this under a lock since tid_used is per
+	 * process and the same process cannot be in hfi1_user_exp_rcv_clear()
+	 * and hfi1_user_exp_rcv_setup() at the same time.
+	 */
+	spin_lock(&fd->tid_lock);
+	if (fd->tid_used + npagesets > fd->tid_limit)
+		pageset_count = fd->tid_limit - fd->tid_used;
+	else
+		pageset_count = npagesets;
+	spin_unlock(&fd->tid_lock);
+
+	if (!pageset_count)
+		goto bail;
+
+	ngroups = pageset_count / dd->rcv_entries.group_size;
+	tidlist = kcalloc(pageset_count, sizeof(*tidlist), GFP_KERNEL);
+	if (!tidlist) {
+		ret = -ENOMEM;
+		goto nomem;
+	}
+
+	tididx = 0;
+
+	/*
+	 * From this point on, we are going to be using shared (between master
+	 * and subcontexts) context resources. We need to take the lock.
+	 */
+	mutex_lock(&uctxt->exp_lock);
+	/*
+	 * The first step is to program the RcvArray entries which are complete
+	 * groups.
+	 */
+	while (ngroups && uctxt->tid_group_list.count) {
+		struct tid_group *grp =
+			tid_group_pop(&uctxt->tid_group_list);
+
+		ret = program_rcvarray(fp, vaddr, grp, pagesets,
+				       pageidx, dd->rcv_entries.group_size,
+				       pages, tidlist, &tididx, &mapped);
+		/*
+		 * If there was a failure to program the RcvArray
+		 * entries for the entire group, reset the grp fields
+		 * and add the grp back to the free group list.
+		 */
+		if (ret <= 0) {
+			tid_group_add_tail(grp, &uctxt->tid_group_list);
+			hfi1_cdbg(TID,
+				  "Failed to program RcvArray group %d", ret);
+			goto unlock;
+		}
+
+		tid_group_add_tail(grp, &uctxt->tid_full_list);
+		ngroups--;
+		pageidx += ret;
+		mapped_pages += mapped;
+	}
+
+	while (pageidx < pageset_count) {
+		struct tid_group *grp, *ptr;
+		/*
+		 * If we don't have any partially used tid groups, check
+		 * if we have empty groups. If so, take one from there and
+		 * put in the partially used list.
+		 */
+		if (!uctxt->tid_used_list.count || need_group) {
+			if (!uctxt->tid_group_list.count)
+				goto unlock;
+
+			grp = tid_group_pop(&uctxt->tid_group_list);
+			tid_group_add_tail(grp, &uctxt->tid_used_list);
+			need_group = 0;
+		}
+		/*
+		 * There is an optimization opportunity here - instead of
+		 * fitting as many page sets as we can, check for a group
+		 * later on in the list that could fit all of them.
+		 */
+		list_for_each_entry_safe(grp, ptr, &uctxt->tid_used_list.list,
+					 list) {
+			unsigned use = min_t(unsigned, pageset_count - pageidx,
+					     grp->size - grp->used);
+
+			ret = program_rcvarray(fp, vaddr, grp, pagesets,
+					       pageidx, use, pages, tidlist,
+					       &tididx, &mapped);
+			if (ret < 0) {
+				hfi1_cdbg(TID,
+					  "Failed to program RcvArray entries %d",
+					  ret);
+				ret = -EFAULT;
+				goto unlock;
+			} else if (ret > 0) {
+				if (grp->used == grp->size)
+					tid_group_move(grp,
+						       &uctxt->tid_used_list,
+						       &uctxt->tid_full_list);
+				pageidx += ret;
+				mapped_pages += mapped;
+				need_group = 0;
+				/* Check if we are done so we break out early */
+				if (pageidx >= pageset_count)
+					break;
+			} else if (WARN_ON(ret == 0)) {
+				/*
+				 * If ret is 0, we did not program any entries
+				 * into this group, which can only happen if
+				 * we've screwed up the accounting somewhere.
+				 * Warn and try to continue.
+				 */
+				need_group = 1;
+			}
+		}
+	}
+unlock:
+	mutex_unlock(&uctxt->exp_lock);
+nomem:
+	hfi1_cdbg(TID, "total mapped: tidpairs:%u pages:%u (%d)", tididx,
+		  mapped_pages, ret);
+	if (tididx) {
+		spin_lock(&fd->tid_lock);
+		fd->tid_used += tididx;
+		spin_unlock(&fd->tid_lock);
+		tinfo->tidcnt = tididx;
+		tinfo->length = mapped_pages * PAGE_SIZE;
+
+		if (copy_to_user((void __user *)(unsigned long)tinfo->tidlist,
+				 tidlist, sizeof(tidlist[0]) * tididx)) {
+			/*
+			 * On failure to copy to the user level, we need to undo
+			 * everything done so far so we don't leak resources.
+			 */
+			tinfo->tidlist = (unsigned long)&tidlist;
+			hfi1_user_exp_rcv_clear(fp, tinfo);
+			tinfo->tidlist = 0;
+			ret = -EFAULT;
+			goto bail;
+		}
+	}
+
+	/*
+	 * If not everything was mapped (due to insufficient RcvArray entries,
+	 * for example), unpin all unmapped pages so we can pin them nex time.
+	 */
+	if (mapped_pages != pinned) {
+		hfi1_release_user_pages(current->mm, &pages[mapped_pages],
+					pinned - mapped_pages,
+					false);
+		fd->tid_n_pinned -= pinned - mapped_pages;
+	}
+bail:
+	kfree(pagesets);
+	kfree(pages);
+	kfree(tidlist);
+	return ret > 0 ? 0 : ret;
+}
+
+int hfi1_user_exp_rcv_clear(struct file *fp, struct hfi1_tid_info *tinfo)
+{
+	int ret = 0;
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	u32 *tidinfo;
+	unsigned tididx;
+
+	tidinfo = kcalloc(tinfo->tidcnt, sizeof(*tidinfo), GFP_KERNEL);
+	if (!tidinfo)
+		return -ENOMEM;
+
+	if (copy_from_user(tidinfo, (void __user *)(unsigned long)
+			   tinfo->tidlist, sizeof(tidinfo[0]) *
+			   tinfo->tidcnt)) {
+		ret = -EFAULT;
+		goto done;
+	}
+
+	mutex_lock(&uctxt->exp_lock);
+	for (tididx = 0; tididx < tinfo->tidcnt; tididx++) {
+		ret = unprogram_rcvarray(fp, tidinfo[tididx], NULL);
+		if (ret) {
+			hfi1_cdbg(TID, "Failed to unprogram rcv array %d",
+				  ret);
+			break;
+		}
+	}
+	spin_lock(&fd->tid_lock);
+	fd->tid_used -= tididx;
+	spin_unlock(&fd->tid_lock);
+	tinfo->tidcnt = tididx;
+	mutex_unlock(&uctxt->exp_lock);
+done:
+	kfree(tidinfo);
+	return ret;
+}
+
+int hfi1_user_exp_rcv_invalid(struct file *fp, struct hfi1_tid_info *tinfo)
+{
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	unsigned long *ev = uctxt->dd->events +
+		(((uctxt->ctxt - uctxt->dd->first_user_ctxt) *
+		  HFI1_MAX_SHARED_CTXTS) + fd->subctxt);
+	u32 *array;
+	int ret = 0;
+
+	if (!fd->invalid_tids)
+		return -EINVAL;
+
+	/*
+	 * copy_to_user() can sleep, which will leave the invalid_lock
+	 * locked and cause the MMU notifier to be blocked on the lock
+	 * for a long time.
+	 * Copy the data to a local buffer so we can release the lock.
+	 */
+	array = kcalloc(uctxt->expected_count, sizeof(*array), GFP_KERNEL);
+	if (!array)
+		return -EFAULT;
+
+	spin_lock(&fd->invalid_lock);
+	if (fd->invalid_tid_idx) {
+		memcpy(array, fd->invalid_tids, sizeof(*array) *
+		       fd->invalid_tid_idx);
+		memset(fd->invalid_tids, 0, sizeof(*fd->invalid_tids) *
+		       fd->invalid_tid_idx);
+		tinfo->tidcnt = fd->invalid_tid_idx;
+		fd->invalid_tid_idx = 0;
+		/*
+		 * Reset the user flag while still holding the lock.
+		 * Otherwise, PSM can miss events.
+		 */
+		clear_bit(_HFI1_EVENT_TID_MMU_NOTIFY_BIT, ev);
+	} else {
+		tinfo->tidcnt = 0;
+	}
+	spin_unlock(&fd->invalid_lock);
+
+	if (tinfo->tidcnt) {
+		if (copy_to_user((void __user *)tinfo->tidlist,
+				 array, sizeof(*array) * tinfo->tidcnt))
+			ret = -EFAULT;
+	}
+	kfree(array);
+
+	return ret;
+}
+
+static u32 find_phys_blocks(struct page **pages, unsigned npages,
+			    struct tid_pageset *list)
+{
+	unsigned pagecount, pageidx, setcount = 0, i;
+	unsigned long pfn, this_pfn;
+
+	if (!npages)
+		return 0;
+
+	/*
+	 * Look for sets of physically contiguous pages in the user buffer.
+	 * This will allow us to optimize Expected RcvArray entry usage by
+	 * using the bigger supported sizes.
+	 */
+	pfn = page_to_pfn(pages[0]);
+	for (pageidx = 0, pagecount = 1, i = 1; i <= npages; i++) {
+		this_pfn = i < npages ? page_to_pfn(pages[i]) : 0;
+
+		/*
+		 * If the pfn's are not sequential, pages are not physically
+		 * contiguous.
+		 */
+		if (this_pfn != ++pfn) {
+			/*
+			 * At this point we have to loop over the set of
+			 * physically contiguous pages and break them down it
+			 * sizes supported by the HW.
+			 * There are two main constraints:
+			 *     1. The max buffer size is MAX_EXPECTED_BUFFER.
+			 *        If the total set size is bigger than that
+			 *        program only a MAX_EXPECTED_BUFFER chunk.
+			 *     2. The buffer size has to be a power of two. If
+			 *        it is not, round down to the closes power of
+			 *        2 and program that size.
+			 */
+			while (pagecount) {
+				int maxpages = pagecount;
+				u32 bufsize = pagecount * PAGE_SIZE;
+
+				if (bufsize > MAX_EXPECTED_BUFFER)
+					maxpages =
+						MAX_EXPECTED_BUFFER >>
+						PAGE_SHIFT;
+				else if (!is_power_of_2(bufsize))
+					maxpages =
+						rounddown_pow_of_two(bufsize) >>
+						PAGE_SHIFT;
+
+				list[setcount].idx = pageidx;
+				list[setcount].count = maxpages;
+				pagecount -= maxpages;
+				pageidx += maxpages;
+				setcount++;
+			}
+			pageidx = i;
+			pagecount = 1;
+			pfn = this_pfn;
+		} else {
+			pagecount++;
+		}
+	}
+	return setcount;
+}
+
+/**
+ * program_rcvarray() - program an RcvArray group with receive buffers
+ * @fp: file pointer
+ * @vaddr: starting user virtual address
+ * @grp: RcvArray group
+ * @sets: array of struct tid_pageset holding information on physically
+ *        contiguous chunks from the user buffer
+ * @start: starting index into sets array
+ * @count: number of struct tid_pageset's to program
+ * @pages: an array of struct page * for the user buffer
+ * @tidlist: the array of u32 elements when the information about the
+ *           programmed RcvArray entries is to be encoded.
+ * @tididx: starting offset into tidlist
+ * @pmapped: (output parameter) number of pages programmed into the RcvArray
+ *           entries.
+ *
+ * This function will program up to 'count' number of RcvArray entries from the
+ * group 'grp'. To make best use of write-combining writes, the function will
+ * perform writes to the unused RcvArray entries which will be ignored by the
+ * HW. Each RcvArray entry will be programmed with a physically contiguous
+ * buffer chunk from the user's virtual buffer.
+ *
+ * Return:
+ * -EINVAL if the requested count is larger than the size of the group,
+ * -ENOMEM or -EFAULT on error from set_rcvarray_entry(), or
+ * number of RcvArray entries programmed.
+ */
+static int program_rcvarray(struct file *fp, unsigned long vaddr,
+			    struct tid_group *grp,
+			    struct tid_pageset *sets,
+			    unsigned start, u16 count, struct page **pages,
+			    u32 *tidlist, unsigned *tididx, unsigned *pmapped)
+{
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_devdata *dd = uctxt->dd;
+	u16 idx;
+	u32 tidinfo = 0, rcventry, useidx = 0;
+	int mapped = 0;
+
+	/* Count should never be larger than the group size */
+	if (count > grp->size)
+		return -EINVAL;
+
+	/* Find the first unused entry in the group */
+	for (idx = 0; idx < grp->size; idx++) {
+		if (!(grp->map & (1 << idx))) {
+			useidx = idx;
+			break;
+		}
+		rcv_array_wc_fill(dd, grp->base + idx);
+	}
+
+	idx = 0;
+	while (idx < count) {
+		u16 npages, pageidx, setidx = start + idx;
+		int ret = 0;
+
+		/*
+		 * If this entry in the group is used, move to the next one.
+		 * If we go past the end of the group, exit the loop.
+		 */
+		if (useidx >= grp->size) {
+			break;
+		} else if (grp->map & (1 << useidx)) {
+			rcv_array_wc_fill(dd, grp->base + useidx);
+			useidx++;
+			continue;
+		}
+
+		rcventry = grp->base + useidx;
+		npages = sets[setidx].count;
+		pageidx = sets[setidx].idx;
+
+		ret = set_rcvarray_entry(fp, vaddr + (pageidx * PAGE_SIZE),
+					 rcventry, grp, pages + pageidx,
+					 npages);
+		if (ret)
+			return ret;
+		mapped += npages;
+
+		tidinfo = rcventry2tidinfo(rcventry - uctxt->expected_base) |
+			EXP_TID_SET(LEN, npages);
+		tidlist[(*tididx)++] = tidinfo;
+		grp->used++;
+		grp->map |= 1 << useidx++;
+		idx++;
+	}
+
+	/* Fill the rest of the group with "blank" writes */
+	for (; useidx < grp->size; useidx++)
+		rcv_array_wc_fill(dd, grp->base + useidx);
+	*pmapped = mapped;
+	return idx;
+}
+
+static int set_rcvarray_entry(struct file *fp, unsigned long vaddr,
+			      u32 rcventry, struct tid_group *grp,
+			      struct page **pages, unsigned npages)
+{
+	int ret;
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct tid_rb_node *node;
+	struct hfi1_devdata *dd = uctxt->dd;
+	struct rb_root *root = &fd->tid_rb_root;
+	dma_addr_t phys;
+
+	/*
+	 * Allocate the node first so we can handle a potential
+	 * failure before we've programmed anything.
+	 */
+	node = kzalloc(sizeof(*node) + (sizeof(struct page *) * npages),
+		       GFP_KERNEL);
+	if (!node)
+		return -ENOMEM;
+
+	phys = pci_map_single(dd->pcidev,
+			      __va(page_to_phys(pages[0])),
+			      npages * PAGE_SIZE, PCI_DMA_FROMDEVICE);
+	if (dma_mapping_error(&dd->pcidev->dev, phys)) {
+		dd_dev_err(dd, "Failed to DMA map Exp Rcv pages 0x%llx\n",
+			   phys);
+		kfree(node);
+		return -EFAULT;
+	}
+
+	node->mmu.addr = vaddr;
+	node->mmu.len = npages * PAGE_SIZE;
+	node->phys = page_to_phys(pages[0]);
+	node->npages = npages;
+	node->rcventry = rcventry;
+	node->dma_addr = phys;
+	node->grp = grp;
+	node->freed = false;
+	memcpy(node->pages, pages, sizeof(struct page *) * npages);
+
+	if (HFI1_CAP_IS_USET(TID_UNMAP))
+		ret = mmu_rb_insert(root, &node->mmu);
+	else
+		ret = hfi1_mmu_rb_insert(root, &node->mmu);
+
+	if (ret) {
+		hfi1_cdbg(TID, "Failed to insert RB node %u 0x%lx, 0x%lx %d",
+			  node->rcventry, node->mmu.addr, node->phys, ret);
+		pci_unmap_single(dd->pcidev, phys, npages * PAGE_SIZE,
+				 PCI_DMA_FROMDEVICE);
+		kfree(node);
+		return -EFAULT;
+	}
+	hfi1_put_tid(dd, rcventry, PT_EXPECTED, phys, ilog2(npages) + 1);
+	trace_hfi1_exp_tid_reg(uctxt->ctxt, fd->subctxt, rcventry, npages,
+			       node->mmu.addr, node->phys, phys);
+	return 0;
+}
+
+static int unprogram_rcvarray(struct file *fp, u32 tidinfo,
+			      struct tid_group **grp)
+{
+	struct hfi1_filedata *fd = fp->private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_devdata *dd = uctxt->dd;
+	struct tid_rb_node *node;
+	u8 tidctrl = EXP_TID_GET(tidinfo, CTRL);
+	u32 tididx = EXP_TID_GET(tidinfo, IDX) << 1, rcventry;
+
+	if (tididx >= uctxt->expected_count) {
+		dd_dev_err(dd, "Invalid RcvArray entry (%u) index for ctxt %u\n",
+			   tididx, uctxt->ctxt);
+		return -EINVAL;
+	}
+
+	if (tidctrl == 0x3)
+		return -EINVAL;
+
+	rcventry = tididx + (tidctrl - 1);
+
+	node = fd->entry_to_rb[rcventry];
+	if (!node || node->rcventry != (uctxt->expected_base + rcventry))
+		return -EBADF;
+	if (HFI1_CAP_IS_USET(TID_UNMAP))
+		mmu_rb_remove(&fd->tid_rb_root, &node->mmu, false);
+	else
+		hfi1_mmu_rb_remove(&fd->tid_rb_root, &node->mmu);
+
+	if (grp)
+		*grp = node->grp;
+	clear_tid_node(fd, fd->subctxt, node);
+	return 0;
+}
+
+static void clear_tid_node(struct hfi1_filedata *fd, u16 subctxt,
+			   struct tid_rb_node *node)
+{
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_devdata *dd = uctxt->dd;
+
+	trace_hfi1_exp_tid_unreg(uctxt->ctxt, fd->subctxt, node->rcventry,
+				 node->npages, node->mmu.addr, node->phys,
+				 node->dma_addr);
+
+	hfi1_put_tid(dd, node->rcventry, PT_INVALID, 0, 0);
+	/*
+	 * Make sure device has seen the write before we unpin the
+	 * pages.
+	 */
+	flush_wc();
+
+	pci_unmap_single(dd->pcidev, node->dma_addr, node->mmu.len,
+			 PCI_DMA_FROMDEVICE);
+	hfi1_release_user_pages(current->mm, node->pages, node->npages, true);
+	fd->tid_n_pinned -= node->npages;
+
+	node->grp->used--;
+	node->grp->map &= ~(1 << (node->rcventry - node->grp->base));
+
+	if (node->grp->used == node->grp->size - 1)
+		tid_group_move(node->grp, &uctxt->tid_full_list,
+			       &uctxt->tid_used_list);
+	else if (!node->grp->used)
+		tid_group_move(node->grp, &uctxt->tid_used_list,
+			       &uctxt->tid_group_list);
+	kfree(node);
+}
+
+static void unlock_exp_tids(struct hfi1_ctxtdata *uctxt,
+			    struct exp_tid_set *set, struct rb_root *root)
+{
+	struct tid_group *grp, *ptr;
+	struct hfi1_filedata *fd = container_of(root, struct hfi1_filedata,
+						tid_rb_root);
+	int i;
+
+	list_for_each_entry_safe(grp, ptr, &set->list, list) {
+		list_del_init(&grp->list);
+
+		for (i = 0; i < grp->size; i++) {
+			if (grp->map & (1 << i)) {
+				u16 rcventry = grp->base + i;
+				struct tid_rb_node *node;
+
+				node = fd->entry_to_rb[rcventry -
+							  uctxt->expected_base];
+				if (!node || node->rcventry != rcventry)
+					continue;
+				if (HFI1_CAP_IS_USET(TID_UNMAP))
+					mmu_rb_remove(&fd->tid_rb_root,
+						      &node->mmu, false);
+				else
+					hfi1_mmu_rb_remove(&fd->tid_rb_root,
+							   &node->mmu);
+				clear_tid_node(fd, -1, node);
+			}
+		}
+	}
+}
+
+static int mmu_rb_invalidate(struct rb_root *root, struct mmu_rb_node *mnode)
+{
+	struct hfi1_filedata *fdata =
+		container_of(root, struct hfi1_filedata, tid_rb_root);
+	struct hfi1_ctxtdata *uctxt = fdata->uctxt;
+	struct tid_rb_node *node =
+		container_of(mnode, struct tid_rb_node, mmu);
+
+	if (node->freed)
+		return 0;
+
+	trace_hfi1_exp_tid_inval(uctxt->ctxt, fdata->subctxt, node->mmu.addr,
+				 node->rcventry, node->npages, node->dma_addr);
+	node->freed = true;
+
+	spin_lock(&fdata->invalid_lock);
+	if (fdata->invalid_tid_idx < uctxt->expected_count) {
+		fdata->invalid_tids[fdata->invalid_tid_idx] =
+			rcventry2tidinfo(node->rcventry - uctxt->expected_base);
+		fdata->invalid_tids[fdata->invalid_tid_idx] |=
+			EXP_TID_SET(LEN, node->npages);
+		if (!fdata->invalid_tid_idx) {
+			unsigned long *ev;
+
+			/*
+			 * hfi1_set_uevent_bits() sets a user event flag
+			 * for all processes. Because calling into the
+			 * driver to process TID cache invalidations is
+			 * expensive and TID cache invalidations are
+			 * handled on a per-process basis, we can
+			 * optimize this to set the flag only for the
+			 * process in question.
+			 */
+			ev = uctxt->dd->events +
+				(((uctxt->ctxt - uctxt->dd->first_user_ctxt) *
+				  HFI1_MAX_SHARED_CTXTS) + fdata->subctxt);
+			set_bit(_HFI1_EVENT_TID_MMU_NOTIFY_BIT, ev);
+		}
+		fdata->invalid_tid_idx++;
+	}
+	spin_unlock(&fdata->invalid_lock);
+	return 0;
+}
+
+static int mmu_rb_insert(struct rb_root *root, struct mmu_rb_node *node)
+{
+	struct hfi1_filedata *fdata =
+		container_of(root, struct hfi1_filedata, tid_rb_root);
+	struct tid_rb_node *tnode =
+		container_of(node, struct tid_rb_node, mmu);
+	u32 base = fdata->uctxt->expected_base;
+
+	fdata->entry_to_rb[tnode->rcventry - base] = tnode;
+	return 0;
+}
+
+static void mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node,
+			  bool notifier)
+{
+	struct hfi1_filedata *fdata =
+		container_of(root, struct hfi1_filedata, tid_rb_root);
+	struct tid_rb_node *tnode =
+		container_of(node, struct tid_rb_node, mmu);
+	u32 base = fdata->uctxt->expected_base;
+
+	fdata->entry_to_rb[tnode->rcventry - base] = NULL;
+}
diff --git a/drivers/staging/rdma/hfi1/user_exp_rcv.h b/drivers/staging/rdma/hfi1/user_exp_rcv.h
index 4f4876e1d353..9bc8d9fba87e 100644
--- a/drivers/staging/rdma/hfi1/user_exp_rcv.h
+++ b/drivers/staging/rdma/hfi1/user_exp_rcv.h
@@ -1,14 +1,13 @@
 #ifndef _HFI1_USER_EXP_RCV_H
 #define _HFI1_USER_EXP_RCV_H
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -20,8 +19,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -50,6 +47,8 @@
  *
  */
 
+#include "hfi.h"
+
 #define EXP_TID_TIDLEN_MASK   0x7FFULL
 #define EXP_TID_TIDLEN_SHIFT  0
 #define EXP_TID_TIDCTRL_MASK  0x3ULL
@@ -71,4 +70,10 @@
 		(tid) |= EXP_TID_SET(field, (value));			\
 	} while (0)
 
+int hfi1_user_exp_rcv_init(struct file *);
+int hfi1_user_exp_rcv_free(struct hfi1_filedata *);
+int hfi1_user_exp_rcv_setup(struct file *, struct hfi1_tid_info *);
+int hfi1_user_exp_rcv_clear(struct file *, struct hfi1_tid_info *);
+int hfi1_user_exp_rcv_invalid(struct file *, struct hfi1_tid_info *);
+
 #endif /* _HFI1_USER_EXP_RCV_H */
diff --git a/drivers/staging/rdma/hfi1/user_pages.c b/drivers/staging/rdma/hfi1/user_pages.c
index 8ebfe9ee0d76..88e10b5f55f1 100644
--- a/drivers/staging/rdma/hfi1/user_pages.c
+++ b/drivers/staging/rdma/hfi1/user_pages.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -51,32 +48,62 @@
 #include <linux/mm.h>
 #include <linux/sched.h>
 #include <linux/device.h>
+#include <linux/module.h>
 
 #include "hfi.h"
 
-/**
- * hfi1_map_page - a safety wrapper around pci_map_page()
+static unsigned long cache_size = 256;
+module_param(cache_size, ulong, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(cache_size, "Send and receive side cache size limit (in MB)");
+
+/*
+ * Determine whether the caller can pin pages.
+ *
+ * This function should be used in the implementation of buffer caches.
+ * The cache implementation should call this function prior to attempting
+ * to pin buffer pages in order to determine whether they should do so.
+ * The function computes cache limits based on the configured ulimit and
+ * cache size. Use of this function is especially important for caches
+ * which are not limited in any other way (e.g. by HW resources) and, thus,
+ * could keeping caching buffers.
  *
  */
-dma_addr_t hfi1_map_page(struct pci_dev *hwdev, struct page *page,
-			 unsigned long offset, size_t size, int direction)
+bool hfi1_can_pin_pages(struct hfi1_devdata *dd, u32 nlocked, u32 npages)
 {
-	return pci_map_page(hwdev, page, offset, size, direction);
-}
-
-int hfi1_acquire_user_pages(unsigned long vaddr, size_t npages, bool writable,
-			    struct page **pages)
-{
-	unsigned long pinned, lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
+	unsigned long ulimit = rlimit(RLIMIT_MEMLOCK), pinned, cache_limit,
+		size = (cache_size * (1UL << 20)); /* convert to bytes */
+	unsigned usr_ctxts = dd->num_rcv_contexts - dd->first_user_ctxt;
 	bool can_lock = capable(CAP_IPC_LOCK);
-	int ret;
+
+	/*
+	 * Calculate per-cache size. The calculation below uses only a quarter
+	 * of the available per-context limit. This leaves space for other
+	 * pinning. Should we worry about shared ctxts?
+	 */
+	cache_limit = (ulimit / usr_ctxts) / 4;
+
+	/* If ulimit isn't set to "unlimited" and is smaller than cache_size. */
+	if (ulimit != (-1UL) && size > cache_limit)
+		size = cache_limit;
+
+	/* Convert to number of pages */
+	size = DIV_ROUND_UP(size, PAGE_SIZE);
 
 	down_read(&current->mm->mmap_sem);
 	pinned = current->mm->pinned_vm;
 	up_read(&current->mm->mmap_sem);
 
-	if (pinned + npages > lock_limit && !can_lock)
-		return -ENOMEM;
+	/* First, check the absolute limit against all pinned pages. */
+	if (pinned + npages >= ulimit && !can_lock)
+		return false;
+
+	return ((nlocked + npages) <= size) || can_lock;
+}
+
+int hfi1_acquire_user_pages(unsigned long vaddr, size_t npages, bool writable,
+			    struct page **pages)
+{
+	int ret;
 
 	ret = get_user_pages_fast(vaddr, npages, writable, pages);
 	if (ret < 0)
@@ -89,7 +116,8 @@ int hfi1_acquire_user_pages(unsigned long vaddr, size_t npages, bool writable,
 	return ret;
 }
 
-void hfi1_release_user_pages(struct page **p, size_t npages, bool dirty)
+void hfi1_release_user_pages(struct mm_struct *mm, struct page **p,
+			     size_t npages, bool dirty)
 {
 	size_t i;
 
@@ -99,9 +127,9 @@ void hfi1_release_user_pages(struct page **p, size_t npages, bool dirty)
 		put_page(p[i]);
 	}
 
-	if (current->mm) { /* during close after signal, mm can be NULL */
-		down_write(&current->mm->mmap_sem);
-		current->mm->pinned_vm -= npages;
-		up_write(&current->mm->mmap_sem);
+	if (mm) { /* during close after signal, mm can be NULL */
+		down_write(&mm->mmap_sem);
+		mm->pinned_vm -= npages;
+		up_write(&mm->mmap_sem);
 	}
 }
diff --git a/drivers/staging/rdma/hfi1/user_sdma.c b/drivers/staging/rdma/hfi1/user_sdma.c
index 9d4f5d6aaf33..ab6b6a42000f 100644
--- a/drivers/staging/rdma/hfi1/user_sdma.c
+++ b/drivers/staging/rdma/hfi1/user_sdma.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -70,6 +67,7 @@
 #include "verbs.h"  /* for the headers */
 #include "common.h" /* for struct hfi1_tid_info */
 #include "trace.h"
+#include "mmu_rb.h"
 
 static uint hfi1_sdma_comp_ring_size = 128;
 module_param_named(sdma_comp_size, hfi1_sdma_comp_ring_size, uint, S_IRUGO);
@@ -146,7 +144,6 @@ MODULE_PARM_DESC(sdma_comp_size, "Size of User SDMA completion ring. Default: 12
 
 /* Last packet in the request */
 #define TXREQ_FLAGS_REQ_LAST_PKT BIT(0)
-#define TXREQ_FLAGS_IOVEC_LAST_PKT BIT(0)
 
 #define SDMA_REQ_IN_USE     0
 #define SDMA_REQ_FOR_THREAD 1
@@ -170,16 +167,28 @@ static unsigned initial_pkt_count = 8;
 #define SDMA_IOWAIT_TIMEOUT 1000 /* in milliseconds */
 
 struct user_sdma_iovec {
+	struct list_head list;
 	struct iovec iov;
 	/* number of pages in this vector */
 	unsigned npages;
 	/* array of pinned pages for this vector */
 	struct page **pages;
-	/* offset into the virtual address space of the vector at
-	 * which we last left off. */
+	/*
+	 * offset into the virtual address space of the vector at
+	 * which we last left off.
+	 */
 	u64 offset;
 };
 
+struct sdma_mmu_node {
+	struct mmu_rb_node rb;
+	struct list_head list;
+	struct hfi1_user_sdma_pkt_q *pq;
+	atomic_t refcount;
+	struct page **pages;
+	unsigned npages;
+};
+
 struct user_sdma_request {
 	struct sdma_req_info info;
 	struct hfi1_user_sdma_pkt_q *pq;
@@ -213,15 +222,6 @@ struct user_sdma_request {
 	 */
 	u8 omfactor;
 	/*
-	 * pointer to the user's mm_struct. We are going to
-	 * get a reference to it so it doesn't get freed
-	 * since we might not be in process context when we
-	 * are processing the iov's.
-	 * Using this mm_struct, we can get vma based on the
-	 * iov's address (find_vma()).
-	 */
-	struct mm_struct *user_mm;
-	/*
 	 * We copy the iovs for this request (based on
 	 * info.iovcnt). These are only the data vectors
 	 */
@@ -238,13 +238,12 @@ struct user_sdma_request {
 	u16 tididx;
 	u32 sent;
 	u64 seqnum;
+	u64 seqcomp;
+	u64 seqsubmitted;
 	struct list_head txps;
-	spinlock_t txcmp_lock;  /* protect txcmp list */
-	struct list_head txcmp;
 	unsigned long flags;
 	/* status of the last txreq completed */
 	int status;
-	struct work_struct worker;
 };
 
 /*
@@ -259,11 +258,6 @@ struct user_sdma_txreq {
 	struct sdma_txreq txreq;
 	struct list_head list;
 	struct user_sdma_request *req;
-	struct {
-		struct user_sdma_iovec *vec;
-		u8 flags;
-	} iovecs[3];
-	int idx;
 	u16 flags;
 	unsigned busycount;
 	u64 seqnum;
@@ -279,21 +273,21 @@ struct user_sdma_txreq {
 
 static int user_sdma_send_pkts(struct user_sdma_request *, unsigned);
 static int num_user_pages(const struct iovec *);
-static void user_sdma_txreq_cb(struct sdma_txreq *, int, int);
-static void user_sdma_delayed_completion(struct work_struct *);
-static void user_sdma_free_request(struct user_sdma_request *);
+static void user_sdma_txreq_cb(struct sdma_txreq *, int);
+static inline void pq_update(struct hfi1_user_sdma_pkt_q *);
+static void user_sdma_free_request(struct user_sdma_request *, bool);
 static int pin_vector_pages(struct user_sdma_request *,
 			    struct user_sdma_iovec *);
-static void unpin_vector_pages(struct user_sdma_request *,
-			       struct user_sdma_iovec *);
+static void unpin_vector_pages(struct mm_struct *, struct page **, unsigned);
 static int check_header_template(struct user_sdma_request *,
 				 struct hfi1_pkt_header *, u32, u32);
 static int set_txreq_header(struct user_sdma_request *,
 			    struct user_sdma_txreq *, u32);
 static int set_txreq_header_ahg(struct user_sdma_request *,
 				struct user_sdma_txreq *, u32);
-static inline void set_comp_state(struct user_sdma_request *,
-					enum hfi1_sdma_comp_state, int);
+static inline void set_comp_state(struct hfi1_user_sdma_pkt_q *,
+				  struct hfi1_user_sdma_comp_q *,
+				  u16, enum hfi1_sdma_comp_state, int);
 static inline u32 set_pkt_bth_psn(__be32, u8, u32);
 static inline u32 get_lrh_len(struct hfi1_pkt_header, u32 len);
 
@@ -303,6 +297,17 @@ static int defer_packet_queue(
 	struct sdma_txreq *,
 	unsigned seq);
 static void activate_packet_queue(struct iowait *, int);
+static bool sdma_rb_filter(struct mmu_rb_node *, unsigned long, unsigned long);
+static int sdma_rb_insert(struct rb_root *, struct mmu_rb_node *);
+static void sdma_rb_remove(struct rb_root *, struct mmu_rb_node *, bool);
+static int sdma_rb_invalidate(struct rb_root *, struct mmu_rb_node *);
+
+static struct mmu_rb_ops sdma_rb_ops = {
+	.filter = sdma_rb_filter,
+	.insert = sdma_rb_insert,
+	.remove = sdma_rb_remove,
+	.invalidate = sdma_rb_invalidate
+};
 
 static int defer_packet_queue(
 	struct sdma_engine *sde,
@@ -380,7 +385,7 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, struct file *fp)
 		goto pq_nomem;
 
 	memsize = sizeof(*pq->reqs) * hfi1_sdma_comp_ring_size;
-	pq->reqs = kmalloc(memsize, GFP_KERNEL);
+	pq->reqs = kzalloc(memsize, GFP_KERNEL);
 	if (!pq->reqs)
 		goto pq_reqs_nomem;
 
@@ -392,9 +397,12 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, struct file *fp)
 	pq->state = SDMA_PKT_Q_INACTIVE;
 	atomic_set(&pq->n_reqs, 0);
 	init_waitqueue_head(&pq->wait);
+	pq->sdma_rb_root = RB_ROOT;
+	INIT_LIST_HEAD(&pq->evict);
+	spin_lock_init(&pq->evict_lock);
 
 	iowait_init(&pq->busy, 0, NULL, defer_packet_queue,
-		    activate_packet_queue);
+		    activate_packet_queue, NULL);
 	pq->reqidx = 0;
 	snprintf(buf, 64, "txreq-kmem-cache-%u-%u-%u", dd->unit, uctxt->ctxt,
 		 fd->subctxt);
@@ -421,6 +429,12 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, struct file *fp)
 	cq->nentries = hfi1_sdma_comp_ring_size;
 	fd->cq = cq;
 
+	ret = hfi1_mmu_rb_register(&pq->sdma_rb_root, &sdma_rb_ops);
+	if (ret) {
+		dd_dev_err(dd, "Failed to register with MMU %d", ret);
+		goto done;
+	}
+
 	spin_lock_irqsave(&uctxt->sdma_qlock, flags);
 	list_add(&pq->list, &uctxt->sdma_queues);
 	spin_unlock_irqrestore(&uctxt->sdma_qlock, flags);
@@ -450,6 +464,7 @@ int hfi1_user_sdma_free_queues(struct hfi1_filedata *fd)
 	hfi1_cdbg(SDMA, "[%u:%u:%u] Freeing user SDMA queues", uctxt->dd->unit,
 		  uctxt->ctxt, fd->subctxt);
 	pq = fd->pq;
+	hfi1_mmu_rb_unregister(&pq->sdma_rb_root);
 	if (pq) {
 		spin_lock_irqsave(&uctxt->sdma_qlock, flags);
 		if (!list_empty(&pq->list))
@@ -476,7 +491,7 @@ int hfi1_user_sdma_free_queues(struct hfi1_filedata *fd)
 int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 				   unsigned long dim, unsigned long *count)
 {
-	int ret = 0, i = 0, sent;
+	int ret = 0, i = 0;
 	struct hfi1_filedata *fd = fp->private_data;
 	struct hfi1_ctxtdata *uctxt = fd->uctxt;
 	struct hfi1_user_sdma_pkt_q *pq = fd->pq;
@@ -502,9 +517,11 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 			  dd->unit, uctxt->ctxt, fd->subctxt, ret);
 		return -EFAULT;
 	}
+
 	trace_hfi1_sdma_user_reqinfo(dd, uctxt->ctxt, fd->subctxt,
 				     (u16 *)&info);
-	if (cq->comps[info.comp_idx].status == QUEUED) {
+	if (cq->comps[info.comp_idx].status == QUEUED ||
+	    test_bit(SDMA_REQ_IN_USE, &pq->reqs[info.comp_idx].flags)) {
 		hfi1_cdbg(SDMA, "[%u:%u:%u] Entry %u is in QUEUED state",
 			  dd->unit, uctxt->ctxt, fd->subctxt,
 			  info.comp_idx);
@@ -531,10 +548,7 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 	req->cq = cq;
 	req->status = -1;
 	INIT_LIST_HEAD(&req->txps);
-	INIT_LIST_HEAD(&req->txcmp);
-	INIT_WORK(&req->worker, user_sdma_delayed_completion);
 
-	spin_lock_init(&req->txcmp_lock);
 	memcpy(&req->info, &info, sizeof(info));
 
 	if (req_opcode(info.ctrl) == EXPECTED)
@@ -593,8 +607,10 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 	}
 
 	req->koffset = le32_to_cpu(req->hdr.kdeth.swdata[6]);
-	/* Calculate the initial TID offset based on the values of
-	   KDETH.OFFSET and KDETH.OM that are passed in. */
+	/*
+	 * Calculate the initial TID offset based on the values of
+	 * KDETH.OFFSET and KDETH.OM that are passed in.
+	 */
 	req->tidoffset = KDETH_GET(req->hdr.kdeth.ver_tid_offset, OFFSET) *
 		(KDETH_GET(req->hdr.kdeth.ver_tid_offset, OM) ?
 		 KDETH_OM_LARGE : KDETH_OM_SMALL);
@@ -603,8 +619,13 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 
 	/* Save all the IO vector structures */
 	while (i < req->data_iovs) {
+		INIT_LIST_HEAD(&req->iovs[i].list);
 		memcpy(&req->iovs[i].iov, iovec + idx++, sizeof(struct iovec));
-		req->iovs[i].offset = 0;
+		ret = pin_vector_pages(req, &req->iovs[i]);
+		if (ret) {
+			req->status = ret;
+			goto free_req;
+		}
 		req->data_len += req->iovs[i++].iov.iov_len;
 	}
 	SDMA_DBG(req, "total data length %u", req->data_len);
@@ -668,52 +689,59 @@ int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
 		}
 	}
 
-	set_comp_state(req, QUEUED, 0);
+	set_comp_state(pq, cq, info.comp_idx, QUEUED, 0);
+	atomic_inc(&pq->n_reqs);
 	/* Send the first N packets in the request to buy us some time */
-	sent = user_sdma_send_pkts(req, pcount);
-	if (unlikely(sent < 0)) {
-		if (sent != -EBUSY) {
-			req->status = sent;
-			set_comp_state(req, ERROR, req->status);
-			return sent;
-		} else
-			sent = 0;
+	ret = user_sdma_send_pkts(req, pcount);
+	if (unlikely(ret < 0 && ret != -EBUSY)) {
+		req->status = ret;
+		goto free_req;
 	}
-	atomic_inc(&pq->n_reqs);
-	xchg(&pq->state, SDMA_PKT_Q_ACTIVE);
 
-	if (sent < req->info.npkts) {
-		/*
-		 * This is a somewhat blocking send implementation.
-		 * The driver will block the caller until all packets of the
-		 * request have been submitted to the SDMA engine. However, it
-		 * will not wait for send completions.
-		 */
-		while (!test_bit(SDMA_REQ_SEND_DONE, &req->flags)) {
-			ret = user_sdma_send_pkts(req, pcount);
-			if (ret < 0) {
-				if (ret != -EBUSY) {
-					req->status = ret;
-					return ret;
-				}
-				wait_event_interruptible_timeout(
-					pq->busy.wait_dma,
-					(pq->state == SDMA_PKT_Q_ACTIVE),
-					msecs_to_jiffies(
-						SDMA_IOWAIT_TIMEOUT));
+	/*
+	 * It is possible that the SDMA engine would have processed all the
+	 * submitted packets by the time we get here. Therefore, only set
+	 * packet queue state to ACTIVE if there are still uncompleted
+	 * requests.
+	 */
+	if (atomic_read(&pq->n_reqs))
+		xchg(&pq->state, SDMA_PKT_Q_ACTIVE);
+
+	/*
+	 * This is a somewhat blocking send implementation.
+	 * The driver will block the caller until all packets of the
+	 * request have been submitted to the SDMA engine. However, it
+	 * will not wait for send completions.
+	 */
+	while (!test_bit(SDMA_REQ_SEND_DONE, &req->flags)) {
+		ret = user_sdma_send_pkts(req, pcount);
+		if (ret < 0) {
+			if (ret != -EBUSY) {
+				req->status = ret;
+				set_bit(SDMA_REQ_DONE_ERROR, &req->flags);
+				if (ACCESS_ONCE(req->seqcomp) ==
+				    req->seqsubmitted - 1)
+					goto free_req;
+				return ret;
 			}
+			wait_event_interruptible_timeout(
+				pq->busy.wait_dma,
+				(pq->state == SDMA_PKT_Q_ACTIVE),
+				msecs_to_jiffies(
+					SDMA_IOWAIT_TIMEOUT));
 		}
-
 	}
 	*count += idx;
 	return 0;
 free_req:
-	user_sdma_free_request(req);
+	user_sdma_free_request(req, true);
+	pq_update(pq);
+	set_comp_state(pq, cq, info.comp_idx, ERROR, req->status);
 	return ret;
 }
 
 static inline u32 compute_data_length(struct user_sdma_request *req,
-					    struct user_sdma_txreq *tx)
+				      struct user_sdma_txreq *tx)
 {
 	/*
 	 * Determine the proper size of the packet data.
@@ -731,8 +759,10 @@ static inline u32 compute_data_length(struct user_sdma_request *req,
 	} else if (req_opcode(req->info.ctrl) == EXPECTED) {
 		u32 tidlen = EXP_TID_GET(req->tids[req->tididx], LEN) *
 			PAGE_SIZE;
-		/* Get the data length based on the remaining space in the
-		 * TID pair. */
+		/*
+		 * Get the data length based on the remaining space in the
+		 * TID pair.
+		 */
 		len = min(tidlen - req->tidoffset, (u32)req->info.fragsize);
 		/* If we've filled up the TID pair, move to the next one. */
 		if (unlikely(!len) && ++req->tididx < req->n_tids &&
@@ -742,12 +772,15 @@ static inline u32 compute_data_length(struct user_sdma_request *req,
 			req->tidoffset = 0;
 			len = min_t(u32, tidlen, req->info.fragsize);
 		}
-		/* Since the TID pairs map entire pages, make sure that we
+		/*
+		 * Since the TID pairs map entire pages, make sure that we
 		 * are not going to try to send more data that we have
-		 * remaining. */
+		 * remaining.
+		 */
 		len = min(len, req->data_len - req->sent);
-	} else
+	} else {
 		len = min(req->data_len - req->sent, (u32)req->info.fragsize);
+	}
 	SDMA_DBG(req, "Data Length = %u", len);
 	return len;
 }
@@ -810,9 +843,7 @@ static int user_sdma_send_pkts(struct user_sdma_request *req, unsigned maxpkts)
 		tx->flags = 0;
 		tx->req = req;
 		tx->busycount = 0;
-		tx->idx = -1;
 		INIT_LIST_HEAD(&tx->list);
-		memset(tx->iovecs, 0, sizeof(tx->iovecs));
 
 		if (req->seqnum == req->info.npkts - 1)
 			tx->flags |= TXREQ_FLAGS_REQ_LAST_PKT;
@@ -833,18 +864,6 @@ static int user_sdma_send_pkts(struct user_sdma_request *req, unsigned maxpkts)
 				WARN_ON(iovec->offset);
 			}
 
-			/*
-			 * This request might include only a header and no user
-			 * data, so pin pages only if there is data and it the
-			 * pages have not been pinned already.
-			 */
-			if (unlikely(!iovec->pages && iovec->iov.iov_len)) {
-				ret = pin_vector_pages(req, iovec);
-				if (ret)
-					goto free_tx;
-			}
-
-			tx->iovecs[++tx->idx].vec = iovec;
 			datalen = compute_data_length(req, tx);
 			if (!datalen) {
 				SDMA_DBG(req,
@@ -934,16 +953,8 @@ static int user_sdma_send_pkts(struct user_sdma_request *req, unsigned maxpkts)
 					      iovec->pages[pageidx],
 					      offset, len);
 			if (ret) {
-				int i;
-
 				SDMA_DBG(req, "SDMA txreq add page failed %d\n",
 					 ret);
-				/* Mark all assigned vectors as complete so they
-				 * are unpinned in the callback. */
-				for (i = tx->idx; i >= 0; i--) {
-					tx->iovecs[i].flags |=
-						TXREQ_FLAGS_IOVEC_LAST_PKT;
-				}
 				goto free_txreq;
 			}
 			iov_offset += len;
@@ -951,19 +962,10 @@ static int user_sdma_send_pkts(struct user_sdma_request *req, unsigned maxpkts)
 			data_sent += len;
 			if (unlikely(queued < datalen &&
 				     pageidx == iovec->npages &&
-				     req->iov_idx < req->data_iovs - 1 &&
-				     tx->idx < ARRAY_SIZE(tx->iovecs))) {
+				     req->iov_idx < req->data_iovs - 1)) {
 				iovec->offset += iov_offset;
-				tx->iovecs[tx->idx].flags |=
-					TXREQ_FLAGS_IOVEC_LAST_PKT;
 				iovec = &req->iovs[++req->iov_idx];
-				if (!iovec->pages) {
-					ret = pin_vector_pages(req, iovec);
-					if (ret)
-						goto free_txreq;
-				}
 				iov_offset = 0;
-				tx->iovecs[++tx->idx].vec = iovec;
 			}
 		}
 		/*
@@ -974,28 +976,21 @@ static int user_sdma_send_pkts(struct user_sdma_request *req, unsigned maxpkts)
 		if (req_opcode(req->info.ctrl) == EXPECTED)
 			req->tidoffset += datalen;
 		req->sent += data_sent;
-		if (req->data_len) {
-			tx->iovecs[tx->idx].vec->offset += iov_offset;
-			/* If we've reached the end of the io vector, mark it
-			 * so the callback can unpin the pages and free it. */
-			if (tx->iovecs[tx->idx].vec->offset ==
-			    tx->iovecs[tx->idx].vec->iov.iov_len)
-				tx->iovecs[tx->idx].flags |=
-					TXREQ_FLAGS_IOVEC_LAST_PKT;
-		}
-
+		if (req->data_len)
+			iovec->offset += iov_offset;
+		list_add_tail(&tx->txreq.list, &req->txps);
 		/*
 		 * It is important to increment this here as it is used to
 		 * generate the BTH.PSN and, therefore, can't be bulk-updated
 		 * outside of the loop.
 		 */
 		tx->seqnum = req->seqnum++;
-		list_add_tail(&tx->txreq.list, &req->txps);
 		npkts++;
 	}
 dosend:
 	ret = sdma_send_txlist(req->sde, &pq->busy, &req->txps);
-	if (list_empty(&req->txps))
+	if (list_empty(&req->txps)) {
+		req->seqsubmitted = req->seqnum;
 		if (req->seqnum == req->info.npkts) {
 			set_bit(SDMA_REQ_SEND_DONE, &req->flags);
 			/*
@@ -1007,6 +1002,10 @@ dosend:
 			if (test_bit(SDMA_REQ_HAVE_AHG, &req->flags))
 				sdma_ahg_free(req->sde, req->ahg_idx);
 		}
+	} else if (ret > 0) {
+		req->seqsubmitted += ret;
+		ret = 0;
+	}
 	return ret;
 
 free_txreq:
@@ -1021,7 +1020,7 @@ free_tx:
  */
 static inline int num_user_pages(const struct iovec *iov)
 {
-	const unsigned long addr  = (unsigned long) iov->iov_base;
+	const unsigned long addr  = (unsigned long)iov->iov_base;
 	const unsigned long len   = iov->iov_len;
 	const unsigned long spage = addr & PAGE_MASK;
 	const unsigned long epage = (addr + len - 1) & PAGE_MASK;
@@ -1029,64 +1028,129 @@ static inline int num_user_pages(const struct iovec *iov)
 	return 1 + ((epage - spage) >> PAGE_SHIFT);
 }
 
-static int pin_vector_pages(struct user_sdma_request *req,
-			    struct user_sdma_iovec *iovec) {
-	int pinned, npages;
+/* Caller must hold pq->evict_lock */
+static u32 sdma_cache_evict(struct hfi1_user_sdma_pkt_q *pq, u32 npages)
+{
+	u32 cleared = 0;
+	struct sdma_mmu_node *node, *ptr;
 
-	npages = num_user_pages(&iovec->iov);
-	iovec->pages = kcalloc(npages, sizeof(*iovec->pages), GFP_KERNEL);
-	if (!iovec->pages) {
-		SDMA_DBG(req, "Failed page array alloc");
-		return -ENOMEM;
+	list_for_each_entry_safe_reverse(node, ptr, &pq->evict, list) {
+		/* Make sure that no one is still using the node. */
+		if (!atomic_read(&node->refcount)) {
+			/*
+			 * Need to use the page count now as the remove callback
+			 * will free the node.
+			 */
+			cleared += node->npages;
+			spin_unlock(&pq->evict_lock);
+			hfi1_mmu_rb_remove(&pq->sdma_rb_root, &node->rb);
+			spin_lock(&pq->evict_lock);
+			if (cleared >= npages)
+				break;
+		}
 	}
+	return cleared;
+}
 
-	/*
-	 * Get a reference to the process's mm so we can use it when
-	 * unpinning the io vectors.
-	 */
-	req->pq->user_mm = get_task_mm(current);
+static int pin_vector_pages(struct user_sdma_request *req,
+			    struct user_sdma_iovec *iovec) {
+	int ret = 0, pinned, npages, cleared;
+	struct page **pages;
+	struct hfi1_user_sdma_pkt_q *pq = req->pq;
+	struct sdma_mmu_node *node = NULL;
+	struct mmu_rb_node *rb_node;
+
+	rb_node = hfi1_mmu_rb_search(&pq->sdma_rb_root,
+				     (unsigned long)iovec->iov.iov_base,
+				     iovec->iov.iov_len);
+	if (rb_node)
+		node = container_of(rb_node, struct sdma_mmu_node, rb);
+
+	if (!node) {
+		node = kzalloc(sizeof(*node), GFP_KERNEL);
+		if (!node)
+			return -ENOMEM;
 
-	pinned = hfi1_acquire_user_pages((unsigned long)iovec->iov.iov_base,
-					 npages, 0, iovec->pages);
+		node->rb.addr = (unsigned long)iovec->iov.iov_base;
+		node->rb.len = iovec->iov.iov_len;
+		node->pq = pq;
+		atomic_set(&node->refcount, 0);
+		INIT_LIST_HEAD(&node->list);
+	}
 
-	if (pinned < 0)
-		return pinned;
+	npages = num_user_pages(&iovec->iov);
+	if (node->npages < npages) {
+		pages = kcalloc(npages, sizeof(*pages), GFP_KERNEL);
+		if (!pages) {
+			SDMA_DBG(req, "Failed page array alloc");
+			ret = -ENOMEM;
+			goto bail;
+		}
+		memcpy(pages, node->pages, node->npages * sizeof(*pages));
+
+		npages -= node->npages;
+retry:
+		if (!hfi1_can_pin_pages(pq->dd, pq->n_locked, npages)) {
+			spin_lock(&pq->evict_lock);
+			cleared = sdma_cache_evict(pq, npages);
+			spin_unlock(&pq->evict_lock);
+			if (cleared >= npages)
+				goto retry;
+		}
+		pinned = hfi1_acquire_user_pages(
+			((unsigned long)iovec->iov.iov_base +
+			 (node->npages * PAGE_SIZE)), npages, 0,
+			pages + node->npages);
+		if (pinned < 0) {
+			kfree(pages);
+			ret = pinned;
+			goto bail;
+		}
+		if (pinned != npages) {
+			unpin_vector_pages(current->mm, pages, pinned);
+			ret = -EFAULT;
+			goto bail;
+		}
+		kfree(node->pages);
+		node->pages = pages;
+		node->npages += pinned;
+		npages = node->npages;
+		spin_lock(&pq->evict_lock);
+		if (!rb_node)
+			list_add(&node->list, &pq->evict);
+		else
+			list_move(&node->list, &pq->evict);
+		pq->n_locked += pinned;
+		spin_unlock(&pq->evict_lock);
+	}
+	iovec->pages = node->pages;
+	iovec->npages = npages;
 
-	iovec->npages = pinned;
-	if (pinned != npages) {
-		SDMA_DBG(req, "Failed to pin pages (%d/%u)", pinned, npages);
-		unpin_vector_pages(req, iovec);
-		return -EFAULT;
+	if (!rb_node) {
+		ret = hfi1_mmu_rb_insert(&req->pq->sdma_rb_root, &node->rb);
+		if (ret) {
+			spin_lock(&pq->evict_lock);
+			list_del(&node->list);
+			pq->n_locked -= node->npages;
+			spin_unlock(&pq->evict_lock);
+			ret = 0;
+			goto bail;
+		}
+	} else {
+		atomic_inc(&node->refcount);
 	}
 	return 0;
+bail:
+	if (!rb_node)
+		kfree(node);
+	return ret;
 }
 
-static void unpin_vector_pages(struct user_sdma_request *req,
-			       struct user_sdma_iovec *iovec)
+static void unpin_vector_pages(struct mm_struct *mm, struct page **pages,
+			       unsigned npages)
 {
-	/*
-	 * Unpinning is done through the workqueue so use the
-	 * process's mm if we have a reference to it.
-	 */
-	if ((current->flags & PF_KTHREAD) && req->pq->user_mm)
-		use_mm(req->pq->user_mm);
-
-	hfi1_release_user_pages(iovec->pages, iovec->npages, 0);
-
-	/*
-	 * Unuse the user's mm (see above) and release the
-	 * reference to it.
-	 */
-	if (req->pq->user_mm) {
-		if (current->flags & PF_KTHREAD)
-			unuse_mm(req->pq->user_mm);
-		mmput(req->pq->user_mm);
-	}
-
-	kfree(iovec->pages);
-	iovec->pages = NULL;
-	iovec->npages = 0;
-	iovec->offset = 0;
+	hfi1_release_user_pages(mm, pages, npages, 0);
+	kfree(pages);
 }
 
 static int check_header_template(struct user_sdma_request *req,
@@ -1209,7 +1273,6 @@ static int set_txreq_header(struct user_sdma_request *req,
 		if (ret)
 			return ret;
 		goto done;
-
 	}
 
 	hdr->bth[2] = cpu_to_be32(
@@ -1219,7 +1282,7 @@ static int set_txreq_header(struct user_sdma_request *req,
 
 	/* Set ACK request on last packet */
 	if (unlikely(tx->flags & TXREQ_FLAGS_REQ_LAST_PKT))
-		hdr->bth[2] |= cpu_to_be32(1UL<<31);
+		hdr->bth[2] |= cpu_to_be32(1UL << 31);
 
 	/* Set the new offset */
 	hdr->kdeth.swdata[6] = cpu_to_le32(req->koffset);
@@ -1233,8 +1296,10 @@ static int set_txreq_header(struct user_sdma_request *req,
 		if ((req->tidoffset) == (EXP_TID_GET(tidval, LEN) *
 					 PAGE_SIZE)) {
 			req->tidoffset = 0;
-			/* Since we don't copy all the TIDs, all at once,
-			 * we have to check again. */
+			/*
+			 * Since we don't copy all the TIDs, all at once,
+			 * we have to check again.
+			 */
 			if (++req->tididx > req->n_tids - 1 ||
 			    !req->tids[req->tididx]) {
 				return -EINVAL;
@@ -1315,8 +1380,10 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 		if ((req->tidoffset) == (EXP_TID_GET(tidval, LEN) *
 					 PAGE_SIZE)) {
 			req->tidoffset = 0;
-			/* Since we don't copy all the TIDs, all at once,
-			 * we have to check again. */
+			/*
+			 * Since we don't copy all the TIDs, all at once,
+			 * we have to check again.
+			 */
 			if (++req->tididx > req->n_tids - 1 ||
 			    !req->tids[req->tididx]) {
 				return -EINVAL;
@@ -1340,8 +1407,9 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 								INTR) >> 16);
 			val &= cpu_to_le16(~(1U << 13));
 			AHG_HEADER_SET(req->ahg, diff, 7, 16, 14, val);
-		} else
+		} else {
 			AHG_HEADER_SET(req->ahg, diff, 7, 16, 12, val);
+		}
 	}
 
 	trace_hfi1_sdma_user_header_ahg(pq->dd, pq->ctxt, pq->subctxt,
@@ -1356,113 +1424,62 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
  * tx request have been processed by the DMA engine. Called in
  * interrupt context.
  */
-static void user_sdma_txreq_cb(struct sdma_txreq *txreq, int status,
-			       int drain)
+static void user_sdma_txreq_cb(struct sdma_txreq *txreq, int status)
 {
 	struct user_sdma_txreq *tx =
 		container_of(txreq, struct user_sdma_txreq, txreq);
 	struct user_sdma_request *req;
-	bool defer;
-	int i;
+	struct hfi1_user_sdma_pkt_q *pq;
+	struct hfi1_user_sdma_comp_q *cq;
+	u16 idx;
 
 	if (!tx->req)
 		return;
 
 	req = tx->req;
-	/*
-	 * If this is the callback for the last packet of the request,
-	 * queue up the request for clean up.
-	 */
-	defer = (tx->seqnum == req->info.npkts - 1);
-
-	/*
-	 * If we have any io vectors associated with this txreq,
-	 * check whether they need to be 'freed'. We can't free them
-	 * here because the unpin function needs to be able to sleep.
-	 */
-	for (i = tx->idx; i >= 0; i--) {
-		if (tx->iovecs[i].flags & TXREQ_FLAGS_IOVEC_LAST_PKT) {
-			defer = true;
-			break;
-		}
-	}
+	pq = req->pq;
+	cq = req->cq;
 
-	req->status = status;
 	if (status != SDMA_TXREQ_S_OK) {
 		SDMA_DBG(req, "SDMA completion with error %d",
 			 status);
 		set_bit(SDMA_REQ_HAS_ERROR, &req->flags);
-		defer = true;
 	}
 
-	/*
-	 * Defer the clean up of the iovectors and the request until later
-	 * so it can be done outside of interrupt context.
-	 */
-	if (defer) {
-		spin_lock(&req->txcmp_lock);
-		list_add_tail(&tx->list, &req->txcmp);
-		spin_unlock(&req->txcmp_lock);
-		schedule_work(&req->worker);
+	req->seqcomp = tx->seqnum;
+	kmem_cache_free(pq->txreq_cache, tx);
+	tx = NULL;
+
+	idx = req->info.comp_idx;
+	if (req->status == -1 && status == SDMA_TXREQ_S_OK) {
+		if (req->seqcomp == req->info.npkts - 1) {
+			req->status = 0;
+			user_sdma_free_request(req, false);
+			pq_update(pq);
+			set_comp_state(pq, cq, idx, COMPLETE, 0);
+		}
 	} else {
-		kmem_cache_free(req->pq->txreq_cache, tx);
+		if (status != SDMA_TXREQ_S_OK)
+			req->status = status;
+		if (req->seqcomp == (ACCESS_ONCE(req->seqsubmitted) - 1) &&
+		    (test_bit(SDMA_REQ_SEND_DONE, &req->flags) ||
+		     test_bit(SDMA_REQ_DONE_ERROR, &req->flags))) {
+			user_sdma_free_request(req, false);
+			pq_update(pq);
+			set_comp_state(pq, cq, idx, ERROR, req->status);
+		}
 	}
 }
 
-static void user_sdma_delayed_completion(struct work_struct *work)
+static inline void pq_update(struct hfi1_user_sdma_pkt_q *pq)
 {
-	struct user_sdma_request *req =
-		container_of(work, struct user_sdma_request, worker);
-	struct hfi1_user_sdma_pkt_q *pq = req->pq;
-	struct user_sdma_txreq *tx = NULL;
-	unsigned long flags;
-	u64 seqnum;
-	int i;
-
-	while (1) {
-		spin_lock_irqsave(&req->txcmp_lock, flags);
-		if (!list_empty(&req->txcmp)) {
-			tx = list_first_entry(&req->txcmp,
-					      struct user_sdma_txreq, list);
-			list_del(&tx->list);
-		}
-		spin_unlock_irqrestore(&req->txcmp_lock, flags);
-		if (!tx)
-			break;
-
-		for (i = tx->idx; i >= 0; i--)
-			if (tx->iovecs[i].flags & TXREQ_FLAGS_IOVEC_LAST_PKT)
-				unpin_vector_pages(req, tx->iovecs[i].vec);
-
-		seqnum = tx->seqnum;
-		kmem_cache_free(pq->txreq_cache, tx);
-		tx = NULL;
-
-		if (req->status != SDMA_TXREQ_S_OK) {
-			if (seqnum == ACCESS_ONCE(req->seqnum) &&
-			    test_bit(SDMA_REQ_DONE_ERROR, &req->flags)) {
-				atomic_dec(&pq->n_reqs);
-				set_comp_state(req, ERROR, req->status);
-				user_sdma_free_request(req);
-				break;
-			}
-		} else {
-			if (seqnum == req->info.npkts - 1) {
-				atomic_dec(&pq->n_reqs);
-				set_comp_state(req, COMPLETE, 0);
-				user_sdma_free_request(req);
-				break;
-			}
-		}
-	}
-
-	if (!atomic_read(&pq->n_reqs)) {
+	if (atomic_dec_and_test(&pq->n_reqs)) {
 		xchg(&pq->state, SDMA_PKT_Q_INACTIVE);
 		wake_up(&pq->wait);
 	}
 }
 
-static void user_sdma_free_request(struct user_sdma_request *req)
+static void user_sdma_free_request(struct user_sdma_request *req, bool unpin)
 {
 	if (!list_empty(&req->txps)) {
 		struct sdma_txreq *t, *p;
@@ -1476,25 +1493,87 @@ static void user_sdma_free_request(struct user_sdma_request *req)
 		}
 	}
 	if (req->data_iovs) {
+		struct sdma_mmu_node *node;
+		struct mmu_rb_node *mnode;
 		int i;
 
-		for (i = 0; i < req->data_iovs; i++)
-			if (req->iovs[i].npages && req->iovs[i].pages)
-				unpin_vector_pages(req, &req->iovs[i]);
+		for (i = 0; i < req->data_iovs; i++) {
+			mnode = hfi1_mmu_rb_search(
+				&req->pq->sdma_rb_root,
+				(unsigned long)req->iovs[i].iov.iov_base,
+				req->iovs[i].iov.iov_len);
+			if (!mnode)
+				continue;
+
+			node = container_of(mnode, struct sdma_mmu_node, rb);
+			if (unpin)
+				hfi1_mmu_rb_remove(&req->pq->sdma_rb_root,
+						   &node->rb);
+			else
+				atomic_dec(&node->refcount);
+		}
 	}
 	kfree(req->tids);
 	clear_bit(SDMA_REQ_IN_USE, &req->flags);
 }
 
-static inline void set_comp_state(struct user_sdma_request *req,
-					enum hfi1_sdma_comp_state state,
-					int ret)
+static inline void set_comp_state(struct hfi1_user_sdma_pkt_q *pq,
+				  struct hfi1_user_sdma_comp_q *cq,
+				  u16 idx, enum hfi1_sdma_comp_state state,
+				  int ret)
 {
-	SDMA_DBG(req, "Setting completion status %u %d", state, ret);
-	req->cq->comps[req->info.comp_idx].status = state;
+	hfi1_cdbg(SDMA, "[%u:%u:%u:%u] Setting completion status %u %d",
+		  pq->dd->unit, pq->ctxt, pq->subctxt, idx, state, ret);
+	cq->comps[idx].status = state;
 	if (state == ERROR)
-		req->cq->comps[req->info.comp_idx].errcode = -ret;
-	trace_hfi1_sdma_user_completion(req->pq->dd, req->pq->ctxt,
-					req->pq->subctxt, req->info.comp_idx,
-					state, ret);
+		cq->comps[idx].errcode = -ret;
+	trace_hfi1_sdma_user_completion(pq->dd, pq->ctxt, pq->subctxt,
+					idx, state, ret);
+}
+
+static bool sdma_rb_filter(struct mmu_rb_node *node, unsigned long addr,
+			   unsigned long len)
+{
+	return (bool)(node->addr == addr);
+}
+
+static int sdma_rb_insert(struct rb_root *root, struct mmu_rb_node *mnode)
+{
+	struct sdma_mmu_node *node =
+		container_of(mnode, struct sdma_mmu_node, rb);
+
+	atomic_inc(&node->refcount);
+	return 0;
+}
+
+static void sdma_rb_remove(struct rb_root *root, struct mmu_rb_node *mnode,
+			   bool notifier)
+{
+	struct sdma_mmu_node *node =
+		container_of(mnode, struct sdma_mmu_node, rb);
+
+	spin_lock(&node->pq->evict_lock);
+	list_del(&node->list);
+	node->pq->n_locked -= node->npages;
+	spin_unlock(&node->pq->evict_lock);
+
+	unpin_vector_pages(notifier ? NULL : current->mm, node->pages,
+			   node->npages);
+	/*
+	 * If called by the MMU notifier, we have to adjust the pinned
+	 * page count ourselves.
+	 */
+	if (notifier)
+		current->mm->pinned_vm -= node->npages;
+	kfree(node);
+}
+
+static int sdma_rb_invalidate(struct rb_root *root, struct mmu_rb_node *mnode)
+{
+	struct sdma_mmu_node *node =
+		container_of(mnode, struct sdma_mmu_node, rb);
+
+	if (!atomic_read(&node->refcount))
+		return 1;
+	return 0;
 }
diff --git a/drivers/staging/rdma/hfi1/user_sdma.h b/drivers/staging/rdma/hfi1/user_sdma.h
index 0afa28508a8a..b9240e351161 100644
--- a/drivers/staging/rdma/hfi1/user_sdma.h
+++ b/drivers/staging/rdma/hfi1/user_sdma.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -69,7 +66,11 @@ struct hfi1_user_sdma_pkt_q {
 	struct iowait busy;
 	unsigned state;
 	wait_queue_head_t wait;
-	struct mm_struct *user_mm;
+	unsigned long unpinned;
+	struct rb_root sdma_rb_root;
+	u32 n_locked;
+	struct list_head evict;
+	spinlock_t evict_lock; /* protect evict and n_locked */
 };
 
 struct hfi1_user_sdma_comp_q {
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 176168614b5a..89f2aad45c1b 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -63,9 +60,9 @@
 #include "device.h"
 #include "trace.h"
 #include "qp.h"
-#include "sdma.h"
+#include "verbs_txreq.h"
 
-unsigned int hfi1_lkey_table_size = 16;
+static unsigned int hfi1_lkey_table_size = 16;
 module_param_named(lkey_table_size, hfi1_lkey_table_size, uint,
 		   S_IRUGO);
 MODULE_PARM_DESC(lkey_table_size,
@@ -124,45 +121,181 @@ unsigned int hfi1_max_srq_wrs = 0x1FFFF;
 module_param_named(max_srq_wrs, hfi1_max_srq_wrs, uint, S_IRUGO);
 MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");
 
+unsigned short piothreshold = 256;
+module_param(piothreshold, ushort, S_IRUGO);
+MODULE_PARM_DESC(piothreshold, "size used to determine sdma vs. pio");
+
+#define COPY_CACHELESS 1
+#define COPY_ADAPTIVE  2
+static unsigned int sge_copy_mode;
+module_param(sge_copy_mode, uint, S_IRUGO);
+MODULE_PARM_DESC(sge_copy_mode,
+		 "Verbs copy mode: 0 use memcpy, 1 use cacheless copy, 2 adapt based on WSS");
+
 static void verbs_sdma_complete(
 	struct sdma_txreq *cookie,
-	int status,
-	int drained);
+	int status);
+
+static int pio_wait(struct rvt_qp *qp,
+		    struct send_context *sc,
+		    struct hfi1_pkt_state *ps,
+		    u32 flag);
 
 /* Length of buffer to create verbs txreq cache name */
 #define TXREQ_NAME_LEN 24
 
+static uint wss_threshold;
+module_param(wss_threshold, uint, S_IRUGO);
+MODULE_PARM_DESC(wss_threshold, "Percentage (1-100) of LLC to use as a threshold for a cacheless copy");
+static uint wss_clean_period = 256;
+module_param(wss_clean_period, uint, S_IRUGO);
+MODULE_PARM_DESC(wss_clean_period, "Count of verbs copies before an entry in the page copy table is cleaned");
+
+/* memory working set size */
+struct hfi1_wss {
+	unsigned long *entries;
+	atomic_t total_count;
+	atomic_t clean_counter;
+	atomic_t clean_entry;
+
+	int threshold;
+	int num_entries;
+	long pages_mask;
+};
+
+static struct hfi1_wss wss;
+
+int hfi1_wss_init(void)
+{
+	long llc_size;
+	long llc_bits;
+	long table_size;
+	long table_bits;
+
+	/* check for a valid percent range - default to 80 if none or invalid */
+	if (wss_threshold < 1 || wss_threshold > 100)
+		wss_threshold = 80;
+	/* reject a wildly large period */
+	if (wss_clean_period > 1000000)
+		wss_clean_period = 256;
+	/* reject a zero period */
+	if (wss_clean_period == 0)
+		wss_clean_period = 1;
+
+	/*
+	 * Calculate the table size - the next power of 2 larger than the
+	 * LLC size.  LLC size is in KiB.
+	 */
+	llc_size = wss_llc_size() * 1024;
+	table_size = roundup_pow_of_two(llc_size);
+
+	/* one bit per page in rounded up table */
+	llc_bits = llc_size / PAGE_SIZE;
+	table_bits = table_size / PAGE_SIZE;
+	wss.pages_mask = table_bits - 1;
+	wss.num_entries = table_bits / BITS_PER_LONG;
+
+	wss.threshold = (llc_bits * wss_threshold) / 100;
+	if (wss.threshold == 0)
+		wss.threshold = 1;
+
+	atomic_set(&wss.clean_counter, wss_clean_period);
+
+	wss.entries = kcalloc(wss.num_entries, sizeof(*wss.entries),
+			      GFP_KERNEL);
+	if (!wss.entries) {
+		hfi1_wss_exit();
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+void hfi1_wss_exit(void)
+{
+	/* coded to handle partially initialized and repeat callers */
+	kfree(wss.entries);
+	wss.entries = NULL;
+}
+
 /*
- * Note that it is OK to post send work requests in the SQE and ERR
- * states; hfi1_do_send() will process them and generate error
- * completions as per IB 1.2 C10-96.
+ * Advance the clean counter.  When the clean period has expired,
+ * clean an entry.
+ *
+ * This is implemented in atomics to avoid locking.  Because multiple
+ * variables are involved, it can be racy which can lead to slightly
+ * inaccurate information.  Since this is only a heuristic, this is
+ * OK.  Any innaccuracies will clean themselves out as the counter
+ * advances.  That said, it is unlikely the entry clean operation will
+ * race - the next possible racer will not start until the next clean
+ * period.
+ *
+ * The clean counter is implemented as a decrement to zero.  When zero
+ * is reached an entry is cleaned.
  */
-const int ib_hfi1_state_ops[IB_QPS_ERR + 1] = {
-	[IB_QPS_RESET] = 0,
-	[IB_QPS_INIT] = HFI1_POST_RECV_OK,
-	[IB_QPS_RTR] = HFI1_POST_RECV_OK | HFI1_PROCESS_RECV_OK,
-	[IB_QPS_RTS] = HFI1_POST_RECV_OK | HFI1_PROCESS_RECV_OK |
-	    HFI1_POST_SEND_OK | HFI1_PROCESS_SEND_OK |
-	    HFI1_PROCESS_NEXT_SEND_OK,
-	[IB_QPS_SQD] = HFI1_POST_RECV_OK | HFI1_PROCESS_RECV_OK |
-	    HFI1_POST_SEND_OK | HFI1_PROCESS_SEND_OK,
-	[IB_QPS_SQE] = HFI1_POST_RECV_OK | HFI1_PROCESS_RECV_OK |
-	    HFI1_POST_SEND_OK | HFI1_FLUSH_SEND,
-	[IB_QPS_ERR] = HFI1_POST_RECV_OK | HFI1_FLUSH_RECV |
-	    HFI1_POST_SEND_OK | HFI1_FLUSH_SEND,
-};
+static void wss_advance_clean_counter(void)
+{
+	int entry;
+	int weight;
+	unsigned long bits;
 
-struct hfi1_ucontext {
-	struct ib_ucontext ibucontext;
-};
+	/* become the cleaner if we decrement the counter to zero */
+	if (atomic_dec_and_test(&wss.clean_counter)) {
+		/*
+		 * Set, not add, the clean period.  This avoids an issue
+		 * where the counter could decrement below the clean period.
+		 * Doing a set can result in lost decrements, slowing the
+		 * clean advance.  Since this a heuristic, this possible
+		 * slowdown is OK.
+		 *
+		 * An alternative is to loop, advancing the counter by a
+		 * clean period until the result is > 0. However, this could
+		 * lead to several threads keeping another in the clean loop.
+		 * This could be mitigated by limiting the number of times
+		 * we stay in the loop.
+		 */
+		atomic_set(&wss.clean_counter, wss_clean_period);
 
-static inline struct hfi1_ucontext *to_iucontext(struct ib_ucontext
-						  *ibucontext)
+		/*
+		 * Uniquely grab the entry to clean and move to next.
+		 * The current entry is always the lower bits of
+		 * wss.clean_entry.  The table size, wss.num_entries,
+		 * is always a power-of-2.
+		 */
+		entry = (atomic_inc_return(&wss.clean_entry) - 1)
+			& (wss.num_entries - 1);
+
+		/* clear the entry and count the bits */
+		bits = xchg(&wss.entries[entry], 0);
+		weight = hweight64((u64)bits);
+		/* only adjust the contended total count if needed */
+		if (weight)
+			atomic_sub(weight, &wss.total_count);
+	}
+}
+
+/*
+ * Insert the given address into the working set array.
+ */
+static void wss_insert(void *address)
 {
-	return container_of(ibucontext, struct hfi1_ucontext, ibucontext);
+	u32 page = ((unsigned long)address >> PAGE_SHIFT) & wss.pages_mask;
+	u32 entry = page / BITS_PER_LONG; /* assumes this ends up a shift */
+	u32 nr = page & (BITS_PER_LONG - 1);
+
+	if (!test_and_set_bit(nr, &wss.entries[entry]))
+		atomic_inc(&wss.total_count);
+
+	wss_advance_clean_counter();
 }
 
-static inline void _hfi1_schedule_send(struct hfi1_qp *qp);
+/*
+ * Is the working set larger than the threshold?
+ */
+static inline int wss_exceeds_threshold(void)
+{
+	return atomic_read(&wss.total_count) >= wss.threshold;
+}
 
 /*
  * Translate ib_wr_opcode into ib_wc_opcode.
@@ -274,14 +407,47 @@ __be64 ib_hfi1_sys_image_guid;
  * @ss: the SGE state
  * @data: the data to copy
  * @length: the length of the data
+ * @copy_last: do a separate copy of the last 8 bytes
  */
 void hfi1_copy_sge(
-	struct hfi1_sge_state *ss,
+	struct rvt_sge_state *ss,
 	void *data, u32 length,
-	int release)
+	int release,
+	int copy_last)
 {
-	struct hfi1_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
+	int in_last = 0;
+	int i;
+	int cacheless_copy = 0;
+
+	if (sge_copy_mode == COPY_CACHELESS) {
+		cacheless_copy = length >= PAGE_SIZE;
+	} else if (sge_copy_mode == COPY_ADAPTIVE) {
+		if (length >= PAGE_SIZE) {
+			/*
+			 * NOTE: this *assumes*:
+			 * o The first vaddr is the dest.
+			 * o If multiple pages, then vaddr is sequential.
+			 */
+			wss_insert(sge->vaddr);
+			if (length >= (2 * PAGE_SIZE))
+				wss_insert(sge->vaddr + PAGE_SIZE);
+
+			cacheless_copy = wss_exceeds_threshold();
+		} else {
+			wss_advance_clean_counter();
+		}
+	}
+	if (copy_last) {
+		if (length > 8) {
+			length -= 8;
+		} else {
+			copy_last = 0;
+			in_last = 1;
+		}
+	}
 
+again:
 	while (length) {
 		u32 len = sge->length;
 
@@ -290,17 +456,25 @@ void hfi1_copy_sge(
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		WARN_ON_ONCE(len == 0);
-		memcpy(sge->vaddr, data, len);
+		if (unlikely(in_last)) {
+			/* enforce byte transfer ordering */
+			for (i = 0; i < len; i++)
+				((u8 *)sge->vaddr)[i] = ((u8 *)data)[i];
+		} else if (cacheless_copy) {
+			cacheless_memcpy(sge->vaddr, data, len);
+		} else {
+			memcpy(sge->vaddr, data, len);
+		}
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (release)
-				hfi1_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= HFI1_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -313,6 +487,13 @@ void hfi1_copy_sge(
 		data += len;
 		length -= len;
 	}
+
+	if (copy_last) {
+		copy_last = 0;
+		in_last = 1;
+		length = 8;
+		goto again;
+	}
 }
 
 /**
@@ -320,9 +501,9 @@ void hfi1_copy_sge(
  * @ss: the SGE state
  * @length: the number of bytes to skip
  */
-void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
+void hfi1_skip_sge(struct rvt_sge_state *ss, u32 length, int release)
 {
-	struct hfi1_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	while (length) {
 		u32 len = sge->length;
@@ -337,11 +518,11 @@ void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
 		sge->sge_length -= len;
 		if (sge->sge_length == 0) {
 			if (release)
-				hfi1_put_mr(sge->mr);
+				rvt_put_mr(sge->mr);
 			if (--ss->num_sge)
 				*sge = *ss->sg_list++;
 		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= HFI1_SEGSZ) {
+			if (++sge->n >= RVT_SEGSZ) {
 				if (++sge->m >= sge->mr->mapsz)
 					break;
 				sge->n = 0;
@@ -355,231 +536,6 @@ void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
 	}
 }
 
-/**
- * post_one_send - post one RC, UC, or UD send work request
- * @qp: the QP to post on
- * @wr: the work request to send
- */
-static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
-{
-	struct hfi1_swqe *wqe;
-	u32 next;
-	int i;
-	int j;
-	int acc;
-	struct hfi1_lkey_table *rkt;
-	struct hfi1_pd *pd;
-	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
-	struct hfi1_pportdata *ppd;
-	struct hfi1_ibport *ibp;
-
-	/* IB spec says that num_sge == 0 is OK. */
-	if (unlikely(wr->num_sge > qp->s_max_sge))
-		return -EINVAL;
-
-	ppd = &dd->pport[qp->port_num - 1];
-	ibp = &ppd->ibport_data;
-
-	/*
-	 * Don't allow RDMA reads or atomic operations on UC or
-	 * undefined operations.
-	 * Make sure buffer is large enough to hold the result for atomics.
-	 */
-	if (qp->ibqp.qp_type == IB_QPT_UC) {
-		if ((unsigned) wr->opcode >= IB_WR_RDMA_READ)
-			return -EINVAL;
-	} else if (qp->ibqp.qp_type != IB_QPT_RC) {
-		/* Check IB_QPT_SMI, IB_QPT_GSI, IB_QPT_UD opcode */
-		if (wr->opcode != IB_WR_SEND &&
-		    wr->opcode != IB_WR_SEND_WITH_IMM)
-			return -EINVAL;
-		/* Check UD destination address PD */
-		if (qp->ibqp.pd != ud_wr(wr)->ah->pd)
-			return -EINVAL;
-	} else if ((unsigned) wr->opcode > IB_WR_ATOMIC_FETCH_AND_ADD)
-		return -EINVAL;
-	else if (wr->opcode >= IB_WR_ATOMIC_CMP_AND_SWP &&
-		   (wr->num_sge == 0 ||
-		    wr->sg_list[0].length < sizeof(u64) ||
-		    wr->sg_list[0].addr & (sizeof(u64) - 1)))
-		return -EINVAL;
-	else if (wr->opcode >= IB_WR_RDMA_READ && !qp->s_max_rd_atomic)
-		return -EINVAL;
-
-	next = qp->s_head + 1;
-	if (next >= qp->s_size)
-		next = 0;
-	if (next == qp->s_last)
-		return -ENOMEM;
-
-	rkt = &to_idev(qp->ibqp.device)->lk_table;
-	pd = to_ipd(qp->ibqp.pd);
-	wqe = get_swqe_ptr(qp, qp->s_head);
-
-
-	if (qp->ibqp.qp_type != IB_QPT_UC &&
-	    qp->ibqp.qp_type != IB_QPT_RC)
-		memcpy(&wqe->ud_wr, ud_wr(wr), sizeof(wqe->ud_wr));
-	else if (wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM ||
-		 wr->opcode == IB_WR_RDMA_WRITE ||
-		 wr->opcode == IB_WR_RDMA_READ)
-		memcpy(&wqe->rdma_wr, rdma_wr(wr), sizeof(wqe->rdma_wr));
-	else if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
-		 wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
-		memcpy(&wqe->atomic_wr, atomic_wr(wr), sizeof(wqe->atomic_wr));
-	else
-		memcpy(&wqe->wr, wr, sizeof(wqe->wr));
-
-	wqe->length = 0;
-	j = 0;
-	if (wr->num_sge) {
-		acc = wr->opcode >= IB_WR_RDMA_READ ?
-			IB_ACCESS_LOCAL_WRITE : 0;
-		for (i = 0; i < wr->num_sge; i++) {
-			u32 length = wr->sg_list[i].length;
-			int ok;
-
-			if (length == 0)
-				continue;
-			ok = hfi1_lkey_ok(rkt, pd, &wqe->sg_list[j],
-					  &wr->sg_list[i], acc);
-			if (!ok)
-				goto bail_inval_free;
-			wqe->length += length;
-			j++;
-		}
-		wqe->wr.num_sge = j;
-	}
-	if (qp->ibqp.qp_type == IB_QPT_UC ||
-	    qp->ibqp.qp_type == IB_QPT_RC) {
-		if (wqe->length > 0x80000000U)
-			goto bail_inval_free;
-	} else {
-		struct hfi1_ah *ah = to_iah(ud_wr(wr)->ah);
-
-		atomic_inc(&ah->refcount);
-	}
-	wqe->ssn = qp->s_ssn++;
-	qp->s_head = next;
-
-	return 0;
-
-bail_inval_free:
-	/* release mr holds */
-	while (j) {
-		struct hfi1_sge *sge = &wqe->sg_list[--j];
-
-		hfi1_put_mr(sge->mr);
-	}
-	return -EINVAL;
-}
-
-/**
- * post_send - post a send on a QP
- * @ibqp: the QP to post the send on
- * @wr: the list of work requests to post
- * @bad_wr: the first bad WR is put here
- *
- * This may be called from interrupt context.
- */
-static int post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
-		     struct ib_send_wr **bad_wr)
-{
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	int err = 0;
-	int call_send;
-	unsigned long flags;
-	unsigned nreq = 0;
-
-	spin_lock_irqsave(&qp->s_lock, flags);
-
-	/* Check that state is OK to post send. */
-	if (unlikely(!(ib_hfi1_state_ops[qp->state] & HFI1_POST_SEND_OK))) {
-		spin_unlock_irqrestore(&qp->s_lock, flags);
-		return -EINVAL;
-	}
-
-	/* sq empty and not list -> call send */
-	call_send = qp->s_head == qp->s_last && !wr->next;
-
-	for (; wr; wr = wr->next) {
-		err = post_one_send(qp, wr);
-		if (unlikely(err)) {
-			*bad_wr = wr;
-			goto bail;
-		}
-		nreq++;
-	}
-bail:
-	spin_unlock_irqrestore(&qp->s_lock, flags);
-	if (nreq && !call_send)
-		_hfi1_schedule_send(qp);
-	if (nreq && call_send)
-		hfi1_do_send(&qp->s_iowait.iowork);
-	return err;
-}
-
-/**
- * post_receive - post a receive on a QP
- * @ibqp: the QP to post the receive on
- * @wr: the WR to post
- * @bad_wr: the first bad WR is put here
- *
- * This may be called from interrupt context.
- */
-static int post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
-			struct ib_recv_wr **bad_wr)
-{
-	struct hfi1_qp *qp = to_iqp(ibqp);
-	struct hfi1_rwq *wq = qp->r_rq.wq;
-	unsigned long flags;
-	int ret;
-
-	/* Check that state is OK to post receive. */
-	if (!(ib_hfi1_state_ops[qp->state] & HFI1_POST_RECV_OK) || !wq) {
-		*bad_wr = wr;
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	for (; wr; wr = wr->next) {
-		struct hfi1_rwqe *wqe;
-		u32 next;
-		int i;
-
-		if ((unsigned) wr->num_sge > qp->r_rq.max_sge) {
-			*bad_wr = wr;
-			ret = -EINVAL;
-			goto bail;
-		}
-
-		spin_lock_irqsave(&qp->r_rq.lock, flags);
-		next = wq->head + 1;
-		if (next >= qp->r_rq.size)
-			next = 0;
-		if (next == wq->tail) {
-			spin_unlock_irqrestore(&qp->r_rq.lock, flags);
-			*bad_wr = wr;
-			ret = -ENOMEM;
-			goto bail;
-		}
-
-		wqe = get_rwqe_ptr(&qp->r_rq, wq->head);
-		wqe->wr_id = wr->wr_id;
-		wqe->num_sge = wr->num_sge;
-		for (i = 0; i < wr->num_sge; i++)
-			wqe->sg_list[i] = wr->sg_list[i];
-		/* Make sure queue entry is written before the head index. */
-		smp_wmb();
-		wq->head = next;
-		spin_unlock_irqrestore(&qp->r_rq.lock, flags);
-	}
-	ret = 0;
-
-bail:
-	return ret;
-}
-
 /*
  * Make sure the QP is ready and able to accept the given opcode.
  */
@@ -587,18 +543,17 @@ static inline int qp_ok(int opcode, struct hfi1_packet *packet)
 {
 	struct hfi1_ibport *ibp;
 
-	if (!(ib_hfi1_state_ops[packet->qp->state] & HFI1_PROCESS_RECV_OK))
+	if (!(ib_rvt_state_ops[packet->qp->state] & RVT_PROCESS_RECV_OK))
 		goto dropit;
 	if (((opcode & OPCODE_QP_MASK) == packet->qp->allowed_ops) ||
 	    (opcode == IB_OPCODE_CNP))
 		return 1;
 dropit:
 	ibp = &packet->rcd->ppd->ibport_data;
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 	return 0;
 }
 
-
 /**
  * hfi1_ib_rcv - process an incoming packet
  * @packet: data packet information
@@ -614,6 +569,7 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 	u32 tlen = packet->tlen;
 	struct hfi1_pportdata *ppd = rcd->ppd;
 	struct hfi1_ibport *ibp = &ppd->ibport_data;
+	struct rvt_dev_info *rdi = &ppd->dd->verbs_dev.rdi;
 	unsigned long flags;
 	u32 qp_num;
 	int lnh;
@@ -622,9 +578,9 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 
 	/* Check for GRH */
 	lnh = be16_to_cpu(hdr->lrh[0]) & 3;
-	if (lnh == HFI1_LRH_BTH)
+	if (lnh == HFI1_LRH_BTH) {
 		packet->ohdr = &hdr->u.oth;
-	else if (lnh == HFI1_LRH_GRH) {
+	} else if (lnh == HFI1_LRH_GRH) {
 		u32 vtf;
 
 		packet->ohdr = &hdr->u.l.oth;
@@ -634,8 +590,9 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 		if ((vtf >> IB_GRH_VERSION_SHIFT) != IB_GRH_VERSION)
 			goto drop;
 		packet->rcv_flags |= HFI1_HAS_GRH;
-	} else
+	} else {
 		goto drop;
+	}
 
 	trace_input_ibhdr(rcd->dd, hdr);
 
@@ -643,17 +600,17 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 	inc_opstats(tlen, &rcd->opstats->stats[opcode]);
 
 	/* Get the destination QP number. */
-	qp_num = be32_to_cpu(packet->ohdr->bth[1]) & HFI1_QPN_MASK;
+	qp_num = be32_to_cpu(packet->ohdr->bth[1]) & RVT_QPN_MASK;
 	lid = be16_to_cpu(hdr->lrh[1]);
-	if (unlikely((lid >= HFI1_MULTICAST_LID_BASE) &&
-	    (lid != HFI1_PERMISSIVE_LID))) {
-		struct hfi1_mcast *mcast;
-		struct hfi1_mcast_qp *p;
+	if (unlikely((lid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) &&
+		     (lid != be16_to_cpu(IB_LID_PERMISSIVE)))) {
+		struct rvt_mcast *mcast;
+		struct rvt_mcast_qp *p;
 
 		if (lnh != HFI1_LRH_GRH)
 			goto drop;
-		mcast = hfi1_mcast_find(ibp, &hdr->u.l.grh.dgid);
-		if (mcast == NULL)
+		mcast = rvt_mcast_find(&ibp->rvp, &hdr->u.l.grh.dgid);
+		if (!mcast)
 			goto drop;
 		list_for_each_entry_rcu(p, &mcast->qp_list, list) {
 			packet->qp = p->qp;
@@ -663,14 +620,14 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 			spin_unlock_irqrestore(&packet->qp->r_lock, flags);
 		}
 		/*
-		 * Notify hfi1_multicast_detach() if it is waiting for us
+		 * Notify rvt_multicast_detach() if it is waiting for us
 		 * to finish.
 		 */
 		if (atomic_dec_return(&mcast->refcount) <= 1)
 			wake_up(&mcast->wait);
 	} else {
 		rcu_read_lock();
-		packet->qp = hfi1_lookup_qpn(ibp, qp_num);
+		packet->qp = rvt_lookup_qpn(rdi, &ibp->rvp, qp_num);
 		if (!packet->qp) {
 			rcu_read_unlock();
 			goto drop;
@@ -684,7 +641,7 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
 	return;
 
 drop:
-	ibp->n_pkt_drops++;
+	ibp->rvp.n_pkt_drops++;
 }
 
 /*
@@ -695,15 +652,17 @@ static void mem_timer(unsigned long data)
 {
 	struct hfi1_ibdev *dev = (struct hfi1_ibdev *)data;
 	struct list_head *list = &dev->memwait;
-	struct hfi1_qp *qp = NULL;
+	struct rvt_qp *qp = NULL;
 	struct iowait *wait;
 	unsigned long flags;
+	struct hfi1_qp_priv *priv;
 
 	write_seqlock_irqsave(&dev->iowait_lock, flags);
 	if (!list_empty(list)) {
 		wait = list_first_entry(list, struct iowait, list);
-		qp = container_of(wait, struct hfi1_qp, s_iowait);
-		list_del_init(&qp->s_iowait.list);
+		qp = iowait_to_qp(wait);
+		priv = qp->priv;
+		list_del_init(&priv->s_iowait.list);
 		/* refcount held until actual wake up */
 		if (!list_empty(list))
 			mod_timer(&dev->mem_timer, jiffies + 1);
@@ -711,12 +670,12 @@ static void mem_timer(unsigned long data)
 	write_sequnlock_irqrestore(&dev->iowait_lock, flags);
 
 	if (qp)
-		hfi1_qp_wakeup(qp, HFI1_S_WAIT_KMEM);
+		hfi1_qp_wakeup(qp, RVT_S_WAIT_KMEM);
 }
 
-void update_sge(struct hfi1_sge_state *ss, u32 length)
+void update_sge(struct rvt_sge_state *ss, u32 length)
 {
-	struct hfi1_sge *sge = &ss->sge;
+	struct rvt_sge *sge = &ss->sge;
 
 	sge->vaddr += length;
 	sge->length -= length;
@@ -725,7 +684,7 @@ void update_sge(struct hfi1_sge_state *ss, u32 length)
 		if (--ss->num_sge)
 			*sge = *ss->sg_list++;
 	} else if (sge->length == 0 && sge->mr->lkey) {
-		if (++sge->n >= HFI1_SEGSZ) {
+		if (++sge->n >= RVT_SEGSZ) {
 			if (++sge->m >= sge->mr->mapsz)
 				return;
 			sge->n = 0;
@@ -735,143 +694,55 @@ void update_sge(struct hfi1_sge_state *ss, u32 length)
 	}
 }
 
-static noinline struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
-						struct hfi1_qp *qp)
-{
-	struct verbs_txreq *tx;
-	unsigned long flags;
-
-	tx = kmem_cache_alloc(dev->verbs_txreq_cache, GFP_ATOMIC);
-	if (!tx) {
-		spin_lock_irqsave(&qp->s_lock, flags);
-		write_seqlock(&dev->iowait_lock);
-		if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK &&
-		    list_empty(&qp->s_iowait.list)) {
-			dev->n_txwait++;
-			qp->s_flags |= HFI1_S_WAIT_TX;
-			list_add_tail(&qp->s_iowait.list, &dev->txwait);
-			trace_hfi1_qpsleep(qp, HFI1_S_WAIT_TX);
-			atomic_inc(&qp->refcount);
-		}
-		qp->s_flags &= ~HFI1_S_BUSY;
-		write_sequnlock(&dev->iowait_lock);
-		spin_unlock_irqrestore(&qp->s_lock, flags);
-		tx = ERR_PTR(-EBUSY);
-	}
-	return tx;
-}
-
-static inline struct verbs_txreq *get_txreq(struct hfi1_ibdev *dev,
-					    struct hfi1_qp *qp)
-{
-	struct verbs_txreq *tx;
-
-	tx = kmem_cache_alloc(dev->verbs_txreq_cache, GFP_ATOMIC);
-	if (!tx) {
-		/* call slow path to get the lock */
-		tx =  __get_txreq(dev, qp);
-		if (IS_ERR(tx))
-			return tx;
-	}
-	tx->qp = qp;
-	return tx;
-}
-
-void hfi1_put_txreq(struct verbs_txreq *tx)
-{
-	struct hfi1_ibdev *dev;
-	struct hfi1_qp *qp;
-	unsigned long flags;
-	unsigned int seq;
-
-	qp = tx->qp;
-	dev = to_idev(qp->ibqp.device);
-
-	if (tx->mr) {
-		hfi1_put_mr(tx->mr);
-		tx->mr = NULL;
-	}
-	sdma_txclean(dd_from_dev(dev), &tx->txreq);
-
-	/* Free verbs_txreq and return to slab cache */
-	kmem_cache_free(dev->verbs_txreq_cache, tx);
-
-	do {
-		seq = read_seqbegin(&dev->iowait_lock);
-		if (!list_empty(&dev->txwait)) {
-			struct iowait *wait;
-
-			write_seqlock_irqsave(&dev->iowait_lock, flags);
-			/* Wake up first QP wanting a free struct */
-			wait = list_first_entry(&dev->txwait, struct iowait,
-						list);
-			qp = container_of(wait, struct hfi1_qp, s_iowait);
-			list_del_init(&qp->s_iowait.list);
-			/* refcount held until actual wake up */
-			write_sequnlock_irqrestore(&dev->iowait_lock, flags);
-			hfi1_qp_wakeup(qp, HFI1_S_WAIT_TX);
-			break;
-		}
-	} while (read_seqretry(&dev->iowait_lock, seq));
-}
-
 /*
  * This is called with progress side lock held.
  */
 /* New API */
 static void verbs_sdma_complete(
 	struct sdma_txreq *cookie,
-	int status,
-	int drained)
+	int status)
 {
 	struct verbs_txreq *tx =
 		container_of(cookie, struct verbs_txreq, txreq);
-	struct hfi1_qp *qp = tx->qp;
+	struct rvt_qp *qp = tx->qp;
 
 	spin_lock(&qp->s_lock);
-	if (tx->wqe)
+	if (tx->wqe) {
 		hfi1_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
-	else if (qp->ibqp.qp_type == IB_QPT_RC) {
+	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		struct hfi1_ib_header *hdr;
 
 		hdr = &tx->phdr.hdr;
 		hfi1_rc_send_complete(qp, hdr);
 	}
-	if (drained) {
-		/*
-		 * This happens when the send engine notes
-		 * a QP in the error state and cannot
-		 * do the flush work until that QP's
-		 * sdma work has finished.
-		 */
-		if (qp->s_flags & HFI1_S_WAIT_DMA) {
-			qp->s_flags &= ~HFI1_S_WAIT_DMA;
-			hfi1_schedule_send(qp);
-		}
-	}
 	spin_unlock(&qp->s_lock);
 
 	hfi1_put_txreq(tx);
 }
 
-static int wait_kmem(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
+static int wait_kmem(struct hfi1_ibdev *dev,
+		     struct rvt_qp *qp,
+		     struct hfi1_pkt_state *ps)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	unsigned long flags;
 	int ret = 0;
 
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
 		write_seqlock(&dev->iowait_lock);
-		if (list_empty(&qp->s_iowait.list)) {
+		list_add_tail(&ps->s_txreq->txreq.list,
+			      &priv->s_iowait.tx_head);
+		if (list_empty(&priv->s_iowait.list)) {
 			if (list_empty(&dev->memwait))
 				mod_timer(&dev->mem_timer, jiffies + 1);
-			qp->s_flags |= HFI1_S_WAIT_KMEM;
-			list_add_tail(&qp->s_iowait.list, &dev->memwait);
-			trace_hfi1_qpsleep(qp, HFI1_S_WAIT_KMEM);
+			qp->s_flags |= RVT_S_WAIT_KMEM;
+			list_add_tail(&priv->s_iowait.list, &dev->memwait);
+			trace_hfi1_qpsleep(qp, RVT_S_WAIT_KMEM);
 			atomic_inc(&qp->refcount);
 		}
 		write_sequnlock(&dev->iowait_lock);
-		qp->s_flags &= ~HFI1_S_BUSY;
+		qp->s_flags &= ~RVT_S_BUSY;
 		ret = -EBUSY;
 	}
 	spin_unlock_irqrestore(&qp->s_lock, flags);
@@ -884,14 +755,14 @@ static int wait_kmem(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
  *
  * Add failures will revert the sge cursor
  */
-static int build_verbs_ulp_payload(
+static noinline int build_verbs_ulp_payload(
 	struct sdma_engine *sde,
-	struct hfi1_sge_state *ss,
+	struct rvt_sge_state *ss,
 	u32 length,
 	struct verbs_txreq *tx)
 {
-	struct hfi1_sge *sg_list = ss->sg_list;
-	struct hfi1_sge sge = ss->sge;
+	struct rvt_sge *sg_list = ss->sg_list;
+	struct rvt_sge sge = ss->sge;
 	u8 num_sge = ss->num_sge;
 	u32 len;
 	int ret = 0;
@@ -928,23 +799,21 @@ bail_txadd:
  * NOTE: DMA mapping is held in the tx until completed in the ring or
  *       the tx desc is freed without having been submitted to the ring
  *
- * This routine insures the following all the helper routine
- * calls succeed.
+ * This routine ensures all the helper routine calls succeed.
  */
 /* New API */
 static int build_verbs_tx_desc(
 	struct sdma_engine *sde,
-	struct hfi1_sge_state *ss,
+	struct rvt_sge_state *ss,
 	u32 length,
 	struct verbs_txreq *tx,
 	struct ahg_ib_header *ahdr,
 	u64 pbc)
 {
 	int ret = 0;
-	struct hfi1_pio_header *phdr;
+	struct hfi1_pio_header *phdr = &tx->phdr;
 	u16 hdrbytes = tx->hdr_dwords << 2;
 
-	phdr = &tx->phdr;
 	if (!ahdr->ahgcount) {
 		ret = sdma_txinit_ahg(
 			&tx->txreq,
@@ -958,29 +827,14 @@ static int build_verbs_tx_desc(
 		if (ret)
 			goto bail_txadd;
 		phdr->pbc = cpu_to_le64(pbc);
-		memcpy(&phdr->hdr, &ahdr->ibh, hdrbytes - sizeof(phdr->pbc));
-		/* add the header */
 		ret = sdma_txadd_kvaddr(
 			sde->dd,
 			&tx->txreq,
-			&tx->phdr,
-			tx->hdr_dwords << 2);
+			phdr,
+			hdrbytes);
 		if (ret)
 			goto bail_txadd;
 	} else {
-		struct hfi1_other_headers *sohdr = &ahdr->ibh.u.oth;
-		struct hfi1_other_headers *dohdr = &phdr->hdr.u.oth;
-
-		/* needed in rc_send_complete() */
-		phdr->hdr.lrh[0] = ahdr->ibh.lrh[0];
-		if ((be16_to_cpu(phdr->hdr.lrh[0]) & 3) == HFI1_LRH_GRH) {
-			sohdr = &ahdr->ibh.u.l.oth;
-			dohdr = &phdr->hdr.u.l.oth;
-		}
-		/* opcode */
-		dohdr->bth[0] = sohdr->bth[0];
-		/* PSN/ACK  */
-		dohdr->bth[2] = sohdr->bth[2];
 		ret = sdma_txinit_ahg(
 			&tx->txreq,
 			ahdr->tx_flags,
@@ -1001,80 +855,75 @@ bail_txadd:
 	return ret;
 }
 
-int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			u64 pbc)
 {
-	struct ahg_ib_header *ahdr = qp->s_hdr;
+	struct hfi1_qp_priv *priv = qp->priv;
+	struct ahg_ib_header *ahdr = priv->s_hdr;
 	u32 hdrwords = qp->s_hdrwords;
-	struct hfi1_sge_state *ss = qp->s_cur_sge;
+	struct rvt_sge_state *ss = qp->s_cur_sge;
 	u32 len = qp->s_cur_size;
 	u32 plen = hdrwords + ((len + 3) >> 2) + 2; /* includes pbc */
 	struct hfi1_ibdev *dev = ps->dev;
 	struct hfi1_pportdata *ppd = ps->ppd;
 	struct verbs_txreq *tx;
-	struct sdma_txreq *stx;
 	u64 pbc_flags = 0;
-	u8 sc5 = qp->s_sc;
+	u8 sc5 = priv->s_sc;
+
 	int ret;
 
-	if (!list_empty(&qp->s_iowait.tx_head)) {
-		stx = list_first_entry(
-			&qp->s_iowait.tx_head,
-			struct sdma_txreq,
-			list);
-		list_del_init(&stx->list);
-		tx = container_of(stx, struct verbs_txreq, txreq);
-		ret = sdma_send_txreq(tx->sde, &qp->s_iowait, stx);
-		if (unlikely(ret == -ECOMM))
+	tx = ps->s_txreq;
+	if (!sdma_txreq_built(&tx->txreq)) {
+		if (likely(pbc == 0)) {
+			u32 vl = sc_to_vlt(dd_from_ibdev(qp->ibqp.device), sc5);
+			/* No vl15 here */
+			/* set PBC_DC_INFO bit (aka SC[4]) in pbc_flags */
+			pbc_flags |= (!!(sc5 & 0x10)) << PBC_DC_INFO_SHIFT;
+
+			pbc = create_pbc(ppd,
+					 pbc_flags,
+					 qp->srate_mbps,
+					 vl,
+					 plen);
+		}
+		tx->wqe = qp->s_wqe;
+		ret = build_verbs_tx_desc(tx->sde, ss, len, tx, ahdr, pbc);
+		if (unlikely(ret))
+			goto bail_build;
+	}
+	ret =  sdma_send_txreq(tx->sde, &priv->s_iowait, &tx->txreq);
+	if (unlikely(ret < 0)) {
+		if (ret == -ECOMM)
 			goto bail_ecomm;
 		return ret;
 	}
-
-	tx = get_txreq(dev, qp);
-	if (IS_ERR(tx))
-		goto bail_tx;
-
-	tx->sde = qp->s_sde;
-
-	if (likely(pbc == 0)) {
-		u32 vl = sc_to_vlt(dd_from_ibdev(qp->ibqp.device), sc5);
-		/* No vl15 here */
-		/* set PBC_DC_INFO bit (aka SC[4]) in pbc_flags */
-		pbc_flags |= (!!(sc5 & 0x10)) << PBC_DC_INFO_SHIFT;
-
-		pbc = create_pbc(ppd, pbc_flags, qp->srate_mbps, vl, plen);
-	}
-	tx->wqe = qp->s_wqe;
-	tx->mr = qp->s_rdma_mr;
-	if (qp->s_rdma_mr)
-		qp->s_rdma_mr = NULL;
-	tx->hdr_dwords = hdrwords + 2;
-	ret = build_verbs_tx_desc(tx->sde, ss, len, tx, ahdr, pbc);
-	if (unlikely(ret))
-		goto bail_build;
-	trace_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &ahdr->ibh);
-	ret =  sdma_send_txreq(tx->sde, &qp->s_iowait, &tx->txreq);
-	if (unlikely(ret == -ECOMM))
-		goto bail_ecomm;
+	trace_sdma_output_ibhdr(dd_from_ibdev(qp->ibqp.device),
+				&ps->s_txreq->phdr.hdr);
 	return ret;
 
 bail_ecomm:
 	/* The current one got "sent" */
 	return 0;
 bail_build:
-	/* kmalloc or mapping fail */
-	hfi1_put_txreq(tx);
-	return wait_kmem(dev, qp);
-bail_tx:
-	return PTR_ERR(tx);
+	ret = wait_kmem(dev, qp, ps);
+	if (!ret) {
+		/* free txreq - bad state */
+		hfi1_put_txreq(ps->s_txreq);
+		ps->s_txreq = NULL;
+	}
+	return ret;
 }
 
 /*
  * If we are now in the error state, return zero to flush the
  * send work request.
  */
-static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
+static int pio_wait(struct rvt_qp *qp,
+		    struct send_context *sc,
+		    struct hfi1_pkt_state *ps,
+		    u32 flag)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_devdata *dd = sc->dd;
 	struct hfi1_ibdev *dev = &dd->verbs_dev;
 	unsigned long flags;
@@ -1087,74 +936,89 @@ static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
 	 * enabling the PIO avail interrupt.
 	 */
 	spin_lock_irqsave(&qp->s_lock, flags);
-	if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
 		write_seqlock(&dev->iowait_lock);
-		if (list_empty(&qp->s_iowait.list)) {
+		list_add_tail(&ps->s_txreq->txreq.list,
+			      &priv->s_iowait.tx_head);
+		if (list_empty(&priv->s_iowait.list)) {
 			struct hfi1_ibdev *dev = &dd->verbs_dev;
 			int was_empty;
 
+			dev->n_piowait += !!(flag & RVT_S_WAIT_PIO);
+			dev->n_piodrain += !!(flag & RVT_S_WAIT_PIO_DRAIN);
 			dev->n_piowait++;
-			qp->s_flags |= HFI1_S_WAIT_PIO;
+			qp->s_flags |= flag;
 			was_empty = list_empty(&sc->piowait);
-			list_add_tail(&qp->s_iowait.list, &sc->piowait);
-			trace_hfi1_qpsleep(qp, HFI1_S_WAIT_PIO);
+			list_add_tail(&priv->s_iowait.list, &sc->piowait);
+			trace_hfi1_qpsleep(qp, RVT_S_WAIT_PIO);
 			atomic_inc(&qp->refcount);
 			/* counting: only call wantpiobuf_intr if first user */
 			if (was_empty)
 				hfi1_sc_wantpiobuf_intr(sc, 1);
 		}
 		write_sequnlock(&dev->iowait_lock);
-		qp->s_flags &= ~HFI1_S_BUSY;
+		qp->s_flags &= ~RVT_S_BUSY;
 		ret = -EBUSY;
 	}
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 	return ret;
 }
 
-struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5)
+static void verbs_pio_complete(void *arg, int code)
 {
-	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
-	struct hfi1_pportdata *ppd = dd->pport + (qp->port_num - 1);
-	u8 vl;
+	struct rvt_qp *qp = (struct rvt_qp *)arg;
+	struct hfi1_qp_priv *priv = qp->priv;
 
-	vl = sc_to_vlt(dd, sc5);
-	if (vl >= ppd->vls_supported && vl != 15)
-		return NULL;
-	return dd->vld[vl].sc;
+	if (iowait_pio_dec(&priv->s_iowait))
+		iowait_drain_wakeup(&priv->s_iowait);
 }
 
-int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			u64 pbc)
 {
-	struct ahg_ib_header *ahdr = qp->s_hdr;
+	struct hfi1_qp_priv *priv = qp->priv;
 	u32 hdrwords = qp->s_hdrwords;
-	struct hfi1_sge_state *ss = qp->s_cur_sge;
+	struct rvt_sge_state *ss = qp->s_cur_sge;
 	u32 len = qp->s_cur_size;
 	u32 dwords = (len + 3) >> 2;
 	u32 plen = hdrwords + dwords + 2; /* includes pbc */
 	struct hfi1_pportdata *ppd = ps->ppd;
-	u32 *hdr = (u32 *)&ahdr->ibh;
+	u32 *hdr = (u32 *)&ps->s_txreq->phdr.hdr;
 	u64 pbc_flags = 0;
-	u32 sc5;
+	u8 sc5;
 	unsigned long flags = 0;
 	struct send_context *sc;
 	struct pio_buf *pbuf;
 	int wc_status = IB_WC_SUCCESS;
+	int ret = 0;
+	pio_release_cb cb = NULL;
+
+	/* only RC/UC use complete */
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+		cb = verbs_pio_complete;
+		break;
+	default:
+		break;
+	}
 
 	/* vl15 special case taken care of in ud.c */
-	sc5 = qp->s_sc;
-	sc = qp_to_send_context(qp, sc5);
+	sc5 = priv->s_sc;
+	sc = ps->s_txreq->psc;
 
-	if (!sc)
-		return -EINVAL;
 	if (likely(pbc == 0)) {
-		u32 vl = sc_to_vlt(dd_from_ibdev(qp->ibqp.device), sc5);
+		u8 vl = sc_to_vlt(dd_from_ibdev(qp->ibqp.device), sc5);
 		/* set PBC_DC_INFO bit (aka SC[4]) in pbc_flags */
 		pbc_flags |= (!!(sc5 & 0x10)) << PBC_DC_INFO_SHIFT;
 		pbc = create_pbc(ppd, pbc_flags, qp->srate_mbps, vl, plen);
 	}
-	pbuf = sc_buffer_alloc(sc, plen, NULL, NULL);
-	if (unlikely(pbuf == NULL)) {
+	if (cb)
+		iowait_pio_inc(&priv->s_iowait);
+	pbuf = sc_buffer_alloc(sc, plen, cb, qp);
+	if (unlikely(!pbuf)) {
+		if (cb)
+			verbs_pio_complete(qp, 0);
 		if (ppd->host_link_state != HLS_UP_ACTIVE) {
 			/*
 			 * If we have filled the PIO buffers to capacity and are
@@ -1174,7 +1038,12 @@ int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 			 * so lets continue to queue the request.
 			 */
 			hfi1_cdbg(PIO, "alloc failed. state active, queuing");
-			return no_bufs_available(qp, sc);
+			ret = pio_wait(qp, sc, ps, RVT_S_WAIT_PIO);
+			if (!ret)
+				/* txreq not queued - free */
+				goto bail;
+			/* tx consumed in wait */
+			return ret;
 		}
 	}
 
@@ -1182,7 +1051,7 @@ int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 		pio_copy(ppd->dd, pbuf, pbc, hdr, hdrwords);
 	} else {
 		if (ss) {
-			seg_pio_copy_start(pbuf, pbc, hdr, hdrwords*4);
+			seg_pio_copy_start(pbuf, pbc, hdr, hdrwords * 4);
 			while (len) {
 				void *addr = ss->sge.vaddr;
 				u32 slen = ss->sge.length;
@@ -1197,12 +1066,8 @@ int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
 		}
 	}
 
-	trace_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &ahdr->ibh);
-
-	if (qp->s_rdma_mr) {
-		hfi1_put_mr(qp->s_rdma_mr);
-		qp->s_rdma_mr = NULL;
-	}
+	trace_pio_output_ibhdr(dd_from_ibdev(qp->ibqp.device),
+			       &ps->s_txreq->phdr.hdr);
 
 pio_bail:
 	if (qp->s_wqe) {
@@ -1211,10 +1076,15 @@ pio_bail:
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		spin_lock_irqsave(&qp->s_lock, flags);
-		hfi1_rc_send_complete(qp, &ahdr->ibh);
+		hfi1_rc_send_complete(qp, &ps->s_txreq->phdr.hdr);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 	}
-	return 0;
+
+	ret = 0;
+
+bail:
+	hfi1_put_txreq(ps->s_txreq);
+	return ret;
 }
 
 /*
@@ -1247,13 +1117,14 @@ static inline int egress_pkey_matches_entry(u16 pkey, u16 ent)
  */
 static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
 				    struct hfi1_ib_header *hdr,
-				    struct hfi1_qp *qp)
+				    struct rvt_qp *qp)
 {
+	struct hfi1_qp_priv *priv = qp->priv;
 	struct hfi1_other_headers *ohdr;
 	struct hfi1_devdata *dd;
 	int i = 0;
 	u16 pkey;
-	u8 lnh, sc5 = qp->s_sc;
+	u8 lnh, sc5 = priv->s_sc;
 
 	if (!(ppd->part_enforce & HFI1_PART_ENFORCE_OUT))
 		return 0;
@@ -1271,14 +1142,14 @@ static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
 	if ((sc5 == 0xf) && ((pkey & PKEY_LOW_15_MASK) != PKEY_LOW_15_MASK))
 		goto bad;
 
-
 	/* Is the pkey = 0x0, or 0x8000? */
 	if ((pkey & PKEY_LOW_15_MASK) == 0)
 		goto bad;
 
 	/* The most likely matching pkey has index qp->s_pkey_index */
 	if (unlikely(!egress_pkey_matches_entry(pkey,
-					ppd->pkeys[qp->s_pkey_index]))) {
+						ppd->pkeys
+						[qp->s_pkey_index]))) {
 		/* no match - try the entire table */
 		for (; i < MAX_PKEY_VALUES; i++) {
 			if (egress_pkey_matches_entry(pkey, ppd->pkeys[i]))
@@ -1302,31 +1173,65 @@ bad:
 }
 
 /**
+ * get_send_routine - choose an egress routine
+ *
+ * Choose an egress routine based on QP type
+ * and size
+ */
+static inline send_routine get_send_routine(struct rvt_qp *qp,
+					    struct verbs_txreq *tx)
+{
+	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+	struct hfi1_qp_priv *priv = qp->priv;
+	struct hfi1_ib_header *h = &tx->phdr.hdr;
+
+	if (unlikely(!(dd->flags & HFI1_HAS_SEND_DMA)))
+		return dd->process_pio_send;
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_SMI:
+		return dd->process_pio_send;
+	case IB_QPT_GSI:
+	case IB_QPT_UD:
+		break;
+	case IB_QPT_RC:
+		if (piothreshold &&
+		    qp->s_cur_size <= min(piothreshold, qp->pmtu) &&
+		    (BIT(get_opcode(h) & 0x1f) & rc_only_opcode) &&
+		    iowait_sdma_pending(&priv->s_iowait) == 0 &&
+		    !sdma_txreq_built(&tx->txreq))
+			return dd->process_pio_send;
+		break;
+	case IB_QPT_UC:
+		if (piothreshold &&
+		    qp->s_cur_size <= min(piothreshold, qp->pmtu) &&
+		    (BIT(get_opcode(h) & 0x1f) & uc_only_opcode) &&
+		    iowait_sdma_pending(&priv->s_iowait) == 0 &&
+		    !sdma_txreq_built(&tx->txreq))
+			return dd->process_pio_send;
+		break;
+	default:
+		break;
+	}
+	return dd->process_dma_send;
+}
+
+/**
  * hfi1_verbs_send - send a packet
  * @qp: the QP to send on
  * @ps: the state of the packet to send
  *
  * Return zero if packet is sent or queued OK.
- * Return non-zero and clear qp->s_flags HFI1_S_BUSY otherwise.
+ * Return non-zero and clear qp->s_flags RVT_S_BUSY otherwise.
  */
-int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps)
+int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 {
 	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
-	struct ahg_ib_header *ahdr = qp->s_hdr;
+	struct hfi1_qp_priv *priv = qp->priv;
+	send_routine sr;
 	int ret;
-	int pio = 0;
-	unsigned long flags = 0;
-
-	/*
-	 * VL15 packets (IB_QPT_SMI) will always use PIO, so we
-	 * can defer SDMA restart until link goes ACTIVE without
-	 * worrying about just how we got there.
-	 */
-	if ((qp->ibqp.qp_type == IB_QPT_SMI) ||
-	    !(dd->flags & HFI1_HAS_SEND_DMA))
-		pio = 1;
 
-	ret = egress_pkey_check(dd->pport, &ahdr->ibh, qp);
+	sr = get_send_routine(qp, ps->s_txreq);
+	ret = egress_pkey_check(dd->pport, &ps->s_txreq->phdr.hdr, qp);
 	if (unlikely(ret)) {
 		/*
 		 * The value we are returning here does not get propagated to
@@ -1336,7 +1241,9 @@ int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps)
 		 * mechanism for handling the errors. So for SDMA we can just
 		 * return.
 		 */
-		if (pio) {
+		if (sr == dd->process_pio_send) {
+			unsigned long flags;
+
 			hfi1_cdbg(PIO, "%s() Failed. Completing with err",
 				  __func__);
 			spin_lock_irqsave(&qp->s_lock, flags);
@@ -1345,71 +1252,57 @@ int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps)
 		}
 		return -EINVAL;
 	}
-
-	if (pio) {
-		ret = dd->process_pio_send(qp, ps, 0);
-	} else {
-#ifdef CONFIG_SDMA_VERBOSITY
-		dd_dev_err(dd, "CONFIG SDMA %s:%d %s()\n",
-			   slashstrip(__FILE__), __LINE__, __func__);
-		dd_dev_err(dd, "SDMA hdrwords = %u, len = %u\n", qp->s_hdrwords,
-			   qp->s_cur_size);
-#endif
-		ret = dd->process_dma_send(qp, ps, 0);
-	}
-
-	return ret;
+	if (sr == dd->process_dma_send && iowait_pio_pending(&priv->s_iowait))
+		return pio_wait(qp,
+				ps->s_txreq->psc,
+				ps,
+				RVT_S_WAIT_PIO_DRAIN);
+	return sr(qp, ps, 0);
 }
 
-static int query_device(struct ib_device *ibdev,
-			struct ib_device_attr *props,
-			struct ib_udata *uhw)
+/**
+ * hfi1_fill_device_attr - Fill in rvt dev info device attributes.
+ * @dd: the device data structure
+ */
+static void hfi1_fill_device_attr(struct hfi1_devdata *dd)
 {
-	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	struct hfi1_ibdev *dev = to_idev(ibdev);
-
-	if (uhw->inlen || uhw->outlen)
-		return -EINVAL;
-	memset(props, 0, sizeof(*props));
-
-	props->device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
-		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
-		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
-		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
-
-	props->page_size_cap = PAGE_SIZE;
-	props->vendor_id =
-		dd->oui1 << 16 | dd->oui2 << 8 | dd->oui3;
-	props->vendor_part_id = dd->pcidev->device;
-	props->hw_ver = dd->minrev;
-	props->sys_image_guid = ib_hfi1_sys_image_guid;
-	props->max_mr_size = ~0ULL;
-	props->max_qp = hfi1_max_qps;
-	props->max_qp_wr = hfi1_max_qp_wrs;
-	props->max_sge = hfi1_max_sges;
-	props->max_sge_rd = hfi1_max_sges;
-	props->max_cq = hfi1_max_cqs;
-	props->max_ah = hfi1_max_ahs;
-	props->max_cqe = hfi1_max_cqes;
-	props->max_mr = dev->lk_table.max;
-	props->max_fmr = dev->lk_table.max;
-	props->max_map_per_fmr = 32767;
-	props->max_pd = hfi1_max_pds;
-	props->max_qp_rd_atom = HFI1_MAX_RDMA_ATOMIC;
-	props->max_qp_init_rd_atom = 255;
-	/* props->max_res_rd_atom */
-	props->max_srq = hfi1_max_srqs;
-	props->max_srq_wr = hfi1_max_srq_wrs;
-	props->max_srq_sge = hfi1_max_srq_sges;
-	/* props->local_ca_ack_delay */
-	props->atomic_cap = IB_ATOMIC_GLOB;
-	props->max_pkeys = hfi1_get_npkeys(dd);
-	props->max_mcast_grp = hfi1_max_mcast_grps;
-	props->max_mcast_qp_attach = hfi1_max_mcast_qp_attached;
-	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
-		props->max_mcast_grp;
-
-	return 0;
+	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
+
+	memset(&rdi->dparms.props, 0, sizeof(rdi->dparms.props));
+
+	rdi->dparms.props.device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
+			IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
+			IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
+			IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
+	rdi->dparms.props.page_size_cap = PAGE_SIZE;
+	rdi->dparms.props.vendor_id = dd->oui1 << 16 | dd->oui2 << 8 | dd->oui3;
+	rdi->dparms.props.vendor_part_id = dd->pcidev->device;
+	rdi->dparms.props.hw_ver = dd->minrev;
+	rdi->dparms.props.sys_image_guid = ib_hfi1_sys_image_guid;
+	rdi->dparms.props.max_mr_size = ~0ULL;
+	rdi->dparms.props.max_qp = hfi1_max_qps;
+	rdi->dparms.props.max_qp_wr = hfi1_max_qp_wrs;
+	rdi->dparms.props.max_sge = hfi1_max_sges;
+	rdi->dparms.props.max_sge_rd = hfi1_max_sges;
+	rdi->dparms.props.max_cq = hfi1_max_cqs;
+	rdi->dparms.props.max_ah = hfi1_max_ahs;
+	rdi->dparms.props.max_cqe = hfi1_max_cqes;
+	rdi->dparms.props.max_mr = rdi->lkey_table.max;
+	rdi->dparms.props.max_fmr = rdi->lkey_table.max;
+	rdi->dparms.props.max_map_per_fmr = 32767;
+	rdi->dparms.props.max_pd = hfi1_max_pds;
+	rdi->dparms.props.max_qp_rd_atom = HFI1_MAX_RDMA_ATOMIC;
+	rdi->dparms.props.max_qp_init_rd_atom = 255;
+	rdi->dparms.props.max_srq = hfi1_max_srqs;
+	rdi->dparms.props.max_srq_wr = hfi1_max_srq_wrs;
+	rdi->dparms.props.max_srq_sge = hfi1_max_srq_sges;
+	rdi->dparms.props.atomic_cap = IB_ATOMIC_GLOB;
+	rdi->dparms.props.max_pkeys = hfi1_get_npkeys(dd);
+	rdi->dparms.props.max_mcast_grp = hfi1_max_mcast_grps;
+	rdi->dparms.props.max_mcast_qp_attach = hfi1_max_mcast_qp_attached;
+	rdi->dparms.props.max_total_mcast_qp_attach =
+					rdi->dparms.props.max_mcast_qp_attach *
+					rdi->dparms.props.max_mcast_grp;
 }
 
 static inline u16 opa_speed_to_ib(u16 in)
@@ -1443,33 +1336,24 @@ static inline u16 opa_width_to_ib(u16 in)
 	}
 }
 
-static int query_port(struct ib_device *ibdev, u8 port,
+static int query_port(struct rvt_dev_info *rdi, u8 port_num,
 		      struct ib_port_attr *props)
 {
-	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	struct hfi1_ibport *ibp = to_iport(ibdev, port);
-	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
+	struct hfi1_ibdev *verbs_dev = dev_from_rdi(rdi);
+	struct hfi1_devdata *dd = dd_from_dev(verbs_dev);
+	struct hfi1_pportdata *ppd = &dd->pport[port_num - 1];
 	u16 lid = ppd->lid;
 
-	memset(props, 0, sizeof(*props));
 	props->lid = lid ? lid : 0;
 	props->lmc = ppd->lmc;
-	props->sm_lid = ibp->sm_lid;
-	props->sm_sl = ibp->sm_sl;
 	/* OPA logical states match IB logical states */
 	props->state = driver_lstate(ppd);
 	props->phys_state = hfi1_ibphys_portstate(ppd);
-	props->port_cap_flags = ibp->port_cap_flags;
 	props->gid_tbl_len = HFI1_GUIDS_PER_PORT;
-	props->max_msg_sz = 0x80000000;
-	props->pkey_tbl_len = hfi1_get_npkeys(dd);
-	props->bad_pkey_cntr = ibp->pkey_violations;
-	props->qkey_viol_cntr = ibp->qkey_violations;
 	props->active_width = (u8)opa_width_to_ib(ppd->link_width_active);
 	/* see rate_show() in ib core/sysfs.c */
 	props->active_speed = (u8)opa_speed_to_ib(ppd->link_speed_active);
 	props->max_vl_num = ppd->vls_supported;
-	props->init_type_reply = 0;
 
 	/* Once we are a "first class" citizen and have added the OPA MTUs to
 	 * the core we can advertise the larger MTU enum to the ULPs, for now
@@ -1483,27 +1367,6 @@ static int query_port(struct ib_device *ibdev, u8 port,
 				      4096 : hfi1_max_mtu), IB_MTU_4096);
 	props->active_mtu = !valid_ib_mtu(ppd->ibmtu) ? props->max_mtu :
 		mtu_to_enum(ppd->ibmtu, IB_MTU_2048);
-	props->subnet_timeout = ibp->subnet_timeout;
-
-	return 0;
-}
-
-static int port_immutable(struct ib_device *ibdev, u8 port_num,
-			  struct ib_port_immutable *immutable)
-{
-	struct ib_port_attr attr;
-	int err;
-
-	err = query_port(ibdev, port_num, &attr);
-	if (err)
-		return err;
-
-	memset(immutable, 0, sizeof(*immutable));
-
-	immutable->pkey_tbl_len = attr.pkey_tbl_len;
-	immutable->gid_tbl_len = attr.gid_tbl_len;
-	immutable->core_cap_flags = RDMA_CORE_PORT_INTEL_OPA;
-	immutable->max_mad_size = OPA_MGMT_MAD_SIZE;
 
 	return 0;
 }
@@ -1547,102 +1410,31 @@ bail:
 	return ret;
 }
 
-static int modify_port(struct ib_device *ibdev, u8 port,
-		       int port_modify_mask, struct ib_port_modify *props)
-{
-	struct hfi1_ibport *ibp = to_iport(ibdev, port);
-	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
-	int ret = 0;
-
-	ibp->port_cap_flags |= props->set_port_cap_mask;
-	ibp->port_cap_flags &= ~props->clr_port_cap_mask;
-	if (props->set_port_cap_mask || props->clr_port_cap_mask)
-		hfi1_cap_mask_chg(ibp);
-	if (port_modify_mask & IB_PORT_SHUTDOWN) {
-		set_link_down_reason(ppd, OPA_LINKDOWN_REASON_UNKNOWN, 0,
-		  OPA_LINKDOWN_REASON_UNKNOWN);
-		ret = set_link_state(ppd, HLS_DN_DOWNDEF);
-	}
-	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
-		ibp->qkey_violations = 0;
-	return ret;
-}
-
-static int query_gid(struct ib_device *ibdev, u8 port,
-		     int index, union ib_gid *gid)
+static int shut_down_port(struct rvt_dev_info *rdi, u8 port_num)
 {
-	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	int ret = 0;
-
-	if (!port || port > dd->num_pports)
-		ret = -EINVAL;
-	else {
-		struct hfi1_ibport *ibp = to_iport(ibdev, port);
-		struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
-
-		gid->global.subnet_prefix = ibp->gid_prefix;
-		if (index == 0)
-			gid->global.interface_id = cpu_to_be64(ppd->guid);
-		else if (index < HFI1_GUIDS_PER_PORT)
-			gid->global.interface_id = ibp->guids[index - 1];
-		else
-			ret = -EINVAL;
-	}
-
-	return ret;
-}
-
-static struct ib_pd *alloc_pd(struct ib_device *ibdev,
-			      struct ib_ucontext *context,
-			      struct ib_udata *udata)
-{
-	struct hfi1_ibdev *dev = to_idev(ibdev);
-	struct hfi1_pd *pd;
-	struct ib_pd *ret;
-
-	/*
-	 * This is actually totally arbitrary.  Some correctness tests
-	 * assume there's a maximum number of PDs that can be allocated.
-	 * We don't actually have this limit, but we fail the test if
-	 * we allow allocations of more than we report for this value.
-	 */
-
-	pd = kmalloc(sizeof(*pd), GFP_KERNEL);
-	if (!pd) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	spin_lock(&dev->n_pds_lock);
-	if (dev->n_pds_allocated == hfi1_max_pds) {
-		spin_unlock(&dev->n_pds_lock);
-		kfree(pd);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	dev->n_pds_allocated++;
-	spin_unlock(&dev->n_pds_lock);
-
-	/* ib_alloc_pd() will initialize pd->ibpd. */
-	pd->user = udata != NULL;
-
-	ret = &pd->ibpd;
+	struct hfi1_ibdev *verbs_dev = dev_from_rdi(rdi);
+	struct hfi1_devdata *dd = dd_from_dev(verbs_dev);
+	struct hfi1_pportdata *ppd = &dd->pport[port_num - 1];
+	int ret;
 
-bail:
+	set_link_down_reason(ppd, OPA_LINKDOWN_REASON_UNKNOWN, 0,
+			     OPA_LINKDOWN_REASON_UNKNOWN);
+	ret = set_link_state(ppd, HLS_DN_DOWNDEF);
 	return ret;
 }
 
-static int dealloc_pd(struct ib_pd *ibpd)
+static int hfi1_get_guid_be(struct rvt_dev_info *rdi, struct rvt_ibport *rvp,
+			    int guid_index, __be64 *guid)
 {
-	struct hfi1_pd *pd = to_ipd(ibpd);
-	struct hfi1_ibdev *dev = to_idev(ibpd->device);
-
-	spin_lock(&dev->n_pds_lock);
-	dev->n_pds_allocated--;
-	spin_unlock(&dev->n_pds_lock);
+	struct hfi1_ibport *ibp = container_of(rvp, struct hfi1_ibport, rvp);
+	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 
-	kfree(pd);
+	if (guid_index == 0)
+		*guid = cpu_to_be64(ppd->guid);
+	else if (guid_index < HFI1_GUIDS_PER_PORT)
+		*guid = ibp->guids[guid_index - 1];
+	else
+		return -EINVAL;
 
 	return 0;
 }
@@ -1657,101 +1449,57 @@ u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah)
 	return ibp->sl_to_sc[ah->sl];
 }
 
-int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
+static int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
 {
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
 	struct hfi1_devdata *dd;
 	u8 sc5;
 
-	/* A multicast address requires a GRH (see ch. 8.4.1). */
-	if (ah_attr->dlid >= HFI1_MULTICAST_LID_BASE &&
-	    ah_attr->dlid != HFI1_PERMISSIVE_LID &&
-	    !(ah_attr->ah_flags & IB_AH_GRH))
-		goto bail;
-	if ((ah_attr->ah_flags & IB_AH_GRH) &&
-	    ah_attr->grh.sgid_index >= HFI1_GUIDS_PER_PORT)
-		goto bail;
-	if (ah_attr->dlid == 0)
-		goto bail;
-	if (ah_attr->port_num < 1 ||
-	    ah_attr->port_num > ibdev->phys_port_cnt)
-		goto bail;
-	if (ah_attr->static_rate != IB_RATE_PORT_CURRENT &&
-	    ib_rate_to_mbps(ah_attr->static_rate) < 0)
-		goto bail;
-	if (ah_attr->sl >= OPA_MAX_SLS)
-		goto bail;
 	/* test the mapping for validity */
 	ibp = to_iport(ibdev, ah_attr->port_num);
 	ppd = ppd_from_ibp(ibp);
 	sc5 = ibp->sl_to_sc[ah_attr->sl];
 	dd = dd_from_ppd(ppd);
 	if (sc_to_vlt(dd, sc5) > num_vls && sc_to_vlt(dd, sc5) != 0xf)
-		goto bail;
+		return -EINVAL;
 	return 0;
-bail:
-	return -EINVAL;
 }
 
-/**
- * create_ah - create an address handle
- * @pd: the protection domain
- * @ah_attr: the attributes of the AH
- *
- * This may be called from interrupt context.
- */
-static struct ib_ah *create_ah(struct ib_pd *pd,
-			       struct ib_ah_attr *ah_attr)
+static void hfi1_notify_new_ah(struct ib_device *ibdev,
+			       struct ib_ah_attr *ah_attr,
+			       struct rvt_ah *ah)
 {
-	struct hfi1_ah *ah;
-	struct ib_ah *ret;
-	struct hfi1_ibdev *dev = to_idev(pd->device);
-	unsigned long flags;
-
-	if (hfi1_check_ah(pd->device, ah_attr)) {
-		ret = ERR_PTR(-EINVAL);
-		goto bail;
-	}
-
-	ah = kmalloc(sizeof(*ah), GFP_ATOMIC);
-	if (!ah) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	spin_lock_irqsave(&dev->n_ahs_lock, flags);
-	if (dev->n_ahs_allocated == hfi1_max_ahs) {
-		spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-		kfree(ah);
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	dev->n_ahs_allocated++;
-	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
-	/* ib_create_ah() will initialize ah->ibah. */
-	ah->attr = *ah_attr;
-	atomic_set(&ah->refcount, 0);
+	struct hfi1_ibport *ibp;
+	struct hfi1_pportdata *ppd;
+	struct hfi1_devdata *dd;
+	u8 sc5;
 
-	ret = &ah->ibah;
+	/*
+	 * Do not trust reading anything from rvt_ah at this point as it is not
+	 * done being setup. We can however modify things which we need to set.
+	 */
 
-bail:
-	return ret;
+	ibp = to_iport(ibdev, ah_attr->port_num);
+	ppd = ppd_from_ibp(ibp);
+	sc5 = ibp->sl_to_sc[ah->attr.sl];
+	dd = dd_from_ppd(ppd);
+	ah->vl = sc_to_vlt(dd, sc5);
+	if (ah->vl < num_vls || ah->vl == 15)
+		ah->log_pmtu = ilog2(dd->vld[ah->vl].mtu);
 }
 
 struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
 {
 	struct ib_ah_attr attr;
 	struct ib_ah *ah = ERR_PTR(-EINVAL);
-	struct hfi1_qp *qp0;
+	struct rvt_qp *qp0;
 
 	memset(&attr, 0, sizeof(attr));
 	attr.dlid = dlid;
 	attr.port_num = ppd_from_ibp(ibp)->port;
 	rcu_read_lock();
-	qp0 = rcu_dereference(ibp->qp[0]);
+	qp0 = rcu_dereference(ibp->rvp.qp[0]);
 	if (qp0)
 		ah = ib_create_ah(qp0->ibqp.pd, &attr);
 	rcu_read_unlock();
@@ -1759,51 +1507,6 @@ struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
 }
 
 /**
- * destroy_ah - destroy an address handle
- * @ibah: the AH to destroy
- *
- * This may be called from interrupt context.
- */
-static int destroy_ah(struct ib_ah *ibah)
-{
-	struct hfi1_ibdev *dev = to_idev(ibah->device);
-	struct hfi1_ah *ah = to_iah(ibah);
-	unsigned long flags;
-
-	if (atomic_read(&ah->refcount) != 0)
-		return -EBUSY;
-
-	spin_lock_irqsave(&dev->n_ahs_lock, flags);
-	dev->n_ahs_allocated--;
-	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
-	kfree(ah);
-
-	return 0;
-}
-
-static int modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
-	struct hfi1_ah *ah = to_iah(ibah);
-
-	if (hfi1_check_ah(ibah->device, ah_attr))
-		return -EINVAL;
-
-	ah->attr = *ah_attr;
-
-	return 0;
-}
-
-static int query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
-	struct hfi1_ah *ah = to_iah(ibah);
-
-	*ah_attr = ah->attr;
-
-	return 0;
-}
-
-/**
  * hfi1_get_npkeys - return the size of the PKEY table for context 0
  * @dd: the hfi1_ib device
  */
@@ -1812,54 +1515,6 @@ unsigned hfi1_get_npkeys(struct hfi1_devdata *dd)
 	return ARRAY_SIZE(dd->pport[0].pkeys);
 }
 
-static int query_pkey(struct ib_device *ibdev, u8 port, u16 index,
-		      u16 *pkey)
-{
-	struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
-	int ret;
-
-	if (index >= hfi1_get_npkeys(dd)) {
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	*pkey = hfi1_get_pkey(to_iport(ibdev, port), index);
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * alloc_ucontext - allocate a ucontest
- * @ibdev: the infiniband device
- * @udata: not used by the driver
- */
-
-static struct ib_ucontext *alloc_ucontext(struct ib_device *ibdev,
-					  struct ib_udata *udata)
-{
-	struct hfi1_ucontext *context;
-	struct ib_ucontext *ret;
-
-	context = kmalloc(sizeof(*context), GFP_KERNEL);
-	if (!context) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
-	ret = &context->ibucontext;
-
-bail:
-	return ret;
-}
-
-static int dealloc_ucontext(struct ib_ucontext *context)
-{
-	kfree(to_iucontext(context));
-	return 0;
-}
-
 static void init_ibport(struct hfi1_pportdata *ppd)
 {
 	struct hfi1_ibport *ibp = &ppd->ibport_data;
@@ -1871,28 +1526,21 @@ static void init_ibport(struct hfi1_pportdata *ppd)
 		ibp->sc_to_sl[i] = i;
 	}
 
-	spin_lock_init(&ibp->lock);
+	spin_lock_init(&ibp->rvp.lock);
 	/* Set the prefix to the default value (see ch. 4.1.1) */
-	ibp->gid_prefix = IB_DEFAULT_GID_PREFIX;
-	ibp->sm_lid = 0;
+	ibp->rvp.gid_prefix = IB_DEFAULT_GID_PREFIX;
+	ibp->rvp.sm_lid = 0;
 	/* Below should only set bits defined in OPA PortInfo.CapabilityMask */
-	ibp->port_cap_flags = IB_PORT_AUTO_MIGR_SUP |
+	ibp->rvp.port_cap_flags = IB_PORT_AUTO_MIGR_SUP |
 		IB_PORT_CAP_MASK_NOTICE_SUP;
-	ibp->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
-	ibp->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
-	ibp->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
-	ibp->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
-	ibp->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
-
-	RCU_INIT_POINTER(ibp->qp[0], NULL);
-	RCU_INIT_POINTER(ibp->qp[1], NULL);
-}
-
-static void verbs_txreq_kmem_cache_ctor(void *obj)
-{
-	struct verbs_txreq *tx = obj;
-
-	memset(tx, 0, sizeof(*tx));
+	ibp->rvp.pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
+	ibp->rvp.pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
+	ibp->rvp.pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
+	ibp->rvp.pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
+	ibp->rvp.pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
+
+	RCU_INIT_POINTER(ibp->rvp.qp[0], NULL);
+	RCU_INIT_POINTER(ibp->rvp.qp[1], NULL);
 }
 
 /**
@@ -1903,74 +1551,26 @@ static void verbs_txreq_kmem_cache_ctor(void *obj)
 int hfi1_register_ib_device(struct hfi1_devdata *dd)
 {
 	struct hfi1_ibdev *dev = &dd->verbs_dev;
-	struct ib_device *ibdev = &dev->ibdev;
+	struct ib_device *ibdev = &dev->rdi.ibdev;
 	struct hfi1_pportdata *ppd = dd->pport;
-	unsigned i, lk_tab_size;
+	unsigned i;
 	int ret;
 	size_t lcpysz = IB_DEVICE_NAME_MAX;
-	u16 descq_cnt;
-	char buf[TXREQ_NAME_LEN];
-
-	ret = hfi1_qp_init(dev);
-	if (ret)
-		goto err_qp_init;
-
 
 	for (i = 0; i < dd->num_pports; i++)
 		init_ibport(ppd + i);
 
 	/* Only need to initialize non-zero fields. */
-	spin_lock_init(&dev->n_pds_lock);
-	spin_lock_init(&dev->n_ahs_lock);
-	spin_lock_init(&dev->n_cqs_lock);
-	spin_lock_init(&dev->n_qps_lock);
-	spin_lock_init(&dev->n_srqs_lock);
-	spin_lock_init(&dev->n_mcast_grps_lock);
+
 	setup_timer(&dev->mem_timer, mem_timer, (unsigned long)dev);
 
-	/*
-	 * The top hfi1_lkey_table_size bits are used to index the
-	 * table.  The lower 8 bits can be owned by the user (copied from
-	 * the LKEY).  The remaining bits act as a generation number or tag.
-	 */
-	spin_lock_init(&dev->lk_table.lock);
-	dev->lk_table.max = 1 << hfi1_lkey_table_size;
-	/* ensure generation is at least 4 bits (keys.c) */
-	if (hfi1_lkey_table_size > MAX_LKEY_TABLE_BITS) {
-		dd_dev_warn(dd, "lkey bits %u too large, reduced to %u\n",
-			      hfi1_lkey_table_size, MAX_LKEY_TABLE_BITS);
-		hfi1_lkey_table_size = MAX_LKEY_TABLE_BITS;
-	}
-	lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
-	dev->lk_table.table = (struct hfi1_mregion __rcu **)
-		vmalloc(lk_tab_size);
-	if (dev->lk_table.table == NULL) {
-		ret = -ENOMEM;
-		goto err_lk;
-	}
-	RCU_INIT_POINTER(dev->dma_mr, NULL);
-	for (i = 0; i < dev->lk_table.max; i++)
-		RCU_INIT_POINTER(dev->lk_table.table[i], NULL);
-	INIT_LIST_HEAD(&dev->pending_mmaps);
-	spin_lock_init(&dev->pending_lock);
 	seqlock_init(&dev->iowait_lock);
-	dev->mmap_offset = PAGE_SIZE;
-	spin_lock_init(&dev->mmap_offset_lock);
 	INIT_LIST_HEAD(&dev->txwait);
 	INIT_LIST_HEAD(&dev->memwait);
 
-	descq_cnt = sdma_get_descq_cnt();
-
-	snprintf(buf, sizeof(buf), "hfi1_%u_vtxreq_cache", dd->unit);
-	/* SLAB_HWCACHE_ALIGN for AHG */
-	dev->verbs_txreq_cache = kmem_cache_create(buf,
-						   sizeof(struct verbs_txreq),
-						   0, SLAB_HWCACHE_ALIGN,
-						   verbs_txreq_kmem_cache_ctor);
-	if (!dev->verbs_txreq_cache) {
-		ret = -ENOMEM;
+	ret = verbs_txreq_init(dev);
+	if (ret)
 		goto err_verbs_txreq;
-	}
 
 	/*
 	 * The system image GUID is supposed to be the same for all
@@ -1983,142 +1583,119 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 	strlcpy(ibdev->name + lcpysz, "_%d", IB_DEVICE_NAME_MAX - lcpysz);
 	ibdev->owner = THIS_MODULE;
 	ibdev->node_guid = cpu_to_be64(ppd->guid);
-	ibdev->uverbs_abi_ver = HFI1_UVERBS_ABI_VERSION;
-	ibdev->uverbs_cmd_mask =
-		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)         |
-		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)        |
-		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)          |
-		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)            |
-		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)          |
-		(1ull << IB_USER_VERBS_CMD_CREATE_AH)           |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_AH)           |
-		(1ull << IB_USER_VERBS_CMD_QUERY_AH)            |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)          |
-		(1ull << IB_USER_VERBS_CMD_REG_MR)              |
-		(1ull << IB_USER_VERBS_CMD_DEREG_MR)            |
-		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
-		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)           |
-		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)          |
-		(1ull << IB_USER_VERBS_CMD_POLL_CQ)             |
-		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)       |
-		(1ull << IB_USER_VERBS_CMD_CREATE_QP)           |
-		(1ull << IB_USER_VERBS_CMD_QUERY_QP)            |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)          |
-		(1ull << IB_USER_VERBS_CMD_POST_SEND)           |
-		(1ull << IB_USER_VERBS_CMD_POST_RECV)           |
-		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)        |
-		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)        |
-		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)          |
-		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)          |
-		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)           |
-		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)         |
-		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
-	ibdev->node_type = RDMA_NODE_IB_CA;
 	ibdev->phys_port_cnt = dd->num_pports;
-	ibdev->num_comp_vectors = 1;
 	ibdev->dma_device = &dd->pcidev->dev;
-	ibdev->query_device = query_device;
 	ibdev->modify_device = modify_device;
-	ibdev->query_port = query_port;
-	ibdev->modify_port = modify_port;
-	ibdev->query_pkey = query_pkey;
-	ibdev->query_gid = query_gid;
-	ibdev->alloc_ucontext = alloc_ucontext;
-	ibdev->dealloc_ucontext = dealloc_ucontext;
-	ibdev->alloc_pd = alloc_pd;
-	ibdev->dealloc_pd = dealloc_pd;
-	ibdev->create_ah = create_ah;
-	ibdev->destroy_ah = destroy_ah;
-	ibdev->modify_ah = modify_ah;
-	ibdev->query_ah = query_ah;
-	ibdev->create_srq = hfi1_create_srq;
-	ibdev->modify_srq = hfi1_modify_srq;
-	ibdev->query_srq = hfi1_query_srq;
-	ibdev->destroy_srq = hfi1_destroy_srq;
-	ibdev->create_qp = hfi1_create_qp;
-	ibdev->modify_qp = hfi1_modify_qp;
-	ibdev->query_qp = hfi1_query_qp;
-	ibdev->destroy_qp = hfi1_destroy_qp;
-	ibdev->post_send = post_send;
-	ibdev->post_recv = post_receive;
-	ibdev->post_srq_recv = hfi1_post_srq_receive;
-	ibdev->create_cq = hfi1_create_cq;
-	ibdev->destroy_cq = hfi1_destroy_cq;
-	ibdev->resize_cq = hfi1_resize_cq;
-	ibdev->poll_cq = hfi1_poll_cq;
-	ibdev->req_notify_cq = hfi1_req_notify_cq;
-	ibdev->get_dma_mr = hfi1_get_dma_mr;
-	ibdev->reg_user_mr = hfi1_reg_user_mr;
-	ibdev->dereg_mr = hfi1_dereg_mr;
-	ibdev->alloc_mr = hfi1_alloc_mr;
-	ibdev->alloc_fmr = hfi1_alloc_fmr;
-	ibdev->map_phys_fmr = hfi1_map_phys_fmr;
-	ibdev->unmap_fmr = hfi1_unmap_fmr;
-	ibdev->dealloc_fmr = hfi1_dealloc_fmr;
-	ibdev->attach_mcast = hfi1_multicast_attach;
-	ibdev->detach_mcast = hfi1_multicast_detach;
+
+	/* keep process mad in the driver */
 	ibdev->process_mad = hfi1_process_mad;
-	ibdev->mmap = hfi1_mmap;
-	ibdev->dma_ops = &hfi1_dma_mapping_ops;
-	ibdev->get_port_immutable = port_immutable;
 
 	strncpy(ibdev->node_desc, init_utsname()->nodename,
 		sizeof(ibdev->node_desc));
 
-	ret = ib_register_device(ibdev, hfi1_create_port_files);
-	if (ret)
-		goto err_reg;
-
-	ret = hfi1_create_agents(dev);
+	/*
+	 * Fill in rvt info object.
+	 */
+	dd->verbs_dev.rdi.driver_f.port_callback = hfi1_create_port_files;
+	dd->verbs_dev.rdi.driver_f.get_card_name = get_card_name;
+	dd->verbs_dev.rdi.driver_f.get_pci_dev = get_pci_dev;
+	dd->verbs_dev.rdi.driver_f.check_ah = hfi1_check_ah;
+	dd->verbs_dev.rdi.driver_f.notify_new_ah = hfi1_notify_new_ah;
+	dd->verbs_dev.rdi.driver_f.get_guid_be = hfi1_get_guid_be;
+	dd->verbs_dev.rdi.driver_f.query_port_state = query_port;
+	dd->verbs_dev.rdi.driver_f.shut_down_port = shut_down_port;
+	dd->verbs_dev.rdi.driver_f.cap_mask_chg = hfi1_cap_mask_chg;
+	/*
+	 * Fill in rvt info device attributes.
+	 */
+	hfi1_fill_device_attr(dd);
+
+	/* queue pair */
+	dd->verbs_dev.rdi.dparms.qp_table_size = hfi1_qp_table_size;
+	dd->verbs_dev.rdi.dparms.qpn_start = 0;
+	dd->verbs_dev.rdi.dparms.qpn_inc = 1;
+	dd->verbs_dev.rdi.dparms.qos_shift = dd->qos_shift;
+	dd->verbs_dev.rdi.dparms.qpn_res_start = kdeth_qp << 16;
+	dd->verbs_dev.rdi.dparms.qpn_res_end =
+	dd->verbs_dev.rdi.dparms.qpn_res_start + 65535;
+	dd->verbs_dev.rdi.dparms.max_rdma_atomic = HFI1_MAX_RDMA_ATOMIC;
+	dd->verbs_dev.rdi.dparms.psn_mask = PSN_MASK;
+	dd->verbs_dev.rdi.dparms.psn_shift = PSN_SHIFT;
+	dd->verbs_dev.rdi.dparms.psn_modify_mask = PSN_MODIFY_MASK;
+	dd->verbs_dev.rdi.dparms.core_cap_flags = RDMA_CORE_PORT_INTEL_OPA;
+	dd->verbs_dev.rdi.dparms.max_mad_size = OPA_MGMT_MAD_SIZE;
+
+	dd->verbs_dev.rdi.driver_f.qp_priv_alloc = qp_priv_alloc;
+	dd->verbs_dev.rdi.driver_f.qp_priv_free = qp_priv_free;
+	dd->verbs_dev.rdi.driver_f.free_all_qps = free_all_qps;
+	dd->verbs_dev.rdi.driver_f.notify_qp_reset = notify_qp_reset;
+	dd->verbs_dev.rdi.driver_f.do_send = hfi1_do_send;
+	dd->verbs_dev.rdi.driver_f.schedule_send = hfi1_schedule_send;
+	dd->verbs_dev.rdi.driver_f.schedule_send_no_lock = _hfi1_schedule_send;
+	dd->verbs_dev.rdi.driver_f.get_pmtu_from_attr = get_pmtu_from_attr;
+	dd->verbs_dev.rdi.driver_f.notify_error_qp = notify_error_qp;
+	dd->verbs_dev.rdi.driver_f.flush_qp_waiters = flush_qp_waiters;
+	dd->verbs_dev.rdi.driver_f.stop_send_queue = stop_send_queue;
+	dd->verbs_dev.rdi.driver_f.quiesce_qp = quiesce_qp;
+	dd->verbs_dev.rdi.driver_f.notify_error_qp = notify_error_qp;
+	dd->verbs_dev.rdi.driver_f.mtu_from_qp = mtu_from_qp;
+	dd->verbs_dev.rdi.driver_f.mtu_to_path_mtu = mtu_to_path_mtu;
+	dd->verbs_dev.rdi.driver_f.check_modify_qp = hfi1_check_modify_qp;
+	dd->verbs_dev.rdi.driver_f.modify_qp = hfi1_modify_qp;
+	dd->verbs_dev.rdi.driver_f.check_send_wqe = hfi1_check_send_wqe;
+
+	/* completeion queue */
+	snprintf(dd->verbs_dev.rdi.dparms.cq_name,
+		 sizeof(dd->verbs_dev.rdi.dparms.cq_name),
+		 "hfi1_cq%d", dd->unit);
+	dd->verbs_dev.rdi.dparms.node = dd->node;
+
+	/* misc settings */
+	dd->verbs_dev.rdi.flags = 0; /* Let rdmavt handle it all */
+	dd->verbs_dev.rdi.dparms.lkey_table_size = hfi1_lkey_table_size;
+	dd->verbs_dev.rdi.dparms.nports = dd->num_pports;
+	dd->verbs_dev.rdi.dparms.npkeys = hfi1_get_npkeys(dd);
+
+	ppd = dd->pport;
+	for (i = 0; i < dd->num_pports; i++, ppd++)
+		rvt_init_port(&dd->verbs_dev.rdi,
+			      &ppd->ibport_data.rvp,
+			      i,
+			      ppd->pkeys);
+
+	ret = rvt_register_device(&dd->verbs_dev.rdi);
 	if (ret)
-		goto err_agents;
+		goto err_verbs_txreq;
 
 	ret = hfi1_verbs_register_sysfs(dd);
 	if (ret)
 		goto err_class;
 
-	goto bail;
+	return ret;
 
 err_class:
-	hfi1_free_agents(dev);
-err_agents:
-	ib_unregister_device(ibdev);
-err_reg:
+	rvt_unregister_device(&dd->verbs_dev.rdi);
 err_verbs_txreq:
-	kmem_cache_destroy(dev->verbs_txreq_cache);
-	vfree(dev->lk_table.table);
-err_lk:
-	hfi1_qp_exit(dev);
-err_qp_init:
+	verbs_txreq_exit(dev);
 	dd_dev_err(dd, "cannot register verbs: %d!\n", -ret);
-bail:
 	return ret;
 }
 
 void hfi1_unregister_ib_device(struct hfi1_devdata *dd)
 {
 	struct hfi1_ibdev *dev = &dd->verbs_dev;
-	struct ib_device *ibdev = &dev->ibdev;
 
 	hfi1_verbs_unregister_sysfs(dd);
 
-	hfi1_free_agents(dev);
-
-	ib_unregister_device(ibdev);
+	rvt_unregister_device(&dd->verbs_dev.rdi);
 
 	if (!list_empty(&dev->txwait))
 		dd_dev_err(dd, "txwait list not empty!\n");
 	if (!list_empty(&dev->memwait))
 		dd_dev_err(dd, "memwait list not empty!\n");
-	if (dev->dma_mr)
-		dd_dev_err(dd, "DMA MR not NULL!\n");
 
-	hfi1_qp_exit(dev);
 	del_timer_sync(&dev->mem_timer);
-	kmem_cache_destroy(dev->verbs_txreq_cache);
-	vfree(dev->lk_table.table);
+	verbs_txreq_exit(dev);
 }
 
 void hfi1_cnp_rcv(struct hfi1_packet *packet)
@@ -2126,7 +1703,7 @@ void hfi1_cnp_rcv(struct hfi1_packet *packet)
 	struct hfi1_ibport *ibp = &packet->rcd->ppd->ibport_data;
 	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
 	struct hfi1_ib_header *hdr = packet->hdr;
-	struct hfi1_qp *qp = packet->qp;
+	struct rvt_qp *qp = packet->qp;
 	u32 lqpn, rqpn = 0;
 	u16 rlid = 0;
 	u8 sl, sc5, sc4_bit, svc_type;
@@ -2149,7 +1726,7 @@ void hfi1_cnp_rcv(struct hfi1_packet *packet)
 		svc_type = IB_CC_SVCTYPE_UD;
 		break;
 	default:
-		ibp->n_pkt_drops++;
+		ibp->rvp.n_pkt_drops++;
 		return;
 	}
 
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 286e468b0479..6c4670fffdbb 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -1,12 +1,11 @@
 /*
+ * Copyright(c) 2015, 2016 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
  *
  * GPL LICENSE SUMMARY
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of version 2 of the GNU General Public License as
  * published by the Free Software Foundation.
@@ -18,8 +17,6 @@
  *
  * BSD LICENSE
  *
- * Copyright(c) 2015 Intel Corporation.
- *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -59,9 +56,13 @@
 #include <linux/workqueue.h>
 #include <linux/kthread.h>
 #include <linux/completion.h>
+#include <linux/slab.h>
 #include <rdma/ib_pack.h>
 #include <rdma/ib_user_verbs.h>
 #include <rdma/ib_mad.h>
+#include <rdma/rdma_vt.h>
+#include <rdma/rdmavt_qp.h>
+#include <rdma/rdmavt_cq.h>
 
 struct hfi1_ctxtdata;
 struct hfi1_pportdata;
@@ -79,12 +80,6 @@ struct hfi1_packet;
  */
 #define HFI1_UVERBS_ABI_VERSION       2
 
-/*
- * Define an ib_cq_notify value that is not valid so we know when CQ
- * notifications are armed.
- */
-#define IB_CQ_NONE      (IB_CQ_NEXT_COMP + 1)
-
 #define IB_SEQ_NAK	(3 << 29)
 
 /* AETH NAK opcode values */
@@ -95,17 +90,6 @@ struct hfi1_packet;
 #define IB_NAK_REMOTE_OPERATIONAL_ERROR 0x63
 #define IB_NAK_INVALID_RD_REQUEST       0x64
 
-/* Flags for checking QP state (see ib_hfi1_state_ops[]) */
-#define HFI1_POST_SEND_OK                0x01
-#define HFI1_POST_RECV_OK                0x02
-#define HFI1_PROCESS_RECV_OK             0x04
-#define HFI1_PROCESS_SEND_OK             0x08
-#define HFI1_PROCESS_NEXT_SEND_OK        0x10
-#define HFI1_FLUSH_SEND			0x20
-#define HFI1_FLUSH_RECV			0x40
-#define HFI1_PROCESS_OR_FLUSH_SEND \
-	(HFI1_PROCESS_SEND_OK | HFI1_FLUSH_SEND)
-
 /* IB Performance Manager status values */
 #define IB_PMA_SAMPLE_STATUS_DONE       0x00
 #define IB_PMA_SAMPLE_STATUS_STARTED    0x01
@@ -208,341 +192,18 @@ struct hfi1_pio_header {
 } __packed;
 
 /*
- * used for force cacheline alignment for AHG
- */
-struct tx_pio_header {
-	struct hfi1_pio_header phdr;
-} ____cacheline_aligned;
-
-/*
- * There is one struct hfi1_mcast for each multicast GID.
- * All attached QPs are then stored as a list of
- * struct hfi1_mcast_qp.
- */
-struct hfi1_mcast_qp {
-	struct list_head list;
-	struct hfi1_qp *qp;
-};
-
-struct hfi1_mcast {
-	struct rb_node rb_node;
-	union ib_gid mgid;
-	struct list_head qp_list;
-	wait_queue_head_t wait;
-	atomic_t refcount;
-	int n_attached;
-};
-
-/* Protection domain */
-struct hfi1_pd {
-	struct ib_pd ibpd;
-	int user;               /* non-zero if created from user space */
-};
-
-/* Address Handle */
-struct hfi1_ah {
-	struct ib_ah ibah;
-	struct ib_ah_attr attr;
-	atomic_t refcount;
-};
-
-/*
- * This structure is used by hfi1_mmap() to validate an offset
- * when an mmap() request is made.  The vm_area_struct then uses
- * this as its vm_private_data.
- */
-struct hfi1_mmap_info {
-	struct list_head pending_mmaps;
-	struct ib_ucontext *context;
-	void *obj;
-	__u64 offset;
-	struct kref ref;
-	unsigned size;
-};
-
-/*
- * This structure is used to contain the head pointer, tail pointer,
- * and completion queue entries as a single memory allocation so
- * it can be mmap'ed into user space.
- */
-struct hfi1_cq_wc {
-	u32 head;               /* index of next entry to fill */
-	u32 tail;               /* index of next ib_poll_cq() entry */
-	union {
-		/* these are actually size ibcq.cqe + 1 */
-		struct ib_uverbs_wc uqueue[0];
-		struct ib_wc kqueue[0];
-	};
-};
-
-/*
- * The completion queue structure.
- */
-struct hfi1_cq {
-	struct ib_cq ibcq;
-	struct kthread_work comptask;
-	struct hfi1_devdata *dd;
-	spinlock_t lock; /* protect changes in this struct */
-	u8 notify;
-	u8 triggered;
-	struct hfi1_cq_wc *queue;
-	struct hfi1_mmap_info *ip;
-};
-
-/*
- * A segment is a linear region of low physical memory.
- * Used by the verbs layer.
- */
-struct hfi1_seg {
-	void *vaddr;
-	size_t length;
-};
-
-/* The number of hfi1_segs that fit in a page. */
-#define HFI1_SEGSZ     (PAGE_SIZE / sizeof(struct hfi1_seg))
-
-struct hfi1_segarray {
-	struct hfi1_seg segs[HFI1_SEGSZ];
-};
-
-struct hfi1_mregion {
-	struct ib_pd *pd;       /* shares refcnt of ibmr.pd */
-	u64 user_base;          /* User's address for this region */
-	u64 iova;               /* IB start address of this region */
-	size_t length;
-	u32 lkey;
-	u32 offset;             /* offset (bytes) to start of region */
-	int access_flags;
-	u32 max_segs;           /* number of hfi1_segs in all the arrays */
-	u32 mapsz;              /* size of the map array */
-	u8  page_shift;         /* 0 - non unform/non powerof2 sizes */
-	u8  lkey_published;     /* in global table */
-	struct completion comp; /* complete when refcount goes to zero */
-	atomic_t refcount;
-	struct hfi1_segarray *map[0];    /* the segments */
-};
-
-/*
- * These keep track of the copy progress within a memory region.
- * Used by the verbs layer.
- */
-struct hfi1_sge {
-	struct hfi1_mregion *mr;
-	void *vaddr;            /* kernel virtual address of segment */
-	u32 sge_length;         /* length of the SGE */
-	u32 length;             /* remaining length of the segment */
-	u16 m;                  /* current index: mr->map[m] */
-	u16 n;                  /* current index: mr->map[m]->segs[n] */
-};
-
-/* Memory region */
-struct hfi1_mr {
-	struct ib_mr ibmr;
-	struct ib_umem *umem;
-	struct hfi1_mregion mr;  /* must be last */
-};
-
-/*
- * Send work request queue entry.
- * The size of the sg_list is determined when the QP is created and stored
- * in qp->s_max_sge.
- */
-struct hfi1_swqe {
-	union {
-		struct ib_send_wr wr;   /* don't use wr.sg_list */
-		struct ib_rdma_wr rdma_wr;
-		struct ib_atomic_wr atomic_wr;
-		struct ib_ud_wr ud_wr;
-	};
-	u32 psn;                /* first packet sequence number */
-	u32 lpsn;               /* last packet sequence number */
-	u32 ssn;                /* send sequence number */
-	u32 length;             /* total length of data in sg_list */
-	struct hfi1_sge sg_list[0];
-};
-
-/*
- * Receive work request queue entry.
- * The size of the sg_list is determined when the QP (or SRQ) is created
- * and stored in qp->r_rq.max_sge (or srq->rq.max_sge).
- */
-struct hfi1_rwqe {
-	u64 wr_id;
-	u8 num_sge;
-	struct ib_sge sg_list[0];
-};
-
-/*
- * This structure is used to contain the head pointer, tail pointer,
- * and receive work queue entries as a single memory allocation so
- * it can be mmap'ed into user space.
- * Note that the wq array elements are variable size so you can't
- * just index into the array to get the N'th element;
- * use get_rwqe_ptr() instead.
+ * hfi1 specific data structures that will be hidden from rvt after the queue
+ * pair is made common
  */
-struct hfi1_rwq {
-	u32 head;               /* new work requests posted to the head */
-	u32 tail;               /* receives pull requests from here. */
-	struct hfi1_rwqe wq[0];
-};
-
-struct hfi1_rq {
-	struct hfi1_rwq *wq;
-	u32 size;               /* size of RWQE array */
-	u8 max_sge;
-	/* protect changes in this struct */
-	spinlock_t lock ____cacheline_aligned_in_smp;
-};
-
-struct hfi1_srq {
-	struct ib_srq ibsrq;
-	struct hfi1_rq rq;
-	struct hfi1_mmap_info *ip;
-	/* send signal when number of RWQEs < limit */
-	u32 limit;
-};
-
-struct hfi1_sge_state {
-	struct hfi1_sge *sg_list;      /* next SGE to be used if any */
-	struct hfi1_sge sge;   /* progress state for the current SGE */
-	u32 total_len;
-	u8 num_sge;
-};
-
-/*
- * This structure holds the information that the send tasklet needs
- * to send a RDMA read response or atomic operation.
- */
-struct hfi1_ack_entry {
-	u8 opcode;
-	u8 sent;
-	u32 psn;
-	u32 lpsn;
-	union {
-		struct hfi1_sge rdma_sge;
-		u64 atomic_data;
-	};
-};
-
-/*
- * Variables prefixed with s_ are for the requester (sender).
- * Variables prefixed with r_ are for the responder (receiver).
- * Variables prefixed with ack_ are for responder replies.
- *
- * Common variables are protected by both r_rq.lock and s_lock in that order
- * which only happens in modify_qp() or changing the QP 'state'.
- */
-struct hfi1_qp {
-	struct ib_qp ibqp;
-	/* read mostly fields above and below */
-	struct ib_ah_attr remote_ah_attr;
-	struct ib_ah_attr alt_ah_attr;
-	struct hfi1_qp __rcu *next;           /* link list for QPN hash table */
-	struct hfi1_swqe *s_wq;  /* send work queue */
-	struct hfi1_mmap_info *ip;
-	struct ahg_ib_header *s_hdr;     /* next packet header to send */
-	/* sc for UC/RC QPs - based on ah for UD */
-	u8 s_sc;
-	unsigned long timeout_jiffies;  /* computed from timeout */
-
-	enum ib_mtu path_mtu;
-	int srate_mbps;		/* s_srate (below) converted to Mbit/s */
-	u32 remote_qpn;
-	u32 pmtu;		/* decoded from path_mtu */
-	u32 qkey;               /* QKEY for this QP (for UD or RD) */
-	u32 s_size;             /* send work queue size */
-	u32 s_rnr_timeout;      /* number of milliseconds for RNR timeout */
-	u32 s_ahgpsn;           /* set to the psn in the copy of the header */
-
-	u8 state;               /* QP state */
-	u8 allowed_ops;		/* high order bits of allowed opcodes */
-	u8 qp_access_flags;
-	u8 alt_timeout;         /* Alternate path timeout for this QP */
-	u8 timeout;             /* Timeout for this QP */
-	u8 s_srate;
-	u8 s_mig_state;
-	u8 port_num;
-	u8 s_pkey_index;        /* PKEY index to use */
-	u8 s_alt_pkey_index;    /* Alternate path PKEY index to use */
-	u8 r_max_rd_atomic;     /* max number of RDMA read/atomic to receive */
-	u8 s_max_rd_atomic;     /* max number of RDMA read/atomic to send */
-	u8 s_retry_cnt;         /* number of times to retry */
-	u8 s_rnr_retry_cnt;
-	u8 r_min_rnr_timer;     /* retry timeout value for RNR NAKs */
-	u8 s_max_sge;           /* size of s_wq->sg_list */
-	u8 s_draining;
-
-	/* start of read/write fields */
-	atomic_t refcount ____cacheline_aligned_in_smp;
-	wait_queue_head_t wait;
-
-
-	struct hfi1_ack_entry s_ack_queue[HFI1_MAX_RDMA_ATOMIC + 1]
-		____cacheline_aligned_in_smp;
-	struct hfi1_sge_state s_rdma_read_sge;
-
-	spinlock_t r_lock ____cacheline_aligned_in_smp;      /* used for APM */
-	unsigned long r_aflags;
-	u64 r_wr_id;            /* ID for current receive WQE */
-	u32 r_ack_psn;          /* PSN for next ACK or atomic ACK */
-	u32 r_len;              /* total length of r_sge */
-	u32 r_rcv_len;          /* receive data len processed */
-	u32 r_psn;              /* expected rcv packet sequence number */
-	u32 r_msn;              /* message sequence number */
-
-	u8 r_adefered;         /* number of acks defered */
-	u8 r_state;             /* opcode of last packet received */
-	u8 r_flags;
-	u8 r_head_ack_queue;    /* index into s_ack_queue[] */
-
-	struct list_head rspwait;       /* link for waiting to respond */
-
-	struct hfi1_sge_state r_sge;     /* current receive data */
-	struct hfi1_rq r_rq;             /* receive work queue */
-
-	spinlock_t s_lock ____cacheline_aligned_in_smp;
-	struct hfi1_sge_state *s_cur_sge;
-	u32 s_flags;
-	struct hfi1_swqe *s_wqe;
-	struct hfi1_sge_state s_sge;     /* current send request data */
-	struct hfi1_mregion *s_rdma_mr;
-	struct sdma_engine *s_sde; /* current sde */
-	u32 s_cur_size;         /* size of send packet in bytes */
-	u32 s_len;              /* total length of s_sge */
-	u32 s_rdma_read_len;    /* total length of s_rdma_read_sge */
-	u32 s_next_psn;         /* PSN for next request */
-	u32 s_last_psn;         /* last response PSN processed */
-	u32 s_sending_psn;      /* lowest PSN that is being sent */
-	u32 s_sending_hpsn;     /* highest PSN that is being sent */
-	u32 s_psn;              /* current packet sequence number */
-	u32 s_ack_rdma_psn;     /* PSN for sending RDMA read responses */
-	u32 s_ack_psn;          /* PSN for acking sends and RDMA writes */
-	u32 s_head;             /* new entries added here */
-	u32 s_tail;             /* next entry to process */
-	u32 s_cur;              /* current work queue entry */
-	u32 s_acked;            /* last un-ACK'ed entry */
-	u32 s_last;             /* last completed entry */
-	u32 s_ssn;              /* SSN of tail entry */
-	u32 s_lsn;              /* limit sequence number (credit) */
-	u16 s_hdrwords;         /* size of s_hdr in 32 bit words */
-	u16 s_rdma_ack_cnt;
-	s8 s_ahgidx;
-	u8 s_state;             /* opcode of last packet sent */
-	u8 s_ack_state;         /* opcode of packet to ACK */
-	u8 s_nak_state;         /* non-zero if NAK is pending */
-	u8 r_nak_state;         /* non-zero if NAK is pending */
-	u8 s_retry;             /* requester retry counter */
-	u8 s_rnr_retry;         /* requester RNR retry counter */
-	u8 s_num_rd_atomic;     /* number of RDMA read/atomic pending */
-	u8 s_tail_ack_queue;    /* index into s_ack_queue[] */
-
-	struct hfi1_sge_state s_ack_rdma_sge;
-	struct timer_list s_timer;
-
+struct hfi1_qp_priv {
+	struct ahg_ib_header *s_hdr;              /* next header to send */
+	struct sdma_engine *s_sde;                /* current sde */
+	struct send_context *s_sendcontext;       /* current sendcontext */
+	u8 s_sc;		                  /* SC[0..4] for next packet */
+	u8 r_adefered;                            /* number of acks defered */
 	struct iowait s_iowait;
-
-	struct hfi1_sge r_sg_list[0] /* verified SGEs */
-		____cacheline_aligned_in_smp;
+	struct timer_list s_rnr_timer;
+	struct rvt_qp *owner;
 };
 
 /*
@@ -553,123 +214,11 @@ struct hfi1_pkt_state {
 	struct hfi1_ibdev *dev;
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
+	struct verbs_txreq *s_txreq;
 };
 
-/*
- * Atomic bit definitions for r_aflags.
- */
-#define HFI1_R_WRID_VALID        0
-#define HFI1_R_REWIND_SGE        1
-
-/*
- * Bit definitions for r_flags.
- */
-#define HFI1_R_REUSE_SGE       0x01
-#define HFI1_R_RDMAR_SEQ       0x02
-/* defer ack until end of interrupt session */
-#define HFI1_R_RSP_DEFERED_ACK 0x04
-/* relay ack to send engine */
-#define HFI1_R_RSP_SEND        0x08
-#define HFI1_R_COMM_EST        0x10
-
-/*
- * Bit definitions for s_flags.
- *
- * HFI1_S_SIGNAL_REQ_WR - set if QP send WRs contain completion signaled
- * HFI1_S_BUSY - send tasklet is processing the QP
- * HFI1_S_TIMER - the RC retry timer is active
- * HFI1_S_ACK_PENDING - an ACK is waiting to be sent after RDMA read/atomics
- * HFI1_S_WAIT_FENCE - waiting for all prior RDMA read or atomic SWQEs
- *                         before processing the next SWQE
- * HFI1_S_WAIT_RDMAR - waiting for a RDMA read or atomic SWQE to complete
- *                         before processing the next SWQE
- * HFI1_S_WAIT_RNR - waiting for RNR timeout
- * HFI1_S_WAIT_SSN_CREDIT - waiting for RC credits to process next SWQE
- * HFI1_S_WAIT_DMA - waiting for send DMA queue to drain before generating
- *                  next send completion entry not via send DMA
- * HFI1_S_WAIT_PIO - waiting for a send buffer to be available
- * HFI1_S_WAIT_TX - waiting for a struct verbs_txreq to be available
- * HFI1_S_WAIT_DMA_DESC - waiting for DMA descriptors to be available
- * HFI1_S_WAIT_KMEM - waiting for kernel memory to be available
- * HFI1_S_WAIT_PSN - waiting for a packet to exit the send DMA queue
- * HFI1_S_WAIT_ACK - waiting for an ACK packet before sending more requests
- * HFI1_S_SEND_ONE - send one packet, request ACK, then wait for ACK
- * HFI1_S_ECN - a BECN was queued to the send engine
- */
-#define HFI1_S_SIGNAL_REQ_WR	0x0001
-#define HFI1_S_BUSY		0x0002
-#define HFI1_S_TIMER		0x0004
-#define HFI1_S_RESP_PENDING	0x0008
-#define HFI1_S_ACK_PENDING	0x0010
-#define HFI1_S_WAIT_FENCE	0x0020
-#define HFI1_S_WAIT_RDMAR	0x0040
-#define HFI1_S_WAIT_RNR		0x0080
-#define HFI1_S_WAIT_SSN_CREDIT	0x0100
-#define HFI1_S_WAIT_DMA		0x0200
-#define HFI1_S_WAIT_PIO		0x0400
-#define HFI1_S_WAIT_TX		0x0800
-#define HFI1_S_WAIT_DMA_DESC	0x1000
-#define HFI1_S_WAIT_KMEM		0x2000
-#define HFI1_S_WAIT_PSN		0x4000
-#define HFI1_S_WAIT_ACK		0x8000
-#define HFI1_S_SEND_ONE		0x10000
-#define HFI1_S_UNLIMITED_CREDIT	0x20000
-#define HFI1_S_AHG_VALID		0x40000
-#define HFI1_S_AHG_CLEAR		0x80000
-#define HFI1_S_ECN		0x100000
-
-/*
- * Wait flags that would prevent any packet type from being sent.
- */
-#define HFI1_S_ANY_WAIT_IO (HFI1_S_WAIT_PIO | HFI1_S_WAIT_TX | \
-	HFI1_S_WAIT_DMA_DESC | HFI1_S_WAIT_KMEM)
-
-/*
- * Wait flags that would prevent send work requests from making progress.
- */
-#define HFI1_S_ANY_WAIT_SEND (HFI1_S_WAIT_FENCE | HFI1_S_WAIT_RDMAR | \
-	HFI1_S_WAIT_RNR | HFI1_S_WAIT_SSN_CREDIT | HFI1_S_WAIT_DMA | \
-	HFI1_S_WAIT_PSN | HFI1_S_WAIT_ACK)
-
-#define HFI1_S_ANY_WAIT (HFI1_S_ANY_WAIT_IO | HFI1_S_ANY_WAIT_SEND)
-
 #define HFI1_PSN_CREDIT  16
 
-/*
- * Since struct hfi1_swqe is not a fixed size, we can't simply index into
- * struct hfi1_qp.s_wq.  This function does the array index computation.
- */
-static inline struct hfi1_swqe *get_swqe_ptr(struct hfi1_qp *qp,
-					     unsigned n)
-{
-	return (struct hfi1_swqe *)((char *)qp->s_wq +
-				     (sizeof(struct hfi1_swqe) +
-				      qp->s_max_sge *
-				      sizeof(struct hfi1_sge)) * n);
-}
-
-/*
- * Since struct hfi1_rwqe is not a fixed size, we can't simply index into
- * struct hfi1_rwq.wq.  This function does the array index computation.
- */
-static inline struct hfi1_rwqe *get_rwqe_ptr(struct hfi1_rq *rq, unsigned n)
-{
-	return (struct hfi1_rwqe *)
-		((char *) rq->wq->wq +
-		 (sizeof(struct hfi1_rwqe) +
-		  rq->max_sge * sizeof(struct ib_sge)) * n);
-}
-
-#define MAX_LKEY_TABLE_BITS 23
-
-struct hfi1_lkey_table {
-	spinlock_t lock; /* protect changes in this struct */
-	u32 next;               /* next unused index (speeds search) */
-	u32 gen;                /* generation count */
-	u32 max;                /* size of the table */
-	struct hfi1_mregion __rcu **table;
-};
-
 struct hfi1_opcode_stats {
 	u64 n_packets;          /* number of packets */
 	u64 n_bytes;            /* total number of bytes */
@@ -690,75 +239,20 @@ static inline void inc_opstats(
 }
 
 struct hfi1_ibport {
-	struct hfi1_qp __rcu *qp[2];
-	struct ib_mad_agent *send_agent;	/* agent for SMI (traps) */
-	struct hfi1_ah *sm_ah;
-	struct hfi1_ah *smi_ah;
-	struct rb_root mcast_tree;
-	spinlock_t lock;		/* protect changes in this struct */
-
-	/* non-zero when timer is set */
-	unsigned long mkey_lease_timeout;
-	unsigned long trap_timeout;
-	__be64 gid_prefix;      /* in network order */
-	__be64 mkey;
+	struct rvt_qp __rcu *qp[2];
+	struct rvt_ibport rvp;
+
 	__be64 guids[HFI1_GUIDS_PER_PORT	- 1];	/* writable GUIDs */
-	u64 tid;		/* TID for traps */
-	u64 n_rc_resends;
-	u64 n_seq_naks;
-	u64 n_rdma_seq;
-	u64 n_rnr_naks;
-	u64 n_other_naks;
-	u64 n_loop_pkts;
-	u64 n_pkt_drops;
-	u64 n_vl15_dropped;
-	u64 n_rc_timeouts;
-	u64 n_dmawait;
-	u64 n_unaligned;
-	u64 n_rc_dupreq;
-	u64 n_rc_seqnak;
-
-	/* Hot-path per CPU counters to avoid cacheline trading to update */
-	u64 z_rc_acks;
-	u64 z_rc_qacks;
-	u64 z_rc_delayed_comp;
-	u64 __percpu *rc_acks;
-	u64 __percpu *rc_qacks;
-	u64 __percpu *rc_delayed_comp;
-
-	u32 port_cap_flags;
-	u32 pma_sample_start;
-	u32 pma_sample_interval;
-	__be16 pma_counter_select[5];
-	u16 pma_tag;
-	u16 pkey_violations;
-	u16 qkey_violations;
-	u16 mkey_violations;
-	u16 mkey_lease_period;
-	u16 sm_lid;
-	u16 repress_traps;
-	u8 sm_sl;
-	u8 mkeyprot;
-	u8 subnet_timeout;
-	u8 vl_high_limit;
+
 	/* the first 16 entries are sl_to_vl for !OPA */
 	u8 sl_to_sc[32];
 	u8 sc_to_sl[32];
 };
 
-
-struct hfi1_qp_ibdev;
 struct hfi1_ibdev {
-	struct ib_device ibdev;
-	struct list_head pending_mmaps;
-	spinlock_t mmap_offset_lock; /* protect mmap_offset */
-	u32 mmap_offset;
-	struct hfi1_mregion __rcu *dma_mr;
-
-	struct hfi1_qp_ibdev *qp_dev;
+	struct rvt_dev_info rdi; /* Must be first */
 
 	/* QP numbers are shared by all IB ports */
-	struct hfi1_lkey_table lk_table;
 	/* protect wait lists */
 	seqlock_t iowait_lock;
 	struct list_head txwait;        /* list for wait verbs_txreq */
@@ -767,26 +261,11 @@ struct hfi1_ibdev {
 	struct kmem_cache *verbs_txreq_cache;
 	struct timer_list mem_timer;
 
-	/* other waiters */
-	spinlock_t pending_lock;
-
 	u64 n_piowait;
+	u64 n_piodrain;
 	u64 n_txwait;
 	u64 n_kmem_wait;
-	u64 n_send_schedule;
-
-	u32 n_pds_allocated;    /* number of PDs allocated for device */
-	spinlock_t n_pds_lock;
-	u32 n_ahs_allocated;    /* number of AHs allocated for device */
-	spinlock_t n_ahs_lock;
-	u32 n_cqs_allocated;    /* number of CQs allocated for device */
-	spinlock_t n_cqs_lock;
-	u32 n_qps_allocated;    /* number of QPs allocated for device */
-	spinlock_t n_qps_lock;
-	u32 n_srqs_allocated;   /* number of SRQs allocated for device */
-	spinlock_t n_srqs_lock;
-	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
-	spinlock_t n_mcast_grps_lock;
+
 #ifdef CONFIG_DEBUG_FS
 	/* per HFI debugfs */
 	struct dentry *hfi1_ibdev_dbg;
@@ -795,66 +274,31 @@ struct hfi1_ibdev {
 #endif
 };
 
-struct hfi1_verbs_counters {
-	u64 symbol_error_counter;
-	u64 link_error_recovery_counter;
-	u64 link_downed_counter;
-	u64 port_rcv_errors;
-	u64 port_rcv_remphys_errors;
-	u64 port_xmit_discards;
-	u64 port_xmit_data;
-	u64 port_rcv_data;
-	u64 port_xmit_packets;
-	u64 port_rcv_packets;
-	u32 local_link_integrity_errors;
-	u32 excessive_buffer_overrun_errors;
-	u32 vl15_dropped;
-};
-
-static inline struct hfi1_mr *to_imr(struct ib_mr *ibmr)
-{
-	return container_of(ibmr, struct hfi1_mr, ibmr);
-}
-
-static inline struct hfi1_pd *to_ipd(struct ib_pd *ibpd)
-{
-	return container_of(ibpd, struct hfi1_pd, ibpd);
-}
-
-static inline struct hfi1_ah *to_iah(struct ib_ah *ibah)
-{
-	return container_of(ibah, struct hfi1_ah, ibah);
-}
-
-static inline struct hfi1_cq *to_icq(struct ib_cq *ibcq)
+static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
 {
-	return container_of(ibcq, struct hfi1_cq, ibcq);
-}
+	struct rvt_dev_info *rdi;
 
-static inline struct hfi1_srq *to_isrq(struct ib_srq *ibsrq)
-{
-	return container_of(ibsrq, struct hfi1_srq, ibsrq);
+	rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
+	return container_of(rdi, struct hfi1_ibdev, rdi);
 }
 
-static inline struct hfi1_qp *to_iqp(struct ib_qp *ibqp)
+static inline struct rvt_qp *iowait_to_qp(struct  iowait *s_iowait)
 {
-	return container_of(ibqp, struct hfi1_qp, ibqp);
-}
+	struct hfi1_qp_priv *priv;
 
-static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
-{
-	return container_of(ibdev, struct hfi1_ibdev, ibdev);
+	priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait);
+	return priv->owner;
 }
 
 /*
  * Send if not busy or waiting for I/O and either
  * a RC response is pending or we can process send work requests.
  */
-static inline int hfi1_send_ok(struct hfi1_qp *qp)
+static inline int hfi1_send_ok(struct rvt_qp *qp)
 {
-	return !(qp->s_flags & (HFI1_S_BUSY | HFI1_S_ANY_WAIT_IO)) &&
-		(qp->s_hdrwords || (qp->s_flags & HFI1_S_RESP_PENDING) ||
-		 !(qp->s_flags & HFI1_S_ANY_WAIT_SEND));
+	return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) &&
+		(qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) ||
+		 !(qp->s_flags & RVT_S_ANY_WAIT_SEND));
 }
 
 /*
@@ -862,7 +306,7 @@ static inline int hfi1_send_ok(struct hfi1_qp *qp)
  */
 void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
 		    u32 qp1, u32 qp2, u16 lid1, u16 lid2);
-void hfi1_cap_mask_chg(struct hfi1_ibport *ibp);
+void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num);
 void hfi1_sys_guid_chg(struct hfi1_ibport *ibp);
 void hfi1_node_desc_chg(struct hfi1_ibport *ibp);
 int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
@@ -870,8 +314,6 @@ int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
 		     const struct ib_mad_hdr *in_mad, size_t in_mad_size,
 		     struct ib_mad_hdr *out_mad, size_t *out_mad_size,
 		     u16 *out_mad_pkey_index);
-int hfi1_create_agents(struct hfi1_ibdev *dev);
-void hfi1_free_agents(struct hfi1_ibdev *dev);
 
 /*
  * The PSN_MASK and PSN_SHIFT allow for
@@ -901,7 +343,7 @@ void hfi1_free_agents(struct hfi1_ibdev *dev);
  */
 static inline int cmp_msn(u32 a, u32 b)
 {
-	return (((int) a) - ((int) b)) << 8;
+	return (((int)a) - ((int)b)) << 8;
 }
 
 /*
@@ -910,7 +352,7 @@ static inline int cmp_msn(u32 a, u32 b)
  */
 static inline int cmp_psn(u32 a, u32 b)
 {
-	return (((int) a) - ((int) b)) << PSN_SHIFT;
+	return (((int)a) - ((int)b)) << PSN_SHIFT;
 }
 
 /*
@@ -929,23 +371,15 @@ static inline u32 delta_psn(u32 a, u32 b)
 	return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT;
 }
 
-struct hfi1_mcast *hfi1_mcast_find(struct hfi1_ibport *ibp, union ib_gid *mgid);
-
-int hfi1_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
-
-int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
-
-int hfi1_mcast_tree_empty(struct hfi1_ibport *ibp);
-
 struct verbs_txreq;
 void hfi1_put_txreq(struct verbs_txreq *tx);
 
-int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps);
+int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 
-void hfi1_copy_sge(struct hfi1_sge_state *ss, void *data, u32 length,
-		   int release);
+void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
+		   int release, int copy_last);
 
-void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release);
+void hfi1_skip_sge(struct rvt_sge_state *ss, u32 length, int release);
 
 void hfi1_cnp_rcv(struct hfi1_packet *packet);
 
@@ -957,147 +391,75 @@ void hfi1_rc_hdrerr(
 	struct hfi1_ctxtdata *rcd,
 	struct hfi1_ib_header *hdr,
 	u32 rcv_flags,
-	struct hfi1_qp *qp);
+	struct rvt_qp *qp);
 
 u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
 
-int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
-
 struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid);
 
 void hfi1_rc_rnr_retry(unsigned long arg);
+void hfi1_add_rnr_timer(struct rvt_qp *qp, u32 to);
+void hfi1_rc_timeout(unsigned long arg);
+void hfi1_del_timers_sync(struct rvt_qp *qp);
+void hfi1_stop_rc_timers(struct rvt_qp *qp);
 
-void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr);
+void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_ib_header *hdr);
 
-void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err);
+void hfi1_rc_error(struct rvt_qp *qp, enum ib_wc_status err);
 
 void hfi1_ud_rcv(struct hfi1_packet *packet);
 
 int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
 
-int hfi1_alloc_lkey(struct hfi1_mregion *mr, int dma_region);
-
-void hfi1_free_lkey(struct hfi1_mregion *mr);
-
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct hfi1_pd *pd,
-		 struct hfi1_sge *isge, struct ib_sge *sge, int acc);
-
-int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
-		 u32 len, u64 vaddr, u32 rkey, int acc);
-
-int hfi1_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
-			  struct ib_recv_wr **bad_wr);
-
-struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
-			       struct ib_srq_init_attr *srq_init_attr,
-			       struct ib_udata *udata);
-
-int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
-		    enum ib_srq_attr_mask attr_mask,
-		    struct ib_udata *udata);
-
-int hfi1_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr);
-
-int hfi1_destroy_srq(struct ib_srq *ibsrq);
-
-int hfi1_cq_init(struct hfi1_devdata *dd);
-
-void hfi1_cq_exit(struct hfi1_devdata *dd);
-
-void hfi1_cq_enter(struct hfi1_cq *cq, struct ib_wc *entry, int sig);
-
-int hfi1_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
-
-struct ib_cq *hfi1_create_cq(
-	struct ib_device *ibdev,
-	const struct ib_cq_init_attr *attr,
-	struct ib_ucontext *context,
-	struct ib_udata *udata);
-
-int hfi1_destroy_cq(struct ib_cq *ibcq);
-
-int hfi1_req_notify_cq(
-	struct ib_cq *ibcq,
-	enum ib_cq_notify_flags notify_flags);
-
-int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
-
-struct ib_mr *hfi1_get_dma_mr(struct ib_pd *pd, int acc);
-
-struct ib_mr *hfi1_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
-			       u64 virt_addr, int mr_access_flags,
-			       struct ib_udata *udata);
+int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only);
 
-int hfi1_dereg_mr(struct ib_mr *ibmr);
+void hfi1_migrate_qp(struct rvt_qp *qp);
 
-struct ib_mr *hfi1_alloc_mr(struct ib_pd *pd,
-			    enum ib_mr_type mr_type,
-			    u32 max_entries);
+int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
+			 int attr_mask, struct ib_udata *udata);
 
-struct ib_fmr *hfi1_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
-			      struct ib_fmr_attr *fmr_attr);
+void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
+		    int attr_mask, struct ib_udata *udata);
 
-int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
-		      int list_len, u64 iova);
+int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
 
-int hfi1_unmap_fmr(struct list_head *fmr_list);
+extern const u32 rc_only_opcode;
+extern const u32 uc_only_opcode;
 
-int hfi1_dealloc_fmr(struct ib_fmr *ibfmr);
-
-static inline void hfi1_get_mr(struct hfi1_mregion *mr)
-{
-	atomic_inc(&mr->refcount);
-}
-
-static inline void hfi1_put_mr(struct hfi1_mregion *mr)
+static inline u8 get_opcode(struct hfi1_ib_header *h)
 {
-	if (unlikely(atomic_dec_and_test(&mr->refcount)))
-		complete(&mr->comp);
-}
+	u16 lnh = be16_to_cpu(h->lrh[0]) & 3;
 
-static inline void hfi1_put_ss(struct hfi1_sge_state *ss)
-{
-	while (ss->num_sge) {
-		hfi1_put_mr(ss->sge.mr);
-		if (--ss->num_sge)
-			ss->sge = *ss->sg_list++;
-	}
+	if (lnh == IB_LNH_IBA_LOCAL)
+		return be32_to_cpu(h->u.oth.bth[0]) >> 24;
+	else
+		return be32_to_cpu(h->u.l.oth.bth[0]) >> 24;
 }
 
-void hfi1_release_mmap_info(struct kref *ref);
-
-struct hfi1_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev, u32 size,
-					     struct ib_ucontext *context,
-					     void *obj);
-
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct hfi1_mmap_info *ip,
-			   u32 size, void *obj);
-
-int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
-
-int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only);
-
 int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
-		       int has_grh, struct hfi1_qp *qp, u32 bth0);
+		       int has_grh, struct rvt_qp *qp, u32 bth0);
 
 u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
 		  struct ib_global_route *grh, u32 hwords, u32 nwords);
 
-void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
-			  u32 bth0, u32 bth2, int middle);
+void hfi1_make_ruc_header(struct rvt_qp *qp, struct hfi1_other_headers *ohdr,
+			  u32 bth0, u32 bth2, int middle,
+			  struct hfi1_pkt_state *ps);
 
-void hfi1_do_send(struct work_struct *work);
+void _hfi1_do_send(struct work_struct *work);
 
-void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
+void hfi1_do_send(struct rvt_qp *qp);
+
+void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 			enum ib_wc_status status);
 
-void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct hfi1_qp *qp, int is_fecn);
+void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct rvt_qp *qp, int is_fecn);
 
-int hfi1_make_rc_req(struct hfi1_qp *qp);
+int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 
-int hfi1_make_uc_req(struct hfi1_qp *qp);
+int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 
-int hfi1_make_ud_req(struct hfi1_qp *qp);
+int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 
 int hfi1_register_ib_device(struct hfi1_devdata *);
 
@@ -1107,24 +469,42 @@ void hfi1_ib_rcv(struct hfi1_packet *packet);
 
 unsigned hfi1_get_npkeys(struct hfi1_devdata *);
 
-int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			u64 pbc);
 
-int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			u64 pbc);
 
-struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5);
+int hfi1_wss_init(void);
+void hfi1_wss_exit(void);
+
+/* platform specific: return the lowest level cache (llc) size, in KiB */
+static inline int wss_llc_size(void)
+{
+	/* assume that the boot CPU value is universal for all CPUs */
+	return boot_cpu_data.x86_cache_size;
+}
+
+/* platform specific: cacheless copy */
+static inline void cacheless_memcpy(void *dst, void *src, size_t n)
+{
+	/*
+	 * Use the only available X64 cacheless copy.  Add a __user cast
+	 * to quiet sparse.  The src agument is already in the kernel so
+	 * there are no security issues.  The extra fault recovery machinery
+	 * is not invoked.
+	 */
+	__copy_user_nocache(dst, (void __user *)src, n, 0);
+}
 
 extern const enum ib_wc_opcode ib_hfi1_wc_opcode[];
 
 extern const u8 hdr_len_by_opcode[];
 
-extern const int ib_hfi1_state_ops[];
+extern const int ib_rvt_state_ops[];
 
 extern __be64 ib_hfi1_sys_image_guid;    /* in network order */
 
-extern unsigned int hfi1_lkey_table_size;
-
 extern unsigned int hfi1_max_cqes;
 
 extern unsigned int hfi1_max_cqs;
@@ -1145,8 +525,8 @@ extern unsigned int hfi1_max_srq_sges;
 
 extern unsigned int hfi1_max_srq_wrs;
 
-extern const u32 ib_hfi1_rnr_table[];
+extern unsigned short piothreshold;
 
-extern struct ib_dma_mapping_ops hfi1_dma_mapping_ops;
+extern const u32 ib_hfi1_rnr_table[];
 
 #endif                          /* HFI1_VERBS_H */
diff --git a/drivers/staging/rdma/hfi1/verbs_txreq.c b/drivers/staging/rdma/hfi1/verbs_txreq.c
new file mode 100644
index 000000000000..bc95c4112c61
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/verbs_txreq.c
@@ -0,0 +1,149 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include "hfi.h"
+#include "verbs_txreq.h"
+#include "qp.h"
+#include "trace.h"
+
+#define TXREQ_LEN 24
+
+void hfi1_put_txreq(struct verbs_txreq *tx)
+{
+	struct hfi1_ibdev *dev;
+	struct rvt_qp *qp;
+	unsigned long flags;
+	unsigned int seq;
+	struct hfi1_qp_priv *priv;
+
+	qp = tx->qp;
+	dev = to_idev(qp->ibqp.device);
+
+	if (tx->mr)
+		rvt_put_mr(tx->mr);
+
+	sdma_txclean(dd_from_dev(dev), &tx->txreq);
+
+	/* Free verbs_txreq and return to slab cache */
+	kmem_cache_free(dev->verbs_txreq_cache, tx);
+
+	do {
+		seq = read_seqbegin(&dev->iowait_lock);
+		if (!list_empty(&dev->txwait)) {
+			struct iowait *wait;
+
+			write_seqlock_irqsave(&dev->iowait_lock, flags);
+			wait = list_first_entry(&dev->txwait, struct iowait,
+						list);
+			qp = iowait_to_qp(wait);
+			priv = qp->priv;
+			list_del_init(&priv->s_iowait.list);
+			/* refcount held until actual wake up */
+			write_sequnlock_irqrestore(&dev->iowait_lock, flags);
+			hfi1_qp_wakeup(qp, RVT_S_WAIT_TX);
+			break;
+		}
+	} while (read_seqretry(&dev->iowait_lock, seq));
+}
+
+struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
+				struct rvt_qp *qp)
+{
+	struct verbs_txreq *tx = ERR_PTR(-EBUSY);
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	write_seqlock(&dev->iowait_lock);
+	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
+		struct hfi1_qp_priv *priv;
+
+		tx = kmem_cache_alloc(dev->verbs_txreq_cache, GFP_ATOMIC);
+		if (tx)
+			goto out;
+		priv = qp->priv;
+		if (list_empty(&priv->s_iowait.list)) {
+			dev->n_txwait++;
+			qp->s_flags |= RVT_S_WAIT_TX;
+			list_add_tail(&priv->s_iowait.list, &dev->txwait);
+			trace_hfi1_qpsleep(qp, RVT_S_WAIT_TX);
+			atomic_inc(&qp->refcount);
+		}
+		qp->s_flags &= ~RVT_S_BUSY;
+	}
+out:
+	write_sequnlock(&dev->iowait_lock);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return tx;
+}
+
+static void verbs_txreq_kmem_cache_ctor(void *obj)
+{
+	struct verbs_txreq *tx = (struct verbs_txreq *)obj;
+
+	memset(tx, 0, sizeof(*tx));
+}
+
+int verbs_txreq_init(struct hfi1_ibdev *dev)
+{
+	char buf[TXREQ_LEN];
+	struct hfi1_devdata *dd = dd_from_dev(dev);
+
+	snprintf(buf, sizeof(buf), "hfi1_%u_vtxreq_cache", dd->unit);
+	dev->verbs_txreq_cache = kmem_cache_create(buf,
+						   sizeof(struct verbs_txreq),
+						   0, SLAB_HWCACHE_ALIGN,
+						   verbs_txreq_kmem_cache_ctor);
+	if (!dev->verbs_txreq_cache)
+		return -ENOMEM;
+	return 0;
+}
+
+void verbs_txreq_exit(struct hfi1_ibdev *dev)
+{
+	kmem_cache_destroy(dev->verbs_txreq_cache);
+	dev->verbs_txreq_cache = NULL;
+}
diff --git a/drivers/staging/rdma/hfi1/verbs_txreq.h b/drivers/staging/rdma/hfi1/verbs_txreq.h
new file mode 100644
index 000000000000..1cf69b2fe4a5
--- /dev/null
+++ b/drivers/staging/rdma/hfi1/verbs_txreq.h
@@ -0,0 +1,116 @@
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef HFI1_VERBS_TXREQ_H
+#define HFI1_VERBS_TXREQ_H
+
+#include <linux/types.h>
+#include <linux/slab.h>
+
+#include "verbs.h"
+#include "sdma_txreq.h"
+#include "iowait.h"
+
+struct verbs_txreq {
+	struct hfi1_pio_header	phdr;
+	struct sdma_txreq       txreq;
+	struct rvt_qp           *qp;
+	struct rvt_swqe         *wqe;
+	struct rvt_mregion	*mr;
+	struct rvt_sge_state    *ss;
+	struct sdma_engine     *sde;
+	struct send_context     *psc;
+	u16                     hdr_dwords;
+};
+
+struct hfi1_ibdev;
+struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
+				struct rvt_qp *qp);
+
+static inline struct verbs_txreq *get_txreq(struct hfi1_ibdev *dev,
+					    struct rvt_qp *qp)
+{
+	struct verbs_txreq *tx;
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	tx = kmem_cache_alloc(dev->verbs_txreq_cache, GFP_ATOMIC);
+	if (unlikely(!tx)) {
+		/* call slow path to get the lock */
+		tx = __get_txreq(dev, qp);
+		if (IS_ERR(tx))
+			return tx;
+	}
+	tx->qp = qp;
+	tx->mr = NULL;
+	tx->sde = priv->s_sde;
+	tx->psc = priv->s_sendcontext;
+	/* so that we can test if the sdma decriptors are there */
+	tx->txreq.num_desc = 0;
+	return tx;
+}
+
+static inline struct sdma_txreq *get_sdma_txreq(struct verbs_txreq *tx)
+{
+	return &tx->txreq;
+}
+
+static inline struct verbs_txreq *get_waiting_verbs_txreq(struct rvt_qp *qp)
+{
+	struct sdma_txreq *stx;
+	struct hfi1_qp_priv *priv = qp->priv;
+
+	stx = iowait_get_txhead(&priv->s_iowait);
+	if (stx)
+		return container_of(stx, struct verbs_txreq, txreq);
+	return NULL;
+}
+
+void hfi1_put_txreq(struct verbs_txreq *tx);
+int verbs_txreq_init(struct hfi1_ibdev *dev);
+void verbs_txreq_exit(struct hfi1_ibdev *dev);
+
+#endif                         /* HFI1_VERBS_TXREQ_H */