Merge tag 'ntb-5.7' of git://github.com/jonmason/ntb

Pull NTB updates from Jon Mason:
 "Bug fixes for a few printing issues, link status detection bug on AMD
  hardware, and a DMA address issue with ntb_perf.

  Also, large series of AMD NTB patches"

* tag 'ntb-5.7' of git://github.com/jonmason/ntb: (21 commits)
  NTB: add pci shutdown handler for AMD NTB
  NTB: send DB event when driver is loaded or un-loaded
  NTB: remove redundant setting of DB valid mask
  NTB: return link up status correctly for PRI and SEC
  NTB: add helper functions to set and clear sideinfo
  NTB: move ntb_ctrl handling to init and deinit
  NTB: handle link up, D0 and D3 events correctly
  NTB: handle link down event correctly
  NTB: remove handling of peer_sta from amd_link_is_up
  NTB: set peer_sta within event handler itself
  NTB: return the side info status from amd_poll_link
  NTB: define a new function to get link status
  NTB: Enable link up and down event notification
  NTB: clear interrupt status register
  NTB: Fix access to link status and control register
  MAINTAINERS: update maintainer list for AMD NTB driver
  NTB: ntb_transport: Use scnprintf() for avoiding potential buffer overflow
  ntb_hw_switchtec: Fix ntb_mw_clear_trans error if size == 0
  ntb_tool: Fix printk format
  NTB: ntb_perf: Fix address err in perf_copy_chunk
  ...

7 files changed, 332 insertions, 98 deletions

diff --git a/MAINTAINERS b/MAINTAINERS
index 1b718c05ad3e..a2b9bbf6e8b2 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -11896,6 +11896,7 @@ F:	scripts/nsdeps
 F:	Documentation/core-api/symbol-namespaces.rst
 
 NTB AMD DRIVER
+M:	Sanjay R Mehta <sanju.mehta@amd.com>
 M:	Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
 L:	linux-ntb@googlegroups.com
 S:	Supported
diff --git a/drivers/ntb/hw/amd/ntb_hw_amd.c b/drivers/ntb/hw/amd/ntb_hw_amd.c
index e52b300b2f5b..9e310e1ad4d0 100644
--- a/drivers/ntb/hw/amd/ntb_hw_amd.c
+++ b/drivers/ntb/hw/amd/ntb_hw_amd.c
@@ -195,26 +195,117 @@ static int amd_ntb_mw_set_trans(struct ntb_dev *ntb, int pidx, int idx,
 	return 0;
 }
 
-static int amd_link_is_up(struct amd_ntb_dev *ndev)
+static int amd_ntb_get_link_status(struct amd_ntb_dev *ndev)
 {
-	if (!ndev->peer_sta)
-		return NTB_LNK_STA_ACTIVE(ndev->cntl_sta);
+	struct pci_dev *pdev = NULL;
+	struct pci_dev *pci_swds = NULL;
+	struct pci_dev *pci_swus = NULL;
+	u32 stat;
+	int rc;
 
-	if (ndev->peer_sta & AMD_LINK_UP_EVENT) {
-		ndev->peer_sta = 0;
-		return 1;
+	if (ndev->ntb.topo == NTB_TOPO_SEC) {
+		/* Locate the pointer to Downstream Switch for this device */
+		pci_swds = pci_upstream_bridge(ndev->ntb.pdev);
+		if (pci_swds) {
+			/*
+			 * Locate the pointer to Upstream Switch for
+			 * the Downstream Switch.
+			 */
+			pci_swus = pci_upstream_bridge(pci_swds);
+			if (pci_swus) {
+				rc = pcie_capability_read_dword(pci_swus,
+								PCI_EXP_LNKCTL,
+								&stat);
+				if (rc)
+					return 0;
+			} else {
+				return 0;
+			}
+		} else {
+			return 0;
+		}
+	} else if (ndev->ntb.topo == NTB_TOPO_PRI) {
+		/*
+		 * For NTB primary, we simply read the Link Status and control
+		 * register of the NTB device itself.
+		 */
+		pdev = ndev->ntb.pdev;
+		rc = pcie_capability_read_dword(pdev, PCI_EXP_LNKCTL, &stat);
+		if (rc)
+			return 0;
+	} else {
+		/* Catch all for everything else */
+		return 0;
 	}
 
-	/* If peer_sta is reset or D0 event, the ISR has
-	 * started a timer to check link status of hardware.
-	 * So here just clear status bit. And if peer_sta is
-	 * D3 or PME_TO, D0/reset event will be happened when
-	 * system wakeup/poweron, so do nothing here.
+	ndev->lnk_sta = stat;
+
+	return 1;
+}
+
+static int amd_link_is_up(struct amd_ntb_dev *ndev)
+{
+	int ret;
+
+	/*
+	 * We consider the link to be up under two conditions:
+	 *
+	 *   - When a link-up event is received. This is indicated by
+	 *     AMD_LINK_UP_EVENT set in peer_sta.
+	 *   - When driver on both sides of the link have been loaded.
+	 *     This is indicated by bit 1 being set in the peer
+	 *     SIDEINFO register.
+	 *
+	 * This function should return 1 when the latter of the above
+	 * two conditions is true.
+	 *
+	 * Now consider the sequence of events - Link-Up event occurs,
+	 * then the peer side driver loads. In this case, we would have
+	 * received LINK_UP event and bit 1 of peer SIDEINFO is also
+	 * set. What happens now if the link goes down? Bit 1 of
+	 * peer SIDEINFO remains set, but LINK_DOWN bit is set in
+	 * peer_sta. So we should return 0 from this function. Not only
+	 * that, we clear bit 1 of peer SIDEINFO to 0, since the peer
+	 * side driver did not even get a chance to clear it before
+	 * the link went down. This can be the case of surprise link
+	 * removal.
+	 *
+	 * LINK_UP event will always occur before the peer side driver
+	 * gets loaded the very first time. So there can be a case when
+	 * the LINK_UP event has occurred, but the peer side driver hasn't
+	 * yet loaded. We return 0 in that case.
+	 *
+	 * There is also a special case when the primary side driver is
+	 * unloaded and then loaded again. Since there is no change in
+	 * the status of NTB secondary in this case, there is no Link-Up
+	 * or Link-Down notification received. We recognize this condition
+	 * with peer_sta being set to 0.
+	 *
+	 * If bit 1 of peer SIDEINFO register is not set, then we
+	 * simply return 0 irrespective of the link up or down status
+	 * set in peer_sta.
 	 */
-	if (ndev->peer_sta & AMD_PEER_RESET_EVENT)
-		ndev->peer_sta &= ~AMD_PEER_RESET_EVENT;
-	else if (ndev->peer_sta & (AMD_PEER_D0_EVENT | AMD_LINK_DOWN_EVENT))
-		ndev->peer_sta = 0;
+	ret = amd_poll_link(ndev);
+	if (ret) {
+		/*
+		 * We need to check the below only for NTB primary. For NTB
+		 * secondary, simply checking the result of PSIDE_INFO
+		 * register will suffice.
+		 */
+		if (ndev->ntb.topo == NTB_TOPO_PRI) {
+			if ((ndev->peer_sta & AMD_LINK_UP_EVENT) ||
+			    (ndev->peer_sta == 0))
+				return ret;
+			else if (ndev->peer_sta & AMD_LINK_DOWN_EVENT) {
+				/* Clear peer sideinfo register */
+				amd_clear_side_info_reg(ndev, true);
+
+				return 0;
+			}
+		} else { /* NTB_TOPO_SEC */
+			return ret;
+		}
+	}
 
 	return 0;
 }
@@ -253,7 +344,6 @@ static int amd_ntb_link_enable(struct ntb_dev *ntb,
 {
 	struct amd_ntb_dev *ndev = ntb_ndev(ntb);
 	void __iomem *mmio = ndev->self_mmio;
-	u32 ntb_ctl;
 
 	/* Enable event interrupt */
 	ndev->int_mask &= ~AMD_EVENT_INTMASK;
@@ -263,10 +353,6 @@ static int amd_ntb_link_enable(struct ntb_dev *ntb,
 		return -EINVAL;
 	dev_dbg(&ntb->pdev->dev, "Enabling Link.\n");
 
-	ntb_ctl = readl(mmio + AMD_CNTL_OFFSET);
-	ntb_ctl |= (PMM_REG_CTL | SMM_REG_CTL);
-	writel(ntb_ctl, mmio + AMD_CNTL_OFFSET);
-
 	return 0;
 }
 
@@ -274,7 +360,6 @@ static int amd_ntb_link_disable(struct ntb_dev *ntb)
 {
 	struct amd_ntb_dev *ndev = ntb_ndev(ntb);
 	void __iomem *mmio = ndev->self_mmio;
-	u32 ntb_ctl;
 
 	/* Disable event interrupt */
 	ndev->int_mask |= AMD_EVENT_INTMASK;
@@ -284,10 +369,6 @@ static int amd_ntb_link_disable(struct ntb_dev *ntb)
 		return -EINVAL;
 	dev_dbg(&ntb->pdev->dev, "Enabling Link.\n");
 
-	ntb_ctl = readl(mmio + AMD_CNTL_OFFSET);
-	ntb_ctl &= ~(PMM_REG_CTL | SMM_REG_CTL);
-	writel(ntb_ctl, mmio + AMD_CNTL_OFFSET);
-
 	return 0;
 }
 
@@ -493,8 +574,6 @@ static void amd_ack_smu(struct amd_ntb_dev *ndev, u32 bit)
 	reg = readl(mmio + AMD_SMUACK_OFFSET);
 	reg |= bit;
 	writel(reg, mmio + AMD_SMUACK_OFFSET);
-
-	ndev->peer_sta |= bit;
 }
 
 static void amd_handle_event(struct amd_ntb_dev *ndev, int vec)
@@ -512,10 +591,16 @@ static void amd_handle_event(struct amd_ntb_dev *ndev, int vec)
 	status &= AMD_EVENT_INTMASK;
 	switch (status) {
 	case AMD_PEER_FLUSH_EVENT:
+		ndev->peer_sta |= AMD_PEER_FLUSH_EVENT;
 		dev_info(dev, "Flush is done.\n");
 		break;
 	case AMD_PEER_RESET_EVENT:
-		amd_ack_smu(ndev, AMD_PEER_RESET_EVENT);
+	case AMD_LINK_DOWN_EVENT:
+		ndev->peer_sta |= status;
+		if (status == AMD_LINK_DOWN_EVENT)
+			ndev->peer_sta &= ~AMD_LINK_UP_EVENT;
+
+		amd_ack_smu(ndev, status);
 
 		/* link down first */
 		ntb_link_event(&ndev->ntb);
@@ -526,7 +611,12 @@ static void amd_handle_event(struct amd_ntb_dev *ndev, int vec)
 	case AMD_PEER_D3_EVENT:
 	case AMD_PEER_PMETO_EVENT:
 	case AMD_LINK_UP_EVENT:
-	case AMD_LINK_DOWN_EVENT:
+		ndev->peer_sta |= status;
+		if (status == AMD_LINK_UP_EVENT)
+			ndev->peer_sta &= ~AMD_LINK_DOWN_EVENT;
+		else if (status == AMD_PEER_D3_EVENT)
+			ndev->peer_sta &= ~AMD_PEER_D0_EVENT;
+
 		amd_ack_smu(ndev, status);
 
 		/* link down */
@@ -540,6 +630,8 @@ static void amd_handle_event(struct amd_ntb_dev *ndev, int vec)
 		if (status & 0x1)
 			dev_info(dev, "Wakeup is done.\n");
 
+		ndev->peer_sta |= AMD_PEER_D0_EVENT;
+		ndev->peer_sta &= ~AMD_PEER_D3_EVENT;
 		amd_ack_smu(ndev, AMD_PEER_D0_EVENT);
 
 		/* start a timer to poll link status */
@@ -550,6 +642,39 @@ static void amd_handle_event(struct amd_ntb_dev *ndev, int vec)
 		dev_info(dev, "event status = 0x%x.\n", status);
 		break;
 	}
+
+	/* Clear the interrupt status */
+	writel(status, mmio + AMD_INTSTAT_OFFSET);
+}
+
+static void amd_handle_db_event(struct amd_ntb_dev *ndev, int vec)
+{
+	struct device *dev = &ndev->ntb.pdev->dev;
+	u64 status;
+
+	status = amd_ntb_db_read(&ndev->ntb);
+
+	dev_dbg(dev, "status = 0x%llx and vec = %d\n", status, vec);
+
+	/*
+	 * Since we had reserved highest order bit of DB for signaling peer of
+	 * a special event, this is the only status bit we should be concerned
+	 * here now.
+	 */
+	if (status & BIT(ndev->db_last_bit)) {
+		ntb_db_clear(&ndev->ntb, BIT(ndev->db_last_bit));
+		/* send link down event notification */
+		ntb_link_event(&ndev->ntb);
+
+		/*
+		 * If we are here, that means the peer has signalled a special
+		 * event which notifies that the peer driver has been
+		 * un-loaded for some reason. Since there is a chance that the
+		 * peer will load its driver again sometime, we schedule link
+		 * polling routine.
+		 */
+		schedule_delayed_work(&ndev->hb_timer, AMD_LINK_HB_TIMEOUT);
+	}
 }
 
 static irqreturn_t ndev_interrupt(struct amd_ntb_dev *ndev, int vec)
@@ -559,8 +684,10 @@ static irqreturn_t ndev_interrupt(struct amd_ntb_dev *ndev, int vec)
 	if (vec > (AMD_DB_CNT - 1) || (ndev->msix_vec_count == 1))
 		amd_handle_event(ndev, vec);
 
-	if (vec < AMD_DB_CNT)
+	if (vec < AMD_DB_CNT) {
+		amd_handle_db_event(ndev, vec);
 		ntb_db_event(&ndev->ntb, vec);
+	}
 
 	return IRQ_HANDLED;
 }
@@ -842,26 +969,18 @@ static inline void ndev_init_struct(struct amd_ntb_dev *ndev,
 static int amd_poll_link(struct amd_ntb_dev *ndev)
 {
 	void __iomem *mmio = ndev->peer_mmio;
-	u32 reg, stat;
-	int rc;
+	u32 reg;
 
 	reg = readl(mmio + AMD_SIDEINFO_OFFSET);
-	reg &= NTB_LIN_STA_ACTIVE_BIT;
+	reg &= AMD_SIDE_READY;
 
 	dev_dbg(&ndev->ntb.pdev->dev, "%s: reg_val = 0x%x.\n", __func__, reg);
 
-	if (reg == ndev->cntl_sta)
-		return 0;
-
 	ndev->cntl_sta = reg;
 
-	rc = pci_read_config_dword(ndev->ntb.pdev,
-				   AMD_LINK_STATUS_OFFSET, &stat);
-	if (rc)
-		return 0;
-	ndev->lnk_sta = stat;
+	amd_ntb_get_link_status(ndev);
 
-	return 1;
+	return ndev->cntl_sta;
 }
 
 static void amd_link_hb(struct work_struct *work)
@@ -880,11 +999,16 @@ static int amd_init_isr(struct amd_ntb_dev *ndev)
 	return ndev_init_isr(ndev, AMD_DB_CNT, AMD_MSIX_VECTOR_CNT);
 }
 
-static void amd_init_side_info(struct amd_ntb_dev *ndev)
+static void amd_set_side_info_reg(struct amd_ntb_dev *ndev, bool peer)
 {
-	void __iomem *mmio = ndev->self_mmio;
+	void __iomem *mmio = NULL;
 	unsigned int reg;
 
+	if (peer)
+		mmio = ndev->peer_mmio;
+	else
+		mmio = ndev->self_mmio;
+
 	reg = readl(mmio + AMD_SIDEINFO_OFFSET);
 	if (!(reg & AMD_SIDE_READY)) {
 		reg |= AMD_SIDE_READY;
@@ -892,11 +1016,16 @@ static void amd_init_side_info(struct amd_ntb_dev *ndev)
 	}
 }
 
-static void amd_deinit_side_info(struct amd_ntb_dev *ndev)
+static void amd_clear_side_info_reg(struct amd_ntb_dev *ndev, bool peer)
 {
-	void __iomem *mmio = ndev->self_mmio;
+	void __iomem *mmio = NULL;
 	unsigned int reg;
 
+	if (peer)
+		mmio = ndev->peer_mmio;
+	else
+		mmio = ndev->self_mmio;
+
 	reg = readl(mmio + AMD_SIDEINFO_OFFSET);
 	if (reg & AMD_SIDE_READY) {
 		reg &= ~AMD_SIDE_READY;
@@ -905,6 +1034,30 @@ static void amd_deinit_side_info(struct amd_ntb_dev *ndev)
 	}
 }
 
+static void amd_init_side_info(struct amd_ntb_dev *ndev)
+{
+	void __iomem *mmio = ndev->self_mmio;
+	u32 ntb_ctl;
+
+	amd_set_side_info_reg(ndev, false);
+
+	ntb_ctl = readl(mmio + AMD_CNTL_OFFSET);
+	ntb_ctl |= (PMM_REG_CTL | SMM_REG_CTL);
+	writel(ntb_ctl, mmio + AMD_CNTL_OFFSET);
+}
+
+static void amd_deinit_side_info(struct amd_ntb_dev *ndev)
+{
+	void __iomem *mmio = ndev->self_mmio;
+	u32 ntb_ctl;
+
+	amd_clear_side_info_reg(ndev, false);
+
+	ntb_ctl = readl(mmio + AMD_CNTL_OFFSET);
+	ntb_ctl &= ~(PMM_REG_CTL | SMM_REG_CTL);
+	writel(ntb_ctl, mmio + AMD_CNTL_OFFSET);
+}
+
 static int amd_init_ntb(struct amd_ntb_dev *ndev)
 {
 	void __iomem *mmio = ndev->self_mmio;
@@ -935,8 +1088,6 @@ static int amd_init_ntb(struct amd_ntb_dev *ndev)
 		return -EINVAL;
 	}
 
-	ndev->db_valid_mask = BIT_ULL(ndev->db_count) - 1;
-
 	/* Mask event interrupts */
 	writel(ndev->int_mask, mmio + AMD_INTMASK_OFFSET);
 
@@ -957,6 +1108,7 @@ static enum ntb_topo amd_get_topo(struct amd_ntb_dev *ndev)
 
 static int amd_init_dev(struct amd_ntb_dev *ndev)
 {
+	void __iomem *mmio = ndev->self_mmio;
 	struct pci_dev *pdev;
 	int rc = 0;
 
@@ -977,6 +1129,25 @@ static int amd_init_dev(struct amd_ntb_dev *ndev)
 	}
 
 	ndev->db_valid_mask = BIT_ULL(ndev->db_count) - 1;
+	/*
+	 * We reserve the highest order bit of the DB register which will
+	 * be used to notify peer when the driver on this side is being
+	 * un-loaded.
+	 */
+	ndev->db_last_bit =
+			find_last_bit((unsigned long *)&ndev->db_valid_mask,
+				      hweight64(ndev->db_valid_mask));
+	writew((u16)~BIT(ndev->db_last_bit), mmio + AMD_DBMASK_OFFSET);
+	/*
+	 * Since now there is one less bit to account for, the DB count
+	 * and DB mask should be adjusted accordingly.
+	 */
+	ndev->db_count -= 1;
+	ndev->db_valid_mask = BIT_ULL(ndev->db_count) - 1;
+
+	/* Enable Link-Up and Link-Down event interrupts */
+	ndev->int_mask &= ~(AMD_LINK_UP_EVENT | AMD_LINK_DOWN_EVENT);
+	writel(ndev->int_mask, mmio + AMD_INTMASK_OFFSET);
 
 	return 0;
 }
@@ -1111,9 +1282,31 @@ static void amd_ntb_pci_remove(struct pci_dev *pdev)
 {
 	struct amd_ntb_dev *ndev = pci_get_drvdata(pdev);
 
+	/*
+	 * Clear the READY bit in SIDEINFO register before sending DB event
+	 * to the peer. This will make sure that when the peer handles the
+	 * DB event, it correctly reads this bit as being 0.
+	 */
+	amd_deinit_side_info(ndev);
+	ntb_peer_db_set(&ndev->ntb, BIT_ULL(ndev->db_last_bit));
 	ntb_unregister_device(&ndev->ntb);
 	ndev_deinit_debugfs(ndev);
+	amd_deinit_dev(ndev);
+	amd_ntb_deinit_pci(ndev);
+	kfree(ndev);
+}
+
+static void amd_ntb_pci_shutdown(struct pci_dev *pdev)
+{
+	struct amd_ntb_dev *ndev = pci_get_drvdata(pdev);
+
+	/* Send link down notification */
+	ntb_link_event(&ndev->ntb);
+
 	amd_deinit_side_info(ndev);
+	ntb_peer_db_set(&ndev->ntb, BIT_ULL(ndev->db_last_bit));
+	ntb_unregister_device(&ndev->ntb);
+	ndev_deinit_debugfs(ndev);
 	amd_deinit_dev(ndev);
 	amd_ntb_deinit_pci(ndev);
 	kfree(ndev);
@@ -1149,6 +1342,7 @@ static struct pci_driver amd_ntb_pci_driver = {
 	.id_table	= amd_ntb_pci_tbl,
 	.probe		= amd_ntb_pci_probe,
 	.remove		= amd_ntb_pci_remove,
+	.shutdown	= amd_ntb_pci_shutdown,
 };
 
 static int __init amd_ntb_pci_driver_init(void)
diff --git a/drivers/ntb/hw/amd/ntb_hw_amd.h b/drivers/ntb/hw/amd/ntb_hw_amd.h
index 139a307147bc..5f337b1572a0 100644
--- a/drivers/ntb/hw/amd/ntb_hw_amd.h
+++ b/drivers/ntb/hw/amd/ntb_hw_amd.h
@@ -53,11 +53,8 @@
 #include <linux/pci.h>
 
 #define AMD_LINK_HB_TIMEOUT	msecs_to_jiffies(1000)
-#define AMD_LINK_STATUS_OFFSET	0x68
-#define NTB_LIN_STA_ACTIVE_BIT	0x00000002
 #define NTB_LNK_STA_SPEED_MASK	0x000F0000
 #define NTB_LNK_STA_WIDTH_MASK	0x03F00000
-#define NTB_LNK_STA_ACTIVE(x)	(!!((x) & NTB_LIN_STA_ACTIVE_BIT))
 #define NTB_LNK_STA_SPEED(x)	(((x) & NTB_LNK_STA_SPEED_MASK) >> 16)
 #define NTB_LNK_STA_WIDTH(x)	(((x) & NTB_LNK_STA_WIDTH_MASK) >> 20)
 
@@ -196,6 +193,7 @@ struct amd_ntb_dev {
 
 	u64 db_valid_mask;
 	u64 db_mask;
+	u64 db_last_bit;
 	u32 int_mask;
 
 	struct msix_entry *msix;
@@ -218,4 +216,8 @@ struct amd_ntb_dev {
 #define ntb_ndev(__ntb) container_of(__ntb, struct amd_ntb_dev, ntb)
 #define hb_ndev(__work) container_of(__work, struct amd_ntb_dev, hb_timer.work)
 
+static void amd_set_side_info_reg(struct amd_ntb_dev *ndev, bool peer);
+static void amd_clear_side_info_reg(struct amd_ntb_dev *ndev, bool peer);
+static int amd_poll_link(struct amd_ntb_dev *ndev);
+
 #endif
diff --git a/drivers/ntb/hw/mscc/ntb_hw_switchtec.c b/drivers/ntb/hw/mscc/ntb_hw_switchtec.c
index 86ffa716eaf2..4c6eb61a6ac6 100644
--- a/drivers/ntb/hw/mscc/ntb_hw_switchtec.c
+++ b/drivers/ntb/hw/mscc/ntb_hw_switchtec.c
@@ -285,7 +285,7 @@ static int switchtec_ntb_mw_set_trans(struct ntb_dev *ntb, int pidx, int widx,
 	if (widx >= switchtec_ntb_mw_count(ntb, pidx))
 		return -EINVAL;
 
-	if (xlate_pos < 12)
+	if (size != 0 && xlate_pos < 12)
 		return -EINVAL;
 
 	if (!IS_ALIGNED(addr, BIT_ULL(xlate_pos))) {
diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
index 00a5d5764993..e6d1f5b298f3 100644
--- a/drivers/ntb/ntb_transport.c
+++ b/drivers/ntb/ntb_transport.c
@@ -481,70 +481,70 @@ static ssize_t debugfs_read(struct file *filp, char __user *ubuf, size_t count,
 		return -ENOMEM;
 
 	out_offset = 0;
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "\nNTB QP stats:\n\n");
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_bytes - \t%llu\n", qp->rx_bytes);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_pkts - \t%llu\n", qp->rx_pkts);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_memcpy - \t%llu\n", qp->rx_memcpy);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_async - \t%llu\n", qp->rx_async);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_ring_empty - %llu\n", qp->rx_ring_empty);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_err_no_buf - %llu\n", qp->rx_err_no_buf);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_err_oflow - \t%llu\n", qp->rx_err_oflow);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_err_ver - \t%llu\n", qp->rx_err_ver);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_buff - \t0x%p\n", qp->rx_buff);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_index - \t%u\n", qp->rx_index);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_max_entry - \t%u\n", qp->rx_max_entry);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "rx_alloc_entry - \t%u\n\n", qp->rx_alloc_entry);
 
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_bytes - \t%llu\n", qp->tx_bytes);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_pkts - \t%llu\n", qp->tx_pkts);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_memcpy - \t%llu\n", qp->tx_memcpy);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_async - \t%llu\n", qp->tx_async);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_ring_full - \t%llu\n", qp->tx_ring_full);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_err_no_buf - %llu\n", qp->tx_err_no_buf);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_mw - \t0x%p\n", qp->tx_mw);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_index (H) - \t%u\n", qp->tx_index);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "RRI (T) - \t%u\n",
 			       qp->remote_rx_info->entry);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "tx_max_entry - \t%u\n", qp->tx_max_entry);
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "free tx - \t%u\n",
 			       ntb_transport_tx_free_entry(qp));
 
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "\n");
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "Using TX DMA - \t%s\n",
 			       qp->tx_dma_chan ? "Yes" : "No");
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "Using RX DMA - \t%s\n",
 			       qp->rx_dma_chan ? "Yes" : "No");
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "QP Link - \t%s\n",
 			       qp->link_is_up ? "Up" : "Down");
-	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+	out_offset += scnprintf(buf + out_offset, out_count - out_offset,
 			       "\n");
 
 	if (out_offset > out_count)
diff --git a/drivers/ntb/test/ntb_perf.c b/drivers/ntb/test/ntb_perf.c
index e9b7c2dfc730..972f6d984f6d 100644
--- a/drivers/ntb/test/ntb_perf.c
+++ b/drivers/ntb/test/ntb_perf.c
@@ -149,7 +149,8 @@ struct perf_peer {
 	u64 outbuf_xlat;
 	resource_size_t outbuf_size;
 	void __iomem *outbuf;
-
+	phys_addr_t out_phys_addr;
+	dma_addr_t dma_dst_addr;
 	/* Inbound MW params */
 	dma_addr_t inbuf_xlat;
 	resource_size_t inbuf_size;
@@ -782,6 +783,10 @@ static int perf_copy_chunk(struct perf_thread *pthr,
 	struct dmaengine_unmap_data *unmap;
 	struct device *dma_dev;
 	int try = 0, ret = 0;
+	struct perf_peer *peer = pthr->perf->test_peer;
+	void __iomem *vbase;
+	void __iomem *dst_vaddr;
+	dma_addr_t dst_dma_addr;
 
 	if (!use_dma) {
 		memcpy_toio(dst, src, len);
@@ -794,6 +799,10 @@ static int perf_copy_chunk(struct perf_thread *pthr,
 				 offset_in_page(dst), len))
 		return -EIO;
 
+	vbase = peer->outbuf;
+	dst_vaddr = dst;
+	dst_dma_addr = peer->dma_dst_addr + (dst_vaddr - vbase);
+
 	unmap = dmaengine_get_unmap_data(dma_dev, 2, GFP_NOWAIT);
 	if (!unmap)
 		return -ENOMEM;
@@ -807,8 +816,7 @@ static int perf_copy_chunk(struct perf_thread *pthr,
 	}
 	unmap->to_cnt = 1;
 
-	unmap->addr[1] = dma_map_page(dma_dev, virt_to_page(dst),
-		offset_in_page(dst), len, DMA_FROM_DEVICE);
+	unmap->addr[1] = dst_dma_addr;
 	if (dma_mapping_error(dma_dev, unmap->addr[1])) {
 		ret = -EIO;
 		goto err_free_resource;
@@ -865,6 +873,7 @@ static int perf_init_test(struct perf_thread *pthr)
 {
 	struct perf_ctx *perf = pthr->perf;
 	dma_cap_mask_t dma_mask;
+	struct perf_peer *peer = pthr->perf->test_peer;
 
 	pthr->src = kmalloc_node(perf->test_peer->outbuf_size, GFP_KERNEL,
 				 dev_to_node(&perf->ntb->dev));
@@ -882,15 +891,33 @@ static int perf_init_test(struct perf_thread *pthr)
 	if (!pthr->dma_chan) {
 		dev_err(&perf->ntb->dev, "%d: Failed to get DMA channel\n",
 			pthr->tidx);
-		atomic_dec(&perf->tsync);
-		wake_up(&perf->twait);
-		kfree(pthr->src);
-		return -ENODEV;
+		goto err_free;
 	}
+	peer->dma_dst_addr =
+		dma_map_resource(pthr->dma_chan->device->dev,
+				 peer->out_phys_addr, peer->outbuf_size,
+				 DMA_FROM_DEVICE, 0);
+	if (dma_mapping_error(pthr->dma_chan->device->dev,
+			      peer->dma_dst_addr)) {
+		dev_err(pthr->dma_chan->device->dev, "%d: Failed to map DMA addr\n",
+			pthr->tidx);
+		peer->dma_dst_addr = 0;
+		dma_release_channel(pthr->dma_chan);
+		goto err_free;
+	}
+	dev_dbg(pthr->dma_chan->device->dev, "%d: Map MMIO %pa to DMA addr %pad\n",
+			pthr->tidx,
+			&peer->out_phys_addr,
+			&peer->dma_dst_addr);
 
 	atomic_set(&pthr->dma_sync, 0);
-
 	return 0;
+
+err_free:
+	atomic_dec(&perf->tsync);
+	wake_up(&perf->twait);
+	kfree(pthr->src);
+	return -ENODEV;
 }
 
 static int perf_run_test(struct perf_thread *pthr)
@@ -978,8 +1005,13 @@ static void perf_clear_test(struct perf_thread *pthr)
 	 * We call it anyway just to be sure of the transfers completion.
 	 */
 	(void)dmaengine_terminate_sync(pthr->dma_chan);
-
-	dma_release_channel(pthr->dma_chan);
+	if (pthr->perf->test_peer->dma_dst_addr)
+		dma_unmap_resource(pthr->dma_chan->device->dev,
+				   pthr->perf->test_peer->dma_dst_addr,
+				   pthr->perf->test_peer->outbuf_size,
+				   DMA_FROM_DEVICE, 0);
+	if (pthr->dma_chan)
+		dma_release_channel(pthr->dma_chan);
 
 no_dma_notify:
 	atomic_dec(&perf->tsync);
@@ -1195,6 +1227,9 @@ static ssize_t perf_dbgfs_read_info(struct file *filep, char __user *ubuf,
 			"\tOut buffer addr 0x%pK\n", peer->outbuf);
 
 		pos += scnprintf(buf + pos, buf_size - pos,
+			"\tOut buff phys addr %pa[p]\n", &peer->out_phys_addr);
+
+		pos += scnprintf(buf + pos, buf_size - pos,
 			"\tOut buffer size %pa\n", &peer->outbuf_size);
 
 		pos += scnprintf(buf + pos, buf_size - pos,
@@ -1388,6 +1423,8 @@ static int perf_setup_peer_mw(struct perf_peer *peer)
 	if (!peer->outbuf)
 		return -ENOMEM;
 
+	peer->out_phys_addr = phys_addr;
+
 	if (max_mw_size && peer->outbuf_size > max_mw_size) {
 		peer->outbuf_size = max_mw_size;
 		dev_warn(&peer->perf->ntb->dev,
diff --git a/drivers/ntb/test/ntb_tool.c b/drivers/ntb/test/ntb_tool.c
index d592c0ffbd19..69da758fe64c 100644
--- a/drivers/ntb/test/ntb_tool.c
+++ b/drivers/ntb/test/ntb_tool.c
@@ -678,19 +678,19 @@ static ssize_t tool_mw_trans_read(struct file *filep, char __user *ubuf,
 			 &inmw->dma_base);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Window Size    \t%pa[p]\n",
+			 "Window Size    \t%pap\n",
 			 &inmw->size);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Alignment      \t%pa[p]\n",
+			 "Alignment      \t%pap\n",
 			 &addr_align);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Size Alignment \t%pa[p]\n",
+			 "Size Alignment \t%pap\n",
 			 &size_align);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Size Max       \t%pa[p]\n",
+			 "Size Max       \t%pap\n",
 			 &size_max);
 
 	ret = simple_read_from_buffer(ubuf, size, offp, buf, off);
@@ -907,16 +907,16 @@ static ssize_t tool_peer_mw_trans_read(struct file *filep, char __user *ubuf,
 			 "Virtual address     \t0x%pK\n", outmw->io_base);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Phys Address        \t%pa[p]\n", &map_base);
+			 "Phys Address        \t%pap\n", &map_base);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Mapping Size        \t%pa[p]\n", &map_size);
+			 "Mapping Size        \t%pap\n", &map_size);
 
 	off += scnprintf(buf + off, buf_size - off,
 			 "Translation Address \t0x%016llx\n", outmw->tr_base);
 
 	off += scnprintf(buf + off, buf_size - off,
-			 "Window Size         \t%pa[p]\n", &outmw->size);
+			 "Window Size         \t%pap\n", &outmw->size);
 
 	ret = simple_read_from_buffer(ubuf, size, offp, buf, off);
 	kfree(buf);