Merge tag 'for-4.20/block-20181021' of git://git.kernel.dk/linux-block

Pull block layer updates from Jens Axboe: "This is the main pull request for block changes for 4.20. This contains: - Series enabling runtime PM for blk-mq (Bart). - Two pull requests from Christoph for NVMe, with items such as; - Better AEN tracking - Multipath improvements - RDMA fixes - Rework of FC for target removal - Fixes for issues identified by static checkers - Fabric cleanups, as prep for TCP transport - Various cleanups and bug fixes - Block merging cleanups (Christoph) - Conversion of drivers to generic DMA mapping API (Christoph) - Series fixing ref count issues with blkcg (Dennis) - Series improving BFQ heuristics (Paolo, et al) - Series improving heuristics for the Kyber IO scheduler (Omar) - Removal of dangerous bio_rewind_iter() API (Ming) - Apply single queue IPI redirection logic to blk-mq (Ming) - Set of fixes and improvements for bcache (Coly et al) - Series closing a hotplug race with sysfs group attributes (Hannes) - Set of patches for lightnvm: - pblk trace support (Hans) - SPDX license header update (Javier) - Tons of refactoring patches to cleanly abstract the 1.2 and 2.0 specs behind a common core interface. (Javier, Matias) - Enable pblk to use a common interface to retrieve chunk metadata (Matias) - Bug fixes (Various) - Set of fixes and updates to the blk IO latency target (Josef) - blk-mq queue number updates fixes (Jianchao) - Convert a bunch of drivers from the old legacy IO interface to blk-mq. This will conclude with the removal of the legacy IO interface itself in 4.21, with the rest of the drivers (me, Omar) - Removal of the DAC960 driver. The SCSI tree will introduce two replacement drivers for this (Hannes)" * tag 'for-4.20/block-20181021' of git://git.kernel.dk/linux-block: (204 commits) block: setup bounce bio_sets properly blkcg: reassociate bios when make_request() is called recursively blkcg: fix edge case for blk_get_rl() under memory pressure nvme-fabrics: move controller options matching to fabrics nvme-rdma: always have a valid trsvcid mtip32xx: fully switch to the generic DMA API rsxx: switch to the generic DMA API umem: switch to the generic DMA API sx8: switch to the generic DMA API sx8: remove dead IF_64BIT_DMA_IS_POSSIBLE code skd: switch to the generic DMA API ubd: remove use of blk_rq_map_sg nvme-pci: remove duplicate check drivers/block: Remove DAC960 driver nvme-pci: fix hot removal during error handling nvmet-fcloop: suppress a compiler warning nvme-core: make implicit seed truncation explicit nvmet-fc: fix kernel-doc headers nvme-fc: rework the request initialization code nvme-fc: introduce struct nvme_fcp_op_w_sgl ...
author: Linus Torvalds 2018-10-22 17:46:08 +0100
committer: Linus Torvalds 2018-10-22 17:46:08 +0100
commit: 6ab9e09238fdfd742fe23b81e2d385a1cab49d9b (patch)
tree: 38c27b6d70c68f1114df2f49f72f1f757b22cd07 /block
parent: 528985117126f11beea339cf39120ee99da04cd2 (diff)
parent: 52990a5fb0c991ecafebdab43138b5ed41376852 (diff)
30 files changed, 1726 insertions, 1010 deletions
diff --git a/block/Kconfig b/block/Kconfig
index 1f2469a0123c..f7045aa47edb 100644
--- a/block/Kconfig
+++ b/block/Kconfig
@@ -74,7 +74,6 @@ config BLK_DEV_BSG
 
 config BLK_DEV_BSGLIB
 	bool "Block layer SG support v4 helper lib"
-	default n
 	select BLK_DEV_BSG
 	select BLK_SCSI_REQUEST
 	help
@@ -107,7 +106,6 @@ config BLK_DEV_ZONED
 config BLK_DEV_THROTTLING
 	bool "Block layer bio throttling support"
 	depends on BLK_CGROUP=y
-	default n
 	---help---
 	Block layer bio throttling support. It can be used to limit
 	the IO rate to a device. IO rate policies are per cgroup and
@@ -119,7 +117,6 @@ config BLK_DEV_THROTTLING
 config BLK_DEV_THROTTLING_LOW
 	bool "Block throttling .low limit interface support (EXPERIMENTAL)"
 	depends on BLK_DEV_THROTTLING
-	default n
 	---help---
 	Add .low limit interface for block throttling. The low limit is a best
 	effort limit to prioritize cgroups. Depending on the setting, the limit
@@ -130,7 +127,6 @@ config BLK_DEV_THROTTLING_LOW
 
 config BLK_CMDLINE_PARSER
 	bool "Block device command line partition parser"
-	default n
 	---help---
 	Enabling this option allows you to specify the partition layout from
 	the kernel boot args.  This is typically of use for embedded devices
@@ -141,7 +137,6 @@ config BLK_CMDLINE_PARSER
 
 config BLK_WBT
 	bool "Enable support for block device writeback throttling"
-	default n
 	---help---
 	Enabling this option enables the block layer to throttle buffered
 	background writeback from the VM, making it more smooth and having
@@ -152,7 +147,6 @@ config BLK_WBT
 config BLK_CGROUP_IOLATENCY
 	bool "Enable support for latency based cgroup IO protection"
 	depends on BLK_CGROUP=y
-	default n
 	---help---
 	Enabling this option enables the .latency interface for IO throttling.
 	The IO controller will attempt to maintain average IO latencies below
@@ -163,7 +157,6 @@ config BLK_CGROUP_IOLATENCY
 
 config BLK_WBT_SQ
 	bool "Single queue writeback throttling"
-	default n
 	depends on BLK_WBT
 	---help---
 	Enable writeback throttling by default on legacy single queue devices
@@ -228,4 +221,7 @@ config BLK_MQ_RDMA
 	depends on BLOCK && INFINIBAND
 	default y
 
+config BLK_PM
+	def_bool BLOCK && PM
+
 source block/Kconfig.iosched
diff --git a/block/Kconfig.iosched b/block/Kconfig.iosched
index a4a8914bf7a4..f95a48b0d7b2 100644
--- a/block/Kconfig.iosched
+++ b/block/Kconfig.iosched
@@ -36,7 +36,6 @@ config IOSCHED_CFQ
 config CFQ_GROUP_IOSCHED
 	bool "CFQ Group Scheduling support"
 	depends on IOSCHED_CFQ && BLK_CGROUP
-	default n
 	---help---
 	  Enable group IO scheduling in CFQ.
 
@@ -82,7 +81,6 @@ config MQ_IOSCHED_KYBER
 
 config IOSCHED_BFQ
 	tristate "BFQ I/O scheduler"
-	default n
 	---help---
 	BFQ I/O scheduler for BLK-MQ. BFQ distributes the bandwidth of
 	of the device among all processes according to their weights,
@@ -94,7 +92,6 @@ config IOSCHED_BFQ
 config BFQ_GROUP_IOSCHED
        bool "BFQ hierarchical scheduling support"
        depends on IOSCHED_BFQ && BLK_CGROUP
-       default n
        ---help---
 
        Enable hierarchical scheduling in BFQ, using the blkio
diff --git a/block/Makefile b/block/Makefile
index 572b33f32c07..27eac600474f 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -37,3 +37,4 @@ obj-$(CONFIG_BLK_WBT)		+= blk-wbt.o
 obj-$(CONFIG_BLK_DEBUG_FS)	+= blk-mq-debugfs.o
 obj-$(CONFIG_BLK_DEBUG_FS_ZONED)+= blk-mq-debugfs-zoned.o
 obj-$(CONFIG_BLK_SED_OPAL)	+= sed-opal.o
+obj-$(CONFIG_BLK_PM)		+= blk-pm.o
diff --git a/block/bfq-cgroup.c b/block/bfq-cgroup.c
index 9fe5952d117d..d9a7916ff0ab 100644
--- a/block/bfq-cgroup.c
+++ b/block/bfq-cgroup.c
@@ -642,7 +642,7 @@ void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio)
 	uint64_t serial_nr;
 
 	rcu_read_lock();
-	serial_nr = bio_blkcg(bio)->css.serial_nr;
+	serial_nr = __bio_blkcg(bio)->css.serial_nr;
 
 	/*
 	 * Check whether blkcg has changed.  The condition may trigger
@@ -651,7 +651,7 @@ void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio)
 	if (unlikely(!bfqd) || likely(bic->blkcg_serial_nr == serial_nr))
 		goto out;
 
-	bfqg = __bfq_bic_change_cgroup(bfqd, bic, bio_blkcg(bio));
+	bfqg = __bfq_bic_change_cgroup(bfqd, bic, __bio_blkcg(bio));
 	/*
 	 * Update blkg_path for bfq_log_* functions. We cache this
 	 * path, and update it here, for the following
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 653100fb719e..6075100f03a5 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -624,12 +624,13 @@ void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq)
 }
 
 /*
- * Tell whether there are active queues or groups with differentiated weights.
+ * Tell whether there are active queues with different weights or
+ * active groups.
  */
-static bool bfq_differentiated_weights(struct bfq_data *bfqd)
+static bool bfq_varied_queue_weights_or_active_groups(struct bfq_data *bfqd)
 {
 	/*
-	 * For weights to differ, at least one of the trees must contain
+	 * For queue weights to differ, queue_weights_tree must contain
 	 * at least two nodes.
 	 */
 	return (!RB_EMPTY_ROOT(&bfqd->queue_weights_tree) &&
@@ -637,9 +638,7 @@ static bool bfq_differentiated_weights(struct bfq_data *bfqd)
 		 bfqd->queue_weights_tree.rb_node->rb_right)
 #ifdef CONFIG_BFQ_GROUP_IOSCHED
 	       ) ||
-	       (!RB_EMPTY_ROOT(&bfqd->group_weights_tree) &&
-		(bfqd->group_weights_tree.rb_node->rb_left ||
-		 bfqd->group_weights_tree.rb_node->rb_right)
+		(bfqd->num_active_groups > 0
 #endif
 	       );
 }
@@ -657,26 +656,25 @@ static bool bfq_differentiated_weights(struct bfq_data *bfqd)
  * 3) all active groups at the same level in the groups tree have the same
  *    number of children.
  *
- * Unfortunately, keeping the necessary state for evaluating exactly the
- * above symmetry conditions would be quite complex and time-consuming.
- * Therefore this function evaluates, instead, the following stronger
- * sub-conditions, for which it is much easier to maintain the needed
- * state:
+ * Unfortunately, keeping the necessary state for evaluating exactly
+ * the last two symmetry sub-conditions above would be quite complex
+ * and time consuming.  Therefore this function evaluates, instead,
+ * only the following stronger two sub-conditions, for which it is
+ * much easier to maintain the needed state:
  * 1) all active queues have the same weight,
- * 2) all active groups have the same weight,
- * 3) all active groups have at most one active child each.
- * In particular, the last two conditions are always true if hierarchical
- * support and the cgroups interface are not enabled, thus no state needs
- * to be maintained in this case.
+ * 2) there are no active groups.
+ * In particular, the last condition is always true if hierarchical
+ * support or the cgroups interface are not enabled, thus no state
+ * needs to be maintained in this case.
  */
 static bool bfq_symmetric_scenario(struct bfq_data *bfqd)
 {
-	return !bfq_differentiated_weights(bfqd);
+	return !bfq_varied_queue_weights_or_active_groups(bfqd);
 }
 
 /*
  * If the weight-counter tree passed as input contains no counter for
- * the weight of the input entity, then add that counter; otherwise just
+ * the weight of the input queue, then add that counter; otherwise just
  * increment the existing counter.
  *
  * Note that weight-counter trees contain few nodes in mostly symmetric
@@ -687,25 +685,25 @@ static bool bfq_symmetric_scenario(struct bfq_data *bfqd)
  * In most scenarios, the rate at which nodes are created/destroyed
  * should be low too.
  */
-void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_entity *entity,
+void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 			  struct rb_root *root)
 {
+	struct bfq_entity *entity = &bfqq->entity;
 	struct rb_node **new = &(root->rb_node), *parent = NULL;
 
 	/*
-	 * Do not insert if the entity is already associated with a
+	 * Do not insert if the queue is already associated with a
 	 * counter, which happens if:
-	 *   1) the entity is associated with a queue,
-	 *   2) a request arrival has caused the queue to become both
+	 *   1) a request arrival has caused the queue to become both
 	 *      non-weight-raised, and hence change its weight, and
 	 *      backlogged; in this respect, each of the two events
 	 *      causes an invocation of this function,
-	 *   3) this is the invocation of this function caused by the
+	 *   2) this is the invocation of this function caused by the
 	 *      second event. This second invocation is actually useless,
 	 *      and we handle this fact by exiting immediately. More
 	 *      efficient or clearer solutions might possibly be adopted.
 	 */
-	if (entity->weight_counter)
+	if (bfqq->weight_counter)
 		return;
 
 	while (*new) {
@@ -715,7 +713,7 @@ void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_entity *entity,
 		parent = *new;
 
 		if (entity->weight == __counter->weight) {
-			entity->weight_counter = __counter;
+			bfqq->weight_counter = __counter;
 			goto inc_counter;
 		}
 		if (entity->weight < __counter->weight)
@@ -724,66 +722,67 @@ void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_entity *entity,
 			new = &((*new)->rb_right);
 	}
 
-	entity->weight_counter = kzalloc(sizeof(struct bfq_weight_counter),
-					 GFP_ATOMIC);
+	bfqq->weight_counter = kzalloc(sizeof(struct bfq_weight_counter),
+				       GFP_ATOMIC);
 
 	/*
 	 * In the unlucky event of an allocation failure, we just
-	 * exit. This will cause the weight of entity to not be
-	 * considered in bfq_differentiated_weights, which, in its
-	 * turn, causes the scenario to be deemed wrongly symmetric in
-	 * case entity's weight would have been the only weight making
-	 * the scenario asymmetric. On the bright side, no unbalance
-	 * will however occur when entity becomes inactive again (the
-	 * invocation of this function is triggered by an activation
-	 * of entity). In fact, bfq_weights_tree_remove does nothing
-	 * if !entity->weight_counter.
+	 * exit. This will cause the weight of queue to not be
+	 * considered in bfq_varied_queue_weights_or_active_groups,
+	 * which, in its turn, causes the scenario to be deemed
+	 * wrongly symmetric in case bfqq's weight would have been
+	 * the only weight making the scenario asymmetric.  On the
+	 * bright side, no unbalance will however occur when bfqq
+	 * becomes inactive again (the invocation of this function
+	 * is triggered by an activation of queue).  In fact,
+	 * bfq_weights_tree_remove does nothing if
+	 * !bfqq->weight_counter.
 	 */
-	if (unlikely(!entity->weight_counter))
+	if (unlikely(!bfqq->weight_counter))
 		return;
 
-	entity->weight_counter->weight = entity->weight;
-	rb_link_node(&entity->weight_counter->weights_node, parent, new);
-	rb_insert_color(&entity->weight_counter->weights_node, root);
+	bfqq->weight_counter->weight = entity->weight;
+	rb_link_node(&bfqq->weight_counter->weights_node, parent, new);
+	rb_insert_color(&bfqq->weight_counter->weights_node, root);
 
 inc_counter:
-	entity->weight_counter->num_active++;
+	bfqq->weight_counter->num_active++;
 }
 
 /*
- * Decrement the weight counter associated with the entity, and, if the
+ * Decrement the weight counter associated with the queue, and, if the
  * counter reaches 0, remove the counter from the tree.
  * See the comments to the function bfq_weights_tree_add() for considerations
  * about overhead.
  */
 void __bfq_weights_tree_remove(struct bfq_data *bfqd,
-			       struct bfq_entity *entity,
+			       struct bfq_queue *bfqq,
 			       struct rb_root *root)
 {
-	if (!entity->weight_counter)
+	if (!bfqq->weight_counter)
 		return;
 
-	entity->weight_counter->num_active--;
-	if (entity->weight_counter->num_active > 0)
+	bfqq->weight_counter->num_active--;
+	if (bfqq->weight_counter->num_active > 0)
 		goto reset_entity_pointer;
 
-	rb_erase(&entity->weight_counter->weights_node, root);
-	kfree(entity->weight_counter);
+	rb_erase(&bfqq->weight_counter->weights_node, root);
+	kfree(bfqq->weight_counter);
 
 reset_entity_pointer:
-	entity->weight_counter = NULL;
+	bfqq->weight_counter = NULL;
 }
 
 /*
- * Invoke __bfq_weights_tree_remove on bfqq and all its inactive
- * parent entities.
+ * Invoke __bfq_weights_tree_remove on bfqq and decrement the number
+ * of active groups for each queue's inactive parent entity.
  */
 void bfq_weights_tree_remove(struct bfq_data *bfqd,
 			     struct bfq_queue *bfqq)
 {
 	struct bfq_entity *entity = bfqq->entity.parent;
 
-	__bfq_weights_tree_remove(bfqd, &bfqq->entity,
+	__bfq_weights_tree_remove(bfqd, bfqq,
 				  &bfqd->queue_weights_tree);
 
 	for_each_entity(entity) {
@@ -797,17 +796,13 @@ void bfq_weights_tree_remove(struct bfq_data *bfqd,
 			 * next_in_service for details on why
 			 * in_service_entity must be checked too).
 			 *
-			 * As a consequence, the weight of entity is
-			 * not to be removed. In addition, if entity
-			 * is active, then its parent entities are
-			 * active as well, and thus their weights are
-			 * not to be removed either. In the end, this
-			 * loop must stop here.
+			 * As a consequence, its parent entities are
+			 * active as well, and thus this loop must
+			 * stop here.
 			 */
 			break;
 		}
-		__bfq_weights_tree_remove(bfqd, entity,
-					  &bfqd->group_weights_tree);
+		bfqd->num_active_groups--;
 	}
 }
 
@@ -3182,6 +3177,13 @@ static unsigned long bfq_bfqq_softrt_next_start(struct bfq_data *bfqd,
 		    jiffies + nsecs_to_jiffies(bfqq->bfqd->bfq_slice_idle) + 4);
 }
 
+static bool bfq_bfqq_injectable(struct bfq_queue *bfqq)
+{
+	return BFQQ_SEEKY(bfqq) && bfqq->wr_coeff == 1 &&
+		blk_queue_nonrot(bfqq->bfqd->queue) &&
+		bfqq->bfqd->hw_tag;
+}
+
 /**
  * bfq_bfqq_expire - expire a queue.
  * @bfqd: device owning the queue.
@@ -3291,6 +3293,8 @@ void bfq_bfqq_expire(struct bfq_data *bfqd,
 	if (ref == 1) /* bfqq is gone, no more actions on it */
 		return;
 
+	bfqq->injected_service = 0;
+
 	/* mark bfqq as waiting a request only if a bic still points to it */
 	if (!bfq_bfqq_busy(bfqq) &&
 	    reason != BFQQE_BUDGET_TIMEOUT &&
@@ -3497,9 +3501,11 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq)
 	 * symmetric scenario where:
 	 * (i)  each of these processes must get the same throughput as
 	 *      the others;
-	 * (ii) all these processes have the same I/O pattern
-		(either sequential or random).
-	 * In fact, in such a scenario, the drive will tend to treat
+	 * (ii) the I/O of each process has the same properties, in
+	 *      terms of locality (sequential or random), direction
+	 *      (reads or writes), request sizes, greediness
+	 *      (from I/O-bound to sporadic), and so on.
+	 * In fact, in such a scenario, the drive tends to treat
 	 * the requests of each of these processes in about the same
 	 * way as the requests of the others, and thus to provide
 	 * each of these processes with about the same throughput
@@ -3508,18 +3514,50 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq)
 	 * certainly needed to guarantee that bfqq receives its
 	 * assigned fraction of the device throughput (see [1] for
 	 * details).
+	 * The problem is that idling may significantly reduce
+	 * throughput with certain combinations of types of I/O and
+	 * devices. An important example is sync random I/O, on flash
+	 * storage with command queueing. So, unless bfqq falls in the
+	 * above cases where idling also boosts throughput, it would
+	 * be important to check conditions (i) and (ii) accurately,
+	 * so as to avoid idling when not strictly needed for service
+	 * guarantees.
+	 *
+	 * Unfortunately, it is extremely difficult to thoroughly
+	 * check condition (ii). And, in case there are active groups,
+	 * it becomes very difficult to check condition (i) too. In
+	 * fact, if there are active groups, then, for condition (i)
+	 * to become false, it is enough that an active group contains
+	 * more active processes or sub-groups than some other active
+	 * group. We address this issue with the following bi-modal
+	 * behavior, implemented in the function
+	 * bfq_symmetric_scenario().
 	 *
-	 * We address this issue by controlling, actually, only the
-	 * symmetry sub-condition (i), i.e., provided that
-	 * sub-condition (i) holds, idling is not performed,
-	 * regardless of whether sub-condition (ii) holds. In other
-	 * words, only if sub-condition (i) holds, then idling is
+	 * If there are active groups, then the scenario is tagged as
+	 * asymmetric, conservatively, without checking any of the
+	 * conditions (i) and (ii). So the device is idled for bfqq.
+	 * This behavior matches also the fact that groups are created
+	 * exactly if controlling I/O (to preserve bandwidth and
+	 * latency guarantees) is a primary concern.
+	 *
+	 * On the opposite end, if there are no active groups, then
+	 * only condition (i) is actually controlled, i.e., provided
+	 * that condition (i) holds, idling is not performed,
+	 * regardless of whether condition (ii) holds. In other words,
+	 * only if condition (i) does not hold, then idling is
 	 * allowed, and the device tends to be prevented from queueing
-	 * many requests, possibly of several processes. The reason
-	 * for not controlling also sub-condition (ii) is that we
-	 * exploit preemption to preserve guarantees in case of
-	 * symmetric scenarios, even if (ii) does not hold, as
-	 * explained in the next two paragraphs.
+	 * many requests, possibly of several processes. Since there
+	 * are no active groups, then, to control condition (i) it is
+	 * enough to check whether all active queues have the same
+	 * weight.
+	 *
+	 * Not checking condition (ii) evidently exposes bfqq to the
+	 * risk of getting less throughput than its fair share.
+	 * However, for queues with the same weight, a further
+	 * mechanism, preemption, mitigates or even eliminates this
+	 * problem. And it does so without consequences on overall
+	 * throughput. This mechanism and its benefits are explained
+	 * in the next three paragraphs.
 	 *
 	 * Even if a queue, say Q, is expired when it remains idle, Q
 	 * can still preempt the new in-service queue if the next
@@ -3533,11 +3571,7 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq)
 	 * idling allows the internal queues of the device to contain
 	 * many requests, and thus to reorder requests, we can rather
 	 * safely assume that the internal scheduler still preserves a
-	 * minimum of mid-term fairness. The motivation for using
-	 * preemption instead of idling is that, by not idling,
-	 * service guarantees are preserved without minimally
-	 * sacrificing throughput. In other words, both a high
-	 * throughput and its desired distribution are obtained.
+	 * minimum of mid-term fairness.
 	 *
 	 * More precisely, this preemption-based, idleless approach
 	 * provides fairness in terms of IOPS, and not sectors per
@@ -3556,22 +3590,27 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq)
 	 * 1024/8 times as high as the service received by the other
 	 * queue.
 	 *
-	 * On the other hand, device idling is performed, and thus
-	 * pure sector-domain guarantees are provided, for the
-	 * following queues, which are likely to need stronger
-	 * throughput guarantees: weight-raised queues, and queues
-	 * with a higher weight than other queues. When such queues
-	 * are active, sub-condition (i) is false, which triggers
-	 * device idling.
+	 * The motivation for using preemption instead of idling (for
+	 * queues with the same weight) is that, by not idling,
+	 * service guarantees are preserved (completely or at least in
+	 * part) without minimally sacrificing throughput. And, if
+	 * there is no active group, then the primary expectation for
+	 * this device is probably a high throughput.
 	 *
-	 * According to the above considerations, the next variable is
-	 * true (only) if sub-condition (i) holds. To compute the
-	 * value of this variable, we not only use the return value of
-	 * the function bfq_symmetric_scenario(), but also check
-	 * whether bfqq is being weight-raised, because
-	 * bfq_symmetric_scenario() does not take into account also
-	 * weight-raised queues (see comments on
-	 * bfq_weights_tree_add()).
+	 * We are now left only with explaining the additional
+	 * compound condition that is checked below for deciding
+	 * whether the scenario is asymmetric. To explain this
+	 * compound condition, we need to add that the function
+	 * bfq_symmetric_scenario checks the weights of only
+	 * non-weight-raised queues, for efficiency reasons (see
+	 * comments on bfq_weights_tree_add()). Then the fact that
+	 * bfqq is weight-raised is checked explicitly here. More
+	 * precisely, the compound condition below takes into account
+	 * also the fact that, even if bfqq is being weight-raised,
+	 * the scenario is still symmetric if all active queues happen
+	 * to be weight-raised. Actually, we should be even more
+	 * precise here, and differentiate between interactive weight
+	 * raising and soft real-time weight raising.
 	 *
 	 * As a side note, it is worth considering that the above
 	 * device-idling countermeasures may however fail in the
@@ -3583,7 +3622,8 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq)
 	 * to let requests be served in the desired order until all
 	 * the requests already queued in the device have been served.
 	 */
-	asymmetric_scenario = bfqq->wr_coeff > 1 ||
+	asymmetric_scenario = (bfqq->wr_coeff > 1 &&
+			       bfqd->wr_busy_queues < bfqd->busy_queues) ||
 		!bfq_symmetric_scenario(bfqd);
 
 	/*
@@ -3629,6 +3669,30 @@ static bool bfq_bfqq_must_idle(struct bfq_queue *bfqq)
 	return RB_EMPTY_ROOT(&bfqq->sort_list) && bfq_better_to_idle(bfqq);
 }
 
+static struct bfq_queue *bfq_choose_bfqq_for_injection(struct bfq_data *bfqd)
+{
+	struct bfq_queue *bfqq;
+
+	/*
+	 * A linear search; but, with a high probability, very few
+	 * steps are needed to find a candidate queue, i.e., a queue
+	 * with enough budget left for its next request. In fact:
+	 * - BFQ dynamically updates the budget of every queue so as
+	 *   to accommodate the expected backlog of the queue;
+	 * - if a queue gets all its requests dispatched as injected
+	 *   service, then the queue is removed from the active list
+	 *   (and re-added only if it gets new requests, but with
+	 *   enough budget for its new backlog).
+	 */
+	list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list)
+		if (!RB_EMPTY_ROOT(&bfqq->sort_list) &&
+		    bfq_serv_to_charge(bfqq->next_rq, bfqq) <=
+		    bfq_bfqq_budget_left(bfqq))
+			return bfqq;
+
+	return NULL;
+}
+
 /*
  * Select a queue for service.  If we have a current queue in service,
  * check whether to continue servicing it, or retrieve and set a new one.
@@ -3710,10 +3774,19 @@ check_queue:
 	 * No requests pending. However, if the in-service queue is idling
 	 * for a new request, or has requests waiting for a completion and
 	 * may idle after their completion, then keep it anyway.
+	 *
+	 * Yet, to boost throughput, inject service from other queues if
+	 * possible.
 	 */
 	if (bfq_bfqq_wait_request(bfqq) ||
 	    (bfqq->dispatched != 0 && bfq_better_to_idle(bfqq))) {
-		bfqq = NULL;
+		if (bfq_bfqq_injectable(bfqq) &&
+		    bfqq->injected_service * bfqq->inject_coeff <
+		    bfqq->entity.service * 10)
+			bfqq = bfq_choose_bfqq_for_injection(bfqd);
+		else
+			bfqq = NULL;
+
 		goto keep_queue;
 	}
 
@@ -3803,6 +3876,14 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
 
 	bfq_dispatch_remove(bfqd->queue, rq);
 
+	if (bfqq != bfqd->in_service_queue) {
+		if (likely(bfqd->in_service_queue))
+			bfqd->in_service_queue->injected_service +=
+				bfq_serv_to_charge(rq, bfqq);
+
+		goto return_rq;
+	}
+
 	/*
 	 * If weight raising has to terminate for bfqq, then next
 	 * function causes an immediate update of bfqq's weight,
@@ -3821,13 +3902,12 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
 	 * belongs to CLASS_IDLE and other queues are waiting for
 	 * service.
 	 */
-	if (bfqd->busy_queues > 1 && bfq_class_idle(bfqq))
-		goto expire;
-
-	return rq;
+	if (!(bfqd->busy_queues > 1 && bfq_class_idle(bfqq)))
+		goto return_rq;
 
-expire:
 	bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED);
+
+return_rq:
 	return rq;
 }
 
@@ -4232,6 +4312,13 @@ static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 			bfq_mark_bfqq_has_short_ttime(bfqq);
 		bfq_mark_bfqq_sync(bfqq);
 		bfq_mark_bfqq_just_created(bfqq);
+		/*
+		 * Aggressively inject a lot of service: up to 90%.
+		 * This coefficient remains constant during bfqq life,
+		 * but this behavior might be changed, after enough
+		 * testing and tuning.
+		 */
+		bfqq->inject_coeff = 1;
 	} else
 		bfq_clear_bfqq_sync(bfqq);
 
@@ -4297,7 +4384,7 @@ static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd,
 
 	rcu_read_lock();
 
-	bfqg = bfq_find_set_group(bfqd, bio_blkcg(bio));
+	bfqg = bfq_find_set_group(bfqd, __bio_blkcg(bio));
 	if (!bfqg) {
 		bfqq = &bfqd->oom_bfqq;
 		goto out;
@@ -5330,7 +5417,7 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e)
 	bfqd->idle_slice_timer.function = bfq_idle_slice_timer;
 
 	bfqd->queue_weights_tree = RB_ROOT;
-	bfqd->group_weights_tree = RB_ROOT;
+	bfqd->num_active_groups = 0;
 
 	INIT_LIST_HEAD(&bfqd->active_list);
 	INIT_LIST_HEAD(&bfqd->idle_list);
diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h
index a8a2e5aca4d4..77651d817ecd 100644
--- a/block/bfq-iosched.h
+++ b/block/bfq-iosched.h
@@ -108,15 +108,14 @@ struct bfq_sched_data {
 };
 
 /**
- * struct bfq_weight_counter - counter of the number of all active entities
+ * struct bfq_weight_counter - counter of the number of all active queues
  *                             with a given weight.
  */
 struct bfq_weight_counter {
-	unsigned int weight; /* weight of the entities this counter refers to */
-	unsigned int num_active; /* nr of active entities with this weight */
+	unsigned int weight; /* weight of the queues this counter refers to */
+	unsigned int num_active; /* nr of active queues with this weight */
 	/*
-	 * Weights tree member (see bfq_data's @queue_weights_tree and
-	 * @group_weights_tree)
+	 * Weights tree member (see bfq_data's @queue_weights_tree)
 	 */
 	struct rb_node weights_node;
 };
@@ -151,8 +150,6 @@ struct bfq_weight_counter {
 struct bfq_entity {
 	/* service_tree member */
 	struct rb_node rb_node;
-	/* pointer to the weight counter associated with this entity */
-	struct bfq_weight_counter *weight_counter;
 
 	/*
 	 * Flag, true if the entity is on a tree (either the active or
@@ -266,6 +263,9 @@ struct bfq_queue {
 	/* entity representing this queue in the scheduler */
 	struct bfq_entity entity;
 
+	/* pointer to the weight counter associated with this entity */
+	struct bfq_weight_counter *weight_counter;
+
 	/* maximum budget allowed from the feedback mechanism */
 	int max_budget;
 	/* budget expiration (in jiffies) */
@@ -351,6 +351,32 @@ struct bfq_queue {
 	unsigned long split_time; /* time of last split */
 
 	unsigned long first_IO_time; /* time of first I/O for this queue */
+
+	/* max service rate measured so far */
+	u32 max_service_rate;
+	/*
+	 * Ratio between the service received by bfqq while it is in
+	 * service, and the cumulative service (of requests of other
+	 * queues) that may be injected while bfqq is empty but still
+	 * in service. To increase precision, the coefficient is
+	 * measured in tenths of unit. Here are some example of (1)
+	 * ratios, (2) resulting percentages of service injected
+	 * w.r.t. to the total service dispatched while bfqq is in
+	 * service, and (3) corresponding values of the coefficient:
+	 * 1 (50%) -> 10
+	 * 2 (33%) -> 20
+	 * 10 (9%) -> 100
+	 * 9.9 (9%) -> 99
+	 * 1.5 (40%) -> 15
+	 * 0.5 (66%) -> 5
+	 * 0.1 (90%) -> 1
+	 *
+	 * So, if the coefficient is lower than 10, then
+	 * injected service is more than bfqq service.
+	 */
+	unsigned int inject_coeff;
+	/* amount of service injected in current service slot */
+	unsigned int injected_service;
 };
 
 /**
@@ -423,14 +449,9 @@ struct bfq_data {
 	 */
 	struct rb_root queue_weights_tree;
 	/*
-	 * rbtree of non-queue @bfq_entity weight counters, sorted by
-	 * weight. Used to keep track of whether all @bfq_groups have
-	 * the same weight. The tree contains one counter for each
-	 * distinct weight associated to some active @bfq_group (see
-	 * the comments to the functions bfq_weights_tree_[add|remove]
-	 * for further details).
+	 * number of groups with requests still waiting for completion
 	 */
-	struct rb_root group_weights_tree;
+	unsigned int num_active_groups;
 
 	/*
 	 * Number of bfq_queues containing requests (including the
@@ -825,10 +846,10 @@ struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync);
 void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync);
 struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic);
 void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq);
-void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_entity *entity,
+void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 			  struct rb_root *root);
 void __bfq_weights_tree_remove(struct bfq_data *bfqd,
-			       struct bfq_entity *entity,
+			       struct bfq_queue *bfqq,
 			       struct rb_root *root);
 void bfq_weights_tree_remove(struct bfq_data *bfqd,
 			     struct bfq_queue *bfqq);
diff --git a/block/bfq-wf2q.c b/block/bfq-wf2q.c
index ae52bff43ce4..476b5a90a5a4 100644
--- a/block/bfq-wf2q.c
+++ b/block/bfq-wf2q.c
@@ -788,25 +788,29 @@ __bfq_entity_update_weight_prio(struct bfq_service_tree *old_st,
 		new_weight = entity->orig_weight *
 			     (bfqq ? bfqq->wr_coeff : 1);
 		/*
-		 * If the weight of the entity changes, remove the entity
-		 * from its old weight counter (if there is a counter
-		 * associated with the entity), and add it to the counter
-		 * associated with its new weight.
+		 * If the weight of the entity changes, and the entity is a
+		 * queue, remove the entity from its old weight counter (if
+		 * there is a counter associated with the entity).
 		 */
 		if (prev_weight != new_weight) {
-			root = bfqq ? &bfqd->queue_weights_tree :
-				      &bfqd->group_weights_tree;
-			__bfq_weights_tree_remove(bfqd, entity, root);
+			if (bfqq) {
+				root = &bfqd->queue_weights_tree;
+				__bfq_weights_tree_remove(bfqd, bfqq, root);
+			} else
+				bfqd->num_active_groups--;
 		}
 		entity->weight = new_weight;
 		/*
-		 * Add the entity to its weights tree only if it is
-		 * not associated with a weight-raised queue.
+		 * Add the entity, if it is not a weight-raised queue,
+		 * to the counter associated with its new weight.
 		 */
-		if (prev_weight != new_weight &&
-		    (bfqq ? bfqq->wr_coeff == 1 : 1))
-			/* If we get here, root has been initialized. */
-			bfq_weights_tree_add(bfqd, entity, root);
+		if (prev_weight != new_weight) {
+			if (bfqq && bfqq->wr_coeff == 1) {
+				/* If we get here, root has been initialized. */
+				bfq_weights_tree_add(bfqd, bfqq, root);
+			} else
+				bfqd->num_active_groups++;
+		}
 
 		new_st->wsum += entity->weight;
 
@@ -1012,9 +1016,9 @@ static void __bfq_activate_entity(struct bfq_entity *entity,
 	if (!bfq_entity_to_bfqq(entity)) { /* bfq_group */
 		struct bfq_group *bfqg =
 			container_of(entity, struct bfq_group, entity);
+		struct bfq_data *bfqd = bfqg->bfqd;
 
-		bfq_weights_tree_add(bfqg->bfqd, entity,
-				     &bfqd->group_weights_tree);
+		bfqd->num_active_groups++;
 	}
 #endif
 
@@ -1181,10 +1185,17 @@ bool __bfq_deactivate_entity(struct bfq_entity *entity, bool ins_into_idle_tree)
 	st = bfq_entity_service_tree(entity);
 	is_in_service = entity == sd->in_service_entity;
 
-	if (is_in_service) {
-		bfq_calc_finish(entity, entity->service);
+	bfq_calc_finish(entity, entity->service);
+
+	if (is_in_service)
 		sd->in_service_entity = NULL;
-	}
+	else
+		/*
+		 * Non in-service entity: nobody will take care of
+		 * resetting its service counter on expiration. Do it
+		 * now.
+		 */
+		entity->service = 0;
 
 	if (entity->tree == &st->active)
 		bfq_active_extract(st, entity);
@@ -1685,7 +1696,7 @@ void bfq_add_bfqq_busy(struct bfq_data *bfqd, struct bfq_queue *bfqq)
 
 	if (!bfqq->dispatched)
 		if (bfqq->wr_coeff == 1)
-			bfq_weights_tree_add(bfqd, &bfqq->entity,
+			bfq_weights_tree_add(bfqd, bfqq,
 					     &bfqd->queue_weights_tree);
 
 	if (bfqq->wr_coeff > 1)
diff --git a/block/bio-integrity.c b/block/bio-integrity.c
index 67b5fb861a51..290af497997b 100644
--- a/block/bio-integrity.c
+++ b/block/bio-integrity.c
@@ -306,6 +306,8 @@ bool bio_integrity_prep(struct bio *bio)
 	if (bio_data_dir(bio) == WRITE) {
 		bio_integrity_process(bio, &bio->bi_iter,
 				      bi->profile->generate_fn);
+	} else {
+		bip->bio_iter = bio->bi_iter;
 	}
 	return true;
 
@@ -331,20 +333,14 @@ static void bio_integrity_verify_fn(struct work_struct *work)
 		container_of(work, struct bio_integrity_payload, bip_work);
 	struct bio *bio = bip->bip_bio;
 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
-	struct bvec_iter iter = bio->bi_iter;
 
 	/*
 	 * At the moment verify is called bio's iterator was advanced
 	 * during split and completion, we need to rewind iterator to
 	 * it's original position.
 	 */
-	if (bio_rewind_iter(bio, &iter, iter.bi_done)) {
-		bio->bi_status = bio_integrity_process(bio, &iter,
-						       bi->profile->verify_fn);
-	} else {
-		bio->bi_status = BLK_STS_IOERR;
-	}
-
+	bio->bi_status = bio_integrity_process(bio, &bip->bio_iter,
+						bi->profile->verify_fn);
 	bio_integrity_free(bio);
 	bio_endio(bio);
 }
diff --git a/block/bio.c b/block/bio.c
index 0093bed81c0e..bbfeb4ee2892 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -609,7 +609,9 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src)
 	bio->bi_iter = bio_src->bi_iter;
 	bio->bi_io_vec = bio_src->bi_io_vec;
 
-	bio_clone_blkcg_association(bio, bio_src);
+	bio_clone_blkg_association(bio, bio_src);
+
+	blkcg_bio_issue_init(bio);
 }
 EXPORT_SYMBOL(__bio_clone_fast);
 
@@ -729,7 +731,7 @@ int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page
 	}
 
 	/* If we may be able to merge these biovecs, force a recount */
-	if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec)))
+	if (bio->bi_vcnt > 1 && biovec_phys_mergeable(q, bvec - 1, bvec))
 		bio_clear_flag(bio, BIO_SEG_VALID);
 
  done:
@@ -827,6 +829,8 @@ int bio_add_page(struct bio *bio, struct page *page,
 }
 EXPORT_SYMBOL(bio_add_page);
 
+#define PAGE_PTRS_PER_BVEC     (sizeof(struct bio_vec) / sizeof(struct page *))
+
 /**
  * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio
  * @bio: bio to add pages to
@@ -839,38 +843,35 @@ EXPORT_SYMBOL(bio_add_page);
  */
 static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 {
-	unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt, idx;
+	unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt;
+	unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt;
 	struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
 	struct page **pages = (struct page **)bv;
+	ssize_t size, left;
+	unsigned len, i;
 	size_t offset;
-	ssize_t size;
+
+	/*
+	 * Move page array up in the allocated memory for the bio vecs as far as
+	 * possible so that we can start filling biovecs from the beginning
+	 * without overwriting the temporary page array.
+	*/
+	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
+	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
 
 	size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
 	if (unlikely(size <= 0))
 		return size ? size : -EFAULT;
-	idx = nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE;
 
-	/*
-	 * Deep magic below:  We need to walk the pinned pages backwards
-	 * because we are abusing the space allocated for the bio_vecs
-	 * for the page array.  Because the bio_vecs are larger than the
-	 * page pointers by definition this will always work.  But it also
-	 * means we can't use bio_add_page, so any changes to it's semantics
-	 * need to be reflected here as well.
-	 */
-	bio->bi_iter.bi_size += size;
-	bio->bi_vcnt += nr_pages;
+	for (left = size, i = 0; left > 0; left -= len, i++) {
+		struct page *page = pages[i];
 
-	while (idx--) {
-		bv[idx].bv_page = pages[idx];
-		bv[idx].bv_len = PAGE_SIZE;
-		bv[idx].bv_offset = 0;
+		len = min_t(size_t, PAGE_SIZE - offset, left);
+		if (WARN_ON_ONCE(bio_add_page(bio, page, len, offset) != len))
+			return -EINVAL;
+		offset = 0;
 	}
 
-	bv[0].bv_offset += offset;
-	bv[0].bv_len -= offset;
-	bv[nr_pages - 1].bv_len -= nr_pages * PAGE_SIZE - offset - size;
-
 	iov_iter_advance(iter, size);
 	return 0;
 }
@@ -1807,7 +1808,6 @@ struct bio *bio_split(struct bio *bio, int sectors,
 		bio_integrity_trim(split);
 
 	bio_advance(bio, split->bi_iter.bi_size);
-	bio->bi_iter.bi_done = 0;
 
 	if (bio_flagged(bio, BIO_TRACE_COMPLETION))
 		bio_set_flag(split, BIO_TRACE_COMPLETION);
@@ -1956,69 +1956,151 @@ EXPORT_SYMBOL(bioset_init_from_src);
 
 #ifdef CONFIG_BLK_CGROUP
 
+/**
+ * bio_associate_blkg - associate a bio with the a blkg
+ * @bio: target bio
+ * @blkg: the blkg to associate
+ *
+ * This tries to associate @bio with the specified blkg.  Association failure
+ * is handled by walking up the blkg tree.  Therefore, the blkg associated can
+ * be anything between @blkg and the root_blkg.  This situation only happens
+ * when a cgroup is dying and then the remaining bios will spill to the closest
+ * alive blkg.
+ *
+ * A reference will be taken on the @blkg and will be released when @bio is
+ * freed.
+ */
+int bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg)
+{
+	if (unlikely(bio->bi_blkg))
+		return -EBUSY;
+	bio->bi_blkg = blkg_tryget_closest(blkg);
+	return 0;
+}
+
+/**
+ * __bio_associate_blkg_from_css - internal blkg association function
+ *
+ * This in the core association function that all association paths rely on.
+ * A blkg reference is taken which is released upon freeing of the bio.
+ */
+static int __bio_associate_blkg_from_css(struct bio *bio,
+					 struct cgroup_subsys_state *css)
+{
+	struct request_queue *q = bio->bi_disk->queue;
+	struct blkcg_gq *blkg;
+	int ret;
+
+	rcu_read_lock();
+
+	if (!css || !css->parent)
+		blkg = q->root_blkg;
+	else
+		blkg = blkg_lookup_create(css_to_blkcg(css), q);
+
+	ret = bio_associate_blkg(bio, blkg);
+
+	rcu_read_unlock();
+	return ret;
+}
+
+/**
+ * bio_associate_blkg_from_css - associate a bio with a specified css
+ * @bio: target bio
+ * @css: target css
+ *
+ * Associate @bio with the blkg found by combining the css's blkg and the
+ * request_queue of the @bio.  This falls back to the queue's root_blkg if
+ * the association fails with the css.
+ */
+int bio_associate_blkg_from_css(struct bio *bio,
+				struct cgroup_subsys_state *css)
+{
+	if (unlikely(bio->bi_blkg))
+		return -EBUSY;
+	return __bio_associate_blkg_from_css(bio, css);
+}
+EXPORT_SYMBOL_GPL(bio_associate_blkg_from_css);
+
 #ifdef CONFIG_MEMCG
 /**
- * bio_associate_blkcg_from_page - associate a bio with the page's blkcg
+ * bio_associate_blkg_from_page - associate a bio with the page's blkg
  * @bio: target bio
  * @page: the page to lookup the blkcg from
  *
- * Associate @bio with the blkcg from @page's owning memcg.  This works like
- * every other associate function wrt references.
+ * Associate @bio with the blkg from @page's owning memcg and the respective
+ * request_queue.  If cgroup_e_css returns NULL, fall back to the queue's
+ * root_blkg.
+ *
+ * Note: this must be called after bio has an associated device.
  */
-int bio_associate_blkcg_from_page(struct bio *bio, struct page *page)
+int bio_associate_blkg_from_page(struct bio *bio, struct page *page)
 {
-	struct cgroup_subsys_state *blkcg_css;
+	struct cgroup_subsys_state *css;
+	int ret;
 
-	if (unlikely(bio->bi_css))
+	if (unlikely(bio->bi_blkg))
 		return -EBUSY;
 	if (!page->mem_cgroup)
 		return 0;
-	blkcg_css = cgroup_get_e_css(page->mem_cgroup->css.cgroup,
-				     &io_cgrp_subsys);
-	bio->bi_css = blkcg_css;
-	return 0;
+
+	rcu_read_lock();
+
+	css = cgroup_e_css(page->mem_cgroup->css.cgroup, &io_cgrp_subsys);
+
+	ret = __bio_associate_blkg_from_css(bio, css);
+
+	rcu_read_unlock();
+	return ret;
 }
 #endif /* CONFIG_MEMCG */
 
 /**
- * bio_associate_blkcg - associate a bio with the specified blkcg
+ * bio_associate_create_blkg - associate a bio with a blkg from q
+ * @q: request_queue where bio is going
  * @bio: target bio
- * @blkcg_css: css of the blkcg to associate
- *
- * Associate @bio with the blkcg specified by @blkcg_css.  Block layer will
- * treat @bio as if it were issued by a task which belongs to the blkcg.
  *
- * This function takes an extra reference of @blkcg_css which will be put
- * when @bio is released.  The caller must own @bio and is responsible for
- * synchronizing calls to this function.
+ * Associate @bio with the blkg found from the bio's css and the request_queue.
+ * If one is not found, bio_lookup_blkg creates the blkg.  This falls back to
+ * the queue's root_blkg if association fails.
  */
-int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css)
+int bio_associate_create_blkg(struct request_queue *q, struct bio *bio)
 {
-	if (unlikely(bio->bi_css))
-		return -EBUSY;
-	css_get(blkcg_css);
-	bio->bi_css = blkcg_css;
-	return 0;
+	struct cgroup_subsys_state *css;
+	int ret = 0;
+
+	/* someone has already associated this bio with a blkg */
+	if (bio->bi_blkg)
+		return ret;
+
+	rcu_read_lock();
+
+	css = blkcg_css();
+
+	ret = __bio_associate_blkg_from_css(bio, css);
+
+	rcu_read_unlock();
+	return ret;
 }
-EXPORT_SYMBOL_GPL(bio_associate_blkcg);
 
 /**
- * bio_associate_blkg - associate a bio with the specified blkg
+ * bio_reassociate_blkg - reassociate a bio with a blkg from q
+ * @q: request_queue where bio is going
  * @bio: target bio
- * @blkg: the blkg to associate
  *
- * Associate @bio with the blkg specified by @blkg.  This is the queue specific
- * blkcg information associated with the @bio, a reference will be taken on the
- * @blkg and will be freed when the bio is freed.
+ * When submitting a bio, multiple recursive calls to make_request() may occur.
+ * This causes the initial associate done in blkcg_bio_issue_check() to be
+ * incorrect and reference the prior request_queue.  This performs reassociation
+ * when this situation happens.
  */
-int bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg)
+int bio_reassociate_blkg(struct request_queue *q, struct bio *bio)
 {
-	if (unlikely(bio->bi_blkg))
-		return -EBUSY;
-	if (!blkg_try_get(blkg))
-		return -ENODEV;
-	bio->bi_blkg = blkg;
-	return 0;
+	if (bio->bi_blkg) {
+		blkg_put(bio->bi_blkg);
+		bio->bi_blkg = NULL;
+	}
+
+	return bio_associate_create_blkg(q, bio);
 }
 
 /**
@@ -2031,10 +2113,6 @@ void bio_disassociate_task(struct bio *bio)
 		put_io_context(bio->bi_ioc);
 		bio->bi_ioc = NULL;
 	}
-	if (bio->bi_css) {
-		css_put(bio->bi_css);
-		bio->bi_css = NULL;
-	}
 	if (bio->bi_blkg) {
 		blkg_put(bio->bi_blkg);
 		bio->bi_blkg = NULL;
@@ -2042,16 +2120,16 @@ void bio_disassociate_task(struct bio *bio)
 }
 
 /**
- * bio_clone_blkcg_association - clone blkcg association from src to dst bio
+ * bio_clone_blkg_association - clone blkg association from src to dst bio
  * @dst: destination bio
  * @src: source bio
  */
-void bio_clone_blkcg_association(struct bio *dst, struct bio *src)
+void bio_clone_blkg_association(struct bio *dst, struct bio *src)
 {
-	if (src->bi_css)
-		WARN_ON(bio_associate_blkcg(dst, src->bi_css));
+	if (src->bi_blkg)
+		bio_associate_blkg(dst, src->bi_blkg);
 }
-EXPORT_SYMBOL_GPL(bio_clone_blkcg_association);
+EXPORT_SYMBOL_GPL(bio_clone_blkg_association);
 #endif /* CONFIG_BLK_CGROUP */
 
 static void __init biovec_init_slabs(void)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index c630e02836a8..992da5592c6e 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -84,6 +84,37 @@ static void blkg_free(struct blkcg_gq *blkg)
 	kfree(blkg);
 }
 
+static void __blkg_release(struct rcu_head *rcu)
+{
+	struct blkcg_gq *blkg = container_of(rcu, struct blkcg_gq, rcu_head);
+
+	percpu_ref_exit(&blkg->refcnt);
+
+	/* release the blkcg and parent blkg refs this blkg has been holding */
+	css_put(&blkg->blkcg->css);
+	if (blkg->parent)
+		blkg_put(blkg->parent);
+
+	wb_congested_put(blkg->wb_congested);
+
+	blkg_free(blkg);
+}
+
+/*
+ * A group is RCU protected, but having an rcu lock does not mean that one
+ * can access all the fields of blkg and assume these are valid.  For
+ * example, don't try to follow throtl_data and request queue links.
+ *
+ * Having a reference to blkg under an rcu allows accesses to only values
+ * local to groups like group stats and group rate limits.
+ */
+static void blkg_release(struct percpu_ref *ref)
+{
+	struct blkcg_gq *blkg = container_of(ref, struct blkcg_gq, refcnt);
+
+	call_rcu(&blkg->rcu_head, __blkg_release);
+}
+
 /**
  * blkg_alloc - allocate a blkg
  * @blkcg: block cgroup the new blkg is associated with
@@ -110,7 +141,6 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
 	blkg->q = q;
 	INIT_LIST_HEAD(&blkg->q_node);
 	blkg->blkcg = blkcg;
-	atomic_set(&blkg->refcnt, 1);
 
 	/* root blkg uses @q->root_rl, init rl only for !root blkgs */
 	if (blkcg != &blkcg_root) {
@@ -217,6 +247,11 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
 		blkg_get(blkg->parent);
 	}
 
+	ret = percpu_ref_init(&blkg->refcnt, blkg_release, 0,
+			      GFP_NOWAIT | __GFP_NOWARN);
+	if (ret)
+		goto err_cancel_ref;
+
 	/* invoke per-policy init */
 	for (i = 0; i < BLKCG_MAX_POLS; i++) {
 		struct blkcg_policy *pol = blkcg_policy[i];
@@ -249,6 +284,8 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
 	blkg_put(blkg);
 	return ERR_PTR(ret);
 
+err_cancel_ref:
+	percpu_ref_exit(&blkg->refcnt);
 err_put_congested:
 	wb_congested_put(wb_congested);
 err_put_css:
@@ -259,7 +296,7 @@ err_free_blkg:
 }
 
 /**
- * blkg_lookup_create - lookup blkg, try to create one if not there
+ * __blkg_lookup_create - lookup blkg, try to create one if not there
  * @blkcg: blkcg of interest
  * @q: request_queue of interest
  *
@@ -268,12 +305,11 @@ err_free_blkg:
  * that all non-root blkg's have access to the parent blkg.  This function
  * should be called under RCU read lock and @q->queue_lock.
  *
- * Returns pointer to the looked up or created blkg on success, ERR_PTR()
- * value on error.  If @q is dead, returns ERR_PTR(-EINVAL).  If @q is not
- * dead and bypassing, returns ERR_PTR(-EBUSY).
+ * Returns the blkg or the closest blkg if blkg_create fails as it walks
+ * down from root.
  */
-struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
-				    struct request_queue *q)
+struct blkcg_gq *__blkg_lookup_create(struct blkcg *blkcg,
+				      struct request_queue *q)
 {
 	struct blkcg_gq *blkg;
 
@@ -285,7 +321,7 @@ struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
 	 * we shouldn't allow anything to go through for a bypassing queue.
 	 */
 	if (unlikely(blk_queue_bypass(q)))
-		return ERR_PTR(blk_queue_dying(q) ? -ENODEV : -EBUSY);
+		return q->root_blkg;
 
 	blkg = __blkg_lookup(blkcg, q, true);
 	if (blkg)
@@ -293,23 +329,58 @@ struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
 
 	/*
 	 * Create blkgs walking down from blkcg_root to @blkcg, so that all
-	 * non-root blkgs have access to their parents.
+	 * non-root blkgs have access to their parents.  Returns the closest
+	 * blkg to the intended blkg should blkg_create() fail.
 	 */
 	while (true) {
 		struct blkcg *pos = blkcg;
 		struct blkcg *parent = blkcg_parent(blkcg);
-
-		while (parent && !__blkg_lookup(parent, q, false)) {
+		struct blkcg_gq *ret_blkg = q->root_blkg;
+
+		while (parent) {
+			blkg = __blkg_lookup(parent, q, false);
+			if (blkg) {
+				/* remember closest blkg */
+				ret_blkg = blkg;
+				break;
+			}
 			pos = parent;
 			parent = blkcg_parent(parent);
 		}
 
 		blkg = blkg_create(pos, q, NULL);
-		if (pos == blkcg || IS_ERR(blkg))
+		if (IS_ERR(blkg))
+			return ret_blkg;
+		if (pos == blkcg)
 			return blkg;
 	}
 }
 
+/**
+ * blkg_lookup_create - find or create a blkg
+ * @blkcg: target block cgroup
+ * @q: target request_queue
+ *
+ * This looks up or creates the blkg representing the unique pair
+ * of the blkcg and the request_queue.
+ */
+struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
+				    struct request_queue *q)
+{
+	struct blkcg_gq *blkg = blkg_lookup(blkcg, q);
+	unsigned long flags;
+
+	if (unlikely(!blkg)) {
+		spin_lock_irqsave(q->queue_lock, flags);
+
+		blkg = __blkg_lookup_create(blkcg, q);
+
+		spin_unlock_irqrestore(q->queue_lock, flags);
+	}
+
+	return blkg;
+}
+
 static void blkg_destroy(struct blkcg_gq *blkg)
 {
 	struct blkcg *blkcg = blkg->blkcg;
@@ -353,7 +424,7 @@ static void blkg_destroy(struct blkcg_gq *blkg)
 	 * Put the reference taken at the time of creation so that when all
 	 * queues are gone, group can be destroyed.
 	 */
-	blkg_put(blkg);
+	percpu_ref_kill(&blkg->refcnt);
 }
 
 /**
@@ -381,29 +452,6 @@ static void blkg_destroy_all(struct request_queue *q)
 }
 
 /*
- * A group is RCU protected, but having an rcu lock does not mean that one
- * can access all the fields of blkg and assume these are valid.  For
- * example, don't try to follow throtl_data and request queue links.
- *
- * Having a reference to blkg under an rcu allows accesses to only values
- * local to groups like group stats and group rate limits.
- */
-void __blkg_release_rcu(struct rcu_head *rcu_head)
-{
-	struct blkcg_gq *blkg = container_of(rcu_head, struct blkcg_gq, rcu_head);
-
-	/* release the blkcg and parent blkg refs this blkg has been holding */
-	css_put(&blkg->blkcg->css);
-	if (blkg->parent)
-		blkg_put(blkg->parent);
-
-	wb_congested_put(blkg->wb_congested);
-
-	blkg_free(blkg);
-}
-EXPORT_SYMBOL_GPL(__blkg_release_rcu);
-
-/*
  * The next function used by blk_queue_for_each_rl().  It's a bit tricky
  * because the root blkg uses @q->root_rl instead of its own rl.
  */
@@ -1748,8 +1796,7 @@ void blkcg_maybe_throttle_current(void)
 	blkg = blkg_lookup(blkcg, q);
 	if (!blkg)
 		goto out;
-	blkg = blkg_try_get(blkg);
-	if (!blkg)
+	if (!blkg_tryget(blkg))
 		goto out;
 	rcu_read_unlock();
 
diff --git a/block/blk-core.c b/block/blk-core.c
index cff0a60ee200..3ed60723e242 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -42,6 +42,7 @@
 #include "blk.h"
 #include "blk-mq.h"
 #include "blk-mq-sched.h"
+#include "blk-pm.h"
 #include "blk-rq-qos.h"
 
 #ifdef CONFIG_DEBUG_FS
@@ -421,24 +422,25 @@ void blk_sync_queue(struct request_queue *q)
 EXPORT_SYMBOL(blk_sync_queue);
 
 /**
- * blk_set_preempt_only - set QUEUE_FLAG_PREEMPT_ONLY
+ * blk_set_pm_only - increment pm_only counter
  * @q: request queue pointer
- *
- * Returns the previous value of the PREEMPT_ONLY flag - 0 if the flag was not
- * set and 1 if the flag was already set.
  */
-int blk_set_preempt_only(struct request_queue *q)
+void blk_set_pm_only(struct request_queue *q)
 {
-	return blk_queue_flag_test_and_set(QUEUE_FLAG_PREEMPT_ONLY, q);
+	atomic_inc(&q->pm_only);
 }
-EXPORT_SYMBOL_GPL(blk_set_preempt_only);
+EXPORT_SYMBOL_GPL(blk_set_pm_only);
 
-void blk_clear_preempt_only(struct request_queue *q)
+void blk_clear_pm_only(struct request_queue *q)
 {
-	blk_queue_flag_clear(QUEUE_FLAG_PREEMPT_ONLY, q);
-	wake_up_all(&q->mq_freeze_wq);
+	int pm_only;
+
+	pm_only = atomic_dec_return(&q->pm_only);
+	WARN_ON_ONCE(pm_only < 0);
+	if (pm_only == 0)
+		wake_up_all(&q->mq_freeze_wq);
 }
-EXPORT_SYMBOL_GPL(blk_clear_preempt_only);
+EXPORT_SYMBOL_GPL(blk_clear_pm_only);
 
 /**
  * __blk_run_queue_uncond - run a queue whether or not it has been stopped
@@ -917,7 +919,7 @@ EXPORT_SYMBOL(blk_alloc_queue);
  */
 int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 {
-	const bool preempt = flags & BLK_MQ_REQ_PREEMPT;
+	const bool pm = flags & BLK_MQ_REQ_PREEMPT;
 
 	while (true) {
 		bool success = false;
@@ -925,11 +927,11 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 		rcu_read_lock();
 		if (percpu_ref_tryget_live(&q->q_usage_counter)) {
 			/*
-			 * The code that sets the PREEMPT_ONLY flag is
-			 * responsible for ensuring that that flag is globally
-			 * visible before the queue is unfrozen.
+			 * The code that increments the pm_only counter is
+			 * responsible for ensuring that that counter is
+			 * globally visible before the queue is unfrozen.
 			 */
-			if (preempt || !blk_queue_preempt_only(q)) {
+			if (pm || !blk_queue_pm_only(q)) {
 				success = true;
 			} else {
 				percpu_ref_put(&q->q_usage_counter);
@@ -954,7 +956,8 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 
 		wait_event(q->mq_freeze_wq,
 			   (atomic_read(&q->mq_freeze_depth) == 0 &&
-			    (preempt || !blk_queue_preempt_only(q))) ||
+			    (pm || (blk_pm_request_resume(q),
+				    !blk_queue_pm_only(q)))) ||
 			   blk_queue_dying(q));
 		if (blk_queue_dying(q))
 			return -ENODEV;
@@ -1051,8 +1054,7 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
 	mutex_init(&q->sysfs_lock);
 	spin_lock_init(&q->__queue_lock);
 
-	if (!q->mq_ops)
-		q->queue_lock = lock ? : &q->__queue_lock;
+	q->queue_lock = lock ? : &q->__queue_lock;
 
 	/*
 	 * A queue starts its life with bypass turned on to avoid
@@ -1160,7 +1162,7 @@ int blk_init_allocated_queue(struct request_queue *q)
 {
 	WARN_ON_ONCE(q->mq_ops);
 
-	q->fq = blk_alloc_flush_queue(q, NUMA_NO_NODE, q->cmd_size);
+	q->fq = blk_alloc_flush_queue(q, NUMA_NO_NODE, q->cmd_size, GFP_KERNEL);
 	if (!q->fq)
 		return -ENOMEM;
 
@@ -1726,16 +1728,6 @@ void part_round_stats(struct request_queue *q, int cpu, struct hd_struct *part)
 }
 EXPORT_SYMBOL_GPL(part_round_stats);
 
-#ifdef CONFIG_PM
-static void blk_pm_put_request(struct request *rq)
-{
-	if (rq->q->dev && !(rq->rq_flags & RQF_PM) && !--rq->q->nr_pending)
-		pm_runtime_mark_last_busy(rq->q->dev);
-}
-#else
-static inline void blk_pm_put_request(struct request *rq) {}
-#endif
-
 void __blk_put_request(struct request_queue *q, struct request *req)
 {
 	req_flags_t rq_flags = req->rq_flags;
@@ -1752,6 +1744,7 @@ void __blk_put_request(struct request_queue *q, struct request *req)
 
 	blk_req_zone_write_unlock(req);
 	blk_pm_put_request(req);
+	blk_pm_mark_last_busy(req);
 
 	elv_completed_request(q, req);
 
@@ -2440,6 +2433,7 @@ blk_qc_t generic_make_request(struct bio *bio)
 			if (q)
 				blk_queue_exit(q);
 			q = bio->bi_disk->queue;
+			bio_reassociate_blkg(q, bio);
 			flags = 0;
 			if (bio->bi_opf & REQ_NOWAIT)
 				flags = BLK_MQ_REQ_NOWAIT;
@@ -2750,30 +2744,6 @@ void blk_account_io_done(struct request *req, u64 now)
 	}
 }
 
-#ifdef CONFIG_PM
-/*
- * Don't process normal requests when queue is suspended
- * or in the process of suspending/resuming
- */
-static bool blk_pm_allow_request(struct request *rq)
-{
-	switch (rq->q->rpm_status) {
-	case RPM_RESUMING:
-	case RPM_SUSPENDING:
-		return rq->rq_flags & RQF_PM;
-	case RPM_SUSPENDED:
-		return false;
-	default:
-		return true;
-	}
-}
-#else
-static bool blk_pm_allow_request(struct request *rq)
-{
-	return true;
-}
-#endif
-
 void blk_account_io_start(struct request *rq, bool new_io)
 {
 	struct hd_struct *part;
@@ -2819,11 +2789,14 @@ static struct request *elv_next_request(struct request_queue *q)
 
 	while (1) {
 		list_for_each_entry(rq, &q->queue_head, queuelist) {
-			if (blk_pm_allow_request(rq))
-				return rq;
-
-			if (rq->rq_flags & RQF_SOFTBARRIER)
-				break;
+#ifdef CONFIG_PM
+			/*
+			 * If a request gets queued in state RPM_SUSPENDED
+			 * then that's a kernel bug.
+			 */
+			WARN_ON_ONCE(q->rpm_status == RPM_SUSPENDED);
+#endif
+			return rq;
 		}
 
 		/*
@@ -3755,191 +3728,6 @@ void blk_finish_plug(struct blk_plug *plug)
 }
 EXPORT_SYMBOL(blk_finish_plug);
 
-#ifdef CONFIG_PM
-/**
- * blk_pm_runtime_init - Block layer runtime PM initialization routine
- * @q: the queue of the device
- * @dev: the device the queue belongs to
- *
- * Description:
- *    Initialize runtime-PM-related fields for @q and start auto suspend for
- *    @dev. Drivers that want to take advantage of request-based runtime PM
- *    should call this function after @dev has been initialized, and its
- *    request queue @q has been allocated, and runtime PM for it can not happen
- *    yet(either due to disabled/forbidden or its usage_count > 0). In most
- *    cases, driver should call this function before any I/O has taken place.
- *
- *    This function takes care of setting up using auto suspend for the device,
- *    the autosuspend delay is set to -1 to make runtime suspend impossible
- *    until an updated value is either set by user or by driver. Drivers do
- *    not need to touch other autosuspend settings.
- *
- *    The block layer runtime PM is request based, so only works for drivers
- *    that use request as their IO unit instead of those directly use bio's.
- */
-void blk_pm_runtime_init(struct request_queue *q, struct device *dev)
-{
-	/* Don't enable runtime PM for blk-mq until it is ready */
-	if (q->mq_ops) {
-		pm_runtime_disable(dev);
-		return;
-	}
-
-	q->dev = dev;
-	q->rpm_status = RPM_ACTIVE;
-	pm_runtime_set_autosuspend_delay(q->dev, -1);
-	pm_runtime_use_autosuspend(q->dev);
-}
-EXPORT_SYMBOL(blk_pm_runtime_init);
-
-/**
- * blk_pre_runtime_suspend - Pre runtime suspend check
- * @q: the queue of the device
- *
- * Description:
- *    This function will check if runtime suspend is allowed for the device
- *    by examining if there are any requests pending in the queue. If there
- *    are requests pending, the device can not be runtime suspended; otherwise,
- *    the queue's status will be updated to SUSPENDING and the driver can
- *    proceed to suspend the device.
- *
- *    For the not allowed case, we mark last busy for the device so that
- *    runtime PM core will try to autosuspend it some time later.
- *
- *    This function should be called near the start of the device's
- *    runtime_suspend callback.
- *
- * Return:
- *    0		- OK to runtime suspend the device
- *    -EBUSY	- Device should not be runtime suspended
- */
-int blk_pre_runtime_suspend(struct request_queue *q)
-{
-	int ret = 0;
-
-	if (!q->dev)
-		return ret;
-
-	spin_lock_irq(q->queue_lock);
-	if (q->nr_pending) {
-		ret = -EBUSY;
-		pm_runtime_mark_last_busy(q->dev);
-	} else {
-		q->rpm_status = RPM_SUSPENDING;
-	}
-	spin_unlock_irq(q->queue_lock);
-	return ret;
-}
-EXPORT_SYMBOL(blk_pre_runtime_suspend);
-
-/**
- * blk_post_runtime_suspend - Post runtime suspend processing
- * @q: the queue of the device
- * @err: return value of the device's runtime_suspend function
- *
- * Description:
- *    Update the queue's runtime status according to the return value of the
- *    device's runtime suspend function and mark last busy for the device so
- *    that PM core will try to auto suspend the device at a later time.
- *
- *    This function should be called near the end of the device's
- *    runtime_suspend callback.
- */
-void blk_post_runtime_suspend(struct request_queue *q, int err)
-{
-	if (!q->dev)
-		return;
-
-	spin_lock_irq(q->queue_lock);
-	if (!err) {
-		q->rpm_status = RPM_SUSPENDED;
-	} else {
-		q->rpm_status = RPM_ACTIVE;
-		pm_runtime_mark_last_busy(q->dev);
-	}
-	spin_unlock_irq(q->queue_lock);
-}
-EXPORT_SYMBOL(blk_post_runtime_suspend);
-
-/**
- * blk_pre_runtime_resume - Pre runtime resume processing
- * @q: the queue of the device
- *
- * Description:
- *    Update the queue's runtime status to RESUMING in preparation for the
- *    runtime resume of the device.
- *
- *    This function should be called near the start of the device's
- *    runtime_resume callback.
- */
-void blk_pre_runtime_resume(struct request_queue *q)
-{
-	if (!q->dev)
-		return;
-
-	spin_lock_irq(q->queue_lock);
-	q->rpm_status = RPM_RESUMING;
-	spin_unlock_irq(q->queue_lock);
-}
-EXPORT_SYMBOL(blk_pre_runtime_resume);
-
-/**
- * blk_post_runtime_resume - Post runtime resume processing
- * @q: the queue of the device
- * @err: return value of the device's runtime_resume function
- *
- * Description:
- *    Update the queue's runtime status according to the return value of the
- *    device's runtime_resume function. If it is successfully resumed, process
- *    the requests that are queued into the device's queue when it is resuming
- *    and then mark last busy and initiate autosuspend for it.
- *
- *    This function should be called near the end of the device's
- *    runtime_resume callback.
- */
-void blk_post_runtime_resume(struct request_queue *q, int err)
-{
-	if (!q->dev)
-		return;
-
-	spin_lock_irq(q->queue_lock);
-	if (!err) {
-		q->rpm_status = RPM_ACTIVE;
-		__blk_run_queue(q);
-		pm_runtime_mark_last_busy(q->dev);
-		pm_request_autosuspend(q->dev);
-	} else {
-		q->rpm_status = RPM_SUSPENDED;
-	}
-	spin_unlock_irq(q->queue_lock);
-}
-EXPORT_SYMBOL(blk_post_runtime_resume);
-
-/**
- * blk_set_runtime_active - Force runtime status of the queue to be active
- * @q: the queue of the device
- *
- * If the device is left runtime suspended during system suspend the resume
- * hook typically resumes the device and corrects runtime status
- * accordingly. However, that does not affect the queue runtime PM status
- * which is still "suspended". This prevents processing requests from the
- * queue.
- *
- * This function can be used in driver's resume hook to correct queue
- * runtime PM status and re-enable peeking requests from the queue. It
- * should be called before first request is added to the queue.
- */
-void blk_set_runtime_active(struct request_queue *q)
-{
-	spin_lock_irq(q->queue_lock);
-	q->rpm_status = RPM_ACTIVE;
-	pm_runtime_mark_last_busy(q->dev);
-	pm_request_autosuspend(q->dev);
-	spin_unlock_irq(q->queue_lock);
-}
-EXPORT_SYMBOL(blk_set_runtime_active);
-#endif
-
 int __init blk_dev_init(void)
 {
 	BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS));
diff --git a/block/blk-flush.c b/block/blk-flush.c
index ce41f666de3e..8b44b86779da 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -566,12 +566,12 @@ int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
 EXPORT_SYMBOL(blkdev_issue_flush);
 
 struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
-		int node, int cmd_size)
+		int node, int cmd_size, gfp_t flags)
 {
 	struct blk_flush_queue *fq;
 	int rq_sz = sizeof(struct request);
 
-	fq = kzalloc_node(sizeof(*fq), GFP_KERNEL, node);
+	fq = kzalloc_node(sizeof(*fq), flags, node);
 	if (!fq)
 		goto fail;
 
@@ -579,7 +579,7 @@ struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
 		spin_lock_init(&fq->mq_flush_lock);
 
 	rq_sz = round_up(rq_sz + cmd_size, cache_line_size());
-	fq->flush_rq = kzalloc_node(rq_sz, GFP_KERNEL, node);
+	fq->flush_rq = kzalloc_node(rq_sz, flags, node);
 	if (!fq->flush_rq)
 		goto fail_rq;
 
diff --git a/block/blk-integrity.c b/block/blk-integrity.c
index 6121611e1316..d1ab089e0919 100644
--- a/block/blk-integrity.c
+++ b/block/blk-integrity.c
@@ -49,12 +49,8 @@ int blk_rq_count_integrity_sg(struct request_queue *q, struct bio *bio)
 	bio_for_each_integrity_vec(iv, bio, iter) {
 
 		if (prev) {
-			if (!BIOVEC_PHYS_MERGEABLE(&ivprv, &iv))
+			if (!biovec_phys_mergeable(q, &ivprv, &iv))
 				goto new_segment;
-
-			if (!BIOVEC_SEG_BOUNDARY(q, &ivprv, &iv))
-				goto new_segment;
-
 			if (seg_size + iv.bv_len > queue_max_segment_size(q))
 				goto new_segment;
 
@@ -95,12 +91,8 @@ int blk_rq_map_integrity_sg(struct request_queue *q, struct bio *bio,
 	bio_for_each_integrity_vec(iv, bio, iter) {
 
 		if (prev) {
-			if (!BIOVEC_PHYS_MERGEABLE(&ivprv, &iv))
+			if (!biovec_phys_mergeable(q, &ivprv, &iv))
 				goto new_segment;
-
-			if (!BIOVEC_SEG_BOUNDARY(q, &ivprv, &iv))
-				goto new_segment;
-
 			if (sg->length + iv.bv_len > queue_max_segment_size(q))
 				goto new_segment;
 
diff --git a/block/blk-iolatency.c b/block/blk-iolatency.c
index 19923f8a029d..35c48d7b8f78 100644
--- a/block/blk-iolatency.c
+++ b/block/blk-iolatency.c
@@ -115,9 +115,22 @@ struct child_latency_info {
 	atomic_t scale_cookie;
 };
 
+struct percentile_stats {
+	u64 total;
+	u64 missed;
+};
+
+struct latency_stat {
+	union {
+		struct percentile_stats ps;
+		struct blk_rq_stat rqs;
+	};
+};
+
 struct iolatency_grp {
 	struct blkg_policy_data pd;
-	struct blk_rq_stat __percpu *stats;
+	struct latency_stat __percpu *stats;
+	struct latency_stat cur_stat;
 	struct blk_iolatency *blkiolat;
 	struct rq_depth rq_depth;
 	struct rq_wait rq_wait;
@@ -132,6 +145,7 @@ struct iolatency_grp {
 	/* Our current number of IO's for the last summation. */
 	u64 nr_samples;
 
+	bool ssd;
 	struct child_latency_info child_lat;
 };
 
@@ -172,6 +186,80 @@ static inline struct blkcg_gq *lat_to_blkg(struct iolatency_grp *iolat)
 	return pd_to_blkg(&iolat->pd);
 }
 
+static inline void latency_stat_init(struct iolatency_grp *iolat,
+				     struct latency_stat *stat)
+{
+	if (iolat->ssd) {
+		stat->ps.total = 0;
+		stat->ps.missed = 0;
+	} else
+		blk_rq_stat_init(&stat->rqs);
+}
+
+static inline void latency_stat_sum(struct iolatency_grp *iolat,
+				    struct latency_stat *sum,
+				    struct latency_stat *stat)
+{
+	if (iolat->ssd) {
+		sum->ps.total += stat->ps.total;
+		sum->ps.missed += stat->ps.missed;
+	} else
+		blk_rq_stat_sum(&sum->rqs, &stat->rqs);
+}
+
+static inline void latency_stat_record_time(struct iolatency_grp *iolat,
+					    u64 req_time)
+{
+	struct latency_stat *stat = get_cpu_ptr(iolat->stats);
+	if (iolat->ssd) {
+		if (req_time >= iolat->min_lat_nsec)
+			stat->ps.missed++;
+		stat->ps.total++;
+	} else
+		blk_rq_stat_add(&stat->rqs, req_time);
+	put_cpu_ptr(stat);
+}
+
+static inline bool latency_sum_ok(struct iolatency_grp *iolat,
+				  struct latency_stat *stat)
+{
+	if (iolat->ssd) {
+		u64 thresh = div64_u64(stat->ps.total, 10);
+		thresh = max(thresh, 1ULL);
+		return stat->ps.missed < thresh;
+	}
+	return stat->rqs.mean <= iolat->min_lat_nsec;
+}
+
+static inline u64 latency_stat_samples(struct iolatency_grp *iolat,
+				       struct latency_stat *stat)
+{
+	if (iolat->ssd)
+		return stat->ps.total;
+	return stat->rqs.nr_samples;
+}
+
+static inline void iolat_update_total_lat_avg(struct iolatency_grp *iolat,
+					      struct latency_stat *stat)
+{
+	int exp_idx;
+
+	if (iolat->ssd)
+		return;
+
+	/*
+	 * CALC_LOAD takes in a number stored in fixed point representation.
+	 * Because we are using this for IO time in ns, the values stored
+	 * are significantly larger than the FIXED_1 denominator (2048).
+	 * Therefore, rounding errors in the calculation are negligible and
+	 * can be ignored.
+	 */
+	exp_idx = min_t(int, BLKIOLATENCY_NR_EXP_FACTORS - 1,
+			div64_u64(iolat->cur_win_nsec,
+				  BLKIOLATENCY_EXP_BUCKET_SIZE));
+	CALC_LOAD(iolat->lat_avg, iolatency_exp_factors[exp_idx], stat->rqs.mean);
+}
+
 static inline bool iolatency_may_queue(struct iolatency_grp *iolat,
 				       wait_queue_entry_t *wait,
 				       bool first_block)
@@ -255,7 +343,7 @@ static void scale_cookie_change(struct blk_iolatency *blkiolat,
 				struct child_latency_info *lat_info,
 				bool up)
 {
-	unsigned long qd = blk_queue_depth(blkiolat->rqos.q);
+	unsigned long qd = blkiolat->rqos.q->nr_requests;
 	unsigned long scale = scale_amount(qd, up);
 	unsigned long old = atomic_read(&lat_info->scale_cookie);
 	unsigned long max_scale = qd << 1;
@@ -295,10 +383,9 @@ static void scale_cookie_change(struct blk_iolatency *blkiolat,
  */
 static void scale_change(struct iolatency_grp *iolat, bool up)
 {
-	unsigned long qd = blk_queue_depth(iolat->blkiolat->rqos.q);
+	unsigned long qd = iolat->blkiolat->rqos.q->nr_requests;
 	unsigned long scale = scale_amount(qd, up);
 	unsigned long old = iolat->rq_depth.max_depth;
-	bool changed = false;
 
 	if (old > qd)
 		old = qd;
@@ -308,15 +395,13 @@ static void scale_change(struct iolatency_grp *iolat, bool up)
 			return;
 
 		if (old < qd) {
-			changed = true;
 			old += scale;
 			old = min(old, qd);
 			iolat->rq_depth.max_depth = old;
 			wake_up_all(&iolat->rq_wait.wait);
 		}
-	} else if (old > 1) {
+	} else {
 		old >>= 1;
-		changed = true;
 		iolat->rq_depth.max_depth = max(old, 1UL);
 	}
 }
@@ -369,7 +454,7 @@ static void check_scale_change(struct iolatency_grp *iolat)
 		 * scale down event.
 		 */
 		samples_thresh = lat_info->nr_samples * 5;
-		samples_thresh = div64_u64(samples_thresh, 100);
+		samples_thresh = max(1ULL, div64_u64(samples_thresh, 100));
 		if (iolat->nr_samples <= samples_thresh)
 			return;
 	}
@@ -395,34 +480,12 @@ static void blkcg_iolatency_throttle(struct rq_qos *rqos, struct bio *bio,
 				     spinlock_t *lock)
 {
 	struct blk_iolatency *blkiolat = BLKIOLATENCY(rqos);
-	struct blkcg *blkcg;
-	struct blkcg_gq *blkg;
-	struct request_queue *q = rqos->q;
+	struct blkcg_gq *blkg = bio->bi_blkg;
 	bool issue_as_root = bio_issue_as_root_blkg(bio);
 
 	if (!blk_iolatency_enabled(blkiolat))
 		return;
 
-	rcu_read_lock();
-	blkcg = bio_blkcg(bio);
-	bio_associate_blkcg(bio, &blkcg->css);
-	blkg = blkg_lookup(blkcg, q);
-	if (unlikely(!blkg)) {
-		if (!lock)
-			spin_lock_irq(q->queue_lock);
-		blkg = blkg_lookup_create(blkcg, q);
-		if (IS_ERR(blkg))
-			blkg = NULL;
-		if (!lock)
-			spin_unlock_irq(q->queue_lock);
-	}
-	if (!blkg)
-		goto out;
-
-	bio_issue_init(&bio->bi_issue, bio_sectors(bio));
-	bio_associate_blkg(bio, blkg);
-out:
-	rcu_read_unlock();
 	while (blkg && blkg->parent) {
 		struct iolatency_grp *iolat = blkg_to_lat(blkg);
 		if (!iolat) {
@@ -443,7 +506,6 @@ static void iolatency_record_time(struct iolatency_grp *iolat,
 				  struct bio_issue *issue, u64 now,
 				  bool issue_as_root)
 {
-	struct blk_rq_stat *rq_stat;
 	u64 start = bio_issue_time(issue);
 	u64 req_time;
 
@@ -469,9 +531,7 @@ static void iolatency_record_time(struct iolatency_grp *iolat,
 		return;
 	}
 
-	rq_stat = get_cpu_ptr(iolat->stats);
-	blk_rq_stat_add(rq_stat, req_time);
-	put_cpu_ptr(rq_stat);
+	latency_stat_record_time(iolat, req_time);
 }
 
 #define BLKIOLATENCY_MIN_ADJUST_TIME (500 * NSEC_PER_MSEC)
@@ -482,17 +542,17 @@ static void iolatency_check_latencies(struct iolatency_grp *iolat, u64 now)
 	struct blkcg_gq *blkg = lat_to_blkg(iolat);
 	struct iolatency_grp *parent;
 	struct child_latency_info *lat_info;
-	struct blk_rq_stat stat;
+	struct latency_stat stat;
 	unsigned long flags;
-	int cpu, exp_idx;
+	int cpu;
 
-	blk_rq_stat_init(&stat);
+	latency_stat_init(iolat, &stat);
 	preempt_disable();
 	for_each_online_cpu(cpu) {
-		struct blk_rq_stat *s;
+		struct latency_stat *s;
 		s = per_cpu_ptr(iolat->stats, cpu);
-		blk_rq_stat_sum(&stat, s);
-		blk_rq_stat_init(s);
+		latency_stat_sum(iolat, &stat, s);
+		latency_stat_init(iolat, s);
 	}
 	preempt_enable();
 
@@ -502,41 +562,36 @@ static void iolatency_check_latencies(struct iolatency_grp *iolat, u64 now)
 
 	lat_info = &parent->child_lat;
 
-	/*
-	 * CALC_LOAD takes in a number stored in fixed point representation.
-	 * Because we are using this for IO time in ns, the values stored
-	 * are significantly larger than the FIXED_1 denominator (2048).
-	 * Therefore, rounding errors in the calculation are negligible and
-	 * can be ignored.
-	 */
-	exp_idx = min_t(int, BLKIOLATENCY_NR_EXP_FACTORS - 1,
-			div64_u64(iolat->cur_win_nsec,
-				  BLKIOLATENCY_EXP_BUCKET_SIZE));
-	CALC_LOAD(iolat->lat_avg, iolatency_exp_factors[exp_idx], stat.mean);
+	iolat_update_total_lat_avg(iolat, &stat);
 
 	/* Everything is ok and we don't need to adjust the scale. */
-	if (stat.mean <= iolat->min_lat_nsec &&
+	if (latency_sum_ok(iolat, &stat) &&
 	    atomic_read(&lat_info->scale_cookie) == DEFAULT_SCALE_COOKIE)
 		return;
 
 	/* Somebody beat us to the punch, just bail. */
 	spin_lock_irqsave(&lat_info->lock, flags);
+
+	latency_stat_sum(iolat, &iolat->cur_stat, &stat);
 	lat_info->nr_samples -= iolat->nr_samples;
-	lat_info->nr_samples += stat.nr_samples;
-	iolat->nr_samples = stat.nr_samples;
+	lat_info->nr_samples += latency_stat_samples(iolat, &iolat->cur_stat);
+	iolat->nr_samples = latency_stat_samples(iolat, &iolat->cur_stat);
 
 	if ((lat_info->last_scale_event >= now ||
-	    now - lat_info->last_scale_event < BLKIOLATENCY_MIN_ADJUST_TIME) &&
-	    lat_info->scale_lat <= iolat->min_lat_nsec)
+	    now - lat_info->last_scale_event < BLKIOLATENCY_MIN_ADJUST_TIME))
 		goto out;
 
-	if (stat.mean <= iolat->min_lat_nsec &&
-	    stat.nr_samples >= BLKIOLATENCY_MIN_GOOD_SAMPLES) {
+	if (latency_sum_ok(iolat, &iolat->cur_stat) &&
+	    latency_sum_ok(iolat, &stat)) {
+		if (latency_stat_samples(iolat, &iolat->cur_stat) <
+		    BLKIOLATENCY_MIN_GOOD_SAMPLES)
+			goto out;
 		if (lat_info->scale_grp == iolat) {
 			lat_info->last_scale_event = now;
 			scale_cookie_change(iolat->blkiolat, lat_info, true);
 		}
-	} else if (stat.mean > iolat->min_lat_nsec) {
+	} else if (lat_info->scale_lat == 0 ||
+		   lat_info->scale_lat >= iolat->min_lat_nsec) {
 		lat_info->last_scale_event = now;
 		if (!lat_info->scale_grp ||
 		    lat_info->scale_lat > iolat->min_lat_nsec) {
@@ -545,6 +600,7 @@ static void iolatency_check_latencies(struct iolatency_grp *iolat, u64 now)
 		}
 		scale_cookie_change(iolat->blkiolat, lat_info, false);
 	}
+	latency_stat_init(iolat, &iolat->cur_stat);
 out:
 	spin_unlock_irqrestore(&lat_info->lock, flags);
 }
@@ -650,7 +706,7 @@ static void blkiolatency_timer_fn(struct timer_list *t)
 		 * We could be exiting, don't access the pd unless we have a
 		 * ref on the blkg.
 		 */
-		if (!blkg_try_get(blkg))
+		if (!blkg_tryget(blkg))
 			continue;
 
 		iolat = blkg_to_lat(blkg);
@@ -761,7 +817,6 @@ static ssize_t iolatency_set_limit(struct kernfs_open_file *of, char *buf,
 {
 	struct blkcg *blkcg = css_to_blkcg(of_css(of));
 	struct blkcg_gq *blkg;
-	struct blk_iolatency *blkiolat;
 	struct blkg_conf_ctx ctx;
 	struct iolatency_grp *iolat;
 	char *p, *tok;
@@ -774,7 +829,6 @@ static ssize_t iolatency_set_limit(struct kernfs_open_file *of, char *buf,
 		return ret;
 
 	iolat = blkg_to_lat(ctx.blkg);
-	blkiolat = iolat->blkiolat;
 	p = ctx.body;
 
 	ret = -EINVAL;
@@ -835,13 +889,43 @@ static int iolatency_print_limit(struct seq_file *sf, void *v)
 	return 0;
 }
 
+static size_t iolatency_ssd_stat(struct iolatency_grp *iolat, char *buf,
+				 size_t size)
+{
+	struct latency_stat stat;
+	int cpu;
+
+	latency_stat_init(iolat, &stat);
+	preempt_disable();
+	for_each_online_cpu(cpu) {
+		struct latency_stat *s;
+		s = per_cpu_ptr(iolat->stats, cpu);
+		latency_stat_sum(iolat, &stat, s);
+	}
+	preempt_enable();
+
+	if (iolat->rq_depth.max_depth == UINT_MAX)
+		return scnprintf(buf, size, " missed=%llu total=%llu depth=max",
+				 (unsigned long long)stat.ps.missed,
+				 (unsigned long long)stat.ps.total);
+	return scnprintf(buf, size, " missed=%llu total=%llu depth=%u",
+			 (unsigned long long)stat.ps.missed,
+			 (unsigned long long)stat.ps.total,
+			 iolat->rq_depth.max_depth);
+}
+
 static size_t iolatency_pd_stat(struct blkg_policy_data *pd, char *buf,
 				size_t size)
 {
 	struct iolatency_grp *iolat = pd_to_lat(pd);
-	unsigned long long avg_lat = div64_u64(iolat->lat_avg, NSEC_PER_USEC);
-	unsigned long long cur_win = div64_u64(iolat->cur_win_nsec, NSEC_PER_MSEC);
+	unsigned long long avg_lat;
+	unsigned long long cur_win;
+
+	if (iolat->ssd)
+		return iolatency_ssd_stat(iolat, buf, size);
 
+	avg_lat = div64_u64(iolat->lat_avg, NSEC_PER_USEC);
+	cur_win = div64_u64(iolat->cur_win_nsec, NSEC_PER_MSEC);
 	if (iolat->rq_depth.max_depth == UINT_MAX)
 		return scnprintf(buf, size, " depth=max avg_lat=%llu win=%llu",
 				 avg_lat, cur_win);
@@ -858,8 +942,8 @@ static struct blkg_policy_data *iolatency_pd_alloc(gfp_t gfp, int node)
 	iolat = kzalloc_node(sizeof(*iolat), gfp, node);
 	if (!iolat)
 		return NULL;
-	iolat->stats = __alloc_percpu_gfp(sizeof(struct blk_rq_stat),
-				       __alignof__(struct blk_rq_stat), gfp);
+	iolat->stats = __alloc_percpu_gfp(sizeof(struct latency_stat),
+				       __alignof__(struct latency_stat), gfp);
 	if (!iolat->stats) {
 		kfree(iolat);
 		return NULL;
@@ -876,15 +960,21 @@ static void iolatency_pd_init(struct blkg_policy_data *pd)
 	u64 now = ktime_to_ns(ktime_get());
 	int cpu;
 
+	if (blk_queue_nonrot(blkg->q))
+		iolat->ssd = true;
+	else
+		iolat->ssd = false;
+
 	for_each_possible_cpu(cpu) {
-		struct blk_rq_stat *stat;
+		struct latency_stat *stat;
 		stat = per_cpu_ptr(iolat->stats, cpu);
-		blk_rq_stat_init(stat);
+		latency_stat_init(iolat, stat);
 	}
 
+	latency_stat_init(iolat, &iolat->cur_stat);
 	rq_wait_init(&iolat->rq_wait);
 	spin_lock_init(&iolat->child_lat.lock);
-	iolat->rq_depth.queue_depth = blk_queue_depth(blkg->q);
+	iolat->rq_depth.queue_depth = blkg->q->nr_requests;
 	iolat->rq_depth.max_depth = UINT_MAX;
 	iolat->rq_depth.default_depth = iolat->rq_depth.queue_depth;
 	iolat->blkiolat = blkiolat;
diff --git a/block/blk-merge.c b/block/blk-merge.c
index aaec38cc37b8..42a46744c11b 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -12,6 +12,69 @@
 
 #include "blk.h"
 
+/*
+ * Check if the two bvecs from two bios can be merged to one segment.  If yes,
+ * no need to check gap between the two bios since the 1st bio and the 1st bvec
+ * in the 2nd bio can be handled in one segment.
+ */
+static inline bool bios_segs_mergeable(struct request_queue *q,
+		struct bio *prev, struct bio_vec *prev_last_bv,
+		struct bio_vec *next_first_bv)
+{
+	if (!biovec_phys_mergeable(q, prev_last_bv, next_first_bv))
+		return false;
+	if (prev->bi_seg_back_size + next_first_bv->bv_len >
+			queue_max_segment_size(q))
+		return false;
+	return true;
+}
+
+static inline bool bio_will_gap(struct request_queue *q,
+		struct request *prev_rq, struct bio *prev, struct bio *next)
+{
+	struct bio_vec pb, nb;
+
+	if (!bio_has_data(prev) || !queue_virt_boundary(q))
+		return false;
+
+	/*
+	 * Don't merge if the 1st bio starts with non-zero offset, otherwise it
+	 * is quite difficult to respect the sg gap limit.  We work hard to
+	 * merge a huge number of small single bios in case of mkfs.
+	 */
+	if (prev_rq)
+		bio_get_first_bvec(prev_rq->bio, &pb);
+	else
+		bio_get_first_bvec(prev, &pb);
+	if (pb.bv_offset)
+		return true;
+
+	/*
+	 * We don't need to worry about the situation that the merged segment
+	 * ends in unaligned virt boundary:
+	 *
+	 * - if 'pb' ends aligned, the merged segment ends aligned
+	 * - if 'pb' ends unaligned, the next bio must include
+	 *   one single bvec of 'nb', otherwise the 'nb' can't
+	 *   merge with 'pb'
+	 */
+	bio_get_last_bvec(prev, &pb);
+	bio_get_first_bvec(next, &nb);
+	if (bios_segs_mergeable(q, prev, &pb, &nb))
+		return false;
+	return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
+}
+
+static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
+{
+	return bio_will_gap(req->q, req, req->biotail, bio);
+}
+
+static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
+{
+	return bio_will_gap(req->q, NULL, bio, req->bio);
+}
+
 static struct bio *blk_bio_discard_split(struct request_queue *q,
 					 struct bio *bio,
 					 struct bio_set *bs,
@@ -134,9 +197,7 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 		if (bvprvp && blk_queue_cluster(q)) {
 			if (seg_size + bv.bv_len > queue_max_segment_size(q))
 				goto new_segment;
-			if (!BIOVEC_PHYS_MERGEABLE(bvprvp, &bv))
-				goto new_segment;
-			if (!BIOVEC_SEG_BOUNDARY(q, bvprvp, &bv))
+			if (!biovec_phys_mergeable(q, bvprvp, &bv))
 				goto new_segment;
 
 			seg_size += bv.bv_len;
@@ -267,9 +328,7 @@ static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
 				if (seg_size + bv.bv_len
 				    > queue_max_segment_size(q))
 					goto new_segment;
-				if (!BIOVEC_PHYS_MERGEABLE(&bvprv, &bv))
-					goto new_segment;
-				if (!BIOVEC_SEG_BOUNDARY(q, &bvprv, &bv))
+				if (!biovec_phys_mergeable(q, &bvprv, &bv))
 					goto new_segment;
 
 				seg_size += bv.bv_len;
@@ -349,17 +408,7 @@ static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
 	bio_get_last_bvec(bio, &end_bv);
 	bio_get_first_bvec(nxt, &nxt_bv);
 
-	if (!BIOVEC_PHYS_MERGEABLE(&end_bv, &nxt_bv))
-		return 0;
-
-	/*
-	 * bio and nxt are contiguous in memory; check if the queue allows
-	 * these two to be merged into one
-	 */
-	if (BIOVEC_SEG_BOUNDARY(q, &end_bv, &nxt_bv))
-		return 1;
-
-	return 0;
+	return biovec_phys_mergeable(q, &end_bv, &nxt_bv);
 }
 
 static inline void
@@ -373,10 +422,7 @@ __blk_segment_map_sg(struct request_queue *q, struct bio_vec *bvec,
 	if (*sg && *cluster) {
 		if ((*sg)->length + nbytes > queue_max_segment_size(q))
 			goto new_segment;
-
-		if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
-			goto new_segment;
-		if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
+		if (!biovec_phys_mergeable(q, bvprv, bvec))
 			goto new_segment;
 
 		(*sg)->length += nbytes;
diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index cb1e6cf7ac48..41b86f50d126 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -102,6 +102,14 @@ static int blk_flags_show(struct seq_file *m, const unsigned long flags,
 	return 0;
 }
 
+static int queue_pm_only_show(void *data, struct seq_file *m)
+{
+	struct request_queue *q = data;
+
+	seq_printf(m, "%d\n", atomic_read(&q->pm_only));
+	return 0;
+}
+
 #define QUEUE_FLAG_NAME(name) [QUEUE_FLAG_##name] = #name
 static const char *const blk_queue_flag_name[] = {
 	QUEUE_FLAG_NAME(QUEUED),
@@ -132,7 +140,6 @@ static const char *const blk_queue_flag_name[] = {
 	QUEUE_FLAG_NAME(REGISTERED),
 	QUEUE_FLAG_NAME(SCSI_PASSTHROUGH),
 	QUEUE_FLAG_NAME(QUIESCED),
-	QUEUE_FLAG_NAME(PREEMPT_ONLY),
 };
 #undef QUEUE_FLAG_NAME
 
@@ -209,6 +216,7 @@ static ssize_t queue_write_hint_store(void *data, const char __user *buf,
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_queue_attrs[] = {
 	{ "poll_stat", 0400, queue_poll_stat_show },
 	{ "requeue_list", 0400, .seq_ops = &queue_requeue_list_seq_ops },
+	{ "pm_only", 0600, queue_pm_only_show, NULL },
 	{ "state", 0600, queue_state_show, queue_state_write },
 	{ "write_hints", 0600, queue_write_hint_show, queue_write_hint_store },
 	{ "zone_wlock", 0400, queue_zone_wlock_show, NULL },
@@ -423,8 +431,7 @@ static void hctx_show_busy_rq(struct request *rq, void *data, bool reserved)
 {
 	const struct show_busy_params *params = data;
 
-	if (blk_mq_map_queue(rq->q, rq->mq_ctx->cpu) == params->hctx &&
-	    blk_mq_rq_state(rq) != MQ_RQ_IDLE)
+	if (blk_mq_map_queue(rq->q, rq->mq_ctx->cpu) == params->hctx)
 		__blk_mq_debugfs_rq_show(params->m,
 					 list_entry_rq(&rq->queuelist));
 }
diff --git a/block/blk-mq-sched.h b/block/blk-mq-sched.h
index 4e028ee42430..8a9544203173 100644
--- a/block/blk-mq-sched.h
+++ b/block/blk-mq-sched.h
@@ -49,12 +49,12 @@ blk_mq_sched_allow_merge(struct request_queue *q, struct request *rq,
 	return true;
 }
 
-static inline void blk_mq_sched_completed_request(struct request *rq)
+static inline void blk_mq_sched_completed_request(struct request *rq, u64 now)
 {
 	struct elevator_queue *e = rq->q->elevator;
 
 	if (e && e->type->ops.mq.completed_request)
-		e->type->ops.mq.completed_request(rq);
+		e->type->ops.mq.completed_request(rq, now);
 }
 
 static inline void blk_mq_sched_started_request(struct request *rq)
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 41317c50a446..cfda95b85d34 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -232,13 +232,26 @@ static bool bt_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
 
 	/*
 	 * We can hit rq == NULL here, because the tagging functions
-	 * test and set the bit before assining ->rqs[].
+	 * test and set the bit before assigning ->rqs[].
 	 */
 	if (rq && rq->q == hctx->queue)
 		iter_data->fn(hctx, rq, iter_data->data, reserved);
 	return true;
 }
 
+/**
+ * bt_for_each - iterate over the requests associated with a hardware queue
+ * @hctx:	Hardware queue to examine.
+ * @bt:		sbitmap to examine. This is either the breserved_tags member
+ *		or the bitmap_tags member of struct blk_mq_tags.
+ * @fn:		Pointer to the function that will be called for each request
+ *		associated with @hctx that has been assigned a driver tag.
+ *		@fn will be called as follows: @fn(@hctx, rq, @data, @reserved)
+ *		where rq is a pointer to a request.
+ * @data:	Will be passed as third argument to @fn.
+ * @reserved:	Indicates whether @bt is the breserved_tags member or the
+ *		bitmap_tags member of struct blk_mq_tags.
+ */
 static void bt_for_each(struct blk_mq_hw_ctx *hctx, struct sbitmap_queue *bt,
 			busy_iter_fn *fn, void *data, bool reserved)
 {
@@ -280,6 +293,18 @@ static bool bt_tags_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
 	return true;
 }
 
+/**
+ * bt_tags_for_each - iterate over the requests in a tag map
+ * @tags:	Tag map to iterate over.
+ * @bt:		sbitmap to examine. This is either the breserved_tags member
+ *		or the bitmap_tags member of struct blk_mq_tags.
+ * @fn:		Pointer to the function that will be called for each started
+ *		request. @fn will be called as follows: @fn(rq, @data,
+ *		@reserved) where rq is a pointer to a request.
+ * @data:	Will be passed as second argument to @fn.
+ * @reserved:	Indicates whether @bt is the breserved_tags member or the
+ *		bitmap_tags member of struct blk_mq_tags.
+ */
 static void bt_tags_for_each(struct blk_mq_tags *tags, struct sbitmap_queue *bt,
 			     busy_tag_iter_fn *fn, void *data, bool reserved)
 {
@@ -294,6 +319,15 @@ static void bt_tags_for_each(struct blk_mq_tags *tags, struct sbitmap_queue *bt,
 		sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
 }
 
+/**
+ * blk_mq_all_tag_busy_iter - iterate over all started requests in a tag map
+ * @tags:	Tag map to iterate over.
+ * @fn:		Pointer to the function that will be called for each started
+ *		request. @fn will be called as follows: @fn(rq, @priv,
+ *		reserved) where rq is a pointer to a request. 'reserved'
+ *		indicates whether or not @rq is a reserved request.
+ * @priv:	Will be passed as second argument to @fn.
+ */
 static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
 		busy_tag_iter_fn *fn, void *priv)
 {
@@ -302,6 +336,15 @@ static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
 	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
 }
 
+/**
+ * blk_mq_tagset_busy_iter - iterate over all started requests in a tag set
+ * @tagset:	Tag set to iterate over.
+ * @fn:		Pointer to the function that will be called for each started
+ *		request. @fn will be called as follows: @fn(rq, @priv,
+ *		reserved) where rq is a pointer to a request. 'reserved'
+ *		indicates whether or not @rq is a reserved request.
+ * @priv:	Will be passed as second argument to @fn.
+ */
 void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
 		busy_tag_iter_fn *fn, void *priv)
 {
@@ -314,6 +357,20 @@ void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
 }
 EXPORT_SYMBOL(blk_mq_tagset_busy_iter);
 
+/**
+ * blk_mq_queue_tag_busy_iter - iterate over all requests with a driver tag
+ * @q:		Request queue to examine.
+ * @fn:		Pointer to the function that will be called for each request
+ *		on @q. @fn will be called as follows: @fn(hctx, rq, @priv,
+ *		reserved) where rq is a pointer to a request and hctx points
+ *		to the hardware queue associated with the request. 'reserved'
+ *		indicates whether or not @rq is a reserved request.
+ * @priv:	Will be passed as third argument to @fn.
+ *
+ * Note: if @q->tag_set is shared with other request queues then @fn will be
+ * called for all requests on all queues that share that tag set and not only
+ * for requests associated with @q.
+ */
 void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 		void *priv)
 {
@@ -321,9 +378,11 @@ void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 	int i;
 
 	/*
-	 * __blk_mq_update_nr_hw_queues will update the nr_hw_queues and
-	 * queue_hw_ctx after freeze the queue, so we use q_usage_counter
-	 * to avoid race with it.
+	 * __blk_mq_update_nr_hw_queues() updates nr_hw_queues and queue_hw_ctx
+	 * while the queue is frozen. So we can use q_usage_counter to avoid
+	 * racing with it. __blk_mq_update_nr_hw_queues() uses
+	 * synchronize_rcu() to ensure this function left the critical section
+	 * below.
 	 */
 	if (!percpu_ref_tryget(&q->q_usage_counter))
 		return;
@@ -332,7 +391,7 @@ void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 		struct blk_mq_tags *tags = hctx->tags;
 
 		/*
-		 * If not software queues are currently mapped to this
+		 * If no software queues are currently mapped to this
 		 * hardware queue, there's nothing to check
 		 */
 		if (!blk_mq_hw_queue_mapped(hctx))
diff --git a/block/blk-mq.c b/block/blk-mq.c
index e3c39ea8e17b..dcf10e39995a 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -33,6 +33,7 @@
 #include "blk-mq.h"
 #include "blk-mq-debugfs.h"
 #include "blk-mq-tag.h"
+#include "blk-pm.h"
 #include "blk-stat.h"
 #include "blk-mq-sched.h"
 #include "blk-rq-qos.h"
@@ -198,7 +199,7 @@ void blk_mq_unfreeze_queue(struct request_queue *q)
 	freeze_depth = atomic_dec_return(&q->mq_freeze_depth);
 	WARN_ON_ONCE(freeze_depth < 0);
 	if (!freeze_depth) {
-		percpu_ref_reinit(&q->q_usage_counter);
+		percpu_ref_resurrect(&q->q_usage_counter);
 		wake_up_all(&q->mq_freeze_wq);
 	}
 }
@@ -475,6 +476,7 @@ static void __blk_mq_free_request(struct request *rq)
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 	const int sched_tag = rq->internal_tag;
 
+	blk_pm_mark_last_busy(rq);
 	if (rq->tag != -1)
 		blk_mq_put_tag(hctx, hctx->tags, ctx, rq->tag);
 	if (sched_tag != -1)
@@ -526,6 +528,9 @@ inline void __blk_mq_end_request(struct request *rq, blk_status_t error)
 		blk_stat_add(rq, now);
 	}
 
+	if (rq->internal_tag != -1)
+		blk_mq_sched_completed_request(rq, now);
+
 	blk_account_io_done(rq, now);
 
 	if (rq->end_io) {
@@ -562,8 +567,20 @@ static void __blk_mq_complete_request(struct request *rq)
 
 	if (!blk_mq_mark_complete(rq))
 		return;
-	if (rq->internal_tag != -1)
-		blk_mq_sched_completed_request(rq);
+
+	/*
+	 * Most of single queue controllers, there is only one irq vector
+	 * for handling IO completion, and the only irq's affinity is set
+	 * as all possible CPUs. On most of ARCHs, this affinity means the
+	 * irq is handled on one specific CPU.
+	 *
+	 * So complete IO reqeust in softirq context in case of single queue
+	 * for not degrading IO performance by irqsoff latency.
+	 */
+	if (rq->q->nr_hw_queues == 1) {
+		__blk_complete_request(rq);
+		return;
+	}
 
 	if (!test_bit(QUEUE_FLAG_SAME_COMP, &rq->q->queue_flags)) {
 		rq->q->softirq_done_fn(rq);
@@ -2137,8 +2154,6 @@ static void blk_mq_exit_hctx(struct request_queue *q,
 		struct blk_mq_tag_set *set,
 		struct blk_mq_hw_ctx *hctx, unsigned int hctx_idx)
 {
-	blk_mq_debugfs_unregister_hctx(hctx);
-
 	if (blk_mq_hw_queue_mapped(hctx))
 		blk_mq_tag_idle(hctx);
 
@@ -2165,6 +2180,7 @@ static void blk_mq_exit_hw_queues(struct request_queue *q,
 	queue_for_each_hw_ctx(q, hctx, i) {
 		if (i == nr_queue)
 			break;
+		blk_mq_debugfs_unregister_hctx(hctx);
 		blk_mq_exit_hctx(q, set, hctx, i);
 	}
 }
@@ -2194,12 +2210,12 @@ static int blk_mq_init_hctx(struct request_queue *q,
 	 * runtime
 	 */
 	hctx->ctxs = kmalloc_array_node(nr_cpu_ids, sizeof(void *),
-					GFP_KERNEL, node);
+			GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY, node);
 	if (!hctx->ctxs)
 		goto unregister_cpu_notifier;
 
-	if (sbitmap_init_node(&hctx->ctx_map, nr_cpu_ids, ilog2(8), GFP_KERNEL,
-			      node))
+	if (sbitmap_init_node(&hctx->ctx_map, nr_cpu_ids, ilog2(8),
+				GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY, node))
 		goto free_ctxs;
 
 	hctx->nr_ctx = 0;
@@ -2212,7 +2228,8 @@ static int blk_mq_init_hctx(struct request_queue *q,
 	    set->ops->init_hctx(hctx, set->driver_data, hctx_idx))
 		goto free_bitmap;
 
-	hctx->fq = blk_alloc_flush_queue(q, hctx->numa_node, set->cmd_size);
+	hctx->fq = blk_alloc_flush_queue(q, hctx->numa_node, set->cmd_size,
+			GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY);
 	if (!hctx->fq)
 		goto exit_hctx;
 
@@ -2222,8 +2239,6 @@ static int blk_mq_init_hctx(struct request_queue *q,
 	if (hctx->flags & BLK_MQ_F_BLOCKING)
 		init_srcu_struct(hctx->srcu);
 
-	blk_mq_debugfs_register_hctx(q, hctx);
-
 	return 0;
 
  free_fq:
@@ -2492,6 +2507,39 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 }
 EXPORT_SYMBOL(blk_mq_init_queue);
 
+/*
+ * Helper for setting up a queue with mq ops, given queue depth, and
+ * the passed in mq ops flags.
+ */
+struct request_queue *blk_mq_init_sq_queue(struct blk_mq_tag_set *set,
+					   const struct blk_mq_ops *ops,
+					   unsigned int queue_depth,
+					   unsigned int set_flags)
+{
+	struct request_queue *q;
+	int ret;
+
+	memset(set, 0, sizeof(*set));
+	set->ops = ops;
+	set->nr_hw_queues = 1;
+	set->queue_depth = queue_depth;
+	set->numa_node = NUMA_NO_NODE;
+	set->flags = set_flags;
+
+	ret = blk_mq_alloc_tag_set(set);
+	if (ret)
+		return ERR_PTR(ret);
+
+	q = blk_mq_init_queue(set);
+	if (IS_ERR(q)) {
+		blk_mq_free_tag_set(set);
+		return q;
+	}
+
+	return q;
+}
+EXPORT_SYMBOL(blk_mq_init_sq_queue);
+
 static int blk_mq_hw_ctx_size(struct blk_mq_tag_set *tag_set)
 {
 	int hw_ctx_size = sizeof(struct blk_mq_hw_ctx);
@@ -2506,48 +2554,90 @@ static int blk_mq_hw_ctx_size(struct blk_mq_tag_set *tag_set)
 	return hw_ctx_size;
 }
 
+static struct blk_mq_hw_ctx *blk_mq_alloc_and_init_hctx(
+		struct blk_mq_tag_set *set, struct request_queue *q,
+		int hctx_idx, int node)
+{
+	struct blk_mq_hw_ctx *hctx;
+
+	hctx = kzalloc_node(blk_mq_hw_ctx_size(set),
+			GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY,
+			node);
+	if (!hctx)
+		return NULL;
+
+	if (!zalloc_cpumask_var_node(&hctx->cpumask,
+				GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY,
+				node)) {
+		kfree(hctx);
+		return NULL;
+	}
+
+	atomic_set(&hctx->nr_active, 0);
+	hctx->numa_node = node;
+	hctx->queue_num = hctx_idx;
+
+	if (blk_mq_init_hctx(q, set, hctx, hctx_idx)) {
+		free_cpumask_var(hctx->cpumask);
+		kfree(hctx);
+		return NULL;
+	}
+	blk_mq_hctx_kobj_init(hctx);
+
+	return hctx;
+}
+
 static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
 						struct request_queue *q)
 {
-	int i, j;
+	int i, j, end;
 	struct blk_mq_hw_ctx **hctxs = q->queue_hw_ctx;
 
-	blk_mq_sysfs_unregister(q);
-
 	/* protect against switching io scheduler  */
 	mutex_lock(&q->sysfs_lock);
 	for (i = 0; i < set->nr_hw_queues; i++) {
 		int node;
-
-		if (hctxs[i])
-			continue;
+		struct blk_mq_hw_ctx *hctx;
 
 		node = blk_mq_hw_queue_to_node(q->mq_map, i);
-		hctxs[i] = kzalloc_node(blk_mq_hw_ctx_size(set),
-					GFP_KERNEL, node);
-		if (!hctxs[i])
-			break;
-
-		if (!zalloc_cpumask_var_node(&hctxs[i]->cpumask, GFP_KERNEL,
-						node)) {
-			kfree(hctxs[i]);
-			hctxs[i] = NULL;
-			break;
-		}
-
-		atomic_set(&hctxs[i]->nr_active, 0);
-		hctxs[i]->numa_node = node;
-		hctxs[i]->queue_num = i;
+		/*
+		 * If the hw queue has been mapped to another numa node,
+		 * we need to realloc the hctx. If allocation fails, fallback
+		 * to use the previous one.
+		 */
+		if (hctxs[i] && (hctxs[i]->numa_node == node))
+			continue;
 
-		if (blk_mq_init_hctx(q, set, hctxs[i], i)) {
-			free_cpumask_var(hctxs[i]->cpumask);
-			kfree(hctxs[i]);
-			hctxs[i] = NULL;
-			break;
+		hctx = blk_mq_alloc_and_init_hctx(set, q, i, node);
+		if (hctx) {
+			if (hctxs[i]) {
+				blk_mq_exit_hctx(q, set, hctxs[i], i);
+				kobject_put(&hctxs[i]->kobj);
+			}
+			hctxs[i] = hctx;
+		} else {
+			if (hctxs[i])
+				pr_warn("Allocate new hctx on node %d fails,\
+						fallback to previous one on node %d\n",
+						node, hctxs[i]->numa_node);
+			else
+				break;
 		}
-		blk_mq_hctx_kobj_init(hctxs[i]);
 	}
-	for (j = i; j < q->nr_hw_queues; j++) {
+	/*
+	 * Increasing nr_hw_queues fails. Free the newly allocated
+	 * hctxs and keep the previous q->nr_hw_queues.
+	 */
+	if (i != set->nr_hw_queues) {
+		j = q->nr_hw_queues;
+		end = i;
+	} else {
+		j = i;
+		end = q->nr_hw_queues;
+		q->nr_hw_queues = set->nr_hw_queues;
+	}
+
+	for (; j < end; j++) {
 		struct blk_mq_hw_ctx *hctx = hctxs[j];
 
 		if (hctx) {
@@ -2559,9 +2649,7 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
 
 		}
 	}
-	q->nr_hw_queues = i;
 	mutex_unlock(&q->sysfs_lock);
-	blk_mq_sysfs_register(q);
 }
 
 struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
@@ -2659,25 +2747,6 @@ void blk_mq_free_queue(struct request_queue *q)
 	blk_mq_exit_hw_queues(q, set, set->nr_hw_queues);
 }
 
-/* Basically redo blk_mq_init_queue with queue frozen */
-static void blk_mq_queue_reinit(struct request_queue *q)
-{
-	WARN_ON_ONCE(!atomic_read(&q->mq_freeze_depth));
-
-	blk_mq_debugfs_unregister_hctxs(q);
-	blk_mq_sysfs_unregister(q);
-
-	/*
-	 * redo blk_mq_init_cpu_queues and blk_mq_init_hw_queues. FIXME: maybe
-	 * we should change hctx numa_node according to the new topology (this
-	 * involves freeing and re-allocating memory, worth doing?)
-	 */
-	blk_mq_map_swqueue(q);
-
-	blk_mq_sysfs_register(q);
-	blk_mq_debugfs_register_hctxs(q);
-}
-
 static int __blk_mq_alloc_rq_maps(struct blk_mq_tag_set *set)
 {
 	int i;
@@ -2964,6 +3033,7 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 {
 	struct request_queue *q;
 	LIST_HEAD(head);
+	int prev_nr_hw_queues;
 
 	lockdep_assert_held(&set->tag_list_lock);
 
@@ -2987,11 +3057,30 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 		if (!blk_mq_elv_switch_none(&head, q))
 			goto switch_back;
 
+	list_for_each_entry(q, &set->tag_list, tag_set_list) {
+		blk_mq_debugfs_unregister_hctxs(q);
+		blk_mq_sysfs_unregister(q);
+	}
+
+	prev_nr_hw_queues = set->nr_hw_queues;
 	set->nr_hw_queues = nr_hw_queues;
 	blk_mq_update_queue_map(set);
+fallback:
 	list_for_each_entry(q, &set->tag_list, tag_set_list) {
 		blk_mq_realloc_hw_ctxs(set, q);
-		blk_mq_queue_reinit(q);
+		if (q->nr_hw_queues != set->nr_hw_queues) {
+			pr_warn("Increasing nr_hw_queues to %d fails, fallback to %d\n",
+					nr_hw_queues, prev_nr_hw_queues);
+			set->nr_hw_queues = prev_nr_hw_queues;
+			blk_mq_map_queues(set);
+			goto fallback;
+		}
+		blk_mq_map_swqueue(q);
+	}
+
+	list_for_each_entry(q, &set->tag_list, tag_set_list) {
+		blk_mq_sysfs_register(q);
+		blk_mq_debugfs_register_hctxs(q);
 	}
 
 switch_back:
diff --git a/block/blk-pm.c b/block/blk-pm.c
new file mode 100644
index 000000000000..f8fdae01bea2
--- /dev/null
+++ b/block/blk-pm.c
@@ -0,0 +1,216 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/blk-mq.h>
+#include <linux/blk-pm.h>
+#include <linux/blkdev.h>
+#include <linux/pm_runtime.h>
+#include "blk-mq.h"
+#include "blk-mq-tag.h"
+
+/**
+ * blk_pm_runtime_init - Block layer runtime PM initialization routine
+ * @q: the queue of the device
+ * @dev: the device the queue belongs to
+ *
+ * Description:
+ *    Initialize runtime-PM-related fields for @q and start auto suspend for
+ *    @dev. Drivers that want to take advantage of request-based runtime PM
+ *    should call this function after @dev has been initialized, and its
+ *    request queue @q has been allocated, and runtime PM for it can not happen
+ *    yet(either due to disabled/forbidden or its usage_count > 0). In most
+ *    cases, driver should call this function before any I/O has taken place.
+ *
+ *    This function takes care of setting up using auto suspend for the device,
+ *    the autosuspend delay is set to -1 to make runtime suspend impossible
+ *    until an updated value is either set by user or by driver. Drivers do
+ *    not need to touch other autosuspend settings.
+ *
+ *    The block layer runtime PM is request based, so only works for drivers
+ *    that use request as their IO unit instead of those directly use bio's.
+ */
+void blk_pm_runtime_init(struct request_queue *q, struct device *dev)
+{
+	q->dev = dev;
+	q->rpm_status = RPM_ACTIVE;
+	pm_runtime_set_autosuspend_delay(q->dev, -1);
+	pm_runtime_use_autosuspend(q->dev);
+}
+EXPORT_SYMBOL(blk_pm_runtime_init);
+
+/**
+ * blk_pre_runtime_suspend - Pre runtime suspend check
+ * @q: the queue of the device
+ *
+ * Description:
+ *    This function will check if runtime suspend is allowed for the device
+ *    by examining if there are any requests pending in the queue. If there
+ *    are requests pending, the device can not be runtime suspended; otherwise,
+ *    the queue's status will be updated to SUSPENDING and the driver can
+ *    proceed to suspend the device.
+ *
+ *    For the not allowed case, we mark last busy for the device so that
+ *    runtime PM core will try to autosuspend it some time later.
+ *
+ *    This function should be called near the start of the device's
+ *    runtime_suspend callback.
+ *
+ * Return:
+ *    0		- OK to runtime suspend the device
+ *    -EBUSY	- Device should not be runtime suspended
+ */
+int blk_pre_runtime_suspend(struct request_queue *q)
+{
+	int ret = 0;
+
+	if (!q->dev)
+		return ret;
+
+	WARN_ON_ONCE(q->rpm_status != RPM_ACTIVE);
+
+	/*
+	 * Increase the pm_only counter before checking whether any
+	 * non-PM blk_queue_enter() calls are in progress to avoid that any
+	 * new non-PM blk_queue_enter() calls succeed before the pm_only
+	 * counter is decreased again.
+	 */
+	blk_set_pm_only(q);
+	ret = -EBUSY;
+	/* Switch q_usage_counter from per-cpu to atomic mode. */
+	blk_freeze_queue_start(q);
+	/*
+	 * Wait until atomic mode has been reached. Since that
+	 * involves calling call_rcu(), it is guaranteed that later
+	 * blk_queue_enter() calls see the pm-only state. See also
+	 * http://lwn.net/Articles/573497/.
+	 */
+	percpu_ref_switch_to_atomic_sync(&q->q_usage_counter);
+	if (percpu_ref_is_zero(&q->q_usage_counter))
+		ret = 0;
+	/* Switch q_usage_counter back to per-cpu mode. */
+	blk_mq_unfreeze_queue(q);
+
+	spin_lock_irq(q->queue_lock);
+	if (ret < 0)
+		pm_runtime_mark_last_busy(q->dev);
+	else
+		q->rpm_status = RPM_SUSPENDING;
+	spin_unlock_irq(q->queue_lock);
+
+	if (ret)
+		blk_clear_pm_only(q);
+
+	return ret;
+}
+EXPORT_SYMBOL(blk_pre_runtime_suspend);
+
+/**
+ * blk_post_runtime_suspend - Post runtime suspend processing
+ * @q: the queue of the device
+ * @err: return value of the device's runtime_suspend function
+ *
+ * Description:
+ *    Update the queue's runtime status according to the return value of the
+ *    device's runtime suspend function and mark last busy for the device so
+ *    that PM core will try to auto suspend the device at a later time.
+ *
+ *    This function should be called near the end of the device's
+ *    runtime_suspend callback.
+ */
+void blk_post_runtime_suspend(struct request_queue *q, int err)
+{
+	if (!q->dev)
+		return;
+
+	spin_lock_irq(q->queue_lock);
+	if (!err) {
+		q->rpm_status = RPM_SUSPENDED;
+	} else {
+		q->rpm_status = RPM_ACTIVE;
+		pm_runtime_mark_last_busy(q->dev);
+	}
+	spin_unlock_irq(q->queue_lock);
+
+	if (err)
+		blk_clear_pm_only(q);
+}
+EXPORT_SYMBOL(blk_post_runtime_suspend);
+
+/**
+ * blk_pre_runtime_resume - Pre runtime resume processing
+ * @q: the queue of the device
+ *
+ * Description:
+ *    Update the queue's runtime status to RESUMING in preparation for the
+ *    runtime resume of the device.
+ *
+ *    This function should be called near the start of the device's
+ *    runtime_resume callback.
+ */
+void blk_pre_runtime_resume(struct request_queue *q)
+{
+	if (!q->dev)
+		return;
+
+	spin_lock_irq(q->queue_lock);
+	q->rpm_status = RPM_RESUMING;
+	spin_unlock_irq(q->queue_lock);
+}
+EXPORT_SYMBOL(blk_pre_runtime_resume);
+
+/**
+ * blk_post_runtime_resume - Post runtime resume processing
+ * @q: the queue of the device
+ * @err: return value of the device's runtime_resume function
+ *
+ * Description:
+ *    Update the queue's runtime status according to the return value of the
+ *    device's runtime_resume function. If it is successfully resumed, process
+ *    the requests that are queued into the device's queue when it is resuming
+ *    and then mark last busy and initiate autosuspend for it.
+ *
+ *    This function should be called near the end of the device's
+ *    runtime_resume callback.
+ */
+void blk_post_runtime_resume(struct request_queue *q, int err)
+{
+	if (!q->dev)
+		return;
+
+	spin_lock_irq(q->queue_lock);
+	if (!err) {
+		q->rpm_status = RPM_ACTIVE;
+		pm_runtime_mark_last_busy(q->dev);
+		pm_request_autosuspend(q->dev);
+	} else {
+		q->rpm_status = RPM_SUSPENDED;
+	}
+	spin_unlock_irq(q->queue_lock);
+
+	if (!err)
+		blk_clear_pm_only(q);
+}
+EXPORT_SYMBOL(blk_post_runtime_resume);
+
+/**
+ * blk_set_runtime_active - Force runtime status of the queue to be active
+ * @q: the queue of the device
+ *
+ * If the device is left runtime suspended during system suspend the resume
+ * hook typically resumes the device and corrects runtime status
+ * accordingly. However, that does not affect the queue runtime PM status
+ * which is still "suspended". This prevents processing requests from the
+ * queue.
+ *
+ * This function can be used in driver's resume hook to correct queue
+ * runtime PM status and re-enable peeking requests from the queue. It
+ * should be called before first request is added to the queue.
+ */
+void blk_set_runtime_active(struct request_queue *q)
+{
+	spin_lock_irq(q->queue_lock);
+	q->rpm_status = RPM_ACTIVE;
+	pm_runtime_mark_last_busy(q->dev);
+	pm_request_autosuspend(q->dev);
+	spin_unlock_irq(q->queue_lock);
+}
+EXPORT_SYMBOL(blk_set_runtime_active);
diff --git a/block/blk-pm.h b/block/blk-pm.h
new file mode 100644
index 000000000000..a8564ea72a41
--- /dev/null
+++ b/block/blk-pm.h
@@ -0,0 +1,69 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef _BLOCK_BLK_PM_H_
+#define _BLOCK_BLK_PM_H_
+
+#include <linux/pm_runtime.h>
+
+#ifdef CONFIG_PM
+static inline void blk_pm_request_resume(struct request_queue *q)
+{
+	if (q->dev && (q->rpm_status == RPM_SUSPENDED ||
+		       q->rpm_status == RPM_SUSPENDING))
+		pm_request_resume(q->dev);
+}
+
+static inline void blk_pm_mark_last_busy(struct request *rq)
+{
+	if (rq->q->dev && !(rq->rq_flags & RQF_PM))
+		pm_runtime_mark_last_busy(rq->q->dev);
+}
+
+static inline void blk_pm_requeue_request(struct request *rq)
+{
+	lockdep_assert_held(rq->q->queue_lock);
+
+	if (rq->q->dev && !(rq->rq_flags & RQF_PM))
+		rq->q->nr_pending--;
+}
+
+static inline void blk_pm_add_request(struct request_queue *q,
+				      struct request *rq)
+{
+	lockdep_assert_held(q->queue_lock);
+
+	if (q->dev && !(rq->rq_flags & RQF_PM))
+		q->nr_pending++;
+}
+
+static inline void blk_pm_put_request(struct request *rq)
+{
+	lockdep_assert_held(rq->q->queue_lock);
+
+	if (rq->q->dev && !(rq->rq_flags & RQF_PM))
+		--rq->q->nr_pending;
+}
+#else
+static inline void blk_pm_request_resume(struct request_queue *q)
+{
+}
+
+static inline void blk_pm_mark_last_busy(struct request *rq)
+{
+}
+
+static inline void blk_pm_requeue_request(struct request *rq)
+{
+}
+
+static inline void blk_pm_add_request(struct request_queue *q,
+				      struct request *rq)
+{
+}
+
+static inline void blk_pm_put_request(struct request *rq)
+{
+}
+#endif
+
+#endif /* _BLOCK_BLK_PM_H_ */
diff --git a/block/blk-softirq.c b/block/blk-softirq.c
index 15c1f5e12eb8..e47a2f751884 100644
--- a/block/blk-softirq.c
+++ b/block/blk-softirq.c
@@ -97,8 +97,8 @@ static int blk_softirq_cpu_dead(unsigned int cpu)
 
 void __blk_complete_request(struct request *req)
 {
-	int ccpu, cpu;
 	struct request_queue *q = req->q;
+	int cpu, ccpu = q->mq_ops ? req->mq_ctx->cpu : req->cpu;
 	unsigned long flags;
 	bool shared = false;
 
@@ -110,8 +110,7 @@ void __blk_complete_request(struct request *req)
 	/*
 	 * Select completion CPU
 	 */
-	if (req->cpu != -1) {
-		ccpu = req->cpu;
+	if (test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags) && ccpu != -1) {
 		if (!test_bit(QUEUE_FLAG_SAME_FORCE, &q->queue_flags))
 			shared = cpus_share_cache(cpu, ccpu);
 	} else
diff --git a/block/blk-stat.c b/block/blk-stat.c
index 7587b1c3caaf..90561af85a62 100644
--- a/block/blk-stat.c
+++ b/block/blk-stat.c
@@ -190,6 +190,7 @@ void blk_stat_enable_accounting(struct request_queue *q)
 	blk_queue_flag_set(QUEUE_FLAG_STATS, q);
 	spin_unlock(&q->stats->lock);
 }
+EXPORT_SYMBOL_GPL(blk_stat_enable_accounting);
 
 struct blk_queue_stats *blk_alloc_queue_stats(void)
 {
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 01d0620a4e4a..4bda70e8db48 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -84,8 +84,7 @@ struct throtl_service_queue {
 	 * RB tree of active children throtl_grp's, which are sorted by
 	 * their ->disptime.
 	 */
-	struct rb_root		pending_tree;	/* RB tree of active tgs */
-	struct rb_node		*first_pending;	/* first node in the tree */
+	struct rb_root_cached	pending_tree;	/* RB tree of active tgs */
 	unsigned int		nr_pending;	/* # queued in the tree */
 	unsigned long		first_pending_disptime;	/* disptime of the first tg */
 	struct timer_list	pending_timer;	/* fires on first_pending_disptime */
@@ -475,7 +474,7 @@ static void throtl_service_queue_init(struct throtl_service_queue *sq)
 {
 	INIT_LIST_HEAD(&sq->queued[0]);
 	INIT_LIST_HEAD(&sq->queued[1]);
-	sq->pending_tree = RB_ROOT;
+	sq->pending_tree = RB_ROOT_CACHED;
 	timer_setup(&sq->pending_timer, throtl_pending_timer_fn, 0);
 }
 
@@ -616,31 +615,23 @@ static void throtl_pd_free(struct blkg_policy_data *pd)
 static struct throtl_grp *
 throtl_rb_first(struct throtl_service_queue *parent_sq)
 {
+	struct rb_node *n;
 	/* Service tree is empty */
 	if (!parent_sq->nr_pending)
 		return NULL;
 
-	if (!parent_sq->first_pending)
-		parent_sq->first_pending = rb_first(&parent_sq->pending_tree);
-
-	if (parent_sq->first_pending)
-		return rb_entry_tg(parent_sq->first_pending);
-
-	return NULL;
-}
-
-static void rb_erase_init(struct rb_node *n, struct rb_root *root)
-{
-	rb_erase(n, root);
-	RB_CLEAR_NODE(n);
+	n = rb_first_cached(&parent_sq->pending_tree);
+	WARN_ON_ONCE(!n);
+	if (!n)
+		return NULL;
+	return rb_entry_tg(n);
 }
 
 static void throtl_rb_erase(struct rb_node *n,
 			    struct throtl_service_queue *parent_sq)
 {
-	if (parent_sq->first_pending == n)
-		parent_sq->first_pending = NULL;
-	rb_erase_init(n, &parent_sq->pending_tree);
+	rb_erase_cached(n, &parent_sq->pending_tree);
+	RB_CLEAR_NODE(n);
 	--parent_sq->nr_pending;
 }
 
@@ -658,11 +649,11 @@ static void update_min_dispatch_time(struct throtl_service_queue *parent_sq)
 static void tg_service_queue_add(struct throtl_grp *tg)
 {
 	struct throtl_service_queue *parent_sq = tg->service_queue.parent_sq;
-	struct rb_node **node = &parent_sq->pending_tree.rb_node;
+	struct rb_node **node = &parent_sq->pending_tree.rb_root.rb_node;
 	struct rb_node *parent = NULL;
 	struct throtl_grp *__tg;
 	unsigned long key = tg->disptime;
-	int left = 1;
+	bool leftmost = true;
 
 	while (*node != NULL) {
 		parent = *node;
@@ -672,15 +663,13 @@ static void tg_service_queue_add(struct throtl_grp *tg)
 			node = &parent->rb_left;
 		else {
 			node = &parent->rb_right;
-			left = 0;
+			leftmost = false;
 		}
 	}
 
-	if (left)
-		parent_sq->first_pending = &tg->rb_node;
-
 	rb_link_node(&tg->rb_node, parent, node);
-	rb_insert_color(&tg->rb_node, &parent_sq->pending_tree);
+	rb_insert_color_cached(&tg->rb_node, &parent_sq->pending_tree,
+			       leftmost);
 }
 
 static void __throtl_enqueue_tg(struct throtl_grp *tg)
@@ -2126,21 +2115,11 @@ static inline void throtl_update_latency_buckets(struct throtl_data *td)
 }
 #endif
 
-static void blk_throtl_assoc_bio(struct throtl_grp *tg, struct bio *bio)
-{
-#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
-	/* fallback to root_blkg if we fail to get a blkg ref */
-	if (bio->bi_css && (bio_associate_blkg(bio, tg_to_blkg(tg)) == -ENODEV))
-		bio_associate_blkg(bio, bio->bi_disk->queue->root_blkg);
-	bio_issue_init(&bio->bi_issue, bio_sectors(bio));
-#endif
-}
-
 bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
 		    struct bio *bio)
 {
 	struct throtl_qnode *qn = NULL;
-	struct throtl_grp *tg = blkg_to_tg(blkg ?: q->root_blkg);
+	struct throtl_grp *tg = blkg_to_tg(blkg);
 	struct throtl_service_queue *sq;
 	bool rw = bio_data_dir(bio);
 	bool throttled = false;
@@ -2159,7 +2138,6 @@ bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
 	if (unlikely(blk_queue_bypass(q)))
 		goto out_unlock;
 
-	blk_throtl_assoc_bio(tg, bio);
 	blk_throtl_update_idletime(tg);
 
 	sq = &tg->service_queue;
diff --git a/block/blk.h b/block/blk.h
index 9db4e389582c..3d2aecba96a4 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -4,6 +4,7 @@
 
 #include <linux/idr.h>
 #include <linux/blk-mq.h>
+#include <xen/xen.h>
 #include "blk-mq.h"
 
 /* Amount of time in which a process may batch requests */
@@ -124,7 +125,7 @@ static inline void __blk_get_queue(struct request_queue *q)
 }
 
 struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
-		int node, int cmd_size);
+		int node, int cmd_size, gfp_t flags);
 void blk_free_flush_queue(struct blk_flush_queue *q);
 
 int blk_init_rl(struct request_list *rl, struct request_queue *q,
@@ -149,6 +150,41 @@ static inline void blk_queue_enter_live(struct request_queue *q)
 	percpu_ref_get(&q->q_usage_counter);
 }
 
+static inline bool biovec_phys_mergeable(struct request_queue *q,
+		struct bio_vec *vec1, struct bio_vec *vec2)
+{
+	unsigned long mask = queue_segment_boundary(q);
+	phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
+	phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
+
+	if (addr1 + vec1->bv_len != addr2)
+		return false;
+	if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2))
+		return false;
+	if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
+		return false;
+	return true;
+}
+
+static inline bool __bvec_gap_to_prev(struct request_queue *q,
+		struct bio_vec *bprv, unsigned int offset)
+{
+	return offset ||
+		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
+}
+
+/*
+ * Check if adding a bio_vec after bprv with offset would create a gap in
+ * the SG list. Most drivers don't care about this, but some do.
+ */
+static inline bool bvec_gap_to_prev(struct request_queue *q,
+		struct bio_vec *bprv, unsigned int offset)
+{
+	if (!queue_virt_boundary(q))
+		return false;
+	return __bvec_gap_to_prev(q, bprv, offset);
+}
+
 #ifdef CONFIG_BLK_DEV_INTEGRITY
 void blk_flush_integrity(void);
 bool __bio_integrity_endio(struct bio *);
@@ -158,7 +194,38 @@ static inline bool bio_integrity_endio(struct bio *bio)
 		return __bio_integrity_endio(bio);
 	return true;
 }
-#else
+
+static inline bool integrity_req_gap_back_merge(struct request *req,
+		struct bio *next)
+{
+	struct bio_integrity_payload *bip = bio_integrity(req->bio);
+	struct bio_integrity_payload *bip_next = bio_integrity(next);
+
+	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
+				bip_next->bip_vec[0].bv_offset);
+}
+
+static inline bool integrity_req_gap_front_merge(struct request *req,
+		struct bio *bio)
+{
+	struct bio_integrity_payload *bip = bio_integrity(bio);
+	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
+
+	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
+				bip_next->bip_vec[0].bv_offset);
+}
+#else /* CONFIG_BLK_DEV_INTEGRITY */
+static inline bool integrity_req_gap_back_merge(struct request *req,
+		struct bio *next)
+{
+	return false;
+}
+static inline bool integrity_req_gap_front_merge(struct request *req,
+		struct bio *bio)
+{
+	return false;
+}
+
 static inline void blk_flush_integrity(void)
 {
 }
@@ -166,7 +233,7 @@ static inline bool bio_integrity_endio(struct bio *bio)
 {
 	return true;
 }
-#endif
+#endif /* CONFIG_BLK_DEV_INTEGRITY */
 
 void blk_timeout_work(struct work_struct *work);
 unsigned long blk_rq_timeout(unsigned long timeout);
diff --git a/block/bounce.c b/block/bounce.c
index bc63b3a2d18c..ec0d99995f5f 100644
--- a/block/bounce.c
+++ b/block/bounce.c
@@ -31,6 +31,24 @@
 static struct bio_set bounce_bio_set, bounce_bio_split;
 static mempool_t page_pool, isa_page_pool;
 
+static void init_bounce_bioset(void)
+{
+	static bool bounce_bs_setup;
+	int ret;
+
+	if (bounce_bs_setup)
+		return;
+
+	ret = bioset_init(&bounce_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
+	BUG_ON(ret);
+	if (bioset_integrity_create(&bounce_bio_set, BIO_POOL_SIZE))
+		BUG_ON(1);
+
+	ret = bioset_init(&bounce_bio_split, BIO_POOL_SIZE, 0, 0);
+	BUG_ON(ret);
+	bounce_bs_setup = true;
+}
+
 #if defined(CONFIG_HIGHMEM)
 static __init int init_emergency_pool(void)
 {
@@ -44,14 +62,7 @@ static __init int init_emergency_pool(void)
 	BUG_ON(ret);
 	pr_info("pool size: %d pages\n", POOL_SIZE);
 
-	ret = bioset_init(&bounce_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
-	BUG_ON(ret);
-	if (bioset_integrity_create(&bounce_bio_set, BIO_POOL_SIZE))
-		BUG_ON(1);
-
-	ret = bioset_init(&bounce_bio_split, BIO_POOL_SIZE, 0, 0);
-	BUG_ON(ret);
-
+	init_bounce_bioset();
 	return 0;
 }
 
@@ -86,6 +97,8 @@ static void *mempool_alloc_pages_isa(gfp_t gfp_mask, void *data)
 	return mempool_alloc_pages(gfp_mask | GFP_DMA, data);
 }
 
+static DEFINE_MUTEX(isa_mutex);
+
 /*
  * gets called "every" time someone init's a queue with BLK_BOUNCE_ISA
  * as the max address, so check if the pool has already been created.
@@ -94,14 +107,20 @@ int init_emergency_isa_pool(void)
 {
 	int ret;
 
-	if (mempool_initialized(&isa_page_pool))
+	mutex_lock(&isa_mutex);
+
+	if (mempool_initialized(&isa_page_pool)) {
+		mutex_unlock(&isa_mutex);
 		return 0;
+	}
 
 	ret = mempool_init(&isa_page_pool, ISA_POOL_SIZE, mempool_alloc_pages_isa,
 			   mempool_free_pages, (void *) 0);
 	BUG_ON(ret);
 
 	pr_info("isa pool size: %d pages\n", ISA_POOL_SIZE);
+	init_bounce_bioset();
+	mutex_unlock(&isa_mutex);
 	return 0;
 }
 
@@ -257,7 +276,9 @@ static struct bio *bounce_clone_bio(struct bio *bio_src, gfp_t gfp_mask,
 		}
 	}
 
-	bio_clone_blkcg_association(bio, bio_src);
+	bio_clone_blkg_association(bio, bio_src);
+
+	blkcg_bio_issue_init(bio);
 
 	return bio;
 }
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 2eb87444b157..6a3d87dd3c1a 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -1644,14 +1644,20 @@ static void cfq_pd_offline(struct blkg_policy_data *pd)
 	int i;
 
 	for (i = 0; i < IOPRIO_BE_NR; i++) {
-		if (cfqg->async_cfqq[0][i])
+		if (cfqg->async_cfqq[0][i]) {
 			cfq_put_queue(cfqg->async_cfqq[0][i]);
-		if (cfqg->async_cfqq[1][i])
+			cfqg->async_cfqq[0][i] = NULL;
+		}
+		if (cfqg->async_cfqq[1][i]) {
 			cfq_put_queue(cfqg->async_cfqq[1][i]);
+			cfqg->async_cfqq[1][i] = NULL;
+		}
 	}
 
-	if (cfqg->async_idle_cfqq)
+	if (cfqg->async_idle_cfqq) {
 		cfq_put_queue(cfqg->async_idle_cfqq);
+		cfqg->async_idle_cfqq = NULL;
+	}
 
 	/*
 	 * @blkg is going offline and will be ignored by
@@ -3753,7 +3759,7 @@ static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
 	uint64_t serial_nr;
 
 	rcu_read_lock();
-	serial_nr = bio_blkcg(bio)->css.serial_nr;
+	serial_nr = __bio_blkcg(bio)->css.serial_nr;
 	rcu_read_unlock();
 
 	/*
@@ -3818,7 +3824,7 @@ cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
 	struct cfq_group *cfqg;
 
 	rcu_read_lock();
-	cfqg = cfq_lookup_cfqg(cfqd, bio_blkcg(bio));
+	cfqg = cfq_lookup_cfqg(cfqd, __bio_blkcg(bio));
 	if (!cfqg) {
 		cfqq = &cfqd->oom_cfqq;
 		goto out;
diff --git a/block/elevator.c b/block/elevator.c
index fae58b2f906f..8fdcd64ae12e 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -41,6 +41,7 @@
 
 #include "blk.h"
 #include "blk-mq-sched.h"
+#include "blk-pm.h"
 #include "blk-wbt.h"
 
 static DEFINE_SPINLOCK(elv_list_lock);
@@ -557,27 +558,6 @@ void elv_bio_merged(struct request_queue *q, struct request *rq,
 		e->type->ops.sq.elevator_bio_merged_fn(q, rq, bio);
 }
 
-#ifdef CONFIG_PM
-static void blk_pm_requeue_request(struct request *rq)
-{
-	if (rq->q->dev && !(rq->rq_flags & RQF_PM))
-		rq->q->nr_pending--;
-}
-
-static void blk_pm_add_request(struct request_queue *q, struct request *rq)
-{
-	if (q->dev && !(rq->rq_flags & RQF_PM) && q->nr_pending++ == 0 &&
-	    (q->rpm_status == RPM_SUSPENDED || q->rpm_status == RPM_SUSPENDING))
-		pm_request_resume(q->dev);
-}
-#else
-static inline void blk_pm_requeue_request(struct request *rq) {}
-static inline void blk_pm_add_request(struct request_queue *q,
-				      struct request *rq)
-{
-}
-#endif
-
 void elv_requeue_request(struct request_queue *q, struct request *rq)
 {
 	/*
diff --git a/block/genhd.c b/block/genhd.c
index be5bab20b2ab..cff6bdf27226 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -567,7 +567,8 @@ static int exact_lock(dev_t devt, void *data)
 	return 0;
 }
 
-static void register_disk(struct device *parent, struct gendisk *disk)
+static void register_disk(struct device *parent, struct gendisk *disk,
+			  const struct attribute_group **groups)
 {
 	struct device *ddev = disk_to_dev(disk);
 	struct block_device *bdev;
@@ -582,6 +583,10 @@ static void register_disk(struct device *parent, struct gendisk *disk)
 	/* delay uevents, until we scanned partition table */
 	dev_set_uevent_suppress(ddev, 1);
 
+	if (groups) {
+		WARN_ON(ddev->groups);
+		ddev->groups = groups;
+	}
 	if (device_add(ddev))
 		return;
 	if (!sysfs_deprecated) {
@@ -647,6 +652,7 @@ exit:
  * __device_add_disk - add disk information to kernel list
  * @parent: parent device for the disk
  * @disk: per-device partitioning information
+ * @groups: Additional per-device sysfs groups
  * @register_queue: register the queue if set to true
  *
  * This function registers the partitioning information in @disk
@@ -655,6 +661,7 @@ exit:
  * FIXME: error handling
  */
 static void __device_add_disk(struct device *parent, struct gendisk *disk,
+			      const struct attribute_group **groups,
 			      bool register_queue)
 {
 	dev_t devt;
@@ -698,7 +705,7 @@ static void __device_add_disk(struct device *parent, struct gendisk *disk,
 		blk_register_region(disk_devt(disk), disk->minors, NULL,
 				    exact_match, exact_lock, disk);
 	}
-	register_disk(parent, disk);
+	register_disk(parent, disk, groups);
 	if (register_queue)
 		blk_register_queue(disk);
 
@@ -712,15 +719,17 @@ static void __device_add_disk(struct device *parent, struct gendisk *disk,
 	blk_integrity_add(disk);
 }
 
-void device_add_disk(struct device *parent, struct gendisk *disk)
+void device_add_disk(struct device *parent, struct gendisk *disk,
+		     const struct attribute_group **groups)
+
 {
-	__device_add_disk(parent, disk, true);
+	__device_add_disk(parent, disk, groups, true);
 }
 EXPORT_SYMBOL(device_add_disk);
 
 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
 {
-	__device_add_disk(parent, disk, false);
+	__device_add_disk(parent, disk, NULL, false);
 }
 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
 
diff --git a/block/kyber-iosched.c b/block/kyber-iosched.c
index a1660bafc912..eccac01a10b6 100644
--- a/block/kyber-iosched.c
+++ b/block/kyber-iosched.c
@@ -29,19 +29,30 @@
 #include "blk-mq-debugfs.h"
 #include "blk-mq-sched.h"
 #include "blk-mq-tag.h"
-#include "blk-stat.h"
 
-/* Scheduling domains. */
+#define CREATE_TRACE_POINTS
+#include <trace/events/kyber.h>
+
+/*
+ * Scheduling domains: the device is divided into multiple domains based on the
+ * request type.
+ */
 enum {
 	KYBER_READ,
-	KYBER_SYNC_WRITE,
-	KYBER_OTHER, /* Async writes, discard, etc. */
+	KYBER_WRITE,
+	KYBER_DISCARD,
+	KYBER_OTHER,
 	KYBER_NUM_DOMAINS,
 };
 
-enum {
-	KYBER_MIN_DEPTH = 256,
+static const char *kyber_domain_names[] = {
+	[KYBER_READ] = "READ",
+	[KYBER_WRITE] = "WRITE",
+	[KYBER_DISCARD] = "DISCARD",
+	[KYBER_OTHER] = "OTHER",
+};
 
+enum {
 	/*
 	 * In order to prevent starvation of synchronous requests by a flood of
 	 * asynchronous requests, we reserve 25% of requests for synchronous
@@ -51,25 +62,87 @@ enum {
 };
 
 /*
- * Initial device-wide depths for each scheduling domain.
+ * Maximum device-wide depth for each scheduling domain.
  *
- * Even for fast devices with lots of tags like NVMe, you can saturate
- * the device with only a fraction of the maximum possible queue depth.
- * So, we cap these to a reasonable value.
+ * Even for fast devices with lots of tags like NVMe, you can saturate the
+ * device with only a fraction of the maximum possible queue depth. So, we cap
+ * these to a reasonable value.
  */
 static const unsigned int kyber_depth[] = {
 	[KYBER_READ] = 256,
-	[KYBER_SYNC_WRITE] = 128,
-	[KYBER_OTHER] = 64,
+	[KYBER_WRITE] = 128,
+	[KYBER_DISCARD] = 64,
+	[KYBER_OTHER] = 16,
 };
 
 /*
- * Scheduling domain batch sizes. We favor reads.
+ * Default latency targets for each scheduling domain.
+ */
+static const u64 kyber_latency_targets[] = {
+	[KYBER_READ] = 2ULL * NSEC_PER_MSEC,
+	[KYBER_WRITE] = 10ULL * NSEC_PER_MSEC,
+	[KYBER_DISCARD] = 5ULL * NSEC_PER_SEC,
+};
+
+/*
+ * Batch size (number of requests we'll dispatch in a row) for each scheduling
+ * domain.
  */
 static const unsigned int kyber_batch_size[] = {
 	[KYBER_READ] = 16,
-	[KYBER_SYNC_WRITE] = 8,
-	[KYBER_OTHER] = 8,
+	[KYBER_WRITE] = 8,
+	[KYBER_DISCARD] = 1,
+	[KYBER_OTHER] = 1,
+};
+
+/*
+ * Requests latencies are recorded in a histogram with buckets defined relative
+ * to the target latency:
+ *
+ * <= 1/4 * target latency
+ * <= 1/2 * target latency
+ * <= 3/4 * target latency
+ * <= target latency
+ * <= 1 1/4 * target latency
+ * <= 1 1/2 * target latency
+ * <= 1 3/4 * target latency
+ * > 1 3/4 * target latency
+ */
+enum {
+	/*
+	 * The width of the latency histogram buckets is
+	 * 1 / (1 << KYBER_LATENCY_SHIFT) * target latency.
+	 */
+	KYBER_LATENCY_SHIFT = 2,
+	/*
+	 * The first (1 << KYBER_LATENCY_SHIFT) buckets are <= target latency,
+	 * thus, "good".
+	 */
+	KYBER_GOOD_BUCKETS = 1 << KYBER_LATENCY_SHIFT,
+	/* There are also (1 << KYBER_LATENCY_SHIFT) "bad" buckets. */
+	KYBER_LATENCY_BUCKETS = 2 << KYBER_LATENCY_SHIFT,
+};
+
+/*
+ * We measure both the total latency and the I/O latency (i.e., latency after
+ * submitting to the device).
+ */
+enum {
+	KYBER_TOTAL_LATENCY,
+	KYBER_IO_LATENCY,
+};
+
+static const char *kyber_latency_type_names[] = {
+	[KYBER_TOTAL_LATENCY] = "total",
+	[KYBER_IO_LATENCY] = "I/O",
+};
+
+/*
+ * Per-cpu latency histograms: total latency and I/O latency for each scheduling
+ * domain except for KYBER_OTHER.
+ */
+struct kyber_cpu_latency {
+	atomic_t buckets[KYBER_OTHER][2][KYBER_LATENCY_BUCKETS];
 };
 
 /*
@@ -88,12 +161,9 @@ struct kyber_ctx_queue {
 struct kyber_queue_data {
 	struct request_queue *q;
 
-	struct blk_stat_callback *cb;
-
 	/*
-	 * The device is divided into multiple scheduling domains based on the
-	 * request type. Each domain has a fixed number of in-flight requests of
-	 * that type device-wide, limited by these tokens.
+	 * Each scheduling domain has a limited number of in-flight requests
+	 * device-wide, limited by these tokens.
 	 */
 	struct sbitmap_queue domain_tokens[KYBER_NUM_DOMAINS];
 
@@ -103,8 +173,19 @@ struct kyber_queue_data {
 	 */
 	unsigned int async_depth;
 
+	struct kyber_cpu_latency __percpu *cpu_latency;
+
+	/* Timer for stats aggregation and adjusting domain tokens. */
+	struct timer_list timer;
+
+	unsigned int latency_buckets[KYBER_OTHER][2][KYBER_LATENCY_BUCKETS];
+
+	unsigned long latency_timeout[KYBER_OTHER];
+
+	int domain_p99[KYBER_OTHER];
+
 	/* Target latencies in nanoseconds. */
-	u64 read_lat_nsec, write_lat_nsec;
+	u64 latency_targets[KYBER_OTHER];
 };
 
 struct kyber_hctx_data {
@@ -124,233 +205,219 @@ static int kyber_domain_wake(wait_queue_entry_t *wait, unsigned mode, int flags,
 
 static unsigned int kyber_sched_domain(unsigned int op)
 {
-	if ((op & REQ_OP_MASK) == REQ_OP_READ)
+	switch (op & REQ_OP_MASK) {
+	case REQ_OP_READ:
 		return KYBER_READ;
-	else if ((op & REQ_OP_MASK) == REQ_OP_WRITE && op_is_sync(op))
-		return KYBER_SYNC_WRITE;
-	else
+	case REQ_OP_WRITE:
+		return KYBER_WRITE;
+	case REQ_OP_DISCARD:
+		return KYBER_DISCARD;
+	default:
 		return KYBER_OTHER;
+	}
 }
 
-enum {
-	NONE = 0,
-	GOOD = 1,
-	GREAT = 2,
-	BAD = -1,
-	AWFUL = -2,
-};
-
-#define IS_GOOD(status) ((status) > 0)
-#define IS_BAD(status) ((status) < 0)
-
-static int kyber_lat_status(struct blk_stat_callback *cb,
-			    unsigned int sched_domain, u64 target)
+static void flush_latency_buckets(struct kyber_queue_data *kqd,
+				  struct kyber_cpu_latency *cpu_latency,
+				  unsigned int sched_domain, unsigned int type)
 {
-	u64 latency;
-
-	if (!cb->stat[sched_domain].nr_samples)
-		return NONE;
+	unsigned int *buckets = kqd->latency_buckets[sched_domain][type];
+	atomic_t *cpu_buckets = cpu_latency->buckets[sched_domain][type];
+	unsigned int bucket;
 
-	latency = cb->stat[sched_domain].mean;
-	if (latency >= 2 * target)
-		return AWFUL;
-	else if (latency > target)
-		return BAD;
-	else if (latency <= target / 2)
-		return GREAT;
-	else /* (latency <= target) */
-		return GOOD;
+	for (bucket = 0; bucket < KYBER_LATENCY_BUCKETS; bucket++)
+		buckets[bucket] += atomic_xchg(&cpu_buckets[bucket], 0);
 }
 
 /*
- * Adjust the read or synchronous write depth given the status of reads and
- * writes. The goal is that the latencies of the two domains are fair (i.e., if
- * one is good, then the other is good).
+ * Calculate the histogram bucket with the given percentile rank, or -1 if there
+ * aren't enough samples yet.
  */
-static void kyber_adjust_rw_depth(struct kyber_queue_data *kqd,
-				  unsigned int sched_domain, int this_status,
-				  int other_status)
+static int calculate_percentile(struct kyber_queue_data *kqd,
+				unsigned int sched_domain, unsigned int type,
+				unsigned int percentile)
 {
-	unsigned int orig_depth, depth;
+	unsigned int *buckets = kqd->latency_buckets[sched_domain][type];
+	unsigned int bucket, samples = 0, percentile_samples;
+
+	for (bucket = 0; bucket < KYBER_LATENCY_BUCKETS; bucket++)
+		samples += buckets[bucket];
+
+	if (!samples)
+		return -1;
 
 	/*
-	 * If this domain had no samples, or reads and writes are both good or
-	 * both bad, don't adjust the depth.
+	 * We do the calculation once we have 500 samples or one second passes
+	 * since the first sample was recorded, whichever comes first.
 	 */
-	if (this_status == NONE ||
-	    (IS_GOOD(this_status) && IS_GOOD(other_status)) ||
-	    (IS_BAD(this_status) && IS_BAD(other_status)))
-		return;
-
-	orig_depth = depth = kqd->domain_tokens[sched_domain].sb.depth;
+	if (!kqd->latency_timeout[sched_domain])
+		kqd->latency_timeout[sched_domain] = max(jiffies + HZ, 1UL);
+	if (samples < 500 &&
+	    time_is_after_jiffies(kqd->latency_timeout[sched_domain])) {
+		return -1;
+	}
+	kqd->latency_timeout[sched_domain] = 0;
 
-	if (other_status == NONE) {
-		depth++;
-	} else {
-		switch (this_status) {
-		case GOOD:
-			if (other_status == AWFUL)
-				depth -= max(depth / 4, 1U);
-			else
-				depth -= max(depth / 8, 1U);
-			break;
-		case GREAT:
-			if (other_status == AWFUL)
-				depth /= 2;
-			else
-				depth -= max(depth / 4, 1U);
+	percentile_samples = DIV_ROUND_UP(samples * percentile, 100);
+	for (bucket = 0; bucket < KYBER_LATENCY_BUCKETS - 1; bucket++) {
+		if (buckets[bucket] >= percentile_samples)
 			break;
-		case BAD:
-			depth++;
-			break;
-		case AWFUL:
-			if (other_status == GREAT)
-				depth += 2;
-			else
-				depth++;
-			break;
-		}
+		percentile_samples -= buckets[bucket];
 	}
+	memset(buckets, 0, sizeof(kqd->latency_buckets[sched_domain][type]));
 
-	depth = clamp(depth, 1U, kyber_depth[sched_domain]);
-	if (depth != orig_depth)
-		sbitmap_queue_resize(&kqd->domain_tokens[sched_domain], depth);
+	trace_kyber_latency(kqd->q, kyber_domain_names[sched_domain],
+			    kyber_latency_type_names[type], percentile,
+			    bucket + 1, 1 << KYBER_LATENCY_SHIFT, samples);
+
+	return bucket;
 }
 
-/*
- * Adjust the depth of other requests given the status of reads and synchronous
- * writes. As long as either domain is doing fine, we don't throttle, but if
- * both domains are doing badly, we throttle heavily.
- */
-static void kyber_adjust_other_depth(struct kyber_queue_data *kqd,
-				     int read_status, int write_status,
-				     bool have_samples)
-{
-	unsigned int orig_depth, depth;
-	int status;
-
-	orig_depth = depth = kqd->domain_tokens[KYBER_OTHER].sb.depth;
-
-	if (read_status == NONE && write_status == NONE) {
-		depth += 2;
-	} else if (have_samples) {
-		if (read_status == NONE)
-			status = write_status;
-		else if (write_status == NONE)
-			status = read_status;
-		else
-			status = max(read_status, write_status);
-		switch (status) {
-		case GREAT:
-			depth += 2;
-			break;
-		case GOOD:
-			depth++;
-			break;
-		case BAD:
-			depth -= max(depth / 4, 1U);
-			break;
-		case AWFUL:
-			depth /= 2;
-			break;
-		}
+static void kyber_resize_domain(struct kyber_queue_data *kqd,
+				unsigned int sched_domain, unsigned int depth)
+{
+	depth = clamp(depth, 1U, kyber_depth[sched_domain]);
+	if (depth != kqd->domain_tokens[sched_domain].sb.depth) {
+		sbitmap_queue_resize(&kqd->domain_tokens[sched_domain], depth);
+		trace_kyber_adjust(kqd->q, kyber_domain_names[sched_domain],
+				   depth);
 	}
-
-	depth = clamp(depth, 1U, kyber_depth[KYBER_OTHER]);
-	if (depth != orig_depth)
-		sbitmap_queue_resize(&kqd->domain_tokens[KYBER_OTHER], depth);
 }
 
-/*
- * Apply heuristics for limiting queue depths based on gathered latency
- * statistics.
- */
-static void kyber_stat_timer_fn(struct blk_stat_callback *cb)
+static void kyber_timer_fn(struct timer_list *t)
 {
-	struct kyber_queue_data *kqd = cb->data;
-	int read_status, write_status;
+	struct kyber_queue_data *kqd = from_timer(kqd, t, timer);
+	unsigned int sched_domain;
+	int cpu;
+	bool bad = false;
+
+	/* Sum all of the per-cpu latency histograms. */
+	for_each_online_cpu(cpu) {
+		struct kyber_cpu_latency *cpu_latency;
+
+		cpu_latency = per_cpu_ptr(kqd->cpu_latency, cpu);
+		for (sched_domain = 0; sched_domain < KYBER_OTHER; sched_domain++) {
+			flush_latency_buckets(kqd, cpu_latency, sched_domain,
+					      KYBER_TOTAL_LATENCY);
+			flush_latency_buckets(kqd, cpu_latency, sched_domain,
+					      KYBER_IO_LATENCY);
+		}
+	}
 
-	read_status = kyber_lat_status(cb, KYBER_READ, kqd->read_lat_nsec);
-	write_status = kyber_lat_status(cb, KYBER_SYNC_WRITE, kqd->write_lat_nsec);
+	/*
+	 * Check if any domains have a high I/O latency, which might indicate
+	 * congestion in the device. Note that we use the p90; we don't want to
+	 * be too sensitive to outliers here.
+	 */
+	for (sched_domain = 0; sched_domain < KYBER_OTHER; sched_domain++) {
+		int p90;
 
-	kyber_adjust_rw_depth(kqd, KYBER_READ, read_status, write_status);
-	kyber_adjust_rw_depth(kqd, KYBER_SYNC_WRITE, write_status, read_status);
-	kyber_adjust_other_depth(kqd, read_status, write_status,
-				 cb->stat[KYBER_OTHER].nr_samples != 0);
+		p90 = calculate_percentile(kqd, sched_domain, KYBER_IO_LATENCY,
+					   90);
+		if (p90 >= KYBER_GOOD_BUCKETS)
+			bad = true;
+	}
 
 	/*
-	 * Continue monitoring latencies if we aren't hitting the targets or
-	 * we're still throttling other requests.
+	 * Adjust the scheduling domain depths. If we determined that there was
+	 * congestion, we throttle all domains with good latencies. Either way,
+	 * we ease up on throttling domains with bad latencies.
 	 */
-	if (!blk_stat_is_active(kqd->cb) &&
-	    ((IS_BAD(read_status) || IS_BAD(write_status) ||
-	      kqd->domain_tokens[KYBER_OTHER].sb.depth < kyber_depth[KYBER_OTHER])))
-		blk_stat_activate_msecs(kqd->cb, 100);
+	for (sched_domain = 0; sched_domain < KYBER_OTHER; sched_domain++) {
+		unsigned int orig_depth, depth;
+		int p99;
+
+		p99 = calculate_percentile(kqd, sched_domain,
+					   KYBER_TOTAL_LATENCY, 99);
+		/*
+		 * This is kind of subtle: different domains will not
+		 * necessarily have enough samples to calculate the latency
+		 * percentiles during the same window, so we have to remember
+		 * the p99 for the next time we observe congestion; once we do,
+		 * we don't want to throttle again until we get more data, so we
+		 * reset it to -1.
+		 */
+		if (bad) {
+			if (p99 < 0)
+				p99 = kqd->domain_p99[sched_domain];
+			kqd->domain_p99[sched_domain] = -1;
+		} else if (p99 >= 0) {
+			kqd->domain_p99[sched_domain] = p99;
+		}
+		if (p99 < 0)
+			continue;
+
+		/*
+		 * If this domain has bad latency, throttle less. Otherwise,
+		 * throttle more iff we determined that there is congestion.
+		 *
+		 * The new depth is scaled linearly with the p99 latency vs the
+		 * latency target. E.g., if the p99 is 3/4 of the target, then
+		 * we throttle down to 3/4 of the current depth, and if the p99
+		 * is 2x the target, then we double the depth.
+		 */
+		if (bad || p99 >= KYBER_GOOD_BUCKETS) {
+			orig_depth = kqd->domain_tokens[sched_domain].sb.depth;
+			depth = (orig_depth * (p99 + 1)) >> KYBER_LATENCY_SHIFT;
+			kyber_resize_domain(kqd, sched_domain, depth);
+		}
+	}
 }
 
-static unsigned int kyber_sched_tags_shift(struct kyber_queue_data *kqd)
+static unsigned int kyber_sched_tags_shift(struct request_queue *q)
 {
 	/*
 	 * All of the hardware queues have the same depth, so we can just grab
 	 * the shift of the first one.
 	 */
-	return kqd->q->queue_hw_ctx[0]->sched_tags->bitmap_tags.sb.shift;
-}
-
-static int kyber_bucket_fn(const struct request *rq)
-{
-	return kyber_sched_domain(rq->cmd_flags);
+	return q->queue_hw_ctx[0]->sched_tags->bitmap_tags.sb.shift;
 }
 
 static struct kyber_queue_data *kyber_queue_data_alloc(struct request_queue *q)
 {
 	struct kyber_queue_data *kqd;
-	unsigned int max_tokens;
 	unsigned int shift;
 	int ret = -ENOMEM;
 	int i;
 
-	kqd = kmalloc_node(sizeof(*kqd), GFP_KERNEL, q->node);
+	kqd = kzalloc_node(sizeof(*kqd), GFP_KERNEL, q->node);
 	if (!kqd)
 		goto err;
+
 	kqd->q = q;
 
-	kqd->cb = blk_stat_alloc_callback(kyber_stat_timer_fn, kyber_bucket_fn,
-					  KYBER_NUM_DOMAINS, kqd);
-	if (!kqd->cb)
+	kqd->cpu_latency = alloc_percpu_gfp(struct kyber_cpu_latency,
+					    GFP_KERNEL | __GFP_ZERO);
+	if (!kqd->cpu_latency)
 		goto err_kqd;
 
-	/*
-	 * The maximum number of tokens for any scheduling domain is at least
-	 * the queue depth of a single hardware queue. If the hardware doesn't
-	 * have many tags, still provide a reasonable number.
-	 */
-	max_tokens = max_t(unsigned int, q->tag_set->queue_depth,
-			   KYBER_MIN_DEPTH);
+	timer_setup(&kqd->timer, kyber_timer_fn, 0);
+
 	for (i = 0; i < KYBER_NUM_DOMAINS; i++) {
 		WARN_ON(!kyber_depth[i]);
 		WARN_ON(!kyber_batch_size[i]);
 		ret = sbitmap_queue_init_node(&kqd->domain_tokens[i],
-					      max_tokens, -1, false, GFP_KERNEL,
-					      q->node);
+					      kyber_depth[i], -1, false,
+					      GFP_KERNEL, q->node);
 		if (ret) {
 			while (--i >= 0)
 				sbitmap_queue_free(&kqd->domain_tokens[i]);
-			goto err_cb;
+			goto err_buckets;
 		}
-		sbitmap_queue_resize(&kqd->domain_tokens[i], kyber_depth[i]);
 	}
 
-	shift = kyber_sched_tags_shift(kqd);
-	kqd->async_depth = (1U << shift) * KYBER_ASYNC_PERCENT / 100U;
+	for (i = 0; i < KYBER_OTHER; i++) {
+		kqd->domain_p99[i] = -1;
+		kqd->latency_targets[i] = kyber_latency_targets[i];
+	}
 
-	kqd->read_lat_nsec = 2000000ULL;
-	kqd->write_lat_nsec = 10000000ULL;
+	shift = kyber_sched_tags_shift(q);
+	kqd->async_depth = (1U << shift) * KYBER_ASYNC_PERCENT / 100U;
 
 	return kqd;
 
-err_cb:
-	blk_stat_free_callback(kqd->cb);
+err_buckets:
+	free_percpu(kqd->cpu_latency);
 err_kqd:
 	kfree(kqd);
 err:
@@ -372,25 +439,24 @@ static int kyber_init_sched(struct request_queue *q, struct elevator_type *e)
 		return PTR_ERR(kqd);
 	}
 
+	blk_stat_enable_accounting(q);
+
 	eq->elevator_data = kqd;
 	q->elevator = eq;
 
-	blk_stat_add_callback(q, kqd->cb);
-
 	return 0;
 }
 
 static void kyber_exit_sched(struct elevator_queue *e)
 {
 	struct kyber_queue_data *kqd = e->elevator_data;
-	struct request_queue *q = kqd->q;
 	int i;
 
-	blk_stat_remove_callback(q, kqd->cb);
+	del_timer_sync(&kqd->timer);
 
 	for (i = 0; i < KYBER_NUM_DOMAINS; i++)
 		sbitmap_queue_free(&kqd->domain_tokens[i]);
-	blk_stat_free_callback(kqd->cb);
+	free_percpu(kqd->cpu_latency);
 	kfree(kqd);
 }
 
@@ -558,41 +624,44 @@ static void kyber_finish_request(struct request *rq)
 	rq_clear_domain_token(kqd, rq);
 }
 
-static void kyber_completed_request(struct request *rq)
+static void add_latency_sample(struct kyber_cpu_latency *cpu_latency,
+			       unsigned int sched_domain, unsigned int type,
+			       u64 target, u64 latency)
 {
-	struct request_queue *q = rq->q;
-	struct kyber_queue_data *kqd = q->elevator->elevator_data;
-	unsigned int sched_domain;
-	u64 now, latency, target;
+	unsigned int bucket;
+	u64 divisor;
 
-	/*
-	 * Check if this request met our latency goal. If not, quickly gather
-	 * some statistics and start throttling.
-	 */
-	sched_domain = kyber_sched_domain(rq->cmd_flags);
-	switch (sched_domain) {
-	case KYBER_READ:
-		target = kqd->read_lat_nsec;
-		break;
-	case KYBER_SYNC_WRITE:
-		target = kqd->write_lat_nsec;
-		break;
-	default:
-		return;
+	if (latency > 0) {
+		divisor = max_t(u64, target >> KYBER_LATENCY_SHIFT, 1);
+		bucket = min_t(unsigned int, div64_u64(latency - 1, divisor),
+			       KYBER_LATENCY_BUCKETS - 1);
+	} else {
+		bucket = 0;
 	}
 
-	/* If we are already monitoring latencies, don't check again. */
-	if (blk_stat_is_active(kqd->cb))
-		return;
+	atomic_inc(&cpu_latency->buckets[sched_domain][type][bucket]);
+}
 
-	now = ktime_get_ns();
-	if (now < rq->io_start_time_ns)
+static void kyber_completed_request(struct request *rq, u64 now)
+{
+	struct kyber_queue_data *kqd = rq->q->elevator->elevator_data;
+	struct kyber_cpu_latency *cpu_latency;
+	unsigned int sched_domain;
+	u64 target;
+
+	sched_domain = kyber_sched_domain(rq->cmd_flags);
+	if (sched_domain == KYBER_OTHER)
 		return;
 
-	latency = now - rq->io_start_time_ns;
+	cpu_latency = get_cpu_ptr(kqd->cpu_latency);
+	target = kqd->latency_targets[sched_domain];
+	add_latency_sample(cpu_latency, sched_domain, KYBER_TOTAL_LATENCY,
+			   target, now - rq->start_time_ns);
+	add_latency_sample(cpu_latency, sched_domain, KYBER_IO_LATENCY, target,
+			   now - rq->io_start_time_ns);
+	put_cpu_ptr(kqd->cpu_latency);
 
-	if (latency > target)
-		blk_stat_activate_msecs(kqd->cb, 10);
+	timer_reduce(&kqd->timer, jiffies + HZ / 10);
 }
 
 struct flush_kcq_data {
@@ -713,6 +782,9 @@ kyber_dispatch_cur_domain(struct kyber_queue_data *kqd,
 			rq_set_domain_token(rq, nr);
 			list_del_init(&rq->queuelist);
 			return rq;
+		} else {
+			trace_kyber_throttled(kqd->q,
+					      kyber_domain_names[khd->cur_domain]);
 		}
 	} else if (sbitmap_any_bit_set(&khd->kcq_map[khd->cur_domain])) {
 		nr = kyber_get_domain_token(kqd, khd, hctx);
@@ -723,6 +795,9 @@ kyber_dispatch_cur_domain(struct kyber_queue_data *kqd,
 			rq_set_domain_token(rq, nr);
 			list_del_init(&rq->queuelist);
 			return rq;
+		} else {
+			trace_kyber_throttled(kqd->q,
+					      kyber_domain_names[khd->cur_domain]);
 		}
 	}
 
@@ -790,17 +865,17 @@ static bool kyber_has_work(struct blk_mq_hw_ctx *hctx)
 	return false;
 }
 
-#define KYBER_LAT_SHOW_STORE(op)					\
-static ssize_t kyber_##op##_lat_show(struct elevator_queue *e,		\
-				     char *page)			\
+#define KYBER_LAT_SHOW_STORE(domain, name)				\
+static ssize_t kyber_##name##_lat_show(struct elevator_queue *e,	\
+				       char *page)			\
 {									\
 	struct kyber_queue_data *kqd = e->elevator_data;		\
 									\
-	return sprintf(page, "%llu\n", kqd->op##_lat_nsec);		\
+	return sprintf(page, "%llu\n", kqd->latency_targets[domain]);	\
 }									\
 									\
-static ssize_t kyber_##op##_lat_store(struct elevator_queue *e,		\
-				      const char *page, size_t count)	\
+static ssize_t kyber_##name##_lat_store(struct elevator_queue *e,	\
+					const char *page, size_t count)	\
 {									\
 	struct kyber_queue_data *kqd = e->elevator_data;		\
 	unsigned long long nsec;					\
@@ -810,12 +885,12 @@ static ssize_t kyber_##op##_lat_store(struct elevator_queue *e,		\
 	if (ret)							\
 		return ret;						\
 									\
-	kqd->op##_lat_nsec = nsec;					\
+	kqd->latency_targets[domain] = nsec;				\
 									\
 	return count;							\
 }
-KYBER_LAT_SHOW_STORE(read);
-KYBER_LAT_SHOW_STORE(write);
+KYBER_LAT_SHOW_STORE(KYBER_READ, read);
+KYBER_LAT_SHOW_STORE(KYBER_WRITE, write);
 #undef KYBER_LAT_SHOW_STORE
 
 #define KYBER_LAT_ATTR(op) __ATTR(op##_lat_nsec, 0644, kyber_##op##_lat_show, kyber_##op##_lat_store)
@@ -882,7 +957,8 @@ static int kyber_##name##_waiting_show(void *data, struct seq_file *m)	\
 	return 0;							\
 }
 KYBER_DEBUGFS_DOMAIN_ATTRS(KYBER_READ, read)
-KYBER_DEBUGFS_DOMAIN_ATTRS(KYBER_SYNC_WRITE, sync_write)
+KYBER_DEBUGFS_DOMAIN_ATTRS(KYBER_WRITE, write)
+KYBER_DEBUGFS_DOMAIN_ATTRS(KYBER_DISCARD, discard)
 KYBER_DEBUGFS_DOMAIN_ATTRS(KYBER_OTHER, other)
 #undef KYBER_DEBUGFS_DOMAIN_ATTRS
 
@@ -900,20 +976,7 @@ static int kyber_cur_domain_show(void *data, struct seq_file *m)
 	struct blk_mq_hw_ctx *hctx = data;
 	struct kyber_hctx_data *khd = hctx->sched_data;
 
-	switch (khd->cur_domain) {
-	case KYBER_READ:
-		seq_puts(m, "READ\n");
-		break;
-	case KYBER_SYNC_WRITE:
-		seq_puts(m, "SYNC_WRITE\n");
-		break;
-	case KYBER_OTHER:
-		seq_puts(m, "OTHER\n");
-		break;
-	default:
-		seq_printf(m, "%u\n", khd->cur_domain);
-		break;
-	}
+	seq_printf(m, "%s\n", kyber_domain_names[khd->cur_domain]);
 	return 0;
 }
 
@@ -930,7 +993,8 @@ static int kyber_batching_show(void *data, struct seq_file *m)
 	{#name "_tokens", 0400, kyber_##name##_tokens_show}
 static const struct blk_mq_debugfs_attr kyber_queue_debugfs_attrs[] = {
 	KYBER_QUEUE_DOMAIN_ATTRS(read),
-	KYBER_QUEUE_DOMAIN_ATTRS(sync_write),
+	KYBER_QUEUE_DOMAIN_ATTRS(write),
+	KYBER_QUEUE_DOMAIN_ATTRS(discard),
 	KYBER_QUEUE_DOMAIN_ATTRS(other),
 	{"async_depth", 0400, kyber_async_depth_show},
 	{},
@@ -942,7 +1006,8 @@ static const struct blk_mq_debugfs_attr kyber_queue_debugfs_attrs[] = {
 	{#name "_waiting", 0400, kyber_##name##_waiting_show}
 static const struct blk_mq_debugfs_attr kyber_hctx_debugfs_attrs[] = {
 	KYBER_HCTX_DOMAIN_ATTRS(read),
-	KYBER_HCTX_DOMAIN_ATTRS(sync_write),
+	KYBER_HCTX_DOMAIN_ATTRS(write),
+	KYBER_HCTX_DOMAIN_ATTRS(discard),
 	KYBER_HCTX_DOMAIN_ATTRS(other),
 	{"cur_domain", 0400, kyber_cur_domain_show},
 	{"batching", 0400, kyber_batching_show},
author	Linus Torvalds	2018-10-22 17:46:08 +0100
committer	Linus Torvalds	2018-10-22 17:46:08 +0100
commit	6ab9e09238fdfd742fe23b81e2d385a1cab49d9b (patch)
tree	38c27b6d70c68f1114df2f49f72f1f757b22cd07 /block
parent	528985117126f11beea339cf39120ee99da04cd2 (diff)
parent	52990a5fb0c991ecafebdab43138b5ed41376852 (diff)