Merge remote-tracking branch 'jens/for-5.10/block' into dm-5.10

DM depends on these block 5.10 commits:

22ada802ede8 block: use lcm_not_zero() when stacking chunk_sectors
07d098e6bbad block: allow 'chunk_sectors' to be non-power-of-2
021a24460dc2 block: add QUEUE_FLAG_NOWAIT
6abc49468eea dm: add support for REQ_NOWAIT and enable it for linear target

Signed-off-by: Mike Snitzer <snitzer@redhat.com>

28 files changed, 2194 insertions, 1250 deletions

diff --git a/block/Kconfig b/block/Kconfig
index bbad5e8bbffe..a2297edfdde8 100644
--- a/block/Kconfig
+++ b/block/Kconfig
@@ -161,8 +161,6 @@ config BLK_WBT_MQ
 	depends on BLK_WBT
 	help
 	Enable writeback throttling by default on multiqueue devices.
-	Multiqueue currently doesn't have support for IO scheduling,
-	enabling this option is recommended.
 
 config BLK_DEBUG_FS
 	bool "Block layer debugging information in debugfs"
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index fa98470df3f0..9e81d1052091 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -4640,6 +4640,9 @@ static bool bfq_has_work(struct blk_mq_hw_ctx *hctx)
 {
 	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
 
+	if (!atomic_read(&hctx->elevator_queued))
+		return false;
+
 	/*
 	 * Avoiding lock: a race on bfqd->busy_queues should cause at
 	 * most a call to dispatch for nothing
@@ -5554,6 +5557,7 @@ static void bfq_insert_requests(struct blk_mq_hw_ctx *hctx,
 		rq = list_first_entry(list, struct request, queuelist);
 		list_del_init(&rq->queuelist);
 		bfq_insert_request(hctx, rq, at_head);
+		atomic_inc(&hctx->elevator_queued);
 	}
 }
 
@@ -5921,6 +5925,7 @@ static void bfq_finish_requeue_request(struct request *rq)
 
 		bfq_completed_request(bfqq, bfqd);
 		bfq_finish_requeue_request_body(bfqq);
+		atomic_dec(&rq->mq_hctx->elevator_queued);
 
 		spin_unlock_irqrestore(&bfqd->lock, flags);
 	} else {
@@ -6360,8 +6365,8 @@ static void bfq_depth_updated(struct blk_mq_hw_ctx *hctx)
 	struct blk_mq_tags *tags = hctx->sched_tags;
 	unsigned int min_shallow;
 
-	min_shallow = bfq_update_depths(bfqd, &tags->bitmap_tags);
-	sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, min_shallow);
+	min_shallow = bfq_update_depths(bfqd, tags->bitmap_tags);
+	sbitmap_queue_min_shallow_depth(tags->bitmap_tags, min_shallow);
 }
 
 static int bfq_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int index)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index c195365c9817..f9b55614d67d 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -119,6 +119,8 @@ static void blkg_async_bio_workfn(struct work_struct *work)
 					     async_bio_work);
 	struct bio_list bios = BIO_EMPTY_LIST;
 	struct bio *bio;
+	struct blk_plug plug;
+	bool need_plug = false;
 
 	/* as long as there are pending bios, @blkg can't go away */
 	spin_lock_bh(&blkg->async_bio_lock);
@@ -126,8 +128,15 @@ static void blkg_async_bio_workfn(struct work_struct *work)
 	bio_list_init(&blkg->async_bios);
 	spin_unlock_bh(&blkg->async_bio_lock);
 
+	/* start plug only when bio_list contains at least 2 bios */
+	if (bios.head && bios.head->bi_next) {
+		need_plug = true;
+		blk_start_plug(&plug);
+	}
 	while ((bio = bio_list_pop(&bios)))
 		submit_bio(bio);
+	if (need_plug)
+		blk_finish_plug(&plug);
 }
 
 /**
@@ -1613,16 +1622,24 @@ static void blkcg_scale_delay(struct blkcg_gq *blkg, u64 now)
 static void blkcg_maybe_throttle_blkg(struct blkcg_gq *blkg, bool use_memdelay)
 {
 	unsigned long pflags;
+	bool clamp;
 	u64 now = ktime_to_ns(ktime_get());
 	u64 exp;
 	u64 delay_nsec = 0;
 	int tok;
 
 	while (blkg->parent) {
-		if (atomic_read(&blkg->use_delay)) {
+		int use_delay = atomic_read(&blkg->use_delay);
+
+		if (use_delay) {
+			u64 this_delay;
+
 			blkcg_scale_delay(blkg, now);
-			delay_nsec = max_t(u64, delay_nsec,
-					   atomic64_read(&blkg->delay_nsec));
+			this_delay = atomic64_read(&blkg->delay_nsec);
+			if (this_delay > delay_nsec) {
+				delay_nsec = this_delay;
+				clamp = use_delay > 0;
+			}
 		}
 		blkg = blkg->parent;
 	}
@@ -1634,10 +1651,13 @@ static void blkcg_maybe_throttle_blkg(struct blkcg_gq *blkg, bool use_memdelay)
 	 * Let's not sleep for all eternity if we've amassed a huge delay.
 	 * Swapping or metadata IO can accumulate 10's of seconds worth of
 	 * delay, and we want userspace to be able to do _something_ so cap the
-	 * delays at 1 second.  If there's 10's of seconds worth of delay then
-	 * the tasks will be delayed for 1 second for every syscall.
+	 * delays at 0.25s. If there's 10's of seconds worth of delay then the
+	 * tasks will be delayed for 0.25 second for every syscall. If
+	 * blkcg_set_delay() was used as indicated by negative use_delay, the
+	 * caller is responsible for regulating the range.
 	 */
-	delay_nsec = min_t(u64, delay_nsec, 250 * NSEC_PER_MSEC);
+	if (clamp)
+		delay_nsec = min_t(u64, delay_nsec, 250 * NSEC_PER_MSEC);
 
 	if (use_memdelay)
 		psi_memstall_enter(&pflags);
diff --git a/block/blk-core.c b/block/blk-core.c
index 10c08ac50697..4884f1e7451b 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -116,8 +116,8 @@ void blk_rq_init(struct request_queue *q, struct request *rq)
 	rq->__sector = (sector_t) -1;
 	INIT_HLIST_NODE(&rq->hash);
 	RB_CLEAR_NODE(&rq->rb_node);
-	rq->tag = -1;
-	rq->internal_tag = -1;
+	rq->tag = BLK_MQ_NO_TAG;
+	rq->internal_tag = BLK_MQ_NO_TAG;
 	rq->start_time_ns = ktime_get_ns();
 	rq->part = NULL;
 	refcount_set(&rq->ref, 1);
@@ -538,11 +538,10 @@ struct request_queue *blk_alloc_queue(int node_id)
 	if (!q->stats)
 		goto fail_stats;
 
-	q->backing_dev_info->ra_pages = VM_READAHEAD_PAGES;
-	q->backing_dev_info->io_pages = VM_READAHEAD_PAGES;
-	q->backing_dev_info->capabilities = BDI_CAP_CGROUP_WRITEBACK;
 	q->node = node_id;
 
+	atomic_set(&q->nr_active_requests_shared_sbitmap, 0);
+
 	timer_setup(&q->backing_dev_info->laptop_mode_wb_timer,
 		    laptop_mode_timer_fn, 0);
 	timer_setup(&q->timeout, blk_rq_timed_out_timer, 0);
@@ -643,162 +642,6 @@ void blk_put_request(struct request *req)
 }
 EXPORT_SYMBOL(blk_put_request);
 
-static void blk_account_io_merge_bio(struct request *req)
-{
-	if (!blk_do_io_stat(req))
-		return;
-
-	part_stat_lock();
-	part_stat_inc(req->part, merges[op_stat_group(req_op(req))]);
-	part_stat_unlock();
-}
-
-bool bio_attempt_back_merge(struct request *req, struct bio *bio,
-		unsigned int nr_segs)
-{
-	const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
-
-	if (!ll_back_merge_fn(req, bio, nr_segs))
-		return false;
-
-	trace_block_bio_backmerge(req->q, req, bio);
-	rq_qos_merge(req->q, req, bio);
-
-	if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
-		blk_rq_set_mixed_merge(req);
-
-	req->biotail->bi_next = bio;
-	req->biotail = bio;
-	req->__data_len += bio->bi_iter.bi_size;
-
-	bio_crypt_free_ctx(bio);
-
-	blk_account_io_merge_bio(req);
-	return true;
-}
-
-bool bio_attempt_front_merge(struct request *req, struct bio *bio,
-		unsigned int nr_segs)
-{
-	const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
-
-	if (!ll_front_merge_fn(req, bio, nr_segs))
-		return false;
-
-	trace_block_bio_frontmerge(req->q, req, bio);
-	rq_qos_merge(req->q, req, bio);
-
-	if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
-		blk_rq_set_mixed_merge(req);
-
-	bio->bi_next = req->bio;
-	req->bio = bio;
-
-	req->__sector = bio->bi_iter.bi_sector;
-	req->__data_len += bio->bi_iter.bi_size;
-
-	bio_crypt_do_front_merge(req, bio);
-
-	blk_account_io_merge_bio(req);
-	return true;
-}
-
-bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
-		struct bio *bio)
-{
-	unsigned short segments = blk_rq_nr_discard_segments(req);
-
-	if (segments >= queue_max_discard_segments(q))
-		goto no_merge;
-	if (blk_rq_sectors(req) + bio_sectors(bio) >
-	    blk_rq_get_max_sectors(req, blk_rq_pos(req)))
-		goto no_merge;
-
-	rq_qos_merge(q, req, bio);
-
-	req->biotail->bi_next = bio;
-	req->biotail = bio;
-	req->__data_len += bio->bi_iter.bi_size;
-	req->nr_phys_segments = segments + 1;
-
-	blk_account_io_merge_bio(req);
-	return true;
-no_merge:
-	req_set_nomerge(q, req);
-	return false;
-}
-
-/**
- * blk_attempt_plug_merge - try to merge with %current's plugged list
- * @q: request_queue new bio is being queued at
- * @bio: new bio being queued
- * @nr_segs: number of segments in @bio
- * @same_queue_rq: pointer to &struct request that gets filled in when
- * another request associated with @q is found on the plug list
- * (optional, may be %NULL)
- *
- * Determine whether @bio being queued on @q can be merged with a request
- * on %current's plugged list.  Returns %true if merge was successful,
- * otherwise %false.
- *
- * Plugging coalesces IOs from the same issuer for the same purpose without
- * going through @q->queue_lock.  As such it's more of an issuing mechanism
- * than scheduling, and the request, while may have elvpriv data, is not
- * added on the elevator at this point.  In addition, we don't have
- * reliable access to the elevator outside queue lock.  Only check basic
- * merging parameters without querying the elevator.
- *
- * Caller must ensure !blk_queue_nomerges(q) beforehand.
- */
-bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
-		unsigned int nr_segs, struct request **same_queue_rq)
-{
-	struct blk_plug *plug;
-	struct request *rq;
-	struct list_head *plug_list;
-
-	plug = blk_mq_plug(q, bio);
-	if (!plug)
-		return false;
-
-	plug_list = &plug->mq_list;
-
-	list_for_each_entry_reverse(rq, plug_list, queuelist) {
-		bool merged = false;
-
-		if (rq->q == q && same_queue_rq) {
-			/*
-			 * Only blk-mq multiple hardware queues case checks the
-			 * rq in the same queue, there should be only one such
-			 * rq in a queue
-			 **/
-			*same_queue_rq = rq;
-		}
-
-		if (rq->q != q || !blk_rq_merge_ok(rq, bio))
-			continue;
-
-		switch (blk_try_merge(rq, bio)) {
-		case ELEVATOR_BACK_MERGE:
-			merged = bio_attempt_back_merge(rq, bio, nr_segs);
-			break;
-		case ELEVATOR_FRONT_MERGE:
-			merged = bio_attempt_front_merge(rq, bio, nr_segs);
-			break;
-		case ELEVATOR_DISCARD_MERGE:
-			merged = bio_attempt_discard_merge(q, rq, bio);
-			break;
-		default:
-			break;
-		}
-
-		if (merged)
-			return true;
-	}
-
-	return false;
-}
-
 static void handle_bad_sector(struct bio *bio, sector_t maxsector)
 {
 	char b[BDEVNAME_SIZE];
@@ -971,9 +814,9 @@ static noinline_for_stack bool submit_bio_checks(struct bio *bio)
 
 	/*
 	 * For a REQ_NOWAIT based request, return -EOPNOTSUPP
-	 * if queue is not a request based queue.
+	 * if queue does not support NOWAIT.
 	 */
-	if ((bio->bi_opf & REQ_NOWAIT) && !queue_is_mq(q))
+	if ((bio->bi_opf & REQ_NOWAIT) && !blk_queue_nowait(q))
 		goto not_supported;
 
 	if (should_fail_bio(bio))
@@ -1301,14 +1144,28 @@ EXPORT_SYMBOL(submit_bio);
  *    limits when retrying requests on other queues. Those requests need
  *    to be checked against the new queue limits again during dispatch.
  */
-static int blk_cloned_rq_check_limits(struct request_queue *q,
+static blk_status_t blk_cloned_rq_check_limits(struct request_queue *q,
 				      struct request *rq)
 {
-	if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) {
+	unsigned int max_sectors = blk_queue_get_max_sectors(q, req_op(rq));
+
+	if (blk_rq_sectors(rq) > max_sectors) {
+		/*
+		 * SCSI device does not have a good way to return if
+		 * Write Same/Zero is actually supported. If a device rejects
+		 * a non-read/write command (discard, write same,etc.) the
+		 * low-level device driver will set the relevant queue limit to
+		 * 0 to prevent blk-lib from issuing more of the offending
+		 * operations. Commands queued prior to the queue limit being
+		 * reset need to be completed with BLK_STS_NOTSUPP to avoid I/O
+		 * errors being propagated to upper layers.
+		 */
+		if (max_sectors == 0)
+			return BLK_STS_NOTSUPP;
+
 		printk(KERN_ERR "%s: over max size limit. (%u > %u)\n",
-			__func__, blk_rq_sectors(rq),
-			blk_queue_get_max_sectors(q, req_op(rq)));
-		return -EIO;
+			__func__, blk_rq_sectors(rq), max_sectors);
+		return BLK_STS_IOERR;
 	}
 
 	/*
@@ -1321,10 +1178,10 @@ static int blk_cloned_rq_check_limits(struct request_queue *q,
 	if (rq->nr_phys_segments > queue_max_segments(q)) {
 		printk(KERN_ERR "%s: over max segments limit. (%hu > %hu)\n",
 			__func__, rq->nr_phys_segments, queue_max_segments(q));
-		return -EIO;
+		return BLK_STS_IOERR;
 	}
 
-	return 0;
+	return BLK_STS_OK;
 }
 
 /**
@@ -1334,8 +1191,11 @@ static int blk_cloned_rq_check_limits(struct request_queue *q,
  */
 blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq)
 {
-	if (blk_cloned_rq_check_limits(q, rq))
-		return BLK_STS_IOERR;
+	blk_status_t ret;
+
+	ret = blk_cloned_rq_check_limits(q, rq);
+	if (ret != BLK_STS_OK)
+		return ret;
 
 	if (rq->rq_disk &&
 	    should_fail_request(&rq->rq_disk->part0, blk_rq_bytes(rq)))
@@ -1461,10 +1321,9 @@ void blk_account_io_start(struct request *rq)
 	part_stat_unlock();
 }
 
-unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
-		unsigned int op)
+static unsigned long __part_start_io_acct(struct hd_struct *part,
+					  unsigned int sectors, unsigned int op)
 {
-	struct hd_struct *part = &disk->part0;
 	const int sgrp = op_stat_group(op);
 	unsigned long now = READ_ONCE(jiffies);
 
@@ -1477,12 +1336,26 @@ unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
 
 	return now;
 }
+
+unsigned long part_start_io_acct(struct gendisk *disk, struct hd_struct **part,
+				 struct bio *bio)
+{
+	*part = disk_map_sector_rcu(disk, bio->bi_iter.bi_sector);
+
+	return __part_start_io_acct(*part, bio_sectors(bio), bio_op(bio));
+}
+EXPORT_SYMBOL_GPL(part_start_io_acct);
+
+unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
+				 unsigned int op)
+{
+	return __part_start_io_acct(&disk->part0, sectors, op);
+}
 EXPORT_SYMBOL(disk_start_io_acct);
 
-void disk_end_io_acct(struct gendisk *disk, unsigned int op,
-		unsigned long start_time)
+static void __part_end_io_acct(struct hd_struct *part, unsigned int op,
+			       unsigned long start_time)
 {
-	struct hd_struct *part = &disk->part0;
 	const int sgrp = op_stat_group(op);
 	unsigned long now = READ_ONCE(jiffies);
 	unsigned long duration = now - start_time;
@@ -1493,6 +1366,20 @@ void disk_end_io_acct(struct gendisk *disk, unsigned int op,
 	part_stat_local_dec(part, in_flight[op_is_write(op)]);
 	part_stat_unlock();
 }
+
+void part_end_io_acct(struct hd_struct *part, struct bio *bio,
+		      unsigned long start_time)
+{
+	__part_end_io_acct(part, bio_op(bio), start_time);
+	hd_struct_put(part);
+}
+EXPORT_SYMBOL_GPL(part_end_io_acct);
+
+void disk_end_io_acct(struct gendisk *disk, unsigned int op,
+		      unsigned long start_time)
+{
+	__part_end_io_acct(&disk->part0, op, start_time);
+}
 EXPORT_SYMBOL(disk_end_io_acct);
 
 /*
diff --git a/block/blk-integrity.c b/block/blk-integrity.c
index c03705cbb9c9..2b36a8f9b813 100644
--- a/block/blk-integrity.c
+++ b/block/blk-integrity.c
@@ -408,7 +408,7 @@ void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template
 	bi->tuple_size = template->tuple_size;
 	bi->tag_size = template->tag_size;
 
-	disk->queue->backing_dev_info->capabilities |= BDI_CAP_STABLE_WRITES;
+	blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue);
 
 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
 	if (disk->queue->ksm) {
@@ -428,7 +428,7 @@ EXPORT_SYMBOL(blk_integrity_register);
  */
 void blk_integrity_unregister(struct gendisk *disk)
 {
-	disk->queue->backing_dev_info->capabilities &= ~BDI_CAP_STABLE_WRITES;
+	blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, disk->queue);
 	memset(&disk->queue->integrity, 0, sizeof(struct blk_integrity));
 }
 EXPORT_SYMBOL(blk_integrity_unregister);
diff --git a/block/blk-iocost.c b/block/blk-iocost.c
index d37b55db2409..b82649c1440b 100644
--- a/block/blk-iocost.c
+++ b/block/blk-iocost.c
@@ -68,7 +68,7 @@
  * gets 300/(100+300) or 75% share, and A0 and A1 equally splits the rest,
  * 12.5% each.  The distribution mechanism only cares about these flattened
  * shares.  They're called hweights (hierarchical weights) and always add
- * upto 1 (HWEIGHT_WHOLE).
+ * upto 1 (WEIGHT_ONE).
  *
  * A given cgroup's vtime runs slower in inverse proportion to its hweight.
  * For example, with 12.5% weight, A0's time runs 8 times slower (100/12.5)
@@ -179,6 +179,8 @@
 #include <linux/parser.h>
 #include <linux/sched/signal.h>
 #include <linux/blk-cgroup.h>
+#include <asm/local.h>
+#include <asm/local64.h>
 #include "blk-rq-qos.h"
 #include "blk-stat.h"
 #include "blk-wbt.h"
@@ -215,36 +217,21 @@ enum {
 	MAX_PERIOD		= USEC_PER_SEC,
 
 	/*
-	 * A cgroup's vtime can run 50% behind the device vtime, which
+	 * iocg->vtime is targeted at 50% behind the device vtime, which
 	 * serves as its IO credit buffer.  Surplus weight adjustment is
 	 * immediately canceled if the vtime margin runs below 10%.
 	 */
-	MARGIN_PCT		= 50,
-	INUSE_MARGIN_PCT	= 10,
+	MARGIN_MIN_PCT		= 10,
+	MARGIN_LOW_PCT		= 20,
+	MARGIN_TARGET_PCT	= 50,
 
-	/* Have some play in waitq timer operations */
-	WAITQ_TIMER_MARGIN_PCT	= 5,
+	INUSE_ADJ_STEP_PCT	= 25,
 
-	/*
-	 * vtime can wrap well within a reasonable uptime when vrate is
-	 * consistently raised.  Don't trust recorded cgroup vtime if the
-	 * period counter indicates that it's older than 5mins.
-	 */
-	VTIME_VALID_DUR		= 300 * USEC_PER_SEC,
-
-	/*
-	 * Remember the past three non-zero usages and use the max for
-	 * surplus calculation.  Three slots guarantee that we remember one
-	 * full period usage from the last active stretch even after
-	 * partial deactivation and re-activation periods.  Don't start
-	 * giving away weight before collecting two data points to prevent
-	 * hweight adjustments based on one partial activation period.
-	 */
-	NR_USAGE_SLOTS		= 3,
-	MIN_VALID_USAGES	= 2,
+	/* Have some play in timer operations */
+	TIMER_SLACK_PCT		= 1,
 
 	/* 1/64k is granular enough and can easily be handled w/ u32 */
-	HWEIGHT_WHOLE		= 1 << 16,
+	WEIGHT_ONE		= 1 << 16,
 
 	/*
 	 * As vtime is used to calculate the cost of each IO, it needs to
@@ -275,16 +262,37 @@ enum {
 	/* unbusy hysterisis */
 	UNBUSY_THR_PCT		= 75,
 
-	/* don't let cmds which take a very long time pin lagging for too long */
-	MAX_LAGGING_PERIODS	= 10,
-
 	/*
-	 * If usage% * 1.25 + 2% is lower than hweight% by more than 3%,
-	 * donate the surplus.
+	 * The effect of delay is indirect and non-linear and a huge amount of
+	 * future debt can accumulate abruptly while unthrottled. Linearly scale
+	 * up delay as debt is going up and then let it decay exponentially.
+	 * This gives us quick ramp ups while delay is accumulating and long
+	 * tails which can help reducing the frequency of debt explosions on
+	 * unthrottle. The parameters are experimentally determined.
+	 *
+	 * The delay mechanism provides adequate protection and behavior in many
+	 * cases. However, this is far from ideal and falls shorts on both
+	 * fronts. The debtors are often throttled too harshly costing a
+	 * significant level of fairness and possibly total work while the
+	 * protection against their impacts on the system can be choppy and
+	 * unreliable.
+	 *
+	 * The shortcoming primarily stems from the fact that, unlike for page
+	 * cache, the kernel doesn't have well-defined back-pressure propagation
+	 * mechanism and policies for anonymous memory. Fully addressing this
+	 * issue will likely require substantial improvements in the area.
 	 */
-	SURPLUS_SCALE_PCT	= 125,			/* * 125% */
-	SURPLUS_SCALE_ABS	= HWEIGHT_WHOLE / 50,	/* + 2% */
-	SURPLUS_MIN_ADJ_DELTA	= HWEIGHT_WHOLE / 33,	/* 3% */
+	MIN_DELAY_THR_PCT	= 500,
+	MAX_DELAY_THR_PCT	= 25000,
+	MIN_DELAY		= 250,
+	MAX_DELAY		= 250 * USEC_PER_MSEC,
+
+	/* halve debts if avg usage over 100ms is under 50% */
+	DFGV_USAGE_PCT		= 50,
+	DFGV_PERIOD		= 100 * USEC_PER_MSEC,
+
+	/* don't let cmds which take a very long time pin lagging for too long */
+	MAX_LAGGING_PERIODS	= 10,
 
 	/* switch iff the conditions are met for longer than this */
 	AUTOP_CYCLE_NSEC	= 10LLU * NSEC_PER_SEC,
@@ -372,9 +380,15 @@ struct ioc_params {
 	u32				too_slow_vrate_pct;
 };
 
+struct ioc_margins {
+	s64				min;
+	s64				low;
+	s64				target;
+};
+
 struct ioc_missed {
-	u32				nr_met;
-	u32				nr_missed;
+	local_t				nr_met;
+	local_t				nr_missed;
 	u32				last_met;
 	u32				last_missed;
 };
@@ -382,7 +396,7 @@ struct ioc_missed {
 struct ioc_pcpu_stat {
 	struct ioc_missed		missed[2];
 
-	u64				rq_wait_ns;
+	local64_t			rq_wait_ns;
 	u64				last_rq_wait_ns;
 };
 
@@ -393,8 +407,9 @@ struct ioc {
 	bool				enabled;
 
 	struct ioc_params		params;
+	struct ioc_margins		margins;
 	u32				period_us;
-	u32				margin_us;
+	u32				timer_slack_ns;
 	u64				vrate_min;
 	u64				vrate_max;
 
@@ -405,18 +420,24 @@ struct ioc {
 
 	enum ioc_running		running;
 	atomic64_t			vtime_rate;
+	u64				vtime_base_rate;
+	s64				vtime_err;
 
 	seqcount_spinlock_t		period_seqcount;
-	u32				period_at;	/* wallclock starttime */
+	u64				period_at;	/* wallclock starttime */
 	u64				period_at_vtime; /* vtime starttime */
 
 	atomic64_t			cur_period;	/* inc'd each period */
 	int				busy_level;	/* saturation history */
 
-	u64				inuse_margin_vtime;
 	bool				weights_updated;
 	atomic_t			hweight_gen;	/* for lazy hweights */
 
+	/* debt forgivness */
+	u64				dfgv_period_at;
+	u64				dfgv_period_rem;
+	u64				dfgv_usage_us_sum;
+
 	u64				autop_too_fast_at;
 	u64				autop_too_slow_at;
 	int				autop_idx;
@@ -424,6 +445,17 @@ struct ioc {
 	bool				user_cost_model:1;
 };
 
+struct iocg_pcpu_stat {
+	local64_t			abs_vusage;
+};
+
+struct iocg_stat {
+	u64				usage_us;
+	u64				wait_us;
+	u64				indebt_us;
+	u64				indelay_us;
+};
+
 /* per device-cgroup pair */
 struct ioc_gq {
 	struct blkg_policy_data		pd;
@@ -443,12 +475,17 @@ struct ioc_gq {
 	 *
 	 * `last_inuse` remembers `inuse` while an iocg is idle to persist
 	 * surplus adjustments.
+	 *
+	 * `inuse` may be adjusted dynamically during period. `saved_*` are used
+	 * to determine and track adjustments.
 	 */
 	u32				cfg_weight;
 	u32				weight;
 	u32				active;
 	u32				inuse;
+
 	u32				last_inuse;
+	s64				saved_margin;
 
 	sector_t			cursor;		/* to detect randio */
 
@@ -461,14 +498,14 @@ struct ioc_gq {
 	 * `vtime_done` is the same but progressed on completion rather
 	 * than issue.  The delta behind `vtime` represents the cost of
 	 * currently in-flight IOs.
-	 *
-	 * `last_vtime` is used to remember `vtime` at the end of the last
-	 * period to calculate utilization.
 	 */
 	atomic64_t			vtime;
 	atomic64_t			done_vtime;
 	u64				abs_vdebt;
-	u64				last_vtime;
+
+	/* current delay in effect and when it started */
+	u64				delay;
+	u64				delay_at;
 
 	/*
 	 * The period this iocg was last active in.  Used for deactivation
@@ -477,21 +514,35 @@ struct ioc_gq {
 	atomic64_t			active_period;
 	struct list_head		active_list;
 
-	/* see __propagate_active_weight() and current_hweight() for details */
+	/* see __propagate_weights() and current_hweight() for details */
 	u64				child_active_sum;
 	u64				child_inuse_sum;
+	u64				child_adjusted_sum;
 	int				hweight_gen;
 	u32				hweight_active;
 	u32				hweight_inuse;
-	bool				has_surplus;
+	u32				hweight_donating;
+	u32				hweight_after_donation;
+
+	struct list_head		walk_list;
+	struct list_head		surplus_list;
 
 	struct wait_queue_head		waitq;
 	struct hrtimer			waitq_timer;
-	struct hrtimer			delay_timer;
 
-	/* usage is recorded as fractions of HWEIGHT_WHOLE */
-	int				usage_idx;
-	u32				usages[NR_USAGE_SLOTS];
+	/* timestamp at the latest activation */
+	u64				activated_at;
+
+	/* statistics */
+	struct iocg_pcpu_stat __percpu	*pcpu_stat;
+	struct iocg_stat		local_stat;
+	struct iocg_stat		desc_stat;
+	struct iocg_stat		last_stat;
+	u64				last_stat_abs_vusage;
+	u64				usage_delta_us;
+	u64				wait_since;
+	u64				indebt_since;
+	u64				indelay_since;
 
 	/* this iocg's depth in the hierarchy and ancestors including self */
 	int				level;
@@ -506,7 +557,7 @@ struct ioc_cgrp {
 
 struct ioc_now {
 	u64				now_ns;
-	u32				now;
+	u64				now;
 	u64				vnow;
 	u64				vrate;
 };
@@ -656,7 +707,7 @@ static struct ioc_cgrp *blkcg_to_iocc(struct blkcg *blkcg)
  */
 static u64 abs_cost_to_cost(u64 abs_cost, u32 hw_inuse)
 {
-	return DIV64_U64_ROUND_UP(abs_cost * HWEIGHT_WHOLE, hw_inuse);
+	return DIV64_U64_ROUND_UP(abs_cost * WEIGHT_ONE, hw_inuse);
 }
 
 /*
@@ -664,18 +715,56 @@ static u64 abs_cost_to_cost(u64 abs_cost, u32 hw_inuse)
  */
 static u64 cost_to_abs_cost(u64 cost, u32 hw_inuse)
 {
-	return DIV64_U64_ROUND_UP(cost * hw_inuse, HWEIGHT_WHOLE);
+	return DIV64_U64_ROUND_UP(cost * hw_inuse, WEIGHT_ONE);
 }
 
-static void iocg_commit_bio(struct ioc_gq *iocg, struct bio *bio, u64 cost)
+static void iocg_commit_bio(struct ioc_gq *iocg, struct bio *bio,
+			    u64 abs_cost, u64 cost)
 {
+	struct iocg_pcpu_stat *gcs;
+
 	bio->bi_iocost_cost = cost;
 	atomic64_add(cost, &iocg->vtime);
+
+	gcs = get_cpu_ptr(iocg->pcpu_stat);
+	local64_add(abs_cost, &gcs->abs_vusage);
+	put_cpu_ptr(gcs);
+}
+
+static void iocg_lock(struct ioc_gq *iocg, bool lock_ioc, unsigned long *flags)
+{
+	if (lock_ioc) {
+		spin_lock_irqsave(&iocg->ioc->lock, *flags);
+		spin_lock(&iocg->waitq.lock);
+	} else {
+		spin_lock_irqsave(&iocg->waitq.lock, *flags);
+	}
+}
+
+static void iocg_unlock(struct ioc_gq *iocg, bool unlock_ioc, unsigned long *flags)
+{
+	if (unlock_ioc) {
+		spin_unlock(&iocg->waitq.lock);
+		spin_unlock_irqrestore(&iocg->ioc->lock, *flags);
+	} else {
+		spin_unlock_irqrestore(&iocg->waitq.lock, *flags);
+	}
 }
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/iocost.h>
 
+static void ioc_refresh_margins(struct ioc *ioc)
+{
+	struct ioc_margins *margins = &ioc->margins;
+	u32 period_us = ioc->period_us;
+	u64 vrate = ioc->vtime_base_rate;
+
+	margins->min = (period_us * MARGIN_MIN_PCT / 100) * vrate;
+	margins->low = (period_us * MARGIN_LOW_PCT / 100) * vrate;
+	margins->target = (period_us * MARGIN_TARGET_PCT / 100) * vrate;
+}
+
 /* latency Qos params changed, update period_us and all the dependent params */
 static void ioc_refresh_period_us(struct ioc *ioc)
 {
@@ -709,9 +798,10 @@ static void ioc_refresh_period_us(struct ioc *ioc)
 
 	/* calculate dependent params */
 	ioc->period_us = period_us;
-	ioc->margin_us = period_us * MARGIN_PCT / 100;
-	ioc->inuse_margin_vtime = DIV64_U64_ROUND_UP(
-			period_us * VTIME_PER_USEC * INUSE_MARGIN_PCT, 100);
+	ioc->timer_slack_ns = div64_u64(
+		(u64)period_us * NSEC_PER_USEC * TIMER_SLACK_PCT,
+		100);
+	ioc_refresh_margins(ioc);
 }
 
 static int ioc_autop_idx(struct ioc *ioc)
@@ -738,8 +828,7 @@ static int ioc_autop_idx(struct ioc *ioc)
 		return idx;
 
 	/* step up/down based on the vrate */
-	vrate_pct = div64_u64(atomic64_read(&ioc->vtime_rate) * 100,
-			      VTIME_PER_USEC);
+	vrate_pct = div64_u64(ioc->vtime_base_rate * 100, VTIME_PER_USEC);
 	now_ns = ktime_get_ns();
 
 	if (p->too_fast_vrate_pct && p->too_fast_vrate_pct <= vrate_pct) {
@@ -847,6 +936,43 @@ static bool ioc_refresh_params(struct ioc *ioc, bool force)
 	return true;
 }
 
+/*
+ * When an iocg accumulates too much vtime or gets deactivated, we throw away
+ * some vtime, which lowers the overall device utilization. As the exact amount
+ * which is being thrown away is known, we can compensate by accelerating the
+ * vrate accordingly so that the extra vtime generated in the current period
+ * matches what got lost.
+ */
+static void ioc_refresh_vrate(struct ioc *ioc, struct ioc_now *now)
+{
+	s64 pleft = ioc->period_at + ioc->period_us - now->now;
+	s64 vperiod = ioc->period_us * ioc->vtime_base_rate;
+	s64 vcomp, vcomp_min, vcomp_max;
+
+	lockdep_assert_held(&ioc->lock);
+
+	/* we need some time left in this period */
+	if (pleft <= 0)
+		goto done;
+
+	/*
+	 * Calculate how much vrate should be adjusted to offset the error.
+	 * Limit the amount of adjustment and deduct the adjusted amount from
+	 * the error.
+	 */
+	vcomp = -div64_s64(ioc->vtime_err, pleft);
+	vcomp_min = -(ioc->vtime_base_rate >> 1);
+	vcomp_max = ioc->vtime_base_rate;
+	vcomp = clamp(vcomp, vcomp_min, vcomp_max);
+
+	ioc->vtime_err += vcomp * pleft;
+
+	atomic64_set(&ioc->vtime_rate, ioc->vtime_base_rate + vcomp);
+done:
+	/* bound how much error can accumulate */
+	ioc->vtime_err = clamp(ioc->vtime_err, -vperiod, vperiod);
+}
+
 /* take a snapshot of the current [v]time and vrate */
 static void ioc_now(struct ioc *ioc, struct ioc_now *now)
 {
@@ -886,16 +1012,25 @@ static void ioc_start_period(struct ioc *ioc, struct ioc_now *now)
 
 /*
  * Update @iocg's `active` and `inuse` to @active and @inuse, update level
- * weight sums and propagate upwards accordingly.
+ * weight sums and propagate upwards accordingly. If @save, the current margin
+ * is saved to be used as reference for later inuse in-period adjustments.
  */
-static void __propagate_active_weight(struct ioc_gq *iocg, u32 active, u32 inuse)
+static void __propagate_weights(struct ioc_gq *iocg, u32 active, u32 inuse,
+				bool save, struct ioc_now *now)
 {
 	struct ioc *ioc = iocg->ioc;
 	int lvl;
 
 	lockdep_assert_held(&ioc->lock);
 
-	inuse = min(active, inuse);
+	inuse = clamp_t(u32, inuse, 1, active);
+
+	iocg->last_inuse = iocg->inuse;
+	if (save)
+		iocg->saved_margin = now->vnow - atomic64_read(&iocg->vtime);
+
+	if (active == iocg->active && inuse == iocg->inuse)
+		return;
 
 	for (lvl = iocg->level - 1; lvl >= 0; lvl--) {
 		struct ioc_gq *parent = iocg->ancestors[lvl];
@@ -933,7 +1068,7 @@ static void __propagate_active_weight(struct ioc_gq *iocg, u32 active, u32 inuse
 	ioc->weights_updated = true;
 }
 
-static void commit_active_weights(struct ioc *ioc)
+static void commit_weights(struct ioc *ioc)
 {
 	lockdep_assert_held(&ioc->lock);
 
@@ -945,10 +1080,11 @@ static void commit_active_weights(struct ioc *ioc)
 	}
 }
 
-static void propagate_active_weight(struct ioc_gq *iocg, u32 active, u32 inuse)
+static void propagate_weights(struct ioc_gq *iocg, u32 active, u32 inuse,
+			      bool save, struct ioc_now *now)
 {
-	__propagate_active_weight(iocg, active, inuse);
-	commit_active_weights(iocg->ioc);
+	__propagate_weights(iocg, active, inuse, save, now);
+	commit_weights(iocg->ioc);
 }
 
 static void current_hweight(struct ioc_gq *iocg, u32 *hw_activep, u32 *hw_inusep)
@@ -964,9 +1100,9 @@ static void current_hweight(struct ioc_gq *iocg, u32 *hw_activep, u32 *hw_inusep
 		goto out;
 
 	/*
-	 * Paired with wmb in commit_active_weights().  If we saw the
-	 * updated hweight_gen, all the weight updates from
-	 * __propagate_active_weight() are visible too.
+	 * Paired with wmb in commit_weights(). If we saw the updated
+	 * hweight_gen, all the weight updates from __propagate_weights() are
+	 * visible too.
 	 *
 	 * We can race with weight updates during calculation and get it
 	 * wrong.  However, hweight_gen would have changed and a future
@@ -975,12 +1111,12 @@ static void current_hweight(struct ioc_gq *iocg, u32 *hw_activep, u32 *hw_inusep
 	 */
 	smp_rmb();
 
-	hwa = hwi = HWEIGHT_WHOLE;
+	hwa = hwi = WEIGHT_ONE;
 	for (lvl = 0; lvl <= iocg->level - 1; lvl++) {
 		struct ioc_gq *parent = iocg->ancestors[lvl];
 		struct ioc_gq *child = iocg->ancestors[lvl + 1];
-		u32 active_sum = READ_ONCE(parent->child_active_sum);
-		u32 inuse_sum = READ_ONCE(parent->child_inuse_sum);
+		u64 active_sum = READ_ONCE(parent->child_active_sum);
+		u64 inuse_sum = READ_ONCE(parent->child_inuse_sum);
 		u32 active = READ_ONCE(child->active);
 		u32 inuse = READ_ONCE(child->inuse);
 
@@ -988,11 +1124,11 @@ static void current_hweight(struct ioc_gq *iocg, u32 *hw_activep, u32 *hw_inusep
 		if (!active_sum || !inuse_sum)
 			continue;
 
-		active_sum = max(active, active_sum);
-		hwa = hwa * active / active_sum;	/* max 16bits * 10000 */
+		active_sum = max_t(u64, active, active_sum);
+		hwa = div64_u64((u64)hwa * active, active_sum);
 
-		inuse_sum = max(inuse, inuse_sum);
-		hwi = hwi * inuse / inuse_sum;		/* max 16bits * 10000 */
+		inuse_sum = max_t(u64, inuse, inuse_sum);
+		hwi = div64_u64((u64)hwi * inuse, inuse_sum);
 	}
 
 	iocg->hweight_active = max_t(u32, hwa, 1);
@@ -1005,7 +1141,33 @@ out:
 		*hw_inusep = iocg->hweight_inuse;
 }
 
-static void weight_updated(struct ioc_gq *iocg)
+/*
+ * Calculate the hweight_inuse @iocg would get with max @inuse assuming all the
+ * other weights stay unchanged.
+ */
+static u32 current_hweight_max(struct ioc_gq *iocg)
+{
+	u32 hwm = WEIGHT_ONE;
+	u32 inuse = iocg->active;
+	u64 child_inuse_sum;
+	int lvl;
+
+	lockdep_assert_held(&iocg->ioc->lock);
+
+	for (lvl = iocg->level - 1; lvl >= 0; lvl--) {
+		struct ioc_gq *parent = iocg->ancestors[lvl];
+		struct ioc_gq *child = iocg->ancestors[lvl + 1];
+
+		child_inuse_sum = parent->child_inuse_sum + inuse - child->inuse;
+		hwm = div64_u64((u64)hwm * inuse, child_inuse_sum);
+		inuse = DIV64_U64_ROUND_UP(parent->active * child_inuse_sum,
+					   parent->child_active_sum);
+	}
+
+	return max_t(u32, hwm, 1);
+}
+
+static void weight_updated(struct ioc_gq *iocg, struct ioc_now *now)
 {
 	struct ioc *ioc = iocg->ioc;
 	struct blkcg_gq *blkg = iocg_to_blkg(iocg);
@@ -1016,16 +1178,15 @@ static void weight_updated(struct ioc_gq *iocg)
 
 	weight = iocg->cfg_weight ?: iocc->dfl_weight;
 	if (weight != iocg->weight && iocg->active)
-		propagate_active_weight(iocg, weight,
-			DIV64_U64_ROUND_UP(iocg->inuse * weight, iocg->weight));
+		propagate_weights(iocg, weight, iocg->inuse, true, now);
 	iocg->weight = weight;
 }
 
 static bool iocg_activate(struct ioc_gq *iocg, struct ioc_now *now)
 {
 	struct ioc *ioc = iocg->ioc;
-	u64 last_period, cur_period, max_period_delta;
-	u64 vtime, vmargin, vmin;
+	u64 last_period, cur_period;
+	u64 vtime, vtarget;
 	int i;
 
 	/*
@@ -1064,22 +1225,15 @@ static bool iocg_activate(struct ioc_gq *iocg, struct ioc_now *now)
 		goto fail_unlock;
 
 	/*
-	 * vtime may wrap when vrate is raised substantially due to
-	 * underestimated IO costs.  Look at the period and ignore its
-	 * vtime if the iocg has been idle for too long.  Also, cap the
-	 * budget it can start with to the margin.
+	 * Always start with the target budget. On deactivation, we throw away
+	 * anything above it.
 	 */
-	max_period_delta = DIV64_U64_ROUND_UP(VTIME_VALID_DUR, ioc->period_us);
+	vtarget = now->vnow - ioc->margins.target;
 	vtime = atomic64_read(&iocg->vtime);
-	vmargin = ioc->margin_us * now->vrate;
-	vmin = now->vnow - vmargin;
 
-	if (last_period + max_period_delta < cur_period ||
-	    time_before64(vtime, vmin)) {
-		atomic64_add(vmin - vtime, &iocg->vtime);
-		atomic64_add(vmin - vtime, &iocg->done_vtime);
-		vtime = vmin;
-	}
+	atomic64_add(vtarget - vtime, &iocg->vtime);
+	atomic64_add(vtarget - vtime, &iocg->done_vtime);
+	vtime = vtarget;
 
 	/*
 	 * Activate, propagate weight and start period timer if not
@@ -1088,16 +1242,19 @@ static bool iocg_activate(struct ioc_gq *iocg, struct ioc_now *now)
 	 */
 	iocg->hweight_gen = atomic_read(&ioc->hweight_gen) - 1;
 	list_add(&iocg->active_list, &ioc->active_iocgs);
-	propagate_active_weight(iocg, iocg->weight,
-				iocg->last_inuse ?: iocg->weight);
+
+	propagate_weights(iocg, iocg->weight,
+			  iocg->last_inuse ?: iocg->weight, true, now);
 
 	TRACE_IOCG_PATH(iocg_activate, iocg, now,
 			last_period, cur_period, vtime);
 
-	iocg->last_vtime = vtime;
+	iocg->activated_at = now->now;
 
 	if (ioc->running == IOC_IDLE) {
 		ioc->running = IOC_RUNNING;
+		ioc->dfgv_period_at = now->now;
+		ioc->dfgv_period_rem = 0;
 		ioc_start_period(ioc, now);
 	}
 
@@ -1110,6 +1267,110 @@ fail_unlock:
 	return false;
 }
 
+static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
+{
+	struct ioc *ioc = iocg->ioc;
+	struct blkcg_gq *blkg = iocg_to_blkg(iocg);
+	u64 tdelta, delay, new_delay;
+	s64 vover, vover_pct;
+	u32 hwa;
+
+	lockdep_assert_held(&iocg->waitq.lock);
+
+	/* calculate the current delay in effect - 1/2 every second */
+	tdelta = now->now - iocg->delay_at;
+	if (iocg->delay)
+		delay = iocg->delay >> div64_u64(tdelta, USEC_PER_SEC);
+	else
+		delay = 0;
+
+	/* calculate the new delay from the debt amount */
+	current_hweight(iocg, &hwa, NULL);
+	vover = atomic64_read(&iocg->vtime) +
+		abs_cost_to_cost(iocg->abs_vdebt, hwa) - now->vnow;
+	vover_pct = div64_s64(100 * vover,
+			      ioc->period_us * ioc->vtime_base_rate);
+
+	if (vover_pct <= MIN_DELAY_THR_PCT)
+		new_delay = 0;
+	else if (vover_pct >= MAX_DELAY_THR_PCT)
+		new_delay = MAX_DELAY;
+	else
+		new_delay = MIN_DELAY +
+			div_u64((MAX_DELAY - MIN_DELAY) *
+				(vover_pct - MIN_DELAY_THR_PCT),
+				MAX_DELAY_THR_PCT - MIN_DELAY_THR_PCT);
+
+	/* pick the higher one and apply */
+	if (new_delay > delay) {
+		iocg->delay = new_delay;
+		iocg->delay_at = now->now;
+		delay = new_delay;
+	}
+
+	if (delay >= MIN_DELAY) {
+		if (!iocg->indelay_since)
+			iocg->indelay_since = now->now;
+		blkcg_set_delay(blkg, delay * NSEC_PER_USEC);
+		return true;
+	} else {
+		if (iocg->indelay_since) {
+			iocg->local_stat.indelay_us += now->now - iocg->indelay_since;
+			iocg->indelay_since = 0;
+		}
+		iocg->delay = 0;
+		blkcg_clear_delay(blkg);
+		return false;
+	}
+}
+
+static void iocg_incur_debt(struct ioc_gq *iocg, u64 abs_cost,
+			    struct ioc_now *now)
+{
+	struct iocg_pcpu_stat *gcs;
+
+	lockdep_assert_held(&iocg->ioc->lock);
+	lockdep_assert_held(&iocg->waitq.lock);
+	WARN_ON_ONCE(list_empty(&iocg->active_list));
+
+	/*
+	 * Once in debt, debt handling owns inuse. @iocg stays at the minimum
+	 * inuse donating all of it share to others until its debt is paid off.
+	 */
+	if (!iocg->abs_vdebt && abs_cost) {
+		iocg->indebt_since = now->now;
+		propagate_weights(iocg, iocg->active, 0, false, now);
+	}
+
+	iocg->abs_vdebt += abs_cost;
+
+	gcs = get_cpu_ptr(iocg->pcpu_stat);
+	local64_add(abs_cost, &gcs->abs_vusage);
+	put_cpu_ptr(gcs);
+}
+
+static void iocg_pay_debt(struct ioc_gq *iocg, u64 abs_vpay,
+			  struct ioc_now *now)
+{
+	lockdep_assert_held(&iocg->ioc->lock);
+	lockdep_assert_held(&iocg->waitq.lock);
+
+	/* make sure that nobody messed with @iocg */
+	WARN_ON_ONCE(list_empty(&iocg->active_list));
+	WARN_ON_ONCE(iocg->inuse > 1);
+
+	iocg->abs_vdebt -= min(abs_vpay, iocg->abs_vdebt);
+
+	/* if debt is paid in full, restore inuse */
+	if (!iocg->abs_vdebt) {
+		iocg->local_stat.indebt_us += now->now - iocg->indebt_since;
+		iocg->indebt_since = 0;
+
+		propagate_weights(iocg, iocg->active, iocg->last_inuse,
+				  false, now);
+	}
+}
+
 static int iocg_wake_fn(struct wait_queue_entry *wq_entry, unsigned mode,
 			int flags, void *key)
 {
@@ -1122,7 +1383,7 @@ static int iocg_wake_fn(struct wait_queue_entry *wq_entry, unsigned mode,
 	if (ctx->vbudget < 0)
 		return -1;
 
-	iocg_commit_bio(ctx->iocg, wait->bio, cost);
+	iocg_commit_bio(ctx->iocg, wait->bio, wait->abs_cost, cost);
 
 	/*
 	 * autoremove_wake_function() removes the wait entry only when it
@@ -1136,132 +1397,106 @@ static int iocg_wake_fn(struct wait_queue_entry *wq_entry, unsigned mode,
 	return 0;
 }
 
-static void iocg_kick_waitq(struct ioc_gq *iocg, struct ioc_now *now)
+/*
+ * Calculate the accumulated budget, pay debt if @pay_debt and wake up waiters
+ * accordingly. When @pay_debt is %true, the caller must be holding ioc->lock in
+ * addition to iocg->waitq.lock.
+ */
+static void iocg_kick_waitq(struct ioc_gq *iocg, bool pay_debt,
+			    struct ioc_now *now)
 {
 	struct ioc *ioc = iocg->ioc;
 	struct iocg_wake_ctx ctx = { .iocg = iocg };
-	u64 margin_ns = (u64)(ioc->period_us *
-			      WAITQ_TIMER_MARGIN_PCT / 100) * NSEC_PER_USEC;
-	u64 vdebt, vshortage, expires, oexpires;
+	u64 vshortage, expires, oexpires;
 	s64 vbudget;
-	u32 hw_inuse;
+	u32 hwa;
 
 	lockdep_assert_held(&iocg->waitq.lock);
 
-	current_hweight(iocg, NULL, &hw_inuse);
+	current_hweight(iocg, &hwa, NULL);
 	vbudget = now->vnow - atomic64_read(&iocg->vtime);
 
 	/* pay off debt */
-	vdebt = abs_cost_to_cost(iocg->abs_vdebt, hw_inuse);
-	if (vdebt && vbudget > 0) {
-		u64 delta = min_t(u64, vbudget, vdebt);
-		u64 abs_delta = min(cost_to_abs_cost(delta, hw_inuse),
-				    iocg->abs_vdebt);
+	if (pay_debt && iocg->abs_vdebt && vbudget > 0) {
+		u64 abs_vbudget = cost_to_abs_cost(vbudget, hwa);
+		u64 abs_vpay = min_t(u64, abs_vbudget, iocg->abs_vdebt);
+		u64 vpay = abs_cost_to_cost(abs_vpay, hwa);
 
-		atomic64_add(delta, &iocg->vtime);
-		atomic64_add(delta, &iocg->done_vtime);
-		iocg->abs_vdebt -= abs_delta;
+		lockdep_assert_held(&ioc->lock);
+
+		atomic64_add(vpay, &iocg->vtime);
+		atomic64_add(vpay, &iocg->done_vtime);
+		iocg_pay_debt(iocg, abs_vpay, now);
+		vbudget -= vpay;
 	}
 
+	if (iocg->abs_vdebt || iocg->delay)
+		iocg_kick_delay(iocg, now);
+
 	/*
-	 * Wake up the ones which are due and see how much vtime we'll need
-	 * for the next one.
+	 * Debt can still be outstanding if we haven't paid all yet or the
+	 * caller raced and called without @pay_debt. Shouldn't wake up waiters
+	 * under debt. Make sure @vbudget reflects the outstanding amount and is
+	 * not positive.
 	 */
-	ctx.hw_inuse = hw_inuse;
-	ctx.vbudget = vbudget - vdebt;
+	if (iocg->abs_vdebt) {
+		s64 vdebt = abs_cost_to_cost(iocg->abs_vdebt, hwa);
+		vbudget = min_t(s64, 0, vbudget - vdebt);
+	}
+
+	/*
+	 * Wake up the ones which are due and see how much vtime we'll need for
+	 * the next one. As paying off debt restores hw_inuse, it must be read
+	 * after the above debt payment.
+	 */
+	ctx.vbudget = vbudget;
+	current_hweight(iocg, NULL, &ctx.hw_inuse);
+
 	__wake_up_locked_key(&iocg->waitq, TASK_NORMAL, &ctx);
-	if (!waitqueue_active(&iocg->waitq))
+
+	if (!waitqueue_active(&iocg->waitq)) {
+		if (iocg->wait_since) {
+			iocg->local_stat.wait_us += now->now - iocg->wait_since;
+			iocg->wait_since = 0;
+		}
 		return;
+	}
+
+	if (!iocg->wait_since)
+		iocg->wait_since = now->now;
+
 	if (WARN_ON_ONCE(ctx.vbudget >= 0))
 		return;
 
-	/* determine next wakeup, add a quarter margin to guarantee chunking */
+	/* determine next wakeup, add a timer margin to guarantee chunking */
 	vshortage = -ctx.vbudget;
 	expires = now->now_ns +
-		DIV64_U64_ROUND_UP(vshortage, now->vrate) * NSEC_PER_USEC;
-	expires += margin_ns / 4;
+		DIV64_U64_ROUND_UP(vshortage, ioc->vtime_base_rate) *
+		NSEC_PER_USEC;
+	expires += ioc->timer_slack_ns;
 
 	/* if already active and close enough, don't bother */
 	oexpires = ktime_to_ns(hrtimer_get_softexpires(&iocg->waitq_timer));
 	if (hrtimer_is_queued(&iocg->waitq_timer) &&
-	    abs(oexpires - expires) <= margin_ns / 4)
+	    abs(oexpires - expires) <= ioc->timer_slack_ns)
 		return;
 
 	hrtimer_start_range_ns(&iocg->waitq_timer, ns_to_ktime(expires),
-			       margin_ns / 4, HRTIMER_MODE_ABS);
+			       ioc->timer_slack_ns, HRTIMER_MODE_ABS);
 }
 
 static enum hrtimer_restart iocg_waitq_timer_fn(struct hrtimer *timer)
 {
 	struct ioc_gq *iocg = container_of(timer, struct ioc_gq, waitq_timer);
+	bool pay_debt = READ_ONCE(iocg->abs_vdebt);
 	struct ioc_now now;
 	unsigned long flags;
 
 	ioc_now(iocg->ioc, &now);
 
-	spin_lock_irqsave(&iocg->waitq.lock, flags);
-	iocg_kick_waitq(iocg, &now);
-	spin_unlock_irqrestore(&iocg->waitq.lock, flags);
-
-	return HRTIMER_NORESTART;
-}
-
-static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
-{
-	struct ioc *ioc = iocg->ioc;
-	struct blkcg_gq *blkg = iocg_to_blkg(iocg);
-	u64 vtime = atomic64_read(&iocg->vtime);
-	u64 vmargin = ioc->margin_us * now->vrate;
-	u64 margin_ns = ioc->margin_us * NSEC_PER_USEC;
-	u64 delta_ns, expires, oexpires;
-	u32 hw_inuse;
-
-	lockdep_assert_held(&iocg->waitq.lock);
-
-	/* debt-adjust vtime */
-	current_hweight(iocg, NULL, &hw_inuse);
-	vtime += abs_cost_to_cost(iocg->abs_vdebt, hw_inuse);
-
-	/*
-	 * Clear or maintain depending on the overage. Non-zero vdebt is what
-	 * guarantees that @iocg is online and future iocg_kick_delay() will
-	 * clear use_delay. Don't leave it on when there's no vdebt.
-	 */
-	if (!iocg->abs_vdebt || time_before_eq64(vtime, now->vnow)) {
-		blkcg_clear_delay(blkg);
-		return false;
-	}
-	if (!atomic_read(&blkg->use_delay) &&
-	    time_before_eq64(vtime, now->vnow + vmargin))
-		return false;
-
-	/* use delay */
-	delta_ns = DIV64_U64_ROUND_UP(vtime - now->vnow,
-				      now->vrate) * NSEC_PER_USEC;
-	blkcg_set_delay(blkg, delta_ns);
-	expires = now->now_ns + delta_ns;
-
-	/* if already active and close enough, don't bother */
-	oexpires = ktime_to_ns(hrtimer_get_softexpires(&iocg->delay_timer));
-	if (hrtimer_is_queued(&iocg->delay_timer) &&
-	    abs(oexpires - expires) <= margin_ns / 4)
-		return true;
-
-	hrtimer_start_range_ns(&iocg->delay_timer, ns_to_ktime(expires),
-			       margin_ns / 4, HRTIMER_MODE_ABS);
-	return true;
-}
-
-static enum hrtimer_restart iocg_delay_timer_fn(struct hrtimer *timer)
-{
-	struct ioc_gq *iocg = container_of(timer, struct ioc_gq, delay_timer);
-	struct ioc_now now;
-	unsigned long flags;
-
-	spin_lock_irqsave(&iocg->waitq.lock, flags);
-	ioc_now(iocg->ioc, &now);
-	iocg_kick_delay(iocg, &now);
-	spin_unlock_irqrestore(&iocg->waitq.lock, flags);
+	iocg_lock(iocg, pay_debt, &flags);
+	iocg_kick_waitq(iocg, pay_debt, &now);
+	iocg_unlock(iocg, pay_debt, &flags);
 
 	return HRTIMER_NORESTART;
 }
@@ -1278,8 +1513,8 @@ static void ioc_lat_stat(struct ioc *ioc, u32 *missed_ppm_ar, u32 *rq_wait_pct_p
 		u64 this_rq_wait_ns;
 
 		for (rw = READ; rw <= WRITE; rw++) {
-			u32 this_met = READ_ONCE(stat->missed[rw].nr_met);
-			u32 this_missed = READ_ONCE(stat->missed[rw].nr_missed);
+			u32 this_met = local_read(&stat->missed[rw].nr_met);
+			u32 this_missed = local_read(&stat->missed[rw].nr_missed);
 
 			nr_met[rw] += this_met - stat->missed[rw].last_met;
 			nr_missed[rw] += this_missed - stat->missed[rw].last_missed;
@@ -1287,7 +1522,7 @@ static void ioc_lat_stat(struct ioc *ioc, u32 *missed_ppm_ar, u32 *rq_wait_pct_p
 			stat->missed[rw].last_missed = this_missed;
 		}
 
-		this_rq_wait_ns = READ_ONCE(stat->rq_wait_ns);
+		this_rq_wait_ns = local64_read(&stat->rq_wait_ns);
 		rq_wait_ns += this_rq_wait_ns - stat->last_rq_wait_ns;
 		stat->last_rq_wait_ns = this_rq_wait_ns;
 	}
@@ -1322,18 +1557,518 @@ static bool iocg_is_idle(struct ioc_gq *iocg)
 	return true;
 }
 
-/* returns usage with margin added if surplus is large enough */
-static u32 surplus_adjusted_hweight_inuse(u32 usage, u32 hw_inuse)
+/*
+ * Call this function on the target leaf @iocg's to build pre-order traversal
+ * list of all the ancestors in @inner_walk. The inner nodes are linked through
+ * ->walk_list and the caller is responsible for dissolving the list after use.
+ */
+static void iocg_build_inner_walk(struct ioc_gq *iocg,
+				  struct list_head *inner_walk)
 {
-	/* add margin */
-	usage = DIV_ROUND_UP(usage * SURPLUS_SCALE_PCT, 100);
-	usage += SURPLUS_SCALE_ABS;
+	int lvl;
 
-	/* don't bother if the surplus is too small */
-	if (usage + SURPLUS_MIN_ADJ_DELTA > hw_inuse)
-		return 0;
+	WARN_ON_ONCE(!list_empty(&iocg->walk_list));
 
-	return usage;
+	/* find the first ancestor which hasn't been visited yet */
+	for (lvl = iocg->level - 1; lvl >= 0; lvl--) {
+		if (!list_empty(&iocg->ancestors[lvl]->walk_list))
+			break;
+	}
+
+	/* walk down and visit the inner nodes to get pre-order traversal */
+	while (++lvl <= iocg->level - 1) {
+		struct ioc_gq *inner = iocg->ancestors[lvl];
+
+		/* record traversal order */
+		list_add_tail(&inner->walk_list, inner_walk);
+	}
+}
+
+/* collect per-cpu counters and propagate the deltas to the parent */
+static void iocg_flush_stat_one(struct ioc_gq *iocg, struct ioc_now *now)
+{
+	struct ioc *ioc = iocg->ioc;
+	struct iocg_stat new_stat;
+	u64 abs_vusage = 0;
+	u64 vusage_delta;
+	int cpu;
+
+	lockdep_assert_held(&iocg->ioc->lock);
+
+	/* collect per-cpu counters */
+	for_each_possible_cpu(cpu) {
+		abs_vusage += local64_read(
+				per_cpu_ptr(&iocg->pcpu_stat->abs_vusage, cpu));
+	}
+	vusage_delta = abs_vusage - iocg->last_stat_abs_vusage;
+	iocg->last_stat_abs_vusage = abs_vusage;
+
+	iocg->usage_delta_us = div64_u64(vusage_delta, ioc->vtime_base_rate);
+	iocg->local_stat.usage_us += iocg->usage_delta_us;
+
+	/* propagate upwards */
+	new_stat.usage_us =
+		iocg->local_stat.usage_us + iocg->desc_stat.usage_us;
+	new_stat.wait_us =
+		iocg->local_stat.wait_us + iocg->desc_stat.wait_us;
+	new_stat.indebt_us =
+		iocg->local_stat.indebt_us + iocg->desc_stat.indebt_us;
+	new_stat.indelay_us =
+		iocg->local_stat.indelay_us + iocg->desc_stat.indelay_us;
+
+	/* propagate the deltas to the parent */
+	if (iocg->level > 0) {
+		struct iocg_stat *parent_stat =
+			&iocg->ancestors[iocg->level - 1]->desc_stat;
+
+		parent_stat->usage_us +=
+			new_stat.usage_us - iocg->last_stat.usage_us;
+		parent_stat->wait_us +=
+			new_stat.wait_us - iocg->last_stat.wait_us;
+		parent_stat->indebt_us +=
+			new_stat.indebt_us - iocg->last_stat.indebt_us;
+		parent_stat->indelay_us +=
+			new_stat.indelay_us - iocg->last_stat.indelay_us;
+	}
+
+	iocg->last_stat = new_stat;
+}
+
+/* get stat counters ready for reading on all active iocgs */
+static void iocg_flush_stat(struct list_head *target_iocgs, struct ioc_now *now)
+{
+	LIST_HEAD(inner_walk);
+	struct ioc_gq *iocg, *tiocg;
+
+	/* flush leaves and build inner node walk list */
+	list_for_each_entry(iocg, target_iocgs, active_list) {
+		iocg_flush_stat_one(iocg, now);
+		iocg_build_inner_walk(iocg, &inner_walk);
+	}
+
+	/* keep flushing upwards by walking the inner list backwards */
+	list_for_each_entry_safe_reverse(iocg, tiocg, &inner_walk, walk_list) {
+		iocg_flush_stat_one(iocg, now);
+		list_del_init(&iocg->walk_list);
+	}
+}
+
+/*
+ * Determine what @iocg's hweight_inuse should be after donating unused
+ * capacity. @hwm is the upper bound and used to signal no donation. This
+ * function also throws away @iocg's excess budget.
+ */
+static u32 hweight_after_donation(struct ioc_gq *iocg, u32 old_hwi, u32 hwm,
+				  u32 usage, struct ioc_now *now)
+{
+	struct ioc *ioc = iocg->ioc;
+	u64 vtime = atomic64_read(&iocg->vtime);
+	s64 excess, delta, target, new_hwi;
+
+	/* debt handling owns inuse for debtors */
+	if (iocg->abs_vdebt)
+		return 1;
+
+	/* see whether minimum margin requirement is met */
+	if (waitqueue_active(&iocg->waitq) ||
+	    time_after64(vtime, now->vnow - ioc->margins.min))
+		return hwm;
+
+	/* throw away excess above target */
+	excess = now->vnow - vtime - ioc->margins.target;
+	if (excess > 0) {
+		atomic64_add(excess, &iocg->vtime);
+		atomic64_add(excess, &iocg->done_vtime);
+		vtime += excess;
+		ioc->vtime_err -= div64_u64(excess * old_hwi, WEIGHT_ONE);
+	}
+
+	/*
+	 * Let's say the distance between iocg's and device's vtimes as a
+	 * fraction of period duration is delta. Assuming that the iocg will
+	 * consume the usage determined above, we want to determine new_hwi so
+	 * that delta equals MARGIN_TARGET at the end of the next period.
+	 *
+	 * We need to execute usage worth of IOs while spending the sum of the
+	 * new budget (1 - MARGIN_TARGET) and the leftover from the last period
+	 * (delta):
+	 *
+	 *   usage = (1 - MARGIN_TARGET + delta) * new_hwi
+	 *
+	 * Therefore, the new_hwi is:
+	 *
+	 *   new_hwi = usage / (1 - MARGIN_TARGET + delta)
+	 */
+	delta = div64_s64(WEIGHT_ONE * (now->vnow - vtime),
+			  now->vnow - ioc->period_at_vtime);
+	target = WEIGHT_ONE * MARGIN_TARGET_PCT / 100;
+	new_hwi = div64_s64(WEIGHT_ONE * usage, WEIGHT_ONE - target + delta);
+
+	return clamp_t(s64, new_hwi, 1, hwm);
+}
+
+/*
+ * For work-conservation, an iocg which isn't using all of its share should
+ * donate the leftover to other iocgs. There are two ways to achieve this - 1.
+ * bumping up vrate accordingly 2. lowering the donating iocg's inuse weight.
+ *
+ * #1 is mathematically simpler but has the drawback of requiring synchronous
+ * global hweight_inuse updates when idle iocg's get activated or inuse weights
+ * change due to donation snapbacks as it has the possibility of grossly
+ * overshooting what's allowed by the model and vrate.
+ *
+ * #2 is inherently safe with local operations. The donating iocg can easily
+ * snap back to higher weights when needed without worrying about impacts on
+ * other nodes as the impacts will be inherently correct. This also makes idle
+ * iocg activations safe. The only effect activations have is decreasing
+ * hweight_inuse of others, the right solution to which is for those iocgs to
+ * snap back to higher weights.
+ *
+ * So, we go with #2. The challenge is calculating how each donating iocg's
+ * inuse should be adjusted to achieve the target donation amounts. This is done
+ * using Andy's method described in the following pdf.
+ *
+ *   https://drive.google.com/file/d/1PsJwxPFtjUnwOY1QJ5AeICCcsL7BM3bo
+ *
+ * Given the weights and target after-donation hweight_inuse values, Andy's
+ * method determines how the proportional distribution should look like at each
+ * sibling level to maintain the relative relationship between all non-donating
+ * pairs. To roughly summarize, it divides the tree into donating and
+ * non-donating parts, calculates global donation rate which is used to
+ * determine the target hweight_inuse for each node, and then derives per-level
+ * proportions.
+ *
+ * The following pdf shows that global distribution calculated this way can be
+ * achieved by scaling inuse weights of donating leaves and propagating the
+ * adjustments upwards proportionally.
+ *
+ *   https://drive.google.com/file/d/1vONz1-fzVO7oY5DXXsLjSxEtYYQbOvsE
+ *
+ * Combining the above two, we can determine how each leaf iocg's inuse should
+ * be adjusted to achieve the target donation.
+ *
+ *   https://drive.google.com/file/d/1WcrltBOSPN0qXVdBgnKm4mdp9FhuEFQN
+ *
+ * The inline comments use symbols from the last pdf.
+ *
+ *   b is the sum of the absolute budgets in the subtree. 1 for the root node.
+ *   f is the sum of the absolute budgets of non-donating nodes in the subtree.
+ *   t is the sum of the absolute budgets of donating nodes in the subtree.
+ *   w is the weight of the node. w = w_f + w_t
+ *   w_f is the non-donating portion of w. w_f = w * f / b
+ *   w_b is the donating portion of w. w_t = w * t / b
+ *   s is the sum of all sibling weights. s = Sum(w) for siblings
+ *   s_f and s_t are the non-donating and donating portions of s.
+ *
+ * Subscript p denotes the parent's counterpart and ' the adjusted value - e.g.
+ * w_pt is the donating portion of the parent's weight and w'_pt the same value
+ * after adjustments. Subscript r denotes the root node's values.
+ */
+static void transfer_surpluses(struct list_head *surpluses, struct ioc_now *now)
+{
+	LIST_HEAD(over_hwa);
+	LIST_HEAD(inner_walk);
+	struct ioc_gq *iocg, *tiocg, *root_iocg;
+	u32 after_sum, over_sum, over_target, gamma;
+
+	/*
+	 * It's pretty unlikely but possible for the total sum of
+	 * hweight_after_donation's to be higher than WEIGHT_ONE, which will
+	 * confuse the following calculations. If such condition is detected,
+	 * scale down everyone over its full share equally to keep the sum below
+	 * WEIGHT_ONE.
+	 */
+	after_sum = 0;
+	over_sum = 0;
+	list_for_each_entry(iocg, surpluses, surplus_list) {
+		u32 hwa;
+
+		current_hweight(iocg, &hwa, NULL);
+		after_sum += iocg->hweight_after_donation;
+
+		if (iocg->hweight_after_donation > hwa) {
+			over_sum += iocg->hweight_after_donation;
+			list_add(&iocg->walk_list, &over_hwa);
+		}
+	}
+
+	if (after_sum >= WEIGHT_ONE) {
+		/*
+		 * The delta should be deducted from the over_sum, calculate
+		 * target over_sum value.
+		 */
+		u32 over_delta = after_sum - (WEIGHT_ONE - 1);
+		WARN_ON_ONCE(over_sum <= over_delta);
+		over_target = over_sum - over_delta;
+	} else {
+		over_target = 0;
+	}
+
+	list_for_each_entry_safe(iocg, tiocg, &over_hwa, walk_list) {
+		if (over_target)
+			iocg->hweight_after_donation =
+				div_u64((u64)iocg->hweight_after_donation *
+					over_target, over_sum);
+		list_del_init(&iocg->walk_list);
+	}
+
+	/*
+	 * Build pre-order inner node walk list and prepare for donation
+	 * adjustment calculations.
+	 */
+	list_for_each_entry(iocg, surpluses, surplus_list) {
+		iocg_build_inner_walk(iocg, &inner_walk);
+	}
+
+	root_iocg = list_first_entry(&inner_walk, struct ioc_gq, walk_list);
+	WARN_ON_ONCE(root_iocg->level > 0);
+
+	list_for_each_entry(iocg, &inner_walk, walk_list) {
+		iocg->child_adjusted_sum = 0;
+		iocg->hweight_donating = 0;
+		iocg->hweight_after_donation = 0;
+	}
+
+	/*
+	 * Propagate the donating budget (b_t) and after donation budget (b'_t)
+	 * up the hierarchy.
+	 */
+	list_for_each_entry(iocg, surpluses, surplus_list) {
+		struct ioc_gq *parent = iocg->ancestors[iocg->level - 1];
+
+		parent->hweight_donating += iocg->hweight_donating;
+		parent->hweight_after_donation += iocg->hweight_after_donation;
+	}
+
+	list_for_each_entry_reverse(iocg, &inner_walk, walk_list) {
+		if (iocg->level > 0) {
+			struct ioc_gq *parent = iocg->ancestors[iocg->level - 1];
+
+			parent->hweight_donating += iocg->hweight_donating;
+			parent->hweight_after_donation += iocg->hweight_after_donation;
+		}
+	}
+
+	/*
+	 * Calculate inner hwa's (b) and make sure the donation values are
+	 * within the accepted ranges as we're doing low res calculations with
+	 * roundups.
+	 */
+	list_for_each_entry(iocg, &inner_walk, walk_list) {
+		if (iocg->level) {
+			struct ioc_gq *parent = iocg->ancestors[iocg->level - 1];
+
+			iocg->hweight_active = DIV64_U64_ROUND_UP(
+				(u64)parent->hweight_active * iocg->active,
+				parent->child_active_sum);
+
+		}
+
+		iocg->hweight_donating = min(iocg->hweight_donating,
+					     iocg->hweight_active);
+		iocg->hweight_after_donation = min(iocg->hweight_after_donation,
+						   iocg->hweight_donating - 1);
+		if (WARN_ON_ONCE(iocg->hweight_active <= 1 ||
+				 iocg->hweight_donating <= 1 ||
+				 iocg->hweight_after_donation == 0)) {
+			pr_warn("iocg: invalid donation weights in ");
+			pr_cont_cgroup_path(iocg_to_blkg(iocg)->blkcg->css.cgroup);
+			pr_cont(": active=%u donating=%u after=%u\n",
+				iocg->hweight_active, iocg->hweight_donating,
+				iocg->hweight_after_donation);
+		}
+	}
+
+	/*
+	 * Calculate the global donation rate (gamma) - the rate to adjust
+	 * non-donating budgets by.
+	 *
+	 * No need to use 64bit multiplication here as the first operand is
+	 * guaranteed to be smaller than WEIGHT_ONE (1<<16).
+	 *
+	 * We know that there are beneficiary nodes and the sum of the donating
+	 * hweights can't be whole; however, due to the round-ups during hweight
+	 * calculations, root_iocg->hweight_donating might still end up equal to
+	 * or greater than whole. Limit the range when calculating the divider.
+	 *
+	 * gamma = (1 - t_r') / (1 - t_r)
+	 */
+	gamma = DIV_ROUND_UP(
+		(WEIGHT_ONE - root_iocg->hweight_after_donation) * WEIGHT_ONE,
+		WEIGHT_ONE - min_t(u32, root_iocg->hweight_donating, WEIGHT_ONE - 1));
+
+	/*
+	 * Calculate adjusted hwi, child_adjusted_sum and inuse for the inner
+	 * nodes.
+	 */
+	list_for_each_entry(iocg, &inner_walk, walk_list) {
+		struct ioc_gq *parent;
+		u32 inuse, wpt, wptp;
+		u64 st, sf;
+
+		if (iocg->level == 0) {
+			/* adjusted weight sum for 1st level: s' = s * b_pf / b'_pf */
+			iocg->child_adjusted_sum = DIV64_U64_ROUND_UP(
+				iocg->child_active_sum * (WEIGHT_ONE - iocg->hweight_donating),
+				WEIGHT_ONE - iocg->hweight_after_donation);
+			continue;
+		}
+
+		parent = iocg->ancestors[iocg->level - 1];
+
+		/* b' = gamma * b_f + b_t' */
+		iocg->hweight_inuse = DIV64_U64_ROUND_UP(
+			(u64)gamma * (iocg->hweight_active - iocg->hweight_donating),
+			WEIGHT_ONE) + iocg->hweight_after_donation;
+
+		/* w' = s' * b' / b'_p */
+		inuse = DIV64_U64_ROUND_UP(
+			(u64)parent->child_adjusted_sum * iocg->hweight_inuse,
+			parent->hweight_inuse);
+
+		/* adjusted weight sum for children: s' = s_f + s_t * w'_pt / w_pt */
+		st = DIV64_U64_ROUND_UP(
+			iocg->child_active_sum * iocg->hweight_donating,
+			iocg->hweight_active);
+		sf = iocg->child_active_sum - st;
+		wpt = DIV64_U64_ROUND_UP(
+			(u64)iocg->active * iocg->hweight_donating,
+			iocg->hweight_active);
+		wptp = DIV64_U64_ROUND_UP(
+			(u64)inuse * iocg->hweight_after_donation,
+			iocg->hweight_inuse);
+
+		iocg->child_adjusted_sum = sf + DIV64_U64_ROUND_UP(st * wptp, wpt);
+	}
+
+	/*
+	 * All inner nodes now have ->hweight_inuse and ->child_adjusted_sum and
+	 * we can finally determine leaf adjustments.
+	 */
+	list_for_each_entry(iocg, surpluses, surplus_list) {
+		struct ioc_gq *parent = iocg->ancestors[iocg->level - 1];
+		u32 inuse;
+
+		/*
+		 * In-debt iocgs participated in the donation calculation with
+		 * the minimum target hweight_inuse. Configuring inuse
+		 * accordingly would work fine but debt handling expects
+		 * @iocg->inuse stay at the minimum and we don't wanna
+		 * interfere.
+		 */
+		if (iocg->abs_vdebt) {
+			WARN_ON_ONCE(iocg->inuse > 1);
+			continue;
+		}
+
+		/* w' = s' * b' / b'_p, note that b' == b'_t for donating leaves */
+		inuse = DIV64_U64_ROUND_UP(
+			parent->child_adjusted_sum * iocg->hweight_after_donation,
+			parent->hweight_inuse);
+
+		TRACE_IOCG_PATH(inuse_transfer, iocg, now,
+				iocg->inuse, inuse,
+				iocg->hweight_inuse,
+				iocg->hweight_after_donation);
+
+		__propagate_weights(iocg, iocg->active, inuse, true, now);
+	}
+
+	/* walk list should be dissolved after use */
+	list_for_each_entry_safe(iocg, tiocg, &inner_walk, walk_list)
+		list_del_init(&iocg->walk_list);
+}
+
+/*
+ * A low weight iocg can amass a large amount of debt, for example, when
+ * anonymous memory gets reclaimed aggressively. If the system has a lot of
+ * memory paired with a slow IO device, the debt can span multiple seconds or
+ * more. If there are no other subsequent IO issuers, the in-debt iocg may end
+ * up blocked paying its debt while the IO device is idle.
+ *
+ * The following protects against such cases. If the device has been
+ * sufficiently idle for a while, the debts are halved and delays are
+ * recalculated.
+ */
+static void ioc_forgive_debts(struct ioc *ioc, u64 usage_us_sum, int nr_debtors,
+			      struct ioc_now *now)
+{
+	struct ioc_gq *iocg;
+	u64 dur, usage_pct, nr_cycles;
+
+	/* if no debtor, reset the cycle */
+	if (!nr_debtors) {
+		ioc->dfgv_period_at = now->now;
+		ioc->dfgv_period_rem = 0;
+		ioc->dfgv_usage_us_sum = 0;
+		return;
+	}
+
+	/*
+	 * Debtors can pass through a lot of writes choking the device and we
+	 * don't want to be forgiving debts while the device is struggling from
+	 * write bursts. If we're missing latency targets, consider the device
+	 * fully utilized.
+	 */
+	if (ioc->busy_level > 0)
+		usage_us_sum = max_t(u64, usage_us_sum, ioc->period_us);
+
+	ioc->dfgv_usage_us_sum += usage_us_sum;
+	if (time_before64(now->now, ioc->dfgv_period_at + DFGV_PERIOD))
+		return;
+
+	/*
+	 * At least DFGV_PERIOD has passed since the last period. Calculate the
+	 * average usage and reset the period counters.
+	 */
+	dur = now->now - ioc->dfgv_period_at;
+	usage_pct = div64_u64(100 * ioc->dfgv_usage_us_sum, dur);
+
+	ioc->dfgv_period_at = now->now;
+	ioc->dfgv_usage_us_sum = 0;
+
+	/* if was too busy, reset everything */
+	if (usage_pct > DFGV_USAGE_PCT) {
+		ioc->dfgv_period_rem = 0;
+		return;
+	}
+
+	/*
+	 * Usage is lower than threshold. Let's forgive some debts. Debt
+	 * forgiveness runs off of the usual ioc timer but its period usually
+	 * doesn't match ioc's. Compensate the difference by performing the
+	 * reduction as many times as would fit in the duration since the last
+	 * run and carrying over the left-over duration in @ioc->dfgv_period_rem
+	 * - if ioc period is 75% of DFGV_PERIOD, one out of three consecutive
+	 * reductions is doubled.
+	 */
+	nr_cycles = dur + ioc->dfgv_period_rem;
+	ioc->dfgv_period_rem = do_div(nr_cycles, DFGV_PERIOD);
+
+	list_for_each_entry(iocg, &ioc->active_iocgs, active_list) {
+		u64 __maybe_unused old_debt, __maybe_unused old_delay;
+
+		if (!iocg->abs_vdebt && !iocg->delay)
+			continue;
+
+		spin_lock(&iocg->waitq.lock);
+
+		old_debt = iocg->abs_vdebt;
+		old_delay = iocg->delay;
+
+		if (iocg->abs_vdebt)
+			iocg->abs_vdebt = iocg->abs_vdebt >> nr_cycles ?: 1;
+		if (iocg->delay)
+			iocg->delay = iocg->delay >> nr_cycles ?: 1;
+
+		iocg_kick_waitq(iocg, true, now);
+
+		TRACE_IOCG_PATH(iocg_forgive_debt, iocg, now, usage_pct,
+				old_debt, iocg->abs_vdebt,
+				old_delay, iocg->delay);
+
+		spin_unlock(&iocg->waitq.lock);
+	}
 }
 
 static void ioc_timer_fn(struct timer_list *timer)
@@ -1341,12 +2076,14 @@ static void ioc_timer_fn(struct timer_list *timer)
 	struct ioc *ioc = container_of(timer, struct ioc, timer);
 	struct ioc_gq *iocg, *tiocg;
 	struct ioc_now now;
-	int nr_surpluses = 0, nr_shortages = 0, nr_lagging = 0;
+	LIST_HEAD(surpluses);
+	int nr_debtors = 0, nr_shortages = 0, nr_lagging = 0;
+	u64 usage_us_sum = 0;
 	u32 ppm_rthr = MILLION - ioc->params.qos[QOS_RPPM];
 	u32 ppm_wthr = MILLION - ioc->params.qos[QOS_WPPM];
 	u32 missed_ppm[2], rq_wait_pct;
 	u64 period_vtime;
-	int prev_busy_level, i;
+	int prev_busy_level;
 
 	/* how were the latencies during the period? */
 	ioc_lat_stat(ioc, missed_ppm, &rq_wait_pct);
@@ -1370,30 +2107,71 @@ static void ioc_timer_fn(struct timer_list *timer)
 	 */
 	list_for_each_entry_safe(iocg, tiocg, &ioc->active_iocgs, active_list) {
 		if (!waitqueue_active(&iocg->waitq) && !iocg->abs_vdebt &&
-		    !iocg_is_idle(iocg))
+		    !iocg->delay && !iocg_is_idle(iocg))
 			continue;
 
 		spin_lock(&iocg->waitq.lock);
 
-		if (waitqueue_active(&iocg->waitq) || iocg->abs_vdebt) {
+		/* flush wait and indebt stat deltas */
+		if (iocg->wait_since) {
+			iocg->local_stat.wait_us += now.now - iocg->wait_since;
+			iocg->wait_since = now.now;
+		}
+		if (iocg->indebt_since) {
+			iocg->local_stat.indebt_us +=
+				now.now - iocg->indebt_since;
+			iocg->indebt_since = now.now;
+		}
+		if (iocg->indelay_since) {
+			iocg->local_stat.indelay_us +=
+				now.now - iocg->indelay_since;
+			iocg->indelay_since = now.now;
+		}
+
+		if (waitqueue_active(&iocg->waitq) || iocg->abs_vdebt ||
+		    iocg->delay) {
 			/* might be oversleeping vtime / hweight changes, kick */
-			iocg_kick_waitq(iocg, &now);
-			iocg_kick_delay(iocg, &now);
+			iocg_kick_waitq(iocg, true, &now);
+			if (iocg->abs_vdebt || iocg->delay)
+				nr_debtors++;
 		} else if (iocg_is_idle(iocg)) {
 			/* no waiter and idle, deactivate */
-			iocg->last_inuse = iocg->inuse;
-			__propagate_active_weight(iocg, 0, 0);
+			u64 vtime = atomic64_read(&iocg->vtime);
+			s64 excess;
+
+			/*
+			 * @iocg has been inactive for a full duration and will
+			 * have a high budget. Account anything above target as
+			 * error and throw away. On reactivation, it'll start
+			 * with the target budget.
+			 */
+			excess = now.vnow - vtime - ioc->margins.target;
+			if (excess > 0) {
+				u32 old_hwi;
+
+				current_hweight(iocg, NULL, &old_hwi);
+				ioc->vtime_err -= div64_u64(excess * old_hwi,
+							    WEIGHT_ONE);
+			}
+
+			__propagate_weights(iocg, 0, 0, false, &now);
 			list_del_init(&iocg->active_list);
 		}
 
 		spin_unlock(&iocg->waitq.lock);
 	}
-	commit_active_weights(ioc);
+	commit_weights(ioc);
+
+	/*
+	 * Wait and indebt stat are flushed above and the donation calculation
+	 * below needs updated usage stat. Let's bring stat up-to-date.
+	 */
+	iocg_flush_stat(&ioc->active_iocgs, &now);
 
-	/* calc usages and see whether some weights need to be moved around */
+	/* calc usage and see whether some weights need to be moved around */
 	list_for_each_entry(iocg, &ioc->active_iocgs, active_list) {
-		u64 vdone, vtime, vusage, vmargin, vmin;
-		u32 hw_active, hw_inuse, usage;
+		u64 vdone, vtime, usage_us, usage_dur;
+		u32 usage, hw_active, hw_inuse;
 
 		/*
 		 * Collect unused and wind vtime closer to vnow to prevent
@@ -1417,116 +2195,73 @@ static void ioc_timer_fn(struct timer_list *timer)
 		    time_before64(vdone, now.vnow - period_vtime))
 			nr_lagging++;
 
-		if (waitqueue_active(&iocg->waitq))
-			vusage = now.vnow - iocg->last_vtime;
-		else if (time_before64(iocg->last_vtime, vtime))
-			vusage = vtime - iocg->last_vtime;
-		else
-			vusage = 0;
-
-		iocg->last_vtime += vusage;
 		/*
-		 * Factor in in-flight vtime into vusage to avoid
-		 * high-latency completions appearing as idle.  This should
-		 * be done after the above ->last_time adjustment.
+		 * Determine absolute usage factoring in in-flight IOs to avoid
+		 * high-latency completions appearing as idle.
 		 */
-		vusage = max(vusage, vtime - vdone);
-
-		/* calculate hweight based usage ratio and record */
-		if (vusage) {
-			usage = DIV64_U64_ROUND_UP(vusage * hw_inuse,
-						   period_vtime);
-			iocg->usage_idx = (iocg->usage_idx + 1) % NR_USAGE_SLOTS;
-			iocg->usages[iocg->usage_idx] = usage;
-		} else {
-			usage = 0;
+		usage_us = iocg->usage_delta_us;
+		usage_us_sum += usage_us;
+
+		if (vdone != vtime) {
+			u64 inflight_us = DIV64_U64_ROUND_UP(
+				cost_to_abs_cost(vtime - vdone, hw_inuse),
+				ioc->vtime_base_rate);
+			usage_us = max(usage_us, inflight_us);
 		}
 
+		/* convert to hweight based usage ratio */
+		if (time_after64(iocg->activated_at, ioc->period_at))
+			usage_dur = max_t(u64, now.now - iocg->activated_at, 1);
+		else
+			usage_dur = max_t(u64, now.now - ioc->period_at, 1);
+
+		usage = clamp_t(u32,
+				DIV64_U64_ROUND_UP(usage_us * WEIGHT_ONE,
+						   usage_dur),
+				1, WEIGHT_ONE);
+
 		/* see whether there's surplus vtime */
-		vmargin = ioc->margin_us * now.vrate;
-		vmin = now.vnow - vmargin;
-
-		iocg->has_surplus = false;
-
-		if (!waitqueue_active(&iocg->waitq) &&
-		    time_before64(vtime, vmin)) {
-			u64 delta = vmin - vtime;
-
-			/* throw away surplus vtime */
-			atomic64_add(delta, &iocg->vtime);
-			atomic64_add(delta, &iocg->done_vtime);
-			iocg->last_vtime += delta;
-			/* if usage is sufficiently low, maybe it can donate */
-			if (surplus_adjusted_hweight_inuse(usage, hw_inuse)) {
-				iocg->has_surplus = true;
-				nr_surpluses++;
-			}
-		} else if (hw_inuse < hw_active) {
-			u32 new_hwi, new_inuse;
+		WARN_ON_ONCE(!list_empty(&iocg->surplus_list));
+		if (hw_inuse < hw_active ||
+		    (!waitqueue_active(&iocg->waitq) &&
+		     time_before64(vtime, now.vnow - ioc->margins.low))) {
+			u32 hwa, old_hwi, hwm, new_hwi;
 
-			/* was donating but might need to take back some */
-			if (waitqueue_active(&iocg->waitq)) {
-				new_hwi = hw_active;
+			/*
+			 * Already donating or accumulated enough to start.
+			 * Determine the donation amount.
+			 */
+			current_hweight(iocg, &hwa, &old_hwi);
+			hwm = current_hweight_max(iocg);
+			new_hwi = hweight_after_donation(iocg, old_hwi, hwm,
+							 usage, &now);
+			if (new_hwi < hwm) {
+				iocg->hweight_donating = hwa;
+				iocg->hweight_after_donation = new_hwi;
+				list_add(&iocg->surplus_list, &surpluses);
 			} else {
-				new_hwi = max(hw_inuse,
-					      usage * SURPLUS_SCALE_PCT / 100 +
-					      SURPLUS_SCALE_ABS);
-			}
+				TRACE_IOCG_PATH(inuse_shortage, iocg, &now,
+						iocg->inuse, iocg->active,
+						iocg->hweight_inuse, new_hwi);
 
-			new_inuse = div64_u64((u64)iocg->inuse * new_hwi,
-					      hw_inuse);
-			new_inuse = clamp_t(u32, new_inuse, 1, iocg->active);
-
-			if (new_inuse > iocg->inuse) {
-				TRACE_IOCG_PATH(inuse_takeback, iocg, &now,
-						iocg->inuse, new_inuse,
-						hw_inuse, new_hwi);
-				__propagate_active_weight(iocg, iocg->weight,
-							  new_inuse);
+				__propagate_weights(iocg, iocg->active,
+						    iocg->active, true, &now);
+				nr_shortages++;
 			}
 		} else {
-			/* genuninely out of vtime */
+			/* genuinely short on vtime */
 			nr_shortages++;
 		}
 	}
 
-	if (!nr_shortages || !nr_surpluses)
-		goto skip_surplus_transfers;
-
-	/* there are both shortages and surpluses, transfer surpluses */
-	list_for_each_entry(iocg, &ioc->active_iocgs, active_list) {
-		u32 usage, hw_active, hw_inuse, new_hwi, new_inuse;
-		int nr_valid = 0;
-
-		if (!iocg->has_surplus)
-			continue;
+	if (!list_empty(&surpluses) && nr_shortages)
+		transfer_surpluses(&surpluses, &now);
 
-		/* base the decision on max historical usage */
-		for (i = 0, usage = 0; i < NR_USAGE_SLOTS; i++) {
-			if (iocg->usages[i]) {
-				usage = max(usage, iocg->usages[i]);
-				nr_valid++;
-			}
-		}
-		if (nr_valid < MIN_VALID_USAGES)
-			continue;
+	commit_weights(ioc);
 
-		current_hweight(iocg, &hw_active, &hw_inuse);
-		new_hwi = surplus_adjusted_hweight_inuse(usage, hw_inuse);
-		if (!new_hwi)
-			continue;
-
-		new_inuse = DIV64_U64_ROUND_UP((u64)iocg->inuse * new_hwi,
-					       hw_inuse);
-		if (new_inuse < iocg->inuse) {
-			TRACE_IOCG_PATH(inuse_giveaway, iocg, &now,
-					iocg->inuse, new_inuse,
-					hw_inuse, new_hwi);
-			__propagate_active_weight(iocg, iocg->weight, new_inuse);
-		}
-	}
-skip_surplus_transfers:
-	commit_active_weights(ioc);
+	/* surplus list should be dissolved after use */
+	list_for_each_entry_safe(iocg, tiocg, &surpluses, surplus_list)
+		list_del_init(&iocg->surplus_list);
 
 	/*
 	 * If q is getting clogged or we're missing too much, we're issuing
@@ -1554,11 +2289,9 @@ skip_surplus_transfers:
 
 			/*
 			 * If there are IOs spanning multiple periods, wait
-			 * them out before pushing the device harder.  If
-			 * there are surpluses, let redistribution work it
-			 * out first.
+			 * them out before pushing the device harder.
 			 */
-			if (!nr_lagging && !nr_surpluses)
+			if (!nr_lagging)
 				ioc->busy_level--;
 		} else {
 			/*
@@ -1577,7 +2310,7 @@ skip_surplus_transfers:
 	ioc->busy_level = clamp(ioc->busy_level, -1000, 1000);
 
 	if (ioc->busy_level > 0 || (ioc->busy_level < 0 && !nr_lagging)) {
-		u64 vrate = atomic64_read(&ioc->vtime_rate);
+		u64 vrate = ioc->vtime_base_rate;
 		u64 vrate_min = ioc->vrate_min, vrate_max = ioc->vrate_max;
 
 		/* rq_wait signal is always reliable, ignore user vrate_min */
@@ -1612,20 +2345,20 @@ skip_surplus_transfers:
 		}
 
 		trace_iocost_ioc_vrate_adj(ioc, vrate, missed_ppm, rq_wait_pct,
-					   nr_lagging, nr_shortages,
-					   nr_surpluses);
+					   nr_lagging, nr_shortages);
 
-		atomic64_set(&ioc->vtime_rate, vrate);
-		ioc->inuse_margin_vtime = DIV64_U64_ROUND_UP(
-			ioc->period_us * vrate * INUSE_MARGIN_PCT, 100);
+		ioc->vtime_base_rate = vrate;
+		ioc_refresh_margins(ioc);
 	} else if (ioc->busy_level != prev_busy_level || nr_lagging) {
 		trace_iocost_ioc_vrate_adj(ioc, atomic64_read(&ioc->vtime_rate),
 					   missed_ppm, rq_wait_pct, nr_lagging,
-					   nr_shortages, nr_surpluses);
+					   nr_shortages);
 	}
 
 	ioc_refresh_params(ioc, false);
 
+	ioc_forgive_debts(ioc, usage_us_sum, nr_debtors, &now);
+
 	/*
 	 * This period is done.  Move onto the next one.  If nothing's
 	 * going on with the device, stop the timer.
@@ -1637,13 +2370,76 @@ skip_surplus_transfers:
 			ioc_start_period(ioc, &now);
 		} else {
 			ioc->busy_level = 0;
+			ioc->vtime_err = 0;
 			ioc->running = IOC_IDLE;
 		}
+
+		ioc_refresh_vrate(ioc, &now);
 	}
 
 	spin_unlock_irq(&ioc->lock);
 }
 
+static u64 adjust_inuse_and_calc_cost(struct ioc_gq *iocg, u64 vtime,
+				      u64 abs_cost, struct ioc_now *now)
+{
+	struct ioc *ioc = iocg->ioc;
+	struct ioc_margins *margins = &ioc->margins;
+	u32 __maybe_unused old_inuse = iocg->inuse, __maybe_unused old_hwi;
+	u32 hwi, adj_step;
+	s64 margin;
+	u64 cost, new_inuse;
+
+	current_hweight(iocg, NULL, &hwi);
+	old_hwi = hwi;
+	cost = abs_cost_to_cost(abs_cost, hwi);
+	margin = now->vnow - vtime - cost;
+
+	/* debt handling owns inuse for debtors */
+	if (iocg->abs_vdebt)
+		return cost;
+
+	/*
+	 * We only increase inuse during period and do so iff the margin has
+	 * deteriorated since the previous adjustment.
+	 */
+	if (margin >= iocg->saved_margin || margin >= margins->low ||
+	    iocg->inuse == iocg->active)
+		return cost;
+
+	spin_lock_irq(&ioc->lock);
+
+	/* we own inuse only when @iocg is in the normal active state */
+	if (iocg->abs_vdebt || list_empty(&iocg->active_list)) {
+		spin_unlock_irq(&ioc->lock);
+		return cost;
+	}
+
+	/*
+	 * Bump up inuse till @abs_cost fits in the existing budget.
+	 * adj_step must be determined after acquiring ioc->lock - we might
+	 * have raced and lost to another thread for activation and could
+	 * be reading 0 iocg->active before ioc->lock which will lead to
+	 * infinite loop.
+	 */
+	new_inuse = iocg->inuse;
+	adj_step = DIV_ROUND_UP(iocg->active * INUSE_ADJ_STEP_PCT, 100);
+	do {
+		new_inuse = new_inuse + adj_step;
+		propagate_weights(iocg, iocg->active, new_inuse, true, now);
+		current_hweight(iocg, NULL, &hwi);
+		cost = abs_cost_to_cost(abs_cost, hwi);
+	} while (time_after64(vtime + cost, now->vnow) &&
+		 iocg->inuse != iocg->active);
+
+	spin_unlock_irq(&ioc->lock);
+
+	TRACE_IOCG_PATH(inuse_adjust, iocg, now,
+			old_inuse, iocg->inuse, old_hwi, hwi);
+
+	return cost;
+}
+
 static void calc_vtime_cost_builtin(struct bio *bio, struct ioc_gq *iocg,
 				    bool is_merge, u64 *costp)
 {
@@ -1725,38 +2521,25 @@ static void ioc_rqos_throttle(struct rq_qos *rqos, struct bio *bio)
 	struct ioc_gq *iocg = blkg_to_iocg(blkg);
 	struct ioc_now now;
 	struct iocg_wait wait;
-	u32 hw_active, hw_inuse;
 	u64 abs_cost, cost, vtime;
+	bool use_debt, ioc_locked;
+	unsigned long flags;
 
 	/* bypass IOs if disabled or for root cgroup */
 	if (!ioc->enabled || !iocg->level)
 		return;
 
-	/* always activate so that even 0 cost IOs get protected to some level */
-	if (!iocg_activate(iocg, &now))
-		return;
-
 	/* calculate the absolute vtime cost */
 	abs_cost = calc_vtime_cost(bio, iocg, false);
 	if (!abs_cost)
 		return;
 
-	iocg->cursor = bio_end_sector(bio);
+	if (!iocg_activate(iocg, &now))
+		return;
 
+	iocg->cursor = bio_end_sector(bio);
 	vtime = atomic64_read(&iocg->vtime);
-	current_hweight(iocg, &hw_active, &hw_inuse);
-
-	if (hw_inuse < hw_active &&
-	    time_after_eq64(vtime + ioc->inuse_margin_vtime, now.vnow)) {
-		TRACE_IOCG_PATH(inuse_reset, iocg, &now,
-				iocg->inuse, iocg->weight, hw_inuse, hw_active);
-		spin_lock_irq(&ioc->lock);
-		propagate_active_weight(iocg, iocg->weight, iocg->weight);
-		spin_unlock_irq(&ioc->lock);
-		current_hweight(iocg, &hw_active, &hw_inuse);
-	}
-
-	cost = abs_cost_to_cost(abs_cost, hw_inuse);
+	cost = adjust_inuse_and_calc_cost(iocg, vtime, abs_cost, &now);
 
 	/*
 	 * If no one's waiting and within budget, issue right away.  The
@@ -1765,21 +2548,32 @@ static void ioc_rqos_throttle(struct rq_qos *rqos, struct bio *bio)
 	 */
 	if (!waitqueue_active(&iocg->waitq) && !iocg->abs_vdebt &&
 	    time_before_eq64(vtime + cost, now.vnow)) {
-		iocg_commit_bio(iocg, bio, cost);
+		iocg_commit_bio(iocg, bio, abs_cost, cost);
 		return;
 	}
 
 	/*
-	 * We activated above but w/o any synchronization. Deactivation is
-	 * synchronized with waitq.lock and we won't get deactivated as long
-	 * as we're waiting or has debt, so we're good if we're activated
-	 * here. In the unlikely case that we aren't, just issue the IO.
+	 * We're over budget. This can be handled in two ways. IOs which may
+	 * cause priority inversions are punted to @ioc->aux_iocg and charged as
+	 * debt. Otherwise, the issuer is blocked on @iocg->waitq. Debt handling
+	 * requires @ioc->lock, waitq handling @iocg->waitq.lock. Determine
+	 * whether debt handling is needed and acquire locks accordingly.
 	 */
-	spin_lock_irq(&iocg->waitq.lock);
+	use_debt = bio_issue_as_root_blkg(bio) || fatal_signal_pending(current);
+	ioc_locked = use_debt || READ_ONCE(iocg->abs_vdebt);
+retry_lock:
+	iocg_lock(iocg, ioc_locked, &flags);
 
+	/*
+	 * @iocg must stay activated for debt and waitq handling. Deactivation
+	 * is synchronized against both ioc->lock and waitq.lock and we won't
+	 * get deactivated as long as we're waiting or has debt, so we're good
+	 * if we're activated here. In the unlikely cases that we aren't, just
+	 * issue the IO.
+	 */
 	if (unlikely(list_empty(&iocg->active_list))) {
-		spin_unlock_irq(&iocg->waitq.lock);
-		iocg_commit_bio(iocg, bio, cost);
+		iocg_unlock(iocg, ioc_locked, &flags);
+		iocg_commit_bio(iocg, bio, abs_cost, cost);
 		return;
 	}
 
@@ -1800,15 +2594,26 @@ static void ioc_rqos_throttle(struct rq_qos *rqos, struct bio *bio)
 	 * clear them and leave @iocg inactive w/ dangling use_delay heavily
 	 * penalizing the cgroup and its descendants.
 	 */
-	if (bio_issue_as_root_blkg(bio) || fatal_signal_pending(current)) {
-		iocg->abs_vdebt += abs_cost;
+	if (use_debt) {
+		iocg_incur_debt(iocg, abs_cost, &now);
 		if (iocg_kick_delay(iocg, &now))
 			blkcg_schedule_throttle(rqos->q,
 					(bio->bi_opf & REQ_SWAP) == REQ_SWAP);
-		spin_unlock_irq(&iocg->waitq.lock);
+		iocg_unlock(iocg, ioc_locked, &flags);
 		return;
 	}
 
+	/* guarantee that iocgs w/ waiters have maximum inuse */
+	if (!iocg->abs_vdebt && iocg->inuse != iocg->active) {
+		if (!ioc_locked) {
+			iocg_unlock(iocg, false, &flags);
+			ioc_locked = true;
+			goto retry_lock;
+		}
+		propagate_weights(iocg, iocg->active, iocg->active, true,
+				  &now);
+	}
+
 	/*
 	 * Append self to the waitq and schedule the wakeup timer if we're
 	 * the first waiter.  The timer duration is calculated based on the
@@ -1829,9 +2634,9 @@ static void ioc_rqos_throttle(struct rq_qos *rqos, struct bio *bio)
 	wait.committed = false;	/* will be set true by waker */
 
 	__add_wait_queue_entry_tail(&iocg->waitq, &wait.wait);
-	iocg_kick_waitq(iocg, &now);
+	iocg_kick_waitq(iocg, ioc_locked, &now);
 
-	spin_unlock_irq(&iocg->waitq.lock);
+	iocg_unlock(iocg, ioc_locked, &flags);
 
 	while (true) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
@@ -1851,8 +2656,7 @@ static void ioc_rqos_merge(struct rq_qos *rqos, struct request *rq,
 	struct ioc *ioc = iocg->ioc;
 	sector_t bio_end = bio_end_sector(bio);
 	struct ioc_now now;
-	u32 hw_inuse;
-	u64 abs_cost, cost;
+	u64 vtime, abs_cost, cost;
 	unsigned long flags;
 
 	/* bypass if disabled or for root cgroup */
@@ -1864,8 +2668,9 @@ static void ioc_rqos_merge(struct rq_qos *rqos, struct request *rq,
 		return;
 
 	ioc_now(ioc, &now);
-	current_hweight(iocg, NULL, &hw_inuse);
-	cost = abs_cost_to_cost(abs_cost, hw_inuse);
+
+	vtime = atomic64_read(&iocg->vtime);
+	cost = adjust_inuse_and_calc_cost(iocg, vtime, abs_cost, &now);
 
 	/* update cursor if backmerging into the request at the cursor */
 	if (blk_rq_pos(rq) < bio_end &&
@@ -1878,7 +2683,7 @@ static void ioc_rqos_merge(struct rq_qos *rqos, struct request *rq,
 	 */
 	if (rq->bio && rq->bio->bi_iocost_cost &&
 	    time_before_eq64(atomic64_read(&iocg->vtime) + cost, now.vnow)) {
-		iocg_commit_bio(iocg, bio, cost);
+		iocg_commit_bio(iocg, bio, abs_cost, cost);
 		return;
 	}
 
@@ -1887,14 +2692,20 @@ static void ioc_rqos_merge(struct rq_qos *rqos, struct request *rq,
 	 * be for the vast majority of cases. See debt handling in
 	 * ioc_rqos_throttle() for details.
 	 */
-	spin_lock_irqsave(&iocg->waitq.lock, flags);
+	spin_lock_irqsave(&ioc->lock, flags);
+	spin_lock(&iocg->waitq.lock);
+
 	if (likely(!list_empty(&iocg->active_list))) {
-		iocg->abs_vdebt += abs_cost;
-		iocg_kick_delay(iocg, &now);
+		iocg_incur_debt(iocg, abs_cost, &now);
+		if (iocg_kick_delay(iocg, &now))
+			blkcg_schedule_throttle(rqos->q,
+					(bio->bi_opf & REQ_SWAP) == REQ_SWAP);
 	} else {
-		iocg_commit_bio(iocg, bio, cost);
+		iocg_commit_bio(iocg, bio, abs_cost, cost);
 	}
-	spin_unlock_irqrestore(&iocg->waitq.lock, flags);
+
+	spin_unlock(&iocg->waitq.lock);
+	spin_unlock_irqrestore(&ioc->lock, flags);
 }
 
 static void ioc_rqos_done_bio(struct rq_qos *rqos, struct bio *bio)
@@ -1908,6 +2719,7 @@ static void ioc_rqos_done_bio(struct rq_qos *rqos, struct bio *bio)
 static void ioc_rqos_done(struct rq_qos *rqos, struct request *rq)
 {
 	struct ioc *ioc = rqos_to_ioc(rqos);
+	struct ioc_pcpu_stat *ccs;
 	u64 on_q_ns, rq_wait_ns, size_nsec;
 	int pidx, rw;
 
@@ -1931,13 +2743,17 @@ static void ioc_rqos_done(struct rq_qos *rqos, struct request *rq)
 	rq_wait_ns = rq->start_time_ns - rq->alloc_time_ns;
 	size_nsec = div64_u64(calc_size_vtime_cost(rq, ioc), VTIME_PER_NSEC);
 
+	ccs = get_cpu_ptr(ioc->pcpu_stat);
+
 	if (on_q_ns <= size_nsec ||
 	    on_q_ns - size_nsec <= ioc->params.qos[pidx] * NSEC_PER_USEC)
-		this_cpu_inc(ioc->pcpu_stat->missed[rw].nr_met);
+		local_inc(&ccs->missed[rw].nr_met);
 	else
-		this_cpu_inc(ioc->pcpu_stat->missed[rw].nr_missed);
+		local_inc(&ccs->missed[rw].nr_missed);
 
-	this_cpu_add(ioc->pcpu_stat->rq_wait_ns, rq_wait_ns);
+	local64_add(rq_wait_ns, &ccs->rq_wait_ns);
+
+	put_cpu_ptr(ccs);
 }
 
 static void ioc_rqos_queue_depth_changed(struct rq_qos *rqos)
@@ -1977,7 +2793,7 @@ static int blk_iocost_init(struct request_queue *q)
 {
 	struct ioc *ioc;
 	struct rq_qos *rqos;
-	int ret;
+	int i, cpu, ret;
 
 	ioc = kzalloc(sizeof(*ioc), GFP_KERNEL);
 	if (!ioc)
@@ -1989,6 +2805,16 @@ static int blk_iocost_init(struct request_queue *q)
 		return -ENOMEM;
 	}
 
+	for_each_possible_cpu(cpu) {
+		struct ioc_pcpu_stat *ccs = per_cpu_ptr(ioc->pcpu_stat, cpu);
+
+		for (i = 0; i < ARRAY_SIZE(ccs->missed); i++) {
+			local_set(&ccs->missed[i].nr_met, 0);
+			local_set(&ccs->missed[i].nr_missed, 0);
+		}
+		local64_set(&ccs->rq_wait_ns, 0);
+	}
+
 	rqos = &ioc->rqos;
 	rqos->id = RQ_QOS_COST;
 	rqos->ops = &ioc_rqos_ops;
@@ -1999,6 +2825,7 @@ static int blk_iocost_init(struct request_queue *q)
 	INIT_LIST_HEAD(&ioc->active_iocgs);
 
 	ioc->running = IOC_IDLE;
+	ioc->vtime_base_rate = VTIME_PER_USEC;
 	atomic64_set(&ioc->vtime_rate, VTIME_PER_USEC);
 	seqcount_spinlock_init(&ioc->period_seqcount, &ioc->lock);
 	ioc->period_at = ktime_to_us(ktime_get());
@@ -2029,7 +2856,7 @@ static struct blkcg_policy_data *ioc_cpd_alloc(gfp_t gfp)
 	if (!iocc)
 		return NULL;
 
-	iocc->dfl_weight = CGROUP_WEIGHT_DFL;
+	iocc->dfl_weight = CGROUP_WEIGHT_DFL * WEIGHT_ONE;
 	return &iocc->cpd;
 }
 
@@ -2048,6 +2875,12 @@ static struct blkg_policy_data *ioc_pd_alloc(gfp_t gfp, struct request_queue *q,
 	if (!iocg)
 		return NULL;
 
+	iocg->pcpu_stat = alloc_percpu_gfp(struct iocg_pcpu_stat, gfp);
+	if (!iocg->pcpu_stat) {
+		kfree(iocg);
+		return NULL;
+	}
+
 	return &iocg->pd;
 }
 
@@ -2067,14 +2900,14 @@ static void ioc_pd_init(struct blkg_policy_data *pd)
 	atomic64_set(&iocg->done_vtime, now.vnow);
 	atomic64_set(&iocg->active_period, atomic64_read(&ioc->cur_period));
 	INIT_LIST_HEAD(&iocg->active_list);
-	iocg->hweight_active = HWEIGHT_WHOLE;
-	iocg->hweight_inuse = HWEIGHT_WHOLE;
+	INIT_LIST_HEAD(&iocg->walk_list);
+	INIT_LIST_HEAD(&iocg->surplus_list);
+	iocg->hweight_active = WEIGHT_ONE;
+	iocg->hweight_inuse = WEIGHT_ONE;
 
 	init_waitqueue_head(&iocg->waitq);
 	hrtimer_init(&iocg->waitq_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
 	iocg->waitq_timer.function = iocg_waitq_timer_fn;
-	hrtimer_init(&iocg->delay_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
-	iocg->delay_timer.function = iocg_delay_timer_fn;
 
 	iocg->level = blkg->blkcg->css.cgroup->level;
 
@@ -2084,7 +2917,7 @@ static void ioc_pd_init(struct blkg_policy_data *pd)
 	}
 
 	spin_lock_irqsave(&ioc->lock, flags);
-	weight_updated(iocg);
+	weight_updated(iocg, &now);
 	spin_unlock_irqrestore(&ioc->lock, flags);
 }
 
@@ -2096,18 +2929,56 @@ static void ioc_pd_free(struct blkg_policy_data *pd)
 
 	if (ioc) {
 		spin_lock_irqsave(&ioc->lock, flags);
+
 		if (!list_empty(&iocg->active_list)) {
-			propagate_active_weight(iocg, 0, 0);
+			struct ioc_now now;
+
+			ioc_now(ioc, &now);
+			propagate_weights(iocg, 0, 0, false, &now);
 			list_del_init(&iocg->active_list);
 		}
+
+		WARN_ON_ONCE(!list_empty(&iocg->walk_list));
+		WARN_ON_ONCE(!list_empty(&iocg->surplus_list));
+
 		spin_unlock_irqrestore(&ioc->lock, flags);
 
 		hrtimer_cancel(&iocg->waitq_timer);
-		hrtimer_cancel(&iocg->delay_timer);
 	}
+	free_percpu(iocg->pcpu_stat);
 	kfree(iocg);
 }
 
+static size_t ioc_pd_stat(struct blkg_policy_data *pd, char *buf, size_t size)
+{
+	struct ioc_gq *iocg = pd_to_iocg(pd);
+	struct ioc *ioc = iocg->ioc;
+	size_t pos = 0;
+
+	if (!ioc->enabled)
+		return 0;
+
+	if (iocg->level == 0) {
+		unsigned vp10k = DIV64_U64_ROUND_CLOSEST(
+			ioc->vtime_base_rate * 10000,
+			VTIME_PER_USEC);
+		pos += scnprintf(buf + pos, size - pos, " cost.vrate=%u.%02u",
+				  vp10k / 100, vp10k % 100);
+	}
+
+	pos += scnprintf(buf + pos, size - pos, " cost.usage=%llu",
+			 iocg->last_stat.usage_us);
+
+	if (blkcg_debug_stats)
+		pos += scnprintf(buf + pos, size - pos,
+				 " cost.wait=%llu cost.indebt=%llu cost.indelay=%llu",
+				 iocg->last_stat.wait_us,
+				 iocg->last_stat.indebt_us,
+				 iocg->last_stat.indelay_us);
+
+	return pos;
+}
+
 static u64 ioc_weight_prfill(struct seq_file *sf, struct blkg_policy_data *pd,
 			     int off)
 {
@@ -2115,7 +2986,7 @@ static u64 ioc_weight_prfill(struct seq_file *sf, struct blkg_policy_data *pd,
 	struct ioc_gq *iocg = pd_to_iocg(pd);
 
 	if (dname && iocg->cfg_weight)
-		seq_printf(sf, "%s %u\n", dname, iocg->cfg_weight);
+		seq_printf(sf, "%s %u\n", dname, iocg->cfg_weight / WEIGHT_ONE);
 	return 0;
 }
 
@@ -2125,7 +2996,7 @@ static int ioc_weight_show(struct seq_file *sf, void *v)
 	struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
 	struct ioc_cgrp *iocc = blkcg_to_iocc(blkcg);
 
-	seq_printf(sf, "default %u\n", iocc->dfl_weight);
+	seq_printf(sf, "default %u\n", iocc->dfl_weight / WEIGHT_ONE);
 	blkcg_print_blkgs(sf, blkcg, ioc_weight_prfill,
 			  &blkcg_policy_iocost, seq_cft(sf)->private, false);
 	return 0;
@@ -2137,6 +3008,7 @@ static ssize_t ioc_weight_write(struct kernfs_open_file *of, char *buf,
 	struct blkcg *blkcg = css_to_blkcg(of_css(of));
 	struct ioc_cgrp *iocc = blkcg_to_iocc(blkcg);
 	struct blkg_conf_ctx ctx;
+	struct ioc_now now;
 	struct ioc_gq *iocg;
 	u32 v;
 	int ret;
@@ -2151,13 +3023,14 @@ static ssize_t ioc_weight_write(struct kernfs_open_file *of, char *buf,
 			return -EINVAL;
 
 		spin_lock(&blkcg->lock);
-		iocc->dfl_weight = v;
+		iocc->dfl_weight = v * WEIGHT_ONE;
 		hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
 			struct ioc_gq *iocg = blkg_to_iocg(blkg);
 
 			if (iocg) {
 				spin_lock_irq(&iocg->ioc->lock);
-				weight_updated(iocg);
+				ioc_now(iocg->ioc, &now);
+				weight_updated(iocg, &now);
 				spin_unlock_irq(&iocg->ioc->lock);
 			}
 		}
@@ -2182,8 +3055,9 @@ static ssize_t ioc_weight_write(struct kernfs_open_file *of, char *buf,
 	}
 
 	spin_lock(&iocg->ioc->lock);
-	iocg->cfg_weight = v;
-	weight_updated(iocg);
+	iocg->cfg_weight = v * WEIGHT_ONE;
+	ioc_now(iocg->ioc, &now);
+	weight_updated(iocg, &now);
 	spin_unlock(&iocg->ioc->lock);
 
 	blkg_conf_finish(&ctx);
@@ -2521,6 +3395,7 @@ static struct blkcg_policy blkcg_policy_iocost = {
 	.pd_alloc_fn	= ioc_pd_alloc,
 	.pd_init_fn	= ioc_pd_init,
 	.pd_free_fn	= ioc_pd_free,
+	.pd_stat_fn	= ioc_pd_stat,
 };
 
 static int __init ioc_init(void)
diff --git a/block/blk-lib.c b/block/blk-lib.c
index 0d1811e57ac7..e90614fd8d6a 100644
--- a/block/blk-lib.c
+++ b/block/blk-lib.c
@@ -64,7 +64,7 @@ int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
 		return -EINVAL;
 
 	/* In case the discard request is in a partition */
-	if (bdev->bd_partno)
+	if (bdev_is_partition(bdev))
 		part_offset = bdev->bd_part->start_sect;
 
 	while (nr_sects) {
diff --git a/block/blk-map.c b/block/blk-map.c
index 6e804892d5ec..21630dccac62 100644
--- a/block/blk-map.c
+++ b/block/blk-map.c
@@ -12,7 +12,8 @@
 #include "blk.h"
 
 struct bio_map_data {
-	int is_our_pages;
+	bool is_our_pages : 1;
+	bool is_null_mapped : 1;
 	struct iov_iter iter;
 	struct iovec iov[];
 };
@@ -108,7 +109,7 @@ static int bio_uncopy_user(struct bio *bio)
 	struct bio_map_data *bmd = bio->bi_private;
 	int ret = 0;
 
-	if (!bio_flagged(bio, BIO_NULL_MAPPED)) {
+	if (!bmd->is_null_mapped) {
 		/*
 		 * if we're in a workqueue, the request is orphaned, so
 		 * don't copy into a random user address space, just free
@@ -126,24 +127,12 @@ static int bio_uncopy_user(struct bio *bio)
 	return ret;
 }
 
-/**
- *	bio_copy_user_iov	-	copy user data to bio
- *	@q:		destination block queue
- *	@map_data:	pointer to the rq_map_data holding pages (if necessary)
- *	@iter:		iovec iterator
- *	@gfp_mask:	memory allocation flags
- *
- *	Prepares and returns a bio for indirect user io, bouncing data
- *	to/from kernel pages as necessary. Must be paired with
- *	call bio_uncopy_user() on io completion.
- */
-static struct bio *bio_copy_user_iov(struct request_queue *q,
-		struct rq_map_data *map_data, struct iov_iter *iter,
-		gfp_t gfp_mask)
+static int bio_copy_user_iov(struct request *rq, struct rq_map_data *map_data,
+		struct iov_iter *iter, gfp_t gfp_mask)
 {
 	struct bio_map_data *bmd;
 	struct page *page;
-	struct bio *bio;
+	struct bio *bio, *bounce_bio;
 	int i = 0, ret;
 	int nr_pages;
 	unsigned int len = iter->count;
@@ -151,14 +140,15 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 
 	bmd = bio_alloc_map_data(iter, gfp_mask);
 	if (!bmd)
-		return ERR_PTR(-ENOMEM);
+		return -ENOMEM;
 
 	/*
 	 * We need to do a deep copy of the iov_iter including the iovecs.
 	 * The caller provided iov might point to an on-stack or otherwise
 	 * shortlived one.
 	 */
-	bmd->is_our_pages = map_data ? 0 : 1;
+	bmd->is_our_pages = !map_data;
+	bmd->is_null_mapped = (map_data && map_data->null_mapped);
 
 	nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE);
 	if (nr_pages > BIO_MAX_PAGES)
@@ -168,8 +158,7 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 	bio = bio_kmalloc(gfp_mask, nr_pages);
 	if (!bio)
 		goto out_bmd;
-
-	ret = 0;
+	bio->bi_opf |= req_op(rq);
 
 	if (map_data) {
 		nr_pages = 1 << map_data->page_order;
@@ -186,7 +175,7 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 		if (map_data) {
 			if (i == map_data->nr_entries * nr_pages) {
 				ret = -ENOMEM;
-				break;
+				goto cleanup;
 			}
 
 			page = map_data->pages[i / nr_pages];
@@ -194,14 +183,14 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 
 			i++;
 		} else {
-			page = alloc_page(q->bounce_gfp | gfp_mask);
+			page = alloc_page(rq->q->bounce_gfp | gfp_mask);
 			if (!page) {
 				ret = -ENOMEM;
-				break;
+				goto cleanup;
 			}
 		}
 
-		if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) {
+		if (bio_add_pc_page(rq->q, bio, page, bytes, offset) < bytes) {
 			if (!map_data)
 				__free_page(page);
 			break;
@@ -211,9 +200,6 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 		offset = 0;
 	}
 
-	if (ret)
-		goto cleanup;
-
 	if (map_data)
 		map_data->offset += bio->bi_iter.bi_size;
 
@@ -233,41 +219,42 @@ static struct bio *bio_copy_user_iov(struct request_queue *q,
 	}
 
 	bio->bi_private = bmd;
-	if (map_data && map_data->null_mapped)
-		bio_set_flag(bio, BIO_NULL_MAPPED);
-	return bio;
+
+	bounce_bio = bio;
+	ret = blk_rq_append_bio(rq, &bounce_bio);
+	if (ret)
+		goto cleanup;
+
+	/*
+	 * We link the bounce buffer in and could have to traverse it later, so
+	 * we have to get a ref to prevent it from being freed
+	 */
+	bio_get(bounce_bio);
+	return 0;
 cleanup:
 	if (!map_data)
 		bio_free_pages(bio);
 	bio_put(bio);
 out_bmd:
 	kfree(bmd);
-	return ERR_PTR(ret);
+	return ret;
 }
 
-/**
- *	bio_map_user_iov - map user iovec into bio
- *	@q:		the struct request_queue for the bio
- *	@iter:		iovec iterator
- *	@gfp_mask:	memory allocation flags
- *
- *	Map the user space address into a bio suitable for io to a block
- *	device. Returns an error pointer in case of error.
- */
-static struct bio *bio_map_user_iov(struct request_queue *q,
-		struct iov_iter *iter, gfp_t gfp_mask)
+static int bio_map_user_iov(struct request *rq, struct iov_iter *iter,
+		gfp_t gfp_mask)
 {
-	unsigned int max_sectors = queue_max_hw_sectors(q);
-	int j;
-	struct bio *bio;
+	unsigned int max_sectors = queue_max_hw_sectors(rq->q);
+	struct bio *bio, *bounce_bio;
 	int ret;
+	int j;
 
 	if (!iov_iter_count(iter))
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES));
 	if (!bio)
-		return ERR_PTR(-ENOMEM);
+		return -ENOMEM;
+	bio->bi_opf |= req_op(rq);
 
 	while (iov_iter_count(iter)) {
 		struct page **pages;
@@ -283,7 +270,7 @@ static struct bio *bio_map_user_iov(struct request_queue *q,
 
 		npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE);
 
-		if (unlikely(offs & queue_dma_alignment(q))) {
+		if (unlikely(offs & queue_dma_alignment(rq->q))) {
 			ret = -EINVAL;
 			j = 0;
 		} else {
@@ -295,7 +282,7 @@ static struct bio *bio_map_user_iov(struct request_queue *q,
 				if (n > bytes)
 					n = bytes;
 
-				if (!bio_add_hw_page(q, bio, page, n, offs,
+				if (!bio_add_hw_page(rq->q, bio, page, n, offs,
 						     max_sectors, &same_page)) {
 					if (same_page)
 						put_page(page);
@@ -319,21 +306,31 @@ static struct bio *bio_map_user_iov(struct request_queue *q,
 			break;
 	}
 
-	bio_set_flag(bio, BIO_USER_MAPPED);
-
 	/*
-	 * subtle -- if bio_map_user_iov() ended up bouncing a bio,
-	 * it would normally disappear when its bi_end_io is run.
-	 * however, we need it for the unmap, so grab an extra
-	 * reference to it
+	 * Subtle: if we end up needing to bounce a bio, it would normally
+	 * disappear when its bi_end_io is run.  However, we need the original
+	 * bio for the unmap, so grab an extra reference to it
 	 */
 	bio_get(bio);
-	return bio;
 
+	bounce_bio = bio;
+	ret = blk_rq_append_bio(rq, &bounce_bio);
+	if (ret)
+		goto out_put_orig;
+
+	/*
+	 * We link the bounce buffer in and could have to traverse it
+	 * later, so we have to get a ref to prevent it from being freed
+	 */
+	bio_get(bounce_bio);
+	return 0;
+
+ out_put_orig:
+	bio_put(bio);
  out_unmap:
 	bio_release_pages(bio, false);
 	bio_put(bio);
-	return ERR_PTR(ret);
+	return ret;
 }
 
 /**
@@ -557,55 +554,6 @@ int blk_rq_append_bio(struct request *rq, struct bio **bio)
 }
 EXPORT_SYMBOL(blk_rq_append_bio);
 
-static int __blk_rq_unmap_user(struct bio *bio)
-{
-	int ret = 0;
-
-	if (bio) {
-		if (bio_flagged(bio, BIO_USER_MAPPED))
-			bio_unmap_user(bio);
-		else
-			ret = bio_uncopy_user(bio);
-	}
-
-	return ret;
-}
-
-static int __blk_rq_map_user_iov(struct request *rq,
-		struct rq_map_data *map_data, struct iov_iter *iter,
-		gfp_t gfp_mask, bool copy)
-{
-	struct request_queue *q = rq->q;
-	struct bio *bio, *orig_bio;
-	int ret;
-
-	if (copy)
-		bio = bio_copy_user_iov(q, map_data, iter, gfp_mask);
-	else
-		bio = bio_map_user_iov(q, iter, gfp_mask);
-
-	if (IS_ERR(bio))
-		return PTR_ERR(bio);
-
-	bio->bi_opf &= ~REQ_OP_MASK;
-	bio->bi_opf |= req_op(rq);
-
-	orig_bio = bio;
-
-	/*
-	 * We link the bounce buffer in and could have to traverse it
-	 * later so we have to get a ref to prevent it from being freed
-	 */
-	ret = blk_rq_append_bio(rq, &bio);
-	if (ret) {
-		__blk_rq_unmap_user(orig_bio);
-		return ret;
-	}
-	bio_get(bio);
-
-	return 0;
-}
-
 /**
  * blk_rq_map_user_iov - map user data to a request, for passthrough requests
  * @q:		request queue where request should be inserted
@@ -649,7 +597,10 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq,
 
 	i = *iter;
 	do {
-		ret =__blk_rq_map_user_iov(rq, map_data, &i, gfp_mask, copy);
+		if (copy)
+			ret = bio_copy_user_iov(rq, map_data, &i, gfp_mask);
+		else
+			ret = bio_map_user_iov(rq, &i, gfp_mask);
 		if (ret)
 			goto unmap_rq;
 		if (!bio)
@@ -700,9 +651,13 @@ int blk_rq_unmap_user(struct bio *bio)
 		if (unlikely(bio_flagged(bio, BIO_BOUNCED)))
 			mapped_bio = bio->bi_private;
 
-		ret2 = __blk_rq_unmap_user(mapped_bio);
-		if (ret2 && !ret)
-			ret = ret2;
+		if (bio->bi_private) {
+			ret2 = bio_uncopy_user(mapped_bio);
+			if (ret2 && !ret)
+				ret = ret2;
+		} else {
+			bio_unmap_user(mapped_bio);
+		}
 
 		mapped_bio = bio;
 		bio = bio->bi_next;
diff --git a/block/blk-merge.c b/block/blk-merge.c
index f685d633bcc9..6ed715835d45 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -11,6 +11,7 @@
 #include <trace/events/block.h>
 
 #include "blk.h"
+#include "blk-rq-qos.h"
 
 static inline bool bio_will_gap(struct request_queue *q,
 		struct request *prev_rq, struct bio *prev, struct bio *next)
@@ -895,3 +896,203 @@ enum elv_merge blk_try_merge(struct request *rq, struct bio *bio)
 		return ELEVATOR_FRONT_MERGE;
 	return ELEVATOR_NO_MERGE;
 }
+
+static void blk_account_io_merge_bio(struct request *req)
+{
+	if (!blk_do_io_stat(req))
+		return;
+
+	part_stat_lock();
+	part_stat_inc(req->part, merges[op_stat_group(req_op(req))]);
+	part_stat_unlock();
+}
+
+enum bio_merge_status bio_attempt_back_merge(struct request *req,
+					     struct bio *bio,
+					     unsigned int nr_segs)
+{
+	const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
+
+	if (!ll_back_merge_fn(req, bio, nr_segs))
+		return BIO_MERGE_FAILED;
+
+	trace_block_bio_backmerge(req->q, req, bio);
+	rq_qos_merge(req->q, req, bio);
+
+	if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
+		blk_rq_set_mixed_merge(req);
+
+	req->biotail->bi_next = bio;
+	req->biotail = bio;
+	req->__data_len += bio->bi_iter.bi_size;
+
+	bio_crypt_free_ctx(bio);
+
+	blk_account_io_merge_bio(req);
+	return BIO_MERGE_OK;
+}
+
+enum bio_merge_status bio_attempt_front_merge(struct request *req,
+					      struct bio *bio,
+					      unsigned int nr_segs)
+{
+	const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
+
+	if (!ll_front_merge_fn(req, bio, nr_segs))
+		return BIO_MERGE_FAILED;
+
+	trace_block_bio_frontmerge(req->q, req, bio);
+	rq_qos_merge(req->q, req, bio);
+
+	if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
+		blk_rq_set_mixed_merge(req);
+
+	bio->bi_next = req->bio;
+	req->bio = bio;
+
+	req->__sector = bio->bi_iter.bi_sector;
+	req->__data_len += bio->bi_iter.bi_size;
+
+	bio_crypt_do_front_merge(req, bio);
+
+	blk_account_io_merge_bio(req);
+	return BIO_MERGE_OK;
+}
+
+enum bio_merge_status bio_attempt_discard_merge(struct request_queue *q,
+						struct request *req,
+						struct bio *bio)
+{
+	unsigned short segments = blk_rq_nr_discard_segments(req);
+
+	if (segments >= queue_max_discard_segments(q))
+		goto no_merge;
+	if (blk_rq_sectors(req) + bio_sectors(bio) >
+	    blk_rq_get_max_sectors(req, blk_rq_pos(req)))
+		goto no_merge;
+
+	rq_qos_merge(q, req, bio);
+
+	req->biotail->bi_next = bio;
+	req->biotail = bio;
+	req->__data_len += bio->bi_iter.bi_size;
+	req->nr_phys_segments = segments + 1;
+
+	blk_account_io_merge_bio(req);
+	return BIO_MERGE_OK;
+no_merge:
+	req_set_nomerge(q, req);
+	return BIO_MERGE_FAILED;
+}
+
+static enum bio_merge_status blk_attempt_bio_merge(struct request_queue *q,
+						   struct request *rq,
+						   struct bio *bio,
+						   unsigned int nr_segs,
+						   bool sched_allow_merge)
+{
+	if (!blk_rq_merge_ok(rq, bio))
+		return BIO_MERGE_NONE;
+
+	switch (blk_try_merge(rq, bio)) {
+	case ELEVATOR_BACK_MERGE:
+		if (!sched_allow_merge || blk_mq_sched_allow_merge(q, rq, bio))
+			return bio_attempt_back_merge(rq, bio, nr_segs);
+		break;
+	case ELEVATOR_FRONT_MERGE:
+		if (!sched_allow_merge || blk_mq_sched_allow_merge(q, rq, bio))
+			return bio_attempt_front_merge(rq, bio, nr_segs);
+		break;
+	case ELEVATOR_DISCARD_MERGE:
+		return bio_attempt_discard_merge(q, rq, bio);
+	default:
+		return BIO_MERGE_NONE;
+	}
+
+	return BIO_MERGE_FAILED;
+}
+
+/**
+ * blk_attempt_plug_merge - try to merge with %current's plugged list
+ * @q: request_queue new bio is being queued at
+ * @bio: new bio being queued
+ * @nr_segs: number of segments in @bio
+ * @same_queue_rq: pointer to &struct request that gets filled in when
+ * another request associated with @q is found on the plug list
+ * (optional, may be %NULL)
+ *
+ * Determine whether @bio being queued on @q can be merged with a request
+ * on %current's plugged list.  Returns %true if merge was successful,
+ * otherwise %false.
+ *
+ * Plugging coalesces IOs from the same issuer for the same purpose without
+ * going through @q->queue_lock.  As such it's more of an issuing mechanism
+ * than scheduling, and the request, while may have elvpriv data, is not
+ * added on the elevator at this point.  In addition, we don't have
+ * reliable access to the elevator outside queue lock.  Only check basic
+ * merging parameters without querying the elevator.
+ *
+ * Caller must ensure !blk_queue_nomerges(q) beforehand.
+ */
+bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
+		unsigned int nr_segs, struct request **same_queue_rq)
+{
+	struct blk_plug *plug;
+	struct request *rq;
+	struct list_head *plug_list;
+
+	plug = blk_mq_plug(q, bio);
+	if (!plug)
+		return false;
+
+	plug_list = &plug->mq_list;
+
+	list_for_each_entry_reverse(rq, plug_list, queuelist) {
+		if (rq->q == q && same_queue_rq) {
+			/*
+			 * Only blk-mq multiple hardware queues case checks the
+			 * rq in the same queue, there should be only one such
+			 * rq in a queue
+			 **/
+			*same_queue_rq = rq;
+		}
+
+		if (rq->q != q)
+			continue;
+
+		if (blk_attempt_bio_merge(q, rq, bio, nr_segs, false) ==
+		    BIO_MERGE_OK)
+			return true;
+	}
+
+	return false;
+}
+
+/*
+ * Iterate list of requests and see if we can merge this bio with any
+ * of them.
+ */
+bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
+			struct bio *bio, unsigned int nr_segs)
+{
+	struct request *rq;
+	int checked = 8;
+
+	list_for_each_entry_reverse(rq, list, queuelist) {
+		if (!checked--)
+			break;
+
+		switch (blk_attempt_bio_merge(q, rq, bio, nr_segs, true)) {
+		case BIO_MERGE_NONE:
+			continue;
+		case BIO_MERGE_OK:
+			return true;
+		case BIO_MERGE_FAILED:
+			return false;
+		}
+
+	}
+
+	return false;
+}
+EXPORT_SYMBOL_GPL(blk_bio_list_merge);
diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 3f09bcb8a6fd..3094542e12ae 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -116,6 +116,7 @@ static const char *const blk_queue_flag_name[] = {
 	QUEUE_FLAG_NAME(SAME_FORCE),
 	QUEUE_FLAG_NAME(DEAD),
 	QUEUE_FLAG_NAME(INIT_DONE),
+	QUEUE_FLAG_NAME(STABLE_WRITES),
 	QUEUE_FLAG_NAME(POLL),
 	QUEUE_FLAG_NAME(WC),
 	QUEUE_FLAG_NAME(FUA),
@@ -240,7 +241,7 @@ static const char *const alloc_policy_name[] = {
 #define HCTX_FLAG_NAME(name) [ilog2(BLK_MQ_F_##name)] = #name
 static const char *const hctx_flag_name[] = {
 	HCTX_FLAG_NAME(SHOULD_MERGE),
-	HCTX_FLAG_NAME(TAG_SHARED),
+	HCTX_FLAG_NAME(TAG_QUEUE_SHARED),
 	HCTX_FLAG_NAME(BLOCKING),
 	HCTX_FLAG_NAME(NO_SCHED),
 	HCTX_FLAG_NAME(STACKING),
@@ -452,11 +453,11 @@ static void blk_mq_debugfs_tags_show(struct seq_file *m,
 		   atomic_read(&tags->active_queues));
 
 	seq_puts(m, "\nbitmap_tags:\n");
-	sbitmap_queue_show(&tags->bitmap_tags, m);
+	sbitmap_queue_show(tags->bitmap_tags, m);
 
 	if (tags->nr_reserved_tags) {
 		seq_puts(m, "\nbreserved_tags:\n");
-		sbitmap_queue_show(&tags->breserved_tags, m);
+		sbitmap_queue_show(tags->breserved_tags, m);
 	}
 }
 
@@ -487,7 +488,7 @@ static int hctx_tags_bitmap_show(void *data, struct seq_file *m)
 	if (res)
 		goto out;
 	if (hctx->tags)
-		sbitmap_bitmap_show(&hctx->tags->bitmap_tags.sb, m);
+		sbitmap_bitmap_show(&hctx->tags->bitmap_tags->sb, m);
 	mutex_unlock(&q->sysfs_lock);
 
 out:
@@ -521,7 +522,7 @@ static int hctx_sched_tags_bitmap_show(void *data, struct seq_file *m)
 	if (res)
 		goto out;
 	if (hctx->sched_tags)
-		sbitmap_bitmap_show(&hctx->sched_tags->bitmap_tags.sb, m);
+		sbitmap_bitmap_show(&hctx->sched_tags->bitmap_tags->sb, m);
 	mutex_unlock(&q->sysfs_lock);
 
 out:
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index d2790e5b06d1..3e9596738852 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -18,21 +18,6 @@
 #include "blk-mq-tag.h"
 #include "blk-wbt.h"
 
-void blk_mq_sched_free_hctx_data(struct request_queue *q,
-				 void (*exit)(struct blk_mq_hw_ctx *))
-{
-	struct blk_mq_hw_ctx *hctx;
-	int i;
-
-	queue_for_each_hw_ctx(q, hctx, i) {
-		if (exit && hctx->sched_data)
-			exit(hctx);
-		kfree(hctx->sched_data);
-		hctx->sched_data = NULL;
-	}
-}
-EXPORT_SYMBOL_GPL(blk_mq_sched_free_hctx_data);
-
 void blk_mq_sched_assign_ioc(struct request *rq)
 {
 	struct request_queue *q = rq->q;
@@ -368,7 +353,7 @@ bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
 	case ELEVATOR_BACK_MERGE:
 		if (!blk_mq_sched_allow_merge(q, rq, bio))
 			return false;
-		if (!bio_attempt_back_merge(rq, bio, nr_segs))
+		if (bio_attempt_back_merge(rq, bio, nr_segs) != BIO_MERGE_OK)
 			return false;
 		*merged_request = attempt_back_merge(q, rq);
 		if (!*merged_request)
@@ -377,86 +362,20 @@ bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
 	case ELEVATOR_FRONT_MERGE:
 		if (!blk_mq_sched_allow_merge(q, rq, bio))
 			return false;
-		if (!bio_attempt_front_merge(rq, bio, nr_segs))
+		if (bio_attempt_front_merge(rq, bio, nr_segs) != BIO_MERGE_OK)
 			return false;
 		*merged_request = attempt_front_merge(q, rq);
 		if (!*merged_request)
 			elv_merged_request(q, rq, ELEVATOR_FRONT_MERGE);
 		return true;
 	case ELEVATOR_DISCARD_MERGE:
-		return bio_attempt_discard_merge(q, rq, bio);
+		return bio_attempt_discard_merge(q, rq, bio) == BIO_MERGE_OK;
 	default:
 		return false;
 	}
 }
 EXPORT_SYMBOL_GPL(blk_mq_sched_try_merge);
 
-/*
- * Iterate list of requests and see if we can merge this bio with any
- * of them.
- */
-bool blk_mq_bio_list_merge(struct request_queue *q, struct list_head *list,
-			   struct bio *bio, unsigned int nr_segs)
-{
-	struct request *rq;
-	int checked = 8;
-
-	list_for_each_entry_reverse(rq, list, queuelist) {
-		bool merged = false;
-
-		if (!checked--)
-			break;
-
-		if (!blk_rq_merge_ok(rq, bio))
-			continue;
-
-		switch (blk_try_merge(rq, bio)) {
-		case ELEVATOR_BACK_MERGE:
-			if (blk_mq_sched_allow_merge(q, rq, bio))
-				merged = bio_attempt_back_merge(rq, bio,
-						nr_segs);
-			break;
-		case ELEVATOR_FRONT_MERGE:
-			if (blk_mq_sched_allow_merge(q, rq, bio))
-				merged = bio_attempt_front_merge(rq, bio,
-						nr_segs);
-			break;
-		case ELEVATOR_DISCARD_MERGE:
-			merged = bio_attempt_discard_merge(q, rq, bio);
-			break;
-		default:
-			continue;
-		}
-
-		return merged;
-	}
-
-	return false;
-}
-EXPORT_SYMBOL_GPL(blk_mq_bio_list_merge);
-
-/*
- * Reverse check our software queue for entries that we could potentially
- * merge with. Currently includes a hand-wavy stop count of 8, to not spend
- * too much time checking for merges.
- */
-static bool blk_mq_attempt_merge(struct request_queue *q,
-				 struct blk_mq_hw_ctx *hctx,
-				 struct blk_mq_ctx *ctx, struct bio *bio,
-				 unsigned int nr_segs)
-{
-	enum hctx_type type = hctx->type;
-
-	lockdep_assert_held(&ctx->lock);
-
-	if (blk_mq_bio_list_merge(q, &ctx->rq_lists[type], bio, nr_segs)) {
-		ctx->rq_merged++;
-		return true;
-	}
-
-	return false;
-}
-
 bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio,
 		unsigned int nr_segs)
 {
@@ -470,14 +389,24 @@ bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio,
 		return e->type->ops.bio_merge(hctx, bio, nr_segs);
 
 	type = hctx->type;
-	if ((hctx->flags & BLK_MQ_F_SHOULD_MERGE) &&
-			!list_empty_careful(&ctx->rq_lists[type])) {
-		/* default per sw-queue merge */
-		spin_lock(&ctx->lock);
-		ret = blk_mq_attempt_merge(q, hctx, ctx, bio, nr_segs);
-		spin_unlock(&ctx->lock);
+	if (!(hctx->flags & BLK_MQ_F_SHOULD_MERGE) ||
+	    list_empty_careful(&ctx->rq_lists[type]))
+		return false;
+
+	/* default per sw-queue merge */
+	spin_lock(&ctx->lock);
+	/*
+	 * Reverse check our software queue for entries that we could
+	 * potentially merge with. Currently includes a hand-wavy stop
+	 * count of 8, to not spend too much time checking for merges.
+	 */
+	if (blk_bio_list_merge(q, &ctx->rq_lists[type], bio, nr_segs)) {
+		ctx->rq_merged++;
+		ret = true;
 	}
 
+	spin_unlock(&ctx->lock);
+
 	return ret;
 }
 
@@ -531,7 +460,7 @@ void blk_mq_sched_insert_request(struct request *rq, bool at_head,
 		goto run;
 	}
 
-	WARN_ON(e && (rq->tag != -1));
+	WARN_ON(e && (rq->tag != BLK_MQ_NO_TAG));
 
 	if (blk_mq_sched_bypass_insert(hctx, !!e, rq)) {
 		/*
@@ -616,9 +545,11 @@ static void blk_mq_sched_free_tags(struct blk_mq_tag_set *set,
 				   struct blk_mq_hw_ctx *hctx,
 				   unsigned int hctx_idx)
 {
+	unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
+
 	if (hctx->sched_tags) {
 		blk_mq_free_rqs(set, hctx->sched_tags, hctx_idx);
-		blk_mq_free_rq_map(hctx->sched_tags);
+		blk_mq_free_rq_map(hctx->sched_tags, flags);
 		hctx->sched_tags = NULL;
 	}
 }
@@ -628,10 +559,12 @@ static int blk_mq_sched_alloc_tags(struct request_queue *q,
 				   unsigned int hctx_idx)
 {
 	struct blk_mq_tag_set *set = q->tag_set;
+	/* Clear HCTX_SHARED so tags are init'ed */
+	unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
 	int ret;
 
 	hctx->sched_tags = blk_mq_alloc_rq_map(set, hctx_idx, q->nr_requests,
-					       set->reserved_tags);
+					       set->reserved_tags, flags);
 	if (!hctx->sched_tags)
 		return -ENOMEM;
 
@@ -649,8 +582,11 @@ static void blk_mq_sched_tags_teardown(struct request_queue *q)
 	int i;
 
 	queue_for_each_hw_ctx(q, hctx, i) {
+		/* Clear HCTX_SHARED so tags are freed */
+		unsigned int flags = hctx->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
+
 		if (hctx->sched_tags) {
-			blk_mq_free_rq_map(hctx->sched_tags);
+			blk_mq_free_rq_map(hctx->sched_tags, flags);
 			hctx->sched_tags = NULL;
 		}
 	}
diff --git a/block/blk-mq-sched.h b/block/blk-mq-sched.h
index e81ca1bf6e10..0476360f05f1 100644
--- a/block/blk-mq-sched.h
+++ b/block/blk-mq-sched.h
@@ -5,9 +5,6 @@
 #include "blk-mq.h"
 #include "blk-mq-tag.h"
 
-void blk_mq_sched_free_hctx_data(struct request_queue *q,
-				 void (*exit)(struct blk_mq_hw_ctx *));
-
 void blk_mq_sched_assign_ioc(struct request *rq);
 
 void blk_mq_sched_request_inserted(struct request *rq);
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 32d82e23b095..9c92053e704d 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -23,9 +23,18 @@
  */
 bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
 {
-	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
-	    !test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
-		atomic_inc(&hctx->tags->active_queues);
+	if (blk_mq_is_sbitmap_shared(hctx->flags)) {
+		struct request_queue *q = hctx->queue;
+		struct blk_mq_tag_set *set = q->tag_set;
+
+		if (!test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags) &&
+		    !test_and_set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
+			atomic_inc(&set->active_queues_shared_sbitmap);
+	} else {
+		if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
+		    !test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
+			atomic_inc(&hctx->tags->active_queues);
+	}
 
 	return true;
 }
@@ -35,9 +44,9 @@ bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
  */
 void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
 {
-	sbitmap_queue_wake_all(&tags->bitmap_tags);
+	sbitmap_queue_wake_all(tags->bitmap_tags);
 	if (include_reserve)
-		sbitmap_queue_wake_all(&tags->breserved_tags);
+		sbitmap_queue_wake_all(tags->breserved_tags);
 }
 
 /*
@@ -47,11 +56,19 @@ void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
 void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
 {
 	struct blk_mq_tags *tags = hctx->tags;
-
-	if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
-		return;
-
-	atomic_dec(&tags->active_queues);
+	struct request_queue *q = hctx->queue;
+	struct blk_mq_tag_set *set = q->tag_set;
+
+	if (blk_mq_is_sbitmap_shared(hctx->flags)) {
+		if (!test_and_clear_bit(QUEUE_FLAG_HCTX_ACTIVE,
+					&q->queue_flags))
+			return;
+		atomic_dec(&set->active_queues_shared_sbitmap);
+	} else {
+		if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
+			return;
+		atomic_dec(&tags->active_queues);
+	}
 
 	blk_mq_tag_wakeup_all(tags, false);
 }
@@ -59,7 +76,8 @@ void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
 static int __blk_mq_get_tag(struct blk_mq_alloc_data *data,
 			    struct sbitmap_queue *bt)
 {
-	if (!data->q->elevator && !hctx_may_queue(data->hctx, bt))
+	if (!data->q->elevator && !(data->flags & BLK_MQ_REQ_RESERVED) &&
+			!hctx_may_queue(data->hctx, bt))
 		return BLK_MQ_NO_TAG;
 
 	if (data->shallow_depth)
@@ -82,10 +100,10 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
 			WARN_ON_ONCE(1);
 			return BLK_MQ_NO_TAG;
 		}
-		bt = &tags->breserved_tags;
+		bt = tags->breserved_tags;
 		tag_offset = 0;
 	} else {
-		bt = &tags->bitmap_tags;
+		bt = tags->bitmap_tags;
 		tag_offset = tags->nr_reserved_tags;
 	}
 
@@ -131,9 +149,9 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
 						data->ctx);
 		tags = blk_mq_tags_from_data(data);
 		if (data->flags & BLK_MQ_REQ_RESERVED)
-			bt = &tags->breserved_tags;
+			bt = tags->breserved_tags;
 		else
-			bt = &tags->bitmap_tags;
+			bt = tags->bitmap_tags;
 
 		/*
 		 * If destination hw queue is changed, fake wake up on
@@ -167,10 +185,10 @@ void blk_mq_put_tag(struct blk_mq_tags *tags, struct blk_mq_ctx *ctx,
 		const int real_tag = tag - tags->nr_reserved_tags;
 
 		BUG_ON(real_tag >= tags->nr_tags);
-		sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
+		sbitmap_queue_clear(tags->bitmap_tags, real_tag, ctx->cpu);
 	} else {
 		BUG_ON(tag >= tags->nr_reserved_tags);
-		sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
+		sbitmap_queue_clear(tags->breserved_tags, tag, ctx->cpu);
 	}
 }
 
@@ -197,7 +215,7 @@ 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 assigning ->rqs[].
 	 */
-	if (rq && rq->q == hctx->queue)
+	if (rq && rq->q == hctx->queue && rq->mq_hctx == hctx)
 		return iter_data->fn(hctx, rq, iter_data->data, reserved);
 	return true;
 }
@@ -298,9 +316,9 @@ static void __blk_mq_all_tag_iter(struct blk_mq_tags *tags,
 	WARN_ON_ONCE(flags & BT_TAG_ITER_RESERVED);
 
 	if (tags->nr_reserved_tags)
-		bt_tags_for_each(tags, &tags->breserved_tags, fn, priv,
+		bt_tags_for_each(tags, tags->breserved_tags, fn, priv,
 				 flags | BT_TAG_ITER_RESERVED);
-	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, flags);
+	bt_tags_for_each(tags, tags->bitmap_tags, fn, priv, flags);
 }
 
 /**
@@ -398,9 +416,7 @@ void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 	/*
 	 * __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.
+	 * racing with it.
 	 */
 	if (!percpu_ref_tryget(&q->q_usage_counter))
 		return;
@@ -416,8 +432,8 @@ void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 			continue;
 
 		if (tags->nr_reserved_tags)
-			bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
-		bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
+			bt_for_each(hctx, tags->breserved_tags, fn, priv, true);
+		bt_for_each(hctx, tags->bitmap_tags, fn, priv, false);
 	}
 	blk_queue_exit(q);
 }
@@ -429,30 +445,64 @@ static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
 				       node);
 }
 
-static struct blk_mq_tags *blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
-						   int node, int alloc_policy)
+static int blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
+				   int node, int alloc_policy)
 {
 	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
 	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
 
-	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
-		goto free_tags;
-	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
-		     node))
+	if (bt_alloc(&tags->__bitmap_tags, depth, round_robin, node))
+		return -ENOMEM;
+	if (bt_alloc(&tags->__breserved_tags, tags->nr_reserved_tags,
+		     round_robin, node))
 		goto free_bitmap_tags;
 
-	return tags;
+	tags->bitmap_tags = &tags->__bitmap_tags;
+	tags->breserved_tags = &tags->__breserved_tags;
+
+	return 0;
 free_bitmap_tags:
-	sbitmap_queue_free(&tags->bitmap_tags);
-free_tags:
-	kfree(tags);
-	return NULL;
+	sbitmap_queue_free(&tags->__bitmap_tags);
+	return -ENOMEM;
+}
+
+int blk_mq_init_shared_sbitmap(struct blk_mq_tag_set *set, unsigned int flags)
+{
+	unsigned int depth = set->queue_depth - set->reserved_tags;
+	int alloc_policy = BLK_MQ_FLAG_TO_ALLOC_POLICY(set->flags);
+	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
+	int i, node = set->numa_node;
+
+	if (bt_alloc(&set->__bitmap_tags, depth, round_robin, node))
+		return -ENOMEM;
+	if (bt_alloc(&set->__breserved_tags, set->reserved_tags,
+		     round_robin, node))
+		goto free_bitmap_tags;
+
+	for (i = 0; i < set->nr_hw_queues; i++) {
+		struct blk_mq_tags *tags = set->tags[i];
+
+		tags->bitmap_tags = &set->__bitmap_tags;
+		tags->breserved_tags = &set->__breserved_tags;
+	}
+
+	return 0;
+free_bitmap_tags:
+	sbitmap_queue_free(&set->__bitmap_tags);
+	return -ENOMEM;
+}
+
+void blk_mq_exit_shared_sbitmap(struct blk_mq_tag_set *set)
+{
+	sbitmap_queue_free(&set->__bitmap_tags);
+	sbitmap_queue_free(&set->__breserved_tags);
 }
 
 struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
 				     unsigned int reserved_tags,
-				     int node, int alloc_policy)
+				     int node, unsigned int flags)
 {
+	int alloc_policy = BLK_MQ_FLAG_TO_ALLOC_POLICY(flags);
 	struct blk_mq_tags *tags;
 
 	if (total_tags > BLK_MQ_TAG_MAX) {
@@ -467,13 +517,22 @@ struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
 	tags->nr_tags = total_tags;
 	tags->nr_reserved_tags = reserved_tags;
 
-	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
+	if (flags & BLK_MQ_F_TAG_HCTX_SHARED)
+		return tags;
+
+	if (blk_mq_init_bitmap_tags(tags, node, alloc_policy) < 0) {
+		kfree(tags);
+		return NULL;
+	}
+	return tags;
 }
 
-void blk_mq_free_tags(struct blk_mq_tags *tags)
+void blk_mq_free_tags(struct blk_mq_tags *tags, unsigned int flags)
 {
-	sbitmap_queue_free(&tags->bitmap_tags);
-	sbitmap_queue_free(&tags->breserved_tags);
+	if (!(flags & BLK_MQ_F_TAG_HCTX_SHARED)) {
+		sbitmap_queue_free(tags->bitmap_tags);
+		sbitmap_queue_free(tags->breserved_tags);
+	}
 	kfree(tags);
 }
 
@@ -492,6 +551,8 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
 	 */
 	if (tdepth > tags->nr_tags) {
 		struct blk_mq_tag_set *set = hctx->queue->tag_set;
+		/* Only sched tags can grow, so clear HCTX_SHARED flag  */
+		unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
 		struct blk_mq_tags *new;
 		bool ret;
 
@@ -506,30 +567,35 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
 			return -EINVAL;
 
 		new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth,
-				tags->nr_reserved_tags);
+				tags->nr_reserved_tags, flags);
 		if (!new)
 			return -ENOMEM;
 		ret = blk_mq_alloc_rqs(set, new, hctx->queue_num, tdepth);
 		if (ret) {
-			blk_mq_free_rq_map(new);
+			blk_mq_free_rq_map(new, flags);
 			return -ENOMEM;
 		}
 
 		blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
-		blk_mq_free_rq_map(*tagsptr);
+		blk_mq_free_rq_map(*tagsptr, flags);
 		*tagsptr = new;
 	} else {
 		/*
 		 * Don't need (or can't) update reserved tags here, they
 		 * remain static and should never need resizing.
 		 */
-		sbitmap_queue_resize(&tags->bitmap_tags,
+		sbitmap_queue_resize(tags->bitmap_tags,
 				tdepth - tags->nr_reserved_tags);
 	}
 
 	return 0;
 }
 
+void blk_mq_tag_resize_shared_sbitmap(struct blk_mq_tag_set *set, unsigned int size)
+{
+	sbitmap_queue_resize(&set->__bitmap_tags, size - set->reserved_tags);
+}
+
 /**
  * blk_mq_unique_tag() - return a tag that is unique queue-wide
  * @rq: request for which to compute a unique tag
diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
index b1acac518c4e..7d3e6b333a4a 100644
--- a/block/blk-mq-tag.h
+++ b/block/blk-mq-tag.h
@@ -2,8 +2,6 @@
 #ifndef INT_BLK_MQ_TAG_H
 #define INT_BLK_MQ_TAG_H
 
-#include "blk-mq.h"
-
 /*
  * Tag address space map.
  */
@@ -13,17 +11,25 @@ struct blk_mq_tags {
 
 	atomic_t active_queues;
 
-	struct sbitmap_queue bitmap_tags;
-	struct sbitmap_queue breserved_tags;
+	struct sbitmap_queue *bitmap_tags;
+	struct sbitmap_queue *breserved_tags;
+
+	struct sbitmap_queue __bitmap_tags;
+	struct sbitmap_queue __breserved_tags;
 
 	struct request **rqs;
 	struct request **static_rqs;
 	struct list_head page_list;
 };
 
+extern struct blk_mq_tags *blk_mq_init_tags(unsigned int nr_tags,
+					unsigned int reserved_tags,
+					int node, unsigned int flags);
+extern void blk_mq_free_tags(struct blk_mq_tags *tags, unsigned int flags);
 
-extern struct blk_mq_tags *blk_mq_init_tags(unsigned int nr_tags, unsigned int reserved_tags, int node, int alloc_policy);
-extern void blk_mq_free_tags(struct blk_mq_tags *tags);
+extern int blk_mq_init_shared_sbitmap(struct blk_mq_tag_set *set,
+				      unsigned int flags);
+extern void blk_mq_exit_shared_sbitmap(struct blk_mq_tag_set *set);
 
 extern unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data);
 extern void blk_mq_put_tag(struct blk_mq_tags *tags, struct blk_mq_ctx *ctx,
@@ -31,6 +37,9 @@ extern void blk_mq_put_tag(struct blk_mq_tags *tags, struct blk_mq_ctx *ctx,
 extern int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
 					struct blk_mq_tags **tags,
 					unsigned int depth, bool can_grow);
+extern void blk_mq_tag_resize_shared_sbitmap(struct blk_mq_tag_set *set,
+					     unsigned int size);
+
 extern void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool);
 void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 		void *priv);
@@ -56,7 +65,7 @@ extern void __blk_mq_tag_idle(struct blk_mq_hw_ctx *);
 
 static inline bool blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
 {
-	if (!(hctx->flags & BLK_MQ_F_TAG_SHARED))
+	if (!(hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED))
 		return false;
 
 	return __blk_mq_tag_busy(hctx);
@@ -64,43 +73,12 @@ static inline bool blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
 
 static inline void blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
 {
-	if (!(hctx->flags & BLK_MQ_F_TAG_SHARED))
+	if (!(hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED))
 		return;
 
 	__blk_mq_tag_idle(hctx);
 }
 
-/*
- * For shared tag users, we track the number of currently active users
- * and attempt to provide a fair share of the tag depth for each of them.
- */
-static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx,
-				  struct sbitmap_queue *bt)
-{
-	unsigned int depth, users;
-
-	if (!hctx || !(hctx->flags & BLK_MQ_F_TAG_SHARED))
-		return true;
-	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
-		return true;
-
-	/*
-	 * Don't try dividing an ant
-	 */
-	if (bt->sb.depth == 1)
-		return true;
-
-	users = atomic_read(&hctx->tags->active_queues);
-	if (!users)
-		return true;
-
-	/*
-	 * Allow at least some tags
-	 */
-	depth = max((bt->sb.depth + users - 1) / users, 4U);
-	return atomic_read(&hctx->nr_active) < depth;
-}
-
 static inline bool blk_mq_tag_is_reserved(struct blk_mq_tags *tags,
 					  unsigned int tag)
 {
diff --git a/block/blk-mq.c b/block/blk-mq.c
index b3d2785eefe9..176698b2285b 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -519,7 +519,7 @@ void blk_mq_free_request(struct request *rq)
 
 	ctx->rq_completed[rq_is_sync(rq)]++;
 	if (rq->rq_flags & RQF_MQ_INFLIGHT)
-		atomic_dec(&hctx->nr_active);
+		__blk_mq_dec_active_requests(hctx);
 
 	if (unlikely(laptop_mode && !blk_rq_is_passthrough(rq)))
 		laptop_io_completion(q->backing_dev_info);
@@ -1096,19 +1096,20 @@ static inline unsigned int queued_to_index(unsigned int queued)
 
 static bool __blk_mq_get_driver_tag(struct request *rq)
 {
-	struct sbitmap_queue *bt = &rq->mq_hctx->tags->bitmap_tags;
+	struct sbitmap_queue *bt = rq->mq_hctx->tags->bitmap_tags;
 	unsigned int tag_offset = rq->mq_hctx->tags->nr_reserved_tags;
 	int tag;
 
 	blk_mq_tag_busy(rq->mq_hctx);
 
 	if (blk_mq_tag_is_reserved(rq->mq_hctx->sched_tags, rq->internal_tag)) {
-		bt = &rq->mq_hctx->tags->breserved_tags;
+		bt = rq->mq_hctx->tags->breserved_tags;
 		tag_offset = 0;
+	} else {
+		if (!hctx_may_queue(rq->mq_hctx, bt))
+			return false;
 	}
 
-	if (!hctx_may_queue(rq->mq_hctx, bt))
-		return false;
 	tag = __sbitmap_queue_get(bt);
 	if (tag == BLK_MQ_NO_TAG)
 		return false;
@@ -1124,10 +1125,10 @@ static bool blk_mq_get_driver_tag(struct request *rq)
 	if (rq->tag == BLK_MQ_NO_TAG && !__blk_mq_get_driver_tag(rq))
 		return false;
 
-	if ((hctx->flags & BLK_MQ_F_TAG_SHARED) &&
+	if ((hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED) &&
 			!(rq->rq_flags & RQF_MQ_INFLIGHT)) {
 		rq->rq_flags |= RQF_MQ_INFLIGHT;
-		atomic_inc(&hctx->nr_active);
+		__blk_mq_inc_active_requests(hctx);
 	}
 	hctx->tags->rqs[rq->tag] = rq;
 	return true;
@@ -1145,7 +1146,7 @@ static int blk_mq_dispatch_wake(wait_queue_entry_t *wait, unsigned mode,
 		struct sbitmap_queue *sbq;
 
 		list_del_init(&wait->entry);
-		sbq = &hctx->tags->bitmap_tags;
+		sbq = hctx->tags->bitmap_tags;
 		atomic_dec(&sbq->ws_active);
 	}
 	spin_unlock(&hctx->dispatch_wait_lock);
@@ -1163,12 +1164,12 @@ static int blk_mq_dispatch_wake(wait_queue_entry_t *wait, unsigned mode,
 static bool blk_mq_mark_tag_wait(struct blk_mq_hw_ctx *hctx,
 				 struct request *rq)
 {
-	struct sbitmap_queue *sbq = &hctx->tags->bitmap_tags;
+	struct sbitmap_queue *sbq = hctx->tags->bitmap_tags;
 	struct wait_queue_head *wq;
 	wait_queue_entry_t *wait;
 	bool ret;
 
-	if (!(hctx->flags & BLK_MQ_F_TAG_SHARED)) {
+	if (!(hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED)) {
 		blk_mq_sched_mark_restart_hctx(hctx);
 
 		/*
@@ -1420,7 +1421,7 @@ out:
 		bool needs_restart;
 		/* For non-shared tags, the RESTART check will suffice */
 		bool no_tag = prep == PREP_DISPATCH_NO_TAG &&
-                        (hctx->flags & BLK_MQ_F_TAG_SHARED);
+			(hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED);
 		bool no_budget_avail = prep == PREP_DISPATCH_NO_BUDGET;
 
 		blk_mq_release_budgets(q, nr_budgets);
@@ -2296,20 +2297,21 @@ void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 	}
 }
 
-void blk_mq_free_rq_map(struct blk_mq_tags *tags)
+void blk_mq_free_rq_map(struct blk_mq_tags *tags, unsigned int flags)
 {
 	kfree(tags->rqs);
 	tags->rqs = NULL;
 	kfree(tags->static_rqs);
 	tags->static_rqs = NULL;
 
-	blk_mq_free_tags(tags);
+	blk_mq_free_tags(tags, flags);
 }
 
 struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 					unsigned int hctx_idx,
 					unsigned int nr_tags,
-					unsigned int reserved_tags)
+					unsigned int reserved_tags,
+					unsigned int flags)
 {
 	struct blk_mq_tags *tags;
 	int node;
@@ -2318,8 +2320,7 @@ struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 	if (node == NUMA_NO_NODE)
 		node = set->numa_node;
 
-	tags = blk_mq_init_tags(nr_tags, reserved_tags, node,
-				BLK_MQ_FLAG_TO_ALLOC_POLICY(set->flags));
+	tags = blk_mq_init_tags(nr_tags, reserved_tags, node, flags);
 	if (!tags)
 		return NULL;
 
@@ -2327,7 +2328,7 @@ struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 				 GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY,
 				 node);
 	if (!tags->rqs) {
-		blk_mq_free_tags(tags);
+		blk_mq_free_tags(tags, flags);
 		return NULL;
 	}
 
@@ -2336,7 +2337,7 @@ struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 					node);
 	if (!tags->static_rqs) {
 		kfree(tags->rqs);
-		blk_mq_free_tags(tags);
+		blk_mq_free_tags(tags, flags);
 		return NULL;
 	}
 
@@ -2660,6 +2661,7 @@ blk_mq_alloc_hctx(struct request_queue *q, struct blk_mq_tag_set *set,
 		goto free_hctx;
 
 	atomic_set(&hctx->nr_active, 0);
+	atomic_set(&hctx->elevator_queued, 0);
 	if (node == NUMA_NO_NODE)
 		node = set->numa_node;
 	hctx->numa_node = node;
@@ -2668,7 +2670,7 @@ blk_mq_alloc_hctx(struct request_queue *q, struct blk_mq_tag_set *set,
 	spin_lock_init(&hctx->lock);
 	INIT_LIST_HEAD(&hctx->dispatch);
 	hctx->queue = q;
-	hctx->flags = set->flags & ~BLK_MQ_F_TAG_SHARED;
+	hctx->flags = set->flags & ~BLK_MQ_F_TAG_QUEUE_SHARED;
 
 	INIT_LIST_HEAD(&hctx->hctx_list);
 
@@ -2745,10 +2747,11 @@ static void blk_mq_init_cpu_queues(struct request_queue *q,
 static bool __blk_mq_alloc_map_and_request(struct blk_mq_tag_set *set,
 					int hctx_idx)
 {
+	unsigned int flags = set->flags;
 	int ret = 0;
 
 	set->tags[hctx_idx] = blk_mq_alloc_rq_map(set, hctx_idx,
-					set->queue_depth, set->reserved_tags);
+					set->queue_depth, set->reserved_tags, flags);
 	if (!set->tags[hctx_idx])
 		return false;
 
@@ -2757,7 +2760,7 @@ static bool __blk_mq_alloc_map_and_request(struct blk_mq_tag_set *set,
 	if (!ret)
 		return true;
 
-	blk_mq_free_rq_map(set->tags[hctx_idx]);
+	blk_mq_free_rq_map(set->tags[hctx_idx], flags);
 	set->tags[hctx_idx] = NULL;
 	return false;
 }
@@ -2765,9 +2768,11 @@ static bool __blk_mq_alloc_map_and_request(struct blk_mq_tag_set *set,
 static void blk_mq_free_map_and_requests(struct blk_mq_tag_set *set,
 					 unsigned int hctx_idx)
 {
+	unsigned int flags = set->flags;
+
 	if (set->tags && set->tags[hctx_idx]) {
 		blk_mq_free_rqs(set, set->tags[hctx_idx], hctx_idx);
-		blk_mq_free_rq_map(set->tags[hctx_idx]);
+		blk_mq_free_rq_map(set->tags[hctx_idx], flags);
 		set->tags[hctx_idx] = NULL;
 	}
 }
@@ -2885,14 +2890,14 @@ static void queue_set_hctx_shared(struct request_queue *q, bool shared)
 
 	queue_for_each_hw_ctx(q, hctx, i) {
 		if (shared)
-			hctx->flags |= BLK_MQ_F_TAG_SHARED;
+			hctx->flags |= BLK_MQ_F_TAG_QUEUE_SHARED;
 		else
-			hctx->flags &= ~BLK_MQ_F_TAG_SHARED;
+			hctx->flags &= ~BLK_MQ_F_TAG_QUEUE_SHARED;
 	}
 }
 
-static void blk_mq_update_tag_set_depth(struct blk_mq_tag_set *set,
-					bool shared)
+static void blk_mq_update_tag_set_shared(struct blk_mq_tag_set *set,
+					 bool shared)
 {
 	struct request_queue *q;
 
@@ -2913,9 +2918,9 @@ static void blk_mq_del_queue_tag_set(struct request_queue *q)
 	list_del(&q->tag_set_list);
 	if (list_is_singular(&set->tag_list)) {
 		/* just transitioned to unshared */
-		set->flags &= ~BLK_MQ_F_TAG_SHARED;
+		set->flags &= ~BLK_MQ_F_TAG_QUEUE_SHARED;
 		/* update existing queue */
-		blk_mq_update_tag_set_depth(set, false);
+		blk_mq_update_tag_set_shared(set, false);
 	}
 	mutex_unlock(&set->tag_list_lock);
 	INIT_LIST_HEAD(&q->tag_set_list);
@@ -2930,12 +2935,12 @@ static void blk_mq_add_queue_tag_set(struct blk_mq_tag_set *set,
 	 * Check to see if we're transitioning to shared (from 1 to 2 queues).
 	 */
 	if (!list_empty(&set->tag_list) &&
-	    !(set->flags & BLK_MQ_F_TAG_SHARED)) {
-		set->flags |= BLK_MQ_F_TAG_SHARED;
+	    !(set->flags & BLK_MQ_F_TAG_QUEUE_SHARED)) {
+		set->flags |= BLK_MQ_F_TAG_QUEUE_SHARED;
 		/* update existing queue */
-		blk_mq_update_tag_set_depth(set, true);
+		blk_mq_update_tag_set_shared(set, true);
 	}
-	if (set->flags & BLK_MQ_F_TAG_SHARED)
+	if (set->flags & BLK_MQ_F_TAG_QUEUE_SHARED)
 		queue_set_hctx_shared(q, true);
 	list_add_tail(&q->tag_set_list, &set->tag_list);
 
@@ -3256,9 +3261,11 @@ static int __blk_mq_alloc_rq_maps(struct blk_mq_tag_set *set)
 {
 	int i;
 
-	for (i = 0; i < set->nr_hw_queues; i++)
+	for (i = 0; i < set->nr_hw_queues; i++) {
 		if (!__blk_mq_alloc_map_and_request(set, i))
 			goto out_unwind;
+		cond_resched();
+	}
 
 	return 0;
 
@@ -3438,11 +3445,23 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
 	if (ret)
 		goto out_free_mq_map;
 
+	if (blk_mq_is_sbitmap_shared(set->flags)) {
+		atomic_set(&set->active_queues_shared_sbitmap, 0);
+
+		if (blk_mq_init_shared_sbitmap(set, set->flags)) {
+			ret = -ENOMEM;
+			goto out_free_mq_rq_maps;
+		}
+	}
+
 	mutex_init(&set->tag_list_lock);
 	INIT_LIST_HEAD(&set->tag_list);
 
 	return 0;
 
+out_free_mq_rq_maps:
+	for (i = 0; i < set->nr_hw_queues; i++)
+		blk_mq_free_map_and_requests(set, i);
 out_free_mq_map:
 	for (i = 0; i < set->nr_maps; i++) {
 		kfree(set->map[i].mq_map);
@@ -3461,6 +3480,9 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
 	for (i = 0; i < set->nr_hw_queues; i++)
 		blk_mq_free_map_and_requests(set, i);
 
+	if (blk_mq_is_sbitmap_shared(set->flags))
+		blk_mq_exit_shared_sbitmap(set);
+
 	for (j = 0; j < set->nr_maps; j++) {
 		kfree(set->map[j].mq_map);
 		set->map[j].mq_map = NULL;
@@ -3497,6 +3519,8 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
 		if (!hctx->sched_tags) {
 			ret = blk_mq_tag_update_depth(hctx, &hctx->tags, nr,
 							false);
+			if (!ret && blk_mq_is_sbitmap_shared(set->flags))
+				blk_mq_tag_resize_shared_sbitmap(set, nr);
 		} else {
 			ret = blk_mq_tag_update_depth(hctx, &hctx->sched_tags,
 							nr, true);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 863a2f3346d4..a52703c98b77 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -53,11 +53,12 @@ struct request *blk_mq_dequeue_from_ctx(struct blk_mq_hw_ctx *hctx,
  */
 void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 		     unsigned int hctx_idx);
-void blk_mq_free_rq_map(struct blk_mq_tags *tags);
+void blk_mq_free_rq_map(struct blk_mq_tags *tags, unsigned int flags);
 struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 					unsigned int hctx_idx,
 					unsigned int nr_tags,
-					unsigned int reserved_tags);
+					unsigned int reserved_tags,
+					unsigned int flags);
 int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 		     unsigned int hctx_idx, unsigned int depth);
 
@@ -158,6 +159,11 @@ struct blk_mq_alloc_data {
 	struct blk_mq_hw_ctx *hctx;
 };
 
+static inline bool blk_mq_is_sbitmap_shared(unsigned int flags)
+{
+	return flags & BLK_MQ_F_TAG_HCTX_SHARED;
+}
+
 static inline struct blk_mq_tags *blk_mq_tags_from_data(struct blk_mq_alloc_data *data)
 {
 	if (data->q->elevator)
@@ -193,6 +199,28 @@ static inline bool blk_mq_get_dispatch_budget(struct request_queue *q)
 	return true;
 }
 
+static inline void __blk_mq_inc_active_requests(struct blk_mq_hw_ctx *hctx)
+{
+	if (blk_mq_is_sbitmap_shared(hctx->flags))
+		atomic_inc(&hctx->queue->nr_active_requests_shared_sbitmap);
+	else
+		atomic_inc(&hctx->nr_active);
+}
+
+static inline void __blk_mq_dec_active_requests(struct blk_mq_hw_ctx *hctx)
+{
+	if (blk_mq_is_sbitmap_shared(hctx->flags))
+		atomic_dec(&hctx->queue->nr_active_requests_shared_sbitmap);
+	else
+		atomic_dec(&hctx->nr_active);
+}
+
+static inline int __blk_mq_active_requests(struct blk_mq_hw_ctx *hctx)
+{
+	if (blk_mq_is_sbitmap_shared(hctx->flags))
+		return atomic_read(&hctx->queue->nr_active_requests_shared_sbitmap);
+	return atomic_read(&hctx->nr_active);
+}
 static inline void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
 					   struct request *rq)
 {
@@ -201,7 +229,7 @@ static inline void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
 
 	if (rq->rq_flags & RQF_MQ_INFLIGHT) {
 		rq->rq_flags &= ~RQF_MQ_INFLIGHT;
-		atomic_dec(&hctx->nr_active);
+		__blk_mq_dec_active_requests(hctx);
 	}
 }
 
@@ -253,4 +281,46 @@ static inline struct blk_plug *blk_mq_plug(struct request_queue *q,
 	return NULL;
 }
 
+/*
+ * For shared tag users, we track the number of currently active users
+ * and attempt to provide a fair share of the tag depth for each of them.
+ */
+static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx,
+				  struct sbitmap_queue *bt)
+{
+	unsigned int depth, users;
+
+	if (!hctx || !(hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED))
+		return true;
+
+	/*
+	 * Don't try dividing an ant
+	 */
+	if (bt->sb.depth == 1)
+		return true;
+
+	if (blk_mq_is_sbitmap_shared(hctx->flags)) {
+		struct request_queue *q = hctx->queue;
+		struct blk_mq_tag_set *set = q->tag_set;
+
+		if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &q->queue_flags))
+			return true;
+		users = atomic_read(&set->active_queues_shared_sbitmap);
+	} else {
+		if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
+			return true;
+		users = atomic_read(&hctx->tags->active_queues);
+	}
+
+	if (!users)
+		return true;
+
+	/*
+	 * Allow at least some tags
+	 */
+	depth = max((bt->sb.depth + users - 1) / users, 4U);
+	return __blk_mq_active_requests(hctx) < depth;
+}
+
+
 #endif
diff --git a/block/blk-settings.c b/block/blk-settings.c
index 76a7e03bcd6c..4f6eb4bb1723 100644
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -172,15 +172,13 @@ EXPORT_SYMBOL(blk_queue_max_hw_sectors);
  *
  * Description:
  *    If a driver doesn't want IOs to cross a given chunk size, it can set
- *    this limit and prevent merging across chunks. Note that the chunk size
- *    must currently be a power-of-2 in sectors. Also note that the block
- *    layer must accept a page worth of data at any offset. So if the
- *    crossing of chunks is a hard limitation in the driver, it must still be
- *    prepared to split single page bios.
+ *    this limit and prevent merging across chunks. Note that the block layer
+ *    must accept a page worth of data at any offset. So if the crossing of
+ *    chunks is a hard limitation in the driver, it must still be prepared
+ *    to split single page bios.
  **/
 void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
 {
-	BUG_ON(!is_power_of_2(chunk_sectors));
 	q->limits.chunk_sectors = chunk_sectors;
 }
 EXPORT_SYMBOL(blk_queue_chunk_sectors);
@@ -374,6 +372,19 @@ void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
 }
 EXPORT_SYMBOL(blk_queue_alignment_offset);
 
+void blk_queue_update_readahead(struct request_queue *q)
+{
+	/*
+	 * For read-ahead of large files to be effective, we need to read ahead
+	 * at least twice the optimal I/O size.
+	 */
+	q->backing_dev_info->ra_pages =
+		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
+	q->backing_dev_info->io_pages =
+		queue_max_sectors(q) >> (PAGE_SHIFT - 9);
+}
+EXPORT_SYMBOL_GPL(blk_queue_update_readahead);
+
 /**
  * blk_limits_io_min - set minimum request size for a device
  * @limits: the queue limits
@@ -452,6 +463,8 @@ EXPORT_SYMBOL(blk_limits_io_opt);
 void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
 {
 	blk_limits_io_opt(&q->limits, opt);
+	q->backing_dev_info->ra_pages =
+		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
 }
 EXPORT_SYMBOL(blk_queue_io_opt);
 
@@ -534,6 +547,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 
 	t->io_min = max(t->io_min, b->io_min);
 	t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
+	t->chunk_sectors = lcm_not_zero(t->chunk_sectors, b->chunk_sectors);
 
 	/* Physical block size a multiple of the logical block size? */
 	if (t->physical_block_size & (t->logical_block_size - 1)) {
@@ -556,6 +570,13 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 		ret = -1;
 	}
 
+	/* chunk_sectors a multiple of the physical block size? */
+	if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
+		t->chunk_sectors = 0;
+		t->misaligned = 1;
+		ret = -1;
+	}
+
 	t->raid_partial_stripes_expensive =
 		max(t->raid_partial_stripes_expensive,
 		    b->raid_partial_stripes_expensive);
@@ -594,10 +615,6 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 			t->discard_granularity;
 	}
 
-	if (b->chunk_sectors)
-		t->chunk_sectors = min_not_zero(t->chunk_sectors,
-						b->chunk_sectors);
-
 	t->zoned = max(t->zoned, b->zoned);
 	return ret;
 }
@@ -629,8 +646,7 @@ void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
 		       top, bottom);
 	}
 
-	t->backing_dev_info->io_pages =
-		t->limits.max_sectors >> (PAGE_SHIFT - 9);
+	blk_queue_update_readahead(disk->queue);
 }
 EXPORT_SYMBOL(disk_stack_limits);
 
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 7dda709f3ccb..76b54c7750b0 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -260,14 +260,14 @@ static ssize_t queue_max_hw_sectors_show(struct request_queue *q, char *page)
 
 #define QUEUE_SYSFS_BIT_FNS(name, flag, neg)				\
 static ssize_t								\
-queue_show_##name(struct request_queue *q, char *page)			\
+queue_##name##_show(struct request_queue *q, char *page)		\
 {									\
 	int bit;							\
 	bit = test_bit(QUEUE_FLAG_##flag, &q->queue_flags);		\
 	return queue_var_show(neg ? !bit : bit, page);			\
 }									\
 static ssize_t								\
-queue_store_##name(struct request_queue *q, const char *page, size_t count) \
+queue_##name##_store(struct request_queue *q, const char *page, size_t count) \
 {									\
 	unsigned long val;						\
 	ssize_t ret;							\
@@ -287,6 +287,7 @@ queue_store_##name(struct request_queue *q, const char *page, size_t count) \
 QUEUE_SYSFS_BIT_FNS(nonrot, NONROT, 1);
 QUEUE_SYSFS_BIT_FNS(random, ADD_RANDOM, 0);
 QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0);
+QUEUE_SYSFS_BIT_FNS(stable_writes, STABLE_WRITES, 0);
 #undef QUEUE_SYSFS_BIT_FNS
 
 static ssize_t queue_zoned_show(struct request_queue *q, char *page)
@@ -547,218 +548,73 @@ static ssize_t queue_dax_show(struct request_queue *q, char *page)
 	return queue_var_show(blk_queue_dax(q), page);
 }
 
-static struct queue_sysfs_entry queue_requests_entry = {
-	.attr = {.name = "nr_requests", .mode = 0644 },
-	.show = queue_requests_show,
-	.store = queue_requests_store,
-};
-
-static struct queue_sysfs_entry queue_ra_entry = {
-	.attr = {.name = "read_ahead_kb", .mode = 0644 },
-	.show = queue_ra_show,
-	.store = queue_ra_store,
-};
-
-static struct queue_sysfs_entry queue_max_sectors_entry = {
-	.attr = {.name = "max_sectors_kb", .mode = 0644 },
-	.show = queue_max_sectors_show,
-	.store = queue_max_sectors_store,
-};
-
-static struct queue_sysfs_entry queue_max_hw_sectors_entry = {
-	.attr = {.name = "max_hw_sectors_kb", .mode = 0444 },
-	.show = queue_max_hw_sectors_show,
-};
-
-static struct queue_sysfs_entry queue_max_segments_entry = {
-	.attr = {.name = "max_segments", .mode = 0444 },
-	.show = queue_max_segments_show,
-};
-
-static struct queue_sysfs_entry queue_max_discard_segments_entry = {
-	.attr = {.name = "max_discard_segments", .mode = 0444 },
-	.show = queue_max_discard_segments_show,
-};
-
-static struct queue_sysfs_entry queue_max_integrity_segments_entry = {
-	.attr = {.name = "max_integrity_segments", .mode = 0444 },
-	.show = queue_max_integrity_segments_show,
-};
-
-static struct queue_sysfs_entry queue_max_segment_size_entry = {
-	.attr = {.name = "max_segment_size", .mode = 0444 },
-	.show = queue_max_segment_size_show,
-};
+#define QUEUE_RO_ENTRY(_prefix, _name)			\
+static struct queue_sysfs_entry _prefix##_entry = {	\
+	.attr	= { .name = _name, .mode = 0444 },	\
+	.show	= _prefix##_show,			\
+};
+
+#define QUEUE_RW_ENTRY(_prefix, _name)			\
+static struct queue_sysfs_entry _prefix##_entry = {	\
+	.attr	= { .name = _name, .mode = 0644 },	\
+	.show	= _prefix##_show,			\
+	.store	= _prefix##_store,			\
+};
+
+QUEUE_RW_ENTRY(queue_requests, "nr_requests");
+QUEUE_RW_ENTRY(queue_ra, "read_ahead_kb");
+QUEUE_RW_ENTRY(queue_max_sectors, "max_sectors_kb");
+QUEUE_RO_ENTRY(queue_max_hw_sectors, "max_hw_sectors_kb");
+QUEUE_RO_ENTRY(queue_max_segments, "max_segments");
+QUEUE_RO_ENTRY(queue_max_integrity_segments, "max_integrity_segments");
+QUEUE_RO_ENTRY(queue_max_segment_size, "max_segment_size");
+QUEUE_RW_ENTRY(elv_iosched, "scheduler");
+
+QUEUE_RO_ENTRY(queue_logical_block_size, "logical_block_size");
+QUEUE_RO_ENTRY(queue_physical_block_size, "physical_block_size");
+QUEUE_RO_ENTRY(queue_chunk_sectors, "chunk_sectors");
+QUEUE_RO_ENTRY(queue_io_min, "minimum_io_size");
+QUEUE_RO_ENTRY(queue_io_opt, "optimal_io_size");
+
+QUEUE_RO_ENTRY(queue_max_discard_segments, "max_discard_segments");
+QUEUE_RO_ENTRY(queue_discard_granularity, "discard_granularity");
+QUEUE_RO_ENTRY(queue_discard_max_hw, "discard_max_hw_bytes");
+QUEUE_RW_ENTRY(queue_discard_max, "discard_max_bytes");
+QUEUE_RO_ENTRY(queue_discard_zeroes_data, "discard_zeroes_data");
+
+QUEUE_RO_ENTRY(queue_write_same_max, "write_same_max_bytes");
+QUEUE_RO_ENTRY(queue_write_zeroes_max, "write_zeroes_max_bytes");
+QUEUE_RO_ENTRY(queue_zone_append_max, "zone_append_max_bytes");
+
+QUEUE_RO_ENTRY(queue_zoned, "zoned");
+QUEUE_RO_ENTRY(queue_nr_zones, "nr_zones");
+QUEUE_RO_ENTRY(queue_max_open_zones, "max_open_zones");
+QUEUE_RO_ENTRY(queue_max_active_zones, "max_active_zones");
+
+QUEUE_RW_ENTRY(queue_nomerges, "nomerges");
+QUEUE_RW_ENTRY(queue_rq_affinity, "rq_affinity");
+QUEUE_RW_ENTRY(queue_poll, "io_poll");
+QUEUE_RW_ENTRY(queue_poll_delay, "io_poll_delay");
+QUEUE_RW_ENTRY(queue_wc, "write_cache");
+QUEUE_RO_ENTRY(queue_fua, "fua");
+QUEUE_RO_ENTRY(queue_dax, "dax");
+QUEUE_RW_ENTRY(queue_io_timeout, "io_timeout");
+QUEUE_RW_ENTRY(queue_wb_lat, "wbt_lat_usec");
 
-static struct queue_sysfs_entry queue_iosched_entry = {
-	.attr = {.name = "scheduler", .mode = 0644 },
-	.show = elv_iosched_show,
-	.store = elv_iosched_store,
-};
+#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
+QUEUE_RW_ENTRY(blk_throtl_sample_time, "throttle_sample_time");
+#endif
 
+/* legacy alias for logical_block_size: */
 static struct queue_sysfs_entry queue_hw_sector_size_entry = {
 	.attr = {.name = "hw_sector_size", .mode = 0444 },
 	.show = queue_logical_block_size_show,
 };
 
-static struct queue_sysfs_entry queue_logical_block_size_entry = {
-	.attr = {.name = "logical_block_size", .mode = 0444 },
-	.show = queue_logical_block_size_show,
-};
-
-static struct queue_sysfs_entry queue_physical_block_size_entry = {
-	.attr = {.name = "physical_block_size", .mode = 0444 },
-	.show = queue_physical_block_size_show,
-};
-
-static struct queue_sysfs_entry queue_chunk_sectors_entry = {
-	.attr = {.name = "chunk_sectors", .mode = 0444 },
-	.show = queue_chunk_sectors_show,
-};
-
-static struct queue_sysfs_entry queue_io_min_entry = {
-	.attr = {.name = "minimum_io_size", .mode = 0444 },
-	.show = queue_io_min_show,
-};
-
-static struct queue_sysfs_entry queue_io_opt_entry = {
-	.attr = {.name = "optimal_io_size", .mode = 0444 },
-	.show = queue_io_opt_show,
-};
-
-static struct queue_sysfs_entry queue_discard_granularity_entry = {
-	.attr = {.name = "discard_granularity", .mode = 0444 },
-	.show = queue_discard_granularity_show,
-};
-
-static struct queue_sysfs_entry queue_discard_max_hw_entry = {
-	.attr = {.name = "discard_max_hw_bytes", .mode = 0444 },
-	.show = queue_discard_max_hw_show,
-};
-
-static struct queue_sysfs_entry queue_discard_max_entry = {
-	.attr = {.name = "discard_max_bytes", .mode = 0644 },
-	.show = queue_discard_max_show,
-	.store = queue_discard_max_store,
-};
-
-static struct queue_sysfs_entry queue_discard_zeroes_data_entry = {
-	.attr = {.name = "discard_zeroes_data", .mode = 0444 },
-	.show = queue_discard_zeroes_data_show,
-};
-
-static struct queue_sysfs_entry queue_write_same_max_entry = {
-	.attr = {.name = "write_same_max_bytes", .mode = 0444 },
-	.show = queue_write_same_max_show,
-};
-
-static struct queue_sysfs_entry queue_write_zeroes_max_entry = {
-	.attr = {.name = "write_zeroes_max_bytes", .mode = 0444 },
-	.show = queue_write_zeroes_max_show,
-};
-
-static struct queue_sysfs_entry queue_zone_append_max_entry = {
-	.attr = {.name = "zone_append_max_bytes", .mode = 0444 },
-	.show = queue_zone_append_max_show,
-};
-
-static struct queue_sysfs_entry queue_nonrot_entry = {
-	.attr = {.name = "rotational", .mode = 0644 },
-	.show = queue_show_nonrot,
-	.store = queue_store_nonrot,
-};
-
-static struct queue_sysfs_entry queue_zoned_entry = {
-	.attr = {.name = "zoned", .mode = 0444 },
-	.show = queue_zoned_show,
-};
-
-static struct queue_sysfs_entry queue_nr_zones_entry = {
-	.attr = {.name = "nr_zones", .mode = 0444 },
-	.show = queue_nr_zones_show,
-};
-
-static struct queue_sysfs_entry queue_max_open_zones_entry = {
-	.attr = {.name = "max_open_zones", .mode = 0444 },
-	.show = queue_max_open_zones_show,
-};
-
-static struct queue_sysfs_entry queue_max_active_zones_entry = {
-	.attr = {.name = "max_active_zones", .mode = 0444 },
-	.show = queue_max_active_zones_show,
-};
-
-static struct queue_sysfs_entry queue_nomerges_entry = {
-	.attr = {.name = "nomerges", .mode = 0644 },
-	.show = queue_nomerges_show,
-	.store = queue_nomerges_store,
-};
-
-static struct queue_sysfs_entry queue_rq_affinity_entry = {
-	.attr = {.name = "rq_affinity", .mode = 0644 },
-	.show = queue_rq_affinity_show,
-	.store = queue_rq_affinity_store,
-};
-
-static struct queue_sysfs_entry queue_iostats_entry = {
-	.attr = {.name = "iostats", .mode = 0644 },
-	.show = queue_show_iostats,
-	.store = queue_store_iostats,
-};
-
-static struct queue_sysfs_entry queue_random_entry = {
-	.attr = {.name = "add_random", .mode = 0644 },
-	.show = queue_show_random,
-	.store = queue_store_random,
-};
-
-static struct queue_sysfs_entry queue_poll_entry = {
-	.attr = {.name = "io_poll", .mode = 0644 },
-	.show = queue_poll_show,
-	.store = queue_poll_store,
-};
-
-static struct queue_sysfs_entry queue_poll_delay_entry = {
-	.attr = {.name = "io_poll_delay", .mode = 0644 },
-	.show = queue_poll_delay_show,
-	.store = queue_poll_delay_store,
-};
-
-static struct queue_sysfs_entry queue_wc_entry = {
-	.attr = {.name = "write_cache", .mode = 0644 },
-	.show = queue_wc_show,
-	.store = queue_wc_store,
-};
-
-static struct queue_sysfs_entry queue_fua_entry = {
-	.attr = {.name = "fua", .mode = 0444 },
-	.show = queue_fua_show,
-};
-
-static struct queue_sysfs_entry queue_dax_entry = {
-	.attr = {.name = "dax", .mode = 0444 },
-	.show = queue_dax_show,
-};
-
-static struct queue_sysfs_entry queue_io_timeout_entry = {
-	.attr = {.name = "io_timeout", .mode = 0644 },
-	.show = queue_io_timeout_show,
-	.store = queue_io_timeout_store,
-};
-
-static struct queue_sysfs_entry queue_wb_lat_entry = {
-	.attr = {.name = "wbt_lat_usec", .mode = 0644 },
-	.show = queue_wb_lat_show,
-	.store = queue_wb_lat_store,
-};
-
-#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
-static struct queue_sysfs_entry throtl_sample_time_entry = {
-	.attr = {.name = "throttle_sample_time", .mode = 0644 },
-	.show = blk_throtl_sample_time_show,
-	.store = blk_throtl_sample_time_store,
-};
-#endif
+QUEUE_RW_ENTRY(queue_nonrot, "rotational");
+QUEUE_RW_ENTRY(queue_iostats, "iostats");
+QUEUE_RW_ENTRY(queue_random, "add_random");
+QUEUE_RW_ENTRY(queue_stable_writes, "stable_writes");
 
 static struct attribute *queue_attrs[] = {
 	&queue_requests_entry.attr,
@@ -769,7 +625,7 @@ static struct attribute *queue_attrs[] = {
 	&queue_max_discard_segments_entry.attr,
 	&queue_max_integrity_segments_entry.attr,
 	&queue_max_segment_size_entry.attr,
-	&queue_iosched_entry.attr,
+	&elv_iosched_entry.attr,
 	&queue_hw_sector_size_entry.attr,
 	&queue_logical_block_size_entry.attr,
 	&queue_physical_block_size_entry.attr,
@@ -791,6 +647,7 @@ static struct attribute *queue_attrs[] = {
 	&queue_nomerges_entry.attr,
 	&queue_rq_affinity_entry.attr,
 	&queue_iostats_entry.attr,
+	&queue_stable_writes_entry.attr,
 	&queue_random_entry.attr,
 	&queue_poll_entry.attr,
 	&queue_wc_entry.attr,
@@ -800,7 +657,7 @@ static struct attribute *queue_attrs[] = {
 	&queue_poll_delay_entry.attr,
 	&queue_io_timeout_entry.attr,
 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
-	&throtl_sample_time_entry.attr,
+	&blk_throtl_sample_time_entry.attr,
 #endif
 	NULL,
 };
@@ -1000,6 +857,8 @@ int blk_register_queue(struct gendisk *disk)
 		percpu_ref_switch_to_percpu(&q->q_usage_counter);
 	}
 
+	blk_queue_update_readahead(q);
+
 	ret = blk_trace_init_sysfs(dev);
 	if (ret)
 		return ret;
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index fee3325edf27..36ba61c5cdbd 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -15,10 +15,10 @@
 #include "blk-cgroup-rwstat.h"
 
 /* Max dispatch from a group in 1 round */
-static int throtl_grp_quantum = 8;
+#define THROTL_GRP_QUANTUM 8
 
 /* Total max dispatch from all groups in one round */
-static int throtl_quantum = 32;
+#define THROTL_QUANTUM 32
 
 /* Throttling is performed over a slice and after that slice is renewed */
 #define DFL_THROTL_SLICE_HD (HZ / 10)
@@ -150,7 +150,7 @@ struct throtl_grp {
 	/* user configured IOPS limits */
 	unsigned int iops_conf[2][LIMIT_CNT];
 
-	/* Number of bytes disptached in current slice */
+	/* Number of bytes dispatched in current slice */
 	uint64_t bytes_disp[2];
 	/* Number of bio's dispatched in current slice */
 	unsigned int io_disp[2];
@@ -852,7 +852,7 @@ static inline void throtl_trim_slice(struct throtl_grp *tg, bool rw)
 	/*
 	 * A bio has been dispatched. Also adjust slice_end. It might happen
 	 * that initially cgroup limit was very low resulting in high
-	 * slice_end, but later limit was bumped up and bio was dispached
+	 * slice_end, but later limit was bumped up and bio was dispatched
 	 * sooner, then we need to reduce slice_end. A high bogus slice_end
 	 * is bad because it does not allow new slice to start.
 	 */
@@ -894,13 +894,19 @@ static inline void throtl_trim_slice(struct throtl_grp *tg, bool rw)
 }
 
 static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
-				  unsigned long *wait)
+				  u32 iops_limit, unsigned long *wait)
 {
 	bool rw = bio_data_dir(bio);
 	unsigned int io_allowed;
 	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
 	u64 tmp;
 
+	if (iops_limit == UINT_MAX) {
+		if (wait)
+			*wait = 0;
+		return true;
+	}
+
 	jiffy_elapsed = jiffies - tg->slice_start[rw];
 
 	/* Round up to the next throttle slice, wait time must be nonzero */
@@ -913,7 +919,7 @@ static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
 	 * have been trimmed.
 	 */
 
-	tmp = (u64)tg_iops_limit(tg, rw) * jiffy_elapsed_rnd;
+	tmp = (u64)iops_limit * jiffy_elapsed_rnd;
 	do_div(tmp, HZ);
 
 	if (tmp > UINT_MAX)
@@ -936,13 +942,19 @@ static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
 }
 
 static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
-				 unsigned long *wait)
+				 u64 bps_limit, unsigned long *wait)
 {
 	bool rw = bio_data_dir(bio);
 	u64 bytes_allowed, extra_bytes, tmp;
 	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
 	unsigned int bio_size = throtl_bio_data_size(bio);
 
+	if (bps_limit == U64_MAX) {
+		if (wait)
+			*wait = 0;
+		return true;
+	}
+
 	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
 
 	/* Slice has just started. Consider one slice interval */
@@ -951,7 +963,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
 
 	jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, tg->td->throtl_slice);
 
-	tmp = tg_bps_limit(tg, rw) * jiffy_elapsed_rnd;
+	tmp = bps_limit * jiffy_elapsed_rnd;
 	do_div(tmp, HZ);
 	bytes_allowed = tmp;
 
@@ -963,7 +975,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
 
 	/* Calc approx time to dispatch */
 	extra_bytes = tg->bytes_disp[rw] + bio_size - bytes_allowed;
-	jiffy_wait = div64_u64(extra_bytes * HZ, tg_bps_limit(tg, rw));
+	jiffy_wait = div64_u64(extra_bytes * HZ, bps_limit);
 
 	if (!jiffy_wait)
 		jiffy_wait = 1;
@@ -987,6 +999,8 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
 {
 	bool rw = bio_data_dir(bio);
 	unsigned long bps_wait = 0, iops_wait = 0, max_wait = 0;
+	u64 bps_limit = tg_bps_limit(tg, rw);
+	u32 iops_limit = tg_iops_limit(tg, rw);
 
 	/*
  	 * Currently whole state machine of group depends on first bio
@@ -998,8 +1012,7 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
 	       bio != throtl_peek_queued(&tg->service_queue.queued[rw]));
 
 	/* If tg->bps = -1, then BW is unlimited */
-	if (tg_bps_limit(tg, rw) == U64_MAX &&
-	    tg_iops_limit(tg, rw) == UINT_MAX) {
+	if (bps_limit == U64_MAX && iops_limit == UINT_MAX) {
 		if (wait)
 			*wait = 0;
 		return true;
@@ -1021,8 +1034,8 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
 				jiffies + tg->td->throtl_slice);
 	}
 
-	if (tg_with_in_bps_limit(tg, bio, &bps_wait) &&
-	    tg_with_in_iops_limit(tg, bio, &iops_wait)) {
+	if (tg_with_in_bps_limit(tg, bio, bps_limit, &bps_wait) &&
+	    tg_with_in_iops_limit(tg, bio, iops_limit, &iops_wait)) {
 		if (wait)
 			*wait = 0;
 		return true;
@@ -1082,7 +1095,7 @@ static void throtl_add_bio_tg(struct bio *bio, struct throtl_qnode *qn,
 	 * If @tg doesn't currently have any bios queued in the same
 	 * direction, queueing @bio can change when @tg should be
 	 * dispatched.  Mark that @tg was empty.  This is automatically
-	 * cleaered on the next tg_update_disptime().
+	 * cleared on the next tg_update_disptime().
 	 */
 	if (!sq->nr_queued[rw])
 		tg->flags |= THROTL_TG_WAS_EMPTY;
@@ -1175,8 +1188,8 @@ static int throtl_dispatch_tg(struct throtl_grp *tg)
 {
 	struct throtl_service_queue *sq = &tg->service_queue;
 	unsigned int nr_reads = 0, nr_writes = 0;
-	unsigned int max_nr_reads = throtl_grp_quantum*3/4;
-	unsigned int max_nr_writes = throtl_grp_quantum - max_nr_reads;
+	unsigned int max_nr_reads = THROTL_GRP_QUANTUM * 3 / 4;
+	unsigned int max_nr_writes = THROTL_GRP_QUANTUM - max_nr_reads;
 	struct bio *bio;
 
 	/* Try to dispatch 75% READS and 25% WRITES */
@@ -1226,7 +1239,7 @@ static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
 		if (sq->nr_queued[0] || sq->nr_queued[1])
 			tg_update_disptime(tg);
 
-		if (nr_disp >= throtl_quantum)
+		if (nr_disp >= THROTL_QUANTUM)
 			break;
 	}
 
@@ -1303,7 +1316,7 @@ again:
 			}
 		}
 	} else {
-		/* reached the top-level, queue issueing */
+		/* reached the top-level, queue issuing */
 		queue_work(kthrotld_workqueue, &td->dispatch_work);
 	}
 out_unlock:
@@ -1314,8 +1327,8 @@ out_unlock:
  * blk_throtl_dispatch_work_fn - work function for throtl_data->dispatch_work
  * @work: work item being executed
  *
- * This function is queued for execution when bio's reach the bio_lists[]
- * of throtl_data->service_queue.  Those bio's are ready and issued by this
+ * This function is queued for execution when bios reach the bio_lists[]
+ * of throtl_data->service_queue.  Those bios are ready and issued by this
  * function.
  */
 static void blk_throtl_dispatch_work_fn(struct work_struct *work)
@@ -1428,8 +1441,8 @@ static void tg_conf_updated(struct throtl_grp *tg, bool global)
 	 * that a group's limit are dropped suddenly and we don't want to
 	 * account recently dispatched IO with new low rate.
 	 */
-	throtl_start_new_slice(tg, 0);
-	throtl_start_new_slice(tg, 1);
+	throtl_start_new_slice(tg, READ);
+	throtl_start_new_slice(tg, WRITE);
 
 	if (tg->flags & THROTL_TG_PENDING) {
 		tg_update_disptime(tg);
@@ -2230,7 +2243,7 @@ again:
 
 		/*
 		 * @bio passed through this layer without being throttled.
-		 * Climb up the ladder.  If we''re already at the top, it
+		 * Climb up the ladder.  If we're already at the top, it
 		 * can be executed directly.
 		 */
 		qn = &tg->qnode_on_parent[rw];
diff --git a/block/blk.h b/block/blk.h
index 49e2928a1632..c08762e10b04 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -29,6 +29,12 @@ struct blk_flush_queue {
 	spinlock_t		mq_flush_lock;
 };
 
+enum bio_merge_status {
+	BIO_MERGE_OK,
+	BIO_MERGE_NONE,
+	BIO_MERGE_FAILED,
+};
+
 extern struct kmem_cache *blk_requestq_cachep;
 extern struct kobj_type blk_queue_ktype;
 extern struct ida blk_queue_ida;
@@ -169,14 +175,19 @@ static inline void blk_integrity_del(struct gendisk *disk)
 unsigned long blk_rq_timeout(unsigned long timeout);
 void blk_add_timer(struct request *req);
 
-bool bio_attempt_front_merge(struct request *req, struct bio *bio,
-		unsigned int nr_segs);
-bool bio_attempt_back_merge(struct request *req, struct bio *bio,
-		unsigned int nr_segs);
-bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
-		struct bio *bio);
+enum bio_merge_status bio_attempt_front_merge(struct request *req,
+					      struct bio *bio,
+					      unsigned int nr_segs);
+enum bio_merge_status bio_attempt_back_merge(struct request *req,
+					     struct bio *bio,
+					     unsigned int nr_segs);
+enum bio_merge_status bio_attempt_discard_merge(struct request_queue *q,
+						struct request *req,
+						struct bio *bio);
 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
 		unsigned int nr_segs, struct request **same_queue_rq);
+bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
+			struct bio *bio, unsigned int nr_segs);
 
 void blk_account_io_start(struct request *req);
 void blk_account_io_done(struct request *req, u64 now);
@@ -350,7 +361,7 @@ char *disk_name(struct gendisk *hd, int partno, char *buf);
 #define ADDPART_FLAG_NONE	0
 #define ADDPART_FLAG_RAID	1
 #define ADDPART_FLAG_WHOLEDISK	2
-void delete_partition(struct gendisk *disk, struct hd_struct *part);
+void delete_partition(struct hd_struct *part);
 int bdev_add_partition(struct block_device *bdev, int partno,
 		sector_t start, sector_t length);
 int bdev_del_partition(struct block_device *bdev, int partno);
diff --git a/block/bsg-lib.c b/block/bsg-lib.c
index d185396d88bb..330fede77271 100644
--- a/block/bsg-lib.c
+++ b/block/bsg-lib.c
@@ -207,7 +207,7 @@ static int bsg_map_buffer(struct bsg_buffer *buf, struct request *req)
 
 	BUG_ON(!req->nr_phys_segments);
 
-	buf->sg_list = kzalloc(sz, GFP_KERNEL);
+	buf->sg_list = kmalloc(sz, GFP_KERNEL);
 	if (!buf->sg_list)
 		return -ENOMEM;
 	sg_init_table(buf->sg_list, req->nr_phys_segments);
diff --git a/block/genhd.c b/block/genhd.c
index 99c64641c314..e5f17f022ec7 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -50,14 +50,13 @@ static void disk_release_events(struct gendisk *disk);
  * zero and will not be set to zero
  */
 void set_capacity_revalidate_and_notify(struct gendisk *disk, sector_t size,
-					bool revalidate)
+					bool update_bdev)
 {
 	sector_t capacity = get_capacity(disk);
 
 	set_capacity(disk, size);
-
-	if (revalidate)
-		revalidate_disk(disk);
+	if (update_bdev)
+		revalidate_disk_size(disk, true);
 
 	if (capacity != size && capacity != 0 && size != 0) {
 		char *envp[] = { "RESIZE=1", NULL };
@@ -86,7 +85,7 @@ char *disk_name(struct gendisk *hd, int partno, char *buf)
 
 const char *bdevname(struct block_device *bdev, char *buf)
 {
-	return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
+	return disk_name(bdev->bd_disk, bdev->bd_partno, buf);
 }
 EXPORT_SYMBOL(bdevname);
 
@@ -110,8 +109,7 @@ static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat)
 	}
 }
 
-static unsigned int part_in_flight(struct request_queue *q,
-		struct hd_struct *part)
+static unsigned int part_in_flight(struct hd_struct *part)
 {
 	unsigned int inflight = 0;
 	int cpu;
@@ -126,8 +124,7 @@ static unsigned int part_in_flight(struct request_queue *q,
 	return inflight;
 }
 
-static void part_in_flight_rw(struct request_queue *q, struct hd_struct *part,
-		unsigned int inflight[2])
+static void part_in_flight_rw(struct hd_struct *part, unsigned int inflight[2])
 {
 	int cpu;
 
@@ -676,11 +673,23 @@ static int exact_lock(dev_t devt, void *data)
 	return 0;
 }
 
+static void disk_scan_partitions(struct gendisk *disk)
+{
+	struct block_device *bdev;
+
+	if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
+		return;
+
+	set_bit(GD_NEED_PART_SCAN, &disk->state);
+	bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
+	if (!IS_ERR(bdev))
+		blkdev_put(bdev, FMODE_READ);
+}
+
 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;
 	struct disk_part_iter piter;
 	struct hd_struct *part;
 	int err;
@@ -722,25 +731,8 @@ static void register_disk(struct device *parent, struct gendisk *disk,
 		return;
 	}
 
-	/* No minors to use for partitions */
-	if (!disk_part_scan_enabled(disk))
-		goto exit;
+	disk_scan_partitions(disk);
 
-	/* No such device (e.g., media were just removed) */
-	if (!get_capacity(disk))
-		goto exit;
-
-	bdev = bdget_disk(disk, 0);
-	if (!bdev)
-		goto exit;
-
-	bdev->bd_invalidated = 1;
-	err = blkdev_get(bdev, FMODE_READ, NULL);
-	if (err < 0)
-		goto exit;
-	blkdev_put(bdev, FMODE_READ);
-
-exit:
 	/* announce disk after possible partitions are created */
 	dev_set_uevent_suppress(ddev, 0);
 	kobject_uevent(&ddev->kobj, KOBJ_ADD);
@@ -913,7 +905,7 @@ void del_gendisk(struct gendisk *disk)
 			     DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
 	while ((part = disk_part_iter_next(&piter))) {
 		invalidate_partition(disk, part->partno);
-		delete_partition(disk, part);
+		delete_partition(part);
 	}
 	disk_part_iter_exit(&piter);
 
@@ -1301,7 +1293,7 @@ ssize_t part_stat_show(struct device *dev,
 	if (queue_is_mq(q))
 		inflight = blk_mq_in_flight(q, p);
 	else
-		inflight = part_in_flight(q, p);
+		inflight = part_in_flight(p);
 
 	return sprintf(buf,
 		"%8lu %8lu %8llu %8u "
@@ -1343,7 +1335,7 @@ ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
 	if (queue_is_mq(q))
 		blk_mq_in_flight_rw(q, p, inflight);
 	else
-		part_in_flight_rw(q, p, inflight);
+		part_in_flight_rw(p, inflight);
 
 	return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
 }
@@ -1623,7 +1615,7 @@ static int diskstats_show(struct seq_file *seqf, void *v)
 		if (queue_is_mq(gp->queue))
 			inflight = blk_mq_in_flight(gp->queue, hd);
 		else
-			inflight = part_in_flight(gp->queue, hd);
+			inflight = part_in_flight(hd);
 
 		seq_printf(seqf, "%4d %7d %s "
 			   "%lu %lu %lu %u "
@@ -1729,45 +1721,48 @@ struct gendisk *__alloc_disk_node(int minors, int node_id)
 	}
 
 	disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
-	if (disk) {
-		disk->part0.dkstats = alloc_percpu(struct disk_stats);
-		if (!disk->part0.dkstats) {
-			kfree(disk);
-			return NULL;
-		}
-		init_rwsem(&disk->lookup_sem);
-		disk->node_id = node_id;
-		if (disk_expand_part_tbl(disk, 0)) {
-			free_percpu(disk->part0.dkstats);
-			kfree(disk);
-			return NULL;
-		}
-		ptbl = rcu_dereference_protected(disk->part_tbl, 1);
-		rcu_assign_pointer(ptbl->part[0], &disk->part0);
+	if (!disk)
+		return NULL;
 
-		/*
-		 * set_capacity() and get_capacity() currently don't use
-		 * seqcounter to read/update the part0->nr_sects. Still init
-		 * the counter as we can read the sectors in IO submission
-		 * patch using seqence counters.
-		 *
-		 * TODO: Ideally set_capacity() and get_capacity() should be
-		 * converted to make use of bd_mutex and sequence counters.
-		 */
-		hd_sects_seq_init(&disk->part0);
-		if (hd_ref_init(&disk->part0)) {
-			hd_free_part(&disk->part0);
-			kfree(disk);
-			return NULL;
-		}
+	disk->part0.dkstats = alloc_percpu(struct disk_stats);
+	if (!disk->part0.dkstats)
+		goto out_free_disk;
 
-		disk->minors = minors;
-		rand_initialize_disk(disk);
-		disk_to_dev(disk)->class = &block_class;
-		disk_to_dev(disk)->type = &disk_type;
-		device_initialize(disk_to_dev(disk));
+	init_rwsem(&disk->lookup_sem);
+	disk->node_id = node_id;
+	if (disk_expand_part_tbl(disk, 0)) {
+		free_percpu(disk->part0.dkstats);
+		goto out_free_disk;
 	}
+
+	ptbl = rcu_dereference_protected(disk->part_tbl, 1);
+	rcu_assign_pointer(ptbl->part[0], &disk->part0);
+
+	/*
+	 * set_capacity() and get_capacity() currently don't use
+	 * seqcounter to read/update the part0->nr_sects. Still init
+	 * the counter as we can read the sectors in IO submission
+	 * patch using seqence counters.
+	 *
+	 * TODO: Ideally set_capacity() and get_capacity() should be
+	 * converted to make use of bd_mutex and sequence counters.
+	 */
+	hd_sects_seq_init(&disk->part0);
+	if (hd_ref_init(&disk->part0))
+		goto out_free_part0;
+
+	disk->minors = minors;
+	rand_initialize_disk(disk);
+	disk_to_dev(disk)->class = &block_class;
+	disk_to_dev(disk)->type = &disk_type;
+	device_initialize(disk_to_dev(disk));
 	return disk;
+
+out_free_part0:
+	hd_free_part(&disk->part0);
+out_free_disk:
+	kfree(disk);
+	return NULL;
 }
 EXPORT_SYMBOL(__alloc_disk_node);
 
@@ -2052,7 +2047,7 @@ void disk_flush_events(struct gendisk *disk, unsigned int mask)
  * CONTEXT:
  * Might sleep.
  */
-unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
+static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
 {
 	struct disk_events *ev = disk->ev;
 	unsigned int pending;
@@ -2090,6 +2085,33 @@ unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
 	return pending;
 }
 
+/**
+ * bdev_check_media_change - check if a removable media has been changed
+ * @bdev: block device to check
+ *
+ * Check whether a removable media has been changed, and attempt to free all
+ * dentries and inodes and invalidates all block device page cache entries in
+ * that case.
+ *
+ * Returns %true if the block device changed, or %false if not.
+ */
+bool bdev_check_media_change(struct block_device *bdev)
+{
+	unsigned int events;
+
+	events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE |
+				   DISK_EVENT_EJECT_REQUEST);
+	if (!(events & DISK_EVENT_MEDIA_CHANGE))
+		return false;
+
+	if (__invalidate_device(bdev, true))
+		pr_warn("VFS: busy inodes on changed media %s\n",
+			bdev->bd_disk->disk_name);
+	set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
+	return true;
+}
+EXPORT_SYMBOL(bdev_check_media_change);
+
 /*
  * Separate this part out so that a different pointer for clearing_ptr can be
  * passed in for disk_clear_events.
diff --git a/block/ioctl.c b/block/ioctl.c
index bdb3bbb253d9..3fbc382eb926 100644
--- a/block/ioctl.c
+++ b/block/ioctl.c
@@ -23,7 +23,7 @@ static int blkpg_do_ioctl(struct block_device *bdev,
 		return -EACCES;
 	if (copy_from_user(&p, upart, sizeof(struct blkpg_partition)))
 		return -EFAULT;
-	if (bdev != bdev->bd_contains)
+	if (bdev_is_partition(bdev))
 		return -EINVAL;
 
 	if (p.pno <= 0)
@@ -94,7 +94,7 @@ static int blkdev_reread_part(struct block_device *bdev)
 {
 	int ret;
 
-	if (!disk_part_scan_enabled(bdev->bd_disk) || bdev != bdev->bd_contains)
+	if (!disk_part_scan_enabled(bdev->bd_disk) || bdev_is_partition(bdev))
 		return -EINVAL;
 	if (!capable(CAP_SYS_ADMIN))
 		return -EACCES;
@@ -112,8 +112,7 @@ static int blk_ioctl_discard(struct block_device *bdev, fmode_t mode,
 	uint64_t range[2];
 	uint64_t start, len;
 	struct request_queue *q = bdev_get_queue(bdev);
-	struct address_space *mapping = bdev->bd_inode->i_mapping;
-
+	int err;
 
 	if (!(mode & FMODE_WRITE))
 		return -EBADF;
@@ -134,7 +133,11 @@ static int blk_ioctl_discard(struct block_device *bdev, fmode_t mode,
 
 	if (start + len > i_size_read(bdev->bd_inode))
 		return -EINVAL;
-	truncate_inode_pages_range(mapping, start, start + len - 1);
+
+	err = truncate_bdev_range(bdev, mode, start, start + len - 1);
+	if (err)
+		return err;
+
 	return blkdev_issue_discard(bdev, start >> 9, len >> 9,
 				    GFP_KERNEL, flags);
 }
@@ -143,8 +146,8 @@ static int blk_ioctl_zeroout(struct block_device *bdev, fmode_t mode,
 		unsigned long arg)
 {
 	uint64_t range[2];
-	struct address_space *mapping;
 	uint64_t start, end, len;
+	int err;
 
 	if (!(mode & FMODE_WRITE))
 		return -EBADF;
@@ -166,8 +169,9 @@ static int blk_ioctl_zeroout(struct block_device *bdev, fmode_t mode,
 		return -EINVAL;
 
 	/* Invalidate the page cache, including dirty pages */
-	mapping = bdev->bd_inode->i_mapping;
-	truncate_inode_pages_range(mapping, start, end);
+	err = truncate_bdev_range(bdev, mode, start, end);
+	if (err)
+		return err;
 
 	return blkdev_issue_zeroout(bdev, start >> 9, len >> 9, GFP_KERNEL,
 			BLKDEV_ZERO_NOUNMAP);
@@ -474,15 +478,14 @@ static int blkdev_bszset(struct block_device *bdev, fmode_t mode,
 	if (get_user(n, argp))
 		return -EFAULT;
 
-	if (!(mode & FMODE_EXCL)) {
-		bdgrab(bdev);
-		if (blkdev_get(bdev, mode | FMODE_EXCL, &bdev) < 0)
-			return -EBUSY;
-	}
+	if (mode & FMODE_EXCL)
+		return set_blocksize(bdev, n);
 
+	if (IS_ERR(blkdev_get_by_dev(bdev->bd_dev, mode | FMODE_EXCL, &bdev)))
+		return -EBUSY;
 	ret = set_blocksize(bdev, n);
-	if (!(mode & FMODE_EXCL))
-		blkdev_put(bdev, mode | FMODE_EXCL);
+	blkdev_put(bdev, mode | FMODE_EXCL);
+
 	return ret;
 }
 
diff --git a/block/ioprio.c b/block/ioprio.c
index 04ebd37966f1..364d2294ba90 100644
--- a/block/ioprio.c
+++ b/block/ioprio.c
@@ -69,7 +69,7 @@ int ioprio_check_cap(int ioprio)
 
 	switch (class) {
 		case IOPRIO_CLASS_RT:
-			if (!capable(CAP_SYS_ADMIN))
+			if (!capable(CAP_SYS_NICE) && !capable(CAP_SYS_ADMIN))
 				return -EPERM;
 			fallthrough;
 			/* rt has prio field too */
diff --git a/block/kyber-iosched.c b/block/kyber-iosched.c
index a38c5ab103d1..dc89199bc8c6 100644
--- a/block/kyber-iosched.c
+++ b/block/kyber-iosched.c
@@ -359,7 +359,7 @@ 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 q->queue_hw_ctx[0]->sched_tags->bitmap_tags.sb.shift;
+	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)
@@ -502,7 +502,7 @@ static int kyber_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int hctx_idx)
 	khd->batching = 0;
 
 	hctx->sched_data = khd;
-	sbitmap_queue_min_shallow_depth(&hctx->sched_tags->bitmap_tags,
+	sbitmap_queue_min_shallow_depth(hctx->sched_tags->bitmap_tags,
 					kqd->async_depth);
 
 	return 0;
@@ -573,7 +573,7 @@ static bool kyber_bio_merge(struct blk_mq_hw_ctx *hctx, struct bio *bio,
 	bool merged;
 
 	spin_lock(&kcq->lock);
-	merged = blk_mq_bio_list_merge(hctx->queue, rq_list, bio, nr_segs);
+	merged = blk_bio_list_merge(hctx->queue, rq_list, bio, nr_segs);
 	spin_unlock(&kcq->lock);
 
 	return merged;
diff --git a/block/mq-deadline.c b/block/mq-deadline.c
index b57470e154c8..800ac902809b 100644
--- a/block/mq-deadline.c
+++ b/block/mq-deadline.c
@@ -386,6 +386,8 @@ static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
 	spin_lock(&dd->lock);
 	rq = __dd_dispatch_request(dd);
 	spin_unlock(&dd->lock);
+	if (rq)
+		atomic_dec(&rq->mq_hctx->elevator_queued);
 
 	return rq;
 }
@@ -533,6 +535,7 @@ static void dd_insert_requests(struct blk_mq_hw_ctx *hctx,
 		rq = list_first_entry(list, struct request, queuelist);
 		list_del_init(&rq->queuelist);
 		dd_insert_request(hctx, rq, at_head);
+		atomic_inc(&hctx->elevator_queued);
 	}
 	spin_unlock(&dd->lock);
 }
@@ -579,6 +582,9 @@ static bool dd_has_work(struct blk_mq_hw_ctx *hctx)
 {
 	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
 
+	if (!atomic_read(&hctx->elevator_queued))
+		return false;
+
 	return !list_empty_careful(&dd->dispatch) ||
 		!list_empty_careful(&dd->fifo_list[0]) ||
 		!list_empty_careful(&dd->fifo_list[1]);
diff --git a/block/partitions/core.c b/block/partitions/core.c
index 722406b841df..5cacbac30107 100644
--- a/block/partitions/core.c
+++ b/block/partitions/core.c
@@ -199,14 +199,20 @@ static ssize_t part_alignment_offset_show(struct device *dev,
 					  struct device_attribute *attr, char *buf)
 {
 	struct hd_struct *p = dev_to_part(dev);
-	return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
+
+	return sprintf(buf, "%u\n",
+		queue_limit_alignment_offset(&part_to_disk(p)->queue->limits,
+				p->start_sect));
 }
 
 static ssize_t part_discard_alignment_show(struct device *dev,
 					   struct device_attribute *attr, char *buf)
 {
 	struct hd_struct *p = dev_to_part(dev);
-	return sprintf(buf, "%u\n", p->discard_alignment);
+
+	return sprintf(buf, "%u\n",
+		queue_limit_discard_alignment(&part_to_disk(p)->queue->limits,
+				p->start_sect));
 }
 
 static DEVICE_ATTR(partition, 0444, part_partition_show, NULL);
@@ -318,8 +324,9 @@ int hd_ref_init(struct hd_struct *part)
  * Must be called either with bd_mutex held, before a disk can be opened or
  * after all disk users are gone.
  */
-void delete_partition(struct gendisk *disk, struct hd_struct *part)
+void delete_partition(struct hd_struct *part)
 {
+	struct gendisk *disk = part_to_disk(part);
 	struct disk_part_tbl *ptbl =
 		rcu_dereference_protected(disk->part_tbl, 1);
 
@@ -327,7 +334,7 @@ void delete_partition(struct gendisk *disk, struct hd_struct *part)
 	 * ->part_tbl is referenced in this part's release handler, so
 	 *  we have to hold the disk device
 	 */
-	get_device(disk_to_dev(part_to_disk(part)));
+	get_device(disk_to_dev(disk));
 	rcu_assign_pointer(ptbl->part[part->partno], NULL);
 	kobject_put(part->holder_dir);
 	device_del(part_to_dev(part));
@@ -405,10 +412,6 @@ static struct hd_struct *add_partition(struct gendisk *disk, int partno,
 	pdev = part_to_dev(p);
 
 	p->start_sect = start;
-	p->alignment_offset =
-		queue_limit_alignment_offset(&disk->queue->limits, start);
-	p->discard_alignment =
-		queue_limit_discard_alignment(&disk->queue->limits, start);
 	p->nr_sects = len;
 	p->partno = partno;
 	p->policy = get_disk_ro(disk);
@@ -554,7 +557,7 @@ int bdev_del_partition(struct block_device *bdev, int partno)
 	sync_blockdev(bdevp);
 	invalidate_bdev(bdevp);
 
-	delete_partition(bdev->bd_disk, part);
+	delete_partition(part);
 	ret = 0;
 out_unlock:
 	mutex_unlock(&bdev->bd_mutex);
@@ -592,8 +595,8 @@ int bdev_resize_partition(struct block_device *bdev, int partno,
 	if (partition_overlaps(bdev->bd_disk, start, length, partno))
 		goto out_unlock;
 
-	part_nr_sects_write(part, (sector_t)length);
-	i_size_write(bdevp->bd_inode, length << SECTOR_SHIFT);
+	part_nr_sects_write(part, length);
+	bd_set_nr_sectors(bdevp, length);
 
 	ret = 0;
 out_unlock:
@@ -634,7 +637,7 @@ int blk_drop_partitions(struct block_device *bdev)
 
 	disk_part_iter_init(&piter, bdev->bd_disk, DISK_PITER_INCL_EMPTY);
 	while ((part = disk_part_iter_next(&piter)))
-		delete_partition(bdev->bd_disk, part);
+		delete_partition(part);
 	disk_part_iter_exit(&piter);
 
 	return 0;
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index ef722f04f88a..d4abd1ed5a2d 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -37,8 +37,6 @@ const unsigned char scsi_command_size_tbl[8] =
 };
 EXPORT_SYMBOL(scsi_command_size_tbl);
 
-#include <scsi/sg.h>
-
 static int sg_get_version(int __user *p)
 {
 	static const int sg_version_num = 30527;
@@ -854,7 +852,7 @@ EXPORT_SYMBOL(scsi_cmd_ioctl);
 
 int scsi_verify_blk_ioctl(struct block_device *bd, unsigned int cmd)
 {
-	if (bd && bd == bd->bd_contains)
+	if (bd && !bdev_is_partition(bd))
 		return 0;
 
 	if (capable(CAP_SYS_RAWIO))