Merge remote-tracking branch 'block/for-next'

15 files changed, 903 insertions, 472 deletions

diff --git a/block/blk-core.c b/block/blk-core.c
index a0e3096c4bb5..a6bd3e702201 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -146,8 +146,8 @@ void blk_dump_rq_flags(struct request *rq, char *msg)
 	printk(KERN_INFO "  sector %llu, nr/cnr %u/%u\n",
 	       (unsigned long long)blk_rq_pos(rq),
 	       blk_rq_sectors(rq), blk_rq_cur_sectors(rq));
-	printk(KERN_INFO "  bio %p, biotail %p, buffer %p, len %u\n",
-	       rq->bio, rq->biotail, rq->buffer, blk_rq_bytes(rq));
+	printk(KERN_INFO "  bio %p, biotail %p, len %u\n",
+	       rq->bio, rq->biotail, blk_rq_bytes(rq));
 
 	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
 		printk(KERN_INFO "  cdb: ");
@@ -251,8 +251,10 @@ void blk_sync_queue(struct request_queue *q)
 		struct blk_mq_hw_ctx *hctx;
 		int i;
 
-		queue_for_each_hw_ctx(q, hctx, i)
-			cancel_delayed_work_sync(&hctx->delayed_work);
+		queue_for_each_hw_ctx(q, hctx, i) {
+			cancel_delayed_work_sync(&hctx->run_work);
+			cancel_delayed_work_sync(&hctx->delay_work);
+		}
 	} else {
 		cancel_delayed_work_sync(&q->delay_work);
 	}
@@ -1231,12 +1233,15 @@ static void add_acct_request(struct request_queue *q, struct request *rq,
 static void part_round_stats_single(int cpu, struct hd_struct *part,
 				    unsigned long now)
 {
+	int inflight;
+
 	if (now == part->stamp)
 		return;
 
-	if (part_in_flight(part)) {
+	inflight = part_in_flight(part);
+	if (inflight) {
 		__part_stat_add(cpu, part, time_in_queue,
-				part_in_flight(part) * (now - part->stamp));
+				inflight * (now - part->stamp));
 		__part_stat_add(cpu, part, io_ticks, (now - part->stamp));
 	}
 	part->stamp = now;
@@ -1360,7 +1365,6 @@ void blk_add_request_payload(struct request *rq, struct page *page,
 
 	rq->__data_len = rq->resid_len = len;
 	rq->nr_phys_segments = 1;
-	rq->buffer = bio_data(bio);
 }
 EXPORT_SYMBOL_GPL(blk_add_request_payload);
 
@@ -1402,12 +1406,6 @@ bool bio_attempt_front_merge(struct request_queue *q, struct request *req,
 	bio->bi_next = req->bio;
 	req->bio = bio;
 
-	/*
-	 * may not be valid. if the low level driver said
-	 * it didn't need a bounce buffer then it better
-	 * not touch req->buffer either...
-	 */
-	req->buffer = bio_data(bio);
 	req->__sector = bio->bi_iter.bi_sector;
 	req->__data_len += bio->bi_iter.bi_size;
 	req->ioprio = ioprio_best(req->ioprio, bio_prio(bio));
@@ -1654,7 +1652,7 @@ static int __init fail_make_request_debugfs(void)
 	struct dentry *dir = fault_create_debugfs_attr("fail_make_request",
 						NULL, &fail_make_request);
 
-	return IS_ERR(dir) ? PTR_ERR(dir) : 0;
+	return PTR_ERR_OR_ZERO(dir);
 }
 
 late_initcall(fail_make_request_debugfs);
@@ -2434,7 +2432,6 @@ bool blk_update_request(struct request *req, int error, unsigned int nr_bytes)
 	}
 
 	req->__data_len -= total_bytes;
-	req->buffer = bio_data(req->bio);
 
 	/* update sector only for requests with clear definition of sector */
 	if (req->cmd_type == REQ_TYPE_FS)
@@ -2503,7 +2500,7 @@ EXPORT_SYMBOL_GPL(blk_unprep_request);
 /*
  * queue lock must be held
  */
-static void blk_finish_request(struct request *req, int error)
+void blk_finish_request(struct request *req, int error)
 {
 	if (blk_rq_tagged(req))
 		blk_queue_end_tag(req->q, req);
@@ -2529,6 +2526,7 @@ static void blk_finish_request(struct request *req, int error)
 		__blk_put_request(req->q, req);
 	}
 }
+EXPORT_SYMBOL(blk_finish_request);
 
 /**
  * blk_end_bidi_request - Complete a bidi request
@@ -2752,10 +2750,9 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
 	/* Bit 0 (R/W) is identical in rq->cmd_flags and bio->bi_rw */
 	rq->cmd_flags |= bio->bi_rw & REQ_WRITE;
 
-	if (bio_has_data(bio)) {
+	if (bio_has_data(bio))
 		rq->nr_phys_segments = bio_phys_segments(q, bio);
-		rq->buffer = bio_data(bio);
-	}
+
 	rq->__data_len = bio->bi_iter.bi_size;
 	rq->bio = rq->biotail = bio;
 
@@ -2831,7 +2828,7 @@ EXPORT_SYMBOL_GPL(blk_rq_unprep_clone);
 
 /*
  * Copy attributes of the original request to the clone request.
- * The actual data parts (e.g. ->cmd, ->buffer, ->sense) are not copied.
+ * The actual data parts (e.g. ->cmd, ->sense) are not copied.
  */
 static void __blk_rq_prep_clone(struct request *dst, struct request *src)
 {
@@ -2857,7 +2854,7 @@ static void __blk_rq_prep_clone(struct request *dst, struct request *src)
  *
  * Description:
  *     Clones bios in @rq_src to @rq, and copies attributes of @rq_src to @rq.
- *     The actual data parts of @rq_src (e.g. ->cmd, ->buffer, ->sense)
+ *     The actual data parts of @rq_src (e.g. ->cmd, ->sense)
  *     are not copied, and copying such parts is the caller's responsibility.
  *     Also, pages which the original bios are pointing to are not copied
  *     and the cloned bios just point same pages.
@@ -2904,19 +2901,26 @@ free_and_out:
 }
 EXPORT_SYMBOL_GPL(blk_rq_prep_clone);
 
-int kblockd_schedule_work(struct request_queue *q, struct work_struct *work)
+int kblockd_schedule_work(struct work_struct *work)
 {
 	return queue_work(kblockd_workqueue, work);
 }
 EXPORT_SYMBOL(kblockd_schedule_work);
 
-int kblockd_schedule_delayed_work(struct request_queue *q,
-			struct delayed_work *dwork, unsigned long delay)
+int kblockd_schedule_delayed_work(struct delayed_work *dwork,
+				  unsigned long delay)
 {
 	return queue_delayed_work(kblockd_workqueue, dwork, delay);
 }
 EXPORT_SYMBOL(kblockd_schedule_delayed_work);
 
+int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork,
+				     unsigned long delay)
+{
+	return queue_delayed_work_on(cpu, kblockd_workqueue, dwork, delay);
+}
+EXPORT_SYMBOL(kblockd_schedule_delayed_work_on);
+
 #define PLUG_MAGIC	0x91827364
 
 /**
diff --git a/block/blk-flush.c b/block/blk-flush.c
index 43e6b4755e9a..ec7a224d6733 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -134,7 +134,7 @@ static void mq_flush_run(struct work_struct *work)
 {
 	struct request *rq;
 
-	rq = container_of(work, struct request, mq_flush_work);
+	rq = container_of(work, struct request, requeue_work);
 
 	memset(&rq->csd, 0, sizeof(rq->csd));
 	blk_mq_insert_request(rq, false, true, false);
@@ -143,8 +143,8 @@ static void mq_flush_run(struct work_struct *work)
 static bool blk_flush_queue_rq(struct request *rq, bool add_front)
 {
 	if (rq->q->mq_ops) {
-		INIT_WORK(&rq->mq_flush_work, mq_flush_run);
-		kblockd_schedule_work(rq->q, &rq->mq_flush_work);
+		INIT_WORK(&rq->requeue_work, mq_flush_run);
+		kblockd_schedule_work(&rq->requeue_work);
 		return false;
 	} else {
 		if (add_front)
@@ -306,23 +306,9 @@ static bool blk_kick_flush(struct request_queue *q)
 	 */
 	q->flush_pending_idx ^= 1;
 
-	if (q->mq_ops) {
-		struct blk_mq_ctx *ctx = first_rq->mq_ctx;
-		struct blk_mq_hw_ctx *hctx = q->mq_ops->map_queue(q, ctx->cpu);
-
-		blk_mq_rq_init(hctx, q->flush_rq);
-		q->flush_rq->mq_ctx = ctx;
-
-		/*
-		 * Reuse the tag value from the fist waiting request,
-		 * with blk-mq the tag is generated during request
-		 * allocation and drivers can rely on it being inside
-		 * the range they asked for.
-		 */
-		q->flush_rq->tag = first_rq->tag;
-	} else {
-		blk_rq_init(q, q->flush_rq);
-	}
+	blk_rq_init(q, q->flush_rq);
+	if (q->mq_ops)
+		blk_mq_clone_flush_request(q->flush_rq, first_rq);
 
 	q->flush_rq->cmd_type = REQ_TYPE_FS;
 	q->flush_rq->cmd_flags = WRITE_FLUSH | REQ_FLUSH_SEQ;
diff --git a/block/blk-iopoll.c b/block/blk-iopoll.c
index c11d24e379e2..d828b44a404b 100644
--- a/block/blk-iopoll.c
+++ b/block/blk-iopoll.c
@@ -64,12 +64,12 @@ EXPORT_SYMBOL(__blk_iopoll_complete);
  *     iopoll handler will not be invoked again before blk_iopoll_sched_prep()
  *     is called.
  **/
-void blk_iopoll_complete(struct blk_iopoll *iopoll)
+void blk_iopoll_complete(struct blk_iopoll *iop)
 {
 	unsigned long flags;
 
 	local_irq_save(flags);
-	__blk_iopoll_complete(iopoll);
+	__blk_iopoll_complete(iop);
 	local_irq_restore(flags);
 }
 EXPORT_SYMBOL(blk_iopoll_complete);
diff --git a/block/blk-map.c b/block/blk-map.c
index f7b22bc21518..f890d4345b0c 100644
--- a/block/blk-map.c
+++ b/block/blk-map.c
@@ -155,7 +155,6 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq,
 	if (!bio_flagged(bio, BIO_USER_MAPPED))
 		rq->cmd_flags |= REQ_COPY_USER;
 
-	rq->buffer = NULL;
 	return 0;
 unmap_rq:
 	blk_rq_unmap_user(bio);
@@ -238,7 +237,6 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq,
 	blk_queue_bounce(q, &bio);
 	bio_get(bio);
 	blk_rq_bio_prep(q, rq, bio);
-	rq->buffer = NULL;
 	return 0;
 }
 EXPORT_SYMBOL(blk_rq_map_user_iov);
@@ -325,7 +323,6 @@ int blk_rq_map_kern(struct request_queue *q, struct request *rq, void *kbuf,
 	}
 
 	blk_queue_bounce(q, &rq->bio);
-	rq->buffer = NULL;
 	return 0;
 }
 EXPORT_SYMBOL(blk_rq_map_kern);
diff --git a/block/blk-mq-cpumap.c b/block/blk-mq-cpumap.c
index 097921329619..5d0f93cf358c 100644
--- a/block/blk-mq-cpumap.c
+++ b/block/blk-mq-cpumap.c
@@ -80,17 +80,17 @@ int blk_mq_update_queue_map(unsigned int *map, unsigned int nr_queues)
 	return 0;
 }
 
-unsigned int *blk_mq_make_queue_map(struct blk_mq_reg *reg)
+unsigned int *blk_mq_make_queue_map(struct blk_mq_tag_set *set)
 {
 	unsigned int *map;
 
 	/* If cpus are offline, map them to first hctx */
 	map = kzalloc_node(sizeof(*map) * num_possible_cpus(), GFP_KERNEL,
-				reg->numa_node);
+				set->numa_node);
 	if (!map)
 		return NULL;
 
-	if (!blk_mq_update_queue_map(map, reg->nr_hw_queues))
+	if (!blk_mq_update_queue_map(map, set->nr_hw_queues))
 		return map;
 
 	kfree(map);
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index b0ba264b0522..8145b5b25b4b 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -203,42 +203,6 @@ static ssize_t blk_mq_hw_sysfs_rq_list_show(struct blk_mq_hw_ctx *hctx,
 	return ret;
 }
 
-static ssize_t blk_mq_hw_sysfs_ipi_show(struct blk_mq_hw_ctx *hctx, char *page)
-{
-	ssize_t ret;
-
-	spin_lock(&hctx->lock);
-	ret = sprintf(page, "%u\n", !!(hctx->flags & BLK_MQ_F_SHOULD_IPI));
-	spin_unlock(&hctx->lock);
-
-	return ret;
-}
-
-static ssize_t blk_mq_hw_sysfs_ipi_store(struct blk_mq_hw_ctx *hctx,
-					 const char *page, size_t len)
-{
-	struct blk_mq_ctx *ctx;
-	unsigned long ret;
-	unsigned int i;
-
-	if (kstrtoul(page, 10, &ret)) {
-		pr_err("blk-mq-sysfs: invalid input '%s'\n", page);
-		return -EINVAL;
-	}
-
-	spin_lock(&hctx->lock);
-	if (ret)
-		hctx->flags |= BLK_MQ_F_SHOULD_IPI;
-	else
-		hctx->flags &= ~BLK_MQ_F_SHOULD_IPI;
-	spin_unlock(&hctx->lock);
-
-	hctx_for_each_ctx(hctx, ctx, i)
-		ctx->ipi_redirect = !!ret;
-
-	return len;
-}
-
 static ssize_t blk_mq_hw_sysfs_tags_show(struct blk_mq_hw_ctx *hctx, char *page)
 {
 	return blk_mq_tag_sysfs_show(hctx->tags, page);
@@ -246,16 +210,12 @@ static ssize_t blk_mq_hw_sysfs_tags_show(struct blk_mq_hw_ctx *hctx, char *page)
 
 static ssize_t blk_mq_hw_sysfs_cpus_show(struct blk_mq_hw_ctx *hctx, char *page)
 {
-	unsigned int i, queue_num, first = 1;
+	unsigned int i, first = 1;
 	ssize_t ret = 0;
 
 	blk_mq_disable_hotplug();
 
-	for_each_online_cpu(i) {
-		queue_num = hctx->queue->mq_map[i];
-		if (queue_num != hctx->queue_num)
-			continue;
-
+	for_each_cpu(i, hctx->cpumask) {
 		if (first)
 			ret += sprintf(ret + page, "%u", i);
 		else
@@ -311,11 +271,6 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_pending = {
 	.attr = {.name = "pending", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_rq_list_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_ipi = {
-	.attr = {.name = "ipi_redirect", .mode = S_IRUGO | S_IWUSR},
-	.show = blk_mq_hw_sysfs_ipi_show,
-	.store = blk_mq_hw_sysfs_ipi_store,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_tags = {
 	.attr = {.name = "tags", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_tags_show,
@@ -330,7 +285,6 @@ static struct attribute *default_hw_ctx_attrs[] = {
 	&blk_mq_hw_sysfs_run.attr,
 	&blk_mq_hw_sysfs_dispatched.attr,
 	&blk_mq_hw_sysfs_pending.attr,
-	&blk_mq_hw_sysfs_ipi.attr,
 	&blk_mq_hw_sysfs_tags.attr,
 	&blk_mq_hw_sysfs_cpus.attr,
 	NULL,
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 83ae96c51a27..8d526a3e02f6 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -1,78 +1,260 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
-#include <linux/percpu_ida.h>
+#include <linux/random.h>
 
 #include <linux/blk-mq.h>
 #include "blk.h"
 #include "blk-mq.h"
 #include "blk-mq-tag.h"
 
+void blk_mq_wait_for_tags(struct blk_mq_tags *tags, struct blk_mq_hw_ctx *hctx,
+			  bool reserved)
+{
+	int tag, zero = 0;
+
+	tag = blk_mq_get_tag(tags, hctx, &zero, __GFP_WAIT, reserved);
+	blk_mq_put_tag(tags, tag, &zero);
+}
+
+static bool bt_has_free_tags(struct blk_mq_bitmap_tags *bt)
+{
+	int i;
+
+	for (i = 0; i < bt->map_nr; i++) {
+		struct blk_mq_bitmap *bm = &bt->map[i];
+		int ret;
+
+		ret = find_first_zero_bit(&bm->word, bm->depth);
+		if (ret < bm->depth)
+			return true;
+	}
+
+	return false;
+}
+
+bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
+{
+	if (!tags)
+		return true;
+
+	return bt_has_free_tags(&tags->bitmap_tags);
+}
+
+static int __bt_get_word(struct blk_mq_bitmap *bm, unsigned int last_tag)
+{
+	int tag, org_last_tag, end;
+
+	org_last_tag = last_tag;
+	end = bm->depth;
+	do {
+restart:
+		tag = find_next_zero_bit(&bm->word, end, last_tag);
+		if (unlikely(tag >= end)) {
+			/*
+			 * We started with an offset, start from 0 to
+			 * exhaust the map.
+			 */
+			if (org_last_tag && last_tag) {
+				end = last_tag;
+				last_tag = 0;
+				goto restart;
+			}
+			return -1;
+		}
+		last_tag = tag + 1;
+	} while (test_and_set_bit_lock(tag, &bm->word));
+
+	return tag;
+}
+
 /*
- * Per tagged queue (tag address space) map
+ * Straight forward bitmap tag implementation, where each bit is a tag
+ * (cleared == free, and set == busy). The small twist is using per-cpu
+ * last_tag caches, which blk-mq stores in the blk_mq_ctx software queue
+ * contexts. This enables us to drastically limit the space searched,
+ * without dirtying an extra shared cacheline like we would if we stored
+ * the cache value inside the shared blk_mq_bitmap_tags structure. On top
+ * of that, each word of tags is in a separate cacheline. This means that
+ * multiple users will tend to stick to different cachelines, at least
+ * until the map is exhausted.
  */
-struct blk_mq_tags {
-	unsigned int nr_tags;
-	unsigned int nr_reserved_tags;
-	unsigned int nr_batch_move;
-	unsigned int nr_max_cache;
+static int __bt_get(struct blk_mq_bitmap_tags *bt, unsigned int *tag_cache)
+{
+	unsigned int last_tag, org_last_tag;
+	int index, i, tag;
+
+	last_tag = org_last_tag = *tag_cache;
+	index = TAG_TO_INDEX(bt, last_tag);
+
+	for (i = 0; i < bt->map_nr; i++) {
+		tag = __bt_get_word(&bt->map[index], TAG_TO_BIT(bt, last_tag));
+		if (tag != -1) {
+			tag += (index << bt->bits_per_word);
+			goto done;
+		}
+
+		last_tag = 0;
+		if (++index >= bt->map_nr)
+			index = 0;
+	}
 
-	struct percpu_ida free_tags;
-	struct percpu_ida reserved_tags;
-};
+	*tag_cache = 0;
+	return -1;
 
-void blk_mq_wait_for_tags(struct blk_mq_tags *tags)
+	/*
+	 * Only update the cache from the allocation path, if we ended
+	 * up using the specific cached tag.
+	 */
+done:
+	if (tag == org_last_tag) {
+		last_tag = tag + 1;
+		if (last_tag >= bt->depth - 1)
+			last_tag = 0;
+
+		*tag_cache = last_tag;
+	}
+
+	return tag;
+}
+
+static inline void bt_index_inc(unsigned int *index)
 {
-	int tag = blk_mq_get_tag(tags, __GFP_WAIT, false);
-	blk_mq_put_tag(tags, tag);
+	*index = (*index + 1) & (BT_WAIT_QUEUES - 1);
 }
 
-bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
+static struct bt_wait_state *bt_wait_ptr(struct blk_mq_bitmap_tags *bt,
+					 struct blk_mq_hw_ctx *hctx)
 {
-	return !tags ||
-		percpu_ida_free_tags(&tags->free_tags, nr_cpu_ids) != 0;
+	struct bt_wait_state *bs;
+
+	if (!hctx)
+		return &bt->bs[0];
+
+	bs = &bt->bs[hctx->wait_index];
+	bt_index_inc(&hctx->wait_index);
+	return bs;
 }
 
-static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp)
+static int bt_get(struct blk_mq_bitmap_tags *bt, struct blk_mq_hw_ctx *hctx,
+		  unsigned int *last_tag, gfp_t gfp)
 {
+	struct bt_wait_state *bs;
+	DEFINE_WAIT(wait);
 	int tag;
 
-	tag = percpu_ida_alloc(&tags->free_tags, (gfp & __GFP_WAIT) ?
-			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
-	if (tag < 0)
-		return BLK_MQ_TAG_FAIL;
-	return tag + tags->nr_reserved_tags;
+	tag = __bt_get(bt, last_tag);
+	if (tag != -1)
+		return tag;
+
+	if (!(gfp & __GFP_WAIT))
+		return -1;
+
+	bs = bt_wait_ptr(bt, hctx);
+	do {
+		bool was_empty;
+
+		was_empty = list_empty(&wait.task_list);
+		prepare_to_wait(&bs->wait, &wait, TASK_UNINTERRUPTIBLE);
+
+		tag = __bt_get(bt, last_tag);
+		if (tag != -1)
+			break;
+
+		if (was_empty)
+			atomic_set(&bs->wait_cnt, bt->wake_cnt);
+
+		io_schedule();
+	} while (1);
+
+	finish_wait(&bs->wait, &wait);
+	return tag;
+}
+
+static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags,
+				     struct blk_mq_hw_ctx *hctx,
+				     unsigned int *last_tag, gfp_t gfp)
+{
+	int tag;
+
+	tag = bt_get(&tags->bitmap_tags, hctx, last_tag, gfp);
+	if (tag >= 0)
+		return tag + tags->nr_reserved_tags;
+
+	return BLK_MQ_TAG_FAIL;
 }
 
 static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_tags *tags,
 					      gfp_t gfp)
 {
-	int tag;
+	int tag, zero = 0;
 
 	if (unlikely(!tags->nr_reserved_tags)) {
 		WARN_ON_ONCE(1);
 		return BLK_MQ_TAG_FAIL;
 	}
 
-	tag = percpu_ida_alloc(&tags->reserved_tags, (gfp & __GFP_WAIT) ?
-			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
+	tag = bt_get(&tags->breserved_tags, NULL, &zero, gfp);
 	if (tag < 0)
 		return BLK_MQ_TAG_FAIL;
+
 	return tag;
 }
 
-unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp, bool reserved)
+unsigned int blk_mq_get_tag(struct blk_mq_tags *tags,
+			    struct blk_mq_hw_ctx *hctx, unsigned int *last_tag,
+			    gfp_t gfp, bool reserved)
 {
 	if (!reserved)
-		return __blk_mq_get_tag(tags, gfp);
+		return __blk_mq_get_tag(tags, hctx, last_tag, gfp);
 
 	return __blk_mq_get_reserved_tag(tags, gfp);
 }
 
+static struct bt_wait_state *bt_wake_ptr(struct blk_mq_bitmap_tags *bt)
+{
+	int i, wake_index;
+
+	wake_index = bt->wake_index;
+	for (i = 0; i < BT_WAIT_QUEUES; i++) {
+		struct bt_wait_state *bs = &bt->bs[wake_index];
+
+		if (waitqueue_active(&bs->wait)) {
+			if (wake_index != bt->wake_index)
+				bt->wake_index = wake_index;
+
+			return bs;
+		}
+
+		bt_index_inc(&wake_index);
+	}
+
+	return NULL;
+}
+
+static void bt_clear_tag(struct blk_mq_bitmap_tags *bt, unsigned int tag)
+{
+	const int index = TAG_TO_INDEX(bt, tag);
+	struct bt_wait_state *bs;
+
+	/*
+	 * The unlock memory barrier need to order access to req in free
+	 * path and clearing tag bit
+	 */
+	clear_bit_unlock(TAG_TO_BIT(bt, tag), &bt->map[index].word);
+
+	bs = bt_wake_ptr(bt);
+	if (bs && atomic_dec_and_test(&bs->wait_cnt)) {
+		atomic_set(&bs->wait_cnt, bt->wake_cnt);
+		bt_index_inc(&bt->wake_index);
+		wake_up(&bs->wait);
+	}
+}
+
 static void __blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag)
 {
 	BUG_ON(tag >= tags->nr_tags);
 
-	percpu_ida_free(&tags->free_tags, tag - tags->nr_reserved_tags);
+	bt_clear_tag(&tags->bitmap_tags, tag);
 }
 
 static void __blk_mq_put_reserved_tag(struct blk_mq_tags *tags,
@@ -80,22 +262,41 @@ static void __blk_mq_put_reserved_tag(struct blk_mq_tags *tags,
 {
 	BUG_ON(tag >= tags->nr_reserved_tags);
 
-	percpu_ida_free(&tags->reserved_tags, tag);
+	bt_clear_tag(&tags->breserved_tags, tag);
 }
 
-void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag)
+void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag,
+		    unsigned int *last_tag)
 {
-	if (tag >= tags->nr_reserved_tags)
-		__blk_mq_put_tag(tags, tag);
-	else
+	if (tag >= tags->nr_reserved_tags) {
+		const int real_tag = tag - tags->nr_reserved_tags;
+
+		__blk_mq_put_tag(tags, real_tag);
+		*last_tag = real_tag;
+	} else
 		__blk_mq_put_reserved_tag(tags, tag);
 }
 
-static int __blk_mq_tag_iter(unsigned id, void *data)
+static void bt_for_each_free(struct blk_mq_bitmap_tags *bt,
+			     unsigned long *free_map, unsigned int off)
 {
-	unsigned long *tag_map = data;
-	__set_bit(id, tag_map);
-	return 0;
+	int i;
+
+	for (i = 0; i < bt->map_nr; i++) {
+		struct blk_mq_bitmap *bm = &bt->map[i];
+		int bit = 0;
+
+		do {
+			bit = find_next_zero_bit(&bm->word, bm->depth, bit);
+			if (bit >= bm->depth)
+				break;
+
+			__set_bit(bit + off, free_map);
+			bit++;
+		} while (1);
+
+		off += (1 << bt->bits_per_word);
+	}
 }
 
 void blk_mq_tag_busy_iter(struct blk_mq_tags *tags,
@@ -109,21 +310,114 @@ void blk_mq_tag_busy_iter(struct blk_mq_tags *tags,
 	if (!tag_map)
 		return;
 
-	percpu_ida_for_each_free(&tags->free_tags, __blk_mq_tag_iter, tag_map);
+	bt_for_each_free(&tags->bitmap_tags, tag_map, tags->nr_reserved_tags);
 	if (tags->nr_reserved_tags)
-		percpu_ida_for_each_free(&tags->reserved_tags, __blk_mq_tag_iter,
-			tag_map);
+		bt_for_each_free(&tags->breserved_tags, tag_map, 0);
 
 	fn(data, tag_map);
 	kfree(tag_map);
 }
 
+static unsigned int bt_unused_tags(struct blk_mq_bitmap_tags *bt)
+{
+	unsigned int i, used;
+
+	for (i = 0, used = 0; i < bt->map_nr; i++) {
+		struct blk_mq_bitmap *bm = &bt->map[i];
+
+		used += bitmap_weight(&bm->word, bm->depth);
+	}
+
+	return bt->depth - used;
+}
+
+static int bt_alloc(struct blk_mq_bitmap_tags *bt, unsigned int depth,
+			int node, bool reserved)
+{
+	int i;
+
+	bt->bits_per_word = ilog2(BITS_PER_LONG);
+
+	/*
+	 * Depth can be zero for reserved tags, that's not a failure
+	 * condition.
+	 */
+	if (depth) {
+		unsigned int nr, i, map_depth, tags_per_word;
+
+		tags_per_word = (1 << bt->bits_per_word);
+
+		/*
+		 * If the tag space is small, shrink the number of tags
+		 * per word so we spread over a few cachelines, at least.
+		 * If less than 4 tags, just forget about it, it's not
+		 * going to work optimally anyway.
+		 */
+		if (depth >= 4) {
+			while (tags_per_word * 4 > depth) {
+				bt->bits_per_word--;
+				tags_per_word = (1 << bt->bits_per_word);
+			}
+		}
+
+		nr = ALIGN(depth, tags_per_word) / tags_per_word;
+		bt->map = kzalloc_node(nr * sizeof(struct blk_mq_bitmap),
+						GFP_KERNEL, node);
+		if (!bt->map)
+			return -ENOMEM;
+
+		bt->map_nr = nr;
+		map_depth = depth;
+		for (i = 0; i < nr; i++) {
+			bt->map[i].depth = min(map_depth, tags_per_word);
+			map_depth -= tags_per_word;
+		}
+	}
+
+	bt->bs = kzalloc(BT_WAIT_QUEUES * sizeof(*bt->bs), GFP_KERNEL);
+	if (!bt->bs) {
+		kfree(bt->map);
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < BT_WAIT_QUEUES; i++)
+		init_waitqueue_head(&bt->bs[i].wait);
+
+	bt->wake_cnt = BT_WAIT_BATCH;
+	if (bt->wake_cnt > depth / 4)
+		bt->wake_cnt = max(1U, depth / 4);
+
+	bt->depth = depth;
+	return 0;
+}
+
+static void bt_free(struct blk_mq_bitmap_tags *bt)
+{
+	kfree(bt->map);
+	kfree(bt->bs);
+}
+
+static struct blk_mq_tags *blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
+						   int node)
+{
+	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
+
+	if (bt_alloc(&tags->bitmap_tags, depth, node, false))
+		goto enomem;
+	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, node, true))
+		goto enomem;
+
+	return tags;
+enomem:
+	bt_free(&tags->bitmap_tags);
+	kfree(tags);
+	return NULL;
+}
+
 struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
 				     unsigned int reserved_tags, int node)
 {
-	unsigned int nr_tags, nr_cache;
 	struct blk_mq_tags *tags;
-	int ret;
 
 	if (total_tags > BLK_MQ_TAG_MAX) {
 		pr_err("blk-mq: tag depth too large\n");
@@ -134,73 +428,43 @@ struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
 	if (!tags)
 		return NULL;
 
-	nr_tags = total_tags - reserved_tags;
-	nr_cache = nr_tags / num_possible_cpus();
-
-	if (nr_cache < BLK_MQ_TAG_CACHE_MIN)
-		nr_cache = BLK_MQ_TAG_CACHE_MIN;
-	else if (nr_cache > BLK_MQ_TAG_CACHE_MAX)
-		nr_cache = BLK_MQ_TAG_CACHE_MAX;
-
 	tags->nr_tags = total_tags;
 	tags->nr_reserved_tags = reserved_tags;
-	tags->nr_max_cache = nr_cache;
-	tags->nr_batch_move = max(1u, nr_cache / 2);
-
-	ret = __percpu_ida_init(&tags->free_tags, tags->nr_tags -
-				tags->nr_reserved_tags,
-				tags->nr_max_cache,
-				tags->nr_batch_move);
-	if (ret)
-		goto err_free_tags;
-
-	if (reserved_tags) {
-		/*
-		 * With max_cahe and batch set to 1, the allocator fallbacks to
-		 * no cached. It's fine reserved tags allocation is slow.
-		 */
-		ret = __percpu_ida_init(&tags->reserved_tags, reserved_tags,
-				1, 1);
-		if (ret)
-			goto err_reserved_tags;
-	}
-
-	return tags;
 
-err_reserved_tags:
-	percpu_ida_destroy(&tags->free_tags);
-err_free_tags:
-	kfree(tags);
-	return NULL;
+	return blk_mq_init_bitmap_tags(tags, node);
 }
 
 void blk_mq_free_tags(struct blk_mq_tags *tags)
 {
-	percpu_ida_destroy(&tags->free_tags);
-	percpu_ida_destroy(&tags->reserved_tags);
+	bt_free(&tags->bitmap_tags);
+	bt_free(&tags->breserved_tags);
 	kfree(tags);
 }
 
+void blk_mq_tag_init_last_tag(struct blk_mq_tags *tags, unsigned int *tag)
+{
+	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
+
+	*tag = prandom_u32() % depth;
+}
+
 ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page)
 {
 	char *orig_page = page;
-	unsigned int cpu;
+	unsigned int free, res;
 
 	if (!tags)
 		return 0;
 
-	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, batch_move=%u,"
-			" max_cache=%u\n", tags->nr_tags, tags->nr_reserved_tags,
-			tags->nr_batch_move, tags->nr_max_cache);
+	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, "
+			"bits_per_word=%u\n",
+			tags->nr_tags, tags->nr_reserved_tags,
+			tags->bitmap_tags.bits_per_word);
 
-	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n",
-			percpu_ida_free_tags(&tags->free_tags, nr_cpu_ids),
-			percpu_ida_free_tags(&tags->reserved_tags, nr_cpu_ids));
+	free = bt_unused_tags(&tags->bitmap_tags);
+	res = bt_unused_tags(&tags->breserved_tags);
 
-	for_each_possible_cpu(cpu) {
-		page += sprintf(page, "  cpu%02u: nr_free=%u\n", cpu,
-				percpu_ida_free_tags(&tags->free_tags, cpu));
-	}
+	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n", free, res);
 
 	return page - orig_page;
 }
diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
index 947ba2c6148e..7aa9f0665489 100644
--- a/block/blk-mq-tag.h
+++ b/block/blk-mq-tag.h
@@ -1,17 +1,61 @@
 #ifndef INT_BLK_MQ_TAG_H
 #define INT_BLK_MQ_TAG_H
 
-struct blk_mq_tags;
+enum {
+	BT_WAIT_QUEUES	= 8,
+	BT_WAIT_BATCH	= 8,
+};
+
+struct bt_wait_state {
+	atomic_t wait_cnt;
+	wait_queue_head_t wait;
+} ____cacheline_aligned_in_smp;
+
+#define TAG_TO_INDEX(bt, tag)	((tag) >> (bt)->bits_per_word)
+#define TAG_TO_BIT(bt, tag)	((tag) & ((1 << (bt)->bits_per_word) - 1))
+
+struct blk_mq_bitmap {
+	unsigned long word;
+	unsigned long depth;
+} ____cacheline_aligned_in_smp;
+
+struct blk_mq_bitmap_tags {
+	unsigned int depth;
+	unsigned int wake_cnt;
+	unsigned int bits_per_word;
+
+	unsigned int map_nr;
+	struct blk_mq_bitmap *map;
+
+	unsigned int wake_index;
+	struct bt_wait_state *bs;
+};
+
+/*
+ * Tag address space map.
+ */
+struct blk_mq_tags {
+	unsigned int nr_tags;
+	unsigned int nr_reserved_tags;
+
+	struct blk_mq_bitmap_tags bitmap_tags;
+	struct blk_mq_bitmap_tags breserved_tags;
+
+	struct request **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);
 extern void blk_mq_free_tags(struct blk_mq_tags *tags);
 
-extern unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp, bool reserved);
-extern void blk_mq_wait_for_tags(struct blk_mq_tags *tags);
-extern void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag);
+extern unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, struct blk_mq_hw_ctx *hctx, unsigned int *last_tag, gfp_t gfp, bool reserved);
+extern void blk_mq_wait_for_tags(struct blk_mq_tags *tags, struct blk_mq_hw_ctx *hctx, bool reserved);
+extern void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag, unsigned int *last_tag);
 extern void blk_mq_tag_busy_iter(struct blk_mq_tags *tags, void (*fn)(void *data, unsigned long *), void *data);
 extern bool blk_mq_has_free_tags(struct blk_mq_tags *tags);
 extern ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page);
+extern void blk_mq_tag_init_last_tag(struct blk_mq_tags *tags, unsigned int *last_tag);
 
 enum {
 	BLK_MQ_TAG_CACHE_MIN	= 1,
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 1d2a9bdbee57..526feee31bff 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -74,16 +74,16 @@ static void blk_mq_hctx_mark_pending(struct blk_mq_hw_ctx *hctx,
 }
 
 static struct request *__blk_mq_alloc_request(struct blk_mq_hw_ctx *hctx,
+					      struct blk_mq_ctx *ctx,
 					      gfp_t gfp, bool reserved)
 {
 	struct request *rq;
 	unsigned int tag;
 
-	tag = blk_mq_get_tag(hctx->tags, gfp, reserved);
+	tag = blk_mq_get_tag(hctx->tags, hctx, &ctx->last_tag, gfp, reserved);
 	if (tag != BLK_MQ_TAG_FAIL) {
-		rq = hctx->rqs[tag];
+		rq = hctx->tags->rqs[tag];
 		rq->tag = tag;
-
 		return rq;
 	}
 
@@ -186,10 +186,54 @@ static void blk_mq_rq_ctx_init(struct request_queue *q, struct blk_mq_ctx *ctx,
 	if (blk_queue_io_stat(q))
 		rw_flags |= REQ_IO_STAT;
 
+	INIT_LIST_HEAD(&rq->queuelist);
+	/* csd/requeue_work/fifo_time is initialized before use */
+	rq->q = q;
 	rq->mq_ctx = ctx;
 	rq->cmd_flags = rw_flags;
+	rq->cmd_type = 0;
+	/* do not touch atomic flags, it needs atomic ops against the timer */
+	rq->cpu = -1;
+	rq->__data_len = 0;
+	rq->__sector = (sector_t) -1;
+	rq->bio = NULL;
+	rq->biotail = NULL;
+	INIT_HLIST_NODE(&rq->hash);
+	RB_CLEAR_NODE(&rq->rb_node);
+	memset(&rq->flush, 0, max(sizeof(rq->flush), sizeof(rq->elv)));
+	rq->rq_disk = NULL;
+	rq->part = NULL;
 	rq->start_time = jiffies;
+#ifdef CONFIG_BLK_CGROUP
+	rq->rl = NULL;
 	set_start_time_ns(rq);
+	rq->io_start_time_ns = 0;
+#endif
+	rq->nr_phys_segments = 0;
+#if defined(CONFIG_BLK_DEV_INTEGRITY)
+	rq->nr_integrity_segments = 0;
+#endif
+	rq->ioprio = 0;
+	rq->special = NULL;
+	/* tag was already set */
+	rq->errors = 0;
+	memset(rq->__cmd, 0, sizeof(rq->__cmd));
+	rq->cmd = rq->__cmd;
+	rq->cmd_len = BLK_MAX_CDB;
+
+	rq->extra_len = 0;
+	rq->sense_len = 0;
+	rq->resid_len = 0;
+	rq->sense = NULL;
+
+	rq->deadline = 0;
+	INIT_LIST_HEAD(&rq->timeout_list);
+	rq->timeout = 0;
+	rq->retries = 0;
+	rq->end_io = NULL;
+	rq->end_io_data = NULL;
+	rq->next_rq = NULL;
+
 	ctx->rq_dispatched[rw_is_sync(rw_flags)]++;
 }
 
@@ -203,18 +247,22 @@ static struct request *blk_mq_alloc_request_pinned(struct request_queue *q,
 		struct blk_mq_ctx *ctx = blk_mq_get_ctx(q);
 		struct blk_mq_hw_ctx *hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
-		rq = __blk_mq_alloc_request(hctx, gfp & ~__GFP_WAIT, reserved);
+		rq = __blk_mq_alloc_request(hctx, ctx, gfp & ~__GFP_WAIT,
+						reserved);
 		if (rq) {
 			blk_mq_rq_ctx_init(q, ctx, rq, rw);
 			break;
 		}
 
-		blk_mq_put_ctx(ctx);
-		if (!(gfp & __GFP_WAIT))
+		if (gfp & __GFP_WAIT) {
+			__blk_mq_run_hw_queue(hctx);
+			blk_mq_put_ctx(ctx);
+		} else {
+			blk_mq_put_ctx(ctx);
 			break;
+		}
 
-		__blk_mq_run_hw_queue(hctx);
-		blk_mq_wait_for_tags(hctx->tags);
+		blk_mq_wait_for_tags(hctx->tags, hctx, reserved);
 	} while (1);
 
 	return rq;
@@ -232,6 +280,7 @@ struct request *blk_mq_alloc_request(struct request_queue *q, int rw, gfp_t gfp)
 		blk_mq_put_ctx(rq->mq_ctx);
 	return rq;
 }
+EXPORT_SYMBOL(blk_mq_alloc_request);
 
 struct request *blk_mq_alloc_reserved_request(struct request_queue *q, int rw,
 					      gfp_t gfp)
@@ -248,26 +297,14 @@ struct request *blk_mq_alloc_reserved_request(struct request_queue *q, int rw,
 }
 EXPORT_SYMBOL(blk_mq_alloc_reserved_request);
 
-/*
- * Re-init and set pdu, if we have it
- */
-void blk_mq_rq_init(struct blk_mq_hw_ctx *hctx, struct request *rq)
-{
-	blk_rq_init(hctx->queue, rq);
-
-	if (hctx->cmd_size)
-		rq->special = blk_mq_rq_to_pdu(rq);
-}
-
 static void __blk_mq_free_request(struct blk_mq_hw_ctx *hctx,
 				  struct blk_mq_ctx *ctx, struct request *rq)
 {
 	const int tag = rq->tag;
 	struct request_queue *q = rq->q;
 
-	blk_mq_rq_init(hctx, rq);
-	blk_mq_put_tag(hctx->tags, tag);
-
+	clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
+	blk_mq_put_tag(hctx->tags, tag, &ctx->last_tag);
 	blk_mq_queue_exit(q);
 }
 
@@ -283,20 +320,47 @@ void blk_mq_free_request(struct request *rq)
 	__blk_mq_free_request(hctx, ctx, rq);
 }
 
-bool blk_mq_end_io_partial(struct request *rq, int error, unsigned int nr_bytes)
+/*
+ * Clone all relevant state from a request that has been put on hold in
+ * the flush state machine into the preallocated flush request that hangs
+ * off the request queue.
+ *
+ * For a driver the flush request should be invisible, that's why we are
+ * impersonating the original request here.
+ */
+void blk_mq_clone_flush_request(struct request *flush_rq,
+		struct request *orig_rq)
 {
-	if (blk_update_request(rq, error, blk_rq_bytes(rq)))
-		return true;
+	struct blk_mq_hw_ctx *hctx =
+		orig_rq->q->mq_ops->map_queue(orig_rq->q, orig_rq->mq_ctx->cpu);
 
+	flush_rq->mq_ctx = orig_rq->mq_ctx;
+	flush_rq->tag = orig_rq->tag;
+	memcpy(blk_mq_rq_to_pdu(flush_rq), blk_mq_rq_to_pdu(orig_rq),
+		hctx->cmd_size);
+}
+
+inline void __blk_mq_end_io(struct request *rq, int error)
+{
 	blk_account_io_done(rq);
 
-	if (rq->end_io)
+	if (rq->end_io) {
 		rq->end_io(rq, error);
-	else
+	} else {
+		if (unlikely(blk_bidi_rq(rq)))
+			blk_mq_free_request(rq->next_rq);
 		blk_mq_free_request(rq);
-	return false;
+	}
 }
-EXPORT_SYMBOL(blk_mq_end_io_partial);
+EXPORT_SYMBOL(__blk_mq_end_io);
+
+void blk_mq_end_io(struct request *rq, int error)
+{
+	if (blk_update_request(rq, error, blk_rq_bytes(rq)))
+		BUG();
+	__blk_mq_end_io(rq, error);
+}
+EXPORT_SYMBOL(blk_mq_end_io);
 
 static void __blk_mq_complete_request_remote(void *data)
 {
@@ -308,15 +372,19 @@ static void __blk_mq_complete_request_remote(void *data)
 void __blk_mq_complete_request(struct request *rq)
 {
 	struct blk_mq_ctx *ctx = rq->mq_ctx;
+	bool shared = false;
 	int cpu;
 
-	if (!ctx->ipi_redirect) {
+	if (!test_bit(QUEUE_FLAG_SAME_COMP, &rq->q->queue_flags)) {
 		rq->q->softirq_done_fn(rq);
 		return;
 	}
 
 	cpu = get_cpu();
-	if (cpu != ctx->cpu && cpu_online(ctx->cpu)) {
+	if (!test_bit(QUEUE_FLAG_SAME_FORCE, &rq->q->queue_flags))
+		shared = cpus_share_cache(cpu, ctx->cpu);
+
+	if (cpu != ctx->cpu && !shared && cpu_online(ctx->cpu)) {
 		rq->csd.func = __blk_mq_complete_request_remote;
 		rq->csd.info = rq;
 		rq->csd.flags = 0;
@@ -350,13 +418,25 @@ static void blk_mq_start_request(struct request *rq, bool last)
 
 	trace_block_rq_issue(q, rq);
 
+	rq->resid_len = blk_rq_bytes(rq);
+	if (unlikely(blk_bidi_rq(rq)))
+		rq->next_rq->resid_len = blk_rq_bytes(rq->next_rq);
+
 	/*
 	 * Just mark start time and set the started bit. Due to memory
 	 * ordering, we know we'll see the correct deadline as long as
 	 * REQ_ATOMIC_STARTED is seen.
 	 */
 	rq->deadline = jiffies + q->rq_timeout;
+
+	/*
+	 * Mark us as started and clear complete. Complete might have been
+	 * set if requeue raced with timeout, which then marked it as
+	 * complete. So be sure to clear complete again when we start
+	 * the request, otherwise we'll ignore the completion event.
+	 */
 	set_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
+	clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
 
 	if (q->dma_drain_size && blk_rq_bytes(rq)) {
 		/*
@@ -378,7 +458,7 @@ static void blk_mq_start_request(struct request *rq, bool last)
 		rq->cmd_flags |= REQ_END;
 }
 
-static void blk_mq_requeue_request(struct request *rq)
+static void __blk_mq_requeue_request(struct request *rq)
 {
 	struct request_queue *q = rq->q;
 
@@ -391,6 +471,22 @@ static void blk_mq_requeue_request(struct request *rq)
 		rq->nr_phys_segments--;
 }
 
+void blk_mq_requeue_request(struct request *rq)
+{
+	__blk_mq_requeue_request(rq);
+	blk_clear_rq_complete(rq);
+
+	BUG_ON(blk_queued_rq(rq));
+	blk_mq_insert_request(rq, true, true, false);
+}
+EXPORT_SYMBOL(blk_mq_requeue_request);
+
+struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag)
+{
+	return tags->rqs[tag];
+}
+EXPORT_SYMBOL(blk_mq_tag_to_rq);
+
 struct blk_mq_timeout_data {
 	struct blk_mq_hw_ctx *hctx;
 	unsigned long *next;
@@ -412,12 +508,13 @@ static void blk_mq_timeout_check(void *__data, unsigned long *free_tags)
 	do {
 		struct request *rq;
 
-		tag = find_next_zero_bit(free_tags, hctx->queue_depth, tag);
-		if (tag >= hctx->queue_depth)
+		tag = find_next_zero_bit(free_tags, hctx->tags->nr_tags, tag);
+		if (tag >= hctx->tags->nr_tags)
 			break;
 
-		rq = hctx->rqs[tag++];
-
+		rq = blk_mq_tag_to_rq(hctx->tags, tag++);
+		if (rq->q != hctx->queue)
+			continue;
 		if (!test_bit(REQ_ATOM_STARTED, &rq->atomic_flags))
 			continue;
 
@@ -442,6 +539,28 @@ static void blk_mq_hw_ctx_check_timeout(struct blk_mq_hw_ctx *hctx,
 	blk_mq_tag_busy_iter(hctx->tags, blk_mq_timeout_check, &data);
 }
 
+static enum blk_eh_timer_return blk_mq_rq_timed_out(struct request *rq)
+{
+	struct request_queue *q = rq->q;
+
+	/*
+	 * We know that complete is set at this point. If STARTED isn't set
+	 * anymore, then the request isn't active and the "timeout" should
+	 * just be ignored. This can happen due to the bitflag ordering.
+	 * Timeout first checks if STARTED is set, and if it is, assumes
+	 * the request is active. But if we race with completion, then
+	 * we both flags will get cleared. So check here again, and ignore
+	 * a timeout event with a request that isn't active.
+	 */
+	if (!test_bit(REQ_ATOM_STARTED, &rq->atomic_flags))
+		return BLK_EH_NOT_HANDLED;
+
+	if (!q->mq_ops->timeout)
+		return BLK_EH_RESET_TIMER;
+
+	return q->mq_ops->timeout(rq);
+}
+
 static void blk_mq_rq_timer(unsigned long data)
 {
 	struct request_queue *q = (struct request_queue *) data;
@@ -495,11 +614,6 @@ static bool blk_mq_attempt_merge(struct request_queue *q,
 	return false;
 }
 
-void blk_mq_add_timer(struct request *rq)
-{
-	__blk_add_timer(rq, NULL);
-}
-
 /*
  * Run this hardware queue, pulling any software queues mapped to it in.
  * Note that this function currently has various problems around ordering
@@ -514,6 +628,8 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 	LIST_HEAD(rq_list);
 	int bit, queued;
 
+	WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
+
 	if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
 		return;
 
@@ -525,7 +641,6 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 	for_each_set_bit(bit, hctx->ctx_map, hctx->nr_ctx) {
 		clear_bit(bit, hctx->ctx_map);
 		ctx = hctx->ctxs[bit];
-		BUG_ON(bit != ctx->index_hw);
 
 		spin_lock(&ctx->lock);
 		list_splice_tail_init(&ctx->rq_list, &rq_list);
@@ -565,13 +680,8 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 			queued++;
 			continue;
 		case BLK_MQ_RQ_QUEUE_BUSY:
-			/*
-			 * FIXME: we should have a mechanism to stop the queue
-			 * like blk_stop_queue, otherwise we will waste cpu
-			 * time
-			 */
 			list_add(&rq->queuelist, &rq_list);
-			blk_mq_requeue_request(rq);
+			__blk_mq_requeue_request(rq);
 			break;
 		default:
 			pr_err("blk-mq: bad return on queue: %d\n", ret);
@@ -601,17 +711,44 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 	}
 }
 
+/*
+ * It'd be great if the workqueue API had a way to pass
+ * in a mask and had some smarts for more clever placement.
+ * For now we just round-robin here, switching for every
+ * BLK_MQ_CPU_WORK_BATCH queued items.
+ */
+static int blk_mq_hctx_next_cpu(struct blk_mq_hw_ctx *hctx)
+{
+	int cpu = hctx->next_cpu;
+
+	if (--hctx->next_cpu_batch <= 0) {
+		int next_cpu;
+
+		next_cpu = cpumask_next(hctx->next_cpu, hctx->cpumask);
+		if (next_cpu >= nr_cpu_ids)
+			next_cpu = cpumask_first(hctx->cpumask);
+
+		hctx->next_cpu = next_cpu;
+		hctx->next_cpu_batch = BLK_MQ_CPU_WORK_BATCH;
+	}
+
+	return cpu;
+}
+
 void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async)
 {
 	if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
 		return;
 
-	if (!async)
+	if (!async && cpumask_test_cpu(smp_processor_id(), hctx->cpumask))
 		__blk_mq_run_hw_queue(hctx);
+	else if (hctx->queue->nr_hw_queues == 1)
+		kblockd_schedule_delayed_work(&hctx->run_work, 0);
 	else {
-		struct request_queue *q = hctx->queue;
+		unsigned int cpu;
 
-		kblockd_schedule_delayed_work(q, &hctx->delayed_work, 0);
+		cpu = blk_mq_hctx_next_cpu(hctx);
+		kblockd_schedule_delayed_work_on(cpu, &hctx->run_work, 0);
 	}
 }
 
@@ -626,14 +763,17 @@ void blk_mq_run_queues(struct request_queue *q, bool async)
 		    test_bit(BLK_MQ_S_STOPPED, &hctx->state))
 			continue;
 
+		preempt_disable();
 		blk_mq_run_hw_queue(hctx, async);
+		preempt_enable();
 	}
 }
 EXPORT_SYMBOL(blk_mq_run_queues);
 
 void blk_mq_stop_hw_queue(struct blk_mq_hw_ctx *hctx)
 {
-	cancel_delayed_work(&hctx->delayed_work);
+	cancel_delayed_work(&hctx->run_work);
+	cancel_delayed_work(&hctx->delay_work);
 	set_bit(BLK_MQ_S_STOPPED, &hctx->state);
 }
 EXPORT_SYMBOL(blk_mq_stop_hw_queue);
@@ -651,11 +791,25 @@ EXPORT_SYMBOL(blk_mq_stop_hw_queues);
 void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx)
 {
 	clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
+
+	preempt_disable();
 	__blk_mq_run_hw_queue(hctx);
+	preempt_enable();
 }
 EXPORT_SYMBOL(blk_mq_start_hw_queue);
 
-void blk_mq_start_stopped_hw_queues(struct request_queue *q)
+void blk_mq_start_hw_queues(struct request_queue *q)
+{
+	struct blk_mq_hw_ctx *hctx;
+	int i;
+
+	queue_for_each_hw_ctx(q, hctx, i)
+		blk_mq_start_hw_queue(hctx);
+}
+EXPORT_SYMBOL(blk_mq_start_hw_queues);
+
+
+void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async)
 {
 	struct blk_mq_hw_ctx *hctx;
 	int i;
@@ -665,19 +819,47 @@ void blk_mq_start_stopped_hw_queues(struct request_queue *q)
 			continue;
 
 		clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
-		blk_mq_run_hw_queue(hctx, true);
+		preempt_disable();
+		blk_mq_run_hw_queue(hctx, async);
+		preempt_enable();
 	}
 }
 EXPORT_SYMBOL(blk_mq_start_stopped_hw_queues);
 
-static void blk_mq_work_fn(struct work_struct *work)
+static void blk_mq_run_work_fn(struct work_struct *work)
 {
 	struct blk_mq_hw_ctx *hctx;
 
-	hctx = container_of(work, struct blk_mq_hw_ctx, delayed_work.work);
+	hctx = container_of(work, struct blk_mq_hw_ctx, run_work.work);
+
 	__blk_mq_run_hw_queue(hctx);
 }
 
+static void blk_mq_delay_work_fn(struct work_struct *work)
+{
+	struct blk_mq_hw_ctx *hctx;
+
+	hctx = container_of(work, struct blk_mq_hw_ctx, delay_work.work);
+
+	if (test_and_clear_bit(BLK_MQ_S_STOPPED, &hctx->state))
+		__blk_mq_run_hw_queue(hctx);
+}
+
+void blk_mq_delay_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs)
+{
+	unsigned long tmo = msecs_to_jiffies(msecs);
+
+	if (hctx->queue->nr_hw_queues == 1)
+		kblockd_schedule_delayed_work(&hctx->delay_work, tmo);
+	else {
+		unsigned int cpu;
+
+		cpu = blk_mq_hctx_next_cpu(hctx);
+		kblockd_schedule_delayed_work_on(cpu, &hctx->delay_work, tmo);
+	}
+}
+EXPORT_SYMBOL(blk_mq_delay_queue);
+
 static void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx,
 				    struct request *rq, bool at_head)
 {
@@ -689,12 +871,13 @@ static void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx,
 		list_add(&rq->queuelist, &ctx->rq_list);
 	else
 		list_add_tail(&rq->queuelist, &ctx->rq_list);
+
 	blk_mq_hctx_mark_pending(hctx, ctx);
 
 	/*
 	 * We do this early, to ensure we are on the right CPU.
 	 */
-	blk_mq_add_timer(rq);
+	blk_add_timer(rq);
 }
 
 void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
@@ -719,10 +902,10 @@ void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
 		spin_unlock(&ctx->lock);
 	}
 
-	blk_mq_put_ctx(current_ctx);
-
 	if (run_queue)
 		blk_mq_run_hw_queue(hctx, async);
+
+	blk_mq_put_ctx(current_ctx);
 }
 
 static void blk_mq_insert_requests(struct request_queue *q,
@@ -758,9 +941,8 @@ static void blk_mq_insert_requests(struct request_queue *q,
 	}
 	spin_unlock(&ctx->lock);
 
-	blk_mq_put_ctx(current_ctx);
-
 	blk_mq_run_hw_queue(hctx, from_schedule);
+	blk_mq_put_ctx(current_ctx);
 }
 
 static int plug_ctx_cmp(void *priv, struct list_head *a, struct list_head *b)
@@ -863,7 +1045,7 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 	if (is_sync)
 		rw |= REQ_SYNC;
 	trace_block_getrq(q, bio, rw);
-	rq = __blk_mq_alloc_request(hctx, GFP_ATOMIC, false);
+	rq = __blk_mq_alloc_request(hctx, ctx, GFP_ATOMIC, false);
 	if (likely(rq))
 		blk_mq_rq_ctx_init(q, ctx, rq, rw);
 	else {
@@ -879,7 +1061,6 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 
 	if (unlikely(is_flush_fua)) {
 		blk_mq_bio_to_request(rq, bio);
-		blk_mq_put_ctx(ctx);
 		blk_insert_flush(rq);
 		goto run_queue;
 	}
@@ -906,18 +1087,23 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 		}
 	}
 
-	spin_lock(&ctx->lock);
-
-	if ((hctx->flags & BLK_MQ_F_SHOULD_MERGE) &&
-	    blk_mq_attempt_merge(q, ctx, bio))
-		__blk_mq_free_request(hctx, ctx, rq);
-	else {
+	if (!(hctx->flags & BLK_MQ_F_SHOULD_MERGE)) {
 		blk_mq_bio_to_request(rq, bio);
+		spin_lock(&ctx->lock);
+insert_rq:
 		__blk_mq_insert_request(hctx, rq, false);
+		spin_unlock(&ctx->lock);
+	} else {
+		spin_lock(&ctx->lock);
+		if (!blk_mq_attempt_merge(q, ctx, bio)) {
+			blk_mq_bio_to_request(rq, bio);
+			goto insert_rq;
+		}
+
+		spin_unlock(&ctx->lock);
+		__blk_mq_free_request(hctx, ctx, rq);
 	}
 
-	spin_unlock(&ctx->lock);
-	blk_mq_put_ctx(ctx);
 
 	/*
 	 * For a SYNC request, send it to the hardware immediately. For an
@@ -926,6 +1112,7 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 	 */
 run_queue:
 	blk_mq_run_hw_queue(hctx, !is_sync || is_flush_fua);
+	blk_mq_put_ctx(ctx);
 }
 
 /*
@@ -937,11 +1124,11 @@ struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *q, const int cpu)
 }
 EXPORT_SYMBOL(blk_mq_map_queue);
 
-struct blk_mq_hw_ctx *blk_mq_alloc_single_hw_queue(struct blk_mq_reg *reg,
+struct blk_mq_hw_ctx *blk_mq_alloc_single_hw_queue(struct blk_mq_tag_set *set,
 						   unsigned int hctx_index)
 {
-	return kmalloc_node(sizeof(struct blk_mq_hw_ctx),
-				GFP_KERNEL | __GFP_ZERO, reg->numa_node);
+	return kzalloc_node(sizeof(struct blk_mq_hw_ctx), GFP_KERNEL,
+				set->numa_node);
 }
 EXPORT_SYMBOL(blk_mq_alloc_single_hw_queue);
 
@@ -993,124 +1180,73 @@ static void blk_mq_hctx_notify(void *data, unsigned long action,
 	blk_mq_hctx_mark_pending(hctx, ctx);
 
 	spin_unlock(&ctx->lock);
-	blk_mq_put_ctx(ctx);
 
 	blk_mq_run_hw_queue(hctx, true);
+	blk_mq_put_ctx(ctx);
 }
 
-static int blk_mq_init_hw_commands(struct blk_mq_hw_ctx *hctx,
-				   int (*init)(void *, struct blk_mq_hw_ctx *,
-					struct request *, unsigned int),
-				   void *data)
-{
-	unsigned int i;
-	int ret = 0;
-
-	for (i = 0; i < hctx->queue_depth; i++) {
-		struct request *rq = hctx->rqs[i];
-
-		ret = init(data, hctx, rq, i);
-		if (ret)
-			break;
-	}
-
-	return ret;
-}
-
-int blk_mq_init_commands(struct request_queue *q,
-			 int (*init)(void *, struct blk_mq_hw_ctx *,
-					struct request *, unsigned int),
-			 void *data)
-{
-	struct blk_mq_hw_ctx *hctx;
-	unsigned int i;
-	int ret = 0;
-
-	queue_for_each_hw_ctx(q, hctx, i) {
-		ret = blk_mq_init_hw_commands(hctx, init, data);
-		if (ret)
-			break;
-	}
-
-	return ret;
-}
-EXPORT_SYMBOL(blk_mq_init_commands);
-
-static void blk_mq_free_hw_commands(struct blk_mq_hw_ctx *hctx,
-				    void (*free)(void *, struct blk_mq_hw_ctx *,
-					struct request *, unsigned int),
-				    void *data)
+static void blk_mq_free_rq_map(struct blk_mq_tag_set *set,
+		struct blk_mq_tags *tags, unsigned int hctx_idx)
 {
-	unsigned int i;
+	struct page *page;
 
-	for (i = 0; i < hctx->queue_depth; i++) {
-		struct request *rq = hctx->rqs[i];
+	if (tags->rqs && set->ops->exit_request) {
+		int i;
 
-		free(data, hctx, rq, i);
+		for (i = 0; i < tags->nr_tags; i++) {
+			if (!tags->rqs[i])
+				continue;
+			set->ops->exit_request(set->driver_data, tags->rqs[i],
+						hctx_idx, i);
+		}
 	}
-}
-
-void blk_mq_free_commands(struct request_queue *q,
-			  void (*free)(void *, struct blk_mq_hw_ctx *,
-					struct request *, unsigned int),
-			  void *data)
-{
-	struct blk_mq_hw_ctx *hctx;
-	unsigned int i;
-
-	queue_for_each_hw_ctx(q, hctx, i)
-		blk_mq_free_hw_commands(hctx, free, data);
-}
-EXPORT_SYMBOL(blk_mq_free_commands);
 
-static void blk_mq_free_rq_map(struct blk_mq_hw_ctx *hctx)
-{
-	struct page *page;
-
-	while (!list_empty(&hctx->page_list)) {
-		page = list_first_entry(&hctx->page_list, struct page, lru);
+	while (!list_empty(&tags->page_list)) {
+		page = list_first_entry(&tags->page_list, struct page, lru);
 		list_del_init(&page->lru);
 		__free_pages(page, page->private);
 	}
 
-	kfree(hctx->rqs);
+	kfree(tags->rqs);
 
-	if (hctx->tags)
-		blk_mq_free_tags(hctx->tags);
+	blk_mq_free_tags(tags);
 }
 
 static size_t order_to_size(unsigned int order)
 {
-	size_t ret = PAGE_SIZE;
-
-	while (order--)
-		ret *= 2;
-
-	return ret;
+	return (size_t)PAGE_SIZE << order;
 }
 
-static int blk_mq_init_rq_map(struct blk_mq_hw_ctx *hctx,
-			      unsigned int reserved_tags, int node)
+static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
+		unsigned int hctx_idx)
 {
+	struct blk_mq_tags *tags;
 	unsigned int i, j, entries_per_page, max_order = 4;
 	size_t rq_size, left;
 
-	INIT_LIST_HEAD(&hctx->page_list);
+	tags = blk_mq_init_tags(set->queue_depth, set->reserved_tags,
+				set->numa_node);
+	if (!tags)
+		return NULL;
 
-	hctx->rqs = kmalloc_node(hctx->queue_depth * sizeof(struct request *),
-					GFP_KERNEL, node);
-	if (!hctx->rqs)
-		return -ENOMEM;
+	INIT_LIST_HEAD(&tags->page_list);
+
+	tags->rqs = kmalloc_node(set->queue_depth * sizeof(struct request *),
+					GFP_KERNEL, set->numa_node);
+	if (!tags->rqs) {
+		blk_mq_free_tags(tags);
+		return NULL;
+	}
 
 	/*
 	 * rq_size is the size of the request plus driver payload, rounded
 	 * to the cacheline size
 	 */
-	rq_size = round_up(sizeof(struct request) + hctx->cmd_size,
+	rq_size = round_up(sizeof(struct request) + set->cmd_size,
 				cache_line_size());
-	left = rq_size * hctx->queue_depth;
+	left = rq_size * set->queue_depth;
 
-	for (i = 0; i < hctx->queue_depth;) {
+	for (i = 0; i < set->queue_depth; ) {
 		int this_order = max_order;
 		struct page *page;
 		int to_do;
@@ -1120,7 +1256,8 @@ static int blk_mq_init_rq_map(struct blk_mq_hw_ctx *hctx,
 			this_order--;
 
 		do {
-			page = alloc_pages_node(node, GFP_KERNEL, this_order);
+			page = alloc_pages_node(set->numa_node, GFP_KERNEL,
+						this_order);
 			if (page)
 				break;
 			if (!this_order--)
@@ -1130,43 +1267,39 @@ static int blk_mq_init_rq_map(struct blk_mq_hw_ctx *hctx,
 		} while (1);
 
 		if (!page)
-			break;
+			goto fail;
 
 		page->private = this_order;
-		list_add_tail(&page->lru, &hctx->page_list);
+		list_add_tail(&page->lru, &tags->page_list);
 
 		p = page_address(page);
 		entries_per_page = order_to_size(this_order) / rq_size;
-		to_do = min(entries_per_page, hctx->queue_depth - i);
+		to_do = min(entries_per_page, set->queue_depth - i);
 		left -= to_do * rq_size;
 		for (j = 0; j < to_do; j++) {
-			hctx->rqs[i] = p;
-			blk_mq_rq_init(hctx, hctx->rqs[i]);
+			tags->rqs[i] = p;
+			if (set->ops->init_request) {
+				if (set->ops->init_request(set->driver_data,
+						tags->rqs[i], hctx_idx, i,
+						set->numa_node))
+					goto fail;
+			}
+
 			p += rq_size;
 			i++;
 		}
 	}
 
-	if (i < (reserved_tags + BLK_MQ_TAG_MIN))
-		goto err_rq_map;
-	else if (i != hctx->queue_depth) {
-		hctx->queue_depth = i;
-		pr_warn("%s: queue depth set to %u because of low memory\n",
-					__func__, i);
-	}
+	return tags;
 
-	hctx->tags = blk_mq_init_tags(hctx->queue_depth, reserved_tags, node);
-	if (!hctx->tags) {
-err_rq_map:
-		blk_mq_free_rq_map(hctx);
-		return -ENOMEM;
-	}
-
-	return 0;
+fail:
+	pr_warn("%s: failed to allocate requests\n", __func__);
+	blk_mq_free_rq_map(set, tags, hctx_idx);
+	return NULL;
 }
 
 static int blk_mq_init_hw_queues(struct request_queue *q,
-				 struct blk_mq_reg *reg, void *driver_data)
+		struct blk_mq_tag_set *set)
 {
 	struct blk_mq_hw_ctx *hctx;
 	unsigned int i, j;
@@ -1180,23 +1313,22 @@ static int blk_mq_init_hw_queues(struct request_queue *q,
 
 		node = hctx->numa_node;
 		if (node == NUMA_NO_NODE)
-			node = hctx->numa_node = reg->numa_node;
+			node = hctx->numa_node = set->numa_node;
 
-		INIT_DELAYED_WORK(&hctx->delayed_work, blk_mq_work_fn);
+		INIT_DELAYED_WORK(&hctx->run_work, blk_mq_run_work_fn);
+		INIT_DELAYED_WORK(&hctx->delay_work, blk_mq_delay_work_fn);
 		spin_lock_init(&hctx->lock);
 		INIT_LIST_HEAD(&hctx->dispatch);
 		hctx->queue = q;
 		hctx->queue_num = i;
-		hctx->flags = reg->flags;
-		hctx->queue_depth = reg->queue_depth;
-		hctx->cmd_size = reg->cmd_size;
+		hctx->flags = set->flags;
+		hctx->cmd_size = set->cmd_size;
 
 		blk_mq_init_cpu_notifier(&hctx->cpu_notifier,
 						blk_mq_hctx_notify, hctx);
 		blk_mq_register_cpu_notifier(&hctx->cpu_notifier);
 
-		if (blk_mq_init_rq_map(hctx, reg->reserved_tags, node))
-			break;
+		hctx->tags = set->tags[i];
 
 		/*
 		 * Allocate space for all possible cpus to avoid allocation in
@@ -1216,8 +1348,8 @@ static int blk_mq_init_hw_queues(struct request_queue *q,
 		hctx->nr_ctx_map = num_maps;
 		hctx->nr_ctx = 0;
 
-		if (reg->ops->init_hctx &&
-		    reg->ops->init_hctx(hctx, driver_data, i))
+		if (set->ops->init_hctx &&
+		    set->ops->init_hctx(hctx, set->driver_data, i))
 			break;
 	}
 
@@ -1231,12 +1363,12 @@ static int blk_mq_init_hw_queues(struct request_queue *q,
 		if (i == j)
 			break;
 
-		if (reg->ops->exit_hctx)
-			reg->ops->exit_hctx(hctx, j);
+		if (set->ops->exit_hctx)
+			set->ops->exit_hctx(hctx, j);
 
 		blk_mq_unregister_cpu_notifier(&hctx->cpu_notifier);
-		blk_mq_free_rq_map(hctx);
 		kfree(hctx->ctxs);
+		kfree(hctx->ctx_map);
 	}
 
 	return 1;
@@ -1258,12 +1390,13 @@ static void blk_mq_init_cpu_queues(struct request_queue *q,
 		__ctx->queue = q;
 
 		/* If the cpu isn't online, the cpu is mapped to first hctx */
-		hctx = q->mq_ops->map_queue(q, i);
-		hctx->nr_ctx++;
-
 		if (!cpu_online(i))
 			continue;
 
+		hctx = q->mq_ops->map_queue(q, i);
+		cpumask_set_cpu(i, hctx->cpumask);
+		hctx->nr_ctx++;
+
 		/*
 		 * Set local node, IFF we have more than one hw queue. If
 		 * not, we remain on the home node of the device
@@ -1280,6 +1413,7 @@ static void blk_mq_map_swqueue(struct request_queue *q)
 	struct blk_mq_ctx *ctx;
 
 	queue_for_each_hw_ctx(q, hctx, i) {
+		cpumask_clear(hctx->cpumask);
 		hctx->nr_ctx = 0;
 	}
 
@@ -1288,59 +1422,55 @@ static void blk_mq_map_swqueue(struct request_queue *q)
 	 */
 	queue_for_each_ctx(q, ctx, i) {
 		/* If the cpu isn't online, the cpu is mapped to first hctx */
+		if (!cpu_online(i))
+			continue;
+
 		hctx = q->mq_ops->map_queue(q, i);
+		cpumask_set_cpu(i, hctx->cpumask);
 		ctx->index_hw = hctx->nr_ctx;
 		hctx->ctxs[hctx->nr_ctx++] = ctx;
 	}
+
+	queue_for_each_hw_ctx(q, hctx, i) {
+		hctx->next_cpu = cpumask_first(hctx->cpumask);
+		hctx->next_cpu_batch = BLK_MQ_CPU_WORK_BATCH;
+	}
 }
 
-struct request_queue *blk_mq_init_queue(struct blk_mq_reg *reg,
-					void *driver_data)
+struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 {
 	struct blk_mq_hw_ctx **hctxs;
 	struct blk_mq_ctx *ctx;
 	struct request_queue *q;
 	int i;
 
-	if (!reg->nr_hw_queues ||
-	    !reg->ops->queue_rq || !reg->ops->map_queue ||
-	    !reg->ops->alloc_hctx || !reg->ops->free_hctx)
-		return ERR_PTR(-EINVAL);
-
-	if (!reg->queue_depth)
-		reg->queue_depth = BLK_MQ_MAX_DEPTH;
-	else if (reg->queue_depth > BLK_MQ_MAX_DEPTH) {
-		pr_err("blk-mq: queuedepth too large (%u)\n", reg->queue_depth);
-		reg->queue_depth = BLK_MQ_MAX_DEPTH;
-	}
-
-	if (reg->queue_depth < (reg->reserved_tags + BLK_MQ_TAG_MIN))
-		return ERR_PTR(-EINVAL);
-
 	ctx = alloc_percpu(struct blk_mq_ctx);
 	if (!ctx)
 		return ERR_PTR(-ENOMEM);
 
-	hctxs = kmalloc_node(reg->nr_hw_queues * sizeof(*hctxs), GFP_KERNEL,
-			reg->numa_node);
+	hctxs = kmalloc_node(set->nr_hw_queues * sizeof(*hctxs), GFP_KERNEL,
+			set->numa_node);
 
 	if (!hctxs)
 		goto err_percpu;
 
-	for (i = 0; i < reg->nr_hw_queues; i++) {
-		hctxs[i] = reg->ops->alloc_hctx(reg, i);
+	for (i = 0; i < set->nr_hw_queues; i++) {
+		hctxs[i] = set->ops->alloc_hctx(set, i);
 		if (!hctxs[i])
 			goto err_hctxs;
 
+		if (!zalloc_cpumask_var(&hctxs[i]->cpumask, GFP_KERNEL))
+			goto err_hctxs;
+
 		hctxs[i]->numa_node = NUMA_NO_NODE;
 		hctxs[i]->queue_num = i;
 	}
 
-	q = blk_alloc_queue_node(GFP_KERNEL, reg->numa_node);
+	q = blk_alloc_queue_node(GFP_KERNEL, set->numa_node);
 	if (!q)
 		goto err_hctxs;
 
-	q->mq_map = blk_mq_make_queue_map(reg);
+	q->mq_map = blk_mq_make_queue_map(set);
 	if (!q->mq_map)
 		goto err_map;
 
@@ -1348,33 +1478,34 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_reg *reg,
 	blk_queue_rq_timeout(q, 30000);
 
 	q->nr_queues = nr_cpu_ids;
-	q->nr_hw_queues = reg->nr_hw_queues;
+	q->nr_hw_queues = set->nr_hw_queues;
 
 	q->queue_ctx = ctx;
 	q->queue_hw_ctx = hctxs;
 
-	q->mq_ops = reg->ops;
+	q->mq_ops = set->ops;
 	q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT;
 
 	q->sg_reserved_size = INT_MAX;
 
 	blk_queue_make_request(q, blk_mq_make_request);
-	blk_queue_rq_timed_out(q, reg->ops->timeout);
-	if (reg->timeout)
-		blk_queue_rq_timeout(q, reg->timeout);
+	blk_queue_rq_timed_out(q, blk_mq_rq_timed_out);
+	if (set->timeout)
+		blk_queue_rq_timeout(q, set->timeout);
 
-	if (reg->ops->complete)
-		blk_queue_softirq_done(q, reg->ops->complete);
+	if (set->ops->complete)
+		blk_queue_softirq_done(q, set->ops->complete);
 
 	blk_mq_init_flush(q);
-	blk_mq_init_cpu_queues(q, reg->nr_hw_queues);
+	blk_mq_init_cpu_queues(q, set->nr_hw_queues);
 
-	q->flush_rq = kzalloc(round_up(sizeof(struct request) + reg->cmd_size,
-				cache_line_size()), GFP_KERNEL);
+	q->flush_rq = kzalloc(round_up(sizeof(struct request) +
+				set->cmd_size, cache_line_size()),
+				GFP_KERNEL);
 	if (!q->flush_rq)
 		goto err_hw;
 
-	if (blk_mq_init_hw_queues(q, reg, driver_data))
+	if (blk_mq_init_hw_queues(q, set))
 		goto err_flush_rq;
 
 	blk_mq_map_swqueue(q);
@@ -1392,10 +1523,11 @@ err_hw:
 err_map:
 	blk_cleanup_queue(q);
 err_hctxs:
-	for (i = 0; i < reg->nr_hw_queues; i++) {
+	for (i = 0; i < set->nr_hw_queues; i++) {
 		if (!hctxs[i])
 			break;
-		reg->ops->free_hctx(hctxs[i], i);
+		free_cpumask_var(hctxs[i]->cpumask);
+		set->ops->free_hctx(hctxs[i], i);
 	}
 	kfree(hctxs);
 err_percpu:
@@ -1412,10 +1544,10 @@ void blk_mq_free_queue(struct request_queue *q)
 	queue_for_each_hw_ctx(q, hctx, i) {
 		kfree(hctx->ctx_map);
 		kfree(hctx->ctxs);
-		blk_mq_free_rq_map(hctx);
 		blk_mq_unregister_cpu_notifier(&hctx->cpu_notifier);
 		if (q->mq_ops->exit_hctx)
 			q->mq_ops->exit_hctx(hctx, i);
+		free_cpumask_var(hctx->cpumask);
 		q->mq_ops->free_hctx(hctx, i);
 	}
 
@@ -1456,10 +1588,10 @@ static int blk_mq_queue_reinit_notify(struct notifier_block *nb,
 	struct request_queue *q;
 
 	/*
-	 * Before new mapping is established, hotadded cpu might already start
-	 * handling requests. This doesn't break anything as we map offline
-	 * CPUs to first hardware queue. We will re-init queue below to get
-	 * optimal settings.
+	 * Before new mappings are established, hotadded cpu might already
+	 * start handling requests. This doesn't break anything as we map
+	 * offline CPUs to first hardware queue. We will re-init the queue
+	 * below to get optimal settings.
 	 */
 	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN &&
 	    action != CPU_ONLINE && action != CPU_ONLINE_FROZEN)
@@ -1472,6 +1604,55 @@ static int blk_mq_queue_reinit_notify(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
+int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
+{
+	int i;
+
+	if (!set->nr_hw_queues)
+		return -EINVAL;
+	if (!set->queue_depth || set->queue_depth > BLK_MQ_MAX_DEPTH)
+		return -EINVAL;
+	if (set->queue_depth < set->reserved_tags + BLK_MQ_TAG_MIN)
+		return -EINVAL;
+
+	if (!set->nr_hw_queues ||
+	    !set->ops->queue_rq || !set->ops->map_queue ||
+	    !set->ops->alloc_hctx || !set->ops->free_hctx)
+		return -EINVAL;
+
+
+	set->tags = kmalloc_node(set->nr_hw_queues *
+				 sizeof(struct blk_mq_tags *),
+				 GFP_KERNEL, set->numa_node);
+	if (!set->tags)
+		goto out;
+
+	for (i = 0; i < set->nr_hw_queues; i++) {
+		set->tags[i] = blk_mq_init_rq_map(set, i);
+		if (!set->tags[i])
+			goto out_unwind;
+	}
+
+	return 0;
+
+out_unwind:
+	while (--i >= 0)
+		blk_mq_free_rq_map(set, set->tags[i], i);
+out:
+	return -ENOMEM;
+}
+EXPORT_SYMBOL(blk_mq_alloc_tag_set);
+
+void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
+{
+	int i;
+
+	for (i = 0; i < set->nr_hw_queues; i++)
+		blk_mq_free_rq_map(set, set->tags[i], i);
+	kfree(set->tags);
+}
+EXPORT_SYMBOL(blk_mq_free_tag_set);
+
 void blk_mq_disable_hotplug(void)
 {
 	mutex_lock(&all_q_mutex);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index ebbe6bac9d61..97cfab9c092f 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -1,6 +1,8 @@
 #ifndef INT_BLK_MQ_H
 #define INT_BLK_MQ_H
 
+struct blk_mq_tag_set;
+
 struct blk_mq_ctx {
 	struct {
 		spinlock_t		lock;
@@ -9,7 +11,8 @@ struct blk_mq_ctx {
 
 	unsigned int		cpu;
 	unsigned int		index_hw;
-	unsigned int		ipi_redirect;
+
+	unsigned int		last_tag ____cacheline_aligned_in_smp;
 
 	/* incremented at dispatch time */
 	unsigned long		rq_dispatched[2];
@@ -20,14 +23,15 @@ struct blk_mq_ctx {
 
 	struct request_queue	*queue;
 	struct kobject		kobj;
-};
+} ____cacheline_aligned_in_smp;
 
 void __blk_mq_complete_request(struct request *rq);
 void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async);
 void blk_mq_init_flush(struct request_queue *q);
 void blk_mq_drain_queue(struct request_queue *q);
 void blk_mq_free_queue(struct request_queue *q);
-void blk_mq_rq_init(struct blk_mq_hw_ctx *hctx, struct request *rq);
+void blk_mq_clone_flush_request(struct request *flush_rq,
+		struct request *orig_rq);
 
 /*
  * CPU hotplug helpers
@@ -45,10 +49,7 @@ void blk_mq_disable_hotplug(void);
 /*
  * CPU -> queue mappings
  */
-struct blk_mq_reg;
-extern unsigned int *blk_mq_make_queue_map(struct blk_mq_reg *reg);
+extern unsigned int *blk_mq_make_queue_map(struct blk_mq_tag_set *set);
 extern int blk_mq_update_queue_map(unsigned int *map, unsigned int nr_queues);
 
-void blk_mq_add_timer(struct request *rq);
-
 #endif
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 033745cd7fba..9353b4683359 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -744,7 +744,7 @@ static inline void throtl_extend_slice(struct throtl_grp *tg, bool rw,
 static bool throtl_slice_used(struct throtl_grp *tg, bool rw)
 {
 	if (time_in_range(jiffies, tg->slice_start[rw], tg->slice_end[rw]))
-		return 0;
+		return false;
 
 	return 1;
 }
@@ -842,7 +842,7 @@ static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
 	if (tg->io_disp[rw] + 1 <= io_allowed) {
 		if (wait)
 			*wait = 0;
-		return 1;
+		return true;
 	}
 
 	/* Calc approx time to dispatch */
@@ -880,7 +880,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
 	if (tg->bytes_disp[rw] + bio->bi_iter.bi_size <= bytes_allowed) {
 		if (wait)
 			*wait = 0;
-		return 1;
+		return true;
 	}
 
 	/* Calc approx time to dispatch */
@@ -923,7 +923,7 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
 	if (tg->bps[rw] == -1 && tg->iops[rw] == -1) {
 		if (wait)
 			*wait = 0;
-		return 1;
+		return true;
 	}
 
 	/*
@@ -1258,7 +1258,7 @@ out_unlock:
  * of throtl_data->service_queue.  Those bio's are ready and issued by this
  * function.
  */
-void blk_throtl_dispatch_work_fn(struct work_struct *work)
+static void blk_throtl_dispatch_work_fn(struct work_struct *work)
 {
 	struct throtl_data *td = container_of(work, struct throtl_data,
 					      dispatch_work);
diff --git a/block/blk-timeout.c b/block/blk-timeout.c
index d96f7061c6fd..448745683d28 100644
--- a/block/blk-timeout.c
+++ b/block/blk-timeout.c
@@ -96,11 +96,7 @@ static void blk_rq_timed_out(struct request *req)
 			__blk_complete_request(req);
 		break;
 	case BLK_EH_RESET_TIMER:
-		if (q->mq_ops)
-			blk_mq_add_timer(req);
-		else
-			blk_add_timer(req);
-
+		blk_add_timer(req);
 		blk_clear_rq_complete(req);
 		break;
 	case BLK_EH_NOT_HANDLED:
@@ -170,7 +166,15 @@ void blk_abort_request(struct request *req)
 }
 EXPORT_SYMBOL_GPL(blk_abort_request);
 
-void __blk_add_timer(struct request *req, struct list_head *timeout_list)
+/**
+ * blk_add_timer - Start timeout timer for a single request
+ * @req:	request that is about to start running.
+ *
+ * Notes:
+ *    Each request has its own timer, and as it is added to the queue, we
+ *    set up the timer. When the request completes, we cancel the timer.
+ */
+void blk_add_timer(struct request *req)
 {
 	struct request_queue *q = req->q;
 	unsigned long expiry;
@@ -188,8 +192,8 @@ void __blk_add_timer(struct request *req, struct list_head *timeout_list)
 		req->timeout = q->rq_timeout;
 
 	req->deadline = jiffies + req->timeout;
-	if (timeout_list)
-		list_add_tail(&req->timeout_list, timeout_list);
+	if (!q->mq_ops)
+		list_add_tail(&req->timeout_list, &req->q->timeout_list);
 
 	/*
 	 * If the timer isn't already pending or this timeout is earlier
@@ -199,21 +203,18 @@ void __blk_add_timer(struct request *req, struct list_head *timeout_list)
 	expiry = round_jiffies_up(req->deadline);
 
 	if (!timer_pending(&q->timeout) ||
-	    time_before(expiry, q->timeout.expires))
-		mod_timer(&q->timeout, expiry);
+	    time_before(expiry, q->timeout.expires)) {
+		unsigned long diff = q->timeout.expires - expiry;
 
-}
+		/*
+		 * Due to added timer slack to group timers, the timer
+		 * will often be a little in front of what we asked for.
+		 * So apply some tolerance here too, otherwise we keep
+		 * modifying the timer because expires for value X
+		 * will be X + something.
+		 */
+		if (diff >= HZ / 2)
+			mod_timer(&q->timeout, expiry);
+	}
 
-/**
- * blk_add_timer - Start timeout timer for a single request
- * @req:	request that is about to start running.
- *
- * Notes:
- *    Each request has its own timer, and as it is added to the queue, we
- *    set up the timer. When the request completes, we cancel the timer.
- */
-void blk_add_timer(struct request *req)
-{
-	__blk_add_timer(req, &req->q->timeout_list);
 }
-
diff --git a/block/blk.h b/block/blk.h
index 1d880f1f957f..79be2cbce7fd 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -37,9 +37,8 @@ bool __blk_end_bidi_request(struct request *rq, int error,
 void blk_rq_timed_out_timer(unsigned long data);
 void blk_rq_check_expired(struct request *rq, unsigned long *next_timeout,
 			  unsigned int *next_set);
-void __blk_add_timer(struct request *req, struct list_head *timeout_list);
+void blk_add_timer(struct request *req);
 void blk_delete_timer(struct request *);
-void blk_add_timer(struct request *);
 
 
 bool bio_attempt_front_merge(struct request_queue *q, struct request *req,
diff --git a/block/bsg.c b/block/bsg.c
index 420a5a9f1b23..e5214c148096 100644
--- a/block/bsg.c
+++ b/block/bsg.c
@@ -1008,7 +1008,7 @@ int bsg_register_queue(struct request_queue *q, struct device *parent,
 	/*
 	 * we need a proper transport to send commands, not a stacked device
 	 */
-	if (!q->request_fn)
+	if (!queue_is_rq_based(q))
 		return 0;
 
 	bcd = &q->bsg_dev;
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index e0985f1955e7..22dffebc7c73 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -908,7 +908,7 @@ static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
 {
 	if (cfqd->busy_queues) {
 		cfq_log(cfqd, "schedule dispatch");
-		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
+		kblockd_schedule_work(&cfqd->unplug_work);
 	}
 }
 
@@ -4460,7 +4460,7 @@ out_free:
 static ssize_t
 cfq_var_show(unsigned int var, char *page)
 {
-	return sprintf(page, "%d\n", var);
+	return sprintf(page, "%u\n", var);
 }
 
 static ssize_t