All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] blk-mq: Revert multiple recent patches
@ 2017-11-07  1:07 ` Bart Van Assche
  0 siblings, 0 replies; 5+ messages in thread
From: Bart Van Assche @ 2017-11-07  1:07 UTC (permalink / raw)
  To: Jens Axboe
  Cc: linux-block, Christoph Hellwig, linux-scsi, Bart Van Assche,
	Martin K . Petersen, Ming Lei

Since we are close to the merge window and since it is too late for
reworking patches and algorithms that have turned out to be flawed,
revert multiple recent patches. Revert patch "blk-mq: don't handle
TAG_SHARED in restart" and all patches that conflict with the revert
of that patch because that patch is based on an incorrect assumption,
namely that the SCSI starved list can be used as a replacement for
the blk-mq queue restarting mechanism. Revert patch "scsi: implement
.get_budget and .put_budget for blk-mq" because it triggers an
infinite loop on my test setup ("scsi 9:0:0:0: rejecting I/O to dead
device" + RCU lockup complaint in blk_mq_sched_dispatch_requests()).

The list of patches that are reverted by this patch is as follows:

blk-mq: don't allocate driver tag upfront for flush rq
blk-mq: move blk_mq_put_driver_tag*() into blk-mq.h
blk-mq-sched: decide how to handle flush rq via RQF_FLUSH_SEQ
blk-flush: use blk_mq_request_bypass_insert()
block: pass 'run_queue' to blk_mq_request_bypass_insert
blk-flush: don't run queue for requests bypassing flush
blk-mq: put the driver tag of nxt rq before first one is requeued
blk-mq: don't handle failure in .get_budget
SCSI: don't get target/host busy_count in scsi_mq_get_budget()
blk-mq: don't restart queue when .get_budget returns BLK_STS_RESOURCE
blk-mq: don't handle TAG_SHARED in restart
scsi: implement .get_budget and .put_budget for blk-mq

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Ming Lei <ming.lei@redhat.com>
---
 block/blk-core.c        |   2 +-
 block/blk-flush.c       |  37 +++------
 block/blk-mq-sched.c    | 200 +++++++++++++++++++++++++++++++++++++-----------
 block/blk-mq-sched.h    |   2 +-
 block/blk-mq.c          | 122 +++++++++++++++++++++++------
 block/blk-mq.h          |  40 +---------
 drivers/scsi/scsi_lib.c |  47 ++++--------
 include/linux/blk-mq.h  |   2 +-
 8 files changed, 285 insertions(+), 167 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 567607f7ea48..25ccddd0b82a 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -2402,7 +2402,7 @@ blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *
 		 * bypass a potential scheduler on the bottom device for
 		 * insert.
 		 */
-		blk_mq_request_bypass_insert(rq, true);
+		blk_mq_request_bypass_insert(rq);
 		return BLK_STS_OK;
 	}
 
diff --git a/block/blk-flush.c b/block/blk-flush.c
index f17170675917..4938bec8cfef 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -231,13 +231,8 @@ static void flush_end_io(struct request *flush_rq, blk_status_t error)
 		/* release the tag's ownership to the req cloned from */
 		spin_lock_irqsave(&fq->mq_flush_lock, flags);
 		hctx = blk_mq_map_queue(q, flush_rq->mq_ctx->cpu);
-		if (!q->elevator) {
-			blk_mq_tag_set_rq(hctx, flush_rq->tag, fq->orig_rq);
-			flush_rq->tag = -1;
-		} else {
-			blk_mq_put_driver_tag_hctx(hctx, flush_rq);
-			flush_rq->internal_tag = -1;
-		}
+		blk_mq_tag_set_rq(hctx, flush_rq->tag, fq->orig_rq);
+		flush_rq->tag = -1;
 	}
 
 	running = &fq->flush_queue[fq->flush_running_idx];
@@ -323,26 +318,19 @@ static bool blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq)
 	blk_rq_init(q, flush_rq);
 
 	/*
-	 * In case of none scheduler, borrow tag from the first request
-	 * since they can't be in flight at the same time. And acquire
-	 * the tag's ownership for flush req.
-	 *
-	 * In case of IO scheduler, flush rq need to borrow scheduler tag
-	 * just for cheating put/get driver tag.
+	 * Borrow tag from the first request since they can't
+	 * be in flight at the same time. And acquire the tag's
+	 * ownership for flush req.
 	 */
 	if (q->mq_ops) {
 		struct blk_mq_hw_ctx *hctx;
 
 		flush_rq->mq_ctx = first_rq->mq_ctx;
+		flush_rq->tag = first_rq->tag;
+		fq->orig_rq = first_rq;
 
-		if (!q->elevator) {
-			fq->orig_rq = first_rq;
-			flush_rq->tag = first_rq->tag;
-			hctx = blk_mq_map_queue(q, first_rq->mq_ctx->cpu);
-			blk_mq_tag_set_rq(hctx, first_rq->tag, flush_rq);
-		} else {
-			flush_rq->internal_tag = first_rq->internal_tag;
-		}
+		hctx = blk_mq_map_queue(q, first_rq->mq_ctx->cpu);
+		blk_mq_tag_set_rq(hctx, first_rq->tag, flush_rq);
 	}
 
 	flush_rq->cmd_flags = REQ_OP_FLUSH | REQ_PREFLUSH;
@@ -406,11 +394,6 @@ static void mq_flush_data_end_io(struct request *rq, blk_status_t error)
 
 	hctx = blk_mq_map_queue(q, ctx->cpu);
 
-	if (q->elevator) {
-		WARN_ON(rq->tag < 0);
-		blk_mq_put_driver_tag_hctx(hctx, rq);
-	}
-
 	/*
 	 * After populating an empty queue, kick it to avoid stall.  Read
 	 * the comment in flush_end_io().
@@ -480,7 +463,7 @@ void blk_insert_flush(struct request *rq)
 	if ((policy & REQ_FSEQ_DATA) &&
 	    !(policy & (REQ_FSEQ_PREFLUSH | REQ_FSEQ_POSTFLUSH))) {
 		if (q->mq_ops)
-			blk_mq_request_bypass_insert(rq, false);
+			blk_mq_sched_insert_request(rq, false, true, false, false);
 		else
 			list_add_tail(&rq->queuelist, &q->queue_head);
 		return;
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index 01a43fed6b8c..df8581bb0a37 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -68,25 +68,29 @@ static void blk_mq_sched_mark_restart_hctx(struct blk_mq_hw_ctx *hctx)
 		set_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
 }
 
-void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)
+static bool blk_mq_sched_restart_hctx(struct blk_mq_hw_ctx *hctx)
 {
 	if (!test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
-		return;
+		return false;
+
+	if (hctx->flags & BLK_MQ_F_TAG_SHARED) {
+		struct request_queue *q = hctx->queue;
 
-	clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
+		if (test_and_clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
+			atomic_dec(&q->shared_hctx_restart);
+	} else
+		clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
 
 	if (blk_mq_hctx_has_pending(hctx)) {
 		blk_mq_run_hw_queue(hctx, true);
-		return;
+		return true;
 	}
+
+	return false;
 }
 
-/*
- * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
- * its queue by itself in its completion handler, so we don't need to
- * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
- */
-static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
+/* return true if hctx need to run again */
+static bool blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	struct elevator_queue *e = q->elevator;
@@ -94,18 +98,23 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 
 	do {
 		struct request *rq;
+		blk_status_t ret;
 
 		if (e->type->ops.mq.has_work &&
 				!e->type->ops.mq.has_work(hctx))
 			break;
 
-		if (!blk_mq_get_dispatch_budget(hctx))
-			break;
+		ret = blk_mq_get_dispatch_budget(hctx);
+		if (ret == BLK_STS_RESOURCE)
+			return true;
 
 		rq = e->type->ops.mq.dispatch_request(hctx);
 		if (!rq) {
 			blk_mq_put_dispatch_budget(hctx);
 			break;
+		} else if (ret != BLK_STS_OK) {
+			blk_mq_end_request(rq, ret);
+			continue;
 		}
 
 		/*
@@ -115,6 +124,8 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 		 */
 		list_add(&rq->queuelist, &rq_list);
 	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));
+
+	return false;
 }
 
 static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
@@ -128,12 +139,8 @@ static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
 	return hctx->ctxs[idx];
 }
 
-/*
- * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
- * its queue by itself in its completion handler, so we don't need to
- * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
- */
-static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
+/* return true if hctx need to run again */
+static bool blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	LIST_HEAD(rq_list);
@@ -141,17 +148,22 @@ static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 
 	do {
 		struct request *rq;
+		blk_status_t ret;
 
 		if (!sbitmap_any_bit_set(&hctx->ctx_map))
 			break;
 
-		if (!blk_mq_get_dispatch_budget(hctx))
-			break;
+		ret = blk_mq_get_dispatch_budget(hctx);
+		if (ret == BLK_STS_RESOURCE)
+			return true;
 
 		rq = blk_mq_dequeue_from_ctx(hctx, ctx);
 		if (!rq) {
 			blk_mq_put_dispatch_budget(hctx);
 			break;
+		} else if (ret != BLK_STS_OK) {
+			blk_mq_end_request(rq, ret);
+			continue;
 		}
 
 		/*
@@ -167,19 +179,22 @@ static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));
 
 	WRITE_ONCE(hctx->dispatch_from, ctx);
+
+	return false;
 }
 
 /* return true if hw queue need to be run again */
-void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
+bool blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	struct elevator_queue *e = q->elevator;
 	const bool has_sched_dispatch = e && e->type->ops.mq.dispatch_request;
 	LIST_HEAD(rq_list);
+	bool run_queue = false;
 
 	/* RCU or SRCU read lock is needed before checking quiesced flag */
 	if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q)))
-		return;
+		return false;
 
 	hctx->run++;
 
@@ -211,12 +226,12 @@ void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 		blk_mq_sched_mark_restart_hctx(hctx);
 		if (blk_mq_dispatch_rq_list(q, &rq_list, false)) {
 			if (has_sched_dispatch)
-				blk_mq_do_dispatch_sched(hctx);
+				run_queue = blk_mq_do_dispatch_sched(hctx);
 			else
-				blk_mq_do_dispatch_ctx(hctx);
+				run_queue = blk_mq_do_dispatch_ctx(hctx);
 		}
 	} else if (has_sched_dispatch) {
-		blk_mq_do_dispatch_sched(hctx);
+		run_queue = blk_mq_do_dispatch_sched(hctx);
 	} else if (q->mq_ops->get_budget) {
 		/*
 		 * If we need to get budget before queuing request, we
@@ -226,11 +241,19 @@ void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 		 * TODO: get more budgets, and dequeue more requests in
 		 * one time.
 		 */
-		blk_mq_do_dispatch_ctx(hctx);
+		run_queue = blk_mq_do_dispatch_ctx(hctx);
 	} else {
 		blk_mq_flush_busy_ctxs(hctx, &rq_list);
 		blk_mq_dispatch_rq_list(q, &rq_list, false);
 	}
+
+	if (run_queue && !blk_mq_sched_needs_restart(hctx) &&
+			!test_bit(BLK_MQ_S_TAG_WAITING, &hctx->state)) {
+		blk_mq_sched_mark_restart_hctx(hctx);
+		return true;
+	}
+
+	return false;
 }
 
 bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
@@ -345,21 +368,98 @@ void blk_mq_sched_request_inserted(struct request *rq)
 EXPORT_SYMBOL_GPL(blk_mq_sched_request_inserted);
 
 static bool blk_mq_sched_bypass_insert(struct blk_mq_hw_ctx *hctx,
-				       bool has_sched,
 				       struct request *rq)
 {
-	/* dispatch flush rq directly */
-	if (rq->rq_flags & RQF_FLUSH_SEQ) {
-		spin_lock(&hctx->lock);
-		list_add(&rq->queuelist, &hctx->dispatch);
-		spin_unlock(&hctx->lock);
-		return true;
+	if (rq->tag == -1) {
+		rq->rq_flags |= RQF_SORTED;
+		return false;
 	}
 
-	if (has_sched)
-		rq->rq_flags |= RQF_SORTED;
+	/*
+	 * If we already have a real request tag, send directly to
+	 * the dispatch list.
+	 */
+	spin_lock(&hctx->lock);
+	list_add(&rq->queuelist, &hctx->dispatch);
+	spin_unlock(&hctx->lock);
+	return true;
+}
 
-	return false;
+/**
+ * list_for_each_entry_rcu_rr - iterate in a round-robin fashion over rcu list
+ * @pos:    loop cursor.
+ * @skip:   the list element that will not be examined. Iteration starts at
+ *          @skip->next.
+ * @head:   head of the list to examine. This list must have at least one
+ *          element, namely @skip.
+ * @member: name of the list_head structure within typeof(*pos).
+ */
+#define list_for_each_entry_rcu_rr(pos, skip, head, member)		\
+	for ((pos) = (skip);						\
+	     (pos = (pos)->member.next != (head) ? list_entry_rcu(	\
+			(pos)->member.next, typeof(*pos), member) :	\
+	      list_entry_rcu((pos)->member.next->next, typeof(*pos), member)), \
+	     (pos) != (skip); )
+
+/*
+ * Called after a driver tag has been freed to check whether a hctx needs to
+ * be restarted. Restarts @hctx if its tag set is not shared. Restarts hardware
+ * queues in a round-robin fashion if the tag set of @hctx is shared with other
+ * hardware queues.
+ */
+void blk_mq_sched_restart(struct blk_mq_hw_ctx *const hctx)
+{
+	struct blk_mq_tags *const tags = hctx->tags;
+	struct blk_mq_tag_set *const set = hctx->queue->tag_set;
+	struct request_queue *const queue = hctx->queue, *q;
+	struct blk_mq_hw_ctx *hctx2;
+	unsigned int i, j;
+
+	if (set->flags & BLK_MQ_F_TAG_SHARED) {
+		/*
+		 * If this is 0, then we know that no hardware queues
+		 * have RESTART marked. We're done.
+		 */
+		if (!atomic_read(&queue->shared_hctx_restart))
+			return;
+
+		rcu_read_lock();
+		list_for_each_entry_rcu_rr(q, queue, &set->tag_list,
+					   tag_set_list) {
+			queue_for_each_hw_ctx(q, hctx2, i)
+				if (hctx2->tags == tags &&
+				    blk_mq_sched_restart_hctx(hctx2))
+					goto done;
+		}
+		j = hctx->queue_num + 1;
+		for (i = 0; i < queue->nr_hw_queues; i++, j++) {
+			if (j == queue->nr_hw_queues)
+				j = 0;
+			hctx2 = queue->queue_hw_ctx[j];
+			if (hctx2->tags == tags &&
+			    blk_mq_sched_restart_hctx(hctx2))
+				break;
+		}
+done:
+		rcu_read_unlock();
+	} else {
+		blk_mq_sched_restart_hctx(hctx);
+	}
+}
+
+/*
+ * Add flush/fua to the queue. If we fail getting a driver tag, then
+ * punt to the requeue list. Requeue will re-invoke us from a context
+ * that's safe to block from.
+ */
+static void blk_mq_sched_insert_flush(struct blk_mq_hw_ctx *hctx,
+				      struct request *rq, bool can_block)
+{
+	if (blk_mq_get_driver_tag(rq, &hctx, can_block)) {
+		blk_insert_flush(rq);
+		blk_mq_run_hw_queue(hctx, true);
+	} else
+		blk_mq_add_to_requeue_list(rq, false, true);
 }
 
 void blk_mq_sched_insert_request(struct request *rq, bool at_head,
@@ -370,15 +470,12 @@ void blk_mq_sched_insert_request(struct request *rq, bool at_head,
 	struct blk_mq_ctx *ctx = rq->mq_ctx;
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 
-	/* flush rq in flush machinery need to be dispatched directly */
-	if (!(rq->rq_flags & RQF_FLUSH_SEQ) && op_is_flush(rq->cmd_flags)) {
-		blk_insert_flush(rq);
-		goto run;
+	if (rq->tag == -1 && op_is_flush(rq->cmd_flags)) {
+		blk_mq_sched_insert_flush(hctx, rq, can_block);
+		return;
 	}
 
-	WARN_ON(e && (rq->tag != -1));
-
-	if (blk_mq_sched_bypass_insert(hctx, !!e, rq))
+	if (e && blk_mq_sched_bypass_insert(hctx, rq))
 		goto run;
 
 	if (e && e->type->ops.mq.insert_requests) {
@@ -404,6 +501,23 @@ void blk_mq_sched_insert_requests(struct request_queue *q,
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 	struct elevator_queue *e = hctx->queue->elevator;
 
+	if (e) {
+		struct request *rq, *next;
+
+		/*
+		 * We bypass requests that already have a driver tag assigned,
+		 * which should only be flushes. Flushes are only ever inserted
+		 * as single requests, so we shouldn't ever hit the
+		 * WARN_ON_ONCE() below (but let's handle it just in case).
+		 */
+		list_for_each_entry_safe(rq, next, list, queuelist) {
+			if (WARN_ON_ONCE(rq->tag != -1)) {
+				list_del_init(&rq->queuelist);
+				blk_mq_sched_bypass_insert(hctx, rq);
+			}
+		}
+	}
+
 	if (e && e->type->ops.mq.insert_requests)
 		e->type->ops.mq.insert_requests(hctx, list, false);
 	else
diff --git a/block/blk-mq-sched.h b/block/blk-mq-sched.h
index 9267d0b7c197..2434061cc5b7 100644
--- a/block/blk-mq-sched.h
+++ b/block/blk-mq-sched.h
@@ -22,7 +22,7 @@ void blk_mq_sched_insert_requests(struct request_queue *q,
 				  struct blk_mq_ctx *ctx,
 				  struct list_head *list, bool run_queue_async);
 
-void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx);
+bool blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx);
 
 int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e);
 void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 9b49a0d16907..4fed71fd74c0 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -653,8 +653,6 @@ static void __blk_mq_requeue_request(struct request *rq)
 {
 	struct request_queue *q = rq->q;
 
-	blk_mq_put_driver_tag(rq);
-
 	trace_block_rq_requeue(q, rq);
 	wbt_requeue(q->rq_wb, &rq->issue_stat);
 	blk_mq_sched_requeue_request(rq);
@@ -1004,6 +1002,62 @@ bool blk_mq_get_driver_tag(struct request *rq, struct blk_mq_hw_ctx **hctx,
 	return rq->tag != -1;
 }
 
+static void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
+				    struct request *rq)
+{
+	blk_mq_put_tag(hctx, hctx->tags, rq->mq_ctx, rq->tag);
+	rq->tag = -1;
+
+	if (rq->rq_flags & RQF_MQ_INFLIGHT) {
+		rq->rq_flags &= ~RQF_MQ_INFLIGHT;
+		atomic_dec(&hctx->nr_active);
+	}
+}
+
+static void blk_mq_put_driver_tag_hctx(struct blk_mq_hw_ctx *hctx,
+				       struct request *rq)
+{
+	if (rq->tag == -1 || rq->internal_tag == -1)
+		return;
+
+	__blk_mq_put_driver_tag(hctx, rq);
+}
+
+static void blk_mq_put_driver_tag(struct request *rq)
+{
+	struct blk_mq_hw_ctx *hctx;
+
+	if (rq->tag == -1 || rq->internal_tag == -1)
+		return;
+
+	hctx = blk_mq_map_queue(rq->q, rq->mq_ctx->cpu);
+	__blk_mq_put_driver_tag(hctx, rq);
+}
+
+/*
+ * If we fail getting a driver tag because all the driver tags are already
+ * assigned and on the dispatch list, BUT the first entry does not have a
+ * tag, then we could deadlock. For that case, move entries with assigned
+ * driver tags to the front, leaving the set of tagged requests in the
+ * same order, and the untagged set in the same order.
+ */
+static bool reorder_tags_to_front(struct list_head *list)
+{
+	struct request *rq, *tmp, *first = NULL;
+
+	list_for_each_entry_safe_reverse(rq, tmp, list, queuelist) {
+		if (rq == first)
+			break;
+		if (rq->tag != -1) {
+			list_move(&rq->queuelist, list);
+			if (!first)
+				first = rq;
+		}
+	}
+
+	return first != NULL;
+}
+
 static int blk_mq_dispatch_wake(wait_queue_entry_t *wait, unsigned mode, int flags,
 				void *key)
 {
@@ -1046,7 +1100,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 		bool got_budget)
 {
 	struct blk_mq_hw_ctx *hctx;
-	struct request *rq, *nxt;
+	struct request *rq;
 	int errors, queued;
 
 	if (list_empty(list))
@@ -1064,6 +1118,9 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 
 		rq = list_first_entry(list, struct request, queuelist);
 		if (!blk_mq_get_driver_tag(rq, &hctx, false)) {
+			if (!queued && reorder_tags_to_front(list))
+				continue;
+
 			/*
 			 * The initial allocation attempt failed, so we need to
 			 * rerun the hardware queue when a tag is freed.
@@ -1086,8 +1143,13 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 			}
 		}
 
-		if (!got_budget && !blk_mq_get_dispatch_budget(hctx))
-			break;
+		if (!got_budget) {
+			ret = blk_mq_get_dispatch_budget(hctx);
+			if (ret == BLK_STS_RESOURCE)
+				break;
+			if (ret != BLK_STS_OK)
+				goto fail_rq;
+		}
 
 		list_del_init(&rq->queuelist);
 
@@ -1100,25 +1162,21 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 		if (list_empty(list))
 			bd.last = true;
 		else {
+			struct request *nxt;
+
 			nxt = list_first_entry(list, struct request, queuelist);
 			bd.last = !blk_mq_get_driver_tag(nxt, NULL, false);
 		}
 
 		ret = q->mq_ops->queue_rq(hctx, &bd);
 		if (ret == BLK_STS_RESOURCE) {
-			/*
-			 * If an I/O scheduler has been configured and we got a
-			 * driver tag for the next request already, free it again.
-			 */
-			if (!list_empty(list)) {
-				nxt = list_first_entry(list, struct request, queuelist);
-				blk_mq_put_driver_tag(nxt);
-			}
+			blk_mq_put_driver_tag_hctx(hctx, rq);
 			list_add(&rq->queuelist, list);
 			__blk_mq_requeue_request(rq);
 			break;
 		}
 
+ fail_rq:
 		if (unlikely(ret != BLK_STS_OK)) {
 			errors++;
 			blk_mq_end_request(rq, BLK_STS_IOERR);
@@ -1135,6 +1193,13 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 	 * that is where we will continue on next queue run.
 	 */
 	if (!list_empty(list)) {
+		/*
+		 * If an I/O scheduler has been configured and we got a driver
+		 * tag for the next request already, free it again.
+		 */
+		rq = list_first_entry(list, struct request, queuelist);
+		blk_mq_put_driver_tag(rq);
+
 		spin_lock(&hctx->lock);
 		list_splice_init(list, &hctx->dispatch);
 		spin_unlock(&hctx->lock);
@@ -1170,6 +1235,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 {
 	int srcu_idx;
+	bool run_queue;
 
 	/*
 	 * We should be running this queue from one of the CPUs that
@@ -1186,15 +1252,18 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 
 	if (!(hctx->flags & BLK_MQ_F_BLOCKING)) {
 		rcu_read_lock();
-		blk_mq_sched_dispatch_requests(hctx);
+		run_queue = blk_mq_sched_dispatch_requests(hctx);
 		rcu_read_unlock();
 	} else {
 		might_sleep();
 
 		srcu_idx = srcu_read_lock(hctx->queue_rq_srcu);
-		blk_mq_sched_dispatch_requests(hctx);
+		run_queue = blk_mq_sched_dispatch_requests(hctx);
 		srcu_read_unlock(hctx->queue_rq_srcu, srcu_idx);
 	}
+
+	if (run_queue)
+		blk_mq_run_hw_queue(hctx, true);
 }
 
 /*
@@ -1440,7 +1509,7 @@ void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
  * Should only be used carefully, when the caller knows we want to
  * bypass a potential IO scheduler on the target device.
  */
-void blk_mq_request_bypass_insert(struct request *rq, bool run_queue)
+void blk_mq_request_bypass_insert(struct request *rq)
 {
 	struct blk_mq_ctx *ctx = rq->mq_ctx;
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(rq->q, ctx->cpu);
@@ -1449,8 +1518,7 @@ void blk_mq_request_bypass_insert(struct request *rq, bool run_queue)
 	list_add_tail(&rq->queuelist, &hctx->dispatch);
 	spin_unlock(&hctx->lock);
 
-	if (run_queue)
-		blk_mq_run_hw_queue(hctx, false);
+	blk_mq_run_hw_queue(hctx, false);
 }
 
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
@@ -1584,10 +1652,12 @@ static void __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 	if (!blk_mq_get_driver_tag(rq, NULL, false))
 		goto insert;
 
-	if (!blk_mq_get_dispatch_budget(hctx)) {
+	ret = blk_mq_get_dispatch_budget(hctx);
+	if (ret == BLK_STS_RESOURCE) {
 		blk_mq_put_driver_tag(rq);
 		goto insert;
-	}
+	} else if (ret != BLK_STS_OK)
+		goto fail_rq;
 
 	new_cookie = request_to_qc_t(hctx, rq);
 
@@ -1605,6 +1675,7 @@ static void __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 		__blk_mq_requeue_request(rq);
 		goto insert;
 	default:
+ fail_rq:
 		*cookie = BLK_QC_T_NONE;
 		blk_mq_end_request(rq, ret);
 		return;
@@ -1678,10 +1749,13 @@ static blk_qc_t blk_mq_make_request(struct request_queue *q, struct bio *bio)
 	if (unlikely(is_flush_fua)) {
 		blk_mq_put_ctx(data.ctx);
 		blk_mq_bio_to_request(rq, bio);
-
-		/* bypass scheduler for flush rq */
-		blk_insert_flush(rq);
-		blk_mq_run_hw_queue(data.hctx, true);
+		if (q->elevator) {
+			blk_mq_sched_insert_request(rq, false, true, true,
+					true);
+		} else {
+			blk_insert_flush(rq);
+			blk_mq_run_hw_queue(data.hctx, true);
+		}
 	} else if (plug && q->nr_hw_queues == 1) {
 		struct request *last = NULL;
 
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 2502f40ccdc0..522b420dedc0 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -2,7 +2,6 @@
 #define INT_BLK_MQ_H
 
 #include "blk-stat.h"
-#include "blk-mq-tag.h"
 
 struct blk_mq_tag_set;
 
@@ -57,7 +56,7 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
  */
 void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
 				bool at_head);
-void blk_mq_request_bypass_insert(struct request *rq, bool run_queue);
+void blk_mq_request_bypass_insert(struct request *rq);
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
 				struct list_head *list);
 
@@ -148,45 +147,14 @@ static inline void blk_mq_put_dispatch_budget(struct blk_mq_hw_ctx *hctx)
 		q->mq_ops->put_budget(hctx);
 }
 
-static inline bool blk_mq_get_dispatch_budget(struct blk_mq_hw_ctx *hctx)
+static inline blk_status_t blk_mq_get_dispatch_budget(
+		struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 
 	if (q->mq_ops->get_budget)
 		return q->mq_ops->get_budget(hctx);
-	return true;
-}
-
-static inline void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
-					   struct request *rq)
-{
-	blk_mq_put_tag(hctx, hctx->tags, rq->mq_ctx, rq->tag);
-	rq->tag = -1;
-
-	if (rq->rq_flags & RQF_MQ_INFLIGHT) {
-		rq->rq_flags &= ~RQF_MQ_INFLIGHT;
-		atomic_dec(&hctx->nr_active);
-	}
-}
-
-static inline void blk_mq_put_driver_tag_hctx(struct blk_mq_hw_ctx *hctx,
-				       struct request *rq)
-{
-	if (rq->tag == -1 || rq->internal_tag == -1)
-		return;
-
-	__blk_mq_put_driver_tag(hctx, rq);
-}
-
-static inline void blk_mq_put_driver_tag(struct request *rq)
-{
-	struct blk_mq_hw_ctx *hctx;
-
-	if (rq->tag == -1 || rq->internal_tag == -1)
-		return;
-
-	hctx = blk_mq_map_queue(rq->q, rq->mq_ctx->cpu);
-	__blk_mq_put_driver_tag(hctx, rq);
+	return BLK_STS_OK;
 }
 
 #endif
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index c537c4d768c1..5634288f244f 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1940,33 +1940,6 @@ static void scsi_mq_done(struct scsi_cmnd *cmd)
 	blk_mq_complete_request(cmd->request);
 }
 
-static void scsi_mq_put_budget(struct blk_mq_hw_ctx *hctx)
-{
-	struct request_queue *q = hctx->queue;
-	struct scsi_device *sdev = q->queuedata;
-
-	atomic_dec(&sdev->device_busy);
-	put_device(&sdev->sdev_gendev);
-}
-
-static bool scsi_mq_get_budget(struct blk_mq_hw_ctx *hctx)
-{
-	struct request_queue *q = hctx->queue;
-	struct scsi_device *sdev = q->queuedata;
-
-	if (!get_device(&sdev->sdev_gendev))
-		goto out;
-	if (!scsi_dev_queue_ready(q, sdev))
-		goto out_put_device;
-
-	return true;
-
-out_put_device:
-	put_device(&sdev->sdev_gendev);
-out:
-	return false;
-}
-
 static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 			 const struct blk_mq_queue_data *bd)
 {
@@ -1980,11 +1953,16 @@ static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 
 	ret = prep_to_mq(scsi_prep_state_check(sdev, req));
 	if (ret != BLK_STS_OK)
-		goto out_put_budget;
+		goto out;
 
 	ret = BLK_STS_RESOURCE;
+	if (!get_device(&sdev->sdev_gendev))
+		goto out;
+
+	if (!scsi_dev_queue_ready(q, sdev))
+		goto out_put_device;
 	if (!scsi_target_queue_ready(shost, sdev))
-		goto out_put_budget;
+		goto out_dec_device_busy;
 	if (!scsi_host_queue_ready(q, shost, sdev))
 		goto out_dec_target_busy;
 
@@ -2015,12 +1993,15 @@ static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 	return BLK_STS_OK;
 
 out_dec_host_busy:
-       atomic_dec(&shost->host_busy);
+	atomic_dec(&shost->host_busy);
 out_dec_target_busy:
 	if (scsi_target(sdev)->can_queue > 0)
 		atomic_dec(&scsi_target(sdev)->target_busy);
-out_put_budget:
-	scsi_mq_put_budget(hctx);
+out_dec_device_busy:
+	atomic_dec(&sdev->device_busy);
+out_put_device:
+	put_device(&sdev->sdev_gendev);
+out:
 	switch (ret) {
 	case BLK_STS_OK:
 		break;
@@ -2224,8 +2205,6 @@ struct request_queue *scsi_old_alloc_queue(struct scsi_device *sdev)
 }
 
 static const struct blk_mq_ops scsi_mq_ops = {
-	.get_budget	= scsi_mq_get_budget,
-	.put_budget	= scsi_mq_put_budget,
 	.queue_rq	= scsi_queue_rq,
 	.complete	= scsi_softirq_done,
 	.timeout	= scsi_timeout,
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 674641527da7..f2e3079eecdd 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -92,7 +92,7 @@ struct blk_mq_queue_data {
 
 typedef blk_status_t (queue_rq_fn)(struct blk_mq_hw_ctx *,
 		const struct blk_mq_queue_data *);
-typedef bool (get_budget_fn)(struct blk_mq_hw_ctx *);
+typedef blk_status_t (get_budget_fn)(struct blk_mq_hw_ctx *);
 typedef void (put_budget_fn)(struct blk_mq_hw_ctx *);
 typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
 typedef int (init_hctx_fn)(struct blk_mq_hw_ctx *, void *, unsigned int);
-- 
2.14.3

^ permalink raw reply related	[flat|nested] 5+ messages in thread
* [PATCH] blk-mq: Revert multiple recent patches
@ 2017-11-07  1:07 ` Bart Van Assche
  0 siblings, 0 replies; 5+ messages in thread
From: Bart Van Assche @ 2017-11-07  1:07 UTC (permalink / raw)
  To: Jens Axboe
  Cc: linux-block, Christoph Hellwig, linux-scsi, Bart Van Assche,
	Martin K . Petersen, Ming Lei

Since we are close to the merge window and since it is too late for
reworking patches and algorithms that have turned out to be flawed,
revert multiple recent patches. Revert patch "blk-mq: don't handle
TAG_SHARED in restart" and all patches that conflict with the revert
of that patch because that patch is based on an incorrect assumption,
namely that the SCSI starved list can be used as a replacement for
the blk-mq queue restarting mechanism. Revert patch "scsi: implement
.get_budget and .put_budget for blk-mq" because it triggers an
infinite loop on my test setup ("scsi 9:0:0:0: rejecting I/O to dead
device" + RCU lockup complaint in blk_mq_sched_dispatch_requests()).

The list of patches that are reverted by this patch is as follows:

blk-mq: don't allocate driver tag upfront for flush rq
blk-mq: move blk_mq_put_driver_tag*() into blk-mq.h
blk-mq-sched: decide how to handle flush rq via RQF_FLUSH_SEQ
blk-flush: use blk_mq_request_bypass_insert()
block: pass 'run_queue' to blk_mq_request_bypass_insert
blk-flush: don't run queue for requests bypassing flush
blk-mq: put the driver tag of nxt rq before first one is requeued
blk-mq: don't handle failure in .get_budget
SCSI: don't get target/host busy_count in scsi_mq_get_budget()
blk-mq: don't restart queue when .get_budget returns BLK_STS_RESOURCE
blk-mq: don't handle TAG_SHARED in restart
scsi: implement .get_budget and .put_budget for blk-mq

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Ming Lei <ming.lei@redhat.com>
---
 block/blk-core.c        |   2 +-
 block/blk-flush.c       |  37 +++------
 block/blk-mq-sched.c    | 200 +++++++++++++++++++++++++++++++++++++-----------
 block/blk-mq-sched.h    |   2 +-
 block/blk-mq.c          | 122 +++++++++++++++++++++++------
 block/blk-mq.h          |  40 +---------
 drivers/scsi/scsi_lib.c |  47 ++++--------
 include/linux/blk-mq.h  |   2 +-
 8 files changed, 285 insertions(+), 167 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 567607f7ea48..25ccddd0b82a 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -2402,7 +2402,7 @@ blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *
 		 * bypass a potential scheduler on the bottom device for
 		 * insert.
 		 */
-		blk_mq_request_bypass_insert(rq, true);
+		blk_mq_request_bypass_insert(rq);
 		return BLK_STS_OK;
 	}
 
diff --git a/block/blk-flush.c b/block/blk-flush.c
index f17170675917..4938bec8cfef 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -231,13 +231,8 @@ static void flush_end_io(struct request *flush_rq, blk_status_t error)
 		/* release the tag's ownership to the req cloned from */
 		spin_lock_irqsave(&fq->mq_flush_lock, flags);
 		hctx = blk_mq_map_queue(q, flush_rq->mq_ctx->cpu);
-		if (!q->elevator) {
-			blk_mq_tag_set_rq(hctx, flush_rq->tag, fq->orig_rq);
-			flush_rq->tag = -1;
-		} else {
-			blk_mq_put_driver_tag_hctx(hctx, flush_rq);
-			flush_rq->internal_tag = -1;
-		}
+		blk_mq_tag_set_rq(hctx, flush_rq->tag, fq->orig_rq);
+		flush_rq->tag = -1;
 	}
 
 	running = &fq->flush_queue[fq->flush_running_idx];
@@ -323,26 +318,19 @@ static bool blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq)
 	blk_rq_init(q, flush_rq);
 
 	/*
-	 * In case of none scheduler, borrow tag from the first request
-	 * since they can't be in flight at the same time. And acquire
-	 * the tag's ownership for flush req.
-	 *
-	 * In case of IO scheduler, flush rq need to borrow scheduler tag
-	 * just for cheating put/get driver tag.
+	 * Borrow tag from the first request since they can't
+	 * be in flight at the same time. And acquire the tag's
+	 * ownership for flush req.
 	 */
 	if (q->mq_ops) {
 		struct blk_mq_hw_ctx *hctx;
 
 		flush_rq->mq_ctx = first_rq->mq_ctx;
+		flush_rq->tag = first_rq->tag;
+		fq->orig_rq = first_rq;
 
-		if (!q->elevator) {
-			fq->orig_rq = first_rq;
-			flush_rq->tag = first_rq->tag;
-			hctx = blk_mq_map_queue(q, first_rq->mq_ctx->cpu);
-			blk_mq_tag_set_rq(hctx, first_rq->tag, flush_rq);
-		} else {
-			flush_rq->internal_tag = first_rq->internal_tag;
-		}
+		hctx = blk_mq_map_queue(q, first_rq->mq_ctx->cpu);
+		blk_mq_tag_set_rq(hctx, first_rq->tag, flush_rq);
 	}
 
 	flush_rq->cmd_flags = REQ_OP_FLUSH | REQ_PREFLUSH;
@@ -406,11 +394,6 @@ static void mq_flush_data_end_io(struct request *rq, blk_status_t error)
 
 	hctx = blk_mq_map_queue(q, ctx->cpu);
 
-	if (q->elevator) {
-		WARN_ON(rq->tag < 0);
-		blk_mq_put_driver_tag_hctx(hctx, rq);
-	}
-
 	/*
 	 * After populating an empty queue, kick it to avoid stall.  Read
 	 * the comment in flush_end_io().
@@ -480,7 +463,7 @@ void blk_insert_flush(struct request *rq)
 	if ((policy & REQ_FSEQ_DATA) &&
 	    !(policy & (REQ_FSEQ_PREFLUSH | REQ_FSEQ_POSTFLUSH))) {
 		if (q->mq_ops)
-			blk_mq_request_bypass_insert(rq, false);
+			blk_mq_sched_insert_request(rq, false, true, false, false);
 		else
 			list_add_tail(&rq->queuelist, &q->queue_head);
 		return;
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index 01a43fed6b8c..df8581bb0a37 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -68,25 +68,29 @@ static void blk_mq_sched_mark_restart_hctx(struct blk_mq_hw_ctx *hctx)
 		set_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
 }
 
-void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)
+static bool blk_mq_sched_restart_hctx(struct blk_mq_hw_ctx *hctx)
 {
 	if (!test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
-		return;
+		return false;
+
+	if (hctx->flags & BLK_MQ_F_TAG_SHARED) {
+		struct request_queue *q = hctx->queue;
 
-	clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
+		if (test_and_clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
+			atomic_dec(&q->shared_hctx_restart);
+	} else
+		clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
 
 	if (blk_mq_hctx_has_pending(hctx)) {
 		blk_mq_run_hw_queue(hctx, true);
-		return;
+		return true;
 	}
+
+	return false;
 }
 
-/*
- * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
- * its queue by itself in its completion handler, so we don't need to
- * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
- */
-static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
+/* return true if hctx need to run again */
+static bool blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	struct elevator_queue *e = q->elevator;
@@ -94,18 +98,23 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 
 	do {
 		struct request *rq;
+		blk_status_t ret;
 
 		if (e->type->ops.mq.has_work &&
 				!e->type->ops.mq.has_work(hctx))
 			break;
 
-		if (!blk_mq_get_dispatch_budget(hctx))
-			break;
+		ret = blk_mq_get_dispatch_budget(hctx);
+		if (ret == BLK_STS_RESOURCE)
+			return true;
 
 		rq = e->type->ops.mq.dispatch_request(hctx);
 		if (!rq) {
 			blk_mq_put_dispatch_budget(hctx);
 			break;
+		} else if (ret != BLK_STS_OK) {
+			blk_mq_end_request(rq, ret);
+			continue;
 		}
 
 		/*
@@ -115,6 +124,8 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
 		 */
 		list_add(&rq->queuelist, &rq_list);
 	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));
+
+	return false;
 }
 
 static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
@@ -128,12 +139,8 @@ static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
 	return hctx->ctxs[idx];
 }
 
-/*
- * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
- * its queue by itself in its completion handler, so we don't need to
- * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
- */
-static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
+/* return true if hctx need to run again */
+static bool blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	LIST_HEAD(rq_list);
@@ -141,17 +148,22 @@ static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 
 	do {
 		struct request *rq;
+		blk_status_t ret;
 
 		if (!sbitmap_any_bit_set(&hctx->ctx_map))
 			break;
 
-		if (!blk_mq_get_dispatch_budget(hctx))
-			break;
+		ret = blk_mq_get_dispatch_budget(hctx);
+		if (ret == BLK_STS_RESOURCE)
+			return true;
 
 		rq = blk_mq_dequeue_from_ctx(hctx, ctx);
 		if (!rq) {
 			blk_mq_put_dispatch_budget(hctx);
 			break;
+		} else if (ret != BLK_STS_OK) {
+			blk_mq_end_request(rq, ret);
+			continue;
 		}
 
 		/*
@@ -167,19 +179,22 @@ static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
 	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));
 
 	WRITE_ONCE(hctx->dispatch_from, ctx);
+
+	return false;
 }
 
 /* return true if hw queue need to be run again */
-void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
+bool blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 	struct elevator_queue *e = q->elevator;
 	const bool has_sched_dispatch = e && e->type->ops.mq.dispatch_request;
 	LIST_HEAD(rq_list);
+	bool run_queue = false;
 
 	/* RCU or SRCU read lock is needed before checking quiesced flag */
 	if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q)))
-		return;
+		return false;
 
 	hctx->run++;
 
@@ -211,12 +226,12 @@ void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 		blk_mq_sched_mark_restart_hctx(hctx);
 		if (blk_mq_dispatch_rq_list(q, &rq_list, false)) {
 			if (has_sched_dispatch)
-				blk_mq_do_dispatch_sched(hctx);
+				run_queue = blk_mq_do_dispatch_sched(hctx);
 			else
-				blk_mq_do_dispatch_ctx(hctx);
+				run_queue = blk_mq_do_dispatch_ctx(hctx);
 		}
 	} else if (has_sched_dispatch) {
-		blk_mq_do_dispatch_sched(hctx);
+		run_queue = blk_mq_do_dispatch_sched(hctx);
 	} else if (q->mq_ops->get_budget) {
 		/*
 		 * If we need to get budget before queuing request, we
@@ -226,11 +241,19 @@ void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 		 * TODO: get more budgets, and dequeue more requests in
 		 * one time.
 		 */
-		blk_mq_do_dispatch_ctx(hctx);
+		run_queue = blk_mq_do_dispatch_ctx(hctx);
 	} else {
 		blk_mq_flush_busy_ctxs(hctx, &rq_list);
 		blk_mq_dispatch_rq_list(q, &rq_list, false);
 	}
+
+	if (run_queue && !blk_mq_sched_needs_restart(hctx) &&
+			!test_bit(BLK_MQ_S_TAG_WAITING, &hctx->state)) {
+		blk_mq_sched_mark_restart_hctx(hctx);
+		return true;
+	}
+
+	return false;
 }
 
 bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
@@ -345,21 +368,98 @@ void blk_mq_sched_request_inserted(struct request *rq)
 EXPORT_SYMBOL_GPL(blk_mq_sched_request_inserted);
 
 static bool blk_mq_sched_bypass_insert(struct blk_mq_hw_ctx *hctx,
-				       bool has_sched,
 				       struct request *rq)
 {
-	/* dispatch flush rq directly */
-	if (rq->rq_flags & RQF_FLUSH_SEQ) {
-		spin_lock(&hctx->lock);
-		list_add(&rq->queuelist, &hctx->dispatch);
-		spin_unlock(&hctx->lock);
-		return true;
+	if (rq->tag == -1) {
+		rq->rq_flags |= RQF_SORTED;
+		return false;
 	}
 
-	if (has_sched)
-		rq->rq_flags |= RQF_SORTED;
+	/*
+	 * If we already have a real request tag, send directly to
+	 * the dispatch list.
+	 */
+	spin_lock(&hctx->lock);
+	list_add(&rq->queuelist, &hctx->dispatch);
+	spin_unlock(&hctx->lock);
+	return true;
+}
 
-	return false;
+/**
+ * list_for_each_entry_rcu_rr - iterate in a round-robin fashion over rcu list
+ * @pos:    loop cursor.
+ * @skip:   the list element that will not be examined. Iteration starts at
+ *          @skip->next.
+ * @head:   head of the list to examine. This list must have at least one
+ *          element, namely @skip.
+ * @member: name of the list_head structure within typeof(*pos).
+ */
+#define list_for_each_entry_rcu_rr(pos, skip, head, member)		\
+	for ((pos) = (skip);						\
+	     (pos = (pos)->member.next != (head) ? list_entry_rcu(	\
+			(pos)->member.next, typeof(*pos), member) :	\
+	      list_entry_rcu((pos)->member.next->next, typeof(*pos), member)), \
+	     (pos) != (skip); )
+
+/*
+ * Called after a driver tag has been freed to check whether a hctx needs to
+ * be restarted. Restarts @hctx if its tag set is not shared. Restarts hardware
+ * queues in a round-robin fashion if the tag set of @hctx is shared with other
+ * hardware queues.
+ */
+void blk_mq_sched_restart(struct blk_mq_hw_ctx *const hctx)
+{
+	struct blk_mq_tags *const tags = hctx->tags;
+	struct blk_mq_tag_set *const set = hctx->queue->tag_set;
+	struct request_queue *const queue = hctx->queue, *q;
+	struct blk_mq_hw_ctx *hctx2;
+	unsigned int i, j;
+
+	if (set->flags & BLK_MQ_F_TAG_SHARED) {
+		/*
+		 * If this is 0, then we know that no hardware queues
+		 * have RESTART marked. We're done.
+		 */
+		if (!atomic_read(&queue->shared_hctx_restart))
+			return;
+
+		rcu_read_lock();
+		list_for_each_entry_rcu_rr(q, queue, &set->tag_list,
+					   tag_set_list) {
+			queue_for_each_hw_ctx(q, hctx2, i)
+				if (hctx2->tags == tags &&
+				    blk_mq_sched_restart_hctx(hctx2))
+					goto done;
+		}
+		j = hctx->queue_num + 1;
+		for (i = 0; i < queue->nr_hw_queues; i++, j++) {
+			if (j == queue->nr_hw_queues)
+				j = 0;
+			hctx2 = queue->queue_hw_ctx[j];
+			if (hctx2->tags == tags &&
+			    blk_mq_sched_restart_hctx(hctx2))
+				break;
+		}
+done:
+		rcu_read_unlock();
+	} else {
+		blk_mq_sched_restart_hctx(hctx);
+	}
+}
+
+/*
+ * Add flush/fua to the queue. If we fail getting a driver tag, then
+ * punt to the requeue list. Requeue will re-invoke us from a context
+ * that's safe to block from.
+ */
+static void blk_mq_sched_insert_flush(struct blk_mq_hw_ctx *hctx,
+				      struct request *rq, bool can_block)
+{
+	if (blk_mq_get_driver_tag(rq, &hctx, can_block)) {
+		blk_insert_flush(rq);
+		blk_mq_run_hw_queue(hctx, true);
+	} else
+		blk_mq_add_to_requeue_list(rq, false, true);
 }
 
 void blk_mq_sched_insert_request(struct request *rq, bool at_head,
@@ -370,15 +470,12 @@ void blk_mq_sched_insert_request(struct request *rq, bool at_head,
 	struct blk_mq_ctx *ctx = rq->mq_ctx;
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 
-	/* flush rq in flush machinery need to be dispatched directly */
-	if (!(rq->rq_flags & RQF_FLUSH_SEQ) && op_is_flush(rq->cmd_flags)) {
-		blk_insert_flush(rq);
-		goto run;
+	if (rq->tag == -1 && op_is_flush(rq->cmd_flags)) {
+		blk_mq_sched_insert_flush(hctx, rq, can_block);
+		return;
 	}
 
-	WARN_ON(e && (rq->tag != -1));
-
-	if (blk_mq_sched_bypass_insert(hctx, !!e, rq))
+	if (e && blk_mq_sched_bypass_insert(hctx, rq))
 		goto run;
 
 	if (e && e->type->ops.mq.insert_requests) {
@@ -404,6 +501,23 @@ void blk_mq_sched_insert_requests(struct request_queue *q,
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 	struct elevator_queue *e = hctx->queue->elevator;
 
+	if (e) {
+		struct request *rq, *next;
+
+		/*
+		 * We bypass requests that already have a driver tag assigned,
+		 * which should only be flushes. Flushes are only ever inserted
+		 * as single requests, so we shouldn't ever hit the
+		 * WARN_ON_ONCE() below (but let's handle it just in case).
+		 */
+		list_for_each_entry_safe(rq, next, list, queuelist) {
+			if (WARN_ON_ONCE(rq->tag != -1)) {
+				list_del_init(&rq->queuelist);
+				blk_mq_sched_bypass_insert(hctx, rq);
+			}
+		}
+	}
+
 	if (e && e->type->ops.mq.insert_requests)
 		e->type->ops.mq.insert_requests(hctx, list, false);
 	else
diff --git a/block/blk-mq-sched.h b/block/blk-mq-sched.h
index 9267d0b7c197..2434061cc5b7 100644
--- a/block/blk-mq-sched.h
+++ b/block/blk-mq-sched.h
@@ -22,7 +22,7 @@ void blk_mq_sched_insert_requests(struct request_queue *q,
 				  struct blk_mq_ctx *ctx,
 				  struct list_head *list, bool run_queue_async);
 
-void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx);
+bool blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx);
 
 int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e);
 void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 9b49a0d16907..4fed71fd74c0 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -653,8 +653,6 @@ static void __blk_mq_requeue_request(struct request *rq)
 {
 	struct request_queue *q = rq->q;
 
-	blk_mq_put_driver_tag(rq);
-
 	trace_block_rq_requeue(q, rq);
 	wbt_requeue(q->rq_wb, &rq->issue_stat);
 	blk_mq_sched_requeue_request(rq);
@@ -1004,6 +1002,62 @@ bool blk_mq_get_driver_tag(struct request *rq, struct blk_mq_hw_ctx **hctx,
 	return rq->tag != -1;
 }
 
+static void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
+				    struct request *rq)
+{
+	blk_mq_put_tag(hctx, hctx->tags, rq->mq_ctx, rq->tag);
+	rq->tag = -1;
+
+	if (rq->rq_flags & RQF_MQ_INFLIGHT) {
+		rq->rq_flags &= ~RQF_MQ_INFLIGHT;
+		atomic_dec(&hctx->nr_active);
+	}
+}
+
+static void blk_mq_put_driver_tag_hctx(struct blk_mq_hw_ctx *hctx,
+				       struct request *rq)
+{
+	if (rq->tag == -1 || rq->internal_tag == -1)
+		return;
+
+	__blk_mq_put_driver_tag(hctx, rq);
+}
+
+static void blk_mq_put_driver_tag(struct request *rq)
+{
+	struct blk_mq_hw_ctx *hctx;
+
+	if (rq->tag == -1 || rq->internal_tag == -1)
+		return;
+
+	hctx = blk_mq_map_queue(rq->q, rq->mq_ctx->cpu);
+	__blk_mq_put_driver_tag(hctx, rq);
+}
+
+/*
+ * If we fail getting a driver tag because all the driver tags are already
+ * assigned and on the dispatch list, BUT the first entry does not have a
+ * tag, then we could deadlock. For that case, move entries with assigned
+ * driver tags to the front, leaving the set of tagged requests in the
+ * same order, and the untagged set in the same order.
+ */
+static bool reorder_tags_to_front(struct list_head *list)
+{
+	struct request *rq, *tmp, *first = NULL;
+
+	list_for_each_entry_safe_reverse(rq, tmp, list, queuelist) {
+		if (rq == first)
+			break;
+		if (rq->tag != -1) {
+			list_move(&rq->queuelist, list);
+			if (!first)
+				first = rq;
+		}
+	}
+
+	return first != NULL;
+}
+
 static int blk_mq_dispatch_wake(wait_queue_entry_t *wait, unsigned mode, int flags,
 				void *key)
 {
@@ -1046,7 +1100,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 		bool got_budget)
 {
 	struct blk_mq_hw_ctx *hctx;
-	struct request *rq, *nxt;
+	struct request *rq;
 	int errors, queued;
 
 	if (list_empty(list))
@@ -1064,6 +1118,9 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 
 		rq = list_first_entry(list, struct request, queuelist);
 		if (!blk_mq_get_driver_tag(rq, &hctx, false)) {
+			if (!queued && reorder_tags_to_front(list))
+				continue;
+
 			/*
 			 * The initial allocation attempt failed, so we need to
 			 * rerun the hardware queue when a tag is freed.
@@ -1086,8 +1143,13 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 			}
 		}
 
-		if (!got_budget && !blk_mq_get_dispatch_budget(hctx))
-			break;
+		if (!got_budget) {
+			ret = blk_mq_get_dispatch_budget(hctx);
+			if (ret == BLK_STS_RESOURCE)
+				break;
+			if (ret != BLK_STS_OK)
+				goto fail_rq;
+		}
 
 		list_del_init(&rq->queuelist);
 
@@ -1100,25 +1162,21 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 		if (list_empty(list))
 			bd.last = true;
 		else {
+			struct request *nxt;
+
 			nxt = list_first_entry(list, struct request, queuelist);
 			bd.last = !blk_mq_get_driver_tag(nxt, NULL, false);
 		}
 
 		ret = q->mq_ops->queue_rq(hctx, &bd);
 		if (ret == BLK_STS_RESOURCE) {
-			/*
-			 * If an I/O scheduler has been configured and we got a
-			 * driver tag for the next request already, free it again.
-			 */
-			if (!list_empty(list)) {
-				nxt = list_first_entry(list, struct request, queuelist);
-				blk_mq_put_driver_tag(nxt);
-			}
+			blk_mq_put_driver_tag_hctx(hctx, rq);
 			list_add(&rq->queuelist, list);
 			__blk_mq_requeue_request(rq);
 			break;
 		}
 
+ fail_rq:
 		if (unlikely(ret != BLK_STS_OK)) {
 			errors++;
 			blk_mq_end_request(rq, BLK_STS_IOERR);
@@ -1135,6 +1193,13 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 	 * that is where we will continue on next queue run.
 	 */
 	if (!list_empty(list)) {
+		/*
+		 * If an I/O scheduler has been configured and we got a driver
+		 * tag for the next request already, free it again.
+		 */
+		rq = list_first_entry(list, struct request, queuelist);
+		blk_mq_put_driver_tag(rq);
+
 		spin_lock(&hctx->lock);
 		list_splice_init(list, &hctx->dispatch);
 		spin_unlock(&hctx->lock);
@@ -1170,6 +1235,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,
 static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 {
 	int srcu_idx;
+	bool run_queue;
 
 	/*
 	 * We should be running this queue from one of the CPUs that
@@ -1186,15 +1252,18 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 
 	if (!(hctx->flags & BLK_MQ_F_BLOCKING)) {
 		rcu_read_lock();
-		blk_mq_sched_dispatch_requests(hctx);
+		run_queue = blk_mq_sched_dispatch_requests(hctx);
 		rcu_read_unlock();
 	} else {
 		might_sleep();
 
 		srcu_idx = srcu_read_lock(hctx->queue_rq_srcu);
-		blk_mq_sched_dispatch_requests(hctx);
+		run_queue = blk_mq_sched_dispatch_requests(hctx);
 		srcu_read_unlock(hctx->queue_rq_srcu, srcu_idx);
 	}
+
+	if (run_queue)
+		blk_mq_run_hw_queue(hctx, true);
 }
 
 /*
@@ -1440,7 +1509,7 @@ void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
  * Should only be used carefully, when the caller knows we want to
  * bypass a potential IO scheduler on the target device.
  */
-void blk_mq_request_bypass_insert(struct request *rq, bool run_queue)
+void blk_mq_request_bypass_insert(struct request *rq)
 {
 	struct blk_mq_ctx *ctx = rq->mq_ctx;
 	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(rq->q, ctx->cpu);
@@ -1449,8 +1518,7 @@ void blk_mq_request_bypass_insert(struct request *rq, bool run_queue)
 	list_add_tail(&rq->queuelist, &hctx->dispatch);
 	spin_unlock(&hctx->lock);
 
-	if (run_queue)
-		blk_mq_run_hw_queue(hctx, false);
+	blk_mq_run_hw_queue(hctx, false);
 }
 
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
@@ -1584,10 +1652,12 @@ static void __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 	if (!blk_mq_get_driver_tag(rq, NULL, false))
 		goto insert;
 
-	if (!blk_mq_get_dispatch_budget(hctx)) {
+	ret = blk_mq_get_dispatch_budget(hctx);
+	if (ret == BLK_STS_RESOURCE) {
 		blk_mq_put_driver_tag(rq);
 		goto insert;
-	}
+	} else if (ret != BLK_STS_OK)
+		goto fail_rq;
 
 	new_cookie = request_to_qc_t(hctx, rq);
 
@@ -1605,6 +1675,7 @@ static void __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 		__blk_mq_requeue_request(rq);
 		goto insert;
 	default:
+ fail_rq:
 		*cookie = BLK_QC_T_NONE;
 		blk_mq_end_request(rq, ret);
 		return;
@@ -1678,10 +1749,13 @@ static blk_qc_t blk_mq_make_request(struct request_queue *q, struct bio *bio)
 	if (unlikely(is_flush_fua)) {
 		blk_mq_put_ctx(data.ctx);
 		blk_mq_bio_to_request(rq, bio);
-
-		/* bypass scheduler for flush rq */
-		blk_insert_flush(rq);
-		blk_mq_run_hw_queue(data.hctx, true);
+		if (q->elevator) {
+			blk_mq_sched_insert_request(rq, false, true, true,
+					true);
+		} else {
+			blk_insert_flush(rq);
+			blk_mq_run_hw_queue(data.hctx, true);
+		}
 	} else if (plug && q->nr_hw_queues == 1) {
 		struct request *last = NULL;
 
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 2502f40ccdc0..522b420dedc0 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -2,7 +2,6 @@
 #define INT_BLK_MQ_H
 
 #include "blk-stat.h"
-#include "blk-mq-tag.h"
 
 struct blk_mq_tag_set;
 
@@ -57,7 +56,7 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
  */
 void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
 				bool at_head);
-void blk_mq_request_bypass_insert(struct request *rq, bool run_queue);
+void blk_mq_request_bypass_insert(struct request *rq);
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
 				struct list_head *list);
 
@@ -148,45 +147,14 @@ static inline void blk_mq_put_dispatch_budget(struct blk_mq_hw_ctx *hctx)
 		q->mq_ops->put_budget(hctx);
 }
 
-static inline bool blk_mq_get_dispatch_budget(struct blk_mq_hw_ctx *hctx)
+static inline blk_status_t blk_mq_get_dispatch_budget(
+		struct blk_mq_hw_ctx *hctx)
 {
 	struct request_queue *q = hctx->queue;
 
 	if (q->mq_ops->get_budget)
 		return q->mq_ops->get_budget(hctx);
-	return true;
-}
-
-static inline void __blk_mq_put_driver_tag(struct blk_mq_hw_ctx *hctx,
-					   struct request *rq)
-{
-	blk_mq_put_tag(hctx, hctx->tags, rq->mq_ctx, rq->tag);
-	rq->tag = -1;
-
-	if (rq->rq_flags & RQF_MQ_INFLIGHT) {
-		rq->rq_flags &= ~RQF_MQ_INFLIGHT;
-		atomic_dec(&hctx->nr_active);
-	}
-}
-
-static inline void blk_mq_put_driver_tag_hctx(struct blk_mq_hw_ctx *hctx,
-				       struct request *rq)
-{
-	if (rq->tag == -1 || rq->internal_tag == -1)
-		return;
-
-	__blk_mq_put_driver_tag(hctx, rq);
-}
-
-static inline void blk_mq_put_driver_tag(struct request *rq)
-{
-	struct blk_mq_hw_ctx *hctx;
-
-	if (rq->tag == -1 || rq->internal_tag == -1)
-		return;
-
-	hctx = blk_mq_map_queue(rq->q, rq->mq_ctx->cpu);
-	__blk_mq_put_driver_tag(hctx, rq);
+	return BLK_STS_OK;
 }
 
 #endif
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index c537c4d768c1..5634288f244f 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1940,33 +1940,6 @@ static void scsi_mq_done(struct scsi_cmnd *cmd)
 	blk_mq_complete_request(cmd->request);
 }
 
-static void scsi_mq_put_budget(struct blk_mq_hw_ctx *hctx)
-{
-	struct request_queue *q = hctx->queue;
-	struct scsi_device *sdev = q->queuedata;
-
-	atomic_dec(&sdev->device_busy);
-	put_device(&sdev->sdev_gendev);
-}
-
-static bool scsi_mq_get_budget(struct blk_mq_hw_ctx *hctx)
-{
-	struct request_queue *q = hctx->queue;
-	struct scsi_device *sdev = q->queuedata;
-
-	if (!get_device(&sdev->sdev_gendev))
-		goto out;
-	if (!scsi_dev_queue_ready(q, sdev))
-		goto out_put_device;
-
-	return true;
-
-out_put_device:
-	put_device(&sdev->sdev_gendev);
-out:
-	return false;
-}
-
 static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 			 const struct blk_mq_queue_data *bd)
 {
@@ -1980,11 +1953,16 @@ static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 
 	ret = prep_to_mq(scsi_prep_state_check(sdev, req));
 	if (ret != BLK_STS_OK)
-		goto out_put_budget;
+		goto out;
 
 	ret = BLK_STS_RESOURCE;
+	if (!get_device(&sdev->sdev_gendev))
+		goto out;
+
+	if (!scsi_dev_queue_ready(q, sdev))
+		goto out_put_device;
 	if (!scsi_target_queue_ready(shost, sdev))
-		goto out_put_budget;
+		goto out_dec_device_busy;
 	if (!scsi_host_queue_ready(q, shost, sdev))
 		goto out_dec_target_busy;
 
@@ -2015,12 +1993,15 @@ static blk_status_t scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 	return BLK_STS_OK;
 
 out_dec_host_busy:
-       atomic_dec(&shost->host_busy);
+	atomic_dec(&shost->host_busy);
 out_dec_target_busy:
 	if (scsi_target(sdev)->can_queue > 0)
 		atomic_dec(&scsi_target(sdev)->target_busy);
-out_put_budget:
-	scsi_mq_put_budget(hctx);
+out_dec_device_busy:
+	atomic_dec(&sdev->device_busy);
+out_put_device:
+	put_device(&sdev->sdev_gendev);
+out:
 	switch (ret) {
 	case BLK_STS_OK:
 		break;
@@ -2224,8 +2205,6 @@ struct request_queue *scsi_old_alloc_queue(struct scsi_device *sdev)
 }
 
 static const struct blk_mq_ops scsi_mq_ops = {
-	.get_budget	= scsi_mq_get_budget,
-	.put_budget	= scsi_mq_put_budget,
 	.queue_rq	= scsi_queue_rq,
 	.complete	= scsi_softirq_done,
 	.timeout	= scsi_timeout,
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 674641527da7..f2e3079eecdd 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -92,7 +92,7 @@ struct blk_mq_queue_data {
 
 typedef blk_status_t (queue_rq_fn)(struct blk_mq_hw_ctx *,
 		const struct blk_mq_queue_data *);
-typedef bool (get_budget_fn)(struct blk_mq_hw_ctx *);
+typedef blk_status_t (get_budget_fn)(struct blk_mq_hw_ctx *);
 typedef void (put_budget_fn)(struct blk_mq_hw_ctx *);
 typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
 typedef int (init_hctx_fn)(struct blk_mq_hw_ctx *, void *, unsigned int);
-- 
2.14.3

^ permalink raw reply related	[flat|nested] 5+ messages in thread
* Re: [PATCH] blk-mq: Revert multiple recent patches
  2017-11-07  1:07 ` Bart Van Assche
  (?)
@ 2017-11-07  2:29 ` Ming Lei
  2017-11-07  2:41   ` Ming Lei
  -1 siblings, 1 reply; 5+ messages in thread
From: Ming Lei @ 2017-11-07  2:29 UTC (permalink / raw)
  To: Bart Van Assche
  Cc: Jens Axboe, linux-block, Christoph Hellwig, linux-scsi,
	Martin K . Petersen

On Mon, Nov 06, 2017 at 05:07:31PM -0800, Bart Van Assche wrote:
> Since we are close to the merge window and since it is too late for
> reworking patches and algorithms that have turned out to be flawed,
> revert multiple recent patches. Revert patch "blk-mq: don't handle
> TAG_SHARED in restart" and all patches that conflict with the revert
> of that patch because that patch is based on an incorrect assumption,
> namely that the SCSI starved list can be used as a replacement for
> the blk-mq queue restarting mechanism. Revert patch "scsi: implement
> .get_budget and .put_budget for blk-mq" because it triggers an
> infinite loop on my test setup ("scsi 9:0:0:0: rejecting I/O to dead
> device" + RCU lockup complaint in blk_mq_sched_dispatch_requests()).

The loop can be fixed by 

	88022d7201e9 blk-mq: don't handle failure in .get_budget


-- 
Ming

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] blk-mq: Revert multiple recent patches
  2017-11-07  2:29 ` Ming Lei
@ 2017-11-07  2:41   ` Ming Lei
  0 siblings, 0 replies; 5+ messages in thread
From: Ming Lei @ 2017-11-07  2:41 UTC (permalink / raw)
  To: Bart Van Assche
  Cc: Jens Axboe, linux-block, Christoph Hellwig, linux-scsi,
	Martin K . Petersen

On Tue, Nov 07, 2017 at 10:29:38AM +0800, Ming Lei wrote:
> On Mon, Nov 06, 2017 at 05:07:31PM -0800, Bart Van Assche wrote:
> > Since we are close to the merge window and since it is too late for
> > reworking patches and algorithms that have turned out to be flawed,
> > revert multiple recent patches. Revert patch "blk-mq: don't handle
> > TAG_SHARED in restart" and all patches that conflict with the revert
> > of that patch because that patch is based on an incorrect assumption,
> > namely that the SCSI starved list can be used as a replacement for
> > the blk-mq queue restarting mechanism. Revert patch "scsi: implement
> > .get_budget and .put_budget for blk-mq" because it triggers an
> > infinite loop on my test setup ("scsi 9:0:0:0: rejecting I/O to dead
> > device" + RCU lockup complaint in blk_mq_sched_dispatch_requests()).
> 
> The loop can be fixed by 
> 
> 	88022d7201e9 blk-mq: don't handle failure in .get_budget

BTW, the exact reason is that the rq to be failed isn't deleted from
'list' in blk_mq_dispatch_rq_list().

Trust me, I have make sure your srp-test won't hang/crash in my test.

Even my test covered more, such as the following report:

	https://marc.info/?t=150988393300001&r=1&w=2

which is a long-term issue which should be from introduction of blk-mq.

Thanks,
Ming

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] blk-mq: Revert multiple recent patches
  2017-11-07  1:07 ` Bart Van Assche
  (?)
  (?)
@ 2017-11-07  2:56 ` Ming Lei
  -1 siblings, 0 replies; 5+ messages in thread
From: Ming Lei @ 2017-11-07  2:56 UTC (permalink / raw)
  To: Bart Van Assche
  Cc: Jens Axboe, linux-block, Christoph Hellwig, linux-scsi,
	Martin K . Petersen

On Mon, Nov 06, 2017 at 05:07:31PM -0800, Bart Van Assche wrote:
> Since we are close to the merge window and since it is too late for
> reworking patches and algorithms that have turned out to be flawed,
> revert multiple recent patches. Revert patch "blk-mq: don't handle
> TAG_SHARED in restart" and all patches that conflict with the revert
> of that patch because that patch is based on an incorrect assumption,
> namely that the SCSI starved list can be used as a replacement for
> the blk-mq queue restarting mechanism. Revert patch "scsi: implement

Why is the replacement correct? Please explain it in details.

-- 
Ming

^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-11-07  2:56 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-11-07  1:07 [PATCH] blk-mq: Revert multiple recent patches Bart Van Assche
2017-11-07  1:07 ` Bart Van Assche
2017-11-07  2:29 ` Ming Lei
2017-11-07  2:41   ` Ming Lei
2017-11-07  2:56 ` Ming Lei

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.