blk-mq: convert to serialize updating nr_requests with update_nr_hwq_lock

[ Upstream commit 626ff4f8ebcb7207f01e7810acb85812ccf06bd8 ]

request_queue->nr_requests can be changed by:

a) switch elevator by updating nr_hw_queues
b) switch elevator by elevator sysfs attribute
c) configue queue sysfs attribute nr_requests

Current lock order is:

1) update_nr_hwq_lock, case a,b
2) freeze_queue
3) elevator_lock, case a,b,c

And update nr_requests is seriablized by elevator_lock() already,
however, in the case c, we'll have to allocate new sched_tags if
nr_requests grow, and do this with elevator_lock held and queue
freezed has the risk of deadlock.

Hence use update_nr_hwq_lock instead, make it possible to allocate
memory if tags grow, meanwhile also prevent nr_requests to be changed
concurrently.

Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Stable-dep-of: b86433721f46 ("blk-mq: fix potential deadlock while nr_requests grown")
Signed-off-by: Sasha Levin <sashal@kernel.org>

1 files changed, 20 insertions, 5 deletions

diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index b61e956a868e..163264e4ec62 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -68,6 +68,7 @@ queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 	int ret, err;
 	unsigned int memflags;
 	struct request_queue *q = disk->queue;
+	struct blk_mq_tag_set *set = q->tag_set;
 
 	if (!queue_is_mq(q))
 		return -EINVAL;
@@ -76,8 +77,11 @@ queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 	if (ret < 0)
 		return ret;
 
-	memflags = blk_mq_freeze_queue(q);
-	mutex_lock(&q->elevator_lock);
+	/*
+	 * Serialize updating nr_requests with concurrent queue_requests_store()
+	 * and switching elevator.
+	 */
+	down_write(&set->update_nr_hwq_lock);
 
 	if (nr == q->nr_requests)
 		goto unlock;
@@ -85,20 +89,31 @@ queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 	if (nr < BLKDEV_MIN_RQ)
 		nr = BLKDEV_MIN_RQ;
 
-	if (nr <= q->tag_set->reserved_tags ||
+	/*
+	 * Switching elevator is protected by update_nr_hwq_lock:
+	 *  - read lock is held from elevator sysfs attribute;
+	 *  - write lock is held from updating nr_hw_queues;
+	 * Hence it's safe to access q->elevator here with write lock held.
+	 */
+	if (nr <= set->reserved_tags ||
 	    (q->elevator && nr > MAX_SCHED_RQ) ||
-	    (!q->elevator && nr > q->tag_set->queue_depth)) {
+	    (!q->elevator && nr > set->queue_depth)) {
 		ret = -EINVAL;
 		goto unlock;
 	}
 
+	memflags = blk_mq_freeze_queue(q);
+	mutex_lock(&q->elevator_lock);
+
 	err = blk_mq_update_nr_requests(disk->queue, nr);
 	if (err)
 		ret = err;
 
-unlock:
 	mutex_unlock(&q->elevator_lock);
 	blk_mq_unfreeze_queue(q, memflags);
+
+unlock:
+	up_write(&set->update_nr_hwq_lock);
 	return ret;
 }