blk-mq: fix potential deadlock while nr_requests grown

Allocate and free sched_tags while queue is freezed can deadlock[1],
this is a long term problem, hence allocate memory before freezing
queue and free memory after queue is unfreezed.

[1] https://lore.kernel.org/all/0659ea8d-a463-47c8-9180-43c719e106eb@linux.ibm.com/
Fixes: e3a2b3f931f5 ("blk-mq: allow changing of queue depth through sysfs")

Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

3 files changed, 34 insertions, 22 deletions

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 299aac458185..4ccf11cadf8c 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -4926,11 +4926,13 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
 }
 EXPORT_SYMBOL(blk_mq_free_tag_set);
 
-int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
+struct elevator_tags *blk_mq_update_nr_requests(struct request_queue *q,
+						struct elevator_tags *et,
+						unsigned int nr)
 {
 	struct blk_mq_tag_set *set = q->tag_set;
+	struct elevator_tags *old_et = NULL;
 	struct blk_mq_hw_ctx *hctx;
-	int ret = 0;
 	unsigned long i;
 
 	blk_mq_quiesce_queue(q);
@@ -4965,24 +4967,18 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
 		}
 	} else {
 		/* Non-shared sched tags, and tags grow */
-		queue_for_each_hw_ctx(q, hctx, i) {
-			if (!hctx->sched_tags)
-				continue;
-			ret = blk_mq_tag_update_depth(hctx, &hctx->sched_tags,
-						      nr);
-			if (ret)
-				goto out;
-		}
+		queue_for_each_hw_ctx(q, hctx, i)
+			hctx->sched_tags = et->tags[i];
+		old_et =  q->elevator->et;
+		q->elevator->et = et;
 	}
 
 	q->nr_requests = nr;
 	if (q->elevator && q->elevator->type->ops.depth_updated)
 		q->elevator->type->ops.depth_updated(q);
 
-out:
 	blk_mq_unquiesce_queue(q);
-
-	return ret;
+	return old_et;
 }
 
 /*
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 3a1d4c37d1bc..0ef96f510477 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -6,6 +6,7 @@
 #include "blk-stat.h"
 
 struct blk_mq_tag_set;
+struct elevator_tags;
 
 struct blk_mq_ctxs {
 	struct kobject kobj;
@@ -45,7 +46,9 @@ void blk_mq_submit_bio(struct bio *bio);
 int blk_mq_poll(struct request_queue *q, blk_qc_t cookie, struct io_comp_batch *iob,
 		unsigned int flags);
 void blk_mq_exit_queue(struct request_queue *q);
-int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr);
+struct elevator_tags *blk_mq_update_nr_requests(struct request_queue *q,
+						struct elevator_tags *tags,
+						unsigned int nr);
 void blk_mq_wake_waiters(struct request_queue *q);
 bool blk_mq_dispatch_rq_list(struct blk_mq_hw_ctx *hctx, struct list_head *,
 			     bool);
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index ff66d1b47169..76c47fe9b8d6 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -64,11 +64,12 @@ static ssize_t queue_requests_show(struct gendisk *disk, char *page)
 static ssize_t
 queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 {
-	unsigned long nr;
-	int ret, err;
-	unsigned int memflags;
 	struct request_queue *q = disk->queue;
 	struct blk_mq_tag_set *set = q->tag_set;
+	struct elevator_tags *et = NULL;
+	unsigned int memflags;
+	unsigned long nr;
+	int ret;
 
 	ret = queue_var_store(&nr, page, count);
 	if (ret < 0)
@@ -99,16 +100,28 @@ queue_requests_store(struct gendisk *disk, const char *page, size_t count)
 		goto unlock;
 	}
 
+	if (!blk_mq_is_shared_tags(set->flags) && q->elevator &&
+	    nr > q->elevator->et->nr_requests) {
+		/*
+		 * Tags will grow, allocate memory before freezing queue to
+		 * prevent deadlock.
+		 */
+		et = blk_mq_alloc_sched_tags(set, q->nr_hw_queues, nr);
+		if (!et) {
+			ret = -ENOMEM;
+			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;
-
+	et = blk_mq_update_nr_requests(q, et, nr);
 	mutex_unlock(&q->elevator_lock);
 	blk_mq_unfreeze_queue(q, memflags);
 
+	if (et)
+		blk_mq_free_sched_tags(et, set);
+
 unlock:
 	up_write(&set->update_nr_hwq_lock);
 	return ret;