block: break pcpu_alloc_mutex dependency on freeze_lock

While nr_hw_update allocates tagset tags it acquires ->pcpu_alloc_mutex
after ->freeze_lock is acquired or queue is frozen. This potentially
creates a circular dependency involving ->fs_reclaim if reclaim is
triggered simultaneously in a code path which first acquires ->pcpu_
alloc_mutex. As the queue is already frozen while nr_hw_queue update
allocates tagsets, the reclaim can't forward progress and thus it could
cause a potential deadlock as reported in lockdep splat[1].

Fix this by pre-allocating tagset tags before we freeze queue during
nr_hw_queue update. Later the allocated tagset tags could be safely
installed and used after queue is frozen.

Reported-by: Yi Zhang <[email protected]>
Closes: https://lore.kernel.org/all/CAHj4cs8F=OV9s3La2kEQ34YndgfZP-B5PHS4Z8_b9euKG6J4mw@mail.gmail.com/ [1]
Signed-off-by: Nilay Shroff <[email protected]>
Reviewed-by: Ming Lei <[email protected]>
Tested-by: Yi Zhang <[email protected]>
Reviewed-by: Yu Kuai <[email protected]>
[axboe: fix brace style issue]
Signed-off-by: Jens Axboe <[email protected]>

Author: shroffni

block/blk-mq.c

@@ -4793,38 +4793,45 @@ static void blk_mq_update_queue_map(struct blk_mq_tag_set *set)
 	}
 }
 
-static int blk_mq_realloc_tag_set_tags(struct blk_mq_tag_set *set,
-				       int new_nr_hw_queues)
+static struct blk_mq_tags **blk_mq_prealloc_tag_set_tags(
+				struct blk_mq_tag_set *set,
+				int new_nr_hw_queues)
 {
 	struct blk_mq_tags **new_tags;
 	int i;
 
 	if (set->nr_hw_queues >= new_nr_hw_queues)
-		goto done;
+		return NULL;
 
 	new_tags = kcalloc_node(new_nr_hw_queues, sizeof(struct blk_mq_tags *),
 				GFP_KERNEL, set->numa_node);
 	if (!new_tags)
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 
 	if (set->tags)
 		memcpy(new_tags, set->tags, set->nr_hw_queues *
 		       sizeof(*set->tags));
-	kfree(set->tags);
-	set->tags = new_tags;
 
 	for (i = set->nr_hw_queues; i < new_nr_hw_queues; i++) {
-		if (!__blk_mq_alloc_map_and_rqs(set, i)) {
-			while (--i >= set->nr_hw_queues)
-				__blk_mq_free_map_and_rqs(set, i);
-			return -ENOMEM;
+		if (blk_mq_is_shared_tags(set->flags)) {
+			new_tags[i] = set->shared_tags;
+		} else {
+			new_tags[i] = blk_mq_alloc_map_and_rqs(set, i,
+					set->queue_depth);
+			if (!new_tags[i])
+				goto out_unwind;
 		}
 		cond_resched();
 	}
 
-done:
-	set->nr_hw_queues = new_nr_hw_queues;
-	return 0;
+	return new_tags;
+out_unwind:
+	while (--i >= set->nr_hw_queues) {
+		if (!blk_mq_is_shared_tags(set->flags))
+			blk_mq_free_map_and_rqs(set, new_tags[i], i);
+	}
+	kfree(new_tags);
+	return ERR_PTR(-ENOMEM);
 }
 
 /*
@@ -5113,6 +5120,7 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 	unsigned int memflags;
 	int i;
 	struct xarray elv_tbl;
+	struct blk_mq_tags **new_tags;
 	bool queues_frozen = false;
 
 	lockdep_assert_held(&set->tag_list_lock);
@@ -5147,11 +5155,18 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 		if (blk_mq_elv_switch_none(q, &elv_tbl))
 			goto switch_back;
 
+	new_tags = blk_mq_prealloc_tag_set_tags(set, nr_hw_queues);
+	if (IS_ERR(new_tags))
+		goto switch_back;
+
 	list_for_each_entry(q, &set->tag_list, tag_set_list)
 		blk_mq_freeze_queue_nomemsave(q);
 	queues_frozen = true;
-	if (blk_mq_realloc_tag_set_tags(set, nr_hw_queues) < 0)
-		goto switch_back;
+	if (new_tags) {
+		kfree(set->tags);
+		set->tags = new_tags;
+	}
+	set->nr_hw_queues = nr_hw_queues;
 
 fallback:
 	blk_mq_update_queue_map(set);