[PATCH] block: unfreeze queue if realloc tag set fails during nr_hw_queues update

[PATCH] block: unfreeze queue if realloc tag set fails during nr_hw_queues update

[Nilay Shroff]

In __blk_mq_update_nr_hw_queues(), the current sequence involves:

1. unregistering sysfs/debugfs attributes
2. freeze the queue
3. reallocating the tag set
4. updating the queue map
5. reallocating hardware contexts
6. updating the elevator (which unfreeze the queue again)
7. re-register sysfs/debugfs attributes

If tag set reallocation fails at step 3, the function skips steps 4–6
and proceeds directly to step 7, re-registering the sysfs/debugfs
attributes without unfreezing the queue first. This is incorrect and
can lead to a system hang or lockdep splat, as the queue remains frozen
and is never properly unfrozen.

This patch addresses the issue by explicitly unfreezing the queue before
re-registering the sysfs/debugfs attributes in the event of a tag set
reallocation failure.

Fixes: 9dc7a882ce96 ("block: move hctx debugfs/sysfs registering out of freezing queue")
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
---
 block/blk-mq.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 4f79a9808fd1..cbc9a9f97a31 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -5002,8 +5002,11 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 	list_for_each_entry(q, &set->tag_list, tag_set_list)
 		blk_mq_freeze_queue_nomemsave(q);
 
-	if (blk_mq_realloc_tag_set_tags(set, nr_hw_queues) < 0)
+	if (blk_mq_realloc_tag_set_tags(set, nr_hw_queues) < 0) {
+		list_for_each_entry(q, &set->tag_list, tag_set_list)
+			blk_mq_unfreeze_queue_nomemrestore(q);
 		goto reregister;
+	}
 
 fallback:
 	blk_mq_update_queue_map(set);
-- 
2.49.0

Re: [PATCH] block: unfreeze queue if realloc tag set fails during nr_hw_queues update

[Ming Lei]

On Mon, May 12, 2025 at 02:43:38PM +0530, Nilay Shroff wrote:
> In __blk_mq_update_nr_hw_queues(), the current sequence involves:
> 
> 1. unregistering sysfs/debugfs attributes
> 2. freeze the queue
> 3. reallocating the tag set
> 4. updating the queue map
> 5. reallocating hardware contexts
> 6. updating the elevator (which unfreeze the queue again)
> 7. re-register sysfs/debugfs attributes
> 
> If tag set reallocation fails at step 3, the function skips steps 4–6
> and proceeds directly to step 7, re-registering the sysfs/debugfs
> attributes without unfreezing the queue first. This is incorrect and
> can lead to a system hang or lockdep splat, as the queue remains frozen
> and is never properly unfrozen.
> 
> This patch addresses the issue by explicitly unfreezing the queue before
> re-registering the sysfs/debugfs attributes in the event of a tag set
> reallocation failure.
> 
> Fixes: 9dc7a882ce96 ("block: move hctx debugfs/sysfs registering out of freezing queue")
> Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
> ---
>  block/blk-mq.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 4f79a9808fd1..cbc9a9f97a31 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -5002,8 +5002,11 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
>  	list_for_each_entry(q, &set->tag_list, tag_set_list)
>  		blk_mq_freeze_queue_nomemsave(q);
>  
> -	if (blk_mq_realloc_tag_set_tags(set, nr_hw_queues) < 0)
> +	if (blk_mq_realloc_tag_set_tags(set, nr_hw_queues) < 0) {
> +		list_for_each_entry(q, &set->tag_list, tag_set_list)
> +			blk_mq_unfreeze_queue_nomemrestore(q);
>  		goto reregister;

Reviewed-by: Ming Lei <ming.lei@redhat.com>

Thanks,
Ming

Re: [PATCH] block: unfreeze queue if realloc tag set fails during nr_hw_queues update

[Jens Axboe]

On Mon, 12 May 2025 14:43:38 +0530, Nilay Shroff wrote:
> In __blk_mq_update_nr_hw_queues(), the current sequence involves:
> 
> 1. unregistering sysfs/debugfs attributes
> 2. freeze the queue
> 3. reallocating the tag set
> 4. updating the queue map
> 5. reallocating hardware contexts
> 6. updating the elevator (which unfreeze the queue again)
> 7. re-register sysfs/debugfs attributes
> 
> [...]

Applied, thanks!

[1/1] block: unfreeze queue if realloc tag set fails during nr_hw_queues update
      commit: 2d8951aee844b7c6d146e25b5c3eebc1c72aa6ff

Best regards,
-- 
Jens Axboe