Re: [PATCH] blk-mq-debugfs: Fix warning about possible deadlock

[Nilay Shroff <nilay@linux.ibm.com>]

On 2/14/26 4:15 AM, Mohamed Khalfella wrote:
> Commit 65d466b62984 ("blk-mq-debugfs: warn about possible deadlock")
> added WARN_ONCE_ON() to debugfs_create_files() to prevent potential
> deadlock if the queue is frozen. We hit this warning with the stacktrace
> below
> 
> WARNING: block/blk-mq-debugfs.c:620 at debugfs_create_files+0x9c/0x1d0
> Workqueue: nvme-wq nvme_tcp_reconnect_ctrl_work [nvme_tcp]
> RIP: 0010:debugfs_create_files+0x9c/0x1d0
> Call Trace:
>  <TASK>
>  blk_mq_debugfs_register_hctx+0x186/0x320
>  blk_mq_debugfs_register_hctxs+0x80/0xa0
>  blk_mq_update_nr_hw_queues+0xc8f/0xd50
>  nvme_tcp_setup_ctrl+0xcfa/0xda0 [nvme_tcp]
>  nvme_tcp_reconnect_ctrl_work+0x51/0x180 [nvme_tcp]
>  process_scheduled_works+0x840/0xd80
>  worker_thread+0x36d/0x4a0
>  kthread+0x366/0x380
>  ret_from_fork+0x7b/0x670
>  ret_from_fork_asm+0x1a/0x30
>  </TASK>
> 
> Here __blk_mq_update_nr_hw_queues() did set PF_MEMALLOC_NOIO flag before
> calling blk_mq_debugfs_register_hctxs() and this should guarantee the
> described deadlock will not happen. Warn only if the queue is frozen and
> NOIO flag is not set.
> 
> Fixes: 65d466b62984 ("blk-mq-debugfs: warn about possible deadlock")
> Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com>
> ---
>  block/blk-mq-debugfs.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
> index faeaa1fc86a7..ec6670dc0a3d 100644
> --- a/block/blk-mq-debugfs.c
> +++ b/block/blk-mq-debugfs.c
> @@ -612,12 +612,14 @@ static void debugfs_create_files(struct request_queue *q, struct dentry *parent,
>  				 void *data,
>  				 const struct blk_mq_debugfs_attr *attr)
>  {
> +	unsigned int pflags = READ_ONCE(current->flags);
> +
>  	lockdep_assert_held(&q->debugfs_mutex);
>  	/*
>  	 * Creating new debugfs entries with queue freezed has the risk of
>  	 * deadlock.
>  	 */
> -	WARN_ON_ONCE(q->mq_freeze_depth != 0);
> +	WARN_ON_ONCE((q->mq_freeze_depth != 0) && !(pflags & PF_MEMALLOC_NOIO));
>  	/*
>  	 * debugfs_mutex should not be nested under other locks that can be
>  	 * grabbed while queue is frozen.
> 
> base-commit: cd7a5651db263b5384aef1950898e5e889425134

This issue is already being addressed here:
https://lore.kernel.org/all/20260214054350.2322436-1-yukuai@fnnas.com/

Thanks,
--Nilay