Re: [PATCH v3 1/1] block/blk-mq: use atomic_t for quiesce_depth to avoid lock contention on RT

On 2026-02-11 22:39:29 [+0200], Ionut Nechita (Wind River) wrote:
> From: Ionut Nechita <[email protected]>
> 
> In RT kernel (PREEMPT_RT), commit 679b1874eba7 ("block: fix ordering
> between checking QUEUE_FLAG_QUIESCED request adding") causes severe
…

> Suggested-by: Sebastian Andrzej Siewior <[email protected]>
> Fixes: 679b1874eba7 ("block: fix ordering between checking QUEUE_FLAG_QUIESCED request adding")
This is
         6bda857bcbb86 ("block: fix ordering between checking QUEUE_FLAG_QUIESCED request adding")

in my tree. I don't have your hash anywhere. The hash at the top in the
email is also affected.

> Cc: [email protected]
> Signed-off-by: Ionut Nechita <[email protected]>

Otherwise
Reviewed-by: Sebastian Andrzej Siewior <[email protected]>

> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -314,21 +314,18 @@ EXPORT_SYMBOL_GPL(blk_mq_quiesce_queue);
>   */
>  void blk_mq_unquiesce_queue(struct request_queue *q)
>  {
> -	unsigned long flags;
> -	bool run_queue = false;
> +	int depth;
>  
> -	spin_lock_irqsave(&q->queue_lock, flags);
> -	if (WARN_ON_ONCE(q->quiesce_depth <= 0)) {
> -		;
> -	} else if (!--q->quiesce_depth) {
> -		blk_queue_flag_clear(QUEUE_FLAG_QUIESCED, q);
> -		run_queue = true;
> -	}
> -	spin_unlock_irqrestore(&q->queue_lock, flags);
> +	depth = atomic_dec_if_positive(&q->quiesce_depth);
> +	if (WARN_ON_ONCE(depth < 0))
> +		return;

Ah. That is cute.

> -	/* dispatch requests which are inserted during quiescing */
> -	if (run_queue)
> +	if (depth == 0) {
> +		/* Ensure the decrement is visible before running queues */
> +		smp_mb__after_atomic();
> +		/* dispatch requests which are inserted during quiescing */
>  		blk_mq_run_hw_queues(q, true);
> +	}
>  }
>  EXPORT_SYMBOL_GPL(blk_mq_unquiesce_queue);
>  

Sebastian