Re: [PATCH v5 1/1] block/blk-mq: use atomic_t for quiesce_depth to avoid lock contention on RT

On 3/3/26 8:37 AM, Ionut Nechita (Wind River) wrote:
> +	/*
> +	 * Ensure the store to quiesce_depth is visible before any
> +	 * subsequent loads in blk_mq_run_hw_queue().
> +	 */
This comment does not make sense to me. How could there be a dependency
of blk_mq_run_hw_queue() on *loads* that happen in 
blk_mq_quiesce_queue_nowait() after atomic_inc(&q->quiesce_depth)?

Bart.