Re: [PATCH] block: fix missing queue put in error path

[Ming Lei <ming.lei@redhat.com>]

On Mon, Nov 15, 2021 at 02:35:09PM -0700, Jens Axboe wrote:
> If we fail the submission queue checks, we don't put the queue afterwards.
> This can cause various issues like stalls on scheduler switch or failure
> to remove the device, or like in the original bug report, timeout waiting
> for the device on reboot/restart.
> 
> While in there, fix a few whitespace discrepancies in the surrounding
> code.
> 
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=215039
> Fixes: b637108a4022 ("blk-mq: fix filesystem I/O request allocation")
> Reported-and-tested-by: Stephen Smith <stephenmsmith@blueyonder.co.uk>
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> 
> ---
> 
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 3ab34c4f20da..5e1c9fd99353 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -2543,8 +2543,7 @@ static struct request *blk_mq_get_new_requests(struct request_queue *q,
>  	return NULL;
>  }
>  
> -static inline bool blk_mq_can_use_cached_rq(struct request *rq,
> -		struct bio *bio)
> +static inline bool blk_mq_can_use_cached_rq(struct request *rq, struct bio *bio)
>  {
>  	if (blk_mq_get_hctx_type(bio->bi_opf) != rq->mq_hctx->type)
>  		return false;
> @@ -2565,7 +2564,6 @@ static inline struct request *blk_mq_get_request(struct request_queue *q,
>  	bool checked = false;
>  
>  	if (plug) {
> -
>  		rq = rq_list_peek(&plug->cached_rq);
>  		if (rq && rq->q == q) {
>  			if (unlikely(!submit_bio_checks(bio)))
> @@ -2587,12 +2585,14 @@ static inline struct request *blk_mq_get_request(struct request_queue *q,
>  fallback:
>  	if (unlikely(bio_queue_enter(bio)))
>  		return NULL;
> -	if (!checked && !submit_bio_checks(bio))
> -		return NULL;
> +	if (unlikely(!checked && !submit_bio_checks(bio)))
> +		goto out_put;
>  	rq = blk_mq_get_new_requests(q, plug, bio, nsegs, same_queue_rq);
> -	if (!rq)
> -		blk_queue_exit(q);
> -	return rq;
> +	if (rq)
> +		return rq;
> +out_put:
> +	blk_queue_exit(q);
> +	return NULL;

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

-- 
Ming