Re: [bug report] kernel BUG at block/blk-mq.c:1042!

[Yi Zhang <yi.zhang@redhat.com>]

On Wed, Oct 20, 2021 at 10:21 PM Jens Axboe <axboe@kernel.dk> wrote:
>
> On 10/19/21 8:29 PM, Yi Zhang wrote:
> > Hello
> >
> > Below issue was triggered with blktests srp/ tests during CKI run on
> > linux-block/for-next, seems it was introduced from [2], and still can
> > be reproduced with [1]
> >
> > [1] 4ff840e57c84 Merge branch 'for-5.16/drivers' into for-next [2]
> > 59d62b58f120 Merge branch 'for-5.16/block' into for-next
>
> Can you see if this helps? I don't think that check is valid anymore,
> there are plenty of cases where queuelist may not be valid just yet,
> but it doesn't mean that it's an issue for requeue. The check is
> meant to catch someone doing requeue when the request is still
> inserted.
>

Yes, the issue was fixed with this patch, feel free to add:
Tested-by: Yi Zhang <yi.zhang@redhat.com>

> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 8b05a8f9bb33..a71aeed7b987 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -1053,7 +1053,6 @@ void blk_mq_requeue_request(struct request *rq, bool kick_requeue_list)
>         /* this request will be re-inserted to io scheduler queue */
>         blk_mq_sched_requeue_request(rq);
>
> -       BUG_ON(!list_empty(&rq->queuelist));
>         blk_mq_add_to_requeue_list(rq, true, kick_requeue_list);
>  }
>  EXPORT_SYMBOL(blk_mq_requeue_request);
>
> --
> Jens Axboe
>


-- 
Best Regards,
  Yi Zhang