On 4/2/18 4:04 PM, Tejun Heo wrote:
> Request abortion is performed by overriding deadline to now and
> scheduling timeout handling immediately. For the latter part, the
> code was using mod_timer(timeout, 0) which can't guarantee that the
> timer runs afterwards. Let's schedule the underlying work item
> directly instead.
>
> This fixes the hangs during probing reported by Sitsofe but it isn't
> yet clear to me how the failure can happen reliably if it's just the
> above described race condition.
>
> Signed-off-by: Tejun Heo <tj@xxxxxxxxxx>
> Reported-by: Sitsofe Wheeler <sitsofe@xxxxxxxxx>
> Reported-by: Meelis Roos <mroos@xxxxxxxx>
> Fixes: 358f70da49d7 ("blk-mq: make blk_abort_request() trigger timeout path")
> Cc: stable@xxxxxxxxxxxxxxx # v4.16
> Link: http://lkml.kernel.org/r/CALjAwxh-PVYFnYFCJpGOja+m5SzZ8Sa4J7ohxdK=r8NyOF-EMA@xxxxxxxxxxxxxx
> Link: http://lkml.kernel.org/r/alpine.LRH.2.21.1802261049140.4893@xxxxxxxxxx
> ---
> Hello,
>
> I don't have the full explanation yet but here's a preliminary patch.
>
> Thanks.
>
> block/blk-timeout.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/block/blk-timeout.c b/block/blk-timeout.c
> index a05e367..f0e6e41 100644
> --- a/block/blk-timeout.c
> +++ b/block/blk-timeout.c
> @@ -165,7 +165,7 @@ void blk_abort_request(struct request *req)
> * No need for fancy synchronizations.
> */
> blk_rq_set_deadline(req, jiffies);
> - mod_timer(&req->q->timeout, 0);
> + kblockd_schedule_work(&req->q->timeout_work);
> } else {
> if (blk_mark_rq_complete(req))
> return;
In any case, it's cleaner than relying on mod_timer(.., 0). If that
doesn't guarantee that the timer runs again, I can see how a race
with the running timer could prevent us from seeing the timeout
after an abort.
I'll apply this, thanks.
--
Jens Axboe
