On 12/20/21 1:36 PM, Keith Busch wrote:
> On Wed, Dec 15, 2021 at 09:24:18AM -0700, Jens Axboe wrote:
>> + /*
>> + * Peek first request and see if we have a ->queue_rqs() hook.
>> + * If we do, we can dispatch the whole plug list in one go. We
>> + * already know at this point that all requests belong to the
>> + * same queue, caller must ensure that's the case.
>> + *
>> + * Since we pass off the full list to the driver at this point,
>> + * we do not increment the active request count for the queue.
>> + * Bypass shared tags for now because of that.
>> + */
>> + if (q->mq_ops->queue_rqs &&
>> + !(rq->mq_hctx->flags & BLK_MQ_F_TAG_QUEUE_SHARED)) {
>> + blk_mq_run_dispatch_ops(q,
>> + q->mq_ops->queue_rqs(&plug->mq_list));
>
> I think we still need to verify the queue isn't quiesced within
> blk_mq_run_dispatch_ops()'s rcu protected area, prior to calling
> .queue_rqs(). Something like below. Or is this supposed to be the
> low-level drivers responsibility now?
Yes, that seems very reasonable, and I'd much rather do that than punt it
to the driver. Care to send it as a real patch?
--
Jens Axboe
