Re: [PATCH v3 2/7] block: Send requeued requests to the I/O scheduler

[Christoph Hellwig <hch@lst.de>]

On Wed, May 24, 2023 at 11:22:00AM -0700, Bart Van Assche wrote:
>>>   static inline void blk_mq_sched_requeue_request(struct request *rq)
>>>   {
>>> -	if (rq->rq_flags & RQF_USE_SCHED) {
>>> -		struct request_queue *q = rq->q;
>>> -		struct elevator_queue *e = q->elevator;
>>> -
>>> -		if (e->type->ops.requeue_request)
>>> -			e->type->ops.requeue_request(rq);
>>> -	}
>>> +	if (rq->rq_flags & RQF_USE_SCHED)
>>> +		rq->rq_flags |= RQF_REQUEUED;
>>>   }
>>
>> I'd drop this helper function if we go down this way.  But maybe
>> we might just want to keep the method.
>
> My understanding is that every .requeue_request() call is followed by a 
> .insert_requests() call and hence that we don't need the .requeue_request() 
> method anymore if the RQF_REQUEUED flag would be introduced?

Yes, but at the same time RQF_REQUEUED no creates global state instead
of just in a callchain.  That's why originally suggest a flag to
->insert_requests instead of leaving state on every request.

>
> Thanks,
>
> Bart.
---end quoted text---