Re: [PATCH V2 1/3] blk-mq: grab rq->refcount before calling ->fn in blk_mq_tagset_busy_iter

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

On Mon, Apr 26, 2021 at 07:28:18PM -0700, Bart Van Assche wrote:
> On 4/26/21 6:45 PM, Ming Lei wrote:
> > + * We grab one request reference before calling @fn and release it after
> > + * @fn returns. So far we don't support to pass the request reference to
> > + * one new conetxt in @fn.
>               ^^^^^^^
>               context?
> 
> > +void blk_mq_put_rq_ref(struct request *rq)
> > +{
> > +	if (is_flush_rq(rq, rq->mq_hctx))
> > +		rq->end_io(rq, 0);
> > +	else if (refcount_dec_and_test(&rq->ref))
> > +		__blk_mq_free_request(rq);
> > +}
> 
> Will rq->end_io() be called twice when a tag iteration function
> encounters a flush request?

It could be, see flush_end_io():

static void flush_end_io(struct request *flush_rq, blk_status_t error)
{
		...
        spin_lock_irqsave(&fq->mq_flush_lock, flags);

        if (!refcount_dec_and_test(&flush_rq->ref)) {
                fq->rq_status = error;
                spin_unlock_irqrestore(&fq->mq_flush_lock, flags);
				return;
		}
		...


Thanks,
Ming