On Thu, Nov 04, 2021 at 12:37:25PM -0600, Jens Axboe wrote:
> On 11/4/21 12:36 PM, Christoph Hellwig wrote:
> >> +static inline bool blk_mq_queue_enter(struct request_queue *q, struct bio *bio)
> >> +{
> >> + if (!blk_try_enter_queue(q, false) && bio_queue_enter(bio))
> >> + return false;
> >> + return true;
> >> +}
> >
> > Didn't we just agree on splitting bio_queue_enter into an inline helper
> > and an out of line slowpath instead?
>
> See cover letter, and I also added to the commit message of this one. I do
> think this approach is better, as bio_queue_enter() itself is just slow
> path and there's no point polluting the code with 90% of what's in there.
>
> Hence I kept it as-is.
Well, let me reword this then: why do you think the above is
blk-mq secific and should not be used by every other caller of
bio_queue_enter as well? In other words, why not rename
bio_queue_enter __bio_queue_enter and make the above the public
bio_queue_enter interface then?
