Re: [RFC PATCH 1/1] block: introduce activity based ioprio

[Jens Axboe <axboe@kernel.dk>]

On 1/18/24 12:48 AM, Zhaoyang Huang wrote:
>>> +static enum dd_prio dd_req_ioprio(struct request *rq)
>>> +{
>>> +     enum dd_prio prio;
>>> +     const u8 ioprio_class = dd_rq_ioclass(rq);
>>> +#ifdef CONFIG_ACTIVITY_BASED_IOPRIO
>>> +     struct bio *bio;
>>> +     struct bio_vec bv;
>>> +     struct bvec_iter iter;
>>> +     struct page *page;
>>> +     int gen = 0;
>>> +     int cnt = 0;
>>> +
>>> +     if (req_op(rq) == REQ_OP_READ) {
>>> +             __rq_for_each_bio(bio, rq) {
>>> +                     bio_for_each_bvec(bv, bio, iter) {
>>> +                             page = bv.bv_page;
>>> +                             gen += PageWorkingset(page) ? 1 : 0;
>>> +                             cnt++;
>>> +                     }
>>> +             }
>>> +             prio = (gen >= cnt / 2) ? ioprio_class_to_prio[IOPRIO_CLASS_RT] :
>>> +                     ioprio_class_to_prio[ioprio_class];
>>> +     } else
>>> +             prio = ioprio_class_to_prio[ioprio_class];
>>> +#else
>>> +     prio = ioprio_class_to_prio[ioprio_class];
>>> +#endif
>>> +     return prio;
>>> +}
>>
>> This is pretty awful imho, you're iterating the pages which isn't
>> exactly cheap. There's also a ternary operator (get rid of it), and
>> magic cnt / 2 which isn't even explained.

> ok. The iterating is on purpose here to not enlarge the bio and
> request structure. The magic number would be replaced by an more
> sensible criteria like MULTI_GEN's thrashing tier things.

It's totally backwards, you're adding code that both dips into parts it
very much should not, and attempting to fix things up after the fact.
Again, not passing any judgement on whether this kind of thing makes
sense, but if it did, you'd do it when adding the page to the bio, as I
mentioned elsewhere in the email. There's no need to grow struct bio, it
already has an ioprio field.

-- 
Jens Axboe