Re: [PATCHv7 1/1] block: introduce content activity based ioprio

[Matthew Wilcox]

On Mon, Feb 05, 2024 at 01:57:05PM +0800, zhaoyang.huang wrote:
> +/*
> + * bio_set_active_ioprio() is helper function for fs to adjust the bio's ioprio via
> + * calculating the content's activity which measured from MGLRU.
> + * The file system should call this function before submit_bio for the buffered
> + * read/write/sync.
> + */
> +#ifdef CONFIG_BLK_CONT_ACT_BASED_IOPRIO
> +void bio_set_active_ioprio(struct bio *bio)
> +{
> +	struct bio_vec bv;
> +	struct bvec_iter iter;
> +	struct page *page;
> +	int class, level, hint;
> +	int activity = 0;
> +	int cnt = 0;
> +
> +	class = IOPRIO_PRIO_CLASS(bio->bi_ioprio);
> +	level = IOPRIO_PRIO_LEVEL(bio->bi_ioprio);
> +	hint = IOPRIO_PRIO_HINT(bio->bi_ioprio);
> +	/*apply legacy ioprio policy on RT task*/
> +	if (task_is_realtime(current)) {
> +		bio->bi_ioprio = IOPRIO_PRIO_VALUE_HINT(IOPRIO_CLASS_RT, level, hint);
> +		return;
> +	}
> +	bio_for_each_bvec(bv, bio, iter) {
> +		page = bv.bv_page;

I gave you the prototype:

: No, stop this.  What the filesystem needs to do is not
: s/bio_add_folio/act_bio_add_folio/.  There needs to be an API to set the
: bio prio; something like:
: 
:         bio_set_active_prio(bio, folio);

Do not iterate over the bio.  Use the folio provided to set the prio.

Or is there some reason this doesn't work?