Re: [RFC PATCH v1.3 04/18] mm/damon/core: handle unreset nr_accesses in damon_nr_accesses_mvsum()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] damon_nr_accesses_mvsum() incorrectly returns stale access data from two aggregation intervals ago during the aggregation reset gap.
--

commit 079525184ff6b3e5e3e7c3387647af444428cb84
Author: SeongJae Park <sj@kernel.org>

mm/damon/core: handle unreset nr_accesses in damon_nr_accesses_mvsum()

This commit addresses an issue in damon_nr_accesses_mvsum() where unreset
nr_accesses can unnecessarily inflate the pseudo moving sum. It adds a check
to return last_nr_accesses directly if the remaining window is full.

> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index bc40a6d10f277..8a7351b4a27c2 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
> @@ -268,6 +268,9 @@ unsigned int damon_nr_accesses_mvsum(struct damon_region *r,
>  			ctx->passed_sample_intervals;
>  	left_window_bp = mult_frac(left_window, 10000, window_len);
>  
> +	if (left_window_bp == 10000)
> +		return r->last_nr_accesses;
> +

[Severity: Medium]
Will this regression return stale access data from two aggregation intervals
ago?

If damon_set_attrs() is executed at the end of an aggregation window, it
advances ctx->next_aggregation_sis but intentionally does not reset
r->nr_accesses.

During this unreset gap, r->nr_accesses correctly retains the fully-aggregated
data of the just-finished window, while r->last_nr_accesses holds the value
from two windows ago.

Because next_aggregation_sis was advanced, left_window_bp evaluates to 10000
in damon_nr_accesses_mvsum(). Does the new check fall back to the
stale last_nr_accesses, skipping the most recent interval's data?

>  	return damon_mvsum(r->nr_accesses, r->last_nr_accesses,
>  			left_window_bp);
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260622142139.30269-1-sj@kernel.org?part=4