Re: [RFC PATCH v1.3 2/7] mm/damon/core: s/damon_max_nr_accesses()/damon_nr_samples_per_aggr()/

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Arithmetic overflow in DAMON access frequency calculations for large sample counts.
--

--- Patch [2]: [RFC PATCH v1.3 2/7] mm/damon/core: s/damon_max_nr_accesses()/damon_nr_samples_per_aggr()/ ---
commit f4e38af8a3753be681be346a7474a71c5bfadd23
Author: SJ Park <sj@kernel.org>

mm/damon/core: s/damon_max_nr_accesses()/damon_nr_samples_per_aggr()/

This commit renames damon_max_nr_accesses() to damon_nr_samples_per_aggr()
to better describe that it returns the number of samples DAMON checks for
each region per aggregation interval.

> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index a15005db82899..177e5abc981ef 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c

[ ... ]

> @@ -880,14 +880,14 @@ static unsigned int damon_age_for_new_attrs(unsigned int age,
>  static unsigned int damon_accesses_bp_to_nr_accesses(
>  		unsigned int accesses_bp, struct damon_attrs *attrs)
>  {
> -	return accesses_bp * damon_max_nr_accesses(attrs) / 10000;
> +	return accesses_bp * damon_nr_samples_per_aggr(attrs) / 10000;
>  }

[Severity: High]
This is a pre-existing issue, but does multiplying accesses_bp by
damon_nr_samples_per_aggr() overflow 32-bit arithmetic?

If the total samples exceed ~429,496 (e.g., by configuring a large ratio
between the aggregation interval and sample interval), this calculation
seems to wrap around, leading to a drastically lower access count threshold.

>  /* Convert nr_accesses to access ratio in bp (per 10,000) */
>  static unsigned int damon_nr_accesses_to_accesses_bp(
>  		unsigned int nr_accesses, struct damon_attrs *attrs)
>  {
> -	return mult_frac(nr_accesses, 10000, damon_max_nr_accesses(attrs));
> +	return mult_frac(nr_accesses, 10000, damon_nr_samples_per_aggr(attrs));
>  }

[Severity: High]
This isn't a bug introduced by this patch, but can this use of mult_frac()
overflow?

The mult_frac macro internally multiplies the remainder of nr_accesses by
10000, which appears to overflow 32-bit math for large interval
configurations.

> diff --git a/mm/damon/lru_sort.c b/mm/damon/lru_sort.c
> index e8c389ad3226f..7e077084cb03a 100644
> --- a/mm/damon/lru_sort.c
> +++ b/mm/damon/lru_sort.c

[ ... ]

> @@ -303,7 +303,7 @@ static int damon_lru_sort_apply_parameters(void)
>  		goto out;
>  
>  	err = -ENOMEM;
> -	hot_thres = damon_max_nr_accesses(&attrs) *
> +	hot_thres = damon_nr_samples_per_aggr(&attrs) *
>  		hot_thres_access_freq / 1000;

[Severity: High]
This is a pre-existing issue, but does this calculation overflow hot_thres?

Multiplication of the samples by hot_thres_access_freq (typically hundreds)
looks like it overflows a 32-bit integer when calculating the hot threshold
if samples exceed ~8.5 million.

> diff --git a/mm/damon/ops-common.c b/mm/damon/ops-common.c
> index 6bdd1cfd3863a..89c694f6e4fa8 100644
> --- a/mm/damon/ops-common.c
> +++ b/mm/damon/ops-common.c

[ ... ]

> @@ -112,7 +112,7 @@ int damon_hot_score(struct damon_ctx *c, struct damon_region *r,
>  	int hotness;
>  
>  	freq_subscore = r->nr_accesses * DAMON_MAX_SUBSCORE /
> -		damon_max_nr_accesses(&c->attrs);
> +		damon_nr_samples_per_aggr(&c->attrs);

[Severity: High]
This isn't a bug introduced by this patch, but could this multiplication
cause an integer overflow?

The r->nr_accesses variable is multiplied by DAMON_MAX_SUBSCORE, which is
100. If nr_accesses exceeds ~42.9 million, it overflows a 32-bit integer,
resulting in a drastically lower or zeroed hotness score. This could lead to
miscalculating region hotness and erroneously reclaiming actively used pages.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260708024855.116614-1-sj@kernel.org?part=2