Re: [RFC PATCH v1.3 09/19] mm/damon/core: disallow probe_hits overflow on attrs only monitoring

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Missing validation in `damon_valid_probe_params()` fails to prevent a 32-bit integer overflow of the weighted sum of probe hits.
--

--- Patch [9]: [RFC PATCH v1.3 09/19] mm/damon/core: disallow probe_hits overflow on attrs only monitoring ---
commit c184df8a3c8632af7a06675128fa4e375050fa8c
Author: SJ Park <sj@kernel.org>
Subject: mm/damon/core: disallow probe_hits overflow on attrs only monitoring

This commit adds validation logic to prevent an 8-bit overflow of
probe_hits when the aggregation interval is much larger than the
sampling interval. It checks the parameters in a new function,
damon_valid_probe_params().

> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index 7d49420ea26c4..4f1425e56950b 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
[ ... ]
> @@ -1332,6 +1332,19 @@ static void damos_set_filters_default_reject(struct damos *s)
>  		damos_filters_default_reject(&s->ops_filters);
>  }
>  
> +static bool damon_valid_probe_params(struct damon_ctx *ctx)
> +{
> +	unsigned long sample_interval;
> +
> +	if (!damon_has_probe_weights(ctx))
> +		return true;
> +
> +	sample_interval = ctx->attrs.sample_interval ? : 1;
> +	if (ctx->attrs.aggr_interval / sample_interval > U8_MAX)
> +		return false;
> +	return true;
> +}

[Severity: Medium]
Since we are adding validation for the 8-bit probe_hits overflow here,
should we also validate that the 32-bit weighted sum of probe hits doesn't
overflow?

If a DAMON context is configured with probes where probe->weight is large
enough, probe->weight * max_probe_hits (up to 255) could exceed UINT_MAX, or
their cumulative sum could overflow. This would result in an incorrect score
being returned by damon_probe_hits_wsum(), causing improper region merges in
damon_merge_regions_of() and degrading monitoring accuracy.

While damon_has_probe_weights() currently returns false making this path
unreachable, would it be safer to add the weighted sum overflow check now to
prevent issues when probe weights are fully implemented?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260710002349.111414-1-sj@kernel.org?part=9