Re: [PATCH 01/18] mm/damon/core: introduce damon_nr_accesses_mvsum()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Calculation of `left_window_bp` and moving sums in `damon_nr_accesses_mvsum()` and `damon_mvsum()` suffers from an integer overflow on 32-bit architectures when `window_len > 429,496`.
--

commit 55e8315faa6caec232a35b71d630c4dd8cbb568f
Author: SJ Park <sj@kernel.org>

mm/damon/core: introduce damon_nr_accesses_mvsum()

Introduce a new DAMON core function, damon_nr_accesses_mvsum(). It returns a
pseudo moving sum value of a given region's nr_accesses for the last
aggregation interval.

> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index 027250e43c66f..fe4c470726d0b 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
> @@ -208,6 +208,68 @@ static struct damon_probe *damon_nth_probe(int n, struct damon_ctx *ctx)
>  	return NULL;
>  }
>  
> +/*
> + * damon_mvsum() - Returns pseudo moving sum value for a time window.

[ ... ]

> +static unsigned long damon_mvsum(unsigned long current_nr,
> +		unsigned long last_nr, unsigned long left_window_bp)
> +{
> +	return current_nr + mult_frac(last_nr, left_window_bp, 10000);

[Severity: Medium]
Could this mult_frac() calculation suffer from an integer overflow on 32-bit
architectures?

If a user configures DAMON with a large aggr_interval and a small
sample_interval such that window_len exceeds 429,496, last_nr could be
equally large. Since unsigned long is 32 bits on those systems, can the
internal multiplication inside mult_frac() silently overflow and result in
incorrect moving sums?

> +}
> +
> +/**
> + * damon_nr_accesses_mvsum() - Returns moving sum access frequency score.

[ ... ]

> +unsigned int damon_nr_accesses_mvsum(struct damon_region *r,
> +		struct damon_ctx *ctx)
> +{
> +	unsigned long sample_interval, aggr_interval;
> +	unsigned long window_len, left_window, left_window_bp;
> +
> +	sample_interval = ctx->attrs.sample_interval ? : 1;
> +	aggr_interval = ctx->attrs.aggr_interval ? : 1;
> +	window_len = aggr_interval / sample_interval;
> +	if (time_after_eq(ctx->passed_sample_intervals,
> +				ctx->next_aggregation_sis))
> +		left_window = 0;
> +	else
> +		left_window = ctx->next_aggregation_sis -
> +			ctx->passed_sample_intervals;
> +	left_window_bp = mult_frac(left_window, 10000, window_len);

[Severity: Medium]
Does this calculation also overflow on 32-bit systems?

The mult_frac() macro calculates (left_window % window_len) * 10000. If
left_window is greater than 429,496, multiplying by 10000 will exceed the
maximum 32-bit unsigned long limit (4,294,967,295), causing a truncated
left_window_bp result.

Should these calculations use 64-bit arithmetic, such as div64_ul(), to
safely support large intervals on 32-bit systems?

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

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260630040812.149729-1-sj@kernel.org?part=1