Re: [PATCH 8/8] mm/damon/core: handle extreme memory state in get_in_active_mem_bp()

[sashiko-bot@kernel.org]

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

Pre-existing issues:
- [High] 32-bit integer overflow in DAMOS metric calculations.
--

commit 4a4658bd5d39473e6ccb07618a88e601987f2acf
Author: SJ Park <sj@kernel.org>

mm/damon/core: handle extreme memory state in get_in_active_mem_bp()

The function damos_get_in_active_mem_bp() uses the sum of active and inactive
memory as a denominator. In extreme environments where both are zero, this
results in a divide by zero error. This patch avoids the issue by changing
the denominator to one if it is zero.

Link: https://lore.kernel.org/20260721034756.147011-1-sj@kernel.org

> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index 3060edf5e4fa6..0df785e72438f 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
> @@ -3006,7 +3006,7 @@ static unsigned int damos_get_in_active_mem_bp(bool active_ratio)
>  		global_node_page_state(NR_LRU_BASE + LRU_ACTIVE_FILE);
>  	inactive = global_node_page_state(NR_LRU_BASE + LRU_INACTIVE_ANON) +
>  		global_node_page_state(NR_LRU_BASE + LRU_INACTIVE_FILE);
> -	total = active + inactive;
> +	total = max(active + inactive, 1);
>  	if (active_ratio)
>  		return mult_frac(active, 10000, total);
>  	return mult_frac(inactive, 10000, total);

[Severity: High]
This isn't a bug introduced by this patch, but it's a pre-existing high
severity issue in the exact function modified.

Could this mult_frac() calculation overflow on 32-bit architectures?

When active or inactive memory exceeds approximately 429,496 pages (about
1.7GB) on a 32-bit system, the intermediate product in mult_frac():
(r * n / d) where r is (active % total) and n is 10000

will overflow the 32-bit unsigned long before division. This wraps around
and causes Damon to apply completely wrong quotas, degrading memory goal
reliability.

Does this same 32-bit overflow issue also affect other metric calculations
in mm/damon/core.c, such as damos_get_node_mem_bp() and
damos_get_node_memcg_used_bp() which similarly compute:
mult_frac(numerator, 10000, i.totalram) ?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260901131850.98037-1-sj@kernel.org?part=8