Re: [RFC PATCH v1.3 01/11] Docs/mm/damon/design: update for DAMOS_QUOTA_NODE_ELIGIBLE_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] Truncation of 64-bit phys_addr_t to 32-bit unsigned long in damos_get_node_eligible_mem_bp leads to incorrect metric calculation on 32-bit systems.
--

commit 569eb38d08bfaf4718faa6440d86cad7ba9478b7
Author: SeongJae Park <sj@kernel.org>

Docs/mm/damon/design: update for DAMOS_QUOTA_NODE_ELIGIBLE_MEM_BP

Commit 9138e27a3bc3 introduced DAMOS_QUOTA_NODE_ELIGIBLE_MEM_BP but forgot
updating the DAMON design document for that. This commit updates the
documentation to include the missing metric.

> diff --git a/Documentation/mm/damon/design.rst b/Documentation/mm/damon/design.rst
> --- a/Documentation/mm/damon/design.rst
> +++ b/Documentation/mm/damon/design.rst
> @@ -686,9 +686,11 @@ mechanism tries to make ``current_value`` of ``target_metric`` be same to
>    (1/10,000).
>  - ``inactive_mem_bp``: Inactive to active + inactive (LRU) memory size ratio in
>    bp (1/10,000).
> +- ``node_eligible_mem_bp``: Scheme target access pattern-eligible memory ratio
> +  of a node in bp (1/10,000).

[Severity: High]
This isn't a bug introduced by this patch, but while reviewing this metric, I
noticed a pre-existing issue in its implementation.

In mm/damon/core.c:damos_get_node_eligible_mem_bp(), 64-bit phys_addr_t byte
counts are cast to 32-bit unsigned long before ratio calculation:

	if (!(unsigned long)total_eligible)
		return 0;

	return mult_frac((unsigned long)node_eligible, 10000,
			(unsigned long)total_eligible);

On 32-bit systems with PAE, if the amount of eligible memory exceeds 4GB,
will this truncation cause the metric to calculate completely incorrect ratios
(e.g., reporting > 10000 basis points) or erroneously return 0 due to modulo
arithmetic?

This could cause the DAMOS quota tuner to adjust quotas wrongly, potentially
resulting in extremely aggressive or completely ineffective memory actions.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260626001644.85825-1-sj@kernel.org?part=1