Re: [PATCH v3 1/2] mm/damon/lru_sort: validate min_region_size to be power of 2

[Liew Rui Yan <aethernet65535@gmail.com>]

Hi Quanmin,

On Fri, 3 Apr 2026 16:59:38 +0800 Quanmin Yan <yanquanmin1@huawei.com> wrote:
> [...]
> 
> I'm a little confused, what does "causing kdamond to terminate
> unexpectedly" mean? The damon_lru_sort_apply_parameters function will
> eventually call damon_commit_ctx, and the power-of-2 check is always
> performed. Is the early check here to prevent some more broken case
> or am I missing something?

terminate unexpectedly means - termination not initiated by the user.
(e.g., not by echo N > enabled)

The issue is not about whether the check exists, but about when it
happens.

In damon_commit_ctx():

    dst->maybe_corrupted = true;
    if (!is_power_of_2(src->min_region_sz))
        return -EINVAL;

Even though -EINVAL is returned, the 'maybe_corrupted' flag has already
been set. When kdamond sees this flag, it terminates.

My patch prevents this by rejecting invalid 'min_region_sz' before
damon_commit_ctx() is called, so 'maybe_corrupted' never gets set for
invalid inputs.

Best regards,
Rui Yan