Re: [PATCH] mm/damon: validate addr_unit to be power of 2

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

Hi SeongJae,

Regarding the kdamond termination, I have an idea:
Should we avoid terminating kdamond just because of invalid parameters
from user input (e.g., non-power-of-2 'min_region_sz' or 'addr_unit')?

Since a terminated kdamond cannot be easily restarted currently (if it
is not terminated by 'echo N > enabled'), we should probably try to
reduce the probability of such unexpected exits caused by user input.

I'm thinking of validating these parameters early in damon_commit_ctx().
If they are invalid, we could reset 'dst->maybe_corrupted' to 'false'
and return -EINVAL.

This way, since no actual modifications to 'dst' were made before the
check, we can safely clear the 'maybe_corrupted' flag. User will then
receive the error but kdamond should be able to keep running with its
existing configuration instead of exiting.

Does this approach make sense to you?

Best regards,
Rui Yan