Re: [Question] mm/damon: conflict between DAMON_STAT and DAMON_LRU_SORT?

[SeongJae Park <sj@kernel.org>]

On Fri, 13 Mar 2026 19:17:02 +0800 Liew Rui Yan <aethernet65535@gmail.com> wrote:

> TL;DR
> =====
> When CONFIG_DAMON_STAT_ENABLED_DEFAULT=y, enabling damon_lru_sort via sysfs
> "enabled" parameter returns -EBUSY. Is this expected behavior?

Yes, it is an expected behavior.  DAMON modules run in an exclusive manner like
this, to avoid interfering others' monitoring results.

I'm planning to make them be able to run together [1], but that's a future
work.  For now, this is the expected behavior.

> If yes, should
> this limitation be documented?

Good point!  Yes, it would be better to be documented.

> 
> Reproduction
> ============
> Environment:
> - Kernel: 7.0.0-rc3+ (x86_64)
> - VM: Virtme-ng (4G RAM, 2 CPUs)
> - Config: CONFIG_DAMON_STAT_ENABLED_DEFAULT=y, CONFIG_DAMON_LRU_SORT=y
> 
> Steps:
> 1. cd /sys/module/damon_lru_sort/parameters
> 2. (optional) tune parameters: hot_thres_access_freq, cold_min_age, etc.
> 3. echo Y > enabled
> 
> Result:
>   bash: echo: write error: Device or resource busy
> 
> Workaround:
>   Disable CONFIG_DAMON_STAT_ENABLED_DEFAULT, or disable damon_stat at runtime
>   before enabling damon_lru_sort.
> 
> Questions
> =========
> 1. Is this resource conflict between damon_stat and damon_lru_sort expected?

Yes, as I mentioned above.

> 2. If yes, should this limitation be documented in:
>    - Documentation/admin-guide/mm/damon/lru_sort.rst
>    - Documentation/admin-guide/mm/damon/stat.rst

Yes.  But, I think 'Special-Purpose Access-aware Kernel Modules' section of
Documentation/mm/damon/design.rst is a better place to document it.

Would you mind sending a patch for that?

> 
> Additional Context
> ==================
> - On major distros (Arch/Fedora), CONFIG_DAMON_STAT_ENABLED_DEFAULT is typically
>   unset (disabled), so this issue may not affect most users.
> - However, for kernel developers or custom builds enabling both modules, clearer
>   documentation or error handling would improve debuggability.

I agree.

[1] DAMON-X in https://lore.kernel.org/20260307210250.204245-1-sj@kernel.org


Thanks,
SJ

[...]