Re: [PATCH] mm/damon/core: delete damon_target when detected invalid

[SeongJae Park <sj@kernel.org>]

On Tue, 09 Dec 2025 17:24:49 +0800 Enze Li <lienze@kylinos.cn> wrote:

> Hi SJ,
> 
> On Mon, Dec 08 2025 at 11:21:45 PM -0800, SeongJae Park wrote:
> 
> > On Tue,  9 Dec 2025 13:57:13 +0800 Enze Li <lienze@kylinos.cn> wrote:
> >
> >> Currently, DAMON does not proactively clean up invalid monitoring
> >> targets during its runtime.  When some monitored targets exit, DAMON
> >> still makes the following unnecessary function calls,
> >> 
> >>   --damon_for_each_target--
> >>   --damon_for_each_region--
> >>       damon_do_apply_schemes
> >>         damos_apply_scheme
> >>           damon_va_apply_scheme
> >>             damos_madvise
> >>               damon_get_mm
> >> 
> >> and it is only in the damon_get_mm() that it may finally discover that
> >> the monitoring target no longer exists.
[...]
> > So, from the beginning part of the patch description, I understand your concern
> > is the unnecessary function calls in kdamond_apply_schemes().  What about
> > checking the target validness using ctx->ops.target_valid() and avoid going
> > deeper if it is invalid?  Seems the damon_for_each_target() loop is a good
> > point to do this.
> >
> > What do you think?
> 
> Excellent suggestion!  Moving the validity check to the
> damon_for_each_target() loop via ctx->ops.target_valid() would indeed
> prevent all unnecessary function calls, which is more efficient and
> architecturally cleaner than my original approach. :)
> 
> I will implement this change shortly.
> 
> By the way, I'll add a "Suggested-by" tag to the commit message, if
> that's okay with you.

Surely that's okay and my honor :)


Thanks,
SJ

[...]