Re: [PATCH v1] mm/damon/stat: roll back context on damon_call() failure

[SeongJae Park]

On Fri, 29 May 2026 11:17:01 -0400 최유호 <dbgh9129@xxxxxxxxx> wrote:

> Dear SJ
> 
> On Fri, 29 May 2026 at 10:44, SeongJae Park <sj@xxxxxxxxxx> wrote:
> 
> > But, having one damon_ctx object in the memory is a real problem?  Why?
> 
> I agree that keeping one `damon_ctx` is unlikely to be a serious problem.
> My intention was not to address memory pressure but to maintain
> failure-path consistency. If initialization fails after the context is
> created and started, the internal state remains in place despite the
> operation returning an error. The cleanup is then deferred until a
> subsequent retry.
> 
> For this reason, I think this is worth fixing as a correctness issue.
> That said, if you feel the current cleanup is sufficient, I can drop
> this patch.

Thank you for clarifying, Yuho.  I think the current cleanup is ok, but I
understand it is confusing to read.  Apparently that confused you at least, and
that's not ok.  What about adding a comment explaining the context?


Thanks,
SJ

[...]