Re: [PATCH v2] mm/oom_kill: show oom eligibility when displaying the current memory state of all tasks

From: Aaron Tomlin
Date: Wed Jun 16 2021 - 16:18:41 EST


On Tue 2021-06-15 14:42 +0200, Michal Hocko wrote:
> This is all true but it is not really clear why that is really a
> problem. Kernel log already contains information about reaped processes
> as they are reported to the log. I fully acknowledge that this is rather
> spartan but on the other hand from years of experience reading oom
> reports I have to say the dump_tasks is the least interesting part of
> the report (while being the most verbose one).

I understand. I suppose, in a situation whereby dump_tasks() output is only
available, for whatever reason, it can provide at least some
insight into what tasks were actually considered not OOM eligible and why.

> All that being said, I am not really opposing extending the information
> although I am a bit worried about leaking too much internal state to the
> log.

Fair enough. That said, I still feel highlighting such "ineligible" tasks
could be useful to the viewer for troubleshooting purposes; we already
display OOM_SCORE_ADJ_MIN. Consider a situation when only a few tasks in
a memcg are displayed as possibly OOM eligible but one had MMF_OOM_SKIP
applied.

In my opinion, perhaps it is better to just exclude such details
altogether. That being said, as you know, we only provide this facility
when one is interested in it anyway i.e., if oom_dump_tasks is enabled.

> What I am asking for here is a justification why this addition is a
> general improvement and how it helps uderstanding oom reports further.
> So please focus on that part.

Sure; albeit, thinking about this more, it does not provide much
understanding in simple isolation.



Kind regards,

--
Aaron Tomlin

Attachment: signature.asc
Description: PGP signature