Re: [PATCH] mm, oom: OOM victims do not need to select next OOM victim unless __GFP_NOFAIL.

[Michal Hocko <mhocko@suse.com>]

On Tue 28-08-18 19:20:32, Tetsuo Handa wrote:
> On 2018/08/20 20:10, Michal Hocko wrote:
> > On Mon 20-08-18 20:02:30, Tetsuo Handa wrote:
> >> On 2018/08/20 19:53, Michal Hocko wrote:
> >>> On Mon 20-08-18 19:37:45, Tetsuo Handa wrote:
> >>>> Commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
> >>>> oom_reaped tasks") changed to select next OOM victim as soon as
> >>>> MMF_OOM_SKIP is set. But since OOM victims can try ALLOC_OOM allocation
> >>>> and then give up (if !memcg OOM) or can use forced charge and then retry
> >>>> (if memcg OOM), OOM victims do not need to select next OOM victim unless
> >>>> they are doing __GFP_NOFAIL allocations.
> >>>
> >>> I do not like this at all. It seems hackish to say the least. And more
> >>> importantly...
> >>>
> >>>> This is a quick mitigation because syzbot is hitting WARN(1) caused by
> >>>> this race window [1]. More robust fix (e.g. make it possible to reclaim
> >>>> more memory before MMF_OOM_SKIP is set, wait for some more after
> >>>> MMF_OOM_SKIP is set) is a future work.
> >>>
> >>> .. there is already a patch (by Johannes) for that warning IIRC.
> >>
> >> You mean http://lkml.kernel.org/r/20180808144515.GA9276@cmpxchg.org ?
> > 
> > Yes
> > 
> >> But I can't find that patch in linux-next.git . And as far as I know,
> >> no patch was sent to linux.git for handling this problem. Therefore,
> >> I wrote this patch so that we can apply for 4.19-rc1.
> > 
> > I am pretty sure Johannes will post them later after merge window
> > closes.
> > 
> 
> But Johannes' patch will not prevent the OOM killer from needlessly selecting
> next OOM victim, will it? I still think we can apply my patch in order to prevent
> the OOM killer from needlessly selecting next OOM victim.

see my feedback on your patch.

-- 
Michal Hocko
SUSE Labs