Re: [PATCH] memcg: bypass the reclaim and oom killer for dying tasks once oom_reaper is done

[Shakeel Butt]

On Wed, Jul 29, 2026 at 06:13:49PM -0700, Andrew Morton wrote:
> On Tue, 28 Jul 2026 19:46:12 -0700 Shakeel Butt <shakeel.butt@xxxxxxxxx> wrote:
> 
> > At Meta, we are seeing instances where an OOM killed job is stuck in the
> > exit path for several hours. In one particular case, the job was stuck
> > for more than 8 hours and I had to manually remove the memory.max limits
> > to allow the process to exit.
> 
> Not seeing the problem here - that sounds like normal kernel behavior ;)
> 
> > Reproduced with 20k threads, each parking a robust futex head on
> > its own zswapped page, OOM-group-killed while a sibling holds mmap_lock
> > for write so the reaper gives up and sets MMF_OOM_SKIP. Tested on
> > next-20260728 and baseline show ~90 seconds exit time while with the
> > patch the exit time reduced to ~3 seconds.
> 
> Very positive.
> 
> But it doesn't sound like you're fully confident that the problem is
> fully solved?

Yes as I am not able to reproduce multi hour stall, so I might be missing
something else. The patch at least fixes the issue which I am able to reproduce.

> 
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -2653,6 +2653,19 @@ static int try_charge_memcg(struct mem_cgroup *memcg, gfp_t gfp_mask,
> >  	if (!gfpflags_allow_blocking(gfp_mask))
> >  		goto nomem;
> >  
> > +	/*
> > +	 * OOM victim still needs to charge memory to exit. OOM reaper should
> > +	 * help but it might fail on mmap_lock contention. If the victim is a
> > +	 * large thread group then all exiting threads might compete on oom_lock
> > +	 * just to learn that there is nothing really killable anymore. Bail
> > +	 * out early and fail the charge to expedite their exit. They are
> > +	 * considered fully reclaimed by the oom reaper and they shouldn't
> > +	 * contribute further charges.
> > +	 */
> > +	if (tsk_is_oom_victim(current) &&
> > +	    mm_flags_test(MMF_OOM_SKIP, current->signal->oom_mm))
> > +		goto nomem;
> > +
> >  	__memcg_memory_event(mem_over_limit, MEMCG_MAX, allow_spinning);
> >  	raised_max_event = true;
> 
> Should we backport this?

I don't have any strong opinion. I wanted to deploy this patch in Meta fleet to
get some exposure. Maybe backporting will give more exposure to it.