Re: [PATCH] mm,oom: use per signal_struct flag rather than clear TIF_MEMDIE

[Oleg Nesterov <oleg@redhat.com>]

On 06/30, Tetsuo Handa wrote:
>
> Michal Hocko wrote:
> > On Wed 29-06-16 22:01:08, Oleg Nesterov wrote:
>
> > > Btw, do we still need this list_for_each_entry(child, &t->children, sibling)
> > > loop in oom_kill_process() ?
> >
> > Well, to be honest, I don't know. This is a heuristic we have been doing
> > for a long time. I do not know how many times it really matters. It can
> > even be harmful in loads where children are created in the same pace OOM
> > killer is killing them. Not sure how likely is that though...
> > Let me think whether we can do something about that.
>
> I'm using that behavior in order to test almost OOM situation. ;)

Can you explain why do we want this behaviour?

Except, again, sysctl_oom_kill_allocating_task, see my reply to Michal.

> By the way, are you going to fix use_mm() race? Currently, we don't wake up
> OOM reaper if some kernel thread is holding a reference to that mm via
> use_mm(). But currently we can hit

Yes, and I already mention this race, and this is why I think we should not
skip kthreads.

> race. I think we need to make use_mm() fail after mark_oom_victim() is called.

Perhaps this makes sense anyway later, but I still think we do not really
care. I'll write another email...

Oleg.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>