On Tue, 30 Aug 2011, Oleg Nesterov wrote:
> > @@ -447,6 +431,9 @@ static int oom_kill_task(struct task_struct *p, struct mem_cgroup *mem)
> > for_each_process(q)
> > if (q->mm == mm && !same_thread_group(q, p) &&
> > !(q->flags & PF_KTHREAD)) {
>
> (I guess this is on top of -mm patch)
>
Yes, it's based on
oom-avoid-killing-kthreads-if-they-assume-the-oom-killed-threads-mm.patch
which I thought would be pushed for the 3.1 rc series, we certainly don't
want to SIGKILL kthreads :)
> > + if (q->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
> > + continue;
> > +
>
> Afaics, this is the only change apart from "removes mm->oom_disable_count
> entirely", looks reasonable to me.
>
Yeah, it's necessary because this loop in oom_kill_task() kills all
user threads in different thread groups unconditionally if they share the
same mm, so we need to ensure that we aren't sending a SIGKILL to anything
that is actually oom disabled. Before, the check in oom_badness() would
have prevented the task (`p' in this function) from being chosen in the
first place.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>