Re: [PATCH] use find_lock_task_mm in memory cgroups oom v2

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Wed, 16 Jun 2010 10:12:20 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> On Wed, Jun 16, 2010 at 5:33 AM, KAMEZAWA Hiroyuki
> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> > On Tue, 15 Jun 2010 18:59:25 +0900
> > Minchan Kim <minchan.kim@gmail.com> wrote:
> >
> >> > -/*
> >> > +/**
> >> > + * find_lock_task_mm - Checking a process which a task belongs to has valid mm
> >> > + * and return a locked task which has a valid pointer to mm.
> >> > + *
> >>
> >> This comment should have been another patch.
> >> BTW, below comment uses "subthread" word.
> >> Personally it's easy to understand function's goal to me. :)
> >>
> >> How about following as?
> >> Checking a process which has any subthread with vaild mm
> >> ....
> >>
> > Sure. thank you. v2 is here. I removed unnecessary parts.
> >
> > ==
> > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> >
> > When the OOM killer scans task, it check a task is under memcg or
> > not when it's called via memcg's context.
> >
> > But, as Oleg pointed out, a thread group leader may have NULL ->mm
> > and task_in_mem_cgroup() may do wrong decision. We have to use
> > find_lock_task_mm() in memcg as generic OOM-Killer does.
> >
> > Changelog:
> > A - removed unnecessary changes in comments.
> >
> 
> mm->owner solves the same problem, but unfortunately we have task
> based selection in OOM killer, so we need this patch. It is quite
> ironic that we find the mm from the task and then eventually the task
> back from mm->owner and then the mem cgroup. If we already know the mm
> from oom_kill.c, I think we can change the function to work off of
> that. mm->owner->cgroup..no?
> 

There is no function as for_each_mm(). There is only for_each_process().

And, generally, there is no way to get a task via mm_struct.
To send signal etc. we need task.
(mm_owner is an option but not always configured and there will be
 complicated problem as vfork() etc.)

I think oom_kill.c has to depend on thread, not mm, unfortunately.


Thanks,
-Kame

--
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>