Re: [BUGFIX][PATCH] memcg: fix oom kill behavior v3

[Daisuke Nishimura]

On Wed, 3 Mar 2010 16:23:04 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> On Wed, 3 Mar 2010 09:38:44 +0900
> KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> 
> > On Wed, 3 Mar 2010 09:26:06 +0900
> > Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> wrote:
> > 
> > > > I'll test this patch all through this night, and check whether it doesn't trigger
> > > > global oom after memcg's oom.
> > > > 
> > > O.K. It works well.
> > > Feel free to add my signs.
> > > 
> > > 	Reviewed-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > > 	Tested-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > > 
> > 
> > Thank you !
> > 
> > I'll apply Balbir's comment and post v3.
> > 
> 
> rebased onto mmotm-Mar2.
> tested on x86-64.
> 
I found a small race problem. This is the fix for it.

===
From: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>

We must avoid making oom_lock of a newly created child be negative.

Signed-off-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
---
 mm/memcontrol.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 3ce8c5b..9e25400 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1272,7 +1272,12 @@ static bool mem_cgroup_oom_lock(struct mem_cgroup *mem)
 
 static int mem_cgroup_oom_unlock_cb(struct mem_cgroup *mem, void *data)
 {
-	atomic_dec(&mem->oom_lock);
+	/*
+	 * There is a small race window where a new child can be created after
+	 * we called mem_cgroup_oom_lock(). Use atomic_add_unless() to avoid
+	 * making oom_lock of such a child be negative.
+	 */
+	atomic_add_unless(&mem->oom_lock, -1, 0);
 	return 0;
 }
 
-- 
1.6.4
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/