Re: [RFC PATCH -v2] mm, oom: introduce oom reaper

[Michal Hocko]

On Sat 28-11-15 13:39:11, Tetsuo Handa wrote:
> Michal Hocko wrote:
> > for write while write but the probability is reduced considerably wrt.
> 
> Is this "while write" garbage?

Fixed

> > Users of mmap_sem which need it for write should be carefully reviewed
> > to use _killable waiting as much as possible and reduce allocations
> > requests done with the lock held to absolute minimum to reduce the risk
> > even further.
> 
> It will be nice if we can have down_write_killable()/down_read_killable().

Yes that is an idea.

> > The API between oom killer and oom reaper is quite trivial. wake_oom_reaper
> > updates mm_to_reap with cmpxchg to guarantee only NUll->mm transition
> 
> NULL->mm

fixed

> > and oom_reaper clear this atomically once it is done with the work.
> 
> Can't oom_reaper() become as compact as below?

Good idea! I think we still need {READ,WRITE}_ONCE to prevent from any
potential mis optimizations, though.

Here is what I did:
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 333953bf4968..b50ce41194b3 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -477,21 +477,11 @@ static void oom_reap_vmas(struct mm_struct *mm)
 
 static int oom_reaper(void *unused)
 {
-	DEFINE_WAIT(wait);
-
 	while (true) {
 		struct mm_struct *mm;
-
-		prepare_to_wait(&oom_reaper_wait, &wait, TASK_UNINTERRUPTIBLE);
-		mm = READ_ONCE(mm_to_reap);
-		if (!mm) {
-			freezable_schedule();
-			finish_wait(&oom_reaper_wait, &wait);
-		} else {
-			finish_wait(&oom_reaper_wait, &wait);
-			oom_reap_vmas(mm);
-			WRITE_ONCE(mm_to_reap, NULL);
-		}
+		wait_event_freezable(oom_reaper_wait, (mm = READ_ONCE(mm_to_reap)));
+		oom_reap_vmas(mm);
+		WRITE_ONCE(mm_to_reap, NULL);
 	}
 
 	return 0;
-- 
Michal Hocko
SUSE Labs
--
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/