[PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Hugh Dickins]

If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
called from __mem_cgroup_same_or_subtree() called from page_referenced():
when processes are exiting, it's easy for mm_match_cgroup() to pass along
a NULL memcg coming from a NULL mm->owner.

Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
False because we cannot know if it was in the hierarchy, but also false
because it's better not to count a reference from an exiting process.

Signed-off-by: Hugh Dickins <hughd@google.com>
---
This a 3.5-rc issue: not needed for stable.

 mm/memcontrol.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- 3.5-rc2/mm/memcontrol.c	2012-05-30 08:17:19.400008280 -0700
+++ linux/mm/memcontrol.c	2012-06-10 08:39:39.618182396 -0700
@@ -1148,7 +1148,7 @@ bool __mem_cgroup_same_or_subtree(const
 {
 	if (root_memcg == memcg)
 		return true;
-	if (!root_memcg->use_hierarchy)
+	if (!root_memcg->use_hierarchy || !memcg)
 		return false;
 	return css_is_ancestor(&memcg->css, &root_memcg->css);
 }

--
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>

Re: [PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Johannes Weiner]

On Sun, Jun 10, 2012 at 11:54:47AM -0700, Hugh Dickins wrote:
> If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
> called from __mem_cgroup_same_or_subtree() called from page_referenced():
> when processes are exiting, it's easy for mm_match_cgroup() to pass along
> a NULL memcg coming from a NULL mm->owner.
> 
> Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
> False because we cannot know if it was in the hierarchy, but also false
> because it's better not to count a reference from an exiting process.
> 
> Signed-off-by: Hugh Dickins <hughd@google.com>

Looks like an older version of the patch that introduced it slipped
into the tree, Konstantin noted this problem during review.  The final
version did

	match = memcg && __mem_cgroup_same_or_subtree(root, memcg);

in the caller because of it.

Do you think it would be cleaner this way, since this is also the
place where that memcg is looked up, and so the "can return NULL"
handling after mem_cgroup_from_task() would be in the same place?

But either way,

Acked-by: Johannes Weiner <hannes@cmpxchg.org>

Thanks, Hugh!

--
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>

Re: [PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Hugh Dickins]

On Mon, 11 Jun 2012, Johannes Weiner wrote:
> On Sun, Jun 10, 2012 at 11:54:47AM -0700, Hugh Dickins wrote:
> > If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
> > called from __mem_cgroup_same_or_subtree() called from page_referenced():
> > when processes are exiting, it's easy for mm_match_cgroup() to pass along
> > a NULL memcg coming from a NULL mm->owner.
> > 
> > Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
> > False because we cannot know if it was in the hierarchy, but also false
> > because it's better not to count a reference from an exiting process.
> > 
> > Signed-off-by: Hugh Dickins <hughd@google.com>
> 
> Looks like an older version of the patch that introduced it slipped
> into the tree, Konstantin noted this problem during review.  The final
> version did
> 
> 	match = memcg && __mem_cgroup_same_or_subtree(root, memcg);
> 
> in the caller because of it.
> 
> Do you think it would be cleaner this way, since this is also the
> place where that memcg is looked up, and so the "can return NULL"
> handling after mem_cgroup_from_task() would be in the same place?

I don't mind, either way.

It depends on whether we add more such uses which could receive a NULL
memcg.  I tend to prefer dealing with rare conditions (which this is)
inside the callee, but common conditions before calling from the caller.

But let's let others decide.

> 
> But either way,
> 
> Acked-by: Johannes Weiner <hannes@cmpxchg.org>
> 
> Thanks, Hugh!

And thank you, Hannes!

Hugh

--
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>

Re: [PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Konstantin Khlebnikov]

Johannes Weiner wrote:
> On Sun, Jun 10, 2012 at 11:54:47AM -0700, Hugh Dickins wrote:
>> If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
>> called from __mem_cgroup_same_or_subtree() called from page_referenced():
>> when processes are exiting, it's easy for mm_match_cgroup() to pass along
>> a NULL memcg coming from a NULL mm->owner.
>>
>> Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
>> False because we cannot know if it was in the hierarchy, but also false
>> because it's better not to count a reference from an exiting process.
>>
>> Signed-off-by: Hugh Dickins<hughd@google.com>
>
> Looks like an older version of the patch that introduced it slipped
> into the tree, Konstantin noted this problem during review.  The final
> version did
>
> 	match = memcg&&  __mem_cgroup_same_or_subtree(root, memcg);
>
> in the caller because of it.
>
> Do you think it would be cleaner this way, since this is also the
> place where that memcg is looked up, and so the "can return NULL"
> handling after mem_cgroup_from_task() would be in the same place?

I agree, it cleaner, but nevertheless:

Acked-by: Konstantin Khlebnikov <khlebnikov@openvz.org>

Thanks, Hugh!

>
> But either way,
>
> Acked-by: Johannes Weiner<hannes@cmpxchg.org>
>
> Thanks, Hugh!
>
> --
> 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>

--
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>

Re: [PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Kamezawa Hiroyuki]

(2012/06/11 3:54), Hugh Dickins wrote:
> If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
> called from __mem_cgroup_same_or_subtree() called from page_referenced():
> when processes are exiting, it's easy for mm_match_cgroup() to pass along
> a NULL memcg coming from a NULL mm->owner.
>
> Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
> False because we cannot know if it was in the hierarchy, but also false
> because it's better not to count a reference from an exiting process.
>
> Signed-off-by: Hugh Dickins<hughd@google.com>

Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

--
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>

Re: [PATCH] memcg: fix use_hierarchy css_is_ancestor oops regression

[Michal Hocko]

On Sun 10-06-12 11:54:47, Hugh Dickins wrote:
> If use_hierarchy is set, reclaim testing soon oopses in css_is_ancestor()
> called from __mem_cgroup_same_or_subtree() called from page_referenced():
> when processes are exiting, it's easy for mm_match_cgroup() to pass along
> a NULL memcg coming from a NULL mm->owner.
> 
> Check for that in __mem_cgroup_same_or_subtree().  Return true or false?
> False because we cannot know if it was in the hierarchy, but also false
> because it's better not to count a reference from an exiting process.
> 
> Signed-off-by: Hugh Dickins <hughd@google.com>

Acked-by: Michal Hocko <mhocko@suse.cz>

Thanks

> ---
> This a 3.5-rc issue: not needed for stable.
> 
>  mm/memcontrol.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> --- 3.5-rc2/mm/memcontrol.c	2012-05-30 08:17:19.400008280 -0700
> +++ linux/mm/memcontrol.c	2012-06-10 08:39:39.618182396 -0700
> @@ -1148,7 +1148,7 @@ bool __mem_cgroup_same_or_subtree(const
>  {
>  	if (root_memcg == memcg)
>  		return true;
> -	if (!root_memcg->use_hierarchy)
> +	if (!root_memcg->use_hierarchy || !memcg)
>  		return false;
>  	return css_is_ancestor(&memcg->css, &root_memcg->css);
>  }

-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic

--
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>