[PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[zangchunxin]

From: Chunxin Zang <zangchunxin@bytedance.com>

On our server, there are about 10k memcg in one machine. They use memory
very frequently. We have observed that drop_caches can take a
considerable amount of time, and can't stop it.

There are two reasons:
1. There is somebody constantly generating more objects to reclaim
   on drop_caches, result the 'freed' always bigger than 10.
2. The process has no chance to process signals.

We can get the following info through 'ps':

  root:~# ps -aux | grep drop
  root  357956 ... R    Aug25 21119854:55 echo 3 > /proc/sys/vm/drop_caches
  root 1771385 ... R    Aug16 21146421:17 echo 3 > /proc/sys/vm/drop_caches

Add a bail out on the fatal signals in the main loop so that the
operation can be terminated by userspace.

Signed-off-by: Chunxin Zang <zangchunxin@bytedance.com>
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
---

	changelogs in v4:
	changelogs in v3: 
	1) Fix some descriptive problems pointed out by Michal Hocko.
		v2 named: mm/vmscan: fix infinite loop in drop_slab_node

	changelogs in v2: 
	1) via check fatal signal break loop.

 mm/vmscan.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index b6d84326bdf2..6b2b5d420510 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -704,6 +704,9 @@ void drop_slab_node(int nid)
 	do {
 		struct mem_cgroup *memcg = NULL;
 
+		if (signal_pending(current))
+			return;
+
 		freed = 0;
 		memcg = mem_cgroup_iter(NULL, NULL, NULL);
 		do {
-- 
2.11.0

Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Michal Hocko]

On Tue 15-09-20 21:27:41, zangchunxin@bytedance.com wrote:
> From: Chunxin Zang <zangchunxin@bytedance.com>
> 
> On our server, there are about 10k memcg in one machine. They use memory
> very frequently. We have observed that drop_caches can take a
> considerable amount of time, and can't stop it.
> 
> There are two reasons:
> 1. There is somebody constantly generating more objects to reclaim
>    on drop_caches, result the 'freed' always bigger than 10.
> 2. The process has no chance to process signals.
> 
> We can get the following info through 'ps':
> 
>   root:~# ps -aux | grep drop
>   root  357956 ... R    Aug25 21119854:55 echo 3 > /proc/sys/vm/drop_caches
>   root 1771385 ... R    Aug16 21146421:17 echo 3 > /proc/sys/vm/drop_caches
> 
> Add a bail out on the fatal signals in the main loop so that the
> operation can be terminated by userspace.
> 
> Signed-off-by: Chunxin Zang <zangchunxin@bytedance.com>
> Signed-off-by: Muchun Song <songmuchun@bytedance.com>

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

> ---
> 
> 	changelogs in v4:
> 	changelogs in v3: 
> 	1) Fix some descriptive problems pointed out by Michal Hocko.
> 		v2 named: mm/vmscan: fix infinite loop in drop_slab_node
> 
> 	changelogs in v2: 
> 	1) via check fatal signal break loop.
> 
>  mm/vmscan.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index b6d84326bdf2..6b2b5d420510 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -704,6 +704,9 @@ void drop_slab_node(int nid)
>  	do {
>  		struct mem_cgroup *memcg = NULL;
>  
> +		if (signal_pending(current))
> +			return;
> +
>  		freed = 0;
>  		memcg = mem_cgroup_iter(NULL, NULL, NULL);
>  		do {
> -- 
> 2.11.0
> 

-- 
Michal Hocko
SUSE Labs

Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Chris Down]

zangchunxin@bytedance.com writes:
>+		if (signal_pending(current))
>+			return;

This doesn't match your patch title, please update it. :-)

After that, you can add:

Acked-by: Chris Down <chris@chrisdown.name>

Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Michal Hocko]

On Tue 15-09-20 15:47:37, Chris Down wrote:
> zangchunxin@bytedance.com writes:
> > +		if (signal_pending(current))
> > +			return;
> 
> This doesn't match your patch title, please update it. :-)

I have to admit I have completely missed this and I think that this
should better be fatal_signal_pending because that would make sure that
the userspace will not see an incomplete operation. This is a general
practice for other bail outs as well.

> After that, you can add:
> 
> Acked-by: Chris Down <chris@chrisdown.name>

-- 
Michal Hocko
SUSE Labs

Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Chris Down]

Michal Hocko writes:
>On Tue 15-09-20 15:47:37, Chris Down wrote:
>> zangchunxin@bytedance.com writes:
>> > +		if (signal_pending(current))
>> > +			return;
>>
>> This doesn't match your patch title, please update it. :-)
>
>I have to admit I have completely missed this and I think that this
>should better be fatal_signal_pending because that would make sure that
>the userspace will not see an incomplete operation. This is a general
>practice for other bail outs as well.

Oh sorry, to be clear, I meant the patch should match the title, not the other 
way around.

Re: [External] Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Chunxin Zang]

On Wed, Sep 16, 2020 at 12:26 AM Chris Down <chris@chrisdown.name> wrote:
>
> Michal Hocko writes:
> >On Tue 15-09-20 15:47:37, Chris Down wrote:
> >> zangchunxin@bytedance.com writes:
> >> > +          if (signal_pending(current))
> >> > +                  return;
> >>
> >> This doesn't match your patch title, please update it. :-)
> >
> >I have to admit I have completely missed this and I think that this
> >should better be fatal_signal_pending because that would make sure that
> >the userspace will not see an incomplete operation. This is a general
> >practice for other bail outs as well.
>
> Oh sorry, to be clear, I meant the patch should match the title, not the other
> way around.

My apologies about that. In my first version of patch, it's
'fatal_signal_pending'.
But in this version, I used the wrong branch. I will update it now.

Best wishes
Chunxin

Re: [External] Re: [PATCH v4] mm/vmscan: add a fatal signals check in drop_slab_node

[Chunxin Zang]

On Wed, Sep 16, 2020 at 12:26 AM Chris Down <chris@chrisdown.name> wrote:
>
> Michal Hocko writes:
> >On Tue 15-09-20 15:47:37, Chris Down wrote:
> >> zangchunxin@bytedance.com writes:
> >> > +          if (signal_pending(current))
> >> > +                  return;
> >>
> >> This doesn't match your patch title, please update it. :-)
> >
> >I have to admit I have completely missed this and I think that this
> >should better be fatal_signal_pending because that would make sure that
> >the userspace will not see an incomplete operation. This is a general
> >practice for other bail outs as well.
>
> Oh sorry, to be clear, I meant the patch should match the title, not the other
> way around.

My apologies about that. In my first version of patch, it's
'fatal_signal_pending'.
But in this version, I used the wrong branch. I will update it now.

Best wishes
Chunxin