* [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers
@ 2022-07-28 9:13 tujinjiang
2022-07-28 11:43 ` Michal Hocko
0 siblings, 1 reply; 3+ messages in thread
From: tujinjiang @ 2022-07-28 9:13 UTC (permalink / raw)
To: mhocko; +Cc: shy828301, akpm, linux-mm, linux-kernel, Jinjiang Tu
From: Jinjiang Tu <tujinjiang@bytedance.com>
When a memcg aware shrinker is registered by register_shrinker(),
shrinker->nr_deferred will not be initialized. But when the shrinker
is unregistered by unregister_shrinker(), shrinker->nr_deferred
will be freed.
Luckily, the memcg aware shrinkers in the current kernel are pre-zeroed.
But a new memcg aware shrinker may be added in the future, and we should
not assume the shrinker is pre-zeroed.
Another unregister API free_prealloced_shrinker() does not assume the
shrinker is pre-zered and free shrinker->nr_deferred only if it is
not memcg aware. So unregister_shrinker() should do like
free_prealloced_shrinker().
Signed-off-by: Jinjiang Tu <tujinjiang@bytedance.com>
---
mm/vmscan.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/mm/vmscan.c b/mm/vmscan.c
index f7d9a683e3a7..f8a9a5349b6e 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -675,8 +675,11 @@ void unregister_shrinker(struct shrinker *shrinker)
down_write(&shrinker_rwsem);
list_del(&shrinker->list);
shrinker->flags &= ~SHRINKER_REGISTERED;
- if (shrinker->flags & SHRINKER_MEMCG_AWARE)
+ if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
unregister_memcg_shrinker(shrinker);
+ up_write(&shrinker_rwsem);
+ return;
+ }
up_write(&shrinker_rwsem);
kfree(shrinker->nr_deferred);
--
2.17.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers
2022-07-28 9:13 [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers tujinjiang
@ 2022-07-28 11:43 ` Michal Hocko
2022-07-29 2:11 ` Jinjiang Tu
0 siblings, 1 reply; 3+ messages in thread
From: Michal Hocko @ 2022-07-28 11:43 UTC (permalink / raw)
To: tujinjiang; +Cc: shy828301, akpm, linux-mm, linux-kernel
On Thu 28-07-22 17:13:41, tujinjiang@bytedance.com wrote:
> From: Jinjiang Tu <tujinjiang@bytedance.com>
>
> When a memcg aware shrinker is registered by register_shrinker(),
> shrinker->nr_deferred will not be initialized. But when the shrinker
> is unregistered by unregister_shrinker(), shrinker->nr_deferred
> will be freed.
>
> Luckily, the memcg aware shrinkers in the current kernel are pre-zeroed.
> But a new memcg aware shrinker may be added in the future, and we should
> not assume the shrinker is pre-zeroed.
>
> Another unregister API free_prealloced_shrinker() does not assume the
> shrinker is pre-zered and free shrinker->nr_deferred only if it is
> not memcg aware. So unregister_shrinker() should do like
> free_prealloced_shrinker().
>
> Signed-off-by: Jinjiang Tu <tujinjiang@bytedance.com>
> ---
> mm/vmscan.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index f7d9a683e3a7..f8a9a5349b6e 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -675,8 +675,11 @@ void unregister_shrinker(struct shrinker *shrinker)
> down_write(&shrinker_rwsem);
> list_del(&shrinker->list);
> shrinker->flags &= ~SHRINKER_REGISTERED;
> - if (shrinker->flags & SHRINKER_MEMCG_AWARE)
> + if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
> unregister_memcg_shrinker(shrinker);
> + up_write(&shrinker_rwsem);
> + return;
> + }
> up_write(&shrinker_rwsem);
>
> kfree(shrinker->nr_deferred);
Can we get rid of the code duplication?
---
diff --git a/mm/vmscan.c b/mm/vmscan.c
index f7d9a683e3a7..308279414fe8 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -632,12 +632,10 @@ int prealloc_shrinker(struct shrinker *shrinker)
return 0;
}
-void free_prealloced_shrinker(struct shrinker *shrinker)
+static void __free_shrinker(struct shrinker *shrinker)
{
if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
- down_write(&shrinker_rwsem);
unregister_memcg_shrinker(shrinker);
- up_write(&shrinker_rwsem);
return;
}
@@ -645,6 +643,13 @@ void free_prealloced_shrinker(struct shrinker *shrinker)
shrinker->nr_deferred = NULL;
}
+void free_prealloced_shrinker(struct shrinker *shrinker)
+{
+ down_write(&shrinker_rwsem);
+ __free_shrinker(shrinker);
+ up_write(&shrinker_rwsem);
+}
+
void register_shrinker_prepared(struct shrinker *shrinker)
{
down_write(&shrinker_rwsem);
@@ -675,12 +680,9 @@ void unregister_shrinker(struct shrinker *shrinker)
down_write(&shrinker_rwsem);
list_del(&shrinker->list);
shrinker->flags &= ~SHRINKER_REGISTERED;
- if (shrinker->flags & SHRINKER_MEMCG_AWARE)
- unregister_memcg_shrinker(shrinker);
- up_write(&shrinker_rwsem);
- kfree(shrinker->nr_deferred);
- shrinker->nr_deferred = NULL;
+ __free_shrinker(shrinker);
+ up_write(&shrinker_rwsem);
}
EXPORT_SYMBOL(unregister_shrinker);
--
Michal Hocko
SUSE Labs
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: Re: [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers
2022-07-28 11:43 ` Michal Hocko
@ 2022-07-29 2:11 ` Jinjiang Tu
0 siblings, 0 replies; 3+ messages in thread
From: Jinjiang Tu @ 2022-07-29 2:11 UTC (permalink / raw)
To: Michal Hocko; +Cc: Yang Shi, akpm, linux-mm, linux-kernel
On Thu, Jul 28, 2022 at 7:43 PM Michal Hocko <mhocko@suse.com> wrote:
>
> On Thu 28-07-22 17:13:41, tujinjiang@bytedance.com wrote:
> > From: Jinjiang Tu <tujinjiang@bytedance.com>
> >
> > When a memcg aware shrinker is registered by register_shrinker(),
> > shrinker->nr_deferred will not be initialized. But when the shrinker
> > is unregistered by unregister_shrinker(), shrinker->nr_deferred
> > will be freed.
> >
> > Luckily, the memcg aware shrinkers in the current kernel are pre-zeroed.
> > But a new memcg aware shrinker may be added in the future, and we should
> > not assume the shrinker is pre-zeroed.
> >
> > Another unregister API free_prealloced_shrinker() does not assume the
> > shrinker is pre-zered and free shrinker->nr_deferred only if it is
> > not memcg aware. So unregister_shrinker() should do like
> > free_prealloced_shrinker().
> >
> > Signed-off-by: Jinjiang Tu <tujinjiang@bytedance.com>
> > ---
> > mm/vmscan.c | 5 ++++-
> > 1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index f7d9a683e3a7..f8a9a5349b6e 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -675,8 +675,11 @@ void unregister_shrinker(struct shrinker *shrinker)
> > down_write(&shrinker_rwsem);
> > list_del(&shrinker->list);
> > shrinker->flags &= ~SHRINKER_REGISTERED;
> > - if (shrinker->flags & SHRINKER_MEMCG_AWARE)
> > + if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
> > unregister_memcg_shrinker(shrinker);
> > + up_write(&shrinker_rwsem);
> > + return;
> > + }
> > up_write(&shrinker_rwsem);
> >
> > kfree(shrinker->nr_deferred);
>
> Can we get rid of the code duplication?
> ---
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index f7d9a683e3a7..308279414fe8 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -632,12 +632,10 @@ int prealloc_shrinker(struct shrinker *shrinker)
> return 0;
> }
>
> -void free_prealloced_shrinker(struct shrinker *shrinker)
> +static void __free_shrinker(struct shrinker *shrinker)
> {
> if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
> - down_write(&shrinker_rwsem);
> unregister_memcg_shrinker(shrinker);
> - up_write(&shrinker_rwsem);
> return;
> }
>
> @@ -645,6 +643,13 @@ void free_prealloced_shrinker(struct shrinker *shrinker)
> shrinker->nr_deferred = NULL;
> }
>
> +void free_prealloced_shrinker(struct shrinker *shrinker)
> +{
> + down_write(&shrinker_rwsem);
> + __free_shrinker(shrinker);
> + up_write(&shrinker_rwsem);
> +}
> +
> void register_shrinker_prepared(struct shrinker *shrinker)
> {
> down_write(&shrinker_rwsem);
> @@ -675,12 +680,9 @@ void unregister_shrinker(struct shrinker *shrinker)
> down_write(&shrinker_rwsem);
> list_del(&shrinker->list);
> shrinker->flags &= ~SHRINKER_REGISTERED;
> - if (shrinker->flags & SHRINKER_MEMCG_AWARE)
> - unregister_memcg_shrinker(shrinker);
> - up_write(&shrinker_rwsem);
>
> - kfree(shrinker->nr_deferred);
> - shrinker->nr_deferred = NULL;
> + __free_shrinker(shrinker);
> + up_write(&shrinker_rwsem);
> }
> EXPORT_SYMBOL(unregister_shrinker);
>
> --
> Michal Hocko
> SUSE Labs
Yes, the code is clearer.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-07-29 2:11 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-28 9:13 [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers tujinjiang
2022-07-28 11:43 ` Michal Hocko
2022-07-29 2:11 ` Jinjiang Tu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).