2021-02-23 09:32:51

by Muchun Song

[permalink] [raw]
Subject: [PATCH] mm: memcontrol: fix get_active_memcg return value

We use a global percpu int_active_memcg variable to store the remote
memcg when we are in the interrupt context. But get_active_memcg always
return the current->active_memcg or root_mem_cgroup. The remote memcg
(set in the interrupt context) is ignored. This is not what we want.
So fix it.

Fixes: 37d5985c003d ("mm: kmem: prepare remote memcg charging infra for interrupt contexts")
Signed-off-by: Muchun Song <[email protected]>
---
mm/memcontrol.c | 10 +++-------
1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index be6bc5044150..bbe25655f7eb 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1061,13 +1061,9 @@ static __always_inline struct mem_cgroup *get_active_memcg(void)

rcu_read_lock();
memcg = active_memcg();
- if (memcg) {
- /* current->active_memcg must hold a ref. */
- if (WARN_ON_ONCE(!css_tryget(&memcg->css)))
- memcg = root_mem_cgroup;
- else
- memcg = current->active_memcg;
- }
+ /* remote memcg must hold a ref. */
+ if (memcg && WARN_ON_ONCE(!css_tryget(&memcg->css)))
+ memcg = root_mem_cgroup;
rcu_read_unlock();

return memcg;
--
2.11.0


2021-02-23 15:23:15

by Shakeel Butt

[permalink] [raw]
Subject: Re: [PATCH] mm: memcontrol: fix get_active_memcg return value

On Tue, Feb 23, 2021 at 1:14 AM Muchun Song <[email protected]> wrote:
>
> We use a global percpu int_active_memcg variable to store the remote
> memcg when we are in the interrupt context. But get_active_memcg always
> return the current->active_memcg or root_mem_cgroup. The remote memcg
> (set in the interrupt context) is ignored. This is not what we want.
> So fix it.
>
> Fixes: 37d5985c003d ("mm: kmem: prepare remote memcg charging infra for interrupt contexts")
> Signed-off-by: Muchun Song <[email protected]>

Good catch.
Cc: [email protected]

Reviewed-by: Shakeel Butt <[email protected]>

> ---
> mm/memcontrol.c | 10 +++-------
> 1 file changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index be6bc5044150..bbe25655f7eb 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -1061,13 +1061,9 @@ static __always_inline struct mem_cgroup *get_active_memcg(void)
>
> rcu_read_lock();
> memcg = active_memcg();
> - if (memcg) {
> - /* current->active_memcg must hold a ref. */
> - if (WARN_ON_ONCE(!css_tryget(&memcg->css)))
> - memcg = root_mem_cgroup;
> - else
> - memcg = current->active_memcg;
> - }
> + /* remote memcg must hold a ref. */
> + if (memcg && WARN_ON_ONCE(!css_tryget(&memcg->css)))
> + memcg = root_mem_cgroup;
> rcu_read_unlock();
>
> return memcg;
> --
> 2.11.0
>

2021-02-23 15:38:30

by Roman Gushchin

[permalink] [raw]
Subject: Re: [PATCH] mm: memcontrol: fix get_active_memcg return value

On Tue, Feb 23, 2021 at 05:11:01PM +0800, Muchun Song wrote:
> We use a global percpu int_active_memcg variable to store the remote
> memcg when we are in the interrupt context. But get_active_memcg always
> return the current->active_memcg or root_mem_cgroup. The remote memcg
> (set in the interrupt context) is ignored. This is not what we want.
> So fix it.
>
> Fixes: 37d5985c003d ("mm: kmem: prepare remote memcg charging infra for interrupt contexts")
> Signed-off-by: Muchun Song <[email protected]>
> ---
> mm/memcontrol.c | 10 +++-------
> 1 file changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index be6bc5044150..bbe25655f7eb 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -1061,13 +1061,9 @@ static __always_inline struct mem_cgroup *get_active_memcg(void)
>
> rcu_read_lock();
> memcg = active_memcg();
> - if (memcg) {
> - /* current->active_memcg must hold a ref. */
> - if (WARN_ON_ONCE(!css_tryget(&memcg->css)))
> - memcg = root_mem_cgroup;
> - else
> - memcg = current->active_memcg;
> - }
> + /* remote memcg must hold a ref. */
> + if (memcg && WARN_ON_ONCE(!css_tryget(&memcg->css)))
> + memcg = root_mem_cgroup;

Ouch, this is a good one!

Reviewed-by: Roman Gushchin <[email protected]>

Thank you!