[PATCH] mm, memcg: Bypass high reclaim iteration for cgroup hierarchy root

[PATCH] mm, memcg: Bypass high reclaim iteration for cgroup hierarchy root

[Chris Down]

The root of the hierarchy cannot have high set, so we will never reclaim
based on it. This makes that clearer and avoids another entry.

Signed-off-by: Chris Down <chris-6Bi1550iOqEnzZ6mRAm98g@public.gmane.org>
Cc: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
Cc: Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>
Cc: Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Cc: linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org
Cc: cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Cc: linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Cc: kernel-team-b10kYP2dOMg@public.gmane.org
---
 mm/memcontrol.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 63bb6a2aab81..ab9d24a657b9 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2232,7 +2232,8 @@ static void reclaim_high(struct mem_cgroup *memcg,
 			continue;
 		memcg_memory_event(memcg, MEMCG_HIGH);
 		try_to_free_mem_cgroup_pages(memcg, nr_pages, gfp_mask, true);
-	} while ((memcg = parent_mem_cgroup(memcg)));
+	} while ((memcg = parent_mem_cgroup(memcg)) &&
+		 !mem_cgroup_is_root(memcg));
 }
 
 static void high_work_func(struct work_struct *work)
-- 
2.25.1

Re: [PATCH] mm, memcg: Bypass high reclaim iteration for cgroup hierarchy root

[Andrew Morton]

On Thu, 12 Mar 2020 16:41:37 +0000 Chris Down <chris-6Bi1550iOqEnzZ6mRAm98g@public.gmane.org> wrote:

> The root of the hierarchy cannot have high set, so we will never reclaim
> based on it. This makes that clearer and avoids another entry.
> 
> ...
>
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2232,7 +2232,8 @@ static void reclaim_high(struct mem_cgroup *memcg,
>  			continue;
>  		memcg_memory_event(memcg, MEMCG_HIGH);
>  		try_to_free_mem_cgroup_pages(memcg, nr_pages, gfp_mask, true);
> -	} while ((memcg = parent_mem_cgroup(memcg)));
> +	} while ((memcg = parent_mem_cgroup(memcg)) &&
> +		 !mem_cgroup_is_root(memcg));
>  }
>  
>  static void high_work_func(struct work_struct *work)

Does someone have time to review this one?

Thanks.

Re: [PATCH] mm, memcg: Bypass high reclaim iteration for cgroup hierarchy root

[Johannes Weiner]

On Thu, Mar 12, 2020 at 04:41:37PM +0000, Chris Down wrote:
> The root of the hierarchy cannot have high set, so we will never reclaim
> based on it. This makes that clearer and avoids another entry.
> 
> Signed-off-by: Chris Down <chris-6Bi1550iOqEnzZ6mRAm98g@public.gmane.org>
> Cc: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
> Cc: Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>
> Cc: Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
> Cc: linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org
> Cc: cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Cc: linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Cc: kernel-team-b10kYP2dOMg@public.gmane.org

This makes sense, memory.high doesn't exist on the root.

And the mem_cgroup_is_root() check, a simple pointer comparison, is
cheaper than reading the page_counter atomic and memcg->high (which we
already know to be PAGE_COUNTER_MAX).

Acked-by: Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>