* Re: [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems
2026-07-15 10:35 [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems Usama Arif
@ 2026-07-16 1:20 ` Baolin Wang
2026-07-20 10:42 ` David Sterba
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Baolin Wang @ 2026-07-16 1:20 UTC (permalink / raw)
To: Usama Arif, Andrew Morton, david, dgc, qi.zheng, brauner, cgroups,
clm, dsterba, hannes, hughd, jack, linux-btrfs, linux-fsdevel,
linux-kernel, linux-mm, mhocko, muchun.song, roman.gushchin,
shakeel.butt, Al Viro, kernel-team
On 7/15/26 6:35 PM, Usama Arif wrote:
> Commit 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects
> in memcg slab shrink") added a check in fs/super.c that skipped every
> ->nr_cached_objects() hook whenever the shrinker was invoked for a
> non-root memcg, on the assumption that none of them honour sc->memcg.
>
> That assumption is wrong for XFS, whose inode-reclaim hook is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab. Encoding a blanket "never memcg-aware" policy in fs/super.c
> short-circuits that path.
>
> Push the check down into the callbacks whose counters really are
> irrelevant to per-memcg reclaim - btrfs_nr_cached_objects() and
> shmem_unused_huge_count() - and drop the fs/super.c gate. Each
> filesystem can now lift the restriction independently if its counter
> later grows memcg awareness, without touching fs/super.c.
>
> Introduce mem_cgroup_shrink_is_root() in <linux/memcontrol.h> so the
> callbacks don't open-code "sc->memcg is NULL or root".
>
> Fixes: 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects in memcg slab shrink")
> Acked-by: Qi Zheng <qi.zheng@linux.dev>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
> Signed-off-by: Usama Arif <usama.arif@linux.dev>
> ---
[snip]
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index e1f46a0016fc..5407e4200460 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -520,6 +520,22 @@ static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
> return (memcg == root_mem_cgroup);
> }
>
> +/**
> + * mem_cgroup_shrink_is_root - is this a global or root-memcg shrink invocation?
> + * @sc: shrink_control describing the current shrinker call
> + *
> + * Returns true when @sc represents a global reclaim shrink (sc->memcg == NULL)
> + * or a root-memcg shrink, i.e. not a per-memcg iteration of
> + * shrink_slab_memcg(). Filesystems whose ->nr_cached_objects()/
> + * ->free_cached_objects() implementations operate on filesystem-global state
> + * and do not honour sc->memcg can use this to early-return 0 in per-memcg
> + * contexts.
> + */
> +static inline bool mem_cgroup_shrink_is_root(struct shrink_control *sc)
> +{
> + return !sc->memcg || mem_cgroup_is_root(sc->memcg);
> +}
> +
> static inline bool obj_cgroup_is_root(const struct obj_cgroup *objcg)
> {
> return objcg->is_root;
> @@ -1071,6 +1087,11 @@ static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
> return true;
> }
>
> +static inline bool mem_cgroup_shrink_is_root(struct shrink_control *sc)
> +{
> + return true;
> +}
> +
> static inline bool obj_cgroup_is_root(const struct obj_cgroup *objcg)
> {
> return true;
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 5789a0f5a346..dc8cd4f563f4 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -846,6 +846,16 @@ static long shmem_unused_huge_count(struct super_block *sb,
> struct shrink_control *sc)
> {
> struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
> +
> + /*
> + * The per-superblock shrinklist is filesystem-global and does not
> + * honour sc->memcg, so it is only meaningful on the global (kswapd or
> + * root direct reclaim) shrink path. Skip the per-memcg iterations of
> + * shrink_slab_memcg() to avoid queueing duplicate global work.
> + */
> + if (!mem_cgroup_shrink_is_root(sc))
> + return 0;
> +
> return READ_ONCE(sbinfo->shrinklist_len);
> }
> #else /* !CONFIG_TRANSPARENT_HUGEPAGE */
For shmem part, LGTM.
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems
2026-07-15 10:35 [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems Usama Arif
2026-07-16 1:20 ` Baolin Wang
@ 2026-07-20 10:42 ` David Sterba
2026-07-30 8:58 ` Qi Zheng
2026-08-10 10:19 ` Usama Arif
3 siblings, 0 replies; 5+ messages in thread
From: David Sterba @ 2026-07-20 10:42 UTC (permalink / raw)
To: Usama Arif
Cc: Andrew Morton, david, dgc, qi.zheng, baolin.wang, brauner,
cgroups, clm, dsterba, hannes, hughd, jack, linux-btrfs,
linux-fsdevel, linux-kernel, linux-mm, mhocko, muchun.song,
roman.gushchin, shakeel.butt, Al Viro, kernel-team
On Wed, Jul 15, 2026 at 03:35:16AM -0700, Usama Arif wrote:
> Commit 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects
> in memcg slab shrink") added a check in fs/super.c that skipped every
> ->nr_cached_objects() hook whenever the shrinker was invoked for a
> non-root memcg, on the assumption that none of them honour sc->memcg.
>
> That assumption is wrong for XFS, whose inode-reclaim hook is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab. Encoding a blanket "never memcg-aware" policy in fs/super.c
> short-circuits that path.
>
> Push the check down into the callbacks whose counters really are
> irrelevant to per-memcg reclaim - btrfs_nr_cached_objects() and
> shmem_unused_huge_count() - and drop the fs/super.c gate. Each
> filesystem can now lift the restriction independently if its counter
> later grows memcg awareness, without touching fs/super.c.
>
> Introduce mem_cgroup_shrink_is_root() in <linux/memcontrol.h> so the
> callbacks don't open-code "sc->memcg is NULL or root".
>
> Fixes: 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects in memcg slab shrink")
> Acked-by: Qi Zheng <qi.zheng@linux.dev>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
> Signed-off-by: Usama Arif <usama.arif@linux.dev>
> ---
> v1 -> v2:
> - Do not gate xfs_fs_nr_cached_objects(); XFS's inode reclaim is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab (Dave Chinner).
> - Add mem_cgroup_shrink_is_root() helper in <linux/memcontrol.h> so the
> filesystem callbacks don't open-code "sc->memcg is NULL or root".
> (Dave Chinner)
> - Add fixes tag (Dave Chinner)
> ---
For
> fs/btrfs/super.c | 10 ++++++++++
Acked-by: David Sterba <dsterba@suse.com>
Thanks.
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems
2026-07-15 10:35 [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems Usama Arif
2026-07-16 1:20 ` Baolin Wang
2026-07-20 10:42 ` David Sterba
@ 2026-07-30 8:58 ` Qi Zheng
2026-08-10 10:19 ` Usama Arif
3 siblings, 0 replies; 5+ messages in thread
From: Qi Zheng @ 2026-07-30 8:58 UTC (permalink / raw)
To: Usama Arif, Andrew Morton, david, dgc, baolin.wang, brauner,
cgroups, clm, dsterba, hannes, hughd, jack, linux-btrfs,
linux-fsdevel, linux-kernel, linux-mm, mhocko, muchun.song,
roman.gushchin, shakeel.butt, Al Viro, kernel-team
On 7/15/26 6:35 PM, Usama Arif wrote:
> Commit 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects
> in memcg slab shrink") added a check in fs/super.c that skipped every
> ->nr_cached_objects() hook whenever the shrinker was invoked for a
> non-root memcg, on the assumption that none of them honour sc->memcg.
>
> That assumption is wrong for XFS, whose inode-reclaim hook is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab. Encoding a blanket "never memcg-aware" policy in fs/super.c
> short-circuits that path.
>
> Push the check down into the callbacks whose counters really are
> irrelevant to per-memcg reclaim - btrfs_nr_cached_objects() and
> shmem_unused_huge_count() - and drop the fs/super.c gate. Each
> filesystem can now lift the restriction independently if its counter
> later grows memcg awareness, without touching fs/super.c.
>
> Introduce mem_cgroup_shrink_is_root() in <linux/memcontrol.h> so the
> callbacks don't open-code "sc->memcg is NULL or root".
>
> Fixes: 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects in memcg slab shrink")
> Acked-by: Qi Zheng <qi.zheng@linux.dev>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
> Signed-off-by: Usama Arif <usama.arif@linux.dev>
> ---
> v1 -> v2:
> - Do not gate xfs_fs_nr_cached_objects(); XFS's inode reclaim is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab (Dave Chinner).
> - Add mem_cgroup_shrink_is_root() helper in <linux/memcontrol.h> so the
> filesystem callbacks don't open-code "sc->memcg is NULL or root".
> (Dave Chinner)
> - Add fixes tag (Dave Chinner)
> ---
> fs/btrfs/super.c | 10 ++++++++++
> fs/super.c | 19 ++-----------------
> include/linux/memcontrol.h | 21 +++++++++++++++++++++
> mm/shmem.c | 10 ++++++++++
> 4 files changed, 43 insertions(+), 17 deletions(-)
>
> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> index a7d804219bec..cc4537435399 100644
> --- a/fs/btrfs/super.c
> +++ b/fs/btrfs/super.c
> @@ -22,6 +22,7 @@
> #include <linux/namei.h>
> #include <linux/miscdevice.h>
> #include <linux/magic.h>
> +#include <linux/memcontrol.h>
> #include <linux/slab.h>
> #include <linux/ratelimit.h>
> #include <linux/crc32c.h>
> @@ -2434,6 +2435,15 @@ static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_contro
> struct btrfs_fs_info *fs_info = btrfs_sb(sb);
> const s64 nr = percpu_counter_read_positive(&fs_info->evictable_extent_maps);
>
> + /*
> + * The evictable extent map counter is filesystem-global and does not
> + * honour sc->memcg, so it is only meaningful on the global (kswapd or
> + * root direct reclaim) shrink path. Skip the per-memcg iterations of
> + * shrink_slab_memcg() to avoid queueing duplicate global work.
> + */
> + if (!mem_cgroup_shrink_is_root(sc))
> + return 0;
> +
> trace_btrfs_extent_map_shrinker_count(fs_info, nr);
>
> return nr;
> diff --git a/fs/super.c b/fs/super.c
> index d2d04a6f4f84..a8fd61136aaf 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -24,7 +24,6 @@
> #include <linux/export.h>
> #include <linux/slab.h>
> #include <linux/blkdev.h>
> -#include <linux/memcontrol.h>
> #include <linux/mount.h>
> #include <linux/security.h>
> #include <linux/writeback.h> /* for the emergency remount stuff */
> @@ -170,19 +169,6 @@ static void super_wake(struct super_block *sb, unsigned int flag)
> wake_up_var(&sb->s_flags);
> }
>
> -/*
> - * The s_op->nr_cached_objects hooks (used for example by btrfs and xfs)
> - * operate on filesystem-global state and ignore sc->memcg. Driving them
> - * from per-memcg shrink_slab_memcg() invocations only burns CPU walking
> - * per-cpu counters and queueing duplicate work: the actual reclaim happens on
> - * the global path (kswapd or root direct reclaim) regardless. Restrict them
> - * to that path.
> - */
> -static inline bool super_fs_objects_eligible(struct shrink_control *sc)
> -{
> - return !sc->memcg || mem_cgroup_is_root(sc->memcg);
> -}
> -
Hi Christian, it looks like you missed this when pushing. The commit
0ef8faff490be (next-20260729) doesn't include this part of the code.
Thanks,
Qi
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems
2026-07-15 10:35 [PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems Usama Arif
` (2 preceding siblings ...)
2026-07-30 8:58 ` Qi Zheng
@ 2026-08-10 10:19 ` Usama Arif
3 siblings, 0 replies; 5+ messages in thread
From: Usama Arif @ 2026-08-10 10:19 UTC (permalink / raw)
To: Usama Arif
Cc: Andrew Morton, david, dgc, qi.zheng, baolin.wang, brauner,
cgroups, clm, dsterba, hannes, hughd, jack, linux-btrfs,
linux-fsdevel, linux-kernel, linux-mm, mhocko, muchun.song,
roman.gushchin, shakeel.butt, Al Viro, kernel-team
On Wed, 15 Jul 2026 03:35:16 -0700 Usama Arif <usama.arif@linux.dev> wrote:
> Commit 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects
> in memcg slab shrink") added a check in fs/super.c that skipped every
> ->nr_cached_objects() hook whenever the shrinker was invoked for a
> non-root memcg, on the assumption that none of them honour sc->memcg.
>
> That assumption is wrong for XFS, whose inode-reclaim hook is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab. Encoding a blanket "never memcg-aware" policy in fs/super.c
> short-circuits that path.
>
> Push the check down into the callbacks whose counters really are
> irrelevant to per-memcg reclaim - btrfs_nr_cached_objects() and
> shmem_unused_huge_count() - and drop the fs/super.c gate. Each
> filesystem can now lift the restriction independently if its counter
> later grows memcg awareness, without touching fs/super.c.
>
> Introduce mem_cgroup_shrink_is_root() in <linux/memcontrol.h> so the
> callbacks don't open-code "sc->memcg is NULL or root".
>
> Fixes: 0baad6f9b997 ("fs/super: skip non-memcg-aware nr_cached_objects in memcg slab shrink")
> Acked-by: Qi Zheng <qi.zheng@linux.dev>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
> Signed-off-by: Usama Arif <usama.arif@linux.dev>
> ---
> v1 -> v2:
> - Do not gate xfs_fs_nr_cached_objects(); XFS's inode reclaim is
> intentionally driven from per-memcg contexts to free memcg-charged
> slab (Dave Chinner).
> - Add mem_cgroup_shrink_is_root() helper in <linux/memcontrol.h> so the
> filesystem callbacks don't open-code "sc->memcg is NULL or root".
> (Dave Chinner)
> - Add fixes tag (Dave Chinner)
> ---
> fs/btrfs/super.c | 10 ++++++++++
> fs/super.c | 19 ++-----------------
> include/linux/memcontrol.h | 21 +++++++++++++++++++++
> mm/shmem.c | 10 ++++++++++
> 4 files changed, 43 insertions(+), 17 deletions(-)
>
> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> index a7d804219bec..cc4537435399 100644
> --- a/fs/btrfs/super.c
> +++ b/fs/btrfs/super.c
> @@ -22,6 +22,7 @@
> #include <linux/namei.h>
> #include <linux/miscdevice.h>
> #include <linux/magic.h>
> +#include <linux/memcontrol.h>
> #include <linux/slab.h>
> #include <linux/ratelimit.h>
> #include <linux/crc32c.h>
> @@ -2434,6 +2435,15 @@ static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_contro
> struct btrfs_fs_info *fs_info = btrfs_sb(sb);
> const s64 nr = percpu_counter_read_positive(&fs_info->evictable_extent_maps);
>
> + /*
> + * The evictable extent map counter is filesystem-global and does not
> + * honour sc->memcg, so it is only meaningful on the global (kswapd or
> + * root direct reclaim) shrink path. Skip the per-memcg iterations of
> + * shrink_slab_memcg() to avoid queueing duplicate global work.
> + */
> + if (!mem_cgroup_shrink_is_root(sc))
> + return 0;
> +
> trace_btrfs_extent_map_shrinker_count(fs_info, nr);
>
> return nr;
> diff --git a/fs/super.c b/fs/super.c
> index d2d04a6f4f84..a8fd61136aaf 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -24,7 +24,6 @@
> #include <linux/export.h>
> #include <linux/slab.h>
> #include <linux/blkdev.h>
> -#include <linux/memcontrol.h>
> #include <linux/mount.h>
> #include <linux/security.h>
> #include <linux/writeback.h> /* for the emergency remount stuff */
> @@ -170,19 +169,6 @@ static void super_wake(struct super_block *sb, unsigned int flag)
> wake_up_var(&sb->s_flags);
> }
>
> -/*
> - * The s_op->nr_cached_objects hooks (used for example by btrfs and xfs)
> - * operate on filesystem-global state and ignore sc->memcg. Driving them
> - * from per-memcg shrink_slab_memcg() invocations only burns CPU walking
> - * per-cpu counters and queueing duplicate work: the actual reclaim happens on
> - * the global path (kswapd or root direct reclaim) regardless. Restrict them
> - * to that path.
> - */
> -static inline bool super_fs_objects_eligible(struct shrink_control *sc)
> -{
> - return !sc->memcg || mem_cgroup_is_root(sc->memcg);
> -}
> -
> /*
> * One thing we have to be careful of with a per-sb shrinker is that we don't
> * drop the last active reference to the superblock from within the shrinker.
> @@ -212,7 +198,7 @@ static unsigned long super_cache_scan(struct shrinker *shrink,
> if (!super_trylock_shared(sb))
> return SHRINK_STOP;
>
> - if (sb->s_op->nr_cached_objects && super_fs_objects_eligible(sc))
> + if (sb->s_op->nr_cached_objects)
> fs_objects = sb->s_op->nr_cached_objects(sb, sc);
>
> inodes = list_lru_shrink_count(&sb->s_inode_lru, sc);
> @@ -273,8 +259,7 @@ static unsigned long super_cache_count(struct shrinker *shrink,
> return 0;
> smp_rmb();
>
> - if (sb->s_op && sb->s_op->nr_cached_objects &&
> - super_fs_objects_eligible(sc))
> + if (sb->s_op && sb->s_op->nr_cached_objects)
> total_objects = sb->s_op->nr_cached_objects(sb, sc);
>
> total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc);
Hi Christian,
It looks like the changes in fs/super.c didnt make it into the patch that was
merged in vfs.fixes [1]. Just wanted to check how we could correct this?
Should Qi (or I) send a followup patch?
Thanks,
Usama
[1] https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git/commit/?h=vfs.fixes&id=0ef8faff490be6aa1a1e5dfcb0c8492689e91c0f
^ permalink raw reply [flat|nested] 5+ messages in thread