[PATCH v2] fs: push nr_cached_objects memcg gating into individual filesystems

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 <[email protected]>
Reviewed-by: Jan Kara <[email protected]>
Reviewed-by: Shakeel Butt <[email protected]>
Signed-off-by: Usama Arif <[email protected]>
---
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);
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 */
-- 
2.53.0-Meta