[PATCH] mm: workingset: compare shadow nodes node-wide without kmem accounting

["T.J. Mercier" <tjmercier@google.com>]

count_shadow_nodes() limits the number of xarray nodes holding shadow
entries to 1/8th of the pages that could be using them. For a memcg, it
counts the LRU and slab pages local to that memcg.

With kmem accounting disabled (cgroup.memory=nokmem), shadow nodes are
not charged to any memcg and __list_lru_init() makes the shadow_nodes
list_lru non-memcg-aware, so all shadow nodes of a NUMA node are on a
single list. shrinker_alloc() falls back to a non-memcg-aware shadow
node shrinker, which shrink_slab() only invokes for the root memcg.
count_shadow_nodes() therefore compares every shadow node on the NUMA
node against the pages local to the root memcg.

When most tasks run in child memcgs, the root memcg only holds a small
fraction of all pages, so the shadow node limit derived from the root
memcg page count is far below the number of shadow nodes on the NUMA
node. Since the shrinker has seeks == 0, every root shrink_slab() call
trims about half of the "excess" shadow nodes. Most refault information
is thrown away, so refaults of recently evicted file folios are not
detected. Refaulted folios lose PG_workingset and the workingset_refault
and workingset_restore stats drop.

In testing on devices with kmem accounting disabled, moving tasks from
the root memcg into per-application memcgs dropped workingset_nodes from
~37,000 to ~900 and raised workingset_nodereclaim 4.6x while
pgsteal_file rose by ~12%.

When kmem accounting is disabled, compare the shadow nodes against the
whole NUMA node, as is done without memcg. Behavior with kmem
accounting enabled is unchanged.

Cc: Jaewon Kim <jaewon31.kim@samsung.com>
Tested-by: Jaeseon Sim <jason.sim@samsung.com> # Galaxy A18 LTE
Tested-by: Youngchan Kim <ycshark.kim@samsung.com> # Galaxy S27
Tested-by: Dmitry Skiba <dskiba@google.com> # Pixel 10
Signed-off-by: T.J. Mercier <tjmercier@google.com>
---
 mm/workingset.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/mm/workingset.c b/mm/workingset.c
index 1504f91cdca5..87d5e55dd684 100644
--- a/mm/workingset.c
+++ b/mm/workingset.c
@@ -682,7 +682,16 @@ static unsigned long count_shadow_nodes(struct shrinker *shrinker,
 	 * PAGE_SIZE / xa_nodes / node_entries * 8 / PAGE_SIZE
 	 */
 #ifdef CONFIG_MEMCG
-	if (sc->memcg) {
+	/*
+	 * Without kmem accounting the shadow_nodes list_lru is not
+	 * memcg-aware. So the count above covers the whole NUMA node, and the
+	 * shrinker only runs for the root memcg, whose local pages can be a
+	 * small fraction of the node's when most tasks run in child memcgs.
+	 * Compare the shadow node count against the memcg page count only when
+	 * the shadow_nodes list_lru is memcg-aware. Otherwise use the NUMA node
+	 * page count as is done without memcg.
+	 */
+	if (sc->memcg && shadow_nodes.memcg_aware) {
 		struct lruvec *lruvec;
 		int i;
 
-- 
2.56.0.455.g6c5a503b38-goog