[Devel] [PATCH vz10] mm/memcontrol: fix cache counter imbalance in mem_cgroup_replace_folio()

[Konstantin Khorenko]

The function was checking cache flag on the new folio instead of the old one,
causing cache counter imbalance when replacing cache pages.
The new folio is freshly allocated and never has the cache flag set, so
we need to copy the flag from the old folio and charge the cache counter
accordingly.

This fixes potential cache accounting issues during page cache replacement
operations like page migration between NUMA nodes.

Fixes: 6a2ca4d515c5 ("mm: Memory cgroup page cache limit")
https://virtuozzo.atlassian.net/browse/VSTOR-111756

Signed-off-by: Konstantin Khorenko <khorenko at virtuozzo.com>

Feature: mm: Memory cgroup page cache limit
---
 mm/memcontrol.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index d0fe610306a0..28f4e518d280 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -5576,6 +5576,12 @@ void mem_cgroup_replace_folio(struct folio *old, struct folio *new)
 			page_counter_charge(&memcg->memsw, nr_pages);
 	}
 
+	if (folio_memcg_cache(old)) {
+		WRITE_ONCE(new->memcg_data,
+			   READ_ONCE(new->memcg_data) | MEMCG_DATA_PGCACHE);
+		page_counter_charge(&memcg->cache, nr_pages);
+	}
+
 	/*
 	 * finist explained the idea behind adding a WARN_ON() here:
 	 * - we do not want to check flags correctness on each flag change
@@ -5596,9 +5602,6 @@ void mem_cgroup_replace_folio(struct folio *old, struct folio *new)
 		 folio_test_swapbacked(new)   || folio_test_swapcache(new) ||
 		 folio_test_mappedtodisk(new) || folio_test_ksm(new)));
 
-	if (folio_memcg_cache(new))
-		page_counter_charge(&memcg->cache, nr_pages);
-
 	css_get(&memcg->css);
 	commit_charge(new, memcg);
 	memcg1_commit_charge(new, memcg);
-- 
2.43.5