[PATCH] mm: vmscan: use per-zone watermark when determine file_is_tiny

[PATCH] mm: vmscan: use per-zone watermark when determine file_is_tiny

[Liu Shixin]

When setting swapiness to 0, the anon pages should be reclaimed if and
only if the value of file_is_tiny is true.

__zone_watermark_ok uses per-zone watermark and lowmem_reserve to determine
whether allocating page from the zone. In the mean time, file_is_tiny is
calculated by per-node watermark. There are inconsistencies between the two
scenarios.

If total free pages on node is enough, then file_is_tiny can not be true,
so the anon pages can not be reclaimed. If the free pages in each zone is
less than watermark + lowmem_reserve, then the allocation will failed too.
Due to lowmem_reserve, these two cases can occur at the same time:

 zone_page_state(zone, NR_FREE_PAGES) < watermark + lowmem_reserve
 node_page_state(pgdat, NR_FREE_PAGES) > total_high_wmark

When both are met, there will be many anon pages that can not be reclaimed
because file_is_tiny is false, and in the same time, the allocation failed
because per-zone watermark is not suitable.

Split the condition (file + free <= high_wmark) to per-zone to fix it.

Reported-and-tested-by: Jinjiang Tu <tujinjiang@huawei.com>
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
---
 mm/vmscan.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index e73e2df8828d..f1dc0dbf1cdb 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -3009,21 +3009,23 @@ static void prepare_scan_count(pg_data_t *pgdat, struct scan_control *sc)
 	 * anon pages.  Try to detect this based on file LRU size.
 	 */
 	if (!cgroup_reclaim(sc)) {
-		unsigned long total_high_wmark = 0;
 		unsigned long free, anon;
 		int z;
 
-		free = sum_zone_node_page_state(pgdat->node_id, NR_FREE_PAGES);
-		file = node_page_state(pgdat, NR_ACTIVE_FILE) +
-			   node_page_state(pgdat, NR_INACTIVE_FILE);
-
 		for (z = 0; z < MAX_NR_ZONES; z++) {
 			struct zone *zone = &pgdat->node_zones[z];
 
 			if (!managed_zone(zone))
 				continue;
 
-			total_high_wmark += high_wmark_pages(zone);
+			free = zone_page_state(zone, NR_FREE_PAGES);
+			file = zone_page_state(zone, NR_ZONE_ACTIVE_FILE) +
+				zone_page_state(zone, NR_ZONE_INACTIVE_FILE);
+
+			if (file + free <= high_wmark_pages(zone)) {
+				sc->file_is_tiny = true;
+				break;
+			}
 		}
 
 		/*
@@ -3033,8 +3035,7 @@ static void prepare_scan_count(pg_data_t *pgdat, struct scan_control *sc)
 		 */
 		anon = node_page_state(pgdat, NR_INACTIVE_ANON);
 
-		sc->file_is_tiny =
-			file + free <= total_high_wmark &&
+		sc->file_is_tiny = sc->file_is_tiny &&
 			!(sc->may_deactivate & DEACTIVATE_ANON) &&
 			anon >> sc->priority;
 	}
-- 
2.25.1