If mlock cannot catch page in LRU then it isn't moved into unevictable lru.
These pages are 'culled' by reclaimer and moved into unevictable lru.
It seems pages locked with MLOCK_ONFAULT always go through this path.
Reclaimer calls try_to_unmap for already isolated pages, thus on this path
we could freely change page to owner of any mlock vma we found in rmap.
This patch passes flag TTU_LRU_ISOLATED into try_to_ummap to move pages
in mmlock_vma_page().
Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>
---
mm/vmscan.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/mm/vmscan.c b/mm/vmscan.c
index bf7a05e8a717..2060f254dd6b 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1345,7 +1345,8 @@ static unsigned long shrink_page_list(struct list_head *page_list,
* processes. Try to unmap it here.
*/
if (page_mapped(page)) {
- enum ttu_flags flags = ttu_flags | TTU_BATCH_FLUSH;
+ enum ttu_flags flags = ttu_flags | TTU_BATCH_FLUSH |
+ TTU_LRU_ISOLATED;
if (unlikely(PageTransHuge(page)))
flags |= TTU_SPLIT_HUGE_PMD;
