After the patches generalizing rmap_walk pulled from upstream recently,
it is safe to call rmap_walk on anonymous pages, but the stale VM_BUG_ON
is still there. It was moved out to rmap_walk_file in the scope of
mainstream commit 9f32624be9435 ("mm/rmap: use rmap_walk() in
page_referenced()"), but I don't want to pull it, because it's quite
intrusive. Instead, let's just pull the bits needed to fix the issue.

https://jira.sw.ru/browse/PSBM-42214

Signed-off-by: Vladimir Davydov <[email protected]>
---
 mm/rmap.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/mm/rmap.c b/mm/rmap.c
index 3eda1ed12070..c97b42dbcf25 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1810,6 +1810,14 @@ static int rmap_walk_file(struct page *page, struct 
rmap_walk_control *rwc)
        struct vm_area_struct *vma;
        int ret = SWAP_AGAIN;
 
+       /*
+        * The page lock not only makes sure that page->mapping cannot
+        * suddenly be NULLified by truncation, it makes sure that the
+        * structure at mapping cannot be freed and reused yet,
+        * so we can safely take mapping->i_mmap_mutex.
+        */
+       VM_BUG_ON(!PageLocked(page));
+
        if (!mapping)
                return ret;
        mutex_lock(&mapping->i_mmap_mutex);
@@ -1841,8 +1849,6 @@ done:
 
 int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
 {
-       VM_BUG_ON(!PageLocked(page));
-
        if (unlikely(PageKsm(page)))
                return rmap_walk_ksm(page, rwc);
        else if (PageAnon(page))
-- 
2.1.4

_______________________________________________
Devel mailing list
[email protected]
https://lists.openvz.org/mailman/listinfo/devel

Reply via email to