rmap_walk_anon() does not use page_lock_anon_vma() for looking up and
locking an anon_vma. One important difference between page_lock_anon_vma()
and rmap_walk_anon() is that the page_lock_anon_vma() takes the RCU lock
before the lookup so that the anon_vma does not disappear. There does not
appear to be locking in place that prevents the anon_vma disappearing before
the spinlock is taken.

This patch puts a rcu_read_lock() around the anon_vma lookup similar to
what page_lock_anon_vma() does to prevent an accidental use-after-free.

Signed-off-by: Mel Gorman <redacted>
---
 mm/rmap.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/mm/rmap.c b/mm/rmap.c
index b468d5f..fb695d3 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c

@@ -1233,9 +1233,10 @@ static int rmap_walk_anon(struct page *page, int (*rmap_one)(struct page *,
 	 * This needs to be reviewed later: avoiding page_lock_anon_vma()
 	 * is risky, and currently limits the usefulness of rmap_walk().
 	 */
+	rcu_read_lock();
 	anon_vma = page_anon_vma(page);
 	if (!anon_vma)
-		return ret;
+		goto out_rcu_unlock;
 	spin_lock(&anon_vma->lock);
 
 	/*

@@ -1256,6 +1257,10 @@ static int rmap_walk_anon(struct page *page, int (*rmap_one)(struct page *,
 
 out_anon_unlock:
 	spin_unlock(&anon_vma->lock);
+
+out_rcu_unlock:
+	rcu_read_unlock();
+
 	return ret;
 }
 

-- 
1.6.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>