slab, kmemleak: stop calling kmemleak_erase() unconditionally
authorJ. R. Okajima <hooanon05@yahoo.co.jp>
Wed, 2 Dec 2009 07:55:49 +0000 (16:55 +0900)
committerPekka Enberg <penberg@cs.helsinki.fi>
Sun, 6 Dec 2009 08:23:05 +0000 (10:23 +0200)
When the gotten object is NULL (probably due to ENOMEM), kmemleak_erase() is
unnecessary here, It just sets NULL to where already is NULL.  Add a condition.

Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
mm/slab.c

index 7dfa481..4e61449 100644 (file)
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -3109,7 +3109,8 @@ static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
         * per-CPU caches is leaked, we need to make sure kmemleak doesn't
         * treat the array pointers as a reference to the object.
         */
-       kmemleak_erase(&ac->entry[ac->avail]);
+       if (objp)
+               kmemleak_erase(&ac->entry[ac->avail]);
        return objp;
 }