Re: linux-next: Tree for June 23

[Pekka J Enberg <penberg@cs.helsinki.fi>]

On Mon, 23 Jun 2008, Vegard Nossum wrote:
> Okay, I can confirm that this happens with today's tree too. I also
> get this (new) warning:
> 
> mm/slub.c: In function 'kmem_cache_defrag':
> mm/slub.c:3046: warning: 'reclaimed' may be used uninitialized in this function
> 
> Otherwise things seem to work good :-)

Christoph, while the warning itself seems harmless, I think we need to do 
something like the following; otherwise the return value for 
kmem_cache_defrag() will be incorrect on NUMA machines that have more than 
one N_NORMAL_MEMORY nodes.

		Pekka

diff --git a/mm/slub.c b/mm/slub.c
index deb8153..7707550 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3043,7 +3043,7 @@ int kmem_cache_defrag(int node)
 		return 0;
 
 	list_for_each_entry(s, &slab_caches, list) {
-		unsigned long reclaimed;
+		unsigned long reclaimed = 0;
 
 		/*
 		 * Defragmentable caches come first. If the slab cache is not
@@ -3056,7 +3056,7 @@ int kmem_cache_defrag(int node)
 			int nid;
 
 			for_each_node_state(nid, N_NORMAL_MEMORY)
-				reclaimed = __kmem_cache_shrink(s, nid,
+				reclaimed += __kmem_cache_shrink(s, nid,
 								MAX_PARTIAL);
 		} else
 			reclaimed = __kmem_cache_shrink(s, node, MAX_PARTIAL);