Re: [ 07/48] mm: fix slab->page _count corruption when using slub

[Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>]

On Mon, Jul 02, 2012 at 08:56:32PM -0300, Herton Ronaldo Krzesinski wrote:
> So, 3.2 didn't have the CONFIG_HAVE_CMPXCHG_DOUBLE and
> CONFIG_HAVE_ALIGNED_STRUCT_PAGE defines, it had instead CMPXCHG_DOUBLE
> and HAVE_ALIGNED_STRUCT_PAGE is a new logic. So the patch is broken for
> 3.2 indeed.

So here is a proposed backport for 3.2, I tested it etc., but please
someone review/ack it as well...

From: Pravin B Shelar <pshelar@nicira.com>
Subject: mm: fix slab->page _count corruption when using slub

commit abca7c4965845924f65d40e0aa1092bdd895e314 upstream.

On arches that do not support this_cpu_cmpxchg_double() slab_lock is used
to do atomic cmpxchg() on double word which contains page->_count.  The
page count can be changed from get_page() or put_page() without taking
slab_lock.  That corrupts page counter.

Fix it by moving page->_count out of cmpxchg_double data.  So that slub
does no change it while updating slub meta-data in struct page.

[akpm@linux-foundation.org: use standard comment layout, tweak comment text]
Reported-by: Amey Bhide <abhide@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[herton: backport to 3.2, use CONFIG_CMPXCHG_DOUBLE]
Signed-off-by: Herton R. Krzesinski <herton@canonical.com>
---
 include/linux/mm_types.h |    9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 5b42f1b..2153639 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -56,8 +56,17 @@ struct page {
 		};
 
 		union {
+#ifdef CONFIG_CMPXCHG_DOUBLE
 			/* Used for cmpxchg_double in slub */
 			unsigned long counters;
+#else
+			/*
+			 * Keep _count separate from slub cmpxchg_double data.
+			 * As the rest of the double word is protected by
+			 * slab_lock but _count is not.
+			 */
+			unsigned counters;
+#endif
 
 			struct {
 
-- 
1.7.9.5


-- 
[]'s
Herton