On Mon, Aug 1, 2016 at 4:55 PM, Andrey Ryabinin [off-list ref] wrote:

On 07/28/2016 06:31 PM, Alexander Potapenko wrote:

quoted

When free_meta_offset is not zero, it is usually aligned on 4 bytes,
because the size of preceding kasan_alloc_meta is aligned on 4 bytes.
As a result, accesses to kasan_free_meta fields may be misaligned.

Signed-off-by: Alexander Potapenko <glider@google.com>
---
 mm/kasan/kasan.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/kasan/kasan.c b/mm/kasan/kasan.c
index 6845f92..0379551 100644
--- a/mm/kasan/kasan.c
+++ b/mm/kasan/kasan.c

@@ -390,7 +390,8 @@ void kasan_cache_create(struct kmem_cache *cache, size_t *size,
      /* Add free meta. */
      if (cache->flags & SLAB_DESTROY_BY_RCU || cache->ctor ||
          cache->object_size < sizeof(struct kasan_free_meta)) {
-             cache->kasan_info.free_meta_offset = *size;
+             cache->kasan_info.free_meta_offset =
+                     ALIGN(*size, sizeof(void *));

This cannot work.

Well, it does, at least on my tests.

I slightly changed metadata layout in http://lkml.kernel.org/g/[ref]
which should also fix UBSAN's complain.

Thanks, will take a look.

quoted

              *size += sizeof(struct kasan_free_meta);
      }
      redzone_adjust = optimal_redzone(cache->object_size) -

-- 
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Matthew Scott Sucherman, Paul Terence Manicle
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

--
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>