On Mon, Sep 26, 2022 at 03:50:43PM +0200, Vlastimil Babka wrote:

On 9/23/22 22:28, Kees Cook wrote:

quoted

Round up allocations with kmalloc_size_roundup() so that mempool's use
of ksize() is always accurate and no special handling of the memory is
needed by KASAN, UBSAN_BOUNDS, nor FORTIFY_SOURCE.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
Signed-off-by: Kees Cook <redacted>
---
  mm/mempool.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/mempool.c b/mm/mempool.c
index 96488b13a1ef..0f3107b28e6b 100644
--- a/mm/mempool.c
+++ b/mm/mempool.c

@@ -526,7 +526,7 @@ EXPORT_SYMBOL(mempool_free_slab);
   */
  void *mempool_kmalloc(gfp_t gfp_mask, void *pool_data)
  {
-	size_t size = (size_t)pool_data;
+	size_t size = kmalloc_size_roundup((size_t)pool_data);

Hm it is kinda wasteful to call into kmalloc_size_roundup for every
allocation that has the same input. We could do it just once in
mempool_init_node() for adjusting pool->pool_data ?

But looking more closely, I wonder why poison_element() and
kasan_unpoison_element() in mm/mempool.c even have to use ksize()/__ksize()
and not just operate on the requested size (again, pool->pool_data). If no
kmalloc mempool's users use ksize() to write beyond requested size, then we
don't have to unpoison/poison that area either?

Yeah, I think that's a fair point. I will adjust this.

-- 
Kees Cook