On Fri, Jul 25, 2025 at 10:48:54AM +0800, Li Qiong wrote:
In object_err(), need dereference the 'object' pointer, it may cause a invalid pointer fault. Use slab_err() instead.
Hi Li Qiong, this patch makes sense to me. But I'd suggest to rephrase it a little bit, like:
mm/slab: avoid deref of free pointer in sanity checks if object is invalid
For debugging purposes, object_err() prints free pointer of the object. However, if check_valid_pointer() returns false for object, `object + s->offset` is also invalid and dereferncing it can lead to a crash. Therefore, avoid dereferencing it and only print the object's address in such cases.
Signed-off-by: Li Qiong liqiong@nfschina.com
Which commit introduced this problem? A Fixes: tag is needed to determine which -stable versions it should be backported to.
And to backport MM patches to -stable, you need to explicitly add 'Cc: stable@vger.kernel.org' to the patch.
mm/slub.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/mm/slub.c b/mm/slub.c index 31e11ef256f9..3a2e57e2e2d7 100644 --- a/mm/slub.c +++ b/mm/slub.c @@ -1587,7 +1587,7 @@ static inline int alloc_consistency_checks(struct kmem_cache *s, return 0; if (!check_valid_pointer(s, slab, object)) {
object_err(s, slab, object, "Freelist Pointer check fails");
slab_err(s, slab, "Freelist Pointer (0x%p) check fails", object);
Can this be slab_err(s, slab, "Invalid object pointer 0x%p", object); to align with free_consistency_checks()?
return 0;}
It might be worth adding a comment in object_err() stating that it should only be called when check_valid_pointer() returns true for object, and a WARN_ON_ONCE(!check_valid_pointer(s, slab, object)) to catch incorrect usages?