Re: [PATCH v3] mm: slub: avoid deref of free pointer in sanity checks if object is invalid

[Harry Yoo <harry.yoo@oracle.com>]

On Tue, Jul 29, 2025 at 04:14:55PM +0800, Li Qiong wrote:
> For debugging, object_err() prints free pointer of the object.
> However, if check_valid_pointer() returns false for a object,
> dereferncing `object + s->offset` can lead to a crash. Therefore,
> print the object's address in such cases.

As the code changed a bit, I think the commit message could better reflect
what this patch actually does.

> Fixes: bb192ed9aa71 ("mm/slub: Convert most struct page to struct slab by spatch")

As Vlastimil mentioned in previous version, this is not the first commit
that introduced this problem.

> Cc: <stable@vger.kernel.org>
> Signed-off-by: Li Qiong <liqiong@nfschina.com>
> ---
> v2:
> - rephrase the commit message, add comment for object_err().
> v3:
> - check object pointer in object_err().
> ---
>  mm/slub.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/slub.c b/mm/slub.c
> index 31e11ef256f9..d3abae5a2193 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -1104,7 +1104,11 @@ static void object_err(struct kmem_cache *s, struct slab *slab,
>  		return;
>  
>  	slab_bug(s, reason);
> -	print_trailer(s, slab, object);
> +	if (!check_valid_pointer(s, slab, object)) {
> +		print_slab_info(slab);
> +		pr_err("invalid object 0x%p\n", object);

Can we just handle this inside print_trailer() because that's the function
that prints the object's free pointer, metadata, etc.?

Also, the message should start with a capital letter.

and "invalid object" sounds misleading because it's the pointer
that is invalid. Perhaps simply "Invalid pointer 0x%p\n"?
(What would be the most comprehensive message here? :P)

> +	} else
> +		print_trailer(s, slab, object);

>  	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
>  
>  	WARN_ON(1);
> @@ -1587,7 +1591,7 @@ 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);
>  		return 0;

Do we really need this hunk after making object_err() resiliant
against wild pointers?

>  	}

-- 
Cheers,
Harry / Hyeonggon