Re: [PATCH v3 1/5] slab: align ZERO_SIZE_PTR to ARCH_KMALLOC_MINALIGN

[David Laight <david.laight.linux@gmail.com>]

On Thu,  3 Sep 2026 22:37:16 +0200
Karl Mehltretter <kmehltretter@gmail.com> wrote:

> The kmalloc entry points are annotated with __assume_kmalloc_alignment
> but return ZERO_SIZE_PTR, currently (void *)16, for zero-size requests.
> This violates the annotation when ARCH_KMALLOC_MINALIGN exceeds 16.
> 
> This can mislead compiler optimizations. Clang's UBSAN_ALIGNMENT detects
> the violation on armv5. GCC and Clang retain the ZERO_OR_NULL_PTR() range
> check but eliminate an exact ZERO_SIZE_PTR comparison after an annotated
> allocation.
> 
> Define ZERO_SIZE_PTR as the greater of 16 and ARCH_KMALLOC_MINALIGN,
> retaining the existing value where it is already aligned. Assert that
> ARCH_KMALLOC_MINALIGN remains below 0x100, the value of LIST_POISON1
> when POISON_POINTER_DELTA is zero, so the sentinel remains distinct
> from that poison pointer.
> 
> Fixes: 94a58c360a45 ("slab.h: sprinkle __assume_aligned attributes")
> Assisted-by: LLM
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
> ---
>  include/linux/slab.h | 12 +++++++++++-
>  1 file changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/slab.h b/include/linux/slab.h
> index cda126def67a..563dadc16d82 100644
> --- a/include/linux/slab.h
> +++ b/include/linux/slab.h
> @@ -262,13 +262,16 @@ enum _slab_flag_bits {
>  
>  /*
>   * ZERO_SIZE_PTR will be returned for zero sized kmalloc requests.
> + * It satisfies the alignment promised by __assume_kmalloc_alignment
> + * and keeps the historic value 16 where that is already aligned.
>   *
>   * Dereferencing ZERO_SIZE_PTR will lead to a distinct access fault.
>   *
>   * ZERO_SIZE_PTR can be passed to kfree though in the same way that NULL can.
>   * Both make kfree a no-op.
>   */
> -#define ZERO_SIZE_PTR ((void *)16)
> +#define ZERO_SIZE_PTR ((void *)(ARCH_KMALLOC_MINALIGN > 16 ? \
> +				ARCH_KMALLOC_MINALIGN : 16))

If ARCH_KMALLOC_MINALIGN is just a constant (I suspect it has to be)
this would be better as:
#if ARCH_KMALLOC_MINALIGN > 16
#define ZERO_SIZE_PTR ((void *)ARCH_KMALLOC_MINALIGN)
#else
#define ARCH_KMALLOC_MINALIGN ((void *)16)
#endif
to avoid bloat at all the expansions.

David

>  
>  #define ZERO_OR_NULL_PTR(x) ((unsigned long)(x) <= \
>  				(unsigned long)ZERO_SIZE_PTR)
> @@ -625,6 +628,13 @@ static inline bool kmem_dump_obj(void *object) { return false; }
>  #define KMALLOC_SHIFT_LOW ilog2(KMALLOC_MIN_SIZE)
>  #endif
>  
> +/*
> + * Keep ZERO_SIZE_PTR at most 128, i.e. below 0x100: LIST_POISON1 is
> + * 0x100 when POISON_POINTER_DELTA is 0, and no architecture currently
> + * has an ARCH_KMALLOC_MINALIGN above 128.
> + */
> +static_assert(ARCH_KMALLOC_MINALIGN < 0x100);
> +
>  /*
>   * Setting ARCH_SLAB_MINALIGN in arch headers allows a different alignment.
>   * Intended for arches that get misalignment faults even for 64 bit integer