On Fri, May 08, 2026 at 04:21:36PM +0200, Marco Elver wrote:
> I think I have a solution for this mess, see below.
>
> I would not send it as 1 series, but only include the slab changes (+
> instruction_pointer.h change to introduce _CODE_LOCATION_) as one
> series, to go through the slab tree. The rest of the patches would go to
> respective arch maintainers.
I'm assuming this will be a follow-up and reviewing patch 1
(and waiting for Jon's thuoghts on patch 2)
> diff --git a/include/linux/instruction_pointer.h
> b/include/linux/instruction_pointer.h
> index aa0b3ffea935..dfe73aafddb8 100644
> --- a/include/linux/instruction_pointer.h
> +++ b/include/linux/instruction_pointer.h
> @@ -8,6 +8,30 @@
>
> #ifndef _THIS_IP_
> #define _THIS_IP_ ({ __label__ __here; __here: (unsigned long)&&__here; })
> +/*
> + * The current generic definition of _THIS_IP_ is considered broken by GCC
> [1]
> + * and Clang [2]. In particular, the address of a label is only expected to
> be
> + * used with a computed goto.
> + *
> + * [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120071
> + * [2] https://github.com/llvm/llvm-project/issues/138272
> + *
> + * Mark it as broken, so that appropriate fallback options can be implemented
> + * for architectures that do not define their won _THIS_IP_.
> + */
> +#define HAS_BROKEN_THIS_IP
> +#endif
As long as _THIS_IP_ is broken on some arches, it cannot be used anyway
when in a general API that can be used by arbitrary users?
Is it something that can be fixed in all arches over time?
> +/*
> + * _CODE_LOCATION_ provides a unique identifier for the current code
> location.
> + * When _THIS_IP_ is broken (generic version), we fall back to a static
> marker
> + * which guarantees uniqueness and resolves to a constant address at link
> time,
> + * avoiding runtime overhead and compiler optimizations breaking it.
> + */
> +#ifdef HAS_BROKEN_THIS_IP
> +#define _CODE_LOCATION_ ({ static const char __here; (unsigned long)&__here;
> })
Nice!
Yes, we don't really need the exact code location
for partitioning kmalloc caches.
IIRC lockdep does a similar thing to define lock classes (unique for
each lock init location)
> +#else
> +#define _CODE_LOCATION_ _THIS_IP_
> #endif
Probably we don't need this fallback?
> #endif /* _LINUX_INSTRUCTION_POINTER_H */
> diff --git a/include/linux/slab.h b/include/linux/slab.h
> index 5e1249e36b0d..a4bf1585411f 100644
> --- a/include/linux/slab.h
> +++ b/include/linux/slab.h
> @@ -503,7 +503,7 @@ int kmem_cache_shrink(struct kmem_cache *s);
> typedef struct { unsigned long v; } kmalloc_token_t;
> #ifdef CONFIG_KMALLOC_PARTITION_RANDOM
> extern unsigned long random_kmalloc_seed;
> -#define __kmalloc_token(...) ((kmalloc_token_t){ .v = _RET_IP_ })
> +#define __kmalloc_token(...) ((kmalloc_token_t){ .v = _CODE_LOCATION_ })
> #elif defined(CONFIG_KMALLOC_PARTITION_TYPED)
> #define __kmalloc_token(...) ((kmalloc_token_t){ .v =
> __builtin_infer_alloc_token(__VA_ARGS__) })
> #endif
--
Cheers,
Harry / Hyeonggon
