Re: [PATCH 2/5] slab: Add __alloc_size attributes for better bounds checking

[Kees Cook <keescook@chromium.org>]

On Tue, Aug 17, 2021 at 10:31:32PM -0700, Joe Perches wrote:
> On Tue, 2021-08-17 at 22:08 -0700, Kees Cook wrote:
> > As already done in GrapheneOS, add the __alloc_size attribute for
> > regular kmalloc interfaces, to provide additional hinting for better
> > bounds checking, assisting CONFIG_FORTIFY_SOURCE and other compiler
> > optimizations.
> []
> > diff --git a/include/linux/slab.h b/include/linux/slab.h
> []
> > @@ -181,7 +181,7 @@ int kmem_cache_shrink(struct kmem_cache *);
> >  /*
> >   * Common kmalloc functions provided by all allocators
> >   */
> > -void * __must_check krealloc(const void *, size_t, gfp_t);
> > +void * __must_check krealloc(const void *, size_t, gfp_t) __alloc_size(2);
> 
> I suggest the __alloc_size attribute be placed at the beginning of the
> function declaration to be more similar to the common __printf attribute
> location uses.

Yeah, any consistent ordering suggestions are welcome here; thank you!
These declarations were all over the place, and trying to follow each
slightly different existing style made my eyes hurt. :)

> __alloc_size(2)
> void * __must_check krealloc(const void *, size_t, gfp_t);
> 
> I really prefer the __must_check to be with the other attribute and that
> function declarations have argument names too like:
> 
> __alloc_size(2) __must_check
> void *krealloc(const void *ptr, size_t size, gfp_t gfp);

I'm happy with whatever makes the most sense.

> but there are a _lot_ of placement of __must_check after the return type
> 
> Lastly __alloc_size should probably be added to checkpatch

Oh, yes! Thanks for the reminder.

> Maybe:
> ---
>  scripts/checkpatch.pl | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> index 161ce7fe5d1e5..1a166b5cf3447 100755
> --- a/scripts/checkpatch.pl
> +++ b/scripts/checkpatch.pl
> @@ -489,7 +489,8 @@ our $Attribute	= qr{
>  			____cacheline_aligned|
>  			____cacheline_aligned_in_smp|
>  			____cacheline_internodealigned_in_smp|
> -			__weak
> +			__weak|
> +			__alloc_size\s*\(\s*\d+\s*(?:,\s*d+\s*){0,5}\)

Why the "{0,5}" bit here? I was expecting just "?". (i.e. it can have
either 1 or 2 arguments.)

>  		  }x;
>  our $Modifier;
>  our $Inline	= qr{inline|__always_inline|noinline|__inline|__inline__};
> 
> 

-- 
Kees Cook