On Sun, Aug 25, 2019 at 04:18:56PM -0700, Kees Cook wrote:
> As we've done with VFS, string operations, etc, reject usercopy sizes
> larger than INT_MAX, which would be nice to have for catching bugs
> related to size calculation overflows[1].
> 
> This adds 10 bytes to x86_64 defconfig text and 1980 bytes to the data
> section:
> 
>    text    data     bss     dec     hex filename
> 19691167        5134320 1646664 26472151        193eed7 vmlinux.before
> 19691177        5136300 1646664 26474141        193f69d vmlinux.after
> 
> [1] https://marc.info/?l=linux-s390&m=156631939010493&w=2
> 
> Suggested-by: Dan Carpenter <dan.carpen...@oracle.com>
> Signed-off-by: Kees Cook <keesc...@chromium.org>

ping! Andrew, can you take this?

-Kees

> ---
>  include/linux/thread_info.h | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h
> index 659a4400517b..e93e249a4e9b 100644
> --- a/include/linux/thread_info.h
> +++ b/include/linux/thread_info.h
> @@ -147,6 +147,8 @@ check_copy_size(const void *addr, size_t bytes, bool 
> is_source)
>                       __bad_copy_to();
>               return false;
>       }
> +     if (WARN_ON_ONCE(bytes > INT_MAX))
> +             return false;
>       check_object_size(addr, bytes, is_source);
>       return true;
>  }
> -- 
> 2.17.1
> 
> 
> -- 
> Kees Cook

-- 
Kees Cook

Reply via email to