On Mon, May 13, 2024 at 04:45:54PM -0700, Bill Wendling wrote:
> Prepare for the coming implementation by GCC and Clang of the
> __counted_by attribute. Flexible array members annotated with
> __counted_by can have their accesses bounds-checked at run-time checking
> via CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE
> (for strcpy/memcpy-family functions).
>
> Cc: Mikulas Patocka <miku...@artax.karlin.mff.cuni.cz>
> Cc: Kees Cook <keesc...@chromium.org>
> Cc: "Gustavo A. R. Silva" <gustavo...@kernel.org>
> Cc: Nathan Chancellor <nat...@kernel.org>
> Cc: Nick Desaulniers <ndesaulni...@google.com>
> Cc: Justin Stitt <justinst...@google.com>
> Cc: linux-ker...@vger.kernel.org
> Cc: linux-hardening@vger.kernel.org
> Cc: l...@lists.linux.dev
> Signed-off-by: Bill Wendling <mo...@google.com>
> ---
> fs/hpfs/hpfs.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/hpfs/hpfs.h b/fs/hpfs/hpfs.h
> index 281dec8f636b..ac137787f1f9 100644
> --- a/fs/hpfs/hpfs.h
> +++ b/fs/hpfs/hpfs.h
> @@ -357,7 +357,7 @@ struct hpfs_dirent {
> u8 ix; /* code page index (of filename), see
> struct code_page_data */
> u8 namelen; /* file name length */
> - u8 name[]; /* file name */
> + u8 name[] __counted_by(namelen); /* file name */
> /* dnode_secno down; btree down pointer, if present,
> follows name on next word boundary, or maybe it
> precedes next dirent, which is on a word boundary. */
Looking through struct hpfs_dirent::name uses, I think everything checks
out. I do see some confusing manipulations in hpfs_add_de(), though. I
*think* it'll be okay, though.
Reviewed-by: Kees Cook <keesc...@chromium.org>
--
Kees Cook
