Re: CVE-2016-7097 causes acl leak

[Jan Kara]

On Sun 11-12-16 16:34:31, Cong Wang wrote:
> On Mon, Dec 5, 2016 at 9:16 AM, Mark Salyzyn <salyzyn@xxxxxxxxxxx> wrote:
> > Commit 073931017b49d9458aa351605b43a7e34598caef has several occurrences of
> > an acl leak.
> >
> > posix_acl_update_mode(inose, &mode, &acl);
> >
> > . . .
> >
> > posix_acl_release(acl);
> >
> >
> > acl is NULLed in posix_acl_update_mode to signal caller to not update the
> > acl; but because it is nulled, it is never released.
> 
> I think you blame the wrong commit, this leak exists before that commit.
> Looks like we should just release it before NULL'ing.

So I agree with you the mentioned commit didn't change anything. I took
care to keep the previous behavior wrt NULLing the acl pointer (obviously I
could have made mistake somewhere but I don't see where). However your
patch is definitely wrong. See e.g. fs/ext2/acl.c: ext2_set_acl() - there we
really want to just clear the pointer. We release the ACL in the caller of
ext2_set_acl().

Mark why do you think we are leaking ACL references and can you be more
specific where exactly it happens?

								Honza

> diff --git a/fs/posix_acl.c b/fs/posix_acl.c
> index 5955220..edd862a 100644
> --- a/fs/posix_acl.c
> +++ b/fs/posix_acl.c
> @@ -648,8 +648,10 @@ int posix_acl_update_mode(struct inode *inode,
> umode_t *mode_p,
>         error = posix_acl_equiv_mode(*acl, &mode);
>         if (error < 0)
>                 return error;
> -       if (error == 0)
> +       if (error == 0) {
> +               posix_acl_release(*acl);
>                 *acl = NULL;
> +       }
>         if (!in_group_p(inode->i_gid) &&
>             !capable_wrt_inode_uidgid(inode, CAP_FSETID))
>                 mode &= ~S_ISGID;
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR