[PATCH v5 3/7] fs: Add fd_install_received() wrapper for __fd_install_received()

[Kees Cook]

On Thu, Jun 18, 2020 at 05:49:19AM +0000, Sargun Dhillon wrote:
> On Wed, Jun 17, 2020 at 03:03:23PM -0700, Kees Cook wrote:
> > [...]
> >  static inline int fd_install_received_user(struct file *file, int __user *ufd,
> >  					   unsigned int o_flags)
> >  {
> > +	if (ufd == NULL)
> > +		return -EFAULT;
> Isn't this *technically* a behvaiour change? Nonetheless, I think this is a much better
> approach than forcing everyone to do null checking, and avoids at least one error case
> where the kernel installs FDs for SCM_RIGHTS, and they're not actualy usable.

So, the only behavior change I see is that the order of sanity checks is
changed.

The loop in scm_detach_fds() is:


        for (i = 0; i < fdmax; i++) {
                err = __scm_install_fd(scm->fp->fp[i], cmsg_data + i, o_flags);
                if (err < 0)
                        break;
        }

Before, __scm_install_fd() does:

        error = security_file_receive(file);
        if (error)
                return error;

        new_fd = get_unused_fd_flags(o_flags);
        if (new_fd < 0)
                return new_fd;

        error = put_user(new_fd, ufd);
        if (error) {
                put_unused_fd(new_fd);
                return error;
        }
	...

After, fd_install_received_user() and __fd_install_received() does:

        if (ufd == NULL)
                return -EFAULT;
	...
        error = security_file_receive(file);
        if (error)
                return error;
	...
                new_fd = get_unused_fd_flags(o_flags);
                if (new_fd < 0)
                        return new_fd;
	...
                error = put_user(new_fd, ufd);
                if (error) {
                        put_unused_fd(new_fd);
                        return error;
                }

i.e. if a caller attempts a receive that is rejected by LSM *and*
includes a NULL userpointer destination, they will get an EFAULT now
instead of an EPERM.

I struggle to imagine a situation where this could possible matter
(both fail, neither installs files). It is only the error code that
is different. I am comfortable making this change and seeing if anyone
screams. If they do, I can restore the v4 "ufd_required" way of doing it.

> Reviewed-by: Sargun Dhillon <sargun at sargun.me>

Thanks!

-- 
Kees Cook