Re: [PATCH for-5.1 2/3] virtiofsd: add container-friendly -o chroot sandboxing option

[Stefan Hajnoczi <stefanha@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1357 bytes --]

On Thu, Jul 23, 2020 at 06:55:38PM +0100, Dr. David Alan Gilbert wrote:
> * Stefan Hajnoczi (stefanha@redhat.com) wrote:
> > On Wed, Jul 22, 2020 at 08:03:18PM +0100, Dr. David Alan Gilbert wrote:
> > > * Stefan Hajnoczi (stefanha@redhat.com) wrote:
> > > > +    /*
> > > > +     * Make the shared directory the file system root so that FUSE_OPEN
> > > > +     * (lo_open()) cannot escape the shared directory by opening a symlink.
> > > > +     *
> > > > +     * It's still possible to escape the chroot via lo->proc_self_fd but that
> > > > +     * requires gaining control of the process first.
> > > > +     */
> > > > +    if (chroot(lo->source) != 0) {
> > > > +        fuse_log(FUSE_LOG_ERR, "chroot(\"%s\"): %m\n", lo->source);
> > > > +        exit(1);
> > > > +    }
> > > 
> > > I'm seeing suggestions that you should drop CAP_SYS_CHROOT after
> > > chroot'ing to stop an old escape (where you create another jail inside
> > > the jail and the kernel then lets you walk outside of the old one).
> > 
> > That's already the case:
> > 
> > 1. setup_seccomp() blocks further chroot(2) calls.
> > 2. setup_capabilities() drops CAP_SYS_CHROOT.
> 
> Ah yes; could you add a comment if you respin; it's not obvious that
> the capability to chroot allows you to break out of an existing chroot
> you're in.

Sure.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]