From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1751954AbaG0MIf (ORCPT <rfc822;w@1wt.eu>);
	Sun, 27 Jul 2014 08:08:35 -0400
Received: from mail-vc0-f172.google.com ([209.85.220.172]:40133 "EHLO
	mail-vc0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751788AbaG0MIc (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Sun, 27 Jul 2014 08:08:32 -0400
MIME-Version: 1.0
In-Reply-To: <CALCETrUGUuYzvQtHLABh78V4fZJmwqb1Lqk-P65VUdRyrDWLug@mail.gmail.com>
References: <1406296033-32693-1-git-send-email-drysdale@google.com>
 <1406296033-32693-11-git-send-email-drysdale@google.com> <53D26355.9020809@redhat.com>
 <CALCETrUGUuYzvQtHLABh78V4fZJmwqb1Lqk-P65VUdRyrDWLug@mail.gmail.com>
From: David Drysdale <drysdale@google.com>
Date: Sun, 27 Jul 2014 13:08:11 +0100
Message-ID: <CAHse=S_tCRsM77EKr=M+zP1pLg2y1EJRG-h=RUJau8Fm6RR2cw@mail.gmail.com>
Subject: Re: [PATCH 10/11] capsicum: prctl(2) to force use of O_BENEATH
To: Andy Lutomirski <luto@amacapital.net>
Cc: Paolo Bonzini <pbonzini@redhat.com>, Al Viro <viro@zeniv.linux.org.uk>,
        LSM List <linux-security-module@vger.kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Paul Moore <paul@paul-moore.com>,
        James Morris <james.l.morris@oracle.com>,
        Kees Cook <keescook@chromium.org>,
        Linux API <linux-api@vger.kernel.org>,
        Meredydd Luff <meredydd@senatehouse.org>,
        Christoph Hellwig <hch@infradead.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Fri, Jul 25, 2014 at 5:00 PM, Andy Lutomirski <luto@amacapital.net> wrote:
>
> On Jul 25, 2014 7:02 AM, "Paolo Bonzini" <pbonzini@redhat.com> wrote:
> >
> > Il 25/07/2014 15:47, David Drysdale ha scritto:
> > > @@ -1996,6 +2013,17 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> > >               if (arg2 || arg3 || arg4 || arg5)
> > >                       return -EINVAL;
> > >               return current->no_new_privs ? 1 : 0;
> > > +     case PR_SET_OPENAT_BENEATH:
> > > +             if (arg2 != 1 || arg4 || arg5)
> > > +                     return -EINVAL;
> > > +             if ((arg3 & ~(PR_SET_OPENAT_BENEATH_TSYNC)) != 0)
> > > +                     return -EINVAL;
> > > +             error = prctl_set_openat_beneath(me, arg3);
> > > +             break;
> > > +     case PR_GET_OPENAT_BENEATH:
> > > +             if (arg2 || arg3 || arg4 || arg5)
> > > +                     return -EINVAL;
> > > +             return me->openat_beneath;
> > >       case PR_GET_THP_DISABLE:
> > >               if (arg2 || arg3 || arg4 || arg5)
> > >                       return -EINVAL;
> > >
> >
> > Why are you always forbidding a change of prctl from 1 to 0?  It should
> > be safe if current->no_new_privs is clear.
>
> I don't immediately see why you're forbidding unsettling it at all.
> If you need it to be sticky, then use seccomp or Capsicum to make it
> sticky.

Good point, that would make the function more generic -- needing to
latch is specific to Capsicum's use of it.

>
> Also, the way implementation is dangerously racy -- if anyone pokes at
> adjacent bitfields without the lock, they can get corrupted.  Try
> basing on Kees' seccomp tree or security-next and using the new atomic
> flags field.

Ah yes, sorry -- I hadn't yet shifted the implementation to line up with
the work you and Kees have put into the seccomp stuff.

>
>
> --Andy
>
> >
> > Do new threads inherit from the parent?
> >
> > Also, I wonder if you need something like this check:
> >
> >         /*
> >          * Installing a seccomp filter requires that the task has
> >          * CAP_SYS_ADMIN in its namespace or be running with no_new_privs.
> >          * This avoids scenarios where unprivileged tasks can affect the
> >          * behavior of privileged children.
> >          */
> >         if (!current->no_new_privs &&
> >             security_capable_noaudit(current_cred(), current_user_ns(),
> >                                      CAP_SYS_ADMIN) != 0)
> >                 return -EACCES;
> >
> > Paolo

Yes, new threads inherit the flag from the parent so the
NNP||CAP_SYS_ADMIN check is probably needed.