Date: Mon, 8 Jan 2018 18:50:28 +0100
From: Ingo Molnar <mingo@kernel.org>
To: Dave Hansen <dave.hansen@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>, Willy Tarreau <w@1wt.eu>,
        linux-kernel@vger.kernel.org, x86@kernel.org,
        gnomes@lxorguk.ukuu.org.uk, torvalds@linux-foundation.org
Subject: Re: [PATCH RFC 3/4] x86/pti: don't mark the user PGD with _PAGE_NX.
Message-ID: <20180108175028.acwe3glhw4rsvdsx@gmail.com>
References: <1515427939-10999-1-git-send-email-w@1wt.eu>
 <1515427939-10999-4-git-send-email-w@1wt.eu>
 <alpine.DEB.2.20.1801081803290.1735@nanos>
 <760b7264-1ae7-bcaa-6d20-f47cc7c7fce1@intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <760b7264-1ae7-bcaa-6d20-f47cc7c7fce1@intel.com>
User-Agent: NeoMutt/20170609 (1.8.3)
Sender: linux-kernel-owner@vger.kernel.org


* Dave Hansen <dave.hansen@intel.com> wrote:

> On 01/08/2018 09:05 AM, Thomas Gleixner wrote:
> > On Mon, 8 Jan 2018, Willy Tarreau wrote:
> >> Since we're going to keep running on the same PGD when returning to
> >> userspace for certain performance-critical tasks, we'll need the user
> >> pages to be executable. So this code disables the extra protection
> >> that was added consisting in marking user pages _PAGE_NX so that this
> >> pgd remains usable for userspace.
> >>
> >> Note: it isn't necessarily the best approach, but one way or another
> >>       if we want to be able to return to userspace from the kernel,
> >>       we'll have to have this executable anyway. Another approach
> >>       might consist in using another pgd for userland+kernel but
> >>       the current core really looks like an extra careful measure
> >>       to catch early bugs if any.
> > 
> > I surely want to keep that as a safety measure. The entry code is simple to
> > get wrong and running with the wrong pagetables by a silly mistake and
> > thereby undoing the protection is surely not what we want.
> > 
> > Need to find a free time slot to think about that.
> 
> This does get immensely easier if we choose a mode at exec() (or fork()
> even) and never change it.  The prctl() _could_ just be a flag to tell
> what your children should do.

Switching PTI on/off for a whole process would be nightmarish.

The simplest model is indeed child inheritance tree propagation - plus perhaps the 
ability for a thread to change its *own* PTI status, which obviously doesn't 
create any deep "process lookup" or cross-CPU complications.

( Note that here I only mean "simple to implement" - we might decide to not offer
  the ABI. )

Thanks,

	Ingo