Re: CVE-2014-4699: Linux ptrace bug

[John Johansen <john.johansen () canonical com>]

On 07/06/2014 04:20 AM, Solar Designer wrote:
> On Sun, Jul 06, 2014 at 04:06:12AM -0400, John Johansen wrote:
> > On 07/05/2014 08:37 PM, Marc Deslauriers wrote:
> > > On 14-07-05 05:22 PM, Yves-Alexis Perez wrote:
> > > > On sam., 2014-07-05 at 22:25 +0400, Solar Designer wrote:
> > > > > "ptrace,x86: force IRET path after a ptrace_stop()"
> > > > > http://kernel.opensuse.org/cgit/kernel/commit/?h=openSUSE-13.1&id=d1f26676dad578a65c94782f0c2bd00b7aa68f1b
> > > > >
> > > > > "CVE-2014-4699 Kernel: x86_64,ptrace: Enforce RIP <= TASK_SIZE_MAX"
> > > > > https://bugzilla.redhat.com/show_bug.cgi?id=1115927
> > > >
> > > > Hmhm, what are the reasons why the mainline (and opensuse) fix
> > > > (b9cd18de4db3c9ffa7e17b0dc0ca99ed5aa4d43a) is to force using IRET
> > > > instead of SYSRET, while distros like Ubuntu and Redhat seem to ???only???
> > > > make sure RIP is canonical?
> > >
> > > AFAIK, our plan is to switch to the upstream fix for the next kernel updates.
> >
> > yep, Ubuntu went with the original patch, as we where mostly through our process
> > when b9cd18de4db3c9ffa7e17b0dc0ca99ed5aa4d43a hit. We decided to do a release with
> > the original patch so we could get something out this weekend, but will switch to
> > b9cd18de4db3c9ffa7e17b0dc0ca99ed5aa4d43a asap
>
> Why don't you keep both patches?  It sounds safer to me to use both.
we may, this is still being evaluated. Generally we try to track what goes into
upstream as each addition out of tree patch carries some cost. Every patch that
isn't upstream gets evaluated on a case by case basis.