Re: [PATCH v2] seccomp: add ptrace options for suspend/resume

[Kees Cook]

On Tue, Jun 9, 2015 at 2:52 PM, Tycho Andersen
<tycho.andersen@xxxxxxxxxxxxx> wrote:
> On Tue, Jun 09, 2015 at 02:45:49PM -0700, Kees Cook wrote:
>> On Tue, Jun 9, 2015 at 2:22 PM, Tycho Andersen
>> <tycho.andersen@xxxxxxxxxxxxx> wrote:
>> > Hi Kees, Andy,
>> >
>> > On Fri, Jun 05, 2015 at 11:16:50PM +0200, Oleg Nesterov wrote:
>> >> Hi Tycho,
>> >>
>> >> On 06/04, Tycho Andersen wrote:
>> >> > > > +#ifdef CONFIG_CHECKPOINT_RESTORE
>> >> > > > +bool may_suspend_seccomp(void)
>> >> > > > +{
>> >> > > > +       if (!capable(CAP_SYS_ADMIN))
>> >> > > > +               return false;
>> >> > > > +
>> >> > > > +       if (current->seccomp.mode != SECCOMP_MODE_DISABLED)
>> >> > > > +               return false;
>> >> > >
>> >> > > Heh. OK, I won't argue with the new check too ;)
>> >> >
>> >> > Actually now that I think about it I agree with you, these checks
>> >> > don't seem necessary. Even inside a user namespace, if you can ptrace
>> >> > a process you can make it do whatever you want irrespective of
>> >> > seccomp, as long as it has the necessary capabilities. Once the
>> >> > seccomp checks are run after ptrace, they'll be enforced so you
>> >> > couldn't have it call whatever you want in the first place.
>> >>
>> >> Good ;)
>> >>
>> >> > Still, perhaps I'm missing something...
>> >>
>> >> Kees, Andy?
>> >
>> > Any thoughts on removing may_suspend_seccomp() all together?
>>
>> As in, just open-code the check? That would be fine by me.
>
> Sorry, I meant getting rid of any checks entirely. Using my argument
> above I've managed to convince myself they don't add any value. You
> guys know a lot more about this than I do, though.

Well, as things stand currently, yes, that check would be redundant.
The fact that ptrace can be used to bypass seccomp is kind of an
accident, though. The design for ptrace-based seccomp managers was
that the manager would do the work, rather than rewriting the syscall
on behalf of the child. I don't think anything actually uses this
effect. It's something we've wanted to fix, though a clean solution
isn't obvious. As a result, I'm cautious to add this behavior in such
a wide open fashion. For now, I'd like to limit the scope of this to
CAP_SYS_ADMIN.

I do think dropping the seccomp.mode check is fine -- this would mean
you could set this flag before the child even added seccomp filters.
So, instead of the function call, maybe just add the capable() call?

-Kees

-- 
Kees Cook
Chrome OS Security
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/