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

[Jann Horn <jann@thejh.net>]

[-- Attachment #1: Type: text/plain, Size: 1473 bytes --]

On Tue, Jun 02, 2015 at 10:02:10PM +0300, Pavel Emelyanov wrote:
> 
> >> +int suspend_seccomp(struct task_struct *task)
> >> +{
> >> +	int ret = -EACCES;
> >> +
> >> +	spin_lock_irq(&task->sighand->siglock);
> >> +
> >> +	if (!capable(CAP_SYS_ADMIN))
> >> +		goto out;
> > 
> > I am puzzled ;) Why do we need ->siglock? And even if we need it, why
> > we can't check CAP_SYS_ADMIN lockless?
> > 
> > And I am not sure I understand why do we need the additional security
> > check, but I leave this to you and Andy.
> > 
> > If you have the rights to trace this task, then you can do anything
> > the tracee could do without the filtering.
> 
> I think _this_ check is required, otherwise the seccomp-ed task (in
> filtered mode) fork-s a child, then this child ptrace-attach to parent
> (allowed) then suspend its seccomd. And -- we have unpriviledged process 
> de-seccomped.

If you can ptrace(), you can already escape from seccomp. See this
section in man 2 seccomp, in the SECCOMP_RET_TRACE section:

              The seccomp check will not be run again after the tracer is
              notified.  (This means that seccomp-based sandboxes must not
              allow use of ptrace(2)—even of other sandboxed processes—
              without extreme care; ptracers can use this mechanism to
              escape from the seccomp sandbox.)

(But I think there have been discussions about changing that behavior in
the future?)

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]