Re: [PATCH hurd 3/5] proc: implement `proc_make_task_namespace'

[Justus Winter]

Quoting David Michael (2014-12-12 23:01:57)
> Hi,
> 
> On Thu, Nov 13, 2014 at 7:26 AM, Justus Winter
> <4winter@informatik.uni-hamburg.de> wrote:
> > @@ -1008,9 +1069,42 @@ S_mach_notify_new_task (mach_port_t notify,
> >        childp = new_proc (task);
> >      }
> >
> > -  /* XXX do something interesting */
> > +  if (MACH_PORT_VALID (parentp->p_task_namespace))
> > +    {
> > +      error_t err;
> > +      /* Tasks in a task namespace are not expected to call
> > +        proc_child, so we do it on their behalf.  */
> > +      mach_port_mod_refs (mach_task_self (), task, MACH_PORT_RIGHT_SEND, 
> > +1);
> > +      err = S_proc_child (parentp, task);
> > +      if (! err)
> > +       /* Relay the notification.  This consumes TASK and PARENT.  */
> > +       return mach_notify_new_task (childp->p_task_namespace, task, 
> > parent);
> > +    }
> >
> >    mach_port_deallocate (mach_task_self (), task);
> >    mach_port_deallocate (mach_task_self (), parent);
> 
> This mach_notify_new_task call in proc/mgt.c seems to be causing a
> linker error.  I added task_notifyUser.o to MIGSTUBS to get around it.
> Is that correct?

Yes, but that issue should be transitional.  It happens whenever we
add RPCs, as the client stubs are distributed in a .so by the libc and
have to be rebuilt.

Justus