[libvirt] [RFC][PATCH] lxc: drop CAP_SYS_BOOT capability to preventrebooting from inside containers

[Serge E. Hallyn]

Quoting Ryota Ozaki (ozaki.ryota at gmail.com):
> Hi Serge,
> 
> On Fri, May 8, 2009 at 11:04 AM, Serge E. Hallyn <serue at us.ibm.com> wrote:
> > Quoting Ryota Ozaki (ozaki.ryota at gmail.com):
> >> Hi Serge,
> >>
> >> On Fri, May 8, 2009 at 9:12 AM, Serge E. Hallyn <serue at us.ibm.com> wrote:
> >> > Quoting Ryota Ozaki (ozaki.ryota at gmail.com):
> >> >> Hi,
> >
> > ...
> >
> >> >> +    for (i = 0 ; i < ARRAY_CARDINALITY(caps) ; i++) {
> >> >> +        if (prctl(PR_CAPBSET_DROP, caps[i].id, 0, 0, 0)) {
> >> >> +            lxcError(NULL, NULL, VIR_ERR_INTERNAL_ERROR,
> >> >> +                     "%s", _("failed to drop %s"), caps[i].name);
> >> >> +            return -1;
> >> >
> >> > Ideally you should also drop it from pI.
> >>
> >> If not drop it, a user in a container could set CAP_SYS_BOOT fI bit of
> >> /bin/reboot on and then the user could gain CAP_SYS_BOOT back through
> >> the fI. Is this understanding right?
> >
> > Yup.
> >
> > Of course most tasks run with pI empty, so it seems unlikely that
> > it would be a problem, but unless the libcap dependecy becomes a
> > problem, it seems worth making sure that doesn't happen.
> 
> Oh, I slightly misread your suggestions, sorry. You are suggesting making
> sure requires dropping a capability in both bounding set AND pI of a process
> and to do so we need an additional package (libcap2 or somewhat) because
> prctl(2) doesn't have the function to drop pI, aren't you?

Yes.

> um, I hope my patch is sufficient as a first step, but ok, I'll try to implement
> the function to drop pI as well and confirm whether it is feasible for libvirt.

Yes, there is nothing wrong with applying your patch as is for now.

thanks,
-serge