Re: [libvirt] [PATCH 9/9] qemu_cgroup: assert threading cgroup layout for machine cgroup
On Fri, 26 Feb 2016 12:02:52 +0000
"Daniel P. Berrange" <berrange(a)redhat.com> wrote:
On Fri, Feb 26, 2016 at 12:57:38PM +0100, Henning Schild wrote:
> On Fri, 26 Feb 2016 11:22:07 +0000
> "Daniel P. Berrange" <berrange(a)redhat.com> wrote:
>
> > On Tue, Feb 23, 2016 at 04:58:44PM +0100, Henning Schild wrote:
> > > Make sure the thread related controls of the machine cgroup
> > > never get any tasks assigned.
> > >
> > > Signed-off-by: Henning Schild <henning.schild(a)siemens.com>
> > > ---
> > > src/qemu/qemu_cgroup.c | 14 ++++++++++++--
> > > 1 file changed, 12 insertions(+), 2 deletions(-)
> >
> > NACK This also won't work for same reason as previous patch
>
> Having that in place can still be useful after we have sorted out
> the random result of what systemd gave us.
> Is the general idea of such an assertion a good idea, and should i
> adopt it according to comments?
> At the moment i just used the assertion mask in the only code-path
> that adds tasks within libvirt. If we have to deal with
> manipulation from the outside, it might be a good idea to introduce
> more assertions based on the mask.
Without patch 4 though, there's nowhere you can put this afaict.
After moving the pid to the emulator cgroup, i can assert that the
parent is now empty and then put it in place. It would assert libvirt
itself does not use the parent group somewhen in the future.
Regards,
Daniel