[lxc-devel] [PATCH] cgroup: re-introduce ns cgroup support

[Serge Hallyn]

Quoting Christian Seiler (christian at iwakd.de):
> Hi again,
> 
> I was just looking at how to best implement the cgroup mount hook.
> Problem now is that the easiest way would be just to create the cgroup
> before the clone() (enter is done afterwards anyway) so that the client
> has access to handler->pid. Unfortunately, handler->pid is needed for
> the ns cgroup renaming scheme you just added.
> 
> Is it OK if I split the function? I.e. have lxc_cgroup_create ignore ns

I'm sorry, I thought I had replied to this last night.  I'm certain I
wrote a response, dunno what happened to it...

> cgroups and have a separate function for the ns cgroup entries? Then

Makes perfect sense to me to do so, yes.

> lxc_cgroup_create could be called before the clone and the ns handling
> function afterwards. And the cgroup mounting hook would mount only the
> root cgroup for ns cgroups anyway. (Or ignore them.)

> Btw., OT: I'm having trouble with lxc-start -d in the current staging;
> the container starts find in the background, but lxc-start doesn't
> terminate. If I kill it with Ctrl+C, the container continues running
> (because it's already forked away) but the foreground lxc-start seems to
> hang completely... lxc-start without -d behaves normally.

Yeah, I just pushed Dwight's commit which should fix that - thanks.

-serge