Comment 0 for bug 1315052

When using the cgroup manager, if an unprivileged user starts a container from one login session, then ssh's back in and tries lxc-attach, that will fail.

The workaround for this is simply to start a container under screen or tmux, then re-attach to that session to lxc-attach.

The proper fix is to use MovePidAbs in lxc-attach to move the current task to the 'full' (relative to proxy) cgroup of the container. This requires a new GetPidCgroupAbs method (which is in cgmanager utopic but not yet in trusty) to find out the proper cgroup to attach to.