Currently, the lxc package ships a "bare-bones" toolset and contains some inaccuracies. I think this can be improved in several ways.

1. Add a systemd service template (lxc@.service). The attached file is taken from the archwiki LXC page except that it does not contain ExecStartPre which remounts / rprivate. The latter is not needed anymore and lxc starts OK without it. The PKGBUILD change is:

install -D -m644 ${srcdir}/lxc@service ${pkgdir}/usr/lib/systemd/system/lxc@service

2. The LXC project has moved from SF to linuxcontainers.org and github. The new URL is "http://linuxcontainers.org" and stable tarballs are at "http://linuxcontainers.org/downloads/$pkgname-${pkgver}.tar.gz".

3. LXC needs that memory cgroup hierachical accounting is enabled, i.e. /sys/fs/cgroup/memory/memory.use_hierarchy is 1. It is 0 by default. Upstream, there is commit 1ea59a which makes lxc write 1 to that file. However, this feature will only be in lxc 1.0 (it was commited ono Oct 25), and now there is a kernel warning each time a container is started. This can be worked around by using a systemd tmpfile config (attached lxc.tmpfiles.d):

install -D -m644 ${srcdir}/lxc.tmpfiles.d ${pkgdir}/usr/lib/tmpfiles.d/lxc.conf

This config can be dropped in 1.0, but even if it's not there seems to be no harm (at least I see no problems running 1.0.0.beta1 with it).

4. There is an incompatibility coming with 1.0.0. The lxc-shutdown binary is gone and lxc-stop changes behavior: new lxc-stop -n ... = lxc-shutdown -n ... (sigpwr is sent to systemd inside container), new lxc-stop -k -n ... = old lxc-stop -n ... (lxc.stopsignal is sent). It took me some time to figure out why systemctl times out on container shutdown. Perhaps, this change can be advertised in the future update 0.9.0 -> 1.0.0, see the attached .install file. Of course, lxc@.service will have to be adjusted as well.

Thanks.