package fails to install: SIOCSIFNETMASK: Cannot assign requested address

It could potentially be a followup failure from the cgroup-lite failure in bug 929381. However, as this was a dist-upgrade, it was running under a lucid kernel which might not yet support the necessary features. It needs to fail gracefully then.

I confirm that this is not a consequence of bug 929381. I cannot reproduce this in a lucid live system, so the SIOCSIFNETMASK might be something special in the dist-upgrader test environment?

I logged into the dist-upgrade tester VM for main-all, and tried to install lxc:

Setting up lxc (0.7.5-3ubuntu20) ...
AppArmor parser error for /etc/apparmor.d/usr.bin.lxc-start in /etc/apparmor.d/usr.bin.lxc-start at line 40: Invalid capability syslog.
 * Starting Linux Containers

dnsmasq: failed to bind listening socket for fe80::ac5d:58ff:fe84:5af4: Address already in use
 * Failed to set up LXC network
invoke-rc.d: initscript lxc, action "start" failed.
dpkg: error processing lxc (--configure):
 subprocess installed post-installation script returned error exit status 2

So this is a different failure, and maybe even not in lxc itself. I think this should also be handled more gracefully.

I'm still trying to reproduce the original error, so that I can set -x the init.d script.

However, in both cases it apparently fails in the same exit path. It seems that the trapping is meant to call undo_network() on failure (which works), then undo_network() goes all the way to the end, and then the script just exits:

.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0

dnsmasq: failed to bind listening socket for fe80::188f:6ff:fe83:141: Address already in use
+ undo_network
+ [ -f /var/run/lxc/dnsmasq.pid ]
+ iptables -t nat -D POSTROUTING -s 10.0.3.0/24 -j MASQUERADE
[...]
+ echo * Failed to set up LXC network
 * Failed to set up LXC network
root@ubuntu:~# echo $?
2

As the script always exits right away after the trap, the only way that I found to do that cleanly is to add an "exit 0":

undo_network() {
        [...]
        log_failure_msg "Failed to set up LXC network"
        exit 0
}

This allows the package to get configured and not break upgrades even if LXC setup fails for some reason. Does that sound acceptable?

exit 0 sounds fine, thanks.

This bug was fixed in the package lxc - 0.7.5-3ubuntu21

---------------
lxc (0.7.5-3ubuntu21) precise; urgency=low

  * debian/lxc.init: Exit cleanly in undo_network(), to avoid the init.d
    script and thus the package installation to fail if the network could not
    be configured for LXC. (LP: #929382)
 -- Martin Pitt <email address hidden> Thu, 09 Feb 2012 16:47:09 +0100