systemd-networkd: Deconfigures bridge after last attached interface disconnects (ie: LXC)

[MelkorLord]

Ubuntu 18.04 Bionic Beaver

An issue was reported to systemd-networkd on GitHub - https://github.com/systemd/systemd/issues/11650 - and all the thorough details are there.

Using anonymous bridges (ie: not attached to physical NICs) is problematic with systemd-networkd as they get deconfigured as soon as the last attached device disconnects (ie: a LXC container). Static IP configuration must be done again manually.

The root cause is that systemd-networkd assumes that every network interface have a carrier signal. There's no notion of carrier signal on anonymous bridges, a case not properly handled by systemd-networkd.

The systemd dev team provided a patch to address the issue and it would be nice to be integrated on the Ubuntu package.

The PR is here : https://github.com/systemd/systemd/commit/93b4dab57e2e13bd804cbee999241be65a443e2e

To get the proper fix, you'll need to combine 2 patches :

- The one from the PR above
- Another patch [1] which fixes a segfault introduced by the PR.

[1] https://github.com/systemd/systemd/pull/11741/commits/a294af6810df3c18909a96b556deadba0e2ab0a9

On my test environment, I rebuild the Ubuntu package with the 2 patches above from "237-3ubuntu10.17" to "237-3ubuntu10.20" and made thorough tests without hiting a single issue so I think we can assume that these 2 patches are safe to use.

[Dan Streetman (ddstreet)]

I tried to reproduce this, but my static ip isn't lost:

root@lp1825470-d:~# grep . /etc/systemd/network/*br0*
/etc/systemd/network/20-br0.netdev:[NetDev]
/etc/systemd/network/20-br0.netdev:Name=br0
/etc/systemd/network/20-br0.netdev:Kind=bridge
/etc/systemd/network/30-br0.network:[Match]
/etc/systemd/network/30-br0.network:Name=br0
/etc/systemd/network/30-br0.network:[Network]
/etc/systemd/network/30-br0.network:Address=1.2.3.4/24
/etc/systemd/network/30-br0.network:LinkLocalAddressing=none

root@lp1825470-d:~# ip a show br0
3: br0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN group default qlen 1000
    link/ether ca:9e:57:58:5f:c4 brd ff:ff:ff:ff:ff:ff
    inet6 fe80::c89e:57ff:fe58:5fc4/64 scope link
       valid_lft forever preferred_lft forever
root@lp1825470-d:~# ip l add test type dummy
root@lp1825470-d:~# ip l set up dev test
root@lp1825470-d:~# ip l set test master br0
root@lp1825470-d:~# ip a show br0
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
    link/ether ca:9e:57:58:5f:c4 brd ff:ff:ff:ff:ff:ff
    inet 1.2.3.4/24 brd 1.2.3.255 scope global br0
       valid_lft forever preferred_lft forever
    inet6 fe80::c89e:57ff:fe58:5fc4/64 scope link
       valid_lft forever preferred_lft forever
root@lp1825470-d:~# ip l del test
root@lp1825470-d:~# ip a show br0
3: br0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN group default qlen 1000
    link/ether ca:9e:57:58:5f:c4 brd ff:ff:ff:ff:ff:ff
    inet6 fe80::c89e:57ff:fe58:5fc4/64 scope link
       valid_lft forever preferred_lft forever
root@lp1825470-d:~# ip l add test type dummy
root@lp1825470-d:~# ip l set up dev test
root@lp1825470-d:~# ip l set test master br0
root@lp1825470-d:~# ip a show br0
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
    link/ether ca:9e:57:58:5f:c4 brd ff:ff:ff:ff:ff:ff
    inet 1.2.3.4/24 brd 1.2.3.255 scope global br0
       valid_lft forever preferred_lft forever
    inet6 fe80::c89e:57ff:fe58:5fc4/64 scope link
       valid_lft forever preferred_lft forever

Are you using systemd-networkd to manage your anonymous bridge ip address? Or are you setting the ip address manually? You mention lxc, do you have a test case involving lxc that reproduces the bug?

[Launchpad Janitor (janitor)]

[Expired for systemd (Ubuntu Bionic) because there has been no activity for 60 days.]

[Launchpad Janitor (janitor)]

[Expired for systemd (Ubuntu Cosmic) because there has been no activity for 60 days.]

[Launchpad Janitor (janitor)]

[Expired for systemd (Ubuntu Eoan) because there has been no activity for 60 days.]

[Launchpad Janitor (janitor)]

[Expired for systemd (Ubuntu Disco) because there has been no activity for 60 days.]