Patch containerd service unit to be After=data.mount on power on

[vasilvas99]

Issue

The kanto-CM service unit being dependent on the data.mount provided a partial fix for the issue of improperly un-mounting the /data partition (most cases when shutdown now and poweroff were used).

However, containerd might continue to use /data even after kanto-cm is stopped, leading to improper un-mounting of the partition when the system was rebooted very soon after startup for example.

Fix

A similar patch for the containerd service unit was provided that makes it depend on the data.mount service on startup (services are stopped in the reverse order on power-off).

Testing

A qemux86-64 and a qemuarm64 leda-distro builds were tested. The system was shut off quickly after a build, rebooted via reboot now and so on. No cases of improper dismount of /data were observed

Power-off logs

...
[  OK  ] Stopped D-Bus System Message Bus.
[  OK  ] Unmounted /run/containerd/…4216-9e02-fcdbe4ee4606/rootfs.
[  OK  ] Unmounted /run/container-management/netns/c8603ff829db.
[  OK  ] Unmounted /run/containerd/…4d72-8442-317765dcaa4f/rootfs.
[  OK  ] Unmounted /run/container-management/netns/4a65d8d364d7.
[  OK  ] Unmounted /run/containerd/…4d60-86f0-f1a5e3fd946e/rootfs.
[  OK  ] Unmounted /run/container-management/netns/66402b492bfa.
[  OK  ] Unmounted /run/containerd/…4090-b7b8-48c3f0a18d4a/rootfs.
[  OK  ] Unmounted /run/container-management/netns/0859b29be63c.
[  OK  ] Unmounted /run/containerd/…46d6-8b65-5022ad5d176b/rootfs.
[  OK  ] Unmounted /run/container-management/netns/516769aeb6b6.
[  OK  ] Unmounted /run/containerd/…47f8-98d9-597686eb8284/rootfs.
[  OK  ] Unmounted /run/container-management/netns/f9d87dd63917.
[  OK  ] Stopped Eclipse Kanto - Container Management.
         Stopping containerd container runtime...
[  OK  ] Stopped containerd container runtime.
[  OK  ] Stopped target Network.
         Unmounting /data...
         Stopping Network Name Resolution...
[  OK  ] Stopped Network Name Resolution.
[  OK  ] Unmounted /data.
...

From the logs it can be observed that the order of services being stopped is:

1. Unmount containerd virtual files at /run/
2. Stop Eclipse kanto-cm service (takes a long time)
3. Stop containerd
4. Unmount /data

[stlachev]

LGTM