fix(dracut): rework timeout for devices added via --mount and --add-device (bsc#1231792)

[aafeijoo-suse]

Currently, dracut adds a systemd dropin that sets an infinite timeout only for the underlying persistent devices of each host device that must be active in the initrd. E.g.:

localhost:~ # dracut -f --stdlog 3 --install "/mnt" --mount "/dev/sda1 /mnt btrfs rw,relatime"
localhost:~ # lsinitrd | grep timeout.conf
-rw-r--r--   2 root     root            0 Oct 28 14:21 etc/systemd/system/dev-disk-by\x2duuid-0d1b24c2\x2df112\x2d48ef\x2d8442\x2d1001cffc92f0.device.d/timeout.conf
localhost:~ # ls -l /dev/disk/by-uuid/0d1b24c2-f112-48ef-8442-1001cffc92f0
lrwxrwxrwx 1 root root 10 Oct 28 11:43 /dev/disk/by-uuid/0d1b24c2-f112-48ef-8442-1001cffc92f0 -> ../../sda1

It can be verified in the emergency shell:

sh-5.2# systemctl show --property JobRunningTimeoutUSec /dev/disk/by-uuid/0d1b24c2-f112-48ef-8442-1001cffc92f0
JobRunningTimeoutUSec=infinity
sh-5.2# systemctl show --property JobRunningTimeoutUSec /dev/sda1
JobRunningTimeoutUSec=1min 30s

This can cause a problem if a user application expects to mount /dev/sda1 in the initrd (that's why it was explicitly added --mount "/dev/sda1 ..."), but for some reason the link is not created within 1min 30s.

(cherry picked from commit dracut-ng/dracut-ng@c79fc8f)

---

I'm not 100% convinced about this patch yet, but since it was accepted upstream we should evaluate whether it deserves to be backported.

[tblume]

I agree to the doubts about the patch.
We might hide other issues (e.g. device link creation) therewith.
On the other hand it is a good fallback to make the system boot.
Would it be feasible to add a warning if the persistent device link cannot be used and we switch to the fallback instead?

[aafeijoo-suse]

I agree to the doubts about the patch. We might hide other issues (e.g. device link creation) therewith. On the other hand it is a good fallback to make the system boot.

The idea is that devices added via --mount and --add-device should be additional user devices. The underlying root device and the symlink pointed by root= on the kernel command line, and other important devices are already added automatically by dracut.

Would it be feasible to add a warning if the persistent device link cannot be used and we switch to the fallback instead?

Do you mean adding a warning at boot? I don't think so... at that point systemd is in charge. The unique effect of this patch is extending the default timeout from 1m 30s to infinity for the symlink added using these user options.