user-runtime-dir unit fails to start

[tinloaf]

Windows version (build number):
20H2 (Build 19042.928)

Linux distribution:
Ubuntu Focal

Genie version:
1.39

Describe the bug
The [email protected] unit fails to start. I see this error in the journal:

Apr 27 08:59:50 LBA-DB-21-wsl systemd[1]: Failed to start User Runtime Directory /run/user/1000.
Apr 27 08:59:50 LBA-DB-21-wsl systemd[1]: [email protected]: Failed with result 'exit-code'.
Apr 27 08:59:50 LBA-DB-21-wsl systemd[1]: [email protected]: Main process exited, code=exited, status=32/n/a
Apr 27 08:59:50 LBA-DB-21-wsl mount[748]: mount: /run/user/1000: special device /mnt/wslg/runtime-dir does not exist.
Apr 27 08:59:50 LBA-DB-21-wsl systemd-logind[289]: New session c1 of user lba.
Apr 27 08:59:50 LBA-DB-21-wsl systemd[1]: Starting User Runtime Directory /run/user/1000...
Apr 27 08:59:50 LBA-DB-21-wsl systemd[1]: Created slice User Slice of UID 1000.

Indeed, the directory /mnt/wslg does not exist. I can see that genie does install an override for user-runtime-dir that mounts /run/user/1000 into /mnt/wslg, but I can't figure out where /mnt/wslg is supposed to come from.

If the bug involves systemctl or a service running under systemd, confirm that you are running inside the bottle:
"inside"

To Reproduce
Just start the bottle (e.g. wsl -- genie -s)

Expected behavior
The service should start. :)

[a-m-s]

I'm seeing this for user 128:

Failed units will now be displayed (systemctl list-units --failed):
  UNIT                         LOAD   ACTIVE SUB    DESCRIPTION
● [email protected] loaded failed failed User Runtime Directory /run/user/128

Same error message about /mnt/wslg/runtime-dir.

[cerebrate]

Both of these errors are unforeseen issues with WSLg support. They'll be fixed in 1.40, releasing later today.

[cerebrate]

And now done.

[tinloaf]

Thanks! However, I think there is still a problem with the start order. This is what I see in my log after starting genie. My systemd's user session fails to start:

> journalctl -ru [email protected]
-- Logs begin at Wed 2021-02-17 09:43:30 CET, end at Wed 2021-04-28 09:01:53 CEST. --
Apr 28 09:00:15 LBA-DB-21-wsl systemd[1]: Failed to start User Manager for UID 1000.
Apr 28 09:00:15 LBA-DB-21-wsl systemd[1]: [email protected]: Failed with result 'exit-code'.
Apr 28 09:00:15 LBA-DB-21-wsl systemd[1]: [email protected]: Main process exited, code=exited, status=219/CGROUP
Apr 28 09:00:15 LBA-DB-21-wsl systemd[1]: Starting User Manager for UID 1000...
Apr 28 08:55:27 LBA-DB-21-wsl systemd[1]: Failed to start User Manager for UID 1000.
Apr 28 08:55:27 LBA-DB-21-wsl systemd[1]: [email protected]: Failed with result 'exit-code'.
Apr 28 08:55:27 LBA-DB-21-wsl systemd[1]: [email protected]: Main process exited, code=exited, status=1/FAILURE
Apr 28 08:55:27 LBA-DB-21-wsl systemd[472]: Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.
Apr 28 08:55:27 LBA-DB-21-wsl systemd[472]: pam_systemd(systemd-user:session): Not setting $XDG_RUNTIME_DIR, as the directory is not in order.
Apr 28 08:55:27 LBA-DB-21-wsl systemd[472]: pam_systemd(systemd-user:session): Runtime directory '/run/user/1000' is not owned by UID 1000, as it should.
Apr 28 08:55:27 LBA-DB-21-wsl systemd[472]: pam_unix(systemd-user:session): session opened for user lba by (uid=0)
Apr 28 08:55:27 LBA-DB-21-wsl systemd[1]: Starting User Manager for UID 1000...
Apr 28 08:54:18 LBA-DB-21-wsl systemd[1]: [email protected]: Job [email protected]/start failed with result 'dependency'.
Apr 28 08:54:18 LBA-DB-21-wsl systemd[1]: Dependency failed for User Manager for UID 1000.
Apr 28 08:50:58 LBA-DB-21-wsl systemd[1]: [email protected]: Job [email protected]/start failed with result 'dependency'.
Apr 28 08:50:58 LBA-DB-21-wsl systemd[1]: Dependency failed for User Manager for UID 1000.

However, the directory looks fine:

root@LBA-DB-21-wsl:/home/lba# ls -l /run/user/
total 0
drwx------ 3 lba lba 60 Apr 28 09:00 1000

And if I just run systemctl restart [email protected], the user session comes up successfully. My guess it that systemd tries to start [email protected] before the directory is mounted?

[cerebrate]

What does a systemctl status on [email protected] look like?

[tinloaf]

I can't reproduce the problem anymore, not even after wsl -t. I'm not sure if there's a race condition somewhere or if this was a one-time glitch. If it happens again, I'll try to get as much logs and status outputs as possible.