FS#54774 - systemd 233.75-3 user service crash upon graphical (lxdm) login

Attached to Project: Arch Linux
Opened by balwierz (balwierz) - Wednesday, 12 July 2017, 10:34 GMT
Last edited by Doug Newgard (Scimmia) - Tuesday, 25 July 2017, 13:49 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To No-one
Architecture x86_64
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
Upon graphical login (from lxdm) systemd user@.service crashes and dumps core. This happens when selected environment is Plasma, MATE or Gnome.
Plasma and MATE display error message "Could not sync environment to dbus." And upon clicking Ok, terminate.

This problem does occur with earlier releasever of systemd and libsystemd, but does not occur with systemd 232-8 or earlier.
Reproducibility on my system is 100%

Strangely I am able to log into Kodi via lxdm.

package versions:
* systemd, libsystemd 233.75-3
* linux 4.11.9-1-ARCH


Journal:
Jul 11 23:26:59 Skylake lxdm-session[580]: pam_unix(lxdm:session): session opened for user dupa2 by (uid=0)
Jul 11 23:26:59 Skylake systemd[1]: Created slice User Slice of dupa2.
Jul 11 23:26:59 Skylake systemd[1]: Starting User Manager for UID 1003...
Jul 11 23:26:59 Skylake systemd-logind[435]: New session c2 of user dupa2.
Jul 11 23:26:59 Skylake systemd[1]: Started Session c2 of user dupa2.
Jul 11 23:26:59 Skylake systemd[581]: pam_unix(systemd-user:session): session opened for user dupa2 by (uid=0)
Jul 11 23:26:59 Skylake systemd[581]: Assertion 'startswith(line, "env=")' failed at src/basic/env-util.c:797, function deserialize_environment(). Aborting.
Jul 11 23:26:59 Skylake systemd[1]: Started Process Core Dump (PID 585/UID 0).
Jul 11 23:26:59 Skylake systemd-coredump[586]: Removed old coredump core.systemd.0.bb7adcc2a42347d9b44114013ecc72ae.561.1499811995000000.lz4.
Jul 11 23:26:59 Skylake systemd[582]: pam_unix(systemd-user:session): session closed for user dupa2
Jul 11 23:26:59 Skylake systemd[1]: Failed to start User Manager for UID 1003.
Jul 11 23:26:59 Skylake systemd[1]: user@1003.service: Unit entered failed state.
Jul 11 23:26:59 Skylake systemd[1]: user@1003.service: Failed with result 'protocol'.
Jul 11 23:26:59 Skylake lxdm-session[580]: pam_systemd(lxdm:session): Failed to create session: Start job for unit user@1003.service failed with 'failed'
Jul 11 23:26:59 Skylake systemd-coredump[586]: Process 581 (systemd) of user 1003 dumped core.

Stack trace of thread 581:
#0 0x00007fd489bdf8c0 raise (libc.so.6)
#1 0x00007fd489be0f72 abort (libc.so.6)
#2 0x00007fd48aaec362 log_assert_failed (libsystemd-shared-233.so)
#3 0x00007fd48aaf734f deserialize_environment (libsystemd-shared-233.so)
#4 0x00007fd48aaf73f8 n/a (libsystemd-shared-233.so)
#5 0x00007fd48aaf4fac execute_directories (libsystemd-shared-233.so)
#6 0x00005639b9b15c6b n/a (systemd)
#7 0x00005639b9b134e0 n/a (systemd)
#8 0x00007fd489bcc4ca __libc_start_main (libc.so.6)
#9 0x00005639b9b14f7a n/a (systemd)


Steps to reproduce:
Update systemd and libsystemd to 233-75, reboot.
Lxdm starts.
Log in as user to any of the following: plasma, MATE, GNOME (both on X.org and wayland)

I have tested it with a freshly created user and the result is also a crash.
I have seen https://bugs.archlinux.org/task/53268, which is not completely dissimilar, but I am getting no quota errors. I am not using quota altogether.

This task depends upon

Closed by  Doug Newgard (Scimmia)
Tuesday, 25 July 2017, 13:49 GMT
Reason for closing:  No response
Comment by balwierz (balwierz) - Wednesday, 12 July 2017, 10:35 GMT
I am able to log in via TTY. That's how I manage to downgrade systemd and libsystemd.
Comment by balwierz (balwierz) - Wednesday, 12 July 2017, 10:46 GMT
I have another Arch installation, with pretty much the same software (but different hardware) and systemd crash does not happen there.
Comment by Alif (alive4ever) - Thursday, 13 July 2017, 14:49 GMT
There is a problem with systemd hierarchy cgroup on 233 and later.

Unless configured with legacy cgroup at compile time, systemd (233) will use unified cgroup as default which causes this error.

I managed to fix my problem[1] by adding boot parameter systemd.legacy_systemd_cgroup_controller=true to force systemd to use legacy cgroup controller.

Can you try this workaround?

[1] https://bbs.archlinux.org/viewtopic.php?id=227953
Comment by Doug Newgard (Scimmia) - Saturday, 22 July 2017, 14:48 GMT
Do you have anything gnupg related in ~/.config/systemd/user/ or any of it's subdirs?

Loading...