Boot fail 2/3 times, apparently related to GDM and Wayland

[plnm]

I am using Debian Testing, and I use the integrated Intel graphics card of an i7-10510U processor.

Since two days, my system fails to start a GDM session for no apparent reason – especially there hasn't been a GDM update in ages. What I have done though is switching between X and Wayland sessions, because I had problems with both at different times. I am currently using X, however GDM seems to start Wayland anyway (sometimes – see below). What is even stranger, the failure to start a session does not happen always, but nearly exactly 2/3 of the time.

This is journalctl -u gdm for a failed boot:

Code: Select all

-- Journal begins at Wed 2021-03-10 10:28:51 GMT, ends at Tue 2021-03-23 13:09:01 GMT. --
Mar 23 12:22:58 ThinkPad systemd[1]: gdm.service: starting held back, waiting for: systemd-user-sessions.service
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Watching D-Bus name 'org.gnome.DisplayManager'.
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: starting held back, waiting for: systemd-user-sessions.service
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now not owned by anyone.
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: About to execute /usr/share/gdm/generate-config
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Forked /usr/share/gdm/generate-config as 1396
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Changed dead -> start-pre
Mar 23 12:22:59 ThinkPad systemd[1]: Starting GNOME Display Manager...
Mar 23 12:22:59 ThinkPad systemd[1396]: gdm.service: Executing: /usr/share/gdm/generate-config
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Child 1396 belongs to gdm.service.
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Control process exited, code=exited, status=0/SUCCESS
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Got final SIGCHLD for state start-pre.
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Passing 0 fds to service
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: About to execute /usr/sbin/gdm3
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Forked /usr/sbin/gdm3 as 1411
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Changed start-pre -> start
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Control group is empty.
Mar 23 12:22:59 ThinkPad systemd[1411]: gdm.service: Executing: /usr/sbin/gdm3
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now owned by :1.11
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Changed start -> running
Mar 23 12:22:59 ThinkPad systemd[1]: gdm.service: Job 139 gdm.service/start finished, result=done
Mar 23 12:22:59 ThinkPad systemd[1]: Started GNOME Display Manager.
Mar 23 12:22:59 ThinkPad gdm-launch-environment][1430]: pam_unix(gdm-launch-environment:session): session opened for user Debian-gdm(uid=118) by (uid=0)
Mar 23 12:23:24 ThinkPad gdm-launch-environment][1430]: pam_systemd(gdm-launch-environment:session): Failed to create session: Connection timed out
Mar 23 12:23:24 ThinkPad /usr/libexec/gdm-wayland-session[2339]: dbus-daemon[2339]: [session uid=118 pid=2339] Activating service name='org.freedesktop.systemd1' requested by ':1.0' (uid=118 pid=2335 comm="/usr/libexec/gdm-wayland-session dbus-run-session ")
Mar 23 12:23:24 ThinkPad /usr/libexec/gdm-wayland-session[2339]: dbus-daemon[2339]: [session uid=118 pid=2339] Activated service 'org.freedesktop.systemd1' failed: Process org.freedesktop.systemd1 exited with status 1
Mar 23 12:23:24 ThinkPad gdm3[1411]: Gdm: Couldn't get type for session 'c1': Kein passendes Gerät bzw. keine passende Adresse gefunden
Mar 23 12:23:24 ThinkPad /usr/libexec/gdm-wayland-session[2335]: Unable to register display with display manager
Mar 23 12:23:24 ThinkPad systemd[1]: gdm.service: Child 2435 belongs to gdm.service.
Mar 23 12:23:24 ThinkPad gdm-launch-environment][1430]: pam_unix(gdm-launch-environment:session): session closed for user Debian-gdm
Mar 23 12:23:24 ThinkPad gdm3[1411]: Gdm: Child process -2335 was already dead.
Mar 23 12:23:24 ThinkPad gdm3[1411]: Gdm: Child process -2335 was already dead.
Mar 23 12:23:25 ThinkPad systemd[1]: gdm.service: Child 2439 belongs to gdm.service.

(Where 'Kein passendes Gerät bzw. keine passende Adresse gefunden' means 'No suitable device or suitable address found.')

And this is the log of a successful session start:

Code: Select all

-- Journal begins at Wed 2021-03-10 10:28:51 GMT, ends at Tue 2021-03-23 13:07:01 GMT. --
Mar 23 12:24:36 ThinkPad systemd[1]: gdm.service: starting held back, waiting for: sysinit.target
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Watching D-Bus name 'org.gnome.DisplayManager'.
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: starting held back, waiting for: systemd-user-sessions.service
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now not owned by anyone.
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: About to execute /usr/share/gdm/generate-config
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Forked /usr/share/gdm/generate-config as 1575
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Changed dead -> start-pre
Mar 23 12:24:37 ThinkPad systemd[1]: Starting GNOME Display Manager...
Mar 23 12:24:37 ThinkPad systemd[1575]: gdm.service: Executing: /usr/share/gdm/generate-config
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Child 1575 belongs to gdm.service.
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Control process exited, code=exited, status=0/SUCCESS
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Got final SIGCHLD for state start-pre.
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Passing 0 fds to service
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: About to execute /usr/sbin/gdm3
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Forked /usr/sbin/gdm3 as 1595
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Changed start-pre -> start
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Control group is empty.
Mar 23 12:24:37 ThinkPad systemd[1595]: gdm.service: Executing: /usr/sbin/gdm3
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now owned by :1.13
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Changed start -> running
Mar 23 12:24:37 ThinkPad systemd[1]: gdm.service: Job 140 gdm.service/start finished, result=done
Mar 23 12:24:37 ThinkPad systemd[1]: Started GNOME Display Manager.
Mar 23 12:24:37 ThinkPad gdm-launch-environment][1629]: pam_unix(gdm-launch-environment:session): session opened for user Debian-gdm(uid=118) by (uid=0)
Mar 23 12:25:06 ThinkPad systemd[1]: gdm.service: Child 3131 belongs to gdm.service.
Mar 23 12:25:09 ThinkPad gdm-password][3401]: gkr-pam: unable to locate daemon control file
Mar 23 12:25:09 ThinkPad gdm-password][3401]: gkr-pam: stashed password to try later in open session
Mar 23 12:25:09 ThinkPad gdm-password][3401]: pam_unix(gdm-password:session): session opened for user cmj(uid=1000) by (uid=0)
Mar 23 12:25:09 ThinkPad gdm-password][3401]: gkr-pam: gnome-keyring-daemon started properly and unlocked keyring
Mar 23 12:25:13 ThinkPad gdm3[1595]: Gdm: Child process -2652 was already dead.
Mar 23 12:36:05 ThinkPad systemd[1]: gdm.service: Watching D-Bus name 'org.gnome.DisplayManager'.
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: Changed dead -> running
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now owned by :1.13
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: Watching D-Bus name 'org.gnome.DisplayManager'.
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: Changed dead -> running
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: Watching D-Bus name 'org.gnome.DisplayManager'.
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: Changed dead -> running
Mar 23 12:36:06 ThinkPad systemd[1]: gdm.service: D-Bus name org.gnome.DisplayManager now owned by :1.13

To me it looks significant that all logs of failing boots have this Wayland error in them, and none of the successful ones have. But I have no clue what that means. Anyone else?

[Head_on_a_Stick]

my system fails to start a GDM session

What actually happens? What do you see? Can you access a console screen or SSH into the system after the "failed" boot?

Please also confirm that you have the firmware-misc-nonfree and intel-microcode packages installed.

[trinidad]

While you were running Xorg did you install any software with synaptic? Was you system originally installed as gnome, or did you add gnome on later?

TC

[plnm]

What actually happens? What do you see? Can you access a console screen or SSH into the system after the "failed" boot?

I do not have an SSH server running, but I can access a console screen and in that I have network access, so SSH would be possible as well I assume.

For what actually happens, I just became very less confident in describing this. Until this point, the system used to either start GDM and show a login screen (around 1/3 of the time) or display a lot of error messages that I strangely can't seem to find in the syslog. After that, not much would happen. At least once I left the computer running for 10 minutes or so and it did not change. However now I am writing this from a boot which appeared to fail, showing the error messages, but then suddenly displayed a login screen. This is getting very ugly to trace.

Please also confirm that you have the firmware-misc-nonfree and intel-microcode packages installed.

I did not have intel-microcode, now I have both, didn't visibly change anything in the frequency of the error.

While you were running Xorg did you install any software with synaptic? Was you system originally installed as gnome, or did you add gnome on later?

I do not even have Synaptic installed, and this is a standard Debian Testing with Gnome as set up by the netinst installer. It has been set up in spring 2020 and has always been updated to the current state of Testing repositories.

[Head_on_a_Stick]

a lot of error messages that I strangely can't seem to find in the syslog

They will be recorded in the systemd journal. Note that Debian's journal is non-persistent by default, see journald.conf(5).

[plnm]

Seems to be a Heisenbug – as soon as I dig deep for finding its cause, it disappears. I now have ~10 successful boots in a row and I cannot continue my research in the now-persistent log …

Still I believe this was most likely caused by something related to Wayland. Since I use my computer productively I will not try and start a Wayland Gnome session now to see if I can reproduce the very inconvenient behaviour described in the first post. I don't know if this qualifies as 'solved' for the topic.