On 18.11.2025 17:21, Max Nikulin wrote:
On 18/11/2025 14:57, Peter Milesson wrote:
There is absolutely nothing in the Xorg.0.log (~/.local/share/xorg/
Xorg.0.log).
"Absolutely nothing" is a bit ambiguous. I am still on bookworm and
in my case path is /var/log/Xorg.0.log.old (without .old it should be
new LightDM greeter session). I hope, you checked file modification
time.
There is nothing in any log indicating non standard behavior, or
errors. There are messages of the session startup in the journal,
and then messages from session shut down, nothing in between.
Have you compared *user* session log with the case when ~/ is a local
directory?
I expect that errors outside of systemd scope should appear in
~/.xsession-errors. There is a chance that something is logged to
/var/log/lightdm.
Xorg.0.log is also properly stored in ~/.local/share/xorg, which
proves that the home folder is mounted rw for the user.
I would rather suspect something behind regular file read-write:
special files, locks, etc.
Ideas to try:
- $HOME on NFS instead of CIFS
- CIFS mounted from command line instead of PAM plugin
- Another lightweight session like fluxbox just to test if the issue
is specific to LXDE
Perhaps there is a way to enable more verbose LXDE logs.
Hi Max,
With absolutely nothing, I really mean absolutely nothing that could
indicate any errors. The timestamp on the
~/.local/share/xorg/Xorg.0.log is the last time the session was run.
When the user profile is locally stored, the LXDE desktop is displayed
as it should, even when the user is authenticated with Kerberos.
Thanks for pointing me at the .xsession-errors file. I missed to check
it, and there are 2 lines that maybe could tell what's wrong. They are
the last lines after session setup:
unix_listener_tmp: bind
"/home/miles/.ssh/agent/s.p5yeigkbUn.agent.LNVFOjW41c": Operation not
supported
main: Couldn't prepare agent socket
There are no files created in the ~/.ssh/agent directory with the
newer kernel.
When I use the older, working kernel, there is a file in the
directory. The last part of the file name changes between the sessions.
Thus it seems that the cifs module has changed substantially from
kernel 6.12.48 to 6.16.3. I haven't got a clue how to allow creation
of sockets in the cifs mount parameters, however. Maybe I should
report it as a bug?
Thanks for pointing me in a direction that maybe can solve the problem.
Best regards,
Peter