[SOLVED] Random xfce log off dump to lightdm screen?
Welcome, Guest.

Author Topic: [SOLVED] Random xfce log off dump to lightdm screen?  (Read 119 times)

Offline CwF

  • Full Member
  • ****
  • Posts: 208
  • Country: us
[SOLVED] Random xfce log off dump to lightdm screen?
« on: May 10, 2018, 01:25:31 AM »
My symptoms are a random log off, a dump to my lightdm screen, with the error from lightdm.log "Seat seat0: Stopping display server, no sessions require it"

This is my Debian Stretch KVM hypervisor running multiple VM's, some with vfio hardware. The issue surfaced with no changes to the host other than VM configuration, including hardware changes and kernel upgrades to ddebian vm's. If there is interest or suspect here, I can detail that later.

There is variation on the actual crash. Sometimes everything open on the host crashes except any xterm's. Other terminals will disappear. Open programs loose unsaved data. Sometimes only a log on is required and nothing crashed. In every case, VM's are all unaffected. VM's on passed video continue without a glitch. With a passed keyboard you'd never know the host logged out. All connections through kvm/spice get terminated. Since most things are happening inside a vm, unsaved work on the host is rare, and since vm's are unaffected I've stopped rebooting since it doesn't help. Since Monday it has logged off 5-6 times.

I'm poking through logs, most things document a result of the session stopping and not clearly why. The problem seems old and rare, I've found threads of this exact issue where xfce and lightdm are the only common elements with my setup. There is no screensaver or power manager software, maybe I should ditch lightdm also. Default Debian  does timeout and blank without those and I use xset manually to change that, been fine. Any ideas?



Post Merge: May 11, 2018, 12:28:21 PM
I'll make it solved hopefully on my next visist..
I was led awry maybe by the only other exact example being so different, my example vs. arch/laptop/virtualbox, I found some bug references in calls to udev that match with my error logs, that match up with the addition of, removal of, and reintroduction of different pcie eth cards. So way over in left field that was a change to the host. The suspect 4 port card that was divvied up has been replaced with a simpler 2 port. How exactly the behavior of that card feeds back into blowing up the xsession on vt7 I'm not sure....

Post Merge: May 13, 2018, 10:30:33 AM
Indeed. After a libvirt request and errors on vm start up I'd get the no crash log off. The open host crash mode seems to occur after a dhcprequest and renewal, every 11 hours or so. Where specifically there's something to fix is beyond me.
« Last Edit: May 13, 2018, 10:30:33 AM by CwF »