Red Hat Bugzilla – Bug 450327
Unlimited respawning of gdm
Last modified: 2014-06-18 04:46:26 EDT
Description of problem:
For as yet unknown reasons, gdm was crashing on my laptop. upstart failed to
limit the respawns, causing gdm to be restarted continually
Version-Release number of selected component (if applicable):
I put the relevant parts of /var/log/messages at
I opened bz 450330 for the gdm problem
We just need to set a respawn limit then. If I could just find documentation on
the syntax :)
I'll have a patch for this when I get up with Scott.
It seems that there is a limit set already. My /etc/event.d/prefdm (which is the
one from initscripts-8.76.2-1.x86_64) has the line 'respawn limit 10 120'
If you get it into this state, (where it's constantly respawning) what's the
output of 'initctl status prefdm'?
(In reply to comment #3)
> It seems that there is a limit set already. My /etc/event.d/prefdm (which is the
> one from initscripts-8.76.2-1.x86_64) has the line 'respawn limit 10 120'
So I just saw. It should be respawning no more than 10 times per 2 minutes.
Could you attach the /etc/event.d/prefdm file for me?
# prefdm - preferred display manager
# Starts gdm/xdm/etc by preference
start on stopped rc5
stop on runlevel [!5]
respawn limit 10 120
exec /etc/X11/prefdm -nodaemon
(In reply to comment #4)
> If you get it into this state, (where it's constantly respawning) what's the
> output of 'initctl status prefdm'?
I couldn't get that information - I can't type anything on a local vt since gdm
takes the screen over faster than I can type, and even when I log in from
another box, the respawning somehow cut the ssh connection. I tried to capture
that output by running the initctl in the background, booting first into
runlevel 3 and then doing a 'telinit 5', but because I had messed with my
homedir in /etc/passwd gdm now does not crash anymore (even after switching back
to the original homedir)
So .. nothing but a mystery. I am wondering if at least the respawning part
could be simulated by replacing /etc/X11/prefdm with a script that just exits.
Could you edit the /etc/event.d/prefdm file and add this line to the end:
post-stop exec sleep 10
That will introduce a 10s delay between each respawn, so should give you ample
time to get a console.
Marking as needinfo, since we still haven't heard back about Scott's requiest.
David, are you still experiencing this issue?
I haven't been able to reproduce this for some time now. Feel free to close it.