Red Hat Bugzilla – Bug 51735
XFree86 locks up machine hard using Revolution IV card and SGI flatpanel
Last modified: 2007-04-18 12:35:51 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.3-12 i686; en-US; rv:0.9.1)
Description of problem:
While using Revolution IV cards and the SGI 1600SW flat panel display,
restarting X (logging out, or killing X) occasionally locks the machine up
Steps to Reproduce:
2.After an extended period of time, log out of X (or kill it)
Actual Results: The machine often locks up hard. Since the monitor does
not use stardard VGA output, it does not recieve signal from the machine
after the lockup, so you cannot check for kernel messages
Expected Results: X should have restarted.
So far I've only tested this with Dell Precision 420 workstations and the
revolution IV cards. I have other machines available for testing if you'd
like be to do any further investigation.
We have a large number of these machines with this particular setup, so the
crash is not machine specific (probably not flakey hardware).
These machines were stable with Redhat 6.2 and 7.0. The issue didn't start
until redhat 7.1, and it continues with the roswell beta. The problem
didn't occur with redhat 6.2 with xfree86 4.0.3, so it's probably not X
I tried running 7.1 with the 2.2.19 kernel, so it's not 2.4 specific.
Perhaps the new glibc?
Attach X server logs and configs using the link below.
Created attachment 28033 [details]
Created attachment 28074 [details]
X server log
Attach your /var/log/messages after a crash.
Created attachment 30214 [details]
/var/log/messages after a crash.
I recommend updating to our current erratum kernel and XFree86, etc.
and seeing if that solves the problem. XFree86-4.1.0-15 has been
released for RHL 7.1
I've applied the new XFree86 RPMs and the latest kernel RPM, and test machines
appear to be stable. I'll send confirmation in a week or so if they're still
As a sidenote, I attempted to debug the old kernel/X by logging over null modem
to another machine, and the machines didn't report any errors when they crashed.
Assuming erratum release fixed the problem as above.