Bug 362731

Description of problem:

When running "glxinfo" on a remote machine over "ssh -Y", the local x86_64
system randomly freezes with a hard lockup requiring a reboot.  No errors are
displayed, and unfortunately no errors are present in the kernel or X server
logs upon system startup either.  The local system is AMD64 with an ATI Radeon
9800Pro AGP card.

The kernel I am using is the F7 test kernel version 2.6.23.1-20.fc7 however I
have had the problem occur under previous kernels as well.  The current official
Fedora kernel release for F7 has an AGP related bug that prevents AGP from
working on this machine, so reproduction on the current official kernel update
isn't possible on this system.  At this point there isn't enough information to
determine whether it is the X server, kernel, etc. at fault so I've just filed
it against xorg-x11-server for the time being, until the problem can be narrowed
down further.


Version-Release number of selected component (if applicable):

xorg-x11-server-Xorg-1.3.0.0-9.fc7
xorg-x11-drv-ati-6.6.3-4.fc7
mesa-libGL-6.5.2-13.fc7
kernel-2.6.23.1-20.fc7


How reproducible:
Sometimes the glxinfo command completes successfully, and it shows that indirect
rendering is available, however other times it just hard locks.  It's more or
less random in nature, however if I keep running glxinfo over and over again,
eventually it will hang usually within 5 attempts.  It hung every time the last
3 times I ran it.   Putting glxinfo in an endless while loop on a remote host
over "ssh -Y" should be a good way to attempt to reproduce.


Steps to Reproduce:
1. start up X on x86_64 with Radeon 9800Pro AGP
2. "ssh -Y remotehost"
3. run glxinfo
  
Actual results:

Local system deadlock within 3 seconds if it does lockup.  glxinfo usually only
gets out one line of output before this happens.


Additional info:
Linux hammer 2.6.23.1-20.fc7 #1 SMP Wed Oct 31 20:31:48 EDT 2007 x86_64 x86_64
x86_64 GNU/Linux

I'm tracking Fedora 7 official kernel updates as well as koji built test kernels
for other unrelated problems, so I'll provide updates here if any kernel updates
cause changes in reproducability at all.

When I get a chance, I'll attach some additional troubleshooting/debugging info
as well.