Every OpenGL application segfaults on Fedora 8 x86_64

[SideshowBOFH]

I'm not sure exactly how long this has been happening, the last 2 or 3 driver releases I think, but every application that's linked with libGL.so immediately segfaults on startup (or soon after startup). e.g:

[me@oblivion ~]$ glxgears
Segmentation fault
[me@oblivion ~]$ glxinfo
name of display: :0.0
Segmentation fault
[me@oblivion ~]$ fglrxinfo
Segmentation fault

I'm running Fedora 8 (x86_64), fully updated (kernel 2.6.23.14-107.fc8, xorg 1.3.0), using the Livna distributed fglrx (8.452 - Catalyst 8.01). Video card is a PowerColor X1950XTX (CrossFire Master, if that makes any difference), Motherboard is an Asus M2R32-MVP (ATI SB600 chipset), 2GB RAM (so probably not running into any PCI memory holes).

I've tried a variety of driver options, ranging from pretty much nothing to just about every single option posted in this forum that is supposed to fix something.

I have seen bug 955 on ati.cchtml.com (http://ati.cchtml.com/show_bug.cgi?id=955) describing similar symptoms, it might be the same thing. I don't think I'll post any extra info on that bug, due to using the 3rd party package of fglrx.

Can anyone help? Has anyone seen this before? Is this biting anyone else?

[seeker010]

if I generate a Fedora package using the scripts or from phorogit, I have the same problem. But if I generate a package for RHEL5_64 and use that for Fedora, everything works ok.

[SideshowBOFH]

How bizarre - I'll have to try it. I wonder what the difference is?

[sobkas]

Running glxgears like this:
LIBGL_DEBUG=verbose glxgears
might bring more info.
Just copy output of this command here.

btw. I think it can't find proper fglrx_dri.so.

[SideshowBOFH]

Thanks for that sobkas - it seems that it IS finding the correct fglrx_dri.so:

[me@oblivion ~]$ LIBGL_DEBUG=verbose glxgears
libGL: XF86DRIGetClientDriverName: 8.45.4 fglrx (screen 0)
libGL: OpenDriver: trying /usr/lib64/dri/fglrx_dri.so
Segmentation fault

[sobkas]

Thanks for that sobkas - it seems that it IS finding the correct fglrx_dri.so:

[me@oblivion ~]$ LIBGL_DEBUG=verbose glxgears
libGL: XF86DRIGetClientDriverName: 8.45.4 fglrx (screen 0)
libGL: OpenDriver: trying /usr/lib64/dri/fglrx_dri.so
Segmentation fault

It segfaults even before trying to open /dev/dri/card*(or while but before printing anything).

Healthy trace looks like this:

http://pastebin.com/d59655f17

Can you post content of the following files:
/var/log/Xorg.0.log
/etc/X11/xorg.conf

and output of:
dmesg
md5sum /usr/lib64/dri/fglrx_dri.so
ldd $(whereis glxgears)
file $(whereis glxgears)
find / -name "fglrx_dri.so" 2>>/dev/null
lsmod
cat /proc/cpuinfo

ps. pastebin.com looks good for this purpose

ps2.
xorg-x11-drv-fglrx-libs is installed?

ps3.

http://fedorasolved.org/video-solutions/ati-yum-livna

http://fedorasolved.org/video-solutions/3rd-pty-video

[SideshowBOFH]

if I generate a Fedora package using the scripts or from phorogit, I have the same problem. But if I generate a package for RHEL5_64 and use that for Fedora, everything works ok.

I tried this, and it works perfectly. However, in the name of science (and trying to get this thing fixed "properly"), I'll follow-up to this post with all the info that you asked for sobkas - it might take a couple of days due to a new kmod-fglrx being released for a kernel that hasn't been released yet.

Edit: note about updated kmod-fglrx.

[SideshowBOFH]

Can you post content of the following files:
/var/log/Xorg.0.log
/etc/X11/xorg.conf

and output of:
dmesg
md5sum /usr/lib64/dri/fglrx_dri.so
ldd $(whereis glxgears)
file $(whereis glxgears)
find / -name "fglrx_dri.so" 2>>/dev/null
lsmod
cat /proc/cpuinfo

I'll go you 2 better - I'll collect this info for both the working and non-working setups, post both, and post a diff of them. That should narrow down where to look.

ps2.
xorg-x11-drv-fglrx-libs is installed?

Yes. At least, they were...

[SideshowBOFH]

Ok, sorry for the delay, having found a solution I just wanted to use it :-D

With the newly released kernel, I needed to screw around with rebuilding modules anyway, so I decided to try the livna fglrx again.

Of course, it works fine now.

Not sure what the difference is, my guess is the upgrade to Catalyst 8.02 fixed a bug in there somewhere. Hopefully this thread will serve as a guide to anyone experiencing the problem.

The 2 solutions are:

1. Upgrade to Catalyst 8.02 (kmod-fglrx-8.455.2-1.8.02.lvn8)
OR
2. Use the installer directly from ATI, and generate a RHEL5 package.