GLX Segfaults 2.6.10 / Xfree86 / 6629

[sebas]

Hi,

I have problems with the glx extension with my 6800 LE (on a A8V deluxe board). It manifests itself by preventing X to start properly. It shows the nvidia logo for a fraction of a second and then goes back to X. Commenting the line Load "glx" from my XF86Config-4 makes X start again, but of course without GLX support. The interesting part of the of XFree86.0.log probably is:

(II) Loading extension NV-GLX
(II) NVIDIA(0): NVIDIA 3D Acceleration Architecture Initialized
(II) NVIDIA(0): Using the NVIDIA 2D acceleration architecture
(==) NVIDIA(0): Backing store disabled
(==) NVIDIA(0): Silken mouse enabled
(**) Option "dpms"
(**) NVIDIA(0): DPMS enabled
(II) Loading extension NV-CONTROL
(==) RandR enabled
(II) Initializing built-in extension MIT-SHM
(II) Initializing built-in extension XInputExtension
(II) Initializing built-in extension XTEST
(II) Initializing built-in extension XKEYBOARD
(II) Initializing built-in extension LBX
(II) Initializing built-in extension XC-APPGROUP
(II) Initializing built-in extension SECURITY
(II) Initializing built-in extension XINERAMA
(II) Initializing built-in extension XFree86-Bigfont
(II) Initializing built-in extension RENDER
(II) Initializing built-in extension RANDR
(II) Initializing extension GLX

   *** If unresolved symbols were reported above, they might not
   *** be the reason for the server aborting.

Fatal server error:
Caught signal 11.  Server aborting

The problem is the same with 2.6.8 and 2.6.10. Note that I had it all running smoothly and fast (about 9000 fps in glxgears) already, but it broke. (I think it was after reinstalling libc6, if that could possibly have to do something with it). I also did not upgrade any package in the meantime, so I suppose it should be possible to have it working in my setup. I am running Debian sid pure64 and I have also installed some 32bits compatibility stuff. I read, that the nvidia-installer might try to install 32bits stuff into the 64bits, directory (Not sure if this is still the case with 6629, though), so I also tried to copy the files from the .run file by hand, making sure 32bit stuff doesn't get mixed with lib64.

I tried uninstalling the NVIDIA driver from the .run file, installing nvidia-glx and friends from http://people.debian.org/~rdonald/index.php, all to no avail. So right now, I am pretty puzzled and already getting short of ideas. for your reference I have posted some logs and information on the following addess: http://vizzzion.org/~sebas/tmp/nvidia/ (XFree86.0.log, strace -f of startx, filelistings of /usr and /lib, XF86Config-4, dmesg, lsmod, .config)

If you need more information, please let me know.

Thanks in advance for your suggestions and help.

[chunkey]

AFAIK, debian pure64 don't have any 64 / 32-Bit split in /usr/X11R6/

May this help you: (as root)

cd /usr/X11R6/
mv lib64 lib64.backup
ln -s lib lib64

and run ./NVIDIA-Linux-x86_64-1.0-6629-pkg2.run (not nvidia-glx) again uncomment/add "load "glx""

Now check again.

(I hope I don't have to edit this once more)...

[sebas]

I tried this, but it didn't help. nvidia-installer spewed some errors, here's the the nvidia-installer.log.

In the meantime I also removed X completely, including xlibs and a couple of other libs. To make sure there's nothing left, I moved /usr/X11R6 out of the way and installed all necessary packages again. The error stays.

I have libc6 2.3.2.ds1-19.0.0.2.pure64 installed, btw, in case that makes any difference. I also added dpkg -l output of packages and versions at the logs directory

\edit
I just tried to install from the NVIDIA*.run package, without ia32 emulation, and I am getting the following error:

-> Running runtime sanity check:
ERROR: The runtime configuration check failed for library
       'libnvidia-tls.so.1.0.6629' (expected:
       '/usr/lib64/tls/libnvidia-tls.so.1', found:
       '/usr/lib64/libnvidia-tls.so.1').  The most likely reason for this is
       that conflicting OpenGL libraries are installed in a location not
       inspected by nvidia-installer.  Please be sure you have uninstalled any
       third-party OpenGL and third-party graphics driver packages.
-> done.
-> Runtime sanity check failed.

(nvidia-installer-64bit-only.log)
Removing /usr/lib64/tls and /usr/lib64/libnvidia-tls.so.1 (and running ldconfig) doesn't make this error go away.

[chunkey]

No. I've also the same libc here: "Version: 2.3.2.ds1-19.0.0.2.pure64".
Anyway, that doesn't solve your problem. Have you ever played with the "--advanced-options"? like "--force-tls=new" or "--force-tls=classic" ?

But you can also use the old "--extract-only" switch. Then the complete
driver will be extracted into the "NVIDIA-Linux-x86_64-1.0-6629-pkg*/". where you can use "make install" to install the driver "by hand"/manually...
(Note: /usr/X11R6/lib64 had to be symlinked to /usr/X11R6/lib!, just all the steps from post one again...).

The last method have also the advantage that you can add some patches when you want to try out 2.6.10 (look in the sticky x86_64 thread for more infos....)

[sebas]

The installer worked without errors when using the --force-tls=new option, also make install would work. Unfortunately after installing, the problem stays the same. I am beginning to believe that there must be some stale symlink around which makes glx barf out. A cd /usr/lib64 && file *|grep 32 didn't show, maybe you have other ideas?

[chunkey]

Unfortunately no, I had only problems getting OpenGL work, and the symlink did the job. Everything else was fine. Now, I can only guess what's wrong with the setup, like having installed some mesa-stuff? Or wait and hope that next driver release will do some black magic...

[sebas]

Hm, yeah. I realized that a clean setup would probably be the easier way, so I spent the afternoon doing that. OpenGL works fine now (used the debian packages for nvidia-glx) and I am happy again.

Thanks for your patience!

[Gonzalo Aguilar]

I have exactly the same problem but 2.6.8 here:

Linux azul1 2.6.8-gentoo-r4 #1 Mon Sep 20 18:47:27 CEST 2004 i686 AMD Athlon(tm) XP 1800+ AuthenticAMD GNU/Linux

I've got the same X log when activating GLX module I had to disable it.

It starts now with old nvidia driver 6211 (I think) but any program I try to run (even glxgears) fails with an "Violación de segmento"=Segment fault

I don't know how this is related cause I cannot find any information in X logs...