gnome-shell crashes x-server after nvidia driver update / old driver not available

[ej999]

hi,

yesterday I performed the following update:

Updated xorg-x11-drv-nvidia-1:304.60-1.fc17.x86_64 @rpmfusion-nonfree-updates
Update 1:304.64-1.fc17.x86_64 @rpmfusion-nonfree-updates
Updated xorg-x11-drv-nvidia-libs-1:304.60-1.fc17.i686 @rpmfusion-nonfree-updates
Updated xorg-x11-drv-nvidia-libs-1:304.60-1.fc17.x86_64 @rpmfusion-nonfree-updates
Update 1:304.64-1.fc17.i686 @rpmfusion-nonfree-updates
Update 1:304.64-1.fc17.x86_64 @rpmfusion-nonfree-updates
now my gnome session seems to be missing any gnome shell extensions and the list in the tweak tool is empty.
trying to enter the gnome settings or starting gome-shell manually causes the following x-server dump:

[ 1029.308]
[ 1029.308] Backtrace:
[ 1029.308] 0: /usr/bin/Xorg (xorg_backtrace+0x36) [0x4652d6]
[ 1029.308] 1: /usr/bin/Xorg (0x400000+0x6a2b9) [0x46a2b9]
[ 1029.308] 2: /lib64/libpthread.so.0 (0x3560e00000+0xefe0) [0x3560e0efe0]
[ 1029.308]
[ 1029.308] Segmentation fault at address (nil)
[ 1029.308]
Fatal server error:
[ 1029.308] Caught signal 11 (Segmentation fault). Server aborting
[ 1029.308]
[ 1029.308]

to check whether the issue is really related to the nvidia driver update, I tried to downgrade to version 304.60 - but it is no longer present in /var/spool/yum and also seems to be missing in the rpmfusion repository :(.
does anybody know how to get the old packages so that I can check whether it is really caused by the new driver or any other idea what may cause this?

update:
I found the old 304.60 driver on an rpmfusion mirror and after the downgrade it's working again. so it's a problem with the 304.64 driver. does anybody else have this problem?


thanks,
ej

[jpprade]

Hello, I updated my system this morning and I have, the same issue and the same stack.

I use kde I can can log in but when i do :

-launch chrome
-launch a 3d program
-start an android avd in eclipse
-opening different kind of application

Xorg crash

[buddha]

Similar problems here, F17 x64, KDE 4.9.3, updated this morning to nvidia driver 304.64,

Can log in to KDE just fine, and run Firefox, but starting Google Chrome causes crash of X

glxinfo gives error, and nvidia-settings crashes when I click on OpenGL/GLX Information

[bryanc806]

I had this issue as well. The latest rpmfind binaries don't setup the libglx.so extension correctly.

fixed with:

cd /usr/lib64/xorg/modules/extensions
mv libglx.so libglx.so-xorg
ln -s ../../../nvidia/xorg/libglx.so .

This is ugly, but works. Hopefully the next build will address this issue.

[ej999]

thanks a lot - the work around also worked for me :-)

[jpprade]

Thanks it worked for me also! How a so huge bug can be pushed in production....

[neekfenwick]

Thank you bryanc806! :D I think I love you.

Wasn't sure if it was the upgrade to kernel 3.6.7-4.fc17.x86_64 or nvidia driver to 304.64-1.fc17.x86_64. I'm running XFCE and launching Chrome brought X down in a heap. Your soft link fixed it, after an X restart.

Was also seeing this typical error message in /var/log/Xorg.0.log:


[ 9.671] (EE) NVIDIA(0): Failed to initialize the GLX module; please check in your X
[ 9.671] (EE) NVIDIA(0): log file that the GLX module has been loaded in your X
[ 9.671] (EE) NVIDIA(0): server, and that the module is the NVIDIA GLX module. If
[ 9.671] (EE) NVIDIA(0): you continue to encounter problems, Please try
[ 9.671] (EE) NVIDIA(0): reinstalling the NVIDIA driver.

I messed around with `yum downgrade` of various packages, but I'm using kmod-nvidia and nothing I did seemed to force the driver to downgrade. Can anyone suggest what command is right to downgrade a nvidia driver package that's pre-packaged against a specific kernel version:


xorg-x11-drv-nvidia-libs-304.64-1.fc17.x86_64
xorg-x11-drv-nvidia-libs-304.64-1.fc17.i686
kmod-nvidia-3.6.7-4.fc17.x86_64-304.64-1.fc17.x86_64
kmod-nvidia-304.64-1.fc17.x86_64
xorg-x11-drv-nvidia-304.64-1.fc17.x86_64

[Yellowman]

Thanks it worked for me also! How a so huge bug can be pushed in production....


The driver worked fine for the rpmfusion developers (I used it for a week, before changing to 310.19).
The real issue here is no user feedback and testing!

[The_Source_HIM]

Instead of mv'ing files just change ModulePath in xorg.conf

Section "Files"
ModulePath "/usr/lib64/nvidia/xorg"
ModulePath "/usr/lib64/xorg/modules"
EndSection

[snagles]

Instead of mv'ing files just change ModulePath in xorg.conf

Section "Files"
ModulePath "/usr/lib64/nvidia/xorg"
ModulePath "/usr/lib64/xorg/modules"
EndSection

Thank you it is work fine for me without any mv'ing

[kopolov]

I think they updated the driver in the RPM fusion repositories
I had the same problem and after checking /var/log/message
I came across the line telling me that the NVIDIDA client and kernel version are different
the client was 304.64 while the kernel was 304.60

this means that the nvidia interface (client) did not match the driver (kernel module)

I looked up how to rebuild the kernel module and it turned out to be pretty easy
sudo yum install akmod-nvidia.x86_64 (I have a 64 bit machine)

as far as I understand (still new at this) the akmod rebuilds the kernel module which of course then matches the client version
anyway, I hope this helps

[ILMostro]

The driver worked fine for the rpmfusion developers (I used it for a week, before changing to 310.19).
The real issue here is no user feedback and testing!

So, how can we be more proactive? I wanna help out nvidia-driver devs.

[tux4me]

Same problem for me, running 32 bit.
Solution is the same, just for 32 bit, it is :

Section "Files"
ModulePath "/usr/lib/nvidia/xorg"
ModulePath "/usr/lib/xorg/modules"
EndSection

[virtualeyes]

Originally Posted by The_Source_HIM

Section "Files"
ModulePath "/usr/lib64/nvidia/xorg"
ModulePath "/usr/lib64/xorg/modules"
EndSection

a caps lock THANK YOU!!! is in order ;-)

What a waste of an evening, this thread should be a sticky.

[ILMostro]

a caps lock THANK YOU!!! is in order ;-)

What a waste of an evening, this thread should be a sticky.

I thought this was pertaining to one specific version of the x-server, namely xorg-x11-drv-nvidia-1:304.60-1.fc17.x86_64 @rpmfusion-nonfree-updates, which had been fixed shortly after it was released. What version do you have? Are you in Fedora17 or 18? Are you sure that your issue was caused by the same update as this thread refers to?

Furthermore, you may want to try the akmods package:

sudo yum install akmod-nvidia.x86_64 as was previously suggested here by kopolov. Note sure if that pulls in the
akmods.noarch package as well, but those two packages should, generally, rebuild the kmod-nvidia kernel packages correctly after a kernel upgrade.