Re: nvidia problems (Sid dist-upgrade)

[Florian Kulzer <florian@molphys.leidenuniv.nl>]

On Sun, Oct 29, 2006 at 19:30:21 +0100, Benjamí Villoslada wrote:
> Hi,
> 
> The yesterday dist-upgrade install two new packages:
> linux-image-2.6.18-1-486 nvidia-kernel-2.6.18-1-486
> 
> (I use a custom kernel)
> 
> Also installs new versions of:
> nvidia-glx nvidia-kernel-source
> 
> After dist-upgrade, xorg doesn't start: «no screens found».
> 
> One new nvidia module-assistant (*) after dist-upgrade doesnt solve the 
> problem.
> (*) http://wiki.debian.org/NvidiaGraphicsDrivers
> 
> Then I've returned to the old versions:
> 
> nvidia-glx 1.0.8774-6
> nvidia-kernel-2.6.16.18-smp-bitassa 1.0.8774-6+20060528.01
> nvidia-kernel-common 20051028+1
> nvidia-kernel-source 1.0.8774-6
> 
> and purge linux-image-2.6.18-1-486, I use my kernel.
> 
> Is the correct way?  I don't understand this new nvidia update. I should use 
> the linux-image-2.6.18-1-486 instead of my kernel?   I'm sorry, don't know 
> deeply how nvidia works. From now I use module-assistant 
> (http://wiki.debian.org/NvidiaGraphicsDrivers) and was very easy :)

I use the nvidia-* packages version 1.0.8776-1 on my Sid system with a
custom kernel (compiled from linux-source-2.6.18 version 2.6.18-3) and I
do not have any problem.

Maybe the old nvidia kernel module was not properly unloaded in your
case and that gave problems when you started X again due to version
mismatch.

I would try this:

Shut down X; install the new nvidia-kernel-source package again.

# module-assiant auto-install nvidia

Make sure the new version of the module is loaded:

$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA Linux x86 Kernel Module  1.0-8776  Mon Oct 16 21:56:04 PDT 2006
GCC version:  gcc version 4.1.2 20061020 (prerelease) (Debian 4.1.1-17)

If you still have the old module loaded try "modprobe -r nvidia"
followed by "modprobe nvidia". If this fails to load the new module then
check the version string of the nvidia.ko file in /lib/modules:

$ strings /lib/modules/$(uname -r)/nvidia/nvidia.ko | grep "nvidia id"
nvidia id: NVIDIA Linux x86 Kernel Module  1.0-8776  Mon Oct 16 21:56:04 PDT 2006

Once you have made sure that the 1.0-8776 module can be loaded, install
the new nvidia-glx package and verify that the libglx.so symlink points
to the right file:

$ file /usr/lib/xorg/modules/extensions/libglx.so
/usr/lib/xorg/modules/extensions/libglx.so: symbolic link to `libglx.so.1.0.8776'

Also check the version of nvidia_drv.o:

$ strings /usr/lib/xorg/modules/drivers/nvidia_drv.o | grep "nvidia id"
nvidia id: NVIDIA X Driver  1.0-8776  Mon Oct 16 21:58:46 PDT 2006

If all versions are consistent then X should start again. If it does not
then run

$ egrep '^\((EE|WW)\)' /var/log/Xorg.0.log

and check the output for clues.

-- 
Regards,
          Florian