Bug#626653: nvidia-kernel-common: freeze because of /dev/nvidia0 without nvidia

[Andreas Beckmann]

On 2011-05-15 23:43, sergio wrote:
> On 05/14/2011 01:44 PM, Andreas Beckmann wrote:
>> Your setup is very special. You will need to reconfigure several things
>> depending on whether nvidia gpu is available or not, e.g.
>> /etc/X11/xorg.conf, the libGL.so.1 selection, ...
> xorg automatically depends video card and uses corresponding driver.
> I don't have any problems with libGL.

xorg does not automatically use the nvidia driver without setting this
in /etc/X11/xorg.conf. Or did you find a way to do so? In that case we
would like to learn this. An xorg.conf with Driver="nvidia" breaks on
other systems.
Please check Xorg.0.log on the nvidia machine to see which driver you
are really using.

nvidia has its own accelerated libGL.so.1 library which replaces the
system one which is from mesa. There is also a custom libglx.so Xorg
module which replaces the one from X.org. Of course these need the
nvidia driver to be functional. While it is now possible to configure
the libGL.so.1 and libglx.so via alternatives, there is nothing that
does autodetection and selects the optimal setup for the current system.

Please send me the output of
update-alternatives --display libGL.so.1

> nvidia doesn't use kms, but I quite shure, that xorg loads nvidia
> driver (if
> nvidia card presents), not udev by accessing /dev/nvidia0 or
> /dev/nvidiactl,
> anyway xorg loads modules for non kms drivers.

In case the device nodes are missing in /dev, does Xorg really load the
kernel module?

Andreas