This has probably been addressed before, but I feel in the mood to share my little but hardly-acquired knowledge. I've spent the last few days configuring Nvidia GeForce 2 boards (PCI and AGP versions) on two Debian 3.0 r0 installations, and here's a short guide on how I did it...
This _should_ work with any TNT2 or GeForce 256/2/3/4 based board. I used kernel 2.2.20, and XFree86 4.1. (?) for this.
Wether you have similar settings or not, let us know if you run into problems that are not covered here.
Make sure you have X (Version 4.x) properly configured, and the needed 3D libraries installed, before following this guide. I'm no expert, I can't tell you wether having these libraries _before_ installing the drivers help. But it's the way I went.
Basically, what I did is get X to run, then use dselect to install the tuxracer package (small game where Tux slides down some icy hills). The needed packages are selected in the proccess, inlduing some X version of the Mesa 3D libraries.
If someone can give more in-depth details regarding the different Mesa 3D libraries, feel free.
1. Get the Nvidia files
First of all, get the two needed files from Nvidia : the GLX driver and the kernel driver. Their current locations is :
GLX : http://download.nvidia.com/XFree86_4....0-2960.tar.gz
kernel : http://download.nvidia.com/XFree86_4....0-2960.tar.gz
You need now to uncompress these archives. Do this by chmoding to the directory where you just saved them. Then run tar xzvf <file_name> with both files
2. Compilation and installation of the kernel driver
Third step : compile and install the kernel driver module. Chmod to the newly created NVIDIA_kernel... directory. Then type the command "make install".
You may receive an error message regarding different compiler version. If your GCC version is close enough to what was used to compile the kernel, you may ignore this by editing the Makefile that's located in the directory. Simply add the following line at the beggining of the file :
You may receive an error message that ends with something like "nv.c:22: linux/modversions.h: No such file or directory". That means you need to install the appropriate kernel headers files before compiling the driver. This can be done with dselect, but first you need to find out the version of your kernel with the command :
Then launch dselect (type in the command dselect). If you need a small introduction to dselect, I suggest the dselect Documentation for Beginners, available at http://www.debian.org/releases/slink...inner.en.html.
In dselect, search for your kernel headers package (use /, then type in headers, then press enter; use \ to cycle through results). When you have selected your package, install it and exit dselect.
You may now receive an error message stating that some file cannot be found. Basically, the Nvidia driver installation looks for the kernel headers in /usr/src/linux, while they are somewhere in /usr/src/kernel-XXXX (XXXX being some version number).
Go to /usr/src, locate your kernel-XXX directory, and then create a symlink called linux to that kernel-XXXX. This is done by typing :
ln -s /usr/src/linux /usr/src/kernel-XXX
Finally, the "make install" command should give you good results.
3. Installation of the GLX driver
Now chmod to your NVIDIA_GLX... directory, and run "make install" as well. This should no get you into much trouble.
4. Edit your XF86Config-4 file
Use vi, or your favorite file editor, to do this. I hate vi, but it's the only one available on pretty much any Unix flavor, even with very minimalist installs. Bit I'm disgressing.
You want to look for the module section and make sure that the follwong line appears :
You also want to remove (or comment out with #) the following lines :
then go to the Device section (pick the relevant one if you have many fo them, like in case of multi-headed setups), and replace the Driver "nv" (or Driver "vesa") line with this :
5. Testing all this
I usually reboot after doing this. See, I come from the Windows world... Seriously though, as we add a kernel module, I suspect it is needed to reboot and reload the kernel. Experts may prove me wrong on that, I'm all for corrections of this information.
When X starts you'll have a Nvidia splash screen (white backgroung, usual green "eye" Nividia logo, and black Nvidia text). This is a good sign.
Now try running TuxRacer (you did install it, right ?) or some other OpenGL game, and see for yourself.
6. So now what ?
Well, I hope this will help newbies like me to get their 3D boards to run under Linux. As stated before, if there's any extra info to add, be sure to let me know : firstname.lastname@example.org
. Who knows, I may find enough inspiration to maintain this kind of "Nvidia with Debian 3.0 mini-howto" embryo and turn it into something more useful to everyone.