Installing a GeForce4 w/ RH 7.3

For quite a while now, I've been working on installing my Gainward GeForce4 Ti4200 under Red Hat 7.3. In fact, its taken so much effort, I decided to document my steps so others may have an easier time of it. Note that much of this information I found at other sites.

Firstly, I have a dual boot with Red Hat and Win XP. This may seem to have nothing to do with a graphics card, but it did influence the process, so I will discuss it breifly. I set this up by first installing XP on a 4 gb NTFS parition, red hat with a 50mb boot, a 3 gb /, and a 1 gb swap partition. The remaining space of my hard drive is formatted in FAT32. During the Red Hat installation, I chose to put LILO on the linux parition, rather than the master boot record. I then used the following to copy LILO onto a floppy:
First, mount a disk in MSDOS format:
mount -t msdos /dev/fd0 /mnt/floppy
Then copy the boot information into a file:
dd if=/dev/hda2 bs=512 count=1 of=/mnt/floppy/linux.bin

I then rebooted into xp, and copied the linux.bin file off the floppy and onto the c drive. I then edited the boot.ini file (its hidden) to include the following line:
C:\linux.bin="Linux"

When I rebooted, the xp bootloader had Linux on its list, and selecting it brought up the LILO srceen, from which I could boot to linux.

Now, I know this sounds more like a dual-boot how to than a vid card installation, but my setup influenced the way I had to install the card, so I felt it was important to discuss it.

Before even begining the driver install, you must update Red Hat's kernel. The verision I had was version 2.4.18-3 and the version required is version 2.4.18-5. To update this, I used red hat's update agent. Do the following at a terminal:
rhn_register
This will register you with the red hat network
up2date
This will begin the update process. Chose to download the newest versions of the kernel (I personally choose to update everything).

Here's where the fun began. The update agent claimed to have installed the new kernel, however, even after a reboot, the command uname -r , which returns the version number of the kernel, continued to show the old version. I finally figured out this was due to the way my dual boot was set up: the LILO file on my xp partition was not being updated to include an option to boot into the new kernel. However, the LILO file on my linux partition was, so the installer detected nothing wrong. Those of you who actually know what your doing with linux now realize I'm a total idiot. After several days of trying this that didn't work, I finally figured it out. By using a floppy to copy the LILO information, using the dd command as stated before, I could update the LILO file on the xp partition. Just overwrite the Linux.bin file I created earlier. This now gave me the option to boot into the new kernel, and uname -r now returned the correct version number.

After all that work, we can now begin the actual driver installation. I went to nvidia.com and dled the following files from the linux drivers section:
NVIDIA_kernel-1.0-2960.tar.gz
NVIDIA_GLX-1.0-2960.tar.gz

I then unpacked the tars:
tar xvzf NVIDIA_kernel-1.0-2960.tar.gz
tar xvzf NVIDIA_GLX-1.0-2960.tar.gz

Now, we must exit Xwindows in order to install the drivers. So, type the following at a command line to ensure that X does not start when we boot up (do this as root):
pico /etc/inittab
Change the line:
id:5:initdefault
by replacing the 5 with a 3. This will put us in runlevel 3 (text only mode) when we reboot. Save the file and exit (ctrl + x, and then type y and enter). Now reboot the computer and log in at text console. Go to the directory you saved the 2 driver files and run the following commands:

cd NVIDIA_kernel-1.0-2960
make
You must run the make install command as root, so we will switch to root with the su command
su
make install
You will be warned that this will taint the kernel. This is ok.

cd NVIDIA_GLX-1.0-2960
make
make install

Now we must edit the XF86Config-4 file:
cd /etc/X11
pico XF86Config-4

Add # in front of the following lines:
Load "GLcore"
Load "dri"

Section "DRI"
Mode 0666
EndSection

And the change the line: Driver "nv" or Driver "vesa" to
Driver "nvidia"

Save the file and exit. Type startx to test the drivers. You should see an nvidia splash screen and Xwindows should load. Everything seems to be operating fine. However, upon reboot, the drivers are no longer working. So do the following:
su
pico /etc/modules.conf
Add the following line:
alias char-majr-195 NVDriver
Save the changes and reboot.

Now test the drivers with the startx command again. If they are working, your done. Remember to change the runlevel back to 5 (follow the same instructions as before, but change from a 3 to 5). Hope this is able to help someone.

Some additional info:
I've found the game Chromium to be useful when telling if OpenGL is configured correctly. When installing, be sure to install the Games package. If your OpenGL is working correctly, the game should play smoothly. If its unplayably slow or does not load at all, your OpenGL setting is not correct.

Wow, I have the same graphics card as you and had no problems installing the drivers. I originaly installed the drivers with my Ge-Force 2 MX. I remember building the drivers from source as the RPM's did not work. But that is all the trouble I had. You tiumphed in the end and that is what is important

excelent job! id like to have all our how toos in some sort of sticky with the first post being a directory of the contents. web based how toos ive found to be incomplete or outdated. like you i had to pice together from several sites to get the whole picture

That is a really awesome idea phantom. Finding enough information to do the install properly took way more time than my installation attempts themselves, thats not the way it should be. I'm sure that if we had a directory of how to's like you suggest, we could really make a lot of people's linux experiences much easier. I know I was about to give up on my linux install when it finally worked.

Wow, I am really too lazy to read that! But I am sure you for got an essential step: cross your fingers while the driver is compiling.

Hey mods, any chance of this being intergrated with the existing NVidia driver setup sticky? I would really like to implement phantom's idea.

I assume you need the kernel source installed to make them? I tried without and it gave me a load of file not founds.

I installed the source that came on the cd, but when i compile it gives me an unresolved symbols (I think thats what it said) but i upgraded the kernel so im guessing i need the latest source too

which is a nice 35mb :/

Also i noticed when i installed the source from the cd it put it in /usr/src/linux-2.4.18-3 but the drivers look for it in /usr/src/linux so i created a link to that dir. Anyone else get that problem?

When updating the kernel, I simply used to "up2date" program included with Red Hat. Get the kernel source from there, not from the CD. The version on the CD is 2.4.18-3. You must use kernel version 2.4.18-5, not 2.4.18-3, to use the latest NVidia drivers. When I ran the update program, it installed everything itself. There is no need to do anything else than what the update program does in order to run the updated kernel. Check the kernel you are currently running with uname -r before installing the drivers to verify that you are running 2.4.18-5. Once I had that kernel running, I used the directions in this thread exactly. I didn't have to create any links to folders inorder to get the drivers running.

Hmmm, thats what I did, i only updated the kernel though, so i dont know if that included the source? I didnt take any of the other updates as im only on 64k connection.

Argh.

I downloaded the source from ftp://updates.redhat.com the 18-5 one, but cant get it to install, i do rpm -ivh kernel_4.whatever and it installs, but the directories arent there and if i do rpm -q it says its not installed.

Is there an up date up2dat that will do the source code as well or should that have been included when i updated the kernel? It was about a 12mb dl to update the kernel, but when i got the src from ftp it was 35mb

thx,

I've heard of problems that people have had if they didn't install the "Kernel Development" package during the install. Somehow it affects their ability to upgrade the kernel later. You might want to try a reinstall, making sure to select kernel development, and see if that helps your problem. Did you sucessfully update the kernel with up2date? Did you verfiy that it is the 2.4.18-5 version? When I used up2date, I installed all the available updates, and I would suggest you do the same. Leave the machine on overnight if you can.

Hmm, ok irll try reinstalling, last time i installed i just selected the development option, but i didnt use the select each individual package page, just the main groups.

I know my current kernel is 2.4.18-5 as it says when i boot to the command prompt. I had to modify my lilo.conf to get it to boot though, as it was still booting from the old kernel, is that normal?

Anyways, I'll try reinstalling tomorow and let u know what happens. Thx for ur help...so far

I didn't have to edit my lilo.conf, somehow the entry for the new kernel was put there. At the time, I assumed it was done by up2date. As long as your running the 2.4.18-5 kernel, I'm not sure what is causing the problem. Give the reinstall a try and see where it takes you. Glad to help as much as I can; I'm still rather new at this.

Wow, it actually works now

Instead of doing a workstation install i did custom and selected kernel development, and then when i updated it listed the new kernal AND the source, which wasnt listed before. It also updated the kernel without me having to modify the lilo.conf. It compiled straight off too. Thanx for help, i can sleep easy tonight

Im kinda liking this linux stuff...

Happy to hear it worked out for you. Fun when it works, ain't it?