nvidia driver with multiple kernels?

[dukeofdevon]

Two questions. I had to recompile my Red Hat 9 kernel (2.4.20-20.9-athlon) to enable usb storage debugging. I ran the nvidia driver updater in the new kernel, which is fine, but now (1) X won't start on the old kernel and (2) the script can't compile the driver on the old kernel since it can't locate nvidia.o due to not knowing where the linux source is. This is odd, because I didn't change the linux source (I compiled from the same kernel source that corresponded to what was running on my computer).

Question 1: Is there no way to have the nvidia driver set up to work in multipe kernels on the same machine?

Question 2: Why can't the script recompile the driver in the old kernel? I gather that the process of making and installing the new kernel changed something that makes it impossible for the nvidia install to work with the old kernel. One annoying possibility would be, I guess, to recompile the original kernel again (Red Hat provides the config files). Surely there's a better way.

Thanks,
duke

[blueworm]

You can have as many kernels as you want.
What you did was...
make modules modules_install. This overwrites /lib/modules/2.4.X
Once you have re-installed the nvidia driver it will work in both kernel as long as they share the same /lib/modules in other words if they are in fact the same kernel but with differrent options built.
If you want to compile different kernel versions, You will have to boot the new kernel and install the kernel module so that it is in the new /lib/modules/2.4.XX.

[dukeofdevon]

Thank you for the reply. You are exactly right, I did a make modules_install after I did make modules.

I'm afraid I'm enough of a novice that I don't understand how to fix the problem. I take it that now that I have different /lib/modules directories, I have to do the latter of your suggestions, boot the the new kernel and install ther kernel module from so that it's in /lib/modules/2.4.blah. I obviously don't understand, because this is what I thought I did, by rerunning the nvidia package in the new kernel. This is what caused X on the old kernel to stop working and also, I guess, prevented me from rerunning the nvidia package in the old kernel. How do I restore X on that kernel?

Also, what *should* I have done? Not done 'make modules_install' if the kernels are builds of the same version?

Thank you!
duke

[blueworm]

if both kernels are 2.4.20-20.9-athlon they will share /lib/modules/2.4.20-20.9-athlon.
Therefore if the nvidia kernel module is installed for 1 it is for both.

[Valheru]

Run the NVIDIA installer script with the --help flag, there's and option there to install it for additional kernels without it uninstalling previously installed drivers for different kernel versions.

[dhill]

But don't You think it should be a question in FAQ... I came a long way to read those posts and I think it's basic security to be able to run an old kernel.

[blueworm]

Originally posted by dhill
But don't You think it should be a question in FAQ... I came a long way to read those posts and I think it's basic security to be able to run an old kernel.

Common sense too.