Unable to install nVidia drivers

atheist · 09-26-2007, 08:24 PM

I've been trying to set up the nVidia drivers from http://www.nvidia.com/object/linux_d...100.14.11.html but they've been giving me shit tons of problem.

Everytime I try to install them I get this message:

Quote:

ERROR: Unable to load the kernel module 'nvidia.ko'. This happens most
frequently when this kernel module was built against the wrong or
improperly configured kernel sources, with a version of gcc that differs
from the one used to build the target kernel, or if a driver such as
rivafb/nvidiafb is present and prevents the NVIDIA kernel module from
obtaining ownership of the NVIDIA graphics device(s).

Please see the log entries 'Kernel module load error' and 'Kernel
messages' at the end of the file '/var/log/nvidia-installer.log' for
more information.

I recompiled my kernel and removed support for nvidiafb (rivafb wasn't added) but I still get the same error message.

Here's the information the error message tells me to check:

Quote:

-> Kernel module load error: insmod: error inserting './usr/src/nv/nvidia.ko':
-1 Invalid module format
-> Kernel messages:
kjournald starting. Commit interval 5 seconds
EXT3 FS on sda5, internal journal
EXT3-fs: mounted filesystem with ordered data mode.
kjournald starting. Commit interval 5 seconds
EXT3 FS on sda6, internal journal
EXT3-fs: mounted filesystem with ordered data mode.
kjournald starting. Commit interval 5 seconds
EXT3 FS on sda3, internal journal
EXT3-fs: mounted filesystem with ordered data mode.
kjournald starting. Commit interval 5 seconds
EXT3 FS on sda7, internal journal
EXT3-fs: mounted filesystem with ordered data mode.
r8169: eth1: link up
r8169: eth1: link up
input: Power Button (FF) as /class/input/input4
ACPI: Power Button (FF) [PWRF]
input: Power Button (CM) as /class/input/input5
ACPI: Power Button (CM) [PWRB]
Linux agpgart interface v0.102 (c) Dave Jones
nvidia: disagrees about version of symbol struct_module
nvidia: disagrees about version of symbol struct_module
NET: Registered protocol family 10
lo: Disabled Privacy Extensions
eth1: no IPv6 routers present
nvidia: disagrees about version of symbol struct_module

Does anyone have the slightest idea of what's going on? Because that doesn't tell me anything... It would be awesome if anyone could help point me in the right direction.

Thanks.

Also:

Quote:

Originally Posted by Paul Sernine

Yeeeeeeeeeesss!!!!

I got it working !

the clue was provided here. It's in spanish, but i'm kind enought to provide a translation:

you need to extract the install script by running "./NVIDIA-Linux-x86-1.0-6111-pkg0.run --extract-only", do not run the installer.

cd to the created folder and edit file ./usr/src/nv/nv.h, you just have to add this line:

#define PM_SUSPEND PM_SAVE_STATE

save, close, and from the installer folder run "sh nvidia-installer"

It worked for me, the nvidia kernel module was compiled and installed without a flaw.

Nevertheless, there is still a little issue. At the beginning of a fresh reboot, if i run startx, X won't load and will complain about the screen configuration section of the XF86Config.

Reading forward in the mentioned post, i find there is a workaround.

When still in the console, you have to su root and run "modprobe nvidia" and "service dm start".

Then you can startx away, it's a bit of a hussle, but i don't reebot very often anyway. Let's expect for future versions to cover the issue.

Have fun.

Didn't work either.

GrapefruiTgirl · 09-26-2007, 09:19 PM

Here's two things to try:

1 - Make sure you are not in the runlevel in which X runs, when you build & install the module. On other words, if X starts in runlevel 5 on your machine, then either use 'gdm stop' or 'telinit 1 then telinit 3' in a root console, or boot up into a non-X runlevel. Then build the driver. When done, start X or telinit 5.

2 - OR: enter your kernel source code folder and type 'make prepare' as root. Now try building the module and installing it, keeping in mind the suggestion in my tip # 1 above. X must not be running when building the nvidia module.

Any luck?

NOTE - if you made any changes suggested in that post you posted by Paul Sernine, undo them. I believe that has nothing to do with your problem.

atheist · 09-26-2007, 09:40 PM

Thanks for the suggestions. I still just get the same error message, though.

GrapefruiTgirl · 09-26-2007, 10:48 PM

An afterthought: Are you certain that the kernel you are running, is the EXACT same as the kernel sourcecode you have on your system, and you have the needed kernel headers installed too? And have you actually built your kernel using that sourcecode?

Another tip - the way I suggested you kill GDM may not work well enough. You may need to do 'killall gdm' after entering init 3. Sorry, if it works, but I don't use Gnome or Ubuntu

and gdm is a bit different than my KDM.

atheist · 09-27-2007, 03:35 PM

Quote:

Originally Posted by GrapefruiTgirl

An afterthought: Are you certain that the kernel you are running, is the EXACT same as the kernel sourcecode you have on your system, and you have the needed kernel headers installed too? And have you actually built your kernel using that sourcecode?

Another tip - the way I suggested you kill GDM may not work well enough. You may need to do 'killall gdm' after entering init 3. Sorry, if it works, but I don't use Gnome or Ubuntu

and gdm is a bit different than my KDM.

Actually, I didn't find any GDM process... I'm actually just using Enlightenment, do you know if there's something I can do about that?

I compiled my kernel from source, though... downloaded the source and the process went like this:

Code:

cd /usr/src/
tar xjf linux-2.6.22.8.tar.bz2
ln -s linux-2.6.22.8 linux
cd linux
make menuconfig
make-kpkg --revision "1" --us --uc --initrd kernel_image kernel_headers
cd ..
dpkg -i linux-image-2.6.22.8_1_i386.deb
dpkg -i linux-headers-2.6.22.8_1_i386.deb

And then I rebooted into my kernel.
This way I should have the correct kernel headers installed, right?

farslayer · 09-27-2007, 03:47 PM

I get Debian kernel headers like this:
aptitude update
aptitude install build-essential module-assistant
m-a update
m-a prepare

When you run the Module assistant prepare command it will automatically download the Debian kernel source/headers for your running kernel and install them properly

If you are running a kernel you compiled from vanilla source downloaded from kernel.org then this wouldn't work for you..

Quite frankly in order to install the nvidia drivers on etch it is not necessary to recompile the Debian kernel. If you prepare the build environment as I showed above, then you can just run the installer from nVidia and it should all fall into place.

atheist · 09-27-2007, 08:45 PM

Thanks for the help, both of you.

What I didn't realize is that dpkg automatically made the new kernel the default one... So everytime I booted I manually selected the old kernel, thinking it was the new one.

I realized this, and the module installed like it should.

Sorry to waste your time.

GrapefruiTgirl · 09-27-2007, 09:14 PM

Not a waste at all, atleast in my case. Like your signature says, '..knowledge is power' and I didn't know that about dpkg.
In the end, we both learned something, and your driver works

and that's what counts.