I have tried everything I can tink of but I have not been able to get the Nvidia driver to compile and install on a 32bit machine with current.

It is not SMP just a bog standard 32bit machine. Worked fine with Slackware 14.1 32 bit.

I get ERROR: Could not insert 'nvidia': Unknown symbol in module or unknown parameter (see dmesg)

There is nothing in dmesg about it so I am stumped.

Any suggestions

Perhaps this thread will help.

This patch

https://github.com/manjaro/packages-...n-symbol.patch

and this procedure

You need to rebuilt your kernel using this mtrr.patch in order to build NVidia 304.131

go to /usr/src/linux-4.4.6
patch -p1 < /path/to/mtrr.patch
cp /boot/config-huge-4.4.6 .config
make oldconfig
make bzImage
cp arch/x86/boot/bzImage /boot/vmlinuz-huge-4.4.6
/sbin/lilo -v
reboot

After that, you can reinstall NVidia driver again.

Fixed it for me.

I guess it should work for the other kernals but I did not try as I had installed the huge to see if it was only smp that was having problems.



Thanks for the advice

It is nice when things work again!

On the dual core CPU box where I have been rebuilding the generic-smp-4.4 kernel, I find that adding '-j3' cuts down the compile time.
After copying the rebuilt kernel into place, you can also do 'make clean' to save some space and clean the source tree.

I wonder if the -j option would have any effect on a single core processor?

This old box has sadly but one core

I wonder if they ever even made any 32bit smp processors ?

Just looked it up and the Intel Core family and the Dual Core Xeon LV would appear to be 32bit

As far as I can tell AMD went straight to 64bit and never looked back

I found on this one anyway just by running the compile at the command prompt instead of the gui it ran a heck of a lot faster. Or it seemed to anyway

Yes a make clean is probably a good idea.

This really should be fixed by Nvidia or I guess in the kernal source code if that stuff is no longer needed

Although I did not really look into what exactly all that stuff was being used for.

Not everyone wants to patch and build a kernal just to load a driver.

Does anyone know if this has been fixed in the 4.7 kernel ?

nope, we need to wait for updates from NVidia since this is not kernel's fault. It's NVidia that should update their drivers

Not much hope of that in this case it is the legacy driver that is causing problems.


Linux Display Driver - x86

Version: 304.131
Release Date: 2015.11.16
Operating System: Linux 32-bit

has not been updated for a while.

Maybe they will fix but I would not hold my breath

I wonder if the fix still works for 4.7

Takes a while to compile and then if it does not work have to go back to 4.6

Works for me with the latest 4.4.7 kernel in 14.2-RC2.

well that sounds hopeful I think I'll try one on a VM as it's about 10 times faster than the actual 32bit hardware

I hope Nvidia fixes it soon I hate to waste time recompiling the kernel for one fix

Well it worked for me too

I did make one in a VM but I think because I gave the VM 2 cpu's and it was running the SMP kernel it did not work. I am not sure about that as I thought I told it to build using the huge_4.4.7 config files That should work but the build of nividia drivers complained about a mismatch of some type.

I rebuilt it on the actual machine and it worked