[PlanetCCRMA] module load - unresolved symbols?

Fernando Pablo Lopez-Lezcano nando@ccrma.Stanford.EDU
Sun Oct 17 15:05:02 2004

On Sun, 2004-10-17 at 14:51, Joe Hartley wrote:
> I've just upgraded my studio system from a 1GHz PIII to a 3GHz P4,
> the motherboard to an ASUS P4P800SE, and my video card from an NVidia 
> FX5200 to an FX5700LE.  It came up just fine with the kernel from the 
> original configuration, a 2.4.26-1.ll.rh90.ccrma kernel and the nvidia 
> module 1.0-6106.
> When I tried running nvidia-install --update, the module compilation
> failed with the following message:
> -> Kernel module load error: ./usr/src/nv/nvidia.o: unresolved symbol
>    smp_call_function
>    ./usr/src/nv/nvidia.o: unresolved symbol smp_num_cpus
>    ./usr/src/nv/nvidia.o: 
> It's always had to build the module before, but it's never failed like
> this.  It's not a multiple CPU mobo, I've just got the one P4, so does
> anyone know why it's looking for an smp symbol, and how to get around
> this?

If hyperthreading is enabled then you are probably running an smp kernel
(do uname -r to see what you are booting into) after all. HT uses spare
cycles in the P4 processor to run two "processors" in parallel,
depending on what software you run you can have from 0% to 15-20% more

-- Fernando