Re: Nvidia package installation problems
On Mon, Dec 23, 2013 at 4:21 AM, Andrei POPESCU
<andreimpopescu@gmail.com> wrote:
> On Du, 22 dec 13, 19:15:25, Jon N wrote:
>>
>> "NWRM: API mismatch: the client has the version 304.108, but this
>> kernel module has the version 319.76. Please make sure that this
>> kernel module and all NVIDIA driver components have the same version."
>>
>> That sounds simple enough, but I have searched all packages I can
>> think make be related in Aptitude and can not find anything that
>> mentions version 304.108. Either they all say 319.76, or a version
>> like '20131102+1 (that's the 'nvidia-kernel-common among others).
>
> You have a leftover kernel module. See if nvidia-kernel-dkms can cleanup
> and install a correct module for you. Make sure you have the matching
> kernel headers for your kernel installed.
>
> Kind regards,
> Andrei
I ran 'dpkg-reconfigure dkms' (which exited after only a few seconds
with no messages) and 'dpkg-reconfigure nvidia-kernel-dkms" which
seemed to do a number of things. But basically it seemed to uninstall
the nvidia module version.319.76 from kernel 3.11-2-686-pae (it said
it was active on this kernel) and then load as new the exact same
module. It even had an error message:
"Error! Module version 319.76 for nvidia-current.ko is not newer than
what is already found in kernel 3.11-2-686-pae (319.76). You may
override by specifying --force"
At the end it says: "DKMS: install completed"
But, it did not fix the problem.
I tried checking the kernel headers using Aptitude and as far as I can
tell I have the following packages for linux-headers installed. These
match the 2 kernels I have installed. BTW, I tried booting the older
kernel and got the same results.
linux-headers-3.11-2-686-pae
linux-headers-3.11-2-common
linux-headers-686-pae
linux-headers-3.10-3-686-pae
linux-headers-3.10-3-common
That got me thinking again about versions. I searched for "304.*" (the
'mystery' version number that showed in some error messages) and found
a whole bunch of 'libcuda' files, all version 304 or older. And a
couple of symlinks setting the newest of those as 'libcuda.so' (or
something similar). I deleted them all as these must have been
leftovers from the downloaded nvidia binary drivers from nvidia. I
didn't see that error message on the next boot, but still no X.
Installing the latest libcuda (319.76) from Aptitude did help either.
The only error message I can find now is from Xorg.0.log:
"(EE) Failed to load /usr/lib/xorg/modules/drivers/nvidia_drv.so:
/usr/lib/xorg/modules/drivers/nvidia_drv.so: undefined symbol:
WindowTable"
Two lines below that is:
"(EE) Failed to load module "nvidia" (loader failed, 7)"
I don't know if the number '7' is significant.
I did start a search for the "... undefined symbol: WindowTable"
message, but so far nothing really stood out as a likely cause.
Thanks,
Jon
Reply to: