I had planned a day of relaxing with Portal 2 but on starting Steam (for the first time in a couple of weeks) I was greeted with the following message in the terminal:
Error: API mismatch: the NVIDIA kernel module has version 270.41.19,
but this NVIDIA driver component has version 270.41.06. Please make
sure that the kernel module and all NVIDIA driver components
have the same version.
I'll confess I don't really know what it's talking about when it says driver. The verion of nvidia-current
is 270.41.19
. I thought that was the driver and module, all in one.
I use the X-SWAT PPA and I have noted that the nvidia-settings
package has boosted to 275.09.07
. As this is just a settings application, I don't think this mismatch has anything to do with this. It's also not the same version as the problem being described.
I'd rather not purge back to the standard Nvidia driver as it's less than stable on my GTX580. I would accept an answer that takes the manual setup and makes it recompile when the kernel recompiles (ie, some DKMS wizardry) but it has to work. I don't want to drop back to text-mode every time I restart after a kernel upgrade.
Edit: Minecraft works without a single complaint about driver versions. Penumbra dies with roughly the same error when entering a game.
Edit: Here's what I'm dealing with in terms of 32bit files. They all seem to be at the right version.
ls -l /usr/lib32/nvidia-current/ | awk '{ print $8 " " $9 " " $10 }'
libcuda.so -> libcuda.so.270.41.19
libcuda.so.1 -> libcuda.so.270.41.19
libcuda.so.270.41.19
libGL.la
libGL.so -> libGL.so.1
libGL.so.1 -> libGL.so.270.41.19
libGL.so.270.41.19
libnvcuvid.so -> libnvcuvid.so.1
libnvcuvid.so.1 -> libnvcuvid.so.270.41.19
libnvcuvid.so.270.41.19
libnvidia-cfg.so -> libnvidia-cfg.so.1
libnvidia-cfg.so.1 -> libnvidia-cfg.so.270.41.19
libnvidia-cfg.so.270.41.19
libnvidia-compiler.so -> libnvidia-compiler.so.1
libnvidia-compiler.so.1 -> libnvidia-compiler.so.270.41.19
libnvidia-compiler.so.270.41.19
libnvidia-glcore.so.270.41.19
libnvidia-ml.so -> libnvidia-ml.so.1
libnvidia-ml.so.1 -> libnvidia-ml.so.270.41.19
libnvidia-ml.so.270.41.19
libnvidia-tls.so.270.41.19
libnvidia-wfb.so.1 -> libnvidia-wfb.so.270.41.19
libnvidia-wfb.so.270.41.19
libOpenCL.so -> libOpenCL.so.1
libOpenCL.so.1 -> libOpenCL.so.1.0
libOpenCL.so.1.0 -> libOpenCL.so.1.0.0
libOpenCL.so.1.0.0
libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.270.41.19
libXvMCNVIDIA.so -> libXvMCNVIDIA.so.270.41.19
libXvMCNVIDIA.so.1 -> libXvMCNVIDIA.so.270.41.19
libXvMCNVIDIA.so.270.41.19
Edit 2: I might have found something.
Lurking in /usr/lib32/
proper (not the nvidia-current
subdir) I found the following:
ls -l /usr/lib32/ | awk '{ print $8 " " $9 " " $10 }' | grep 270
libcuda.so.1 -> libcuda.so.270.41.06
libcuda.so.270.41.06
libGL.so.1 -> libGL.so.270.41.06
libGL.so.270.41.06
libnvidia-compiler.so.270.41.06
libnvidia-glcore.so.270.41.06
libnvidia-tls.so.270.41.06
libvdpau.so.1 -> libvdpau.so.270.41.06
libvdpau.so.270.41.06
libvdpau_trace.so -> vdpau/libvdpau_trace.so.270.41.06
For some reason everything in /usr/lib32/nvidia-current/
is the right version but there's a load of cruft in /usr/lib32/
that might be ruining the party.
Edit 3: Trying to track down which package owns these files has failed:
find /usr/lib32 -iname '*270.41.06*' -exec dpkg -S "{}" \;
dpkg-query: no path found matching pattern /usr/lib32/libnvidia-compiler.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/libGL.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/vdpau/libvdpau_nvidia.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/vdpau/libvdpau_trace.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/libvdpau.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/tls/libnvidia-tls.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/libnvidia-tls.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/libcuda.so.270.41.06.
dpkg-query: no path found matching pattern /usr/lib32/libnvidia-glcore.so.270.41.06.
Any tips on how I should deal with these broken versions? Delete them? Delete then symlink to ./nvidia-current/
versions?
/usr/lib32
. Check if they look like the same version as the primary 64-bit ones. If they differ, find out what package owns them and look for an update.