Bug#800554: nvidia-kernel-dkms: kernel module won't build for 3.16.0-4-amd64 (missing rule in Makefile)
Luca Boccassi
luca.boccassi at gmail.com
Wed Sep 30 23:55:33 UTC 2015
On Wed, 2015-09-30 at 18:43 -0400, Norman Ramsey wrote:
> Thank you for your timely and helpful response! Unfortunately, as you
> will see below, the issue is not quite settled.
>
> > From the system info, I gather that you are running i386 userspace with
> > an amd64 kernel, is that correct?
>
> Correct.
>
> > If that's the case, please see 799960 and the relative discussion for
> > more information. [1]
>
> I've read this discussion, but I can't claim to have understood it all.
>
> > As it was written by Andreas in that thread, support was dropped for
> > building amd64 kernel with the i386 userspace packages because
> > amd64:i386 kernel is no longer available in unstable.
>
> A most unfortunate decision. I really like the combination of i386
> userspace with amd64 kernel. The cost of losing it would be pretty
> high. But I realize this is off topic.
>
> > As he suggested, you should be able to build the module by
> > installing the package nvidia-kernel-dkms:amd64 thanks to
> > multiarch.
>
> I can indeed build the module. However, there is a minor problem and
> a major problem.
>
> - The minor problem is that nvidia-kernel-dkms:amd64 appears to be
> incompatible with nvidia-kernel-dkms:i386, which means I lose the
> i386 kernel module. This outcome is a little unfortunate.
> I can't remember enough about multiarch to know if there is a
> setting to enable the two packages to coexist---or if there would
> be problems in this case. But certainly there is no conflict in
> /lib/modules.
This conflict is expected, as the package is marked as Multiarch:
foreign [1], since it installs the sources to /usr/src/<nvidia-version>
and there would be a conflict.
> - The X server will not start. I have attached the Xorg.0.log.
>
> For the moment I have reverted to the i686 kernel so I can get some
> work done. But if you would like me to reinstall the
> nvidia-kernel-dkms:amd64 and re-run reportbug so you can see the
> entire non-working configuration, please let me know. I will be happy
> to do it.
Now that is quite strange, I've never seen that error:
/dev/dri/card0: failed to set DRM interface version 1.4: Inappropriate ioctl for device
[ 260.044] (--) PCI:*(0:1:0:0) 10de:1401:1043:8520 rev 161, Mem @ 0xf6000000/16777216, 0xe0000000/268435456, 0xf0000000/33554432, I/O @ 0x0000e000/128, BIOS @ 0x????????/524288
The only thing that comes up in Google is bug #764639. I don't suppose
the blacklisting of nouveau is missing? Having the output of reportbug
might help.
Also trying manually from a TTY to load the module and start X might
give more clues.
But I have to stress that this is a bit uncharted territory, I have
never tried a setup with different archs for kernel and userspace, so I
am not sure I'll be able to help. I can try :-)
Kind regards,
Luca Boccassi
[1] https://wiki.ubuntu.com/MultiarchSpec#Binary_package_control_fields
[2] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=764639
More information about the pkg-nvidia-devel
mailing list