Bug#959210: no automatic Xorg configuration for NVidia

[Andreas Beckmann]

On 01/05/2020 14.12, Eduard Bloch wrote:
> $ modprobe -v nvidia-drm
> $ modinfo nvidia-drm
> modinfo: ERROR: Module nvidia-drm not found.
> 
> Now what?

Yesterdays log showed these to be avaialble:

/lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current-modeset.ko
/lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current-uvm.ko
/lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current-drm.ko
/lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current.ko

> # modprobe -v nvidia-drm
> # modprobe -v nvidia-current-drm
> insmod /lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current.ko
> insmod /lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current-modeset.ko
> insmod /lib/modules/5.6.0-trunk-amd64/updates/dkms/nvidia-current-drm.ko
> 
> So, your hint from above was only partly helpful. Maybe this should be
> brought to modutils maintainers, this difference should not exist.

That's where /etc/modprobe.d/nvidia.conf comes into play.
We allow all different driver series to be installed concurrently. All
have their modules named "nvidia" etc. and the driver expects the module
with this name. To avoid file conflicts, we rename all the modules (in
the file system, not the internal name) and let the nvidia alternative
install the matching /etc/modprobe.d/nvidia.conf to allow loading the
correct module by the canonical name.
So a 'modprobe nvidia' will load nvidia-current.ko, but lsmod will
report afterwards that 'nvidia' is loaded.

On 01/05/2020 14.14, Eduard Bloch wrote:
> BTW, I also found this in the logs. Is that a problem?
>
> pr 30 22:12:02 zombie systemd[1]: Starting NVIDIA Persistence Daemon...
> Apr 30 22:12:02 zombie nvidia-persistenced[17237]: Started (17237)
> Apr 30 22:12:02 zombie nvidia-persistenced[17237]: Failed to open
libnvidia-cfg.so.1: libnvidia-cfg.so.1: cannot open shared object file:
No such file or directory
> Apr 30 22:12:02 zombie nvidia-persistenced[17236]: nvidia-persistenced
failed to initialize. Check syslog for more details.
> Apr 30 22:12:02 zombie systemd[1]: nvidia-persistenced.service:
Control process exited, code=exited, status=1/FAILURE
> Apr 30 22:12:02 zombie nvidia-persistenced[17237]: Shutdown (17237)
> Apr 30 22:12:02 zombie systemd[1]: nvidia-persistenced.service: Failed
with result 'exit-code'.
> Apr 30 22:12:02 zombie systemd[1]: Failed to start NVIDIA Persistence
Daemon.

nvidia-persistenced is not really needed, but libnvidia-cfg.so.1 should
be available (it was in yesterdays log).
/usr/lib/<triplet>/libnvidia-cfg.so.1 is handled by nesting the glx and
nvidia alternative.
nvidia-persistenced might fail to start if the kernel module is not
loaded (either because it does not exist or because no supported
hardware was found), check dmesg output. Perhaps libnvidia-cfg.so.1
refuses to load without the kernel module being loaded? (iirc there is a
bug report about installing nvidia-persistenced failing on non-nvidia
systems)


Andreas

PS: the recommended way to install and activate the nvidia driver on a
"normal" system (recommends enabled, default kernel installed via
linux-image-<arch> metapackage) is to install the nvidia-driver (or
nvidia-{legacy-ABCxx,tesla-XYZ}-driver) metapackage, it should pull in
everything that is needed and after reboot Xorg will start using the
nvidia driver - this seems to work very well for fresh installations

PPS: I really do like long grown systems and the fact that there is no
need to reinstall Debian for new major releases.