Public bug reported:

The intended behavior of `ubuntu-drivers` has always been mysterious to
me. Here are a few examples:

(1) It is not clear to me what --gpgpu is intended to do. The help
output simply says:

Options:
  --gpgpu              gpgpu drivers

According to https://ubuntu.com/server/docs/nvidia-drivers-installation:


> Check the available drivers for your hardware
> For desktop:
> 
> sudo ubuntu-drivers list
> or, for servers:
> 
> sudo ubuntu-drivers list --gpgpu
```

But both commands list the same set of packages - just in a different
order:

$ sudo ubuntu-drivers list
nvidia-driver-550-open, (kernel modules provided by 
linux-modules-nvidia-550-open-generic)
nvidia-driver-470-server, (kernel modules provided by 
linux-modules-nvidia-470-server-generic)
nvidia-driver-535-open, (kernel modules provided by 
linux-modules-nvidia-535-open-generic)
nvidia-driver-535-server-open, (kernel modules provided by 
linux-modules-nvidia-535-server-open-generic)
nvidia-driver-550, (kernel modules provided by linux-modules-nvidia-550-generic)
nvidia-driver-535-server, (kernel modules provided by 
linux-modules-nvidia-535-server-generic)
nvidia-driver-470, (kernel modules provided by linux-modules-nvidia-470-generic)
nvidia-driver-535, (kernel modules provided by linux-modules-nvidia-535-generic)

$ sudo ubuntu-drivers list --gpgpu
nvidia-driver-470-server, (kernel modules provided by 
linux-modules-nvidia-470-server-generic)
nvidia-driver-535-open, (kernel modules provided by 
linux-modules-nvidia-535-open-generic)
nvidia-driver-550, (kernel modules provided by linux-modules-nvidia-550-generic)
nvidia-driver-535-server, (kernel modules provided by 
linux-modules-nvidia-535-server-generic)
nvidia-driver-470, (kernel modules provided by linux-modules-nvidia-470-generic)
nvidia-driver-550-open, (kernel modules provided by 
linux-modules-nvidia-550-open-generic)
nvidia-driver-535, (kernel modules provided by linux-modules-nvidia-535-generic)
nvidia-driver-535-server-open, (kernel modules provided by 
linux-modules-nvidia-535-server-open-generic)


But there's no indication that the order means anything. `sudo ubuntu-drivers 
install --gpgpu` on this system will install 
nvidia-headless-no-dkms-535-server. Which, notably, installs no kernel drivers 
(neither DKMS nor signed) on my system. `sudo ubuntu-drivers install`, OTOH, 
will install nvidia-driver-550 linux-modules-nvidia-550-generic.

(2) According to https://ubuntu.com/server/docs/nvidia-drivers-
installation, ubuntu-drivers "always tries to install signed drivers
which are known to work with Secure Boot." But, if there isn't an l-r-m
package available for the current kernel, it will fall back to a -dkms
package. It seems like that would be the case in the window between
pushing out a new nvidia-graphics-drivers package and l-r-m's having
been built against it. Maybe that archive state "shouldn't happen" - but
if this mode is documented to install signed drivers, then unavailable
signed drivers should be an error.


(3) There's no option to automatically install the best "-open" variant. There 
is a `--free-only` option, but that filters out all nvidia drivers.


Suggestions:

From what I can tell, the `--gpgpu` actually intends to install drivers
for a headless system. (Maybe it is just a bug that it installs no
driver on my system?) Assuming that is the intent, then `--headless`
seems like a better option name. Perhaps we could add `--headless` as an
alias for `--gpgpu`... and maybe deprecate --gpgpu?

Could we add a `--server|--desktop` flag so a user can explicitly choose
the server variant? I realize that `--server` and `--headless` seem
similar - but we do provide the full graphics stack for the -server
variant drivers, and that does make sense on some systems (DGX A100
Station, for example). Again, documentation could clarify the
difference.

Could we allow the -open variants to be installed with --free-only? Or
could we add a flag to select the -open variant, and document the
difference between that and --free-only?

** Affects: ubuntu-drivers-common (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2081881

Title:
  nvidia driver installation modes are unclear and in conflict w/ the
  server guide

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/ubuntu-drivers-common/+bug/2081881/+subscriptions


-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to