gridhead / nvidia-auto-installer-for-fedora-linux Goto Github PK

Steam and many other third party helpers like DXVK require x86 dependencies to be present.

The following GPUs are no longer supported with proprietary drivers from RPM Fusion and the installer should exit out without installation when these are detected.

• Legacy Geforce 6/7
• Legacy Geforce 5 (FX series)
• Legacy Geforce 2 through Geforce 4

Specific support is needed for

• Proprietary 390.xx drivers for the legacy discrete cards
• Proprietary CUDA support upto 9.2 for the legacy discrete cards

Need contributors to test and confirm the tool's workability in MATE-Compiz spin.

hi,
i have a issue when installing. Here is my prompt message:

[ # ] NVIDIA AUTOINSTALLER FOR FEDORA 32 AND ABOVE
[ ✘ ] Installer cannot proceed without root privileges
[ ✘ ] Leaving installer
[fenrir@localhost Downloads]$ sudo ./NVAutoInstFedora32 
[sudo] password for fenrir: 
[ # ] NVIDIA AUTOINSTALLER FOR FEDORA 32 AND ABOVE
[ # ] GATHERING ACTIVE INFORMATION...
[ ✔ ] Host information was gathered
      System: Linux v5.6.14-300.fc32.x86_64
      Hostname: localhost.localdomain
      Version: #1 SMP Wed May 20 20:47:32 UTC 2020
      Distribution: Fedora 32 x86_64
[ ✔ ] Supported OS detected - Fedora 32 x86_64
[ # ] CHECKING FOR GPU COMPATIBILITY...
[ ✔ ] Compatibility infomation was obtained
      00:02.0 VGA compatible controller: Intel Corporation HD Graphics 630 (rev 04)
      01:00.0 VGA compatible controller: NVIDIA Corporation GP107M [GeForce GTX 1050 Mobile] (rev a1)
[ ✔ ] An active NVIDIA GPU was detected!
[ ✔ ] An Optimus Dual GPU setup was detected!
[ # ] LOOKING FOR EXISTING DRIVER PACKAGES...
[ ! ] No existing NVIDIA drivers were detected!
[ ! ] Do you wish to install the drivers? (Y/N) y
[ # ] ATTEMPTING CONNECTION TO RPM FUSION...
[ ✔ ] Connection to RPM Fusion server was estabilished
[ ! ] RPM Fusion repository for Proprietary NVIDIA Driver not detected
[ ! ] Do you wish to fetch packages from this repository? (Y/N) y
[ # ] FETCHING REPOSITORY DATA...
[ ✘ ] RPM Fusion NVIDIA repository could not be enabled
[ ✘ ] Leaving installer

How do I fixed it?
Thank you.`

A command line option would be provided for users who want to install CUDA tools using the installer.

Need contributors to test and confirm the tool's workability in LXQt spin.

RPM Fusion support ffmpeg compiled with NVENC/NVDEC with Fedora 25 and later. You need to have a recent NVIDIA card (see the support matrix) and install the cuda sub-package.

This cannot be installed if CUDA is not detected and if the GPU is not recent.

Need contributors to test and confirm the tool's workability in SOAS spin.

Need contributors to test and confirm the tool's workability in Cinnamon spin.

Specific support is needed for

• Proprietary 340.xx drivers for the legacy discrete cards
• Proprietary CUDA support upto 6.5 for the legacy discrete cards

I ran

sudo nvautoinstall --rpmadd
sudo nvautoinstall --nvrepo
sudo nvautoinstall --driver

And it said that the driver installed successfully. I restarted my system and found out that the nvidia drivers aren't loaded, instead the nouveau driver was loaded.

How can i fix this?
Im running Fedora 34.
My graphics card is a GTX970.

When packaged using PyInstaller, exit() is not recognized

Traceback (most recent call last):
  File "MainFunction.py", line 50, in <module>
  File "MainFunction.py", line 47, in main
NameError: name 'exit' is not defined
[5385] Failed to execute script MainFunction

Need contributors to test and confirm the tool's workability in KDE Plasma spin.

As shown in the image, the installation continues even when the kernel forcing has resulted in a failure.

There are certain situations where people using Tensorflow and other such libraries require a specific version of CUDA. Albeit the installer provides the most recent one, CUDA installation of multiple versions can coexist (with minor replacements with the dependencies used by either) so the users should be given a choice to do that.

There are certain information not provided correctly on the README.md.

• The binary is not executable from the get-go so requires a chmod +x NVAutoInstFedora32 first
• The documentation does not state about root permission and how to obtain it

As shown in the image, the tool proceeds further from installation operation even when it has exited out before/without completion.

The modified kernel modules are likely to be loaded up on the next boot up

@lcts, now that the most recent changes are merged, please go ahead and build a package with the latest changes.

Need contributors to test and confirm the tool's workability in Xfce spin.

installed without any problems, but I got this error on the boot. Nvidia app won't open. Im using Lenovo ideapad Gaming 3-15ARH05 Laptop - Amd Ryzen 7 4800H and Gtx1650ti.
screenshot: https://www.hizliresim.com/rzmzq4v

For now, it is up to user's discretion as to where they would want to use this tool. There is no way to check the host OS and tell if it is Fedora or not.

Text coloring is supported by click. We can safely remove colorama after writing codes which allow for colors using click.

The tool is currently not working of Fedora 33 beta

Reduce the mandatory sleep time while doing so.

Task has limited timeframe. Any help is appreciated.

The server at developer.download.nvidia.com could not be pinged to and returns a 404 response code. Due to the failure of the ping operation, the installer cannot continue. Pinging for the availability of the repo link is suggested as it is much more effective.

Also, the repository link has changed from http://developer.download.nvidia.com/compute/cuda/repos/fedora29/x86_64/cuda-fedora29.repo to http://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo - So the change must be reflected to avoid fetching obsolete versions of CUDA.

In the spirit of getting more contributors for the project, the above documents would be appreciated.

This might need your attention @lcts. There is stuff changed in one of the branches which do not run on Fedora 32 Workstation when they are packaged into an executable binary by PyInstaller. On searching deep, I figured that it is the Python version to blame as PyInstaller does not provide support for Python 3.8 (or above).

Any solution for this?

As @ferdnyc remarked, checking for newlines on an lspci output is truly a shabby way to detect an Optimus setup. This must be reworked upon to facilitate better detection with corner cases.

Need contributors to test and confirm the tool's workability in LXDE spin.

Hello,

First, thank you very much for this super tool.

This is what I did :

• Fresh installation of fedora 32.
• SELinux off (same problem than issue : "Generalize and test the tool in KDE Plasma spin #5")
• dnf update
• installed nvidia driver (all steps)
• reboot
• installed some tools like VSC.
• reboot
• installed KDE and Xfce
• reboot
• dnf update (error)

I received that message :

Erreur : Erreur de la transaction de test :
le fichier /usr/lib/udev/rules.d/60-nvidia.rules de l'installation de nvidia-kmod-common-3:440.95.01-1.fc29.x86_64 entre en conflit avec le fichier du paquet xorg-x11-drv-nvidia-3:440.82-1.fc32.x86_64

I don't understand that my update try to update from fc32 to nvidia driver fc29?

I'm stuck and unable to update my environment.

Thank's to help me

Sylvain

You new release is an archive... not the usual script... So what do you want us to do with it?
Thanks

[CTRL]+[C] should be handled gracefully.

The initial releases should be constrained to Workstation only by making changes in the CONF file but should slowly expand to include other notable spins as well.

Not that, it should be compatible with Arch Linux anyway but still, the compatibility check should run fine in all major distributions. Arch Linux does not seem to have a distribution version which causes the error.

Simply execute the following

# akmods --force
# dracut --force

I don't know if this is the right place but on Fedora 34 with kernel above 5.12, laptop display are still not working after installing the drivers. Funny thing is that an external monitor connected over hdmi is working.
HP ProBook 450 G7, I7-10510U CPU, Nvidia GeForce MX250 OS: Fedora release 34 (Thirty Four)

Not sure it concerns your script because I don't recall if your script does take into account the vdpau plugin.
But What I can see on my install after a few upgrades through dnf, is that the symbolic link of /usr/bin/libvdpau_nvidia.so doesn't point to the correct lib
here : lrwxrwxrwx. 1 root root 31 Jul 21 21:01 libvdpau_nvidia.so -> vdpau/libvdpau_nvidia.so.440.36

and should be:
lrwxrwxrwx. 1 root root 31 Jul 21 21:01 libvdpau_nvidia.so -> vdpau/llibvdpau_nvidia.so.450.66
or
the generic symbolic link already in place :
lrwxrwxrwx. 1 root root 31 Jul 21 21:01 libvdpau_nvidia.so -> vdpau/llibvdpau_nvidia.so.1

The textual content of the standard messages need not be themed to be white. Let the terminal decide the color at that time or else the white overlay would become exclusive to black background terminals only.

Scrape the NVIDIA website during the installer initialization and fetch the card names which are officially supported by NVIDIA.

P.S. I think it would be better off to be done from the RPMFusion website if they maintain a list of supported cards for their proprietary NVIDIA drivers.

The N input stays unrecognised during the package check confirmation

Ok so i folllowed the steps outlined by @sylvoie in this thread and I had driver failing and falling back to noveau.

then i cd-d to '/usr/sbin' and used 'akmods --kernels $(uname -r)' to rebuild kernel which prompted me to install akmods which apparently weren't installed

After that i rebooted and it went straight to login screen without errors, finally nvidia-smi outputed that the driver is installed.

So once again the exact steps:

sudo ./NVAutoInstFedora32 --rpmadd
sudo ./NVAutoInstFedora32 --driver
sudo ./NVAutoInstFedora32 --x86lib
sudo ./NVAutoInstFedora32 --nvrepo
sudo ./NVAutoInstFedora32 --ffmpeg
sudo ./NVAutoInstFedora32 --vidacc
reboot
sudo ./NVAutoInstFedora32 --plcuda
sudo ./NVAutoInstFedora32 --vulkan
reboot

--

cd /usr/sbin

sudo akmods --kernels $(uname -r)

Install package 'akmods' to provide command 'akmods'? [N/y] y

sudo akmods --kernels $(uname -r)

Checking kmods exist for 5.9.10-200.fc33.x86_64 [ OK ]
Building and installing nvidia-kmod [ OK ]

reboot.