[BUG] Lutris does not open and does not recognise GPU

[TheRealOne78]

NVIDIA Open GPU Kernel Modules Version

515.48.07 - Gentoo kernel-open build

Does this happen with the proprietary driver (of the same version) as well?

I cannot test this

Operating System and Version

Gentoo Base System release 2.8

Kernel Release

5.18.1-gentoo-r1 - configured and compiled by myself

Hardware: GPU

NVIDIA GeForce RTX 2060 (UUID: GPU-466269ac-80e7-d5cf-f317-6157a8421d30)

Describe the bug

After emerging the gentoo driver for nvidia with the kernel-open flag and restarted, I couldn't open lutris anymore.
From the dumped info, it seems like lutris can't recognise this GPU

Lutris log:

2022-06-02 19:26:26,589: No GPU available on this system!
2022-06-02 19:26:27,273: Starting Lutris 0.5.10.1
2022-06-02 19:26:27,399: Writing full game list from MAME to /home/therealone/.cache/lutris/
2022-06-02 19:26:27,402: Couldn't get any output for mame -listxml
2022-06-02 19:26:27,402: MAME did not write anything to /home/therealone/.cache/lutris/mame/
2022-06-02 19:26:27,404: Using NVIDIA drivers for for Open
2022-06-02 19:26:27,404: GPU: NVIDIA GeForce RTX 2060
2022-06-02 19:26:27,404: No GPU available on this system!
Traceback (most recent call last):
  File "/usr/lib/python3.9/site-packages/lutris/gui/application.py", line 409, in do_command
    run_all_checks()
  File "/usr/lib/python3.9/site-packages/lutris/startup.py", line 156, in run_all_checks
    check_driver()
  File "/usr/lib/python3.9/site-packages/lutris/startup.py", line 81, in check_driver
    if drivers.is_outdated():
  File "/usr/lib/python3.9/site-packages/lutris/util/graphics/drivers.py", line 112, in is_o
    major_version = int(driver_version.split(".")[0])
ValueError: invalid literal for int() with base 10: 'for'
2022-06-02 19:26:27,405: Shutting down Lutris

To Reproduce

1. Emerge version 515.48.07 of x11-drivers/nvidia-drivers with the kernel-open flag, restart and login
2. Open lutris
3. Notice it doesn't open so open a terminal
4. Run lutris and get this dump error info:

2022-06-02 19:26:26,589: No GPU available on this system!
2022-06-02 19:26:27,273: Starting Lutris 0.5.10.1
2022-06-02 19:26:27,399: Writing full game list from MAME to /home/therealone/.cache/lutris/
2022-06-02 19:26:27,402: Couldn't get any output for mame -listxml
2022-06-02 19:26:27,402: MAME did not write anything to /home/therealone/.cache/lutris/mame/
2022-06-02 19:26:27,404: Using NVIDIA drivers for for Open
2022-06-02 19:26:27,404: GPU: NVIDIA GeForce RTX 2060
2022-06-02 19:26:27,404: No GPU available on this system!
Traceback (most recent call last):
  File "/usr/lib/python3.9/site-packages/lutris/gui/application.py", line 409, in do_command
    run_all_checks()
  File "/usr/lib/python3.9/site-packages/lutris/startup.py", line 156, in run_all_checks
    check_driver()
  File "/usr/lib/python3.9/site-packages/lutris/startup.py", line 81, in check_driver
    if drivers.is_outdated():
  File "/usr/lib/python3.9/site-packages/lutris/util/graphics/drivers.py", line 112, in is_o
    major_version = int(driver_version.split(".")[0])
ValueError: invalid literal for int() with base 10: 'for'
2022-06-02 19:26:27,405: Shutting down Lutris

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

More Info

No response

[mtijanic]

Thanks for the report. Looks to be the same issue as #272 , with the RM_FREE RPC timing out.

[maximilianovermeyer]

This is just a bug in lutris:
lutris/lutris#4290

[DaVikingMan]

For anyone who had experienced this bug, this has been fixed in the latest version(if using arch, you may want to build it from source or install it from AUR as the patch wasn't upstreamed by the devs).

[TheRealOne78]

I don't think lutris devs updating it's program would really solve the real bug.
It seems like some programs don't recognise the NVIDIA GPU driver.
Making all the devs in the world update their code to be able to see the new nvidia open source driver is simply not a good solution, maybe some there would be some abandoned programs as well.
In my opinion, instead of making all people change their code, a better solution would be to fix the open source driver to show itself just like the proprietary driver.

For example, wazzapp also doesn't work, giving me the following error.

[28941:0607/172949.236856:FATAL:gpu_data_manager_impl_private.cc(445)] GPU process isn't usable. Goodbye.
/bin/whatsapp: line 2: 28941 Trace/breakpoint trap   /opt/wazzapp/wazzapp

As a side note, wazzapp was also abandoned a week ago.

[maximilianovermeyer]

Look at the error message and the code in question. It's just a string parsing error because lutris did not recognize the version part. They expected it to come directly after the word "Module".

If you move "for x86_64" to the end, this will break the next project assuming the version number is the last part of the string. Then, there is a project assuming it's always the fifth word... There will always be problems, one way or the other.

NVIDIA UNIX Open Kernel Module for x86_64 515.43.04

[TheRealOne78]

So basically the proprietary version would be NVIDIA UNIX Kernel Module for x86_64 515.43.04 and the open source version would be NVIDIA UNIX Open Kernel Module for x86_64 515.43.04?

If yes, what would happen if instead of NVIDIA UNIX Open Kernel Module for x86_64 515.43.04 it would be NVIDIA UNIX Kernel Module for x86_64 515.43.04-open or add another number character to the version number that would differentiate the propritary kernel from the open kernel? For example add a 0 at the begining 0515.43.04 or something like this? Or add at the end of the version a .0 in case some programs requires a version or later: 515.43.04.0

Honestly, changing completly the version scheme sounds like a bad idea, but in the same time the version scheme was already changed by nvidia and it is a real issue, that in my opinion, it could be partially solved by thinking how generally developers write their code.

[maximilianovermeyer]

https://github.com/lutris/lutris/blob/master/lutris/util/graphics/drivers.py

This would crash the new solution again and likely some other projects too. There now is a good solution by @aaronp24 in the lutris bug, this seems to be clean and not that error-prone.

[aritger]

Sorry that the version format here tripped the lutris version parser. It sounds like this is resolved in top of tree of lutris. I'm going to go ahead and close this Issue. Please reopen it if the fix in lutris is not sufficient.