Arch Linux User Repository

For testing users, contributors: check out https://github.com/archlinux-jerry/nvidia-340xx
Warning: force push may happen to match the aur tree

Edit:
The driver works and the steps bellow aren't requared. I've trusted the people saying there is an issue without confirming it myself.

Original:
Here is a workaround I've used (it's based on lmh69's comment).

I've extracted 'drm_legacy.h' from 'linux-headers-6.7.arch3-1-x86_64.pkg.tar.zst' to '/usr/lib/modules/6.8.4-arch1-1/build/include/drm/drm_legacy.h'. Then I've reinstalled both 'linux-headers' (to run dkms) [and 'linux' to run mkinitcpio] you can run them manually but this way was easier for me.

I've tested just a few things: vdpau, glx, Xorg.0.log and dmesg. I haven't found any issues so far.

IMPORTANT: I am still with 'nvidia-340xx-dkms 340.108-35'. I've never updated. It won't work with 'nvidia-340xx-dkms 340.108-36'.

I don't know if it's wrong but works for me:

$ uname -r
6.8.4-arch1-1

$ glxgears
Running synchronized to the vertical refresh.  The framerate should be
approximately the same as the monitor refresh rate.
303 frames in 5.0 seconds = 60.491 FPS
300 frames in 5.0 seconds = 59.972 FPS

$ mpv ...
 (+) Video --vid=1 (*) (h264 1280x720 25.000fps)
 (+) Audio --aid=1 (*) (aac 2ch 44100Hz)
Using hardware decoding (vdpau).
AO: [pulse] 44100Hz stereo 2ch float
VO: [gpu] 1280x720 vdpau[yuv420p]
Exiting... (Quit)

Although I have installed nvidia-340xx-dkms 340.108-36 under 6.8.1-arch1-1 kernel, every time I'm trying to launch any "opengl" application raises this:

$ kitty
X Error of failed request:  BadDrawable (invalid Pixmap or Window parameter)
  Major opcode of failed request:  155 (NV-GLX)
  Minor opcode of failed request:  4 ()
  Resource id in failed request:  0x1000009
  Serial number of failed request:  241
  Current serial number in output stream:  241

@lmh69 @ JerryXiao Who maintains the patches? I mean who writes them? I guess someone has to help him with 'drm' thing.

I just checked the code changes in the kernel code and did a fast test, restoring the removed drm_legacy.h file from the lastest 6.7 kernel (I paste it in a new folder and changed the code to reference to it:

include <path_to_the_file/drm_legacy.h>

Seems to be working ok. I got a little bit worst fps in a first test I'm doing but could be something random... need to check more. Xorg.log, dmesg, nvidia logs seems the same as with 6.7 series and the previous ones.

@lmh69 Thanks. Well, that's sad news. Anyway, I guess I'm not using any h/w acceleration. At least I disable it everywhere like the browser etc.

@TheExplorer "What is 'drm functionality' and how does it affect the driver's performance?"
No 3d hardware acceleration here. Vdpau and 2D acceleration seems to be working...

[    72.370] (II) "glx" will be loaded. This was enabled by default and also specifie
d in the config file.
[    72.370] (II) LoadModule: "glx"
[    72.374] (II) Loading /usr/lib/xorg/modules/extensions/libglx.so
[    72.503] (II) Module glx: vendor="X.Org Foundation"
[    72.503]    compiled for 1.21.1.11, module version = 1.0.0
[    72.503]    ABI class: X.Org Server Extension, version 10.0
...
[    72.503] (II) LoadModule: "nvidia"
[    72.506] (II) Loading /usr/lib/xorg/modules/drivers/nvidia_drv.so
[    72.517] (II) Module nvidia: vendor="NVIDIA Corporation"
[    72.517]    compiled for 4.0.2, module version = 1.0.0
[    72.517]    Module class: X.Org Video Driver
...
[    72.527] (**) NVIDIA(0): Enabling 2D acceleration
[    72.527] (**) NVIDIA(0): ConnectedMonitor string: "DFP-1"
[    72.527] (EE) NVIDIA(0): Failed to initialize the GLX module; please check in your X
[    72.527] (EE) NVIDIA(0):     log file that the GLX module has been loaded in your X
[    72.527] (EE) NVIDIA(0):     server, and that the module is the NVIDIA GLX module.  If
[    72.527] (EE) NVIDIA(0):     you continue to encounter problems, Please try
[    72.527] (EE) NVIDIA(0):     reinstalling the NVIDIA driver.
[    73.576] (II) NVIDIA(0): Using ConnectedMonitor string "DFP-1".
...
[    75.119] (II) Loading sub module "dri2"
[    75.119] (II) LoadModule: "dri2"
[    75.119] (II) Module "dri2" already built-in
[    75.119] (II) NVIDIA(0): [DRI2] Setup complete
[    75.119] (II) NVIDIA(0): [DRI2]   VDPAU driver: nvidia
...
[    75.123] (II) AIGLX: Screen 0 is not DRI2 capable
[    75.357] (II) IGLX: Loaded and initialized swrast
[    75.357] (II) GLX: Initialized DRISWRAST GL provider for screen 0

$ glxinfo 
name of display: :0
X Error of failed request:  BadWindow (invalid Window parameter)
  Major opcode of failed request:  153 (NV-GLX)
  Minor opcode of failed request:  4 ()
  Resource id in failed request:  0x220000a
  Serial number of failed request:  59
  Current serial number in output stream:  59

@JerryXiao Thank you for the info. Cheers!

On second thought, spending too much effort maintaining this package is not the best use of time. Go fix nouveau reclocking for NV50 cards if you really had that time. This driver will eventually bit rot into nothing sooner or later.

I would expect no performance loss. In terms of drm the 340xx implementation has little to no real use as far as I can tell. You don't get to run anything wayland/kms whatsoever. If all you need is X then the nvidia ddx & opengl implementation is right inside the package for you.

@TheExplorer The answer to your first question is not hard to find: DRM in this context is not Digital “Rights” (or, in the words of the FSF, Restrictions) Management but a component called Direct Rendering Manager. See https://bbs.archlinux.org/viewtopic.php?id=275936 and https://man.archlinux.org/man/drm.7.en – Regarding your second question, I don’t know.

I have now changed my pacman.conf to prevent my Linux packages from updating. Would love to help in some way but I don’t think I have the required resources or knowledge. I have some experience in C development but that is basically from the 1990s and I know nothing about developing graphics drivers or working with Linux kernel code. I assume this would involve much more than adapting to a few changes in function signatures.