| 21 May 2025 |
 matthewcroughan | > nvidia.o: warning: objtool: _nv045909rm+0x77: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv045853rm+0x7f: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035412rm+0x3e: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035411rm+0x27: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035415rm+0x37: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035414rm+0x24: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035413rm+0xca: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035416rm+0x193: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv035418rm+0x17f: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv013509rm+0x5d: 'naked' return found in MITIGATION_RETHUNK build
> nvidia.o: warning: objtool: _nv046215rm+0x12f: 'naked' return found in MITIGATION_RETHUNK build
> MODPOST Module.symvers
> WARNING: modpost: missing MODULE_DESCRIPTION() in nvidia.o
> WARNING: modpost: missing MODULE_DESCRIPTION() in nvidia-uvm.o
> WARNING: modpost: missing MODULE_DESCRIPTION() in nvidia-modeset.o
> WARNING: modpost: missing MODULE_DESCRIPTION() in nvidia-drm.o
> ERROR: modpost: GPL-incompatible module nvidia.ko uses GPL-only symbol '__vma_start_write'
> ERROR: modpost: GPL-incompatible module nvidia-drm.ko uses GPL-only symbol '__vma_start_write'
> make[4]: *** [/nix/store/f0rfxq2w8m53gfwpspr3irafqv06xjf6-linux-6.15-rc7-dev/lib/modules/6.15.0-rc7/source/scripts/Makefile.modpost:147: Module.symvers] Error 1
> make[3]: *** [/nix/store/f0rfxq2w8m53gfwpspr3irafqv06xjf6-linux-6.15-rc7-dev/lib/modules/6.15.0-rc7/source/Makefile:1953: modpost] Error 2
> make[2]: *** [/nix/store/f0rfxq2w8m53gfwpspr3irafqv06xjf6-linux-6.15-rc7-dev/lib/modules/6.15.0-rc7/source/Makefile:248: __sub-make] Error 2
> make[2]: Leaving directory '/build/NVIDIA-Linux-x86_64-570.153.02/kernel'
> make[1]: *** [Makefile:248: __sub-make] Error 2
> make[1]: Leaving directory '/nix/store/f0rfxq2w8m53gfwpspr3irafqv06xjf6-linux-6.15-rc7-dev/lib/modules/6.15.0-rc7/source'
| 17:27:04 |
| matthewcroughan | just so buggy | 17:27:07 |
 K900 | Yes | 17:34:31 |
| K900 | There's no stable kernel API | 17:35:05 |
| K900 | So any out of tree driver can and will break | 17:35:10 |
| matthewcroughan | And there's no way to opt out of the open nvidia driver really anymore | 17:40:00 |
| matthewcroughan | so that's pretty bad | 17:40:03 |
| matthewcroughan | it's like zfs but non-optional | 17:40:35 |
| K900 | There is, you can just run Nouveau | 17:42:10 |
| matthewcroughan | rofl | 17:42:19 |
 olivia | nvk is genuinely pretty good these days if you have a turing+card. If you haven't tried it recently you might be surprised | 19:54:32 |
| olivia | if this thing you're trying to run is cuda only that won't work though | 19:54:39 |
| olivia | * if this thing you're trying to run is cuda-only that won't work though | 19:54:45 |
 ElvishJerricco | I'm guessing nvk is vulkan only, and that that's ok as long as you only play windows / proton games because they're almost universally adapted to vulkan? | 20:13:49 |
| olivia | it's vulkan but mesa 25.1 defaults to zink/nvk for opengl on Turing cards now | 20:32:07 |
 srestegosaurio (on tchncs.de) | In reply to @olivia:computer.surgery
nvk is genuinely pretty good these days if you have a turing+card. If you haven't tried it recently you might be surprised Are there any docs on how to set it up? I feel very lost when touching NVIDIA related settings. | 23:18:06 |
| srestegosaurio (on tchncs.de) | In reply to @olivia:computer.surgery
nvk is genuinely pretty good these days if you have a turing+card. If you haven't tried it recently you might be surprised * | 23:18:14 |
| ElvishJerricco | it sounds like if it's anything like amdgpu you just remove any nvidia related code from your config and mesa will do the right thing on its own (as long as you're on a nixpkgs revision with a new enough mesa) | 23:20:06 |
| ElvishJerricco | but I'm only guessing | 23:20:53 |
| srestegosaurio (on tchncs.de) | In reply to @elvishjerricco:matrix.org
it sounds like if it's anything like amdgpu you just remove any nvidia related code from your config and mesa will do the right thing on its own (as long as you're on a nixpkgs revision with a new enough mesa) I think it used to be like that so seams reasonable. | 23:21:32 |
| srestegosaurio (on tchncs.de) | And what about nvidia-offload? | 23:22:01 |
| olivia | yeah, this is accurate | 23:22:42 |
| olivia | specifically you'll get zink-by-default if you have https://github.com/NixOS/nixpkgs/pull/399388 (which was merged 12 hours ago) | 23:23:14 |
| olivia | yeah, you'll want to remove that | 23:23:57 |
| ElvishJerricco | I'm guessing it's the same story with nvidia-offload? i.e. Mesa just already knows how to use another GPU for rendering and lets the app choose? | 23:24:21 |
| srestegosaurio (on tchncs.de) | In reply to @olivia:computer.surgery
yeah, you'll want to remove that Yeah, waht I meant is how should I go about running things on the dGPU. | 23:24:41 |
| srestegosaurio (on tchncs.de) | Isn't there any magic ENVVAR/command? | 23:25:14 |
| ElvishJerricco | In reply to @srestegosaurio:tchncs.de
Yeah, waht I meant is how should I go about running things on the dGPU. If it works like amdgpu then yea games will just automatically use it by default because they'll ask the driver for the biggest GPU. And if you want to force one you set DRM_PRIME or whatever | 23:25:38 |
| olivia | DRI_PRIME is what people usually use | 23:25:48 |
| olivia | https://docs.mesa3d.org/envvars.html#envvar-DRI_PRIME | 23:25:59 |
