search

Frogging-Family / nvidia-all

Nvidia driver latest to 396 series AIO installer
790 stars 69 forks source link

Nvidia 530.41.03 won't build with kernel 6.4 #177

Closed ryanmusante closed 1 year ago

ryanmusante commented 1 year ago

Related to post below, nv 530.41.03 will not build with kernel 6.4. Others having same problem?

https://forums.developer.nvidia.com/t/error-compiling-with-recent-mainline-kernel-linux-6-4-rc1/252732

thojohns commented 1 year ago

The linked thread already contains a solution. Granted, it's a solution for 470.xx, but the code that needs changing is the same - it's only the line numbers that differ, although you'd also want to modify files in /kernel-dkms instead of /kernel.

I threw together a quick patch for 530.41.03 based on the 470 patch, which you'll need to put in your patches folder. I also updated the PKGBUILD to use it.

I'm refraining from submitting a pull request with this fix because A) it's not a general solution - every driver version pre-535 will probably need a patch like this unless Nvidia puts out minor releases with a similar fix, and B) I've been having a hell of a time trying to update the PKGBUILD with a section for 6.4 patches - I initially added a tweaked 470 patch as well but no matter what conditionals I set, the script kept trying to apply both the 470 and 530 patches, resulting in a build failure.

EDIT: updated the PKGBUILD because I left a trailing newline off (or GH clipped it) when uploading the patch, causing the MD5 checksum to fail.

ryanmusante commented 1 year ago

Ahh, nice. When I get the chance I will test it

Drogobo commented 1 year ago

this did fix the compile errors.

ryanmusante commented 1 year ago

This needs to be taken in upstream, should be posted to Nvidia developer forums?

thojohns commented 1 year ago

I believe Nvidia has already implemented their own similar fix - obviously, 535 compiles without issue, and if you look at the open kernel modules you'll even see that their fix is slightly more elegant despite accomplishing the exact same result, but there's also a workaround included in the 470.199 driver, so it does seem like it's being backported. I don't expect any future releases for 470, 530, 535, or otherwise will suffer from this issue in particular.