You are not logged in.
1st link below.
Online
I did something hope this okay
https://gitea.awptechnologies.co/awptec … s/Arch-Dev
Last edited by awptechnologies (2024-09-13 02:49:56)
Offline
I use gitea for private version control but i opened this repository to document this journey.
Offline
just click explore and it's the only repository there.
Offline
/tmp/selfgz847/NVIDIA-Linux-x86_64-535.161.07-grid/kernel/nvidia-drm/nvidia-drm-gem.c:115:16: error: initialization of 'int (*)(struct drm_gem_object *, struct iosys_map *)' from incompatible pointer type 'void * (*)(struct drm_gem_object *)' [-Wincompatible-pointer-types]
Online
With your export
export EXTRA_CFLAGS="-Wimplicit-function-declaration -Wincompatible-pointer-types"
I added the second log to same link just .log-2
https://gitea.awptechnologies.co/awptec … s/Arch-Dev
Last edited by awptechnologies (2024-09-13 02:49:41)
Offline
It still failed. Gonna try loqs next.
Offline
loqs failed as well
export EXTRA_CFLAGS="-Wno-error=implicit-function-declaration -Wno-error=incompatible-pointer-types"
logs in link .log-3
https://gitea.awptechnologies.co/awptec … s/Arch-Dev
Last edited by awptechnologies (2024-09-13 02:49:09)
Offline
Just incase you need it i also uploaded the specific driver i am trying to install here as well
https://gitea.awptechnologies.co/awptec … s/Arch-Dev
Last edited by awptechnologies (2024-09-13 02:48:48)
Offline
They all still have the gcc errors, ie EXTRA_CFLAGS are not applied by the installer.
Install gcc13 "export CC=/usr/bin/cc-13" and hope for the best
The normal nvidia installers have a feature to rebuild themselves w/ patches, check the "--help" of your installer.
You would then extract it, isolate the CFLAGS in the makefile and adjust them accordingly, generate a diff from that and feed that to the patch self-rebuild feature (along maybe the 6.10 patches)
Online
These didn't work ethier.
I extracted the driver and put it in the git. Will you tell me what i need to edit. Also there is a readme that shows options of the driver.
https://gitea.awptechnologies.co/awptec … s/Arch-Dev
Link changed and i changed it in previous posts as well.
Offline
If there is anything wrong with the git repo let me know. Uncompressed the driver took up alot.
Offline
Any Luck?
Offline
Just a heads up, I spent the better part of last week on a phone and limited throughput and had no chance to look at this at all so far.
Online
Okay thank you for getting back to me. I totally understand. I will keep that repository up indefinitely in case other users have this issue. And look forward to your wisdom lol.
Offline
Do you think its worth me trying again with the gpu patch we did before. I save backups after i got that working. Maybe the new gnome with better wayland support will make a differenece?
Offline
https://gitea.awptechnologies.co/awptec … kefile#L51
Install gcc13 "export CC=/usr/bin/cc-13" and hope for the best
These didn't work ethier.
How exactly did that not work?
You can probably also just run edit kernel*/{Makefile,Kbuild} to change the compiler version or add the | EXTRA_CFLAGS += "-Wimplicit-function-declaration -Wincompatible-pointer-types" | there and the run ./nvidia-installer from there.
nb. that unpatched, that driver version will only build against the LTS kernel - otherwise you'll have to manually apply the 6.10 patch from the AUR package and the 0003-Fix-conftest-to-use-nv_drm_gem_vmap-which-has-the-se.patch from https://0x0.st/XJ0g.diff
patch -Np1 -i "linux-6.10.patch"
patch -Np1 -i "0003-Fix-conftest-to-use-nv_drm_gem_vmap-which-has-the-se.patch" # provided you copied them into the extracted installer path
before running ./nvidia-installer
Alternatively go back to https://bbs.archlinux.org/viewtopic.php … 1#p2191131 and just copy the nvidia-gridd daemon and script and https://gitea.awptechnologies.co/awptec … dd.service into the system manually.
Online
If it was your machine which route would you go? I would Prefer to stay on latest and greatest kernel so i guess that gets rid of using unpatched driver. I have a snapshot of the previous build with gpu working so i guess i should try adding the nvidia-gridd service first. But i like the idea of a fresh install with the patches and editing my makefile and kbuild. Im still struggling to understand exactly what todo if i take that route. What do you mean run edit kernel? is that a program that exists in arch? and where do i add the extra_cflags? When running the patch commands do i have to install the .diff file and reference it? still a little lost. Sorry i have been researching non stop and still figuring all of this out.
Offline
So i copied patch files in to root of extracted drivers and patched it. Says successful
Since i didn't know exactly where to put the extra cflags i just ran
export EXTRA_CFLAGS="-Wno-error=implicit-function-declaration -Wno-error=incompatible-pointer-types"
and since i didn't know how to change compiler version i ran
export CC=/usr/bin/cc-13
Seems to have worked now.
Offline
Im gonna let VM sit overnight and make sure the lag is fixed. After that im going to revert to snapshot before old driver patch and do a fresh install since i have messed around with this install so much.
When i do that which files and where do i change compiler version and add cflags so i dont need to run export.
Or do you suggest i do it like i listed above?
Offline
What do you mean run edit kernel? i… Since i didn't know exactly where to put the extra cflags… Seems to have worked now.
You can probably also just run edit kernel*/{Makefile,Kbuild}
there're two directories, kernel and kernel-open and in them you'll find "Makefile" and "Kbuild" files where you would have changed CC and extended the EXTRA_CFLAGS, though I con't understand why the exports would not have worked before.
Either way, using gcc13 should™ actually not be necessary w/ the EXTRA_CFLAGS in place.
You should™ be able to run ./makeself.sh to turn the patched and edited sources into a new installer tarball script (but that's technically not necessary, you can keep using the extracted driver, archive it as a regular tarball or whatever you like.
Online
So they didnt work before because i didn't implement the patch. They only worked after i installed patch.
At first i tried only the patch and that did not work.
After running the exports it worked.
And from what your saying with the cflags in place there is no need to install gcc13 before install?
Also do i need to edit both the "Makefile" and "Kbuild" in both kernel and kernel-open?
Would be awesome if you registered as a user on my git server and pushed a commit to those files. Also with that being said your more then welcome to use it freely.
Offline
The files aren't very long, you'll likely just have to edit the Kbuild for the EXTRA_CFLAGS and there's a bunch of lines setting that variable.
You made it this far, you'll make it over the finish line.
Online
can i just do the export method for now until i figure that out
Offline
So just an update and a few concerns. I got it to work. Even dkms was able to work moving to the new 6.11 kernel. Only concern i have is how i had to do it. I had to go into pacman.conf and ignore the nvidia utils pkg from updating. First of all i find it wierd that this driver of mine refuses to work without the nvidia-utils package. What does it contain? Anyways after removing nvidia-dkms and ignoring the nvidia-utils package i was able to upgrade and driver is still fully functional.
Without nvidia-utils the driver installs but the vm refuses to restart or show nvidia-smi.
With nvidia-utils and allowing it to update i get a library conflict and again nvidia-smi doesn't work.
When ignoring package system updates fine as well as retaining nvidia-smi and full gpu function.
Any Ideas?
Offline