Skip to content

Conversation

lukaszsobala
Copy link
Contributor

@lukaszsobala lukaszsobala commented Aug 24, 2025

The amdgpu module never compiled with the Rockchip Linux kernel 6.1. These changes are enough to make it compile and are related to DRM scheduling. Testing with 1 AMD card (on Radxa Rock 2F) lead to no display output so far, but the module is loading properly and I am planning to do more tests.

@hbiyik
Copy link

hbiyik commented Sep 8, 2025

most likely amdgpu was broken when i was backporting panthor initially. I tried to maintain the patches intact as much as i could but it is possible that i might have missed some stuff in amdgpu, nvidia, intel and adreno drivers.

Fixes should be trivial though.

@lukaszsobala
Copy link
Contributor Author

lukaszsobala commented Sep 8, 2025

Fixes should be trivial though.

So what about these particular fixes presented here? I'm a bit confused with your comment.

The kernel builds using actions, but I don't think the module was enabled for the build. Locally it does build with the module

@hbiyik
Copy link

hbiyik commented Sep 8, 2025

I'm a bit confused with your comment.

I am just trying to say, "sorry i broke something which was already working :)" in the bsp kernel.

The issue originates from hbiyik/linux@212bbed this commit which is the one i backported when i was backporting the panthor driver.

in armbian tree all panthor changes are squashed into single commit and the change is here c4f16b4

The fundemantal change in hbiyik/linux@212bbed is that drm_sched_job_init now expects additional "credits" argument. i have backported the change to almost every driver and also amdgpu but i have forgotten the file "amd_gpu_cs.c". And also nouveau driver is also missing, so this might be another issue that i might have created.

what about these particular fixes presented here

So such interface changes might be missing in the drm subsystem time to time, and i think this patch seems clean the mess i have caused. It is of course the best to run the driver with hardware with the driver, i can not judge if it is ok or not just by looking at it, drm subsystem is crazy at times.

@lukaszsobala
Copy link
Contributor Author

I understand. You are the expert here, I was just "vibe coding" to get the module to compile. I have little programming knowledge. So hopefully someone can check this. The changes are indeed minor.

@lukaszsobala
Copy link
Contributor Author

lukaszsobala commented Sep 14, 2025

Hopefully these changes are applied now.
I'll try compiling with nouveau too to see if it's broken as well, as @hbiyik suggested. The cards won't work well yet, but maybe some day :)
EDIT: nouveau builds.

@lukaszsobala
Copy link
Contributor Author

lukaszsobala commented Sep 15, 2025

oh no, have I done something stupid? I was looking for the changes requested by @amazingfate after pulling and nothing came up in vscode. Previously, they appeared. I clicked this button to merge branch by mistake. Please tell me what to do if I broke something – and why can't I see the requested changes this time.
Please don't hate me :_)

@amazingfate
Copy link
Collaborator

oh no, have I done something stupid? I was looking for the changes requested by @amazingfate after pulling and nothing came up in vscode. Previously, they appeared. I clicked this button to merge branch by mistake. Please tell me what to do if I broke something – and why can't I see the requested changes this time. Please don't hate me :_)

This is fine, don't worry.

@amazingfate amazingfate merged commit f07e924 into armbian:rk-6.1-rkr5.1 Sep 17, 2025
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants