17 Fresh AMDGPU DC Patches Posted Today

[Grim85]

Would you really trust a fish? https://github.com/M-Bab

Open source projects are one thing, but kernel binary from a totally random Github account?

Yeah, don't use it myself. But it is an option floating out there for ubuntu users

I'd recommend the Fedora one over it any day

[PuckPoltergeist]

Last time I checked, there was no need to merge any changes. IIRC, anything that doesn't apply has been already applied; I noticed a few patch hunks fail, but I believe they can be ignored.

If it helps people out, I can mirror my kernel code on github.

I'm speaking about merging the amd-staging-4.9 branch into linux-stable/linux-4.9.y. This results in merge conflicts like this:
drivers/gpu/drm/amd/amdgpu/si_dpm.c

       case CHIP_HAINAN:
               if (((adev->pdev->revision == 0x81) &&
                       (adev->pdev->device == 0x6660)) ||
                   ((adev->pdev->revision == 0x83) &&
                       ((adev->pdev->device == 0x6660) ||
                       (adev->pdev->device == 0x6663) ||
                       (adev->pdev->device == 0x6665) ||
<<<<<<< HEAD
                       (adev->pdev->device == 0x6667))) ||
                   ((adev->pdev->revision == 0xc3) &&
                       (adev->pdev->device == 0x6665)))
=======
                        (adev->pdev->device == 0x6667))))
>>>>>>> 888eff335db092f83c807e239189e3df969d1caf
                       chip_name = "hainan_k";

You can't ignore this, caust this wouldn't compile. And for fixing those semantic changes, you need knowledge about the code.

[Mystro256]

I'm speaking about merging the amd-staging-4.9 branch into linux-stable/linux-4.9.y. This results in merge conflicts like this:
drivers/gpu/drm/amd/amdgpu/si_dpm.c

case CHIP_HAINAN:
if (((adev->pdev->revision == 0x81) &&
(adev->pdev->device == 0x6660)) ||
((adev->pdev->revision == 0x83) &&
((adev->pdev->device == 0x6660) ||
(adev->pdev->device == 0x6663) ||
(adev->pdev->device == 0x6665) ||
<<<<<<< HEAD
(adev->pdev->device == 0x6667))) ||
((adev->pdev->revision == 0xc3) &&
(adev->pdev->device == 0x6665)))
=======
(adev->pdev->device == 0x6667))))
>>>>>>> 888eff335db092f83c807e239189e3df969d1caf
chip_name = "hainan_k";

You can't ignore this, caust this wouldn't compile. And for fixing those semantic changes, you need knowledge about the code.

Fair enough, I'll consider mirroring my sources somewhere to help the non-fedora users out. I need to reorganize a few things, so I'll look into it this weekend.

EDIT: I wasn't reading the posts above, looks like someone beat me to it:
https://github.com/M-Bab/linux-kernel-amdgpu

[Mystro256]

Yeah, don't use it myself. But it is an option floating out there for ubuntu users

I'd recommend the Fedora one over it any day

LEW21 Well looks like the source is there, so people can build it if they want:

https://github.com/M-Bab/linux-kernel-amdgpu

[automorphism]

I'm speaking about merging the amd-staging-4.9 branch into linux-stable/linux-4.9.y. This results in merge conflicts like this:
drivers/gpu/drm/amd/amdgpu/si_dpm.c

case CHIP_HAINAN:
if (((adev->pdev->revision == 0x81) &&
(adev->pdev->device == 0x6660)) ||
((adev->pdev->revision == 0x83) &&
((adev->pdev->device == 0x6660) ||
(adev->pdev->device == 0x6663) ||
(adev->pdev->device == 0x6665) ||
<<<<<<< HEAD
(adev->pdev->device == 0x6667))) ||
((adev->pdev->revision == 0xc3) &&
(adev->pdev->device == 0x6665)))
=======
(adev->pdev->device == 0x6667))))
>>>>>>> 888eff335db092f83c807e239189e3df969d1caf
chip_name = "hainan_k";

You can't ignore this, caust this wouldn't compile. And for fixing those semantic changes, you need knowledge about the code.

I'm assuming this problem must arise from the order in which git tries to merge the patches. Because it works fine if you first download the amd-staging-4.9 branch, and then apply the 4.9.18.patch file from kernel.org with the patch command, rather than trying to merge them with git. Everything in the kernel.org 4.9.0->4.9.18 patch file either applies without complaint or has already been applied in amd-staging-4.9, meaning that you can safely ignore those hunks. I was also able to apply all of the other patches in gentoo-sources without any issues since they deal with unrelated parts of the kernel. I'm currently using a kernel compiled in this manner as I type this so I can confirm that it also compiles without any issues.

[PuckPoltergeist]

I'm assuming this problem must arise from the order in which git tries to merge the patches.

possibly, I haven't tested it

Because it works fine if you first download the amd-staging-4.9 branch, and then apply the 4.9.18.patch file from kernel.org with the patch command, rather than trying to merge them with git. Everything in the kernel.org 4.9.0->4.9.18 patch file either applies without complaint or has already been applied in amd-staging-4.9, meaning that you can safely ignore those hunks.

No, you can't. It would not compile. So as long as you don't have any conflicts, it's ok. If you get some, you have to solve them.