linux-firmware: amdgpu regression causing freezes, fix available upstream

[aviallon]

Nixpkgs version

• Stable (25.11)

Describe the bug

Exactly what is described here: https://gitlab.freedesktop.org/drm/amd/-/issues/4737
There is a regression in amdgpu's firmware that causes SMU crashes, which manifests as bad stuttering on the screen.

Steps to reproduce

• Use a relatively recent AMD GPU (RX 7900 XT for instance)
• Use linux-firmware 20251125
• Stress the CPU (using stress-ng for instance)

(more details here: https://gitlab.freedesktop.org/drm/amd/-/issues/4737#note_3215978)

Expected behaviour

My screen does not completely freeze under load.

Screenshots

No response

Relevant log output

déc. 01 21:50:19 luke-skywalker-nixos kernel: [drm] PCIE GART of 512M enabled (table at 0x00000084FEB00000).
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: PSP is resuming...
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: reserve 0x1300000 from 0x84fc000000 for PSP TMR
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: RAP: optional rap ta ucode is not available
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: SECUREDISPLAY: optional securedisplay ta ucode is not available
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: SMU is resuming...
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: smu driver if version = 0x0000003d, smu fw if version = 0x00000040, smu fw program = 0, smu fw version = 0x004e8200 (78.130.0)
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: SMU driver if version not matched
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: SMU is resumed successfully!
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: [drm] DMUB hardware initialized: version=0x07002F00
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: [drm] Cannot find any crtc or sizes
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring gfx_0.0.0 uses VM inv eng 0 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.0.0 uses VM inv eng 1 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.1.0 uses VM inv eng 4 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.2.0 uses VM inv eng 6 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.3.0 uses VM inv eng 7 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.0.1 uses VM inv eng 8 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.1.1 uses VM inv eng 9 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.2.1 uses VM inv eng 10 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring comp_1.3.1 uses VM inv eng 11 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring sdma0 uses VM inv eng 12 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring sdma1 uses VM inv eng 13 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring vcn_unified_0 uses VM inv eng 0 on hub 8
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring vcn_unified_1 uses VM inv eng 1 on hub 8
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring jpeg_dec uses VM inv eng 4 on hub 8
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: amdgpu: ring mes_kiq_3.1.0 uses VM inv eng 14 on hub 0
déc. 01 21:50:19 luke-skywalker-nixos kernel: amdgpu 0000:10:00.0: [drm] Cannot find any crtc or sizes

Additional context

No response

System metadata

• system: "x86_64-linux"
• host os: Linux 6.17.9-cachyos, NixOS, 25.11 (Xantusia), 25.11pre-git
• multi-user?: yes
• sandbox: yes
• version: nix-env (Nix) 2.32.4
• nixpkgs: /nix/store/ycr2g3rk022wld5cnpxjxjlkqfpnpa6l-source

Notify maintainers

@fpletz @K900

---

Note for maintainers: Please tag this issue in your pull request description. (i.e. Resolves #ISSUE.)

I assert that this issue is relevant for Nixpkgs

• I assert that this is a bug and not a support request.
• I assert that this is not a duplicate of an existing issue.
• I assert that I have read the NixOS Code of Conduct and agree to abide by it.

Is this issue important to you?

Add a 👍 reaction to issues you find important.

[Michael-D-Hart]

I was just about to open an issue on this...

This commit updates the firmware version: 9c7701d

Looking at the kernel/git/firmware/linux-firmware.git I see only amdgpu updates moving from 20251111 to 20251125 .

Pinning your flake nikpkgs to the state at the previous commit keeps the system stable. And by stable I mean being able to run for more than 10 minutes without crashing in my case. It is unusable.

flake.nix

{
   inputs = {
      # nixpkgs.url = "github:NixOS/nixpkgs/nixos-25.05"; # deprecated
      nixpkgs.url = "github:NixOS/nixpkgs/df3793a20f2e3f62359faf5eb084d3bb88347682"; # works fine, commit before firmware update
      # nixpkgs.url = "github:NixOS/nixpkgs/9c7701dcc7e55cce484ed9b86993314e393866ed"; # breaks things,  display driver crashes, system crashes, reboots, display goes black
      # nixpkgs.url = "github:NixOS/nixpkgs/nixos-25.11"; # latest version, usage conditioned by firmware override
      ...
   }
   ...
}

That also means you will stay on 25.05.

To update to the latest 25.11 and not pin my flake to that commit I have an override:

configuration.nix

{ pkgs, ... }:{
 ....
  hardware.firmware = [
    (pkgs.linux-firmware.overrideAttrs (old: {
      version = "20251111";
      src = pkgs.fetchurl {
        # https://www.kernel.org/pub/linux/kernel/firmware/
        url = "https://www.kernel.org/pub/linux/kernel/firmware/linux-firmware-20251111.tar.gz";
        # > nix-prefetch-url https://www.kernel.org/pub/linux/kernel/firmware/linux-firmware-20251111.tar.gz
        sha256 = "0rp2ah8drcnl7fh9vbawa8p8c9lhvn1d8zkl48ckj20vba0maz2g";
      };
    }))
  ];
...
}

However, some things seem to not work. For example reboot or dd commands seem to break. I'm not sure this is a issue because of the override or another unreleated nixos package issue. Nevermind that, it was toybox messing with my commands.

[chrispouliot]

I wonder if this is the same issue I was reporting in the drm repo here
I could only see "page_flip timeout" in the logs but it sounds identical to my experience. Complete freezing of the entire system until I reboot, both on a 9070 XT and an 890M iGPU.

Do I need to update anything specifically on NixOS for when a new amdgpu firmware comes out, or is that handled normally? I am on stable 25.11 currently.

My logs only seem to have a page_flip timeout so perhaps its a separate issue with the screen / DE freezing.

[nixos-discourse]

This issue has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/amd-apu-graphics-driver-resets-periodically-on-wayland-after-upgrading-to-25-11/72982/2

[nixos-discourse]

This issue has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/graphics-breaks-after-rebuild-on-nixos-25-11/72975/16

[json1c]

I’m also experiencing this regression after rebuilding my system to nixpkgs-25.11

дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: Dumping IP State Completed
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: [drm] AMDGPU device coredump file has been created
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: [drm] Check your /sys/class/drm/card2/device/devcoredump/data
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring gfx_0.1.0 timeout, signaled seq=14894, emitted seq=14896
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu:  Process .kwin_wayland-w pid 2517 thread kwin_wayla:cs0 pid 2555
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: Starting gfx_0.1.0 ring reset
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: Ring gfx_0.1.0 reset failed
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: amdgpu: GPU reset begin!
дек 06 13:18:39 legion kernel: amdgpu 0000:06:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB error - collecting diagnostic data
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: MODE2 reset
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: GPU reset succeeded, trying to resume
дек 06 13:18:40 legion kernel: [drm] PCIE GART of 1024M enabled (table at 0x000000F41FC00000).
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: PSP is resuming...
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: reserve 0xa00000 from 0xf41e000000 for PSP TMR
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: RAS: optional ras ta ucode is not available
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: RAP: optional rap ta ucode is not available
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: SECUREDISPLAY: optional securedisplay ta ucode is not available
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: SMU is resuming...
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: SMU is resumed successfully!
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: kiq ring mec 2 pipe 1 q 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: [drm] DMUB hardware initialized: version=0x05002C00
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring gfx_0.0.0 uses VM inv eng 0 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring gfx_0.1.0 uses VM inv eng 1 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.0.0 uses VM inv eng 4 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.1.0 uses VM inv eng 5 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.2.0 uses VM inv eng 6 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.3.0 uses VM inv eng 7 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.0.1 uses VM inv eng 8 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.1.1 uses VM inv eng 9 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.2.1 uses VM inv eng 10 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring comp_1.3.1 uses VM inv eng 11 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring kiq_0.2.1.0 uses VM inv eng 12 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring sdma0 uses VM inv eng 13 on hub 0
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring vcn_dec_0 uses VM inv eng 0 on hub 8
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring vcn_enc_0.0 uses VM inv eng 1 on hub 8
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring vcn_enc_0.1 uses VM inv eng 4 on hub 8
дек 06 13:18:40 legion kernel: amdgpu 0000:06:00.0: amdgpu: ring jpeg_dec uses VM inv eng 5 on hub 8

дек 06 13:18:40 legion kernel: [drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Failed to initialize parser -125!
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost
дек 06 13:18:40 legion kwin_wayland[2517]: Rendering a layer failed!
дек 06 13:18:40 legion kwin_wayland[2517]: Failed to find a working output layer configuration! Enabled layers:
дек 06 13:18:40 legion kwin_wayland[2517]: src QRectF(0,0 2560x1600) -> dst QRect(0,0 2560x1600)
дек 06 13:18:40 legion kwin_wayland[2517]: src QRectF(0,0 256x256) -> dst QRect(2552,-2 256x256)
дек 06 13:18:40 legion kwin_wayland[2517]: 0x2: GL_CONTEXT_LOST in context lost

[ToRRent1812]

Same issue, very annoying. Is there a non-flake way of reverting the firmware?

[Michael-D-Hart]

@ToRRent1812 I'm not an expert with nixos, but I believe all you have to do is try my fix without the flake part. I think it should still be valid.

[ToRRent1812]

@Michael-D-Hart Oh, you are right, I missed that part.
Thank You kindly.

[aviallon]

@K900 we are in dire need of some maintainer intervention here.
Imho, this is a bug of moderately high severity (because it is a regression, and renders many systems — amdgpu users — mostly unusable)

[K900]

I am still not sure how many systems are affected, because none of mine are. There should hopefully be a new linux-firmware tag in the next few days (it's about time for one), if that doesn't happen, we can revert.

[aviallon]

I believe having more than one GPU increases the problem.
I personally overriden my linux-firmware version to a specific git commit after a revert of the problematic file.

[make-42]

I'm also experiencing this.

I'd been racking my brain as to why my system would freeze as soon as I moved my mouse on GDM (on boot), I did everything from changing login managers, trying to run Niri through the tty to no avail. This seems to be the issue I'm facing.

System is with an RX 9070 XT.

[actuday6418]

On Ryzen 5 7600 with no discrete GPU, @Michael-D-Hart 's fix

To update to the latest 25.11 and not pin my flake to that commit I have an override:

configuration.nix

{ pkgs, ... }:{
 ....
  hardware.firmware = [
    (pkgs.linux-firmware.overrideAttrs (old: {
      version = "20251111";
      src = pkgs.fetchurl {
        # https://www.kernel.org/pub/linux/kernel/firmware/
        url = "https://www.kernel.org/pub/linux/kernel/firmware/linux-firmware-20251111.tar.gz";
        # > nix-prefetch-url https://www.kernel.org/pub/linux/kernel/firmware/linux-firmware-20251111.tar.gz
        sha256 = "0rp2ah8drcnl7fh9vbawa8p8c9lhvn1d8zkl48ckj20vba0maz2g";
      };
    }))
  ];
...
}

fixes it for me, but I still have rare stutters that last maybe 500ms each time, accompanied by kernel: [drm] Fence fallback timer expired on ring gfx_0.1.0 in journalctl.

[ToRRent1812]

I reverted to 20251111 and still get the same crashes, mainly during gameplay when gpu is at max. so I assume it needs to go even earlier, I will try 20251021.