Recently I've been having dgpu crashes that take out the whole computer, and then I get behavior where the dgpu doesn't show up or load properly when rebooting or coming back online. The support folks at framework don't seem to know what's going on. They say it's an OS config error. It happens on both ubuntu and fedora live disks though, as well as my main nixos install. It started happening about 2-3 weeks after I made my latest config change and I hadn't changed or updated anything since then. Here's the only related dmesg stuff I can find on it.

[    3.828129] amdgpu 0000:03:00.0: amdgpu: get invalid ip discovery binary signature
[    3.828166] [drm:amdgpu_discovery_set_ip_blocks [amdgpu]] *ERROR* amdgpu_discovery_init failed
[    3.828419] amdgpu 0000:03:00.0: amdgpu: Fatal error during GPU init
[    3.828477] amdgpu 0000:03:00.0: amdgpu: amdgpu: finishing device.
[    3.828621] amdgpu 0000:03:00.0: probe with driver amdgpu failed with error -22

Anyone seen this or know what's going on? Happy to provide further logs too.

  services.udev = {
    enable = true;
    extraRules = ''
      ACTION=="add", SUBSYSTEM=="acpi", DRIVERS=="button", ATTRS{hid}=="PNP0C0D", ATTR{power/wakeup}="disabled"
      ACTION=="add", SUBSYSTEM=="serio", DRIVERS=="atkbd", ATTR{power/wakeup}="disabled"
    '';

  systemd.tmpfiles.rules =
    builtins.map (dev: "w+ /sys/bus/pci/devices/${dev}/power/wakeup - - - - disabled")
      [
        "0000:00:02.2" # GPP6
        "0000:00:03.1" # GP11
        "0000:00:04.1" # GP12
        "0000:c1:00.3" # XHC0
        "0000:c1:00.4" # XHC1
        "0000:c3:00.3" # XHC3
        "0000:c3:00.4" # XHC4
        "0000:c3:00.5" # NHI0
        "0000:c3:00.6" # NHI1
      ];

This is my current, non-working attempt FWIW