Bug 2265568

After updating today, my desktop became unusable due to the recent firmware updates. The fans spin at maximum speed after a minute or two, with nothing obvious in system logs, even with no programs open.

AMD Ryzen 9 7950X3D
AMD Radeon RX 7900 XTX

I don't see anything that stands out in system logs. However, when the fan went nuts, I saw that the reported fan speed was not available, so I'm guessing that something either in the firmware or kernel stops reporting the fan speed, so it kicks into some emergency mode?

(Good thing I'm using Silverblue so I can roll back in the meantime. Too bad for everyone else who would be affected, who either don't know how or can't roll back.)

The likely culprit is probably part of this update. So far, I'm guessing either amd-gpu-firmware or amd-ucode-firmware, unless something in the kernel broke the interface to the AMD firmware.

These were all packages that were upgraded between versions (old -> new, according to rpm-ostree):

  amd-gpu-firmware 20240115-2.fc39 -> 20240220-1.fc39
  amd-ucode-firmware 20240115-2.fc39 -> 20240220-1.fc39
  atheros-firmware 20240115-2.fc39 -> 20240220-1.fc39
  brcmfmac-firmware 20240115-2.fc39 -> 20240220-1.fc39
  cirrus-audio-firmware 20240115-2.fc39 -> 20240220-1.fc39
  conmon 2:2.1.8-2.fc39 -> 2:2.1.10-1.fc39
  gvisor-tap-vsock 6:0.7.2-1.fc39 -> 6:0.7.3-1.fc39
  gvisor-tap-vsock-gvforwarder 6:0.7.2-1.fc39 -> 6:0.7.3-1.fc39
  intel-audio-firmware 20240115-2.fc39 -> 20240220-1.fc39
  intel-gpu-firmware 20240115-2.fc39 -> 20240220-1.fc39
  iwlegacy-firmware 20240115-2.fc39 -> 20240220-1.fc39
  iwlwifi-dvm-firmware 20240115-2.fc39 -> 20240220-1.fc39
  iwlwifi-mvm-firmware 20240115-2.fc39 -> 20240220-1.fc39
  kernel 6.7.4-200.fc39 -> 6.7.5-200.fc39
  kernel-core 6.7.4-200.fc39 -> 6.7.5-200.fc39
  kernel-modules 6.7.4-200.fc39 -> 6.7.5-200.fc39
  kernel-modules-core 6.7.4-200.fc39 -> 6.7.5-200.fc39
  kernel-modules-extra 6.7.4-200.fc39 -> 6.7.5-200.fc39
  libertas-firmware 20240115-2.fc39 -> 20240220-1.fc39
  libgphoto2 2.5.30-5.fc39 -> 2.5.31-1.fc39
  linux-firmware 20240115-2.fc39 -> 20240220-1.fc39
  linux-firmware-whence 20240115-2.fc39 -> 20240220-1.fc39
  mt7xxx-firmware 20240115-2.fc39 -> 20240220-1.fc39
  nvidia-gpu-firmware 20240115-2.fc39 -> 20240220-1.fc39
  nxpwireless-firmware 20240115-2.fc39 -> 20240220-1.fc39
  openvpn 2.6.8-1.fc39 -> 2.6.9-1.fc39
  realtek-firmware 20240115-2.fc39 -> 20240220-1.fc39
  tiwilink-firmware 20240115-2.fc39 -> 20240220-1.fc39


Reproducible: Always

Actual Results:  
Fans run at maximum speed. I'm unsure if it's just the GPU or all the system fans. It's very, very loud and probably not good for the hardware to run like this.

Expected Results:  
Fans should adjust automatically, as before.

A warm reboot doesn't solve the problem, even if I reboot into an older Silverblue deployment without the new firmware. The only workaround for now is to completely power off the machine and boot into the old version.

As this has the potential to break my computer if I run with the new firmware, I am a bit weary at trying to debug this too much further. But I'm happy to try relatively safe things that can be done within the timespan of a minute or two.