Gnome Shell crash and GPU ring timeout on AMD GPU when using Brave browser (Fedora 42)

Ask Fedora
1

Hi everyone

I’m experiencing repeated crashes of the GNOME Shell environment due to GPU timeouts related to the amdgpu driver when using the Brave browser under Fedora 42. This crash results in a complete loss of the graphical session and often forces a logout or reboot.

Below is an excerpt from journalctl -b -1 -p 3 showing the relevant logs from the crash:

Apr 21 08:49:29 fedora kernel: amdgpu 0000:01:00.0: amdgpu: ring gfx timeout, signaled seq=58600, emitted seq=58602
Apr 21 08:49:29 fedora kernel: amdgpu 0000:01:00.0: amdgpu: Process information: process brave pid 3625 thread brave:cs0 pid 3664
Apr 21 08:49:32 fedora systemd-coredump[6215]: [🡕] Process 2154 (gnome-shell) of user 1000 dumped core.
                                               
Module /usr/bin/gnome-shell from rpm gnome-shell-48.1-1.fc42.x86_64
...

Additional context:

  • The problem consistently happens while using Brave (installed via RPM) with hardware acceleration enabled.
  • Disabling hardware acceleration in Brave seems to prevent the crash.
  • GPU model: AMD Radeon™ R7 200 Series

Steps to reproduce:

  1. Use Fedora 42 with GNOME + Wayland session.
  2. Launch Brave browser with hardware acceleration enabled.
  3. Open a few tabs and use the browser normally (e.g., video playback or animations).
  4. After a while, the screen freezes, and the graphical session crashes.

Would appreciate any help or insight into whether this is a bug in Mesa, amdgpu, Wayland, or GNOME Shell itself. Let me know if you need full journalctl output or coredumpctl logs.

Thanks!