The desktop freezes during a normal work session: both monitors stop updating, mouse/keyboard are dead, the machine stays powered, a hard power-cycle is required. This is not a shutdown/reboot hang.
Hardware / OS: - Ubuntu 26.04 LTS, kernel 7.0.0-34-generic, XFCE on X11 - AMD Ryzen 7 7700X, iGPU Raphael (amdgpu, PCI 1002:164e) drives the displays (DisplayPort + HDMI, 2x1080p, 3840x1080 desktop) - NVIDIA GeForce RTX 5070 (PCI 10de:2f04), open kernel module 595.91.07 - xrandr providers: NVIDIA-0 = Source Output; AMD = Sink Output + Source Offload - glxinfo renderer = NVIDIA GeForce RTX 5070 (also with DRI_PRIME=0 and DRI_PRIME=1) - Mesa 26.0.3 Observed trigger (same night, graphical session already open): - pCloud Drive AppImage 2.3.0 (Electron 39.8.10 / Chromium 142), command line --no-sandbox --ozone-platform=x11, GPU acceleration enabled - Telegram Desktop Flatpak 7.2.9 (GNOME runtime 50) which installed two Flatpak NVIDIA GL stacks at once: org.freedesktop.Platform.GL.nvidia-595-71-05 and nvidia-595-91-07 (plus GL32 595-71-05) Kernel log (hostname stripped): overnight / frozen session: amdgpu 0000:0f:00.0: [drm] REG_WAIT timeout 1us * 100 tries - dcn31_program_compbuf_size line:141 WARNING: drivers/gpu/drm/amd/amdgpu/../display/dc/hubbub/dcn31/dcn31_hubbub.c:151 at dcn31_program_compbuf_size ... kworker amdgpu 0000:0f:00.0: [drm] REG_WAIT timeout 1us * 100 tries - dcn31_program_compbuf_size line:142 INFO: task pcloud.bin blocked for more than 122 seconds (D state) later the same day, both screens froze: amdgpu 0000:0f:00.0: [drm] *ERROR* [CRTC:364:crtc-0] hw_done or flip_done timed out amdgpu 0000:0f:00.0: [drm] *ERROR* [CRTC:368:crtc-1] hw_done or flip_done timed out Question: what mitigation is recommended when amdgpu owns the displays, GLX is on NVIDIA, Electron 39 is GPU-accelerated, and Flatpak has two NVIDIA GL runtimes? Thanks.
