https://bugs.kde.org/show_bug.cgi?id=519877
Stephan Kulow <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Resolution|--- |WORKSFORME Status|REPORTED |RESOLVED --- Comment #3 from Stephan Kulow <[email protected]> --- You know what? I asked a friend! 🔍 What's Likely Happening Based on your report, this is almost certainly a GPU driver / Mesa / kernel-level crash, not a KPatience application bug per se. Here's the nerdy breakdown: Why KPatience specifically? KPatience uses Qt5 with OpenGL/Vulkan rendering for its card animations. When you drag a card, it triggers: Continuous GPU draw calls (animated card following your mouse) Compositing via KWin (or in your case, GNOME's Mutter compositor on Wayland) Rapid, repeated DMA buffer flips on the iGPU The crash mid-drag is a classic symptom of a GPU command buffer overflow, fence timeout, or TLB shootdown in the amdgpu kernel driver — which causes the kernel to either: Hang (GPU reset fails → system freeze) Panic (unrecoverable GPU fault → kernel panic → reboot) Why does it happen across multiple AMD iGPU generations? You mentioned this happened on: Vega (Ryzen 5 2400G) — Ubuntu 18.04 Vega (Ryzen 5 5600G) — Ubuntu 22.04 RDNA3 / Phoenix (Ryzen 7 7840HS) — Ubuntu 24.04 This strongly suggests a persistent bug in how the amdgpu driver handles the specific rendering pattern that KPatience's card drag animation produces — likely related to shared VRAM management on APUs (where system RAM is used as VRAM). 🛠️ Debugging Steps & Workarounds Step 1 — Enable Persistent Kernel Crash Logs Since you can't get logs after a crash, set up pstore / kdump to capture them: # Install kdump tools sudo apt install linux-crashdump # Enable pstore (survives reboots) sudo mkdir -p /sys/fs/pstore Also add this to your kernel boot parameters to get crash data written to disk: sudo nano /etc/default/grub Find GRUB_CMDLINE_LINUX_DEFAULT and add pstore.backend=efi to it, then: sudo update-grub After a crash and reboot, check: ls /sys/fs/pstore/ # or sudo dmesg | grep -i "amdgpu\|gpu\|hang\|reset\|fence\|timeout" | tail -50 Step 2 — Enable amdgpu Debug Logging Before Playing Run this before launching KPatience: # Enable amdgpu debug output sudo sh -c 'echo 0xffffffff > /sys/module/amdgpu/parameters/debug_mask' # Watch the kernel log live in a separate terminal sudo journalctl -f -k | grep -i "amdgpu\|drm\|gpu\|fence\|hang\|reset" Then launch KPatience and play until it crashes. If the system survives long enough, you might catch the error. After reboot: sudo journalctl -b -1 -k | grep -i "amdgpu\|drm\|fence\|hang\|reset\|error" The -b -1 flag reads the previous boot's kernel log — this is your best friend here! 🎯 Step 3 — Check Previous Boot Logs Right Now You can already do this without any setup: # Check if previous boot logs exist sudo journalctl -b -1 -k --no-pager | tail -100 # Specifically hunt for GPU issues sudo journalctl -b -1 -k | grep -iE "amdgpu|drm|gpu|fence|hang|reset|panic|oom|killed" | tail -50 Step 4 — Try Forcing Software Rendering for KPatience This is a workaround to confirm the GPU is the culprit. Launch KPatience with software rendering: LIBGL_ALWAYS_SOFTWARE=1 kpat If it never crashes in software rendering mode, that's your smoking gun — it's 100% a GPU driver issue. Step 5 — Try Disabling Vsync / Compositor Sync # Launch with different rendering hints QT_QPA_PLATFORM=xcb kpat # or try forcing X11 instead of Wayland Or try running KPatience under XWayland explicitly by launching it from an X11 session temporarily. Step 6 — Mesa & Kernel Version Check Your Mesa 25.2.8 is very recent (bleeding edge for 24.04!). Check if there are known regressions: # Check exact versions glxinfo | grep "OpenGL version" vainfo 2>/dev/null | head -5 vulkaninfo 2>/dev/null | grep "driverVersion" | head -3 # Check amdgpu firmware version sudo dmesg | grep -i "amdgpu.*firmware\|amdgpu.*version" | head -10 Step 7 — Undervolting / Power Cap Consideration Your system is undervolted with a 35W TDP cap. This is a potential suspect! When KPatience triggers rapid GPU draw calls, the GPU might: Request a power state transition Hit the power cap Cause a voltage droop below stable threshold Trigger a GPU fault Try temporarily removing the power cap and testing: # Check current power settings cat /sys/class/drm/card0/device/hwmon/hwmon*/power1_cap # Temporarily raise or remove cap (replace hwmon0 with your actual hwmon) sudo sh -c 'echo 0 > /sys/class/drm/card0/device/hwmon/hwmon0/power1_cap_max' Or test with your undervolting tool disabled entirely for a session. 🐛 Where to Report This Since this is reproducible across multiple AMD APU generations, this is definitely worth reporting upstream: Mesa bug tracker: https://gitlab.freedesktop.org/mesa/mesa/-/issues Linux kernel amdgpu: https://gitlab.freedesktop.org/drm/amd/-/issues -- You are receiving this mail because: You are watching all bug changes.
