https://bugs.kde.org/show_bug.cgi?id=527654
--- Comment #1 from [email protected] --- Additional info (had an LLM dig through my logs): KWin tears down the whole GPU on resume, even though the kernel never removes it. After resume, KWin gets ~10 udev "change" events for card1 within ~3s, then: kwin_wayland: Received change event for monitored drm device "/dev/dri/card1" (x9) kwin_wayland: Removing output KWin::DrmOutput(...) (x4, all outputs) kwin_wayland: Removing GPU "/dev/dri/card1" kwin_wayland: drmGetDeviceFromDevId failed! No such file or directory (x~20) kwin_wayland: adding GPU /dev/dri/card1 kwin_wayland: New output on GPU /dev/dri/card1: VA24D / VG248 / E2442 / VA24D The kernel log for the same resume shows no DRM device removal, no NVRM/Xid errors, and /dev/dri/card1 keeps the same node (226:1). So the GPU removal seems to come from how KWin handles the change events, not from the driver. Consequences while outputs are gone: - Every Qt client logs "There are no outputs - creating placeholder screen" - kscreenlocker_greet segfaults in libQt6Quick.so.6.12.0 and KWin restarts the lock (probably also why the lock screen shows the default 6.8 wallpaper) Reproduces on every resume with kwin 6.8.1-1 and the Arch rebuild 6.8.1-2. The problem started with the Plasma 6.7.5 -> 6.8.0 upgrade (with NVIDIA 615.71 -> 615.78 in the same update). -- You are receiving this mail because: You are watching all bug changes.
