Dear Daniel Here is a deeper protocol that I made with the help of mistral. Hope that gives further informations. The problem was NOT solved with a reinstall... here it is - best, andreas
___________ [IMPORTANT] gbm_create_device() EINVAL on Intel Alder Lake (8086:46a8) - GDM/gnome-shell fails at boot - NOT FIXED BY FULL REINSTALL --- === SYSTEM INFORMATION (fresh Ubuntu 26.04 LTS install) === Ubuntu: 26.04 LTS (Resolute Raccoon) Kernel: 7.0.0-28-generic GNOME Shell: 50.1-0ubuntu1.1 libmutter-18-0: 50.1-0ubuntu2.2 mutter-common: 50.1-0ubuntu2.2 libgbm1: 26.0.3-1ubuntu1 mesa-libgallium: 26.0.3-1ubuntu1 libegl-mesa0: 26.0.3-1ubuntu1 GPU: Intel Alder Lake-UP3 GT2 Iris Xe Graphics (PCI ID: 8086:46a8) Driver: i915 (kernel) + Mesa 26.0.3 --- === SYMPTOMS === 1. GDM starts but gnome-shell greeter crashes 3x with: "Failed to open gpu '/dev/dri/card1': Failed to initialize render device for /dev/dri/card1: Failed to create gbm device: Das Argument ist ungültig" "EGLStream render device requires an EGL display" "Failed to setup: No GPUs found" 2. GDM gives up: "GdmLocalDisplayFactory: maximum number of display failures reached. Giving up." 3. Result: BLACK SCREEN WITH FROZEN CURSOR (top-left), no login screen. --- === WORKAROUNDS (confirmed working) === - systemctl set-default multi-user.target → boots to TTY - sway (Wayland compositor) → WORKS PERFECTLY (proves GPU/kernel/Wayland stack is healthy) - sudo systemctl start gdm.service → WORKS INTERMITTENTLY (sometimes starts fine) --- === CRITICAL NEW EVIDENCE === Full Ubuntu 26.04 reinstall was performed on 2026-08-05 (as suggested by Daniel van Vugt in comment #12). THE BUG RECURRED ON 2026-08-06 WITH IDENTICAL SYMPTOMS. This disproves the "installation corruption" theory. The bug is either: - A persistent system-level state (systemd/logind/udev) - A genuine upstream bug in mutter/Mesa - A hardware/firmware state not cleared by reinstall --- === INTERMITTENT BEHAVIOR (new proof) === On 2026-08-06 (post-reinstall): - 18:06:59 → gdm.service start → CRASH (classic "Failed to open gpu" signature) - 18:22:36 → gdm.service start → SUCCESS (clean "Added device '/dev/dri/card1' ... Created gbm renderer" logs) This confirms: THE BUG IS INTERMITTENT, NOT A HARD BLOCKER. --- === PREVIOUS DIAGNOSIS (summary) === ✅ RULED OUT: - Hardware defect (Live USB test: GNOME starts perfectly) - Kernel bug (tested 7.0.0-27 and 7.0.0-28) - DRM master lock (cat /sys/kernel/debug/dri/1/clients → 0 clients) - AppArmor sandboxing (profile removed, bug persists) - Package version skew (apt install --reinstall full stack, bug persists) - Systemd sandboxing (no DeviceAllow/PrivateDevices in unit file) - Udev/seat tags (correctly tagged) - EGLStream leak (only 50_mesa.json in egl_vendor.d) - i386 package collisions (dpkg -V clean) 🔬 DEEP DIAGNOSIS (via gdb & strace): - gdb breakpoint on gbm_create_device(): * File descriptor passed to function is VALID (/dev/dri/card1, drm-driver: i915) * NO ioctl() calls inside function → pure userspace logic fails * Function reads PCI sysfs attributes 4x identically (all succeed) → still returns EINVAL - MESA_DEBUG=1 EGL_LOG_LEVEL=debug: * "EGL user error 0x3001 (EGL_NOT_INITIALIZED) in eglInitialize: DRI2: failed to load driver" (repeated 3x) --- === REPRODUCTION STEPS === 1. Set default boot target to graphical: sudo systemctl set-default graphical.target 2. Reboot: sudo reboot 3. After crash, collect logs: journalctl -b -0 -u gdm.service -o short-precise | grep -iE "gdm|gnome-shell|gbm|Failed" journalctl -b -0 -k | grep -iE "i915|drm|gpu" 4. Test workarounds: sudo systemctl set-default multi-user.target sudo reboot # Then manually: sway & # Always works # OR sudo systemctl start gdm.service # Sometimes works --- === TIMELINE === 2026-07-06: Mass purge cascade (apt remove libheif1 → purged gnome-shell, gdm3, ubuntu-desktop, etc.) 2026-07-07 09:32:56: First occurrence of bug (previously worked) 2026-07-08: multi-user.target set as workaround 2026-07-14: gdb tracing → gbm_create_device() fails without I/O 2026-07-21: Live USB test → GNOME starts perfectly → installation-specific 2026-07-25: Launchpad bug #2160652 filed 2026-08-02: Backup & reinstall scripts created 2026-08-05: Full Ubuntu 26.04 reinstall 2026-08-06: BUG RECURRED → REINSTALL DID NOT FIX --- === QUESTIONS FOR DEVELOPERS === 1. Is this a known issue with Intel Alder Lake + GNOME 50 + Mesa 26.0.3? 2. Are there similar bug reports we missed? 3. Is there a debug mode for gbm_create_device() (e.g., GBM_DEBUG=1)? 4. Should we test with strace -f -e trace=all on gnome-shell --mode=gdm? 5. Are there patches in newer mutter/Mesa versions (e.g., GNOME 52)? 6. Is multi-user.target + sway an acceptable temporary workaround? --- === ATTACHMENTS (if possible) === - /var/log/journal/ (GDM crash logs) - ~/mesa_debug.log (from MESA_DEBUG=1 EGL_LOG_LEVEL=debug LIBGL_DEBUG=verbose) - ~/gbm_debug.log (from gdb tracing session) --- === SUMMARY === This is an intermittent, installation-specific bug in mutter/Mesa's gbm_create_device() that is NOT fixed by: - Package reinstalls - Full system reinstall - Kernel updates - AppArmor removal The bug: - BLOCKS GDM/GNOME at boot (frozen cursor) - IS INTERMITTENT (sometimes gdm.service starts fine) - DOES NOT AFFECT sway/Live USB (GPU/Wayland stack is healthy) -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2160652 Title: gnome-shell/mutter 50.1 fails to start on Intel Alder Lake iGPU: gbm_create_device() returns EINVAL ("Failed to open gpu ... No GPUs found" To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/gnome-shell/+bug/2160652/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
