Public bug reported:
### Summary Ubuntu 24.04.4 LTS on a GNOME Wayland session is showing repeated Mutter / GNOME Shell instability on a multi-monitor AMDGPU setup. The most visible incidents were: - `2026-05-22`: severe desktop degradation while Chrome/YouTube was active; - `2026-05-25 16:59-17:00 BST`: crash/reset context while Boomaga manual duplex was open; - recurring lower-severity assertions later in the same `2026-05-25` session during ordinary app opens. The recurring markers are: - `meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed` - monitor/work-area assertions after display-layout churn - `Invalid window geometry for xdg_surface` - kernel `REG_WAIT timeout 1us * 100 tries - dcn32_program_compbuf_size` This looks like a Mutter/Wayland/display-topology problem that may be exposed or amplified by the AMDGPU DCN path and/or extension/app trigger contexts. ### Environment - Ubuntu: `24.04.4 LTS` - Session: `Wayland` - GNOME Shell package: `46.0-0ubuntu6~24.04.14` - GNOME Shell runtime: `GNOME Shell 46.0` - Mutter runtime from journal: `Running GNOME Shell (using mutter 46.2) as a Wayland display server` - Installed Mutter libraries: - `libmutter-14-0 46.2-1ubuntu0.24.04.15` - `mutter-common 46.2-1ubuntu0.24.04.15` - `mutter-common-bin 46.2-1ubuntu0.24.04.15` - Kernel: `6.17.0-29-generic` - Mesa Vulkan drivers: `25.2.8-0ubuntu0.24.04.1` - libdrm: `2.4.125-1ubuntu0.1~24.04.1` - Firmware package: `linux-firmware 20240318.git3b128b60-0ubuntu2.27` - Primary display GPU: `0000:14:00.0 AMD/ATI Navi 33 [Radeon Pro W7500]` - Secondary AMD GPU also present: `0000:03:00.0 AMD/ATI device 1002:7551` ### Display Topology At incident time on `2026-05-25`, GNOME had four logical 4K outputs active on the W7500: - `DP-4` - `DP-5` - `DP-6` - `DP-7` (Hisense TV path) After the crash, current live state showed: - active logical monitors: `DP-4`, `DP-5`, `DP-6` - `DP-7` remained physically `connected` in `/sys/class/drm`, but was no longer in the active GNOME layout That matters because the logs went quiet after DP-7 was removed from the logical layout, even though the cable/TV path remained electrically present. ### Symptoms - GNOME Shell / desktop degradation - missing dock/panel/desktop icons in prior incidents - Power Off dialog failure in earlier related incident - on `2026-05-25`, abrupt crash/reset context around `17:00 BST` - after reboot, ordinary app opens still triggered repeated Mutter assertions ### Reproduction / Context This is not yet reduced to a minimal upstream reproducer. Current strongest context is: 1. GNOME Wayland session 2. Radeon Pro W7500 driving a heavy multi-4K layout 3. DP-7 / Hisense TV path present; before mitigation it was also an active 4K logical monitor 4. One incident occurred while Boomaga manual duplex UI was open 5. Repeated `meta_window_set_stack_position_no_sync` assertions are also reproducible later from ordinary app launches such as Extension Manager and LibreOffice, so Boomaga does not look like the sole cause ### Key Log Excerpts From the crash window on boot `-1` (`2026-05-25`): ```text May 25 16:55:01 org.boomaga[...] QSocketNotifier: Can only be used with threads started with QThread May 25 16:57:06 gnome-shell[...] meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed May 25 16:59:02 org.boomaga[...] QSocketNotifier: Can only be used with threads started with QThread May 25 17:00:12 gnome-shell[...] Could not create transient scope for PID ... No such process ``` Immediately after reboot on boot `0`: ```text May 25 17:07:32 gnome-shell[...] Running GNOME Shell (using mutter 46.2) as a Wayland display server May 25 17:07:36 kernel: amdgpu 0000:14:00.0: amdgpu: [drm] REG_WAIT timeout 1us * 100 tries - dcn32_program_compbuf_size line:147 May 25 17:10:17 gnome-shell[...] meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed May 25 17:10:44 gnome-shell[...] Invalid window geometry for xdg_surface@49. Ignoring for now, but this will result in client termination in the future. May 25 17:22:34 gnome-shell[...] Invalid window geometry for xdg_surface@30. Ignoring for now, but this will result in client termination in the future. ``` After disabling DP-7 as a logical display, GNOME briefly logged display-layout assertions during the reconfiguration itself: ```text May 25 18:17:27 gnome-shell[...] meta_monitor_manager_get_logical_monitor_from_number: assertion '(unsigned int) number < g_list_length (manager->logical_monitors)' failed May 25 18:17:27 gnome-shell[...] meta_workspace_get_work_area_for_monitor: assertion 'logical_monitor != NULL' failed May 25 18:17:27 gnome-shell[...] meta_display_get_monitor_in_fullscreen: assertion 'monitor >= 0 && monitor < n_logical_monitors' failed ``` Later in the same session, even with DP-7 out of the logical layout, ordinary app opens still triggered the stack-position assertion: ```text May 25 18:34:13 gnome-shell[...] meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed May 25 18:41:06 gnome-shell[...] meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed May 25 18:45:56 gnome-shell[...] meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed May 25 18:46:14 gnome-shell[...] JS ERROR: TypeError: this.actor is null ``` ### Mitigation Observation The strongest mitigation observation so far is: - when `DP-7` / the Hisense TV path was removed from the active GNOME logical layout, the severe AMDGPU/DCN and Mutter marker stream stopped for the following observation window; - however, lower-level `meta_window_set_stack_position_no_sync` assertions were still reproducible from ordinary app launches afterward. That suggests the active four-display topology, especially with DP-7 included, is a strong trigger or amplifier, but not the whole story. ### Extensions Present Enabled during current investigation: - `[email protected]` - `just-perfection-desktop@just-perfection` - `[email protected]` - `[email protected]` - `[email protected]` - `[email protected]` - `[email protected]` - `[email protected]` I have not yet reduced this to a minimal "extensions disabled" reproducer, so this is being filed against Ubuntu Mutter first rather than upstream GNOME. ### Crash Artifacts - No fresh same-day `/var/crash` artifact was found for this incident. - The only present crash file was older: - `/var/crash/_usr_bin_gnome-shell.1000.crash` dated `2026-05-17` - `coredumpctl` is not installed on this machine. ### Related Local Investigation Notes - `docs/complex_tasks/ops-2026-05-25-gnome-boomaga-display-stability-suspects.md` - `docs/complex_tasks/ops-2026-05-22-gnome-amdgpu-chrome-display-crash.md` - `docs/complex_tasks/ops-2026-05-14-s3-resume-hard-poweroff.md` - `docs/complex_tasks/ops-2026-05-08-gnome-session-crash.md` ### What I am asking Please advise whether this looks like: 1. a known Mutter issue in Ubuntu's 46.2 packaging, 2. a Mutter/Wayland handling issue triggered by this monitor topology, 3. an AMDGPU/DCN interaction that should instead move to another Ubuntu package, or 4. something that should only be taken upstream after a more minimal repro. I can attach the saved Apport report and provide more focused journal excerpts if there is a preferred package or duplicate to follow. ## Local Artifacts - Offline Apport bundle: `/tmp/mutter-2026-05-25.apport` ** Affects: mutter (Ubuntu) Importance: Undecided Status: New ** Attachment added: "mutter-2026-05-25.apport" https://bugs.launchpad.net/bugs/2154222/+attachment/5973440/+files/mutter-2026-05-25.apport -- You received this bug notification because you are a member of Ubuntu Desktop Bugs, which is subscribed to mutter in Ubuntu. https://bugs.launchpad.net/bugs/2154222 Title: Ubuntu 24.04.4 LTS on a GNOME Wayland session is showing repeated Mutter / GNOME Shell instability on a multi-monitor AMDGPU setup To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/mutter/+bug/2154222/+subscriptions -- desktop-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/desktop-bugs
