I have been seeing occasional deadlocks since upgrading to Plasma 6. The problem first appeared after upgrading from Plasma 5 (don’t remember the minor version) to Plasma 6.4.5. I upgraded to Plasma 6.5 yesterday and it’s just happened again.
The frequency of the problem is random, averaging about once every 2 or 3 days. The only pattern that I have seen is that it always happens when clicking on some UI element. This morning, it happened when I hovered over an icon on the icons-only task manager. I have also seen it happen when various applications create a new window.
Once the problem happens, the UI becomes unresponsive and all clicks and key presses seem to be ignored, with one exception. The exception is that if I use ctrl-alt-del, it shows the screen to select log out, sleep, restart, etc. That screen itself is then unresponsive and I cannot interact with the computer at all, except to do a hard reset. This results in a loss of any work in progress.
After this morning’s deadlock, I ran the following command, and only 3 lines were shown before the reboot:
user@host:~$ journalctl --user -u plasma-plasmashell.service -S 2025-10-24 Oct 24 06:08:55 fly-by-night plasmashell[7495]: kf.plasma.quick: Exposed with no visual parent. Window positioning broken. Oct 24 06:08:55 fly-by-night plasmashell[7495]: kf.plasma.quick: Exposed with no visual parent. Window positioning broken. Oct 24 06:08:55 fly-by-night plasmashell[7495]: **kf.plasma.quick: Exposed with no visual parent. Window positioning broken.
**
I would appreciate any suggestions either for a resolution, troubleshooting steps to try, or any suggestions for logging or other data collection to help narrow it down.
Deadlocks are due either to kernel/drivers bugs, X11 or kwin (in X11 term, the window manager). But here since ctrl-alt-del works, the kernel is probably fine and X11 too.
But since sddm (the login manager) is also concerned, it means it is related to X11.
The input seems to be in shambles.
Still the kwin logs might have something interesting.
journalctl --user -t kwin_x11 > kwin-x11.log
Then share the file.
There is an option to show the logs up until the last crash. But the crash might not be in KWin. KWin could see detect something wrong happening and log something.
The worst thing that plasmashell can do is to show a black screen with a cursor on top.
Did you try a plasma Wayland session recently ? It probably won’t have the same issue.
I should have been a bit more clear about the deadlock. The kernel is still running. The mouse pointer moves, and I have a some widgets that show CPU usage, CPU temperature, memory usage, disk I/O and network I/O that are still updating. The mouse pointer does not change when I move over window borders, etc. I have not used a debugger to actually confirm that there are deadlocked threads. “Non-responsive” is a better description.
Regarding Wayland, I cannot use it. One of my main applications is an astrophotography application called PixInsight that does not support Wayland and is effectively unusable under XWayland due to numerous application UI elements appearing and disappearing as the mouse moves, so I have to use x11.
I have created the kwin_x11 log that you requested. Unfortunately, the forum does not allow me to post links. I just created a login this morning, and I suspect that I don’t have full user permissions yet. Here is the URL minus the HTTPS protocol:
Oct 24 06:14:22 fly-by-night kwin_x11[2286]: The X11 connection broke: I/O error (code 1)
Oct 24 06:14:22 fly-by-night kwin_x11[2286]: XIO: fatal IO error 4 (Interrupted system call) on X server “:0” Oct 24 06:14:22 fly-by-night kwin_x11[2286]: after 35873 requests (35873 known processed) with 0 events remaining.
Oct 24 06:14:22 fly-by-night kwin_x11[2286]: ASSERT: “m_cache.isEmpty()” in file ./src/scene/workspacescene_opengl.cpp, line 125
Oct 24 06:14:22 fly-by-night kwin_x11[2286]: Application::crashHandler() called with signal 6; recent crashes: 1 Oct 24 06:14:23 fly-by-night kwin_x11[2286]: KCrash: Application ‘kwin_x11’ crashing… crashRecursionCounter = 2
The ASSERT: "m_cache.isEmpty()" in file ./src/scene/workspacescene_opengl.cpp, line 125 is quite clear and can help to debug this.
The next step would be the stack trace, the command will take a hot minute to run and should download the debug symbols as needed (if they don’t download turn on debuginfod on your system and rerun):
coredumpctl debug kwin_x11 -A "-ex bt -ex quit" |& tee backtrace.txt
The coredumpctl command results in “No match found.” If I do “coredump list”, it shows one with the path /usr/bin/plasmashell, but it’s dated 2025-10-22. It is the most recent item listed.
Is there something that I need to do in order to ensure that a dump file is created next time? Since this problem is random, I cannot induce it. It usually takes a few days after booting before it happens. Sometimes, it might be up to a week.
I am not able to file a bug report because when I try to log into the link that you provided above, it fails with the following error message:
”The login or password you entered is not valid.”
If I visit bugs.kde.org directly, it shows me the home page, but when I click on the link to file a bug, it fails the same way. Are separate accounts required for the forum and the bug reporting system?