plasmashell 6 is unstable when using PREEMPT_RT kernel

Help
,
1

I spent a long time troubleshooting this issue, but nothing actualy found. So just share my observations.

My environment:

  • Distro: Debian Unstable (Sid)

  • Desktop: Plasma 6 Wayland

  • Kernel: linux-image-rt-amd64 (PREEMPT_RT)

  • Display: Dual monitors, Internal display set to 165% scaling

In daily use, plasma-plasmashell.service might restart 3-5 times by systemd in a day, but log just show:
systemd\[2116\]: plasma-plasmashell.service: Main process exited, code=killed, status=9/KILL

I tried rebuilding the plasmashell applet config from scratch, but the issue persisted.

I recently switched back to the generic kernel (linux-image-amd64 with lazy/full dynamic preempt). The issue are gone and the desktop is perfectly stable now.

This is strange, as most people don’t use RT kernels for desktop.

1 Like
2

Quite the opposite, this means it is under-tested.
But there isn’t a reason why it shouldn’t work, there is probably a data-race appearing then.

What does this command return ; make sure to have debuginfod enabled (Guidelines and HOWTOs/Debugging/How to create useful crash reports - KDE Community Wiki):

coredumpctl debug plasmashell -A "-ex bt -ex quit" |& tee backtrace.txt
3

For this issue, only left messages in journalctl(signaled sigkill) and no coredump. systemdrestart plasmashellinstantly.

I approve of data-race in this issue. It appears to be triggered mostly when a new notification arrives.

4

You can enable additional log output by plasma using KDebugSettings - Applications de KDE
This should help us determine where the error happens.

You might need to restart plasmashell for the settings to take effect, I am not sure.