https://bugs.kde.org/show_bug.cgi?id=522578
Bug ID: 522578
Summary: Issue 1 — `systemctl --user restart
plasma-kwin_wayland.service` crashes the whole session
instead of gracefully restarting the compositor
Classification: Plasma
Product: kwin
Version First 6.7.0
Reported In:
Platform: Neon
OS: Linux
Status: REPORTED
Severity: normal
Priority: NOR
Component: core
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---
Created attachment 193986
--> https://bugs.kde.org/attachment.cgi?id=193986&action=edit
auto-generated by CLAUDE via SSH , because i can't see anything into MONITOR
# KDE Bug Report — multi-GPU Wayland session instability (KDE Neon / Plasma
6.7.0)
This file describes **three separate reproducible issues**. On bugs.kde.org
each should be
filed as its **own** report (one issue per ticket). Pick the component from the
section header.
---
## Environment
| Item | Value |
|---|---|
| Distro | KDE neon User Edition (based on Ubuntu 24.04 noble) |
| Plasma | 6.7.0 (`plasmashell --version`, `kwin_wayland --version`) |
| kglobalaccel | 6.26.0 |
| KDE Connect | 26.04.2 |
| SDDM | Xorg greeter (default) |
| Kernel | 6.17.0-35-generic |
| Session | Plasma Wayland |
| GPUs | **Two AMD GPUs:** |
| | • `03:00.0` Navi 21 — Radeon RX 6900 XT (discrete, `/dev/dri/card1`,
renderD128) — **no display attached** |
| | • `0e:00.0` Cezanne — Radeon Vega APU (integrated, `/dev/dri/card2`,
renderD129) — **monitor on HDMI-A-2** |
| Display | Single monitor, 1920×1080@60, connected to **card2-HDMI-A-2 (Vega
APU)** |
| Driver | amdgpu (in-tree) |
Key consequence: the **display is on the *secondary* GPU** (integrated Vega),
while the
primary discrete GPU (RX 6900 XT) has no attached display. This is the common
factor in
issues #2 and #3.
---
## Issue 1 — `systemctl --user restart plasma-kwin_wayland.service` crashes the
whole session instead of gracefully restarting the compositor
**Component:** `kwin` (Wayland compositor) / `plasma-workspace`
### Summary
On Plasma 6, restarting the kwin Wayland compositor via its systemd user unit
is expected to
restart the compositor in place (native Wayland clients survive). On this setup
it instead
**tears down the entire Plasma session** — a cascade of Plasma services
core-dump and the
user is dropped back to SDDM.
### Steps to reproduce
1. Log in to a Plasma Wayland session (multi-GPU setup as above).
2. Run `systemctl --user restart plasma-kwin_wayland.service`.
### Expected
kwin restarts; native Wayland windows reconnect and survive; session continues.
### Actual
The session dies. From the user journal (`journalctl --user`), all within the
same second:
```
systemd[1688]: Stopping plasma-kwin_wayland.service - KDE Wayland Compositor...
systemd[1688]: Stopped plasma-kwin_wayland.service - KDE Wayland Compositor.
systemd[1688]: Starting plasma-kwin_wayland.service - KDE Wayland Compositor...
systemd[1688]: Started plasma-kwin_wayland.service - KDE Wayland Compositor.
kwin_wayland[...]: No backend specified, automatically choosing Wayland because
WAYLAND_DISPLAY is set
systemd[1688]: plasma-kaccess.service: Failed with result
'core-dump'.
systemd[1688]: plasma-xembedsniproxy.service: Failed with result
'core-dump'.
systemd[1688]: plasma-plasmashell.service: Failed with result
'core-dump'.
systemd[1688]: plasma-kded6.service: Failed with result
'core-dump'.
systemd[1688]: plasma-xdg-desktop-portal-kde.service: Failed with result
'core-dump'.
systemd[1688]: plasma-powerdevil.service: Failed with result
'core-dump'.
… then services repeatedly time out and the session returns to SDDM.
```
### Notes
A crash dump was recorded for kwin_wayland (`coredumpctl list` shows a
`kwin_wayland`
SIGTRAP core). DrKonqi could not associate KCrash metadata.
---
--
You are receiving this mail because:
You are watching all bug changes.
