https://bugs.kde.org/show_bug.cgi?id=521757
Bug ID: 521757
Summary: plasmashell 6.7.0 pegs one CPU core after update to
6.7.0; perf shows KActivities::Stats / SQLite;
clearing kactivitymanagerd resources fixes it
Classification: Plasma
Product: plasmashell
Version First 6.7.0
Reported In:
Platform: CachyOS
OS: Linux
Status: REPORTED
Severity: normal
Priority: NOR
Component: Activities in general
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Target Milestone: 1.0
DESCRIPTION
After upgrading to Plasma 6.7.0, plasmashell immediately started using 100% of
one CPU core on login/restart. Stopping plasma-plasmashell.service immediately
removed the CPU load. Profiling showed significant time in
libPlasmaActivitiesStats.so.6.7.0, libQt6Sql.so.6.11.1, and libqsqlite.so.
Moving ~/.local/share/kactivitymanagerd/resources out of the way and recreating
the directory fixed the issue.
This looks like a plasmashell / KActivitiesStats / SQLite resource-history
issue, possibly triggered by existing activity/resource data after the Plasma
6.7.0 update? I don't know enough about KDE Plasma for more info. Debugging was
largely together with chatGPT, the fix works.
STEPS TO REPRODUCE
1. Take an existing cachyOS system (sorry, I didn't simplify it to a minimal
example)
2. Upgrade to 6.7.0
3. See CPU 100% use a single core
Reproducible on my existing user profile before clearing
~/.local/share/kactivitymanagerd/resources.
I no longer have the active broken state after applying the workaround, but I
kept a backup of the old resources directory created during the workaround. I
can provide it privately if needed, depending on whether it contains sensitive
recent-file/activity data.
OBSERVED RESULT
Immediately after starting or restarting plasmashell, it goes to 100% CPU usage
and stays there.
Restarting the service:
systemctl --user restart plasma-plasmashell.service
causes the CPU load to return immediately.
Stopping the service:
systemctl --user stop plasma-plasmashell.service
immediately removes the CPU load.
EXPECTED RESULT
plasmashell should not peg one CPU core indefinitely because of existing
KActivities resource/history data.
SOFTWARE/OS VERSIONS
Operating System (available in the Info Center app, or by running `kinfo` in a
terminal window):
Distribution: CachyOS
Session: Wayland
Hardware: Framework laptop with AMD Radeon 780M / Phoenix1 iGPU using amdgpu
Relevant package versions:
plasmashell 6.7.0
plasma-workspace 6.7.0-1.1
plasma-desktop 6.7.0-1.1
qt6-base 6.11.1-1
mesa 2:26.1.2-1
linux-cachyos 7.0.12-1
Graphics stack:
GPU: AMD Radeon 780M Graphics, driver amdgpu
OpenGL renderer: AMD Radeon 780M Graphics (radeonsi phoenix ACO DRM 3.64
7.0.12-1-cachyos)
Mesa: 26.1.2
Wayland compositor: kwin_wayland
ADDITIONAL INFORMATION
The most relevant diagnostic clue is the perf output pointing at:
libPlasmaActivitiesStats.so.6.7.0
KActivities::Stats::ResultSet::~ResultSet()
libQt6Sql.so.6.11.1
QSqlQuery::~QSqlQuery()
libqsqlite.so
and the fact that moving ~/.local/share/kactivitymanagerd/resources fixed the
CPU loop.
--
You are receiving this mail because:
You are watching all bug changes.
