https://bugs.kde.org/show_bug.cgi?id=522477
Bug ID: 522477
Summary: drkonqi-coredump-launcher crash-loops and fills the
disk when triggered with no active graphical session
Classification: Applications
Product: drkonqi
Version First 6.7.1
Reported In:
Platform: Fedora RPMs
OS: Linux
Status: REPORTED
Severity: major
Priority: NOR
Component: general
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---
Created attachment 193924
--> https://bugs.kde.org/attachment.cgi?id=193924&action=edit
Coredump of drkonqi-coredump-launcher
DESCRIPTION
drkonqi-coredump-launcher gets into an infinite crash loop when it's invoked
with no active graphical session (e.g. sitting at the SDDM greeter, or a
text-console login with no Plasma running). The launcher is a QGuiApplication
and aborts on startup with no display; because it's launched by the
systemd-coredump pipeline, its own crash generates a core that re-triggers the
pipeline, which launches it again, repeating ~2.5/second and filling
`/var/lib/systemd/coredump` and the journal (GBs) until stopped.
STEPS TO REPRODUCE
1. On a machine with KDE installed, log out to a text VT (do not log into
SDDM/Plasma, so there is no graphical session).
2. Cause one core to be handled by the launcher, e.g. run it once:
`env -i XDG_RUNTIME_DIR=/run/user/$(id -u) HOME="$HOME"
/usr/libexec/drkonqi-coredump-launcher`
3. Watch: `coredumpctl list | grep drkonqi-coredump-launcher`
OBSERVED RESULT
An endless, self-sustaining loop of drkonqi-coredump-launcher crashes (Signal 6
/ SIGABRT). The systemd user-unit instance counter climbs:
drkonqi-coredump-launcher@0-…, @1-…, @2-… Each crash spawns the next. It fills
disk/journal until I mask the units. Backtrace shows qFatal in main() at
coredump/launcher/main.cpp: line 248 (QGuiApplication ctor, no QPA platform)
and line 253 (qFatal "Not exactly one fd passed by systemd. Quel malheur!").
EXPECTED RESULT
The launcher should exit cleanly (non-fatally) when there's no graphical
session / no data to show, and the pipeline should never reprocess the
launcher's own crashes. A coredump handler must not be able to loop on its own
failure.
SOFTWARE/OS VERSIONS
Operating System: Fedora Linux 44
KDE Plasma Version: 6.7.1
KDE Frameworks Version: 6.27.0
Qt Version: 6.11.1
ADDITIONAL INFORMATION
Package: plasma-drkonqi-6.7.1-1.fc44. rpm -V is clean.
Reproduced on two separate Fedora 44 machines on 6.7.1: it loops only when
there is no active graphical session; with a normal Plasma session it never
loops.
Workaround: `sudo systemctl mask [email protected]
drkonqi-coredump-launcher.socket`
Attaching a coredump of one launcher crash.
--
You are receiving this mail because:
You are watching all bug changes.
