Bug 510390

Summary: Plasma Discover invocation is randomly blocked, until reboot, by an EAGAIN on a FUTEX call that requires 300 s to fail
Product: [Applications] Discover Reporter: Roke Julian Lockhart Beedell <4wy78uwh>
Component: discoverAssignee: Plasma Bugs List <plasma-bugs-null>
Status: RESOLVED DUPLICATE    
Severity: normal CC: aleixpol, john.liptrot, kdedev, nate
Priority: NOR    
Version First Reported In: 6.4.5   
Target Milestone: ---   
Platform: Fedora RPMs   
OS: Linux   
URL: https://discuss.kde.org/t/why-cant-i-open-discover/39560/5?u=rokejulianlockhart
See Also: https://bugs.kde.org/show_bug.cgi?id=507217
Latest Commit: Version Fixed/Implemented In:
Sentry Crash Report:
Attachments: The Output Of `time strace -Ttr plasma-discover`
The `perf.data`, And Its `.perfparser` Counterpart, Created By `perf record --debuginfod --call-graph dwarf plasma-discover`

Description Roke Julian Lockhart Beedell 2025-10-08 13:04:37 UTC
Created attachment 185601 [details]
The Output Of `time strace -Ttr plasma-discover`

# SUMMARY

Apparently randomly, I (and other users, including my brother, per https://discuss.kde.org/t/why-cant-i-open-discover/39560/5?u=rokejulianlockhart) become unable to invoke Discover, until we invoke `qdbus-qt6 org.kde.Shutdown /Shutdown org.kde.Shutdown.logoutAndReboot` (`org.kde.Shutdown.logout` might also work, but I've yet to confirm, due to how rarely this reproduces).

# STEPS TO REPRODUCE

Invoke `plasma-discover` (from `$Env:PATH`, in `bash`).

# OBSERVED RESULT

It closes with an `EAGAIN` of "Resource temporarily unavailable", but requires an inordinate amount of time to:

> ~~~
> real    5m8.252s
> user    0m0.307s
> sys     0m0.273s
> ~~~

Most notable is:

> ~~~
> 13:15:37 (+     0.000029) futex(0x55bcc7b568d8, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, FUTEX_BITSET_MATCH_ANY) = 0 <299.662141>
> 13:20:37 (+   299.662324) futex(0x55bcc7b56890, FUTEX_WAKE_PRIVATE, 1) = 0 <0.000085>
> ~~~

I've seen this reach 300 s, but that's a mere difference of 1 s.

# EXPECTED RESULT

It should invoke.

# SOFTWARE/OS VERSIONS

> ~~~
> Operating System: Fedora Linux 42
> KDE Plasma Version: 6.4.5
> KDE Frameworks Version: 6.18.0
> Qt Version: 6.9.2
> Kernel Version: 6.16.9-200.fc42.x86_64 (64-bit)
> Graphics Platform: Wayland
> ~~~

The hardware likely doesn't matter, due to the amount of hardware that this has occurred on.
Comment 1 Roke Julian Lockhart Beedell 2025-10-08 13:23:30 UTC
Created attachment 185602 [details]
The `perf.data`, And Its `.perfparser` Counterpart, Created By `perf record --debuginfod --call-graph dwarf plasma-discover`

Luckily, this is easy to accurately profile, because it closes itself. However, per https://discussion.fedoraproject.org/t/why-does-perf-lack-debuginfod-capabilities/167812, the perf file *may* lack symbols (although doesn't appear to).
Comment 2 Nate Graham 2025-10-08 19:51:20 UTC
When this happens, is there already a `plasma-discover` process running in the background, with no window?
Comment 3 Roke Julian Lockhart Beedell 2025-10-08 22:00:41 UTC
(In reply to Nate Graham from comment #2)

> When this happens, is there already a `plasma-discover` process running in
> the background, with no window?

I believe so, because I did see the system tray icon (whose context menu refused to invoke)! I've neglected to check System Monitor when it reproduces, so that's the best I can provide until it reproduces next.

Tangentially, when this occurs, the Task Manager icon is always stuck at approximately 10 % progress. At least, that's what its green underlay indicates.
Comment 4 Nate Graham 2025-10-09 15:43:05 UTC
Based on the symptoms, I'm declaring this a duplicate of Bug 507217.

*** This bug has been marked as a duplicate of bug 507217 ***
Comment 5 Bug Janitor Service 2025-10-26 21:07:16 UTC
A possibly relevant merge request was started @ https://invent.kde.org/plasma/discover/-/merge_requests/1194