Bug 410262

Summary: plasmashell uses 100% cpu
Product: [Plasma] plasmashell Reporter: Kevin Gilbert <kmg952>
Component: generalAssignee: David Edmundson <kde>
Status: RESOLVED WORKSFORME    
Severity: normal CC: plasma-bugs-null
Priority: NOR    
Version First Reported In: 5.15.5   
Target Milestone: 1.0   
Platform: Other   
OS: Linux   
Latest Commit: Version Fixed In:
Sentry Crash Report:
Attachments: Output from perf report

Description Kevin Gilbert 2019-07-27 01:36:05 UTC 
SUMMARY
As for the summary, plasmashell constantly uses 100% of a cou


STEPS TO REPRODUCE
Don't know

OBSERVED RESULT
100% cpu via top

EXPECTED RESULT


SOFTWARE/OS VERSIONS
Linux/KDE Plasma: 
(available in About System)
KDE Plasma Version: 
KDE Frameworks Version: 
Qt Version: 

kf5-plasma.x86_64 5.59.0-1.fc30
qt.x86_64 1:4.8.7-49.fc30

Note: this is a stock standard, up to date Fedora 30 system

ADDITIONAL INFORMATION
There are lots of items in a Google search regarding this bug. I have tried lots of them to no avail.

It also appears to be a common problem over many years. Can we please fix this once and for all?
Comment 1 David Edmundson 2019-07-28 18:03:50 UTC 
I can't do anything with the little information provided.

Please try and get a report in hotspot

Failing that run "sudo gdb --pid `pidof plasmashell` wait for it to attach and type "bt" - and paste the output here. Then type "continue" in that shell and repeat a few times.

Please try and note some conditions when this occurs. New notification for example?
Comment 2 Kevin Gilbert 2019-07-28 21:04:48 UTC 
Firstly, brevity: sending from phone because hdd problems on laptop.

Understand re lack of docs - didn't know what to provide. 

Further details after hdd replaced.

Lastly, in this context, what is "hotspot"?
Comment 3 Christoph Feck 2019-07-29 00:41:00 UTC 
https://github.com/KDAB/hotspot
Comment 4 Kevin Gilbert 2019-07-29 01:05:36 UTC 
OK - will install hotfix when hdd replaced. Obviously no updates till then (early next week, hopefully - I live in a remote area)
Comment 5 Kevin Gilbert 2019-08-03 23:40:00 UTC 
Well, back on the air and, my, wasn't it fun rebuilding your machine back up from bare metal. But to business:

'hotspot' appears to me to be just a front-end to 'perf' - and it (hotspot) crashed when I tried to run the recording phase. It did, however, give me enough to run:

# perf record -o perf.data --call-graph dwarf --aio --sample-cpu --mmap-pages 32M --pid 2201
^C[ perf record: Woken up 94 times to write data ]
[ perf record: Captured and wrote 1527.975 MB perf.data (189804 samples) ]

(Note that perf crashes when you specify 1G to the --mmap-pages parameter.)

That gave me 155GB of data. I then ran:

$ perf report -g --children -i perf.data

which produced a pretty output that _may_ be of use, see the attachment.

Cheers,
Kevin
Comment 6 Kevin Gilbert 2019-08-03 23:41:29 UTC 
Created attachment 121918 [details]
Output from perf report
Comment 7 Christoph Feck 2019-08-14 01:44:59 UTC 
New information was added with comment #5; changing status for inspection.
Comment 8 David Edmundson 2019-08-15 21:53:10 UTC 
Deadlock resizing some text.

We did have a report of this happening with notification popups that should be gone with Plasma 5.16

Please reopen if it's an issue.