https://bugs.kde.org/show_bug.cgi?id=424671

            Bug ID: 424671
           Summary: Wrong CPU usage reported after OOM kill
           Product: ksysguard
           Version: 5.19.3
          Platform: Other
                OS: Linux
            Status: REPORTED
          Severity: normal
          Priority: NOR
         Component: general
          Assignee: ksysguard-b...@kde.org
          Reporter: matheus.scattolinanse...@outlook.com
                CC: plasma-b...@kde.org
  Target Milestone: ---

Created attachment 130403
  --> https://bugs.kde.org/attachment.cgi?id=130403&action=edit
comparison between ksysguard and htop

SUMMARY
After an "out of memory" kill is triggered, ksysguard will report incorrect CPU
usage. It is happening without any userspace tool for OOM management, and I've
only managed to fix it by restarting the session, closing ksysguard and opening
it again will keep showing incorrect data.
"wrong" as in higher than reported in other tool, such as 100% in some of the
cores, all of them or just one, depending on what the program was doing before
stopping.
I'm using htop and gnome-system-monitor for comparison.

STEPS TO REPRODUCE
1. Open enough applications to fill up the memory. One single program filling
the entire memory works best, like a game.
2. Wait for the kernel to kill the application.
3. ksysguard will report wrong CPU usage info from now on.

OBSERVED RESULT
wronge CPU usage information

EXPECTED RESULT
CPU usage according to what is displayed on the 'Process Table' tab

SOFTWARE/OS VERSIONS
Windows: 
macOS: 
Linux/KDE Plasma: 5.7.10
(available in About System)
KDE Plasma Version: 5.19.3
KDE Frameworks Version: 5.72
Qt Version: 5.15

ADDITIONAL INFORMATION
the OOM kill logs the following, plus some stack trace:

jul 25 22:55:46 hades kernel: Out of memory: Killed process 2651 (Cities.x64)
total-vm:13314264kB, anon-rss:3085924kB, file-rss:0kB, shmem-rss:9716kB,
UID:1000 pgtables:8536kB oom_score_adj:0

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to