In general, i used ftrace to figure out latency sources, below is what i did, you can give it a try
trace-cmd start -e irq -e sched -e timer -p function -l '*spin*'; cyclictest -p 90 -m -c 0 -n -i 100 -h 100 -q -l 1000000 -a -b 80 --tracemark
trace-cmd extract trace-cmd report -l > your_tracecmd_report.txt On 2020年06月29日 09:08, sunner.xu--- via Xenomai wrote:
Dear all, I have do something in xenomai3 latency program(linux4.9.24+xenomai3.0.5 and close all function that affect real-time). and found that it run about 41mins , a big overrun is happen as follows: 1) why the big overrun happen event cpu occupy is low(40%)? 2) how can solve the problem, and trace the big overrun? Thanks! -------------- next part -------------- A non-text attachment was scrubbed... Name: Catch635C.jpg Type: image/jpeg Size: 825532 bytes Desc: not available URL: <http://xenomai.org/pipermail/xenomai/attachments/20200629/2b8800f5/attachment.jpg>