Hello!
I've been using Powerlink for quite a while now, and, it seems to me, have encountered a very srange bug.
In a period of time, which depends on powerlink cycle, a lag occurs, which leads to powerlink cycle violation error. On high-performance systems this error may occur once in several days or even weeks of uptime with powerlink cycle ~1ms, but on low-performance systems it occurs quite often. I made a system performance trace during one of theese events, but still doubt what causes this error.
I suppose it is a tlb_flush error, which makes system flush whole tlb buffer, containing some valuable data for oplk-threads. But I'm not shure. If you have ever encountered same problems or know some workarrounds for this case - please, write it to me.
Cpu trace is made with kernelshark (Linux)
System specifications are : Fedora 20 Linux localhost.localdomain 4.4.1-rt6 #1 SMP PREEMPT RT Fri Apr 8 16:07:05 MSK 2016 x86_64 x86_64 x86_64 GNU/Linux; hardware specs are in lshw.html file.
Powelink cycle time is 2.5ms.
I have two questions regarding your issue.
1) Is it a requirement for your application to use a 4.4 kernel?
I am wondering because this is a fairly new kernel version and it could be that there are some timing issues or it's not very stable in general.
2) Could you please attach the traces to a SourceForge post?
Unfortunately I can't access the files your uploaded due to company proxy settings.
Best Regards,
Norman
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello!
I've been using Powerlink for quite a while now, and, it seems to me, have encountered a very srange bug.
In a period of time, which depends on powerlink cycle, a lag occurs, which leads to powerlink cycle violation error. On high-performance systems this error may occur once in several days or even weeks of uptime with powerlink cycle ~1ms, but on low-performance systems it occurs quite often. I made a system performance trace during one of theese events, but still doubt what causes this error.
I suppose it is a tlb_flush error, which makes system flush whole tlb buffer, containing some valuable data for oplk-threads. But I'm not shure. If you have ever encountered same problems or know some workarrounds for this case - please, write it to me.
Cpu trace is made with kernelshark (Linux)
System specifications are : Fedora 20 Linux localhost.localdomain 4.4.1-rt6 #1 SMP PREEMPT RT Fri Apr 8 16:07:05 MSK 2016 x86_64 x86_64 x86_64 GNU/Linux; hardware specs are in lshw.html file.
Powelink cycle time is 2.5ms.
Bogdan, MIREA
lshw:https://yadi.sk/d/zeCM7tI-rAxtR
trace:https://yadi.sk/d/q___HdDErAxst
Last edit: Bogdan Slaschov 2016-04-22
Hello Bogdan,
I have two questions regarding your issue.
1) Is it a requirement for your application to use a 4.4 kernel?
I am wondering because this is a fairly new kernel version and it could be that there are some timing issues or it's not very stable in general.
2) Could you please attach the traces to a SourceForge post?
Unfortunately I can't access the files your uploaded due to company proxy settings.
Best Regards,
Norman