[etherlab-users] skipped/unmatched datagrams - at startup - safe to ignore?

Matthieu Bec mbec at gmto.org
Tue Jun 26 23:55:24 CEST 2012


Yes I'm using RT_PREEMPT, but the ethercat layer of my application is 
implemented as a kernel module. It starts with a preset cycle time, the 
'workaround' lets me change it by ioctl.

I tried different bus topologies, number of terminals, different number 
of domains, mostly same results. Oddly I often get the same numbers 
reloading the kernel module (not quite always but it strucked me as 
weirdly recurring for a diagnostic printed every second with a 10kHz 
cycle): something like 24 UNMATCHED, 16 SKIPPED, usually happens twice 
and then goes away.

it's quite possible I have a regression in my kernel app layer 
somewhere. Will post once I find something. Thanks.


On 6/26/12 1:31 AM, Florian Pose wrote:
> Hash: SHA1
> Am 23.06.2012 01:26, schrieb Matthieu Bec:
>> I'm now able to get reliable clean starts (no more startup
>> warnings) by ramping up my cycle time in increments; like 1, 2, 4,
>> 8, 10 kHz over a short period.
>> Is anyone using similar "warm-up" technique?
> No. I think, you should check if your RT app is causing latencies at
> startup. Are you using RT-PREEMPT? Did you avoid causing pagefaults by
> locking the memory and cause an initial pagefault before cyclic operation?
> - --
> Best regards,
> Florian Pose
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
> iEYEARECAAYFAk/pc3kACgkQABFOIMygR8zwUACgoZeCkw9O26hjm0XlQ5Y+ztpf
> j2gAn2+ws0scMQBVLOyKg7es/ogBFLDU
> =XijT
> _______________________________________________
> etherlab-users mailing list
> etherlab-users at etherlab.org
> http://lists.etherlab.org/mailman/listinfo/etherlab-users

More information about the Etherlab-users mailing list