Re: [Xenomai-help] Latencies due to RT-Socket-CAN register accesses

[Jan Kiszka <jan.kiszka@domain.hid>]

[-- Attachment #1: Type: text/plain, Size: 1797 bytes --]

roland Tollenaar wrote:
> Hi
> 
>> The latency killers:
>> - SMI: check dmesg to see if you get a warning saying that an
>>   SMI-enabled chipset was detected. If yes,
> Yes get this message.
> 
>> read the TROUBLESHOOTING. If
> Done what it says here. No change. My latency is also less than 100 us
> so this is probably not the problem. Tried anyway but as mentioned, no
> difference.
> 
>> - USB: disable legacy USB in BIOS configuration, sometimes it is not
> My BIOS settings do not allow me to switch off the USB legacy. At
> least I cannot find the place where this is to be done.
> 
> 
>>   enough, and you have to _enable_ USB in kernel configuration, and load
>>   the USB modules (that is at least uhci-hcd, ohci-hcd, ehci-hcd and
>>   usbcore) so that they disable BIOS legacy USB emulation.
> These ARE enabled.
> 
>> - X-window: if you see high latencies only when the X server is running,
>>   the X-server is probably causing them. A workaround is proposed in the
>>   TROUBLESHOOTING file.
> Ran latency without X running. No difference.
> 
>>
>> In case all that fails, enable the I-pipe tracer, run latency with the
>> -f option, and check /proc/ipipe/trace/frozen once you observed the high
>> latency in latency.
> Have done this and attached the result. I have looked at it myself but
> cannot interpret the meaning.
> 
> A bit of help would be appreciated.

Two conclusions:

 - You are running your kernel as i586 without TSC support - suboptimal,
   costs you a few micros.
 - The reported latency perfectly matches the trace, nothing
   pathological there. The trace looks like this: Timer fired,
   measurement task woken up, two interrupts squeeze themselves between
   wakeup and time stamp acquisition. All sane.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]