[ntp:questions] Peridoc anomalies in loopstats file ...

unruh unruh at wormhole.physics.ubc.ca
Tue Feb 1 18:14:24 UTC 2011


On 2011-02-01, Martin Burnicki <martin.burnicki at meinberg.de> wrote:
> unruh wrote:
>> On 2011-01-27, Martin Burnicki <martin.burnicki at meinberg.de> wrote:
>>> Hal,
>>>
>>> Hal Murray wrote:
>>>> In article <g7do08-bma.ln1 at gateway.py.meinberg.de>,
>>>>  Heiko Gerstung <heiko.removethistext.gerstung at meinberg.de> writes:
>>>> 
>>>>> Mmmh, the polling cycle I use is 8s (patched ntpd), and if you look
>>>>> at this sample sequence ...:
>>>> 
>>>> Does the problem happen without that patch?
>>>
>>> The patch is only to allow minpoll 4 instead of minpoll 6.
>> 
>> 8 sec is poll 3, not poll 4.
>
> Of course, I know. Should have been thinking before I typed ... ;-)
>
>> Again, if you run ntp without the patch, does it work?
>> You may thing that "all" you changed was the minpoll stuff, but that
>> might be wrong-- you might have changed more, or your change might have
>> had unintended consequences. (Not that any of this should be unknown to
>> you).
>
> Seems to turn out this is a problem with a PPS signal supplied via a (real)
> serial port. Sometimes it simply takes too long until the PPS IRQ can be
> serviced, and thus the time stamp taken in the IRQ service routine is way
> too late.
>
> Looks like this is due to a hard disk driver module whose IRQ level (14) has
> higher priority than the serial port's IRQ level (4). Using a different IRQ
> (11) for the serial port, or using a different hard disk driver module
> seems to fix this problem.

I thought that the priority went from low to high numbers-- ie irq4 has
higher priority than 14. (That is why the timer interrupt is the lowest
number). It sounds to me like  you might have a hardware problem. 
And to imagine that an interrupt might take many milliseconds to be
serviced is just bizzare. 
I tend to like parallel port interrupts with a home built interrupt
handler better, but admit I have never done extensive timing tests to
back up my intuition. 

>
> We're still investigating.
>
> Martin




More information about the questions mailing list