On Wed, Sep 14, 2011 at 11:07 AM, Fran <[email protected]> wrote: > Chris, > > That is a good point, I hadn't thought of reference clocks case(s) and > interrupt level processing which will be at a higher priority than the > ntpd process. I was thinking all network connections between ntpd > processes. > > That then made me think about later stages of ntpd processing where > ntpd has decided to adjust the clock some amount. What if pre-emptions > happen in that segment of processing, where the adjustment value was > calculated but not actually applied yet ?
So what you say is that NTP calculates the error in the clock at some point in time. Then adjusts the clock at some other point in time because of preemption. I think in theory you are correct but task preemption will cause at most a 1/10 second delay and more likely closer to 1/100th second. The adjtime function works at uS resoltion. I doubt there would be a full microsecond difference in only 1/10th second. What we are talking about here is not the error in the clock but the rate of growth of the error in the clock. - Chris Albertson Redondo Beach, California _______________________________________________ questions mailing list [email protected] http://lists.ntp.org/listinfo/questions
