Re: [Linuxptp-users] Time Jumps whole seconds only

[Jacob K. <jac...@in...>]

On 9/1/2020 9:48 AM, Mark Priest wrote:
> While characterizing the response of linuxptp to time jumps I discovered
> the following behavior. Both Master and slave are running linuxPTP but
> are different hardware architectures.
> When the time jump occurs (greater than the step_threshold on the
> master) the time adjusts on the slave via a jump but only the whole
> seconds are adjusted.  This results in the time being still off by the
> fractional amount (less than 1 second). This difference is then slewed
> to at the normal slew rate.  In my specific case (assuming random time
> jumps) this results in an average of .5s needing to be slewed.  Why
> isn't the time jumped to exactly the new time? Is there a way to change
> the behavior to jump to exactly the new time?  The time required to slew
> causes critical data to be considered unusable.
> 
> Example:
> *Before Jump:* Master time is 10:10:12.800 slave time is 10:10:10.300
> (Time difference is 2.5s)
> *After Jump:*    Master time is 10:10:12.800 slave time is 10:10:12.300
> (Time difference is 0.5s)
> 
> Thanks,
> Mark

What mode is ptp4l running in? HW? SW? If in hardware mode we'd need to
know the hardware and what driver it is using.

Thanks,
Jake