Re: [Linuxptp-users] Hardware PTP clock synchronization

[Richard C. <ric...@gm...>]

On Wed, Aug 07, 2013 at 09:43:45AM +0300, Alex Gavrilov wrote:
> 
> I use assembler code like this to read time registers:
> 
> .model large
> 
> test_data segment "FAR_DATA" public use16

You are running this program under DOS?

If so, why?

> ================================================================================
> 
> This is output from software mode:
> 
> [root@lab32 linuxptp-1.3]# ./ptp4l -i p16p1 -m -s -2 -S
                                        ^^^^^
Why does this interface name look so strange?

> ptp4l[3662.740]: port 1: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[3662.740]: port 0: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[3663.635]: port 1: new foreign master ece555.fffe.2de639-2
> ptp4l[3667.295]: selected best master clock ece555.fffe.2de639
> ptp4l[3667.295]: port 1: LISTENING to UNCALIBRATED on RS_SLAVE
> ptp4l[3667.433]: port 1: minimum delay request interval 2^3
> ptp4l[3674.614]: master offset -3600882843951 s0 freq      +0 path delay  19917808

Path delay is estimated at 20 milliseconds! Are your packets going
through a router?

> ptp4l[3675.529]: master offset -3600965414318 s0 freq      +0 path delay  19917808
> ptp4l[3676.444]: master offset -3601047983089 s0 freq      +0 path delay  19917808
> ptp4l[3677.359]: master offset -3601130555972 s0 freq      +0 path delay  19917808
> ptp4l[3678.274]: master offset -3601213123527 s0 freq      +0 path delay  19917808
> ptp4l[3679.189]: master offset -3601295694079 s0 freq      +0 path delay  19917808
> ptp4l[3680.104]: master offset -3601378264480 s0 freq      +0 path delay  19917808
> ptp4l[3681.019]: master offset -3601460837143 s0 freq      +0 path delay  19917808
> ptp4l[3681.934]: master offset -3601543405035 s0 freq      +0 path delay  19917808
> ptp4l[3682.849]: master offset -3601617688457 s0 freq      +0 path delay  11630558
> ptp4l[3683.889]: master offset -3601575495165 s0 freq      +0 path delay  11630558
> ptp4l[3684.970]: master offset -3601491859633 s0 freq      +0 path delay  11630558
> ptp4l[3686.051]: master offset -3601408225417 s0 freq      +0 path delay  11630558
> ptp4l[3687.132]: master offset -3601324587614 s0 freq      +0 path delay  11630558
> ptp4l[3688.213]: master offset -3601240952059 s0 freq      +0 path delay  11630558
> ptp4l[3689.294]: master offset -3601157315298 s0 freq      +0 path delay  11630558
> ptp4l[3690.376]: master offset -3601073679242 s0 freq      +0 path delay  11630558
> ptp4l[3691.457]: master offset -3600990048190 s1 freq -499999 path delay  11630558

The frequency offset is greater than 500 ppm. What is your system
clock? Can it really be so badly out of tune?

> ptp4l[3692.538]: master offset   83635426 s2 freq +499999 path delay  11630558
> ptp4l[3692.538]: port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED
> ptp4l[3693.618]: master offset  166278466 s2 freq +499999 path delay  11630558
> ptp4l[3694.698]: master offset  248915872 s2 freq +499999 path delay  11630558
> ptp4l[3695.778]: master offset  331556238 s2 freq +499999 path delay  11630558
> ptp4l[3696.858]: master offset  414192925 s2 freq +499999 path delay  11630558
> ptp4l[3697.939]: master offset  496831388 s2 freq +499999 path delay  11630558
> ptp4l[3699.019]: master offset  579467338 s2 freq +499999 path delay  11630558

The offset keeps growing, and the frequency adjustment is maxed
out. This means one (or more) of the following:

- The network delay and jitter are huge.
- The local time stamps are broken (can't think how this could
  happen with SW time stamps).
- The remote time stamps are broken (master is broken).
- The local oscillator is unusable.

Are you able to synchronize this system using ntp? radclock? ptpd?

Is this system running as a guest in a virtual machine?

Thanks,
Richard