PTP can't handle large initial time differences

[andrewsg3]

v0.1.1 controller
v0.1.1-14-g8fe04dbe camera

Controllers have RTC batteries, modules do not. This means that after a long period offline the controller will have a very different (and more accurate) system clock than modules. In my experience, PTP cannot cope with large initial discrepancies in times - it will never catch up.

Think this is correlated with these faults in the logs

Dec 13 20:29:11 saviour-camera python[26563]: ERROR - src.modules.ptp - PTP error detected: Dec 13 20:28:49 saviour-camera ptp4l[296542]: [72670.725] port 1: UNCALIBRATED to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)

Indeed running check ready shows the offsets are incredibly high

Jan 05 16:34:28 saviour-camera python[26563]: INFO - modules.module - A check failed: _check_ptp, PTP not synchronized: offset 1972708326653114.0μs (max: 1000000.0μs)

Maybe there needs to be a manual, initial large sync?

[andrewsg3]

Could the controller act as an NTP server?

I'm currently getting around this bug by putting the module online and reactivating NTP (normally disabled due to conflicts with PTP).

If we could configure NTP for PTP coexistence, and have the controller act as NTP server (modules will normally be offline), NTP might be able to account for large changes while PTP acts on small ones.

[andrewsg3]

The problem seems to have been with the controller.

Although the controller was presenting as having correct system time ( $ date reported accurate time ) the phc2sys logs told a different story.

Jan 06 15:33:57 dev-controller phc2sys[1636]: [109471.411] eth0 sys offset -2035013609371772 s2 freq -64000000 delay 37
I restarted ptp4l and phc2sys on the controller, and then on the module. They are now syncing correctly.