Error "clock jumped backward or running slower than expected!"

[JohnRama]

• Environment LS1028ARDB

• SW

  • OpenIL
  • master branch

• network setup

  • Linux PC(Intel I210)
  • Summit X440-t8 (Avnu certified switch)
  • LS1028 ARDB

• Problem

  • How to duplicate

    1. configure system
       • Linux PC to grand master
       • LS1028 ARDB to slave
       • comamnd line $ptp4l -i eno0 -p /dev/ptp0 -f /etc/ptp4l_cfg/gPTP.cfg -m -l 7 |& tee tmp.log
       • By adjusting the priority1 parameter of BMCA.
    2. confirm time is synchronized by testptp under linux kernel tree.
       • $testptp -g clock time: 1500002055.106458152 or Thu Jul 13 23:14:15 2017
    3. modify the grand master's clock
       • $testptp -T 1595964364

  • Problem

  • LS1028 ARDB is not synchronized. (Following command result is not changed to 1595964364)

    • $testptp -g clock time: 1500002055.106458152 or Thu Jul 13 23:14:15 2017

  • Following message appeared (log attached) ptp4l[521866.541]: clockcheck: clock jumped backward or running slower than expected!

  • Problem never fixed until restarting ptp4l.

  Would you please help on this ? Thanks !!

John tmp.log

[vladimiroltean]

Hi John, Does it help if you add the 'step_threshold 0.00002' setting to the gPTP.cfg file? Thanks, -Vladimir

[JohnRama]

Great thanks, Vladmir

Does it help if you add the 'step_threshold 0.00002' setting to the gPTP.cfg file? This helped a lot !!

However, if I keep repeated to modifying the time between 1500002055 and 1595964364 many times, sometimes same problem happens (probability is about 1 time per 20 trial) with the same message. ptp4l[521866.541]: clockcheck: clock jumped backward or running slower than expected!

So looks like there is some corner case...

P.S This problem never happens with ptp4l (version2.0) running on my Linux PC(I210).

John

[vladimiroltean]

What is the maximum frequency adjustment in parts per billion for enetc, and what is it for i210?

./testptp -c -d /dev/ptp0
capabilities:
  999999999 maximum frequency adjustment (ppb)
  0 programmable alarms
  2 external time stamp channels
  0 programmable periodic signals
  1 pulse per second
  0 programmable pins
  0 cross timestamping
  0 adjust_phase

If the max freq adjustment for i210 is below the default sanity frequency limit value of 200000000, then that's your answer. So, with enetc, what happens if you configure max_frequency and sanity_freq_limit both to 999999999? It'll no longer complain, that's for sure, but the question is: will it work?

[JohnRama]

Hi Vladimir,

Thank you for your reply.

On 2020/07/29 17:25, Vladimir Oltean wrote:

What is the maximum frequency adjustment in parts per billion for enetc, and what is it for i210?

./testptp -c -d /dev/ptp0 capabilities: 999999999 maximum frequency adjustment (ppb) 0 programmable alarms 2 external time stamp channels 0 programmable periodic signals 1 pulse per second 0 programmable pins 0 cross timestamping 0 adjust_phase

LS1028ARDB

$./testptp -c capabilities: 999999999 maximum frequency adjustment (ppb) 0 programmable alarms 2 external time stamp channels 0 programmable periodic signals 1 pulse per second 0 programmable pins 0 cross timestamping

Linux PC (I210)

$testptp -c -d /dev/ptp1 capabilities: 62499999 maximum frequency adjustment (ppb) 0 programmable alarms 2 external time stamp channels 2 programmable periodic signals 1 pulse per second 4 programmable pins 0 cross timestamping

If the max freq adjustment for i210 is below the default sanity frequency limit value of 200000000, then that's your answer. Yes, the max freq adjustment for I210 is 62499999, which is below 200000000.

So, with enetc, what happens if you configure |max_frequency| and |sanity_freq_limit| both to 999999999? It'll no longer complain, that's for sure, but the question is: will it work? Setting the 999999999 to both max_frequency and sanity_freq_limit resulted in the same problem.

Using the same number with I210(max_frequency to 62499999 and sanity_freq_limit to 200000000) resulted in the problem fixed. (at least I never see the problem so far.) $testptp -g always return the synced time.

I'm curious what was the root cause of this problem(who was wrong anyway ?? Changing GM time so much is not allowed ??), so I read the man page of ptp for the parameter explanation, and tried to understand the code, but coding of frequency check part of linuxptp is little bit complicated to me to understand the mechanism in a short time.

So if you or someone could explain a little bit more, it's really appreciated. But at least problem was fixed. So thank you very much !!

John

[justdomyself]

hello, i change my param from 999999999 to 62499999 on my slave device, the erros message is missing, but the ptp can not sync time from master?