Re: [chrony-dev] SW/HW timestamping on Linux

[Denny Page <dennypage@xxxxxx>]

I’ve pushed a new version of ethtscal that greatly improves the accuracy/stability of results. With runs of 50000 packets, I’m seeing consistency of +-5ns on loopback. If you pulled an older version, I recommend pulling a new one. 

Here is a sample run with i211 chips with loopback cable:

Iterating over 100000 packets (minimum runtime 200 seconds)

Tx interface igb0, speed 100Mb, adddress 0:8:a2:9:5b:8b

Rx interface igb1, speed 100Mb, adddress 0:8:a2:9:5b:8a

Tx avg ptp offset -200414826ns, 15 of 25 samples used

Rx avg ptp offset -96810099ns, 15 of 25 samples used

Misc times:

  Tx before send  ->  after send       8108ns

  Tx before send  ->  tx timestamp    15481ns

  Tx after send   ->  before poll    224578ns

  Tx after send   ->  after poll     231414ns

  Rx timestamp    ->  poll return     27296ns

Compensation values:

  Tx timestamp                            0ns

  Rx timestamp                            0ns

  Switch port to port                     0ns

  Cable length                            1ns  (0.3m)

  Total                                   1ns

Tx -> Rx timestamp:

  Uncompensated                        3175ns  (stddev 110ns)

  Compensated                          3175ns

Connection types:       Expected        Error

  Regular switch          6721ns      -3546ns

  Cut-through switch      2081ns       1094ns

  loopback cable             1ns       3174ns

Intel’s predicted result is 3177ns.

Denny