[Linuxptp-users] Confused about one-step P2P OC versus EEE1588v2 spec

[Janusz U. <j.u...@el...>]

Hi.


In Linux kernel HWTSTAMP_TX_ONESTEP_P2P is introduced as feature 
superset of HWTSTAMP_TX_ONESTEP_SYNC for peer delay. However only few 
device drivers and its hardware support full P2P one-step timestamping.

It is not clear if one-step SYNC and two-step Pdelay method in P2P can 
be mixed together (by single OC/BC master) when HWTSTAMP_TX_ONESTEP_SYNC 
is supported only:

- According to IEEE1588-2008 specification (11.4.3) peer delay responder 
one-step clock "SHALL" (key word) update Pdelay_resp on the fly so it 
seems they can't be mixed.

- Moreover "one-step clock" term suggests it is clock property, not port 
timestamping ability. On the other hand it seems artifical requirement.

1. Can you clarify for one-step P2P what is met in the field (eg. PTP 
Plugfest) ?

2. And what is formal IEEE1588 requirement/statement ?

3. Is such P2P mix supported/handled properly by TC and OC/BC slaves? It 
looks supported (both OC slave and master) by linuxptp 4.0 and 3.1 
(port.c::process_pdelay_req()).

4. The latest gPTP and power profile have recommendation for one-step. 
Do they allow/mean one-step SYNC only or full P2P with Pdelay ?


best regards
Janusz