On Fri, Apr 17, 2020 at 03:32:09PM +0200, Kamil Alkhouri wrote:
> Additionally, I would like to raise the same discussion started earlier
> by Michael. Some hardware already supports two PTP clocks, one can be
> free running and used to transport 802.1AS synchronization while the
> other is a synchronized clock that can be used by TSN as an example. Do
> you think it is worth it to add this feature to linuxptp?
It depends on how. If I understand correctly, nothing changes in the
main PTP stack in ptp4l, but you do need a helper program to
discipline the second TSN clock. (There is one ready, see below)
> Of course, this will require extra efforts outside the linuxptp circle
> since we will need to define two PHCs per interface and pass
> cross-timestamps up through the kernel. Would you please give us your
> expert opinion on this?
cross-timestamps are only for system-clock to PHC use.
Neither of your clocks are the system-clock, so that won't work.
Instead, make both clocks PHC, and use the periodic-outputs/extts API
to generate time stamps from one clock to another.
If you do it that way, you get the servo program for free!
See our WIP branch that adds GM PPS and cross-PHC PPS synchronization
support:
https://github.com/richardcochran/linuxptp/tree/ts2phc-integration-11
The status of that branch is that I'm about to post it for review on
the mailing list.
Thanks,
Richard
_______________________________________________
Linuxptp-devel mailing list
Linuxptp-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
