Boundary clock support for ptpd

[Wolfgang Grandegger]

Ronciak, John wrote:
> I'm not sure what you are looking for.  The Intel HW that can assist in timestamping 1588 packets is per port.  Which means it really can't be used across ports as for a boundary clock.  This has been brought up as a possible feature in future devices but it hasn't even been accepted for them at this time.

The timestamping hardware of the Gianfar ethernet controller on
Freescale processors like MPC8313/5 or P2020 use a *common* high
precision clock for both ethernet ports.

> Can you use a SW only solution for your boundary clock?  I think the PTPd could do that.

Yes, I was able to implement a boundary clock for daisy-chained MPC8313
boards by running:

  # ptpd-fsl -b eth0
  # ptpd-fsl -b eth1 -n "_ATL1" -p -t

eth0 connects the previous and eth1 the next board in the daisy chain.
With this setup, I measured +-150ns jitter on the second and +-300ns on
the third board relative to the time of the first board. Unfortunately,
the jitter gets doubled for each board added to the chain. To improve
that, I'm considering to switch to v2. Shorter sync intervals will help
and v2 would allow to implement a transparent clock allowing
point-to-point (or end-to-end) PTP synchronization.

I appreciate any comment or suggestion.

Wolfgang.

>> -----Original Message-----
>> From: ptpd-bounces at lists.infradead.org 
>> [mailto:ptpd-bounces at lists.infradead.org] On Behalf Of 
>> Wolfgang Grandegger
>> Sent: Friday, April 16, 2010 7:09 AM
>> To: gnn at neville-neil.com
>> Cc: ptpd at lists.infradead.org
>> Subject: Re: Boundary clock support for ptpd
>>
>> gnn wrote:
>>> On 4/15/10 15:54 , Wolfgang Grandegger wrote:
>>>> Hello,
>>>>
>>>> anybody listening?
>>>>
>>> Yes, just a bit slow.
>>>
>>>> Wolfgang Grandegger wrote:
>>>>> Hello,
>>>>>
>>>>> I'm currently evaluating ptpd-1.0.0 for IEEE1588 time 
>> syncronization
>>>>> with MPC8313 boards. I realized that there is an effort to 
>> support PTPv2
>>>>> (at 
>> http://ptpd.svn.sourceforge.net/viewvc/ptpd/branches/v2/). Could
>>>>> somebody give a brief status on what is already 
>> working/supported. I'm
>>>>> especially looking for boundary clock support (master and 
>> slave on the
>>>>> same board).
>>>> In the meantime I made some progress. I cascaded three 
>> MPC8313 boards
>>>> and measured jitters of +-150ns on the second and +-300ns 
>> on the third
>>>> board with ptpd *v1*. I will now try with v2 allowing for 
>> shorter sync
>>>> intervals, which should reduce the jitter significantly.
>>>>
>>>> I also found some interesting related link. Apart from the PTPDv2
>>>> implementation mentioned above, there is another one, 
>> especially for the
>>>> MPC8313, mentioned here:
>>>>
>>>>
>> http://forums.freescale.com/t5/Other-Microcontrollers/Open-sour
>> ce-for-IEEE-1588-PTP-v2-supporting-MPC831x-to-be/m-p/53829;jses
>> sionid=9DF51308EB4B68D57F509D77E77C8DD9
>>>> http://code.google.com/p/ptpv2d/
>>>>
>>>> As I see it, this one also lacks support for recent kernel versions
>>>> using the timestamping interface:
>>>>
>>>> http://lxr.linux.no/linux/Documentation/networking/timestamping.txt
>>>>
>>>> Is there already an effort supporting that generic 
>> interface in ptpd or
>>>> ptpdv2?
>>>
>>> I have not looked at that, but I will try to check that out.
>> It's used for PTP time stamping in a few netdev drivers:
>>
>> http://lxr.linux.no/#linux+v2.6.33/drivers/net/igb/igb_main.c
>> http://marc.info/?l=linux-netdev&m=126389931509102&w=4
>>
>> Wolfgang.
>>
>>
>> _______________________________________________
>> Ptpd mailing list
>> Ptpd at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/ptpd
>>
> _______________________________________________
> Ptpd mailing list
> Ptpd at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/ptpd
> 
>