[time-nuts] Thunderbolt gpsd > ntpd failure.

[Scott Mace]

Here is what I did to get it to work:

http://www.febo.com/pipermail/time-nuts/2008-August/032726.html
http://www.febo.com/pipermail/time-nuts/2008-August/033106.html
http://www.febo.com/pipermail/time-nuts/2008-August/032803.html

I have a better patch for leap seconds that actually works correctly but 
have not had a chance to post it yet.

You need to make sure your kernel has the sysvshm option:
options         SYSVSHM                 # SYSV-style shared memory

then finally make sure that ntpd was compiled to support the SHM refclock.

I have a couple of thunderbolts working with net4501s using the TMR1IN
mod:  http://www.febo.com/pages/soekris/

I used a TAPR fatpps to stretch the pulse so the net4501 would be happy.

I start gpsd before ntpd

ntp.conf has:
server 127.127.28.0 minpoll 4 maxpoll 4 prefer
fudge 127.127.28.0 time1 -0.017 refid GPS

server 127.127.22.0 minpoll 4 maxpoll 4

tos mindist 0.050


	Scott

On 04/30/2009 08:50 PM, John Murphy wrote:
> I'm trying to configure ntpd (4.2.4p5-a FreeBSD-7.1) to use the
> shared memory driver to which gpsd should be writing timestamps.
> All looks good on the gpsd side, after building it for 8N1, as
> I can see what looks like ntpshm writing to shared memory:
>
> gpsd: TSIP pkt_id = 0x41, packetlen= 14
> gpsd: TSIP packet id 0x41 length 10: 48cf974905f941700000
> gpsd: ntpshm_put: Clock: 1241129131 @ 1241129132.166348
> gpsd: GPS Time 425146.281250 1529 15.000000
>
> But, with ntp.conf entries like:
>
> tos mindist 0.010
> server 127.127.28.0 minpoll 4 maxpoll 4
> fudge 127.127.28.0 time1 0.420 refid GPS
> server 127.127.22.0 minpoll 4 maxpoll 4
>
> The log shows:
>
> ntpd 4.2.4p5-a Thu Jan  1 09:59:32 UTC 2009 (1)
> refclock_newpeer: clock type 28 invalid
> configuration of 127.127.28.0 failed
>
> I'm also using the PPS output of the Thunderbolt, hence the
> 127.127.22.0 entry, and the 10MHz up converted to 33.333333MHz
> for the soekris net4501 CPU clock. Both of those working well.
>
> Do I need to initiate the shared memory some how? I feel I may
> have missed something vital, which may be obvious to someone.
>