Subject: Re: ntp and pps
To: Andreas Wrede <firstname.lastname@example.org>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
Date: 05/12/2000 11:50:16
In message <Pine.NEB.email@example.com>,
Andreas Wrede writes:
>I am trying to use a GPS clock with ntp 4 in -current on a i386
>ntpd 4.0.99i Tue May 9 00:20:10 EDT 2000 (1)
>using kernel phase-lock loop 0001
>refclock_ioctl: time_pps_kcbind failed: Bad address
>refclock_open: fd 10 ioctl failed: Bad address
>configuration of 127.127.20.0 failed
the basic problem here is that the personw ho implemented the PPS code
(me) didnt have a true refclock to test...
Netbsd's PPS-API was tested with a test program, and I generated
pulses from another machine; but htey werent accurate enough to try
and turn on the KCBIND code. Looks like that is what's biting you.
For now, I suggest making sure that options PPS_SYNC gets into a
header file that really is included by sys/dev/ic/com.c, and adding
code to silenlty ignore the KCBIND call.
It should be a no-op on NetBSD, since the only consumer we have is
Dave Mills' hardpps() code.
I will fix this as soon as I get camera-ready copy done for
SIGCOMM2000. Since my personal circumstances have changed, i'll get a
GPS refclock, as well.