Problem with GPSD refclock

Rich Wales richw at
Thu Jan 26 03:59:51 UTC 2017

Eric S. Raymond wrote:

> That looks like it ought to work, but my suspicion falls on the gpsd
> refclock driver.  It has never worked really right; that's a problem
> we inherited from Classic and are trying to fix.  We recomend using
> GPSD's SHM interface. (For this purpose you want to run it as root.)
> These lines work for me:
> # GPS Serial data reference (NTP0)
> refclock shm unit 0 refid GPS
> # GPS PPS reference (NTP1)
> refclock shm unit 1 refid PPS

Still no joy -- the refclocks are silently ignored -- I see no errors in
/var/log/syslog, but the refclocks are nowhere to be seen when I list
peers via ntpq.

If I go back to ntpd 4.2.8 and use the SHM refclock (, it
works fine.

And if I use the GPSD/JSON refclock ( in ntpd 4.2.8 --
specifying mode 1 (strict operation, producing clock samples only if PPS
is working) -- it also works fine for me, with no stability problems at
all as far as I can see.  Note, as I mentioned a little while ago, that
my GPS is attached to my system via a real RS232 serial port on the mobo
(not one of those flaky serial-over-USB add-on cards), so the PPS signal
from the GPS is being handled properly.

Rich Wales
richw at

More information about the users mailing list