Re: NTP issue aarch64

[Gary E. Miller]

Yo Nick!

On Mon, 3 Jun 2024 09:48:49 +0100
Nick Taylor <nicktaylor@dataskill.uk> wrote:

> On 30/05/2024 20:34, Gary E. Miller wrote:
> > Oh, boy.  Not a lot of details there.  Please run gpsdenuginfo, as
> > root, ans share the results.  Get it from git head, or the web:
> >
> >      https://gpsd.io/gpsdebuginfo
> >
> >  
> Attaching output as gpsdebug1.log

Here is one big issue.  You are running gpsd 3.25.1~dev:

    + gpsd -V
    gpsd: 3.25.1~dev (revision release-3.25-571-gc12daabed)

But you have the Python gps module from 3.22:

    + python3 -c import gps; print(gps.__version__)
    3.22

Mixing client and server versions is not supported.  You need to remove
all remnants of older gpsd stuff.

But not your PPS problem.

Also, Python 3.9 has been unmaintainted for 2 years now:

    + python3 -V
    Python 3.9.2

    https://peps.python.org/pep-0596/

But not your PPS problem.

This is odd:

    795 ?        S<s    0:14 /usr/sbin/gpsd /dev/ttymxc3

You did NOT run gpsd with the same options for the gpsdebuginfo, and
for capturing the debug log.  So gpsdebuginfo is not showing me all that
I want to see from it.  Please try again, after fixing the issues already
identified.

> >      root@A14LKN:~# gpsd -D9 -N /dev/ttymxc5
> >
> > You forgot the -n option.  Without -n, gpsd waits for a client to
> > connect before reading from the GNSS receiver.  When gpsd is not
> >  
> Ran my debug again this time including -n
> 
> debug2.log.xz attached

Here is why gpsd is not connecting to chronyd with a socket:

    gpsd:PROG: NTP:/dev/ttymxc3 chrony socket /run/chrony.clk.ttymxc3.sock 
doesn't exist

But not your PPS problem.

It is opening SHM(0) and SHM(1):

    gpsd:PROG: NTP:SHM: ntpshm_alloc(/dev/ttymxc3), sourcetype 2 shm_clock 
using SHM(0)

    gpsd:PROG: NTP:SHM: ntpshm_alloc(/dev/ttymxc3), sourcetype 2 shm_pps using 
SHM(1)

Your gpsd built without KPPS support:

    gpsd:WARN: KPPS:/dev/ttymxc3 no HAVE_SYS_TIMEPPS_H, PPS accuracy will suffer

Why did you not build with KPPS?  That is part of your PPS problem.

Then this is the last PPS related message in the debug log:

    gpsd:PROG: PPS:/dev/ttymxc3 thread launched

The driver for /dev/ttymxc3 is doing something odd. Or rather, not doing
anything.  Does it support TIOCMIWAIT??

Can you please run as root:

        # ppscheck --seconds 10 /dev/ttymxc3


I suggest you rebuild with KPPS enabled.  That usually requires the
ppstools package, or whatever supplies /usr/include/sys/timepps.h for
your distro.

I do see a problem in gpsd:

gpsd:SPIN: CORE: parse_packet() = {NTPTIME|ONLINE|PACKET|TIME}
gpsd:DATA: CORE: gpsd_poll(/dev/ttymxc3) {NTPTIME|ONLINE|PACKET|TIME}
gpsd:DATA: CORE: packet type 11 from /dev/ttymxc3 with 
{NTPTIME|ONLINE|PACKET|TIME}
gpsd:DATA: all_reports(): changed {NTPTIME|ONLINE|PACKET|TIME}
gpsd:DATA: NTP: no fix

For some reason gpsd is not accepting the time from the u-blox binary for
sending to SHM<(0).  I'll have to go look at that.  Can you send me a
few seconds of the raw ubx binary when that happens?

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        gem@rellim.com  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin

pgpraDglGKN2W.pgp
Description: OpenPGP digital signature