AW: PPS stopped once => stopped forever?

[Wielant, Ulrich]

Hi Gary,

> > >Anything in dmesg after the failure?
> >
> > No, just:
> > [1279323.350188] pps pps0: removed
> > [1279473.387246] pps pps0: new PPS source serial0 [1279473.387273] pps
> > pps0: source "/dev/ttyS0" added
>
> Odd.  Must be another systemd bug.
>
> > >Note that SHM(0) and SHM(1) are root only, but your gpsd has already
> > >dropped root.  So gpsd can no longer open pps, SHM(0) or SHM(1).
> >
> > Ok, understood. Is there any way to change that?
>
> No.  But you can use SHM(2) and SHM(3).  That is not a good idea.

Sorry, I don't understand what you recommend.

>
> > I am using chrony. Does this problem disappear by using sockets
> > instead of SHM?
>
> Nope.  It is a design decision.
>

Ok.
This means gpsd will never forward the PPS signal to chrony/ntp if the GPS 
receiver is started after gpsd.
I have tested that:
- Start gpsd
- Wait some seconds
- Switch on power of GPS receiver
=> pps_monitor exits and time sync info is never forwarded to chrony/ntp.

Do you see any chance to handle such a case without regular external checks if 
gpsd is still providing time info to ntp/chrony and a restart of gpsd if the 
time info is no longer updated to SHM?
Would it be possible to implement a new variant so that gpsd does not exit 
pps_monitor on error so that it can continue providing PPS information after 
reconnect? This may be enabled by a new gpsd parameter to keep the old default 
behaviour.

Thx!
Ulrich
LEONARDO Germany GmbH
Sitz der Gesellschaft / Registered Office: Neuss
Registergericht / Register Court: Neuss HRB 17453
Geschäftsführer / Managing Director: Andrea Gaggelli