[gpsd-users] Using gpio-based kernelpps with gpsd

[Robert Calhoun]

Gary, this is probably a question for you based on your post "RFC 2783"
to gpsd-dev:

http://lists.nongnu.org/archive/html/gpsd-dev/2015-03/msg00036.html

Is there a way to tell gpsd directly what pps device to use in gpsd 3.15?

I see some code and comments in ppsthread.c that suggest this is possible,
but I can't figure out how to do it. From ppsthread.c:
                   
                   
     * Some Linuxes, like the RasbPi's, have PPS devices preexisting.
                   
     * Other OS have no way to automatically determine the proper
/dev/ppsX.         
         
     * Allow user to pass in an explicit PPS device path.
                   
     */
    if (strncmp(pps_thread->devicename, "/dev/pps", 8) == 0)
    (void)strncpy(path, pps_thread->devicename, sizeof(path));
    else {...


But best as I can tell from how gpsmon calls it, pps_thread->devicename
is always going to be a serial device like /dev/ttyS0, not a pps device, so
we're never going to get a match on the code above. I have added the
path of my serial device to the .path property of the pps driver based
on the RFC 2783 discussion but I don't think that gets read anywhere in the
current code. Am I missing something or should I start modifying
ppsthread.c?

Thanks,

Rob Calhoun



More details if helpful:

I am trying to get gpsd + kernel pps working on a custom Cortex A8 TI SoC
based
board that has an uBlox NEO-6 chipset. I am cross-compiling gpsd using the
yocto
build system.

The uBlox' time pulse is wired into a gpio on the SoC. There is a gpio
driver 
(such as used with the Raspberry Pi) in newer kernels, but as we are using
2.6.34
where there is no such driver, I wrote my own. I might have goofed that
up, 
but it seems to be working fine. (Exception: I don't have a
/sys/devices/virtual
directory with my kernel, only /sys/class/pps/pps0 and /dev/pps0.)

/sys/class/pps/pps0 $ cat assert
1436805817.992079574#487
/sys/class/pps/pps0 $ cat assert
1436805818.992079574#488

While the kernel PPS feature seems to be working,  gpsd can't find it
because it
seems to assume the time pulse signal is always on the serial tty.

I've tried gpsd 3.14 based on the openembedded recipe:

http://cgit.openembedded.org/meta-openembedded/tree/meta-oe/recipes-navigat
ion/gpsd?h=master


and also 3.15 because there are a lot of kernel pps related changes going
from
gpsd 3.14 -> 3.15. I think I have gpsd (3.15) configured appropriately.
From
gpsd --help:

The following driver types are compiled into this gpsd instance:
    NMEA0183
    (...)
  * NMEA2000
  * JSON slave driver
    PPS
# n: mode switch, b: speed switch, c: rate switch, *: non-NMEA packet type.
# Socket export enabled.
# Shared memory export enabled.
# DBUS export enabled
# Time service features enabled.
# PPS enabled.


...but when I run gpsmon, init_kernel_pps() in ppsthread only looks at
/dev/ttyS3 for KPPS, which doesn't have it, so it returns failure.

$ gpsmon -D9 /dev/ttyS3 /dev/pps0 -l /var/log/gpsd
gpsmon:INFO: startup at 2015-07-13T22:35:49.000Z (1436826949)
gpsmon:INFO: opening GPS data source type 2 at '/dev/ttyS3'
gpsmon:INFO: speed 57600, 8N1
gpsmon:SPIN: open(/dev/ttyS3) -> 4 in gpsd_serial_open()
gpsmon:PROG: no probe matched...
gpsmon:INFO: gpsd_activate(1): activated GPS (fd 4)
gpsmon:INFO: KPPS:/dev/ttyS3 cannot set PPS line discipline Invalid
argument
gpsmon:WARN: KPPS:/dev/ttyS3 kernel PPS unavailable, PPS accuracy will
suffer
gpsmon:PROG: PPS:/dev/ttyS3 thread launched

I might be able to route the gpio inside the SoC so it looks like it is
associated
with /dev/ttyS3, but that seems like a hack given that /dev/pps0 works
fine as-is.


(side note for a different email to gpsd-dev: 3.15 makes fewer sysroot
assumptions
than 3.14 and requires a lot less patching to cross-compile--great!---but
the 
python-related code is still looking at the host's sysroot; I just
disabled python
via the scons options to get it to compile.)


p.s. Just 9 hours until New Horizons reaches Pluto!