[LinuxPPS] linuxpps 2.6.23.1 report

gnu not unix gnu at wraith.sf.ca.us
Thu Oct 25 01:42:49 CEST 2007


Hi Folks--

I've installed linux 2.6.23.1 patched with the 
patch off the linuxpps web site. I also used the
nmea patch on ntp-dev-4.2.5p83 source.

The system sees both serial and parallel pps
sources. I am running ntpd with the nmea patch
and it does seem to be converging (very slowly).

The ppstest program picked up the parallel port 
transitions fine. Once there's an .ATOM. patch
I can try it on the parallel port. In the meantime
I'm using the shm driver with the parallel port
and ntpd.

Sample billboard:

root at thrall.wraith.sf.ca.us[1056] ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
*GPS_NMEA(1)     .GPS.            0 l   65   64  377    0.000  -30.382   0.602
 SHM(0)          .SHM.            0 l 238m   16    0    0.000  -39.577   0.000
+smidge.wraith.s .GPS.            1 u   57   64  377    0.238  -31.111   0.499
+ring.wraith.sf. .GPS.            1 u   55   64  377    0.125  -30.851   0.335
+wraith.wraith.s .SHM.            1 u   60   64  377    0.162  -31.111   0.474

The large offset causes the shm userland code to
report "out of range" and not deliver timestamps.
Once the offset works down some more the shm code
will begin to deliver timestamps again.

The slow, slow convergence is due to the linux
kernel having made major changes to the internal
time stuff while working towards HPET support.

It will be very interesting to watch the plots and
see how well the new setup works. If it is stable
I expect single digit microsecond performance.

Nice!

../Steven
ps. thrall system plots are at:

    http://www.wraith.sf.ca.us/ntp/rrd/index.html




More information about the LinuxPPS mailing list