[ntp:hackers] processing a PPS siganl with no kernel support
loretta_1958 at yahoo.com
Fri Apr 24 20:22:59 UTC 2009
I have written a reference clock driver for a GPS currently being designed, i.e. I am simulating the time-of-day message and PPS it will provide.
I am doing this for a proprietary Unix-like OS, so I have no kernel support for the PPS, no Line Disciplines or Streams Modules (HAVE_TERMIOS is defined). My driver is being built for ntp-4.2.4p4.
I noticed the "Bugzilla Bug 577 Let ATOM driver be the sole-source of PPS support".
Since "HAVE_PPSAPI" is not defined for me, refclock_atom.c boils down to a start, shutdown, and poll function + the pps_sample() function that I need to call.
Am I correct, that this is a scenario where the "handling" (opening the I/O port, processing the interrupt, etc.) of the PPS needs to be in the driver handling the associated time-of-day signal? The handling will end with the call to pps_sample() for each computed offset?
More information about the hackers