aboutsummaryrefslogtreecommitdiff
path: root/contrib/ntp/ntpd/refclock_jupiter.c
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/ntp/ntpd/refclock_jupiter.c')
-rw-r--r--contrib/ntp/ntpd/refclock_jupiter.c1262
1 files changed, 1262 insertions, 0 deletions
diff --git a/contrib/ntp/ntpd/refclock_jupiter.c b/contrib/ntp/ntpd/refclock_jupiter.c
new file mode 100644
index 000000000000..f22714bdf792
--- /dev/null
+++ b/contrib/ntp/ntpd/refclock_jupiter.c
@@ -0,0 +1,1262 @@
+/*
+ * Copyright (c) 1997, 1998
+ * The Regents of the University of California. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ * must display the following acknowledgement:
+ * This product includes software developed by the University of
+ * California, Lawrence Berkeley Laboratory.
+ * 4. The name of the University may not be used to endorse or promote
+ * products derived from this software without specific prior
+ * written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#ifdef HAVE_CONFIG_H
+# include <config.h>
+#endif
+
+#if defined(REFCLOCK) && defined(CLOCK_JUPITER) && defined(PPS)
+
+#include <stdio.h>
+#include <ctype.h>
+#include <sys/time.h>
+
+#include "ntpd.h"
+#include "ntp_io.h"
+#include "ntp_refclock.h"
+#include "ntp_unixtime.h"
+#include "ntp_stdlib.h"
+#include "ntp_calendar.h"
+
+#include "jupiter.h"
+
+#include <sys/ppsclock.h>
+
+#ifdef XNTP_BIG_ENDIAN
+#define getshort(s) ((((s) & 0xff) << 8) | (((s) >> 8) & 0xff))
+#define putshort(s) ((((s) & 0xff) << 8) | (((s) >> 8) & 0xff))
+#else
+#define getshort(s) (s)
+#define putshort(s) (s)
+#endif
+
+/* XXX */
+#ifdef sun
+char *strerror(int);
+#endif
+
+/*
+ * This driver supports the Rockwell Jupiter GPS Receiver board
+ * adapted to precision timing applications. It requires the
+ * ppsclock line discipline or streams module described in the
+ * Line Disciplines and Streams Drivers page. It also requires a
+ * gadget box and 1-PPS level converter, such as described in the
+ * Pulse-per-second (PPS) Signal Interfacing page.
+ *
+ * It may work (with minor modifications) with other Rockwell GPS
+ * receivers such as the CityTracker.
+ */
+
+/*
+ * GPS Definitions
+ */
+#define DEVICE "/dev/gps%d" /* device name and unit */
+#define SPEED232 B9600 /* baud */
+
+/*
+ * The number of raw samples which we acquire to derive a single estimate.
+ * NSAMPLES ideally should not exceed the default poll interval 64.
+ * NKEEP must be a power of 2 to simplify the averaging process.
+ */
+#define NSAMPLES 64
+#define NKEEP 8
+#define REFCLOCKMAXDISPERSE .25 /* max sample dispersion */
+
+/*
+ * Radio interface parameters
+ */
+#define PRECISION (-18) /* precision assumed (about 4 us) */
+#define REFID "GPS\0" /* reference id */
+#define DESCRIPTION "Rockwell Jupiter GPS Receiver" /* who we are */
+#define DEFFUDGETIME 0 /* default fudge time (ms) */
+
+/* Unix timestamp for the GPS epoch: January 6, 1980 */
+#define GPS_EPOCH 315964800
+
+/* Double short to unsigned int */
+#define DS2UI(p) ((getshort((p)[1]) << 16) | getshort((p)[0]))
+
+/* Double short to signed int */
+#define DS2I(p) ((getshort((p)[1]) << 16) | getshort((p)[0]))
+
+/* One week's worth of seconds */
+#define WEEKSECS (7 * 24 * 60 * 60)
+
+/*
+ * Jupiter unit control structure.
+ */
+struct jupiterunit {
+ u_int pollcnt; /* poll message counter */
+ u_int polled; /* Hand in a time sample? */
+ u_int lastserial; /* last pps serial number */
+ struct ppsclockev ppsev; /* PPS control structure */
+ u_int gweek; /* current GPS week number */
+ u_int32 lastsweek; /* last seconds into GPS week */
+ u_int32 timecode; /* current ntp timecode */
+ u_int32 stime; /* used to detect firmware bug */
+ int wantid; /* don't reconfig on channel id msg */
+ u_int moving; /* mobile platform? */
+ u_long sloppyclockflag; /* fudge flags */
+ u_int known; /* position known yet? */
+ int coderecv; /* total received samples */
+ int nkeep; /* number of samples to preserve */
+ int rshift; /* number of rshifts for division */
+ l_fp filter[NSAMPLES]; /* offset filter */
+ l_fp lastref; /* last reference timestamp */
+ u_short sbuf[512]; /* local input buffer */
+ int ssize; /* space used in sbuf */
+};
+
+/*
+ * Function prototypes
+ */
+static void jupiter_canmsg P((struct peer *, u_int));
+static u_short jupiter_cksum P((u_short *, u_int));
+#ifdef QSORT_USES_VOID_P
+ int jupiter_cmpl_fp P((const void *, const void *));
+#else
+ int jupiter_cmpl_fp P((const l_fp *, const l_fp *));
+#endif /* not QSORT_USES_VOID_P */
+static void jupiter_config P((struct peer *));
+static void jupiter_debug P((struct peer *, char *, ...))
+ __attribute__ ((format (printf, 2, 3)));
+static char * jupiter_offset P((struct peer *));
+static char * jupiter_parse_t P((struct peer *, u_short *));
+static void jupiter_platform P((struct peer *, u_int));
+static void jupiter_poll P((int, struct peer *));
+static int jupiter_pps P((struct peer *));
+static char * jupiter_process P((struct peer *));
+static int jupiter_recv P((struct peer *));
+static void jupiter_receive P((register struct recvbuf *rbufp));
+static void jupiter_reqmsg P((struct peer *, u_int, u_int));
+static void jupiter_reqonemsg P((struct peer *, u_int));
+static char * jupiter_send P((struct peer *, struct jheader *));
+static void jupiter_shutdown P((int, struct peer *));
+static int jupiter_start P((int, struct peer *));
+static int jupiter_ttyinit P((struct peer *, int));
+
+/*
+ * Transfer vector
+ */
+struct refclock refclock_jupiter = {
+ jupiter_start, /* start up driver */
+ jupiter_shutdown, /* shut down driver */
+ jupiter_poll, /* transmit poll message */
+ noentry, /* (clock control) */
+ noentry, /* (clock init) */
+ noentry, /* (clock buginfo) */
+ NOFLAGS /* not used */
+};
+
+/*
+ * jupiter_start - open the devices and initialize data for processing
+ */
+static int
+jupiter_start(
+ register int unit,
+ register struct peer *peer
+ )
+{
+ struct refclockproc *pp;
+ register struct jupiterunit *up;
+ register int fd;
+ char gpsdev[20];
+
+ /*
+ * Open serial port
+ */
+ (void)sprintf(gpsdev, DEVICE, unit);
+ fd = open(gpsdev, O_RDWR
+#ifdef O_NONBLOCK
+ | O_NONBLOCK
+#endif
+ , 0);
+ if (fd < 0) {
+ jupiter_debug(peer, "jupiter_start: open %s: %s\n",
+ gpsdev, strerror(errno));
+ return (0);
+ }
+ if (!jupiter_ttyinit(peer, fd))
+ return (0);
+
+ /* Allocate unit structure */
+ if ((up = (struct jupiterunit *)
+ emalloc(sizeof(struct jupiterunit))) == NULL) {
+ (void) close(fd);
+ return (0);
+ }
+ memset((char *)up, 0, sizeof(struct jupiterunit));
+ pp = peer->procptr;
+ pp->io.clock_recv = jupiter_receive;
+ pp->io.srcclock = (caddr_t)peer;
+ pp->io.datalen = 0;
+ pp->io.fd = fd;
+ if (!io_addclock(&pp->io)) {
+ (void) close(fd);
+ free(up);
+ return (0);
+ }
+ pp->unitptr = (caddr_t)up;
+
+ /*
+ * Initialize miscellaneous variables
+ */
+ peer->precision = PRECISION;
+ pp->clockdesc = DESCRIPTION;
+ memcpy((char *)&pp->refid, REFID, 4);
+
+
+ /* Ensure the receiver is properly configured */
+ jupiter_config(peer);
+
+ /* Turn on pulse gathering by requesting the first sample */
+ if (ioctl(fd, CIOGETEV, (caddr_t)&up->ppsev) < 0) {
+ jupiter_debug(peer, "jupiter_ttyinit: CIOGETEV: %s\n",
+ strerror(errno));
+ (void) close(fd);
+ free(up);
+ return (0);
+ }
+ up->lastserial = up->ppsev.serial;
+ memset(&up->ppsev, 0, sizeof(up->ppsev));
+ return (1);
+}
+
+/*
+ * jupiter_shutdown - shut down the clock
+ */
+static void
+jupiter_shutdown(register int unit, register struct peer *peer)
+{
+ register struct jupiterunit *up;
+ struct refclockproc *pp;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+ io_closeclock(&pp->io);
+ free(up);
+}
+
+/*
+ * jupiter_config - Configure the receiver
+ */
+static void
+jupiter_config(register struct peer *peer)
+{
+ register int i;
+ register struct jupiterunit *up;
+ register struct refclockproc *pp;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * Initialize the unit variables
+ *
+ * STRANGE BEHAVIOUR WARNING: The fudge flags are not available
+ * at the time jupiter_start is called. These are set later,
+ * and so the code must be prepared to handle changing flags.
+ */
+ up->sloppyclockflag = pp->sloppyclockflag;
+ if (pp->sloppyclockflag & CLK_FLAG2) {
+ up->moving = 1; /* Receiver on mobile platform */
+ msyslog(LOG_DEBUG, "jupiter_config: mobile platform");
+ } else {
+ up->moving = 0; /* Static Installation */
+ }
+
+ /* XXX fludge flags don't make the trip from the config to here... */
+#ifdef notdef
+ /* Configure for trailing edge triggers */
+#ifdef CIOSETTET
+ i = ((pp->sloppyclockflag & CLK_FLAG3) != 0);
+ jupiter_debug(peer, "jupiter_configure: (sloppyclockflag 0x%lx)\n",
+ pp->sloppyclockflag);
+ if (ioctl(pp->io.fd, CIOSETTET, (char *)&i) < 0)
+ msyslog(LOG_DEBUG, "jupiter_configure: CIOSETTET %d: %m", i);
+#else
+ if (pp->sloppyclockflag & CLK_FLAG3)
+ msyslog(LOG_DEBUG, "jupiter_configure: \
+No kernel support for trailing edge trigger");
+#endif
+#endif
+
+ up->pollcnt = 2;
+ up->polled = 0;
+ up->known = 0;
+ up->gweek = 0;
+ up->lastsweek = 2 * WEEKSECS;
+ up->timecode = 0;
+ up->stime = 0;
+ up->ssize = 0;
+ up->coderecv = 0;
+ up->nkeep = NKEEP;
+ if (up->nkeep > NSAMPLES)
+ up->nkeep = NSAMPLES;
+ if (up->nkeep >= 1)
+ up->rshift = 0;
+ if (up->nkeep >= 2)
+ up->rshift = 1;
+ if (up->nkeep >= 4)
+ up->rshift = 2;
+ if (up->nkeep >= 8)
+ up->rshift = 3;
+ if (up->nkeep >= 16)
+ up->rshift = 4;
+ if (up->nkeep >= 32)
+ up->rshift = 5;
+ if (up->nkeep >= 64)
+ up->rshift = 6;
+ up->nkeep = 1;
+ i = up->rshift;
+ while (i > 0) {
+ up->nkeep *= 2;
+ i--;
+ }
+
+ /* Stop outputting all messages */
+ jupiter_canmsg(peer, JUPITER_ALL);
+
+ /* Request the receiver id so we can syslog the firmware version */
+ jupiter_reqonemsg(peer, JUPITER_O_ID);
+
+ /* Flag that this the id was requested (so we don't get called again) */
+ up->wantid = 1;
+
+ /* Request perodic time mark pulse messages */
+ jupiter_reqmsg(peer, JUPITER_O_PULSE, 1);
+
+ /* Set application platform type */
+ if (up->moving)
+ jupiter_platform(peer, JUPITER_I_PLAT_MED);
+ else
+ jupiter_platform(peer, JUPITER_I_PLAT_LOW);
+}
+
+/*
+ * jupiter_poll - jupiter watchdog routine
+ */
+static void
+jupiter_poll(register int unit, register struct peer *peer)
+{
+ register struct jupiterunit *up;
+ register struct refclockproc *pp;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * You don't need to poll this clock. It puts out timecodes
+ * once per second. If asked for a timestamp, take note.
+ * The next time a timecode comes in, it will be fed back.
+ */
+
+ /*
+ * If we haven't had a response in a while, reset the receiver.
+ */
+ if (up->pollcnt > 0) {
+ up->pollcnt--;
+ } else {
+ refclock_report(peer, CEVNT_TIMEOUT);
+
+ /* Request the receiver id to trigger a reconfig */
+ jupiter_reqonemsg(peer, JUPITER_O_ID);
+ up->wantid = 0;
+ }
+
+ /*
+ * polled every 64 seconds. Ask jupiter_receive to hand in
+ * a timestamp.
+ */
+ up->polled = 1;
+ pp->polls++;
+}
+
+/*
+ * jupiter_receive - receive gps data
+ * Gag me!
+ */
+static void
+jupiter_receive(register struct recvbuf *rbufp)
+{
+ register int bpcnt, cc, size, ppsret;
+ register u_int32 last_timecode, laststime;
+ register char *cp;
+ register u_char *bp;
+ register u_short *sp;
+ register u_long sloppyclockflag;
+ register struct jupiterunit *up;
+ register struct jid *ip;
+ register struct jheader *hp;
+ register struct refclockproc *pp;
+ register struct peer *peer;
+
+ /* Initialize pointers and read the timecode and timestamp */
+ peer = (struct peer *)rbufp->recv_srcclock;
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * If operating mode has been changed, then reinitialize the receiver
+ * before doing anything else.
+ */
+/* XXX Sloppy clock flags are broken!! */
+ sloppyclockflag = up->sloppyclockflag;
+ up->sloppyclockflag = pp->sloppyclockflag;
+ if ((pp->sloppyclockflag & CLK_FLAG2) !=
+ (sloppyclockflag & CLK_FLAG2)) {
+ jupiter_debug(peer,
+ "jupiter_receive: mode switch: reset receiver\n");
+ jupiter_config(peer);
+ return;
+ }
+
+ up->pollcnt = 2;
+
+ bp = (u_char *)rbufp->recv_buffer;
+ bpcnt = rbufp->recv_length;
+
+ /* This shouldn't happen */
+ if (bpcnt > sizeof(up->sbuf) - up->ssize)
+ bpcnt = sizeof(up->sbuf) - up->ssize;
+
+ /* Append to input buffer */
+ memcpy((u_char *)up->sbuf + up->ssize, bp, bpcnt);
+ up->ssize += bpcnt;
+
+ /* While there's at least a header and we parse a intact message */
+ while (up->ssize > sizeof(*hp) && (cc = jupiter_recv(peer)) > 0) {
+ hp = (struct jheader *)up->sbuf;
+ sp = (u_short *)(hp + 1);
+ size = cc - sizeof(*hp);
+ switch (getshort(hp->id)) {
+
+ case JUPITER_O_PULSE:
+ if (size != sizeof(struct jpulse)) {
+ jupiter_debug(peer,
+ "jupiter_receive: pulse: len %d != %u\n",
+ size, (int)sizeof(struct jpulse));
+ refclock_report(peer, CEVNT_BADREPLY);
+ break;
+ }
+
+ /*
+ * There appears to be a firmware bug related
+ * to the pulse message; in addition to the one
+ * per second messages, we get an extra pulse
+ * message once an hour (on the anniversary of
+ * the cold start). It seems to come 200 ms
+ * after the one requested. So if we've seen a
+ * pulse message in the last 210 ms, we skip
+ * this one.
+ */
+ laststime = up->stime;
+ up->stime = DS2UI(((struct jpulse *)sp)->stime);
+ if (laststime != 0 && up->stime - laststime <= 21) {
+ jupiter_debug(peer, "jupiter_receive: \
+avoided firmware bug (stime %.2f, laststime %.2f)\n",
+ (double)up->stime * 0.01, (double)laststime * 0.01);
+ break;
+ }
+
+ /* Retrieve pps timestamp */
+ ppsret = jupiter_pps(peer);
+
+ /* Parse timecode (even when there's no pps) */
+ last_timecode = up->timecode;
+ if ((cp = jupiter_parse_t(peer, sp)) != NULL) {
+ jupiter_debug(peer,
+ "jupiter_receive: pulse: %s\n", cp);
+ break;
+ }
+
+ /* Bail if we didn't get a pps timestamp */
+ if (ppsret)
+ break;
+
+ /* Bail if we don't have the last timecode yet */
+ if (last_timecode == 0)
+ break;
+
+ /* Add the new sample to a median filter */
+ if ((cp = jupiter_offset(peer)) != NULL) {
+ jupiter_debug(peer,
+ "jupiter_receive: offset: %s\n", cp);
+ refclock_report(peer, CEVNT_BADTIME);
+ break;
+ }
+
+ /*
+ * The clock will blurt a timecode every second
+ * but we only want one when polled. If we
+ * havn't been polled, bail out.
+ */
+ if (!up->polled)
+ break;
+
+ /*
+ * It's a live one! Remember this time.
+ */
+ pp->lasttime = current_time;
+
+ /*
+ * Determine the reference clock offset and
+ * dispersion. NKEEP of NSAMPLE offsets are
+ * passed through a median filter.
+ * Save the (filtered) offset and dispersion in
+ * pp->offset and pp->disp.
+ */
+ if ((cp = jupiter_process(peer)) != NULL) {
+ jupiter_debug(peer,
+ "jupiter_receive: process: %s\n", cp);
+ refclock_report(peer, CEVNT_BADTIME);
+ break;
+ }
+ /*
+ * Return offset and dispersion to control
+ * module. We use lastrec as both the reference
+ * time and receive time in order to avoid
+ * being cute, like setting the reference time
+ * later than the receive time, which may cause
+ * a paranoid protocol module to chuck out the
+ * data.
+ */
+ jupiter_debug(peer,
+ "jupiter_receive: process time: \
+%4d-%03d %02d:%02d:%02d at %s, %s\n",
+ pp->year, pp->day,
+ pp->hour, pp->minute, pp->second,
+ prettydate(&pp->lastrec), lfptoa(&pp->offset, 6));
+
+ refclock_receive(peer);
+
+ /*
+ * We have succeeded in answering the poll.
+ * Turn off the flag and return
+ */
+ up->polled = 0;
+ break;
+
+ case JUPITER_O_ID:
+ if (size != sizeof(struct jid)) {
+ jupiter_debug(peer,
+ "jupiter_receive: id: len %d != %u\n",
+ size, (int)sizeof(struct jid));
+ refclock_report(peer, CEVNT_BADREPLY);
+ break;
+ }
+ /*
+ * If we got this message because the Jupiter
+ * just powered up, it needs to be reconfigured.
+ */
+ ip = (struct jid *)sp;
+ jupiter_debug(peer,
+ "jupiter_receive: >> %s chan ver %s, %s (%s)\n",
+ ip->chans, ip->vers, ip->date, ip->opts);
+ msyslog(LOG_DEBUG,
+ "jupiter_receive: %s chan ver %s, %s (%s)\n",
+ ip->chans, ip->vers, ip->date, ip->opts);
+ if (up->wantid)
+ up->wantid = 0;
+ else {
+ jupiter_debug(peer,
+ "jupiter_receive: reset receiver\n");
+ jupiter_config(peer);
+ /* Rese since jupiter_config() just zeroed it */
+ up->ssize = cc;
+ }
+ break;
+
+ default:
+ jupiter_debug(peer,
+ "jupiter_receive: >> unknown message id %d\n",
+ getshort(hp->id));
+ break;
+ }
+ up->ssize -= cc;
+ if (up->ssize < 0) {
+ fprintf(stderr, "jupiter_recv: negative ssize!\n");
+ abort();
+ } else if (up->ssize > 0)
+ memcpy(up->sbuf, (u_char *)up->sbuf + cc, up->ssize);
+ }
+ record_clock_stats(&peer->srcadr, "<timecode is binary>");
+}
+
+/*
+ * jupiter_offset - Calculate the offset, and add to the rolling filter.
+ */
+static char *
+jupiter_offset(register struct peer *peer)
+{
+ register struct jupiterunit *up;
+ register struct refclockproc *pp;
+ register int i;
+ l_fp offset;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * Calculate the offset
+ */
+ if (!clocktime(pp->day, pp->hour, pp->minute, pp->second, GMT,
+ pp->lastrec.l_ui, &pp->yearstart, &offset.l_ui)) {
+ return ("jupiter_process: clocktime failed");
+ }
+ if (pp->usec) {
+ TVUTOTSF(pp->usec, offset.l_uf);
+ } else {
+ MSUTOTSF(pp->msec, offset.l_uf);
+ }
+ L_ADD(&offset, &pp->fudgetime1);
+ up->lastref = offset; /* save last reference time */
+ L_SUB(&offset, &pp->lastrec); /* form true offset */
+
+ /*
+ * A rolling filter. Initialize first time around.
+ */
+ i = ((up->coderecv)) % NSAMPLES;
+
+ up->filter[i] = offset;
+ if (up->coderecv == 0)
+ for (i = 1; (u_int) i < NSAMPLES; i++)
+ up->filter[i] = up->filter[0];
+ up->coderecv++;
+
+ return (NULL);
+}
+
+/*
+ * jupiter_process - process the sample from the clock,
+ * passing it through a median filter and optionally averaging
+ * the samples. Returns offset and dispersion in "up" structure.
+ */
+static char *
+jupiter_process(register struct peer *peer)
+{
+ register struct jupiterunit *up;
+ register struct refclockproc *pp;
+ register int i, n;
+ register int j, k;
+ l_fp offset, median, lftmp;
+ u_fp disp;
+ l_fp off[NSAMPLES];
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * Copy the raw offsets and sort into ascending order
+ */
+ for (i = 0; i < NSAMPLES; i++)
+ off[i] = up->filter[i];
+ qsort((char *)off, NSAMPLES, sizeof(l_fp), jupiter_cmpl_fp);
+
+ /*
+ * Reject the furthest from the median of NSAMPLES samples until
+ * NKEEP samples remain.
+ */
+ i = 0;
+ n = NSAMPLES;
+ while ((n - i) > up->nkeep) {
+ lftmp = off[n - 1];
+ median = off[(n + i) / 2];
+ L_SUB(&lftmp, &median);
+ L_SUB(&median, &off[i]);
+ if (L_ISHIS(&median, &lftmp)) {
+ /* reject low end */
+ i++;
+ } else {
+ /* reject high end */
+ n--;
+ }
+ }
+
+ /*
+ * Copy key values to the billboard to measure performance.
+ */
+ pp->lastref = up->lastref;
+ pp->coderecv = up->coderecv;
+ pp->filter[0] = off[0]; /* smallest offset */
+ pp->filter[1] = off[NSAMPLES-1]; /* largest offset */
+ for (j = 2, k = i; k < n; j++, k++)
+ pp->filter[j] = off[k]; /* offsets actually examined */
+
+ /*
+ * Compute the dispersion based on the difference between the
+ * extremes of the remaining offsets. Add to this the time since
+ * the last clock update, which represents the dispersion
+ * increase with time. We know that NTP_MAXSKEW is 16. If the
+ * sum is greater than the allowed sample dispersion, bail out.
+ * If the loop is unlocked, return the most recent offset;
+ * otherwise, return the median offset.
+ */
+ lftmp = off[n - 1];
+ L_SUB(&lftmp, &off[i]);
+ disp = LFPTOFP(&lftmp);
+ if (disp > REFCLOCKMAXDISPERSE)
+ return ("Maximum dispersion exceeded");
+
+ /*
+ * Now compute the offset estimate. If fudge flag 1
+ * is set, average the remainder, otherwise pick the
+ * median.
+ */
+ if (pp->sloppyclockflag & CLK_FLAG1) {
+ L_CLR(&lftmp);
+ while (i < n) {
+ L_ADD(&lftmp, &off[i]);
+ i++;
+ }
+ i = up->rshift;
+ while (i > 0) {
+ L_RSHIFT(&lftmp);
+ i--;
+ }
+ offset = lftmp;
+ } else {
+ i = (n + i) / 2;
+ offset = off[i];
+ }
+
+ /*
+ * The payload: filtered offset and dispersion.
+ */
+
+ pp->offset = offset;
+ pp->disp = disp;
+
+ return (NULL);
+
+}
+
+/* Compare two l_fp's, used with qsort() */
+int
+#ifdef QSORT_USES_VOID_P
+jupiter_cmpl_fp(register const void *p1, register const void *p2)
+#else
+jupiter_cmpl_fp(register const l_fp *fp1, register const l_fp *fp2)
+#endif
+{
+#ifdef QSORT_USES_VOID_P
+ register const l_fp *fp1 = (const l_fp *)p1;
+ register const l_fp *fp2 = (const l_fp *)p2;
+#endif
+
+ if (!L_ISGEQ(fp1, fp2))
+ return (-1);
+ if (L_ISEQU(fp1, fp2))
+ return (0);
+ return (1);
+}
+
+static char *
+jupiter_parse_t(register struct peer *peer, register u_short *sp)
+{
+ register struct refclockproc *pp;
+ register struct jupiterunit *up;
+ register struct tm *tm;
+ register char *cp;
+ register struct jpulse *jp;
+ register struct calendar *jt;
+ register u_int32 sweek;
+ register u_int32 last_timecode;
+ register u_short flags;
+ time_t t;
+ struct calendar cal;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+ jp = (struct jpulse *)sp;
+
+ /* The timecode is presented as seconds into the current GPS week */
+ sweek = DS2UI(jp->sweek);
+
+ /*
+ * If we don't know the current GPS week, calculate it from the
+ * current time. (It's too bad they didn't include this
+ * important value in the pulse message). We'd like to pick it
+ * up from one of the other messages like gpos or chan but they
+ * don't appear to be synchronous with time keeping and changes
+ * too soon (something like 10 seconds before the new GPS
+ * week).
+ *
+ * If we already know the current GPS week, increment it when
+ * we wrap into a new week.
+ */
+ if (up->gweek == 0)
+ up->gweek = (time(NULL) - GPS_EPOCH) / WEEKSECS;
+ else if (sweek == 0 && up->lastsweek == WEEKSECS - 1) {
+ ++up->gweek;
+ jupiter_debug(peer,
+ "jupiter_parse_t: NEW gps week %u\n", up->gweek);
+ }
+
+ /*
+ * See if the sweek stayed the same (this happens when there is
+ * no pps pulse).
+ *
+ * Otherwise, look for time warps:
+ *
+ * - we have stored at least one lastsweek and
+ * - the sweek didn't increase by one and
+ * - we didn't wrap to a new GPS week
+ *
+ * Then we warped.
+ */
+ if (up->lastsweek == sweek)
+ jupiter_debug(peer,
+ "jupiter_parse_t: gps sweek not incrementing (%d)\n",
+ sweek);
+ else if (up->lastsweek != 2 * WEEKSECS &&
+ up->lastsweek + 1 != sweek &&
+ !(sweek == 0 && up->lastsweek == WEEKSECS - 1))
+ jupiter_debug(peer,
+ "jupiter_parse_t: gps sweek jumped (was %d, now %d)\n",
+ up->lastsweek, sweek);
+ up->lastsweek = sweek;
+
+ /* This timecode describes next pulse */
+ last_timecode = up->timecode;
+ up->timecode = (u_int32)JAN_1970 +
+ GPS_EPOCH + (up->gweek * WEEKSECS) + sweek;
+
+ if (last_timecode == 0)
+ /* XXX debugging */
+ jupiter_debug(peer,
+ "jupiter_parse_t: UTC <none> (gweek/sweek %u/%u)\n",
+ up->gweek, sweek);
+ else {
+ /* XXX debugging */
+ t = last_timecode - (u_int32)JAN_1970;
+ tm = gmtime(&t);
+ cp = asctime(tm);
+
+ jupiter_debug(peer,
+ "jupiter_parse_t: UTC %.24s (gweek/sweek %u/%u)\n",
+ cp, up->gweek, sweek);
+
+ /* Billboard last_timecode (which is now the current time) */
+ jt = &cal;
+ caljulian(last_timecode, jt);
+ pp = peer->procptr;
+ pp->year = jt->year;
+ pp->day = jt->yearday;
+ pp->hour = jt->hour;
+ pp->minute = jt->minute;
+ pp->second = jt->second;
+ pp->msec = 0;
+ pp->usec = 0;
+ }
+
+ /* XXX debugging */
+ tm = gmtime(&up->ppsev.tv.tv_sec);
+ cp = asctime(tm);
+ flags = getshort(jp->flags);
+ jupiter_debug(peer,
+ "jupiter_parse_t: PPS %.19s.%06lu %.4s (serial %u)%s\n",
+ cp, up->ppsev.tv.tv_usec, cp + 20, up->ppsev.serial,
+ (flags & JUPITER_O_PULSE_VALID) == 0 ?
+ " NOT VALID" : "");
+
+ /* Toss if not designated "valid" by the gps */
+ if ((flags & JUPITER_O_PULSE_VALID) == 0) {
+ refclock_report(peer, CEVNT_BADTIME);
+ return ("time mark not valid");
+ }
+
+ /* We better be sync'ed to UTC... */
+ if ((flags & JUPITER_O_PULSE_UTC) == 0) {
+ refclock_report(peer, CEVNT_BADTIME);
+ return ("time mark not sync'ed to UTC");
+ }
+
+ return (NULL);
+}
+
+/*
+ * Process a PPS signal, returning a timestamp.
+ */
+static int
+jupiter_pps(register struct peer *peer)
+{
+ register struct refclockproc *pp;
+ register struct jupiterunit *up;
+ register int firsttime;
+ struct timeval ntp_tv;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /*
+ * Grab the timestamp of the PPS signal.
+ */
+ firsttime = (up->ppsev.tv.tv_sec == 0);
+ if (ioctl(pp->io.fd, CIOGETEV, (caddr_t)&up->ppsev) < 0) {
+ /* XXX Actually, if this fails, we're pretty much screwed */
+ jupiter_debug(peer, "jupiter_pps: CIOGETEV: %s\n",
+ strerror(errno));
+ refclock_report(peer, CEVNT_FAULT);
+ return (1);
+ }
+
+ /*
+ * Check pps serial number against last one
+ */
+ if (!firsttime && up->lastserial + 1 != up->ppsev.serial) {
+ if (up->ppsev.serial == up->lastserial)
+ jupiter_debug(peer, "jupiter_pps: no new pps event\n");
+ else
+ jupiter_debug(peer,
+ "jupiter_pps: missed %d pps events\n",
+ up->ppsev.serial - up->lastserial - 1);
+ up->lastserial = up->ppsev.serial;
+ refclock_report(peer, CEVNT_FAULT);
+ return (1);
+ }
+ up->lastserial = up->ppsev.serial;
+
+ /*
+ * Return the timestamp in pp->lastrec
+ */
+ ntp_tv = up->ppsev.tv;
+ ntp_tv.tv_sec += (u_int32)JAN_1970;
+ TVTOTS(&ntp_tv, &pp->lastrec);
+
+ return (0);
+}
+
+/*
+ * jupiter_debug - print debug messages
+ */
+#if __STDC__
+static void
+jupiter_debug(struct peer *peer, char *fmt, ...)
+#else
+static void
+jupiter_debug(peer, fmt, va_alist)
+ struct peer *peer;
+ char *fmt;
+#endif
+{
+ va_list ap;
+
+ if (debug) {
+
+#if __STDC__
+ va_start(ap, fmt);
+#else
+ va_start(ap);
+#endif
+ /*
+ * Print debug message to stdout
+ * In the future, we may want to get get more creative...
+ */
+ vfprintf(stderr, fmt, ap);
+
+ va_end(ap);
+ }
+}
+
+/* Checksum and transmit a message to the Jupiter */
+static char *
+jupiter_send(register struct peer *peer, register struct jheader *hp)
+{
+ register u_int len, size;
+ register int cc;
+ register u_short *sp;
+ static char errstr[132];
+
+ size = sizeof(*hp);
+ hp->hsum = putshort(jupiter_cksum((u_short *)hp,
+ (size / sizeof(u_short)) - 1));
+ len = getshort(hp->len);
+ if (len > 0) {
+ sp = (u_short *)(hp + 1);
+ sp[len] = putshort(jupiter_cksum(sp, len));
+ size += (len + 1) * sizeof(u_short);
+ }
+
+ if ((cc = write(peer->procptr->io.fd, (char *)hp, size)) < 0) {
+ (void)sprintf(errstr, "write: %s", strerror(errno));
+ return (errstr);
+ } else if (cc != size) {
+ (void)sprintf(errstr, "short write (%d != %d)", cc, size);
+ return (errstr);
+ }
+ return (NULL);
+}
+
+/* Request periodic message output */
+static struct {
+ struct jheader jheader;
+ struct jrequest jrequest;
+} reqmsg = {
+ { putshort(JUPITER_SYNC), 0,
+ putshort((sizeof(struct jrequest) / sizeof(u_short)) - 1),
+ 0, putshort(JUPITER_FLAG_REQUEST | JUPITER_FLAG_NAK |
+ JUPITER_FLAG_CONN | JUPITER_FLAG_LOG), 0 },
+ { 0, 0, 0, 0 }
+};
+
+/* An interval of zero means to output on trigger */
+static void
+jupiter_reqmsg(register struct peer *peer, register u_int id,
+ register u_int interval)
+{
+ register struct jheader *hp;
+ register struct jrequest *rp;
+ register char *cp;
+
+ hp = &reqmsg.jheader;
+ hp->id = putshort(id);
+ rp = &reqmsg.jrequest;
+ rp->trigger = putshort(interval == 0);
+ rp->interval = putshort(interval);
+ if ((cp = jupiter_send(peer, hp)) != NULL)
+ jupiter_debug(peer, "jupiter_reqmsg: %u: %s\n", id, cp);
+}
+
+/* Cancel periodic message output */
+static struct jheader canmsg = {
+ putshort(JUPITER_SYNC), 0, 0, 0,
+ putshort(JUPITER_FLAG_REQUEST | JUPITER_FLAG_NAK | JUPITER_FLAG_DISC),
+ 0
+};
+
+static void
+jupiter_canmsg(register struct peer *peer, register u_int id)
+{
+ register struct jheader *hp;
+ register char *cp;
+
+ hp = &canmsg;
+ hp->id = putshort(id);
+ if ((cp = jupiter_send(peer, hp)) != NULL)
+ jupiter_debug(peer, "jupiter_canmsg: %u: %s\n", id, cp);
+}
+
+/* Request a single message output */
+static struct jheader reqonemsg = {
+ putshort(JUPITER_SYNC), 0, 0, 0,
+ putshort(JUPITER_FLAG_REQUEST | JUPITER_FLAG_NAK | JUPITER_FLAG_QUERY),
+ 0
+};
+
+static void
+jupiter_reqonemsg(register struct peer *peer, register u_int id)
+{
+ register struct jheader *hp;
+ register char *cp;
+
+ hp = &reqonemsg;
+ hp->id = putshort(id);
+ if ((cp = jupiter_send(peer, hp)) != NULL)
+ jupiter_debug(peer, "jupiter_reqonemsg: %u: %s\n", id, cp);
+}
+
+/* Set the platform dynamics */
+static struct {
+ struct jheader jheader;
+ struct jplat jplat;
+} platmsg = {
+ { putshort(JUPITER_SYNC), putshort(JUPITER_I_PLAT),
+ putshort((sizeof(struct jplat) / sizeof(u_short)) - 1), 0,
+ putshort(JUPITER_FLAG_REQUEST | JUPITER_FLAG_NAK), 0 },
+ { 0, 0, 0 }
+};
+
+static void
+jupiter_platform(register struct peer *peer, register u_int platform)
+{
+ register struct jheader *hp;
+ register struct jplat *pp;
+ register char *cp;
+
+ hp = &platmsg.jheader;
+ pp = &platmsg.jplat;
+ pp->platform = putshort(platform);
+ if ((cp = jupiter_send(peer, hp)) != NULL)
+ jupiter_debug(peer, "jupiter_platform: %u: %s\n", platform, cp);
+}
+
+/* Checksum "len" shorts */
+static u_short
+jupiter_cksum(register u_short *sp, register u_int len)
+{
+ register u_short sum, x;
+
+ sum = 0;
+ while (len-- > 0) {
+ x = *sp++;
+ sum += getshort(x);
+ }
+ return (~sum + 1);
+}
+
+/* Return the size of the next message (or zero if we don't have it all yet) */
+static int
+jupiter_recv(register struct peer *peer)
+{
+ register int n, len, size, cc;
+ register struct refclockproc *pp;
+ register struct jupiterunit *up;
+ register struct jheader *hp;
+ register u_char *bp;
+ register u_short *sp;
+
+ pp = peer->procptr;
+ up = (struct jupiterunit *)pp->unitptr;
+
+ /* Must have at least a header's worth */
+ cc = sizeof(*hp);
+ size = up->ssize;
+ if (size < cc)
+ return (0);
+
+ /* Search for the sync short if missing */
+ sp = up->sbuf;
+ hp = (struct jheader *)sp;
+ if (getshort(hp->sync) != JUPITER_SYNC) {
+ /* Wasn't at the front, sync up */
+ jupiter_debug(peer, "syncing");
+ bp = (u_char *)sp;
+ n = size;
+ while (n >= 2) {
+ if (bp[0] != (JUPITER_SYNC & 0xff)) {
+ jupiter_debug(peer, "{0x%x}", bp[0]);
+ ++bp;
+ --n;
+ continue;
+ }
+ if (bp[1] == ((JUPITER_SYNC >> 8) & 0xff))
+ break;
+ jupiter_debug(peer, "{0x%x 0x%x}", bp[0], bp[1]);
+ bp += 2;
+ n -= 2;
+ }
+ jupiter_debug(peer, "\n");
+ /* Shuffle data to front of input buffer */
+ if (n > 0)
+ memcpy(sp, bp, n);
+ size = n;
+ up->ssize = size;
+ if (size < cc || hp->sync != JUPITER_SYNC)
+ return (0);
+ }
+
+ if (jupiter_cksum(sp, (cc / sizeof(u_short) - 1)) !=
+ getshort(hp->hsum)) {
+ jupiter_debug(peer, "jupiter_recv: bad header checksum!\n");
+ /* This is drastic but checksum errors should be rare */
+ up->ssize = 0;
+ return (0);
+ }
+
+ /* Check for a payload */
+ len = getshort(hp->len);
+ if (len > 0) {
+ n = (len + 1) * sizeof(u_short);
+ /* Not enough data yet */
+ if (size < cc + n)
+ return (0);
+
+ /* Check payload checksum */
+ sp = (u_short *)(hp + 1);
+ if (jupiter_cksum(sp, len) != getshort(sp[len])) {
+ jupiter_debug(peer,
+ "jupiter_recv: bad payload checksum!\n");
+ /* This is drastic but checksum errors should be rare */
+ up->ssize = 0;
+ return (0);
+ }
+ cc += n;
+ }
+ return (cc);
+}
+
+static int
+jupiter_ttyinit(register struct peer *peer, register int fd)
+{
+ struct termios termios;
+
+ memset((char *)&termios, 0, sizeof(termios));
+ if (cfsetispeed(&termios, B9600) < 0 ||
+ cfsetospeed(&termios, B9600) < 0) {
+ jupiter_debug(peer,
+ "jupiter_ttyinit: cfsetispeed/cfgetospeed: %s\n",
+ strerror(errno));
+ return (0);
+ }
+#ifdef HAVE_CFMAKERAW
+ cfmakeraw(&termios);
+#else
+ termios.c_iflag &= ~(IMAXBEL | IXOFF | INPCK | BRKINT | PARMRK |
+ ISTRIP | INLCR | IGNCR | ICRNL | IXON | IGNPAR);
+ termios.c_iflag |= IGNBRK;
+ termios.c_oflag &= ~OPOST;
+ termios.c_lflag &= ~(ECHO | ECHOE | ECHOK | ECHONL | ICANON | ISIG |
+ IEXTEN | NOFLSH | TOSTOP | PENDIN);
+ termios.c_cflag &= ~(CSIZE | PARENB);
+ termios.c_cflag |= CS8 | CREAD;
+ termios.c_cc[VMIN] = 1;
+#endif
+ termios.c_cflag |= CLOCAL;
+ if (tcsetattr(fd, TCSANOW, &termios) < 0) {
+ jupiter_debug(peer, "jupiter_ttyinit: tcsetattr: %s\n",
+ strerror(errno));
+ return (0);
+ }
+
+#ifdef TIOCSPPS
+ if (ioctl(fd, TIOCSPPS, (char *)&fdpps) < 0) {
+ jupiter_debug(peer, "jupiter_ttyinit: TIOCSPPS: %s\n",
+ strerror(errno));
+ return (0);
+ }
+#endif
+#ifdef I_PUSH
+ if (ioctl(fd, I_PUSH, "ppsclock") < 0) {
+ jupiter_debug(peer, "jupiter_ttyinit: push ppsclock: %s\n",
+ strerror(errno));
+ return (0);
+ }
+#endif
+
+ return (1);
+}
+
+#else /* not (REFCLOCK && CLOCK_JUPITER && PPS) */
+int refclock_jupiter_bs;
+#endif /* not (REFCLOCK && CLOCK_JUPITER && PPS) */