/*      $NetBSD: ntp_proto.c,v 1.20 2024/08/18 20:47:17 christos Exp $  */

/*
* ntp_proto.c - NTP version 4 protocol machinery
*
* ATTENTION: Get approval from Harlan on all changes to this file!
*          (Harlan will be discussing these changes with Dave Mills.)
*
*/
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#include "ntpd.h"
#include "ntp_stdlib.h"
#include "ntp_unixtime.h"
#include "ntp_control.h"
#include "ntp_string.h"
#include "ntp_leapsec.h"
#include "ntp_psl.h"
#include "refidsmear.h"

#include <stdio.h>
#ifdef HAVE_LIBSCF_H
#include <libscf.h>
#endif
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif

/* [Bug 3031] define automatic broadcastdelay cutoff preset */
#ifndef BDELAY_DEFAULT
# define BDELAY_DEFAULT (-0.050)
#endif

#define SRVFUZ_SHIFT    6       /* 64 seconds */
#define SRVRSP_FUZZ(x)                                  \
       do {                                            \
               x.l_uf = 0;                             \
               x.l_ui &= ~((1 << SRVFUZ_SHIFT) - 1U);  \
       } while (FALSE)

/*
* This macro defines the authentication state. If x is 1 authentication
* is required; otherwise it is optional.
*/
#define AUTH(x, y)      ((x) ? (y) == AUTH_OK \
                            : (y) == AUTH_OK || (y) == AUTH_NONE)

typedef enum
auth_state {
       AUTH_UNKNOWN = -1,      /* Unknown */
       AUTH_NONE,              /* authentication not required */
       AUTH_OK,                /* authentication OK */
       AUTH_ERROR,             /* authentication error */
       AUTH_CRYPTO             /* crypto_NAK */
} auth_code;

/*
* Set up Kiss Code values
*/

typedef enum
kiss_codes {
       NOKISS,                         /* No Kiss Code */
       RATEKISS,                       /* Rate limit Kiss Code */
       DENYKISS,                       /* Deny Kiss */
       RSTRKISS,                       /* Restricted Kiss */
       XKISS                           /* Experimental Kiss */
} kiss_code;

typedef enum
nak_error_codes {
       NONAK,                          /* No NAK seen */
       INVALIDNAK,                     /* NAK cannot be used */
       VALIDNAK                        /* NAK is valid */
} nak_code;

/*
* traffic shaping parameters
*/
#define NTP_IBURST      6       /* packets in iburst */
#define RESP_DELAY      1       /* refclock burst delay (s) */

/*
* pool soliciting restriction duration (s)
*/
#define POOL_SOLICIT_WINDOW     8

/*
* flag bits propagated from pool/manycast to individual peers
*/
#define POOL_FLAG_PMASK         (FLAG_IBURST | FLAG_NOSELECT)

/*
* peer_select groups statistics for a peer used by clock_select() and
* clock_cluster().
*/
typedef struct peer_select_tag {
       struct peer *   peer;
       double          synch;  /* sync distance */
       double          error;  /* jitter */
       double          seljit; /* selection jitter */
} peer_select;

/*
* System variables are declared here. Unless specified otherwise, all
* times are in seconds.
*/
u_char  sys_leap;               /* system leap indicator, use set_sys_leap() to change this */
u_char  xmt_leap;               /* leap indicator sent in client requests, set up by set_sys_leap() */
u_char  sys_stratum;            /* system stratum */
s_char  sys_precision;          /* local clock precision (log2 s) */
double  sys_rootdelay;          /* roundtrip delay to root (primary source) */
double  sys_rootdisp;           /* dispersion to root (primary source) */
double  prev_rootdisp;          /* previous root dispersion */
double  p2_rootdisp;            /* previous previous root dispersion */
u_int32 sys_refid;              /* reference id (network byte order) */
l_fp    sys_reftime;            /* last update time */
l_fp    prev_reftime;           /* previous sys_reftime */
l_fp    p2_reftime;             /* previous previous sys_reftime */
u_long  prev_time;              /* "current_time" when saved prev_time */
u_long  p2_time;                /* previous prev_time */
struct  peer *sys_peer;         /* current peer */

#ifdef LEAP_SMEAR
struct leap_smear_info leap_smear;
#endif
int leap_sec_in_progress;

/*
* Rate controls. Leaky buckets are used to throttle the packet
* transmission rates in order to protect busy servers such as at NIST
* and USNO. There is a counter for each association and another for KoD
* packets. The association counter decrements each second, but not
* below zero. Each time a packet is sent the counter is incremented by
* a configurable value representing the average interval between
* packets. A packet is delayed as long as the counter is greater than
* zero. Note this does not affect the time value computations.
*/
/*
* Nonspecified system state variables
*/
int     sys_bclient;            /* broadcast client enable */
int     sys_mclient;            /* multicast client enable */
double  sys_bdelay;             /* broadcast client default delay */
int     sys_authenticate;       /* requre authentication for config */
l_fp    sys_authdelay;          /* authentication delay */
double  sys_offset;     /* current local clock offset */
double  sys_mindisp = MINDISPERSE; /* minimum distance (s) */
double  sys_maxdist = MAXDISTANCE; /* selection threshold */
double  sys_jitter;             /* system jitter */
u_long  sys_epoch;              /* last clock update time */
static  double sys_clockhop;    /* clockhop threshold */
static int leap_vote_ins;       /* leap consensus for insert */
static int leap_vote_del;       /* leap consensus for delete */
keyid_t sys_private;            /* private value for session seed */
int     sys_manycastserver;     /* respond to manycast client pkts */
int     ntp_mode7;              /* respond to ntpdc (mode7) */
int     peer_ntpdate;           /* active peers in ntpdate mode */
int     sys_survivors;          /* truest of the truechimers */
char    *sys_ident = NULL;      /* identity scheme */

/*
* TOS and multicast mapping stuff
*/
int     sys_floor = 0;          /* cluster stratum floor */
u_char  sys_bcpollbstep = 0;    /* Broadcast Poll backstep gate */
int     sys_ceiling = STRATUM_UNSPEC - 1; /* cluster stratum ceiling */
int     sys_minsane = 1;        /* minimum candidates */
int     sys_minclock = NTP_MINCLOCK; /* minimum candidates */
int     sys_maxclock = NTP_MAXCLOCK; /* maximum candidates */
int     sys_cohort = 0;         /* cohort switch */
int     sys_orphan = STRATUM_UNSPEC + 1; /* orphan stratum */
int     sys_orphwait = NTP_ORPHWAIT; /* orphan wait */
int     sys_beacon = BEACON;    /* manycast beacon interval */
u_int   sys_ttlmax;             /* max ttl mapping vector index */
u_char  sys_ttl[MAX_TTL];       /* ttl mapping vector */

/*
* Statistics counters - first the good, then the bad
*/
u_long  sys_stattime;           /* elapsed time */
u_long  sys_received;           /* packets received */
u_long  sys_processed;          /* packets for this host */
u_long  sys_newversion;         /* current version */
u_long  sys_oldversion;         /* old version */
u_long  sys_restricted;         /* access denied */
u_long  sys_badlength;          /* bad length or format */
u_long  sys_badauth;            /* bad authentication */
u_long  sys_declined;           /* declined */
u_long  sys_limitrejected;      /* rate exceeded */
u_long  sys_kodsent;            /* KoD sent */

/*
* Mechanism knobs: how soon do we peer_clear() or unpeer()?
*
* The default way is "on-receipt".  If this was a packet from a
* well-behaved source, on-receipt will offer the fastest recovery.
* If this was from a DoS attack, the default way makes it easier
* for a bad-guy to DoS us.  So look and see what bites you harder
* and choose according to your environment.
*/
int peer_clear_digest_early     = 1;    /* bad digest (TEST5) and Autokey */
int unpeer_crypto_early         = 1;    /* bad crypto (TEST9) */
int unpeer_crypto_nak_early     = 1;    /* crypto_NAK (TEST5) */
int unpeer_digest_early         = 1;    /* bad digest (TEST5) */

int dynamic_interleave = DYNAMIC_INTERLEAVE;    /* Bug 2978 mitigation */

int kiss_code_check(u_char hisleap, u_char hisstratum, u_char hismode, u_int32 refid);
nak_code        valid_NAK       (struct peer *peer, struct recvbuf *rbufp, u_char hismode);
static  double  root_distance   (struct peer *);
static  void    clock_combine   (peer_select *, int, int);
static  void    peer_xmit       (struct peer *);
static  void    fast_xmit       (struct recvbuf *, int, keyid_t, int);
static  void    pool_xmit       (struct peer *);
static  void    clock_update    (struct peer *);
static  void    measure_precision(void);
static  double  measure_tick_fuzz(void);
static  int     local_refid     (struct peer *);
static  int     peer_unfit      (struct peer *);
#ifdef AUTOKEY
static  int     group_test      (char *, char *);
#endif /* AUTOKEY */
#ifdef WORKER
void    pool_name_resolved      (int, int, void *, const char *,
                                const char *, const struct addrinfo *,
                                const struct addrinfo *);
#endif /* WORKER */

const char *    amtoa           (int am);


void
set_sys_leap(
       u_char new_sys_leap
       )
{
       sys_leap = new_sys_leap;
       xmt_leap = sys_leap;

       /*
        * Under certain conditions we send faked leap bits to clients, so
        * eventually change xmt_leap below, but never change LEAP_NOTINSYNC.
        */
       if (xmt_leap != LEAP_NOTINSYNC) {
               if (leap_sec_in_progress) {
                       /* always send "not sync" */
                       xmt_leap = LEAP_NOTINSYNC;
               }
#ifdef LEAP_SMEAR
               else {
                       /*
                        * If leap smear is enabled in general we must
                        * never send a leap second warning to clients,
                        * so make sure we only send "in sync".
                        */
                       if (leap_smear.enabled)
                               xmt_leap = LEAP_NOWARNING;
               }
#endif  /* LEAP_SMEAR */
       }
}


/*
* Kiss Code check
*/
int
kiss_code_check(
       u_char hisleap,
       u_char hisstratum,
       u_char hismode,
       u_int32 refid
       )
{

       if (   hismode == MODE_SERVER
           && hisleap == LEAP_NOTINSYNC
           && hisstratum == STRATUM_UNSPEC) {
               if(memcmp(&refid,"RATE", 4) == 0) {
                       return (RATEKISS);
               } else if(memcmp(&refid,"DENY", 4) == 0) {
                       return (DENYKISS);
               } else if(memcmp(&refid,"RSTR", 4) == 0) {
                       return (RSTRKISS);
               } else if(memcmp(&refid,"X", 1) == 0) {
                       return (XKISS);
               }
       }
       return (NOKISS);
}


/*
* Check that NAK is valid
*/
nak_code
valid_NAK(
         struct peer *peer,
         struct recvbuf *rbufp,
         u_char hismode
         )
{
       int             base_packet_length = MIN_V4_PKT_LEN;
       int             remainder_size;
       struct pkt *    rpkt;
       int             keyid;
       l_fp            p_org;  /* origin timestamp */
       const l_fp *    myorg;  /* selected peer origin */

       /*
        * Check to see if there is something beyond the basic packet
        */
       if (rbufp->recv_length == base_packet_length) {
               return NONAK;
       }

       remainder_size = rbufp->recv_length - base_packet_length;
       /*
        * Is this a potential NAK?
        */
       if (remainder_size != 4) {
               return NONAK;
       }

       /*
        * Only server responses can contain NAK's
        */

       if (hismode != MODE_SERVER &&
           hismode != MODE_ACTIVE &&
           hismode != MODE_PASSIVE
           ) {
               return INVALIDNAK;
       }

       /*
        * Make sure that the extra field in the packet is all zeros
        */
       rpkt = &rbufp->recv_pkt;
       keyid = ntohl(((u_int32 *)rpkt)[base_packet_length / 4]);
       if (keyid != 0) {
               return INVALIDNAK;
       }

       /*
        * During the first few packets of the autokey dance there will
        * not (yet) be a keyid, but in this case FLAG_SKEY is set.
        * So the NAK is invalid if either there's no peer, or
        * if the keyid is 0 and FLAG_SKEY is not set.
        */
       if (!peer || (!peer->keyid && !(peer->flags & FLAG_SKEY))) {
               return INVALIDNAK;
       }

       /*
        * The ORIGIN must match, or this cannot be a valid NAK, either.
        */

       if (FLAG_LOOPNONCE & peer->flags) {
               myorg = &peer->nonce;
       } else {
               if (peer->flip > 0) {
                       myorg = &peer->borg;
               } else {
                       myorg = &peer->aorg;
               }
       }

       NTOHL_FP(&rpkt->org, &p_org);

       if (L_ISZERO(&p_org) ||
           L_ISZERO( myorg) ||
           !L_ISEQU(&p_org, myorg)) {
               return INVALIDNAK;
       }

       /* If we ever passed all that checks, we should be safe. Well,
        * as safe as we can ever be with an unauthenticated crypto-nak.
        */
       return VALIDNAK;
}


/*
* transmit - transmit procedure called by poll timeout
*/
void
transmit(
       struct peer *peer       /* peer structure pointer */
       )
{
       u_char  hpoll;

       /*
        * The polling state machine. There are two kinds of machines,
        * those that never expect a reply (broadcast and manycast
        * server modes) and those that do (all other modes). The dance
        * is intricate...
        */
       hpoll = peer->hpoll;

       /*
        * If we haven't received anything (even if unsync) since last
        * send, reset ppoll.
        */
       if (peer->outdate > peer->timelastrec && !peer->reach)
               peer->ppoll = peer->maxpoll;

       /*
        * In broadcast mode the poll interval is never changed from
        * minpoll.
        */
       if (peer->cast_flags & (MDF_BCAST | MDF_MCAST)) {
               peer->outdate = current_time;
               poll_update(peer, hpoll, 0);
               if (sys_leap != LEAP_NOTINSYNC)
                       peer_xmit(peer);
               return;
       }

       /*
        * In manycast mode we start with unity ttl. The ttl is
        * increased by one for each poll until either sys_maxclock
        * servers have been found or the maximum ttl is reached. When
        * sys_maxclock servers are found we stop polling until one or
        * more servers have timed out or until less than sys_minclock
        * associations turn up. In this case additional better servers
        * are dragged in and preempt the existing ones.  Once every
        * sys_beacon seconds we are to transmit unconditionally, but
        * this code is not quite right -- peer->unreach counts polls
        * and is being compared with sys_beacon, so the beacons happen
        * every sys_beacon polls.
        */
       if (peer->cast_flags & MDF_ACAST) {
               peer->outdate = current_time;
               poll_update(peer, hpoll, 0);
               if (peer->unreach > sys_beacon) {
                       peer->unreach = 0;
                       peer->ttl = 0;
                       peer_xmit(peer);
               } else if (   sys_survivors < sys_minclock
                          || peer_associations < sys_maxclock) {
                       if (peer->ttl < sys_ttlmax)
                               peer->ttl++;
                       peer_xmit(peer);
               }
               peer->unreach++;
               return;
       }

       /*
        * Pool associations transmit unicast solicitations when there
        * are less than a hard limit of 2 * sys_maxclock associations,
        * and either less than sys_minclock survivors or less than
        * sys_maxclock associations.  The hard limit prevents unbounded
        * growth in associations if the system clock or network quality
        * result in survivor count dipping below sys_minclock often.
        * This was observed testing with pool, where sys_maxclock == 12
        * resulted in 60 associations without the hard limit.  A
        * similar hard limit on manycastclient ephemeral associations
        * may be appropriate.
        */
       if (peer->cast_flags & MDF_POOL) {
               peer->outdate = current_time;
               poll_update(peer, hpoll, 0);
               if (   (peer_associations <= 2 * sys_maxclock)
                   && (   peer_associations < sys_maxclock
                       || sys_survivors < sys_minclock))
                       pool_xmit(peer);
               return;
       }

       /* [Bug 3851] drop pool servers which can no longer be reached. */
       if (MDF_PCLNT & peer->cast_flags) {
               if (   (IS_IPV6(&peer->srcadr) && !nonlocal_v6_addr_up)
                   || !nonlocal_v4_addr_up) {
                       unpeer(peer);
                       return;
               }
       }

        /*
        * In unicast modes the dance is much more intricate. It is
        * designed to back off whenever possible to minimize network
        * traffic.
        */
       if (peer->burst == 0) {
               u_char oreach;

               /*
                * Update the reachability status. If not heard for
                * three consecutive polls, stuff infinity in the clock
                * filter.
                */
               oreach = peer->reach;
               peer->outdate = current_time;
               peer->unreach++;
               peer->reach <<= 1;
               if (!peer->reach) {

                       /*
                        * Here the peer is unreachable. If it was
                        * previously reachable raise a trap. Send a
                        * burst if enabled.
                        */
                       clock_filter(peer, 0., 0., MAXDISPERSE);
                       if (oreach) {
                               peer_unfit(peer);
                               report_event(PEVNT_UNREACH, peer, NULL);
                       }
                       if (   (peer->flags & FLAG_IBURST)
                           && peer->retry == 0)
                               peer->retry = NTP_RETRY;
               } else {

                       /*
                        * Here the peer is reachable. Send a burst if
                        * enabled and the peer is fit.  Reset unreach
                        * for persistent and ephemeral associations.
                        * Unreach is also reset for survivors in
                        * clock_select().
                        */
                       hpoll = sys_poll;
                       if (!(peer->flags & FLAG_PREEMPT))
                               peer->unreach = 0;
                       if (   (peer->flags & FLAG_BURST)
                           && peer->retry == 0
                           && !peer_unfit(peer))
                               peer->retry = NTP_RETRY;
               }

               /*
                * Watch for timeout.  If ephemeral, toss the rascal;
                * otherwise, bump the poll interval. Note the
                * poll_update() routine will clamp it to maxpoll.
                * If preemptible and we have more peers than maxclock,
                * and this peer has the minimum score of preemptibles,
                * demobilize.
                */
               if (peer->unreach >= NTP_UNREACH) {
                       hpoll++;
                       /* ephemeral: no FLAG_CONFIG nor FLAG_PREEMPT */
                       if (!(peer->flags & (FLAG_CONFIG | FLAG_PREEMPT))) {
                               report_event(PEVNT_RESTART, peer, "timeout");
                               peer_clear(peer, "TIME");
                               unpeer(peer);
                               return;
                       }
                       if (   (peer->flags & FLAG_PREEMPT)
                           && (peer_associations > sys_maxclock)
                           && score_all(peer)) {
                               report_event(PEVNT_RESTART, peer, "timeout");
                               peer_clear(peer, "TIME");
                               unpeer(peer);
                               return;
                       }
               }
       } else {
               peer->burst--;
               if (peer->burst == 0) {

                       /*
                        * If ntpdate mode and the clock has not been
                        * set and all peers have completed the burst,
                        * we declare a successful failure.
                        */
                       if (mode_ntpdate) {
                               peer_ntpdate--;
                               if (peer_ntpdate == 0) {
                                       msyslog(LOG_NOTICE,
                                           "ntpd: no servers found");
                                       if (!msyslog_term)
                                               printf(
                                                   "ntpd: no servers found\n");
                                       exit (0);
                               }
                       }
               }
       }
       if (peer->retry > 0)
               peer->retry--;

       /*
        * Do not transmit if in broadcast client mode.
        */
       poll_update(peer, hpoll, (peer->hmode == MODE_CLIENT));
       if (peer->hmode != MODE_BCLIENT)
               peer_xmit(peer);

       return;
}


#ifdef DEBUG
const char *
amtoa(
       int am
       )
{
       char *bp;

       switch(am) {
           case AM_ERR:        return "AM_ERR";
           case AM_NOMATCH:    return "AM_NOMATCH";
           case AM_PROCPKT:    return "AM_PROCPKT";
           case AM_BCST:       return "AM_BCST";
           case AM_FXMIT:      return "AM_FXMIT";
           case AM_MANYCAST:   return "AM_MANYCAST";
           case AM_NEWPASS:    return "AM_NEWPASS";
           case AM_NEWBCL:     return "AM_NEWBCL";
           case AM_POSSBCL:    return "AM_POSSBCL";
           default:
               LIB_GETBUF(bp);
               snprintf(bp, LIB_BUFLENGTH, "AM_#%d", am);
               return bp;
       }
}
#endif  /* DEBUG */


/*
* receive - receive procedure called for each packet received
*/
void
receive(
       struct recvbuf *rbufp
       )
{
       register struct peer *peer;     /* peer structure pointer */
       register struct pkt *pkt;       /* receive packet pointer */
       u_char  hisversion;             /* packet version */
       u_char  hisleap;                /* packet leap indicator */
       u_char  hismode;                /* packet mode */
       u_char  hisstratum;             /* packet stratum */
       r4addr  r4a;                    /* address restrictions */
       u_short restrict_mask;          /* restrict bits */
       const char *hm_str;             /* hismode string */
       int     kissCode = NOKISS;      /* Kiss Code */
       int     has_mac;                /* length of MAC field */
       int     authlen;                /* offset of MAC field */
       auth_code is_authentic = AUTH_UNKNOWN;  /* Was AUTH_NONE */
       nak_code crypto_nak_test;       /* result of crypto-NAK check */
       int     retcode = AM_NOMATCH;   /* match code */
       keyid_t skeyid = 0;             /* key IDs */
       u_int32 opcode = 0;             /* extension field opcode */
       sockaddr_u *dstadr_sin;         /* active runway */
       u_char  cast_flags;             /* MDF_* flags for newpeer() */
       struct peer *peer2;             /* aux peer structure pointer */
       endpt   *match_ep;              /* newpeer() local address */
       l_fp    p_org;                  /* origin timestamp */
       l_fp    p_rec;                  /* receive timestamp */
       l_fp    p_xmt;                  /* transmit timestamp */
#ifdef DEBUG
       const char *am_str;             /* association match string */
#endif
#ifdef AUTOKEY
       char    hostname[NTP_MAXSTRLEN + 1];
       char    *groupname = NULL;
       struct autokey *ap;             /* autokey structure pointer */
       int     rval;                   /* cookie snatcher */
       keyid_t pkeyid = 0, tkeyid = 0; /* key IDs */
#endif  /* AUTOKEY */
#ifdef HAVE_NTP_SIGND
       static unsigned char zero_key[16];
#endif /* HAVE_NTP_SIGND */

       /*
        * Note that there are many places we do not call record_raw_stats().
        *
        * We only want to call it *after* we've sent a response, or perhaps
        * when we've decided to drop a packet.
        */

       /*
        * Monitor the packet and get restrictions. Note that the packet
        * length for control and private mode packets must be checked
        * by the service routines. Some restrictions have to be handled
        * later in order to generate a kiss-o'-death packet.
        */
       /*
        * Bogus port check is before anything, since it probably
        * reveals a clogging attack. Likewise the mimimum packet size
        * of 2 bytes (for mode 6/7) must be checked first.
        */
       sys_received++;
       if (0 == SRCPORT(&rbufp->recv_srcadr) || rbufp->recv_length < 2) {
               sys_badlength++;
               return;                         /* bogus port / length */
       }
       restrictions(&rbufp->recv_srcadr, &r4a);
       restrict_mask = r4a.rflags;

       pkt = &rbufp->recv_pkt;
       hisversion = PKT_VERSION(pkt->li_vn_mode);
       hismode = (int)PKT_MODE(pkt->li_vn_mode);

       if (restrict_mask & RES_IGNORE) {
               DPRINTF(2, ("receive: drop: RES_IGNORE\n"));
               sys_restricted++;
               return;                         /* ignore everything */
       }
       if (hismode == MODE_PRIVATE) {
               if (!ntp_mode7 || (restrict_mask & RES_NOQUERY)) {
                       DPRINTF(2, ("receive: drop: !mode7 or RES_NOQUERY\n"));
                       sys_restricted++;
                       return;                 /* no query private */
               }
               process_private(rbufp, !(RES_NOMODIFY & restrict_mask));
               return;
       }
       if (hismode == MODE_CONTROL) {
               if (restrict_mask & RES_NOQUERY) {
                       DPRINTF(2, ("receive: drop: RES_NOQUERY\n"));
                       sys_restricted++;
                       return;                 /* no query control */
               }
               process_control(rbufp, restrict_mask);
               return;
       }
       if (restrict_mask & RES_DONTSERVE) {
               DPRINTF(2, ("receive: drop: RES_DONTSERVE\n"));
               sys_restricted++;
               return;                         /* no time serve */
       }


       /* If we arrive here, we should have a standard NTP packet. We
        * check that the minimum size is available and fetch some more
        * items from the packet once we can be sure they are indeed
        * there.
        */
       if ((size_t)rbufp->recv_length < LEN_PKT_NOMAC) {
               sys_badlength++;
               return;                         /* bogus length */
       }

       hisleap = PKT_LEAP(pkt->li_vn_mode);
       hisstratum = PKT_TO_STRATUM(pkt->stratum);
       DEBUG_INSIST(0 != hisstratum);  /* paranoia check PKT_TO_STRATUM result */
                                       /* TODO: this should be in a unit test */
       DPRINTF(1, ("receive: at %ld %s<-%s ippeerlimit %d mode %d iflags %s "
                   "restrict %s org 0x%x.%08x xmt 0x%x.%08x\n",
                   current_time, stoa(&rbufp->dstadr->sin),
                   stoa(&rbufp->recv_srcadr), r4a.ippeerlimit, hismode,
                   iflags_str(rbufp->dstadr->flags),
                   rflags_str(restrict_mask),
                   ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                   ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf)));

       /*
        * This is for testing. If restricted drop ten percent of
        * surviving packets.
        */
       if (restrict_mask & RES_FLAKE) {
               if (ntp_uurandom() < .1) {
                       DPRINTF(2, ("receive: drop: RES_FLAKE\n"));
                       sys_restricted++;
                       return;                 /* no flakeway */
               }
       }

       /*
       ** Format Layer Checks
       **
       ** Validate the packet format.  The packet size, packet header,
       ** and any extension field lengths are checked.  We identify
       ** the beginning of the MAC, to identify the upper limit of
       ** of the hash computation.
       **
       ** In case of a format layer check violation, the packet is
       ** discarded with no further processing.
       */

       /*
        * Version check must be after the query packets, since they
        * intentionally use an early version.
        */
       if (hisversion == NTP_VERSION) {
               sys_newversion++;               /* new version */
       } else if (   !(restrict_mask & RES_VERSION)
                  && hisversion >= NTP_OLDVERSION) {
               sys_oldversion++;               /* previous version */
       } else {
               DPRINTF(2, ("receive: drop: RES_VERSION\n"));
               sys_badlength++;
               return;                         /* old version */
       }

       /*
        * Figure out his mode and validate the packet. This has some
        * legacy raunch that probably should be removed. In very early
        * NTP versions mode 0 was equivalent to what later versions
        * would interpret as client mode.
        */
       if (hismode == MODE_UNSPEC) {
               if (hisversion == NTP_OLDVERSION) {
                       hismode = MODE_CLIENT;
               } else {
                       DPRINTF(2, ("receive: drop: MODE_UNSPEC\n"));
                       sys_badlength++;
                       return;                 /* invalid mode */
               }
       }

       /*
        * Validate the poll interval in the packet.
        * 0 probably indicates a data-minimized packet.
        * A valid poll interval is required for RATEKISS, where
        * a value of 0 is not allowed.  We check for this below.
        *
        * There might be arguments against this check.  If you have
        * any of these arguments, please let us know.
        *
        * At this point, the packet cannot be a mode[67] packet.
        */
       if (   pkt->ppoll
           && (   (NTP_MINPOLL > pkt->ppoll)
               || (NTP_MAXPOLL < pkt->ppoll)
              )
          ) {
#ifdef BUG3870
               DPRINTF(2, ("receive: drop: Invalid ppoll (%d) from %s\n",
                               pkt->ppoll, stoa(&rbufp->recv_srcadr)));
               sys_badlength++;
               return;                 /* invalid packet poll */
#else
               DPRINTF(2, ("receive: info: Invalid ppoll (%d) from %s\n",
                               pkt->ppoll, stoa(&rbufp->recv_srcadr)));
#endif
       }

       /*
        * Parse the extension field if present. We figure out whether
        * an extension field is present by measuring the MAC size. If
        * the number of words following the packet header is 0, no MAC
        * is present and the packet is not authenticated. If 1, the
        * packet is a crypto-NAK; if 3, the packet is authenticated
        * with DES; if 5, the packet is authenticated with MD5; if 6,
        * the packet is authenticated with SHA. If 2 or * 4, the packet
        * is a runt and discarded forthwith. If greater than 6, an
        * extension field is present, so we subtract the length of the
        * field and go around again.
        *
        * Note the above description is lame.  We should/could also check
        * the two bytes that make up the EF type and subtype, and then
        * check the two bytes that tell us the EF length.  A legacy MAC
        * has a 4 byte keyID, and for conforming symmetric keys its value
        * must be <= 64k, meaning the top two bytes will always be zero.
        * Since the EF Type of 0 is reserved/unused, there's no way a
        * conforming legacy MAC could ever be misinterpreted as an EF.
        *
        * There is more, but this isn't the place to document it.
        */

       authlen = LEN_PKT_NOMAC;
       has_mac = rbufp->recv_length - authlen;
       while (has_mac > 0) {
               u_int32 len;
#ifdef AUTOKEY
               u_int32 hostlen;
               struct exten *ep;
#endif /*AUTOKEY */

               if (has_mac % 4 != 0 || has_mac < (int)MIN_MAC_LEN) {
                       DPRINTF(2, ("receive: drop: bad post-packet length\n"));
                       sys_badlength++;
                       return;                 /* bad length */
               }
               /*
                * This next test is clearly wrong - it needlessly
                * prohibits short EFs (which don't yet exist)
                */
               if (has_mac <= (int)MAX_MAC_LEN) {
                       skeyid = ntohl(((u_int32 *)pkt)[authlen / 4]);
                       break;

               } else {
                       opcode = ntohl(((u_int32 *)pkt)[authlen / 4]);
                       len = opcode & 0xffff;
                       if (   len % 4 != 0
                           || len < 4
                           || (int)len + authlen > rbufp->recv_length) {
                               DPRINTF(2, ("receive: drop: bad EF length\n"));
                               sys_badlength++;
                               return;         /* bad length */
                       }
#ifdef AUTOKEY
                       /*
                        * Extract calling group name for later.  If
                        * sys_groupname is non-NULL, there must be
                        * a group name provided to elicit a response.
                        */
                       if (   (opcode & 0x3fff0000) == CRYPTO_ASSOC
                           && sys_groupname != NULL) {
                               ep = (struct exten *)&((u_int32 *)pkt)[authlen / 4];
                               hostlen = ntohl(ep->vallen);
                               if (   hostlen >= sizeof(hostname)
                                   || hostlen > len -
                                               offsetof(struct exten, pkt)) {
                                       DPRINTF(2, ("receive: drop: bad autokey hostname length\n"));
                                       sys_badlength++;
                                       return;         /* bad length */
                               }
                               memcpy(hostname, &ep->pkt, hostlen);
                               hostname[hostlen] = '\0';
                               groupname = strchr(hostname, '@');
                               if (groupname == NULL) {
                                       DPRINTF(2, ("receive: drop: empty autokey groupname\n"));
                                       sys_declined++;
                                       return;
                               }
                               groupname++;
                       }
#endif /* AUTOKEY */
                       authlen += len;
                       has_mac -= len;
               }
       }

       /*
        * If has_mac is < 0 we had a malformed packet.
        */
       if (has_mac < 0) {
               DPRINTF(2, ("receive: drop: post-packet under-read\n"));
               sys_badlength++;
               return;         /* bad length */
       }

       /*
       ** Packet Data Verification Layer
       **
       ** This layer verifies the packet data content.  If
       ** authentication is required, a MAC must be present.
       ** If a MAC is present, it must validate.
       ** Crypto-NAK?  Look - a shiny thing!
       **
       ** If authentication fails, we're done.
       */

       /*
        * If authentication is explicitly required, a MAC must be present.
        */
       if (restrict_mask & RES_DONTTRUST && has_mac == 0) {
               DPRINTF(2, ("receive: drop: RES_DONTTRUST\n"));
               sys_restricted++;
               return;                         /* access denied */
       }

       /*
        * Update the MRU list and finger the cloggers. It can be a
        * little expensive, so turn it off for production use.
        * RES_LIMITED and RES_KOD will be cleared in the returned
        * restrict_mask unless one or both actions are warranted.
        */
       restrict_mask = ntp_monitor(rbufp, restrict_mask);
       if (restrict_mask & RES_LIMITED) {
               sys_limitrejected++;
               if (   !(restrict_mask & RES_KOD)
                   || MODE_BROADCAST == hismode
                   || MODE_SERVER == hismode) {
                       if (MODE_SERVER == hismode) {
                               DPRINTF(1, ("Possibly self-induced rate limiting of MODE_SERVER from %s\n",
                                       stoa(&rbufp->recv_srcadr)));
                       } else {
                               DPRINTF(2, ("receive: drop: RES_KOD\n"));
                       }
                       return;                 /* rate exceeded */
               }
               if (hismode == MODE_CLIENT) {
                       fast_xmit(rbufp, MODE_SERVER, skeyid,
                           restrict_mask);
               } else {
                       fast_xmit(rbufp, MODE_ACTIVE, skeyid,
                           restrict_mask);
               }
               return;                         /* rate exceeded */
       }
       restrict_mask &= ~RES_KOD;

       /*
        * We have tossed out as many buggy packets as possible early in
        * the game to reduce the exposure to a clogging attack. Now we
        * have to burn some cycles to find the association and
        * authenticate the packet if required. Note that we burn only
        * digest cycles, again to reduce exposure. There may be no
        * matching association and that's okay.
        *
        * More on the autokey mambo. Normally the local interface is
        * found when the association was mobilized with respect to a
        * designated remote address. We assume packets arriving from
        * the remote address arrive via this interface and the local
        * address used to construct the autokey is the unicast address
        * of the interface. However, if the sender is a broadcaster,
        * the interface broadcast address is used instead.
        * Notwithstanding this technobabble, if the sender is a
        * multicaster, the broadcast address is null, so we use the
        * unicast address anyway. Don't ask.
        */

       peer = findpeer(rbufp,  hismode, &retcode);
       dstadr_sin = &rbufp->dstadr->sin;
       NTOHL_FP(&pkt->org, &p_org);
       NTOHL_FP(&pkt->rec, &p_rec);
       NTOHL_FP(&pkt->xmt, &p_xmt);
       hm_str = modetoa(hismode);
#ifdef DEBUG
       am_str = amtoa(retcode);
#endif

       /*
        * Authentication is conditioned by three switches:
        *
        * NOPEER  (RES_NOPEER) do not mobilize an association unless
        *         authenticated
        * NOTRUST (RES_DONTTRUST) do not allow access unless
        *         authenticated (implies NOPEER)
        * enable  (sys_authenticate) master NOPEER switch, by default
        *         on
        *
        * The NOPEER and NOTRUST can be specified on a per-client basis
        * using the restrict command. The enable switch if on implies
        * NOPEER for all clients. There are four outcomes:
        *
        * NONE    The packet has no MAC.
        * OK      the packet has a MAC and authentication succeeds
        * ERROR   the packet has a MAC and authentication fails
        * CRYPTO  crypto-NAK. The MAC has four octets only.
        *
        * Note: The AUTH(x, y) macro is used to filter outcomes. If x
        * is zero, acceptable outcomes of y are NONE and OK. If x is
        * one, the only acceptable outcome of y is OK.
        */
       crypto_nak_test = valid_NAK(peer, rbufp, hismode);

       /*
        * Drop any invalid crypto-NAKs
        */
       if (crypto_nak_test == INVALIDNAK) {
               report_event(PEVNT_AUTH, peer, "Invalid_NAK");
               if (0 != peer) {
                       peer->badNAK++;
               }
               msyslog(LOG_ERR, "Invalid-NAK error at %ld %s<-%s",
                       current_time, stoa(dstadr_sin), stoa(&rbufp->recv_srcadr));
               return;
       }

       if (has_mac == 0) {
               restrict_mask &= ~RES_MSSNTP;
               is_authentic = AUTH_NONE; /* not required */
               DPRINTF(1, ("receive: at %ld %s<-%s mode %d/%s:%s len %d org 0x%x.%08x xmt 0x%x.%08x NOMAC\n",
                           current_time, stoa(dstadr_sin),
                           stoa(&rbufp->recv_srcadr), hismode, hm_str, am_str,
                           authlen,
                           ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                           ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf)));
       } else if (crypto_nak_test == VALIDNAK) {
               restrict_mask &= ~RES_MSSNTP;
               is_authentic = AUTH_CRYPTO; /* crypto-NAK */
               DPRINTF(1, ("receive: at %ld %s<-%s mode %d/%s:%s keyid %08x len %d auth %d org 0x%x.%08x xmt 0x%x.%08x CRYPTONAK\n",
                           current_time, stoa(dstadr_sin),
                           stoa(&rbufp->recv_srcadr), hismode, hm_str, am_str,
                           skeyid, authlen + has_mac, is_authentic,
                           ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                           ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf)));

#ifdef HAVE_NTP_SIGND
               /*
                * If the signature is 20 bytes long, the last 16 of
                * which are zero, then this is a Microsoft client
                * wanting AD-style authentication of the server's
                * reply.
                *
                * This is described in Microsoft's WSPP docs, in MS-SNTP:
                * http://msdn.microsoft.com/en-us/library/cc212930.aspx
                */
       } else if (   has_mac == MAX_MD5_LEN
                  && (restrict_mask & RES_MSSNTP)
                  && (retcode == AM_FXMIT || retcode == AM_NEWPASS)
                  && (memcmp(zero_key, (char *)pkt + authlen + 4,
                             MAX_MD5_LEN - 4) == 0)) {
               is_authentic = AUTH_NONE;
               DPRINTF(1, ("receive: at %ld %s<-%s mode %d/%s:%s len %d org %x.%08x xmt %x.%08x SIGND\n",
                           current_time, stoa(dstadr_sin),
                           stoa(&rbufp->recv_srcadr), hismode, hm_str, am_str,
                           authlen,
                           ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                           ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf)));
#endif /* HAVE_NTP_SIGND */

       } else {
               /*
                * has_mac is not 0
                * Not a VALID_NAK
                * Not an MS-SNTP SIGND packet
                *
                * So there is a MAC here.
                */

               restrict_mask &= ~RES_MSSNTP;
#ifdef AUTOKEY
               /*
                * For autokey modes, generate the session key
                * and install in the key cache. Use the socket
                * broadcast or unicast address as appropriate.
                */
               if (crypto_flags && skeyid > NTP_MAXKEY) {

                       /*
                        * More on the autokey dance (AKD). A cookie is
                        * constructed from public and private values.
                        * For broadcast packets, the cookie is public
                        * (zero). For packets that match no
                        * association, the cookie is hashed from the
                        * addresses and private value. For server
                        * packets, the cookie was previously obtained
                        * from the server. For symmetric modes, the
                        * cookie was previously constructed using an
                        * agreement protocol; however, should PKI be
                        * unavailable, we construct a fake agreement as
                        * the EXOR of the peer and host cookies.
                        *
                        * hismode      ephemeral       persistent
                        * =======================================
                        * active       0               cookie#
                        * passive      0%              cookie#
                        * client       sys cookie      0%
                        * server       0%              sys cookie
                        * broadcast    0               0
                        *
                        * # if unsync, 0
                        * % can't happen
                        */
                       if (has_mac < (int)MAX_MD5_LEN) {
                               DPRINTF(2, ("receive: drop: MD5 digest too short\n"));
                               sys_badauth++;
                               return;
                       }
                       if (hismode == MODE_BROADCAST) {

                               /*
                                * For broadcaster, use the interface
                                * broadcast address when available;
                                * otherwise, use the unicast address
                                * found when the association was
                                * mobilized. However, if this is from
                                * the wildcard interface, game over.
                                */
                               if (   crypto_flags
                                   && rbufp->dstadr ==
                                      ANY_INTERFACE_CHOOSE(&rbufp->recv_srcadr)) {
                                       DPRINTF(2, ("receive: drop: BCAST from wildcard\n"));
                                       sys_restricted++;
                                       return;         /* no wildcard */
                               }
                               pkeyid = 0;
                               if (!SOCK_UNSPEC(&rbufp->dstadr->bcast))
                                       dstadr_sin =
                                           &rbufp->dstadr->bcast;
                       } else if (peer == NULL) {
                               pkeyid = session_key(
                                   &rbufp->recv_srcadr, dstadr_sin, 0,
                                   sys_private, 0);
                       } else {
                               pkeyid = peer->pcookie;
                       }

                       /*
                        * The session key includes both the public
                        * values and cookie. In case of an extension
                        * field, the cookie used for authentication
                        * purposes is zero. Note the hash is saved for
                        * use later in the autokey mambo.
                        */
                       if (authlen > (int)LEN_PKT_NOMAC && pkeyid != 0) {
                               session_key(&rbufp->recv_srcadr,
                                   dstadr_sin, skeyid, 0, 2);
                               tkeyid = session_key(
                                   &rbufp->recv_srcadr, dstadr_sin,
                                   skeyid, pkeyid, 0);
                       } else {
                               tkeyid = session_key(
                                   &rbufp->recv_srcadr, dstadr_sin,
                                   skeyid, pkeyid, 2);
                       }

               }
#endif  /* AUTOKEY */

               /*
                * Compute the cryptosum. Note a clogging attack may
                * succeed in bloating the key cache. If an autokey,
                * purge it immediately, since we won't be needing it
                * again. If the packet is authentic, it can mobilize an
                * association. Note that there is no key zero.
                */
               if (!authdecrypt(skeyid, (u_int32 *)pkt, authlen,
                   has_mac))
                       is_authentic = AUTH_ERROR;
               else
                       is_authentic = AUTH_OK;
#ifdef AUTOKEY
               if (crypto_flags && skeyid > NTP_MAXKEY)
                       authtrust(skeyid, 0);
#endif  /* AUTOKEY */
               DPRINTF(1, ("receive: at %ld %s<-%s mode %d/%s:%s keyid %08x len %d auth %d org 0x%x.%08x xmt 0x%x.%08x MAC\n",
                           current_time, stoa(dstadr_sin),
                           stoa(&rbufp->recv_srcadr), hismode, hm_str, am_str,
                           skeyid, authlen + has_mac, is_authentic,
                           ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                           ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf)));
       }


       /*
        * Bug 3454:
        *
        * Now come at this from a different perspective:
        * - If we expect a MAC and it's not there, we drop it.
        * - If we expect one keyID and get another, we drop it.
        * - If we have a MAC ahd it hasn't been validated yet, try.
        * - if the provided MAC doesn't validate, we drop it.
        *
        * There might be more to this.
        */
       if (0 != peer && 0 != peer->keyid) {
               /* Should we msyslog() any of these? */

               /*
                * This should catch:
                * - no keyID where one is expected,
                * - different keyID than what we expect.
                */
               if (peer->keyid != skeyid) {
                       DPRINTF(2, ("receive: drop: Wanted keyID %d, got %d from %s\n",
                                   peer->keyid, skeyid,
                                   stoa(&rbufp->recv_srcadr)));
                       sys_restricted++;
                       return;                 /* drop: access denied */
               }

               /*
                * if has_mac != 0 ...
                * - If it has not yet been validated, do so.
                *   (under what circumstances might that happen?)
                * - if missing or bad MAC, log and drop.
                */
               if (0 != has_mac) {
                       if (is_authentic == AUTH_UNKNOWN) {
                               /* How can this happen? */
                               DPRINTF(2, ("receive: 3454 check: AUTH_UNKNOWN from %s\n",
                                   stoa(&rbufp->recv_srcadr)));
                               if (!authdecrypt(skeyid, (u_int32 *)pkt, authlen,
                                   has_mac)) {
                                       /* MAC invalid or not found */
                                       is_authentic = AUTH_ERROR;
                               } else {
                                       is_authentic = AUTH_OK;
                               }
                       }
                       if (is_authentic != AUTH_OK) {
                               DPRINTF(2, ("receive: drop: missing or bad MAC from %s\n",
                                           stoa(&rbufp->recv_srcadr)));
                               sys_restricted++;
                               return;         /* drop: access denied */
                       }
               }
       }
       /**/

       /*
       ** On-Wire Protocol Layer
       **
       ** Verify protocol operations consistent with the on-wire protocol.
       ** The protocol discards bogus and duplicate packets as well as
       ** minimizes disruptions doe to protocol restarts and dropped
       ** packets.  The operations are controlled by two timestamps:
       ** the transmit timestamp saved in the client state variables,
       ** and the origin timestamp in the server packet header.  The
       ** comparison of these two timestamps is called the loopback test.
       ** The transmit timestamp functions as a nonce to verify that the
       ** response corresponds to the original request.  The transmit
       ** timestamp also serves to discard replays of the most recent
       ** packet.  Upon failure of either test, the packet is discarded
       ** with no further action.
       */

       /*
        * The association matching rules are implemented by a set of
        * routines and an association table. A packet matching an
        * association is processed by the peer process for that
        * association. If there are no errors, an ephemeral association
        * is mobilized: a broadcast packet mobilizes a broadcast client
        * aassociation; a manycast server packet mobilizes a manycast
        * client association; a symmetric active packet mobilizes a
        * symmetric passive association.
        */
       DPRINTF(1, ("receive: MATCH_ASSOC dispatch: mode %d/%s:%s \n",
               hismode, hm_str, am_str));
       switch (retcode) {

       /*
        * This is a client mode packet not matching any association. If
        * an ordinary client, simply toss a server mode packet back
        * over the fence. If a manycast client, we have to work a
        * little harder.
        *
        * There are cases here where we do not call record_raw_stats().
        */
       case AM_FXMIT:

               /*
                * If authentication OK, send a server reply; otherwise,
                * send a crypto-NAK.
                */
               if (!(rbufp->dstadr->flags & INT_MCASTOPEN)) {
                       /* HMS: would be nice to log FAST_XMIT|BADAUTH|RESTRICTED */
                       record_raw_stats(&rbufp->recv_srcadr,
                           &rbufp->dstadr->sin,
                           &p_org, &p_rec, &p_xmt, &rbufp->recv_time,
                           PKT_LEAP(pkt->li_vn_mode),
                           PKT_VERSION(pkt->li_vn_mode),
                           PKT_MODE(pkt->li_vn_mode),
                           PKT_TO_STRATUM(pkt->stratum),
                           pkt->ppoll,
                           pkt->precision,
                           FPTOD(NTOHS_FP(pkt->rootdelay)),
                           FPTOD(NTOHS_FP(pkt->rootdisp)),
                           pkt->refid,
                           rbufp->recv_length - MIN_V4_PKT_LEN, (u_char *)&pkt->exten);

                       if (AUTH(restrict_mask & RES_DONTTRUST,
                          is_authentic)) {
                               /* Bug 3596: Do we want to fuzz the reftime? */
                               fast_xmit(rbufp, MODE_SERVER, skeyid,
                                   restrict_mask);
                       } else if (is_authentic == AUTH_ERROR) {
                               /* Bug 3596: Do we want to fuzz the reftime? */
                               fast_xmit(rbufp, MODE_SERVER, 0,
                                   restrict_mask);
                               sys_badauth++;
                       } else {
                               DPRINTF(2, ("receive: AM_FXMIT drop: !mcast restricted\n"));
                               sys_restricted++;
                       }

                       return;                 /* hooray */
               }

               /*
                * This must be manycast. Do not respond if not
                * configured as a manycast server.
                */
               if (!sys_manycastserver) {
                       DPRINTF(2, ("receive: AM_FXMIT drop: Not manycastserver\n"));
                       sys_restricted++;
                       return;                 /* not enabled */
               }

#ifdef AUTOKEY
               /*
                * Do not respond if not the same group.
                */
               if (group_test(groupname, NULL)) {
                       DPRINTF(2, ("receive: AM_FXMIT drop: empty groupname\n"));
                       sys_declined++;
                       return;
               }
#endif /* AUTOKEY */

               /*
                * Do not respond if we are not synchronized or our
                * stratum is greater than the manycaster or the
                * manycaster has already synchronized to us.
                */
               if (   sys_leap == LEAP_NOTINSYNC
                   || sys_stratum >= hisstratum
                   || (!sys_cohort && sys_stratum == hisstratum + 1)
                   || rbufp->dstadr->addr_refid == pkt->refid
#           ifdef WORDS_BIGENDIAN       /* see local_refid() comment */
                   || (   IS_IPV6(&rbufp->dstadr->sin)
                       &&rbufp->dstadr->old_refid ==  pkt->refid)
#           endif
                                                                 ) {
                       DPRINTF(2, ("receive: sys leap: %0x, sys_stratum %d > hisstratum+1 %d, !sys_cohort %d && sys_stratum == hisstratum+1, loop refid %#x == pkt refid %#x\n", sys_leap, sys_stratum, hisstratum + 1, !sys_cohort, rbufp->dstadr->addr_refid, pkt->refid));
                       DPRINTF(2, ("receive: AM_FXMIT drop: LEAP_NOTINSYNC || stratum || loop\n"));
                       sys_declined++;
                       return;                 /* no help */
               }

               /*
                * Do not respond if the packet came into an IPv6 link-local
                * address on an interface where we also have a usable
                * global address, to avoid duplicate associations.
                */
               if (INT_LL_OF_GLOB & rbufp->dstadr->flags) {
                       DPRINTF(2, ("receive: declining manycast solicitation on link-local IPv6\n"));
                       sys_declined++;
                       return;
               }

               /*
                * Respond only if authentication succeeds. Don't do a
                * crypto-NAK, as that would not be useful.
                */
               if (AUTH(restrict_mask & RES_DONTTRUST, is_authentic)) {
                       record_raw_stats(&rbufp->recv_srcadr,
                           &rbufp->dstadr->sin,
                           &p_org, &p_rec, &p_xmt, &rbufp->recv_time,
                           PKT_LEAP(pkt->li_vn_mode),
                           PKT_VERSION(pkt->li_vn_mode),
                           PKT_MODE(pkt->li_vn_mode),
                           PKT_TO_STRATUM(pkt->stratum),
                           pkt->ppoll,
                           pkt->precision,
                           FPTOD(NTOHS_FP(pkt->rootdelay)),
                           FPTOD(NTOHS_FP(pkt->rootdisp)),
                           pkt->refid,
                           rbufp->recv_length - MIN_V4_PKT_LEN, (u_char *)&pkt->exten);

                       /* Bug 3596: Do we want to fuzz the reftime? */
                       fast_xmit(rbufp, MODE_SERVER, skeyid,
                           restrict_mask);
               }
               return;                         /* hooray */

       /*
        * This is a server mode packet returned in response to a client
        * mode packet sent to a multicast group address (for
        * manycastclient) or to a unicast address (for pool). The
        * origin timestamp is a good nonce to reliably associate the
        * reply with what was sent. If there is no match, that's
        * curious and could be an intruder attempting to clog, so we
        * just ignore it.
        *
        * If the packet is authentic and the manycastclient or pool
        * association is found, we mobilize a client association and
        * copy pertinent variables from the manycastclient or pool
        * association to the new client association. If not, just
        * ignore the packet.
        *
        * There is an implosion hazard at the manycast client, since
        * the manycast servers send the server packet immediately. If
        * the guy is already here, don't fire up a duplicate.
        *
        * There are cases here where we do not call record_raw_stats().
        */
       case AM_MANYCAST:

#ifdef AUTOKEY
               /*
                * Do not respond if not the same group.
                */
               if (group_test(groupname, NULL)) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: empty groupname\n"));
                       sys_declined++;
                       return;
               }
#endif /* AUTOKEY */
               /* Do not spin up duplicate manycast associations */
               if (INT_LL_OF_GLOB & rbufp->dstadr->flags) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: link-local server\n"));
                       sys_declined++;
                       return;
               }
               if ((peer2 = findmanycastpeer(rbufp)) == NULL) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: No manycast peer\n"));
                       sys_restricted++;
                       return;                 /* not enabled */
               }
               if (!AUTH(  (!(peer2->cast_flags & MDF_POOL)
                            && sys_authenticate)
                         || (restrict_mask & (RES_NOPEER |
                             RES_DONTTRUST)), is_authentic)
                   /* MC: RES_NOEPEER? */
                  ) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: bad auth || (NOPEER|DONTTRUST)\n"));
                       sys_restricted++;
                       return;                 /* access denied */
               }

               /*
                * Do not respond if unsynchronized or stratum is below
                * the floor or at or above the ceiling.
                */
               if (   hisleap == LEAP_NOTINSYNC
                   || hisstratum < sys_floor
                   || hisstratum >= sys_ceiling) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: unsync/stratum\n"));
                       sys_declined++;
                       return;                 /* no help */
               }
               cast_flags = MDF_UCAST;
               if (MDF_POOL & peer2->cast_flags) {
                       cast_flags |= MDF_PCLNT;
               }
               peer = newpeer(&rbufp->recv_srcadr, NULL, rbufp->dstadr,
                              r4a.ippeerlimit, MODE_CLIENT, hisversion,
                              peer2->minpoll, peer2->maxpoll,
                              (FLAG_PREEMPT | (POOL_FLAG_PMASK & peer2->flags)),
                              cast_flags, 0, skeyid, sys_ident);
               if (NULL == peer) {
                       DPRINTF(2, ("receive: AM_MANYCAST drop: duplicate\n"));
                       sys_declined++;
                       return;                 /* ignore duplicate */
               }

               /*
                * After each ephemeral pool association is spun,
                * accelerate the next poll for the pool solicitor so
                * the pool will fill promptly.
                */
               if (peer2->cast_flags & MDF_POOL)
                       peer2->nextdate = current_time + 1;

               /*
                * Further processing of the solicitation response would
                * simply detect its origin timestamp as bogus for the
                * brand-new association (it matches the prototype
                * association) and tinker with peer->nextdate delaying
                * first sync.
                */
               return;         /* solicitation response handled */

       /*
        * This is the first packet received from a broadcast server. If
        * the packet is authentic and we are enabled as broadcast
        * client, mobilize a broadcast client association. We don't
        * kiss any frogs here.
        *
        * There are cases here where we do not call record_raw_stats().
        */
       case AM_NEWBCL:

#ifdef AUTOKEY
               /*
                * Do not respond if not the same group.
                */
               if (group_test(groupname, sys_ident)) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: groupname mismatch\n"));
                       sys_declined++;
                       return;
               }
#endif /* AUTOKEY */
               if (!sys_bclient && !sys_mclient) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: not a bclient/mclient\n"));
                       sys_restricted++;
                       return;                 /* not enabled */
               }
               if (!AUTH(sys_authenticate | (restrict_mask &
                         (RES_NOPEER | RES_DONTTRUST)), is_authentic)
                   /* NEWBCL: RES_NOEPEER? */
                  ) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: AUTH failed\n"));
                       sys_restricted++;
                       return;                 /* access denied */
               }

               /*
                * Do not respond if unsynchronized or stratum is below
                * the floor or at or above the ceiling.
                */
               if (   hisleap == LEAP_NOTINSYNC
                   || hisstratum < sys_floor
                   || hisstratum >= sys_ceiling) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: Unsync or bad stratum\n"));
                       sys_declined++;
                       return;                 /* no help */
               }

#ifdef AUTOKEY
               /*
                * Do not respond if Autokey and the opcode is not a
                * CRYPTO_ASSOC response with association ID.
                */
               if (   crypto_flags && skeyid > NTP_MAXKEY
                   && (opcode & 0xffff0000) != (CRYPTO_ASSOC | CRYPTO_RESP)) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: Autokey but not CRYPTO_ASSOC\n"));
                       sys_declined++;
                       return;                 /* protocol error */
               }
#endif  /* AUTOKEY */

               /*
                * Broadcasts received via a multicast address may
                * arrive after a unicast volley has begun
                * with the same remote address.  newpeer() will not
                * find duplicate associations on other local endpoints
                * if a non-NULL endpoint is supplied.  multicastclient
                * ephemeral associations are unique across all local
                * endpoints.
                */
               if (!(INT_MCASTOPEN & rbufp->dstadr->flags))
                       match_ep = rbufp->dstadr;
               else
                       match_ep = NULL;

               /*
                * Determine whether to execute the initial volley.
                */
               if (sys_bdelay > 0.0) {
#ifdef AUTOKEY
                       /*
                        * If a two-way exchange is not possible,
                        * neither is Autokey.
                        */
                       if (crypto_flags && skeyid > NTP_MAXKEY) {
                               sys_restricted++;
                               DPRINTF(2, ("receive: AM_NEWBCL drop: Autokey but not 2-way\n"));
                               return;         /* no autokey */
                       }
#endif  /* AUTOKEY */

                       /*
                        * Do not execute the volley. Start out in
                        * broadcast client mode.
                        */
                       peer = newpeer(&rbufp->recv_srcadr, NULL, match_ep,
                           r4a.ippeerlimit, MODE_BCLIENT, hisversion,
                           pkt->ppoll, pkt->ppoll,
                           FLAG_PREEMPT, MDF_BCLNT, 0, skeyid, sys_ident);
                       if (NULL == peer) {
                               DPRINTF(2, ("receive: AM_NEWBCL drop: duplicate\n"));
                               sys_restricted++;
                               return;         /* ignore duplicate */

                       } else {
                               peer->delay = sys_bdelay;
                               peer->bxmt = p_xmt;
                       }
                       break;
               }

               /*
                * Execute the initial volley in order to calibrate the
                * propagation delay and run the Autokey protocol.
                *
                * Note that the minpoll is taken from the broadcast
                * packet, normally 6 (64 s) and that the poll interval
                * is fixed at this value.
                */
               peer = newpeer(&rbufp->recv_srcadr, NULL, match_ep,
                              r4a.ippeerlimit, MODE_CLIENT, hisversion,
                              pkt->ppoll, pkt->ppoll,
                              FLAG_BC_VOL | FLAG_IBURST | FLAG_PREEMPT, MDF_BCLNT,
                              0, skeyid, sys_ident);
               if (NULL == peer) {
                       DPRINTF(2, ("receive: AM_NEWBCL drop: empty newpeer() failed\n"));
                       sys_restricted++;
                       return;                 /* ignore duplicate */
               }
               peer->bxmt = p_xmt;
#ifdef AUTOKEY
               if (skeyid > NTP_MAXKEY)
                       crypto_recv(peer, rbufp);
#endif  /* AUTOKEY */

               return;                         /* hooray */

       /*
        * This is the first packet received from a potential ephemeral
        * symmetric active peer.  First, deal with broken Windows clients.
        * Then, if NOEPEER is enabled, drop it.  If the packet meets our
        * authenticty requirements and is the first he sent, mobilize
        * a passive association.
        * Otherwise, kiss the frog.
        *
        * There are cases here where we do not call record_raw_stats().
        */
       case AM_NEWPASS:

               DEBUG_REQUIRE(MODE_ACTIVE == hismode);

#ifdef AUTOKEY
               /*
                * Do not respond if not the same group.
                */
               if (group_test(groupname, sys_ident)) {
                       DPRINTF(2, ("receive: AM_NEWPASS drop: Autokey group mismatch\n"));
                       sys_declined++;
                       return;
               }
#endif /* AUTOKEY */
               if (!AUTH(sys_authenticate | (restrict_mask &
                         (RES_NOPEER | RES_DONTTRUST)), is_authentic)
                  ) {
                       /*
                        * If authenticated but cannot mobilize an
                        * association, send a symmetric passive
                        * response without mobilizing an association.
                        * This is for drat broken Windows clients. See
                        * Microsoft KB 875424 for preferred workaround.
                        */
                       if (AUTH(restrict_mask & RES_DONTTRUST,
                                is_authentic)) {
                               fast_xmit(rbufp, MODE_PASSIVE, skeyid,
                                   restrict_mask);
                               return;                 /* hooray */
                       }
                       /* HMS: Why is this next set of lines a feature? */
                       if (is_authentic == AUTH_ERROR) {
                               fast_xmit(rbufp, MODE_PASSIVE, 0,
                                   restrict_mask);
                               sys_restricted++;
                               return;
                       }

                       if (restrict_mask & RES_NOEPEER) {
                               DPRINTF(2, ("receive: AM_NEWPASS drop: NOEPEER\n"));
                               sys_declined++;
                               return;
                       }

                       /* [Bug 2941]
                        * If we got here, the packet isn't part of an
                        * existing association, either isn't correctly
                        * authenticated or it is but we are refusing
                        * ephemeral peer requests, and it didn't meet
                        * either of the previous two special cases so we
                        * should just drop it on the floor.  For example,
                        * crypto-NAKs (is_authentic == AUTH_CRYPTO)
                        * will make it this far.  This is just
                        * debug-printed and not logged to avoid log
                        * flooding.
                        */
                       DPRINTF(2, ("receive: at %ld refusing to mobilize passive association"
                                   " with unknown peer %s mode %d/%s:%s keyid %08x len %d auth %d\n",
                                   current_time, stoa(&rbufp->recv_srcadr),
                                   hismode, hm_str, am_str, skeyid,
                                   (authlen + has_mac), is_authentic));
                       sys_declined++;
                       return;
               }

               if (restrict_mask & RES_NOEPEER) {
                       DPRINTF(2, ("receive: AM_NEWPASS drop: NOEPEER\n"));
                       sys_declined++;
                       return;
               }

               /*
                * Do not respond if synchronized and if stratum is
                * below the floor or at or above the ceiling. Note,
                * this allows an unsynchronized peer to synchronize to
                * us. It would be very strange if he did and then was
                * nipped, but that could only happen if we were
                * operating at the top end of the range.  It also means
                * we will spin an ephemeral association in response to
                * MODE_ACTIVE KoDs, which will time out eventually.
                */
               if (   hisleap != LEAP_NOTINSYNC
                      && (hisstratum < sys_floor || hisstratum >= sys_ceiling)) {
                       DPRINTF(2, ("receive: AM_NEWPASS drop: Remote stratum (%d) out of range\n",
                                       hisstratum));
                       sys_declined++;
                       return;                 /* no help */
               }

               /*
                * The message is correctly authenticated and allowed.
                * Mobilize a symmetric passive association, if we won't
                * exceed the ippeerlimit.
                */
               if ((peer = newpeer(&rbufp->recv_srcadr, NULL, rbufp->dstadr,
                                   r4a.ippeerlimit, MODE_PASSIVE, hisversion,
                                   pkt->ppoll, NTP_MAXDPOLL, 0, MDF_UCAST, 0,
                                   skeyid, sys_ident)) == NULL) {
                       DPRINTF(2, ("receive: AM_NEWPASS drop: newpeer() failed\n"));
                       sys_declined++;
                       return;                 /* ignore duplicate */
               }
               break;


       /*
        * Process regular packet. Nothing special.
        *
        * There are cases here where we do not call record_raw_stats().
        */
       case AM_PROCPKT:

#ifdef AUTOKEY
               /*
                * Do not respond if not the same group.
                */
               if (group_test(groupname, peer->ident)) {
                       DPRINTF(2, ("receive: AM_PROCPKT drop: Autokey group mismatch\n"));
                       sys_declined++;
                       return;
               }
#endif /* AUTOKEY */

               if (MODE_BROADCAST == hismode) {
                       int     bail = 0;
                       l_fp    tdiff;
                       u_long  deadband;

                       DPRINTF(2, ("receive: PROCPKT/BROADCAST: prev pkt %ld seconds ago, ppoll: %d, %d secs\n",
                                   (current_time - peer->timelastrec),
                                   peer->ppoll, (1 << peer->ppoll)
                                   ));
                       /* Things we can check:
                        *
                        * Did the poll interval change?
                        * Is the poll interval in the packet in-range?
                        * Did this packet arrive too soon?
                        * Is the timestamp in this packet monotonic
                        *  with respect to the previous packet?
                        */

                       /* This is noteworthy, not error-worthy */
                       if (pkt->ppoll != peer->ppoll) {
                               msyslog(LOG_INFO, "receive: broadcast poll from %s changed from %u to %u",
                                       stoa(&rbufp->recv_srcadr),
                                       peer->ppoll, pkt->ppoll);
                       }

                       /* This is error-worthy */
                       if (   pkt->ppoll < peer->minpoll
                           || pkt->ppoll > peer->maxpoll) {
                               msyslog(LOG_INFO, "receive: broadcast poll of %u from %s is out-of-range (%d to %d)!",
                                       pkt->ppoll, stoa(&rbufp->recv_srcadr),
                                       peer->minpoll, peer->maxpoll);
                               ++bail;
                       }

                       /* too early? worth an error, too!
                        *
                        * [Bug 3113] Ensure that at least one poll
                        * interval has elapsed since the last **clean**
                        * packet was received.  We limit the check to
                        * **clean** packets to prevent replayed packets
                        * and incorrectly authenticated packets, which
                        * we'll discard, from being used to create a
                        * denial of service condition.
                        */
                       deadband = (1u << pkt->ppoll);
                       if (FLAG_BC_VOL & peer->flags)
                               deadband -= 3;  /* allow greater fuzz after volley */
                       if ((current_time - peer->timereceived) < deadband) {
                               msyslog(LOG_INFO, "receive: broadcast packet from %s arrived after %lu, not %lu seconds!",
                                       stoa(&rbufp->recv_srcadr),
                                       (current_time - peer->timereceived),
                                       deadband);
                               ++bail;
                       }

                       /* Alert if time from the server is non-monotonic.
                        *
                        * [Bug 3114] is about Broadcast mode replay DoS.
                        *
                        * Broadcast mode *assumes* a trusted network.
                        * Even so, it's nice to be robust in the face
                        * of attacks.
                        *
                        * If we get an authenticated broadcast packet
                        * with an "earlier" timestamp, it means one of
                        * two things:
                        *
                        * - the broadcast server had a backward step.
                        *
                        * - somebody is trying a replay attack.
                        *
                        * deadband: By default, we assume the broadcast
                        * network is trustable, so we take our accepted
                        * broadcast packets as we receive them.  But
                        * some folks might want to take additional poll
                        * delays before believing a backward step.
                        */
                       if (sys_bcpollbstep) {
                               /* pkt->ppoll or peer->ppoll ? */
                               deadband = (1u << pkt->ppoll)
                                          * sys_bcpollbstep + 2;
                       } else {
                               deadband = 0;
                       }

                       if (L_ISZERO(&peer->bxmt)) {
                               tdiff.l_ui = tdiff.l_uf = 0;
                       } else {
                               tdiff = p_xmt;
                               L_SUB(&tdiff, &peer->bxmt);
                       }
                       if (   tdiff.l_i < 0
                           && (current_time - peer->timereceived) < deadband)
                       {
                               msyslog(LOG_INFO, "receive: broadcast packet from %s contains non-monotonic timestamp: 0x%x.%08x -> 0x%x.%08x",
                                       stoa(&rbufp->recv_srcadr),
                                       peer->bxmt.l_ui, peer->bxmt.l_uf,
                                       p_xmt.l_ui, p_xmt.l_uf
                                       );
                               ++bail;
                       }

                       if (bail) {
                               DPRINTF(2, ("receive: AM_PROCPKT drop: bail\n"));
                               peer->timelastrec = current_time;
                               sys_declined++;
                               return;
                       }
               }

               break;

       /*
        * A passive packet matches a passive association. This is
        * usually the result of reconfiguring a client on the fly. As
        * this association might be legitimate and this packet an
        * attempt to deny service, just ignore it.
        */
       case AM_ERR:
               DPRINTF(2, ("receive: AM_ERR drop.\n"));
               sys_declined++;
               return;

       /*
        * For everything else there is the bit bucket.
        */
       default:
               DPRINTF(2, ("receive: default drop.\n"));
               sys_declined++;
               return;
       }

#ifdef AUTOKEY
       /*
        * If the association is configured for Autokey, the packet must
        * have a public key ID; if not, the packet must have a
        * symmetric key ID.
        */
       if (   is_authentic != AUTH_CRYPTO
           && (   ((peer->flags & FLAG_SKEY) && skeyid <= NTP_MAXKEY)
               || (!(peer->flags & FLAG_SKEY) && skeyid > NTP_MAXKEY))) {
               DPRINTF(2, ("receive: drop: Autokey but wrong/bad auth\n"));
               sys_badauth++;
               return;
       }
#endif  /* AUTOKEY */

       peer->received++;
       peer->flash &= ~PKT_TEST_MASK;
       if (peer->flags & FLAG_XBOGUS) {
               peer->flags &= ~FLAG_XBOGUS;
               peer->flash |= TEST3;
       }

       /*
        * Next comes a rigorous schedule of timestamp checking. If the
        * transmit timestamp is zero, the server has not initialized in
        * interleaved modes or is horribly broken.
        *
        * A KoD packet we pay attention to cannot have a 0 transmit
        * timestamp.
        */

       kissCode = kiss_code_check(hisleap, hisstratum, hismode, pkt->refid);

       if (L_ISZERO(&p_xmt)) {
               peer->flash |= TEST3;                   /* unsynch */
               if (kissCode != NOKISS) {               /* KoD packet */
                       peer->bogusorg++;               /* for TEST2 or TEST3 */
                       msyslog(LOG_INFO,
                               "receive: Unexpected zero transmit timestamp in KoD from %s",
                               ntoa(&peer->srcadr));
                       return;
               }

       /*
        * If the transmit timestamp duplicates our previous one, the
        * packet is a replay. This prevents the bad guys from replaying
        * the most recent packet, authenticated or not.
        */
       } else if (   ((FLAG_LOOPNONCE & peer->flags) && L_ISEQU(&peer->nonce, &p_xmt))
                  || (!(FLAG_LOOPNONCE & peer->flags) && L_ISEQU(&peer->xmt, &p_xmt))
       ) {
               DPRINTF(2, ("receive: drop: Duplicate xmit\n"));
               peer->flash |= TEST1;                   /* duplicate */
               peer->oldpkt++;
               return;

       /*
        * If this is a broadcast mode packet, make sure hisstratum
        * is appropriate.  Don't do anything else here - we wait to
        * see if this is an interleave broadcast packet until after
        * we've validated the MAC that SHOULD be provided.
        *
        * hisstratum cannot be 0 - see assertion above.
        * If hisstratum is 15, then we'll advertise as UNSPEC but
        * at least we'll be able to sync with the broadcast server.
        */
       } else if (hismode == MODE_BROADCAST) {
               /* 0 is unexpected too, and impossible */
               if (STRATUM_UNSPEC <= hisstratum) {
                       /* Is this a ++sys_declined or ??? */
                       msyslog(LOG_INFO,
                               "receive: Unexpected stratum (%d) in broadcast from %s",
                               hisstratum, ntoa(&peer->srcadr));
                       return;
               }

       /*
        * Basic KoD validation checking:
        *
        * KoD packets are a mixed-blessing.  Forged KoD packets
        * are DoS attacks.  There are rare situations where we might
        * get a valid KoD response, though.  Since KoD packets are
        * a special case that complicate the checks we do next, we
        * handle the basic KoD checks here.
        *
        * Note that we expect the incoming KoD packet to have its
        * (nonzero) org, rec, and xmt timestamps set to the xmt timestamp
        * that we have previously sent out.  Watch interleave mode.
        */
       } else if (kissCode != NOKISS) {
               DEBUG_INSIST(!L_ISZERO(&p_xmt));
               if (   L_ISZERO(&p_org)         /* We checked p_xmt above */
                   || L_ISZERO(&p_rec)) {
                       peer->bogusorg++;
                       msyslog(LOG_INFO,
                               "receive: KoD packet from %s has a zero org or rec timestamp.  Ignoring.",
                               ntoa(&peer->srcadr));
                       return;
               }

               if (   !L_ISEQU(&p_xmt, &p_org)
                   || !L_ISEQU(&p_xmt, &p_rec)) {
                       peer->bogusorg++;
                       msyslog(LOG_INFO,
                               "receive: KoD packet from %s has inconsistent xmt/org/rec timestamps.  Ignoring.",
                               ntoa(&peer->srcadr));
                       return;
               }

               /* Be conservative */
               if (peer->flip == 0 && !L_ISEQU(&p_org, &peer->aorg)) {
                       peer->bogusorg++;
                       msyslog(LOG_INFO,
                               "receive: flip 0 KoD origin timestamp 0x%x.%08x from %s does not match 0x%x.%08x - ignoring.",
                               p_org.l_ui, p_org.l_uf,
                               ntoa(&peer->srcadr),
                               peer->aorg.l_ui, peer->aorg.l_uf);
                       return;
               } else if (peer->flip == 1 && !L_ISEQU(&p_org, &peer->borg)) {
                       peer->bogusorg++;
                       msyslog(LOG_INFO,
                               "receive: flip 1 KoD origin timestamp 0x%x.%08x from %s does not match interleave 0x%x.%08x - ignoring.",
                               p_org.l_ui, p_org.l_uf,
                               ntoa(&peer->srcadr),
                               peer->borg.l_ui, peer->borg.l_uf);
                       return;
               }

               /*
                * Basic mode checks:
                *
                * If there is no origin timestamp, it's either an initial
                * packet or we've already received a response to our query.
                * Of course, should 'aorg' be all-zero because this really
                * was the original transmit timestamp, we'll ignore this
                * reply.  There is a window of one nanosecond once every
                * 136 years' time where this is possible.  We currently
                * ignore this situation, as a completely zero timestamp
                * is (quietly?) disallowed.
                *
                * Otherwise, check for bogus packet in basic mode.
                * If it is bogus, switch to interleaved mode and
                * resynchronize, but only after confirming the packet is
                * not bogus in symmetric interleaved mode.
                *
                * This could also mean somebody is forging packets claiming
                * to be from us, attempting to cause our server to KoD us.
                *
                * We have earlier asserted that hisstratum cannot be 0.
                * If hisstratum is STRATUM_UNSPEC, it means he's not sync'd.
                */

               /* XXX: FLAG_LOOPNONCE */
               DEBUG_INSIST(0 == (FLAG_LOOPNONCE & peer->flags));

               if (RATEKISS == kissCode) {
                       msyslog(LOG_INFO, "RATE KoD from %s poll %u",
                               ntoa(&peer->srcadr), 1u << pkt->ppoll);
               } else {
                       msyslog(LOG_INFO, "KoD %s from %s",
                               refid_str(pkt->refid, -1),
                               ntoa(&peer->srcadr));
               }
       } else if (peer->flip == 0) {
               if (0) {
               } else if (L_ISZERO(&p_org)) {
                       const char *action;

#ifdef BUG3361
                       msyslog(LOG_INFO,
                               "receive: BUG 3361: Clearing peer->aorg ");
                       L_CLR(&peer->aorg);
                       /* Clear peer->nonce, too? */
#endif
                       /**/
                       switch (hismode) {
                               /* We allow 0org for: */
                           case UCHAR_MAX:
                               action = "Allow";
                               break;
                               /* We disallow 0org for: */
                           case MODE_UNSPEC:
                           case MODE_ACTIVE:
                           case MODE_PASSIVE:
                           case MODE_CLIENT:
                           case MODE_SERVER:
                           case MODE_BROADCAST:
                               action = "Drop";
                               peer->bogusorg++;
                               peer->flash |= TEST2;   /* bogus */
                               break;
                           default:
                               action = "";    /* for cranky compilers / MSVC */
                               INSIST(!"receive(): impossible hismode");
                               break;
                       }
                       /**/
                       msyslog(LOG_INFO,
                               "receive: %s 0 origin timestamp from %s@%s xmt 0x%x.%08x",
                               action, hm_str, ntoa(&peer->srcadr),
                               ntohl(pkt->xmt.l_ui), ntohl(pkt->xmt.l_uf));
               } else if (   L_ISZERO(&peer->aorg) && MODE_CLIENT != hismode
                          && !memcmp("STEP", &peer->refid, 4)) {
                       /* response came in just after we stepped clock, normal */
               } else if (!L_ISEQU(&p_org, &peer->aorg)) {
                       /* are there cases here where we should bail? */
                       /* Should we set TEST2 if we decide to try xleave? */
                       peer->bogusorg++;
                       peer->flash |= TEST2;   /* bogus */
                       msyslog(LOG_INFO,
                               "duplicate or replay: org 0x%x.%08x does not match 0x%x.%08x from %s@%s",
                               ntohl(pkt->org.l_ui), ntohl(pkt->org.l_uf),
                               peer->aorg.l_ui, peer->aorg.l_uf,
                               hm_str, ntoa(&peer->srcadr));
                       if (  !L_ISZERO(&peer->dst)
                           && L_ISEQU(&p_org, &peer->dst)) {
                               /* Might be the start of an interleave */
                               if (dynamic_interleave) {
                                       peer->flip = 1;
                                       report_event(PEVNT_XLEAVE, peer, NULL);
                               } else {
                                       msyslog(LOG_INFO,
                                               "receive: Dynamic interleave from %s@%s denied",
                                               hm_str, ntoa(&peer->srcadr));
                               }
                       }
               } else {
                       L_CLR(&peer->aorg);
                       /* XXX: FLAG_LOOPNONCE */
               }

       /*
        * Check for valid nonzero timestamp fields.
        */
       } else if (   L_ISZERO(&p_org)
                  || L_ISZERO(&p_rec)
                  || L_ISZERO(&peer->dst)) {
               peer->flash |= TEST3;           /* unsynch */

       /*
        * Check for bogus packet in interleaved symmetric mode. This
        * can happen if a packet is lost, duplicated or crossed. If
        * found, flip and resynchronize.
        */
       } else if (   !L_ISZERO(&peer->dst)
                  && !L_ISEQU(&p_org, &peer->dst)) {
               DPRINTF(2, ("receive: drop: Bogus packet in interleaved symmetric mode\n"));
               peer->bogusorg++;
               peer->flags |= FLAG_XBOGUS;
               peer->flash |= TEST2;           /* bogus */
#ifdef BUG3453
               return; /* Bogus packet, we are done */
#endif
       }

       /**/

       /*
        * If this is a crypto_NAK, the server cannot authenticate a
        * client packet. The server might have just changed keys. Clear
        * the association and restart the protocol.
        */
       if (crypto_nak_test == VALIDNAK) {
               report_event(PEVNT_AUTH, peer, "crypto_NAK");
               peer->flash |= TEST5;           /* bad auth */
               peer->badauth++;
               if (peer->flags & FLAG_PREEMPT) {
                       if (unpeer_crypto_nak_early) {
                               unpeer(peer);
                       }
                       DPRINTF(2, ("receive: drop: PREEMPT crypto_NAK\n"));
                       return;
               }
#ifdef AUTOKEY
               if (peer->crypto) {
                       peer_clear(peer, "AUTH");
               }
#endif  /* AUTOKEY */
               DPRINTF(2, ("receive: drop: crypto_NAK\n"));
               return;

       /*
        * If the digest fails or it's missing for authenticated
        * associations, the client cannot authenticate a server
        * reply to a client packet previously sent. The loopback check
        * is designed to avoid a bait-and-switch attack, which was
        * possible in past versions. If symmetric modes, return a
        * crypto-NAK. The peer should restart the protocol.
        */
       } else if (!AUTH(peer->keyid || has_mac ||
                        (restrict_mask & RES_DONTTRUST), is_authentic)) {

               if (peer->flash & PKT_TEST_MASK) {
                       msyslog(LOG_INFO,
                               "receive: Bad auth in packet with bad timestamps from %s denied - spoof?",
                               ntoa(&peer->srcadr));
                       return;
               }

               report_event(PEVNT_AUTH, peer, "digest");
               peer->flash |= TEST5;           /* bad auth */
               peer->badauth++;
               if (   has_mac
                   && (   hismode == MODE_ACTIVE
                       || hismode == MODE_PASSIVE))
                       fast_xmit(rbufp, MODE_ACTIVE, 0, restrict_mask);
               if (peer->flags & FLAG_PREEMPT) {
                       if (unpeer_digest_early) {
                               unpeer(peer);
                       }
               }
#ifdef AUTOKEY
               else if (peer_clear_digest_early && peer->crypto) {
                       peer_clear(peer, "AUTH");
               }
#endif  /* AUTOKEY */
               DPRINTF(2, ("receive: drop: Bad or missing AUTH\n"));
               return;
       }

       /*
        * For broadcast packets:
        *
        * HMS: This next line never made much sense to me, even
        * when it was up higher:
        *   If an initial volley, bail out now and let the
        *   client do its stuff.
        *
        * If the packet has not failed authentication, then
        * - if the origin timestamp is nonzero this is an
        *   interleaved broadcast, so restart the protocol.
        * - else, this is not an interleaved broadcast packet.
        */
       if (hismode == MODE_BROADCAST) {
               if (   is_authentic == AUTH_OK
                   || is_authentic == AUTH_NONE) {
                       if (!L_ISZERO(&p_org)) {
                               if (!(peer->flags & FLAG_XB)) {
                                       msyslog(LOG_INFO,
                                               "receive: Broadcast server at %s is in interleave mode",
                                               ntoa(&peer->srcadr));
                                       peer->flags |= FLAG_XB;
                                       peer->aorg = p_xmt;
                                       peer->borg = rbufp->recv_time;
                                       report_event(PEVNT_XLEAVE, peer, NULL);
                                       return;
                               }
                       } else if (peer->flags & FLAG_XB) {
                               msyslog(LOG_INFO,
                                       "receive: Broadcast server at %s is no longer in interleave mode",
                                       ntoa(&peer->srcadr));
                               peer->flags &= ~FLAG_XB;
                       }
               } else {
                       msyslog(LOG_INFO,
                               "receive: Bad broadcast auth (%d) from %s",
                               is_authentic, ntoa(&peer->srcadr));
               }

               /*
                * Now that we know the packet is correctly authenticated,
                * update peer->bxmt.
                */
               peer->bxmt = p_xmt;
       }


       /*
       ** Update the state variables.
       */
       if (peer->flip == 0) {
               if (hismode != MODE_BROADCAST)
                       peer->rec = p_xmt;
               peer->dst = rbufp->recv_time;
       }
       peer->xmt = p_xmt;

       /*
        * Set the peer ppoll to the maximum of the packet ppoll and the
        * peer minpoll. If a kiss-o'-death, set the peer minpoll to
        * this maximum and advance the headway to give the sender some
        * headroom. Very intricate.
        */

       /*
        * Check for any kiss codes. Note this is only used when a server
        * responds to a packet request.
        */

       /*
        * Check to see if this is a RATE Kiss Code
        * Currently this kiss code will accept whatever valid poll
        * rate that the server sends
        */
       if (   (NTP_MINPOLL > pkt->ppoll)
           || (NTP_MAXPOLL < pkt->ppoll)
          ) {
               DPRINTF(2, ("RATEKISS: Invalid ppoll (%d) from %s\n",
                               pkt->ppoll, stoa(&rbufp->recv_srcadr)));
               sys_badlength++;
               return;                 /* invalid packet poll */
       }
       peer->ppoll = max(peer->minpoll, pkt->ppoll);
       if (kissCode == RATEKISS) {
               peer->selbroken++;      /* Increment the KoD count */
               report_event(PEVNT_RATE, peer, NULL);
               if (pkt->ppoll > peer->minpoll)
                       peer->minpoll = peer->ppoll;
               peer->burst = peer->retry = 0;
               peer->throttle = (NTP_SHIFT + 1) * (1 << peer->minpoll);
               poll_update(peer, pkt->ppoll, 0);
               return;                         /* kiss-o'-death */
       }
       if (kissCode != NOKISS) {
               peer->selbroken++;      /* Increment the KoD count */
               return;         /* Drop any other kiss code packets */
       }


       /*
        * XXX
        */


       /*
        * If:
        *      - this is a *cast (uni-, broad-, or m-) server packet
        *      - and it's symmetric-key authenticated
        * then see if the sender's IP is trusted for this keyid.
        * If it is, great - nothing special to do here.
        * Otherwise, we should report and bail.
        *
        * Autokey-authenticated packets are accepted.
        */

       switch (hismode) {
           case MODE_SERVER:           /* server mode */
           case MODE_BROADCAST:        /* broadcast mode */
           case MODE_ACTIVE:           /* symmetric active mode */
           case MODE_PASSIVE:          /* symmetric passive mode */
               if (   is_authentic == AUTH_OK
                   && skeyid
                   && skeyid <= NTP_MAXKEY
                   && !authistrustedip(skeyid, &peer->srcadr)) {
                       report_event(PEVNT_AUTH, peer, "authIP");
                       peer->badauth++;
                       return;
               }
               break;

           case MODE_CLIENT:           /* client mode */
#if 0           /* At this point, MODE_CONTROL is overloaded by MODE_BCLIENT */
           case MODE_CONTROL:          /* control mode */
#endif
           case MODE_PRIVATE:          /* private mode */
           case MODE_BCLIENT:          /* broadcast client mode */
               break;

           case MODE_UNSPEC:           /* unspecified (old version) */
           default:
               msyslog(LOG_INFO,
                       "receive: Unexpected mode (%d) in packet from %s",
                       hismode, ntoa(&peer->srcadr));
               break;
       }


       /*
        * That was hard and I am sweaty, but the packet is squeaky
        * clean. Get on with real work.
        */
       peer->timereceived = current_time;
       peer->timelastrec = current_time;
       if (is_authentic == AUTH_OK)
               peer->flags |= FLAG_AUTHENTIC;
       else
               peer->flags &= ~FLAG_AUTHENTIC;

#ifdef AUTOKEY
       /*
        * More autokey dance. The rules of the cha-cha are as follows:
        *
        * 1. If there is no key or the key is not auto, do nothing.
        *
        * 2. If this packet is in response to the one just previously
        *    sent or from a broadcast server, do the extension fields.
        *    Otherwise, assume bogosity and bail out.
        *
        * 3. If an extension field contains a verified signature, it is
        *    self-authenticated and we sit the dance.
        *
        * 4. If this is a server reply, check only to see that the
        *    transmitted key ID matches the received key ID.
        *
        * 5. Check to see that one or more hashes of the current key ID
        *    matches the previous key ID or ultimate original key ID
        *    obtained from the broadcaster or symmetric peer. If no
        *    match, sit the dance and call for new autokey values.
        *
        * In case of crypto error, fire the orchestra, stop dancing and
        * restart the protocol.
        */
       if (peer->flags & FLAG_SKEY) {
               /*
                * Decrement remaining autokey hashes. This isn't
                * perfect if a packet is lost, but results in no harm.
                */
               ap = (struct autokey *)peer->recval.ptr;
               if (ap != NULL) {
                       if (ap->seq > 0)
                               ap->seq--;
               }
               peer->flash |= TEST8;
               rval = crypto_recv(peer, rbufp);
               if (rval == XEVNT_OK) {
                       peer->unreach = 0;
               } else {
                       if (rval == XEVNT_ERR) {
                               report_event(PEVNT_RESTART, peer,
                                   "crypto error");
                               peer_clear(peer, "CRYP");
                               peer->flash |= TEST9;   /* bad crypt */
                               if (peer->flags & FLAG_PREEMPT) {
                                       if (unpeer_crypto_early) {
                                               unpeer(peer);
                                       }
                               }
                       }
                       return;
               }

               /*
                * If server mode, verify the receive key ID matches
                * the transmit key ID.
                */
               if (hismode == MODE_SERVER) {
                       if (skeyid == peer->keyid)
                               peer->flash &= ~TEST8;

               /*
                * If an extension field is present, verify only that it
                * has been correctly signed. We don't need a sequence
                * check here, but the sequence continues.
                */
               } else if (!(peer->flash & TEST8)) {
                       peer->pkeyid = skeyid;

               /*
                * Now the fun part. Here, skeyid is the current ID in
                * the packet, pkeyid is the ID in the last packet and
                * tkeyid is the hash of skeyid. If the autokey values
                * have not been received, this is an automatic error.
                * If so, check that the tkeyid matches pkeyid. If not,
                * hash tkeyid and try again. If the number of hashes
                * exceeds the number remaining in the sequence, declare
                * a successful failure and refresh the autokey values.
                */
               } else if (ap != NULL) {
                       int i;

                       for (i = 0; ; i++) {
                               if (   tkeyid == peer->pkeyid
                                   || tkeyid == ap->key) {
                                       peer->flash &= ~TEST8;
                                       peer->pkeyid = skeyid;
                                       ap->seq -= i;
                                       break;
                               }
                               if (i > ap->seq) {
                                       peer->crypto &=
                                           ~CRYPTO_FLAG_AUTO;
                                       break;
                               }
                               tkeyid = session_key(
                                   &rbufp->recv_srcadr, dstadr_sin,
                                   tkeyid, pkeyid, 0);
                       }
                       if (peer->flash & TEST8)
                               report_event(PEVNT_AUTH, peer, "keylist");
               }
               if (!(peer->crypto & CRYPTO_FLAG_PROV)) /* test 9 */
                       peer->flash |= TEST8;   /* bad autokey */

               /*
                * The maximum lifetime of the protocol is about one
                * week before restarting the Autokey protocol to
                * refresh certificates and leapseconds values.
                */
               if (current_time > peer->refresh) {
                       report_event(PEVNT_RESTART, peer,
                           "crypto refresh");
                       peer_clear(peer, "TIME");
                       return;
               }
       }
#endif  /* AUTOKEY */

       /*
        * The dance is complete and the flash bits have been lit. Toss
        * the packet over the fence for processing, which may light up
        * more flashers. Leave if the packet is not good.
        */
       process_packet(peer, pkt, rbufp->recv_length);
       /* Bug 2734: TEST3 prevents initial interleave sync */
       if ((~TEST3 & peer->flash) & PKT_TEST_MASK) {
               return;
       }

       /* [bug 3592] Update poll. Ideally this should not happen in a
        * receive branch, but too much is going on here... at least we
        * do it only if the packet was good!
        */
       poll_update(peer, peer->hpoll, (peer->hmode == MODE_CLIENT));

       /*
        * In interleaved mode update the state variables. Also adjust the
        * transmit phase to avoid crossover.
        */
       if (peer->flip != 0) {
               peer->rec = p_rec;
               peer->dst = rbufp->recv_time;
               if (peer->nextdate - current_time < (1U << min(peer->ppoll,
                   peer->hpoll)) / 2)
                       peer->nextdate++;
               else
                       peer->nextdate--;
       }
}


/*
* process_packet - Packet Procedure, a la Section 3.4.4 of RFC-1305
*      Or almost, at least.  If we're in here we have a reasonable
*      expectation that we will be having a long term
*      relationship with this host.
*/
void
process_packet(
       register struct peer *peer,
       register struct pkt *pkt,
       u_int   len
       )
{
       double  t34, t21;
       double  p_offset, p_del, p_disp;
       l_fp    p_rec, p_xmt, p_org, p_reftime, ci;
       u_char  pmode, pleap, pversion, pstratum;
#ifdef ASSYM
       int     itemp;
       double  etemp, ftemp, td;
#endif /* ASSYM */

       p_del = FPTOD(NTOHS_FP(pkt->rootdelay));
       p_offset = 0;
       p_disp = FPTOD(NTOHS_FP(pkt->rootdisp));
       NTOHL_FP(&pkt->reftime, &p_reftime);
       NTOHL_FP(&pkt->org, &p_org);
       NTOHL_FP(&pkt->rec, &p_rec);
       NTOHL_FP(&pkt->xmt, &p_xmt);
       pmode = PKT_MODE(pkt->li_vn_mode);
       pleap = PKT_LEAP(pkt->li_vn_mode);
       pversion = PKT_VERSION(pkt->li_vn_mode);
       pstratum = PKT_TO_STRATUM(pkt->stratum);

       /*
        * Verify the server is synchronized; that is, the leap bits,
        * stratum and root distance are valid.
        */
       if (   pleap == LEAP_NOTINSYNC          /* test 6 */
           || pstratum < sys_floor || pstratum >= sys_ceiling)
               peer->flash |= TEST6;           /* bad synch or strat */
       if (p_del / 2 + p_disp >= MAXDISPERSE)  /* test 7 */
               peer->flash |= TEST7;           /* bad header */

       /*
        * If any tests fail at this point, the packet is discarded.
        * Note that some flashers may have already been set in the
        * receive() routine.
        */
       if (peer->flash & PKT_TEST_MASK) {
               peer->seldisptoolarge++;
               DPRINTF(1, ("packet: flash header %04x\n",
                           peer->flash));
               /* [Bug 3592] do *not* update poll on bad packets! */
               return;
       }

       /*
        * update stats, now that we really handle this packet:
        */
       sys_processed++;
       peer->processed++;

       /*
        * Capture the header values in the client/peer association..
        */
       record_raw_stats(&peer->srcadr,
           peer->dstadr ? &peer->dstadr->sin : NULL,
           &p_org, &p_rec, &p_xmt, &peer->dst,
           pleap, pversion, pmode, pstratum, pkt->ppoll, pkt->precision,
           p_del, p_disp, pkt->refid,
           len - MIN_V4_PKT_LEN, (u_char *)&pkt->exten);
       peer->leap = pleap;
       peer->stratum = min(pstratum, STRATUM_UNSPEC);
       peer->pmode = pmode;
       peer->precision = pkt->precision;
       peer->rootdelay = p_del;
       peer->rootdisp = p_disp;
       peer->refid = pkt->refid;               /* network byte order */
       peer->reftime = p_reftime;

       /*
        * First, if either burst mode is armed, enable the burst.
        * Compute the headway for the next packet and delay if
        * necessary to avoid exceeding the threshold.
        */
       if (peer->retry > 0) {
               peer->retry = 0;
               if (peer->reach)
                       peer->burst = min(1 << (peer->hpoll -
                           peer->minpoll), NTP_SHIFT) - 1;
               else
                       peer->burst = NTP_IBURST - 1;
               if (peer->burst > 0)
                       peer->nextdate = current_time;
       }

       /*
        * If the peer was previously unreachable, raise a trap. In any
        * case, mark it reachable.
        */
       if (!peer->reach) {
               report_event(PEVNT_REACH, peer, NULL);
               peer->timereachable = current_time;
       }
       peer->reach |= 1;

       /*
        * For a client/server association, calculate the clock offset,
        * roundtrip delay and dispersion. The equations are reordered
        * from the spec for more efficient use of temporaries. For a
        * broadcast association, offset the last measurement by the
        * computed delay during the client/server volley. Note the
        * computation of dispersion includes the system precision plus
        * that due to the frequency error since the origin time.
        *
        * It is very important to respect the hazards of overflow. The
        * only permitted operation on raw timestamps is subtraction,
        * where the result is a signed quantity spanning from 68 years
        * in the past to 68 years in the future. To avoid loss of
        * precision, these calculations are done using 64-bit integer
        * arithmetic. However, the offset and delay calculations are
        * sums and differences of these first-order differences, which
        * if done using 64-bit integer arithmetic, would be valid over
        * only half that span. Since the typical first-order
        * differences are usually very small, they are converted to 64-
        * bit doubles and all remaining calculations done in floating-
        * double arithmetic. This preserves the accuracy while
        * retaining the 68-year span.
        *
        * There are three interleaving schemes, basic, interleaved
        * symmetric and interleaved broadcast. The timestamps are
        * idioscyncratically different. See the onwire briefing/white
        * paper at www.eecis.udel.edu/~mills for details.
        *
        * Interleaved symmetric mode
        * t1 = peer->aorg/borg, t2 = peer->rec, t3 = p_xmt,
        * t4 = peer->dst
        */
       if (peer->flip != 0) {
               ci = p_xmt;                             /* t3 - t4 */
               L_SUB(&ci, &peer->dst);
               LFPTOD(&ci, t34);
               ci = p_rec;                             /* t2 - t1 */
               if (peer->flip > 0)
                       L_SUB(&ci, &peer->borg);
               else
                       L_SUB(&ci, &peer->aorg);
               LFPTOD(&ci, t21);
               p_del = t21 - t34;
               p_offset = (t21 + t34) / 2.;
               if (p_del < 0 || p_del > 1.) {
                       mprintf_event(PEVNT_XERR, peer,
                                     "t21 %.9f t34 %.9f", t21, t34);
                       return;
               }

       /*
        * Broadcast modes
        */
       } else if (peer->pmode == MODE_BROADCAST) {

               /*
                * Interleaved broadcast mode. Use interleaved timestamps.
                * t1 = peer->borg, t2 = p_org, t3 = p_org, t4 = aorg
                */
               if (peer->flags & FLAG_XB) {
                       ci = p_org;                     /* delay */
                       L_SUB(&ci, &peer->aorg);
                       LFPTOD(&ci, t34);
                       ci = p_org;                     /* t2 - t1 */
                       L_SUB(&ci, &peer->borg);
                       LFPTOD(&ci, t21);
                       peer->aorg = p_xmt;
                       peer->borg = peer->dst;
                       if (t34 < 0 || t34 > 1.) {
                               /* drop all if in the initial volley */
                               if (FLAG_BC_VOL & peer->flags)
                                       goto bcc_init_volley_fail;
                               mprintf_event(PEVNT_XERR, peer,
                                             "offset %.9f delay %.9f",
                                             t21, t34);
                               return;
                       }
                       p_offset = t21;
                       peer->xleave = t34;

               /*
                * Basic broadcast - use direct timestamps.
                * t3 = p_xmt, t4 = peer->dst
                */
               } else {
                       ci = p_xmt;             /* t3 - t4 */
                       L_SUB(&ci, &peer->dst);
                       LFPTOD(&ci, t34);
                       p_offset = t34;
               }

               /*
                * When calibration is complete and the clock is
                * synchronized, the bias is calculated as the difference
                * between the unicast timestamp and the broadcast
                * timestamp. This works for both basic and interleaved
                * modes.
                * [Bug 3031] Don't keep this peer when the delay
                * calculation gives reason to suspect clock steps.
                * This is assumed for delays > 50ms.
                */
               if (FLAG_BC_VOL & peer->flags) {
                       peer->flags &= ~FLAG_BC_VOL;
                       peer->delay = fabs(peer->offset - p_offset) * 2;
                       DPRINTF(2, ("broadcast volley: initial delay=%.6f\n",
                               peer->delay));
                       if (peer->delay > fabs(sys_bdelay)) {
               bcc_init_volley_fail:
                               DPRINTF(2, ("%s", "broadcast volley: initial delay exceeds limit\n"));
                               unpeer(peer);
                               return;
                       }
               }
               peer->nextdate = current_time + (1u << peer->ppoll) - 2u;
               p_del = peer->delay;
               p_offset += p_del / 2;


       /*
        * Basic mode, otherwise known as the old fashioned way.
        *
        * t1 = p_org, t2 = p_rec, t3 = p_xmt, t4 = peer->dst
        */
       } else {
               ci = p_xmt;                             /* t3 - t4 */
               L_SUB(&ci, &peer->dst);
               LFPTOD(&ci, t34);
               ci = p_rec;                             /* t2 - t1 */
               L_SUB(&ci, &p_org);
               LFPTOD(&ci, t21);
               p_del = fabs(t21 - t34);
               p_offset = (t21 + t34) / 2.;
       }
       p_del = max(p_del, LOGTOD(sys_precision));
       p_disp = LOGTOD(sys_precision) + LOGTOD(peer->precision) +
           clock_phi * p_del;

#if ASSYM
       /*
        * This code calculates the outbound and inbound data rates by
        * measuring the differences between timestamps at different
        * packet lengths. This is helpful in cases of large asymmetric
        * delays commonly experienced on deep space communication
        * links.
        */
       if (peer->t21_last > 0 && peer->t34_bytes > 0) {
               itemp = peer->t21_bytes - peer->t21_last;
               if (itemp > 25) {
                       etemp = t21 - peer->t21;
                       if (fabs(etemp) > 1e-6) {
                               ftemp = itemp / etemp;
                               if (ftemp > 1000.)
                                       peer->r21 = ftemp;
                       }
               }
               itemp = len - peer->t34_bytes;
               if (itemp > 25) {
                       etemp = -t34 - peer->t34;
                       if (fabs(etemp) > 1e-6) {
                               ftemp = itemp / etemp;
                               if (ftemp > 1000.)
                                       peer->r34 = ftemp;
                       }
               }
       }

       /*
        * The following section compensates for different data rates on
        * the outbound (d21) and inbound (t34) directions. To do this,
        * it finds t such that r21 * t - r34 * (d - t) = 0, where d is
        * the roundtrip delay. Then it calculates the correction as a
        * fraction of d.
        */
       peer->t21 = t21;
       peer->t21_last = peer->t21_bytes;
       peer->t34 = -t34;
       peer->t34_bytes = len;
       DPRINTF(2, ("packet: t21 %.9lf %d t34 %.9lf %d\n", peer->t21,
                   peer->t21_bytes, peer->t34, peer->t34_bytes));
       if (peer->r21 > 0 && peer->r34 > 0 && p_del > 0) {
               if (peer->pmode != MODE_BROADCAST)
                       td = (peer->r34 / (peer->r21 + peer->r34) -
                           .5) * p_del;
               else
                       td = 0;

               /*
                * Unfortunately, in many cases the errors are
                * unacceptable, so for the present the rates are not
                * used. In future, we might find conditions where the
                * calculations are useful, so this should be considered
                * a work in progress.
                */
               t21 -= td;
               t34 -= td;
               DPRINTF(2, ("packet: del %.6lf r21 %.1lf r34 %.1lf %.6lf\n",
                           p_del, peer->r21 / 1e3, peer->r34 / 1e3,
                           td));
       }
#endif /* ASSYM */

       /*
        * That was awesome. Now hand off to the clock filter.
        */
       clock_filter(peer, p_offset + peer->bias, p_del, p_disp);

       /*
        * If we are in broadcast calibrate mode, return to broadcast
        * client mode when the client is fit and the autokey dance is
        * complete.
        */
       if (   (FLAG_BC_VOL & peer->flags)
           && MODE_CLIENT == peer->hmode
           && !(TEST11 & peer_unfit(peer))) {  /* distance exceeded */
#ifdef AUTOKEY
               if (peer->flags & FLAG_SKEY) {
                       if (!(~peer->crypto & CRYPTO_FLAG_ALL))
                               peer->hmode = MODE_BCLIENT;
               } else {
                       peer->hmode = MODE_BCLIENT;
               }
#else   /* !AUTOKEY follows */
               peer->hmode = MODE_BCLIENT;
#endif  /* !AUTOKEY */
       }
}


/*
* clock_update - Called at system process update intervals.
*/
static void
clock_update(
       struct peer *peer       /* peer structure pointer */
       )
{
       double  dtemp;
       l_fp    now;
#ifdef HAVE_LIBSCF_H
       char    *fmri;
#endif /* HAVE_LIBSCF_H */

       /*
        * Update the system state variables. We do this very carefully,
        * as the poll interval might need to be clamped differently.
        */
       sys_peer = peer;
       sys_epoch = peer->epoch;
       if (sys_poll < peer->minpoll)
               sys_poll = peer->minpoll;
       if (sys_poll > peer->maxpoll)
               sys_poll = peer->maxpoll;
       poll_update(peer, sys_poll, 0);
       sys_stratum = min(peer->stratum + 1, STRATUM_UNSPEC);
       if (   peer->stratum == STRATUM_REFCLOCK
           || peer->stratum == STRATUM_UNSPEC)
               sys_refid = peer->refid;
       else
               sys_refid = addr2refid(&peer->srcadr);
       /*
        * Root Dispersion (E) is defined (in RFC 5905) as:
        *
        * E = p.epsilon_r + p.epsilon + p.psi + PHI*(s.t - p.t) + |THETA|
        *
        * where:
        *  p.epsilon_r is the PollProc's root dispersion
        *  p.epsilon   is the PollProc's dispersion
        *  p.psi       is the PollProc's jitter
        *  THETA       is the combined offset
        *
        * NB: Think Hard about where these numbers come from and
        * what they mean.  When did peer->update happen?  Has anything
        * interesting happened since then?  What values are the most
        * defensible?  Why?
        *
        * DLM thinks this equation is probably the best of all worse choices.
        */
       dtemp   = peer->rootdisp
               + peer->disp
               + sys_jitter
               + clock_phi * (current_time - peer->update)
               + fabs(sys_offset);

       p2_rootdisp = prev_rootdisp;
       prev_rootdisp = sys_rootdisp;
       if (dtemp > sys_mindisp)
               sys_rootdisp = dtemp;
       else
               sys_rootdisp = sys_mindisp;

       sys_rootdelay = peer->delay + peer->rootdelay;

       p2_reftime = prev_reftime;
       p2_time = prev_time;

       prev_reftime = sys_reftime;
       prev_time = current_time + 64 + (rand() & 0x3f);        /* 64-127 s */

       sys_reftime = peer->dst;

       DPRINTF(1, ("clock_update: at %lu sample %lu associd %d\n",
                   current_time, peer->epoch, peer->associd));

       /*
        * Comes now the moment of truth. Crank the clock discipline and
        * see what comes out.
        */
       switch (local_clock(peer, sys_offset)) {

       /*
        * Clock exceeds panic threshold. Life as we know it ends.
        */
       case -1:
               msyslog(LOG_ERR, "Clock offset exceeds panic threshold.");
#ifdef HAVE_LIBSCF_H
               /*
                * For Solaris enter the maintenance mode.
                */
               if ((fmri = getenv("SMF_FMRI")) != NULL) {
                       if (smf_maintain_instance(fmri, 0) < 0) {
                               msyslog(LOG_ERR, "smf_maintain_instance: %s",
                                                scf_strerror(scf_error()));
                               exit(1);
                       }
                       /*
                        * Sleep until SMF kills us.
                        */
                       msyslog(LOG_ERR, "%s placed into maintenance. "
                               "Set system clock by hand before clearing.",
                               fmri);
                       for (;;)
                               pause();
               }
#endif /* HAVE_LIBSCF_H */
               msyslog(LOG_ERR, "Set system clock by hand.");
               exit (-1);
               /* not reached */

       /*
        * Clock was stepped. Flush all time values of all peers.
        */
       case 2:
               clear_all();
               set_sys_leap(LEAP_NOTINSYNC);
               sys_stratum = STRATUM_UNSPEC;
               memcpy(&sys_refid, "STEP", 4);
               sys_rootdelay = 0;
               p2_rootdisp = 0;
               prev_rootdisp = 0;
               sys_rootdisp = 0;
               L_CLR(&p2_reftime);     /* Should we clear p2_reftime? */
               L_CLR(&prev_reftime);   /* Should we clear prev_reftime? */
               L_CLR(&sys_reftime);
               sys_jitter = LOGTOD(sys_precision);
               leapsec_reset_frame();
               break;

       /*
        * Clock was slewed. Handle the leapsecond stuff.
        */
       case 1:

               /*
                * If this is the first time the clock is set, reset the
                * leap bits. If crypto, the timer will goose the setup
                * process.
                */
               if (sys_leap == LEAP_NOTINSYNC) {
                       set_sys_leap(LEAP_NOWARNING);
#ifdef AUTOKEY
                       if (crypto_flags)
                               crypto_update();
#endif  /* AUTOKEY */

               }

               /*
                * If there is no leap second pending and the number of
                * survivor leap bits is greater than half the number of
                * survivors, try to schedule a leap for the end of the
                * current month. (This only works if no leap second for
                * that range is in the table, so doing this more than
                * once is mostly harmless.)
                */
               if (leapsec == LSPROX_NOWARN) {
                       if (   leap_vote_ins > leap_vote_del
                           && leap_vote_ins > sys_survivors / 2) {
                               get_systime(&now);
                               leapsec_add_dyn(TRUE, now.l_ui, NULL);
                       }
                       if (   leap_vote_del > leap_vote_ins
                           && leap_vote_del > sys_survivors / 2) {
                               get_systime(&now);
                               leapsec_add_dyn(FALSE, now.l_ui, NULL);
                       }
               }
               break;

       /*
        * Popcorn spike or step threshold exceeded. Pretend it never
        * happened.
        */
       default:
               break;
       }
}


/*
* poll_update - update peer poll interval
*/
void
poll_update(
       struct peer *peer,      /* peer structure pointer */
       u_char  mpoll,
       u_char  skewpoll
       )
{
       u_long  next, utemp, limit;
       u_char  hpoll;

       /*
        * This routine figures out when the next poll should be sent.
        * That turns out to be wickedly complicated. One problem is
        * that sometimes the time for the next poll is in the past when
        * the poll interval is reduced. We watch out for races here
        * between the receive process and the poll process.
        *
        * Clamp the poll interval between minpoll and maxpoll.
        */
       hpoll = max(min(peer->maxpoll, mpoll), peer->minpoll);

#ifdef AUTOKEY
       /*
        * If during the crypto protocol the poll interval has changed,
        * the lifetimes in the key list are probably bogus. Purge the
        * the key list and regenerate it later.
        */
       if ((peer->flags & FLAG_SKEY) && hpoll != peer->hpoll)
               key_expire(peer);
#endif  /* AUTOKEY */
       peer->hpoll = hpoll;

       /*
        * There are three variables important for poll scheduling, the
        * current time (current_time), next scheduled time (nextdate)
        * and the earliest time (utemp). The earliest time is 2 s
        * seconds, but could be more due to rate management. When
        * sending in a burst, use the earliest time. When not in a
        * burst but with a reply pending, send at the earliest time
        * unless the next scheduled time has not advanced. This can
        * only happen if multiple replies are pending in the same
        * response interval. Otherwise, send at the later of the next
        * scheduled time and the earliest time.
        *
        * Now we figure out if there is an override. If a burst is in
        * progress and we get called from the receive process, just
        * slink away. If called from the poll process, delay 1 s for a
        * reference clock, otherwise 2 s.
        */
       utemp = current_time + max(peer->throttle - (NTP_SHIFT - 1) *
           (1 << peer->minpoll), ntp_minpkt);

       /*[Bug 3592] avoid unlimited postpone of next poll */
       limit = (2u << hpoll);
       if (limit > 64)
               limit -= (limit >> 2);
       limit += peer->outdate;
       if (limit < current_time)
               limit = current_time;

       if (peer->burst > 0) {
               if (peer->nextdate > current_time)
                       return;
#ifdef REFCLOCK
               else if (peer->flags & FLAG_REFCLOCK)
                       peer->nextdate = current_time + RESP_DELAY;
#endif /* REFCLOCK */
               else
                       peer->nextdate = utemp;

#ifdef AUTOKEY
       /*
        * If a burst is not in progress and a crypto response message
        * is pending, delay 2 s, but only if this is a new interval.
        */
       } else if (peer->cmmd != NULL) {
               if (peer->nextdate > current_time) {
                       if (peer->nextdate + ntp_minpkt != utemp)
                               peer->nextdate = utemp;
               } else {
                       peer->nextdate = utemp;
               }
#endif  /* AUTOKEY */

       /*
        * The ordinary case. If a retry, use minpoll; if unreachable,
        * use host poll; otherwise, use the minimum of host and peer
        * polls; In other words, oversampling is okay but
        * understampling is evil. Use the maximum of this value and the
        * headway. If the average headway is greater than the headway
        * threshold, increase the headway by the minimum interval.
        */
       } else {
               if (peer->retry > 0)
                       hpoll = peer->minpoll;
               else
                       hpoll = min(peer->ppoll, peer->hpoll);
#ifdef REFCLOCK
               if (peer->flags & FLAG_REFCLOCK)
                       next = 1 << hpoll;
               else
#endif /* REFCLOCK */
                       next = ((0x1000UL | (ntp_random() & 0x0ff)) <<
                           hpoll) >> 12;
               next += peer->outdate;
               /* XXX: bug3596: Deal with poll skew list? */
               if (skewpoll) {
                       psl_item psi;

                       if (0 == get_pollskew(hpoll, &psi)) {
                               int sub = psi.sub;
                               int qty = psi.qty;
                               int msk = psi.msk;
                               int val;

                               if (   0 != sub
                                   || 0 != qty) {
                                       do {
                                               val = ntp_random() & msk;
                                       } while (val > qty);

                                       next -= sub;
                                       next += val;
                               }
                       } else {
                               /* get_pollskew() already logged this */
                       }
               }
               if (next > utemp)
                       peer->nextdate = next;
               else
                       peer->nextdate = utemp;
               if (peer->throttle > (1 << peer->minpoll))
                       peer->nextdate += ntp_minpkt;
       }

       /*[Bug 3592] avoid unlimited postpone of next poll */
       if (peer->nextdate > limit) {
               DPRINTF(1, ("poll_update: clamp reached; limit %lu next %lu\n",
                           limit, peer->nextdate));
               peer->nextdate = limit;
       }
       DPRINTF(2, ("poll_update: at %lu %s poll %d burst %d retry %d head %d early %lu next %lu\n",
                   current_time, ntoa(&peer->srcadr), peer->hpoll,
                   peer->burst, peer->retry, peer->throttle,
                   utemp - current_time, peer->nextdate -
                   current_time));
}


/*
* peer_clear - clear peer filter registers.  See Section 3.4.8 of the
* spec.
*/
void
peer_clear(
       struct peer *peer,              /* peer structure */
       const char *ident               /* tally lights */
       )
{
       static u_long   earliest;
       u_char          u;
       l_fp            bxmt = peer->bxmt;      /* bcast clients retain this! */

#ifdef AUTOKEY
       /*
        * If cryptographic credentials have been acquired, toss them to
        * Valhalla. Note that autokeys are ephemeral, in that they are
        * tossed immediately upon use. Therefore, the keylist can be
        * purged anytime without needing to preserve random keys. Note
        * that, if the peer is purged, the cryptographic variables are
        * purged, too. This makes it much harder to sneak in some
        * unauthenticated data in the clock filter.
        */
       key_expire(peer);
       if (peer->iffval != NULL)
               BN_free(peer->iffval);
       value_free(&peer->cookval);
       value_free(&peer->recval);
       value_free(&peer->encrypt);
       value_free(&peer->sndval);
       if (peer->cmmd != NULL)
               free(peer->cmmd);
       if (peer->subject != NULL)
               free(peer->subject);
       if (peer->issuer != NULL)
               free(peer->issuer);
#endif /* AUTOKEY */

       /*
        * Clear all values, including the optional crypto values above.
        */
       memset(CLEAR_TO_ZERO(peer), 0, LEN_CLEAR_TO_ZERO(peer));
       peer->ppoll = peer->maxpoll;
       peer->hpoll = peer->minpoll;
       peer->disp = MAXDISPERSE;
       peer->flash = peer_unfit(peer);
       peer->jitter = LOGTOD(sys_precision);

       /* Don't throw away our broadcast replay protection */
       if (peer->hmode == MODE_BCLIENT)
               peer->bxmt = bxmt;

       /*
        * If interleave mode, initialize the alternate origin switch.
        */
       if (peer->flags & FLAG_XLEAVE)
               peer->flip = 1;
       for (u = 0; u < NTP_SHIFT; u++) {
               peer->filter_order[u] = u;
               peer->filter_disp[u] = MAXDISPERSE;
       }
#ifdef REFCLOCK
       if (!(peer->flags & FLAG_REFCLOCK)) {
#endif
               peer->leap = LEAP_NOTINSYNC;
               peer->stratum = STRATUM_UNSPEC;
               memcpy(&peer->refid, ident, 4);
#ifdef REFCLOCK
       } else {
               /* Clear refclock sample filter */
               peer->procptr->codeproc = 0;
               peer->procptr->coderecv = 0;
       }
#endif

       /*
        * During initialization use the association count to spread out
        * the polls at one-second intervals. Unconfigured associations'
        * first poll is delayed by the "discard minimum" plus 1 to avoid
        * rate limiting. Other post-startup new or cleared associations
        * randomize the first poll over the minimum poll interval to
        * avoid implosion.
        */
       peer->nextdate = peer->update = peer->outdate = current_time;
       if (initializing) {
               peer->nextdate += peer_associations;
       } else if (!(FLAG_CONFIG & peer->flags)) {
               peer->nextdate += ntp_minpkt + 1;
               /* space out manycastclient first polls */
               if (peer->nextdate < earliest) {
                       peer->nextdate = earliest;
               }
               earliest = peer->nextdate + 1;
       } else {
               peer->nextdate += ntp_random() % (1 << peer->minpoll);
       }
#ifdef AUTOKEY
       peer->refresh = current_time + (1 << NTP_REFRESH);
#endif  /* AUTOKEY */
       DPRINTF(1, ("peer_clear: at %ld next %ld associd %d refid %s\n",
                   current_time, peer->nextdate, peer->associd,
                   ident));
}


/*
* clock_filter - add incoming clock sample to filter register and run
*                the filter procedure to find the best sample.
*/
void
clock_filter(
       struct peer *peer,              /* peer structure pointer */
       double  sample_offset,          /* clock offset */
       double  sample_delay,           /* roundtrip delay */
       double  sample_disp             /* dispersion */
       )
{
       double  dst[NTP_SHIFT];         /* distance vector */
       u_char  ord[NTP_SHIFT];         /* index vector */
       short   i, j;
       u_char  k, m;
       double  dtemp, etemp;

       /*
        * A sample consists of the offset, delay, dispersion and epoch
        * of arrival. The offset and delay are determined by the on-
        * wire protocol. The dispersion grows from the last outbound
        * packet to the arrival of this one increased by the sum of the
        * peer precision and the system precision as required by the
        * error budget. First, shift the new arrival into the shift
        * register discarding the oldest one.
        */
       j = peer->filter_nextpt;
       peer->filter_offset[j] = sample_offset;
       peer->filter_delay[j] = sample_delay;
       peer->filter_disp[j] = sample_disp;
       peer->filter_epoch[j] = current_time;
       j = (j + 1) % NTP_SHIFT;
       peer->filter_nextpt = (u_char)j;

       /*
        * Update dispersions since the last update and at the same
        * time initialize the distance and index lists. Since samples
        * become increasingly uncorrelated beyond the Allan intercept,
        * only under exceptional cases will an older sample be used.
        * Therefore, the distance list uses a compound metric. If the
        * dispersion is greater than the maximum dispersion, clamp the
        * distance at that value. If the time since the last update is
        * less than the Allan intercept use the delay; otherwise, use
        * the sum of the delay and dispersion.
        */
       dtemp = clock_phi * (current_time - peer->update);
       peer->update = current_time;
       for (i = NTP_SHIFT - 1; i >= 0; i--) {
               if (i != 0)
                       peer->filter_disp[j] += dtemp;
               if (peer->filter_disp[j] >= MAXDISPERSE) {
                       peer->filter_disp[j] = MAXDISPERSE;
                       dst[i] = MAXDISPERSE;
               } else if (peer->update - peer->filter_epoch[j] >
                   (u_long)ULOGTOD(allan_xpt)) {
                       dst[i] = peer->filter_delay[j] +
                           peer->filter_disp[j];
               } else {
                       dst[i] = peer->filter_delay[j];
               }
               ord[i] = (u_char)j;
               j = (j + 1) % NTP_SHIFT;
       }

       /*
        * If the clock has stabilized, sort the samples by distance.
        */
       if (freq_cnt == 0) {
               for (i = 1; i < NTP_SHIFT; i++) {
                       for (j = 0; j < i; j++) {
                               if (dst[j] > dst[i]) {
                                       k = ord[j];
                                       ord[j] = ord[i];
                                       ord[i] = k;
                                       etemp = dst[j];
                                       dst[j] = dst[i];
                                       dst[i] = etemp;
                               }
                       }
               }
       }

       /*
        * Copy the index list to the association structure so ntpq
        * can see it later. Prune the distance list to leave only
        * samples less than the maximum dispersion, which disfavors
        * uncorrelated samples older than the Allan intercept. To
        * further improve the jitter estimate, of the remainder leave
        * only samples less than the maximum distance, but keep at
        * least two samples for jitter calculation.
        */
       m = 0;
       for (i = 0; i < NTP_SHIFT; i++) {
               peer->filter_order[i] = ord[i];
               if (   dst[i] >= MAXDISPERSE
                   || (m >= 2 && dst[i] >= sys_maxdist))
                       continue;
               m++;
       }

       /*
        * Compute the dispersion and jitter. The dispersion is weighted
        * exponentially by NTP_FWEIGHT (0.5) so it is normalized close
        * to 1.0. The jitter is the RMS differences relative to the
        * lowest delay sample.
        */
       peer->disp = peer->jitter = 0;
       k = ord[0];
       for (i = NTP_SHIFT - 1; i >= 0; i--) {
               j = ord[i];
               peer->disp = NTP_FWEIGHT * (  peer->disp
                                           + peer->filter_disp[j]);
               if (i < m) {
                       peer->jitter += DIFF(peer->filter_offset[j],
                                            peer->filter_offset[k]);
               }
       }

       /*
        * If no acceptable samples remain in the shift register,
        * quietly tiptoe home leaving only the dispersion. Otherwise,
        * save the offset, delay and jitter. Note the jitter must not
        * be less than the precision.
        */
       if (0 == m) {
               clock_select();
               return;
       }
       etemp = fabs(peer->offset - peer->filter_offset[k]);
       peer->offset = peer->filter_offset[k];
       peer->delay = peer->filter_delay[k];
       if (m > 1) {
               peer->jitter /= m - 1;
       }
       peer->jitter = max(SQRT(peer->jitter), LOGTOD(sys_precision));

       /*
        * If the the new sample and the current sample are both valid
        * and the difference between their offsets exceeds CLOCK_SGATE
        * (3) times the jitter and the interval between them is less
        * than twice the host poll interval, consider the new sample
        * a popcorn spike and ignore it.
        */
       if (   peer->disp < sys_maxdist
           && peer->filter_disp[k] < sys_maxdist
           && etemp > CLOCK_SGATE * peer->jitter
           &&   peer->filter_epoch[k] - peer->epoch
              < 2. * ULOGTOD(peer->hpoll)) {
               mprintf_event(PEVNT_POPCORN, peer, "%.9f s", etemp);
               return;
       }

       /*
        * A new minimum sample is useful only if it is later than the
        * last one used. In this design the maximum lifetime of any
        * sample is not greater than NTP_SHIFT (8) times the poll
        * interval, so the maximum interval between minimum samples is
        * NTP_SHIFT packets.
        */
       if (peer->filter_epoch[k] <= peer->epoch) {
               DPRINTF(2, ("clock_filter: old sample %lu s\n",
                           current_time - peer->filter_epoch[k]));
               return;
       }
       peer->epoch = peer->filter_epoch[k];

       /*
        * The mitigated sample statistics are saved for later
        * processing. If not synchronized or not in a burst, tickle the
        * clock select algorithm.
        */
       record_peer_stats(&peer->srcadr, ctlpeerstatus(peer), peer->offset,
                         peer->delay, peer->disp, peer->jitter);
       DPRINTF(1, ("clock_filter: n %hu off %.9f del %.9f dsp %.9f jit %.9f\n",
                   (u_short)m, peer->offset, peer->delay, peer->disp,
                   peer->jitter));
       if (0 == peer->burst || LEAP_NOTINSYNC == sys_leap) {
               clock_select();
       }
}


/*
* clock_select - find the pick-of-the-litter clock
*
* LOCKCLOCK: (1) If the local clock is the prefer peer, it will always
* be enabled, even if declared falseticker, (2) only the prefer peer
* can be selected as the system peer, (3) if the external source is
* down, the system leap bits are set to 11 and the stratum set to
* infinity.
*/
void
clock_select(void)
{
       struct peer *peer;
       int     i, j, k, n;
       int     nlist, nl2;
       int     allow;
       int     speer;
       double  d, e, f, g;
       double  high, low;
       double  speermet;
       double  lastresort_dist = MAXDISPERSE;
       double  orphmet = 2.0 * U_INT32_MAX; /* 2x is greater than */
       struct endpoint endp;
       struct peer *osys_peer;
       struct peer *sys_prefer = NULL; /* prefer peer */
       struct peer *typesystem = NULL;
       struct peer *typelastresort = NULL;
       struct peer *typeorphan = NULL;
#ifdef REFCLOCK
       struct peer *typeacts = NULL;
       struct peer *typelocal = NULL;
       struct peer *typepps = NULL;
#endif /* REFCLOCK */
       static struct endpoint *endpoint = NULL;
       static int *indx = NULL;
       static peer_select *peers = NULL;
       static u_int endpoint_size = 0;
       static u_int peers_size = 0;
       static u_int indx_size = 0;
       size_t octets;

       /*
        * Initialize and create endpoint, index and peer lists big
        * enough to handle all associations.
        */
       osys_peer = sys_peer;
       sys_survivors = 0;
#ifdef LOCKCLOCK
       set_sys_leap(LEAP_NOTINSYNC);
       sys_stratum = STRATUM_UNSPEC;
       memcpy(&sys_refid, "DOWN", 4);
#endif /* LOCKCLOCK */

       /*
        * Allocate dynamic space depending on the number of
        * associations.
        */
       nlist = 1;
       for (peer = peer_list; peer != NULL; peer = peer->p_link)
               nlist++;
       endpoint_size = ALIGNED_SIZE(nlist * 2 * sizeof(*endpoint));
       peers_size = ALIGNED_SIZE(nlist * sizeof(*peers));
       indx_size = ALIGNED_SIZE(nlist * 2 * sizeof(*indx));
       octets = endpoint_size + peers_size + indx_size;
       endpoint = erealloc(endpoint, octets);
       peers = INC_ALIGNED_PTR(endpoint, endpoint_size);
       indx = INC_ALIGNED_PTR(peers, peers_size);

       /*
        * Initially, we populate the island with all the rifraff peers
        * that happen to be lying around. Those with seriously
        * defective clocks are immediately booted off the island. Then,
        * the falsetickers are culled and put to sea. The truechimers
        * remaining are subject to repeated rounds where the most
        * unpopular at each round is kicked off. When the population
        * has dwindled to sys_minclock, the survivors split a million
        * bucks and collectively crank the chimes.
        */
       nlist = nl2 = 0;        /* none yet */
       for (peer = peer_list; peer != NULL; peer = peer->p_link) {
               peer->new_status = CTL_PST_SEL_REJECT;

               /*
                * Leave the island immediately if the peer is
                * unfit to synchronize.
                */
               if (peer_unfit(peer)) {
                       continue;
               }

               /*
                * If we have never been synchronised, look for any peer
                * which has ever been synchronised and pick the one which
                * has the lowest root distance. This can be used as a last
                * resort if all else fails. Once we get an initial sync
                * with this peer, sys_reftime gets set and so this
                * function becomes disabled.
                */
               if (L_ISZERO(&sys_reftime)) {
                       d = root_distance(peer);
                       if (!L_ISZERO(&peer->reftime) && d < lastresort_dist) {
                               typelastresort = peer;
                               lastresort_dist = d;
                       }
               }

               /*
                * If this peer is an orphan parent, elect the
                * one with the lowest metric defined as the
                * IPv4 address or the first 64 bits of the
                * hashed IPv6 address.  To ensure convergence
                * on the same selected orphan, consider as
                * well that this system may have the lowest
                * metric and be the orphan parent.  If this
                * system wins, sys_peer will be NULL to trigger
                * orphan mode in timer().
                */
               if (peer->stratum == sys_orphan) {
                       u_int32 localmet;
                       u_int32 peermet;

                       if (peer->dstadr != NULL)
                               localmet = ntohl(peer->dstadr->addr_refid);
                       else
                               localmet = U_INT32_MAX;
                       peermet = ntohl(addr2refid(&peer->srcadr));
                       if (peermet < localmet && peermet < orphmet) {
                               typeorphan = peer;
                               orphmet = peermet;
                       }
                       continue;
               }

               /*
                * If this peer could have the orphan parent
                * as a synchronization ancestor, exclude it
                * from selection to avoid forming a
                * synchronization loop within the orphan mesh,
                * triggering stratum climb to infinity
                * instability.  Peers at stratum higher than
                * the orphan stratum could have the orphan
                * parent in ancestry so are excluded.
                * See http://bugs.ntp.org/2050
                */
               if (peer->stratum > sys_orphan) {
                       continue;
               }
#ifdef REFCLOCK
               /*
                * The following are special cases. We deal
                * with them later.
                */
               if (!(peer->flags & FLAG_PREFER)) {
                       switch (peer->refclktype) {
                       case REFCLK_LOCALCLOCK:
                               if (   current_time > orphwait
                                   && typelocal == NULL)
                                       typelocal = peer;
                               continue;

                       case REFCLK_ACTS:
                               if (   current_time > orphwait
                                   && typeacts == NULL)
                                       typeacts = peer;
                               continue;
                       }
               }
#endif /* REFCLOCK */

               /*
                * If we get this far, the peer can stay on the
                * island, but does not yet have the immunity
                * idol.
                */
               peer->new_status = CTL_PST_SEL_SANE;
               f = root_distance(peer);
               peers[nlist].peer = peer;
               peers[nlist].error = peer->jitter;
               peers[nlist].synch = f;
               nlist++;

               /*
                * Insert each interval endpoint on the unsorted
                * endpoint[] list.
                */
               e = peer->offset;
               endpoint[nl2].type = -1;        /* lower end */
               endpoint[nl2].val = e - f;
               nl2++;
               endpoint[nl2].type = 1;         /* upper end */
               endpoint[nl2].val = e + f;
               nl2++;
       }
       /*
        * Construct sorted indx[] of endpoint[] indexes ordered by
        * offset.
        */
       for (i = 0; i < nl2; i++)
               indx[i] = i;
       for (i = 0; i < nl2; i++) {
               endp = endpoint[indx[i]];
               e = endp.val;
               k = i;
               for (j = i + 1; j < nl2; j++) {
                       endp = endpoint[indx[j]];
                       if (endp.val < e) {
                               e = endp.val;
                               k = j;
                       }
               }
               if (k != i) {
                       j = indx[k];
                       indx[k] = indx[i];
                       indx[i] = j;
               }
       }
       for (i = 0; i < nl2; i++)
               DPRINTF(3, ("select: endpoint %2d %.6f\n",
                       endpoint[indx[i]].type, endpoint[indx[i]].val));

       /*
        * This is the actual algorithm that cleaves the truechimers
        * from the falsetickers. The original algorithm was described
        * in Keith Marzullo's dissertation, but has been modified for
        * better accuracy.
        *
        * Briefly put, we first assume there are no falsetickers, then
        * scan the candidate list first from the low end upwards and
        * then from the high end downwards. The scans stop when the
        * number of intersections equals the number of candidates less
        * the number of falsetickers. If this doesn't happen for a
        * given number of falsetickers, we bump the number of
        * falsetickers and try again. If the number of falsetickers
        * becomes equal to or greater than half the number of
        * candidates, the Albanians have won the Byzantine wars and
        * correct synchronization is not possible.
        *
        * Here, nlist is the number of candidates and allow is the
        * number of falsetickers. Upon exit, the truechimers are the
        * survivors with offsets not less than low and not greater than
        * high. There may be none of them.
        */
       low = 1e9;
       high = -1e9;
       for (allow = 0; 2 * allow < nlist; allow++) {

               /*
                * Bound the interval (low, high) as the smallest
                * interval containing points from the most sources.
                */
               n = 0;
               for (i = 0; i < nl2; i++) {
                       low = endpoint[indx[i]].val;
                       n -= endpoint[indx[i]].type;
                       if (n >= nlist - allow)
                               break;
               }
               n = 0;
               for (j = nl2 - 1; j >= 0; j--) {
                       high = endpoint[indx[j]].val;
                       n += endpoint[indx[j]].type;
                       if (n >= nlist - allow)
                               break;
               }

               /*
                * If an interval containing truechimers is found, stop.
                * If not, increase the number of falsetickers and go
                * around again.
                */
               if (high > low)
                       break;
       }

       /*
        * Clustering algorithm. Whittle candidate list of falsetickers,
        * who leave the island immediately. The TRUE peer is always a
        * truechimer. We must leave at least one peer to collect the
        * million bucks.
        *
        * We assert the correct time is contained in the interval, but
        * the best offset estimate for the interval might not be
        * contained in the interval. For this purpose, a truechimer is
        * defined as the midpoint of an interval that overlaps the
        * intersection interval.
        */
       j = 0;
       for (i = 0; i < nlist; i++) {
               double  h;

               peer = peers[i].peer;
               h = peers[i].synch;
               if ((   high <= low
                    || peer->offset + h < low
                    || peer->offset - h > high
                   ) && !(peer->flags & FLAG_TRUE))
                       continue;

#ifdef REFCLOCK
               /*
                * Eligible PPS peers must survive the intersection
                * algorithm. Use the first one found, but don't
                * include any of them in the cluster population.
                */
               if (peer->flags & FLAG_PPS) {
                       if (typepps == NULL)
                               typepps = peer;
                       if (!(peer->flags & FLAG_TSTAMP_PPS))
                               continue;
               }
#endif /* REFCLOCK */

               if (j != i)
                       peers[j] = peers[i];
               j++;
       }
       nlist = j;

       /*
        * If no survivors remain at this point, check if the modem
        * driver, local driver or orphan parent in that order. If so,
        * nominate the first one found as the only survivor.
        * Otherwise, give up and leave the island to the rats.
        */
       if (nlist == 0) {
               peers[0].error = 0;
               peers[0].synch = sys_mindisp;
#ifdef REFCLOCK
               if (typeacts != NULL) {
                       peers[0].peer = typeacts;
                       nlist = 1;
               } else if (typelocal != NULL) {
                       peers[0].peer = typelocal;
                       nlist = 1;
               } else
#endif /* REFCLOCK */
               if (typeorphan != NULL) {
                       peers[0].peer = typeorphan;
                       nlist = 1;
               } else if (typelastresort != NULL) {
                       peers[0].peer = typelastresort;
                       nlist = 1;
               }
       }

       /*
        * Mark the candidates at this point as truechimers.
        */
       for (i = 0; i < nlist; i++) {
               peers[i].peer->new_status = CTL_PST_SEL_SELCAND;
               DPRINTF(2, ("select: survivor %s %f\n",
                       stoa(&peers[i].peer->srcadr), peers[i].synch));
       }

       /*
        * Now, vote outliers off the island by select jitter weighted
        * by root distance. Continue voting as long as there are more
        * than sys_minclock survivors and the select jitter of the peer
        * with the worst metric is greater than the minimum peer
        * jitter. Stop if we are about to discard a TRUE or PREFER
        * peer, who of course have the immunity idol.
        */
       while (1) {
               d = 1e9;
               e = -1e9;
               g = 0;
               k = 0;
               for (i = 0; i < nlist; i++) {
                       if (peers[i].error < d)
                               d = peers[i].error;
                       peers[i].seljit = 0;
                       if (nlist > 1) {
                               f = 0;
                               for (j = 0; j < nlist; j++)
                                       f += DIFF(peers[j].peer->offset,
                                           peers[i].peer->offset);
                               peers[i].seljit = SQRT(f / (nlist - 1));
                       }
                       if (peers[i].seljit * peers[i].synch > e) {
                               g = peers[i].seljit;
                               e = peers[i].seljit * peers[i].synch;
                               k = i;
                       }
               }
               g = max(g, LOGTOD(sys_precision));
               if (   nlist <= max(1, sys_minclock)
                   || g <= d
                   || ((FLAG_TRUE | FLAG_PREFER) & peers[k].peer->flags))
                       break;

               DPRINTF(3, ("select: drop %s seljit %.9f jit %.9f\n",
                       ntoa(&peers[k].peer->srcadr), g, d));
               if (nlist > sys_maxclock)
                       peers[k].peer->new_status = CTL_PST_SEL_EXCESS;
               for (j = k + 1; j < nlist; j++)
                       peers[j - 1] = peers[j];
               nlist--;
       }

       /*
        * What remains is a list usually not greater than sys_minclock
        * peers. Note that unsynchronized peers cannot survive this
        * far.  Count and mark these survivors.
        *
        * While at it, count the number of leap warning bits found.
        * This will be used later to vote the system leap warning bit.
        * If a leap warning bit is found on a reference clock, the vote
        * is always won.
        *
        * Choose the system peer using a hybrid metric composed of the
        * selection jitter scaled by the root distance augmented by
        * stratum scaled by sys_mindisp (.001 by default). The goal of
        * the small stratum factor is to avoid clockhop between a
        * reference clock and a network peer which has a refclock and
        * is using an older ntpd, which does not floor sys_rootdisp at
        * sys_mindisp.
        *
        * In contrast, ntpd 4.2.6 and earlier used stratum primarily
        * in selecting the system peer, using a weight of 1 second of
        * additional root distance per stratum.  This heavy bias is no
        * longer appropriate, as the scaled root distance provides a
        * more rational metric carrying the cumulative error budget.
        */
       e = 1e9;
       speer = 0;
       leap_vote_ins = 0;
       leap_vote_del = 0;
       for (i = 0; i < nlist; i++) {
               peer = peers[i].peer;
               peer->unreach = 0;
               peer->new_status = CTL_PST_SEL_SYNCCAND;
               sys_survivors++;
               if (peer->leap == LEAP_ADDSECOND) {
                       if (peer->flags & FLAG_REFCLOCK)
                               leap_vote_ins = nlist;
                       else if (leap_vote_ins < nlist)
                               leap_vote_ins++;
               }
               if (peer->leap == LEAP_DELSECOND) {
                       if (peer->flags & FLAG_REFCLOCK)
                               leap_vote_del = nlist;
                       else if (leap_vote_del < nlist)
                               leap_vote_del++;
               }
               if (peer->flags & FLAG_PREFER)
                       sys_prefer = peer;
               speermet = peers[i].seljit * peers[i].synch +
                   peer->stratum * sys_mindisp;
               if (speermet < e) {
                       e = speermet;
                       speer = i;
               }
       }

       /*
        * Unless there are at least sys_misane survivors, leave the
        * building dark. Otherwise, do a clockhop dance. Ordinarily,
        * use the selected survivor speer. However, if the current
        * system peer is not speer, stay with the current system peer
        * as long as it doesn't get too old or too ugly.
        */
       if (nlist > 0 && nlist >= sys_minsane) {
               double  x;

               typesystem = peers[speer].peer;
               if (osys_peer == NULL || osys_peer == typesystem) {
                       sys_clockhop = 0;
               } else if ((x = fabs(typesystem->offset -
                   osys_peer->offset)) < sys_mindisp) {
                       if (0 == sys_clockhop)
                               sys_clockhop = sys_mindisp;
                       else
                               sys_clockhop *= .5;
                       DPRINTF(1, ("select: clockhop %d %.9f %.9f\n",
                               j, x, sys_clockhop));
                       if (x < sys_clockhop)
                               typesystem = osys_peer;
                       else
                               sys_clockhop = 0;
               } else {
                       sys_clockhop = 0;
               }
       }

       /*
        * Mitigation rules of the game. We have the pick of the
        * litter in typesystem if any survivors are left. If
        * there is a prefer peer, use its offset and jitter.
        * Otherwise, use the combined offset and jitter of all kitters.
        */
       if (typesystem != NULL) {
               if (sys_prefer == NULL) {
                       typesystem->new_status = CTL_PST_SEL_SYSPEER;
                       clock_combine(peers, sys_survivors, speer);
               } else {
                       typesystem = sys_prefer;
                       sys_clockhop = 0;
                       typesystem->new_status = CTL_PST_SEL_SYSPEER;
                       sys_offset = typesystem->offset;
                       sys_jitter = typesystem->jitter;
               }
               DPRINTF(1, ("select: combine offset %.9f jitter %.9f\n",
                       sys_offset, sys_jitter));
       }
#ifdef REFCLOCK
       /*
        * If a PPS driver is lit and the combined offset is less than
        * 0.4 s, select the driver as the PPS peer and use its offset
        * and jitter. However, if this is the atom driver, use it only
        * if there is a prefer peer or there are no survivors and none
        * are required.
        */
       if (   typepps != NULL
           && fabs(sys_offset) < 0.4
           && (   typepps->refclktype != REFCLK_ATOM_PPS
               || (   typepps->refclktype == REFCLK_ATOM_PPS
                   && (   sys_prefer != NULL
                       || (typesystem == NULL && sys_minsane == 0))))) {
               typesystem = typepps;
               sys_clockhop = 0;
               typesystem->new_status = CTL_PST_SEL_PPS;
               sys_offset = typesystem->offset;
               sys_jitter = typesystem->jitter;
               DPRINTF(1, ("select: pps offset %.9f jitter %.9f\n",
                       sys_offset, sys_jitter));
       }
#endif /* REFCLOCK */

       /*
        * If there are no survivors at this point, there is no
        * system peer. If so and this is an old update, keep the
        * current statistics, but do not update the clock.
        */
       if (typesystem == NULL) {
               if (osys_peer != NULL) {
                       orphwait = current_time + sys_orphwait;
                       report_event(EVNT_NOPEER, NULL, NULL);
               }
               sys_peer = NULL;
               for (peer = peer_list; peer != NULL; peer = peer->p_link)
                       peer->status = peer->new_status;
               return;
       }

       /*
        * Do not use old data, as this may mess up the clock discipline
        * stability.
        */
       if (typesystem->epoch <= sys_epoch)
               return;

       /*
        * We have found the alpha male. Wind the clock.
        */
       if (osys_peer != typesystem)
               report_event(PEVNT_NEWPEER, typesystem, NULL);
       for (peer = peer_list; peer != NULL; peer = peer->p_link)
               peer->status = peer->new_status;
       clock_update(typesystem);
}


static void
clock_combine(
       peer_select *   peers,  /* survivor list */
       int             npeers, /* number of survivors */
       int             syspeer /* index of sys.peer */
       )
{
       int     i;
       double  x, y, z, w;

       y = z = w = 0;
       for (i = 0; i < npeers; i++) {
               x = 1. / peers[i].synch;
               y += x;
               z += x * peers[i].peer->offset;
               w += x * DIFF(peers[i].peer->offset,
                   peers[syspeer].peer->offset);
       }
       sys_offset = z / y;
       sys_jitter = SQRT(w / y + SQUARE(peers[syspeer].seljit));
}


/*
* root_distance - compute synchronization distance from peer to root
*/
static double
root_distance(
       struct peer *peer       /* peer structure pointer */
       )
{
       double  dtemp;

       /*
        * Root Distance (LAMBDA) is defined as:
        * (delta + DELTA)/2 + epsilon + EPSILON + D
        *
        * where:
        *  delta   is the round-trip delay
        *  DELTA   is the root delay
        *  epsilon is the peer dispersion
        *          + (15 usec each second)
        *  EPSILON is the root dispersion
        *  D       is sys_jitter
        *
        * NB: Think hard about why we are using these values, and what
        * the alternatives are, and the various pros/cons.
        *
        * DLM thinks these are probably the best choices from any of the
        * other worse choices.
        */
       dtemp = (peer->delay + peer->rootdelay) / 2
               + peer->disp
                 + clock_phi * (current_time - peer->update)
               + peer->rootdisp
               + peer->jitter;
       /*
        * Careful squeak here. The value returned must be greater than
        * the minimum root dispersion in order to avoid clockhop with
        * highly precise reference clocks. Note that the root distance
        * cannot exceed the sys_maxdist, as this is the cutoff by the
        * selection algorithm.
        */
       if (dtemp < sys_mindisp)
               dtemp = sys_mindisp;
       return (dtemp);
}


/*
* peer_xmit - send packet for persistent association.
*/
static void
peer_xmit(
       struct peer *peer       /* peer structure pointer */
       )
{
       struct pkt xpkt;        /* transmit packet */
       size_t  sendlen, authlen;
       keyid_t xkeyid = 0;     /* transmit key ID */
       l_fp    xmt_tx, xmt_ty;

       if (!peer->dstadr) {    /* can't send */
               return;
       }
       xpkt.li_vn_mode = PKT_LI_VN_MODE(sys_leap, peer->version,
           peer->hmode);
       xpkt.stratum = STRATUM_TO_PKT(sys_stratum);
       xpkt.ppoll = peer->hpoll;
       xpkt.precision = sys_precision;
       xpkt.refid = sys_refid;
       xpkt.rootdelay = HTONS_FP(DTOFP(sys_rootdelay));
       xpkt.rootdisp =  HTONS_FP(DTOUFP(sys_rootdisp));
       /* Use sys_reftime for peer exchanges */
       HTONL_FP(&sys_reftime, &xpkt.reftime);
       HTONL_FP(&peer->rec, &xpkt.org);
       HTONL_FP(&peer->dst, &xpkt.rec);

       /*
        * If the received packet contains a MAC, the transmitted packet
        * is authenticated and contains a MAC. If not, the transmitted
        * packet is not authenticated.
        *
        * It is most important when autokey is in use that the local
        * interface IP address be known before the first packet is
        * sent. Otherwise, it is not possible to compute a correct MAC
        * the recipient will accept. Thus, the I/O semantics have to do
        * a little more work. In particular, the wildcard interface
        * might not be usable.
        */
       sendlen = LEN_PKT_NOMAC;
       if (
#ifdef AUTOKEY
           !(peer->flags & FLAG_SKEY) &&
#endif  /* !AUTOKEY */
           peer->keyid == 0) {

               /*
                * Transmit a-priori timestamps
                */
               get_systime(&xmt_tx);
               if (peer->flip == 0) {  /* basic mode */
                       peer->aorg = xmt_tx;
                       HTONL_FP(&xmt_tx, &xpkt.xmt);
               } else {                /* interleaved modes */
                       if (peer->hmode == MODE_BROADCAST) { /* bcst */
                               HTONL_FP(&xmt_tx, &xpkt.xmt);
                               if (peer->flip > 0)
                                       HTONL_FP(&peer->borg,
                                           &xpkt.org);
                               else
                                       HTONL_FP(&peer->aorg,
                                           &xpkt.org);
                       } else {        /* symmetric */
                               if (peer->flip > 0)
                                       HTONL_FP(&peer->borg,
                                           &xpkt.xmt);
                               else
                                       HTONL_FP(&peer->aorg,
                                           &xpkt.xmt);
                       }
               }
               peer->t21_bytes = sendlen;
               sendpkt(&peer->srcadr, peer->dstadr,
                       sys_ttl[(peer->ttl >= sys_ttlmax) ? sys_ttlmax : peer->ttl],
                       &xpkt, sendlen);
               peer->sent++;
               peer->throttle += (1 << peer->minpoll) - 2;

               /*
                * Capture a-posteriori timestamps
                */
               get_systime(&xmt_ty);
               if (peer->flip != 0) {          /* interleaved modes */
                       if (peer->flip > 0)
                               peer->aorg = xmt_ty;
                       else
                               peer->borg = xmt_ty;
                       peer->flip = -peer->flip;
               }
               L_SUB(&xmt_ty, &xmt_tx);
               LFPTOD(&xmt_ty, peer->xleave);
               DPRINTF(1, ("peer_xmit: at %ld %s->%s mode %d len %zu xmt 0x%x.%08x\n",
                           current_time, latoa(peer->dstadr),
                           stoa(&peer->srcadr), peer->hmode, sendlen,
                           xmt_tx.l_ui, xmt_tx.l_uf));
               return;
       }

       /*
        * Authentication is enabled, so the transmitted packet must be
        * authenticated. If autokey is enabled, fuss with the various
        * modes; otherwise, symmetric key cryptography is used.
        */
#ifdef AUTOKEY
       if (peer->flags & FLAG_SKEY) {
               struct exten *exten;    /* extension field */

               /*
                * The Public Key Dance (PKD): Cryptographic credentials
                * are contained in extension fields, each including a
                * 4-octet length/code word followed by a 4-octet
                * association ID and optional additional data. Optional
                * data includes a 4-octet data length field followed by
                * the data itself. Request messages are sent from a
                * configured association; response messages can be sent
                * from a configured association or can take the fast
                * path without ever matching an association. Response
                * messages have the same code as the request, but have
                * a response bit and possibly an error bit set. In this
                * implementation, a message may contain no more than
                * one command and one or more responses.
                *
                * Cryptographic session keys include both a public and
                * a private componet. Request and response messages
                * using extension fields are always sent with the
                * private component set to zero. Packets without
                * extension fields indlude the private component when
                * the session key is generated.
                */
               while (1) {

                       /*
                        * Allocate and initialize a keylist if not
                        * already done. Then, use the list in inverse
                        * order, discarding keys once used. Keep the
                        * latest key around until the next one, so
                        * clients can use client/server packets to
                        * compute propagation delay.
                        *
                        * Note that once a key is used from the list,
                        * it is retained in the key cache until the
                        * next key is used. This is to allow a client
                        * to retrieve the encrypted session key
                        * identifier to verify authenticity.
                        *
                        * If for some reason a key is no longer in the
                        * key cache, a birthday has happened or the key
                        * has expired, so the pseudo-random sequence is
                        * broken. In that case, purge the keylist and
                        * regenerate it.
                        */
                       if (peer->keynumber == 0)
                               make_keylist(peer, peer->dstadr);
                       else
                               peer->keynumber--;
                       xkeyid = peer->keylist[peer->keynumber];
                       if (authistrusted(xkeyid))
                               break;
                       else
                               key_expire(peer);
               }
               peer->keyid = xkeyid;
               exten = NULL;
               switch (peer->hmode) {

               /*
                * In broadcast server mode the autokey values are
                * required by the broadcast clients. Push them when a
                * new keylist is generated; otherwise, push the
                * association message so the client can request them at
                * other times.
                */
               case MODE_BROADCAST:
                       if (peer->flags & FLAG_ASSOC)
                               exten = crypto_args(peer, CRYPTO_AUTO |
                                   CRYPTO_RESP, peer->associd, NULL);
                       else
                               exten = crypto_args(peer, CRYPTO_ASSOC |
                                   CRYPTO_RESP, peer->associd, NULL);
                       break;

               /*
                * In symmetric modes the parameter, certificate,
                * identity, cookie and autokey exchanges are
                * required. The leapsecond exchange is optional. But, a
                * peer will not believe the other peer until the other
                * peer has synchronized, so the certificate exchange
                * might loop until then. If a peer finds a broken
                * autokey sequence, it uses the autokey exchange to
                * retrieve the autokey values. In any case, if a new
                * keylist is generated, the autokey values are pushed.
                */
               case MODE_ACTIVE:
               case MODE_PASSIVE:

                       /*
                        * Parameter, certificate and identity.
                        */
                       if (!peer->crypto)
                               exten = crypto_args(peer, CRYPTO_ASSOC,
                                   peer->associd, hostval.ptr);
                       else if (!(peer->crypto & CRYPTO_FLAG_CERT))
                               exten = crypto_args(peer, CRYPTO_CERT,
                                   peer->associd, peer->issuer);
                       else if (!(peer->crypto & CRYPTO_FLAG_VRFY))
                               exten = crypto_args(peer,
                                   crypto_ident(peer), peer->associd,
                                   NULL);

                       /*
                        * Cookie and autokey. We request the cookie
                        * only when the this peer and the other peer
                        * are synchronized. But, this peer needs the
                        * autokey values when the cookie is zero. Any
                        * time we regenerate the key list, we offer the
                        * autokey values without being asked. If for
                        * some reason either peer finds a broken
                        * autokey sequence, the autokey exchange is
                        * used to retrieve the autokey values.
                        */
                       else if (   sys_leap != LEAP_NOTINSYNC
                                && peer->leap != LEAP_NOTINSYNC
                                && !(peer->crypto & CRYPTO_FLAG_COOK))
                               exten = crypto_args(peer, CRYPTO_COOK,
                                   peer->associd, NULL);
                       else if (!(peer->crypto & CRYPTO_FLAG_AUTO))
                               exten = crypto_args(peer, CRYPTO_AUTO,
                                   peer->associd, NULL);
                       else if (   peer->flags & FLAG_ASSOC
                                && peer->crypto & CRYPTO_FLAG_SIGN)
                               exten = crypto_args(peer, CRYPTO_AUTO |
                                   CRYPTO_RESP, peer->assoc, NULL);

                       /*
                        * Wait for clock sync, then sign the
                        * certificate and retrieve the leapsecond
                        * values.
                        */
                       else if (sys_leap == LEAP_NOTINSYNC)
                               break;

                       else if (!(peer->crypto & CRYPTO_FLAG_SIGN))
                               exten = crypto_args(peer, CRYPTO_SIGN,
                                   peer->associd, hostval.ptr);
                       else if (!(peer->crypto & CRYPTO_FLAG_LEAP))
                               exten = crypto_args(peer, CRYPTO_LEAP,
                                   peer->associd, NULL);
                       break;

               /*
                * In client mode the parameter, certificate, identity,
                * cookie and sign exchanges are required. The
                * leapsecond exchange is optional. If broadcast client
                * mode the same exchanges are required, except that the
                * autokey exchange is substitutes for the cookie
                * exchange, since the cookie is always zero. If the
                * broadcast client finds a broken autokey sequence, it
                * uses the autokey exchange to retrieve the autokey
                * values.
                */
               case MODE_CLIENT:

                       /*
                        * Parameter, certificate and identity.
                        */
                       if (!peer->crypto)
                               exten = crypto_args(peer, CRYPTO_ASSOC,
                                   peer->associd, hostval.ptr);
                       else if (!(peer->crypto & CRYPTO_FLAG_CERT))
                               exten = crypto_args(peer, CRYPTO_CERT,
                                   peer->associd, peer->issuer);
                       else if (!(peer->crypto & CRYPTO_FLAG_VRFY))
                               exten = crypto_args(peer,
                                   crypto_ident(peer), peer->associd,
                                   NULL);

                       /*
                        * Cookie and autokey. These are requests, but
                        * we use the peer association ID with autokey
                        * rather than our own.
                        */
                       else if (!(peer->crypto & CRYPTO_FLAG_COOK))
                               exten = crypto_args(peer, CRYPTO_COOK,
                                   peer->associd, NULL);
                       else if (!(peer->crypto & CRYPTO_FLAG_AUTO))
                               exten = crypto_args(peer, CRYPTO_AUTO,
                                   peer->assoc, NULL);

                       /*
                        * Wait for clock sync, then sign the
                        * certificate and retrieve the leapsecond
                        * values.
                        */
                       else if (sys_leap == LEAP_NOTINSYNC)
                               break;

                       else if (!(peer->crypto & CRYPTO_FLAG_SIGN))
                               exten = crypto_args(peer, CRYPTO_SIGN,
                                   peer->associd, hostval.ptr);
                       else if (!(peer->crypto & CRYPTO_FLAG_LEAP))
                               exten = crypto_args(peer, CRYPTO_LEAP,
                                   peer->associd, NULL);
                       break;
               }

               /*
                * Add a queued extension field if present. This is
                * always a request message, so the reply ID is already
                * in the message. If an error occurs, the error bit is
                * lit in the response.
                */
               if (peer->cmmd != NULL) {
                       u_int32 temp32;

                       temp32 = CRYPTO_RESP;
                       peer->cmmd->opcode |= htonl(temp32);
                       sendlen += crypto_xmit(peer, &xpkt, NULL,
                           sendlen, peer->cmmd, 0);
                       free(peer->cmmd);
                       peer->cmmd = NULL;
               }

               /*
                * Add an extension field created above. All but the
                * autokey response message are request messages.
                */
               if (exten != NULL) {
                       if (exten->opcode != 0)
                               sendlen += crypto_xmit(peer, &xpkt,
                                   NULL, sendlen, exten, 0);
                       free(exten);
               }

               /*
                * Calculate the next session key. Since extension
                * fields are present, the cookie value is zero.
                */
               if (sendlen > (int)LEN_PKT_NOMAC) {
                       session_key(&peer->dstadr->sin, &peer->srcadr,
                           xkeyid, 0, 2);
               }
       }
#endif  /* AUTOKEY */

       /*
        * Transmit a-priori timestamps
        */
       get_systime(&xmt_tx);
       if (peer->flip == 0) {          /* basic mode */
               peer->aorg = xmt_tx;
               HTONL_FP(&xmt_tx, &xpkt.xmt);
       } else {                        /* interleaved modes */
               if (peer->hmode == MODE_BROADCAST) { /* bcst */
                       HTONL_FP(&xmt_tx, &xpkt.xmt);
                       if (peer->flip > 0)
                               HTONL_FP(&peer->borg, &xpkt.org);
                       else
                               HTONL_FP(&peer->aorg, &xpkt.org);
               } else {                /* symmetric */
                       if (peer->flip > 0)
                               HTONL_FP(&peer->borg, &xpkt.xmt);
                       else
                               HTONL_FP(&peer->aorg, &xpkt.xmt);
               }
       }
       xkeyid = peer->keyid;
       authlen = authencrypt(xkeyid, (u_int32 *)&xpkt, sendlen);
       if (authlen == 0) {
               report_event(PEVNT_AUTH, peer, "no key");
               peer->flash |= TEST5;           /* auth error */
               peer->badauth++;
               return;
       }
       sendlen += authlen;
#ifdef AUTOKEY
       if (xkeyid > NTP_MAXKEY)
               authtrust(xkeyid, 0);
#endif  /* AUTOKEY */
       if (sendlen > sizeof(xpkt)) {
               msyslog(LOG_ERR, "peer_xmit: buffer overflow %u", (u_int)sendlen);
               exit(EX_SOFTWARE);
       }
       peer->t21_bytes = sendlen;
       sendpkt(&peer->srcadr, peer->dstadr,
               sys_ttl[(peer->ttl >= sys_ttlmax) ? sys_ttlmax : peer->ttl],
               &xpkt, sendlen);
       peer->sent++;
       peer->throttle += (1 << peer->minpoll) - 2;

       /*
        * Capture a-posteriori timestamps
        */
       get_systime(&xmt_ty);
       if (peer->flip != 0) {                  /* interleaved modes */
               if (peer->flip > 0)
                       peer->aorg = xmt_ty;
               else
                       peer->borg = xmt_ty;
               peer->flip = -peer->flip;
       }
       L_SUB(&xmt_ty, &xmt_tx);
       LFPTOD(&xmt_ty, peer->xleave);
#ifdef AUTOKEY
       DPRINTF(1, ("peer_xmit: at %ld %s->%s mode %d keyid %08x len %zu index %d\n",
                   current_time, latoa(peer->dstadr), stoa(&peer->srcadr),
                   peer->hmode, xkeyid, sendlen, peer->keynumber));
#else   /* !AUTOKEY follows */
       DPRINTF(1, ("peer_xmit: at %ld %s->%s mode %d keyid %08x len %zu\n",
                   current_time, peer->dstadr ?
                   ntoa(&peer->dstadr->sin) : "-",
                   ntoa(&peer->srcadr), peer->hmode, xkeyid, sendlen));
#endif  /* !AUTOKEY */

       return;
}


#ifdef LEAP_SMEAR

static void
leap_smear_add_offs(
       l_fp *t,
       l_fp *t_recv
       )
{

       L_ADD(t, &leap_smear.offset);

       /*
       ** XXX: Should the smear be added to the root dispersion?
       */

       return;
}

#endif /* LEAP_SMEAR */


/*
* fast_xmit - Send packet for nonpersistent association. Note that
* neither the source or destination can be a broadcast address.
*/
static void
fast_xmit(
       struct recvbuf* rbufp,  /* receive packet pointer */
       int     xmode,          /* receive mode */  /* XXX: HMS: really? */
       keyid_t xkeyid,         /* transmit key ID */
       int     flags           /* restrict mask */
)
{
       struct pkt xpkt;        /* transmit packet structure */
       struct pkt* rpkt;       /* receive packet structure */
       l_fp    xmt_tx, xmt_ty;
       size_t  sendlen;
#ifdef AUTOKEY
       u_int32 temp32;
#endif

       /*
        * Initialize transmit packet header fields from the receive
        * buffer provided. We leave the fields intact as received, but
        * set the peer poll at the maximum of the receive peer poll and
        * the system minimum poll (ntp_minpoll). This is for KoD rate
        * control and not strictly specification compliant, but doesn't
        * break anything.
        */
       rpkt = &rbufp->recv_pkt;
       /*
        * If the packet was received on an endpoint open only on
        * a multicast address, the response needs to go out from
        * a unicast endpoint.
        */
#ifndef MULTICAST_NONEWSOCKET
       if (rbufp->dstadr->flags & INT_MCASTOPEN) {
               rbufp->dstadr = findinterface(&rbufp->recv_srcadr);
               if (NULL == rbufp->dstadr ||
                   ANY_INTERFACE_CHOOSE(&rbufp->recv_srcadr) /* wildcard */
                     == rbufp->dstadr) {
                       DPRINTF(2, ("No unicast local address found for"
                                   " reply to %s mcast.",
                                   stoa(&rbufp->recv_srcadr)));
                       return;
               }
       }
#endif

       /*
        * If this is a kiss-o'-death (KoD) packet, show leap
        * unsynchronized, stratum zero, reference ID the four-character
        * kiss code and (???) system root delay. Note we don't reveal
        * the local time, so these packets can't be used for
        * synchronization.
        */
       if (flags & RES_KOD) {
               sys_kodsent++;
               xpkt.li_vn_mode = PKT_LI_VN_MODE(LEAP_NOTINSYNC,
                   PKT_VERSION(rpkt->li_vn_mode), xmode);
               xpkt.stratum = STRATUM_PKT_UNSPEC;
               xpkt.ppoll = max(rpkt->ppoll, ntp_minpoll);
               xpkt.precision = rpkt->precision;
               memcpy(&xpkt.refid, "RATE", 4);
               xpkt.rootdelay = rpkt->rootdelay;
               xpkt.rootdisp = rpkt->rootdisp;
               xpkt.reftime = rpkt->reftime;
               xpkt.org = rpkt->xmt;
               xpkt.rec = rpkt->xmt;
               xpkt.xmt = rpkt->xmt;

       /*
        * This is a normal packet. Use the system variables.
        */
       } else {
               double this_rootdisp;
               l_fp this_ref_time;

#ifdef LEAP_SMEAR
               /*
                * Make copies of the variables which can be affected by smearing.
                */
               l_fp this_recv_time;
#endif

               /*
                * If we are inside the leap smear interval we add
                * the current smear offset to:
                * - the packet receive time,
                * - the packet transmit time,
                * - and eventually to the reftime to make sure the
                *   reftime isn't later than the transmit/receive times.
                */
               xpkt.li_vn_mode = PKT_LI_VN_MODE(xmt_leap,
                   PKT_VERSION(rpkt->li_vn_mode), xmode);

               xpkt.stratum = STRATUM_TO_PKT(sys_stratum);
               xpkt.ppoll = max(rpkt->ppoll, ntp_minpoll);
               xpkt.precision = sys_precision;
               xpkt.refid = sys_refid;
               xpkt.rootdelay = HTONS_FP(DTOFP(sys_rootdelay));

               /*
               ** Server Response Fuzzing
               **
               ** Which values do we want to use for reftime and rootdisp?
               */

               if (   MODE_SERVER == xmode
                   && RES_SRVRSPFUZ & flags) {
                       if (current_time < p2_time) {
                               this_ref_time = p2_reftime;
                               this_rootdisp = p2_rootdisp;
                       } else if (current_time < prev_time) {
                               this_ref_time = prev_reftime;
                               this_rootdisp = prev_rootdisp;
                       } else {
                               this_ref_time = sys_reftime;
                               this_rootdisp = sys_rootdisp;
                       }

                       SRVRSP_FUZZ(this_ref_time);
               } else {
                       this_ref_time = sys_reftime;
                       this_rootdisp = sys_rootdisp;
               }

               /*
               ** ROOT DISPERSION
               */

               xpkt.rootdisp = HTONS_FP(DTOUFP(this_rootdisp));

               /*
               ** REFTIME
               */

#ifdef LEAP_SMEAR
               if (leap_smear.in_progress) {
                       /* adjust the reftime by the same amount as the
                        * leap smear, as we don't want to risk the
                        * reftime being later than the transmit time.
                        */
                       leap_smear_add_offs(&this_ref_time, NULL);
               }
#endif

               HTONL_FP(&this_ref_time, &xpkt.reftime);

               /*
               ** REFID
               */

#ifdef LEAP_SMEAR
               if (leap_smear.in_progress) {
                       xpkt.refid = convertLFPToRefID(leap_smear.offset);
                       DPRINTF(2, ("fast_xmit: leap_smear.in_progress: refid %8x, smear %s\n",
                               ntohl(xpkt.refid),
                               lfptoa(&leap_smear.offset, 8)
                               ));
               }
#endif

               /*
               ** ORIGIN
               */

               xpkt.org = rpkt->xmt;

               /*
               ** RECEIVE
               */
#ifdef LEAP_SMEAR
               this_recv_time = rbufp->recv_time;
               if (leap_smear.in_progress)
                       leap_smear_add_offs(&this_recv_time, NULL);
               HTONL_FP(&this_recv_time, &xpkt.rec);
#else
               HTONL_FP(&rbufp->recv_time, &xpkt.rec);
#endif

               /*
               ** TRANSMIT
               */

               get_systime(&xmt_tx);
#ifdef LEAP_SMEAR
               if (leap_smear.in_progress)
                       leap_smear_add_offs(&xmt_tx, &this_recv_time);
#endif
               HTONL_FP(&xmt_tx, &xpkt.xmt);
       }

#ifdef HAVE_NTP_SIGND
       if (flags & RES_MSSNTP) {
               send_via_ntp_signd(rbufp, xmode, xkeyid, flags, &xpkt);
               return;
       }
#endif /* HAVE_NTP_SIGND */

       /*
        * If the received packet contains a MAC, the transmitted packet
        * is authenticated and contains a MAC. If not, the transmitted
        * packet is not authenticated.
        */
       sendlen = LEN_PKT_NOMAC;
       if ((size_t)rbufp->recv_length == sendlen) {
               sendpkt(&rbufp->recv_srcadr, rbufp->dstadr, 0, &xpkt,
                   sendlen);
               DPRINTF(1, ("fast_xmit: at %ld %s->%s mode %d len %lu\n",
                           current_time, stoa(&rbufp->dstadr->sin),
                           stoa(&rbufp->recv_srcadr), xmode,
                           (u_long)sendlen));
               return;
       }

       /*
        * The received packet contains a MAC, so the transmitted packet
        * must be authenticated. For symmetric key cryptography, use
        * the predefined and trusted symmetric keys to generate the
        * cryptosum. For autokey cryptography, use the server private
        * value to generate the cookie, which is unique for every
        * source-destination-key ID combination.
        */
#ifdef AUTOKEY
       if (xkeyid > NTP_MAXKEY) {
               keyid_t cookie;

               /*
                * The only way to get here is a reply to a legitimate
                * client request message, so the mode must be
                * MODE_SERVER. If an extension field is present, there
                * can be only one and that must be a command. Do what
                * needs, but with private value of zero so the poor
                * jerk can decode it. If no extension field is present,
                * use the cookie to generate the session key.
                */
               cookie = session_key(&rbufp->recv_srcadr,
                   &rbufp->dstadr->sin, 0, sys_private, 0);
               if ((size_t)rbufp->recv_length > sendlen + MAX_MAC_LEN) {
                       session_key(&rbufp->dstadr->sin,
                           &rbufp->recv_srcadr, xkeyid, 0, 2);
                       temp32 = CRYPTO_RESP;
                       rpkt->exten[0] |= htonl(temp32);
                       sendlen += crypto_xmit(NULL, &xpkt, rbufp,
                           sendlen, (struct exten *)rpkt->exten,
                           cookie);
               } else {
                       session_key(&rbufp->dstadr->sin,
                           &rbufp->recv_srcadr, xkeyid, cookie, 2);
               }
       }
#endif  /* AUTOKEY */
       get_systime(&xmt_tx);
       sendlen += authencrypt(xkeyid, (u_int32 *)&xpkt, sendlen);
#ifdef AUTOKEY
       if (xkeyid > NTP_MAXKEY)
               authtrust(xkeyid, 0);
#endif  /* AUTOKEY */
       sendpkt(&rbufp->recv_srcadr, rbufp->dstadr, 0, &xpkt, sendlen);
       get_systime(&xmt_ty);
       L_SUB(&xmt_ty, &xmt_tx);
       sys_authdelay = xmt_ty;
       DPRINTF(1, ("fast_xmit: at %ld %s->%s mode %d keyid %08x len %lu\n",
                   current_time, ntoa(&rbufp->dstadr->sin),
                   ntoa(&rbufp->recv_srcadr), xmode, xkeyid,
                   (u_long)sendlen));
}


/*
* pool_xmit - resolve hostname or send unicast solicitation for pool.
*/
static void
pool_xmit(
       struct peer *pool       /* pool solicitor association */
       )
{
#ifdef WORKER
       struct pkt      xpkt;   /* transmit packet structure */
       struct addrinfo hints;
       int             rc;
       endpt *         lcladr;
       sockaddr_u *    rmtadr;
       u_short         af;
       struct peer *   p;
       l_fp            xmt_tx;

       DEBUG_REQUIRE(pool);
       if (NULL == pool->ai) {
               if (pool->addrs != NULL) {
                       /* free() is used with copy_addrinfo_list() */
                       free(pool->addrs);
                       pool->addrs = NULL;
               }
               af = AF(&pool->srcadr);
               if (   (AF_INET == af && !nonlocal_v4_addr_up)
                   || (AF_INET6 == af && !nonlocal_v6_addr_up)
                   || (   AF_UNSPEC == af
                       && !nonlocal_v4_addr_up
                       && !nonlocal_v6_addr_up)) {

                       /* POOL DNS query would be useless [Bug 3845] */
                       return;
               }
               ZERO(hints);
               hints.ai_family = AF(&pool->srcadr);
               hints.ai_socktype = SOCK_DGRAM;
               hints.ai_protocol = IPPROTO_UDP;
               /* ignore getaddrinfo_sometime() errors, we will retry */
               rc = getaddrinfo_sometime(
                       pool->hostname,
                       "ntp",
                       &hints,
                       0,                      /* no retry */
                       &pool_name_resolved,
                       (void *)(intptr_t)pool->associd);
               if (!rc)
                       DPRINTF(1, ("pool DNS lookup %s started\n",
                               pool->hostname));
               else
                       msyslog(LOG_ERR,
                               "unable to start pool DNS %s: %m",
                               pool->hostname);
               return;
       }

       do {
               /* copy_addrinfo_list ai_addr points to a sockaddr_u */
               rmtadr = (sockaddr_u *)(void *)pool->ai->ai_addr;
               pool->ai = pool->ai->ai_next;
               /* do not solicit when hopeless [Bug 3845] */
               if (   (IS_IPV4(rmtadr) && !nonlocal_v4_addr_up)
                   || (IS_IPV6(rmtadr) && !nonlocal_v6_addr_up)) {
                       continue;
               }
               p = findexistingpeer(rmtadr, NULL, NULL, MODE_CLIENT, 0, NULL);
       } while (p != NULL && pool->ai != NULL);
       if (p != NULL) {
               return; /* out of addresses, re-query DNS next poll */
       }
       restrict_source(rmtadr, FALSE, 1 + POOL_SOLICIT_WINDOW);
       lcladr = findinterface(rmtadr);
       memset(&xpkt, 0, sizeof(xpkt));
       xpkt.li_vn_mode = PKT_LI_VN_MODE(sys_leap, pool->version,
                                        MODE_CLIENT);
       xpkt.stratum = STRATUM_TO_PKT(sys_stratum);
       xpkt.ppoll = pool->hpoll;
       xpkt.precision = sys_precision;
       xpkt.refid = sys_refid;
       xpkt.rootdelay = HTONS_FP(DTOFP(sys_rootdelay));
       xpkt.rootdisp = HTONS_FP(DTOUFP(sys_rootdisp));
       /* Bug 3596: What are the pros/cons of using sys_reftime here? */
       HTONL_FP(&sys_reftime, &xpkt.reftime);

       get_systime(&xmt_tx);
       pool->aorg = xmt_tx;

       if (FLAG_LOOPNONCE & pool->flags) {
               l_fp nonce;

               do {
                       nonce.l_ui = ntp_random();
               } while (0 == nonce.l_ui);
               do {
                       nonce.l_uf = ntp_random();
               } while (0 == nonce.l_uf);
               pool->nonce = nonce;
               HTONL_FP(&nonce, &xpkt.xmt);
       } else {
               L_CLR(&pool->nonce);
               HTONL_FP(&xmt_tx, &xpkt.xmt);
       }
       pool->sent++;
       pool->throttle += (1 << pool->minpoll) - 2;
       DPRINTF(1, ("pool_xmit: at %ld %s->%s pool\n",
                   current_time, latoa(lcladr), stoa(rmtadr)));
       msyslog(LOG_INFO, "Soliciting pool server %s", stoa(rmtadr));
       sendpkt(rmtadr, lcladr,
               sys_ttl[(pool->ttl >= sys_ttlmax) ? sys_ttlmax : pool->ttl],
               &xpkt, LEN_PKT_NOMAC);
#endif  /* WORKER */
}


#ifdef AUTOKEY
       /*
        * group_test - test if this is the same group
        *
        * host         assoc           return          action
        * none         none            0               mobilize *
        * none         group           0               mobilize *
        * group        none            0               mobilize *
        * group        group           1               mobilize
        * group        different       1               ignore
        * * ignore if notrust
        */
int
group_test(
       char    *grp,
       char    *ident
       )
{
       if (grp == NULL)
               return (0);

       if (strcmp(grp, sys_groupname) == 0)
               return (0);

       if (ident == NULL)
               return (1);

       if (strcmp(grp, ident) == 0)
               return (0);

       return (1);
}
#endif /* AUTOKEY */


#ifdef WORKER
void
pool_name_resolved(
       int                     rescode,
       int                     gai_errno,
       void *                  context,
       const char *            name,
       const char *            service,
       const struct addrinfo * hints,
       const struct addrinfo * res
       )
{
       struct peer *   pool;   /* pool solicitor association */
       associd_t       assoc;

       if (rescode) {
               msyslog(LOG_ERR,
                       "error resolving pool %s: %s (%d)",
                       name, gai_strerror(rescode), rescode);
               return;
       }

       assoc = (associd_t)(intptr_t)context;
       pool = findpeerbyassoc(assoc);
       if (NULL == pool) {
               msyslog(LOG_ERR,
                       "Could not find assoc %u for pool DNS %s",
                       assoc, name);
               return;
       }
       DPRINTF(1, ("pool DNS %s completed\n", name));
       pool->addrs = copy_addrinfo_list(res);
       pool->ai = pool->addrs;
       pool_xmit(pool);

}
#endif  /* WORKER */


#ifdef AUTOKEY
/*
* key_expire - purge the key list
*/
void
key_expire(
       struct peer *peer       /* peer structure pointer */
       )
{
       int i;

       if (peer->keylist != NULL) {
               for (i = 0; i <= peer->keynumber; i++)
                       authtrust(peer->keylist[i], 0);
               free(peer->keylist);
               peer->keylist = NULL;
       }
       value_free(&peer->sndval);
       peer->keynumber = 0;
       peer->flags &= ~FLAG_ASSOC;
       DPRINTF(1, ("key_expire: at %lu associd %d\n", current_time,
                   peer->associd));
}
#endif  /* AUTOKEY */


/*
* local_refid(peer) - Check peer refid to avoid selecting peers
*                     currently synced to this ntpd.
* Note that until 4.2.8p18 and 4.3.1XX ntpd calculated the IPv6
* refid differently on different-endian systems.  It now calculates
* the refid the same on both, the same way it did on little-endian
* in the past.  On big-endian systems, ntpd also calculates a
* byte-swapped version of each of its IPv6 local addresses' refids,
* as endpt.old_refid and also detects a loop when seeing it.  This
* ensures new BE ntpd will detect loops interoperating with older
* BE ntpd, and keeps the more-common LE old ntpd code detecting
* loops with IPv6 refids correctly.  Thanks to Hal Murray for
* the byte-swapping idea.
*/
static int
local_refid(
       struct peer *   p
       )
{
       endpt * unicast_ep;

       if (p->dstadr != NULL && !(INT_MCASTIF & p->dstadr->flags))
               unicast_ep = p->dstadr;
       else
               unicast_ep = findinterface(&p->srcadr);

       if (unicast_ep != NULL
           && (   p->refid == unicast_ep->addr_refid
#ifdef WORDS_BIGENDIAN
               || (   IS_IPV6(&unicast_ep->sin)
                   && p->refid == unicast_ep->old_refid)
#endif
                                                        )) {
               return TRUE;
       } else {
               return FALSE;
       }
}


/*
* Determine if the peer is unfit for synchronization
*
* A peer is unfit for synchronization if
* > TEST10 bad leap or stratum below floor or at or above ceiling
* > TEST11 root distance exceeded for remote peer
* > TEST12 a direct or indirect synchronization loop would form
* > TEST13 unreachable or noselect
*/
int                             /* FALSE if fit, TRUE if unfit */
peer_unfit(
       struct peer *peer       /* peer structure pointer */
       )
{
       int     rval = 0;

       /*
        * A stratum error occurs if (1) the server has never been
        * synchronized, (2) the server stratum is below the floor or
        * greater than or equal to the ceiling.
        */
       if (   peer->leap == LEAP_NOTINSYNC
           || peer->stratum < sys_floor
           || peer->stratum >= sys_ceiling) {
               rval |= TEST10;         /* bad synch or stratum */
       }

       /*
        * A distance error for a remote peer occurs if the root
        * distance is greater than or equal to the distance threshold
        * plus the increment due to one host poll interval.
        */
       if (   !(peer->flags & FLAG_REFCLOCK)
           && root_distance(peer) >= sys_maxdist
                                     + clock_phi * ULOGTOD(peer->hpoll)) {
               rval |= TEST11;         /* distance exceeded */
       }

       /*
        * A loop error occurs if the remote peer is synchronized to the
        * local peer or if the remote peer is synchronized to the same
        * server as the local peer but only if the remote peer is
        * neither a reference clock nor an orphan.
        */
       if (peer->stratum > 1 && local_refid(peer)) {
               rval |= TEST12;         /* synchronization loop */
       }

       /*
        * An unreachable error occurs if the server is unreachable or
        * the noselect bit is set.
        */
       if (!peer->reach || (peer->flags & FLAG_NOSELECT)) {
               rval |= TEST13;         /* unreachable */
       }

       peer->flash &= ~PEER_TEST_MASK;
       peer->flash |= rval;
       return (rval);
}


/*
* Find the precision of this particular machine
*/
#define MINSTEP         20e-9   /* minimum clock increment (s) */
#define MAXSTEP         1       /* maximum clock increment (s) */
#define MINCHANGES      12      /* minimum number of step samples */
#define MAXLOOPS        ((int)(1. / MINSTEP))   /* avoid infinite loop */

/*
* This routine measures the system precision defined as the minimum of
* a sequence of differences between successive readings of the system
* clock. However, if a difference is less than MINSTEP, the clock has
* been read more than once during a clock tick and the difference is
* ignored. We set MINSTEP greater than zero in case something happens
* like a cache miss, and to tolerate underlying system clocks which
* ensure each reading is strictly greater than prior readings while
* using an underlying stepping (not interpolated) clock.
*
* sys_tick and sys_precision represent the time to read the clock for
* systems with high-precision clocks, and the tick interval or step
* size for lower-precision stepping clocks.
*
* This routine also measures the time to read the clock on stepping
* system clocks by counting the number of readings between changes of
* the underlying clock.  With either type of clock, the minimum time
* to read the clock is saved as sys_fuzz, and used to ensure the
* get_systime() readings always increase and are fuzzed below sys_fuzz.
*/
void
measure_precision(void)
{
       /*
        * With sys_fuzz set to zero, get_systime() fuzzing of low bits
        * is effectively disabled.  trunc_os_clock is FALSE to disable
        * get_ostime() simulation of a low-precision system clock.
        */
       set_sys_fuzz(0.);
       trunc_os_clock = FALSE;
       measured_tick = measure_tick_fuzz();
       set_sys_tick_precision(measured_tick);
       msyslog(LOG_INFO, "proto: precision = %.3f usec (%d)",
               sys_tick * 1e6, sys_precision);
       if (sys_fuzz < sys_tick) {
               msyslog(LOG_NOTICE, "proto: fuzz beneath %.3f usec",
                       sys_fuzz * 1e6);
       }
}


/*
* measure_tick_fuzz()
*
* measures the minimum time to read the clock (stored in sys_fuzz)
* and returns the tick, the larger of the minimum increment observed
* between successive clock readings and the time to read the clock.
*/
double
measure_tick_fuzz(void)
{
       l_fp    minstep;        /* MINSTEP as l_fp */
       l_fp    val;            /* current seconds fraction */
       l_fp    last;           /* last seconds fraction */
       l_fp    ldiff;          /* val - last */
       double  tick;           /* computed tick value */
       double  diff;
       long    repeats;
       long    max_repeats;
       int     changes;
       int     i;              /* log2 precision */

       tick = MAXSTEP;
       max_repeats = 0;
       repeats = 0;
       changes = 0;
       DTOLFP(MINSTEP, &minstep);
       get_systime(&last);
       for (i = 0; i < MAXLOOPS && changes < MINCHANGES; i++) {
               get_systime(&val);
               ldiff = val;
               L_SUB(&ldiff, &last);
               last = val;
               if (L_ISGT(&ldiff, &minstep)) {
                       max_repeats = max(repeats, max_repeats);
                       repeats = 0;
                       changes++;
                       LFPTOD(&ldiff, diff);
                       tick = min(diff, tick);
               } else {
                       repeats++;
               }
       }
       if (changes < MINCHANGES) {
               msyslog(LOG_ERR, "Fatal error: precision could not be measured (MINSTEP too large?)");
               exit(1);
       }

       if (0 == max_repeats) {
               set_sys_fuzz(tick);
       } else {
               set_sys_fuzz(tick / max_repeats);
       }

       return tick;
}


void
set_sys_tick_precision(
       double tick
       )
{
       int i;

       if (tick > 1.) {
               msyslog(LOG_ERR,
                       "unsupported tick %.3f > 1s ignored", tick);
               return;
       }
       if (tick < measured_tick) {
               msyslog(LOG_ERR,
                       "proto: tick %.3f less than measured tick %.3f, ignored",
                       tick, measured_tick);
               return;
       } else if (tick > measured_tick) {
               trunc_os_clock = TRUE;
               msyslog(LOG_NOTICE,
                       "proto: truncating system clock to multiples of %.9f",
                       tick);
       }
       sys_tick = tick;

       /*
        * Find the nearest power of two.
        */
       for (i = 0; tick <= 1; i--)
               tick *= 2;
       if (tick - 1 > 1 - tick / 2)
               i++;

       sys_precision = (s_char)i;
}


/*
* init_proto - initialize the protocol module's data
*/
void
init_proto(void)
{
       l_fp    dummy;
       int     i;

       /*
        * Fill in the sys_* stuff.  Default is don't listen to
        * broadcasting, require authentication.
        */
       set_sys_leap(LEAP_NOTINSYNC);
       sys_stratum = STRATUM_UNSPEC;
       memcpy(&sys_refid, "INIT", 4);
       sys_peer = NULL;
       sys_rootdelay = 0;
       sys_rootdisp = 0;
       L_CLR(&sys_reftime);
       sys_jitter = 0;
       measure_precision();
       get_systime(&dummy);
       sys_survivors = 0;
       sys_manycastserver = 0;
       sys_bclient = 0;
       sys_mclient = 0;
       sys_bdelay = BDELAY_DEFAULT;    /*[Bug 3031] delay cutoff */
       sys_authenticate = 1;
       sys_stattime = current_time;
       orphwait = current_time + sys_orphwait;
       proto_clr_stats();
       for (i = 0; i < MAX_TTL; ++i)
               sys_ttl[i] = (u_char)((i * 256) / MAX_TTL);
       sys_ttlmax = (MAX_TTL - 1);
       hardpps_enable = 0;
       stats_control = 1;
}


/*
* proto_config - configure the protocol module
*/
void
proto_config(
       int     item,
       u_long  value,
       double  dvalue,
       sockaddr_u *svalue
       )
{
       /*
        * Figure out what he wants to change, then do it
        */
       DPRINTF(2, ("proto_config: code %d value %lu dvalue %lf\n",
                   item, value, dvalue));

       switch (item) {

       /*
        * enable and disable commands - arguments are Boolean.
        */
       case PROTO_AUTHENTICATE: /* authentication (auth) */
               sys_authenticate = value;
               break;

       case PROTO_BROADCLIENT: /* broadcast client (bclient) */
               sys_bclient = (int)value;
               if (!sys_bclient)
                       io_unsetbclient();
               else
                       io_setbclient();
               break;

#ifdef REFCLOCK
       case PROTO_CAL:         /* refclock calibrate (calibrate) */
               cal_enable = value;
               break;
#endif /* REFCLOCK */

       case PROTO_KERNEL:      /* kernel discipline (kernel) */
               select_loop(value);
               break;

       case PROTO_MONITOR:     /* monitoring (monitor) */
               if (value)
                       mon_start(MON_ON);
               else {
                       mon_stop(MON_ON);
                       if (mon_enabled)
                               msyslog(LOG_WARNING,
                                       "restrict: 'monitor' cannot be disabled while 'limited' is enabled");
               }
               break;

       case PROTO_NTP:         /* NTP discipline (ntp) */
               ntp_enable = value;
               break;

       case PROTO_MODE7:       /* mode7 management (ntpdc) */
               ntp_mode7 = value;
               break;

       case PROTO_PPS:         /* PPS discipline (pps) */
               hardpps_enable = value;
               break;

       case PROTO_FILEGEN:     /* statistics (stats) */
               stats_control = value;
               break;

       /*
        * tos command - arguments are double, sometimes cast to int
        */

       case PROTO_BCPOLLBSTEP: /* Broadcast Poll Backstep gate (bcpollbstep) */
               sys_bcpollbstep = (u_char)dvalue;
               break;

       case PROTO_BEACON:      /* manycast beacon (beacon) */
               sys_beacon = (int)dvalue;
               break;

       case PROTO_BROADDELAY:  /* default broadcast delay (bdelay) */
               sys_bdelay = (dvalue ? dvalue : BDELAY_DEFAULT);
               break;

       case PROTO_CEILING:     /* stratum ceiling (ceiling) */
               sys_ceiling = (int)dvalue;
               break;

       case PROTO_COHORT:      /* cohort switch (cohort) */
               sys_cohort = (int)dvalue;
               break;

       case PROTO_FLOOR:       /* stratum floor (floor) */
               sys_floor = (int)dvalue;
               break;

       case PROTO_MAXCLOCK:    /* maximum candidates (maxclock) */
               sys_maxclock = (int)dvalue;
               break;

       case PROTO_MAXDIST:     /* select threshold (maxdist) */
               sys_maxdist = dvalue;
               break;

       case PROTO_CALLDELAY:   /* modem call delay (mdelay) */
               break;          /* NOT USED */

       case PROTO_MINCLOCK:    /* minimum candidates (minclock) */
               sys_minclock = (int)dvalue;
               break;

       case PROTO_MINDISP:     /* minimum distance (mindist) */
               sys_mindisp = dvalue;
               break;

       case PROTO_MINSANE:     /* minimum survivors (minsane) */
               sys_minsane = (int)dvalue;
               break;

       case PROTO_ORPHAN:      /* orphan stratum (orphan) */
               sys_orphan = (int)dvalue;
               break;

       case PROTO_ORPHWAIT:    /* orphan wait (orphwait) */
               orphwait -= sys_orphwait;
               sys_orphwait = (dvalue >= 1) ? (int)dvalue : NTP_ORPHWAIT;
               orphwait += sys_orphwait;
               break;

       /*
        * Miscellaneous commands
        */
       case PROTO_MULTICAST_ADD: /* add group address */
               if (svalue != NULL)
                       io_multicast_add(svalue);
               sys_mclient = 1;
               break;

       case PROTO_MULTICAST_DEL: /* delete group address */
               if (svalue != NULL)
                       io_multicast_del(svalue);
               break;

       /*
        * Peer_clear Early policy choices
        */

       case PROTO_PCEDIGEST:   /* Digest */
               peer_clear_digest_early = value;
               break;

       /*
        * Unpeer Early policy choices
        */

       case PROTO_UECRYPTO:    /* Crypto */
               unpeer_crypto_early = value;
               break;

       case PROTO_UECRYPTONAK: /* Crypto_NAK */
               unpeer_crypto_nak_early = value;
               break;

       case PROTO_UEDIGEST:    /* Digest */
               unpeer_digest_early = value;
               break;

       default:
               msyslog(LOG_NOTICE,
                   "proto: unsupported option %d", item);
       }
}


/*
* proto_clr_stats - clear protocol stat counters
*/
void
proto_clr_stats(void)
{
       sys_stattime = current_time;
       sys_received = 0;
       sys_processed = 0;
       sys_newversion = 0;
       sys_oldversion = 0;
       sys_declined = 0;
       sys_restricted = 0;
       sys_badlength = 0;
       sys_badauth = 0;
       sys_limitrejected = 0;
       sys_kodsent = 0;
       sys_lamport = 0;
       sys_tsrounding = 0;
}