2 * Copyright (c) 1983, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 #include "pathnames.h"
41 __COPYRIGHT("@(#) Copyright (c) 1983, 1988, 1993 "
42 "The Regents of the University of California."
43 " All rights reserved.");
47 #elif defined(__FreeBSD__)
50 __RCSID("$Revision: 2.31 $");
51 #ident "$Revision: 2.31 $"
56 naddr myaddr; /* system address */
57 static char myname[MAXHOSTNAMELEN+1];
61 int supplier; /* supply or broadcast updates */
63 static int ipforwarding = 1; /* kernel forwarding on */
65 static int default_gateway; /* 1=advertise default */
66 static int background = 1;
67 int ridhosts; /* 1=reduce host routes */
68 int mhome; /* 1=want multi-homed host route */
69 int advertise_mhome; /* 1=must continue advertising it */
70 int auth_ok = 1; /* 1=ignore auth if we do not care */
72 struct timeval epoch; /* when started */
74 static struct timeval prev_clk;
75 static int usec_fudge;
76 struct timeval now; /* current idea of time */
81 static struct timeval next_bcast; /* next general broadcast */
82 struct timeval no_flash = { /* inhibit flash update */
83 EPOCH+SUPPLY_INTERVAL, 0
86 static struct timeval flush_kern_timer;
90 int rip_sock = -1; /* RIP socket */
91 const struct interface *rip_sock_mcast; /* current multicast interface */
92 int rt_sock; /* routing socket */
96 static int get_rip_sock(naddr, int);
97 static void timevalsub(struct timeval *, struct timeval *, struct timeval *);
98 static void sigalrm(int s UNUSED);
99 static void sigterm(int sig);
109 struct timeval wtime, t2;
113 struct interface *ifp;
118 /* Some shells are badly broken and send SIGHUP to backgrounded
121 signal(SIGHUP, SIG_IGN);
123 openlog("routed", LOG_PID, LOG_DAEMON);
126 gettimeofday(&clk, 0);
129 epoch.tv_sec -= EPOCH;
131 now_stale = EPOCH - STALE_TIME;
132 now_expire = EPOCH - EXPIRE_TIME;
133 now_garbage = EPOCH - GARBAGE_TIME;
136 (void)gethostname(myname, sizeof(myname)-1);
137 (void)gethost(myname, &myaddr);
139 while ((n = getopt(argc, argv, "sqdghmAtvT:F:P:")) != -1) {
156 memset(&parm, 0, sizeof(parm));
157 parm.parm_d_metric = 1;
158 cp = check_parms(&parm);
160 msglog("bad -g: %s", cp);
165 case 'h': /* suppress extra host routes */
169 case 'm': /* advertise host route */
170 mhome = 1; /* on multi-homed hosts */
174 /* Ignore authentication if we do not care.
175 * Crazy as it is, that is what RFC 1723 requires.
188 case 'F': /* minimal routes for SLIP */
190 p = strchr(optarg,',');
191 if (p && *p != '\0') {
192 n = (int)strtoul(p+1, &q, 0);
194 && n <= HOPCNT_INFINITY-1
198 if (!getnet(optarg, &p_net, &p_mask)) {
199 msglog("bad network; \"-F %s\"",
203 memset(&parm, 0, sizeof(parm));
204 parm.parm_net = p_net;
205 parm.parm_mask = p_mask;
206 parm.parm_d_metric = n;
207 cp = check_parms(&parm);
209 msglog("bad -F: %s", cp);
213 /* handle arbitrary parameters.
216 cp = parse_parms(q, 0);
218 msglog("%s in \"-P %s\"", cp, optarg);
223 /* display version */
225 msglog("version 2.31");
235 if (tracename == 0 && argc >= 1) {
239 if (tracename != 0 && tracename[0] == '\0')
243 logbad(0, "usage: routed [-sqdghmAtv] [-T tracefile]"
244 " [-F net[,metric]] [-P parms]");
246 if (geteuid() != 0) {
249 logbad(0, "requires UID 0");
255 mib[3] = IPCTL_FORWARDING;
256 len = sizeof(ipforwarding);
257 if (sysctl(mib, 4, &ipforwarding, &len, 0, 0) < 0)
258 LOGERR("sysctl(IPCTL_FORWARDING)");
262 msglog("-s incompatible with ipforwarding=0");
263 if (default_gateway) {
264 msglog("-g incompatible with ipforwarding=0");
270 if (default_gateway) {
271 if (supplier_set && !supplier) {
272 msglog("-g and -q incompatible");
280 signal(SIGALRM, sigalrm);
282 signal(SIGHUP, sigterm); /* SIGHUP fatal during debugging */
283 signal(SIGTERM, sigterm);
284 signal(SIGINT, sigterm);
285 signal(SIGUSR1, sigtrace_on);
286 signal(SIGUSR2, sigtrace_off);
288 /* get into the background */
290 if (0 > _daemonize(background ? 0 : (_DF_NOCHDIR|_DF_NOFORK),
291 STDIN_FILENO, STDOUT_FILENO, STDERR_FILENO))
292 BADERR(0, "_daemonize()");
294 if (background && daemon(0, 1) < 0)
295 BADERR(0,"daemon()");
298 #if defined(__NetBSD__)
305 srandom((int)(clk.tv_sec ^ clk.tv_usec ^ mypid));
308 /* prepare socket connected to the kernel.
310 rt_sock = socket(AF_ROUTE, SOCK_RAW, 0);
312 BADERR(1,"rt_sock = socket()");
313 if (fcntl(rt_sock, F_SETFL, O_NONBLOCK) == -1)
314 logbad(1, "fcntl(rt_sock) O_NONBLOCK: %s", strerror(errno));
316 if (setsockopt(rt_sock, SOL_SOCKET,SO_USELOOPBACK,
317 &off,sizeof(off)) < 0)
318 LOGERR("setsockopt(SO_USELOOPBACK,0)");
323 if (tracename != 0) {
324 strncpy(inittracename, tracename, sizeof(inittracename)-1);
325 set_tracefile(inittracename, "%s", -1);
327 tracelevel_msg("%s", -1); /* turn on tracing to stdio */
332 /* initialize radix tree */
335 /* Pick a random part of the second for our output to minimize
338 * Start broadcasting after hearing from other routers, and
339 * at a random time so a bunch of systems do not get synchronized
340 * after a power failure.
342 intvl_random(&next_bcast, EPOCH+MIN_WAITTIME, EPOCH+SUPPLY_INTERVAL);
343 age_timer.tv_usec = next_bcast.tv_usec;
344 age_timer.tv_sec = EPOCH+MIN_WAITTIME;
345 rdisc_timer = next_bcast;
346 ifinit_timer.tv_usec = next_bcast.tv_usec;
348 /* Collect an initial view of the world by checking the interface
349 * configuration and the kludge file.
358 /* Now turn off stdio if not tracing */
359 if (new_tracelevel == 0)
360 trace_close(background);
362 /* Loop forever, listening and broadcasting.
366 gettimeofday(&clk, 0);
367 if (prev_clk.tv_sec == clk.tv_sec
368 && prev_clk.tv_usec == clk.tv_usec+usec_fudge) {
369 /* Much of `routed` depends on time always advancing.
370 * On systems that do not guarantee that gettimeofday()
371 * produces unique timestamps even if called within
372 * a single tick, use trickery like that in classic
375 clk.tv_usec += ++usec_fudge;
380 timevalsub(&t2, &clk, &prev_clk);
382 || t2.tv_sec > wtime.tv_sec + 5) {
383 /* Deal with time changes before other
384 * housekeeping to keep everything straight.
389 trace_act("time changed by %d sec", (int)dt);
393 timevalsub(&now, &clk, &epoch);
394 now_stale = now.tv_sec - STALE_TIME;
395 now_expire = now.tv_sec - EXPIRE_TIME;
396 now_garbage = now.tv_sec - GARBAGE_TIME;
398 /* deal with signals that should affect tracing */
404 trace_off("exiting with signal %d", stopint);
408 /* look for new or dead interfaces */
409 timevalsub(&wtime, &ifinit_timer, &now);
410 if (wtime.tv_sec <= 0) {
417 /* Check the kernel table occassionally for mysteriously
420 timevalsub(&t2, &flush_kern_timer, &now);
421 if (t2.tv_sec <= 0) {
423 flush_kern_timer.tv_sec = (now.tv_sec
424 + CHECK_QUIET_INTERVAL);
427 if (timercmp(&t2, &wtime, <))
430 /* If it is time, then broadcast our routes.
432 if (supplier || advertise_mhome) {
433 timevalsub(&t2, &next_bcast, &now);
434 if (t2.tv_sec <= 0) {
435 /* Synchronize the aging and broadcast
436 * timers to minimize awakenings
442 /* It is desirable to send routing updates
443 * regularly. So schedule the next update
444 * 30 seconds after the previous one was
445 * scheduled, instead of 30 seconds after
446 * the previous update was finished.
447 * Even if we just started after discovering
448 * a 2nd interface or were otherwise delayed,
449 * pick a 30-second aniversary of the
450 * original broadcast time.
452 n = 1 + (0-t2.tv_sec)/SUPPLY_INTERVAL;
453 next_bcast.tv_sec += n*SUPPLY_INTERVAL;
458 if (timercmp(&t2, &wtime, <))
462 /* If we need a flash update, either do it now or
463 * set the delay to end when it is time.
465 * If we are within MIN_WAITTIME seconds of a full update,
470 && no_flash.tv_sec+MIN_WAITTIME < next_bcast.tv_sec) {
471 /* accurate to the millisecond */
472 if (!timercmp(&no_flash, &now, >))
474 timevalsub(&t2, &no_flash, &now);
475 if (timercmp(&t2, &wtime, <))
479 /* trigger the main aging timer.
481 timevalsub(&t2, &age_timer, &now);
482 if (t2.tv_sec <= 0) {
486 if (timercmp(&t2, &wtime, <))
489 /* update the kernel routing table
491 timevalsub(&t2, &need_kern, &now);
492 if (t2.tv_sec <= 0) {
496 if (timercmp(&t2, &wtime, <))
499 /* take care of router discovery,
500 * but do it in the correct the millisecond
502 if (!timercmp(&rdisc_timer, &now, >)) {
506 timevalsub(&t2, &rdisc_timer, &now);
507 if (timercmp(&t2, &wtime, <))
511 /* wait for input or a timer to expire.
515 n = select(sock_max, &ibits, 0, 0, &wtime);
517 if (n < 0 && errno != EINTR && errno != EAGAIN)
522 if (FD_ISSET(rt_sock, &ibits)) {
526 if (rdisc_sock >= 0 && FD_ISSET(rdisc_sock, &ibits)) {
530 if (rip_sock >= 0 && FD_ISSET(rip_sock, &ibits)) {
531 read_rip(rip_sock, 0);
535 LIST_FOREACH(ifp, &ifnet, int_list) {
538 if (ifp->int_rip_sock >= 0
539 && FD_ISSET(ifp->int_rip_sock, &ibits)) {
540 read_rip(ifp->int_rip_sock, ifp);
550 sigalrm(int s UNUSED)
552 /* Historically, SIGALRM would cause the daemon to check for
553 * new and broken interfaces.
555 ifinit_timer.tv_sec = now.tv_sec;
556 trace_act("SIGALRM");
560 /* watch for fatal signals */
565 (void)signal(sig, SIG_DFL); /* catch it only once */
572 struct interface *ifp;
578 FD_SET(rt_sock, &fdbits);
579 if (sock_max <= rt_sock)
580 sock_max = rt_sock+1;
582 FD_SET(rip_sock, &fdbits);
583 if (sock_max <= rip_sock)
584 sock_max = rip_sock+1;
586 LIST_FOREACH(ifp, &ifnet, int_list) {
587 if (ifp->int_rip_sock >= 0) {
588 FD_SET(ifp->int_rip_sock, &fdbits);
589 if (sock_max <= ifp->int_rip_sock)
590 sock_max = ifp->int_rip_sock+1;
593 if (rdisc_sock >= 0) {
594 FD_SET(rdisc_sock, &fdbits);
595 if (sock_max <= rdisc_sock)
596 sock_max = rdisc_sock+1;
606 #define MIN_SOCKBUF (4*1024)
609 if (fcntl(sock, F_SETFL, O_NONBLOCK) == -1)
610 logbad(1, "fcntl(%s) O_NONBLOCK: %s",
611 name, strerror(errno));
613 if (setsockopt(sock, SOL_SOCKET,SO_BROADCAST, &on,sizeof(on)) < 0)
614 msglog("setsockopt(%s,SO_BROADCAST): %s",
615 name, strerror(errno));
616 #ifdef USE_PASSIFNAME
618 if (setsockopt(sock, SOL_SOCKET, SO_PASSIFNAME, &on,sizeof(on)) < 0)
619 msglog("setsockopt(%s,SO_PASSIFNAME): %s",
620 name, strerror(errno));
623 if (rbuf >= MIN_SOCKBUF) {
624 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
625 &rbuf, sizeof(rbuf)) < 0)
626 msglog("setsockopt(%s,SO_RCVBUF=%d): %s",
627 name, rbuf, strerror(errno));
629 for (rbuf = 60*1024; ; rbuf -= 4096) {
630 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
631 &rbuf, sizeof(rbuf)) == 0) {
632 trace_act("RCVBUF=%d", rbuf);
635 if (rbuf < MIN_SOCKBUF) {
636 msglog("setsockopt(%s,SO_RCVBUF = %d): %s",
637 name, rbuf, strerror(errno));
647 static int /* <0 or file descriptor */
648 get_rip_sock(naddr addr,
649 int serious) /* 1=failure to bind is serious */
651 struct sockaddr_in rsin;
656 if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
657 BADERR(1,"rip_sock = socket()");
659 memset(&rsin, 0, sizeof(rsin));
661 rsin.sin_len = sizeof(rsin);
663 rsin.sin_family = AF_INET;
664 rsin.sin_port = htons(RIP_PORT);
665 rsin.sin_addr.s_addr = addr;
666 if (bind(s, (struct sockaddr *)&rsin, sizeof(rsin)) < 0) {
668 BADERR(errno != EADDRINUSE, "bind(rip_sock)");
671 fix_sock(s,"rip_sock");
674 if (setsockopt(s, IPPROTO_IP, IP_MULTICAST_TTL,
675 &ttl, sizeof(ttl)) < 0)
676 DBGERR(1,"rip_sock setsockopt(IP_MULTICAST_TTL)");
682 /* turn off main RIP socket */
686 struct interface *ifp;
690 if (rip_sock >= 0 && !mhome) {
691 trace_act("turn off RIP");
693 (void)close(rip_sock);
696 /* get non-broadcast sockets to listen to queries.
698 LIST_FOREACH(ifp, &ifnet, int_list) {
699 if (ifp->int_state & IS_REMOTE)
701 if (ifp->int_rip_sock < 0) {
702 addr = ((ifp->int_if_flags & IFF_POINTOPOINT)
705 ifp->int_rip_sock = get_rip_sock(addr, 0);
716 /* turn on RIP multicast input via an interface
719 rip_mcast_on(struct interface *ifp)
722 struct sockaddr_in *sin;
724 if (!IS_RIP_IN_OFF(ifp->int_state)
725 && (ifp->int_if_flags & IFF_MULTICAST)
726 && !(ifp->int_state & IS_ALIAS)) {
727 memset(&gr, 0, sizeof(gr));
728 gr.gr_interface = ifp->int_index;
729 sin = (struct sockaddr_in *)&gr.gr_group;
730 sin->sin_family = AF_INET;
732 sin->sin_len = sizeof(struct sockaddr_in);
734 sin->sin_addr.s_addr = htonl(INADDR_RIP_GROUP);
735 if (setsockopt(rip_sock, IPPROTO_IP, MCAST_JOIN_GROUP,
736 &gr, sizeof(gr)) < 0)
737 LOGERR("setsockopt(MCAST_JOIN_GROUP RIP)");
742 /* Prepare socket used for RIP.
745 rip_on(struct interface *ifp)
747 /* If the main RIP socket is already alive, only start receiving
748 * multicasts for this interface.
756 /* If the main RIP socket is off and it makes sense to turn it on,
757 * then turn it on for all of the interfaces.
758 * It makes sense if either router discovery is off, or if
759 * router discover is on and at most one interface is doing RIP.
761 if (rip_interfaces > 0 && (!rdisc_ok || rip_interfaces > 1)) {
762 trace_act("turn on RIP");
764 /* Close all of the query sockets so that we can open
765 * the main socket. SO_REUSEPORT is not a solution,
766 * since that would let two daemons bind to the broadcast
769 LIST_FOREACH(ifp, &ifnet, int_list) {
770 if (ifp->int_rip_sock >= 0) {
771 (void)close(ifp->int_rip_sock);
772 ifp->int_rip_sock = -1;
776 rip_sock = get_rip_sock(INADDR_ANY, 1);
779 /* Do not advertise anything until we have heard something
781 if (next_bcast.tv_sec < now.tv_sec+MIN_WAITTIME)
782 next_bcast.tv_sec = now.tv_sec+MIN_WAITTIME;
784 LIST_FOREACH(ifp, &ifnet, int_list) {
785 ifp->int_query_time = NEVER;
788 ifinit_timer.tv_sec = now.tv_sec;
791 && !(ifp->int_state & IS_REMOTE)
792 && ifp->int_rip_sock < 0) {
793 /* RIP is off, so ensure there are sockets on which
794 * to listen for queries.
796 ifp->int_rip_sock = get_rip_sock(ifp->int_addr, 0);
803 /* die if malloc(3) fails
806 rtmalloc(size_t size,
809 void *p = malloc(size);
811 logbad(1,"malloc(%lu) failed in %s", (u_long)size, msg);
816 /* get a random instant in an interval
819 intvl_random(struct timeval *tp, /* put value here */
820 u_long lo, /* value is after this second */
821 u_long hi) /* and before this */
823 tp->tv_sec = (time_t)(hi == lo
825 : (lo + random() % ((hi - lo))));
826 tp->tv_usec = random() % 1000000;
831 timevaladd(struct timeval *t1,
835 t1->tv_sec += t2->tv_sec;
836 if ((t1->tv_usec += t2->tv_usec) >= 1000000) {
838 t1->tv_usec -= 1000000;
846 timevalsub(struct timeval *t1,
850 t1->tv_sec = t2->tv_sec - t3->tv_sec;
851 if ((t1->tv_usec = t2->tv_usec - t3->tv_usec) < 0) {
853 t1->tv_usec += 1000000;
858 /* put a message into the system log
861 msglog(const char *p, ...)
868 vsyslog(LOG_ERR, p, args);
871 if (ftrace == stdout)
872 (void)fputs("routed: ", ftrace);
874 (void)vfprintf(ftrace, p, args);
876 (void)fputc('\n', ftrace);
881 /* Put a message about a bad system into the system log if
882 * we have not complained about it recently.
884 * It is desirable to complain about all bad systems, but not too often.
885 * In the worst case, it is not practical to keep track of all bad systems.
886 * For example, there can be many systems with the wrong password.
889 msglim(struct msg_limit *lim, naddr addr, const char *p, ...)
893 struct msg_sub *ms1, *ms;
896 /* look for the oldest slot in the table
897 * or the slot for the bad router.
899 ms = ms1 = lim->subs;
900 for (i = MSG_SUBJECT_N; ; i--, ms1++) {
902 /* Reuse a slot at most once every 10 minutes.
904 if (lim->reuse > now.tv_sec) {
908 lim->reuse = now.tv_sec + 10*60;
912 if (ms->addr == addr) {
913 /* Repeat a complaint about a given system at
916 if (ms->until > now.tv_sec)
920 if (ms->until < ms1->until)
925 ms->until = now.tv_sec + 60*60; /* 60 minutes */
928 for (p1 = p; *p1 == ' '; p1++)
931 vsyslog(LOG_ERR, p1, args);
935 /* always display the message if tracing */
938 (void)vfprintf(ftrace, p, args);
940 (void)fputc('\n', ftrace);
946 logbad(int dump, const char *p, ...)
953 vsyslog(LOG_ERR, p, args);
955 (void)fputs("routed: ", stderr);
957 (void)vfprintf(stderr, p, args);
959 (void)fputs("; giving up\n",stderr);
960 (void)fflush(stderr);