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 */
71 int insecure; /* Reply to special queries or not */
73 struct timeval epoch; /* when started */
75 static struct timeval prev_clk;
76 static int usec_fudge;
77 struct timeval now; /* current idea of time */
82 static struct timeval next_bcast; /* next general broadcast */
83 struct timeval no_flash = { /* inhibit flash update */
84 EPOCH+SUPPLY_INTERVAL, 0
87 static struct timeval flush_kern_timer;
91 int rip_sock = -1; /* RIP socket */
92 const struct interface *rip_sock_mcast; /* current multicast interface */
93 int rt_sock; /* routing socket */
97 static int get_rip_sock(naddr, int);
98 static void timevalsub(struct timeval *, struct timeval *, struct timeval *);
99 static void sigalrm(int s UNUSED);
100 static void sigterm(int sig);
110 struct timeval wtime, t2;
114 struct interface *ifp;
119 /* Some shells are badly broken and send SIGHUP to backgrounded
122 signal(SIGHUP, SIG_IGN);
124 openlog("routed", LOG_PID, LOG_DAEMON);
127 gettimeofday(&clk, 0);
130 epoch.tv_sec -= EPOCH;
132 now_stale = EPOCH - STALE_TIME;
133 now_expire = EPOCH - EXPIRE_TIME;
134 now_garbage = EPOCH - GARBAGE_TIME;
137 (void)gethostname(myname, sizeof(myname)-1);
138 (void)gethost(myname, &myaddr);
140 while ((n = getopt(argc, argv, "isqdghmAtvT:F:P:")) != -1) {
160 memset(&parm, 0, sizeof(parm));
161 parm.parm_d_metric = 1;
162 cp = check_parms(&parm);
164 msglog("bad -g: %s", cp);
169 case 'h': /* suppress extra host routes */
173 case 'm': /* advertise host route */
174 mhome = 1; /* on multi-homed hosts */
178 /* Ignore authentication if we do not care.
179 * Crazy as it is, that is what RFC 1723 requires.
192 case 'F': /* minimal routes for SLIP */
194 p = strchr(optarg,',');
195 if (p && *p != '\0') {
196 n = (int)strtoul(p+1, &q, 0);
198 && n <= HOPCNT_INFINITY-1
202 if (!getnet(optarg, &p_net, &p_mask)) {
203 msglog("bad network; \"-F %s\"",
207 memset(&parm, 0, sizeof(parm));
208 parm.parm_net = p_net;
209 parm.parm_mask = p_mask;
210 parm.parm_d_metric = n;
211 cp = check_parms(&parm);
213 msglog("bad -F: %s", cp);
217 /* handle arbitrary parameters.
220 cp = parse_parms(q, 0);
222 msglog("%s in \"-P %s\"", cp, optarg);
227 /* display version */
229 msglog("version 2.31");
239 if (tracename == 0 && argc >= 1) {
243 if (tracename != 0 && tracename[0] == '\0')
247 logbad(0, "usage: routed [-sqdghmAtv] [-T tracefile]"
248 " [-F net[,metric]] [-P parms]");
250 if (geteuid() != 0) {
253 logbad(0, "requires UID 0");
259 mib[3] = IPCTL_FORWARDING;
260 len = sizeof(ipforwarding);
261 if (sysctl(mib, 4, &ipforwarding, &len, 0, 0) < 0)
262 LOGERR("sysctl(IPCTL_FORWARDING)");
266 msglog("-s incompatible with ipforwarding=0");
267 if (default_gateway) {
268 msglog("-g incompatible with ipforwarding=0");
274 if (default_gateway) {
275 if (supplier_set && !supplier) {
276 msglog("-g and -q incompatible");
284 signal(SIGALRM, sigalrm);
286 signal(SIGHUP, sigterm); /* SIGHUP fatal during debugging */
287 signal(SIGTERM, sigterm);
288 signal(SIGINT, sigterm);
289 signal(SIGUSR1, sigtrace_on);
290 signal(SIGUSR2, sigtrace_off);
292 /* get into the background */
294 if (0 > _daemonize(background ? 0 : (_DF_NOCHDIR|_DF_NOFORK),
295 STDIN_FILENO, STDOUT_FILENO, STDERR_FILENO))
296 BADERR(0, "_daemonize()");
298 if (background && daemon(0, 1) < 0)
299 BADERR(0,"daemon()");
302 #if defined(__NetBSD__)
309 srandom((int)(clk.tv_sec ^ clk.tv_usec ^ mypid));
312 /* prepare socket connected to the kernel.
314 rt_sock = socket(AF_ROUTE, SOCK_RAW, 0);
316 BADERR(1,"rt_sock = socket()");
317 if (fcntl(rt_sock, F_SETFL, O_NONBLOCK) == -1)
318 logbad(1, "fcntl(rt_sock) O_NONBLOCK: %s", strerror(errno));
320 if (setsockopt(rt_sock, SOL_SOCKET,SO_USELOOPBACK,
321 &off,sizeof(off)) < 0)
322 LOGERR("setsockopt(SO_USELOOPBACK,0)");
327 if (tracename != 0) {
328 strncpy(inittracename, tracename, sizeof(inittracename)-1);
329 set_tracefile(inittracename, "%s", -1);
331 tracelevel_msg("%s", -1); /* turn on tracing to stdio */
336 /* initialize radix tree */
339 /* Pick a random part of the second for our output to minimize
342 * Start broadcasting after hearing from other routers, and
343 * at a random time so a bunch of systems do not get synchronized
344 * after a power failure.
346 intvl_random(&next_bcast, EPOCH+MIN_WAITTIME, EPOCH+SUPPLY_INTERVAL);
347 age_timer.tv_usec = next_bcast.tv_usec;
348 age_timer.tv_sec = EPOCH+MIN_WAITTIME;
349 rdisc_timer = next_bcast;
350 ifinit_timer.tv_usec = next_bcast.tv_usec;
352 /* Collect an initial view of the world by checking the interface
353 * configuration and the kludge file.
362 /* Now turn off stdio if not tracing */
363 if (new_tracelevel == 0)
364 trace_close(background);
366 /* Loop forever, listening and broadcasting.
370 gettimeofday(&clk, 0);
371 if (prev_clk.tv_sec == clk.tv_sec
372 && prev_clk.tv_usec == clk.tv_usec+usec_fudge) {
373 /* Much of `routed` depends on time always advancing.
374 * On systems that do not guarantee that gettimeofday()
375 * produces unique timestamps even if called within
376 * a single tick, use trickery like that in classic
379 clk.tv_usec += ++usec_fudge;
384 timevalsub(&t2, &clk, &prev_clk);
386 || t2.tv_sec > wtime.tv_sec + 5) {
387 /* Deal with time changes before other
388 * housekeeping to keep everything straight.
393 trace_act("time changed by %d sec", (int)dt);
397 timevalsub(&now, &clk, &epoch);
398 now_stale = now.tv_sec - STALE_TIME;
399 now_expire = now.tv_sec - EXPIRE_TIME;
400 now_garbage = now.tv_sec - GARBAGE_TIME;
402 /* deal with signals that should affect tracing */
408 trace_off("exiting with signal %d", stopint);
412 /* look for new or dead interfaces */
413 timevalsub(&wtime, &ifinit_timer, &now);
414 if (wtime.tv_sec <= 0) {
421 /* Check the kernel table occasionally for mysteriously
424 timevalsub(&t2, &flush_kern_timer, &now);
425 if (t2.tv_sec <= 0) {
427 flush_kern_timer.tv_sec = (now.tv_sec
428 + CHECK_QUIET_INTERVAL);
431 if (timercmp(&t2, &wtime, <))
434 /* If it is time, then broadcast our routes.
436 if (supplier || advertise_mhome) {
437 timevalsub(&t2, &next_bcast, &now);
438 if (t2.tv_sec <= 0) {
439 /* Synchronize the aging and broadcast
440 * timers to minimize awakenings
446 /* It is desirable to send routing updates
447 * regularly. So schedule the next update
448 * 30 seconds after the previous one was
449 * scheduled, instead of 30 seconds after
450 * the previous update was finished.
451 * Even if we just started after discovering
452 * a 2nd interface or were otherwise delayed,
453 * pick a 30-second anniversary of the
454 * original broadcast time.
456 n = 1 + (0-t2.tv_sec)/SUPPLY_INTERVAL;
457 next_bcast.tv_sec += n*SUPPLY_INTERVAL;
462 if (timercmp(&t2, &wtime, <))
466 /* If we need a flash update, either do it now or
467 * set the delay to end when it is time.
469 * If we are within MIN_WAITTIME seconds of a full update,
474 && no_flash.tv_sec+MIN_WAITTIME < next_bcast.tv_sec) {
475 /* accurate to the millisecond */
476 if (!timercmp(&no_flash, &now, >))
478 timevalsub(&t2, &no_flash, &now);
479 if (timercmp(&t2, &wtime, <))
483 /* trigger the main aging timer.
485 timevalsub(&t2, &age_timer, &now);
486 if (t2.tv_sec <= 0) {
490 if (timercmp(&t2, &wtime, <))
493 /* update the kernel routing table
495 timevalsub(&t2, &need_kern, &now);
496 if (t2.tv_sec <= 0) {
500 if (timercmp(&t2, &wtime, <))
503 /* take care of router discovery,
504 * but do it in the correct the millisecond
506 if (!timercmp(&rdisc_timer, &now, >)) {
510 timevalsub(&t2, &rdisc_timer, &now);
511 if (timercmp(&t2, &wtime, <))
515 /* wait for input or a timer to expire.
519 n = select(sock_max, &ibits, 0, 0, &wtime);
521 if (n < 0 && errno != EINTR && errno != EAGAIN)
526 if (FD_ISSET(rt_sock, &ibits)) {
530 if (rdisc_sock >= 0 && FD_ISSET(rdisc_sock, &ibits)) {
534 if (rip_sock >= 0 && FD_ISSET(rip_sock, &ibits)) {
535 read_rip(rip_sock, 0);
539 LIST_FOREACH(ifp, &ifnet, int_list) {
542 if (ifp->int_rip_sock >= 0
543 && FD_ISSET(ifp->int_rip_sock, &ibits)) {
544 read_rip(ifp->int_rip_sock, ifp);
554 sigalrm(int s UNUSED)
556 /* Historically, SIGALRM would cause the daemon to check for
557 * new and broken interfaces.
559 ifinit_timer.tv_sec = now.tv_sec;
560 trace_act("SIGALRM");
564 /* watch for fatal signals */
569 (void)signal(sig, SIG_DFL); /* catch it only once */
576 struct interface *ifp;
582 FD_SET(rt_sock, &fdbits);
583 if (sock_max <= rt_sock)
584 sock_max = rt_sock+1;
586 FD_SET(rip_sock, &fdbits);
587 if (sock_max <= rip_sock)
588 sock_max = rip_sock+1;
590 LIST_FOREACH(ifp, &ifnet, int_list) {
591 if (ifp->int_rip_sock >= 0) {
592 FD_SET(ifp->int_rip_sock, &fdbits);
593 if (sock_max <= ifp->int_rip_sock)
594 sock_max = ifp->int_rip_sock+1;
597 if (rdisc_sock >= 0) {
598 FD_SET(rdisc_sock, &fdbits);
599 if (sock_max <= rdisc_sock)
600 sock_max = rdisc_sock+1;
610 #define MIN_SOCKBUF (4*1024)
613 if (fcntl(sock, F_SETFL, O_NONBLOCK) == -1)
614 logbad(1, "fcntl(%s) O_NONBLOCK: %s",
615 name, strerror(errno));
617 if (setsockopt(sock, SOL_SOCKET,SO_BROADCAST, &on,sizeof(on)) < 0)
618 msglog("setsockopt(%s,SO_BROADCAST): %s",
619 name, strerror(errno));
620 #ifdef USE_PASSIFNAME
622 if (setsockopt(sock, SOL_SOCKET, SO_PASSIFNAME, &on,sizeof(on)) < 0)
623 msglog("setsockopt(%s,SO_PASSIFNAME): %s",
624 name, strerror(errno));
627 if (rbuf >= MIN_SOCKBUF) {
628 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
629 &rbuf, sizeof(rbuf)) < 0)
630 msglog("setsockopt(%s,SO_RCVBUF=%d): %s",
631 name, rbuf, strerror(errno));
633 for (rbuf = 60*1024; ; rbuf -= 4096) {
634 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
635 &rbuf, sizeof(rbuf)) == 0) {
636 trace_act("RCVBUF=%d", rbuf);
639 if (rbuf < MIN_SOCKBUF) {
640 msglog("setsockopt(%s,SO_RCVBUF = %d): %s",
641 name, rbuf, strerror(errno));
651 static int /* <0 or file descriptor */
652 get_rip_sock(naddr addr,
653 int serious) /* 1=failure to bind is serious */
655 struct sockaddr_in rsin;
660 if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
661 BADERR(1,"rip_sock = socket()");
663 memset(&rsin, 0, sizeof(rsin));
665 rsin.sin_len = sizeof(rsin);
667 rsin.sin_family = AF_INET;
668 rsin.sin_port = htons(RIP_PORT);
669 rsin.sin_addr.s_addr = addr;
670 if (bind(s, (struct sockaddr *)&rsin, sizeof(rsin)) < 0) {
672 BADERR(errno != EADDRINUSE, "bind(rip_sock)");
675 fix_sock(s,"rip_sock");
678 if (setsockopt(s, IPPROTO_IP, IP_MULTICAST_TTL,
679 &ttl, sizeof(ttl)) < 0)
680 DBGERR(1,"rip_sock setsockopt(IP_MULTICAST_TTL)");
686 /* turn off main RIP socket */
690 struct interface *ifp;
694 if (rip_sock >= 0 && !mhome) {
695 trace_act("turn off RIP");
697 (void)close(rip_sock);
700 /* get non-broadcast sockets to listen to queries.
702 LIST_FOREACH(ifp, &ifnet, int_list) {
703 if (ifp->int_state & IS_REMOTE)
705 if (ifp->int_rip_sock < 0) {
706 addr = ((ifp->int_if_flags & IFF_POINTOPOINT)
709 ifp->int_rip_sock = get_rip_sock(addr, 0);
720 /* turn on RIP multicast input via an interface
723 rip_mcast_on(struct interface *ifp)
726 struct sockaddr_in *sin;
728 if (!IS_RIP_IN_OFF(ifp->int_state)
729 && (ifp->int_if_flags & IFF_MULTICAST)
730 && !(ifp->int_state & IS_ALIAS)) {
731 memset(&gr, 0, sizeof(gr));
732 gr.gr_interface = ifp->int_index;
733 sin = (struct sockaddr_in *)&gr.gr_group;
734 sin->sin_family = AF_INET;
736 sin->sin_len = sizeof(struct sockaddr_in);
738 sin->sin_addr.s_addr = htonl(INADDR_RIP_GROUP);
739 if (setsockopt(rip_sock, IPPROTO_IP, MCAST_JOIN_GROUP,
740 &gr, sizeof(gr)) < 0)
741 LOGERR("setsockopt(MCAST_JOIN_GROUP RIP)");
746 /* Prepare socket used for RIP.
749 rip_on(struct interface *ifp)
751 /* If the main RIP socket is already alive, only start receiving
752 * multicasts for this interface.
760 /* If the main RIP socket is off and it makes sense to turn it on,
761 * then turn it on for all of the interfaces.
762 * It makes sense if either router discovery is off, or if
763 * router discover is on and at most one interface is doing RIP.
765 if (rip_interfaces > 0 && (!rdisc_ok || rip_interfaces > 1)) {
766 trace_act("turn on RIP");
768 /* Close all of the query sockets so that we can open
769 * the main socket. SO_REUSEPORT is not a solution,
770 * since that would let two daemons bind to the broadcast
773 LIST_FOREACH(ifp, &ifnet, int_list) {
774 if (ifp->int_rip_sock >= 0) {
775 (void)close(ifp->int_rip_sock);
776 ifp->int_rip_sock = -1;
780 rip_sock = get_rip_sock(INADDR_ANY, 1);
783 /* Do not advertise anything until we have heard something
785 if (next_bcast.tv_sec < now.tv_sec+MIN_WAITTIME)
786 next_bcast.tv_sec = now.tv_sec+MIN_WAITTIME;
788 LIST_FOREACH(ifp, &ifnet, int_list) {
789 ifp->int_query_time = NEVER;
792 ifinit_timer.tv_sec = now.tv_sec;
795 && !(ifp->int_state & IS_REMOTE)
796 && ifp->int_rip_sock < 0) {
797 /* RIP is off, so ensure there are sockets on which
798 * to listen for queries.
800 ifp->int_rip_sock = get_rip_sock(ifp->int_addr, 0);
807 /* die if malloc(3) fails
810 rtmalloc(size_t size,
813 void *p = malloc(size);
815 logbad(1,"malloc(%lu) failed in %s", (u_long)size, msg);
820 /* get a random instant in an interval
823 intvl_random(struct timeval *tp, /* put value here */
824 u_long lo, /* value is after this second */
825 u_long hi) /* and before this */
827 tp->tv_sec = (time_t)(hi == lo
829 : (lo + random() % ((hi - lo))));
830 tp->tv_usec = random() % 1000000;
835 timevaladd(struct timeval *t1,
839 t1->tv_sec += t2->tv_sec;
840 if ((t1->tv_usec += t2->tv_usec) >= 1000000) {
842 t1->tv_usec -= 1000000;
850 timevalsub(struct timeval *t1,
854 t1->tv_sec = t2->tv_sec - t3->tv_sec;
855 if ((t1->tv_usec = t2->tv_usec - t3->tv_usec) < 0) {
857 t1->tv_usec += 1000000;
862 /* put a message into the system log
865 msglog(const char *p, ...)
872 vsyslog(LOG_ERR, p, args);
875 if (ftrace == stdout)
876 (void)fputs("routed: ", ftrace);
878 (void)vfprintf(ftrace, p, args);
880 (void)fputc('\n', ftrace);
885 /* Put a message about a bad system into the system log if
886 * we have not complained about it recently.
888 * It is desirable to complain about all bad systems, but not too often.
889 * In the worst case, it is not practical to keep track of all bad systems.
890 * For example, there can be many systems with the wrong password.
893 msglim(struct msg_limit *lim, naddr addr, const char *p, ...)
897 struct msg_sub *ms1, *ms;
900 /* look for the oldest slot in the table
901 * or the slot for the bad router.
903 ms = ms1 = lim->subs;
904 for (i = MSG_SUBJECT_N; ; i--, ms1++) {
906 /* Reuse a slot at most once every 10 minutes.
908 if (lim->reuse > now.tv_sec) {
912 lim->reuse = now.tv_sec + 10*60;
916 if (ms->addr == addr) {
917 /* Repeat a complaint about a given system at
920 if (ms->until > now.tv_sec)
924 if (ms->until < ms1->until)
929 ms->until = now.tv_sec + 60*60; /* 60 minutes */
932 for (p1 = p; *p1 == ' '; p1++)
935 vsyslog(LOG_ERR, p1, args);
939 /* always display the message if tracing */
942 (void)vfprintf(ftrace, p, args);
944 (void)fputc('\n', ftrace);
950 logbad(int dump, const char *p, ...)
957 vsyslog(LOG_ERR, p, args);
959 (void)fputs("routed: ", stderr);
961 (void)vfprintf(stderr, p, args);
963 (void)fputs("; giving up\n",stderr);
964 (void)fflush(stderr);