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\n"
42 "The Regents of the University of California."
43 " All rights reserved.\n");
47 #elif defined(__FreeBSD__)
50 __RCSID("$Revision: 2.27 $");
51 #ident "$Revision: 2.27 $"
56 naddr myaddr; /* system address */
57 char myname[MAXHOSTNAMELEN+1];
61 int supplier; /* supply or broadcast updates */
63 int ipforwarding = 1; /* kernel forwarding on */
65 int default_gateway; /* 1=advertise default */
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 */
73 struct timeval clk, prev_clk;
74 static int usec_fudge;
75 struct timeval now; /* current idea of time */
80 struct timeval next_bcast; /* next general broadcast */
81 struct timeval no_flash = { /* inhibit flash update */
82 EPOCH+SUPPLY_INTERVAL, 0
85 struct timeval flush_kern_timer;
89 int rip_sock = -1; /* RIP socket */
90 struct interface *rip_sock_mcast; /* current multicast interface */
91 int rt_sock; /* routing socket */
95 static int get_rip_sock(naddr, int);
96 static void timevalsub(struct timeval *, struct timeval *, struct timeval *);
106 struct timeval wtime, t2;
110 struct interface *ifp;
115 /* Some shells are badly broken and send SIGHUP to backgrounded
118 signal(SIGHUP, SIG_IGN);
120 openlog("routed", LOG_PID, LOG_DAEMON);
123 gettimeofday(&clk, 0);
126 epoch.tv_sec -= EPOCH;
128 now_stale = EPOCH - STALE_TIME;
129 now_expire = EPOCH - EXPIRE_TIME;
130 now_garbage = EPOCH - GARBAGE_TIME;
133 (void)gethostname(myname, sizeof(myname)-1);
134 (void)gethost(myname, &myaddr);
136 while ((n = getopt(argc, argv, "sqdghmpAtvT:F:P:")) != -1) {
153 memset(&parm, 0, sizeof(parm));
154 parm.parm_d_metric = 1;
155 cp = check_parms(&parm);
157 msglog("bad -g: %s", cp);
162 case 'h': /* suppress extra host routes */
166 case 'm': /* advertise host route */
167 mhome = 1; /* on multi-homed hosts */
171 /* Ignore authentication if we do not care.
172 * Crazy as it is, that is what RFC 1723 requires.
185 case 'F': /* minimal routes for SLIP */
187 p = strchr(optarg,',');
188 if (p && *p != '\0') {
189 n = (int)strtoul(p+1, &q, 0);
191 && n <= HOPCNT_INFINITY-1
195 if (!getnet(optarg, &p_net, &p_mask)) {
196 msglog("bad network; \"-F %s\"",
200 memset(&parm, 0, sizeof(parm));
201 parm.parm_net = p_net;
202 parm.parm_mask = p_mask;
203 parm.parm_d_metric = n;
204 cp = check_parms(&parm);
206 msglog("bad -F: %s", cp);
210 /* handle arbitrary parameters.
213 cp = parse_parms(q, 0);
215 msglog("%s in \"-P %s\"", cp, optarg);
220 /* display version */
222 msglog("version 2.25");
232 if (tracename == 0 && argc >= 1) {
236 if (tracename != 0 && tracename[0] == '\0')
240 logbad(0, "usage: routed [-sqdghmpAtv] [-T tracefile]"
241 " [-F net[,metric]] [-P parms]");
243 if (geteuid() != 0) {
246 logbad(0, "requires UID 0");
252 mib[3] = IPCTL_FORWARDING;
253 len = sizeof(ipforwarding);
254 if (sysctl(mib, 4, &ipforwarding, &len, 0, 0) < 0)
255 LOGERR("sysctl(IPCTL_FORWARDING)");
259 msglog("-s incompatible with ipforwarding=0");
260 if (default_gateway) {
261 msglog("-g incompatible with ipforwarding=0");
267 if (default_gateway) {
268 if (supplier_set && !supplier) {
269 msglog("-g and -q incompatible");
277 signal(SIGALRM, sigalrm);
279 signal(SIGHUP, sigterm); /* SIGHUP fatal during debugging */
280 signal(SIGTERM, sigterm);
281 signal(SIGINT, sigterm);
282 signal(SIGUSR1, sigtrace_on);
283 signal(SIGUSR2, sigtrace_off);
285 /* get into the background */
287 if (0 > _daemonize(background ? 0 : (_DF_NOCHDIR|_DF_NOFORK),
288 STDIN_FILENO, STDOUT_FILENO, STDERR_FILENO))
289 BADERR(0, "_daemonize()");
291 if (background && daemon(0, 1) < 0)
292 BADERR(0,"daemon()");
295 #if defined(__NetBSD__)
302 srandom((int)(clk.tv_sec ^ clk.tv_usec ^ mypid));
305 /* prepare socket connected to the kernel.
307 rt_sock = socket(AF_ROUTE, SOCK_RAW, 0);
309 BADERR(1,"rt_sock = socket()");
310 if (fcntl(rt_sock, F_SETFL, O_NONBLOCK) == -1)
311 logbad(1, "fcntl(rt_sock) O_NONBLOCK: %s", strerror(errno));
313 if (setsockopt(rt_sock, SOL_SOCKET,SO_USELOOPBACK,
314 &off,sizeof(off)) < 0)
315 LOGERR("setsockopt(SO_USELOOPBACK,0)");
320 if (tracename != 0) {
321 strncpy(inittracename, tracename, sizeof(inittracename)-1);
322 set_tracefile(inittracename, "%s", -1);
324 tracelevel_msg("%s", -1); /* turn on tracing to stdio */
329 /* initialize radix tree */
332 /* Pick a random part of the second for our output to minimize
335 * Start broadcasting after hearing from other routers, and
336 * at a random time so a bunch of systems do not get synchronized
337 * after a power failure.
339 intvl_random(&next_bcast, EPOCH+MIN_WAITTIME, EPOCH+SUPPLY_INTERVAL);
340 age_timer.tv_usec = next_bcast.tv_usec;
341 age_timer.tv_sec = EPOCH+MIN_WAITTIME;
342 rdisc_timer = next_bcast;
343 ifinit_timer.tv_usec = next_bcast.tv_usec;
345 /* Collect an initial view of the world by checking the interface
346 * configuration and the kludge file.
355 /* Now turn off stdio if not tracing */
356 if (new_tracelevel == 0)
357 trace_close(background);
359 /* Loop forever, listening and broadcasting.
363 gettimeofday(&clk, 0);
364 if (prev_clk.tv_sec == clk.tv_sec
365 && prev_clk.tv_usec == clk.tv_usec+usec_fudge) {
366 /* Much of `routed` depends on time always advancing.
367 * On systems that do not guarantee that gettimeofday()
368 * produces unique timestamps even if called within
369 * a single tick, use trickery like that in classic
372 clk.tv_usec += ++usec_fudge;
377 timevalsub(&t2, &clk, &prev_clk);
379 || t2.tv_sec > wtime.tv_sec + 5) {
380 /* Deal with time changes before other
381 * housekeeping to keep everything straight.
386 trace_act("time changed by %d sec", (int)dt);
390 timevalsub(&now, &clk, &epoch);
391 now_stale = now.tv_sec - STALE_TIME;
392 now_expire = now.tv_sec - EXPIRE_TIME;
393 now_garbage = now.tv_sec - GARBAGE_TIME;
395 /* deal with signals that should affect tracing */
401 trace_off("exiting with signal %d", stopint);
405 /* look for new or dead interfaces */
406 timevalsub(&wtime, &ifinit_timer, &now);
407 if (wtime.tv_sec <= 0) {
414 /* Check the kernel table occassionally for mysteriously
417 timevalsub(&t2, &flush_kern_timer, &now);
418 if (t2.tv_sec <= 0) {
420 flush_kern_timer.tv_sec = (now.tv_sec
421 + CHECK_QUIET_INTERVAL);
424 if (timercmp(&t2, &wtime, <))
427 /* If it is time, then broadcast our routes.
429 if (supplier || advertise_mhome) {
430 timevalsub(&t2, &next_bcast, &now);
431 if (t2.tv_sec <= 0) {
432 /* Synchronize the aging and broadcast
433 * timers to minimize awakenings
439 /* It is desirable to send routing updates
440 * regularly. So schedule the next update
441 * 30 seconds after the previous one was
442 * scheduled, instead of 30 seconds after
443 * the previous update was finished.
444 * Even if we just started after discovering
445 * a 2nd interface or were otherwise delayed,
446 * pick a 30-second aniversary of the
447 * original broadcast time.
449 n = 1 + (0-t2.tv_sec)/SUPPLY_INTERVAL;
450 next_bcast.tv_sec += n*SUPPLY_INTERVAL;
455 if (timercmp(&t2, &wtime, <))
459 /* If we need a flash update, either do it now or
460 * set the delay to end when it is time.
462 * If we are within MIN_WAITTIME seconds of a full update,
467 && no_flash.tv_sec+MIN_WAITTIME < next_bcast.tv_sec) {
468 /* accurate to the millisecond */
469 if (!timercmp(&no_flash, &now, >))
471 timevalsub(&t2, &no_flash, &now);
472 if (timercmp(&t2, &wtime, <))
476 /* trigger the main aging timer.
478 timevalsub(&t2, &age_timer, &now);
479 if (t2.tv_sec <= 0) {
483 if (timercmp(&t2, &wtime, <))
486 /* update the kernel routing table
488 timevalsub(&t2, &need_kern, &now);
489 if (t2.tv_sec <= 0) {
493 if (timercmp(&t2, &wtime, <))
496 /* take care of router discovery,
497 * but do it in the correct the millisecond
499 if (!timercmp(&rdisc_timer, &now, >)) {
503 timevalsub(&t2, &rdisc_timer, &now);
504 if (timercmp(&t2, &wtime, <))
508 /* wait for input or a timer to expire.
512 n = select(sock_max, &ibits, 0, 0, &wtime);
514 if (n < 0 && errno != EINTR && errno != EAGAIN)
519 if (FD_ISSET(rt_sock, &ibits)) {
523 if (rdisc_sock >= 0 && FD_ISSET(rdisc_sock, &ibits)) {
527 if (rip_sock >= 0 && FD_ISSET(rip_sock, &ibits)) {
528 read_rip(rip_sock, 0);
532 for (ifp = ifnet; n > 0 && 0 != ifp; ifp = ifp->int_next) {
533 if (ifp->int_rip_sock >= 0
534 && FD_ISSET(ifp->int_rip_sock, &ibits)) {
535 read_rip(ifp->int_rip_sock, ifp);
545 sigalrm(int s UNUSED)
547 /* Historically, SIGALRM would cause the daemon to check for
548 * new and broken interfaces.
550 ifinit_timer.tv_sec = now.tv_sec;
551 trace_act("SIGALRM");
555 /* watch for fatal signals */
560 (void)signal(sig, SIG_DFL); /* catch it only once */
567 struct interface *ifp;
573 FD_SET(rt_sock, &fdbits);
574 if (sock_max <= rt_sock)
575 sock_max = rt_sock+1;
577 FD_SET(rip_sock, &fdbits);
578 if (sock_max <= rip_sock)
579 sock_max = rip_sock+1;
581 for (ifp = ifnet; 0 != ifp; ifp = ifp->int_next) {
582 if (ifp->int_rip_sock >= 0) {
583 FD_SET(ifp->int_rip_sock, &fdbits);
584 if (sock_max <= ifp->int_rip_sock)
585 sock_max = ifp->int_rip_sock+1;
588 if (rdisc_sock >= 0) {
589 FD_SET(rdisc_sock, &fdbits);
590 if (sock_max <= rdisc_sock)
591 sock_max = rdisc_sock+1;
601 #define MIN_SOCKBUF (4*1024)
604 if (fcntl(sock, F_SETFL, O_NONBLOCK) == -1)
605 logbad(1, "fcntl(%s) O_NONBLOCK: %s",
606 name, strerror(errno));
608 if (setsockopt(sock, SOL_SOCKET,SO_BROADCAST, &on,sizeof(on)) < 0)
609 msglog("setsockopt(%s,SO_BROADCAST): %s",
610 name, strerror(errno));
611 #ifdef USE_PASSIFNAME
613 if (setsockopt(sock, SOL_SOCKET, SO_PASSIFNAME, &on,sizeof(on)) < 0)
614 msglog("setsockopt(%s,SO_PASSIFNAME): %s",
615 name, strerror(errno));
618 if (rbuf >= MIN_SOCKBUF) {
619 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
620 &rbuf, sizeof(rbuf)) < 0)
621 msglog("setsockopt(%s,SO_RCVBUF=%d): %s",
622 name, rbuf, strerror(errno));
624 for (rbuf = 60*1024; ; rbuf -= 4096) {
625 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
626 &rbuf, sizeof(rbuf)) == 0) {
627 trace_act("RCVBUF=%d", rbuf);
630 if (rbuf < MIN_SOCKBUF) {
631 msglog("setsockopt(%s,SO_RCVBUF = %d): %s",
632 name, rbuf, strerror(errno));
642 static int /* <0 or file descriptor */
643 get_rip_sock(naddr addr,
644 int serious) /* 1=failure to bind is serious */
646 struct sockaddr_in rsin;
651 if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
652 BADERR(1,"rip_sock = socket()");
654 memset(&rsin, 0, sizeof(rsin));
656 rsin.sin_len = sizeof(rsin);
658 rsin.sin_family = AF_INET;
659 rsin.sin_port = htons(RIP_PORT);
660 rsin.sin_addr.s_addr = addr;
661 if (bind(s, (struct sockaddr *)&rsin, sizeof(rsin)) < 0) {
663 BADERR(errno != EADDRINUSE, "bind(rip_sock)");
666 fix_sock(s,"rip_sock");
669 if (setsockopt(s, IPPROTO_IP, IP_MULTICAST_TTL,
670 &ttl, sizeof(ttl)) < 0)
671 DBGERR(1,"rip_sock setsockopt(IP_MULTICAST_TTL)");
677 /* turn off main RIP socket */
681 struct interface *ifp;
685 if (rip_sock >= 0 && !mhome) {
686 trace_act("turn off RIP");
688 (void)close(rip_sock);
691 /* get non-broadcast sockets to listen to queries.
693 for (ifp = ifnet; ifp != 0; ifp = ifp->int_next) {
694 if (ifp->int_state & IS_REMOTE)
696 if (ifp->int_rip_sock < 0) {
697 addr = ((ifp->int_if_flags & IFF_POINTOPOINT)
700 ifp->int_rip_sock = get_rip_sock(addr, 0);
711 /* turn on RIP multicast input via an interface
714 rip_mcast_on(struct interface *ifp)
718 if (!IS_RIP_IN_OFF(ifp->int_state)
719 && (ifp->int_if_flags & IFF_MULTICAST)
721 && !(ifp->int_if_flags & IFF_POINTOPOINT)
723 && !(ifp->int_state & IS_ALIAS)) {
724 m.imr_multiaddr.s_addr = htonl(INADDR_RIP_GROUP);
726 m.imr_interface.s_addr = htonl(ifp->int_index);
728 m.imr_interface.s_addr = ((ifp->int_if_flags & IFF_POINTOPOINT)
732 if (setsockopt(rip_sock,IPPROTO_IP, IP_ADD_MEMBERSHIP,
734 LOGERR("setsockopt(IP_ADD_MEMBERSHIP RIP)");
739 /* Prepare socket used for RIP.
742 rip_on(struct interface *ifp)
744 /* If the main RIP socket is already alive, only start receiving
745 * multicasts for this interface.
753 /* If the main RIP socket is off and it makes sense to turn it on,
754 * then turn it on for all of the interfaces.
755 * It makes sense if either router discovery is off, or if
756 * router discover is on and at most one interface is doing RIP.
758 if (rip_interfaces > 0 && (!rdisc_ok || rip_interfaces > 1)) {
759 trace_act("turn on RIP");
761 /* Close all of the query sockets so that we can open
762 * the main socket. SO_REUSEPORT is not a solution,
763 * since that would let two daemons bind to the broadcast
766 for (ifp = ifnet; ifp != 0; ifp = ifp->int_next) {
767 if (ifp->int_rip_sock >= 0) {
768 (void)close(ifp->int_rip_sock);
769 ifp->int_rip_sock = -1;
773 rip_sock = get_rip_sock(INADDR_ANY, 1);
776 /* Do not advertise anything until we have heard something
778 if (next_bcast.tv_sec < now.tv_sec+MIN_WAITTIME)
779 next_bcast.tv_sec = now.tv_sec+MIN_WAITTIME;
781 for (ifp = ifnet; ifp != 0; ifp = ifp->int_next) {
782 ifp->int_query_time = NEVER;
785 ifinit_timer.tv_sec = now.tv_sec;
788 && !(ifp->int_state & IS_REMOTE)
789 && ifp->int_rip_sock < 0) {
790 /* RIP is off, so ensure there are sockets on which
791 * to listen for queries.
793 ifp->int_rip_sock = get_rip_sock(ifp->int_addr, 0);
800 /* die if malloc(3) fails
803 rtmalloc(size_t size,
806 void *p = malloc(size);
808 logbad(1,"malloc(%lu) failed in %s", (u_long)size, msg);
813 /* get a random instant in an interval
816 intvl_random(struct timeval *tp, /* put value here */
817 u_long lo, /* value is after this second */
818 u_long hi) /* and before this */
820 tp->tv_sec = (time_t)(hi == lo
822 : (lo + random() % ((hi - lo))));
823 tp->tv_usec = random() % 1000000;
828 timevaladd(struct timeval *t1,
832 t1->tv_sec += t2->tv_sec;
833 if ((t1->tv_usec += t2->tv_usec) >= 1000000) {
835 t1->tv_usec -= 1000000;
843 timevalsub(struct timeval *t1,
847 t1->tv_sec = t2->tv_sec - t3->tv_sec;
848 if ((t1->tv_usec = t2->tv_usec - t3->tv_usec) < 0) {
850 t1->tv_usec += 1000000;
855 /* put a message into the system log
858 msglog(const char *p, ...)
865 vsyslog(LOG_ERR, p, args);
868 if (ftrace == stdout)
869 (void)fputs("routed: ", ftrace);
870 (void)vfprintf(ftrace, p, args);
871 (void)fputc('\n', ftrace);
877 /* Put a message about a bad system into the system log if
878 * we have not complained about it recently.
880 * It is desirable to complain about all bad systems, but not too often.
881 * In the worst case, it is not practical to keep track of all bad systems.
882 * For example, there can be many systems with the wrong password.
885 msglim(struct msg_limit *lim, naddr addr, const char *p, ...)
889 struct msg_sub *ms1, *ms;
894 /* look for the oldest slot in the table
895 * or the slot for the bad router.
897 ms = ms1 = lim->subs;
898 for (i = MSG_SUBJECT_N; ; i--, ms1++) {
900 /* Reuse a slot at most once every 10 minutes.
902 if (lim->reuse > now.tv_sec) {
906 lim->reuse = now.tv_sec + 10*60;
910 if (ms->addr == addr) {
911 /* Repeat a complaint about a given system at
914 if (ms->until > now.tv_sec)
918 if (ms->until < ms1->until)
923 ms->until = now.tv_sec + 60*60; /* 60 minutes */
926 for (p1 = p; *p1 == ' '; p1++)
928 vsyslog(LOG_ERR, p1, args);
931 /* always display the message if tracing */
933 (void)vfprintf(ftrace, p, args);
934 (void)fputc('\n', ftrace);
941 logbad(int dump, const char *p, ...)
948 vsyslog(LOG_ERR, p, args);
950 (void)fputs("routed: ", stderr);
951 (void)vfprintf(stderr, p, args);
952 (void)fputs("; giving up\n",stderr);
953 (void)fflush(stderr);