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__)
307 /* prepare socket connected to the kernel.
309 rt_sock = socket(AF_ROUTE, SOCK_RAW, 0);
311 BADERR(1,"rt_sock = socket()");
312 if (fcntl(rt_sock, F_SETFL, O_NONBLOCK) == -1)
313 logbad(1, "fcntl(rt_sock) O_NONBLOCK: %s", strerror(errno));
315 if (setsockopt(rt_sock, SOL_SOCKET,SO_USELOOPBACK,
316 &off,sizeof(off)) < 0)
317 LOGERR("setsockopt(SO_USELOOPBACK,0)");
322 if (tracename != 0) {
323 strncpy(inittracename, tracename, sizeof(inittracename)-1);
324 set_tracefile(inittracename, "%s", -1);
326 tracelevel_msg("%s", -1); /* turn on tracing to stdio */
331 /* initialize radix tree */
334 /* Pick a random part of the second for our output to minimize
337 * Start broadcasting after hearing from other routers, and
338 * at a random time so a bunch of systems do not get synchronized
339 * after a power failure.
341 intvl_random(&next_bcast, EPOCH+MIN_WAITTIME, EPOCH+SUPPLY_INTERVAL);
342 age_timer.tv_usec = next_bcast.tv_usec;
343 age_timer.tv_sec = EPOCH+MIN_WAITTIME;
344 rdisc_timer = next_bcast;
345 ifinit_timer.tv_usec = next_bcast.tv_usec;
347 /* Collect an initial view of the world by checking the interface
348 * configuration and the kludge file.
357 /* Now turn off stdio if not tracing */
358 if (new_tracelevel == 0)
359 trace_close(background);
361 /* Loop forever, listening and broadcasting.
365 gettimeofday(&clk, 0);
366 if (prev_clk.tv_sec == clk.tv_sec
367 && prev_clk.tv_usec == clk.tv_usec+usec_fudge) {
368 /* Much of `routed` depends on time always advancing.
369 * On systems that do not guarantee that gettimeofday()
370 * produces unique timestamps even if called within
371 * a single tick, use trickery like that in classic
374 clk.tv_usec += ++usec_fudge;
379 timevalsub(&t2, &clk, &prev_clk);
381 || t2.tv_sec > wtime.tv_sec + 5) {
382 /* Deal with time changes before other
383 * housekeeping to keep everything straight.
388 trace_act("time changed by %d sec", (int)dt);
392 timevalsub(&now, &clk, &epoch);
393 now_stale = now.tv_sec - STALE_TIME;
394 now_expire = now.tv_sec - EXPIRE_TIME;
395 now_garbage = now.tv_sec - GARBAGE_TIME;
397 /* deal with signals that should affect tracing */
403 trace_off("exiting with signal %d", stopint);
407 /* look for new or dead interfaces */
408 timevalsub(&wtime, &ifinit_timer, &now);
409 if (wtime.tv_sec <= 0) {
416 /* Check the kernel table occasionally for mysteriously
419 timevalsub(&t2, &flush_kern_timer, &now);
420 if (t2.tv_sec <= 0) {
422 flush_kern_timer.tv_sec = (now.tv_sec
423 + CHECK_QUIET_INTERVAL);
426 if (timercmp(&t2, &wtime, <))
429 /* If it is time, then broadcast our routes.
431 if (supplier || advertise_mhome) {
432 timevalsub(&t2, &next_bcast, &now);
433 if (t2.tv_sec <= 0) {
434 /* Synchronize the aging and broadcast
435 * timers to minimize awakenings
441 /* It is desirable to send routing updates
442 * regularly. So schedule the next update
443 * 30 seconds after the previous one was
444 * scheduled, instead of 30 seconds after
445 * the previous update was finished.
446 * Even if we just started after discovering
447 * a 2nd interface or were otherwise delayed,
448 * pick a 30-second anniversary of the
449 * original broadcast time.
451 n = 1 + (0-t2.tv_sec)/SUPPLY_INTERVAL;
452 next_bcast.tv_sec += n*SUPPLY_INTERVAL;
457 if (timercmp(&t2, &wtime, <))
461 /* If we need a flash update, either do it now or
462 * set the delay to end when it is time.
464 * If we are within MIN_WAITTIME seconds of a full update,
469 && no_flash.tv_sec+MIN_WAITTIME < next_bcast.tv_sec) {
470 /* accurate to the millisecond */
471 if (!timercmp(&no_flash, &now, >))
473 timevalsub(&t2, &no_flash, &now);
474 if (timercmp(&t2, &wtime, <))
478 /* trigger the main aging timer.
480 timevalsub(&t2, &age_timer, &now);
481 if (t2.tv_sec <= 0) {
485 if (timercmp(&t2, &wtime, <))
488 /* update the kernel routing table
490 timevalsub(&t2, &need_kern, &now);
491 if (t2.tv_sec <= 0) {
495 if (timercmp(&t2, &wtime, <))
498 /* take care of router discovery,
499 * but do it in the correct the millisecond
501 if (!timercmp(&rdisc_timer, &now, >)) {
505 timevalsub(&t2, &rdisc_timer, &now);
506 if (timercmp(&t2, &wtime, <))
510 /* wait for input or a timer to expire.
514 n = select(sock_max, &ibits, 0, 0, &wtime);
516 if (n < 0 && errno != EINTR && errno != EAGAIN)
521 if (FD_ISSET(rt_sock, &ibits)) {
525 if (rdisc_sock >= 0 && FD_ISSET(rdisc_sock, &ibits)) {
529 if (rip_sock >= 0 && FD_ISSET(rip_sock, &ibits)) {
530 read_rip(rip_sock, 0);
534 LIST_FOREACH(ifp, &ifnet, int_list) {
537 if (ifp->int_rip_sock >= 0
538 && FD_ISSET(ifp->int_rip_sock, &ibits)) {
539 read_rip(ifp->int_rip_sock, ifp);
549 sigalrm(int s UNUSED)
551 /* Historically, SIGALRM would cause the daemon to check for
552 * new and broken interfaces.
554 ifinit_timer.tv_sec = now.tv_sec;
555 trace_act("SIGALRM");
559 /* watch for fatal signals */
564 (void)signal(sig, SIG_DFL); /* catch it only once */
571 struct interface *ifp;
577 FD_SET(rt_sock, &fdbits);
578 if (sock_max <= rt_sock)
579 sock_max = rt_sock+1;
581 FD_SET(rip_sock, &fdbits);
582 if (sock_max <= rip_sock)
583 sock_max = rip_sock+1;
585 LIST_FOREACH(ifp, &ifnet, int_list) {
586 if (ifp->int_rip_sock >= 0) {
587 FD_SET(ifp->int_rip_sock, &fdbits);
588 if (sock_max <= ifp->int_rip_sock)
589 sock_max = ifp->int_rip_sock+1;
592 if (rdisc_sock >= 0) {
593 FD_SET(rdisc_sock, &fdbits);
594 if (sock_max <= rdisc_sock)
595 sock_max = rdisc_sock+1;
605 #define MIN_SOCKBUF (4*1024)
608 if (fcntl(sock, F_SETFL, O_NONBLOCK) == -1)
609 logbad(1, "fcntl(%s) O_NONBLOCK: %s",
610 name, strerror(errno));
612 if (setsockopt(sock, SOL_SOCKET,SO_BROADCAST, &on,sizeof(on)) < 0)
613 msglog("setsockopt(%s,SO_BROADCAST): %s",
614 name, strerror(errno));
615 #ifdef USE_PASSIFNAME
617 if (setsockopt(sock, SOL_SOCKET, SO_PASSIFNAME, &on,sizeof(on)) < 0)
618 msglog("setsockopt(%s,SO_PASSIFNAME): %s",
619 name, strerror(errno));
622 if (rbuf >= MIN_SOCKBUF) {
623 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
624 &rbuf, sizeof(rbuf)) < 0)
625 msglog("setsockopt(%s,SO_RCVBUF=%d): %s",
626 name, rbuf, strerror(errno));
628 for (rbuf = 60*1024; ; rbuf -= 4096) {
629 if (setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
630 &rbuf, sizeof(rbuf)) == 0) {
631 trace_act("RCVBUF=%d", rbuf);
634 if (rbuf < MIN_SOCKBUF) {
635 msglog("setsockopt(%s,SO_RCVBUF = %d): %s",
636 name, rbuf, strerror(errno));
646 static int /* <0 or file descriptor */
647 get_rip_sock(naddr addr,
648 int serious) /* 1=failure to bind is serious */
650 struct sockaddr_in rsin;
655 if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
656 BADERR(1,"rip_sock = socket()");
658 memset(&rsin, 0, sizeof(rsin));
660 rsin.sin_len = sizeof(rsin);
662 rsin.sin_family = AF_INET;
663 rsin.sin_port = htons(RIP_PORT);
664 rsin.sin_addr.s_addr = addr;
665 if (bind(s, (struct sockaddr *)&rsin, sizeof(rsin)) < 0) {
667 BADERR(errno != EADDRINUSE, "bind(rip_sock)");
670 fix_sock(s,"rip_sock");
673 if (setsockopt(s, IPPROTO_IP, IP_MULTICAST_TTL,
674 &ttl, sizeof(ttl)) < 0)
675 DBGERR(1,"rip_sock setsockopt(IP_MULTICAST_TTL)");
681 /* turn off main RIP socket */
685 struct interface *ifp;
689 if (rip_sock >= 0 && !mhome) {
690 trace_act("turn off RIP");
692 (void)close(rip_sock);
695 /* get non-broadcast sockets to listen to queries.
697 LIST_FOREACH(ifp, &ifnet, int_list) {
698 if (ifp->int_state & IS_REMOTE)
700 if (ifp->int_rip_sock < 0) {
701 addr = ((ifp->int_if_flags & IFF_POINTOPOINT)
704 ifp->int_rip_sock = get_rip_sock(addr, 0);
715 /* turn on RIP multicast input via an interface
718 rip_mcast_on(struct interface *ifp)
721 struct sockaddr_in *sin;
723 if (!IS_RIP_IN_OFF(ifp->int_state)
724 && (ifp->int_if_flags & IFF_MULTICAST)
725 && !(ifp->int_state & IS_ALIAS)) {
726 memset(&gr, 0, sizeof(gr));
727 gr.gr_interface = ifp->int_index;
728 sin = (struct sockaddr_in *)&gr.gr_group;
729 sin->sin_family = AF_INET;
731 sin->sin_len = sizeof(struct sockaddr_in);
733 sin->sin_addr.s_addr = htonl(INADDR_RIP_GROUP);
734 if (setsockopt(rip_sock, IPPROTO_IP, MCAST_JOIN_GROUP,
735 &gr, sizeof(gr)) < 0)
736 LOGERR("setsockopt(MCAST_JOIN_GROUP RIP)");
741 /* Prepare socket used for RIP.
744 rip_on(struct interface *ifp)
746 /* If the main RIP socket is already alive, only start receiving
747 * multicasts for this interface.
755 /* If the main RIP socket is off and it makes sense to turn it on,
756 * then turn it on for all of the interfaces.
757 * It makes sense if either router discovery is off, or if
758 * router discover is on and at most one interface is doing RIP.
760 if (rip_interfaces > 0 && (!rdisc_ok || rip_interfaces > 1)) {
761 trace_act("turn on RIP");
763 /* Close all of the query sockets so that we can open
764 * the main socket. SO_REUSEPORT is not a solution,
765 * since that would let two daemons bind to the broadcast
768 LIST_FOREACH(ifp, &ifnet, int_list) {
769 if (ifp->int_rip_sock >= 0) {
770 (void)close(ifp->int_rip_sock);
771 ifp->int_rip_sock = -1;
775 rip_sock = get_rip_sock(INADDR_ANY, 1);
776 rip_sock_mcast = NULL;
778 /* Do not advertise anything until we have heard something
780 if (next_bcast.tv_sec < now.tv_sec+MIN_WAITTIME)
781 next_bcast.tv_sec = now.tv_sec+MIN_WAITTIME;
783 LIST_FOREACH(ifp, &ifnet, int_list) {
784 ifp->int_query_time = NEVER;
787 ifinit_timer.tv_sec = now.tv_sec;
789 } else if (ifp != NULL
790 && !(ifp->int_state & IS_REMOTE)
791 && ifp->int_rip_sock < 0) {
792 /* RIP is off, so ensure there are sockets on which
793 * to listen for queries.
795 ifp->int_rip_sock = get_rip_sock(ifp->int_addr, 0);
802 /* die if malloc(3) fails
805 rtmalloc(size_t size,
808 void *p = malloc(size);
810 logbad(1,"malloc(%lu) failed in %s", (u_long)size, msg);
815 /* get a random instant in an interval
818 intvl_random(struct timeval *tp, /* put value here */
819 u_long lo, /* value is after this second */
820 u_long hi) /* and before this */
822 tp->tv_sec = (time_t)(hi == lo
824 : (lo + arc4random_uniform(1 + hi - lo)));
825 tp->tv_usec = arc4random_uniform(1000000);
830 timevaladd(struct timeval *t1,
834 t1->tv_sec += t2->tv_sec;
835 if ((t1->tv_usec += t2->tv_usec) >= 1000000) {
837 t1->tv_usec -= 1000000;
845 timevalsub(struct timeval *t1,
849 t1->tv_sec = t2->tv_sec - t3->tv_sec;
850 if ((t1->tv_usec = t2->tv_usec - t3->tv_usec) < 0) {
852 t1->tv_usec += 1000000;
857 /* put a message into the system log
860 msglog(const char *p, ...)
867 vsyslog(LOG_ERR, p, args);
869 if (ftrace != NULL) {
870 if (ftrace == stdout)
871 (void)fputs("routed: ", ftrace);
873 (void)vfprintf(ftrace, p, args);
875 (void)fputc('\n', ftrace);
880 /* Put a message about a bad system into the system log if
881 * we have not complained about it recently.
883 * It is desirable to complain about all bad systems, but not too often.
884 * In the worst case, it is not practical to keep track of all bad systems.
885 * For example, there can be many systems with the wrong password.
888 msglim(struct msg_limit *lim, naddr addr, const char *p, ...)
892 struct msg_sub *ms1, *ms;
895 /* look for the oldest slot in the table
896 * or the slot for the bad router.
898 ms = ms1 = lim->subs;
899 for (i = MSG_SUBJECT_N; ; i--, ms1++) {
901 /* Reuse a slot at most once every 10 minutes.
903 if (lim->reuse > now.tv_sec) {
907 lim->reuse = now.tv_sec + 10*60;
911 if (ms->addr == addr) {
912 /* Repeat a complaint about a given system at
915 if (ms->until > now.tv_sec)
919 if (ms->until < ms1->until)
924 ms->until = now.tv_sec + 60*60; /* 60 minutes */
927 for (p1 = p; *p1 == ' '; p1++)
930 vsyslog(LOG_ERR, p1, args);
934 /* always display the message if tracing */
935 if (ftrace != NULL) {
937 (void)vfprintf(ftrace, p, args);
939 (void)fputc('\n', ftrace);
945 logbad(int dump, const char *p, ...)
952 vsyslog(LOG_ERR, p, args);
954 (void)fputs("routed: ", stderr);
956 (void)vfprintf(stderr, p, args);
958 (void)fputs("; giving up\n",stderr);
959 (void)fflush(stderr);