1/*
2 * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
3 * Use is subject to license terms.
4 *
5 * Copyright (c) 1983, 1988, 1993
6 *	The Regents of the University of California.  All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 *    must display the following acknowledgment:
18 *	This product includes software developed by the University of
19 *	California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * $FreeBSD: src/sbin/routed/output.c,v 1.7 2000/08/11 08:24:38 sheldonh Exp $
37 */
38
39#pragma ident	"%Z%%M%	%I%	%E% SMI"
40
41#include "defs.h"
42#include <md5.h>
43#include <alloca.h>
44
45uint_t update_seqno;
46
47
48/*
49 * walk the tree of routes with this for output
50 */
51static	struct {
52	struct sockaddr_in to;
53	in_addr_t	to_mask;
54	in_addr_t	to_net;
55	in_addr_t	to_std_mask;
56	in_addr_t	to_std_net;
57	struct interface *ifp;		/* usually output interface */
58	struct auth	*a;
59	uint8_t		metric;		/* adjust metrics by interface */
60	uint32_t	npackets;
61	uint32_t	gen_limit;
62#define	WS_GEN_LIMIT_MAX	1024
63	uint16_t	state;
64#define	WS_ST_FLASH	0x001	/* send only changed routes */
65#define	WS_ST_RIP2_ALL	0x002	/* send full featured RIPv2 */
66#define	WS_ST_AG	0x004	/* ok to aggregate subnets */
67#define	WS_ST_SUPER_AG	0x008	/* ok to aggregate networks */
68#define	WS_ST_QUERY	0x010	/* responding to a query */
69#define	WS_ST_TO_ON_NET	0x020	/* sending onto one of our nets */
70#define	WS_ST_DEFAULT	0x040	/* faking a default */
71} ws;
72
73/* A buffer for what can be heard by both RIPv1 and RIPv2 listeners */
74struct ws_buf v12buf;
75static	union pkt_buf ripv12_buf;
76
77/* Another for only RIPv2 listeners */
78static	struct ws_buf v2buf;
79static	union pkt_buf rip_v2_buf;
80
81
82
83void
84bufinit(void)
85{
86	ripv12_buf.rip.rip_cmd = RIPCMD_RESPONSE;
87	v12buf.buf = &ripv12_buf.rip;
88	v12buf.base = &v12buf.buf->rip_nets[0];
89
90	rip_v2_buf.rip.rip_cmd = RIPCMD_RESPONSE;
91	rip_v2_buf.rip.rip_vers = RIPv2;
92	v2buf.buf = &rip_v2_buf.rip;
93	v2buf.base = &v2buf.buf->rip_nets[0];
94}
95
96
97/*
98 * Send the contents of the global buffer via the non-multicast socket
99 */
100int					/* <0 on failure */
101output(enum output_type type,
102    struct sockaddr_in *dst,		/* send to here */
103    struct interface *ifp,
104    struct rip *buf,
105    int size)			/* this many bytes */
106{
107	struct sockaddr_in sin;
108	int flags;
109	const char *msg;
110	int res;
111	int ifindex;
112	struct in_addr addr;
113
114	sin = *dst;
115	if (sin.sin_port == 0)
116		sin.sin_port = htons(RIP_PORT);
117
118	flags = 0;
119
120	if (ifp == NULL && type == OUT_MULTICAST) {
121		msglog("Cannot send RIP message to %s",
122		    inet_ntoa(sin.sin_addr));
123		return (-1);
124	}
125
126	switch (type) {
127	case OUT_QUERY:
128		msg = "Answer Query";
129		break;
130	case OUT_UNICAST:
131		msg = "Send";
132		flags = MSG_DONTROUTE;
133		break;
134	case OUT_BROADCAST:
135		msg = "Send bcast";
136		break;
137	case OUT_MULTICAST:
138		msg = "Send mcast";
139		break;
140
141	case NO_OUT_MULTICAST:
142	case NO_OUT_RIPV2:
143	default:
144#ifdef DEBUG
145		abort();
146#endif
147		return (-1);
148	}
149
150	/*
151	 * IP_PKTINFO overrides IP_MULTICAST_IF, so we don't set ifindex
152	 * for multicast traffic.
153	 */
154	ifindex = (type != OUT_MULTICAST && type != OUT_QUERY &&
155	    ifp != NULL && ifp->int_phys != NULL) ?
156	    ifp->int_phys->phyi_index : 0;
157
158	if (rip_sock_interface != ifp) {
159		/*
160		 * For multicast, we have to choose the source
161		 * address.  This is either the local address
162		 * (non-point-to-point) or the remote address.
163		 */
164		if (ifp != NULL) {
165			addr.s_addr = (ifp->int_if_flags & IFF_POINTOPOINT) ?
166			    ifp->int_dstaddr : ifp->int_addr;
167			if (type == OUT_MULTICAST &&
168			    setsockopt(rip_sock, IPPROTO_IP,
169			    IP_MULTICAST_IF, &addr, sizeof (addr)) == -1) {
170				LOGERR("setsockopt(rip_sock, IP_MULTICAST_IF)");
171				return (-1);
172			}
173		}
174		rip_sock_interface = ifp;
175	}
176
177	trace_rip(msg, "to", &sin, ifp, buf, size);
178
179	res = sendtoif(rip_sock, buf, size, flags, &sin, ifindex);
180	if (res < 0 && (ifp == NULL || !(ifp->int_state & IS_BROKE))) {
181		writelog(LOG_WARNING, "%s sendto(%s%s%s.%d): %s", msg,
182		    ifp != NULL ? ifp->int_name : "",
183		    ifp != NULL ? ", " : "",
184		    inet_ntoa(sin.sin_addr),
185		    ntohs(sin.sin_port),
186		    rip_strerror(errno));
187	}
188
189	return (res);
190}
191
192/*
193 * Semantically identical to sendto(), but sends the message through a
194 * specific interface (if ifindex is non-zero) using IP_PKTINFO.
195 */
196int
197sendtoif(int fd, const void *buf, uint_t bufsize, uint_t flags,
198    struct sockaddr_in *sinp, uint_t ifindex)
199{
200	struct iovec iov;
201	struct msghdr msg;
202	struct cmsghdr *cmsgp;
203	struct in_pktinfo *ipip;
204
205	iov.iov_base = (void *)buf;
206	iov.iov_len = bufsize;
207
208	(void) memset(&msg, 0, sizeof (struct msghdr));
209	msg.msg_name = (struct sockaddr *)sinp;
210	msg.msg_namelen = sizeof (struct sockaddr_in);
211	msg.msg_iov = &iov;
212	msg.msg_iovlen = 1;
213
214	if (ifindex != 0) {
215		/*
216		 * We can't precisely predict the alignment padding we'll
217		 * need, so we allocate the maximum alignment and then
218		 * use CMSG_NXTHDR() to fix it up at the end.
219		 */
220		msg.msg_controllen = sizeof (*cmsgp) + _MAX_ALIGNMENT +
221		    sizeof (*ipip) + _MAX_ALIGNMENT + sizeof (*cmsgp);
222		msg.msg_control = alloca(msg.msg_controllen);
223
224		cmsgp = CMSG_FIRSTHDR(&msg);
225		ipip = (void *)CMSG_DATA(cmsgp);
226		(void) memset(ipip, 0, sizeof (struct in_pktinfo));
227		ipip->ipi_ifindex = ifindex;
228		cmsgp->cmsg_len = (caddr_t)(ipip + 1) - (caddr_t)cmsgp;
229		cmsgp->cmsg_type = IP_PKTINFO;
230		cmsgp->cmsg_level = IPPROTO_IP;
231
232		/*
233		 * Correct the control message length.
234		 */
235		cmsgp = CMSG_NXTHDR(&msg, cmsgp);
236		msg.msg_controllen = (caddr_t)cmsgp - (caddr_t)msg.msg_control;
237	}
238
239	return (sendmsg(fd, &msg, flags));
240}
241
242/*
243 * Find the first key for a packet to send.
244 * Try for a key that is eligible and has not expired, but settle for
245 * the last key if they have all expired.
246 * If no key is ready yet, give up.
247 */
248struct auth *
249find_auth(struct interface *ifp)
250{
251	struct auth *ap, *res = NULL;
252	int i;
253
254
255	if (ifp == NULL)
256		return (NULL);
257
258	if ((ap = ifp->int_auth) == NULL)
259		return (NULL);
260
261	for (i = 0; i < MAX_AUTH_KEYS; i++, ap++) {
262		/* stop looking after the last key */
263		if (ap->type == RIP_AUTH_NONE)
264			break;
265
266		/* ignore keys that are not ready yet */
267		if ((ulong_t)ap->start > (ulong_t)clk.tv_sec)
268			continue;
269
270		if ((ulong_t)ap->end < (ulong_t)clk.tv_sec) {
271			/* note best expired password as a fall-back */
272			if (res == NULL ||
273			    (((ulong_t)ap->end > (ulong_t)res->end)) &&
274			    ((ulong_t)res->end < (ulong_t)clk.tv_sec))
275				res = ap;
276			continue;
277		}
278
279		/* note key with the best future */
280		if (res == NULL || (ulong_t)res->end < (ulong_t)ap->end)
281			res = ap;
282	}
283	return (res);
284}
285
286
287void
288clr_ws_buf(struct ws_buf *wb, struct auth *ap)
289{
290	struct netauth *na;
291
292	wb->lim = wb->base + NETS_LEN;
293	wb->n = wb->base;
294	(void) memset(wb->n, 0, NETS_LEN*sizeof (*wb->n));
295
296	/*
297	 * (start to) install authentication if appropriate
298	 */
299	if (ap == NULL)
300		return;
301
302	na = (struct netauth *)wb->n;
303	if (ap->type == RIP_AUTH_PW) {
304		na->a_family = RIP_AF_AUTH;
305		na->a_type = RIP_AUTH_PW;
306		(void) memcpy(na->au.au_pw, ap->key, sizeof (na->au.au_pw));
307		wb->n++;
308
309	} else if (ap->type ==  RIP_AUTH_MD5) {
310		na->a_family = RIP_AF_AUTH;
311		na->a_type = RIP_AUTH_MD5;
312		na->au.a_md5.md5_keyid = ap->keyid;
313		na->au.a_md5.md5_auth_len = RIP_AUTH_MD5_LEN;
314		na->au.a_md5.md5_seqno = htonl(clk.tv_sec);
315		wb->n++;
316		wb->lim--;		/* make room for trailer */
317	}
318}
319
320
321void
322end_md5_auth(struct ws_buf *wb, struct auth *ap)
323{
324	struct netauth *na, *na2;
325	MD5_CTX md5_ctx;
326	int len;
327
328	na = (struct netauth *)wb->base;
329	na2 = (struct netauth *)wb->n;
330	len = (char *)na2-(char *)wb->buf;
331	na2->a_family = RIP_AF_AUTH;
332	na2->a_type = RIP_AUTH_TRAILER;
333	na->au.a_md5.md5_pkt_len = htons(len);
334	MD5Init(&md5_ctx);
335	/* len+4 to include auth trailer's family/type in MD5 sum */
336	MD5Update(&md5_ctx, (uchar_t *)wb->buf, len + 4);
337	MD5Update(&md5_ctx, ap->key, RIP_AUTH_MD5_LEN);
338	MD5Final(na2->au.au_pw, &md5_ctx);
339	wb->n++;
340}
341
342
343/*
344 * Send the buffer
345 */
346static void
347supply_write(struct ws_buf *wb)
348{
349	/*
350	 * Output multicast only if legal.
351	 * If we would multicast and it would be illegal, then discard the
352	 * packet.
353	 */
354	switch (wb->type) {
355	case NO_OUT_MULTICAST:
356		trace_pkt("skip multicast to %s because impossible",
357		    naddr_ntoa(ws.to.sin_addr.s_addr));
358		break;
359	case NO_OUT_RIPV2:
360		break;
361	default:
362		if (ws.a != NULL && ws.a->type == RIP_AUTH_MD5)
363			end_md5_auth(wb, ws.a);
364		if (output(wb->type, &ws.to, ws.ifp, wb->buf,
365		    ((char *)wb->n - (char *)wb->buf)) < 0 && ws.ifp != NULL)
366			if_sick(ws.ifp, _B_FALSE);
367		ws.npackets++;
368		break;
369	}
370
371	clr_ws_buf(wb, ws.a);
372}
373
374
375/*
376 * Put an entry into the packet
377 */
378static void
379supply_out(struct ag_info *ag)
380{
381	uint32_t dstcount;
382	in_addr_t mask, v1_mask, dst_h, ddst_h = 0;
383	struct ws_buf *wb;
384
385
386	/*
387	 * Skip this route if doing a flash update and it and the routes
388	 * it aggregates have not changed recently.
389	 */
390	if (ag->ag_seqno < update_seqno && (ws.state & WS_ST_FLASH))
391		return;
392
393	dst_h = ag->ag_dst_h;
394	mask = ag->ag_mask;
395	v1_mask = ripv1_mask_host(htonl(dst_h),
396	    (ws.state & WS_ST_TO_ON_NET) ? ws.ifp : NULL);
397	dstcount = 0;
398
399	/*
400	 * If we are sending RIPv2 packets that cannot (or must not) be
401	 * heard by RIPv1 listeners, do not worry about sub- or supernets.
402	 * Subnets (from other networks) can only be sent via multicast.
403	 * A pair of subnet routes might have been promoted so that they
404	 * are legal to send by RIPv1.
405	 * If RIPv1 is off, use the multicast buffer.
406	 */
407	if ((ws.state & WS_ST_RIP2_ALL) ||
408	    ((ag->ag_state & AGS_RIPV2) && v1_mask != mask)) {
409		/* use the RIPv2-only buffer */
410		wb = &v2buf;
411
412	} else {
413		/*
414		 * use the RIPv1-or-RIPv2 buffer
415		 */
416		wb = &v12buf;
417
418		/*
419		 * Convert supernet route into corresponding set of network
420		 * routes for RIPv1, but leave non-contiguous netmasks
421		 * to ag_check().
422		 */
423		if (v1_mask > mask &&
424		    mask + (mask & -mask) == 0) {
425			ddst_h = v1_mask & -v1_mask;
426			dstcount = (v1_mask & ~mask)/ddst_h;
427
428			if (dstcount > ws.gen_limit) {
429				/*
430				 * Punt if we would have to generate an
431				 * unreasonable number of routes.
432				 */
433				if (TRACECONTENTS)
434					trace_misc("sending %s-->%s as 1"
435					    " instead of %d routes",
436					    addrname(htonl(dst_h), mask, 1),
437					    naddr_ntoa(ws.to.sin_addr.s_addr),
438					    dstcount + 1);
439				dstcount = 0;
440
441			} else {
442				mask = v1_mask;
443				ws.gen_limit -= dstcount;
444			}
445		}
446	}
447
448	do {
449		wb->n->n_family = RIP_AF_INET;
450		wb->n->n_dst = htonl(dst_h);
451		/*
452		 * If the route is from router-discovery or we are
453		 * shutting down, or this is a broken/sick interface,
454		 * admit only a bad metric.
455		 */
456		wb->n->n_metric = ((stopint || ag->ag_metric < 1 ||
457		    (ag->ag_ifp && (ag->ag_ifp->int_state &
458		    (IS_BROKE|IS_SICK)))) ? HOPCNT_INFINITY : ag->ag_metric);
459		wb->n->n_metric = htonl(wb->n->n_metric);
460		/*
461		 * Any non-zero bits in the supposedly unused RIPv1 fields
462		 * cause the old `routed` to ignore the route.
463		 * That means the mask and so forth cannot be sent
464		 * in the hybrid RIPv1/RIPv2 mode.
465		 */
466		if (ws.state & WS_ST_RIP2_ALL) {
467			if (ag->ag_nhop != 0 &&
468			    ((ws.state & WS_ST_QUERY) ||
469			    (ag->ag_nhop != ws.ifp->int_addr &&
470			    on_net(ag->ag_nhop, ws.ifp->int_net,
471			    ws.ifp->int_mask)) &&
472			    ifwithaddr(ag->ag_nhop, _B_FALSE, _B_FALSE) ==
473			    NULL))
474				wb->n->n_nhop = ag->ag_nhop;
475			wb->n->n_mask = htonl(mask);
476			wb->n->n_tag = ag->ag_tag;
477		}
478		dst_h += ddst_h;
479
480		if (++wb->n >= wb->lim)
481			supply_write(wb);
482	} while (dstcount-- > 0);
483}
484
485
486/*
487 * Supply one route from the table
488 */
489/* ARGSUSED */
490static int
491walk_supply(struct radix_node *rn, void *argp)
492{
493#define	RT ((struct rt_entry *)rn)
494	ushort_t ags;
495	uint8_t metric, pref;
496	in_addr_t dst, nhop;
497	struct rt_spare *rts;
498	uint_t sparecount;
499
500
501	/*
502	 * Do not advertise external remote interfaces or passive interfaces.
503	 */
504	if ((RT->rt_state & RS_IF) && RT->rt_ifp != NULL &&
505	    (RT->rt_ifp->int_state & IS_PASSIVE) &&
506	    !(RT->rt_state & RS_MHOME))
507		return (0);
508	/*
509	 * Do not advertise routes learnt from /etc/gateways.
510	 */
511	if (RT->rt_spares[0].rts_origin == RO_FILE)
512		return (0);
513
514	/*
515	 * Do not advertise routes which would lead to forwarding on a
516	 * non-forwarding interface.
517	 */
518	if (RT->rt_state & RS_NOPROPAGATE)
519		return (0);
520
521	/*
522	 * If being quiet about our ability to forward, then
523	 * do not say anything unless responding to a query,
524	 * except about our main interface.
525	 */
526	if (!should_supply(NULL) && !(ws.state & WS_ST_QUERY) &&
527	    !(RT->rt_state & RS_MHOME))
528		return (0);
529
530	dst = RT->rt_dst;
531
532	/*
533	 * do not collide with the fake default route
534	 */
535	if (dst == RIP_DEFAULT && (ws.state & WS_ST_DEFAULT))
536		return (0);
537
538	if (RT->rt_state & RS_NET_SYN) {
539		if (RT->rt_state & RS_NET_INT) {
540			/*
541			 * Do not send manual synthetic network routes
542			 * into the subnet.
543			 */
544			if (on_net(ws.to.sin_addr.s_addr,
545			    ntohl(dst), RT->rt_mask))
546				return (0);
547
548		} else {
549			/*
550			 * Do not send automatic synthetic network routes
551			 * if they are not needed because no RIPv1 listeners
552			 * can hear them.
553			 */
554			if (ws.state & WS_ST_RIP2_ALL)
555				return (0);
556
557			/*
558			 * Do not send automatic synthetic network routes to
559			 * the real subnet.
560			 */
561			if (on_net(ws.to.sin_addr.s_addr,
562			    ntohl(dst), RT->rt_mask))
563				return (0);
564		}
565		nhop = 0;
566
567	} else {
568		/*
569		 * Advertise the next hop if this is not a route for one
570		 * of our interfaces and the next hop is on the same
571		 * network as the target.
572		 * The final determination is made by supply_out().
573		 */
574		if (!(RT->rt_state & RS_IF) && !(RT->rt_state & RS_MHOME) &&
575		    RT->rt_gate != loopaddr)
576			nhop = RT->rt_gate;
577		else
578			nhop = 0;
579	}
580
581	metric = RT->rt_metric;
582	ags = 0;
583
584	if (!RT_ISHOST(RT)) {
585		/*
586		 * Always suppress network routes into other, existing
587		 * network routes
588		 */
589		ags |= AGS_SUPPRESS;
590
591		/*
592		 * Generate supernets if allowed.
593		 * If we can be heard by RIPv1 systems, we will
594		 * later convert back to ordinary nets.
595		 * This unifies dealing with received supernets.
596		 */
597		if ((ws.state & WS_ST_AG) && ((RT->rt_state & RS_SUBNET) ||
598		    (ws.state & WS_ST_SUPER_AG)))
599			ags |= AGS_AGGREGATE;
600	} else if (!(RT->rt_state & RS_MHOME)) {
601		/*
602		 * We should always suppress (into existing network routes)
603		 * the host routes for the local end of our point-to-point
604		 * links.
605		 * If we are suppressing host routes in general, then do so.
606		 * Avoid advertising host routes onto their own network,
607		 * where they should be handled by proxy-ARP.
608		 */
609		if ((RT->rt_state & RS_LOCAL) || ridhosts ||
610		    on_net(dst, ws.to_net, ws.to_mask))
611			ags |= AGS_SUPPRESS;
612
613		/*
614		 * Aggregate stray host routes into network routes if allowed.
615		 * We cannot aggregate host routes into small network routes
616		 * without confusing RIPv1 listeners into thinking the
617		 * network routes are host routes.
618		 */
619		if ((ws.state & WS_ST_AG) && (ws.state & WS_ST_RIP2_ALL))
620			ags |= AGS_AGGREGATE;
621	}
622
623	/*
624	 * Do not send RIPv1 advertisements of subnets to other
625	 * networks. If possible, multicast them by RIPv2.
626	 */
627	if ((RT->rt_state & RS_SUBNET) && !(ws.state & WS_ST_RIP2_ALL) &&
628	    !on_net(dst, ws.to_std_net, ws.to_std_mask))
629		ags |= AGS_RIPV2 | AGS_AGGREGATE;
630
631
632	/*
633	 * Do not send a route back to where it came from, except in
634	 * response to a query.  This is "split-horizon".  That means not
635	 * advertising back to the same network	and so via the same interface.
636	 *
637	 * We want to suppress routes that might have been fragmented
638	 * from this route by a RIPv1 router and sent back to us, and so we
639	 * cannot forget this route here.  Let the split-horizon route
640	 * suppress the fragmented routes and then itself be forgotten.
641	 *
642	 * Include the routes for both ends of point-to-point interfaces
643	 * among those suppressed by split-horizon, since the other side
644	 * should knows them as well as we do.
645	 *
646	 * Notice spare routes with the same metric that we are about to
647	 * advertise, to split the horizon on redundant, inactive paths.
648	 */
649	if (ws.ifp != NULL && !(ws.state & WS_ST_QUERY) &&
650	    (ws.state & WS_ST_TO_ON_NET) && (!(RT->rt_state & RS_IF) ||
651	    (ws.ifp->int_if_flags & IFF_POINTOPOINT))) {
652		for (rts = RT->rt_spares, sparecount = 0;
653		    sparecount < RT->rt_num_spares; sparecount++, rts++) {
654			if (rts->rts_metric > metric || rts->rts_ifp != ws.ifp)
655				continue;
656
657			/*
658			 * If we do not mark the route with AGS_SPLIT_HZ here,
659			 * it will be poisoned-reverse, or advertised back
660			 * toward its source with an infinite metric.
661			 * If we have recently advertised the route with a
662			 * better metric than we now have, then we should
663			 * poison-reverse the route before suppressing it for
664			 * split-horizon.
665			 *
666			 * In almost all cases, if there is no spare for the
667			 * route then it is either old and dead or a brand
668			 * new route. If it is brand new, there is no need
669			 * for poison-reverse. If it is old and dead, it
670			 * is already poisoned.
671			 */
672			if (RT->rt_poison_time < now_expire ||
673			    RT->rt_poison_metric >= metric ||
674			    RT->rt_spares[1].rts_gate == 0) {
675				ags |= AGS_SPLIT_HZ;
676				ags &= ~AGS_SUPPRESS;
677			}
678			metric = HOPCNT_INFINITY;
679			break;
680		}
681	}
682
683	/*
684	 * Keep track of the best metric with which the
685	 * route has been advertised recently.
686	 */
687	if (RT->rt_poison_metric >= metric ||
688	    RT->rt_poison_time < now_expire) {
689		RT->rt_poison_time = now.tv_sec;
690		RT->rt_poison_metric = metric;
691	}
692
693	/*
694	 * Adjust the outgoing metric by the cost of the link.
695	 * Avoid aggregation when a route is counting to infinity.
696	 */
697	pref = RT->rt_poison_metric + ws.metric;
698	metric += ws.metric;
699
700	/*
701	 * If this is a static route pointing to the same interface
702	 * upon which we are sending out the RIP RESPONSE
703	 * adjust the preference so that we don't aggregate into this
704	 * route. Note that the maximum possible hop count on a route
705	 * per RFC 2453 is 16 (HOPCNT_INFINITY)
706	 */
707	if ((RT->rt_state & RS_STATIC) && (ws.ifp == RT->rt_ifp))
708		pref = (HOPCNT_INFINITY+1);
709
710	/*
711	 * Do not advertise stable routes that will be ignored,
712	 * unless we are answering a query.
713	 * If the route recently was advertised with a metric that
714	 * would have been less than infinity through this interface,
715	 * we need to continue to advertise it in order to poison it.
716	 */
717	if (metric >= HOPCNT_INFINITY) {
718		if (!(ws.state & WS_ST_QUERY) && (pref >= HOPCNT_INFINITY ||
719		    RT->rt_poison_time < now_garbage))
720			return (0);
721
722		metric = HOPCNT_INFINITY;
723	}
724
725	/*
726	 * supply this route out on the wire- we only care about dest/mask
727	 * and so can ignore all rt_spares[i] with i > 0
728	 */
729	ag_check(dst, RT->rt_mask, 0, RT->rt_ifp, nhop, metric, pref,
730	    RT->rt_seqno, RT->rt_tag, ags, supply_out);
731	return (0);
732#undef RT
733}
734
735
736/*
737 * Supply dst with the contents of the routing tables.
738 * If this won't fit in one packet, chop it up into several.
739 */
740void
741supply(struct sockaddr_in *dst,
742    struct interface *ifp,	/* output interface */
743    enum output_type type,
744    int flash,			/* 1=flash update */
745    int vers,			/* RIP version */
746    boolean_t passwd_ok)	/* OK to include cleartext password */
747{
748	struct rt_entry *rt;
749	uint8_t def_metric;
750
751
752	ws.state = 0;
753	ws.gen_limit = WS_GEN_LIMIT_MAX;
754
755	ws.to = *dst;
756	ws.to_std_mask = std_mask(ws.to.sin_addr.s_addr);
757	ws.to_std_net = ntohl(ws.to.sin_addr.s_addr) & ws.to_std_mask;
758
759	if (ifp != NULL) {
760		ws.to_mask = ifp->int_mask;
761		ws.to_net = ifp->int_net;
762		if (on_net(ws.to.sin_addr.s_addr, ws.to_net, ws.to_mask) ||
763		    type == OUT_MULTICAST)
764			ws.state |= WS_ST_TO_ON_NET;
765
766	} else {
767		ws.to_mask = ripv1_mask_net(ws.to.sin_addr.s_addr, NULL);
768		ws.to_net = ntohl(ws.to.sin_addr.s_addr) & ws.to_mask;
769		rt = rtfind(dst->sin_addr.s_addr);
770		if (rt != NULL)
771			ifp = rt->rt_ifp;
772		else
773			return;
774	}
775
776	ws.npackets = 0;
777	if (flash)
778		ws.state |= WS_ST_FLASH;
779
780	ws.ifp = ifp;
781
782	/*
783	 * Routes in the table were already adjusted by their respective
784	 * destination interface costs (which are zero by default) on
785	 * input.  The following is the value by which each route's metric
786	 * will be bumped up on output.
787	 */
788	ws.metric = 1;
789
790	ripv12_buf.rip.rip_vers = vers;
791
792	switch (type) {
793	case OUT_MULTICAST:
794		if (ifp->int_if_flags & IFF_MULTICAST)
795			v2buf.type = OUT_MULTICAST;
796		else
797			v2buf.type = NO_OUT_MULTICAST;
798		v12buf.type = OUT_BROADCAST;
799		break;
800
801	case OUT_QUERY:
802		ws.state |= WS_ST_QUERY;
803		/* FALLTHROUGH */
804	case OUT_BROADCAST:
805	case OUT_UNICAST:
806		v2buf.type = (vers == RIPv2) ? type : NO_OUT_RIPV2;
807		v12buf.type = type;
808		break;
809
810	case NO_OUT_MULTICAST:
811	case NO_OUT_RIPV2:
812		return;			/* no output */
813	}
814
815	if (vers == RIPv2) {
816		/* full RIPv2 only if cannot be heard by RIPv1 listeners */
817		if (type != OUT_BROADCAST)
818			ws.state |= WS_ST_RIP2_ALL;
819		if ((ws.state & WS_ST_QUERY) || !(ws.state & WS_ST_TO_ON_NET)) {
820			ws.state |= (WS_ST_AG | WS_ST_SUPER_AG);
821		} else if (ifp == NULL || !(ifp->int_state & IS_NO_AG)) {
822			ws.state |= WS_ST_AG;
823			if (type != OUT_BROADCAST && (ifp == NULL ||
824			    !(ifp->int_state & IS_NO_SUPER_AG)))
825				ws.state |= WS_ST_SUPER_AG;
826		}
827
828		/* See if this packet needs authenticating */
829		ws.a = find_auth(ifp);
830		if (!passwd_ok && ws.a != NULL && ws.a->type == RIP_AUTH_PW)
831			ws.a = NULL;
832		if (ws.a != NULL && (ulong_t)ws.a->end < (ulong_t)clk.tv_sec &&
833		    !ws.a->warnedflag) {
834			/*
835			 * If the best key is an expired one, we may as
836			 * well use it.  Log this event.
837			 */
838			writelog(LOG_WARNING,
839			    "Using expired auth while transmitting to %s",
840			    naddr_ntoa(ws.to.sin_addr.s_addr));
841			ws.a->warnedflag = 1;
842		}
843	} else {
844		ws.a = NULL;
845	}
846
847	clr_ws_buf(&v12buf, ws.a);
848	clr_ws_buf(&v2buf, ws.a);
849
850	/*
851	 * Fake a default route if asked and if there is not already
852	 * a better, real default route.
853	 */
854	if (should_supply(NULL) && (def_metric = ifp->int_d_metric) != 0) {
855		if (NULL == (rt = rtget(RIP_DEFAULT, 0)) ||
856		    rt->rt_metric+ws.metric >= def_metric) {
857			ws.state |= WS_ST_DEFAULT;
858			ag_check(0, 0, 0, NULL, 0, def_metric, def_metric,
859			    0, 0, 0, supply_out);
860		} else {
861			def_metric = rt->rt_metric+ws.metric;
862		}
863
864		/*
865		 * If both RIPv2 and the poor-man's router discovery
866		 * kludge are on, arrange to advertise an extra
867		 * default route via RIPv1.
868		 */
869		if ((ws.state & WS_ST_RIP2_ALL) &&
870		    (ifp->int_state & IS_PM_RDISC)) {
871			ripv12_buf.rip.rip_vers = RIPv1;
872			v12buf.n->n_family = RIP_AF_INET;
873			v12buf.n->n_dst = htonl(RIP_DEFAULT);
874			v12buf.n->n_metric = htonl(def_metric);
875			v12buf.n++;
876		}
877	}
878
879	(void) rn_walktree(rhead, walk_supply, NULL);
880	ag_flush(0, 0, supply_out);
881
882	/*
883	 * Flush the packet buffers, provided they are not empty and
884	 * do not contain only the password.
885	 */
886	if (v12buf.n != v12buf.base &&
887	    (v12buf.n > v12buf.base+1 ||
888	    v12buf.base->n_family != RIP_AF_AUTH))
889		supply_write(&v12buf);
890	if (v2buf.n != v2buf.base && (v2buf.n > v2buf.base+1 ||
891	    v2buf.base->n_family != RIP_AF_AUTH))
892		supply_write(&v2buf);
893
894	/*
895	 * If we sent nothing and this is an answer to a query, send
896	 * an empty buffer.
897	 */
898	if (ws.npackets == 0 && (ws.state & WS_ST_QUERY)) {
899		supply_write(&v2buf);
900		if (ws.npackets == 0)
901			supply_write(&v12buf);
902	}
903}
904
905
906/*
907 * send all of the routing table or just do a flash update
908 */
909void
910rip_bcast(int flash)
911{
912	static struct sockaddr_in dst = {AF_INET};
913	struct interface *ifp;
914	enum output_type type;
915	int vers;
916	struct timeval rtime;
917
918
919	need_flash = _B_FALSE;
920	intvl_random(&rtime, MIN_WAITTIME, MAX_WAITTIME);
921	no_flash = rtime;
922	timevaladd(&no_flash, &now);
923
924	if (!rip_enabled)
925		return;
926
927	trace_act("send %s and inhibit dynamic updates for %.3f sec",
928	    flash ? "dynamic update" : "all routes",
929	    rtime.tv_sec + ((double)rtime.tv_usec)/1000000.0);
930
931	for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) {
932		/*
933		 * Skip interfaces not doing RIP or for which IP
934		 * forwarding isn't turned on.  Skip duplicate
935		 * interfaces, we don't want to generate duplicate
936		 * packets.  Do try broken interfaces to see if they
937		 * have healed.
938		 */
939		if (IS_RIP_OUT_OFF(ifp->int_state) ||
940		    (ifp->int_state & IS_DUP) ||
941		    !IS_IFF_ROUTING(ifp->int_if_flags))
942			continue;
943
944		/* skip turned off interfaces */
945		if (!IS_IFF_UP(ifp->int_if_flags))
946			continue;
947
948		/* skip interfaces we shouldn't use */
949		if (IS_IFF_QUIET(ifp->int_if_flags))
950			continue;
951
952		vers = (ifp->int_state & IS_NO_RIPV1_OUT) ? RIPv2 : RIPv1;
953		dst.sin_addr.s_addr = ifp->int_ripout_addr;
954
955		/*
956		 * Ignore the interface if it's not broadcast,
957		 * point-to-point, or remote.  It must be non-broadcast
958		 * multiaccess, and therefore unsupported.
959		 */
960		if (!(ifp->int_if_flags & (IFF_BROADCAST | IFF_POINTOPOINT)) &&
961		    !(ifp->int_state & IS_REMOTE))
962			continue;
963
964		type = (ifp->int_if_flags & IFF_BROADCAST) ?
965		    OUT_BROADCAST : OUT_UNICAST;
966		if (vers == RIPv2 && (ifp->int_if_flags & IFF_MULTICAST) &&
967		    !(ifp->int_state & IS_NO_RIP_MCAST))
968			type = OUT_MULTICAST;
969
970		supply(&dst, ifp, type, flash, vers, _B_TRUE);
971	}
972
973	update_seqno++;			/* all routes are up to date */
974}
975
976
977/*
978 * Ask for routes
979 * Do it only once to an interface, and not even after the interface
980 * was broken and recovered.
981 */
982void
983rip_query(void)
984{
985	static struct sockaddr_in dst = {AF_INET};
986	struct interface *ifp;
987	struct rip buf;
988	enum output_type type;
989
990
991	if (!rip_enabled)
992		return;
993
994	(void) memset(&buf, 0, sizeof (buf));
995
996	for (ifp = ifnet; ifp; ifp = ifp->int_next) {
997		/*
998		 * Skip interfaces those already queried.  Do not ask
999		 * via interfaces through which we don't accept input.
1000		 * Do not ask via interfaces that cannot send RIP
1001		 * packets.  Don't send queries on duplicate
1002		 * interfaces, that would generate duplicate packets
1003		 * on link.  Do try broken interfaces to see if they
1004		 * have healed.
1005		 */
1006		if (IS_RIP_IN_OFF(ifp->int_state) ||
1007		    (ifp->int_state & IS_DUP) ||
1008		    ifp->int_query_time != NEVER)
1009			continue;
1010
1011		/* skip turned off interfaces */
1012		if (!IS_IFF_UP(ifp->int_if_flags))
1013			continue;
1014
1015		/* skip interfaces we shouldn't use */
1016		if (IS_IFF_QUIET(ifp->int_if_flags))
1017			continue;
1018
1019		/*
1020		 * Ignore the interface if it's not broadcast,
1021		 * point-to-point, or remote.  It must be non-broadcast
1022		 * multiaccess, and therefore unsupported.
1023		 */
1024		if (!(ifp->int_if_flags & (IFF_BROADCAST | IFF_POINTOPOINT)) &&
1025		    !(ifp->int_state & IS_REMOTE))
1026			continue;
1027
1028		buf.rip_cmd = RIPCMD_REQUEST;
1029		buf.rip_nets[0].n_family = RIP_AF_UNSPEC;
1030		buf.rip_nets[0].n_metric = htonl(HOPCNT_INFINITY);
1031
1032		/*
1033		 * Send a RIPv1 query only if allowed and if we will
1034		 * listen to RIPv1 routers.
1035		 */
1036		if ((ifp->int_state & IS_NO_RIPV1_OUT) ||
1037		    (ifp->int_state & IS_NO_RIPV1_IN)) {
1038			buf.rip_vers = RIPv2;
1039		} else {
1040			buf.rip_vers = RIPv1;
1041		}
1042
1043		dst.sin_addr.s_addr = ifp->int_ripout_addr;
1044
1045		type = (ifp->int_if_flags & IFF_BROADCAST) ?
1046		    OUT_BROADCAST : OUT_UNICAST;
1047		if (buf.rip_vers == RIPv2 &&
1048		    (ifp->int_if_flags & IFF_MULTICAST) &&
1049		    !(ifp->int_state & IS_NO_RIP_MCAST))
1050			type = OUT_MULTICAST;
1051
1052		ifp->int_query_time = now.tv_sec+SUPPLY_INTERVAL;
1053		if (output(type, &dst, ifp, &buf, sizeof (buf)) < 0)
1054			if_sick(ifp, _B_FALSE);
1055	}
1056}
1057