xref: /openbsd-src/sys/netinet/if_ether.c (revision 5b859c19fe53bbea08f5c342e0a4470e99f883e1)
1 /*	$OpenBSD: if_ether.c,v 1.137 2014/11/01 21:40:38 mpi Exp $	*/
2 /*	$NetBSD: if_ether.c,v 1.31 1996/05/11 12:59:58 mycroft Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 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. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
33  */
34 
35 /*
36  * Ethernet address resolution protocol.
37  * TODO:
38  *	add "inuse/lock" bit (or ref. count) along with valid bit
39  */
40 
41 #ifdef INET
42 #include "carp.h"
43 
44 #include "bridge.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/timeout.h>
51 #include <sys/kernel.h>
52 #include <sys/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/pool.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/route.h>
59 #include <net/if_types.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_var.h>
63 #include <netinet/if_ether.h>
64 #if NCARP > 0
65 #include <netinet/ip_carp.h>
66 #endif
67 #if NBRIDGE > 0
68 #include <net/if_bridge.h>
69 #endif
70 
71 #define SDL(s) ((struct sockaddr_dl *)s)
72 #define SRP(s) ((struct sockaddr_inarp *)s)
73 
74 /*
75  * ARP trailer negotiation.  Trailer protocol is not IP specific,
76  * but ARP request/response use IP addresses.
77  */
78 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
79 
80 /* timer values */
81 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
82 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
83 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
84 #define	rt_expire rt_rmx.rmx_expire
85 
86 void arptfree(struct llinfo_arp *);
87 void arptimer(void *);
88 struct llinfo_arp *arplookup(u_int32_t, int, int, u_int);
89 void in_arpinput(struct mbuf *);
90 
91 LIST_HEAD(, llinfo_arp) llinfo_arp;
92 struct	pool arp_pool;		/* pool for llinfo_arp structures */
93 struct	ifqueue arpintrq;
94 int	arp_inuse, arp_allocated;
95 int	arp_maxtries = 5;
96 int	arpinit_done;
97 int	la_hold_total;
98 
99 #ifdef NFSCLIENT
100 /* revarp state */
101 struct in_addr revarp_myip, revarp_srvip;
102 int revarp_finished;
103 int revarp_in_progress;
104 struct ifnet *revarp_ifp;
105 #endif /* NFSCLIENT */
106 
107 #ifdef DDB
108 
109 void	db_print_sa(struct sockaddr *);
110 void	db_print_ifa(struct ifaddr *);
111 void	db_print_llinfo(caddr_t);
112 int	db_show_radix_node(struct radix_node *, void *, u_int);
113 #endif
114 
115 static const struct sockaddr_dl null_sdl = { sizeof(null_sdl), AF_LINK };
116 
117 /*
118  * Timeout routine.  Age arp_tab entries periodically.
119  */
120 /* ARGSUSED */
121 void
122 arptimer(void *arg)
123 {
124 	struct timeout *to = (struct timeout *)arg;
125 	int s;
126 	struct llinfo_arp *la, *nla;
127 
128 	s = splsoftnet();
129 	timeout_add_sec(to, arpt_prune);
130 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
131 		struct rtentry *rt = la->la_rt;
132 
133 		nla = LIST_NEXT(la, la_list);
134 		if (rt->rt_expire && rt->rt_expire <= time_second)
135 			arptfree(la); /* timer has expired; clear */
136 	}
137 	splx(s);
138 }
139 
140 void
141 arp_rtrequest(int req, struct rtentry *rt)
142 {
143 	struct sockaddr *gate = rt->rt_gateway;
144 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
145 	struct ifnet *ifp = rt->rt_ifp;
146 	struct ifaddr *ifa;
147 	struct mbuf *m;
148 
149 	if (!arpinit_done) {
150 		static struct timeout arptimer_to;
151 
152 		arpinit_done = 1;
153 		pool_init(&arp_pool, sizeof(struct llinfo_arp), 0, 0, 0, "arp",
154 		    NULL);
155 		IFQ_SET_MAXLEN(&arpintrq, 50);	/* XXX hate magic numbers */
156 		/*
157 		 * We generate expiration times from time.tv_sec
158 		 * so avoid accidently creating permanent routes.
159 		 */
160 		if (time_second == 0) {
161 			time_second++;
162 		}
163 
164 		timeout_set(&arptimer_to, arptimer, &arptimer_to);
165 		timeout_add_sec(&arptimer_to, 1);
166 	}
167 
168 	if (rt->rt_flags & RTF_GATEWAY)
169 		return;
170 
171 	switch (req) {
172 
173 	case RTM_ADD:
174 		/*
175 		 * XXX: If this is a manually added route to interface
176 		 * such as older version of routed or gated might provide,
177 		 * restore cloning bit.
178 		 */
179 		if ((rt->rt_flags & RTF_HOST) == 0 && rt_mask(rt) &&
180 		    satosin(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
181 			rt->rt_flags |= RTF_CLONING;
182 		if (rt->rt_flags & RTF_CLONING ||
183 		    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !la)) {
184 			/*
185 			 * Case 1: This route should come from a route to iface.
186 			 */
187 			rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl,
188 			    ifp->if_rdomain);
189 			gate = rt->rt_gateway;
190 			SDL(gate)->sdl_type = ifp->if_type;
191 			SDL(gate)->sdl_index = ifp->if_index;
192 			/*
193 			 * Give this route an expiration time, even though
194 			 * it's a "permanent" route, so that routes cloned
195 			 * from it do not need their expiration time set.
196 			 */
197 			rt->rt_expire = time_second;
198 			if ((rt->rt_flags & RTF_CLONING) != 0)
199 				break;
200 		}
201 		/* Announce a new entry if requested. */
202 		if (rt->rt_flags & RTF_ANNOUNCE)
203 			arprequest(ifp,
204 			    &satosin(rt_key(rt))->sin_addr.s_addr,
205 			    &satosin(rt_key(rt))->sin_addr.s_addr,
206 			    (u_char *)LLADDR(SDL(gate)));
207 		/*FALLTHROUGH*/
208 	case RTM_RESOLVE:
209 		if (gate->sa_family != AF_LINK ||
210 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
211 			log(LOG_DEBUG, "%s: bad gateway value: %s\n", __func__,
212 			    ifp->if_xname);
213 			break;
214 		}
215 		SDL(gate)->sdl_type = ifp->if_type;
216 		SDL(gate)->sdl_index = ifp->if_index;
217 		if (la != 0)
218 			break; /* This happens on a route change */
219 		/*
220 		 * Case 2:  This route may come from cloning, or a manual route
221 		 * add with a LL address.
222 		 */
223 		la = pool_get(&arp_pool, PR_NOWAIT | PR_ZERO);
224 		rt->rt_llinfo = (caddr_t)la;
225 		if (la == NULL) {
226 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
227 			break;
228 		}
229 		arp_inuse++;
230 		arp_allocated++;
231 		la->la_rt = rt;
232 		rt->rt_flags |= RTF_LLINFO;
233 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
234 
235 		/*
236 		 * Routes to broadcast addresses must be incomplete
237 		 * arp entries so that they won't be picked up, but
238 		 * since we expect them to always be present in the
239 		 * routing table, make sure arptimer() won't free
240 		 * them.
241 		 */
242 		if (rt->rt_flags & RTF_BROADCAST) {
243 			rt->rt_expire = 0;
244 			break;
245 		}
246 
247 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
248 			if ((ifa->ifa_addr->sa_family == AF_INET) &&
249 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr ==
250 			    satosin(rt_key(rt))->sin_addr.s_addr)
251 				break;
252 		}
253 		if (ifa) {
254 			rt->rt_expire = 0;
255 			SDL(gate)->sdl_alen = ETHER_ADDR_LEN;
256 			memcpy(LLADDR(SDL(gate)),
257 			    ((struct arpcom *)ifp)->ac_enaddr, ETHER_ADDR_LEN);
258 
259 			/*
260 			 * XXX Since lo0 is in the default rdomain we
261 			 * should not (ab)use it for any route related
262 			 * to an interface of a different rdomain.
263 			 */
264 			rt->rt_ifp = lo0ifp;
265 
266 			/*
267 			 * make sure to set rt->rt_ifa to the interface
268 			 * address we are using, otherwise we will have trouble
269 			 * with source address selection.
270 			 */
271 			if (ifa != rt->rt_ifa) {
272 				ifafree(rt->rt_ifa);
273 				ifa->ifa_refcnt++;
274 				rt->rt_ifa = ifa;
275 			}
276 		}
277 		break;
278 
279 	case RTM_DELETE:
280 		if (la == 0)
281 			break;
282 		arp_inuse--;
283 		LIST_REMOVE(la, la_list);
284 		rt->rt_llinfo = 0;
285 		rt->rt_flags &= ~RTF_LLINFO;
286 		while ((m = la->la_hold_head) != NULL) {
287 			la->la_hold_head = la->la_hold_head->m_nextpkt;
288 			la_hold_total--;
289 			m_freem(m);
290 		}
291 		pool_put(&arp_pool, la);
292 	}
293 }
294 
295 /*
296  * Broadcast an ARP request. Caller specifies:
297  *	- arp header source ip address
298  *	- arp header target ip address
299  *	- arp header source ethernet address
300  */
301 void
302 arprequest(struct ifnet *ifp, u_int32_t *sip, u_int32_t *tip, u_int8_t *enaddr)
303 {
304 	struct mbuf *m;
305 	struct ether_header *eh;
306 	struct ether_arp *ea;
307 	struct sockaddr sa;
308 
309 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
310 		return;
311 	m->m_len = sizeof(*ea);
312 	m->m_pkthdr.len = sizeof(*ea);
313 	m->m_pkthdr.ph_rtableid = ifp->if_rdomain;
314 	MH_ALIGN(m, sizeof(*ea));
315 	ea = mtod(m, struct ether_arp *);
316 	eh = (struct ether_header *)sa.sa_data;
317 	memset(ea, 0, sizeof(*ea));
318 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
319 	eh->ether_type = htons(ETHERTYPE_ARP);	/* if_output will not swap */
320 	ea->arp_hrd = htons(ARPHRD_ETHER);
321 	ea->arp_pro = htons(ETHERTYPE_IP);
322 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
323 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
324 	ea->arp_op = htons(ARPOP_REQUEST);
325 	memcpy(eh->ether_shost, enaddr, sizeof(eh->ether_shost));
326 	memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
327 	memcpy(ea->arp_spa, sip, sizeof(ea->arp_spa));
328 	memcpy(ea->arp_tpa, tip, sizeof(ea->arp_tpa));
329 	sa.sa_family = pseudo_AF_HDRCMPLT;
330 	sa.sa_len = sizeof(sa);
331 	m->m_flags |= M_BCAST;
332 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
333 }
334 
335 /*
336  * Resolve an IP address into an ethernet address.  If success,
337  * desten is filled in.  If there is no entry in arptab,
338  * set one up and broadcast a request for the IP address.
339  * Hold onto this mbuf and resend it once the address
340  * is finally resolved.  A return value of 1 indicates
341  * that desten has been filled in and the packet should be sent
342  * normally; a 0 return indicates that the packet has been
343  * taken over here, either now or for later transmission.
344  */
345 int
346 arpresolve(struct arpcom *ac, struct rtentry *rt, struct mbuf *m,
347     struct sockaddr *dst, u_char *desten)
348 {
349 	struct llinfo_arp *la;
350 	struct sockaddr_dl *sdl;
351 	struct mbuf *mh;
352 	char addr[INET_ADDRSTRLEN];
353 
354 	if (m->m_flags & M_BCAST) {	/* broadcast */
355 		memcpy(desten, etherbroadcastaddr, sizeof(etherbroadcastaddr));
356 		return (1);
357 	}
358 	if (m->m_flags & M_MCAST) {	/* multicast */
359 		ETHER_MAP_IP_MULTICAST(&satosin(dst)->sin_addr, desten);
360 		return (1);
361 	}
362 	if (rt) {
363 		la = (struct llinfo_arp *)rt->rt_llinfo;
364 		if (la == NULL)
365 			log(LOG_DEBUG, "arpresolve: %s: route without link "
366 			    "local address\n", inet_ntop(AF_INET,
367 				&satosin(dst)->sin_addr, addr, sizeof(addr)));
368 	} else {
369 		if ((la = arplookup(satosin(dst)->sin_addr.s_addr, 1, 0,
370 		    ac->ac_if.if_rdomain)) != NULL)
371 			rt = la->la_rt;
372 		else
373 			log(LOG_DEBUG,
374 			    "arpresolve: %s: can't allocate llinfo\n",
375 			    inet_ntop(AF_INET, &satosin(dst)->sin_addr,
376 				addr, sizeof(addr)));
377 	}
378 	if (la == 0 || rt == 0) {
379 		m_freem(m);
380 		return (0);
381 	}
382 	sdl = SDL(rt->rt_gateway);
383 	/*
384 	 * Check the address family and length is valid, the address
385 	 * is resolved; otherwise, try to resolve.
386 	 */
387 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
388 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
389 		memcpy(desten, LLADDR(sdl), sdl->sdl_alen);
390 		return 1;
391 	}
392 	if (((struct ifnet *)ac)->if_flags & IFF_NOARP) {
393 		m_freem(m);
394 		return 0;
395 	}
396 
397 	/*
398 	 * There is an arptab entry, but no ethernet address
399 	 * response yet. Insert mbuf in hold queue if below limit
400 	 * if above the limit free the queue without queuing the new packet.
401 	 */
402 	if (la_hold_total < MAX_HOLD_TOTAL && la_hold_total < nmbclust / 64) {
403 		if (la->la_hold_count >= MAX_HOLD_QUEUE) {
404 			mh = la->la_hold_head;
405 			la->la_hold_head = la->la_hold_head->m_nextpkt;
406 			if (mh == la->la_hold_tail)
407 				la->la_hold_tail = NULL;
408 			la->la_hold_count--;
409 			la_hold_total--;
410 			m_freem(mh);
411 		}
412 		if (la->la_hold_tail == NULL)
413 			la->la_hold_head = m;
414 		else
415 			la->la_hold_tail->m_nextpkt = m;
416 		la->la_hold_tail = m;
417 		la->la_hold_count++;
418 		la_hold_total++;
419 	} else {
420 		while ((mh = la->la_hold_head) != NULL) {
421 			la->la_hold_head =
422 			    la->la_hold_head->m_nextpkt;
423 			la_hold_total--;
424 			m_freem(mh);
425 		}
426 		la->la_hold_tail = NULL;
427 		la->la_hold_count = 0;
428 		m_freem(m);
429 	}
430 
431 	/*
432 	 * Re-send the ARP request when appropriate.
433 	 */
434 #ifdef	DIAGNOSTIC
435 	if (rt->rt_expire == 0) {
436 		/* This should never happen. (Should it? -gwr) */
437 		printf("arpresolve: unresolved and rt_expire == 0\n");
438 		/* Set expiration time to now (expired). */
439 		rt->rt_expire = time_second;
440 	}
441 #endif
442 	if (rt->rt_expire) {
443 		rt->rt_flags &= ~RTF_REJECT;
444 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
445 			rt->rt_expire = time_second;
446 			if (la->la_asked++ < arp_maxtries)
447 				arprequest(&ac->ac_if,
448 				    &satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr,
449 				    &satosin(dst)->sin_addr.s_addr,
450 #if NCARP > 0
451 				    (rt->rt_ifp->if_type == IFT_CARP) ?
452 					((struct arpcom *) rt->rt_ifp->if_softc
453 					)->ac_enaddr :
454 #endif
455 				    ac->ac_enaddr);
456 			else {
457 				rt->rt_flags |= RTF_REJECT;
458 				rt->rt_expire += arpt_down;
459 				la->la_asked = 0;
460 				while ((mh = la->la_hold_head) != NULL) {
461 					la->la_hold_head =
462 					    la->la_hold_head->m_nextpkt;
463 					la_hold_total--;
464 					m_freem(mh);
465 				}
466 				la->la_hold_tail = NULL;
467 				la->la_hold_count = 0;
468 			}
469 		}
470 	}
471 	return (0);
472 }
473 
474 /*
475  * Common length and type checks are done here,
476  * then the protocol-specific routine is called.
477  */
478 void
479 arpintr(void)
480 {
481 	struct mbuf *m;
482 	struct arphdr *ar;
483 	int s, len;
484 
485 	for (;;) {
486 		s = splnet();
487 		IF_DEQUEUE(&arpintrq, m);
488 		splx(s);
489 		if (m == NULL)
490 			break;
491 #ifdef DIAGNOSTIC
492 		if ((m->m_flags & M_PKTHDR) == 0)
493 			panic("arpintr");
494 #endif
495 
496 		len = sizeof(struct arphdr);
497 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
498 			continue;
499 
500 		ar = mtod(m, struct arphdr *);
501 		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) {
502 			m_freem(m);
503 			continue;
504 		}
505 
506 		len += 2 * (ar->ar_hln + ar->ar_pln);
507 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
508 			continue;
509 
510 		switch (ntohs(ar->ar_pro)) {
511 		case ETHERTYPE_IP:
512 		case ETHERTYPE_IPTRAILERS:
513 			in_arpinput(m);
514 			continue;
515 		}
516 		m_freem(m);
517 	}
518 }
519 
520 /*
521  * ARP for Internet protocols on Ethernet.
522  * Algorithm is that given in RFC 826.
523  * In addition, a sanity check is performed on the sender
524  * protocol address, to catch impersonators.
525  * We no longer handle negotiations for use of trailer protocol:
526  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
527  * along with IP replies if we wanted trailers sent to us,
528  * and also sent them in response to IP replies.
529  * This allowed either end to announce the desire to receive
530  * trailer packets.
531  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
532  * but formerly didn't normally send requests.
533  */
534 void
535 in_arpinput(struct mbuf *m)
536 {
537 	struct ether_arp *ea;
538 	struct ifnet *ifp = m->m_pkthdr.rcvif;
539 	struct arpcom *ac = (struct arpcom *)ifp;
540 	struct ether_header *eh;
541 	struct llinfo_arp *la = 0;
542 	struct rtentry *rt;
543 	struct ifaddr *ifa;
544 	struct sockaddr_dl *sdl;
545 	struct sockaddr sa;
546 	struct in_addr isaddr, itaddr, myaddr;
547 	struct mbuf *mh, *mt;
548 	u_int8_t *enaddr = NULL;
549 #if NCARP > 0
550 	u_int8_t *ether_shost = NULL;
551 #endif
552 	char addr[INET_ADDRSTRLEN];
553 	int op;
554 
555 	ea = mtod(m, struct ether_arp *);
556 	op = ntohs(ea->arp_op);
557 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
558 		goto out;
559 #if notyet
560 	if ((op == ARPOP_REPLY) && (m->m_flags & (M_BCAST|M_MCAST))) {
561 		log(LOG_ERR,
562 		    "arp: received reply to broadcast or multicast address\n");
563 		goto out;
564 	}
565 #endif
566 
567 	memcpy(&itaddr, ea->arp_tpa, sizeof(itaddr));
568 	memcpy(&isaddr, ea->arp_spa, sizeof(isaddr));
569 
570 	/* First try: check target against our addresses */
571 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
572 		if (ifa->ifa_addr->sa_family != AF_INET)
573 			continue;
574 
575 		if (itaddr.s_addr != ifatoia(ifa)->ia_addr.sin_addr.s_addr)
576 			continue;
577 
578 #if NCARP > 0
579 		if (ifp->if_type == IFT_CARP &&
580 		    ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
581 		    (IFF_UP|IFF_RUNNING))) {
582 			if (op == ARPOP_REPLY)
583 				break;
584 			if (carp_iamatch(ifatoia(ifa), ea->arp_sha,
585 			    &enaddr, &ether_shost))
586 				break;
587 			else
588 				goto out;
589 		}
590 #endif
591 		break;
592 	}
593 
594 	/* Second try: check source against our addresses */
595 	if (ifa == NULL) {
596 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
597 			if (ifa->ifa_addr->sa_family != AF_INET)
598 				continue;
599 
600 			if (isaddr.s_addr ==
601 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr)
602 				break;
603 		}
604 	}
605 
606 	/* Third try: not one of our addresses, just find an usable ia */
607 	if (ifa == NULL) {
608 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
609 			if (ifa->ifa_addr->sa_family == AF_INET)
610 				break;
611 		}
612 	}
613 
614 	if (ifa == NULL)
615 		goto out;
616 
617 	if (!enaddr)
618 		enaddr = ac->ac_enaddr;
619 	myaddr = ifatoia(ifa)->ia_addr.sin_addr;
620 
621 	if (!memcmp(ea->arp_sha, enaddr, sizeof(ea->arp_sha)))
622 		goto out;	/* it's from me, ignore it. */
623 	if (ETHER_IS_MULTICAST(&ea->arp_sha[0]))
624 		if (!memcmp(ea->arp_sha, etherbroadcastaddr,
625 		    sizeof (ea->arp_sha))) {
626 			inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
627 			log(LOG_ERR, "arp: ether address is broadcast for "
628 			    "IP address %s!\n", addr);
629 			goto out;
630 		}
631 	if (myaddr.s_addr && isaddr.s_addr == myaddr.s_addr) {
632 		inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
633 		log(LOG_ERR,
634 		   "duplicate IP address %s sent from ethernet address %s\n",
635 		   addr, ether_sprintf(ea->arp_sha));
636 		itaddr = myaddr;
637 		goto reply;
638 	}
639 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0,
640 	    rtable_l2(m->m_pkthdr.ph_rtableid));
641 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
642 		if (sdl->sdl_alen) {
643 		    if (memcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
644 			if (rt->rt_flags &
645 			    (RTF_PERMANENT_ARP|RTF_LOCAL|RTF_BROADCAST)) {
646 				inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
647 				log(LOG_WARNING,
648 				   "arp: attempt to overwrite permanent "
649 				   "entry for %s by %s on %s\n", addr,
650 				   ether_sprintf(ea->arp_sha),
651 				   ac->ac_if.if_xname);
652 				goto out;
653 			} else if (rt->rt_ifp != &ac->ac_if) {
654 #if NCARP > 0
655 				if (ac->ac_if.if_type != IFT_CARP)
656 #endif
657 				{
658 					inet_ntop(AF_INET, &isaddr,
659 					    addr, sizeof(addr));
660 					log(LOG_WARNING,
661 					   "arp: attempt to overwrite entry for"
662 					   " %s on %s by %s on %s\n", addr,
663 					   rt->rt_ifp->if_xname,
664 					   ether_sprintf(ea->arp_sha),
665 					   ac->ac_if.if_xname);
666 				}
667 				goto out;
668 			} else {
669 				inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
670 				log(LOG_INFO,
671 				   "arp info overwritten for %s by %s on %s\n",
672 				   addr,
673 				   ether_sprintf(ea->arp_sha),
674 				   ac->ac_if.if_xname);
675 				rt->rt_expire = 1; /* no longer static */
676 			}
677 		    }
678 		} else if (rt->rt_ifp != &ac->ac_if &&
679 #if NBRIDGE > 0
680 		    !SAME_BRIDGE(ac->ac_if.if_bridgeport,
681 		    rt->rt_ifp->if_bridgeport) &&
682 #endif
683 #if NCARP > 0
684 		    !(rt->rt_ifp->if_type == IFT_CARP &&
685 		    rt->rt_ifp->if_carpdev == &ac->ac_if) &&
686 		    !(ac->ac_if.if_type == IFT_CARP &&
687 		    ac->ac_if.if_carpdev == rt->rt_ifp) &&
688 #endif
689 		    1) {
690 			inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
691 			log(LOG_WARNING,
692 			    "arp: attempt to add entry for %s "
693 			    "on %s by %s on %s\n", addr,
694 			    rt->rt_ifp->if_xname,
695 			    ether_sprintf(ea->arp_sha),
696 			    ac->ac_if.if_xname);
697 			goto out;
698 		}
699  		sdl->sdl_alen = sizeof(ea->arp_sha);
700 		memcpy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
701 		if (rt->rt_expire)
702 			rt->rt_expire = time_second + arpt_keep;
703 		rt->rt_flags &= ~RTF_REJECT;
704 		la->la_asked = 0;
705 		rt_sendmsg(rt, RTM_RESOLVE, rt->rt_ifp->if_rdomain);
706 		while ((mh = la->la_hold_head) != NULL) {
707 			if ((la->la_hold_head = mh->m_nextpkt) == NULL)
708 				la->la_hold_tail = NULL;
709 			la->la_hold_count--;
710 			la_hold_total--;
711 			mt = la->la_hold_tail;
712 
713 			(*ac->ac_if.if_output)(&ac->ac_if, mh, rt_key(rt), rt);
714 
715 			if (la->la_hold_tail == mh) {
716 				/* mbuf is back in queue. Discard. */
717 				la->la_hold_tail = mt;
718 				if (la->la_hold_tail)
719 					la->la_hold_tail->m_nextpkt = NULL;
720 				else
721 					la->la_hold_head = NULL;
722 				la->la_hold_count--;
723 				la_hold_total--;
724 				m_freem(mh);
725 			}
726 		}
727 	}
728 reply:
729 	if (op != ARPOP_REQUEST) {
730 out:
731 		m_freem(m);
732 		return;
733 	}
734 	if (itaddr.s_addr == myaddr.s_addr) {
735 		/* I am the target */
736 		memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
737 		memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
738 	} else {
739 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY,
740 		    rtable_l2(m->m_pkthdr.ph_rtableid));
741 		if (la == 0)
742 			goto out;
743 		rt = la->la_rt;
744 		if (rt->rt_ifp->if_type == IFT_CARP && ifp->if_type != IFT_CARP)
745 			goto out;
746 		memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
747 		sdl = SDL(rt->rt_gateway);
748 		memcpy(ea->arp_sha, LLADDR(sdl), sizeof(ea->arp_sha));
749 	}
750 
751 	memcpy(ea->arp_tpa, ea->arp_spa, sizeof(ea->arp_spa));
752 	memcpy(ea->arp_spa, &itaddr, sizeof(ea->arp_spa));
753 	ea->arp_op = htons(ARPOP_REPLY);
754 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
755 	eh = (struct ether_header *)sa.sa_data;
756 	memcpy(eh->ether_dhost, ea->arp_tha, sizeof(eh->ether_dhost));
757 #if NCARP > 0
758 	if (ether_shost)
759 		enaddr = ether_shost;
760 #endif
761 	memcpy(eh->ether_shost, enaddr, sizeof(eh->ether_shost));
762 
763 	eh->ether_type = htons(ETHERTYPE_ARP);
764 	sa.sa_family = pseudo_AF_HDRCMPLT;
765 	sa.sa_len = sizeof(sa);
766 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
767 	return;
768 }
769 
770 /*
771  * Free an arp entry.
772  */
773 void
774 arptfree(struct llinfo_arp *la)
775 {
776 	struct rtentry *rt = la->la_rt;
777 	struct sockaddr_dl *sdl;
778 	u_int tid = 0;
779 
780 	if (rt == NULL)
781 		panic("arptfree");
782 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
783 	    sdl->sdl_family == AF_LINK) {
784 		sdl->sdl_alen = 0;
785 		la->la_asked = 0;
786 		rt->rt_flags &= ~RTF_REJECT;
787 		return;
788 	}
789 
790 	if (rt->rt_ifp)
791 		tid = rt->rt_ifp->if_rdomain;
792 
793 	rtdeletemsg(rt, tid);
794 }
795 
796 /*
797  * Lookup or enter a new address in arptab.
798  */
799 struct llinfo_arp *
800 arplookup(u_int32_t addr, int create, int proxy, u_int tableid)
801 {
802 	struct rtentry *rt;
803 	struct sockaddr_inarp sin;
804 	int flags;
805 
806 	memset(&sin, 0, sizeof(sin));
807 	sin.sin_len = sizeof(sin);
808 	sin.sin_family = AF_INET;
809 	sin.sin_addr.s_addr = addr;
810 	sin.sin_other = proxy ? SIN_PROXY : 0;
811 	flags = (create) ? (RT_REPORT|RT_RESOLVE) : 0;
812 
813 	rt = rtalloc((struct sockaddr *)&sin, flags, tableid);
814 	if (rt == 0)
815 		return (0);
816 	rt->rt_refcnt--;
817 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
818 	    rt->rt_gateway->sa_family != AF_LINK) {
819 		if (create) {
820 			if (rt->rt_refcnt <= 0 &&
821 			    (rt->rt_flags & RTF_CLONED) != 0) {
822 				struct rt_addrinfo info;
823 
824 				memset(&info, 0, sizeof(info));
825 				info.rti_info[RTAX_DST] = rt_key(rt);
826 				info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
827 				info.rti_info[RTAX_NETMASK] = rt_mask(rt);
828 
829 				rtrequest1(RTM_DELETE, &info, rt->rt_priority,
830 				    NULL, tableid);
831 			}
832 		}
833 		return (0);
834 	}
835 	return ((struct llinfo_arp *)rt->rt_llinfo);
836 }
837 
838 /*
839  * Check whether we do proxy ARP for this address and we point to ourselves.
840  */
841 int
842 arpproxy(struct in_addr in, u_int rdomain)
843 {
844 	struct llinfo_arp *la;
845 	struct ifnet *ifp;
846 	int found = 0;
847 
848 	la = arplookup(in.s_addr, 0, SIN_PROXY, rdomain);
849 	if (la == NULL)
850 		return (0);
851 
852 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
853 		if (ifp->if_rdomain != rdomain)
854 			continue;
855 
856 		if (!memcmp(LLADDR((struct sockaddr_dl *)la->la_rt->rt_gateway),
857 		    LLADDR(ifp->if_sadl),
858 		    ETHER_ADDR_LEN)) {
859 		    	found = 1;
860 		    	break;
861 		}
862 	}
863 
864 	return (found);
865 }
866 
867 void
868 arp_ifinit(struct arpcom *ac, struct ifaddr *ifa)
869 {
870 
871 	/* Warn the user if another station has this IP address. */
872 	arprequest(&ac->ac_if,
873 	    &satosin(ifa->ifa_addr)->sin_addr.s_addr,
874 	    &satosin(ifa->ifa_addr)->sin_addr.s_addr,
875 	    ac->ac_enaddr);
876 	ifa->ifa_rtrequest = arp_rtrequest;
877 }
878 
879 /*
880  * Called from Ethernet interrupt handlers
881  * when ether packet type ETHERTYPE_REVARP
882  * is received.  Common length and type checks are done here,
883  * then the protocol-specific routine is called.
884  */
885 void
886 revarpinput(struct mbuf *m)
887 {
888 	struct arphdr *ar;
889 
890 	if (m->m_len < sizeof(struct arphdr))
891 		goto out;
892 	ar = mtod(m, struct arphdr *);
893 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
894 		goto out;
895 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
896 		goto out;
897 	switch (ntohs(ar->ar_pro)) {
898 
899 	case ETHERTYPE_IP:
900 	case ETHERTYPE_IPTRAILERS:
901 		in_revarpinput(m);
902 		return;
903 
904 	default:
905 		break;
906 	}
907 out:
908 	m_freem(m);
909 }
910 
911 /*
912  * RARP for Internet protocols on Ethernet.
913  * Algorithm is that given in RFC 903.
914  * We are only using for bootstrap purposes to get an ip address for one of
915  * our interfaces.  Thus we support no user-interface.
916  *
917  * Since the contents of the RARP reply are specific to the interface that
918  * sent the request, this code must ensure that they are properly associated.
919  *
920  * Note: also supports ARP via RARP packets, per the RFC.
921  */
922 void
923 in_revarpinput(struct mbuf *m)
924 {
925 #ifdef NFSCLIENT
926 	struct ifnet *ifp;
927 #endif /* NFSCLIENT */
928 	struct ether_arp *ar;
929 	int op;
930 
931 	ar = mtod(m, struct ether_arp *);
932 	op = ntohs(ar->arp_op);
933 	switch (op) {
934 	case ARPOP_REQUEST:
935 	case ARPOP_REPLY:	/* per RFC */
936 		in_arpinput(m);
937 		return;
938 	case ARPOP_REVREPLY:
939 		break;
940 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
941 	default:
942 		goto out;
943 	}
944 #ifdef NFSCLIENT
945 	if (!revarp_in_progress)
946 		goto out;
947 	ifp = m->m_pkthdr.rcvif;
948 	if (ifp != revarp_ifp) /* !same interface */
949 		goto out;
950 	if (revarp_finished)
951 		goto wake;
952 	if (memcmp(ar->arp_tha, ((struct arpcom *)ifp)->ac_enaddr,
953 	    sizeof(ar->arp_tha)))
954 		goto out;
955 	memcpy(&revarp_srvip, ar->arp_spa, sizeof(revarp_srvip));
956 	memcpy(&revarp_myip, ar->arp_tpa, sizeof(revarp_myip));
957 	revarp_finished = 1;
958 wake:	/* Do wakeup every time in case it was missed. */
959 	wakeup((caddr_t)&revarp_myip);
960 #endif
961 
962 out:
963 	m_freem(m);
964 }
965 
966 /*
967  * Send a RARP request for the ip address of the specified interface.
968  * The request should be RFC 903-compliant.
969  */
970 void
971 revarprequest(struct ifnet *ifp)
972 {
973 	struct sockaddr sa;
974 	struct mbuf *m;
975 	struct ether_header *eh;
976 	struct ether_arp *ea;
977 	struct arpcom *ac = (struct arpcom *)ifp;
978 
979 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
980 		return;
981 	m->m_len = sizeof(*ea);
982 	m->m_pkthdr.len = sizeof(*ea);
983 	MH_ALIGN(m, sizeof(*ea));
984 	ea = mtod(m, struct ether_arp *);
985 	eh = (struct ether_header *)sa.sa_data;
986 	memset(ea, 0, sizeof(*ea));
987 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
988 	eh->ether_type = htons(ETHERTYPE_REVARP);
989 	ea->arp_hrd = htons(ARPHRD_ETHER);
990 	ea->arp_pro = htons(ETHERTYPE_IP);
991 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
992 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
993 	ea->arp_op = htons(ARPOP_REVREQUEST);
994 	memcpy(eh->ether_shost, ac->ac_enaddr, sizeof(ea->arp_tha));
995 	memcpy(ea->arp_sha, ac->ac_enaddr, sizeof(ea->arp_sha));
996 	memcpy(ea->arp_tha, ac->ac_enaddr, sizeof(ea->arp_tha));
997 	sa.sa_family = pseudo_AF_HDRCMPLT;
998 	sa.sa_len = sizeof(sa);
999 	m->m_flags |= M_BCAST;
1000 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
1001 }
1002 
1003 #ifdef NFSCLIENT
1004 /*
1005  * RARP for the ip address of the specified interface, but also
1006  * save the ip address of the server that sent the answer.
1007  * Timeout if no response is received.
1008  */
1009 int
1010 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1011     struct in_addr *clnt_in)
1012 {
1013 	int result, count = 20;
1014 
1015 	if (revarp_finished)
1016 		return EIO;
1017 
1018 	revarp_ifp = ifp;
1019 	revarp_in_progress = 1;
1020 	while (count--) {
1021 		revarprequest(ifp);
1022 		result = tsleep((caddr_t)&revarp_myip, PSOCK, "revarp", hz/2);
1023 		if (result != EWOULDBLOCK)
1024 			break;
1025 	}
1026 	revarp_in_progress = 0;
1027 	if (!revarp_finished)
1028 		return ENETUNREACH;
1029 
1030 	memcpy(serv_in, &revarp_srvip, sizeof(*serv_in));
1031 	memcpy(clnt_in, &revarp_myip, sizeof(*clnt_in));
1032 	return 0;
1033 }
1034 
1035 /* For compatibility: only saves interface address. */
1036 int
1037 revarpwhoami(struct in_addr *in, struct ifnet *ifp)
1038 {
1039 	struct in_addr server;
1040 	return (revarpwhoarewe(ifp, &server, in));
1041 }
1042 #endif /* NFSCLIENT */
1043 
1044 #ifdef DDB
1045 
1046 #include <machine/db_machdep.h>
1047 #include <ddb/db_interface.h>
1048 #include <ddb/db_output.h>
1049 
1050 void
1051 db_print_sa(struct sockaddr *sa)
1052 {
1053 	int len;
1054 	u_char *p;
1055 
1056 	if (sa == 0) {
1057 		db_printf("[NULL]");
1058 		return;
1059 	}
1060 
1061 	p = (u_char *)sa;
1062 	len = sa->sa_len;
1063 	db_printf("[");
1064 	while (len > 0) {
1065 		db_printf("%d", *p);
1066 		p++;
1067 		len--;
1068 		if (len)
1069 			db_printf(",");
1070 	}
1071 	db_printf("]\n");
1072 }
1073 
1074 void
1075 db_print_ifa(struct ifaddr *ifa)
1076 {
1077 	if (ifa == 0)
1078 		return;
1079 	db_printf("  ifa_addr=");
1080 	db_print_sa(ifa->ifa_addr);
1081 	db_printf("  ifa_dsta=");
1082 	db_print_sa(ifa->ifa_dstaddr);
1083 	db_printf("  ifa_mask=");
1084 	db_print_sa(ifa->ifa_netmask);
1085 	db_printf("  flags=0x%x, refcnt=%d, metric=%d\n",
1086 	    ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric);
1087 }
1088 
1089 void
1090 db_print_llinfo(caddr_t li)
1091 {
1092 	struct llinfo_arp *la;
1093 
1094 	if (li == 0)
1095 		return;
1096 	la = (struct llinfo_arp *)li;
1097 	db_printf("  la_rt=%p la_hold_head=%p, la_asked=0x%lx\n",
1098 	    la->la_rt, la->la_hold_head, la->la_asked);
1099 }
1100 
1101 /*
1102  * Function to pass to rn_walktree().
1103  * Return non-zero error to abort walk.
1104  */
1105 int
1106 db_show_radix_node(struct radix_node *rn, void *w, u_int id)
1107 {
1108 	struct rtentry *rt = (struct rtentry *)rn;
1109 
1110 	db_printf("rtentry=%p", rt);
1111 
1112 	db_printf(" flags=0x%x refcnt=%d use=%llu expire=%lld rtableid=%u\n",
1113 	    rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id);
1114 
1115 	db_printf(" key="); db_print_sa(rt_key(rt));
1116 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1117 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1118 
1119 	db_printf(" ifp=%p ", rt->rt_ifp);
1120 	if (rt->rt_ifp)
1121 		db_printf("(%s)", rt->rt_ifp->if_xname);
1122 	else
1123 		db_printf("(NULL)");
1124 
1125 	db_printf(" ifa=%p\n", rt->rt_ifa);
1126 	db_print_ifa(rt->rt_ifa);
1127 
1128 	db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo);
1129 	db_print_llinfo(rt->rt_llinfo);
1130 	return (0);
1131 }
1132 
1133 /*
1134  * Function to print all the route trees.
1135  * Use this from ddb:  "call db_show_arptab"
1136  */
1137 int
1138 db_show_arptab(void)
1139 {
1140 	struct radix_node_head *rnh;
1141 	rnh = rtable_get(0, AF_INET);
1142 	db_printf("Route tree for AF_INET\n");
1143 	if (rnh == NULL) {
1144 		db_printf(" (not initialized)\n");
1145 		return (0);
1146 	}
1147 	rn_walktree(rnh, db_show_radix_node, NULL);
1148 	return (0);
1149 }
1150 #endif
1151 #endif /* INET */
1152