xref: /openbsd-src/sys/netinet/if_ether.c (revision 91f110e064cd7c194e59e019b83bb7496c1c84d4)
1 /*	$OpenBSD: if_ether.c,v 1.122 2014/03/18 14:55:49 mikeb 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 
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/route.h>
57 #include <net/if_types.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_var.h>
61 #include <netinet/if_ether.h>
62 #if NCARP > 0
63 #include <netinet/ip_carp.h>
64 #endif
65 #if NBRIDGE > 0
66 #include <net/if_bridge.h>
67 #endif
68 
69 #define SDL(s) ((struct sockaddr_dl *)s)
70 #define SRP(s) ((struct sockaddr_inarp *)s)
71 
72 /*
73  * ARP trailer negotiation.  Trailer protocol is not IP specific,
74  * but ARP request/response use IP addresses.
75  */
76 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
77 
78 /* timer values */
79 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
80 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
81 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
82 #define	rt_expire rt_rmx.rmx_expire
83 
84 void arptfree(struct llinfo_arp *);
85 void arptimer(void *);
86 struct llinfo_arp *arplookup(u_int32_t, int, int, u_int);
87 void in_arpinput(struct mbuf *);
88 
89 LIST_HEAD(, llinfo_arp) llinfo_arp;
90 struct	ifqueue arpintrq;
91 int	arp_inuse, arp_allocated;
92 int	arp_maxtries = 5;
93 int	useloopback = 1;	/* use loopback interface for local traffic */
94 int	arpinit_done;
95 int	la_hold_total;
96 
97 #ifdef NFSCLIENT
98 /* revarp state */
99 struct in_addr revarp_myip, revarp_srvip;
100 int revarp_finished;
101 int revarp_in_progress;
102 struct ifnet *revarp_ifp;
103 #endif /* NFSCLIENT */
104 
105 #ifdef DDB
106 
107 void	db_print_sa(struct sockaddr *);
108 void	db_print_ifa(struct ifaddr *);
109 void	db_print_llinfo(caddr_t);
110 int	db_show_radix_node(struct radix_node *, void *, u_int);
111 #endif
112 
113 static const struct sockaddr_dl null_sdl = { sizeof(null_sdl), AF_LINK };
114 
115 /*
116  * Timeout routine.  Age arp_tab entries periodically.
117  */
118 /* ARGSUSED */
119 void
120 arptimer(void *arg)
121 {
122 	struct timeout *to = (struct timeout *)arg;
123 	int s;
124 	struct llinfo_arp *la, *nla;
125 
126 	s = splsoftnet();
127 	timeout_add_sec(to, arpt_prune);
128 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
129 		struct rtentry *rt = la->la_rt;
130 
131 		nla = LIST_NEXT(la, la_list);
132 		if (rt->rt_expire && rt->rt_expire <= time_second)
133 			arptfree(la); /* timer has expired; clear */
134 	}
135 	splx(s);
136 }
137 
138 void
139 arp_rtrequest(int req, struct rtentry *rt)
140 {
141 	struct sockaddr *gate = rt->rt_gateway;
142 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
143 	struct ifnet *ifp = rt->rt_ifp;
144 	struct ifaddr *ifa;
145 	struct mbuf *m;
146 
147 	if (!arpinit_done) {
148 		static struct timeout arptimer_to;
149 
150 		arpinit_done = 1;
151 		IFQ_SET_MAXLEN(&arpintrq, 50);	/* XXX hate magic numbers */
152 		/*
153 		 * We generate expiration times from time.tv_sec
154 		 * so avoid accidently creating permanent routes.
155 		 */
156 		if (time_second == 0) {
157 			time_second++;
158 		}
159 
160 		timeout_set(&arptimer_to, arptimer, &arptimer_to);
161 		timeout_add_sec(&arptimer_to, 1);
162 	}
163 
164 	if (rt->rt_flags & RTF_GATEWAY)
165 		return;
166 
167 	switch (req) {
168 
169 	case RTM_ADD:
170 		/*
171 		 * XXX: If this is a manually added route to interface
172 		 * such as older version of routed or gated might provide,
173 		 * restore cloning bit.
174 		 */
175 		if ((rt->rt_flags & RTF_HOST) == 0 && rt_mask(rt) &&
176 		    satosin(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
177 			rt->rt_flags |= RTF_CLONING;
178 		if (rt->rt_flags & RTF_CLONING) {
179 			/*
180 			 * Case 1: This route should come from a route to iface.
181 			 */
182 			rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl,
183 			    ifp->if_rdomain);
184 			gate = rt->rt_gateway;
185 			SDL(gate)->sdl_type = ifp->if_type;
186 			SDL(gate)->sdl_index = ifp->if_index;
187 			/*
188 			 * Give this route an expiration time, even though
189 			 * it's a "permanent" route, so that routes cloned
190 			 * from it do not need their expiration time set.
191 			 */
192 			rt->rt_expire = time_second;
193 			break;
194 		}
195 		/* Announce a new entry if requested. */
196 		if (rt->rt_flags & RTF_ANNOUNCE)
197 			arprequest(ifp,
198 			    &satosin(rt_key(rt))->sin_addr.s_addr,
199 			    &satosin(rt_key(rt))->sin_addr.s_addr,
200 			    (u_char *)LLADDR(SDL(gate)));
201 		/*FALLTHROUGH*/
202 	case RTM_RESOLVE:
203 		if (gate->sa_family != AF_LINK ||
204 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
205 			log(LOG_DEBUG, "%s: bad gateway value: %s\n", __func__,
206 			    ifp->if_xname);
207 			break;
208 		}
209 		SDL(gate)->sdl_type = ifp->if_type;
210 		SDL(gate)->sdl_index = ifp->if_index;
211 		if (la != 0)
212 			break; /* This happens on a route change */
213 		/*
214 		 * Case 2:  This route may come from cloning, or a manual route
215 		 * add with a LL address.
216 		 */
217 		la = malloc(sizeof(*la), M_RTABLE, M_NOWAIT | M_ZERO);
218 		rt->rt_llinfo = (caddr_t)la;
219 		if (la == NULL) {
220 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
221 			break;
222 		}
223 		arp_inuse++;
224 		arp_allocated++;
225 		la->la_rt = rt;
226 		rt->rt_flags |= RTF_LLINFO;
227 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
228 
229 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
230 			if ((ifa->ifa_addr->sa_family == AF_INET) &&
231 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr ==
232 			    satosin(rt_key(rt))->sin_addr.s_addr)
233 				break;
234 		}
235 		if (ifa) {
236 			/*
237 			 * This test used to be
238 			 *	if (lo0ifp->if_flags & IFF_UP)
239 			 * It allowed local traffic to be forced through
240 			 * the hardware by configuring the loopback down.
241 			 * However, it causes problems during network
242 			 * configuration for boards that can't receive
243 			 * packets they send.  It is now necessary to clear
244 			 * "useloopback" and remove the route to force
245 			 * traffic out to the hardware.
246 			 *
247 			 * In 4.4BSD, the above "if" statement checked
248 			 * rt->rt_ifa against rt_key(rt).  It was changed
249 			 * to the current form so that we can provide a
250 			 * better support for multiple IPv4 addresses on a
251 			 * interface.
252 			 */
253 			rt->rt_expire = 0;
254 			SDL(gate)->sdl_alen = ETHER_ADDR_LEN;
255 			memcpy(LLADDR(SDL(gate)),
256 			    ((struct arpcom *)ifp)->ac_enaddr, ETHER_ADDR_LEN);
257 			if (useloopback)
258 				rt->rt_ifp = lo0ifp;
259 			/*
260 			 * make sure to set rt->rt_ifa to the interface
261 			 * address we are using, otherwise we will have trouble
262 			 * with source address selection.
263 			 */
264 			if (ifa != rt->rt_ifa) {
265 				ifafree(rt->rt_ifa);
266 				ifa->ifa_refcnt++;
267 				rt->rt_ifa = ifa;
268 			}
269 		}
270 		break;
271 
272 	case RTM_DELETE:
273 		if (la == 0)
274 			break;
275 		arp_inuse--;
276 		LIST_REMOVE(la, la_list);
277 		rt->rt_llinfo = 0;
278 		rt->rt_flags &= ~RTF_LLINFO;
279 		while ((m = la->la_hold_head) != NULL) {
280 			la->la_hold_head = la->la_hold_head->m_nextpkt;
281 			la_hold_total--;
282 			m_freem(m);
283 		}
284 		free(la, M_RTABLE);
285 	}
286 }
287 
288 /*
289  * Broadcast an ARP request. Caller specifies:
290  *	- arp header source ip address
291  *	- arp header target ip address
292  *	- arp header source ethernet address
293  */
294 void
295 arprequest(struct ifnet *ifp, u_int32_t *sip, u_int32_t *tip, u_int8_t *enaddr)
296 {
297 	struct mbuf *m;
298 	struct ether_header *eh;
299 	struct ether_arp *ea;
300 	struct sockaddr sa;
301 
302 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
303 		return;
304 	m->m_len = sizeof(*ea);
305 	m->m_pkthdr.len = sizeof(*ea);
306 	m->m_pkthdr.rdomain = ifp->if_rdomain;
307 	MH_ALIGN(m, sizeof(*ea));
308 	ea = mtod(m, struct ether_arp *);
309 	eh = (struct ether_header *)sa.sa_data;
310 	memset(ea, 0, sizeof(*ea));
311 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
312 	eh->ether_type = htons(ETHERTYPE_ARP);	/* if_output will not swap */
313 	ea->arp_hrd = htons(ARPHRD_ETHER);
314 	ea->arp_pro = htons(ETHERTYPE_IP);
315 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
316 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
317 	ea->arp_op = htons(ARPOP_REQUEST);
318 	memcpy(eh->ether_shost, enaddr, sizeof(eh->ether_shost));
319 	memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
320 	memcpy(ea->arp_spa, sip, sizeof(ea->arp_spa));
321 	memcpy(ea->arp_tpa, tip, sizeof(ea->arp_tpa));
322 	sa.sa_family = pseudo_AF_HDRCMPLT;
323 	sa.sa_len = sizeof(sa);
324 	m->m_flags |= M_BCAST;
325 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
326 }
327 
328 /*
329  * Resolve an IP address into an ethernet address.  If success,
330  * desten is filled in.  If there is no entry in arptab,
331  * set one up and broadcast a request for the IP address.
332  * Hold onto this mbuf and resend it once the address
333  * is finally resolved.  A return value of 1 indicates
334  * that desten has been filled in and the packet should be sent
335  * normally; a 0 return indicates that the packet has been
336  * taken over here, either now or for later transmission.
337  */
338 int
339 arpresolve(struct arpcom *ac, struct rtentry *rt, struct mbuf *m,
340     struct sockaddr *dst, u_char *desten)
341 {
342 	struct llinfo_arp *la;
343 	struct sockaddr_dl *sdl;
344 	struct mbuf *mh;
345 	char addr[INET_ADDRSTRLEN];
346 
347 	if (m->m_flags & M_BCAST) {	/* broadcast */
348 		memcpy(desten, etherbroadcastaddr, sizeof(etherbroadcastaddr));
349 		return (1);
350 	}
351 	if (m->m_flags & M_MCAST) {	/* multicast */
352 		ETHER_MAP_IP_MULTICAST(&satosin(dst)->sin_addr, desten);
353 		return (1);
354 	}
355 	if (rt) {
356 		la = (struct llinfo_arp *)rt->rt_llinfo;
357 		if (la == NULL)
358 			log(LOG_DEBUG, "arpresolve: %s: route without link "
359 			    "local address\n", inet_ntop(AF_INET,
360 				&satosin(dst)->sin_addr, addr, sizeof(addr)));
361 	} else {
362 		if ((la = arplookup(satosin(dst)->sin_addr.s_addr, RT_REPORT, 0,
363 		    ac->ac_if.if_rdomain)) != NULL)
364 			rt = la->la_rt;
365 		else
366 			log(LOG_DEBUG,
367 			    "arpresolve: %s: can't allocate llinfo\n",
368 			    inet_ntop(AF_INET, &satosin(dst)->sin_addr,
369 				addr, sizeof(addr)));
370 	}
371 	if (la == 0 || rt == 0) {
372 		m_freem(m);
373 		return (0);
374 	}
375 	sdl = SDL(rt->rt_gateway);
376 	/*
377 	 * Check the address family and length is valid, the address
378 	 * is resolved; otherwise, try to resolve.
379 	 */
380 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
381 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
382 		memcpy(desten, LLADDR(sdl), sdl->sdl_alen);
383 		return 1;
384 	}
385 	if (((struct ifnet *)ac)->if_flags & IFF_NOARP) {
386 		m_freem(m);
387 		return 0;
388 	}
389 
390 	/*
391 	 * There is an arptab entry, but no ethernet address
392 	 * response yet. Insert mbuf in hold queue if below limit
393 	 * if above the limit free the queue without queuing the new packet.
394 	 */
395 	if (la_hold_total < MAX_HOLD_TOTAL && la_hold_total < nmbclust / 64) {
396 		if (la->la_hold_count >= MAX_HOLD_QUEUE) {
397 			mh = la->la_hold_head;
398 			la->la_hold_head = la->la_hold_head->m_nextpkt;
399 			if (mh == la->la_hold_tail)
400 				la->la_hold_tail = NULL;
401 			la->la_hold_count--;
402 			la_hold_total--;
403 			m_freem(mh);
404 		}
405 		if (la->la_hold_tail == NULL)
406 			la->la_hold_head = m;
407 		else
408 			la->la_hold_tail->m_nextpkt = m;
409 		la->la_hold_tail = m;
410 		la->la_hold_count++;
411 		la_hold_total++;
412 	} else {
413 		while ((mh = la->la_hold_head) != NULL) {
414 			la->la_hold_head =
415 			    la->la_hold_head->m_nextpkt;
416 			la_hold_total--;
417 			m_freem(mh);
418 		}
419 		la->la_hold_tail = NULL;
420 		la->la_hold_count = 0;
421 		m_freem(m);
422 	}
423 
424 	/*
425 	 * Re-send the ARP request when appropriate.
426 	 */
427 #ifdef	DIAGNOSTIC
428 	if (rt->rt_expire == 0) {
429 		/* This should never happen. (Should it? -gwr) */
430 		printf("arpresolve: unresolved and rt_expire == 0\n");
431 		/* Set expiration time to now (expired). */
432 		rt->rt_expire = time_second;
433 	}
434 #endif
435 	if (rt->rt_expire) {
436 		rt->rt_flags &= ~RTF_REJECT;
437 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
438 			rt->rt_expire = time_second;
439 			if (la->la_asked++ < arp_maxtries)
440 				arprequest(&ac->ac_if,
441 				    &satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr,
442 				    &satosin(dst)->sin_addr.s_addr,
443 #if NCARP > 0
444 				    (rt->rt_ifp->if_type == IFT_CARP) ?
445 					((struct arpcom *) rt->rt_ifp->if_softc
446 					)->ac_enaddr :
447 #endif
448 				    ac->ac_enaddr);
449 			else {
450 				rt->rt_flags |= RTF_REJECT;
451 				rt->rt_expire += arpt_down;
452 				la->la_asked = 0;
453 				while ((mh = la->la_hold_head) != NULL) {
454 					la->la_hold_head =
455 					    la->la_hold_head->m_nextpkt;
456 					la_hold_total--;
457 					m_freem(mh);
458 				}
459 				la->la_hold_tail = NULL;
460 				la->la_hold_count = 0;
461 			}
462 		}
463 	}
464 	return (0);
465 }
466 
467 /*
468  * Common length and type checks are done here,
469  * then the protocol-specific routine is called.
470  */
471 void
472 arpintr(void)
473 {
474 	struct mbuf *m;
475 	struct arphdr *ar;
476 	int s, len;
477 
478 	for (;;) {
479 		s = splnet();
480 		IF_DEQUEUE(&arpintrq, m);
481 		splx(s);
482 		if (m == NULL)
483 			break;
484 #ifdef DIAGNOSTIC
485 		if ((m->m_flags & M_PKTHDR) == 0)
486 			panic("arpintr");
487 #endif
488 
489 		len = sizeof(struct arphdr);
490 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
491 			continue;
492 
493 		ar = mtod(m, struct arphdr *);
494 		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) {
495 			m_freem(m);
496 			continue;
497 		}
498 
499 		len += 2 * (ar->ar_hln + ar->ar_pln);
500 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
501 			continue;
502 
503 		switch (ntohs(ar->ar_pro)) {
504 		case ETHERTYPE_IP:
505 		case ETHERTYPE_IPTRAILERS:
506 			in_arpinput(m);
507 			continue;
508 		}
509 		m_freem(m);
510 	}
511 }
512 
513 /*
514  * ARP for Internet protocols on Ethernet.
515  * Algorithm is that given in RFC 826.
516  * In addition, a sanity check is performed on the sender
517  * protocol address, to catch impersonators.
518  * We no longer handle negotiations for use of trailer protocol:
519  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
520  * along with IP replies if we wanted trailers sent to us,
521  * and also sent them in response to IP replies.
522  * This allowed either end to announce the desire to receive
523  * trailer packets.
524  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
525  * but formerly didn't normally send requests.
526  */
527 void
528 in_arpinput(struct mbuf *m)
529 {
530 	struct ether_arp *ea;
531 	struct ifnet *ifp = m->m_pkthdr.rcvif;
532 	struct arpcom *ac = (struct arpcom *)ifp;
533 	struct ether_header *eh;
534 	struct llinfo_arp *la = 0;
535 	struct rtentry *rt;
536 	struct ifaddr *ifa;
537 	struct sockaddr_dl *sdl;
538 	struct sockaddr sa;
539 	struct in_addr isaddr, itaddr, myaddr;
540 	struct mbuf *mh, *mt;
541 	u_int8_t *enaddr = NULL;
542 #if NCARP > 0
543 	u_int8_t *ether_shost = NULL;
544 #endif
545 	char addr[INET_ADDRSTRLEN];
546 	int op;
547 
548 	ea = mtod(m, struct ether_arp *);
549 	op = ntohs(ea->arp_op);
550 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
551 		goto out;
552 #if notyet
553 	if ((op == ARPOP_REPLY) && (m->m_flags & (M_BCAST|M_MCAST))) {
554 		log(LOG_ERR,
555 		    "arp: received reply to broadcast or multicast address\n");
556 		goto out;
557 	}
558 #endif
559 
560 	memcpy(&itaddr, ea->arp_tpa, sizeof(itaddr));
561 	memcpy(&isaddr, ea->arp_spa, sizeof(isaddr));
562 
563 	/* First try: check target against our addresses */
564 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
565 		if (ifa->ifa_addr->sa_family != AF_INET)
566 			continue;
567 
568 		if (itaddr.s_addr != ifatoia(ifa)->ia_addr.sin_addr.s_addr)
569 			continue;
570 
571 #if NCARP > 0
572 		if (ifp->if_type == IFT_CARP &&
573 		    ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
574 		    (IFF_UP|IFF_RUNNING))) {
575 			if (op == ARPOP_REPLY)
576 				break;
577 			if (carp_iamatch(ifatoia(ifa), ea->arp_sha,
578 			    &enaddr, &ether_shost))
579 				break;
580 			else
581 				goto out;
582 		}
583 #endif
584 		break;
585 	}
586 
587 	/* Second try: check source against our addresses */
588 	if (ifa == NULL) {
589 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
590 			if (ifa->ifa_addr->sa_family != AF_INET)
591 				continue;
592 
593 			if (isaddr.s_addr ==
594 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr)
595 				break;
596 		}
597 	}
598 
599 	/* Third try: not one of our addresses, just find an usable ia */
600 	if (ifa == NULL) {
601 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
602 			if (ifa->ifa_addr->sa_family == AF_INET)
603 				break;
604 		}
605 	}
606 
607 	if (ifa == NULL)
608 		goto out;
609 
610 	if (!enaddr)
611 		enaddr = ac->ac_enaddr;
612 	myaddr = ifatoia(ifa)->ia_addr.sin_addr;
613 
614 	if (!memcmp(ea->arp_sha, enaddr, sizeof(ea->arp_sha)))
615 		goto out;	/* it's from me, ignore it. */
616 	if (ETHER_IS_MULTICAST(&ea->arp_sha[0]))
617 		if (!memcmp(ea->arp_sha, etherbroadcastaddr,
618 		    sizeof (ea->arp_sha))) {
619 			inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
620 			log(LOG_ERR, "arp: ether address is broadcast for "
621 			    "IP address %s!\n", addr);
622 			goto out;
623 		}
624 	if (myaddr.s_addr && isaddr.s_addr == myaddr.s_addr) {
625 		inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
626 		log(LOG_ERR,
627 		   "duplicate IP address %s sent from ethernet address %s\n",
628 		   addr, ether_sprintf(ea->arp_sha));
629 		itaddr = myaddr;
630 		goto reply;
631 	}
632 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0,
633 	    rtable_l2(m->m_pkthdr.rdomain));
634 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
635 		if (sdl->sdl_alen) {
636 		    if (memcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
637 			if (rt->rt_flags & RTF_PERMANENT_ARP) {
638 				inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
639 				log(LOG_WARNING,
640 				   "arp: attempt to overwrite permanent "
641 				   "entry for %s by %s on %s\n", addr,
642 				   ether_sprintf(ea->arp_sha),
643 				   ac->ac_if.if_xname);
644 				goto out;
645 			} else if (rt->rt_ifp != &ac->ac_if) {
646 #if NCARP > 0
647 				if (ac->ac_if.if_type != IFT_CARP)
648 #endif
649 				{
650 					inet_ntop(AF_INET, &isaddr,
651 					    addr, sizeof(addr));
652 					log(LOG_WARNING,
653 					   "arp: attempt to overwrite entry for"
654 					   " %s on %s by %s on %s\n", addr,
655 					   rt->rt_ifp->if_xname,
656 					   ether_sprintf(ea->arp_sha),
657 					   ac->ac_if.if_xname);
658 				}
659 				goto out;
660 			} else {
661 				inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
662 				log(LOG_INFO,
663 				   "arp info overwritten for %s by %s on %s\n",
664 				   addr,
665 				   ether_sprintf(ea->arp_sha),
666 				   ac->ac_if.if_xname);
667 				rt->rt_expire = 1; /* no longer static */
668 			}
669 		    }
670 		} else if (rt->rt_ifp != &ac->ac_if &&
671 #if NBRIDGE > 0
672 		    !SAME_BRIDGE(ac->ac_if.if_bridgeport,
673 		    rt->rt_ifp->if_bridgeport) &&
674 #endif
675 #if NCARP > 0
676 		    !(rt->rt_ifp->if_type == IFT_CARP &&
677 		    rt->rt_ifp->if_carpdev == &ac->ac_if) &&
678 		    !(ac->ac_if.if_type == IFT_CARP &&
679 		    ac->ac_if.if_carpdev == rt->rt_ifp) &&
680 #endif
681 		    1) {
682 			inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
683 			log(LOG_WARNING,
684 			    "arp: attempt to add entry for %s "
685 			    "on %s by %s on %s\n", addr,
686 			    rt->rt_ifp->if_xname,
687 			    ether_sprintf(ea->arp_sha),
688 			    ac->ac_if.if_xname);
689 			goto out;
690 		}
691  		sdl->sdl_alen = sizeof(ea->arp_sha);
692 		memcpy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
693 		if (rt->rt_expire)
694 			rt->rt_expire = time_second + arpt_keep;
695 		rt->rt_flags &= ~RTF_REJECT;
696 		la->la_asked = 0;
697 		rt_sendmsg(rt, RTM_RESOLVE, rt->rt_ifp->if_rdomain);
698 		while ((mh = la->la_hold_head) != NULL) {
699 			if ((la->la_hold_head = mh->m_nextpkt) == NULL)
700 				la->la_hold_tail = NULL;
701 			la->la_hold_count--;
702 			la_hold_total--;
703 			mt = la->la_hold_tail;
704 
705 			(*ac->ac_if.if_output)(&ac->ac_if, mh, rt_key(rt), rt);
706 
707 			if (la->la_hold_tail == mh) {
708 				/* mbuf is back in queue. Discard. */
709 				la->la_hold_tail = mt;
710 				if (la->la_hold_tail)
711 					la->la_hold_tail->m_nextpkt = NULL;
712 				else
713 					la->la_hold_head = NULL;
714 				la->la_hold_count--;
715 				la_hold_total--;
716 				m_freem(mh);
717 			}
718 		}
719 	}
720 reply:
721 	if (op != ARPOP_REQUEST) {
722 out:
723 		m_freem(m);
724 		return;
725 	}
726 	if (itaddr.s_addr == myaddr.s_addr) {
727 		/* I am the target */
728 		memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
729 		memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
730 	} else {
731 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY,
732 		    rtable_l2(m->m_pkthdr.rdomain));
733 		if (la == 0)
734 			goto out;
735 		rt = la->la_rt;
736 		if (rt->rt_ifp->if_type == IFT_CARP && ifp->if_type != IFT_CARP)
737 			goto out;
738 		memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
739 		sdl = SDL(rt->rt_gateway);
740 		memcpy(ea->arp_sha, LLADDR(sdl), sizeof(ea->arp_sha));
741 	}
742 
743 	memcpy(ea->arp_tpa, ea->arp_spa, sizeof(ea->arp_spa));
744 	memcpy(ea->arp_spa, &itaddr, sizeof(ea->arp_spa));
745 	ea->arp_op = htons(ARPOP_REPLY);
746 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
747 	eh = (struct ether_header *)sa.sa_data;
748 	memcpy(eh->ether_dhost, ea->arp_tha, sizeof(eh->ether_dhost));
749 #if NCARP > 0
750 	if (ether_shost)
751 		enaddr = ether_shost;
752 #endif
753 	memcpy(eh->ether_shost, enaddr, sizeof(eh->ether_shost));
754 
755 	eh->ether_type = htons(ETHERTYPE_ARP);
756 	sa.sa_family = pseudo_AF_HDRCMPLT;
757 	sa.sa_len = sizeof(sa);
758 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
759 	return;
760 }
761 
762 /*
763  * Free an arp entry.
764  */
765 void
766 arptfree(struct llinfo_arp *la)
767 {
768 	struct rtentry *rt = la->la_rt;
769 	struct sockaddr_dl *sdl;
770 	struct rt_addrinfo info;
771 	u_int tid = 0;
772 
773 	if (rt == NULL)
774 		panic("arptfree");
775 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
776 	    sdl->sdl_family == AF_LINK) {
777 		sdl->sdl_alen = 0;
778 		la->la_asked = 0;
779 		rt->rt_flags &= ~RTF_REJECT;
780 		return;
781 	}
782 	memset(&info, 0, sizeof(info));
783 	info.rti_info[RTAX_DST] = rt_key(rt);
784 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
785 
786 	if (rt->rt_ifp)
787 		tid = rt->rt_ifp->if_rdomain;
788 
789 	rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL, tid);
790 }
791 
792 /*
793  * Lookup or enter a new address in arptab.
794  */
795 struct llinfo_arp *
796 arplookup(u_int32_t addr, int create, int proxy, u_int tableid)
797 {
798 	struct rtentry *rt;
799 	struct sockaddr_inarp sin;
800 
801 	memset(&sin, 0, sizeof(sin));
802 	sin.sin_len = sizeof(sin);
803 	sin.sin_family = AF_INET;
804 	sin.sin_addr.s_addr = addr;
805 	sin.sin_other = proxy ? SIN_PROXY : 0;
806 	rt = rtalloc1((struct sockaddr *)&sin, create, tableid);
807 	if (rt == 0)
808 		return (0);
809 	rt->rt_refcnt--;
810 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
811 	    rt->rt_gateway->sa_family != AF_LINK) {
812 		if (create) {
813 			if (rt->rt_refcnt <= 0 &&
814 			    (rt->rt_flags & RTF_CLONED) != 0) {
815 				struct rt_addrinfo info;
816 
817 				memset(&info, 0, sizeof(info));
818 				info.rti_info[RTAX_DST] = rt_key(rt);
819 				info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
820 				info.rti_info[RTAX_NETMASK] = rt_mask(rt);
821 
822 				rtrequest1(RTM_DELETE, &info, rt->rt_priority,
823 				    NULL, tableid);
824 			}
825 		}
826 		return (0);
827 	}
828 	return ((struct llinfo_arp *)rt->rt_llinfo);
829 }
830 
831 /*
832  * Check whether we do proxy ARP for this address and we point to ourselves.
833  */
834 int
835 arpproxy(struct in_addr in, u_int rdomain)
836 {
837 	struct llinfo_arp *la;
838 	struct ifnet *ifp;
839 	int found = 0;
840 
841 	la = arplookup(in.s_addr, 0, SIN_PROXY, rdomain);
842 	if (la == NULL)
843 		return (0);
844 
845 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
846 		if (ifp->if_rdomain != rdomain)
847 			continue;
848 
849 		if (!memcmp(LLADDR((struct sockaddr_dl *)la->la_rt->rt_gateway),
850 		    LLADDR(ifp->if_sadl),
851 		    ETHER_ADDR_LEN)) {
852 		    	found = 1;
853 		    	break;
854 		}
855 	}
856 
857 	return (found);
858 }
859 
860 void
861 arp_ifinit(struct arpcom *ac, struct ifaddr *ifa)
862 {
863 
864 	/* Warn the user if another station has this IP address. */
865 	arprequest(&ac->ac_if,
866 	    &satosin(ifa->ifa_addr)->sin_addr.s_addr,
867 	    &satosin(ifa->ifa_addr)->sin_addr.s_addr,
868 	    ac->ac_enaddr);
869 	ifa->ifa_rtrequest = arp_rtrequest;
870 	ifa->ifa_flags |= RTF_CLONING;
871 }
872 
873 /*
874  * Called from Ethernet interrupt handlers
875  * when ether packet type ETHERTYPE_REVARP
876  * is received.  Common length and type checks are done here,
877  * then the protocol-specific routine is called.
878  */
879 void
880 revarpinput(struct mbuf *m)
881 {
882 	struct arphdr *ar;
883 
884 	if (m->m_len < sizeof(struct arphdr))
885 		goto out;
886 	ar = mtod(m, struct arphdr *);
887 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
888 		goto out;
889 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
890 		goto out;
891 	switch (ntohs(ar->ar_pro)) {
892 
893 	case ETHERTYPE_IP:
894 	case ETHERTYPE_IPTRAILERS:
895 		in_revarpinput(m);
896 		return;
897 
898 	default:
899 		break;
900 	}
901 out:
902 	m_freem(m);
903 }
904 
905 /*
906  * RARP for Internet protocols on Ethernet.
907  * Algorithm is that given in RFC 903.
908  * We are only using for bootstrap purposes to get an ip address for one of
909  * our interfaces.  Thus we support no user-interface.
910  *
911  * Since the contents of the RARP reply are specific to the interface that
912  * sent the request, this code must ensure that they are properly associated.
913  *
914  * Note: also supports ARP via RARP packets, per the RFC.
915  */
916 void
917 in_revarpinput(struct mbuf *m)
918 {
919 #ifdef NFSCLIENT
920 	struct ifnet *ifp;
921 #endif /* NFSCLIENT */
922 	struct ether_arp *ar;
923 	int op;
924 
925 	ar = mtod(m, struct ether_arp *);
926 	op = ntohs(ar->arp_op);
927 	switch (op) {
928 	case ARPOP_REQUEST:
929 	case ARPOP_REPLY:	/* per RFC */
930 		in_arpinput(m);
931 		return;
932 	case ARPOP_REVREPLY:
933 		break;
934 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
935 	default:
936 		goto out;
937 	}
938 #ifdef NFSCLIENT
939 	if (!revarp_in_progress)
940 		goto out;
941 	ifp = m->m_pkthdr.rcvif;
942 	if (ifp != revarp_ifp) /* !same interface */
943 		goto out;
944 	if (revarp_finished)
945 		goto wake;
946 	if (memcmp(ar->arp_tha, ((struct arpcom *)ifp)->ac_enaddr,
947 	    sizeof(ar->arp_tha)))
948 		goto out;
949 	memcpy(&revarp_srvip, ar->arp_spa, sizeof(revarp_srvip));
950 	memcpy(&revarp_myip, ar->arp_tpa, sizeof(revarp_myip));
951 	revarp_finished = 1;
952 wake:	/* Do wakeup every time in case it was missed. */
953 	wakeup((caddr_t)&revarp_myip);
954 #endif
955 
956 out:
957 	m_freem(m);
958 }
959 
960 /*
961  * Send a RARP request for the ip address of the specified interface.
962  * The request should be RFC 903-compliant.
963  */
964 void
965 revarprequest(struct ifnet *ifp)
966 {
967 	struct sockaddr sa;
968 	struct mbuf *m;
969 	struct ether_header *eh;
970 	struct ether_arp *ea;
971 	struct arpcom *ac = (struct arpcom *)ifp;
972 
973 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
974 		return;
975 	m->m_len = sizeof(*ea);
976 	m->m_pkthdr.len = sizeof(*ea);
977 	MH_ALIGN(m, sizeof(*ea));
978 	ea = mtod(m, struct ether_arp *);
979 	eh = (struct ether_header *)sa.sa_data;
980 	memset(ea, 0, sizeof(*ea));
981 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
982 	eh->ether_type = htons(ETHERTYPE_REVARP);
983 	ea->arp_hrd = htons(ARPHRD_ETHER);
984 	ea->arp_pro = htons(ETHERTYPE_IP);
985 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
986 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
987 	ea->arp_op = htons(ARPOP_REVREQUEST);
988 	memcpy(eh->ether_shost, ac->ac_enaddr, sizeof(ea->arp_tha));
989 	memcpy(ea->arp_sha, ac->ac_enaddr, sizeof(ea->arp_sha));
990 	memcpy(ea->arp_tha, ac->ac_enaddr, sizeof(ea->arp_tha));
991 	sa.sa_family = pseudo_AF_HDRCMPLT;
992 	sa.sa_len = sizeof(sa);
993 	m->m_flags |= M_BCAST;
994 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
995 }
996 
997 #ifdef NFSCLIENT
998 /*
999  * RARP for the ip address of the specified interface, but also
1000  * save the ip address of the server that sent the answer.
1001  * Timeout if no response is received.
1002  */
1003 int
1004 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1005     struct in_addr *clnt_in)
1006 {
1007 	int result, count = 20;
1008 
1009 	if (revarp_finished)
1010 		return EIO;
1011 
1012 	revarp_ifp = ifp;
1013 	revarp_in_progress = 1;
1014 	while (count--) {
1015 		revarprequest(ifp);
1016 		result = tsleep((caddr_t)&revarp_myip, PSOCK, "revarp", hz/2);
1017 		if (result != EWOULDBLOCK)
1018 			break;
1019 	}
1020 	revarp_in_progress = 0;
1021 	if (!revarp_finished)
1022 		return ENETUNREACH;
1023 
1024 	memcpy(serv_in, &revarp_srvip, sizeof(*serv_in));
1025 	memcpy(clnt_in, &revarp_myip, sizeof(*clnt_in));
1026 	return 0;
1027 }
1028 
1029 /* For compatibility: only saves interface address. */
1030 int
1031 revarpwhoami(struct in_addr *in, struct ifnet *ifp)
1032 {
1033 	struct in_addr server;
1034 	return (revarpwhoarewe(ifp, &server, in));
1035 }
1036 #endif /* NFSCLIENT */
1037 
1038 #ifdef DDB
1039 
1040 #include <machine/db_machdep.h>
1041 #include <ddb/db_interface.h>
1042 #include <ddb/db_output.h>
1043 
1044 void
1045 db_print_sa(struct sockaddr *sa)
1046 {
1047 	int len;
1048 	u_char *p;
1049 
1050 	if (sa == 0) {
1051 		db_printf("[NULL]");
1052 		return;
1053 	}
1054 
1055 	p = (u_char *)sa;
1056 	len = sa->sa_len;
1057 	db_printf("[");
1058 	while (len > 0) {
1059 		db_printf("%d", *p);
1060 		p++;
1061 		len--;
1062 		if (len)
1063 			db_printf(",");
1064 	}
1065 	db_printf("]\n");
1066 }
1067 
1068 void
1069 db_print_ifa(struct ifaddr *ifa)
1070 {
1071 	if (ifa == 0)
1072 		return;
1073 	db_printf("  ifa_addr=");
1074 	db_print_sa(ifa->ifa_addr);
1075 	db_printf("  ifa_dsta=");
1076 	db_print_sa(ifa->ifa_dstaddr);
1077 	db_printf("  ifa_mask=");
1078 	db_print_sa(ifa->ifa_netmask);
1079 	db_printf("  flags=0x%x, refcnt=%d, metric=%d\n",
1080 	    ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric);
1081 }
1082 
1083 void
1084 db_print_llinfo(caddr_t li)
1085 {
1086 	struct llinfo_arp *la;
1087 
1088 	if (li == 0)
1089 		return;
1090 	la = (struct llinfo_arp *)li;
1091 	db_printf("  la_rt=%p la_hold_head=%p, la_asked=0x%lx\n",
1092 	    la->la_rt, la->la_hold_head, la->la_asked);
1093 }
1094 
1095 /*
1096  * Function to pass to rn_walktree().
1097  * Return non-zero error to abort walk.
1098  */
1099 int
1100 db_show_radix_node(struct radix_node *rn, void *w, u_int id)
1101 {
1102 	struct rtentry *rt = (struct rtentry *)rn;
1103 
1104 	db_printf("rtentry=%p", rt);
1105 
1106 	db_printf(" flags=0x%x refcnt=%d use=%llu expire=%lld rtableid=%u\n",
1107 	    rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id);
1108 
1109 	db_printf(" key="); db_print_sa(rt_key(rt));
1110 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1111 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1112 
1113 	db_printf(" ifp=%p ", rt->rt_ifp);
1114 	if (rt->rt_ifp)
1115 		db_printf("(%s)", rt->rt_ifp->if_xname);
1116 	else
1117 		db_printf("(NULL)");
1118 
1119 	db_printf(" ifa=%p\n", rt->rt_ifa);
1120 	db_print_ifa(rt->rt_ifa);
1121 
1122 	db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo);
1123 	db_print_llinfo(rt->rt_llinfo);
1124 	return (0);
1125 }
1126 
1127 /*
1128  * Function to print all the route trees.
1129  * Use this from ddb:  "call db_show_arptab"
1130  */
1131 int
1132 db_show_arptab(void)
1133 {
1134 	struct radix_node_head *rnh;
1135 	rnh = rtable_get(0, AF_INET);
1136 	db_printf("Route tree for AF_INET\n");
1137 	if (rnh == NULL) {
1138 		db_printf(" (not initialized)\n");
1139 		return (0);
1140 	}
1141 	rn_walktree(rnh, db_show_radix_node, NULL);
1142 	return (0);
1143 }
1144 #endif
1145 #endif /* INET */
1146