xref: /openbsd-src/sys/netinet/if_ether.c (revision aa5e9e10509ffd51558f081f01cd78bfa3c4f2a5)
1 /*	$OpenBSD: if_ether.c,v 1.101 2013/03/28 23:10:05 tedu 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_fddi.h>
58 #include <net/if_types.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_var.h>
62 #include <netinet/if_ether.h>
63 #if NCARP > 0
64 #include <netinet/ip_carp.h>
65 #endif
66 #if NBRIDGE > 0
67 #include <net/if_bridge.h>
68 #endif
69 
70 #define SIN(s) ((struct sockaddr_in *)s)
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	ifqueue arpintrq;
93 int	arp_inuse, arp_allocated;
94 int	arp_maxtries = 5;
95 int	useloopback = 1;	/* use loopback interface for local traffic */
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 /*
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 /*
139  * Parallel to llc_rtrequest.
140  */
141 void
142 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
143 {
144 	struct sockaddr *gate = rt->rt_gateway;
145 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
146 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
147 	struct in_ifaddr *ia;
148 	struct ifaddr *ifa;
149 	struct mbuf *m;
150 
151 	if (!arpinit_done) {
152 		static struct timeout arptimer_to;
153 
154 		arpinit_done = 1;
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 		if (req != RTM_ADD)
170 			return;
171 
172 		/*
173 		 * linklayers with particular link MTU limitation.  it is a bit
174 		 * awkward to have FDDI handling here, we should split ARP from
175 		 * netinet/if_ether.c like NetBSD does.
176 		 */
177 		switch (rt->rt_ifp->if_type) {
178 		case IFT_FDDI:
179 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
180 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
181 			break;
182 		}
183 
184 		return;
185 	}
186 
187 	switch (req) {
188 
189 	case RTM_ADD:
190 		/*
191 		 * XXX: If this is a manually added route to interface
192 		 * such as older version of routed or gated might provide,
193 		 * restore cloning bit.
194 		 */
195 		if ((rt->rt_flags & RTF_HOST) == 0 &&
196 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
197 			rt->rt_flags |= RTF_CLONING;
198 		if (rt->rt_flags & RTF_CLONING) {
199 			/*
200 			 * Case 1: This route should come from a route to iface.
201 			 */
202 			rt_setgate(rt, rt_key(rt),
203 			    (struct sockaddr *)&null_sdl,
204 			    rt->rt_ifp->if_rdomain);
205 			gate = rt->rt_gateway;
206 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
207 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
208 			/*
209 			 * Give this route an expiration time, even though
210 			 * it's a "permanent" route, so that routes cloned
211 			 * from it do not need their expiration time set.
212 			 */
213 			rt->rt_expire = time_second;
214 			/*
215 			 * linklayers with particular link MTU limitation.
216 			 */
217 			switch (rt->rt_ifp->if_type) {
218 			case IFT_FDDI:
219 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
220 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
221 				     (rt->rt_rmx.rmx_mtu == 0 &&
222 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
223 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
224 				break;
225 			}
226 			break;
227 		}
228 		/* Announce a new entry if requested. */
229 		if (rt->rt_flags & RTF_ANNOUNCE)
230 			arprequest(rt->rt_ifp,
231 			    &SIN(rt_key(rt))->sin_addr.s_addr,
232 			    &SIN(rt_key(rt))->sin_addr.s_addr,
233 			    (u_char *)LLADDR(SDL(gate)));
234 		/*FALLTHROUGH*/
235 	case RTM_RESOLVE:
236 		if (gate->sa_family != AF_LINK ||
237 		    gate->sa_len < sizeof(null_sdl)) {
238 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
239 			break;
240 		}
241 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
242 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
243 		if (la != 0)
244 			break; /* This happens on a route change */
245 		/*
246 		 * Case 2:  This route may come from cloning, or a manual route
247 		 * add with a LL address.
248 		 */
249 		R_Malloc(la, struct llinfo_arp *, sizeof(*la));
250 		rt->rt_llinfo = (caddr_t)la;
251 		if (la == 0) {
252 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
253 			break;
254 		}
255 		arp_inuse++, arp_allocated++;
256 		Bzero(la, sizeof(*la));
257 		la->la_rt = rt;
258 		rt->rt_flags |= RTF_LLINFO;
259 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
260 
261 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
262 			if (ia->ia_ifp == rt->rt_ifp &&
263 			    SIN(rt_key(rt))->sin_addr.s_addr ==
264 			    (IA_SIN(ia))->sin_addr.s_addr)
265 				break;
266 		}
267 		if (ia) {
268 			/*
269 			 * This test used to be
270 			 *	if (lo0ifp->if_flags & IFF_UP)
271 			 * It allowed local traffic to be forced through
272 			 * the hardware by configuring the loopback down.
273 			 * However, it causes problems during network
274 			 * configuration for boards that can't receive
275 			 * packets they send.  It is now necessary to clear
276 			 * "useloopback" and remove the route to force
277 			 * traffic out to the hardware.
278 			 *
279 			 * In 4.4BSD, the above "if" statement checked
280 			 * rt->rt_ifa against rt_key(rt).  It was changed
281 			 * to the current form so that we can provide a
282 			 * better support for multiple IPv4 addresses on a
283 			 * interface.
284 			 */
285 			rt->rt_expire = 0;
286 			Bcopy(((struct arpcom *)rt->rt_ifp)->ac_enaddr,
287 			    LLADDR(SDL(gate)),
288 			    SDL(gate)->sdl_alen = ETHER_ADDR_LEN);
289 			if (useloopback)
290 				rt->rt_ifp = lo0ifp;
291 			/*
292 			 * make sure to set rt->rt_ifa to the interface
293 			 * address we are using, otherwise we will have trouble
294 			 * with source address selection.
295 			 */
296 			ifa = &ia->ia_ifa;
297 			if (ifa != rt->rt_ifa) {
298 				ifafree(rt->rt_ifa);
299 				ifa->ifa_refcnt++;
300 				rt->rt_ifa = ifa;
301 			}
302 		}
303 		break;
304 
305 	case RTM_DELETE:
306 		if (la == 0)
307 			break;
308 		arp_inuse--;
309 		LIST_REMOVE(la, la_list);
310 		rt->rt_llinfo = 0;
311 		rt->rt_flags &= ~RTF_LLINFO;
312 		while ((m = la->la_hold_head) != NULL) {
313 			la->la_hold_head = la->la_hold_head->m_nextpkt;
314 			la_hold_total--;
315 			m_freem(m);
316 		}
317 		Free((caddr_t)la);
318 	}
319 }
320 
321 /*
322  * Broadcast an ARP request. Caller specifies:
323  *	- arp header source ip address
324  *	- arp header target ip address
325  *	- arp header source ethernet address
326  */
327 void
328 arprequest(struct ifnet *ifp, u_int32_t *sip, u_int32_t *tip, u_int8_t *enaddr)
329 {
330 	struct mbuf *m;
331 	struct ether_header *eh;
332 	struct ether_arp *ea;
333 	struct sockaddr sa;
334 
335 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
336 		return;
337 	m->m_len = sizeof(*ea);
338 	m->m_pkthdr.len = sizeof(*ea);
339 	m->m_pkthdr.rdomain = ifp->if_rdomain;
340 	MH_ALIGN(m, sizeof(*ea));
341 	ea = mtod(m, struct ether_arp *);
342 	eh = (struct ether_header *)sa.sa_data;
343 	bzero((caddr_t)ea, sizeof (*ea));
344 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
345 	    sizeof(eh->ether_dhost));
346 	eh->ether_type = htons(ETHERTYPE_ARP);	/* if_output will not swap */
347 	ea->arp_hrd = htons(ARPHRD_ETHER);
348 	ea->arp_pro = htons(ETHERTYPE_IP);
349 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
350 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
351 	ea->arp_op = htons(ARPOP_REQUEST);
352 	bcopy((caddr_t)enaddr, (caddr_t)eh->ether_shost,
353 	      sizeof(eh->ether_shost));
354 	bcopy((caddr_t)enaddr, (caddr_t)ea->arp_sha, sizeof(ea->arp_sha));
355 	bcopy((caddr_t)sip, (caddr_t)ea->arp_spa, sizeof(ea->arp_spa));
356 	bcopy((caddr_t)tip, (caddr_t)ea->arp_tpa, sizeof(ea->arp_tpa));
357 	sa.sa_family = pseudo_AF_HDRCMPLT;
358 	sa.sa_len = sizeof(sa);
359 	m->m_flags |= M_BCAST;
360 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
361 }
362 
363 /*
364  * Resolve an IP address into an ethernet address.  If success,
365  * desten is filled in.  If there is no entry in arptab,
366  * set one up and broadcast a request for the IP address.
367  * Hold onto this mbuf and resend it once the address
368  * is finally resolved.  A return value of 1 indicates
369  * that desten has been filled in and the packet should be sent
370  * normally; a 0 return indicates that the packet has been
371  * taken over here, either now or for later transmission.
372  */
373 int
374 arpresolve(struct arpcom *ac, struct rtentry *rt, struct mbuf *m,
375     struct sockaddr *dst, u_char *desten)
376 {
377 	struct llinfo_arp *la;
378 	struct sockaddr_dl *sdl;
379 	struct mbuf *mh;
380 
381 	if (m->m_flags & M_BCAST) {	/* broadcast */
382 		bcopy((caddr_t)etherbroadcastaddr, (caddr_t)desten,
383 		    sizeof(etherbroadcastaddr));
384 		return (1);
385 	}
386 	if (m->m_flags & M_MCAST) {	/* multicast */
387 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
388 		return (1);
389 	}
390 	if (rt) {
391 		la = (struct llinfo_arp *)rt->rt_llinfo;
392 		if (la == NULL)
393 			log(LOG_DEBUG, "arpresolve: %s: route without link "
394 			    "local address\n", inet_ntoa(SIN(dst)->sin_addr));
395 	} else {
396 		if ((la = arplookup(SIN(dst)->sin_addr.s_addr, RT_REPORT, 0,
397 		    ac->ac_if.if_rdomain)) != NULL)
398 			rt = la->la_rt;
399 		else
400 			log(LOG_DEBUG,
401 			    "arpresolve: %s: can't allocate llinfo\n",
402 			    inet_ntoa(SIN(dst)->sin_addr));
403 	}
404 	if (la == 0 || rt == 0) {
405 		m_freem(m);
406 		return (0);
407 	}
408 	sdl = SDL(rt->rt_gateway);
409 	/*
410 	 * Check the address family and length is valid, the address
411 	 * is resolved; otherwise, try to resolve.
412 	 */
413 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
414 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
415 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
416 		return 1;
417 	}
418 	if (((struct ifnet *)ac)->if_flags & IFF_NOARP) {
419 		m_freem(m);
420 		return 0;
421 	}
422 
423 	/*
424 	 * There is an arptab entry, but no ethernet address
425 	 * response yet. Insert mbuf in hold queue if below limit
426 	 * if above the limit free the queue without queuing the new packet.
427 	 */
428 	if (la_hold_total < MAX_HOLD_TOTAL && la_hold_total < nmbclust / 64) {
429 		if (la->la_hold_count >= MAX_HOLD_QUEUE) {
430 			mh = la->la_hold_head;
431 			la->la_hold_head = la->la_hold_head->m_nextpkt;
432 			if (mh == la->la_hold_tail)
433 				la->la_hold_tail = NULL;
434 			la->la_hold_count--;
435 			la_hold_total--;
436 			m_freem(mh);
437 		}
438 		if (la->la_hold_tail == NULL)
439 			la->la_hold_head = m;
440 		else
441 			la->la_hold_tail->m_nextpkt = m;
442 		la->la_hold_tail = m;
443 		la->la_hold_count++;
444 		la_hold_total++;
445 	} else {
446 		while ((mh = la->la_hold_head) != NULL) {
447 			la->la_hold_head =
448 			    la->la_hold_head->m_nextpkt;
449 			la_hold_total--;
450 			m_freem(mh);
451 		}
452 		la->la_hold_tail = NULL;
453 		la->la_hold_count = 0;
454 		m_freem(m);
455 	}
456 
457 	/*
458 	 * Re-send the ARP request when appropriate.
459 	 */
460 #ifdef	DIAGNOSTIC
461 	if (rt->rt_expire == 0) {
462 		/* This should never happen. (Should it? -gwr) */
463 		printf("arpresolve: unresolved and rt_expire == 0\n");
464 		/* Set expiration time to now (expired). */
465 		rt->rt_expire = time_second;
466 	}
467 #endif
468 	if (rt->rt_expire) {
469 		rt->rt_flags &= ~RTF_REJECT;
470 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
471 			rt->rt_expire = time_second;
472 			if (la->la_asked++ < arp_maxtries)
473 				arprequest(&ac->ac_if,
474 				    &(SIN(rt->rt_ifa->ifa_addr)->sin_addr.s_addr),
475 				    &(SIN(dst)->sin_addr.s_addr),
476 #if NCARP > 0
477 				    (rt->rt_ifp->if_type == IFT_CARP) ?
478 					((struct arpcom *) rt->rt_ifp->if_softc
479 					)->ac_enaddr :
480 #endif
481 				    ac->ac_enaddr);
482 			else {
483 				rt->rt_flags |= RTF_REJECT;
484 				rt->rt_expire += arpt_down;
485 				la->la_asked = 0;
486 				while ((mh = la->la_hold_head) != NULL) {
487 					la->la_hold_head =
488 					    la->la_hold_head->m_nextpkt;
489 					la_hold_total--;
490 					m_freem(mh);
491 				}
492 				la->la_hold_tail = NULL;
493 				la->la_hold_count = 0;
494 			}
495 		}
496 	}
497 	return (0);
498 }
499 
500 /*
501  * Common length and type checks are done here,
502  * then the protocol-specific routine is called.
503  */
504 void
505 arpintr(void)
506 {
507 	struct mbuf *m;
508 	struct arphdr *ar;
509 	int s, len;
510 
511 	for (;;) {
512 		s = splnet();
513 		IF_DEQUEUE(&arpintrq, m);
514 		splx(s);
515 		if (m == NULL)
516 			break;
517 #ifdef DIAGNOSTIC
518 		if ((m->m_flags & M_PKTHDR) == 0)
519 			panic("arpintr");
520 #endif
521 
522 		len = sizeof(struct arphdr);
523 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
524 			continue;
525 
526 		ar = mtod(m, struct arphdr *);
527 		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) {
528 			m_freem(m);
529 			continue;
530 		}
531 
532 		len += 2 * (ar->ar_hln + ar->ar_pln);
533 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
534 			continue;
535 
536 		switch (ntohs(ar->ar_pro)) {
537 		case ETHERTYPE_IP:
538 		case ETHERTYPE_IPTRAILERS:
539 			in_arpinput(m);
540 			continue;
541 		}
542 		m_freem(m);
543 	}
544 }
545 
546 /*
547  * ARP for Internet protocols on Ethernet.
548  * Algorithm is that given in RFC 826.
549  * In addition, a sanity check is performed on the sender
550  * protocol address, to catch impersonators.
551  * We no longer handle negotiations for use of trailer protocol:
552  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
553  * along with IP replies if we wanted trailers sent to us,
554  * and also sent them in response to IP replies.
555  * This allowed either end to announce the desire to receive
556  * trailer packets.
557  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
558  * but formerly didn't normally send requests.
559  */
560 void
561 in_arpinput(struct mbuf *m)
562 {
563 	struct ether_arp *ea;
564 	struct arpcom *ac = (struct arpcom *)m->m_pkthdr.rcvif;
565 	struct ether_header *eh;
566 	struct llinfo_arp *la = 0;
567 	struct rtentry *rt;
568 	struct in_ifaddr *ia;
569 	struct sockaddr_dl *sdl;
570 	struct sockaddr sa;
571 	struct in_addr isaddr, itaddr, myaddr;
572 	struct mbuf *mh, *mt;
573 	u_int8_t *enaddr = NULL;
574 #if NCARP > 0
575 	u_int8_t *ether_shost = NULL;
576 #endif
577 	int op;
578 
579 	ea = mtod(m, struct ether_arp *);
580 	op = ntohs(ea->arp_op);
581 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
582 		goto out;
583 #if notyet
584 	if ((op == ARPOP_REPLY) && (m->m_flags & (M_BCAST|M_MCAST))) {
585 		log(LOG_ERR,
586 		    "arp: received reply to broadcast or multicast address\n");
587 		goto out;
588 	}
589 #endif
590 
591 	bcopy((caddr_t)ea->arp_tpa, (caddr_t)&itaddr, sizeof(itaddr));
592 	bcopy((caddr_t)ea->arp_spa, (caddr_t)&isaddr, sizeof(isaddr));
593 
594 	/* First try: check target against our addresses */
595 	TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
596 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
597 			continue;
598 
599 #if NCARP > 0
600 		if (ia->ia_ifp->if_type == IFT_CARP &&
601 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
602 		    (IFF_UP|IFF_RUNNING))) {
603 			if (ia->ia_ifp == m->m_pkthdr.rcvif) {
604 				if (op == ARPOP_REPLY)
605 					break;
606 				if (carp_iamatch(ia, ea->arp_sha,
607 				    &enaddr, &ether_shost))
608 					break;
609 				else
610 					goto out;
611 			}
612 		} else
613 #endif
614 			if (ia->ia_ifp == m->m_pkthdr.rcvif)
615 				break;
616 	}
617 
618 	/* Second try: check source against our addresses */
619 	if (ia == NULL) {
620 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
621 			if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
622 				continue;
623 			if (ia->ia_ifp == m->m_pkthdr.rcvif)
624 				break;
625 		}
626 	}
627 
628 	/* Third try: not one of our addresses, just find an usable ia */
629 	if (ia == NULL) {
630 		struct ifaddr *ifa;
631 
632 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
633 			if (ifa->ifa_addr->sa_family == AF_INET)
634 				break;
635 		}
636 		if (ifa)
637 			ia = (struct in_ifaddr *)ifa;
638 	}
639 
640 	if (ia == NULL)
641 		goto out;
642 
643 	if (!enaddr)
644 		enaddr = ac->ac_enaddr;
645 	myaddr = ia->ia_addr.sin_addr;
646 
647 	if (!bcmp((caddr_t)ea->arp_sha, enaddr, sizeof (ea->arp_sha)))
648 		goto out;	/* it's from me, ignore it. */
649 	if (ETHER_IS_MULTICAST(&ea->arp_sha[0]))
650 		if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)etherbroadcastaddr,
651 		    sizeof (ea->arp_sha))) {
652 			log(LOG_ERR, "arp: ether address is broadcast for "
653 			    "IP address %s!\n", inet_ntoa(isaddr));
654 			goto out;
655 		}
656 	if (myaddr.s_addr && isaddr.s_addr == myaddr.s_addr) {
657 		log(LOG_ERR,
658 		   "duplicate IP address %s sent from ethernet address %s\n",
659 		   inet_ntoa(isaddr), ether_sprintf(ea->arp_sha));
660 		itaddr = myaddr;
661 		goto reply;
662 	}
663 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0,
664 	    rtable_l2(m->m_pkthdr.rdomain));
665 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
666 		if (sdl->sdl_alen) {
667 		    if (bcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
668 			if (rt->rt_flags & RTF_PERMANENT_ARP) {
669 				log(LOG_WARNING,
670 				   "arp: attempt to overwrite permanent "
671 				   "entry for %s by %s on %s\n",
672 				   inet_ntoa(isaddr),
673 				   ether_sprintf(ea->arp_sha),
674 				   ac->ac_if.if_xname);
675 				goto out;
676 			} else if (rt->rt_ifp != &ac->ac_if) {
677 #if NCARP > 0
678 				if (ac->ac_if.if_type != IFT_CARP)
679 #endif
680 					log(LOG_WARNING,
681 					   "arp: attempt to overwrite entry for"
682 					   " %s on %s by %s on %s\n",
683 					   inet_ntoa(isaddr),
684 					   rt->rt_ifp->if_xname,
685 					   ether_sprintf(ea->arp_sha),
686 					   ac->ac_if.if_xname);
687 				goto out;
688 			} else {
689 				log(LOG_INFO,
690 				   "arp info overwritten for %s by %s on %s\n",
691 				   inet_ntoa(isaddr),
692 				   ether_sprintf(ea->arp_sha),
693 				   ac->ac_if.if_xname);
694 				rt->rt_expire = 1; /* no longer static */
695 			}
696 		    }
697 		} else if (rt->rt_ifp != &ac->ac_if &&
698 #if NBRIDGE > 0
699 		    !SAME_BRIDGE(ac->ac_if.if_bridgeport,
700 		    rt->rt_ifp->if_bridgeport) &&
701 #endif
702 #if NCARP > 0
703 		    !(rt->rt_ifp->if_type == IFT_CARP &&
704 		    rt->rt_ifp->if_carpdev == &ac->ac_if) &&
705 		    !(ac->ac_if.if_type == IFT_CARP &&
706 		    ac->ac_if.if_carpdev == rt->rt_ifp) &&
707 #endif
708 		    1) {
709 			log(LOG_WARNING,
710 			    "arp: attempt to add entry for %s "
711 			    "on %s by %s on %s\n",
712 			    inet_ntoa(isaddr), rt->rt_ifp->if_xname,
713 			    ether_sprintf(ea->arp_sha),
714 			    ac->ac_if.if_xname);
715 			goto out;
716 		}
717 		bcopy(ea->arp_sha, LLADDR(sdl),
718 		    sdl->sdl_alen = sizeof(ea->arp_sha));
719 		if (rt->rt_expire)
720 			rt->rt_expire = time_second + arpt_keep;
721 		rt->rt_flags &= ~RTF_REJECT;
722 		la->la_asked = 0;
723 		while ((mh = la->la_hold_head) != NULL) {
724 			if ((la->la_hold_head = mh->m_nextpkt) == NULL)
725 				la->la_hold_tail = NULL;
726 			la->la_hold_count--;
727 			la_hold_total--;
728 			mt = la->la_hold_tail;
729 
730 			(*ac->ac_if.if_output)(&ac->ac_if, mh, rt_key(rt), rt);
731 
732 			if (la->la_hold_tail == mh) {
733 				/* mbuf is back in queue. Discard. */
734 				la->la_hold_tail = mt;
735 				if (la->la_hold_tail)
736 					la->la_hold_tail->m_nextpkt = NULL;
737 				else
738 					la->la_hold_head = NULL;
739 				la->la_hold_count--;
740 				la_hold_total--;
741 				m_freem(mh);
742 			}
743 		}
744 	}
745 reply:
746 	if (op != ARPOP_REQUEST) {
747 out:
748 		m_freem(m);
749 		return;
750 	}
751 	if (itaddr.s_addr == myaddr.s_addr) {
752 		/* I am the target */
753 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
754 		bcopy(enaddr, ea->arp_sha, sizeof(ea->arp_sha));
755 	} else {
756 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY,
757 		    rtable_l2(m->m_pkthdr.rdomain));
758 		if (la == 0)
759 			goto out;
760 		rt = la->la_rt;
761 		if (rt->rt_ifp->if_type == IFT_CARP &&
762 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
763 			goto out;
764 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
765 		sdl = SDL(rt->rt_gateway);
766 		bcopy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
767 	}
768 
769 	bcopy(ea->arp_spa, ea->arp_tpa, sizeof(ea->arp_spa));
770 	bcopy(&itaddr, ea->arp_spa, sizeof(ea->arp_spa));
771 	ea->arp_op = htons(ARPOP_REPLY);
772 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
773 	eh = (struct ether_header *)sa.sa_data;
774 	bcopy(ea->arp_tha, eh->ether_dhost, sizeof(eh->ether_dhost));
775 #if NCARP > 0
776 	if (ether_shost)
777 		enaddr = ether_shost;
778 #endif
779 	bcopy(enaddr, eh->ether_shost, sizeof(eh->ether_shost));
780 
781 	eh->ether_type = htons(ETHERTYPE_ARP);
782 	sa.sa_family = pseudo_AF_HDRCMPLT;
783 	sa.sa_len = sizeof(sa);
784 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
785 	return;
786 }
787 
788 /*
789  * Free an arp entry.
790  */
791 void
792 arptfree(struct llinfo_arp *la)
793 {
794 	struct rtentry *rt = la->la_rt;
795 	struct sockaddr_dl *sdl;
796 	struct rt_addrinfo info;
797 	u_int tid = 0;
798 
799 	if (rt == 0)
800 		panic("arptfree");
801 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
802 	    sdl->sdl_family == AF_LINK) {
803 		sdl->sdl_alen = 0;
804 		la->la_asked = 0;
805 		rt->rt_flags &= ~RTF_REJECT;
806 		return;
807 	}
808 	bzero(&info, sizeof(info));
809 	info.rti_info[RTAX_DST] = rt_key(rt);
810 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
811 
812 	if (rt->rt_ifp)
813 		tid = rt->rt_ifp->if_rdomain;
814 
815 	rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL, tid);
816 }
817 
818 /*
819  * Lookup or enter a new address in arptab.
820  */
821 struct llinfo_arp *
822 arplookup(u_int32_t addr, int create, int proxy, u_int tableid)
823 {
824 	struct rtentry *rt;
825 	static struct sockaddr_inarp sin;
826 
827 	sin.sin_len = sizeof(sin);
828 	sin.sin_family = AF_INET;
829 	sin.sin_addr.s_addr = addr;
830 	sin.sin_other = proxy ? SIN_PROXY : 0;
831 	rt = rtalloc1((struct sockaddr *)&sin, create, tableid);
832 	if (rt == 0)
833 		return (0);
834 	rt->rt_refcnt--;
835 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
836 	    rt->rt_gateway->sa_family != AF_LINK) {
837 		if (create) {
838 			if (rt->rt_refcnt <= 0 &&
839 			    (rt->rt_flags & RTF_CLONED) != 0) {
840 				struct rt_addrinfo info;
841 
842 				bzero(&info, sizeof(info));
843 				info.rti_info[RTAX_DST] = rt_key(rt);
844 				info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
845 				info.rti_info[RTAX_NETMASK] = rt_mask(rt);
846 
847 				rtrequest1(RTM_DELETE, &info, rt->rt_priority,
848 				    NULL, tableid);
849 			}
850 		}
851 		return (0);
852 	}
853 	return ((struct llinfo_arp *)rt->rt_llinfo);
854 }
855 
856 void
857 arp_ifinit(struct arpcom *ac, struct ifaddr *ifa)
858 {
859 
860 	/* Warn the user if another station has this IP address. */
861 	arprequest(&ac->ac_if,
862 	    &(IA_SIN(ifa)->sin_addr.s_addr),
863 	    &(IA_SIN(ifa)->sin_addr.s_addr),
864 	    ac->ac_enaddr);
865 	ifa->ifa_rtrequest = arp_rtrequest;
866 	ifa->ifa_flags |= RTF_CLONING;
867 }
868 
869 /*
870  * Called from Ethernet interrupt handlers
871  * when ether packet type ETHERTYPE_REVARP
872  * is received.  Common length and type checks are done here,
873  * then the protocol-specific routine is called.
874  */
875 void
876 revarpinput(struct mbuf *m)
877 {
878 	struct arphdr *ar;
879 
880 	if (m->m_len < sizeof(struct arphdr))
881 		goto out;
882 	ar = mtod(m, struct arphdr *);
883 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
884 		goto out;
885 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
886 		goto out;
887 	switch (ntohs(ar->ar_pro)) {
888 
889 	case ETHERTYPE_IP:
890 	case ETHERTYPE_IPTRAILERS:
891 		in_revarpinput(m);
892 		return;
893 
894 	default:
895 		break;
896 	}
897 out:
898 	m_freem(m);
899 }
900 
901 /*
902  * RARP for Internet protocols on Ethernet.
903  * Algorithm is that given in RFC 903.
904  * We are only using for bootstrap purposes to get an ip address for one of
905  * our interfaces.  Thus we support no user-interface.
906  *
907  * Since the contents of the RARP reply are specific to the interface that
908  * sent the request, this code must ensure that they are properly associated.
909  *
910  * Note: also supports ARP via RARP packets, per the RFC.
911  */
912 void
913 in_revarpinput(struct mbuf *m)
914 {
915 #ifdef NFSCLIENT
916 	struct ifnet *ifp;
917 #endif /* NFSCLIENT */
918 	struct ether_arp *ar;
919 	int op;
920 
921 	ar = mtod(m, struct ether_arp *);
922 	op = ntohs(ar->arp_op);
923 	switch (op) {
924 	case ARPOP_REQUEST:
925 	case ARPOP_REPLY:	/* per RFC */
926 		in_arpinput(m);
927 		return;
928 	case ARPOP_REVREPLY:
929 		break;
930 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
931 	default:
932 		goto out;
933 	}
934 #ifdef NFSCLIENT
935 	if (!revarp_in_progress)
936 		goto out;
937 	ifp = m->m_pkthdr.rcvif;
938 	if (ifp != revarp_ifp) /* !same interface */
939 		goto out;
940 	if (revarp_finished)
941 		goto wake;
942 	if (bcmp(ar->arp_tha, ((struct arpcom *)ifp)->ac_enaddr,
943 	    sizeof(ar->arp_tha)))
944 		goto out;
945 	bcopy((caddr_t)ar->arp_spa, (caddr_t)&revarp_srvip,
946 	    sizeof(revarp_srvip));
947 	bcopy((caddr_t)ar->arp_tpa, (caddr_t)&revarp_myip,
948 	    sizeof(revarp_myip));
949 	revarp_finished = 1;
950 wake:	/* Do wakeup every time in case it was missed. */
951 	wakeup((caddr_t)&revarp_myip);
952 #endif
953 
954 out:
955 	m_freem(m);
956 }
957 
958 /*
959  * Send a RARP request for the ip address of the specified interface.
960  * The request should be RFC 903-compliant.
961  */
962 void
963 revarprequest(struct ifnet *ifp)
964 {
965 	struct sockaddr sa;
966 	struct mbuf *m;
967 	struct ether_header *eh;
968 	struct ether_arp *ea;
969 	struct arpcom *ac = (struct arpcom *)ifp;
970 
971 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
972 		return;
973 	m->m_len = sizeof(*ea);
974 	m->m_pkthdr.len = sizeof(*ea);
975 	MH_ALIGN(m, sizeof(*ea));
976 	ea = mtod(m, struct ether_arp *);
977 	eh = (struct ether_header *)sa.sa_data;
978 	bzero((caddr_t)ea, sizeof(*ea));
979 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
980 	    sizeof(eh->ether_dhost));
981 	eh->ether_type = htons(ETHERTYPE_REVARP);
982 	ea->arp_hrd = htons(ARPHRD_ETHER);
983 	ea->arp_pro = htons(ETHERTYPE_IP);
984 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
985 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
986 	ea->arp_op = htons(ARPOP_REVREQUEST);
987 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)eh->ether_shost,
988 	   sizeof(ea->arp_tha));
989 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_sha,
990 	   sizeof(ea->arp_sha));
991 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_tha,
992 	   sizeof(ea->arp_tha));
993 	sa.sa_family = pseudo_AF_HDRCMPLT;
994 	sa.sa_len = sizeof(sa);
995 	m->m_flags |= M_BCAST;
996 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
997 }
998 
999 #ifdef NFSCLIENT
1000 /*
1001  * RARP for the ip address of the specified interface, but also
1002  * save the ip address of the server that sent the answer.
1003  * Timeout if no response is received.
1004  */
1005 int
1006 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1007     struct in_addr *clnt_in)
1008 {
1009 	int result, count = 20;
1010 
1011 	if (revarp_finished)
1012 		return EIO;
1013 
1014 	revarp_ifp = ifp;
1015 	revarp_in_progress = 1;
1016 	while (count--) {
1017 		revarprequest(ifp);
1018 		result = tsleep((caddr_t)&revarp_myip, PSOCK, "revarp", hz/2);
1019 		if (result != EWOULDBLOCK)
1020 			break;
1021 	}
1022 	revarp_in_progress = 0;
1023 	if (!revarp_finished)
1024 		return ENETUNREACH;
1025 
1026 	bcopy((caddr_t)&revarp_srvip, serv_in, sizeof(*serv_in));
1027 	bcopy((caddr_t)&revarp_myip, clnt_in, sizeof(*clnt_in));
1028 	return 0;
1029 }
1030 
1031 /* For compatibility: only saves interface address. */
1032 int
1033 revarpwhoami(struct in_addr *in, struct ifnet *ifp)
1034 {
1035 	struct in_addr server;
1036 	return (revarpwhoarewe(ifp, &server, in));
1037 }
1038 #endif /* NFSCLIENT */
1039 
1040 #ifdef DDB
1041 
1042 #include <machine/db_machdep.h>
1043 #include <ddb/db_interface.h>
1044 #include <ddb/db_output.h>
1045 
1046 void
1047 db_print_sa(struct sockaddr *sa)
1048 {
1049 	int len;
1050 	u_char *p;
1051 
1052 	if (sa == 0) {
1053 		db_printf("[NULL]");
1054 		return;
1055 	}
1056 
1057 	p = (u_char *)sa;
1058 	len = sa->sa_len;
1059 	db_printf("[");
1060 	while (len > 0) {
1061 		db_printf("%d", *p);
1062 		p++;
1063 		len--;
1064 		if (len)
1065 			db_printf(",");
1066 	}
1067 	db_printf("]\n");
1068 }
1069 
1070 void
1071 db_print_ifa(struct ifaddr *ifa)
1072 {
1073 	if (ifa == 0)
1074 		return;
1075 	db_printf("  ifa_addr=");
1076 	db_print_sa(ifa->ifa_addr);
1077 	db_printf("  ifa_dsta=");
1078 	db_print_sa(ifa->ifa_dstaddr);
1079 	db_printf("  ifa_mask=");
1080 	db_print_sa(ifa->ifa_netmask);
1081 	db_printf("  flags=0x%x, refcnt=%d, metric=%d\n",
1082 	    ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric);
1083 }
1084 
1085 void
1086 db_print_llinfo(caddr_t li)
1087 {
1088 	struct llinfo_arp *la;
1089 
1090 	if (li == 0)
1091 		return;
1092 	la = (struct llinfo_arp *)li;
1093 	db_printf("  la_rt=%p la_hold_head=%p, la_asked=0x%lx\n",
1094 	    la->la_rt, la->la_hold_head, la->la_asked);
1095 }
1096 
1097 /*
1098  * Function to pass to rn_walktree().
1099  * Return non-zero error to abort walk.
1100  */
1101 int
1102 db_show_radix_node(struct radix_node *rn, void *w, u_int id)
1103 {
1104 	struct rtentry *rt = (struct rtentry *)rn;
1105 
1106 	db_printf("rtentry=%p", rt);
1107 
1108 	db_printf(" flags=0x%x refcnt=%d use=%llu expire=%u rtableid %u\n",
1109 	    rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id);
1110 
1111 	db_printf(" key="); db_print_sa(rt_key(rt));
1112 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1113 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1114 
1115 	db_printf(" ifp=%p ", rt->rt_ifp);
1116 	if (rt->rt_ifp)
1117 		db_printf("(%s)", rt->rt_ifp->if_xname);
1118 	else
1119 		db_printf("(NULL)");
1120 
1121 	db_printf(" ifa=%p\n", rt->rt_ifa);
1122 	db_print_ifa(rt->rt_ifa);
1123 
1124 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1125 
1126 	db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo);
1127 	db_print_llinfo(rt->rt_llinfo);
1128 	return (0);
1129 }
1130 
1131 /*
1132  * Function to print all the route trees.
1133  * Use this from ddb:  "call db_show_arptab"
1134  */
1135 int
1136 db_show_arptab(void)
1137 {
1138 	struct radix_node_head *rnh;
1139 	rnh = rt_gettable(AF_INET, 0);
1140 	db_printf("Route tree for AF_INET\n");
1141 	if (rnh == NULL) {
1142 		db_printf(" (not initialized)\n");
1143 		return (0);
1144 	}
1145 	rn_walktree(rnh, db_show_radix_node, NULL);
1146 	return (0);
1147 }
1148 #endif
1149 #endif /* INET */
1150