xref: /openbsd-src/sys/netinet/if_ether.c (revision d874cce4b1d9fe6b41c9e4f2117a77d8a4a37b92)
1 /*	$OpenBSD: if_ether.c,v 1.75 2008/06/06 13:29:43 henning 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/kernel.h>
51 #include <sys/syslog.h>
52 #include <sys/proc.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 
67 #define SIN(s) ((struct sockaddr_in *)s)
68 #define SDL(s) ((struct sockaddr_dl *)s)
69 #define SRP(s) ((struct sockaddr_inarp *)s)
70 
71 /*
72  * ARP trailer negotiation.  Trailer protocol is not IP specific,
73  * but ARP request/response use IP addresses.
74  */
75 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
76 
77 /* timer values */
78 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
79 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
80 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
81 #define	rt_expire rt_rmx.rmx_expire
82 
83 void arptfree(struct llinfo_arp *);
84 void arptimer(void *);
85 struct llinfo_arp *arplookup(u_int32_t, int, int);
86 void in_arpinput(struct mbuf *);
87 
88 LIST_HEAD(, llinfo_arp) llinfo_arp;
89 struct	ifqueue arpintrq = {0, 0, 0, 50};
90 int	arp_inuse, arp_allocated, arp_intimer;
91 int	arp_maxtries = 5;
92 int	useloopback = 1;	/* use loopback interface for local traffic */
93 int	arpinit_done = 0;
94 
95 /* revarp state */
96 struct in_addr myip, srv_ip;
97 int myip_initialized = 0;
98 int revarp_in_progress = 0;
99 struct ifnet *myip_ifp = NULL;
100 
101 #ifdef DDB
102 #include <uvm/uvm_extern.h>
103 
104 void	db_print_sa(struct sockaddr *);
105 void	db_print_ifa(struct ifaddr *);
106 void	db_print_llinfo(caddr_t);
107 int	db_show_radix_node(struct radix_node *, void *);
108 #endif
109 
110 /*
111  * Timeout routine.  Age arp_tab entries periodically.
112  */
113 /* ARGSUSED */
114 void
115 arptimer(arg)
116 	void *arg;
117 {
118 	struct timeout *to = (struct timeout *)arg;
119 	int s;
120 	struct llinfo_arp *la, *nla;
121 
122 	s = splsoftnet();
123 	timeout_add(to, arpt_prune * hz);
124 	for (la = LIST_FIRST(&llinfo_arp); la != LIST_END(&llinfo_arp);
125 	    la = nla) {
126 		struct rtentry *rt = la->la_rt;
127 
128 		nla = LIST_NEXT(la, la_list);
129 		if (rt->rt_expire && rt->rt_expire <= time_second)
130 			arptfree(la); /* timer has expired; clear */
131 	}
132 	splx(s);
133 }
134 
135 /*
136  * Parallel to llc_rtrequest.
137  */
138 void
139 arp_rtrequest(req, rt, info)
140 	int req;
141 	struct rtentry *rt;
142 	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 
150 	if (!arpinit_done) {
151 		static struct timeout arptimer_to;
152 
153 		arpinit_done = 1;
154 		/*
155 		 * We generate expiration times from time.tv_sec
156 		 * so avoid accidently creating permanent routes.
157 		 */
158 		if (time_second == 0) {
159 			time_second++;
160 		}
161 
162 		timeout_set(&arptimer_to, arptimer, &arptimer_to);
163 		timeout_add(&arptimer_to, hz);
164 	}
165 
166 	if (rt->rt_flags & RTF_GATEWAY) {
167 		if (req != RTM_ADD)
168 			return;
169 
170 		/*
171 		 * linklayers with particular link MTU limitation.  it is a bit
172 		 * awkward to have FDDI handling here, we should split ARP from
173 		 * netinet/if_ether.c like NetBSD does.
174 		 */
175 		switch (rt->rt_ifp->if_type) {
176 		case IFT_FDDI:
177 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
178 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
179 			break;
180 		}
181 
182 		return;
183 	}
184 
185 	switch (req) {
186 
187 	case RTM_ADD:
188 		/*
189 		 * XXX: If this is a manually added route to interface
190 		 * such as older version of routed or gated might provide,
191 		 * restore cloning bit.
192 		 */
193 		if ((rt->rt_flags & RTF_HOST) == 0 &&
194 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
195 			rt->rt_flags |= RTF_CLONING;
196 		if (rt->rt_flags & RTF_CLONING) {
197 			/*
198 			 * Case 1: This route should come from a route to iface.
199 			 */
200 			rt_setgate(rt, rt_key(rt),
201 			    (struct sockaddr *)&null_sdl, 0);
202 			gate = rt->rt_gateway;
203 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
204 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
205 			/*
206 			 * Give this route an expiration time, even though
207 			 * it's a "permanent" route, so that routes cloned
208 			 * from it do not need their expiration time set.
209 			 */
210 			rt->rt_expire = time_second;
211 			/*
212 			 * linklayers with particular link MTU limitation.
213 			 */
214 			switch (rt->rt_ifp->if_type) {
215 			case IFT_FDDI:
216 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
217 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
218 				     (rt->rt_rmx.rmx_mtu == 0 &&
219 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
220 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
221 				break;
222 			}
223 			break;
224 		}
225 		/* Announce a new entry if requested. */
226 		if (rt->rt_flags & RTF_ANNOUNCE)
227 			arprequest(rt->rt_ifp,
228 			    &SIN(rt_key(rt))->sin_addr.s_addr,
229 			    &SIN(rt_key(rt))->sin_addr.s_addr,
230 			    (u_char *)LLADDR(SDL(gate)));
231 		/*FALLTHROUGH*/
232 	case RTM_RESOLVE:
233 		if (gate->sa_family != AF_LINK ||
234 		    gate->sa_len < sizeof(null_sdl)) {
235 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
236 			break;
237 		}
238 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
239 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
240 		if (la != 0)
241 			break; /* This happens on a route change */
242 		/*
243 		 * Case 2:  This route may come from cloning, or a manual route
244 		 * add with a LL address.
245 		 */
246 		R_Malloc(la, struct llinfo_arp *, sizeof(*la));
247 		rt->rt_llinfo = (caddr_t)la;
248 		if (la == 0) {
249 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
250 			break;
251 		}
252 		arp_inuse++, arp_allocated++;
253 		Bzero(la, sizeof(*la));
254 		la->la_rt = rt;
255 		rt->rt_flags |= RTF_LLINFO;
256 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
257 
258 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
259 			if (ia->ia_ifp == rt->rt_ifp &&
260 			    SIN(rt_key(rt))->sin_addr.s_addr ==
261 			    (IA_SIN(ia))->sin_addr.s_addr)
262 				break;
263 		}
264 		if (ia) {
265 			/*
266 			 * This test used to be
267 			 *	if (lo0ifp->if_flags & IFF_UP)
268 			 * It allowed local traffic to be forced through
269 			 * the hardware by configuring the loopback down.
270 			 * However, it causes problems during network
271 			 * configuration for boards that can't receive
272 			 * packets they send.  It is now necessary to clear
273 			 * "useloopback" and remove the route to force
274 			 * traffic out to the hardware.
275 			 *
276 			 * In 4.4BSD, the above "if" statement checked
277 			 * rt->rt_ifa against rt_key(rt).  It was changed
278 			 * to the current form so that we can provide a
279 			 * better support for multiple IPv4 addresses on a
280 			 * interface.
281 			 */
282 			rt->rt_expire = 0;
283 			Bcopy(((struct arpcom *)rt->rt_ifp)->ac_enaddr,
284 			    LLADDR(SDL(gate)),
285 			    SDL(gate)->sdl_alen = ETHER_ADDR_LEN);
286 			if (useloopback)
287 				rt->rt_ifp = lo0ifp;
288 			/*
289 			 * make sure to set rt->rt_ifa to the interface
290 			 * address we are using, otherwise we will have trouble
291 			 * with source address selection.
292 			 */
293 			ifa = &ia->ia_ifa;
294 			if (ifa != rt->rt_ifa) {
295 				IFAFREE(rt->rt_ifa);
296 				ifa->ifa_refcnt++;
297 				rt->rt_ifa = ifa;
298 			}
299 		}
300 		break;
301 
302 	case RTM_DELETE:
303 		if (la == 0)
304 			break;
305 		arp_inuse--;
306 		LIST_REMOVE(la, la_list);
307 		rt->rt_llinfo = 0;
308 		rt->rt_flags &= ~RTF_LLINFO;
309 		if (la->la_hold)
310 			m_freem(la->la_hold);
311 		Free((caddr_t)la);
312 	}
313 }
314 
315 /*
316  * Broadcast an ARP request. Caller specifies:
317  *	- arp header source ip address
318  *	- arp header target ip address
319  *	- arp header source ethernet address
320  */
321 void
322 arprequest(ifp, sip, tip, enaddr)
323 	struct ifnet *ifp;
324 	u_int32_t *sip, *tip;
325 	u_int8_t *enaddr;
326 {
327 	struct mbuf *m;
328 	struct ether_header *eh;
329 	struct ether_arp *ea;
330 	struct sockaddr sa;
331 
332 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
333 		return;
334 	m->m_len = sizeof(*ea);
335 	m->m_pkthdr.len = sizeof(*ea);
336 	MH_ALIGN(m, sizeof(*ea));
337 	ea = mtod(m, struct ether_arp *);
338 	eh = (struct ether_header *)sa.sa_data;
339 	bzero((caddr_t)ea, sizeof (*ea));
340 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
341 	    sizeof(eh->ether_dhost));
342 	eh->ether_type = htons(ETHERTYPE_ARP);	/* if_output will not swap */
343 	ea->arp_hrd = htons(ARPHRD_ETHER);
344 	ea->arp_pro = htons(ETHERTYPE_IP);
345 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
346 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
347 	ea->arp_op = htons(ARPOP_REQUEST);
348 	bcopy((caddr_t)enaddr, (caddr_t)eh->ether_shost,
349 	      sizeof(eh->ether_shost));
350 	bcopy((caddr_t)enaddr, (caddr_t)ea->arp_sha, sizeof(ea->arp_sha));
351 	bcopy((caddr_t)sip, (caddr_t)ea->arp_spa, sizeof(ea->arp_spa));
352 	bcopy((caddr_t)tip, (caddr_t)ea->arp_tpa, sizeof(ea->arp_tpa));
353 	sa.sa_family = pseudo_AF_HDRCMPLT;
354 	sa.sa_len = sizeof(sa);
355 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
356 }
357 
358 /*
359  * Resolve an IP address into an ethernet address.  If success,
360  * desten is filled in.  If there is no entry in arptab,
361  * set one up and broadcast a request for the IP address.
362  * Hold onto this mbuf and resend it once the address
363  * is finally resolved.  A return value of 1 indicates
364  * that desten has been filled in and the packet should be sent
365  * normally; a 0 return indicates that the packet has been
366  * taken over here, either now or for later transmission.
367  */
368 int
369 arpresolve(ac, rt, m, dst, desten)
370 	struct arpcom *ac;
371 	struct rtentry *rt;
372 	struct mbuf *m;
373 	struct sockaddr *dst;
374 	u_char *desten;
375 {
376 	struct llinfo_arp *la;
377 	struct sockaddr_dl *sdl;
378 
379 	if (m->m_flags & M_BCAST) {	/* broadcast */
380 		bcopy((caddr_t)etherbroadcastaddr, (caddr_t)desten,
381 		    sizeof(etherbroadcastaddr));
382 		return (1);
383 	}
384 	if (m->m_flags & M_MCAST) {	/* multicast */
385 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
386 		return (1);
387 	}
388 	if (rt) {
389 		la = (struct llinfo_arp *)rt->rt_llinfo;
390 		if (la == NULL)
391 			log(LOG_DEBUG, "arpresolve: %s: route without link "
392 			    "local address\n", inet_ntoa(SIN(dst)->sin_addr));
393 	} else {
394 		if ((la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0)) != NULL)
395 			rt = la->la_rt;
396 		else
397 			log(LOG_DEBUG,
398 			    "arpresolve: %s: can't allocate llinfo\n",
399 			    inet_ntoa(SIN(dst)->sin_addr));
400 	}
401 	if (la == 0 || rt == 0) {
402 		m_freem(m);
403 		return (0);
404 	}
405 	sdl = SDL(rt->rt_gateway);
406 	/*
407 	 * Check the address family and length is valid, the address
408 	 * is resolved; otherwise, try to resolve.
409 	 */
410 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
411 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
412 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
413 		return 1;
414 	}
415 	if (((struct ifnet *)ac)->if_flags & IFF_NOARP)
416 		return 0;
417 
418 	/*
419 	 * There is an arptab entry, but no ethernet address
420 	 * response yet.  Replace the held mbuf with this
421 	 * latest one.
422 	 */
423 	if (la->la_hold)
424 		m_freem(la->la_hold);
425 	la->la_hold = m;
426 	/*
427 	 * Re-send the ARP request when appropriate.
428 	 */
429 #ifdef	DIAGNOSTIC
430 	if (rt->rt_expire == 0) {
431 		/* This should never happen. (Should it? -gwr) */
432 		printf("arpresolve: unresolved and rt_expire == 0\n");
433 		/* Set expiration time to now (expired). */
434 		rt->rt_expire = time_second;
435 	}
436 #endif
437 	if (rt->rt_expire) {
438 		rt->rt_flags &= ~RTF_REJECT;
439 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
440 			rt->rt_expire = time_second;
441 			if (la->la_asked++ < arp_maxtries)
442 				arprequest(&ac->ac_if,
443 				    &(SIN(rt->rt_ifa->ifa_addr)->sin_addr.s_addr),
444 				    &(SIN(dst)->sin_addr.s_addr),
445 #if NCARP > 0
446 				    (rt->rt_ifp->if_type == IFT_CARP) ?
447 					((struct arpcom *) rt->rt_ifp->if_softc
448 					)->ac_enaddr :
449 #endif
450 				    ac->ac_enaddr);
451 			else {
452 				rt->rt_flags |= RTF_REJECT;
453 				rt->rt_expire += arpt_down;
454 				la->la_asked = 0;
455 			}
456 		}
457 	}
458 	return (0);
459 }
460 
461 /*
462  * Common length and type checks are done here,
463  * then the protocol-specific routine is called.
464  */
465 void
466 arpintr()
467 {
468 	struct mbuf *m;
469 	struct arphdr *ar;
470 	int s, len;
471 
472 	for (;;) {
473 		s = splnet();
474 		IF_DEQUEUE(&arpintrq, m);
475 		splx(s);
476 		if (m == NULL)
477 			break;
478 #ifdef DIAGNOSTIC
479 		if ((m->m_flags & M_PKTHDR) == 0)
480 			panic("arpintr");
481 #endif
482 
483 		len = sizeof(struct arphdr);
484 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
485 			continue;
486 
487 		ar = mtod(m, struct arphdr *);
488 		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) {
489 			m_freem(m);
490 			continue;
491 		}
492 
493 		len += 2 * (ar->ar_hln + ar->ar_pln);
494 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
495 			continue;
496 
497 		switch (ntohs(ar->ar_pro)) {
498 		case ETHERTYPE_IP:
499 		case ETHERTYPE_IPTRAILERS:
500 			in_arpinput(m);
501 			continue;
502 		}
503 		m_freem(m);
504 	}
505 }
506 
507 /*
508  * ARP for Internet protocols on Ethernet.
509  * Algorithm is that given in RFC 826.
510  * In addition, a sanity check is performed on the sender
511  * protocol address, to catch impersonators.
512  * We no longer handle negotiations for use of trailer protocol:
513  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
514  * along with IP replies if we wanted trailers sent to us,
515  * and also sent them in response to IP replies.
516  * This allowed either end to announce the desire to receive
517  * trailer packets.
518  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
519  * but formerly didn't normally send requests.
520  */
521 void
522 in_arpinput(m)
523 	struct mbuf *m;
524 {
525 	struct ether_arp *ea;
526 	struct arpcom *ac = (struct arpcom *)m->m_pkthdr.rcvif;
527 	struct ether_header *eh;
528 	struct llinfo_arp *la = 0;
529 	struct rtentry *rt;
530 	struct in_ifaddr *ia;
531 #if NBRIDGE > 0
532 	struct in_ifaddr *bridge_ia = NULL;
533 #endif
534 	struct sockaddr_dl *sdl;
535 	struct sockaddr sa;
536 	struct in_addr isaddr, itaddr, myaddr;
537 	u_int8_t *enaddr = NULL;
538 #if NCARP > 0
539 	u_int8_t *ether_shost = NULL;
540 #endif
541 	int op;
542 
543 	ea = mtod(m, struct ether_arp *);
544 	op = ntohs(ea->arp_op);
545 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
546 		goto out;
547 #if notyet
548 	if ((op == ARPOP_REPLY) && (m->m_flags & (M_BCAST|M_MCAST))) {
549 		log(LOG_ERR,
550 		    "arp: received reply to broadcast or multicast address\n");
551 		goto out;
552 	}
553 #endif
554 
555 	bcopy((caddr_t)ea->arp_tpa, (caddr_t)&itaddr, sizeof(itaddr));
556 	bcopy((caddr_t)ea->arp_spa, (caddr_t)&isaddr, sizeof(isaddr));
557 
558 	TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
559 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
560 			continue;
561 
562 #if NCARP > 0
563 		if (ia->ia_ifp->if_type == IFT_CARP &&
564 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
565 		    (IFF_UP|IFF_RUNNING))) {
566 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
567 			    (op == ARPOP_REPLY  ||
568 			    carp_iamatch(ia, ea->arp_sha,
569 			    &enaddr, &ether_shost)))
570 				break;
571 		} else
572 #endif
573 			if (ia->ia_ifp == m->m_pkthdr.rcvif)
574 				break;
575 #if NBRIDGE > 0
576 		/*
577 		 * If the interface we received the packet on
578 		 * is part of a bridge, check to see if we need
579 		 * to "bridge" the packet to ourselves at this
580 		 * layer.  Note we still prefer a perfect match,
581 		 * but allow this weaker match if necessary.
582 		 */
583 		if (m->m_pkthdr.rcvif->if_bridge != NULL) {
584 			if (m->m_pkthdr.rcvif->if_bridge ==
585 			    ia->ia_ifp->if_bridge)
586 				bridge_ia = ia;
587 #if NCARP > 0
588 			else if (ia->ia_ifp->if_carpdev != NULL &&
589 			    m->m_pkthdr.rcvif->if_bridge ==
590 			    ia->ia_ifp->if_carpdev->if_bridge &&
591 			    carp_iamatch(ia, ea->arp_sha,
592 			    &enaddr, &ether_shost))
593 				bridge_ia = ia;
594 #endif
595 		}
596 #endif
597 	}
598 
599 #if NBRIDGE > 0
600 	if (ia == NULL && bridge_ia != NULL) {
601 		ia = bridge_ia;
602 		ac = (struct arpcom *)bridge_ia->ia_ifp;
603 	}
604 #endif
605 
606 	if (ia == NULL) {
607 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
608 			if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
609 				continue;
610 			if (ia->ia_ifp == m->m_pkthdr.rcvif)
611 				break;
612 		}
613 	}
614 
615 	if (ia == NULL && m->m_pkthdr.rcvif->if_type != IFT_CARP) {
616 		struct ifaddr *ifa;
617 
618 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
619 			if (ifa->ifa_addr->sa_family == AF_INET)
620 				break;
621 		}
622 		if (ifa)
623 			ia = (struct in_ifaddr *)ifa;
624 	}
625 
626 	if (ia == NULL)
627 		goto out;
628 
629 	if (!enaddr)
630 		enaddr = ac->ac_enaddr;
631 	myaddr = ia->ia_addr.sin_addr;
632 
633 	if (!bcmp((caddr_t)ea->arp_sha, enaddr, sizeof (ea->arp_sha)))
634 		goto out;	/* it's from me, ignore it. */
635 	if (ETHER_IS_MULTICAST (&ea->arp_sha[0]))
636 		if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)etherbroadcastaddr,
637 		    sizeof (ea->arp_sha))) {
638 			log(LOG_ERR, "arp: ether address is broadcast for "
639 			    "IP address %s!\n", inet_ntoa(isaddr));
640 			goto out;
641 		}
642 	if (myaddr.s_addr && isaddr.s_addr == myaddr.s_addr) {
643 		log(LOG_ERR,
644 		   "duplicate IP address %s sent from ethernet address %s\n",
645 		   inet_ntoa(isaddr), ether_sprintf(ea->arp_sha));
646 		itaddr = myaddr;
647 		goto reply;
648 	}
649 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
650 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
651 		if (sdl->sdl_alen) {
652 		    if (bcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
653 			if (rt->rt_flags & RTF_PERMANENT_ARP) {
654 				log(LOG_WARNING,
655 				   "arp: attempt to overwrite permanent "
656 				   "entry for %s by %s on %s\n",
657 				   inet_ntoa(isaddr),
658 				   ether_sprintf(ea->arp_sha),
659 				   ac->ac_if.if_xname);
660 				goto out;
661 			} else if (rt->rt_ifp != &ac->ac_if) {
662 				log(LOG_WARNING,
663 				   "arp: attempt to overwrite entry for %s "
664 				   "on %s by %s on %s\n",
665 				   inet_ntoa(isaddr), rt->rt_ifp->if_xname,
666 				   ether_sprintf(ea->arp_sha),
667 				   ac->ac_if.if_xname);
668 				goto out;
669 			} else {
670 				log(LOG_INFO,
671 				   "arp info overwritten for %s by %s on %s\n",
672 				   inet_ntoa(isaddr),
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 && !(ac->ac_if.if_bridge &&
679 		    (rt->rt_ifp->if_bridge == ac->ac_if.if_bridge)) &&
680 		    !(rt->rt_ifp->if_type == IFT_CARP &&
681 		    rt->rt_ifp->if_carpdev == &ac->ac_if) &&
682 		    !(ac->ac_if.if_type == IFT_CARP &&
683 		    ac->ac_if.if_carpdev == rt->rt_ifp)) {
684 		    log(LOG_WARNING,
685 			"arp: attempt to add entry for %s "
686 			"on %s by %s on %s\n",
687 			inet_ntoa(isaddr), rt->rt_ifp->if_xname,
688 			ether_sprintf(ea->arp_sha),
689 			ac->ac_if.if_xname);
690 		    goto out;
691 		}
692 		bcopy(ea->arp_sha, LLADDR(sdl),
693 		    sdl->sdl_alen = sizeof(ea->arp_sha));
694 		if (rt->rt_expire)
695 			rt->rt_expire = time_second + arpt_keep;
696 		rt->rt_flags &= ~RTF_REJECT;
697 		la->la_asked = 0;
698 		if (la->la_hold) {
699 			struct mbuf *n = la->la_hold;
700 			la->la_hold = NULL;
701 			(*ac->ac_if.if_output)(&ac->ac_if, n, rt_key(rt), rt);
702 			if (la->la_hold == n) {
703 				/* n is back in la_hold. Discard. */
704 				m_freem(la->la_hold);
705 				la->la_hold = NULL;
706 			}
707 		}
708 	}
709 reply:
710 	if (op != ARPOP_REQUEST) {
711 	out:
712 		m_freem(m);
713 		return;
714 	}
715 	if (itaddr.s_addr == myaddr.s_addr) {
716 		/* I am the target */
717 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
718 		bcopy(enaddr, ea->arp_sha, sizeof(ea->arp_sha));
719 	} else {
720 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
721 		if (la == 0)
722 			goto out;
723 		rt = la->la_rt;
724 		if (rt->rt_ifp->if_type == IFT_CARP &&
725 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
726 			goto out;
727 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
728 		sdl = SDL(rt->rt_gateway);
729 		bcopy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
730 	}
731 
732 	bcopy(ea->arp_spa, ea->arp_tpa, sizeof(ea->arp_spa));
733 	bcopy(&itaddr, ea->arp_spa, sizeof(ea->arp_spa));
734 	ea->arp_op = htons(ARPOP_REPLY);
735 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
736 	eh = (struct ether_header *)sa.sa_data;
737 	bcopy(ea->arp_tha, eh->ether_dhost, sizeof(eh->ether_dhost));
738 #if NCARP > 0
739 	if (ether_shost)
740 		enaddr = ether_shost;
741 #endif
742 	bcopy(enaddr, eh->ether_shost, sizeof(eh->ether_shost));
743 
744 	eh->ether_type = htons(ETHERTYPE_ARP);
745 	sa.sa_family = pseudo_AF_HDRCMPLT;
746 	sa.sa_len = sizeof(sa);
747 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
748 	return;
749 }
750 
751 /*
752  * Free an arp entry.
753  */
754 void
755 arptfree(la)
756 	struct llinfo_arp *la;
757 {
758 	struct rtentry *rt = la->la_rt;
759 	struct sockaddr_dl *sdl;
760 	struct rt_addrinfo info;
761 
762 	if (rt == 0)
763 		panic("arptfree");
764 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
765 	    sdl->sdl_family == AF_LINK) {
766 		sdl->sdl_alen = 0;
767 		la->la_asked = 0;
768 		rt->rt_flags &= ~RTF_REJECT;
769 		return;
770 	}
771 	bzero(&info, sizeof(info));
772 	info.rti_info[RTAX_DST] = rt_key(rt);
773 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
774 
775 	rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL, 0);
776 }
777 
778 /*
779  * Lookup or enter a new address in arptab.
780  */
781 struct llinfo_arp *
782 arplookup(addr, create, proxy)
783 	u_int32_t addr;
784 	int create, proxy;
785 {
786 	struct rtentry *rt;
787 	static struct sockaddr_inarp sin;
788 
789 	sin.sin_len = sizeof(sin);
790 	sin.sin_family = AF_INET;
791 	sin.sin_addr.s_addr = addr;
792 	sin.sin_other = proxy ? SIN_PROXY : 0;
793 	rt = rtalloc1(sintosa(&sin), create, 0);
794 	if (rt == 0)
795 		return (0);
796 	rt->rt_refcnt--;
797 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
798 	    rt->rt_gateway->sa_family != AF_LINK) {
799 		if (create) {
800 			if (rt->rt_refcnt <= 0 &&
801 			    (rt->rt_flags & RTF_CLONED) != 0) {
802 				struct rt_addrinfo info;
803 
804 				bzero(&info, sizeof(info));
805 				info.rti_info[RTAX_DST] = rt_key(rt);
806 				info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
807 				info.rti_info[RTAX_NETMASK] = rt_mask(rt);
808 
809 				rtrequest1(RTM_DELETE, &info, rt->rt_priority,
810 				    NULL, 0);
811 			}
812 		}
813 		return (0);
814 	}
815 	return ((struct llinfo_arp *)rt->rt_llinfo);
816 }
817 
818 int
819 arpioctl(cmd, data)
820 	u_long cmd;
821 	caddr_t data;
822 {
823 
824 	return (EOPNOTSUPP);
825 }
826 
827 void
828 arp_ifinit(ac, ifa)
829 	struct arpcom *ac;
830 	struct ifaddr *ifa;
831 {
832 
833 	/* Warn the user if another station has this IP address. */
834 	arprequest(&ac->ac_if,
835 	    &(IA_SIN(ifa)->sin_addr.s_addr),
836 	    &(IA_SIN(ifa)->sin_addr.s_addr),
837 	    ac->ac_enaddr);
838 	ifa->ifa_rtrequest = arp_rtrequest;
839 	ifa->ifa_flags |= RTF_CLONING;
840 }
841 
842 /*
843  * Called from Ethernet interrupt handlers
844  * when ether packet type ETHERTYPE_REVARP
845  * is received.  Common length and type checks are done here,
846  * then the protocol-specific routine is called.
847  */
848 void
849 revarpinput(m)
850 	struct mbuf *m;
851 {
852 	struct arphdr *ar;
853 
854 	if (m->m_len < sizeof(struct arphdr))
855 		goto out;
856 	ar = mtod(m, struct arphdr *);
857 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
858 		goto out;
859 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
860 		goto out;
861 	switch (ntohs(ar->ar_pro)) {
862 
863 	case ETHERTYPE_IP:
864 	case ETHERTYPE_IPTRAILERS:
865 		in_revarpinput(m);
866 		return;
867 
868 	default:
869 		break;
870 	}
871 out:
872 	m_freem(m);
873 }
874 
875 /*
876  * RARP for Internet protocols on Ethernet.
877  * Algorithm is that given in RFC 903.
878  * We are only using for bootstrap purposes to get an ip address for one of
879  * our interfaces.  Thus we support no user-interface.
880  *
881  * Since the contents of the RARP reply are specific to the interface that
882  * sent the request, this code must ensure that they are properly associated.
883  *
884  * Note: also supports ARP via RARP packets, per the RFC.
885  */
886 void
887 in_revarpinput(m)
888 	struct mbuf *m;
889 {
890 	struct ifnet *ifp;
891 	struct ether_arp *ar;
892 	int op;
893 
894 	ar = mtod(m, struct ether_arp *);
895 	op = ntohs(ar->arp_op);
896 	switch (op) {
897 	case ARPOP_REQUEST:
898 	case ARPOP_REPLY:	/* per RFC */
899 		in_arpinput(m);
900 		return;
901 	case ARPOP_REVREPLY:
902 		break;
903 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
904 	default:
905 		goto out;
906 	}
907 	if (!revarp_in_progress)
908 		goto out;
909 	ifp = m->m_pkthdr.rcvif;
910 	if (ifp != myip_ifp) /* !same interface */
911 		goto out;
912 	if (myip_initialized)
913 		goto wake;
914 	if (bcmp(ar->arp_tha, ((struct arpcom *)ifp)->ac_enaddr,
915 	    sizeof(ar->arp_tha)))
916 		goto out;
917 	bcopy((caddr_t)ar->arp_spa, (caddr_t)&srv_ip, sizeof(srv_ip));
918 	bcopy((caddr_t)ar->arp_tpa, (caddr_t)&myip, sizeof(myip));
919 	myip_initialized = 1;
920 wake:	/* Do wakeup every time in case it was missed. */
921 	wakeup((caddr_t)&myip);
922 
923 out:
924 	m_freem(m);
925 }
926 
927 /*
928  * Send a RARP request for the ip address of the specified interface.
929  * The request should be RFC 903-compliant.
930  */
931 void
932 revarprequest(ifp)
933 	struct ifnet *ifp;
934 {
935 	struct sockaddr sa;
936 	struct mbuf *m;
937 	struct ether_header *eh;
938 	struct ether_arp *ea;
939 	struct arpcom *ac = (struct arpcom *)ifp;
940 
941 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
942 		return;
943 	m->m_len = sizeof(*ea);
944 	m->m_pkthdr.len = sizeof(*ea);
945 	MH_ALIGN(m, sizeof(*ea));
946 	ea = mtod(m, struct ether_arp *);
947 	eh = (struct ether_header *)sa.sa_data;
948 	bzero((caddr_t)ea, sizeof(*ea));
949 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
950 	    sizeof(eh->ether_dhost));
951 	eh->ether_type = htons(ETHERTYPE_REVARP);
952 	ea->arp_hrd = htons(ARPHRD_ETHER);
953 	ea->arp_pro = htons(ETHERTYPE_IP);
954 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
955 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
956 	ea->arp_op = htons(ARPOP_REVREQUEST);
957 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)eh->ether_shost,
958 	   sizeof(ea->arp_tha));
959 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_sha,
960 	   sizeof(ea->arp_sha));
961 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_tha,
962 	   sizeof(ea->arp_tha));
963 	sa.sa_family = pseudo_AF_HDRCMPLT;
964 	sa.sa_len = sizeof(sa);
965 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
966 }
967 
968 /*
969  * RARP for the ip address of the specified interface, but also
970  * save the ip address of the server that sent the answer.
971  * Timeout if no response is received.
972  */
973 int
974 revarpwhoarewe(ifp, serv_in, clnt_in)
975 	struct ifnet *ifp;
976 	struct in_addr *serv_in;
977 	struct in_addr *clnt_in;
978 {
979 	int result, count = 20;
980 
981 	if (myip_initialized)
982 		return EIO;
983 
984 	myip_ifp = ifp;
985 	revarp_in_progress = 1;
986 	while (count--) {
987 		revarprequest(ifp);
988 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
989 		if (result != EWOULDBLOCK)
990 			break;
991 	}
992 	revarp_in_progress = 0;
993 	if (!myip_initialized)
994 		return ENETUNREACH;
995 
996 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
997 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
998 	return 0;
999 }
1000 
1001 /* For compatibility: only saves interface address. */
1002 int
1003 revarpwhoami(in, ifp)
1004 	struct in_addr *in;
1005 	struct ifnet *ifp;
1006 {
1007 	struct in_addr server;
1008 	return (revarpwhoarewe(ifp, &server, in));
1009 }
1010 
1011 
1012 #ifdef DDB
1013 
1014 #include <machine/db_machdep.h>
1015 #include <ddb/db_interface.h>
1016 #include <ddb/db_output.h>
1017 
1018 void
1019 db_print_sa(sa)
1020 	struct sockaddr *sa;
1021 {
1022 	int len;
1023 	u_char *p;
1024 
1025 	if (sa == 0) {
1026 		db_printf("[NULL]");
1027 		return;
1028 	}
1029 
1030 	p = (u_char *)sa;
1031 	len = sa->sa_len;
1032 	db_printf("[");
1033 	while (len > 0) {
1034 		db_printf("%d", *p);
1035 		p++;
1036 		len--;
1037 		if (len)
1038 			db_printf(",");
1039 	}
1040 	db_printf("]\n");
1041 }
1042 
1043 void
1044 db_print_ifa(ifa)
1045 	struct ifaddr *ifa;
1046 {
1047 	if (ifa == 0)
1048 		return;
1049 	db_printf("  ifa_addr=");
1050 	db_print_sa(ifa->ifa_addr);
1051 	db_printf("  ifa_dsta=");
1052 	db_print_sa(ifa->ifa_dstaddr);
1053 	db_printf("  ifa_mask=");
1054 	db_print_sa(ifa->ifa_netmask);
1055 	db_printf("  flags=0x%x, refcnt=%d, metric=%d\n",
1056 	    ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric);
1057 }
1058 
1059 void
1060 db_print_llinfo(li)
1061 	caddr_t li;
1062 {
1063 	struct llinfo_arp *la;
1064 
1065 	if (li == 0)
1066 		return;
1067 	la = (struct llinfo_arp *)li;
1068 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1069 	    la->la_rt, la->la_hold, la->la_asked);
1070 }
1071 
1072 /*
1073  * Function to pass to rn_walktree().
1074  * Return non-zero error to abort walk.
1075  */
1076 int
1077 db_show_radix_node(rn, w)
1078 	struct radix_node *rn;
1079 	void *w;
1080 {
1081 	struct rtentry *rt = (struct rtentry *)rn;
1082 
1083 	db_printf("rtentry=%p", rt);
1084 
1085 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1086 	    rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire);
1087 
1088 	db_printf(" key="); db_print_sa(rt_key(rt));
1089 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1090 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1091 
1092 	db_printf(" ifp=%p ", rt->rt_ifp);
1093 	if (rt->rt_ifp)
1094 		db_printf("(%s)", rt->rt_ifp->if_xname);
1095 	else
1096 		db_printf("(NULL)");
1097 
1098 	db_printf(" ifa=%p\n", rt->rt_ifa);
1099 	db_print_ifa(rt->rt_ifa);
1100 
1101 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1102 
1103 	db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo);
1104 	db_print_llinfo(rt->rt_llinfo);
1105 	return (0);
1106 }
1107 
1108 /*
1109  * Function to print all the route trees.
1110  * Use this from ddb:  "call db_show_arptab"
1111  */
1112 int
1113 db_show_arptab()
1114 {
1115 	struct radix_node_head *rnh;
1116 	rnh = rt_gettable(AF_INET, 0);
1117 	db_printf("Route tree for AF_INET\n");
1118 	if (rnh == NULL) {
1119 		db_printf(" (not initialized)\n");
1120 		return (0);
1121 	}
1122 	rn_walktree(rnh, db_show_radix_node, NULL);
1123 	return (0);
1124 }
1125 #endif
1126 #endif /* INET */
1127