xref: /openbsd-src/sys/netinet/if_ether.c (revision 0eea0d082377cb9c3ec583313dc4d52b7b6a4d6d)
1 /*	$OpenBSD: if_ether.c,v 1.54 2004/06/21 23:50:37 tholo 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);
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 = AF_UNSPEC;
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 	else {
391 		if ((la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0)) != NULL)
392 			rt = la->la_rt;
393 	}
394 	if (la == 0 || rt == 0) {
395 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo\n");
396 		m_freem(m);
397 		return (0);
398 	}
399 	sdl = SDL(rt->rt_gateway);
400 	/*
401 	 * Check the address family and length is valid, the address
402 	 * is resolved; otherwise, try to resolve.
403 	 */
404 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
405 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
406 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
407 		return 1;
408 	}
409 	if (((struct ifnet *)ac)->if_flags & IFF_NOARP)
410 		return 0;
411 
412 	/*
413 	 * There is an arptab entry, but no ethernet address
414 	 * response yet.  Replace the held mbuf with this
415 	 * latest one.
416 	 */
417 	if (la->la_hold)
418 		m_freem(la->la_hold);
419 	la->la_hold = m;
420 	/*
421 	 * Re-send the ARP request when appropriate.
422 	 */
423 #ifdef	DIAGNOSTIC
424 	if (rt->rt_expire == 0) {
425 		/* This should never happen. (Should it? -gwr) */
426 		printf("arpresolve: unresolved and rt_expire == 0\n");
427 		/* Set expiration time to now (expired). */
428 		rt->rt_expire = time_second;
429 	}
430 #endif
431 	if (rt->rt_expire) {
432 		rt->rt_flags &= ~RTF_REJECT;
433 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
434 			rt->rt_expire = time_second;
435 			if (la->la_asked++ < arp_maxtries)
436 				arprequest(&ac->ac_if,
437 				    &(SIN(rt->rt_ifa->ifa_addr)->sin_addr.s_addr),
438 				    &(SIN(dst)->sin_addr.s_addr),
439 				    ac->ac_enaddr);
440 			else {
441 				rt->rt_flags |= RTF_REJECT;
442 				rt->rt_expire += arpt_down;
443 				la->la_asked = 0;
444 			}
445 		}
446 	}
447 	return (0);
448 }
449 
450 /*
451  * Common length and type checks are done here,
452  * then the protocol-specific routine is called.
453  */
454 void
455 arpintr()
456 {
457 	struct mbuf *m;
458 	struct arphdr *ar;
459 	int s, len;
460 
461 	while (arpintrq.ifq_head) {
462 		s = splimp();
463 		IF_DEQUEUE(&arpintrq, m);
464 		splx(s);
465 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
466 			panic("arpintr");
467 
468 		len = sizeof(struct arphdr);
469 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
470 			continue;
471 
472 		ar = mtod(m, struct arphdr *);
473 		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) {
474 			m_freem(m);
475 			continue;
476 		}
477 
478 		len += 2 * (ar->ar_hln + ar->ar_pln);
479 		if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
480 			continue;
481 
482 		switch (ntohs(ar->ar_pro)) {
483 		case ETHERTYPE_IP:
484 		case ETHERTYPE_IPTRAILERS:
485 			in_arpinput(m);
486 			continue;
487 		}
488 		m_freem(m);
489 	}
490 }
491 
492 /*
493  * ARP for Internet protocols on Ethernet.
494  * Algorithm is that given in RFC 826.
495  * In addition, a sanity check is performed on the sender
496  * protocol address, to catch impersonators.
497  * We no longer handle negotiations for use of trailer protocol:
498  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
499  * along with IP replies if we wanted trailers sent to us,
500  * and also sent them in response to IP replies.
501  * This allowed either end to announce the desire to receive
502  * trailer packets.
503  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
504  * but formerly didn't normally send requests.
505  */
506 void
507 in_arpinput(m)
508 	struct mbuf *m;
509 {
510 	struct ether_arp *ea;
511 	struct arpcom *ac = (struct arpcom *)m->m_pkthdr.rcvif;
512 	struct ether_header *eh;
513 	struct llinfo_arp *la = 0;
514 	struct rtentry *rt;
515 	struct in_ifaddr *ia;
516 #if NBRIDGE > 0
517 	struct in_ifaddr *bridge_ia = NULL;
518 #endif
519 	struct sockaddr_dl *sdl;
520 	struct sockaddr sa;
521 	struct in_addr isaddr, itaddr, myaddr;
522 	u_int8_t *enaddr = NULL;
523 	int op;
524 
525 	ea = mtod(m, struct ether_arp *);
526 	op = ntohs(ea->arp_op);
527 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
528 		goto out;
529 #if notyet
530 	if ((op == ARPOP_REPLY) && (m->m_flags & (M_BCAST|M_MCAST))) {
531 		log(LOG_ERR,
532 		    "arp: received reply to broadcast or multicast address\n");
533 		goto out;
534 	}
535 #endif
536 
537 	bcopy((caddr_t)ea->arp_tpa, (caddr_t)&itaddr, sizeof(itaddr));
538 	bcopy((caddr_t)ea->arp_spa, (caddr_t)&isaddr, sizeof(isaddr));
539 
540 	TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
541 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
542 			continue;
543 
544 		if (ia->ia_ifp == m->m_pkthdr.rcvif)
545 			break;
546 #if NBRIDGE > 0
547 		/*
548 		 * If the interface we received the packet on
549 		 * is part of a bridge, check to see if we need
550 		 * to "bridge" the packet to ourselves at this
551 		 * layer.  Note we still prefer a perfect match,
552 		 * but allow this weaker match if necessary.
553 		 */
554 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
555 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
556 			bridge_ia = ia;
557 #endif
558 
559 #if NCARP > 0
560 		if (ac->ac_if.if_carp) {
561 			if (carp_iamatch(ac->ac_if.if_carp, ia,
562 			    &isaddr, &enaddr))
563 				break;
564 		}
565 #endif
566 	}
567 
568 #if NBRIDGE > 0
569 	if (ia == NULL && bridge_ia != NULL) {
570 		ia = bridge_ia;
571 		ac = (struct arpcom *)bridge_ia->ia_ifp;
572 	}
573 #endif
574 
575 	if (ia == NULL) {
576 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
577 			if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
578 				continue;
579 			if (ia->ia_ifp == m->m_pkthdr.rcvif)
580 				break;
581 		}
582 	}
583 
584 	if (ia == NULL) {
585 		struct ifaddr *ifa;
586 
587 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
588 			if (ifa->ifa_addr->sa_family == AF_INET)
589 				break;
590 		}
591 		if (ifa)
592 			ia = (struct in_ifaddr *)ifa;
593 	}
594 
595 	if (ia == NULL)
596 		goto out;
597 
598 	if (!enaddr)
599 		enaddr = ac->ac_enaddr;
600 	myaddr = ia->ia_addr.sin_addr;
601 
602 	if (!bcmp((caddr_t)ea->arp_sha, enaddr, sizeof (ea->arp_sha)))
603 		goto out;	/* it's from me, ignore it. */
604 	if (ETHER_IS_MULTICAST (&ea->arp_sha[0]))
605 		if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)etherbroadcastaddr,
606 		    sizeof (ea->arp_sha))) {
607 			log(LOG_ERR, "arp: ether address is broadcast for "
608 			    "IP address %s!\n", inet_ntoa(isaddr));
609 			goto out;
610 		}
611 	if (myaddr.s_addr && isaddr.s_addr == myaddr.s_addr) {
612 		log(LOG_ERR,
613 		   "duplicate IP address %s sent from ethernet address %s\n",
614 		   inet_ntoa(isaddr), ether_sprintf(ea->arp_sha));
615 		itaddr = myaddr;
616 		goto reply;
617 	}
618 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
619 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
620 		if (sdl->sdl_alen) {
621 		    if (bcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
622 		  	if (rt->rt_flags & RTF_PERMANENT_ARP) {
623 				log(LOG_WARNING,
624 				   "arp: attempt to overwrite permanent "
625 				   "entry for %s by %s on %s\n",
626 				   inet_ntoa(isaddr),
627 				   ether_sprintf(ea->arp_sha),
628 				   ac->ac_if.if_xname);
629 				goto out;
630 			} else if (rt->rt_ifp != &ac->ac_if) {
631 			        log(LOG_WARNING,
632 				   "arp: attempt to overwrite entry for %s "
633 				   "on %s by %s on %s\n",
634 				   inet_ntoa(isaddr), rt->rt_ifp->if_xname,
635 				   ether_sprintf(ea->arp_sha),
636 				   ac->ac_if.if_xname);
637 				goto out;
638 			} else {
639 				log(LOG_INFO,
640 				   "arp info overwritten for %s by %s on %s\n",
641 			    	   inet_ntoa(isaddr),
642 				   ether_sprintf(ea->arp_sha),
643 				   ac->ac_if.if_xname);
644 				rt->rt_expire = 1; /* no longer static */
645 			}
646 		    }
647 		} else if (rt->rt_ifp != &ac->ac_if && !(ac->ac_if.if_bridge &&
648 		    (rt->rt_ifp->if_bridge == ac->ac_if.if_bridge))) {
649 		    log(LOG_WARNING,
650 			"arp: attempt to add entry for %s "
651 			"on %s by %s on %s\n",
652 			inet_ntoa(isaddr), rt->rt_ifp->if_xname,
653 			ether_sprintf(ea->arp_sha),
654 			ac->ac_if.if_xname);
655 		    goto out;
656 		}
657 		bcopy(ea->arp_sha, LLADDR(sdl),
658 		    sdl->sdl_alen = sizeof(ea->arp_sha));
659 		if (rt->rt_expire)
660 			rt->rt_expire = time_second + arpt_keep;
661 		rt->rt_flags &= ~RTF_REJECT;
662 		la->la_asked = 0;
663 		if (la->la_hold) {
664 			(*ac->ac_if.if_output)(&ac->ac_if, la->la_hold,
665 				rt_key(rt), rt);
666 			la->la_hold = 0;
667 		}
668 	}
669 reply:
670 	if (op != ARPOP_REQUEST) {
671 	out:
672 		m_freem(m);
673 		return;
674 	}
675 	if (itaddr.s_addr == myaddr.s_addr) {
676 		/* I am the target */
677 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
678 		bcopy(enaddr, ea->arp_sha, sizeof(ea->arp_sha));
679 	} else {
680 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
681 		if (la == 0)
682 			goto out;
683 		rt = la->la_rt;
684 		bcopy(ea->arp_sha, ea->arp_tha, sizeof(ea->arp_sha));
685 		sdl = SDL(rt->rt_gateway);
686 		bcopy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
687 	}
688 
689 	bcopy(ea->arp_spa, ea->arp_tpa, sizeof(ea->arp_spa));
690 	bcopy(&itaddr, ea->arp_spa, sizeof(ea->arp_spa));
691 	ea->arp_op = htons(ARPOP_REPLY);
692 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
693 	eh = (struct ether_header *)sa.sa_data;
694 	bcopy(ea->arp_tha, eh->ether_dhost, sizeof(eh->ether_dhost));
695 	bcopy(enaddr, eh->ether_shost, sizeof(eh->ether_shost));
696 	eh->ether_type = htons(ETHERTYPE_ARP);
697 	sa.sa_family = AF_UNSPEC;
698 	sa.sa_len = sizeof(sa);
699 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
700 	return;
701 }
702 
703 /*
704  * Free an arp entry.
705  */
706 void
707 arptfree(la)
708 	struct llinfo_arp *la;
709 {
710 	struct rtentry *rt = la->la_rt;
711 	struct sockaddr_dl *sdl;
712 
713 	if (rt == 0)
714 		panic("arptfree");
715 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
716 	    sdl->sdl_family == AF_LINK) {
717 		sdl->sdl_alen = 0;
718 		la->la_asked = 0;
719 		rt->rt_flags &= ~RTF_REJECT;
720 		return;
721 	}
722 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
723 	    0, (struct rtentry **)0);
724 }
725 
726 /*
727  * Lookup or enter a new address in arptab.
728  */
729 struct llinfo_arp *
730 arplookup(addr, create, proxy)
731 	u_int32_t addr;
732 	int create, proxy;
733 {
734 	struct rtentry *rt;
735 	static struct sockaddr_inarp sin;
736 
737 	sin.sin_len = sizeof(sin);
738 	sin.sin_family = AF_INET;
739 	sin.sin_addr.s_addr = addr;
740 	sin.sin_other = proxy ? SIN_PROXY : 0;
741 	rt = rtalloc1(sintosa(&sin), create);
742 	if (rt == 0)
743 		return (0);
744 	rt->rt_refcnt--;
745 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
746 	    rt->rt_gateway->sa_family != AF_LINK) {
747 		if (create) {
748 			log(LOG_DEBUG,
749 			    "arplookup: unable to enter address for %s\n",
750 			    inet_ntoa(sin.sin_addr));
751 			if (rt->rt_refcnt <= 0 &&
752 			    (rt->rt_flags & RTF_CLONED) != 0) {
753 				rtrequest(RTM_DELETE,
754 				    (struct sockaddr *)rt_key(rt),
755 				    rt->rt_gateway, rt_mask(rt), rt->rt_flags,
756 				    0);
757 			}
758 		}
759 		return (0);
760 	}
761 	return ((struct llinfo_arp *)rt->rt_llinfo);
762 }
763 
764 int
765 arpioctl(cmd, data)
766 	u_long cmd;
767 	caddr_t data;
768 {
769 
770 	return (EOPNOTSUPP);
771 }
772 
773 void
774 arp_ifinit(ac, ifa)
775 	struct arpcom *ac;
776 	struct ifaddr *ifa;
777 {
778 
779 	/* Warn the user if another station has this IP address. */
780 	arprequest(&ac->ac_if,
781 	    &(IA_SIN(ifa)->sin_addr.s_addr),
782 	    &(IA_SIN(ifa)->sin_addr.s_addr),
783 	    ac->ac_enaddr);
784 	ifa->ifa_rtrequest = arp_rtrequest;
785 	ifa->ifa_flags |= RTF_CLONING;
786 }
787 
788 /*
789  * Called from Ethernet interrupt handlers
790  * when ether packet type ETHERTYPE_REVARP
791  * is received.  Common length and type checks are done here,
792  * then the protocol-specific routine is called.
793  */
794 void
795 revarpinput(m)
796 	struct mbuf *m;
797 {
798 	struct arphdr *ar;
799 
800 	if (m->m_len < sizeof(struct arphdr))
801 		goto out;
802 	ar = mtod(m, struct arphdr *);
803 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
804 		goto out;
805 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
806 		goto out;
807 	switch (ntohs(ar->ar_pro)) {
808 
809 	case ETHERTYPE_IP:
810 	case ETHERTYPE_IPTRAILERS:
811 		in_revarpinput(m);
812 		return;
813 
814 	default:
815 		break;
816 	}
817 out:
818 	m_freem(m);
819 }
820 
821 /*
822  * RARP for Internet protocols on Ethernet.
823  * Algorithm is that given in RFC 903.
824  * We are only using for bootstrap purposes to get an ip address for one of
825  * our interfaces.  Thus we support no user-interface.
826  *
827  * Since the contents of the RARP reply are specific to the interface that
828  * sent the request, this code must ensure that they are properly associated.
829  *
830  * Note: also supports ARP via RARP packets, per the RFC.
831  */
832 void
833 in_revarpinput(m)
834 	struct mbuf *m;
835 {
836 	struct ifnet *ifp;
837 	struct ether_arp *ar;
838 	int op;
839 
840 	ar = mtod(m, struct ether_arp *);
841 	op = ntohs(ar->arp_op);
842 	switch (op) {
843 	case ARPOP_REQUEST:
844 	case ARPOP_REPLY:	/* per RFC */
845 		in_arpinput(m);
846 		return;
847 	case ARPOP_REVREPLY:
848 		break;
849 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
850 	default:
851 		goto out;
852 	}
853 	if (!revarp_in_progress)
854 		goto out;
855 	ifp = m->m_pkthdr.rcvif;
856 	if (ifp != myip_ifp) /* !same interface */
857 		goto out;
858 	if (myip_initialized)
859 		goto wake;
860 	if (bcmp(ar->arp_tha, ((struct arpcom *)ifp)->ac_enaddr,
861 	    sizeof(ar->arp_tha)))
862 		goto out;
863 	bcopy((caddr_t)ar->arp_spa, (caddr_t)&srv_ip, sizeof(srv_ip));
864 	bcopy((caddr_t)ar->arp_tpa, (caddr_t)&myip, sizeof(myip));
865 	myip_initialized = 1;
866 wake:	/* Do wakeup every time in case it was missed. */
867 	wakeup((caddr_t)&myip);
868 
869 out:
870 	m_freem(m);
871 }
872 
873 /*
874  * Send a RARP request for the ip address of the specified interface.
875  * The request should be RFC 903-compliant.
876  */
877 void
878 revarprequest(ifp)
879 	struct ifnet *ifp;
880 {
881 	struct sockaddr sa;
882 	struct mbuf *m;
883 	struct ether_header *eh;
884 	struct ether_arp *ea;
885 	struct arpcom *ac = (struct arpcom *)ifp;
886 
887 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
888 		return;
889 	m->m_len = sizeof(*ea);
890 	m->m_pkthdr.len = sizeof(*ea);
891 	MH_ALIGN(m, sizeof(*ea));
892 	ea = mtod(m, struct ether_arp *);
893 	eh = (struct ether_header *)sa.sa_data;
894 	bzero((caddr_t)ea, sizeof(*ea));
895 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
896 	    sizeof(eh->ether_dhost));
897 	eh->ether_type = htons(ETHERTYPE_REVARP);
898 	ea->arp_hrd = htons(ARPHRD_ETHER);
899 	ea->arp_pro = htons(ETHERTYPE_IP);
900 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
901 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
902 	ea->arp_op = htons(ARPOP_REVREQUEST);
903 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)eh->ether_shost,
904 	   sizeof(ea->arp_tha));
905 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_sha,
906 	   sizeof(ea->arp_sha));
907 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_tha,
908 	   sizeof(ea->arp_tha));
909 	sa.sa_family = AF_UNSPEC;
910 	sa.sa_len = sizeof(sa);
911 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
912 }
913 
914 /*
915  * RARP for the ip address of the specified interface, but also
916  * save the ip address of the server that sent the answer.
917  * Timeout if no response is received.
918  */
919 int
920 revarpwhoarewe(ifp, serv_in, clnt_in)
921 	struct ifnet *ifp;
922 	struct in_addr *serv_in;
923 	struct in_addr *clnt_in;
924 {
925 	int result, count = 20;
926 
927 	if (myip_initialized)
928 		return EIO;
929 
930 	myip_ifp = ifp;
931 	revarp_in_progress = 1;
932 	while (count--) {
933 		revarprequest(ifp);
934 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
935 		if (result != EWOULDBLOCK)
936 			break;
937 	}
938 	revarp_in_progress = 0;
939 	if (!myip_initialized)
940 		return ENETUNREACH;
941 
942 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
943 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
944 	return 0;
945 }
946 
947 /* For compatibility: only saves interface address. */
948 int
949 revarpwhoami(in, ifp)
950 	struct in_addr *in;
951 	struct ifnet *ifp;
952 {
953 	struct in_addr server;
954 	return (revarpwhoarewe(ifp, &server, in));
955 }
956 
957 
958 #ifdef DDB
959 
960 #include <machine/db_machdep.h>
961 #include <ddb/db_interface.h>
962 #include <ddb/db_output.h>
963 
964 void
965 db_print_sa(sa)
966 	struct sockaddr *sa;
967 {
968 	int len;
969 	u_char *p;
970 
971 	if (sa == 0) {
972 		db_printf("[NULL]");
973 		return;
974 	}
975 
976 	p = (u_char *)sa;
977 	len = sa->sa_len;
978 	db_printf("[");
979 	while (len > 0) {
980 		db_printf("%d", *p);
981 		p++;
982 		len--;
983 		if (len)
984 			db_printf(",");
985 	}
986 	db_printf("]\n");
987 }
988 
989 void
990 db_print_ifa(ifa)
991 	struct ifaddr *ifa;
992 {
993 	if (ifa == 0)
994 		return;
995 	db_printf("  ifa_addr=");
996 	db_print_sa(ifa->ifa_addr);
997 	db_printf("  ifa_dsta=");
998 	db_print_sa(ifa->ifa_dstaddr);
999 	db_printf("  ifa_mask=");
1000 	db_print_sa(ifa->ifa_netmask);
1001 	db_printf("  flags=0x%x, refcnt=%d, metric=%d\n",
1002 	    ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric);
1003 }
1004 
1005 void
1006 db_print_llinfo(li)
1007 	caddr_t li;
1008 {
1009 	struct llinfo_arp *la;
1010 
1011 	if (li == 0)
1012 		return;
1013 	la = (struct llinfo_arp *)li;
1014 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1015 	    la->la_rt, la->la_hold, la->la_asked);
1016 }
1017 
1018 /*
1019  * Function to pass to rn_walktree().
1020  * Return non-zero error to abort walk.
1021  */
1022 int
1023 db_show_radix_node(rn, w)
1024 	struct radix_node *rn;
1025 	void *w;
1026 {
1027 	struct rtentry *rt = (struct rtentry *)rn;
1028 
1029 	db_printf("rtentry=%p", rt);
1030 
1031 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1032 	    rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire);
1033 
1034 	db_printf(" key="); db_print_sa(rt_key(rt));
1035 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1036 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1037 
1038 	db_printf(" ifp=%p ", rt->rt_ifp);
1039 	if (rt->rt_ifp)
1040 		db_printf("(%s)", rt->rt_ifp->if_xname);
1041 	else
1042 		db_printf("(NULL)");
1043 
1044 	db_printf(" ifa=%p\n", rt->rt_ifa);
1045 	db_print_ifa(rt->rt_ifa);
1046 
1047 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1048 
1049 	db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo);
1050 	db_print_llinfo(rt->rt_llinfo);
1051 	return (0);
1052 }
1053 
1054 /*
1055  * Function to print all the route trees.
1056  * Use this from ddb:  "call db_show_arptab"
1057  */
1058 int
1059 db_show_arptab()
1060 {
1061 	struct radix_node_head *rnh;
1062 	rnh = rt_tables[AF_INET];
1063 	db_printf("Route tree for AF_INET\n");
1064 	if (rnh == NULL) {
1065 		db_printf(" (not initialized)\n");
1066 		return (0);
1067 	}
1068 	rn_walktree(rnh, db_show_radix_node, NULL);
1069 	return (0);
1070 }
1071 #endif
1072 #endif /* INET */
1073