xref: /freebsd/sys/netinet/if_ether.c (revision d2387d42b8da231a5b95cbc313825fb2aadf26f6)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
34  * $FreeBSD$
35  */
36 
37 /*
38  * Ethernet address resolution protocol.
39  * TODO:
40  *	add "inuse/lock" bit (or ref. count) along with valid bit
41  */
42 
43 #include "opt_inet.h"
44 #include "opt_bdg.h"
45 #include "opt_mac.h"
46 
47 #include <sys/param.h>
48 #include <sys/kernel.h>
49 #include <sys/queue.h>
50 #include <sys/sysctl.h>
51 #include <sys/systm.h>
52 #include <sys/mac.h>
53 #include <sys/mbuf.h>
54 #include <sys/malloc.h>
55 #include <sys/socket.h>
56 #include <sys/syslog.h>
57 
58 #include <net/if.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/route.h>
62 #include <net/netisr.h>
63 #include <net/if_llc.h>
64 #ifdef BRIDGE
65 #include <net/ethernet.h>
66 #include <net/bridge.h>
67 #endif
68 
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/if_ether.h>
72 
73 #include <net/if_arc.h>
74 #include <net/iso88025.h>
75 
76 #define SIN(s) ((struct sockaddr_in *)s)
77 #define SDL(s) ((struct sockaddr_dl *)s)
78 
79 SYSCTL_DECL(_net_link_ether);
80 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 
82 /* timer values */
83 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
84 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
85 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
86 
87 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
88 	   &arpt_prune, 0, "");
89 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
90 	   &arpt_keep, 0, "");
91 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
92 	   &arpt_down, 0, "");
93 
94 #define	rt_expire rt_rmx.rmx_expire
95 
96 struct llinfo_arp {
97 	LIST_ENTRY(llinfo_arp) la_le;
98 	struct	rtentry *la_rt;
99 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
100 	u_short	la_preempt;	/* countdown for pre-expiry arps */
101 	u_short	la_asked;	/* #times we QUERIED following expiration */
102 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
103 };
104 
105 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
106 
107 static struct	ifqueue arpintrq;
108 static int	arp_allocated;
109 static int	arpinit_done;
110 
111 static int	arp_maxtries = 5;
112 static int	useloopback = 1; /* use loopback interface for local traffic */
113 static int	arp_proxyall = 0;
114 static struct callout arp_callout;
115 
116 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
117 	   &arp_maxtries, 0, "");
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
119 	   &useloopback, 0, "");
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
121 	   &arp_proxyall, 0, "");
122 
123 static void	arp_init(void);
124 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
125 static void	arprequest(struct ifnet *,
126 			struct in_addr *, struct in_addr *, u_char *);
127 static void	arpintr(struct mbuf *);
128 static void	arptfree(struct llinfo_arp *);
129 static void	arptimer(void *);
130 static struct llinfo_arp
131 		*arplookup(u_long, int, int);
132 #ifdef INET
133 static void	in_arpinput(struct mbuf *);
134 #endif
135 
136 /*
137  * Timeout routine.  Age arp_tab entries periodically.
138  */
139 /* ARGSUSED */
140 static void
141 arptimer(ignored_arg)
142 	void *ignored_arg;
143 {
144 	struct llinfo_arp *la, *ola;
145 
146 	RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]);
147 	la = LIST_FIRST(&llinfo_arp);
148 	while (la != NULL) {
149 		struct rtentry *rt = la->la_rt;
150 		ola = la;
151 		la = LIST_NEXT(la, la_le);
152 		if (rt->rt_expire && rt->rt_expire <= time_second)
153 			arptfree(ola);		/* timer has expired, clear */
154 	}
155 	RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]);
156 
157 	callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL);
158 }
159 
160 /*
161  * Parallel to llc_rtrequest.
162  */
163 static void
164 arp_rtrequest(req, rt, info)
165 	int req;
166 	struct rtentry *rt;
167 	struct rt_addrinfo *info;
168 {
169 	struct sockaddr *gate;
170 	struct llinfo_arp *la;
171 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
172 
173 	RT_LOCK_ASSERT(rt);
174 
175 	if (!arpinit_done) {
176 		arpinit_done = 1;
177 		callout_reset(&arp_callout, hz, arptimer, NULL);
178 	}
179 	if (rt->rt_flags & RTF_GATEWAY)
180 		return;
181 	gate = rt->rt_gateway;
182 	la = (struct llinfo_arp *)rt->rt_llinfo;
183 	switch (req) {
184 
185 	case RTM_ADD:
186 		/*
187 		 * XXX: If this is a manually added route to interface
188 		 * such as older version of routed or gated might provide,
189 		 * restore cloning bit.
190 		 */
191 		if ((rt->rt_flags & RTF_HOST) == 0 &&
192 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
193 			rt->rt_flags |= RTF_CLONING;
194 		if (rt->rt_flags & RTF_CLONING) {
195 			/*
196 			 * Case 1: This route should come from a route to iface.
197 			 */
198 			rt_setgate(rt, rt_key(rt),
199 					(struct sockaddr *)&null_sdl);
200 			gate = rt->rt_gateway;
201 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
202 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
203 			rt->rt_expire = time_second;
204 			break;
205 		}
206 		/* Announce a new entry if requested. */
207 		if (rt->rt_flags & RTF_ANNOUNCE)
208 			arprequest(rt->rt_ifp,
209 			    &SIN(rt_key(rt))->sin_addr,
210 			    &SIN(rt_key(rt))->sin_addr,
211 			    (u_char *)LLADDR(SDL(gate)));
212 		/*FALLTHROUGH*/
213 	case RTM_RESOLVE:
214 		if (gate->sa_family != AF_LINK ||
215 		    gate->sa_len < sizeof(null_sdl)) {
216 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
217 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
218 			    (gate->sa_family != AF_LINK) ?
219 			    " (!AF_LINK)": "");
220 			break;
221 		}
222 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
223 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
224 		if (la != 0)
225 			break; /* This happens on a route change */
226 		/*
227 		 * Case 2:  This route may come from cloning, or a manual route
228 		 * add with a LL address.
229 		 */
230 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
231 		rt->rt_llinfo = (caddr_t)la;
232 		if (la == 0) {
233 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
234 			break;
235 		}
236 		arp_allocated++;
237 		la->la_rt = rt;
238 		rt->rt_flags |= RTF_LLINFO;
239 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
240 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
241 
242 #ifdef INET
243 		/*
244 		 * This keeps the multicast addresses from showing up
245 		 * in `arp -a' listings as unresolved.  It's not actually
246 		 * functional.  Then the same for broadcast.
247 		 */
248 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
249 		    rt->rt_ifp->if_type != IFT_ARCNET) {
250 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
251 					       LLADDR(SDL(gate)));
252 			SDL(gate)->sdl_alen = 6;
253 			rt->rt_expire = 0;
254 		}
255 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
256 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
257 			       rt->rt_ifp->if_addrlen);
258 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
259 			rt->rt_expire = 0;
260 		}
261 #endif
262 
263 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
264 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
265 		    /*
266 		     * This test used to be
267 		     *	if (loif.if_flags & IFF_UP)
268 		     * It allowed local traffic to be forced
269 		     * through the hardware by configuring the loopback down.
270 		     * However, it causes problems during network configuration
271 		     * for boards that can't receive packets they send.
272 		     * It is now necessary to clear "useloopback" and remove
273 		     * the route to force traffic out to the hardware.
274 		     */
275 			rt->rt_expire = 0;
276 			Bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
277 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
278 			if (useloopback)
279 				rt->rt_ifp = loif;
280 
281 		}
282 		break;
283 
284 	case RTM_DELETE:
285 		if (la == 0)
286 			break;
287 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
288 		LIST_REMOVE(la, la_le);
289 		rt->rt_llinfo = 0;
290 		rt->rt_flags &= ~RTF_LLINFO;
291 		if (la->la_hold)
292 			m_freem(la->la_hold);
293 		Free((caddr_t)la);
294 	}
295 }
296 
297 /*
298  * Broadcast an ARP request. Caller specifies:
299  *	- arp header source ip address
300  *	- arp header target ip address
301  *	- arp header source ethernet address
302  */
303 static void
304 arprequest(ifp, sip, tip, enaddr)
305 	struct ifnet *ifp;
306 	struct in_addr *sip, *tip;
307 	u_char *enaddr;
308 {
309 	struct mbuf *m;
310 	struct arphdr *ah;
311 	struct sockaddr sa;
312 
313 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
314 		return;
315 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
316 		2*ifp->if_data.ifi_addrlen;
317 	m->m_pkthdr.len = m->m_len;
318 	MH_ALIGN(m, m->m_len);
319 	ah = mtod(m, struct arphdr *);
320 	bzero((caddr_t)ah, m->m_len);
321 #ifdef MAC
322 	mac_create_mbuf_linklayer(ifp, m);
323 #endif
324 	ah->ar_pro = htons(ETHERTYPE_IP);
325 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
326 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
327 	ah->ar_op = htons(ARPOP_REQUEST);
328 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
329 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
330 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
331 	sa.sa_family = AF_ARP;
332 	sa.sa_len = 2;
333 	m->m_flags |= M_BCAST;
334 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
335 
336 	return;
337 }
338 
339 /*
340  * Resolve an IP address into an ethernet address.  If success,
341  * desten is filled in.  If there is no entry in arptab,
342  * set one up and broadcast a request for the IP address.
343  * Hold onto this mbuf and resend it once the address
344  * is finally resolved.  A return value of 1 indicates
345  * that desten has been filled in and the packet should be sent
346  * normally; a 0 return indicates that the packet has been
347  * taken over here, either now or for later transmission.
348  */
349 int
350 arpresolve(ifp, rt, m, dst, desten, rt0)
351 	struct ifnet *ifp;
352 	struct rtentry *rt;
353 	struct mbuf *m;
354 	struct sockaddr *dst;
355 	u_char *desten;
356 	struct rtentry *rt0;
357 {
358 	struct llinfo_arp *la = 0;
359 	struct sockaddr_dl *sdl;
360 
361 	if (m->m_flags & M_BCAST) {	/* broadcast */
362 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
363 		return (1);
364 	}
365 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
366 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
367 		return(1);
368 	}
369 	if (rt)
370 		la = (struct llinfo_arp *)rt->rt_llinfo;
371 	if (la == 0) {
372 		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
373 		if (la)
374 			rt = la->la_rt;
375 	}
376 	if (la == 0 || rt == 0) {
377 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
378 			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
379 				rt ? "rt" : "");
380 		m_freem(m);
381 		return (0);
382 	}
383 	sdl = SDL(rt->rt_gateway);
384 	/*
385 	 * Check the address family and length is valid, the address
386 	 * is resolved; otherwise, try to resolve.
387 	 */
388 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
389 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
390 		/*
391 		 * If entry has an expiry time and it is approaching,
392 		 * see if we need to send an ARP request within this
393 		 * arpt_down interval.
394 		 */
395 		if ((rt->rt_expire != 0) &&
396 		    (time_second + la->la_preempt > rt->rt_expire)) {
397 			arprequest(ifp,
398 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
399 				   &SIN(dst)->sin_addr,
400 				   IF_LLADDR(ifp));
401 			la->la_preempt--;
402 		}
403 
404 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
405 		return 1;
406 	}
407 	/*
408 	 * If ARP is disabled or static on this interface, stop.
409 	 * XXX
410 	 * Probably should not allocate empty llinfo struct if we are
411 	 * not going to be sending out an arp request.
412 	 */
413 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
414 		m_freem(m);
415 		return (0);
416 	}
417 	/*
418 	 * There is an arptab entry, but no ethernet address
419 	 * response yet.  Replace the held mbuf with this
420 	 * latest one.
421 	 */
422 	if (la->la_hold)
423 		m_freem(la->la_hold);
424 	la->la_hold = m;
425 	if (rt->rt_expire) {
426 		RT_LOCK(rt);
427 		rt->rt_flags &= ~RTF_REJECT;
428 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
429 			rt->rt_expire = time_second;
430 			if (la->la_asked++ < arp_maxtries) {
431 				arprequest(ifp,
432 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
433 					   &SIN(dst)->sin_addr,
434 					   IF_LLADDR(ifp));
435 			} else {
436 				rt->rt_flags |= RTF_REJECT;
437 				rt->rt_expire += arpt_down;
438 				la->la_asked = 0;
439 				la->la_preempt = arp_maxtries;
440 			}
441 
442 		}
443 		RT_UNLOCK(rt);
444 	}
445 	return (0);
446 }
447 
448 /*
449  * Common length and type checks are done here,
450  * then the protocol-specific routine is called.
451  */
452 static void
453 arpintr(struct mbuf *m)
454 {
455 	struct arphdr *ar;
456 
457 	if (!arpinit_done) {
458 		/* NB: this race should not matter */
459 		arpinit_done = 1;
460 		callout_reset(&arp_callout, hz, arptimer, NULL);
461 	}
462 	if (m->m_len < sizeof(struct arphdr) &&
463 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
464 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
465 		return;
466 	}
467 	ar = mtod(m, struct arphdr *);
468 
469 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
470 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
471 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET) {
472 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
473 		    (unsigned char *)&ar->ar_hrd, "");
474 		m_freem(m);
475 		return;
476 	}
477 
478 	if (m->m_len < arphdr_len(ar)) {
479 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
480 			log(LOG_ERR, "arp: runt packet\n");
481 			m_freem(m);
482 			return;
483 		}
484 		ar = mtod(m, struct arphdr *);
485 	}
486 
487 	switch (ntohs(ar->ar_pro)) {
488 #ifdef INET
489 	case ETHERTYPE_IP:
490 		in_arpinput(m);
491 		return;
492 #endif
493 	}
494 	m_freem(m);
495 }
496 
497 #ifdef INET
498 /*
499  * ARP for Internet protocols on 10 Mb/s Ethernet.
500  * Algorithm is that given in RFC 826.
501  * In addition, a sanity check is performed on the sender
502  * protocol address, to catch impersonators.
503  * We no longer handle negotiations for use of trailer protocol:
504  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
505  * along with IP replies if we wanted trailers sent to us,
506  * and also sent them in response to IP replies.
507  * This allowed either end to announce the desire to receive
508  * trailer packets.
509  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
510  * but formerly didn't normally send requests.
511  */
512 static int log_arp_wrong_iface = 1;
513 static int log_arp_movements = 1;
514 
515 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
516 	&log_arp_wrong_iface, 0,
517 	"log arp packets arriving on the wrong interface");
518 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
519         &log_arp_movements, 0,
520         "log arp replies from MACs different than the one in the cache");
521 
522 
523 static void
524 in_arpinput(m)
525 	struct mbuf *m;
526 {
527 	struct arphdr *ah;
528 	struct ifnet *ifp = m->m_pkthdr.rcvif;
529 	struct iso88025_header *th = (struct iso88025_header *)0;
530 	struct iso88025_sockaddr_dl_data *trld;
531 	struct llinfo_arp *la = 0;
532 	struct rtentry *rt;
533 	struct ifaddr *ifa;
534 	struct in_ifaddr *ia;
535 	struct sockaddr_dl *sdl;
536 	struct sockaddr sa;
537 	struct in_addr isaddr, itaddr, myaddr;
538 	int op, rif_len;
539 	int req_len;
540 
541 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
542 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
543 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
544 		return;
545 	}
546 
547 	ah = mtod(m, struct arphdr *);
548 	op = ntohs(ah->ar_op);
549 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
550 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
551 #ifdef BRIDGE
552 #define BRIDGE_TEST (do_bridge)
553 #else
554 #define BRIDGE_TEST (0) /* cc will optimise the test away */
555 #endif
556 	/*
557 	 * For a bridge, we want to check the address irrespective
558 	 * of the receive interface. (This will change slightly
559 	 * when we have clusters of interfaces).
560 	 */
561 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash)
562 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
563 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
564 			goto match;
565 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
566 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
567 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
568 			goto match;
569 	/*
570 	 * No match, use the first inet address on the receive interface
571 	 * as a dummy address for the rest of the function.
572 	 */
573 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
574 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
575 			ia = ifatoia(ifa);
576 			goto match;
577 		}
578 	/*
579 	 * If bridging, fall back to using any inet address.
580 	 */
581 	if (!BRIDGE_TEST ||
582 	    (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) {
583 		m_freem(m);
584 		return;
585 	}
586 match:
587 	myaddr = ia->ia_addr.sin_addr;
588 	if (!bcmp(ar_sha(ah), IF_LLADDR(ifp), ifp->if_addrlen)) {
589 		m_freem(m);	/* it's from me, ignore it. */
590 		return;
591 	}
592 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
593 		log(LOG_ERR,
594 		    "arp: link address is broadcast for IP address %s!\n",
595 		    inet_ntoa(isaddr));
596 		m_freem(m);
597 		return;
598 	}
599 	if (isaddr.s_addr == myaddr.s_addr) {
600 		log(LOG_ERR,
601 		   "arp: %*D is using my IP address %s!\n",
602 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
603 		   inet_ntoa(isaddr));
604 		itaddr = myaddr;
605 		goto reply;
606 	}
607 	if (ifp->if_flags & IFF_STATICARP)
608 		goto reply;
609 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
610 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
611 		/* the following is not an error when doing bridging */
612 		if (!BRIDGE_TEST && rt->rt_ifp != ifp) {
613 			if (log_arp_wrong_iface)
614 				log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
615 				    inet_ntoa(isaddr),
616 				    rt->rt_ifp->if_xname,
617 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
618 				    ifp->if_xname);
619 			goto reply;
620 		}
621 		if (sdl->sdl_alen &&
622 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
623 			if (rt->rt_expire) {
624 			    if (log_arp_movements)
625 			        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
626 				    inet_ntoa(isaddr),
627 				    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
628 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
629 				    ifp->if_xname);
630 			} else {
631 			    log(LOG_ERR,
632 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
633 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
634 				inet_ntoa(isaddr), ifp->if_xname);
635 			    goto reply;
636 			}
637 		}
638 		/*
639 		 * sanity check for the address length.
640 		 * XXX this does not work for protocols with variable address
641 		 * length. -is
642 		 */
643 		if (sdl->sdl_alen &&
644 		    sdl->sdl_alen != ah->ar_hln) {
645 			log(LOG_WARNING,
646 			    "arp from %*D: new addr len %d, was %d",
647 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
648 			    ah->ar_hln, sdl->sdl_alen);
649 		}
650 		if (ifp->if_addrlen != ah->ar_hln) {
651 			log(LOG_WARNING,
652 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
653 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
654 			    ah->ar_hln, ifp->if_addrlen);
655 			goto reply;
656 		}
657 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
658 		    sdl->sdl_alen = ah->ar_hln);
659 		/*
660 		 * If we receive an arp from a token-ring station over
661 		 * a token-ring nic then try to save the source
662 		 * routing info.
663 		 */
664 		if (ifp->if_type == IFT_ISO88025) {
665 			th = (struct iso88025_header *)m->m_pkthdr.header;
666 			trld = SDL_ISO88025(sdl);
667 			rif_len = TR_RCF_RIFLEN(th->rcf);
668 			if ((th->iso88025_shost[0] & TR_RII) &&
669 			    (rif_len > 2)) {
670 				trld->trld_rcf = th->rcf;
671 				trld->trld_rcf ^= htons(TR_RCF_DIR);
672 				memcpy(trld->trld_route, th->rd, rif_len - 2);
673 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
674 				/*
675 				 * Set up source routing information for
676 				 * reply packet (XXX)
677 				 */
678 				m->m_data -= rif_len;
679 				m->m_len  += rif_len;
680 				m->m_pkthdr.len += rif_len;
681 			} else {
682 				th->iso88025_shost[0] &= ~TR_RII;
683 				trld->trld_rcf = 0;
684 			}
685 			m->m_data -= 8;
686 			m->m_len  += 8;
687 			m->m_pkthdr.len += 8;
688 			th->rcf = trld->trld_rcf;
689 		}
690 		RT_LOCK(rt);
691 		if (rt->rt_expire)
692 			rt->rt_expire = time_second + arpt_keep;
693 		rt->rt_flags &= ~RTF_REJECT;
694 		RT_UNLOCK(rt);
695 		la->la_asked = 0;
696 		la->la_preempt = arp_maxtries;
697 		if (la->la_hold) {
698 			(*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt);
699 			la->la_hold = 0;
700 		}
701 	}
702 reply:
703 	if (op != ARPOP_REQUEST) {
704 		m_freem(m);
705 		return;
706 	}
707 	if (itaddr.s_addr == myaddr.s_addr) {
708 		/* I am the target */
709 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
710 		(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
711 	} else {
712 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
713 		if (la == NULL) {
714 			struct sockaddr_in sin;
715 
716 			if (!arp_proxyall) {
717 				m_freem(m);
718 				return;
719 			}
720 
721 			bzero(&sin, sizeof sin);
722 			sin.sin_family = AF_INET;
723 			sin.sin_len = sizeof sin;
724 			sin.sin_addr = itaddr;
725 
726 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
727 			if (!rt) {
728 				m_freem(m);
729 				return;
730 			}
731 			/*
732 			 * Don't send proxies for nodes on the same interface
733 			 * as this one came out of, or we'll get into a fight
734 			 * over who claims what Ether address.
735 			 */
736 			if (rt->rt_ifp == ifp) {
737 				rtfree(rt);
738 				m_freem(m);
739 				return;
740 			}
741 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
742 			(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
743 			rtfree(rt);
744 
745 			/*
746 			 * Also check that the node which sent the ARP packet
747 			 * is on the the interface we expect it to be on. This
748 			 * avoids ARP chaos if an interface is connected to the
749 			 * wrong network.
750 			 */
751 			sin.sin_addr = isaddr;
752 
753 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
754 			if (!rt) {
755 				m_freem(m);
756 				return;
757 			}
758 			if (rt->rt_ifp != ifp) {
759 				log(LOG_INFO, "arp_proxy: ignoring request"
760 				    " from %s via %s, expecting %s\n",
761 				    inet_ntoa(isaddr), ifp->if_xname,
762 				    rt->rt_ifp->if_xname);
763 				rtfree(rt);
764 				m_freem(m);
765 				return;
766 			}
767 			rtfree(rt);
768 
769 #ifdef DEBUG_PROXY
770 			printf("arp: proxying for %s\n",
771 			       inet_ntoa(itaddr));
772 #endif
773 		} else {
774 			rt = la->la_rt;
775 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
776 			sdl = SDL(rt->rt_gateway);
777 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
778 		}
779 	}
780 
781 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
782 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
783 	ah->ar_op = htons(ARPOP_REPLY);
784 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
785 	m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
786 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
787 	m->m_pkthdr.len = m->m_len;
788 	sa.sa_family = AF_ARP;
789 	sa.sa_len = 2;
790 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
791 	return;
792 }
793 #endif
794 
795 /*
796  * Free an arp entry.
797  */
798 static void
799 arptfree(la)
800 	struct llinfo_arp *la;
801 {
802 	struct rtentry *rt = la->la_rt;
803 	struct sockaddr_dl *sdl;
804 
805 	if (rt == 0)
806 		panic("arptfree");
807 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
808 	    sdl->sdl_family == AF_LINK) {
809 		sdl->sdl_alen = 0;
810 		la->la_preempt = la->la_asked = 0;
811 		RT_LOCK(rt);		/* XXX needed or move higher? */
812 		rt->rt_flags &= ~RTF_REJECT;
813 		RT_UNLOCK(rt);
814 		return;
815 	}
816 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
817 			0, (struct rtentry **)0);
818 }
819 /*
820  * Lookup or enter a new address in arptab.
821  */
822 static struct llinfo_arp *
823 arplookup(addr, create, proxy)
824 	u_long addr;
825 	int create, proxy;
826 {
827 	struct rtentry *rt;
828 	struct sockaddr_inarp sin;
829 	const char *why = 0;
830 
831 	bzero(&sin, sizeof(sin));
832 	sin.sin_len = sizeof(sin);
833 	sin.sin_family = AF_INET;
834 	sin.sin_addr.s_addr = addr;
835 	if (proxy)
836 		sin.sin_other = SIN_PROXY;
837 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
838 	if (rt == 0)
839 		return (0);
840 
841 	if (rt->rt_flags & RTF_GATEWAY)
842 		why = "host is not on local network";
843 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
844 		why = "could not allocate llinfo";
845 	else if (rt->rt_gateway->sa_family != AF_LINK)
846 		why = "gateway route is not ours";
847 
848 	if (why) {
849 #define	ISDYNCLONE(_rt) \
850 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
851 		if (create)
852 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
853 			    inet_ntoa(sin.sin_addr), why);
854 		/*
855 		 * If there are no references to this Layer 2 route,
856 		 * and it is a cloned route, and not static, and
857 		 * arplookup() is creating the route, then purge
858 		 * it from the routing table as it is probably bogus.
859 		 */
860 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
861 			rtexpunge(rt);
862 		RTFREE_LOCKED(rt);
863 		return (0);
864 #undef ISDYNCLONE
865 	} else {
866 		RT_REMREF(rt);
867 		RT_UNLOCK(rt);
868 		return ((struct llinfo_arp *)rt->rt_llinfo);
869 	}
870 }
871 
872 void
873 arp_ifinit(ifp, ifa)
874 	struct ifnet *ifp;
875 	struct ifaddr *ifa;
876 {
877 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
878 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
879 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
880 	ifa->ifa_rtrequest = arp_rtrequest;
881 	ifa->ifa_flags |= RTF_CLONING;
882 }
883 
884 static void
885 arp_init(void)
886 {
887 
888 	arpintrq.ifq_maxlen = 50;
889 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
890 	LIST_INIT(&llinfo_arp);
891 	callout_init(&arp_callout, CALLOUT_MPSAFE);
892 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
893 }
894 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
895