xref: /freebsd/sys/netinet/if_ether.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30  * $FreeBSD$
31  */
32 
33 /*
34  * Ethernet address resolution protocol.
35  * TODO:
36  *	add "inuse/lock" bit (or ref. count) along with valid bit
37  */
38 
39 #include "opt_inet.h"
40 #include "opt_mac.h"
41 #include "opt_carp.h"
42 
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/queue.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/socket.h>
51 #include <sys/syslog.h>
52 
53 #include <net/if.h>
54 #include <net/if_dl.h>
55 #include <net/if_types.h>
56 #include <net/route.h>
57 #include <net/netisr.h>
58 #include <net/if_llc.h>
59 #include <net/ethernet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_var.h>
63 #include <netinet/if_ether.h>
64 
65 #include <net/if_arc.h>
66 #include <net/iso88025.h>
67 
68 #ifdef DEV_CARP
69 #include <netinet/ip_carp.h>
70 #endif
71 
72 #include <security/mac/mac_framework.h>
73 
74 #define SIN(s) ((struct sockaddr_in *)s)
75 #define SDL(s) ((struct sockaddr_dl *)s)
76 
77 SYSCTL_DECL(_net_link_ether);
78 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
79 
80 /* timer values */
81 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
82 
83 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
84 	   &arpt_keep, 0, "ARP entry lifetime in seconds");
85 
86 #define	rt_expire rt_rmx.rmx_expire
87 
88 struct llinfo_arp {
89 	struct	callout la_timer;
90 	struct	rtentry *la_rt;
91 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
92 	u_short	la_preempt;	/* countdown for pre-expiry arps */
93 	u_short	la_asked;	/* # requests sent */
94 };
95 
96 static struct	ifqueue arpintrq;
97 static int	arp_allocated;
98 
99 static int	arp_maxtries = 5;
100 static int	useloopback = 1; /* use loopback interface for local traffic */
101 static int	arp_proxyall = 0;
102 
103 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
104 	   &arp_maxtries, 0, "ARP resolution attempts before returning error");
105 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
106 	   &useloopback, 0, "Use the loopback interface for local traffic");
107 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
108 	   &arp_proxyall, 0, "Enable proxy ARP for all suitable requests");
109 
110 static void	arp_init(void);
111 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
112 static void	arprequest(struct ifnet *,
113 			struct in_addr *, struct in_addr *, u_char *);
114 static void	arpintr(struct mbuf *);
115 static void	arptimer(void *);
116 static struct rtentry
117 		*arplookup(u_long, int, int);
118 #ifdef INET
119 static void	in_arpinput(struct mbuf *);
120 #endif
121 
122 /*
123  * Timeout routine.
124  */
125 static void
126 arptimer(void *arg)
127 {
128 	struct rtentry *rt = (struct rtentry *)arg;
129 
130 	RT_LOCK_ASSERT(rt);
131 	/*
132 	 * The lock is needed to close a theoretical race
133 	 * between spontaneous expiry and intentional removal.
134 	 * We still got an extra reference on rtentry, so can
135 	 * safely pass pointers to its contents.
136 	 */
137 	RT_UNLOCK(rt);
138 
139 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
140 }
141 
142 /*
143  * Parallel to llc_rtrequest.
144  */
145 static void
146 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
147 {
148 	struct sockaddr *gate;
149 	struct llinfo_arp *la;
150 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
151 	struct in_ifaddr *ia;
152 	struct ifaddr *ifa;
153 
154 	RT_LOCK_ASSERT(rt);
155 
156 	if (rt->rt_flags & RTF_GATEWAY)
157 		return;
158 	gate = rt->rt_gateway;
159 	la = (struct llinfo_arp *)rt->rt_llinfo;
160 	switch (req) {
161 
162 	case RTM_ADD:
163 		/*
164 		 * XXX: If this is a manually added route to interface
165 		 * such as older version of routed or gated might provide,
166 		 * restore cloning bit.
167 		 */
168 		if ((rt->rt_flags & RTF_HOST) == 0 &&
169 		    rt_mask(rt) != NULL &&
170 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
171 			rt->rt_flags |= RTF_CLONING;
172 		if (rt->rt_flags & RTF_CLONING) {
173 			/*
174 			 * Case 1: This route should come from a route to iface.
175 			 */
176 			rt_setgate(rt, rt_key(rt),
177 					(struct sockaddr *)&null_sdl);
178 			gate = rt->rt_gateway;
179 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
180 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
181 			rt->rt_expire = time_uptime;
182 			break;
183 		}
184 		/* Announce a new entry if requested. */
185 		if (rt->rt_flags & RTF_ANNOUNCE)
186 			arprequest(rt->rt_ifp,
187 			    &SIN(rt_key(rt))->sin_addr,
188 			    &SIN(rt_key(rt))->sin_addr,
189 			    (u_char *)LLADDR(SDL(gate)));
190 		/*FALLTHROUGH*/
191 	case RTM_RESOLVE:
192 		if (gate->sa_family != AF_LINK ||
193 		    gate->sa_len < sizeof(null_sdl)) {
194 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
195 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
196 			    (gate->sa_family != AF_LINK) ?
197 			    " (!AF_LINK)": "");
198 			break;
199 		}
200 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
201 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
202 		if (la != 0)
203 			break; /* This happens on a route change */
204 		/*
205 		 * Case 2:  This route may come from cloning, or a manual route
206 		 * add with a LL address.
207 		 */
208 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
209 		rt->rt_llinfo = (caddr_t)la;
210 		if (la == 0) {
211 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
212 			break;
213 		}
214 		arp_allocated++;
215 		/*
216 		 * We are storing a route entry outside of radix tree. So,
217 		 * it can be found and accessed by other means than radix
218 		 * lookup. The routing code assumes that any rtentry detached
219 		 * from radix can be destroyed safely. To prevent this, we
220 		 * add an additional reference.
221 		 */
222 		RT_ADDREF(rt);
223 		la->la_rt = rt;
224 		rt->rt_flags |= RTF_LLINFO;
225 		callout_init_mtx(&la->la_timer, &rt->rt_mtx,
226 		    CALLOUT_RETURNUNLOCKED);
227 
228 #ifdef INET
229 		/*
230 		 * This keeps the multicast addresses from showing up
231 		 * in `arp -a' listings as unresolved.  It's not actually
232 		 * functional.  Then the same for broadcast.
233 		 */
234 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
235 		    rt->rt_ifp->if_type != IFT_ARCNET) {
236 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
237 					       LLADDR(SDL(gate)));
238 			SDL(gate)->sdl_alen = 6;
239 			rt->rt_expire = 0;
240 		}
241 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
242 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
243 			       rt->rt_ifp->if_addrlen);
244 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
245 			rt->rt_expire = 0;
246 		}
247 #endif
248 
249 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
250 			if (ia->ia_ifp == rt->rt_ifp &&
251 			    SIN(rt_key(rt))->sin_addr.s_addr ==
252 			    (IA_SIN(ia))->sin_addr.s_addr)
253 				break;
254 		}
255 		if (ia) {
256 		    /*
257 		     * This test used to be
258 		     *	if (loif.if_flags & IFF_UP)
259 		     * It allowed local traffic to be forced
260 		     * through the hardware by configuring the loopback down.
261 		     * However, it causes problems during network configuration
262 		     * for boards that can't receive packets they send.
263 		     * It is now necessary to clear "useloopback" and remove
264 		     * the route to force traffic out to the hardware.
265 		     */
266 			rt->rt_expire = 0;
267 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
268 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
269 			if (useloopback) {
270 				rt->rt_ifp = loif;
271 				rt->rt_rmx.rmx_mtu = loif->if_mtu;
272 			}
273 
274 		    /*
275 		     * make sure to set rt->rt_ifa to the interface
276 		     * address we are using, otherwise we will have trouble
277 		     * with source address selection.
278 		     */
279 			ifa = &ia->ia_ifa;
280 			if (ifa != rt->rt_ifa) {
281 				IFAFREE(rt->rt_ifa);
282 				IFAREF(ifa);
283 				rt->rt_ifa = ifa;
284 			}
285 		}
286 		break;
287 
288 	case RTM_DELETE:
289 		if (la == NULL)	/* XXX: at least CARP does this. */
290 			break;
291 		callout_stop(&la->la_timer);
292 		rt->rt_llinfo = NULL;
293 		rt->rt_flags &= ~RTF_LLINFO;
294 		RT_REMREF(rt);
295 		if (la->la_hold)
296 			m_freem(la->la_hold);
297 		Free((caddr_t)la);
298 	}
299 }
300 
301 /*
302  * Broadcast an ARP request. Caller specifies:
303  *	- arp header source ip address
304  *	- arp header target ip address
305  *	- arp header source ethernet address
306  */
307 static void
308 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
309     u_char *enaddr)
310 {
311 	struct mbuf *m;
312 	struct arphdr *ah;
313 	struct sockaddr sa;
314 
315 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
316 		return;
317 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
318 		2*ifp->if_data.ifi_addrlen;
319 	m->m_pkthdr.len = m->m_len;
320 	MH_ALIGN(m, m->m_len);
321 	ah = mtod(m, struct arphdr *);
322 	bzero((caddr_t)ah, m->m_len);
323 #ifdef MAC
324 	mac_create_mbuf_linklayer(ifp, m);
325 #endif
326 	ah->ar_pro = htons(ETHERTYPE_IP);
327 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
328 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
329 	ah->ar_op = htons(ARPOP_REQUEST);
330 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
331 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
332 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
333 	sa.sa_family = AF_ARP;
334 	sa.sa_len = 2;
335 	m->m_flags |= M_BCAST;
336 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
337 
338 	return;
339 }
340 
341 /*
342  * Resolve an IP address into an ethernet address.
343  * On input:
344  *    ifp is the interface we use
345  *    dst is the next hop,
346  *    rt0 is the route to the final destination (possibly useless)
347  *    m is the mbuf
348  *    desten is where we want the address.
349  *
350  * On success, desten is filled in and the function returns 0;
351  * If the packet must be held pending resolution, we return EWOULDBLOCK
352  * On other errors, we return the corresponding error code.
353  */
354 int
355 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
356     struct sockaddr *dst, u_char *desten)
357 {
358 	struct llinfo_arp *la = NULL;
359 	struct rtentry *rt = NULL;
360 	struct sockaddr_dl *sdl;
361 	int error;
362 
363 	if (m->m_flags & M_BCAST) {	/* broadcast */
364 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
365 		return (0);
366 	}
367 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
368 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
369 		return (0);
370 	}
371 
372 	if (rt0 != NULL) {
373 		error = rt_check(&rt, &rt0, dst);
374 		if (error) {
375 			m_freem(m);
376 			return error;
377 		}
378 		la = (struct llinfo_arp *)rt->rt_llinfo;
379 		if (la == NULL)
380 			RT_UNLOCK(rt);
381 	}
382 	if (la == NULL) {
383 		/*
384 		 * We enter this block in case if rt0 was NULL,
385 		 * or if rt found by rt_check() didn't have llinfo.
386 		 */
387 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
388 		if (rt == NULL) {
389 			log(LOG_DEBUG,
390 			    "arpresolve: can't allocate route for %s\n",
391 			    inet_ntoa(SIN(dst)->sin_addr));
392 			m_freem(m);
393 			return (EINVAL); /* XXX */
394 		}
395 		la = (struct llinfo_arp *)rt->rt_llinfo;
396 		if (la == NULL) {
397 			RT_UNLOCK(rt);
398 			log(LOG_DEBUG,
399 			    "arpresolve: can't allocate llinfo for %s\n",
400 			    inet_ntoa(SIN(dst)->sin_addr));
401 			m_freem(m);
402 			return (EINVAL); /* XXX */
403 		}
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_uptime) &&
411 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
412 
413 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
414 
415 		/*
416 		 * If entry has an expiry time and it is approaching,
417 		 * send an ARP request.
418 		 */
419 		if ((rt->rt_expire != 0) &&
420 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
421 			struct in_addr sin =
422 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
423 
424 			la->la_preempt--;
425 			RT_UNLOCK(rt);
426 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
427 			    IF_LLADDR(ifp));
428 			return (0);
429 		}
430 
431 		RT_UNLOCK(rt);
432 		return (0);
433 	}
434 	/*
435 	 * If ARP is disabled or static on this interface, stop.
436 	 * XXX
437 	 * Probably should not allocate empty llinfo struct if we are
438 	 * not going to be sending out an arp request.
439 	 */
440 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
441 		RT_UNLOCK(rt);
442 		m_freem(m);
443 		return (EINVAL);
444 	}
445 	/*
446 	 * There is an arptab entry, but no ethernet address
447 	 * response yet.  Replace the held mbuf with this
448 	 * latest one.
449 	 */
450 	if (la->la_hold)
451 		m_freem(la->la_hold);
452 	la->la_hold = m;
453 
454 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
455 
456 	/*
457 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
458 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
459 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
460 	 * ARP request, but not faster than one request per second.
461 	 */
462 	if (la->la_asked < arp_maxtries)
463 		error = EWOULDBLOCK;	/* First request. */
464 	else
465 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
466 
467 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
468 		struct in_addr sin =
469 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
470 
471 		rt->rt_expire = time_uptime;
472 		callout_reset(&la->la_timer, hz, arptimer, rt);
473 		la->la_asked++;
474 		RT_UNLOCK(rt);
475 
476 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
477 		    IF_LLADDR(ifp));
478 	} else
479 		RT_UNLOCK(rt);
480 
481 	return (error);
482 }
483 
484 /*
485  * Common length and type checks are done here,
486  * then the protocol-specific routine is called.
487  */
488 static void
489 arpintr(struct mbuf *m)
490 {
491 	struct arphdr *ar;
492 
493 	if (m->m_len < sizeof(struct arphdr) &&
494 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
495 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
496 		return;
497 	}
498 	ar = mtod(m, struct arphdr *);
499 
500 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
501 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
502 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
503 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
504 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
505 		    (unsigned char *)&ar->ar_hrd, "");
506 		m_freem(m);
507 		return;
508 	}
509 
510 	if (m->m_len < arphdr_len(ar)) {
511 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
512 			log(LOG_ERR, "arp: runt packet\n");
513 			m_freem(m);
514 			return;
515 		}
516 		ar = mtod(m, struct arphdr *);
517 	}
518 
519 	switch (ntohs(ar->ar_pro)) {
520 #ifdef INET
521 	case ETHERTYPE_IP:
522 		in_arpinput(m);
523 		return;
524 #endif
525 	}
526 	m_freem(m);
527 }
528 
529 #ifdef INET
530 /*
531  * ARP for Internet protocols on 10 Mb/s Ethernet.
532  * Algorithm is that given in RFC 826.
533  * In addition, a sanity check is performed on the sender
534  * protocol address, to catch impersonators.
535  * We no longer handle negotiations for use of trailer protocol:
536  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
537  * along with IP replies if we wanted trailers sent to us,
538  * and also sent them in response to IP replies.
539  * This allowed either end to announce the desire to receive
540  * trailer packets.
541  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
542  * but formerly didn't normally send requests.
543  */
544 static int log_arp_wrong_iface = 1;
545 static int log_arp_movements = 1;
546 static int log_arp_permanent_modify = 1;
547 
548 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
549 	&log_arp_wrong_iface, 0,
550 	"log arp packets arriving on the wrong interface");
551 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
552         &log_arp_movements, 0,
553         "log arp replies from MACs different than the one in the cache");
554 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
555         &log_arp_permanent_modify, 0,
556         "log arp replies from MACs different than the one in the permanent arp entry");
557 
558 
559 static void
560 in_arpinput(struct mbuf *m)
561 {
562 	struct arphdr *ah;
563 	struct ifnet *ifp = m->m_pkthdr.rcvif;
564 	struct llinfo_arp *la;
565 	struct rtentry *rt;
566 	struct ifaddr *ifa;
567 	struct in_ifaddr *ia;
568 	struct sockaddr_dl *sdl;
569 	struct sockaddr sa;
570 	struct in_addr isaddr, itaddr, myaddr;
571 	struct mbuf *hold;
572 	u_int8_t *enaddr = NULL;
573 	int op, rif_len;
574 	int req_len;
575 	int bridged = 0;
576 #ifdef DEV_CARP
577 	int carp_match = 0;
578 #endif
579 
580 	if (ifp->if_bridge)
581 		bridged = 1;
582 
583 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
584 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
585 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
586 		return;
587 	}
588 
589 	ah = mtod(m, struct arphdr *);
590 	op = ntohs(ah->ar_op);
591 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
592 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
593 
594 	/*
595 	 * For a bridge, we want to check the address irrespective
596 	 * of the receive interface. (This will change slightly
597 	 * when we have clusters of interfaces).
598 	 * If the interface does not match, but the recieving interface
599 	 * is part of carp, we call carp_iamatch to see if this is a
600 	 * request for the virtual host ip.
601 	 * XXX: This is really ugly!
602 	 */
603 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
604 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
605 		    (ia->ia_ifp == ifp)) &&
606 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
607 			goto match;
608 #ifdef DEV_CARP
609 		if (ifp->if_carp != NULL &&
610 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
611 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
612 			carp_match = 1;
613 			goto match;
614 		}
615 #endif
616 	}
617 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
618 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
619 		    (ia->ia_ifp == ifp)) &&
620 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
621 			goto match;
622 	/*
623 	 * No match, use the first inet address on the receive interface
624 	 * as a dummy address for the rest of the function.
625 	 */
626 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
627 		if (ifa->ifa_addr->sa_family == AF_INET) {
628 			ia = ifatoia(ifa);
629 			goto match;
630 		}
631 	/*
632 	 * If bridging, fall back to using any inet address.
633 	 */
634 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
635 		goto drop;
636 match:
637 	if (!enaddr)
638 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
639 	myaddr = ia->ia_addr.sin_addr;
640 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
641 		goto drop;	/* it's from me, ignore it. */
642 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
643 		log(LOG_ERR,
644 		    "arp: link address is broadcast for IP address %s!\n",
645 		    inet_ntoa(isaddr));
646 		goto drop;
647 	}
648 	/*
649 	 * Warn if another host is using the same IP address, but only if the
650 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
651 	 * case we suppress the warning to avoid false positive complaints of
652 	 * potential misconfiguration.
653 	 */
654 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
655 		log(LOG_ERR,
656 		   "arp: %*D is using my IP address %s!\n",
657 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
658 		   inet_ntoa(isaddr));
659 		itaddr = myaddr;
660 		goto reply;
661 	}
662 	if (ifp->if_flags & IFF_STATICARP)
663 		goto reply;
664 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
665 	if (rt != NULL) {
666 		la = (struct llinfo_arp *)rt->rt_llinfo;
667 		if (la == NULL) {
668 			RT_UNLOCK(rt);
669 			goto reply;
670 		}
671 	} else
672 		goto reply;
673 
674 	/* The following is not an error when doing bridging. */
675 	if (!bridged && rt->rt_ifp != ifp
676 #ifdef DEV_CARP
677 	    && (ifp->if_type != IFT_CARP || !carp_match)
678 #endif
679 							) {
680 		if (log_arp_wrong_iface)
681 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
682 			    inet_ntoa(isaddr),
683 			    rt->rt_ifp->if_xname,
684 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
685 			    ifp->if_xname);
686 		RT_UNLOCK(rt);
687 		goto reply;
688 	}
689 	sdl = SDL(rt->rt_gateway);
690 	if (sdl->sdl_alen &&
691 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
692 		if (rt->rt_expire) {
693 		    if (log_arp_movements)
694 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
695 			    inet_ntoa(isaddr),
696 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
697 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
698 			    ifp->if_xname);
699 		} else {
700 			RT_UNLOCK(rt);
701 			if (log_arp_permanent_modify)
702 				log(LOG_ERR, "arp: %*D attempts to modify "
703 				    "permanent entry for %s on %s\n",
704 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
705 				    inet_ntoa(isaddr), ifp->if_xname);
706 			goto reply;
707 		}
708 	}
709 	/*
710 	 * sanity check for the address length.
711 	 * XXX this does not work for protocols with variable address
712 	 * length. -is
713 	 */
714 	if (sdl->sdl_alen &&
715 	    sdl->sdl_alen != ah->ar_hln) {
716 		log(LOG_WARNING,
717 		    "arp from %*D: new addr len %d, was %d",
718 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
719 		    ah->ar_hln, sdl->sdl_alen);
720 	}
721 	if (ifp->if_addrlen != ah->ar_hln) {
722 		log(LOG_WARNING,
723 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
724 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
725 		    ah->ar_hln, ifp->if_addrlen);
726 		RT_UNLOCK(rt);
727 		goto reply;
728 	}
729 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
730 	    sdl->sdl_alen = ah->ar_hln);
731 	/*
732 	 * If we receive an arp from a token-ring station over
733 	 * a token-ring nic then try to save the source
734 	 * routing info.
735 	 */
736 	if (ifp->if_type == IFT_ISO88025) {
737 		struct iso88025_header *th = NULL;
738 		struct iso88025_sockaddr_dl_data *trld;
739 
740 		th = (struct iso88025_header *)m->m_pkthdr.header;
741 		trld = SDL_ISO88025(sdl);
742 		rif_len = TR_RCF_RIFLEN(th->rcf);
743 		if ((th->iso88025_shost[0] & TR_RII) &&
744 		    (rif_len > 2)) {
745 			trld->trld_rcf = th->rcf;
746 			trld->trld_rcf ^= htons(TR_RCF_DIR);
747 			memcpy(trld->trld_route, th->rd, rif_len - 2);
748 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
749 			/*
750 			 * Set up source routing information for
751 			 * reply packet (XXX)
752 			 */
753 			m->m_data -= rif_len;
754 			m->m_len  += rif_len;
755 			m->m_pkthdr.len += rif_len;
756 		} else {
757 			th->iso88025_shost[0] &= ~TR_RII;
758 			trld->trld_rcf = 0;
759 		}
760 		m->m_data -= 8;
761 		m->m_len  += 8;
762 		m->m_pkthdr.len += 8;
763 		th->rcf = trld->trld_rcf;
764 	}
765 	if (rt->rt_expire) {
766 		rt->rt_expire = time_uptime + arpt_keep;
767 		callout_reset(&la->la_timer, hz * arpt_keep, arptimer, rt);
768 	}
769 	la->la_asked = 0;
770 	la->la_preempt = arp_maxtries;
771 	hold = la->la_hold;
772 	la->la_hold = NULL;
773 	RT_UNLOCK(rt);
774 	if (hold != NULL)
775 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
776 
777 reply:
778 	if (op != ARPOP_REQUEST)
779 		goto drop;
780 	if (itaddr.s_addr == myaddr.s_addr) {
781 		/* I am the target */
782 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
783 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
784 	} else {
785 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
786 		if (rt == NULL) {
787 			struct sockaddr_in sin;
788 
789 			if (!arp_proxyall)
790 				goto drop;
791 
792 			bzero(&sin, sizeof sin);
793 			sin.sin_family = AF_INET;
794 			sin.sin_len = sizeof sin;
795 			sin.sin_addr = itaddr;
796 
797 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
798 			if (!rt)
799 				goto drop;
800 			/*
801 			 * Don't send proxies for nodes on the same interface
802 			 * as this one came out of, or we'll get into a fight
803 			 * over who claims what Ether address.
804 			 */
805 			if (rt->rt_ifp == ifp) {
806 				rtfree(rt);
807 				goto drop;
808 			}
809 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
810 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
811 			rtfree(rt);
812 
813 			/*
814 			 * Also check that the node which sent the ARP packet
815 			 * is on the the interface we expect it to be on. This
816 			 * avoids ARP chaos if an interface is connected to the
817 			 * wrong network.
818 			 */
819 			sin.sin_addr = isaddr;
820 
821 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
822 			if (!rt)
823 				goto drop;
824 			if (rt->rt_ifp != ifp) {
825 				log(LOG_INFO, "arp_proxy: ignoring request"
826 				    " from %s via %s, expecting %s\n",
827 				    inet_ntoa(isaddr), ifp->if_xname,
828 				    rt->rt_ifp->if_xname);
829 				rtfree(rt);
830 				goto drop;
831 			}
832 			rtfree(rt);
833 
834 #ifdef DEBUG_PROXY
835 			printf("arp: proxying for %s\n",
836 			       inet_ntoa(itaddr));
837 #endif
838 		} else {
839 			/*
840 			 * Return proxied ARP replies only on the interface
841 			 * or bridge cluster where this network resides.
842 			 * Otherwise we may conflict with the host we are
843 			 * proxying for.
844 			 */
845 			if (rt->rt_ifp != ifp &&
846 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
847 			    ifp->if_bridge == NULL)) {
848 				RT_UNLOCK(rt);
849 				goto drop;
850 			}
851 			sdl = SDL(rt->rt_gateway);
852 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
853 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
854 			RT_UNLOCK(rt);
855 		}
856 	}
857 
858 	if (itaddr.s_addr == myaddr.s_addr &&
859 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
860 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
861 #ifdef DEBUG_LINKLOCAL
862 		printf("arp: sending reply for link-local addr %s\n",
863 		    inet_ntoa(itaddr));
864 #endif
865 		m->m_flags |= M_BCAST;
866 		m->m_flags &= ~M_MCAST;
867 	} else {
868 		/* default behaviour; never reply by broadcast. */
869 		m->m_flags &= ~(M_BCAST|M_MCAST);
870 	}
871 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
872 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
873 	ah->ar_op = htons(ARPOP_REPLY);
874 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
875 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
876 	m->m_pkthdr.len = m->m_len;
877 	sa.sa_family = AF_ARP;
878 	sa.sa_len = 2;
879 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
880 	return;
881 
882 drop:
883 	m_freem(m);
884 }
885 #endif
886 
887 /*
888  * Lookup or enter a new address in arptab.
889  */
890 static struct rtentry *
891 arplookup(u_long addr, int create, int proxy)
892 {
893 	struct rtentry *rt;
894 	struct sockaddr_inarp sin;
895 	const char *why = 0;
896 
897 	bzero(&sin, sizeof(sin));
898 	sin.sin_len = sizeof(sin);
899 	sin.sin_family = AF_INET;
900 	sin.sin_addr.s_addr = addr;
901 	if (proxy)
902 		sin.sin_other = SIN_PROXY;
903 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
904 	if (rt == 0)
905 		return (0);
906 
907 	if (rt->rt_flags & RTF_GATEWAY)
908 		why = "host is not on local network";
909 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
910 		why = "could not allocate llinfo";
911 	else if (rt->rt_gateway->sa_family != AF_LINK)
912 		why = "gateway route is not ours";
913 
914 	if (why) {
915 #define	ISDYNCLONE(_rt) \
916 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
917 		if (create)
918 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
919 			    inet_ntoa(sin.sin_addr), why);
920 		/*
921 		 * If there are no references to this Layer 2 route,
922 		 * and it is a cloned route, and not static, and
923 		 * arplookup() is creating the route, then purge
924 		 * it from the routing table as it is probably bogus.
925 		 */
926 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
927 			rtexpunge(rt);
928 		RTFREE_LOCKED(rt);
929 		return (0);
930 #undef ISDYNCLONE
931 	} else {
932 		RT_REMREF(rt);
933 		return (rt);
934 	}
935 }
936 
937 void
938 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
939 {
940 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
941 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
942 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
943 	ifa->ifa_rtrequest = arp_rtrequest;
944 	ifa->ifa_flags |= RTF_CLONING;
945 }
946 
947 void
948 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
949 {
950 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
951 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
952 				&IA_SIN(ifa)->sin_addr, enaddr);
953 	ifa->ifa_rtrequest = arp_rtrequest;
954 	ifa->ifa_flags |= RTF_CLONING;
955 }
956 
957 static void
958 arp_init(void)
959 {
960 
961 	arpintrq.ifq_maxlen = 50;
962 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
963 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
964 }
965 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
966