xref: /freebsd/sys/netinet/if_ether.c (revision ab2043b81eaba0d7d7769b4a58b2b6d17bc464a3)
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  */
31 
32 /*
33  * Ethernet address resolution protocol.
34  * TODO:
35  *	add "inuse/lock" bit (or ref. count) along with valid bit
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.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/proc.h>
51 #include <sys/socket.h>
52 #include <sys/syslog.h>
53 
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/netisr.h>
58 #include <net/if_llc.h>
59 #include <net/ethernet.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_var.h>
65 #include <net/if_llatbl.h>
66 #include <netinet/if_ether.h>
67 #ifdef INET
68 #include <netinet/ip_carp.h>
69 #endif
70 
71 #include <net/if_arc.h>
72 #include <net/iso88025.h>
73 
74 #include <security/mac/mac_framework.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 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
82 
83 /* timer values */
84 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
85 						 * minutes */
86 static VNET_DEFINE(int, arp_maxtries) = 5;
87 VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
88 					 * local traffic */
89 static VNET_DEFINE(int, arp_proxyall) = 0;
90 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
91 						 * 20 seconds */
92 VNET_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
93 
94 static VNET_DEFINE(int, arp_maxhold) = 1;
95 
96 #define	V_arpt_keep		VNET(arpt_keep)
97 #define	V_arpt_down		VNET(arpt_down)
98 #define	V_arp_maxtries		VNET(arp_maxtries)
99 #define	V_arp_proxyall		VNET(arp_proxyall)
100 #define	V_arpstat		VNET(arpstat)
101 #define	V_arp_maxhold		VNET(arp_maxhold)
102 
103 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
104 	&VNET_NAME(arpt_keep), 0,
105 	"ARP entry lifetime in seconds");
106 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
107 	&VNET_NAME(arp_maxtries), 0,
108 	"ARP resolution attempts before returning error");
109 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
110 	&VNET_NAME(useloopback), 0,
111 	"Use the loopback interface for local traffic");
112 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
113 	&VNET_NAME(arp_proxyall), 0,
114 	"Enable proxy ARP for all suitable requests");
115 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW,
116 	&VNET_NAME(arpt_down), 0,
117 	"Incomplete ARP entry lifetime in seconds");
118 SYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW,
119 	&VNET_NAME(arpstat), arpstat,
120 	"ARP statistics (struct arpstat, net/if_arp.h)");
121 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW,
122 	&VNET_NAME(arp_maxhold), 0,
123 	"Number of packets to hold per ARP entry");
124 
125 static void	arp_init(void);
126 static void	arpintr(struct mbuf *);
127 static void	arptimer(void *);
128 #ifdef INET
129 static void	in_arpinput(struct mbuf *);
130 #endif
131 
132 static const struct netisr_handler arp_nh = {
133 	.nh_name = "arp",
134 	.nh_handler = arpintr,
135 	.nh_proto = NETISR_ARP,
136 	.nh_policy = NETISR_POLICY_SOURCE,
137 };
138 
139 #ifdef AF_INET
140 /*
141  * called by in_ifscrub to remove entry from the table when
142  * the interface goes away
143  */
144 void
145 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
146 {
147 	struct sockaddr_in addr4;
148 
149 	bzero((void *)&addr4, sizeof(addr4));
150 	addr4.sin_len    = sizeof(addr4);
151 	addr4.sin_family = AF_INET;
152 	addr4.sin_addr.s_addr = addr;
153 	IF_AFDATA_LOCK(ifp);
154 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
155 	    (struct sockaddr *)&addr4);
156 	IF_AFDATA_UNLOCK(ifp);
157 }
158 #endif
159 
160 /*
161  * Timeout routine.  Age arp_tab entries periodically.
162  */
163 static void
164 arptimer(void *arg)
165 {
166 	struct llentry *lle = (struct llentry *)arg;
167 	struct ifnet *ifp;
168 	size_t pkts_dropped;
169 
170 	if (lle->la_flags & LLE_STATIC) {
171 		LLE_WUNLOCK(lle);
172 		return;
173 	}
174 
175 	ifp = lle->lle_tbl->llt_ifp;
176 	CURVNET_SET(ifp->if_vnet);
177 
178 	if (lle->la_flags != LLE_DELETED) {
179 		int evt;
180 
181 		if (lle->la_flags & LLE_VALID)
182 			evt = LLENTRY_EXPIRED;
183 		else
184 			evt = LLENTRY_TIMEDOUT;
185 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
186 	}
187 
188 	callout_stop(&lle->la_timer);
189 
190 	/* XXX: LOR avoidance. We still have ref on lle. */
191 	LLE_WUNLOCK(lle);
192 	IF_AFDATA_LOCK(ifp);
193 	LLE_WLOCK(lle);
194 
195 	LLE_REMREF(lle);
196 	pkts_dropped = llentry_free(lle);
197 	IF_AFDATA_UNLOCK(ifp);
198 	ARPSTAT_ADD(dropped, pkts_dropped);
199 	ARPSTAT_INC(timeouts);
200 	CURVNET_RESTORE();
201 }
202 
203 /*
204  * Broadcast an ARP request. Caller specifies:
205  *	- arp header source ip address
206  *	- arp header target ip address
207  *	- arp header source ethernet address
208  */
209 void
210 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
211     u_char *enaddr)
212 {
213 	struct mbuf *m;
214 	struct arphdr *ah;
215 	struct sockaddr sa;
216 	u_char *carpaddr = NULL;
217 
218 	if (sip == NULL) {
219 		/*
220 		 * The caller did not supply a source address, try to find
221 		 * a compatible one among those assigned to this interface.
222 		 */
223 		struct ifaddr *ifa;
224 
225 		IF_ADDR_RLOCK(ifp);
226 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
227 			if (ifa->ifa_addr->sa_family != AF_INET)
228 				continue;
229 
230 			if (ifa->ifa_carp) {
231 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
232 					continue;
233 				sip = &IA_SIN(ifa)->sin_addr;
234 			} else {
235 				carpaddr = NULL;
236 				sip = &IA_SIN(ifa)->sin_addr;
237 			}
238 
239 			if (0 == ((sip->s_addr ^ tip->s_addr) &
240 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
241 				break;  /* found it. */
242 		}
243 		IF_ADDR_RUNLOCK(ifp);
244 		if (sip == NULL) {
245 			printf("%s: cannot find matching address\n", __func__);
246 			return;
247 		}
248 	}
249 	if (enaddr == NULL)
250 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
251 
252 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
253 		return;
254 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
255 		2*ifp->if_data.ifi_addrlen;
256 	m->m_pkthdr.len = m->m_len;
257 	MH_ALIGN(m, m->m_len);
258 	ah = mtod(m, struct arphdr *);
259 	bzero((caddr_t)ah, m->m_len);
260 #ifdef MAC
261 	mac_netinet_arp_send(ifp, m);
262 #endif
263 	ah->ar_pro = htons(ETHERTYPE_IP);
264 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
265 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
266 	ah->ar_op = htons(ARPOP_REQUEST);
267 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
268 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
269 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
270 	sa.sa_family = AF_ARP;
271 	sa.sa_len = 2;
272 	m->m_flags |= M_BCAST;
273 	(*ifp->if_output)(ifp, m, &sa, NULL);
274 	ARPSTAT_INC(txrequests);
275 }
276 
277 /*
278  * Resolve an IP address into an ethernet address.
279  * On input:
280  *    ifp is the interface we use
281  *    rt0 is the route to the final destination (possibly useless)
282  *    m is the mbuf. May be NULL if we don't have a packet.
283  *    dst is the next hop,
284  *    desten is where we want the address.
285  *
286  * On success, desten is filled in and the function returns 0;
287  * If the packet must be held pending resolution, we return EWOULDBLOCK
288  * On other errors, we return the corresponding error code.
289  * Note that m_freem() handles NULL.
290  */
291 int
292 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
293 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
294 {
295 	struct llentry *la = 0;
296 	u_int flags = 0;
297 	struct mbuf *curr = NULL;
298 	struct mbuf *next = NULL;
299 	int error, renew;
300 
301 	*lle = NULL;
302 	if (m != NULL) {
303 		if (m->m_flags & M_BCAST) {
304 			/* broadcast */
305 			(void)memcpy(desten,
306 			    ifp->if_broadcastaddr, ifp->if_addrlen);
307 			return (0);
308 		}
309 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
310 			/* multicast */
311 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
312 			return (0);
313 		}
314 	}
315 retry:
316 	IF_AFDATA_RLOCK(ifp);
317 	la = lla_lookup(LLTABLE(ifp), flags, dst);
318 	IF_AFDATA_RUNLOCK(ifp);
319 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
320 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
321 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
322 		IF_AFDATA_WLOCK(ifp);
323 		la = lla_lookup(LLTABLE(ifp), flags, dst);
324 		IF_AFDATA_WUNLOCK(ifp);
325 	}
326 	if (la == NULL) {
327 		if (flags & LLE_CREATE)
328 			log(LOG_DEBUG,
329 			    "arpresolve: can't allocate llinfo for %s\n",
330 			    inet_ntoa(SIN(dst)->sin_addr));
331 		m_freem(m);
332 		return (EINVAL);
333 	}
334 
335 	if ((la->la_flags & LLE_VALID) &&
336 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
337 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
338 		/*
339 		 * If entry has an expiry time and it is approaching,
340 		 * see if we need to send an ARP request within this
341 		 * arpt_down interval.
342 		 */
343 		if (!(la->la_flags & LLE_STATIC) &&
344 		    time_uptime + la->la_preempt > la->la_expire) {
345 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
346 			la->la_preempt--;
347 		}
348 
349 		*lle = la;
350 		error = 0;
351 		goto done;
352 	}
353 
354 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
355 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
356 		    inet_ntoa(SIN(dst)->sin_addr));
357 		m_freem(m);
358 		error = EINVAL;
359 		goto done;
360 	}
361 
362 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
363 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
364 		flags |= LLE_EXCLUSIVE;
365 		LLE_RUNLOCK(la);
366 		goto retry;
367 	}
368 	/*
369 	 * There is an arptab entry, but no ethernet address
370 	 * response yet.  Add the mbuf to the list, dropping
371 	 * the oldest packet if we have exceeded the system
372 	 * setting.
373 	 */
374 	if (m != NULL) {
375 		if (la->la_numheld >= V_arp_maxhold) {
376 			if (la->la_hold != NULL) {
377 				next = la->la_hold->m_nextpkt;
378 				m_freem(la->la_hold);
379 				la->la_hold = next;
380 				la->la_numheld--;
381 				ARPSTAT_INC(dropped);
382 			}
383 		}
384 		if (la->la_hold != NULL) {
385 			curr = la->la_hold;
386 			while (curr->m_nextpkt != NULL)
387 				curr = curr->m_nextpkt;
388 			curr->m_nextpkt = m;
389 		} else
390 			la->la_hold = m;
391 		la->la_numheld++;
392 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
393 			flags &= ~LLE_EXCLUSIVE;
394 			LLE_DOWNGRADE(la);
395 		}
396 
397 	}
398 	/*
399 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
400 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
401 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
402 	 * ARP request, but not faster than one request per second.
403 	 */
404 	if (la->la_asked < V_arp_maxtries)
405 		error = EWOULDBLOCK;	/* First request. */
406 	else
407 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
408 		    EHOSTUNREACH : EHOSTDOWN;
409 
410 	if (renew) {
411 		int canceled;
412 
413 		LLE_ADDREF(la);
414 		la->la_expire = time_uptime;
415 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
416 		    arptimer, la);
417 		if (canceled)
418 			LLE_REMREF(la);
419 		la->la_asked++;
420 		LLE_WUNLOCK(la);
421 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
422 		return (error);
423 	}
424 done:
425 	if (flags & LLE_EXCLUSIVE)
426 		LLE_WUNLOCK(la);
427 	else
428 		LLE_RUNLOCK(la);
429 	return (error);
430 }
431 
432 /*
433  * Common length and type checks are done here,
434  * then the protocol-specific routine is called.
435  */
436 static void
437 arpintr(struct mbuf *m)
438 {
439 	struct arphdr *ar;
440 
441 	if (m->m_len < sizeof(struct arphdr) &&
442 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
443 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
444 		return;
445 	}
446 	ar = mtod(m, struct arphdr *);
447 
448 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
449 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
450 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
451 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
452 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
453 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
454 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
455 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
456 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
457 		m_freem(m);
458 		return;
459 	}
460 
461 	if (m->m_len < arphdr_len(ar)) {
462 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
463 			log(LOG_NOTICE, "arp: runt packet\n");
464 			m_freem(m);
465 			return;
466 		}
467 		ar = mtod(m, struct arphdr *);
468 	}
469 
470 	ARPSTAT_INC(received);
471 	switch (ntohs(ar->ar_pro)) {
472 #ifdef INET
473 	case ETHERTYPE_IP:
474 		in_arpinput(m);
475 		return;
476 #endif
477 	}
478 	m_freem(m);
479 }
480 
481 #ifdef INET
482 /*
483  * ARP for Internet protocols on 10 Mb/s Ethernet.
484  * Algorithm is that given in RFC 826.
485  * In addition, a sanity check is performed on the sender
486  * protocol address, to catch impersonators.
487  * We no longer handle negotiations for use of trailer protocol:
488  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
489  * along with IP replies if we wanted trailers sent to us,
490  * and also sent them in response to IP replies.
491  * This allowed either end to announce the desire to receive
492  * trailer packets.
493  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
494  * but formerly didn't normally send requests.
495  */
496 static int log_arp_wrong_iface = 1;
497 static int log_arp_movements = 1;
498 static int log_arp_permanent_modify = 1;
499 static int allow_multicast = 0;
500 
501 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
502 	&log_arp_wrong_iface, 0,
503 	"log arp packets arriving on the wrong interface");
504 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
505 	&log_arp_movements, 0,
506 	"log arp replies from MACs different than the one in the cache");
507 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
508 	&log_arp_permanent_modify, 0,
509 	"log arp replies from MACs different than the one in the permanent arp entry");
510 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
511 	&allow_multicast, 0, "accept multicast addresses");
512 
513 static void
514 in_arpinput(struct mbuf *m)
515 {
516 	struct arphdr *ah;
517 	struct ifnet *ifp = m->m_pkthdr.rcvif;
518 	struct llentry *la = NULL;
519 	struct rtentry *rt;
520 	struct ifaddr *ifa;
521 	struct in_ifaddr *ia;
522 	struct sockaddr sa;
523 	struct in_addr isaddr, itaddr, myaddr;
524 	u_int8_t *enaddr = NULL;
525 	int op, flags;
526 	int req_len;
527 	int bridged = 0, is_bridge = 0;
528 	int carped;
529 	struct sockaddr_in sin;
530 	sin.sin_len = sizeof(struct sockaddr_in);
531 	sin.sin_family = AF_INET;
532 	sin.sin_addr.s_addr = 0;
533 
534 	if (ifp->if_bridge)
535 		bridged = 1;
536 	if (ifp->if_type == IFT_BRIDGE)
537 		is_bridge = 1;
538 
539 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
540 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
541 		log(LOG_NOTICE, "in_arp: runt packet -- m_pullup failed\n");
542 		return;
543 	}
544 
545 	ah = mtod(m, struct arphdr *);
546 	/*
547 	 * ARP is only for IPv4 so we can reject packets with
548 	 * a protocol length not equal to an IPv4 address.
549 	 */
550 	if (ah->ar_pln != sizeof(struct in_addr)) {
551 		log(LOG_NOTICE, "in_arp: requested protocol length != %zu\n",
552 		    sizeof(struct in_addr));
553 		return;
554 	}
555 
556 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
557 		log(LOG_NOTICE, "arp: %*D is multicast\n",
558 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
559 		return;
560 	}
561 
562 	op = ntohs(ah->ar_op);
563 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
564 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
565 
566 	if (op == ARPOP_REPLY)
567 		ARPSTAT_INC(rxreplies);
568 
569 	/*
570 	 * For a bridge, we want to check the address irrespective
571 	 * of the receive interface. (This will change slightly
572 	 * when we have clusters of interfaces).
573 	 */
574 	IN_IFADDR_RLOCK();
575 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
576 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
577 		    ia->ia_ifp == ifp) &&
578 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
579 		    (ia->ia_ifa.ifa_carp == NULL ||
580 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
581 			ifa_ref(&ia->ia_ifa);
582 			IN_IFADDR_RUNLOCK();
583 			goto match;
584 		}
585 	}
586 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
587 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
588 		    ia->ia_ifp == ifp) &&
589 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
590 			ifa_ref(&ia->ia_ifa);
591 			IN_IFADDR_RUNLOCK();
592 			goto match;
593 		}
594 
595 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
596   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
597   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
598   addr == ia->ia_addr.sin_addr.s_addr)
599 	/*
600 	 * Check the case when bridge shares its MAC address with
601 	 * some of its children, so packets are claimed by bridge
602 	 * itself (bridge_input() does it first), but they are really
603 	 * meant to be destined to the bridge member.
604 	 */
605 	if (is_bridge) {
606 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
607 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
608 				ifa_ref(&ia->ia_ifa);
609 				ifp = ia->ia_ifp;
610 				IN_IFADDR_RUNLOCK();
611 				goto match;
612 			}
613 		}
614 	}
615 #undef BDG_MEMBER_MATCHES_ARP
616 	IN_IFADDR_RUNLOCK();
617 
618 	/*
619 	 * No match, use the first inet address on the receive interface
620 	 * as a dummy address for the rest of the function.
621 	 */
622 	IF_ADDR_RLOCK(ifp);
623 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
624 		if (ifa->ifa_addr->sa_family == AF_INET &&
625 		    (ifa->ifa_carp == NULL ||
626 		    (*carp_iamatch_p)(ifa, &enaddr))) {
627 			ia = ifatoia(ifa);
628 			ifa_ref(ifa);
629 			IF_ADDR_RUNLOCK(ifp);
630 			goto match;
631 		}
632 	IF_ADDR_RUNLOCK(ifp);
633 
634 	/*
635 	 * If bridging, fall back to using any inet address.
636 	 */
637 	IN_IFADDR_RLOCK();
638 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
639 		IN_IFADDR_RUNLOCK();
640 		goto drop;
641 	}
642 	ifa_ref(&ia->ia_ifa);
643 	IN_IFADDR_RUNLOCK();
644 match:
645 	if (!enaddr)
646 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
647 	carped = (ia->ia_ifa.ifa_carp != NULL);
648 	myaddr = ia->ia_addr.sin_addr;
649 	ifa_free(&ia->ia_ifa);
650 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
651 		goto drop;	/* it's from me, ignore it. */
652 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
653 		log(LOG_NOTICE,
654 		    "arp: link address is broadcast for IP address %s!\n",
655 		    inet_ntoa(isaddr));
656 		goto drop;
657 	}
658 	/*
659 	 * Warn if another host is using the same IP address, but only if the
660 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
661 	 * case we suppress the warning to avoid false positive complaints of
662 	 * potential misconfiguration.
663 	 */
664 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
665 	    myaddr.s_addr != 0) {
666 		log(LOG_ERR, "arp: %*D is using my IP address %s on %s!\n",
667 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
668 		   inet_ntoa(isaddr), ifp->if_xname);
669 		itaddr = myaddr;
670 		ARPSTAT_INC(dupips);
671 		goto reply;
672 	}
673 	if (ifp->if_flags & IFF_STATICARP)
674 		goto reply;
675 
676 	bzero(&sin, sizeof(sin));
677 	sin.sin_len = sizeof(struct sockaddr_in);
678 	sin.sin_family = AF_INET;
679 	sin.sin_addr = isaddr;
680 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
681 	flags |= LLE_EXCLUSIVE;
682 	IF_AFDATA_LOCK(ifp);
683 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
684 	IF_AFDATA_UNLOCK(ifp);
685 	if (la != NULL) {
686 		/* the following is not an error when doing bridging */
687 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
688 			if (log_arp_wrong_iface)
689 				log(LOG_WARNING, "arp: %s is on %s "
690 				    "but got reply from %*D on %s\n",
691 				    inet_ntoa(isaddr),
692 				    la->lle_tbl->llt_ifp->if_xname,
693 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
694 				    ifp->if_xname);
695 			LLE_WUNLOCK(la);
696 			goto reply;
697 		}
698 		if ((la->la_flags & LLE_VALID) &&
699 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
700 			if (la->la_flags & LLE_STATIC) {
701 				LLE_WUNLOCK(la);
702 				if (log_arp_permanent_modify)
703 					log(LOG_ERR,
704 					    "arp: %*D attempts to modify "
705 					    "permanent entry for %s on %s\n",
706 					    ifp->if_addrlen,
707 					    (u_char *)ar_sha(ah), ":",
708 					    inet_ntoa(isaddr), ifp->if_xname);
709 				goto reply;
710 			}
711 			if (log_arp_movements) {
712 				log(LOG_INFO, "arp: %s moved from %*D "
713 				    "to %*D on %s\n",
714 				    inet_ntoa(isaddr),
715 				    ifp->if_addrlen,
716 				    (u_char *)&la->ll_addr, ":",
717 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
718 				    ifp->if_xname);
719 			}
720 		}
721 
722 		if (ifp->if_addrlen != ah->ar_hln) {
723 			LLE_WUNLOCK(la);
724 			log(LOG_WARNING, "arp from %*D: addr len: new %d, "
725 			    "i/f %d (ignored)\n", ifp->if_addrlen,
726 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
727 			    ifp->if_addrlen);
728 			goto drop;
729 		}
730 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
731 		la->la_flags |= LLE_VALID;
732 
733 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
734 
735 		if (!(la->la_flags & LLE_STATIC)) {
736 			int canceled;
737 
738 			LLE_ADDREF(la);
739 			la->la_expire = time_uptime + V_arpt_keep;
740 			canceled = callout_reset(&la->la_timer,
741 			    hz * V_arpt_keep, arptimer, la);
742 			if (canceled)
743 				LLE_REMREF(la);
744 		}
745 		la->la_asked = 0;
746 		la->la_preempt = V_arp_maxtries;
747 		/*
748 		 * The packets are all freed within the call to the output
749 		 * routine.
750 		 *
751 		 * NB: The lock MUST be released before the call to the
752 		 * output routine.
753 		 */
754 		if (la->la_hold != NULL) {
755 			struct mbuf *m_hold, *m_hold_next;
756 
757 			m_hold = la->la_hold;
758 			la->la_hold = NULL;
759 			la->la_numheld = 0;
760 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
761 			LLE_WUNLOCK(la);
762 			for (; m_hold != NULL; m_hold = m_hold_next) {
763 				m_hold_next = m_hold->m_nextpkt;
764 				m_hold->m_nextpkt = NULL;
765 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
766 			}
767 		} else
768 			LLE_WUNLOCK(la);
769 	}
770 reply:
771 	if (op != ARPOP_REQUEST)
772 		goto drop;
773 	ARPSTAT_INC(rxrequests);
774 
775 	if (itaddr.s_addr == myaddr.s_addr) {
776 		/* Shortcut.. the receiving interface is the target. */
777 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
778 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
779 	} else {
780 		struct llentry *lle = NULL;
781 
782 		sin.sin_addr = itaddr;
783 		IF_AFDATA_LOCK(ifp);
784 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
785 		IF_AFDATA_UNLOCK(ifp);
786 
787 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
788 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
789 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
790 			LLE_RUNLOCK(lle);
791 		} else {
792 
793 			if (lle != NULL)
794 				LLE_RUNLOCK(lle);
795 
796 			if (!V_arp_proxyall)
797 				goto drop;
798 
799 			sin.sin_addr = itaddr;
800 			/* XXX MRT use table 0 for arp reply  */
801 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
802 			if (!rt)
803 				goto drop;
804 
805 			/*
806 			 * Don't send proxies for nodes on the same interface
807 			 * as this one came out of, or we'll get into a fight
808 			 * over who claims what Ether address.
809 			 */
810 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
811 				RTFREE_LOCKED(rt);
812 				goto drop;
813 			}
814 			RTFREE_LOCKED(rt);
815 
816 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
817 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
818 
819 			/*
820 			 * Also check that the node which sent the ARP packet
821 			 * is on the interface we expect it to be on. This
822 			 * avoids ARP chaos if an interface is connected to the
823 			 * wrong network.
824 			 */
825 			sin.sin_addr = isaddr;
826 
827 			/* XXX MRT use table 0 for arp checks */
828 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
829 			if (!rt)
830 				goto drop;
831 			if (rt->rt_ifp != ifp) {
832 				log(LOG_INFO, "arp_proxy: ignoring request"
833 				    " from %s via %s, expecting %s\n",
834 				    inet_ntoa(isaddr), ifp->if_xname,
835 				    rt->rt_ifp->if_xname);
836 				RTFREE_LOCKED(rt);
837 				goto drop;
838 			}
839 			RTFREE_LOCKED(rt);
840 
841 #ifdef DEBUG_PROXY
842 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
843 #endif
844 		}
845 	}
846 
847 	if (itaddr.s_addr == myaddr.s_addr &&
848 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
849 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
850 #ifdef DEBUG_LINKLOCAL
851 		printf("arp: sending reply for link-local addr %s\n",
852 		    inet_ntoa(itaddr));
853 #endif
854 		m->m_flags |= M_BCAST;
855 		m->m_flags &= ~M_MCAST;
856 	} else {
857 		/* default behaviour; never reply by broadcast. */
858 		m->m_flags &= ~(M_BCAST|M_MCAST);
859 	}
860 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
861 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
862 	ah->ar_op = htons(ARPOP_REPLY);
863 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
864 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
865 	m->m_pkthdr.len = m->m_len;
866 	m->m_pkthdr.rcvif = NULL;
867 	sa.sa_family = AF_ARP;
868 	sa.sa_len = 2;
869 	(*ifp->if_output)(ifp, m, &sa, NULL);
870 	ARPSTAT_INC(txreplies);
871 	return;
872 
873 drop:
874 	m_freem(m);
875 }
876 #endif
877 
878 void
879 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
880 {
881 	struct llentry *lle;
882 
883 	if (ifa->ifa_carp != NULL)
884 		return;
885 
886 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
887 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
888 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
889 		/*
890 		 * interface address is considered static entry
891 		 * because the output of the arp utility shows
892 		 * that L2 entry as permanent
893 		 */
894 		IF_AFDATA_LOCK(ifp);
895 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
896 				 (struct sockaddr *)IA_SIN(ifa));
897 		IF_AFDATA_UNLOCK(ifp);
898 		if (lle == NULL)
899 			log(LOG_INFO, "arp_ifinit: cannot create arp "
900 			    "entry for interface address\n");
901 		else
902 			LLE_RUNLOCK(lle);
903 	}
904 	ifa->ifa_rtrequest = NULL;
905 }
906 
907 void
908 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
909 {
910 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
911 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
912 				&IA_SIN(ifa)->sin_addr, enaddr);
913 	ifa->ifa_rtrequest = NULL;
914 }
915 
916 static void
917 arp_init(void)
918 {
919 
920 	netisr_register(&arp_nh);
921 }
922 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
923