xref: /freebsd/sys/netinet/if_ether.c (revision 641a6cfb86023499caafe26a4d821a0b885cf00b)
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 ifnet *ifp;
167 	struct llentry   *lle;
168 	int pkts_dropped;
169 
170 	KASSERT(arg != NULL, ("%s: arg NULL", __func__));
171 	lle = (struct llentry *)arg;
172 	ifp = lle->lle_tbl->llt_ifp;
173 	CURVNET_SET(ifp->if_vnet);
174 	IF_AFDATA_LOCK(ifp);
175 	LLE_WLOCK(lle);
176 	if (lle->la_flags & LLE_STATIC)
177 		LLE_WUNLOCK(lle);
178 	else {
179 		if (!callout_pending(&lle->la_timer) &&
180 		    callout_active(&lle->la_timer)) {
181 			callout_stop(&lle->la_timer);
182 			LLE_REMREF(lle);
183 
184 			if (lle->la_flags != LLE_DELETED) {
185 				int evt;
186 
187 				if (lle->la_flags & LLE_VALID)
188 					evt = LLENTRY_EXPIRED;
189 				else
190 					evt = LLENTRY_TIMEDOUT;
191 				EVENTHANDLER_INVOKE(lle_event, lle, evt);
192 			}
193 
194 			pkts_dropped = llentry_free(lle);
195 			ARPSTAT_ADD(dropped, pkts_dropped);
196 			ARPSTAT_INC(timeouts);
197 		} else {
198 #ifdef DIAGNOSTIC
199 			struct sockaddr *l3addr = L3_ADDR(lle);
200 			log(LOG_INFO,
201 			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
202 			    inet_ntoa(
203 			        ((const struct sockaddr_in *)l3addr)->sin_addr));
204 #endif
205 			LLE_WUNLOCK(lle);
206 		}
207 	}
208 	IF_AFDATA_UNLOCK(ifp);
209 	CURVNET_RESTORE();
210 }
211 
212 /*
213  * Broadcast an ARP request. Caller specifies:
214  *	- arp header source ip address
215  *	- arp header target ip address
216  *	- arp header source ethernet address
217  */
218 void
219 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
220     u_char *enaddr)
221 {
222 	struct mbuf *m;
223 	struct arphdr *ah;
224 	struct sockaddr sa;
225 	u_char *carpaddr = NULL;
226 
227 	if (sip == NULL) {
228 		/*
229 		 * The caller did not supply a source address, try to find
230 		 * a compatible one among those assigned to this interface.
231 		 */
232 		struct ifaddr *ifa;
233 
234 		IF_ADDR_RLOCK(ifp);
235 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
236 			if (ifa->ifa_addr->sa_family != AF_INET)
237 				continue;
238 
239 			if (ifa->ifa_carp) {
240 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
241 					continue;
242 				sip = &IA_SIN(ifa)->sin_addr;
243 			} else {
244 				carpaddr = NULL;
245 				sip = &IA_SIN(ifa)->sin_addr;
246 			}
247 
248 			if (0 == ((sip->s_addr ^ tip->s_addr) &
249 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
250 				break;  /* found it. */
251 		}
252 		IF_ADDR_RUNLOCK(ifp);
253 		if (sip == NULL) {
254 			printf("%s: cannot find matching address\n", __func__);
255 			return;
256 		}
257 	}
258 	if (enaddr == NULL)
259 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
260 
261 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
262 		return;
263 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
264 		2*ifp->if_data.ifi_addrlen;
265 	m->m_pkthdr.len = m->m_len;
266 	MH_ALIGN(m, m->m_len);
267 	ah = mtod(m, struct arphdr *);
268 	bzero((caddr_t)ah, m->m_len);
269 #ifdef MAC
270 	mac_netinet_arp_send(ifp, m);
271 #endif
272 	ah->ar_pro = htons(ETHERTYPE_IP);
273 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
274 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
275 	ah->ar_op = htons(ARPOP_REQUEST);
276 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
277 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
278 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
279 	sa.sa_family = AF_ARP;
280 	sa.sa_len = 2;
281 	m->m_flags |= M_BCAST;
282 	(*ifp->if_output)(ifp, m, &sa, NULL);
283 	ARPSTAT_INC(txrequests);
284 }
285 
286 /*
287  * Resolve an IP address into an ethernet address.
288  * On input:
289  *    ifp is the interface we use
290  *    rt0 is the route to the final destination (possibly useless)
291  *    m is the mbuf. May be NULL if we don't have a packet.
292  *    dst is the next hop,
293  *    desten is where we want the address.
294  *
295  * On success, desten is filled in and the function returns 0;
296  * If the packet must be held pending resolution, we return EWOULDBLOCK
297  * On other errors, we return the corresponding error code.
298  * Note that m_freem() handles NULL.
299  */
300 int
301 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
302 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
303 {
304 	struct llentry *la = 0;
305 	u_int flags = 0;
306 	struct mbuf *curr = NULL;
307 	struct mbuf *next = NULL;
308 	int error, renew;
309 
310 	*lle = NULL;
311 	if (m != NULL) {
312 		if (m->m_flags & M_BCAST) {
313 			/* broadcast */
314 			(void)memcpy(desten,
315 			    ifp->if_broadcastaddr, ifp->if_addrlen);
316 			return (0);
317 		}
318 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
319 			/* multicast */
320 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
321 			return (0);
322 		}
323 	}
324 retry:
325 	IF_AFDATA_RLOCK(ifp);
326 	la = lla_lookup(LLTABLE(ifp), flags, dst);
327 	IF_AFDATA_RUNLOCK(ifp);
328 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
329 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
330 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
331 		IF_AFDATA_WLOCK(ifp);
332 		la = lla_lookup(LLTABLE(ifp), flags, dst);
333 		IF_AFDATA_WUNLOCK(ifp);
334 	}
335 	if (la == NULL) {
336 		if (flags & LLE_CREATE)
337 			log(LOG_DEBUG,
338 			    "arpresolve: can't allocate llinfo for %s\n",
339 			    inet_ntoa(SIN(dst)->sin_addr));
340 		m_freem(m);
341 		return (EINVAL);
342 	}
343 
344 	if ((la->la_flags & LLE_VALID) &&
345 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
346 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
347 		/*
348 		 * If entry has an expiry time and it is approaching,
349 		 * see if we need to send an ARP request within this
350 		 * arpt_down interval.
351 		 */
352 		if (!(la->la_flags & LLE_STATIC) &&
353 		    time_uptime + la->la_preempt > la->la_expire) {
354 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
355 			la->la_preempt--;
356 		}
357 
358 		*lle = la;
359 		error = 0;
360 		goto done;
361 	}
362 
363 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
364 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
365 		    inet_ntoa(SIN(dst)->sin_addr));
366 		m_freem(m);
367 		error = EINVAL;
368 		goto done;
369 	}
370 
371 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
372 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
373 		flags |= LLE_EXCLUSIVE;
374 		LLE_RUNLOCK(la);
375 		goto retry;
376 	}
377 	/*
378 	 * There is an arptab entry, but no ethernet address
379 	 * response yet.  Add the mbuf to the list, dropping
380 	 * the oldest packet if we have exceeded the system
381 	 * setting.
382 	 */
383 	if (m != NULL) {
384 		if (la->la_numheld >= V_arp_maxhold) {
385 			if (la->la_hold != NULL) {
386 				next = la->la_hold->m_nextpkt;
387 				m_freem(la->la_hold);
388 				la->la_hold = next;
389 				la->la_numheld--;
390 				ARPSTAT_INC(dropped);
391 			}
392 		}
393 		if (la->la_hold != NULL) {
394 			curr = la->la_hold;
395 			while (curr->m_nextpkt != NULL)
396 				curr = curr->m_nextpkt;
397 			curr->m_nextpkt = m;
398 		} else
399 			la->la_hold = m;
400 		la->la_numheld++;
401 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
402 			flags &= ~LLE_EXCLUSIVE;
403 			LLE_DOWNGRADE(la);
404 		}
405 
406 	}
407 	/*
408 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
409 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
410 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
411 	 * ARP request, but not faster than one request per second.
412 	 */
413 	if (la->la_asked < V_arp_maxtries)
414 		error = EWOULDBLOCK;	/* First request. */
415 	else
416 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
417 		    EHOSTUNREACH : EHOSTDOWN;
418 
419 	if (renew) {
420 		int canceled;
421 
422 		LLE_ADDREF(la);
423 		la->la_expire = time_uptime;
424 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
425 		    arptimer, la);
426 		if (canceled)
427 			LLE_REMREF(la);
428 		la->la_asked++;
429 		LLE_WUNLOCK(la);
430 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
431 		return (error);
432 	}
433 done:
434 	if (flags & LLE_EXCLUSIVE)
435 		LLE_WUNLOCK(la);
436 	else
437 		LLE_RUNLOCK(la);
438 	return (error);
439 }
440 
441 /*
442  * Common length and type checks are done here,
443  * then the protocol-specific routine is called.
444  */
445 static void
446 arpintr(struct mbuf *m)
447 {
448 	struct arphdr *ar;
449 
450 	if (m->m_len < sizeof(struct arphdr) &&
451 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
452 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
453 		return;
454 	}
455 	ar = mtod(m, struct arphdr *);
456 
457 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
458 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
459 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
460 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
461 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
462 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
463 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
464 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
465 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
466 		m_freem(m);
467 		return;
468 	}
469 
470 	if (m->m_len < arphdr_len(ar)) {
471 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
472 			log(LOG_NOTICE, "arp: runt packet\n");
473 			m_freem(m);
474 			return;
475 		}
476 		ar = mtod(m, struct arphdr *);
477 	}
478 
479 	ARPSTAT_INC(received);
480 	switch (ntohs(ar->ar_pro)) {
481 #ifdef INET
482 	case ETHERTYPE_IP:
483 		in_arpinput(m);
484 		return;
485 #endif
486 	}
487 	m_freem(m);
488 }
489 
490 #ifdef INET
491 /*
492  * ARP for Internet protocols on 10 Mb/s Ethernet.
493  * Algorithm is that given in RFC 826.
494  * In addition, a sanity check is performed on the sender
495  * protocol address, to catch impersonators.
496  * We no longer handle negotiations for use of trailer protocol:
497  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
498  * along with IP replies if we wanted trailers sent to us,
499  * and also sent them in response to IP replies.
500  * This allowed either end to announce the desire to receive
501  * trailer packets.
502  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
503  * but formerly didn't normally send requests.
504  */
505 static int log_arp_wrong_iface = 1;
506 static int log_arp_movements = 1;
507 static int log_arp_permanent_modify = 1;
508 
509 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
510 	&log_arp_wrong_iface, 0,
511 	"log arp packets arriving on the wrong interface");
512 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
513         &log_arp_movements, 0,
514         "log arp replies from MACs different than the one in the cache");
515 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
516         &log_arp_permanent_modify, 0,
517         "log arp replies from MACs different than the one in the permanent arp entry");
518 
519 
520 static void
521 in_arpinput(struct mbuf *m)
522 {
523 	struct arphdr *ah;
524 	struct ifnet *ifp = m->m_pkthdr.rcvif;
525 	struct llentry *la = NULL;
526 	struct rtentry *rt;
527 	struct ifaddr *ifa;
528 	struct in_ifaddr *ia;
529 	struct sockaddr sa;
530 	struct in_addr isaddr, itaddr, myaddr;
531 	u_int8_t *enaddr = NULL;
532 	int op, flags;
533 	int req_len;
534 	int bridged = 0, is_bridge = 0;
535 	int carped;
536 	struct sockaddr_in sin;
537 	sin.sin_len = sizeof(struct sockaddr_in);
538 	sin.sin_family = AF_INET;
539 	sin.sin_addr.s_addr = 0;
540 
541 	if (ifp->if_bridge)
542 		bridged = 1;
543 	if (ifp->if_type == IFT_BRIDGE)
544 		is_bridge = 1;
545 
546 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
547 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
548 		log(LOG_NOTICE, "in_arp: runt packet -- m_pullup failed\n");
549 		return;
550 	}
551 
552 	ah = mtod(m, struct arphdr *);
553 	/*
554 	 * ARP is only for IPv4 so we can reject packets with
555 	 * a protocol length not equal to an IPv4 address.
556 	 */
557 	if (ah->ar_pln != sizeof(struct in_addr)) {
558 		log(LOG_NOTICE, "in_arp: requested protocol length != %zu\n",
559 		    sizeof(struct in_addr));
560 		return;
561 	}
562 
563 	if (ETHER_IS_MULTICAST(ar_sha(ah))) {
564 		log(LOG_NOTICE, "in_arp: %*D is multicast\n",
565 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
566 		return;
567 	}
568 
569 	op = ntohs(ah->ar_op);
570 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
571 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
572 
573 	if (op == ARPOP_REPLY)
574 		ARPSTAT_INC(rxreplies);
575 
576 	/*
577 	 * For a bridge, we want to check the address irrespective
578 	 * of the receive interface. (This will change slightly
579 	 * when we have clusters of interfaces).
580 	 */
581 	IN_IFADDR_RLOCK();
582 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
583 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
584 		    ia->ia_ifp == ifp) &&
585 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
586 		    (ia->ia_ifa.ifa_carp == NULL ||
587 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
588 			ifa_ref(&ia->ia_ifa);
589 			IN_IFADDR_RUNLOCK();
590 			goto match;
591 		}
592 	}
593 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
594 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
595 		    ia->ia_ifp == ifp) &&
596 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
597 			ifa_ref(&ia->ia_ifa);
598 			IN_IFADDR_RUNLOCK();
599 			goto match;
600 		}
601 
602 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
603   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
604   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
605   addr == ia->ia_addr.sin_addr.s_addr)
606 	/*
607 	 * Check the case when bridge shares its MAC address with
608 	 * some of its children, so packets are claimed by bridge
609 	 * itself (bridge_input() does it first), but they are really
610 	 * meant to be destined to the bridge member.
611 	 */
612 	if (is_bridge) {
613 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
614 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
615 				ifa_ref(&ia->ia_ifa);
616 				ifp = ia->ia_ifp;
617 				IN_IFADDR_RUNLOCK();
618 				goto match;
619 			}
620 		}
621 	}
622 #undef BDG_MEMBER_MATCHES_ARP
623 	IN_IFADDR_RUNLOCK();
624 
625 	/*
626 	 * No match, use the first inet address on the receive interface
627 	 * as a dummy address for the rest of the function.
628 	 */
629 	IF_ADDR_RLOCK(ifp);
630 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
631 		if (ifa->ifa_addr->sa_family == AF_INET &&
632 		    (ifa->ifa_carp == NULL ||
633 		    (*carp_iamatch_p)(ifa, &enaddr))) {
634 			ia = ifatoia(ifa);
635 			ifa_ref(ifa);
636 			IF_ADDR_RUNLOCK(ifp);
637 			goto match;
638 		}
639 	IF_ADDR_RUNLOCK(ifp);
640 
641 	/*
642 	 * If bridging, fall back to using any inet address.
643 	 */
644 	IN_IFADDR_RLOCK();
645 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
646 		IN_IFADDR_RUNLOCK();
647 		goto drop;
648 	}
649 	ifa_ref(&ia->ia_ifa);
650 	IN_IFADDR_RUNLOCK();
651 match:
652 	if (!enaddr)
653 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
654 	carped = (ia->ia_ifa.ifa_carp != NULL);
655 	myaddr = ia->ia_addr.sin_addr;
656 	ifa_free(&ia->ia_ifa);
657 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
658 		goto drop;	/* it's from me, ignore it. */
659 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
660 		log(LOG_NOTICE,
661 		    "arp: link address is broadcast for IP address %s!\n",
662 		    inet_ntoa(isaddr));
663 		goto drop;
664 	}
665 	/*
666 	 * Warn if another host is using the same IP address, but only if the
667 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
668 	 * case we suppress the warning to avoid false positive complaints of
669 	 * potential misconfiguration.
670 	 */
671 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
672 	    myaddr.s_addr != 0) {
673 		log(LOG_ERR, "arp: %*D is using my IP address %s on %s!\n",
674 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
675 		   inet_ntoa(isaddr), ifp->if_xname);
676 		itaddr = myaddr;
677 		ARPSTAT_INC(dupips);
678 		goto reply;
679 	}
680 	if (ifp->if_flags & IFF_STATICARP)
681 		goto reply;
682 
683 	bzero(&sin, sizeof(sin));
684 	sin.sin_len = sizeof(struct sockaddr_in);
685 	sin.sin_family = AF_INET;
686 	sin.sin_addr = isaddr;
687 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
688 	flags |= LLE_EXCLUSIVE;
689 	IF_AFDATA_LOCK(ifp);
690 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
691 	IF_AFDATA_UNLOCK(ifp);
692 	if (la != NULL) {
693 		/* the following is not an error when doing bridging */
694 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
695 			if (log_arp_wrong_iface)
696 				log(LOG_WARNING, "arp: %s is on %s "
697 				    "but got reply from %*D on %s\n",
698 				    inet_ntoa(isaddr),
699 				    la->lle_tbl->llt_ifp->if_xname,
700 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
701 				    ifp->if_xname);
702 			LLE_WUNLOCK(la);
703 			goto reply;
704 		}
705 		if ((la->la_flags & LLE_VALID) &&
706 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
707 			if (la->la_flags & LLE_STATIC) {
708 				LLE_WUNLOCK(la);
709 				if (log_arp_permanent_modify)
710 					log(LOG_ERR,
711 					    "arp: %*D attempts to modify "
712 					    "permanent entry for %s on %s\n",
713 					    ifp->if_addrlen,
714 					    (u_char *)ar_sha(ah), ":",
715 					    inet_ntoa(isaddr), ifp->if_xname);
716 				goto reply;
717 			}
718 			if (log_arp_movements) {
719 			        log(LOG_INFO, "arp: %s moved from %*D "
720 				    "to %*D on %s\n",
721 				    inet_ntoa(isaddr),
722 				    ifp->if_addrlen,
723 				    (u_char *)&la->ll_addr, ":",
724 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
725 				    ifp->if_xname);
726 			}
727 		}
728 
729 		if (ifp->if_addrlen != ah->ar_hln) {
730 			LLE_WUNLOCK(la);
731 			log(LOG_WARNING, "arp from %*D: addr len: new %d, "
732 			    "i/f %d (ignored)\n", ifp->if_addrlen,
733 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
734 			    ifp->if_addrlen);
735 			goto drop;
736 		}
737 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
738 		la->la_flags |= LLE_VALID;
739 
740 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
741 
742 		if (!(la->la_flags & LLE_STATIC)) {
743 			int canceled;
744 
745 			LLE_ADDREF(la);
746 			la->la_expire = time_uptime + V_arpt_keep;
747 			canceled = callout_reset(&la->la_timer,
748 			    hz * V_arpt_keep, arptimer, la);
749 			if (canceled)
750 				LLE_REMREF(la);
751 		}
752 		la->la_asked = 0;
753 		la->la_preempt = V_arp_maxtries;
754 		/*
755 		 * The packets are all freed within the call to the output
756 		 * routine.
757 		 *
758 		 * NB: The lock MUST be released before the call to the
759 		 * output routine.
760 		 */
761 		if (la->la_hold != NULL) {
762 			struct mbuf *m_hold, *m_hold_next;
763 
764 			m_hold = la->la_hold;
765 			la->la_hold = NULL;
766 			la->la_numheld = 0;
767 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
768 			LLE_WUNLOCK(la);
769 			for (; m_hold != NULL; m_hold = m_hold_next) {
770 				m_hold_next = m_hold->m_nextpkt;
771 				m_hold->m_nextpkt = NULL;
772 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
773 			}
774 		} else
775 			LLE_WUNLOCK(la);
776 	}
777 reply:
778 	if (op != ARPOP_REQUEST)
779 		goto drop;
780 	ARPSTAT_INC(rxrequests);
781 
782 	if (itaddr.s_addr == myaddr.s_addr) {
783 		/* Shortcut.. the receiving interface is the target. */
784 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
785 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
786 	} else {
787 		struct llentry *lle = NULL;
788 
789 		sin.sin_addr = itaddr;
790 		IF_AFDATA_LOCK(ifp);
791 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
792 		IF_AFDATA_UNLOCK(ifp);
793 
794 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
795 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
796 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
797 			LLE_RUNLOCK(lle);
798 		} else {
799 
800 			if (lle != NULL)
801 				LLE_RUNLOCK(lle);
802 
803 			if (!V_arp_proxyall)
804 				goto drop;
805 
806 			sin.sin_addr = itaddr;
807 			/* XXX MRT use table 0 for arp reply  */
808 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
809 			if (!rt)
810 				goto drop;
811 
812 			/*
813 			 * Don't send proxies for nodes on the same interface
814 			 * as this one came out of, or we'll get into a fight
815 			 * over who claims what Ether address.
816 			 */
817 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
818 				RTFREE_LOCKED(rt);
819 				goto drop;
820 			}
821 			RTFREE_LOCKED(rt);
822 
823 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
824 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
825 
826 			/*
827 			 * Also check that the node which sent the ARP packet
828 			 * is on the interface we expect it to be on. This
829 			 * avoids ARP chaos if an interface is connected to the
830 			 * wrong network.
831 			 */
832 			sin.sin_addr = isaddr;
833 
834 			/* XXX MRT use table 0 for arp checks */
835 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
836 			if (!rt)
837 				goto drop;
838 			if (rt->rt_ifp != ifp) {
839 				log(LOG_INFO, "arp_proxy: ignoring request"
840 				    " from %s via %s, expecting %s\n",
841 				    inet_ntoa(isaddr), ifp->if_xname,
842 				    rt->rt_ifp->if_xname);
843 				RTFREE_LOCKED(rt);
844 				goto drop;
845 			}
846 			RTFREE_LOCKED(rt);
847 
848 #ifdef DEBUG_PROXY
849 			printf("arp: proxying for %s\n",
850 			       inet_ntoa(itaddr));
851 #endif
852 		}
853 	}
854 
855 	if (itaddr.s_addr == myaddr.s_addr &&
856 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
857 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
858 #ifdef DEBUG_LINKLOCAL
859 		printf("arp: sending reply for link-local addr %s\n",
860 		    inet_ntoa(itaddr));
861 #endif
862 		m->m_flags |= M_BCAST;
863 		m->m_flags &= ~M_MCAST;
864 	} else {
865 		/* default behaviour; never reply by broadcast. */
866 		m->m_flags &= ~(M_BCAST|M_MCAST);
867 	}
868 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
869 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
870 	ah->ar_op = htons(ARPOP_REPLY);
871 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
872 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
873 	m->m_pkthdr.len = m->m_len;
874 	m->m_pkthdr.rcvif = NULL;
875 	sa.sa_family = AF_ARP;
876 	sa.sa_len = 2;
877 	(*ifp->if_output)(ifp, m, &sa, NULL);
878 	ARPSTAT_INC(txreplies);
879 	return;
880 
881 drop:
882 	m_freem(m);
883 }
884 #endif
885 
886 void
887 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
888 {
889 	struct llentry *lle;
890 
891 	if (ifa->ifa_carp != NULL)
892 		return;
893 
894 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
895 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
896 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
897 		/*
898 		 * interface address is considered static entry
899 		 * because the output of the arp utility shows
900 		 * that L2 entry as permanent
901 		 */
902 		IF_AFDATA_LOCK(ifp);
903 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
904 				 (struct sockaddr *)IA_SIN(ifa));
905 		IF_AFDATA_UNLOCK(ifp);
906 		if (lle == NULL)
907 			log(LOG_INFO, "arp_ifinit: cannot create arp "
908 			    "entry for interface address\n");
909 		else
910 			LLE_RUNLOCK(lle);
911 	}
912 	ifa->ifa_rtrequest = NULL;
913 }
914 
915 void
916 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
917 {
918 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
919 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
920 				&IA_SIN(ifa)->sin_addr, enaddr);
921 	ifa->ifa_rtrequest = NULL;
922 }
923 
924 static void
925 arp_init(void)
926 {
927 
928 	netisr_register(&arp_nh);
929 }
930 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
931