xref: /freebsd/sys/netinet/if_ether.c (revision 47e8d432d5db395e2812c03093b70227e48a1c6d)
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) ((const 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 
169 	if (lle->la_flags & LLE_STATIC) {
170 		LLE_WUNLOCK(lle);
171 		return;
172 	}
173 
174 	ifp = lle->lle_tbl->llt_ifp;
175 	CURVNET_SET(ifp->if_vnet);
176 
177 	if (lle->la_flags != LLE_DELETED) {
178 		int evt;
179 
180 		if (lle->la_flags & LLE_VALID)
181 			evt = LLENTRY_EXPIRED;
182 		else
183 			evt = LLENTRY_TIMEDOUT;
184 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
185 	}
186 
187 	callout_stop(&lle->la_timer);
188 
189 	/* XXX: LOR avoidance. We still have ref on lle. */
190 	LLE_WUNLOCK(lle);
191 	IF_AFDATA_LOCK(ifp);
192 	LLE_WLOCK(lle);
193 
194 	/* Guard against race with other llentry_free(). */
195 	if (lle->la_flags & LLE_LINKED) {
196 		size_t pkts_dropped;
197 
198 		LLE_REMREF(lle);
199 		pkts_dropped = llentry_free(lle);
200 		ARPSTAT_ADD(dropped, pkts_dropped);
201 	} else
202 		LLE_FREE_LOCKED(lle);
203 
204 	IF_AFDATA_UNLOCK(ifp);
205 
206 	ARPSTAT_INC(timeouts);
207 
208 	CURVNET_RESTORE();
209 }
210 
211 /*
212  * Broadcast an ARP request. Caller specifies:
213  *	- arp header source ip address
214  *	- arp header target ip address
215  *	- arp header source ethernet address
216  */
217 void
218 arprequest(struct ifnet *ifp, const struct in_addr *sip,
219     const struct in_addr *tip, u_char *enaddr)
220 {
221 	struct mbuf *m;
222 	struct arphdr *ah;
223 	struct sockaddr sa;
224 	u_char *carpaddr = NULL;
225 
226 	if (sip == NULL) {
227 		/*
228 		 * The caller did not supply a source address, try to find
229 		 * a compatible one among those assigned to this interface.
230 		 */
231 		struct ifaddr *ifa;
232 
233 		IF_ADDR_RLOCK(ifp);
234 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
235 			if (ifa->ifa_addr->sa_family != AF_INET)
236 				continue;
237 
238 			if (ifa->ifa_carp) {
239 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
240 					continue;
241 				sip = &IA_SIN(ifa)->sin_addr;
242 			} else {
243 				carpaddr = NULL;
244 				sip = &IA_SIN(ifa)->sin_addr;
245 			}
246 
247 			if (0 == ((sip->s_addr ^ tip->s_addr) &
248 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
249 				break;  /* found it. */
250 		}
251 		IF_ADDR_RUNLOCK(ifp);
252 		if (sip == NULL) {
253 			printf("%s: cannot find matching address\n", __func__);
254 			return;
255 		}
256 	}
257 	if (enaddr == NULL)
258 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
259 
260 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
261 		return;
262 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
263 		2*ifp->if_data.ifi_addrlen;
264 	m->m_pkthdr.len = m->m_len;
265 	MH_ALIGN(m, m->m_len);
266 	ah = mtod(m, struct arphdr *);
267 	bzero((caddr_t)ah, m->m_len);
268 #ifdef MAC
269 	mac_netinet_arp_send(ifp, m);
270 #endif
271 	ah->ar_pro = htons(ETHERTYPE_IP);
272 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
273 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
274 	ah->ar_op = htons(ARPOP_REQUEST);
275 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
276 	bcopy(sip, ar_spa(ah), ah->ar_pln);
277 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
278 	sa.sa_family = AF_ARP;
279 	sa.sa_len = 2;
280 	m->m_flags |= M_BCAST;
281 	(*ifp->if_output)(ifp, m, &sa, NULL);
282 	ARPSTAT_INC(txrequests);
283 }
284 
285 /*
286  * Resolve an IP address into an ethernet address.
287  * On input:
288  *    ifp is the interface we use
289  *    rt0 is the route to the final destination (possibly useless)
290  *    m is the mbuf. May be NULL if we don't have a packet.
291  *    dst is the next hop,
292  *    desten is where we want the address.
293  *
294  * On success, desten is filled in and the function returns 0;
295  * If the packet must be held pending resolution, we return EWOULDBLOCK
296  * On other errors, we return the corresponding error code.
297  * Note that m_freem() handles NULL.
298  */
299 int
300 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
301 	const struct sockaddr *dst, u_char *desten, struct llentry **lle)
302 {
303 	struct llentry *la = 0;
304 	u_int flags = 0;
305 	struct mbuf *curr = NULL;
306 	struct mbuf *next = NULL;
307 	int error, renew;
308 
309 	*lle = NULL;
310 	if (m != NULL) {
311 		if (m->m_flags & M_BCAST) {
312 			/* broadcast */
313 			(void)memcpy(desten,
314 			    ifp->if_broadcastaddr, ifp->if_addrlen);
315 			return (0);
316 		}
317 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
318 			/* multicast */
319 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
320 			return (0);
321 		}
322 	}
323 retry:
324 	IF_AFDATA_RLOCK(ifp);
325 	la = lla_lookup(LLTABLE(ifp), flags, dst);
326 	IF_AFDATA_RUNLOCK(ifp);
327 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
328 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
329 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
330 		IF_AFDATA_WLOCK(ifp);
331 		la = lla_lookup(LLTABLE(ifp), flags, dst);
332 		IF_AFDATA_WUNLOCK(ifp);
333 	}
334 	if (la == NULL) {
335 		if (flags & LLE_CREATE)
336 			log(LOG_DEBUG,
337 			    "arpresolve: can't allocate llinfo for %s\n",
338 			    inet_ntoa(SIN(dst)->sin_addr));
339 		m_freem(m);
340 		return (EINVAL);
341 	}
342 
343 	if ((la->la_flags & LLE_VALID) &&
344 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
345 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
346 		/*
347 		 * If entry has an expiry time and it is approaching,
348 		 * see if we need to send an ARP request within this
349 		 * arpt_down interval.
350 		 */
351 		if (!(la->la_flags & LLE_STATIC) &&
352 		    time_uptime + la->la_preempt > la->la_expire) {
353 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
354 			la->la_preempt--;
355 		}
356 
357 		*lle = la;
358 		error = 0;
359 		goto done;
360 	}
361 
362 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
363 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
364 		    inet_ntoa(SIN(dst)->sin_addr));
365 		m_freem(m);
366 		error = EINVAL;
367 		goto done;
368 	}
369 
370 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
371 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
372 		flags |= LLE_EXCLUSIVE;
373 		LLE_RUNLOCK(la);
374 		goto retry;
375 	}
376 	/*
377 	 * There is an arptab entry, but no ethernet address
378 	 * response yet.  Add the mbuf to the list, dropping
379 	 * the oldest packet if we have exceeded the system
380 	 * setting.
381 	 */
382 	if (m != NULL) {
383 		if (la->la_numheld >= V_arp_maxhold) {
384 			if (la->la_hold != NULL) {
385 				next = la->la_hold->m_nextpkt;
386 				m_freem(la->la_hold);
387 				la->la_hold = next;
388 				la->la_numheld--;
389 				ARPSTAT_INC(dropped);
390 			}
391 		}
392 		if (la->la_hold != NULL) {
393 			curr = la->la_hold;
394 			while (curr->m_nextpkt != NULL)
395 				curr = curr->m_nextpkt;
396 			curr->m_nextpkt = m;
397 		} else
398 			la->la_hold = m;
399 		la->la_numheld++;
400 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
401 			flags &= ~LLE_EXCLUSIVE;
402 			LLE_DOWNGRADE(la);
403 		}
404 
405 	}
406 	/*
407 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
408 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
409 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
410 	 * ARP request, but not faster than one request per second.
411 	 */
412 	if (la->la_asked < V_arp_maxtries)
413 		error = EWOULDBLOCK;	/* First request. */
414 	else
415 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
416 		    EHOSTUNREACH : EHOSTDOWN;
417 
418 	if (renew) {
419 		int canceled;
420 
421 		LLE_ADDREF(la);
422 		la->la_expire = time_uptime;
423 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
424 		    arptimer, la);
425 		if (canceled)
426 			LLE_REMREF(la);
427 		la->la_asked++;
428 		LLE_WUNLOCK(la);
429 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
430 		return (error);
431 	}
432 done:
433 	if (flags & LLE_EXCLUSIVE)
434 		LLE_WUNLOCK(la);
435 	else
436 		LLE_RUNLOCK(la);
437 	return (error);
438 }
439 
440 /*
441  * Common length and type checks are done here,
442  * then the protocol-specific routine is called.
443  */
444 static void
445 arpintr(struct mbuf *m)
446 {
447 	struct arphdr *ar;
448 
449 	if (m->m_len < sizeof(struct arphdr) &&
450 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
451 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
452 		return;
453 	}
454 	ar = mtod(m, struct arphdr *);
455 
456 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
457 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
458 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
459 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
460 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
461 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
462 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
463 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
464 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
465 		m_freem(m);
466 		return;
467 	}
468 
469 	if (m->m_len < arphdr_len(ar)) {
470 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
471 			log(LOG_NOTICE, "arp: runt packet\n");
472 			m_freem(m);
473 			return;
474 		}
475 		ar = mtod(m, struct arphdr *);
476 	}
477 
478 	ARPSTAT_INC(received);
479 	switch (ntohs(ar->ar_pro)) {
480 #ifdef INET
481 	case ETHERTYPE_IP:
482 		in_arpinput(m);
483 		return;
484 #endif
485 	}
486 	m_freem(m);
487 }
488 
489 #ifdef INET
490 /*
491  * ARP for Internet protocols on 10 Mb/s Ethernet.
492  * Algorithm is that given in RFC 826.
493  * In addition, a sanity check is performed on the sender
494  * protocol address, to catch impersonators.
495  * We no longer handle negotiations for use of trailer protocol:
496  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
497  * along with IP replies if we wanted trailers sent to us,
498  * and also sent them in response to IP replies.
499  * This allowed either end to announce the desire to receive
500  * trailer packets.
501  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
502  * but formerly didn't normally send requests.
503  */
504 static int log_arp_wrong_iface = 1;
505 static int log_arp_movements = 1;
506 static int log_arp_permanent_modify = 1;
507 static int allow_multicast = 0;
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 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
519 	&allow_multicast, 0, "accept multicast addresses");
520 
521 static void
522 in_arpinput(struct mbuf *m)
523 {
524 	struct arphdr *ah;
525 	struct ifnet *ifp = m->m_pkthdr.rcvif;
526 	struct llentry *la = NULL;
527 	struct rtentry *rt;
528 	struct ifaddr *ifa;
529 	struct in_ifaddr *ia;
530 	struct sockaddr sa;
531 	struct in_addr isaddr, itaddr, myaddr;
532 	u_int8_t *enaddr = NULL;
533 	int op, flags;
534 	int req_len;
535 	int bridged = 0, is_bridge = 0;
536 	int carped;
537 	struct sockaddr_in sin;
538 	sin.sin_len = sizeof(struct sockaddr_in);
539 	sin.sin_family = AF_INET;
540 	sin.sin_addr.s_addr = 0;
541 
542 	if (ifp->if_bridge)
543 		bridged = 1;
544 	if (ifp->if_type == IFT_BRIDGE)
545 		is_bridge = 1;
546 
547 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
548 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
549 		log(LOG_NOTICE, "in_arp: runt packet -- m_pullup failed\n");
550 		return;
551 	}
552 
553 	ah = mtod(m, struct arphdr *);
554 	/*
555 	 * ARP is only for IPv4 so we can reject packets with
556 	 * a protocol length not equal to an IPv4 address.
557 	 */
558 	if (ah->ar_pln != sizeof(struct in_addr)) {
559 		log(LOG_NOTICE, "in_arp: requested protocol length != %zu\n",
560 		    sizeof(struct in_addr));
561 		goto drop;
562 	}
563 
564 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
565 		log(LOG_NOTICE, "arp: %*D is multicast\n",
566 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
567 		goto drop;
568 	}
569 
570 	op = ntohs(ah->ar_op);
571 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
572 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
573 
574 	if (op == ARPOP_REPLY)
575 		ARPSTAT_INC(rxreplies);
576 
577 	/*
578 	 * For a bridge, we want to check the address irrespective
579 	 * of the receive interface. (This will change slightly
580 	 * when we have clusters of interfaces).
581 	 */
582 	IN_IFADDR_RLOCK();
583 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
584 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
585 		    ia->ia_ifp == ifp) &&
586 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
587 		    (ia->ia_ifa.ifa_carp == NULL ||
588 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
589 			ifa_ref(&ia->ia_ifa);
590 			IN_IFADDR_RUNLOCK();
591 			goto match;
592 		}
593 	}
594 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
595 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
596 		    ia->ia_ifp == ifp) &&
597 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
598 			ifa_ref(&ia->ia_ifa);
599 			IN_IFADDR_RUNLOCK();
600 			goto match;
601 		}
602 
603 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
604   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
605   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
606   addr == ia->ia_addr.sin_addr.s_addr)
607 	/*
608 	 * Check the case when bridge shares its MAC address with
609 	 * some of its children, so packets are claimed by bridge
610 	 * itself (bridge_input() does it first), but they are really
611 	 * meant to be destined to the bridge member.
612 	 */
613 	if (is_bridge) {
614 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
615 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
616 				ifa_ref(&ia->ia_ifa);
617 				ifp = ia->ia_ifp;
618 				IN_IFADDR_RUNLOCK();
619 				goto match;
620 			}
621 		}
622 	}
623 #undef BDG_MEMBER_MATCHES_ARP
624 	IN_IFADDR_RUNLOCK();
625 
626 	/*
627 	 * No match, use the first inet address on the receive interface
628 	 * as a dummy address for the rest of the function.
629 	 */
630 	IF_ADDR_RLOCK(ifp);
631 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
632 		if (ifa->ifa_addr->sa_family == AF_INET &&
633 		    (ifa->ifa_carp == NULL ||
634 		    (*carp_iamatch_p)(ifa, &enaddr))) {
635 			ia = ifatoia(ifa);
636 			ifa_ref(ifa);
637 			IF_ADDR_RUNLOCK(ifp);
638 			goto match;
639 		}
640 	IF_ADDR_RUNLOCK(ifp);
641 
642 	/*
643 	 * If bridging, fall back to using any inet address.
644 	 */
645 	IN_IFADDR_RLOCK();
646 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
647 		IN_IFADDR_RUNLOCK();
648 		goto drop;
649 	}
650 	ifa_ref(&ia->ia_ifa);
651 	IN_IFADDR_RUNLOCK();
652 match:
653 	if (!enaddr)
654 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
655 	carped = (ia->ia_ifa.ifa_carp != NULL);
656 	myaddr = ia->ia_addr.sin_addr;
657 	ifa_free(&ia->ia_ifa);
658 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
659 		goto drop;	/* it's from me, ignore it. */
660 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
661 		log(LOG_NOTICE,
662 		    "arp: link address is broadcast for IP address %s!\n",
663 		    inet_ntoa(isaddr));
664 		goto drop;
665 	}
666 	/*
667 	 * Warn if another host is using the same IP address, but only if the
668 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
669 	 * case we suppress the warning to avoid false positive complaints of
670 	 * potential misconfiguration.
671 	 */
672 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
673 	    myaddr.s_addr != 0) {
674 		log(LOG_ERR, "arp: %*D is using my IP address %s on %s!\n",
675 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
676 		   inet_ntoa(isaddr), ifp->if_xname);
677 		itaddr = myaddr;
678 		ARPSTAT_INC(dupips);
679 		goto reply;
680 	}
681 	if (ifp->if_flags & IFF_STATICARP)
682 		goto reply;
683 
684 	bzero(&sin, sizeof(sin));
685 	sin.sin_len = sizeof(struct sockaddr_in);
686 	sin.sin_family = AF_INET;
687 	sin.sin_addr = isaddr;
688 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
689 	flags |= LLE_EXCLUSIVE;
690 	IF_AFDATA_LOCK(ifp);
691 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
692 	IF_AFDATA_UNLOCK(ifp);
693 	if (la != NULL) {
694 		/* the following is not an error when doing bridging */
695 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
696 			if (log_arp_wrong_iface)
697 				log(LOG_WARNING, "arp: %s is on %s "
698 				    "but got reply from %*D on %s\n",
699 				    inet_ntoa(isaddr),
700 				    la->lle_tbl->llt_ifp->if_xname,
701 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
702 				    ifp->if_xname);
703 			LLE_WUNLOCK(la);
704 			goto reply;
705 		}
706 		if ((la->la_flags & LLE_VALID) &&
707 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
708 			if (la->la_flags & LLE_STATIC) {
709 				LLE_WUNLOCK(la);
710 				if (log_arp_permanent_modify)
711 					log(LOG_ERR,
712 					    "arp: %*D attempts to modify "
713 					    "permanent entry for %s on %s\n",
714 					    ifp->if_addrlen,
715 					    (u_char *)ar_sha(ah), ":",
716 					    inet_ntoa(isaddr), ifp->if_xname);
717 				goto reply;
718 			}
719 			if (log_arp_movements) {
720 				log(LOG_INFO, "arp: %s moved from %*D "
721 				    "to %*D on %s\n",
722 				    inet_ntoa(isaddr),
723 				    ifp->if_addrlen,
724 				    (u_char *)&la->ll_addr, ":",
725 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
726 				    ifp->if_xname);
727 			}
728 		}
729 
730 		if (ifp->if_addrlen != ah->ar_hln) {
731 			LLE_WUNLOCK(la);
732 			log(LOG_WARNING, "arp from %*D: addr len: new %d, "
733 			    "i/f %d (ignored)\n", ifp->if_addrlen,
734 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
735 			    ifp->if_addrlen);
736 			goto drop;
737 		}
738 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
739 		la->la_flags |= LLE_VALID;
740 
741 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
742 
743 		if (!(la->la_flags & LLE_STATIC)) {
744 			int canceled;
745 
746 			LLE_ADDREF(la);
747 			la->la_expire = time_uptime + V_arpt_keep;
748 			canceled = callout_reset(&la->la_timer,
749 			    hz * V_arpt_keep, arptimer, la);
750 			if (canceled)
751 				LLE_REMREF(la);
752 		}
753 		la->la_asked = 0;
754 		la->la_preempt = V_arp_maxtries;
755 		/*
756 		 * The packets are all freed within the call to the output
757 		 * routine.
758 		 *
759 		 * NB: The lock MUST be released before the call to the
760 		 * output routine.
761 		 */
762 		if (la->la_hold != NULL) {
763 			struct mbuf *m_hold, *m_hold_next;
764 
765 			m_hold = la->la_hold;
766 			la->la_hold = NULL;
767 			la->la_numheld = 0;
768 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
769 			LLE_WUNLOCK(la);
770 			for (; m_hold != NULL; m_hold = m_hold_next) {
771 				m_hold_next = m_hold->m_nextpkt;
772 				m_hold->m_nextpkt = NULL;
773 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
774 			}
775 		} else
776 			LLE_WUNLOCK(la);
777 	}
778 reply:
779 	if (op != ARPOP_REQUEST)
780 		goto drop;
781 	ARPSTAT_INC(rxrequests);
782 
783 	if (itaddr.s_addr == myaddr.s_addr) {
784 		/* Shortcut.. the receiving interface is the target. */
785 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
786 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
787 	} else {
788 		struct llentry *lle = NULL;
789 
790 		sin.sin_addr = itaddr;
791 		IF_AFDATA_LOCK(ifp);
792 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
793 		IF_AFDATA_UNLOCK(ifp);
794 
795 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
796 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
797 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
798 			LLE_RUNLOCK(lle);
799 		} else {
800 
801 			if (lle != NULL)
802 				LLE_RUNLOCK(lle);
803 
804 			if (!V_arp_proxyall)
805 				goto drop;
806 
807 			sin.sin_addr = itaddr;
808 			/* XXX MRT use table 0 for arp reply  */
809 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
810 			if (!rt)
811 				goto drop;
812 
813 			/*
814 			 * Don't send proxies for nodes on the same interface
815 			 * as this one came out of, or we'll get into a fight
816 			 * over who claims what Ether address.
817 			 */
818 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
819 				RTFREE_LOCKED(rt);
820 				goto drop;
821 			}
822 			RTFREE_LOCKED(rt);
823 
824 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
825 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
826 
827 			/*
828 			 * Also check that the node which sent the ARP packet
829 			 * is on the interface we expect it to be on. This
830 			 * avoids ARP chaos if an interface is connected to the
831 			 * wrong network.
832 			 */
833 			sin.sin_addr = isaddr;
834 
835 			/* XXX MRT use table 0 for arp checks */
836 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
837 			if (!rt)
838 				goto drop;
839 			if (rt->rt_ifp != ifp) {
840 				log(LOG_INFO, "arp_proxy: ignoring request"
841 				    " from %s via %s, expecting %s\n",
842 				    inet_ntoa(isaddr), ifp->if_xname,
843 				    rt->rt_ifp->if_xname);
844 				RTFREE_LOCKED(rt);
845 				goto drop;
846 			}
847 			RTFREE_LOCKED(rt);
848 
849 #ifdef DEBUG_PROXY
850 			printf("arp: proxying for %s\n", 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