xref: /freebsd/sys/netinet/if_ether.c (revision b7c60aadbbd5c846a250c05791fe7406d6d78bf4)
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 			pkts_dropped = llentry_free(lle);
184 			ARPSTAT_ADD(dropped, pkts_dropped);
185 			ARPSTAT_INC(timeouts);
186 		} else {
187 #ifdef DIAGNOSTIC
188 			struct sockaddr *l3addr = L3_ADDR(lle);
189 			log(LOG_INFO,
190 			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
191 			    inet_ntoa(
192 			        ((const struct sockaddr_in *)l3addr)->sin_addr));
193 #endif
194 			LLE_WUNLOCK(lle);
195 		}
196 	}
197 	IF_AFDATA_UNLOCK(ifp);
198 	CURVNET_RESTORE();
199 }
200 
201 /*
202  * Broadcast an ARP request. Caller specifies:
203  *	- arp header source ip address
204  *	- arp header target ip address
205  *	- arp header source ethernet address
206  */
207 void
208 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
209     u_char *enaddr)
210 {
211 	struct mbuf *m;
212 	struct arphdr *ah;
213 	struct sockaddr sa;
214 	u_char *carpaddr = NULL;
215 
216 	if (sip == NULL) {
217 		/*
218 		 * The caller did not supply a source address, try to find
219 		 * a compatible one among those assigned to this interface.
220 		 */
221 		struct ifaddr *ifa;
222 
223 		IF_ADDR_RLOCK(ifp);
224 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
225 			if (ifa->ifa_addr->sa_family != AF_INET)
226 				continue;
227 
228 			if (ifa->ifa_carp) {
229 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
230 					continue;
231 				sip = &IA_SIN(ifa)->sin_addr;
232 			} else {
233 				carpaddr = NULL;
234 				sip = &IA_SIN(ifa)->sin_addr;
235 			}
236 
237 			if (0 == ((sip->s_addr ^ tip->s_addr) &
238 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
239 				break;  /* found it. */
240 		}
241 		IF_ADDR_RUNLOCK(ifp);
242 		if (sip == NULL) {
243 			printf("%s: cannot find matching address\n", __func__);
244 			return;
245 		}
246 	}
247 	if (enaddr == NULL)
248 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
249 
250 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
251 		return;
252 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
253 		2*ifp->if_data.ifi_addrlen;
254 	m->m_pkthdr.len = m->m_len;
255 	MH_ALIGN(m, m->m_len);
256 	ah = mtod(m, struct arphdr *);
257 	bzero((caddr_t)ah, m->m_len);
258 #ifdef MAC
259 	mac_netinet_arp_send(ifp, m);
260 #endif
261 	ah->ar_pro = htons(ETHERTYPE_IP);
262 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
263 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
264 	ah->ar_op = htons(ARPOP_REQUEST);
265 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
266 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
267 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
268 	sa.sa_family = AF_ARP;
269 	sa.sa_len = 2;
270 	m->m_flags |= M_BCAST;
271 	(*ifp->if_output)(ifp, m, &sa, NULL);
272 	ARPSTAT_INC(txrequests);
273 }
274 
275 /*
276  * Resolve an IP address into an ethernet address.
277  * On input:
278  *    ifp is the interface we use
279  *    rt0 is the route to the final destination (possibly useless)
280  *    m is the mbuf. May be NULL if we don't have a packet.
281  *    dst is the next hop,
282  *    desten is where we want the address.
283  *
284  * On success, desten is filled in and the function returns 0;
285  * If the packet must be held pending resolution, we return EWOULDBLOCK
286  * On other errors, we return the corresponding error code.
287  * Note that m_freem() handles NULL.
288  */
289 int
290 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
291 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
292 {
293 	struct llentry *la = 0;
294 	u_int flags = 0;
295 	struct mbuf *curr = NULL;
296 	struct mbuf *next = NULL;
297 	int error, renew;
298 
299 	*lle = NULL;
300 	if (m != NULL) {
301 		if (m->m_flags & M_BCAST) {
302 			/* broadcast */
303 			(void)memcpy(desten,
304 			    ifp->if_broadcastaddr, ifp->if_addrlen);
305 			return (0);
306 		}
307 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
308 			/* multicast */
309 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
310 			return (0);
311 		}
312 	}
313 retry:
314 	IF_AFDATA_RLOCK(ifp);
315 	la = lla_lookup(LLTABLE(ifp), flags, dst);
316 	IF_AFDATA_RUNLOCK(ifp);
317 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
318 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
319 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
320 		IF_AFDATA_WLOCK(ifp);
321 		la = lla_lookup(LLTABLE(ifp), flags, dst);
322 		IF_AFDATA_WUNLOCK(ifp);
323 	}
324 	if (la == NULL) {
325 		if (flags & LLE_CREATE)
326 			log(LOG_DEBUG,
327 			    "arpresolve: can't allocate llinfo for %s\n",
328 			    inet_ntoa(SIN(dst)->sin_addr));
329 		m_freem(m);
330 		return (EINVAL);
331 	}
332 
333 	if ((la->la_flags & LLE_VALID) &&
334 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
335 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
336 		/*
337 		 * If entry has an expiry time and it is approaching,
338 		 * see if we need to send an ARP request within this
339 		 * arpt_down interval.
340 		 */
341 		if (!(la->la_flags & LLE_STATIC) &&
342 		    time_uptime + la->la_preempt > la->la_expire) {
343 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
344 			la->la_preempt--;
345 		}
346 
347 		*lle = la;
348 		error = 0;
349 		goto done;
350 	}
351 
352 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
353 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
354 		    inet_ntoa(SIN(dst)->sin_addr));
355 		m_freem(m);
356 		error = EINVAL;
357 		goto done;
358 	}
359 
360 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
361 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
362 		flags |= LLE_EXCLUSIVE;
363 		LLE_RUNLOCK(la);
364 		goto retry;
365 	}
366 	/*
367 	 * There is an arptab entry, but no ethernet address
368 	 * response yet.  Add the mbuf to the list, dropping
369 	 * the oldest packet if we have exceeded the system
370 	 * setting.
371 	 */
372 	if (m != NULL) {
373 		if (la->la_numheld >= V_arp_maxhold) {
374 			if (la->la_hold != NULL) {
375 				next = la->la_hold->m_nextpkt;
376 				m_freem(la->la_hold);
377 				la->la_hold = next;
378 				la->la_numheld--;
379 				ARPSTAT_INC(dropped);
380 			}
381 		}
382 		if (la->la_hold != NULL) {
383 			curr = la->la_hold;
384 			while (curr->m_nextpkt != NULL)
385 				curr = curr->m_nextpkt;
386 			curr->m_nextpkt = m;
387 		} else
388 			la->la_hold = m;
389 		la->la_numheld++;
390 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
391 			flags &= ~LLE_EXCLUSIVE;
392 			LLE_DOWNGRADE(la);
393 		}
394 
395 	}
396 	/*
397 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
398 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
399 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
400 	 * ARP request, but not faster than one request per second.
401 	 */
402 	if (la->la_asked < V_arp_maxtries)
403 		error = EWOULDBLOCK;	/* First request. */
404 	else
405 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
406 		    EHOSTUNREACH : EHOSTDOWN;
407 
408 	if (renew) {
409 		int canceled;
410 
411 		LLE_ADDREF(la);
412 		la->la_expire = time_uptime;
413 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
414 		    arptimer, la);
415 		if (canceled)
416 			LLE_REMREF(la);
417 		la->la_asked++;
418 		LLE_WUNLOCK(la);
419 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
420 		return (error);
421 	}
422 done:
423 	if (flags & LLE_EXCLUSIVE)
424 		LLE_WUNLOCK(la);
425 	else
426 		LLE_RUNLOCK(la);
427 	return (error);
428 }
429 
430 /*
431  * Common length and type checks are done here,
432  * then the protocol-specific routine is called.
433  */
434 static void
435 arpintr(struct mbuf *m)
436 {
437 	struct arphdr *ar;
438 
439 	if (m->m_len < sizeof(struct arphdr) &&
440 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
441 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
442 		return;
443 	}
444 	ar = mtod(m, struct arphdr *);
445 
446 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
447 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
448 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
449 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
450 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
451 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
452 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
453 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
454 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
455 		m_freem(m);
456 		return;
457 	}
458 
459 	if (m->m_len < arphdr_len(ar)) {
460 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
461 			log(LOG_NOTICE, "arp: runt packet\n");
462 			m_freem(m);
463 			return;
464 		}
465 		ar = mtod(m, struct arphdr *);
466 	}
467 
468 	ARPSTAT_INC(received);
469 	switch (ntohs(ar->ar_pro)) {
470 #ifdef INET
471 	case ETHERTYPE_IP:
472 		in_arpinput(m);
473 		return;
474 #endif
475 	}
476 	m_freem(m);
477 }
478 
479 #ifdef INET
480 /*
481  * ARP for Internet protocols on 10 Mb/s Ethernet.
482  * Algorithm is that given in RFC 826.
483  * In addition, a sanity check is performed on the sender
484  * protocol address, to catch impersonators.
485  * We no longer handle negotiations for use of trailer protocol:
486  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
487  * along with IP replies if we wanted trailers sent to us,
488  * and also sent them in response to IP replies.
489  * This allowed either end to announce the desire to receive
490  * trailer packets.
491  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
492  * but formerly didn't normally send requests.
493  */
494 static int log_arp_wrong_iface = 1;
495 static int log_arp_movements = 1;
496 static int log_arp_permanent_modify = 1;
497 
498 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
499 	&log_arp_wrong_iface, 0,
500 	"log arp packets arriving on the wrong interface");
501 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
502         &log_arp_movements, 0,
503         "log arp replies from MACs different than the one in the cache");
504 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
505         &log_arp_permanent_modify, 0,
506         "log arp replies from MACs different than the one in the permanent arp entry");
507 
508 
509 static void
510 in_arpinput(struct mbuf *m)
511 {
512 	struct arphdr *ah;
513 	struct ifnet *ifp = m->m_pkthdr.rcvif;
514 	struct llentry *la = NULL;
515 	struct rtentry *rt;
516 	struct ifaddr *ifa;
517 	struct in_ifaddr *ia;
518 	struct sockaddr sa;
519 	struct in_addr isaddr, itaddr, myaddr;
520 	u_int8_t *enaddr = NULL;
521 	int op, flags;
522 	int req_len;
523 	int bridged = 0, is_bridge = 0;
524 	int carped;
525 	struct sockaddr_in sin;
526 	sin.sin_len = sizeof(struct sockaddr_in);
527 	sin.sin_family = AF_INET;
528 	sin.sin_addr.s_addr = 0;
529 
530 	if (ifp->if_bridge)
531 		bridged = 1;
532 	if (ifp->if_type == IFT_BRIDGE)
533 		is_bridge = 1;
534 
535 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
536 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
537 		log(LOG_NOTICE, "in_arp: runt packet -- m_pullup failed\n");
538 		return;
539 	}
540 
541 	ah = mtod(m, struct arphdr *);
542 	/*
543 	 * ARP is only for IPv4 so we can reject packets with
544 	 * a protocol length not equal to an IPv4 address.
545 	 */
546 	if (ah->ar_pln != sizeof(struct in_addr)) {
547 		log(LOG_NOTICE, "in_arp: requested protocol length != %zu\n",
548 		    sizeof(struct in_addr));
549 		return;
550 	}
551 
552 	if (ETHER_IS_MULTICAST(ar_sha(ah))) {
553 		log(LOG_NOTICE, "in_arp: %*D is multicast\n",
554 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
555 		return;
556 	}
557 
558 	op = ntohs(ah->ar_op);
559 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
560 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
561 
562 	if (op == ARPOP_REPLY)
563 		ARPSTAT_INC(rxreplies);
564 
565 	/*
566 	 * For a bridge, we want to check the address irrespective
567 	 * of the receive interface. (This will change slightly
568 	 * when we have clusters of interfaces).
569 	 */
570 	IN_IFADDR_RLOCK();
571 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
572 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
573 		    ia->ia_ifp == ifp) &&
574 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
575 		    (ia->ia_ifa.ifa_carp == NULL ||
576 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
577 			ifa_ref(&ia->ia_ifa);
578 			IN_IFADDR_RUNLOCK();
579 			goto match;
580 		}
581 	}
582 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
583 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
584 		    ia->ia_ifp == ifp) &&
585 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
586 			ifa_ref(&ia->ia_ifa);
587 			IN_IFADDR_RUNLOCK();
588 			goto match;
589 		}
590 
591 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
592   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
593   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
594   addr == ia->ia_addr.sin_addr.s_addr)
595 	/*
596 	 * Check the case when bridge shares its MAC address with
597 	 * some of its children, so packets are claimed by bridge
598 	 * itself (bridge_input() does it first), but they are really
599 	 * meant to be destined to the bridge member.
600 	 */
601 	if (is_bridge) {
602 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
603 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
604 				ifa_ref(&ia->ia_ifa);
605 				ifp = ia->ia_ifp;
606 				IN_IFADDR_RUNLOCK();
607 				goto match;
608 			}
609 		}
610 	}
611 #undef BDG_MEMBER_MATCHES_ARP
612 	IN_IFADDR_RUNLOCK();
613 
614 	/*
615 	 * No match, use the first inet address on the receive interface
616 	 * as a dummy address for the rest of the function.
617 	 */
618 	IF_ADDR_RLOCK(ifp);
619 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
620 		if (ifa->ifa_addr->sa_family == AF_INET &&
621 		    (ifa->ifa_carp == NULL ||
622 		    (*carp_iamatch_p)(ifa, &enaddr))) {
623 			ia = ifatoia(ifa);
624 			ifa_ref(ifa);
625 			IF_ADDR_RUNLOCK(ifp);
626 			goto match;
627 		}
628 	IF_ADDR_RUNLOCK(ifp);
629 
630 	/*
631 	 * If bridging, fall back to using any inet address.
632 	 */
633 	IN_IFADDR_RLOCK();
634 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
635 		IN_IFADDR_RUNLOCK();
636 		goto drop;
637 	}
638 	ifa_ref(&ia->ia_ifa);
639 	IN_IFADDR_RUNLOCK();
640 match:
641 	if (!enaddr)
642 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
643 	carped = (ia->ia_ifa.ifa_carp != NULL);
644 	myaddr = ia->ia_addr.sin_addr;
645 	ifa_free(&ia->ia_ifa);
646 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
647 		goto drop;	/* it's from me, ignore it. */
648 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
649 		log(LOG_NOTICE,
650 		    "arp: link address is broadcast for IP address %s!\n",
651 		    inet_ntoa(isaddr));
652 		goto drop;
653 	}
654 	/*
655 	 * Warn if another host is using the same IP address, but only if the
656 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
657 	 * case we suppress the warning to avoid false positive complaints of
658 	 * potential misconfiguration.
659 	 */
660 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
661 	    myaddr.s_addr != 0) {
662 		log(LOG_ERR, "arp: %*D is using my IP address %s on %s!\n",
663 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
664 		   inet_ntoa(isaddr), ifp->if_xname);
665 		itaddr = myaddr;
666 		ARPSTAT_INC(dupips);
667 		goto reply;
668 	}
669 	if (ifp->if_flags & IFF_STATICARP)
670 		goto reply;
671 
672 	bzero(&sin, sizeof(sin));
673 	sin.sin_len = sizeof(struct sockaddr_in);
674 	sin.sin_family = AF_INET;
675 	sin.sin_addr = isaddr;
676 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
677 	flags |= LLE_EXCLUSIVE;
678 	IF_AFDATA_LOCK(ifp);
679 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
680 	IF_AFDATA_UNLOCK(ifp);
681 	if (la != NULL) {
682 		/* the following is not an error when doing bridging */
683 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
684 			if (log_arp_wrong_iface)
685 				log(LOG_WARNING, "arp: %s is on %s "
686 				    "but got reply from %*D on %s\n",
687 				    inet_ntoa(isaddr),
688 				    la->lle_tbl->llt_ifp->if_xname,
689 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
690 				    ifp->if_xname);
691 			LLE_WUNLOCK(la);
692 			goto reply;
693 		}
694 		if ((la->la_flags & LLE_VALID) &&
695 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
696 			if (la->la_flags & LLE_STATIC) {
697 				LLE_WUNLOCK(la);
698 				if (log_arp_permanent_modify)
699 					log(LOG_ERR,
700 					    "arp: %*D attempts to modify "
701 					    "permanent entry for %s on %s\n",
702 					    ifp->if_addrlen,
703 					    (u_char *)ar_sha(ah), ":",
704 					    inet_ntoa(isaddr), ifp->if_xname);
705 				goto reply;
706 			}
707 			if (log_arp_movements) {
708 			        log(LOG_INFO, "arp: %s moved from %*D "
709 				    "to %*D on %s\n",
710 				    inet_ntoa(isaddr),
711 				    ifp->if_addrlen,
712 				    (u_char *)&la->ll_addr, ":",
713 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
714 				    ifp->if_xname);
715 			}
716 		}
717 
718 		if (ifp->if_addrlen != ah->ar_hln) {
719 			LLE_WUNLOCK(la);
720 			log(LOG_WARNING, "arp from %*D: addr len: new %d, "
721 			    "i/f %d (ignored)\n", ifp->if_addrlen,
722 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
723 			    ifp->if_addrlen);
724 			goto drop;
725 		}
726 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
727 		la->la_flags |= LLE_VALID;
728 
729 		EVENTHANDLER_INVOKE(arp_update_event, la);
730 
731 		if (!(la->la_flags & LLE_STATIC)) {
732 			int canceled;
733 
734 			LLE_ADDREF(la);
735 			la->la_expire = time_uptime + V_arpt_keep;
736 			canceled = callout_reset(&la->la_timer,
737 			    hz * V_arpt_keep, arptimer, la);
738 			if (canceled)
739 				LLE_REMREF(la);
740 		}
741 		la->la_asked = 0;
742 		la->la_preempt = V_arp_maxtries;
743 		/*
744 		 * The packets are all freed within the call to the output
745 		 * routine.
746 		 *
747 		 * NB: The lock MUST be released before the call to the
748 		 * output routine.
749 		 */
750 		if (la->la_hold != NULL) {
751 			struct mbuf *m_hold, *m_hold_next;
752 
753 			m_hold = la->la_hold;
754 			la->la_hold = NULL;
755 			la->la_numheld = 0;
756 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
757 			LLE_WUNLOCK(la);
758 			for (; m_hold != NULL; m_hold = m_hold_next) {
759 				m_hold_next = m_hold->m_nextpkt;
760 				m_hold->m_nextpkt = NULL;
761 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
762 			}
763 		} else
764 			LLE_WUNLOCK(la);
765 	}
766 reply:
767 	if (op != ARPOP_REQUEST)
768 		goto drop;
769 	ARPSTAT_INC(rxrequests);
770 
771 	if (itaddr.s_addr == myaddr.s_addr) {
772 		/* Shortcut.. the receiving interface is the target. */
773 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
774 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
775 	} else {
776 		struct llentry *lle = NULL;
777 
778 		sin.sin_addr = itaddr;
779 		IF_AFDATA_LOCK(ifp);
780 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
781 		IF_AFDATA_UNLOCK(ifp);
782 
783 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
784 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
785 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
786 			LLE_RUNLOCK(lle);
787 		} else {
788 
789 			if (lle != NULL)
790 				LLE_RUNLOCK(lle);
791 
792 			if (!V_arp_proxyall)
793 				goto drop;
794 
795 			sin.sin_addr = itaddr;
796 			/* XXX MRT use table 0 for arp reply  */
797 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
798 			if (!rt)
799 				goto drop;
800 
801 			/*
802 			 * Don't send proxies for nodes on the same interface
803 			 * as this one came out of, or we'll get into a fight
804 			 * over who claims what Ether address.
805 			 */
806 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
807 				RTFREE_LOCKED(rt);
808 				goto drop;
809 			}
810 			RTFREE_LOCKED(rt);
811 
812 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
813 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
814 
815 			/*
816 			 * Also check that the node which sent the ARP packet
817 			 * is on the interface we expect it to be on. This
818 			 * avoids ARP chaos if an interface is connected to the
819 			 * wrong network.
820 			 */
821 			sin.sin_addr = isaddr;
822 
823 			/* XXX MRT use table 0 for arp checks */
824 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
825 			if (!rt)
826 				goto drop;
827 			if (rt->rt_ifp != ifp) {
828 				log(LOG_INFO, "arp_proxy: ignoring request"
829 				    " from %s via %s, expecting %s\n",
830 				    inet_ntoa(isaddr), ifp->if_xname,
831 				    rt->rt_ifp->if_xname);
832 				RTFREE_LOCKED(rt);
833 				goto drop;
834 			}
835 			RTFREE_LOCKED(rt);
836 
837 #ifdef DEBUG_PROXY
838 			printf("arp: proxying for %s\n",
839 			       inet_ntoa(itaddr));
840 #endif
841 		}
842 	}
843 
844 	if (itaddr.s_addr == myaddr.s_addr &&
845 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
846 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
847 #ifdef DEBUG_LINKLOCAL
848 		printf("arp: sending reply for link-local addr %s\n",
849 		    inet_ntoa(itaddr));
850 #endif
851 		m->m_flags |= M_BCAST;
852 		m->m_flags &= ~M_MCAST;
853 	} else {
854 		/* default behaviour; never reply by broadcast. */
855 		m->m_flags &= ~(M_BCAST|M_MCAST);
856 	}
857 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
858 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
859 	ah->ar_op = htons(ARPOP_REPLY);
860 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
861 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
862 	m->m_pkthdr.len = m->m_len;
863 	m->m_pkthdr.rcvif = NULL;
864 	sa.sa_family = AF_ARP;
865 	sa.sa_len = 2;
866 	(*ifp->if_output)(ifp, m, &sa, NULL);
867 	ARPSTAT_INC(txreplies);
868 	return;
869 
870 drop:
871 	m_freem(m);
872 }
873 #endif
874 
875 void
876 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
877 {
878 	struct llentry *lle;
879 
880 	if (ifa->ifa_carp != NULL)
881 		return;
882 
883 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
884 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
885 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
886 		/*
887 		 * interface address is considered static entry
888 		 * because the output of the arp utility shows
889 		 * that L2 entry as permanent
890 		 */
891 		IF_AFDATA_LOCK(ifp);
892 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
893 				 (struct sockaddr *)IA_SIN(ifa));
894 		IF_AFDATA_UNLOCK(ifp);
895 		if (lle == NULL)
896 			log(LOG_INFO, "arp_ifinit: cannot create arp "
897 			    "entry for interface address\n");
898 		else
899 			LLE_RUNLOCK(lle);
900 	}
901 	ifa->ifa_rtrequest = NULL;
902 }
903 
904 void
905 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
906 {
907 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
908 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
909 				&IA_SIN(ifa)->sin_addr, enaddr);
910 	ifa->ifa_rtrequest = NULL;
911 }
912 
913 static void
914 arp_init(void)
915 {
916 
917 	netisr_register(&arp_nh);
918 }
919 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
920