xref: /freebsd/sys/netinet/if_ether.c (revision b2db760808f74bb53c232900091c9da801ebbfcc)
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 #if defined(INET) || defined(INET6)
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 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 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 static VNET_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
93 
94 #define	V_arpt_keep		VNET(arpt_keep)
95 #define	V_arpt_down		VNET(arpt_down)
96 #define	V_arp_maxtries		VNET(arp_maxtries)
97 #define	V_arp_proxyall		VNET(arp_proxyall)
98 #define	V_arpstat		VNET(arpstat)
99 
100 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
101 	&VNET_NAME(arpt_keep), 0,
102 	"ARP entry lifetime in seconds");
103 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
104 	&VNET_NAME(arp_maxtries), 0,
105 	"ARP resolution attempts before returning error");
106 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
107 	&VNET_NAME(useloopback), 0,
108 	"Use the loopback interface for local traffic");
109 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
110 	&VNET_NAME(arp_proxyall), 0,
111 	"Enable proxy ARP for all suitable requests");
112 SYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW,
113 	&VNET_NAME(arpstat), arpstat,
114 	"ARP statistics (struct arpstat, net/if_arp.h)");
115 
116 static void	arp_init(void);
117 void		arprequest(struct ifnet *,
118 			struct in_addr *, struct in_addr *, u_char *);
119 static void	arpintr(struct mbuf *);
120 static void	arptimer(void *);
121 #ifdef INET
122 static void	in_arpinput(struct mbuf *);
123 #endif
124 
125 static const struct netisr_handler arp_nh = {
126 	.nh_name = "arp",
127 	.nh_handler = arpintr,
128 	.nh_proto = NETISR_ARP,
129 	.nh_policy = NETISR_POLICY_SOURCE,
130 };
131 
132 #ifdef AF_INET
133 void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
134 
135 /*
136  * called by in_ifscrub to remove entry from the table when
137  * the interface goes away
138  */
139 void
140 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
141 {
142 	struct sockaddr_in addr4;
143 
144 	bzero((void *)&addr4, sizeof(addr4));
145 	addr4.sin_len    = sizeof(addr4);
146 	addr4.sin_family = AF_INET;
147 	addr4.sin_addr.s_addr = addr;
148 	CURVNET_SET(ifp->if_vnet);
149 	IF_AFDATA_LOCK(ifp);
150 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
151 	    (struct sockaddr *)&addr4);
152 	IF_AFDATA_UNLOCK(ifp);
153 	CURVNET_RESTORE();
154 }
155 #endif
156 
157 /*
158  * Timeout routine.  Age arp_tab entries periodically.
159  */
160 static void
161 arptimer(void *arg)
162 {
163 	struct ifnet *ifp;
164 	struct llentry   *lle;
165 
166 	KASSERT(arg != NULL, ("%s: arg NULL", __func__));
167 	lle = (struct llentry *)arg;
168 	ifp = lle->lle_tbl->llt_ifp;
169 	CURVNET_SET(ifp->if_vnet);
170 	IF_AFDATA_LOCK(ifp);
171 	LLE_WLOCK(lle);
172 	if (lle->la_flags & LLE_STATIC)
173 		LLE_WUNLOCK(lle);
174 	else {
175 		if (!callout_pending(&lle->la_timer) &&
176 		    callout_active(&lle->la_timer)) {
177 			callout_stop(&lle->la_timer);
178 			LLE_REMREF(lle);
179 			(void) llentry_free(lle);
180 			ARPSTAT_INC(timeouts);
181 		}
182 #ifdef DIAGNOSTIC
183 		else {
184 			struct sockaddr *l3addr = L3_ADDR(lle);
185 			log(LOG_INFO,
186 			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
187 			    inet_ntoa(
188 			        ((const struct sockaddr_in *)l3addr)->sin_addr));
189 		}
190 #endif
191 	}
192 	IF_AFDATA_UNLOCK(ifp);
193 	CURVNET_RESTORE();
194 }
195 
196 /*
197  * Broadcast an ARP request. Caller specifies:
198  *	- arp header source ip address
199  *	- arp header target ip address
200  *	- arp header source ethernet address
201  */
202 void
203 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
204     u_char *enaddr)
205 {
206 	struct mbuf *m;
207 	struct arphdr *ah;
208 	struct sockaddr sa;
209 
210 	if (sip == NULL) {
211 		/* XXX don't believe this can happen (or explain why) */
212 		/*
213 		 * The caller did not supply a source address, try to find
214 		 * a compatible one among those assigned to this interface.
215 		 */
216 		struct ifaddr *ifa;
217 
218 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
219 			if (!ifa->ifa_addr ||
220 			    ifa->ifa_addr->sa_family != AF_INET)
221 				continue;
222 			sip = &SIN(ifa->ifa_addr)->sin_addr;
223 			if (0 == ((sip->s_addr ^ tip->s_addr) &
224 			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
225 				break;  /* found it. */
226 		}
227 		if (sip == NULL) {
228 			printf("%s: cannot find matching address\n", __func__);
229 			return;
230 		}
231 	}
232 
233 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
234 		return;
235 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
236 		2*ifp->if_data.ifi_addrlen;
237 	m->m_pkthdr.len = m->m_len;
238 	MH_ALIGN(m, m->m_len);
239 	ah = mtod(m, struct arphdr *);
240 	bzero((caddr_t)ah, m->m_len);
241 #ifdef MAC
242 	mac_netinet_arp_send(ifp, m);
243 #endif
244 	ah->ar_pro = htons(ETHERTYPE_IP);
245 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
246 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
247 	ah->ar_op = htons(ARPOP_REQUEST);
248 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
249 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
250 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
251 	sa.sa_family = AF_ARP;
252 	sa.sa_len = 2;
253 	m->m_flags |= M_BCAST;
254 	(*ifp->if_output)(ifp, m, &sa, NULL);
255 	ARPSTAT_INC(txrequests);
256 }
257 
258 /*
259  * Resolve an IP address into an ethernet address.
260  * On input:
261  *    ifp is the interface we use
262  *    rt0 is the route to the final destination (possibly useless)
263  *    m is the mbuf. May be NULL if we don't have a packet.
264  *    dst is the next hop,
265  *    desten is where we want the address.
266  *
267  * On success, desten is filled in and the function returns 0;
268  * If the packet must be held pending resolution, we return EWOULDBLOCK
269  * On other errors, we return the corresponding error code.
270  * Note that m_freem() handles NULL.
271  */
272 int
273 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
274 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
275 {
276 	struct llentry *la = 0;
277 	u_int flags = 0;
278 	int error, renew;
279 
280 	*lle = NULL;
281 	if (m != NULL) {
282 		if (m->m_flags & M_BCAST) {
283 			/* broadcast */
284 			(void)memcpy(desten,
285 			    ifp->if_broadcastaddr, ifp->if_addrlen);
286 			return (0);
287 		}
288 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
289 			/* multicast */
290 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
291 			return (0);
292 		}
293 	}
294 	/* XXXXX
295 	 */
296 retry:
297 	IF_AFDATA_RLOCK(ifp);
298 	la = lla_lookup(LLTABLE(ifp), flags, dst);
299 	IF_AFDATA_RUNLOCK(ifp);
300 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
301 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
302 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
303 		IF_AFDATA_WLOCK(ifp);
304 		la = lla_lookup(LLTABLE(ifp), flags, dst);
305 		IF_AFDATA_WUNLOCK(ifp);
306 	}
307 	if (la == NULL) {
308 		if (flags & LLE_CREATE)
309 			log(LOG_DEBUG,
310 			    "arpresolve: can't allocate llinfo for %s\n",
311 			    inet_ntoa(SIN(dst)->sin_addr));
312 		m_freem(m);
313 		return (EINVAL);
314 	}
315 
316 	if ((la->la_flags & LLE_VALID) &&
317 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_second)) {
318 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
319 		/*
320 		 * If entry has an expiry time and it is approaching,
321 		 * see if we need to send an ARP request within this
322 		 * arpt_down interval.
323 		 */
324 		if (!(la->la_flags & LLE_STATIC) &&
325 		    time_second + la->la_preempt > la->la_expire) {
326 			arprequest(ifp, NULL,
327 			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
328 
329 			la->la_preempt--;
330 		}
331 
332 		*lle = la;
333 		error = 0;
334 		goto done;
335 	}
336 
337 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
338 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
339 		    inet_ntoa(SIN(dst)->sin_addr));
340 		m_freem(m);
341 		error = EINVAL;
342 		goto done;
343 	}
344 
345 	renew = (la->la_asked == 0 || la->la_expire != time_second);
346 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
347 		flags |= LLE_EXCLUSIVE;
348 		LLE_RUNLOCK(la);
349 		goto retry;
350 	}
351 	/*
352 	 * There is an arptab entry, but no ethernet address
353 	 * response yet.  Replace the held mbuf with this
354 	 * latest one.
355 	 */
356 	if (m != NULL) {
357 		if (la->la_hold != NULL) {
358 			m_freem(la->la_hold);
359 			ARPSTAT_INC(dropped);
360 		}
361 		la->la_hold = m;
362 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
363 			flags &= ~LLE_EXCLUSIVE;
364 			LLE_DOWNGRADE(la);
365 		}
366 
367 	}
368 	/*
369 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
370 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
371 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
372 	 * ARP request, but not faster than one request per second.
373 	 */
374 	if (la->la_asked < V_arp_maxtries)
375 		error = EWOULDBLOCK;	/* First request. */
376 	else
377 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
378 		    EHOSTUNREACH : EHOSTDOWN;
379 
380 	if (renew) {
381 		int canceled;
382 
383 		LLE_ADDREF(la);
384 		la->la_expire = time_second + V_arpt_down;
385 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
386 		    arptimer, la);
387 		if (canceled)
388 			LLE_REMREF(la);
389 		la->la_asked++;
390 		LLE_WUNLOCK(la);
391 		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
392 		    IF_LLADDR(ifp));
393 		return (error);
394 	}
395 done:
396 	if (flags & LLE_EXCLUSIVE)
397 		LLE_WUNLOCK(la);
398 	else
399 		LLE_RUNLOCK(la);
400 	return (error);
401 }
402 
403 /*
404  * Common length and type checks are done here,
405  * then the protocol-specific routine is called.
406  */
407 static void
408 arpintr(struct mbuf *m)
409 {
410 	struct arphdr *ar;
411 
412 	if (m->m_len < sizeof(struct arphdr) &&
413 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
414 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
415 		return;
416 	}
417 	ar = mtod(m, struct arphdr *);
418 
419 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
420 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
421 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
422 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
423 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
424 		    (unsigned char *)&ar->ar_hrd, "");
425 		m_freem(m);
426 		return;
427 	}
428 
429 	if (m->m_len < arphdr_len(ar)) {
430 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
431 			log(LOG_ERR, "arp: runt packet\n");
432 			m_freem(m);
433 			return;
434 		}
435 		ar = mtod(m, struct arphdr *);
436 	}
437 
438 	ARPSTAT_INC(received);
439 	switch (ntohs(ar->ar_pro)) {
440 #ifdef INET
441 	case ETHERTYPE_IP:
442 		in_arpinput(m);
443 		return;
444 #endif
445 	}
446 	m_freem(m);
447 }
448 
449 #ifdef INET
450 /*
451  * ARP for Internet protocols on 10 Mb/s Ethernet.
452  * Algorithm is that given in RFC 826.
453  * In addition, a sanity check is performed on the sender
454  * protocol address, to catch impersonators.
455  * We no longer handle negotiations for use of trailer protocol:
456  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
457  * along with IP replies if we wanted trailers sent to us,
458  * and also sent them in response to IP replies.
459  * This allowed either end to announce the desire to receive
460  * trailer packets.
461  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
462  * but formerly didn't normally send requests.
463  */
464 static int log_arp_wrong_iface = 1;
465 static int log_arp_movements = 1;
466 static int log_arp_permanent_modify = 1;
467 
468 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
469 	&log_arp_wrong_iface, 0,
470 	"log arp packets arriving on the wrong interface");
471 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
472         &log_arp_movements, 0,
473         "log arp replies from MACs different than the one in the cache");
474 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
475         &log_arp_permanent_modify, 0,
476         "log arp replies from MACs different than the one in the permanent arp entry");
477 
478 
479 static void
480 in_arpinput(struct mbuf *m)
481 {
482 	struct arphdr *ah;
483 	struct ifnet *ifp = m->m_pkthdr.rcvif;
484 	struct llentry *la = NULL;
485 	struct rtentry *rt;
486 	struct ifaddr *ifa;
487 	struct in_ifaddr *ia;
488 	struct mbuf *hold;
489 	struct sockaddr sa;
490 	struct in_addr isaddr, itaddr, myaddr;
491 	u_int8_t *enaddr = NULL;
492 	int op, flags;
493 	int req_len;
494 	int bridged = 0, is_bridge = 0;
495 	int carp_match = 0;
496 	struct sockaddr_in sin;
497 	sin.sin_len = sizeof(struct sockaddr_in);
498 	sin.sin_family = AF_INET;
499 	sin.sin_addr.s_addr = 0;
500 
501 	if (ifp->if_bridge)
502 		bridged = 1;
503 	if (ifp->if_type == IFT_BRIDGE)
504 		is_bridge = 1;
505 
506 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
507 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
508 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
509 		return;
510 	}
511 
512 	ah = mtod(m, struct arphdr *);
513 	op = ntohs(ah->ar_op);
514 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
515 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
516 
517 	if (op == ARPOP_REPLY)
518 		ARPSTAT_INC(rxreplies);
519 
520 	/*
521 	 * For a bridge, we want to check the address irrespective
522 	 * of the receive interface. (This will change slightly
523 	 * when we have clusters of interfaces).
524 	 * If the interface does not match, but the recieving interface
525 	 * is part of carp, we call carp_iamatch to see if this is a
526 	 * request for the virtual host ip.
527 	 * XXX: This is really ugly!
528 	 */
529 	IN_IFADDR_RLOCK();
530 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
531 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
532 		    ia->ia_ifp == ifp) &&
533 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
534 			ifa_ref(&ia->ia_ifa);
535 			IN_IFADDR_RUNLOCK();
536 			goto match;
537 		}
538 		if (ifp->if_carp != NULL &&
539 		    (*carp_iamatch_p)(ifp, ia, &isaddr, &enaddr) &&
540 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
541 			carp_match = 1;
542 			ifa_ref(&ia->ia_ifa);
543 			IN_IFADDR_RUNLOCK();
544 			goto match;
545 		}
546 	}
547 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
548 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
549 		    ia->ia_ifp == ifp) &&
550 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
551 			ifa_ref(&ia->ia_ifa);
552 			IN_IFADDR_RUNLOCK();
553 			goto match;
554 		}
555 
556 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
557   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
558   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
559   addr == ia->ia_addr.sin_addr.s_addr)
560 	/*
561 	 * Check the case when bridge shares its MAC address with
562 	 * some of its children, so packets are claimed by bridge
563 	 * itself (bridge_input() does it first), but they are really
564 	 * meant to be destined to the bridge member.
565 	 */
566 	if (is_bridge) {
567 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
568 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
569 				ifa_ref(&ia->ia_ifa);
570 				ifp = ia->ia_ifp;
571 				IN_IFADDR_RUNLOCK();
572 				goto match;
573 			}
574 		}
575 	}
576 #undef BDG_MEMBER_MATCHES_ARP
577 	IN_IFADDR_RUNLOCK();
578 
579 	/*
580 	 * No match, use the first inet address on the receive interface
581 	 * as a dummy address for the rest of the function.
582 	 */
583 	IF_ADDR_LOCK(ifp);
584 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
585 		if (ifa->ifa_addr->sa_family == AF_INET) {
586 			ia = ifatoia(ifa);
587 			ifa_ref(ifa);
588 			IF_ADDR_UNLOCK(ifp);
589 			goto match;
590 		}
591 	IF_ADDR_UNLOCK(ifp);
592 
593 	/*
594 	 * If bridging, fall back to using any inet address.
595 	 */
596 	IN_IFADDR_RLOCK();
597 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
598 		IN_IFADDR_RUNLOCK();
599 		goto drop;
600 	}
601 	ifa_ref(&ia->ia_ifa);
602 	IN_IFADDR_RUNLOCK();
603 match:
604 	if (!enaddr)
605 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
606 	myaddr = ia->ia_addr.sin_addr;
607 	ifa_free(&ia->ia_ifa);
608 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
609 		goto drop;	/* it's from me, ignore it. */
610 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
611 		log(LOG_ERR,
612 		    "arp: link address is broadcast for IP address %s!\n",
613 		    inet_ntoa(isaddr));
614 		goto drop;
615 	}
616 	/*
617 	 * Warn if another host is using the same IP address, but only if the
618 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
619 	 * case we suppress the warning to avoid false positive complaints of
620 	 * potential misconfiguration.
621 	 */
622 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
623 		log(LOG_ERR,
624 		   "arp: %*D is using my IP address %s on %s!\n",
625 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
626 		   inet_ntoa(isaddr), ifp->if_xname);
627 		itaddr = myaddr;
628 		ARPSTAT_INC(dupips);
629 		goto reply;
630 	}
631 	if (ifp->if_flags & IFF_STATICARP)
632 		goto reply;
633 
634 	bzero(&sin, sizeof(sin));
635 	sin.sin_len = sizeof(struct sockaddr_in);
636 	sin.sin_family = AF_INET;
637 	sin.sin_addr = isaddr;
638 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
639 	flags |= LLE_EXCLUSIVE;
640 	IF_AFDATA_LOCK(ifp);
641 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
642 	IF_AFDATA_UNLOCK(ifp);
643 	if (la != NULL) {
644 		/* the following is not an error when doing bridging */
645 		if (!bridged && la->lle_tbl->llt_ifp != ifp && !carp_match) {
646 			if (log_arp_wrong_iface)
647 				log(LOG_ERR, "arp: %s is on %s "
648 				    "but got reply from %*D on %s\n",
649 				    inet_ntoa(isaddr),
650 				    la->lle_tbl->llt_ifp->if_xname,
651 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
652 				    ifp->if_xname);
653 			LLE_WUNLOCK(la);
654 			goto reply;
655 		}
656 		if ((la->la_flags & LLE_VALID) &&
657 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
658 			if (la->la_flags & LLE_STATIC) {
659 				LLE_WUNLOCK(la);
660 				log(LOG_ERR,
661 				    "arp: %*D attempts to modify permanent "
662 				    "entry for %s on %s\n",
663 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
664 				    inet_ntoa(isaddr), ifp->if_xname);
665 				goto reply;
666 			}
667 			if (log_arp_movements) {
668 			        log(LOG_INFO, "arp: %s moved from %*D "
669 				    "to %*D on %s\n",
670 				    inet_ntoa(isaddr),
671 				    ifp->if_addrlen,
672 				    (u_char *)&la->ll_addr, ":",
673 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
674 				    ifp->if_xname);
675 			}
676 		}
677 
678 		if (ifp->if_addrlen != ah->ar_hln) {
679 			LLE_WUNLOCK(la);
680 			log(LOG_WARNING,
681 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
682 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
683 			    ah->ar_hln, ifp->if_addrlen);
684 			goto reply;
685 		}
686 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
687 		la->la_flags |= LLE_VALID;
688 
689 		EVENTHANDLER_INVOKE(arp_update_event, la);
690 
691 		if (!(la->la_flags & LLE_STATIC)) {
692 			int canceled;
693 
694 			LLE_ADDREF(la);
695 			la->la_expire = time_second + V_arpt_keep;
696 			canceled = callout_reset(&la->la_timer,
697 			    hz * V_arpt_keep, arptimer, la);
698 			if (canceled)
699 				LLE_REMREF(la);
700 		}
701 		la->la_asked = 0;
702 		la->la_preempt = V_arp_maxtries;
703 		hold = la->la_hold;
704 		if (hold != NULL) {
705 			la->la_hold = NULL;
706 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
707 		}
708 		LLE_WUNLOCK(la);
709 		if (hold != NULL)
710 			(*ifp->if_output)(ifp, hold, &sa, NULL);
711 	}
712 reply:
713 	if (op != ARPOP_REQUEST)
714 		goto drop;
715 	ARPSTAT_INC(rxrequests);
716 
717 	if (itaddr.s_addr == myaddr.s_addr) {
718 		/* Shortcut.. the receiving interface is the target. */
719 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
720 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
721 	} else {
722 		struct llentry *lle = NULL;
723 
724 		sin.sin_addr = itaddr;
725 		IF_AFDATA_LOCK(ifp);
726 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
727 		IF_AFDATA_UNLOCK(ifp);
728 
729 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
730 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
731 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
732 			LLE_RUNLOCK(lle);
733 		} else {
734 
735 			if (lle != NULL)
736 				LLE_RUNLOCK(lle);
737 
738 			if (!V_arp_proxyall)
739 				goto drop;
740 
741 			sin.sin_addr = itaddr;
742 			/* XXX MRT use table 0 for arp reply  */
743 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
744 			if (!rt)
745 				goto drop;
746 
747 			/*
748 			 * Don't send proxies for nodes on the same interface
749 			 * as this one came out of, or we'll get into a fight
750 			 * over who claims what Ether address.
751 			 */
752 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
753 				RTFREE_LOCKED(rt);
754 				goto drop;
755 			}
756 			RTFREE_LOCKED(rt);
757 
758 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
759 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
760 
761 			/*
762 			 * Also check that the node which sent the ARP packet
763 			 * is on the the interface we expect it to be on. This
764 			 * avoids ARP chaos if an interface is connected to the
765 			 * wrong network.
766 			 */
767 			sin.sin_addr = isaddr;
768 
769 			/* XXX MRT use table 0 for arp checks */
770 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
771 			if (!rt)
772 				goto drop;
773 			if (rt->rt_ifp != ifp) {
774 				log(LOG_INFO, "arp_proxy: ignoring request"
775 				    " from %s via %s, expecting %s\n",
776 				    inet_ntoa(isaddr), ifp->if_xname,
777 				    rt->rt_ifp->if_xname);
778 				RTFREE_LOCKED(rt);
779 				goto drop;
780 			}
781 			RTFREE_LOCKED(rt);
782 
783 #ifdef DEBUG_PROXY
784 			printf("arp: proxying for %s\n",
785 			       inet_ntoa(itaddr));
786 #endif
787 		}
788 	}
789 
790 	if (itaddr.s_addr == myaddr.s_addr &&
791 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
792 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
793 #ifdef DEBUG_LINKLOCAL
794 		printf("arp: sending reply for link-local addr %s\n",
795 		    inet_ntoa(itaddr));
796 #endif
797 		m->m_flags |= M_BCAST;
798 		m->m_flags &= ~M_MCAST;
799 	} else {
800 		/* default behaviour; never reply by broadcast. */
801 		m->m_flags &= ~(M_BCAST|M_MCAST);
802 	}
803 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
804 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
805 	ah->ar_op = htons(ARPOP_REPLY);
806 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
807 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
808 	m->m_pkthdr.len = m->m_len;
809 	sa.sa_family = AF_ARP;
810 	sa.sa_len = 2;
811 	(*ifp->if_output)(ifp, m, &sa, NULL);
812 	ARPSTAT_INC(txreplies);
813 	return;
814 
815 drop:
816 	m_freem(m);
817 }
818 #endif
819 
820 void
821 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
822 {
823 	struct llentry *lle;
824 
825 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
826 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
827 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
828 		/*
829 		 * interface address is considered static entry
830 		 * because the output of the arp utility shows
831 		 * that L2 entry as permanent
832 		 */
833 		IF_AFDATA_LOCK(ifp);
834 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
835 				 (struct sockaddr *)IA_SIN(ifa));
836 		IF_AFDATA_UNLOCK(ifp);
837 		if (lle == NULL)
838 			log(LOG_INFO, "arp_ifinit: cannot create arp "
839 			    "entry for interface address\n");
840 		else
841 			LLE_RUNLOCK(lle);
842 	}
843 	ifa->ifa_rtrequest = NULL;
844 }
845 
846 void
847 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
848 {
849 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
850 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
851 				&IA_SIN(ifa)->sin_addr, enaddr);
852 	ifa->ifa_rtrequest = NULL;
853 }
854 
855 static void
856 arp_init(void)
857 {
858 
859 	netisr_register(&arp_nh);
860 }
861 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
862