xref: /freebsd/sys/netinet/if_ether.c (revision 5b628e0c26d11e34ac50d8f83971db6450c9d872)
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 #include "opt_carp.h"
43 
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/queue.h>
47 #include <sys/sysctl.h>
48 #include <sys/systm.h>
49 #include <sys/mbuf.h>
50 #include <sys/malloc.h>
51 #include <sys/proc.h>
52 #include <sys/socket.h>
53 #include <sys/syslog.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/netisr.h>
59 #include <net/if_llc.h>
60 #include <net/ethernet.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <net/if_llatbl.h>
67 #include <netinet/if_ether.h>
68 
69 #include <net/if_arc.h>
70 #include <net/iso88025.h>
71 
72 #ifdef DEV_CARP
73 #include <netinet/ip_carp.h>
74 #endif
75 
76 #include <security/mac/mac_framework.h>
77 
78 #define SIN(s) ((struct sockaddr_in *)s)
79 #define SDL(s) ((struct sockaddr_dl *)s)
80 
81 SYSCTL_DECL(_net_link_ether);
82 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
83 
84 VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
85 					 * local traffic */
86 
87 /* timer values */
88 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
89 						 * minutes */
90 static VNET_DEFINE(int, arp_maxtries) = 5;
91 static VNET_DEFINE(int, arp_proxyall);
92 
93 #define	V_arpt_keep		VNET(arpt_keep)
94 #define	V_arp_maxtries		VNET(arp_maxtries)
95 #define	V_arp_proxyall		VNET(arp_proxyall)
96 
97 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
98 	&VNET_NAME(arpt_keep), 0,
99 	"ARP entry lifetime in seconds");
100 
101 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
102 	&VNET_NAME(arp_maxtries), 0,
103 	"ARP resolution attempts before returning error");
104 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
105 	&VNET_NAME(useloopback), 0,
106 	"Use the loopback interface for local traffic");
107 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
108 	&VNET_NAME(arp_proxyall), 0,
109 	"Enable proxy ARP for all suitable requests");
110 
111 static void	arp_init(void);
112 void		arprequest(struct ifnet *,
113 			struct in_addr *, struct in_addr *, u_char *);
114 static void	arpintr(struct mbuf *);
115 static void	arptimer(void *);
116 #ifdef INET
117 static void	in_arpinput(struct mbuf *);
118 #endif
119 
120 static const struct netisr_handler arp_nh = {
121 	.nh_name = "arp",
122 	.nh_handler = arpintr,
123 	.nh_proto = NETISR_ARP,
124 	.nh_policy = NETISR_POLICY_SOURCE,
125 };
126 
127 #ifdef AF_INET
128 void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
129 
130 /*
131  * called by in_ifscrub to remove entry from the table when
132  * the interface goes away
133  */
134 void
135 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
136 {
137 	struct sockaddr_in addr4;
138 
139 	bzero((void *)&addr4, sizeof(addr4));
140 	addr4.sin_len    = sizeof(addr4);
141 	addr4.sin_family = AF_INET;
142 	addr4.sin_addr.s_addr = addr;
143 	CURVNET_SET(ifp->if_vnet);
144 	IF_AFDATA_LOCK(ifp);
145 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
146 	    (struct sockaddr *)&addr4);
147 	IF_AFDATA_UNLOCK(ifp);
148 	CURVNET_RESTORE();
149 }
150 #endif
151 
152 /*
153  * Timeout routine.  Age arp_tab entries periodically.
154  */
155 static void
156 arptimer(void *arg)
157 {
158 	struct ifnet *ifp;
159 	struct llentry   *lle = (struct llentry *)arg;
160 
161 	if (lle == NULL) {
162 		panic("%s: NULL entry!\n", __func__);
163 		return;
164 	}
165 	ifp = lle->lle_tbl->llt_ifp;
166 	IF_AFDATA_LOCK(ifp);
167 	LLE_WLOCK(lle);
168 	if (((lle->la_flags & LLE_DELETED)
169 		|| (time_second >= lle->la_expire))
170 	    && (!callout_pending(&lle->la_timer) &&
171 		callout_active(&lle->la_timer)))
172 		(void) llentry_free(lle);
173 	else {
174 		/*
175 		 * Still valid, just drop our reference
176 		 */
177 		LLE_FREE_LOCKED(lle);
178 	}
179 	IF_AFDATA_UNLOCK(ifp);
180 }
181 
182 /*
183  * Broadcast an ARP request. Caller specifies:
184  *	- arp header source ip address
185  *	- arp header target ip address
186  *	- arp header source ethernet address
187  */
188 void
189 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
190     u_char *enaddr)
191 {
192 	struct mbuf *m;
193 	struct arphdr *ah;
194 	struct sockaddr sa;
195 
196 	if (sip == NULL) {
197 		/* XXX don't believe this can happen (or explain why) */
198 		/*
199 		 * The caller did not supply a source address, try to find
200 		 * a compatible one among those assigned to this interface.
201 		 */
202 		struct ifaddr *ifa;
203 
204 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
205 			if (!ifa->ifa_addr ||
206 			    ifa->ifa_addr->sa_family != AF_INET)
207 				continue;
208 			sip = &SIN(ifa->ifa_addr)->sin_addr;
209 			if (0 == ((sip->s_addr ^ tip->s_addr) &
210 			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
211 				break;  /* found it. */
212 		}
213 		if (sip == NULL) {
214 			printf("%s: cannot find matching address\n", __func__);
215 			return;
216 		}
217 	}
218 
219 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
220 		return;
221 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
222 		2*ifp->if_data.ifi_addrlen;
223 	m->m_pkthdr.len = m->m_len;
224 	MH_ALIGN(m, m->m_len);
225 	ah = mtod(m, struct arphdr *);
226 	bzero((caddr_t)ah, m->m_len);
227 #ifdef MAC
228 	mac_netinet_arp_send(ifp, m);
229 #endif
230 	ah->ar_pro = htons(ETHERTYPE_IP);
231 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
232 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
233 	ah->ar_op = htons(ARPOP_REQUEST);
234 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
235 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
236 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
237 	sa.sa_family = AF_ARP;
238 	sa.sa_len = 2;
239 	m->m_flags |= M_BCAST;
240 	(*ifp->if_output)(ifp, m, &sa, NULL);
241 }
242 
243 /*
244  * Resolve an IP address into an ethernet address.
245  * On input:
246  *    ifp is the interface we use
247  *    rt0 is the route to the final destination (possibly useless)
248  *    m is the mbuf. May be NULL if we don't have a packet.
249  *    dst is the next hop,
250  *    desten is where we want the address.
251  *
252  * On success, desten is filled in and the function returns 0;
253  * If the packet must be held pending resolution, we return EWOULDBLOCK
254  * On other errors, we return the corresponding error code.
255  * Note that m_freem() handles NULL.
256  */
257 int
258 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
259 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
260 {
261 	struct llentry *la = 0;
262 	u_int flags = 0;
263 	int error, renew;
264 
265 	*lle = NULL;
266 	if (m != NULL) {
267 		if (m->m_flags & M_BCAST) {
268 			/* broadcast */
269 			(void)memcpy(desten,
270 			    ifp->if_broadcastaddr, ifp->if_addrlen);
271 			return (0);
272 		}
273 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
274 			/* multicast */
275 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
276 			return (0);
277 		}
278 	}
279 	/* XXXXX
280 	 */
281 retry:
282 	IF_AFDATA_RLOCK(ifp);
283 	la = lla_lookup(LLTABLE(ifp), flags, dst);
284 	IF_AFDATA_RUNLOCK(ifp);
285 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
286 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
287 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
288 		IF_AFDATA_WLOCK(ifp);
289 		la = lla_lookup(LLTABLE(ifp), flags, dst);
290 		IF_AFDATA_WUNLOCK(ifp);
291 	}
292 	if (la == NULL) {
293 		if (flags & LLE_CREATE)
294 			log(LOG_DEBUG,
295 			    "arpresolve: can't allocate llinfo for %s\n",
296 			    inet_ntoa(SIN(dst)->sin_addr));
297 		m_freem(m);
298 		return (EINVAL);
299 	}
300 
301 	if ((la->la_flags & LLE_VALID) &&
302 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
303 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
304 		/*
305 		 * If entry has an expiry time and it is approaching,
306 		 * see if we need to send an ARP request within this
307 		 * arpt_down interval.
308 		 */
309 		if (!(la->la_flags & LLE_STATIC) &&
310 		    time_uptime + la->la_preempt > la->la_expire) {
311 			arprequest(ifp, NULL,
312 			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
313 
314 			la->la_preempt--;
315 		}
316 
317 		*lle = la;
318 		error = 0;
319 		goto done;
320 	}
321 
322 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
323 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
324 		    inet_ntoa(SIN(dst)->sin_addr));
325 		m_freem(m);
326 		error = EINVAL;
327 		goto done;
328 	}
329 
330 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
331 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
332 		flags |= LLE_EXCLUSIVE;
333 		LLE_RUNLOCK(la);
334 		goto retry;
335 	}
336 	/*
337 	 * There is an arptab entry, but no ethernet address
338 	 * response yet.  Replace the held mbuf with this
339 	 * latest one.
340 	 */
341 	if (m != NULL) {
342 		if (la->la_hold != NULL)
343 			m_freem(la->la_hold);
344 		la->la_hold = m;
345 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
346 			flags &= ~LLE_EXCLUSIVE;
347 			LLE_DOWNGRADE(la);
348 		}
349 
350 	}
351 	/*
352 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
353 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
354 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
355 	 * ARP request, but not faster than one request per second.
356 	 */
357 	if (la->la_asked < V_arp_maxtries)
358 		error = EWOULDBLOCK;	/* First request. */
359 	else
360 		error =
361 		    (rt0->rt_flags & RTF_GATEWAY) ? EHOSTDOWN : EHOSTUNREACH;
362 
363 	if (renew) {
364 		LLE_ADDREF(la);
365 		la->la_expire = time_uptime;
366 		callout_reset(&la->la_timer, hz, arptimer, la);
367 		la->la_asked++;
368 		LLE_WUNLOCK(la);
369 		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
370 		    IF_LLADDR(ifp));
371 		return (error);
372 	}
373 done:
374 	if (flags & LLE_EXCLUSIVE)
375 		LLE_WUNLOCK(la);
376 	else
377 		LLE_RUNLOCK(la);
378 	return (error);
379 }
380 
381 /*
382  * Common length and type checks are done here,
383  * then the protocol-specific routine is called.
384  */
385 static void
386 arpintr(struct mbuf *m)
387 {
388 	struct arphdr *ar;
389 
390 	if (m->m_len < sizeof(struct arphdr) &&
391 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
392 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
393 		return;
394 	}
395 	ar = mtod(m, struct arphdr *);
396 
397 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
398 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
399 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
400 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
401 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
402 		    (unsigned char *)&ar->ar_hrd, "");
403 		m_freem(m);
404 		return;
405 	}
406 
407 	if (m->m_len < arphdr_len(ar)) {
408 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
409 			log(LOG_ERR, "arp: runt packet\n");
410 			m_freem(m);
411 			return;
412 		}
413 		ar = mtod(m, struct arphdr *);
414 	}
415 
416 	switch (ntohs(ar->ar_pro)) {
417 #ifdef INET
418 	case ETHERTYPE_IP:
419 		in_arpinput(m);
420 		return;
421 #endif
422 	}
423 	m_freem(m);
424 }
425 
426 #ifdef INET
427 /*
428  * ARP for Internet protocols on 10 Mb/s Ethernet.
429  * Algorithm is that given in RFC 826.
430  * In addition, a sanity check is performed on the sender
431  * protocol address, to catch impersonators.
432  * We no longer handle negotiations for use of trailer protocol:
433  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
434  * along with IP replies if we wanted trailers sent to us,
435  * and also sent them in response to IP replies.
436  * This allowed either end to announce the desire to receive
437  * trailer packets.
438  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
439  * but formerly didn't normally send requests.
440  */
441 static int log_arp_wrong_iface = 1;
442 static int log_arp_movements = 1;
443 static int log_arp_permanent_modify = 1;
444 
445 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
446 	&log_arp_wrong_iface, 0,
447 	"log arp packets arriving on the wrong interface");
448 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
449         &log_arp_movements, 0,
450         "log arp replies from MACs different than the one in the cache");
451 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
452         &log_arp_permanent_modify, 0,
453         "log arp replies from MACs different than the one in the permanent arp entry");
454 
455 
456 static void
457 in_arpinput(struct mbuf *m)
458 {
459 	struct arphdr *ah;
460 	struct ifnet *ifp = m->m_pkthdr.rcvif;
461 	struct llentry *la = NULL;
462 	struct rtentry *rt;
463 	struct ifaddr *ifa;
464 	struct in_ifaddr *ia;
465 	struct mbuf *hold;
466 	struct sockaddr sa;
467 	struct in_addr isaddr, itaddr, myaddr;
468 	u_int8_t *enaddr = NULL;
469 	int op, flags;
470 	int req_len;
471 	int bridged = 0, is_bridge = 0;
472 #ifdef DEV_CARP
473 	int carp_match = 0;
474 #endif
475 	struct sockaddr_in sin;
476 	sin.sin_len = sizeof(struct sockaddr_in);
477 	sin.sin_family = AF_INET;
478 	sin.sin_addr.s_addr = 0;
479 
480 	if (ifp->if_bridge)
481 		bridged = 1;
482 	if (ifp->if_type == IFT_BRIDGE)
483 		is_bridge = 1;
484 
485 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
486 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
487 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
488 		return;
489 	}
490 
491 	ah = mtod(m, struct arphdr *);
492 	op = ntohs(ah->ar_op);
493 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
494 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
495 
496 	/*
497 	 * For a bridge, we want to check the address irrespective
498 	 * of the receive interface. (This will change slightly
499 	 * when we have clusters of interfaces).
500 	 * If the interface does not match, but the recieving interface
501 	 * is part of carp, we call carp_iamatch to see if this is a
502 	 * request for the virtual host ip.
503 	 * XXX: This is really ugly!
504 	 */
505 	IN_IFADDR_RLOCK();
506 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
507 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
508 		    ia->ia_ifp == ifp) &&
509 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
510 			ifa_ref(&ia->ia_ifa);
511 			IN_IFADDR_RUNLOCK();
512 			goto match;
513 		}
514 #ifdef DEV_CARP
515 		if (ifp->if_carp != NULL &&
516 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
517 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
518 			carp_match = 1;
519 			ifa_ref(&ia->ia_ifa);
520 			IN_IFADDR_RUNLOCK();
521 			goto match;
522 		}
523 #endif
524 	}
525 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
526 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
527 		    ia->ia_ifp == ifp) &&
528 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
529 			ifa_ref(&ia->ia_ifa);
530 			IN_IFADDR_RUNLOCK();
531 			goto match;
532 		}
533 
534 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
535   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
536   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
537   addr == ia->ia_addr.sin_addr.s_addr)
538 	/*
539 	 * Check the case when bridge shares its MAC address with
540 	 * some of its children, so packets are claimed by bridge
541 	 * itself (bridge_input() does it first), but they are really
542 	 * meant to be destined to the bridge member.
543 	 */
544 	if (is_bridge) {
545 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
546 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
547 				ifa_ref(&ia->ia_ifa);
548 				ifp = ia->ia_ifp;
549 				IN_IFADDR_RUNLOCK();
550 				goto match;
551 			}
552 		}
553 	}
554 #undef BDG_MEMBER_MATCHES_ARP
555 	IN_IFADDR_RUNLOCK();
556 
557 	/*
558 	 * No match, use the first inet address on the receive interface
559 	 * as a dummy address for the rest of the function.
560 	 */
561 	IF_ADDR_LOCK(ifp);
562 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
563 		if (ifa->ifa_addr->sa_family == AF_INET) {
564 			ia = ifatoia(ifa);
565 			ifa_ref(ifa);
566 			IF_ADDR_UNLOCK(ifp);
567 			goto match;
568 		}
569 	IF_ADDR_UNLOCK(ifp);
570 
571 	/*
572 	 * If bridging, fall back to using any inet address.
573 	 */
574 	IN_IFADDR_RLOCK();
575 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
576 		IN_IFADDR_RUNLOCK();
577 		goto drop;
578 	}
579 	ifa_ref(&ia->ia_ifa);
580 	IN_IFADDR_RUNLOCK();
581 match:
582 	if (!enaddr)
583 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
584 	myaddr = ia->ia_addr.sin_addr;
585 	ifa_free(&ia->ia_ifa);
586 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
587 		goto drop;	/* it's from me, ignore it. */
588 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
589 		log(LOG_ERR,
590 		    "arp: link address is broadcast for IP address %s!\n",
591 		    inet_ntoa(isaddr));
592 		goto drop;
593 	}
594 	/*
595 	 * Warn if another host is using the same IP address, but only if the
596 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
597 	 * case we suppress the warning to avoid false positive complaints of
598 	 * potential misconfiguration.
599 	 */
600 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
601 		log(LOG_ERR,
602 		   "arp: %*D is using my IP address %s on %s!\n",
603 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
604 		   inet_ntoa(isaddr), ifp->if_xname);
605 		itaddr = myaddr;
606 		goto reply;
607 	}
608 	if (ifp->if_flags & IFF_STATICARP)
609 		goto reply;
610 
611 	bzero(&sin, sizeof(sin));
612 	sin.sin_len = sizeof(struct sockaddr_in);
613 	sin.sin_family = AF_INET;
614 	sin.sin_addr = isaddr;
615 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
616 	flags |= LLE_EXCLUSIVE;
617 	IF_AFDATA_LOCK(ifp);
618 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
619 	IF_AFDATA_UNLOCK(ifp);
620 	if (la != NULL) {
621 		/* the following is not an error when doing bridging */
622 		if (!bridged && la->lle_tbl->llt_ifp != ifp
623 #ifdef DEV_CARP
624 		    && (ifp->if_type != IFT_CARP || !carp_match)
625 #endif
626 			) {
627 			if (log_arp_wrong_iface)
628 				log(LOG_ERR, "arp: %s is on %s "
629 				    "but got reply from %*D on %s\n",
630 				    inet_ntoa(isaddr),
631 				    la->lle_tbl->llt_ifp->if_xname,
632 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
633 				    ifp->if_xname);
634 			LLE_WUNLOCK(la);
635 			goto reply;
636 		}
637 		if ((la->la_flags & LLE_VALID) &&
638 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
639 			if (la->la_flags & LLE_STATIC) {
640 				LLE_WUNLOCK(la);
641 				log(LOG_ERR,
642 				    "arp: %*D attempts to modify permanent "
643 				    "entry for %s on %s\n",
644 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
645 				    inet_ntoa(isaddr), ifp->if_xname);
646 				goto reply;
647 			}
648 			if (log_arp_movements) {
649 			        log(LOG_INFO, "arp: %s moved from %*D "
650 				    "to %*D on %s\n",
651 				    inet_ntoa(isaddr),
652 				    ifp->if_addrlen,
653 				    (u_char *)&la->ll_addr, ":",
654 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
655 				    ifp->if_xname);
656 			}
657 		}
658 
659 		if (ifp->if_addrlen != ah->ar_hln) {
660 			LLE_WUNLOCK(la);
661 			log(LOG_WARNING,
662 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
663 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
664 			    ah->ar_hln, ifp->if_addrlen);
665 			goto reply;
666 		}
667 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
668 		la->la_flags |= LLE_VALID;
669 
670 		if (!(la->la_flags & LLE_STATIC)) {
671 			la->la_expire = time_uptime + V_arpt_keep;
672 			callout_reset(&la->la_timer, hz * V_arpt_keep,
673 			    arptimer, la);
674 		}
675 		la->la_asked = 0;
676 		la->la_preempt = V_arp_maxtries;
677 		hold = la->la_hold;
678 		if (hold != NULL) {
679 			la->la_hold = NULL;
680 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
681 		}
682 		LLE_WUNLOCK(la);
683 		if (hold != NULL)
684 			(*ifp->if_output)(ifp, hold, &sa, NULL);
685 	}
686 reply:
687 	if (op != ARPOP_REQUEST)
688 		goto drop;
689 
690 	if (itaddr.s_addr == myaddr.s_addr) {
691 		/* Shortcut.. the receiving interface is the target. */
692 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
693 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
694 	} else {
695 		struct llentry *lle = NULL;
696 
697 		if (!V_arp_proxyall)
698 			goto drop;
699 
700 		sin.sin_addr = itaddr;
701 		/* XXX MRT use table 0 for arp reply  */
702 		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
703 		if (!rt)
704 			goto drop;
705 
706 		/*
707 		 * Don't send proxies for nodes on the same interface
708 		 * as this one came out of, or we'll get into a fight
709 		 * over who claims what Ether address.
710 		 */
711 		if (!rt->rt_ifp || rt->rt_ifp == ifp) {
712 			RTFREE_LOCKED(rt);
713 			goto drop;
714 		}
715 		IF_AFDATA_LOCK(rt->rt_ifp);
716 		lle = lla_lookup(LLTABLE(rt->rt_ifp), 0, (struct sockaddr *)&sin);
717 		IF_AFDATA_UNLOCK(rt->rt_ifp);
718 		RTFREE_LOCKED(rt);
719 
720 		if (lle != NULL) {
721 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
722 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
723 			LLE_RUNLOCK(lle);
724 		} else
725 			goto drop;
726 
727 		/*
728 		 * Also check that the node which sent the ARP packet
729 		 * is on the the interface we expect it to be on. This
730 		 * avoids ARP chaos if an interface is connected to the
731 		 * wrong network.
732 		 */
733 		sin.sin_addr = isaddr;
734 
735 		/* XXX MRT use table 0 for arp checks */
736 		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
737 		if (!rt)
738 			goto drop;
739 		if (rt->rt_ifp != ifp) {
740 			log(LOG_INFO, "arp_proxy: ignoring request"
741 			    " from %s via %s, expecting %s\n",
742 			    inet_ntoa(isaddr), ifp->if_xname,
743 			    rt->rt_ifp->if_xname);
744 			RTFREE_LOCKED(rt);
745 			goto drop;
746 		}
747 		RTFREE_LOCKED(rt);
748 
749 #ifdef DEBUG_PROXY
750 		printf("arp: proxying for %s\n",
751 		       inet_ntoa(itaddr));
752 #endif
753 	}
754 
755 	if (itaddr.s_addr == myaddr.s_addr &&
756 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
757 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
758 #ifdef DEBUG_LINKLOCAL
759 		printf("arp: sending reply for link-local addr %s\n",
760 		    inet_ntoa(itaddr));
761 #endif
762 		m->m_flags |= M_BCAST;
763 		m->m_flags &= ~M_MCAST;
764 	} else {
765 		/* default behaviour; never reply by broadcast. */
766 		m->m_flags &= ~(M_BCAST|M_MCAST);
767 	}
768 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
769 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
770 	ah->ar_op = htons(ARPOP_REPLY);
771 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
772 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
773 	m->m_pkthdr.len = m->m_len;
774 	sa.sa_family = AF_ARP;
775 	sa.sa_len = 2;
776 	(*ifp->if_output)(ifp, m, &sa, NULL);
777 	return;
778 
779 drop:
780 	m_freem(m);
781 }
782 #endif
783 
784 void
785 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
786 {
787 	struct llentry *lle;
788 
789 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
790 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
791 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
792 		/*
793 		 * interface address is considered static entry
794 		 * because the output of the arp utility shows
795 		 * that L2 entry as permanent
796 		 */
797 		IF_AFDATA_LOCK(ifp);
798 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
799 				 (struct sockaddr *)IA_SIN(ifa));
800 		IF_AFDATA_UNLOCK(ifp);
801 		if (lle == NULL)
802 			log(LOG_INFO, "arp_ifinit: cannot create arp "
803 			    "entry for interface address\n");
804 		else
805 			LLE_RUNLOCK(lle);
806 	}
807 	ifa->ifa_rtrequest = NULL;
808 }
809 
810 void
811 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
812 {
813 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
814 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
815 				&IA_SIN(ifa)->sin_addr, enaddr);
816 	ifa->ifa_rtrequest = NULL;
817 }
818 
819 static void
820 arp_init(void)
821 {
822 
823 	netisr_register(&arp_nh);
824 }
825 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
826