xref: /freebsd/sys/netinet/if_ether.c (revision aa79fe245de7616cda41b69a296a5ce209c95c45)
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 #include <sys/vimage.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/route.h>
60 #include <net/netisr.h>
61 #include <net/if_llc.h>
62 #include <net/ethernet.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <net/if_llatbl.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/vinet.h>
70 
71 #include <net/if_arc.h>
72 #include <net/iso88025.h>
73 
74 #ifdef DEV_CARP
75 #include <netinet/ip_carp.h>
76 #endif
77 
78 #include <security/mac/mac_framework.h>
79 
80 #define SIN(s) ((struct sockaddr_in *)s)
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 SYSCTL_DECL(_net_link_ether);
84 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
85 
86 /* timer values */
87 #ifdef VIMAGE_GLOBALS
88 static int	arpt_keep; /* once resolved, good for 20 more minutes */
89 static int	arp_maxtries;
90 int	useloopback; /* use loopback interface for local traffic */
91 static int	arp_proxyall;
92 #endif
93 
94 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, max_age,
95     CTLFLAG_RW, arpt_keep, 0, "ARP entry lifetime in seconds");
96 
97 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries,
98 	CTLFLAG_RW, arp_maxtries, 0,
99 	"ARP resolution attempts before returning error");
100 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, useloopback,
101 	CTLFLAG_RW, useloopback, 0,
102 	"Use the loopback interface for local traffic");
103 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall,
104 	CTLFLAG_RW, arp_proxyall, 0,
105 	"Enable proxy ARP for all suitable requests");
106 
107 static void	arp_init(void);
108 static int	arp_iattach(const void *);
109 void		arprequest(struct ifnet *,
110 			struct in_addr *, struct in_addr *, u_char *);
111 static void	arpintr(struct mbuf *);
112 static void	arptimer(void *);
113 #ifdef INET
114 static void	in_arpinput(struct mbuf *);
115 #endif
116 
117 static const struct netisr_handler arp_nh = {
118 	.nh_name = "arp",
119 	.nh_handler = arpintr,
120 	.nh_proto = NETISR_ARP,
121 	.nh_policy = NETISR_POLICY_SOURCE,
122 };
123 
124 #ifndef VIMAGE_GLOBALS
125 static const vnet_modinfo_t vnet_arp_modinfo = {
126 	.vmi_id		= VNET_MOD_ARP,
127 	.vmi_name	= "arp",
128 	.vmi_dependson	= VNET_MOD_INET,
129 	.vmi_iattach	= arp_iattach
130 };
131 #endif /* !VIMAGE_GLOBALS */
132 
133 #ifdef AF_INET
134 void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
135 
136 /*
137  * called by in_ifscrub to remove entry from the table when
138  * the interface goes away
139  */
140 void
141 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
142 {
143 	struct sockaddr_in addr4;
144 
145 	bzero((void *)&addr4, sizeof(addr4));
146 	addr4.sin_len    = sizeof(addr4);
147 	addr4.sin_family = AF_INET;
148 	addr4.sin_addr.s_addr = addr;
149 	CURVNET_SET(ifp->if_vnet);
150 	IF_AFDATA_LOCK(ifp);
151 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
152 	    (struct sockaddr *)&addr4);
153 	IF_AFDATA_UNLOCK(ifp);
154 	CURVNET_RESTORE();
155 }
156 #endif
157 
158 /*
159  * Timeout routine.  Age arp_tab entries periodically.
160  */
161 static void
162 arptimer(void *arg)
163 {
164 	struct ifnet *ifp;
165 	struct llentry   *lle = (struct llentry *)arg;
166 
167 	if (lle == NULL) {
168 		panic("%s: NULL entry!\n", __func__);
169 		return;
170 	}
171 	ifp = lle->lle_tbl->llt_ifp;
172 	IF_AFDATA_LOCK(ifp);
173 	LLE_WLOCK(lle);
174 	if (((lle->la_flags & LLE_DELETED)
175 		|| (time_second >= lle->la_expire))
176 	    && (!callout_pending(&lle->la_timer) &&
177 		callout_active(&lle->la_timer)))
178 		(void) llentry_free(lle);
179 	else {
180 		/*
181 		 * Still valid, just drop our reference
182 		 */
183 		LLE_FREE_LOCKED(lle);
184 	}
185 	IF_AFDATA_UNLOCK(ifp);
186 }
187 
188 /*
189  * Broadcast an ARP request. Caller specifies:
190  *	- arp header source ip address
191  *	- arp header target ip address
192  *	- arp header source ethernet address
193  */
194 void
195 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
196     u_char *enaddr)
197 {
198 	struct mbuf *m;
199 	struct arphdr *ah;
200 	struct sockaddr sa;
201 
202 	if (sip == NULL) {
203 		/* XXX don't believe this can happen (or explain why) */
204 		/*
205 		 * The caller did not supply a source address, try to find
206 		 * a compatible one among those assigned to this interface.
207 		 */
208 		struct ifaddr *ifa;
209 
210 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
211 			if (!ifa->ifa_addr ||
212 			    ifa->ifa_addr->sa_family != AF_INET)
213 				continue;
214 			sip = &SIN(ifa->ifa_addr)->sin_addr;
215 			if (0 == ((sip->s_addr ^ tip->s_addr) &
216 			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
217 				break;  /* found it. */
218 		}
219 		if (sip == NULL) {
220 			printf("%s: cannot find matching address\n", __func__);
221 			return;
222 		}
223 	}
224 
225 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
226 		return;
227 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
228 		2*ifp->if_data.ifi_addrlen;
229 	m->m_pkthdr.len = m->m_len;
230 	MH_ALIGN(m, m->m_len);
231 	ah = mtod(m, struct arphdr *);
232 	bzero((caddr_t)ah, m->m_len);
233 #ifdef MAC
234 	mac_netinet_arp_send(ifp, m);
235 #endif
236 	ah->ar_pro = htons(ETHERTYPE_IP);
237 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
238 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
239 	ah->ar_op = htons(ARPOP_REQUEST);
240 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
241 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
242 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
243 	sa.sa_family = AF_ARP;
244 	sa.sa_len = 2;
245 	m->m_flags |= M_BCAST;
246 	(*ifp->if_output)(ifp, m, &sa, NULL);
247 }
248 
249 /*
250  * Resolve an IP address into an ethernet address.
251  * On input:
252  *    ifp is the interface we use
253  *    rt0 is the route to the final destination (possibly useless)
254  *    m is the mbuf. May be NULL if we don't have a packet.
255  *    dst is the next hop,
256  *    desten is where we want the address.
257  *
258  * On success, desten is filled in and the function returns 0;
259  * If the packet must be held pending resolution, we return EWOULDBLOCK
260  * On other errors, we return the corresponding error code.
261  * Note that m_freem() handles NULL.
262  */
263 int
264 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
265 	struct sockaddr *dst, u_char *desten, struct llentry **lle)
266 {
267 	INIT_VNET_INET(ifp->if_vnet);
268 	struct llentry *la = 0;
269 	u_int flags = 0;
270 	int error, renew;
271 
272 	*lle = NULL;
273 	if (m != NULL) {
274 		if (m->m_flags & M_BCAST) {
275 			/* broadcast */
276 			(void)memcpy(desten,
277 			    ifp->if_broadcastaddr, ifp->if_addrlen);
278 			return (0);
279 		}
280 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
281 			/* multicast */
282 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
283 			return (0);
284 		}
285 	}
286 	/* XXXXX
287 	 */
288 retry:
289 	IF_AFDATA_RLOCK(ifp);
290 	la = lla_lookup(LLTABLE(ifp), flags, dst);
291 	IF_AFDATA_RUNLOCK(ifp);
292 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
293 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
294 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
295 		IF_AFDATA_WLOCK(ifp);
296 		la = lla_lookup(LLTABLE(ifp), flags, dst);
297 		IF_AFDATA_WUNLOCK(ifp);
298 	}
299 	if (la == NULL) {
300 		if (flags & LLE_CREATE)
301 			log(LOG_DEBUG,
302 			    "arpresolve: can't allocate llinfo for %s\n",
303 			    inet_ntoa(SIN(dst)->sin_addr));
304 		m_freem(m);
305 		return (EINVAL);
306 	}
307 
308 	if ((la->la_flags & LLE_VALID) &&
309 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
310 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
311 		/*
312 		 * If entry has an expiry time and it is approaching,
313 		 * see if we need to send an ARP request within this
314 		 * arpt_down interval.
315 		 */
316 		if (!(la->la_flags & LLE_STATIC) &&
317 		    time_uptime + la->la_preempt > la->la_expire) {
318 			arprequest(ifp, NULL,
319 			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
320 
321 			la->la_preempt--;
322 		}
323 
324 		*lle = la;
325 		error = 0;
326 		goto done;
327 	}
328 
329 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
330 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
331 		    inet_ntoa(SIN(dst)->sin_addr));
332 		m_freem(m);
333 		error = EINVAL;
334 		goto done;
335 	}
336 
337 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
338 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
339 		flags |= LLE_EXCLUSIVE;
340 		LLE_RUNLOCK(la);
341 		goto retry;
342 	}
343 	/*
344 	 * There is an arptab entry, but no ethernet address
345 	 * response yet.  Replace the held mbuf with this
346 	 * latest one.
347 	 */
348 	if (m != NULL) {
349 		if (la->la_hold != NULL)
350 			m_freem(la->la_hold);
351 		la->la_hold = m;
352 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
353 			flags &= ~LLE_EXCLUSIVE;
354 			LLE_DOWNGRADE(la);
355 		}
356 
357 	}
358 	/*
359 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
360 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
361 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
362 	 * ARP request, but not faster than one request per second.
363 	 */
364 	if (la->la_asked < V_arp_maxtries)
365 		error = EWOULDBLOCK;	/* First request. */
366 	else
367 		error =
368 		    (rt0->rt_flags & RTF_GATEWAY) ? EHOSTDOWN : EHOSTUNREACH;
369 
370 	if (renew) {
371 		LLE_ADDREF(la);
372 		la->la_expire = time_uptime;
373 		callout_reset(&la->la_timer, hz, arptimer, la);
374 		la->la_asked++;
375 		LLE_WUNLOCK(la);
376 		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
377 		    IF_LLADDR(ifp));
378 		return (error);
379 	}
380 done:
381 	if (flags & LLE_EXCLUSIVE)
382 		LLE_WUNLOCK(la);
383 	else
384 		LLE_RUNLOCK(la);
385 	return (error);
386 }
387 
388 /*
389  * Common length and type checks are done here,
390  * then the protocol-specific routine is called.
391  */
392 static void
393 arpintr(struct mbuf *m)
394 {
395 	struct arphdr *ar;
396 
397 	if (m->m_len < sizeof(struct arphdr) &&
398 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
399 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
400 		return;
401 	}
402 	ar = mtod(m, struct arphdr *);
403 
404 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
405 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
406 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
407 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
408 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
409 		    (unsigned char *)&ar->ar_hrd, "");
410 		m_freem(m);
411 		return;
412 	}
413 
414 	if (m->m_len < arphdr_len(ar)) {
415 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
416 			log(LOG_ERR, "arp: runt packet\n");
417 			m_freem(m);
418 			return;
419 		}
420 		ar = mtod(m, struct arphdr *);
421 	}
422 
423 	switch (ntohs(ar->ar_pro)) {
424 #ifdef INET
425 	case ETHERTYPE_IP:
426 		in_arpinput(m);
427 		return;
428 #endif
429 	}
430 	m_freem(m);
431 }
432 
433 #ifdef INET
434 /*
435  * ARP for Internet protocols on 10 Mb/s Ethernet.
436  * Algorithm is that given in RFC 826.
437  * In addition, a sanity check is performed on the sender
438  * protocol address, to catch impersonators.
439  * We no longer handle negotiations for use of trailer protocol:
440  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
441  * along with IP replies if we wanted trailers sent to us,
442  * and also sent them in response to IP replies.
443  * This allowed either end to announce the desire to receive
444  * trailer packets.
445  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
446  * but formerly didn't normally send requests.
447  */
448 static int log_arp_wrong_iface = 1;
449 static int log_arp_movements = 1;
450 static int log_arp_permanent_modify = 1;
451 
452 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
453 	&log_arp_wrong_iface, 0,
454 	"log arp packets arriving on the wrong interface");
455 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
456         &log_arp_movements, 0,
457         "log arp replies from MACs different than the one in the cache");
458 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
459         &log_arp_permanent_modify, 0,
460         "log arp replies from MACs different than the one in the permanent arp entry");
461 
462 
463 static void
464 in_arpinput(struct mbuf *m)
465 {
466 	struct arphdr *ah;
467 	struct ifnet *ifp = m->m_pkthdr.rcvif;
468 	struct llentry *la = NULL;
469 	struct rtentry *rt;
470 	struct ifaddr *ifa;
471 	struct in_ifaddr *ia;
472 	struct sockaddr sa;
473 	struct in_addr isaddr, itaddr, myaddr;
474 	u_int8_t *enaddr = NULL;
475 	int op, flags;
476 	struct mbuf *m0;
477 	int req_len;
478 	int bridged = 0, is_bridge = 0;
479 #ifdef DEV_CARP
480 	int carp_match = 0;
481 #endif
482 	struct sockaddr_in sin;
483 	sin.sin_len = sizeof(struct sockaddr_in);
484 	sin.sin_family = AF_INET;
485 	sin.sin_addr.s_addr = 0;
486 	INIT_VNET_INET(ifp->if_vnet);
487 
488 	if (ifp->if_bridge)
489 		bridged = 1;
490 	if (ifp->if_type == IFT_BRIDGE)
491 		is_bridge = 1;
492 
493 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
494 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
495 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
496 		return;
497 	}
498 
499 	ah = mtod(m, struct arphdr *);
500 	op = ntohs(ah->ar_op);
501 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
502 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
503 
504 	/*
505 	 * For a bridge, we want to check the address irrespective
506 	 * of the receive interface. (This will change slightly
507 	 * when we have clusters of interfaces).
508 	 * If the interface does not match, but the recieving interface
509 	 * is part of carp, we call carp_iamatch to see if this is a
510 	 * request for the virtual host ip.
511 	 * XXX: This is really ugly!
512 	 */
513 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
514 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
515 		    ia->ia_ifp == ifp) &&
516 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
517 			goto match;
518 #ifdef DEV_CARP
519 		if (ifp->if_carp != NULL &&
520 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
521 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
522 			carp_match = 1;
523 			goto match;
524 		}
525 #endif
526 	}
527 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
528 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
529 		    ia->ia_ifp == ifp) &&
530 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
531 			goto match;
532 
533 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
534   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
535   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
536   addr == ia->ia_addr.sin_addr.s_addr)
537 	/*
538 	 * Check the case when bridge shares its MAC address with
539 	 * some of its children, so packets are claimed by bridge
540 	 * itself (bridge_input() does it first), but they are really
541 	 * meant to be destined to the bridge member.
542 	 */
543 	if (is_bridge) {
544 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
545 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
546 				ifp = ia->ia_ifp;
547 				goto match;
548 			}
549 		}
550 	}
551 #undef BDG_MEMBER_MATCHES_ARP
552 
553 	/*
554 	 * No match, use the first inet address on the receive interface
555 	 * as a dummy address for the rest of the function.
556 	 */
557 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
558 		if (ifa->ifa_addr->sa_family == AF_INET) {
559 			ia = ifatoia(ifa);
560 			goto match;
561 		}
562 	/*
563 	 * If bridging, fall back to using any inet address.
564 	 */
565 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL)
566 		goto drop;
567 match:
568 	if (!enaddr)
569 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
570 	myaddr = ia->ia_addr.sin_addr;
571 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
572 		goto drop;	/* it's from me, ignore it. */
573 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
574 		log(LOG_ERR,
575 		    "arp: link address is broadcast for IP address %s!\n",
576 		    inet_ntoa(isaddr));
577 		goto drop;
578 	}
579 	/*
580 	 * Warn if another host is using the same IP address, but only if the
581 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
582 	 * case we suppress the warning to avoid false positive complaints of
583 	 * potential misconfiguration.
584 	 */
585 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
586 		log(LOG_ERR,
587 		   "arp: %*D is using my IP address %s on %s!\n",
588 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
589 		   inet_ntoa(isaddr), ifp->if_xname);
590 		itaddr = myaddr;
591 		goto reply;
592 	}
593 	if (ifp->if_flags & IFF_STATICARP)
594 		goto reply;
595 
596 	bzero(&sin, sizeof(sin));
597 	sin.sin_len = sizeof(struct sockaddr_in);
598 	sin.sin_family = AF_INET;
599 	sin.sin_addr = isaddr;
600 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
601 	flags |= LLE_EXCLUSIVE;
602 	IF_AFDATA_LOCK(ifp);
603 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
604 	IF_AFDATA_UNLOCK(ifp);
605 	if (la != NULL) {
606 		/* the following is not an error when doing bridging */
607 		if (!bridged && la->lle_tbl->llt_ifp != ifp
608 #ifdef DEV_CARP
609 		    && (ifp->if_type != IFT_CARP || !carp_match)
610 #endif
611 			) {
612 			if (log_arp_wrong_iface)
613 				log(LOG_ERR, "arp: %s is on %s "
614 				    "but got reply from %*D on %s\n",
615 				    inet_ntoa(isaddr),
616 				    la->lle_tbl->llt_ifp->if_xname,
617 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
618 				    ifp->if_xname);
619 			goto reply;
620 		}
621 		if ((la->la_flags & LLE_VALID) &&
622 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
623 			if (la->la_flags & LLE_STATIC) {
624 				log(LOG_ERR,
625 				    "arp: %*D attempts to modify permanent "
626 				    "entry for %s on %s\n",
627 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
628 				    inet_ntoa(isaddr), ifp->if_xname);
629 				goto reply;
630 			}
631 			if (log_arp_movements) {
632 			        log(LOG_INFO, "arp: %s moved from %*D "
633 				    "to %*D on %s\n",
634 				    inet_ntoa(isaddr),
635 				    ifp->if_addrlen,
636 				    (u_char *)&la->ll_addr, ":",
637 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
638 				    ifp->if_xname);
639 			}
640 		}
641 
642 		if (ifp->if_addrlen != ah->ar_hln) {
643 			log(LOG_WARNING,
644 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
645 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
646 			    ah->ar_hln, ifp->if_addrlen);
647 			goto reply;
648 		}
649 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
650 		la->la_flags |= LLE_VALID;
651 
652 		if (!(la->la_flags & LLE_STATIC)) {
653 			la->la_expire = time_uptime + V_arpt_keep;
654 			callout_reset(&la->la_timer, hz * V_arpt_keep,
655 			    arptimer, la);
656 		}
657 		la->la_asked = 0;
658 		la->la_preempt = V_arp_maxtries;
659 		if (la->la_hold != NULL) {
660 			m0 = la->la_hold;
661 			la->la_hold = 0;
662 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
663 			LLE_WUNLOCK(la);
664 
665 			(*ifp->if_output)(ifp, m0, &sa, NULL);
666 			return;
667 		}
668 	}
669 reply:
670 	if (op != ARPOP_REQUEST)
671 		goto drop;
672 
673 	if (itaddr.s_addr == myaddr.s_addr) {
674 		/* Shortcut.. the receiving interface is the target. */
675 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
676 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
677 	} else {
678 		struct llentry *lle = NULL;
679 
680 		if (!V_arp_proxyall)
681 			goto drop;
682 
683 		sin.sin_addr = itaddr;
684 		/* XXX MRT use table 0 for arp reply  */
685 		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
686 		if (!rt)
687 			goto drop;
688 
689 		/*
690 		 * Don't send proxies for nodes on the same interface
691 		 * as this one came out of, or we'll get into a fight
692 		 * over who claims what Ether address.
693 		 */
694 		if (!rt->rt_ifp || rt->rt_ifp == ifp) {
695 			RTFREE_LOCKED(rt);
696 			goto drop;
697 		}
698 		IF_AFDATA_LOCK(rt->rt_ifp);
699 		lle = lla_lookup(LLTABLE(rt->rt_ifp), 0, (struct sockaddr *)&sin);
700 		IF_AFDATA_UNLOCK(rt->rt_ifp);
701 		RTFREE_LOCKED(rt);
702 
703 		if (lle != NULL) {
704 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
705 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
706 			LLE_RUNLOCK(lle);
707 		} else
708 			goto drop;
709 
710 		/*
711 		 * Also check that the node which sent the ARP packet
712 		 * is on the the interface we expect it to be on. This
713 		 * avoids ARP chaos if an interface is connected to the
714 		 * wrong network.
715 		 */
716 		sin.sin_addr = isaddr;
717 
718 		/* XXX MRT use table 0 for arp checks */
719 		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
720 		if (!rt)
721 			goto drop;
722 		if (rt->rt_ifp != ifp) {
723 			log(LOG_INFO, "arp_proxy: ignoring request"
724 			    " from %s via %s, expecting %s\n",
725 			    inet_ntoa(isaddr), ifp->if_xname,
726 			    rt->rt_ifp->if_xname);
727 			RTFREE_LOCKED(rt);
728 			goto drop;
729 		}
730 		RTFREE_LOCKED(rt);
731 
732 #ifdef DEBUG_PROXY
733 		printf("arp: proxying for %s\n",
734 		       inet_ntoa(itaddr));
735 #endif
736 	}
737 
738 	if (la != NULL)
739 		LLE_WUNLOCK(la);
740 	if (itaddr.s_addr == myaddr.s_addr &&
741 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
742 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
743 #ifdef DEBUG_LINKLOCAL
744 		printf("arp: sending reply for link-local addr %s\n",
745 		    inet_ntoa(itaddr));
746 #endif
747 		m->m_flags |= M_BCAST;
748 		m->m_flags &= ~M_MCAST;
749 	} else {
750 		/* default behaviour; never reply by broadcast. */
751 		m->m_flags &= ~(M_BCAST|M_MCAST);
752 	}
753 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
754 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
755 	ah->ar_op = htons(ARPOP_REPLY);
756 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
757 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
758 	m->m_pkthdr.len = m->m_len;
759 	sa.sa_family = AF_ARP;
760 	sa.sa_len = 2;
761 	(*ifp->if_output)(ifp, m, &sa, NULL);
762 	return;
763 
764 drop:
765 	if (la != NULL)
766 		LLE_WUNLOCK(la);
767 	m_freem(m);
768 }
769 #endif
770 
771 void
772 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
773 {
774 	struct llentry *lle;
775 
776 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
777 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
778 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
779 		/*
780 		 * interface address is considered static entry
781 		 * because the output of the arp utility shows
782 		 * that L2 entry as permanent
783 		 */
784 		IF_AFDATA_LOCK(ifp);
785 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
786 				 (struct sockaddr *)IA_SIN(ifa));
787 		IF_AFDATA_UNLOCK(ifp);
788 		if (lle == NULL)
789 			log(LOG_INFO, "arp_ifinit: cannot create arp "
790 			    "entry for interface address\n");
791 		else
792 			LLE_RUNLOCK(lle);
793 	}
794 	ifa->ifa_rtrequest = NULL;
795 }
796 
797 void
798 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
799 {
800 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
801 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
802 				&IA_SIN(ifa)->sin_addr, enaddr);
803 	ifa->ifa_rtrequest = NULL;
804 }
805 
806 static int
807 arp_iattach(const void *unused __unused)
808 {
809 	INIT_VNET_INET(curvnet);
810 
811 	V_arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
812 	V_arp_maxtries = 5;
813 	V_useloopback = 1; /* use loopback interface for local traffic */
814 	V_arp_proxyall = 0;
815 
816 	return (0);
817 }
818 
819 static void
820 arp_init(void)
821 {
822 
823 #ifndef VIMAGE_GLOBALS
824 	vnet_mod_register(&vnet_arp_modinfo);
825 #else
826 	arp_iattach(NULL);
827 #endif
828 
829 	netisr_register(&arp_nh);
830 }
831 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
832