xref: /freebsd/sys/netinet/if_ether.c (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
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_var.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 #ifdef INET
69 #include <netinet/ip_carp.h>
70 #endif
71 
72 #include <net/if_arc.h>
73 #include <net/iso88025.h>
74 
75 #include <security/mac/mac_framework.h>
76 
77 #define SIN(s) ((const struct sockaddr_in *)(s))
78 #define SDL(s) ((struct sockaddr_dl *)s)
79 
80 SYSCTL_DECL(_net_link_ether);
81 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
82 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
83 
84 /* timer values */
85 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
86 						 * minutes */
87 static VNET_DEFINE(int, arp_maxtries) = 5;
88 static VNET_DEFINE(int, arp_proxyall) = 0;
89 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
90 						 * 20 seconds */
91 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
92 VNET_PCPUSTAT_SYSINIT(arpstat);
93 
94 #ifdef VIMAGE
95 VNET_PCPUSTAT_SYSUNINIT(arpstat);
96 #endif /* VIMAGE */
97 
98 static VNET_DEFINE(int, arp_maxhold) = 1;
99 
100 #define	V_arpt_keep		VNET(arpt_keep)
101 #define	V_arpt_down		VNET(arpt_down)
102 #define	V_arp_maxtries		VNET(arp_maxtries)
103 #define	V_arp_proxyall		VNET(arp_proxyall)
104 #define	V_arp_maxhold		VNET(arp_maxhold)
105 
106 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW,
107 	&VNET_NAME(arpt_keep), 0,
108 	"ARP entry lifetime in seconds");
109 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW,
110 	&VNET_NAME(arp_maxtries), 0,
111 	"ARP resolution attempts before returning error");
112 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW,
113 	&VNET_NAME(arp_proxyall), 0,
114 	"Enable proxy ARP for all suitable requests");
115 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW,
116 	&VNET_NAME(arpt_down), 0,
117 	"Incomplete ARP entry lifetime in seconds");
118 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
119     arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW,
121 	&VNET_NAME(arp_maxhold), 0,
122 	"Number of packets to hold per ARP entry");
123 
124 static void	arp_init(void);
125 static void	arpintr(struct mbuf *);
126 static void	arptimer(void *);
127 #ifdef INET
128 static void	in_arpinput(struct mbuf *);
129 #endif
130 
131 static const struct netisr_handler arp_nh = {
132 	.nh_name = "arp",
133 	.nh_handler = arpintr,
134 	.nh_proto = NETISR_ARP,
135 	.nh_policy = NETISR_POLICY_SOURCE,
136 };
137 
138 #ifdef AF_INET
139 /*
140  * called by in_scrubprefix() to remove entry from the table when
141  * the interface goes away
142  */
143 void
144 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
145 {
146 	struct sockaddr_in addr4;
147 
148 	bzero((void *)&addr4, sizeof(addr4));
149 	addr4.sin_len    = sizeof(addr4);
150 	addr4.sin_family = AF_INET;
151 	addr4.sin_addr.s_addr = addr;
152 	IF_AFDATA_RLOCK(ifp);
153 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
154 	    (struct sockaddr *)&addr4);
155 	IF_AFDATA_RUNLOCK(ifp);
156 }
157 #endif
158 
159 /*
160  * Timeout routine.  Age arp_tab entries periodically.
161  */
162 static void
163 arptimer(void *arg)
164 {
165 	struct llentry *lle = (struct llentry *)arg;
166 	struct ifnet *ifp;
167 
168 	if (lle->la_flags & LLE_STATIC) {
169 		LLE_WUNLOCK(lle);
170 		return;
171 	}
172 
173 	ifp = lle->lle_tbl->llt_ifp;
174 	CURVNET_SET(ifp->if_vnet);
175 
176 	if ((lle->la_flags & LLE_DELETED) == 0) {
177 		int evt;
178 
179 		if (lle->la_flags & LLE_VALID)
180 			evt = LLENTRY_EXPIRED;
181 		else
182 			evt = LLENTRY_TIMEDOUT;
183 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
184 	}
185 
186 	callout_stop(&lle->la_timer);
187 
188 	/* XXX: LOR avoidance. We still have ref on lle. */
189 	LLE_WUNLOCK(lle);
190 	IF_AFDATA_LOCK(ifp);
191 	LLE_WLOCK(lle);
192 
193 	/* Guard against race with other llentry_free(). */
194 	if (lle->la_flags & LLE_LINKED) {
195 		size_t pkts_dropped;
196 
197 		LLE_REMREF(lle);
198 		pkts_dropped = llentry_free(lle);
199 		ARPSTAT_ADD(dropped, pkts_dropped);
200 	} else
201 		LLE_FREE_LOCKED(lle);
202 
203 	IF_AFDATA_UNLOCK(ifp);
204 
205 	ARPSTAT_INC(timeouts);
206 
207 	CURVNET_RESTORE();
208 }
209 
210 /*
211  * Broadcast an ARP request. Caller specifies:
212  *	- arp header source ip address
213  *	- arp header target ip address
214  *	- arp header source ethernet address
215  */
216 void
217 arprequest(struct ifnet *ifp, const struct in_addr *sip,
218     const struct in_addr *tip, u_char *enaddr)
219 {
220 	struct mbuf *m;
221 	struct arphdr *ah;
222 	struct sockaddr sa;
223 	u_char *carpaddr = NULL;
224 
225 	if (sip == NULL) {
226 		/*
227 		 * The caller did not supply a source address, try to find
228 		 * a compatible one among those assigned to this interface.
229 		 */
230 		struct ifaddr *ifa;
231 
232 		IF_ADDR_RLOCK(ifp);
233 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
234 			if (ifa->ifa_addr->sa_family != AF_INET)
235 				continue;
236 
237 			if (ifa->ifa_carp) {
238 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
239 					continue;
240 				sip = &IA_SIN(ifa)->sin_addr;
241 			} else {
242 				carpaddr = NULL;
243 				sip = &IA_SIN(ifa)->sin_addr;
244 			}
245 
246 			if (0 == ((sip->s_addr ^ tip->s_addr) &
247 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
248 				break;  /* found it. */
249 		}
250 		IF_ADDR_RUNLOCK(ifp);
251 		if (sip == NULL) {
252 			printf("%s: cannot find matching address\n", __func__);
253 			return;
254 		}
255 	}
256 	if (enaddr == NULL)
257 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
258 
259 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
260 		return;
261 	m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
262 		2 * ifp->if_addrlen;
263 	m->m_pkthdr.len = m->m_len;
264 	MH_ALIGN(m, m->m_len);
265 	ah = mtod(m, struct arphdr *);
266 	bzero((caddr_t)ah, m->m_len);
267 #ifdef MAC
268 	mac_netinet_arp_send(ifp, m);
269 #endif
270 	ah->ar_pro = htons(ETHERTYPE_IP);
271 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
272 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
273 	ah->ar_op = htons(ARPOP_REQUEST);
274 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
275 	bcopy(sip, ar_spa(ah), ah->ar_pln);
276 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
277 	sa.sa_family = AF_ARP;
278 	sa.sa_len = 2;
279 	m->m_flags |= M_BCAST;
280 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
281 	(*ifp->if_output)(ifp, m, &sa, NULL);
282 	ARPSTAT_INC(txrequests);
283 }
284 
285 /*
286  * Resolve an IP address into an ethernet address.
287  * On input:
288  *    ifp is the interface we use
289  *    is_gw != if @dst represents gateway to some destination
290  *    m is the mbuf. May be NULL if we don't have a packet.
291  *    dst is the next hop,
292  *    desten is where we want the address.
293  *    flags returns lle entry flags.
294  *
295  * On success, desten and flags are filled in and the function returns 0;
296  * If the packet must be held pending resolution, we return EWOULDBLOCK
297  * On other errors, we return the corresponding error code.
298  * Note that m_freem() handles NULL.
299  */
300 int
301 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
302 	const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
303 {
304 	struct llentry *la = 0;
305 	u_int flags = 0;
306 	struct mbuf *curr = NULL;
307 	struct mbuf *next = NULL;
308 	int error, renew;
309 
310 	if (pflags != NULL)
311 		*pflags = 0;
312 
313 	if (m != NULL) {
314 		if (m->m_flags & M_BCAST) {
315 			/* broadcast */
316 			(void)memcpy(desten,
317 			    ifp->if_broadcastaddr, ifp->if_addrlen);
318 			return (0);
319 		}
320 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
321 			/* multicast */
322 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
323 			return (0);
324 		}
325 	}
326 retry:
327 	IF_AFDATA_RLOCK(ifp);
328 	la = lla_lookup(LLTABLE(ifp), flags, dst);
329 	IF_AFDATA_RUNLOCK(ifp);
330 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
331 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
332 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
333 		IF_AFDATA_WLOCK(ifp);
334 		la = lla_lookup(LLTABLE(ifp), flags, dst);
335 		IF_AFDATA_WUNLOCK(ifp);
336 	}
337 	if (la == NULL) {
338 		if (flags & LLE_CREATE)
339 			log(LOG_DEBUG,
340 			    "arpresolve: can't allocate llinfo for %s on %s\n",
341 			    inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname);
342 		m_freem(m);
343 		return (EINVAL);
344 	}
345 
346 	if ((la->la_flags & LLE_VALID) &&
347 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
348 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
349 		/*
350 		 * If entry has an expiry time and it is approaching,
351 		 * see if we need to send an ARP request within this
352 		 * arpt_down interval.
353 		 */
354 		if (!(la->la_flags & LLE_STATIC) &&
355 		    time_uptime + la->la_preempt > la->la_expire) {
356 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
357 			la->la_preempt--;
358 		}
359 
360 		if (pflags != NULL)
361 			*pflags = la->la_flags;
362 		error = 0;
363 		goto done;
364 	}
365 
366 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
367 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
368 		    inet_ntoa(SIN(dst)->sin_addr));
369 		m_freem(m);
370 		error = EINVAL;
371 		goto done;
372 	}
373 
374 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
375 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
376 		flags |= LLE_EXCLUSIVE;
377 		LLE_RUNLOCK(la);
378 		goto retry;
379 	}
380 	/*
381 	 * There is an arptab entry, but no ethernet address
382 	 * response yet.  Add the mbuf to the list, dropping
383 	 * the oldest packet if we have exceeded the system
384 	 * setting.
385 	 */
386 	if (m != NULL) {
387 		if (la->la_numheld >= V_arp_maxhold) {
388 			if (la->la_hold != NULL) {
389 				next = la->la_hold->m_nextpkt;
390 				m_freem(la->la_hold);
391 				la->la_hold = next;
392 				la->la_numheld--;
393 				ARPSTAT_INC(dropped);
394 			}
395 		}
396 		if (la->la_hold != NULL) {
397 			curr = la->la_hold;
398 			while (curr->m_nextpkt != NULL)
399 				curr = curr->m_nextpkt;
400 			curr->m_nextpkt = m;
401 		} else
402 			la->la_hold = m;
403 		la->la_numheld++;
404 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
405 			flags &= ~LLE_EXCLUSIVE;
406 			LLE_DOWNGRADE(la);
407 		}
408 
409 	}
410 	/*
411 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
412 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
413 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
414 	 * ARP request, but not faster than one request per second.
415 	 */
416 	if (la->la_asked < V_arp_maxtries)
417 		error = EWOULDBLOCK;	/* First request. */
418 	else
419 		error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN;
420 
421 	if (renew) {
422 		int canceled;
423 
424 		LLE_ADDREF(la);
425 		la->la_expire = time_uptime;
426 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
427 		    arptimer, la);
428 		if (canceled)
429 			LLE_REMREF(la);
430 		la->la_asked++;
431 		LLE_WUNLOCK(la);
432 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
433 		return (error);
434 	}
435 done:
436 	if (flags & LLE_EXCLUSIVE)
437 		LLE_WUNLOCK(la);
438 	else
439 		LLE_RUNLOCK(la);
440 	return (error);
441 }
442 
443 /*
444  * Common length and type checks are done here,
445  * then the protocol-specific routine is called.
446  */
447 static void
448 arpintr(struct mbuf *m)
449 {
450 	struct arphdr *ar;
451 
452 	if (m->m_len < sizeof(struct arphdr) &&
453 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
454 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
455 		return;
456 	}
457 	ar = mtod(m, struct arphdr *);
458 
459 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
460 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
461 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
462 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
463 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
464 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
465 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
466 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
467 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
468 		m_freem(m);
469 		return;
470 	}
471 
472 	if (m->m_len < arphdr_len(ar)) {
473 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
474 			log(LOG_NOTICE, "arp: runt packet\n");
475 			m_freem(m);
476 			return;
477 		}
478 		ar = mtod(m, struct arphdr *);
479 	}
480 
481 	ARPSTAT_INC(received);
482 	switch (ntohs(ar->ar_pro)) {
483 #ifdef INET
484 	case ETHERTYPE_IP:
485 		in_arpinput(m);
486 		return;
487 #endif
488 	}
489 	m_freem(m);
490 }
491 
492 #ifdef INET
493 /*
494  * ARP for Internet protocols on 10 Mb/s Ethernet.
495  * Algorithm is that given in RFC 826.
496  * In addition, a sanity check is performed on the sender
497  * protocol address, to catch impersonators.
498  * We no longer handle negotiations for use of trailer protocol:
499  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
500  * along with IP replies if we wanted trailers sent to us,
501  * and also sent them in response to IP replies.
502  * This allowed either end to announce the desire to receive
503  * trailer packets.
504  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
505  * but formerly didn't normally send requests.
506  */
507 static int log_arp_wrong_iface = 1;
508 static int log_arp_movements = 1;
509 static int log_arp_permanent_modify = 1;
510 static int allow_multicast = 0;
511 static struct timeval arp_lastlog;
512 static int arp_curpps;
513 static int arp_maxpps = 1;
514 
515 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
516 	&log_arp_wrong_iface, 0,
517 	"log arp packets arriving on the wrong interface");
518 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
519 	&log_arp_movements, 0,
520 	"log arp replies from MACs different than the one in the cache");
521 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
522 	&log_arp_permanent_modify, 0,
523 	"log arp replies from MACs different than the one in the permanent arp entry");
524 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
525 	&allow_multicast, 0, "accept multicast addresses");
526 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
527 	CTLFLAG_RW, &arp_maxpps, 0,
528 	"Maximum number of remotely triggered ARP messages that can be "
529 	"logged per second");
530 
531 #define	ARP_LOG(pri, ...)	do {					\
532 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
533 		log((pri), "arp: " __VA_ARGS__);			\
534 } while (0)
535 
536 static void
537 in_arpinput(struct mbuf *m)
538 {
539 	struct arphdr *ah;
540 	struct ifnet *ifp = m->m_pkthdr.rcvif;
541 	struct llentry *la = NULL;
542 	struct rtentry *rt;
543 	struct ifaddr *ifa;
544 	struct in_ifaddr *ia;
545 	struct sockaddr sa;
546 	struct in_addr isaddr, itaddr, myaddr;
547 	u_int8_t *enaddr = NULL;
548 	int op, flags;
549 	int req_len;
550 	int bridged = 0, is_bridge = 0;
551 	int carped;
552 	struct sockaddr_in sin;
553 	sin.sin_len = sizeof(struct sockaddr_in);
554 	sin.sin_family = AF_INET;
555 	sin.sin_addr.s_addr = 0;
556 
557 	if (ifp->if_bridge)
558 		bridged = 1;
559 	if (ifp->if_type == IFT_BRIDGE)
560 		is_bridge = 1;
561 
562 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
563 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
564 		ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
565 		return;
566 	}
567 
568 	ah = mtod(m, struct arphdr *);
569 	/*
570 	 * ARP is only for IPv4 so we can reject packets with
571 	 * a protocol length not equal to an IPv4 address.
572 	 */
573 	if (ah->ar_pln != sizeof(struct in_addr)) {
574 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
575 		    sizeof(struct in_addr));
576 		goto drop;
577 	}
578 
579 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
580 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
581 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
582 		goto drop;
583 	}
584 
585 	op = ntohs(ah->ar_op);
586 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
587 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
588 
589 	if (op == ARPOP_REPLY)
590 		ARPSTAT_INC(rxreplies);
591 
592 	/*
593 	 * For a bridge, we want to check the address irrespective
594 	 * of the receive interface. (This will change slightly
595 	 * when we have clusters of interfaces).
596 	 */
597 	IN_IFADDR_RLOCK();
598 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
599 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
600 		    ia->ia_ifp == ifp) &&
601 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
602 		    (ia->ia_ifa.ifa_carp == NULL ||
603 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
604 			ifa_ref(&ia->ia_ifa);
605 			IN_IFADDR_RUNLOCK();
606 			goto match;
607 		}
608 	}
609 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
610 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
611 		    ia->ia_ifp == ifp) &&
612 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
613 			ifa_ref(&ia->ia_ifa);
614 			IN_IFADDR_RUNLOCK();
615 			goto match;
616 		}
617 
618 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
619   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
620   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
621   addr == ia->ia_addr.sin_addr.s_addr)
622 	/*
623 	 * Check the case when bridge shares its MAC address with
624 	 * some of its children, so packets are claimed by bridge
625 	 * itself (bridge_input() does it first), but they are really
626 	 * meant to be destined to the bridge member.
627 	 */
628 	if (is_bridge) {
629 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
630 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
631 				ifa_ref(&ia->ia_ifa);
632 				ifp = ia->ia_ifp;
633 				IN_IFADDR_RUNLOCK();
634 				goto match;
635 			}
636 		}
637 	}
638 #undef BDG_MEMBER_MATCHES_ARP
639 	IN_IFADDR_RUNLOCK();
640 
641 	/*
642 	 * No match, use the first inet address on the receive interface
643 	 * as a dummy address for the rest of the function.
644 	 */
645 	IF_ADDR_RLOCK(ifp);
646 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
647 		if (ifa->ifa_addr->sa_family == AF_INET &&
648 		    (ifa->ifa_carp == NULL ||
649 		    (*carp_iamatch_p)(ifa, &enaddr))) {
650 			ia = ifatoia(ifa);
651 			ifa_ref(ifa);
652 			IF_ADDR_RUNLOCK(ifp);
653 			goto match;
654 		}
655 	IF_ADDR_RUNLOCK(ifp);
656 
657 	/*
658 	 * If bridging, fall back to using any inet address.
659 	 */
660 	IN_IFADDR_RLOCK();
661 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
662 		IN_IFADDR_RUNLOCK();
663 		goto drop;
664 	}
665 	ifa_ref(&ia->ia_ifa);
666 	IN_IFADDR_RUNLOCK();
667 match:
668 	if (!enaddr)
669 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
670 	carped = (ia->ia_ifa.ifa_carp != NULL);
671 	myaddr = ia->ia_addr.sin_addr;
672 	ifa_free(&ia->ia_ifa);
673 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
674 		goto drop;	/* it's from me, ignore it. */
675 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
676 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
677 		    "%s!\n", inet_ntoa(isaddr));
678 		goto drop;
679 	}
680 	/*
681 	 * Warn if another host is using the same IP address, but only if the
682 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
683 	 * case we suppress the warning to avoid false positive complaints of
684 	 * potential misconfiguration.
685 	 */
686 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
687 	    myaddr.s_addr != 0) {
688 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
689 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
690 		   inet_ntoa(isaddr), ifp->if_xname);
691 		itaddr = myaddr;
692 		ARPSTAT_INC(dupips);
693 		goto reply;
694 	}
695 	if (ifp->if_flags & IFF_STATICARP)
696 		goto reply;
697 
698 	bzero(&sin, sizeof(sin));
699 	sin.sin_len = sizeof(struct sockaddr_in);
700 	sin.sin_family = AF_INET;
701 	sin.sin_addr = isaddr;
702 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
703 	flags |= LLE_EXCLUSIVE;
704 	IF_AFDATA_LOCK(ifp);
705 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
706 	IF_AFDATA_UNLOCK(ifp);
707 	if (la != NULL) {
708 		/* the following is not an error when doing bridging */
709 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
710 			if (log_arp_wrong_iface)
711 				ARP_LOG(LOG_WARNING, "%s is on %s "
712 				    "but got reply from %*D on %s\n",
713 				    inet_ntoa(isaddr),
714 				    la->lle_tbl->llt_ifp->if_xname,
715 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
716 				    ifp->if_xname);
717 			LLE_WUNLOCK(la);
718 			goto reply;
719 		}
720 		if ((la->la_flags & LLE_VALID) &&
721 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
722 			if (la->la_flags & LLE_STATIC) {
723 				LLE_WUNLOCK(la);
724 				if (log_arp_permanent_modify)
725 					ARP_LOG(LOG_ERR,
726 					    "%*D attempts to modify "
727 					    "permanent entry for %s on %s\n",
728 					    ifp->if_addrlen,
729 					    (u_char *)ar_sha(ah), ":",
730 					    inet_ntoa(isaddr), ifp->if_xname);
731 				goto reply;
732 			}
733 			if (log_arp_movements) {
734 				ARP_LOG(LOG_INFO, "%s moved from %*D "
735 				    "to %*D on %s\n",
736 				    inet_ntoa(isaddr),
737 				    ifp->if_addrlen,
738 				    (u_char *)&la->ll_addr, ":",
739 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
740 				    ifp->if_xname);
741 			}
742 		}
743 
744 		if (ifp->if_addrlen != ah->ar_hln) {
745 			LLE_WUNLOCK(la);
746 			ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
747 			    "i/f %d (ignored)\n", ifp->if_addrlen,
748 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
749 			    ifp->if_addrlen);
750 			goto drop;
751 		}
752 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
753 		la->la_flags |= LLE_VALID;
754 
755 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
756 
757 		if (!(la->la_flags & LLE_STATIC)) {
758 			int canceled;
759 
760 			LLE_ADDREF(la);
761 			la->la_expire = time_uptime + V_arpt_keep;
762 			canceled = callout_reset(&la->la_timer,
763 			    hz * V_arpt_keep, arptimer, la);
764 			if (canceled)
765 				LLE_REMREF(la);
766 		}
767 		la->la_asked = 0;
768 		la->la_preempt = V_arp_maxtries;
769 		/*
770 		 * The packets are all freed within the call to the output
771 		 * routine.
772 		 *
773 		 * NB: The lock MUST be released before the call to the
774 		 * output routine.
775 		 */
776 		if (la->la_hold != NULL) {
777 			struct mbuf *m_hold, *m_hold_next;
778 
779 			m_hold = la->la_hold;
780 			la->la_hold = NULL;
781 			la->la_numheld = 0;
782 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
783 			LLE_WUNLOCK(la);
784 			for (; m_hold != NULL; m_hold = m_hold_next) {
785 				m_hold_next = m_hold->m_nextpkt;
786 				m_hold->m_nextpkt = NULL;
787 				/* Avoid confusing lower layers. */
788 				m_clrprotoflags(m_hold);
789 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
790 			}
791 		} else
792 			LLE_WUNLOCK(la);
793 	}
794 reply:
795 	if (op != ARPOP_REQUEST)
796 		goto drop;
797 	ARPSTAT_INC(rxrequests);
798 
799 	if (itaddr.s_addr == myaddr.s_addr) {
800 		/* Shortcut.. the receiving interface is the target. */
801 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
802 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
803 	} else {
804 		struct llentry *lle = NULL;
805 
806 		sin.sin_addr = itaddr;
807 		IF_AFDATA_RLOCK(ifp);
808 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
809 		IF_AFDATA_RUNLOCK(ifp);
810 
811 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
812 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
813 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
814 			LLE_RUNLOCK(lle);
815 		} else {
816 
817 			if (lle != NULL)
818 				LLE_RUNLOCK(lle);
819 
820 			if (!V_arp_proxyall)
821 				goto drop;
822 
823 			sin.sin_addr = itaddr;
824 			/* XXX MRT use table 0 for arp reply  */
825 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
826 			if (!rt)
827 				goto drop;
828 
829 			/*
830 			 * Don't send proxies for nodes on the same interface
831 			 * as this one came out of, or we'll get into a fight
832 			 * over who claims what Ether address.
833 			 */
834 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
835 				RTFREE_LOCKED(rt);
836 				goto drop;
837 			}
838 			RTFREE_LOCKED(rt);
839 
840 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
841 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
842 
843 			/*
844 			 * Also check that the node which sent the ARP packet
845 			 * is on the interface we expect it to be on. This
846 			 * avoids ARP chaos if an interface is connected to the
847 			 * wrong network.
848 			 */
849 			sin.sin_addr = isaddr;
850 
851 			/* XXX MRT use table 0 for arp checks */
852 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
853 			if (!rt)
854 				goto drop;
855 			if (rt->rt_ifp != ifp) {
856 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
857 				    " from %s via %s, expecting %s\n",
858 				    inet_ntoa(isaddr), ifp->if_xname,
859 				    rt->rt_ifp->if_xname);
860 				RTFREE_LOCKED(rt);
861 				goto drop;
862 			}
863 			RTFREE_LOCKED(rt);
864 
865 #ifdef DEBUG_PROXY
866 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
867 #endif
868 		}
869 	}
870 
871 	if (itaddr.s_addr == myaddr.s_addr &&
872 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
873 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
874 #ifdef DEBUG_LINKLOCAL
875 		printf("arp: sending reply for link-local addr %s\n",
876 		    inet_ntoa(itaddr));
877 #endif
878 		m->m_flags |= M_BCAST;
879 		m->m_flags &= ~M_MCAST;
880 	} else {
881 		/* default behaviour; never reply by broadcast. */
882 		m->m_flags &= ~(M_BCAST|M_MCAST);
883 	}
884 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
885 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
886 	ah->ar_op = htons(ARPOP_REPLY);
887 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
888 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
889 	m->m_pkthdr.len = m->m_len;
890 	m->m_pkthdr.rcvif = NULL;
891 	sa.sa_family = AF_ARP;
892 	sa.sa_len = 2;
893 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
894 	(*ifp->if_output)(ifp, m, &sa, NULL);
895 	ARPSTAT_INC(txreplies);
896 	return;
897 
898 drop:
899 	m_freem(m);
900 }
901 #endif
902 
903 void
904 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
905 {
906 	struct llentry *lle;
907 
908 	if (ifa->ifa_carp != NULL)
909 		return;
910 
911 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
912 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
913 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
914 		/*
915 		 * interface address is considered static entry
916 		 * because the output of the arp utility shows
917 		 * that L2 entry as permanent
918 		 */
919 		IF_AFDATA_LOCK(ifp);
920 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
921 				 (struct sockaddr *)IA_SIN(ifa));
922 		IF_AFDATA_UNLOCK(ifp);
923 		if (lle == NULL)
924 			log(LOG_INFO, "arp_ifinit: cannot create arp "
925 			    "entry for interface address\n");
926 		else
927 			LLE_RUNLOCK(lle);
928 	}
929 	ifa->ifa_rtrequest = NULL;
930 }
931 
932 void
933 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
934 {
935 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
936 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
937 				&IA_SIN(ifa)->sin_addr, enaddr);
938 	ifa->ifa_rtrequest = NULL;
939 }
940 
941 static void
942 arp_init(void)
943 {
944 
945 	netisr_register(&arp_nh);
946 }
947 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
948