xref: /freebsd/sys/netinet/if_ether.c (revision f4b37ed0f8b307b1f3f0f630ca725d68f1dff30d)
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_WLOCK(ifp);
153 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
154 	    (struct sockaddr *)&addr4);
155 	IF_AFDATA_WUNLOCK(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 		return;
170 	}
171 	LLE_WLOCK(lle);
172 	if (callout_pending(&lle->la_timer)) {
173 		/*
174 		 * Here we are a bit odd here in the treatment of
175 		 * active/pending. If the pending bit is set, it got
176 		 * rescheduled before I ran. The active
177 		 * bit we ignore, since if it was stopped
178 		 * in ll_tablefree() and was currently running
179 		 * it would have return 0 so the code would
180 		 * not have deleted it since the callout could
181 		 * not be stopped so we want to go through
182 		 * with the delete here now. If the callout
183 		 * was restarted, the pending bit will be back on and
184 		 * we just want to bail since the callout_reset would
185 		 * return 1 and our reference would have been removed
186 		 * by arpresolve() below.
187 		 */
188 		LLE_WUNLOCK(lle);
189  		return;
190  	}
191 	ifp = lle->lle_tbl->llt_ifp;
192 	CURVNET_SET(ifp->if_vnet);
193 
194 	if ((lle->la_flags & LLE_DELETED) == 0) {
195 		int evt;
196 
197 		if (lle->la_flags & LLE_VALID)
198 			evt = LLENTRY_EXPIRED;
199 		else
200 			evt = LLENTRY_TIMEDOUT;
201 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
202 	}
203 
204 	callout_stop(&lle->la_timer);
205 
206 	/* XXX: LOR avoidance. We still have ref on lle. */
207 	LLE_WUNLOCK(lle);
208 	IF_AFDATA_LOCK(ifp);
209 	LLE_WLOCK(lle);
210 
211 	/* Guard against race with other llentry_free(). */
212 	if (lle->la_flags & LLE_LINKED) {
213 		size_t pkts_dropped;
214 
215 		LLE_REMREF(lle);
216 		pkts_dropped = llentry_free(lle);
217 		ARPSTAT_ADD(dropped, pkts_dropped);
218 	} else
219 		LLE_FREE_LOCKED(lle);
220 
221 	IF_AFDATA_UNLOCK(ifp);
222 
223 	ARPSTAT_INC(timeouts);
224 
225 	CURVNET_RESTORE();
226 }
227 
228 /*
229  * Broadcast an ARP request. Caller specifies:
230  *	- arp header source ip address
231  *	- arp header target ip address
232  *	- arp header source ethernet address
233  */
234 void
235 arprequest(struct ifnet *ifp, const struct in_addr *sip,
236     const struct in_addr *tip, u_char *enaddr)
237 {
238 	struct mbuf *m;
239 	struct arphdr *ah;
240 	struct sockaddr sa;
241 	u_char *carpaddr = NULL;
242 
243 	if (sip == NULL) {
244 		/*
245 		 * The caller did not supply a source address, try to find
246 		 * a compatible one among those assigned to this interface.
247 		 */
248 		struct ifaddr *ifa;
249 
250 		IF_ADDR_RLOCK(ifp);
251 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
252 			if (ifa->ifa_addr->sa_family != AF_INET)
253 				continue;
254 
255 			if (ifa->ifa_carp) {
256 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
257 					continue;
258 				sip = &IA_SIN(ifa)->sin_addr;
259 			} else {
260 				carpaddr = NULL;
261 				sip = &IA_SIN(ifa)->sin_addr;
262 			}
263 
264 			if (0 == ((sip->s_addr ^ tip->s_addr) &
265 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
266 				break;  /* found it. */
267 		}
268 		IF_ADDR_RUNLOCK(ifp);
269 		if (sip == NULL) {
270 			printf("%s: cannot find matching address\n", __func__);
271 			return;
272 		}
273 	}
274 	if (enaddr == NULL)
275 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
276 
277 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
278 		return;
279 	m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
280 		2 * ifp->if_addrlen;
281 	m->m_pkthdr.len = m->m_len;
282 	M_ALIGN(m, m->m_len);
283 	ah = mtod(m, struct arphdr *);
284 	bzero((caddr_t)ah, m->m_len);
285 #ifdef MAC
286 	mac_netinet_arp_send(ifp, m);
287 #endif
288 	ah->ar_pro = htons(ETHERTYPE_IP);
289 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
290 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
291 	ah->ar_op = htons(ARPOP_REQUEST);
292 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
293 	bcopy(sip, ar_spa(ah), ah->ar_pln);
294 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
295 	sa.sa_family = AF_ARP;
296 	sa.sa_len = 2;
297 	m->m_flags |= M_BCAST;
298 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
299 	(*ifp->if_output)(ifp, m, &sa, NULL);
300 	ARPSTAT_INC(txrequests);
301 }
302 
303 /*
304  * Resolve an IP address into an ethernet address.
305  * On input:
306  *    ifp is the interface we use
307  *    is_gw != if @dst represents gateway to some destination
308  *    m is the mbuf. May be NULL if we don't have a packet.
309  *    dst is the next hop,
310  *    desten is where we want the address.
311  *    flags returns lle entry flags.
312  *
313  * On success, desten and flags are filled in and the function returns 0;
314  * If the packet must be held pending resolution, we return EWOULDBLOCK
315  * On other errors, we return the corresponding error code.
316  * Note that m_freem() handles NULL.
317  */
318 int
319 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
320 	const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
321 {
322 	struct llentry *la = 0;
323 	u_int flags = 0;
324 	struct mbuf *curr = NULL;
325 	struct mbuf *next = NULL;
326 	int error, renew;
327 
328 	if (pflags != NULL)
329 		*pflags = 0;
330 
331 	if (m != NULL) {
332 		if (m->m_flags & M_BCAST) {
333 			/* broadcast */
334 			(void)memcpy(desten,
335 			    ifp->if_broadcastaddr, ifp->if_addrlen);
336 			return (0);
337 		}
338 		if (m->m_flags & M_MCAST) {
339 			/* multicast */
340 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
341 			return (0);
342 		}
343 	}
344 retry:
345 	IF_AFDATA_RLOCK(ifp);
346 	la = lla_lookup(LLTABLE(ifp), flags, dst);
347 	IF_AFDATA_RUNLOCK(ifp);
348 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
349 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
350 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
351 		IF_AFDATA_WLOCK(ifp);
352 		la = lla_lookup(LLTABLE(ifp), flags, dst);
353 		IF_AFDATA_WUNLOCK(ifp);
354 	}
355 	if (la == NULL) {
356 		if (flags & LLE_CREATE)
357 			log(LOG_DEBUG,
358 			    "arpresolve: can't allocate llinfo for %s on %s\n",
359 			    inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname);
360 		m_freem(m);
361 		return (EINVAL);
362 	}
363 
364 	if ((la->la_flags & LLE_VALID) &&
365 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
366 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
367 		renew = 0;
368 		/*
369 		 * If entry has an expiry time and it is approaching,
370 		 * see if we need to send an ARP request within this
371 		 * arpt_down interval.
372 		 */
373 		if (!(la->la_flags & LLE_STATIC) &&
374 		    time_uptime + la->la_preempt > la->la_expire) {
375 			renew = 1;
376 			la->la_preempt--;
377 		}
378 
379 		if (pflags != NULL)
380 			*pflags = la->la_flags;
381 
382 		if (flags & LLE_EXCLUSIVE)
383 			LLE_WUNLOCK(la);
384 		else
385 			LLE_RUNLOCK(la);
386 
387 		if (renew == 1)
388 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
389 
390 		return (0);
391 	}
392 
393 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
394 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
395 		    inet_ntoa(SIN(dst)->sin_addr));
396 		m_freem(m);
397 		error = EINVAL;
398 		goto done;
399 	}
400 
401 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
402 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
403 		flags |= LLE_EXCLUSIVE;
404 		LLE_RUNLOCK(la);
405 		goto retry;
406 	}
407 	/*
408 	 * There is an arptab entry, but no ethernet address
409 	 * response yet.  Add the mbuf to the list, dropping
410 	 * the oldest packet if we have exceeded the system
411 	 * setting.
412 	 */
413 	if (m != NULL) {
414 		if (la->la_numheld >= V_arp_maxhold) {
415 			if (la->la_hold != NULL) {
416 				next = la->la_hold->m_nextpkt;
417 				m_freem(la->la_hold);
418 				la->la_hold = next;
419 				la->la_numheld--;
420 				ARPSTAT_INC(dropped);
421 			}
422 		}
423 		if (la->la_hold != NULL) {
424 			curr = la->la_hold;
425 			while (curr->m_nextpkt != NULL)
426 				curr = curr->m_nextpkt;
427 			curr->m_nextpkt = m;
428 		} else
429 			la->la_hold = m;
430 		la->la_numheld++;
431 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
432 			flags &= ~LLE_EXCLUSIVE;
433 			LLE_DOWNGRADE(la);
434 		}
435 
436 	}
437 	/*
438 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
439 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
440 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
441 	 * ARP request, but not faster than one request per second.
442 	 */
443 	if (la->la_asked < V_arp_maxtries)
444 		error = EWOULDBLOCK;	/* First request. */
445 	else
446 		error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN;
447 
448 	if (renew) {
449 		int canceled;
450 
451 		LLE_ADDREF(la);
452 		la->la_expire = time_uptime;
453 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
454 		    arptimer, la);
455 		if (canceled)
456 			LLE_REMREF(la);
457 		la->la_asked++;
458 		LLE_WUNLOCK(la);
459 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
460 		return (error);
461 	}
462 done:
463 	if (flags & LLE_EXCLUSIVE)
464 		LLE_WUNLOCK(la);
465 	else
466 		LLE_RUNLOCK(la);
467 	return (error);
468 }
469 
470 /*
471  * Common length and type checks are done here,
472  * then the protocol-specific routine is called.
473  */
474 static void
475 arpintr(struct mbuf *m)
476 {
477 	struct arphdr *ar;
478 
479 	if (m->m_len < sizeof(struct arphdr) &&
480 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
481 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
482 		return;
483 	}
484 	ar = mtod(m, struct arphdr *);
485 
486 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
487 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
488 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
489 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
490 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
491 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
492 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
493 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
494 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
495 		m_freem(m);
496 		return;
497 	}
498 
499 	if (m->m_len < arphdr_len(ar)) {
500 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
501 			log(LOG_NOTICE, "arp: runt packet\n");
502 			m_freem(m);
503 			return;
504 		}
505 		ar = mtod(m, struct arphdr *);
506 	}
507 
508 	ARPSTAT_INC(received);
509 	switch (ntohs(ar->ar_pro)) {
510 #ifdef INET
511 	case ETHERTYPE_IP:
512 		in_arpinput(m);
513 		return;
514 #endif
515 	}
516 	m_freem(m);
517 }
518 
519 #ifdef INET
520 /*
521  * ARP for Internet protocols on 10 Mb/s Ethernet.
522  * Algorithm is that given in RFC 826.
523  * In addition, a sanity check is performed on the sender
524  * protocol address, to catch impersonators.
525  * We no longer handle negotiations for use of trailer protocol:
526  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
527  * along with IP replies if we wanted trailers sent to us,
528  * and also sent them in response to IP replies.
529  * This allowed either end to announce the desire to receive
530  * trailer packets.
531  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
532  * but formerly didn't normally send requests.
533  */
534 static int log_arp_wrong_iface = 1;
535 static int log_arp_movements = 1;
536 static int log_arp_permanent_modify = 1;
537 static int allow_multicast = 0;
538 static struct timeval arp_lastlog;
539 static int arp_curpps;
540 static int arp_maxpps = 1;
541 
542 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
543 	&log_arp_wrong_iface, 0,
544 	"log arp packets arriving on the wrong interface");
545 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
546 	&log_arp_movements, 0,
547 	"log arp replies from MACs different than the one in the cache");
548 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
549 	&log_arp_permanent_modify, 0,
550 	"log arp replies from MACs different than the one in the permanent arp entry");
551 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
552 	&allow_multicast, 0, "accept multicast addresses");
553 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
554 	CTLFLAG_RW, &arp_maxpps, 0,
555 	"Maximum number of remotely triggered ARP messages that can be "
556 	"logged per second");
557 
558 #define	ARP_LOG(pri, ...)	do {					\
559 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
560 		log((pri), "arp: " __VA_ARGS__);			\
561 } while (0)
562 
563 static void
564 in_arpinput(struct mbuf *m)
565 {
566 	struct arphdr *ah;
567 	struct ifnet *ifp = m->m_pkthdr.rcvif;
568 	struct llentry *la = NULL;
569 	struct rtentry *rt;
570 	struct ifaddr *ifa;
571 	struct in_ifaddr *ia;
572 	struct sockaddr sa;
573 	struct in_addr isaddr, itaddr, myaddr;
574 	u_int8_t *enaddr = NULL;
575 	int op, flags;
576 	int req_len;
577 	int bridged = 0, is_bridge = 0;
578 	int carped;
579 	struct sockaddr_in sin;
580 	sin.sin_len = sizeof(struct sockaddr_in);
581 	sin.sin_family = AF_INET;
582 	sin.sin_addr.s_addr = 0;
583 
584 	if (ifp->if_bridge)
585 		bridged = 1;
586 	if (ifp->if_type == IFT_BRIDGE)
587 		is_bridge = 1;
588 
589 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
590 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
591 		ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
592 		return;
593 	}
594 
595 	ah = mtod(m, struct arphdr *);
596 	/*
597 	 * ARP is only for IPv4 so we can reject packets with
598 	 * a protocol length not equal to an IPv4 address.
599 	 */
600 	if (ah->ar_pln != sizeof(struct in_addr)) {
601 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
602 		    sizeof(struct in_addr));
603 		goto drop;
604 	}
605 
606 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
607 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
608 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
609 		goto drop;
610 	}
611 
612 	op = ntohs(ah->ar_op);
613 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
614 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
615 
616 	if (op == ARPOP_REPLY)
617 		ARPSTAT_INC(rxreplies);
618 
619 	/*
620 	 * For a bridge, we want to check the address irrespective
621 	 * of the receive interface. (This will change slightly
622 	 * when we have clusters of interfaces).
623 	 */
624 	IN_IFADDR_RLOCK();
625 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
626 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
627 		    ia->ia_ifp == ifp) &&
628 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
629 		    (ia->ia_ifa.ifa_carp == NULL ||
630 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
631 			ifa_ref(&ia->ia_ifa);
632 			IN_IFADDR_RUNLOCK();
633 			goto match;
634 		}
635 	}
636 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
637 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
638 		    ia->ia_ifp == ifp) &&
639 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
640 			ifa_ref(&ia->ia_ifa);
641 			IN_IFADDR_RUNLOCK();
642 			goto match;
643 		}
644 
645 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
646   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
647   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
648   addr == ia->ia_addr.sin_addr.s_addr)
649 	/*
650 	 * Check the case when bridge shares its MAC address with
651 	 * some of its children, so packets are claimed by bridge
652 	 * itself (bridge_input() does it first), but they are really
653 	 * meant to be destined to the bridge member.
654 	 */
655 	if (is_bridge) {
656 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
657 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
658 				ifa_ref(&ia->ia_ifa);
659 				ifp = ia->ia_ifp;
660 				IN_IFADDR_RUNLOCK();
661 				goto match;
662 			}
663 		}
664 	}
665 #undef BDG_MEMBER_MATCHES_ARP
666 	IN_IFADDR_RUNLOCK();
667 
668 	/*
669 	 * No match, use the first inet address on the receive interface
670 	 * as a dummy address for the rest of the function.
671 	 */
672 	IF_ADDR_RLOCK(ifp);
673 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
674 		if (ifa->ifa_addr->sa_family == AF_INET &&
675 		    (ifa->ifa_carp == NULL ||
676 		    (*carp_iamatch_p)(ifa, &enaddr))) {
677 			ia = ifatoia(ifa);
678 			ifa_ref(ifa);
679 			IF_ADDR_RUNLOCK(ifp);
680 			goto match;
681 		}
682 	IF_ADDR_RUNLOCK(ifp);
683 
684 	/*
685 	 * If bridging, fall back to using any inet address.
686 	 */
687 	IN_IFADDR_RLOCK();
688 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
689 		IN_IFADDR_RUNLOCK();
690 		goto drop;
691 	}
692 	ifa_ref(&ia->ia_ifa);
693 	IN_IFADDR_RUNLOCK();
694 match:
695 	if (!enaddr)
696 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
697 	carped = (ia->ia_ifa.ifa_carp != NULL);
698 	myaddr = ia->ia_addr.sin_addr;
699 	ifa_free(&ia->ia_ifa);
700 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
701 		goto drop;	/* it's from me, ignore it. */
702 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
703 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
704 		    "%s!\n", inet_ntoa(isaddr));
705 		goto drop;
706 	}
707 	/*
708 	 * Warn if another host is using the same IP address, but only if the
709 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
710 	 * case we suppress the warning to avoid false positive complaints of
711 	 * potential misconfiguration.
712 	 */
713 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
714 	    myaddr.s_addr != 0) {
715 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
716 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
717 		   inet_ntoa(isaddr), ifp->if_xname);
718 		itaddr = myaddr;
719 		ARPSTAT_INC(dupips);
720 		goto reply;
721 	}
722 	if (ifp->if_flags & IFF_STATICARP)
723 		goto reply;
724 
725 	bzero(&sin, sizeof(sin));
726 	sin.sin_len = sizeof(struct sockaddr_in);
727 	sin.sin_family = AF_INET;
728 	sin.sin_addr = isaddr;
729 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
730 	flags |= LLE_EXCLUSIVE;
731 	IF_AFDATA_LOCK(ifp);
732 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
733 	IF_AFDATA_UNLOCK(ifp);
734 	if (la != NULL) {
735 		/* the following is not an error when doing bridging */
736 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
737 			if (log_arp_wrong_iface)
738 				ARP_LOG(LOG_WARNING, "%s is on %s "
739 				    "but got reply from %*D on %s\n",
740 				    inet_ntoa(isaddr),
741 				    la->lle_tbl->llt_ifp->if_xname,
742 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
743 				    ifp->if_xname);
744 			LLE_WUNLOCK(la);
745 			goto reply;
746 		}
747 		if ((la->la_flags & LLE_VALID) &&
748 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
749 			if (la->la_flags & LLE_STATIC) {
750 				LLE_WUNLOCK(la);
751 				if (log_arp_permanent_modify)
752 					ARP_LOG(LOG_ERR,
753 					    "%*D attempts to modify "
754 					    "permanent entry for %s on %s\n",
755 					    ifp->if_addrlen,
756 					    (u_char *)ar_sha(ah), ":",
757 					    inet_ntoa(isaddr), ifp->if_xname);
758 				goto reply;
759 			}
760 			if (log_arp_movements) {
761 				ARP_LOG(LOG_INFO, "%s moved from %*D "
762 				    "to %*D on %s\n",
763 				    inet_ntoa(isaddr),
764 				    ifp->if_addrlen,
765 				    (u_char *)&la->ll_addr, ":",
766 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
767 				    ifp->if_xname);
768 			}
769 		}
770 
771 		if (ifp->if_addrlen != ah->ar_hln) {
772 			LLE_WUNLOCK(la);
773 			ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
774 			    "i/f %d (ignored)\n", ifp->if_addrlen,
775 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
776 			    ifp->if_addrlen);
777 			goto drop;
778 		}
779 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
780 		la->la_flags |= LLE_VALID;
781 
782 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
783 
784 		if (!(la->la_flags & LLE_STATIC)) {
785 			int canceled;
786 
787 			LLE_ADDREF(la);
788 			la->la_expire = time_uptime + V_arpt_keep;
789 			canceled = callout_reset(&la->la_timer,
790 			    hz * V_arpt_keep, arptimer, la);
791 			if (canceled)
792 				LLE_REMREF(la);
793 		}
794 		la->la_asked = 0;
795 		la->la_preempt = V_arp_maxtries;
796 		/*
797 		 * The packets are all freed within the call to the output
798 		 * routine.
799 		 *
800 		 * NB: The lock MUST be released before the call to the
801 		 * output routine.
802 		 */
803 		if (la->la_hold != NULL) {
804 			struct mbuf *m_hold, *m_hold_next;
805 
806 			m_hold = la->la_hold;
807 			la->la_hold = NULL;
808 			la->la_numheld = 0;
809 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
810 			LLE_WUNLOCK(la);
811 			for (; m_hold != NULL; m_hold = m_hold_next) {
812 				m_hold_next = m_hold->m_nextpkt;
813 				m_hold->m_nextpkt = NULL;
814 				/* Avoid confusing lower layers. */
815 				m_clrprotoflags(m_hold);
816 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
817 			}
818 		} else
819 			LLE_WUNLOCK(la);
820 	}
821 reply:
822 	if (op != ARPOP_REQUEST)
823 		goto drop;
824 	ARPSTAT_INC(rxrequests);
825 
826 	if (itaddr.s_addr == myaddr.s_addr) {
827 		/* Shortcut.. the receiving interface is the target. */
828 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
829 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
830 	} else {
831 		struct llentry *lle = NULL;
832 
833 		sin.sin_addr = itaddr;
834 		IF_AFDATA_RLOCK(ifp);
835 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
836 		IF_AFDATA_RUNLOCK(ifp);
837 
838 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
839 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
840 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
841 			LLE_RUNLOCK(lle);
842 		} else {
843 
844 			if (lle != NULL)
845 				LLE_RUNLOCK(lle);
846 
847 			if (!V_arp_proxyall)
848 				goto drop;
849 
850 			sin.sin_addr = itaddr;
851 			/* XXX MRT use table 0 for arp reply  */
852 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
853 			if (!rt)
854 				goto drop;
855 
856 			/*
857 			 * Don't send proxies for nodes on the same interface
858 			 * as this one came out of, or we'll get into a fight
859 			 * over who claims what Ether address.
860 			 */
861 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
862 				RTFREE_LOCKED(rt);
863 				goto drop;
864 			}
865 			RTFREE_LOCKED(rt);
866 
867 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
868 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
869 
870 			/*
871 			 * Also check that the node which sent the ARP packet
872 			 * is on the interface we expect it to be on. This
873 			 * avoids ARP chaos if an interface is connected to the
874 			 * wrong network.
875 			 */
876 			sin.sin_addr = isaddr;
877 
878 			/* XXX MRT use table 0 for arp checks */
879 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
880 			if (!rt)
881 				goto drop;
882 			if (rt->rt_ifp != ifp) {
883 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
884 				    " from %s via %s, expecting %s\n",
885 				    inet_ntoa(isaddr), ifp->if_xname,
886 				    rt->rt_ifp->if_xname);
887 				RTFREE_LOCKED(rt);
888 				goto drop;
889 			}
890 			RTFREE_LOCKED(rt);
891 
892 #ifdef DEBUG_PROXY
893 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
894 #endif
895 		}
896 	}
897 
898 	if (itaddr.s_addr == myaddr.s_addr &&
899 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
900 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
901 #ifdef DEBUG_LINKLOCAL
902 		printf("arp: sending reply for link-local addr %s\n",
903 		    inet_ntoa(itaddr));
904 #endif
905 		m->m_flags |= M_BCAST;
906 		m->m_flags &= ~M_MCAST;
907 	} else {
908 		/* default behaviour; never reply by broadcast. */
909 		m->m_flags &= ~(M_BCAST|M_MCAST);
910 	}
911 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
912 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
913 	ah->ar_op = htons(ARPOP_REPLY);
914 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
915 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
916 	m->m_pkthdr.len = m->m_len;
917 	m->m_pkthdr.rcvif = NULL;
918 	sa.sa_family = AF_ARP;
919 	sa.sa_len = 2;
920 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
921 	(*ifp->if_output)(ifp, m, &sa, NULL);
922 	ARPSTAT_INC(txreplies);
923 	return;
924 
925 drop:
926 	m_freem(m);
927 }
928 #endif
929 
930 void
931 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
932 {
933 	struct llentry *lle;
934 
935 	if (ifa->ifa_carp != NULL)
936 		return;
937 
938 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
939 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
940 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
941 		/*
942 		 * interface address is considered static entry
943 		 * because the output of the arp utility shows
944 		 * that L2 entry as permanent
945 		 */
946 		IF_AFDATA_LOCK(ifp);
947 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
948 				 (struct sockaddr *)IA_SIN(ifa));
949 		IF_AFDATA_UNLOCK(ifp);
950 		if (lle == NULL)
951 			log(LOG_INFO, "arp_ifinit: cannot create arp "
952 			    "entry for interface address\n");
953 		else
954 			LLE_RUNLOCK(lle);
955 	}
956 	ifa->ifa_rtrequest = NULL;
957 }
958 
959 void
960 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
961 {
962 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
963 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
964 				&IA_SIN(ifa)->sin_addr, enaddr);
965 	ifa->ifa_rtrequest = NULL;
966 }
967 
968 static void
969 arp_init(void)
970 {
971 
972 	netisr_register(&arp_nh);
973 }
974 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
975