xref: /freebsd/sys/netinet/if_ether.c (revision ce3adf4362fcca6a43e500b2531f0038adbfbd21)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30  */
31 
32 /*
33  * Ethernet address resolution protocol.
34  * TODO:
35  *	add "inuse/lock" bit (or ref. count) along with valid bit
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/queue.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/syslog.h>
53 
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/netisr.h>
58 #include <net/if_llc.h>
59 #include <net/ethernet.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_var.h>
65 #include <net/if_llatbl.h>
66 #include <netinet/if_ether.h>
67 #ifdef INET
68 #include <netinet/ip_carp.h>
69 #endif
70 
71 #include <net/if_arc.h>
72 #include <net/iso88025.h>
73 
74 #include <security/mac/mac_framework.h>
75 
76 #define SIN(s) ((const struct sockaddr_in *)(s))
77 #define SDL(s) ((struct sockaddr_dl *)s)
78 
79 SYSCTL_DECL(_net_link_ether);
80 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
82 
83 /* timer values */
84 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
85 						 * minutes */
86 static VNET_DEFINE(int, arp_maxtries) = 5;
87 VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
88 					 * local traffic */
89 static VNET_DEFINE(int, arp_proxyall) = 0;
90 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
91 						 * 20 seconds */
92 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
93 VNET_PCPUSTAT_SYSINIT(arpstat);
94 
95 #ifdef VIMAGE
96 VNET_PCPUSTAT_SYSUNINIT(arpstat);
97 #endif /* VIMAGE */
98 
99 static VNET_DEFINE(int, arp_maxhold) = 1;
100 
101 #define	V_arpt_keep		VNET(arpt_keep)
102 #define	V_arpt_down		VNET(arpt_down)
103 #define	V_arp_maxtries		VNET(arp_maxtries)
104 #define	V_arp_proxyall		VNET(arp_proxyall)
105 #define	V_arp_maxhold		VNET(arp_maxhold)
106 
107 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
108 	&VNET_NAME(arpt_keep), 0,
109 	"ARP entry lifetime in seconds");
110 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
111 	&VNET_NAME(arp_maxtries), 0,
112 	"ARP resolution attempts before returning error");
113 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
114 	&VNET_NAME(useloopback), 0,
115 	"Use the loopback interface for local traffic");
116 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
117 	&VNET_NAME(arp_proxyall), 0,
118 	"Enable proxy ARP for all suitable requests");
119 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW,
120 	&VNET_NAME(arpt_down), 0,
121 	"Incomplete ARP entry lifetime in seconds");
122 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
123     arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
124 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW,
125 	&VNET_NAME(arp_maxhold), 0,
126 	"Number of packets to hold per ARP entry");
127 
128 static void	arp_init(void);
129 static void	arpintr(struct mbuf *);
130 static void	arptimer(void *);
131 #ifdef INET
132 static void	in_arpinput(struct mbuf *);
133 #endif
134 
135 static const struct netisr_handler arp_nh = {
136 	.nh_name = "arp",
137 	.nh_handler = arpintr,
138 	.nh_proto = NETISR_ARP,
139 	.nh_policy = NETISR_POLICY_SOURCE,
140 };
141 
142 #ifdef AF_INET
143 /*
144  * called by in_ifscrub to remove entry from the table when
145  * the interface goes away
146  */
147 void
148 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
149 {
150 	struct sockaddr_in addr4;
151 
152 	bzero((void *)&addr4, sizeof(addr4));
153 	addr4.sin_len    = sizeof(addr4);
154 	addr4.sin_family = AF_INET;
155 	addr4.sin_addr.s_addr = addr;
156 	IF_AFDATA_LOCK(ifp);
157 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
158 	    (struct sockaddr *)&addr4);
159 	IF_AFDATA_UNLOCK(ifp);
160 }
161 #endif
162 
163 /*
164  * Timeout routine.  Age arp_tab entries periodically.
165  */
166 static void
167 arptimer(void *arg)
168 {
169 	struct llentry *lle = (struct llentry *)arg;
170 	struct ifnet *ifp;
171 
172 	if (lle->la_flags & LLE_STATIC) {
173 		LLE_WUNLOCK(lle);
174 		return;
175 	}
176 
177 	ifp = lle->lle_tbl->llt_ifp;
178 	CURVNET_SET(ifp->if_vnet);
179 
180 	if ((lle->la_flags & LLE_DELETED) == 0) {
181 		int evt;
182 
183 		if (lle->la_flags & LLE_VALID)
184 			evt = LLENTRY_EXPIRED;
185 		else
186 			evt = LLENTRY_TIMEDOUT;
187 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
188 	}
189 
190 	callout_stop(&lle->la_timer);
191 
192 	/* XXX: LOR avoidance. We still have ref on lle. */
193 	LLE_WUNLOCK(lle);
194 	IF_AFDATA_LOCK(ifp);
195 	LLE_WLOCK(lle);
196 
197 	/* Guard against race with other llentry_free(). */
198 	if (lle->la_flags & LLE_LINKED) {
199 		size_t pkts_dropped;
200 
201 		LLE_REMREF(lle);
202 		pkts_dropped = llentry_free(lle);
203 		ARPSTAT_ADD(dropped, pkts_dropped);
204 	} else
205 		LLE_FREE_LOCKED(lle);
206 
207 	IF_AFDATA_UNLOCK(ifp);
208 
209 	ARPSTAT_INC(timeouts);
210 
211 	CURVNET_RESTORE();
212 }
213 
214 /*
215  * Broadcast an ARP request. Caller specifies:
216  *	- arp header source ip address
217  *	- arp header target ip address
218  *	- arp header source ethernet address
219  */
220 void
221 arprequest(struct ifnet *ifp, const struct in_addr *sip,
222     const struct in_addr *tip, u_char *enaddr)
223 {
224 	struct mbuf *m;
225 	struct arphdr *ah;
226 	struct sockaddr sa;
227 	u_char *carpaddr = NULL;
228 
229 	if (sip == NULL) {
230 		/*
231 		 * The caller did not supply a source address, try to find
232 		 * a compatible one among those assigned to this interface.
233 		 */
234 		struct ifaddr *ifa;
235 
236 		IF_ADDR_RLOCK(ifp);
237 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
238 			if (ifa->ifa_addr->sa_family != AF_INET)
239 				continue;
240 
241 			if (ifa->ifa_carp) {
242 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
243 					continue;
244 				sip = &IA_SIN(ifa)->sin_addr;
245 			} else {
246 				carpaddr = NULL;
247 				sip = &IA_SIN(ifa)->sin_addr;
248 			}
249 
250 			if (0 == ((sip->s_addr ^ tip->s_addr) &
251 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
252 				break;  /* found it. */
253 		}
254 		IF_ADDR_RUNLOCK(ifp);
255 		if (sip == NULL) {
256 			printf("%s: cannot find matching address\n", __func__);
257 			return;
258 		}
259 	}
260 	if (enaddr == NULL)
261 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
262 
263 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
264 		return;
265 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
266 		2*ifp->if_data.ifi_addrlen;
267 	m->m_pkthdr.len = m->m_len;
268 	MH_ALIGN(m, m->m_len);
269 	ah = mtod(m, struct arphdr *);
270 	bzero((caddr_t)ah, m->m_len);
271 #ifdef MAC
272 	mac_netinet_arp_send(ifp, m);
273 #endif
274 	ah->ar_pro = htons(ETHERTYPE_IP);
275 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
276 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
277 	ah->ar_op = htons(ARPOP_REQUEST);
278 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
279 	bcopy(sip, ar_spa(ah), ah->ar_pln);
280 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
281 	sa.sa_family = AF_ARP;
282 	sa.sa_len = 2;
283 	m->m_flags |= M_BCAST;
284 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
285 	(*ifp->if_output)(ifp, m, &sa, NULL);
286 	ARPSTAT_INC(txrequests);
287 }
288 
289 /*
290  * Resolve an IP address into an ethernet address.
291  * On input:
292  *    ifp is the interface we use
293  *    rt0 is the route to the final destination (possibly useless)
294  *    m is the mbuf. May be NULL if we don't have a packet.
295  *    dst is the next hop,
296  *    desten is where we want the address.
297  *
298  * On success, desten is filled in and the function returns 0;
299  * If the packet must be held pending resolution, we return EWOULDBLOCK
300  * On other errors, we return the corresponding error code.
301  * Note that m_freem() handles NULL.
302  */
303 int
304 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
305 	const struct sockaddr *dst, u_char *desten, struct llentry **lle)
306 {
307 	struct llentry *la = 0;
308 	u_int flags = 0;
309 	struct mbuf *curr = NULL;
310 	struct mbuf *next = NULL;
311 	int error, renew;
312 
313 	*lle = NULL;
314 	if (m != NULL) {
315 		if (m->m_flags & M_BCAST) {
316 			/* broadcast */
317 			(void)memcpy(desten,
318 			    ifp->if_broadcastaddr, ifp->if_addrlen);
319 			return (0);
320 		}
321 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
322 			/* multicast */
323 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
324 			return (0);
325 		}
326 	}
327 retry:
328 	IF_AFDATA_RLOCK(ifp);
329 	la = lla_lookup(LLTABLE(ifp), flags, dst);
330 	IF_AFDATA_RUNLOCK(ifp);
331 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
332 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
333 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
334 		IF_AFDATA_WLOCK(ifp);
335 		la = lla_lookup(LLTABLE(ifp), flags, dst);
336 		IF_AFDATA_WUNLOCK(ifp);
337 	}
338 	if (la == NULL) {
339 		if (flags & LLE_CREATE)
340 			log(LOG_DEBUG,
341 			    "arpresolve: can't allocate llinfo for %s\n",
342 			    inet_ntoa(SIN(dst)->sin_addr));
343 		m_freem(m);
344 		return (EINVAL);
345 	}
346 
347 	if ((la->la_flags & LLE_VALID) &&
348 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
349 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
350 		/*
351 		 * If entry has an expiry time and it is approaching,
352 		 * see if we need to send an ARP request within this
353 		 * arpt_down interval.
354 		 */
355 		if (!(la->la_flags & LLE_STATIC) &&
356 		    time_uptime + la->la_preempt > la->la_expire) {
357 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
358 			la->la_preempt--;
359 		}
360 
361 		*lle = la;
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 = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
420 		    EHOSTUNREACH : EHOSTDOWN;
421 
422 	if (renew) {
423 		int canceled;
424 
425 		LLE_ADDREF(la);
426 		la->la_expire = time_uptime;
427 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
428 		    arptimer, la);
429 		if (canceled)
430 			LLE_REMREF(la);
431 		la->la_asked++;
432 		LLE_WUNLOCK(la);
433 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
434 		return (error);
435 	}
436 done:
437 	if (flags & LLE_EXCLUSIVE)
438 		LLE_WUNLOCK(la);
439 	else
440 		LLE_RUNLOCK(la);
441 	return (error);
442 }
443 
444 /*
445  * Common length and type checks are done here,
446  * then the protocol-specific routine is called.
447  */
448 static void
449 arpintr(struct mbuf *m)
450 {
451 	struct arphdr *ar;
452 
453 	if (m->m_len < sizeof(struct arphdr) &&
454 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
455 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
456 		return;
457 	}
458 	ar = mtod(m, struct arphdr *);
459 
460 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
461 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
462 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
463 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
464 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
465 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
466 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
467 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
468 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
469 		m_freem(m);
470 		return;
471 	}
472 
473 	if (m->m_len < arphdr_len(ar)) {
474 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
475 			log(LOG_NOTICE, "arp: runt packet\n");
476 			m_freem(m);
477 			return;
478 		}
479 		ar = mtod(m, struct arphdr *);
480 	}
481 
482 	ARPSTAT_INC(received);
483 	switch (ntohs(ar->ar_pro)) {
484 #ifdef INET
485 	case ETHERTYPE_IP:
486 		in_arpinput(m);
487 		return;
488 #endif
489 	}
490 	m_freem(m);
491 }
492 
493 #ifdef INET
494 /*
495  * ARP for Internet protocols on 10 Mb/s Ethernet.
496  * Algorithm is that given in RFC 826.
497  * In addition, a sanity check is performed on the sender
498  * protocol address, to catch impersonators.
499  * We no longer handle negotiations for use of trailer protocol:
500  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
501  * along with IP replies if we wanted trailers sent to us,
502  * and also sent them in response to IP replies.
503  * This allowed either end to announce the desire to receive
504  * trailer packets.
505  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
506  * but formerly didn't normally send requests.
507  */
508 static int log_arp_wrong_iface = 1;
509 static int log_arp_movements = 1;
510 static int log_arp_permanent_modify = 1;
511 static int allow_multicast = 0;
512 static struct timeval arp_lastlog;
513 static int arp_curpps;
514 static int arp_maxpps = 1;
515 
516 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
517 	&log_arp_wrong_iface, 0,
518 	"log arp packets arriving on the wrong interface");
519 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
520 	&log_arp_movements, 0,
521 	"log arp replies from MACs different than the one in the cache");
522 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
523 	&log_arp_permanent_modify, 0,
524 	"log arp replies from MACs different than the one in the permanent arp entry");
525 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
526 	&allow_multicast, 0, "accept multicast addresses");
527 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
528 	CTLFLAG_RW, &arp_maxpps, 0,
529 	"Maximum number of remotely triggered ARP messages that can be "
530 	"logged per second");
531 
532 #define	ARP_LOG(pri, ...)	do {					\
533 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
534 		log((pri), "arp: " __VA_ARGS__);			\
535 } while (0)
536 
537 static void
538 in_arpinput(struct mbuf *m)
539 {
540 	struct arphdr *ah;
541 	struct ifnet *ifp = m->m_pkthdr.rcvif;
542 	struct llentry *la = NULL;
543 	struct rtentry *rt;
544 	struct ifaddr *ifa;
545 	struct in_ifaddr *ia;
546 	struct sockaddr sa;
547 	struct in_addr isaddr, itaddr, myaddr;
548 	u_int8_t *enaddr = NULL;
549 	int op, flags;
550 	int req_len;
551 	int bridged = 0, is_bridge = 0;
552 	int carped;
553 	struct sockaddr_in sin;
554 	sin.sin_len = sizeof(struct sockaddr_in);
555 	sin.sin_family = AF_INET;
556 	sin.sin_addr.s_addr = 0;
557 
558 	if (ifp->if_bridge)
559 		bridged = 1;
560 	if (ifp->if_type == IFT_BRIDGE)
561 		is_bridge = 1;
562 
563 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
564 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
565 		ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
566 		return;
567 	}
568 
569 	ah = mtod(m, struct arphdr *);
570 	/*
571 	 * ARP is only for IPv4 so we can reject packets with
572 	 * a protocol length not equal to an IPv4 address.
573 	 */
574 	if (ah->ar_pln != sizeof(struct in_addr)) {
575 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
576 		    sizeof(struct in_addr));
577 		goto drop;
578 	}
579 
580 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
581 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
582 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
583 		goto drop;
584 	}
585 
586 	op = ntohs(ah->ar_op);
587 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
588 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
589 
590 	if (op == ARPOP_REPLY)
591 		ARPSTAT_INC(rxreplies);
592 
593 	/*
594 	 * For a bridge, we want to check the address irrespective
595 	 * of the receive interface. (This will change slightly
596 	 * when we have clusters of interfaces).
597 	 */
598 	IN_IFADDR_RLOCK();
599 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
600 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
601 		    ia->ia_ifp == ifp) &&
602 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
603 		    (ia->ia_ifa.ifa_carp == NULL ||
604 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
605 			ifa_ref(&ia->ia_ifa);
606 			IN_IFADDR_RUNLOCK();
607 			goto match;
608 		}
609 	}
610 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
611 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
612 		    ia->ia_ifp == ifp) &&
613 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
614 			ifa_ref(&ia->ia_ifa);
615 			IN_IFADDR_RUNLOCK();
616 			goto match;
617 		}
618 
619 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
620   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
621   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
622   addr == ia->ia_addr.sin_addr.s_addr)
623 	/*
624 	 * Check the case when bridge shares its MAC address with
625 	 * some of its children, so packets are claimed by bridge
626 	 * itself (bridge_input() does it first), but they are really
627 	 * meant to be destined to the bridge member.
628 	 */
629 	if (is_bridge) {
630 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
631 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
632 				ifa_ref(&ia->ia_ifa);
633 				ifp = ia->ia_ifp;
634 				IN_IFADDR_RUNLOCK();
635 				goto match;
636 			}
637 		}
638 	}
639 #undef BDG_MEMBER_MATCHES_ARP
640 	IN_IFADDR_RUNLOCK();
641 
642 	/*
643 	 * No match, use the first inet address on the receive interface
644 	 * as a dummy address for the rest of the function.
645 	 */
646 	IF_ADDR_RLOCK(ifp);
647 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
648 		if (ifa->ifa_addr->sa_family == AF_INET &&
649 		    (ifa->ifa_carp == NULL ||
650 		    (*carp_iamatch_p)(ifa, &enaddr))) {
651 			ia = ifatoia(ifa);
652 			ifa_ref(ifa);
653 			IF_ADDR_RUNLOCK(ifp);
654 			goto match;
655 		}
656 	IF_ADDR_RUNLOCK(ifp);
657 
658 	/*
659 	 * If bridging, fall back to using any inet address.
660 	 */
661 	IN_IFADDR_RLOCK();
662 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
663 		IN_IFADDR_RUNLOCK();
664 		goto drop;
665 	}
666 	ifa_ref(&ia->ia_ifa);
667 	IN_IFADDR_RUNLOCK();
668 match:
669 	if (!enaddr)
670 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
671 	carped = (ia->ia_ifa.ifa_carp != NULL);
672 	myaddr = ia->ia_addr.sin_addr;
673 	ifa_free(&ia->ia_ifa);
674 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
675 		goto drop;	/* it's from me, ignore it. */
676 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
677 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
678 		    "%s!\n", inet_ntoa(isaddr));
679 		goto drop;
680 	}
681 	/*
682 	 * Warn if another host is using the same IP address, but only if the
683 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
684 	 * case we suppress the warning to avoid false positive complaints of
685 	 * potential misconfiguration.
686 	 */
687 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
688 	    myaddr.s_addr != 0) {
689 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
690 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
691 		   inet_ntoa(isaddr), ifp->if_xname);
692 		itaddr = myaddr;
693 		ARPSTAT_INC(dupips);
694 		goto reply;
695 	}
696 	if (ifp->if_flags & IFF_STATICARP)
697 		goto reply;
698 
699 	bzero(&sin, sizeof(sin));
700 	sin.sin_len = sizeof(struct sockaddr_in);
701 	sin.sin_family = AF_INET;
702 	sin.sin_addr = isaddr;
703 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
704 	flags |= LLE_EXCLUSIVE;
705 	IF_AFDATA_LOCK(ifp);
706 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
707 	IF_AFDATA_UNLOCK(ifp);
708 	if (la != NULL) {
709 		/* the following is not an error when doing bridging */
710 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
711 			if (log_arp_wrong_iface)
712 				ARP_LOG(LOG_WARNING, "%s is on %s "
713 				    "but got reply from %*D on %s\n",
714 				    inet_ntoa(isaddr),
715 				    la->lle_tbl->llt_ifp->if_xname,
716 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
717 				    ifp->if_xname);
718 			LLE_WUNLOCK(la);
719 			goto reply;
720 		}
721 		if ((la->la_flags & LLE_VALID) &&
722 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
723 			if (la->la_flags & LLE_STATIC) {
724 				LLE_WUNLOCK(la);
725 				if (log_arp_permanent_modify)
726 					ARP_LOG(LOG_ERR,
727 					    "%*D attempts to modify "
728 					    "permanent entry for %s on %s\n",
729 					    ifp->if_addrlen,
730 					    (u_char *)ar_sha(ah), ":",
731 					    inet_ntoa(isaddr), ifp->if_xname);
732 				goto reply;
733 			}
734 			if (log_arp_movements) {
735 				ARP_LOG(LOG_INFO, "%s moved from %*D "
736 				    "to %*D on %s\n",
737 				    inet_ntoa(isaddr),
738 				    ifp->if_addrlen,
739 				    (u_char *)&la->ll_addr, ":",
740 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
741 				    ifp->if_xname);
742 			}
743 		}
744 
745 		if (ifp->if_addrlen != ah->ar_hln) {
746 			LLE_WUNLOCK(la);
747 			ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
748 			    "i/f %d (ignored)\n", ifp->if_addrlen,
749 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
750 			    ifp->if_addrlen);
751 			goto drop;
752 		}
753 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
754 		la->la_flags |= LLE_VALID;
755 
756 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
757 
758 		if (!(la->la_flags & LLE_STATIC)) {
759 			int canceled;
760 
761 			LLE_ADDREF(la);
762 			la->la_expire = time_uptime + V_arpt_keep;
763 			canceled = callout_reset(&la->la_timer,
764 			    hz * V_arpt_keep, arptimer, la);
765 			if (canceled)
766 				LLE_REMREF(la);
767 		}
768 		la->la_asked = 0;
769 		la->la_preempt = V_arp_maxtries;
770 		/*
771 		 * The packets are all freed within the call to the output
772 		 * routine.
773 		 *
774 		 * NB: The lock MUST be released before the call to the
775 		 * output routine.
776 		 */
777 		if (la->la_hold != NULL) {
778 			struct mbuf *m_hold, *m_hold_next;
779 
780 			m_hold = la->la_hold;
781 			la->la_hold = NULL;
782 			la->la_numheld = 0;
783 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
784 			LLE_WUNLOCK(la);
785 			for (; m_hold != NULL; m_hold = m_hold_next) {
786 				m_hold_next = m_hold->m_nextpkt;
787 				m_hold->m_nextpkt = NULL;
788 				/* Avoid confusing lower layers. */
789 				m_clrprotoflags(m_hold);
790 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
791 			}
792 		} else
793 			LLE_WUNLOCK(la);
794 	}
795 reply:
796 	if (op != ARPOP_REQUEST)
797 		goto drop;
798 	ARPSTAT_INC(rxrequests);
799 
800 	if (itaddr.s_addr == myaddr.s_addr) {
801 		/* Shortcut.. the receiving interface is the target. */
802 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
803 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
804 	} else {
805 		struct llentry *lle = NULL;
806 
807 		sin.sin_addr = itaddr;
808 		IF_AFDATA_LOCK(ifp);
809 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
810 		IF_AFDATA_UNLOCK(ifp);
811 
812 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
813 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
814 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
815 			LLE_RUNLOCK(lle);
816 		} else {
817 
818 			if (lle != NULL)
819 				LLE_RUNLOCK(lle);
820 
821 			if (!V_arp_proxyall)
822 				goto drop;
823 
824 			sin.sin_addr = itaddr;
825 			/* XXX MRT use table 0 for arp reply  */
826 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
827 			if (!rt)
828 				goto drop;
829 
830 			/*
831 			 * Don't send proxies for nodes on the same interface
832 			 * as this one came out of, or we'll get into a fight
833 			 * over who claims what Ether address.
834 			 */
835 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
836 				RTFREE_LOCKED(rt);
837 				goto drop;
838 			}
839 			RTFREE_LOCKED(rt);
840 
841 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
842 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
843 
844 			/*
845 			 * Also check that the node which sent the ARP packet
846 			 * is on the interface we expect it to be on. This
847 			 * avoids ARP chaos if an interface is connected to the
848 			 * wrong network.
849 			 */
850 			sin.sin_addr = isaddr;
851 
852 			/* XXX MRT use table 0 for arp checks */
853 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
854 			if (!rt)
855 				goto drop;
856 			if (rt->rt_ifp != ifp) {
857 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
858 				    " from %s via %s, expecting %s\n",
859 				    inet_ntoa(isaddr), ifp->if_xname,
860 				    rt->rt_ifp->if_xname);
861 				RTFREE_LOCKED(rt);
862 				goto drop;
863 			}
864 			RTFREE_LOCKED(rt);
865 
866 #ifdef DEBUG_PROXY
867 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
868 #endif
869 		}
870 	}
871 
872 	if (itaddr.s_addr == myaddr.s_addr &&
873 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
874 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
875 #ifdef DEBUG_LINKLOCAL
876 		printf("arp: sending reply for link-local addr %s\n",
877 		    inet_ntoa(itaddr));
878 #endif
879 		m->m_flags |= M_BCAST;
880 		m->m_flags &= ~M_MCAST;
881 	} else {
882 		/* default behaviour; never reply by broadcast. */
883 		m->m_flags &= ~(M_BCAST|M_MCAST);
884 	}
885 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
886 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
887 	ah->ar_op = htons(ARPOP_REPLY);
888 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
889 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
890 	m->m_pkthdr.len = m->m_len;
891 	m->m_pkthdr.rcvif = NULL;
892 	sa.sa_family = AF_ARP;
893 	sa.sa_len = 2;
894 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
895 	(*ifp->if_output)(ifp, m, &sa, NULL);
896 	ARPSTAT_INC(txreplies);
897 	return;
898 
899 drop:
900 	m_freem(m);
901 }
902 #endif
903 
904 void
905 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
906 {
907 	struct llentry *lle;
908 
909 	if (ifa->ifa_carp != NULL)
910 		return;
911 
912 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
913 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
914 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
915 		/*
916 		 * interface address is considered static entry
917 		 * because the output of the arp utility shows
918 		 * that L2 entry as permanent
919 		 */
920 		IF_AFDATA_LOCK(ifp);
921 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
922 				 (struct sockaddr *)IA_SIN(ifa));
923 		IF_AFDATA_UNLOCK(ifp);
924 		if (lle == NULL)
925 			log(LOG_INFO, "arp_ifinit: cannot create arp "
926 			    "entry for interface address\n");
927 		else
928 			LLE_RUNLOCK(lle);
929 	}
930 	ifa->ifa_rtrequest = NULL;
931 }
932 
933 void
934 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
935 {
936 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
937 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
938 				&IA_SIN(ifa)->sin_addr, enaddr);
939 	ifa->ifa_rtrequest = NULL;
940 }
941 
942 static void
943 arp_init(void)
944 {
945 
946 	netisr_register(&arp_nh);
947 }
948 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
949