xref: /freebsd/sys/netinet/if_ether.c (revision 09d325677d53a12c79a43664ff29871e92247629)
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 VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
89 					 * local traffic */
90 static VNET_DEFINE(int, arp_proxyall) = 0;
91 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
92 						 * 20 seconds */
93 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
94 VNET_PCPUSTAT_SYSINIT(arpstat);
95 
96 #ifdef VIMAGE
97 VNET_PCPUSTAT_SYSUNINIT(arpstat);
98 #endif /* VIMAGE */
99 
100 static VNET_DEFINE(int, arp_maxhold) = 1;
101 
102 #define	V_arpt_keep		VNET(arpt_keep)
103 #define	V_arpt_down		VNET(arpt_down)
104 #define	V_arp_maxtries		VNET(arp_maxtries)
105 #define	V_arp_proxyall		VNET(arp_proxyall)
106 #define	V_arp_maxhold		VNET(arp_maxhold)
107 
108 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
109 	&VNET_NAME(arpt_keep), 0,
110 	"ARP entry lifetime in seconds");
111 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
112 	&VNET_NAME(arp_maxtries), 0,
113 	"ARP resolution attempts before returning error");
114 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
115 	&VNET_NAME(useloopback), 0,
116 	"Use the loopback interface for local traffic");
117 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
118 	&VNET_NAME(arp_proxyall), 0,
119 	"Enable proxy ARP for all suitable requests");
120 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW,
121 	&VNET_NAME(arpt_down), 0,
122 	"Incomplete ARP entry lifetime in seconds");
123 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
124     arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
125 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW,
126 	&VNET_NAME(arp_maxhold), 0,
127 	"Number of packets to hold per ARP entry");
128 
129 static void	arp_init(void);
130 static void	arpintr(struct mbuf *);
131 static void	arptimer(void *);
132 #ifdef INET
133 static void	in_arpinput(struct mbuf *);
134 #endif
135 
136 static const struct netisr_handler arp_nh = {
137 	.nh_name = "arp",
138 	.nh_handler = arpintr,
139 	.nh_proto = NETISR_ARP,
140 	.nh_policy = NETISR_POLICY_SOURCE,
141 };
142 
143 #ifdef AF_INET
144 /*
145  * called by in_scrubprefix() to remove entry from the table when
146  * the interface goes away
147  */
148 void
149 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
150 {
151 	struct sockaddr_in addr4;
152 
153 	bzero((void *)&addr4, sizeof(addr4));
154 	addr4.sin_len    = sizeof(addr4);
155 	addr4.sin_family = AF_INET;
156 	addr4.sin_addr.s_addr = addr;
157 	IF_AFDATA_LOCK(ifp);
158 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
159 	    (struct sockaddr *)&addr4);
160 	IF_AFDATA_UNLOCK(ifp);
161 }
162 #endif
163 
164 /*
165  * Timeout routine.  Age arp_tab entries periodically.
166  */
167 static void
168 arptimer(void *arg)
169 {
170 	struct llentry *lle = (struct llentry *)arg;
171 	struct ifnet *ifp;
172 
173 	if (lle->la_flags & LLE_STATIC) {
174 		LLE_WUNLOCK(lle);
175 		return;
176 	}
177 
178 	ifp = lle->lle_tbl->llt_ifp;
179 	CURVNET_SET(ifp->if_vnet);
180 
181 	if ((lle->la_flags & LLE_DELETED) == 0) {
182 		int evt;
183 
184 		if (lle->la_flags & LLE_VALID)
185 			evt = LLENTRY_EXPIRED;
186 		else
187 			evt = LLENTRY_TIMEDOUT;
188 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
189 	}
190 
191 	callout_stop(&lle->la_timer);
192 
193 	/* XXX: LOR avoidance. We still have ref on lle. */
194 	LLE_WUNLOCK(lle);
195 	IF_AFDATA_LOCK(ifp);
196 	LLE_WLOCK(lle);
197 
198 	/* Guard against race with other llentry_free(). */
199 	if (lle->la_flags & LLE_LINKED) {
200 		size_t pkts_dropped;
201 
202 		LLE_REMREF(lle);
203 		pkts_dropped = llentry_free(lle);
204 		ARPSTAT_ADD(dropped, pkts_dropped);
205 	} else
206 		LLE_FREE_LOCKED(lle);
207 
208 	IF_AFDATA_UNLOCK(ifp);
209 
210 	ARPSTAT_INC(timeouts);
211 
212 	CURVNET_RESTORE();
213 }
214 
215 /*
216  * Broadcast an ARP request. Caller specifies:
217  *	- arp header source ip address
218  *	- arp header target ip address
219  *	- arp header source ethernet address
220  */
221 void
222 arprequest(struct ifnet *ifp, const struct in_addr *sip,
223     const struct in_addr *tip, u_char *enaddr)
224 {
225 	struct mbuf *m;
226 	struct arphdr *ah;
227 	struct sockaddr sa;
228 	u_char *carpaddr = NULL;
229 
230 	if (sip == NULL) {
231 		/*
232 		 * The caller did not supply a source address, try to find
233 		 * a compatible one among those assigned to this interface.
234 		 */
235 		struct ifaddr *ifa;
236 
237 		IF_ADDR_RLOCK(ifp);
238 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
239 			if (ifa->ifa_addr->sa_family != AF_INET)
240 				continue;
241 
242 			if (ifa->ifa_carp) {
243 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
244 					continue;
245 				sip = &IA_SIN(ifa)->sin_addr;
246 			} else {
247 				carpaddr = NULL;
248 				sip = &IA_SIN(ifa)->sin_addr;
249 			}
250 
251 			if (0 == ((sip->s_addr ^ tip->s_addr) &
252 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
253 				break;  /* found it. */
254 		}
255 		IF_ADDR_RUNLOCK(ifp);
256 		if (sip == NULL) {
257 			printf("%s: cannot find matching address\n", __func__);
258 			return;
259 		}
260 	}
261 	if (enaddr == NULL)
262 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
263 
264 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
265 		return;
266 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
267 		2*ifp->if_data.ifi_addrlen;
268 	m->m_pkthdr.len = m->m_len;
269 	MH_ALIGN(m, m->m_len);
270 	ah = mtod(m, struct arphdr *);
271 	bzero((caddr_t)ah, m->m_len);
272 #ifdef MAC
273 	mac_netinet_arp_send(ifp, m);
274 #endif
275 	ah->ar_pro = htons(ETHERTYPE_IP);
276 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
277 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
278 	ah->ar_op = htons(ARPOP_REQUEST);
279 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
280 	bcopy(sip, ar_spa(ah), ah->ar_pln);
281 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
282 	sa.sa_family = AF_ARP;
283 	sa.sa_len = 2;
284 	m->m_flags |= M_BCAST;
285 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
286 	(*ifp->if_output)(ifp, m, &sa, NULL);
287 	ARPSTAT_INC(txrequests);
288 }
289 
290 /*
291  * Resolve an IP address into an ethernet address.
292  * On input:
293  *    ifp is the interface we use
294  *    rt0 is the route to the final destination (possibly useless)
295  *    m is the mbuf. May be NULL if we don't have a packet.
296  *    dst is the next hop,
297  *    desten is where we want the address.
298  *
299  * On success, desten is filled in and the function returns 0;
300  * If the packet must be held pending resolution, we return EWOULDBLOCK
301  * On other errors, we return the corresponding error code.
302  * Note that m_freem() handles NULL.
303  */
304 int
305 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
306 	const struct sockaddr *dst, u_char *desten, struct llentry **lle)
307 {
308 	struct llentry *la = 0;
309 	u_int flags = 0;
310 	struct mbuf *curr = NULL;
311 	struct mbuf *next = NULL;
312 	int error, renew;
313 
314 	*lle = NULL;
315 	if (m != NULL) {
316 		if (m->m_flags & M_BCAST) {
317 			/* broadcast */
318 			(void)memcpy(desten,
319 			    ifp->if_broadcastaddr, ifp->if_addrlen);
320 			return (0);
321 		}
322 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
323 			/* multicast */
324 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
325 			return (0);
326 		}
327 	}
328 retry:
329 	IF_AFDATA_RLOCK(ifp);
330 	la = lla_lookup(LLTABLE(ifp), flags, dst);
331 	IF_AFDATA_RUNLOCK(ifp);
332 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
333 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
334 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
335 		IF_AFDATA_WLOCK(ifp);
336 		la = lla_lookup(LLTABLE(ifp), flags, dst);
337 		IF_AFDATA_WUNLOCK(ifp);
338 	}
339 	if (la == NULL) {
340 		if (flags & LLE_CREATE)
341 			log(LOG_DEBUG,
342 			    "arpresolve: can't allocate llinfo for %s\n",
343 			    inet_ntoa(SIN(dst)->sin_addr));
344 		m_freem(m);
345 		return (EINVAL);
346 	}
347 
348 	if ((la->la_flags & LLE_VALID) &&
349 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
350 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
351 		/*
352 		 * If entry has an expiry time and it is approaching,
353 		 * see if we need to send an ARP request within this
354 		 * arpt_down interval.
355 		 */
356 		if (!(la->la_flags & LLE_STATIC) &&
357 		    time_uptime + la->la_preempt > la->la_expire) {
358 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
359 			la->la_preempt--;
360 		}
361 
362 		*lle = la;
363 		error = 0;
364 		goto done;
365 	}
366 
367 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
368 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
369 		    inet_ntoa(SIN(dst)->sin_addr));
370 		m_freem(m);
371 		error = EINVAL;
372 		goto done;
373 	}
374 
375 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
376 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
377 		flags |= LLE_EXCLUSIVE;
378 		LLE_RUNLOCK(la);
379 		goto retry;
380 	}
381 	/*
382 	 * There is an arptab entry, but no ethernet address
383 	 * response yet.  Add the mbuf to the list, dropping
384 	 * the oldest packet if we have exceeded the system
385 	 * setting.
386 	 */
387 	if (m != NULL) {
388 		if (la->la_numheld >= V_arp_maxhold) {
389 			if (la->la_hold != NULL) {
390 				next = la->la_hold->m_nextpkt;
391 				m_freem(la->la_hold);
392 				la->la_hold = next;
393 				la->la_numheld--;
394 				ARPSTAT_INC(dropped);
395 			}
396 		}
397 		if (la->la_hold != NULL) {
398 			curr = la->la_hold;
399 			while (curr->m_nextpkt != NULL)
400 				curr = curr->m_nextpkt;
401 			curr->m_nextpkt = m;
402 		} else
403 			la->la_hold = m;
404 		la->la_numheld++;
405 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
406 			flags &= ~LLE_EXCLUSIVE;
407 			LLE_DOWNGRADE(la);
408 		}
409 
410 	}
411 	/*
412 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
413 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
414 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
415 	 * ARP request, but not faster than one request per second.
416 	 */
417 	if (la->la_asked < V_arp_maxtries)
418 		error = EWOULDBLOCK;	/* First request. */
419 	else
420 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
421 		    EHOSTUNREACH : EHOSTDOWN;
422 
423 	if (renew) {
424 		int canceled;
425 
426 		LLE_ADDREF(la);
427 		la->la_expire = time_uptime;
428 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
429 		    arptimer, la);
430 		if (canceled)
431 			LLE_REMREF(la);
432 		la->la_asked++;
433 		LLE_WUNLOCK(la);
434 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
435 		return (error);
436 	}
437 done:
438 	if (flags & LLE_EXCLUSIVE)
439 		LLE_WUNLOCK(la);
440 	else
441 		LLE_RUNLOCK(la);
442 	return (error);
443 }
444 
445 /*
446  * Common length and type checks are done here,
447  * then the protocol-specific routine is called.
448  */
449 static void
450 arpintr(struct mbuf *m)
451 {
452 	struct arphdr *ar;
453 
454 	if (m->m_len < sizeof(struct arphdr) &&
455 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
456 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
457 		return;
458 	}
459 	ar = mtod(m, struct arphdr *);
460 
461 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
462 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
463 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
464 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
465 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
466 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
467 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
468 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
469 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
470 		m_freem(m);
471 		return;
472 	}
473 
474 	if (m->m_len < arphdr_len(ar)) {
475 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
476 			log(LOG_NOTICE, "arp: runt packet\n");
477 			m_freem(m);
478 			return;
479 		}
480 		ar = mtod(m, struct arphdr *);
481 	}
482 
483 	ARPSTAT_INC(received);
484 	switch (ntohs(ar->ar_pro)) {
485 #ifdef INET
486 	case ETHERTYPE_IP:
487 		in_arpinput(m);
488 		return;
489 #endif
490 	}
491 	m_freem(m);
492 }
493 
494 #ifdef INET
495 /*
496  * ARP for Internet protocols on 10 Mb/s Ethernet.
497  * Algorithm is that given in RFC 826.
498  * In addition, a sanity check is performed on the sender
499  * protocol address, to catch impersonators.
500  * We no longer handle negotiations for use of trailer protocol:
501  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
502  * along with IP replies if we wanted trailers sent to us,
503  * and also sent them in response to IP replies.
504  * This allowed either end to announce the desire to receive
505  * trailer packets.
506  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
507  * but formerly didn't normally send requests.
508  */
509 static int log_arp_wrong_iface = 1;
510 static int log_arp_movements = 1;
511 static int log_arp_permanent_modify = 1;
512 static int allow_multicast = 0;
513 static struct timeval arp_lastlog;
514 static int arp_curpps;
515 static int arp_maxpps = 1;
516 
517 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
518 	&log_arp_wrong_iface, 0,
519 	"log arp packets arriving on the wrong interface");
520 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
521 	&log_arp_movements, 0,
522 	"log arp replies from MACs different than the one in the cache");
523 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
524 	&log_arp_permanent_modify, 0,
525 	"log arp replies from MACs different than the one in the permanent arp entry");
526 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
527 	&allow_multicast, 0, "accept multicast addresses");
528 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
529 	CTLFLAG_RW, &arp_maxpps, 0,
530 	"Maximum number of remotely triggered ARP messages that can be "
531 	"logged per second");
532 
533 #define	ARP_LOG(pri, ...)	do {					\
534 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
535 		log((pri), "arp: " __VA_ARGS__);			\
536 } while (0)
537 
538 static void
539 in_arpinput(struct mbuf *m)
540 {
541 	struct arphdr *ah;
542 	struct ifnet *ifp = m->m_pkthdr.rcvif;
543 	struct llentry *la = NULL;
544 	struct rtentry *rt;
545 	struct ifaddr *ifa;
546 	struct in_ifaddr *ia;
547 	struct sockaddr sa;
548 	struct in_addr isaddr, itaddr, myaddr;
549 	u_int8_t *enaddr = NULL;
550 	int op, flags;
551 	int req_len;
552 	int bridged = 0, is_bridge = 0;
553 	int carped;
554 	struct sockaddr_in sin;
555 	sin.sin_len = sizeof(struct sockaddr_in);
556 	sin.sin_family = AF_INET;
557 	sin.sin_addr.s_addr = 0;
558 
559 	if (ifp->if_bridge)
560 		bridged = 1;
561 	if (ifp->if_type == IFT_BRIDGE)
562 		is_bridge = 1;
563 
564 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
565 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
566 		ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
567 		return;
568 	}
569 
570 	ah = mtod(m, struct arphdr *);
571 	/*
572 	 * ARP is only for IPv4 so we can reject packets with
573 	 * a protocol length not equal to an IPv4 address.
574 	 */
575 	if (ah->ar_pln != sizeof(struct in_addr)) {
576 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
577 		    sizeof(struct in_addr));
578 		goto drop;
579 	}
580 
581 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
582 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
583 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
584 		goto drop;
585 	}
586 
587 	op = ntohs(ah->ar_op);
588 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
589 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
590 
591 	if (op == ARPOP_REPLY)
592 		ARPSTAT_INC(rxreplies);
593 
594 	/*
595 	 * For a bridge, we want to check the address irrespective
596 	 * of the receive interface. (This will change slightly
597 	 * when we have clusters of interfaces).
598 	 */
599 	IN_IFADDR_RLOCK();
600 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
601 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
602 		    ia->ia_ifp == ifp) &&
603 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
604 		    (ia->ia_ifa.ifa_carp == NULL ||
605 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
606 			ifa_ref(&ia->ia_ifa);
607 			IN_IFADDR_RUNLOCK();
608 			goto match;
609 		}
610 	}
611 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
612 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
613 		    ia->ia_ifp == ifp) &&
614 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
615 			ifa_ref(&ia->ia_ifa);
616 			IN_IFADDR_RUNLOCK();
617 			goto match;
618 		}
619 
620 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
621   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
622   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
623   addr == ia->ia_addr.sin_addr.s_addr)
624 	/*
625 	 * Check the case when bridge shares its MAC address with
626 	 * some of its children, so packets are claimed by bridge
627 	 * itself (bridge_input() does it first), but they are really
628 	 * meant to be destined to the bridge member.
629 	 */
630 	if (is_bridge) {
631 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
632 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
633 				ifa_ref(&ia->ia_ifa);
634 				ifp = ia->ia_ifp;
635 				IN_IFADDR_RUNLOCK();
636 				goto match;
637 			}
638 		}
639 	}
640 #undef BDG_MEMBER_MATCHES_ARP
641 	IN_IFADDR_RUNLOCK();
642 
643 	/*
644 	 * No match, use the first inet address on the receive interface
645 	 * as a dummy address for the rest of the function.
646 	 */
647 	IF_ADDR_RLOCK(ifp);
648 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
649 		if (ifa->ifa_addr->sa_family == AF_INET &&
650 		    (ifa->ifa_carp == NULL ||
651 		    (*carp_iamatch_p)(ifa, &enaddr))) {
652 			ia = ifatoia(ifa);
653 			ifa_ref(ifa);
654 			IF_ADDR_RUNLOCK(ifp);
655 			goto match;
656 		}
657 	IF_ADDR_RUNLOCK(ifp);
658 
659 	/*
660 	 * If bridging, fall back to using any inet address.
661 	 */
662 	IN_IFADDR_RLOCK();
663 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
664 		IN_IFADDR_RUNLOCK();
665 		goto drop;
666 	}
667 	ifa_ref(&ia->ia_ifa);
668 	IN_IFADDR_RUNLOCK();
669 match:
670 	if (!enaddr)
671 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
672 	carped = (ia->ia_ifa.ifa_carp != NULL);
673 	myaddr = ia->ia_addr.sin_addr;
674 	ifa_free(&ia->ia_ifa);
675 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
676 		goto drop;	/* it's from me, ignore it. */
677 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
678 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
679 		    "%s!\n", inet_ntoa(isaddr));
680 		goto drop;
681 	}
682 	/*
683 	 * Warn if another host is using the same IP address, but only if the
684 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
685 	 * case we suppress the warning to avoid false positive complaints of
686 	 * potential misconfiguration.
687 	 */
688 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
689 	    myaddr.s_addr != 0) {
690 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
691 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
692 		   inet_ntoa(isaddr), ifp->if_xname);
693 		itaddr = myaddr;
694 		ARPSTAT_INC(dupips);
695 		goto reply;
696 	}
697 	if (ifp->if_flags & IFF_STATICARP)
698 		goto reply;
699 
700 	bzero(&sin, sizeof(sin));
701 	sin.sin_len = sizeof(struct sockaddr_in);
702 	sin.sin_family = AF_INET;
703 	sin.sin_addr = isaddr;
704 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
705 	flags |= LLE_EXCLUSIVE;
706 	IF_AFDATA_LOCK(ifp);
707 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
708 	IF_AFDATA_UNLOCK(ifp);
709 	if (la != NULL) {
710 		/* the following is not an error when doing bridging */
711 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
712 			if (log_arp_wrong_iface)
713 				ARP_LOG(LOG_WARNING, "%s is on %s "
714 				    "but got reply from %*D on %s\n",
715 				    inet_ntoa(isaddr),
716 				    la->lle_tbl->llt_ifp->if_xname,
717 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
718 				    ifp->if_xname);
719 			LLE_WUNLOCK(la);
720 			goto reply;
721 		}
722 		if ((la->la_flags & LLE_VALID) &&
723 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
724 			if (la->la_flags & LLE_STATIC) {
725 				LLE_WUNLOCK(la);
726 				if (log_arp_permanent_modify)
727 					ARP_LOG(LOG_ERR,
728 					    "%*D attempts to modify "
729 					    "permanent entry for %s on %s\n",
730 					    ifp->if_addrlen,
731 					    (u_char *)ar_sha(ah), ":",
732 					    inet_ntoa(isaddr), ifp->if_xname);
733 				goto reply;
734 			}
735 			if (log_arp_movements) {
736 				ARP_LOG(LOG_INFO, "%s moved from %*D "
737 				    "to %*D on %s\n",
738 				    inet_ntoa(isaddr),
739 				    ifp->if_addrlen,
740 				    (u_char *)&la->ll_addr, ":",
741 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
742 				    ifp->if_xname);
743 			}
744 		}
745 
746 		if (ifp->if_addrlen != ah->ar_hln) {
747 			LLE_WUNLOCK(la);
748 			ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
749 			    "i/f %d (ignored)\n", ifp->if_addrlen,
750 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
751 			    ifp->if_addrlen);
752 			goto drop;
753 		}
754 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
755 		la->la_flags |= LLE_VALID;
756 
757 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
758 
759 		if (!(la->la_flags & LLE_STATIC)) {
760 			int canceled;
761 
762 			LLE_ADDREF(la);
763 			la->la_expire = time_uptime + V_arpt_keep;
764 			canceled = callout_reset(&la->la_timer,
765 			    hz * V_arpt_keep, arptimer, la);
766 			if (canceled)
767 				LLE_REMREF(la);
768 		}
769 		la->la_asked = 0;
770 		la->la_preempt = V_arp_maxtries;
771 		/*
772 		 * The packets are all freed within the call to the output
773 		 * routine.
774 		 *
775 		 * NB: The lock MUST be released before the call to the
776 		 * output routine.
777 		 */
778 		if (la->la_hold != NULL) {
779 			struct mbuf *m_hold, *m_hold_next;
780 
781 			m_hold = la->la_hold;
782 			la->la_hold = NULL;
783 			la->la_numheld = 0;
784 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
785 			LLE_WUNLOCK(la);
786 			for (; m_hold != NULL; m_hold = m_hold_next) {
787 				m_hold_next = m_hold->m_nextpkt;
788 				m_hold->m_nextpkt = NULL;
789 				/* Avoid confusing lower layers. */
790 				m_clrprotoflags(m_hold);
791 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
792 			}
793 		} else
794 			LLE_WUNLOCK(la);
795 	}
796 reply:
797 	if (op != ARPOP_REQUEST)
798 		goto drop;
799 	ARPSTAT_INC(rxrequests);
800 
801 	if (itaddr.s_addr == myaddr.s_addr) {
802 		/* Shortcut.. the receiving interface is the target. */
803 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
804 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
805 	} else {
806 		struct llentry *lle = NULL;
807 
808 		sin.sin_addr = itaddr;
809 		IF_AFDATA_LOCK(ifp);
810 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
811 		IF_AFDATA_UNLOCK(ifp);
812 
813 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
814 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
815 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
816 			LLE_RUNLOCK(lle);
817 		} else {
818 
819 			if (lle != NULL)
820 				LLE_RUNLOCK(lle);
821 
822 			if (!V_arp_proxyall)
823 				goto drop;
824 
825 			sin.sin_addr = itaddr;
826 			/* XXX MRT use table 0 for arp reply  */
827 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
828 			if (!rt)
829 				goto drop;
830 
831 			/*
832 			 * Don't send proxies for nodes on the same interface
833 			 * as this one came out of, or we'll get into a fight
834 			 * over who claims what Ether address.
835 			 */
836 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
837 				RTFREE_LOCKED(rt);
838 				goto drop;
839 			}
840 			RTFREE_LOCKED(rt);
841 
842 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
843 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
844 
845 			/*
846 			 * Also check that the node which sent the ARP packet
847 			 * is on the interface we expect it to be on. This
848 			 * avoids ARP chaos if an interface is connected to the
849 			 * wrong network.
850 			 */
851 			sin.sin_addr = isaddr;
852 
853 			/* XXX MRT use table 0 for arp checks */
854 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
855 			if (!rt)
856 				goto drop;
857 			if (rt->rt_ifp != ifp) {
858 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
859 				    " from %s via %s, expecting %s\n",
860 				    inet_ntoa(isaddr), ifp->if_xname,
861 				    rt->rt_ifp->if_xname);
862 				RTFREE_LOCKED(rt);
863 				goto drop;
864 			}
865 			RTFREE_LOCKED(rt);
866 
867 #ifdef DEBUG_PROXY
868 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
869 #endif
870 		}
871 	}
872 
873 	if (itaddr.s_addr == myaddr.s_addr &&
874 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
875 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
876 #ifdef DEBUG_LINKLOCAL
877 		printf("arp: sending reply for link-local addr %s\n",
878 		    inet_ntoa(itaddr));
879 #endif
880 		m->m_flags |= M_BCAST;
881 		m->m_flags &= ~M_MCAST;
882 	} else {
883 		/* default behaviour; never reply by broadcast. */
884 		m->m_flags &= ~(M_BCAST|M_MCAST);
885 	}
886 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
887 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
888 	ah->ar_op = htons(ARPOP_REPLY);
889 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
890 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
891 	m->m_pkthdr.len = m->m_len;
892 	m->m_pkthdr.rcvif = NULL;
893 	sa.sa_family = AF_ARP;
894 	sa.sa_len = 2;
895 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
896 	(*ifp->if_output)(ifp, m, &sa, NULL);
897 	ARPSTAT_INC(txreplies);
898 	return;
899 
900 drop:
901 	m_freem(m);
902 }
903 #endif
904 
905 void
906 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
907 {
908 	struct llentry *lle;
909 
910 	if (ifa->ifa_carp != NULL)
911 		return;
912 
913 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
914 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
915 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
916 		/*
917 		 * interface address is considered static entry
918 		 * because the output of the arp utility shows
919 		 * that L2 entry as permanent
920 		 */
921 		IF_AFDATA_LOCK(ifp);
922 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
923 				 (struct sockaddr *)IA_SIN(ifa));
924 		IF_AFDATA_UNLOCK(ifp);
925 		if (lle == NULL)
926 			log(LOG_INFO, "arp_ifinit: cannot create arp "
927 			    "entry for interface address\n");
928 		else
929 			LLE_RUNLOCK(lle);
930 	}
931 	ifa->ifa_rtrequest = NULL;
932 }
933 
934 void
935 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
936 {
937 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
938 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
939 				&IA_SIN(ifa)->sin_addr, enaddr);
940 	ifa->ifa_rtrequest = NULL;
941 }
942 
943 static void
944 arp_init(void)
945 {
946 
947 	netisr_register(&arp_nh);
948 }
949 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
950