xref: /freebsd/sys/netinet/if_ether.c (revision ff0ba87247820afbdfdc1b307c803f7923d0e4d3)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30  */
31 
32 /*
33  * Ethernet address resolution protocol.
34  * TODO:
35  *	add "inuse/lock" bit (or ref. count) along with valid bit
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/queue.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/syslog.h>
53 
54 #include <net/if.h>
55 #include <net/if_var.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/netisr.h>
59 #include <net/if_llc.h>
60 #include <net/ethernet.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <net/if_llatbl.h>
67 #include <netinet/if_ether.h>
68 #ifdef INET
69 #include <netinet/ip_carp.h>
70 #endif
71 
72 #include <net/if_arc.h>
73 #include <net/iso88025.h>
74 
75 #include <security/mac/mac_framework.h>
76 
77 #define SIN(s) ((const struct sockaddr_in *)(s))
78 #define SDL(s) ((struct sockaddr_dl *)s)
79 
80 SYSCTL_DECL(_net_link_ether);
81 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
82 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
83 
84 /* timer values */
85 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
86 						 * minutes */
87 static VNET_DEFINE(int, arp_maxtries) = 5;
88 static VNET_DEFINE(int, arp_proxyall) = 0;
89 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
90 						 * 20 seconds */
91 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
92 VNET_PCPUSTAT_SYSINIT(arpstat);
93 
94 #ifdef VIMAGE
95 VNET_PCPUSTAT_SYSUNINIT(arpstat);
96 #endif /* VIMAGE */
97 
98 static VNET_DEFINE(int, arp_maxhold) = 1;
99 
100 #define	V_arpt_keep		VNET(arpt_keep)
101 #define	V_arpt_down		VNET(arpt_down)
102 #define	V_arp_maxtries		VNET(arp_maxtries)
103 #define	V_arp_proxyall		VNET(arp_proxyall)
104 #define	V_arp_maxhold		VNET(arp_maxhold)
105 
106 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW,
107 	&VNET_NAME(arpt_keep), 0,
108 	"ARP entry lifetime in seconds");
109 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW,
110 	&VNET_NAME(arp_maxtries), 0,
111 	"ARP resolution attempts before returning error");
112 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW,
113 	&VNET_NAME(arp_proxyall), 0,
114 	"Enable proxy ARP for all suitable requests");
115 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW,
116 	&VNET_NAME(arpt_down), 0,
117 	"Incomplete ARP entry lifetime in seconds");
118 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
119     arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW,
121 	&VNET_NAME(arp_maxhold), 0,
122 	"Number of packets to hold per ARP entry");
123 
124 static void	arp_init(void);
125 static void	arpintr(struct mbuf *);
126 static void	arptimer(void *);
127 #ifdef INET
128 static void	in_arpinput(struct mbuf *);
129 #endif
130 
131 static const struct netisr_handler arp_nh = {
132 	.nh_name = "arp",
133 	.nh_handler = arpintr,
134 	.nh_proto = NETISR_ARP,
135 	.nh_policy = NETISR_POLICY_SOURCE,
136 };
137 
138 #ifdef AF_INET
139 /*
140  * called by in_scrubprefix() to remove entry from the table when
141  * the interface goes away
142  */
143 void
144 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
145 {
146 	struct sockaddr_in addr4;
147 
148 	bzero((void *)&addr4, sizeof(addr4));
149 	addr4.sin_len    = sizeof(addr4);
150 	addr4.sin_family = AF_INET;
151 	addr4.sin_addr.s_addr = addr;
152 	IF_AFDATA_RLOCK(ifp);
153 	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
154 	    (struct sockaddr *)&addr4);
155 	IF_AFDATA_RUNLOCK(ifp);
156 }
157 #endif
158 
159 /*
160  * Timeout routine.  Age arp_tab entries periodically.
161  */
162 static void
163 arptimer(void *arg)
164 {
165 	struct llentry *lle = (struct llentry *)arg;
166 	struct ifnet *ifp;
167 
168 	if (lle->la_flags & LLE_STATIC) {
169 		LLE_WUNLOCK(lle);
170 		return;
171 	}
172 
173 	ifp = lle->lle_tbl->llt_ifp;
174 	CURVNET_SET(ifp->if_vnet);
175 
176 	if ((lle->la_flags & LLE_DELETED) == 0) {
177 		int evt;
178 
179 		if (lle->la_flags & LLE_VALID)
180 			evt = LLENTRY_EXPIRED;
181 		else
182 			evt = LLENTRY_TIMEDOUT;
183 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
184 	}
185 
186 	callout_stop(&lle->la_timer);
187 
188 	/* XXX: LOR avoidance. We still have ref on lle. */
189 	LLE_WUNLOCK(lle);
190 	IF_AFDATA_LOCK(ifp);
191 	LLE_WLOCK(lle);
192 
193 	/* Guard against race with other llentry_free(). */
194 	if (lle->la_flags & LLE_LINKED) {
195 		size_t pkts_dropped;
196 
197 		LLE_REMREF(lle);
198 		pkts_dropped = llentry_free(lle);
199 		ARPSTAT_ADD(dropped, pkts_dropped);
200 	} else
201 		LLE_FREE_LOCKED(lle);
202 
203 	IF_AFDATA_UNLOCK(ifp);
204 
205 	ARPSTAT_INC(timeouts);
206 
207 	CURVNET_RESTORE();
208 }
209 
210 /*
211  * Broadcast an ARP request. Caller specifies:
212  *	- arp header source ip address
213  *	- arp header target ip address
214  *	- arp header source ethernet address
215  */
216 void
217 arprequest(struct ifnet *ifp, const struct in_addr *sip,
218     const struct in_addr *tip, u_char *enaddr)
219 {
220 	struct mbuf *m;
221 	struct arphdr *ah;
222 	struct sockaddr sa;
223 	u_char *carpaddr = NULL;
224 
225 	if (sip == NULL) {
226 		/*
227 		 * The caller did not supply a source address, try to find
228 		 * a compatible one among those assigned to this interface.
229 		 */
230 		struct ifaddr *ifa;
231 
232 		IF_ADDR_RLOCK(ifp);
233 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
234 			if (ifa->ifa_addr->sa_family != AF_INET)
235 				continue;
236 
237 			if (ifa->ifa_carp) {
238 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
239 					continue;
240 				sip = &IA_SIN(ifa)->sin_addr;
241 			} else {
242 				carpaddr = NULL;
243 				sip = &IA_SIN(ifa)->sin_addr;
244 			}
245 
246 			if (0 == ((sip->s_addr ^ tip->s_addr) &
247 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
248 				break;  /* found it. */
249 		}
250 		IF_ADDR_RUNLOCK(ifp);
251 		if (sip == NULL) {
252 			printf("%s: cannot find matching address\n", __func__);
253 			return;
254 		}
255 	}
256 	if (enaddr == NULL)
257 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
258 
259 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
260 		return;
261 	m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
262 		2 * ifp->if_addrlen;
263 	m->m_pkthdr.len = m->m_len;
264 	MH_ALIGN(m, m->m_len);
265 	ah = mtod(m, struct arphdr *);
266 	bzero((caddr_t)ah, m->m_len);
267 #ifdef MAC
268 	mac_netinet_arp_send(ifp, m);
269 #endif
270 	ah->ar_pro = htons(ETHERTYPE_IP);
271 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
272 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
273 	ah->ar_op = htons(ARPOP_REQUEST);
274 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
275 	bcopy(sip, ar_spa(ah), ah->ar_pln);
276 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
277 	sa.sa_family = AF_ARP;
278 	sa.sa_len = 2;
279 	m->m_flags |= M_BCAST;
280 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
281 	(*ifp->if_output)(ifp, m, &sa, NULL);
282 	ARPSTAT_INC(txrequests);
283 }
284 
285 /*
286  * Resolve an IP address into an ethernet address.
287  * On input:
288  *    ifp is the interface we use
289  *    rt0 is the route to the final destination (possibly useless)
290  *    m is the mbuf. May be NULL if we don't have a packet.
291  *    dst is the next hop,
292  *    desten is where we want the address.
293  *
294  * On success, desten is filled in and the function returns 0;
295  * If the packet must be held pending resolution, we return EWOULDBLOCK
296  * On other errors, we return the corresponding error code.
297  * Note that m_freem() handles NULL.
298  */
299 int
300 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
301 	const struct sockaddr *dst, u_char *desten, struct llentry **lle)
302 {
303 	struct llentry *la = 0;
304 	u_int flags = 0;
305 	struct mbuf *curr = NULL;
306 	struct mbuf *next = NULL;
307 	int error, renew;
308 
309 	*lle = NULL;
310 	if (m != NULL) {
311 		if (m->m_flags & M_BCAST) {
312 			/* broadcast */
313 			(void)memcpy(desten,
314 			    ifp->if_broadcastaddr, ifp->if_addrlen);
315 			return (0);
316 		}
317 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
318 			/* multicast */
319 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
320 			return (0);
321 		}
322 	}
323 retry:
324 	IF_AFDATA_RLOCK(ifp);
325 	la = lla_lookup(LLTABLE(ifp), flags, dst);
326 	IF_AFDATA_RUNLOCK(ifp);
327 	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
328 	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
329 		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
330 		IF_AFDATA_WLOCK(ifp);
331 		la = lla_lookup(LLTABLE(ifp), flags, dst);
332 		IF_AFDATA_WUNLOCK(ifp);
333 	}
334 	if (la == NULL) {
335 		if (flags & LLE_CREATE)
336 			log(LOG_DEBUG,
337 			    "arpresolve: can't allocate llinfo for %s on %s\n",
338 			    inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname);
339 		m_freem(m);
340 		return (EINVAL);
341 	}
342 
343 	if ((la->la_flags & LLE_VALID) &&
344 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
345 		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
346 		/*
347 		 * If entry has an expiry time and it is approaching,
348 		 * see if we need to send an ARP request within this
349 		 * arpt_down interval.
350 		 */
351 		if (!(la->la_flags & LLE_STATIC) &&
352 		    time_uptime + la->la_preempt > la->la_expire) {
353 			arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
354 			la->la_preempt--;
355 		}
356 
357 		*lle = la;
358 		error = 0;
359 		goto done;
360 	}
361 
362 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
363 		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
364 		    inet_ntoa(SIN(dst)->sin_addr));
365 		m_freem(m);
366 		error = EINVAL;
367 		goto done;
368 	}
369 
370 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
371 	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
372 		flags |= LLE_EXCLUSIVE;
373 		LLE_RUNLOCK(la);
374 		goto retry;
375 	}
376 	/*
377 	 * There is an arptab entry, but no ethernet address
378 	 * response yet.  Add the mbuf to the list, dropping
379 	 * the oldest packet if we have exceeded the system
380 	 * setting.
381 	 */
382 	if (m != NULL) {
383 		if (la->la_numheld >= V_arp_maxhold) {
384 			if (la->la_hold != NULL) {
385 				next = la->la_hold->m_nextpkt;
386 				m_freem(la->la_hold);
387 				la->la_hold = next;
388 				la->la_numheld--;
389 				ARPSTAT_INC(dropped);
390 			}
391 		}
392 		if (la->la_hold != NULL) {
393 			curr = la->la_hold;
394 			while (curr->m_nextpkt != NULL)
395 				curr = curr->m_nextpkt;
396 			curr->m_nextpkt = m;
397 		} else
398 			la->la_hold = m;
399 		la->la_numheld++;
400 		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
401 			flags &= ~LLE_EXCLUSIVE;
402 			LLE_DOWNGRADE(la);
403 		}
404 
405 	}
406 	/*
407 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
408 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
409 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
410 	 * ARP request, but not faster than one request per second.
411 	 */
412 	if (la->la_asked < V_arp_maxtries)
413 		error = EWOULDBLOCK;	/* First request. */
414 	else
415 		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
416 		    EHOSTUNREACH : EHOSTDOWN;
417 
418 	if (renew) {
419 		int canceled;
420 
421 		LLE_ADDREF(la);
422 		la->la_expire = time_uptime;
423 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
424 		    arptimer, la);
425 		if (canceled)
426 			LLE_REMREF(la);
427 		la->la_asked++;
428 		LLE_WUNLOCK(la);
429 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
430 		return (error);
431 	}
432 done:
433 	if (flags & LLE_EXCLUSIVE)
434 		LLE_WUNLOCK(la);
435 	else
436 		LLE_RUNLOCK(la);
437 	return (error);
438 }
439 
440 /*
441  * Common length and type checks are done here,
442  * then the protocol-specific routine is called.
443  */
444 static void
445 arpintr(struct mbuf *m)
446 {
447 	struct arphdr *ar;
448 
449 	if (m->m_len < sizeof(struct arphdr) &&
450 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
451 		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
452 		return;
453 	}
454 	ar = mtod(m, struct arphdr *);
455 
456 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
457 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
458 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
459 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
460 	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
461 		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
462 		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
463 		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
464 		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
465 		m_freem(m);
466 		return;
467 	}
468 
469 	if (m->m_len < arphdr_len(ar)) {
470 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
471 			log(LOG_NOTICE, "arp: runt packet\n");
472 			m_freem(m);
473 			return;
474 		}
475 		ar = mtod(m, struct arphdr *);
476 	}
477 
478 	ARPSTAT_INC(received);
479 	switch (ntohs(ar->ar_pro)) {
480 #ifdef INET
481 	case ETHERTYPE_IP:
482 		in_arpinput(m);
483 		return;
484 #endif
485 	}
486 	m_freem(m);
487 }
488 
489 #ifdef INET
490 /*
491  * ARP for Internet protocols on 10 Mb/s Ethernet.
492  * Algorithm is that given in RFC 826.
493  * In addition, a sanity check is performed on the sender
494  * protocol address, to catch impersonators.
495  * We no longer handle negotiations for use of trailer protocol:
496  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
497  * along with IP replies if we wanted trailers sent to us,
498  * and also sent them in response to IP replies.
499  * This allowed either end to announce the desire to receive
500  * trailer packets.
501  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
502  * but formerly didn't normally send requests.
503  */
504 static int log_arp_wrong_iface = 1;
505 static int log_arp_movements = 1;
506 static int log_arp_permanent_modify = 1;
507 static int allow_multicast = 0;
508 static struct timeval arp_lastlog;
509 static int arp_curpps;
510 static int arp_maxpps = 1;
511 
512 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
513 	&log_arp_wrong_iface, 0,
514 	"log arp packets arriving on the wrong interface");
515 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
516 	&log_arp_movements, 0,
517 	"log arp replies from MACs different than the one in the cache");
518 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
519 	&log_arp_permanent_modify, 0,
520 	"log arp replies from MACs different than the one in the permanent arp entry");
521 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
522 	&allow_multicast, 0, "accept multicast addresses");
523 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
524 	CTLFLAG_RW, &arp_maxpps, 0,
525 	"Maximum number of remotely triggered ARP messages that can be "
526 	"logged per second");
527 
528 #define	ARP_LOG(pri, ...)	do {					\
529 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
530 		log((pri), "arp: " __VA_ARGS__);			\
531 } while (0)
532 
533 static void
534 in_arpinput(struct mbuf *m)
535 {
536 	struct arphdr *ah;
537 	struct ifnet *ifp = m->m_pkthdr.rcvif;
538 	struct llentry *la = NULL;
539 	struct rtentry *rt;
540 	struct ifaddr *ifa;
541 	struct in_ifaddr *ia;
542 	struct sockaddr sa;
543 	struct in_addr isaddr, itaddr, myaddr;
544 	u_int8_t *enaddr = NULL;
545 	int op, flags;
546 	int req_len;
547 	int bridged = 0, is_bridge = 0;
548 	int carped;
549 	struct sockaddr_in sin;
550 	sin.sin_len = sizeof(struct sockaddr_in);
551 	sin.sin_family = AF_INET;
552 	sin.sin_addr.s_addr = 0;
553 
554 	if (ifp->if_bridge)
555 		bridged = 1;
556 	if (ifp->if_type == IFT_BRIDGE)
557 		is_bridge = 1;
558 
559 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
560 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
561 		ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
562 		return;
563 	}
564 
565 	ah = mtod(m, struct arphdr *);
566 	/*
567 	 * ARP is only for IPv4 so we can reject packets with
568 	 * a protocol length not equal to an IPv4 address.
569 	 */
570 	if (ah->ar_pln != sizeof(struct in_addr)) {
571 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
572 		    sizeof(struct in_addr));
573 		goto drop;
574 	}
575 
576 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
577 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
578 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
579 		goto drop;
580 	}
581 
582 	op = ntohs(ah->ar_op);
583 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
584 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
585 
586 	if (op == ARPOP_REPLY)
587 		ARPSTAT_INC(rxreplies);
588 
589 	/*
590 	 * For a bridge, we want to check the address irrespective
591 	 * of the receive interface. (This will change slightly
592 	 * when we have clusters of interfaces).
593 	 */
594 	IN_IFADDR_RLOCK();
595 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
596 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
597 		    ia->ia_ifp == ifp) &&
598 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
599 		    (ia->ia_ifa.ifa_carp == NULL ||
600 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
601 			ifa_ref(&ia->ia_ifa);
602 			IN_IFADDR_RUNLOCK();
603 			goto match;
604 		}
605 	}
606 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
607 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
608 		    ia->ia_ifp == ifp) &&
609 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
610 			ifa_ref(&ia->ia_ifa);
611 			IN_IFADDR_RUNLOCK();
612 			goto match;
613 		}
614 
615 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
616   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
617   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
618   addr == ia->ia_addr.sin_addr.s_addr)
619 	/*
620 	 * Check the case when bridge shares its MAC address with
621 	 * some of its children, so packets are claimed by bridge
622 	 * itself (bridge_input() does it first), but they are really
623 	 * meant to be destined to the bridge member.
624 	 */
625 	if (is_bridge) {
626 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
627 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
628 				ifa_ref(&ia->ia_ifa);
629 				ifp = ia->ia_ifp;
630 				IN_IFADDR_RUNLOCK();
631 				goto match;
632 			}
633 		}
634 	}
635 #undef BDG_MEMBER_MATCHES_ARP
636 	IN_IFADDR_RUNLOCK();
637 
638 	/*
639 	 * No match, use the first inet address on the receive interface
640 	 * as a dummy address for the rest of the function.
641 	 */
642 	IF_ADDR_RLOCK(ifp);
643 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
644 		if (ifa->ifa_addr->sa_family == AF_INET &&
645 		    (ifa->ifa_carp == NULL ||
646 		    (*carp_iamatch_p)(ifa, &enaddr))) {
647 			ia = ifatoia(ifa);
648 			ifa_ref(ifa);
649 			IF_ADDR_RUNLOCK(ifp);
650 			goto match;
651 		}
652 	IF_ADDR_RUNLOCK(ifp);
653 
654 	/*
655 	 * If bridging, fall back to using any inet address.
656 	 */
657 	IN_IFADDR_RLOCK();
658 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
659 		IN_IFADDR_RUNLOCK();
660 		goto drop;
661 	}
662 	ifa_ref(&ia->ia_ifa);
663 	IN_IFADDR_RUNLOCK();
664 match:
665 	if (!enaddr)
666 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
667 	carped = (ia->ia_ifa.ifa_carp != NULL);
668 	myaddr = ia->ia_addr.sin_addr;
669 	ifa_free(&ia->ia_ifa);
670 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
671 		goto drop;	/* it's from me, ignore it. */
672 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
673 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
674 		    "%s!\n", inet_ntoa(isaddr));
675 		goto drop;
676 	}
677 	/*
678 	 * Warn if another host is using the same IP address, but only if the
679 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
680 	 * case we suppress the warning to avoid false positive complaints of
681 	 * potential misconfiguration.
682 	 */
683 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
684 	    myaddr.s_addr != 0) {
685 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
686 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
687 		   inet_ntoa(isaddr), ifp->if_xname);
688 		itaddr = myaddr;
689 		ARPSTAT_INC(dupips);
690 		goto reply;
691 	}
692 	if (ifp->if_flags & IFF_STATICARP)
693 		goto reply;
694 
695 	bzero(&sin, sizeof(sin));
696 	sin.sin_len = sizeof(struct sockaddr_in);
697 	sin.sin_family = AF_INET;
698 	sin.sin_addr = isaddr;
699 	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
700 	flags |= LLE_EXCLUSIVE;
701 	IF_AFDATA_LOCK(ifp);
702 	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
703 	IF_AFDATA_UNLOCK(ifp);
704 	if (la != NULL) {
705 		/* the following is not an error when doing bridging */
706 		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
707 			if (log_arp_wrong_iface)
708 				ARP_LOG(LOG_WARNING, "%s is on %s "
709 				    "but got reply from %*D on %s\n",
710 				    inet_ntoa(isaddr),
711 				    la->lle_tbl->llt_ifp->if_xname,
712 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
713 				    ifp->if_xname);
714 			LLE_WUNLOCK(la);
715 			goto reply;
716 		}
717 		if ((la->la_flags & LLE_VALID) &&
718 		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
719 			if (la->la_flags & LLE_STATIC) {
720 				LLE_WUNLOCK(la);
721 				if (log_arp_permanent_modify)
722 					ARP_LOG(LOG_ERR,
723 					    "%*D attempts to modify "
724 					    "permanent entry for %s on %s\n",
725 					    ifp->if_addrlen,
726 					    (u_char *)ar_sha(ah), ":",
727 					    inet_ntoa(isaddr), ifp->if_xname);
728 				goto reply;
729 			}
730 			if (log_arp_movements) {
731 				ARP_LOG(LOG_INFO, "%s moved from %*D "
732 				    "to %*D on %s\n",
733 				    inet_ntoa(isaddr),
734 				    ifp->if_addrlen,
735 				    (u_char *)&la->ll_addr, ":",
736 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
737 				    ifp->if_xname);
738 			}
739 		}
740 
741 		if (ifp->if_addrlen != ah->ar_hln) {
742 			LLE_WUNLOCK(la);
743 			ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
744 			    "i/f %d (ignored)\n", ifp->if_addrlen,
745 			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
746 			    ifp->if_addrlen);
747 			goto drop;
748 		}
749 		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
750 		la->la_flags |= LLE_VALID;
751 
752 		EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
753 
754 		if (!(la->la_flags & LLE_STATIC)) {
755 			int canceled;
756 
757 			LLE_ADDREF(la);
758 			la->la_expire = time_uptime + V_arpt_keep;
759 			canceled = callout_reset(&la->la_timer,
760 			    hz * V_arpt_keep, arptimer, la);
761 			if (canceled)
762 				LLE_REMREF(la);
763 		}
764 		la->la_asked = 0;
765 		la->la_preempt = V_arp_maxtries;
766 		/*
767 		 * The packets are all freed within the call to the output
768 		 * routine.
769 		 *
770 		 * NB: The lock MUST be released before the call to the
771 		 * output routine.
772 		 */
773 		if (la->la_hold != NULL) {
774 			struct mbuf *m_hold, *m_hold_next;
775 
776 			m_hold = la->la_hold;
777 			la->la_hold = NULL;
778 			la->la_numheld = 0;
779 			memcpy(&sa, L3_ADDR(la), sizeof(sa));
780 			LLE_WUNLOCK(la);
781 			for (; m_hold != NULL; m_hold = m_hold_next) {
782 				m_hold_next = m_hold->m_nextpkt;
783 				m_hold->m_nextpkt = NULL;
784 				/* Avoid confusing lower layers. */
785 				m_clrprotoflags(m_hold);
786 				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
787 			}
788 		} else
789 			LLE_WUNLOCK(la);
790 	}
791 reply:
792 	if (op != ARPOP_REQUEST)
793 		goto drop;
794 	ARPSTAT_INC(rxrequests);
795 
796 	if (itaddr.s_addr == myaddr.s_addr) {
797 		/* Shortcut.. the receiving interface is the target. */
798 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
799 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
800 	} else {
801 		struct llentry *lle = NULL;
802 
803 		sin.sin_addr = itaddr;
804 		IF_AFDATA_RLOCK(ifp);
805 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
806 		IF_AFDATA_RUNLOCK(ifp);
807 
808 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
809 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
810 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
811 			LLE_RUNLOCK(lle);
812 		} else {
813 
814 			if (lle != NULL)
815 				LLE_RUNLOCK(lle);
816 
817 			if (!V_arp_proxyall)
818 				goto drop;
819 
820 			sin.sin_addr = itaddr;
821 			/* XXX MRT use table 0 for arp reply  */
822 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
823 			if (!rt)
824 				goto drop;
825 
826 			/*
827 			 * Don't send proxies for nodes on the same interface
828 			 * as this one came out of, or we'll get into a fight
829 			 * over who claims what Ether address.
830 			 */
831 			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
832 				RTFREE_LOCKED(rt);
833 				goto drop;
834 			}
835 			RTFREE_LOCKED(rt);
836 
837 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
838 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
839 
840 			/*
841 			 * Also check that the node which sent the ARP packet
842 			 * is on the interface we expect it to be on. This
843 			 * avoids ARP chaos if an interface is connected to the
844 			 * wrong network.
845 			 */
846 			sin.sin_addr = isaddr;
847 
848 			/* XXX MRT use table 0 for arp checks */
849 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
850 			if (!rt)
851 				goto drop;
852 			if (rt->rt_ifp != ifp) {
853 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
854 				    " from %s via %s, expecting %s\n",
855 				    inet_ntoa(isaddr), ifp->if_xname,
856 				    rt->rt_ifp->if_xname);
857 				RTFREE_LOCKED(rt);
858 				goto drop;
859 			}
860 			RTFREE_LOCKED(rt);
861 
862 #ifdef DEBUG_PROXY
863 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
864 #endif
865 		}
866 	}
867 
868 	if (itaddr.s_addr == myaddr.s_addr &&
869 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
870 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
871 #ifdef DEBUG_LINKLOCAL
872 		printf("arp: sending reply for link-local addr %s\n",
873 		    inet_ntoa(itaddr));
874 #endif
875 		m->m_flags |= M_BCAST;
876 		m->m_flags &= ~M_MCAST;
877 	} else {
878 		/* default behaviour; never reply by broadcast. */
879 		m->m_flags &= ~(M_BCAST|M_MCAST);
880 	}
881 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
882 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
883 	ah->ar_op = htons(ARPOP_REPLY);
884 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
885 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
886 	m->m_pkthdr.len = m->m_len;
887 	m->m_pkthdr.rcvif = NULL;
888 	sa.sa_family = AF_ARP;
889 	sa.sa_len = 2;
890 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
891 	(*ifp->if_output)(ifp, m, &sa, NULL);
892 	ARPSTAT_INC(txreplies);
893 	return;
894 
895 drop:
896 	m_freem(m);
897 }
898 #endif
899 
900 void
901 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
902 {
903 	struct llentry *lle;
904 
905 	if (ifa->ifa_carp != NULL)
906 		return;
907 
908 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
909 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
910 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
911 		/*
912 		 * interface address is considered static entry
913 		 * because the output of the arp utility shows
914 		 * that L2 entry as permanent
915 		 */
916 		IF_AFDATA_LOCK(ifp);
917 		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
918 				 (struct sockaddr *)IA_SIN(ifa));
919 		IF_AFDATA_UNLOCK(ifp);
920 		if (lle == NULL)
921 			log(LOG_INFO, "arp_ifinit: cannot create arp "
922 			    "entry for interface address\n");
923 		else
924 			LLE_RUNLOCK(lle);
925 	}
926 	ifa->ifa_rtrequest = NULL;
927 }
928 
929 void
930 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
931 {
932 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
933 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
934 				&IA_SIN(ifa)->sin_addr, enaddr);
935 	ifa->ifa_rtrequest = NULL;
936 }
937 
938 static void
939 arp_init(void)
940 {
941 
942 	netisr_register(&arp_nh);
943 }
944 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
945