xref: /freebsd/sys/netinet/if_ether.c (revision 30d239bc4c510432e65a84fa1c14ed67a3ab1c92)
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 #include "opt_mac.h"
43 #include "opt_carp.h"
44 
45 #include <sys/param.h>
46 #include <sys/kernel.h>
47 #include <sys/queue.h>
48 #include <sys/sysctl.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/malloc.h>
52 #include <sys/socket.h>
53 #include <sys/syslog.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 #include <net/netisr.h>
60 #include <net/if_llc.h>
61 #include <net/ethernet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_var.h>
65 #include <netinet/if_ether.h>
66 
67 #include <net/if_arc.h>
68 #include <net/iso88025.h>
69 
70 #ifdef DEV_CARP
71 #include <netinet/ip_carp.h>
72 #endif
73 
74 #include <security/mac/mac_framework.h>
75 
76 #define SIN(s) ((struct sockaddr_in *)s)
77 #define SDL(s) ((struct sockaddr_dl *)s)
78 
79 SYSCTL_DECL(_net_link_ether);
80 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 
82 /* timer values */
83 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
84 
85 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
86 	   &arpt_keep, 0, "ARP entry lifetime in seconds");
87 
88 #define	rt_expire rt_rmx.rmx_expire
89 
90 struct llinfo_arp {
91 	struct	callout la_timer;
92 	struct	rtentry *la_rt;
93 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
94 	u_short	la_preempt;	/* countdown for pre-expiry arps */
95 	u_short	la_asked;	/* # requests sent */
96 };
97 
98 static struct	ifqueue arpintrq;
99 static int	arp_allocated;
100 
101 static int	arp_maxtries = 5;
102 static int	useloopback = 1; /* use loopback interface for local traffic */
103 static int	arp_proxyall = 0;
104 
105 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
106 	   &arp_maxtries, 0, "ARP resolution attempts before returning error");
107 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
108 	   &useloopback, 0, "Use the loopback interface for local traffic");
109 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
110 	   &arp_proxyall, 0, "Enable proxy ARP for all suitable requests");
111 
112 static void	arp_init(void);
113 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
114 static void	arprequest(struct ifnet *,
115 			struct in_addr *, struct in_addr *, u_char *);
116 static void	arpintr(struct mbuf *);
117 static void	arptimer(void *);
118 static struct rtentry
119 		*arplookup(u_long, int, int);
120 #ifdef INET
121 static void	in_arpinput(struct mbuf *);
122 #endif
123 
124 /*
125  * Timeout routine.
126  */
127 static void
128 arptimer(void *arg)
129 {
130 	struct rtentry *rt = (struct rtentry *)arg;
131 
132 	RT_LOCK_ASSERT(rt);
133 	/*
134 	 * The lock is needed to close a theoretical race
135 	 * between spontaneous expiry and intentional removal.
136 	 * We still got an extra reference on rtentry, so can
137 	 * safely pass pointers to its contents.
138 	 */
139 	RT_UNLOCK(rt);
140 
141 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
142 }
143 
144 /*
145  * Parallel to llc_rtrequest.
146  */
147 static void
148 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
149 {
150 	struct sockaddr *gate;
151 	struct llinfo_arp *la;
152 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
153 	struct in_ifaddr *ia;
154 	struct ifaddr *ifa;
155 
156 	RT_LOCK_ASSERT(rt);
157 
158 	if (rt->rt_flags & RTF_GATEWAY)
159 		return;
160 	gate = rt->rt_gateway;
161 	la = (struct llinfo_arp *)rt->rt_llinfo;
162 	switch (req) {
163 
164 	case RTM_ADD:
165 		/*
166 		 * XXX: If this is a manually added route to interface
167 		 * such as older version of routed or gated might provide,
168 		 * restore cloning bit.
169 		 */
170 		if ((rt->rt_flags & RTF_HOST) == 0 &&
171 		    rt_mask(rt) != NULL &&
172 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
173 			rt->rt_flags |= RTF_CLONING;
174 		if (rt->rt_flags & RTF_CLONING) {
175 			/*
176 			 * Case 1: This route should come from a route to iface.
177 			 */
178 			rt_setgate(rt, rt_key(rt),
179 					(struct sockaddr *)&null_sdl);
180 			gate = rt->rt_gateway;
181 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
182 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
183 			rt->rt_expire = time_uptime;
184 			break;
185 		}
186 		/* Announce a new entry if requested. */
187 		if (rt->rt_flags & RTF_ANNOUNCE)
188 			arprequest(rt->rt_ifp,
189 			    &SIN(rt_key(rt))->sin_addr,
190 			    &SIN(rt_key(rt))->sin_addr,
191 			    (u_char *)LLADDR(SDL(gate)));
192 		/*FALLTHROUGH*/
193 	case RTM_RESOLVE:
194 		if (gate->sa_family != AF_LINK ||
195 		    gate->sa_len < sizeof(null_sdl)) {
196 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
197 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
198 			    (gate->sa_family != AF_LINK) ?
199 			    " (!AF_LINK)": "");
200 			break;
201 		}
202 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
203 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
204 		if (la != 0)
205 			break; /* This happens on a route change */
206 		/*
207 		 * Case 2:  This route may come from cloning, or a manual route
208 		 * add with a LL address.
209 		 */
210 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
211 		rt->rt_llinfo = (caddr_t)la;
212 		if (la == 0) {
213 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
214 			break;
215 		}
216 		arp_allocated++;
217 		/*
218 		 * We are storing a route entry outside of radix tree. So,
219 		 * it can be found and accessed by other means than radix
220 		 * lookup. The routing code assumes that any rtentry detached
221 		 * from radix can be destroyed safely. To prevent this, we
222 		 * add an additional reference.
223 		 */
224 		RT_ADDREF(rt);
225 		la->la_rt = rt;
226 		rt->rt_flags |= RTF_LLINFO;
227 		callout_init_mtx(&la->la_timer, &rt->rt_mtx,
228 		    CALLOUT_RETURNUNLOCKED);
229 
230 #ifdef INET
231 		/*
232 		 * This keeps the multicast addresses from showing up
233 		 * in `arp -a' listings as unresolved.  It's not actually
234 		 * functional.  Then the same for broadcast.
235 		 */
236 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
237 		    rt->rt_ifp->if_type != IFT_ARCNET) {
238 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
239 					       LLADDR(SDL(gate)));
240 			SDL(gate)->sdl_alen = 6;
241 			rt->rt_expire = 0;
242 		}
243 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
244 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
245 			       rt->rt_ifp->if_addrlen);
246 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
247 			rt->rt_expire = 0;
248 		}
249 #endif
250 
251 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
252 			if (ia->ia_ifp == rt->rt_ifp &&
253 			    SIN(rt_key(rt))->sin_addr.s_addr ==
254 			    (IA_SIN(ia))->sin_addr.s_addr)
255 				break;
256 		}
257 		if (ia) {
258 		    /*
259 		     * This test used to be
260 		     *	if (loif.if_flags & IFF_UP)
261 		     * It allowed local traffic to be forced
262 		     * through the hardware by configuring the loopback down.
263 		     * However, it causes problems during network configuration
264 		     * for boards that can't receive packets they send.
265 		     * It is now necessary to clear "useloopback" and remove
266 		     * the route to force traffic out to the hardware.
267 		     */
268 			rt->rt_expire = 0;
269 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
270 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
271 			if (useloopback) {
272 				rt->rt_ifp = loif;
273 				rt->rt_rmx.rmx_mtu = loif->if_mtu;
274 			}
275 
276 		    /*
277 		     * make sure to set rt->rt_ifa to the interface
278 		     * address we are using, otherwise we will have trouble
279 		     * with source address selection.
280 		     */
281 			ifa = &ia->ia_ifa;
282 			if (ifa != rt->rt_ifa) {
283 				IFAFREE(rt->rt_ifa);
284 				IFAREF(ifa);
285 				rt->rt_ifa = ifa;
286 			}
287 		}
288 		break;
289 
290 	case RTM_DELETE:
291 		if (la == NULL)	/* XXX: at least CARP does this. */
292 			break;
293 		callout_stop(&la->la_timer);
294 		rt->rt_llinfo = NULL;
295 		rt->rt_flags &= ~RTF_LLINFO;
296 		RT_REMREF(rt);
297 		if (la->la_hold)
298 			m_freem(la->la_hold);
299 		Free((caddr_t)la);
300 	}
301 }
302 
303 /*
304  * Broadcast an ARP request. Caller specifies:
305  *	- arp header source ip address
306  *	- arp header target ip address
307  *	- arp header source ethernet address
308  */
309 static void
310 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
311     u_char *enaddr)
312 {
313 	struct mbuf *m;
314 	struct arphdr *ah;
315 	struct sockaddr sa;
316 
317 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
318 		return;
319 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
320 		2*ifp->if_data.ifi_addrlen;
321 	m->m_pkthdr.len = m->m_len;
322 	MH_ALIGN(m, m->m_len);
323 	ah = mtod(m, struct arphdr *);
324 	bzero((caddr_t)ah, m->m_len);
325 #ifdef MAC
326 	mac_create_mbuf_linklayer(ifp, m);
327 #endif
328 	ah->ar_pro = htons(ETHERTYPE_IP);
329 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
330 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
331 	ah->ar_op = htons(ARPOP_REQUEST);
332 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
333 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
334 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
335 	sa.sa_family = AF_ARP;
336 	sa.sa_len = 2;
337 	m->m_flags |= M_BCAST;
338 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
339 
340 	return;
341 }
342 
343 /*
344  * Resolve an IP address into an ethernet address.
345  * On input:
346  *    ifp is the interface we use
347  *    dst is the next hop,
348  *    rt0 is the route to the final destination (possibly useless)
349  *    m is the mbuf
350  *    desten is where we want the address.
351  *
352  * On success, desten is filled in and the function returns 0;
353  * If the packet must be held pending resolution, we return EWOULDBLOCK
354  * On other errors, we return the corresponding error code.
355  */
356 int
357 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
358     struct sockaddr *dst, u_char *desten)
359 {
360 	struct llinfo_arp *la = NULL;
361 	struct rtentry *rt = NULL;
362 	struct sockaddr_dl *sdl;
363 	int error;
364 
365 	if (m->m_flags & M_BCAST) {	/* broadcast */
366 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
367 		return (0);
368 	}
369 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
370 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
371 		return (0);
372 	}
373 
374 	if (rt0 != NULL) {
375 		error = rt_check(&rt, &rt0, dst);
376 		if (error) {
377 			m_freem(m);
378 			return error;
379 		}
380 		la = (struct llinfo_arp *)rt->rt_llinfo;
381 		if (la == NULL)
382 			RT_UNLOCK(rt);
383 	}
384 	if (la == NULL) {
385 		/*
386 		 * We enter this block in case if rt0 was NULL,
387 		 * or if rt found by rt_check() didn't have llinfo.
388 		 */
389 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
390 		if (rt == NULL) {
391 			log(LOG_DEBUG,
392 			    "arpresolve: can't allocate route for %s\n",
393 			    inet_ntoa(SIN(dst)->sin_addr));
394 			m_freem(m);
395 			return (EINVAL); /* XXX */
396 		}
397 		la = (struct llinfo_arp *)rt->rt_llinfo;
398 		if (la == NULL) {
399 			RT_UNLOCK(rt);
400 			log(LOG_DEBUG,
401 			    "arpresolve: can't allocate llinfo for %s\n",
402 			    inet_ntoa(SIN(dst)->sin_addr));
403 			m_freem(m);
404 			return (EINVAL); /* XXX */
405 		}
406 	}
407 	sdl = SDL(rt->rt_gateway);
408 	/*
409 	 * Check the address family and length is valid, the address
410 	 * is resolved; otherwise, try to resolve.
411 	 */
412 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
413 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
414 
415 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
416 
417 		/*
418 		 * If entry has an expiry time and it is approaching,
419 		 * send an ARP request.
420 		 */
421 		if ((rt->rt_expire != 0) &&
422 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
423 			struct in_addr sin =
424 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
425 
426 			la->la_preempt--;
427 			RT_UNLOCK(rt);
428 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
429 			    IF_LLADDR(ifp));
430 			return (0);
431 		}
432 
433 		RT_UNLOCK(rt);
434 		return (0);
435 	}
436 	/*
437 	 * If ARP is disabled or static on this interface, stop.
438 	 * XXX
439 	 * Probably should not allocate empty llinfo struct if we are
440 	 * not going to be sending out an arp request.
441 	 */
442 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
443 		RT_UNLOCK(rt);
444 		m_freem(m);
445 		return (EINVAL);
446 	}
447 	/*
448 	 * There is an arptab entry, but no ethernet address
449 	 * response yet.  Replace the held mbuf with this
450 	 * latest one.
451 	 */
452 	if (la->la_hold)
453 		m_freem(la->la_hold);
454 	la->la_hold = m;
455 
456 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
457 
458 	/*
459 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
460 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
461 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
462 	 * ARP request, but not faster than one request per second.
463 	 */
464 	if (la->la_asked < arp_maxtries)
465 		error = EWOULDBLOCK;	/* First request. */
466 	else
467 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
468 
469 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
470 		struct in_addr sin =
471 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
472 
473 		rt->rt_expire = time_uptime;
474 		callout_reset(&la->la_timer, hz, arptimer, rt);
475 		la->la_asked++;
476 		RT_UNLOCK(rt);
477 
478 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
479 		    IF_LLADDR(ifp));
480 	} else
481 		RT_UNLOCK(rt);
482 
483 	return (error);
484 }
485 
486 /*
487  * Common length and type checks are done here,
488  * then the protocol-specific routine is called.
489  */
490 static void
491 arpintr(struct mbuf *m)
492 {
493 	struct arphdr *ar;
494 
495 	if (m->m_len < sizeof(struct arphdr) &&
496 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
497 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
498 		return;
499 	}
500 	ar = mtod(m, struct arphdr *);
501 
502 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
503 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
504 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
505 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
506 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
507 		    (unsigned char *)&ar->ar_hrd, "");
508 		m_freem(m);
509 		return;
510 	}
511 
512 	if (m->m_len < arphdr_len(ar)) {
513 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
514 			log(LOG_ERR, "arp: runt packet\n");
515 			m_freem(m);
516 			return;
517 		}
518 		ar = mtod(m, struct arphdr *);
519 	}
520 
521 	switch (ntohs(ar->ar_pro)) {
522 #ifdef INET
523 	case ETHERTYPE_IP:
524 		in_arpinput(m);
525 		return;
526 #endif
527 	}
528 	m_freem(m);
529 }
530 
531 #ifdef INET
532 /*
533  * ARP for Internet protocols on 10 Mb/s Ethernet.
534  * Algorithm is that given in RFC 826.
535  * In addition, a sanity check is performed on the sender
536  * protocol address, to catch impersonators.
537  * We no longer handle negotiations for use of trailer protocol:
538  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
539  * along with IP replies if we wanted trailers sent to us,
540  * and also sent them in response to IP replies.
541  * This allowed either end to announce the desire to receive
542  * trailer packets.
543  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
544  * but formerly didn't normally send requests.
545  */
546 static int log_arp_wrong_iface = 1;
547 static int log_arp_movements = 1;
548 static int log_arp_permanent_modify = 1;
549 
550 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
551 	&log_arp_wrong_iface, 0,
552 	"log arp packets arriving on the wrong interface");
553 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
554         &log_arp_movements, 0,
555         "log arp replies from MACs different than the one in the cache");
556 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
557         &log_arp_permanent_modify, 0,
558         "log arp replies from MACs different than the one in the permanent arp entry");
559 
560 
561 static void
562 in_arpinput(struct mbuf *m)
563 {
564 	struct arphdr *ah;
565 	struct ifnet *ifp = m->m_pkthdr.rcvif;
566 	struct llinfo_arp *la;
567 	struct rtentry *rt;
568 	struct ifaddr *ifa;
569 	struct in_ifaddr *ia;
570 	struct sockaddr_dl *sdl;
571 	struct sockaddr sa;
572 	struct in_addr isaddr, itaddr, myaddr;
573 	struct mbuf *hold;
574 	u_int8_t *enaddr = NULL;
575 	int op, rif_len;
576 	int req_len;
577 	int bridged = 0;
578 #ifdef DEV_CARP
579 	int carp_match = 0;
580 #endif
581 
582 	if (ifp->if_bridge)
583 		bridged = 1;
584 
585 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
586 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
587 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
588 		return;
589 	}
590 
591 	ah = mtod(m, struct arphdr *);
592 	op = ntohs(ah->ar_op);
593 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
594 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
595 
596 	/*
597 	 * For a bridge, we want to check the address irrespective
598 	 * of the receive interface. (This will change slightly
599 	 * when we have clusters of interfaces).
600 	 * If the interface does not match, but the recieving interface
601 	 * is part of carp, we call carp_iamatch to see if this is a
602 	 * request for the virtual host ip.
603 	 * XXX: This is really ugly!
604 	 */
605 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
606 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
607 		    (ia->ia_ifp == ifp)) &&
608 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
609 			goto match;
610 #ifdef DEV_CARP
611 		if (ifp->if_carp != NULL &&
612 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
613 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
614 			carp_match = 1;
615 			goto match;
616 		}
617 #endif
618 	}
619 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
620 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
621 		    (ia->ia_ifp == ifp)) &&
622 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
623 			goto match;
624 	/*
625 	 * No match, use the first inet address on the receive interface
626 	 * as a dummy address for the rest of the function.
627 	 */
628 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
629 		if (ifa->ifa_addr->sa_family == AF_INET) {
630 			ia = ifatoia(ifa);
631 			goto match;
632 		}
633 	/*
634 	 * If bridging, fall back to using any inet address.
635 	 */
636 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
637 		goto drop;
638 match:
639 	if (!enaddr)
640 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
641 	myaddr = ia->ia_addr.sin_addr;
642 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
643 		goto drop;	/* it's from me, ignore it. */
644 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
645 		log(LOG_ERR,
646 		    "arp: link address is broadcast for IP address %s!\n",
647 		    inet_ntoa(isaddr));
648 		goto drop;
649 	}
650 	/*
651 	 * Warn if another host is using the same IP address, but only if the
652 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
653 	 * case we suppress the warning to avoid false positive complaints of
654 	 * potential misconfiguration.
655 	 */
656 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
657 		log(LOG_ERR,
658 		   "arp: %*D is using my IP address %s!\n",
659 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
660 		   inet_ntoa(isaddr));
661 		itaddr = myaddr;
662 		goto reply;
663 	}
664 	if (ifp->if_flags & IFF_STATICARP)
665 		goto reply;
666 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
667 	if (rt != NULL) {
668 		la = (struct llinfo_arp *)rt->rt_llinfo;
669 		if (la == NULL) {
670 			RT_UNLOCK(rt);
671 			goto reply;
672 		}
673 	} else
674 		goto reply;
675 
676 	/* The following is not an error when doing bridging. */
677 	if (!bridged && rt->rt_ifp != ifp
678 #ifdef DEV_CARP
679 	    && (ifp->if_type != IFT_CARP || !carp_match)
680 #endif
681 							) {
682 		if (log_arp_wrong_iface)
683 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
684 			    inet_ntoa(isaddr),
685 			    rt->rt_ifp->if_xname,
686 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
687 			    ifp->if_xname);
688 		RT_UNLOCK(rt);
689 		goto reply;
690 	}
691 	sdl = SDL(rt->rt_gateway);
692 	if (sdl->sdl_alen &&
693 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
694 		if (rt->rt_expire) {
695 		    if (log_arp_movements)
696 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
697 			    inet_ntoa(isaddr),
698 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
699 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
700 			    ifp->if_xname);
701 		} else {
702 			RT_UNLOCK(rt);
703 			if (log_arp_permanent_modify)
704 				log(LOG_ERR, "arp: %*D attempts to modify "
705 				    "permanent entry for %s on %s\n",
706 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
707 				    inet_ntoa(isaddr), ifp->if_xname);
708 			goto reply;
709 		}
710 	}
711 	/*
712 	 * sanity check for the address length.
713 	 * XXX this does not work for protocols with variable address
714 	 * length. -is
715 	 */
716 	if (sdl->sdl_alen &&
717 	    sdl->sdl_alen != ah->ar_hln) {
718 		log(LOG_WARNING,
719 		    "arp from %*D: new addr len %d, was %d",
720 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
721 		    ah->ar_hln, sdl->sdl_alen);
722 	}
723 	if (ifp->if_addrlen != ah->ar_hln) {
724 		log(LOG_WARNING,
725 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
726 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
727 		    ah->ar_hln, ifp->if_addrlen);
728 		RT_UNLOCK(rt);
729 		goto reply;
730 	}
731 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
732 	    sdl->sdl_alen = ah->ar_hln);
733 	/*
734 	 * If we receive an arp from a token-ring station over
735 	 * a token-ring nic then try to save the source
736 	 * routing info.
737 	 */
738 	if (ifp->if_type == IFT_ISO88025) {
739 		struct iso88025_header *th = NULL;
740 		struct iso88025_sockaddr_dl_data *trld;
741 
742 		th = (struct iso88025_header *)m->m_pkthdr.header;
743 		trld = SDL_ISO88025(sdl);
744 		rif_len = TR_RCF_RIFLEN(th->rcf);
745 		if ((th->iso88025_shost[0] & TR_RII) &&
746 		    (rif_len > 2)) {
747 			trld->trld_rcf = th->rcf;
748 			trld->trld_rcf ^= htons(TR_RCF_DIR);
749 			memcpy(trld->trld_route, th->rd, rif_len - 2);
750 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
751 			/*
752 			 * Set up source routing information for
753 			 * reply packet (XXX)
754 			 */
755 			m->m_data -= rif_len;
756 			m->m_len  += rif_len;
757 			m->m_pkthdr.len += rif_len;
758 		} else {
759 			th->iso88025_shost[0] &= ~TR_RII;
760 			trld->trld_rcf = 0;
761 		}
762 		m->m_data -= 8;
763 		m->m_len  += 8;
764 		m->m_pkthdr.len += 8;
765 		th->rcf = trld->trld_rcf;
766 	}
767 	if (rt->rt_expire) {
768 		rt->rt_expire = time_uptime + arpt_keep;
769 		callout_reset(&la->la_timer, hz * arpt_keep, arptimer, rt);
770 	}
771 	la->la_asked = 0;
772 	la->la_preempt = arp_maxtries;
773 	hold = la->la_hold;
774 	la->la_hold = NULL;
775 	RT_UNLOCK(rt);
776 	if (hold != NULL)
777 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
778 
779 reply:
780 	if (op != ARPOP_REQUEST)
781 		goto drop;
782 	if (itaddr.s_addr == myaddr.s_addr) {
783 		/* I am the target */
784 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
785 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
786 	} else {
787 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
788 		if (rt == NULL) {
789 			struct sockaddr_in sin;
790 
791 			if (!arp_proxyall)
792 				goto drop;
793 
794 			bzero(&sin, sizeof sin);
795 			sin.sin_family = AF_INET;
796 			sin.sin_len = sizeof sin;
797 			sin.sin_addr = itaddr;
798 
799 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
800 			if (!rt)
801 				goto drop;
802 			/*
803 			 * Don't send proxies for nodes on the same interface
804 			 * as this one came out of, or we'll get into a fight
805 			 * over who claims what Ether address.
806 			 */
807 			if (rt->rt_ifp == ifp) {
808 				rtfree(rt);
809 				goto drop;
810 			}
811 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
812 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
813 			rtfree(rt);
814 
815 			/*
816 			 * Also check that the node which sent the ARP packet
817 			 * is on the the interface we expect it to be on. This
818 			 * avoids ARP chaos if an interface is connected to the
819 			 * wrong network.
820 			 */
821 			sin.sin_addr = isaddr;
822 
823 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
824 			if (!rt)
825 				goto drop;
826 			if (rt->rt_ifp != ifp) {
827 				log(LOG_INFO, "arp_proxy: ignoring request"
828 				    " from %s via %s, expecting %s\n",
829 				    inet_ntoa(isaddr), ifp->if_xname,
830 				    rt->rt_ifp->if_xname);
831 				rtfree(rt);
832 				goto drop;
833 			}
834 			rtfree(rt);
835 
836 #ifdef DEBUG_PROXY
837 			printf("arp: proxying for %s\n",
838 			       inet_ntoa(itaddr));
839 #endif
840 		} else {
841 			/*
842 			 * Return proxied ARP replies only on the interface
843 			 * or bridge cluster where this network resides.
844 			 * Otherwise we may conflict with the host we are
845 			 * proxying for.
846 			 */
847 			if (rt->rt_ifp != ifp &&
848 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
849 			    ifp->if_bridge == NULL)) {
850 				RT_UNLOCK(rt);
851 				goto drop;
852 			}
853 			sdl = SDL(rt->rt_gateway);
854 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
855 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
856 			RT_UNLOCK(rt);
857 		}
858 	}
859 
860 	if (itaddr.s_addr == myaddr.s_addr &&
861 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
862 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
863 #ifdef DEBUG_LINKLOCAL
864 		printf("arp: sending reply for link-local addr %s\n",
865 		    inet_ntoa(itaddr));
866 #endif
867 		m->m_flags |= M_BCAST;
868 		m->m_flags &= ~M_MCAST;
869 	} else {
870 		/* default behaviour; never reply by broadcast. */
871 		m->m_flags &= ~(M_BCAST|M_MCAST);
872 	}
873 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
874 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
875 	ah->ar_op = htons(ARPOP_REPLY);
876 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
877 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
878 	m->m_pkthdr.len = m->m_len;
879 	sa.sa_family = AF_ARP;
880 	sa.sa_len = 2;
881 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
882 	return;
883 
884 drop:
885 	m_freem(m);
886 }
887 #endif
888 
889 /*
890  * Lookup or enter a new address in arptab.
891  */
892 static struct rtentry *
893 arplookup(u_long addr, int create, int proxy)
894 {
895 	struct rtentry *rt;
896 	struct sockaddr_inarp sin;
897 	const char *why = 0;
898 
899 	bzero(&sin, sizeof(sin));
900 	sin.sin_len = sizeof(sin);
901 	sin.sin_family = AF_INET;
902 	sin.sin_addr.s_addr = addr;
903 	if (proxy)
904 		sin.sin_other = SIN_PROXY;
905 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
906 	if (rt == 0)
907 		return (0);
908 
909 	if (rt->rt_flags & RTF_GATEWAY)
910 		why = "host is not on local network";
911 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
912 		why = "could not allocate llinfo";
913 	else if (rt->rt_gateway->sa_family != AF_LINK)
914 		why = "gateway route is not ours";
915 
916 	if (why) {
917 #define	ISDYNCLONE(_rt) \
918 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
919 		if (create)
920 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
921 			    inet_ntoa(sin.sin_addr), why);
922 		/*
923 		 * If there are no references to this Layer 2 route,
924 		 * and it is a cloned route, and not static, and
925 		 * arplookup() is creating the route, then purge
926 		 * it from the routing table as it is probably bogus.
927 		 */
928 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
929 			rtexpunge(rt);
930 		RTFREE_LOCKED(rt);
931 		return (0);
932 #undef ISDYNCLONE
933 	} else {
934 		RT_REMREF(rt);
935 		return (rt);
936 	}
937 }
938 
939 void
940 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
941 {
942 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
943 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
944 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
945 	ifa->ifa_rtrequest = arp_rtrequest;
946 	ifa->ifa_flags |= RTF_CLONING;
947 }
948 
949 void
950 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
951 {
952 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
953 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
954 				&IA_SIN(ifa)->sin_addr, enaddr);
955 	ifa->ifa_rtrequest = arp_rtrequest;
956 	ifa->ifa_flags |= RTF_CLONING;
957 }
958 
959 static void
960 arp_init(void)
961 {
962 
963 	arpintrq.ifq_maxlen = 50;
964 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
965 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
966 }
967 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
968