xref: /freebsd/sys/netinet/if_ether.c (revision d429ea332342fcb98d27a350d0c4944bf9aec3f9)
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  * $FreeBSD$
31  */
32 
33 /*
34  * Ethernet address resolution protocol.
35  * TODO:
36  *	add "inuse/lock" bit (or ref. count) along with valid bit
37  */
38 
39 #include "opt_inet.h"
40 #include "opt_bdg.h"
41 #include "opt_mac.h"
42 #include "opt_carp.h"
43 
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/queue.h>
47 #include <sys/sysctl.h>
48 #include <sys/systm.h>
49 #include <sys/mac.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 #include <net/bridge.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <netinet/if_ether.h>
67 
68 #include <net/if_arc.h>
69 #include <net/iso88025.h>
70 
71 #ifdef DEV_CARP
72 #include <netinet/ip_carp.h>
73 #endif
74 
75 #define SIN(s) ((struct sockaddr_in *)s)
76 #define SDL(s) ((struct sockaddr_dl *)s)
77 
78 SYSCTL_DECL(_net_link_ether);
79 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
80 
81 /* timer values */
82 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
83 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
84 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
85 
86 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
87 	   &arpt_prune, 0, "");
88 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
89 	   &arpt_keep, 0, "");
90 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
91 	   &arpt_down, 0, "");
92 
93 #define	rt_expire rt_rmx.rmx_expire
94 
95 struct llinfo_arp {
96 	LIST_ENTRY(llinfo_arp) la_le;
97 	struct	rtentry *la_rt;
98 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
99 	u_short	la_preempt;	/* countdown for pre-expiry arps */
100 	u_short	la_asked;	/* #times we QUERIED following expiration */
101 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
102 };
103 
104 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
105 
106 static struct	ifqueue arpintrq;
107 static int	arp_allocated;
108 
109 static int	arp_maxtries = 5;
110 static int	useloopback = 1; /* use loopback interface for local traffic */
111 static int	arp_proxyall = 0;
112 static struct callout arp_callout;
113 
114 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
115 	   &arp_maxtries, 0, "");
116 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
117 	   &useloopback, 0, "");
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
119 	   &arp_proxyall, 0, "");
120 
121 static void	arp_init(void);
122 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
123 static void	arprequest(struct ifnet *,
124 			struct in_addr *, struct in_addr *, u_char *);
125 static void	arpintr(struct mbuf *);
126 static void	arptfree(struct llinfo_arp *);
127 static void	arptimer(void *);
128 static struct llinfo_arp
129 		*arplookup(u_long, int, int);
130 #ifdef INET
131 static void	in_arpinput(struct mbuf *);
132 #endif
133 
134 /*
135  * Timeout routine.  Age arp_tab entries periodically.
136  */
137 /* ARGSUSED */
138 static void
139 arptimer(ignored_arg)
140 	void *ignored_arg;
141 {
142 	struct llinfo_arp *la, *ola;
143 
144 	RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]);
145 	la = LIST_FIRST(&llinfo_arp);
146 	while (la != NULL) {
147 		struct rtentry *rt = la->la_rt;
148 		ola = la;
149 		la = LIST_NEXT(la, la_le);
150 		if (rt->rt_expire && rt->rt_expire <= time_second)
151 			arptfree(ola);		/* timer has expired, clear */
152 	}
153 	RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]);
154 
155 	callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL);
156 }
157 
158 /*
159  * Parallel to llc_rtrequest.
160  */
161 static void
162 arp_rtrequest(req, rt, info)
163 	int req;
164 	struct rtentry *rt;
165 	struct rt_addrinfo *info;
166 {
167 	struct sockaddr *gate;
168 	struct llinfo_arp *la;
169 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
170 	struct in_ifaddr *ia;
171 	struct ifaddr *ifa;
172 
173 	RT_LOCK_ASSERT(rt);
174 
175 	if (rt->rt_flags & RTF_GATEWAY)
176 		return;
177 	gate = rt->rt_gateway;
178 	la = (struct llinfo_arp *)rt->rt_llinfo;
179 	switch (req) {
180 
181 	case RTM_ADD:
182 		/*
183 		 * XXX: If this is a manually added route to interface
184 		 * such as older version of routed or gated might provide,
185 		 * restore cloning bit.
186 		 */
187 		if ((rt->rt_flags & RTF_HOST) == 0 &&
188 		    rt_mask(rt) != NULL &&
189 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
190 			rt->rt_flags |= RTF_CLONING;
191 		if (rt->rt_flags & RTF_CLONING) {
192 			/*
193 			 * Case 1: This route should come from a route to iface.
194 			 */
195 			rt_setgate(rt, rt_key(rt),
196 					(struct sockaddr *)&null_sdl);
197 			gate = rt->rt_gateway;
198 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
199 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
200 			rt->rt_expire = time_second;
201 			break;
202 		}
203 		/* Announce a new entry if requested. */
204 		if (rt->rt_flags & RTF_ANNOUNCE)
205 			arprequest(rt->rt_ifp,
206 			    &SIN(rt_key(rt))->sin_addr,
207 			    &SIN(rt_key(rt))->sin_addr,
208 			    (u_char *)LLADDR(SDL(gate)));
209 		/*FALLTHROUGH*/
210 	case RTM_RESOLVE:
211 		if (gate->sa_family != AF_LINK ||
212 		    gate->sa_len < sizeof(null_sdl)) {
213 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
214 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
215 			    (gate->sa_family != AF_LINK) ?
216 			    " (!AF_LINK)": "");
217 			break;
218 		}
219 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
220 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
221 		if (la != 0)
222 			break; /* This happens on a route change */
223 		/*
224 		 * Case 2:  This route may come from cloning, or a manual route
225 		 * add with a LL address.
226 		 */
227 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
228 		rt->rt_llinfo = (caddr_t)la;
229 		if (la == 0) {
230 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
231 			break;
232 		}
233 		arp_allocated++;
234 		la->la_rt = rt;
235 		rt->rt_flags |= RTF_LLINFO;
236 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
237 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
238 
239 #ifdef INET
240 		/*
241 		 * This keeps the multicast addresses from showing up
242 		 * in `arp -a' listings as unresolved.  It's not actually
243 		 * functional.  Then the same for broadcast.
244 		 */
245 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
246 		    rt->rt_ifp->if_type != IFT_ARCNET) {
247 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
248 					       LLADDR(SDL(gate)));
249 			SDL(gate)->sdl_alen = 6;
250 			rt->rt_expire = 0;
251 		}
252 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
253 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
254 			       rt->rt_ifp->if_addrlen);
255 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
256 			rt->rt_expire = 0;
257 		}
258 #endif
259 
260 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
261 			if (ia->ia_ifp == rt->rt_ifp &&
262 			    SIN(rt_key(rt))->sin_addr.s_addr ==
263 			    (IA_SIN(ia))->sin_addr.s_addr)
264 				break;
265 		}
266 		if (ia) {
267 		    /*
268 		     * This test used to be
269 		     *	if (loif.if_flags & IFF_UP)
270 		     * It allowed local traffic to be forced
271 		     * through the hardware by configuring the loopback down.
272 		     * However, it causes problems during network configuration
273 		     * for boards that can't receive packets they send.
274 		     * It is now necessary to clear "useloopback" and remove
275 		     * the route to force traffic out to the hardware.
276 		     */
277 			rt->rt_expire = 0;
278 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
279 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
280 			if (useloopback)
281 				rt->rt_ifp = loif;
282 
283 		    /*
284 		     * make sure to set rt->rt_ifa to the interface
285 		     * address we are using, otherwise we will have trouble
286 		     * with source address selection.
287 		     */
288 			ifa = &ia->ia_ifa;
289 			if (ifa != rt->rt_ifa) {
290 				IFAFREE(rt->rt_ifa);
291 				IFAREF(ifa);
292 				rt->rt_ifa = ifa;
293 			}
294 		}
295 		break;
296 
297 	case RTM_DELETE:
298 		if (la == 0)
299 			break;
300 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
301 		LIST_REMOVE(la, la_le);
302 		rt->rt_llinfo = 0;
303 		rt->rt_flags &= ~RTF_LLINFO;
304 		if (la->la_hold)
305 			m_freem(la->la_hold);
306 		Free((caddr_t)la);
307 	}
308 }
309 
310 /*
311  * Broadcast an ARP request. Caller specifies:
312  *	- arp header source ip address
313  *	- arp header target ip address
314  *	- arp header source ethernet address
315  */
316 static void
317 arprequest(ifp, sip, tip, enaddr)
318 	struct ifnet *ifp;
319 	struct in_addr *sip, *tip;
320 	u_char *enaddr;
321 {
322 	struct mbuf *m;
323 	struct arphdr *ah;
324 	struct sockaddr sa;
325 
326 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
327 		return;
328 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
329 		2*ifp->if_data.ifi_addrlen;
330 	m->m_pkthdr.len = m->m_len;
331 	MH_ALIGN(m, m->m_len);
332 	ah = mtod(m, struct arphdr *);
333 	bzero((caddr_t)ah, m->m_len);
334 #ifdef MAC
335 	mac_create_mbuf_linklayer(ifp, m);
336 #endif
337 	ah->ar_pro = htons(ETHERTYPE_IP);
338 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
339 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
340 	ah->ar_op = htons(ARPOP_REQUEST);
341 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
342 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
343 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
344 	sa.sa_family = AF_ARP;
345 	sa.sa_len = 2;
346 	m->m_flags |= M_BCAST;
347 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
348 
349 	return;
350 }
351 
352 /*
353  * Resolve an IP address into an ethernet address.
354  * On input:
355  *    ifp is the interface we use
356  *    dst is the next hop,
357  *    rt0 is the route to the final destination (possibly useless)
358  *    m is the mbuf
359  *    desten is where we want the address.
360  *
361  * On success, desten is filled in and the function returns 0;
362  * If the packet must be held pending resolution, we return EWOULDBLOCK
363  * On other errors, we return the corresponding error code.
364  */
365 int
366 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
367 	struct sockaddr *dst, u_char *desten)
368 {
369 	struct llinfo_arp *la = 0;
370 	struct sockaddr_dl *sdl;
371 	int error;
372 	struct rtentry *rt;
373 
374 	error = rt_check(&rt, &rt0, dst);
375 	if (error) {
376 		m_freem(m);
377 		return error;
378 	}
379 
380 	if (m->m_flags & M_BCAST) {	/* broadcast */
381 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
382 		return (0);
383 	}
384 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
385 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
386 		return (0);
387 	}
388 	if (rt)
389 		la = (struct llinfo_arp *)rt->rt_llinfo;
390 	if (la == 0) {
391 		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
392 		if (la)
393 			rt = la->la_rt;
394 	}
395 	if (la == 0 || rt == 0) {
396 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
397 			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
398 				rt ? "rt" : "");
399 		m_freem(m);
400 		return (EINVAL); /* XXX */
401 	}
402 	sdl = SDL(rt->rt_gateway);
403 	/*
404 	 * Check the address family and length is valid, the address
405 	 * is resolved; otherwise, try to resolve.
406 	 */
407 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
408 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
409 		/*
410 		 * If entry has an expiry time and it is approaching,
411 		 * see if we need to send an ARP request within this
412 		 * arpt_down interval.
413 		 */
414 		if ((rt->rt_expire != 0) &&
415 		    (time_second + la->la_preempt > rt->rt_expire)) {
416 			arprequest(ifp,
417 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
418 				   &SIN(dst)->sin_addr,
419 				   IF_LLADDR(ifp));
420 			la->la_preempt--;
421 		}
422 
423 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
424 		return (0);
425 	}
426 	/*
427 	 * If ARP is disabled or static on this interface, stop.
428 	 * XXX
429 	 * Probably should not allocate empty llinfo struct if we are
430 	 * not going to be sending out an arp request.
431 	 */
432 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
433 		m_freem(m);
434 		return (EINVAL);
435 	}
436 	/*
437 	 * There is an arptab entry, but no ethernet address
438 	 * response yet.  Replace the held mbuf with this
439 	 * latest one.
440 	 */
441 	if (la->la_hold)
442 		m_freem(la->la_hold);
443 	la->la_hold = m;
444 	if (rt->rt_expire) {
445 		RT_LOCK(rt);
446 		rt->rt_flags &= ~RTF_REJECT;
447 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
448 			rt->rt_expire = time_second;
449 			if (la->la_asked++ < arp_maxtries) {
450 				arprequest(ifp,
451 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
452 					   &SIN(dst)->sin_addr,
453 					   IF_LLADDR(ifp));
454 			} else {
455 				rt->rt_flags |= RTF_REJECT;
456 				rt->rt_expire += arpt_down;
457 				la->la_asked = 0;
458 				la->la_preempt = arp_maxtries;
459 			}
460 
461 		}
462 		RT_UNLOCK(rt);
463 	}
464 	return (EWOULDBLOCK);
465 }
466 
467 /*
468  * Common length and type checks are done here,
469  * then the protocol-specific routine is called.
470  */
471 static void
472 arpintr(struct mbuf *m)
473 {
474 	struct arphdr *ar;
475 
476 	if (m->m_len < sizeof(struct arphdr) &&
477 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
478 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
479 		return;
480 	}
481 	ar = mtod(m, struct arphdr *);
482 
483 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
484 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
485 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
486 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
487 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
488 		    (unsigned char *)&ar->ar_hrd, "");
489 		m_freem(m);
490 		return;
491 	}
492 
493 	if (m->m_len < arphdr_len(ar)) {
494 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
495 			log(LOG_ERR, "arp: runt packet\n");
496 			m_freem(m);
497 			return;
498 		}
499 		ar = mtod(m, struct arphdr *);
500 	}
501 
502 	switch (ntohs(ar->ar_pro)) {
503 #ifdef INET
504 	case ETHERTYPE_IP:
505 		in_arpinput(m);
506 		return;
507 #endif
508 	}
509 	m_freem(m);
510 }
511 
512 #ifdef INET
513 /*
514  * ARP for Internet protocols on 10 Mb/s Ethernet.
515  * Algorithm is that given in RFC 826.
516  * In addition, a sanity check is performed on the sender
517  * protocol address, to catch impersonators.
518  * We no longer handle negotiations for use of trailer protocol:
519  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
520  * along with IP replies if we wanted trailers sent to us,
521  * and also sent them in response to IP replies.
522  * This allowed either end to announce the desire to receive
523  * trailer packets.
524  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
525  * but formerly didn't normally send requests.
526  */
527 static int log_arp_wrong_iface = 1;
528 static int log_arp_movements = 1;
529 
530 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
531 	&log_arp_wrong_iface, 0,
532 	"log arp packets arriving on the wrong interface");
533 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
534         &log_arp_movements, 0,
535         "log arp replies from MACs different than the one in the cache");
536 
537 
538 static void
539 in_arpinput(m)
540 	struct mbuf *m;
541 {
542 	struct arphdr *ah;
543 	struct ifnet *ifp = m->m_pkthdr.rcvif;
544 	struct iso88025_header *th = (struct iso88025_header *)0;
545 	struct iso88025_sockaddr_dl_data *trld;
546 	struct llinfo_arp *la = 0;
547 	struct rtentry *rt;
548 	struct ifaddr *ifa;
549 	struct in_ifaddr *ia;
550 	struct sockaddr_dl *sdl;
551 	struct sockaddr sa;
552 	struct in_addr isaddr, itaddr, myaddr;
553 	u_int8_t *enaddr = NULL;
554 	int op, rif_len;
555 	int req_len;
556 	int bridged = 0;
557 #ifdef DEV_CARP
558 	int carp_match = 0;
559 #endif
560 
561 	if (do_bridge || ifp->if_bridge)
562 		bridged = 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 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
567 		return;
568 	}
569 
570 	ah = mtod(m, struct arphdr *);
571 	op = ntohs(ah->ar_op);
572 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
573 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
574 
575 	/*
576 	 * For a bridge, we want to check the address irrespective
577 	 * of the receive interface. (This will change slightly
578 	 * when we have clusters of interfaces).
579 	 * If the interface does not match, but the recieving interface
580 	 * is part of carp, we call carp_iamatch to see if this is a
581 	 * request for the virtual host ip.
582 	 * XXX: This is really ugly!
583 	 */
584 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
585 		if ((bridged || (ia->ia_ifp == ifp)) &&
586 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
587 			goto match;
588 #ifdef DEV_CARP
589 		if (ifp->if_carp != NULL &&
590 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
591 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
592 			carp_match = 1;
593 			goto match;
594 		}
595 #endif
596 	}
597 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
598 		if ((bridged || (ia->ia_ifp == ifp)) &&
599 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
600 			goto match;
601 	/*
602 	 * No match, use the first inet address on the receive interface
603 	 * as a dummy address for the rest of the function.
604 	 */
605 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
606 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
607 			ia = ifatoia(ifa);
608 			goto match;
609 		}
610 	/*
611 	 * If bridging, fall back to using any inet address.
612 	 */
613 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
614 		goto drop;
615 match:
616 	if (!enaddr)
617 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
618 	myaddr = ia->ia_addr.sin_addr;
619 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
620 		goto drop;	/* it's from me, ignore it. */
621 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
622 		log(LOG_ERR,
623 		    "arp: link address is broadcast for IP address %s!\n",
624 		    inet_ntoa(isaddr));
625 		goto drop;
626 	}
627 	/*
628 	 * Warn if another host is using the same IP address, but only if the
629 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
630 	 * case we suppress the warning to avoid false positive complaints of
631 	 * potential misconfiguration.
632 	 */
633 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
634 		log(LOG_ERR,
635 		   "arp: %*D is using my IP address %s!\n",
636 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
637 		   inet_ntoa(isaddr));
638 		itaddr = myaddr;
639 		goto reply;
640 	}
641 	if (ifp->if_flags & IFF_STATICARP)
642 		goto reply;
643 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
644 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
645 		/* the following is not an error when doing bridging */
646 		if (!bridged && rt->rt_ifp != ifp
647 #ifdef DEV_CARP
648 		    && (ifp->if_type != IFT_CARP || !carp_match)
649 #endif
650 								) {
651 			if (log_arp_wrong_iface)
652 				log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
653 				    inet_ntoa(isaddr),
654 				    rt->rt_ifp->if_xname,
655 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
656 				    ifp->if_xname);
657 			goto reply;
658 		}
659 		if (sdl->sdl_alen &&
660 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
661 			if (rt->rt_expire) {
662 			    if (log_arp_movements)
663 			        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
664 				    inet_ntoa(isaddr),
665 				    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
666 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
667 				    ifp->if_xname);
668 			} else {
669 			    log(LOG_ERR,
670 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
671 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
672 				inet_ntoa(isaddr), ifp->if_xname);
673 			    goto reply;
674 			}
675 		}
676 		/*
677 		 * sanity check for the address length.
678 		 * XXX this does not work for protocols with variable address
679 		 * length. -is
680 		 */
681 		if (sdl->sdl_alen &&
682 		    sdl->sdl_alen != ah->ar_hln) {
683 			log(LOG_WARNING,
684 			    "arp from %*D: new addr len %d, was %d",
685 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
686 			    ah->ar_hln, sdl->sdl_alen);
687 		}
688 		if (ifp->if_addrlen != ah->ar_hln) {
689 			log(LOG_WARNING,
690 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
691 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
692 			    ah->ar_hln, ifp->if_addrlen);
693 			goto reply;
694 		}
695 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
696 		    sdl->sdl_alen = ah->ar_hln);
697 		/*
698 		 * If we receive an arp from a token-ring station over
699 		 * a token-ring nic then try to save the source
700 		 * routing info.
701 		 */
702 		if (ifp->if_type == IFT_ISO88025) {
703 			th = (struct iso88025_header *)m->m_pkthdr.header;
704 			trld = SDL_ISO88025(sdl);
705 			rif_len = TR_RCF_RIFLEN(th->rcf);
706 			if ((th->iso88025_shost[0] & TR_RII) &&
707 			    (rif_len > 2)) {
708 				trld->trld_rcf = th->rcf;
709 				trld->trld_rcf ^= htons(TR_RCF_DIR);
710 				memcpy(trld->trld_route, th->rd, rif_len - 2);
711 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
712 				/*
713 				 * Set up source routing information for
714 				 * reply packet (XXX)
715 				 */
716 				m->m_data -= rif_len;
717 				m->m_len  += rif_len;
718 				m->m_pkthdr.len += rif_len;
719 			} else {
720 				th->iso88025_shost[0] &= ~TR_RII;
721 				trld->trld_rcf = 0;
722 			}
723 			m->m_data -= 8;
724 			m->m_len  += 8;
725 			m->m_pkthdr.len += 8;
726 			th->rcf = trld->trld_rcf;
727 		}
728 		RT_LOCK(rt);
729 		if (rt->rt_expire)
730 			rt->rt_expire = time_second + arpt_keep;
731 		rt->rt_flags &= ~RTF_REJECT;
732 		RT_UNLOCK(rt);
733 		la->la_asked = 0;
734 		la->la_preempt = arp_maxtries;
735 		if (la->la_hold) {
736 			(*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt);
737 			la->la_hold = 0;
738 		}
739 	}
740 reply:
741 	if (op != ARPOP_REQUEST)
742 		goto drop;
743 	if (itaddr.s_addr == myaddr.s_addr) {
744 		/* I am the target */
745 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
746 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
747 	} else {
748 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
749 		if (la == NULL) {
750 			struct sockaddr_in sin;
751 
752 			if (!arp_proxyall)
753 				goto drop;
754 
755 			bzero(&sin, sizeof sin);
756 			sin.sin_family = AF_INET;
757 			sin.sin_len = sizeof sin;
758 			sin.sin_addr = itaddr;
759 
760 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
761 			if (!rt)
762 				goto drop;
763 			/*
764 			 * Don't send proxies for nodes on the same interface
765 			 * as this one came out of, or we'll get into a fight
766 			 * over who claims what Ether address.
767 			 */
768 			if (rt->rt_ifp == ifp) {
769 				rtfree(rt);
770 				goto drop;
771 			}
772 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
773 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
774 			rtfree(rt);
775 
776 			/*
777 			 * Also check that the node which sent the ARP packet
778 			 * is on the the interface we expect it to be on. This
779 			 * avoids ARP chaos if an interface is connected to the
780 			 * wrong network.
781 			 */
782 			sin.sin_addr = isaddr;
783 
784 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
785 			if (!rt)
786 				goto drop;
787 			if (rt->rt_ifp != ifp) {
788 				log(LOG_INFO, "arp_proxy: ignoring request"
789 				    " from %s via %s, expecting %s\n",
790 				    inet_ntoa(isaddr), ifp->if_xname,
791 				    rt->rt_ifp->if_xname);
792 				rtfree(rt);
793 				goto drop;
794 			}
795 			rtfree(rt);
796 
797 #ifdef DEBUG_PROXY
798 			printf("arp: proxying for %s\n",
799 			       inet_ntoa(itaddr));
800 #endif
801 		} else {
802 			rt = la->la_rt;
803 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
804 			sdl = SDL(rt->rt_gateway);
805 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
806 		}
807 	}
808 
809 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
810 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
811 	ah->ar_op = htons(ARPOP_REPLY);
812 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
813 	m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
814 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
815 	m->m_pkthdr.len = m->m_len;
816 	sa.sa_family = AF_ARP;
817 	sa.sa_len = 2;
818 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
819 	return;
820 
821 drop:
822 	m_freem(m);
823 }
824 #endif
825 
826 /*
827  * Free an arp entry.
828  */
829 static void
830 arptfree(la)
831 	struct llinfo_arp *la;
832 {
833 	struct rtentry *rt = la->la_rt;
834 	struct sockaddr_dl *sdl;
835 
836 	if (rt == 0)
837 		panic("arptfree");
838 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
839 	    sdl->sdl_family == AF_LINK) {
840 		sdl->sdl_alen = 0;
841 		la->la_preempt = la->la_asked = 0;
842 		RT_LOCK(rt);		/* XXX needed or move higher? */
843 		rt->rt_flags &= ~RTF_REJECT;
844 		RT_UNLOCK(rt);
845 		return;
846 	}
847 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
848 			0, (struct rtentry **)0);
849 }
850 /*
851  * Lookup or enter a new address in arptab.
852  */
853 static struct llinfo_arp *
854 arplookup(addr, create, proxy)
855 	u_long addr;
856 	int create, proxy;
857 {
858 	struct rtentry *rt;
859 	struct sockaddr_inarp sin;
860 	const char *why = 0;
861 
862 	bzero(&sin, sizeof(sin));
863 	sin.sin_len = sizeof(sin);
864 	sin.sin_family = AF_INET;
865 	sin.sin_addr.s_addr = addr;
866 	if (proxy)
867 		sin.sin_other = SIN_PROXY;
868 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
869 	if (rt == 0)
870 		return (0);
871 
872 	if (rt->rt_flags & RTF_GATEWAY)
873 		why = "host is not on local network";
874 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
875 		why = "could not allocate llinfo";
876 	else if (rt->rt_gateway->sa_family != AF_LINK)
877 		why = "gateway route is not ours";
878 
879 	if (why) {
880 #define	ISDYNCLONE(_rt) \
881 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
882 		if (create)
883 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
884 			    inet_ntoa(sin.sin_addr), why);
885 		/*
886 		 * If there are no references to this Layer 2 route,
887 		 * and it is a cloned route, and not static, and
888 		 * arplookup() is creating the route, then purge
889 		 * it from the routing table as it is probably bogus.
890 		 */
891 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
892 			rtexpunge(rt);
893 		RTFREE_LOCKED(rt);
894 		return (0);
895 #undef ISDYNCLONE
896 	} else {
897 		RT_REMREF(rt);
898 		RT_UNLOCK(rt);
899 		return ((struct llinfo_arp *)rt->rt_llinfo);
900 	}
901 }
902 
903 void
904 arp_ifinit(ifp, ifa)
905 	struct ifnet *ifp;
906 	struct ifaddr *ifa;
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 	ifa->ifa_rtrequest = arp_rtrequest;
912 	ifa->ifa_flags |= RTF_CLONING;
913 }
914 
915 void
916 arp_ifinit2(ifp, ifa, enaddr)
917 	struct ifnet *ifp;
918 	struct ifaddr *ifa;
919 	u_char *enaddr;
920 {
921 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
922 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
923 				&IA_SIN(ifa)->sin_addr, enaddr);
924 	ifa->ifa_rtrequest = arp_rtrequest;
925 	ifa->ifa_flags |= RTF_CLONING;
926 }
927 
928 static void
929 arp_init(void)
930 {
931 
932 	arpintrq.ifq_maxlen = 50;
933 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
934 	LIST_INIT(&llinfo_arp);
935 	callout_init(&arp_callout, CALLOUT_MPSAFE);
936 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
937 	callout_reset(&arp_callout, hz, arptimer, NULL);
938 }
939 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
940