xref: /freebsd/sys/netinet/if_ether.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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_netinet_arp_send(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  *    rt0 is the route to the final destination (possibly useless)
348  *    m is the mbuf. May be NULL if we don't have a packet.
349  *    dst is the next hop,
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  * Note that m_freem() handles NULL.
356  */
357 int
358 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
359     struct sockaddr *dst, u_char *desten)
360 {
361 	struct llinfo_arp *la = NULL;
362 	struct rtentry *rt = NULL;
363 	struct sockaddr_dl *sdl;
364 	int error;
365 
366 	if (m) {
367 		if (m->m_flags & M_BCAST) {
368 			/* broadcast */
369 			(void)memcpy(desten,
370 			    ifp->if_broadcastaddr, ifp->if_addrlen);
371 			return (0);
372 		}
373 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
374 			/* multicast */
375 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
376 			return (0);
377 		}
378 	}
379 
380 	if (rt0 != NULL) {
381 		error = rt_check(&rt, &rt0, dst);
382 		if (error) {
383 			m_freem(m);
384 			return error;
385 		}
386 		la = (struct llinfo_arp *)rt->rt_llinfo;
387 		if (la == NULL)
388 			RT_UNLOCK(rt);
389 	}
390 	if (la == NULL) {
391 		/*
392 		 * We enter this block in case if rt0 was NULL,
393 		 * or if rt found by rt_check() didn't have llinfo.
394 		 */
395 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
396 		if (rt == NULL) {
397 			log(LOG_DEBUG,
398 			    "arpresolve: can't allocate route for %s\n",
399 			    inet_ntoa(SIN(dst)->sin_addr));
400 			m_freem(m);
401 			return (EINVAL); /* XXX */
402 		}
403 		la = (struct llinfo_arp *)rt->rt_llinfo;
404 		if (la == NULL) {
405 			RT_UNLOCK(rt);
406 			log(LOG_DEBUG,
407 			    "arpresolve: can't allocate llinfo for %s\n",
408 			    inet_ntoa(SIN(dst)->sin_addr));
409 			m_freem(m);
410 			return (EINVAL); /* XXX */
411 		}
412 	}
413 	sdl = SDL(rt->rt_gateway);
414 	/*
415 	 * Check the address family and length is valid, the address
416 	 * is resolved; otherwise, try to resolve.
417 	 */
418 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
419 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
420 
421 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
422 
423 		/*
424 		 * If entry has an expiry time and it is approaching,
425 		 * send an ARP request.
426 		 */
427 		if ((rt->rt_expire != 0) &&
428 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
429 			struct in_addr sin =
430 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
431 
432 			la->la_preempt--;
433 			RT_UNLOCK(rt);
434 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
435 			    IF_LLADDR(ifp));
436 			return (0);
437 		}
438 
439 		RT_UNLOCK(rt);
440 		return (0);
441 	}
442 	/*
443 	 * If ARP is disabled or static on this interface, stop.
444 	 * XXX
445 	 * Probably should not allocate empty llinfo struct if we are
446 	 * not going to be sending out an arp request.
447 	 */
448 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
449 		RT_UNLOCK(rt);
450 		m_freem(m);
451 		return (EINVAL);
452 	}
453 	/*
454 	 * There is an arptab entry, but no ethernet address
455 	 * response yet.  Replace the held mbuf with this
456 	 * latest one.
457 	 */
458 	if (m) {
459 		if (la->la_hold)
460 			m_freem(la->la_hold);
461 		la->la_hold = m;
462 	}
463 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
464 
465 	/*
466 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
467 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
468 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
469 	 * ARP request, but not faster than one request per second.
470 	 */
471 	if (la->la_asked < arp_maxtries)
472 		error = EWOULDBLOCK;	/* First request. */
473 	else
474 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
475 
476 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
477 		struct in_addr sin =
478 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
479 
480 		rt->rt_expire = time_uptime;
481 		callout_reset(&la->la_timer, hz, arptimer, rt);
482 		la->la_asked++;
483 		RT_UNLOCK(rt);
484 
485 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
486 		    IF_LLADDR(ifp));
487 	} else
488 		RT_UNLOCK(rt);
489 
490 	return (error);
491 }
492 
493 /*
494  * Common length and type checks are done here,
495  * then the protocol-specific routine is called.
496  */
497 static void
498 arpintr(struct mbuf *m)
499 {
500 	struct arphdr *ar;
501 
502 	if (m->m_len < sizeof(struct arphdr) &&
503 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
504 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
505 		return;
506 	}
507 	ar = mtod(m, struct arphdr *);
508 
509 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
510 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
511 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
512 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
513 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
514 		    (unsigned char *)&ar->ar_hrd, "");
515 		m_freem(m);
516 		return;
517 	}
518 
519 	if (m->m_len < arphdr_len(ar)) {
520 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
521 			log(LOG_ERR, "arp: runt packet\n");
522 			m_freem(m);
523 			return;
524 		}
525 		ar = mtod(m, struct arphdr *);
526 	}
527 
528 	switch (ntohs(ar->ar_pro)) {
529 #ifdef INET
530 	case ETHERTYPE_IP:
531 		in_arpinput(m);
532 		return;
533 #endif
534 	}
535 	m_freem(m);
536 }
537 
538 #ifdef INET
539 /*
540  * ARP for Internet protocols on 10 Mb/s Ethernet.
541  * Algorithm is that given in RFC 826.
542  * In addition, a sanity check is performed on the sender
543  * protocol address, to catch impersonators.
544  * We no longer handle negotiations for use of trailer protocol:
545  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
546  * along with IP replies if we wanted trailers sent to us,
547  * and also sent them in response to IP replies.
548  * This allowed either end to announce the desire to receive
549  * trailer packets.
550  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
551  * but formerly didn't normally send requests.
552  */
553 static int log_arp_wrong_iface = 1;
554 static int log_arp_movements = 1;
555 static int log_arp_permanent_modify = 1;
556 
557 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
558 	&log_arp_wrong_iface, 0,
559 	"log arp packets arriving on the wrong interface");
560 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
561         &log_arp_movements, 0,
562         "log arp replies from MACs different than the one in the cache");
563 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
564         &log_arp_permanent_modify, 0,
565         "log arp replies from MACs different than the one in the permanent arp entry");
566 
567 
568 static void
569 in_arpinput(struct mbuf *m)
570 {
571 	struct arphdr *ah;
572 	struct ifnet *ifp = m->m_pkthdr.rcvif;
573 	struct llinfo_arp *la;
574 	struct rtentry *rt;
575 	struct ifaddr *ifa;
576 	struct in_ifaddr *ia;
577 	struct sockaddr_dl *sdl;
578 	struct sockaddr sa;
579 	struct in_addr isaddr, itaddr, myaddr;
580 	struct mbuf *hold;
581 	u_int8_t *enaddr = NULL;
582 	int op, rif_len;
583 	int req_len;
584 	int bridged = 0;
585 #ifdef DEV_CARP
586 	int carp_match = 0;
587 #endif
588 	struct sockaddr_in sin;
589 	sin.sin_len = sizeof(struct sockaddr_in);
590 	sin.sin_family = AF_INET;
591 	sin.sin_addr.s_addr = 0;
592 
593 	if (ifp->if_bridge)
594 		bridged = 1;
595 
596 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
597 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
598 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
599 		return;
600 	}
601 
602 	ah = mtod(m, struct arphdr *);
603 	op = ntohs(ah->ar_op);
604 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
605 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
606 
607 	/*
608 	 * For a bridge, we want to check the address irrespective
609 	 * of the receive interface. (This will change slightly
610 	 * when we have clusters of interfaces).
611 	 * If the interface does not match, but the recieving interface
612 	 * is part of carp, we call carp_iamatch to see if this is a
613 	 * request for the virtual host ip.
614 	 * XXX: This is really ugly!
615 	 */
616 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
617 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
618 		    (ia->ia_ifp == ifp)) &&
619 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
620 			goto match;
621 #ifdef DEV_CARP
622 		if (ifp->if_carp != NULL &&
623 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
624 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
625 			carp_match = 1;
626 			goto match;
627 		}
628 #endif
629 	}
630 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
631 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
632 		    (ia->ia_ifp == ifp)) &&
633 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
634 			goto match;
635 	/*
636 	 * No match, use the first inet address on the receive interface
637 	 * as a dummy address for the rest of the function.
638 	 */
639 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
640 		if (ifa->ifa_addr->sa_family == AF_INET) {
641 			ia = ifatoia(ifa);
642 			goto match;
643 		}
644 	/*
645 	 * If bridging, fall back to using any inet address.
646 	 */
647 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
648 		goto drop;
649 match:
650 	if (!enaddr)
651 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
652 	myaddr = ia->ia_addr.sin_addr;
653 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
654 		goto drop;	/* it's from me, ignore it. */
655 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
656 		log(LOG_ERR,
657 		    "arp: link address is broadcast for IP address %s!\n",
658 		    inet_ntoa(isaddr));
659 		goto drop;
660 	}
661 	/*
662 	 * Warn if another host is using the same IP address, but only if the
663 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
664 	 * case we suppress the warning to avoid false positive complaints of
665 	 * potential misconfiguration.
666 	 */
667 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
668 		log(LOG_ERR,
669 		   "arp: %*D is using my IP address %s on %s!\n",
670 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
671 		   inet_ntoa(isaddr), ifp->if_xname);
672 		itaddr = myaddr;
673 		goto reply;
674 	}
675 	if (ifp->if_flags & IFF_STATICARP)
676 		goto reply;
677 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
678 	if (rt != NULL) {
679 		sin.sin_addr.s_addr = isaddr.s_addr;
680 		EVENTHANDLER_INVOKE(route_arp_update_event, rt,
681 		    ar_sha(ah), (struct sockaddr *)&sin);
682 
683 		la = (struct llinfo_arp *)rt->rt_llinfo;
684 		if (la == NULL) {
685 			RT_UNLOCK(rt);
686 			goto reply;
687 		}
688 	} else
689 		goto reply;
690 
691 	/* The following is not an error when doing bridging. */
692 	if (!bridged && rt->rt_ifp != ifp
693 #ifdef DEV_CARP
694 	    && (ifp->if_type != IFT_CARP || !carp_match)
695 #endif
696 							) {
697 		if (log_arp_wrong_iface)
698 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
699 			    inet_ntoa(isaddr),
700 			    rt->rt_ifp->if_xname,
701 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
702 			    ifp->if_xname);
703 		RT_UNLOCK(rt);
704 		goto reply;
705 	}
706 	sdl = SDL(rt->rt_gateway);
707 	if (sdl->sdl_alen &&
708 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
709 		if (rt->rt_expire) {
710 		    if (log_arp_movements)
711 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
712 			    inet_ntoa(isaddr),
713 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
714 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
715 			    ifp->if_xname);
716 		} else {
717 			RT_UNLOCK(rt);
718 			if (log_arp_permanent_modify)
719 				log(LOG_ERR, "arp: %*D attempts to modify "
720 				    "permanent entry for %s on %s\n",
721 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
722 				    inet_ntoa(isaddr), ifp->if_xname);
723 			goto reply;
724 		}
725 	}
726 	/*
727 	 * sanity check for the address length.
728 	 * XXX this does not work for protocols with variable address
729 	 * length. -is
730 	 */
731 	if (sdl->sdl_alen &&
732 	    sdl->sdl_alen != ah->ar_hln) {
733 		log(LOG_WARNING,
734 		    "arp from %*D: new addr len %d, was %d",
735 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
736 		    ah->ar_hln, sdl->sdl_alen);
737 	}
738 	if (ifp->if_addrlen != ah->ar_hln) {
739 		log(LOG_WARNING,
740 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
741 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
742 		    ah->ar_hln, ifp->if_addrlen);
743 		RT_UNLOCK(rt);
744 		goto reply;
745 	}
746 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
747 	    sdl->sdl_alen = ah->ar_hln);
748 	/*
749 	 * If we receive an arp from a token-ring station over
750 	 * a token-ring nic then try to save the source
751 	 * routing info.
752 	 */
753 	if (ifp->if_type == IFT_ISO88025) {
754 		struct iso88025_header *th = NULL;
755 		struct iso88025_sockaddr_dl_data *trld;
756 
757 		th = (struct iso88025_header *)m->m_pkthdr.header;
758 		trld = SDL_ISO88025(sdl);
759 		rif_len = TR_RCF_RIFLEN(th->rcf);
760 		if ((th->iso88025_shost[0] & TR_RII) &&
761 		    (rif_len > 2)) {
762 			trld->trld_rcf = th->rcf;
763 			trld->trld_rcf ^= htons(TR_RCF_DIR);
764 			memcpy(trld->trld_route, th->rd, rif_len - 2);
765 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
766 			/*
767 			 * Set up source routing information for
768 			 * reply packet (XXX)
769 			 */
770 			m->m_data -= rif_len;
771 			m->m_len  += rif_len;
772 			m->m_pkthdr.len += rif_len;
773 		} else {
774 			th->iso88025_shost[0] &= ~TR_RII;
775 			trld->trld_rcf = 0;
776 		}
777 		m->m_data -= 8;
778 		m->m_len  += 8;
779 		m->m_pkthdr.len += 8;
780 		th->rcf = trld->trld_rcf;
781 	}
782 	if (rt->rt_expire) {
783 		rt->rt_expire = time_uptime + arpt_keep;
784 		callout_reset(&la->la_timer, hz * arpt_keep, arptimer, rt);
785 	}
786 	la->la_asked = 0;
787 	la->la_preempt = arp_maxtries;
788 	hold = la->la_hold;
789 	la->la_hold = NULL;
790 	RT_UNLOCK(rt);
791 	if (hold != NULL)
792 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
793 
794 reply:
795 	if (op != ARPOP_REQUEST)
796 		goto drop;
797 	if (itaddr.s_addr == myaddr.s_addr) {
798 		/* I am the target */
799 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
800 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
801 	} else {
802 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
803 		if (rt == NULL) {
804 			struct sockaddr_in sin;
805 
806 			if (!arp_proxyall)
807 				goto drop;
808 
809 			bzero(&sin, sizeof sin);
810 			sin.sin_family = AF_INET;
811 			sin.sin_len = sizeof sin;
812 			sin.sin_addr = itaddr;
813 
814 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
815 			if (!rt)
816 				goto drop;
817 			/*
818 			 * Don't send proxies for nodes on the same interface
819 			 * as this one came out of, or we'll get into a fight
820 			 * over who claims what Ether address.
821 			 */
822 			if (rt->rt_ifp == ifp) {
823 				rtfree(rt);
824 				goto drop;
825 			}
826 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
827 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
828 			rtfree(rt);
829 
830 			/*
831 			 * Also check that the node which sent the ARP packet
832 			 * is on the the interface we expect it to be on. This
833 			 * avoids ARP chaos if an interface is connected to the
834 			 * wrong network.
835 			 */
836 			sin.sin_addr = isaddr;
837 
838 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
839 			if (!rt)
840 				goto drop;
841 			if (rt->rt_ifp != ifp) {
842 				log(LOG_INFO, "arp_proxy: ignoring request"
843 				    " from %s via %s, expecting %s\n",
844 				    inet_ntoa(isaddr), ifp->if_xname,
845 				    rt->rt_ifp->if_xname);
846 				rtfree(rt);
847 				goto drop;
848 			}
849 			rtfree(rt);
850 
851 #ifdef DEBUG_PROXY
852 			printf("arp: proxying for %s\n",
853 			       inet_ntoa(itaddr));
854 #endif
855 		} else {
856 			/*
857 			 * Return proxied ARP replies only on the interface
858 			 * or bridge cluster where this network resides.
859 			 * Otherwise we may conflict with the host we are
860 			 * proxying for.
861 			 */
862 			if (rt->rt_ifp != ifp &&
863 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
864 			    ifp->if_bridge == NULL)) {
865 				RT_UNLOCK(rt);
866 				goto drop;
867 			}
868 			sdl = SDL(rt->rt_gateway);
869 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
870 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
871 			RT_UNLOCK(rt);
872 		}
873 	}
874 
875 	if (itaddr.s_addr == myaddr.s_addr &&
876 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
877 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
878 #ifdef DEBUG_LINKLOCAL
879 		printf("arp: sending reply for link-local addr %s\n",
880 		    inet_ntoa(itaddr));
881 #endif
882 		m->m_flags |= M_BCAST;
883 		m->m_flags &= ~M_MCAST;
884 	} else {
885 		/* default behaviour; never reply by broadcast. */
886 		m->m_flags &= ~(M_BCAST|M_MCAST);
887 	}
888 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
889 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
890 	ah->ar_op = htons(ARPOP_REPLY);
891 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
892 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
893 	m->m_pkthdr.len = m->m_len;
894 	sa.sa_family = AF_ARP;
895 	sa.sa_len = 2;
896 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
897 	return;
898 
899 drop:
900 	m_freem(m);
901 }
902 #endif
903 
904 /*
905  * Lookup or enter a new address in arptab.
906  */
907 static struct rtentry *
908 arplookup(u_long addr, int create, int proxy)
909 {
910 	struct rtentry *rt;
911 	struct sockaddr_inarp sin;
912 	const char *why = 0;
913 
914 	bzero(&sin, sizeof(sin));
915 	sin.sin_len = sizeof(sin);
916 	sin.sin_family = AF_INET;
917 	sin.sin_addr.s_addr = addr;
918 	if (proxy)
919 		sin.sin_other = SIN_PROXY;
920 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
921 	if (rt == 0)
922 		return (0);
923 
924 	if (rt->rt_flags & RTF_GATEWAY)
925 		why = "host is not on local network";
926 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
927 		why = "could not allocate llinfo";
928 	else if (rt->rt_gateway->sa_family != AF_LINK)
929 		why = "gateway route is not ours";
930 
931 	if (why) {
932 #define	ISDYNCLONE(_rt) \
933 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
934 		if (create)
935 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
936 			    inet_ntoa(sin.sin_addr), why);
937 		/*
938 		 * If there are no references to this Layer 2 route,
939 		 * and it is a cloned route, and not static, and
940 		 * arplookup() is creating the route, then purge
941 		 * it from the routing table as it is probably bogus.
942 		 */
943 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
944 			rtexpunge(rt);
945 		RTFREE_LOCKED(rt);
946 		return (0);
947 #undef ISDYNCLONE
948 	} else {
949 		RT_REMREF(rt);
950 		return (rt);
951 	}
952 }
953 
954 void
955 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
956 {
957 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
958 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
959 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
960 	ifa->ifa_rtrequest = arp_rtrequest;
961 	ifa->ifa_flags |= RTF_CLONING;
962 }
963 
964 void
965 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
966 {
967 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
968 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
969 				&IA_SIN(ifa)->sin_addr, enaddr);
970 	ifa->ifa_rtrequest = arp_rtrequest;
971 	ifa->ifa_flags |= RTF_CLONING;
972 }
973 
974 static void
975 arp_init(void)
976 {
977 
978 	arpintrq.ifq_maxlen = 50;
979 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
980 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
981 }
982 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
983