xref: /freebsd/sys/netinet/if_ether.c (revision 9336e0699bda8a301cd2bfa37106b6ec5e32012e)
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  *    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 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
456 
457 	/*
458 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
459 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
460 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
461 	 * ARP request, but not faster than one request per second.
462 	 */
463 	if (la->la_asked < arp_maxtries)
464 		error = EWOULDBLOCK;	/* First request. */
465 	else
466 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
467 
468 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
469 		struct in_addr sin =
470 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
471 
472 		rt->rt_expire = time_uptime;
473 		callout_reset(&la->la_timer, hz, arptimer, rt);
474 		la->la_asked++;
475 		RT_UNLOCK(rt);
476 
477 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
478 		    IF_LLADDR(ifp));
479 	} else
480 		RT_UNLOCK(rt);
481 
482 	return (error);
483 }
484 
485 
486 int
487 arpresolve2(struct ifnet *ifp, struct rtentry *rt0, struct sockaddr *dst,
488     u_char *desten)
489 {
490 	struct llinfo_arp *la = NULL;
491 	struct rtentry *rt = NULL;
492 	struct sockaddr_dl *sdl;
493 	int error;
494 
495 	if (rt0 != NULL) {
496 		error = rt_check(&rt, &rt0, dst);
497 		if (error)
498 			return (error);
499 
500 		la = (struct llinfo_arp *)rt->rt_llinfo;
501 		if (la == NULL)
502 			RT_UNLOCK(rt);
503 	}
504 	if (la == NULL) {
505 		/*
506 		 * We enter this block in case if rt0 was NULL,
507 		 * or if rt found by rt_check() didn't have llinfo.
508 		 */
509 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
510 		if (rt == NULL) {
511 			log(LOG_DEBUG,
512 			    "arpresolve: can't allocate route for %s\n",
513 			    inet_ntoa(SIN(dst)->sin_addr));
514 			return (EINVAL); /* XXX */
515 		}
516 		la = (struct llinfo_arp *)rt->rt_llinfo;
517 		if (la == NULL) {
518 			RT_UNLOCK(rt);
519 			log(LOG_DEBUG,
520 			    "arpresolve: can't allocate llinfo for %s\n",
521 			    inet_ntoa(SIN(dst)->sin_addr));
522 			return (EINVAL); /* XXX */
523 		}
524 	}
525 	sdl = SDL(rt->rt_gateway);
526 	/*
527 	 * Check the address family and length is valid, the address
528 	 * is resolved; otherwise, try to resolve.
529 	 */
530 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
531 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
532 
533 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
534 
535 		/*
536 		 * If entry has an expiry time and it is approaching,
537 		 * send an ARP request.
538 		 */
539 		if ((rt->rt_expire != 0) &&
540 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
541 			struct in_addr sin =
542 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
543 
544 			la->la_preempt--;
545 			RT_UNLOCK(rt);
546 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
547 			    IF_LLADDR(ifp));
548 			return (0);
549 		}
550 
551 		RT_UNLOCK(rt);
552 		return (0);
553 	}
554 	/*
555 	 * If ARP is disabled or static on this interface, stop.
556 	 * XXX
557 	 * Probably should not allocate empty llinfo struct if we are
558 	 * not going to be sending out an arp request.
559 	 */
560 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
561 		RT_UNLOCK(rt);
562 		return (EINVAL);
563 	}
564 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
565 
566 	/*
567 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
568 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
569 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
570 	 * ARP request, but not faster than one request per second.
571 	 */
572 	if (la->la_asked < arp_maxtries)
573 		error = EWOULDBLOCK;	/* First request. */
574 	else
575 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
576 
577 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
578 		struct in_addr sin =
579 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
580 
581 		rt->rt_expire = time_uptime;
582 		callout_reset(&la->la_timer, hz, arptimer, rt);
583 		la->la_asked++;
584 		RT_UNLOCK(rt);
585 
586 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
587 		    IF_LLADDR(ifp));
588 	} else
589 		RT_UNLOCK(rt);
590 
591 	return (error);
592 }
593 
594 /*
595  * Common length and type checks are done here,
596  * then the protocol-specific routine is called.
597  */
598 static void
599 arpintr(struct mbuf *m)
600 {
601 	struct arphdr *ar;
602 
603 	if (m->m_len < sizeof(struct arphdr) &&
604 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
605 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
606 		return;
607 	}
608 	ar = mtod(m, struct arphdr *);
609 
610 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
611 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
612 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
613 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
614 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
615 		    (unsigned char *)&ar->ar_hrd, "");
616 		m_freem(m);
617 		return;
618 	}
619 
620 	if (m->m_len < arphdr_len(ar)) {
621 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
622 			log(LOG_ERR, "arp: runt packet\n");
623 			m_freem(m);
624 			return;
625 		}
626 		ar = mtod(m, struct arphdr *);
627 	}
628 
629 	switch (ntohs(ar->ar_pro)) {
630 #ifdef INET
631 	case ETHERTYPE_IP:
632 		in_arpinput(m);
633 		return;
634 #endif
635 	}
636 	m_freem(m);
637 }
638 
639 #ifdef INET
640 /*
641  * ARP for Internet protocols on 10 Mb/s Ethernet.
642  * Algorithm is that given in RFC 826.
643  * In addition, a sanity check is performed on the sender
644  * protocol address, to catch impersonators.
645  * We no longer handle negotiations for use of trailer protocol:
646  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
647  * along with IP replies if we wanted trailers sent to us,
648  * and also sent them in response to IP replies.
649  * This allowed either end to announce the desire to receive
650  * trailer packets.
651  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
652  * but formerly didn't normally send requests.
653  */
654 static int log_arp_wrong_iface = 1;
655 static int log_arp_movements = 1;
656 static int log_arp_permanent_modify = 1;
657 
658 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
659 	&log_arp_wrong_iface, 0,
660 	"log arp packets arriving on the wrong interface");
661 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
662         &log_arp_movements, 0,
663         "log arp replies from MACs different than the one in the cache");
664 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
665         &log_arp_permanent_modify, 0,
666         "log arp replies from MACs different than the one in the permanent arp entry");
667 
668 
669 static void
670 in_arpinput(struct mbuf *m)
671 {
672 	struct arphdr *ah;
673 	struct ifnet *ifp = m->m_pkthdr.rcvif;
674 	struct llinfo_arp *la;
675 	struct rtentry *rt;
676 	struct ifaddr *ifa;
677 	struct in_ifaddr *ia;
678 	struct sockaddr_dl *sdl;
679 	struct sockaddr sa;
680 	struct in_addr isaddr, itaddr, myaddr;
681 	struct mbuf *hold;
682 	u_int8_t *enaddr = NULL;
683 	int op, rif_len;
684 	int req_len;
685 	int bridged = 0;
686 #ifdef DEV_CARP
687 	int carp_match = 0;
688 #endif
689 	struct sockaddr_in sin;
690 	sin.sin_len = sizeof(struct sockaddr_in);
691 	sin.sin_family = AF_INET;
692 	sin.sin_addr.s_addr = 0;
693 
694 	if (ifp->if_bridge)
695 		bridged = 1;
696 
697 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
698 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
699 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
700 		return;
701 	}
702 
703 	ah = mtod(m, struct arphdr *);
704 	op = ntohs(ah->ar_op);
705 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
706 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
707 
708 	/*
709 	 * For a bridge, we want to check the address irrespective
710 	 * of the receive interface. (This will change slightly
711 	 * when we have clusters of interfaces).
712 	 * If the interface does not match, but the recieving interface
713 	 * is part of carp, we call carp_iamatch to see if this is a
714 	 * request for the virtual host ip.
715 	 * XXX: This is really ugly!
716 	 */
717 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
718 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
719 		    (ia->ia_ifp == ifp)) &&
720 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
721 			goto match;
722 #ifdef DEV_CARP
723 		if (ifp->if_carp != NULL &&
724 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
725 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
726 			carp_match = 1;
727 			goto match;
728 		}
729 #endif
730 	}
731 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
732 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
733 		    (ia->ia_ifp == ifp)) &&
734 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
735 			goto match;
736 	/*
737 	 * No match, use the first inet address on the receive interface
738 	 * as a dummy address for the rest of the function.
739 	 */
740 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
741 		if (ifa->ifa_addr->sa_family == AF_INET) {
742 			ia = ifatoia(ifa);
743 			goto match;
744 		}
745 	/*
746 	 * If bridging, fall back to using any inet address.
747 	 */
748 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
749 		goto drop;
750 match:
751 	if (!enaddr)
752 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
753 	myaddr = ia->ia_addr.sin_addr;
754 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
755 		goto drop;	/* it's from me, ignore it. */
756 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
757 		log(LOG_ERR,
758 		    "arp: link address is broadcast for IP address %s!\n",
759 		    inet_ntoa(isaddr));
760 		goto drop;
761 	}
762 	/*
763 	 * Warn if another host is using the same IP address, but only if the
764 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
765 	 * case we suppress the warning to avoid false positive complaints of
766 	 * potential misconfiguration.
767 	 */
768 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
769 		log(LOG_ERR,
770 		   "arp: %*D is using my IP address %s on %s!\n",
771 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
772 		   inet_ntoa(isaddr), ifp->if_xname);
773 		itaddr = myaddr;
774 		goto reply;
775 	}
776 	if (ifp->if_flags & IFF_STATICARP)
777 		goto reply;
778 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
779 	if (rt != NULL) {
780 		sin.sin_addr.s_addr = isaddr.s_addr;
781 		EVENTHANDLER_INVOKE(route_arp_update_event, rt,
782 		    ar_sha(ah), (struct sockaddr *)&sin);
783 
784 		la = (struct llinfo_arp *)rt->rt_llinfo;
785 		if (la == NULL) {
786 			RT_UNLOCK(rt);
787 			goto reply;
788 		}
789 	} else
790 		goto reply;
791 
792 	/* The following is not an error when doing bridging. */
793 	if (!bridged && rt->rt_ifp != ifp
794 #ifdef DEV_CARP
795 	    && (ifp->if_type != IFT_CARP || !carp_match)
796 #endif
797 							) {
798 		if (log_arp_wrong_iface)
799 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
800 			    inet_ntoa(isaddr),
801 			    rt->rt_ifp->if_xname,
802 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
803 			    ifp->if_xname);
804 		RT_UNLOCK(rt);
805 		goto reply;
806 	}
807 	sdl = SDL(rt->rt_gateway);
808 	if (sdl->sdl_alen &&
809 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
810 		if (rt->rt_expire) {
811 		    if (log_arp_movements)
812 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
813 			    inet_ntoa(isaddr),
814 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
815 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
816 			    ifp->if_xname);
817 		} else {
818 			RT_UNLOCK(rt);
819 			if (log_arp_permanent_modify)
820 				log(LOG_ERR, "arp: %*D attempts to modify "
821 				    "permanent entry for %s on %s\n",
822 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
823 				    inet_ntoa(isaddr), ifp->if_xname);
824 			goto reply;
825 		}
826 	}
827 	/*
828 	 * sanity check for the address length.
829 	 * XXX this does not work for protocols with variable address
830 	 * length. -is
831 	 */
832 	if (sdl->sdl_alen &&
833 	    sdl->sdl_alen != ah->ar_hln) {
834 		log(LOG_WARNING,
835 		    "arp from %*D: new addr len %d, was %d",
836 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
837 		    ah->ar_hln, sdl->sdl_alen);
838 	}
839 	if (ifp->if_addrlen != ah->ar_hln) {
840 		log(LOG_WARNING,
841 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
842 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
843 		    ah->ar_hln, ifp->if_addrlen);
844 		RT_UNLOCK(rt);
845 		goto reply;
846 	}
847 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
848 	    sdl->sdl_alen = ah->ar_hln);
849 	/*
850 	 * If we receive an arp from a token-ring station over
851 	 * a token-ring nic then try to save the source
852 	 * routing info.
853 	 */
854 	if (ifp->if_type == IFT_ISO88025) {
855 		struct iso88025_header *th = NULL;
856 		struct iso88025_sockaddr_dl_data *trld;
857 
858 		th = (struct iso88025_header *)m->m_pkthdr.header;
859 		trld = SDL_ISO88025(sdl);
860 		rif_len = TR_RCF_RIFLEN(th->rcf);
861 		if ((th->iso88025_shost[0] & TR_RII) &&
862 		    (rif_len > 2)) {
863 			trld->trld_rcf = th->rcf;
864 			trld->trld_rcf ^= htons(TR_RCF_DIR);
865 			memcpy(trld->trld_route, th->rd, rif_len - 2);
866 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
867 			/*
868 			 * Set up source routing information for
869 			 * reply packet (XXX)
870 			 */
871 			m->m_data -= rif_len;
872 			m->m_len  += rif_len;
873 			m->m_pkthdr.len += rif_len;
874 		} else {
875 			th->iso88025_shost[0] &= ~TR_RII;
876 			trld->trld_rcf = 0;
877 		}
878 		m->m_data -= 8;
879 		m->m_len  += 8;
880 		m->m_pkthdr.len += 8;
881 		th->rcf = trld->trld_rcf;
882 	}
883 	if (rt->rt_expire) {
884 		rt->rt_expire = time_uptime + arpt_keep;
885 		callout_reset(&la->la_timer, hz * arpt_keep, arptimer, rt);
886 	}
887 	la->la_asked = 0;
888 	la->la_preempt = arp_maxtries;
889 	hold = la->la_hold;
890 	la->la_hold = NULL;
891 	RT_UNLOCK(rt);
892 	if (hold != NULL)
893 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
894 
895 reply:
896 	if (op != ARPOP_REQUEST)
897 		goto drop;
898 	if (itaddr.s_addr == myaddr.s_addr) {
899 		/* I am the target */
900 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
901 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
902 	} else {
903 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
904 		if (rt == NULL) {
905 			struct sockaddr_in sin;
906 
907 			if (!arp_proxyall)
908 				goto drop;
909 
910 			bzero(&sin, sizeof sin);
911 			sin.sin_family = AF_INET;
912 			sin.sin_len = sizeof sin;
913 			sin.sin_addr = itaddr;
914 
915 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
916 			if (!rt)
917 				goto drop;
918 			/*
919 			 * Don't send proxies for nodes on the same interface
920 			 * as this one came out of, or we'll get into a fight
921 			 * over who claims what Ether address.
922 			 */
923 			if (rt->rt_ifp == ifp) {
924 				rtfree(rt);
925 				goto drop;
926 			}
927 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
928 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
929 			rtfree(rt);
930 
931 			/*
932 			 * Also check that the node which sent the ARP packet
933 			 * is on the the interface we expect it to be on. This
934 			 * avoids ARP chaos if an interface is connected to the
935 			 * wrong network.
936 			 */
937 			sin.sin_addr = isaddr;
938 
939 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
940 			if (!rt)
941 				goto drop;
942 			if (rt->rt_ifp != ifp) {
943 				log(LOG_INFO, "arp_proxy: ignoring request"
944 				    " from %s via %s, expecting %s\n",
945 				    inet_ntoa(isaddr), ifp->if_xname,
946 				    rt->rt_ifp->if_xname);
947 				rtfree(rt);
948 				goto drop;
949 			}
950 			rtfree(rt);
951 
952 #ifdef DEBUG_PROXY
953 			printf("arp: proxying for %s\n",
954 			       inet_ntoa(itaddr));
955 #endif
956 		} else {
957 			/*
958 			 * Return proxied ARP replies only on the interface
959 			 * or bridge cluster where this network resides.
960 			 * Otherwise we may conflict with the host we are
961 			 * proxying for.
962 			 */
963 			if (rt->rt_ifp != ifp &&
964 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
965 			    ifp->if_bridge == NULL)) {
966 				RT_UNLOCK(rt);
967 				goto drop;
968 			}
969 			sdl = SDL(rt->rt_gateway);
970 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
971 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
972 			RT_UNLOCK(rt);
973 		}
974 	}
975 
976 	if (itaddr.s_addr == myaddr.s_addr &&
977 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
978 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
979 #ifdef DEBUG_LINKLOCAL
980 		printf("arp: sending reply for link-local addr %s\n",
981 		    inet_ntoa(itaddr));
982 #endif
983 		m->m_flags |= M_BCAST;
984 		m->m_flags &= ~M_MCAST;
985 	} else {
986 		/* default behaviour; never reply by broadcast. */
987 		m->m_flags &= ~(M_BCAST|M_MCAST);
988 	}
989 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
990 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
991 	ah->ar_op = htons(ARPOP_REPLY);
992 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
993 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
994 	m->m_pkthdr.len = m->m_len;
995 	sa.sa_family = AF_ARP;
996 	sa.sa_len = 2;
997 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
998 	return;
999 
1000 drop:
1001 	m_freem(m);
1002 }
1003 #endif
1004 
1005 /*
1006  * Lookup or enter a new address in arptab.
1007  */
1008 static struct rtentry *
1009 arplookup(u_long addr, int create, int proxy)
1010 {
1011 	struct rtentry *rt;
1012 	struct sockaddr_inarp sin;
1013 	const char *why = 0;
1014 
1015 	bzero(&sin, sizeof(sin));
1016 	sin.sin_len = sizeof(sin);
1017 	sin.sin_family = AF_INET;
1018 	sin.sin_addr.s_addr = addr;
1019 	if (proxy)
1020 		sin.sin_other = SIN_PROXY;
1021 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
1022 	if (rt == 0)
1023 		return (0);
1024 
1025 	if (rt->rt_flags & RTF_GATEWAY)
1026 		why = "host is not on local network";
1027 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
1028 		why = "could not allocate llinfo";
1029 	else if (rt->rt_gateway->sa_family != AF_LINK)
1030 		why = "gateway route is not ours";
1031 
1032 	if (why) {
1033 #define	ISDYNCLONE(_rt) \
1034 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
1035 		if (create)
1036 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1037 			    inet_ntoa(sin.sin_addr), why);
1038 		/*
1039 		 * If there are no references to this Layer 2 route,
1040 		 * and it is a cloned route, and not static, and
1041 		 * arplookup() is creating the route, then purge
1042 		 * it from the routing table as it is probably bogus.
1043 		 */
1044 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
1045 			rtexpunge(rt);
1046 		RTFREE_LOCKED(rt);
1047 		return (0);
1048 #undef ISDYNCLONE
1049 	} else {
1050 		RT_REMREF(rt);
1051 		return (rt);
1052 	}
1053 }
1054 
1055 void
1056 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1057 {
1058 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
1059 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1060 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1061 	ifa->ifa_rtrequest = arp_rtrequest;
1062 	ifa->ifa_flags |= RTF_CLONING;
1063 }
1064 
1065 void
1066 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
1067 {
1068 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
1069 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1070 				&IA_SIN(ifa)->sin_addr, enaddr);
1071 	ifa->ifa_rtrequest = arp_rtrequest;
1072 	ifa->ifa_flags |= RTF_CLONING;
1073 }
1074 
1075 static void
1076 arp_init(void)
1077 {
1078 
1079 	arpintrq.ifq_maxlen = 50;
1080 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
1081 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
1082 }
1083 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1084