xref: /openbsd-src/sys/netinet/if_ether.c (revision 897fc685943471cf985a0fe38ba076ea6fe74fa5)
1 /*	$OpenBSD: if_ether.c,v 1.235 2018/03/31 15:07:09 stsp Exp $	*/
2 /*	$NetBSD: if_ether.c,v 1.31 1996/05/11 12:59:58 mycroft Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
33  */
34 
35 /*
36  * Ethernet address resolution protocol.
37  * TODO:
38  *	add "inuse/lock" bit (or ref. count) along with valid bit
39  */
40 
41 #include "carp.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/mbuf.h>
46 #include <sys/socket.h>
47 #include <sys/timeout.h>
48 #include <sys/kernel.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
51 #include <sys/pool.h>
52 
53 #include <net/if.h>
54 #include <net/if_var.h>
55 #include <net/if_dl.h>
56 #include <net/route.h>
57 #include <net/if_types.h>
58 #include <net/netisr.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_var.h>
62 #include <netinet/if_ether.h>
63 #if NCARP > 0
64 #include <netinet/ip_carp.h>
65 #endif
66 
67 struct llinfo_arp {
68 	LIST_ENTRY(llinfo_arp)	 la_list;
69 	struct rtentry		*la_rt;		/* backpointer to rtentry */
70 	long			 la_asked;	/* last time we QUERIED */
71 	struct mbuf_list	 la_ml;		/* packet hold queue */
72 };
73 #define LA_HOLD_QUEUE 10
74 #define LA_HOLD_TOTAL 100
75 
76 /* timer values */
77 int	arpt_prune = (5 * 60);	/* walk list every 5 minutes */
78 int	arpt_keep = (20 * 60);	/* once resolved, cache for 20 minutes */
79 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
80 
81 struct mbuf *arppullup(struct mbuf *m);
82 void arpinvalidate(struct rtentry *);
83 void arptfree(struct rtentry *);
84 void arptimer(void *);
85 struct rtentry *arplookup(struct in_addr *, int, int, unsigned int);
86 void in_arpinput(struct ifnet *, struct mbuf *);
87 void in_revarpinput(struct ifnet *, struct mbuf *);
88 int arpcache(struct ifnet *, struct ether_arp *, struct rtentry *);
89 void arpreply(struct ifnet *, struct mbuf *, struct in_addr *, uint8_t *,
90     unsigned int);
91 
92 struct niqueue arpinq = NIQUEUE_INITIALIZER(50, NETISR_ARP);
93 
94 LIST_HEAD(, llinfo_arp) arp_list;
95 struct	pool arp_pool;		/* pool for llinfo_arp structures */
96 int	arp_maxtries = 5;
97 int	arpinit_done;
98 int	la_hold_total;
99 
100 #ifdef NFSCLIENT
101 /* revarp state */
102 struct in_addr revarp_myip, revarp_srvip;
103 int revarp_finished;
104 unsigned int revarp_ifidx;
105 #endif /* NFSCLIENT */
106 
107 /*
108  * Timeout routine.  Age arp_tab entries periodically.
109  */
110 /* ARGSUSED */
111 void
112 arptimer(void *arg)
113 {
114 	struct timeout *to = (struct timeout *)arg;
115 	struct llinfo_arp *la, *nla;
116 
117 	NET_LOCK();
118 	timeout_add_sec(to, arpt_prune);
119 	LIST_FOREACH_SAFE(la, &arp_list, la_list, nla) {
120 		struct rtentry *rt = la->la_rt;
121 
122 		if (rt->rt_expire && rt->rt_expire <= time_uptime)
123 			arptfree(rt); /* timer has expired; clear */
124 	}
125 	NET_UNLOCK();
126 }
127 
128 void
129 arp_rtrequest(struct ifnet *ifp, int req, struct rtentry *rt)
130 {
131 	struct sockaddr *gate = rt->rt_gateway;
132 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
133 	struct ifaddr *ifa;
134 
135 	if (!arpinit_done) {
136 		static struct timeout arptimer_to;
137 
138 		arpinit_done = 1;
139 		pool_init(&arp_pool, sizeof(struct llinfo_arp), 0,
140 		    IPL_SOFTNET, 0, "arp", NULL);
141 
142 		timeout_set_proc(&arptimer_to, arptimer, &arptimer_to);
143 		timeout_add_sec(&arptimer_to, 1);
144 	}
145 
146 	if (ISSET(rt->rt_flags, RTF_GATEWAY|RTF_BROADCAST|RTF_MULTICAST))
147 		return;
148 
149 	switch (req) {
150 
151 	case RTM_ADD:
152 		if (rt->rt_flags & RTF_CLONING ||
153 		    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !la)) {
154 			/*
155 			 * Give this route an expiration time, even though
156 			 * it's a "permanent" route, so that routes cloned
157 			 * from it do not need their expiration time set.
158 			 */
159 			rt->rt_expire = time_uptime;
160 			if ((rt->rt_flags & RTF_CLONING) != 0)
161 				break;
162 		}
163 		/*
164 		 * Announce a new entry if requested or warn the user
165 		 * if another station has this IP address.
166 		 */
167 		if (rt->rt_flags & (RTF_ANNOUNCE|RTF_LOCAL))
168 			arprequest(ifp,
169 			    &satosin(rt_key(rt))->sin_addr.s_addr,
170 			    &satosin(rt_key(rt))->sin_addr.s_addr,
171 			    (u_char *)LLADDR(satosdl(gate)));
172 		/*FALLTHROUGH*/
173 	case RTM_RESOLVE:
174 		if (gate->sa_family != AF_LINK ||
175 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
176 			log(LOG_DEBUG, "%s: bad gateway value: %s\n", __func__,
177 			    ifp->if_xname);
178 			break;
179 		}
180 		satosdl(gate)->sdl_type = ifp->if_type;
181 		satosdl(gate)->sdl_index = ifp->if_index;
182 		if (la != 0)
183 			break; /* This happens on a route change */
184 		/*
185 		 * Case 2:  This route may come from cloning, or a manual route
186 		 * add with a LL address.
187 		 */
188 		la = pool_get(&arp_pool, PR_NOWAIT | PR_ZERO);
189 		rt->rt_llinfo = (caddr_t)la;
190 		if (la == NULL) {
191 			log(LOG_DEBUG, "%s: pool get failed\n", __func__);
192 			break;
193 		}
194 
195 		ml_init(&la->la_ml);
196 		la->la_rt = rt;
197 		rt->rt_flags |= RTF_LLINFO;
198 		LIST_INSERT_HEAD(&arp_list, la, la_list);
199 
200 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
201 			if ((ifa->ifa_addr->sa_family == AF_INET) &&
202 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr ==
203 			    satosin(rt_key(rt))->sin_addr.s_addr)
204 				break;
205 		}
206 		if (ifa) {
207 			KASSERT(ifa == rt->rt_ifa);
208 			rt->rt_expire = 0;
209 		}
210 		break;
211 
212 	case RTM_DELETE:
213 		if (la == NULL)
214 			break;
215 		LIST_REMOVE(la, la_list);
216 		rt->rt_llinfo = NULL;
217 		rt->rt_flags &= ~RTF_LLINFO;
218 		la_hold_total -= ml_purge(&la->la_ml);
219 		pool_put(&arp_pool, la);
220 		break;
221 
222 	case RTM_INVALIDATE:
223 		if (!ISSET(rt->rt_flags, RTF_LOCAL))
224 			arpinvalidate(rt);
225 		break;
226 	}
227 }
228 
229 /*
230  * Broadcast an ARP request. Caller specifies:
231  *	- arp header source ip address
232  *	- arp header target ip address
233  *	- arp header source ethernet address
234  */
235 void
236 arprequest(struct ifnet *ifp, u_int32_t *sip, u_int32_t *tip, u_int8_t *enaddr)
237 {
238 	struct mbuf *m;
239 	struct ether_header *eh;
240 	struct ether_arp *ea;
241 	struct sockaddr sa;
242 
243 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
244 		return;
245 	m->m_len = sizeof(*ea);
246 	m->m_pkthdr.len = sizeof(*ea);
247 	m->m_pkthdr.ph_rtableid = ifp->if_rdomain;
248 	m->m_pkthdr.pf.prio = ifp->if_llprio;
249 	MH_ALIGN(m, sizeof(*ea));
250 	ea = mtod(m, struct ether_arp *);
251 	eh = (struct ether_header *)sa.sa_data;
252 	memset(ea, 0, sizeof(*ea));
253 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
254 	eh->ether_type = htons(ETHERTYPE_ARP);	/* if_output will not swap */
255 	ea->arp_hrd = htons(ARPHRD_ETHER);
256 	ea->arp_pro = htons(ETHERTYPE_IP);
257 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
258 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
259 	ea->arp_op = htons(ARPOP_REQUEST);
260 	memcpy(eh->ether_shost, enaddr, sizeof(eh->ether_shost));
261 	memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
262 	memcpy(ea->arp_spa, sip, sizeof(ea->arp_spa));
263 	memcpy(ea->arp_tpa, tip, sizeof(ea->arp_tpa));
264 	sa.sa_family = pseudo_AF_HDRCMPLT;
265 	sa.sa_len = sizeof(sa);
266 	m->m_flags |= M_BCAST;
267 	ifp->if_output(ifp, m, &sa, NULL);
268 }
269 
270 void
271 arpreply(struct ifnet *ifp, struct mbuf *m, struct in_addr *sip, uint8_t *eaddr,
272     unsigned int rdomain)
273 {
274 	struct ether_header *eh;
275 	struct ether_arp *ea;
276 	struct sockaddr sa;
277 
278 	m_resethdr(m);
279 	m->m_pkthdr.ph_rtableid = rdomain;
280 
281 	ea = mtod(m, struct ether_arp *);
282 	ea->arp_op = htons(ARPOP_REPLY);
283 	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
284 
285 	/* We're replying to a request. */
286 	memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
287 	memcpy(ea->arp_tpa, ea->arp_spa, sizeof(ea->arp_spa));
288 
289 	memcpy(ea->arp_sha, eaddr, sizeof(ea->arp_sha));
290 	memcpy(ea->arp_spa, sip, sizeof(ea->arp_spa));
291 
292 	eh = (struct ether_header *)sa.sa_data;
293 	memcpy(eh->ether_dhost, ea->arp_tha, sizeof(eh->ether_dhost));
294 	memcpy(eh->ether_shost, eaddr, sizeof(eh->ether_shost));
295 	eh->ether_type = htons(ETHERTYPE_ARP);
296 	sa.sa_family = pseudo_AF_HDRCMPLT;
297 	sa.sa_len = sizeof(sa);
298 	ifp->if_output(ifp, m, &sa, NULL);
299 }
300 
301 /*
302  * Resolve an IP address into an ethernet address.  If success,
303  * desten is filled in.  If there is no entry in arptab,
304  * set one up and broadcast a request for the IP address.
305  * Hold onto this mbuf and resend it once the address
306  * is finally resolved.  A return value of 0 indicates
307  * that desten has been filled in and the packet should be sent
308  * normally; A return value of EAGAIN indicates that the packet
309  * has been taken over here, either now or for later transmission.
310  * Any other return value indicates an error.
311  */
312 int
313 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
314     struct sockaddr *dst, u_char *desten)
315 {
316 	struct arpcom *ac = (struct arpcom *)ifp;
317 	struct llinfo_arp *la = NULL;
318 	struct sockaddr_dl *sdl;
319 	struct rtentry *rt = NULL;
320 	char addr[INET_ADDRSTRLEN];
321 
322 	if (m->m_flags & M_BCAST) {	/* broadcast */
323 		memcpy(desten, etherbroadcastaddr, sizeof(etherbroadcastaddr));
324 		return (0);
325 	}
326 	if (m->m_flags & M_MCAST) {	/* multicast */
327 		ETHER_MAP_IP_MULTICAST(&satosin(dst)->sin_addr, desten);
328 		return (0);
329 	}
330 
331 	rt = rt_getll(rt0);
332 
333 	if (ISSET(rt->rt_flags, RTF_REJECT) &&
334 	    (rt->rt_expire == 0 || time_uptime < rt->rt_expire)) {
335 		m_freem(m);
336 		return (rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
337 	}
338 
339 	if (!ISSET(rt->rt_flags, RTF_LLINFO)) {
340 		log(LOG_DEBUG, "%s: %s: route contains no arp information\n",
341 		    __func__, inet_ntop(AF_INET, &satosin(rt_key(rt))->sin_addr,
342 		    addr, sizeof(addr)));
343 		m_freem(m);
344 		return (EINVAL);
345 	}
346 
347 	sdl = satosdl(rt->rt_gateway);
348 	if (sdl->sdl_alen > 0 && sdl->sdl_alen != ETHER_ADDR_LEN) {
349 		log(LOG_DEBUG, "%s: %s: incorrect arp information\n", __func__,
350 		    inet_ntop(AF_INET, &satosin(dst)->sin_addr,
351 			addr, sizeof(addr)));
352 		goto bad;
353 	}
354 
355 	/*
356 	 * Check the address family and length is valid, the address
357 	 * is resolved; otherwise, try to resolve.
358 	 */
359 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
360 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
361 		memcpy(desten, LLADDR(sdl), sdl->sdl_alen);
362 		return (0);
363 	}
364 
365 	if (ifp->if_flags & (IFF_NOARP|IFF_STATICARP))
366 		goto bad;
367 
368 	/*
369 	 * There is an arptab entry, but no ethernet address
370 	 * response yet. Insert mbuf in hold queue if below limit
371 	 * if above the limit free the queue without queuing the new packet.
372 	 */
373 	la = (struct llinfo_arp *)rt->rt_llinfo;
374 	KASSERT(la != NULL);
375 	if (la_hold_total < LA_HOLD_TOTAL && la_hold_total < nmbclust / 64) {
376 		struct mbuf *mh;
377 
378 		if (ml_len(&la->la_ml) >= LA_HOLD_QUEUE) {
379 			mh = ml_dequeue(&la->la_ml);
380 			la_hold_total--;
381 			m_freem(mh);
382 		}
383 		ml_enqueue(&la->la_ml, m);
384 		la_hold_total++;
385 	} else {
386 		la_hold_total -= ml_purge(&la->la_ml);
387 		m_freem(m);
388 	}
389 
390 	/*
391 	 * Re-send the ARP request when appropriate.
392 	 */
393 #ifdef	DIAGNOSTIC
394 	if (rt->rt_expire == 0) {
395 		/* This should never happen. (Should it? -gwr) */
396 		printf("%s: unresolved and rt_expire == 0\n", __func__);
397 		/* Set expiration time to now (expired). */
398 		rt->rt_expire = time_uptime;
399 	}
400 #endif
401 	if (rt->rt_expire) {
402 		rt->rt_flags &= ~RTF_REJECT;
403 		if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
404 			rt->rt_expire = time_uptime;
405 			if (la->la_asked++ < arp_maxtries)
406 				arprequest(ifp,
407 				    &satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr,
408 				    &satosin(dst)->sin_addr.s_addr,
409 				    ac->ac_enaddr);
410 			else {
411 				rt->rt_flags |= RTF_REJECT;
412 				rt->rt_expire += arpt_down;
413 				la->la_asked = 0;
414 				la_hold_total -= ml_purge(&la->la_ml);
415 			}
416 		}
417 	}
418 
419 	return (EAGAIN);
420 
421 bad:
422 	m_freem(m);
423 	return (EINVAL);
424 }
425 
426 struct mbuf *
427 arppullup(struct mbuf *m)
428 {
429 	struct arphdr *ar;
430 	int len;
431 
432 #ifdef DIAGNOSTIC
433 	if ((m->m_flags & M_PKTHDR) == 0)
434 		panic("arp without packet header");
435 #endif
436 
437 	len = sizeof(struct arphdr);
438 	if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
439 		return NULL;
440 
441 	ar = mtod(m, struct arphdr *);
442 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER ||
443 	    ntohs(ar->ar_pro) != ETHERTYPE_IP ||
444 	    ar->ar_hln != ETHER_ADDR_LEN ||
445 	    ar->ar_pln != sizeof(struct in_addr)) {
446 		m_freem(m);
447 		return NULL;
448 	}
449 
450 	len += 2 * (ar->ar_hln + ar->ar_pln);
451 	if (m->m_len < len && (m = m_pullup(m, len)) == NULL)
452 		return NULL;
453 
454 	return m;
455 }
456 
457 /*
458  * Common length and type checks are done here,
459  * then the protocol-specific routine is called.
460  */
461 void
462 arpinput(struct ifnet *ifp, struct mbuf *m)
463 {
464 	if ((m = arppullup(m)) == NULL)
465 		return;
466 	niq_enqueue(&arpinq, m);
467 }
468 
469 void
470 arpintr(void)
471 {
472 	struct mbuf_list ml;
473 	struct mbuf *m;
474 	struct ifnet *ifp;
475 
476 	niq_delist(&arpinq, &ml);
477 
478 	while ((m = ml_dequeue(&ml)) != NULL) {
479 		ifp = if_get(m->m_pkthdr.ph_ifidx);
480 
481 		if (ifp != NULL)
482 			in_arpinput(ifp, m);
483 		else
484 			m_freem(m);
485 
486 		if_put(ifp);
487 	}
488 }
489 
490 /*
491  * ARP for Internet protocols on Ethernet, RFC 826.
492  * In addition, a sanity check is performed on the sender
493  * protocol address, to catch impersonators.
494  */
495 void
496 in_arpinput(struct ifnet *ifp, struct mbuf *m)
497 {
498 	struct ether_arp *ea;
499 	struct rtentry *rt = NULL;
500 	struct sockaddr_in sin;
501 	struct in_addr isaddr, itaddr;
502 	char addr[INET_ADDRSTRLEN];
503 	int op, target = 0;
504 	unsigned int rdomain;
505 
506 	rdomain = rtable_l2(m->m_pkthdr.ph_rtableid);
507 
508 	ea = mtod(m, struct ether_arp *);
509 	op = ntohs(ea->arp_op);
510 	if ((op != ARPOP_REQUEST) && (op != ARPOP_REPLY))
511 		goto out;
512 
513 	memcpy(&itaddr, ea->arp_tpa, sizeof(itaddr));
514 	memcpy(&isaddr, ea->arp_spa, sizeof(isaddr));
515 	memset(&sin, 0, sizeof(sin));
516 	sin.sin_len = sizeof(sin);
517 	sin.sin_family = AF_INET;
518 
519 	if (ETHER_IS_MULTICAST(&ea->arp_sha[0]) &&
520 	    !memcmp(ea->arp_sha, etherbroadcastaddr, sizeof(ea->arp_sha))) {
521 		inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
522 		log(LOG_ERR, "arp: ether address is broadcast for IP address "
523 		    "%s!\n", addr);
524 		goto out;
525 	}
526 
527 	if (!memcmp(ea->arp_sha, LLADDR(ifp->if_sadl), sizeof(ea->arp_sha)))
528 		goto out;	/* it's from me, ignore it. */
529 
530 	/* Check target against our interface addresses. */
531 	sin.sin_addr = itaddr;
532 	rt = rtalloc(sintosa(&sin), 0, rdomain);
533 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL) &&
534 	    rt->rt_ifidx == ifp->if_index)
535 		target = 1;
536 	rtfree(rt);
537 	rt = NULL;
538 
539 #if NCARP > 0
540 	if (target && op == ARPOP_REQUEST && ifp->if_type == IFT_CARP &&
541 	    !carp_iamatch(ifp))
542 		goto out;
543 #endif
544 
545 	/* Do we have an ARP cache for the sender? Create if we are target. */
546 	rt = arplookup(&isaddr, target, 0, rdomain);
547 
548 	/* Check sender against our interface addresses. */
549 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL) &&
550 	    rt->rt_ifidx == ifp->if_index && isaddr.s_addr != INADDR_ANY) {
551 		inet_ntop(AF_INET, &isaddr, addr, sizeof(addr));
552 		log(LOG_ERR, "duplicate IP address %s sent from ethernet "
553 		    "address %s\n", addr, ether_sprintf(ea->arp_sha));
554 		itaddr = isaddr;
555 	} else if (rt != NULL) {
556 		int error;
557 
558 		KERNEL_LOCK();
559 		error = arpcache(ifp, ea, rt);
560 		KERNEL_UNLOCK();
561 		if (error)
562 			goto out;
563 	}
564 
565 	if (op == ARPOP_REQUEST) {
566 		uint8_t *eaddr;
567 
568 		if (target) {
569 			/* We already have all info for the reply */
570 			eaddr = LLADDR(ifp->if_sadl);
571 		} else {
572 			rtfree(rt);
573 			rt = arplookup(&itaddr, 0, SIN_PROXY, rdomain);
574 			/*
575 			 * Protect from possible duplicates, only owner
576 			 * should respond
577 			 */
578 			if ((rt == NULL) || (rt->rt_ifidx != ifp->if_index))
579 				goto out;
580 			eaddr = LLADDR(satosdl(rt->rt_gateway));
581 		}
582 		arpreply(ifp, m, &itaddr, eaddr, rdomain);
583 		rtfree(rt);
584 		return;
585 	}
586 
587 out:
588 	rtfree(rt);
589 	m_freem(m);
590 }
591 
592 int
593 arpcache(struct ifnet *ifp, struct ether_arp *ea, struct rtentry *rt)
594 {
595 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
596 	struct sockaddr_dl *sdl = satosdl(rt->rt_gateway);
597 	struct in_addr *spa = (struct in_addr *)ea->arp_spa;
598 	char addr[INET_ADDRSTRLEN];
599 	struct ifnet *rifp;
600 	unsigned int len;
601 	int changed = 0;
602 
603 	KERNEL_ASSERT_LOCKED();
604 	KASSERT(sdl != NULL);
605 
606 	/*
607 	 * This can happen if the entry has been deleted by another CPU
608 	 * after we found it.
609 	 */
610 	if (la == NULL)
611 		return (0);
612 
613 	if (sdl->sdl_alen > 0) {
614 		if (memcmp(ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
615 			if (ISSET(rt->rt_flags, RTF_PERMANENT_ARP|RTF_LOCAL)) {
616 				inet_ntop(AF_INET, spa, addr, sizeof(addr));
617 				log(LOG_WARNING, "arp: attempt to overwrite "
618 				   "permanent entry for %s by %s on %s\n", addr,
619 				   ether_sprintf(ea->arp_sha), ifp->if_xname);
620 				return (-1);
621 			} else if (rt->rt_ifidx != ifp->if_index) {
622 #if NCARP > 0
623 				if (ifp->if_type != IFT_CARP)
624 #endif
625 				{
626 					rifp = if_get(rt->rt_ifidx);
627 					if (rifp == NULL)
628 						return (-1);
629 					inet_ntop(AF_INET, spa, addr,
630 					    sizeof(addr));
631 					log(LOG_WARNING, "arp: attempt to "
632 					    "overwrite entry for %s on %s by "
633 					    "%s on %s\n", addr, rifp->if_xname,
634 					    ether_sprintf(ea->arp_sha),
635 					    ifp->if_xname);
636 					if_put(rifp);
637 				}
638 				return (-1);
639 			} else {
640 				inet_ntop(AF_INET, spa, addr, sizeof(addr));
641 				log(LOG_INFO, "arp info overwritten for %s by "
642 				    "%s on %s\n", addr,
643 				    ether_sprintf(ea->arp_sha), ifp->if_xname);
644 				rt->rt_expire = 1;/* no longer static */
645 			}
646 			changed = 1;
647 		}
648 	} else if (!if_isconnected(ifp, rt->rt_ifidx)) {
649 		rifp = if_get(rt->rt_ifidx);
650 		if (rifp == NULL)
651 			return (-1);
652 		inet_ntop(AF_INET, spa, addr, sizeof(addr));
653 		log(LOG_WARNING, "arp: attempt to add entry for %s on %s by %s"
654 		    " on %s\n", addr, rifp->if_xname,
655 		    ether_sprintf(ea->arp_sha), ifp->if_xname);
656 		if_put(rifp);
657 		return (-1);
658 	}
659 	sdl->sdl_alen = sizeof(ea->arp_sha);
660 	memcpy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
661 	if (rt->rt_expire)
662 		rt->rt_expire = time_uptime + arpt_keep;
663 	rt->rt_flags &= ~RTF_REJECT;
664 
665 	/* Notify userland that an ARP resolution has been done. */
666 	if (la->la_asked || changed) {
667 		KERNEL_LOCK();
668 		rtm_send(rt, RTM_RESOLVE, 0, ifp->if_rdomain);
669 		KERNEL_UNLOCK();
670 	}
671 
672 	la->la_asked = 0;
673 	while ((len = ml_len(&la->la_ml)) != 0) {
674 		struct mbuf *mh;
675 
676 		mh = ml_dequeue(&la->la_ml);
677 		la_hold_total--;
678 
679 		ifp->if_output(ifp, mh, rt_key(rt), rt);
680 
681 		if (ml_len(&la->la_ml) == len) {
682 			/* mbuf is back in queue. Discard. */
683 			while ((mh = ml_dequeue(&la->la_ml)) != NULL) {
684 				la_hold_total--;
685 				m_freem(mh);
686 			}
687 			break;
688 		}
689 	}
690 
691 	return (0);
692 }
693 
694 void
695 arpinvalidate(struct rtentry *rt)
696 {
697 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
698 	struct sockaddr_dl *sdl = satosdl(rt->rt_gateway);
699 
700 	la_hold_total -= ml_purge(&la->la_ml);
701 	sdl->sdl_alen = 0;
702 	la->la_asked = 0;
703 }
704 
705 /*
706  * Free an arp entry.
707  */
708 void
709 arptfree(struct rtentry *rt)
710 {
711 	struct ifnet *ifp;
712 
713 	KASSERT(!ISSET(rt->rt_flags, RTF_LOCAL));
714 	arpinvalidate(rt);
715 
716 	ifp = if_get(rt->rt_ifidx);
717 	KASSERT(ifp != NULL);
718 	if (!ISSET(rt->rt_flags, RTF_STATIC|RTF_CACHED))
719 		rtdeletemsg(rt, ifp, ifp->if_rdomain);
720 	if_put(ifp);
721 }
722 
723 /*
724  * Lookup or enter a new address in arptab.
725  */
726 struct rtentry *
727 arplookup(struct in_addr *inp, int create, int proxy, u_int tableid)
728 {
729 	struct rtentry *rt;
730 	struct sockaddr_inarp sin;
731 	int flags;
732 
733 	memset(&sin, 0, sizeof(sin));
734 	sin.sin_len = sizeof(sin);
735 	sin.sin_family = AF_INET;
736 	sin.sin_addr.s_addr = inp->s_addr;
737 	sin.sin_other = proxy ? SIN_PROXY : 0;
738 	flags = (create) ? RT_RESOLVE : 0;
739 
740 	rt = rtalloc((struct sockaddr *)&sin, flags, tableid);
741 	if (!rtisvalid(rt) || ISSET(rt->rt_flags, RTF_GATEWAY) ||
742 	    !ISSET(rt->rt_flags, RTF_LLINFO) ||
743 	    rt->rt_gateway->sa_family != AF_LINK) {
744 		rtfree(rt);
745 		return (NULL);
746 	}
747 
748 	if (proxy && !ISSET(rt->rt_flags, RTF_ANNOUNCE)) {
749 		while ((rt = rtable_iterate(rt)) != NULL) {
750 			if (ISSET(rt->rt_flags, RTF_ANNOUNCE)) {
751 				break;
752 			}
753 		}
754 	}
755 
756 	return (rt);
757 }
758 
759 /*
760  * Check whether we do proxy ARP for this address and we point to ourselves.
761  */
762 int
763 arpproxy(struct in_addr in, unsigned int rtableid)
764 {
765 	struct sockaddr_dl *sdl;
766 	struct rtentry *rt;
767 	struct ifnet *ifp;
768 	int found = 0;
769 
770 	rt = arplookup(&in, 0, SIN_PROXY, rtableid);
771 	if (!rtisvalid(rt)) {
772 		rtfree(rt);
773 		return (0);
774 	}
775 
776 	/* Check that arp information are correct. */
777 	sdl = satosdl(rt->rt_gateway);
778 	if (sdl->sdl_alen != ETHER_ADDR_LEN) {
779 		rtfree(rt);
780 		return (0);
781 	}
782 
783 	ifp = if_get(rt->rt_ifidx);
784 	if (ifp == NULL) {
785 		rtfree(rt);
786 		return (0);
787 	}
788 
789 	if (!memcmp(LLADDR(sdl), LLADDR(ifp->if_sadl), sdl->sdl_alen))
790 		found = 1;
791 
792 	if_put(ifp);
793 	rtfree(rt);
794 	return (found);
795 }
796 
797 /*
798  * Called from Ethernet interrupt handlers
799  * when ether packet type ETHERTYPE_REVARP
800  * is received.  Common length and type checks are done here,
801  * then the protocol-specific routine is called.
802  */
803 void
804 revarpinput(struct ifnet *ifp, struct mbuf *m)
805 {
806 	if ((m = arppullup(m)) == NULL)
807 		return;
808 	in_revarpinput(ifp, m);
809 }
810 
811 /*
812  * RARP for Internet protocols on Ethernet.
813  * Algorithm is that given in RFC 903.
814  * We are only using for bootstrap purposes to get an ip address for one of
815  * our interfaces.  Thus we support no user-interface.
816  *
817  * Since the contents of the RARP reply are specific to the interface that
818  * sent the request, this code must ensure that they are properly associated.
819  *
820  * Note: also supports ARP via RARP packets, per the RFC.
821  */
822 void
823 in_revarpinput(struct ifnet *ifp, struct mbuf *m)
824 {
825 	struct ether_arp *ar;
826 	int op;
827 
828 	ar = mtod(m, struct ether_arp *);
829 	op = ntohs(ar->arp_op);
830 	switch (op) {
831 	case ARPOP_REQUEST:
832 	case ARPOP_REPLY:	/* per RFC */
833 		niq_enqueue(&arpinq, m);
834 		return;
835 	case ARPOP_REVREPLY:
836 		break;
837 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
838 	default:
839 		goto out;
840 	}
841 #ifdef NFSCLIENT
842 	if (revarp_ifidx == 0)
843 		goto out;
844 	if (revarp_ifidx != m->m_pkthdr.ph_ifidx) /* !same interface */
845 		goto out;
846 	if (revarp_finished)
847 		goto wake;
848 	if (memcmp(ar->arp_tha, LLADDR(ifp->if_sadl), sizeof(ar->arp_tha)))
849 		goto out;
850 	memcpy(&revarp_srvip, ar->arp_spa, sizeof(revarp_srvip));
851 	memcpy(&revarp_myip, ar->arp_tpa, sizeof(revarp_myip));
852 	revarp_finished = 1;
853 wake:	/* Do wakeup every time in case it was missed. */
854 	wakeup((caddr_t)&revarp_myip);
855 #endif /* NFSCLIENT */
856 
857 out:
858 	m_freem(m);
859 }
860 
861 /*
862  * Send a RARP request for the ip address of the specified interface.
863  * The request should be RFC 903-compliant.
864  */
865 void
866 revarprequest(struct ifnet *ifp)
867 {
868 	struct sockaddr sa;
869 	struct mbuf *m;
870 	struct ether_header *eh;
871 	struct ether_arp *ea;
872 	struct arpcom *ac = (struct arpcom *)ifp;
873 
874 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
875 		return;
876 	m->m_len = sizeof(*ea);
877 	m->m_pkthdr.len = sizeof(*ea);
878 	m->m_pkthdr.pf.prio = ifp->if_llprio;
879 	MH_ALIGN(m, sizeof(*ea));
880 	ea = mtod(m, struct ether_arp *);
881 	eh = (struct ether_header *)sa.sa_data;
882 	memset(ea, 0, sizeof(*ea));
883 	memcpy(eh->ether_dhost, etherbroadcastaddr, sizeof(eh->ether_dhost));
884 	eh->ether_type = htons(ETHERTYPE_REVARP);
885 	ea->arp_hrd = htons(ARPHRD_ETHER);
886 	ea->arp_pro = htons(ETHERTYPE_IP);
887 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
888 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
889 	ea->arp_op = htons(ARPOP_REVREQUEST);
890 	memcpy(eh->ether_shost, ac->ac_enaddr, sizeof(ea->arp_tha));
891 	memcpy(ea->arp_sha, ac->ac_enaddr, sizeof(ea->arp_sha));
892 	memcpy(ea->arp_tha, ac->ac_enaddr, sizeof(ea->arp_tha));
893 	sa.sa_family = pseudo_AF_HDRCMPLT;
894 	sa.sa_len = sizeof(sa);
895 	m->m_flags |= M_BCAST;
896 	ifp->if_output(ifp, m, &sa, NULL);
897 }
898 
899 #ifdef NFSCLIENT
900 /*
901  * RARP for the ip address of the specified interface, but also
902  * save the ip address of the server that sent the answer.
903  * Timeout if no response is received.
904  */
905 int
906 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
907     struct in_addr *clnt_in)
908 {
909 	int result, count = 20;
910 
911 	if (revarp_finished)
912 		return EIO;
913 
914 	revarp_ifidx = ifp->if_index;
915 	while (count--) {
916 		revarprequest(ifp);
917 		result = tsleep((caddr_t)&revarp_myip, PSOCK, "revarp", hz/2);
918 		if (result != EWOULDBLOCK)
919 			break;
920 	}
921 	revarp_ifidx = 0;
922 	if (!revarp_finished)
923 		return ENETUNREACH;
924 
925 	memcpy(serv_in, &revarp_srvip, sizeof(*serv_in));
926 	memcpy(clnt_in, &revarp_myip, sizeof(*clnt_in));
927 	return 0;
928 }
929 
930 /* For compatibility: only saves interface address. */
931 int
932 revarpwhoami(struct in_addr *in, struct ifnet *ifp)
933 {
934 	struct in_addr server;
935 	return (revarpwhoarewe(ifp, &server, in));
936 }
937 #endif /* NFSCLIENT */
938