xref: /netbsd-src/sys/netinet/if_arp.c (revision 21e37cc72a480a47828990a439cde7ac9ffaf0c6)
1 /*	$NetBSD: if_arp.c,v 1.98 2004/05/25 04:33:59 atatat Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Public Access Networks Corporation ("Panix").  It was developed under
9  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1988, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the University nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  *
68  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
69  */
70 
71 /*
72  * Ethernet address resolution protocol.
73  * TODO:
74  *	add "inuse/lock" bit (or ref. count) along with valid bit
75  */
76 
77 #include <sys/cdefs.h>
78 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.98 2004/05/25 04:33:59 atatat Exp $");
79 
80 #include "opt_ddb.h"
81 #include "opt_inet.h"
82 
83 #ifdef INET
84 
85 #include "bridge.h"
86 
87 #include <sys/param.h>
88 #include <sys/systm.h>
89 #include <sys/callout.h>
90 #include <sys/malloc.h>
91 #include <sys/mbuf.h>
92 #include <sys/socket.h>
93 #include <sys/time.h>
94 #include <sys/kernel.h>
95 #include <sys/errno.h>
96 #include <sys/ioctl.h>
97 #include <sys/syslog.h>
98 #include <sys/proc.h>
99 #include <sys/protosw.h>
100 #include <sys/domain.h>
101 #include <sys/sysctl.h>
102 
103 #include <net/ethertypes.h>
104 #include <net/if.h>
105 #include <net/if_dl.h>
106 #include <net/if_token.h>
107 #include <net/if_types.h>
108 #include <net/route.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/in_var.h>
113 #include <netinet/ip.h>
114 #include <netinet/if_inarp.h>
115 
116 #include "loop.h"
117 #include "arc.h"
118 #if NARC > 0
119 #include <net/if_arc.h>
120 #endif
121 #include "fddi.h"
122 #if NFDDI > 0
123 #include <net/if_fddi.h>
124 #endif
125 #include "token.h"
126 
127 #define SIN(s) ((struct sockaddr_in *)s)
128 #define SDL(s) ((struct sockaddr_dl *)s)
129 #define SRP(s) ((struct sockaddr_inarp *)s)
130 
131 /*
132  * ARP trailer negotiation.  Trailer protocol is not IP specific,
133  * but ARP request/response use IP addresses.
134  */
135 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
136 
137 /* timer values */
138 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
139 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
140 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
141 int	arpt_refresh = (5*60);	/* time left before refreshing */
142 #define	rt_expire rt_rmx.rmx_expire
143 #define	rt_pksent rt_rmx.rmx_pksent
144 
145 extern	struct domain arpdomain;
146 
147 static	void arprequest __P((struct ifnet *,
148 	    struct in_addr *, struct in_addr *, u_int8_t *));
149 static	void arptfree __P((struct llinfo_arp *));
150 static	void arptimer __P((void *));
151 static	struct llinfo_arp *arplookup __P((struct mbuf *, struct in_addr *,
152 					  int, int));
153 static	void in_arpinput __P((struct mbuf *));
154 
155 #if NLOOP > 0
156 extern	struct ifnet loif[NLOOP];
157 #endif
158 LIST_HEAD(, llinfo_arp) llinfo_arp;
159 struct	ifqueue arpintrq = {0, 0, 0, 50};
160 int	arp_inuse, arp_allocated, arp_intimer;
161 int	arp_maxtries = 5;
162 int	useloopback = 1;	/* use loopback interface for local traffic */
163 int	arpinit_done = 0;
164 
165 struct	arpstat arpstat;
166 struct	callout arptimer_ch;
167 
168 
169 /* revarp state */
170 static struct	in_addr myip, srv_ip;
171 static int	myip_initialized = 0;
172 static int	revarp_in_progress = 0;
173 static struct	ifnet *myip_ifp = NULL;
174 
175 #ifdef DDB
176 static void db_print_sa __P((const struct sockaddr *));
177 static void db_print_ifa __P((struct ifaddr *));
178 static void db_print_llinfo __P((caddr_t));
179 static int db_show_radix_node __P((struct radix_node *, void *));
180 #endif
181 
182 /*
183  * this should be elsewhere.
184  */
185 
186 static char *
187 lla_snprintf __P((u_int8_t *, int));
188 
189 static char *
190 lla_snprintf(adrp, len)
191 	u_int8_t *adrp;
192 	int len;
193 {
194 #define NUMBUFS 3
195 	static char buf[NUMBUFS][16*3];
196 	static int bnum = 0;
197 	static const char hexdigits[] = {
198 	    '0','1','2','3','4','5','6','7',
199 	    '8','9','a','b','c','d','e','f'
200 	};
201 
202 	int i;
203 	char *p;
204 
205 	p = buf[bnum];
206 
207 	*p++ = hexdigits[(*adrp)>>4];
208 	*p++ = hexdigits[(*adrp++)&0xf];
209 
210 	for (i=1; i<len && i<16; i++) {
211 		*p++ = ':';
212 		*p++ = hexdigits[(*adrp)>>4];
213 		*p++ = hexdigits[(*adrp++)&0xf];
214 	}
215 
216 	*p = 0;
217 	p = buf[bnum];
218 	bnum = (bnum + 1) % NUMBUFS;
219 	return p;
220 }
221 
222 const struct protosw arpsw[] = {
223 	{ 0, 0, 0, 0,
224 	  0, 0, 0, 0,
225 	  0,
226 	  0, 0, 0, arp_drain,
227 	}
228 };
229 
230 
231 struct domain arpdomain =
232 { 	PF_ARP,  "arp", 0, 0, 0,
233 	arpsw, &arpsw[sizeof(arpsw)/sizeof(arpsw[0])]
234 };
235 
236 /*
237  * ARP table locking.
238  *
239  * to prevent lossage vs. the arp_drain routine (which may be called at
240  * any time, including in a device driver context), we do two things:
241  *
242  * 1) manipulation of la->la_hold is done at splnet() (for all of
243  * about two instructions).
244  *
245  * 2) manipulation of the arp table's linked list is done under the
246  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
247  * while the arp table is locked, we punt and try again later.
248  */
249 
250 static int	arp_locked;
251 static __inline int arp_lock_try __P((int));
252 static __inline void arp_unlock __P((void));
253 
254 static __inline int
255 arp_lock_try(int recurse)
256 {
257 	int s;
258 
259 	/*
260 	 * Use splvm() -- we're blocking things that would cause
261 	 * mbuf allocation.
262 	 */
263 	s = splvm();
264 	if (!recurse && arp_locked) {
265 		splx(s);
266 		return (0);
267 	}
268 	arp_locked++;
269 	splx(s);
270 	return (1);
271 }
272 
273 static __inline void
274 arp_unlock()
275 {
276 	int s;
277 
278 	s = splvm();
279 	arp_locked--;
280 	splx(s);
281 }
282 
283 #ifdef DIAGNOSTIC
284 #define	ARP_LOCK(recurse)						\
285 do {									\
286 	if (arp_lock_try(recurse) == 0) {				\
287 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
288 		panic("arp_lock");					\
289 	}								\
290 } while (/*CONSTCOND*/ 0)
291 #define	ARP_LOCK_CHECK()						\
292 do {									\
293 	if (arp_locked == 0) {						\
294 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
295 		panic("arp lock check");				\
296 	}								\
297 } while (/*CONSTCOND*/ 0)
298 #else
299 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
300 #define	ARP_LOCK_CHECK()	/* nothing */
301 #endif
302 
303 #define	ARP_UNLOCK()		arp_unlock()
304 
305 /*
306  * ARP protocol drain routine.  Called when memory is in short supply.
307  * Called at splvm();
308  */
309 
310 void
311 arp_drain()
312 {
313 	struct llinfo_arp *la, *nla;
314 	int count = 0;
315 	struct mbuf *mold;
316 
317 	if (arp_lock_try(0) == 0) {
318 		printf("arp_drain: locked; punting\n");
319 		return;
320 	}
321 
322 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
323 		nla = LIST_NEXT(la, la_list);
324 
325 		mold = la->la_hold;
326 		la->la_hold = 0;
327 
328 		if (mold) {
329 			m_freem(mold);
330 			count++;
331 		}
332 	}
333 	ARP_UNLOCK();
334 	arpstat.as_dfrdropped += count;
335 }
336 
337 
338 /*
339  * Timeout routine.  Age arp_tab entries periodically.
340  */
341 /* ARGSUSED */
342 static void
343 arptimer(arg)
344 	void *arg;
345 {
346 	int s;
347 	struct llinfo_arp *la, *nla;
348 
349 	s = splsoftnet();
350 
351 	if (arp_lock_try(0) == 0) {
352 		/* get it later.. */
353 		splx(s);
354 		return;
355 	}
356 
357 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
358 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
359 		struct rtentry *rt = la->la_rt;
360 
361 		nla = LIST_NEXT(la, la_list);
362 		if (rt->rt_expire == 0)
363 			continue;
364 		if ((rt->rt_expire - time.tv_sec) < arpt_refresh &&
365 		    rt->rt_pksent > (time.tv_sec - arpt_keep)) {
366 			/*
367 			 * If the entry has been used during since last
368 			 * refresh, try to renew it before deleting.
369 			 */
370 			arprequest(rt->rt_ifp,
371 			    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
372 			    &SIN(rt_key(rt))->sin_addr,
373 			    LLADDR(rt->rt_ifp->if_sadl));
374 		} else if (rt->rt_expire <= time.tv_sec)
375 			arptfree(la); /* timer has expired; clear */
376 	}
377 
378 	ARP_UNLOCK();
379 
380 	splx(s);
381 }
382 
383 /*
384  * Parallel to llc_rtrequest.
385  */
386 void
387 arp_rtrequest(req, rt, info)
388 	int req;
389 	struct rtentry *rt;
390 	struct rt_addrinfo *info;
391 {
392 	struct sockaddr *gate = rt->rt_gateway;
393 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
394 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
395 	size_t allocsize;
396 	struct mbuf *mold;
397 	int s;
398 	struct in_ifaddr *ia;
399 	struct ifaddr *ifa;
400 
401 	if (!arpinit_done) {
402 		arpinit_done = 1;
403 		/*
404 		 * We generate expiration times from time.tv_sec
405 		 * so avoid accidently creating permanent routes.
406 		 */
407 		if (time.tv_sec == 0) {
408 			time.tv_sec++;
409 		}
410 		callout_init(&arptimer_ch);
411 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
412 	}
413 
414 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
415 		if (req != RTM_ADD)
416 			return;
417 
418 		/*
419 		 * linklayers with particular link MTU limitation.
420 		 */
421 		switch(rt->rt_ifp->if_type) {
422 #if NFDDI > 0
423 		case IFT_FDDI:
424 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
425 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
426 			break;
427 #endif
428 #if NARC > 0
429 		case IFT_ARCNET:
430 		    {
431 			int arcipifmtu;
432 
433 			if (rt->rt_ifp->if_flags & IFF_LINK0)
434 				arcipifmtu = arc_ipmtu;
435 			else
436 				arcipifmtu = ARCMTU;
437 			if (rt->rt_ifp->if_mtu > arcipifmtu)
438 				rt->rt_rmx.rmx_mtu = arcipifmtu;
439 			break;
440 		    }
441 #endif
442 		}
443 		return;
444 	}
445 
446 	ARP_LOCK(1);		/* we may already be locked here. */
447 
448 	switch (req) {
449 
450 	case RTM_ADD:
451 		/*
452 		 * XXX: If this is a manually added route to interface
453 		 * such as older version of routed or gated might provide,
454 		 * restore cloning bit.
455 		 */
456 		if ((rt->rt_flags & RTF_HOST) == 0 &&
457 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
458 			rt->rt_flags |= RTF_CLONING;
459 		if (rt->rt_flags & RTF_CLONING) {
460 			/*
461 			 * Case 1: This route should come from a route to iface.
462 			 */
463 			rt_setgate(rt, rt_key(rt),
464 					(struct sockaddr *)&null_sdl);
465 			gate = rt->rt_gateway;
466 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
467 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
468 			/*
469 			 * Give this route an expiration time, even though
470 			 * it's a "permanent" route, so that routes cloned
471 			 * from it do not need their expiration time set.
472 			 */
473 			rt->rt_expire = time.tv_sec;
474 			/*
475 			 * linklayers with particular link MTU limitation.
476 			 */
477 			switch (rt->rt_ifp->if_type) {
478 #if NFDDI > 0
479 			case IFT_FDDI:
480 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
481 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
482 				     (rt->rt_rmx.rmx_mtu == 0 &&
483 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
484 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
485 				break;
486 #endif
487 #if NARC > 0
488 			case IFT_ARCNET:
489 			    {
490 				int arcipifmtu;
491 				if (rt->rt_ifp->if_flags & IFF_LINK0)
492 					arcipifmtu = arc_ipmtu;
493 				else
494 					arcipifmtu = ARCMTU;
495 
496 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
497 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
498 				     (rt->rt_rmx.rmx_mtu == 0 &&
499 				      rt->rt_ifp->if_mtu > arcipifmtu)))
500 					rt->rt_rmx.rmx_mtu = arcipifmtu;
501 				break;
502 			    }
503 #endif
504 			}
505 			break;
506 		}
507 		/* Announce a new entry if requested. */
508 		if (rt->rt_flags & RTF_ANNOUNCE)
509 			arprequest(rt->rt_ifp,
510 			    &SIN(rt_key(rt))->sin_addr,
511 			    &SIN(rt_key(rt))->sin_addr,
512 			    (u_char *)LLADDR(SDL(gate)));
513 		/*FALLTHROUGH*/
514 	case RTM_RESOLVE:
515 		if (gate->sa_family != AF_LINK ||
516 		    gate->sa_len < sizeof(null_sdl)) {
517 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
518 			break;
519 		}
520 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
521 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
522 		if (la != 0)
523 			break; /* This happens on a route change */
524 		/*
525 		 * Case 2:  This route may come from cloning, or a manual route
526 		 * add with a LL address.
527 		 */
528 		switch (SDL(gate)->sdl_type) {
529 #if NTOKEN > 0
530 		case IFT_ISO88025:
531 			allocsize = sizeof(*la) + sizeof(struct token_rif);
532 			break;
533 #endif /* NTOKEN > 0 */
534 		default:
535 			allocsize = sizeof(*la);
536 		}
537 		R_Malloc(la, struct llinfo_arp *, allocsize);
538 		rt->rt_llinfo = (caddr_t)la;
539 		if (la == 0) {
540 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
541 			break;
542 		}
543 		arp_inuse++, arp_allocated++;
544 		Bzero(la, allocsize);
545 		la->la_rt = rt;
546 		rt->rt_flags |= RTF_LLINFO;
547 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
548 
549 		INADDR_TO_IA(SIN(rt_key(rt))->sin_addr, ia);
550 		while (ia && ia->ia_ifp != rt->rt_ifp)
551 			NEXT_IA_WITH_SAME_ADDR(ia);
552 		if (ia) {
553 			/*
554 			 * This test used to be
555 			 *	if (loif.if_flags & IFF_UP)
556 			 * It allowed local traffic to be forced through
557 			 * the hardware by configuring the loopback down.
558 			 * However, it causes problems during network
559 			 * configuration for boards that can't receive
560 			 * packets they send.  It is now necessary to clear
561 			 * "useloopback" and remove the route to force
562 			 * traffic out to the hardware.
563 			 *
564 			 * In 4.4BSD, the above "if" statement checked
565 			 * rt->rt_ifa against rt_key(rt).  It was changed
566 			 * to the current form so that we can provide a
567 			 * better support for multiple IPv4 addresses on a
568 			 * interface.
569 			 */
570 			rt->rt_expire = 0;
571 			Bcopy(LLADDR(rt->rt_ifp->if_sadl),
572 			    LLADDR(SDL(gate)),
573 			    SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
574 #if NLOOP > 0
575 			if (useloopback)
576 				rt->rt_ifp = &loif[0];
577 #endif
578 			/*
579 			 * make sure to set rt->rt_ifa to the interface
580 			 * address we are using, otherwise we will have trouble
581 			 * with source address selection.
582 			 */
583 			ifa = &ia->ia_ifa;
584 			if (ifa != rt->rt_ifa) {
585 				IFAFREE(rt->rt_ifa);
586 				IFAREF(ifa);
587 				rt->rt_ifa = ifa;
588 			}
589 		}
590 		break;
591 
592 	case RTM_DELETE:
593 		if (la == 0)
594 			break;
595 		arp_inuse--;
596 		LIST_REMOVE(la, la_list);
597 		rt->rt_llinfo = 0;
598 		rt->rt_flags &= ~RTF_LLINFO;
599 
600 		s = splnet();
601 		mold = la->la_hold;
602 		la->la_hold = 0;
603 		splx(s);
604 
605 		if (mold)
606 			m_freem(mold);
607 
608 		Free((caddr_t)la);
609 	}
610 	ARP_UNLOCK();
611 }
612 
613 /*
614  * Broadcast an ARP request. Caller specifies:
615  *	- arp header source ip address
616  *	- arp header target ip address
617  *	- arp header source ethernet address
618  */
619 static void
620 arprequest(ifp, sip, tip, enaddr)
621 	struct ifnet *ifp;
622 	struct in_addr *sip, *tip;
623 	u_int8_t *enaddr;
624 {
625 	struct mbuf *m;
626 	struct arphdr *ah;
627 	struct sockaddr sa;
628 
629 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
630 		return;
631 	MCLAIM(m, &arpdomain.dom_mowner);
632 	switch (ifp->if_type) {
633 	case IFT_IEEE1394:
634 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
635 		    ifp->if_addrlen;
636 		break;
637 	default:
638 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
639 		    2 * ifp->if_addrlen;
640 		break;
641 	}
642 	m->m_pkthdr.len = m->m_len;
643 	MH_ALIGN(m, m->m_len);
644 	ah = mtod(m, struct arphdr *);
645 	bzero((caddr_t)ah, m->m_len);
646 	switch (ifp->if_type) {
647 	case IFT_IEEE1394:	/* RFC2734 */
648 		/* fill it now for ar_tpa computation */
649 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
650 		break;
651 	default:
652 		/* ifp->if_output will fill ar_hrd */
653 		break;
654 	}
655 	ah->ar_pro = htons(ETHERTYPE_IP);
656 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
657 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
658 	ah->ar_op = htons(ARPOP_REQUEST);
659 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
660 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
661 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
662 	sa.sa_family = AF_ARP;
663 	sa.sa_len = 2;
664 	m->m_flags |= M_BCAST;
665 	arpstat.as_sndtotal++;
666 	arpstat.as_sndrequest++;
667 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
668 }
669 
670 /*
671  * Resolve an IP address into an ethernet address.  If success,
672  * desten is filled in.  If there is no entry in arptab,
673  * set one up and broadcast a request for the IP address.
674  * Hold onto this mbuf and resend it once the address
675  * is finally resolved.  A return value of 1 indicates
676  * that desten has been filled in and the packet should be sent
677  * normally; a 0 return indicates that the packet has been
678  * taken over here, either now or for later transmission.
679  */
680 int
681 arpresolve(ifp, rt, m, dst, desten)
682 	struct ifnet *ifp;
683 	struct rtentry *rt;
684 	struct mbuf *m;
685 	struct sockaddr *dst;
686 	u_char *desten;
687 {
688 	struct llinfo_arp *la;
689 	struct sockaddr_dl *sdl;
690 	struct mbuf *mold;
691 	int s;
692 
693 	if (rt)
694 		la = (struct llinfo_arp *)rt->rt_llinfo;
695 	else {
696 		if ((la = arplookup(m, &SIN(dst)->sin_addr, 1, 0)) != NULL)
697 			rt = la->la_rt;
698 	}
699 	if (la == 0 || rt == 0) {
700 		arpstat.as_allocfail++;
701 		log(LOG_DEBUG,
702 		    "arpresolve: can't allocate llinfo on %s for %s\n",
703 		    ifp->if_xname, in_fmtaddr(SIN(dst)->sin_addr));
704 		m_freem(m);
705 		return (0);
706 	}
707 	sdl = SDL(rt->rt_gateway);
708 	/*
709 	 * Check the address family and length is valid, the address
710 	 * is resolved; otherwise, try to resolve.
711 	 */
712 	if ((rt->rt_expire == 0 || rt->rt_expire > time.tv_sec) &&
713 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
714 		bcopy(LLADDR(sdl), desten,
715 		    min(sdl->sdl_alen, ifp->if_addrlen));
716 		rt->rt_pksent = time.tv_sec; /* Time for last pkt sent */
717 		return 1;
718 	}
719 	/*
720 	 * There is an arptab entry, but no ethernet address
721 	 * response yet.  Replace the held mbuf with this
722 	 * latest one.
723 	 */
724 
725 	arpstat.as_dfrtotal++;
726 	s = splnet();
727 	mold = la->la_hold;
728 	la->la_hold = m;
729 	splx(s);
730 
731 	if (mold) {
732 		arpstat.as_dfrdropped++;
733 		m_freem(mold);
734 	}
735 
736 	/*
737 	 * Re-send the ARP request when appropriate.
738 	 */
739 #ifdef	DIAGNOSTIC
740 	if (rt->rt_expire == 0) {
741 		/* This should never happen. (Should it? -gwr) */
742 		printf("arpresolve: unresolved and rt_expire == 0\n");
743 		/* Set expiration time to now (expired). */
744 		rt->rt_expire = time.tv_sec;
745 	}
746 #endif
747 	if (rt->rt_expire) {
748 		rt->rt_flags &= ~RTF_REJECT;
749 		if (la->la_asked == 0 || rt->rt_expire != time.tv_sec) {
750 			rt->rt_expire = time.tv_sec;
751 			if (la->la_asked++ < arp_maxtries)
752 				arprequest(ifp,
753 				    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
754 				    &SIN(dst)->sin_addr,
755 				    LLADDR(ifp->if_sadl));
756 			else {
757 				rt->rt_flags |= RTF_REJECT;
758 				rt->rt_expire += arpt_down;
759 				la->la_asked = 0;
760 			}
761 		}
762 	}
763 	return (0);
764 }
765 
766 /*
767  * Common length and type checks are done here,
768  * then the protocol-specific routine is called.
769  */
770 void
771 arpintr()
772 {
773 	struct mbuf *m;
774 	struct arphdr *ar;
775 	int s;
776 	int arplen;
777 
778 	while (arpintrq.ifq_head) {
779 		s = splnet();
780 		IF_DEQUEUE(&arpintrq, m);
781 		splx(s);
782 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
783 			panic("arpintr");
784 
785 		MCLAIM(m, &arpdomain.dom_mowner);
786 		arpstat.as_rcvtotal++;
787 
788 		/*
789 		 * First, make sure we have at least struct arphdr.
790 		 */
791 		if (m->m_len < sizeof(struct arphdr) ||
792 		    (ar = mtod(m, struct arphdr *)) == NULL)
793 			goto badlen;
794 
795 		switch (m->m_pkthdr.rcvif->if_type) {
796 		case IFT_IEEE1394:
797 			arplen = sizeof(struct arphdr) +
798 			    ar->ar_hln + 2 * ar->ar_pln;
799 			break;
800 		default:
801 			arplen = sizeof(struct arphdr) +
802 			    2 * ar->ar_hln + 2 * ar->ar_pln;
803 			break;
804 		}
805 
806 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
807 		    m->m_len >= arplen)
808 			switch (ntohs(ar->ar_pro)) {
809 			case ETHERTYPE_IP:
810 			case ETHERTYPE_IPTRAILERS:
811 				in_arpinput(m);
812 				continue;
813 			default:
814 				arpstat.as_rcvbadproto++;
815 			}
816 		else {
817 badlen:
818 			arpstat.as_rcvbadlen++;
819 		}
820 		m_freem(m);
821 	}
822 }
823 
824 /*
825  * ARP for Internet protocols on 10 Mb/s Ethernet.
826  * Algorithm is that given in RFC 826.
827  * In addition, a sanity check is performed on the sender
828  * protocol address, to catch impersonators.
829  * We no longer handle negotiations for use of trailer protocol:
830  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
831  * along with IP replies if we wanted trailers sent to us,
832  * and also sent them in response to IP replies.
833  * This allowed either end to announce the desire to receive
834  * trailer packets.
835  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
836  * but formerly didn't normally send requests.
837  */
838 static void
839 in_arpinput(m)
840 	struct mbuf *m;
841 {
842 	struct arphdr *ah;
843 	struct ifnet *ifp = m->m_pkthdr.rcvif;
844 	struct llinfo_arp *la = 0;
845 	struct rtentry  *rt;
846 	struct in_ifaddr *ia;
847 #if NBRIDGE > 0
848 	struct in_ifaddr *bridge_ia = NULL;
849 #endif
850 	struct sockaddr_dl *sdl;
851 	struct sockaddr sa;
852 	struct in_addr isaddr, itaddr, myaddr;
853 	int op;
854 	struct mbuf *mold;
855 	int s;
856 
857 	ah = mtod(m, struct arphdr *);
858 	op = ntohs(ah->ar_op);
859 
860 	/*
861 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
862 	 * ar_tpa().
863 	 */
864 	switch (ifp->if_type) {
865 	case IFT_IEEE1394:
866 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
867 			;
868 		else {
869 			/* XXX this is to make sure we compute ar_tha right */
870 			/* XXX check ar_hrd more strictly? */
871 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
872 		}
873 		break;
874 	default:
875 		/* XXX check ar_hrd? */
876 		break;
877 	}
878 
879 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
880 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
881 
882 	if (m->m_flags & (M_BCAST|M_MCAST))
883 		arpstat.as_rcvmcast++;
884 
885 	/*
886 	 * If the target IP address is zero, ignore the packet.
887 	 * This prevents the code below from tring to answer
888 	 * when we are using IP address zero (booting).
889 	 */
890 	if (in_nullhost(itaddr)) {
891 		arpstat.as_rcvzerotpa++;
892 		goto out;
893 	}
894 
895 	/*
896 	 * If the source IP address is zero, this is most likely a
897 	 * confused host trying to use IP address zero. (Windoze?)
898 	 * XXX: Should we bother trying to reply to these?
899 	 */
900 	if (in_nullhost(isaddr)) {
901 		arpstat.as_rcvzerospa++;
902 		goto out;
903 	}
904 
905 	/*
906 	 * Search for a matching interface address
907 	 * or any address on the interface to use
908 	 * as a dummy address in the rest of this function
909 	 */
910 	INADDR_TO_IA(itaddr, ia);
911 	while (ia != NULL) {
912 		if (ia->ia_ifp == m->m_pkthdr.rcvif)
913 			break;
914 
915 #if NBRIDGE > 0
916 		/*
917 		 * If the interface we received the packet on
918 		 * is part of a bridge, check to see if we need
919 		 * to "bridge" the packet to ourselves at this
920 		 * layer.  Note we still prefer a perfect match,
921 		 * but allow this weaker match if necessary.
922 		 */
923 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
924 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
925 			bridge_ia = ia;
926 #endif /* NBRIDGE > 0 */
927 
928 		NEXT_IA_WITH_SAME_ADDR(ia);
929 	}
930 
931 #if NBRIDGE > 0
932 	if (ia == NULL && bridge_ia != NULL) {
933 		ia = bridge_ia;
934 		ifp = bridge_ia->ia_ifp;
935 	}
936 #endif
937 
938 	if (ia == NULL) {
939 		INADDR_TO_IA(isaddr, ia);
940 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
941 			NEXT_IA_WITH_SAME_ADDR(ia);
942 
943 		if (ia == NULL) {
944 			IFP_TO_IA(ifp, ia);
945 			if (ia == NULL) {
946 				arpstat.as_rcvnoint++;
947 				goto out;
948 			}
949 		}
950 	}
951 
952 	myaddr = ia->ia_addr.sin_addr;
953 
954 	/* XXX checks for bridge case? */
955 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
956 	    ifp->if_addrlen)) {
957 		arpstat.as_rcvlocalsha++;
958 		goto out;	/* it's from me, ignore it. */
959 	}
960 
961 	/* XXX checks for bridge case? */
962 	if (!bcmp((caddr_t)ar_sha(ah), (caddr_t)ifp->if_broadcastaddr,
963 	    ifp->if_addrlen)) {
964 		arpstat.as_rcvbcastsha++;
965 		log(LOG_ERR,
966 		    "%s: arp: link address is broadcast for IP address %s!\n",
967 		    ifp->if_xname, in_fmtaddr(isaddr));
968 		goto out;
969 	}
970 
971 	if (in_hosteq(isaddr, myaddr)) {
972 		arpstat.as_rcvlocalspa++;
973 		log(LOG_ERR,
974 		   "duplicate IP address %s sent from link address %s\n",
975 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
976 		itaddr = myaddr;
977 		goto reply;
978 	}
979 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
980 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
981 		if (sdl->sdl_alen &&
982 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
983 			if (rt->rt_flags & RTF_STATIC) {
984 				arpstat.as_rcvoverperm++;
985 				log(LOG_INFO,
986 				    "%s tried to overwrite permanent arp info"
987 				    " for %s\n",
988 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
989 				    in_fmtaddr(isaddr));
990 				goto out;
991 			} else if (rt->rt_ifp != ifp) {
992 				arpstat.as_rcvoverint++;
993 				log(LOG_INFO,
994 				    "%s on %s tried to overwrite "
995 				    "arp info for %s on %s\n",
996 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
997 				    ifp->if_xname, in_fmtaddr(isaddr),
998 				    rt->rt_ifp->if_xname);
999 				    goto out;
1000 			} else {
1001 				arpstat.as_rcvover++;
1002 				log(LOG_INFO,
1003 				    "arp info overwritten for %s by %s\n",
1004 				    in_fmtaddr(isaddr),
1005 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1006 			}
1007 		}
1008 		/*
1009 		 * sanity check for the address length.
1010 		 * XXX this does not work for protocols with variable address
1011 		 * length. -is
1012 		 */
1013 		if (sdl->sdl_alen &&
1014 		    sdl->sdl_alen != ah->ar_hln) {
1015 			arpstat.as_rcvlenchg++;
1016 			log(LOG_WARNING,
1017 			    "arp from %s: new addr len %d, was %d",
1018 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1019 		}
1020 		if (ifp->if_addrlen != ah->ar_hln) {
1021 			arpstat.as_rcvbadlen++;
1022 			log(LOG_WARNING,
1023 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1024 			    in_fmtaddr(isaddr), ah->ar_hln,
1025 			    ifp->if_addrlen);
1026 			goto reply;
1027 		}
1028 #if NTOKEN > 0
1029 		/*
1030 		 * XXX uses m_data and assumes the complete answer including
1031 		 * XXX token-ring headers is in the same buf
1032 		 */
1033 		if (ifp->if_type == IFT_ISO88025) {
1034 			struct token_header *trh;
1035 
1036 			trh = (struct token_header *)M_TRHSTART(m);
1037 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1038 				struct token_rif	*rif;
1039 				size_t	riflen;
1040 
1041 				rif = TOKEN_RIF(trh);
1042 				riflen = (ntohs(rif->tr_rcf) &
1043 				    TOKEN_RCF_LEN_MASK) >> 8;
1044 
1045 				if (riflen > 2 &&
1046 				    riflen < sizeof(struct token_rif) &&
1047 				    (riflen & 1) == 0) {
1048 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1049 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1050 					bcopy(rif, TOKEN_RIF(la), riflen);
1051 				}
1052 			}
1053 		}
1054 #endif /* NTOKEN > 0 */
1055 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
1056 		    sdl->sdl_alen = ah->ar_hln);
1057 		if (rt->rt_expire)
1058 			rt->rt_expire = time.tv_sec + arpt_keep;
1059 		rt->rt_flags &= ~RTF_REJECT;
1060 		la->la_asked = 0;
1061 
1062 		s = splnet();
1063 		mold = la->la_hold;
1064 		la->la_hold = 0;
1065 		splx(s);
1066 
1067 		if (mold) {
1068 			arpstat.as_dfrsent++;
1069 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
1070 		}
1071 	}
1072 reply:
1073 	if (op != ARPOP_REQUEST) {
1074 		if (op == ARPOP_REPLY)
1075 			arpstat.as_rcvreply++;
1076 	out:
1077 		m_freem(m);
1078 		return;
1079 	}
1080 	arpstat.as_rcvrequest++;
1081 	if (in_hosteq(itaddr, myaddr)) {
1082 		/* I am the target */
1083 		if (ar_tha(ah))
1084 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1085 			    ah->ar_hln);
1086 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1087 	} else {
1088 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1089 		if (la == 0)
1090 			goto out;
1091 		rt = la->la_rt;
1092 		if (ar_tha(ah))
1093 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1094 			    ah->ar_hln);
1095 		sdl = SDL(rt->rt_gateway);
1096 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
1097 	}
1098 
1099 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
1100 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
1101 	ah->ar_op = htons(ARPOP_REPLY);
1102 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1103 	switch (ifp->if_type) {
1104 	case IFT_IEEE1394:
1105 		/*
1106 		 * ieee1394 arp reply is broadcast
1107 		 */
1108 		m->m_flags &= ~M_MCAST;
1109 		m->m_flags |= M_BCAST;
1110 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1111 		break;
1112 
1113 	default:
1114 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1115 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1116 		break;
1117 	}
1118 	m->m_pkthdr.len = m->m_len;
1119 	sa.sa_family = AF_ARP;
1120 	sa.sa_len = 2;
1121 	arpstat.as_sndtotal++;
1122 	arpstat.as_sndreply++;
1123 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1124 	return;
1125 }
1126 
1127 /*
1128  * Free an arp entry.
1129  */
1130 static void
1131 arptfree(la)
1132 	struct llinfo_arp *la;
1133 {
1134 	struct rtentry *rt = la->la_rt;
1135 	struct sockaddr_dl *sdl;
1136 
1137 	ARP_LOCK_CHECK();
1138 
1139 	if (rt == 0)
1140 		panic("arptfree");
1141 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
1142 	    sdl->sdl_family == AF_LINK) {
1143 		sdl->sdl_alen = 0;
1144 		la->la_asked = 0;
1145 		rt->rt_flags &= ~RTF_REJECT;
1146 		return;
1147 	}
1148 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
1149 	    0, (struct rtentry **)0);
1150 }
1151 
1152 /*
1153  * Lookup or enter a new address in arptab.
1154  */
1155 static struct llinfo_arp *
1156 arplookup(m, addr, create, proxy)
1157 	struct mbuf *m;
1158 	struct in_addr *addr;
1159 	int create, proxy;
1160 {
1161 	struct arphdr *ah;
1162 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1163 	struct rtentry *rt;
1164 	static struct sockaddr_inarp sin;
1165 	const char *why = 0;
1166 
1167 	ah = mtod(m, struct arphdr *);
1168 	sin.sin_len = sizeof(sin);
1169 	sin.sin_family = AF_INET;
1170 	sin.sin_addr = *addr;
1171 	sin.sin_other = proxy ? SIN_PROXY : 0;
1172 	rt = rtalloc1(sintosa(&sin), create);
1173 	if (rt == 0)
1174 		return (0);
1175 	rt->rt_refcnt--;
1176 
1177 	if (rt->rt_flags & RTF_GATEWAY)
1178 		why = "host is not on local network";
1179 	else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1180 		arpstat.as_allocfail++;
1181 		why = "could not allocate llinfo";
1182 	} else if (rt->rt_gateway->sa_family != AF_LINK)
1183 		why = "gateway route is not ours";
1184 	else
1185 		return ((struct llinfo_arp *)rt->rt_llinfo);
1186 
1187 	if (create) {
1188 		log(LOG_DEBUG, "arplookup: unable to enter address"
1189 		    " for %s@%s on %s (%s)\n",
1190 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1191 		    (ifp) ? ifp->if_xname : 0, why);
1192 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1193 			rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
1194 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1195 		}
1196 	}
1197 	return (0);
1198 }
1199 
1200 int
1201 arpioctl(cmd, data)
1202 	u_long cmd;
1203 	caddr_t data;
1204 {
1205 
1206 	return (EOPNOTSUPP);
1207 }
1208 
1209 void
1210 arp_ifinit(ifp, ifa)
1211 	struct ifnet *ifp;
1212 	struct ifaddr *ifa;
1213 {
1214 	struct in_addr *ip;
1215 
1216 	/*
1217 	 * Warn the user if another station has this IP address,
1218 	 * but only if the interface IP address is not zero.
1219 	 */
1220 	ip = &IA_SIN(ifa)->sin_addr;
1221 	if (!in_nullhost(*ip))
1222 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
1223 
1224 	ifa->ifa_rtrequest = arp_rtrequest;
1225 	ifa->ifa_flags |= RTF_CLONING;
1226 }
1227 
1228 /*
1229  * Called from 10 Mb/s Ethernet interrupt handlers
1230  * when ether packet type ETHERTYPE_REVARP
1231  * is received.  Common length and type checks are done here,
1232  * then the protocol-specific routine is called.
1233  */
1234 void
1235 revarpinput(m)
1236 	struct mbuf *m;
1237 {
1238 	struct arphdr *ar;
1239 
1240 	if (m->m_len < sizeof(struct arphdr))
1241 		goto out;
1242 	ar = mtod(m, struct arphdr *);
1243 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1244 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1245 		goto out;
1246 #endif
1247 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1248 		goto out;
1249 	switch (ntohs(ar->ar_pro)) {
1250 	case ETHERTYPE_IP:
1251 	case ETHERTYPE_IPTRAILERS:
1252 		in_revarpinput(m);
1253 		return;
1254 
1255 	default:
1256 		break;
1257 	}
1258 out:
1259 	m_freem(m);
1260 }
1261 
1262 /*
1263  * RARP for Internet protocols on 10 Mb/s Ethernet.
1264  * Algorithm is that given in RFC 903.
1265  * We are only using for bootstrap purposes to get an ip address for one of
1266  * our interfaces.  Thus we support no user-interface.
1267  *
1268  * Since the contents of the RARP reply are specific to the interface that
1269  * sent the request, this code must ensure that they are properly associated.
1270  *
1271  * Note: also supports ARP via RARP packets, per the RFC.
1272  */
1273 void
1274 in_revarpinput(m)
1275 	struct mbuf *m;
1276 {
1277 	struct ifnet *ifp;
1278 	struct arphdr *ah;
1279 	int op;
1280 
1281 	ah = mtod(m, struct arphdr *);
1282 	op = ntohs(ah->ar_op);
1283 
1284 	switch (m->m_pkthdr.rcvif->if_type) {
1285 	case IFT_IEEE1394:
1286 		/* ARP without target hardware address is not supported */
1287 		goto out;
1288 	default:
1289 		break;
1290 	}
1291 
1292 	switch (op) {
1293 	case ARPOP_REQUEST:
1294 	case ARPOP_REPLY:	/* per RFC */
1295 		in_arpinput(m);
1296 		return;
1297 	case ARPOP_REVREPLY:
1298 		break;
1299 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1300 	default:
1301 		goto out;
1302 	}
1303 	if (!revarp_in_progress)
1304 		goto out;
1305 	ifp = m->m_pkthdr.rcvif;
1306 	if (ifp != myip_ifp) /* !same interface */
1307 		goto out;
1308 	if (myip_initialized)
1309 		goto wake;
1310 	if (bcmp(ar_tha(ah), LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1311 		goto out;
1312 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1313 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1314 	myip_initialized = 1;
1315 wake:	/* Do wakeup every time in case it was missed. */
1316 	wakeup((caddr_t)&myip);
1317 
1318 out:
1319 	m_freem(m);
1320 }
1321 
1322 /*
1323  * Send a RARP request for the ip address of the specified interface.
1324  * The request should be RFC 903-compliant.
1325  */
1326 void
1327 revarprequest(ifp)
1328 	struct ifnet *ifp;
1329 {
1330 	struct sockaddr sa;
1331 	struct mbuf *m;
1332 	struct arphdr *ah;
1333 
1334 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1335 		return;
1336 	MCLAIM(m, &arpdomain.dom_mowner);
1337 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1338 	    2*ifp->if_addrlen;
1339 	m->m_pkthdr.len = m->m_len;
1340 	MH_ALIGN(m, m->m_len);
1341 	ah = mtod(m, struct arphdr *);
1342 	bzero((caddr_t)ah, m->m_len);
1343 	ah->ar_pro = htons(ETHERTYPE_IP);
1344 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1345 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1346 	ah->ar_op = htons(ARPOP_REVREQUEST);
1347 
1348 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1349 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_tha(ah), ah->ar_hln);
1350 
1351 	sa.sa_family = AF_ARP;
1352 	sa.sa_len = 2;
1353 	m->m_flags |= M_BCAST;
1354 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1355 
1356 }
1357 
1358 /*
1359  * RARP for the ip address of the specified interface, but also
1360  * save the ip address of the server that sent the answer.
1361  * Timeout if no response is received.
1362  */
1363 int
1364 revarpwhoarewe(ifp, serv_in, clnt_in)
1365 	struct ifnet *ifp;
1366 	struct in_addr *serv_in;
1367 	struct in_addr *clnt_in;
1368 {
1369 	int result, count = 20;
1370 
1371 	myip_initialized = 0;
1372 	myip_ifp = ifp;
1373 
1374 	revarp_in_progress = 1;
1375 	while (count--) {
1376 		revarprequest(ifp);
1377 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1378 		if (result != EWOULDBLOCK)
1379 			break;
1380 	}
1381 	revarp_in_progress = 0;
1382 
1383 	if (!myip_initialized)
1384 		return ENETUNREACH;
1385 
1386 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1387 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1388 	return 0;
1389 }
1390 
1391 
1392 
1393 #ifdef DDB
1394 
1395 #include <machine/db_machdep.h>
1396 #include <ddb/db_interface.h>
1397 #include <ddb/db_output.h>
1398 static void
1399 db_print_sa(sa)
1400 	const struct sockaddr *sa;
1401 {
1402 	int len;
1403 	u_char *p;
1404 
1405 	if (sa == 0) {
1406 		db_printf("[NULL]");
1407 		return;
1408 	}
1409 
1410 	p = (u_char*)sa;
1411 	len = sa->sa_len;
1412 	db_printf("[");
1413 	while (len > 0) {
1414 		db_printf("%d", *p);
1415 		p++; len--;
1416 		if (len) db_printf(",");
1417 	}
1418 	db_printf("]\n");
1419 }
1420 static void
1421 db_print_ifa(ifa)
1422 	struct ifaddr *ifa;
1423 {
1424 	if (ifa == 0)
1425 		return;
1426 	db_printf("  ifa_addr=");
1427 	db_print_sa(ifa->ifa_addr);
1428 	db_printf("  ifa_dsta=");
1429 	db_print_sa(ifa->ifa_dstaddr);
1430 	db_printf("  ifa_mask=");
1431 	db_print_sa(ifa->ifa_netmask);
1432 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1433 			  ifa->ifa_flags,
1434 			  ifa->ifa_refcnt,
1435 			  ifa->ifa_metric);
1436 }
1437 static void
1438 db_print_llinfo(li)
1439 	caddr_t li;
1440 {
1441 	struct llinfo_arp *la;
1442 
1443 	if (li == 0)
1444 		return;
1445 	la = (struct llinfo_arp *)li;
1446 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1447 			  la->la_rt, la->la_hold, la->la_asked);
1448 }
1449 /*
1450  * Function to pass to rn_walktree().
1451  * Return non-zero error to abort walk.
1452  */
1453 static int
1454 db_show_radix_node(rn, w)
1455 	struct radix_node *rn;
1456 	void *w;
1457 {
1458 	struct rtentry *rt = (struct rtentry *)rn;
1459 
1460 	db_printf("rtentry=%p", rt);
1461 
1462 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1463 			  rt->rt_flags, rt->rt_refcnt,
1464 			  rt->rt_use, rt->rt_expire);
1465 
1466 	db_printf(" key="); db_print_sa(rt_key(rt));
1467 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1468 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1469 
1470 	db_printf(" ifp=%p ", rt->rt_ifp);
1471 	if (rt->rt_ifp)
1472 		db_printf("(%s)", rt->rt_ifp->if_xname);
1473 	else
1474 		db_printf("(NULL)");
1475 
1476 	db_printf(" ifa=%p\n", rt->rt_ifa);
1477 	db_print_ifa(rt->rt_ifa);
1478 
1479 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1480 
1481 	db_printf(" gwroute=%p llinfo=%p\n",
1482 			  rt->rt_gwroute, rt->rt_llinfo);
1483 	db_print_llinfo(rt->rt_llinfo);
1484 
1485 	return (0);
1486 }
1487 /*
1488  * Function to print all the route trees.
1489  * Use this from ddb:  "show arptab"
1490  */
1491 void
1492 db_show_arptab(addr, have_addr, count, modif)
1493 	db_expr_t	addr;
1494 	int		have_addr;
1495 	db_expr_t	count;
1496 	char *		modif;
1497 {
1498 	struct radix_node_head *rnh;
1499 	rnh = rt_tables[AF_INET];
1500 	db_printf("Route tree for AF_INET\n");
1501 	if (rnh == NULL) {
1502 		db_printf(" (not initialized)\n");
1503 		return;
1504 	}
1505 	rn_walktree(rnh, db_show_radix_node, NULL);
1506 	return;
1507 }
1508 #endif
1509 
1510 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1511 {
1512 	struct sysctlnode *node;
1513 
1514 	sysctl_createv(clog, 0, NULL, NULL,
1515 			CTLFLAG_PERMANENT,
1516 			CTLTYPE_NODE, "net", NULL,
1517 			NULL, 0, NULL, 0,
1518 			CTL_NET, CTL_EOL);
1519 	sysctl_createv(clog, 0, NULL, NULL,
1520 			CTLFLAG_PERMANENT,
1521 			CTLTYPE_NODE, "inet", NULL,
1522 			NULL, 0, NULL, 0,
1523 			CTL_NET, PF_INET, CTL_EOL);
1524 	sysctl_createv(clog, 0, NULL, &node,
1525 			CTLFLAG_PERMANENT,
1526 			CTLTYPE_NODE, "arp",
1527 			SYSCTL_DESCR("Address Resolution Protocol"),
1528 			NULL, 0, NULL, 0,
1529 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1530 
1531 	sysctl_createv(clog, 0, NULL, NULL,
1532 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1533 			CTLTYPE_INT, "prune",
1534 			SYSCTL_DESCR("ARP cache pruning interval"),
1535 			NULL, 0, &arpt_prune, 0,
1536 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1537 
1538 	sysctl_createv(clog, 0, NULL, NULL,
1539 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1540 			CTLTYPE_INT, "keep",
1541 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1542 			NULL, 0, &arpt_keep, 0,
1543 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1544 
1545 	sysctl_createv(clog, 0, NULL, NULL,
1546 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1547 			CTLTYPE_INT, "down",
1548 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1549 			NULL, 0, &arpt_down, 0,
1550 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1551 
1552 	sysctl_createv(clog, 0, NULL, NULL,
1553 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1554 			CTLTYPE_INT, "refresh",
1555 			SYSCTL_DESCR("ARP entry refresh interval"),
1556 			NULL, 0, &arpt_refresh, 0,
1557 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1558 }
1559 
1560 #endif /* INET */
1561