xref: /netbsd-src/sys/netinet/if_arp.c (revision db6316d1518382eecd2fdbe55a1205e0620a1b35)
1 /*	$NetBSD: if_arp.c,v 1.100 2004/12/04 16:10:25 peter 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 1.100 2004/12/04 16:10:25 peter 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 "arc.h"
117 #if NARC > 0
118 #include <net/if_arc.h>
119 #endif
120 #include "fddi.h"
121 #if NFDDI > 0
122 #include <net/if_fddi.h>
123 #endif
124 #include "token.h"
125 
126 #define SIN(s) ((struct sockaddr_in *)s)
127 #define SDL(s) ((struct sockaddr_dl *)s)
128 #define SRP(s) ((struct sockaddr_inarp *)s)
129 
130 /*
131  * ARP trailer negotiation.  Trailer protocol is not IP specific,
132  * but ARP request/response use IP addresses.
133  */
134 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
135 
136 /* timer values */
137 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
138 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
139 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
140 int	arpt_refresh = (5*60);	/* time left before refreshing */
141 #define	rt_expire rt_rmx.rmx_expire
142 #define	rt_pksent rt_rmx.rmx_pksent
143 
144 extern	struct domain arpdomain;
145 
146 static	void arprequest __P((struct ifnet *,
147 	    struct in_addr *, struct in_addr *, u_int8_t *));
148 static	void arptfree __P((struct llinfo_arp *));
149 static	void arptimer __P((void *));
150 static	struct llinfo_arp *arplookup __P((struct mbuf *, struct in_addr *,
151 					  int, int));
152 static	void in_arpinput __P((struct mbuf *));
153 
154 LIST_HEAD(, llinfo_arp) llinfo_arp;
155 struct	ifqueue arpintrq = {0, 0, 0, 50};
156 int	arp_inuse, arp_allocated, arp_intimer;
157 int	arp_maxtries = 5;
158 int	useloopback = 1;	/* use loopback interface for local traffic */
159 int	arpinit_done = 0;
160 
161 struct	arpstat arpstat;
162 struct	callout arptimer_ch;
163 
164 
165 /* revarp state */
166 static struct	in_addr myip, srv_ip;
167 static int	myip_initialized = 0;
168 static int	revarp_in_progress = 0;
169 static struct	ifnet *myip_ifp = NULL;
170 
171 #ifdef DDB
172 static void db_print_sa __P((const struct sockaddr *));
173 static void db_print_ifa __P((struct ifaddr *));
174 static void db_print_llinfo __P((caddr_t));
175 static int db_show_radix_node __P((struct radix_node *, void *));
176 #endif
177 
178 /*
179  * this should be elsewhere.
180  */
181 
182 static char *
183 lla_snprintf __P((u_int8_t *, int));
184 
185 static char *
186 lla_snprintf(adrp, len)
187 	u_int8_t *adrp;
188 	int len;
189 {
190 #define NUMBUFS 3
191 	static char buf[NUMBUFS][16*3];
192 	static int bnum = 0;
193 	static const char hexdigits[] = {
194 	    '0','1','2','3','4','5','6','7',
195 	    '8','9','a','b','c','d','e','f'
196 	};
197 
198 	int i;
199 	char *p;
200 
201 	p = buf[bnum];
202 
203 	*p++ = hexdigits[(*adrp)>>4];
204 	*p++ = hexdigits[(*adrp++)&0xf];
205 
206 	for (i=1; i<len && i<16; i++) {
207 		*p++ = ':';
208 		*p++ = hexdigits[(*adrp)>>4];
209 		*p++ = hexdigits[(*adrp++)&0xf];
210 	}
211 
212 	*p = 0;
213 	p = buf[bnum];
214 	bnum = (bnum + 1) % NUMBUFS;
215 	return p;
216 }
217 
218 const struct protosw arpsw[] = {
219 	{ 0, 0, 0, 0,
220 	  0, 0, 0, 0,
221 	  0,
222 	  0, 0, 0, arp_drain,
223 	}
224 };
225 
226 
227 struct domain arpdomain =
228 { 	PF_ARP,  "arp", 0, 0, 0,
229 	arpsw, &arpsw[sizeof(arpsw)/sizeof(arpsw[0])]
230 };
231 
232 /*
233  * ARP table locking.
234  *
235  * to prevent lossage vs. the arp_drain routine (which may be called at
236  * any time, including in a device driver context), we do two things:
237  *
238  * 1) manipulation of la->la_hold is done at splnet() (for all of
239  * about two instructions).
240  *
241  * 2) manipulation of the arp table's linked list is done under the
242  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
243  * while the arp table is locked, we punt and try again later.
244  */
245 
246 static int	arp_locked;
247 static __inline int arp_lock_try __P((int));
248 static __inline void arp_unlock __P((void));
249 
250 static __inline int
251 arp_lock_try(int recurse)
252 {
253 	int s;
254 
255 	/*
256 	 * Use splvm() -- we're blocking things that would cause
257 	 * mbuf allocation.
258 	 */
259 	s = splvm();
260 	if (!recurse && arp_locked) {
261 		splx(s);
262 		return (0);
263 	}
264 	arp_locked++;
265 	splx(s);
266 	return (1);
267 }
268 
269 static __inline void
270 arp_unlock()
271 {
272 	int s;
273 
274 	s = splvm();
275 	arp_locked--;
276 	splx(s);
277 }
278 
279 #ifdef DIAGNOSTIC
280 #define	ARP_LOCK(recurse)						\
281 do {									\
282 	if (arp_lock_try(recurse) == 0) {				\
283 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
284 		panic("arp_lock");					\
285 	}								\
286 } while (/*CONSTCOND*/ 0)
287 #define	ARP_LOCK_CHECK()						\
288 do {									\
289 	if (arp_locked == 0) {						\
290 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
291 		panic("arp lock check");				\
292 	}								\
293 } while (/*CONSTCOND*/ 0)
294 #else
295 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
296 #define	ARP_LOCK_CHECK()	/* nothing */
297 #endif
298 
299 #define	ARP_UNLOCK()		arp_unlock()
300 
301 /*
302  * ARP protocol drain routine.  Called when memory is in short supply.
303  * Called at splvm();
304  */
305 
306 void
307 arp_drain()
308 {
309 	struct llinfo_arp *la, *nla;
310 	int count = 0;
311 	struct mbuf *mold;
312 
313 	if (arp_lock_try(0) == 0) {
314 		printf("arp_drain: locked; punting\n");
315 		return;
316 	}
317 
318 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
319 		nla = LIST_NEXT(la, la_list);
320 
321 		mold = la->la_hold;
322 		la->la_hold = 0;
323 
324 		if (mold) {
325 			m_freem(mold);
326 			count++;
327 		}
328 	}
329 	ARP_UNLOCK();
330 	arpstat.as_dfrdropped += count;
331 }
332 
333 
334 /*
335  * Timeout routine.  Age arp_tab entries periodically.
336  */
337 /* ARGSUSED */
338 static void
339 arptimer(arg)
340 	void *arg;
341 {
342 	int s;
343 	struct llinfo_arp *la, *nla;
344 
345 	s = splsoftnet();
346 
347 	if (arp_lock_try(0) == 0) {
348 		/* get it later.. */
349 		splx(s);
350 		return;
351 	}
352 
353 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
354 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
355 		struct rtentry *rt = la->la_rt;
356 
357 		nla = LIST_NEXT(la, la_list);
358 		if (rt->rt_expire == 0)
359 			continue;
360 		if ((rt->rt_expire - time.tv_sec) < arpt_refresh &&
361 		    rt->rt_pksent > (time.tv_sec - arpt_keep)) {
362 			/*
363 			 * If the entry has been used during since last
364 			 * refresh, try to renew it before deleting.
365 			 */
366 			arprequest(rt->rt_ifp,
367 			    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
368 			    &SIN(rt_key(rt))->sin_addr,
369 			    LLADDR(rt->rt_ifp->if_sadl));
370 		} else if (rt->rt_expire <= time.tv_sec)
371 			arptfree(la); /* timer has expired; clear */
372 	}
373 
374 	ARP_UNLOCK();
375 
376 	splx(s);
377 }
378 
379 /*
380  * Parallel to llc_rtrequest.
381  */
382 void
383 arp_rtrequest(req, rt, info)
384 	int req;
385 	struct rtentry *rt;
386 	struct rt_addrinfo *info;
387 {
388 	struct sockaddr *gate = rt->rt_gateway;
389 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
390 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
391 	size_t allocsize;
392 	struct mbuf *mold;
393 	int s;
394 	struct in_ifaddr *ia;
395 	struct ifaddr *ifa;
396 
397 	if (!arpinit_done) {
398 		arpinit_done = 1;
399 		/*
400 		 * We generate expiration times from time.tv_sec
401 		 * so avoid accidently creating permanent routes.
402 		 */
403 		if (time.tv_sec == 0) {
404 			time.tv_sec++;
405 		}
406 		callout_init(&arptimer_ch);
407 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
408 	}
409 
410 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
411 		if (req != RTM_ADD)
412 			return;
413 
414 		/*
415 		 * linklayers with particular link MTU limitation.
416 		 */
417 		switch(rt->rt_ifp->if_type) {
418 #if NFDDI > 0
419 		case IFT_FDDI:
420 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
421 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
422 			break;
423 #endif
424 #if NARC > 0
425 		case IFT_ARCNET:
426 		    {
427 			int arcipifmtu;
428 
429 			if (rt->rt_ifp->if_flags & IFF_LINK0)
430 				arcipifmtu = arc_ipmtu;
431 			else
432 				arcipifmtu = ARCMTU;
433 			if (rt->rt_ifp->if_mtu > arcipifmtu)
434 				rt->rt_rmx.rmx_mtu = arcipifmtu;
435 			break;
436 		    }
437 #endif
438 		}
439 		return;
440 	}
441 
442 	ARP_LOCK(1);		/* we may already be locked here. */
443 
444 	switch (req) {
445 
446 	case RTM_ADD:
447 		/*
448 		 * XXX: If this is a manually added route to interface
449 		 * such as older version of routed or gated might provide,
450 		 * restore cloning bit.
451 		 */
452 		if ((rt->rt_flags & RTF_HOST) == 0 &&
453 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
454 			rt->rt_flags |= RTF_CLONING;
455 		if (rt->rt_flags & RTF_CLONING) {
456 			/*
457 			 * Case 1: This route should come from a route to iface.
458 			 */
459 			rt_setgate(rt, rt_key(rt),
460 					(struct sockaddr *)&null_sdl);
461 			gate = rt->rt_gateway;
462 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
463 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
464 			/*
465 			 * Give this route an expiration time, even though
466 			 * it's a "permanent" route, so that routes cloned
467 			 * from it do not need their expiration time set.
468 			 */
469 			rt->rt_expire = time.tv_sec;
470 			/*
471 			 * linklayers with particular link MTU limitation.
472 			 */
473 			switch (rt->rt_ifp->if_type) {
474 #if NFDDI > 0
475 			case IFT_FDDI:
476 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
477 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
478 				     (rt->rt_rmx.rmx_mtu == 0 &&
479 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
480 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
481 				break;
482 #endif
483 #if NARC > 0
484 			case IFT_ARCNET:
485 			    {
486 				int arcipifmtu;
487 				if (rt->rt_ifp->if_flags & IFF_LINK0)
488 					arcipifmtu = arc_ipmtu;
489 				else
490 					arcipifmtu = ARCMTU;
491 
492 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
493 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
494 				     (rt->rt_rmx.rmx_mtu == 0 &&
495 				      rt->rt_ifp->if_mtu > arcipifmtu)))
496 					rt->rt_rmx.rmx_mtu = arcipifmtu;
497 				break;
498 			    }
499 #endif
500 			}
501 			break;
502 		}
503 		/* Announce a new entry if requested. */
504 		if (rt->rt_flags & RTF_ANNOUNCE)
505 			arprequest(rt->rt_ifp,
506 			    &SIN(rt_key(rt))->sin_addr,
507 			    &SIN(rt_key(rt))->sin_addr,
508 			    (u_char *)LLADDR(SDL(gate)));
509 		/*FALLTHROUGH*/
510 	case RTM_RESOLVE:
511 		if (gate->sa_family != AF_LINK ||
512 		    gate->sa_len < sizeof(null_sdl)) {
513 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
514 			break;
515 		}
516 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
517 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
518 		if (la != 0)
519 			break; /* This happens on a route change */
520 		/*
521 		 * Case 2:  This route may come from cloning, or a manual route
522 		 * add with a LL address.
523 		 */
524 		switch (SDL(gate)->sdl_type) {
525 #if NTOKEN > 0
526 		case IFT_ISO88025:
527 			allocsize = sizeof(*la) + sizeof(struct token_rif);
528 			break;
529 #endif /* NTOKEN > 0 */
530 		default:
531 			allocsize = sizeof(*la);
532 		}
533 		R_Malloc(la, struct llinfo_arp *, allocsize);
534 		rt->rt_llinfo = (caddr_t)la;
535 		if (la == 0) {
536 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
537 			break;
538 		}
539 		arp_inuse++, arp_allocated++;
540 		Bzero(la, allocsize);
541 		la->la_rt = rt;
542 		rt->rt_flags |= RTF_LLINFO;
543 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
544 
545 		INADDR_TO_IA(SIN(rt_key(rt))->sin_addr, ia);
546 		while (ia && ia->ia_ifp != rt->rt_ifp)
547 			NEXT_IA_WITH_SAME_ADDR(ia);
548 		if (ia) {
549 			/*
550 			 * This test used to be
551 			 *	if (lo0ifp->if_flags & IFF_UP)
552 			 * It allowed local traffic to be forced through
553 			 * the hardware by configuring the loopback down.
554 			 * However, it causes problems during network
555 			 * configuration for boards that can't receive
556 			 * packets they send.  It is now necessary to clear
557 			 * "useloopback" and remove the route to force
558 			 * traffic out to the hardware.
559 			 *
560 			 * In 4.4BSD, the above "if" statement checked
561 			 * rt->rt_ifa against rt_key(rt).  It was changed
562 			 * to the current form so that we can provide a
563 			 * better support for multiple IPv4 addresses on a
564 			 * interface.
565 			 */
566 			rt->rt_expire = 0;
567 			Bcopy(LLADDR(rt->rt_ifp->if_sadl),
568 			    LLADDR(SDL(gate)),
569 			    SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
570 			if (useloopback)
571 				rt->rt_ifp = lo0ifp;
572 			/*
573 			 * make sure to set rt->rt_ifa to the interface
574 			 * address we are using, otherwise we will have trouble
575 			 * with source address selection.
576 			 */
577 			ifa = &ia->ia_ifa;
578 			if (ifa != rt->rt_ifa) {
579 				IFAFREE(rt->rt_ifa);
580 				IFAREF(ifa);
581 				rt->rt_ifa = ifa;
582 			}
583 		}
584 		break;
585 
586 	case RTM_DELETE:
587 		if (la == 0)
588 			break;
589 		arp_inuse--;
590 		LIST_REMOVE(la, la_list);
591 		rt->rt_llinfo = 0;
592 		rt->rt_flags &= ~RTF_LLINFO;
593 
594 		s = splnet();
595 		mold = la->la_hold;
596 		la->la_hold = 0;
597 		splx(s);
598 
599 		if (mold)
600 			m_freem(mold);
601 
602 		Free((caddr_t)la);
603 	}
604 	ARP_UNLOCK();
605 }
606 
607 /*
608  * Broadcast an ARP request. Caller specifies:
609  *	- arp header source ip address
610  *	- arp header target ip address
611  *	- arp header source ethernet address
612  */
613 static void
614 arprequest(ifp, sip, tip, enaddr)
615 	struct ifnet *ifp;
616 	struct in_addr *sip, *tip;
617 	u_int8_t *enaddr;
618 {
619 	struct mbuf *m;
620 	struct arphdr *ah;
621 	struct sockaddr sa;
622 
623 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
624 		return;
625 	MCLAIM(m, &arpdomain.dom_mowner);
626 	switch (ifp->if_type) {
627 	case IFT_IEEE1394:
628 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
629 		    ifp->if_addrlen;
630 		break;
631 	default:
632 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
633 		    2 * ifp->if_addrlen;
634 		break;
635 	}
636 	m->m_pkthdr.len = m->m_len;
637 	MH_ALIGN(m, m->m_len);
638 	ah = mtod(m, struct arphdr *);
639 	bzero((caddr_t)ah, m->m_len);
640 	switch (ifp->if_type) {
641 	case IFT_IEEE1394:	/* RFC2734 */
642 		/* fill it now for ar_tpa computation */
643 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
644 		break;
645 	default:
646 		/* ifp->if_output will fill ar_hrd */
647 		break;
648 	}
649 	ah->ar_pro = htons(ETHERTYPE_IP);
650 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
651 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
652 	ah->ar_op = htons(ARPOP_REQUEST);
653 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
654 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
655 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
656 	sa.sa_family = AF_ARP;
657 	sa.sa_len = 2;
658 	m->m_flags |= M_BCAST;
659 	arpstat.as_sndtotal++;
660 	arpstat.as_sndrequest++;
661 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
662 }
663 
664 /*
665  * Resolve an IP address into an ethernet address.  If success,
666  * desten is filled in.  If there is no entry in arptab,
667  * set one up and broadcast a request for the IP address.
668  * Hold onto this mbuf and resend it once the address
669  * is finally resolved.  A return value of 1 indicates
670  * that desten has been filled in and the packet should be sent
671  * normally; a 0 return indicates that the packet has been
672  * taken over here, either now or for later transmission.
673  */
674 int
675 arpresolve(ifp, rt, m, dst, desten)
676 	struct ifnet *ifp;
677 	struct rtentry *rt;
678 	struct mbuf *m;
679 	struct sockaddr *dst;
680 	u_char *desten;
681 {
682 	struct llinfo_arp *la;
683 	struct sockaddr_dl *sdl;
684 	struct mbuf *mold;
685 	int s;
686 
687 	if (rt)
688 		la = (struct llinfo_arp *)rt->rt_llinfo;
689 	else {
690 		if ((la = arplookup(m, &SIN(dst)->sin_addr, 1, 0)) != NULL)
691 			rt = la->la_rt;
692 	}
693 	if (la == 0 || rt == 0) {
694 		arpstat.as_allocfail++;
695 		log(LOG_DEBUG,
696 		    "arpresolve: can't allocate llinfo on %s for %s\n",
697 		    ifp->if_xname, in_fmtaddr(SIN(dst)->sin_addr));
698 		m_freem(m);
699 		return (0);
700 	}
701 	sdl = SDL(rt->rt_gateway);
702 	/*
703 	 * Check the address family and length is valid, the address
704 	 * is resolved; otherwise, try to resolve.
705 	 */
706 	if ((rt->rt_expire == 0 || rt->rt_expire > time.tv_sec) &&
707 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
708 		bcopy(LLADDR(sdl), desten,
709 		    min(sdl->sdl_alen, ifp->if_addrlen));
710 		rt->rt_pksent = time.tv_sec; /* Time for last pkt sent */
711 		return 1;
712 	}
713 	/*
714 	 * There is an arptab entry, but no ethernet address
715 	 * response yet.  Replace the held mbuf with this
716 	 * latest one.
717 	 */
718 
719 	arpstat.as_dfrtotal++;
720 	s = splnet();
721 	mold = la->la_hold;
722 	la->la_hold = m;
723 	splx(s);
724 
725 	if (mold) {
726 		arpstat.as_dfrdropped++;
727 		m_freem(mold);
728 	}
729 
730 	/*
731 	 * Re-send the ARP request when appropriate.
732 	 */
733 #ifdef	DIAGNOSTIC
734 	if (rt->rt_expire == 0) {
735 		/* This should never happen. (Should it? -gwr) */
736 		printf("arpresolve: unresolved and rt_expire == 0\n");
737 		/* Set expiration time to now (expired). */
738 		rt->rt_expire = time.tv_sec;
739 	}
740 #endif
741 	if (rt->rt_expire) {
742 		rt->rt_flags &= ~RTF_REJECT;
743 		if (la->la_asked == 0 || rt->rt_expire != time.tv_sec) {
744 			rt->rt_expire = time.tv_sec;
745 			if (la->la_asked++ < arp_maxtries)
746 				arprequest(ifp,
747 				    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
748 				    &SIN(dst)->sin_addr,
749 				    LLADDR(ifp->if_sadl));
750 			else {
751 				rt->rt_flags |= RTF_REJECT;
752 				rt->rt_expire += arpt_down;
753 				la->la_asked = 0;
754 			}
755 		}
756 	}
757 	return (0);
758 }
759 
760 /*
761  * Common length and type checks are done here,
762  * then the protocol-specific routine is called.
763  */
764 void
765 arpintr()
766 {
767 	struct mbuf *m;
768 	struct arphdr *ar;
769 	int s;
770 	int arplen;
771 
772 	while (arpintrq.ifq_head) {
773 		s = splnet();
774 		IF_DEQUEUE(&arpintrq, m);
775 		splx(s);
776 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
777 			panic("arpintr");
778 
779 		MCLAIM(m, &arpdomain.dom_mowner);
780 		arpstat.as_rcvtotal++;
781 
782 		/*
783 		 * First, make sure we have at least struct arphdr.
784 		 */
785 		if (m->m_len < sizeof(struct arphdr) ||
786 		    (ar = mtod(m, struct arphdr *)) == NULL)
787 			goto badlen;
788 
789 		switch (m->m_pkthdr.rcvif->if_type) {
790 		case IFT_IEEE1394:
791 			arplen = sizeof(struct arphdr) +
792 			    ar->ar_hln + 2 * ar->ar_pln;
793 			break;
794 		default:
795 			arplen = sizeof(struct arphdr) +
796 			    2 * ar->ar_hln + 2 * ar->ar_pln;
797 			break;
798 		}
799 
800 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
801 		    m->m_len >= arplen)
802 			switch (ntohs(ar->ar_pro)) {
803 			case ETHERTYPE_IP:
804 			case ETHERTYPE_IPTRAILERS:
805 				in_arpinput(m);
806 				continue;
807 			default:
808 				arpstat.as_rcvbadproto++;
809 			}
810 		else {
811 badlen:
812 			arpstat.as_rcvbadlen++;
813 		}
814 		m_freem(m);
815 	}
816 }
817 
818 /*
819  * ARP for Internet protocols on 10 Mb/s Ethernet.
820  * Algorithm is that given in RFC 826.
821  * In addition, a sanity check is performed on the sender
822  * protocol address, to catch impersonators.
823  * We no longer handle negotiations for use of trailer protocol:
824  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
825  * along with IP replies if we wanted trailers sent to us,
826  * and also sent them in response to IP replies.
827  * This allowed either end to announce the desire to receive
828  * trailer packets.
829  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
830  * but formerly didn't normally send requests.
831  */
832 static void
833 in_arpinput(m)
834 	struct mbuf *m;
835 {
836 	struct arphdr *ah;
837 	struct ifnet *ifp = m->m_pkthdr.rcvif;
838 	struct llinfo_arp *la = 0;
839 	struct rtentry  *rt;
840 	struct in_ifaddr *ia;
841 #if NBRIDGE > 0
842 	struct in_ifaddr *bridge_ia = NULL;
843 #endif
844 	struct sockaddr_dl *sdl;
845 	struct sockaddr sa;
846 	struct in_addr isaddr, itaddr, myaddr;
847 	int op;
848 	struct mbuf *mold;
849 	int s;
850 
851 	ah = mtod(m, struct arphdr *);
852 	op = ntohs(ah->ar_op);
853 
854 	/*
855 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
856 	 * ar_tpa().
857 	 */
858 	switch (ifp->if_type) {
859 	case IFT_IEEE1394:
860 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
861 			;
862 		else {
863 			/* XXX this is to make sure we compute ar_tha right */
864 			/* XXX check ar_hrd more strictly? */
865 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
866 		}
867 		break;
868 	default:
869 		/* XXX check ar_hrd? */
870 		break;
871 	}
872 
873 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
874 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
875 
876 	if (m->m_flags & (M_BCAST|M_MCAST))
877 		arpstat.as_rcvmcast++;
878 
879 	/*
880 	 * If the target IP address is zero, ignore the packet.
881 	 * This prevents the code below from tring to answer
882 	 * when we are using IP address zero (booting).
883 	 */
884 	if (in_nullhost(itaddr)) {
885 		arpstat.as_rcvzerotpa++;
886 		goto out;
887 	}
888 
889 	/*
890 	 * If the source IP address is zero, this is most likely a
891 	 * confused host trying to use IP address zero. (Windoze?)
892 	 * XXX: Should we bother trying to reply to these?
893 	 */
894 	if (in_nullhost(isaddr)) {
895 		arpstat.as_rcvzerospa++;
896 		goto out;
897 	}
898 
899 	/*
900 	 * Search for a matching interface address
901 	 * or any address on the interface to use
902 	 * as a dummy address in the rest of this function
903 	 */
904 	INADDR_TO_IA(itaddr, ia);
905 	while (ia != NULL) {
906 		if (ia->ia_ifp == m->m_pkthdr.rcvif)
907 			break;
908 
909 #if NBRIDGE > 0
910 		/*
911 		 * If the interface we received the packet on
912 		 * is part of a bridge, check to see if we need
913 		 * to "bridge" the packet to ourselves at this
914 		 * layer.  Note we still prefer a perfect match,
915 		 * but allow this weaker match if necessary.
916 		 */
917 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
918 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
919 			bridge_ia = ia;
920 #endif /* NBRIDGE > 0 */
921 
922 		NEXT_IA_WITH_SAME_ADDR(ia);
923 	}
924 
925 #if NBRIDGE > 0
926 	if (ia == NULL && bridge_ia != NULL) {
927 		ia = bridge_ia;
928 		ifp = bridge_ia->ia_ifp;
929 	}
930 #endif
931 
932 	if (ia == NULL) {
933 		INADDR_TO_IA(isaddr, ia);
934 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
935 			NEXT_IA_WITH_SAME_ADDR(ia);
936 
937 		if (ia == NULL) {
938 			IFP_TO_IA(ifp, ia);
939 			if (ia == NULL) {
940 				arpstat.as_rcvnoint++;
941 				goto out;
942 			}
943 		}
944 	}
945 
946 	myaddr = ia->ia_addr.sin_addr;
947 
948 	/* XXX checks for bridge case? */
949 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
950 	    ifp->if_addrlen)) {
951 		arpstat.as_rcvlocalsha++;
952 		goto out;	/* it's from me, ignore it. */
953 	}
954 
955 	/* XXX checks for bridge case? */
956 	if (!bcmp((caddr_t)ar_sha(ah), (caddr_t)ifp->if_broadcastaddr,
957 	    ifp->if_addrlen)) {
958 		arpstat.as_rcvbcastsha++;
959 		log(LOG_ERR,
960 		    "%s: arp: link address is broadcast for IP address %s!\n",
961 		    ifp->if_xname, in_fmtaddr(isaddr));
962 		goto out;
963 	}
964 
965 	if (in_hosteq(isaddr, myaddr)) {
966 		arpstat.as_rcvlocalspa++;
967 		log(LOG_ERR,
968 		   "duplicate IP address %s sent from link address %s\n",
969 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
970 		itaddr = myaddr;
971 		goto reply;
972 	}
973 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
974 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
975 		if (sdl->sdl_alen &&
976 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
977 			if (rt->rt_flags & RTF_STATIC) {
978 				arpstat.as_rcvoverperm++;
979 				log(LOG_INFO,
980 				    "%s tried to overwrite permanent arp info"
981 				    " for %s\n",
982 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
983 				    in_fmtaddr(isaddr));
984 				goto out;
985 			} else if (rt->rt_ifp != ifp) {
986 				arpstat.as_rcvoverint++;
987 				log(LOG_INFO,
988 				    "%s on %s tried to overwrite "
989 				    "arp info for %s on %s\n",
990 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
991 				    ifp->if_xname, in_fmtaddr(isaddr),
992 				    rt->rt_ifp->if_xname);
993 				    goto out;
994 			} else {
995 				arpstat.as_rcvover++;
996 				log(LOG_INFO,
997 				    "arp info overwritten for %s by %s\n",
998 				    in_fmtaddr(isaddr),
999 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1000 			}
1001 		}
1002 		/*
1003 		 * sanity check for the address length.
1004 		 * XXX this does not work for protocols with variable address
1005 		 * length. -is
1006 		 */
1007 		if (sdl->sdl_alen &&
1008 		    sdl->sdl_alen != ah->ar_hln) {
1009 			arpstat.as_rcvlenchg++;
1010 			log(LOG_WARNING,
1011 			    "arp from %s: new addr len %d, was %d",
1012 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1013 		}
1014 		if (ifp->if_addrlen != ah->ar_hln) {
1015 			arpstat.as_rcvbadlen++;
1016 			log(LOG_WARNING,
1017 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1018 			    in_fmtaddr(isaddr), ah->ar_hln,
1019 			    ifp->if_addrlen);
1020 			goto reply;
1021 		}
1022 #if NTOKEN > 0
1023 		/*
1024 		 * XXX uses m_data and assumes the complete answer including
1025 		 * XXX token-ring headers is in the same buf
1026 		 */
1027 		if (ifp->if_type == IFT_ISO88025) {
1028 			struct token_header *trh;
1029 
1030 			trh = (struct token_header *)M_TRHSTART(m);
1031 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1032 				struct token_rif	*rif;
1033 				size_t	riflen;
1034 
1035 				rif = TOKEN_RIF(trh);
1036 				riflen = (ntohs(rif->tr_rcf) &
1037 				    TOKEN_RCF_LEN_MASK) >> 8;
1038 
1039 				if (riflen > 2 &&
1040 				    riflen < sizeof(struct token_rif) &&
1041 				    (riflen & 1) == 0) {
1042 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1043 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1044 					bcopy(rif, TOKEN_RIF(la), riflen);
1045 				}
1046 			}
1047 		}
1048 #endif /* NTOKEN > 0 */
1049 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
1050 		    sdl->sdl_alen = ah->ar_hln);
1051 		if (rt->rt_expire)
1052 			rt->rt_expire = time.tv_sec + arpt_keep;
1053 		rt->rt_flags &= ~RTF_REJECT;
1054 		la->la_asked = 0;
1055 
1056 		s = splnet();
1057 		mold = la->la_hold;
1058 		la->la_hold = 0;
1059 		splx(s);
1060 
1061 		if (mold) {
1062 			arpstat.as_dfrsent++;
1063 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
1064 		}
1065 	}
1066 reply:
1067 	if (op != ARPOP_REQUEST) {
1068 		if (op == ARPOP_REPLY)
1069 			arpstat.as_rcvreply++;
1070 	out:
1071 		m_freem(m);
1072 		return;
1073 	}
1074 	arpstat.as_rcvrequest++;
1075 	if (in_hosteq(itaddr, myaddr)) {
1076 		/* I am the target */
1077 		if (ar_tha(ah))
1078 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1079 			    ah->ar_hln);
1080 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1081 	} else {
1082 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1083 		if (la == 0)
1084 			goto out;
1085 		rt = la->la_rt;
1086 		if (ar_tha(ah))
1087 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1088 			    ah->ar_hln);
1089 		sdl = SDL(rt->rt_gateway);
1090 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
1091 	}
1092 
1093 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
1094 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
1095 	ah->ar_op = htons(ARPOP_REPLY);
1096 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1097 	switch (ifp->if_type) {
1098 	case IFT_IEEE1394:
1099 		/*
1100 		 * ieee1394 arp reply is broadcast
1101 		 */
1102 		m->m_flags &= ~M_MCAST;
1103 		m->m_flags |= M_BCAST;
1104 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1105 		break;
1106 
1107 	default:
1108 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1109 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1110 		break;
1111 	}
1112 	m->m_pkthdr.len = m->m_len;
1113 	sa.sa_family = AF_ARP;
1114 	sa.sa_len = 2;
1115 	arpstat.as_sndtotal++;
1116 	arpstat.as_sndreply++;
1117 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1118 	return;
1119 }
1120 
1121 /*
1122  * Free an arp entry.
1123  */
1124 static void
1125 arptfree(la)
1126 	struct llinfo_arp *la;
1127 {
1128 	struct rtentry *rt = la->la_rt;
1129 	struct sockaddr_dl *sdl;
1130 
1131 	ARP_LOCK_CHECK();
1132 
1133 	if (rt == 0)
1134 		panic("arptfree");
1135 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
1136 	    sdl->sdl_family == AF_LINK) {
1137 		sdl->sdl_alen = 0;
1138 		la->la_asked = 0;
1139 		rt->rt_flags &= ~RTF_REJECT;
1140 		return;
1141 	}
1142 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
1143 	    0, (struct rtentry **)0);
1144 }
1145 
1146 /*
1147  * Lookup or enter a new address in arptab.
1148  */
1149 static struct llinfo_arp *
1150 arplookup(m, addr, create, proxy)
1151 	struct mbuf *m;
1152 	struct in_addr *addr;
1153 	int create, proxy;
1154 {
1155 	struct arphdr *ah;
1156 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1157 	struct rtentry *rt;
1158 	static struct sockaddr_inarp sin;
1159 	const char *why = 0;
1160 
1161 	ah = mtod(m, struct arphdr *);
1162 	sin.sin_len = sizeof(sin);
1163 	sin.sin_family = AF_INET;
1164 	sin.sin_addr = *addr;
1165 	sin.sin_other = proxy ? SIN_PROXY : 0;
1166 	rt = rtalloc1(sintosa(&sin), create);
1167 	if (rt == 0)
1168 		return (0);
1169 	rt->rt_refcnt--;
1170 
1171 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO &&
1172 	    rt->rt_gateway->sa_family == AF_LINK)
1173 		return ((struct llinfo_arp *)rt->rt_llinfo);
1174 
1175 
1176 
1177 	if (create) {
1178 		if (rt->rt_flags & RTF_GATEWAY)
1179 			why = "host is not on local network";
1180 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1181 			arpstat.as_allocfail++;
1182 			why = "could not allocate llinfo";
1183 		} else
1184 			why = "gateway route is not ours";
1185 		log(LOG_DEBUG, "arplookup: unable to enter address"
1186 		    " for %s@%s on %s (%s)\n",
1187 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1188 		    (ifp) ? ifp->if_xname : 0, why);
1189 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1190 			rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
1191 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1192 		}
1193 	}
1194 	return (0);
1195 }
1196 
1197 int
1198 arpioctl(cmd, data)
1199 	u_long cmd;
1200 	caddr_t data;
1201 {
1202 
1203 	return (EOPNOTSUPP);
1204 }
1205 
1206 void
1207 arp_ifinit(ifp, ifa)
1208 	struct ifnet *ifp;
1209 	struct ifaddr *ifa;
1210 {
1211 	struct in_addr *ip;
1212 
1213 	/*
1214 	 * Warn the user if another station has this IP address,
1215 	 * but only if the interface IP address is not zero.
1216 	 */
1217 	ip = &IA_SIN(ifa)->sin_addr;
1218 	if (!in_nullhost(*ip))
1219 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
1220 
1221 	ifa->ifa_rtrequest = arp_rtrequest;
1222 	ifa->ifa_flags |= RTF_CLONING;
1223 }
1224 
1225 /*
1226  * Called from 10 Mb/s Ethernet interrupt handlers
1227  * when ether packet type ETHERTYPE_REVARP
1228  * is received.  Common length and type checks are done here,
1229  * then the protocol-specific routine is called.
1230  */
1231 void
1232 revarpinput(m)
1233 	struct mbuf *m;
1234 {
1235 	struct arphdr *ar;
1236 
1237 	if (m->m_len < sizeof(struct arphdr))
1238 		goto out;
1239 	ar = mtod(m, struct arphdr *);
1240 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1241 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1242 		goto out;
1243 #endif
1244 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1245 		goto out;
1246 	switch (ntohs(ar->ar_pro)) {
1247 	case ETHERTYPE_IP:
1248 	case ETHERTYPE_IPTRAILERS:
1249 		in_revarpinput(m);
1250 		return;
1251 
1252 	default:
1253 		break;
1254 	}
1255 out:
1256 	m_freem(m);
1257 }
1258 
1259 /*
1260  * RARP for Internet protocols on 10 Mb/s Ethernet.
1261  * Algorithm is that given in RFC 903.
1262  * We are only using for bootstrap purposes to get an ip address for one of
1263  * our interfaces.  Thus we support no user-interface.
1264  *
1265  * Since the contents of the RARP reply are specific to the interface that
1266  * sent the request, this code must ensure that they are properly associated.
1267  *
1268  * Note: also supports ARP via RARP packets, per the RFC.
1269  */
1270 void
1271 in_revarpinput(m)
1272 	struct mbuf *m;
1273 {
1274 	struct ifnet *ifp;
1275 	struct arphdr *ah;
1276 	int op;
1277 
1278 	ah = mtod(m, struct arphdr *);
1279 	op = ntohs(ah->ar_op);
1280 
1281 	switch (m->m_pkthdr.rcvif->if_type) {
1282 	case IFT_IEEE1394:
1283 		/* ARP without target hardware address is not supported */
1284 		goto out;
1285 	default:
1286 		break;
1287 	}
1288 
1289 	switch (op) {
1290 	case ARPOP_REQUEST:
1291 	case ARPOP_REPLY:	/* per RFC */
1292 		in_arpinput(m);
1293 		return;
1294 	case ARPOP_REVREPLY:
1295 		break;
1296 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1297 	default:
1298 		goto out;
1299 	}
1300 	if (!revarp_in_progress)
1301 		goto out;
1302 	ifp = m->m_pkthdr.rcvif;
1303 	if (ifp != myip_ifp) /* !same interface */
1304 		goto out;
1305 	if (myip_initialized)
1306 		goto wake;
1307 	if (bcmp(ar_tha(ah), LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1308 		goto out;
1309 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1310 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1311 	myip_initialized = 1;
1312 wake:	/* Do wakeup every time in case it was missed. */
1313 	wakeup((caddr_t)&myip);
1314 
1315 out:
1316 	m_freem(m);
1317 }
1318 
1319 /*
1320  * Send a RARP request for the ip address of the specified interface.
1321  * The request should be RFC 903-compliant.
1322  */
1323 void
1324 revarprequest(ifp)
1325 	struct ifnet *ifp;
1326 {
1327 	struct sockaddr sa;
1328 	struct mbuf *m;
1329 	struct arphdr *ah;
1330 
1331 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1332 		return;
1333 	MCLAIM(m, &arpdomain.dom_mowner);
1334 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1335 	    2*ifp->if_addrlen;
1336 	m->m_pkthdr.len = m->m_len;
1337 	MH_ALIGN(m, m->m_len);
1338 	ah = mtod(m, struct arphdr *);
1339 	bzero((caddr_t)ah, m->m_len);
1340 	ah->ar_pro = htons(ETHERTYPE_IP);
1341 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1342 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1343 	ah->ar_op = htons(ARPOP_REVREQUEST);
1344 
1345 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1346 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_tha(ah), ah->ar_hln);
1347 
1348 	sa.sa_family = AF_ARP;
1349 	sa.sa_len = 2;
1350 	m->m_flags |= M_BCAST;
1351 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1352 
1353 }
1354 
1355 /*
1356  * RARP for the ip address of the specified interface, but also
1357  * save the ip address of the server that sent the answer.
1358  * Timeout if no response is received.
1359  */
1360 int
1361 revarpwhoarewe(ifp, serv_in, clnt_in)
1362 	struct ifnet *ifp;
1363 	struct in_addr *serv_in;
1364 	struct in_addr *clnt_in;
1365 {
1366 	int result, count = 20;
1367 
1368 	myip_initialized = 0;
1369 	myip_ifp = ifp;
1370 
1371 	revarp_in_progress = 1;
1372 	while (count--) {
1373 		revarprequest(ifp);
1374 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1375 		if (result != EWOULDBLOCK)
1376 			break;
1377 	}
1378 	revarp_in_progress = 0;
1379 
1380 	if (!myip_initialized)
1381 		return ENETUNREACH;
1382 
1383 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1384 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1385 	return 0;
1386 }
1387 
1388 
1389 
1390 #ifdef DDB
1391 
1392 #include <machine/db_machdep.h>
1393 #include <ddb/db_interface.h>
1394 #include <ddb/db_output.h>
1395 static void
1396 db_print_sa(sa)
1397 	const struct sockaddr *sa;
1398 {
1399 	int len;
1400 	u_char *p;
1401 
1402 	if (sa == 0) {
1403 		db_printf("[NULL]");
1404 		return;
1405 	}
1406 
1407 	p = (u_char*)sa;
1408 	len = sa->sa_len;
1409 	db_printf("[");
1410 	while (len > 0) {
1411 		db_printf("%d", *p);
1412 		p++; len--;
1413 		if (len) db_printf(",");
1414 	}
1415 	db_printf("]\n");
1416 }
1417 static void
1418 db_print_ifa(ifa)
1419 	struct ifaddr *ifa;
1420 {
1421 	if (ifa == 0)
1422 		return;
1423 	db_printf("  ifa_addr=");
1424 	db_print_sa(ifa->ifa_addr);
1425 	db_printf("  ifa_dsta=");
1426 	db_print_sa(ifa->ifa_dstaddr);
1427 	db_printf("  ifa_mask=");
1428 	db_print_sa(ifa->ifa_netmask);
1429 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1430 			  ifa->ifa_flags,
1431 			  ifa->ifa_refcnt,
1432 			  ifa->ifa_metric);
1433 }
1434 static void
1435 db_print_llinfo(li)
1436 	caddr_t li;
1437 {
1438 	struct llinfo_arp *la;
1439 
1440 	if (li == 0)
1441 		return;
1442 	la = (struct llinfo_arp *)li;
1443 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1444 			  la->la_rt, la->la_hold, la->la_asked);
1445 }
1446 /*
1447  * Function to pass to rn_walktree().
1448  * Return non-zero error to abort walk.
1449  */
1450 static int
1451 db_show_radix_node(rn, w)
1452 	struct radix_node *rn;
1453 	void *w;
1454 {
1455 	struct rtentry *rt = (struct rtentry *)rn;
1456 
1457 	db_printf("rtentry=%p", rt);
1458 
1459 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1460 			  rt->rt_flags, rt->rt_refcnt,
1461 			  rt->rt_use, rt->rt_expire);
1462 
1463 	db_printf(" key="); db_print_sa(rt_key(rt));
1464 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1465 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1466 
1467 	db_printf(" ifp=%p ", rt->rt_ifp);
1468 	if (rt->rt_ifp)
1469 		db_printf("(%s)", rt->rt_ifp->if_xname);
1470 	else
1471 		db_printf("(NULL)");
1472 
1473 	db_printf(" ifa=%p\n", rt->rt_ifa);
1474 	db_print_ifa(rt->rt_ifa);
1475 
1476 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1477 
1478 	db_printf(" gwroute=%p llinfo=%p\n",
1479 			  rt->rt_gwroute, rt->rt_llinfo);
1480 	db_print_llinfo(rt->rt_llinfo);
1481 
1482 	return (0);
1483 }
1484 /*
1485  * Function to print all the route trees.
1486  * Use this from ddb:  "show arptab"
1487  */
1488 void
1489 db_show_arptab(addr, have_addr, count, modif)
1490 	db_expr_t	addr;
1491 	int		have_addr;
1492 	db_expr_t	count;
1493 	char *		modif;
1494 {
1495 	struct radix_node_head *rnh;
1496 	rnh = rt_tables[AF_INET];
1497 	db_printf("Route tree for AF_INET\n");
1498 	if (rnh == NULL) {
1499 		db_printf(" (not initialized)\n");
1500 		return;
1501 	}
1502 	rn_walktree(rnh, db_show_radix_node, NULL);
1503 	return;
1504 }
1505 #endif
1506 
1507 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1508 {
1509 	struct sysctlnode *node;
1510 
1511 	sysctl_createv(clog, 0, NULL, NULL,
1512 			CTLFLAG_PERMANENT,
1513 			CTLTYPE_NODE, "net", NULL,
1514 			NULL, 0, NULL, 0,
1515 			CTL_NET, CTL_EOL);
1516 	sysctl_createv(clog, 0, NULL, NULL,
1517 			CTLFLAG_PERMANENT,
1518 			CTLTYPE_NODE, "inet", NULL,
1519 			NULL, 0, NULL, 0,
1520 			CTL_NET, PF_INET, CTL_EOL);
1521 	sysctl_createv(clog, 0, NULL, &node,
1522 			CTLFLAG_PERMANENT,
1523 			CTLTYPE_NODE, "arp",
1524 			SYSCTL_DESCR("Address Resolution Protocol"),
1525 			NULL, 0, NULL, 0,
1526 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1527 
1528 	sysctl_createv(clog, 0, NULL, NULL,
1529 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1530 			CTLTYPE_INT, "prune",
1531 			SYSCTL_DESCR("ARP cache pruning interval"),
1532 			NULL, 0, &arpt_prune, 0,
1533 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1534 
1535 	sysctl_createv(clog, 0, NULL, NULL,
1536 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1537 			CTLTYPE_INT, "keep",
1538 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1539 			NULL, 0, &arpt_keep, 0,
1540 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1541 
1542 	sysctl_createv(clog, 0, NULL, NULL,
1543 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1544 			CTLTYPE_INT, "down",
1545 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1546 			NULL, 0, &arpt_down, 0,
1547 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1548 
1549 	sysctl_createv(clog, 0, NULL, NULL,
1550 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1551 			CTLTYPE_INT, "refresh",
1552 			SYSCTL_DESCR("ARP entry refresh interval"),
1553 			NULL, 0, &arpt_refresh, 0,
1554 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1555 }
1556 
1557 #endif /* INET */
1558