xref: /netbsd-src/sys/netinet/if_arp.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: if_arp.c,v 1.114 2006/10/12 01:32:37 christos 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 1.114 2006/10/12 01:32:37 christos 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/timetc.h>
95 #include <sys/kernel.h>
96 #include <sys/errno.h>
97 #include <sys/ioctl.h>
98 #include <sys/syslog.h>
99 #include <sys/proc.h>
100 #include <sys/protosw.h>
101 #include <sys/domain.h>
102 #include <sys/sysctl.h>
103 
104 #include <net/ethertypes.h>
105 #include <net/if.h>
106 #include <net/if_dl.h>
107 #include <net/if_token.h>
108 #include <net/if_types.h>
109 #include <net/if_ether.h>
110 #include <net/route.h>
111 
112 #include <netinet/in.h>
113 #include <netinet/in_systm.h>
114 #include <netinet/in_var.h>
115 #include <netinet/ip.h>
116 #include <netinet/if_inarp.h>
117 
118 #include "arc.h"
119 #if NARC > 0
120 #include <net/if_arc.h>
121 #endif
122 #include "fddi.h"
123 #if NFDDI > 0
124 #include <net/if_fddi.h>
125 #endif
126 #include "token.h"
127 #include "carp.h"
128 #if NCARP > 0
129 #include <netinet/ip_carp.h>
130 #endif
131 
132 #define SIN(s) ((struct sockaddr_in *)s)
133 #define SDL(s) ((struct sockaddr_dl *)s)
134 #define SRP(s) ((struct sockaddr_inarp *)s)
135 
136 /*
137  * ARP trailer negotiation.  Trailer protocol is not IP specific,
138  * but ARP request/response use IP addresses.
139  */
140 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
141 
142 /* timer values */
143 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
144 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
145 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
146 int	arpt_refresh = (5*60);	/* time left before refreshing */
147 #define	rt_expire rt_rmx.rmx_expire
148 #define	rt_pksent rt_rmx.rmx_pksent
149 
150 static	void arptfree(struct llinfo_arp *);
151 static	void arptimer(void *);
152 static	struct llinfo_arp *arplookup(struct mbuf *, struct in_addr *,
153 					  int, int);
154 static	void in_arpinput(struct mbuf *);
155 
156 LIST_HEAD(, llinfo_arp) llinfo_arp;
157 struct	ifqueue arpintrq = {
158 	.ifq_head = NULL,
159 	.ifq_tail = NULL,
160 	.ifq_len = 0,
161 	.ifq_maxlen = 50,
162 	.ifq_drops = 0,
163 };
164 int	arp_inuse, arp_allocated, arp_intimer;
165 int	arp_maxtries = 5;
166 int	useloopback = 1;	/* use loopback interface for local traffic */
167 int	arpinit_done = 0;
168 
169 struct	arpstat arpstat;
170 struct	callout arptimer_ch;
171 
172 
173 /* revarp state */
174 struct	in_addr myip, srv_ip;
175 int	myip_initialized = 0;
176 int	revarp_in_progress = 0;
177 struct	ifnet *myip_ifp = NULL;
178 
179 #ifdef DDB
180 static void db_print_sa(const struct sockaddr *);
181 static void db_print_ifa(struct ifaddr *);
182 static void db_print_llinfo(caddr_t);
183 static int db_show_radix_node(struct radix_node *, void *);
184 #endif
185 
186 /*
187  * this should be elsewhere.
188  */
189 
190 static char *
191 lla_snprintf(u_int8_t *, int);
192 
193 static char *
194 lla_snprintf(u_int8_t *adrp, int len)
195 {
196 #define NUMBUFS 3
197 	static char buf[NUMBUFS][16*3];
198 	static int bnum = 0;
199 
200 	int i;
201 	char *p;
202 
203 	p = buf[bnum];
204 
205 	*p++ = hexdigits[(*adrp)>>4];
206 	*p++ = hexdigits[(*adrp++)&0xf];
207 
208 	for (i=1; i<len && i<16; i++) {
209 		*p++ = ':';
210 		*p++ = hexdigits[(*adrp)>>4];
211 		*p++ = hexdigits[(*adrp++)&0xf];
212 	}
213 
214 	*p = 0;
215 	p = buf[bnum];
216 	bnum = (bnum + 1) % NUMBUFS;
217 	return p;
218 }
219 
220 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
221 
222 const struct protosw arpsw[] = {
223 	{ 0, &arpdomain, 0, 0,
224 	  0, 0, 0, 0,
225 	  0,
226 	  0, 0, 0, arp_drain,
227 	}
228 };
229 
230 
231 struct domain arpdomain = {
232 	.dom_family = PF_ARP,
233 	.dom_name = "arp",
234 	.dom_protosw = arpsw,
235 	.dom_protoswNPROTOSW = &arpsw[sizeof(arpsw)/sizeof(arpsw[0])],
236 };
237 
238 /*
239  * ARP table locking.
240  *
241  * to prevent lossage vs. the arp_drain routine (which may be called at
242  * any time, including in a device driver context), we do two things:
243  *
244  * 1) manipulation of la->la_hold is done at splnet() (for all of
245  * about two instructions).
246  *
247  * 2) manipulation of the arp table's linked list is done under the
248  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
249  * while the arp table is locked, we punt and try again later.
250  */
251 
252 static int	arp_locked;
253 static inline int arp_lock_try(int);
254 static inline void arp_unlock(void);
255 
256 static inline int
257 arp_lock_try(int recurse)
258 {
259 	int s;
260 
261 	/*
262 	 * Use splvm() -- we're blocking things that would cause
263 	 * mbuf allocation.
264 	 */
265 	s = splvm();
266 	if (!recurse && arp_locked) {
267 		splx(s);
268 		return (0);
269 	}
270 	arp_locked++;
271 	splx(s);
272 	return (1);
273 }
274 
275 static inline void
276 arp_unlock(void)
277 {
278 	int s;
279 
280 	s = splvm();
281 	arp_locked--;
282 	splx(s);
283 }
284 
285 #ifdef DIAGNOSTIC
286 #define	ARP_LOCK(recurse)						\
287 do {									\
288 	if (arp_lock_try(recurse) == 0) {				\
289 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
290 		panic("arp_lock");					\
291 	}								\
292 } while (/*CONSTCOND*/ 0)
293 #define	ARP_LOCK_CHECK()						\
294 do {									\
295 	if (arp_locked == 0) {						\
296 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
297 		panic("arp lock check");				\
298 	}								\
299 } while (/*CONSTCOND*/ 0)
300 #else
301 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
302 #define	ARP_LOCK_CHECK()	/* nothing */
303 #endif
304 
305 #define	ARP_UNLOCK()		arp_unlock()
306 
307 /*
308  * ARP protocol drain routine.  Called when memory is in short supply.
309  * Called at splvm();
310  */
311 
312 void
313 arp_drain(void)
314 {
315 	struct llinfo_arp *la, *nla;
316 	int count = 0;
317 	struct mbuf *mold;
318 
319 	if (arp_lock_try(0) == 0) {
320 		printf("arp_drain: locked; punting\n");
321 		return;
322 	}
323 
324 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
325 		nla = LIST_NEXT(la, la_list);
326 
327 		mold = la->la_hold;
328 		la->la_hold = 0;
329 
330 		if (mold) {
331 			m_freem(mold);
332 			count++;
333 		}
334 	}
335 	ARP_UNLOCK();
336 	arpstat.as_dfrdropped += count;
337 }
338 
339 
340 /*
341  * Timeout routine.  Age arp_tab entries periodically.
342  */
343 /* ARGSUSED */
344 static void
345 arptimer(void *arg __unused)
346 {
347 	int s;
348 	struct llinfo_arp *la, *nla;
349 
350 	s = splsoftnet();
351 
352 	if (arp_lock_try(0) == 0) {
353 		/* get it later.. */
354 		splx(s);
355 		return;
356 	}
357 
358 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
359 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
360 		struct rtentry *rt = la->la_rt;
361 
362 		nla = LIST_NEXT(la, la_list);
363 		if (rt->rt_expire == 0)
364 			continue;
365 		if ((rt->rt_expire - time_second) < arpt_refresh &&
366 		    rt->rt_pksent > (time_second - arpt_keep)) {
367 			/*
368 			 * If the entry has been used during since last
369 			 * refresh, try to renew it before deleting.
370 			 */
371 			arprequest(rt->rt_ifp,
372 			    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
373 			    &SIN(rt_key(rt))->sin_addr,
374 			    LLADDR(rt->rt_ifp->if_sadl));
375 		} else if (rt->rt_expire <= time_second)
376 			arptfree(la); /* timer has expired; clear */
377 	}
378 
379 	ARP_UNLOCK();
380 
381 	splx(s);
382 }
383 
384 /*
385  * Parallel to llc_rtrequest.
386  */
387 void
388 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info __unused)
389 {
390 	struct sockaddr *gate = rt->rt_gateway;
391 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
392 	static const struct sockaddr_dl null_sdl = {
393 		.sdl_len = sizeof(null_sdl),
394 		.sdl_family = AF_LINK,
395 	};
396 	size_t allocsize;
397 	struct mbuf *mold;
398 	int s;
399 	struct in_ifaddr *ia;
400 	struct ifaddr *ifa;
401 
402 	if (!arpinit_done) {
403 		arpinit_done = 1;
404 		/*
405 		 * We generate expiration times from time_second
406 		 * so avoid accidently creating permanent routes.
407 		 */
408 		if (time_second == 0) {
409 #ifdef __HAVE_TIMECOUNTER
410 			struct timespec ts;
411 			ts.tv_sec = 1;
412 			ts.tv_nsec = 0;
413 			tc_setclock(&ts);
414 #else /* !__HAVE_TIMECOUNTER */
415 			time.tv_sec++;
416 #endif /* !__HAVE_TIMECOUNTER */
417 		}
418 		callout_init(&arptimer_ch);
419 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
420 	}
421 
422 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
423 		if (req != RTM_ADD)
424 			return;
425 
426 		/*
427 		 * linklayers with particular link MTU limitation.
428 		 */
429 		switch(rt->rt_ifp->if_type) {
430 #if NFDDI > 0
431 		case IFT_FDDI:
432 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
433 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
434 			break;
435 #endif
436 #if NARC > 0
437 		case IFT_ARCNET:
438 		    {
439 			int arcipifmtu;
440 
441 			if (rt->rt_ifp->if_flags & IFF_LINK0)
442 				arcipifmtu = arc_ipmtu;
443 			else
444 				arcipifmtu = ARCMTU;
445 			if (rt->rt_ifp->if_mtu > arcipifmtu)
446 				rt->rt_rmx.rmx_mtu = arcipifmtu;
447 			break;
448 		    }
449 #endif
450 		}
451 		return;
452 	}
453 
454 	ARP_LOCK(1);		/* we may already be locked here. */
455 
456 	switch (req) {
457 
458 	case RTM_ADD:
459 		/*
460 		 * XXX: If this is a manually added route to interface
461 		 * such as older version of routed or gated might provide,
462 		 * restore cloning bit.
463 		 */
464 		if ((rt->rt_flags & RTF_HOST) == 0 &&
465 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
466 			rt->rt_flags |= RTF_CLONING;
467 		if (rt->rt_flags & RTF_CLONING) {
468 			/*
469 			 * Case 1: This route should come from a route to iface.
470 			 */
471 			rt_setgate(rt, rt_key(rt),
472 			    (const struct sockaddr *)&null_sdl);
473 			gate = rt->rt_gateway;
474 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
475 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
476 			/*
477 			 * Give this route an expiration time, even though
478 			 * it's a "permanent" route, so that routes cloned
479 			 * from it do not need their expiration time set.
480 			 */
481 			rt->rt_expire = time_second;
482 			/*
483 			 * linklayers with particular link MTU limitation.
484 			 */
485 			switch (rt->rt_ifp->if_type) {
486 #if NFDDI > 0
487 			case IFT_FDDI:
488 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
489 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
490 				     (rt->rt_rmx.rmx_mtu == 0 &&
491 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
492 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
493 				break;
494 #endif
495 #if NARC > 0
496 			case IFT_ARCNET:
497 			    {
498 				int arcipifmtu;
499 				if (rt->rt_ifp->if_flags & IFF_LINK0)
500 					arcipifmtu = arc_ipmtu;
501 				else
502 					arcipifmtu = ARCMTU;
503 
504 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
505 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
506 				     (rt->rt_rmx.rmx_mtu == 0 &&
507 				      rt->rt_ifp->if_mtu > arcipifmtu)))
508 					rt->rt_rmx.rmx_mtu = arcipifmtu;
509 				break;
510 			    }
511 #endif
512 			}
513 			break;
514 		}
515 		/* Announce a new entry if requested. */
516 		if (rt->rt_flags & RTF_ANNOUNCE)
517 			arprequest(rt->rt_ifp,
518 			    &SIN(rt_key(rt))->sin_addr,
519 			    &SIN(rt_key(rt))->sin_addr,
520 			    (u_char *)LLADDR(SDL(gate)));
521 		/*FALLTHROUGH*/
522 	case RTM_RESOLVE:
523 		if (gate->sa_family != AF_LINK ||
524 		    gate->sa_len < sizeof(null_sdl)) {
525 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
526 			break;
527 		}
528 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
529 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
530 		if (la != 0)
531 			break; /* This happens on a route change */
532 		/*
533 		 * Case 2:  This route may come from cloning, or a manual route
534 		 * add with a LL address.
535 		 */
536 		switch (SDL(gate)->sdl_type) {
537 #if NTOKEN > 0
538 		case IFT_ISO88025:
539 			allocsize = sizeof(*la) + sizeof(struct token_rif);
540 			break;
541 #endif /* NTOKEN > 0 */
542 		default:
543 			allocsize = sizeof(*la);
544 		}
545 		R_Malloc(la, struct llinfo_arp *, allocsize);
546 		rt->rt_llinfo = (caddr_t)la;
547 		if (la == 0) {
548 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
549 			break;
550 		}
551 		arp_inuse++, arp_allocated++;
552 		Bzero(la, allocsize);
553 		la->la_rt = rt;
554 		rt->rt_flags |= RTF_LLINFO;
555 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
556 
557 		INADDR_TO_IA(SIN(rt_key(rt))->sin_addr, ia);
558 		while (ia && ia->ia_ifp != rt->rt_ifp)
559 			NEXT_IA_WITH_SAME_ADDR(ia);
560 		if (ia) {
561 			/*
562 			 * This test used to be
563 			 *	if (lo0ifp->if_flags & IFF_UP)
564 			 * It allowed local traffic to be forced through
565 			 * the hardware by configuring the loopback down.
566 			 * However, it causes problems during network
567 			 * configuration for boards that can't receive
568 			 * packets they send.  It is now necessary to clear
569 			 * "useloopback" and remove the route to force
570 			 * traffic out to the hardware.
571 			 *
572 			 * In 4.4BSD, the above "if" statement checked
573 			 * rt->rt_ifa against rt_key(rt).  It was changed
574 			 * to the current form so that we can provide a
575 			 * better support for multiple IPv4 addresses on a
576 			 * interface.
577 			 */
578 			rt->rt_expire = 0;
579 			Bcopy(LLADDR(rt->rt_ifp->if_sadl),
580 			    LLADDR(SDL(gate)),
581 			    SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
582 			if (useloopback)
583 				rt->rt_ifp = lo0ifp;
584 			/*
585 			 * make sure to set rt->rt_ifa to the interface
586 			 * address we are using, otherwise we will have trouble
587 			 * with source address selection.
588 			 */
589 			ifa = &ia->ia_ifa;
590 			if (ifa != rt->rt_ifa) {
591 				IFAFREE(rt->rt_ifa);
592 				IFAREF(ifa);
593 				rt->rt_ifa = ifa;
594 			}
595 		}
596 		break;
597 
598 	case RTM_DELETE:
599 		if (la == 0)
600 			break;
601 		arp_inuse--;
602 		LIST_REMOVE(la, la_list);
603 		rt->rt_llinfo = 0;
604 		rt->rt_flags &= ~RTF_LLINFO;
605 
606 		s = splnet();
607 		mold = la->la_hold;
608 		la->la_hold = 0;
609 		splx(s);
610 
611 		if (mold)
612 			m_freem(mold);
613 
614 		Free((caddr_t)la);
615 	}
616 	ARP_UNLOCK();
617 }
618 
619 /*
620  * Broadcast an ARP request. Caller specifies:
621  *	- arp header source ip address
622  *	- arp header target ip address
623  *	- arp header source ethernet address
624  */
625 void
626 arprequest(struct ifnet *ifp,
627     struct in_addr *sip, struct in_addr *tip, u_int8_t *enaddr)
628 {
629 	struct mbuf *m;
630 	struct arphdr *ah;
631 	struct sockaddr sa;
632 
633 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
634 		return;
635 	MCLAIM(m, &arpdomain.dom_mowner);
636 	switch (ifp->if_type) {
637 	case IFT_IEEE1394:
638 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
639 		    ifp->if_addrlen;
640 		break;
641 	default:
642 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
643 		    2 * ifp->if_addrlen;
644 		break;
645 	}
646 	m->m_pkthdr.len = m->m_len;
647 	MH_ALIGN(m, m->m_len);
648 	ah = mtod(m, struct arphdr *);
649 	bzero((caddr_t)ah, m->m_len);
650 	switch (ifp->if_type) {
651 	case IFT_IEEE1394:	/* RFC2734 */
652 		/* fill it now for ar_tpa computation */
653 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
654 		break;
655 	default:
656 		/* ifp->if_output will fill ar_hrd */
657 		break;
658 	}
659 	ah->ar_pro = htons(ETHERTYPE_IP);
660 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
661 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
662 	ah->ar_op = htons(ARPOP_REQUEST);
663 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
664 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
665 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
666 	sa.sa_family = AF_ARP;
667 	sa.sa_len = 2;
668 	m->m_flags |= M_BCAST;
669 	arpstat.as_sndtotal++;
670 	arpstat.as_sndrequest++;
671 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
672 }
673 
674 /*
675  * Resolve an IP address into an ethernet address.  If success,
676  * desten is filled in.  If there is no entry in arptab,
677  * set one up and broadcast a request for the IP address.
678  * Hold onto this mbuf and resend it once the address
679  * is finally resolved.  A return value of 1 indicates
680  * that desten has been filled in and the packet should be sent
681  * normally; a 0 return indicates that the packet has been
682  * taken over here, either now or for later transmission.
683  */
684 int
685 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
686     struct sockaddr *dst, 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_second) &&
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_second; /* 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_second;
745 	}
746 #endif
747 	if (rt->rt_expire) {
748 		rt->rt_flags &= ~RTF_REJECT;
749 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
750 			rt->rt_expire = time_second;
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 #if NCARP > 0
756 				    (rt->rt_ifp->if_type == IFT_CARP) ?
757 				    LLADDR(rt->rt_ifp->if_sadl):
758 #endif
759 				    LLADDR(ifp->if_sadl));
760 			else {
761 				rt->rt_flags |= RTF_REJECT;
762 				rt->rt_expire += arpt_down;
763 				la->la_asked = 0;
764 			}
765 		}
766 	}
767 	return (0);
768 }
769 
770 /*
771  * Common length and type checks are done here,
772  * then the protocol-specific routine is called.
773  */
774 void
775 arpintr(void)
776 {
777 	struct mbuf *m;
778 	struct arphdr *ar;
779 	int s;
780 	int arplen;
781 
782 	while (arpintrq.ifq_head) {
783 		s = splnet();
784 		IF_DEQUEUE(&arpintrq, m);
785 		splx(s);
786 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
787 			panic("arpintr");
788 
789 		MCLAIM(m, &arpdomain.dom_mowner);
790 		arpstat.as_rcvtotal++;
791 
792 		/*
793 		 * First, make sure we have at least struct arphdr.
794 		 */
795 		if (m->m_len < sizeof(struct arphdr) ||
796 		    (ar = mtod(m, struct arphdr *)) == NULL)
797 			goto badlen;
798 
799 		switch (m->m_pkthdr.rcvif->if_type) {
800 		case IFT_IEEE1394:
801 			arplen = sizeof(struct arphdr) +
802 			    ar->ar_hln + 2 * ar->ar_pln;
803 			break;
804 		default:
805 			arplen = sizeof(struct arphdr) +
806 			    2 * ar->ar_hln + 2 * ar->ar_pln;
807 			break;
808 		}
809 
810 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
811 		    m->m_len >= arplen)
812 			switch (ntohs(ar->ar_pro)) {
813 			case ETHERTYPE_IP:
814 			case ETHERTYPE_IPTRAILERS:
815 				in_arpinput(m);
816 				continue;
817 			default:
818 				arpstat.as_rcvbadproto++;
819 			}
820 		else {
821 badlen:
822 			arpstat.as_rcvbadlen++;
823 		}
824 		m_freem(m);
825 	}
826 }
827 
828 /*
829  * ARP for Internet protocols on 10 Mb/s Ethernet.
830  * Algorithm is that given in RFC 826.
831  * In addition, a sanity check is performed on the sender
832  * protocol address, to catch impersonators.
833  * We no longer handle negotiations for use of trailer protocol:
834  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
835  * along with IP replies if we wanted trailers sent to us,
836  * and also sent them in response to IP replies.
837  * This allowed either end to announce the desire to receive
838  * trailer packets.
839  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
840  * but formerly didn't normally send requests.
841  */
842 static void
843 in_arpinput(struct mbuf *m)
844 {
845 	struct arphdr *ah;
846 	struct ifnet *ifp = m->m_pkthdr.rcvif;
847 	struct llinfo_arp *la = 0;
848 	struct rtentry  *rt;
849 	struct in_ifaddr *ia;
850 #if NBRIDGE > 0
851 	struct in_ifaddr *bridge_ia = NULL;
852 #endif
853 #if NCARP > 0
854 	u_int32_t count = 0, index = 0;
855 #endif
856 	struct sockaddr_dl *sdl;
857 	struct sockaddr sa;
858 	struct in_addr isaddr, itaddr, myaddr;
859 	int op;
860 	struct mbuf *mold;
861 	caddr_t tha;
862 	int s;
863 
864 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
865 		goto out;
866 	ah = mtod(m, struct arphdr *);
867 	op = ntohs(ah->ar_op);
868 
869 	/*
870 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
871 	 * ar_tpa().
872 	 */
873 	switch (ifp->if_type) {
874 	case IFT_IEEE1394:
875 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
876 			;
877 		else {
878 			/* XXX this is to make sure we compute ar_tha right */
879 			/* XXX check ar_hrd more strictly? */
880 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
881 		}
882 		break;
883 	default:
884 		/* XXX check ar_hrd? */
885 		break;
886 	}
887 
888 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
889 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
890 
891 	if (m->m_flags & (M_BCAST|M_MCAST))
892 		arpstat.as_rcvmcast++;
893 
894 	/*
895 	 * If the target IP address is zero, ignore the packet.
896 	 * This prevents the code below from tring to answer
897 	 * when we are using IP address zero (booting).
898 	 */
899 	if (in_nullhost(itaddr)) {
900 		arpstat.as_rcvzerotpa++;
901 		goto out;
902 	}
903 
904 	/*
905 	 * If the source IP address is zero, this is most likely a
906 	 * confused host trying to use IP address zero. (Windoze?)
907 	 * XXX: Should we bother trying to reply to these?
908 	 */
909 	if (in_nullhost(isaddr)) {
910 		arpstat.as_rcvzerospa++;
911 		goto out;
912 	}
913 
914 	/*
915 	 * Search for a matching interface address
916 	 * or any address on the interface to use
917 	 * as a dummy address in the rest of this function
918 	 */
919 
920 	INADDR_TO_IA(itaddr, ia);
921 	while (ia != NULL) {
922 #if NCARP > 0
923 		if (ia->ia_ifp->if_type == IFT_CARP &&
924 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
925 		    (IFF_UP|IFF_RUNNING))) {
926 			index++;
927 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
928 			    carp_iamatch(ia, ar_sha(ah),
929 			    &count, index)) {
930 				break;
931 				}
932 		} else
933 #endif
934 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
935 				break;
936 #if NBRIDGE > 0
937 		/*
938 		 * If the interface we received the packet on
939 		 * is part of a bridge, check to see if we need
940 		 * to "bridge" the packet to ourselves at this
941 		 * layer.  Note we still prefer a perfect match,
942 		 * but allow this weaker match if necessary.
943 		 */
944 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
945 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
946 			bridge_ia = ia;
947 #endif /* NBRIDGE > 0 */
948 
949 		NEXT_IA_WITH_SAME_ADDR(ia);
950 	}
951 
952 #if NBRIDGE > 0
953 	if (ia == NULL && bridge_ia != NULL) {
954 		ia = bridge_ia;
955 		ifp = bridge_ia->ia_ifp;
956 	}
957 #endif
958 
959 	if (ia == NULL) {
960 		INADDR_TO_IA(isaddr, ia);
961 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
962 			NEXT_IA_WITH_SAME_ADDR(ia);
963 
964 		if (ia == NULL) {
965 			IFP_TO_IA(ifp, ia);
966 			if (ia == NULL) {
967 				arpstat.as_rcvnoint++;
968 				goto out;
969 			}
970 		}
971 	}
972 
973 	myaddr = ia->ia_addr.sin_addr;
974 
975 	/* XXX checks for bridge case? */
976 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
977 	    ifp->if_addrlen)) {
978 		arpstat.as_rcvlocalsha++;
979 		goto out;	/* it's from me, ignore it. */
980 	}
981 
982 	/* XXX checks for bridge case? */
983 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
984 		arpstat.as_rcvbcastsha++;
985 		log(LOG_ERR,
986 		    "%s: arp: link address is broadcast for IP address %s!\n",
987 		    ifp->if_xname, in_fmtaddr(isaddr));
988 		goto out;
989 	}
990 
991 	if (in_hosteq(isaddr, myaddr)) {
992 		arpstat.as_rcvlocalspa++;
993 		log(LOG_ERR,
994 		   "duplicate IP address %s sent from link address %s\n",
995 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
996 		itaddr = myaddr;
997 		goto reply;
998 	}
999 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1000 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
1001 		if (sdl->sdl_alen &&
1002 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
1003 			if (rt->rt_flags & RTF_STATIC) {
1004 				arpstat.as_rcvoverperm++;
1005 				log(LOG_INFO,
1006 				    "%s tried to overwrite permanent arp info"
1007 				    " for %s\n",
1008 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1009 				    in_fmtaddr(isaddr));
1010 				goto out;
1011 			} else if (rt->rt_ifp != ifp) {
1012 				arpstat.as_rcvoverint++;
1013 				log(LOG_INFO,
1014 				    "%s on %s tried to overwrite "
1015 				    "arp info for %s on %s\n",
1016 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1017 				    ifp->if_xname, in_fmtaddr(isaddr),
1018 				    rt->rt_ifp->if_xname);
1019 				    goto out;
1020 			} else {
1021 				arpstat.as_rcvover++;
1022 				log(LOG_INFO,
1023 				    "arp info overwritten for %s by %s\n",
1024 				    in_fmtaddr(isaddr),
1025 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1026 			}
1027 		}
1028 		/*
1029 		 * sanity check for the address length.
1030 		 * XXX this does not work for protocols with variable address
1031 		 * length. -is
1032 		 */
1033 		if (sdl->sdl_alen &&
1034 		    sdl->sdl_alen != ah->ar_hln) {
1035 			arpstat.as_rcvlenchg++;
1036 			log(LOG_WARNING,
1037 			    "arp from %s: new addr len %d, was %d",
1038 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1039 		}
1040 		if (ifp->if_addrlen != ah->ar_hln) {
1041 			arpstat.as_rcvbadlen++;
1042 			log(LOG_WARNING,
1043 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1044 			    in_fmtaddr(isaddr), ah->ar_hln,
1045 			    ifp->if_addrlen);
1046 			goto reply;
1047 		}
1048 #if NTOKEN > 0
1049 		/*
1050 		 * XXX uses m_data and assumes the complete answer including
1051 		 * XXX token-ring headers is in the same buf
1052 		 */
1053 		if (ifp->if_type == IFT_ISO88025) {
1054 			struct token_header *trh;
1055 
1056 			trh = (struct token_header *)M_TRHSTART(m);
1057 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1058 				struct token_rif	*rif;
1059 				size_t	riflen;
1060 
1061 				rif = TOKEN_RIF(trh);
1062 				riflen = (ntohs(rif->tr_rcf) &
1063 				    TOKEN_RCF_LEN_MASK) >> 8;
1064 
1065 				if (riflen > 2 &&
1066 				    riflen < sizeof(struct token_rif) &&
1067 				    (riflen & 1) == 0) {
1068 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1069 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1070 					bcopy(rif, TOKEN_RIF(la), riflen);
1071 				}
1072 			}
1073 		}
1074 #endif /* NTOKEN > 0 */
1075 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
1076 		    sdl->sdl_alen = ah->ar_hln);
1077 		if (rt->rt_expire)
1078 			rt->rt_expire = time_second + arpt_keep;
1079 		rt->rt_flags &= ~RTF_REJECT;
1080 		la->la_asked = 0;
1081 
1082 		s = splnet();
1083 		mold = la->la_hold;
1084 		la->la_hold = 0;
1085 		splx(s);
1086 
1087 		if (mold) {
1088 			arpstat.as_dfrsent++;
1089 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
1090 		}
1091 	}
1092 reply:
1093 	if (op != ARPOP_REQUEST) {
1094 		if (op == ARPOP_REPLY)
1095 			arpstat.as_rcvreply++;
1096 	out:
1097 		m_freem(m);
1098 		return;
1099 	}
1100 	arpstat.as_rcvrequest++;
1101 	if (in_hosteq(itaddr, myaddr)) {
1102 		/* I am the target */
1103 		tha = ar_tha(ah);
1104 		if (tha)
1105 			bcopy((caddr_t)ar_sha(ah), tha, ah->ar_hln);
1106 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1107 	} else {
1108 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1109 		if (la == 0)
1110 			goto out;
1111 		rt = la->la_rt;
1112 		if (rt->rt_ifp->if_type == IFT_CARP &&
1113 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1114 			goto out;
1115 		tha = ar_tha(ah);
1116 		if (tha)
1117 			bcopy((caddr_t)ar_sha(ah), tha, ah->ar_hln);
1118 		sdl = SDL(rt->rt_gateway);
1119 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
1120 	}
1121 
1122 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
1123 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
1124 	ah->ar_op = htons(ARPOP_REPLY);
1125 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1126 	switch (ifp->if_type) {
1127 	case IFT_IEEE1394:
1128 		/*
1129 		 * ieee1394 arp reply is broadcast
1130 		 */
1131 		m->m_flags &= ~M_MCAST;
1132 		m->m_flags |= M_BCAST;
1133 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1134 		break;
1135 
1136 	default:
1137 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1138 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1139 		break;
1140 	}
1141 	m->m_pkthdr.len = m->m_len;
1142 	sa.sa_family = AF_ARP;
1143 	sa.sa_len = 2;
1144 	arpstat.as_sndtotal++;
1145 	arpstat.as_sndreply++;
1146 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1147 	return;
1148 }
1149 
1150 /*
1151  * Free an arp entry.
1152  */
1153 static void arptfree(struct llinfo_arp *la)
1154 {
1155 	struct rtentry *rt = la->la_rt;
1156 	struct sockaddr_dl *sdl;
1157 
1158 	ARP_LOCK_CHECK();
1159 
1160 	if (rt == 0)
1161 		panic("arptfree");
1162 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
1163 	    sdl->sdl_family == AF_LINK) {
1164 		sdl->sdl_alen = 0;
1165 		la->la_asked = 0;
1166 		rt->rt_flags &= ~RTF_REJECT;
1167 		return;
1168 	}
1169 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
1170 	    0, (struct rtentry **)0);
1171 }
1172 
1173 /*
1174  * Lookup or enter a new address in arptab.
1175  */
1176 static struct llinfo_arp *
1177 arplookup(struct mbuf *m, struct in_addr *addr, int create, int proxy)
1178 {
1179 	struct arphdr *ah;
1180 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1181 	struct rtentry *rt;
1182 	static struct sockaddr_inarp sin;
1183 	const char *why = 0;
1184 
1185 	ah = mtod(m, struct arphdr *);
1186 	sin.sin_len = sizeof(sin);
1187 	sin.sin_family = AF_INET;
1188 	sin.sin_addr = *addr;
1189 	sin.sin_other = proxy ? SIN_PROXY : 0;
1190 	rt = rtalloc1(sintosa(&sin), create);
1191 	if (rt == 0)
1192 		return (0);
1193 	rt->rt_refcnt--;
1194 
1195 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO &&
1196 	    rt->rt_gateway->sa_family == AF_LINK)
1197 		return ((struct llinfo_arp *)rt->rt_llinfo);
1198 
1199 
1200 
1201 	if (create) {
1202 		if (rt->rt_flags & RTF_GATEWAY)
1203 			why = "host is not on local network";
1204 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1205 			arpstat.as_allocfail++;
1206 			why = "could not allocate llinfo";
1207 		} else
1208 			why = "gateway route is not ours";
1209 		log(LOG_DEBUG, "arplookup: unable to enter address"
1210 		    " for %s@%s on %s (%s)\n",
1211 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1212 		    (ifp) ? ifp->if_xname : 0, why);
1213 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1214 			rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
1215 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1216 		}
1217 	}
1218 	return (0);
1219 }
1220 
1221 int
1222 arpioctl(u_long cmd __unused, caddr_t data __unused)
1223 {
1224 
1225 	return (EOPNOTSUPP);
1226 }
1227 
1228 void
1229 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1230 {
1231 	struct in_addr *ip;
1232 
1233 	/*
1234 	 * Warn the user if another station has this IP address,
1235 	 * but only if the interface IP address is not zero.
1236 	 */
1237 	ip = &IA_SIN(ifa)->sin_addr;
1238 	if (!in_nullhost(*ip))
1239 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
1240 
1241 	ifa->ifa_rtrequest = arp_rtrequest;
1242 	ifa->ifa_flags |= RTF_CLONING;
1243 }
1244 
1245 /*
1246  * Called from 10 Mb/s Ethernet interrupt handlers
1247  * when ether packet type ETHERTYPE_REVARP
1248  * is received.  Common length and type checks are done here,
1249  * then the protocol-specific routine is called.
1250  */
1251 void
1252 revarpinput(struct mbuf *m)
1253 {
1254 	struct arphdr *ar;
1255 
1256 	if (m->m_len < sizeof(struct arphdr))
1257 		goto out;
1258 	ar = mtod(m, struct arphdr *);
1259 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1260 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1261 		goto out;
1262 #endif
1263 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1264 		goto out;
1265 	switch (ntohs(ar->ar_pro)) {
1266 	case ETHERTYPE_IP:
1267 	case ETHERTYPE_IPTRAILERS:
1268 		in_revarpinput(m);
1269 		return;
1270 
1271 	default:
1272 		break;
1273 	}
1274 out:
1275 	m_freem(m);
1276 }
1277 
1278 /*
1279  * RARP for Internet protocols on 10 Mb/s Ethernet.
1280  * Algorithm is that given in RFC 903.
1281  * We are only using for bootstrap purposes to get an ip address for one of
1282  * our interfaces.  Thus we support no user-interface.
1283  *
1284  * Since the contents of the RARP reply are specific to the interface that
1285  * sent the request, this code must ensure that they are properly associated.
1286  *
1287  * Note: also supports ARP via RARP packets, per the RFC.
1288  */
1289 void
1290 in_revarpinput(struct mbuf *m)
1291 {
1292 	struct ifnet *ifp;
1293 	struct arphdr *ah;
1294 	caddr_t tha;
1295 	int op;
1296 
1297 	ah = mtod(m, struct arphdr *);
1298 	op = ntohs(ah->ar_op);
1299 
1300 	switch (m->m_pkthdr.rcvif->if_type) {
1301 	case IFT_IEEE1394:
1302 		/* ARP without target hardware address is not supported */
1303 		goto out;
1304 	default:
1305 		break;
1306 	}
1307 
1308 	switch (op) {
1309 	case ARPOP_REQUEST:
1310 	case ARPOP_REPLY:	/* per RFC */
1311 		in_arpinput(m);
1312 		return;
1313 	case ARPOP_REVREPLY:
1314 		break;
1315 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1316 	default:
1317 		goto out;
1318 	}
1319 	if (!revarp_in_progress)
1320 		goto out;
1321 	ifp = m->m_pkthdr.rcvif;
1322 	if (ifp != myip_ifp) /* !same interface */
1323 		goto out;
1324 	if (myip_initialized)
1325 		goto wake;
1326 	tha = ar_tha(ah);
1327 	KASSERT(tha);
1328 	if (bcmp(tha, LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1329 		goto out;
1330 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1331 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1332 	myip_initialized = 1;
1333 wake:	/* Do wakeup every time in case it was missed. */
1334 	wakeup((caddr_t)&myip);
1335 
1336 out:
1337 	m_freem(m);
1338 }
1339 
1340 /*
1341  * Send a RARP request for the ip address of the specified interface.
1342  * The request should be RFC 903-compliant.
1343  */
1344 void
1345 revarprequest(struct ifnet *ifp)
1346 {
1347 	struct sockaddr sa;
1348 	struct mbuf *m;
1349 	struct arphdr *ah;
1350 	caddr_t tha;
1351 
1352 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1353 		return;
1354 	MCLAIM(m, &arpdomain.dom_mowner);
1355 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1356 	    2*ifp->if_addrlen;
1357 	m->m_pkthdr.len = m->m_len;
1358 	MH_ALIGN(m, m->m_len);
1359 	ah = mtod(m, struct arphdr *);
1360 	bzero((caddr_t)ah, m->m_len);
1361 	ah->ar_pro = htons(ETHERTYPE_IP);
1362 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1363 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1364 	ah->ar_op = htons(ARPOP_REVREQUEST);
1365 
1366 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1367 	tha = ar_tha(ah);
1368 	KASSERT(tha);
1369 	bcopy(LLADDR(ifp->if_sadl), tha, ah->ar_hln);
1370 
1371 	sa.sa_family = AF_ARP;
1372 	sa.sa_len = 2;
1373 	m->m_flags |= M_BCAST;
1374 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1375 
1376 }
1377 
1378 /*
1379  * RARP for the ip address of the specified interface, but also
1380  * save the ip address of the server that sent the answer.
1381  * Timeout if no response is received.
1382  */
1383 int
1384 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1385     struct in_addr *clnt_in)
1386 {
1387 	int result, count = 20;
1388 
1389 	myip_initialized = 0;
1390 	myip_ifp = ifp;
1391 
1392 	revarp_in_progress = 1;
1393 	while (count--) {
1394 		revarprequest(ifp);
1395 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1396 		if (result != EWOULDBLOCK)
1397 			break;
1398 	}
1399 	revarp_in_progress = 0;
1400 
1401 	if (!myip_initialized)
1402 		return ENETUNREACH;
1403 
1404 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1405 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1406 	return 0;
1407 }
1408 
1409 
1410 
1411 #ifdef DDB
1412 
1413 #include <machine/db_machdep.h>
1414 #include <ddb/db_interface.h>
1415 #include <ddb/db_output.h>
1416 
1417 static void
1418 db_print_sa(const struct sockaddr *sa)
1419 {
1420 	int len;
1421 	const u_char *p;
1422 
1423 	if (sa == 0) {
1424 		db_printf("[NULL]");
1425 		return;
1426 	}
1427 
1428 	p = (const u_char *)sa;
1429 	len = sa->sa_len;
1430 	db_printf("[");
1431 	while (len > 0) {
1432 		db_printf("%d", *p);
1433 		p++; len--;
1434 		if (len) db_printf(",");
1435 	}
1436 	db_printf("]\n");
1437 }
1438 
1439 static void
1440 db_print_ifa(struct ifaddr *ifa)
1441 {
1442 	if (ifa == 0)
1443 		return;
1444 	db_printf("  ifa_addr=");
1445 	db_print_sa(ifa->ifa_addr);
1446 	db_printf("  ifa_dsta=");
1447 	db_print_sa(ifa->ifa_dstaddr);
1448 	db_printf("  ifa_mask=");
1449 	db_print_sa(ifa->ifa_netmask);
1450 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1451 			  ifa->ifa_flags,
1452 			  ifa->ifa_refcnt,
1453 			  ifa->ifa_metric);
1454 }
1455 
1456 static void
1457 db_print_llinfo(caddr_t li)
1458 {
1459 	struct llinfo_arp *la;
1460 
1461 	if (li == 0)
1462 		return;
1463 	la = (struct llinfo_arp *)li;
1464 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1465 			  la->la_rt, la->la_hold, la->la_asked);
1466 }
1467 
1468 /*
1469  * Function to pass to rn_walktree().
1470  * Return non-zero error to abort walk.
1471  */
1472 static int
1473 db_show_radix_node(struct radix_node *rn, void *w __unused)
1474 {
1475 	struct rtentry *rt = (struct rtentry *)rn;
1476 
1477 	db_printf("rtentry=%p", rt);
1478 
1479 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1480 			  rt->rt_flags, rt->rt_refcnt,
1481 			  rt->rt_use, rt->rt_expire);
1482 
1483 	db_printf(" key="); db_print_sa(rt_key(rt));
1484 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1485 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1486 
1487 	db_printf(" ifp=%p ", rt->rt_ifp);
1488 	if (rt->rt_ifp)
1489 		db_printf("(%s)", rt->rt_ifp->if_xname);
1490 	else
1491 		db_printf("(NULL)");
1492 
1493 	db_printf(" ifa=%p\n", rt->rt_ifa);
1494 	db_print_ifa(rt->rt_ifa);
1495 
1496 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1497 
1498 	db_printf(" gwroute=%p llinfo=%p\n",
1499 			  rt->rt_gwroute, rt->rt_llinfo);
1500 	db_print_llinfo(rt->rt_llinfo);
1501 
1502 	return (0);
1503 }
1504 
1505 /*
1506  * Function to print all the route trees.
1507  * Use this from ddb:  "show arptab"
1508  */
1509 void
1510 db_show_arptab(db_expr_t addr __unused, int have_addr __unused,
1511     db_expr_t count __unused, const char *modif __unused)
1512 {
1513 	struct radix_node_head *rnh;
1514 	rnh = rt_tables[AF_INET];
1515 	db_printf("Route tree for AF_INET\n");
1516 	if (rnh == NULL) {
1517 		db_printf(" (not initialized)\n");
1518 		return;
1519 	}
1520 	rn_walktree(rnh, db_show_radix_node, NULL);
1521 	return;
1522 }
1523 #endif
1524 
1525 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1526 {
1527 	const struct sysctlnode *node;
1528 
1529 	sysctl_createv(clog, 0, NULL, NULL,
1530 			CTLFLAG_PERMANENT,
1531 			CTLTYPE_NODE, "net", NULL,
1532 			NULL, 0, NULL, 0,
1533 			CTL_NET, CTL_EOL);
1534 	sysctl_createv(clog, 0, NULL, NULL,
1535 			CTLFLAG_PERMANENT,
1536 			CTLTYPE_NODE, "inet", NULL,
1537 			NULL, 0, NULL, 0,
1538 			CTL_NET, PF_INET, CTL_EOL);
1539 	sysctl_createv(clog, 0, NULL, &node,
1540 			CTLFLAG_PERMANENT,
1541 			CTLTYPE_NODE, "arp",
1542 			SYSCTL_DESCR("Address Resolution Protocol"),
1543 			NULL, 0, NULL, 0,
1544 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1545 
1546 	sysctl_createv(clog, 0, NULL, NULL,
1547 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1548 			CTLTYPE_INT, "prune",
1549 			SYSCTL_DESCR("ARP cache pruning interval"),
1550 			NULL, 0, &arpt_prune, 0,
1551 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1552 
1553 	sysctl_createv(clog, 0, NULL, NULL,
1554 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1555 			CTLTYPE_INT, "keep",
1556 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1557 			NULL, 0, &arpt_keep, 0,
1558 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1559 
1560 	sysctl_createv(clog, 0, NULL, NULL,
1561 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1562 			CTLTYPE_INT, "down",
1563 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1564 			NULL, 0, &arpt_down, 0,
1565 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1566 
1567 	sysctl_createv(clog, 0, NULL, NULL,
1568 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1569 			CTLTYPE_INT, "refresh",
1570 			SYSCTL_DESCR("ARP entry refresh interval"),
1571 			NULL, 0, &arpt_refresh, 0,
1572 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1573 }
1574 
1575 #endif /* INET */
1576