xref: /netbsd-src/sys/netinet/if_arp.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: if_arp.c,v 1.106 2005/06/20 02:49:18 atatat Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Public Access Networks Corporation ("Panix").  It was developed under
9  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1988, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the University nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  *
68  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
69  */
70 
71 /*
72  * Ethernet address resolution protocol.
73  * TODO:
74  *	add "inuse/lock" bit (or ref. count) along with valid bit
75  */
76 
77 #include <sys/cdefs.h>
78 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.106 2005/06/20 02:49:18 atatat Exp $");
79 
80 #include "opt_ddb.h"
81 #include "opt_inet.h"
82 
83 #ifdef INET
84 
85 #include "bridge.h"
86 
87 #include <sys/param.h>
88 #include <sys/systm.h>
89 #include <sys/callout.h>
90 #include <sys/malloc.h>
91 #include <sys/mbuf.h>
92 #include <sys/socket.h>
93 #include <sys/time.h>
94 #include <sys/kernel.h>
95 #include <sys/errno.h>
96 #include <sys/ioctl.h>
97 #include <sys/syslog.h>
98 #include <sys/proc.h>
99 #include <sys/protosw.h>
100 #include <sys/domain.h>
101 #include <sys/sysctl.h>
102 
103 #include <net/ethertypes.h>
104 #include <net/if.h>
105 #include <net/if_dl.h>
106 #include <net/if_token.h>
107 #include <net/if_types.h>
108 #include <net/route.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/in_var.h>
113 #include <netinet/ip.h>
114 #include <netinet/if_inarp.h>
115 
116 #include "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 static	void arprequest(struct ifnet *,
145 	    struct in_addr *, struct in_addr *, u_int8_t *);
146 static	void arptfree(struct llinfo_arp *);
147 static	void arptimer(void *);
148 static	struct llinfo_arp *arplookup(struct mbuf *, struct in_addr *,
149 					  int, int);
150 static	void in_arpinput(struct mbuf *);
151 
152 LIST_HEAD(, llinfo_arp) llinfo_arp;
153 struct	ifqueue arpintrq = {0, 0, 0, 50};
154 int	arp_inuse, arp_allocated, arp_intimer;
155 int	arp_maxtries = 5;
156 int	useloopback = 1;	/* use loopback interface for local traffic */
157 int	arpinit_done = 0;
158 
159 struct	arpstat arpstat;
160 struct	callout arptimer_ch;
161 
162 
163 /* revarp state */
164 static struct	in_addr myip, srv_ip;
165 static int	myip_initialized = 0;
166 static int	revarp_in_progress = 0;
167 static struct	ifnet *myip_ifp = NULL;
168 
169 #ifdef DDB
170 static void db_print_sa(const struct sockaddr *);
171 static void db_print_ifa(struct ifaddr *);
172 static void db_print_llinfo(caddr_t);
173 static int db_show_radix_node(struct radix_node *, void *);
174 #endif
175 
176 /*
177  * this should be elsewhere.
178  */
179 
180 static char *
181 lla_snprintf(u_int8_t *, int);
182 
183 static char *
184 lla_snprintf(u_int8_t *adrp, int len)
185 {
186 #define NUMBUFS 3
187 	static char buf[NUMBUFS][16*3];
188 	static int bnum = 0;
189 
190 	int i;
191 	char *p;
192 
193 	p = buf[bnum];
194 
195 	*p++ = hexdigits[(*adrp)>>4];
196 	*p++ = hexdigits[(*adrp++)&0xf];
197 
198 	for (i=1; i<len && i<16; i++) {
199 		*p++ = ':';
200 		*p++ = hexdigits[(*adrp)>>4];
201 		*p++ = hexdigits[(*adrp++)&0xf];
202 	}
203 
204 	*p = 0;
205 	p = buf[bnum];
206 	bnum = (bnum + 1) % NUMBUFS;
207 	return p;
208 }
209 
210 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
211 
212 const struct protosw arpsw[] = {
213 	{ 0, &arpdomain, 0, 0,
214 	  0, 0, 0, 0,
215 	  0,
216 	  0, 0, 0, arp_drain,
217 	}
218 };
219 
220 
221 struct domain arpdomain =
222 { 	PF_ARP,  "arp", 0, 0, 0,
223 	arpsw, &arpsw[sizeof(arpsw)/sizeof(arpsw[0])]
224 };
225 
226 /*
227  * ARP table locking.
228  *
229  * to prevent lossage vs. the arp_drain routine (which may be called at
230  * any time, including in a device driver context), we do two things:
231  *
232  * 1) manipulation of la->la_hold is done at splnet() (for all of
233  * about two instructions).
234  *
235  * 2) manipulation of the arp table's linked list is done under the
236  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
237  * while the arp table is locked, we punt and try again later.
238  */
239 
240 static int	arp_locked;
241 static __inline int arp_lock_try(int);
242 static __inline void arp_unlock(void);
243 
244 static __inline int
245 arp_lock_try(int recurse)
246 {
247 	int s;
248 
249 	/*
250 	 * Use splvm() -- we're blocking things that would cause
251 	 * mbuf allocation.
252 	 */
253 	s = splvm();
254 	if (!recurse && arp_locked) {
255 		splx(s);
256 		return (0);
257 	}
258 	arp_locked++;
259 	splx(s);
260 	return (1);
261 }
262 
263 static __inline void
264 arp_unlock(void)
265 {
266 	int s;
267 
268 	s = splvm();
269 	arp_locked--;
270 	splx(s);
271 }
272 
273 #ifdef DIAGNOSTIC
274 #define	ARP_LOCK(recurse)						\
275 do {									\
276 	if (arp_lock_try(recurse) == 0) {				\
277 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
278 		panic("arp_lock");					\
279 	}								\
280 } while (/*CONSTCOND*/ 0)
281 #define	ARP_LOCK_CHECK()						\
282 do {									\
283 	if (arp_locked == 0) {						\
284 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
285 		panic("arp lock check");				\
286 	}								\
287 } while (/*CONSTCOND*/ 0)
288 #else
289 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
290 #define	ARP_LOCK_CHECK()	/* nothing */
291 #endif
292 
293 #define	ARP_UNLOCK()		arp_unlock()
294 
295 /*
296  * ARP protocol drain routine.  Called when memory is in short supply.
297  * Called at splvm();
298  */
299 
300 void
301 arp_drain(void)
302 {
303 	struct llinfo_arp *la, *nla;
304 	int count = 0;
305 	struct mbuf *mold;
306 
307 	if (arp_lock_try(0) == 0) {
308 		printf("arp_drain: locked; punting\n");
309 		return;
310 	}
311 
312 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
313 		nla = LIST_NEXT(la, la_list);
314 
315 		mold = la->la_hold;
316 		la->la_hold = 0;
317 
318 		if (mold) {
319 			m_freem(mold);
320 			count++;
321 		}
322 	}
323 	ARP_UNLOCK();
324 	arpstat.as_dfrdropped += count;
325 }
326 
327 
328 /*
329  * Timeout routine.  Age arp_tab entries periodically.
330  */
331 /* ARGSUSED */
332 static void
333 arptimer(void *arg)
334 {
335 	int s;
336 	struct llinfo_arp *la, *nla;
337 
338 	s = splsoftnet();
339 
340 	if (arp_lock_try(0) == 0) {
341 		/* get it later.. */
342 		splx(s);
343 		return;
344 	}
345 
346 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
347 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
348 		struct rtentry *rt = la->la_rt;
349 
350 		nla = LIST_NEXT(la, la_list);
351 		if (rt->rt_expire == 0)
352 			continue;
353 		if ((rt->rt_expire - time.tv_sec) < arpt_refresh &&
354 		    rt->rt_pksent > (time.tv_sec - arpt_keep)) {
355 			/*
356 			 * If the entry has been used during since last
357 			 * refresh, try to renew it before deleting.
358 			 */
359 			arprequest(rt->rt_ifp,
360 			    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
361 			    &SIN(rt_key(rt))->sin_addr,
362 			    LLADDR(rt->rt_ifp->if_sadl));
363 		} else if (rt->rt_expire <= time.tv_sec)
364 			arptfree(la); /* timer has expired; clear */
365 	}
366 
367 	ARP_UNLOCK();
368 
369 	splx(s);
370 }
371 
372 /*
373  * Parallel to llc_rtrequest.
374  */
375 void
376 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
377 {
378 	struct sockaddr *gate = rt->rt_gateway;
379 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
380 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
381 	size_t allocsize;
382 	struct mbuf *mold;
383 	int s;
384 	struct in_ifaddr *ia;
385 	struct ifaddr *ifa;
386 
387 	if (!arpinit_done) {
388 		arpinit_done = 1;
389 		/*
390 		 * We generate expiration times from time.tv_sec
391 		 * so avoid accidently creating permanent routes.
392 		 */
393 		if (time.tv_sec == 0) {
394 			time.tv_sec++;
395 		}
396 		callout_init(&arptimer_ch);
397 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
398 	}
399 
400 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
401 		if (req != RTM_ADD)
402 			return;
403 
404 		/*
405 		 * linklayers with particular link MTU limitation.
406 		 */
407 		switch(rt->rt_ifp->if_type) {
408 #if NFDDI > 0
409 		case IFT_FDDI:
410 			if (rt->rt_ifp->if_mtu > FDDIIPMTU)
411 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
412 			break;
413 #endif
414 #if NARC > 0
415 		case IFT_ARCNET:
416 		    {
417 			int arcipifmtu;
418 
419 			if (rt->rt_ifp->if_flags & IFF_LINK0)
420 				arcipifmtu = arc_ipmtu;
421 			else
422 				arcipifmtu = ARCMTU;
423 			if (rt->rt_ifp->if_mtu > arcipifmtu)
424 				rt->rt_rmx.rmx_mtu = arcipifmtu;
425 			break;
426 		    }
427 #endif
428 		}
429 		return;
430 	}
431 
432 	ARP_LOCK(1);		/* we may already be locked here. */
433 
434 	switch (req) {
435 
436 	case RTM_ADD:
437 		/*
438 		 * XXX: If this is a manually added route to interface
439 		 * such as older version of routed or gated might provide,
440 		 * restore cloning bit.
441 		 */
442 		if ((rt->rt_flags & RTF_HOST) == 0 &&
443 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
444 			rt->rt_flags |= RTF_CLONING;
445 		if (rt->rt_flags & RTF_CLONING) {
446 			/*
447 			 * Case 1: This route should come from a route to iface.
448 			 */
449 			rt_setgate(rt, rt_key(rt),
450 					(struct sockaddr *)&null_sdl);
451 			gate = rt->rt_gateway;
452 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
453 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
454 			/*
455 			 * Give this route an expiration time, even though
456 			 * it's a "permanent" route, so that routes cloned
457 			 * from it do not need their expiration time set.
458 			 */
459 			rt->rt_expire = time.tv_sec;
460 			/*
461 			 * linklayers with particular link MTU limitation.
462 			 */
463 			switch (rt->rt_ifp->if_type) {
464 #if NFDDI > 0
465 			case IFT_FDDI:
466 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
467 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
468 				     (rt->rt_rmx.rmx_mtu == 0 &&
469 				      rt->rt_ifp->if_mtu > FDDIIPMTU)))
470 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
471 				break;
472 #endif
473 #if NARC > 0
474 			case IFT_ARCNET:
475 			    {
476 				int arcipifmtu;
477 				if (rt->rt_ifp->if_flags & IFF_LINK0)
478 					arcipifmtu = arc_ipmtu;
479 				else
480 					arcipifmtu = ARCMTU;
481 
482 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
483 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
484 				     (rt->rt_rmx.rmx_mtu == 0 &&
485 				      rt->rt_ifp->if_mtu > arcipifmtu)))
486 					rt->rt_rmx.rmx_mtu = arcipifmtu;
487 				break;
488 			    }
489 #endif
490 			}
491 			break;
492 		}
493 		/* Announce a new entry if requested. */
494 		if (rt->rt_flags & RTF_ANNOUNCE)
495 			arprequest(rt->rt_ifp,
496 			    &SIN(rt_key(rt))->sin_addr,
497 			    &SIN(rt_key(rt))->sin_addr,
498 			    (u_char *)LLADDR(SDL(gate)));
499 		/*FALLTHROUGH*/
500 	case RTM_RESOLVE:
501 		if (gate->sa_family != AF_LINK ||
502 		    gate->sa_len < sizeof(null_sdl)) {
503 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
504 			break;
505 		}
506 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
507 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
508 		if (la != 0)
509 			break; /* This happens on a route change */
510 		/*
511 		 * Case 2:  This route may come from cloning, or a manual route
512 		 * add with a LL address.
513 		 */
514 		switch (SDL(gate)->sdl_type) {
515 #if NTOKEN > 0
516 		case IFT_ISO88025:
517 			allocsize = sizeof(*la) + sizeof(struct token_rif);
518 			break;
519 #endif /* NTOKEN > 0 */
520 		default:
521 			allocsize = sizeof(*la);
522 		}
523 		R_Malloc(la, struct llinfo_arp *, allocsize);
524 		rt->rt_llinfo = (caddr_t)la;
525 		if (la == 0) {
526 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
527 			break;
528 		}
529 		arp_inuse++, arp_allocated++;
530 		Bzero(la, allocsize);
531 		la->la_rt = rt;
532 		rt->rt_flags |= RTF_LLINFO;
533 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
534 
535 		INADDR_TO_IA(SIN(rt_key(rt))->sin_addr, ia);
536 		while (ia && ia->ia_ifp != rt->rt_ifp)
537 			NEXT_IA_WITH_SAME_ADDR(ia);
538 		if (ia) {
539 			/*
540 			 * This test used to be
541 			 *	if (lo0ifp->if_flags & IFF_UP)
542 			 * It allowed local traffic to be forced through
543 			 * the hardware by configuring the loopback down.
544 			 * However, it causes problems during network
545 			 * configuration for boards that can't receive
546 			 * packets they send.  It is now necessary to clear
547 			 * "useloopback" and remove the route to force
548 			 * traffic out to the hardware.
549 			 *
550 			 * In 4.4BSD, the above "if" statement checked
551 			 * rt->rt_ifa against rt_key(rt).  It was changed
552 			 * to the current form so that we can provide a
553 			 * better support for multiple IPv4 addresses on a
554 			 * interface.
555 			 */
556 			rt->rt_expire = 0;
557 			Bcopy(LLADDR(rt->rt_ifp->if_sadl),
558 			    LLADDR(SDL(gate)),
559 			    SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
560 			if (useloopback)
561 				rt->rt_ifp = lo0ifp;
562 			/*
563 			 * make sure to set rt->rt_ifa to the interface
564 			 * address we are using, otherwise we will have trouble
565 			 * with source address selection.
566 			 */
567 			ifa = &ia->ia_ifa;
568 			if (ifa != rt->rt_ifa) {
569 				IFAFREE(rt->rt_ifa);
570 				IFAREF(ifa);
571 				rt->rt_ifa = ifa;
572 			}
573 		}
574 		break;
575 
576 	case RTM_DELETE:
577 		if (la == 0)
578 			break;
579 		arp_inuse--;
580 		LIST_REMOVE(la, la_list);
581 		rt->rt_llinfo = 0;
582 		rt->rt_flags &= ~RTF_LLINFO;
583 
584 		s = splnet();
585 		mold = la->la_hold;
586 		la->la_hold = 0;
587 		splx(s);
588 
589 		if (mold)
590 			m_freem(mold);
591 
592 		Free((caddr_t)la);
593 	}
594 	ARP_UNLOCK();
595 }
596 
597 /*
598  * Broadcast an ARP request. Caller specifies:
599  *	- arp header source ip address
600  *	- arp header target ip address
601  *	- arp header source ethernet address
602  */
603 static void
604 arprequest(struct ifnet *ifp,
605     struct in_addr *sip, struct in_addr *tip, u_int8_t *enaddr)
606 {
607 	struct mbuf *m;
608 	struct arphdr *ah;
609 	struct sockaddr sa;
610 
611 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
612 		return;
613 	MCLAIM(m, &arpdomain.dom_mowner);
614 	switch (ifp->if_type) {
615 	case IFT_IEEE1394:
616 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
617 		    ifp->if_addrlen;
618 		break;
619 	default:
620 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
621 		    2 * ifp->if_addrlen;
622 		break;
623 	}
624 	m->m_pkthdr.len = m->m_len;
625 	MH_ALIGN(m, m->m_len);
626 	ah = mtod(m, struct arphdr *);
627 	bzero((caddr_t)ah, m->m_len);
628 	switch (ifp->if_type) {
629 	case IFT_IEEE1394:	/* RFC2734 */
630 		/* fill it now for ar_tpa computation */
631 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
632 		break;
633 	default:
634 		/* ifp->if_output will fill ar_hrd */
635 		break;
636 	}
637 	ah->ar_pro = htons(ETHERTYPE_IP);
638 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
639 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
640 	ah->ar_op = htons(ARPOP_REQUEST);
641 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
642 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
643 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
644 	sa.sa_family = AF_ARP;
645 	sa.sa_len = 2;
646 	m->m_flags |= M_BCAST;
647 	arpstat.as_sndtotal++;
648 	arpstat.as_sndrequest++;
649 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
650 }
651 
652 /*
653  * Resolve an IP address into an ethernet address.  If success,
654  * desten is filled in.  If there is no entry in arptab,
655  * set one up and broadcast a request for the IP address.
656  * Hold onto this mbuf and resend it once the address
657  * is finally resolved.  A return value of 1 indicates
658  * that desten has been filled in and the packet should be sent
659  * normally; a 0 return indicates that the packet has been
660  * taken over here, either now or for later transmission.
661  */
662 int
663 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
664     struct sockaddr *dst, u_char *desten)
665 {
666 	struct llinfo_arp *la;
667 	struct sockaddr_dl *sdl;
668 	struct mbuf *mold;
669 	int s;
670 
671 	if (rt)
672 		la = (struct llinfo_arp *)rt->rt_llinfo;
673 	else {
674 		if ((la = arplookup(m, &SIN(dst)->sin_addr, 1, 0)) != NULL)
675 			rt = la->la_rt;
676 	}
677 	if (la == 0 || rt == 0) {
678 		arpstat.as_allocfail++;
679 		log(LOG_DEBUG,
680 		    "arpresolve: can't allocate llinfo on %s for %s\n",
681 		    ifp->if_xname, in_fmtaddr(SIN(dst)->sin_addr));
682 		m_freem(m);
683 		return (0);
684 	}
685 	sdl = SDL(rt->rt_gateway);
686 	/*
687 	 * Check the address family and length is valid, the address
688 	 * is resolved; otherwise, try to resolve.
689 	 */
690 	if ((rt->rt_expire == 0 || rt->rt_expire > time.tv_sec) &&
691 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
692 		bcopy(LLADDR(sdl), desten,
693 		    min(sdl->sdl_alen, ifp->if_addrlen));
694 		rt->rt_pksent = time.tv_sec; /* Time for last pkt sent */
695 		return 1;
696 	}
697 	/*
698 	 * There is an arptab entry, but no ethernet address
699 	 * response yet.  Replace the held mbuf with this
700 	 * latest one.
701 	 */
702 
703 	arpstat.as_dfrtotal++;
704 	s = splnet();
705 	mold = la->la_hold;
706 	la->la_hold = m;
707 	splx(s);
708 
709 	if (mold) {
710 		arpstat.as_dfrdropped++;
711 		m_freem(mold);
712 	}
713 
714 	/*
715 	 * Re-send the ARP request when appropriate.
716 	 */
717 #ifdef	DIAGNOSTIC
718 	if (rt->rt_expire == 0) {
719 		/* This should never happen. (Should it? -gwr) */
720 		printf("arpresolve: unresolved and rt_expire == 0\n");
721 		/* Set expiration time to now (expired). */
722 		rt->rt_expire = time.tv_sec;
723 	}
724 #endif
725 	if (rt->rt_expire) {
726 		rt->rt_flags &= ~RTF_REJECT;
727 		if (la->la_asked == 0 || rt->rt_expire != time.tv_sec) {
728 			rt->rt_expire = time.tv_sec;
729 			if (la->la_asked++ < arp_maxtries)
730 				arprequest(ifp,
731 				    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
732 				    &SIN(dst)->sin_addr,
733 				    LLADDR(ifp->if_sadl));
734 			else {
735 				rt->rt_flags |= RTF_REJECT;
736 				rt->rt_expire += arpt_down;
737 				la->la_asked = 0;
738 			}
739 		}
740 	}
741 	return (0);
742 }
743 
744 /*
745  * Common length and type checks are done here,
746  * then the protocol-specific routine is called.
747  */
748 void
749 arpintr(void)
750 {
751 	struct mbuf *m;
752 	struct arphdr *ar;
753 	int s;
754 	int arplen;
755 
756 	while (arpintrq.ifq_head) {
757 		s = splnet();
758 		IF_DEQUEUE(&arpintrq, m);
759 		splx(s);
760 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
761 			panic("arpintr");
762 
763 		MCLAIM(m, &arpdomain.dom_mowner);
764 		arpstat.as_rcvtotal++;
765 
766 		/*
767 		 * First, make sure we have at least struct arphdr.
768 		 */
769 		if (m->m_len < sizeof(struct arphdr) ||
770 		    (ar = mtod(m, struct arphdr *)) == NULL)
771 			goto badlen;
772 
773 		switch (m->m_pkthdr.rcvif->if_type) {
774 		case IFT_IEEE1394:
775 			arplen = sizeof(struct arphdr) +
776 			    ar->ar_hln + 2 * ar->ar_pln;
777 			break;
778 		default:
779 			arplen = sizeof(struct arphdr) +
780 			    2 * ar->ar_hln + 2 * ar->ar_pln;
781 			break;
782 		}
783 
784 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
785 		    m->m_len >= arplen)
786 			switch (ntohs(ar->ar_pro)) {
787 			case ETHERTYPE_IP:
788 			case ETHERTYPE_IPTRAILERS:
789 				in_arpinput(m);
790 				continue;
791 			default:
792 				arpstat.as_rcvbadproto++;
793 			}
794 		else {
795 badlen:
796 			arpstat.as_rcvbadlen++;
797 		}
798 		m_freem(m);
799 	}
800 }
801 
802 /*
803  * ARP for Internet protocols on 10 Mb/s Ethernet.
804  * Algorithm is that given in RFC 826.
805  * In addition, a sanity check is performed on the sender
806  * protocol address, to catch impersonators.
807  * We no longer handle negotiations for use of trailer protocol:
808  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
809  * along with IP replies if we wanted trailers sent to us,
810  * and also sent them in response to IP replies.
811  * This allowed either end to announce the desire to receive
812  * trailer packets.
813  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
814  * but formerly didn't normally send requests.
815  */
816 static void
817 in_arpinput(struct mbuf *m)
818 {
819 	struct arphdr *ah;
820 	struct ifnet *ifp = m->m_pkthdr.rcvif;
821 	struct llinfo_arp *la = 0;
822 	struct rtentry  *rt;
823 	struct in_ifaddr *ia;
824 #if NBRIDGE > 0
825 	struct in_ifaddr *bridge_ia = NULL;
826 #endif
827 	struct sockaddr_dl *sdl;
828 	struct sockaddr sa;
829 	struct in_addr isaddr, itaddr, myaddr;
830 	int op;
831 	struct mbuf *mold;
832 	int s;
833 
834 	ah = mtod(m, struct arphdr *);
835 	op = ntohs(ah->ar_op);
836 
837 	/*
838 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
839 	 * ar_tpa().
840 	 */
841 	switch (ifp->if_type) {
842 	case IFT_IEEE1394:
843 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
844 			;
845 		else {
846 			/* XXX this is to make sure we compute ar_tha right */
847 			/* XXX check ar_hrd more strictly? */
848 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
849 		}
850 		break;
851 	default:
852 		/* XXX check ar_hrd? */
853 		break;
854 	}
855 
856 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
857 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
858 
859 	if (m->m_flags & (M_BCAST|M_MCAST))
860 		arpstat.as_rcvmcast++;
861 
862 	/*
863 	 * If the target IP address is zero, ignore the packet.
864 	 * This prevents the code below from tring to answer
865 	 * when we are using IP address zero (booting).
866 	 */
867 	if (in_nullhost(itaddr)) {
868 		arpstat.as_rcvzerotpa++;
869 		goto out;
870 	}
871 
872 	/*
873 	 * If the source IP address is zero, this is most likely a
874 	 * confused host trying to use IP address zero. (Windoze?)
875 	 * XXX: Should we bother trying to reply to these?
876 	 */
877 	if (in_nullhost(isaddr)) {
878 		arpstat.as_rcvzerospa++;
879 		goto out;
880 	}
881 
882 	/*
883 	 * Search for a matching interface address
884 	 * or any address on the interface to use
885 	 * as a dummy address in the rest of this function
886 	 */
887 	INADDR_TO_IA(itaddr, ia);
888 	while (ia != NULL) {
889 		if (ia->ia_ifp == m->m_pkthdr.rcvif)
890 			break;
891 
892 #if NBRIDGE > 0
893 		/*
894 		 * If the interface we received the packet on
895 		 * is part of a bridge, check to see if we need
896 		 * to "bridge" the packet to ourselves at this
897 		 * layer.  Note we still prefer a perfect match,
898 		 * but allow this weaker match if necessary.
899 		 */
900 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
901 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
902 			bridge_ia = ia;
903 #endif /* NBRIDGE > 0 */
904 
905 		NEXT_IA_WITH_SAME_ADDR(ia);
906 	}
907 
908 #if NBRIDGE > 0
909 	if (ia == NULL && bridge_ia != NULL) {
910 		ia = bridge_ia;
911 		ifp = bridge_ia->ia_ifp;
912 	}
913 #endif
914 
915 	if (ia == NULL) {
916 		INADDR_TO_IA(isaddr, ia);
917 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
918 			NEXT_IA_WITH_SAME_ADDR(ia);
919 
920 		if (ia == NULL) {
921 			IFP_TO_IA(ifp, ia);
922 			if (ia == NULL) {
923 				arpstat.as_rcvnoint++;
924 				goto out;
925 			}
926 		}
927 	}
928 
929 	myaddr = ia->ia_addr.sin_addr;
930 
931 	/* XXX checks for bridge case? */
932 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
933 	    ifp->if_addrlen)) {
934 		arpstat.as_rcvlocalsha++;
935 		goto out;	/* it's from me, ignore it. */
936 	}
937 
938 	/* XXX checks for bridge case? */
939 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
940 		arpstat.as_rcvbcastsha++;
941 		log(LOG_ERR,
942 		    "%s: arp: link address is broadcast for IP address %s!\n",
943 		    ifp->if_xname, in_fmtaddr(isaddr));
944 		goto out;
945 	}
946 
947 	if (in_hosteq(isaddr, myaddr)) {
948 		arpstat.as_rcvlocalspa++;
949 		log(LOG_ERR,
950 		   "duplicate IP address %s sent from link address %s\n",
951 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
952 		itaddr = myaddr;
953 		goto reply;
954 	}
955 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
956 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
957 		if (sdl->sdl_alen &&
958 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
959 			if (rt->rt_flags & RTF_STATIC) {
960 				arpstat.as_rcvoverperm++;
961 				log(LOG_INFO,
962 				    "%s tried to overwrite permanent arp info"
963 				    " for %s\n",
964 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
965 				    in_fmtaddr(isaddr));
966 				goto out;
967 			} else if (rt->rt_ifp != ifp) {
968 				arpstat.as_rcvoverint++;
969 				log(LOG_INFO,
970 				    "%s on %s tried to overwrite "
971 				    "arp info for %s on %s\n",
972 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
973 				    ifp->if_xname, in_fmtaddr(isaddr),
974 				    rt->rt_ifp->if_xname);
975 				    goto out;
976 			} else {
977 				arpstat.as_rcvover++;
978 				log(LOG_INFO,
979 				    "arp info overwritten for %s by %s\n",
980 				    in_fmtaddr(isaddr),
981 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
982 			}
983 		}
984 		/*
985 		 * sanity check for the address length.
986 		 * XXX this does not work for protocols with variable address
987 		 * length. -is
988 		 */
989 		if (sdl->sdl_alen &&
990 		    sdl->sdl_alen != ah->ar_hln) {
991 			arpstat.as_rcvlenchg++;
992 			log(LOG_WARNING,
993 			    "arp from %s: new addr len %d, was %d",
994 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
995 		}
996 		if (ifp->if_addrlen != ah->ar_hln) {
997 			arpstat.as_rcvbadlen++;
998 			log(LOG_WARNING,
999 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1000 			    in_fmtaddr(isaddr), ah->ar_hln,
1001 			    ifp->if_addrlen);
1002 			goto reply;
1003 		}
1004 #if NTOKEN > 0
1005 		/*
1006 		 * XXX uses m_data and assumes the complete answer including
1007 		 * XXX token-ring headers is in the same buf
1008 		 */
1009 		if (ifp->if_type == IFT_ISO88025) {
1010 			struct token_header *trh;
1011 
1012 			trh = (struct token_header *)M_TRHSTART(m);
1013 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1014 				struct token_rif	*rif;
1015 				size_t	riflen;
1016 
1017 				rif = TOKEN_RIF(trh);
1018 				riflen = (ntohs(rif->tr_rcf) &
1019 				    TOKEN_RCF_LEN_MASK) >> 8;
1020 
1021 				if (riflen > 2 &&
1022 				    riflen < sizeof(struct token_rif) &&
1023 				    (riflen & 1) == 0) {
1024 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1025 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1026 					bcopy(rif, TOKEN_RIF(la), riflen);
1027 				}
1028 			}
1029 		}
1030 #endif /* NTOKEN > 0 */
1031 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
1032 		    sdl->sdl_alen = ah->ar_hln);
1033 		if (rt->rt_expire)
1034 			rt->rt_expire = time.tv_sec + arpt_keep;
1035 		rt->rt_flags &= ~RTF_REJECT;
1036 		la->la_asked = 0;
1037 
1038 		s = splnet();
1039 		mold = la->la_hold;
1040 		la->la_hold = 0;
1041 		splx(s);
1042 
1043 		if (mold) {
1044 			arpstat.as_dfrsent++;
1045 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
1046 		}
1047 	}
1048 reply:
1049 	if (op != ARPOP_REQUEST) {
1050 		if (op == ARPOP_REPLY)
1051 			arpstat.as_rcvreply++;
1052 	out:
1053 		m_freem(m);
1054 		return;
1055 	}
1056 	arpstat.as_rcvrequest++;
1057 	if (in_hosteq(itaddr, myaddr)) {
1058 		/* I am the target */
1059 		if (ar_tha(ah))
1060 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1061 			    ah->ar_hln);
1062 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1063 	} else {
1064 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1065 		if (la == 0)
1066 			goto out;
1067 		rt = la->la_rt;
1068 		if (ar_tha(ah))
1069 			bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah),
1070 			    ah->ar_hln);
1071 		sdl = SDL(rt->rt_gateway);
1072 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
1073 	}
1074 
1075 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
1076 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
1077 	ah->ar_op = htons(ARPOP_REPLY);
1078 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1079 	switch (ifp->if_type) {
1080 	case IFT_IEEE1394:
1081 		/*
1082 		 * ieee1394 arp reply is broadcast
1083 		 */
1084 		m->m_flags &= ~M_MCAST;
1085 		m->m_flags |= M_BCAST;
1086 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1087 		break;
1088 
1089 	default:
1090 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1091 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1092 		break;
1093 	}
1094 	m->m_pkthdr.len = m->m_len;
1095 	sa.sa_family = AF_ARP;
1096 	sa.sa_len = 2;
1097 	arpstat.as_sndtotal++;
1098 	arpstat.as_sndreply++;
1099 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1100 	return;
1101 }
1102 
1103 /*
1104  * Free an arp entry.
1105  */
1106 static void arptfree(struct llinfo_arp *la)
1107 {
1108 	struct rtentry *rt = la->la_rt;
1109 	struct sockaddr_dl *sdl;
1110 
1111 	ARP_LOCK_CHECK();
1112 
1113 	if (rt == 0)
1114 		panic("arptfree");
1115 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
1116 	    sdl->sdl_family == AF_LINK) {
1117 		sdl->sdl_alen = 0;
1118 		la->la_asked = 0;
1119 		rt->rt_flags &= ~RTF_REJECT;
1120 		return;
1121 	}
1122 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
1123 	    0, (struct rtentry **)0);
1124 }
1125 
1126 /*
1127  * Lookup or enter a new address in arptab.
1128  */
1129 static struct llinfo_arp *
1130 arplookup(struct mbuf *m, struct in_addr *addr, int create, int proxy)
1131 {
1132 	struct arphdr *ah;
1133 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1134 	struct rtentry *rt;
1135 	static struct sockaddr_inarp sin;
1136 	const char *why = 0;
1137 
1138 	ah = mtod(m, struct arphdr *);
1139 	sin.sin_len = sizeof(sin);
1140 	sin.sin_family = AF_INET;
1141 	sin.sin_addr = *addr;
1142 	sin.sin_other = proxy ? SIN_PROXY : 0;
1143 	rt = rtalloc1(sintosa(&sin), create);
1144 	if (rt == 0)
1145 		return (0);
1146 	rt->rt_refcnt--;
1147 
1148 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO &&
1149 	    rt->rt_gateway->sa_family == AF_LINK)
1150 		return ((struct llinfo_arp *)rt->rt_llinfo);
1151 
1152 
1153 
1154 	if (create) {
1155 		if (rt->rt_flags & RTF_GATEWAY)
1156 			why = "host is not on local network";
1157 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1158 			arpstat.as_allocfail++;
1159 			why = "could not allocate llinfo";
1160 		} else
1161 			why = "gateway route is not ours";
1162 		log(LOG_DEBUG, "arplookup: unable to enter address"
1163 		    " for %s@%s on %s (%s)\n",
1164 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1165 		    (ifp) ? ifp->if_xname : 0, why);
1166 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1167 			rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
1168 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1169 		}
1170 	}
1171 	return (0);
1172 }
1173 
1174 int
1175 arpioctl(u_long cmd, caddr_t data)
1176 {
1177 
1178 	return (EOPNOTSUPP);
1179 }
1180 
1181 void
1182 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1183 {
1184 	struct in_addr *ip;
1185 
1186 	/*
1187 	 * Warn the user if another station has this IP address,
1188 	 * but only if the interface IP address is not zero.
1189 	 */
1190 	ip = &IA_SIN(ifa)->sin_addr;
1191 	if (!in_nullhost(*ip))
1192 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
1193 
1194 	ifa->ifa_rtrequest = arp_rtrequest;
1195 	ifa->ifa_flags |= RTF_CLONING;
1196 }
1197 
1198 /*
1199  * Called from 10 Mb/s Ethernet interrupt handlers
1200  * when ether packet type ETHERTYPE_REVARP
1201  * is received.  Common length and type checks are done here,
1202  * then the protocol-specific routine is called.
1203  */
1204 void
1205 revarpinput(struct mbuf *m)
1206 {
1207 	struct arphdr *ar;
1208 
1209 	if (m->m_len < sizeof(struct arphdr))
1210 		goto out;
1211 	ar = mtod(m, struct arphdr *);
1212 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1213 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1214 		goto out;
1215 #endif
1216 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1217 		goto out;
1218 	switch (ntohs(ar->ar_pro)) {
1219 	case ETHERTYPE_IP:
1220 	case ETHERTYPE_IPTRAILERS:
1221 		in_revarpinput(m);
1222 		return;
1223 
1224 	default:
1225 		break;
1226 	}
1227 out:
1228 	m_freem(m);
1229 }
1230 
1231 /*
1232  * RARP for Internet protocols on 10 Mb/s Ethernet.
1233  * Algorithm is that given in RFC 903.
1234  * We are only using for bootstrap purposes to get an ip address for one of
1235  * our interfaces.  Thus we support no user-interface.
1236  *
1237  * Since the contents of the RARP reply are specific to the interface that
1238  * sent the request, this code must ensure that they are properly associated.
1239  *
1240  * Note: also supports ARP via RARP packets, per the RFC.
1241  */
1242 void
1243 in_revarpinput(struct mbuf *m)
1244 {
1245 	struct ifnet *ifp;
1246 	struct arphdr *ah;
1247 	int op;
1248 
1249 	ah = mtod(m, struct arphdr *);
1250 	op = ntohs(ah->ar_op);
1251 
1252 	switch (m->m_pkthdr.rcvif->if_type) {
1253 	case IFT_IEEE1394:
1254 		/* ARP without target hardware address is not supported */
1255 		goto out;
1256 	default:
1257 		break;
1258 	}
1259 
1260 	switch (op) {
1261 	case ARPOP_REQUEST:
1262 	case ARPOP_REPLY:	/* per RFC */
1263 		in_arpinput(m);
1264 		return;
1265 	case ARPOP_REVREPLY:
1266 		break;
1267 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1268 	default:
1269 		goto out;
1270 	}
1271 	if (!revarp_in_progress)
1272 		goto out;
1273 	ifp = m->m_pkthdr.rcvif;
1274 	if (ifp != myip_ifp) /* !same interface */
1275 		goto out;
1276 	if (myip_initialized)
1277 		goto wake;
1278 	if (bcmp(ar_tha(ah), LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1279 		goto out;
1280 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1281 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1282 	myip_initialized = 1;
1283 wake:	/* Do wakeup every time in case it was missed. */
1284 	wakeup((caddr_t)&myip);
1285 
1286 out:
1287 	m_freem(m);
1288 }
1289 
1290 /*
1291  * Send a RARP request for the ip address of the specified interface.
1292  * The request should be RFC 903-compliant.
1293  */
1294 void
1295 revarprequest(struct ifnet *ifp)
1296 {
1297 	struct sockaddr sa;
1298 	struct mbuf *m;
1299 	struct arphdr *ah;
1300 
1301 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1302 		return;
1303 	MCLAIM(m, &arpdomain.dom_mowner);
1304 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1305 	    2*ifp->if_addrlen;
1306 	m->m_pkthdr.len = m->m_len;
1307 	MH_ALIGN(m, m->m_len);
1308 	ah = mtod(m, struct arphdr *);
1309 	bzero((caddr_t)ah, m->m_len);
1310 	ah->ar_pro = htons(ETHERTYPE_IP);
1311 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1312 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1313 	ah->ar_op = htons(ARPOP_REVREQUEST);
1314 
1315 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1316 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_tha(ah), ah->ar_hln);
1317 
1318 	sa.sa_family = AF_ARP;
1319 	sa.sa_len = 2;
1320 	m->m_flags |= M_BCAST;
1321 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1322 
1323 }
1324 
1325 /*
1326  * RARP for the ip address of the specified interface, but also
1327  * save the ip address of the server that sent the answer.
1328  * Timeout if no response is received.
1329  */
1330 int
1331 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1332     struct in_addr *clnt_in)
1333 {
1334 	int result, count = 20;
1335 
1336 	myip_initialized = 0;
1337 	myip_ifp = ifp;
1338 
1339 	revarp_in_progress = 1;
1340 	while (count--) {
1341 		revarprequest(ifp);
1342 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1343 		if (result != EWOULDBLOCK)
1344 			break;
1345 	}
1346 	revarp_in_progress = 0;
1347 
1348 	if (!myip_initialized)
1349 		return ENETUNREACH;
1350 
1351 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1352 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1353 	return 0;
1354 }
1355 
1356 
1357 
1358 #ifdef DDB
1359 
1360 #include <machine/db_machdep.h>
1361 #include <ddb/db_interface.h>
1362 #include <ddb/db_output.h>
1363 
1364 static void
1365 db_print_sa(const struct sockaddr *sa)
1366 {
1367 	int len;
1368 	const u_char *p;
1369 
1370 	if (sa == 0) {
1371 		db_printf("[NULL]");
1372 		return;
1373 	}
1374 
1375 	p = (const u_char *)sa;
1376 	len = sa->sa_len;
1377 	db_printf("[");
1378 	while (len > 0) {
1379 		db_printf("%d", *p);
1380 		p++; len--;
1381 		if (len) db_printf(",");
1382 	}
1383 	db_printf("]\n");
1384 }
1385 
1386 static void
1387 db_print_ifa(struct ifaddr *ifa)
1388 {
1389 	if (ifa == 0)
1390 		return;
1391 	db_printf("  ifa_addr=");
1392 	db_print_sa(ifa->ifa_addr);
1393 	db_printf("  ifa_dsta=");
1394 	db_print_sa(ifa->ifa_dstaddr);
1395 	db_printf("  ifa_mask=");
1396 	db_print_sa(ifa->ifa_netmask);
1397 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1398 			  ifa->ifa_flags,
1399 			  ifa->ifa_refcnt,
1400 			  ifa->ifa_metric);
1401 }
1402 
1403 static void
1404 db_print_llinfo(caddr_t li)
1405 {
1406 	struct llinfo_arp *la;
1407 
1408 	if (li == 0)
1409 		return;
1410 	la = (struct llinfo_arp *)li;
1411 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1412 			  la->la_rt, la->la_hold, la->la_asked);
1413 }
1414 
1415 /*
1416  * Function to pass to rn_walktree().
1417  * Return non-zero error to abort walk.
1418  */
1419 static int
1420 db_show_radix_node(struct radix_node *rn, void *w)
1421 {
1422 	struct rtentry *rt = (struct rtentry *)rn;
1423 
1424 	db_printf("rtentry=%p", rt);
1425 
1426 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1427 			  rt->rt_flags, rt->rt_refcnt,
1428 			  rt->rt_use, rt->rt_expire);
1429 
1430 	db_printf(" key="); db_print_sa(rt_key(rt));
1431 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1432 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1433 
1434 	db_printf(" ifp=%p ", rt->rt_ifp);
1435 	if (rt->rt_ifp)
1436 		db_printf("(%s)", rt->rt_ifp->if_xname);
1437 	else
1438 		db_printf("(NULL)");
1439 
1440 	db_printf(" ifa=%p\n", rt->rt_ifa);
1441 	db_print_ifa(rt->rt_ifa);
1442 
1443 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1444 
1445 	db_printf(" gwroute=%p llinfo=%p\n",
1446 			  rt->rt_gwroute, rt->rt_llinfo);
1447 	db_print_llinfo(rt->rt_llinfo);
1448 
1449 	return (0);
1450 }
1451 
1452 /*
1453  * Function to print all the route trees.
1454  * Use this from ddb:  "show arptab"
1455  */
1456 void
1457 db_show_arptab(db_expr_t addr, int have_addr, db_expr_t count, const char *modif)
1458 {
1459 	struct radix_node_head *rnh;
1460 	rnh = rt_tables[AF_INET];
1461 	db_printf("Route tree for AF_INET\n");
1462 	if (rnh == NULL) {
1463 		db_printf(" (not initialized)\n");
1464 		return;
1465 	}
1466 	rn_walktree(rnh, db_show_radix_node, NULL);
1467 	return;
1468 }
1469 #endif
1470 
1471 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1472 {
1473 	const struct sysctlnode *node;
1474 
1475 	sysctl_createv(clog, 0, NULL, NULL,
1476 			CTLFLAG_PERMANENT,
1477 			CTLTYPE_NODE, "net", NULL,
1478 			NULL, 0, NULL, 0,
1479 			CTL_NET, CTL_EOL);
1480 	sysctl_createv(clog, 0, NULL, NULL,
1481 			CTLFLAG_PERMANENT,
1482 			CTLTYPE_NODE, "inet", NULL,
1483 			NULL, 0, NULL, 0,
1484 			CTL_NET, PF_INET, CTL_EOL);
1485 	sysctl_createv(clog, 0, NULL, &node,
1486 			CTLFLAG_PERMANENT,
1487 			CTLTYPE_NODE, "arp",
1488 			SYSCTL_DESCR("Address Resolution Protocol"),
1489 			NULL, 0, NULL, 0,
1490 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1491 
1492 	sysctl_createv(clog, 0, NULL, NULL,
1493 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1494 			CTLTYPE_INT, "prune",
1495 			SYSCTL_DESCR("ARP cache pruning interval"),
1496 			NULL, 0, &arpt_prune, 0,
1497 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1498 
1499 	sysctl_createv(clog, 0, NULL, NULL,
1500 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1501 			CTLTYPE_INT, "keep",
1502 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1503 			NULL, 0, &arpt_keep, 0,
1504 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1505 
1506 	sysctl_createv(clog, 0, NULL, NULL,
1507 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1508 			CTLTYPE_INT, "down",
1509 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1510 			NULL, 0, &arpt_down, 0,
1511 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1512 
1513 	sysctl_createv(clog, 0, NULL, NULL,
1514 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1515 			CTLTYPE_INT, "refresh",
1516 			SYSCTL_DESCR("ARP entry refresh interval"),
1517 			NULL, 0, &arpt_refresh, 0,
1518 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1519 }
1520 
1521 #endif /* INET */
1522