xref: /netbsd-src/sys/netinet/if_arp.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: if_arp.c,v 1.117 2006/11/24 19:37:03 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 1.117 2006/11/24 19:37:03 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)
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)
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 accidentally 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 				rt_replace_ifa(rt, ifa);
592 		}
593 		break;
594 
595 	case RTM_DELETE:
596 		if (la == 0)
597 			break;
598 		arp_inuse--;
599 		LIST_REMOVE(la, la_list);
600 		rt->rt_llinfo = 0;
601 		rt->rt_flags &= ~RTF_LLINFO;
602 
603 		s = splnet();
604 		mold = la->la_hold;
605 		la->la_hold = 0;
606 		splx(s);
607 
608 		if (mold)
609 			m_freem(mold);
610 
611 		Free((caddr_t)la);
612 	}
613 	ARP_UNLOCK();
614 }
615 
616 /*
617  * Broadcast an ARP request. Caller specifies:
618  *	- arp header source ip address
619  *	- arp header target ip address
620  *	- arp header source ethernet address
621  */
622 void
623 arprequest(struct ifnet *ifp,
624     struct in_addr *sip, struct in_addr *tip, u_int8_t *enaddr)
625 {
626 	struct mbuf *m;
627 	struct arphdr *ah;
628 	struct sockaddr sa;
629 
630 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
631 		return;
632 	MCLAIM(m, &arpdomain.dom_mowner);
633 	switch (ifp->if_type) {
634 	case IFT_IEEE1394:
635 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
636 		    ifp->if_addrlen;
637 		break;
638 	default:
639 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
640 		    2 * ifp->if_addrlen;
641 		break;
642 	}
643 	m->m_pkthdr.len = m->m_len;
644 	MH_ALIGN(m, m->m_len);
645 	ah = mtod(m, struct arphdr *);
646 	bzero((caddr_t)ah, m->m_len);
647 	switch (ifp->if_type) {
648 	case IFT_IEEE1394:	/* RFC2734 */
649 		/* fill it now for ar_tpa computation */
650 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
651 		break;
652 	default:
653 		/* ifp->if_output will fill ar_hrd */
654 		break;
655 	}
656 	ah->ar_pro = htons(ETHERTYPE_IP);
657 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
658 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
659 	ah->ar_op = htons(ARPOP_REQUEST);
660 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
661 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
662 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
663 	sa.sa_family = AF_ARP;
664 	sa.sa_len = 2;
665 	m->m_flags |= M_BCAST;
666 	arpstat.as_sndtotal++;
667 	arpstat.as_sndrequest++;
668 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
669 }
670 
671 /*
672  * Resolve an IP address into an ethernet address.  If success,
673  * desten is filled in.  If there is no entry in arptab,
674  * set one up and broadcast a request for the IP address.
675  * Hold onto this mbuf and resend it once the address
676  * is finally resolved.  A return value of 1 indicates
677  * that desten has been filled in and the packet should be sent
678  * normally; a 0 return indicates that the packet has been
679  * taken over here, either now or for later transmission.
680  */
681 int
682 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
683     struct sockaddr *dst, u_char *desten)
684 {
685 	struct llinfo_arp *la;
686 	struct sockaddr_dl *sdl;
687 	struct mbuf *mold;
688 	int s;
689 
690 	if (rt)
691 		la = (struct llinfo_arp *)rt->rt_llinfo;
692 	else {
693 		if ((la = arplookup(m, &SIN(dst)->sin_addr, 1, 0)) != NULL)
694 			rt = la->la_rt;
695 	}
696 	if (la == 0 || rt == 0) {
697 		arpstat.as_allocfail++;
698 		log(LOG_DEBUG,
699 		    "arpresolve: can't allocate llinfo on %s for %s\n",
700 		    ifp->if_xname, in_fmtaddr(SIN(dst)->sin_addr));
701 		m_freem(m);
702 		return (0);
703 	}
704 	sdl = SDL(rt->rt_gateway);
705 	/*
706 	 * Check the address family and length is valid, the address
707 	 * is resolved; otherwise, try to resolve.
708 	 */
709 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
710 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
711 		bcopy(LLADDR(sdl), desten,
712 		    min(sdl->sdl_alen, ifp->if_addrlen));
713 		rt->rt_pksent = time_second; /* Time for last pkt sent */
714 		return 1;
715 	}
716 	/*
717 	 * There is an arptab entry, but no ethernet address
718 	 * response yet.  Replace the held mbuf with this
719 	 * latest one.
720 	 */
721 
722 	arpstat.as_dfrtotal++;
723 	s = splnet();
724 	mold = la->la_hold;
725 	la->la_hold = m;
726 	splx(s);
727 
728 	if (mold) {
729 		arpstat.as_dfrdropped++;
730 		m_freem(mold);
731 	}
732 
733 	/*
734 	 * Re-send the ARP request when appropriate.
735 	 */
736 #ifdef	DIAGNOSTIC
737 	if (rt->rt_expire == 0) {
738 		/* This should never happen. (Should it? -gwr) */
739 		printf("arpresolve: unresolved and rt_expire == 0\n");
740 		/* Set expiration time to now (expired). */
741 		rt->rt_expire = time_second;
742 	}
743 #endif
744 	if (rt->rt_expire) {
745 		rt->rt_flags &= ~RTF_REJECT;
746 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
747 			rt->rt_expire = time_second;
748 			if (la->la_asked++ < arp_maxtries)
749 				arprequest(ifp,
750 				    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
751 				    &SIN(dst)->sin_addr,
752 #if NCARP > 0
753 				    (rt->rt_ifp->if_type == IFT_CARP) ?
754 				    LLADDR(rt->rt_ifp->if_sadl):
755 #endif
756 				    LLADDR(ifp->if_sadl));
757 			else {
758 				rt->rt_flags |= RTF_REJECT;
759 				rt->rt_expire += arpt_down;
760 				la->la_asked = 0;
761 			}
762 		}
763 	}
764 	return (0);
765 }
766 
767 /*
768  * Common length and type checks are done here,
769  * then the protocol-specific routine is called.
770  */
771 void
772 arpintr(void)
773 {
774 	struct mbuf *m;
775 	struct arphdr *ar;
776 	int s;
777 	int arplen;
778 
779 	while (arpintrq.ifq_head) {
780 		s = splnet();
781 		IF_DEQUEUE(&arpintrq, m);
782 		splx(s);
783 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
784 			panic("arpintr");
785 
786 		MCLAIM(m, &arpdomain.dom_mowner);
787 		arpstat.as_rcvtotal++;
788 
789 		/*
790 		 * First, make sure we have at least struct arphdr.
791 		 */
792 		if (m->m_len < sizeof(struct arphdr) ||
793 		    (ar = mtod(m, struct arphdr *)) == NULL)
794 			goto badlen;
795 
796 		switch (m->m_pkthdr.rcvif->if_type) {
797 		case IFT_IEEE1394:
798 			arplen = sizeof(struct arphdr) +
799 			    ar->ar_hln + 2 * ar->ar_pln;
800 			break;
801 		default:
802 			arplen = sizeof(struct arphdr) +
803 			    2 * ar->ar_hln + 2 * ar->ar_pln;
804 			break;
805 		}
806 
807 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
808 		    m->m_len >= arplen)
809 			switch (ntohs(ar->ar_pro)) {
810 			case ETHERTYPE_IP:
811 			case ETHERTYPE_IPTRAILERS:
812 				in_arpinput(m);
813 				continue;
814 			default:
815 				arpstat.as_rcvbadproto++;
816 			}
817 		else {
818 badlen:
819 			arpstat.as_rcvbadlen++;
820 		}
821 		m_freem(m);
822 	}
823 }
824 
825 /*
826  * ARP for Internet protocols on 10 Mb/s Ethernet.
827  * Algorithm is that given in RFC 826.
828  * In addition, a sanity check is performed on the sender
829  * protocol address, to catch impersonators.
830  * We no longer handle negotiations for use of trailer protocol:
831  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
832  * along with IP replies if we wanted trailers sent to us,
833  * and also sent them in response to IP replies.
834  * This allowed either end to announce the desire to receive
835  * trailer packets.
836  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
837  * but formerly didn't normally send requests.
838  */
839 static void
840 in_arpinput(struct mbuf *m)
841 {
842 	struct arphdr *ah;
843 	struct ifnet *ifp = m->m_pkthdr.rcvif;
844 	struct llinfo_arp *la = 0;
845 	struct rtentry  *rt;
846 	struct in_ifaddr *ia;
847 #if NBRIDGE > 0
848 	struct in_ifaddr *bridge_ia = NULL;
849 #endif
850 #if NCARP > 0
851 	u_int32_t count = 0, index = 0;
852 #endif
853 	struct sockaddr_dl *sdl;
854 	struct sockaddr sa;
855 	struct in_addr isaddr, itaddr, myaddr;
856 	int op;
857 	struct mbuf *mold;
858 	caddr_t tha;
859 	int s;
860 
861 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
862 		goto out;
863 	ah = mtod(m, struct arphdr *);
864 	op = ntohs(ah->ar_op);
865 
866 	/*
867 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
868 	 * ar_tpa().
869 	 */
870 	switch (ifp->if_type) {
871 	case IFT_IEEE1394:
872 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
873 			;
874 		else {
875 			/* XXX this is to make sure we compute ar_tha right */
876 			/* XXX check ar_hrd more strictly? */
877 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
878 		}
879 		break;
880 	default:
881 		/* XXX check ar_hrd? */
882 		break;
883 	}
884 
885 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
886 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
887 
888 	if (m->m_flags & (M_BCAST|M_MCAST))
889 		arpstat.as_rcvmcast++;
890 
891 	/*
892 	 * If the target IP address is zero, ignore the packet.
893 	 * This prevents the code below from tring to answer
894 	 * when we are using IP address zero (booting).
895 	 */
896 	if (in_nullhost(itaddr)) {
897 		arpstat.as_rcvzerotpa++;
898 		goto out;
899 	}
900 
901 	/*
902 	 * If the source IP address is zero, this is most likely a
903 	 * confused host trying to use IP address zero. (Windoze?)
904 	 * XXX: Should we bother trying to reply to these?
905 	 */
906 	if (in_nullhost(isaddr)) {
907 		arpstat.as_rcvzerospa++;
908 		goto out;
909 	}
910 
911 	/*
912 	 * Search for a matching interface address
913 	 * or any address on the interface to use
914 	 * as a dummy address in the rest of this function
915 	 */
916 
917 	INADDR_TO_IA(itaddr, ia);
918 	while (ia != NULL) {
919 #if NCARP > 0
920 		if (ia->ia_ifp->if_type == IFT_CARP &&
921 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
922 		    (IFF_UP|IFF_RUNNING))) {
923 			index++;
924 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
925 			    carp_iamatch(ia, ar_sha(ah),
926 			    &count, index)) {
927 				break;
928 				}
929 		} else
930 #endif
931 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
932 				break;
933 #if NBRIDGE > 0
934 		/*
935 		 * If the interface we received the packet on
936 		 * is part of a bridge, check to see if we need
937 		 * to "bridge" the packet to ourselves at this
938 		 * layer.  Note we still prefer a perfect match,
939 		 * but allow this weaker match if necessary.
940 		 */
941 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
942 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
943 			bridge_ia = ia;
944 #endif /* NBRIDGE > 0 */
945 
946 		NEXT_IA_WITH_SAME_ADDR(ia);
947 	}
948 
949 #if NBRIDGE > 0
950 	if (ia == NULL && bridge_ia != NULL) {
951 		ia = bridge_ia;
952 		ifp = bridge_ia->ia_ifp;
953 	}
954 #endif
955 
956 	if (ia == NULL) {
957 		INADDR_TO_IA(isaddr, ia);
958 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
959 			NEXT_IA_WITH_SAME_ADDR(ia);
960 
961 		if (ia == NULL) {
962 			IFP_TO_IA(ifp, ia);
963 			if (ia == NULL) {
964 				arpstat.as_rcvnoint++;
965 				goto out;
966 			}
967 		}
968 	}
969 
970 	myaddr = ia->ia_addr.sin_addr;
971 
972 	/* XXX checks for bridge case? */
973 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
974 	    ifp->if_addrlen)) {
975 		arpstat.as_rcvlocalsha++;
976 		goto out;	/* it's from me, ignore it. */
977 	}
978 
979 	/* XXX checks for bridge case? */
980 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
981 		arpstat.as_rcvbcastsha++;
982 		log(LOG_ERR,
983 		    "%s: arp: link address is broadcast for IP address %s!\n",
984 		    ifp->if_xname, in_fmtaddr(isaddr));
985 		goto out;
986 	}
987 
988 	if (in_hosteq(isaddr, myaddr)) {
989 		arpstat.as_rcvlocalspa++;
990 		log(LOG_ERR,
991 		   "duplicate IP address %s sent from link address %s\n",
992 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
993 		itaddr = myaddr;
994 		goto reply;
995 	}
996 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
997 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
998 		if (sdl->sdl_alen &&
999 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
1000 			if (rt->rt_flags & RTF_STATIC) {
1001 				arpstat.as_rcvoverperm++;
1002 				log(LOG_INFO,
1003 				    "%s tried to overwrite permanent arp info"
1004 				    " for %s\n",
1005 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1006 				    in_fmtaddr(isaddr));
1007 				goto out;
1008 			} else if (rt->rt_ifp != ifp) {
1009 				arpstat.as_rcvoverint++;
1010 				log(LOG_INFO,
1011 				    "%s on %s tried to overwrite "
1012 				    "arp info for %s on %s\n",
1013 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1014 				    ifp->if_xname, in_fmtaddr(isaddr),
1015 				    rt->rt_ifp->if_xname);
1016 				    goto out;
1017 			} else {
1018 				arpstat.as_rcvover++;
1019 				log(LOG_INFO,
1020 				    "arp info overwritten for %s by %s\n",
1021 				    in_fmtaddr(isaddr),
1022 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1023 			}
1024 		}
1025 		/*
1026 		 * sanity check for the address length.
1027 		 * XXX this does not work for protocols with variable address
1028 		 * length. -is
1029 		 */
1030 		if (sdl->sdl_alen &&
1031 		    sdl->sdl_alen != ah->ar_hln) {
1032 			arpstat.as_rcvlenchg++;
1033 			log(LOG_WARNING,
1034 			    "arp from %s: new addr len %d, was %d",
1035 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1036 		}
1037 		if (ifp->if_addrlen != ah->ar_hln) {
1038 			arpstat.as_rcvbadlen++;
1039 			log(LOG_WARNING,
1040 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1041 			    in_fmtaddr(isaddr), ah->ar_hln,
1042 			    ifp->if_addrlen);
1043 			goto reply;
1044 		}
1045 #if NTOKEN > 0
1046 		/*
1047 		 * XXX uses m_data and assumes the complete answer including
1048 		 * XXX token-ring headers is in the same buf
1049 		 */
1050 		if (ifp->if_type == IFT_ISO88025) {
1051 			struct token_header *trh;
1052 
1053 			trh = (struct token_header *)M_TRHSTART(m);
1054 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1055 				struct token_rif	*rif;
1056 				size_t	riflen;
1057 
1058 				rif = TOKEN_RIF(trh);
1059 				riflen = (ntohs(rif->tr_rcf) &
1060 				    TOKEN_RCF_LEN_MASK) >> 8;
1061 
1062 				if (riflen > 2 &&
1063 				    riflen < sizeof(struct token_rif) &&
1064 				    (riflen & 1) == 0) {
1065 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1066 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1067 					bcopy(rif, TOKEN_RIF(la), riflen);
1068 				}
1069 			}
1070 		}
1071 #endif /* NTOKEN > 0 */
1072 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
1073 		    sdl->sdl_alen = ah->ar_hln);
1074 		if (rt->rt_expire)
1075 			rt->rt_expire = time_second + arpt_keep;
1076 		rt->rt_flags &= ~RTF_REJECT;
1077 		la->la_asked = 0;
1078 
1079 		s = splnet();
1080 		mold = la->la_hold;
1081 		la->la_hold = 0;
1082 		splx(s);
1083 
1084 		if (mold) {
1085 			arpstat.as_dfrsent++;
1086 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
1087 		}
1088 	}
1089 reply:
1090 	if (op != ARPOP_REQUEST) {
1091 		if (op == ARPOP_REPLY)
1092 			arpstat.as_rcvreply++;
1093 	out:
1094 		m_freem(m);
1095 		return;
1096 	}
1097 	arpstat.as_rcvrequest++;
1098 	if (in_hosteq(itaddr, myaddr)) {
1099 		/* I am the target */
1100 		tha = ar_tha(ah);
1101 		if (tha)
1102 			bcopy((caddr_t)ar_sha(ah), tha, ah->ar_hln);
1103 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1104 	} else {
1105 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1106 		if (la == 0)
1107 			goto out;
1108 		rt = la->la_rt;
1109 		if (rt->rt_ifp->if_type == IFT_CARP &&
1110 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1111 			goto out;
1112 		tha = ar_tha(ah);
1113 		if (tha)
1114 			bcopy((caddr_t)ar_sha(ah), tha, ah->ar_hln);
1115 		sdl = SDL(rt->rt_gateway);
1116 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
1117 	}
1118 
1119 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
1120 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
1121 	ah->ar_op = htons(ARPOP_REPLY);
1122 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1123 	switch (ifp->if_type) {
1124 	case IFT_IEEE1394:
1125 		/*
1126 		 * ieee1394 arp reply is broadcast
1127 		 */
1128 		m->m_flags &= ~M_MCAST;
1129 		m->m_flags |= M_BCAST;
1130 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1131 		break;
1132 
1133 	default:
1134 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1135 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1136 		break;
1137 	}
1138 	m->m_pkthdr.len = m->m_len;
1139 	sa.sa_family = AF_ARP;
1140 	sa.sa_len = 2;
1141 	arpstat.as_sndtotal++;
1142 	arpstat.as_sndreply++;
1143 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1144 	return;
1145 }
1146 
1147 /*
1148  * Free an arp entry.
1149  */
1150 static void arptfree(struct llinfo_arp *la)
1151 {
1152 	struct rtentry *rt = la->la_rt;
1153 	struct sockaddr_dl *sdl;
1154 
1155 	ARP_LOCK_CHECK();
1156 
1157 	if (rt == 0)
1158 		panic("arptfree");
1159 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
1160 	    sdl->sdl_family == AF_LINK) {
1161 		sdl->sdl_alen = 0;
1162 		la->la_asked = 0;
1163 		rt->rt_flags &= ~RTF_REJECT;
1164 		return;
1165 	}
1166 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
1167 	    0, (struct rtentry **)0);
1168 }
1169 
1170 /*
1171  * Lookup or enter a new address in arptab.
1172  */
1173 static struct llinfo_arp *
1174 arplookup(struct mbuf *m, struct in_addr *addr, int create, int proxy)
1175 {
1176 	struct arphdr *ah;
1177 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1178 	struct rtentry *rt;
1179 	static struct sockaddr_inarp sin;
1180 	const char *why = 0;
1181 
1182 	ah = mtod(m, struct arphdr *);
1183 	sin.sin_len = sizeof(sin);
1184 	sin.sin_family = AF_INET;
1185 	sin.sin_addr = *addr;
1186 	sin.sin_other = proxy ? SIN_PROXY : 0;
1187 	rt = rtalloc1(sintosa(&sin), create);
1188 	if (rt == 0)
1189 		return (0);
1190 	rt->rt_refcnt--;
1191 
1192 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO &&
1193 	    rt->rt_gateway->sa_family == AF_LINK)
1194 		return ((struct llinfo_arp *)rt->rt_llinfo);
1195 
1196 
1197 
1198 	if (create) {
1199 		if (rt->rt_flags & RTF_GATEWAY)
1200 			why = "host is not on local network";
1201 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1202 			arpstat.as_allocfail++;
1203 			why = "could not allocate llinfo";
1204 		} else
1205 			why = "gateway route is not ours";
1206 		log(LOG_DEBUG, "arplookup: unable to enter address"
1207 		    " for %s@%s on %s (%s)\n",
1208 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1209 		    (ifp) ? ifp->if_xname : 0, why);
1210 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1211 			rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
1212 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1213 		}
1214 	}
1215 	return (0);
1216 }
1217 
1218 int
1219 arpioctl(u_long cmd, caddr_t data)
1220 {
1221 
1222 	return (EOPNOTSUPP);
1223 }
1224 
1225 void
1226 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1227 {
1228 	struct in_addr *ip;
1229 
1230 	/*
1231 	 * Warn the user if another station has this IP address,
1232 	 * but only if the interface IP address is not zero.
1233 	 */
1234 	ip = &IA_SIN(ifa)->sin_addr;
1235 	if (!in_nullhost(*ip))
1236 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
1237 
1238 	ifa->ifa_rtrequest = arp_rtrequest;
1239 	ifa->ifa_flags |= RTF_CLONING;
1240 }
1241 
1242 /*
1243  * Called from 10 Mb/s Ethernet interrupt handlers
1244  * when ether packet type ETHERTYPE_REVARP
1245  * is received.  Common length and type checks are done here,
1246  * then the protocol-specific routine is called.
1247  */
1248 void
1249 revarpinput(struct mbuf *m)
1250 {
1251 	struct arphdr *ar;
1252 
1253 	if (m->m_len < sizeof(struct arphdr))
1254 		goto out;
1255 	ar = mtod(m, struct arphdr *);
1256 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1257 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1258 		goto out;
1259 #endif
1260 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1261 		goto out;
1262 	switch (ntohs(ar->ar_pro)) {
1263 	case ETHERTYPE_IP:
1264 	case ETHERTYPE_IPTRAILERS:
1265 		in_revarpinput(m);
1266 		return;
1267 
1268 	default:
1269 		break;
1270 	}
1271 out:
1272 	m_freem(m);
1273 }
1274 
1275 /*
1276  * RARP for Internet protocols on 10 Mb/s Ethernet.
1277  * Algorithm is that given in RFC 903.
1278  * We are only using for bootstrap purposes to get an ip address for one of
1279  * our interfaces.  Thus we support no user-interface.
1280  *
1281  * Since the contents of the RARP reply are specific to the interface that
1282  * sent the request, this code must ensure that they are properly associated.
1283  *
1284  * Note: also supports ARP via RARP packets, per the RFC.
1285  */
1286 void
1287 in_revarpinput(struct mbuf *m)
1288 {
1289 	struct ifnet *ifp;
1290 	struct arphdr *ah;
1291 	caddr_t tha;
1292 	int op;
1293 
1294 	ah = mtod(m, struct arphdr *);
1295 	op = ntohs(ah->ar_op);
1296 
1297 	switch (m->m_pkthdr.rcvif->if_type) {
1298 	case IFT_IEEE1394:
1299 		/* ARP without target hardware address is not supported */
1300 		goto out;
1301 	default:
1302 		break;
1303 	}
1304 
1305 	switch (op) {
1306 	case ARPOP_REQUEST:
1307 	case ARPOP_REPLY:	/* per RFC */
1308 		in_arpinput(m);
1309 		return;
1310 	case ARPOP_REVREPLY:
1311 		break;
1312 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1313 	default:
1314 		goto out;
1315 	}
1316 	if (!revarp_in_progress)
1317 		goto out;
1318 	ifp = m->m_pkthdr.rcvif;
1319 	if (ifp != myip_ifp) /* !same interface */
1320 		goto out;
1321 	if (myip_initialized)
1322 		goto wake;
1323 	tha = ar_tha(ah);
1324 	KASSERT(tha);
1325 	if (bcmp(tha, LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1326 		goto out;
1327 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1328 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1329 	myip_initialized = 1;
1330 wake:	/* Do wakeup every time in case it was missed. */
1331 	wakeup((caddr_t)&myip);
1332 
1333 out:
1334 	m_freem(m);
1335 }
1336 
1337 /*
1338  * Send a RARP request for the ip address of the specified interface.
1339  * The request should be RFC 903-compliant.
1340  */
1341 void
1342 revarprequest(struct ifnet *ifp)
1343 {
1344 	struct sockaddr sa;
1345 	struct mbuf *m;
1346 	struct arphdr *ah;
1347 	caddr_t tha;
1348 
1349 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1350 		return;
1351 	MCLAIM(m, &arpdomain.dom_mowner);
1352 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1353 	    2*ifp->if_addrlen;
1354 	m->m_pkthdr.len = m->m_len;
1355 	MH_ALIGN(m, m->m_len);
1356 	ah = mtod(m, struct arphdr *);
1357 	bzero((caddr_t)ah, m->m_len);
1358 	ah->ar_pro = htons(ETHERTYPE_IP);
1359 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1360 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1361 	ah->ar_op = htons(ARPOP_REVREQUEST);
1362 
1363 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1364 	tha = ar_tha(ah);
1365 	KASSERT(tha);
1366 	bcopy(LLADDR(ifp->if_sadl), tha, ah->ar_hln);
1367 
1368 	sa.sa_family = AF_ARP;
1369 	sa.sa_len = 2;
1370 	m->m_flags |= M_BCAST;
1371 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1372 
1373 }
1374 
1375 /*
1376  * RARP for the ip address of the specified interface, but also
1377  * save the ip address of the server that sent the answer.
1378  * Timeout if no response is received.
1379  */
1380 int
1381 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1382     struct in_addr *clnt_in)
1383 {
1384 	int result, count = 20;
1385 
1386 	myip_initialized = 0;
1387 	myip_ifp = ifp;
1388 
1389 	revarp_in_progress = 1;
1390 	while (count--) {
1391 		revarprequest(ifp);
1392 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1393 		if (result != EWOULDBLOCK)
1394 			break;
1395 	}
1396 	revarp_in_progress = 0;
1397 
1398 	if (!myip_initialized)
1399 		return ENETUNREACH;
1400 
1401 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1402 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1403 	return 0;
1404 }
1405 
1406 
1407 
1408 #ifdef DDB
1409 
1410 #include <machine/db_machdep.h>
1411 #include <ddb/db_interface.h>
1412 #include <ddb/db_output.h>
1413 
1414 static void
1415 db_print_sa(const struct sockaddr *sa)
1416 {
1417 	int len;
1418 	const u_char *p;
1419 
1420 	if (sa == 0) {
1421 		db_printf("[NULL]");
1422 		return;
1423 	}
1424 
1425 	p = (const u_char *)sa;
1426 	len = sa->sa_len;
1427 	db_printf("[");
1428 	while (len > 0) {
1429 		db_printf("%d", *p);
1430 		p++; len--;
1431 		if (len) db_printf(",");
1432 	}
1433 	db_printf("]\n");
1434 }
1435 
1436 static void
1437 db_print_ifa(struct ifaddr *ifa)
1438 {
1439 	if (ifa == 0)
1440 		return;
1441 	db_printf("  ifa_addr=");
1442 	db_print_sa(ifa->ifa_addr);
1443 	db_printf("  ifa_dsta=");
1444 	db_print_sa(ifa->ifa_dstaddr);
1445 	db_printf("  ifa_mask=");
1446 	db_print_sa(ifa->ifa_netmask);
1447 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1448 			  ifa->ifa_flags,
1449 			  ifa->ifa_refcnt,
1450 			  ifa->ifa_metric);
1451 }
1452 
1453 static void
1454 db_print_llinfo(caddr_t li)
1455 {
1456 	struct llinfo_arp *la;
1457 
1458 	if (li == 0)
1459 		return;
1460 	la = (struct llinfo_arp *)li;
1461 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1462 			  la->la_rt, la->la_hold, la->la_asked);
1463 }
1464 
1465 /*
1466  * Function to pass to rn_walktree().
1467  * Return non-zero error to abort walk.
1468  */
1469 static int
1470 db_show_radix_node(struct radix_node *rn, void *w)
1471 {
1472 	struct rtentry *rt = (struct rtentry *)rn;
1473 
1474 	db_printf("rtentry=%p", rt);
1475 
1476 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1477 			  rt->rt_flags, rt->rt_refcnt,
1478 			  rt->rt_use, rt->rt_expire);
1479 
1480 	db_printf(" key="); db_print_sa(rt_key(rt));
1481 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1482 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1483 
1484 	db_printf(" ifp=%p ", rt->rt_ifp);
1485 	if (rt->rt_ifp)
1486 		db_printf("(%s)", rt->rt_ifp->if_xname);
1487 	else
1488 		db_printf("(NULL)");
1489 
1490 	db_printf(" ifa=%p\n", rt->rt_ifa);
1491 	db_print_ifa(rt->rt_ifa);
1492 
1493 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1494 
1495 	db_printf(" gwroute=%p llinfo=%p\n",
1496 			  rt->rt_gwroute, rt->rt_llinfo);
1497 	db_print_llinfo(rt->rt_llinfo);
1498 
1499 	return (0);
1500 }
1501 
1502 /*
1503  * Function to print all the route trees.
1504  * Use this from ddb:  "show arptab"
1505  */
1506 void
1507 db_show_arptab(db_expr_t addr, int have_addr,
1508     db_expr_t count, const char *modif)
1509 {
1510 	struct radix_node_head *rnh;
1511 	rnh = rt_tables[AF_INET];
1512 	db_printf("Route tree for AF_INET\n");
1513 	if (rnh == NULL) {
1514 		db_printf(" (not initialized)\n");
1515 		return;
1516 	}
1517 	rn_walktree(rnh, db_show_radix_node, NULL);
1518 	return;
1519 }
1520 #endif
1521 
1522 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1523 {
1524 	const struct sysctlnode *node;
1525 
1526 	sysctl_createv(clog, 0, NULL, NULL,
1527 			CTLFLAG_PERMANENT,
1528 			CTLTYPE_NODE, "net", NULL,
1529 			NULL, 0, NULL, 0,
1530 			CTL_NET, CTL_EOL);
1531 	sysctl_createv(clog, 0, NULL, NULL,
1532 			CTLFLAG_PERMANENT,
1533 			CTLTYPE_NODE, "inet", NULL,
1534 			NULL, 0, NULL, 0,
1535 			CTL_NET, PF_INET, CTL_EOL);
1536 	sysctl_createv(clog, 0, NULL, &node,
1537 			CTLFLAG_PERMANENT,
1538 			CTLTYPE_NODE, "arp",
1539 			SYSCTL_DESCR("Address Resolution Protocol"),
1540 			NULL, 0, NULL, 0,
1541 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1542 
1543 	sysctl_createv(clog, 0, NULL, NULL,
1544 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1545 			CTLTYPE_INT, "prune",
1546 			SYSCTL_DESCR("ARP cache pruning interval"),
1547 			NULL, 0, &arpt_prune, 0,
1548 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1549 
1550 	sysctl_createv(clog, 0, NULL, NULL,
1551 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1552 			CTLTYPE_INT, "keep",
1553 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1554 			NULL, 0, &arpt_keep, 0,
1555 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1556 
1557 	sysctl_createv(clog, 0, NULL, NULL,
1558 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1559 			CTLTYPE_INT, "down",
1560 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1561 			NULL, 0, &arpt_down, 0,
1562 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1563 
1564 	sysctl_createv(clog, 0, NULL, NULL,
1565 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1566 			CTLTYPE_INT, "refresh",
1567 			SYSCTL_DESCR("ARP entry refresh interval"),
1568 			NULL, 0, &arpt_refresh, 0,
1569 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1570 }
1571 
1572 #endif /* INET */
1573