xref: /netbsd-src/sys/netinet/if_arp.c (revision 0df165c04d0a9ca1adde9ed2b890344c937954a6)
1 /*	$NetBSD: if_arp.c,v 1.129 2007/11/14 01:11:14 cube 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 1.129 2007/11/14 01:11:14 cube 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 "arcnet.h"
119 #if NARCNET > 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 SRP(s) ((struct sockaddr_inarp *)s)
134 
135 /*
136  * ARP trailer negotiation.  Trailer protocol is not IP specific,
137  * but ARP request/response use IP addresses.
138  */
139 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
140 
141 /* timer values */
142 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
143 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
144 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
145 int	arpt_refresh = (5*60);	/* time left before refreshing */
146 #define	rt_expire rt_rmx.rmx_expire
147 #define	rt_pksent rt_rmx.rmx_pksent
148 
149 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
150 	    const struct sockaddr *);
151 static	void arptfree(struct llinfo_arp *);
152 static	void arptimer(void *);
153 static	struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *,
154 				      int, int, struct rtentry *);
155 static	struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *,
156 					  int, int);
157 static	void in_arpinput(struct mbuf *);
158 
159 LIST_HEAD(, llinfo_arp) llinfo_arp;
160 struct	ifqueue arpintrq = {
161 	.ifq_head = NULL,
162 	.ifq_tail = NULL,
163 	.ifq_len = 0,
164 	.ifq_maxlen = 50,
165 	.ifq_drops = 0,
166 };
167 int	arp_inuse, arp_allocated, arp_intimer;
168 int	arp_maxtries = 5;
169 int	useloopback = 1;	/* use loopback interface for local traffic */
170 int	arpinit_done = 0;
171 
172 struct	arpstat arpstat;
173 struct	callout arptimer_ch;
174 
175 /* revarp state */
176 struct	in_addr myip, srv_ip;
177 int	myip_initialized = 0;
178 int	revarp_in_progress = 0;
179 struct	ifnet *myip_ifp = NULL;
180 
181 #ifdef DDB
182 static void db_print_sa(const struct sockaddr *);
183 static void db_print_ifa(struct ifaddr *);
184 static void db_print_llinfo(void *);
185 static int db_show_rtentry(struct rtentry *, void *);
186 #endif
187 
188 /*
189  * this should be elsewhere.
190  */
191 
192 static char *
193 lla_snprintf(u_int8_t *, int);
194 
195 static char *
196 lla_snprintf(u_int8_t *adrp, int len)
197 {
198 #define NUMBUFS 3
199 	static char buf[NUMBUFS][16*3];
200 	static int bnum = 0;
201 
202 	int i;
203 	char *p;
204 
205 	p = buf[bnum];
206 
207 	*p++ = hexdigits[(*adrp)>>4];
208 	*p++ = hexdigits[(*adrp++)&0xf];
209 
210 	for (i=1; i<len && i<16; i++) {
211 		*p++ = ':';
212 		*p++ = hexdigits[(*adrp)>>4];
213 		*p++ = hexdigits[(*adrp++)&0xf];
214 	}
215 
216 	*p = 0;
217 	p = buf[bnum];
218 	bnum = (bnum + 1) % NUMBUFS;
219 	return p;
220 }
221 
222 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
223 
224 const struct protosw arpsw[] = {
225 	{ 0, &arpdomain, 0, 0,
226 	  0, 0, 0, 0,
227 	  0,
228 	  0, 0, 0, arp_drain,
229 	}
230 };
231 
232 
233 struct domain arpdomain = {
234 	.dom_family = PF_ARP,
235 	.dom_name = "arp",
236 	.dom_protosw = arpsw,
237 	.dom_protoswNPROTOSW = &arpsw[sizeof(arpsw)/sizeof(arpsw[0])],
238 };
239 
240 /*
241  * ARP table locking.
242  *
243  * to prevent lossage vs. the arp_drain routine (which may be called at
244  * any time, including in a device driver context), we do two things:
245  *
246  * 1) manipulation of la->la_hold is done at splnet() (for all of
247  * about two instructions).
248  *
249  * 2) manipulation of the arp table's linked list is done under the
250  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
251  * while the arp table is locked, we punt and try again later.
252  */
253 
254 static int	arp_locked;
255 static inline int arp_lock_try(int);
256 static inline void arp_unlock(void);
257 
258 static inline int
259 arp_lock_try(int recurse)
260 {
261 	int s;
262 
263 	/*
264 	 * Use splvm() -- we're blocking things that would cause
265 	 * mbuf allocation.
266 	 */
267 	s = splvm();
268 	if (!recurse && arp_locked) {
269 		splx(s);
270 		return (0);
271 	}
272 	arp_locked++;
273 	splx(s);
274 	return (1);
275 }
276 
277 static inline void
278 arp_unlock(void)
279 {
280 	int s;
281 
282 	s = splvm();
283 	arp_locked--;
284 	splx(s);
285 }
286 
287 #ifdef DIAGNOSTIC
288 #define	ARP_LOCK(recurse)						\
289 do {									\
290 	if (arp_lock_try(recurse) == 0) {				\
291 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
292 		panic("arp_lock");					\
293 	}								\
294 } while (/*CONSTCOND*/ 0)
295 #define	ARP_LOCK_CHECK()						\
296 do {									\
297 	if (arp_locked == 0) {						\
298 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
299 		panic("arp lock check");				\
300 	}								\
301 } while (/*CONSTCOND*/ 0)
302 #else
303 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
304 #define	ARP_LOCK_CHECK()	/* nothing */
305 #endif
306 
307 #define	ARP_UNLOCK()		arp_unlock()
308 
309 /*
310  * ARP protocol drain routine.  Called when memory is in short supply.
311  * Called at splvm();
312  */
313 
314 void
315 arp_drain(void)
316 {
317 	struct llinfo_arp *la, *nla;
318 	int count = 0;
319 	struct mbuf *mold;
320 
321 	if (arp_lock_try(0) == 0) {
322 		printf("arp_drain: locked; punting\n");
323 		return;
324 	}
325 
326 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
327 		nla = LIST_NEXT(la, la_list);
328 
329 		mold = la->la_hold;
330 		la->la_hold = 0;
331 
332 		if (mold) {
333 			m_freem(mold);
334 			count++;
335 		}
336 	}
337 	ARP_UNLOCK();
338 	arpstat.as_dfrdropped += count;
339 }
340 
341 
342 /*
343  * Timeout routine.  Age arp_tab entries periodically.
344  */
345 /* ARGSUSED */
346 static void
347 arptimer(void *arg)
348 {
349 	int s;
350 	struct llinfo_arp *la, *nla;
351 
352 	s = splsoftnet();
353 
354 	if (arp_lock_try(0) == 0) {
355 		/* get it later.. */
356 		splx(s);
357 		return;
358 	}
359 
360 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
361 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
362 		struct rtentry *rt = la->la_rt;
363 
364 		nla = LIST_NEXT(la, la_list);
365 		if (rt->rt_expire == 0)
366 			continue;
367 		if ((rt->rt_expire - time_second) < arpt_refresh &&
368 		    rt->rt_pksent > (time_second - arpt_keep)) {
369 			/*
370 			 * If the entry has been used during since last
371 			 * refresh, try to renew it before deleting.
372 			 */
373 			arprequest(rt->rt_ifp,
374 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
375 			    &satocsin(rt_getkey(rt))->sin_addr,
376 			    CLLADDR(rt->rt_ifp->if_sadl));
377 		} else if (rt->rt_expire <= time_second)
378 			arptfree(la); /* timer has expired; clear */
379 	}
380 
381 	ARP_UNLOCK();
382 
383 	splx(s);
384 }
385 
386 /*
387  * We set the gateway for RTF_CLONING routes to a "prototype"
388  * link-layer sockaddr whose interface type (if_type) and interface
389  * index (if_index) fields are prepared.
390  */
391 static struct sockaddr *
392 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
393     const struct sockaddr *netmask)
394 {
395 	const struct ifnet *ifp = rt->rt_ifp;
396 	uint8_t namelen = strlen(ifp->if_xname);
397 	uint8_t addrlen = ifp->if_addrlen;
398 
399 	/*
400 	 * XXX: If this is a manually added route to interface
401 	 * such as older version of routed or gated might provide,
402 	 * restore cloning bit.
403 	 */
404 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
405 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
406 		rt->rt_flags |= RTF_CLONING;
407 	if (rt->rt_flags & RTF_CLONING) {
408 		union {
409 			struct sockaddr sa;
410 			struct sockaddr_storage ss;
411 			struct sockaddr_dl sdl;
412 		} u;
413 		/*
414 		 * Case 1: This route should come from a route to iface.
415 		 */
416 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
417 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
418 		rt_setgate(rt, &u.sa);
419 		gate = rt->rt_gateway;
420 	}
421 	return gate;
422 }
423 
424 /*
425  * Parallel to llc_rtrequest.
426  */
427 void
428 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
429 {
430 	struct sockaddr *gate = rt->rt_gateway;
431 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
432 	size_t allocsize;
433 	struct mbuf *mold;
434 	int s;
435 	struct in_ifaddr *ia;
436 	struct ifaddr *ifa;
437 	struct ifnet *ifp = rt->rt_ifp;
438 	uint8_t namelen = strlen(ifp->if_xname);
439 	uint8_t addrlen = ifp->if_addrlen;
440 
441 	if (!arpinit_done) {
442 		arpinit_done = 1;
443 		/*
444 		 * We generate expiration times from time_second
445 		 * so avoid accidentally creating permanent routes.
446 		 */
447 		if (time_second == 0) {
448 #ifdef __HAVE_TIMECOUNTER
449 			struct timespec ts;
450 			ts.tv_sec = 1;
451 			ts.tv_nsec = 0;
452 			tc_setclock(&ts);
453 #else /* !__HAVE_TIMECOUNTER */
454 			time.tv_sec++;
455 #endif /* !__HAVE_TIMECOUNTER */
456 		}
457 		callout_init(&arptimer_ch, 0);
458 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
459 	}
460 
461 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
462 		if (req != RTM_ADD)
463 			return;
464 
465 		/*
466 		 * linklayers with particular link MTU limitation.
467 		 */
468 		switch(ifp->if_type) {
469 #if NFDDI > 0
470 		case IFT_FDDI:
471 			if (ifp->if_mtu > FDDIIPMTU)
472 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
473 			break;
474 #endif
475 #if NARC > 0
476 		case IFT_ARCNET:
477 		    {
478 			int arcipifmtu;
479 
480 			if (ifp->if_flags & IFF_LINK0)
481 				arcipifmtu = arc_ipmtu;
482 			else
483 				arcipifmtu = ARCMTU;
484 			if (ifp->if_mtu > arcipifmtu)
485 				rt->rt_rmx.rmx_mtu = arcipifmtu;
486 			break;
487 		    }
488 #endif
489 		}
490 		return;
491 	}
492 
493 	ARP_LOCK(1);		/* we may already be locked here. */
494 
495 	switch (req) {
496 
497 	case RTM_SETGATE:
498 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
499 		break;
500 	case RTM_ADD:
501 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
502 		if (rt->rt_flags & RTF_CLONING) {
503 			/*
504 			 * Give this route an expiration time, even though
505 			 * it's a "permanent" route, so that routes cloned
506 			 * from it do not need their expiration time set.
507 			 */
508 			rt->rt_expire = time_second;
509 			/*
510 			 * linklayers with particular link MTU limitation.
511 			 */
512 			switch (ifp->if_type) {
513 #if NFDDI > 0
514 			case IFT_FDDI:
515 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
516 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
517 				     (rt->rt_rmx.rmx_mtu == 0 &&
518 				      ifp->if_mtu > FDDIIPMTU)))
519 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
520 				break;
521 #endif
522 #if NARC > 0
523 			case IFT_ARCNET:
524 			    {
525 				int arcipifmtu;
526 				if (ifp->if_flags & IFF_LINK0)
527 					arcipifmtu = arc_ipmtu;
528 				else
529 					arcipifmtu = ARCMTU;
530 
531 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
532 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
533 				     (rt->rt_rmx.rmx_mtu == 0 &&
534 				      ifp->if_mtu > arcipifmtu)))
535 					rt->rt_rmx.rmx_mtu = arcipifmtu;
536 				break;
537 			    }
538 #endif
539 			}
540 			break;
541 		}
542 		/* Announce a new entry if requested. */
543 		if (rt->rt_flags & RTF_ANNOUNCE)
544 			arprequest(ifp,
545 			    &satocsin(rt_getkey(rt))->sin_addr,
546 			    &satocsin(rt_getkey(rt))->sin_addr,
547 			    CLLADDR(satocsdl(gate)));
548 		/*FALLTHROUGH*/
549 	case RTM_RESOLVE:
550 		if (gate->sa_family != AF_LINK ||
551 		    gate->sa_len < sockaddr_dl_measure(namelen, addrlen)) {
552 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
553 			break;
554 		}
555 		satosdl(gate)->sdl_type = ifp->if_type;
556 		satosdl(gate)->sdl_index = ifp->if_index;
557 		if (la != 0)
558 			break; /* This happens on a route change */
559 		/*
560 		 * Case 2:  This route may come from cloning, or a manual route
561 		 * add with a LL address.
562 		 */
563 		switch (satocsdl(gate)->sdl_type) {
564 #if NTOKEN > 0
565 		case IFT_ISO88025:
566 			allocsize = sizeof(*la) + sizeof(struct token_rif);
567 			break;
568 #endif /* NTOKEN > 0 */
569 		default:
570 			allocsize = sizeof(*la);
571 		}
572 		R_Malloc(la, struct llinfo_arp *, allocsize);
573 		rt->rt_llinfo = (void *)la;
574 		if (la == 0) {
575 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
576 			break;
577 		}
578 		arp_inuse++, arp_allocated++;
579 		Bzero(la, allocsize);
580 		la->la_rt = rt;
581 		rt->rt_flags |= RTF_LLINFO;
582 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
583 
584 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
585 		while (ia && ia->ia_ifp != ifp)
586 			NEXT_IA_WITH_SAME_ADDR(ia);
587 		if (ia) {
588 			/*
589 			 * This test used to be
590 			 *	if (lo0ifp->if_flags & IFF_UP)
591 			 * It allowed local traffic to be forced through
592 			 * the hardware by configuring the loopback down.
593 			 * However, it causes problems during network
594 			 * configuration for boards that can't receive
595 			 * packets they send.  It is now necessary to clear
596 			 * "useloopback" and remove the route to force
597 			 * traffic out to the hardware.
598 			 *
599 			 * In 4.4BSD, the above "if" statement checked
600 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
601 			 * to the current form so that we can provide a
602 			 * better support for multiple IPv4 addresses on a
603 			 * interface.
604 			 */
605 			rt->rt_expire = 0;
606 			(void)sockaddr_dl_setaddr(satosdl(gate), gate->sa_len,
607 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen);
608 			if (useloopback)
609 				ifp = rt->rt_ifp = lo0ifp;
610 			/*
611 			 * make sure to set rt->rt_ifa to the interface
612 			 * address we are using, otherwise we will have trouble
613 			 * with source address selection.
614 			 */
615 			ifa = &ia->ia_ifa;
616 			if (ifa != rt->rt_ifa)
617 				rt_replace_ifa(rt, ifa);
618 		}
619 		break;
620 
621 	case RTM_DELETE:
622 		if (la == 0)
623 			break;
624 		arp_inuse--;
625 		LIST_REMOVE(la, la_list);
626 		rt->rt_llinfo = 0;
627 		rt->rt_flags &= ~RTF_LLINFO;
628 
629 		s = splnet();
630 		mold = la->la_hold;
631 		la->la_hold = 0;
632 		splx(s);
633 
634 		if (mold)
635 			m_freem(mold);
636 
637 		Free((void *)la);
638 	}
639 	ARP_UNLOCK();
640 }
641 
642 /*
643  * Broadcast an ARP request. Caller specifies:
644  *	- arp header source ip address
645  *	- arp header target ip address
646  *	- arp header source ethernet address
647  */
648 void
649 arprequest(struct ifnet *ifp,
650     const struct in_addr *sip, const struct in_addr *tip,
651     const u_int8_t *enaddr)
652 {
653 	struct mbuf *m;
654 	struct arphdr *ah;
655 	struct sockaddr sa;
656 
657 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
658 		return;
659 	MCLAIM(m, &arpdomain.dom_mowner);
660 	switch (ifp->if_type) {
661 	case IFT_IEEE1394:
662 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
663 		    ifp->if_addrlen;
664 		break;
665 	default:
666 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
667 		    2 * ifp->if_addrlen;
668 		break;
669 	}
670 	m->m_pkthdr.len = m->m_len;
671 	MH_ALIGN(m, m->m_len);
672 	ah = mtod(m, struct arphdr *);
673 	bzero((void *)ah, m->m_len);
674 	switch (ifp->if_type) {
675 	case IFT_IEEE1394:	/* RFC2734 */
676 		/* fill it now for ar_tpa computation */
677 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
678 		break;
679 	default:
680 		/* ifp->if_output will fill ar_hrd */
681 		break;
682 	}
683 	ah->ar_pro = htons(ETHERTYPE_IP);
684 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
685 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
686 	ah->ar_op = htons(ARPOP_REQUEST);
687 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
688 	memcpy(ar_spa(ah), sip, ah->ar_pln);
689 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
690 	sa.sa_family = AF_ARP;
691 	sa.sa_len = 2;
692 	m->m_flags |= M_BCAST;
693 	arpstat.as_sndtotal++;
694 	arpstat.as_sndrequest++;
695 	(*ifp->if_output)(ifp, m, &sa, NULL);
696 }
697 
698 /*
699  * Resolve an IP address into an ethernet address.  If success,
700  * desten is filled in.  If there is no entry in arptab,
701  * set one up and broadcast a request for the IP address.
702  * Hold onto this mbuf and resend it once the address
703  * is finally resolved.  A return value of 1 indicates
704  * that desten has been filled in and the packet should be sent
705  * normally; a 0 return indicates that the packet has been
706  * taken over here, either now or for later transmission.
707  */
708 int
709 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
710     const struct sockaddr *dst, u_char *desten)
711 {
712 	struct llinfo_arp *la;
713 	const struct sockaddr_dl *sdl;
714 	struct mbuf *mold;
715 	int s;
716 
717 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
718 		rt = la->la_rt;
719 
720 	if (la == 0 || rt == 0) {
721 		arpstat.as_allocfail++;
722 		log(LOG_DEBUG,
723 		    "arpresolve: can't allocate llinfo on %s for %s\n",
724 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
725 		m_freem(m);
726 		return (0);
727 	}
728 	sdl = satocsdl(rt->rt_gateway);
729 	/*
730 	 * Check the address family and length is valid, the address
731 	 * is resolved; otherwise, try to resolve.
732 	 */
733 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
734 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
735 		bcopy(CLLADDR(sdl), desten,
736 		    min(sdl->sdl_alen, ifp->if_addrlen));
737 		rt->rt_pksent = time_second; /* Time for last pkt sent */
738 		return 1;
739 	}
740 	/*
741 	 * There is an arptab entry, but no ethernet address
742 	 * response yet.  Replace the held mbuf with this
743 	 * latest one.
744 	 */
745 
746 	arpstat.as_dfrtotal++;
747 	s = splnet();
748 	mold = la->la_hold;
749 	la->la_hold = m;
750 	splx(s);
751 
752 	if (mold) {
753 		arpstat.as_dfrdropped++;
754 		m_freem(mold);
755 	}
756 
757 	/*
758 	 * Re-send the ARP request when appropriate.
759 	 */
760 #ifdef	DIAGNOSTIC
761 	if (rt->rt_expire == 0) {
762 		/* This should never happen. (Should it? -gwr) */
763 		printf("arpresolve: unresolved and rt_expire == 0\n");
764 		/* Set expiration time to now (expired). */
765 		rt->rt_expire = time_second;
766 	}
767 #endif
768 	if (rt->rt_expire) {
769 		rt->rt_flags &= ~RTF_REJECT;
770 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
771 			rt->rt_expire = time_second;
772 			if (la->la_asked++ < arp_maxtries)
773 				arprequest(ifp,
774 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
775 				    &satocsin(dst)->sin_addr,
776 #if NCARP > 0
777 				    (rt->rt_ifp->if_type == IFT_CARP) ?
778 				    CLLADDR(rt->rt_ifp->if_sadl):
779 #endif
780 				    CLLADDR(ifp->if_sadl));
781 			else {
782 				rt->rt_flags |= RTF_REJECT;
783 				rt->rt_expire += arpt_down;
784 				la->la_asked = 0;
785 			}
786 		}
787 	}
788 	return (0);
789 }
790 
791 /*
792  * Common length and type checks are done here,
793  * then the protocol-specific routine is called.
794  */
795 void
796 arpintr(void)
797 {
798 	struct mbuf *m;
799 	struct arphdr *ar;
800 	int s;
801 	int arplen;
802 
803 	while (arpintrq.ifq_head) {
804 		s = splnet();
805 		IF_DEQUEUE(&arpintrq, m);
806 		splx(s);
807 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
808 			panic("arpintr");
809 
810 		MCLAIM(m, &arpdomain.dom_mowner);
811 		arpstat.as_rcvtotal++;
812 
813 		/*
814 		 * First, make sure we have at least struct arphdr.
815 		 */
816 		if (m->m_len < sizeof(struct arphdr) ||
817 		    (ar = mtod(m, struct arphdr *)) == NULL)
818 			goto badlen;
819 
820 		switch (m->m_pkthdr.rcvif->if_type) {
821 		case IFT_IEEE1394:
822 			arplen = sizeof(struct arphdr) +
823 			    ar->ar_hln + 2 * ar->ar_pln;
824 			break;
825 		default:
826 			arplen = sizeof(struct arphdr) +
827 			    2 * ar->ar_hln + 2 * ar->ar_pln;
828 			break;
829 		}
830 
831 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
832 		    m->m_len >= arplen)
833 			switch (ntohs(ar->ar_pro)) {
834 			case ETHERTYPE_IP:
835 			case ETHERTYPE_IPTRAILERS:
836 				in_arpinput(m);
837 				continue;
838 			default:
839 				arpstat.as_rcvbadproto++;
840 			}
841 		else {
842 badlen:
843 			arpstat.as_rcvbadlen++;
844 		}
845 		m_freem(m);
846 	}
847 }
848 
849 /*
850  * ARP for Internet protocols on 10 Mb/s Ethernet.
851  * Algorithm is that given in RFC 826.
852  * In addition, a sanity check is performed on the sender
853  * protocol address, to catch impersonators.
854  * We no longer handle negotiations for use of trailer protocol:
855  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
856  * along with IP replies if we wanted trailers sent to us,
857  * and also sent them in response to IP replies.
858  * This allowed either end to announce the desire to receive
859  * trailer packets.
860  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
861  * but formerly didn't normally send requests.
862  */
863 static void
864 in_arpinput(struct mbuf *m)
865 {
866 	struct arphdr *ah;
867 	struct ifnet *ifp = m->m_pkthdr.rcvif;
868 	struct llinfo_arp *la = 0;
869 	struct rtentry  *rt;
870 	struct in_ifaddr *ia;
871 #if NBRIDGE > 0
872 	struct in_ifaddr *bridge_ia = NULL;
873 #endif
874 #if NCARP > 0
875 	u_int32_t count = 0, index = 0;
876 #endif
877 	struct sockaddr_dl *sdl;
878 	struct sockaddr sa;
879 	struct in_addr isaddr, itaddr, myaddr;
880 	int op;
881 	struct mbuf *mold;
882 	void *tha;
883 	int s;
884 
885 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
886 		goto out;
887 	ah = mtod(m, struct arphdr *);
888 	op = ntohs(ah->ar_op);
889 
890 	/*
891 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
892 	 * ar_tpa().
893 	 */
894 	switch (ifp->if_type) {
895 	case IFT_IEEE1394:
896 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
897 			;
898 		else {
899 			/* XXX this is to make sure we compute ar_tha right */
900 			/* XXX check ar_hrd more strictly? */
901 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
902 		}
903 		break;
904 	default:
905 		/* XXX check ar_hrd? */
906 		break;
907 	}
908 
909 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
910 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
911 
912 	if (m->m_flags & (M_BCAST|M_MCAST))
913 		arpstat.as_rcvmcast++;
914 
915 	/*
916 	 * If the target IP address is zero, ignore the packet.
917 	 * This prevents the code below from tring to answer
918 	 * when we are using IP address zero (booting).
919 	 */
920 	if (in_nullhost(itaddr)) {
921 		arpstat.as_rcvzerotpa++;
922 		goto out;
923 	}
924 
925 	/*
926 	 * If the source IP address is zero, this is most likely a
927 	 * confused host trying to use IP address zero. (Windoze?)
928 	 * XXX: Should we bother trying to reply to these?
929 	 */
930 	if (in_nullhost(isaddr)) {
931 		arpstat.as_rcvzerospa++;
932 		goto out;
933 	}
934 
935 	/*
936 	 * Search for a matching interface address
937 	 * or any address on the interface to use
938 	 * as a dummy address in the rest of this function
939 	 */
940 
941 	INADDR_TO_IA(itaddr, ia);
942 	while (ia != NULL) {
943 #if NCARP > 0
944 		if (ia->ia_ifp->if_type == IFT_CARP &&
945 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
946 		    (IFF_UP|IFF_RUNNING))) {
947 			index++;
948 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
949 			    carp_iamatch(ia, ar_sha(ah),
950 			    &count, index)) {
951 				break;
952 				}
953 		} else
954 #endif
955 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
956 				break;
957 #if NBRIDGE > 0
958 		/*
959 		 * If the interface we received the packet on
960 		 * is part of a bridge, check to see if we need
961 		 * to "bridge" the packet to ourselves at this
962 		 * layer.  Note we still prefer a perfect match,
963 		 * but allow this weaker match if necessary.
964 		 */
965 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
966 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
967 			bridge_ia = ia;
968 #endif /* NBRIDGE > 0 */
969 
970 		NEXT_IA_WITH_SAME_ADDR(ia);
971 	}
972 
973 #if NBRIDGE > 0
974 	if (ia == NULL && bridge_ia != NULL) {
975 		ia = bridge_ia;
976 		ifp = bridge_ia->ia_ifp;
977 	}
978 #endif
979 
980 	if (ia == NULL) {
981 		INADDR_TO_IA(isaddr, ia);
982 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
983 			NEXT_IA_WITH_SAME_ADDR(ia);
984 
985 		if (ia == NULL) {
986 			IFP_TO_IA(ifp, ia);
987 			if (ia == NULL) {
988 				arpstat.as_rcvnoint++;
989 				goto out;
990 			}
991 		}
992 	}
993 
994 	myaddr = ia->ia_addr.sin_addr;
995 
996 	/* XXX checks for bridge case? */
997 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
998 		arpstat.as_rcvlocalsha++;
999 		goto out;	/* it's from me, ignore it. */
1000 	}
1001 
1002 	/* XXX checks for bridge case? */
1003 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1004 		arpstat.as_rcvbcastsha++;
1005 		log(LOG_ERR,
1006 		    "%s: arp: link address is broadcast for IP address %s!\n",
1007 		    ifp->if_xname, in_fmtaddr(isaddr));
1008 		goto out;
1009 	}
1010 
1011 	if (in_hosteq(isaddr, myaddr)) {
1012 		arpstat.as_rcvlocalspa++;
1013 		log(LOG_ERR,
1014 		   "duplicate IP address %s sent from link address %s\n",
1015 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
1016 		itaddr = myaddr;
1017 		goto reply;
1018 	}
1019 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1020 	if (la && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
1021 		if (sdl->sdl_alen &&
1022 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
1023 			if (rt->rt_flags & RTF_STATIC) {
1024 				arpstat.as_rcvoverperm++;
1025 				log(LOG_INFO,
1026 				    "%s tried to overwrite permanent arp info"
1027 				    " for %s\n",
1028 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1029 				    in_fmtaddr(isaddr));
1030 				goto out;
1031 			} else if (rt->rt_ifp != ifp) {
1032 				arpstat.as_rcvoverint++;
1033 				log(LOG_INFO,
1034 				    "%s on %s tried to overwrite "
1035 				    "arp info for %s on %s\n",
1036 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1037 				    ifp->if_xname, in_fmtaddr(isaddr),
1038 				    rt->rt_ifp->if_xname);
1039 				    goto out;
1040 			} else {
1041 				arpstat.as_rcvover++;
1042 				log(LOG_INFO,
1043 				    "arp info overwritten for %s by %s\n",
1044 				    in_fmtaddr(isaddr),
1045 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1046 			}
1047 		}
1048 		/*
1049 		 * sanity check for the address length.
1050 		 * XXX this does not work for protocols with variable address
1051 		 * length. -is
1052 		 */
1053 		if (sdl->sdl_alen &&
1054 		    sdl->sdl_alen != ah->ar_hln) {
1055 			arpstat.as_rcvlenchg++;
1056 			log(LOG_WARNING,
1057 			    "arp from %s: new addr len %d, was %d",
1058 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1059 		}
1060 		if (ifp->if_addrlen != ah->ar_hln) {
1061 			arpstat.as_rcvbadlen++;
1062 			log(LOG_WARNING,
1063 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1064 			    in_fmtaddr(isaddr), ah->ar_hln,
1065 			    ifp->if_addrlen);
1066 			goto reply;
1067 		}
1068 #if NTOKEN > 0
1069 		/*
1070 		 * XXX uses m_data and assumes the complete answer including
1071 		 * XXX token-ring headers is in the same buf
1072 		 */
1073 		if (ifp->if_type == IFT_ISO88025) {
1074 			struct token_header *trh;
1075 
1076 			trh = (struct token_header *)M_TRHSTART(m);
1077 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1078 				struct token_rif	*rif;
1079 				size_t	riflen;
1080 
1081 				rif = TOKEN_RIF(trh);
1082 				riflen = (ntohs(rif->tr_rcf) &
1083 				    TOKEN_RCF_LEN_MASK) >> 8;
1084 
1085 				if (riflen > 2 &&
1086 				    riflen < sizeof(struct token_rif) &&
1087 				    (riflen & 1) == 0) {
1088 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1089 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1090 					bcopy(rif, TOKEN_RIF(la), riflen);
1091 				}
1092 			}
1093 		}
1094 #endif /* NTOKEN > 0 */
1095 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
1096 		    ah->ar_hln);
1097 		if (rt->rt_expire)
1098 			rt->rt_expire = time_second + arpt_keep;
1099 		rt->rt_flags &= ~RTF_REJECT;
1100 		la->la_asked = 0;
1101 
1102 		s = splnet();
1103 		mold = la->la_hold;
1104 		la->la_hold = 0;
1105 		splx(s);
1106 
1107 		if (mold) {
1108 			arpstat.as_dfrsent++;
1109 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
1110 		}
1111 	}
1112 reply:
1113 	if (op != ARPOP_REQUEST) {
1114 		if (op == ARPOP_REPLY)
1115 			arpstat.as_rcvreply++;
1116 	out:
1117 		m_freem(m);
1118 		return;
1119 	}
1120 	arpstat.as_rcvrequest++;
1121 	if (in_hosteq(itaddr, myaddr)) {
1122 		/* I am the target */
1123 		tha = ar_tha(ah);
1124 		if (tha)
1125 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1126 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1127 	} else {
1128 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1129 		if (la == 0)
1130 			goto out;
1131 		rt = la->la_rt;
1132 		if (rt->rt_ifp->if_type == IFT_CARP &&
1133 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1134 			goto out;
1135 		tha = ar_tha(ah);
1136 		if (tha)
1137 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1138 		sdl = satosdl(rt->rt_gateway);
1139 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
1140 	}
1141 
1142 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1143 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1144 	ah->ar_op = htons(ARPOP_REPLY);
1145 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1146 	switch (ifp->if_type) {
1147 	case IFT_IEEE1394:
1148 		/*
1149 		 * ieee1394 arp reply is broadcast
1150 		 */
1151 		m->m_flags &= ~M_MCAST;
1152 		m->m_flags |= M_BCAST;
1153 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1154 		break;
1155 
1156 	default:
1157 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1158 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1159 		break;
1160 	}
1161 	m->m_pkthdr.len = m->m_len;
1162 	sa.sa_family = AF_ARP;
1163 	sa.sa_len = 2;
1164 	arpstat.as_sndtotal++;
1165 	arpstat.as_sndreply++;
1166 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1167 	return;
1168 }
1169 
1170 /*
1171  * Free an arp entry.
1172  */
1173 static void arptfree(struct llinfo_arp *la)
1174 {
1175 	struct rtentry *rt = la->la_rt;
1176 	struct sockaddr_dl *sdl;
1177 
1178 	ARP_LOCK_CHECK();
1179 
1180 	if (rt == 0)
1181 		panic("arptfree");
1182 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
1183 	    sdl->sdl_family == AF_LINK) {
1184 		sdl->sdl_alen = 0;
1185 		la->la_asked = 0;
1186 		rt->rt_flags &= ~RTF_REJECT;
1187 		return;
1188 	}
1189 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1190 }
1191 
1192 static struct llinfo_arp *
1193 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
1194 {
1195 	return arplookup1(m, addr, create, proxy, NULL);
1196 }
1197 
1198 /*
1199  * Lookup or enter a new address in arptab.
1200  */
1201 static struct llinfo_arp *
1202 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
1203     struct rtentry *rt0)
1204 {
1205 	struct arphdr *ah;
1206 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1207 	struct rtentry *rt;
1208 	static struct sockaddr_inarp sin;
1209 	const char *why = 0;
1210 
1211 	ah = mtod(m, struct arphdr *);
1212 	if (rt0 == NULL) {
1213 		sin.sin_len = sizeof(sin);
1214 		sin.sin_family = AF_INET;
1215 		sin.sin_addr = *addr;
1216 		sin.sin_other = proxy ? SIN_PROXY : 0;
1217 		rt = rtalloc1(sintosa(&sin), create);
1218 		if (rt == NULL)
1219 			return (NULL);
1220 		rt->rt_refcnt--;
1221 	} else
1222 		rt = rt0;
1223 
1224 #define	IS_LLINFO(__rt)							  \
1225 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
1226 	 (__rt)->rt_gateway->sa_family == AF_LINK)
1227 
1228 
1229 	if (IS_LLINFO(rt))
1230 		return ((struct llinfo_arp *)rt->rt_llinfo);
1231 
1232 	if (create) {
1233 		if (rt->rt_flags & RTF_GATEWAY)
1234 			why = "host is not on local network";
1235 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1236 			arpstat.as_allocfail++;
1237 			why = "could not allocate llinfo";
1238 		} else
1239 			why = "gateway route is not ours";
1240 		log(LOG_DEBUG, "arplookup: unable to enter address"
1241 		    " for %s@%s on %s (%s)\n",
1242 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1243 		    (ifp) ? ifp->if_xname : 0, why);
1244 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1245 			rtrequest(RTM_DELETE, rt_getkey(rt),
1246 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1247 		}
1248 	}
1249 	return (0);
1250 }
1251 
1252 int
1253 arpioctl(u_long cmd, void *data)
1254 {
1255 
1256 	return (EOPNOTSUPP);
1257 }
1258 
1259 void
1260 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1261 {
1262 	struct in_addr *ip;
1263 
1264 	/*
1265 	 * Warn the user if another station has this IP address,
1266 	 * but only if the interface IP address is not zero.
1267 	 */
1268 	ip = &IA_SIN(ifa)->sin_addr;
1269 	if (!in_nullhost(*ip))
1270 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
1271 
1272 	ifa->ifa_rtrequest = arp_rtrequest;
1273 	ifa->ifa_flags |= RTF_CLONING;
1274 }
1275 
1276 /*
1277  * Called from 10 Mb/s Ethernet interrupt handlers
1278  * when ether packet type ETHERTYPE_REVARP
1279  * is received.  Common length and type checks are done here,
1280  * then the protocol-specific routine is called.
1281  */
1282 void
1283 revarpinput(struct mbuf *m)
1284 {
1285 	struct arphdr *ar;
1286 
1287 	if (m->m_len < sizeof(struct arphdr))
1288 		goto out;
1289 	ar = mtod(m, struct arphdr *);
1290 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1291 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1292 		goto out;
1293 #endif
1294 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1295 		goto out;
1296 	switch (ntohs(ar->ar_pro)) {
1297 	case ETHERTYPE_IP:
1298 	case ETHERTYPE_IPTRAILERS:
1299 		in_revarpinput(m);
1300 		return;
1301 
1302 	default:
1303 		break;
1304 	}
1305 out:
1306 	m_freem(m);
1307 }
1308 
1309 /*
1310  * RARP for Internet protocols on 10 Mb/s Ethernet.
1311  * Algorithm is that given in RFC 903.
1312  * We are only using for bootstrap purposes to get an ip address for one of
1313  * our interfaces.  Thus we support no user-interface.
1314  *
1315  * Since the contents of the RARP reply are specific to the interface that
1316  * sent the request, this code must ensure that they are properly associated.
1317  *
1318  * Note: also supports ARP via RARP packets, per the RFC.
1319  */
1320 void
1321 in_revarpinput(struct mbuf *m)
1322 {
1323 	struct ifnet *ifp;
1324 	struct arphdr *ah;
1325 	void *tha;
1326 	int op;
1327 
1328 	ah = mtod(m, struct arphdr *);
1329 	op = ntohs(ah->ar_op);
1330 
1331 	switch (m->m_pkthdr.rcvif->if_type) {
1332 	case IFT_IEEE1394:
1333 		/* ARP without target hardware address is not supported */
1334 		goto out;
1335 	default:
1336 		break;
1337 	}
1338 
1339 	switch (op) {
1340 	case ARPOP_REQUEST:
1341 	case ARPOP_REPLY:	/* per RFC */
1342 		in_arpinput(m);
1343 		return;
1344 	case ARPOP_REVREPLY:
1345 		break;
1346 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1347 	default:
1348 		goto out;
1349 	}
1350 	if (!revarp_in_progress)
1351 		goto out;
1352 	ifp = m->m_pkthdr.rcvif;
1353 	if (ifp != myip_ifp) /* !same interface */
1354 		goto out;
1355 	if (myip_initialized)
1356 		goto wake;
1357 	tha = ar_tha(ah);
1358 	KASSERT(tha);
1359 	if (bcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1360 		goto out;
1361 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1362 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1363 	myip_initialized = 1;
1364 wake:	/* Do wakeup every time in case it was missed. */
1365 	wakeup((void *)&myip);
1366 
1367 out:
1368 	m_freem(m);
1369 }
1370 
1371 /*
1372  * Send a RARP request for the ip address of the specified interface.
1373  * The request should be RFC 903-compliant.
1374  */
1375 void
1376 revarprequest(struct ifnet *ifp)
1377 {
1378 	struct sockaddr sa;
1379 	struct mbuf *m;
1380 	struct arphdr *ah;
1381 	void *tha;
1382 
1383 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1384 		return;
1385 	MCLAIM(m, &arpdomain.dom_mowner);
1386 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1387 	    2*ifp->if_addrlen;
1388 	m->m_pkthdr.len = m->m_len;
1389 	MH_ALIGN(m, m->m_len);
1390 	ah = mtod(m, struct arphdr *);
1391 	bzero((void *)ah, m->m_len);
1392 	ah->ar_pro = htons(ETHERTYPE_IP);
1393 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1394 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1395 	ah->ar_op = htons(ARPOP_REVREQUEST);
1396 
1397 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1398 	tha = ar_tha(ah);
1399 	KASSERT(tha);
1400 	bcopy(CLLADDR(ifp->if_sadl), tha, ah->ar_hln);
1401 
1402 	sa.sa_family = AF_ARP;
1403 	sa.sa_len = 2;
1404 	m->m_flags |= M_BCAST;
1405 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1406 
1407 }
1408 
1409 /*
1410  * RARP for the ip address of the specified interface, but also
1411  * save the ip address of the server that sent the answer.
1412  * Timeout if no response is received.
1413  */
1414 int
1415 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1416     struct in_addr *clnt_in)
1417 {
1418 	int result, count = 20;
1419 
1420 	myip_initialized = 0;
1421 	myip_ifp = ifp;
1422 
1423 	revarp_in_progress = 1;
1424 	while (count--) {
1425 		revarprequest(ifp);
1426 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1427 		if (result != EWOULDBLOCK)
1428 			break;
1429 	}
1430 	revarp_in_progress = 0;
1431 
1432 	if (!myip_initialized)
1433 		return ENETUNREACH;
1434 
1435 	bcopy((void *)&srv_ip, serv_in, sizeof(*serv_in));
1436 	bcopy((void *)&myip, clnt_in, sizeof(*clnt_in));
1437 	return 0;
1438 }
1439 
1440 
1441 
1442 #ifdef DDB
1443 
1444 #include <machine/db_machdep.h>
1445 #include <ddb/db_interface.h>
1446 #include <ddb/db_output.h>
1447 
1448 static void
1449 db_print_sa(const struct sockaddr *sa)
1450 {
1451 	int len;
1452 	const u_char *p;
1453 
1454 	if (sa == 0) {
1455 		db_printf("[NULL]");
1456 		return;
1457 	}
1458 
1459 	p = (const u_char *)sa;
1460 	len = sa->sa_len;
1461 	db_printf("[");
1462 	while (len > 0) {
1463 		db_printf("%d", *p);
1464 		p++; len--;
1465 		if (len) db_printf(",");
1466 	}
1467 	db_printf("]\n");
1468 }
1469 
1470 static void
1471 db_print_ifa(struct ifaddr *ifa)
1472 {
1473 	if (ifa == 0)
1474 		return;
1475 	db_printf("  ifa_addr=");
1476 	db_print_sa(ifa->ifa_addr);
1477 	db_printf("  ifa_dsta=");
1478 	db_print_sa(ifa->ifa_dstaddr);
1479 	db_printf("  ifa_mask=");
1480 	db_print_sa(ifa->ifa_netmask);
1481 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1482 			  ifa->ifa_flags,
1483 			  ifa->ifa_refcnt,
1484 			  ifa->ifa_metric);
1485 }
1486 
1487 static void
1488 db_print_llinfo(void *li)
1489 {
1490 	struct llinfo_arp *la;
1491 
1492 	if (li == 0)
1493 		return;
1494 	la = (struct llinfo_arp *)li;
1495 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1496 			  la->la_rt, la->la_hold, la->la_asked);
1497 }
1498 
1499 /*
1500  * Function to pass to rt_walktree().
1501  * Return non-zero error to abort walk.
1502  */
1503 static int
1504 db_show_rtentry(struct rtentry *rt, void *w)
1505 {
1506 	db_printf("rtentry=%p", rt);
1507 
1508 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1509 			  rt->rt_flags, rt->rt_refcnt,
1510 			  rt->rt_use, rt->rt_expire);
1511 
1512 	db_printf(" key="); db_print_sa(rt_getkey(rt));
1513 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1514 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1515 
1516 	db_printf(" ifp=%p ", rt->rt_ifp);
1517 	if (rt->rt_ifp)
1518 		db_printf("(%s)", rt->rt_ifp->if_xname);
1519 	else
1520 		db_printf("(NULL)");
1521 
1522 	db_printf(" ifa=%p\n", rt->rt_ifa);
1523 	db_print_ifa(rt->rt_ifa);
1524 
1525 	db_printf(" gwroute=%p llinfo=%p\n",
1526 			  rt->rt_gwroute, rt->rt_llinfo);
1527 	db_print_llinfo(rt->rt_llinfo);
1528 
1529 	return (0);
1530 }
1531 
1532 /*
1533  * Function to print all the route trees.
1534  * Use this from ddb:  "show arptab"
1535  */
1536 void
1537 db_show_arptab(db_expr_t addr, bool have_addr,
1538     db_expr_t count, const char *modif)
1539 {
1540 	rt_walktree(AF_INET, db_show_rtentry, NULL);
1541 }
1542 #endif
1543 
1544 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1545 {
1546 	const struct sysctlnode *node;
1547 
1548 	sysctl_createv(clog, 0, NULL, NULL,
1549 			CTLFLAG_PERMANENT,
1550 			CTLTYPE_NODE, "net", NULL,
1551 			NULL, 0, NULL, 0,
1552 			CTL_NET, CTL_EOL);
1553 	sysctl_createv(clog, 0, NULL, NULL,
1554 			CTLFLAG_PERMANENT,
1555 			CTLTYPE_NODE, "inet", NULL,
1556 			NULL, 0, NULL, 0,
1557 			CTL_NET, PF_INET, CTL_EOL);
1558 	sysctl_createv(clog, 0, NULL, &node,
1559 			CTLFLAG_PERMANENT,
1560 			CTLTYPE_NODE, "arp",
1561 			SYSCTL_DESCR("Address Resolution Protocol"),
1562 			NULL, 0, NULL, 0,
1563 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1564 
1565 	sysctl_createv(clog, 0, NULL, NULL,
1566 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1567 			CTLTYPE_INT, "prune",
1568 			SYSCTL_DESCR("ARP cache pruning interval"),
1569 			NULL, 0, &arpt_prune, 0,
1570 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1571 
1572 	sysctl_createv(clog, 0, NULL, NULL,
1573 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1574 			CTLTYPE_INT, "keep",
1575 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1576 			NULL, 0, &arpt_keep, 0,
1577 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1578 
1579 	sysctl_createv(clog, 0, NULL, NULL,
1580 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1581 			CTLTYPE_INT, "down",
1582 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1583 			NULL, 0, &arpt_down, 0,
1584 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1585 
1586 	sysctl_createv(clog, 0, NULL, NULL,
1587 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1588 			CTLTYPE_INT, "refresh",
1589 			SYSCTL_DESCR("ARP entry refresh interval"),
1590 			NULL, 0, &arpt_refresh, 0,
1591 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1592 }
1593 
1594 #endif /* INET */
1595