xref: /netbsd-src/sys/netinet/if_arp.c (revision 89c5a767f8fc7a4633b2d409966e2becbb98ff92)
1 /*	$NetBSD: if_arp.c,v 1.66 1999/09/25 17:49:29 is Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998 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. All advertising materials mentioning features or use of this software
53  *    must display the following acknowledgement:
54  *	This product includes software developed by the University of
55  *	California, Berkeley and its contributors.
56  * 4. Neither the name of the University nor the names of its contributors
57  *    may be used to endorse or promote products derived from this software
58  *    without specific prior written permission.
59  *
60  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70  * SUCH DAMAGE.
71  *
72  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
73  */
74 
75 /*
76  * Ethernet address resolution protocol.
77  * TODO:
78  *	add "inuse/lock" bit (or ref. count) along with valid bit
79  */
80 
81 #include "opt_ddb.h"
82 #include "opt_inet.h"
83 
84 #ifdef INET
85 
86 #include <sys/param.h>
87 #include <sys/systm.h>
88 #include <sys/malloc.h>
89 #include <sys/mbuf.h>
90 #include <sys/socket.h>
91 #include <sys/time.h>
92 #include <sys/kernel.h>
93 #include <sys/errno.h>
94 #include <sys/ioctl.h>
95 #include <sys/syslog.h>
96 #include <sys/proc.h>
97 #include <sys/protosw.h>
98 #include <sys/domain.h>
99 
100 #include <net/ethertypes.h>
101 #include <net/if.h>
102 #include <net/if_dl.h>
103 #include <net/if_token.h>
104 #include <net/if_types.h>
105 #include <net/route.h>
106 
107 
108 #include <netinet/in.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/if_inarp.h>
113 
114 #include "loop.h"
115 #include "arc.h"
116 #if NARC > 0
117 #include <net/if_arc.h>
118 #endif
119 #include "fddi.h"
120 #if NFDDI > 0
121 #include <net/if_fddi.h>
122 #endif
123 #include "token.h"
124 #include "token.h"
125 
126 #define SIN(s) ((struct sockaddr_in *)s)
127 #define SDL(s) ((struct sockaddr_dl *)s)
128 #define SRP(s) ((struct sockaddr_inarp *)s)
129 
130 /*
131  * ARP trailer negotiation.  Trailer protocol is not IP specific,
132  * but ARP request/response use IP addresses.
133  */
134 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
135 
136 /* timer values */
137 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
138 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
139 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
140 #define	rt_expire rt_rmx.rmx_expire
141 
142 static	void arprequest __P((struct ifnet *,
143 	    struct in_addr *, struct in_addr *, u_int8_t *));
144 static	void arptfree __P((struct llinfo_arp *));
145 static	void arptimer __P((void *));
146 static	struct llinfo_arp *arplookup __P((struct in_addr *, int, int));
147 static	void in_arpinput __P((struct mbuf *));
148 
149 #if NLOOP > 0
150 extern	struct ifnet loif[NLOOP];
151 #endif
152 LIST_HEAD(, llinfo_arp) llinfo_arp;
153 struct	ifqueue arpintrq = {0, 0, 0, 50};
154 int	arp_inuse, arp_allocated, arp_intimer;
155 int	arp_maxtries = 5;
156 int	useloopback = 1;	/* use loopback interface for local traffic */
157 int	arpinit_done = 0;
158 
159 /* revarp state */
160 static struct	in_addr myip, srv_ip;
161 static int	myip_initialized = 0;
162 static int	revarp_in_progress = 0;
163 static struct	ifnet *myip_ifp = NULL;
164 
165 #ifdef DDB
166 static void db_print_sa __P((struct sockaddr *));
167 static void db_print_ifa __P((struct ifaddr *));
168 static void db_print_llinfo __P((caddr_t));
169 static int db_show_radix_node __P((struct radix_node *, void *));
170 #endif
171 
172 /*
173  * this should be elsewhere.
174  */
175 
176 static char *
177 lla_snprintf __P((u_int8_t *, int));
178 
179 static char *
180 lla_snprintf(adrp, len)
181 	u_int8_t *adrp;
182 	int len;
183 {
184 	static char buf[16*3];
185 	static const char hexdigits[] = {
186 	    '0','1','2','3','4','5','6','7',
187 	    '8','9','a','b','c','d','e','f'
188 	};
189 
190 	int i;
191 	char *p;
192 
193 	p = buf;
194 
195 	*p++ = hexdigits[(*adrp)>>4];
196 	*p++ = hexdigits[(*adrp++)&0xf];
197 
198 	for (i=1; i<len && i<16; i++) {
199 		*p++ = ':';
200 		*p++ = hexdigits[(*adrp)>>4];
201 		*p++ = hexdigits[(*adrp++)&0xf];
202 	}
203 
204 	*p = 0;
205 	return buf;
206 }
207 
208 struct protosw arpsw[] = {
209 	{ 0, 0, 0, 0,
210 	  0, 0, 0, 0,
211 	  0,
212 	  0, 0, 0, arp_drain,
213 	}
214 };
215 
216 
217 struct domain arpdomain =
218 { 	PF_ARP,  "arp", 0, 0, 0,
219 	arpsw, &arpsw[sizeof(arpsw)/sizeof(arpsw[0])]
220 };
221 
222 /*
223  * ARP table locking.
224  *
225  * to prevent lossage vs. the arp_drain routine (which may be called at
226  * any time, including in a device driver context), we do two things:
227  *
228  * 1) manipulation of la->la_hold is done at splimp() (for all of
229  * about two instructions).
230  *
231  * 2) manipulation of the arp table's linked list is done under the
232  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
233  * while the arp table is locked, we punt and try again later.
234  */
235 
236 int	arp_locked;
237 
238 static __inline int arp_lock_try __P((int));
239 static __inline void arp_unlock __P((void));
240 
241 static __inline int
242 arp_lock_try(int recurse)
243 {
244 	int s;
245 
246 	s = splimp();
247 	if (!recurse && arp_locked) {
248 		splx(s);
249 		return (0);
250 	}
251 	arp_locked++;
252 	splx(s);
253 	return (1);
254 }
255 
256 static __inline void
257 arp_unlock()
258 {
259 	int s;
260 
261 	s = splimp();
262 	arp_locked--;
263 	splx(s);
264 }
265 
266 #ifdef DIAGNOSTIC
267 #define	ARP_LOCK(recurse)						\
268 do {									\
269 	if (arp_lock_try(recurse) == 0) {				\
270 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
271 		panic("arp_lock");					\
272 	}								\
273 } while (0)
274 #define	ARP_LOCK_CHECK()						\
275 do {									\
276 	if (arp_locked == 0) {						\
277 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
278 		panic("arp lock check");				\
279 	}								\
280 } while (0)
281 #else
282 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
283 #define	ARP_LOCK_CHECK()	/* nothing */
284 #endif
285 
286 #define	ARP_UNLOCK()		arp_unlock()
287 
288 /*
289  * ARP protocol drain routine.  Called when memory is in short supply.
290  * Called at splimp();
291  */
292 
293 void
294 arp_drain()
295 {
296 	register struct llinfo_arp *la, *nla;
297 	int count = 0;
298 	struct mbuf *mold;
299 
300 	if (arp_lock_try(0) == 0) {
301 		printf("arp_drain: locked; punting\n");
302 		return;
303 	}
304 
305 	for (la = llinfo_arp.lh_first; la != 0; la = nla) {
306 		nla = la->la_list.le_next;
307 
308 		mold = la->la_hold;
309 		la->la_hold = 0;
310 
311 		if (mold) {
312 			m_freem(mold);
313 			count++;
314 		}
315 	}
316 	ARP_UNLOCK();
317 }
318 
319 
320 /*
321  * Timeout routine.  Age arp_tab entries periodically.
322  */
323 /* ARGSUSED */
324 static void
325 arptimer(arg)
326 	void *arg;
327 {
328 	int s;
329 	register struct llinfo_arp *la, *nla;
330 
331 	s = splsoftnet();
332 
333 	if (arp_lock_try(0) == 0) {
334 		/* get it later.. */
335 		splx(s);
336 		return;
337 	}
338 
339 	timeout(arptimer, NULL, arpt_prune * hz);
340 	for (la = llinfo_arp.lh_first; la != 0; la = nla) {
341 		register struct rtentry *rt = la->la_rt;
342 
343 		nla = la->la_list.le_next;
344 		if (rt->rt_expire && rt->rt_expire <= time.tv_sec)
345 			arptfree(la); /* timer has expired; clear */
346 	}
347 
348 	ARP_UNLOCK();
349 
350 	splx(s);
351 }
352 
353 /*
354  * Parallel to llc_rtrequest.
355  */
356 void
357 arp_rtrequest(req, rt, sa)
358 	int req;
359 	register struct rtentry *rt;
360 	struct sockaddr *sa;
361 {
362 	register struct sockaddr *gate = rt->rt_gateway;
363 	register struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
364 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
365 	size_t allocsize;
366 	struct mbuf *mold;
367 	int s;
368 
369 	if (!arpinit_done) {
370 		arpinit_done = 1;
371 		/*
372 		 * We generate expiration times from time.tv_sec
373 		 * so avoid accidently creating permanent routes.
374 		 */
375 		if (time.tv_sec == 0) {
376 			time.tv_sec++;
377 		}
378 		timeout(arptimer, (caddr_t)0, hz);
379 	}
380 	if (rt->rt_flags & RTF_GATEWAY)
381 		return;
382 
383 	ARP_LOCK(1);		/* we may already be locked here. */
384 
385 	switch (req) {
386 
387 	case RTM_ADD:
388 		/*
389 		 * XXX: If this is a manually added route to interface
390 		 * such as older version of routed or gated might provide,
391 		 * restore cloning bit.
392 		 */
393 		if ((rt->rt_flags & RTF_HOST) == 0 &&
394 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
395 			rt->rt_flags |= RTF_CLONING;
396 		if (rt->rt_flags & RTF_CLONING) {
397 			/*
398 			 * Case 1: This route should come from a route to iface.
399 			 */
400 			rt_setgate(rt, rt_key(rt),
401 					(struct sockaddr *)&null_sdl);
402 			gate = rt->rt_gateway;
403 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
404 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
405 			/*
406 			 * Give this route an expiration time, even though
407 			 * it's a "permanent" route, so that routes cloned
408 			 * from it do not need their expiration time set.
409 			 */
410 			rt->rt_expire = time.tv_sec;
411 #if NFDDI > 0
412 			if (rt->rt_ifp->if_type == IFT_FDDI
413 			    && (rt->rt_rmx.rmx_mtu > FDDIIPMTU
414 				|| (rt->rt_rmx.rmx_mtu == 0
415 				    && rt->rt_ifp->if_mtu > FDDIIPMTU))) {
416 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
417 			}
418 #endif
419 #if NARC > 0
420 			if (rt->rt_ifp->if_type == IFT_ARCNET) {
421 				int arcipifmtu;
422 
423 				if (rt->rt_ifp->if_flags & IFF_LINK0)
424 					arcipifmtu = arc_ipmtu;
425 				else
426 					arcipifmtu = ARCMTU;
427 
428 			    	if (rt->rt_rmx.rmx_mtu > arcipifmtu ||
429 				    (rt->rt_rmx.rmx_mtu == 0 &&
430 				     rt->rt_ifp->if_mtu > arcipifmtu))
431 
432 					rt->rt_rmx.rmx_mtu = arcipifmtu;
433 			}
434 #endif
435 			break;
436 		}
437 		/* Announce a new entry if requested. */
438 		if (rt->rt_flags & RTF_ANNOUNCE)
439 			arprequest(rt->rt_ifp,
440 			    &SIN(rt_key(rt))->sin_addr,
441 			    &SIN(rt_key(rt))->sin_addr,
442 			    (u_char *)LLADDR(SDL(gate)));
443 		/*FALLTHROUGH*/
444 	case RTM_RESOLVE:
445 		if (gate->sa_family != AF_LINK ||
446 		    gate->sa_len < sizeof(null_sdl)) {
447 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
448 			break;
449 		}
450 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
451 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
452 		if (la != 0)
453 			break; /* This happens on a route change */
454 		/*
455 		 * Case 2:  This route may come from cloning, or a manual route
456 		 * add with a LL address.
457 		 */
458 		switch (SDL(gate)->sdl_type) {
459 #if NTOKEN > 0
460 		case IFT_ISO88025:
461 			allocsize = sizeof(*la) + sizeof(struct token_rif);
462 			break;
463 #endif /* NTOKEN > 0 */
464 		default:
465 			allocsize = sizeof(*la);
466 		}
467 		R_Malloc(la, struct llinfo_arp *, allocsize);
468 		rt->rt_llinfo = (caddr_t)la;
469 		if (la == 0) {
470 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
471 			break;
472 		}
473 		arp_inuse++, arp_allocated++;
474 		Bzero(la, allocsize);
475 		la->la_rt = rt;
476 		rt->rt_flags |= RTF_LLINFO;
477 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
478 		if (in_hosteq(SIN(rt_key(rt))->sin_addr,
479 		    (IA_SIN(rt->rt_ifa))->sin_addr)) {
480 			/*
481 			 * This test used to be
482 			 *	if (loif.if_flags & IFF_UP)
483 			 * It allowed local traffic to be forced through
484 			 * the hardware by configuring the loopback down.
485 			 * However, it causes problems during network
486 			 * configuration for boards that can't receive
487 			 * packets they send.  It is now necessary to clear
488 			 * "useloopback" and remove the route to force
489 			 * traffic out to the hardware.
490 			 */
491 			rt->rt_expire = 0;
492 			Bcopy(LLADDR(rt->rt_ifp->if_sadl),
493 			    LLADDR(SDL(gate)),
494 			    SDL(gate)->sdl_alen =
495 			    rt->rt_ifp->if_data.ifi_addrlen);
496 #if NLOOP > 0
497 			if (useloopback)
498 				rt->rt_ifp = &loif[0];
499 #endif
500 		}
501 		break;
502 
503 	case RTM_DELETE:
504 		if (la == 0)
505 			break;
506 		arp_inuse--;
507 		LIST_REMOVE(la, la_list);
508 		rt->rt_llinfo = 0;
509 		rt->rt_flags &= ~RTF_LLINFO;
510 
511 		s = splimp();
512 		mold = la->la_hold;
513 		la->la_hold = 0;
514 		splx(s);
515 
516 		if (mold)
517 			m_freem(mold);
518 
519 		Free((caddr_t)la);
520 	}
521 	ARP_UNLOCK();
522 }
523 
524 /*
525  * Broadcast an ARP request. Caller specifies:
526  *	- arp header source ip address
527  *	- arp header target ip address
528  *	- arp header source ethernet address
529  */
530 static void
531 arprequest(ifp, sip, tip, enaddr)
532 	register struct ifnet *ifp;
533 	register struct in_addr *sip, *tip;
534 	register u_int8_t *enaddr;
535 {
536 	register struct mbuf *m;
537 	struct arphdr *ah;
538 	struct sockaddr sa;
539 
540 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
541 		return;
542 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
543 	    2*ifp->if_data.ifi_addrlen;
544 	m->m_pkthdr.len = m->m_len;
545 	MH_ALIGN(m, m->m_len);
546 	ah = mtod(m, struct arphdr *);
547 	bzero((caddr_t)ah, m->m_len);
548 	ah->ar_pro = htons(ETHERTYPE_IP);
549 	ah->ar_hln = ifp->if_data.ifi_addrlen;	/* hardware address length */
550 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
551 	ah->ar_op = htons(ARPOP_REQUEST);
552 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
553 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
554 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
555 	sa.sa_family = AF_ARP;
556 	sa.sa_len = 2;
557 	m->m_flags |= M_BCAST;
558 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
559 }
560 
561 /*
562  * Resolve an IP address into an ethernet address.  If success,
563  * desten is filled in.  If there is no entry in arptab,
564  * set one up and broadcast a request for the IP address.
565  * Hold onto this mbuf and resend it once the address
566  * is finally resolved.  A return value of 1 indicates
567  * that desten has been filled in and the packet should be sent
568  * normally; a 0 return indicates that the packet has been
569  * taken over here, either now or for later transmission.
570  */
571 int
572 arpresolve(ifp, rt, m, dst, desten)
573 	register struct ifnet *ifp;
574 	register struct rtentry *rt;
575 	struct mbuf *m;
576 	register struct sockaddr *dst;
577 	register u_char *desten;
578 {
579 	register struct llinfo_arp *la;
580 	struct sockaddr_dl *sdl;
581 	struct mbuf *mold;
582 	int s;
583 
584 	if (rt)
585 		la = (struct llinfo_arp *)rt->rt_llinfo;
586 	else {
587 		if ((la = arplookup(&SIN(dst)->sin_addr, 1, 0)) != NULL)
588 			rt = la->la_rt;
589 	}
590 	if (la == 0 || rt == 0) {
591 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo\n");
592 		m_freem(m);
593 		return (0);
594 	}
595 	sdl = SDL(rt->rt_gateway);
596 	/*
597 	 * Check the address family and length is valid, the address
598 	 * is resolved; otherwise, try to resolve.
599 	 */
600 	if ((rt->rt_expire == 0 || rt->rt_expire > time.tv_sec) &&
601 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
602 		bcopy(LLADDR(sdl), desten,
603 		    min(sdl->sdl_alen, ifp->if_data.ifi_addrlen));
604 		return 1;
605 	}
606 	/*
607 	 * There is an arptab entry, but no ethernet address
608 	 * response yet.  Replace the held mbuf with this
609 	 * latest one.
610 	 */
611 
612 	s = splimp();
613 	mold = la->la_hold;
614 	la->la_hold = m;
615 	splx(s);
616 
617 	if (mold)
618 		m_freem(mold);
619 
620 
621 	/*
622 	 * Re-send the ARP request when appropriate.
623 	 */
624 #ifdef	DIAGNOSTIC
625 	if (rt->rt_expire == 0) {
626 		/* This should never happen. (Should it? -gwr) */
627 		printf("arpresolve: unresolved and rt_expire == 0\n");
628 		/* Set expiration time to now (expired). */
629 		rt->rt_expire = time.tv_sec;
630 	}
631 #endif
632 	if (rt->rt_expire) {
633 		rt->rt_flags &= ~RTF_REJECT;
634 		if (la->la_asked == 0 || rt->rt_expire != time.tv_sec) {
635 			rt->rt_expire = time.tv_sec;
636 			if (la->la_asked++ < arp_maxtries)
637 				arprequest(ifp,
638 				    &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
639 				    &SIN(dst)->sin_addr,
640 				    LLADDR(ifp->if_sadl));
641 			else {
642 				rt->rt_flags |= RTF_REJECT;
643 				rt->rt_expire += arpt_down;
644 				la->la_asked = 0;
645 			}
646 		}
647 	}
648 	return (0);
649 }
650 
651 /*
652  * Common length and type checks are done here,
653  * then the protocol-specific routine is called.
654  */
655 void
656 arpintr()
657 {
658 	register struct mbuf *m;
659 	register struct arphdr *ar;
660 	int s;
661 
662 	while (arpintrq.ifq_head) {
663 		s = splimp();
664 		IF_DEQUEUE(&arpintrq, m);
665 		splx(s);
666 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
667 			panic("arpintr");
668 
669 		if (m->m_len >= sizeof(struct arphdr) &&
670 		    (ar = mtod(m, struct arphdr *)) &&
671 		    /* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
672 		    m->m_len >=
673 		      sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
674 			switch (ntohs(ar->ar_pro)) {
675 
676 			case ETHERTYPE_IP:
677 			case ETHERTYPE_IPTRAILERS:
678 				in_arpinput(m);
679 				continue;
680 			}
681 		m_freem(m);
682 	}
683 }
684 
685 /*
686  * ARP for Internet protocols on 10 Mb/s Ethernet.
687  * Algorithm is that given in RFC 826.
688  * In addition, a sanity check is performed on the sender
689  * protocol address, to catch impersonators.
690  * We no longer handle negotiations for use of trailer protocol:
691  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
692  * along with IP replies if we wanted trailers sent to us,
693  * and also sent them in response to IP replies.
694  * This allowed either end to announce the desire to receive
695  * trailer packets.
696  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
697  * but formerly didn't normally send requests.
698  */
699 static void
700 in_arpinput(m)
701 	struct mbuf *m;
702 {
703 	struct arphdr *ah;
704 	register struct ifnet *ifp = m->m_pkthdr.rcvif;
705 	register struct llinfo_arp *la = 0;
706 	register struct rtentry  *rt;
707 	struct in_ifaddr *ia;
708 	struct sockaddr_dl *sdl;
709 	struct sockaddr sa;
710 	struct in_addr isaddr, itaddr, myaddr;
711 	int op;
712 	struct mbuf *mold;
713 	int s;
714 
715 
716 	ah = mtod(m, struct arphdr *);
717 	op = ntohs(ah->ar_op);
718 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&isaddr, sizeof (isaddr));
719 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&itaddr, sizeof (itaddr));
720 
721 	/*
722 	 * If the target IP address is zero, ignore the packet.
723 	 * This prevents the code below from tring to answer
724 	 * when we are using IP address zero (booting).
725 	 */
726 	if (in_nullhost(itaddr))
727 		goto out;
728 
729 	/*
730 	 * If the source IP address is zero, this is most likely a
731 	 * confused host trying to use IP address zero. (Windoze?)
732 	 * XXX: Should we bother trying to reply to these?
733 	 */
734 	if (in_nullhost(isaddr))
735 		goto out;
736 
737 	/*
738 	 * Search for a matching interface address
739 	 * or any address on the interface to use
740 	 * as a dummy address in the rest of this function
741 	 */
742 	INADDR_TO_IA(itaddr, ia);
743 	while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
744 		NEXT_IA_WITH_SAME_ADDR(ia);
745 
746 	if (ia == NULL) {
747 		INADDR_TO_IA(isaddr, ia);
748 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
749 			NEXT_IA_WITH_SAME_ADDR(ia);
750 
751 		if (ia == NULL) {
752 			IFP_TO_IA(ifp, ia);
753 			if (ia == NULL)
754 				goto out;
755 		}
756 	}
757 
758 	myaddr = ia->ia_addr.sin_addr;
759 
760 	if (!bcmp((caddr_t)ar_sha(ah), LLADDR(ifp->if_sadl),
761 	    ifp->if_data.ifi_addrlen))
762 		goto out;	/* it's from me, ignore it. */
763 
764 	if (!bcmp((caddr_t)ar_sha(ah), (caddr_t)ifp->if_broadcastaddr,
765 	    ifp->if_data.ifi_addrlen)) {
766 		log(LOG_ERR,
767 		    "%s: arp: link address is broadcast for IP address %s!\n",
768 		    ifp->if_xname, in_fmtaddr(isaddr));
769 		goto out;
770 	}
771 
772 	if (in_hosteq(isaddr, myaddr)) {
773 		log(LOG_ERR,
774 		   "duplicate IP address %s sent from link address %s\n",
775 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
776 		itaddr = myaddr;
777 		goto reply;
778 	}
779 	la = arplookup(&isaddr, in_hosteq(itaddr, myaddr), 0);
780 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
781 		if (sdl->sdl_alen &&
782 		    bcmp((caddr_t)ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
783 			if (rt->rt_flags & RTF_STATIC) {
784 				log(LOG_INFO,
785 				    "%s tried to overwrite permanent arp info"
786 				    " for %s\n",
787 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
788 				    in_fmtaddr(isaddr));
789 				goto out;
790 			} else if (rt->rt_ifp != ifp) {
791 				log(LOG_INFO,
792 				    "%s on %s tried to overwrite "
793 				    "arp info for %s on %s\n",
794 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
795 				    ifp->if_xname, in_fmtaddr(isaddr),
796 				    rt->rt_ifp->if_xname);
797 				    goto out;
798 			} else {
799 				log(LOG_INFO,
800 				    "arp info overwritten for %s by %s\n",
801 				    in_fmtaddr(isaddr),
802 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
803 			}
804 		}
805 		/*
806 		 * sanity check for the address length.
807 		 * XXX this does not work for protocols with variable address
808 		 * length. -is
809 		 */
810 		if (sdl->sdl_alen &&
811 		    sdl->sdl_alen != ah->ar_hln) {
812 			log(LOG_WARNING,
813 			    "arp from %s: new addr len %d, was %d",
814 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
815 		}
816 		if (ifp->if_data.ifi_addrlen != ah->ar_hln) {
817 			log(LOG_WARNING,
818 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
819 			    in_fmtaddr(isaddr), ah->ar_hln,
820 			    ifp->if_data.ifi_addrlen);
821 			goto reply;
822 		}
823 #if NTOKEN > 0
824 		/*
825 		 * XXX uses m_data and assumes the complete answer including
826 		 * XXX token-ring headers is in the same buf
827 		 */
828 		if (ifp->if_type == IFT_ISO88025) {
829 			struct token_header *trh;
830 
831 			trh = (struct token_header *)M_TRHSTART(m);
832 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
833 				struct token_rif	*rif;
834 				size_t	riflen;
835 
836 				rif = TOKEN_RIF(trh);
837 				riflen = (ntohs(rif->tr_rcf) &
838 				    TOKEN_RCF_LEN_MASK) >> 8;
839 
840 				if (riflen > 2 &&
841 				    riflen < sizeof(struct token_rif) &&
842 				    (riflen & 1) == 0) {
843 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
844 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
845 					bcopy(rif, TOKEN_RIF(la), riflen);
846 				}
847 			}
848 		}
849 #endif /* NTOKEN > 0 */
850 		bcopy((caddr_t)ar_sha(ah), LLADDR(sdl),
851 		    sdl->sdl_alen = ah->ar_hln);
852 		if (rt->rt_expire)
853 			rt->rt_expire = time.tv_sec + arpt_keep;
854 		rt->rt_flags &= ~RTF_REJECT;
855 		la->la_asked = 0;
856 
857 		s = splimp();
858 		mold = la->la_hold;
859 		la->la_hold = 0;
860 		splx(s);
861 
862 		if (mold)
863 			(*ifp->if_output)(ifp, mold, rt_key(rt), rt);
864 	}
865 reply:
866 	if (op != ARPOP_REQUEST) {
867 	out:
868 		m_freem(m);
869 		return;
870 	}
871 	if (in_hosteq(itaddr, myaddr)) {
872 		/* I am the target */
873 		bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah), ah->ar_hln);
874 		bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
875 	} else {
876 		la = arplookup(&itaddr, 0, SIN_PROXY);
877 		if (la == 0)
878 			goto out;
879 		rt = la->la_rt;
880 		bcopy((caddr_t)ar_sha(ah), (caddr_t)ar_tha(ah), ah->ar_hln);
881 		sdl = SDL(rt->rt_gateway);
882 		bcopy(LLADDR(sdl), (caddr_t)ar_sha(ah), ah->ar_hln);
883 	}
884 
885 	bcopy((caddr_t)ar_spa(ah), (caddr_t)ar_tpa(ah), ah->ar_pln);
886 	bcopy((caddr_t)&itaddr, (caddr_t)ar_spa(ah), ah->ar_pln);
887 	ah->ar_op = htons(ARPOP_REPLY);
888 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
889 	m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
890 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
891 	m->m_pkthdr.len = m->m_len;
892 	sa.sa_family = AF_ARP;
893 	sa.sa_len = 2;
894 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
895 	return;
896 }
897 
898 /*
899  * Free an arp entry.
900  */
901 static void
902 arptfree(la)
903 	register struct llinfo_arp *la;
904 {
905 	register struct rtentry *rt = la->la_rt;
906 	register struct sockaddr_dl *sdl;
907 
908 	ARP_LOCK_CHECK();
909 
910 	if (rt == 0)
911 		panic("arptfree");
912 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
913 	    sdl->sdl_family == AF_LINK) {
914 		sdl->sdl_alen = 0;
915 		la->la_asked = 0;
916 		rt->rt_flags &= ~RTF_REJECT;
917 		return;
918 	}
919 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
920 	    0, (struct rtentry **)0);
921 }
922 
923 /*
924  * Lookup or enter a new address in arptab.
925  */
926 static struct llinfo_arp *
927 arplookup(addr, create, proxy)
928 	struct in_addr *addr;
929 	int create, proxy;
930 {
931 	register struct rtentry *rt;
932 	static struct sockaddr_inarp sin;
933 	const char *why = 0;
934 
935 	sin.sin_len = sizeof(sin);
936 	sin.sin_family = AF_INET;
937 	sin.sin_addr = *addr;
938 	sin.sin_other = proxy ? SIN_PROXY : 0;
939 	rt = rtalloc1(sintosa(&sin), create);
940 	if (rt == 0)
941 		return (0);
942 	rt->rt_refcnt--;
943 
944 	if (rt->rt_flags & RTF_GATEWAY)
945 		why = "host is not on local network";
946 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
947 		why = "could not allocate llinfo";
948 	else if (rt->rt_gateway->sa_family != AF_LINK)
949 		why = "gateway route is not ours";
950 	else
951 		return ((struct llinfo_arp *)rt->rt_llinfo);
952 
953 	if (create)
954 		log(LOG_DEBUG, "arplookup: unable to enter address"
955 		    " for %s (%s)\n",
956 		    in_fmtaddr(*addr), why);
957 	return (0);
958 }
959 
960 int
961 arpioctl(cmd, data)
962 	u_long cmd;
963 	caddr_t data;
964 {
965 
966 	return (EOPNOTSUPP);
967 }
968 
969 void
970 arp_ifinit(ifp, ifa)
971 	struct ifnet *ifp;
972 	struct ifaddr *ifa;
973 {
974 	struct in_addr *ip;
975 
976 	/*
977 	 * Warn the user if another station has this IP address,
978 	 * but only if the interface IP address is not zero.
979 	 */
980 	ip = &IA_SIN(ifa)->sin_addr;
981 	if (!in_nullhost(*ip))
982 		arprequest(ifp, ip, ip, LLADDR(ifp->if_sadl));
983 
984 	ifa->ifa_rtrequest = arp_rtrequest;
985 	ifa->ifa_flags |= RTF_CLONING;
986 }
987 
988 /*
989  * Called from 10 Mb/s Ethernet interrupt handlers
990  * when ether packet type ETHERTYPE_REVARP
991  * is received.  Common length and type checks are done here,
992  * then the protocol-specific routine is called.
993  */
994 void
995 revarpinput(m)
996 	struct mbuf *m;
997 {
998 	struct arphdr *ar;
999 
1000 	if (m->m_len < sizeof(struct arphdr))
1001 		goto out;
1002 	ar = mtod(m, struct arphdr *);
1003 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1004 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1005 		goto out;
1006 #endif
1007 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1008 		goto out;
1009 	switch (ntohs(ar->ar_pro)) {
1010 
1011 	case ETHERTYPE_IP:
1012 	case ETHERTYPE_IPTRAILERS:
1013 		in_revarpinput(m);
1014 		return;
1015 
1016 	default:
1017 		break;
1018 	}
1019 out:
1020 	m_freem(m);
1021 }
1022 
1023 /*
1024  * RARP for Internet protocols on 10 Mb/s Ethernet.
1025  * Algorithm is that given in RFC 903.
1026  * We are only using for bootstrap purposes to get an ip address for one of
1027  * our interfaces.  Thus we support no user-interface.
1028  *
1029  * Since the contents of the RARP reply are specific to the interface that
1030  * sent the request, this code must ensure that they are properly associated.
1031  *
1032  * Note: also supports ARP via RARP packets, per the RFC.
1033  */
1034 void
1035 in_revarpinput(m)
1036 	struct mbuf *m;
1037 {
1038 	struct ifnet *ifp;
1039 	struct arphdr *ah;
1040 	int op;
1041 
1042 	ah = mtod(m, struct arphdr *);
1043 	op = ntohs(ah->ar_op);
1044 	switch (op) {
1045 	case ARPOP_REQUEST:
1046 	case ARPOP_REPLY:	/* per RFC */
1047 		in_arpinput(m);
1048 		return;
1049 	case ARPOP_REVREPLY:
1050 		break;
1051 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1052 	default:
1053 		goto out;
1054 	}
1055 	if (!revarp_in_progress)
1056 		goto out;
1057 	ifp = m->m_pkthdr.rcvif;
1058 	if (ifp != myip_ifp) /* !same interface */
1059 		goto out;
1060 	if (myip_initialized)
1061 		goto wake;
1062 	if (bcmp(ar_tha(ah), LLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1063 		goto out;
1064 	bcopy((caddr_t)ar_spa(ah), (caddr_t)&srv_ip, sizeof(srv_ip));
1065 	bcopy((caddr_t)ar_tpa(ah), (caddr_t)&myip, sizeof(myip));
1066 	myip_initialized = 1;
1067 wake:	/* Do wakeup every time in case it was missed. */
1068 	wakeup((caddr_t)&myip);
1069 
1070 out:
1071 	m_freem(m);
1072 }
1073 
1074 /*
1075  * Send a RARP request for the ip address of the specified interface.
1076  * The request should be RFC 903-compliant.
1077  */
1078 void
1079 revarprequest(ifp)
1080 	struct ifnet *ifp;
1081 {
1082 	struct sockaddr sa;
1083 	struct mbuf *m;
1084 	struct arphdr *ah;
1085 
1086 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1087 		return;
1088 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1089 	    2*ifp->if_data.ifi_addrlen;
1090 	m->m_pkthdr.len = m->m_len;
1091 	MH_ALIGN(m, m->m_len);
1092 	ah = mtod(m, struct arphdr *);
1093 	bzero((caddr_t)ah, m->m_len);
1094 	ah->ar_pro = htons(ETHERTYPE_IP);
1095 	ah->ar_hln = ifp->if_data.ifi_addrlen;	/* hardware address length */
1096 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1097 	ah->ar_op = htons(ARPOP_REVREQUEST);
1098 
1099 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_sha(ah), ah->ar_hln);
1100 	bcopy(LLADDR(ifp->if_sadl), (caddr_t)ar_tha(ah), ah->ar_hln);
1101 
1102 	sa.sa_family = AF_ARP;
1103 	sa.sa_len = 2;
1104 	m->m_flags |= M_BCAST;
1105 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1106 
1107 }
1108 
1109 /*
1110  * RARP for the ip address of the specified interface, but also
1111  * save the ip address of the server that sent the answer.
1112  * Timeout if no response is received.
1113  */
1114 int
1115 revarpwhoarewe(ifp, serv_in, clnt_in)
1116 	struct ifnet *ifp;
1117 	struct in_addr *serv_in;
1118 	struct in_addr *clnt_in;
1119 {
1120 	int result, count = 20;
1121 
1122 	myip_initialized = 0;
1123 	myip_ifp = ifp;
1124 
1125 	revarp_in_progress = 1;
1126 	while (count--) {
1127 		revarprequest(ifp);
1128 		result = tsleep((caddr_t)&myip, PSOCK, "revarp", hz/2);
1129 		if (result != EWOULDBLOCK)
1130 			break;
1131 	}
1132 	revarp_in_progress = 0;
1133 
1134 	if (!myip_initialized)
1135 		return ENETUNREACH;
1136 
1137 	bcopy((caddr_t)&srv_ip, serv_in, sizeof(*serv_in));
1138 	bcopy((caddr_t)&myip, clnt_in, sizeof(*clnt_in));
1139 	return 0;
1140 }
1141 
1142 
1143 
1144 #ifdef DDB
1145 
1146 #include <machine/db_machdep.h>
1147 #include <ddb/db_interface.h>
1148 #include <ddb/db_output.h>
1149 static void
1150 db_print_sa(sa)
1151 	struct sockaddr *sa;
1152 {
1153 	int len;
1154 	u_char *p;
1155 
1156 	if (sa == 0) {
1157 		db_printf("[NULL]");
1158 		return;
1159 	}
1160 
1161 	p = (u_char*)sa;
1162 	len = sa->sa_len;
1163 	db_printf("[");
1164 	while (len > 0) {
1165 		db_printf("%d", *p);
1166 		p++; len--;
1167 		if (len) db_printf(",");
1168 	}
1169 	db_printf("]\n");
1170 }
1171 static void
1172 db_print_ifa(ifa)
1173 	struct ifaddr *ifa;
1174 {
1175 	if (ifa == 0)
1176 		return;
1177 	db_printf("  ifa_addr=");
1178 	db_print_sa(ifa->ifa_addr);
1179 	db_printf("  ifa_dsta=");
1180 	db_print_sa(ifa->ifa_dstaddr);
1181 	db_printf("  ifa_mask=");
1182 	db_print_sa(ifa->ifa_netmask);
1183 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1184 			  ifa->ifa_flags,
1185 			  ifa->ifa_refcnt,
1186 			  ifa->ifa_metric);
1187 }
1188 static void
1189 db_print_llinfo(li)
1190 	caddr_t li;
1191 {
1192 	struct llinfo_arp *la;
1193 
1194 	if (li == 0)
1195 		return;
1196 	la = (struct llinfo_arp *)li;
1197 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1198 			  la->la_rt, la->la_hold, la->la_asked);
1199 }
1200 /*
1201  * Function to pass to rn_walktree().
1202  * Return non-zero error to abort walk.
1203  */
1204 static int
1205 db_show_radix_node(rn, w)
1206 	struct radix_node *rn;
1207 	void *w;
1208 {
1209 	struct rtentry *rt = (struct rtentry *)rn;
1210 
1211 	db_printf("rtentry=%p", rt);
1212 
1213 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1214 			  rt->rt_flags, rt->rt_refcnt,
1215 			  rt->rt_use, rt->rt_expire);
1216 
1217 	db_printf(" key="); db_print_sa(rt_key(rt));
1218 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1219 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1220 
1221 	db_printf(" ifp=%p ", rt->rt_ifp);
1222 	if (rt->rt_ifp)
1223 		db_printf("(%s)", rt->rt_ifp->if_xname);
1224 	else
1225 		db_printf("(NULL)");
1226 
1227 	db_printf(" ifa=%p\n", rt->rt_ifa);
1228 	db_print_ifa(rt->rt_ifa);
1229 
1230 	db_printf(" genmask="); db_print_sa(rt->rt_genmask);
1231 
1232 	db_printf(" gwroute=%p llinfo=%p\n",
1233 			  rt->rt_gwroute, rt->rt_llinfo);
1234 	db_print_llinfo(rt->rt_llinfo);
1235 
1236 	return (0);
1237 }
1238 /*
1239  * Function to print all the route trees.
1240  * Use this from ddb:  "call db_show_arptab"
1241  */
1242 int
1243 db_show_arptab()
1244 {
1245 	struct radix_node_head *rnh;
1246 	rnh = rt_tables[AF_INET];
1247 	db_printf("Route tree for AF_INET\n");
1248 	if (rnh == NULL) {
1249 		db_printf(" (not initialized)\n");
1250 		return (0);
1251 	}
1252 	rn_walktree(rnh, db_show_radix_node, NULL);
1253 	return (0);
1254 }
1255 #endif
1256 #endif /* INET */
1257 
1258