xref: /netbsd-src/sys/netinet/if_arp.c (revision 11a6dbe72840351315e0652b2fc6663628c84cad)
1 /*	$NetBSD: if_arp.c,v 1.135 2008/04/28 20:24:09 martin Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998, 2000, 2008 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  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
62  */
63 
64 /*
65  * Ethernet address resolution protocol.
66  * TODO:
67  *	add "inuse/lock" bit (or ref. count) along with valid bit
68  */
69 
70 #include <sys/cdefs.h>
71 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.135 2008/04/28 20:24:09 martin Exp $");
72 
73 #include "opt_ddb.h"
74 #include "opt_inet.h"
75 
76 #ifdef INET
77 
78 #include "bridge.h"
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/callout.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/socket.h>
86 #include <sys/time.h>
87 #include <sys/timetc.h>
88 #include <sys/kernel.h>
89 #include <sys/errno.h>
90 #include <sys/ioctl.h>
91 #include <sys/syslog.h>
92 #include <sys/proc.h>
93 #include <sys/protosw.h>
94 #include <sys/domain.h>
95 #include <sys/sysctl.h>
96 #include <sys/socketvar.h>
97 #include <sys/percpu.h>
98 
99 #include <net/ethertypes.h>
100 #include <net/if.h>
101 #include <net/if_dl.h>
102 #include <net/if_token.h>
103 #include <net/if_types.h>
104 #include <net/if_ether.h>
105 #include <net/route.h>
106 #include <net/net_stats.h>
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 "arcnet.h"
115 #if NARCNET > 0
116 #include <net/if_arc.h>
117 #endif
118 #include "fddi.h"
119 #if NFDDI > 0
120 #include <net/if_fddi.h>
121 #endif
122 #include "token.h"
123 #include "carp.h"
124 #if NCARP > 0
125 #include <netinet/ip_carp.h>
126 #endif
127 
128 #define SIN(s) ((struct sockaddr_in *)s)
129 #define SRP(s) ((struct sockaddr_inarp *)s)
130 
131 /*
132  * ARP trailer negotiation.  Trailer protocol is not IP specific,
133  * but ARP request/response use IP addresses.
134  */
135 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
136 
137 /* timer values */
138 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
139 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
140 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
141 int	arpt_refresh = (5*60);	/* time left before refreshing */
142 #define	rt_expire rt_rmx.rmx_expire
143 #define	rt_pksent rt_rmx.rmx_pksent
144 
145 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
146 	    const struct sockaddr *);
147 static	void arptfree(struct llinfo_arp *);
148 static	void arptimer(void *);
149 static	struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *,
150 				      int, int, struct rtentry *);
151 static	struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *,
152 					  int, int);
153 static	void in_arpinput(struct mbuf *);
154 
155 LIST_HEAD(, llinfo_arp) llinfo_arp;
156 struct	ifqueue arpintrq = {
157 	.ifq_head = NULL,
158 	.ifq_tail = NULL,
159 	.ifq_len = 0,
160 	.ifq_maxlen = 50,
161 	.ifq_drops = 0,
162 };
163 int	arp_inuse, arp_allocated, arp_intimer;
164 int	arp_maxtries = 5;
165 int	useloopback = 1;	/* use loopback interface for local traffic */
166 int	arpinit_done = 0;
167 
168 static percpu_t *arpstat_percpu;
169 
170 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
171 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
172 
173 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
174 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
175 
176 struct	callout arptimer_ch;
177 
178 /* revarp state */
179 struct	in_addr myip, srv_ip;
180 int	myip_initialized = 0;
181 int	revarp_in_progress = 0;
182 struct	ifnet *myip_ifp = NULL;
183 
184 #ifdef DDB
185 static void db_print_sa(const struct sockaddr *);
186 static void db_print_ifa(struct ifaddr *);
187 static void db_print_llinfo(void *);
188 static int db_show_rtentry(struct rtentry *, void *);
189 #endif
190 
191 /*
192  * this should be elsewhere.
193  */
194 
195 static char *
196 lla_snprintf(u_int8_t *, int);
197 
198 static char *
199 lla_snprintf(u_int8_t *adrp, int len)
200 {
201 #define NUMBUFS 3
202 	static char buf[NUMBUFS][16*3];
203 	static int bnum = 0;
204 
205 	int i;
206 	char *p;
207 
208 	p = buf[bnum];
209 
210 	*p++ = hexdigits[(*adrp)>>4];
211 	*p++ = hexdigits[(*adrp++)&0xf];
212 
213 	for (i=1; i<len && i<16; i++) {
214 		*p++ = ':';
215 		*p++ = hexdigits[(*adrp)>>4];
216 		*p++ = hexdigits[(*adrp++)&0xf];
217 	}
218 
219 	*p = 0;
220 	p = buf[bnum];
221 	bnum = (bnum + 1) % NUMBUFS;
222 	return p;
223 }
224 
225 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
226 
227 const struct protosw arpsw[] = {
228 	{ .pr_type = 0,
229 	  .pr_domain = &arpdomain,
230 	  .pr_protocol = 0,
231 	  .pr_flags = 0,
232 	  .pr_input = 0,
233 	  .pr_output = 0,
234 	  .pr_ctlinput = 0,
235 	  .pr_ctloutput = 0,
236 	  .pr_usrreq =  0,
237 	  .pr_init = arp_init,
238 	  .pr_fasttimo = 0,
239 	  .pr_slowtimo = 0,
240 	  .pr_drain = arp_drain,
241 	}
242 };
243 
244 
245 struct domain arpdomain = {
246 	.dom_family = PF_ARP,
247 	.dom_name = "arp",
248 	.dom_protosw = arpsw,
249 	.dom_protoswNPROTOSW = &arpsw[sizeof(arpsw)/sizeof(arpsw[0])],
250 };
251 
252 /*
253  * ARP table locking.
254  *
255  * to prevent lossage vs. the arp_drain routine (which may be called at
256  * any time, including in a device driver context), we do two things:
257  *
258  * 1) manipulation of la->la_hold is done at splnet() (for all of
259  * about two instructions).
260  *
261  * 2) manipulation of the arp table's linked list is done under the
262  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
263  * while the arp table is locked, we punt and try again later.
264  */
265 
266 static int	arp_locked;
267 static inline int arp_lock_try(int);
268 static inline void arp_unlock(void);
269 
270 static inline int
271 arp_lock_try(int recurse)
272 {
273 	int s;
274 
275 	/*
276 	 * Use splvm() -- we're blocking things that would cause
277 	 * mbuf allocation.
278 	 */
279 	s = splvm();
280 	if (!recurse && arp_locked) {
281 		splx(s);
282 		return (0);
283 	}
284 	arp_locked++;
285 	splx(s);
286 	return (1);
287 }
288 
289 static inline void
290 arp_unlock(void)
291 {
292 	int s;
293 
294 	s = splvm();
295 	arp_locked--;
296 	splx(s);
297 }
298 
299 #ifdef DIAGNOSTIC
300 #define	ARP_LOCK(recurse)						\
301 do {									\
302 	if (arp_lock_try(recurse) == 0) {				\
303 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
304 		panic("arp_lock");					\
305 	}								\
306 } while (/*CONSTCOND*/ 0)
307 #define	ARP_LOCK_CHECK()						\
308 do {									\
309 	if (arp_locked == 0) {						\
310 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
311 		panic("arp lock check");				\
312 	}								\
313 } while (/*CONSTCOND*/ 0)
314 #else
315 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
316 #define	ARP_LOCK_CHECK()	/* nothing */
317 #endif
318 
319 #define	ARP_UNLOCK()		arp_unlock()
320 
321 void
322 arp_init(void)
323 {
324 
325 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
326 }
327 
328 /*
329  * ARP protocol drain routine.  Called when memory is in short supply.
330  * Called at splvm();
331  */
332 void
333 arp_drain(void)
334 {
335 	struct llinfo_arp *la, *nla;
336 	int count = 0;
337 	struct mbuf *mold;
338 
339 	mutex_enter(softnet_lock);
340 	KERNEL_LOCK(1, NULL);
341 
342 	if (arp_lock_try(0) == 0) {
343 		printf("arp_drain: locked; punting\n");
344 		KERNEL_UNLOCK_ONE(NULL);
345 		mutex_exit(softnet_lock);
346 		return;
347 	}
348 
349 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
350 		nla = LIST_NEXT(la, la_list);
351 
352 		mold = la->la_hold;
353 		la->la_hold = 0;
354 
355 		if (mold) {
356 			m_freem(mold);
357 			count++;
358 		}
359 	}
360 	ARP_UNLOCK();
361 	ARP_STATADD(ARP_STAT_DFRDROPPED, count);
362 	KERNEL_UNLOCK_ONE(NULL);
363 	mutex_exit(softnet_lock);
364 }
365 
366 
367 /*
368  * Timeout routine.  Age arp_tab entries periodically.
369  */
370 /* ARGSUSED */
371 static void
372 arptimer(void *arg)
373 {
374 	struct llinfo_arp *la, *nla;
375 
376 	mutex_enter(softnet_lock);
377 	KERNEL_LOCK(1, NULL);
378 
379 	if (arp_lock_try(0) == 0) {
380 		/* get it later.. */
381 		KERNEL_UNLOCK_ONE(NULL);
382 		mutex_exit(softnet_lock);
383 		return;
384 	}
385 
386 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
387 	for (la = LIST_FIRST(&llinfo_arp); la != 0; la = nla) {
388 		struct rtentry *rt = la->la_rt;
389 
390 		nla = LIST_NEXT(la, la_list);
391 		if (rt->rt_expire == 0)
392 			continue;
393 		if ((rt->rt_expire - time_second) < arpt_refresh &&
394 		    rt->rt_pksent > (time_second - arpt_keep)) {
395 			/*
396 			 * If the entry has been used during since last
397 			 * refresh, try to renew it before deleting.
398 			 */
399 			arprequest(rt->rt_ifp,
400 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
401 			    &satocsin(rt_getkey(rt))->sin_addr,
402 			    CLLADDR(rt->rt_ifp->if_sadl));
403 		} else if (rt->rt_expire <= time_second)
404 			arptfree(la); /* timer has expired; clear */
405 	}
406 
407 	ARP_UNLOCK();
408 
409 	KERNEL_UNLOCK_ONE(NULL);
410 	mutex_exit(softnet_lock);
411 }
412 
413 /*
414  * We set the gateway for RTF_CLONING routes to a "prototype"
415  * link-layer sockaddr whose interface type (if_type) and interface
416  * index (if_index) fields are prepared.
417  */
418 static struct sockaddr *
419 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
420     const struct sockaddr *netmask)
421 {
422 	const struct ifnet *ifp = rt->rt_ifp;
423 	uint8_t namelen = strlen(ifp->if_xname);
424 	uint8_t addrlen = ifp->if_addrlen;
425 
426 	/*
427 	 * XXX: If this is a manually added route to interface
428 	 * such as older version of routed or gated might provide,
429 	 * restore cloning bit.
430 	 */
431 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
432 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
433 		rt->rt_flags |= RTF_CLONING;
434 	if (rt->rt_flags & RTF_CLONING) {
435 		union {
436 			struct sockaddr sa;
437 			struct sockaddr_storage ss;
438 			struct sockaddr_dl sdl;
439 		} u;
440 		/*
441 		 * Case 1: This route should come from a route to iface.
442 		 */
443 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
444 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
445 		rt_setgate(rt, &u.sa);
446 		gate = rt->rt_gateway;
447 	}
448 	return gate;
449 }
450 
451 /*
452  * Parallel to llc_rtrequest.
453  */
454 void
455 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
456 {
457 	struct sockaddr *gate = rt->rt_gateway;
458 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
459 	size_t allocsize;
460 	struct mbuf *mold;
461 	int s;
462 	struct in_ifaddr *ia;
463 	struct ifaddr *ifa;
464 	struct ifnet *ifp = rt->rt_ifp;
465 	uint8_t namelen = strlen(ifp->if_xname);
466 	uint8_t addrlen = ifp->if_addrlen;
467 
468 	if (!arpinit_done) {
469 		arpinit_done = 1;
470 		/*
471 		 * We generate expiration times from time_second
472 		 * so avoid accidentally creating permanent routes.
473 		 */
474 		if (time_second == 0) {
475 			struct timespec ts;
476 			ts.tv_sec = 1;
477 			ts.tv_nsec = 0;
478 			tc_setclock(&ts);
479 		}
480 		callout_init(&arptimer_ch, CALLOUT_MPSAFE);
481 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
482 	}
483 
484 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
485 		if (req != RTM_ADD)
486 			return;
487 
488 		/*
489 		 * linklayers with particular link MTU limitation.
490 		 */
491 		switch(ifp->if_type) {
492 #if NFDDI > 0
493 		case IFT_FDDI:
494 			if (ifp->if_mtu > FDDIIPMTU)
495 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
496 			break;
497 #endif
498 #if NARC > 0
499 		case IFT_ARCNET:
500 		    {
501 			int arcipifmtu;
502 
503 			if (ifp->if_flags & IFF_LINK0)
504 				arcipifmtu = arc_ipmtu;
505 			else
506 				arcipifmtu = ARCMTU;
507 			if (ifp->if_mtu > arcipifmtu)
508 				rt->rt_rmx.rmx_mtu = arcipifmtu;
509 			break;
510 		    }
511 #endif
512 		}
513 		return;
514 	}
515 
516 	ARP_LOCK(1);		/* we may already be locked here. */
517 
518 	switch (req) {
519 
520 	case RTM_SETGATE:
521 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
522 		break;
523 	case RTM_ADD:
524 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
525 		if (rt->rt_flags & RTF_CLONING) {
526 			/*
527 			 * Give this route an expiration time, even though
528 			 * it's a "permanent" route, so that routes cloned
529 			 * from it do not need their expiration time set.
530 			 */
531 			rt->rt_expire = time_second;
532 			/*
533 			 * linklayers with particular link MTU limitation.
534 			 */
535 			switch (ifp->if_type) {
536 #if NFDDI > 0
537 			case IFT_FDDI:
538 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
539 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
540 				     (rt->rt_rmx.rmx_mtu == 0 &&
541 				      ifp->if_mtu > FDDIIPMTU)))
542 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
543 				break;
544 #endif
545 #if NARC > 0
546 			case IFT_ARCNET:
547 			    {
548 				int arcipifmtu;
549 				if (ifp->if_flags & IFF_LINK0)
550 					arcipifmtu = arc_ipmtu;
551 				else
552 					arcipifmtu = ARCMTU;
553 
554 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
555 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
556 				     (rt->rt_rmx.rmx_mtu == 0 &&
557 				      ifp->if_mtu > arcipifmtu)))
558 					rt->rt_rmx.rmx_mtu = arcipifmtu;
559 				break;
560 			    }
561 #endif
562 			}
563 			break;
564 		}
565 		/* Announce a new entry if requested. */
566 		if (rt->rt_flags & RTF_ANNOUNCE)
567 			arprequest(ifp,
568 			    &satocsin(rt_getkey(rt))->sin_addr,
569 			    &satocsin(rt_getkey(rt))->sin_addr,
570 			    CLLADDR(satocsdl(gate)));
571 		/*FALLTHROUGH*/
572 	case RTM_RESOLVE:
573 		if (gate->sa_family != AF_LINK ||
574 		    gate->sa_len < sockaddr_dl_measure(namelen, addrlen)) {
575 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
576 			break;
577 		}
578 		satosdl(gate)->sdl_type = ifp->if_type;
579 		satosdl(gate)->sdl_index = ifp->if_index;
580 		if (la != 0)
581 			break; /* This happens on a route change */
582 		/*
583 		 * Case 2:  This route may come from cloning, or a manual route
584 		 * add with a LL address.
585 		 */
586 		switch (satocsdl(gate)->sdl_type) {
587 #if NTOKEN > 0
588 		case IFT_ISO88025:
589 			allocsize = sizeof(*la) + sizeof(struct token_rif);
590 			break;
591 #endif /* NTOKEN > 0 */
592 		default:
593 			allocsize = sizeof(*la);
594 		}
595 		R_Malloc(la, struct llinfo_arp *, allocsize);
596 		rt->rt_llinfo = (void *)la;
597 		if (la == 0) {
598 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
599 			break;
600 		}
601 		arp_inuse++, arp_allocated++;
602 		Bzero(la, allocsize);
603 		la->la_rt = rt;
604 		rt->rt_flags |= RTF_LLINFO;
605 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
606 
607 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
608 		while (ia && ia->ia_ifp != ifp)
609 			NEXT_IA_WITH_SAME_ADDR(ia);
610 		if (ia) {
611 			/*
612 			 * This test used to be
613 			 *	if (lo0ifp->if_flags & IFF_UP)
614 			 * It allowed local traffic to be forced through
615 			 * the hardware by configuring the loopback down.
616 			 * However, it causes problems during network
617 			 * configuration for boards that can't receive
618 			 * packets they send.  It is now necessary to clear
619 			 * "useloopback" and remove the route to force
620 			 * traffic out to the hardware.
621 			 *
622 			 * In 4.4BSD, the above "if" statement checked
623 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
624 			 * to the current form so that we can provide a
625 			 * better support for multiple IPv4 addresses on a
626 			 * interface.
627 			 */
628 			rt->rt_expire = 0;
629 			(void)sockaddr_dl_setaddr(satosdl(gate), gate->sa_len,
630 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen);
631 			if (useloopback)
632 				ifp = rt->rt_ifp = lo0ifp;
633 			/*
634 			 * make sure to set rt->rt_ifa to the interface
635 			 * address we are using, otherwise we will have trouble
636 			 * with source address selection.
637 			 */
638 			ifa = &ia->ia_ifa;
639 			if (ifa != rt->rt_ifa)
640 				rt_replace_ifa(rt, ifa);
641 		}
642 		break;
643 
644 	case RTM_DELETE:
645 		if (la == 0)
646 			break;
647 		arp_inuse--;
648 		LIST_REMOVE(la, la_list);
649 		rt->rt_llinfo = 0;
650 		rt->rt_flags &= ~RTF_LLINFO;
651 
652 		s = splnet();
653 		mold = la->la_hold;
654 		la->la_hold = 0;
655 		splx(s);
656 
657 		if (mold)
658 			m_freem(mold);
659 
660 		Free((void *)la);
661 	}
662 	ARP_UNLOCK();
663 }
664 
665 /*
666  * Broadcast an ARP request. Caller specifies:
667  *	- arp header source ip address
668  *	- arp header target ip address
669  *	- arp header source ethernet address
670  */
671 void
672 arprequest(struct ifnet *ifp,
673     const struct in_addr *sip, const struct in_addr *tip,
674     const u_int8_t *enaddr)
675 {
676 	struct mbuf *m;
677 	struct arphdr *ah;
678 	struct sockaddr sa;
679 	uint64_t *arps;
680 
681 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
682 		return;
683 	MCLAIM(m, &arpdomain.dom_mowner);
684 	switch (ifp->if_type) {
685 	case IFT_IEEE1394:
686 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
687 		    ifp->if_addrlen;
688 		break;
689 	default:
690 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
691 		    2 * ifp->if_addrlen;
692 		break;
693 	}
694 	m->m_pkthdr.len = m->m_len;
695 	MH_ALIGN(m, m->m_len);
696 	ah = mtod(m, struct arphdr *);
697 	bzero((void *)ah, m->m_len);
698 	switch (ifp->if_type) {
699 	case IFT_IEEE1394:	/* RFC2734 */
700 		/* fill it now for ar_tpa computation */
701 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
702 		break;
703 	default:
704 		/* ifp->if_output will fill ar_hrd */
705 		break;
706 	}
707 	ah->ar_pro = htons(ETHERTYPE_IP);
708 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
709 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
710 	ah->ar_op = htons(ARPOP_REQUEST);
711 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
712 	memcpy(ar_spa(ah), sip, ah->ar_pln);
713 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
714 	sa.sa_family = AF_ARP;
715 	sa.sa_len = 2;
716 	m->m_flags |= M_BCAST;
717 	arps = ARP_STAT_GETREF();
718 	arps[ARP_STAT_SNDTOTAL]++;
719 	arps[ARP_STAT_SENDREQUEST]++;
720 	ARP_STAT_PUTREF();
721 	(*ifp->if_output)(ifp, m, &sa, NULL);
722 }
723 
724 /*
725  * Resolve an IP address into an ethernet address.  If success,
726  * desten is filled in.  If there is no entry in arptab,
727  * set one up and broadcast a request for the IP address.
728  * Hold onto this mbuf and resend it once the address
729  * is finally resolved.  A return value of 1 indicates
730  * that desten has been filled in and the packet should be sent
731  * normally; a 0 return indicates that the packet has been
732  * taken over here, either now or for later transmission.
733  */
734 int
735 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
736     const struct sockaddr *dst, u_char *desten)
737 {
738 	struct llinfo_arp *la;
739 	const struct sockaddr_dl *sdl;
740 	struct mbuf *mold;
741 	int s;
742 
743 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
744 		rt = la->la_rt;
745 
746 	if (la == 0 || rt == 0) {
747 		ARP_STATINC(ARP_STAT_ALLOCFAIL);
748 		log(LOG_DEBUG,
749 		    "arpresolve: can't allocate llinfo on %s for %s\n",
750 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
751 		m_freem(m);
752 		return (0);
753 	}
754 	sdl = satocsdl(rt->rt_gateway);
755 	/*
756 	 * Check the address family and length is valid, the address
757 	 * is resolved; otherwise, try to resolve.
758 	 */
759 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
760 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
761 		bcopy(CLLADDR(sdl), desten,
762 		    min(sdl->sdl_alen, ifp->if_addrlen));
763 		rt->rt_pksent = time_second; /* Time for last pkt sent */
764 		return 1;
765 	}
766 	/*
767 	 * There is an arptab entry, but no ethernet address
768 	 * response yet.  Replace the held mbuf with this
769 	 * latest one.
770 	 */
771 
772 	ARP_STATINC(ARP_STAT_DFRTOTAL);
773 	s = splnet();
774 	mold = la->la_hold;
775 	la->la_hold = m;
776 	splx(s);
777 
778 	if (mold) {
779 		ARP_STATINC(ARP_STAT_DFRDROPPED);
780 		m_freem(mold);
781 	}
782 
783 	/*
784 	 * Re-send the ARP request when appropriate.
785 	 */
786 #ifdef	DIAGNOSTIC
787 	if (rt->rt_expire == 0) {
788 		/* This should never happen. (Should it? -gwr) */
789 		printf("arpresolve: unresolved and rt_expire == 0\n");
790 		/* Set expiration time to now (expired). */
791 		rt->rt_expire = time_second;
792 	}
793 #endif
794 	if (rt->rt_expire) {
795 		rt->rt_flags &= ~RTF_REJECT;
796 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
797 			rt->rt_expire = time_second;
798 			if (la->la_asked++ < arp_maxtries)
799 				arprequest(ifp,
800 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
801 				    &satocsin(dst)->sin_addr,
802 #if NCARP > 0
803 				    (rt->rt_ifp->if_type == IFT_CARP) ?
804 				    CLLADDR(rt->rt_ifp->if_sadl):
805 #endif
806 				    CLLADDR(ifp->if_sadl));
807 			else {
808 				rt->rt_flags |= RTF_REJECT;
809 				rt->rt_expire += arpt_down;
810 				la->la_asked = 0;
811 			}
812 		}
813 	}
814 	return (0);
815 }
816 
817 /*
818  * Common length and type checks are done here,
819  * then the protocol-specific routine is called.
820  */
821 void
822 arpintr(void)
823 {
824 	struct mbuf *m;
825 	struct arphdr *ar;
826 	int s;
827 	int arplen;
828 
829 	mutex_enter(softnet_lock);
830 	KERNEL_LOCK(1, NULL);
831 	while (arpintrq.ifq_head) {
832 		s = splnet();
833 		IF_DEQUEUE(&arpintrq, m);
834 		splx(s);
835 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
836 			panic("arpintr");
837 
838 		MCLAIM(m, &arpdomain.dom_mowner);
839 		ARP_STATINC(ARP_STAT_RCVTOTAL);
840 
841 		/*
842 		 * First, make sure we have at least struct arphdr.
843 		 */
844 		if (m->m_len < sizeof(struct arphdr) ||
845 		    (ar = mtod(m, struct arphdr *)) == NULL)
846 			goto badlen;
847 
848 		switch (m->m_pkthdr.rcvif->if_type) {
849 		case IFT_IEEE1394:
850 			arplen = sizeof(struct arphdr) +
851 			    ar->ar_hln + 2 * ar->ar_pln;
852 			break;
853 		default:
854 			arplen = sizeof(struct arphdr) +
855 			    2 * ar->ar_hln + 2 * ar->ar_pln;
856 			break;
857 		}
858 
859 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
860 		    m->m_len >= arplen)
861 			switch (ntohs(ar->ar_pro)) {
862 			case ETHERTYPE_IP:
863 			case ETHERTYPE_IPTRAILERS:
864 				in_arpinput(m);
865 				continue;
866 			default:
867 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
868 			}
869 		else {
870 badlen:
871 			ARP_STATINC(ARP_STAT_RCVBADLEN);
872 		}
873 		m_freem(m);
874 	}
875 	KERNEL_UNLOCK_ONE(NULL);
876 	mutex_exit(softnet_lock);
877 }
878 
879 /*
880  * ARP for Internet protocols on 10 Mb/s Ethernet.
881  * Algorithm is that given in RFC 826.
882  * In addition, a sanity check is performed on the sender
883  * protocol address, to catch impersonators.
884  * We no longer handle negotiations for use of trailer protocol:
885  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
886  * along with IP replies if we wanted trailers sent to us,
887  * and also sent them in response to IP replies.
888  * This allowed either end to announce the desire to receive
889  * trailer packets.
890  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
891  * but formerly didn't normally send requests.
892  */
893 static void
894 in_arpinput(struct mbuf *m)
895 {
896 	struct arphdr *ah;
897 	struct ifnet *ifp = m->m_pkthdr.rcvif;
898 	struct llinfo_arp *la = 0;
899 	struct rtentry  *rt;
900 	struct in_ifaddr *ia;
901 #if NBRIDGE > 0
902 	struct in_ifaddr *bridge_ia = NULL;
903 #endif
904 #if NCARP > 0
905 	u_int32_t count = 0, index = 0;
906 #endif
907 	struct sockaddr_dl *sdl;
908 	struct sockaddr sa;
909 	struct in_addr isaddr, itaddr, myaddr;
910 	int op;
911 	struct mbuf *mold;
912 	void *tha;
913 	int s;
914 	uint64_t *arps;
915 
916 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
917 		goto out;
918 	ah = mtod(m, struct arphdr *);
919 	op = ntohs(ah->ar_op);
920 
921 	/*
922 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
923 	 * ar_tpa().
924 	 */
925 	switch (ifp->if_type) {
926 	case IFT_IEEE1394:
927 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
928 			;
929 		else {
930 			/* XXX this is to make sure we compute ar_tha right */
931 			/* XXX check ar_hrd more strictly? */
932 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
933 		}
934 		break;
935 	default:
936 		/* XXX check ar_hrd? */
937 		break;
938 	}
939 
940 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
941 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
942 
943 	if (m->m_flags & (M_BCAST|M_MCAST))
944 		ARP_STATINC(ARP_STAT_RCVMCAST);
945 
946 	/*
947 	 * If the target IP address is zero, ignore the packet.
948 	 * This prevents the code below from tring to answer
949 	 * when we are using IP address zero (booting).
950 	 */
951 	if (in_nullhost(itaddr)) {
952 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
953 		goto out;
954 	}
955 
956 	/*
957 	 * If the source IP address is zero, this is most likely a
958 	 * confused host trying to use IP address zero. (Windoze?)
959 	 * XXX: Should we bother trying to reply to these?
960 	 */
961 	if (in_nullhost(isaddr)) {
962 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
963 		goto out;
964 	}
965 
966 	/*
967 	 * Search for a matching interface address
968 	 * or any address on the interface to use
969 	 * as a dummy address in the rest of this function
970 	 */
971 
972 	INADDR_TO_IA(itaddr, ia);
973 	while (ia != NULL) {
974 #if NCARP > 0
975 		if (ia->ia_ifp->if_type == IFT_CARP &&
976 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
977 		    (IFF_UP|IFF_RUNNING))) {
978 			index++;
979 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
980 			    carp_iamatch(ia, ar_sha(ah),
981 			    &count, index)) {
982 				break;
983 				}
984 		} else
985 #endif
986 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
987 				break;
988 #if NBRIDGE > 0
989 		/*
990 		 * If the interface we received the packet on
991 		 * is part of a bridge, check to see if we need
992 		 * to "bridge" the packet to ourselves at this
993 		 * layer.  Note we still prefer a perfect match,
994 		 * but allow this weaker match if necessary.
995 		 */
996 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
997 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
998 			bridge_ia = ia;
999 #endif /* NBRIDGE > 0 */
1000 
1001 		NEXT_IA_WITH_SAME_ADDR(ia);
1002 	}
1003 
1004 #if NBRIDGE > 0
1005 	if (ia == NULL && bridge_ia != NULL) {
1006 		ia = bridge_ia;
1007 		ifp = bridge_ia->ia_ifp;
1008 	}
1009 #endif
1010 
1011 	if (ia == NULL) {
1012 		INADDR_TO_IA(isaddr, ia);
1013 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
1014 			NEXT_IA_WITH_SAME_ADDR(ia);
1015 
1016 		if (ia == NULL) {
1017 			IFP_TO_IA(ifp, ia);
1018 			if (ia == NULL) {
1019 				ARP_STATINC(ARP_STAT_RCVNOINT);
1020 				goto out;
1021 			}
1022 		}
1023 	}
1024 
1025 	myaddr = ia->ia_addr.sin_addr;
1026 
1027 	/* XXX checks for bridge case? */
1028 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
1029 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1030 		goto out;	/* it's from me, ignore it. */
1031 	}
1032 
1033 	/* XXX checks for bridge case? */
1034 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1035 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1036 		log(LOG_ERR,
1037 		    "%s: arp: link address is broadcast for IP address %s!\n",
1038 		    ifp->if_xname, in_fmtaddr(isaddr));
1039 		goto out;
1040 	}
1041 
1042 	if (in_hosteq(isaddr, myaddr)) {
1043 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1044 		log(LOG_ERR,
1045 		   "duplicate IP address %s sent from link address %s\n",
1046 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
1047 		itaddr = myaddr;
1048 		goto reply;
1049 	}
1050 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1051 	if (la && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
1052 		if (sdl->sdl_alen &&
1053 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
1054 			if (rt->rt_flags & RTF_STATIC) {
1055 				ARP_STATINC(ARP_STAT_RCVOVERPERM);
1056 				log(LOG_INFO,
1057 				    "%s tried to overwrite permanent arp info"
1058 				    " for %s\n",
1059 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1060 				    in_fmtaddr(isaddr));
1061 				goto out;
1062 			} else if (rt->rt_ifp != ifp) {
1063 				ARP_STATINC(ARP_STAT_RCVOVERINT);
1064 				log(LOG_INFO,
1065 				    "%s on %s tried to overwrite "
1066 				    "arp info for %s on %s\n",
1067 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1068 				    ifp->if_xname, in_fmtaddr(isaddr),
1069 				    rt->rt_ifp->if_xname);
1070 				    goto out;
1071 			} else {
1072 				ARP_STATINC(ARP_STAT_RCVOVER);
1073 				log(LOG_INFO,
1074 				    "arp info overwritten for %s by %s\n",
1075 				    in_fmtaddr(isaddr),
1076 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1077 			}
1078 		}
1079 		/*
1080 		 * sanity check for the address length.
1081 		 * XXX this does not work for protocols with variable address
1082 		 * length. -is
1083 		 */
1084 		if (sdl->sdl_alen &&
1085 		    sdl->sdl_alen != ah->ar_hln) {
1086 			ARP_STATINC(ARP_STAT_RCVLENCHG);
1087 			log(LOG_WARNING,
1088 			    "arp from %s: new addr len %d, was %d",
1089 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1090 		}
1091 		if (ifp->if_addrlen != ah->ar_hln) {
1092 			ARP_STATINC(ARP_STAT_RCVBADLEN);
1093 			log(LOG_WARNING,
1094 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1095 			    in_fmtaddr(isaddr), ah->ar_hln,
1096 			    ifp->if_addrlen);
1097 			goto reply;
1098 		}
1099 #if NTOKEN > 0
1100 		/*
1101 		 * XXX uses m_data and assumes the complete answer including
1102 		 * XXX token-ring headers is in the same buf
1103 		 */
1104 		if (ifp->if_type == IFT_ISO88025) {
1105 			struct token_header *trh;
1106 
1107 			trh = (struct token_header *)M_TRHSTART(m);
1108 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1109 				struct token_rif	*rif;
1110 				size_t	riflen;
1111 
1112 				rif = TOKEN_RIF(trh);
1113 				riflen = (ntohs(rif->tr_rcf) &
1114 				    TOKEN_RCF_LEN_MASK) >> 8;
1115 
1116 				if (riflen > 2 &&
1117 				    riflen < sizeof(struct token_rif) &&
1118 				    (riflen & 1) == 0) {
1119 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1120 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1121 					bcopy(rif, TOKEN_RIF(la), riflen);
1122 				}
1123 			}
1124 		}
1125 #endif /* NTOKEN > 0 */
1126 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
1127 		    ah->ar_hln);
1128 		if (rt->rt_expire)
1129 			rt->rt_expire = time_second + arpt_keep;
1130 		rt->rt_flags &= ~RTF_REJECT;
1131 		la->la_asked = 0;
1132 
1133 		s = splnet();
1134 		mold = la->la_hold;
1135 		la->la_hold = 0;
1136 		splx(s);
1137 
1138 		if (mold) {
1139 			ARP_STATINC(ARP_STAT_DFRSENT);
1140 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
1141 		}
1142 	}
1143 reply:
1144 	if (op != ARPOP_REQUEST) {
1145 		if (op == ARPOP_REPLY)
1146 			ARP_STATINC(ARP_STAT_RCVREPLY);
1147 	out:
1148 		m_freem(m);
1149 		return;
1150 	}
1151 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1152 	if (in_hosteq(itaddr, myaddr)) {
1153 		/* I am the target */
1154 		tha = ar_tha(ah);
1155 		if (tha)
1156 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1157 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1158 	} else {
1159 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1160 		if (la == 0)
1161 			goto out;
1162 		rt = la->la_rt;
1163 		if (rt->rt_ifp->if_type == IFT_CARP &&
1164 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1165 			goto out;
1166 		tha = ar_tha(ah);
1167 		if (tha)
1168 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1169 		sdl = satosdl(rt->rt_gateway);
1170 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
1171 	}
1172 
1173 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1174 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1175 	ah->ar_op = htons(ARPOP_REPLY);
1176 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1177 	switch (ifp->if_type) {
1178 	case IFT_IEEE1394:
1179 		/*
1180 		 * ieee1394 arp reply is broadcast
1181 		 */
1182 		m->m_flags &= ~M_MCAST;
1183 		m->m_flags |= M_BCAST;
1184 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1185 		break;
1186 
1187 	default:
1188 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1189 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1190 		break;
1191 	}
1192 	m->m_pkthdr.len = m->m_len;
1193 	sa.sa_family = AF_ARP;
1194 	sa.sa_len = 2;
1195 	arps = ARP_STAT_GETREF();
1196 	arps[ARP_STAT_SNDTOTAL]++;
1197 	arps[ARP_STAT_SNDREPLY]++;
1198 	ARP_STAT_PUTREF();
1199 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1200 	return;
1201 }
1202 
1203 /*
1204  * Free an arp entry.
1205  */
1206 static void arptfree(struct llinfo_arp *la)
1207 {
1208 	struct rtentry *rt = la->la_rt;
1209 	struct sockaddr_dl *sdl;
1210 
1211 	ARP_LOCK_CHECK();
1212 
1213 	if (rt == 0)
1214 		panic("arptfree");
1215 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
1216 	    sdl->sdl_family == AF_LINK) {
1217 		sdl->sdl_alen = 0;
1218 		la->la_asked = 0;
1219 		rt->rt_flags &= ~RTF_REJECT;
1220 		return;
1221 	}
1222 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1223 }
1224 
1225 static struct llinfo_arp *
1226 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
1227 {
1228 	return arplookup1(m, addr, create, proxy, NULL);
1229 }
1230 
1231 /*
1232  * Lookup or enter a new address in arptab.
1233  */
1234 static struct llinfo_arp *
1235 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
1236     struct rtentry *rt0)
1237 {
1238 	struct arphdr *ah;
1239 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1240 	struct rtentry *rt;
1241 	static struct sockaddr_inarp sin;
1242 	const char *why = 0;
1243 
1244 	ah = mtod(m, struct arphdr *);
1245 	if (rt0 == NULL) {
1246 		sin.sin_len = sizeof(sin);
1247 		sin.sin_family = AF_INET;
1248 		sin.sin_addr = *addr;
1249 		sin.sin_other = proxy ? SIN_PROXY : 0;
1250 		rt = rtalloc1(sintosa(&sin), create);
1251 		if (rt == NULL)
1252 			return (NULL);
1253 		rt->rt_refcnt--;
1254 	} else
1255 		rt = rt0;
1256 
1257 #define	IS_LLINFO(__rt)							  \
1258 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
1259 	 (__rt)->rt_gateway->sa_family == AF_LINK)
1260 
1261 
1262 	if (IS_LLINFO(rt))
1263 		return ((struct llinfo_arp *)rt->rt_llinfo);
1264 
1265 	if (create) {
1266 		if (rt->rt_flags & RTF_GATEWAY)
1267 			why = "host is not on local network";
1268 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1269 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
1270 			why = "could not allocate llinfo";
1271 		} else
1272 			why = "gateway route is not ours";
1273 		log(LOG_DEBUG, "arplookup: unable to enter address"
1274 		    " for %s@%s on %s (%s)\n",
1275 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1276 		    (ifp) ? ifp->if_xname : 0, why);
1277 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1278 			rtrequest(RTM_DELETE, rt_getkey(rt),
1279 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1280 		}
1281 	}
1282 	return (0);
1283 }
1284 
1285 int
1286 arpioctl(u_long cmd, void *data)
1287 {
1288 
1289 	return (EOPNOTSUPP);
1290 }
1291 
1292 void
1293 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1294 {
1295 	struct in_addr *ip;
1296 
1297 	/*
1298 	 * Warn the user if another station has this IP address,
1299 	 * but only if the interface IP address is not zero.
1300 	 */
1301 	ip = &IA_SIN(ifa)->sin_addr;
1302 	if (!in_nullhost(*ip))
1303 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
1304 
1305 	ifa->ifa_rtrequest = arp_rtrequest;
1306 	ifa->ifa_flags |= RTF_CLONING;
1307 }
1308 
1309 /*
1310  * Called from 10 Mb/s Ethernet interrupt handlers
1311  * when ether packet type ETHERTYPE_REVARP
1312  * is received.  Common length and type checks are done here,
1313  * then the protocol-specific routine is called.
1314  */
1315 void
1316 revarpinput(struct mbuf *m)
1317 {
1318 	struct arphdr *ar;
1319 
1320 	if (m->m_len < sizeof(struct arphdr))
1321 		goto out;
1322 	ar = mtod(m, struct arphdr *);
1323 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1324 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1325 		goto out;
1326 #endif
1327 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1328 		goto out;
1329 	switch (ntohs(ar->ar_pro)) {
1330 	case ETHERTYPE_IP:
1331 	case ETHERTYPE_IPTRAILERS:
1332 		in_revarpinput(m);
1333 		return;
1334 
1335 	default:
1336 		break;
1337 	}
1338 out:
1339 	m_freem(m);
1340 }
1341 
1342 /*
1343  * RARP for Internet protocols on 10 Mb/s Ethernet.
1344  * Algorithm is that given in RFC 903.
1345  * We are only using for bootstrap purposes to get an ip address for one of
1346  * our interfaces.  Thus we support no user-interface.
1347  *
1348  * Since the contents of the RARP reply are specific to the interface that
1349  * sent the request, this code must ensure that they are properly associated.
1350  *
1351  * Note: also supports ARP via RARP packets, per the RFC.
1352  */
1353 void
1354 in_revarpinput(struct mbuf *m)
1355 {
1356 	struct ifnet *ifp;
1357 	struct arphdr *ah;
1358 	void *tha;
1359 	int op;
1360 
1361 	ah = mtod(m, struct arphdr *);
1362 	op = ntohs(ah->ar_op);
1363 
1364 	switch (m->m_pkthdr.rcvif->if_type) {
1365 	case IFT_IEEE1394:
1366 		/* ARP without target hardware address is not supported */
1367 		goto out;
1368 	default:
1369 		break;
1370 	}
1371 
1372 	switch (op) {
1373 	case ARPOP_REQUEST:
1374 	case ARPOP_REPLY:	/* per RFC */
1375 		in_arpinput(m);
1376 		return;
1377 	case ARPOP_REVREPLY:
1378 		break;
1379 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1380 	default:
1381 		goto out;
1382 	}
1383 	if (!revarp_in_progress)
1384 		goto out;
1385 	ifp = m->m_pkthdr.rcvif;
1386 	if (ifp != myip_ifp) /* !same interface */
1387 		goto out;
1388 	if (myip_initialized)
1389 		goto wake;
1390 	tha = ar_tha(ah);
1391 	KASSERT(tha);
1392 	if (bcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1393 		goto out;
1394 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1395 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1396 	myip_initialized = 1;
1397 wake:	/* Do wakeup every time in case it was missed. */
1398 	wakeup((void *)&myip);
1399 
1400 out:
1401 	m_freem(m);
1402 }
1403 
1404 /*
1405  * Send a RARP request for the ip address of the specified interface.
1406  * The request should be RFC 903-compliant.
1407  */
1408 void
1409 revarprequest(struct ifnet *ifp)
1410 {
1411 	struct sockaddr sa;
1412 	struct mbuf *m;
1413 	struct arphdr *ah;
1414 	void *tha;
1415 
1416 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1417 		return;
1418 	MCLAIM(m, &arpdomain.dom_mowner);
1419 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1420 	    2*ifp->if_addrlen;
1421 	m->m_pkthdr.len = m->m_len;
1422 	MH_ALIGN(m, m->m_len);
1423 	ah = mtod(m, struct arphdr *);
1424 	bzero((void *)ah, m->m_len);
1425 	ah->ar_pro = htons(ETHERTYPE_IP);
1426 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1427 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1428 	ah->ar_op = htons(ARPOP_REVREQUEST);
1429 
1430 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1431 	tha = ar_tha(ah);
1432 	KASSERT(tha);
1433 	bcopy(CLLADDR(ifp->if_sadl), tha, ah->ar_hln);
1434 
1435 	sa.sa_family = AF_ARP;
1436 	sa.sa_len = 2;
1437 	m->m_flags |= M_BCAST;
1438 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1439 
1440 }
1441 
1442 /*
1443  * RARP for the ip address of the specified interface, but also
1444  * save the ip address of the server that sent the answer.
1445  * Timeout if no response is received.
1446  */
1447 int
1448 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1449     struct in_addr *clnt_in)
1450 {
1451 	int result, count = 20;
1452 
1453 	myip_initialized = 0;
1454 	myip_ifp = ifp;
1455 
1456 	revarp_in_progress = 1;
1457 	while (count--) {
1458 		revarprequest(ifp);
1459 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1460 		if (result != EWOULDBLOCK)
1461 			break;
1462 	}
1463 	revarp_in_progress = 0;
1464 
1465 	if (!myip_initialized)
1466 		return ENETUNREACH;
1467 
1468 	bcopy((void *)&srv_ip, serv_in, sizeof(*serv_in));
1469 	bcopy((void *)&myip, clnt_in, sizeof(*clnt_in));
1470 	return 0;
1471 }
1472 
1473 
1474 
1475 #ifdef DDB
1476 
1477 #include <machine/db_machdep.h>
1478 #include <ddb/db_interface.h>
1479 #include <ddb/db_output.h>
1480 
1481 static void
1482 db_print_sa(const struct sockaddr *sa)
1483 {
1484 	int len;
1485 	const u_char *p;
1486 
1487 	if (sa == 0) {
1488 		db_printf("[NULL]");
1489 		return;
1490 	}
1491 
1492 	p = (const u_char *)sa;
1493 	len = sa->sa_len;
1494 	db_printf("[");
1495 	while (len > 0) {
1496 		db_printf("%d", *p);
1497 		p++; len--;
1498 		if (len) db_printf(",");
1499 	}
1500 	db_printf("]\n");
1501 }
1502 
1503 static void
1504 db_print_ifa(struct ifaddr *ifa)
1505 {
1506 	if (ifa == 0)
1507 		return;
1508 	db_printf("  ifa_addr=");
1509 	db_print_sa(ifa->ifa_addr);
1510 	db_printf("  ifa_dsta=");
1511 	db_print_sa(ifa->ifa_dstaddr);
1512 	db_printf("  ifa_mask=");
1513 	db_print_sa(ifa->ifa_netmask);
1514 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1515 			  ifa->ifa_flags,
1516 			  ifa->ifa_refcnt,
1517 			  ifa->ifa_metric);
1518 }
1519 
1520 static void
1521 db_print_llinfo(void *li)
1522 {
1523 	struct llinfo_arp *la;
1524 
1525 	if (li == 0)
1526 		return;
1527 	la = (struct llinfo_arp *)li;
1528 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1529 			  la->la_rt, la->la_hold, la->la_asked);
1530 }
1531 
1532 /*
1533  * Function to pass to rt_walktree().
1534  * Return non-zero error to abort walk.
1535  */
1536 static int
1537 db_show_rtentry(struct rtentry *rt, void *w)
1538 {
1539 	db_printf("rtentry=%p", rt);
1540 
1541 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1542 			  rt->rt_flags, rt->rt_refcnt,
1543 			  rt->rt_use, rt->rt_expire);
1544 
1545 	db_printf(" key="); db_print_sa(rt_getkey(rt));
1546 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1547 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1548 
1549 	db_printf(" ifp=%p ", rt->rt_ifp);
1550 	if (rt->rt_ifp)
1551 		db_printf("(%s)", rt->rt_ifp->if_xname);
1552 	else
1553 		db_printf("(NULL)");
1554 
1555 	db_printf(" ifa=%p\n", rt->rt_ifa);
1556 	db_print_ifa(rt->rt_ifa);
1557 
1558 	db_printf(" gwroute=%p llinfo=%p\n",
1559 			  rt->rt_gwroute, rt->rt_llinfo);
1560 	db_print_llinfo(rt->rt_llinfo);
1561 
1562 	return (0);
1563 }
1564 
1565 /*
1566  * Function to print all the route trees.
1567  * Use this from ddb:  "show arptab"
1568  */
1569 void
1570 db_show_arptab(db_expr_t addr, bool have_addr,
1571     db_expr_t count, const char *modif)
1572 {
1573 	rt_walktree(AF_INET, db_show_rtentry, NULL);
1574 }
1575 #endif
1576 
1577 static int
1578 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1579 {
1580 	netstat_sysctl_context ctx;
1581 	uint64_t arps[ARP_NSTATS];
1582 
1583 	ctx.ctx_stat = arpstat_percpu;
1584 	ctx.ctx_counters = arps;
1585 	ctx.ctx_ncounters = ARP_NSTATS;
1586 	return (NETSTAT_SYSCTL(&ctx));
1587 }
1588 
1589 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1590 {
1591 	const struct sysctlnode *node;
1592 
1593 	sysctl_createv(clog, 0, NULL, NULL,
1594 			CTLFLAG_PERMANENT,
1595 			CTLTYPE_NODE, "net", NULL,
1596 			NULL, 0, NULL, 0,
1597 			CTL_NET, CTL_EOL);
1598 	sysctl_createv(clog, 0, NULL, NULL,
1599 			CTLFLAG_PERMANENT,
1600 			CTLTYPE_NODE, "inet", NULL,
1601 			NULL, 0, NULL, 0,
1602 			CTL_NET, PF_INET, CTL_EOL);
1603 	sysctl_createv(clog, 0, NULL, &node,
1604 			CTLFLAG_PERMANENT,
1605 			CTLTYPE_NODE, "arp",
1606 			SYSCTL_DESCR("Address Resolution Protocol"),
1607 			NULL, 0, NULL, 0,
1608 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1609 
1610 	sysctl_createv(clog, 0, NULL, NULL,
1611 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1612 			CTLTYPE_INT, "prune",
1613 			SYSCTL_DESCR("ARP cache pruning interval"),
1614 			NULL, 0, &arpt_prune, 0,
1615 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1616 
1617 	sysctl_createv(clog, 0, NULL, NULL,
1618 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1619 			CTLTYPE_INT, "keep",
1620 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1621 			NULL, 0, &arpt_keep, 0,
1622 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1623 
1624 	sysctl_createv(clog, 0, NULL, NULL,
1625 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1626 			CTLTYPE_INT, "down",
1627 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1628 			NULL, 0, &arpt_down, 0,
1629 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1630 
1631 	sysctl_createv(clog, 0, NULL, NULL,
1632 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1633 			CTLTYPE_INT, "refresh",
1634 			SYSCTL_DESCR("ARP entry refresh interval"),
1635 			NULL, 0, &arpt_refresh, 0,
1636 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1637 
1638 	sysctl_createv(clog, 0, NULL, NULL,
1639 			CTLFLAG_PERMANENT,
1640 			CTLTYPE_STRUCT, "stats",
1641 			SYSCTL_DESCR("ARP statistics"),
1642 			sysctl_net_inet_arp_stats, 0, NULL, 0,
1643 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1644 }
1645 
1646 #endif /* INET */
1647