xref: /netbsd-src/sys/netinet/if_arp.c (revision 37afb7eb6895c833050f8bfb1d1bb2f99f332539)
1 /*	$NetBSD: if_arp.c,v 1.170 2015/07/15 08:49:15 ozaki-r 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 1.170 2015/07/15 08:49:15 ozaki-r 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 #include <sys/cprng.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/if_ether.h>
106 #include <net/net_osdep.h>
107 #include <net/route.h>
108 #include <net/net_stats.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/in_var.h>
113 #include <netinet/ip.h>
114 #include <netinet/if_inarp.h>
115 
116 #include "arcnet.h"
117 #if NARCNET > 0
118 #include <net/if_arc.h>
119 #endif
120 #include "fddi.h"
121 #if NFDDI > 0
122 #include <net/if_fddi.h>
123 #endif
124 #include "token.h"
125 #include "carp.h"
126 #if NCARP > 0
127 #include <netinet/ip_carp.h>
128 #endif
129 
130 #define SIN(s) ((struct sockaddr_in *)s)
131 #define SRP(s) ((struct sockaddr_inarp *)s)
132 
133 /*
134  * ARP trailer negotiation.  Trailer protocol is not IP specific,
135  * but ARP request/response use IP addresses.
136  */
137 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
138 
139 /* timer values */
140 static int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
141 static int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
142 static int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
143 static int	arpt_refresh = (5*60);	/* time left before refreshing */
144 #define	rt_expire rt_rmx.rmx_expire
145 #define	rt_pksent rt_rmx.rmx_pksent
146 
147 int		ip_dad_count = PROBE_NUM;
148 #ifdef ARP_DEBUG
149 static int	arp_debug = 1;
150 #else
151 static int	arp_debug = 0;
152 #endif
153 #define arplog(x)	do { if (arp_debug) log x; } while (/*CONSTCOND*/ 0)
154 
155 static	void arp_init(void);
156 
157 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
158 	    const struct sockaddr *);
159 static	void arptfree(struct llinfo_arp *);
160 static	void arptimer(void *);
161 static	struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *,
162 				      int, int, struct rtentry *);
163 static	struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *,
164 					  int, int);
165 static	void in_arpinput(struct mbuf *);
166 static	void in_revarpinput(struct mbuf *);
167 static	void revarprequest(struct ifnet *);
168 
169 static	void arp_drainstub(void);
170 
171 static void arp_dad_timer(struct ifaddr *);
172 static void arp_dad_start(struct ifaddr *);
173 static void arp_dad_stop(struct ifaddr *);
174 static void arp_dad_duplicated(struct ifaddr *);
175 
176 LIST_HEAD(llinfo_arpq, llinfo_arp) llinfo_arp;
177 struct	ifqueue arpintrq = {
178 	.ifq_head = NULL,
179 	.ifq_tail = NULL,
180 	.ifq_len = 0,
181 	.ifq_maxlen = 50,
182 	.ifq_drops = 0,
183 };
184 static int	arp_inuse, arp_allocated;
185 static int	arp_maxtries = 5;
186 static int	useloopback = 1;	/* use loopback interface for local traffic */
187 static int	arpinit_done = 0;
188 
189 static percpu_t *arpstat_percpu;
190 
191 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
192 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
193 
194 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
195 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
196 
197 struct	callout arptimer_ch;
198 
199 /* revarp state */
200 static struct	in_addr myip, srv_ip;
201 static int	myip_initialized = 0;
202 static int	revarp_in_progress = 0;
203 static struct	ifnet *myip_ifp = NULL;
204 
205 #ifdef DDB
206 static void db_print_sa(const struct sockaddr *);
207 static void db_print_ifa(struct ifaddr *);
208 static void db_print_llinfo(void *);
209 static int db_show_rtentry(struct rtentry *, void *);
210 #endif
211 
212 static int arp_drainwanted;
213 
214 static int log_movements = 1;
215 static int log_permanent_modify = 1;
216 static int log_wrong_iface = 1;
217 static int log_unknown_network = 1;
218 
219 /*
220  * this should be elsewhere.
221  */
222 
223 static char *
224 lla_snprintf(u_int8_t *, int);
225 
226 static char *
227 lla_snprintf(u_int8_t *adrp, int len)
228 {
229 #define NUMBUFS 3
230 	static char buf[NUMBUFS][16*3];
231 	static int bnum = 0;
232 
233 	int i;
234 	char *p;
235 
236 	p = buf[bnum];
237 
238 	*p++ = hexdigits[(*adrp)>>4];
239 	*p++ = hexdigits[(*adrp++)&0xf];
240 
241 	for (i=1; i<len && i<16; i++) {
242 		*p++ = ':';
243 		*p++ = hexdigits[(*adrp)>>4];
244 		*p++ = hexdigits[(*adrp++)&0xf];
245 	}
246 
247 	*p = 0;
248 	p = buf[bnum];
249 	bnum = (bnum + 1) % NUMBUFS;
250 	return p;
251 }
252 
253 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
254 
255 static void
256 arp_fasttimo(void)
257 {
258 	if (arp_drainwanted) {
259 		arp_drain();
260 		arp_drainwanted = 0;
261 	}
262 }
263 
264 const struct protosw arpsw[] = {
265 	{ .pr_type = 0,
266 	  .pr_domain = &arpdomain,
267 	  .pr_protocol = 0,
268 	  .pr_flags = 0,
269 	  .pr_input = 0,
270 	  .pr_output = 0,
271 	  .pr_ctlinput = 0,
272 	  .pr_ctloutput = 0,
273 	  .pr_usrreqs = 0,
274 	  .pr_init = arp_init,
275 	  .pr_fasttimo = arp_fasttimo,
276 	  .pr_slowtimo = 0,
277 	  .pr_drain = arp_drainstub,
278 	}
279 };
280 
281 struct domain arpdomain = {
282 	.dom_family = PF_ARP,
283 	.dom_name = "arp",
284 	.dom_protosw = arpsw,
285 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
286 };
287 
288 /*
289  * ARP table locking.
290  *
291  * to prevent lossage vs. the arp_drain routine (which may be called at
292  * any time, including in a device driver context), we do two things:
293  *
294  * 1) manipulation of la->la_hold is done at splnet() (for all of
295  * about two instructions).
296  *
297  * 2) manipulation of the arp table's linked list is done under the
298  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
299  * while the arp table is locked, we punt and try again later.
300  */
301 
302 static int	arp_locked;
303 static inline int arp_lock_try(int);
304 static inline void arp_unlock(void);
305 
306 static inline int
307 arp_lock_try(int recurse)
308 {
309 	int s;
310 
311 	/*
312 	 * Use splvm() -- we're blocking things that would cause
313 	 * mbuf allocation.
314 	 */
315 	s = splvm();
316 	if (!recurse && arp_locked) {
317 		splx(s);
318 		return 0;
319 	}
320 	arp_locked++;
321 	splx(s);
322 	return 1;
323 }
324 
325 static inline void
326 arp_unlock(void)
327 {
328 	int s;
329 
330 	s = splvm();
331 	arp_locked--;
332 	splx(s);
333 }
334 
335 #ifdef DIAGNOSTIC
336 #define	ARP_LOCK(recurse)						\
337 do {									\
338 	if (arp_lock_try(recurse) == 0) {				\
339 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
340 		panic("arp_lock");					\
341 	}								\
342 } while (/*CONSTCOND*/ 0)
343 #define	ARP_LOCK_CHECK()						\
344 do {									\
345 	if (arp_locked == 0) {						\
346 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
347 		panic("arp lock check");				\
348 	}								\
349 } while (/*CONSTCOND*/ 0)
350 #else
351 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
352 #define	ARP_LOCK_CHECK()	/* nothing */
353 #endif
354 
355 #define	ARP_UNLOCK()		arp_unlock()
356 
357 static void sysctl_net_inet_arp_setup(struct sysctllog **);
358 
359 void
360 arp_init(void)
361 {
362 
363 	sysctl_net_inet_arp_setup(NULL);
364 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
365 }
366 
367 static void
368 arp_drainstub(void)
369 {
370 	arp_drainwanted = 1;
371 }
372 
373 /*
374  * ARP protocol drain routine.  Called when memory is in short supply.
375  * Called at splvm();  don't acquire softnet_lock as can be called from
376  * hardware interrupt handlers.
377  */
378 void
379 arp_drain(void)
380 {
381 	struct llinfo_arp *la;
382 	int count = 0;
383 	struct mbuf *mold;
384 
385 	KERNEL_LOCK(1, NULL);
386 
387 	if (arp_lock_try(0) == 0) {
388 		KERNEL_UNLOCK_ONE(NULL);
389 		return;
390 	}
391 
392 	LIST_FOREACH(la, &llinfo_arp, la_list) {
393 		mold = la->la_hold;
394 		la->la_hold = NULL;
395 
396 		if (mold) {
397 			m_freem(mold);
398 			count++;
399 		}
400 	}
401 	ARP_UNLOCK();
402 	ARP_STATADD(ARP_STAT_DFRDROPPED, count);
403 	KERNEL_UNLOCK_ONE(NULL);
404 }
405 
406 
407 /*
408  * Timeout routine.  Age arp_tab entries periodically.
409  */
410 /* ARGSUSED */
411 static void
412 arptimer(void *arg)
413 {
414 	struct llinfo_arp *la, *nla;
415 
416 	mutex_enter(softnet_lock);
417 	KERNEL_LOCK(1, NULL);
418 
419 	if (arp_lock_try(0) == 0) {
420 		/* get it later.. */
421 		KERNEL_UNLOCK_ONE(NULL);
422 		mutex_exit(softnet_lock);
423 		return;
424 	}
425 
426 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
427 	LIST_FOREACH_SAFE(la, &llinfo_arp, la_list, nla) {
428 		struct rtentry *rt = la->la_rt;
429 
430 		if (rt->rt_expire == 0)
431 			continue;
432 		if ((rt->rt_expire - time_second) < arpt_refresh &&
433 		    rt->rt_pksent > (time_second - arpt_keep)) {
434 			/*
435 			 * If the entry has been used during since last
436 			 * refresh, try to renew it before deleting.
437 			 */
438 			arprequest(rt->rt_ifp,
439 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
440 			    &satocsin(rt_getkey(rt))->sin_addr,
441 			    CLLADDR(rt->rt_ifp->if_sadl));
442 		} else if (rt->rt_expire <= time_second)
443 			arptfree(la); /* timer has expired; clear */
444 	}
445 
446 	ARP_UNLOCK();
447 
448 	KERNEL_UNLOCK_ONE(NULL);
449 	mutex_exit(softnet_lock);
450 }
451 
452 /*
453  * We set the gateway for RTF_CLONING routes to a "prototype"
454  * link-layer sockaddr whose interface type (if_type) and interface
455  * index (if_index) fields are prepared.
456  */
457 static struct sockaddr *
458 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
459     const struct sockaddr *netmask)
460 {
461 	const struct ifnet *ifp = rt->rt_ifp;
462 	uint8_t namelen = strlen(ifp->if_xname);
463 	uint8_t addrlen = ifp->if_addrlen;
464 
465 	/*
466 	 * XXX: If this is a manually added route to interface
467 	 * such as older version of routed or gated might provide,
468 	 * restore cloning bit.
469 	 */
470 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
471 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
472 		rt->rt_flags |= RTF_CLONING;
473 	if (rt->rt_flags & RTF_CLONING ||
474 	    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !rt->rt_llinfo))
475 	{
476 		union {
477 			struct sockaddr sa;
478 			struct sockaddr_storage ss;
479 			struct sockaddr_dl sdl;
480 		} u;
481 		/*
482 		 * Case 1: This route should come from a route to iface.
483 		 */
484 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
485 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
486 		rt_setgate(rt, &u.sa);
487 		gate = rt->rt_gateway;
488 	}
489 	return gate;
490 }
491 
492 /*
493  * Parallel to llc_rtrequest.
494  */
495 void
496 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
497 {
498 	struct sockaddr *gate = rt->rt_gateway;
499 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
500 	size_t allocsize;
501 	struct mbuf *mold;
502 	int s;
503 	struct in_ifaddr *ia;
504 	struct ifaddr *ifa;
505 	struct ifnet *ifp = rt->rt_ifp;
506 
507 	if (!arpinit_done) {
508 		arpinit_done = 1;
509 		/*
510 		 * We generate expiration times from time_second
511 		 * so avoid accidentally creating permanent routes.
512 		 */
513 		if (time_second == 0) {
514 			struct timespec ts;
515 			ts.tv_sec = 1;
516 			ts.tv_nsec = 0;
517 			tc_setclock(&ts);
518 		}
519 		callout_init(&arptimer_ch, CALLOUT_MPSAFE);
520 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
521 	}
522 
523 	if (req == RTM_LLINFO_UPD) {
524 		struct in_addr *in;
525 
526 		if ((ifa = info->rti_ifa) == NULL)
527 			return;
528 
529 		in = &ifatoia(ifa)->ia_addr.sin_addr;
530 
531 		if (ifatoia(ifa)->ia4_flags &
532 		    (IN_IFF_NOTREADY | IN_IFF_DETACHED))
533 		{
534 			arplog((LOG_DEBUG, "arp_request: %s not ready\n",
535 			   in_fmtaddr(*in)));
536 			return;
537 		}
538 
539 		arprequest(ifa->ifa_ifp, in, in,
540 		    CLLADDR(ifa->ifa_ifp->if_sadl));
541 		return;
542 	}
543 
544 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
545 		if (req != RTM_ADD)
546 			return;
547 
548 		/*
549 		 * linklayers with particular link MTU limitation.
550 		 */
551 		switch(ifp->if_type) {
552 #if NFDDI > 0
553 		case IFT_FDDI:
554 			if (ifp->if_mtu > FDDIIPMTU)
555 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
556 			break;
557 #endif
558 #if NARCNET > 0
559 		case IFT_ARCNET:
560 		    {
561 			int arcipifmtu;
562 
563 			if (ifp->if_flags & IFF_LINK0)
564 				arcipifmtu = arc_ipmtu;
565 			else
566 				arcipifmtu = ARCMTU;
567 			if (ifp->if_mtu > arcipifmtu)
568 				rt->rt_rmx.rmx_mtu = arcipifmtu;
569 			break;
570 		    }
571 #endif
572 		}
573 		return;
574 	}
575 
576 	ARP_LOCK(1);		/* we may already be locked here. */
577 
578 	switch (req) {
579 	case RTM_SETGATE:
580 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
581 		break;
582 	case RTM_ADD:
583 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
584 		if (rt->rt_flags & RTF_CLONING ||
585 		    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !la))
586 		{
587 			/*
588 			 * Give this route an expiration time, even though
589 			 * it's a "permanent" route, so that routes cloned
590 			 * from it do not need their expiration time set.
591 			 */
592 			rt->rt_expire = time_second;
593 			/*
594 			 * linklayers with particular link MTU limitation.
595 			 */
596 			switch (ifp->if_type) {
597 #if NFDDI > 0
598 			case IFT_FDDI:
599 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
600 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
601 				     (rt->rt_rmx.rmx_mtu == 0 &&
602 				      ifp->if_mtu > FDDIIPMTU)))
603 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
604 				break;
605 #endif
606 #if NARCNET > 0
607 			case IFT_ARCNET:
608 			    {
609 				int arcipifmtu;
610 				if (ifp->if_flags & IFF_LINK0)
611 					arcipifmtu = arc_ipmtu;
612 				else
613 					arcipifmtu = ARCMTU;
614 
615 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
616 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
617 				     (rt->rt_rmx.rmx_mtu == 0 &&
618 				      ifp->if_mtu > arcipifmtu)))
619 					rt->rt_rmx.rmx_mtu = arcipifmtu;
620 				break;
621 			    }
622 #endif
623 			}
624 			if (rt->rt_flags & RTF_CLONING)
625 				break;
626 		}
627 		/* Announce a new entry if requested. */
628 		if (rt->rt_flags & RTF_ANNOUNCE) {
629 			INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
630 			while (ia && ia->ia_ifp != ifp)
631 				NEXT_IA_WITH_SAME_ADDR(ia);
632 			if (ia == NULL ||
633 			    ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
634 				;
635 			else
636 				arprequest(ifp,
637 				    &satocsin(rt_getkey(rt))->sin_addr,
638 				    &satocsin(rt_getkey(rt))->sin_addr,
639 				    CLLADDR(satocsdl(gate)));
640 		}
641 		/*FALLTHROUGH*/
642 	case RTM_RESOLVE:
643 		if (gate->sa_family != AF_LINK ||
644 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
645 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
646 			break;
647 		}
648 
649 		satosdl(gate)->sdl_type = ifp->if_type;
650 		satosdl(gate)->sdl_index = ifp->if_index;
651 		if (la != NULL)
652 			break; /* This happens on a route change */
653 
654 		/* If the route is for a broadcast address mark it as such.
655 		 * This way we can avoid an expensive call to in_broadcast()
656 		 * in ip_output() most of the time (because the route passed
657 		 * to ip_output() is almost always a host route). */
658 		if (rt->rt_flags & RTF_HOST &&
659 		    !(rt->rt_flags & RTF_BROADCAST) &&
660 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
661 			rt->rt_flags |= RTF_BROADCAST;
662 		/* There is little point in resolving the broadcast address */
663 		if (rt->rt_flags & RTF_BROADCAST)
664 			break;
665 
666 		/*
667 		 * Case 2:  This route may come from cloning, or a manual route
668 		 * add with a LL address.
669 		 */
670 		switch (ifp->if_type) {
671 #if NTOKEN > 0
672 		case IFT_ISO88025:
673 			allocsize = sizeof(*la) + sizeof(struct token_rif);
674 			break;
675 #endif /* NTOKEN > 0 */
676 		default:
677 			allocsize = sizeof(*la);
678 		}
679 		R_Malloc(la, struct llinfo_arp *, allocsize);
680 		rt->rt_llinfo = (void *)la;
681 		if (la == NULL) {
682 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
683 			break;
684 		}
685 		arp_inuse++, arp_allocated++;
686 		memset(la, 0, allocsize);
687 		la->la_rt = rt;
688 		rt->rt_flags |= RTF_LLINFO;
689 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
690 
691 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
692 		while (ia && ia->ia_ifp != ifp)
693 			NEXT_IA_WITH_SAME_ADDR(ia);
694 		if (ia) {
695 			/*
696 			 * This test used to be
697 			 *	if (lo0ifp->if_flags & IFF_UP)
698 			 * It allowed local traffic to be forced through
699 			 * the hardware by configuring the loopback down.
700 			 * However, it causes problems during network
701 			 * configuration for boards that can't receive
702 			 * packets they send.  It is now necessary to clear
703 			 * "useloopback" and remove the route to force
704 			 * traffic out to the hardware.
705 			 *
706 			 * In 4.4BSD, the above "if" statement checked
707 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
708 			 * to the current form so that we can provide a
709 			 * better support for multiple IPv4 addresses on a
710 			 * interface.
711 			 */
712 			rt->rt_expire = 0;
713 			if (sockaddr_dl_init(satosdl(gate), gate->sa_len,
714 			    ifp->if_index, ifp->if_type, NULL, 0,
715 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen) == NULL) {
716 				panic("%s(%s): sockaddr_dl_init cannot fail",
717 				    __func__, ifp->if_xname);
718 			}
719 			if (useloopback)
720 				ifp = rt->rt_ifp = lo0ifp;
721 			/*
722 			 * make sure to set rt->rt_ifa to the interface
723 			 * address we are using, otherwise we will have trouble
724 			 * with source address selection.
725 			 */
726 			ifa = &ia->ia_ifa;
727 			if (ifa != rt->rt_ifa)
728 				rt_replace_ifa(rt, ifa);
729 		}
730 		break;
731 
732 	case RTM_DELETE:
733 		if (la == NULL)
734 			break;
735 		arp_inuse--;
736 		LIST_REMOVE(la, la_list);
737 		rt->rt_llinfo = NULL;
738 		rt->rt_flags &= ~RTF_LLINFO;
739 
740 		s = splnet();
741 		mold = la->la_hold;
742 		la->la_hold = NULL;
743 		splx(s);
744 
745 		if (mold)
746 			m_freem(mold);
747 
748 		Free((void *)la);
749 	}
750 	ARP_UNLOCK();
751 }
752 
753 /*
754  * Broadcast an ARP request. Caller specifies:
755  *	- arp header source ip address
756  *	- arp header target ip address
757  *	- arp header source ethernet address
758  */
759 void
760 arprequest(struct ifnet *ifp,
761     const struct in_addr *sip, const struct in_addr *tip,
762     const u_int8_t *enaddr)
763 {
764 	struct mbuf *m;
765 	struct arphdr *ah;
766 	struct sockaddr sa;
767 	uint64_t *arps;
768 
769 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
770 		return;
771 	MCLAIM(m, &arpdomain.dom_mowner);
772 	switch (ifp->if_type) {
773 	case IFT_IEEE1394:
774 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
775 		    ifp->if_addrlen;
776 		break;
777 	default:
778 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
779 		    2 * ifp->if_addrlen;
780 		break;
781 	}
782 	m->m_pkthdr.len = m->m_len;
783 	MH_ALIGN(m, m->m_len);
784 	ah = mtod(m, struct arphdr *);
785 	memset(ah, 0, m->m_len);
786 	switch (ifp->if_type) {
787 	case IFT_IEEE1394:	/* RFC2734 */
788 		/* fill it now for ar_tpa computation */
789 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
790 		break;
791 	default:
792 		/* ifp->if_output will fill ar_hrd */
793 		break;
794 	}
795 	ah->ar_pro = htons(ETHERTYPE_IP);
796 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
797 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
798 	ah->ar_op = htons(ARPOP_REQUEST);
799 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
800 	memcpy(ar_spa(ah), sip, ah->ar_pln);
801 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
802 	sa.sa_family = AF_ARP;
803 	sa.sa_len = 2;
804 	m->m_flags |= M_BCAST;
805 	arps = ARP_STAT_GETREF();
806 	arps[ARP_STAT_SNDTOTAL]++;
807 	arps[ARP_STAT_SENDREQUEST]++;
808 	ARP_STAT_PUTREF();
809 	(*ifp->if_output)(ifp, m, &sa, NULL);
810 }
811 
812 /*
813  * Resolve an IP address into an ethernet address.  If success,
814  * desten is filled in.  If there is no entry in arptab,
815  * set one up and broadcast a request for the IP address.
816  * Hold onto this mbuf and resend it once the address
817  * is finally resolved.  A return value of 1 indicates
818  * that desten has been filled in and the packet should be sent
819  * normally; a 0 return indicates that the packet has been
820  * taken over here, either now or for later transmission.
821  */
822 int
823 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
824     const struct sockaddr *dst, u_char *desten)
825 {
826 	struct llinfo_arp *la;
827 	const struct sockaddr_dl *sdl;
828 	struct mbuf *mold;
829 	int s;
830 
831 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
832 		rt = la->la_rt;
833 
834 	if (la == NULL || rt == NULL) {
835 		ARP_STATINC(ARP_STAT_ALLOCFAIL);
836 		log(LOG_DEBUG,
837 		    "arpresolve: can't allocate llinfo on %s for %s\n",
838 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
839 		m_freem(m);
840 		return 0;
841 	}
842 	sdl = satocsdl(rt->rt_gateway);
843 	/*
844 	 * Check the address family and length is valid, the address
845 	 * is resolved; otherwise, try to resolve.
846 	 */
847 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
848 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
849 		memcpy(desten, CLLADDR(sdl),
850 		    min(sdl->sdl_alen, ifp->if_addrlen));
851 		rt->rt_pksent = time_second; /* Time for last pkt sent */
852 		return 1;
853 	}
854 	/*
855 	 * There is an arptab entry, but no ethernet address
856 	 * response yet.  Replace the held mbuf with this
857 	 * latest one.
858 	 */
859 
860 	ARP_STATINC(ARP_STAT_DFRTOTAL);
861 	s = splnet();
862 	mold = la->la_hold;
863 	la->la_hold = m;
864 	splx(s);
865 
866 	if (mold) {
867 		ARP_STATINC(ARP_STAT_DFRDROPPED);
868 		m_freem(mold);
869 	}
870 
871 	/*
872 	 * Re-send the ARP request when appropriate.
873 	 */
874 #ifdef	DIAGNOSTIC
875 	if (rt->rt_expire == 0) {
876 		/* This should never happen. (Should it? -gwr) */
877 		printf("arpresolve: unresolved and rt_expire == 0\n");
878 		/* Set expiration time to now (expired). */
879 		rt->rt_expire = time_second;
880 	}
881 #endif
882 	if (rt->rt_expire) {
883 		rt->rt_flags &= ~RTF_REJECT;
884 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
885 			rt->rt_expire = time_second;
886 			if (la->la_asked++ < arp_maxtries) {
887 				arprequest(ifp,
888 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
889 				    &satocsin(dst)->sin_addr,
890 #if NCARP > 0
891 				    (rt->rt_ifp->if_type == IFT_CARP) ?
892 				    CLLADDR(rt->rt_ifp->if_sadl):
893 #endif
894 				    CLLADDR(ifp->if_sadl));
895 			} else {
896 				rt->rt_flags |= RTF_REJECT;
897 				rt->rt_expire += arpt_down;
898 				la->la_asked = 0;
899 			}
900 		}
901 	}
902 	return 0;
903 }
904 
905 /*
906  * Common length and type checks are done here,
907  * then the protocol-specific routine is called.
908  */
909 void
910 arpintr(void)
911 {
912 	struct mbuf *m;
913 	struct arphdr *ar;
914 	int s;
915 	int arplen;
916 
917 	mutex_enter(softnet_lock);
918 	KERNEL_LOCK(1, NULL);
919 	while (arpintrq.ifq_head) {
920 		s = splnet();
921 		IF_DEQUEUE(&arpintrq, m);
922 		splx(s);
923 		if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
924 			panic("arpintr");
925 
926 		MCLAIM(m, &arpdomain.dom_mowner);
927 		ARP_STATINC(ARP_STAT_RCVTOTAL);
928 
929 		/*
930 		 * First, make sure we have at least struct arphdr.
931 		 */
932 		if (m->m_len < sizeof(struct arphdr) ||
933 		    (ar = mtod(m, struct arphdr *)) == NULL)
934 			goto badlen;
935 
936 		switch (m->m_pkthdr.rcvif->if_type) {
937 		case IFT_IEEE1394:
938 			arplen = sizeof(struct arphdr) +
939 			    ar->ar_hln + 2 * ar->ar_pln;
940 			break;
941 		default:
942 			arplen = sizeof(struct arphdr) +
943 			    2 * ar->ar_hln + 2 * ar->ar_pln;
944 			break;
945 		}
946 
947 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
948 		    m->m_len >= arplen)
949 			switch (ntohs(ar->ar_pro)) {
950 			case ETHERTYPE_IP:
951 			case ETHERTYPE_IPTRAILERS:
952 				in_arpinput(m);
953 				continue;
954 			default:
955 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
956 			}
957 		else {
958 badlen:
959 			ARP_STATINC(ARP_STAT_RCVBADLEN);
960 		}
961 		m_freem(m);
962 	}
963 	KERNEL_UNLOCK_ONE(NULL);
964 	mutex_exit(softnet_lock);
965 }
966 
967 /*
968  * ARP for Internet protocols on 10 Mb/s Ethernet.
969  * Algorithm is that given in RFC 826.
970  * In addition, a sanity check is performed on the sender
971  * protocol address, to catch impersonators.
972  * We no longer handle negotiations for use of trailer protocol:
973  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
974  * along with IP replies if we wanted trailers sent to us,
975  * and also sent them in response to IP replies.
976  * This allowed either end to announce the desire to receive
977  * trailer packets.
978  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
979  * but formerly didn't normally send requests.
980  */
981 static void
982 in_arpinput(struct mbuf *m)
983 {
984 	struct arphdr *ah;
985 	struct ifnet *ifp = m->m_pkthdr.rcvif;
986 	struct llinfo_arp *la = NULL;
987 	struct rtentry  *rt;
988 	struct in_ifaddr *ia;
989 #if NBRIDGE > 0
990 	struct in_ifaddr *bridge_ia = NULL;
991 #endif
992 #if NCARP > 0
993 	u_int32_t count = 0, index = 0;
994 #endif
995 	struct sockaddr_dl *sdl;
996 	struct sockaddr sa;
997 	struct in_addr isaddr, itaddr, myaddr;
998 	int op;
999 	struct mbuf *mold;
1000 	void *tha;
1001 	int s;
1002 	uint64_t *arps;
1003 
1004 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
1005 		goto out;
1006 	ah = mtod(m, struct arphdr *);
1007 	op = ntohs(ah->ar_op);
1008 
1009 	/*
1010 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
1011 	 * ar_tpa().
1012 	 */
1013 	switch (ifp->if_type) {
1014 	case IFT_IEEE1394:
1015 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
1016 			;
1017 		else {
1018 			/* XXX this is to make sure we compute ar_tha right */
1019 			/* XXX check ar_hrd more strictly? */
1020 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
1021 		}
1022 		break;
1023 	default:
1024 		/* XXX check ar_hrd? */
1025 		break;
1026 	}
1027 
1028 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
1029 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
1030 
1031 	if (m->m_flags & (M_BCAST|M_MCAST))
1032 		ARP_STATINC(ARP_STAT_RCVMCAST);
1033 
1034 
1035 	/*
1036 	 * Search for a matching interface address
1037 	 * or any address on the interface to use
1038 	 * as a dummy address in the rest of this function
1039 	 */
1040 
1041 	INADDR_TO_IA(itaddr, ia);
1042 	while (ia != NULL) {
1043 #if NCARP > 0
1044 		if (ia->ia_ifp->if_type == IFT_CARP &&
1045 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1046 		    (IFF_UP|IFF_RUNNING))) {
1047 			index++;
1048 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
1049 			    carp_iamatch(ia, ar_sha(ah),
1050 			    &count, index)) {
1051 				break;
1052 				}
1053 		} else
1054 #endif
1055 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
1056 				break;
1057 #if NBRIDGE > 0
1058 		/*
1059 		 * If the interface we received the packet on
1060 		 * is part of a bridge, check to see if we need
1061 		 * to "bridge" the packet to ourselves at this
1062 		 * layer.  Note we still prefer a perfect match,
1063 		 * but allow this weaker match if necessary.
1064 		 */
1065 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
1066 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
1067 			bridge_ia = ia;
1068 #endif /* NBRIDGE > 0 */
1069 
1070 		NEXT_IA_WITH_SAME_ADDR(ia);
1071 	}
1072 
1073 #if NBRIDGE > 0
1074 	if (ia == NULL && bridge_ia != NULL) {
1075 		ia = bridge_ia;
1076 		ifp = bridge_ia->ia_ifp;
1077 	}
1078 #endif
1079 
1080 	if (ia == NULL) {
1081 		INADDR_TO_IA(isaddr, ia);
1082 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
1083 			NEXT_IA_WITH_SAME_ADDR(ia);
1084 
1085 		if (ia == NULL) {
1086 			IFP_TO_IA(ifp, ia);
1087 			if (ia == NULL) {
1088 				ARP_STATINC(ARP_STAT_RCVNOINT);
1089 				goto out;
1090 			}
1091 		}
1092 	}
1093 
1094 	myaddr = ia->ia_addr.sin_addr;
1095 
1096 	/* XXX checks for bridge case? */
1097 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
1098 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1099 		goto out;	/* it's from me, ignore it. */
1100 	}
1101 
1102 	/* XXX checks for bridge case? */
1103 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1104 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1105 		log(LOG_ERR,
1106 		    "%s: arp: link address is broadcast for IP address %s!\n",
1107 		    ifp->if_xname, in_fmtaddr(isaddr));
1108 		goto out;
1109 	}
1110 
1111 	/*
1112 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
1113 	 */
1114 	if (in_nullhost(isaddr))
1115 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
1116 	else if (in_hosteq(isaddr, myaddr))
1117 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1118 
1119 	if (in_nullhost(itaddr))
1120 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
1121 
1122 	/* DAD check, RFC 5227 2.1.1, Probe Details */
1123 	if (in_hosteq(isaddr, myaddr) ||
1124 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)))
1125 	{
1126 		/* If our address is tentative, mark it as duplicated */
1127 		if (ia->ia4_flags & IN_IFF_TENTATIVE)
1128 			arp_dad_duplicated((struct ifaddr *)ia);
1129 		/* If our address is unuseable, don't reply */
1130 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1131 			goto out;
1132 	}
1133 
1134 	/*
1135 	 * If the target IP address is zero, ignore the packet.
1136 	 * This prevents the code below from tring to answer
1137 	 * when we are using IP address zero (booting).
1138 	 */
1139 	if (in_nullhost(itaddr))
1140 		goto out;
1141 
1142 	if (in_nullhost(isaddr))
1143 		goto reply;
1144 
1145 	if (in_hosteq(isaddr, myaddr)) {
1146 		log(LOG_ERR,
1147 		   "duplicate IP address %s sent from link address %s\n",
1148 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
1149 		itaddr = myaddr;
1150 		goto reply;
1151 	}
1152 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1153 	if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
1154 		if (sdl->sdl_alen &&
1155 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
1156 			if (rt->rt_flags & RTF_STATIC) {
1157 				ARP_STATINC(ARP_STAT_RCVOVERPERM);
1158 				if (!log_permanent_modify)
1159 					goto out;
1160 				log(LOG_INFO,
1161 				    "%s tried to overwrite permanent arp info"
1162 				    " for %s\n",
1163 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1164 				    in_fmtaddr(isaddr));
1165 				goto out;
1166 			} else if (rt->rt_ifp != ifp) {
1167 				ARP_STATINC(ARP_STAT_RCVOVERINT);
1168 				if (!log_wrong_iface)
1169 					goto out;
1170 				log(LOG_INFO,
1171 				    "%s on %s tried to overwrite "
1172 				    "arp info for %s on %s\n",
1173 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1174 				    ifp->if_xname, in_fmtaddr(isaddr),
1175 				    rt->rt_ifp->if_xname);
1176 				    goto out;
1177 			} else {
1178 				ARP_STATINC(ARP_STAT_RCVOVER);
1179 				if (log_movements)
1180 					log(LOG_INFO, "arp info overwritten "
1181 					    "for %s by %s\n",
1182 					    in_fmtaddr(isaddr),
1183 					    lla_snprintf(ar_sha(ah),
1184 					    ah->ar_hln));
1185 			}
1186 		}
1187 		/*
1188 		 * sanity check for the address length.
1189 		 * XXX this does not work for protocols with variable address
1190 		 * length. -is
1191 		 */
1192 		if (sdl->sdl_alen &&
1193 		    sdl->sdl_alen != ah->ar_hln) {
1194 			ARP_STATINC(ARP_STAT_RCVLENCHG);
1195 			log(LOG_WARNING,
1196 			    "arp from %s: new addr len %d, was %d\n",
1197 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1198 		}
1199 		if (ifp->if_addrlen != ah->ar_hln) {
1200 			ARP_STATINC(ARP_STAT_RCVBADLEN);
1201 			log(LOG_WARNING,
1202 			    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
1203 			    in_fmtaddr(isaddr), ah->ar_hln,
1204 			    ifp->if_addrlen);
1205 			goto reply;
1206 		}
1207 #if NTOKEN > 0
1208 		/*
1209 		 * XXX uses m_data and assumes the complete answer including
1210 		 * XXX token-ring headers is in the same buf
1211 		 */
1212 		if (ifp->if_type == IFT_ISO88025) {
1213 			struct token_header *trh;
1214 
1215 			trh = (struct token_header *)M_TRHSTART(m);
1216 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1217 				struct token_rif	*rif;
1218 				size_t	riflen;
1219 
1220 				rif = TOKEN_RIF(trh);
1221 				riflen = (ntohs(rif->tr_rcf) &
1222 				    TOKEN_RCF_LEN_MASK) >> 8;
1223 
1224 				if (riflen > 2 &&
1225 				    riflen < sizeof(struct token_rif) &&
1226 				    (riflen & 1) == 0) {
1227 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1228 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1229 					memcpy(TOKEN_RIF(la), rif, riflen);
1230 				}
1231 			}
1232 		}
1233 #endif /* NTOKEN > 0 */
1234 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
1235 		    ah->ar_hln);
1236 		if (rt->rt_expire)
1237 			rt->rt_expire = time_second + arpt_keep;
1238 		rt->rt_flags &= ~RTF_REJECT;
1239 		la->la_asked = 0;
1240 
1241 		s = splnet();
1242 		mold = la->la_hold;
1243 		la->la_hold = NULL;
1244 		splx(s);
1245 
1246 		if (mold) {
1247 			ARP_STATINC(ARP_STAT_DFRSENT);
1248 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
1249 		}
1250 	}
1251 reply:
1252 	if (op != ARPOP_REQUEST) {
1253 		if (op == ARPOP_REPLY)
1254 			ARP_STATINC(ARP_STAT_RCVREPLY);
1255 	out:
1256 		m_freem(m);
1257 		return;
1258 	}
1259 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1260 	if (in_hosteq(itaddr, myaddr)) {
1261 		/* If our address is unuseable, don't reply */
1262 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1263 			goto out;
1264 		/* I am the target */
1265 		tha = ar_tha(ah);
1266 		if (tha)
1267 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1268 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1269 	} else {
1270 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1271 		if (la == NULL)
1272 			goto out;
1273 		rt = la->la_rt;
1274 		if (rt->rt_ifp->if_type == IFT_CARP &&
1275 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1276 			goto out;
1277 		tha = ar_tha(ah);
1278 		if (tha)
1279 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1280 		sdl = satosdl(rt->rt_gateway);
1281 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
1282 	}
1283 
1284 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1285 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1286 	ah->ar_op = htons(ARPOP_REPLY);
1287 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1288 	switch (ifp->if_type) {
1289 	case IFT_IEEE1394:
1290 		/*
1291 		 * ieee1394 arp reply is broadcast
1292 		 */
1293 		m->m_flags &= ~M_MCAST;
1294 		m->m_flags |= M_BCAST;
1295 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1296 		break;
1297 
1298 	default:
1299 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1300 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1301 		break;
1302 	}
1303 	m->m_pkthdr.len = m->m_len;
1304 	sa.sa_family = AF_ARP;
1305 	sa.sa_len = 2;
1306 	arps = ARP_STAT_GETREF();
1307 	arps[ARP_STAT_SNDTOTAL]++;
1308 	arps[ARP_STAT_SNDREPLY]++;
1309 	ARP_STAT_PUTREF();
1310 	(*ifp->if_output)(ifp, m, &sa, NULL);
1311 	return;
1312 }
1313 
1314 /*
1315  * Free an arp entry.
1316  */
1317 static void arptfree(struct llinfo_arp *la)
1318 {
1319 	struct rtentry *rt = la->la_rt;
1320 	struct sockaddr_dl *sdl;
1321 
1322 	ARP_LOCK_CHECK();
1323 
1324 	if (rt == NULL)
1325 		panic("arptfree");
1326 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
1327 	    sdl->sdl_family == AF_LINK) {
1328 		sdl->sdl_alen = 0;
1329 		la->la_asked = 0;
1330 		rt->rt_flags &= ~RTF_REJECT;
1331 		return;
1332 	}
1333 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1334 }
1335 
1336 static struct llinfo_arp *
1337 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
1338 {
1339 	return arplookup1(m, addr, create, proxy, NULL);
1340 }
1341 
1342 /*
1343  * Lookup or enter a new address in arptab.
1344  */
1345 static struct llinfo_arp *
1346 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
1347     struct rtentry *rt0)
1348 {
1349 	struct arphdr *ah;
1350 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1351 	struct rtentry *rt;
1352 	struct sockaddr_inarp sin;
1353 	const char *why = NULL;
1354 
1355 	ah = mtod(m, struct arphdr *);
1356 	if (rt0 == NULL) {
1357 		memset(&sin, 0, sizeof(sin));
1358 		sin.sin_len = sizeof(sin);
1359 		sin.sin_family = AF_INET;
1360 		sin.sin_addr = *addr;
1361 		sin.sin_other = proxy ? SIN_PROXY : 0;
1362 		rt = rtalloc1(sintosa(&sin), create);
1363 		if (rt == NULL)
1364 			return NULL;
1365 		rt->rt_refcnt--;
1366 	} else
1367 		rt = rt0;
1368 
1369 #define	IS_LLINFO(__rt)							  \
1370 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
1371 	 (__rt)->rt_gateway->sa_family == AF_LINK)
1372 
1373 
1374 	if (IS_LLINFO(rt))
1375 		return (struct llinfo_arp *)rt->rt_llinfo;
1376 
1377 	if (create) {
1378 		if (rt->rt_flags & RTF_GATEWAY) {
1379 			if (log_unknown_network)
1380 				why = "host is not on local network";
1381 		} else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1382 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
1383 			why = "could not allocate llinfo";
1384 		} else
1385 			why = "gateway route is not ours";
1386 		if (why) {
1387 			log(LOG_DEBUG, "arplookup: unable to enter address"
1388 			    " for %s@%s on %s (%s)\n", in_fmtaddr(*addr),
1389 			    lla_snprintf(ar_sha(ah), ah->ar_hln),
1390 			    (ifp) ? ifp->if_xname : "null", why);
1391 		}
1392 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1393 			rtrequest(RTM_DELETE, rt_getkey(rt),
1394 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1395 		}
1396 	}
1397 	return NULL;
1398 }
1399 
1400 int
1401 arpioctl(u_long cmd, void *data)
1402 {
1403 
1404 	return EOPNOTSUPP;
1405 }
1406 
1407 void
1408 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1409 {
1410 	struct in_addr *ip;
1411 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1412 
1413 	/*
1414 	 * Warn the user if another station has this IP address,
1415 	 * but only if the interface IP address is not zero.
1416 	 */
1417 	ip = &IA_SIN(ifa)->sin_addr;
1418 	if (!in_nullhost(*ip) &&
1419 	    (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) == 0)
1420 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
1421 
1422 	ifa->ifa_rtrequest = arp_rtrequest;
1423 	ifa->ifa_flags |= RTF_CLONING;
1424 
1425 	/* ARP will handle DAD for this address. */
1426 	if (ia->ia4_flags & IN_IFF_TRYTENTATIVE) {
1427 		ia->ia4_flags |= IN_IFF_TENTATIVE;
1428 		ia->ia_dad_start = arp_dad_start;
1429 		ia->ia_dad_stop = arp_dad_stop;
1430 	}
1431 }
1432 
1433 TAILQ_HEAD(dadq_head, dadq);
1434 struct dadq {
1435 	TAILQ_ENTRY(dadq) dad_list;
1436 	struct ifaddr *dad_ifa;
1437 	int dad_count;		/* max ARP to send */
1438 	int dad_arp_tcount;	/* # of trials to send ARP */
1439 	int dad_arp_ocount;	/* ARP sent so far */
1440 	int dad_arp_announce;	/* max ARP announcements */
1441 	int dad_arp_acount;	/* # of announcements */
1442 	struct callout dad_timer_ch;
1443 };
1444 MALLOC_JUSTDEFINE(M_IPARP, "ARP DAD", "ARP DAD Structure");
1445 
1446 static struct dadq_head dadq;
1447 static int dad_init = 0;
1448 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1449 
1450 static struct dadq *
1451 arp_dad_find(struct ifaddr *ifa)
1452 {
1453 	struct dadq *dp;
1454 
1455 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1456 		if (dp->dad_ifa == ifa)
1457 			return dp;
1458 	}
1459 	return NULL;
1460 }
1461 
1462 static void
1463 arp_dad_starttimer(struct dadq *dp, int ticks)
1464 {
1465 
1466 	callout_reset(&dp->dad_timer_ch, ticks,
1467 	    (void (*)(void *))arp_dad_timer, (void *)dp->dad_ifa);
1468 }
1469 
1470 static void
1471 arp_dad_stoptimer(struct dadq *dp)
1472 {
1473 
1474 	callout_stop(&dp->dad_timer_ch);
1475 }
1476 
1477 static void
1478 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1479 {
1480 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1481 	struct ifnet *ifp = ifa->ifa_ifp;
1482 	struct in_addr sip;
1483 
1484 	dp->dad_arp_tcount++;
1485 	if ((ifp->if_flags & IFF_UP) == 0)
1486 		return;
1487 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1488 		return;
1489 
1490 	dp->dad_arp_tcount = 0;
1491 	dp->dad_arp_ocount++;
1492 
1493 	memset(&sip, 0, sizeof(sip));
1494 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1495 	    CLLADDR(ifa->ifa_ifp->if_sadl));
1496 }
1497 
1498 /*
1499  * Start Duplicate Address Detection (DAD) for specified interface address.
1500  */
1501 static void
1502 arp_dad_start(struct ifaddr *ifa)
1503 {
1504 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1505 	struct dadq *dp;
1506 
1507 	if (!dad_init) {
1508 		TAILQ_INIT(&dadq);
1509 		dad_init++;
1510 	}
1511 
1512 	/*
1513 	 * If we don't need DAD, don't do it.
1514 	 * - DAD is disabled (ip_dad_count == 0)
1515 	 */
1516 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1517 		log(LOG_DEBUG,
1518 			"arp_dad_start: called with non-tentative address "
1519 			"%s(%s)\n",
1520 			in_fmtaddr(ia->ia_addr.sin_addr),
1521 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1522 		return;
1523 	}
1524 	if (!ip_dad_count) {
1525 		struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
1526 
1527 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1528 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1529 		arprequest(ifa->ifa_ifp, ip, ip,
1530 		    CLLADDR(ifa->ifa_ifp->if_sadl));
1531 		return;
1532 	}
1533 	if (ifa->ifa_ifp == NULL)
1534 		panic("arp_dad_start: ifa->ifa_ifp == NULL");
1535 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1536 		return;
1537 	if (arp_dad_find(ifa) != NULL) {
1538 		/* DAD already in progress */
1539 		return;
1540 	}
1541 
1542 	dp = malloc(sizeof(*dp), M_IPARP, M_NOWAIT);
1543 	if (dp == NULL) {
1544 		log(LOG_ERR, "arp_dad_start: memory allocation failed for "
1545 			"%s(%s)\n",
1546 			in_fmtaddr(ia->ia_addr.sin_addr),
1547 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1548 		return;
1549 	}
1550 	memset(dp, 0, sizeof(*dp));
1551 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1552 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1553 
1554 	arplog((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1555 	    in_fmtaddr(ia->ia_addr.sin_addr)));
1556 
1557 	/*
1558 	 * Send ARP packet for DAD, ip_dad_count times.
1559 	 * Note that we must delay the first transmission.
1560 	 */
1561 	dp->dad_ifa = ifa;
1562 	ifaref(ifa);	/* just for safety */
1563 	dp->dad_count = ip_dad_count;
1564 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1565 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1566 
1567 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1568 }
1569 
1570 /*
1571  * terminate DAD unconditionally.  used for address removals.
1572  */
1573 static void
1574 arp_dad_stop(struct ifaddr *ifa)
1575 {
1576 	struct dadq *dp;
1577 
1578 	if (!dad_init)
1579 		return;
1580 	dp = arp_dad_find(ifa);
1581 	if (dp == NULL) {
1582 		/* DAD wasn't started yet */
1583 		return;
1584 	}
1585 
1586 	arp_dad_stoptimer(dp);
1587 
1588 	TAILQ_REMOVE(&dadq, dp, dad_list);
1589 	free(dp, M_IPARP);
1590 	dp = NULL;
1591 	ifafree(ifa);
1592 }
1593 
1594 static void
1595 arp_dad_timer(struct ifaddr *ifa)
1596 {
1597 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1598 	struct dadq *dp;
1599 	struct in_addr *ip;
1600 
1601 	mutex_enter(softnet_lock);
1602 	KERNEL_LOCK(1, NULL);
1603 
1604 	/* Sanity check */
1605 	if (ia == NULL) {
1606 		log(LOG_ERR, "arp_dad_timer: called with null parameter\n");
1607 		goto done;
1608 	}
1609 	dp = arp_dad_find(ifa);
1610 	if (dp == NULL) {
1611 		log(LOG_ERR, "arp_dad_timer: DAD structure not found\n");
1612 		goto done;
1613 	}
1614 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1615 		log(LOG_ERR, "nd4_dad_timer: called with duplicate address "
1616 			"%s(%s)\n",
1617 			in_fmtaddr(ia->ia_addr.sin_addr),
1618 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1619 		goto done;
1620 	}
1621 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0){
1622 		log(LOG_ERR, "arp_dad_timer: called with non-tentative address "
1623 			"%s(%s)\n",
1624 			in_fmtaddr(ia->ia_addr.sin_addr),
1625 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1626 		goto done;
1627 	}
1628 
1629 	/* timeouted with IFF_{RUNNING,UP} check */
1630 	if (dp->dad_arp_tcount > dad_maxtry) {
1631 		arplog((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1632 			if_name(ifa->ifa_ifp)));
1633 
1634 		TAILQ_REMOVE(&dadq, dp, dad_list);
1635 		free(dp, M_IPARP);
1636 		dp = NULL;
1637 		ifafree(ifa);
1638 		goto done;
1639 	}
1640 
1641 	/* Need more checks? */
1642 	if (dp->dad_arp_ocount < dp->dad_count) {
1643 		int adelay;
1644 
1645 		/*
1646 		 * We have more ARP to go.  Send ARP packet for DAD.
1647 		 */
1648 		arp_dad_output(dp, ifa);
1649 		if (dp->dad_arp_ocount < dp->dad_count)
1650 			adelay = (PROBE_MIN * hz) +
1651 			    (cprng_fast32() %
1652 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1653 		else
1654 			adelay = ANNOUNCE_WAIT * hz;
1655 		arp_dad_starttimer(dp, adelay);
1656 		goto done;
1657 	} else if (dp->dad_arp_acount == 0) {
1658 		/*
1659 		 * We are done with DAD.
1660 		 * No duplicate address found.
1661 		 */
1662 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1663 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1664 		arplog((LOG_DEBUG,
1665 		    "%s: DAD complete for %s - no duplicates found\n",
1666 		    if_name(ifa->ifa_ifp),
1667 		    in_fmtaddr(ia->ia_addr.sin_addr)));
1668 		dp->dad_arp_announce = ANNOUNCE_NUM;
1669 		goto announce;
1670 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1671 announce:
1672 		/*
1673 		 * Announce the address.
1674 		 */
1675 		ip = &IA_SIN(ifa)->sin_addr;
1676 		arprequest(ifa->ifa_ifp, ip, ip,
1677 		    CLLADDR(ifa->ifa_ifp->if_sadl));
1678 		dp->dad_arp_acount++;
1679 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1680 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1681 			goto done;
1682 		}
1683 		arplog((LOG_DEBUG,
1684 		    "%s: ARP announcement complete for %s\n",
1685 		    if_name(ifa->ifa_ifp),
1686 		    in_fmtaddr(ia->ia_addr.sin_addr)));
1687 	}
1688 
1689 	TAILQ_REMOVE(&dadq, dp, dad_list);
1690 	free(dp, M_IPARP);
1691 	dp = NULL;
1692 	ifafree(ifa);
1693 
1694 done:
1695 	KERNEL_UNLOCK_ONE(NULL);
1696 	mutex_exit(softnet_lock);
1697 }
1698 
1699 static void
1700 arp_dad_duplicated(struct ifaddr *ifa)
1701 {
1702 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1703 	struct ifnet *ifp;
1704 	struct dadq *dp;
1705 
1706 	dp = arp_dad_find(ifa);
1707 	if (dp == NULL) {
1708 		log(LOG_ERR, "arp_dad_duplicated: DAD structure not found\n");
1709 		return;
1710 	}
1711 
1712 	ifp = ifa->ifa_ifp;
1713 	log(LOG_ERR, "%s: DAD detected duplicate IPv4 address %s: "
1714 	    "ARP out=%d\n",
1715 	    if_name(ifp), in_fmtaddr(ia->ia_addr.sin_addr),
1716 	    dp->dad_arp_ocount);
1717 
1718 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1719 	ia->ia4_flags |= IN_IFF_DUPLICATED;
1720 
1721 	/* We are done with DAD, with duplicated address found. (failure) */
1722 	arp_dad_stoptimer(dp);
1723 
1724 	/* Inform the routing socket that DAD has completed */
1725 	rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1726 
1727 	TAILQ_REMOVE(&dadq, dp, dad_list);
1728 	free(dp, M_IPARP);
1729 	dp = NULL;
1730 	ifafree(ifa);
1731 }
1732 
1733 /*
1734  * Called from 10 Mb/s Ethernet interrupt handlers
1735  * when ether packet type ETHERTYPE_REVARP
1736  * is received.  Common length and type checks are done here,
1737  * then the protocol-specific routine is called.
1738  */
1739 void
1740 revarpinput(struct mbuf *m)
1741 {
1742 	struct arphdr *ar;
1743 
1744 	if (m->m_len < sizeof(struct arphdr))
1745 		goto out;
1746 	ar = mtod(m, struct arphdr *);
1747 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1748 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1749 		goto out;
1750 #endif
1751 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1752 		goto out;
1753 	switch (ntohs(ar->ar_pro)) {
1754 	case ETHERTYPE_IP:
1755 	case ETHERTYPE_IPTRAILERS:
1756 		in_revarpinput(m);
1757 		return;
1758 
1759 	default:
1760 		break;
1761 	}
1762 out:
1763 	m_freem(m);
1764 }
1765 
1766 /*
1767  * RARP for Internet protocols on 10 Mb/s Ethernet.
1768  * Algorithm is that given in RFC 903.
1769  * We are only using for bootstrap purposes to get an ip address for one of
1770  * our interfaces.  Thus we support no user-interface.
1771  *
1772  * Since the contents of the RARP reply are specific to the interface that
1773  * sent the request, this code must ensure that they are properly associated.
1774  *
1775  * Note: also supports ARP via RARP packets, per the RFC.
1776  */
1777 void
1778 in_revarpinput(struct mbuf *m)
1779 {
1780 	struct ifnet *ifp;
1781 	struct arphdr *ah;
1782 	void *tha;
1783 	int op;
1784 
1785 	ah = mtod(m, struct arphdr *);
1786 	op = ntohs(ah->ar_op);
1787 
1788 	switch (m->m_pkthdr.rcvif->if_type) {
1789 	case IFT_IEEE1394:
1790 		/* ARP without target hardware address is not supported */
1791 		goto out;
1792 	default:
1793 		break;
1794 	}
1795 
1796 	switch (op) {
1797 	case ARPOP_REQUEST:
1798 	case ARPOP_REPLY:	/* per RFC */
1799 		in_arpinput(m);
1800 		return;
1801 	case ARPOP_REVREPLY:
1802 		break;
1803 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1804 	default:
1805 		goto out;
1806 	}
1807 	if (!revarp_in_progress)
1808 		goto out;
1809 	ifp = m->m_pkthdr.rcvif;
1810 	if (ifp != myip_ifp) /* !same interface */
1811 		goto out;
1812 	if (myip_initialized)
1813 		goto wake;
1814 	tha = ar_tha(ah);
1815 	if (tha == NULL)
1816 		goto out;
1817 	if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1818 		goto out;
1819 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1820 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1821 	myip_initialized = 1;
1822 wake:	/* Do wakeup every time in case it was missed. */
1823 	wakeup((void *)&myip);
1824 
1825 out:
1826 	m_freem(m);
1827 }
1828 
1829 /*
1830  * Send a RARP request for the ip address of the specified interface.
1831  * The request should be RFC 903-compliant.
1832  */
1833 void
1834 revarprequest(struct ifnet *ifp)
1835 {
1836 	struct sockaddr sa;
1837 	struct mbuf *m;
1838 	struct arphdr *ah;
1839 	void *tha;
1840 
1841 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1842 		return;
1843 	MCLAIM(m, &arpdomain.dom_mowner);
1844 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1845 	    2*ifp->if_addrlen;
1846 	m->m_pkthdr.len = m->m_len;
1847 	MH_ALIGN(m, m->m_len);
1848 	ah = mtod(m, struct arphdr *);
1849 	memset(ah, 0, m->m_len);
1850 	ah->ar_pro = htons(ETHERTYPE_IP);
1851 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1852 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1853 	ah->ar_op = htons(ARPOP_REVREQUEST);
1854 
1855 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1856 	tha = ar_tha(ah);
1857 	if (tha == NULL) {
1858 		m_free(m);
1859 		return;
1860 	}
1861 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1862 
1863 	sa.sa_family = AF_ARP;
1864 	sa.sa_len = 2;
1865 	m->m_flags |= M_BCAST;
1866 
1867 	KERNEL_LOCK(1, NULL);
1868 	(*ifp->if_output)(ifp, m, &sa, NULL);
1869 	KERNEL_UNLOCK_ONE(NULL);
1870 }
1871 
1872 /*
1873  * RARP for the ip address of the specified interface, but also
1874  * save the ip address of the server that sent the answer.
1875  * Timeout if no response is received.
1876  */
1877 int
1878 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1879     struct in_addr *clnt_in)
1880 {
1881 	int result, count = 20;
1882 
1883 	myip_initialized = 0;
1884 	myip_ifp = ifp;
1885 
1886 	revarp_in_progress = 1;
1887 	while (count--) {
1888 		revarprequest(ifp);
1889 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1890 		if (result != EWOULDBLOCK)
1891 			break;
1892 	}
1893 	revarp_in_progress = 0;
1894 
1895 	if (!myip_initialized)
1896 		return ENETUNREACH;
1897 
1898 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1899 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1900 	return 0;
1901 }
1902 
1903 
1904 
1905 #ifdef DDB
1906 
1907 #include <machine/db_machdep.h>
1908 #include <ddb/db_interface.h>
1909 #include <ddb/db_output.h>
1910 
1911 static void
1912 db_print_sa(const struct sockaddr *sa)
1913 {
1914 	int len;
1915 	const u_char *p;
1916 
1917 	if (sa == NULL) {
1918 		db_printf("[NULL]");
1919 		return;
1920 	}
1921 
1922 	p = (const u_char *)sa;
1923 	len = sa->sa_len;
1924 	db_printf("[");
1925 	while (len > 0) {
1926 		db_printf("%d", *p);
1927 		p++; len--;
1928 		if (len) db_printf(",");
1929 	}
1930 	db_printf("]\n");
1931 }
1932 
1933 static void
1934 db_print_ifa(struct ifaddr *ifa)
1935 {
1936 	if (ifa == NULL)
1937 		return;
1938 	db_printf("  ifa_addr=");
1939 	db_print_sa(ifa->ifa_addr);
1940 	db_printf("  ifa_dsta=");
1941 	db_print_sa(ifa->ifa_dstaddr);
1942 	db_printf("  ifa_mask=");
1943 	db_print_sa(ifa->ifa_netmask);
1944 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1945 			  ifa->ifa_flags,
1946 			  ifa->ifa_refcnt,
1947 			  ifa->ifa_metric);
1948 }
1949 
1950 static void
1951 db_print_llinfo(void *li)
1952 {
1953 	struct llinfo_arp *la;
1954 
1955 	if (li == NULL)
1956 		return;
1957 	la = (struct llinfo_arp *)li;
1958 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1959 			  la->la_rt, la->la_hold, la->la_asked);
1960 }
1961 
1962 /*
1963  * Function to pass to rt_walktree().
1964  * Return non-zero error to abort walk.
1965  */
1966 static int
1967 db_show_rtentry(struct rtentry *rt, void *w)
1968 {
1969 	db_printf("rtentry=%p", rt);
1970 
1971 	db_printf(" flags=0x%x refcnt=%d use=%"PRId64" expire=%"PRId64"\n",
1972 			  rt->rt_flags, rt->rt_refcnt,
1973 			  rt->rt_use, (uint64_t)rt->rt_expire);
1974 
1975 	db_printf(" key="); db_print_sa(rt_getkey(rt));
1976 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1977 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1978 
1979 	db_printf(" ifp=%p ", rt->rt_ifp);
1980 	if (rt->rt_ifp)
1981 		db_printf("(%s)", rt->rt_ifp->if_xname);
1982 	else
1983 		db_printf("(NULL)");
1984 
1985 	db_printf(" ifa=%p\n", rt->rt_ifa);
1986 	db_print_ifa(rt->rt_ifa);
1987 
1988 	db_printf(" gwroute=%p llinfo=%p\n",
1989 			  rt->rt_gwroute, rt->rt_llinfo);
1990 	db_print_llinfo(rt->rt_llinfo);
1991 
1992 	return 0;
1993 }
1994 
1995 /*
1996  * Function to print all the route trees.
1997  * Use this from ddb:  "show arptab"
1998  */
1999 void
2000 db_show_arptab(db_expr_t addr, bool have_addr,
2001     db_expr_t count, const char *modif)
2002 {
2003 	rt_walktree(AF_INET, db_show_rtentry, NULL);
2004 }
2005 #endif
2006 
2007 static int
2008 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
2009 {
2010 
2011 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2012 }
2013 
2014 static void
2015 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2016 {
2017 	const struct sysctlnode *node;
2018 
2019 	sysctl_createv(clog, 0, NULL, NULL,
2020 			CTLFLAG_PERMANENT,
2021 			CTLTYPE_NODE, "inet", NULL,
2022 			NULL, 0, NULL, 0,
2023 			CTL_NET, PF_INET, CTL_EOL);
2024 	sysctl_createv(clog, 0, NULL, &node,
2025 			CTLFLAG_PERMANENT,
2026 			CTLTYPE_NODE, "arp",
2027 			SYSCTL_DESCR("Address Resolution Protocol"),
2028 			NULL, 0, NULL, 0,
2029 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2030 
2031 	sysctl_createv(clog, 0, NULL, NULL,
2032 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2033 			CTLTYPE_INT, "prune",
2034 			SYSCTL_DESCR("ARP cache pruning interval in seconds"),
2035 			NULL, 0, &arpt_prune, 0,
2036 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2037 
2038 	sysctl_createv(clog, 0, NULL, NULL,
2039 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2040 			CTLTYPE_INT, "keep",
2041 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
2042 			NULL, 0, &arpt_keep, 0,
2043 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2044 
2045 	sysctl_createv(clog, 0, NULL, NULL,
2046 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2047 			CTLTYPE_INT, "down",
2048 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
2049 			NULL, 0, &arpt_down, 0,
2050 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2051 
2052 	sysctl_createv(clog, 0, NULL, NULL,
2053 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2054 			CTLTYPE_INT, "refresh",
2055 			SYSCTL_DESCR("ARP entry refresh interval"),
2056 			NULL, 0, &arpt_refresh, 0,
2057 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2058 
2059 	sysctl_createv(clog, 0, NULL, NULL,
2060 			CTLFLAG_PERMANENT,
2061 			CTLTYPE_STRUCT, "stats",
2062 			SYSCTL_DESCR("ARP statistics"),
2063 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2064 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2065 
2066 	sysctl_createv(clog, 0, NULL, NULL,
2067 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2068 			CTLTYPE_INT, "log_movements",
2069 			SYSCTL_DESCR("log ARP replies from MACs different than"
2070 			    " the one in the cache"),
2071 			NULL, 0, &log_movements, 0,
2072 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2073 
2074 	sysctl_createv(clog, 0, NULL, NULL,
2075 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2076 			CTLTYPE_INT, "log_permanent_modify",
2077 			SYSCTL_DESCR("log ARP replies from MACs different than"
2078 			    " the one in the permanent arp entry"),
2079 			NULL, 0, &log_permanent_modify, 0,
2080 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2081 
2082 	sysctl_createv(clog, 0, NULL, NULL,
2083 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2084 			CTLTYPE_INT, "log_wrong_iface",
2085 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2086 			    " interface"),
2087 			NULL, 0, &log_wrong_iface, 0,
2088 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2089 
2090 	sysctl_createv(clog, 0, NULL, NULL,
2091 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2092 			CTLTYPE_INT, "log_unknown_network",
2093 			SYSCTL_DESCR("log ARP packets from non-local network"),
2094 			NULL, 0, &log_unknown_network, 0,
2095 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2096 
2097 	sysctl_createv(clog, 0, NULL, NULL,
2098 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2099 		       CTLTYPE_INT, "debug",
2100 		       SYSCTL_DESCR("Enable ARP DAD debug output"),
2101 		       NULL, 0, &arp_debug, 0,
2102 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2103 }
2104 
2105 #endif /* INET */
2106