xref: /netbsd-src/sys/netinet/if_arp.c (revision cc576e1d8e4f4078fd4e81238abca9fca216f6ec)
1 /*	$NetBSD: if_arp.c,v 1.240 2017/01/24 07:09:24 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 1.240 2017/01/24 07:09:24 ozaki-r Exp $");
72 
73 #ifdef _KERNEL_OPT
74 #include "opt_ddb.h"
75 #include "opt_inet.h"
76 #include "opt_net_mpsafe.h"
77 #endif
78 
79 #ifdef INET
80 
81 #include "arp.h"
82 #include "bridge.h"
83 
84 #include <sys/param.h>
85 #include <sys/systm.h>
86 #include <sys/callout.h>
87 #include <sys/malloc.h>
88 #include <sys/mbuf.h>
89 #include <sys/socket.h>
90 #include <sys/time.h>
91 #include <sys/timetc.h>
92 #include <sys/kernel.h>
93 #include <sys/errno.h>
94 #include <sys/ioctl.h>
95 #include <sys/syslog.h>
96 #include <sys/proc.h>
97 #include <sys/protosw.h>
98 #include <sys/domain.h>
99 #include <sys/sysctl.h>
100 #include <sys/socketvar.h>
101 #include <sys/percpu.h>
102 #include <sys/cprng.h>
103 #include <sys/kmem.h>
104 
105 #include <net/ethertypes.h>
106 #include <net/if.h>
107 #include <net/if_dl.h>
108 #include <net/if_token.h>
109 #include <net/if_types.h>
110 #include <net/if_ether.h>
111 #include <net/if_llatbl.h>
112 #include <net/net_osdep.h>
113 #include <net/route.h>
114 #include <net/net_stats.h>
115 
116 #include <netinet/in.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip.h>
120 #include <netinet/if_inarp.h>
121 
122 #include "arcnet.h"
123 #if NARCNET > 0
124 #include <net/if_arc.h>
125 #endif
126 #include "fddi.h"
127 #if NFDDI > 0
128 #include <net/if_fddi.h>
129 #endif
130 #include "token.h"
131 #include "carp.h"
132 #if NCARP > 0
133 #include <netinet/ip_carp.h>
134 #endif
135 
136 #define SIN(s) ((struct sockaddr_in *)s)
137 #define SRP(s) ((struct sockaddr_inarp *)s)
138 
139 /*
140  * ARP trailer negotiation.  Trailer protocol is not IP specific,
141  * but ARP request/response use IP addresses.
142  */
143 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
144 
145 /* timer values */
146 static int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
147 static int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
148 static int	arp_maxhold = 1;	/* number of packets to hold per ARP entry */
149 #define	rt_expire rt_rmx.rmx_expire
150 #define	rt_pksent rt_rmx.rmx_pksent
151 
152 int		ip_dad_count = PROBE_NUM;
153 #ifdef ARP_DEBUG
154 int		arp_debug = 1;
155 #else
156 int		arp_debug = 0;
157 #endif
158 
159 static	void arp_init(void);
160 
161 static	void arprequest(struct ifnet *,
162     const struct in_addr *, const struct in_addr *,
163     const u_int8_t *);
164 static	void arpannounce1(struct ifaddr *);
165 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
166 	    const struct sockaddr *);
167 static	void arptimer(void *);
168 static	void arp_settimer(struct llentry *, int);
169 static	struct llentry *arplookup(struct ifnet *, struct mbuf *,
170 	    const struct in_addr *, const struct sockaddr *, int);
171 static	struct llentry *arpcreate(struct ifnet *, struct mbuf *,
172 	    const struct in_addr *, const struct sockaddr *, int);
173 static	void in_arpinput(struct mbuf *);
174 static	void in_revarpinput(struct mbuf *);
175 static	void revarprequest(struct ifnet *);
176 
177 static	void arp_drainstub(void);
178 
179 static void arp_dad_timer(struct ifaddr *);
180 static void arp_dad_start(struct ifaddr *);
181 static void arp_dad_stop(struct ifaddr *);
182 static void arp_dad_duplicated(struct ifaddr *, const char *);
183 
184 static void arp_init_llentry(struct ifnet *, struct llentry *);
185 #if NTOKEN > 0
186 static void arp_free_llentry_tokenring(struct llentry *);
187 #endif
188 
189 struct	ifqueue arpintrq = {
190 	.ifq_head = NULL,
191 	.ifq_tail = NULL,
192 	.ifq_len = 0,
193 	.ifq_maxlen = 50,
194 	.ifq_drops = 0,
195 };
196 static int	arp_maxtries = 5;
197 static int	useloopback = 1;	/* use loopback interface for local traffic */
198 
199 static percpu_t *arpstat_percpu;
200 
201 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
202 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
203 
204 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
205 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
206 
207 /* revarp state */
208 static struct	in_addr myip, srv_ip;
209 static int	myip_initialized = 0;
210 static int	revarp_in_progress = 0;
211 static struct	ifnet *myip_ifp = NULL;
212 
213 static int arp_drainwanted;
214 
215 static int log_movements = 1;
216 static int log_permanent_modify = 1;
217 static int log_wrong_iface = 1;
218 static int log_unknown_network = 1;
219 
220 /*
221  * this should be elsewhere.
222  */
223 
224 #define	LLA_ADDRSTRLEN	(16 * 3)
225 
226 static char *
227 lla_snprintf(char *, u_int8_t *, int);
228 
229 static char *
230 lla_snprintf(char *dst, u_int8_t *adrp, int len)
231 {
232 	int i;
233 	char *p;
234 
235 	p = dst;
236 
237 	*p++ = hexdigits[(*adrp) >> 4];
238 	*p++ = hexdigits[(*adrp++) & 0xf];
239 
240 	for (i = 1; i < len && i < 16; i++) {
241 		*p++ = ':';
242 		*p++ = hexdigits[(*adrp) >> 4];
243 		*p++ = hexdigits[(*adrp++) & 0xf];
244 	}
245 
246 	*p = 0;
247 	return dst;
248 }
249 
250 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
251 
252 static void
253 arp_fasttimo(void)
254 {
255 	if (arp_drainwanted) {
256 		arp_drain();
257 		arp_drainwanted = 0;
258 	}
259 }
260 
261 const struct protosw arpsw[] = {
262 	{ .pr_type = 0,
263 	  .pr_domain = &arpdomain,
264 	  .pr_protocol = 0,
265 	  .pr_flags = 0,
266 	  .pr_input = 0,
267 	  .pr_ctlinput = 0,
268 	  .pr_ctloutput = 0,
269 	  .pr_usrreqs = 0,
270 	  .pr_init = arp_init,
271 	  .pr_fasttimo = arp_fasttimo,
272 	  .pr_slowtimo = 0,
273 	  .pr_drain = arp_drainstub,
274 	}
275 };
276 
277 struct domain arpdomain = {
278 	.dom_family = PF_ARP,
279 	.dom_name = "arp",
280 	.dom_protosw = arpsw,
281 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
282 };
283 
284 static void sysctl_net_inet_arp_setup(struct sysctllog **);
285 
286 void
287 arp_init(void)
288 {
289 
290 	sysctl_net_inet_arp_setup(NULL);
291 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
292 	IFQ_LOCK_INIT(&arpintrq);
293 }
294 
295 static void
296 arp_drainstub(void)
297 {
298 	arp_drainwanted = 1;
299 }
300 
301 /*
302  * ARP protocol drain routine.  Called when memory is in short supply.
303  * Called at splvm();  don't acquire softnet_lock as can be called from
304  * hardware interrupt handlers.
305  */
306 void
307 arp_drain(void)
308 {
309 
310 	lltable_drain(AF_INET);
311 }
312 
313 static void
314 arptimer(void *arg)
315 {
316 	struct llentry *lle = arg;
317 	struct ifnet *ifp;
318 
319 	if (lle == NULL)
320 		return;
321 
322 	if (lle->la_flags & LLE_STATIC)
323 		return;
324 
325 	LLE_WLOCK(lle);
326 	if (callout_pending(&lle->la_timer)) {
327 		/*
328 		 * Here we are a bit odd here in the treatment of
329 		 * active/pending. If the pending bit is set, it got
330 		 * rescheduled before I ran. The active
331 		 * bit we ignore, since if it was stopped
332 		 * in ll_tablefree() and was currently running
333 		 * it would have return 0 so the code would
334 		 * not have deleted it since the callout could
335 		 * not be stopped so we want to go through
336 		 * with the delete here now. If the callout
337 		 * was restarted, the pending bit will be back on and
338 		 * we just want to bail since the callout_reset would
339 		 * return 1 and our reference would have been removed
340 		 * by arpresolve() below.
341 		 */
342 		LLE_WUNLOCK(lle);
343 		return;
344 	}
345 	ifp = lle->lle_tbl->llt_ifp;
346 
347 	callout_stop(&lle->la_timer);
348 
349 	/* XXX: LOR avoidance. We still have ref on lle. */
350 	LLE_WUNLOCK(lle);
351 
352 	IF_AFDATA_LOCK(ifp);
353 	LLE_WLOCK(lle);
354 
355 	/* Guard against race with other llentry_free(). */
356 	if (lle->la_flags & LLE_LINKED) {
357 		size_t pkts_dropped;
358 
359 		LLE_REMREF(lle);
360 		pkts_dropped = llentry_free(lle);
361 		ARP_STATADD(ARP_STAT_DFRDROPPED, pkts_dropped);
362 		ARP_STATADD(ARP_STAT_DFRTOTAL, pkts_dropped);
363 	} else {
364 		LLE_FREE_LOCKED(lle);
365 	}
366 
367 	IF_AFDATA_UNLOCK(ifp);
368 }
369 
370 static void
371 arp_settimer(struct llentry *la, int sec)
372 {
373 
374 	LLE_WLOCK_ASSERT(la);
375 	LLE_ADDREF(la);
376 	callout_reset(&la->la_timer, hz * sec, arptimer, la);
377 }
378 
379 /*
380  * We set the gateway for RTF_CLONING routes to a "prototype"
381  * link-layer sockaddr whose interface type (if_type) and interface
382  * index (if_index) fields are prepared.
383  */
384 static struct sockaddr *
385 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
386     const struct sockaddr *netmask)
387 {
388 	const struct ifnet *ifp = rt->rt_ifp;
389 	uint8_t namelen = strlen(ifp->if_xname);
390 	uint8_t addrlen = ifp->if_addrlen;
391 
392 	/*
393 	 * XXX: If this is a manually added route to interface
394 	 * such as older version of routed or gated might provide,
395 	 * restore cloning bit.
396 	 */
397 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
398 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
399 		rt->rt_flags |= RTF_CONNECTED;
400 
401 	if ((rt->rt_flags & (RTF_CONNECTED | RTF_LOCAL))) {
402 		union {
403 			struct sockaddr sa;
404 			struct sockaddr_storage ss;
405 			struct sockaddr_dl sdl;
406 		} u;
407 		/*
408 		 * Case 1: This route should come from a route to iface.
409 		 */
410 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
411 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
412 		rt_setgate(rt, &u.sa);
413 		gate = rt->rt_gateway;
414 	}
415 	return gate;
416 }
417 
418 static void
419 arp_init_llentry(struct ifnet *ifp, struct llentry *lle)
420 {
421 
422 	switch (ifp->if_type) {
423 #if NTOKEN > 0
424 	case IFT_ISO88025:
425 		lle->la_opaque = kmem_intr_alloc(sizeof(struct token_rif),
426 		    KM_NOSLEEP);
427 		lle->lle_ll_free = arp_free_llentry_tokenring;
428 		break;
429 #endif
430 	}
431 }
432 
433 #if NTOKEN > 0
434 static void
435 arp_free_llentry_tokenring(struct llentry *lle)
436 {
437 
438 	kmem_intr_free(lle->la_opaque, sizeof(struct token_rif));
439 }
440 #endif
441 
442 /*
443  * Parallel to llc_rtrequest.
444  */
445 void
446 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
447 {
448 	struct sockaddr *gate = rt->rt_gateway;
449 	struct in_ifaddr *ia;
450 	struct ifaddr *ifa;
451 	struct ifnet *ifp = rt->rt_ifp;
452 	int bound;
453 	int s;
454 
455 	if (req == RTM_LLINFO_UPD) {
456 		if ((ifa = info->rti_ifa) != NULL)
457 			arpannounce1(ifa);
458 		return;
459 	}
460 
461 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
462 		if (req != RTM_ADD)
463 			return;
464 
465 		/*
466 		 * linklayers with particular link MTU limitation.
467 		 */
468 		switch(ifp->if_type) {
469 #if NFDDI > 0
470 		case IFT_FDDI:
471 			if (ifp->if_mtu > FDDIIPMTU)
472 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
473 			break;
474 #endif
475 #if NARCNET > 0
476 		case IFT_ARCNET:
477 		    {
478 			int arcipifmtu;
479 
480 			if (ifp->if_flags & IFF_LINK0)
481 				arcipifmtu = arc_ipmtu;
482 			else
483 				arcipifmtu = ARCMTU;
484 			if (ifp->if_mtu > arcipifmtu)
485 				rt->rt_rmx.rmx_mtu = arcipifmtu;
486 			break;
487 		    }
488 #endif
489 		}
490 		return;
491 	}
492 
493 	switch (req) {
494 	case RTM_SETGATE:
495 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
496 		break;
497 	case RTM_ADD:
498 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
499 		if (gate == NULL) {
500 			log(LOG_ERR, "%s: arp_setgate failed\n", __func__);
501 			break;
502 		}
503 		if ((rt->rt_flags & RTF_CONNECTED) ||
504 		    (rt->rt_flags & RTF_LOCAL)) {
505 			/*
506 			 * Give this route an expiration time, even though
507 			 * it's a "permanent" route, so that routes cloned
508 			 * from it do not need their expiration time set.
509 			 */
510 			KASSERT(time_uptime != 0);
511 			rt->rt_expire = time_uptime;
512 			/*
513 			 * linklayers with particular link MTU limitation.
514 			 */
515 			switch (ifp->if_type) {
516 #if NFDDI > 0
517 			case IFT_FDDI:
518 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
519 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
520 				     (rt->rt_rmx.rmx_mtu == 0 &&
521 				      ifp->if_mtu > FDDIIPMTU)))
522 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
523 				break;
524 #endif
525 #if NARCNET > 0
526 			case IFT_ARCNET:
527 			    {
528 				int arcipifmtu;
529 				if (ifp->if_flags & IFF_LINK0)
530 					arcipifmtu = arc_ipmtu;
531 				else
532 					arcipifmtu = ARCMTU;
533 
534 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
535 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
536 				     (rt->rt_rmx.rmx_mtu == 0 &&
537 				      ifp->if_mtu > arcipifmtu)))
538 					rt->rt_rmx.rmx_mtu = arcipifmtu;
539 				break;
540 			    }
541 #endif
542 			}
543 			if (rt->rt_flags & RTF_CONNECTED)
544 				break;
545 		}
546 
547 		bound = curlwp_bind();
548 		/* Announce a new entry if requested. */
549 		if (rt->rt_flags & RTF_ANNOUNCE) {
550 			struct psref psref;
551 			ia = in_get_ia_on_iface_psref(
552 			    satocsin(rt_getkey(rt))->sin_addr, ifp, &psref);
553 			if (ia != NULL) {
554 				arpannounce(ifp, &ia->ia_ifa,
555 				    CLLADDR(satocsdl(gate)));
556 				ia4_release(ia, &psref);
557 			}
558 		}
559 
560 		if (gate->sa_family != AF_LINK ||
561 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
562 			log(LOG_DEBUG, "%s: bad gateway value\n", __func__);
563 			goto out;
564 		}
565 
566 		satosdl(gate)->sdl_type = ifp->if_type;
567 		satosdl(gate)->sdl_index = ifp->if_index;
568 
569 		/* If the route is for a broadcast address mark it as such.
570 		 * This way we can avoid an expensive call to in_broadcast()
571 		 * in ip_output() most of the time (because the route passed
572 		 * to ip_output() is almost always a host route). */
573 		if (rt->rt_flags & RTF_HOST &&
574 		    !(rt->rt_flags & RTF_BROADCAST) &&
575 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
576 			rt->rt_flags |= RTF_BROADCAST;
577 		/* There is little point in resolving the broadcast address */
578 		if (rt->rt_flags & RTF_BROADCAST)
579 			goto out;
580 
581 		/*
582 		 * When called from rt_ifa_addlocal, we cannot depend on that
583 		 * the address (rt_getkey(rt)) exits in the address list of the
584 		 * interface. So check RTF_LOCAL instead.
585 		 */
586 		if (rt->rt_flags & RTF_LOCAL) {
587 			rt->rt_expire = 0;
588 			if (useloopback) {
589 				rt->rt_ifp = lo0ifp;
590 				rt->rt_rmx.rmx_mtu = 0;
591 			}
592 			goto out;
593 		}
594 
595 		s = pserialize_read_enter();
596 		ia = in_get_ia_on_iface(satocsin(rt_getkey(rt))->sin_addr, ifp);
597 		if (ia == NULL) {
598 			pserialize_read_exit(s);
599 			goto out;
600 		}
601 
602 		rt->rt_expire = 0;
603 		if (useloopback) {
604 			rt->rt_ifp = lo0ifp;
605 			rt->rt_rmx.rmx_mtu = 0;
606 		}
607 		rt->rt_flags |= RTF_LOCAL;
608 		/*
609 		 * make sure to set rt->rt_ifa to the interface
610 		 * address we are using, otherwise we will have trouble
611 		 * with source address selection.
612 		 */
613 		ifa = &ia->ia_ifa;
614 		if (ifa != rt->rt_ifa)
615 			/* Assume it doesn't sleep */
616 			rt_replace_ifa(rt, ifa);
617 		pserialize_read_exit(s);
618 	out:
619 		curlwp_bindx(bound);
620 		break;
621 	}
622 }
623 
624 /*
625  * Broadcast an ARP request. Caller specifies:
626  *	- arp header source ip address
627  *	- arp header target ip address
628  *	- arp header source ethernet address
629  */
630 static void
631 arprequest(struct ifnet *ifp,
632     const struct in_addr *sip, const struct in_addr *tip,
633     const u_int8_t *enaddr)
634 {
635 	struct mbuf *m;
636 	struct arphdr *ah;
637 	struct sockaddr sa;
638 	uint64_t *arps;
639 
640 	KASSERT(sip != NULL);
641 	KASSERT(tip != NULL);
642 	KASSERT(enaddr != NULL);
643 
644 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
645 		return;
646 	MCLAIM(m, &arpdomain.dom_mowner);
647 	switch (ifp->if_type) {
648 	case IFT_IEEE1394:
649 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
650 		    ifp->if_addrlen;
651 		break;
652 	default:
653 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
654 		    2 * ifp->if_addrlen;
655 		break;
656 	}
657 	m->m_pkthdr.len = m->m_len;
658 	MH_ALIGN(m, m->m_len);
659 	ah = mtod(m, struct arphdr *);
660 	memset(ah, 0, m->m_len);
661 	switch (ifp->if_type) {
662 	case IFT_IEEE1394:	/* RFC2734 */
663 		/* fill it now for ar_tpa computation */
664 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
665 		break;
666 	default:
667 		/* ifp->if_output will fill ar_hrd */
668 		break;
669 	}
670 	ah->ar_pro = htons(ETHERTYPE_IP);
671 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
672 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
673 	ah->ar_op = htons(ARPOP_REQUEST);
674 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
675 	memcpy(ar_spa(ah), sip, ah->ar_pln);
676 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
677 	sa.sa_family = AF_ARP;
678 	sa.sa_len = 2;
679 	m->m_flags |= M_BCAST;
680 	arps = ARP_STAT_GETREF();
681 	arps[ARP_STAT_SNDTOTAL]++;
682 	arps[ARP_STAT_SENDREQUEST]++;
683 	ARP_STAT_PUTREF();
684 	if_output_lock(ifp, ifp, m, &sa, NULL);
685 }
686 
687 void
688 arpannounce(struct ifnet *ifp, struct ifaddr *ifa, const uint8_t *enaddr)
689 {
690 	struct in_ifaddr *ia = ifatoia(ifa);
691 	struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
692 
693 	if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) {
694 		ARPLOG(LOG_DEBUG, "%s not ready\n", ARPLOGADDR(*ip));
695 		return;
696 	}
697 	arprequest(ifp, ip, ip, enaddr);
698 }
699 
700 static void
701 arpannounce1(struct ifaddr *ifa)
702 {
703 
704 	arpannounce(ifa->ifa_ifp, ifa, CLLADDR(ifa->ifa_ifp->if_sadl));
705 }
706 
707 /*
708  * Resolve an IP address into an ethernet address.  If success,
709  * desten is filled in.  If there is no entry in arptab,
710  * set one up and broadcast a request for the IP address.
711  * Hold onto this mbuf and resend it once the address
712  * is finally resolved.  A return value of 0 indicates
713  * that desten has been filled in and the packet should be sent
714  * normally; a return value of EWOULDBLOCK indicates that the packet has been
715  * held pending resolution.
716  * Any other value indicates an error.
717  */
718 int
719 arpresolve(struct ifnet *ifp, const struct rtentry *rt, struct mbuf *m,
720     const struct sockaddr *dst, void *desten, size_t destlen)
721 {
722 	struct llentry *la;
723 	const char *create_lookup;
724 	bool renew;
725 	int error;
726 
727 	KASSERT(m != NULL);
728 
729 	la = arplookup(ifp, m, NULL, dst, 0);
730 	if (la == NULL)
731 		goto notfound;
732 
733 	if ((la->la_flags & LLE_VALID) &&
734 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
735 		KASSERT(destlen >= ifp->if_addrlen);
736 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
737 		LLE_RUNLOCK(la);
738 		return 0;
739 	}
740 
741 notfound:
742 #ifdef IFF_STATICARP /* FreeBSD */
743 #define _IFF_NOARP (IFF_NOARP | IFF_STATICARP)
744 #else
745 #define _IFF_NOARP IFF_NOARP
746 #endif
747 	if (ifp->if_flags & _IFF_NOARP) {
748 		if (la != NULL)
749 			LLE_RUNLOCK(la);
750 		error = ENOTSUP;
751 		goto bad;
752 	}
753 #undef _IFF_NOARP
754 	if (la == NULL) {
755 		create_lookup = "create";
756 		IF_AFDATA_WLOCK(ifp);
757 		la = lla_create(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
758 		IF_AFDATA_WUNLOCK(ifp);
759 		if (la == NULL)
760 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
761 		else
762 			arp_init_llentry(ifp, la);
763 	} else if (LLE_TRY_UPGRADE(la) == 0) {
764 		create_lookup = "lookup";
765 		LLE_RUNLOCK(la);
766 		IF_AFDATA_RLOCK(ifp);
767 		la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
768 		IF_AFDATA_RUNLOCK(ifp);
769 	}
770 
771 	error = EINVAL;
772 	if (la == NULL) {
773 		log(LOG_DEBUG,
774 		    "%s: failed to %s llentry for %s on %s\n",
775 		    __func__, create_lookup, inet_ntoa(satocsin(dst)->sin_addr),
776 		    ifp->if_xname);
777 		goto bad;
778 	}
779 
780 	if ((la->la_flags & LLE_VALID) &&
781 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime))
782 	{
783 		KASSERT(destlen >= ifp->if_addrlen);
784 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
785 		renew = false;
786 		/*
787 		 * If entry has an expiry time and it is approaching,
788 		 * see if we need to send an ARP request within this
789 		 * arpt_down interval.
790 		 */
791 		if (!(la->la_flags & LLE_STATIC) &&
792 		    time_uptime + la->la_preempt > la->la_expire)
793 		{
794 			renew = true;
795 			la->la_preempt--;
796 		}
797 
798 		LLE_WUNLOCK(la);
799 
800 		if (renew) {
801 			const u_int8_t *enaddr =
802 #if NCARP > 0
803 			    (ifp->if_type == IFT_CARP) ?
804 			    CLLADDR(ifp->if_sadl):
805 #endif
806 			    CLLADDR(ifp->if_sadl);
807 			arprequest(ifp,
808 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
809 			    &satocsin(dst)->sin_addr, enaddr);
810 		}
811 
812 		return 0;
813 	}
814 
815 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
816 		LLE_RUNLOCK(la);
817 		log(LOG_DEBUG, "%s: ouch, empty static llinfo for %s\n",
818 		    __func__, inet_ntoa(satocsin(dst)->sin_addr));
819 		error = EINVAL;
820 		goto bad;
821 	}
822 
823 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
824 
825 	/*
826 	 * There is an arptab entry, but no ethernet address
827 	 * response yet.  Add the mbuf to the list, dropping
828 	 * the oldest packet if we have exceeded the system
829 	 * setting.
830 	 */
831 	LLE_WLOCK_ASSERT(la);
832 	if (la->la_numheld >= arp_maxhold) {
833 		if (la->la_hold != NULL) {
834 			struct mbuf *next = la->la_hold->m_nextpkt;
835 			m_freem(la->la_hold);
836 			la->la_hold = next;
837 			la->la_numheld--;
838 			ARP_STATINC(ARP_STAT_DFRDROPPED);
839 			ARP_STATINC(ARP_STAT_DFRTOTAL);
840 		}
841 	}
842 	if (la->la_hold != NULL) {
843 		struct mbuf *curr = la->la_hold;
844 		while (curr->m_nextpkt != NULL)
845 			curr = curr->m_nextpkt;
846 		curr->m_nextpkt = m;
847 	} else
848 		la->la_hold = m;
849 	la->la_numheld++;
850 	if (!renew)
851 		LLE_DOWNGRADE(la);
852 
853 	/*
854 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
855 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
856 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
857 	 * ARP request, but not faster than one request per second.
858 	 */
859 	if (la->la_asked < arp_maxtries)
860 		error = EWOULDBLOCK;	/* First request. */
861 	else
862 		error = (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
863 		    EHOSTUNREACH : EHOSTDOWN;
864 
865 	if (renew) {
866 		const u_int8_t *enaddr =
867 #if NCARP > 0
868 		    (rt != NULL && rt->rt_ifp->if_type == IFT_CARP) ?
869 		    CLLADDR(rt->rt_ifp->if_sadl):
870 #endif
871 		    CLLADDR(ifp->if_sadl);
872 		la->la_expire = time_uptime;
873 		arp_settimer(la, arpt_down);
874 		la->la_asked++;
875 		LLE_WUNLOCK(la);
876 
877 		if (rt != NULL) {
878 			arprequest(ifp, &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
879 			    &satocsin(dst)->sin_addr, enaddr);
880 		} else {
881 			struct sockaddr_in sin;
882 			struct rtentry *_rt;
883 
884 			sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
885 
886 			/* XXX */
887 			_rt = rtalloc1((struct sockaddr *)&sin, 0);
888 			if (_rt == NULL)
889 				goto bad;
890 			arprequest(ifp,
891 			    &satocsin(_rt->rt_ifa->ifa_addr)->sin_addr,
892 			    &satocsin(dst)->sin_addr, enaddr);
893 			rt_unref(_rt);
894 		}
895 		return error;
896 	}
897 
898 	LLE_RUNLOCK(la);
899 	return error;
900 
901 bad:
902 	m_freem(m);
903 	return error;
904 }
905 
906 /*
907  * Common length and type checks are done here,
908  * then the protocol-specific routine is called.
909  */
910 void
911 arpintr(void)
912 {
913 	struct mbuf *m;
914 	struct arphdr *ar;
915 	int s;
916 	int arplen;
917 
918 #ifndef NET_MPSAFE
919 	mutex_enter(softnet_lock);
920 	KERNEL_LOCK(1, NULL);
921 #endif
922 	for (;;) {
923 		struct ifnet *rcvif;
924 
925 		IFQ_LOCK(&arpintrq);
926 		IF_DEQUEUE(&arpintrq, m);
927 		IFQ_UNLOCK(&arpintrq);
928 		if (m == NULL)
929 			goto out;
930 		if ((m->m_flags & M_PKTHDR) == 0)
931 			panic("arpintr");
932 
933 		MCLAIM(m, &arpdomain.dom_mowner);
934 		ARP_STATINC(ARP_STAT_RCVTOTAL);
935 
936 		/*
937 		 * First, make sure we have at least struct arphdr.
938 		 */
939 		if (m->m_len < sizeof(struct arphdr) ||
940 		    (ar = mtod(m, struct arphdr *)) == NULL)
941 			goto badlen;
942 
943 		rcvif = m_get_rcvif(m, &s);
944 		switch (rcvif->if_type) {
945 		case IFT_IEEE1394:
946 			arplen = sizeof(struct arphdr) +
947 			    ar->ar_hln + 2 * ar->ar_pln;
948 			break;
949 		default:
950 			arplen = sizeof(struct arphdr) +
951 			    2 * ar->ar_hln + 2 * ar->ar_pln;
952 			break;
953 		}
954 		m_put_rcvif(rcvif, &s);
955 
956 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
957 		    m->m_len >= arplen)
958 			switch (ntohs(ar->ar_pro)) {
959 			case ETHERTYPE_IP:
960 			case ETHERTYPE_IPTRAILERS:
961 				in_arpinput(m);
962 				continue;
963 			default:
964 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
965 			}
966 		else {
967 badlen:
968 			ARP_STATINC(ARP_STAT_RCVBADLEN);
969 		}
970 		m_freem(m);
971 	}
972 out:
973 #ifndef NET_MPSAFE
974 	KERNEL_UNLOCK_ONE(NULL);
975 	mutex_exit(softnet_lock);
976 #else
977 	return; /* XXX gcc */
978 #endif
979 }
980 
981 /*
982  * ARP for Internet protocols on 10 Mb/s Ethernet. Algorithm is that given in
983  * RFC 826. In addition, a sanity check is performed on the sender protocol
984  * address, to catch impersonators.
985  *
986  * We no longer handle negotiations for use of trailer protocol: formerly, ARP
987  * replied for protocol type ETHERTYPE_TRAIL sent along with IP replies if we
988  * wanted trailers sent to us, and also sent them in response to IP replies.
989  * This allowed either end to announce the desire to receive trailer packets.
990  *
991  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, but
992  * formerly didn't normally send requests.
993  */
994 static void
995 in_arpinput(struct mbuf *m)
996 {
997 	struct arphdr *ah;
998 	struct ifnet *ifp, *rcvif = NULL;
999 	struct llentry *la = NULL;
1000 	struct in_ifaddr *ia = NULL;
1001 #if NBRIDGE > 0
1002 	struct in_ifaddr *bridge_ia = NULL;
1003 #endif
1004 #if NCARP > 0
1005 	u_int32_t count = 0, index = 0;
1006 #endif
1007 	struct sockaddr sa;
1008 	struct in_addr isaddr, itaddr, myaddr;
1009 	int op;
1010 	void *tha;
1011 	uint64_t *arps;
1012 	struct psref psref, psref_ia;
1013 	int s;
1014 	char llabuf[LLA_ADDRSTRLEN];
1015 	char ipbuf[INET_ADDRSTRLEN];
1016 
1017 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
1018 		goto out;
1019 	ah = mtod(m, struct arphdr *);
1020 	op = ntohs(ah->ar_op);
1021 
1022 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
1023 	if (__predict_false(rcvif == NULL))
1024 		goto drop;
1025 
1026 	/*
1027 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or ar_tpa().
1028 	 * XXX check ar_hrd more strictly?
1029 	 */
1030 	switch (ifp->if_type) {
1031 	case IFT_IEEE1394:
1032 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
1033 			;
1034 		else {
1035 			/* XXX this is to make sure we compute ar_tha right */
1036 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
1037 		}
1038 		break;
1039 	default:
1040 		break;
1041 	}
1042 
1043 	if (ah->ar_pln != sizeof(struct in_addr))
1044 		goto drop;
1045 
1046 	memcpy(&isaddr, ar_spa(ah), sizeof(isaddr));
1047 	memcpy(&itaddr, ar_tpa(ah), sizeof(itaddr));
1048 
1049 	if (m->m_flags & (M_BCAST|M_MCAST))
1050 		ARP_STATINC(ARP_STAT_RCVMCAST);
1051 
1052 	/*
1053 	 * Search for a matching interface address
1054 	 * or any address on the interface to use
1055 	 * as a dummy address in the rest of this function
1056 	 */
1057 	s = pserialize_read_enter();
1058 	IN_ADDRHASH_READER_FOREACH(ia, itaddr.s_addr) {
1059 		if (!in_hosteq(ia->ia_addr.sin_addr, itaddr))
1060 			continue;
1061 #if NCARP > 0
1062 		if (ia->ia_ifp->if_type == IFT_CARP &&
1063 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1064 		    (IFF_UP|IFF_RUNNING))) {
1065 			index++;
1066 			/* XXX: ar_hln? */
1067 			if (ia->ia_ifp == rcvif && (ah->ar_hln >= 6) &&
1068 			    carp_iamatch(ia, ar_sha(ah),
1069 			    &count, index)) {
1070 				break;
1071 			}
1072 		} else
1073 #endif
1074 		if (ia->ia_ifp == rcvif)
1075 			break;
1076 #if NBRIDGE > 0
1077 		/*
1078 		 * If the interface we received the packet on
1079 		 * is part of a bridge, check to see if we need
1080 		 * to "bridge" the packet to ourselves at this
1081 		 * layer.  Note we still prefer a perfect match,
1082 		 * but allow this weaker match if necessary.
1083 		 */
1084 		if (rcvif->if_bridge != NULL &&
1085 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
1086 			bridge_ia = ia;
1087 #endif /* NBRIDGE > 0 */
1088 	}
1089 
1090 #if NBRIDGE > 0
1091 	if (ia == NULL && bridge_ia != NULL) {
1092 		ia = bridge_ia;
1093 		m_put_rcvif_psref(rcvif, &psref);
1094 		rcvif = NULL;
1095 		/* FIXME */
1096 		ifp = bridge_ia->ia_ifp;
1097 	}
1098 #endif
1099 	if (ia != NULL)
1100 		ia4_acquire(ia, &psref_ia);
1101 	pserialize_read_exit(s);
1102 
1103 	if (ah->ar_hln != ifp->if_addrlen) {
1104 		ARP_STATINC(ARP_STAT_RCVBADLEN);
1105 		log(LOG_WARNING,
1106 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
1107 		    in_fmtaddr(ipbuf, isaddr), ah->ar_hln, ifp->if_addrlen);
1108 		goto out;
1109 	}
1110 
1111 	if (ia == NULL) {
1112 		ia = in_get_ia_on_iface_psref(isaddr, rcvif, &psref_ia);
1113 		if (ia == NULL) {
1114 			ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
1115 			if (ia == NULL) {
1116 				ARP_STATINC(ARP_STAT_RCVNOINT);
1117 				goto out;
1118 			}
1119 		}
1120 	}
1121 
1122 	myaddr = ia->ia_addr.sin_addr;
1123 
1124 	/* XXX checks for bridge case? */
1125 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
1126 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1127 		goto out;	/* it's from me, ignore it. */
1128 	}
1129 
1130 	/* XXX checks for bridge case? */
1131 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1132 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1133 		log(LOG_ERR,
1134 		    "%s: arp: link address is broadcast for IP address %s!\n",
1135 		    ifp->if_xname, in_fmtaddr(ipbuf, isaddr));
1136 		goto out;
1137 	}
1138 
1139 	/*
1140 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
1141 	 */
1142 	if (in_nullhost(isaddr))
1143 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
1144 	else if (in_hosteq(isaddr, myaddr))
1145 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1146 
1147 	/*
1148 	 * If the target IP address is zero, ignore the packet.
1149 	 * This prevents the code below from trying to answer
1150 	 * when we are using IP address zero (booting).
1151 	 */
1152 	if (in_nullhost(itaddr)) {
1153 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
1154 		goto out;
1155 	}
1156 
1157 	/* DAD check, RFC 5227 */
1158 	if (in_hosteq(isaddr, myaddr) ||
1159 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)))
1160 	{
1161 		arp_dad_duplicated((struct ifaddr *)ia,
1162 		    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln));
1163 		goto out;
1164 	}
1165 
1166 	if (in_nullhost(isaddr))
1167 		goto reply;
1168 
1169 	if (in_hosteq(itaddr, myaddr))
1170 		la = arpcreate(ifp, m, &isaddr, NULL, 1);
1171 	else
1172 		la = arplookup(ifp, m, &isaddr, NULL, 1);
1173 	if (la == NULL)
1174 		goto reply;
1175 
1176 	if ((la->la_flags & LLE_VALID) &&
1177 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
1178 		if (la->la_flags & LLE_STATIC) {
1179 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
1180 			if (!log_permanent_modify)
1181 				goto out;
1182 			log(LOG_INFO,
1183 			    "%s tried to overwrite permanent arp info"
1184 			    " for %s\n",
1185 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1186 			    in_fmtaddr(ipbuf, isaddr));
1187 			goto out;
1188 		} else if (la->lle_tbl->llt_ifp != ifp) {
1189 			/* XXX should not happen? */
1190 			ARP_STATINC(ARP_STAT_RCVOVERINT);
1191 			if (!log_wrong_iface)
1192 				goto out;
1193 			log(LOG_INFO,
1194 			    "%s on %s tried to overwrite "
1195 			    "arp info for %s on %s\n",
1196 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1197 			    ifp->if_xname, in_fmtaddr(ipbuf, isaddr),
1198 			    la->lle_tbl->llt_ifp->if_xname);
1199 				goto out;
1200 		} else {
1201 			ARP_STATINC(ARP_STAT_RCVOVER);
1202 			if (log_movements)
1203 				log(LOG_INFO, "arp info overwritten "
1204 				    "for %s by %s\n",
1205 				    in_fmtaddr(ipbuf, isaddr),
1206 				    lla_snprintf(llabuf, ar_sha(ah),
1207 				    ah->ar_hln));
1208 		}
1209 	}
1210 
1211 	/* XXX llentry should have addrlen? */
1212 #if 0
1213 	/*
1214 	 * sanity check for the address length.
1215 	 * XXX this does not work for protocols with variable address
1216 	 * length. -is
1217 	 */
1218 	if (sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
1219 		ARP_STATINC(ARP_STAT_RCVLENCHG);
1220 		log(LOG_WARNING,
1221 		    "arp from %s: new addr len %d, was %d\n",
1222 		    in_fmtaddr(ipbuf, isaddr), ah->ar_hln, sdl->sdl_alen);
1223 	}
1224 #endif
1225 
1226 #if NTOKEN > 0
1227 	/*
1228 	 * XXX uses m_data and assumes the complete answer including
1229 	 * XXX token-ring headers is in the same buf
1230 	 */
1231 	if (ifp->if_type == IFT_ISO88025) {
1232 		struct token_header *trh;
1233 
1234 		trh = (struct token_header *)M_TRHSTART(m);
1235 		if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1236 			struct token_rif *rif;
1237 			size_t riflen;
1238 
1239 			rif = TOKEN_RIF(trh);
1240 			riflen = (ntohs(rif->tr_rcf) &
1241 			    TOKEN_RCF_LEN_MASK) >> 8;
1242 
1243 			if (riflen > 2 &&
1244 			    riflen < sizeof(struct token_rif) &&
1245 			    (riflen & 1) == 0) {
1246 				rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1247 				rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1248 				memcpy(TOKEN_RIF_LLE(la), rif, riflen);
1249 			}
1250 		}
1251 	}
1252 #endif /* NTOKEN > 0 */
1253 
1254 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1255 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1256 	la->la_flags |= LLE_VALID;
1257 	if ((la->la_flags & LLE_STATIC) == 0) {
1258 		la->la_expire = time_uptime + arpt_keep;
1259 		arp_settimer(la, arpt_keep);
1260 	}
1261 	la->la_asked = 0;
1262 	/* rt->rt_flags &= ~RTF_REJECT; */
1263 
1264 	if (la->la_hold != NULL) {
1265 		int n = la->la_numheld;
1266 		struct mbuf *m_hold, *m_hold_next;
1267 		struct sockaddr_in sin;
1268 
1269 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1270 
1271 		m_hold = la->la_hold;
1272 		la->la_hold = NULL;
1273 		la->la_numheld = 0;
1274 		/*
1275 		 * We have to unlock here because if_output would call
1276 		 * arpresolve
1277 		 */
1278 		LLE_WUNLOCK(la);
1279 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1280 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1281 		for (; m_hold != NULL; m_hold = m_hold_next) {
1282 			m_hold_next = m_hold->m_nextpkt;
1283 			m_hold->m_nextpkt = NULL;
1284 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1285 		}
1286 	} else
1287 		LLE_WUNLOCK(la);
1288 	la = NULL;
1289 
1290 reply:
1291 	if (la != NULL) {
1292 		LLE_WUNLOCK(la);
1293 		la = NULL;
1294 	}
1295 	if (op != ARPOP_REQUEST) {
1296 		if (op == ARPOP_REPLY)
1297 			ARP_STATINC(ARP_STAT_RCVREPLY);
1298 		goto out;
1299 	}
1300 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1301 	if (in_hosteq(itaddr, myaddr)) {
1302 		/* If our address is unusable, don't reply */
1303 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1304 			goto out;
1305 		/* I am the target */
1306 		tha = ar_tha(ah);
1307 		if (tha)
1308 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1309 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1310 	} else {
1311 		/* Proxy ARP */
1312 		struct llentry *lle = NULL;
1313 		struct sockaddr_in sin;
1314 #if NCARP > 0
1315 		struct ifnet *_rcvif = m_get_rcvif(m, &s);
1316 		if (ifp->if_type == IFT_CARP && _rcvif->if_type != IFT_CARP)
1317 			goto out;
1318 		m_put_rcvif(_rcvif, &s);
1319 #endif
1320 
1321 		tha = ar_tha(ah);
1322 
1323 		sockaddr_in_init(&sin, &itaddr, 0);
1324 
1325 		IF_AFDATA_RLOCK(ifp);
1326 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1327 		IF_AFDATA_RUNLOCK(ifp);
1328 
1329 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1330 			if (tha)
1331 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1332 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1333 			LLE_RUNLOCK(lle);
1334 		} else {
1335 			if (lle != NULL)
1336 				LLE_RUNLOCK(lle);
1337 			goto drop;
1338 		}
1339 	}
1340 	ia4_release(ia, &psref_ia);
1341 
1342 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1343 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1344 	ah->ar_op = htons(ARPOP_REPLY);
1345 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1346 	switch (ifp->if_type) {
1347 	case IFT_IEEE1394:
1348 		/*
1349 		 * ieee1394 arp reply is broadcast
1350 		 */
1351 		m->m_flags &= ~M_MCAST;
1352 		m->m_flags |= M_BCAST;
1353 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1354 		break;
1355 	default:
1356 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1357 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1358 		break;
1359 	}
1360 	m->m_pkthdr.len = m->m_len;
1361 	sa.sa_family = AF_ARP;
1362 	sa.sa_len = 2;
1363 	arps = ARP_STAT_GETREF();
1364 	arps[ARP_STAT_SNDTOTAL]++;
1365 	arps[ARP_STAT_SNDREPLY]++;
1366 	ARP_STAT_PUTREF();
1367 	if_output_lock(ifp, ifp, m, &sa, NULL);
1368 	if (rcvif != NULL)
1369 		m_put_rcvif_psref(rcvif, &psref);
1370 	return;
1371 
1372 out:
1373 	if (la != NULL)
1374 		LLE_WUNLOCK(la);
1375 drop:
1376 	if (ia != NULL)
1377 		ia4_release(ia, &psref_ia);
1378 	if (rcvif != NULL)
1379 		m_put_rcvif_psref(rcvif, &psref);
1380 	m_freem(m);
1381 }
1382 
1383 /*
1384  * Lookup or a new address in arptab.
1385  */
1386 static struct llentry *
1387 arplookup(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
1388     const struct sockaddr *sa, int wlock)
1389 {
1390 	struct sockaddr_in sin;
1391 	struct llentry *la;
1392 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1393 
1394 
1395 	if (sa == NULL) {
1396 		KASSERT(addr != NULL);
1397 		sockaddr_in_init(&sin, addr, 0);
1398 		sa = sintocsa(&sin);
1399 	}
1400 
1401 	IF_AFDATA_RLOCK(ifp);
1402 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1403 	IF_AFDATA_RUNLOCK(ifp);
1404 
1405 	return la;
1406 }
1407 
1408 static struct llentry *
1409 arpcreate(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
1410     const struct sockaddr *sa, int wlock)
1411 {
1412 	struct sockaddr_in sin;
1413 	struct llentry *la;
1414 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1415 
1416 	if (sa == NULL) {
1417 		KASSERT(addr != NULL);
1418 		sockaddr_in_init(&sin, addr, 0);
1419 		sa = sintocsa(&sin);
1420 	}
1421 
1422 	la = arplookup(ifp, m, addr, sa, wlock);
1423 
1424 	if (la == NULL) {
1425 		IF_AFDATA_WLOCK(ifp);
1426 		la = lla_create(LLTABLE(ifp), flags, sa);
1427 		IF_AFDATA_WUNLOCK(ifp);
1428 
1429 		if (la != NULL)
1430 			arp_init_llentry(ifp, la);
1431 	}
1432 
1433 	return la;
1434 }
1435 
1436 int
1437 arpioctl(u_long cmd, void *data)
1438 {
1439 
1440 	return EOPNOTSUPP;
1441 }
1442 
1443 void
1444 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1445 {
1446 	struct in_addr *ip;
1447 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1448 
1449 	/*
1450 	 * Warn the user if another station has this IP address,
1451 	 * but only if the interface IP address is not zero.
1452 	 */
1453 	ip = &IA_SIN(ifa)->sin_addr;
1454 	if (!in_nullhost(*ip) &&
1455 	    (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) == 0) {
1456 		struct llentry *lle;
1457 
1458 		/*
1459 		 * interface address is considered static entry
1460 		 * because the output of the arp utility shows
1461 		 * that L2 entry as permanent
1462 		 */
1463 		IF_AFDATA_WLOCK(ifp);
1464 		lle = lla_create(LLTABLE(ifp), (LLE_IFADDR | LLE_STATIC),
1465 				 (struct sockaddr *)IA_SIN(ifa));
1466 		IF_AFDATA_WUNLOCK(ifp);
1467 		if (lle == NULL)
1468 			log(LOG_INFO, "%s: cannot create arp entry for"
1469 			    " interface address\n", __func__);
1470 		else {
1471 			arp_init_llentry(ifp, lle);
1472 			LLE_RUNLOCK(lle);
1473 		}
1474 	}
1475 
1476 	ifa->ifa_rtrequest = arp_rtrequest;
1477 	ifa->ifa_flags |= RTF_CONNECTED;
1478 
1479 	/* ARP will handle DAD for this address. */
1480 	if (in_nullhost(*ip)) {
1481 		if (ia->ia_dad_stop != NULL)	/* safety */
1482 			ia->ia_dad_stop(ifa);
1483 		ia->ia_dad_start = NULL;
1484 		ia->ia_dad_stop = NULL;
1485 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1486 	} else {
1487 		ia->ia_dad_start = arp_dad_start;
1488 		ia->ia_dad_stop = arp_dad_stop;
1489 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE)
1490 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1491 		else
1492 			arpannounce1(ifa);
1493 	}
1494 }
1495 
1496 TAILQ_HEAD(dadq_head, dadq);
1497 struct dadq {
1498 	TAILQ_ENTRY(dadq) dad_list;
1499 	struct ifaddr *dad_ifa;
1500 	int dad_count;		/* max ARP to send */
1501 	int dad_arp_tcount;	/* # of trials to send ARP */
1502 	int dad_arp_ocount;	/* ARP sent so far */
1503 	int dad_arp_announce;	/* max ARP announcements */
1504 	int dad_arp_acount;	/* # of announcements */
1505 	struct callout dad_timer_ch;
1506 };
1507 MALLOC_JUSTDEFINE(M_IPARP, "ARP DAD", "ARP DAD Structure");
1508 
1509 static struct dadq_head dadq;
1510 static int dad_init = 0;
1511 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1512 static kmutex_t arp_dad_lock;
1513 
1514 static struct dadq *
1515 arp_dad_find(struct ifaddr *ifa)
1516 {
1517 	struct dadq *dp;
1518 
1519 	KASSERT(mutex_owned(&arp_dad_lock));
1520 
1521 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1522 		if (dp->dad_ifa == ifa)
1523 			return dp;
1524 	}
1525 	return NULL;
1526 }
1527 
1528 static void
1529 arp_dad_starttimer(struct dadq *dp, int ticks)
1530 {
1531 
1532 	callout_reset(&dp->dad_timer_ch, ticks,
1533 	    (void (*)(void *))arp_dad_timer, (void *)dp->dad_ifa);
1534 }
1535 
1536 static void
1537 arp_dad_stoptimer(struct dadq *dp)
1538 {
1539 
1540 #ifdef NET_MPSAFE
1541 	callout_halt(&dp->dad_timer_ch, NULL);
1542 #else
1543 	callout_halt(&dp->dad_timer_ch, softnet_lock);
1544 #endif
1545 }
1546 
1547 static void
1548 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1549 {
1550 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1551 	struct ifnet *ifp = ifa->ifa_ifp;
1552 	struct in_addr sip;
1553 
1554 	dp->dad_arp_tcount++;
1555 	if ((ifp->if_flags & IFF_UP) == 0)
1556 		return;
1557 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1558 		return;
1559 
1560 	dp->dad_arp_tcount = 0;
1561 	dp->dad_arp_ocount++;
1562 
1563 	memset(&sip, 0, sizeof(sip));
1564 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1565 	    CLLADDR(ifa->ifa_ifp->if_sadl));
1566 }
1567 
1568 /*
1569  * Start Duplicate Address Detection (DAD) for specified interface address.
1570  */
1571 static void
1572 arp_dad_start(struct ifaddr *ifa)
1573 {
1574 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1575 	struct dadq *dp;
1576 	char ipbuf[INET_ADDRSTRLEN];
1577 
1578 	if (!dad_init) {
1579 		TAILQ_INIT(&dadq);
1580 		mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1581 		dad_init++;
1582 	}
1583 
1584 	/*
1585 	 * If we don't need DAD, don't do it.
1586 	 * - DAD is disabled (ip_dad_count == 0)
1587 	 */
1588 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1589 		log(LOG_DEBUG,
1590 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1591 		    in_fmtaddr(ipbuf, ia->ia_addr.sin_addr),
1592 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1593 		return;
1594 	}
1595 	if (!ip_dad_count) {
1596 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1597 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1598 		arpannounce1(ifa);
1599 		return;
1600 	}
1601 	KASSERT(ifa->ifa_ifp != NULL);
1602 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1603 		return;
1604 
1605 	mutex_enter(&arp_dad_lock);
1606 	if (arp_dad_find(ifa) != NULL) {
1607 		mutex_exit(&arp_dad_lock);
1608 		/* DAD already in progress */
1609 		return;
1610 	}
1611 
1612 	dp = malloc(sizeof(*dp), M_IPARP, M_NOWAIT);
1613 	if (dp == NULL) {
1614 		mutex_exit(&arp_dad_lock);
1615 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1616 		    __func__, in_fmtaddr(ipbuf, ia->ia_addr.sin_addr),
1617 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1618 		return;
1619 	}
1620 	memset(dp, 0, sizeof(*dp));
1621 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1622 
1623 	/*
1624 	 * Send ARP packet for DAD, ip_dad_count times.
1625 	 * Note that we must delay the first transmission.
1626 	 */
1627 	dp->dad_ifa = ifa;
1628 	ifaref(ifa);	/* just for safety */
1629 	dp->dad_count = ip_dad_count;
1630 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1631 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1632 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1633 
1634 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1635 	    ARPLOGADDR(ia->ia_addr.sin_addr));
1636 
1637 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1638 
1639 	mutex_exit(&arp_dad_lock);
1640 }
1641 
1642 /*
1643  * terminate DAD unconditionally.  used for address removals.
1644  */
1645 static void
1646 arp_dad_stop(struct ifaddr *ifa)
1647 {
1648 	struct dadq *dp;
1649 
1650 	if (!dad_init)
1651 		return;
1652 
1653 	mutex_enter(&arp_dad_lock);
1654 	dp = arp_dad_find(ifa);
1655 	if (dp == NULL) {
1656 		mutex_exit(&arp_dad_lock);
1657 		/* DAD wasn't started yet */
1658 		return;
1659 	}
1660 
1661 	/* Prevent the timer from running anymore. */
1662 	TAILQ_REMOVE(&dadq, dp, dad_list);
1663 	mutex_exit(&arp_dad_lock);
1664 
1665 	arp_dad_stoptimer(dp);
1666 
1667 	free(dp, M_IPARP);
1668 	dp = NULL;
1669 	ifafree(ifa);
1670 }
1671 
1672 static void
1673 arp_dad_timer(struct ifaddr *ifa)
1674 {
1675 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1676 	struct dadq *dp;
1677 	char ipbuf[INET_ADDRSTRLEN];
1678 
1679 #ifndef NET_MPSAFE
1680 	mutex_enter(softnet_lock);
1681 	KERNEL_LOCK(1, NULL);
1682 #endif
1683 	mutex_enter(&arp_dad_lock);
1684 
1685 	/* Sanity check */
1686 	if (ia == NULL) {
1687 		log(LOG_ERR, "%s: called with null parameter\n", __func__);
1688 		goto done;
1689 	}
1690 	dp = arp_dad_find(ifa);
1691 	if (dp == NULL) {
1692 		/* DAD seems to be stopping, so do nothing. */
1693 		goto done;
1694 	}
1695 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1696 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1697 		    __func__, in_fmtaddr(ipbuf, ia->ia_addr.sin_addr),
1698 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1699 		goto done;
1700 	}
1701 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1702 	{
1703 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1704 		    __func__, in_fmtaddr(ipbuf, ia->ia_addr.sin_addr),
1705 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1706 		goto done;
1707 	}
1708 
1709 	/* timeouted with IFF_{RUNNING,UP} check */
1710 	if (dp->dad_arp_tcount > dad_maxtry) {
1711 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1712 		    if_name(ifa->ifa_ifp));
1713 
1714 		TAILQ_REMOVE(&dadq, dp, dad_list);
1715 		free(dp, M_IPARP);
1716 		dp = NULL;
1717 		ifafree(ifa);
1718 		goto done;
1719 	}
1720 
1721 	/* Need more checks? */
1722 	if (dp->dad_arp_ocount < dp->dad_count) {
1723 		int adelay;
1724 
1725 		/*
1726 		 * We have more ARP to go.  Send ARP packet for DAD.
1727 		 */
1728 		arp_dad_output(dp, ifa);
1729 		if (dp->dad_arp_ocount < dp->dad_count)
1730 			adelay = (PROBE_MIN * hz) +
1731 			    (cprng_fast32() %
1732 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1733 		else
1734 			adelay = ANNOUNCE_WAIT * hz;
1735 		arp_dad_starttimer(dp, adelay);
1736 		goto done;
1737 	} else if (dp->dad_arp_acount == 0) {
1738 		/*
1739 		 * We are done with DAD.
1740 		 * No duplicate address found.
1741 		 */
1742 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1743 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1744 		ARPLOG(LOG_DEBUG,
1745 		    "%s: DAD complete for %s - no duplicates found\n",
1746 		    if_name(ifa->ifa_ifp), ARPLOGADDR(ia->ia_addr.sin_addr));
1747 		dp->dad_arp_announce = ANNOUNCE_NUM;
1748 		goto announce;
1749 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1750 announce:
1751 		/*
1752 		 * Announce the address.
1753 		 */
1754 		arpannounce1(ifa);
1755 		dp->dad_arp_acount++;
1756 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1757 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1758 			goto done;
1759 		}
1760 		ARPLOG(LOG_DEBUG,
1761 		    "%s: ARP announcement complete for %s\n",
1762 		    if_name(ifa->ifa_ifp), ARPLOGADDR(ia->ia_addr.sin_addr));
1763 	}
1764 
1765 	TAILQ_REMOVE(&dadq, dp, dad_list);
1766 	free(dp, M_IPARP);
1767 	dp = NULL;
1768 	ifafree(ifa);
1769 
1770 done:
1771 	mutex_exit(&arp_dad_lock);
1772 #ifndef NET_MPSAFE
1773 	KERNEL_UNLOCK_ONE(NULL);
1774 	mutex_exit(softnet_lock);
1775 #endif
1776 }
1777 
1778 static void
1779 arp_dad_duplicated(struct ifaddr *ifa, const char *sha)
1780 {
1781 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1782 	struct ifnet *ifp = ifa->ifa_ifp;
1783 	char ipbuf[INET_ADDRSTRLEN];
1784 	const char *iastr;
1785 
1786 	iastr = in_fmtaddr(ipbuf, ia->ia_addr.sin_addr);
1787 
1788 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1789 		log(LOG_ERR,
1790 		    "%s: DAD duplicate address %s from %s\n",
1791 		    if_name(ifp), iastr, sha);
1792 	} else if (ia->ia_dad_defended == 0 ||
1793 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1794 		ia->ia_dad_defended = time_uptime;
1795 		arpannounce1(ifa);
1796 		log(LOG_ERR,
1797 		    "%s: DAD defended address %s from %s\n",
1798 		    if_name(ifp), iastr, sha);
1799 		return;
1800 	} else {
1801 		/* If DAD is disabled, just report the duplicate. */
1802 		if (ip_dad_count == 0) {
1803 			log(LOG_ERR,
1804 			    "%s: DAD ignoring duplicate address %s from %s\n",
1805 			    if_name(ifp), iastr, sha);
1806 			return;
1807 		}
1808 		log(LOG_ERR,
1809 		    "%s: DAD defence failed for %s from %s\n",
1810 		    if_name(ifp), iastr, sha);
1811 	}
1812 
1813 	arp_dad_stop(ifa);
1814 
1815 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1816 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1817 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1818 		/* Inform the routing socket of the duplicate address */
1819 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1820 	}
1821 }
1822 
1823 /*
1824  * Called from 10 Mb/s Ethernet interrupt handlers
1825  * when ether packet type ETHERTYPE_REVARP
1826  * is received.  Common length and type checks are done here,
1827  * then the protocol-specific routine is called.
1828  */
1829 void
1830 revarpinput(struct mbuf *m)
1831 {
1832 	struct arphdr *ar;
1833 
1834 	if (m->m_len < sizeof(struct arphdr))
1835 		goto out;
1836 	ar = mtod(m, struct arphdr *);
1837 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1838 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1839 		goto out;
1840 #endif
1841 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1842 		goto out;
1843 	switch (ntohs(ar->ar_pro)) {
1844 	case ETHERTYPE_IP:
1845 	case ETHERTYPE_IPTRAILERS:
1846 		in_revarpinput(m);
1847 		return;
1848 
1849 	default:
1850 		break;
1851 	}
1852 out:
1853 	m_freem(m);
1854 }
1855 
1856 /*
1857  * RARP for Internet protocols on 10 Mb/s Ethernet.
1858  * Algorithm is that given in RFC 903.
1859  * We are only using for bootstrap purposes to get an ip address for one of
1860  * our interfaces.  Thus we support no user-interface.
1861  *
1862  * Since the contents of the RARP reply are specific to the interface that
1863  * sent the request, this code must ensure that they are properly associated.
1864  *
1865  * Note: also supports ARP via RARP packets, per the RFC.
1866  */
1867 void
1868 in_revarpinput(struct mbuf *m)
1869 {
1870 	struct arphdr *ah;
1871 	void *tha;
1872 	int op;
1873 	struct ifnet *rcvif;
1874 	int s;
1875 
1876 	ah = mtod(m, struct arphdr *);
1877 	op = ntohs(ah->ar_op);
1878 
1879 	rcvif = m_get_rcvif(m, &s);
1880 	switch (rcvif->if_type) {
1881 	case IFT_IEEE1394:
1882 		/* ARP without target hardware address is not supported */
1883 		goto out;
1884 	default:
1885 		break;
1886 	}
1887 
1888 	switch (op) {
1889 	case ARPOP_REQUEST:
1890 	case ARPOP_REPLY:	/* per RFC */
1891 		m_put_rcvif(rcvif, &s);
1892 		in_arpinput(m);
1893 		return;
1894 	case ARPOP_REVREPLY:
1895 		break;
1896 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1897 	default:
1898 		goto out;
1899 	}
1900 	if (!revarp_in_progress)
1901 		goto out;
1902 	if (rcvif != myip_ifp) /* !same interface */
1903 		goto out;
1904 	if (myip_initialized)
1905 		goto wake;
1906 	tha = ar_tha(ah);
1907 	if (tha == NULL)
1908 		goto out;
1909 	if (ah->ar_pln != sizeof(struct in_addr))
1910 		goto out;
1911 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1912 		goto out;
1913 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1914 		goto out;
1915 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1916 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1917 	myip_initialized = 1;
1918 wake:	/* Do wakeup every time in case it was missed. */
1919 	wakeup((void *)&myip);
1920 
1921 out:
1922 	m_put_rcvif(rcvif, &s);
1923 	m_freem(m);
1924 }
1925 
1926 /*
1927  * Send a RARP request for the ip address of the specified interface.
1928  * The request should be RFC 903-compliant.
1929  */
1930 static void
1931 revarprequest(struct ifnet *ifp)
1932 {
1933 	struct sockaddr sa;
1934 	struct mbuf *m;
1935 	struct arphdr *ah;
1936 	void *tha;
1937 
1938 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1939 		return;
1940 	MCLAIM(m, &arpdomain.dom_mowner);
1941 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1942 	    2*ifp->if_addrlen;
1943 	m->m_pkthdr.len = m->m_len;
1944 	MH_ALIGN(m, m->m_len);
1945 	ah = mtod(m, struct arphdr *);
1946 	memset(ah, 0, m->m_len);
1947 	ah->ar_pro = htons(ETHERTYPE_IP);
1948 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1949 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1950 	ah->ar_op = htons(ARPOP_REVREQUEST);
1951 
1952 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1953 	tha = ar_tha(ah);
1954 	if (tha == NULL) {
1955 		m_free(m);
1956 		return;
1957 	}
1958 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1959 
1960 	sa.sa_family = AF_ARP;
1961 	sa.sa_len = 2;
1962 	m->m_flags |= M_BCAST;
1963 
1964 	if_output_lock(ifp, ifp, m, &sa, NULL);
1965 }
1966 
1967 /*
1968  * RARP for the ip address of the specified interface, but also
1969  * save the ip address of the server that sent the answer.
1970  * Timeout if no response is received.
1971  */
1972 int
1973 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1974     struct in_addr *clnt_in)
1975 {
1976 	int result, count = 20;
1977 
1978 	myip_initialized = 0;
1979 	myip_ifp = ifp;
1980 
1981 	revarp_in_progress = 1;
1982 	while (count--) {
1983 		revarprequest(ifp);
1984 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1985 		if (result != EWOULDBLOCK)
1986 			break;
1987 	}
1988 	revarp_in_progress = 0;
1989 
1990 	if (!myip_initialized)
1991 		return ENETUNREACH;
1992 
1993 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1994 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1995 	return 0;
1996 }
1997 
1998 void
1999 arp_stat_add(int type, uint64_t count)
2000 {
2001 	ARP_STATADD(type, count);
2002 }
2003 
2004 static int
2005 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
2006 {
2007 
2008 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2009 }
2010 
2011 static void
2012 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2013 {
2014 	const struct sysctlnode *node;
2015 
2016 	sysctl_createv(clog, 0, NULL, NULL,
2017 			CTLFLAG_PERMANENT,
2018 			CTLTYPE_NODE, "inet", NULL,
2019 			NULL, 0, NULL, 0,
2020 			CTL_NET, PF_INET, CTL_EOL);
2021 	sysctl_createv(clog, 0, NULL, &node,
2022 			CTLFLAG_PERMANENT,
2023 			CTLTYPE_NODE, "arp",
2024 			SYSCTL_DESCR("Address Resolution Protocol"),
2025 			NULL, 0, NULL, 0,
2026 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2027 
2028 	sysctl_createv(clog, 0, NULL, NULL,
2029 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2030 			CTLTYPE_INT, "keep",
2031 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
2032 			NULL, 0, &arpt_keep, 0,
2033 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2034 
2035 	sysctl_createv(clog, 0, NULL, NULL,
2036 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2037 			CTLTYPE_INT, "down",
2038 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
2039 			NULL, 0, &arpt_down, 0,
2040 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2041 
2042 	sysctl_createv(clog, 0, NULL, NULL,
2043 			CTLFLAG_PERMANENT,
2044 			CTLTYPE_STRUCT, "stats",
2045 			SYSCTL_DESCR("ARP statistics"),
2046 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2047 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2048 
2049 	sysctl_createv(clog, 0, NULL, NULL,
2050 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2051 			CTLTYPE_INT, "log_movements",
2052 			SYSCTL_DESCR("log ARP replies from MACs different than"
2053 			    " the one in the cache"),
2054 			NULL, 0, &log_movements, 0,
2055 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2056 
2057 	sysctl_createv(clog, 0, NULL, NULL,
2058 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2059 			CTLTYPE_INT, "log_permanent_modify",
2060 			SYSCTL_DESCR("log ARP replies from MACs different than"
2061 			    " the one in the permanent arp entry"),
2062 			NULL, 0, &log_permanent_modify, 0,
2063 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2064 
2065 	sysctl_createv(clog, 0, NULL, NULL,
2066 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2067 			CTLTYPE_INT, "log_wrong_iface",
2068 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2069 			    " interface"),
2070 			NULL, 0, &log_wrong_iface, 0,
2071 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2072 
2073 	sysctl_createv(clog, 0, NULL, NULL,
2074 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2075 			CTLTYPE_INT, "log_unknown_network",
2076 			SYSCTL_DESCR("log ARP packets from non-local network"),
2077 			NULL, 0, &log_unknown_network, 0,
2078 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2079 
2080 	sysctl_createv(clog, 0, NULL, NULL,
2081 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2082 		       CTLTYPE_INT, "debug",
2083 		       SYSCTL_DESCR("Enable ARP DAD debug output"),
2084 		       NULL, 0, &arp_debug, 0,
2085 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2086 }
2087 
2088 #endif /* INET */
2089