xref: /netbsd-src/sys/netinet/if_arp.c (revision 16dce51364ebe8aeafbae46bc5aa167b8115bc45)
1 /*	$NetBSD: if_arp.c,v 1.273 2018/04/11 05:38:47 maxv 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 1.273 2018/04/11 05:38:47 maxv 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/kmem.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 /*
137  * ARP trailer negotiation.  Trailer protocol is not IP specific,
138  * but ARP request/response use IP addresses.
139  */
140 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
141 
142 /* timer values */
143 static int arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
144 static int arpt_down = 20;		/* once declared down, don't send for 20 secs */
145 static int arp_maxhold = 1;	/* number of packets to hold per ARP entry */
146 #define	rt_expire rt_rmx.rmx_expire
147 #define	rt_pksent rt_rmx.rmx_pksent
148 
149 int ip_dad_count = PROBE_NUM;
150 #ifdef ARP_DEBUG
151 int arp_debug = 1;
152 #else
153 int arp_debug = 0;
154 #endif
155 
156 static void arp_init(void);
157 
158 static void arprequest(struct ifnet *,
159     const struct in_addr *, const struct in_addr *,
160     const uint8_t *);
161 static void arpannounce1(struct ifaddr *);
162 static struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
163     const struct sockaddr *);
164 static void arptimer(void *);
165 static void arp_settimer(struct llentry *, int);
166 static struct llentry *arplookup(struct ifnet *,
167     const struct in_addr *, const struct sockaddr *, int);
168 static struct llentry *arpcreate(struct ifnet *,
169     const struct in_addr *, const struct sockaddr *, int);
170 static void in_arpinput(struct mbuf *);
171 static void in_revarpinput(struct mbuf *);
172 static void revarprequest(struct ifnet *);
173 
174 static void arp_drainstub(void);
175 
176 struct dadq;
177 static void arp_dad_timer(struct dadq *);
178 static void arp_dad_start(struct ifaddr *);
179 static void arp_dad_stop(struct ifaddr *);
180 static void arp_dad_duplicated(struct ifaddr *, const char *);
181 
182 static void arp_init_llentry(struct ifnet *, struct llentry *);
183 #if NTOKEN > 0
184 static void arp_free_llentry_tokenring(struct llentry *);
185 #endif
186 
187 struct ifqueue arpintrq = {
188 	.ifq_head = NULL,
189 	.ifq_tail = NULL,
190 	.ifq_len = 0,
191 	.ifq_maxlen = 50,
192 	.ifq_drops = 0,
193 };
194 static int arp_maxtries = 5;
195 static int useloopback = 1;	/* use loopback interface for local traffic */
196 
197 static percpu_t *arpstat_percpu;
198 
199 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
200 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
201 
202 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
203 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
204 
205 /* revarp state */
206 static struct in_addr myip, srv_ip;
207 static int myip_initialized = 0;
208 static int revarp_in_progress = 0;
209 static struct ifnet *myip_ifp = NULL;
210 
211 static int arp_drainwanted;
212 
213 static int log_movements = 1;
214 static int log_permanent_modify = 1;
215 static int log_wrong_iface = 1;
216 static int log_unknown_network = 1;
217 
218 /*
219  * this should be elsewhere.
220  */
221 
222 #define	LLA_ADDRSTRLEN	(16 * 3)
223 
224 static char *
225 lla_snprintf(char *, const u_int8_t *, int);
226 
227 static char *
228 lla_snprintf(char *dst, const u_int8_t *adrp, int len)
229 {
230 	int i;
231 	char *p;
232 
233 	p = dst;
234 
235 	*p++ = hexdigits[(*adrp) >> 4];
236 	*p++ = hexdigits[(*adrp++) & 0xf];
237 
238 	for (i = 1; i < len && i < 16; i++) {
239 		*p++ = ':';
240 		*p++ = hexdigits[(*adrp) >> 4];
241 		*p++ = hexdigits[(*adrp++) & 0xf];
242 	}
243 
244 	*p = 0;
245 	return dst;
246 }
247 
248 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
249 
250 static void
251 arp_fasttimo(void)
252 {
253 	if (arp_drainwanted) {
254 		arp_drain();
255 		arp_drainwanted = 0;
256 	}
257 }
258 
259 const struct protosw arpsw[] = {
260 	{
261 		.pr_type = 0,
262 		.pr_domain = &arpdomain,
263 		.pr_protocol = 0,
264 		.pr_flags = 0,
265 		.pr_input = 0,
266 		.pr_ctlinput = 0,
267 		.pr_ctloutput = 0,
268 		.pr_usrreqs = 0,
269 		.pr_init = arp_init,
270 		.pr_fasttimo = arp_fasttimo,
271 		.pr_slowtimo = 0,
272 		.pr_drain = arp_drainstub,
273 	}
274 };
275 
276 struct domain arpdomain = {
277 	.dom_family = PF_ARP,
278 	.dom_name = "arp",
279 	.dom_protosw = arpsw,
280 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
281 };
282 
283 static void sysctl_net_inet_arp_setup(struct sysctllog **);
284 
285 void
286 arp_init(void)
287 {
288 
289 	sysctl_net_inet_arp_setup(NULL);
290 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
291 	IFQ_LOCK_INIT(&arpintrq);
292 }
293 
294 static void
295 arp_drainstub(void)
296 {
297 	arp_drainwanted = 1;
298 }
299 
300 /*
301  * ARP protocol drain routine.  Called when memory is in short supply.
302  * Called at splvm();  don't acquire softnet_lock as can be called from
303  * hardware interrupt handlers.
304  */
305 void
306 arp_drain(void)
307 {
308 
309 	lltable_drain(AF_INET);
310 }
311 
312 static void
313 arptimer(void *arg)
314 {
315 	struct llentry *lle = arg;
316 	struct ifnet *ifp;
317 
318 	KASSERT((lle->la_flags & LLE_STATIC) == 0);
319 
320 	LLE_WLOCK(lle);
321 
322 	/*
323 	 * This shortcut is required to avoid trying to touch ifp that may be
324 	 * being destroyed.
325 	 */
326 	if ((lle->la_flags & LLE_LINKED) == 0) {
327 		LLE_FREE_LOCKED(lle);
328 		return;
329 	}
330 
331 	ifp = lle->lle_tbl->llt_ifp;
332 
333 	/* XXX: LOR avoidance. We still have ref on lle. */
334 	LLE_WUNLOCK(lle);
335 
336 	IF_AFDATA_LOCK(ifp);
337 	LLE_WLOCK(lle);
338 
339 	/* Guard against race with other llentry_free(). */
340 	if (lle->la_flags & LLE_LINKED) {
341 		size_t pkts_dropped;
342 
343 		LLE_REMREF(lle);
344 		pkts_dropped = llentry_free(lle);
345 		ARP_STATADD(ARP_STAT_DFRDROPPED, pkts_dropped);
346 		ARP_STATADD(ARP_STAT_DFRTOTAL, pkts_dropped);
347 	} else {
348 		LLE_FREE_LOCKED(lle);
349 	}
350 
351 	IF_AFDATA_UNLOCK(ifp);
352 }
353 
354 static void
355 arp_settimer(struct llentry *la, int sec)
356 {
357 
358 	LLE_WLOCK_ASSERT(la);
359 	KASSERT((la->la_flags & LLE_STATIC) == 0);
360 
361 	/*
362 	 * We have to take care of a reference leak which occurs if
363 	 * callout_reset overwrites a pending callout schedule.  Unfortunately
364 	 * we don't have a mean to know the overwrite, so we need to know it
365 	 * using callout_stop.  We need to call callout_pending first to exclude
366 	 * the case that the callout has never been scheduled.
367 	 */
368 	if (callout_pending(&la->la_timer)) {
369 		bool expired = callout_stop(&la->la_timer);
370 		if (!expired)
371 			/* A pending callout schedule is canceled. */
372 			LLE_REMREF(la);
373 	}
374 	LLE_ADDREF(la);
375 	callout_reset(&la->la_timer, hz * sec, arptimer, la);
376 }
377 
378 /*
379  * We set the gateway for RTF_CLONING routes to a "prototype"
380  * link-layer sockaddr whose interface type (if_type) and interface
381  * index (if_index) fields are prepared.
382  */
383 static struct sockaddr *
384 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
385     const struct sockaddr *netmask)
386 {
387 	const struct ifnet *ifp = rt->rt_ifp;
388 	uint8_t namelen = strlen(ifp->if_xname);
389 	uint8_t addrlen = ifp->if_addrlen;
390 
391 	/*
392 	 * XXX: If this is a manually added route to interface
393 	 * such as older version of routed or gated might provide,
394 	 * restore cloning bit.
395 	 */
396 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
397 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
398 		rt->rt_flags |= RTF_CONNECTED;
399 
400 	if ((rt->rt_flags & (RTF_CONNECTED | RTF_LOCAL))) {
401 		union {
402 			struct sockaddr sa;
403 			struct sockaddr_storage ss;
404 			struct sockaddr_dl sdl;
405 		} u;
406 		/*
407 		 * Case 1: This route should come from a route to iface.
408 		 */
409 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
410 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
411 		rt_setgate(rt, &u.sa);
412 		gate = rt->rt_gateway;
413 	}
414 	return gate;
415 }
416 
417 static void
418 arp_init_llentry(struct ifnet *ifp, struct llentry *lle)
419 {
420 
421 	switch (ifp->if_type) {
422 #if NTOKEN > 0
423 	case IFT_ISO88025:
424 		lle->la_opaque = kmem_intr_alloc(sizeof(struct token_rif),
425 		    KM_NOSLEEP);
426 		lle->lle_ll_free = arp_free_llentry_tokenring;
427 		break;
428 #endif
429 	}
430 }
431 
432 #if NTOKEN > 0
433 static void
434 arp_free_llentry_tokenring(struct llentry *lle)
435 {
436 
437 	kmem_intr_free(lle->la_opaque, sizeof(struct token_rif));
438 }
439 #endif
440 
441 /*
442  * Parallel to llc_rtrequest.
443  */
444 void
445 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
446 {
447 	struct sockaddr *gate = rt->rt_gateway;
448 	struct in_ifaddr *ia;
449 	struct ifaddr *ifa;
450 	struct ifnet *ifp = rt->rt_ifp;
451 	int bound;
452 	int s;
453 
454 	if (req == RTM_LLINFO_UPD) {
455 		if ((ifa = info->rti_ifa) != NULL)
456 			arpannounce1(ifa);
457 		return;
458 	}
459 
460 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
461 		if (req != RTM_ADD)
462 			return;
463 
464 		/*
465 		 * linklayers with particular link MTU limitation.
466 		 */
467 		switch(ifp->if_type) {
468 #if NFDDI > 0
469 		case IFT_FDDI:
470 			if (ifp->if_mtu > FDDIIPMTU)
471 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
472 			break;
473 #endif
474 #if NARCNET > 0
475 		case IFT_ARCNET:
476 		    {
477 			int arcipifmtu;
478 
479 			if (ifp->if_flags & IFF_LINK0)
480 				arcipifmtu = arc_ipmtu;
481 			else
482 				arcipifmtu = ARCMTU;
483 			if (ifp->if_mtu > arcipifmtu)
484 				rt->rt_rmx.rmx_mtu = arcipifmtu;
485 			break;
486 		    }
487 #endif
488 		}
489 		return;
490 	}
491 
492 	switch (req) {
493 	case RTM_SETGATE:
494 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
495 		break;
496 	case RTM_ADD:
497 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
498 		if (gate == NULL) {
499 			log(LOG_ERR, "%s: arp_setgate failed\n", __func__);
500 			break;
501 		}
502 		if ((rt->rt_flags & RTF_CONNECTED) ||
503 		    (rt->rt_flags & RTF_LOCAL)) {
504 			/*
505 			 * Give this route an expiration time, even though
506 			 * it's a "permanent" route, so that routes cloned
507 			 * from it do not need their expiration time set.
508 			 */
509 			KASSERT(time_uptime != 0);
510 			rt->rt_expire = time_uptime;
511 			/*
512 			 * linklayers with particular link MTU limitation.
513 			 */
514 			switch (ifp->if_type) {
515 #if NFDDI > 0
516 			case IFT_FDDI:
517 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
518 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
519 				     (rt->rt_rmx.rmx_mtu == 0 &&
520 				      ifp->if_mtu > FDDIIPMTU)))
521 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
522 				break;
523 #endif
524 #if NARCNET > 0
525 			case IFT_ARCNET:
526 			    {
527 				int arcipifmtu;
528 				if (ifp->if_flags & IFF_LINK0)
529 					arcipifmtu = arc_ipmtu;
530 				else
531 					arcipifmtu = ARCMTU;
532 
533 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
534 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
535 				     (rt->rt_rmx.rmx_mtu == 0 &&
536 				      ifp->if_mtu > arcipifmtu)))
537 					rt->rt_rmx.rmx_mtu = arcipifmtu;
538 				break;
539 			    }
540 #endif
541 			}
542 			if (rt->rt_flags & RTF_CONNECTED)
543 				break;
544 		}
545 
546 		bound = curlwp_bind();
547 		/* Announce a new entry if requested. */
548 		if (rt->rt_flags & RTF_ANNOUNCE) {
549 			struct psref psref;
550 			ia = in_get_ia_on_iface_psref(
551 			    satocsin(rt_getkey(rt))->sin_addr, ifp, &psref);
552 			if (ia != NULL) {
553 				arpannounce(ifp, &ia->ia_ifa,
554 				    CLLADDR(satocsdl(gate)));
555 				ia4_release(ia, &psref);
556 			}
557 		}
558 
559 		if (gate->sa_family != AF_LINK ||
560 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
561 			log(LOG_DEBUG, "%s: bad gateway value\n", __func__);
562 			goto out;
563 		}
564 
565 		satosdl(gate)->sdl_type = ifp->if_type;
566 		satosdl(gate)->sdl_index = ifp->if_index;
567 
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 		 */
574 		if (rt->rt_flags & RTF_HOST &&
575 		    !(rt->rt_flags & RTF_BROADCAST) &&
576 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
577 			rt->rt_flags |= RTF_BROADCAST;
578 		/* There is little point in resolving the broadcast address */
579 		if (rt->rt_flags & RTF_BROADCAST)
580 			goto out;
581 
582 		/*
583 		 * When called from rt_ifa_addlocal, we cannot depend on that
584 		 * the address (rt_getkey(rt)) exits in the address list of the
585 		 * interface. So check RTF_LOCAL instead.
586 		 */
587 		if (rt->rt_flags & RTF_LOCAL) {
588 			rt->rt_expire = 0;
589 			if (useloopback) {
590 				rt->rt_ifp = lo0ifp;
591 				rt->rt_rmx.rmx_mtu = 0;
592 			}
593 			goto out;
594 		}
595 
596 		s = pserialize_read_enter();
597 		ia = in_get_ia_on_iface(satocsin(rt_getkey(rt))->sin_addr, ifp);
598 		if (ia == NULL) {
599 			pserialize_read_exit(s);
600 			goto out;
601 		}
602 
603 		rt->rt_expire = 0;
604 		if (useloopback) {
605 			rt->rt_ifp = lo0ifp;
606 			rt->rt_rmx.rmx_mtu = 0;
607 		}
608 		rt->rt_flags |= RTF_LOCAL;
609 		/*
610 		 * make sure to set rt->rt_ifa to the interface
611 		 * address we are using, otherwise we will have trouble
612 		 * with source address selection.
613 		 */
614 		ifa = &ia->ia_ifa;
615 		if (ifa != rt->rt_ifa)
616 			/* Assume it doesn't sleep */
617 			rt_replace_ifa(rt, ifa);
618 		pserialize_read_exit(s);
619 	out:
620 		curlwp_bindx(bound);
621 		break;
622 	}
623 }
624 
625 /*
626  * Broadcast an ARP request. Caller specifies:
627  *	- arp header source ip address
628  *	- arp header target ip address
629  *	- arp header source ethernet address
630  */
631 static void
632 arprequest(struct ifnet *ifp,
633     const struct in_addr *sip, const struct in_addr *tip,
634     const uint8_t *enaddr)
635 {
636 	struct mbuf *m;
637 	struct arphdr *ah;
638 	struct sockaddr sa;
639 	uint64_t *arps;
640 
641 	KASSERT(sip != NULL);
642 	KASSERT(tip != NULL);
643 	KASSERT(enaddr != NULL);
644 
645 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
646 		return;
647 	MCLAIM(m, &arpdomain.dom_mowner);
648 	switch (ifp->if_type) {
649 	case IFT_IEEE1394:
650 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
651 		    ifp->if_addrlen;
652 		break;
653 	default:
654 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
655 		    2 * ifp->if_addrlen;
656 		break;
657 	}
658 	m->m_pkthdr.len = m->m_len;
659 	MH_ALIGN(m, m->m_len);
660 	ah = mtod(m, struct arphdr *);
661 	memset(ah, 0, m->m_len);
662 	switch (ifp->if_type) {
663 	case IFT_IEEE1394:	/* RFC2734 */
664 		/* fill it now for ar_tpa computation */
665 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
666 		break;
667 	default:
668 		/* ifp->if_output will fill ar_hrd */
669 		break;
670 	}
671 	ah->ar_pro = htons(ETHERTYPE_IP);
672 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
673 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
674 	ah->ar_op = htons(ARPOP_REQUEST);
675 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
676 	memcpy(ar_spa(ah), sip, ah->ar_pln);
677 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
678 	sa.sa_family = AF_ARP;
679 	sa.sa_len = 2;
680 	m->m_flags |= M_BCAST;
681 	arps = ARP_STAT_GETREF();
682 	arps[ARP_STAT_SNDTOTAL]++;
683 	arps[ARP_STAT_SENDREQUEST]++;
684 	ARP_STAT_PUTREF();
685 	if_output_lock(ifp, ifp, m, &sa, NULL);
686 }
687 
688 void
689 arpannounce(struct ifnet *ifp, struct ifaddr *ifa, const uint8_t *enaddr)
690 {
691 	struct in_ifaddr *ia = ifatoia(ifa);
692 	struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
693 
694 	if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) {
695 		ARPLOG(LOG_DEBUG, "%s not ready\n", ARPLOGADDR(ip));
696 		return;
697 	}
698 	arprequest(ifp, ip, ip, enaddr);
699 }
700 
701 static void
702 arpannounce1(struct ifaddr *ifa)
703 {
704 
705 	arpannounce(ifa->ifa_ifp, ifa, CLLADDR(ifa->ifa_ifp->if_sadl));
706 }
707 
708 /*
709  * Resolve an IP address into an ethernet address.  If success, desten is
710  * filled in. If there is no entry in arptab, set one up and broadcast a
711  * request for the IP address. Hold onto this mbuf and resend it once the
712  * address is finally resolved.
713  *
714  * A return value of 0 indicates that desten has been filled in and the packet
715  * should be sent normally; a return value of EWOULDBLOCK indicates that the
716  * packet has been held pending resolution. Any other value indicates an
717  * error.
718  */
719 int
720 arpresolve(struct ifnet *ifp, const struct rtentry *rt, struct mbuf *m,
721     const struct sockaddr *dst, void *desten, size_t destlen)
722 {
723 	struct llentry *la;
724 	const char *create_lookup;
725 	bool renew;
726 	int error;
727 	struct ifnet *origifp = ifp;
728 
729 #if NCARP > 0
730 	if (rt != NULL && rt->rt_ifp->if_type == IFT_CARP)
731 		ifp = rt->rt_ifp;
732 #endif
733 
734 	KASSERT(m != NULL);
735 
736 	la = arplookup(ifp, NULL, dst, 0);
737 	if (la == NULL)
738 		goto notfound;
739 
740 	if ((la->la_flags & LLE_VALID) &&
741 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
742 		KASSERT(destlen >= ifp->if_addrlen);
743 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
744 		LLE_RUNLOCK(la);
745 		return 0;
746 	}
747 
748 notfound:
749 	if (ifp->if_flags & IFF_NOARP) {
750 		if (la != NULL)
751 			LLE_RUNLOCK(la);
752 		error = ENOTSUP;
753 		goto bad;
754 	}
755 
756 	if (la == NULL) {
757 		struct rtentry *_rt;
758 
759 		create_lookup = "create";
760 		_rt = rtalloc1(dst, 0);
761 		IF_AFDATA_WLOCK(ifp);
762 		la = lla_create(LLTABLE(ifp), LLE_EXCLUSIVE, dst, _rt);
763 		IF_AFDATA_WUNLOCK(ifp);
764 		if (_rt != NULL)
765 			rt_unref(_rt);
766 		if (la == NULL)
767 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
768 		else {
769 			struct sockaddr_in sin;
770 
771 			arp_init_llentry(ifp, la);
772 			sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
773 			rt_clonedmsg(sintosa(&sin), ifp, rt);
774 		}
775 	} else if (LLE_TRY_UPGRADE(la) == 0) {
776 		create_lookup = "lookup";
777 		LLE_RUNLOCK(la);
778 		IF_AFDATA_RLOCK(ifp);
779 		la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
780 		IF_AFDATA_RUNLOCK(ifp);
781 	}
782 
783 	error = EINVAL;
784 	if (la == NULL) {
785 		log(LOG_DEBUG,
786 		    "%s: failed to %s llentry for %s on %s\n",
787 		    __func__, create_lookup, inet_ntoa(satocsin(dst)->sin_addr),
788 		    ifp->if_xname);
789 		goto bad;
790 	}
791 
792 	if ((la->la_flags & LLE_VALID) &&
793 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime))
794 	{
795 		KASSERT(destlen >= ifp->if_addrlen);
796 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
797 		renew = false;
798 		/*
799 		 * If entry has an expiry time and it is approaching,
800 		 * see if we need to send an ARP request within this
801 		 * arpt_down interval.
802 		 */
803 		if (!(la->la_flags & LLE_STATIC) &&
804 		    time_uptime + la->la_preempt > la->la_expire)
805 		{
806 			renew = true;
807 			la->la_preempt--;
808 		}
809 
810 		LLE_WUNLOCK(la);
811 
812 		if (renew) {
813 			const uint8_t *enaddr = CLLADDR(ifp->if_sadl);
814 			arprequest(origifp,
815 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
816 			    &satocsin(dst)->sin_addr, enaddr);
817 		}
818 
819 		return 0;
820 	}
821 
822 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
823 		LLE_RUNLOCK(la);
824 		log(LOG_DEBUG, "%s: ouch, empty static llinfo for %s\n",
825 		    __func__, inet_ntoa(satocsin(dst)->sin_addr));
826 		error = EINVAL;
827 		goto bad;
828 	}
829 
830 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
831 
832 	/*
833 	 * There is an arptab entry, but no ethernet address
834 	 * response yet.  Add the mbuf to the list, dropping
835 	 * the oldest packet if we have exceeded the system
836 	 * setting.
837 	 */
838 	LLE_WLOCK_ASSERT(la);
839 	if (la->la_numheld >= arp_maxhold) {
840 		if (la->la_hold != NULL) {
841 			struct mbuf *next = la->la_hold->m_nextpkt;
842 			m_freem(la->la_hold);
843 			la->la_hold = next;
844 			la->la_numheld--;
845 			ARP_STATINC(ARP_STAT_DFRDROPPED);
846 			ARP_STATINC(ARP_STAT_DFRTOTAL);
847 		}
848 	}
849 	if (la->la_hold != NULL) {
850 		struct mbuf *curr = la->la_hold;
851 		while (curr->m_nextpkt != NULL)
852 			curr = curr->m_nextpkt;
853 		curr->m_nextpkt = m;
854 	} else
855 		la->la_hold = m;
856 	la->la_numheld++;
857 	if (!renew)
858 		LLE_DOWNGRADE(la);
859 
860 	/*
861 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
862 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
863 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
864 	 * ARP request, but not faster than one request per second.
865 	 */
866 	if (la->la_asked < arp_maxtries)
867 		error = EWOULDBLOCK;	/* First request. */
868 	else
869 		error = (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
870 		    EHOSTUNREACH : EHOSTDOWN;
871 
872 	if (renew) {
873 		const uint8_t *enaddr = CLLADDR(ifp->if_sadl);
874 		la->la_expire = time_uptime;
875 		arp_settimer(la, arpt_down);
876 		la->la_asked++;
877 		LLE_WUNLOCK(la);
878 
879 		if (rt != NULL) {
880 			arprequest(origifp,
881 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
882 			    &satocsin(dst)->sin_addr, enaddr);
883 		} else {
884 			struct sockaddr_in sin;
885 			struct rtentry *_rt;
886 
887 			sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
888 
889 			/* XXX */
890 			_rt = rtalloc1((struct sockaddr *)&sin, 0);
891 			if (_rt == NULL)
892 				goto bad;
893 			arprequest(origifp,
894 			    &satocsin(_rt->rt_ifa->ifa_addr)->sin_addr,
895 			    &satocsin(dst)->sin_addr, enaddr);
896 			rt_unref(_rt);
897 		}
898 		return error;
899 	}
900 
901 	LLE_RUNLOCK(la);
902 	return error;
903 
904 bad:
905 	m_freem(m);
906 	return error;
907 }
908 
909 /*
910  * Common length and type checks are done here,
911  * then the protocol-specific routine is called.
912  */
913 void
914 arpintr(void)
915 {
916 	struct mbuf *m;
917 	struct arphdr *ar;
918 	int s;
919 	int arplen;
920 
921 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
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 		arplen = sizeof(struct arphdr);
937 		if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
938 			goto badlen;
939 		ar = mtod(m, struct arphdr *);
940 
941 		rcvif = m_get_rcvif(m, &s);
942 		if (__predict_false(rcvif == NULL)) {
943 			ARP_STATINC(ARP_STAT_RCVNOINT);
944 			goto free;
945 		}
946 
947 		/*
948 		 * We don't want non-IEEE1394 ARP packets on IEEE1394
949 		 * interfaces, and vice versa. Our life depends on that.
950 		 */
951 		switch (rcvif->if_type) {
952 		case IFT_IEEE1394:
953 			if (ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
954 				m_put_rcvif(rcvif, &s);
955 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
956 				goto free;
957 			}
958 
959 			arplen = sizeof(struct arphdr) +
960 			    ar->ar_hln + 2 * ar->ar_pln;
961 			break;
962 		default:
963 			if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394) {
964 				m_put_rcvif(rcvif, &s);
965 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
966 				goto free;
967 			}
968 
969 			arplen = sizeof(struct arphdr) +
970 			    2 * ar->ar_hln + 2 * ar->ar_pln;
971 			break;
972 		}
973 
974 		m_put_rcvif(rcvif, &s);
975 
976 		if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
977 			goto badlen;
978 		ar = mtod(m, struct arphdr *);
979 
980 		switch (ntohs(ar->ar_pro)) {
981 		case ETHERTYPE_IP:
982 		case ETHERTYPE_IPTRAILERS:
983 			in_arpinput(m);
984 			continue;
985 		default:
986 			ARP_STATINC(ARP_STAT_RCVBADPROTO);
987 			goto free;
988 		}
989 
990 badlen:
991 		ARP_STATINC(ARP_STAT_RCVBADLEN);
992 free:
993 		m_freem(m);
994 	}
995 
996 out:
997 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
998 	return; /* XXX gcc */
999 }
1000 
1001 /*
1002  * ARP for Internet protocols on 10 Mb/s Ethernet. Algorithm is that given in
1003  * RFC 826. In addition, a sanity check is performed on the sender protocol
1004  * address, to catch impersonators.
1005  *
1006  * We no longer handle negotiations for use of trailer protocol: formerly, ARP
1007  * replied for protocol type ETHERTYPE_TRAIL sent along with IP replies if we
1008  * wanted trailers sent to us, and also sent them in response to IP replies.
1009  * This allowed either end to announce the desire to receive trailer packets.
1010  *
1011  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, but
1012  * formerly didn't normally send requests.
1013  */
1014 static void
1015 in_arpinput(struct mbuf *m)
1016 {
1017 	struct arphdr *ah;
1018 	struct ifnet *ifp, *rcvif = NULL;
1019 	struct llentry *la = NULL;
1020 	struct in_ifaddr *ia = NULL;
1021 #if NBRIDGE > 0
1022 	struct in_ifaddr *bridge_ia = NULL;
1023 #endif
1024 #if NCARP > 0
1025 	uint32_t count = 0, index = 0;
1026 #endif
1027 	struct sockaddr sa;
1028 	struct in_addr isaddr, itaddr, myaddr;
1029 	int op;
1030 	void *tha;
1031 	uint64_t *arps;
1032 	struct psref psref, psref_ia;
1033 	int s;
1034 	char llabuf[LLA_ADDRSTRLEN];
1035 	char ipbuf[INET_ADDRSTRLEN];
1036 	bool do_dad;
1037 
1038 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
1039 		goto out;
1040 	ah = mtod(m, struct arphdr *);
1041 	op = ntohs(ah->ar_op);
1042 
1043 	if (ah->ar_pln != sizeof(struct in_addr))
1044 		goto out;
1045 
1046 	ifp = if_get_bylla(ar_sha(ah), ah->ar_hln, &psref);
1047 	if (ifp) {
1048 		/* it's from me, ignore it. */
1049 		if_put(ifp, &psref);
1050 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1051 		goto out;
1052 	}
1053 
1054 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
1055 	if (__predict_false(rcvif == NULL))
1056 		goto out;
1057 	if (rcvif->if_flags & IFF_NOARP)
1058 		goto out;
1059 
1060 	memcpy(&isaddr, ar_spa(ah), sizeof(isaddr));
1061 	memcpy(&itaddr, ar_tpa(ah), sizeof(itaddr));
1062 
1063 	if (m->m_flags & (M_BCAST|M_MCAST))
1064 		ARP_STATINC(ARP_STAT_RCVMCAST);
1065 
1066 	/*
1067 	 * Search for a matching interface address
1068 	 * or any address on the interface to use
1069 	 * as a dummy address in the rest of this function.
1070 	 *
1071 	 * If the target IP address is zero then try and find
1072 	 * the sender address for DAD.
1073 	 */
1074 	myaddr = in_nullhost(itaddr) ? isaddr : itaddr;
1075 	s = pserialize_read_enter();
1076 	IN_ADDRHASH_READER_FOREACH(ia, myaddr.s_addr) {
1077 		if (!in_hosteq(ia->ia_addr.sin_addr, myaddr))
1078 			continue;
1079 #if NCARP > 0
1080 		if (ia->ia_ifp->if_type == IFT_CARP &&
1081 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1082 		    (IFF_UP|IFF_RUNNING))) {
1083 			index++;
1084 			/* XXX: ar_hln? */
1085 			if (ia->ia_ifp == rcvif && (ah->ar_hln >= 6) &&
1086 			    carp_iamatch(ia, ar_sha(ah),
1087 			    &count, index)) {
1088 				break;
1089 			}
1090 		} else
1091 #endif
1092 		if (ia->ia_ifp == rcvif)
1093 			break;
1094 #if NBRIDGE > 0
1095 		/*
1096 		 * If the interface we received the packet on
1097 		 * is part of a bridge, check to see if we need
1098 		 * to "bridge" the packet to ourselves at this
1099 		 * layer.  Note we still prefer a perfect match,
1100 		 * but allow this weaker match if necessary.
1101 		 */
1102 		if (rcvif->if_bridge != NULL &&
1103 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
1104 			bridge_ia = ia;
1105 #endif
1106 	}
1107 
1108 #if NBRIDGE > 0
1109 	if (ia == NULL && bridge_ia != NULL) {
1110 		ia = bridge_ia;
1111 		m_put_rcvif_psref(rcvif, &psref);
1112 		rcvif = NULL;
1113 		/* FIXME */
1114 		ifp = bridge_ia->ia_ifp;
1115 	}
1116 #endif
1117 	if (ia != NULL)
1118 		ia4_acquire(ia, &psref_ia);
1119 	pserialize_read_exit(s);
1120 
1121 	if (ah->ar_hln != ifp->if_addrlen) {
1122 		ARP_STATINC(ARP_STAT_RCVBADLEN);
1123 		log(LOG_WARNING,
1124 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
1125 		    IN_PRINT(ipbuf, &isaddr), ah->ar_hln, ifp->if_addrlen);
1126 		goto out;
1127 	}
1128 
1129 	/* Only do DaD if we have a matching address. */
1130 	do_dad = (ia != NULL);
1131 
1132 	if (ia == NULL) {
1133 		ia = in_get_ia_on_iface_psref(isaddr, rcvif, &psref_ia);
1134 		if (ia == NULL) {
1135 			ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
1136 			if (ia == NULL) {
1137 				ARP_STATINC(ARP_STAT_RCVNOINT);
1138 				goto out;
1139 			}
1140 		}
1141 	}
1142 
1143 	myaddr = ia->ia_addr.sin_addr;
1144 
1145 	/* XXX checks for bridge case? */
1146 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1147 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1148 		log(LOG_ERR,
1149 		    "%s: arp: link address is broadcast for IP address %s!\n",
1150 		    ifp->if_xname, IN_PRINT(ipbuf, &isaddr));
1151 		goto out;
1152 	}
1153 
1154 	/*
1155 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
1156 	 */
1157 	if (in_nullhost(isaddr))
1158 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
1159 	else if (in_hosteq(isaddr, myaddr))
1160 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1161 
1162 	if (in_nullhost(itaddr))
1163 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
1164 
1165 	/*
1166 	 * DAD check, RFC 5227.
1167 	 * Collision on sender address is always a duplicate.
1168 	 * Collision on target address is only a duplicate IF
1169 	 * the sender address is the null host (ie a DAD probe) AND
1170 	 * our address is in the TENTATIVE state.
1171 	 * DUPLICATED state is also checked so that processing stops here
1172 	 * and an error can be logged.
1173 	 */
1174 	if (do_dad &&
1175 	    (in_hosteq(isaddr, myaddr) ||
1176 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)
1177 	    && ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DUPLICATED))))
1178 	{
1179 		arp_dad_duplicated((struct ifaddr *)ia,
1180 		    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln));
1181 		goto out;
1182 	}
1183 
1184 	/*
1185 	 * If the target IP address is zero, ignore the packet.
1186 	 * This prevents the code below from trying to answer
1187 	 * when we are using IP address zero (booting).
1188 	 */
1189 	if (in_nullhost(itaddr))
1190 		goto out;
1191 
1192 	if (in_nullhost(isaddr))
1193 		goto reply;
1194 
1195 	if (in_hosteq(itaddr, myaddr))
1196 		la = arpcreate(ifp, &isaddr, NULL, 1);
1197 	else
1198 		la = arplookup(ifp, &isaddr, NULL, 1);
1199 	if (la == NULL)
1200 		goto reply;
1201 
1202 	if ((la->la_flags & LLE_VALID) &&
1203 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
1204 		if (la->la_flags & LLE_STATIC) {
1205 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
1206 			if (!log_permanent_modify)
1207 				goto out;
1208 			log(LOG_INFO,
1209 			    "%s tried to overwrite permanent arp info"
1210 			    " for %s\n",
1211 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1212 			    IN_PRINT(ipbuf, &isaddr));
1213 			goto out;
1214 		} else if (la->lle_tbl->llt_ifp != ifp) {
1215 			/* XXX should not happen? */
1216 			ARP_STATINC(ARP_STAT_RCVOVERINT);
1217 			if (!log_wrong_iface)
1218 				goto out;
1219 			log(LOG_INFO,
1220 			    "%s on %s tried to overwrite "
1221 			    "arp info for %s on %s\n",
1222 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1223 			    ifp->if_xname, IN_PRINT(ipbuf, &isaddr),
1224 			    la->lle_tbl->llt_ifp->if_xname);
1225 				goto out;
1226 		} else {
1227 			ARP_STATINC(ARP_STAT_RCVOVER);
1228 			if (log_movements)
1229 				log(LOG_INFO, "arp info overwritten "
1230 				    "for %s by %s\n",
1231 				    IN_PRINT(ipbuf, &isaddr),
1232 				    lla_snprintf(llabuf, ar_sha(ah),
1233 				    ah->ar_hln));
1234 		}
1235 	}
1236 
1237 	KASSERT(ifp->if_sadl->sdl_alen == ifp->if_addrlen);
1238 
1239 #if NTOKEN > 0
1240 	/*
1241 	 * XXX uses m_data and assumes the complete answer including
1242 	 * XXX token-ring headers is in the same buf
1243 	 */
1244 	if (ifp->if_type == IFT_ISO88025) {
1245 		struct token_header *trh;
1246 
1247 		trh = (struct token_header *)M_TRHSTART(m);
1248 		if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1249 			struct token_rif *rif;
1250 			size_t riflen;
1251 
1252 			rif = TOKEN_RIF(trh);
1253 			riflen = (ntohs(rif->tr_rcf) &
1254 			    TOKEN_RCF_LEN_MASK) >> 8;
1255 
1256 			if (riflen > 2 &&
1257 			    riflen < sizeof(struct token_rif) &&
1258 			    (riflen & 1) == 0) {
1259 				rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1260 				rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1261 				memcpy(TOKEN_RIF_LLE(la), rif, riflen);
1262 			}
1263 		}
1264 	}
1265 #endif
1266 
1267 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1268 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1269 	la->la_flags |= LLE_VALID;
1270 	if ((la->la_flags & LLE_STATIC) == 0) {
1271 		la->la_expire = time_uptime + arpt_keep;
1272 		arp_settimer(la, arpt_keep);
1273 	}
1274 	la->la_asked = 0;
1275 	/* rt->rt_flags &= ~RTF_REJECT; */
1276 
1277 	if (la->la_hold != NULL) {
1278 		int n = la->la_numheld;
1279 		struct mbuf *m_hold, *m_hold_next;
1280 		struct sockaddr_in sin;
1281 
1282 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1283 
1284 		m_hold = la->la_hold;
1285 		la->la_hold = NULL;
1286 		la->la_numheld = 0;
1287 		/*
1288 		 * We have to unlock here because if_output would call
1289 		 * arpresolve
1290 		 */
1291 		LLE_WUNLOCK(la);
1292 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1293 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1294 		for (; m_hold != NULL; m_hold = m_hold_next) {
1295 			m_hold_next = m_hold->m_nextpkt;
1296 			m_hold->m_nextpkt = NULL;
1297 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1298 		}
1299 	} else
1300 		LLE_WUNLOCK(la);
1301 	la = NULL;
1302 
1303 reply:
1304 	if (la != NULL) {
1305 		LLE_WUNLOCK(la);
1306 		la = NULL;
1307 	}
1308 	if (op != ARPOP_REQUEST) {
1309 		if (op == ARPOP_REPLY)
1310 			ARP_STATINC(ARP_STAT_RCVREPLY);
1311 		goto out;
1312 	}
1313 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1314 	if (in_hosteq(itaddr, myaddr)) {
1315 		/* If our address is unusable, don't reply */
1316 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1317 			goto out;
1318 		/* I am the target */
1319 		tha = ar_tha(ah);
1320 		if (tha)
1321 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1322 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1323 	} else {
1324 		/* Proxy ARP */
1325 		struct llentry *lle = NULL;
1326 		struct sockaddr_in sin;
1327 
1328 #if NCARP > 0
1329 		if (ifp->if_type == IFT_CARP) {
1330 			struct ifnet *_rcvif = m_get_rcvif(m, &s);
1331 			int iftype = 0;
1332 			if (__predict_true(_rcvif != NULL))
1333 				iftype = _rcvif->if_type;
1334 			m_put_rcvif(_rcvif, &s);
1335 			if (iftype != IFT_CARP)
1336 				goto out;
1337 		}
1338 #endif
1339 
1340 		tha = ar_tha(ah);
1341 
1342 		sockaddr_in_init(&sin, &itaddr, 0);
1343 
1344 		IF_AFDATA_RLOCK(ifp);
1345 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1346 		IF_AFDATA_RUNLOCK(ifp);
1347 
1348 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1349 			if (tha)
1350 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1351 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1352 			LLE_RUNLOCK(lle);
1353 		} else {
1354 			if (lle != NULL)
1355 				LLE_RUNLOCK(lle);
1356 			goto out;
1357 		}
1358 	}
1359 	ia4_release(ia, &psref_ia);
1360 
1361 	/*
1362 	 * XXX XXX: Here we're recycling the mbuf. But the mbuf could have
1363 	 * other mbufs in its chain, and just overwriting m->m_pkthdr.len
1364 	 * would be wrong in this case (the length becomes smaller than the
1365 	 * real chain size).
1366 	 *
1367 	 * This can theoretically cause bugs in the lower layers (drivers,
1368 	 * and L2encap), in some corner cases.
1369 	 */
1370 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1371 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1372 	ah->ar_op = htons(ARPOP_REPLY);
1373 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1374 	switch (ifp->if_type) {
1375 	case IFT_IEEE1394:
1376 		/* ieee1394 arp reply is broadcast */
1377 		m->m_flags &= ~M_MCAST;
1378 		m->m_flags |= M_BCAST;
1379 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1380 		break;
1381 	default:
1382 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1383 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1384 		break;
1385 	}
1386 	m->m_pkthdr.len = m->m_len;
1387 	sa.sa_family = AF_ARP;
1388 	sa.sa_len = 2;
1389 	arps = ARP_STAT_GETREF();
1390 	arps[ARP_STAT_SNDTOTAL]++;
1391 	arps[ARP_STAT_SNDREPLY]++;
1392 	ARP_STAT_PUTREF();
1393 	if_output_lock(ifp, ifp, m, &sa, NULL);
1394 	if (rcvif != NULL)
1395 		m_put_rcvif_psref(rcvif, &psref);
1396 	return;
1397 
1398 out:
1399 	if (la != NULL)
1400 		LLE_WUNLOCK(la);
1401 	if (ia != NULL)
1402 		ia4_release(ia, &psref_ia);
1403 	if (rcvif != NULL)
1404 		m_put_rcvif_psref(rcvif, &psref);
1405 	m_freem(m);
1406 }
1407 
1408 /*
1409  * Lookup or a new address in arptab.
1410  */
1411 static struct llentry *
1412 arplookup(struct ifnet *ifp, const struct in_addr *addr,
1413     const struct sockaddr *sa, int wlock)
1414 {
1415 	struct sockaddr_in sin;
1416 	struct llentry *la;
1417 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1418 
1419 	if (sa == NULL) {
1420 		KASSERT(addr != NULL);
1421 		sockaddr_in_init(&sin, addr, 0);
1422 		sa = sintocsa(&sin);
1423 	}
1424 
1425 	IF_AFDATA_RLOCK(ifp);
1426 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1427 	IF_AFDATA_RUNLOCK(ifp);
1428 
1429 	return la;
1430 }
1431 
1432 static struct llentry *
1433 arpcreate(struct ifnet *ifp, const struct in_addr *addr,
1434     const struct sockaddr *sa, int wlock)
1435 {
1436 	struct sockaddr_in sin;
1437 	struct llentry *la;
1438 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1439 
1440 	if (sa == NULL) {
1441 		KASSERT(addr != NULL);
1442 		sockaddr_in_init(&sin, addr, 0);
1443 		sa = sintocsa(&sin);
1444 	}
1445 
1446 	la = arplookup(ifp, addr, sa, wlock);
1447 
1448 	if (la == NULL) {
1449 		struct rtentry *rt;
1450 
1451 		rt = rtalloc1(sa, 0);
1452 		IF_AFDATA_WLOCK(ifp);
1453 		la = lla_create(LLTABLE(ifp), flags, sa, rt);
1454 		IF_AFDATA_WUNLOCK(ifp);
1455 		if (rt != NULL)
1456 			rt_unref(rt);
1457 
1458 		if (la != NULL)
1459 			arp_init_llentry(ifp, la);
1460 	}
1461 
1462 	return la;
1463 }
1464 
1465 int
1466 arpioctl(u_long cmd, void *data)
1467 {
1468 
1469 	return EOPNOTSUPP;
1470 }
1471 
1472 void
1473 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1474 {
1475 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1476 
1477 	ifa->ifa_rtrequest = arp_rtrequest;
1478 	ifa->ifa_flags |= RTF_CONNECTED;
1479 
1480 	/* ARP will handle DAD for this address. */
1481 	if (in_nullhost(IA_SIN(ifa)->sin_addr)) {
1482 		if (ia->ia_dad_stop != NULL)	/* safety */
1483 			ia->ia_dad_stop(ifa);
1484 		ia->ia_dad_start = NULL;
1485 		ia->ia_dad_stop = NULL;
1486 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1487 	} else {
1488 		ia->ia_dad_start = arp_dad_start;
1489 		ia->ia_dad_stop = arp_dad_stop;
1490 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE && ip_dad_count > 0)
1491 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1492 		else
1493 			arpannounce1(ifa);
1494 	}
1495 }
1496 
1497 TAILQ_HEAD(dadq_head, dadq);
1498 struct dadq {
1499 	TAILQ_ENTRY(dadq) dad_list;
1500 	struct ifaddr *dad_ifa;
1501 	int dad_count;		/* max ARP to send */
1502 	int dad_arp_tcount;	/* # of trials to send ARP */
1503 	int dad_arp_ocount;	/* ARP sent so far */
1504 	int dad_arp_announce;	/* max ARP announcements */
1505 	int dad_arp_acount;	/* # of announcements */
1506 	struct callout dad_timer_ch;
1507 };
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, dp);
1534 }
1535 
1536 static void
1537 arp_dad_stoptimer(struct dadq *dp)
1538 {
1539 
1540 	KASSERT(mutex_owned(&arp_dad_lock));
1541 
1542 	TAILQ_REMOVE(&dadq, dp, dad_list);
1543 	/* Tell the timer that dp is being destroyed. */
1544 	dp->dad_ifa = NULL;
1545 	callout_halt(&dp->dad_timer_ch, &arp_dad_lock);
1546 }
1547 
1548 static void
1549 arp_dad_destroytimer(struct dadq *dp)
1550 {
1551 
1552 	callout_destroy(&dp->dad_timer_ch);
1553 	KASSERT(dp->dad_ifa == NULL);
1554 	kmem_intr_free(dp, sizeof(*dp));
1555 }
1556 
1557 static void
1558 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1559 {
1560 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1561 	struct ifnet *ifp = ifa->ifa_ifp;
1562 	struct in_addr sip;
1563 
1564 	dp->dad_arp_tcount++;
1565 	if ((ifp->if_flags & IFF_UP) == 0)
1566 		return;
1567 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1568 		return;
1569 
1570 	dp->dad_arp_tcount = 0;
1571 	dp->dad_arp_ocount++;
1572 
1573 	memset(&sip, 0, sizeof(sip));
1574 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1575 	    CLLADDR(ifa->ifa_ifp->if_sadl));
1576 }
1577 
1578 /*
1579  * Start Duplicate Address Detection (DAD) for specified interface address.
1580  */
1581 static void
1582 arp_dad_start(struct ifaddr *ifa)
1583 {
1584 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1585 	struct dadq *dp;
1586 	char ipbuf[INET_ADDRSTRLEN];
1587 
1588 	if (!dad_init) {
1589 		TAILQ_INIT(&dadq);
1590 		mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1591 		dad_init++;
1592 	}
1593 
1594 	/*
1595 	 * If we don't need DAD, don't do it.
1596 	 * - DAD is disabled (ip_dad_count == 0)
1597 	 */
1598 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1599 		log(LOG_DEBUG,
1600 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1601 		    IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1602 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1603 		return;
1604 	}
1605 	if (!ip_dad_count) {
1606 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1607 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1608 		arpannounce1(ifa);
1609 		return;
1610 	}
1611 	KASSERT(ifa->ifa_ifp != NULL);
1612 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1613 		return;
1614 
1615 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1616 
1617 	mutex_enter(&arp_dad_lock);
1618 	if (arp_dad_find(ifa) != NULL) {
1619 		mutex_exit(&arp_dad_lock);
1620 		/* DAD already in progress */
1621 		if (dp != NULL)
1622 			kmem_intr_free(dp, sizeof(*dp));
1623 		return;
1624 	}
1625 
1626 	if (dp == NULL) {
1627 		mutex_exit(&arp_dad_lock);
1628 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1629 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1630 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1631 		return;
1632 	}
1633 
1634 	/*
1635 	 * Send ARP packet for DAD, ip_dad_count times.
1636 	 * Note that we must delay the first transmission.
1637 	 */
1638 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1639 	dp->dad_ifa = ifa;
1640 	ifaref(ifa);	/* just for safety */
1641 	dp->dad_count = ip_dad_count;
1642 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1643 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1644 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1645 
1646 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1647 	    ARPLOGADDR(&ia->ia_addr.sin_addr));
1648 
1649 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1650 
1651 	mutex_exit(&arp_dad_lock);
1652 }
1653 
1654 /*
1655  * terminate DAD unconditionally.  used for address removals.
1656  */
1657 static void
1658 arp_dad_stop(struct ifaddr *ifa)
1659 {
1660 	struct dadq *dp;
1661 
1662 	if (!dad_init)
1663 		return;
1664 
1665 	mutex_enter(&arp_dad_lock);
1666 	dp = arp_dad_find(ifa);
1667 	if (dp == NULL) {
1668 		mutex_exit(&arp_dad_lock);
1669 		/* DAD wasn't started yet */
1670 		return;
1671 	}
1672 
1673 	arp_dad_stoptimer(dp);
1674 
1675 	mutex_exit(&arp_dad_lock);
1676 
1677 	arp_dad_destroytimer(dp);
1678 	ifafree(ifa);
1679 }
1680 
1681 static void
1682 arp_dad_timer(struct dadq *dp)
1683 {
1684 	struct ifaddr *ifa;
1685 	struct in_ifaddr *ia;
1686 	char ipbuf[INET_ADDRSTRLEN];
1687 	bool need_free = false;
1688 
1689 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1690 	mutex_enter(&arp_dad_lock);
1691 
1692 	ifa = dp->dad_ifa;
1693 	if (ifa == NULL) {
1694 		/* dp is being destroyed by someone.  Do nothing. */
1695 		goto done;
1696 	}
1697 
1698 	ia = (struct in_ifaddr *)ifa;
1699 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1700 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1701 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1702 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1703 		goto done;
1704 	}
1705 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1706 	{
1707 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1708 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1709 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1710 		goto done;
1711 	}
1712 
1713 	/* timeouted with IFF_{RUNNING,UP} check */
1714 	if (dp->dad_arp_tcount > dad_maxtry) {
1715 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1716 		    if_name(ifa->ifa_ifp));
1717 
1718 		arp_dad_stoptimer(dp);
1719 		need_free = true;
1720 		goto done;
1721 	}
1722 
1723 	/* Need more checks? */
1724 	if (dp->dad_arp_ocount < dp->dad_count) {
1725 		int adelay;
1726 
1727 		/*
1728 		 * We have more ARP to go.  Send ARP packet for DAD.
1729 		 */
1730 		arp_dad_output(dp, ifa);
1731 		if (dp->dad_arp_ocount < dp->dad_count)
1732 			adelay = (PROBE_MIN * hz) +
1733 			    (cprng_fast32() %
1734 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1735 		else
1736 			adelay = ANNOUNCE_WAIT * hz;
1737 		arp_dad_starttimer(dp, adelay);
1738 		goto done;
1739 	} else if (dp->dad_arp_acount == 0) {
1740 		/*
1741 		 * We are done with DAD.
1742 		 * No duplicate address found.
1743 		 */
1744 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1745 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1746 		ARPLOG(LOG_DEBUG,
1747 		    "%s: DAD complete for %s - no duplicates found\n",
1748 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1749 		dp->dad_arp_announce = ANNOUNCE_NUM;
1750 		goto announce;
1751 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1752 announce:
1753 		/*
1754 		 * Announce the address.
1755 		 */
1756 		arpannounce1(ifa);
1757 		dp->dad_arp_acount++;
1758 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1759 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1760 			goto done;
1761 		}
1762 		ARPLOG(LOG_DEBUG,
1763 		    "%s: ARP announcement complete for %s\n",
1764 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1765 	}
1766 
1767 	arp_dad_stoptimer(dp);
1768 	need_free = true;
1769 done:
1770 	mutex_exit(&arp_dad_lock);
1771 
1772 	if (need_free) {
1773 		arp_dad_destroytimer(dp);
1774 		KASSERT(ifa != NULL);
1775 		ifafree(ifa);
1776 	}
1777 
1778 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1779 }
1780 
1781 static void
1782 arp_dad_duplicated(struct ifaddr *ifa, const char *sha)
1783 {
1784 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1785 	struct ifnet *ifp = ifa->ifa_ifp;
1786 	char ipbuf[INET_ADDRSTRLEN];
1787 	const char *iastr;
1788 
1789 	iastr = IN_PRINT(ipbuf, &ia->ia_addr.sin_addr);
1790 
1791 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1792 		log(LOG_ERR,
1793 		    "%s: DAD duplicate address %s from %s\n",
1794 		    if_name(ifp), iastr, sha);
1795 	} else if (ia->ia_dad_defended == 0 ||
1796 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1797 		ia->ia_dad_defended = time_uptime;
1798 		arpannounce1(ifa);
1799 		log(LOG_ERR,
1800 		    "%s: DAD defended address %s from %s\n",
1801 		    if_name(ifp), iastr, sha);
1802 		return;
1803 	} else {
1804 		/* If DAD is disabled, just report the duplicate. */
1805 		if (ip_dad_count == 0) {
1806 			log(LOG_ERR,
1807 			    "%s: DAD ignoring duplicate address %s from %s\n",
1808 			    if_name(ifp), iastr, sha);
1809 			return;
1810 		}
1811 		log(LOG_ERR,
1812 		    "%s: DAD defence failed for %s from %s\n",
1813 		    if_name(ifp), iastr, sha);
1814 	}
1815 
1816 	arp_dad_stop(ifa);
1817 
1818 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1819 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1820 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1821 		/* Inform the routing socket of the duplicate address */
1822 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1823 	}
1824 }
1825 
1826 /*
1827  * Called from 10 Mb/s Ethernet interrupt handlers
1828  * when ether packet type ETHERTYPE_REVARP
1829  * is received.  Common length and type checks are done here,
1830  * then the protocol-specific routine is called.
1831  */
1832 void
1833 revarpinput(struct mbuf *m)
1834 {
1835 	struct arphdr *ar;
1836 	int arplen;
1837 
1838 	arplen = sizeof(struct arphdr);
1839 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1840 		return;
1841 	ar = mtod(m, struct arphdr *);
1842 
1843 	if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394) {
1844 		goto out;
1845 	}
1846 
1847 	arplen = sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln);
1848 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1849 		return;
1850 	ar = mtod(m, struct arphdr *);
1851 
1852 	switch (ntohs(ar->ar_pro)) {
1853 	case ETHERTYPE_IP:
1854 	case ETHERTYPE_IPTRAILERS:
1855 		in_revarpinput(m);
1856 		return;
1857 
1858 	default:
1859 		break;
1860 	}
1861 
1862 out:
1863 	m_freem(m);
1864 }
1865 
1866 /*
1867  * RARP for Internet protocols on 10 Mb/s Ethernet.
1868  * Algorithm is that given in RFC 903.
1869  * We are only using for bootstrap purposes to get an ip address for one of
1870  * our interfaces.  Thus we support no user-interface.
1871  *
1872  * Since the contents of the RARP reply are specific to the interface that
1873  * sent the request, this code must ensure that they are properly associated.
1874  *
1875  * Note: also supports ARP via RARP packets, per the RFC.
1876  */
1877 void
1878 in_revarpinput(struct mbuf *m)
1879 {
1880 	struct arphdr *ah;
1881 	void *tha;
1882 	int op;
1883 	struct ifnet *rcvif;
1884 	int s;
1885 
1886 	ah = mtod(m, struct arphdr *);
1887 	op = ntohs(ah->ar_op);
1888 
1889 	rcvif = m_get_rcvif(m, &s);
1890 	if (__predict_false(rcvif == NULL))
1891 		goto out;
1892 	if (rcvif->if_flags & IFF_NOARP)
1893 		goto out;
1894 
1895 	switch (rcvif->if_type) {
1896 	case IFT_IEEE1394:
1897 		/* ARP without target hardware address is not supported */
1898 		goto out;
1899 	default:
1900 		break;
1901 	}
1902 
1903 	switch (op) {
1904 	case ARPOP_REQUEST:
1905 	case ARPOP_REPLY:	/* per RFC */
1906 		m_put_rcvif(rcvif, &s);
1907 		in_arpinput(m);
1908 		return;
1909 	case ARPOP_REVREPLY:
1910 		break;
1911 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1912 	default:
1913 		goto out;
1914 	}
1915 	if (!revarp_in_progress)
1916 		goto out;
1917 	if (rcvif != myip_ifp) /* !same interface */
1918 		goto out;
1919 	if (myip_initialized)
1920 		goto wake;
1921 	tha = ar_tha(ah);
1922 	if (tha == NULL)
1923 		goto out;
1924 	if (ah->ar_pln != sizeof(struct in_addr))
1925 		goto out;
1926 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1927 		goto out;
1928 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1929 		goto out;
1930 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1931 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1932 	myip_initialized = 1;
1933 wake:	/* Do wakeup every time in case it was missed. */
1934 	wakeup((void *)&myip);
1935 
1936 out:
1937 	m_put_rcvif(rcvif, &s);
1938 	m_freem(m);
1939 }
1940 
1941 /*
1942  * Send a RARP request for the ip address of the specified interface.
1943  * The request should be RFC 903-compliant.
1944  */
1945 static void
1946 revarprequest(struct ifnet *ifp)
1947 {
1948 	struct sockaddr sa;
1949 	struct mbuf *m;
1950 	struct arphdr *ah;
1951 	void *tha;
1952 
1953 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1954 		return;
1955 	MCLAIM(m, &arpdomain.dom_mowner);
1956 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1957 	    2*ifp->if_addrlen;
1958 	m->m_pkthdr.len = m->m_len;
1959 	MH_ALIGN(m, m->m_len);
1960 	ah = mtod(m, struct arphdr *);
1961 	memset(ah, 0, m->m_len);
1962 	ah->ar_pro = htons(ETHERTYPE_IP);
1963 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1964 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1965 	ah->ar_op = htons(ARPOP_REVREQUEST);
1966 
1967 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1968 	tha = ar_tha(ah);
1969 	if (tha == NULL) {
1970 		m_free(m);
1971 		return;
1972 	}
1973 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1974 
1975 	sa.sa_family = AF_ARP;
1976 	sa.sa_len = 2;
1977 	m->m_flags |= M_BCAST;
1978 
1979 	if_output_lock(ifp, ifp, m, &sa, NULL);
1980 }
1981 
1982 /*
1983  * RARP for the ip address of the specified interface, but also
1984  * save the ip address of the server that sent the answer.
1985  * Timeout if no response is received.
1986  */
1987 int
1988 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1989     struct in_addr *clnt_in)
1990 {
1991 	int result, count = 20;
1992 
1993 	myip_initialized = 0;
1994 	myip_ifp = ifp;
1995 
1996 	revarp_in_progress = 1;
1997 	while (count--) {
1998 		revarprequest(ifp);
1999 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
2000 		if (result != EWOULDBLOCK)
2001 			break;
2002 	}
2003 	revarp_in_progress = 0;
2004 
2005 	if (!myip_initialized)
2006 		return ENETUNREACH;
2007 
2008 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
2009 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
2010 	return 0;
2011 }
2012 
2013 void
2014 arp_stat_add(int type, uint64_t count)
2015 {
2016 	ARP_STATADD(type, count);
2017 }
2018 
2019 static int
2020 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
2021 {
2022 
2023 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2024 }
2025 
2026 static void
2027 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2028 {
2029 	const struct sysctlnode *node;
2030 
2031 	sysctl_createv(clog, 0, NULL, NULL,
2032 			CTLFLAG_PERMANENT,
2033 			CTLTYPE_NODE, "inet", NULL,
2034 			NULL, 0, NULL, 0,
2035 			CTL_NET, PF_INET, CTL_EOL);
2036 	sysctl_createv(clog, 0, NULL, &node,
2037 			CTLFLAG_PERMANENT,
2038 			CTLTYPE_NODE, "arp",
2039 			SYSCTL_DESCR("Address Resolution Protocol"),
2040 			NULL, 0, NULL, 0,
2041 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2042 
2043 	sysctl_createv(clog, 0, NULL, NULL,
2044 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2045 			CTLTYPE_INT, "keep",
2046 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
2047 			NULL, 0, &arpt_keep, 0,
2048 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2049 
2050 	sysctl_createv(clog, 0, NULL, NULL,
2051 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2052 			CTLTYPE_INT, "down",
2053 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
2054 			NULL, 0, &arpt_down, 0,
2055 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2056 
2057 	sysctl_createv(clog, 0, NULL, NULL,
2058 			CTLFLAG_PERMANENT,
2059 			CTLTYPE_STRUCT, "stats",
2060 			SYSCTL_DESCR("ARP statistics"),
2061 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2062 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2063 
2064 	sysctl_createv(clog, 0, NULL, NULL,
2065 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2066 			CTLTYPE_INT, "log_movements",
2067 			SYSCTL_DESCR("log ARP replies from MACs different than"
2068 			    " the one in the cache"),
2069 			NULL, 0, &log_movements, 0,
2070 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2071 
2072 	sysctl_createv(clog, 0, NULL, NULL,
2073 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2074 			CTLTYPE_INT, "log_permanent_modify",
2075 			SYSCTL_DESCR("log ARP replies from MACs different than"
2076 			    " the one in the permanent arp entry"),
2077 			NULL, 0, &log_permanent_modify, 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, "log_wrong_iface",
2083 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2084 			    " interface"),
2085 			NULL, 0, &log_wrong_iface, 0,
2086 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2087 
2088 	sysctl_createv(clog, 0, NULL, NULL,
2089 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2090 			CTLTYPE_INT, "log_unknown_network",
2091 			SYSCTL_DESCR("log ARP packets from non-local network"),
2092 			NULL, 0, &log_unknown_network, 0,
2093 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2094 
2095 	sysctl_createv(clog, 0, NULL, NULL,
2096 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2097 			CTLTYPE_INT, "debug",
2098 			SYSCTL_DESCR("Enable ARP DAD debug output"),
2099 			NULL, 0, &arp_debug, 0,
2100 			CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2101 }
2102 
2103 #endif /* INET */
2104