xref: /netbsd-src/sys/netinet/if_arp.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: if_arp.c,v 1.255 2017/11/17 07:37:12 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 1.255 2017/11/17 07:37:12 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/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 u_int8_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 *, struct mbuf *,
167 	    const struct in_addr *, const struct sockaddr *, int);
168 static	struct llentry *arpcreate(struct ifnet *, struct mbuf *,
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 static void arp_dad_timer(struct ifaddr *);
177 static void arp_dad_start(struct ifaddr *);
178 static void arp_dad_stop(struct ifaddr *);
179 static void arp_dad_duplicated(struct ifaddr *, const char *);
180 
181 static void arp_init_llentry(struct ifnet *, struct llentry *);
182 #if NTOKEN > 0
183 static void arp_free_llentry_tokenring(struct llentry *);
184 #endif
185 
186 struct	ifqueue arpintrq = {
187 	.ifq_head = NULL,
188 	.ifq_tail = NULL,
189 	.ifq_len = 0,
190 	.ifq_maxlen = 50,
191 	.ifq_drops = 0,
192 };
193 static int	arp_maxtries = 5;
194 static int	useloopback = 1;	/* use loopback interface for local traffic */
195 
196 static percpu_t *arpstat_percpu;
197 
198 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
199 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
200 
201 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
202 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
203 
204 /* revarp state */
205 static struct	in_addr myip, srv_ip;
206 static int	myip_initialized = 0;
207 static int	revarp_in_progress = 0;
208 static struct	ifnet *myip_ifp = NULL;
209 
210 static int arp_drainwanted;
211 
212 static int log_movements = 1;
213 static int log_permanent_modify = 1;
214 static int log_wrong_iface = 1;
215 static int log_unknown_network = 1;
216 
217 /*
218  * this should be elsewhere.
219  */
220 
221 #define	LLA_ADDRSTRLEN	(16 * 3)
222 
223 static char *
224 lla_snprintf(char *, u_int8_t *, int);
225 
226 static char *
227 lla_snprintf(char *dst, u_int8_t *adrp, int len)
228 {
229 	int i;
230 	char *p;
231 
232 	p = dst;
233 
234 	*p++ = hexdigits[(*adrp) >> 4];
235 	*p++ = hexdigits[(*adrp++) & 0xf];
236 
237 	for (i = 1; i < len && i < 16; i++) {
238 		*p++ = ':';
239 		*p++ = hexdigits[(*adrp) >> 4];
240 		*p++ = hexdigits[(*adrp++) & 0xf];
241 	}
242 
243 	*p = 0;
244 	return dst;
245 }
246 
247 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
248 
249 static void
250 arp_fasttimo(void)
251 {
252 	if (arp_drainwanted) {
253 		arp_drain();
254 		arp_drainwanted = 0;
255 	}
256 }
257 
258 const struct protosw arpsw[] = {
259 	{ .pr_type = 0,
260 	  .pr_domain = &arpdomain,
261 	  .pr_protocol = 0,
262 	  .pr_flags = 0,
263 	  .pr_input = 0,
264 	  .pr_ctlinput = 0,
265 	  .pr_ctloutput = 0,
266 	  .pr_usrreqs = 0,
267 	  .pr_init = arp_init,
268 	  .pr_fasttimo = arp_fasttimo,
269 	  .pr_slowtimo = 0,
270 	  .pr_drain = arp_drainstub,
271 	}
272 };
273 
274 struct domain arpdomain = {
275 	.dom_family = PF_ARP,
276 	.dom_name = "arp",
277 	.dom_protosw = arpsw,
278 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
279 };
280 
281 static void sysctl_net_inet_arp_setup(struct sysctllog **);
282 
283 void
284 arp_init(void)
285 {
286 
287 	sysctl_net_inet_arp_setup(NULL);
288 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
289 	IFQ_LOCK_INIT(&arpintrq);
290 }
291 
292 static void
293 arp_drainstub(void)
294 {
295 	arp_drainwanted = 1;
296 }
297 
298 /*
299  * ARP protocol drain routine.  Called when memory is in short supply.
300  * Called at splvm();  don't acquire softnet_lock as can be called from
301  * hardware interrupt handlers.
302  */
303 void
304 arp_drain(void)
305 {
306 
307 	lltable_drain(AF_INET);
308 }
309 
310 static void
311 arptimer(void *arg)
312 {
313 	struct llentry *lle = arg;
314 	struct ifnet *ifp;
315 
316 	if (lle == NULL)
317 		return;
318 
319 	if (lle->la_flags & LLE_STATIC)
320 		return;
321 
322 	LLE_WLOCK(lle);
323 	if (callout_pending(&lle->la_timer)) {
324 		/*
325 		 * Here we are a bit odd here in the treatment of
326 		 * active/pending. If the pending bit is set, it got
327 		 * rescheduled before I ran. The active
328 		 * bit we ignore, since if it was stopped
329 		 * in ll_tablefree() and was currently running
330 		 * it would have return 0 so the code would
331 		 * not have deleted it since the callout could
332 		 * not be stopped so we want to go through
333 		 * with the delete here now. If the callout
334 		 * was restarted, the pending bit will be back on and
335 		 * we just want to bail since the callout_reset would
336 		 * return 1 and our reference would have been removed
337 		 * by arpresolve() below.
338 		 */
339 		LLE_WUNLOCK(lle);
340 		return;
341 	}
342 	ifp = lle->lle_tbl->llt_ifp;
343 
344 	callout_stop(&lle->la_timer);
345 
346 	/* XXX: LOR avoidance. We still have ref on lle. */
347 	LLE_WUNLOCK(lle);
348 
349 	IF_AFDATA_LOCK(ifp);
350 	LLE_WLOCK(lle);
351 
352 	/* Guard against race with other llentry_free(). */
353 	if (lle->la_flags & LLE_LINKED) {
354 		size_t pkts_dropped;
355 
356 		LLE_REMREF(lle);
357 		pkts_dropped = llentry_free(lle);
358 		ARP_STATADD(ARP_STAT_DFRDROPPED, pkts_dropped);
359 		ARP_STATADD(ARP_STAT_DFRTOTAL, pkts_dropped);
360 	} else {
361 		LLE_FREE_LOCKED(lle);
362 	}
363 
364 	IF_AFDATA_UNLOCK(ifp);
365 }
366 
367 static void
368 arp_settimer(struct llentry *la, int sec)
369 {
370 
371 	LLE_WLOCK_ASSERT(la);
372 	LLE_ADDREF(la);
373 	callout_reset(&la->la_timer, hz * sec, arptimer, la);
374 }
375 
376 /*
377  * We set the gateway for RTF_CLONING routes to a "prototype"
378  * link-layer sockaddr whose interface type (if_type) and interface
379  * index (if_index) fields are prepared.
380  */
381 static struct sockaddr *
382 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
383     const struct sockaddr *netmask)
384 {
385 	const struct ifnet *ifp = rt->rt_ifp;
386 	uint8_t namelen = strlen(ifp->if_xname);
387 	uint8_t addrlen = ifp->if_addrlen;
388 
389 	/*
390 	 * XXX: If this is a manually added route to interface
391 	 * such as older version of routed or gated might provide,
392 	 * restore cloning bit.
393 	 */
394 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
395 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
396 		rt->rt_flags |= RTF_CONNECTED;
397 
398 	if ((rt->rt_flags & (RTF_CONNECTED | RTF_LOCAL))) {
399 		union {
400 			struct sockaddr sa;
401 			struct sockaddr_storage ss;
402 			struct sockaddr_dl sdl;
403 		} u;
404 		/*
405 		 * Case 1: This route should come from a route to iface.
406 		 */
407 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
408 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
409 		rt_setgate(rt, &u.sa);
410 		gate = rt->rt_gateway;
411 	}
412 	return gate;
413 }
414 
415 static void
416 arp_init_llentry(struct ifnet *ifp, struct llentry *lle)
417 {
418 
419 	switch (ifp->if_type) {
420 #if NTOKEN > 0
421 	case IFT_ISO88025:
422 		lle->la_opaque = kmem_intr_alloc(sizeof(struct token_rif),
423 		    KM_NOSLEEP);
424 		lle->lle_ll_free = arp_free_llentry_tokenring;
425 		break;
426 #endif
427 	}
428 }
429 
430 #if NTOKEN > 0
431 static void
432 arp_free_llentry_tokenring(struct llentry *lle)
433 {
434 
435 	kmem_intr_free(lle->la_opaque, sizeof(struct token_rif));
436 }
437 #endif
438 
439 /*
440  * Parallel to llc_rtrequest.
441  */
442 void
443 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
444 {
445 	struct sockaddr *gate = rt->rt_gateway;
446 	struct in_ifaddr *ia;
447 	struct ifaddr *ifa;
448 	struct ifnet *ifp = rt->rt_ifp;
449 	int bound;
450 	int s;
451 
452 	if (req == RTM_LLINFO_UPD) {
453 		if ((ifa = info->rti_ifa) != NULL)
454 			arpannounce1(ifa);
455 		return;
456 	}
457 
458 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
459 		if (req != RTM_ADD)
460 			return;
461 
462 		/*
463 		 * linklayers with particular link MTU limitation.
464 		 */
465 		switch(ifp->if_type) {
466 #if NFDDI > 0
467 		case IFT_FDDI:
468 			if (ifp->if_mtu > FDDIIPMTU)
469 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
470 			break;
471 #endif
472 #if NARCNET > 0
473 		case IFT_ARCNET:
474 		    {
475 			int arcipifmtu;
476 
477 			if (ifp->if_flags & IFF_LINK0)
478 				arcipifmtu = arc_ipmtu;
479 			else
480 				arcipifmtu = ARCMTU;
481 			if (ifp->if_mtu > arcipifmtu)
482 				rt->rt_rmx.rmx_mtu = arcipifmtu;
483 			break;
484 		    }
485 #endif
486 		}
487 		return;
488 	}
489 
490 	switch (req) {
491 	case RTM_SETGATE:
492 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
493 		break;
494 	case RTM_ADD:
495 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
496 		if (gate == NULL) {
497 			log(LOG_ERR, "%s: arp_setgate failed\n", __func__);
498 			break;
499 		}
500 		if ((rt->rt_flags & RTF_CONNECTED) ||
501 		    (rt->rt_flags & RTF_LOCAL)) {
502 			/*
503 			 * Give this route an expiration time, even though
504 			 * it's a "permanent" route, so that routes cloned
505 			 * from it do not need their expiration time set.
506 			 */
507 			KASSERT(time_uptime != 0);
508 			rt->rt_expire = time_uptime;
509 			/*
510 			 * linklayers with particular link MTU limitation.
511 			 */
512 			switch (ifp->if_type) {
513 #if NFDDI > 0
514 			case IFT_FDDI:
515 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
516 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
517 				     (rt->rt_rmx.rmx_mtu == 0 &&
518 				      ifp->if_mtu > FDDIIPMTU)))
519 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
520 				break;
521 #endif
522 #if NARCNET > 0
523 			case IFT_ARCNET:
524 			    {
525 				int arcipifmtu;
526 				if (ifp->if_flags & IFF_LINK0)
527 					arcipifmtu = arc_ipmtu;
528 				else
529 					arcipifmtu = ARCMTU;
530 
531 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
532 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
533 				     (rt->rt_rmx.rmx_mtu == 0 &&
534 				      ifp->if_mtu > arcipifmtu)))
535 					rt->rt_rmx.rmx_mtu = arcipifmtu;
536 				break;
537 			    }
538 #endif
539 			}
540 			if (rt->rt_flags & RTF_CONNECTED)
541 				break;
542 		}
543 
544 		bound = curlwp_bind();
545 		/* Announce a new entry if requested. */
546 		if (rt->rt_flags & RTF_ANNOUNCE) {
547 			struct psref psref;
548 			ia = in_get_ia_on_iface_psref(
549 			    satocsin(rt_getkey(rt))->sin_addr, ifp, &psref);
550 			if (ia != NULL) {
551 				arpannounce(ifp, &ia->ia_ifa,
552 				    CLLADDR(satocsdl(gate)));
553 				ia4_release(ia, &psref);
554 			}
555 		}
556 
557 		if (gate->sa_family != AF_LINK ||
558 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
559 			log(LOG_DEBUG, "%s: bad gateway value\n", __func__);
560 			goto out;
561 		}
562 
563 		satosdl(gate)->sdl_type = ifp->if_type;
564 		satosdl(gate)->sdl_index = ifp->if_index;
565 
566 		/* If the route is for a broadcast address mark it as such.
567 		 * This way we can avoid an expensive call to in_broadcast()
568 		 * in ip_output() most of the time (because the route passed
569 		 * to ip_output() is almost always a host route). */
570 		if (rt->rt_flags & RTF_HOST &&
571 		    !(rt->rt_flags & RTF_BROADCAST) &&
572 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
573 			rt->rt_flags |= RTF_BROADCAST;
574 		/* There is little point in resolving the broadcast address */
575 		if (rt->rt_flags & RTF_BROADCAST)
576 			goto out;
577 
578 		/*
579 		 * When called from rt_ifa_addlocal, we cannot depend on that
580 		 * the address (rt_getkey(rt)) exits in the address list of the
581 		 * interface. So check RTF_LOCAL instead.
582 		 */
583 		if (rt->rt_flags & RTF_LOCAL) {
584 			rt->rt_expire = 0;
585 			if (useloopback) {
586 				rt->rt_ifp = lo0ifp;
587 				rt->rt_rmx.rmx_mtu = 0;
588 			}
589 			goto out;
590 		}
591 
592 		s = pserialize_read_enter();
593 		ia = in_get_ia_on_iface(satocsin(rt_getkey(rt))->sin_addr, ifp);
594 		if (ia == NULL) {
595 			pserialize_read_exit(s);
596 			goto out;
597 		}
598 
599 		rt->rt_expire = 0;
600 		if (useloopback) {
601 			rt->rt_ifp = lo0ifp;
602 			rt->rt_rmx.rmx_mtu = 0;
603 		}
604 		rt->rt_flags |= RTF_LOCAL;
605 		/*
606 		 * make sure to set rt->rt_ifa to the interface
607 		 * address we are using, otherwise we will have trouble
608 		 * with source address selection.
609 		 */
610 		ifa = &ia->ia_ifa;
611 		if (ifa != rt->rt_ifa)
612 			/* Assume it doesn't sleep */
613 			rt_replace_ifa(rt, ifa);
614 		pserialize_read_exit(s);
615 	out:
616 		curlwp_bindx(bound);
617 		break;
618 	}
619 }
620 
621 /*
622  * Broadcast an ARP request. Caller specifies:
623  *	- arp header source ip address
624  *	- arp header target ip address
625  *	- arp header source ethernet address
626  */
627 static void
628 arprequest(struct ifnet *ifp,
629     const struct in_addr *sip, const struct in_addr *tip,
630     const u_int8_t *enaddr)
631 {
632 	struct mbuf *m;
633 	struct arphdr *ah;
634 	struct sockaddr sa;
635 	uint64_t *arps;
636 
637 	KASSERT(sip != NULL);
638 	KASSERT(tip != NULL);
639 	KASSERT(enaddr != NULL);
640 
641 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
642 		return;
643 	MCLAIM(m, &arpdomain.dom_mowner);
644 	switch (ifp->if_type) {
645 	case IFT_IEEE1394:
646 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
647 		    ifp->if_addrlen;
648 		break;
649 	default:
650 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
651 		    2 * ifp->if_addrlen;
652 		break;
653 	}
654 	m->m_pkthdr.len = m->m_len;
655 	MH_ALIGN(m, m->m_len);
656 	ah = mtod(m, struct arphdr *);
657 	memset(ah, 0, m->m_len);
658 	switch (ifp->if_type) {
659 	case IFT_IEEE1394:	/* RFC2734 */
660 		/* fill it now for ar_tpa computation */
661 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
662 		break;
663 	default:
664 		/* ifp->if_output will fill ar_hrd */
665 		break;
666 	}
667 	ah->ar_pro = htons(ETHERTYPE_IP);
668 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
669 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
670 	ah->ar_op = htons(ARPOP_REQUEST);
671 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
672 	memcpy(ar_spa(ah), sip, ah->ar_pln);
673 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
674 	sa.sa_family = AF_ARP;
675 	sa.sa_len = 2;
676 	m->m_flags |= M_BCAST;
677 	arps = ARP_STAT_GETREF();
678 	arps[ARP_STAT_SNDTOTAL]++;
679 	arps[ARP_STAT_SENDREQUEST]++;
680 	ARP_STAT_PUTREF();
681 	if_output_lock(ifp, ifp, m, &sa, NULL);
682 }
683 
684 void
685 arpannounce(struct ifnet *ifp, struct ifaddr *ifa, const uint8_t *enaddr)
686 {
687 	struct in_ifaddr *ia = ifatoia(ifa);
688 	struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
689 
690 	if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) {
691 		ARPLOG(LOG_DEBUG, "%s not ready\n", ARPLOGADDR(ip));
692 		return;
693 	}
694 	arprequest(ifp, ip, ip, enaddr);
695 }
696 
697 static void
698 arpannounce1(struct ifaddr *ifa)
699 {
700 
701 	arpannounce(ifa->ifa_ifp, ifa, CLLADDR(ifa->ifa_ifp->if_sadl));
702 }
703 
704 /*
705  * Resolve an IP address into an ethernet address.  If success,
706  * desten is filled in.  If there is no entry in arptab,
707  * set one up and broadcast a request for the IP address.
708  * Hold onto this mbuf and resend it once the address
709  * is finally resolved.  A return value of 0 indicates
710  * that desten has been filled in and the packet should be sent
711  * normally; a return value of EWOULDBLOCK indicates that the packet has been
712  * held pending resolution.
713  * Any other value indicates an error.
714  */
715 int
716 arpresolve(struct ifnet *ifp, const struct rtentry *rt, struct mbuf *m,
717     const struct sockaddr *dst, void *desten, size_t destlen)
718 {
719 	struct llentry *la;
720 	const char *create_lookup;
721 	bool renew;
722 	int error;
723 	struct ifnet *origifp = ifp;
724 #if NCARP > 0
725 	if (rt != NULL && rt->rt_ifp->if_type == IFT_CARP)
726 		ifp = rt->rt_ifp;
727 #endif
728 
729 	KASSERT(m != NULL);
730 
731 	la = arplookup(ifp, m, NULL, dst, 0);
732 	if (la == NULL)
733 		goto notfound;
734 
735 	if ((la->la_flags & LLE_VALID) &&
736 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
737 		KASSERT(destlen >= ifp->if_addrlen);
738 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
739 		LLE_RUNLOCK(la);
740 		return 0;
741 	}
742 
743 notfound:
744 #ifdef IFF_STATICARP /* FreeBSD */
745 #define _IFF_NOARP (IFF_NOARP | IFF_STATICARP)
746 #else
747 #define _IFF_NOARP IFF_NOARP
748 #endif
749 	if (ifp->if_flags & _IFF_NOARP) {
750 		if (la != NULL)
751 			LLE_RUNLOCK(la);
752 		error = ENOTSUP;
753 		goto bad;
754 	}
755 #undef _IFF_NOARP
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 u_int8_t *enaddr =
814 			    CLLADDR(ifp->if_sadl);
815 			arprequest(origifp,
816 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
817 			    &satocsin(dst)->sin_addr, enaddr);
818 		}
819 
820 		return 0;
821 	}
822 
823 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
824 		LLE_RUNLOCK(la);
825 		log(LOG_DEBUG, "%s: ouch, empty static llinfo for %s\n",
826 		    __func__, inet_ntoa(satocsin(dst)->sin_addr));
827 		error = EINVAL;
828 		goto bad;
829 	}
830 
831 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
832 
833 	/*
834 	 * There is an arptab entry, but no ethernet address
835 	 * response yet.  Add the mbuf to the list, dropping
836 	 * the oldest packet if we have exceeded the system
837 	 * setting.
838 	 */
839 	LLE_WLOCK_ASSERT(la);
840 	if (la->la_numheld >= arp_maxhold) {
841 		if (la->la_hold != NULL) {
842 			struct mbuf *next = la->la_hold->m_nextpkt;
843 			m_freem(la->la_hold);
844 			la->la_hold = next;
845 			la->la_numheld--;
846 			ARP_STATINC(ARP_STAT_DFRDROPPED);
847 			ARP_STATINC(ARP_STAT_DFRTOTAL);
848 		}
849 	}
850 	if (la->la_hold != NULL) {
851 		struct mbuf *curr = la->la_hold;
852 		while (curr->m_nextpkt != NULL)
853 			curr = curr->m_nextpkt;
854 		curr->m_nextpkt = m;
855 	} else
856 		la->la_hold = m;
857 	la->la_numheld++;
858 	if (!renew)
859 		LLE_DOWNGRADE(la);
860 
861 	/*
862 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
863 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
864 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
865 	 * ARP request, but not faster than one request per second.
866 	 */
867 	if (la->la_asked < arp_maxtries)
868 		error = EWOULDBLOCK;	/* First request. */
869 	else
870 		error = (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
871 		    EHOSTUNREACH : EHOSTDOWN;
872 
873 	if (renew) {
874 		const u_int8_t *enaddr =
875 		    CLLADDR(ifp->if_sadl);
876 		la->la_expire = time_uptime;
877 		arp_settimer(la, arpt_down);
878 		la->la_asked++;
879 		LLE_WUNLOCK(la);
880 
881 		if (rt != NULL) {
882 			arprequest(origifp,
883 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
884 			    &satocsin(dst)->sin_addr, enaddr);
885 		} else {
886 			struct sockaddr_in sin;
887 			struct rtentry *_rt;
888 
889 			sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
890 
891 			/* XXX */
892 			_rt = rtalloc1((struct sockaddr *)&sin, 0);
893 			if (_rt == NULL)
894 				goto bad;
895 			arprequest(origifp,
896 			    &satocsin(_rt->rt_ifa->ifa_addr)->sin_addr,
897 			    &satocsin(dst)->sin_addr, enaddr);
898 			rt_unref(_rt);
899 		}
900 		return error;
901 	}
902 
903 	LLE_RUNLOCK(la);
904 	return error;
905 
906 bad:
907 	m_freem(m);
908 	return error;
909 }
910 
911 /*
912  * Common length and type checks are done here,
913  * then the protocol-specific routine is called.
914  */
915 void
916 arpintr(void)
917 {
918 	struct mbuf *m;
919 	struct arphdr *ar;
920 	int s;
921 	int arplen;
922 
923 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
924 	for (;;) {
925 		struct ifnet *rcvif;
926 
927 		IFQ_LOCK(&arpintrq);
928 		IF_DEQUEUE(&arpintrq, m);
929 		IFQ_UNLOCK(&arpintrq);
930 		if (m == NULL)
931 			goto out;
932 		if ((m->m_flags & M_PKTHDR) == 0)
933 			panic("arpintr");
934 
935 		MCLAIM(m, &arpdomain.dom_mowner);
936 		ARP_STATINC(ARP_STAT_RCVTOTAL);
937 
938 		/*
939 		 * First, make sure we have at least struct arphdr.
940 		 */
941 		if (m->m_len < sizeof(struct arphdr) ||
942 		    (ar = mtod(m, struct arphdr *)) == NULL)
943 			goto badlen;
944 
945 		rcvif = m_get_rcvif(m, &s);
946 		if (__predict_false(rcvif == NULL)) {
947 			ARP_STATINC(ARP_STAT_RCVNOINT);
948 			goto free;
949 		}
950 		switch (rcvif->if_type) {
951 		case IFT_IEEE1394:
952 			arplen = sizeof(struct arphdr) +
953 			    ar->ar_hln + 2 * ar->ar_pln;
954 			break;
955 		default:
956 			arplen = sizeof(struct arphdr) +
957 			    2 * ar->ar_hln + 2 * ar->ar_pln;
958 			break;
959 		}
960 		m_put_rcvif(rcvif, &s);
961 
962 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
963 		    m->m_len >= arplen)
964 			switch (ntohs(ar->ar_pro)) {
965 			case ETHERTYPE_IP:
966 			case ETHERTYPE_IPTRAILERS:
967 				in_arpinput(m);
968 				continue;
969 			default:
970 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
971 			}
972 		else {
973 badlen:
974 			ARP_STATINC(ARP_STAT_RCVBADLEN);
975 		}
976 free:
977 		m_freem(m);
978 	}
979 out:
980 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
981 	return; /* XXX gcc */
982 }
983 
984 /*
985  * ARP for Internet protocols on 10 Mb/s Ethernet. Algorithm is that given in
986  * RFC 826. In addition, a sanity check is performed on the sender protocol
987  * address, to catch impersonators.
988  *
989  * We no longer handle negotiations for use of trailer protocol: formerly, ARP
990  * replied for protocol type ETHERTYPE_TRAIL sent along with IP replies if we
991  * wanted trailers sent to us, and also sent them in response to IP replies.
992  * This allowed either end to announce the desire to receive trailer packets.
993  *
994  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, but
995  * formerly didn't normally send requests.
996  */
997 static void
998 in_arpinput(struct mbuf *m)
999 {
1000 	struct arphdr *ah;
1001 	struct ifnet *ifp, *rcvif = NULL;
1002 	struct llentry *la = NULL;
1003 	struct in_ifaddr *ia = NULL;
1004 #if NBRIDGE > 0
1005 	struct in_ifaddr *bridge_ia = NULL;
1006 #endif
1007 #if NCARP > 0
1008 	u_int32_t count = 0, index = 0;
1009 #endif
1010 	struct sockaddr sa;
1011 	struct in_addr isaddr, itaddr, myaddr;
1012 	int op;
1013 	void *tha;
1014 	uint64_t *arps;
1015 	struct psref psref, psref_ia;
1016 	int s;
1017 	char llabuf[LLA_ADDRSTRLEN];
1018 	char ipbuf[INET_ADDRSTRLEN];
1019 	bool do_dad;
1020 
1021 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
1022 		goto out;
1023 	ah = mtod(m, struct arphdr *);
1024 	op = ntohs(ah->ar_op);
1025 
1026 	if (ah->ar_pln != sizeof(struct in_addr))
1027 		goto out;
1028 
1029 	ifp = if_get_bylla(ar_sha(ah), ah->ar_hln, &psref);
1030 	if (ifp) {
1031 		if_put(ifp, &psref);
1032 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1033 		goto out;	/* it's from me, ignore it. */
1034 	}
1035 
1036 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
1037 	if (__predict_false(rcvif == NULL))
1038 		goto drop;
1039 
1040 	/*
1041 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or ar_tpa().
1042 	 * XXX check ar_hrd more strictly?
1043 	 */
1044 	switch (ifp->if_type) {
1045 	case IFT_IEEE1394:
1046 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
1047 			;
1048 		else {
1049 			/* XXX this is to make sure we compute ar_tha right */
1050 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
1051 		}
1052 		break;
1053 	default:
1054 		break;
1055 	}
1056 
1057 	memcpy(&isaddr, ar_spa(ah), sizeof(isaddr));
1058 	memcpy(&itaddr, ar_tpa(ah), sizeof(itaddr));
1059 
1060 	if (m->m_flags & (M_BCAST|M_MCAST))
1061 		ARP_STATINC(ARP_STAT_RCVMCAST);
1062 
1063 	/*
1064 	 * Search for a matching interface address
1065 	 * or any address on the interface to use
1066 	 * as a dummy address in the rest of this function.
1067 	 *
1068 	 * If the target IP address is zero then try and find
1069 	 * the sender address for DAD.
1070 	 */
1071 	myaddr = in_nullhost(itaddr) ? isaddr : itaddr;
1072 	s = pserialize_read_enter();
1073 	IN_ADDRHASH_READER_FOREACH(ia, myaddr.s_addr) {
1074 		if (!in_hosteq(ia->ia_addr.sin_addr, myaddr))
1075 			continue;
1076 #if NCARP > 0
1077 		if (ia->ia_ifp->if_type == IFT_CARP &&
1078 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1079 		    (IFF_UP|IFF_RUNNING))) {
1080 			index++;
1081 			/* XXX: ar_hln? */
1082 			if (ia->ia_ifp == rcvif && (ah->ar_hln >= 6) &&
1083 			    carp_iamatch(ia, ar_sha(ah),
1084 			    &count, index)) {
1085 				break;
1086 			}
1087 		} else
1088 #endif
1089 		if (ia->ia_ifp == rcvif)
1090 			break;
1091 #if NBRIDGE > 0
1092 		/*
1093 		 * If the interface we received the packet on
1094 		 * is part of a bridge, check to see if we need
1095 		 * to "bridge" the packet to ourselves at this
1096 		 * layer.  Note we still prefer a perfect match,
1097 		 * but allow this weaker match if necessary.
1098 		 */
1099 		if (rcvif->if_bridge != NULL &&
1100 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
1101 			bridge_ia = ia;
1102 #endif /* NBRIDGE > 0 */
1103 	}
1104 
1105 #if NBRIDGE > 0
1106 	if (ia == NULL && bridge_ia != NULL) {
1107 		ia = bridge_ia;
1108 		m_put_rcvif_psref(rcvif, &psref);
1109 		rcvif = NULL;
1110 		/* FIXME */
1111 		ifp = bridge_ia->ia_ifp;
1112 	}
1113 #endif
1114 	if (ia != NULL)
1115 		ia4_acquire(ia, &psref_ia);
1116 	pserialize_read_exit(s);
1117 
1118 	if (ah->ar_hln != ifp->if_addrlen) {
1119 		ARP_STATINC(ARP_STAT_RCVBADLEN);
1120 		log(LOG_WARNING,
1121 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
1122 		    IN_PRINT(ipbuf, &isaddr), ah->ar_hln, ifp->if_addrlen);
1123 		goto out;
1124 	}
1125 
1126 	/* Only do DaD if we have a matching address. */
1127 	do_dad = (ia != NULL);
1128 
1129 	if (ia == NULL) {
1130 		ia = in_get_ia_on_iface_psref(isaddr, rcvif, &psref_ia);
1131 		if (ia == NULL) {
1132 			ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
1133 			if (ia == NULL) {
1134 				ARP_STATINC(ARP_STAT_RCVNOINT);
1135 				goto out;
1136 			}
1137 		}
1138 	}
1139 
1140 	myaddr = ia->ia_addr.sin_addr;
1141 
1142 	/* XXX checks for bridge case? */
1143 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1144 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1145 		log(LOG_ERR,
1146 		    "%s: arp: link address is broadcast for IP address %s!\n",
1147 		    ifp->if_xname, IN_PRINT(ipbuf, &isaddr));
1148 		goto out;
1149 	}
1150 
1151 	/*
1152 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
1153 	 */
1154 	if (in_nullhost(isaddr))
1155 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
1156 	else if (in_hosteq(isaddr, myaddr))
1157 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1158 
1159 	if (in_nullhost(itaddr))
1160 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
1161 
1162 	/*
1163 	 * DAD check, RFC 5227.
1164 	 * Collision on sender address is always a duplicate.
1165 	 * Collision on target address is only a duplicate IF
1166 	 * the sender address is the null host (ie a DAD probe) AND
1167 	 * our address is in the TENTATIVE state.
1168 	 * DUPLICATED state is also checked so that processing stops here
1169 	 * and an error can be logged.
1170 	 */
1171 	if (do_dad &&
1172 	    (in_hosteq(isaddr, myaddr) ||
1173 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)
1174 	    && ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DUPLICATED))))
1175 	{
1176 		arp_dad_duplicated((struct ifaddr *)ia,
1177 		    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln));
1178 		goto out;
1179 	}
1180 
1181 	/*
1182 	 * If the target IP address is zero, ignore the packet.
1183 	 * This prevents the code below from trying to answer
1184 	 * when we are using IP address zero (booting).
1185 	 */
1186 	if (in_nullhost(itaddr))
1187 		goto out;
1188 
1189 	if (in_nullhost(isaddr))
1190 		goto reply;
1191 
1192 	if (in_hosteq(itaddr, myaddr))
1193 		la = arpcreate(ifp, m, &isaddr, NULL, 1);
1194 	else
1195 		la = arplookup(ifp, m, &isaddr, NULL, 1);
1196 	if (la == NULL)
1197 		goto reply;
1198 
1199 	if ((la->la_flags & LLE_VALID) &&
1200 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
1201 		if (la->la_flags & LLE_STATIC) {
1202 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
1203 			if (!log_permanent_modify)
1204 				goto out;
1205 			log(LOG_INFO,
1206 			    "%s tried to overwrite permanent arp info"
1207 			    " for %s\n",
1208 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1209 			    IN_PRINT(ipbuf, &isaddr));
1210 			goto out;
1211 		} else if (la->lle_tbl->llt_ifp != ifp) {
1212 			/* XXX should not happen? */
1213 			ARP_STATINC(ARP_STAT_RCVOVERINT);
1214 			if (!log_wrong_iface)
1215 				goto out;
1216 			log(LOG_INFO,
1217 			    "%s on %s tried to overwrite "
1218 			    "arp info for %s on %s\n",
1219 			    lla_snprintf(llabuf, ar_sha(ah), ah->ar_hln),
1220 			    ifp->if_xname, IN_PRINT(ipbuf, &isaddr),
1221 			    la->lle_tbl->llt_ifp->if_xname);
1222 				goto out;
1223 		} else {
1224 			ARP_STATINC(ARP_STAT_RCVOVER);
1225 			if (log_movements)
1226 				log(LOG_INFO, "arp info overwritten "
1227 				    "for %s by %s\n",
1228 				    IN_PRINT(ipbuf, &isaddr),
1229 				    lla_snprintf(llabuf, ar_sha(ah),
1230 				    ah->ar_hln));
1231 		}
1232 	}
1233 
1234 	/* XXX llentry should have addrlen? */
1235 #if 0
1236 	/*
1237 	 * sanity check for the address length.
1238 	 * XXX this does not work for protocols with variable address
1239 	 * length. -is
1240 	 */
1241 	if (sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
1242 		ARP_STATINC(ARP_STAT_RCVLENCHG);
1243 		log(LOG_WARNING,
1244 		    "arp from %s: new addr len %d, was %d\n",
1245 		    IN_PRINT(ipbuf, &isaddr), ah->ar_hln, sdl->sdl_alen);
1246 	}
1247 #endif
1248 
1249 #if NTOKEN > 0
1250 	/*
1251 	 * XXX uses m_data and assumes the complete answer including
1252 	 * XXX token-ring headers is in the same buf
1253 	 */
1254 	if (ifp->if_type == IFT_ISO88025) {
1255 		struct token_header *trh;
1256 
1257 		trh = (struct token_header *)M_TRHSTART(m);
1258 		if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1259 			struct token_rif *rif;
1260 			size_t riflen;
1261 
1262 			rif = TOKEN_RIF(trh);
1263 			riflen = (ntohs(rif->tr_rcf) &
1264 			    TOKEN_RCF_LEN_MASK) >> 8;
1265 
1266 			if (riflen > 2 &&
1267 			    riflen < sizeof(struct token_rif) &&
1268 			    (riflen & 1) == 0) {
1269 				rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1270 				rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1271 				memcpy(TOKEN_RIF_LLE(la), rif, riflen);
1272 			}
1273 		}
1274 	}
1275 #endif /* NTOKEN > 0 */
1276 
1277 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1278 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1279 	la->la_flags |= LLE_VALID;
1280 	if ((la->la_flags & LLE_STATIC) == 0) {
1281 		la->la_expire = time_uptime + arpt_keep;
1282 		arp_settimer(la, arpt_keep);
1283 	}
1284 	la->la_asked = 0;
1285 	/* rt->rt_flags &= ~RTF_REJECT; */
1286 
1287 	if (la->la_hold != NULL) {
1288 		int n = la->la_numheld;
1289 		struct mbuf *m_hold, *m_hold_next;
1290 		struct sockaddr_in sin;
1291 
1292 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1293 
1294 		m_hold = la->la_hold;
1295 		la->la_hold = NULL;
1296 		la->la_numheld = 0;
1297 		/*
1298 		 * We have to unlock here because if_output would call
1299 		 * arpresolve
1300 		 */
1301 		LLE_WUNLOCK(la);
1302 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1303 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1304 		for (; m_hold != NULL; m_hold = m_hold_next) {
1305 			m_hold_next = m_hold->m_nextpkt;
1306 			m_hold->m_nextpkt = NULL;
1307 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1308 		}
1309 	} else
1310 		LLE_WUNLOCK(la);
1311 	la = NULL;
1312 
1313 reply:
1314 	if (la != NULL) {
1315 		LLE_WUNLOCK(la);
1316 		la = NULL;
1317 	}
1318 	if (op != ARPOP_REQUEST) {
1319 		if (op == ARPOP_REPLY)
1320 			ARP_STATINC(ARP_STAT_RCVREPLY);
1321 		goto out;
1322 	}
1323 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1324 	if (in_hosteq(itaddr, myaddr)) {
1325 		/* If our address is unusable, don't reply */
1326 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1327 			goto out;
1328 		/* I am the target */
1329 		tha = ar_tha(ah);
1330 		if (tha)
1331 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1332 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1333 	} else {
1334 		/* Proxy ARP */
1335 		struct llentry *lle = NULL;
1336 		struct sockaddr_in sin;
1337 #if NCARP > 0
1338 		if (ifp->if_type == IFT_CARP) {
1339 			struct ifnet *_rcvif = m_get_rcvif(m, &s);
1340 			int iftype = 0;
1341 			if (__predict_true(_rcvif != NULL))
1342 				iftype = _rcvif->if_type;
1343 			m_put_rcvif(_rcvif, &s);
1344 			if (iftype != IFT_CARP)
1345 				goto out;
1346 		}
1347 #endif
1348 
1349 		tha = ar_tha(ah);
1350 
1351 		sockaddr_in_init(&sin, &itaddr, 0);
1352 
1353 		IF_AFDATA_RLOCK(ifp);
1354 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1355 		IF_AFDATA_RUNLOCK(ifp);
1356 
1357 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1358 			if (tha)
1359 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1360 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1361 			LLE_RUNLOCK(lle);
1362 		} else {
1363 			if (lle != NULL)
1364 				LLE_RUNLOCK(lle);
1365 			goto drop;
1366 		}
1367 	}
1368 	ia4_release(ia, &psref_ia);
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 		/*
1377 		 * ieee1394 arp reply is broadcast
1378 		 */
1379 		m->m_flags &= ~M_MCAST;
1380 		m->m_flags |= M_BCAST;
1381 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1382 		break;
1383 	default:
1384 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1385 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1386 		break;
1387 	}
1388 	m->m_pkthdr.len = m->m_len;
1389 	sa.sa_family = AF_ARP;
1390 	sa.sa_len = 2;
1391 	arps = ARP_STAT_GETREF();
1392 	arps[ARP_STAT_SNDTOTAL]++;
1393 	arps[ARP_STAT_SNDREPLY]++;
1394 	ARP_STAT_PUTREF();
1395 	if_output_lock(ifp, ifp, m, &sa, NULL);
1396 	if (rcvif != NULL)
1397 		m_put_rcvif_psref(rcvif, &psref);
1398 	return;
1399 
1400 out:
1401 	if (la != NULL)
1402 		LLE_WUNLOCK(la);
1403 drop:
1404 	if (ia != NULL)
1405 		ia4_release(ia, &psref_ia);
1406 	if (rcvif != NULL)
1407 		m_put_rcvif_psref(rcvif, &psref);
1408 	m_freem(m);
1409 }
1410 
1411 /*
1412  * Lookup or a new address in arptab.
1413  */
1414 static struct llentry *
1415 arplookup(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
1416     const struct sockaddr *sa, int wlock)
1417 {
1418 	struct sockaddr_in sin;
1419 	struct llentry *la;
1420 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1421 
1422 
1423 	if (sa == NULL) {
1424 		KASSERT(addr != NULL);
1425 		sockaddr_in_init(&sin, addr, 0);
1426 		sa = sintocsa(&sin);
1427 	}
1428 
1429 	IF_AFDATA_RLOCK(ifp);
1430 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1431 	IF_AFDATA_RUNLOCK(ifp);
1432 
1433 	return la;
1434 }
1435 
1436 static struct llentry *
1437 arpcreate(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
1438     const struct sockaddr *sa, int wlock)
1439 {
1440 	struct sockaddr_in sin;
1441 	struct llentry *la;
1442 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1443 
1444 	if (sa == NULL) {
1445 		KASSERT(addr != NULL);
1446 		sockaddr_in_init(&sin, addr, 0);
1447 		sa = sintocsa(&sin);
1448 	}
1449 
1450 	la = arplookup(ifp, m, addr, sa, wlock);
1451 
1452 	if (la == NULL) {
1453 		struct rtentry *rt;
1454 
1455 		rt = rtalloc1(sa, 0);
1456 		IF_AFDATA_WLOCK(ifp);
1457 		la = lla_create(LLTABLE(ifp), flags, sa, rt);
1458 		IF_AFDATA_WUNLOCK(ifp);
1459 		if (rt != NULL)
1460 			rt_unref(rt);
1461 
1462 		if (la != NULL)
1463 			arp_init_llentry(ifp, la);
1464 	}
1465 
1466 	return la;
1467 }
1468 
1469 int
1470 arpioctl(u_long cmd, void *data)
1471 {
1472 
1473 	return EOPNOTSUPP;
1474 }
1475 
1476 void
1477 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1478 {
1479 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1480 
1481 	ifa->ifa_rtrequest = arp_rtrequest;
1482 	ifa->ifa_flags |= RTF_CONNECTED;
1483 
1484 	/* ARP will handle DAD for this address. */
1485 	if (in_nullhost(IA_SIN(ifa)->sin_addr)) {
1486 		if (ia->ia_dad_stop != NULL)	/* safety */
1487 			ia->ia_dad_stop(ifa);
1488 		ia->ia_dad_start = NULL;
1489 		ia->ia_dad_stop = NULL;
1490 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1491 	} else {
1492 		ia->ia_dad_start = arp_dad_start;
1493 		ia->ia_dad_stop = arp_dad_stop;
1494 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE && ip_dad_count > 0)
1495 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1496 		else
1497 			arpannounce1(ifa);
1498 	}
1499 }
1500 
1501 TAILQ_HEAD(dadq_head, dadq);
1502 struct dadq {
1503 	TAILQ_ENTRY(dadq) dad_list;
1504 	struct ifaddr *dad_ifa;
1505 	int dad_count;		/* max ARP to send */
1506 	int dad_arp_tcount;	/* # of trials to send ARP */
1507 	int dad_arp_ocount;	/* ARP sent so far */
1508 	int dad_arp_announce;	/* max ARP announcements */
1509 	int dad_arp_acount;	/* # of announcements */
1510 	struct callout dad_timer_ch;
1511 };
1512 
1513 static struct dadq_head dadq;
1514 static int dad_init = 0;
1515 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1516 static kmutex_t arp_dad_lock;
1517 
1518 static struct dadq *
1519 arp_dad_find(struct ifaddr *ifa)
1520 {
1521 	struct dadq *dp;
1522 
1523 	KASSERT(mutex_owned(&arp_dad_lock));
1524 
1525 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1526 		if (dp->dad_ifa == ifa)
1527 			return dp;
1528 	}
1529 	return NULL;
1530 }
1531 
1532 static void
1533 arp_dad_starttimer(struct dadq *dp, int ticks)
1534 {
1535 
1536 	callout_reset(&dp->dad_timer_ch, ticks,
1537 	    (void (*)(void *))arp_dad_timer, (void *)dp->dad_ifa);
1538 }
1539 
1540 static void
1541 arp_dad_stoptimer(struct dadq *dp)
1542 {
1543 
1544 #ifdef NET_MPSAFE
1545 	callout_halt(&dp->dad_timer_ch, NULL);
1546 #else
1547 	callout_halt(&dp->dad_timer_ch, softnet_lock);
1548 #endif
1549 }
1550 
1551 static void
1552 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1553 {
1554 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1555 	struct ifnet *ifp = ifa->ifa_ifp;
1556 	struct in_addr sip;
1557 
1558 	dp->dad_arp_tcount++;
1559 	if ((ifp->if_flags & IFF_UP) == 0)
1560 		return;
1561 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1562 		return;
1563 
1564 	dp->dad_arp_tcount = 0;
1565 	dp->dad_arp_ocount++;
1566 
1567 	memset(&sip, 0, sizeof(sip));
1568 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1569 	    CLLADDR(ifa->ifa_ifp->if_sadl));
1570 }
1571 
1572 /*
1573  * Start Duplicate Address Detection (DAD) for specified interface address.
1574  */
1575 static void
1576 arp_dad_start(struct ifaddr *ifa)
1577 {
1578 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1579 	struct dadq *dp;
1580 	char ipbuf[INET_ADDRSTRLEN];
1581 
1582 	if (!dad_init) {
1583 		TAILQ_INIT(&dadq);
1584 		mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1585 		dad_init++;
1586 	}
1587 
1588 	/*
1589 	 * If we don't need DAD, don't do it.
1590 	 * - DAD is disabled (ip_dad_count == 0)
1591 	 */
1592 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1593 		log(LOG_DEBUG,
1594 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1595 		    IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1596 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1597 		return;
1598 	}
1599 	if (!ip_dad_count) {
1600 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1601 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1602 		arpannounce1(ifa);
1603 		return;
1604 	}
1605 	KASSERT(ifa->ifa_ifp != NULL);
1606 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1607 		return;
1608 
1609 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1610 
1611 	mutex_enter(&arp_dad_lock);
1612 	if (arp_dad_find(ifa) != NULL) {
1613 		mutex_exit(&arp_dad_lock);
1614 		/* DAD already in progress */
1615 		if (dp != NULL)
1616 			kmem_intr_free(dp, sizeof(*dp));
1617 		return;
1618 	}
1619 
1620 	if (dp == NULL) {
1621 		mutex_exit(&arp_dad_lock);
1622 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1623 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1624 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1625 		return;
1626 	}
1627 
1628 	/*
1629 	 * Send ARP packet for DAD, ip_dad_count times.
1630 	 * Note that we must delay the first transmission.
1631 	 */
1632 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1633 	dp->dad_ifa = ifa;
1634 	ifaref(ifa);	/* just for safety */
1635 	dp->dad_count = ip_dad_count;
1636 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1637 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1638 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1639 
1640 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1641 	    ARPLOGADDR(&ia->ia_addr.sin_addr));
1642 
1643 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1644 
1645 	mutex_exit(&arp_dad_lock);
1646 }
1647 
1648 /*
1649  * terminate DAD unconditionally.  used for address removals.
1650  */
1651 static void
1652 arp_dad_stop(struct ifaddr *ifa)
1653 {
1654 	struct dadq *dp;
1655 
1656 	if (!dad_init)
1657 		return;
1658 
1659 	mutex_enter(&arp_dad_lock);
1660 	dp = arp_dad_find(ifa);
1661 	if (dp == NULL) {
1662 		mutex_exit(&arp_dad_lock);
1663 		/* DAD wasn't started yet */
1664 		return;
1665 	}
1666 
1667 	/* Prevent the timer from running anymore. */
1668 	TAILQ_REMOVE(&dadq, dp, dad_list);
1669 	mutex_exit(&arp_dad_lock);
1670 
1671 	arp_dad_stoptimer(dp);
1672 
1673 	kmem_intr_free(dp, sizeof(*dp));
1674 	ifafree(ifa);
1675 }
1676 
1677 static void
1678 arp_dad_timer(struct ifaddr *ifa)
1679 {
1680 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1681 	struct dadq *dp;
1682 	char ipbuf[INET_ADDRSTRLEN];
1683 	bool need_free = false;
1684 
1685 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
1686 	mutex_enter(&arp_dad_lock);
1687 
1688 	/* Sanity check */
1689 	if (ia == NULL) {
1690 		log(LOG_ERR, "%s: called with null parameter\n", __func__);
1691 		goto done;
1692 	}
1693 	dp = arp_dad_find(ifa);
1694 	if (dp == NULL) {
1695 		/* DAD seems to be stopping, so do nothing. */
1696 		goto done;
1697 	}
1698 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1699 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1700 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1701 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1702 		goto done;
1703 	}
1704 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1705 	{
1706 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1707 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1708 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1709 		goto done;
1710 	}
1711 
1712 	/* timeouted with IFF_{RUNNING,UP} check */
1713 	if (dp->dad_arp_tcount > dad_maxtry) {
1714 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1715 		    if_name(ifa->ifa_ifp));
1716 
1717 		TAILQ_REMOVE(&dadq, dp, dad_list);
1718 		need_free = true;
1719 		goto done;
1720 	}
1721 
1722 	/* Need more checks? */
1723 	if (dp->dad_arp_ocount < dp->dad_count) {
1724 		int adelay;
1725 
1726 		/*
1727 		 * We have more ARP to go.  Send ARP packet for DAD.
1728 		 */
1729 		arp_dad_output(dp, ifa);
1730 		if (dp->dad_arp_ocount < dp->dad_count)
1731 			adelay = (PROBE_MIN * hz) +
1732 			    (cprng_fast32() %
1733 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1734 		else
1735 			adelay = ANNOUNCE_WAIT * hz;
1736 		arp_dad_starttimer(dp, adelay);
1737 		goto done;
1738 	} else if (dp->dad_arp_acount == 0) {
1739 		/*
1740 		 * We are done with DAD.
1741 		 * No duplicate address found.
1742 		 */
1743 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1744 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1745 		ARPLOG(LOG_DEBUG,
1746 		    "%s: DAD complete for %s - no duplicates found\n",
1747 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1748 		dp->dad_arp_announce = ANNOUNCE_NUM;
1749 		goto announce;
1750 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1751 announce:
1752 		/*
1753 		 * Announce the address.
1754 		 */
1755 		arpannounce1(ifa);
1756 		dp->dad_arp_acount++;
1757 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1758 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1759 			goto done;
1760 		}
1761 		ARPLOG(LOG_DEBUG,
1762 		    "%s: ARP announcement complete for %s\n",
1763 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1764 	}
1765 
1766 	TAILQ_REMOVE(&dadq, dp, dad_list);
1767 	need_free = true;
1768 done:
1769 	mutex_exit(&arp_dad_lock);
1770 
1771 	if (need_free) {
1772 		kmem_intr_free(dp, sizeof(*dp));
1773 		ifafree(ifa);
1774 	}
1775 
1776 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1777 }
1778 
1779 static void
1780 arp_dad_duplicated(struct ifaddr *ifa, const char *sha)
1781 {
1782 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1783 	struct ifnet *ifp = ifa->ifa_ifp;
1784 	char ipbuf[INET_ADDRSTRLEN];
1785 	const char *iastr;
1786 
1787 	iastr = IN_PRINT(ipbuf, &ia->ia_addr.sin_addr);
1788 
1789 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1790 		log(LOG_ERR,
1791 		    "%s: DAD duplicate address %s from %s\n",
1792 		    if_name(ifp), iastr, sha);
1793 	} else if (ia->ia_dad_defended == 0 ||
1794 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1795 		ia->ia_dad_defended = time_uptime;
1796 		arpannounce1(ifa);
1797 		log(LOG_ERR,
1798 		    "%s: DAD defended address %s from %s\n",
1799 		    if_name(ifp), iastr, sha);
1800 		return;
1801 	} else {
1802 		/* If DAD is disabled, just report the duplicate. */
1803 		if (ip_dad_count == 0) {
1804 			log(LOG_ERR,
1805 			    "%s: DAD ignoring duplicate address %s from %s\n",
1806 			    if_name(ifp), iastr, sha);
1807 			return;
1808 		}
1809 		log(LOG_ERR,
1810 		    "%s: DAD defence failed for %s from %s\n",
1811 		    if_name(ifp), iastr, sha);
1812 	}
1813 
1814 	arp_dad_stop(ifa);
1815 
1816 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1817 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1818 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1819 		/* Inform the routing socket of the duplicate address */
1820 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1821 	}
1822 }
1823 
1824 /*
1825  * Called from 10 Mb/s Ethernet interrupt handlers
1826  * when ether packet type ETHERTYPE_REVARP
1827  * is received.  Common length and type checks are done here,
1828  * then the protocol-specific routine is called.
1829  */
1830 void
1831 revarpinput(struct mbuf *m)
1832 {
1833 	struct arphdr *ar;
1834 
1835 	if (m->m_len < sizeof(struct arphdr))
1836 		goto out;
1837 	ar = mtod(m, struct arphdr *);
1838 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1839 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1840 		goto out;
1841 #endif
1842 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1843 		goto out;
1844 	switch (ntohs(ar->ar_pro)) {
1845 	case ETHERTYPE_IP:
1846 	case ETHERTYPE_IPTRAILERS:
1847 		in_revarpinput(m);
1848 		return;
1849 
1850 	default:
1851 		break;
1852 	}
1853 out:
1854 	m_freem(m);
1855 }
1856 
1857 /*
1858  * RARP for Internet protocols on 10 Mb/s Ethernet.
1859  * Algorithm is that given in RFC 903.
1860  * We are only using for bootstrap purposes to get an ip address for one of
1861  * our interfaces.  Thus we support no user-interface.
1862  *
1863  * Since the contents of the RARP reply are specific to the interface that
1864  * sent the request, this code must ensure that they are properly associated.
1865  *
1866  * Note: also supports ARP via RARP packets, per the RFC.
1867  */
1868 void
1869 in_revarpinput(struct mbuf *m)
1870 {
1871 	struct arphdr *ah;
1872 	void *tha;
1873 	int op;
1874 	struct ifnet *rcvif;
1875 	int s;
1876 
1877 	ah = mtod(m, struct arphdr *);
1878 	op = ntohs(ah->ar_op);
1879 
1880 	rcvif = m_get_rcvif(m, &s);
1881 	if (__predict_false(rcvif == NULL))
1882 		goto out;
1883 	switch (rcvif->if_type) {
1884 	case IFT_IEEE1394:
1885 		/* ARP without target hardware address is not supported */
1886 		goto out;
1887 	default:
1888 		break;
1889 	}
1890 
1891 	switch (op) {
1892 	case ARPOP_REQUEST:
1893 	case ARPOP_REPLY:	/* per RFC */
1894 		m_put_rcvif(rcvif, &s);
1895 		in_arpinput(m);
1896 		return;
1897 	case ARPOP_REVREPLY:
1898 		break;
1899 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1900 	default:
1901 		goto out;
1902 	}
1903 	if (!revarp_in_progress)
1904 		goto out;
1905 	if (rcvif != myip_ifp) /* !same interface */
1906 		goto out;
1907 	if (myip_initialized)
1908 		goto wake;
1909 	tha = ar_tha(ah);
1910 	if (tha == NULL)
1911 		goto out;
1912 	if (ah->ar_pln != sizeof(struct in_addr))
1913 		goto out;
1914 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1915 		goto out;
1916 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1917 		goto out;
1918 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1919 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1920 	myip_initialized = 1;
1921 wake:	/* Do wakeup every time in case it was missed. */
1922 	wakeup((void *)&myip);
1923 
1924 out:
1925 	m_put_rcvif(rcvif, &s);
1926 	m_freem(m);
1927 }
1928 
1929 /*
1930  * Send a RARP request for the ip address of the specified interface.
1931  * The request should be RFC 903-compliant.
1932  */
1933 static void
1934 revarprequest(struct ifnet *ifp)
1935 {
1936 	struct sockaddr sa;
1937 	struct mbuf *m;
1938 	struct arphdr *ah;
1939 	void *tha;
1940 
1941 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1942 		return;
1943 	MCLAIM(m, &arpdomain.dom_mowner);
1944 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1945 	    2*ifp->if_addrlen;
1946 	m->m_pkthdr.len = m->m_len;
1947 	MH_ALIGN(m, m->m_len);
1948 	ah = mtod(m, struct arphdr *);
1949 	memset(ah, 0, m->m_len);
1950 	ah->ar_pro = htons(ETHERTYPE_IP);
1951 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1952 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1953 	ah->ar_op = htons(ARPOP_REVREQUEST);
1954 
1955 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1956 	tha = ar_tha(ah);
1957 	if (tha == NULL) {
1958 		m_free(m);
1959 		return;
1960 	}
1961 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1962 
1963 	sa.sa_family = AF_ARP;
1964 	sa.sa_len = 2;
1965 	m->m_flags |= M_BCAST;
1966 
1967 	if_output_lock(ifp, ifp, m, &sa, NULL);
1968 }
1969 
1970 /*
1971  * RARP for the ip address of the specified interface, but also
1972  * save the ip address of the server that sent the answer.
1973  * Timeout if no response is received.
1974  */
1975 int
1976 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1977     struct in_addr *clnt_in)
1978 {
1979 	int result, count = 20;
1980 
1981 	myip_initialized = 0;
1982 	myip_ifp = ifp;
1983 
1984 	revarp_in_progress = 1;
1985 	while (count--) {
1986 		revarprequest(ifp);
1987 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1988 		if (result != EWOULDBLOCK)
1989 			break;
1990 	}
1991 	revarp_in_progress = 0;
1992 
1993 	if (!myip_initialized)
1994 		return ENETUNREACH;
1995 
1996 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1997 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1998 	return 0;
1999 }
2000 
2001 void
2002 arp_stat_add(int type, uint64_t count)
2003 {
2004 	ARP_STATADD(type, count);
2005 }
2006 
2007 static int
2008 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
2009 {
2010 
2011 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2012 }
2013 
2014 static void
2015 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2016 {
2017 	const struct sysctlnode *node;
2018 
2019 	sysctl_createv(clog, 0, NULL, NULL,
2020 			CTLFLAG_PERMANENT,
2021 			CTLTYPE_NODE, "inet", NULL,
2022 			NULL, 0, NULL, 0,
2023 			CTL_NET, PF_INET, CTL_EOL);
2024 	sysctl_createv(clog, 0, NULL, &node,
2025 			CTLFLAG_PERMANENT,
2026 			CTLTYPE_NODE, "arp",
2027 			SYSCTL_DESCR("Address Resolution Protocol"),
2028 			NULL, 0, NULL, 0,
2029 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2030 
2031 	sysctl_createv(clog, 0, NULL, NULL,
2032 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2033 			CTLTYPE_INT, "keep",
2034 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
2035 			NULL, 0, &arpt_keep, 0,
2036 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2037 
2038 	sysctl_createv(clog, 0, NULL, NULL,
2039 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2040 			CTLTYPE_INT, "down",
2041 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
2042 			NULL, 0, &arpt_down, 0,
2043 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2044 
2045 	sysctl_createv(clog, 0, NULL, NULL,
2046 			CTLFLAG_PERMANENT,
2047 			CTLTYPE_STRUCT, "stats",
2048 			SYSCTL_DESCR("ARP statistics"),
2049 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2050 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2051 
2052 	sysctl_createv(clog, 0, NULL, NULL,
2053 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2054 			CTLTYPE_INT, "log_movements",
2055 			SYSCTL_DESCR("log ARP replies from MACs different than"
2056 			    " the one in the cache"),
2057 			NULL, 0, &log_movements, 0,
2058 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2059 
2060 	sysctl_createv(clog, 0, NULL, NULL,
2061 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2062 			CTLTYPE_INT, "log_permanent_modify",
2063 			SYSCTL_DESCR("log ARP replies from MACs different than"
2064 			    " the one in the permanent arp entry"),
2065 			NULL, 0, &log_permanent_modify, 0,
2066 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2067 
2068 	sysctl_createv(clog, 0, NULL, NULL,
2069 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2070 			CTLTYPE_INT, "log_wrong_iface",
2071 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2072 			    " interface"),
2073 			NULL, 0, &log_wrong_iface, 0,
2074 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2075 
2076 	sysctl_createv(clog, 0, NULL, NULL,
2077 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2078 			CTLTYPE_INT, "log_unknown_network",
2079 			SYSCTL_DESCR("log ARP packets from non-local network"),
2080 			NULL, 0, &log_unknown_network, 0,
2081 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2082 
2083 	sysctl_createv(clog, 0, NULL, NULL,
2084 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2085 		       CTLTYPE_INT, "debug",
2086 		       SYSCTL_DESCR("Enable ARP DAD debug output"),
2087 		       NULL, 0, &arp_debug, 0,
2088 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2089 }
2090 
2091 #endif /* INET */
2092