xref: /netbsd-src/sys/netinet/if_arp.c (revision 8e33eff89e26cf71871ead62f0d5063e1313c33a)
1 /*	$NetBSD: if_arp.c,v 1.315 2024/09/09 07:25:32 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 1.315 2024/09/09 07:25:32 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_types.h>
109 #include <net/if_ether.h>
110 #include <net/if_llatbl.h>
111 #include <net/nd.h>
112 #include <net/route.h>
113 #include <net/net_stats.h>
114 
115 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/if_inarp.h>
120 
121 #include "arcnet.h"
122 #if NARCNET > 0
123 #include <net/if_arc.h>
124 #endif
125 #include "carp.h"
126 #if NCARP > 0
127 #include <netinet/ip_carp.h>
128 #endif
129 
130 /*
131  * ARP trailer negotiation.  Trailer protocol is not IP specific,
132  * but ARP request/response use IP addresses.
133  */
134 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
135 
136 /* timers */
137 static int arp_reachable = REACHABLE_TIME;
138 static int arp_retrans = RETRANS_TIMER;
139 static int arp_perform_nud = 1;
140 
141 static bool arp_nud_enabled(struct ifnet *);
142 static unsigned int arp_llinfo_reachable(struct ifnet *);
143 static unsigned int arp_llinfo_retrans(struct ifnet *);
144 static union l3addr *arp_llinfo_holdsrc(struct llentry *, union l3addr *);
145 static void arp_llinfo_output(struct ifnet *, const union l3addr *,
146     const union l3addr *, const uint8_t *, const union l3addr *);
147 static void arp_llinfo_missed(struct ifnet *, const union l3addr *,
148     int16_t, struct mbuf *);
149 static void arp_free(struct llentry *, int);
150 
151 static struct nd_domain arp_nd_domain = {
152 	.nd_family = AF_INET,
153 	.nd_delay = 5,		/* delay first probe time 5 second */
154 	.nd_mmaxtries = 3,	/* maximum broadcast query */
155 	.nd_umaxtries = 3,	/* maximum unicast query */
156 	.nd_retransmultiple = BACKOFF_MULTIPLE,
157 	.nd_maxretrans = MAX_RETRANS_TIMER,
158 	.nd_maxnudhint = 0,	/* max # of subsequent upper layer hints */
159 	.nd_maxqueuelen = 1,	/* max # of packets in unresolved ND entries */
160 	.nd_nud_enabled = arp_nud_enabled,
161 	.nd_reachable = arp_llinfo_reachable,
162 	.nd_retrans = arp_llinfo_retrans,
163 	.nd_holdsrc = arp_llinfo_holdsrc,
164 	.nd_output = arp_llinfo_output,
165 	.nd_missed = arp_llinfo_missed,
166 	.nd_free = arp_free,
167 };
168 
169 int ip_dad_count = PROBE_NUM;
170 #ifdef ARP_DEBUG
171 int arp_debug = 1;
172 #else
173 int arp_debug = 0;
174 #endif
175 
176 static void arp_init(void);
177 static void arp_dad_init(void);
178 
179 static void arprequest(struct ifnet *,
180     const struct in_addr *, const struct in_addr *,
181     const uint8_t *, const uint8_t *);
182 static void arpannounce1(struct ifaddr *);
183 static struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
184     const struct sockaddr *);
185 static struct llentry *arpcreate(struct ifnet *,
186     const struct in_addr *, const struct sockaddr *, int);
187 static void in_arpinput(struct mbuf *);
188 static void in_revarpinput(struct mbuf *);
189 static void revarprequest(struct ifnet *);
190 
191 static void arp_drainstub(void);
192 
193 struct dadq;
194 static void arp_dad_timer(struct dadq *);
195 static void arp_dad_start(struct ifaddr *);
196 static void arp_dad_stop(struct ifaddr *);
197 static void arp_dad_duplicated(struct ifaddr *, const struct sockaddr_dl *);
198 
199 #define	ARP_MAXQLEN	50
200 pktqueue_t *		arp_pktq		__read_mostly;
201 
202 static int useloopback = 1;	/* use loopback interface for local traffic */
203 
204 static percpu_t *arpstat_percpu;
205 
206 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
207 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
208 
209 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
210 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
211 
212 /* revarp state */
213 static struct in_addr myip, srv_ip;
214 static int myip_initialized = 0;
215 static int revarp_in_progress = 0;
216 static struct ifnet *myip_ifp = NULL;
217 
218 static int arp_drainwanted;
219 
220 static int log_movements = 0;
221 static int log_permanent_modify = 1;
222 static int log_wrong_iface = 1;
223 
224 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
225 
226 static void
227 arp_fasttimo(void)
228 {
229 	if (arp_drainwanted) {
230 		arp_drain();
231 		arp_drainwanted = 0;
232 	}
233 }
234 
235 static const struct protosw arpsw[] = {
236 	{
237 		.pr_type = 0,
238 		.pr_domain = &arpdomain,
239 		.pr_protocol = 0,
240 		.pr_flags = 0,
241 		.pr_input = 0,
242 		.pr_ctlinput = 0,
243 		.pr_ctloutput = 0,
244 		.pr_usrreqs = 0,
245 		.pr_init = arp_init,
246 		.pr_fasttimo = arp_fasttimo,
247 		.pr_slowtimo = 0,
248 		.pr_drain = arp_drainstub,
249 	}
250 };
251 
252 struct domain arpdomain = {
253 	.dom_family = PF_ARP,
254 	.dom_name = "arp",
255 	.dom_protosw = arpsw,
256 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
257 #ifdef MBUFTRACE
258 	.dom_mowner = MOWNER_INIT("internet", "arp"),
259 #endif
260 };
261 
262 static void sysctl_net_inet_arp_setup(struct sysctllog **);
263 
264 void
265 arp_init(void)
266 {
267 
268 	arp_pktq = pktq_create(ARP_MAXQLEN, arpintr, NULL);
269 	KASSERT(arp_pktq != NULL);
270 
271 	sysctl_net_inet_arp_setup(NULL);
272 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
273 
274 #ifdef MBUFTRACE
275 	MOWNER_ATTACH(&arpdomain.dom_mowner);
276 #endif
277 
278 	nd_attach_domain(&arp_nd_domain);
279 	arp_dad_init();
280 }
281 
282 static void
283 arp_drainstub(void)
284 {
285 	arp_drainwanted = 1;
286 }
287 
288 /*
289  * ARP protocol drain routine.  Called when memory is in short supply.
290  * Called at splvm();  don't acquire softnet_lock as can be called from
291  * hardware interrupt handlers.
292  */
293 void
294 arp_drain(void)
295 {
296 
297 	lltable_drain(AF_INET);
298 }
299 
300 /*
301  * We set the gateway for RTF_CLONING routes to a "prototype"
302  * link-layer sockaddr whose interface type (if_type) and interface
303  * index (if_index) fields are prepared.
304  */
305 static struct sockaddr *
306 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
307     const struct sockaddr *netmask)
308 {
309 	const struct ifnet *ifp = rt->rt_ifp;
310 	uint8_t namelen = strlen(ifp->if_xname);
311 	uint8_t addrlen = ifp->if_addrlen;
312 
313 	/*
314 	 * XXX: If this is a manually added route to interface
315 	 * such as older version of routed or gated might provide,
316 	 * restore cloning bit.
317 	 */
318 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
319 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
320 		rt->rt_flags |= RTF_CONNECTED;
321 
322 	if ((rt->rt_flags & (RTF_CONNECTED | RTF_LOCAL))) {
323 		union {
324 			struct sockaddr sa;
325 			struct sockaddr_storage ss;
326 			struct sockaddr_dl sdl;
327 		} u;
328 		/*
329 		 * Case 1: This route should come from a route to iface.
330 		 */
331 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
332 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
333 		rt_setgate(rt, &u.sa);
334 		gate = rt->rt_gateway;
335 	}
336 	return gate;
337 }
338 
339 /*
340  * Parallel to llc_rtrequest.
341  */
342 void
343 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
344 {
345 	struct sockaddr *gate = rt->rt_gateway;
346 	struct in_ifaddr *ia;
347 	struct ifaddr *ifa;
348 	struct ifnet *ifp = rt->rt_ifp;
349 	int bound;
350 	int s;
351 
352 	if (req == RTM_LLINFO_UPD) {
353 		if ((ifa = info->rti_ifa) != NULL)
354 			arpannounce1(ifa);
355 		return;
356 	}
357 
358 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
359 		if (req != RTM_ADD)
360 			return;
361 
362 		/*
363 		 * linklayers with particular link MTU limitation.
364 		 */
365 		switch(ifp->if_type) {
366 #if NARCNET > 0
367 		case IFT_ARCNET:
368 		    {
369 			int arcipifmtu;
370 
371 			if (ifp->if_flags & IFF_LINK0)
372 				arcipifmtu = arc_ipmtu;
373 			else
374 				arcipifmtu = ARCMTU;
375 			if (ifp->if_mtu > arcipifmtu)
376 				rt->rt_rmx.rmx_mtu = arcipifmtu;
377 			break;
378 		    }
379 #endif
380 		}
381 		return;
382 	}
383 
384 	switch (req) {
385 	case RTM_SETGATE:
386 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
387 		break;
388 	case RTM_ADD:
389 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
390 		if (gate == NULL) {
391 			log(LOG_ERR, "%s: arp_setgate failed\n", __func__);
392 			break;
393 		}
394 		if ((rt->rt_flags & RTF_CONNECTED) ||
395 		    (rt->rt_flags & RTF_LOCAL)) {
396 			/*
397 			 * linklayers with particular link MTU limitation.
398 			 */
399 			switch (ifp->if_type) {
400 #if NARCNET > 0
401 			case IFT_ARCNET:
402 			    {
403 				int arcipifmtu;
404 				if (ifp->if_flags & IFF_LINK0)
405 					arcipifmtu = arc_ipmtu;
406 				else
407 					arcipifmtu = ARCMTU;
408 
409 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
410 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
411 				     (rt->rt_rmx.rmx_mtu == 0 &&
412 				      ifp->if_mtu > arcipifmtu)))
413 					rt->rt_rmx.rmx_mtu = arcipifmtu;
414 				break;
415 			    }
416 #endif
417 			}
418 			if (rt->rt_flags & RTF_CONNECTED)
419 				break;
420 		}
421 
422 		bound = curlwp_bind();
423 		/* Announce a new entry if requested. */
424 		if (rt->rt_flags & RTF_ANNOUNCE) {
425 			struct psref psref;
426 			ia = in_get_ia_on_iface_psref(
427 			    satocsin(rt_getkey(rt))->sin_addr, ifp, &psref);
428 			if (ia != NULL) {
429 				arpannounce(ifp, &ia->ia_ifa,
430 				    CLLADDR(satocsdl(gate)));
431 				ia4_release(ia, &psref);
432 			}
433 		}
434 
435 		if (gate->sa_family != AF_LINK ||
436 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
437 			log(LOG_DEBUG, "%s: bad gateway value\n", __func__);
438 			goto out;
439 		}
440 
441 		satosdl(gate)->sdl_type = ifp->if_type;
442 		satosdl(gate)->sdl_index = ifp->if_index;
443 
444 		/*
445 		 * If the route is for a broadcast address mark it as such.
446 		 * This way we can avoid an expensive call to in_broadcast()
447 		 * in ip_output() most of the time (because the route passed
448 		 * to ip_output() is almost always a host route).
449 		 */
450 		if (rt->rt_flags & RTF_HOST &&
451 		    !(rt->rt_flags & RTF_BROADCAST) &&
452 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
453 			rt->rt_flags |= RTF_BROADCAST;
454 		/* There is little point in resolving the broadcast address */
455 		if (rt->rt_flags & RTF_BROADCAST)
456 			goto out;
457 
458 		/*
459 		 * When called from rt_ifa_addlocal, we cannot depend on that
460 		 * the address (rt_getkey(rt)) exits in the address list of the
461 		 * interface. So check RTF_LOCAL instead.
462 		 */
463 		if (rt->rt_flags & RTF_LOCAL) {
464 			if (useloopback) {
465 				rt->rt_ifp = lo0ifp;
466 				rt->rt_rmx.rmx_mtu = 0;
467 			}
468 			goto out;
469 		}
470 
471 		s = pserialize_read_enter();
472 		ia = in_get_ia_on_iface(satocsin(rt_getkey(rt))->sin_addr, ifp);
473 		if (ia == NULL) {
474 			pserialize_read_exit(s);
475 			goto out;
476 		}
477 
478 		if (useloopback) {
479 			rt->rt_ifp = lo0ifp;
480 			rt->rt_rmx.rmx_mtu = 0;
481 		}
482 		rt->rt_flags |= RTF_LOCAL;
483 
484 		if (ISSET(info->rti_flags, RTF_DONTCHANGEIFA)) {
485 			pserialize_read_exit(s);
486 			goto out;
487 		}
488 		/*
489 		 * make sure to set rt->rt_ifa to the interface
490 		 * address we are using, otherwise we will have trouble
491 		 * with source address selection.
492 		 */
493 		ifa = &ia->ia_ifa;
494 		if (ifa != rt->rt_ifa)
495 			/* Assume it doesn't sleep */
496 			rt_replace_ifa(rt, ifa);
497 		pserialize_read_exit(s);
498 	out:
499 		curlwp_bindx(bound);
500 		break;
501 	}
502 }
503 
504 /*
505  * Broadcast an ARP request. Caller specifies:
506  *	- arp header source ip address
507  *	- arp header target ip address
508  *	- arp header source ethernet address
509  */
510 static void
511 arprequest(struct ifnet *ifp,
512     const struct in_addr *sip, const struct in_addr *tip,
513     const uint8_t *saddr, const uint8_t *taddr)
514 {
515 	struct mbuf *m;
516 	struct arphdr *ah;
517 	struct sockaddr sa;
518 	net_stat_ref_t arps;
519 
520 	KASSERT(sip != NULL);
521 	KASSERT(tip != NULL);
522 	KASSERT(saddr != NULL);
523 
524 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
525 		return;
526 	MCLAIM(m, &arpdomain.dom_mowner);
527 	switch (ifp->if_type) {
528 	case IFT_IEEE1394:
529 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
530 		    ifp->if_addrlen;
531 		break;
532 	default:
533 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
534 		    2 * ifp->if_addrlen;
535 		break;
536 	}
537 	m->m_pkthdr.len = m->m_len;
538 	m_align(m, m->m_len);
539 	ah = mtod(m, struct arphdr *);
540 	memset(ah, 0, m->m_len);
541 	switch (ifp->if_type) {
542 	case IFT_IEEE1394:	/* RFC2734 */
543 		/* fill it now for ar_tpa computation */
544 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
545 		break;
546 	default:
547 		/* ifp->if_output will fill ar_hrd */
548 		break;
549 	}
550 	ah->ar_pro = htons(ETHERTYPE_IP);
551 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
552 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
553 	ah->ar_op = htons(ARPOP_REQUEST);
554 	memcpy(ar_sha(ah), saddr, ah->ar_hln);
555 	if (taddr == NULL)
556 		m->m_flags |= M_BCAST;
557 	else
558 		memcpy(ar_tha(ah), taddr, ah->ar_hln);
559 	memcpy(ar_spa(ah), sip, ah->ar_pln);
560 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
561 	sa.sa_family = AF_ARP;
562 	sa.sa_len = 2;
563 	arps = ARP_STAT_GETREF();
564 	_NET_STATINC_REF(arps, ARP_STAT_SNDTOTAL);
565 	_NET_STATINC_REF(arps, ARP_STAT_SENDREQUEST);
566 	ARP_STAT_PUTREF();
567 	if_output_lock(ifp, ifp, m, &sa, NULL);
568 }
569 
570 void
571 arpannounce(struct ifnet *ifp, struct ifaddr *ifa, const uint8_t *enaddr)
572 {
573 	struct in_ifaddr *ia = ifatoia(ifa);
574 	struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
575 
576 	if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) {
577 		ARPLOG(LOG_DEBUG, "%s not ready\n", ARPLOGADDR(ip));
578 		return;
579 	}
580 	arprequest(ifp, ip, ip, enaddr, NULL);
581 }
582 
583 static void
584 arpannounce1(struct ifaddr *ifa)
585 {
586 
587 	arpannounce(ifa->ifa_ifp, ifa, CLLADDR(ifa->ifa_ifp->if_sadl));
588 }
589 
590 /*
591  * Resolve an IP address into an ethernet address.  If success, desten is
592  * filled in. If there is no entry in arptab, set one up and broadcast a
593  * request for the IP address. Hold onto this mbuf and resend it once the
594  * address is finally resolved.
595  *
596  * A return value of 0 indicates that desten has been filled in and the packet
597  * should be sent normally; a return value of EWOULDBLOCK indicates that the
598  * packet has been held pending resolution. Any other value indicates an
599  * error.
600  */
601 int
602 arpresolve(struct ifnet *ifp, const struct rtentry *rt, struct mbuf *m,
603     const struct sockaddr *dst, void *desten, size_t destlen)
604 {
605 	struct llentry *la;
606 	const char *create_lookup;
607 	int error;
608 
609 #if NCARP > 0
610 	if (rt != NULL && rt->rt_ifp->if_type == IFT_CARP)
611 		ifp = rt->rt_ifp;
612 #endif
613 
614 	KASSERT(m != NULL);
615 
616 	la = arplookup(ifp, NULL, dst, 0);
617 	if (la == NULL)
618 		goto notfound;
619 
620 	if (la->la_flags & LLE_VALID && la->ln_state == ND_LLINFO_REACHABLE) {
621 		KASSERT(destlen >= ifp->if_addrlen);
622 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
623 		LLE_RUNLOCK(la);
624 		return 0;
625 	}
626 
627 notfound:
628 	if (ifp->if_flags & IFF_NOARP) {
629 		if (la != NULL)
630 			LLE_RUNLOCK(la);
631 		error = ENOTSUP;
632 		goto bad;
633 	}
634 
635 	if (la == NULL) {
636 		struct rtentry *_rt;
637 
638 		create_lookup = "create";
639 		_rt = rtalloc1(dst, 0);
640 		IF_AFDATA_WLOCK(ifp);
641 		la = lla_create(LLTABLE(ifp), LLE_EXCLUSIVE, dst, _rt);
642 		IF_AFDATA_WUNLOCK(ifp);
643 		if (_rt != NULL)
644 			rt_unref(_rt);
645 		if (la == NULL)
646 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
647 		else
648 			la->ln_state = ND_LLINFO_NOSTATE;
649 	} else if (LLE_TRY_UPGRADE(la) == 0) {
650 		create_lookup = "lookup";
651 		LLE_RUNLOCK(la);
652 		IF_AFDATA_RLOCK(ifp);
653 		la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
654 		IF_AFDATA_RUNLOCK(ifp);
655 	}
656 
657 	error = EINVAL;
658 	if (la == NULL) {
659 		log(LOG_DEBUG,
660 		    "%s: failed to %s llentry for %s on %s\n",
661 		    __func__, create_lookup, inet_ntoa(satocsin(dst)->sin_addr),
662 		    ifp->if_xname);
663 		goto bad;
664 	}
665 
666 	error = nd_resolve(la, rt, m, desten, destlen);
667 	return error;
668 
669 bad:
670 	m_freem(m);
671 	return error;
672 }
673 
674 /*
675  * Common length and type checks are done here,
676  * then the protocol-specific routine is called.
677  */
678 void
679 arpintr(void *arg __unused)
680 {
681 	struct mbuf *m;
682 	struct arphdr *ar;
683 	int s;
684 	int arplen;
685 	struct ifnet *rcvif;
686 	bool badhrd;
687 
688 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
689 	while ((m = pktq_dequeue(arp_pktq)) != NULL) {
690 		if ((m->m_flags & M_PKTHDR) == 0)
691 			panic("arpintr");
692 
693 		MCLAIM(m, &arpdomain.dom_mowner);
694 		ARP_STATINC(ARP_STAT_RCVTOTAL);
695 
696 		if (__predict_false(m->m_len < sizeof(*ar))) {
697 			if ((m = m_pullup(m, sizeof(*ar))) == NULL)
698 				goto badlen;
699 		}
700 		ar = mtod(m, struct arphdr *);
701 		KASSERT(ACCESSIBLE_POINTER(ar, struct arphdr));
702 
703 		rcvif = m_get_rcvif(m, &s);
704 		if (__predict_false(rcvif == NULL)) {
705 			ARP_STATINC(ARP_STAT_RCVNOINT);
706 			goto free;
707 		}
708 
709 		/*
710 		 * We don't want non-IEEE1394 ARP packets on IEEE1394
711 		 * interfaces, and vice versa. Our life depends on that.
712 		 */
713 		if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394)
714 			badhrd = rcvif->if_type != IFT_IEEE1394;
715 		else
716 			badhrd = rcvif->if_type == IFT_IEEE1394;
717 
718 		m_put_rcvif(rcvif, &s);
719 
720 		if (badhrd) {
721 			ARP_STATINC(ARP_STAT_RCVBADPROTO);
722 			goto free;
723 		}
724 
725 		arplen = sizeof(*ar) + 2 * ar->ar_hln + 2 * ar->ar_pln;
726 		if (__predict_false(m->m_len < arplen)) {
727 			if ((m = m_pullup(m, arplen)) == NULL)
728 				goto badlen;
729 			ar = mtod(m, struct arphdr *);
730 			KASSERT(ACCESSIBLE_POINTER(ar, struct arphdr));
731 		}
732 
733 		switch (ntohs(ar->ar_pro)) {
734 		case ETHERTYPE_IP:
735 		case ETHERTYPE_IPTRAILERS:
736 			in_arpinput(m);
737 			continue;
738 		default:
739 			ARP_STATINC(ARP_STAT_RCVBADPROTO);
740 			goto free;
741 		}
742 
743 badlen:
744 		ARP_STATINC(ARP_STAT_RCVBADLEN);
745 free:
746 		m_freem(m);
747 	}
748 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
749 	return; /* XXX gcc */
750 }
751 
752 /*
753  * ARP for Internet protocols on 10 Mb/s Ethernet. Algorithm is that given in
754  * RFC 826. In addition, a sanity check is performed on the sender protocol
755  * address, to catch impersonators.
756  *
757  * We no longer handle negotiations for use of trailer protocol: formerly, ARP
758  * replied for protocol type ETHERTYPE_TRAIL sent along with IP replies if we
759  * wanted trailers sent to us, and also sent them in response to IP replies.
760  * This allowed either end to announce the desire to receive trailer packets.
761  *
762  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, but
763  * formerly didn't normally send requests.
764  */
765 static void
766 in_arpinput(struct mbuf *m)
767 {
768 	struct arphdr *ah;
769 	struct ifnet *ifp, *rcvif = NULL;
770 	struct llentry *la = NULL;
771 	struct in_ifaddr *ia = NULL;
772 #if NBRIDGE > 0
773 	struct in_ifaddr *bridge_ia = NULL;
774 #endif
775 #if NCARP > 0
776 	uint32_t count = 0, index = 0;
777 #endif
778 	struct sockaddr sa;
779 	struct in_addr isaddr, itaddr, myaddr;
780 	int op, rt_cmd, new_state = 0;
781 	void *tha;
782 	net_stat_ref_t arps;
783 	struct psref psref, psref_ia;
784 	int s;
785 	char ipbuf[INET_ADDRSTRLEN];
786 	bool find_source, do_dad;
787 
788 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
789 		goto out;
790 	ah = mtod(m, struct arphdr *);
791 	op = ntohs(ah->ar_op);
792 
793 	if (ah->ar_pln != sizeof(struct in_addr))
794 		goto out;
795 
796 	ifp = if_get_bylla(ar_sha(ah), ah->ar_hln, &psref);
797 	if (ifp) {
798 		/* it's from me, ignore it. */
799 		if_put(ifp, &psref);
800 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
801 		goto out;
802 	}
803 
804 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
805 	if (__predict_false(rcvif == NULL))
806 		goto out;
807 	if (rcvif->if_flags & IFF_NOARP)
808 		goto out;
809 
810 	memcpy(&isaddr, ar_spa(ah), sizeof(isaddr));
811 	memcpy(&itaddr, ar_tpa(ah), sizeof(itaddr));
812 
813 	if (m->m_flags & (M_BCAST|M_MCAST))
814 		ARP_STATINC(ARP_STAT_RCVMCAST);
815 
816 	/*
817 	 * Search for a matching interface address
818 	 * or any address on the interface to use
819 	 * as a dummy address in the rest of this function.
820 	 *
821 	 * First try and find the source address for early
822 	 * duplicate address detection.
823 	 */
824 	if (in_nullhost(isaddr)) {
825 		if (in_nullhost(itaddr)) /* very bogus ARP */
826 			goto out;
827 		find_source = false;
828 		myaddr = itaddr;
829 	} else {
830 		find_source = true;
831 		myaddr = isaddr;
832 	}
833 	s = pserialize_read_enter();
834 again:
835 	IN_ADDRHASH_READER_FOREACH(ia, myaddr.s_addr) {
836 		if (!in_hosteq(ia->ia_addr.sin_addr, myaddr))
837 			continue;
838 #if NCARP > 0
839 		if (ia->ia_ifp->if_type == IFT_CARP &&
840 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
841 		    (IFF_UP|IFF_RUNNING))) {
842 			index++;
843 			/* XXX: ar_hln? */
844 			if (ia->ia_ifp == rcvif && (ah->ar_hln >= 6) &&
845 			    carp_iamatch(ia, ar_sha(ah),
846 			    &count, index)) {
847 				break;
848 			}
849 		} else
850 #endif
851 		if (ia->ia_ifp == rcvif)
852 			break;
853 #if NBRIDGE > 0
854 		/*
855 		 * If the interface we received the packet on
856 		 * is part of a bridge, check to see if we need
857 		 * to "bridge" the packet to ourselves at this
858 		 * layer.  Note we still prefer a perfect match,
859 		 * but allow this weaker match if necessary.
860 		 */
861 		if (rcvif->if_bridge != NULL &&
862 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
863 			bridge_ia = ia;
864 #endif
865 	}
866 
867 #if NBRIDGE > 0
868 	if (ia == NULL && bridge_ia != NULL) {
869 		ia = bridge_ia;
870 		m_put_rcvif_psref(rcvif, &psref);
871 		rcvif = NULL;
872 		/* FIXME */
873 		ifp = bridge_ia->ia_ifp;
874 	}
875 #endif
876 
877 	/* If we failed to find the source address then find
878 	 * the target address. */
879 	if (ia == NULL && find_source && !in_nullhost(itaddr)) {
880 		find_source = false;
881 		myaddr = itaddr;
882 		goto again;
883 	}
884 
885 	if (ia != NULL)
886 		ia4_acquire(ia, &psref_ia);
887 	pserialize_read_exit(s);
888 
889 	if (ah->ar_hln != ifp->if_addrlen) {
890 		ARP_STATINC(ARP_STAT_RCVBADLEN);
891 		log(LOG_WARNING,
892 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
893 		    IN_PRINT(ipbuf, &isaddr), ah->ar_hln, ifp->if_addrlen);
894 		goto out;
895 	}
896 
897 	/* Only do DaD if we have a matching address. */
898 	do_dad = (ia != NULL);
899 
900 	if (ia == NULL) {
901 		ia = in_get_ia_on_iface_psref(isaddr, rcvif, &psref_ia);
902 		if (ia == NULL) {
903 			ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
904 			if (ia == NULL) {
905 				ARP_STATINC(ARP_STAT_RCVNOINT);
906 				goto out;
907 			}
908 		}
909 	}
910 
911 	myaddr = ia->ia_addr.sin_addr;
912 
913 	/* XXX checks for bridge case? */
914 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
915 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
916 		log(LOG_ERR,
917 		    "%s: arp: link address is broadcast for IP address %s!\n",
918 		    ifp->if_xname, IN_PRINT(ipbuf, &isaddr));
919 		goto out;
920 	}
921 
922 	/*
923 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
924 	 */
925 	if (in_nullhost(isaddr))
926 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
927 	else if (in_hosteq(isaddr, myaddr)) {
928 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
929 		/* This is the original behavior prior to supporting IPv4 DAD */
930 		if (!ip_dad_enabled()) {
931 			char llabuf[LLA_ADDRSTRLEN];
932 			log(LOG_ERR,
933 			    "duplicate IP address %s sent from link address %s\n",
934 			    IN_PRINT(ipbuf, &isaddr),
935 			    lla_snprintf(llabuf, sizeof(llabuf), ar_sha(ah),
936 			                 ah->ar_hln));
937 			itaddr = myaddr;
938 			goto reply;
939 		}
940 	}
941 
942 	if (in_nullhost(itaddr))
943 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
944 
945 	/*
946 	 * DAD check, RFC 5227.
947 	 * ARP sender hardware address must match the interface
948 	 * address of the interface sending the packet.
949 	 * Collision on sender address is always a duplicate.
950 	 * Collision on target address is only a duplicate
951 	 * IF the sender address is the null host (ie a DAD probe)
952 	 * AND the message was broadcast
953 	 * AND our address is either tentative or duplicated
954 	 * If it was unicast then it's a valid Unicast Poll from RFC 1122.
955 	 */
956 	if (ip_dad_enabled() && do_dad &&
957 	    (in_hosteq(isaddr, myaddr) ||
958 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr) &&
959 	     m->m_flags & M_BCAST &&
960 	     ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DUPLICATED))))
961 	{
962 		struct m_tag *mtag;
963 
964 		mtag = m_tag_find(m, PACKET_TAG_ETHERNET_SRC);
965 		if (mtag == NULL || (ah->ar_hln == ETHER_ADDR_LEN &&
966 		    memcmp(mtag + 1, ar_sha(ah), ah->ar_hln) == 0)) {
967 			struct sockaddr_dl sdl, *sdlp;
968 
969 			sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
970 			    ifp->if_index, ifp->if_type,
971 			    NULL, 0, ar_sha(ah), ah->ar_hln);
972 			arp_dad_duplicated((struct ifaddr *)ia, sdlp);
973 			goto out;
974 		}
975 	}
976 
977 	/*
978 	 * If the target IP address is zero, ignore the packet.
979 	 * This prevents the code below from trying to answer
980 	 * when we are using IP address zero (booting).
981 	 */
982 	if (in_nullhost(itaddr))
983 		goto out;
984 
985 	if (in_nullhost(isaddr))
986 		goto reply;
987 
988 	if (in_hosteq(itaddr, myaddr))
989 		la = arpcreate(ifp, &isaddr, NULL, 1);
990 	else
991 		la = arplookup(ifp, &isaddr, NULL, 1);
992 	if (la == NULL)
993 		goto reply;
994 
995 	if ((la->la_flags & LLE_VALID) &&
996 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen))
997 	{
998 		char llabuf[LLA_ADDRSTRLEN], *llastr;
999 
1000 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
1001 		    ar_sha(ah), ah->ar_hln);
1002 
1003 		if (la->la_flags & LLE_STATIC) {
1004 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
1005 			if (!log_permanent_modify)
1006 				goto out;
1007 			log(LOG_INFO,
1008 			    "%s tried to overwrite permanent arp info"
1009 			    " for %s\n", llastr, IN_PRINT(ipbuf, &isaddr));
1010 			goto out;
1011 		} else if (la->lle_tbl->llt_ifp != ifp) {
1012 			/* XXX should not happen? */
1013 			ARP_STATINC(ARP_STAT_RCVOVERINT);
1014 			if (!log_wrong_iface)
1015 				goto out;
1016 			log(LOG_INFO,
1017 			    "%s on %s tried to overwrite "
1018 			    "arp info for %s on %s\n",
1019 			    llastr,
1020 			    ifp->if_xname, IN_PRINT(ipbuf, &isaddr),
1021 			    la->lle_tbl->llt_ifp->if_xname);
1022 				goto out;
1023 		} else {
1024 			ARP_STATINC(ARP_STAT_RCVOVER);
1025 			if (log_movements)
1026 				log(LOG_INFO, "arp info overwritten "
1027 				    "for %s by %s\n",
1028 				    IN_PRINT(ipbuf, &isaddr), llastr);
1029 		}
1030 		rt_cmd = RTM_CHANGE;
1031 		new_state = ND_LLINFO_STALE;
1032 	} else {
1033 		if (op == ARPOP_REPLY && in_hosteq(itaddr, myaddr)) {
1034 			/* This was a solicited ARP reply. */
1035 			la->ln_byhint = 0;
1036 			new_state = ND_LLINFO_REACHABLE;
1037 		} else if (op == ARPOP_REQUEST &&
1038 		           (la->ln_state == ND_LLINFO_NOSTATE ||
1039 			    la->ln_state == ND_LLINFO_INCOMPLETE)) {
1040 			/*
1041 			 * If an ARP request comes but there is no entry
1042 			 * and a new one has been created or an entry exists
1043 			 * but incomplete, make it stale to allow to send
1044 			 * packets to the requester without an ARP resolution.
1045 			 */
1046 			la->ln_byhint = 0;
1047 			new_state = ND_LLINFO_STALE;
1048 		}
1049 		rt_cmd = la->la_flags & LLE_VALID ? 0 : RTM_ADD;
1050 	}
1051 
1052 	KASSERT(ifp->if_sadl->sdl_alen == ifp->if_addrlen);
1053 
1054 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1055 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1056 	la->la_flags |= LLE_VALID;
1057 	la->ln_asked = 0;
1058 	if (new_state != 0) {
1059 		la->ln_state = new_state;
1060 
1061 		if (new_state != ND_LLINFO_REACHABLE ||
1062 		    !(la->la_flags & LLE_STATIC))
1063 		{
1064 			int timer = ND_TIMER_GC;
1065 
1066 			if (new_state == ND_LLINFO_REACHABLE)
1067 				timer = ND_TIMER_REACHABLE;
1068 			nd_set_timer(la, timer);
1069 		}
1070 	}
1071 
1072 	if (rt_cmd != 0) {
1073 		struct sockaddr_in sin;
1074 
1075 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1076 		rt_clonedmsg(rt_cmd, NULL, sintosa(&sin), ar_sha(ah), ifp);
1077 	}
1078 
1079 	if (la->la_hold != NULL) {
1080 		int n = la->la_numheld;
1081 		struct mbuf *m_hold, *m_hold_next;
1082 		struct sockaddr_in sin;
1083 
1084 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1085 
1086 		m_hold = la->la_hold;
1087 		la->la_hold = NULL;
1088 		la->la_numheld = 0;
1089 		/*
1090 		 * We have to unlock here because if_output would call
1091 		 * arpresolve
1092 		 */
1093 		LLE_WUNLOCK(la);
1094 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1095 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1096 		for (; m_hold != NULL; m_hold = m_hold_next) {
1097 			m_hold_next = m_hold->m_nextpkt;
1098 			m_hold->m_nextpkt = NULL;
1099 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1100 		}
1101 	} else
1102 		LLE_WUNLOCK(la);
1103 	la = NULL;
1104 
1105 reply:
1106 	if (la != NULL) {
1107 		LLE_WUNLOCK(la);
1108 		la = NULL;
1109 	}
1110 	if (op != ARPOP_REQUEST) {
1111 		if (op == ARPOP_REPLY)
1112 			ARP_STATINC(ARP_STAT_RCVREPLY);
1113 		goto out;
1114 	}
1115 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1116 	if (in_hosteq(itaddr, myaddr)) {
1117 		/* If our address is unusable, don't reply */
1118 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1119 			goto out;
1120 		/* I am the target */
1121 		tha = ar_tha(ah);
1122 		if (tha)
1123 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1124 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1125 	} else {
1126 		/* Proxy ARP */
1127 		struct llentry *lle = NULL;
1128 		struct sockaddr_in sin;
1129 
1130 #if NCARP > 0
1131 		if (ifp->if_type == IFT_CARP) {
1132 			struct ifnet *_rcvif = m_get_rcvif(m, &s);
1133 			int iftype = 0;
1134 			if (__predict_true(_rcvif != NULL))
1135 				iftype = _rcvif->if_type;
1136 			m_put_rcvif(_rcvif, &s);
1137 			if (iftype != IFT_CARP)
1138 				goto out;
1139 		}
1140 #endif
1141 
1142 		tha = ar_tha(ah);
1143 
1144 		sockaddr_in_init(&sin, &itaddr, 0);
1145 
1146 		IF_AFDATA_RLOCK(ifp);
1147 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1148 		IF_AFDATA_RUNLOCK(ifp);
1149 
1150 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1151 			if (tha)
1152 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1153 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1154 			LLE_RUNLOCK(lle);
1155 		} else {
1156 			if (lle != NULL)
1157 				LLE_RUNLOCK(lle);
1158 			goto out;
1159 		}
1160 	}
1161 	ia4_release(ia, &psref_ia);
1162 
1163 	/*
1164 	 * XXX XXX: Here we're recycling the mbuf. But the mbuf could have
1165 	 * other mbufs in its chain, and just overwriting m->m_pkthdr.len
1166 	 * would be wrong in this case (the length becomes smaller than the
1167 	 * real chain size).
1168 	 *
1169 	 * This can theoretically cause bugs in the lower layers (drivers,
1170 	 * and L2encap), in some corner cases.
1171 	 */
1172 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1173 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1174 	ah->ar_op = htons(ARPOP_REPLY);
1175 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1176 	switch (ifp->if_type) {
1177 	case IFT_IEEE1394:
1178 		/* ieee1394 arp reply is broadcast */
1179 		m->m_flags &= ~M_MCAST;
1180 		m->m_flags |= M_BCAST;
1181 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1182 		break;
1183 	default:
1184 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1185 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1186 		break;
1187 	}
1188 	m->m_pkthdr.len = m->m_len;
1189 	sa.sa_family = AF_ARP;
1190 	sa.sa_len = 2;
1191 	arps = ARP_STAT_GETREF();
1192 	_NET_STATINC_REF(arps, ARP_STAT_SNDTOTAL);
1193 	_NET_STATINC_REF(arps, ARP_STAT_SNDREPLY);
1194 	ARP_STAT_PUTREF();
1195 	if_output_lock(ifp, ifp, m, &sa, NULL);
1196 	if (rcvif != NULL)
1197 		m_put_rcvif_psref(rcvif, &psref);
1198 	return;
1199 
1200 out:
1201 	if (la != NULL)
1202 		LLE_WUNLOCK(la);
1203 	if (ia != NULL)
1204 		ia4_release(ia, &psref_ia);
1205 	if (rcvif != NULL)
1206 		m_put_rcvif_psref(rcvif, &psref);
1207 	m_freem(m);
1208 }
1209 
1210 /*
1211  * Lookup or a new address in arptab.
1212  */
1213 struct llentry *
1214 arplookup(struct ifnet *ifp, const struct in_addr *addr,
1215     const struct sockaddr *sa, int wlock)
1216 {
1217 	struct sockaddr_in sin;
1218 	struct llentry *la;
1219 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1220 
1221 	if (sa == NULL) {
1222 		KASSERT(addr != NULL);
1223 		sockaddr_in_init(&sin, addr, 0);
1224 		sa = sintocsa(&sin);
1225 	}
1226 
1227 	IF_AFDATA_RLOCK(ifp);
1228 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1229 	IF_AFDATA_RUNLOCK(ifp);
1230 
1231 	return la;
1232 }
1233 
1234 static struct llentry *
1235 arpcreate(struct ifnet *ifp, const struct in_addr *addr,
1236     const struct sockaddr *sa, int wlock)
1237 {
1238 	struct sockaddr_in sin;
1239 	struct llentry *la;
1240 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1241 
1242 	if (sa == NULL) {
1243 		KASSERT(addr != NULL);
1244 		sockaddr_in_init(&sin, addr, 0);
1245 		sa = sintocsa(&sin);
1246 	}
1247 
1248 	la = arplookup(ifp, addr, sa, wlock);
1249 
1250 	if (la == NULL) {
1251 		struct rtentry *rt;
1252 
1253 		rt = rtalloc1(sa, 0);
1254 		IF_AFDATA_WLOCK(ifp);
1255 		la = lla_create(LLTABLE(ifp), flags, sa, rt);
1256 		IF_AFDATA_WUNLOCK(ifp);
1257 		if (rt != NULL)
1258 			rt_unref(rt);
1259 
1260 		if (la != NULL)
1261 			la->ln_state = ND_LLINFO_NOSTATE;
1262 	}
1263 
1264 	return la;
1265 }
1266 
1267 int
1268 arpioctl(u_long cmd, void *data)
1269 {
1270 
1271 	return EOPNOTSUPP;
1272 }
1273 
1274 void
1275 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1276 {
1277 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1278 
1279 	ifa->ifa_rtrequest = arp_rtrequest;
1280 	ifa->ifa_flags |= RTF_CONNECTED;
1281 
1282 	/* ARP will handle DAD for this address. */
1283 	if (in_nullhost(IA_SIN(ifa)->sin_addr)) {
1284 		if (ia->ia_dad_stop != NULL)	/* safety */
1285 			ia->ia_dad_stop(ifa);
1286 		ia->ia_dad_start = NULL;
1287 		ia->ia_dad_stop = NULL;
1288 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1289 	} else {
1290 		ia->ia_dad_start = arp_dad_start;
1291 		ia->ia_dad_stop = arp_dad_stop;
1292 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE && ip_dad_enabled())
1293 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1294 		else
1295 			arpannounce1(ifa);
1296 	}
1297 }
1298 
1299 static bool
1300 arp_nud_enabled(__unused struct ifnet *ifp)
1301 {
1302 
1303 	return arp_perform_nud != 0;
1304 }
1305 
1306 static unsigned int
1307 arp_llinfo_reachable(__unused struct ifnet *ifp)
1308 {
1309 
1310 	return arp_reachable;
1311 }
1312 
1313 static unsigned int
1314 arp_llinfo_retrans(__unused struct ifnet *ifp)
1315 {
1316 
1317 	return arp_retrans;
1318 }
1319 
1320 /*
1321  * Gets source address of the first packet in hold queue
1322  * and stores it in @src.
1323  * Returns pointer to @src (if hold queue is not empty) or NULL.
1324  */
1325 static union l3addr *
1326 arp_llinfo_holdsrc(struct llentry *ln, union l3addr *src)
1327 {
1328 	struct ip *ip;
1329 
1330 	if (ln == NULL || ln->ln_hold == NULL)
1331 		return NULL;
1332 
1333 	/*
1334 	 * assuming every packet in ln_hold has the same IP header
1335 	 */
1336 	ip = mtod(ln->ln_hold, struct ip *);
1337 	/* XXX pullup? */
1338 	if (sizeof(*ip) < ln->ln_hold->m_len)
1339 		src->addr4 = ip->ip_src;
1340 	else
1341 		src = NULL;
1342 
1343 	return src;
1344 }
1345 
1346 static void
1347 arp_llinfo_output(struct ifnet *ifp, __unused const union l3addr *daddr,
1348     const union l3addr *taddr, const uint8_t *tlladdr,
1349     const union l3addr *hsrc)
1350 {
1351 	struct in_addr tip = taddr->addr4, sip = zeroin_addr;
1352 	const uint8_t *slladdr = CLLADDR(ifp->if_sadl);
1353 
1354 	if (hsrc != NULL) {
1355 		struct in_ifaddr *ia;
1356 		struct psref psref;
1357 
1358 		ia = in_get_ia_on_iface_psref(hsrc->addr4, ifp, &psref);
1359 		if (ia != NULL) {
1360 			sip = hsrc->addr4;
1361 			ia4_release(ia, &psref);
1362 		}
1363 	}
1364 
1365 	if (sip.s_addr == INADDR_ANY) {
1366 		struct sockaddr_in dst;
1367 		struct rtentry *rt;
1368 
1369 		sockaddr_in_init(&dst, &tip, 0);
1370 		rt = rtalloc1(sintosa(&dst), 0);
1371 		if (rt != NULL) {
1372 			if (rt->rt_ifp == ifp &&
1373 			    rt->rt_ifa != NULL &&
1374 			    rt->rt_ifa->ifa_addr->sa_family == AF_INET)
1375 				sip = satosin(rt->rt_ifa->ifa_addr)->sin_addr;
1376 			rt_unref(rt);
1377 		}
1378 		if (sip.s_addr == INADDR_ANY) {
1379 			char ipbuf[INET_ADDRSTRLEN];
1380 
1381 			log(LOG_DEBUG, "%s: source can't be "
1382 			    "determined: dst=%s\n", __func__,
1383 			    IN_PRINT(ipbuf, &tip));
1384 			return;
1385 		}
1386 	}
1387 
1388 	arprequest(ifp, &sip, &tip, slladdr, tlladdr);
1389 }
1390 
1391 
1392 static void
1393 arp_llinfo_missed(struct ifnet *ifp, const union l3addr *taddr,
1394     __unused int16_t type, struct mbuf *m)
1395 {
1396 	struct in_addr mdaddr = zeroin_addr;
1397 	struct sockaddr_in dsin, tsin;
1398 	struct sockaddr *sa;
1399 
1400 	if (m != NULL) {
1401 		struct ip *ip = mtod(m, struct ip *);
1402 
1403 		if (sizeof(*ip) < m->m_len)
1404 			mdaddr = ip->ip_src;
1405 
1406 		/* ip_input() will send ICMP_UNREACH_HOST, not us. */
1407 		m_freem(m);
1408 	}
1409 
1410 	if (mdaddr.s_addr != INADDR_ANY) {
1411 		sockaddr_in_init(&dsin, &mdaddr, 0);
1412 		sa = sintosa(&dsin);
1413 	} else
1414 		sa = NULL;
1415 
1416 	sockaddr_in_init(&tsin, &taddr->addr4, 0);
1417 	rt_clonedmsg(RTM_MISS, sa, sintosa(&tsin), NULL, ifp);
1418 }
1419 
1420 static void
1421 arp_free(struct llentry *ln, int gc)
1422 {
1423 	struct ifnet *ifp;
1424 
1425 	KASSERT(ln != NULL);
1426 	LLE_WLOCK_ASSERT(ln);
1427 
1428 	ifp = ln->lle_tbl->llt_ifp;
1429 
1430 	if (ln->la_flags & LLE_VALID || gc) {
1431 		struct sockaddr_in sin;
1432 		const char *lladdr;
1433 
1434 		sockaddr_in_init(&sin, &ln->r_l3addr.addr4, 0);
1435 		lladdr = ln->la_flags & LLE_VALID ?
1436 		    (const char *)&ln->ll_addr : NULL;
1437 		rt_clonedmsg(RTM_DELETE, NULL, sintosa(&sin), lladdr, ifp);
1438 	}
1439 
1440 	/*
1441 	 * Save to unlock. We still hold an extra reference and will not
1442 	 * free(9) in llentry_free() if someone else holds one as well.
1443 	 */
1444 	LLE_WUNLOCK(ln);
1445 	IF_AFDATA_LOCK(ifp);
1446 	LLE_WLOCK(ln);
1447 
1448 	lltable_free_entry(LLTABLE(ifp), ln);
1449 
1450 	IF_AFDATA_UNLOCK(ifp);
1451 }
1452 
1453 /*
1454  * Upper-layer reachability hint for Neighbor Unreachability Detection.
1455  *
1456  * XXX cost-effective methods?
1457  */
1458 void
1459 arp_nud_hint(struct rtentry *rt)
1460 {
1461 	struct llentry *ln;
1462 	struct ifnet *ifp;
1463 
1464 	if (rt == NULL)
1465 		return;
1466 
1467 	ifp = rt->rt_ifp;
1468 	ln = arplookup(ifp, NULL, rt_getkey(rt), 1);
1469 	nd_nud_hint(ln);
1470 }
1471 
1472 TAILQ_HEAD(dadq_head, dadq);
1473 struct dadq {
1474 	TAILQ_ENTRY(dadq) dad_list;
1475 	struct ifaddr *dad_ifa;
1476 	int dad_count;		/* max ARP to send */
1477 	int dad_arp_tcount;	/* # of trials to send ARP */
1478 	int dad_arp_ocount;	/* ARP sent so far */
1479 	int dad_arp_announce;	/* max ARP announcements */
1480 	int dad_arp_acount;	/* # of announcements */
1481 	struct callout dad_timer_ch;
1482 };
1483 
1484 static struct dadq_head dadq;
1485 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1486 static kmutex_t arp_dad_lock;
1487 
1488 static void
1489 arp_dad_init(void)
1490 {
1491 
1492 	TAILQ_INIT(&dadq);
1493 	mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1494 }
1495 
1496 static struct dadq *
1497 arp_dad_find(struct ifaddr *ifa)
1498 {
1499 	struct dadq *dp;
1500 
1501 	KASSERT(mutex_owned(&arp_dad_lock));
1502 
1503 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1504 		if (dp->dad_ifa == ifa)
1505 			return dp;
1506 	}
1507 	return NULL;
1508 }
1509 
1510 static void
1511 arp_dad_starttimer(struct dadq *dp, int ticks)
1512 {
1513 
1514 	callout_reset(&dp->dad_timer_ch, ticks,
1515 	    (void (*)(void *))arp_dad_timer, dp);
1516 }
1517 
1518 static void
1519 arp_dad_stoptimer(struct dadq *dp)
1520 {
1521 
1522 	KASSERT(mutex_owned(&arp_dad_lock));
1523 
1524 	TAILQ_REMOVE(&dadq, dp, dad_list);
1525 	/* Tell the timer that dp is being destroyed. */
1526 	dp->dad_ifa = NULL;
1527 	callout_halt(&dp->dad_timer_ch, &arp_dad_lock);
1528 }
1529 
1530 static void
1531 arp_dad_destroytimer(struct dadq *dp)
1532 {
1533 
1534 	callout_destroy(&dp->dad_timer_ch);
1535 	KASSERT(dp->dad_ifa == NULL);
1536 	kmem_intr_free(dp, sizeof(*dp));
1537 }
1538 
1539 static void
1540 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1541 {
1542 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1543 	struct ifnet *ifp = ifa->ifa_ifp;
1544 	struct in_addr sip;
1545 
1546 	dp->dad_arp_tcount++;
1547 	if ((ifp->if_flags & IFF_UP) == 0)
1548 		return;
1549 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1550 		return;
1551 
1552 	dp->dad_arp_tcount = 0;
1553 	dp->dad_arp_ocount++;
1554 
1555 	memset(&sip, 0, sizeof(sip));
1556 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1557 	    CLLADDR(ifa->ifa_ifp->if_sadl), NULL);
1558 }
1559 
1560 /*
1561  * Start Duplicate Address Detection (DAD) for specified interface address.
1562  */
1563 static void
1564 arp_dad_start(struct ifaddr *ifa)
1565 {
1566 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1567 	struct dadq *dp;
1568 	char ipbuf[INET_ADDRSTRLEN];
1569 
1570 	/*
1571 	 * If we don't need DAD, don't do it.
1572 	 * - DAD is disabled
1573 	 */
1574 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1575 		log(LOG_DEBUG,
1576 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1577 		    IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1578 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1579 		return;
1580 	}
1581 	if (!ip_dad_enabled()) {
1582 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1583 		rt_addrmsg(RTM_NEWADDR, ifa);
1584 		arpannounce1(ifa);
1585 		return;
1586 	}
1587 	KASSERT(ifa->ifa_ifp != NULL);
1588 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1589 		return;
1590 
1591 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1592 
1593 	mutex_enter(&arp_dad_lock);
1594 	if (arp_dad_find(ifa) != NULL) {
1595 		mutex_exit(&arp_dad_lock);
1596 		/* DAD already in progress */
1597 		if (dp != NULL)
1598 			kmem_intr_free(dp, sizeof(*dp));
1599 		return;
1600 	}
1601 
1602 	if (dp == NULL) {
1603 		mutex_exit(&arp_dad_lock);
1604 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1605 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1606 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1607 		return;
1608 	}
1609 
1610 	/*
1611 	 * Send ARP packet for DAD, ip_dad_count times.
1612 	 * Note that we must delay the first transmission.
1613 	 */
1614 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1615 	dp->dad_ifa = ifa;
1616 	ifaref(ifa);	/* just for safety */
1617 	dp->dad_count = ip_dad_count;
1618 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1619 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1620 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1621 
1622 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1623 	    ARPLOGADDR(&ia->ia_addr.sin_addr));
1624 
1625 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1626 
1627 	mutex_exit(&arp_dad_lock);
1628 }
1629 
1630 /*
1631  * terminate DAD unconditionally.  used for address removals.
1632  */
1633 static void
1634 arp_dad_stop(struct ifaddr *ifa)
1635 {
1636 	struct dadq *dp;
1637 
1638 	mutex_enter(&arp_dad_lock);
1639 	dp = arp_dad_find(ifa);
1640 	if (dp == NULL) {
1641 		mutex_exit(&arp_dad_lock);
1642 		/* DAD wasn't started yet */
1643 		return;
1644 	}
1645 
1646 	arp_dad_stoptimer(dp);
1647 
1648 	mutex_exit(&arp_dad_lock);
1649 
1650 	arp_dad_destroytimer(dp);
1651 	ifafree(ifa);
1652 }
1653 
1654 static void
1655 arp_dad_timer(struct dadq *dp)
1656 {
1657 	struct ifaddr *ifa;
1658 	struct in_ifaddr *ia;
1659 	char ipbuf[INET_ADDRSTRLEN];
1660 	bool need_free = false;
1661 
1662 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1663 	mutex_enter(&arp_dad_lock);
1664 
1665 	ifa = dp->dad_ifa;
1666 	if (ifa == NULL) {
1667 		/* dp is being destroyed by someone.  Do nothing. */
1668 		goto done;
1669 	}
1670 
1671 	ia = (struct in_ifaddr *)ifa;
1672 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1673 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1674 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1675 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1676 		goto done;
1677 	}
1678 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1679 	{
1680 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1681 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1682 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1683 		goto done;
1684 	}
1685 
1686 	/* timeouted with IFF_{RUNNING,UP} check */
1687 	if (dp->dad_arp_tcount > dad_maxtry) {
1688 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1689 		    if_name(ifa->ifa_ifp));
1690 
1691 		arp_dad_stoptimer(dp);
1692 		need_free = true;
1693 		goto done;
1694 	}
1695 
1696 	/* Need more checks? */
1697 	if (dp->dad_arp_ocount < dp->dad_count) {
1698 		int adelay;
1699 
1700 		/*
1701 		 * We have more ARP to go.  Send ARP packet for DAD.
1702 		 */
1703 		arp_dad_output(dp, ifa);
1704 		if (dp->dad_arp_ocount < dp->dad_count)
1705 			adelay = (PROBE_MIN * hz) +
1706 			    (cprng_fast32() %
1707 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1708 		else
1709 			adelay = ANNOUNCE_WAIT * hz;
1710 		arp_dad_starttimer(dp, adelay);
1711 		goto done;
1712 	} else if (dp->dad_arp_acount == 0) {
1713 		/*
1714 		 * We are done with DAD.
1715 		 * No duplicate address found.
1716 		 */
1717 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1718 		rt_addrmsg(RTM_NEWADDR, ifa);
1719 		ARPLOG(LOG_DEBUG,
1720 		    "%s: DAD complete for %s - no duplicates found\n",
1721 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1722 		dp->dad_arp_announce = ANNOUNCE_NUM;
1723 		goto announce;
1724 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1725 announce:
1726 		/*
1727 		 * Announce the address.
1728 		 */
1729 		arpannounce1(ifa);
1730 		dp->dad_arp_acount++;
1731 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1732 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1733 			goto done;
1734 		}
1735 		ARPLOG(LOG_DEBUG,
1736 		    "%s: ARP announcement complete for %s\n",
1737 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1738 	}
1739 
1740 	arp_dad_stoptimer(dp);
1741 	need_free = true;
1742 done:
1743 	mutex_exit(&arp_dad_lock);
1744 
1745 	if (need_free) {
1746 		arp_dad_destroytimer(dp);
1747 		KASSERT(ifa != NULL);
1748 		ifafree(ifa);
1749 	}
1750 
1751 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1752 }
1753 
1754 static void
1755 arp_dad_duplicated(struct ifaddr *ifa, const struct sockaddr_dl *from)
1756 {
1757 	struct in_ifaddr *ia = ifatoia(ifa);
1758 	struct ifnet *ifp = ifa->ifa_ifp;
1759 	char ipbuf[INET_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN];
1760 	const char *iastr, *llastr;
1761 
1762 	iastr = IN_PRINT(ipbuf, &ia->ia_addr.sin_addr);
1763 	if (__predict_false(from == NULL))
1764 		llastr = NULL;
1765 	else
1766 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
1767 		    CLLADDR(from), from->sdl_alen);
1768 
1769 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1770 		log(LOG_ERR,
1771 		    "%s: DAD duplicate address %s from %s\n",
1772 		    if_name(ifp), iastr, llastr);
1773 	} else if (ia->ia_dad_defended == 0 ||
1774 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1775 		ia->ia_dad_defended = time_uptime;
1776 		arpannounce1(ifa);
1777 		log(LOG_ERR,
1778 		    "%s: DAD defended address %s from %s\n",
1779 		    if_name(ifp), iastr, llastr);
1780 		return;
1781 	} else {
1782 		/* If DAD is disabled, just report the duplicate. */
1783 		if (!ip_dad_enabled()) {
1784 			log(LOG_ERR,
1785 			    "%s: DAD ignoring duplicate address %s from %s\n",
1786 			    if_name(ifp), iastr, llastr);
1787 			return;
1788 		}
1789 		log(LOG_ERR,
1790 		    "%s: DAD defence failed for %s from %s\n",
1791 		    if_name(ifp), iastr, llastr);
1792 	}
1793 
1794 	arp_dad_stop(ifa);
1795 
1796 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1797 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1798 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1799 		/* Inform the routing socket of the duplicate address */
1800 		rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1801 	}
1802 }
1803 
1804 /*
1805  * Called from 10 Mb/s Ethernet interrupt handlers
1806  * when ether packet type ETHERTYPE_REVARP
1807  * is received.  Common length and type checks are done here,
1808  * then the protocol-specific routine is called.
1809  */
1810 void
1811 revarpinput(struct mbuf *m)
1812 {
1813 	struct arphdr *ar;
1814 	int arplen;
1815 
1816 	arplen = sizeof(struct arphdr);
1817 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1818 		return;
1819 	ar = mtod(m, struct arphdr *);
1820 
1821 	if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394) {
1822 		goto out;
1823 	}
1824 
1825 	arplen = sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln);
1826 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1827 		return;
1828 	ar = mtod(m, struct arphdr *);
1829 
1830 	switch (ntohs(ar->ar_pro)) {
1831 	case ETHERTYPE_IP:
1832 	case ETHERTYPE_IPTRAILERS:
1833 		in_revarpinput(m);
1834 		return;
1835 
1836 	default:
1837 		break;
1838 	}
1839 
1840 out:
1841 	m_freem(m);
1842 }
1843 
1844 /*
1845  * RARP for Internet protocols on 10 Mb/s Ethernet.
1846  * Algorithm is that given in RFC 903.
1847  * We are only using for bootstrap purposes to get an ip address for one of
1848  * our interfaces.  Thus we support no user-interface.
1849  *
1850  * Since the contents of the RARP reply are specific to the interface that
1851  * sent the request, this code must ensure that they are properly associated.
1852  *
1853  * Note: also supports ARP via RARP packets, per the RFC.
1854  */
1855 void
1856 in_revarpinput(struct mbuf *m)
1857 {
1858 	struct arphdr *ah;
1859 	void *tha;
1860 	int op;
1861 	struct ifnet *rcvif;
1862 	int s;
1863 
1864 	ah = mtod(m, struct arphdr *);
1865 	op = ntohs(ah->ar_op);
1866 
1867 	rcvif = m_get_rcvif(m, &s);
1868 	if (__predict_false(rcvif == NULL))
1869 		goto out;
1870 	if (rcvif->if_flags & IFF_NOARP)
1871 		goto out;
1872 
1873 	switch (rcvif->if_type) {
1874 	case IFT_IEEE1394:
1875 		/* ARP without target hardware address is not supported */
1876 		goto out;
1877 	default:
1878 		break;
1879 	}
1880 
1881 	switch (op) {
1882 	case ARPOP_REQUEST:
1883 	case ARPOP_REPLY:	/* per RFC */
1884 		m_put_rcvif(rcvif, &s);
1885 		in_arpinput(m);
1886 		return;
1887 	case ARPOP_REVREPLY:
1888 		break;
1889 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1890 	default:
1891 		goto out;
1892 	}
1893 	if (!revarp_in_progress)
1894 		goto out;
1895 	if (rcvif != myip_ifp) /* !same interface */
1896 		goto out;
1897 	if (myip_initialized)
1898 		goto wake;
1899 	tha = ar_tha(ah);
1900 	if (tha == NULL)
1901 		goto out;
1902 	if (ah->ar_pln != sizeof(struct in_addr))
1903 		goto out;
1904 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1905 		goto out;
1906 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1907 		goto out;
1908 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1909 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1910 	myip_initialized = 1;
1911 wake:	/* Do wakeup every time in case it was missed. */
1912 	wakeup((void *)&myip);
1913 
1914 out:
1915 	m_put_rcvif(rcvif, &s);
1916 	m_freem(m);
1917 }
1918 
1919 /*
1920  * Send a RARP request for the ip address of the specified interface.
1921  * The request should be RFC 903-compliant.
1922  */
1923 static void
1924 revarprequest(struct ifnet *ifp)
1925 {
1926 	struct sockaddr sa;
1927 	struct mbuf *m;
1928 	struct arphdr *ah;
1929 	void *tha;
1930 
1931 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1932 		return;
1933 	MCLAIM(m, &arpdomain.dom_mowner);
1934 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1935 	    2*ifp->if_addrlen;
1936 	m->m_pkthdr.len = m->m_len;
1937 	m_align(m, m->m_len);
1938 	ah = mtod(m, struct arphdr *);
1939 	memset(ah, 0, m->m_len);
1940 	ah->ar_pro = htons(ETHERTYPE_IP);
1941 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1942 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1943 	ah->ar_op = htons(ARPOP_REVREQUEST);
1944 
1945 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1946 	tha = ar_tha(ah);
1947 	if (tha == NULL) {
1948 		m_free(m);
1949 		return;
1950 	}
1951 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1952 
1953 	sa.sa_family = AF_ARP;
1954 	sa.sa_len = 2;
1955 	m->m_flags |= M_BCAST;
1956 
1957 	if_output_lock(ifp, ifp, m, &sa, NULL);
1958 }
1959 
1960 /*
1961  * RARP for the ip address of the specified interface, but also
1962  * save the ip address of the server that sent the answer.
1963  * Timeout if no response is received.
1964  */
1965 int
1966 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1967     struct in_addr *clnt_in)
1968 {
1969 	int result, count = 20;
1970 
1971 	myip_initialized = 0;
1972 	myip_ifp = ifp;
1973 
1974 	revarp_in_progress = 1;
1975 	while (count--) {
1976 		revarprequest(ifp);
1977 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1978 		if (result != EWOULDBLOCK)
1979 			break;
1980 	}
1981 	revarp_in_progress = 0;
1982 
1983 	if (!myip_initialized)
1984 		return ENETUNREACH;
1985 
1986 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1987 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1988 	return 0;
1989 }
1990 
1991 void
1992 arp_stat_add(int type, uint64_t count)
1993 {
1994 	ARP_STATADD(type, count);
1995 }
1996 
1997 static int
1998 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1999 {
2000 
2001 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2002 }
2003 
2004 static void
2005 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2006 {
2007 	const struct sysctlnode *node;
2008 
2009 	sysctl_createv(clog, 0, NULL, NULL,
2010 			CTLFLAG_PERMANENT,
2011 			CTLTYPE_NODE, "inet", NULL,
2012 			NULL, 0, NULL, 0,
2013 			CTL_NET, PF_INET, CTL_EOL);
2014 	sysctl_createv(clog, 0, NULL, &node,
2015 			CTLFLAG_PERMANENT,
2016 			CTLTYPE_NODE, "arp",
2017 			SYSCTL_DESCR("Address Resolution Protocol"),
2018 			NULL, 0, NULL, 0,
2019 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2020 
2021 	sysctl_createv(clog, 0, NULL, NULL,
2022 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2023 		       CTLTYPE_INT, "nd_delay",
2024 		       SYSCTL_DESCR("First probe delay time"),
2025 		       NULL, 0, &arp_nd_domain.nd_delay, 0,
2026 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2027 	sysctl_createv(clog, 0, NULL, NULL,
2028 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2029 		       CTLTYPE_INT, "nd_bmaxtries",
2030 		       SYSCTL_DESCR("Number of broadcast discovery attempts"),
2031 		       NULL, 0, &arp_nd_domain.nd_mmaxtries, 0,
2032 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2033 	sysctl_createv(clog, 0, NULL, NULL,
2034 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2035 		       CTLTYPE_INT, "nd_umaxtries",
2036 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2037 		       NULL, 0, &arp_nd_domain.nd_umaxtries, 0,
2038 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2039 	sysctl_createv(clog, 0, NULL, NULL,
2040 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2041 		       CTLTYPE_INT, "nd_reachable",
2042 		       SYSCTL_DESCR("Reachable time"),
2043 		       NULL, 0, &arp_reachable, 0,
2044 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2045 	sysctl_createv(clog, 0, NULL, NULL,
2046 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2047 		       CTLTYPE_INT, "nd_retrans",
2048 		       SYSCTL_DESCR("Retransmission time"),
2049 		       NULL, 0, &arp_retrans, 0,
2050 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2051 	sysctl_createv(clog, 0, NULL, NULL,
2052 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2053 		       CTLTYPE_INT, "nd_nud",
2054 		       SYSCTL_DESCR("Perform neighbour unreachability detection"),
2055 		       NULL, 0, &arp_perform_nud, 0,
2056 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2057 	sysctl_createv(clog, 0, NULL, NULL,
2058 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2059 		       CTLTYPE_INT, "nd_maxnudhint",
2060 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2061 		       NULL, 0, &arp_nd_domain.nd_maxnudhint, 0,
2062 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2063 	sysctl_createv(clog, 0, NULL, NULL,
2064 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2065 		       CTLTYPE_INT, "maxqueuelen",
2066 		       SYSCTL_DESCR("max packet queue len for a unresolved ARP"),
2067 		       NULL, 1, &arp_nd_domain.nd_maxqueuelen, 0,
2068 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2069 
2070 	sysctl_createv(clog, 0, NULL, NULL,
2071 			CTLFLAG_PERMANENT,
2072 			CTLTYPE_STRUCT, "stats",
2073 			SYSCTL_DESCR("ARP statistics"),
2074 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2075 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2076 
2077 	sysctl_createv(clog, 0, NULL, NULL,
2078 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2079 			CTLTYPE_INT, "log_movements",
2080 			SYSCTL_DESCR("log ARP replies from MACs different than"
2081 			    " the one in the cache"),
2082 			NULL, 0, &log_movements, 0,
2083 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2084 
2085 	sysctl_createv(clog, 0, NULL, NULL,
2086 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2087 			CTLTYPE_INT, "log_permanent_modify",
2088 			SYSCTL_DESCR("log ARP replies from MACs different than"
2089 			    " the one in the permanent arp entry"),
2090 			NULL, 0, &log_permanent_modify, 0,
2091 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2092 
2093 	sysctl_createv(clog, 0, NULL, NULL,
2094 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2095 			CTLTYPE_INT, "log_wrong_iface",
2096 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2097 			    " interface"),
2098 			NULL, 0, &log_wrong_iface, 0,
2099 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2100 
2101 	sysctl_createv(clog, 0, NULL, NULL,
2102 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2103 			CTLTYPE_INT, "debug",
2104 			SYSCTL_DESCR("Enable ARP DAD debug output"),
2105 			NULL, 0, &arp_debug, 0,
2106 			CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2107 }
2108 
2109 #endif /* INET */
2110