xref: /netbsd-src/sys/netinet/if_arp.c (revision aef5eb5f59cdfe8314f1b5f78ac04eb144e44010)
1 /*	$NetBSD: if_arp.c,v 1.308 2022/09/03 01:35:03 thorpej 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 1.308 2022/09/03 01:35:03 thorpej 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 	uint64_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 	arps[ARP_STAT_SNDTOTAL]++;
565 	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 	uint64_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 
930 	if (in_nullhost(itaddr))
931 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
932 
933 	/*
934 	 * DAD check, RFC 5227.
935 	 * Collision on sender address is always a duplicate.
936 	 * Collision on target address is only a duplicate
937 	 * IF the sender address is the null host (ie a DAD probe)
938 	 * AND the message was broadcast
939 	 * AND our address is either tentative or duplicated
940 	 * If it was unicast then it's a valid Unicast Poll from RFC 1122.
941 	 */
942 	if (do_dad &&
943 	    (in_hosteq(isaddr, myaddr) ||
944 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr) &&
945 	     m->m_flags & M_BCAST &&
946 	     ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DUPLICATED))))
947 	{
948 		struct sockaddr_dl sdl, *sdlp;
949 
950 		sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
951 		    ifp->if_index, ifp->if_type,
952 		    NULL, 0, ar_sha(ah), ah->ar_hln);
953 		arp_dad_duplicated((struct ifaddr *)ia, sdlp);
954 		goto out;
955 	}
956 
957 	/*
958 	 * If the target IP address is zero, ignore the packet.
959 	 * This prevents the code below from trying to answer
960 	 * when we are using IP address zero (booting).
961 	 */
962 	if (in_nullhost(itaddr))
963 		goto out;
964 
965 	if (in_nullhost(isaddr))
966 		goto reply;
967 
968 	if (in_hosteq(itaddr, myaddr))
969 		la = arpcreate(ifp, &isaddr, NULL, 1);
970 	else
971 		la = arplookup(ifp, &isaddr, NULL, 1);
972 	if (la == NULL)
973 		goto reply;
974 
975 	if ((la->la_flags & LLE_VALID) &&
976 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen))
977 	{
978 		char llabuf[LLA_ADDRSTRLEN], *llastr;
979 
980 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
981 		    ar_sha(ah), ah->ar_hln);
982 
983 		if (la->la_flags & LLE_STATIC) {
984 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
985 			if (!log_permanent_modify)
986 				goto out;
987 			log(LOG_INFO,
988 			    "%s tried to overwrite permanent arp info"
989 			    " for %s\n", llastr, IN_PRINT(ipbuf, &isaddr));
990 			goto out;
991 		} else if (la->lle_tbl->llt_ifp != ifp) {
992 			/* XXX should not happen? */
993 			ARP_STATINC(ARP_STAT_RCVOVERINT);
994 			if (!log_wrong_iface)
995 				goto out;
996 			log(LOG_INFO,
997 			    "%s on %s tried to overwrite "
998 			    "arp info for %s on %s\n",
999 			    llastr,
1000 			    ifp->if_xname, IN_PRINT(ipbuf, &isaddr),
1001 			    la->lle_tbl->llt_ifp->if_xname);
1002 				goto out;
1003 		} else {
1004 			ARP_STATINC(ARP_STAT_RCVOVER);
1005 			if (log_movements)
1006 				log(LOG_INFO, "arp info overwritten "
1007 				    "for %s by %s\n",
1008 				    IN_PRINT(ipbuf, &isaddr), llastr);
1009 		}
1010 		rt_cmd = RTM_CHANGE;
1011 		new_state = ND_LLINFO_STALE;
1012 	} else {
1013 		if (op == ARPOP_REPLY && in_hosteq(itaddr, myaddr)) {
1014 			/* This was a solicited ARP reply. */
1015 			la->ln_byhint = 0;
1016 			new_state = ND_LLINFO_REACHABLE;
1017 		}
1018 		rt_cmd = la->la_flags & LLE_VALID ? 0 : RTM_ADD;
1019 	}
1020 
1021 	KASSERT(ifp->if_sadl->sdl_alen == ifp->if_addrlen);
1022 
1023 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1024 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1025 	la->la_flags |= LLE_VALID;
1026 	la->ln_asked = 0;
1027 	if (new_state != 0) {
1028 		la->ln_state = new_state;
1029 
1030 		if (new_state != ND_LLINFO_REACHABLE ||
1031 		    !(la->la_flags & LLE_STATIC))
1032 		{
1033 			int timer = ND_TIMER_GC;
1034 
1035 			if (new_state == ND_LLINFO_REACHABLE)
1036 				timer = ND_TIMER_REACHABLE;
1037 			nd_set_timer(la, timer);
1038 		}
1039 	}
1040 
1041 	if (rt_cmd != 0) {
1042 		struct sockaddr_in sin;
1043 
1044 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1045 		rt_clonedmsg(rt_cmd, NULL, sintosa(&sin), ar_sha(ah), ifp);
1046 	}
1047 
1048 	if (la->la_hold != NULL) {
1049 		int n = la->la_numheld;
1050 		struct mbuf *m_hold, *m_hold_next;
1051 		struct sockaddr_in sin;
1052 
1053 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1054 
1055 		m_hold = la->la_hold;
1056 		la->la_hold = NULL;
1057 		la->la_numheld = 0;
1058 		/*
1059 		 * We have to unlock here because if_output would call
1060 		 * arpresolve
1061 		 */
1062 		LLE_WUNLOCK(la);
1063 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1064 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1065 		for (; m_hold != NULL; m_hold = m_hold_next) {
1066 			m_hold_next = m_hold->m_nextpkt;
1067 			m_hold->m_nextpkt = NULL;
1068 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1069 		}
1070 	} else
1071 		LLE_WUNLOCK(la);
1072 	la = NULL;
1073 
1074 reply:
1075 	if (la != NULL) {
1076 		LLE_WUNLOCK(la);
1077 		la = NULL;
1078 	}
1079 	if (op != ARPOP_REQUEST) {
1080 		if (op == ARPOP_REPLY)
1081 			ARP_STATINC(ARP_STAT_RCVREPLY);
1082 		goto out;
1083 	}
1084 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1085 	if (in_hosteq(itaddr, myaddr)) {
1086 		/* If our address is unusable, don't reply */
1087 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1088 			goto out;
1089 		/* I am the target */
1090 		tha = ar_tha(ah);
1091 		if (tha)
1092 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1093 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1094 	} else {
1095 		/* Proxy ARP */
1096 		struct llentry *lle = NULL;
1097 		struct sockaddr_in sin;
1098 
1099 #if NCARP > 0
1100 		if (ifp->if_type == IFT_CARP) {
1101 			struct ifnet *_rcvif = m_get_rcvif(m, &s);
1102 			int iftype = 0;
1103 			if (__predict_true(_rcvif != NULL))
1104 				iftype = _rcvif->if_type;
1105 			m_put_rcvif(_rcvif, &s);
1106 			if (iftype != IFT_CARP)
1107 				goto out;
1108 		}
1109 #endif
1110 
1111 		tha = ar_tha(ah);
1112 
1113 		sockaddr_in_init(&sin, &itaddr, 0);
1114 
1115 		IF_AFDATA_RLOCK(ifp);
1116 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1117 		IF_AFDATA_RUNLOCK(ifp);
1118 
1119 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1120 			if (tha)
1121 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1122 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1123 			LLE_RUNLOCK(lle);
1124 		} else {
1125 			if (lle != NULL)
1126 				LLE_RUNLOCK(lle);
1127 			goto out;
1128 		}
1129 	}
1130 	ia4_release(ia, &psref_ia);
1131 
1132 	/*
1133 	 * XXX XXX: Here we're recycling the mbuf. But the mbuf could have
1134 	 * other mbufs in its chain, and just overwriting m->m_pkthdr.len
1135 	 * would be wrong in this case (the length becomes smaller than the
1136 	 * real chain size).
1137 	 *
1138 	 * This can theoretically cause bugs in the lower layers (drivers,
1139 	 * and L2encap), in some corner cases.
1140 	 */
1141 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1142 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1143 	ah->ar_op = htons(ARPOP_REPLY);
1144 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1145 	switch (ifp->if_type) {
1146 	case IFT_IEEE1394:
1147 		/* ieee1394 arp reply is broadcast */
1148 		m->m_flags &= ~M_MCAST;
1149 		m->m_flags |= M_BCAST;
1150 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1151 		break;
1152 	default:
1153 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1154 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1155 		break;
1156 	}
1157 	m->m_pkthdr.len = m->m_len;
1158 	sa.sa_family = AF_ARP;
1159 	sa.sa_len = 2;
1160 	arps = ARP_STAT_GETREF();
1161 	arps[ARP_STAT_SNDTOTAL]++;
1162 	arps[ARP_STAT_SNDREPLY]++;
1163 	ARP_STAT_PUTREF();
1164 	if_output_lock(ifp, ifp, m, &sa, NULL);
1165 	if (rcvif != NULL)
1166 		m_put_rcvif_psref(rcvif, &psref);
1167 	return;
1168 
1169 out:
1170 	if (la != NULL)
1171 		LLE_WUNLOCK(la);
1172 	if (ia != NULL)
1173 		ia4_release(ia, &psref_ia);
1174 	if (rcvif != NULL)
1175 		m_put_rcvif_psref(rcvif, &psref);
1176 	m_freem(m);
1177 }
1178 
1179 /*
1180  * Lookup or a new address in arptab.
1181  */
1182 struct llentry *
1183 arplookup(struct ifnet *ifp, const struct in_addr *addr,
1184     const struct sockaddr *sa, int wlock)
1185 {
1186 	struct sockaddr_in sin;
1187 	struct llentry *la;
1188 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1189 
1190 	if (sa == NULL) {
1191 		KASSERT(addr != NULL);
1192 		sockaddr_in_init(&sin, addr, 0);
1193 		sa = sintocsa(&sin);
1194 	}
1195 
1196 	IF_AFDATA_RLOCK(ifp);
1197 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1198 	IF_AFDATA_RUNLOCK(ifp);
1199 
1200 	return la;
1201 }
1202 
1203 static struct llentry *
1204 arpcreate(struct ifnet *ifp, const struct in_addr *addr,
1205     const struct sockaddr *sa, int wlock)
1206 {
1207 	struct sockaddr_in sin;
1208 	struct llentry *la;
1209 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1210 
1211 	if (sa == NULL) {
1212 		KASSERT(addr != NULL);
1213 		sockaddr_in_init(&sin, addr, 0);
1214 		sa = sintocsa(&sin);
1215 	}
1216 
1217 	la = arplookup(ifp, addr, sa, wlock);
1218 
1219 	if (la == NULL) {
1220 		struct rtentry *rt;
1221 
1222 		rt = rtalloc1(sa, 0);
1223 		IF_AFDATA_WLOCK(ifp);
1224 		la = lla_create(LLTABLE(ifp), flags, sa, rt);
1225 		IF_AFDATA_WUNLOCK(ifp);
1226 		if (rt != NULL)
1227 			rt_unref(rt);
1228 
1229 		if (la != NULL)
1230 			la->ln_state = ND_LLINFO_NOSTATE;
1231 	}
1232 
1233 	return la;
1234 }
1235 
1236 int
1237 arpioctl(u_long cmd, void *data)
1238 {
1239 
1240 	return EOPNOTSUPP;
1241 }
1242 
1243 void
1244 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1245 {
1246 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1247 
1248 	ifa->ifa_rtrequest = arp_rtrequest;
1249 	ifa->ifa_flags |= RTF_CONNECTED;
1250 
1251 	/* ARP will handle DAD for this address. */
1252 	if (in_nullhost(IA_SIN(ifa)->sin_addr)) {
1253 		if (ia->ia_dad_stop != NULL)	/* safety */
1254 			ia->ia_dad_stop(ifa);
1255 		ia->ia_dad_start = NULL;
1256 		ia->ia_dad_stop = NULL;
1257 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1258 	} else {
1259 		ia->ia_dad_start = arp_dad_start;
1260 		ia->ia_dad_stop = arp_dad_stop;
1261 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE && ip_dad_enabled())
1262 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1263 		else
1264 			arpannounce1(ifa);
1265 	}
1266 }
1267 
1268 static bool
1269 arp_nud_enabled(__unused struct ifnet *ifp)
1270 {
1271 
1272 	return arp_perform_nud != 0;
1273 }
1274 
1275 static unsigned int
1276 arp_llinfo_reachable(__unused struct ifnet *ifp)
1277 {
1278 
1279 	return arp_reachable;
1280 }
1281 
1282 static unsigned int
1283 arp_llinfo_retrans(__unused struct ifnet *ifp)
1284 {
1285 
1286 	return arp_retrans;
1287 }
1288 
1289 /*
1290  * Gets source address of the first packet in hold queue
1291  * and stores it in @src.
1292  * Returns pointer to @src (if hold queue is not empty) or NULL.
1293  */
1294 static union l3addr *
1295 arp_llinfo_holdsrc(struct llentry *ln, union l3addr *src)
1296 {
1297 	struct ip *ip;
1298 
1299 	if (ln == NULL || ln->ln_hold == NULL)
1300 		return NULL;
1301 
1302 	/*
1303 	 * assuming every packet in ln_hold has the same IP header
1304 	 */
1305 	ip = mtod(ln->ln_hold, struct ip *);
1306 	/* XXX pullup? */
1307 	if (sizeof(*ip) < ln->ln_hold->m_len)
1308 		src->addr4 = ip->ip_src;
1309 	else
1310 		src = NULL;
1311 
1312 	return src;
1313 }
1314 
1315 static void
1316 arp_llinfo_output(struct ifnet *ifp, __unused const union l3addr *daddr,
1317     const union l3addr *taddr, const uint8_t *tlladdr,
1318     const union l3addr *hsrc)
1319 {
1320 	struct in_addr tip = taddr->addr4, sip = zeroin_addr;
1321 	const uint8_t *slladdr = CLLADDR(ifp->if_sadl);
1322 
1323 	if (hsrc != NULL) {
1324 		struct in_ifaddr *ia;
1325 		struct psref psref;
1326 
1327 		ia = in_get_ia_on_iface_psref(hsrc->addr4, ifp, &psref);
1328 		if (ia != NULL) {
1329 			sip = hsrc->addr4;
1330 			ia4_release(ia, &psref);
1331 		}
1332 	}
1333 
1334 	if (sip.s_addr == INADDR_ANY) {
1335 		struct sockaddr_in dst;
1336 		struct rtentry *rt;
1337 
1338 		sockaddr_in_init(&dst, &tip, 0);
1339 		rt = rtalloc1(sintosa(&dst), 0);
1340 		if (rt != NULL) {
1341 			if (rt->rt_ifp == ifp &&
1342 			    rt->rt_ifa != NULL &&
1343 			    rt->rt_ifa->ifa_addr->sa_family == AF_INET)
1344 				sip = satosin(rt->rt_ifa->ifa_addr)->sin_addr;
1345 			rt_unref(rt);
1346 		}
1347 		if (sip.s_addr == INADDR_ANY) {
1348 			char ipbuf[INET_ADDRSTRLEN];
1349 
1350 			log(LOG_DEBUG, "source can't be "
1351 			    "determined: dst=%s\n",
1352 			    IN_PRINT(ipbuf, &tip));
1353 			return;
1354 		}
1355 	}
1356 
1357 	arprequest(ifp, &sip, &tip, slladdr, tlladdr);
1358 }
1359 
1360 
1361 static void
1362 arp_llinfo_missed(struct ifnet *ifp, const union l3addr *taddr,
1363     __unused int16_t type, struct mbuf *m)
1364 {
1365 	struct in_addr mdaddr = zeroin_addr;
1366 	struct sockaddr_in dsin, tsin;
1367 	struct sockaddr *sa;
1368 
1369 	if (m != NULL) {
1370 		struct ip *ip = mtod(m, struct ip *);
1371 
1372 		if (sizeof(*ip) < m->m_len)
1373 			mdaddr = ip->ip_src;
1374 
1375 		/* ip_input() will send ICMP_UNREACH_HOST, not us. */
1376 		m_freem(m);
1377 	}
1378 
1379 	if (mdaddr.s_addr != INADDR_ANY) {
1380 		sockaddr_in_init(&dsin, &mdaddr, 0);
1381 		sa = sintosa(&dsin);
1382 	} else
1383 		sa = NULL;
1384 
1385 	sockaddr_in_init(&tsin, &taddr->addr4, 0);
1386 	rt_clonedmsg(RTM_MISS, sa, sintosa(&tsin), NULL, ifp);
1387 }
1388 
1389 static void
1390 arp_free(struct llentry *ln, int gc)
1391 {
1392 	struct ifnet *ifp;
1393 
1394 	KASSERT(ln != NULL);
1395 	LLE_WLOCK_ASSERT(ln);
1396 
1397 	ifp = ln->lle_tbl->llt_ifp;
1398 
1399 	if (ln->la_flags & LLE_VALID || gc) {
1400 		struct sockaddr_in sin;
1401 		const char *lladdr;
1402 
1403 		sockaddr_in_init(&sin, &ln->r_l3addr.addr4, 0);
1404 		lladdr = ln->la_flags & LLE_VALID ?
1405 		    (const char *)&ln->ll_addr : NULL;
1406 		rt_clonedmsg(RTM_DELETE, NULL, sintosa(&sin), lladdr, ifp);
1407 	}
1408 
1409 	/*
1410 	 * Save to unlock. We still hold an extra reference and will not
1411 	 * free(9) in llentry_free() if someone else holds one as well.
1412 	 */
1413 	LLE_WUNLOCK(ln);
1414 	IF_AFDATA_LOCK(ifp);
1415 	LLE_WLOCK(ln);
1416 
1417 	lltable_free_entry(LLTABLE(ifp), ln);
1418 
1419 	IF_AFDATA_UNLOCK(ifp);
1420 }
1421 
1422 /*
1423  * Upper-layer reachability hint for Neighbor Unreachability Detection.
1424  *
1425  * XXX cost-effective methods?
1426  */
1427 void
1428 arp_nud_hint(struct rtentry *rt)
1429 {
1430 	struct llentry *ln;
1431 	struct ifnet *ifp;
1432 
1433 	if (rt == NULL)
1434 		return;
1435 
1436 	ifp = rt->rt_ifp;
1437 	ln = arplookup(ifp, NULL, rt_getkey(rt), 1);
1438 	nd_nud_hint(ln);
1439 }
1440 
1441 TAILQ_HEAD(dadq_head, dadq);
1442 struct dadq {
1443 	TAILQ_ENTRY(dadq) dad_list;
1444 	struct ifaddr *dad_ifa;
1445 	int dad_count;		/* max ARP to send */
1446 	int dad_arp_tcount;	/* # of trials to send ARP */
1447 	int dad_arp_ocount;	/* ARP sent so far */
1448 	int dad_arp_announce;	/* max ARP announcements */
1449 	int dad_arp_acount;	/* # of announcements */
1450 	struct callout dad_timer_ch;
1451 };
1452 
1453 static struct dadq_head dadq;
1454 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1455 static kmutex_t arp_dad_lock;
1456 
1457 static void
1458 arp_dad_init(void)
1459 {
1460 
1461 	TAILQ_INIT(&dadq);
1462 	mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1463 }
1464 
1465 static struct dadq *
1466 arp_dad_find(struct ifaddr *ifa)
1467 {
1468 	struct dadq *dp;
1469 
1470 	KASSERT(mutex_owned(&arp_dad_lock));
1471 
1472 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1473 		if (dp->dad_ifa == ifa)
1474 			return dp;
1475 	}
1476 	return NULL;
1477 }
1478 
1479 static void
1480 arp_dad_starttimer(struct dadq *dp, int ticks)
1481 {
1482 
1483 	callout_reset(&dp->dad_timer_ch, ticks,
1484 	    (void (*)(void *))arp_dad_timer, dp);
1485 }
1486 
1487 static void
1488 arp_dad_stoptimer(struct dadq *dp)
1489 {
1490 
1491 	KASSERT(mutex_owned(&arp_dad_lock));
1492 
1493 	TAILQ_REMOVE(&dadq, dp, dad_list);
1494 	/* Tell the timer that dp is being destroyed. */
1495 	dp->dad_ifa = NULL;
1496 	callout_halt(&dp->dad_timer_ch, &arp_dad_lock);
1497 }
1498 
1499 static void
1500 arp_dad_destroytimer(struct dadq *dp)
1501 {
1502 
1503 	callout_destroy(&dp->dad_timer_ch);
1504 	KASSERT(dp->dad_ifa == NULL);
1505 	kmem_intr_free(dp, sizeof(*dp));
1506 }
1507 
1508 static void
1509 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1510 {
1511 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1512 	struct ifnet *ifp = ifa->ifa_ifp;
1513 	struct in_addr sip;
1514 
1515 	dp->dad_arp_tcount++;
1516 	if ((ifp->if_flags & IFF_UP) == 0)
1517 		return;
1518 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1519 		return;
1520 
1521 	dp->dad_arp_tcount = 0;
1522 	dp->dad_arp_ocount++;
1523 
1524 	memset(&sip, 0, sizeof(sip));
1525 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1526 	    CLLADDR(ifa->ifa_ifp->if_sadl), NULL);
1527 }
1528 
1529 /*
1530  * Start Duplicate Address Detection (DAD) for specified interface address.
1531  */
1532 static void
1533 arp_dad_start(struct ifaddr *ifa)
1534 {
1535 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1536 	struct dadq *dp;
1537 	char ipbuf[INET_ADDRSTRLEN];
1538 
1539 	/*
1540 	 * If we don't need DAD, don't do it.
1541 	 * - DAD is disabled
1542 	 */
1543 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1544 		log(LOG_DEBUG,
1545 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1546 		    IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1547 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1548 		return;
1549 	}
1550 	if (!ip_dad_enabled()) {
1551 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1552 		rt_addrmsg(RTM_NEWADDR, ifa);
1553 		arpannounce1(ifa);
1554 		return;
1555 	}
1556 	KASSERT(ifa->ifa_ifp != NULL);
1557 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1558 		return;
1559 
1560 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1561 
1562 	mutex_enter(&arp_dad_lock);
1563 	if (arp_dad_find(ifa) != NULL) {
1564 		mutex_exit(&arp_dad_lock);
1565 		/* DAD already in progress */
1566 		if (dp != NULL)
1567 			kmem_intr_free(dp, sizeof(*dp));
1568 		return;
1569 	}
1570 
1571 	if (dp == NULL) {
1572 		mutex_exit(&arp_dad_lock);
1573 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1574 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1575 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1576 		return;
1577 	}
1578 
1579 	/*
1580 	 * Send ARP packet for DAD, ip_dad_count times.
1581 	 * Note that we must delay the first transmission.
1582 	 */
1583 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1584 	dp->dad_ifa = ifa;
1585 	ifaref(ifa);	/* just for safety */
1586 	dp->dad_count = ip_dad_count;
1587 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1588 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1589 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1590 
1591 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1592 	    ARPLOGADDR(&ia->ia_addr.sin_addr));
1593 
1594 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1595 
1596 	mutex_exit(&arp_dad_lock);
1597 }
1598 
1599 /*
1600  * terminate DAD unconditionally.  used for address removals.
1601  */
1602 static void
1603 arp_dad_stop(struct ifaddr *ifa)
1604 {
1605 	struct dadq *dp;
1606 
1607 	mutex_enter(&arp_dad_lock);
1608 	dp = arp_dad_find(ifa);
1609 	if (dp == NULL) {
1610 		mutex_exit(&arp_dad_lock);
1611 		/* DAD wasn't started yet */
1612 		return;
1613 	}
1614 
1615 	arp_dad_stoptimer(dp);
1616 
1617 	mutex_exit(&arp_dad_lock);
1618 
1619 	arp_dad_destroytimer(dp);
1620 	ifafree(ifa);
1621 }
1622 
1623 static void
1624 arp_dad_timer(struct dadq *dp)
1625 {
1626 	struct ifaddr *ifa;
1627 	struct in_ifaddr *ia;
1628 	char ipbuf[INET_ADDRSTRLEN];
1629 	bool need_free = false;
1630 
1631 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1632 	mutex_enter(&arp_dad_lock);
1633 
1634 	ifa = dp->dad_ifa;
1635 	if (ifa == NULL) {
1636 		/* dp is being destroyed by someone.  Do nothing. */
1637 		goto done;
1638 	}
1639 
1640 	ia = (struct in_ifaddr *)ifa;
1641 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1642 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1643 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1644 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1645 		goto done;
1646 	}
1647 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1648 	{
1649 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1650 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1651 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1652 		goto done;
1653 	}
1654 
1655 	/* timeouted with IFF_{RUNNING,UP} check */
1656 	if (dp->dad_arp_tcount > dad_maxtry) {
1657 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1658 		    if_name(ifa->ifa_ifp));
1659 
1660 		arp_dad_stoptimer(dp);
1661 		need_free = true;
1662 		goto done;
1663 	}
1664 
1665 	/* Need more checks? */
1666 	if (dp->dad_arp_ocount < dp->dad_count) {
1667 		int adelay;
1668 
1669 		/*
1670 		 * We have more ARP to go.  Send ARP packet for DAD.
1671 		 */
1672 		arp_dad_output(dp, ifa);
1673 		if (dp->dad_arp_ocount < dp->dad_count)
1674 			adelay = (PROBE_MIN * hz) +
1675 			    (cprng_fast32() %
1676 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1677 		else
1678 			adelay = ANNOUNCE_WAIT * hz;
1679 		arp_dad_starttimer(dp, adelay);
1680 		goto done;
1681 	} else if (dp->dad_arp_acount == 0) {
1682 		/*
1683 		 * We are done with DAD.
1684 		 * No duplicate address found.
1685 		 */
1686 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1687 		rt_addrmsg(RTM_NEWADDR, ifa);
1688 		ARPLOG(LOG_DEBUG,
1689 		    "%s: DAD complete for %s - no duplicates found\n",
1690 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1691 		dp->dad_arp_announce = ANNOUNCE_NUM;
1692 		goto announce;
1693 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1694 announce:
1695 		/*
1696 		 * Announce the address.
1697 		 */
1698 		arpannounce1(ifa);
1699 		dp->dad_arp_acount++;
1700 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1701 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1702 			goto done;
1703 		}
1704 		ARPLOG(LOG_DEBUG,
1705 		    "%s: ARP announcement complete for %s\n",
1706 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1707 	}
1708 
1709 	arp_dad_stoptimer(dp);
1710 	need_free = true;
1711 done:
1712 	mutex_exit(&arp_dad_lock);
1713 
1714 	if (need_free) {
1715 		arp_dad_destroytimer(dp);
1716 		KASSERT(ifa != NULL);
1717 		ifafree(ifa);
1718 	}
1719 
1720 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1721 }
1722 
1723 static void
1724 arp_dad_duplicated(struct ifaddr *ifa, const struct sockaddr_dl *from)
1725 {
1726 	struct in_ifaddr *ia = ifatoia(ifa);
1727 	struct ifnet *ifp = ifa->ifa_ifp;
1728 	char ipbuf[INET_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN];
1729 	const char *iastr, *llastr;
1730 
1731 	iastr = IN_PRINT(ipbuf, &ia->ia_addr.sin_addr);
1732 	if (__predict_false(from == NULL))
1733 		llastr = NULL;
1734 	else
1735 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
1736 		    CLLADDR(from), from->sdl_alen);
1737 
1738 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1739 		log(LOG_ERR,
1740 		    "%s: DAD duplicate address %s from %s\n",
1741 		    if_name(ifp), iastr, llastr);
1742 	} else if (ia->ia_dad_defended == 0 ||
1743 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1744 		ia->ia_dad_defended = time_uptime;
1745 		arpannounce1(ifa);
1746 		log(LOG_ERR,
1747 		    "%s: DAD defended address %s from %s\n",
1748 		    if_name(ifp), iastr, llastr);
1749 		return;
1750 	} else {
1751 		/* If DAD is disabled, just report the duplicate. */
1752 		if (!ip_dad_enabled()) {
1753 			log(LOG_ERR,
1754 			    "%s: DAD ignoring duplicate address %s from %s\n",
1755 			    if_name(ifp), iastr, llastr);
1756 			return;
1757 		}
1758 		log(LOG_ERR,
1759 		    "%s: DAD defence failed for %s from %s\n",
1760 		    if_name(ifp), iastr, llastr);
1761 	}
1762 
1763 	arp_dad_stop(ifa);
1764 
1765 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1766 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1767 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1768 		/* Inform the routing socket of the duplicate address */
1769 		rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1770 	}
1771 }
1772 
1773 /*
1774  * Called from 10 Mb/s Ethernet interrupt handlers
1775  * when ether packet type ETHERTYPE_REVARP
1776  * is received.  Common length and type checks are done here,
1777  * then the protocol-specific routine is called.
1778  */
1779 void
1780 revarpinput(struct mbuf *m)
1781 {
1782 	struct arphdr *ar;
1783 	int arplen;
1784 
1785 	arplen = sizeof(struct arphdr);
1786 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1787 		return;
1788 	ar = mtod(m, struct arphdr *);
1789 
1790 	if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394) {
1791 		goto out;
1792 	}
1793 
1794 	arplen = sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln);
1795 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1796 		return;
1797 	ar = mtod(m, struct arphdr *);
1798 
1799 	switch (ntohs(ar->ar_pro)) {
1800 	case ETHERTYPE_IP:
1801 	case ETHERTYPE_IPTRAILERS:
1802 		in_revarpinput(m);
1803 		return;
1804 
1805 	default:
1806 		break;
1807 	}
1808 
1809 out:
1810 	m_freem(m);
1811 }
1812 
1813 /*
1814  * RARP for Internet protocols on 10 Mb/s Ethernet.
1815  * Algorithm is that given in RFC 903.
1816  * We are only using for bootstrap purposes to get an ip address for one of
1817  * our interfaces.  Thus we support no user-interface.
1818  *
1819  * Since the contents of the RARP reply are specific to the interface that
1820  * sent the request, this code must ensure that they are properly associated.
1821  *
1822  * Note: also supports ARP via RARP packets, per the RFC.
1823  */
1824 void
1825 in_revarpinput(struct mbuf *m)
1826 {
1827 	struct arphdr *ah;
1828 	void *tha;
1829 	int op;
1830 	struct ifnet *rcvif;
1831 	int s;
1832 
1833 	ah = mtod(m, struct arphdr *);
1834 	op = ntohs(ah->ar_op);
1835 
1836 	rcvif = m_get_rcvif(m, &s);
1837 	if (__predict_false(rcvif == NULL))
1838 		goto out;
1839 	if (rcvif->if_flags & IFF_NOARP)
1840 		goto out;
1841 
1842 	switch (rcvif->if_type) {
1843 	case IFT_IEEE1394:
1844 		/* ARP without target hardware address is not supported */
1845 		goto out;
1846 	default:
1847 		break;
1848 	}
1849 
1850 	switch (op) {
1851 	case ARPOP_REQUEST:
1852 	case ARPOP_REPLY:	/* per RFC */
1853 		m_put_rcvif(rcvif, &s);
1854 		in_arpinput(m);
1855 		return;
1856 	case ARPOP_REVREPLY:
1857 		break;
1858 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1859 	default:
1860 		goto out;
1861 	}
1862 	if (!revarp_in_progress)
1863 		goto out;
1864 	if (rcvif != myip_ifp) /* !same interface */
1865 		goto out;
1866 	if (myip_initialized)
1867 		goto wake;
1868 	tha = ar_tha(ah);
1869 	if (tha == NULL)
1870 		goto out;
1871 	if (ah->ar_pln != sizeof(struct in_addr))
1872 		goto out;
1873 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1874 		goto out;
1875 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1876 		goto out;
1877 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1878 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1879 	myip_initialized = 1;
1880 wake:	/* Do wakeup every time in case it was missed. */
1881 	wakeup((void *)&myip);
1882 
1883 out:
1884 	m_put_rcvif(rcvif, &s);
1885 	m_freem(m);
1886 }
1887 
1888 /*
1889  * Send a RARP request for the ip address of the specified interface.
1890  * The request should be RFC 903-compliant.
1891  */
1892 static void
1893 revarprequest(struct ifnet *ifp)
1894 {
1895 	struct sockaddr sa;
1896 	struct mbuf *m;
1897 	struct arphdr *ah;
1898 	void *tha;
1899 
1900 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1901 		return;
1902 	MCLAIM(m, &arpdomain.dom_mowner);
1903 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1904 	    2*ifp->if_addrlen;
1905 	m->m_pkthdr.len = m->m_len;
1906 	m_align(m, m->m_len);
1907 	ah = mtod(m, struct arphdr *);
1908 	memset(ah, 0, m->m_len);
1909 	ah->ar_pro = htons(ETHERTYPE_IP);
1910 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1911 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1912 	ah->ar_op = htons(ARPOP_REVREQUEST);
1913 
1914 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1915 	tha = ar_tha(ah);
1916 	if (tha == NULL) {
1917 		m_free(m);
1918 		return;
1919 	}
1920 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1921 
1922 	sa.sa_family = AF_ARP;
1923 	sa.sa_len = 2;
1924 	m->m_flags |= M_BCAST;
1925 
1926 	if_output_lock(ifp, ifp, m, &sa, NULL);
1927 }
1928 
1929 /*
1930  * RARP for the ip address of the specified interface, but also
1931  * save the ip address of the server that sent the answer.
1932  * Timeout if no response is received.
1933  */
1934 int
1935 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1936     struct in_addr *clnt_in)
1937 {
1938 	int result, count = 20;
1939 
1940 	myip_initialized = 0;
1941 	myip_ifp = ifp;
1942 
1943 	revarp_in_progress = 1;
1944 	while (count--) {
1945 		revarprequest(ifp);
1946 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1947 		if (result != EWOULDBLOCK)
1948 			break;
1949 	}
1950 	revarp_in_progress = 0;
1951 
1952 	if (!myip_initialized)
1953 		return ENETUNREACH;
1954 
1955 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1956 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1957 	return 0;
1958 }
1959 
1960 void
1961 arp_stat_add(int type, uint64_t count)
1962 {
1963 	ARP_STATADD(type, count);
1964 }
1965 
1966 static int
1967 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1968 {
1969 
1970 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
1971 }
1972 
1973 static void
1974 sysctl_net_inet_arp_setup(struct sysctllog **clog)
1975 {
1976 	const struct sysctlnode *node;
1977 
1978 	sysctl_createv(clog, 0, NULL, NULL,
1979 			CTLFLAG_PERMANENT,
1980 			CTLTYPE_NODE, "inet", NULL,
1981 			NULL, 0, NULL, 0,
1982 			CTL_NET, PF_INET, CTL_EOL);
1983 	sysctl_createv(clog, 0, NULL, &node,
1984 			CTLFLAG_PERMANENT,
1985 			CTLTYPE_NODE, "arp",
1986 			SYSCTL_DESCR("Address Resolution Protocol"),
1987 			NULL, 0, NULL, 0,
1988 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1989 
1990 	sysctl_createv(clog, 0, NULL, NULL,
1991 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1992 		       CTLTYPE_INT, "nd_delay",
1993 		       SYSCTL_DESCR("First probe delay time"),
1994 		       NULL, 0, &arp_nd_domain.nd_delay, 0,
1995 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1996 	sysctl_createv(clog, 0, NULL, NULL,
1997 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1998 		       CTLTYPE_INT, "nd_bmaxtries",
1999 		       SYSCTL_DESCR("Number of broadcast discovery attempts"),
2000 		       NULL, 0, &arp_nd_domain.nd_mmaxtries, 0,
2001 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2002 	sysctl_createv(clog, 0, NULL, NULL,
2003 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2004 		       CTLTYPE_INT, "nd_umaxtries",
2005 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2006 		       NULL, 0, &arp_nd_domain.nd_umaxtries, 0,
2007 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2008 	sysctl_createv(clog, 0, NULL, NULL,
2009 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2010 		       CTLTYPE_INT, "nd_reachable",
2011 		       SYSCTL_DESCR("Reachable time"),
2012 		       NULL, 0, &arp_reachable, 0,
2013 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2014 	sysctl_createv(clog, 0, NULL, NULL,
2015 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2016 		       CTLTYPE_INT, "nd_retrans",
2017 		       SYSCTL_DESCR("Retransmission time"),
2018 		       NULL, 0, &arp_retrans, 0,
2019 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2020 	sysctl_createv(clog, 0, NULL, NULL,
2021 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2022 		       CTLTYPE_INT, "nd_nud",
2023 		       SYSCTL_DESCR("Perform neighbour unreachability detection"),
2024 		       NULL, 0, &arp_perform_nud, 0,
2025 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2026 	sysctl_createv(clog, 0, NULL, NULL,
2027 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2028 		       CTLTYPE_INT, "nd_maxnudhint",
2029 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2030 		       NULL, 0, &arp_nd_domain.nd_maxnudhint, 0,
2031 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2032 	sysctl_createv(clog, 0, NULL, NULL,
2033 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2034 		       CTLTYPE_INT, "maxqueuelen",
2035 		       SYSCTL_DESCR("max packet queue len for a unresolved ARP"),
2036 		       NULL, 1, &arp_nd_domain.nd_maxqueuelen, 0,
2037 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2038 
2039 	sysctl_createv(clog, 0, NULL, NULL,
2040 			CTLFLAG_PERMANENT,
2041 			CTLTYPE_STRUCT, "stats",
2042 			SYSCTL_DESCR("ARP statistics"),
2043 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2044 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2045 
2046 	sysctl_createv(clog, 0, NULL, NULL,
2047 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2048 			CTLTYPE_INT, "log_movements",
2049 			SYSCTL_DESCR("log ARP replies from MACs different than"
2050 			    " the one in the cache"),
2051 			NULL, 0, &log_movements, 0,
2052 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2053 
2054 	sysctl_createv(clog, 0, NULL, NULL,
2055 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2056 			CTLTYPE_INT, "log_permanent_modify",
2057 			SYSCTL_DESCR("log ARP replies from MACs different than"
2058 			    " the one in the permanent arp entry"),
2059 			NULL, 0, &log_permanent_modify, 0,
2060 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2061 
2062 	sysctl_createv(clog, 0, NULL, NULL,
2063 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2064 			CTLTYPE_INT, "log_wrong_iface",
2065 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2066 			    " interface"),
2067 			NULL, 0, &log_wrong_iface, 0,
2068 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2069 
2070 	sysctl_createv(clog, 0, NULL, NULL,
2071 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2072 			CTLTYPE_INT, "debug",
2073 			SYSCTL_DESCR("Enable ARP DAD debug output"),
2074 			NULL, 0, &arp_debug, 0,
2075 			CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2076 }
2077 
2078 #endif /* INET */
2079