xref: /netbsd-src/sys/netinet/if_arp.c (revision 3117ece4fc4a4ca4489ba793710b60b0d26bab6c)
1 /*	$NetBSD: if_arp.c,v 1.316 2024/10/04 23:31:06 roy 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 1.316 2024/10/04 23:31:06 roy 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 = 1;
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 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
797 	if (__predict_false(rcvif == NULL))
798 		goto out;
799 	if (rcvif->if_flags & IFF_NOARP)
800 		goto out;
801 
802 	memcpy(&isaddr, ar_spa(ah), sizeof(isaddr));
803 	memcpy(&itaddr, ar_tpa(ah), sizeof(itaddr));
804 
805 	if (m->m_flags & (M_BCAST|M_MCAST))
806 		ARP_STATINC(ARP_STAT_RCVMCAST);
807 
808 	/*
809 	 * Search for a matching interface address
810 	 * or any address on the interface to use
811 	 * as a dummy address in the rest of this function.
812 	 *
813 	 * First try and find the source address for early
814 	 * duplicate address detection.
815 	 */
816 	if (in_nullhost(isaddr)) {
817 		if (in_nullhost(itaddr)) /* very bogus ARP */
818 			goto out;
819 		find_source = false;
820 		myaddr = itaddr;
821 	} else {
822 		find_source = true;
823 		myaddr = isaddr;
824 	}
825 	s = pserialize_read_enter();
826 again:
827 	IN_ADDRHASH_READER_FOREACH(ia, myaddr.s_addr) {
828 		if (!in_hosteq(ia->ia_addr.sin_addr, myaddr))
829 			continue;
830 #if NCARP > 0
831 		if (ia->ia_ifp->if_type == IFT_CARP &&
832 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
833 		    (IFF_UP|IFF_RUNNING))) {
834 			index++;
835 			/* XXX: ar_hln? */
836 			if (ia->ia_ifp == rcvif && (ah->ar_hln >= 6) &&
837 			    carp_iamatch(ia, ar_sha(ah),
838 			    &count, index)) {
839 				break;
840 			}
841 		} else
842 #endif
843 		if (ia->ia_ifp == rcvif)
844 			break;
845 #if NBRIDGE > 0
846 		/*
847 		 * If the interface we received the packet on
848 		 * is part of a bridge, check to see if we need
849 		 * to "bridge" the packet to ourselves at this
850 		 * layer.  Note we still prefer a perfect match,
851 		 * but allow this weaker match if necessary.
852 		 */
853 		if (rcvif->if_bridge != NULL &&
854 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
855 			bridge_ia = ia;
856 #endif
857 	}
858 
859 #if NBRIDGE > 0
860 	if (ia == NULL && bridge_ia != NULL) {
861 		ia = bridge_ia;
862 		m_put_rcvif_psref(rcvif, &psref);
863 		rcvif = NULL;
864 		/* FIXME */
865 		ifp = bridge_ia->ia_ifp;
866 	}
867 #endif
868 
869 	/* If we failed to find the source address then find
870 	 * the target address. */
871 	if (ia == NULL && find_source && !in_nullhost(itaddr)) {
872 		find_source = false;
873 		myaddr = itaddr;
874 		goto again;
875 	}
876 
877 	if (ia != NULL)
878 		ia4_acquire(ia, &psref_ia);
879 	pserialize_read_exit(s);
880 
881 	if (ah->ar_hln != ifp->if_addrlen) {
882 		ARP_STATINC(ARP_STAT_RCVBADLEN);
883 		log(LOG_WARNING,
884 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
885 		    IN_PRINT(ipbuf, &isaddr), ah->ar_hln, ifp->if_addrlen);
886 		goto out;
887 	}
888 
889 	/* Only do DaD if we have a matching address. */
890 	do_dad = (ia != NULL);
891 
892 	if (ia == NULL) {
893 		ia = in_get_ia_on_iface_psref(isaddr, rcvif, &psref_ia);
894 		if (ia == NULL) {
895 			ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
896 			if (ia == NULL) {
897 				ARP_STATINC(ARP_STAT_RCVNOINT);
898 				goto out;
899 			}
900 		}
901 	}
902 
903 	myaddr = ia->ia_addr.sin_addr;
904 
905 	/* XXX checks for bridge case? */
906 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
907 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
908 		goto out;	/* it's from me, ignore it. */
909 	}
910 
911 	/* XXX checks for bridge case? */
912 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
913 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
914 		log(LOG_ERR,
915 		    "%s: arp: link address is broadcast for IP address %s!\n",
916 		    ifp->if_xname, IN_PRINT(ipbuf, &isaddr));
917 		goto out;
918 	}
919 
920 	/*
921 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
922 	 */
923 	if (in_nullhost(isaddr))
924 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
925 	else if (in_hosteq(isaddr, myaddr)) {
926 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
927 		/* This is the original behavior prior to supporting IPv4 DAD */
928 		if (!ip_dad_enabled()) {
929 			char llabuf[LLA_ADDRSTRLEN];
930 			log(LOG_ERR,
931 			    "duplicate IP address %s sent from link address %s\n",
932 			    IN_PRINT(ipbuf, &isaddr),
933 			    lla_snprintf(llabuf, sizeof(llabuf), ar_sha(ah),
934 			                 ah->ar_hln));
935 			itaddr = myaddr;
936 			goto reply;
937 		}
938 	}
939 
940 	if (in_nullhost(itaddr))
941 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
942 
943 	/*
944 	 * DAD check, RFC 5227.
945 	 * ARP sender hardware address must match the interface
946 	 * address of the interface sending the packet.
947 	 * Collision on sender address is always a duplicate.
948 	 * Collision on target address is only a duplicate
949 	 * IF the sender address is the null host (ie a DAD probe)
950 	 * AND the message was broadcast
951 	 * AND our address is either tentative or duplicated
952 	 * If it was unicast then it's a valid Unicast Poll from RFC 1122.
953 	 */
954 	if (ip_dad_enabled() && do_dad &&
955 	    (in_hosteq(isaddr, myaddr) ||
956 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr) &&
957 	     m->m_flags & M_BCAST &&
958 	     ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DUPLICATED))))
959 	{
960 		struct m_tag *mtag;
961 
962 		mtag = m_tag_find(m, PACKET_TAG_ETHERNET_SRC);
963 		if (mtag == NULL || (ah->ar_hln == ETHER_ADDR_LEN &&
964 		    memcmp(mtag + 1, ar_sha(ah), ah->ar_hln) == 0)) {
965 			struct sockaddr_dl sdl, *sdlp;
966 
967 			sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
968 			    ifp->if_index, ifp->if_type,
969 			    NULL, 0, ar_sha(ah), ah->ar_hln);
970 			arp_dad_duplicated((struct ifaddr *)ia, sdlp);
971 			goto out;
972 		}
973 	}
974 
975 	/*
976 	 * If the target IP address is zero, ignore the packet.
977 	 * This prevents the code below from trying to answer
978 	 * when we are using IP address zero (booting).
979 	 */
980 	if (in_nullhost(itaddr))
981 		goto out;
982 
983 	if (in_nullhost(isaddr))
984 		goto reply;
985 
986 	if (in_hosteq(itaddr, myaddr))
987 		la = arpcreate(ifp, &isaddr, NULL, 1);
988 	else
989 		la = arplookup(ifp, &isaddr, NULL, 1);
990 	if (la == NULL)
991 		goto reply;
992 
993 	if ((la->la_flags & LLE_VALID) &&
994 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen))
995 	{
996 		char llabuf[LLA_ADDRSTRLEN], *llastr;
997 
998 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
999 		    ar_sha(ah), ah->ar_hln);
1000 
1001 		if (la->la_flags & LLE_STATIC) {
1002 			ARP_STATINC(ARP_STAT_RCVOVERPERM);
1003 			if (!log_permanent_modify)
1004 				goto out;
1005 			log(LOG_INFO,
1006 			    "%s tried to overwrite permanent arp info"
1007 			    " for %s\n", llastr, IN_PRINT(ipbuf, &isaddr));
1008 			goto out;
1009 		} else if (la->lle_tbl->llt_ifp != ifp) {
1010 			/* XXX should not happen? */
1011 			ARP_STATINC(ARP_STAT_RCVOVERINT);
1012 			if (!log_wrong_iface)
1013 				goto out;
1014 			log(LOG_INFO,
1015 			    "%s on %s tried to overwrite "
1016 			    "arp info for %s on %s\n",
1017 			    llastr,
1018 			    ifp->if_xname, IN_PRINT(ipbuf, &isaddr),
1019 			    la->lle_tbl->llt_ifp->if_xname);
1020 				goto out;
1021 		} else {
1022 			ARP_STATINC(ARP_STAT_RCVOVER);
1023 			if (log_movements)
1024 				log(LOG_INFO, "arp info overwritten "
1025 				    "for %s by %s\n",
1026 				    IN_PRINT(ipbuf, &isaddr), llastr);
1027 		}
1028 		rt_cmd = RTM_CHANGE;
1029 		new_state = ND_LLINFO_STALE;
1030 	} else {
1031 		if (op == ARPOP_REPLY && in_hosteq(itaddr, myaddr)) {
1032 			/* This was a solicited ARP reply. */
1033 			la->ln_byhint = 0;
1034 			new_state = ND_LLINFO_REACHABLE;
1035 		} else if (op == ARPOP_REQUEST &&
1036 		           (la->ln_state == ND_LLINFO_NOSTATE ||
1037 			    la->ln_state == ND_LLINFO_INCOMPLETE)) {
1038 			/*
1039 			 * If an ARP request comes but there is no entry
1040 			 * and a new one has been created or an entry exists
1041 			 * but incomplete, make it stale to allow to send
1042 			 * packets to the requester without an ARP resolution.
1043 			 */
1044 			la->ln_byhint = 0;
1045 			new_state = ND_LLINFO_STALE;
1046 		}
1047 		rt_cmd = la->la_flags & LLE_VALID ? 0 : RTM_ADD;
1048 	}
1049 
1050 	KASSERT(ifp->if_sadl->sdl_alen == ifp->if_addrlen);
1051 
1052 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
1053 	memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
1054 	la->la_flags |= LLE_VALID;
1055 	la->ln_asked = 0;
1056 	if (new_state != 0) {
1057 		la->ln_state = new_state;
1058 
1059 		if (new_state != ND_LLINFO_REACHABLE ||
1060 		    !(la->la_flags & LLE_STATIC))
1061 		{
1062 			int timer = ND_TIMER_GC;
1063 
1064 			if (new_state == ND_LLINFO_REACHABLE)
1065 				timer = ND_TIMER_REACHABLE;
1066 			nd_set_timer(la, timer);
1067 		}
1068 	}
1069 
1070 	if (rt_cmd != 0) {
1071 		struct sockaddr_in sin;
1072 
1073 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1074 		rt_clonedmsg(rt_cmd, NULL, sintosa(&sin), ar_sha(ah), ifp);
1075 	}
1076 
1077 	if (la->la_hold != NULL) {
1078 		int n = la->la_numheld;
1079 		struct mbuf *m_hold, *m_hold_next;
1080 		struct sockaddr_in sin;
1081 
1082 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
1083 
1084 		m_hold = la->la_hold;
1085 		la->la_hold = NULL;
1086 		la->la_numheld = 0;
1087 		/*
1088 		 * We have to unlock here because if_output would call
1089 		 * arpresolve
1090 		 */
1091 		LLE_WUNLOCK(la);
1092 		ARP_STATADD(ARP_STAT_DFRSENT, n);
1093 		ARP_STATADD(ARP_STAT_DFRTOTAL, n);
1094 		for (; m_hold != NULL; m_hold = m_hold_next) {
1095 			m_hold_next = m_hold->m_nextpkt;
1096 			m_hold->m_nextpkt = NULL;
1097 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
1098 		}
1099 	} else
1100 		LLE_WUNLOCK(la);
1101 	la = NULL;
1102 
1103 reply:
1104 	if (la != NULL) {
1105 		LLE_WUNLOCK(la);
1106 		la = NULL;
1107 	}
1108 	if (op != ARPOP_REQUEST) {
1109 		if (op == ARPOP_REPLY)
1110 			ARP_STATINC(ARP_STAT_RCVREPLY);
1111 		goto out;
1112 	}
1113 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1114 	if (in_hosteq(itaddr, myaddr)) {
1115 		/* If our address is unusable, don't reply */
1116 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
1117 			goto out;
1118 		/* I am the target */
1119 		tha = ar_tha(ah);
1120 		if (tha)
1121 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1122 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1123 	} else {
1124 		/* Proxy ARP */
1125 		struct llentry *lle = NULL;
1126 		struct sockaddr_in sin;
1127 
1128 #if NCARP > 0
1129 		if (ifp->if_type == IFT_CARP) {
1130 			struct ifnet *_rcvif = m_get_rcvif(m, &s);
1131 			int iftype = 0;
1132 			if (__predict_true(_rcvif != NULL))
1133 				iftype = _rcvif->if_type;
1134 			m_put_rcvif(_rcvif, &s);
1135 			if (iftype != IFT_CARP)
1136 				goto out;
1137 		}
1138 #endif
1139 
1140 		tha = ar_tha(ah);
1141 
1142 		sockaddr_in_init(&sin, &itaddr, 0);
1143 
1144 		IF_AFDATA_RLOCK(ifp);
1145 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1146 		IF_AFDATA_RUNLOCK(ifp);
1147 
1148 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1149 			if (tha)
1150 				memcpy(tha, ar_sha(ah), ah->ar_hln);
1151 			memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
1152 			LLE_RUNLOCK(lle);
1153 		} else {
1154 			if (lle != NULL)
1155 				LLE_RUNLOCK(lle);
1156 			goto out;
1157 		}
1158 	}
1159 	ia4_release(ia, &psref_ia);
1160 
1161 	/*
1162 	 * XXX XXX: Here we're recycling the mbuf. But the mbuf could have
1163 	 * other mbufs in its chain, and just overwriting m->m_pkthdr.len
1164 	 * would be wrong in this case (the length becomes smaller than the
1165 	 * real chain size).
1166 	 *
1167 	 * This can theoretically cause bugs in the lower layers (drivers,
1168 	 * and L2encap), in some corner cases.
1169 	 */
1170 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1171 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1172 	ah->ar_op = htons(ARPOP_REPLY);
1173 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1174 	switch (ifp->if_type) {
1175 	case IFT_IEEE1394:
1176 		/* ieee1394 arp reply is broadcast */
1177 		m->m_flags &= ~M_MCAST;
1178 		m->m_flags |= M_BCAST;
1179 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1180 		break;
1181 	default:
1182 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1183 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1184 		break;
1185 	}
1186 	m->m_pkthdr.len = m->m_len;
1187 	sa.sa_family = AF_ARP;
1188 	sa.sa_len = 2;
1189 	arps = ARP_STAT_GETREF();
1190 	_NET_STATINC_REF(arps, ARP_STAT_SNDTOTAL);
1191 	_NET_STATINC_REF(arps, ARP_STAT_SNDREPLY);
1192 	ARP_STAT_PUTREF();
1193 	if_output_lock(ifp, ifp, m, &sa, NULL);
1194 	if (rcvif != NULL)
1195 		m_put_rcvif_psref(rcvif, &psref);
1196 	return;
1197 
1198 out:
1199 	if (la != NULL)
1200 		LLE_WUNLOCK(la);
1201 	if (ia != NULL)
1202 		ia4_release(ia, &psref_ia);
1203 	if (rcvif != NULL)
1204 		m_put_rcvif_psref(rcvif, &psref);
1205 	m_freem(m);
1206 }
1207 
1208 /*
1209  * Lookup or a new address in arptab.
1210  */
1211 struct llentry *
1212 arplookup(struct ifnet *ifp, const struct in_addr *addr,
1213     const struct sockaddr *sa, int wlock)
1214 {
1215 	struct sockaddr_in sin;
1216 	struct llentry *la;
1217 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1218 
1219 	if (sa == NULL) {
1220 		KASSERT(addr != NULL);
1221 		sockaddr_in_init(&sin, addr, 0);
1222 		sa = sintocsa(&sin);
1223 	}
1224 
1225 	IF_AFDATA_RLOCK(ifp);
1226 	la = lla_lookup(LLTABLE(ifp), flags, sa);
1227 	IF_AFDATA_RUNLOCK(ifp);
1228 
1229 	return la;
1230 }
1231 
1232 static struct llentry *
1233 arpcreate(struct ifnet *ifp, const struct in_addr *addr,
1234     const struct sockaddr *sa, int wlock)
1235 {
1236 	struct sockaddr_in sin;
1237 	struct llentry *la;
1238 	int flags = wlock ? LLE_EXCLUSIVE : 0;
1239 
1240 	if (sa == NULL) {
1241 		KASSERT(addr != NULL);
1242 		sockaddr_in_init(&sin, addr, 0);
1243 		sa = sintocsa(&sin);
1244 	}
1245 
1246 	la = arplookup(ifp, addr, sa, wlock);
1247 
1248 	if (la == NULL) {
1249 		struct rtentry *rt;
1250 
1251 		rt = rtalloc1(sa, 0);
1252 		IF_AFDATA_WLOCK(ifp);
1253 		la = lla_create(LLTABLE(ifp), flags, sa, rt);
1254 		IF_AFDATA_WUNLOCK(ifp);
1255 		if (rt != NULL)
1256 			rt_unref(rt);
1257 
1258 		if (la != NULL)
1259 			la->ln_state = ND_LLINFO_NOSTATE;
1260 	}
1261 
1262 	return la;
1263 }
1264 
1265 int
1266 arpioctl(u_long cmd, void *data)
1267 {
1268 
1269 	return EOPNOTSUPP;
1270 }
1271 
1272 void
1273 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1274 {
1275 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1276 
1277 	ifa->ifa_rtrequest = arp_rtrequest;
1278 	ifa->ifa_flags |= RTF_CONNECTED;
1279 
1280 	/* ARP will handle DAD for this address. */
1281 	if (in_nullhost(IA_SIN(ifa)->sin_addr)) {
1282 		if (ia->ia_dad_stop != NULL)	/* safety */
1283 			ia->ia_dad_stop(ifa);
1284 		ia->ia_dad_start = NULL;
1285 		ia->ia_dad_stop = NULL;
1286 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1287 	} else {
1288 		ia->ia_dad_start = arp_dad_start;
1289 		ia->ia_dad_stop = arp_dad_stop;
1290 		if (ia->ia4_flags & IN_IFF_TRYTENTATIVE && ip_dad_enabled())
1291 			ia->ia4_flags |= IN_IFF_TENTATIVE;
1292 		else
1293 			arpannounce1(ifa);
1294 	}
1295 }
1296 
1297 static bool
1298 arp_nud_enabled(__unused struct ifnet *ifp)
1299 {
1300 
1301 	return arp_perform_nud != 0;
1302 }
1303 
1304 static unsigned int
1305 arp_llinfo_reachable(__unused struct ifnet *ifp)
1306 {
1307 
1308 	return arp_reachable;
1309 }
1310 
1311 static unsigned int
1312 arp_llinfo_retrans(__unused struct ifnet *ifp)
1313 {
1314 
1315 	return arp_retrans;
1316 }
1317 
1318 /*
1319  * Gets source address of the first packet in hold queue
1320  * and stores it in @src.
1321  * Returns pointer to @src (if hold queue is not empty) or NULL.
1322  */
1323 static union l3addr *
1324 arp_llinfo_holdsrc(struct llentry *ln, union l3addr *src)
1325 {
1326 	struct ip *ip;
1327 
1328 	if (ln == NULL || ln->ln_hold == NULL)
1329 		return NULL;
1330 
1331 	/*
1332 	 * assuming every packet in ln_hold has the same IP header
1333 	 */
1334 	ip = mtod(ln->ln_hold, struct ip *);
1335 	/* XXX pullup? */
1336 	if (sizeof(*ip) < ln->ln_hold->m_len)
1337 		src->addr4 = ip->ip_src;
1338 	else
1339 		src = NULL;
1340 
1341 	return src;
1342 }
1343 
1344 static void
1345 arp_llinfo_output(struct ifnet *ifp, __unused const union l3addr *daddr,
1346     const union l3addr *taddr, const uint8_t *tlladdr,
1347     const union l3addr *hsrc)
1348 {
1349 	struct in_addr tip = taddr->addr4, sip = zeroin_addr;
1350 	const uint8_t *slladdr = CLLADDR(ifp->if_sadl);
1351 
1352 	if (hsrc != NULL) {
1353 		struct in_ifaddr *ia;
1354 		struct psref psref;
1355 
1356 		ia = in_get_ia_on_iface_psref(hsrc->addr4, ifp, &psref);
1357 		if (ia != NULL) {
1358 			sip = hsrc->addr4;
1359 			ia4_release(ia, &psref);
1360 		}
1361 	}
1362 
1363 	if (sip.s_addr == INADDR_ANY) {
1364 		struct sockaddr_in dst;
1365 		struct rtentry *rt;
1366 
1367 		sockaddr_in_init(&dst, &tip, 0);
1368 		rt = rtalloc1(sintosa(&dst), 0);
1369 		if (rt != NULL) {
1370 			if (rt->rt_ifp == ifp &&
1371 			    rt->rt_ifa != NULL &&
1372 			    rt->rt_ifa->ifa_addr->sa_family == AF_INET)
1373 				sip = satosin(rt->rt_ifa->ifa_addr)->sin_addr;
1374 			rt_unref(rt);
1375 		}
1376 		if (sip.s_addr == INADDR_ANY) {
1377 			char ipbuf[INET_ADDRSTRLEN];
1378 
1379 			log(LOG_DEBUG, "%s: source can't be "
1380 			    "determined: dst=%s\n", __func__,
1381 			    IN_PRINT(ipbuf, &tip));
1382 			return;
1383 		}
1384 	}
1385 
1386 	arprequest(ifp, &sip, &tip, slladdr, tlladdr);
1387 }
1388 
1389 
1390 static void
1391 arp_llinfo_missed(struct ifnet *ifp, const union l3addr *taddr,
1392     __unused int16_t type, struct mbuf *m)
1393 {
1394 	struct in_addr mdaddr = zeroin_addr;
1395 	struct sockaddr_in dsin, tsin;
1396 	struct sockaddr *sa;
1397 
1398 	if (m != NULL) {
1399 		struct ip *ip = mtod(m, struct ip *);
1400 
1401 		if (sizeof(*ip) < m->m_len)
1402 			mdaddr = ip->ip_src;
1403 
1404 		/* ip_input() will send ICMP_UNREACH_HOST, not us. */
1405 		m_freem(m);
1406 	}
1407 
1408 	if (mdaddr.s_addr != INADDR_ANY) {
1409 		sockaddr_in_init(&dsin, &mdaddr, 0);
1410 		sa = sintosa(&dsin);
1411 	} else
1412 		sa = NULL;
1413 
1414 	sockaddr_in_init(&tsin, &taddr->addr4, 0);
1415 	rt_clonedmsg(RTM_MISS, sa, sintosa(&tsin), NULL, ifp);
1416 }
1417 
1418 static void
1419 arp_free(struct llentry *ln, int gc)
1420 {
1421 	struct ifnet *ifp;
1422 
1423 	KASSERT(ln != NULL);
1424 	LLE_WLOCK_ASSERT(ln);
1425 
1426 	ifp = ln->lle_tbl->llt_ifp;
1427 
1428 	if (ln->la_flags & LLE_VALID || gc) {
1429 		struct sockaddr_in sin;
1430 		const char *lladdr;
1431 
1432 		sockaddr_in_init(&sin, &ln->r_l3addr.addr4, 0);
1433 		lladdr = ln->la_flags & LLE_VALID ?
1434 		    (const char *)&ln->ll_addr : NULL;
1435 		rt_clonedmsg(RTM_DELETE, NULL, sintosa(&sin), lladdr, ifp);
1436 	}
1437 
1438 	/*
1439 	 * Save to unlock. We still hold an extra reference and will not
1440 	 * free(9) in llentry_free() if someone else holds one as well.
1441 	 */
1442 	LLE_WUNLOCK(ln);
1443 	IF_AFDATA_LOCK(ifp);
1444 	LLE_WLOCK(ln);
1445 
1446 	lltable_free_entry(LLTABLE(ifp), ln);
1447 
1448 	IF_AFDATA_UNLOCK(ifp);
1449 }
1450 
1451 /*
1452  * Upper-layer reachability hint for Neighbor Unreachability Detection.
1453  *
1454  * XXX cost-effective methods?
1455  */
1456 void
1457 arp_nud_hint(struct rtentry *rt)
1458 {
1459 	struct llentry *ln;
1460 	struct ifnet *ifp;
1461 
1462 	if (rt == NULL)
1463 		return;
1464 
1465 	ifp = rt->rt_ifp;
1466 	ln = arplookup(ifp, NULL, rt_getkey(rt), 1);
1467 	nd_nud_hint(ln);
1468 }
1469 
1470 TAILQ_HEAD(dadq_head, dadq);
1471 struct dadq {
1472 	TAILQ_ENTRY(dadq) dad_list;
1473 	struct ifaddr *dad_ifa;
1474 	int dad_count;		/* max ARP to send */
1475 	int dad_arp_tcount;	/* # of trials to send ARP */
1476 	int dad_arp_ocount;	/* ARP sent so far */
1477 	int dad_arp_announce;	/* max ARP announcements */
1478 	int dad_arp_acount;	/* # of announcements */
1479 	struct callout dad_timer_ch;
1480 };
1481 
1482 static struct dadq_head dadq;
1483 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
1484 static kmutex_t arp_dad_lock;
1485 
1486 static void
1487 arp_dad_init(void)
1488 {
1489 
1490 	TAILQ_INIT(&dadq);
1491 	mutex_init(&arp_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1492 }
1493 
1494 static struct dadq *
1495 arp_dad_find(struct ifaddr *ifa)
1496 {
1497 	struct dadq *dp;
1498 
1499 	KASSERT(mutex_owned(&arp_dad_lock));
1500 
1501 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1502 		if (dp->dad_ifa == ifa)
1503 			return dp;
1504 	}
1505 	return NULL;
1506 }
1507 
1508 static void
1509 arp_dad_starttimer(struct dadq *dp, int ticks)
1510 {
1511 
1512 	callout_reset(&dp->dad_timer_ch, ticks,
1513 	    (void (*)(void *))arp_dad_timer, dp);
1514 }
1515 
1516 static void
1517 arp_dad_stoptimer(struct dadq *dp)
1518 {
1519 
1520 	KASSERT(mutex_owned(&arp_dad_lock));
1521 
1522 	TAILQ_REMOVE(&dadq, dp, dad_list);
1523 	/* Tell the timer that dp is being destroyed. */
1524 	dp->dad_ifa = NULL;
1525 	callout_halt(&dp->dad_timer_ch, &arp_dad_lock);
1526 }
1527 
1528 static void
1529 arp_dad_destroytimer(struct dadq *dp)
1530 {
1531 
1532 	callout_destroy(&dp->dad_timer_ch);
1533 	KASSERT(dp->dad_ifa == NULL);
1534 	kmem_intr_free(dp, sizeof(*dp));
1535 }
1536 
1537 static void
1538 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
1539 {
1540 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1541 	struct ifnet *ifp = ifa->ifa_ifp;
1542 	struct in_addr sip;
1543 
1544 	dp->dad_arp_tcount++;
1545 	if ((ifp->if_flags & IFF_UP) == 0)
1546 		return;
1547 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1548 		return;
1549 
1550 	dp->dad_arp_tcount = 0;
1551 	dp->dad_arp_ocount++;
1552 
1553 	memset(&sip, 0, sizeof(sip));
1554 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
1555 	    CLLADDR(ifa->ifa_ifp->if_sadl), NULL);
1556 }
1557 
1558 /*
1559  * Start Duplicate Address Detection (DAD) for specified interface address.
1560  */
1561 static void
1562 arp_dad_start(struct ifaddr *ifa)
1563 {
1564 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
1565 	struct dadq *dp;
1566 	char ipbuf[INET_ADDRSTRLEN];
1567 
1568 	/*
1569 	 * If we don't need DAD, don't do it.
1570 	 * - DAD is disabled
1571 	 */
1572 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
1573 		log(LOG_DEBUG,
1574 		    "%s: called with non-tentative address %s(%s)\n", __func__,
1575 		    IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1576 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1577 		return;
1578 	}
1579 	if (!ip_dad_enabled()) {
1580 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1581 		rt_addrmsg(RTM_NEWADDR, ifa);
1582 		arpannounce1(ifa);
1583 		return;
1584 	}
1585 	KASSERT(ifa->ifa_ifp != NULL);
1586 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1587 		return;
1588 
1589 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1590 
1591 	mutex_enter(&arp_dad_lock);
1592 	if (arp_dad_find(ifa) != NULL) {
1593 		mutex_exit(&arp_dad_lock);
1594 		/* DAD already in progress */
1595 		if (dp != NULL)
1596 			kmem_intr_free(dp, sizeof(*dp));
1597 		return;
1598 	}
1599 
1600 	if (dp == NULL) {
1601 		mutex_exit(&arp_dad_lock);
1602 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1603 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1604 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1605 		return;
1606 	}
1607 
1608 	/*
1609 	 * Send ARP packet for DAD, ip_dad_count times.
1610 	 * Note that we must delay the first transmission.
1611 	 */
1612 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1613 	dp->dad_ifa = ifa;
1614 	ifaref(ifa);	/* just for safety */
1615 	dp->dad_count = ip_dad_count;
1616 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
1617 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
1618 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1619 
1620 	ARPLOG(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1621 	    ARPLOGADDR(&ia->ia_addr.sin_addr));
1622 
1623 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
1624 
1625 	mutex_exit(&arp_dad_lock);
1626 }
1627 
1628 /*
1629  * terminate DAD unconditionally.  used for address removals.
1630  */
1631 static void
1632 arp_dad_stop(struct ifaddr *ifa)
1633 {
1634 	struct dadq *dp;
1635 
1636 	mutex_enter(&arp_dad_lock);
1637 	dp = arp_dad_find(ifa);
1638 	if (dp == NULL) {
1639 		mutex_exit(&arp_dad_lock);
1640 		/* DAD wasn't started yet */
1641 		return;
1642 	}
1643 
1644 	arp_dad_stoptimer(dp);
1645 
1646 	mutex_exit(&arp_dad_lock);
1647 
1648 	arp_dad_destroytimer(dp);
1649 	ifafree(ifa);
1650 }
1651 
1652 static void
1653 arp_dad_timer(struct dadq *dp)
1654 {
1655 	struct ifaddr *ifa;
1656 	struct in_ifaddr *ia;
1657 	char ipbuf[INET_ADDRSTRLEN];
1658 	bool need_free = false;
1659 
1660 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1661 	mutex_enter(&arp_dad_lock);
1662 
1663 	ifa = dp->dad_ifa;
1664 	if (ifa == NULL) {
1665 		/* dp is being destroyed by someone.  Do nothing. */
1666 		goto done;
1667 	}
1668 
1669 	ia = (struct in_ifaddr *)ifa;
1670 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1671 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
1672 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1673 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1674 		goto done;
1675 	}
1676 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
1677 	{
1678 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
1679 		    __func__, IN_PRINT(ipbuf, &ia->ia_addr.sin_addr),
1680 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1681 		goto done;
1682 	}
1683 
1684 	/* timeouted with IFF_{RUNNING,UP} check */
1685 	if (dp->dad_arp_tcount > dad_maxtry) {
1686 		ARPLOG(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1687 		    if_name(ifa->ifa_ifp));
1688 
1689 		arp_dad_stoptimer(dp);
1690 		need_free = true;
1691 		goto done;
1692 	}
1693 
1694 	/* Need more checks? */
1695 	if (dp->dad_arp_ocount < dp->dad_count) {
1696 		int adelay;
1697 
1698 		/*
1699 		 * We have more ARP to go.  Send ARP packet for DAD.
1700 		 */
1701 		arp_dad_output(dp, ifa);
1702 		if (dp->dad_arp_ocount < dp->dad_count)
1703 			adelay = (PROBE_MIN * hz) +
1704 			    (cprng_fast32() %
1705 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
1706 		else
1707 			adelay = ANNOUNCE_WAIT * hz;
1708 		arp_dad_starttimer(dp, adelay);
1709 		goto done;
1710 	} else if (dp->dad_arp_acount == 0) {
1711 		/*
1712 		 * We are done with DAD.
1713 		 * No duplicate address found.
1714 		 */
1715 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1716 		rt_addrmsg(RTM_NEWADDR, ifa);
1717 		ARPLOG(LOG_DEBUG,
1718 		    "%s: DAD complete for %s - no duplicates found\n",
1719 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1720 		dp->dad_arp_announce = ANNOUNCE_NUM;
1721 		goto announce;
1722 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
1723 announce:
1724 		/*
1725 		 * Announce the address.
1726 		 */
1727 		arpannounce1(ifa);
1728 		dp->dad_arp_acount++;
1729 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
1730 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
1731 			goto done;
1732 		}
1733 		ARPLOG(LOG_DEBUG,
1734 		    "%s: ARP announcement complete for %s\n",
1735 		    if_name(ifa->ifa_ifp), ARPLOGADDR(&ia->ia_addr.sin_addr));
1736 	}
1737 
1738 	arp_dad_stoptimer(dp);
1739 	need_free = true;
1740 done:
1741 	mutex_exit(&arp_dad_lock);
1742 
1743 	if (need_free) {
1744 		arp_dad_destroytimer(dp);
1745 		KASSERT(ifa != NULL);
1746 		ifafree(ifa);
1747 	}
1748 
1749 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1750 }
1751 
1752 static void
1753 arp_dad_duplicated(struct ifaddr *ifa, const struct sockaddr_dl *from)
1754 {
1755 	struct in_ifaddr *ia = ifatoia(ifa);
1756 	struct ifnet *ifp = ifa->ifa_ifp;
1757 	char ipbuf[INET_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN];
1758 	const char *iastr, *llastr;
1759 
1760 	iastr = IN_PRINT(ipbuf, &ia->ia_addr.sin_addr);
1761 	if (__predict_false(from == NULL))
1762 		llastr = NULL;
1763 	else
1764 		llastr = lla_snprintf(llabuf, sizeof(llabuf),
1765 		    CLLADDR(from), from->sdl_alen);
1766 
1767 	if (ia->ia4_flags & (IN_IFF_TENTATIVE|IN_IFF_DUPLICATED)) {
1768 		log(LOG_ERR,
1769 		    "%s: DAD duplicate address %s from %s\n",
1770 		    if_name(ifp), iastr, llastr);
1771 	} else if (ia->ia_dad_defended == 0 ||
1772 		   ia->ia_dad_defended < time_uptime - DEFEND_INTERVAL) {
1773 		ia->ia_dad_defended = time_uptime;
1774 		arpannounce1(ifa);
1775 		log(LOG_ERR,
1776 		    "%s: DAD defended address %s from %s\n",
1777 		    if_name(ifp), iastr, llastr);
1778 		return;
1779 	} else {
1780 		/* If DAD is disabled, just report the duplicate. */
1781 		if (!ip_dad_enabled()) {
1782 			log(LOG_ERR,
1783 			    "%s: DAD ignoring duplicate address %s from %s\n",
1784 			    if_name(ifp), iastr, llastr);
1785 			return;
1786 		}
1787 		log(LOG_ERR,
1788 		    "%s: DAD defence failed for %s from %s\n",
1789 		    if_name(ifp), iastr, llastr);
1790 	}
1791 
1792 	arp_dad_stop(ifa);
1793 
1794 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1795 	if ((ia->ia4_flags & IN_IFF_DUPLICATED) == 0) {
1796 		ia->ia4_flags |= IN_IFF_DUPLICATED;
1797 		/* Inform the routing socket of the duplicate address */
1798 		rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1799 	}
1800 }
1801 
1802 /*
1803  * Called from 10 Mb/s Ethernet interrupt handlers
1804  * when ether packet type ETHERTYPE_REVARP
1805  * is received.  Common length and type checks are done here,
1806  * then the protocol-specific routine is called.
1807  */
1808 void
1809 revarpinput(struct mbuf *m)
1810 {
1811 	struct arphdr *ar;
1812 	int arplen;
1813 
1814 	arplen = sizeof(struct arphdr);
1815 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1816 		return;
1817 	ar = mtod(m, struct arphdr *);
1818 
1819 	if (ntohs(ar->ar_hrd) == ARPHRD_IEEE1394) {
1820 		goto out;
1821 	}
1822 
1823 	arplen = sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln);
1824 	if (m->m_len < arplen && (m = m_pullup(m, arplen)) == NULL)
1825 		return;
1826 	ar = mtod(m, struct arphdr *);
1827 
1828 	switch (ntohs(ar->ar_pro)) {
1829 	case ETHERTYPE_IP:
1830 	case ETHERTYPE_IPTRAILERS:
1831 		in_revarpinput(m);
1832 		return;
1833 
1834 	default:
1835 		break;
1836 	}
1837 
1838 out:
1839 	m_freem(m);
1840 }
1841 
1842 /*
1843  * RARP for Internet protocols on 10 Mb/s Ethernet.
1844  * Algorithm is that given in RFC 903.
1845  * We are only using for bootstrap purposes to get an ip address for one of
1846  * our interfaces.  Thus we support no user-interface.
1847  *
1848  * Since the contents of the RARP reply are specific to the interface that
1849  * sent the request, this code must ensure that they are properly associated.
1850  *
1851  * Note: also supports ARP via RARP packets, per the RFC.
1852  */
1853 void
1854 in_revarpinput(struct mbuf *m)
1855 {
1856 	struct arphdr *ah;
1857 	void *tha;
1858 	int op;
1859 	struct ifnet *rcvif;
1860 	int s;
1861 
1862 	ah = mtod(m, struct arphdr *);
1863 	op = ntohs(ah->ar_op);
1864 
1865 	rcvif = m_get_rcvif(m, &s);
1866 	if (__predict_false(rcvif == NULL))
1867 		goto out;
1868 	if (rcvif->if_flags & IFF_NOARP)
1869 		goto out;
1870 
1871 	switch (rcvif->if_type) {
1872 	case IFT_IEEE1394:
1873 		/* ARP without target hardware address is not supported */
1874 		goto out;
1875 	default:
1876 		break;
1877 	}
1878 
1879 	switch (op) {
1880 	case ARPOP_REQUEST:
1881 	case ARPOP_REPLY:	/* per RFC */
1882 		m_put_rcvif(rcvif, &s);
1883 		in_arpinput(m);
1884 		return;
1885 	case ARPOP_REVREPLY:
1886 		break;
1887 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1888 	default:
1889 		goto out;
1890 	}
1891 	if (!revarp_in_progress)
1892 		goto out;
1893 	if (rcvif != myip_ifp) /* !same interface */
1894 		goto out;
1895 	if (myip_initialized)
1896 		goto wake;
1897 	tha = ar_tha(ah);
1898 	if (tha == NULL)
1899 		goto out;
1900 	if (ah->ar_pln != sizeof(struct in_addr))
1901 		goto out;
1902 	if (ah->ar_hln != rcvif->if_sadl->sdl_alen)
1903 		goto out;
1904 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
1905 		goto out;
1906 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1907 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1908 	myip_initialized = 1;
1909 wake:	/* Do wakeup every time in case it was missed. */
1910 	wakeup((void *)&myip);
1911 
1912 out:
1913 	m_put_rcvif(rcvif, &s);
1914 	m_freem(m);
1915 }
1916 
1917 /*
1918  * Send a RARP request for the ip address of the specified interface.
1919  * The request should be RFC 903-compliant.
1920  */
1921 static void
1922 revarprequest(struct ifnet *ifp)
1923 {
1924 	struct sockaddr sa;
1925 	struct mbuf *m;
1926 	struct arphdr *ah;
1927 	void *tha;
1928 
1929 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1930 		return;
1931 	MCLAIM(m, &arpdomain.dom_mowner);
1932 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1933 	    2*ifp->if_addrlen;
1934 	m->m_pkthdr.len = m->m_len;
1935 	m_align(m, m->m_len);
1936 	ah = mtod(m, struct arphdr *);
1937 	memset(ah, 0, m->m_len);
1938 	ah->ar_pro = htons(ETHERTYPE_IP);
1939 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1940 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1941 	ah->ar_op = htons(ARPOP_REVREQUEST);
1942 
1943 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1944 	tha = ar_tha(ah);
1945 	if (tha == NULL) {
1946 		m_free(m);
1947 		return;
1948 	}
1949 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1950 
1951 	sa.sa_family = AF_ARP;
1952 	sa.sa_len = 2;
1953 	m->m_flags |= M_BCAST;
1954 
1955 	if_output_lock(ifp, ifp, m, &sa, NULL);
1956 }
1957 
1958 /*
1959  * RARP for the ip address of the specified interface, but also
1960  * save the ip address of the server that sent the answer.
1961  * Timeout if no response is received.
1962  */
1963 int
1964 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1965     struct in_addr *clnt_in)
1966 {
1967 	int result, count = 20;
1968 
1969 	myip_initialized = 0;
1970 	myip_ifp = ifp;
1971 
1972 	revarp_in_progress = 1;
1973 	while (count--) {
1974 		revarprequest(ifp);
1975 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1976 		if (result != EWOULDBLOCK)
1977 			break;
1978 	}
1979 	revarp_in_progress = 0;
1980 
1981 	if (!myip_initialized)
1982 		return ENETUNREACH;
1983 
1984 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1985 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1986 	return 0;
1987 }
1988 
1989 void
1990 arp_stat_add(int type, uint64_t count)
1991 {
1992 	ARP_STATADD(type, count);
1993 }
1994 
1995 static int
1996 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1997 {
1998 
1999 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
2000 }
2001 
2002 static void
2003 sysctl_net_inet_arp_setup(struct sysctllog **clog)
2004 {
2005 	const struct sysctlnode *node;
2006 
2007 	sysctl_createv(clog, 0, NULL, NULL,
2008 			CTLFLAG_PERMANENT,
2009 			CTLTYPE_NODE, "inet", NULL,
2010 			NULL, 0, NULL, 0,
2011 			CTL_NET, PF_INET, CTL_EOL);
2012 	sysctl_createv(clog, 0, NULL, &node,
2013 			CTLFLAG_PERMANENT,
2014 			CTLTYPE_NODE, "arp",
2015 			SYSCTL_DESCR("Address Resolution Protocol"),
2016 			NULL, 0, NULL, 0,
2017 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2018 
2019 	sysctl_createv(clog, 0, NULL, NULL,
2020 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2021 		       CTLTYPE_INT, "nd_delay",
2022 		       SYSCTL_DESCR("First probe delay time"),
2023 		       NULL, 0, &arp_nd_domain.nd_delay, 0,
2024 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2025 	sysctl_createv(clog, 0, NULL, NULL,
2026 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2027 		       CTLTYPE_INT, "nd_bmaxtries",
2028 		       SYSCTL_DESCR("Number of broadcast discovery attempts"),
2029 		       NULL, 0, &arp_nd_domain.nd_mmaxtries, 0,
2030 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2031 	sysctl_createv(clog, 0, NULL, NULL,
2032 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2033 		       CTLTYPE_INT, "nd_umaxtries",
2034 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2035 		       NULL, 0, &arp_nd_domain.nd_umaxtries, 0,
2036 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2037 	sysctl_createv(clog, 0, NULL, NULL,
2038 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2039 		       CTLTYPE_INT, "nd_reachable",
2040 		       SYSCTL_DESCR("Reachable time"),
2041 		       NULL, 0, &arp_reachable, 0,
2042 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2043 	sysctl_createv(clog, 0, NULL, NULL,
2044 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2045 		       CTLTYPE_INT, "nd_retrans",
2046 		       SYSCTL_DESCR("Retransmission time"),
2047 		       NULL, 0, &arp_retrans, 0,
2048 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2049 	sysctl_createv(clog, 0, NULL, NULL,
2050 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2051 		       CTLTYPE_INT, "nd_nud",
2052 		       SYSCTL_DESCR("Perform neighbour unreachability detection"),
2053 		       NULL, 0, &arp_perform_nud, 0,
2054 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2055 	sysctl_createv(clog, 0, NULL, NULL,
2056 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2057 		       CTLTYPE_INT, "nd_maxnudhint",
2058 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2059 		       NULL, 0, &arp_nd_domain.nd_maxnudhint, 0,
2060 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2061 	sysctl_createv(clog, 0, NULL, NULL,
2062 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2063 		       CTLTYPE_INT, "maxqueuelen",
2064 		       SYSCTL_DESCR("max packet queue len for a unresolved ARP"),
2065 		       NULL, 1, &arp_nd_domain.nd_maxqueuelen, 0,
2066 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2067 
2068 	sysctl_createv(clog, 0, NULL, NULL,
2069 			CTLFLAG_PERMANENT,
2070 			CTLTYPE_STRUCT, "stats",
2071 			SYSCTL_DESCR("ARP statistics"),
2072 			sysctl_net_inet_arp_stats, 0, NULL, 0,
2073 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2074 
2075 	sysctl_createv(clog, 0, NULL, NULL,
2076 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2077 			CTLTYPE_INT, "log_movements",
2078 			SYSCTL_DESCR("log ARP replies from MACs different than"
2079 			    " the one in the cache"),
2080 			NULL, 0, &log_movements, 0,
2081 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2082 
2083 	sysctl_createv(clog, 0, NULL, NULL,
2084 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2085 			CTLTYPE_INT, "log_permanent_modify",
2086 			SYSCTL_DESCR("log ARP replies from MACs different than"
2087 			    " the one in the permanent arp entry"),
2088 			NULL, 0, &log_permanent_modify, 0,
2089 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2090 
2091 	sysctl_createv(clog, 0, NULL, NULL,
2092 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2093 			CTLTYPE_INT, "log_wrong_iface",
2094 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
2095 			    " interface"),
2096 			NULL, 0, &log_wrong_iface, 0,
2097 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2098 
2099 	sysctl_createv(clog, 0, NULL, NULL,
2100 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2101 			CTLTYPE_INT, "debug",
2102 			SYSCTL_DESCR("Enable ARP DAD debug output"),
2103 			NULL, 0, &arp_debug, 0,
2104 			CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
2105 }
2106 
2107 #endif /* INET */
2108