if_ether.c revision 206481
1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30 */
31
32/*
33 * Ethernet address resolution protocol.
34 * TODO:
35 *	add "inuse/lock" bit (or ref. count) along with valid bit
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/netinet/if_ether.c 206481 2010-04-11 16:04:08Z bz $");
40
41#include "opt_inet.h"
42#include "opt_carp.h"
43
44#include <sys/param.h>
45#include <sys/kernel.h>
46#include <sys/queue.h>
47#include <sys/sysctl.h>
48#include <sys/systm.h>
49#include <sys/mbuf.h>
50#include <sys/malloc.h>
51#include <sys/proc.h>
52#include <sys/socket.h>
53#include <sys/syslog.h>
54
55#include <net/if.h>
56#include <net/if_dl.h>
57#include <net/if_types.h>
58#include <net/netisr.h>
59#include <net/if_llc.h>
60#include <net/ethernet.h>
61#include <net/route.h>
62#include <net/vnet.h>
63
64#include <netinet/in.h>
65#include <netinet/in_var.h>
66#include <net/if_llatbl.h>
67#include <netinet/if_ether.h>
68
69#include <net/if_arc.h>
70#include <net/iso88025.h>
71
72#ifdef DEV_CARP
73#include <netinet/ip_carp.h>
74#endif
75
76#include <security/mac/mac_framework.h>
77
78#define SIN(s) ((struct sockaddr_in *)s)
79#define SDL(s) ((struct sockaddr_dl *)s)
80
81SYSCTL_DECL(_net_link_ether);
82SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
83SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
84
85VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
86					 * local traffic */
87
88/* timer values */
89static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
90						 * minutes */
91static VNET_DEFINE(int, arpt_down) = 20;      /* keep incomplete entries for
92					       * 20 seconds */
93static VNET_DEFINE(int, arp_maxtries) = 5;
94static VNET_DEFINE(int, arp_proxyall);
95static VNET_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
96
97#define	V_arpt_keep		VNET(arpt_keep)
98#define	V_arpt_down		VNET(arpt_down)
99#define	V_arp_maxtries		VNET(arp_maxtries)
100#define	V_arp_proxyall		VNET(arp_proxyall)
101#define	V_arpstat		VNET(arpstat)
102
103SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
104	&VNET_NAME(arpt_keep), 0,
105	"ARP entry lifetime in seconds");
106
107SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
108	&VNET_NAME(arp_maxtries), 0,
109	"ARP resolution attempts before returning error");
110SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
111	&VNET_NAME(useloopback), 0,
112	"Use the loopback interface for local traffic");
113SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
114	&VNET_NAME(arp_proxyall), 0,
115	"Enable proxy ARP for all suitable requests");
116SYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW,
117	&VNET_NAME(arpstat), arpstat,
118	"ARP statistics (struct arpstat, net/if_arp.h)");
119
120static void	arp_init(void);
121void		arprequest(struct ifnet *,
122			struct in_addr *, struct in_addr *, u_char *);
123static void	arpintr(struct mbuf *);
124static void	arptimer(void *);
125#ifdef INET
126static void	in_arpinput(struct mbuf *);
127#endif
128
129static const struct netisr_handler arp_nh = {
130	.nh_name = "arp",
131	.nh_handler = arpintr,
132	.nh_proto = NETISR_ARP,
133	.nh_policy = NETISR_POLICY_SOURCE,
134};
135
136#ifdef AF_INET
137void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
138
139/*
140 * called by in_ifscrub to remove entry from the table when
141 * the interface goes away
142 */
143void
144arp_ifscrub(struct ifnet *ifp, uint32_t addr)
145{
146	struct sockaddr_in addr4;
147
148	bzero((void *)&addr4, sizeof(addr4));
149	addr4.sin_len    = sizeof(addr4);
150	addr4.sin_family = AF_INET;
151	addr4.sin_addr.s_addr = addr;
152	CURVNET_SET(ifp->if_vnet);
153	IF_AFDATA_LOCK(ifp);
154	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
155	    (struct sockaddr *)&addr4);
156	IF_AFDATA_UNLOCK(ifp);
157	CURVNET_RESTORE();
158}
159#endif
160
161/*
162 * Timeout routine.  Age arp_tab entries periodically.
163 */
164static void
165arptimer(void *arg)
166{
167	struct ifnet *ifp;
168	struct llentry   *lle = (struct llentry *)arg;
169
170	if (lle == NULL) {
171		panic("%s: NULL entry!\n", __func__);
172		return;
173	}
174	ifp = lle->lle_tbl->llt_ifp;
175	CURVNET_SET(ifp->if_vnet);
176	IF_AFDATA_LOCK(ifp);
177	LLE_WLOCK(lle);
178	if (lle->la_flags & LLE_STATIC)
179		LLE_WUNLOCK(lle);
180	else {
181		if (!callout_pending(&lle->la_timer) &&
182		    callout_active(&lle->la_timer)) {
183			callout_stop(&lle->la_timer);
184			LLE_REMREF(lle);
185			(void) llentry_free(lle);
186			ARPSTAT_INC(timeouts);
187		}
188#ifdef DIAGNOSTIC
189		else {
190			struct sockaddr *l3addr = L3_ADDR(lle);
191			log(LOG_INFO,
192			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
193			    inet_ntoa(
194			        ((const struct sockaddr_in *)l3addr)->sin_addr));
195		}
196#endif
197	}
198	IF_AFDATA_UNLOCK(ifp);
199	CURVNET_RESTORE();
200}
201
202/*
203 * Broadcast an ARP request. Caller specifies:
204 *	- arp header source ip address
205 *	- arp header target ip address
206 *	- arp header source ethernet address
207 */
208void
209arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
210    u_char *enaddr)
211{
212	struct mbuf *m;
213	struct arphdr *ah;
214	struct sockaddr sa;
215
216	if (sip == NULL) {
217		/* XXX don't believe this can happen (or explain why) */
218		/*
219		 * The caller did not supply a source address, try to find
220		 * a compatible one among those assigned to this interface.
221		 */
222		struct ifaddr *ifa;
223
224		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
225			if (!ifa->ifa_addr ||
226			    ifa->ifa_addr->sa_family != AF_INET)
227				continue;
228			sip = &SIN(ifa->ifa_addr)->sin_addr;
229			if (0 == ((sip->s_addr ^ tip->s_addr) &
230			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
231				break;  /* found it. */
232		}
233		if (sip == NULL) {
234			printf("%s: cannot find matching address\n", __func__);
235			return;
236		}
237	}
238
239	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
240		return;
241	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
242		2*ifp->if_data.ifi_addrlen;
243	m->m_pkthdr.len = m->m_len;
244	MH_ALIGN(m, m->m_len);
245	ah = mtod(m, struct arphdr *);
246	bzero((caddr_t)ah, m->m_len);
247#ifdef MAC
248	mac_netinet_arp_send(ifp, m);
249#endif
250	ah->ar_pro = htons(ETHERTYPE_IP);
251	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
252	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
253	ah->ar_op = htons(ARPOP_REQUEST);
254	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
255	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
256	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
257	sa.sa_family = AF_ARP;
258	sa.sa_len = 2;
259	m->m_flags |= M_BCAST;
260	(*ifp->if_output)(ifp, m, &sa, NULL);
261	ARPSTAT_INC(txrequests);
262}
263
264/*
265 * Resolve an IP address into an ethernet address.
266 * On input:
267 *    ifp is the interface we use
268 *    rt0 is the route to the final destination (possibly useless)
269 *    m is the mbuf. May be NULL if we don't have a packet.
270 *    dst is the next hop,
271 *    desten is where we want the address.
272 *
273 * On success, desten is filled in and the function returns 0;
274 * If the packet must be held pending resolution, we return EWOULDBLOCK
275 * On other errors, we return the corresponding error code.
276 * Note that m_freem() handles NULL.
277 */
278int
279arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
280	struct sockaddr *dst, u_char *desten, struct llentry **lle)
281{
282	struct llentry *la = 0;
283	u_int flags = 0;
284	int error, renew;
285
286	*lle = NULL;
287	if (m != NULL) {
288		if (m->m_flags & M_BCAST) {
289			/* broadcast */
290			(void)memcpy(desten,
291			    ifp->if_broadcastaddr, ifp->if_addrlen);
292			return (0);
293		}
294		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
295			/* multicast */
296			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
297			return (0);
298		}
299	}
300	/* XXXXX
301	 */
302retry:
303	IF_AFDATA_RLOCK(ifp);
304	la = lla_lookup(LLTABLE(ifp), flags, dst);
305	IF_AFDATA_RUNLOCK(ifp);
306	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
307	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
308		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
309		IF_AFDATA_WLOCK(ifp);
310		la = lla_lookup(LLTABLE(ifp), flags, dst);
311		IF_AFDATA_WUNLOCK(ifp);
312	}
313	if (la == NULL) {
314		if (flags & LLE_CREATE)
315			log(LOG_DEBUG,
316			    "arpresolve: can't allocate llinfo for %s\n",
317			    inet_ntoa(SIN(dst)->sin_addr));
318		m_freem(m);
319		return (EINVAL);
320	}
321
322	if ((la->la_flags & LLE_VALID) &&
323	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_second)) {
324		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
325		/*
326		 * If entry has an expiry time and it is approaching,
327		 * see if we need to send an ARP request within this
328		 * arpt_down interval.
329		 */
330		if (!(la->la_flags & LLE_STATIC) &&
331		    time_second + la->la_preempt > la->la_expire) {
332			arprequest(ifp, NULL,
333			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
334
335			la->la_preempt--;
336		}
337
338		*lle = la;
339		error = 0;
340		goto done;
341	}
342
343	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
344		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
345		    inet_ntoa(SIN(dst)->sin_addr));
346		m_freem(m);
347		error = EINVAL;
348		goto done;
349	}
350
351	renew = (la->la_asked == 0 || la->la_expire != time_second);
352	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
353		flags |= LLE_EXCLUSIVE;
354		LLE_RUNLOCK(la);
355		goto retry;
356	}
357	/*
358	 * There is an arptab entry, but no ethernet address
359	 * response yet.  Replace the held mbuf with this
360	 * latest one.
361	 */
362	if (m != NULL) {
363		if (la->la_hold != NULL) {
364			m_freem(la->la_hold);
365			ARPSTAT_INC(dropped);
366		}
367		la->la_hold = m;
368		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
369			flags &= ~LLE_EXCLUSIVE;
370			LLE_DOWNGRADE(la);
371		}
372
373	}
374	/*
375	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
376	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
377	 * if we have already sent arp_maxtries ARP requests. Retransmit the
378	 * ARP request, but not faster than one request per second.
379	 */
380	if (la->la_asked < V_arp_maxtries)
381		error = EWOULDBLOCK;	/* First request. */
382	else
383		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
384		    EHOSTUNREACH : EHOSTDOWN;
385
386	if (renew) {
387		int canceled;
388
389		LLE_ADDREF(la);
390		la->la_expire = time_second + V_arpt_down;
391		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
392		    arptimer, la);
393		if (canceled)
394			LLE_REMREF(la);
395		la->la_asked++;
396		LLE_WUNLOCK(la);
397		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
398		    IF_LLADDR(ifp));
399		return (error);
400	}
401done:
402	if (flags & LLE_EXCLUSIVE)
403		LLE_WUNLOCK(la);
404	else
405		LLE_RUNLOCK(la);
406	return (error);
407}
408
409/*
410 * Common length and type checks are done here,
411 * then the protocol-specific routine is called.
412 */
413static void
414arpintr(struct mbuf *m)
415{
416	struct arphdr *ar;
417
418	if (m->m_len < sizeof(struct arphdr) &&
419	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
420		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
421		return;
422	}
423	ar = mtod(m, struct arphdr *);
424
425	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
426	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
427	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
428	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
429		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
430		    (unsigned char *)&ar->ar_hrd, "");
431		m_freem(m);
432		return;
433	}
434
435	if (m->m_len < arphdr_len(ar)) {
436		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
437			log(LOG_ERR, "arp: runt packet\n");
438			m_freem(m);
439			return;
440		}
441		ar = mtod(m, struct arphdr *);
442	}
443
444	ARPSTAT_INC(received);
445	switch (ntohs(ar->ar_pro)) {
446#ifdef INET
447	case ETHERTYPE_IP:
448		in_arpinput(m);
449		return;
450#endif
451	}
452	m_freem(m);
453}
454
455#ifdef INET
456/*
457 * ARP for Internet protocols on 10 Mb/s Ethernet.
458 * Algorithm is that given in RFC 826.
459 * In addition, a sanity check is performed on the sender
460 * protocol address, to catch impersonators.
461 * We no longer handle negotiations for use of trailer protocol:
462 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
463 * along with IP replies if we wanted trailers sent to us,
464 * and also sent them in response to IP replies.
465 * This allowed either end to announce the desire to receive
466 * trailer packets.
467 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
468 * but formerly didn't normally send requests.
469 */
470static int log_arp_wrong_iface = 1;
471static int log_arp_movements = 1;
472static int log_arp_permanent_modify = 1;
473
474SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
475	&log_arp_wrong_iface, 0,
476	"log arp packets arriving on the wrong interface");
477SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
478        &log_arp_movements, 0,
479        "log arp replies from MACs different than the one in the cache");
480SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
481        &log_arp_permanent_modify, 0,
482        "log arp replies from MACs different than the one in the permanent arp entry");
483
484
485static void
486in_arpinput(struct mbuf *m)
487{
488	struct arphdr *ah;
489	struct ifnet *ifp = m->m_pkthdr.rcvif;
490	struct llentry *la = NULL;
491	struct rtentry *rt;
492	struct ifaddr *ifa;
493	struct in_ifaddr *ia;
494	struct mbuf *hold;
495	struct sockaddr sa;
496	struct in_addr isaddr, itaddr, myaddr;
497	u_int8_t *enaddr = NULL;
498	int op, flags;
499	int req_len;
500	int bridged = 0, is_bridge = 0;
501#ifdef DEV_CARP
502	int carp_match = 0;
503#endif
504	struct sockaddr_in sin;
505	sin.sin_len = sizeof(struct sockaddr_in);
506	sin.sin_family = AF_INET;
507	sin.sin_addr.s_addr = 0;
508
509	if (ifp->if_bridge)
510		bridged = 1;
511	if (ifp->if_type == IFT_BRIDGE)
512		is_bridge = 1;
513
514	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
515	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
516		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
517		return;
518	}
519
520	ah = mtod(m, struct arphdr *);
521	op = ntohs(ah->ar_op);
522	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
523	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
524
525	if (op == ARPOP_REPLY)
526		ARPSTAT_INC(rxreplies);
527
528	/*
529	 * For a bridge, we want to check the address irrespective
530	 * of the receive interface. (This will change slightly
531	 * when we have clusters of interfaces).
532	 * If the interface does not match, but the recieving interface
533	 * is part of carp, we call carp_iamatch to see if this is a
534	 * request for the virtual host ip.
535	 * XXX: This is really ugly!
536	 */
537	IN_IFADDR_RLOCK();
538	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
539		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
540		    ia->ia_ifp == ifp) &&
541		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
542			ifa_ref(&ia->ia_ifa);
543			IN_IFADDR_RUNLOCK();
544			goto match;
545		}
546#ifdef DEV_CARP
547		if (ifp->if_carp != NULL &&
548		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
549		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
550			carp_match = 1;
551			ifa_ref(&ia->ia_ifa);
552			IN_IFADDR_RUNLOCK();
553			goto match;
554		}
555#endif
556	}
557	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
558		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
559		    ia->ia_ifp == ifp) &&
560		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
561			ifa_ref(&ia->ia_ifa);
562			IN_IFADDR_RUNLOCK();
563			goto match;
564		}
565
566#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
567  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
568  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
569  addr == ia->ia_addr.sin_addr.s_addr)
570	/*
571	 * Check the case when bridge shares its MAC address with
572	 * some of its children, so packets are claimed by bridge
573	 * itself (bridge_input() does it first), but they are really
574	 * meant to be destined to the bridge member.
575	 */
576	if (is_bridge) {
577		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
578			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
579				ifa_ref(&ia->ia_ifa);
580				ifp = ia->ia_ifp;
581				IN_IFADDR_RUNLOCK();
582				goto match;
583			}
584		}
585	}
586#undef BDG_MEMBER_MATCHES_ARP
587	IN_IFADDR_RUNLOCK();
588
589	/*
590	 * No match, use the first inet address on the receive interface
591	 * as a dummy address for the rest of the function.
592	 */
593	IF_ADDR_LOCK(ifp);
594	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
595		if (ifa->ifa_addr->sa_family == AF_INET) {
596			ia = ifatoia(ifa);
597			ifa_ref(ifa);
598			IF_ADDR_UNLOCK(ifp);
599			goto match;
600		}
601	IF_ADDR_UNLOCK(ifp);
602
603	/*
604	 * If bridging, fall back to using any inet address.
605	 */
606	IN_IFADDR_RLOCK();
607	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
608		IN_IFADDR_RUNLOCK();
609		goto drop;
610	}
611	ifa_ref(&ia->ia_ifa);
612	IN_IFADDR_RUNLOCK();
613match:
614	if (!enaddr)
615		enaddr = (u_int8_t *)IF_LLADDR(ifp);
616	myaddr = ia->ia_addr.sin_addr;
617	ifa_free(&ia->ia_ifa);
618	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
619		goto drop;	/* it's from me, ignore it. */
620	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
621		log(LOG_ERR,
622		    "arp: link address is broadcast for IP address %s!\n",
623		    inet_ntoa(isaddr));
624		goto drop;
625	}
626	/*
627	 * Warn if another host is using the same IP address, but only if the
628	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
629	 * case we suppress the warning to avoid false positive complaints of
630	 * potential misconfiguration.
631	 */
632	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
633		log(LOG_ERR,
634		   "arp: %*D is using my IP address %s on %s!\n",
635		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
636		   inet_ntoa(isaddr), ifp->if_xname);
637		itaddr = myaddr;
638		ARPSTAT_INC(dupips);
639		goto reply;
640	}
641	if (ifp->if_flags & IFF_STATICARP)
642		goto reply;
643
644	bzero(&sin, sizeof(sin));
645	sin.sin_len = sizeof(struct sockaddr_in);
646	sin.sin_family = AF_INET;
647	sin.sin_addr = isaddr;
648	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
649	flags |= LLE_EXCLUSIVE;
650	IF_AFDATA_LOCK(ifp);
651	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
652	IF_AFDATA_UNLOCK(ifp);
653	if (la != NULL) {
654		/* the following is not an error when doing bridging */
655		if (!bridged && la->lle_tbl->llt_ifp != ifp
656#ifdef DEV_CARP
657		    && (ifp->if_type != IFT_CARP || !carp_match)
658#endif
659			) {
660			if (log_arp_wrong_iface)
661				log(LOG_ERR, "arp: %s is on %s "
662				    "but got reply from %*D on %s\n",
663				    inet_ntoa(isaddr),
664				    la->lle_tbl->llt_ifp->if_xname,
665				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
666				    ifp->if_xname);
667			LLE_WUNLOCK(la);
668			goto reply;
669		}
670		if ((la->la_flags & LLE_VALID) &&
671		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
672			if (la->la_flags & LLE_STATIC) {
673				LLE_WUNLOCK(la);
674				log(LOG_ERR,
675				    "arp: %*D attempts to modify permanent "
676				    "entry for %s on %s\n",
677				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
678				    inet_ntoa(isaddr), ifp->if_xname);
679				goto reply;
680			}
681			if (log_arp_movements) {
682			        log(LOG_INFO, "arp: %s moved from %*D "
683				    "to %*D on %s\n",
684				    inet_ntoa(isaddr),
685				    ifp->if_addrlen,
686				    (u_char *)&la->ll_addr, ":",
687				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
688				    ifp->if_xname);
689			}
690		}
691
692		if (ifp->if_addrlen != ah->ar_hln) {
693			LLE_WUNLOCK(la);
694			log(LOG_WARNING,
695			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
696			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
697			    ah->ar_hln, ifp->if_addrlen);
698			goto reply;
699		}
700		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
701		la->la_flags |= LLE_VALID;
702
703		EVENTHANDLER_INVOKE(arp_update_event, la);
704
705		if (!(la->la_flags & LLE_STATIC)) {
706			int canceled;
707
708			LLE_ADDREF(la);
709			la->la_expire = time_second + V_arpt_keep;
710			canceled = callout_reset(&la->la_timer,
711			    hz * V_arpt_keep, arptimer, la);
712			if (canceled)
713				LLE_REMREF(la);
714		}
715		la->la_asked = 0;
716		la->la_preempt = V_arp_maxtries;
717		hold = la->la_hold;
718		if (hold != NULL) {
719			la->la_hold = NULL;
720			memcpy(&sa, L3_ADDR(la), sizeof(sa));
721		}
722		LLE_WUNLOCK(la);
723		if (hold != NULL)
724			(*ifp->if_output)(ifp, hold, &sa, NULL);
725	}
726reply:
727	if (op != ARPOP_REQUEST)
728		goto drop;
729	ARPSTAT_INC(rxrequests);
730
731	if (itaddr.s_addr == myaddr.s_addr) {
732		/* Shortcut.. the receiving interface is the target. */
733		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
734		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
735	} else {
736		struct llentry *lle = NULL;
737
738		sin.sin_addr = itaddr;
739		IF_AFDATA_LOCK(ifp);
740		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
741		IF_AFDATA_UNLOCK(ifp);
742
743		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
744			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
745			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
746			LLE_RUNLOCK(lle);
747		} else {
748
749			if (lle != NULL)
750				LLE_RUNLOCK(lle);
751
752			if (!V_arp_proxyall)
753				goto drop;
754
755			sin.sin_addr = itaddr;
756			/* XXX MRT use table 0 for arp reply  */
757			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
758			if (!rt)
759				goto drop;
760
761			/*
762			 * Don't send proxies for nodes on the same interface
763			 * as this one came out of, or we'll get into a fight
764			 * over who claims what Ether address.
765			 */
766			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
767				RTFREE_LOCKED(rt);
768				goto drop;
769			}
770			RTFREE_LOCKED(rt);
771
772			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
773			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
774
775			/*
776			 * Also check that the node which sent the ARP packet
777			 * is on the the interface we expect it to be on. This
778			 * avoids ARP chaos if an interface is connected to the
779			 * wrong network.
780			 */
781			sin.sin_addr = isaddr;
782
783			/* XXX MRT use table 0 for arp checks */
784			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
785			if (!rt)
786				goto drop;
787			if (rt->rt_ifp != ifp) {
788				log(LOG_INFO, "arp_proxy: ignoring request"
789				    " from %s via %s, expecting %s\n",
790				    inet_ntoa(isaddr), ifp->if_xname,
791				    rt->rt_ifp->if_xname);
792				RTFREE_LOCKED(rt);
793				goto drop;
794			}
795			RTFREE_LOCKED(rt);
796
797#ifdef DEBUG_PROXY
798			printf("arp: proxying for %s\n",
799			       inet_ntoa(itaddr));
800#endif
801		}
802	}
803
804	if (itaddr.s_addr == myaddr.s_addr &&
805	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
806		/* RFC 3927 link-local IPv4; always reply by broadcast. */
807#ifdef DEBUG_LINKLOCAL
808		printf("arp: sending reply for link-local addr %s\n",
809		    inet_ntoa(itaddr));
810#endif
811		m->m_flags |= M_BCAST;
812		m->m_flags &= ~M_MCAST;
813	} else {
814		/* default behaviour; never reply by broadcast. */
815		m->m_flags &= ~(M_BCAST|M_MCAST);
816	}
817	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
818	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
819	ah->ar_op = htons(ARPOP_REPLY);
820	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
821	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
822	m->m_pkthdr.len = m->m_len;
823	sa.sa_family = AF_ARP;
824	sa.sa_len = 2;
825	(*ifp->if_output)(ifp, m, &sa, NULL);
826	ARPSTAT_INC(txreplies);
827	return;
828
829drop:
830	m_freem(m);
831}
832#endif
833
834void
835arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
836{
837	struct llentry *lle;
838
839	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
840		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
841				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
842		/*
843		 * interface address is considered static entry
844		 * because the output of the arp utility shows
845		 * that L2 entry as permanent
846		 */
847		IF_AFDATA_LOCK(ifp);
848		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
849				 (struct sockaddr *)IA_SIN(ifa));
850		IF_AFDATA_UNLOCK(ifp);
851		if (lle == NULL)
852			log(LOG_INFO, "arp_ifinit: cannot create arp "
853			    "entry for interface address\n");
854		else
855			LLE_RUNLOCK(lle);
856	}
857	ifa->ifa_rtrequest = NULL;
858}
859
860void
861arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
862{
863	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
864		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
865				&IA_SIN(ifa)->sin_addr, enaddr);
866	ifa->ifa_rtrequest = NULL;
867}
868
869static void
870arp_init(void)
871{
872
873	netisr_register(&arp_nh);
874}
875SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
876