if_ether.c revision 207369
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 207369 2010-04-29 11:52:42Z 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
85/* timer values */
86static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
87						 * minutes */
88static VNET_DEFINE(int, arp_maxtries) = 5;
89VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
90					 * local traffic */
91static VNET_DEFINE(int, arp_proxyall) = 0;
92static VNET_DEFINE(int, arpt_down) = 20;      /* keep incomplete entries for
93					       * 20 seconds */
94static VNET_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
95
96#define	V_arpt_keep		VNET(arpt_keep)
97#define	V_arpt_down		VNET(arpt_down)
98#define	V_arp_maxtries		VNET(arp_maxtries)
99#define	V_arp_proxyall		VNET(arp_proxyall)
100#define	V_arpstat		VNET(arpstat)
101
102SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
103	&VNET_NAME(arpt_keep), 0,
104	"ARP entry lifetime in seconds");
105SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
106	&VNET_NAME(arp_maxtries), 0,
107	"ARP resolution attempts before returning error");
108SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
109	&VNET_NAME(useloopback), 0,
110	"Use the loopback interface for local traffic");
111SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
112	&VNET_NAME(arp_proxyall), 0,
113	"Enable proxy ARP for all suitable requests");
114SYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW,
115	&VNET_NAME(arpstat), arpstat,
116	"ARP statistics (struct arpstat, net/if_arp.h)");
117
118static void	arp_init(void);
119void		arprequest(struct ifnet *,
120			struct in_addr *, struct in_addr *, u_char *);
121static void	arpintr(struct mbuf *);
122static void	arptimer(void *);
123#ifdef INET
124static void	in_arpinput(struct mbuf *);
125#endif
126
127static const struct netisr_handler arp_nh = {
128	.nh_name = "arp",
129	.nh_handler = arpintr,
130	.nh_proto = NETISR_ARP,
131	.nh_policy = NETISR_POLICY_SOURCE,
132};
133
134#ifdef AF_INET
135void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
136
137/*
138 * called by in_ifscrub to remove entry from the table when
139 * the interface goes away
140 */
141void
142arp_ifscrub(struct ifnet *ifp, uint32_t addr)
143{
144	struct sockaddr_in addr4;
145
146	bzero((void *)&addr4, sizeof(addr4));
147	addr4.sin_len    = sizeof(addr4);
148	addr4.sin_family = AF_INET;
149	addr4.sin_addr.s_addr = addr;
150	CURVNET_SET(ifp->if_vnet);
151	IF_AFDATA_LOCK(ifp);
152	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
153	    (struct sockaddr *)&addr4);
154	IF_AFDATA_UNLOCK(ifp);
155	CURVNET_RESTORE();
156}
157#endif
158
159/*
160 * Timeout routine.  Age arp_tab entries periodically.
161 */
162static void
163arptimer(void *arg)
164{
165	struct ifnet *ifp;
166	struct llentry   *lle = (struct llentry *)arg;
167
168	if (lle == NULL) {
169		panic("%s: NULL entry!\n", __func__);
170		return;
171	}
172	ifp = lle->lle_tbl->llt_ifp;
173	CURVNET_SET(ifp->if_vnet);
174	IF_AFDATA_LOCK(ifp);
175	LLE_WLOCK(lle);
176	if (lle->la_flags & LLE_STATIC)
177		LLE_WUNLOCK(lle);
178	else {
179		if (!callout_pending(&lle->la_timer) &&
180		    callout_active(&lle->la_timer)) {
181			callout_stop(&lle->la_timer);
182			LLE_REMREF(lle);
183			(void) llentry_free(lle);
184			ARPSTAT_INC(timeouts);
185		}
186#ifdef DIAGNOSTIC
187		else {
188			struct sockaddr *l3addr = L3_ADDR(lle);
189			log(LOG_INFO,
190			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
191			    inet_ntoa(
192			        ((const struct sockaddr_in *)l3addr)->sin_addr));
193		}
194#endif
195	}
196	IF_AFDATA_UNLOCK(ifp);
197	CURVNET_RESTORE();
198}
199
200/*
201 * Broadcast an ARP request. Caller specifies:
202 *	- arp header source ip address
203 *	- arp header target ip address
204 *	- arp header source ethernet address
205 */
206void
207arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
208    u_char *enaddr)
209{
210	struct mbuf *m;
211	struct arphdr *ah;
212	struct sockaddr sa;
213
214	if (sip == NULL) {
215		/* XXX don't believe this can happen (or explain why) */
216		/*
217		 * The caller did not supply a source address, try to find
218		 * a compatible one among those assigned to this interface.
219		 */
220		struct ifaddr *ifa;
221
222		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
223			if (!ifa->ifa_addr ||
224			    ifa->ifa_addr->sa_family != AF_INET)
225				continue;
226			sip = &SIN(ifa->ifa_addr)->sin_addr;
227			if (0 == ((sip->s_addr ^ tip->s_addr) &
228			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
229				break;  /* found it. */
230		}
231		if (sip == NULL) {
232			printf("%s: cannot find matching address\n", __func__);
233			return;
234		}
235	}
236
237	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
238		return;
239	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
240		2*ifp->if_data.ifi_addrlen;
241	m->m_pkthdr.len = m->m_len;
242	MH_ALIGN(m, m->m_len);
243	ah = mtod(m, struct arphdr *);
244	bzero((caddr_t)ah, m->m_len);
245#ifdef MAC
246	mac_netinet_arp_send(ifp, m);
247#endif
248	ah->ar_pro = htons(ETHERTYPE_IP);
249	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
250	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
251	ah->ar_op = htons(ARPOP_REQUEST);
252	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
253	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
254	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
255	sa.sa_family = AF_ARP;
256	sa.sa_len = 2;
257	m->m_flags |= M_BCAST;
258	(*ifp->if_output)(ifp, m, &sa, NULL);
259	ARPSTAT_INC(txrequests);
260}
261
262/*
263 * Resolve an IP address into an ethernet address.
264 * On input:
265 *    ifp is the interface we use
266 *    rt0 is the route to the final destination (possibly useless)
267 *    m is the mbuf. May be NULL if we don't have a packet.
268 *    dst is the next hop,
269 *    desten is where we want the address.
270 *
271 * On success, desten is filled in and the function returns 0;
272 * If the packet must be held pending resolution, we return EWOULDBLOCK
273 * On other errors, we return the corresponding error code.
274 * Note that m_freem() handles NULL.
275 */
276int
277arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
278	struct sockaddr *dst, u_char *desten, struct llentry **lle)
279{
280	struct llentry *la = 0;
281	u_int flags = 0;
282	int error, renew;
283
284	*lle = NULL;
285	if (m != NULL) {
286		if (m->m_flags & M_BCAST) {
287			/* broadcast */
288			(void)memcpy(desten,
289			    ifp->if_broadcastaddr, ifp->if_addrlen);
290			return (0);
291		}
292		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
293			/* multicast */
294			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
295			return (0);
296		}
297	}
298	/* XXXXX
299	 */
300retry:
301	IF_AFDATA_RLOCK(ifp);
302	la = lla_lookup(LLTABLE(ifp), flags, dst);
303	IF_AFDATA_RUNLOCK(ifp);
304	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
305	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
306		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
307		IF_AFDATA_WLOCK(ifp);
308		la = lla_lookup(LLTABLE(ifp), flags, dst);
309		IF_AFDATA_WUNLOCK(ifp);
310	}
311	if (la == NULL) {
312		if (flags & LLE_CREATE)
313			log(LOG_DEBUG,
314			    "arpresolve: can't allocate llinfo for %s\n",
315			    inet_ntoa(SIN(dst)->sin_addr));
316		m_freem(m);
317		return (EINVAL);
318	}
319
320	if ((la->la_flags & LLE_VALID) &&
321	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_second)) {
322		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
323		/*
324		 * If entry has an expiry time and it is approaching,
325		 * see if we need to send an ARP request within this
326		 * arpt_down interval.
327		 */
328		if (!(la->la_flags & LLE_STATIC) &&
329		    time_second + la->la_preempt > la->la_expire) {
330			arprequest(ifp, NULL,
331			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
332
333			la->la_preempt--;
334		}
335
336		*lle = la;
337		error = 0;
338		goto done;
339	}
340
341	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
342		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
343		    inet_ntoa(SIN(dst)->sin_addr));
344		m_freem(m);
345		error = EINVAL;
346		goto done;
347	}
348
349	renew = (la->la_asked == 0 || la->la_expire != time_second);
350	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
351		flags |= LLE_EXCLUSIVE;
352		LLE_RUNLOCK(la);
353		goto retry;
354	}
355	/*
356	 * There is an arptab entry, but no ethernet address
357	 * response yet.  Replace the held mbuf with this
358	 * latest one.
359	 */
360	if (m != NULL) {
361		if (la->la_hold != NULL) {
362			m_freem(la->la_hold);
363			ARPSTAT_INC(dropped);
364		}
365		la->la_hold = m;
366		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
367			flags &= ~LLE_EXCLUSIVE;
368			LLE_DOWNGRADE(la);
369		}
370
371	}
372	/*
373	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
374	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
375	 * if we have already sent arp_maxtries ARP requests. Retransmit the
376	 * ARP request, but not faster than one request per second.
377	 */
378	if (la->la_asked < V_arp_maxtries)
379		error = EWOULDBLOCK;	/* First request. */
380	else
381		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
382		    EHOSTUNREACH : EHOSTDOWN;
383
384	if (renew) {
385		int canceled;
386
387		LLE_ADDREF(la);
388		la->la_expire = time_second + V_arpt_down;
389		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
390		    arptimer, la);
391		if (canceled)
392			LLE_REMREF(la);
393		la->la_asked++;
394		LLE_WUNLOCK(la);
395		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
396		    IF_LLADDR(ifp));
397		return (error);
398	}
399done:
400	if (flags & LLE_EXCLUSIVE)
401		LLE_WUNLOCK(la);
402	else
403		LLE_RUNLOCK(la);
404	return (error);
405}
406
407/*
408 * Common length and type checks are done here,
409 * then the protocol-specific routine is called.
410 */
411static void
412arpintr(struct mbuf *m)
413{
414	struct arphdr *ar;
415
416	if (m->m_len < sizeof(struct arphdr) &&
417	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
418		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
419		return;
420	}
421	ar = mtod(m, struct arphdr *);
422
423	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
424	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
425	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
426	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
427		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
428		    (unsigned char *)&ar->ar_hrd, "");
429		m_freem(m);
430		return;
431	}
432
433	if (m->m_len < arphdr_len(ar)) {
434		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
435			log(LOG_ERR, "arp: runt packet\n");
436			m_freem(m);
437			return;
438		}
439		ar = mtod(m, struct arphdr *);
440	}
441
442	ARPSTAT_INC(received);
443	switch (ntohs(ar->ar_pro)) {
444#ifdef INET
445	case ETHERTYPE_IP:
446		in_arpinput(m);
447		return;
448#endif
449	}
450	m_freem(m);
451}
452
453#ifdef INET
454/*
455 * ARP for Internet protocols on 10 Mb/s Ethernet.
456 * Algorithm is that given in RFC 826.
457 * In addition, a sanity check is performed on the sender
458 * protocol address, to catch impersonators.
459 * We no longer handle negotiations for use of trailer protocol:
460 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
461 * along with IP replies if we wanted trailers sent to us,
462 * and also sent them in response to IP replies.
463 * This allowed either end to announce the desire to receive
464 * trailer packets.
465 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
466 * but formerly didn't normally send requests.
467 */
468static int log_arp_wrong_iface = 1;
469static int log_arp_movements = 1;
470static int log_arp_permanent_modify = 1;
471
472SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
473	&log_arp_wrong_iface, 0,
474	"log arp packets arriving on the wrong interface");
475SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
476        &log_arp_movements, 0,
477        "log arp replies from MACs different than the one in the cache");
478SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
479        &log_arp_permanent_modify, 0,
480        "log arp replies from MACs different than the one in the permanent arp entry");
481
482
483static void
484in_arpinput(struct mbuf *m)
485{
486	struct arphdr *ah;
487	struct ifnet *ifp = m->m_pkthdr.rcvif;
488	struct llentry *la = NULL;
489	struct rtentry *rt;
490	struct ifaddr *ifa;
491	struct in_ifaddr *ia;
492	struct mbuf *hold;
493	struct sockaddr sa;
494	struct in_addr isaddr, itaddr, myaddr;
495	u_int8_t *enaddr = NULL;
496	int op, flags;
497	int req_len;
498	int bridged = 0, is_bridge = 0;
499#ifdef DEV_CARP
500	int carp_match = 0;
501#endif
502	struct sockaddr_in sin;
503	sin.sin_len = sizeof(struct sockaddr_in);
504	sin.sin_family = AF_INET;
505	sin.sin_addr.s_addr = 0;
506
507	if (ifp->if_bridge)
508		bridged = 1;
509	if (ifp->if_type == IFT_BRIDGE)
510		is_bridge = 1;
511
512	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
513	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
514		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
515		return;
516	}
517
518	ah = mtod(m, struct arphdr *);
519	op = ntohs(ah->ar_op);
520	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
521	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
522
523	if (op == ARPOP_REPLY)
524		ARPSTAT_INC(rxreplies);
525
526	/*
527	 * For a bridge, we want to check the address irrespective
528	 * of the receive interface. (This will change slightly
529	 * when we have clusters of interfaces).
530	 * If the interface does not match, but the recieving interface
531	 * is part of carp, we call carp_iamatch to see if this is a
532	 * request for the virtual host ip.
533	 * XXX: This is really ugly!
534	 */
535	IN_IFADDR_RLOCK();
536	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
537		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
538		    ia->ia_ifp == ifp) &&
539		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
540			ifa_ref(&ia->ia_ifa);
541			IN_IFADDR_RUNLOCK();
542			goto match;
543		}
544#ifdef DEV_CARP
545		if (ifp->if_carp != NULL &&
546		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
547		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
548			carp_match = 1;
549			ifa_ref(&ia->ia_ifa);
550			IN_IFADDR_RUNLOCK();
551			goto match;
552		}
553#endif
554	}
555	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
556		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
557		    ia->ia_ifp == ifp) &&
558		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
559			ifa_ref(&ia->ia_ifa);
560			IN_IFADDR_RUNLOCK();
561			goto match;
562		}
563
564#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
565  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
566  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
567  addr == ia->ia_addr.sin_addr.s_addr)
568	/*
569	 * Check the case when bridge shares its MAC address with
570	 * some of its children, so packets are claimed by bridge
571	 * itself (bridge_input() does it first), but they are really
572	 * meant to be destined to the bridge member.
573	 */
574	if (is_bridge) {
575		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
576			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
577				ifa_ref(&ia->ia_ifa);
578				ifp = ia->ia_ifp;
579				IN_IFADDR_RUNLOCK();
580				goto match;
581			}
582		}
583	}
584#undef BDG_MEMBER_MATCHES_ARP
585	IN_IFADDR_RUNLOCK();
586
587	/*
588	 * No match, use the first inet address on the receive interface
589	 * as a dummy address for the rest of the function.
590	 */
591	IF_ADDR_LOCK(ifp);
592	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
593		if (ifa->ifa_addr->sa_family == AF_INET) {
594			ia = ifatoia(ifa);
595			ifa_ref(ifa);
596			IF_ADDR_UNLOCK(ifp);
597			goto match;
598		}
599	IF_ADDR_UNLOCK(ifp);
600
601	/*
602	 * If bridging, fall back to using any inet address.
603	 */
604	IN_IFADDR_RLOCK();
605	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
606		IN_IFADDR_RUNLOCK();
607		goto drop;
608	}
609	ifa_ref(&ia->ia_ifa);
610	IN_IFADDR_RUNLOCK();
611match:
612	if (!enaddr)
613		enaddr = (u_int8_t *)IF_LLADDR(ifp);
614	myaddr = ia->ia_addr.sin_addr;
615	ifa_free(&ia->ia_ifa);
616	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
617		goto drop;	/* it's from me, ignore it. */
618	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
619		log(LOG_ERR,
620		    "arp: link address is broadcast for IP address %s!\n",
621		    inet_ntoa(isaddr));
622		goto drop;
623	}
624	/*
625	 * Warn if another host is using the same IP address, but only if the
626	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
627	 * case we suppress the warning to avoid false positive complaints of
628	 * potential misconfiguration.
629	 */
630	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
631		log(LOG_ERR,
632		   "arp: %*D is using my IP address %s on %s!\n",
633		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
634		   inet_ntoa(isaddr), ifp->if_xname);
635		itaddr = myaddr;
636		ARPSTAT_INC(dupips);
637		goto reply;
638	}
639	if (ifp->if_flags & IFF_STATICARP)
640		goto reply;
641
642	bzero(&sin, sizeof(sin));
643	sin.sin_len = sizeof(struct sockaddr_in);
644	sin.sin_family = AF_INET;
645	sin.sin_addr = isaddr;
646	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
647	flags |= LLE_EXCLUSIVE;
648	IF_AFDATA_LOCK(ifp);
649	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
650	IF_AFDATA_UNLOCK(ifp);
651	if (la != NULL) {
652		/* the following is not an error when doing bridging */
653		if (!bridged && la->lle_tbl->llt_ifp != ifp
654#ifdef DEV_CARP
655		    && (ifp->if_type != IFT_CARP || !carp_match)
656#endif
657			) {
658			if (log_arp_wrong_iface)
659				log(LOG_ERR, "arp: %s is on %s "
660				    "but got reply from %*D on %s\n",
661				    inet_ntoa(isaddr),
662				    la->lle_tbl->llt_ifp->if_xname,
663				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
664				    ifp->if_xname);
665			LLE_WUNLOCK(la);
666			goto reply;
667		}
668		if ((la->la_flags & LLE_VALID) &&
669		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
670			if (la->la_flags & LLE_STATIC) {
671				LLE_WUNLOCK(la);
672				log(LOG_ERR,
673				    "arp: %*D attempts to modify permanent "
674				    "entry for %s on %s\n",
675				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
676				    inet_ntoa(isaddr), ifp->if_xname);
677				goto reply;
678			}
679			if (log_arp_movements) {
680			        log(LOG_INFO, "arp: %s moved from %*D "
681				    "to %*D on %s\n",
682				    inet_ntoa(isaddr),
683				    ifp->if_addrlen,
684				    (u_char *)&la->ll_addr, ":",
685				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
686				    ifp->if_xname);
687			}
688		}
689
690		if (ifp->if_addrlen != ah->ar_hln) {
691			LLE_WUNLOCK(la);
692			log(LOG_WARNING,
693			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
694			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
695			    ah->ar_hln, ifp->if_addrlen);
696			goto reply;
697		}
698		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
699		la->la_flags |= LLE_VALID;
700
701		EVENTHANDLER_INVOKE(arp_update_event, la);
702
703		if (!(la->la_flags & LLE_STATIC)) {
704			int canceled;
705
706			LLE_ADDREF(la);
707			la->la_expire = time_second + V_arpt_keep;
708			canceled = callout_reset(&la->la_timer,
709			    hz * V_arpt_keep, arptimer, la);
710			if (canceled)
711				LLE_REMREF(la);
712		}
713		la->la_asked = 0;
714		la->la_preempt = V_arp_maxtries;
715		hold = la->la_hold;
716		if (hold != NULL) {
717			la->la_hold = NULL;
718			memcpy(&sa, L3_ADDR(la), sizeof(sa));
719		}
720		LLE_WUNLOCK(la);
721		if (hold != NULL)
722			(*ifp->if_output)(ifp, hold, &sa, NULL);
723	}
724reply:
725	if (op != ARPOP_REQUEST)
726		goto drop;
727	ARPSTAT_INC(rxrequests);
728
729	if (itaddr.s_addr == myaddr.s_addr) {
730		/* Shortcut.. the receiving interface is the target. */
731		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
732		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
733	} else {
734		struct llentry *lle = NULL;
735
736		sin.sin_addr = itaddr;
737		IF_AFDATA_LOCK(ifp);
738		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
739		IF_AFDATA_UNLOCK(ifp);
740
741		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
742			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
743			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
744			LLE_RUNLOCK(lle);
745		} else {
746
747			if (lle != NULL)
748				LLE_RUNLOCK(lle);
749
750			if (!V_arp_proxyall)
751				goto drop;
752
753			sin.sin_addr = itaddr;
754			/* XXX MRT use table 0 for arp reply  */
755			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
756			if (!rt)
757				goto drop;
758
759			/*
760			 * Don't send proxies for nodes on the same interface
761			 * as this one came out of, or we'll get into a fight
762			 * over who claims what Ether address.
763			 */
764			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
765				RTFREE_LOCKED(rt);
766				goto drop;
767			}
768			RTFREE_LOCKED(rt);
769
770			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
771			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
772
773			/*
774			 * Also check that the node which sent the ARP packet
775			 * is on the the interface we expect it to be on. This
776			 * avoids ARP chaos if an interface is connected to the
777			 * wrong network.
778			 */
779			sin.sin_addr = isaddr;
780
781			/* XXX MRT use table 0 for arp checks */
782			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
783			if (!rt)
784				goto drop;
785			if (rt->rt_ifp != ifp) {
786				log(LOG_INFO, "arp_proxy: ignoring request"
787				    " from %s via %s, expecting %s\n",
788				    inet_ntoa(isaddr), ifp->if_xname,
789				    rt->rt_ifp->if_xname);
790				RTFREE_LOCKED(rt);
791				goto drop;
792			}
793			RTFREE_LOCKED(rt);
794
795#ifdef DEBUG_PROXY
796			printf("arp: proxying for %s\n",
797			       inet_ntoa(itaddr));
798#endif
799		}
800	}
801
802	if (itaddr.s_addr == myaddr.s_addr &&
803	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
804		/* RFC 3927 link-local IPv4; always reply by broadcast. */
805#ifdef DEBUG_LINKLOCAL
806		printf("arp: sending reply for link-local addr %s\n",
807		    inet_ntoa(itaddr));
808#endif
809		m->m_flags |= M_BCAST;
810		m->m_flags &= ~M_MCAST;
811	} else {
812		/* default behaviour; never reply by broadcast. */
813		m->m_flags &= ~(M_BCAST|M_MCAST);
814	}
815	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
816	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
817	ah->ar_op = htons(ARPOP_REPLY);
818	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
819	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
820	m->m_pkthdr.len = m->m_len;
821	sa.sa_family = AF_ARP;
822	sa.sa_len = 2;
823	(*ifp->if_output)(ifp, m, &sa, NULL);
824	ARPSTAT_INC(txreplies);
825	return;
826
827drop:
828	m_freem(m);
829}
830#endif
831
832void
833arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
834{
835	struct llentry *lle;
836
837	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
838		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
839				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
840		/*
841		 * interface address is considered static entry
842		 * because the output of the arp utility shows
843		 * that L2 entry as permanent
844		 */
845		IF_AFDATA_LOCK(ifp);
846		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
847				 (struct sockaddr *)IA_SIN(ifa));
848		IF_AFDATA_UNLOCK(ifp);
849		if (lle == NULL)
850			log(LOG_INFO, "arp_ifinit: cannot create arp "
851			    "entry for interface address\n");
852		else
853			LLE_RUNLOCK(lle);
854	}
855	ifa->ifa_rtrequest = NULL;
856}
857
858void
859arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
860{
861	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
862		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
863				&IA_SIN(ifa)->sin_addr, enaddr);
864	ifa->ifa_rtrequest = NULL;
865}
866
867static void
868arp_init(void)
869{
870
871	netisr_register(&arp_nh);
872}
873SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
874