if_ether.c revision 223261
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 223261 2011-06-18 13:54:36Z bz $");
40
41#include "opt_inet.h"
42
43#include <sys/param.h>
44#include <sys/kernel.h>
45#include <sys/queue.h>
46#include <sys/sysctl.h>
47#include <sys/systm.h>
48#include <sys/mbuf.h>
49#include <sys/malloc.h>
50#include <sys/proc.h>
51#include <sys/socket.h>
52#include <sys/syslog.h>
53
54#include <net/if.h>
55#include <net/if_dl.h>
56#include <net/if_types.h>
57#include <net/netisr.h>
58#include <net/if_llc.h>
59#include <net/ethernet.h>
60#include <net/route.h>
61#include <net/vnet.h>
62
63#include <netinet/in.h>
64#include <netinet/in_var.h>
65#include <net/if_llatbl.h>
66#include <netinet/if_ether.h>
67#if defined(INET) || defined(INET6)
68#include <netinet/ip_carp.h>
69#endif
70
71#include <net/if_arc.h>
72#include <net/iso88025.h>
73
74#include <security/mac/mac_framework.h>
75
76#define SIN(s) ((struct sockaddr_in *)s)
77#define SDL(s) ((struct sockaddr_dl *)s)
78
79SYSCTL_DECL(_net_link_ether);
80SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
82
83/* timer values */
84static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
85						 * minutes */
86static VNET_DEFINE(int, arp_maxtries) = 5;
87VNET_DEFINE(int, useloopback) = 1;	/* use loopback interface for
88					 * local traffic */
89static VNET_DEFINE(int, arp_proxyall) = 0;
90static VNET_DEFINE(int, arpt_down) = 20;      /* keep incomplete entries for
91					       * 20 seconds */
92VNET_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
93
94static VNET_DEFINE(int, arp_maxhold) = 1;
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#define	V_arp_maxhold		VNET(arp_maxhold)
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");
106SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
107	&VNET_NAME(arp_maxtries), 0,
108	"ARP resolution attempts before returning error");
109SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
110	&VNET_NAME(useloopback), 0,
111	"Use the loopback interface for local traffic");
112SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
113	&VNET_NAME(arp_proxyall), 0,
114	"Enable proxy ARP for all suitable requests");
115SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW,
116	&VNET_NAME(arpt_down), 0,
117	"Incomplete ARP entry lifetime in seconds");
118SYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW,
119	&VNET_NAME(arpstat), arpstat,
120	"ARP statistics (struct arpstat, net/if_arp.h)");
121SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW,
122	&VNET_NAME(arp_maxhold), 0,
123	"Number of packets to hold per ARP entry");
124
125static void	arp_init(void);
126void		arprequest(struct ifnet *,
127			struct in_addr *, struct in_addr *, u_char *);
128static void	arpintr(struct mbuf *);
129static void	arptimer(void *);
130#ifdef INET
131static void	in_arpinput(struct mbuf *);
132#endif
133
134static const struct netisr_handler arp_nh = {
135	.nh_name = "arp",
136	.nh_handler = arpintr,
137	.nh_proto = NETISR_ARP,
138	.nh_policy = NETISR_POLICY_SOURCE,
139};
140
141#ifdef AF_INET
142void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
143
144/*
145 * called by in_ifscrub to remove entry from the table when
146 * the interface goes away
147 */
148void
149arp_ifscrub(struct ifnet *ifp, uint32_t addr)
150{
151	struct sockaddr_in addr4;
152
153	bzero((void *)&addr4, sizeof(addr4));
154	addr4.sin_len    = sizeof(addr4);
155	addr4.sin_family = AF_INET;
156	addr4.sin_addr.s_addr = addr;
157	IF_AFDATA_LOCK(ifp);
158	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
159	    (struct sockaddr *)&addr4);
160	IF_AFDATA_UNLOCK(ifp);
161}
162#endif
163
164/*
165 * Timeout routine.  Age arp_tab entries periodically.
166 */
167static void
168arptimer(void *arg)
169{
170	struct ifnet *ifp;
171	struct llentry   *lle;
172	int pkts_dropped;
173
174	KASSERT(arg != NULL, ("%s: arg NULL", __func__));
175	lle = (struct llentry *)arg;
176	ifp = lle->lle_tbl->llt_ifp;
177	CURVNET_SET(ifp->if_vnet);
178	IF_AFDATA_LOCK(ifp);
179	LLE_WLOCK(lle);
180	if (lle->la_flags & LLE_STATIC)
181		LLE_WUNLOCK(lle);
182	else {
183		if (!callout_pending(&lle->la_timer) &&
184		    callout_active(&lle->la_timer)) {
185			callout_stop(&lle->la_timer);
186			LLE_REMREF(lle);
187			pkts_dropped = llentry_free(lle);
188			ARPSTAT_ADD(dropped, pkts_dropped);
189			ARPSTAT_INC(timeouts);
190		} else {
191#ifdef DIAGNOSTIC
192			struct sockaddr *l3addr = L3_ADDR(lle);
193			log(LOG_INFO,
194			    "arptimer issue: %p, IPv4 address: \"%s\"\n", lle,
195			    inet_ntoa(
196			        ((const struct sockaddr_in *)l3addr)->sin_addr));
197#endif
198			LLE_WUNLOCK(lle);
199		}
200	}
201	IF_AFDATA_UNLOCK(ifp);
202	CURVNET_RESTORE();
203}
204
205/*
206 * Broadcast an ARP request. Caller specifies:
207 *	- arp header source ip address
208 *	- arp header target ip address
209 *	- arp header source ethernet address
210 */
211void
212arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
213    u_char *enaddr)
214{
215	struct mbuf *m;
216	struct arphdr *ah;
217	struct sockaddr sa;
218
219	if (sip == NULL) {
220		/* XXX don't believe this can happen (or explain why) */
221		/*
222		 * The caller did not supply a source address, try to find
223		 * a compatible one among those assigned to this interface.
224		 */
225		struct ifaddr *ifa;
226
227		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
228			if (!ifa->ifa_addr ||
229			    ifa->ifa_addr->sa_family != AF_INET)
230				continue;
231			sip = &SIN(ifa->ifa_addr)->sin_addr;
232			if (0 == ((sip->s_addr ^ tip->s_addr) &
233			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
234				break;  /* found it. */
235		}
236		if (sip == NULL) {
237			printf("%s: cannot find matching address\n", __func__);
238			return;
239		}
240	}
241
242	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
243		return;
244	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
245		2*ifp->if_data.ifi_addrlen;
246	m->m_pkthdr.len = m->m_len;
247	MH_ALIGN(m, m->m_len);
248	ah = mtod(m, struct arphdr *);
249	bzero((caddr_t)ah, m->m_len);
250#ifdef MAC
251	mac_netinet_arp_send(ifp, m);
252#endif
253	ah->ar_pro = htons(ETHERTYPE_IP);
254	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
255	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
256	ah->ar_op = htons(ARPOP_REQUEST);
257	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
258	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
259	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
260	sa.sa_family = AF_ARP;
261	sa.sa_len = 2;
262	m->m_flags |= M_BCAST;
263	(*ifp->if_output)(ifp, m, &sa, NULL);
264	ARPSTAT_INC(txrequests);
265}
266
267/*
268 * Resolve an IP address into an ethernet address.
269 * On input:
270 *    ifp is the interface we use
271 *    rt0 is the route to the final destination (possibly useless)
272 *    m is the mbuf. May be NULL if we don't have a packet.
273 *    dst is the next hop,
274 *    desten is where we want the address.
275 *
276 * On success, desten is filled in and the function returns 0;
277 * If the packet must be held pending resolution, we return EWOULDBLOCK
278 * On other errors, we return the corresponding error code.
279 * Note that m_freem() handles NULL.
280 */
281int
282arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
283	struct sockaddr *dst, u_char *desten, struct llentry **lle)
284{
285	struct llentry *la = 0;
286	u_int flags = 0;
287	struct mbuf *curr = NULL;
288	struct mbuf *next = NULL;
289	int error, renew;
290
291	*lle = NULL;
292	if (m != NULL) {
293		if (m->m_flags & M_BCAST) {
294			/* broadcast */
295			(void)memcpy(desten,
296			    ifp->if_broadcastaddr, ifp->if_addrlen);
297			return (0);
298		}
299		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
300			/* multicast */
301			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
302			return (0);
303		}
304	}
305retry:
306	IF_AFDATA_RLOCK(ifp);
307	la = lla_lookup(LLTABLE(ifp), flags, dst);
308	IF_AFDATA_RUNLOCK(ifp);
309	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
310	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
311		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
312		IF_AFDATA_WLOCK(ifp);
313		la = lla_lookup(LLTABLE(ifp), flags, dst);
314		IF_AFDATA_WUNLOCK(ifp);
315	}
316	if (la == NULL) {
317		if (flags & LLE_CREATE)
318			log(LOG_DEBUG,
319			    "arpresolve: can't allocate llinfo for %s\n",
320			    inet_ntoa(SIN(dst)->sin_addr));
321		m_freem(m);
322		return (EINVAL);
323	}
324
325	if ((la->la_flags & LLE_VALID) &&
326	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
327		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
328		/*
329		 * If entry has an expiry time and it is approaching,
330		 * see if we need to send an ARP request within this
331		 * arpt_down interval.
332		 */
333		if (!(la->la_flags & LLE_STATIC) &&
334		    time_uptime + la->la_preempt > la->la_expire) {
335			arprequest(ifp, NULL,
336			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
337
338			la->la_preempt--;
339		}
340
341		*lle = la;
342		error = 0;
343		goto done;
344	}
345
346	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
347		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
348		    inet_ntoa(SIN(dst)->sin_addr));
349		m_freem(m);
350		error = EINVAL;
351		goto done;
352	}
353
354	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
355	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
356		flags |= LLE_EXCLUSIVE;
357		LLE_RUNLOCK(la);
358		goto retry;
359	}
360	/*
361	 * There is an arptab entry, but no ethernet address
362	 * response yet.  Add the mbuf to the list, dropping
363	 * the oldest packet if we have exceeded the system
364	 * setting.
365	 */
366	if (m != NULL) {
367		if (la->la_numheld >= V_arp_maxhold) {
368			if (la->la_hold != NULL) {
369				next = la->la_hold->m_nextpkt;
370				m_freem(la->la_hold);
371				la->la_hold = next;
372				la->la_numheld--;
373				ARPSTAT_INC(dropped);
374			}
375		}
376		if (la->la_hold != NULL) {
377			curr = la->la_hold;
378			while (curr->m_nextpkt != NULL)
379				curr = curr->m_nextpkt;
380			curr->m_nextpkt = m;
381		} else
382			la->la_hold = m;
383		la->la_numheld++;
384		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
385			flags &= ~LLE_EXCLUSIVE;
386			LLE_DOWNGRADE(la);
387		}
388
389	}
390	/*
391	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
392	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
393	 * if we have already sent arp_maxtries ARP requests. Retransmit the
394	 * ARP request, but not faster than one request per second.
395	 */
396	if (la->la_asked < V_arp_maxtries)
397		error = EWOULDBLOCK;	/* First request. */
398	else
399		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
400		    EHOSTUNREACH : EHOSTDOWN;
401
402	if (renew) {
403		int canceled;
404
405		LLE_ADDREF(la);
406		la->la_expire = time_uptime;
407		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
408		    arptimer, la);
409		if (canceled)
410			LLE_REMREF(la);
411		la->la_asked++;
412		LLE_WUNLOCK(la);
413		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
414		    IF_LLADDR(ifp));
415		return (error);
416	}
417done:
418	if (flags & LLE_EXCLUSIVE)
419		LLE_WUNLOCK(la);
420	else
421		LLE_RUNLOCK(la);
422	return (error);
423}
424
425/*
426 * Common length and type checks are done here,
427 * then the protocol-specific routine is called.
428 */
429static void
430arpintr(struct mbuf *m)
431{
432	struct arphdr *ar;
433
434	if (m->m_len < sizeof(struct arphdr) &&
435	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
436		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
437		return;
438	}
439	ar = mtod(m, struct arphdr *);
440
441	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
442	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
443	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
444	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
445	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
446		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
447		    (unsigned char *)&ar->ar_hrd, "");
448		m_freem(m);
449		return;
450	}
451
452	if (m->m_len < arphdr_len(ar)) {
453		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
454			log(LOG_ERR, "arp: runt packet\n");
455			m_freem(m);
456			return;
457		}
458		ar = mtod(m, struct arphdr *);
459	}
460
461	ARPSTAT_INC(received);
462	switch (ntohs(ar->ar_pro)) {
463#ifdef INET
464	case ETHERTYPE_IP:
465		in_arpinput(m);
466		return;
467#endif
468	}
469	m_freem(m);
470}
471
472#ifdef INET
473/*
474 * ARP for Internet protocols on 10 Mb/s Ethernet.
475 * Algorithm is that given in RFC 826.
476 * In addition, a sanity check is performed on the sender
477 * protocol address, to catch impersonators.
478 * We no longer handle negotiations for use of trailer protocol:
479 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
480 * along with IP replies if we wanted trailers sent to us,
481 * and also sent them in response to IP replies.
482 * This allowed either end to announce the desire to receive
483 * trailer packets.
484 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
485 * but formerly didn't normally send requests.
486 */
487static int log_arp_wrong_iface = 1;
488static int log_arp_movements = 1;
489static int log_arp_permanent_modify = 1;
490
491SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
492	&log_arp_wrong_iface, 0,
493	"log arp packets arriving on the wrong interface");
494SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
495        &log_arp_movements, 0,
496        "log arp replies from MACs different than the one in the cache");
497SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
498        &log_arp_permanent_modify, 0,
499        "log arp replies from MACs different than the one in the permanent arp entry");
500
501
502static void
503in_arpinput(struct mbuf *m)
504{
505	struct arphdr *ah;
506	struct ifnet *ifp = m->m_pkthdr.rcvif;
507	struct llentry *la = NULL;
508	struct rtentry *rt;
509	struct ifaddr *ifa;
510	struct in_ifaddr *ia;
511	struct sockaddr sa;
512	struct in_addr isaddr, itaddr, myaddr;
513	u_int8_t *enaddr = NULL;
514	int op, flags;
515	int req_len;
516	int bridged = 0, is_bridge = 0;
517	int carp_match = 0;
518	struct sockaddr_in sin;
519	sin.sin_len = sizeof(struct sockaddr_in);
520	sin.sin_family = AF_INET;
521	sin.sin_addr.s_addr = 0;
522
523	if (ifp->if_bridge)
524		bridged = 1;
525	if (ifp->if_type == IFT_BRIDGE)
526		is_bridge = 1;
527
528	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
529	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
530		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
531		return;
532	}
533
534	ah = mtod(m, struct arphdr *);
535	/*
536	 * ARP is only for IPv4 so we can reject packets with
537	 * a protocol length not equal to an IPv4 address.
538	 */
539	if (ah->ar_pln != sizeof(struct in_addr)) {
540		log(LOG_ERR, "in_arp: requested protocol length != %zu\n",
541		    sizeof(struct in_addr));
542		return;
543	}
544
545	if (ETHER_IS_MULTICAST(ar_sha(ah))) {
546		log(LOG_ERR, "in_arp: source hardware address is multicast.");
547		return;
548	}
549
550	op = ntohs(ah->ar_op);
551	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
552	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
553
554	if (op == ARPOP_REPLY)
555		ARPSTAT_INC(rxreplies);
556
557	/*
558	 * For a bridge, we want to check the address irrespective
559	 * of the receive interface. (This will change slightly
560	 * when we have clusters of interfaces).
561	 * If the interface does not match, but the recieving interface
562	 * is part of carp, we call carp_iamatch to see if this is a
563	 * request for the virtual host ip.
564	 * XXX: This is really ugly!
565	 */
566	IN_IFADDR_RLOCK();
567	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
568		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
569		    ia->ia_ifp == ifp) &&
570		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
571			ifa_ref(&ia->ia_ifa);
572			IN_IFADDR_RUNLOCK();
573			goto match;
574		}
575		if (ifp->if_carp != NULL &&
576		    (*carp_iamatch_p)(ifp, ia, &isaddr, &enaddr) &&
577		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
578			carp_match = 1;
579			ifa_ref(&ia->ia_ifa);
580			IN_IFADDR_RUNLOCK();
581			goto match;
582		}
583	}
584	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
585		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
586		    ia->ia_ifp == ifp) &&
587		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
588			ifa_ref(&ia->ia_ifa);
589			IN_IFADDR_RUNLOCK();
590			goto match;
591		}
592
593#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
594  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
595  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
596  addr == ia->ia_addr.sin_addr.s_addr)
597	/*
598	 * Check the case when bridge shares its MAC address with
599	 * some of its children, so packets are claimed by bridge
600	 * itself (bridge_input() does it first), but they are really
601	 * meant to be destined to the bridge member.
602	 */
603	if (is_bridge) {
604		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
605			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
606				ifa_ref(&ia->ia_ifa);
607				ifp = ia->ia_ifp;
608				IN_IFADDR_RUNLOCK();
609				goto match;
610			}
611		}
612	}
613#undef BDG_MEMBER_MATCHES_ARP
614	IN_IFADDR_RUNLOCK();
615
616	/*
617	 * No match, use the first inet address on the receive interface
618	 * as a dummy address for the rest of the function.
619	 */
620	IF_ADDR_LOCK(ifp);
621	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
622		if (ifa->ifa_addr->sa_family == AF_INET) {
623			ia = ifatoia(ifa);
624			ifa_ref(ifa);
625			IF_ADDR_UNLOCK(ifp);
626			goto match;
627		}
628	IF_ADDR_UNLOCK(ifp);
629
630	/*
631	 * If bridging, fall back to using any inet address.
632	 */
633	IN_IFADDR_RLOCK();
634	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
635		IN_IFADDR_RUNLOCK();
636		goto drop;
637	}
638	ifa_ref(&ia->ia_ifa);
639	IN_IFADDR_RUNLOCK();
640match:
641	if (!enaddr)
642		enaddr = (u_int8_t *)IF_LLADDR(ifp);
643	myaddr = ia->ia_addr.sin_addr;
644	ifa_free(&ia->ia_ifa);
645	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
646		goto drop;	/* it's from me, ignore it. */
647	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
648		log(LOG_ERR,
649		    "arp: link address is broadcast for IP address %s!\n",
650		    inet_ntoa(isaddr));
651		goto drop;
652	}
653	/*
654	 * Warn if another host is using the same IP address, but only if the
655	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
656	 * case we suppress the warning to avoid false positive complaints of
657	 * potential misconfiguration.
658	 */
659	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
660		log(LOG_ERR,
661		   "arp: %*D is using my IP address %s on %s!\n",
662		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
663		   inet_ntoa(isaddr), ifp->if_xname);
664		itaddr = myaddr;
665		ARPSTAT_INC(dupips);
666		goto reply;
667	}
668	if (ifp->if_flags & IFF_STATICARP)
669		goto reply;
670
671	bzero(&sin, sizeof(sin));
672	sin.sin_len = sizeof(struct sockaddr_in);
673	sin.sin_family = AF_INET;
674	sin.sin_addr = isaddr;
675	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
676	flags |= LLE_EXCLUSIVE;
677	IF_AFDATA_LOCK(ifp);
678	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
679	IF_AFDATA_UNLOCK(ifp);
680	if (la != NULL) {
681		/* the following is not an error when doing bridging */
682		if (!bridged && la->lle_tbl->llt_ifp != ifp && !carp_match) {
683			if (log_arp_wrong_iface)
684				log(LOG_ERR, "arp: %s is on %s "
685				    "but got reply from %*D on %s\n",
686				    inet_ntoa(isaddr),
687				    la->lle_tbl->llt_ifp->if_xname,
688				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
689				    ifp->if_xname);
690			LLE_WUNLOCK(la);
691			goto reply;
692		}
693		if ((la->la_flags & LLE_VALID) &&
694		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
695			if (la->la_flags & LLE_STATIC) {
696				LLE_WUNLOCK(la);
697				log(LOG_ERR,
698				    "arp: %*D attempts to modify permanent "
699				    "entry for %s on %s\n",
700				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
701				    inet_ntoa(isaddr), ifp->if_xname);
702				goto reply;
703			}
704			if (log_arp_movements) {
705			        log(LOG_INFO, "arp: %s moved from %*D "
706				    "to %*D on %s\n",
707				    inet_ntoa(isaddr),
708				    ifp->if_addrlen,
709				    (u_char *)&la->ll_addr, ":",
710				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
711				    ifp->if_xname);
712			}
713		}
714
715		if (ifp->if_addrlen != ah->ar_hln) {
716			LLE_WUNLOCK(la);
717			log(LOG_WARNING,
718			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
719			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
720			    ah->ar_hln, ifp->if_addrlen);
721			goto drop;
722		}
723		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
724		la->la_flags |= LLE_VALID;
725
726		EVENTHANDLER_INVOKE(arp_update_event, la);
727
728		if (!(la->la_flags & LLE_STATIC)) {
729			int canceled;
730
731			LLE_ADDREF(la);
732			la->la_expire = time_uptime + V_arpt_keep;
733			canceled = callout_reset(&la->la_timer,
734			    hz * V_arpt_keep, arptimer, la);
735			if (canceled)
736				LLE_REMREF(la);
737		}
738		la->la_asked = 0;
739		la->la_preempt = V_arp_maxtries;
740		/*
741		 * The packets are all freed within the call to the output
742		 * routine.
743		 *
744		 * NB: The lock MUST be released before the call to the
745		 * output routine.
746		 */
747		if (la->la_hold != NULL) {
748			struct mbuf *m_hold, *m_hold_next;
749
750			m_hold = la->la_hold;
751			la->la_hold = NULL;
752			la->la_numheld = 0;
753			memcpy(&sa, L3_ADDR(la), sizeof(sa));
754			LLE_WUNLOCK(la);
755			for (; m_hold != NULL; m_hold = m_hold_next) {
756				m_hold_next = m_hold->m_nextpkt;
757				m_hold->m_nextpkt = NULL;
758				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
759			}
760		} else
761			LLE_WUNLOCK(la);
762	}
763reply:
764	if (op != ARPOP_REQUEST)
765		goto drop;
766	ARPSTAT_INC(rxrequests);
767
768	if (itaddr.s_addr == myaddr.s_addr) {
769		/* Shortcut.. the receiving interface is the target. */
770		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
771		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
772	} else {
773		struct llentry *lle = NULL;
774
775		sin.sin_addr = itaddr;
776		IF_AFDATA_LOCK(ifp);
777		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
778		IF_AFDATA_UNLOCK(ifp);
779
780		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
781			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
782			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
783			LLE_RUNLOCK(lle);
784		} else {
785
786			if (lle != NULL)
787				LLE_RUNLOCK(lle);
788
789			if (!V_arp_proxyall)
790				goto drop;
791
792			sin.sin_addr = itaddr;
793			/* XXX MRT use table 0 for arp reply  */
794			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
795			if (!rt)
796				goto drop;
797
798			/*
799			 * Don't send proxies for nodes on the same interface
800			 * as this one came out of, or we'll get into a fight
801			 * over who claims what Ether address.
802			 */
803			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
804				RTFREE_LOCKED(rt);
805				goto drop;
806			}
807			RTFREE_LOCKED(rt);
808
809			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
810			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
811
812			/*
813			 * Also check that the node which sent the ARP packet
814			 * is on the interface we expect it to be on. This
815			 * avoids ARP chaos if an interface is connected to the
816			 * wrong network.
817			 */
818			sin.sin_addr = isaddr;
819
820			/* XXX MRT use table 0 for arp checks */
821			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
822			if (!rt)
823				goto drop;
824			if (rt->rt_ifp != ifp) {
825				log(LOG_INFO, "arp_proxy: ignoring request"
826				    " from %s via %s, expecting %s\n",
827				    inet_ntoa(isaddr), ifp->if_xname,
828				    rt->rt_ifp->if_xname);
829				RTFREE_LOCKED(rt);
830				goto drop;
831			}
832			RTFREE_LOCKED(rt);
833
834#ifdef DEBUG_PROXY
835			printf("arp: proxying for %s\n",
836			       inet_ntoa(itaddr));
837#endif
838		}
839	}
840
841	if (itaddr.s_addr == myaddr.s_addr &&
842	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
843		/* RFC 3927 link-local IPv4; always reply by broadcast. */
844#ifdef DEBUG_LINKLOCAL
845		printf("arp: sending reply for link-local addr %s\n",
846		    inet_ntoa(itaddr));
847#endif
848		m->m_flags |= M_BCAST;
849		m->m_flags &= ~M_MCAST;
850	} else {
851		/* default behaviour; never reply by broadcast. */
852		m->m_flags &= ~(M_BCAST|M_MCAST);
853	}
854	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
855	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
856	ah->ar_op = htons(ARPOP_REPLY);
857	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
858	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
859	m->m_pkthdr.len = m->m_len;
860	sa.sa_family = AF_ARP;
861	sa.sa_len = 2;
862	(*ifp->if_output)(ifp, m, &sa, NULL);
863	ARPSTAT_INC(txreplies);
864	return;
865
866drop:
867	m_freem(m);
868}
869#endif
870
871void
872arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
873{
874	struct llentry *lle;
875
876	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
877		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
878				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
879		/*
880		 * interface address is considered static entry
881		 * because the output of the arp utility shows
882		 * that L2 entry as permanent
883		 */
884		IF_AFDATA_LOCK(ifp);
885		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
886				 (struct sockaddr *)IA_SIN(ifa));
887		IF_AFDATA_UNLOCK(ifp);
888		if (lle == NULL)
889			log(LOG_INFO, "arp_ifinit: cannot create arp "
890			    "entry for interface address\n");
891		else
892			LLE_RUNLOCK(lle);
893	}
894	ifa->ifa_rtrequest = NULL;
895}
896
897void
898arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
899{
900	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
901		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
902				&IA_SIN(ifa)->sin_addr, enaddr);
903	ifa->ifa_rtrequest = NULL;
904}
905
906static void
907arp_init(void)
908{
909
910	netisr_register(&arp_nh);
911}
912SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
913