if_ether.c revision 228571
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 228571 2011-12-16 12:16:56Z glebius $");
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);
80static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81static SYSCTL_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
142/*
143 * called by in_ifscrub to remove entry from the table when
144 * the interface goes away
145 */
146void
147arp_ifscrub(struct ifnet *ifp, uint32_t addr)
148{
149	struct sockaddr_in addr4;
150
151	bzero((void *)&addr4, sizeof(addr4));
152	addr4.sin_len    = sizeof(addr4);
153	addr4.sin_family = AF_INET;
154	addr4.sin_addr.s_addr = addr;
155	IF_AFDATA_LOCK(ifp);
156	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
157	    (struct sockaddr *)&addr4);
158	IF_AFDATA_UNLOCK(ifp);
159}
160#endif
161
162/*
163 * Timeout routine.  Age arp_tab entries periodically.
164 */
165static void
166arptimer(void *arg)
167{
168	struct ifnet *ifp;
169	struct llentry   *lle;
170	int pkts_dropped;
171
172	KASSERT(arg != NULL, ("%s: arg NULL", __func__));
173	lle = (struct llentry *)arg;
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			pkts_dropped = llentry_free(lle);
186			ARPSTAT_ADD(dropped, pkts_dropped);
187			ARPSTAT_INC(timeouts);
188		} else {
189#ifdef DIAGNOSTIC
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#endif
196			LLE_WUNLOCK(lle);
197		}
198	}
199	IF_AFDATA_UNLOCK(ifp);
200	CURVNET_RESTORE();
201}
202
203/*
204 * Broadcast an ARP request. Caller specifies:
205 *	- arp header source ip address
206 *	- arp header target ip address
207 *	- arp header source ethernet address
208 */
209void
210arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
211    u_char *enaddr)
212{
213	struct mbuf *m;
214	struct arphdr *ah;
215	struct sockaddr sa;
216
217	if (sip == NULL) {
218		/* XXX don't believe this can happen (or explain why) */
219		/*
220		 * The caller did not supply a source address, try to find
221		 * a compatible one among those assigned to this interface.
222		 */
223		struct ifaddr *ifa;
224
225		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
226			if (!ifa->ifa_addr ||
227			    ifa->ifa_addr->sa_family != AF_INET)
228				continue;
229			sip = &SIN(ifa->ifa_addr)->sin_addr;
230			if (0 == ((sip->s_addr ^ tip->s_addr) &
231			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
232				break;  /* found it. */
233		}
234		if (sip == NULL) {
235			printf("%s: cannot find matching address\n", __func__);
236			return;
237		}
238	}
239
240	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
241		return;
242	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
243		2*ifp->if_data.ifi_addrlen;
244	m->m_pkthdr.len = m->m_len;
245	MH_ALIGN(m, m->m_len);
246	ah = mtod(m, struct arphdr *);
247	bzero((caddr_t)ah, m->m_len);
248#ifdef MAC
249	mac_netinet_arp_send(ifp, m);
250#endif
251	ah->ar_pro = htons(ETHERTYPE_IP);
252	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
253	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
254	ah->ar_op = htons(ARPOP_REQUEST);
255	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
256	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
257	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
258	sa.sa_family = AF_ARP;
259	sa.sa_len = 2;
260	m->m_flags |= M_BCAST;
261	(*ifp->if_output)(ifp, m, &sa, NULL);
262	ARPSTAT_INC(txrequests);
263}
264
265/*
266 * Resolve an IP address into an ethernet address.
267 * On input:
268 *    ifp is the interface we use
269 *    rt0 is the route to the final destination (possibly useless)
270 *    m is the mbuf. May be NULL if we don't have a packet.
271 *    dst is the next hop,
272 *    desten is where we want the address.
273 *
274 * On success, desten is filled in and the function returns 0;
275 * If the packet must be held pending resolution, we return EWOULDBLOCK
276 * On other errors, we return the corresponding error code.
277 * Note that m_freem() handles NULL.
278 */
279int
280arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
281	struct sockaddr *dst, u_char *desten, struct llentry **lle)
282{
283	struct llentry *la = 0;
284	u_int flags = 0;
285	struct mbuf *curr = NULL;
286	struct mbuf *next = NULL;
287	int error, renew;
288
289	*lle = NULL;
290	if (m != NULL) {
291		if (m->m_flags & M_BCAST) {
292			/* broadcast */
293			(void)memcpy(desten,
294			    ifp->if_broadcastaddr, ifp->if_addrlen);
295			return (0);
296		}
297		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
298			/* multicast */
299			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
300			return (0);
301		}
302	}
303retry:
304	IF_AFDATA_RLOCK(ifp);
305	la = lla_lookup(LLTABLE(ifp), flags, dst);
306	IF_AFDATA_RUNLOCK(ifp);
307	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
308	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
309		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
310		IF_AFDATA_WLOCK(ifp);
311		la = lla_lookup(LLTABLE(ifp), flags, dst);
312		IF_AFDATA_WUNLOCK(ifp);
313	}
314	if (la == NULL) {
315		if (flags & LLE_CREATE)
316			log(LOG_DEBUG,
317			    "arpresolve: can't allocate llinfo for %s\n",
318			    inet_ntoa(SIN(dst)->sin_addr));
319		m_freem(m);
320		return (EINVAL);
321	}
322
323	if ((la->la_flags & LLE_VALID) &&
324	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
325		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
326		/*
327		 * If entry has an expiry time and it is approaching,
328		 * see if we need to send an ARP request within this
329		 * arpt_down interval.
330		 */
331		if (!(la->la_flags & LLE_STATIC) &&
332		    time_uptime + la->la_preempt > la->la_expire) {
333			arprequest(ifp, NULL,
334			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
335
336			la->la_preempt--;
337		}
338
339		*lle = la;
340		error = 0;
341		goto done;
342	}
343
344	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
345		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
346		    inet_ntoa(SIN(dst)->sin_addr));
347		m_freem(m);
348		error = EINVAL;
349		goto done;
350	}
351
352	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
353	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
354		flags |= LLE_EXCLUSIVE;
355		LLE_RUNLOCK(la);
356		goto retry;
357	}
358	/*
359	 * There is an arptab entry, but no ethernet address
360	 * response yet.  Add the mbuf to the list, dropping
361	 * the oldest packet if we have exceeded the system
362	 * setting.
363	 */
364	if (m != NULL) {
365		if (la->la_numheld >= V_arp_maxhold) {
366			if (la->la_hold != NULL) {
367				next = la->la_hold->m_nextpkt;
368				m_freem(la->la_hold);
369				la->la_hold = next;
370				la->la_numheld--;
371				ARPSTAT_INC(dropped);
372			}
373		}
374		if (la->la_hold != NULL) {
375			curr = la->la_hold;
376			while (curr->m_nextpkt != NULL)
377				curr = curr->m_nextpkt;
378			curr->m_nextpkt = m;
379		} else
380			la->la_hold = m;
381		la->la_numheld++;
382		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
383			flags &= ~LLE_EXCLUSIVE;
384			LLE_DOWNGRADE(la);
385		}
386
387	}
388	/*
389	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
390	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
391	 * if we have already sent arp_maxtries ARP requests. Retransmit the
392	 * ARP request, but not faster than one request per second.
393	 */
394	if (la->la_asked < V_arp_maxtries)
395		error = EWOULDBLOCK;	/* First request. */
396	else
397		error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
398		    EHOSTUNREACH : EHOSTDOWN;
399
400	if (renew) {
401		int canceled;
402
403		LLE_ADDREF(la);
404		la->la_expire = time_uptime;
405		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
406		    arptimer, la);
407		if (canceled)
408			LLE_REMREF(la);
409		la->la_asked++;
410		LLE_WUNLOCK(la);
411		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
412		    IF_LLADDR(ifp));
413		return (error);
414	}
415done:
416	if (flags & LLE_EXCLUSIVE)
417		LLE_WUNLOCK(la);
418	else
419		LLE_RUNLOCK(la);
420	return (error);
421}
422
423/*
424 * Common length and type checks are done here,
425 * then the protocol-specific routine is called.
426 */
427static void
428arpintr(struct mbuf *m)
429{
430	struct arphdr *ar;
431
432	if (m->m_len < sizeof(struct arphdr) &&
433	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
434		log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
435		return;
436	}
437	ar = mtod(m, struct arphdr *);
438
439	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
440	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
441	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
442	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
443	    ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
444		log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
445		    " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
446		    ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
447		    ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
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_NOTICE, "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 carped;
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_NOTICE, "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_NOTICE, "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_NOTICE, "in_arp: %*D is multicast\n",
547		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
548		return;
549	}
550
551	op = ntohs(ah->ar_op);
552	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
553	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
554
555	if (op == ARPOP_REPLY)
556		ARPSTAT_INC(rxreplies);
557
558	/*
559	 * For a bridge, we want to check the address irrespective
560	 * of the receive interface. (This will change slightly
561	 * when we have clusters of interfaces).
562	 */
563	IN_IFADDR_RLOCK();
564	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
565		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
566		    ia->ia_ifp == ifp) &&
567		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
568		    (ia->ia_ifa.ifa_carp == NULL ||
569		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
570			ifa_ref(&ia->ia_ifa);
571			IN_IFADDR_RUNLOCK();
572			goto match;
573		}
574	}
575	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
576		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
577		    ia->ia_ifp == ifp) &&
578		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
579			ifa_ref(&ia->ia_ifa);
580			IN_IFADDR_RUNLOCK();
581			goto match;
582		}
583
584#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
585  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
586  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
587  addr == ia->ia_addr.sin_addr.s_addr)
588	/*
589	 * Check the case when bridge shares its MAC address with
590	 * some of its children, so packets are claimed by bridge
591	 * itself (bridge_input() does it first), but they are really
592	 * meant to be destined to the bridge member.
593	 */
594	if (is_bridge) {
595		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
596			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
597				ifa_ref(&ia->ia_ifa);
598				ifp = ia->ia_ifp;
599				IN_IFADDR_RUNLOCK();
600				goto match;
601			}
602		}
603	}
604#undef BDG_MEMBER_MATCHES_ARP
605	IN_IFADDR_RUNLOCK();
606
607	/*
608	 * No match, use the first inet address on the receive interface
609	 * as a dummy address for the rest of the function.
610	 */
611	IF_ADDR_LOCK(ifp);
612	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
613		if (ifa->ifa_addr->sa_family == AF_INET) {
614			ia = ifatoia(ifa);
615			ifa_ref(ifa);
616			IF_ADDR_UNLOCK(ifp);
617			goto match;
618		}
619	IF_ADDR_UNLOCK(ifp);
620
621	/*
622	 * If bridging, fall back to using any inet address.
623	 */
624	IN_IFADDR_RLOCK();
625	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
626		IN_IFADDR_RUNLOCK();
627		goto drop;
628	}
629	ifa_ref(&ia->ia_ifa);
630	IN_IFADDR_RUNLOCK();
631match:
632	if (!enaddr)
633		enaddr = (u_int8_t *)IF_LLADDR(ifp);
634	carped = (ia->ia_ifa.ifa_carp != NULL);
635	myaddr = ia->ia_addr.sin_addr;
636	ifa_free(&ia->ia_ifa);
637	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
638		goto drop;	/* it's from me, ignore it. */
639	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
640		log(LOG_NOTICE,
641		    "arp: link address is broadcast for IP address %s!\n",
642		    inet_ntoa(isaddr));
643		goto drop;
644	}
645	/*
646	 * Warn if another host is using the same IP address, but only if the
647	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
648	 * case we suppress the warning to avoid false positive complaints of
649	 * potential misconfiguration.
650	 */
651	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
652	    myaddr.s_addr != 0) {
653		log(LOG_ERR, "arp: %*D is using my IP address %s on %s!\n",
654		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
655		   inet_ntoa(isaddr), ifp->if_xname);
656		itaddr = myaddr;
657		ARPSTAT_INC(dupips);
658		goto reply;
659	}
660	if (ifp->if_flags & IFF_STATICARP)
661		goto reply;
662
663	bzero(&sin, sizeof(sin));
664	sin.sin_len = sizeof(struct sockaddr_in);
665	sin.sin_family = AF_INET;
666	sin.sin_addr = isaddr;
667	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
668	flags |= LLE_EXCLUSIVE;
669	IF_AFDATA_LOCK(ifp);
670	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
671	IF_AFDATA_UNLOCK(ifp);
672	if (la != NULL) {
673		/* the following is not an error when doing bridging */
674		if (!bridged && la->lle_tbl->llt_ifp != ifp) {
675			if (log_arp_wrong_iface)
676				log(LOG_WARNING, "arp: %s is on %s "
677				    "but got reply from %*D on %s\n",
678				    inet_ntoa(isaddr),
679				    la->lle_tbl->llt_ifp->if_xname,
680				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
681				    ifp->if_xname);
682			LLE_WUNLOCK(la);
683			goto reply;
684		}
685		if ((la->la_flags & LLE_VALID) &&
686		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
687			if (la->la_flags & LLE_STATIC) {
688				LLE_WUNLOCK(la);
689				if (log_arp_permanent_modify)
690					log(LOG_ERR,
691					    "arp: %*D attempts to modify "
692					    "permanent entry for %s on %s\n",
693					    ifp->if_addrlen,
694					    (u_char *)ar_sha(ah), ":",
695					    inet_ntoa(isaddr), ifp->if_xname);
696				goto reply;
697			}
698			if (log_arp_movements) {
699			        log(LOG_INFO, "arp: %s moved from %*D "
700				    "to %*D on %s\n",
701				    inet_ntoa(isaddr),
702				    ifp->if_addrlen,
703				    (u_char *)&la->ll_addr, ":",
704				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
705				    ifp->if_xname);
706			}
707		}
708
709		if (ifp->if_addrlen != ah->ar_hln) {
710			LLE_WUNLOCK(la);
711			log(LOG_WARNING, "arp from %*D: addr len: new %d, "
712			    "i/f %d (ignored)\n", ifp->if_addrlen,
713			    (u_char *) ar_sha(ah), ":", ah->ar_hln,
714			    ifp->if_addrlen);
715			goto drop;
716		}
717		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
718		la->la_flags |= LLE_VALID;
719
720		EVENTHANDLER_INVOKE(arp_update_event, la);
721
722		if (!(la->la_flags & LLE_STATIC)) {
723			int canceled;
724
725			LLE_ADDREF(la);
726			la->la_expire = time_uptime + V_arpt_keep;
727			canceled = callout_reset(&la->la_timer,
728			    hz * V_arpt_keep, arptimer, la);
729			if (canceled)
730				LLE_REMREF(la);
731		}
732		la->la_asked = 0;
733		la->la_preempt = V_arp_maxtries;
734		/*
735		 * The packets are all freed within the call to the output
736		 * routine.
737		 *
738		 * NB: The lock MUST be released before the call to the
739		 * output routine.
740		 */
741		if (la->la_hold != NULL) {
742			struct mbuf *m_hold, *m_hold_next;
743
744			m_hold = la->la_hold;
745			la->la_hold = NULL;
746			la->la_numheld = 0;
747			memcpy(&sa, L3_ADDR(la), sizeof(sa));
748			LLE_WUNLOCK(la);
749			for (; m_hold != NULL; m_hold = m_hold_next) {
750				m_hold_next = m_hold->m_nextpkt;
751				m_hold->m_nextpkt = NULL;
752				(*ifp->if_output)(ifp, m_hold, &sa, NULL);
753			}
754		} else
755			LLE_WUNLOCK(la);
756	}
757reply:
758	if (op != ARPOP_REQUEST)
759		goto drop;
760	ARPSTAT_INC(rxrequests);
761
762	if (itaddr.s_addr == myaddr.s_addr) {
763		/* Shortcut.. the receiving interface is the target. */
764		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
765		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
766	} else {
767		struct llentry *lle = NULL;
768
769		sin.sin_addr = itaddr;
770		IF_AFDATA_LOCK(ifp);
771		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
772		IF_AFDATA_UNLOCK(ifp);
773
774		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
775			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
776			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
777			LLE_RUNLOCK(lle);
778		} else {
779
780			if (lle != NULL)
781				LLE_RUNLOCK(lle);
782
783			if (!V_arp_proxyall)
784				goto drop;
785
786			sin.sin_addr = itaddr;
787			/* XXX MRT use table 0 for arp reply  */
788			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
789			if (!rt)
790				goto drop;
791
792			/*
793			 * Don't send proxies for nodes on the same interface
794			 * as this one came out of, or we'll get into a fight
795			 * over who claims what Ether address.
796			 */
797			if (!rt->rt_ifp || rt->rt_ifp == ifp) {
798				RTFREE_LOCKED(rt);
799				goto drop;
800			}
801			RTFREE_LOCKED(rt);
802
803			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
804			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
805
806			/*
807			 * Also check that the node which sent the ARP packet
808			 * is on the interface we expect it to be on. This
809			 * avoids ARP chaos if an interface is connected to the
810			 * wrong network.
811			 */
812			sin.sin_addr = isaddr;
813
814			/* XXX MRT use table 0 for arp checks */
815			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
816			if (!rt)
817				goto drop;
818			if (rt->rt_ifp != ifp) {
819				log(LOG_INFO, "arp_proxy: ignoring request"
820				    " from %s via %s, expecting %s\n",
821				    inet_ntoa(isaddr), ifp->if_xname,
822				    rt->rt_ifp->if_xname);
823				RTFREE_LOCKED(rt);
824				goto drop;
825			}
826			RTFREE_LOCKED(rt);
827
828#ifdef DEBUG_PROXY
829			printf("arp: proxying for %s\n",
830			       inet_ntoa(itaddr));
831#endif
832		}
833	}
834
835	if (itaddr.s_addr == myaddr.s_addr &&
836	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
837		/* RFC 3927 link-local IPv4; always reply by broadcast. */
838#ifdef DEBUG_LINKLOCAL
839		printf("arp: sending reply for link-local addr %s\n",
840		    inet_ntoa(itaddr));
841#endif
842		m->m_flags |= M_BCAST;
843		m->m_flags &= ~M_MCAST;
844	} else {
845		/* default behaviour; never reply by broadcast. */
846		m->m_flags &= ~(M_BCAST|M_MCAST);
847	}
848	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
849	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
850	ah->ar_op = htons(ARPOP_REPLY);
851	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
852	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
853	m->m_pkthdr.len = m->m_len;
854	m->m_pkthdr.rcvif = NULL;
855	sa.sa_family = AF_ARP;
856	sa.sa_len = 2;
857	(*ifp->if_output)(ifp, m, &sa, NULL);
858	ARPSTAT_INC(txreplies);
859	return;
860
861drop:
862	m_freem(m);
863}
864#endif
865
866void
867arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
868{
869	struct llentry *lle;
870
871	if (ifa->ifa_carp != NULL)
872		return;
873
874	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
875		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
876				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
877		/*
878		 * interface address is considered static entry
879		 * because the output of the arp utility shows
880		 * that L2 entry as permanent
881		 */
882		IF_AFDATA_LOCK(ifp);
883		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
884				 (struct sockaddr *)IA_SIN(ifa));
885		IF_AFDATA_UNLOCK(ifp);
886		if (lle == NULL)
887			log(LOG_INFO, "arp_ifinit: cannot create arp "
888			    "entry for interface address\n");
889		else
890			LLE_RUNLOCK(lle);
891	}
892	ifa->ifa_rtrequest = NULL;
893}
894
895void
896arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
897{
898	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
899		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
900				&IA_SIN(ifa)->sin_addr, enaddr);
901	ifa->ifa_rtrequest = NULL;
902}
903
904static void
905arp_init(void)
906{
907
908	netisr_register(&arp_nh);
909}
910SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
911