if_ether.c revision 189592
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 189592 2009-03-09 17:53:05Z bms $");
40
41#include "opt_inet.h"
42#include "opt_route.h"
43#include "opt_mac.h"
44#include "opt_carp.h"
45
46#include <sys/param.h>
47#include <sys/kernel.h>
48#include <sys/queue.h>
49#include <sys/sysctl.h>
50#include <sys/systm.h>
51#include <sys/mbuf.h>
52#include <sys/malloc.h>
53#include <sys/proc.h>
54#include <sys/socket.h>
55#include <sys/syslog.h>
56#include <sys/vimage.h>
57
58#include <net/if.h>
59#include <net/if_dl.h>
60#include <net/if_types.h>
61#include <net/route.h>
62#include <net/netisr.h>
63#include <net/if_llc.h>
64#include <net/ethernet.h>
65#include <net/vnet.h>
66
67#include <netinet/in.h>
68#include <netinet/in_var.h>
69#include <net/if_llatbl.h>
70#include <netinet/if_ether.h>
71#include <netinet/vinet.h>
72
73#include <net/if_arc.h>
74#include <net/iso88025.h>
75
76#ifdef DEV_CARP
77#include <netinet/ip_carp.h>
78#endif
79
80#include <security/mac/mac_framework.h>
81
82#define SIN(s) ((struct sockaddr_in *)s)
83#define SDL(s) ((struct sockaddr_dl *)s)
84#define LLTABLE(ifp)	\
85	((struct in_ifinfo *)(ifp)->if_afdata[AF_INET])->ii_llt
86
87SYSCTL_DECL(_net_link_ether);
88SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
89
90/* timer values */
91#ifdef VIMAGE_GLOBALS
92static int	arpt_keep; /* once resolved, good for 20 more minutes */
93static int	arp_maxtries;
94int	useloopback; /* use loopback interface for local traffic */
95static int	arp_proxyall;
96#endif
97
98SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, max_age,
99    CTLFLAG_RW, arpt_keep, 0, "ARP entry lifetime in seconds");
100
101static struct	ifqueue arpintrq;
102
103SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries,
104	CTLFLAG_RW, arp_maxtries, 0,
105	"ARP resolution attempts before returning error");
106SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, useloopback,
107	CTLFLAG_RW, useloopback, 0,
108	"Use the loopback interface for local traffic");
109SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall,
110	CTLFLAG_RW, arp_proxyall, 0,
111	"Enable proxy ARP for all suitable requests");
112
113static void	arp_init(void);
114void		arprequest(struct ifnet *,
115			struct in_addr *, struct in_addr *, u_char *);
116static void	arpintr(struct mbuf *);
117static void	arptimer(void *);
118#ifdef INET
119static void	in_arpinput(struct mbuf *);
120#endif
121
122#ifdef AF_INET
123void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
124
125/*
126 * called by in_ifscrub to remove entry from the table when
127 * the interface goes away
128 */
129void
130arp_ifscrub(struct ifnet *ifp, uint32_t addr)
131{
132	struct sockaddr_in addr4;
133
134	bzero((void *)&addr4, sizeof(addr4));
135	addr4.sin_len    = sizeof(addr4);
136	addr4.sin_family = AF_INET;
137	addr4.sin_addr.s_addr = addr;
138	IF_AFDATA_LOCK(ifp);
139	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
140	    (struct sockaddr *)&addr4);
141	IF_AFDATA_UNLOCK(ifp);
142}
143#endif
144
145/*
146 * Timeout routine.  Age arp_tab entries periodically.
147 */
148static void
149arptimer(void *arg)
150{
151	struct ifnet *ifp;
152	struct llentry   *lle = (struct llentry *)arg;
153
154	if (lle == NULL) {
155		panic("%s: NULL entry!\n", __func__);
156		return;
157	}
158	ifp = lle->lle_tbl->llt_ifp;
159	IF_AFDATA_LOCK(ifp);
160	LLE_WLOCK(lle);
161	if (((lle->la_flags & LLE_DELETED)
162		|| (time_second >= lle->la_expire))
163	    && (!callout_pending(&lle->la_timer) &&
164		callout_active(&lle->la_timer)))
165		(void) llentry_free(lle);
166	else {
167		/*
168		 * Still valid, just drop our reference
169		 */
170		LLE_FREE_LOCKED(lle);
171	}
172	IF_AFDATA_UNLOCK(ifp);
173}
174
175/*
176 * Broadcast an ARP request. Caller specifies:
177 *	- arp header source ip address
178 *	- arp header target ip address
179 *	- arp header source ethernet address
180 */
181void
182arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
183    u_char *enaddr)
184{
185	struct mbuf *m;
186	struct arphdr *ah;
187	struct sockaddr sa;
188
189	if (sip == NULL) {
190		/* XXX don't believe this can happen (or explain why) */
191		/*
192		 * The caller did not supply a source address, try to find
193		 * a compatible one among those assigned to this interface.
194		 */
195		struct ifaddr *ifa;
196
197		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
198			if (!ifa->ifa_addr ||
199			    ifa->ifa_addr->sa_family != AF_INET)
200				continue;
201			sip = &SIN(ifa->ifa_addr)->sin_addr;
202			if (0 == ((sip->s_addr ^ tip->s_addr) &
203			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
204				break;  /* found it. */
205		}
206		if (sip == NULL) {
207			printf("%s: cannot find matching address\n", __func__);
208			return;
209		}
210	}
211
212	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
213		return;
214	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
215		2*ifp->if_data.ifi_addrlen;
216	m->m_pkthdr.len = m->m_len;
217	MH_ALIGN(m, m->m_len);
218	ah = mtod(m, struct arphdr *);
219	bzero((caddr_t)ah, m->m_len);
220#ifdef MAC
221	mac_netinet_arp_send(ifp, m);
222#endif
223	ah->ar_pro = htons(ETHERTYPE_IP);
224	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
225	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
226	ah->ar_op = htons(ARPOP_REQUEST);
227	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
228	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
229	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
230	sa.sa_family = AF_ARP;
231	sa.sa_len = 2;
232	m->m_flags |= M_BCAST;
233	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
234}
235
236/*
237 * Resolve an IP address into an ethernet address.
238 * On input:
239 *    ifp is the interface we use
240 *    rt0 is the route to the final destination (possibly useless)
241 *    m is the mbuf. May be NULL if we don't have a packet.
242 *    dst is the next hop,
243 *    desten is where we want the address.
244 *
245 * On success, desten is filled in and the function returns 0;
246 * If the packet must be held pending resolution, we return EWOULDBLOCK
247 * On other errors, we return the corresponding error code.
248 * Note that m_freem() handles NULL.
249 */
250int
251arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
252	struct sockaddr *dst, u_char *desten, struct llentry **lle)
253{
254	INIT_VNET_INET(ifp->if_vnet);
255	struct llentry *la = 0;
256	u_int flags = 0;
257	int error, renew;
258
259	*lle = NULL;
260	if (m != NULL) {
261		if (m->m_flags & M_BCAST) {
262			/* broadcast */
263			(void)memcpy(desten,
264			    ifp->if_broadcastaddr, ifp->if_addrlen);
265			return (0);
266		}
267		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
268			/* multicast */
269			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
270			return (0);
271		}
272	}
273	/* XXXXX
274	 */
275retry:
276	IF_AFDATA_RLOCK(ifp);
277	la = lla_lookup(LLTABLE(ifp), flags, dst);
278	IF_AFDATA_RUNLOCK(ifp);
279	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
280	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
281		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
282		IF_AFDATA_WLOCK(ifp);
283		la = lla_lookup(LLTABLE(ifp), flags, dst);
284		IF_AFDATA_WUNLOCK(ifp);
285	}
286	if (la == NULL) {
287		if (flags & LLE_CREATE)
288			log(LOG_DEBUG,
289			    "arpresolve: can't allocate llinfo for %s\n",
290			    inet_ntoa(SIN(dst)->sin_addr));
291		m_freem(m);
292		return (EINVAL);
293	}
294
295	if ((la->la_flags & LLE_VALID) &&
296	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
297		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
298		/*
299		 * If entry has an expiry time and it is approaching,
300		 * see if we need to send an ARP request within this
301		 * arpt_down interval.
302		 */
303		if (!(la->la_flags & LLE_STATIC) &&
304		    time_uptime + la->la_preempt > la->la_expire) {
305			arprequest(ifp, NULL,
306			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
307
308			la->la_preempt--;
309		}
310
311		*lle = la;
312		error = 0;
313		goto done;
314	}
315
316	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
317		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
318		    inet_ntoa(SIN(dst)->sin_addr));
319		m_freem(m);
320		error = EINVAL;
321		goto done;
322	}
323
324	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
325	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
326		flags |= LLE_EXCLUSIVE;
327		LLE_RUNLOCK(la);
328		goto retry;
329	}
330	/*
331	 * There is an arptab entry, but no ethernet address
332	 * response yet.  Replace the held mbuf with this
333	 * latest one.
334	 */
335	if (m != NULL) {
336		if (la->la_hold != NULL)
337			m_freem(la->la_hold);
338		la->la_hold = m;
339		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
340			flags &= ~LLE_EXCLUSIVE;
341			LLE_DOWNGRADE(la);
342		}
343
344	}
345	/*
346	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
347	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
348	 * if we have already sent arp_maxtries ARP requests. Retransmit the
349	 * ARP request, but not faster than one request per second.
350	 */
351	if (la->la_asked < V_arp_maxtries)
352		error = EWOULDBLOCK;	/* First request. */
353	else
354		error =
355		    (rt0->rt_flags & RTF_GATEWAY) ? EHOSTDOWN : EHOSTUNREACH;
356
357	if (renew) {
358		LLE_ADDREF(la);
359		la->la_expire = time_uptime;
360		callout_reset(&la->la_timer, hz, arptimer, la);
361		la->la_asked++;
362		LLE_WUNLOCK(la);
363		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
364		    IF_LLADDR(ifp));
365		return (error);
366	}
367done:
368	if (flags & LLE_EXCLUSIVE)
369		LLE_WUNLOCK(la);
370	else
371		LLE_RUNLOCK(la);
372	return (error);
373}
374
375/*
376 * Common length and type checks are done here,
377 * then the protocol-specific routine is called.
378 */
379static void
380arpintr(struct mbuf *m)
381{
382	struct arphdr *ar;
383
384	if (m->m_len < sizeof(struct arphdr) &&
385	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
386		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
387		return;
388	}
389	ar = mtod(m, struct arphdr *);
390
391	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
392	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
393	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
394	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
395		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
396		    (unsigned char *)&ar->ar_hrd, "");
397		m_freem(m);
398		return;
399	}
400
401	if (m->m_len < arphdr_len(ar)) {
402		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
403			log(LOG_ERR, "arp: runt packet\n");
404			m_freem(m);
405			return;
406		}
407		ar = mtod(m, struct arphdr *);
408	}
409
410	switch (ntohs(ar->ar_pro)) {
411#ifdef INET
412	case ETHERTYPE_IP:
413		in_arpinput(m);
414		return;
415#endif
416	}
417	m_freem(m);
418}
419
420#ifdef INET
421/*
422 * ARP for Internet protocols on 10 Mb/s Ethernet.
423 * Algorithm is that given in RFC 826.
424 * In addition, a sanity check is performed on the sender
425 * protocol address, to catch impersonators.
426 * We no longer handle negotiations for use of trailer protocol:
427 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
428 * along with IP replies if we wanted trailers sent to us,
429 * and also sent them in response to IP replies.
430 * This allowed either end to announce the desire to receive
431 * trailer packets.
432 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
433 * but formerly didn't normally send requests.
434 */
435static int log_arp_wrong_iface = 1;
436static int log_arp_movements = 1;
437static int log_arp_permanent_modify = 1;
438
439SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
440	&log_arp_wrong_iface, 0,
441	"log arp packets arriving on the wrong interface");
442SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
443        &log_arp_movements, 0,
444        "log arp replies from MACs different than the one in the cache");
445SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
446        &log_arp_permanent_modify, 0,
447        "log arp replies from MACs different than the one in the permanent arp entry");
448
449
450static void
451in_arpinput(struct mbuf *m)
452{
453	struct arphdr *ah;
454	struct ifnet *ifp = m->m_pkthdr.rcvif;
455	struct llentry *la = NULL;
456	struct rtentry *rt;
457	struct ifaddr *ifa;
458	struct in_ifaddr *ia;
459	struct sockaddr sa;
460	struct in_addr isaddr, itaddr, myaddr;
461	u_int8_t *enaddr = NULL;
462	int op, flags;
463	struct mbuf *m0;
464	int req_len;
465	int bridged = 0, is_bridge = 0;
466#ifdef DEV_CARP
467	int carp_match = 0;
468#endif
469	struct sockaddr_in sin;
470	sin.sin_len = sizeof(struct sockaddr_in);
471	sin.sin_family = AF_INET;
472	sin.sin_addr.s_addr = 0;
473	INIT_VNET_INET(ifp->if_vnet);
474
475	if (ifp->if_bridge)
476		bridged = 1;
477	if (ifp->if_type == IFT_BRIDGE)
478		is_bridge = 1;
479
480	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
481	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
482		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
483		return;
484	}
485
486	ah = mtod(m, struct arphdr *);
487	op = ntohs(ah->ar_op);
488	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
489	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
490
491	/*
492	 * For a bridge, we want to check the address irrespective
493	 * of the receive interface. (This will change slightly
494	 * when we have clusters of interfaces).
495	 * If the interface does not match, but the recieving interface
496	 * is part of carp, we call carp_iamatch to see if this is a
497	 * request for the virtual host ip.
498	 * XXX: This is really ugly!
499	 */
500	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
501		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
502		    ia->ia_ifp == ifp) &&
503		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
504			goto match;
505#ifdef DEV_CARP
506		if (ifp->if_carp != NULL &&
507		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
508		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
509			carp_match = 1;
510			goto match;
511		}
512#endif
513	}
514	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
515		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
516		    ia->ia_ifp == ifp) &&
517		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
518			goto match;
519
520#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
521  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
522  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
523  addr == ia->ia_addr.sin_addr.s_addr)
524	/*
525	 * Check the case when bridge shares its MAC address with
526	 * some of its children, so packets are claimed by bridge
527	 * itself (bridge_input() does it first), but they are really
528	 * meant to be destined to the bridge member.
529	 */
530	if (is_bridge) {
531		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
532			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
533				ifp = ia->ia_ifp;
534				goto match;
535			}
536		}
537	}
538#undef BDG_MEMBER_MATCHES_ARP
539
540	/*
541	 * No match, use the first inet address on the receive interface
542	 * as a dummy address for the rest of the function.
543	 */
544	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
545		if (ifa->ifa_addr->sa_family == AF_INET) {
546			ia = ifatoia(ifa);
547			goto match;
548		}
549	/*
550	 * If bridging, fall back to using any inet address.
551	 */
552	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL)
553		goto drop;
554match:
555	if (!enaddr)
556		enaddr = (u_int8_t *)IF_LLADDR(ifp);
557	myaddr = ia->ia_addr.sin_addr;
558	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
559		goto drop;	/* it's from me, ignore it. */
560	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
561		log(LOG_ERR,
562		    "arp: link address is broadcast for IP address %s!\n",
563		    inet_ntoa(isaddr));
564		goto drop;
565	}
566	/*
567	 * Warn if another host is using the same IP address, but only if the
568	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
569	 * case we suppress the warning to avoid false positive complaints of
570	 * potential misconfiguration.
571	 */
572	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
573		log(LOG_ERR,
574		   "arp: %*D is using my IP address %s on %s!\n",
575		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
576		   inet_ntoa(isaddr), ifp->if_xname);
577		itaddr = myaddr;
578		goto reply;
579	}
580	if (ifp->if_flags & IFF_STATICARP)
581		goto reply;
582
583	bzero(&sin, sizeof(sin));
584	sin.sin_len = sizeof(struct sockaddr_in);
585	sin.sin_family = AF_INET;
586	sin.sin_addr = isaddr;
587	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
588	flags |= LLE_EXCLUSIVE;
589	IF_AFDATA_LOCK(ifp);
590	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
591	IF_AFDATA_UNLOCK(ifp);
592	if (la != NULL) {
593		/* the following is not an error when doing bridging */
594		if (!bridged && la->lle_tbl->llt_ifp != ifp
595#ifdef DEV_CARP
596		    && (ifp->if_type != IFT_CARP || !carp_match)
597#endif
598			) {
599			if (log_arp_wrong_iface)
600				log(LOG_ERR, "arp: %s is on %s "
601				    "but got reply from %*D on %s\n",
602				    inet_ntoa(isaddr),
603				    la->lle_tbl->llt_ifp->if_xname,
604				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
605				    ifp->if_xname);
606			goto reply;
607		}
608		if ((la->la_flags & LLE_VALID) &&
609		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
610			if (la->la_flags & LLE_STATIC) {
611				log(LOG_ERR,
612				    "arp: %*D attempts to modify permanent "
613				    "entry for %s on %s\n",
614				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
615				    inet_ntoa(isaddr), ifp->if_xname);
616				goto reply;
617			}
618			if (log_arp_movements) {
619			        log(LOG_INFO, "arp: %s moved from %*D "
620				    "to %*D on %s\n",
621				    inet_ntoa(isaddr),
622				    ifp->if_addrlen,
623				    (u_char *)&la->ll_addr, ":",
624				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
625				    ifp->if_xname);
626			}
627		}
628
629		if (ifp->if_addrlen != ah->ar_hln) {
630			log(LOG_WARNING,
631			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
632			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
633			    ah->ar_hln, ifp->if_addrlen);
634			goto reply;
635		}
636		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
637		la->la_flags |= LLE_VALID;
638
639		if (!(la->la_flags & LLE_STATIC)) {
640			la->la_expire = time_uptime + V_arpt_keep;
641			callout_reset(&la->la_timer, hz * V_arpt_keep,
642			    arptimer, la);
643		}
644		la->la_asked = 0;
645		la->la_preempt = V_arp_maxtries;
646		if (la->la_hold != NULL) {
647			m0 = la->la_hold;
648			la->la_hold = 0;
649			memcpy(&sa, L3_ADDR(la), sizeof(sa));
650			LLE_WUNLOCK(la);
651
652			(*ifp->if_output)(ifp, m0, &sa, NULL);
653			return;
654		}
655	}
656reply:
657	if (op != ARPOP_REQUEST)
658		goto drop;
659
660	if (itaddr.s_addr == myaddr.s_addr) {
661		/* Shortcut.. the receiving interface is the target. */
662		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
663		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
664	} else {
665		struct llentry *lle = NULL;
666
667		if (!V_arp_proxyall)
668			goto drop;
669
670		sin.sin_addr = itaddr;
671		/* XXX MRT use table 0 for arp reply  */
672		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
673		if (!rt)
674			goto drop;
675
676		/*
677		 * Don't send proxies for nodes on the same interface
678		 * as this one came out of, or we'll get into a fight
679		 * over who claims what Ether address.
680		 */
681		if (!rt->rt_ifp || rt->rt_ifp == ifp) {
682			RTFREE_LOCKED(rt);
683			goto drop;
684		}
685		IF_AFDATA_LOCK(rt->rt_ifp);
686		lle = lla_lookup(LLTABLE(rt->rt_ifp), 0, (struct sockaddr *)&sin);
687		IF_AFDATA_UNLOCK(rt->rt_ifp);
688		RTFREE_LOCKED(rt);
689
690		if (lle != NULL) {
691			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
692			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
693			LLE_RUNLOCK(lle);
694		} else
695			goto drop;
696
697		/*
698		 * Also check that the node which sent the ARP packet
699		 * is on the the interface we expect it to be on. This
700		 * avoids ARP chaos if an interface is connected to the
701		 * wrong network.
702		 */
703		sin.sin_addr = isaddr;
704
705		/* XXX MRT use table 0 for arp checks */
706		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
707		if (!rt)
708			goto drop;
709		if (rt->rt_ifp != ifp) {
710			log(LOG_INFO, "arp_proxy: ignoring request"
711			    " from %s via %s, expecting %s\n",
712			    inet_ntoa(isaddr), ifp->if_xname,
713			    rt->rt_ifp->if_xname);
714			RTFREE_LOCKED(rt);
715			goto drop;
716		}
717		RTFREE_LOCKED(rt);
718
719#ifdef DEBUG_PROXY
720		printf("arp: proxying for %s\n",
721		       inet_ntoa(itaddr));
722#endif
723	}
724
725	if (la != NULL)
726		LLE_WUNLOCK(la);
727	if (itaddr.s_addr == myaddr.s_addr &&
728	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
729		/* RFC 3927 link-local IPv4; always reply by broadcast. */
730#ifdef DEBUG_LINKLOCAL
731		printf("arp: sending reply for link-local addr %s\n",
732		    inet_ntoa(itaddr));
733#endif
734		m->m_flags |= M_BCAST;
735		m->m_flags &= ~M_MCAST;
736	} else {
737		/* default behaviour; never reply by broadcast. */
738		m->m_flags &= ~(M_BCAST|M_MCAST);
739	}
740	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
741	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
742	ah->ar_op = htons(ARPOP_REPLY);
743	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
744	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
745	m->m_pkthdr.len = m->m_len;
746	sa.sa_family = AF_ARP;
747	sa.sa_len = 2;
748	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
749	return;
750
751drop:
752	if (la != NULL)
753		LLE_WUNLOCK(la);
754	m_freem(m);
755}
756#endif
757
758void
759arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
760{
761	struct llentry *lle;
762
763	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
764		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
765				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
766		/*
767		 * interface address is considered static entry
768		 * because the output of the arp utility shows
769		 * that L2 entry as permanent
770		 */
771		IF_AFDATA_LOCK(ifp);
772		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
773				 (struct sockaddr *)IA_SIN(ifa));
774		IF_AFDATA_UNLOCK(ifp);
775		if (lle == NULL)
776			log(LOG_INFO, "arp_ifinit: cannot create arp "
777			    "entry for interface address\n");
778		else
779			LLE_RUNLOCK(lle);
780	}
781	ifa->ifa_rtrequest = NULL;
782}
783
784void
785arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
786{
787	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
788		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
789				&IA_SIN(ifa)->sin_addr, enaddr);
790	ifa->ifa_rtrequest = NULL;
791}
792
793static void
794arp_init(void)
795{
796	INIT_VNET_INET(curvnet);
797
798	V_arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
799	V_arp_maxtries = 5;
800	V_useloopback = 1; /* use loopback interface for local traffic */
801	V_arp_proxyall = 0;
802
803	arpintrq.ifq_maxlen = 50;
804	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
805	netisr_register(NETISR_ARP, arpintr, &arpintrq, 0);
806}
807SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
808