if_ether.c revision 194837
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 194837 2009-06-24 14:49:26Z rwatson $");
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#include <sys/vimage.h>
55
56#include <net/if.h>
57#include <net/if_dl.h>
58#include <net/if_types.h>
59#include <net/netisr.h>
60#include <net/if_llc.h>
61#include <net/ethernet.h>
62#include <net/route.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#include <netinet/vinet.h>
69
70#include <net/if_arc.h>
71#include <net/iso88025.h>
72
73#ifdef DEV_CARP
74#include <netinet/ip_carp.h>
75#endif
76
77#include <security/mac/mac_framework.h>
78
79#define SIN(s) ((struct sockaddr_in *)s)
80#define SDL(s) ((struct sockaddr_dl *)s)
81
82SYSCTL_DECL(_net_link_ether);
83SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
84
85/* timer values */
86#ifdef VIMAGE_GLOBALS
87static int	arpt_keep; /* once resolved, good for 20 more minutes */
88static int	arp_maxtries;
89int	useloopback; /* use loopback interface for local traffic */
90static int	arp_proxyall;
91#endif
92
93SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, max_age,
94    CTLFLAG_RW, arpt_keep, 0, "ARP entry lifetime in seconds");
95
96SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries,
97	CTLFLAG_RW, arp_maxtries, 0,
98	"ARP resolution attempts before returning error");
99SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, useloopback,
100	CTLFLAG_RW, useloopback, 0,
101	"Use the loopback interface for local traffic");
102SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall,
103	CTLFLAG_RW, arp_proxyall, 0,
104	"Enable proxy ARP for all suitable requests");
105
106static void	arp_init(void);
107static int	arp_iattach(const void *);
108void		arprequest(struct ifnet *,
109			struct in_addr *, struct in_addr *, u_char *);
110static void	arpintr(struct mbuf *);
111static void	arptimer(void *);
112#ifdef INET
113static void	in_arpinput(struct mbuf *);
114#endif
115
116static const struct netisr_handler arp_nh = {
117	.nh_name = "arp",
118	.nh_handler = arpintr,
119	.nh_proto = NETISR_ARP,
120	.nh_policy = NETISR_POLICY_SOURCE,
121};
122
123#ifndef VIMAGE_GLOBALS
124static const vnet_modinfo_t vnet_arp_modinfo = {
125	.vmi_id		= VNET_MOD_ARP,
126	.vmi_name	= "arp",
127	.vmi_dependson	= VNET_MOD_INET,
128	.vmi_iattach	= arp_iattach
129};
130#endif /* !VIMAGE_GLOBALS */
131
132#ifdef AF_INET
133void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
134
135/*
136 * called by in_ifscrub to remove entry from the table when
137 * the interface goes away
138 */
139void
140arp_ifscrub(struct ifnet *ifp, uint32_t addr)
141{
142	struct sockaddr_in addr4;
143
144	bzero((void *)&addr4, sizeof(addr4));
145	addr4.sin_len    = sizeof(addr4);
146	addr4.sin_family = AF_INET;
147	addr4.sin_addr.s_addr = addr;
148	CURVNET_SET(ifp->if_vnet);
149	IF_AFDATA_LOCK(ifp);
150	lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
151	    (struct sockaddr *)&addr4);
152	IF_AFDATA_UNLOCK(ifp);
153	CURVNET_RESTORE();
154}
155#endif
156
157/*
158 * Timeout routine.  Age arp_tab entries periodically.
159 */
160static void
161arptimer(void *arg)
162{
163	struct ifnet *ifp;
164	struct llentry   *lle = (struct llentry *)arg;
165
166	if (lle == NULL) {
167		panic("%s: NULL entry!\n", __func__);
168		return;
169	}
170	ifp = lle->lle_tbl->llt_ifp;
171	IF_AFDATA_LOCK(ifp);
172	LLE_WLOCK(lle);
173	if (((lle->la_flags & LLE_DELETED)
174		|| (time_second >= lle->la_expire))
175	    && (!callout_pending(&lle->la_timer) &&
176		callout_active(&lle->la_timer)))
177		(void) llentry_free(lle);
178	else {
179		/*
180		 * Still valid, just drop our reference
181		 */
182		LLE_FREE_LOCKED(lle);
183	}
184	IF_AFDATA_UNLOCK(ifp);
185}
186
187/*
188 * Broadcast an ARP request. Caller specifies:
189 *	- arp header source ip address
190 *	- arp header target ip address
191 *	- arp header source ethernet address
192 */
193void
194arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr  *tip,
195    u_char *enaddr)
196{
197	struct mbuf *m;
198	struct arphdr *ah;
199	struct sockaddr sa;
200
201	if (sip == NULL) {
202		/* XXX don't believe this can happen (or explain why) */
203		/*
204		 * The caller did not supply a source address, try to find
205		 * a compatible one among those assigned to this interface.
206		 */
207		struct ifaddr *ifa;
208
209		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
210			if (!ifa->ifa_addr ||
211			    ifa->ifa_addr->sa_family != AF_INET)
212				continue;
213			sip = &SIN(ifa->ifa_addr)->sin_addr;
214			if (0 == ((sip->s_addr ^ tip->s_addr) &
215			    SIN(ifa->ifa_netmask)->sin_addr.s_addr) )
216				break;  /* found it. */
217		}
218		if (sip == NULL) {
219			printf("%s: cannot find matching address\n", __func__);
220			return;
221		}
222	}
223
224	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
225		return;
226	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
227		2*ifp->if_data.ifi_addrlen;
228	m->m_pkthdr.len = m->m_len;
229	MH_ALIGN(m, m->m_len);
230	ah = mtod(m, struct arphdr *);
231	bzero((caddr_t)ah, m->m_len);
232#ifdef MAC
233	mac_netinet_arp_send(ifp, m);
234#endif
235	ah->ar_pro = htons(ETHERTYPE_IP);
236	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
237	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
238	ah->ar_op = htons(ARPOP_REQUEST);
239	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
240	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
241	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
242	sa.sa_family = AF_ARP;
243	sa.sa_len = 2;
244	m->m_flags |= M_BCAST;
245	(*ifp->if_output)(ifp, m, &sa, NULL);
246}
247
248/*
249 * Resolve an IP address into an ethernet address.
250 * On input:
251 *    ifp is the interface we use
252 *    rt0 is the route to the final destination (possibly useless)
253 *    m is the mbuf. May be NULL if we don't have a packet.
254 *    dst is the next hop,
255 *    desten is where we want the address.
256 *
257 * On success, desten is filled in and the function returns 0;
258 * If the packet must be held pending resolution, we return EWOULDBLOCK
259 * On other errors, we return the corresponding error code.
260 * Note that m_freem() handles NULL.
261 */
262int
263arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
264	struct sockaddr *dst, u_char *desten, struct llentry **lle)
265{
266	INIT_VNET_INET(ifp->if_vnet);
267	struct llentry *la = 0;
268	u_int flags = 0;
269	int error, renew;
270
271	*lle = NULL;
272	if (m != NULL) {
273		if (m->m_flags & M_BCAST) {
274			/* broadcast */
275			(void)memcpy(desten,
276			    ifp->if_broadcastaddr, ifp->if_addrlen);
277			return (0);
278		}
279		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
280			/* multicast */
281			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
282			return (0);
283		}
284	}
285	/* XXXXX
286	 */
287retry:
288	IF_AFDATA_RLOCK(ifp);
289	la = lla_lookup(LLTABLE(ifp), flags, dst);
290	IF_AFDATA_RUNLOCK(ifp);
291	if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
292	    && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
293		flags |= (LLE_CREATE | LLE_EXCLUSIVE);
294		IF_AFDATA_WLOCK(ifp);
295		la = lla_lookup(LLTABLE(ifp), flags, dst);
296		IF_AFDATA_WUNLOCK(ifp);
297	}
298	if (la == NULL) {
299		if (flags & LLE_CREATE)
300			log(LOG_DEBUG,
301			    "arpresolve: can't allocate llinfo for %s\n",
302			    inet_ntoa(SIN(dst)->sin_addr));
303		m_freem(m);
304		return (EINVAL);
305	}
306
307	if ((la->la_flags & LLE_VALID) &&
308	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
309		bcopy(&la->ll_addr, desten, ifp->if_addrlen);
310		/*
311		 * If entry has an expiry time and it is approaching,
312		 * see if we need to send an ARP request within this
313		 * arpt_down interval.
314		 */
315		if (!(la->la_flags & LLE_STATIC) &&
316		    time_uptime + la->la_preempt > la->la_expire) {
317			arprequest(ifp, NULL,
318			    &SIN(dst)->sin_addr, IF_LLADDR(ifp));
319
320			la->la_preempt--;
321		}
322
323		*lle = la;
324		error = 0;
325		goto done;
326	}
327
328	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
329		log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
330		    inet_ntoa(SIN(dst)->sin_addr));
331		m_freem(m);
332		error = EINVAL;
333		goto done;
334	}
335
336	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
337	if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
338		flags |= LLE_EXCLUSIVE;
339		LLE_RUNLOCK(la);
340		goto retry;
341	}
342	/*
343	 * There is an arptab entry, but no ethernet address
344	 * response yet.  Replace the held mbuf with this
345	 * latest one.
346	 */
347	if (m != NULL) {
348		if (la->la_hold != NULL)
349			m_freem(la->la_hold);
350		la->la_hold = m;
351		if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
352			flags &= ~LLE_EXCLUSIVE;
353			LLE_DOWNGRADE(la);
354		}
355
356	}
357	/*
358	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
359	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
360	 * if we have already sent arp_maxtries ARP requests. Retransmit the
361	 * ARP request, but not faster than one request per second.
362	 */
363	if (la->la_asked < V_arp_maxtries)
364		error = EWOULDBLOCK;	/* First request. */
365	else
366		error =
367		    (rt0->rt_flags & RTF_GATEWAY) ? EHOSTDOWN : EHOSTUNREACH;
368
369	if (renew) {
370		LLE_ADDREF(la);
371		la->la_expire = time_uptime;
372		callout_reset(&la->la_timer, hz, arptimer, la);
373		la->la_asked++;
374		LLE_WUNLOCK(la);
375		arprequest(ifp, NULL, &SIN(dst)->sin_addr,
376		    IF_LLADDR(ifp));
377		return (error);
378	}
379done:
380	if (flags & LLE_EXCLUSIVE)
381		LLE_WUNLOCK(la);
382	else
383		LLE_RUNLOCK(la);
384	return (error);
385}
386
387/*
388 * Common length and type checks are done here,
389 * then the protocol-specific routine is called.
390 */
391static void
392arpintr(struct mbuf *m)
393{
394	struct arphdr *ar;
395
396	if (m->m_len < sizeof(struct arphdr) &&
397	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
398		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
399		return;
400	}
401	ar = mtod(m, struct arphdr *);
402
403	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
404	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
405	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
406	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
407		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
408		    (unsigned char *)&ar->ar_hrd, "");
409		m_freem(m);
410		return;
411	}
412
413	if (m->m_len < arphdr_len(ar)) {
414		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
415			log(LOG_ERR, "arp: runt packet\n");
416			m_freem(m);
417			return;
418		}
419		ar = mtod(m, struct arphdr *);
420	}
421
422	switch (ntohs(ar->ar_pro)) {
423#ifdef INET
424	case ETHERTYPE_IP:
425		in_arpinput(m);
426		return;
427#endif
428	}
429	m_freem(m);
430}
431
432#ifdef INET
433/*
434 * ARP for Internet protocols on 10 Mb/s Ethernet.
435 * Algorithm is that given in RFC 826.
436 * In addition, a sanity check is performed on the sender
437 * protocol address, to catch impersonators.
438 * We no longer handle negotiations for use of trailer protocol:
439 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
440 * along with IP replies if we wanted trailers sent to us,
441 * and also sent them in response to IP replies.
442 * This allowed either end to announce the desire to receive
443 * trailer packets.
444 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
445 * but formerly didn't normally send requests.
446 */
447static int log_arp_wrong_iface = 1;
448static int log_arp_movements = 1;
449static int log_arp_permanent_modify = 1;
450
451SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
452	&log_arp_wrong_iface, 0,
453	"log arp packets arriving on the wrong interface");
454SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
455        &log_arp_movements, 0,
456        "log arp replies from MACs different than the one in the cache");
457SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
458        &log_arp_permanent_modify, 0,
459        "log arp replies from MACs different than the one in the permanent arp entry");
460
461
462static void
463in_arpinput(struct mbuf *m)
464{
465	struct arphdr *ah;
466	struct ifnet *ifp = m->m_pkthdr.rcvif;
467	struct llentry *la = NULL;
468	struct rtentry *rt;
469	struct ifaddr *ifa;
470	struct in_ifaddr *ia;
471	struct sockaddr sa;
472	struct in_addr isaddr, itaddr, myaddr;
473	u_int8_t *enaddr = NULL;
474	int op, flags;
475	struct mbuf *m0;
476	int req_len;
477	int bridged = 0, is_bridge = 0;
478#ifdef DEV_CARP
479	int carp_match = 0;
480#endif
481	struct sockaddr_in sin;
482	sin.sin_len = sizeof(struct sockaddr_in);
483	sin.sin_family = AF_INET;
484	sin.sin_addr.s_addr = 0;
485	INIT_VNET_INET(ifp->if_vnet);
486
487	if (ifp->if_bridge)
488		bridged = 1;
489	if (ifp->if_type == IFT_BRIDGE)
490		is_bridge = 1;
491
492	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
493	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
494		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
495		return;
496	}
497
498	ah = mtod(m, struct arphdr *);
499	op = ntohs(ah->ar_op);
500	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
501	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
502
503	/*
504	 * For a bridge, we want to check the address irrespective
505	 * of the receive interface. (This will change slightly
506	 * when we have clusters of interfaces).
507	 * If the interface does not match, but the recieving interface
508	 * is part of carp, we call carp_iamatch to see if this is a
509	 * request for the virtual host ip.
510	 * XXX: This is really ugly!
511	 */
512	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
513		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
514		    ia->ia_ifp == ifp) &&
515		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
516			ifa_ref(&ia->ia_ifa);
517			goto match;
518		}
519#ifdef DEV_CARP
520		if (ifp->if_carp != NULL &&
521		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
522		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
523			carp_match = 1;
524			ifa_ref(&ia->ia_ifa);
525			goto match;
526		}
527#endif
528	}
529	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
530		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
531		    ia->ia_ifp == ifp) &&
532		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
533			ifa_ref(&ia->ia_ifa);
534			goto match;
535		}
536
537#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
538  (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
539  !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
540  addr == ia->ia_addr.sin_addr.s_addr)
541	/*
542	 * Check the case when bridge shares its MAC address with
543	 * some of its children, so packets are claimed by bridge
544	 * itself (bridge_input() does it first), but they are really
545	 * meant to be destined to the bridge member.
546	 */
547	if (is_bridge) {
548		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
549			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
550				ifa_ref(&ia->ia_ifa);
551				ifp = ia->ia_ifp;
552				goto match;
553			}
554		}
555	}
556#undef BDG_MEMBER_MATCHES_ARP
557
558	/*
559	 * No match, use the first inet address on the receive interface
560	 * as a dummy address for the rest of the function.
561	 */
562	IF_ADDR_LOCK(ifp);
563	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
564		if (ifa->ifa_addr->sa_family == AF_INET) {
565			ia = ifatoia(ifa);
566			ifa_ref(ifa);
567			IF_ADDR_UNLOCK(ifp);
568			goto match;
569		}
570	IF_ADDR_UNLOCK(ifp);
571
572	/*
573	 * If bridging, fall back to using any inet address.
574	 */
575	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL)
576		goto drop;
577	ifa_ref(&ia->ia_ifa);
578match:
579	if (!enaddr)
580		enaddr = (u_int8_t *)IF_LLADDR(ifp);
581	myaddr = ia->ia_addr.sin_addr;
582	ifa_free(&ia->ia_ifa);
583	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
584		goto drop;	/* it's from me, ignore it. */
585	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
586		log(LOG_ERR,
587		    "arp: link address is broadcast for IP address %s!\n",
588		    inet_ntoa(isaddr));
589		goto drop;
590	}
591	/*
592	 * Warn if another host is using the same IP address, but only if the
593	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
594	 * case we suppress the warning to avoid false positive complaints of
595	 * potential misconfiguration.
596	 */
597	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
598		log(LOG_ERR,
599		   "arp: %*D is using my IP address %s on %s!\n",
600		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
601		   inet_ntoa(isaddr), ifp->if_xname);
602		itaddr = myaddr;
603		goto reply;
604	}
605	if (ifp->if_flags & IFF_STATICARP)
606		goto reply;
607
608	bzero(&sin, sizeof(sin));
609	sin.sin_len = sizeof(struct sockaddr_in);
610	sin.sin_family = AF_INET;
611	sin.sin_addr = isaddr;
612	flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
613	flags |= LLE_EXCLUSIVE;
614	IF_AFDATA_LOCK(ifp);
615	la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
616	IF_AFDATA_UNLOCK(ifp);
617	if (la != NULL) {
618		/* the following is not an error when doing bridging */
619		if (!bridged && la->lle_tbl->llt_ifp != ifp
620#ifdef DEV_CARP
621		    && (ifp->if_type != IFT_CARP || !carp_match)
622#endif
623			) {
624			if (log_arp_wrong_iface)
625				log(LOG_ERR, "arp: %s is on %s "
626				    "but got reply from %*D on %s\n",
627				    inet_ntoa(isaddr),
628				    la->lle_tbl->llt_ifp->if_xname,
629				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
630				    ifp->if_xname);
631			goto reply;
632		}
633		if ((la->la_flags & LLE_VALID) &&
634		    bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
635			if (la->la_flags & LLE_STATIC) {
636				log(LOG_ERR,
637				    "arp: %*D attempts to modify permanent "
638				    "entry for %s on %s\n",
639				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
640				    inet_ntoa(isaddr), ifp->if_xname);
641				goto reply;
642			}
643			if (log_arp_movements) {
644			        log(LOG_INFO, "arp: %s moved from %*D "
645				    "to %*D on %s\n",
646				    inet_ntoa(isaddr),
647				    ifp->if_addrlen,
648				    (u_char *)&la->ll_addr, ":",
649				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
650				    ifp->if_xname);
651			}
652		}
653
654		if (ifp->if_addrlen != ah->ar_hln) {
655			log(LOG_WARNING,
656			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
657			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
658			    ah->ar_hln, ifp->if_addrlen);
659			goto reply;
660		}
661		(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
662		la->la_flags |= LLE_VALID;
663
664		if (!(la->la_flags & LLE_STATIC)) {
665			la->la_expire = time_uptime + V_arpt_keep;
666			callout_reset(&la->la_timer, hz * V_arpt_keep,
667			    arptimer, la);
668		}
669		la->la_asked = 0;
670		la->la_preempt = V_arp_maxtries;
671		if (la->la_hold != NULL) {
672			m0 = la->la_hold;
673			la->la_hold = 0;
674			memcpy(&sa, L3_ADDR(la), sizeof(sa));
675			LLE_WUNLOCK(la);
676
677			(*ifp->if_output)(ifp, m0, &sa, NULL);
678			return;
679		}
680	}
681reply:
682	if (op != ARPOP_REQUEST)
683		goto drop;
684
685	if (itaddr.s_addr == myaddr.s_addr) {
686		/* Shortcut.. the receiving interface is the target. */
687		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
688		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
689	} else {
690		struct llentry *lle = NULL;
691
692		if (!V_arp_proxyall)
693			goto drop;
694
695		sin.sin_addr = itaddr;
696		/* XXX MRT use table 0 for arp reply  */
697		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
698		if (!rt)
699			goto drop;
700
701		/*
702		 * Don't send proxies for nodes on the same interface
703		 * as this one came out of, or we'll get into a fight
704		 * over who claims what Ether address.
705		 */
706		if (!rt->rt_ifp || rt->rt_ifp == ifp) {
707			RTFREE_LOCKED(rt);
708			goto drop;
709		}
710		IF_AFDATA_LOCK(rt->rt_ifp);
711		lle = lla_lookup(LLTABLE(rt->rt_ifp), 0, (struct sockaddr *)&sin);
712		IF_AFDATA_UNLOCK(rt->rt_ifp);
713		RTFREE_LOCKED(rt);
714
715		if (lle != NULL) {
716			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
717			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
718			LLE_RUNLOCK(lle);
719		} else
720			goto drop;
721
722		/*
723		 * Also check that the node which sent the ARP packet
724		 * is on the the interface we expect it to be on. This
725		 * avoids ARP chaos if an interface is connected to the
726		 * wrong network.
727		 */
728		sin.sin_addr = isaddr;
729
730		/* XXX MRT use table 0 for arp checks */
731		rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
732		if (!rt)
733			goto drop;
734		if (rt->rt_ifp != ifp) {
735			log(LOG_INFO, "arp_proxy: ignoring request"
736			    " from %s via %s, expecting %s\n",
737			    inet_ntoa(isaddr), ifp->if_xname,
738			    rt->rt_ifp->if_xname);
739			RTFREE_LOCKED(rt);
740			goto drop;
741		}
742		RTFREE_LOCKED(rt);
743
744#ifdef DEBUG_PROXY
745		printf("arp: proxying for %s\n",
746		       inet_ntoa(itaddr));
747#endif
748	}
749
750	if (la != NULL)
751		LLE_WUNLOCK(la);
752	if (itaddr.s_addr == myaddr.s_addr &&
753	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
754		/* RFC 3927 link-local IPv4; always reply by broadcast. */
755#ifdef DEBUG_LINKLOCAL
756		printf("arp: sending reply for link-local addr %s\n",
757		    inet_ntoa(itaddr));
758#endif
759		m->m_flags |= M_BCAST;
760		m->m_flags &= ~M_MCAST;
761	} else {
762		/* default behaviour; never reply by broadcast. */
763		m->m_flags &= ~(M_BCAST|M_MCAST);
764	}
765	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
766	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
767	ah->ar_op = htons(ARPOP_REPLY);
768	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
769	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
770	m->m_pkthdr.len = m->m_len;
771	sa.sa_family = AF_ARP;
772	sa.sa_len = 2;
773	(*ifp->if_output)(ifp, m, &sa, NULL);
774	return;
775
776drop:
777	if (la != NULL)
778		LLE_WUNLOCK(la);
779	m_freem(m);
780}
781#endif
782
783void
784arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
785{
786	struct llentry *lle;
787
788	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
789		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
790				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
791		/*
792		 * interface address is considered static entry
793		 * because the output of the arp utility shows
794		 * that L2 entry as permanent
795		 */
796		IF_AFDATA_LOCK(ifp);
797		lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
798				 (struct sockaddr *)IA_SIN(ifa));
799		IF_AFDATA_UNLOCK(ifp);
800		if (lle == NULL)
801			log(LOG_INFO, "arp_ifinit: cannot create arp "
802			    "entry for interface address\n");
803		else
804			LLE_RUNLOCK(lle);
805	}
806	ifa->ifa_rtrequest = NULL;
807}
808
809void
810arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
811{
812	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
813		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
814				&IA_SIN(ifa)->sin_addr, enaddr);
815	ifa->ifa_rtrequest = NULL;
816}
817
818static int
819arp_iattach(const void *unused __unused)
820{
821	INIT_VNET_INET(curvnet);
822
823	V_arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
824	V_arp_maxtries = 5;
825	V_useloopback = 1; /* use loopback interface for local traffic */
826	V_arp_proxyall = 0;
827
828	return (0);
829}
830
831static void
832arp_init(void)
833{
834
835#ifndef VIMAGE_GLOBALS
836	vnet_mod_register(&vnet_arp_modinfo);
837#else
838	arp_iattach(NULL);
839#endif
840
841	netisr_register(&arp_nh);
842}
843SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
844