if_ether.c revision 74851
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 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
34 * $FreeBSD: head/sys/netinet/if_ether.c 74851 2001-03-27 12:34:58Z yar $
35 */
36
37/*
38 * Ethernet address resolution protocol.
39 * TODO:
40 *	add "inuse/lock" bit (or ref. count) along with valid bit
41 */
42
43#include "opt_inet.h"
44#include "opt_bdg.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/socket.h>
54#include <sys/syslog.h>
55
56#include <net/if.h>
57#include <net/if_dl.h>
58#include <net/if_types.h>
59#include <net/route.h>
60#include <net/netisr.h>
61#include <net/if_llc.h>
62#ifdef BRIDGE
63#include <net/ethernet.h>
64#include <net/bridge.h>
65#endif
66
67#include <netinet/in.h>
68#include <netinet/in_var.h>
69#include <netinet/if_ether.h>
70
71#include <net/iso88025.h>
72
73#define SIN(s) ((struct sockaddr_in *)s)
74#define SDL(s) ((struct sockaddr_dl *)s)
75
76SYSCTL_DECL(_net_link_ether);
77SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
78
79/* timer values */
80static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
81static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
82static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
83
84SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
85	   &arpt_prune, 0, "");
86SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
87	   &arpt_keep, 0, "");
88SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
89	   &arpt_down, 0, "");
90
91#define	rt_expire rt_rmx.rmx_expire
92
93struct llinfo_arp {
94	LIST_ENTRY(llinfo_arp) la_le;
95	struct	rtentry *la_rt;
96	struct	mbuf *la_hold;		/* last packet until resolved/timeout */
97	long	la_asked;		/* last time we QUERIED for this addr */
98#define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
99};
100
101static	LIST_HEAD(, llinfo_arp) llinfo_arp;
102
103struct	ifqueue arpintrq;
104static int	arp_inuse, arp_allocated;
105
106static int	arp_maxtries = 5;
107static int	useloopback = 1; /* use loopback interface for local traffic */
108static int	arp_proxyall = 0;
109
110SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
111	   &arp_maxtries, 0, "");
112SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
113	   &useloopback, 0, "");
114SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
115	   &arp_proxyall, 0, "");
116
117static void	arp_init __P((void));
118static void	arp_rtrequest __P((int, struct rtentry *, struct sockaddr *));
119static void	arprequest __P((struct arpcom *,
120			struct in_addr *, struct in_addr *, u_char *));
121static void	arpintr __P((void));
122static void	arptfree __P((struct llinfo_arp *));
123static void	arptimer __P((void *));
124static struct llinfo_arp
125		*arplookup __P((u_long, int, int));
126#ifdef INET
127static void	in_arpinput __P((struct mbuf *));
128#endif
129
130/*
131 * Timeout routine.  Age arp_tab entries periodically.
132 */
133/* ARGSUSED */
134static void
135arptimer(ignored_arg)
136	void *ignored_arg;
137{
138	int s = splnet();
139	register struct llinfo_arp *la = LIST_FIRST(&llinfo_arp);
140	struct llinfo_arp *ola;
141
142	timeout(arptimer, (caddr_t)0, arpt_prune * hz);
143	while ((ola = la) != 0) {
144		register struct rtentry *rt = la->la_rt;
145		la = LIST_NEXT(la, la_le);
146		if (rt->rt_expire && rt->rt_expire <= time_second)
147			arptfree(ola); /* timer has expired, clear */
148	}
149	splx(s);
150}
151
152/*
153 * Parallel to llc_rtrequest.
154 */
155static void
156arp_rtrequest(req, rt, sa)
157	int req;
158	register struct rtentry *rt;
159	struct sockaddr *sa;
160{
161	register struct sockaddr *gate = rt->rt_gateway;
162	register struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
163	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
164	static int arpinit_done;
165
166	if (!arpinit_done) {
167		arpinit_done = 1;
168		LIST_INIT(&llinfo_arp);
169		timeout(arptimer, (caddr_t)0, hz);
170		register_netisr(NETISR_ARP, arpintr);
171	}
172	if (rt->rt_flags & RTF_GATEWAY)
173		return;
174	switch (req) {
175
176	case RTM_ADD:
177		/*
178		 * XXX: If this is a manually added route to interface
179		 * such as older version of routed or gated might provide,
180		 * restore cloning bit.
181		 */
182		if ((rt->rt_flags & RTF_HOST) == 0 &&
183		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
184			rt->rt_flags |= RTF_CLONING;
185		if (rt->rt_flags & RTF_CLONING) {
186			/*
187			 * Case 1: This route should come from a route to iface.
188			 */
189			rt_setgate(rt, rt_key(rt),
190					(struct sockaddr *)&null_sdl);
191			gate = rt->rt_gateway;
192			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
193			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
194			rt->rt_expire = time_second;
195			break;
196		}
197		/* Announce a new entry if requested. */
198		if (rt->rt_flags & RTF_ANNOUNCE)
199			arprequest((struct arpcom *)rt->rt_ifp,
200			    &SIN(rt_key(rt))->sin_addr,
201			    &SIN(rt_key(rt))->sin_addr,
202			    (u_char *)LLADDR(SDL(gate)));
203		/*FALLTHROUGH*/
204	case RTM_RESOLVE:
205		if (gate->sa_family != AF_LINK ||
206		    gate->sa_len < sizeof(null_sdl)) {
207			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
208			break;
209		}
210		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
211		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
212		if (la != 0)
213			break; /* This happens on a route change */
214		/*
215		 * Case 2:  This route may come from cloning, or a manual route
216		 * add with a LL address.
217		 */
218		R_Malloc(la, struct llinfo_arp *, sizeof(*la));
219		rt->rt_llinfo = (caddr_t)la;
220		if (la == 0) {
221			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
222			break;
223		}
224		arp_inuse++, arp_allocated++;
225		Bzero(la, sizeof(*la));
226		la->la_rt = rt;
227		rt->rt_flags |= RTF_LLINFO;
228		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
229
230#ifdef INET
231		/*
232		 * This keeps the multicast addresses from showing up
233		 * in `arp -a' listings as unresolved.  It's not actually
234		 * functional.  Then the same for broadcast.
235		 */
236		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
237			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
238					       LLADDR(SDL(gate)));
239			SDL(gate)->sdl_alen = 6;
240			rt->rt_expire = 0;
241		}
242		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
243			memcpy(LLADDR(SDL(gate)), etherbroadcastaddr, 6);
244			SDL(gate)->sdl_alen = 6;
245			rt->rt_expire = 0;
246		}
247#endif
248
249		if (SIN(rt_key(rt))->sin_addr.s_addr ==
250		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
251		    /*
252		     * This test used to be
253		     *	if (loif.if_flags & IFF_UP)
254		     * It allowed local traffic to be forced
255		     * through the hardware by configuring the loopback down.
256		     * However, it causes problems during network configuration
257		     * for boards that can't receive packets they send.
258		     * It is now necessary to clear "useloopback" and remove
259		     * the route to force traffic out to the hardware.
260		     */
261			rt->rt_expire = 0;
262			Bcopy(((struct arpcom *)rt->rt_ifp)->ac_enaddr,
263				LLADDR(SDL(gate)), SDL(gate)->sdl_alen = 6);
264			if (useloopback)
265				rt->rt_ifp = loif;
266
267		}
268		break;
269
270	case RTM_DELETE:
271		if (la == 0)
272			break;
273		arp_inuse--;
274		LIST_REMOVE(la, la_le);
275		rt->rt_llinfo = 0;
276		rt->rt_flags &= ~RTF_LLINFO;
277		if (la->la_hold)
278			m_freem(la->la_hold);
279		Free((caddr_t)la);
280	}
281}
282
283/*
284 * Broadcast an ARP request. Caller specifies:
285 *	- arp header source ip address
286 *	- arp header target ip address
287 *	- arp header source ethernet address
288 */
289static void
290arprequest(ac, sip, tip, enaddr)
291	register struct arpcom *ac;
292	register struct in_addr *sip, *tip;
293	register u_char *enaddr;
294{
295	register struct mbuf *m;
296	register struct ether_header *eh;
297	register struct ether_arp *ea;
298	struct sockaddr sa;
299	static u_char	llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
300				   LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
301
302	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
303		return;
304	m->m_pkthdr.rcvif = (struct ifnet *)0;
305	switch (ac->ac_if.if_type) {
306	case IFT_ISO88025:
307		m->m_len = sizeof(*ea) + sizeof(llcx);
308		m->m_pkthdr.len = sizeof(*ea) + sizeof(llcx);
309		MH_ALIGN(m, sizeof(*ea) + sizeof(llcx));
310		(void)memcpy(mtod(m, caddr_t), llcx, sizeof(llcx));
311		(void)memcpy(sa.sa_data, etherbroadcastaddr, 6);
312		(void)memcpy(sa.sa_data + 6, enaddr, 6);
313		sa.sa_data[6] |= TR_RII;
314		sa.sa_data[12] = TR_AC;
315		sa.sa_data[13] = TR_LLC_FRAME;
316		ea = (struct ether_arp *)(mtod(m, char *) + sizeof(llcx));
317		bzero((caddr_t)ea, sizeof (*ea));
318		ea->arp_hrd = htons(ARPHRD_IEEE802);
319		break;
320	case IFT_FDDI:
321	case IFT_ETHER:
322		/*
323		 * This may not be correct for types not explicitly
324		 * listed, but this is our best guess
325		 */
326	default:
327		m->m_len = sizeof(*ea);
328		m->m_pkthdr.len = sizeof(*ea);
329		MH_ALIGN(m, sizeof(*ea));
330		ea = mtod(m, struct ether_arp *);
331		eh = (struct ether_header *)sa.sa_data;
332		bzero((caddr_t)ea, sizeof (*ea));
333		/* if_output will not swap */
334		eh->ether_type = htons(ETHERTYPE_ARP);
335		(void)memcpy(eh->ether_dhost, etherbroadcastaddr,
336		    sizeof(eh->ether_dhost));
337		ea->arp_hrd = htons(ARPHRD_ETHER);
338		break;
339	}
340	ea->arp_pro = htons(ETHERTYPE_IP);
341	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
342	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
343	ea->arp_op = htons(ARPOP_REQUEST);
344	(void)memcpy(ea->arp_sha, enaddr, sizeof(ea->arp_sha));
345	(void)memcpy(ea->arp_spa, sip, sizeof(ea->arp_spa));
346	(void)memcpy(ea->arp_tpa, tip, sizeof(ea->arp_tpa));
347	sa.sa_family = AF_UNSPEC;
348	sa.sa_len = sizeof(sa);
349	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
350}
351
352/*
353 * Resolve an IP address into an ethernet address.  If success,
354 * desten is filled in.  If there is no entry in arptab,
355 * set one up and broadcast a request for the IP address.
356 * Hold onto this mbuf and resend it once the address
357 * is finally resolved.  A return value of 1 indicates
358 * that desten has been filled in and the packet should be sent
359 * normally; a 0 return indicates that the packet has been
360 * taken over here, either now or for later transmission.
361 */
362int
363arpresolve(ac, rt, m, dst, desten, rt0)
364	register struct arpcom *ac;
365	register struct rtentry *rt;
366	struct mbuf *m;
367	register struct sockaddr *dst;
368	register u_char *desten;
369	struct rtentry *rt0;
370{
371	register struct llinfo_arp *la = 0;
372	struct sockaddr_dl *sdl;
373
374	if (m->m_flags & M_BCAST) {	/* broadcast */
375		(void)memcpy(desten, etherbroadcastaddr, sizeof(etherbroadcastaddr));
376		return (1);
377	}
378	if (m->m_flags & M_MCAST) {	/* multicast */
379		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
380		return(1);
381	}
382	if (rt)
383		la = (struct llinfo_arp *)rt->rt_llinfo;
384	if (la == 0) {
385		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
386		if (la)
387			rt = la->la_rt;
388	}
389	if (la == 0 || rt == 0) {
390		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
391			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
392				rt ? "rt" : "");
393		m_freem(m);
394		return (0);
395	}
396	sdl = SDL(rt->rt_gateway);
397	/*
398	 * Check the address family and length is valid, the address
399	 * is resolved; otherwise, try to resolve.
400	 */
401	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
402	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
403		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
404		return 1;
405	}
406	/*
407	 * There is an arptab entry, but no ethernet address
408	 * response yet.  Replace the held mbuf with this
409	 * latest one.
410	 */
411	if (la->la_hold)
412		m_freem(la->la_hold);
413	la->la_hold = m;
414	if (rt->rt_expire) {
415		rt->rt_flags &= ~RTF_REJECT;
416		if (la->la_asked == 0 || rt->rt_expire != time_second) {
417			rt->rt_expire = time_second;
418			if (la->la_asked++ < arp_maxtries)
419			    arprequest(ac,
420			        &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
421				&SIN(dst)->sin_addr, ac->ac_enaddr);
422			else {
423				rt->rt_flags |= RTF_REJECT;
424				rt->rt_expire += arpt_down;
425				la->la_asked = 0;
426			}
427
428		}
429	}
430	return (0);
431}
432
433/*
434 * Common length and type checks are done here,
435 * then the protocol-specific routine is called.
436 */
437static void
438arpintr()
439{
440	register struct mbuf *m;
441	register struct arphdr *ar;
442	int s;
443
444	while (arpintrq.ifq_head) {
445		s = splimp();
446		IF_DEQUEUE(&arpintrq, m);
447		splx(s);
448		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
449			panic("arpintr");
450
451                if (m->m_len < sizeof(struct arphdr) &&
452                    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
453			log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
454			continue;
455		}
456		ar = mtod(m, struct arphdr *);
457
458		if (ntohs(ar->ar_hrd) != ARPHRD_ETHER
459		    && ntohs(ar->ar_hrd) != ARPHRD_IEEE802) {
460			log(LOG_ERR,
461			    "arp: unknown hardware address format (0x%2D)\n",
462			    (unsigned char *)&ar->ar_hrd, "");
463			m_freem(m);
464			continue;
465		}
466
467		if (m->m_pkthdr.len < sizeof(struct arphdr) + 2 * ar->ar_hln
468		    + 2 * ar->ar_pln) {
469			log(LOG_ERR, "arp: runt packet\n");
470			m_freem(m);
471			continue;
472		}
473
474		switch (ntohs(ar->ar_pro)) {
475#ifdef INET
476			case ETHERTYPE_IP:
477				in_arpinput(m);
478				continue;
479#endif
480		}
481		m_freem(m);
482	}
483}
484
485#ifdef INET
486/*
487 * ARP for Internet protocols on 10 Mb/s Ethernet.
488 * Algorithm is that given in RFC 826.
489 * In addition, a sanity check is performed on the sender
490 * protocol address, to catch impersonators.
491 * We no longer handle negotiations for use of trailer protocol:
492 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
493 * along with IP replies if we wanted trailers sent to us,
494 * and also sent them in response to IP replies.
495 * This allowed either end to announce the desire to receive
496 * trailer packets.
497 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
498 * but formerly didn't normally send requests.
499 */
500static int log_arp_wrong_iface = 1;
501
502SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
503	&log_arp_wrong_iface, 0,
504	"log arp packets arriving on the wrong interface");
505
506static void
507in_arpinput(m)
508	struct mbuf *m;
509{
510	register struct ether_arp *ea;
511	register struct arpcom *ac = (struct arpcom *)m->m_pkthdr.rcvif;
512	struct ether_header *eh;
513	struct iso88025_header *th = (struct iso88025_header *)0;
514	register struct llinfo_arp *la = 0;
515	register struct rtentry *rt;
516	struct in_ifaddr *ia, *maybe_ia = 0;
517	struct sockaddr_dl *sdl;
518	struct sockaddr sa;
519	struct in_addr isaddr, itaddr, myaddr;
520	int op, rif_len;
521
522	if (m->m_len < sizeof(struct ether_arp) &&
523	    (m = m_pullup(m, sizeof(struct ether_arp))) == NULL) {
524		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
525		return;
526	}
527
528	ea = mtod(m, struct ether_arp *);
529	op = ntohs(ea->arp_op);
530	(void)memcpy(&isaddr, ea->arp_spa, sizeof (isaddr));
531	(void)memcpy(&itaddr, ea->arp_tpa, sizeof (itaddr));
532	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
533		/*
534		 * For a bridge, we want to check the address irrespective
535		 * of the receive interface. (This will change slightly
536		 * when we have clusters of interfaces).
537		 */
538#ifdef BRIDGE
539#define BRIDGE_TEST (do_bridge)
540#else
541#define BRIDGE_TEST (0) /* cc will optimise the test away */
542#endif
543		if ((BRIDGE_TEST) || (ia->ia_ifp == &ac->ac_if)) {
544			maybe_ia = ia;
545			if ((itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) ||
546			     (isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
547				break;
548			}
549		}
550	}
551	if (maybe_ia == 0) {
552		m_freem(m);
553		return;
554	}
555	myaddr = ia ? ia->ia_addr.sin_addr : maybe_ia->ia_addr.sin_addr;
556	if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)ac->ac_enaddr,
557	    sizeof (ea->arp_sha))) {
558		m_freem(m);	/* it's from me, ignore it. */
559		return;
560	}
561	if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)etherbroadcastaddr,
562	    sizeof (ea->arp_sha))) {
563		log(LOG_ERR,
564		    "arp: ether address is broadcast for IP address %s!\n",
565		    inet_ntoa(isaddr));
566		m_freem(m);
567		return;
568	}
569	if (isaddr.s_addr == myaddr.s_addr) {
570		log(LOG_ERR,
571		   "arp: %6D is using my IP address %s!\n",
572		   ea->arp_sha, ":", inet_ntoa(isaddr));
573		itaddr = myaddr;
574		goto reply;
575	}
576	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
577	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
578		/* the following is not an error when doing bridging */
579		if (!BRIDGE_TEST && rt->rt_ifp != &ac->ac_if) {
580			if (log_arp_wrong_iface)
581				log(LOG_ERR, "arp: %s is on %s%d but got reply from %6D on %s%d\n",
582				    inet_ntoa(isaddr),
583				    rt->rt_ifp->if_name, rt->rt_ifp->if_unit,
584				    ea->arp_sha, ":",
585				    ac->ac_if.if_name, ac->ac_if.if_unit);
586			goto reply;
587		}
588		if (sdl->sdl_alen &&
589		    bcmp((caddr_t)ea->arp_sha, LLADDR(sdl), sdl->sdl_alen)) {
590			if (rt->rt_expire)
591			    log(LOG_INFO, "arp: %s moved from %6D to %6D on %s%d\n",
592				inet_ntoa(isaddr), (u_char *)LLADDR(sdl), ":",
593				ea->arp_sha, ":",
594				ac->ac_if.if_name, ac->ac_if.if_unit);
595			else {
596			    log(LOG_ERR,
597				"arp: %6D attempts to modify permanent entry for %s on %s%d\n",
598				ea->arp_sha, ":", inet_ntoa(isaddr),
599				ac->ac_if.if_name, ac->ac_if.if_unit);
600			    goto reply;
601			}
602		}
603		(void)memcpy(LLADDR(sdl), ea->arp_sha, sizeof(ea->arp_sha));
604		sdl->sdl_alen = sizeof(ea->arp_sha);
605                sdl->sdl_rcf = (u_short)0;
606		/*
607		 * If we receive an arp from a token-ring station over
608		 * a token-ring nic then try to save the source
609		 * routing info.
610		 */
611		if (ac->ac_if.if_type == IFT_ISO88025) {
612			th = (struct iso88025_header *)m->m_pkthdr.header;
613			rif_len = TR_RCF_RIFLEN(th->rcf);
614			if ((th->iso88025_shost[0] & TR_RII) &&
615			    (rif_len > 2)) {
616				sdl->sdl_rcf = th->rcf;
617				sdl->sdl_rcf ^= htons(TR_RCF_DIR);
618				memcpy(sdl->sdl_route, th->rd, rif_len - 2);
619				sdl->sdl_rcf &= ~htons(TR_RCF_BCST_MASK);
620				/*
621				 * Set up source routing information for
622				 * reply packet (XXX)
623				 */
624				m->m_data -= rif_len;
625				m->m_len  += rif_len;
626				m->m_pkthdr.len += rif_len;
627			} else {
628				th->iso88025_shost[0] &= ~TR_RII;
629			}
630			m->m_data -= 8;
631			m->m_len  += 8;
632			m->m_pkthdr.len += 8;
633			th->rcf = sdl->sdl_rcf;
634		} else {
635			sdl->sdl_rcf = (u_short)0;
636		}
637		if (rt->rt_expire)
638			rt->rt_expire = time_second + arpt_keep;
639		rt->rt_flags &= ~RTF_REJECT;
640		la->la_asked = 0;
641		if (la->la_hold) {
642			(*ac->ac_if.if_output)(&ac->ac_if, la->la_hold,
643				rt_key(rt), rt);
644			la->la_hold = 0;
645		}
646	}
647reply:
648	if (op != ARPOP_REQUEST) {
649		m_freem(m);
650		return;
651	}
652	if (itaddr.s_addr == myaddr.s_addr) {
653		/* I am the target */
654		(void)memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
655		(void)memcpy(ea->arp_sha, ac->ac_enaddr, sizeof(ea->arp_sha));
656	} else {
657		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
658		if (la == NULL) {
659			struct sockaddr_in sin;
660
661			if (!arp_proxyall) {
662				m_freem(m);
663				return;
664			}
665
666			bzero(&sin, sizeof sin);
667			sin.sin_family = AF_INET;
668			sin.sin_len = sizeof sin;
669			sin.sin_addr = itaddr;
670
671			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
672			if (!rt) {
673				m_freem(m);
674				return;
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 == &ac->ac_if) {
682				rtfree(rt);
683				m_freem(m);
684				return;
685			}
686			(void)memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
687			(void)memcpy(ea->arp_sha, ac->ac_enaddr, sizeof(ea->arp_sha));
688			rtfree(rt);
689
690			/*
691			 * Also check that the node which sent the ARP packet
692			 * is on the the interface we expect it to be on. This
693			 * avoids ARP chaos if an interface is connected to the
694			 * wrong network.
695			 */
696			sin.sin_addr = isaddr;
697
698			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
699			if (!rt) {
700				m_freem(m);
701				return;
702			}
703			if (rt->rt_ifp != &ac->ac_if) {
704				log(LOG_INFO, "arp_proxy: ignoring request"
705				    " from %s via %s%d, expecting %s%d\n",
706				    inet_ntoa(isaddr), ac->ac_if.if_name,
707				    ac->ac_if.if_unit, rt->rt_ifp->if_name,
708				    rt->rt_ifp->if_unit);
709				rtfree(rt);
710				m_freem(m);
711				return;
712			}
713			rtfree(rt);
714
715#ifdef DEBUG_PROXY
716			printf("arp: proxying for %s\n",
717			       inet_ntoa(itaddr));
718#endif
719		} else {
720			rt = la->la_rt;
721			(void)memcpy(ea->arp_tha, ea->arp_sha, sizeof(ea->arp_sha));
722			sdl = SDL(rt->rt_gateway);
723			(void)memcpy(ea->arp_sha, LLADDR(sdl), sizeof(ea->arp_sha));
724		}
725	}
726
727	(void)memcpy(ea->arp_tpa, ea->arp_spa, sizeof(ea->arp_spa));
728	(void)memcpy(ea->arp_spa, &itaddr, sizeof(ea->arp_spa));
729	ea->arp_op = htons(ARPOP_REPLY);
730	ea->arp_pro = htons(ETHERTYPE_IP); /* let's be sure! */
731	switch (ac->ac_if.if_type) {
732	case IFT_ISO88025:
733		/* Re-arrange the source/dest address */
734		memcpy(th->iso88025_dhost, th->iso88025_shost,
735		    sizeof(th->iso88025_dhost));
736		memcpy(th->iso88025_shost, ac->ac_enaddr,
737		    sizeof(th->iso88025_shost));
738		/* Set the source routing bit if neccesary */
739		if (th->iso88025_dhost[0] & TR_RII) {
740			th->iso88025_dhost[0] &= ~TR_RII;
741			if (TR_RCF_RIFLEN(th->rcf) > 2)
742				th->iso88025_shost[0] |= TR_RII;
743		}
744		/* Copy the addresses, ac and fc into sa_data */
745		memcpy(sa.sa_data, th->iso88025_dhost,
746		    sizeof(th->iso88025_dhost) * 2);
747		sa.sa_data[(sizeof(th->iso88025_dhost) * 2)] = TR_AC;
748		sa.sa_data[(sizeof(th->iso88025_dhost) * 2) + 1] = TR_LLC_FRAME;
749		break;
750	case IFT_ETHER:
751	case IFT_FDDI:
752	/*
753	 * May not be correct for types not explictly
754	 * listed, but it is our best guess.
755	 */
756	default:
757		eh = (struct ether_header *)sa.sa_data;
758		(void)memcpy(eh->ether_dhost, ea->arp_tha,
759		    sizeof(eh->ether_dhost));
760		eh->ether_type = htons(ETHERTYPE_ARP);
761		break;
762	}
763	sa.sa_family = AF_UNSPEC;
764	sa.sa_len = sizeof(sa);
765	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa, (struct rtentry *)0);
766	return;
767}
768#endif
769
770/*
771 * Free an arp entry.
772 */
773static void
774arptfree(la)
775	register struct llinfo_arp *la;
776{
777	register struct rtentry *rt = la->la_rt;
778	register struct sockaddr_dl *sdl;
779	if (rt == 0)
780		panic("arptfree");
781	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
782	    sdl->sdl_family == AF_LINK) {
783		sdl->sdl_alen = 0;
784		la->la_asked = 0;
785		rt->rt_flags &= ~RTF_REJECT;
786		return;
787	}
788	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
789			0, (struct rtentry **)0);
790}
791/*
792 * Lookup or enter a new address in arptab.
793 */
794static struct llinfo_arp *
795arplookup(addr, create, proxy)
796	u_long addr;
797	int create, proxy;
798{
799	register struct rtentry *rt;
800	static struct sockaddr_inarp sin = {sizeof(sin), AF_INET };
801	const char *why = 0;
802
803	sin.sin_addr.s_addr = addr;
804	sin.sin_other = proxy ? SIN_PROXY : 0;
805	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
806	if (rt == 0)
807		return (0);
808	rt->rt_refcnt--;
809
810	if (rt->rt_flags & RTF_GATEWAY)
811		why = "host is not on local network";
812	else if ((rt->rt_flags & RTF_LLINFO) == 0)
813		why = "could not allocate llinfo";
814	else if (rt->rt_gateway->sa_family != AF_LINK)
815		why = "gateway route is not ours";
816
817	if (why && create) {
818		log(LOG_DEBUG, "arplookup %s failed: %s\n",
819		    inet_ntoa(sin.sin_addr), why);
820		return 0;
821	} else if (why) {
822		return 0;
823	}
824	return ((struct llinfo_arp *)rt->rt_llinfo);
825}
826
827void
828arp_ifinit(ac, ifa)
829	struct arpcom *ac;
830	struct ifaddr *ifa;
831{
832	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
833		arprequest(ac, &IA_SIN(ifa)->sin_addr,
834			       &IA_SIN(ifa)->sin_addr, ac->ac_enaddr);
835	ifa->ifa_rtrequest = arp_rtrequest;
836	ifa->ifa_flags |= RTF_CLONING;
837}
838
839static void
840arp_init(void)
841{
842
843	arpintrq.ifq_maxlen = 50;
844	mtx_init(&arpintrq.ifq_mtx, "arp_inq", MTX_DEF);
845}
846
847SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
848