xref: /dflybsd-src/sys/netinet/if_ether.c (revision bc76a771df54af7e361532b257cecc26227736b4)
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: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $
35  * $DragonFly: src/sys/netinet/if_ether.c,v 1.11 2004/04/09 22:34:10 hsu Exp $
36  */
37 
38 /*
39  * Ethernet address resolution protocol.
40  * TODO:
41  *	add "inuse/lock" bit (or ref. count) along with valid bit
42  */
43 
44 #include "opt_inet.h"
45 #include "opt_bdg.h"
46 
47 #include <sys/param.h>
48 #include <sys/kernel.h>
49 #include <sys/queue.h>
50 #include <sys/sysctl.h>
51 #include <sys/systm.h>
52 #include <sys/mbuf.h>
53 #include <sys/malloc.h>
54 #include <sys/socket.h>
55 #include <sys/syslog.h>
56 
57 #include <sys/thread2.h>
58 #include <sys/msgport2.h>
59 
60 #include <net/if.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/if_llc.h>
66 #ifdef BRIDGE
67 #include <net/ethernet.h>
68 #include <net/bridge/bridge.h>
69 #endif
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/if_ether.h>
74 
75 #include <net/if_arc.h>
76 #include <net/iso88025.h>
77 
78 #define SIN(s) ((struct sockaddr_in *)s)
79 #define SDL(s) ((struct sockaddr_dl *)s)
80 
81 SYSCTL_DECL(_net_link_ether);
82 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
83 
84 /* timer values */
85 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
86 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
87 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
88 
89 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
90 	   &arpt_prune, 0, "");
91 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
92 	   &arpt_keep, 0, "");
93 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
94 	   &arpt_down, 0, "");
95 
96 #define	rt_expire rt_rmx.rmx_expire
97 
98 struct llinfo_arp {
99 	LIST_ENTRY(llinfo_arp) la_le;
100 	struct	rtentry *la_rt;
101 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
102 	u_short	la_preempt;	/* #times we QUERIED before entry expiration */
103 	u_short	la_asked;	/* #times we QUERIED following expiration */
104 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
105 };
106 
107 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
108 
109 static int	arp_inuse, arp_allocated, arpinit_done;
110 
111 static int	arp_maxtries = 5;
112 static int	useloopback = 1; /* use loopback interface for local traffic */
113 static int	arp_proxyall = 0;
114 
115 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
116 	   &arp_maxtries, 0, "");
117 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
118 	   &useloopback, 0, "");
119 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
120 	   &arp_proxyall, 0, "");
121 
122 static void	arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
123 static void	arprequest (struct ifnet *,
124 			struct in_addr *, struct in_addr *, u_char *);
125 static void	arpintr(struct netmsg *);
126 static void	arptfree (struct llinfo_arp *);
127 static void	arptimer (void *);
128 static struct llinfo_arp
129 		*arplookup (u_long, int, int);
130 #ifdef INET
131 static void	in_arpinput (struct mbuf *);
132 #endif
133 
134 u_char arcbroadcastaddr = 0;
135 
136 #define IF_BCASTADDR(ifp)					\
137 	((ifp)->if_type == IFT_ARCNET ?				\
138 		&arcbroadcastaddr : etherbroadcastaddr)
139 
140 /*
141  * Timeout routine.  Age arp_tab entries periodically.
142  */
143 /* ARGSUSED */
144 static void
145 arptimer(ignored_arg)
146 	void *ignored_arg;
147 {
148 	int s = splnet();
149 	struct llinfo_arp *la = LIST_FIRST(&llinfo_arp);
150 	struct llinfo_arp *ola;
151 
152 	timeout(arptimer, (caddr_t)0, arpt_prune * hz);
153 	while ((ola = la) != 0) {
154 		struct rtentry *rt = la->la_rt;
155 		la = LIST_NEXT(la, la_le);
156 		if (rt->rt_expire && rt->rt_expire <= time_second)
157 			arptfree(ola); /* timer has expired, clear */
158 	}
159 	splx(s);
160 }
161 
162 /*
163  * Parallel to llc_rtrequest.
164  */
165 static void
166 arp_rtrequest(req, rt, info)
167 	int req;
168 	struct rtentry *rt;
169 	struct rt_addrinfo *info;
170 {
171 	struct sockaddr *gate = rt->rt_gateway;
172 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
173 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
174 
175 	if (!arpinit_done) {
176 		arpinit_done = 1;
177 		timeout(arptimer, (caddr_t)0, hz);
178 	}
179 	if (rt->rt_flags & RTF_GATEWAY)
180 		return;
181 	switch (req) {
182 
183 	case RTM_ADD:
184 		/*
185 		 * XXX: If this is a manually added route to interface
186 		 * such as older version of routed or gated might provide,
187 		 * restore cloning bit.
188 		 */
189 		if ((rt->rt_flags & RTF_HOST) == 0 &&
190 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
191 			rt->rt_flags |= RTF_CLONING;
192 		if (rt->rt_flags & RTF_CLONING) {
193 			/*
194 			 * Case 1: This route should come from a route to iface.
195 			 */
196 			rt_setgate(rt, rt_key(rt),
197 					(struct sockaddr *)&null_sdl);
198 			gate = rt->rt_gateway;
199 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
200 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
201 			rt->rt_expire = time_second;
202 			break;
203 		}
204 		/* Announce a new entry if requested. */
205 		if (rt->rt_flags & RTF_ANNOUNCE)
206 			arprequest(rt->rt_ifp,
207 			    &SIN(rt_key(rt))->sin_addr,
208 			    &SIN(rt_key(rt))->sin_addr,
209 			    (u_char *)LLADDR(SDL(gate)));
210 		/*FALLTHROUGH*/
211 	case RTM_RESOLVE:
212 		if (gate->sa_family != AF_LINK ||
213 		    gate->sa_len < sizeof(null_sdl)) {
214 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
215 			break;
216 		}
217 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
218 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
219 		if (la != 0)
220 			break; /* This happens on a route change */
221 		/*
222 		 * Case 2:  This route may come from cloning, or a manual route
223 		 * add with a LL address.
224 		 */
225 		R_Malloc(la, struct llinfo_arp *, sizeof(*la));
226 		rt->rt_llinfo = (caddr_t)la;
227 		if (la == 0) {
228 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
229 			break;
230 		}
231 		arp_inuse++, arp_allocated++;
232 		Bzero(la, sizeof(*la));
233 		la->la_rt = rt;
234 		rt->rt_flags |= RTF_LLINFO;
235 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
236 
237 #ifdef INET
238 		/*
239 		 * This keeps the multicast addresses from showing up
240 		 * in `arp -a' listings as unresolved.  It's not actually
241 		 * functional.  Then the same for broadcast.
242 		 */
243 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))
244 		&&  rt->rt_ifp->if_type != IFT_ARCNET) {
245 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
246 					       LLADDR(SDL(gate)));
247 			SDL(gate)->sdl_alen = 6;
248 			rt->rt_expire = 0;
249 		}
250 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
251 			memcpy(LLADDR(SDL(gate)), IF_BCASTADDR(rt->rt_ifp),
252 			       rt->rt_ifp->if_addrlen);
253 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
254 			rt->rt_expire = 0;
255 		}
256 #endif
257 
258 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
259 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
260 		    /*
261 		     * This test used to be
262 		     *	if (loif.if_flags & IFF_UP)
263 		     * It allowed local traffic to be forced
264 		     * through the hardware by configuring the loopback down.
265 		     * However, it causes problems during network configuration
266 		     * for boards that can't receive packets they send.
267 		     * It is now necessary to clear "useloopback" and remove
268 		     * the route to force traffic out to the hardware.
269 		     */
270 			rt->rt_expire = 0;
271 			Bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
272 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
273 			if (useloopback)
274 				rt->rt_ifp = loif;
275 
276 		}
277 		break;
278 
279 	case RTM_DELETE:
280 		if (la == 0)
281 			break;
282 		arp_inuse--;
283 		LIST_REMOVE(la, la_le);
284 		rt->rt_llinfo = 0;
285 		rt->rt_flags &= ~RTF_LLINFO;
286 		if (la->la_hold)
287 			m_freem(la->la_hold);
288 		Free((caddr_t)la);
289 	}
290 }
291 
292 /*
293  * Broadcast an ARP request. Caller specifies:
294  *	- arp header source ip address
295  *	- arp header target ip address
296  *	- arp header source ethernet address
297  */
298 static void
299 arprequest(ifp, sip, tip, enaddr)
300 	struct ifnet *ifp;
301 	struct in_addr *sip, *tip;
302 	u_char *enaddr;
303 {
304 	struct mbuf *m;
305 	struct ether_header *eh;
306 	struct arc_header *arh;
307 	struct arphdr *ah;
308 	struct sockaddr sa;
309 	static u_char	llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
310 				   LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
311 	u_short ar_hrd;
312 
313 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
314 		return;
315 	m->m_pkthdr.rcvif = (struct ifnet *)0;
316 	switch (ifp->if_type) {
317 	case IFT_ARCNET:
318 		ar_hrd = htons(ARPHRD_ARCNET);
319 
320 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
321 		m->m_pkthdr.len = m->m_len;
322 		MH_ALIGN(m, m->m_len);
323 
324 		arh = (struct arc_header *)sa.sa_data;
325 		arh->arc_dhost = arcbroadcastaddr;
326 		arh->arc_type = ARCTYPE_ARP;
327 
328 		ah = mtod(m, struct arphdr *);
329 		break;
330 
331 	case IFT_ISO88025:
332 		ar_hrd = htons(ARPHRD_IEEE802);
333 
334 		m->m_len = sizeof(llcx) +
335 		    arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
336 		m->m_pkthdr.len = m->m_len;
337 		MH_ALIGN(m, m->m_len);
338 
339 		(void)memcpy(mtod(m, caddr_t), llcx, sizeof(llcx));
340 		(void)memcpy(sa.sa_data, etherbroadcastaddr, 6);
341 		(void)memcpy(sa.sa_data + 6, enaddr, 6);
342 		sa.sa_data[6] |= TR_RII;
343 		sa.sa_data[12] = TR_AC;
344 		sa.sa_data[13] = TR_LLC_FRAME;
345 
346 		ah = (struct arphdr *)(mtod(m, char *) + sizeof(llcx));
347 		break;
348 	case IFT_FDDI:
349 	case IFT_ETHER:
350 		/*
351 		 * This may not be correct for types not explicitly
352 		 * listed, but this is our best guess
353 		 */
354 	default:
355 		ar_hrd = htons(ARPHRD_ETHER);
356 
357 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
358 		m->m_pkthdr.len = m->m_len;
359 		MH_ALIGN(m, m->m_len);
360 
361 		eh = (struct ether_header *)sa.sa_data;
362 		/* if_output will not swap */
363 		eh->ether_type = htons(ETHERTYPE_ARP);
364 		(void)memcpy(eh->ether_dhost, etherbroadcastaddr,
365 		    sizeof(eh->ether_dhost));
366 
367 		ah = mtod(m, struct arphdr *);
368 		break;
369 	}
370 
371 	ah->ar_hrd = ar_hrd;
372 	ah->ar_pro = htons(ETHERTYPE_IP);
373 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
374 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
375 	ah->ar_op = htons(ARPOP_REQUEST);
376 	(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
377 	memset(ar_tha(ah), 0, ah->ar_hln);
378 	(void)memcpy(ar_spa(ah), sip, ah->ar_pln);
379 	(void)memcpy(ar_tpa(ah), tip, ah->ar_pln);
380 
381 	sa.sa_family = AF_UNSPEC;
382 	sa.sa_len = sizeof(sa);
383 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
384 }
385 
386 /*
387  * Resolve an IP address into an ethernet address.  If success,
388  * desten is filled in.  If there is no entry in arptab,
389  * set one up and broadcast a request for the IP address.
390  * Hold onto this mbuf and resend it once the address
391  * is finally resolved.  A return value of 1 indicates
392  * that desten has been filled in and the packet should be sent
393  * normally; a 0 return indicates that the packet has been
394  * taken over here, either now or for later transmission.
395  */
396 int
397 arpresolve(ifp, rt, m, dst, desten, rt0)
398 	struct ifnet *ifp;
399 	struct rtentry *rt;
400 	struct mbuf *m;
401 	struct sockaddr *dst;
402 	u_char *desten;
403 	struct rtentry *rt0;
404 {
405 	struct llinfo_arp *la = 0;
406 	struct sockaddr_dl *sdl;
407 
408 	if (m->m_flags & M_BCAST) {	/* broadcast */
409 		(void)memcpy(desten, IF_BCASTADDR(ifp), ifp->if_addrlen);
410 		return (1);
411 	}
412 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
413 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
414 		return(1);
415 	}
416 	if (rt)
417 		la = (struct llinfo_arp *)rt->rt_llinfo;
418 	if (la == 0) {
419 		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
420 		if (la)
421 			rt = la->la_rt;
422 	}
423 	if (la == 0 || rt == 0) {
424 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
425 			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
426 				rt ? "rt" : "");
427 		m_freem(m);
428 		return (0);
429 	}
430 	sdl = SDL(rt->rt_gateway);
431 	/*
432 	 * Check the address family and length is valid, the address
433 	 * is resolved; otherwise, try to resolve.
434 	 */
435 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
436 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
437 		/*
438 		 * If entry has an expiry time and it is approaching,
439 		 * see if we need to send an ARP request within this
440 		 * arpt_down interval.
441 		 */
442 		if ((rt->rt_expire != 0) &&
443 		    (time_second + (arp_maxtries - la->la_preempt) * arpt_down
444 		     > rt->rt_expire)) {
445 			arprequest(ifp,
446 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
447 				   &SIN(dst)->sin_addr,
448 				   IF_LLADDR(ifp));
449 			la->la_preempt++;
450 		}
451 
452 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
453 		return 1;
454 	}
455 	/*
456 	 * If ARP is disabled on this interface, stop.
457 	 * XXX
458 	 * Probably should not allocate empty llinfo struct if we are
459 	 * not going to be sending out an arp request.
460 	 */
461 	if (ifp->if_flags & IFF_NOARP) {
462 		m_freem(m);
463 		return (0);
464 	}
465 	/*
466 	 * There is an arptab entry, but no ethernet address
467 	 * response yet.  Replace the held mbuf with this
468 	 * latest one.
469 	 */
470 	if (la->la_hold)
471 		m_freem(la->la_hold);
472 	la->la_hold = m;
473 	if (rt->rt_expire) {
474 		rt->rt_flags &= ~RTF_REJECT;
475 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
476 			rt->rt_expire = time_second;
477 			if (la->la_asked++ < arp_maxtries) {
478 				arprequest(ifp,
479 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
480 					   &SIN(dst)->sin_addr,
481 					   IF_LLADDR(ifp));
482 			} else {
483 				rt->rt_flags |= RTF_REJECT;
484 				rt->rt_expire += arpt_down;
485 				la->la_preempt = la->la_asked = 0;
486 			}
487 
488 		}
489 	}
490 	return (0);
491 }
492 
493 /*
494  * Common length and type checks are done here,
495  * then the protocol-specific routine is called.
496  */
497 static void
498 arpintr(struct netmsg *msg)
499 {
500 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
501 	struct arphdr *ar;
502 	u_short ar_hrd;
503 
504 	if (m->m_len < sizeof(struct arphdr) &&
505 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
506 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
507 		return;
508 	}
509 	ar = mtod(m, struct arphdr *);
510 
511 	ar_hrd = ntohs(ar->ar_hrd);
512 	if (ar_hrd != ARPHRD_ETHER &&
513 	    ar_hrd != ARPHRD_IEEE802 &&
514 	    ar_hrd != ARPHRD_ARCNET) {
515 		log(LOG_ERR,
516 		    "arp: unknown hardware address format (0x%2D)\n",
517 		    (unsigned char *)&ar->ar_hrd, "");
518 		m_freem(m);
519 		return;
520 	}
521 
522 	if (m->m_pkthdr.len < arphdr_len(ar) &&
523 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
524 		log(LOG_ERR, "arp: runt packet\n");
525 		m_freem(m);
526 		return;
527 	}
528 
529 	switch (ntohs(ar->ar_pro)) {
530 #ifdef INET
531 		case ETHERTYPE_IP:
532 			in_arpinput(m);
533 			return;
534 #endif
535 	}
536 	m_freem(m);
537 	lwkt_replymsg(&msg->nm_lmsg, 0);
538 }
539 
540 #ifdef INET
541 /*
542  * ARP for Internet protocols on 10 Mb/s Ethernet.
543  * Algorithm is that given in RFC 826.
544  * In addition, a sanity check is performed on the sender
545  * protocol address, to catch impersonators.
546  * We no longer handle negotiations for use of trailer protocol:
547  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
548  * along with IP replies if we wanted trailers sent to us,
549  * and also sent them in response to IP replies.
550  * This allowed either end to announce the desire to receive
551  * trailer packets.
552  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
553  * but formerly didn't normally send requests.
554  */
555 static int log_arp_wrong_iface = 1;
556 
557 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
558 	&log_arp_wrong_iface, 0,
559 	"log arp packets arriving on the wrong interface");
560 
561 static void
562 in_arpinput(m)
563 	struct mbuf *m;
564 {
565 	struct arphdr *ah;
566 	struct ifnet *ifp = m->m_pkthdr.rcvif;
567 	struct ether_header *eh;
568 	struct arc_header *arh;
569 	struct iso88025_header *th = (struct iso88025_header *)0;
570 	struct iso88025_sockaddr_dl_data *trld;
571 	struct llinfo_arp *la = 0;
572 	struct rtentry *rt;
573 	struct ifaddr *ifa;
574 	struct in_ifaddr *ia;
575 	struct sockaddr_dl *sdl;
576 	struct sockaddr sa;
577 	struct in_addr isaddr, itaddr, myaddr;
578 	int op, rif_len;
579 	int req_len;
580 
581 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
582 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
583 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
584 		return;
585 	}
586 
587 	ah = mtod(m, struct arphdr *);
588 	op = ntohs(ah->ar_op);
589 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
590 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
591 #ifdef BRIDGE
592 #define BRIDGE_TEST (do_bridge)
593 #else
594 #define BRIDGE_TEST (0) /* cc will optimise the test away */
595 #endif
596 	/*
597 	 * For a bridge, we want to check the address irrespective
598 	 * of the receive interface. (This will change slightly
599 	 * when we have clusters of interfaces).
600 	 */
601 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash)
602 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
603 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
604 			goto match;
605 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
606 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
607 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
608 			goto match;
609 	/*
610 	 * No match, use the first inet address on the receive interface
611 	 * as a dummy address for the rest of the function.
612 	 */
613 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
614 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
615 			ia = ifatoia(ifa);
616 			goto match;
617 		}
618 	/*
619 	 * If bridging, fall back to using any inet address.
620 	 * This is probably incorrect, the right way being try to match
621 	 * addresses for interfaces in the same cluster, so if we
622 	 * get here we should always drop the packet.
623 	 */
624 	if (!BRIDGE_TEST ||
625 	    (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) {
626 		m_freem(m);
627 		return;
628 	}
629 match:
630 	myaddr = ia->ia_addr.sin_addr;
631 	if (!bcmp(ar_sha(ah), IF_LLADDR(ifp), ifp->if_addrlen)) {
632 		m_freem(m);	/* it's from me, ignore it. */
633 		return;
634 	}
635 	if (!bcmp(ar_sha(ah), IF_BCASTADDR(ifp), ifp->if_addrlen)) {
636 		log(LOG_ERR,
637 		    "arp: link address is broadcast for IP address %s!\n",
638 		    inet_ntoa(isaddr));
639 		m_freem(m);
640 		return;
641 	}
642 	if (isaddr.s_addr == myaddr.s_addr) {
643 		log(LOG_ERR,
644 		   "arp: %*D is using my IP address %s!\n",
645 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
646 		   inet_ntoa(isaddr));
647 		itaddr = myaddr;
648 		goto reply;
649 	}
650 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
651 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
652 		/* the following is not an error when doing bridging */
653 		if (!BRIDGE_TEST && rt->rt_ifp != ifp) {
654 		    if (log_arp_wrong_iface)
655 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
656 			    inet_ntoa(isaddr),
657 			    rt->rt_ifp->if_xname,
658 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
659 			    ifp->if_xname);
660 		    goto reply;
661 		}
662 		if (sdl->sdl_alen &&
663 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
664 			if (rt->rt_expire)
665 			    log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
666 				inet_ntoa(isaddr),
667 				ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
668 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
669 				ifp->if_xname);
670 			else {
671 			    log(LOG_ERR,
672 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
673 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
674 				inet_ntoa(isaddr), ifp->if_xname);
675 			    goto reply;
676 			}
677 		}
678 		/*
679 		 * sanity check for the address length.
680 		 * XXX this does not work for protocols with variable address
681 		 * length. -is
682 		 */
683 		if (sdl->sdl_alen &&
684 		    sdl->sdl_alen != ah->ar_hln) {
685 			log(LOG_WARNING,
686 			    "arp from %*D: new addr len %d, was %d",
687 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
688 			    ah->ar_hln, sdl->sdl_alen);
689 		}
690 		if (ifp->if_addrlen != ah->ar_hln) {
691 			log(LOG_WARNING,
692 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
693 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
694 			    ah->ar_hln, ifp->if_addrlen);
695 			goto reply;
696 		}
697 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
698 		    sdl->sdl_alen = ah->ar_hln);
699 		/*
700 		 * If we receive an arp from a token-ring station over
701 		 * a token-ring nic then try to save the source
702 		 * routing info.
703 		 */
704 		if (ifp->if_type == IFT_ISO88025) {
705 			th = (struct iso88025_header *)m->m_pkthdr.header;
706 			trld = SDL_ISO88025(sdl);
707 			rif_len = TR_RCF_RIFLEN(th->rcf);
708 			if ((th->iso88025_shost[0] & TR_RII) &&
709 			    (rif_len > 2)) {
710 				trld->trld_rcf = th->rcf;
711 				trld->trld_rcf ^= htons(TR_RCF_DIR);
712 				memcpy(trld->trld_route, th->rd, rif_len - 2);
713 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
714 				/*
715 				 * Set up source routing information for
716 				 * reply packet (XXX)
717 				 */
718 				m->m_data -= rif_len;
719 				m->m_len  += rif_len;
720 				m->m_pkthdr.len += rif_len;
721 			} else {
722 				th->iso88025_shost[0] &= ~TR_RII;
723 				trld->trld_rcf = 0;
724 			}
725 			m->m_data -= 8;
726 			m->m_len  += 8;
727 			m->m_pkthdr.len += 8;
728 			th->rcf = trld->trld_rcf;
729 		}
730 		if (rt->rt_expire)
731 			rt->rt_expire = time_second + arpt_keep;
732 		rt->rt_flags &= ~RTF_REJECT;
733 		la->la_preempt = la->la_asked = 0;
734 		if (la->la_hold) {
735 			(*ifp->if_output)(ifp, la->la_hold,
736 				rt_key(rt), rt);
737 			la->la_hold = 0;
738 		}
739 	}
740 reply:
741 	if (op != ARPOP_REQUEST) {
742 		m_freem(m);
743 		return;
744 	}
745 	if (itaddr.s_addr == myaddr.s_addr) {
746 		/* I am the target */
747 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
748 		(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
749 	} else {
750 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
751 		if (la == NULL) {
752 			struct sockaddr_in sin;
753 
754 			if (!arp_proxyall) {
755 				m_freem(m);
756 				return;
757 			}
758 
759 			bzero(&sin, sizeof sin);
760 			sin.sin_family = AF_INET;
761 			sin.sin_len = sizeof sin;
762 			sin.sin_addr = itaddr;
763 
764 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
765 			if (!rt) {
766 				m_freem(m);
767 				return;
768 			}
769 			/*
770 			 * Don't send proxies for nodes on the same interface
771 			 * as this one came out of, or we'll get into a fight
772 			 * over who claims what Ether address.
773 			 */
774 			if (rt->rt_ifp == ifp) {
775 				rtfree(rt);
776 				m_freem(m);
777 				return;
778 			}
779 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
780 			(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
781 			rtfree(rt);
782 #ifdef DEBUG_PROXY
783 			printf("arp: proxying for %s\n",
784 			       inet_ntoa(itaddr));
785 #endif
786 		} else {
787 			rt = la->la_rt;
788 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
789 			sdl = SDL(rt->rt_gateway);
790 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
791 		}
792 	}
793 
794 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
795 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
796 	ah->ar_op = htons(ARPOP_REPLY);
797 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
798 	switch (ifp->if_type) {
799 	case IFT_ARCNET:
800 		arh = (struct arc_header *)sa.sa_data;
801 		arh->arc_dhost = *ar_tha(ah);
802 		arh->arc_type = ARCTYPE_ARP;
803 		break;
804 
805 	case IFT_ISO88025:
806 		/* Re-arrange the source/dest address */
807 		memcpy(th->iso88025_dhost, th->iso88025_shost,
808 		    sizeof(th->iso88025_dhost));
809 		memcpy(th->iso88025_shost, IF_LLADDR(ifp),
810 		    sizeof(th->iso88025_shost));
811 		/* Set the source routing bit if neccesary */
812 		if (th->iso88025_dhost[0] & TR_RII) {
813 			th->iso88025_dhost[0] &= ~TR_RII;
814 			if (TR_RCF_RIFLEN(th->rcf) > 2)
815 				th->iso88025_shost[0] |= TR_RII;
816 		}
817 		/* Copy the addresses, ac and fc into sa_data */
818 		memcpy(sa.sa_data, th->iso88025_dhost,
819 		    sizeof(th->iso88025_dhost) * 2);
820 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2)] = TR_AC;
821 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2) + 1] = TR_LLC_FRAME;
822 		break;
823 	case IFT_ETHER:
824 	case IFT_FDDI:
825 	/*
826 	 * May not be correct for types not explictly
827 	 * listed, but it is our best guess.
828 	 */
829 	default:
830 		eh = (struct ether_header *)sa.sa_data;
831 		(void)memcpy(eh->ether_dhost, ar_tha(ah),
832 		    sizeof(eh->ether_dhost));
833 		eh->ether_type = htons(ETHERTYPE_ARP);
834 		break;
835 	}
836 	sa.sa_family = AF_UNSPEC;
837 	sa.sa_len = sizeof(sa);
838 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
839 	return;
840 }
841 #endif
842 
843 /*
844  * Free an arp entry.
845  */
846 static void
847 arptfree(la)
848 	struct llinfo_arp *la;
849 {
850 	struct rtentry *rt = la->la_rt;
851 	struct sockaddr_dl *sdl;
852 	if (rt == 0)
853 		panic("arptfree");
854 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
855 	    sdl->sdl_family == AF_LINK) {
856 		sdl->sdl_alen = 0;
857 		la->la_preempt = la->la_asked = 0;
858 		rt->rt_flags &= ~RTF_REJECT;
859 		return;
860 	}
861 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
862 			0, (struct rtentry **)0);
863 }
864 /*
865  * Lookup or enter a new address in arptab.
866  */
867 static struct llinfo_arp *
868 arplookup(addr, create, proxy)
869 	u_long addr;
870 	int create, proxy;
871 {
872 	struct rtentry *rt;
873 	static struct sockaddr_inarp sin = {sizeof(sin), AF_INET };
874 	const char *why = 0;
875 
876 	sin.sin_addr.s_addr = addr;
877 	sin.sin_other = proxy ? SIN_PROXY : 0;
878 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
879 	if (rt == 0)
880 		return (0);
881 	rt->rt_refcnt--;
882 
883 	if (rt->rt_flags & RTF_GATEWAY)
884 		why = "host is not on local network";
885 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
886 		why = "could not allocate llinfo";
887 	else if (rt->rt_gateway->sa_family != AF_LINK)
888 		why = "gateway route is not ours";
889 
890 	if (why) {
891 		if (create) {
892 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
893 			    inet_ntoa(sin.sin_addr), why);
894 		}
895 
896 		/* if there are no references to this route, purge it */
897 		if (rt->rt_refcnt <= 0 &&
898 		    (rt->rt_flags & RTF_WASCLONED) == RTF_WASCLONED) {
899 			    rtrequest(RTM_DELETE,
900 				(struct sockaddr *)rt_key(rt), rt->rt_gateway,
901 				rt_mask(rt), rt->rt_flags, 0);
902 		}
903 		return (0);
904 	}
905 	return ((struct llinfo_arp *)rt->rt_llinfo);
906 }
907 
908 void
909 arp_ifinit(ifp, ifa)
910 	struct ifnet *ifp;
911 	struct ifaddr *ifa;
912 {
913 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
914 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
915 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
916 	ifa->ifa_rtrequest = arp_rtrequest;
917 	ifa->ifa_flags |= RTF_CLONING;
918 }
919 
920 static void
921 arp_init(void)
922 {
923 	LIST_INIT(&llinfo_arp);
924 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr);
925 }
926 
927 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
928