xref: /dragonfly/sys/netinet/if_ether.c (revision 28c7b939)
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 1.9 2004/01/06 03:17:27 dillon 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 <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/route.h>
61 #include <net/netisr.h>
62 #include <net/if_llc.h>
63 #ifdef BRIDGE
64 #include <net/ethernet.h>
65 #include <net/bridge/bridge.h>
66 #endif
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/if_ether.h>
71 
72 #include <net/if_arc.h>
73 #include <net/iso88025.h>
74 
75 #define SIN(s) ((struct sockaddr_in *)s)
76 #define SDL(s) ((struct sockaddr_dl *)s)
77 
78 SYSCTL_DECL(_net_link_ether);
79 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
80 
81 /* timer values */
82 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
83 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
84 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
85 
86 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
87 	   &arpt_prune, 0, "");
88 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
89 	   &arpt_keep, 0, "");
90 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
91 	   &arpt_down, 0, "");
92 
93 #define	rt_expire rt_rmx.rmx_expire
94 
95 struct llinfo_arp {
96 	LIST_ENTRY(llinfo_arp) la_le;
97 	struct	rtentry *la_rt;
98 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
99 	u_short	la_preempt;	/* #times we QUERIED before entry expiration */
100 	u_short	la_asked;	/* #times we QUERIED following expiration */
101 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
102 };
103 
104 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
105 
106 static int	arp_inuse, arp_allocated, arpinit_done;
107 
108 static int	arp_maxtries = 5;
109 static int	useloopback = 1; /* use loopback interface for local traffic */
110 static int	arp_proxyall = 0;
111 
112 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
113 	   &arp_maxtries, 0, "");
114 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
115 	   &useloopback, 0, "");
116 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
117 	   &arp_proxyall, 0, "");
118 
119 static void	arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
120 static void	arprequest (struct ifnet *,
121 			struct in_addr *, struct in_addr *, u_char *);
122 static void	arpintr(struct mbuf *);
123 static void	arptfree (struct llinfo_arp *);
124 static void	arptimer (void *);
125 static struct llinfo_arp
126 		*arplookup (u_long, int, int);
127 #ifdef INET
128 static void	in_arpinput (struct mbuf *);
129 #endif
130 
131 u_char arcbroadcastaddr = 0;
132 
133 #define IF_BCASTADDR(ifp)					\
134 	((ifp)->if_type == IFT_ARCNET ?				\
135 		&arcbroadcastaddr : etherbroadcastaddr)
136 
137 /*
138  * Timeout routine.  Age arp_tab entries periodically.
139  */
140 /* ARGSUSED */
141 static void
142 arptimer(ignored_arg)
143 	void *ignored_arg;
144 {
145 	int s = splnet();
146 	struct llinfo_arp *la = LIST_FIRST(&llinfo_arp);
147 	struct llinfo_arp *ola;
148 
149 	timeout(arptimer, (caddr_t)0, arpt_prune * hz);
150 	while ((ola = la) != 0) {
151 		struct rtentry *rt = la->la_rt;
152 		la = LIST_NEXT(la, la_le);
153 		if (rt->rt_expire && rt->rt_expire <= time_second)
154 			arptfree(ola); /* timer has expired, clear */
155 	}
156 	splx(s);
157 }
158 
159 /*
160  * Parallel to llc_rtrequest.
161  */
162 static void
163 arp_rtrequest(req, rt, info)
164 	int req;
165 	struct rtentry *rt;
166 	struct rt_addrinfo *info;
167 {
168 	struct sockaddr *gate = rt->rt_gateway;
169 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
170 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
171 
172 	if (!arpinit_done) {
173 		arpinit_done = 1;
174 		timeout(arptimer, (caddr_t)0, hz);
175 	}
176 	if (rt->rt_flags & RTF_GATEWAY)
177 		return;
178 	switch (req) {
179 
180 	case RTM_ADD:
181 		/*
182 		 * XXX: If this is a manually added route to interface
183 		 * such as older version of routed or gated might provide,
184 		 * restore cloning bit.
185 		 */
186 		if ((rt->rt_flags & RTF_HOST) == 0 &&
187 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
188 			rt->rt_flags |= RTF_CLONING;
189 		if (rt->rt_flags & RTF_CLONING) {
190 			/*
191 			 * Case 1: This route should come from a route to iface.
192 			 */
193 			rt_setgate(rt, rt_key(rt),
194 					(struct sockaddr *)&null_sdl);
195 			gate = rt->rt_gateway;
196 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
197 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
198 			rt->rt_expire = time_second;
199 			break;
200 		}
201 		/* Announce a new entry if requested. */
202 		if (rt->rt_flags & RTF_ANNOUNCE)
203 			arprequest(rt->rt_ifp,
204 			    &SIN(rt_key(rt))->sin_addr,
205 			    &SIN(rt_key(rt))->sin_addr,
206 			    (u_char *)LLADDR(SDL(gate)));
207 		/*FALLTHROUGH*/
208 	case RTM_RESOLVE:
209 		if (gate->sa_family != AF_LINK ||
210 		    gate->sa_len < sizeof(null_sdl)) {
211 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
212 			break;
213 		}
214 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
215 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
216 		if (la != 0)
217 			break; /* This happens on a route change */
218 		/*
219 		 * Case 2:  This route may come from cloning, or a manual route
220 		 * add with a LL address.
221 		 */
222 		R_Malloc(la, struct llinfo_arp *, sizeof(*la));
223 		rt->rt_llinfo = (caddr_t)la;
224 		if (la == 0) {
225 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
226 			break;
227 		}
228 		arp_inuse++, arp_allocated++;
229 		Bzero(la, sizeof(*la));
230 		la->la_rt = rt;
231 		rt->rt_flags |= RTF_LLINFO;
232 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
233 
234 #ifdef INET
235 		/*
236 		 * This keeps the multicast addresses from showing up
237 		 * in `arp -a' listings as unresolved.  It's not actually
238 		 * functional.  Then the same for broadcast.
239 		 */
240 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))
241 		&&  rt->rt_ifp->if_type != IFT_ARCNET) {
242 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
243 					       LLADDR(SDL(gate)));
244 			SDL(gate)->sdl_alen = 6;
245 			rt->rt_expire = 0;
246 		}
247 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
248 			memcpy(LLADDR(SDL(gate)), IF_BCASTADDR(rt->rt_ifp),
249 			       rt->rt_ifp->if_addrlen);
250 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
251 			rt->rt_expire = 0;
252 		}
253 #endif
254 
255 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
256 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
257 		    /*
258 		     * This test used to be
259 		     *	if (loif.if_flags & IFF_UP)
260 		     * It allowed local traffic to be forced
261 		     * through the hardware by configuring the loopback down.
262 		     * However, it causes problems during network configuration
263 		     * for boards that can't receive packets they send.
264 		     * It is now necessary to clear "useloopback" and remove
265 		     * the route to force traffic out to the hardware.
266 		     */
267 			rt->rt_expire = 0;
268 			Bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
269 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
270 			if (useloopback)
271 				rt->rt_ifp = loif;
272 
273 		}
274 		break;
275 
276 	case RTM_DELETE:
277 		if (la == 0)
278 			break;
279 		arp_inuse--;
280 		LIST_REMOVE(la, la_le);
281 		rt->rt_llinfo = 0;
282 		rt->rt_flags &= ~RTF_LLINFO;
283 		if (la->la_hold)
284 			m_freem(la->la_hold);
285 		Free((caddr_t)la);
286 	}
287 }
288 
289 /*
290  * Broadcast an ARP request. Caller specifies:
291  *	- arp header source ip address
292  *	- arp header target ip address
293  *	- arp header source ethernet address
294  */
295 static void
296 arprequest(ifp, sip, tip, enaddr)
297 	struct ifnet *ifp;
298 	struct in_addr *sip, *tip;
299 	u_char *enaddr;
300 {
301 	struct mbuf *m;
302 	struct ether_header *eh;
303 	struct arc_header *arh;
304 	struct arphdr *ah;
305 	struct sockaddr sa;
306 	static u_char	llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
307 				   LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
308 	u_short ar_hrd;
309 
310 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
311 		return;
312 	m->m_pkthdr.rcvif = (struct ifnet *)0;
313 	switch (ifp->if_type) {
314 	case IFT_ARCNET:
315 		ar_hrd = htons(ARPHRD_ARCNET);
316 
317 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
318 		m->m_pkthdr.len = m->m_len;
319 		MH_ALIGN(m, m->m_len);
320 
321 		arh = (struct arc_header *)sa.sa_data;
322 		arh->arc_dhost = arcbroadcastaddr;
323 		arh->arc_type = ARCTYPE_ARP;
324 
325 		ah = mtod(m, struct arphdr *);
326 		break;
327 
328 	case IFT_ISO88025:
329 		ar_hrd = htons(ARPHRD_IEEE802);
330 
331 		m->m_len = sizeof(llcx) +
332 		    arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
333 		m->m_pkthdr.len = m->m_len;
334 		MH_ALIGN(m, m->m_len);
335 
336 		(void)memcpy(mtod(m, caddr_t), llcx, sizeof(llcx));
337 		(void)memcpy(sa.sa_data, etherbroadcastaddr, 6);
338 		(void)memcpy(sa.sa_data + 6, enaddr, 6);
339 		sa.sa_data[6] |= TR_RII;
340 		sa.sa_data[12] = TR_AC;
341 		sa.sa_data[13] = TR_LLC_FRAME;
342 
343 		ah = (struct arphdr *)(mtod(m, char *) + sizeof(llcx));
344 		break;
345 	case IFT_FDDI:
346 	case IFT_ETHER:
347 		/*
348 		 * This may not be correct for types not explicitly
349 		 * listed, but this is our best guess
350 		 */
351 	default:
352 		ar_hrd = htons(ARPHRD_ETHER);
353 
354 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
355 		m->m_pkthdr.len = m->m_len;
356 		MH_ALIGN(m, m->m_len);
357 
358 		eh = (struct ether_header *)sa.sa_data;
359 		/* if_output will not swap */
360 		eh->ether_type = htons(ETHERTYPE_ARP);
361 		(void)memcpy(eh->ether_dhost, etherbroadcastaddr,
362 		    sizeof(eh->ether_dhost));
363 
364 		ah = mtod(m, struct arphdr *);
365 		break;
366 	}
367 
368 	ah->ar_hrd = ar_hrd;
369 	ah->ar_pro = htons(ETHERTYPE_IP);
370 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
371 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
372 	ah->ar_op = htons(ARPOP_REQUEST);
373 	(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
374 	memset(ar_tha(ah), 0, ah->ar_hln);
375 	(void)memcpy(ar_spa(ah), sip, ah->ar_pln);
376 	(void)memcpy(ar_tpa(ah), tip, ah->ar_pln);
377 
378 	sa.sa_family = AF_UNSPEC;
379 	sa.sa_len = sizeof(sa);
380 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
381 }
382 
383 /*
384  * Resolve an IP address into an ethernet address.  If success,
385  * desten is filled in.  If there is no entry in arptab,
386  * set one up and broadcast a request for the IP address.
387  * Hold onto this mbuf and resend it once the address
388  * is finally resolved.  A return value of 1 indicates
389  * that desten has been filled in and the packet should be sent
390  * normally; a 0 return indicates that the packet has been
391  * taken over here, either now or for later transmission.
392  */
393 int
394 arpresolve(ifp, rt, m, dst, desten, rt0)
395 	struct ifnet *ifp;
396 	struct rtentry *rt;
397 	struct mbuf *m;
398 	struct sockaddr *dst;
399 	u_char *desten;
400 	struct rtentry *rt0;
401 {
402 	struct llinfo_arp *la = 0;
403 	struct sockaddr_dl *sdl;
404 
405 	if (m->m_flags & M_BCAST) {	/* broadcast */
406 		(void)memcpy(desten, IF_BCASTADDR(ifp), ifp->if_addrlen);
407 		return (1);
408 	}
409 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
410 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
411 		return(1);
412 	}
413 	if (rt)
414 		la = (struct llinfo_arp *)rt->rt_llinfo;
415 	if (la == 0) {
416 		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
417 		if (la)
418 			rt = la->la_rt;
419 	}
420 	if (la == 0 || rt == 0) {
421 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
422 			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
423 				rt ? "rt" : "");
424 		m_freem(m);
425 		return (0);
426 	}
427 	sdl = SDL(rt->rt_gateway);
428 	/*
429 	 * Check the address family and length is valid, the address
430 	 * is resolved; otherwise, try to resolve.
431 	 */
432 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
433 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
434 		/*
435 		 * If entry has an expiry time and it is approaching,
436 		 * see if we need to send an ARP request within this
437 		 * arpt_down interval.
438 		 */
439 		if ((rt->rt_expire != 0) &&
440 		    (time_second + (arp_maxtries - la->la_preempt) * arpt_down
441 		     > rt->rt_expire)) {
442 			arprequest(ifp,
443 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
444 				   &SIN(dst)->sin_addr,
445 				   IF_LLADDR(ifp));
446 			la->la_preempt++;
447 		}
448 
449 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
450 		return 1;
451 	}
452 	/*
453 	 * If ARP is disabled on this interface, stop.
454 	 * XXX
455 	 * Probably should not allocate empty llinfo struct if we are
456 	 * not going to be sending out an arp request.
457 	 */
458 	if (ifp->if_flags & IFF_NOARP) {
459 		m_freem(m);
460 		return (0);
461 	}
462 	/*
463 	 * There is an arptab entry, but no ethernet address
464 	 * response yet.  Replace the held mbuf with this
465 	 * latest one.
466 	 */
467 	if (la->la_hold)
468 		m_freem(la->la_hold);
469 	la->la_hold = m;
470 	if (rt->rt_expire) {
471 		rt->rt_flags &= ~RTF_REJECT;
472 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
473 			rt->rt_expire = time_second;
474 			if (la->la_asked++ < arp_maxtries) {
475 				arprequest(ifp,
476 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
477 					   &SIN(dst)->sin_addr,
478 					   IF_LLADDR(ifp));
479 			} else {
480 				rt->rt_flags |= RTF_REJECT;
481 				rt->rt_expire += arpt_down;
482 				la->la_preempt = la->la_asked = 0;
483 			}
484 
485 		}
486 	}
487 	return (0);
488 }
489 
490 /*
491  * Common length and type checks are done here,
492  * then the protocol-specific routine is called.
493  */
494 static void
495 arpintr(struct mbuf *m)
496 {
497 	struct arphdr *ar;
498 	u_short ar_hrd;
499 
500 	if (m->m_len < sizeof(struct arphdr) &&
501 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
502 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
503 		return;
504 	}
505 	ar = mtod(m, struct arphdr *);
506 
507 	ar_hrd = ntohs(ar->ar_hrd);
508 	if (ar_hrd != ARPHRD_ETHER &&
509 	    ar_hrd != ARPHRD_IEEE802 &&
510 	    ar_hrd != ARPHRD_ARCNET) {
511 		log(LOG_ERR,
512 		    "arp: unknown hardware address format (0x%2D)\n",
513 		    (unsigned char *)&ar->ar_hrd, "");
514 		m_freem(m);
515 		return;
516 	}
517 
518 	if (m->m_pkthdr.len < arphdr_len(ar) &&
519 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
520 		log(LOG_ERR, "arp: runt packet\n");
521 		m_freem(m);
522 		return;
523 	}
524 
525 	switch (ntohs(ar->ar_pro)) {
526 #ifdef INET
527 		case ETHERTYPE_IP:
528 			in_arpinput(m);
529 			return;
530 #endif
531 	}
532 	m_freem(m);
533 }
534 
535 #ifdef INET
536 /*
537  * ARP for Internet protocols on 10 Mb/s Ethernet.
538  * Algorithm is that given in RFC 826.
539  * In addition, a sanity check is performed on the sender
540  * protocol address, to catch impersonators.
541  * We no longer handle negotiations for use of trailer protocol:
542  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
543  * along with IP replies if we wanted trailers sent to us,
544  * and also sent them in response to IP replies.
545  * This allowed either end to announce the desire to receive
546  * trailer packets.
547  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
548  * but formerly didn't normally send requests.
549  */
550 static int log_arp_wrong_iface = 1;
551 
552 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
553 	&log_arp_wrong_iface, 0,
554 	"log arp packets arriving on the wrong interface");
555 
556 static void
557 in_arpinput(m)
558 	struct mbuf *m;
559 {
560 	struct arphdr *ah;
561 	struct ifnet *ifp = m->m_pkthdr.rcvif;
562 	struct ether_header *eh;
563 	struct arc_header *arh;
564 	struct iso88025_header *th = (struct iso88025_header *)0;
565 	struct iso88025_sockaddr_dl_data *trld;
566 	struct llinfo_arp *la = 0;
567 	struct rtentry *rt;
568 	struct ifaddr *ifa;
569 	struct in_ifaddr *ia;
570 	struct sockaddr_dl *sdl;
571 	struct sockaddr sa;
572 	struct in_addr isaddr, itaddr, myaddr;
573 	int op, rif_len;
574 	int req_len;
575 
576 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
577 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
578 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
579 		return;
580 	}
581 
582 	ah = mtod(m, struct arphdr *);
583 	op = ntohs(ah->ar_op);
584 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
585 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
586 #ifdef BRIDGE
587 #define BRIDGE_TEST (do_bridge)
588 #else
589 #define BRIDGE_TEST (0) /* cc will optimise the test away */
590 #endif
591 	/*
592 	 * For a bridge, we want to check the address irrespective
593 	 * of the receive interface. (This will change slightly
594 	 * when we have clusters of interfaces).
595 	 */
596 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash)
597 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
598 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
599 			goto match;
600 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
601 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
602 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
603 			goto match;
604 	/*
605 	 * No match, use the first inet address on the receive interface
606 	 * as a dummy address for the rest of the function.
607 	 */
608 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
609 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
610 			ia = ifatoia(ifa);
611 			goto match;
612 		}
613 	/*
614 	 * If bridging, fall back to using any inet address.
615 	 * This is probably incorrect, the right way being try to match
616 	 * addresses for interfaces in the same cluster, so if we
617 	 * get here we should always drop the packet.
618 	 */
619 	if (!BRIDGE_TEST ||
620 	    (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) {
621 		m_freem(m);
622 		return;
623 	}
624 match:
625 	myaddr = ia->ia_addr.sin_addr;
626 	if (!bcmp(ar_sha(ah), IF_LLADDR(ifp), ifp->if_addrlen)) {
627 		m_freem(m);	/* it's from me, ignore it. */
628 		return;
629 	}
630 	if (!bcmp(ar_sha(ah), IF_BCASTADDR(ifp), ifp->if_addrlen)) {
631 		log(LOG_ERR,
632 		    "arp: link address is broadcast for IP address %s!\n",
633 		    inet_ntoa(isaddr));
634 		m_freem(m);
635 		return;
636 	}
637 	if (isaddr.s_addr == myaddr.s_addr) {
638 		log(LOG_ERR,
639 		   "arp: %*D is using my IP address %s!\n",
640 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
641 		   inet_ntoa(isaddr));
642 		itaddr = myaddr;
643 		goto reply;
644 	}
645 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
646 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
647 		/* the following is not an error when doing bridging */
648 		if (!BRIDGE_TEST && rt->rt_ifp != ifp) {
649 		    if (log_arp_wrong_iface)
650 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
651 			    inet_ntoa(isaddr),
652 			    rt->rt_ifp->if_xname,
653 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
654 			    ifp->if_xname);
655 		    goto reply;
656 		}
657 		if (sdl->sdl_alen &&
658 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
659 			if (rt->rt_expire)
660 			    log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
661 				inet_ntoa(isaddr),
662 				ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
663 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
664 				ifp->if_xname);
665 			else {
666 			    log(LOG_ERR,
667 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
668 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
669 				inet_ntoa(isaddr), ifp->if_xname);
670 			    goto reply;
671 			}
672 		}
673 		/*
674 		 * sanity check for the address length.
675 		 * XXX this does not work for protocols with variable address
676 		 * length. -is
677 		 */
678 		if (sdl->sdl_alen &&
679 		    sdl->sdl_alen != ah->ar_hln) {
680 			log(LOG_WARNING,
681 			    "arp from %*D: new addr len %d, was %d",
682 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
683 			    ah->ar_hln, sdl->sdl_alen);
684 		}
685 		if (ifp->if_addrlen != ah->ar_hln) {
686 			log(LOG_WARNING,
687 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
688 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
689 			    ah->ar_hln, ifp->if_addrlen);
690 			goto reply;
691 		}
692 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
693 		    sdl->sdl_alen = ah->ar_hln);
694 		/*
695 		 * If we receive an arp from a token-ring station over
696 		 * a token-ring nic then try to save the source
697 		 * routing info.
698 		 */
699 		if (ifp->if_type == IFT_ISO88025) {
700 			th = (struct iso88025_header *)m->m_pkthdr.header;
701 			trld = SDL_ISO88025(sdl);
702 			rif_len = TR_RCF_RIFLEN(th->rcf);
703 			if ((th->iso88025_shost[0] & TR_RII) &&
704 			    (rif_len > 2)) {
705 				trld->trld_rcf = th->rcf;
706 				trld->trld_rcf ^= htons(TR_RCF_DIR);
707 				memcpy(trld->trld_route, th->rd, rif_len - 2);
708 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
709 				/*
710 				 * Set up source routing information for
711 				 * reply packet (XXX)
712 				 */
713 				m->m_data -= rif_len;
714 				m->m_len  += rif_len;
715 				m->m_pkthdr.len += rif_len;
716 			} else {
717 				th->iso88025_shost[0] &= ~TR_RII;
718 				trld->trld_rcf = 0;
719 			}
720 			m->m_data -= 8;
721 			m->m_len  += 8;
722 			m->m_pkthdr.len += 8;
723 			th->rcf = trld->trld_rcf;
724 		}
725 		if (rt->rt_expire)
726 			rt->rt_expire = time_second + arpt_keep;
727 		rt->rt_flags &= ~RTF_REJECT;
728 		la->la_preempt = la->la_asked = 0;
729 		if (la->la_hold) {
730 			(*ifp->if_output)(ifp, la->la_hold,
731 				rt_key(rt), rt);
732 			la->la_hold = 0;
733 		}
734 	}
735 reply:
736 	if (op != ARPOP_REQUEST) {
737 		m_freem(m);
738 		return;
739 	}
740 	if (itaddr.s_addr == myaddr.s_addr) {
741 		/* I am the target */
742 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
743 		(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
744 	} else {
745 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
746 		if (la == NULL) {
747 			struct sockaddr_in sin;
748 
749 			if (!arp_proxyall) {
750 				m_freem(m);
751 				return;
752 			}
753 
754 			bzero(&sin, sizeof sin);
755 			sin.sin_family = AF_INET;
756 			sin.sin_len = sizeof sin;
757 			sin.sin_addr = itaddr;
758 
759 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
760 			if (!rt) {
761 				m_freem(m);
762 				return;
763 			}
764 			/*
765 			 * Don't send proxies for nodes on the same interface
766 			 * as this one came out of, or we'll get into a fight
767 			 * over who claims what Ether address.
768 			 */
769 			if (rt->rt_ifp == ifp) {
770 				rtfree(rt);
771 				m_freem(m);
772 				return;
773 			}
774 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
775 			(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
776 			rtfree(rt);
777 #ifdef DEBUG_PROXY
778 			printf("arp: proxying for %s\n",
779 			       inet_ntoa(itaddr));
780 #endif
781 		} else {
782 			rt = la->la_rt;
783 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
784 			sdl = SDL(rt->rt_gateway);
785 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
786 		}
787 	}
788 
789 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
790 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
791 	ah->ar_op = htons(ARPOP_REPLY);
792 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
793 	switch (ifp->if_type) {
794 	case IFT_ARCNET:
795 		arh = (struct arc_header *)sa.sa_data;
796 		arh->arc_dhost = *ar_tha(ah);
797 		arh->arc_type = ARCTYPE_ARP;
798 		break;
799 
800 	case IFT_ISO88025:
801 		/* Re-arrange the source/dest address */
802 		memcpy(th->iso88025_dhost, th->iso88025_shost,
803 		    sizeof(th->iso88025_dhost));
804 		memcpy(th->iso88025_shost, IF_LLADDR(ifp),
805 		    sizeof(th->iso88025_shost));
806 		/* Set the source routing bit if neccesary */
807 		if (th->iso88025_dhost[0] & TR_RII) {
808 			th->iso88025_dhost[0] &= ~TR_RII;
809 			if (TR_RCF_RIFLEN(th->rcf) > 2)
810 				th->iso88025_shost[0] |= TR_RII;
811 		}
812 		/* Copy the addresses, ac and fc into sa_data */
813 		memcpy(sa.sa_data, th->iso88025_dhost,
814 		    sizeof(th->iso88025_dhost) * 2);
815 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2)] = TR_AC;
816 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2) + 1] = TR_LLC_FRAME;
817 		break;
818 	case IFT_ETHER:
819 	case IFT_FDDI:
820 	/*
821 	 * May not be correct for types not explictly
822 	 * listed, but it is our best guess.
823 	 */
824 	default:
825 		eh = (struct ether_header *)sa.sa_data;
826 		(void)memcpy(eh->ether_dhost, ar_tha(ah),
827 		    sizeof(eh->ether_dhost));
828 		eh->ether_type = htons(ETHERTYPE_ARP);
829 		break;
830 	}
831 	sa.sa_family = AF_UNSPEC;
832 	sa.sa_len = sizeof(sa);
833 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
834 	return;
835 }
836 #endif
837 
838 /*
839  * Free an arp entry.
840  */
841 static void
842 arptfree(la)
843 	struct llinfo_arp *la;
844 {
845 	struct rtentry *rt = la->la_rt;
846 	struct sockaddr_dl *sdl;
847 	if (rt == 0)
848 		panic("arptfree");
849 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
850 	    sdl->sdl_family == AF_LINK) {
851 		sdl->sdl_alen = 0;
852 		la->la_preempt = la->la_asked = 0;
853 		rt->rt_flags &= ~RTF_REJECT;
854 		return;
855 	}
856 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
857 			0, (struct rtentry **)0);
858 }
859 /*
860  * Lookup or enter a new address in arptab.
861  */
862 static struct llinfo_arp *
863 arplookup(addr, create, proxy)
864 	u_long addr;
865 	int create, proxy;
866 {
867 	struct rtentry *rt;
868 	static struct sockaddr_inarp sin = {sizeof(sin), AF_INET };
869 	const char *why = 0;
870 
871 	sin.sin_addr.s_addr = addr;
872 	sin.sin_other = proxy ? SIN_PROXY : 0;
873 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
874 	if (rt == 0)
875 		return (0);
876 	rt->rt_refcnt--;
877 
878 	if (rt->rt_flags & RTF_GATEWAY)
879 		why = "host is not on local network";
880 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
881 		why = "could not allocate llinfo";
882 	else if (rt->rt_gateway->sa_family != AF_LINK)
883 		why = "gateway route is not ours";
884 
885 	if (why) {
886 		if (create) {
887 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
888 			    inet_ntoa(sin.sin_addr), why);
889 		}
890 
891 		/* if there are no references to this route, purge it */
892 		if (rt->rt_refcnt <= 0 &&
893 		    (rt->rt_flags & RTF_WASCLONED) == RTF_WASCLONED) {
894 			    rtrequest(RTM_DELETE,
895 				(struct sockaddr *)rt_key(rt), rt->rt_gateway,
896 				rt_mask(rt), rt->rt_flags, 0);
897 		}
898 		return (0);
899 	}
900 	return ((struct llinfo_arp *)rt->rt_llinfo);
901 }
902 
903 void
904 arp_ifinit(ifp, ifa)
905 	struct ifnet *ifp;
906 	struct ifaddr *ifa;
907 {
908 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
909 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
910 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
911 	ifa->ifa_rtrequest = arp_rtrequest;
912 	ifa->ifa_flags |= RTF_CLONING;
913 }
914 
915 static void
916 arp_init(void)
917 {
918 	LIST_INIT(&llinfo_arp);
919 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr);
920 }
921 
922 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
923