xref: /dragonfly/sys/netinet/if_ether.c (revision 6bd457ed)
1 /*
2  * Copyright (c) 2004, 2005 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Jeffrey M. Hsu.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of The DragonFly Project nor the names of its
16  *    contributors may be used to endorse or promote products derived
17  *    from this software without specific, prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 2004, 2005 Jeffrey M. Hsu.  All rights reserved.
35  *
36  * License terms: all terms for the DragonFly license above plus the following:
37  *
38  * 4. All advertising materials mentioning features or use of this software
39  *    must display the following acknowledgement:
40  *
41  *	This product includes software developed by Jeffrey M. Hsu
42  *	for the DragonFly Project.
43  *
44  *    This requirement may be waived with permission from Jeffrey Hsu.
45  *    Permission will be granted to any DragonFly user for free.
46  *    This requirement will sunset and may be removed on Jan 31, 2006,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1982, 1986, 1988, 1993
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
84  * $FreeBSD: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $
85  * $DragonFly: src/sys/netinet/if_ether.c,v 1.27 2005/06/02 23:52:42 dillon Exp $
86  */
87 
88 /*
89  * Ethernet address resolution protocol.
90  * TODO:
91  *	add "inuse/lock" bit (or ref. count) along with valid bit
92  */
93 
94 #include "opt_inet.h"
95 #include "opt_bdg.h"
96 
97 #include <sys/param.h>
98 #include <sys/kernel.h>
99 #include <sys/queue.h>
100 #include <sys/sysctl.h>
101 #include <sys/systm.h>
102 #include <sys/mbuf.h>
103 #include <sys/malloc.h>
104 #include <sys/socket.h>
105 #include <sys/syslog.h>
106 
107 #include <sys/thread2.h>
108 #include <sys/msgport2.h>
109 
110 #include <net/if.h>
111 #include <net/if_dl.h>
112 #include <net/if_types.h>
113 #include <net/route.h>
114 #include <net/netisr.h>
115 #include <net/if_llc.h>
116 #ifdef BRIDGE
117 #include <net/ethernet.h>
118 #include <net/bridge/bridge.h>
119 #endif
120 
121 #include <netinet/in.h>
122 #include <netinet/in_var.h>
123 #include <netinet/if_ether.h>
124 
125 #include <net/if_arc.h>
126 #include <net/iso88025.h>
127 
128 #define SIN(s) ((struct sockaddr_in *)s)
129 #define SDL(s) ((struct sockaddr_dl *)s)
130 
131 SYSCTL_DECL(_net_link_ether);
132 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
133 
134 /* timer values */
135 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
136 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
137 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
138 
139 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
140 	   &arpt_prune, 0, "");
141 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
142 	   &arpt_keep, 0, "");
143 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
144 	   &arpt_down, 0, "");
145 
146 #define	rt_expire	rt_rmx.rmx_expire
147 
148 struct llinfo_arp {
149 	LIST_ENTRY(llinfo_arp) la_le;
150 	struct	rtentry *la_rt;
151 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
152 	u_short	la_preempt;	/* countdown for pre-expiry arps */
153 	u_short	la_asked;	/* #times we QUERIED following expiration */
154 };
155 
156 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
157 
158 static int	arp_maxtries = 5;
159 static int	useloopback = 1; /* use loopback interface for local traffic */
160 static int	arp_proxyall = 0;
161 
162 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
163 	   &arp_maxtries, 0, "");
164 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
165 	   &useloopback, 0, "");
166 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
167 	   &arp_proxyall, 0, "");
168 
169 static void	arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
170 static void	arprequest (struct ifnet *,
171 			struct in_addr *, struct in_addr *, u_char *);
172 static int	arpintr(struct netmsg *);
173 static void	arptfree (struct llinfo_arp *);
174 static void	arptimer (void *);
175 static struct llinfo_arp
176 		*arplookup (in_addr_t addr, boolean_t create, boolean_t proxy);
177 #ifdef INET
178 static void	in_arpinput (struct mbuf *);
179 #endif
180 
181 static struct callout	arptimer_ch;
182 
183 /*
184  * Timeout routine.  Age arp_tab entries periodically.
185  */
186 /* ARGSUSED */
187 static void
188 arptimer(void *ignored_arg)
189 {
190 	struct llinfo_arp *la, *nla;
191 
192 	crit_enter();
193 	LIST_FOREACH_MUTABLE(la, &llinfo_arp, la_le, nla) {
194 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
195 			arptfree(la);	/* might remove la from llinfo_arp! */
196 	}
197 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
198 	crit_exit();
199 }
200 
201 /*
202  * Parallel to llc_rtrequest.
203  */
204 static void
205 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
206 {
207 	struct sockaddr *gate = rt->rt_gateway;
208 	struct llinfo_arp *la = rt->rt_llinfo;
209 
210 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
211 	static boolean_t arpinit_done;
212 
213 	if (!arpinit_done) {
214 		arpinit_done = TRUE;
215 		callout_init(&arptimer_ch);
216 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
217 	}
218 	if (rt->rt_flags & RTF_GATEWAY)
219 		return;
220 
221 	switch (req) {
222 	case RTM_ADD:
223 		/*
224 		 * XXX: If this is a manually added route to interface
225 		 * such as older version of routed or gated might provide,
226 		 * restore cloning bit.
227 		 */
228 		if (!(rt->rt_flags & RTF_HOST) &&
229 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
230 			rt->rt_flags |= RTF_CLONING;
231 		if (rt->rt_flags & RTF_CLONING) {
232 			/*
233 			 * Case 1: This route should come from a route to iface.
234 			 */
235 			rt_setgate(rt, rt_key(rt),
236 				   (struct sockaddr *)&null_sdl);
237 			gate = rt->rt_gateway;
238 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
239 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
240 			rt->rt_expire = time_second;
241 			break;
242 		}
243 		/* Announce a new entry if requested. */
244 		if (rt->rt_flags & RTF_ANNOUNCE)
245 			arprequest(rt->rt_ifp,
246 			    &SIN(rt_key(rt))->sin_addr,
247 			    &SIN(rt_key(rt))->sin_addr,
248 			    LLADDR(SDL(gate)));
249 		/*FALLTHROUGH*/
250 	case RTM_RESOLVE:
251 		if (gate->sa_family != AF_LINK ||
252 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
253 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
254 			break;
255 		}
256 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
257 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
258 		if (la != NULL)
259 			break; /* This happens on a route change */
260 		/*
261 		 * Case 2:  This route may come from cloning, or a manual route
262 		 * add with a LL address.
263 		 */
264 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
265 		rt->rt_llinfo = la;
266 		if (la == NULL) {
267 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
268 			break;
269 		}
270 		bzero(la, sizeof *la);
271 		la->la_rt = rt;
272 		rt->rt_flags |= RTF_LLINFO;
273 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
274 
275 #ifdef INET
276 		/*
277 		 * This keeps the multicast addresses from showing up
278 		 * in `arp -a' listings as unresolved.  It's not actually
279 		 * functional.  Then the same for broadcast.
280 		 */
281 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
282 		    rt->rt_ifp->if_type != IFT_ARCNET) {
283 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
284 					       LLADDR(SDL(gate)));
285 			SDL(gate)->sdl_alen = 6;
286 			rt->rt_expire = 0;
287 		}
288 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
289 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
290 			       rt->rt_ifp->if_addrlen);
291 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
292 			rt->rt_expire = 0;
293 		}
294 #endif
295 
296 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
297 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
298 			/*
299 			 * This test used to be
300 			 *	if (loif.if_flags & IFF_UP)
301 			 * It allowed local traffic to be forced
302 			 * through the hardware by configuring the
303 			 * loopback down.  However, it causes problems
304 			 * during network configuration for boards
305 			 * that can't receive packets they send.  It
306 			 * is now necessary to clear "useloopback" and
307 			 * remove the route to force traffic out to
308 			 * the hardware.
309 			 */
310 			rt->rt_expire = 0;
311 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
312 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
313 			if (useloopback)
314 				rt->rt_ifp = loif;
315 		}
316 		break;
317 
318 	case RTM_DELETE:
319 		if (la == NULL)
320 			break;
321 		LIST_REMOVE(la, la_le);
322 		rt->rt_llinfo = NULL;
323 		rt->rt_flags &= ~RTF_LLINFO;
324 		if (la->la_hold != NULL)
325 			m_freem(la->la_hold);
326 		Free(la);
327 	}
328 }
329 
330 /*
331  * Broadcast an ARP request. Caller specifies:
332  *	- arp header source ip address
333  *	- arp header target ip address
334  *	- arp header source ethernet address
335  */
336 static void
337 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
338 	   u_char *enaddr)
339 {
340 	struct mbuf *m;
341 	struct ether_header *eh;
342 	struct arc_header *arh;
343 	struct arphdr *ah;
344 	struct sockaddr sa;
345 	static u_char llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
346 				 LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
347 	u_short ar_hrd;
348 
349 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
350 		return;
351 	m->m_pkthdr.rcvif = (struct ifnet *)NULL;
352 
353 	switch (ifp->if_type) {
354 	case IFT_ARCNET:
355 		ar_hrd = htons(ARPHRD_ARCNET);
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 		arh = (struct arc_header *)sa.sa_data;
362 		arh->arc_dhost = ifp->if_broadcastaddr[0];
363 		arh->arc_type = ARCTYPE_ARP;
364 
365 		ah = mtod(m, struct arphdr *);
366 		break;
367 
368 	case IFT_ISO88025:
369 		ar_hrd = htons(ARPHRD_IEEE802);
370 
371 		m->m_len = (sizeof llcx) +
372 		    arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
373 		m->m_pkthdr.len = m->m_len;
374 		MH_ALIGN(m, m->m_len);
375 
376 		memcpy(mtod(m, caddr_t), llcx, sizeof llcx);
377 		memcpy(sa.sa_data, ifp->if_broadcastaddr, ifp->if_addrlen);
378 		memcpy(sa.sa_data + 6, enaddr, 6);
379 		sa.sa_data[6] |= TR_RII;
380 		sa.sa_data[12] = TR_AC;
381 		sa.sa_data[13] = TR_LLC_FRAME;
382 
383 		ah = (struct arphdr *)(mtod(m, char *) + sizeof llcx);
384 		break;
385 	case IFT_FDDI:
386 	case IFT_ETHER:
387 		/*
388 		 * This may not be correct for types not explicitly
389 		 * listed, but this is our best guess
390 		 */
391 	default:
392 		ar_hrd = htons(ARPHRD_ETHER);
393 
394 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
395 		m->m_pkthdr.len = m->m_len;
396 		MH_ALIGN(m, m->m_len);
397 
398 		eh = (struct ether_header *)sa.sa_data;
399 		/* if_output() will not swap */
400 		eh->ether_type = htons(ETHERTYPE_ARP);
401 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
402 
403 		ah = mtod(m, struct arphdr *);
404 		break;
405 	}
406 
407 	ah->ar_hrd = ar_hrd;
408 	ah->ar_pro = htons(ETHERTYPE_IP);
409 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
410 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
411 	ah->ar_op = htons(ARPOP_REQUEST);
412 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
413 	memset(ar_tha(ah), 0, ah->ar_hln);
414 	memcpy(ar_spa(ah), sip, ah->ar_pln);
415 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
416 
417 	sa.sa_family = AF_UNSPEC;
418 	sa.sa_len = sizeof sa;
419 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)NULL);
420 }
421 
422 /*
423  * Resolve an IP address into an ethernet address.  If success,
424  * desten is filled in.  If there is no entry in arptab,
425  * set one up and broadcast a request for the IP address.
426  * Hold onto this mbuf and resend it once the address
427  * is finally resolved.  A return value of 1 indicates
428  * that desten has been filled in and the packet should be sent
429  * normally; a 0 return indicates that the packet has been
430  * taken over here, either now or for later transmission.
431  */
432 int
433 arpresolve(
434 	struct ifnet *ifp,
435 	struct rtentry *rt0,
436 	struct mbuf *m,
437 	struct sockaddr *dst,
438 	u_char *desten)
439 {
440 	struct rtentry *rt;
441 	struct llinfo_arp *la = NULL;
442 	struct sockaddr_dl *sdl;
443 
444 	if (m->m_flags & M_BCAST) {	/* broadcast */
445 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
446 		return (1);
447 	}
448 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
449 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
450 		return (1);
451 	}
452 	if (rt0 != NULL) {
453 		if (rt_llroute(dst, rt0, &rt) != 0) {
454 			m_freem(m);
455 			return 0;
456 		}
457 		la = rt->rt_llinfo;
458 	}
459 	if (la == NULL) {
460 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
461 		if (la != NULL)
462 			rt = la->la_rt;
463 	}
464 	if (la == NULL || rt == NULL) {
465 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
466 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
467 		    rt ? "rt" : "");
468 		m_freem(m);
469 		return (0);
470 	}
471 	sdl = SDL(rt->rt_gateway);
472 	/*
473 	 * Check the address family and length is valid, the address
474 	 * is resolved; otherwise, try to resolve.
475 	 */
476 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
477 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
478 		/*
479 		 * If entry has an expiry time and it is approaching,
480 		 * see if we need to send an ARP request within this
481 		 * arpt_down interval.
482 		 */
483 		if ((rt->rt_expire != 0) &&
484 		    (time_second + la->la_preempt > rt->rt_expire)) {
485 			arprequest(ifp,
486 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
487 				   &SIN(dst)->sin_addr,
488 				   IF_LLADDR(ifp));
489 			la->la_preempt--;
490 		}
491 
492 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
493 		return 1;
494 	}
495 	/*
496 	 * If ARP is disabled on this interface, stop.
497 	 * XXX
498 	 * Probably should not allocate empty llinfo struct if we are
499 	 * not going to be sending out an arp request.
500 	 */
501 	if (ifp->if_flags & IFF_NOARP) {
502 		m_freem(m);
503 		return (0);
504 	}
505 	/*
506 	 * There is an arptab entry, but no ethernet address
507 	 * response yet.  Replace the held mbuf with this
508 	 * latest one.
509 	 */
510 	if (la->la_hold != NULL)
511 		m_freem(la->la_hold);
512 	la->la_hold = m;
513 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
514 		rt->rt_flags &= ~RTF_REJECT;
515 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
516 			rt->rt_expire = time_second;
517 			if (la->la_asked++ < arp_maxtries) {
518 				arprequest(ifp,
519 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
520 					   &SIN(dst)->sin_addr,
521 					   IF_LLADDR(ifp));
522 			} else {
523 				rt->rt_flags |= RTF_REJECT;
524 				rt->rt_expire += arpt_down;
525 				la->la_asked = 0;
526 				la->la_preempt = arp_maxtries;
527 			}
528 
529 		}
530 	}
531 	return (0);
532 }
533 
534 /*
535  * Common length and type checks are done here,
536  * then the protocol-specific routine is called.
537  */
538 static int
539 arpintr(struct netmsg *msg)
540 {
541 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
542 	struct arphdr *ar;
543 	u_short ar_hrd;
544 
545 	if (m->m_len < sizeof(struct arphdr) &&
546 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
547 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
548 		goto out2;
549 	}
550 	ar = mtod(m, struct arphdr *);
551 
552 	ar_hrd = ntohs(ar->ar_hrd);
553 	if (ar_hrd != ARPHRD_ETHER &&
554 	    ar_hrd != ARPHRD_IEEE802 &&
555 	    ar_hrd != ARPHRD_ARCNET) {
556 		log(LOG_ERR,
557 		    "arp: unknown hardware address format (0x%2D)\n",
558 		    (unsigned char *)&ar->ar_hrd, "");
559 		goto out1;
560 	}
561 
562 	if (m->m_pkthdr.len < arphdr_len(ar) &&
563 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
564 		log(LOG_ERR, "arp: runt packet\n");
565 		goto out1;
566 	}
567 
568 	switch (ntohs(ar->ar_pro)) {
569 #ifdef INET
570 		case ETHERTYPE_IP:
571 			in_arpinput(m);
572 			goto out2;
573 #endif
574 	}
575 out1:
576 	m_freem(m);
577 out2:
578 	lwkt_replymsg(&msg->nm_lmsg, 0);
579 	return(EASYNC);
580 }
581 
582 #ifdef INET
583 /*
584  * ARP for Internet protocols on 10 Mb/s Ethernet.
585  * Algorithm is that given in RFC 826.
586  * In addition, a sanity check is performed on the sender
587  * protocol address, to catch impersonators.
588  * We no longer handle negotiations for use of trailer protocol:
589  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
590  * along with IP replies if we wanted trailers sent to us,
591  * and also sent them in response to IP replies.
592  * This allowed either end to announce the desire to receive
593  * trailer packets.
594  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
595  * but formerly didn't normally send requests.
596  */
597 static int log_arp_wrong_iface = 1;
598 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
599 	&log_arp_wrong_iface, 0,
600 	"log arp packets arriving on the wrong interface");
601 
602 static void
603 in_arpinput(struct mbuf *m)
604 {
605 	struct arphdr *ah;
606 	struct ifnet *ifp = m->m_pkthdr.rcvif;
607 	struct ether_header *eh;
608 	struct arc_header *arh;
609 	struct iso88025_header *th = (struct iso88025_header *)NULL;
610 	struct iso88025_sockaddr_dl_data *trld;
611 	struct llinfo_arp *la = NULL;
612 	struct rtentry *rt;
613 	struct ifaddr *ifa;
614 	struct in_ifaddr *ia;
615 	struct sockaddr_dl *sdl;
616 	struct sockaddr sa;
617 	struct in_addr isaddr, itaddr, myaddr;
618 	int op, rif_len;
619 	int req_len;
620 
621 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
622 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
623 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
624 		return;
625 	}
626 
627 	ah = mtod(m, struct arphdr *);
628 	op = ntohs(ah->ar_op);
629 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
630 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
631 #ifdef BRIDGE
632 #define BRIDGE_TEST (do_bridge)
633 #else
634 #define BRIDGE_TEST (0) /* cc will optimise the test away */
635 #endif
636 	/*
637 	 * For a bridge, we want to check the address irrespective
638 	 * of the receive interface. (This will change slightly
639 	 * when we have clusters of interfaces).
640 	 */
641 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash)
642 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
643 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
644 			goto match;
645 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
646 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
647 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
648 			goto match;
649 	/*
650 	 * No match, use the first inet address on the receive interface
651 	 * as a dummy address for the rest of the function.
652 	 */
653 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
654 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
655 			ia = ifatoia(ifa);
656 			goto match;
657 		}
658 	/*
659 	 * If bridging, fall back to using any inet address.
660 	 * This is probably incorrect, the right way being try to match
661 	 * addresses for interfaces in the same cluster, so if we
662 	 * get here we should always drop the packet.
663 	 */
664 	if (!BRIDGE_TEST ||
665 	    (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) {
666 		m_freem(m);
667 		return;
668 	}
669 match:
670 	myaddr = ia->ia_addr.sin_addr;
671 	if (!bcmp(ar_sha(ah), IF_LLADDR(ifp), ifp->if_addrlen)) {
672 		m_freem(m);	/* it's from me, ignore it. */
673 		return;
674 	}
675 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
676 		log(LOG_ERR,
677 		    "arp: link address is broadcast for IP address %s!\n",
678 		    inet_ntoa(isaddr));
679 		m_freem(m);
680 		return;
681 	}
682 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
683 		log(LOG_ERR,
684 		   "arp: %*D is using my IP address %s!\n",
685 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
686 		   inet_ntoa(isaddr));
687 		itaddr = myaddr;
688 		goto reply;
689 	}
690 	la = arplookup(isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr), FALSE);
691 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
692 		/* the following is not an error when doing bridging */
693 		if (!BRIDGE_TEST && rt->rt_ifp != ifp) {
694 		    if (log_arp_wrong_iface)
695 			log(LOG_ERR,
696 			    "arp: %s is on %s but got reply from %*D on %s\n",
697 			    inet_ntoa(isaddr),
698 			    rt->rt_ifp->if_xname,
699 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
700 			    ifp->if_xname);
701 		    goto reply;
702 		}
703 		if (sdl->sdl_alen &&
704 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
705 			if (rt->rt_expire)
706 			    log(LOG_INFO,
707 				"arp: %s moved from %*D to %*D on %s\n",
708 				inet_ntoa(isaddr),
709 				ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
710 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
711 				ifp->if_xname);
712 			else {
713 			    log(LOG_ERR,
714 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
715 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
716 				inet_ntoa(isaddr), ifp->if_xname);
717 			    goto reply;
718 			}
719 		}
720 		/*
721 		 * sanity check for the address length.
722 		 * XXX this does not work for protocols with variable address
723 		 * length. -is
724 		 */
725 		if (sdl->sdl_alen &&
726 		    sdl->sdl_alen != ah->ar_hln) {
727 			log(LOG_WARNING,
728 			    "arp from %*D: new addr len %d, was %d",
729 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
730 			    ah->ar_hln, sdl->sdl_alen);
731 		}
732 		if (ifp->if_addrlen != ah->ar_hln) {
733 			log(LOG_WARNING,
734 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
735 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
736 			    ah->ar_hln, ifp->if_addrlen);
737 			goto reply;
738 		}
739 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
740 		/*
741 		 * If we receive an arp from a token-ring station over
742 		 * a token-ring nic then try to save the source
743 		 * routing info.
744 		 */
745 		if (ifp->if_type == IFT_ISO88025) {
746 			th = (struct iso88025_header *)m->m_pkthdr.header;
747 			trld = SDL_ISO88025(sdl);
748 			rif_len = TR_RCF_RIFLEN(th->rcf);
749 			if ((th->iso88025_shost[0] & TR_RII) &&
750 			    (rif_len > 2)) {
751 				trld->trld_rcf = th->rcf;
752 				trld->trld_rcf ^= htons(TR_RCF_DIR);
753 				memcpy(trld->trld_route, th->rd, rif_len - 2);
754 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
755 				/*
756 				 * Set up source routing information for
757 				 * reply packet (XXX)
758 				 */
759 				m->m_data -= rif_len;
760 				m->m_len  += rif_len;
761 				m->m_pkthdr.len += rif_len;
762 			} else {
763 				th->iso88025_shost[0] &= ~TR_RII;
764 				trld->trld_rcf = 0;
765 			}
766 			m->m_data -= 8;
767 			m->m_len  += 8;
768 			m->m_pkthdr.len += 8;
769 			th->rcf = trld->trld_rcf;
770 		}
771 		if (rt->rt_expire)
772 			rt->rt_expire = time_second + arpt_keep;
773 		rt->rt_flags &= ~RTF_REJECT;
774 		la->la_asked = 0;
775 		la->la_preempt = arp_maxtries;
776 		if (la->la_hold != NULL) {
777 			m_adj(la->la_hold, sizeof(struct ether_header));
778 			(*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt);
779 			la->la_hold = NULL;
780 		}
781 	}
782 reply:
783 	if (op != ARPOP_REQUEST) {
784 		m_freem(m);
785 		return;
786 	}
787 	if (itaddr.s_addr == myaddr.s_addr) {
788 		/* I am the target */
789 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
790 		memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
791 	} else {
792 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
793 		if (la == NULL) {
794 			struct sockaddr_in sin;
795 
796 			if (!arp_proxyall) {
797 				m_freem(m);
798 				return;
799 			}
800 
801 			bzero(&sin, sizeof sin);
802 			sin.sin_family = AF_INET;
803 			sin.sin_len = sizeof sin;
804 			sin.sin_addr = itaddr;
805 
806 			rt = rtpurelookup((struct sockaddr *)&sin);
807 			if (rt == NULL) {
808 				m_freem(m);
809 				return;
810 			}
811 			--rt->rt_refcnt;
812 			/*
813 			 * Don't send proxies for nodes on the same interface
814 			 * as this one came out of, or we'll get into a fight
815 			 * over who claims what Ether address.
816 			 */
817 			if (rt->rt_ifp == ifp) {
818 				m_freem(m);
819 				return;
820 			}
821 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
822 			memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
823 #ifdef DEBUG_PROXY
824 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
825 #endif
826 		} else {
827 			rt = la->la_rt;
828 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
829 			sdl = SDL(rt->rt_gateway);
830 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
831 		}
832 	}
833 
834 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
835 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
836 	ah->ar_op = htons(ARPOP_REPLY);
837 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
838 	switch (ifp->if_type) {
839 	case IFT_ARCNET:
840 		arh = (struct arc_header *)sa.sa_data;
841 		arh->arc_dhost = *ar_tha(ah);
842 		arh->arc_type = ARCTYPE_ARP;
843 		break;
844 	case IFT_ISO88025:
845 		/* Re-arrange the source/dest address */
846 		memcpy(th->iso88025_dhost, th->iso88025_shost,
847 		    sizeof th->iso88025_dhost);
848 		memcpy(th->iso88025_shost, IF_LLADDR(ifp),
849 		    sizeof th->iso88025_shost);
850 		/* Set the source routing bit if neccesary */
851 		if (th->iso88025_dhost[0] & TR_RII) {
852 			th->iso88025_dhost[0] &= ~TR_RII;
853 			if (TR_RCF_RIFLEN(th->rcf) > 2)
854 				th->iso88025_shost[0] |= TR_RII;
855 		}
856 		/* Copy the addresses, ac and fc into sa_data */
857 		memcpy(sa.sa_data, th->iso88025_dhost,
858 		    (sizeof th->iso88025_dhost) * 2);
859 		sa.sa_data[(sizeof th->iso88025_dhost) * 2] = TR_AC;
860 		sa.sa_data[(sizeof th->iso88025_dhost) * 2 + 1] = TR_LLC_FRAME;
861 		break;
862 	case IFT_ETHER:
863 	case IFT_FDDI:
864 	/*
865 	 * May not be correct for types not explictly
866 	 * listed, but it is our best guess.
867 	 */
868 	default:
869 		eh = (struct ether_header *)sa.sa_data;
870 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
871 		eh->ether_type = htons(ETHERTYPE_ARP);
872 		break;
873 	}
874 	sa.sa_family = AF_UNSPEC;
875 	sa.sa_len = sizeof sa;
876 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
877 	return;
878 }
879 #endif
880 
881 /*
882  * Free an arp entry.  If the arp entry is actively referenced or represents
883  * a static entry we only clear it back to an unresolved state, otherwise
884  * we destroy the entry entirely.
885  *
886  * Note that static entries are created when route add ... -interface is used
887  * to create an interface route to a (direct) destination.
888  */
889 static void
890 arptfree(struct llinfo_arp *la)
891 {
892 	struct rtentry *rt = la->la_rt;
893 	struct sockaddr_dl *sdl;
894 
895 	if (rt == NULL)
896 		panic("arptfree");
897 	sdl = SDL(rt->rt_gateway);
898 	if (sdl != NULL &&
899 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
900 	     (rt->rt_flags & RTF_STATIC))) {
901 		sdl->sdl_alen = 0;
902 		la->la_preempt = la->la_asked = 0;
903 		rt->rt_flags &= ~RTF_REJECT;
904 		return;
905 	}
906 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
907 }
908 
909 /*
910  * Lookup or enter a new address in arptab.
911  */
912 static struct llinfo_arp *
913 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
914 {
915 	struct rtentry *rt;
916 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
917 	const char *why = NULL;
918 
919 	sin.sin_addr.s_addr = addr;
920 	sin.sin_other = proxy ? SIN_PROXY : 0;
921 	if (create)
922 		rt = rtlookup((struct sockaddr *)&sin);
923 	else
924 		rt = rtpurelookup((struct sockaddr *)&sin);
925 	if (rt == NULL)
926 		return (NULL);
927 	rt->rt_refcnt--;
928 
929 	if (rt->rt_flags & RTF_GATEWAY)
930 		why = "host is not on local network";
931 	else if (!(rt->rt_flags & RTF_LLINFO))
932 		why = "could not allocate llinfo";
933 	else if (rt->rt_gateway->sa_family != AF_LINK)
934 		why = "gateway route is not ours";
935 
936 	if (why) {
937 		if (create) {
938 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
939 			    inet_ntoa(sin.sin_addr), why);
940 		}
941 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
942 			/* No references to this route.  Purge it. */
943 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
944 				  rt_mask(rt), rt->rt_flags, NULL);
945 		}
946 		return (NULL);
947 	}
948 	return (rt->rt_llinfo);
949 }
950 
951 void
952 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
953 {
954 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
955 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
956 			   IF_LLADDR(ifp));
957 	ifa->ifa_rtrequest = arp_rtrequest;
958 	ifa->ifa_flags |= RTF_CLONING;
959 }
960 
961 static void
962 arp_init(void)
963 {
964 	LIST_INIT(&llinfo_arp);
965 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr);
966 }
967 
968 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
969