xref: /dflybsd-src/sys/netinet/if_ether.c (revision d797dc5d823699c41734581020450ed2a7edf045)
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) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
66  * $FreeBSD: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $
67  * $DragonFly: src/sys/netinet/if_ether.c,v 1.53 2008/09/27 11:29:47 sephe Exp $
68  */
69 
70 /*
71  * Ethernet address resolution protocol.
72  * TODO:
73  *	add "inuse/lock" bit (or ref. count) along with valid bit
74  */
75 
76 #include "opt_inet.h"
77 #include "opt_carp.h"
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/queue.h>
82 #include <sys/sysctl.h>
83 #include <sys/systm.h>
84 #include <sys/mbuf.h>
85 #include <sys/malloc.h>
86 #include <sys/socket.h>
87 #include <sys/syslog.h>
88 
89 #include <net/if.h>
90 #include <net/if_dl.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 #include <net/netisr.h>
94 #include <net/if_llc.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 #include <netinet/if_ether.h>
99 
100 #include <sys/thread2.h>
101 #include <sys/msgport2.h>
102 #include <net/netmsg2.h>
103 
104 #ifdef CARP
105 #include <netinet/ip_carp.h>
106 #endif
107 
108 #define SIN(s) ((struct sockaddr_in *)s)
109 #define SDL(s) ((struct sockaddr_dl *)s)
110 
111 SYSCTL_DECL(_net_link_ether);
112 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
113 
114 /* timer values */
115 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
116 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
117 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
118 
119 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
120 	   &arpt_prune, 0, "");
121 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
122 	   &arpt_keep, 0, "");
123 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
124 	   &arpt_down, 0, "");
125 
126 #define	rt_expire	rt_rmx.rmx_expire
127 
128 struct llinfo_arp {
129 	LIST_ENTRY(llinfo_arp) la_le;
130 	struct	rtentry *la_rt;
131 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
132 	u_short	la_preempt;	/* countdown for pre-expiry arps */
133 	u_short	la_asked;	/* #times we QUERIED following expiration */
134 };
135 
136 static	LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
137 
138 static int	arp_maxtries = 5;
139 static int	useloopback = 1; /* use loopback interface for local traffic */
140 static int	arp_proxyall = 0;
141 
142 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
143 	   &arp_maxtries, 0, "");
144 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
145 	   &useloopback, 0, "");
146 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
147 	   &arp_proxyall, 0, "");
148 
149 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
150 static void	arprequest(struct ifnet *, struct in_addr *, struct in_addr *,
151 			   u_char *);
152 static void	arpintr(struct netmsg *);
153 static void	arptfree(struct llinfo_arp *);
154 static void	arptimer(void *);
155 static struct llinfo_arp *
156 		arplookup(in_addr_t, boolean_t, boolean_t);
157 #ifdef INET
158 static void	in_arpinput(struct mbuf *);
159 #endif
160 
161 static struct callout	arptimer_ch[MAXCPU];
162 
163 /*
164  * Timeout routine.  Age arp_tab entries periodically.
165  */
166 /* ARGSUSED */
167 static void
168 arptimer(void *ignored_arg)
169 {
170 	struct llinfo_arp *la, *nla;
171 
172 	crit_enter();
173 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
174 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
175 			arptfree(la);
176 	}
177 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
178 	crit_exit();
179 }
180 
181 /*
182  * Parallel to llc_rtrequest.
183  */
184 static void
185 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
186 {
187 	struct sockaddr *gate = rt->rt_gateway;
188 	struct llinfo_arp *la = rt->rt_llinfo;
189 
190 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
191 	static boolean_t arpinit_done[MAXCPU];
192 
193 	if (!arpinit_done[mycpuid]) {
194 		arpinit_done[mycpuid] = TRUE;
195 		callout_init(&arptimer_ch[mycpuid]);
196 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
197 	}
198 	if (rt->rt_flags & RTF_GATEWAY)
199 		return;
200 
201 	switch (req) {
202 	case RTM_ADD:
203 		/*
204 		 * XXX: If this is a manually added route to interface
205 		 * such as older version of routed or gated might provide,
206 		 * restore cloning bit.
207 		 */
208 		if (!(rt->rt_flags & RTF_HOST) &&
209 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
210 			rt->rt_flags |= RTF_CLONING;
211 		if (rt->rt_flags & RTF_CLONING) {
212 			/*
213 			 * Case 1: This route should come from a route to iface.
214 			 */
215 			rt_setgate(rt, rt_key(rt),
216 				   (struct sockaddr *)&null_sdl);
217 			gate = rt->rt_gateway;
218 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
219 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
220 			rt->rt_expire = time_second;
221 			break;
222 		}
223 		/* Announce a new entry if requested. */
224 		if (rt->rt_flags & RTF_ANNOUNCE) {
225 			arprequest(rt->rt_ifp,
226 			    &SIN(rt_key(rt))->sin_addr,
227 			    &SIN(rt_key(rt))->sin_addr,
228 			    LLADDR(SDL(gate)));
229 		}
230 		/*FALLTHROUGH*/
231 	case RTM_RESOLVE:
232 		if (gate->sa_family != AF_LINK ||
233 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
234 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
235 			break;
236 		}
237 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
238 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
239 		if (la != NULL)
240 			break; /* This happens on a route change */
241 		/*
242 		 * Case 2:  This route may come from cloning, or a manual route
243 		 * add with a LL address.
244 		 */
245 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
246 		rt->rt_llinfo = la;
247 		if (la == NULL) {
248 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
249 			break;
250 		}
251 		bzero(la, sizeof *la);
252 		la->la_rt = rt;
253 		rt->rt_flags |= RTF_LLINFO;
254 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
255 
256 #ifdef INET
257 		/*
258 		 * This keeps the multicast addresses from showing up
259 		 * in `arp -a' listings as unresolved.  It's not actually
260 		 * functional.  Then the same for broadcast.
261 		 */
262 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
263 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
264 					       LLADDR(SDL(gate)));
265 			SDL(gate)->sdl_alen = 6;
266 			rt->rt_expire = 0;
267 		}
268 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
269 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
270 			       rt->rt_ifp->if_addrlen);
271 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
272 			rt->rt_expire = 0;
273 		}
274 #endif
275 
276 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
277 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
278 			/*
279 			 * This test used to be
280 			 *	if (loif.if_flags & IFF_UP)
281 			 * It allowed local traffic to be forced
282 			 * through the hardware by configuring the
283 			 * loopback down.  However, it causes problems
284 			 * during network configuration for boards
285 			 * that can't receive packets they send.  It
286 			 * is now necessary to clear "useloopback" and
287 			 * remove the route to force traffic out to
288 			 * the hardware.
289 			 */
290 			rt->rt_expire = 0;
291 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
292 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
293 			if (useloopback)
294 				rt->rt_ifp = loif;
295 		}
296 		break;
297 
298 	case RTM_DELETE:
299 		if (la == NULL)
300 			break;
301 		LIST_REMOVE(la, la_le);
302 		rt->rt_llinfo = NULL;
303 		rt->rt_flags &= ~RTF_LLINFO;
304 		if (la->la_hold != NULL)
305 			m_freem(la->la_hold);
306 		Free(la);
307 		break;
308 	}
309 }
310 
311 /*
312  * Broadcast an ARP request. Caller specifies:
313  *	- arp header source ip address
314  *	- arp header target ip address
315  *	- arp header source ethernet address
316  */
317 static void
318 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
319 	   u_char *enaddr)
320 {
321 	struct mbuf *m;
322 	struct ether_header *eh;
323 	struct arphdr *ah;
324 	struct sockaddr sa;
325 	u_short ar_hrd;
326 
327 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
328 		return;
329 	m->m_pkthdr.rcvif = (struct ifnet *)NULL;
330 
331 	switch (ifp->if_type) {
332 	case IFT_ETHER:
333 		/*
334 		 * This may not be correct for types not explicitly
335 		 * listed, but this is our best guess
336 		 */
337 	default:
338 		ar_hrd = htons(ARPHRD_ETHER);
339 
340 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
341 		m->m_pkthdr.len = m->m_len;
342 		MH_ALIGN(m, m->m_len);
343 
344 		eh = (struct ether_header *)sa.sa_data;
345 		/* if_output() will not swap */
346 		eh->ether_type = htons(ETHERTYPE_ARP);
347 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
348 
349 		ah = mtod(m, struct arphdr *);
350 		break;
351 	}
352 
353 	ah->ar_hrd = ar_hrd;
354 	ah->ar_pro = htons(ETHERTYPE_IP);
355 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
356 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
357 	ah->ar_op = htons(ARPOP_REQUEST);
358 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
359 	memset(ar_tha(ah), 0, ah->ar_hln);
360 	memcpy(ar_spa(ah), sip, ah->ar_pln);
361 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
362 
363 	sa.sa_family = AF_UNSPEC;
364 	sa.sa_len = sizeof sa;
365 	ifp->if_output(ifp, m, &sa, NULL);
366 }
367 
368 /*
369  * Resolve an IP address into an ethernet address.  If success,
370  * desten is filled in.  If there is no entry in arptab,
371  * set one up and broadcast a request for the IP address.
372  * Hold onto this mbuf and resend it once the address
373  * is finally resolved.  A return value of 1 indicates
374  * that desten has been filled in and the packet should be sent
375  * normally; a 0 return indicates that the packet has been
376  * taken over here, either now or for later transmission.
377  */
378 int
379 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
380 	   struct sockaddr *dst, u_char *desten)
381 {
382 	struct rtentry *rt;
383 	struct llinfo_arp *la = NULL;
384 	struct sockaddr_dl *sdl;
385 
386 	if (m->m_flags & M_BCAST) {	/* broadcast */
387 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
388 		return (1);
389 	}
390 	if (m->m_flags & M_MCAST) {/* multicast */
391 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
392 		return (1);
393 	}
394 	if (rt0 != NULL) {
395 		if (rt_llroute(dst, rt0, &rt) != 0) {
396 			m_freem(m);
397 			return 0;
398 		}
399 		la = rt->rt_llinfo;
400 	}
401 	if (la == NULL) {
402 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
403 		if (la != NULL)
404 			rt = la->la_rt;
405 	}
406 	if (la == NULL || rt == NULL) {
407 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
408 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
409 		    rt ? "rt" : "");
410 		m_freem(m);
411 		return (0);
412 	}
413 	sdl = SDL(rt->rt_gateway);
414 	/*
415 	 * Check the address family and length is valid, the address
416 	 * is resolved; otherwise, try to resolve.
417 	 */
418 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
419 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
420 		/*
421 		 * If entry has an expiry time and it is approaching,
422 		 * see if we need to send an ARP request within this
423 		 * arpt_down interval.
424 		 */
425 		if ((rt->rt_expire != 0) &&
426 		    (time_second + la->la_preempt > rt->rt_expire)) {
427 			arprequest(ifp,
428 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
429 				   &SIN(dst)->sin_addr,
430 				   IF_LLADDR(ifp));
431 			la->la_preempt--;
432 		}
433 
434 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
435 		return 1;
436 	}
437 	/*
438 	 * If ARP is disabled on this interface, stop.
439 	 * XXX
440 	 * Probably should not allocate empty llinfo struct if we are
441 	 * not going to be sending out an arp request.
442 	 */
443 	if (ifp->if_flags & IFF_NOARP) {
444 		m_freem(m);
445 		return (0);
446 	}
447 	/*
448 	 * There is an arptab entry, but no ethernet address
449 	 * response yet.  Replace the held mbuf with this
450 	 * latest one.
451 	 */
452 	if (la->la_hold != NULL)
453 		m_freem(la->la_hold);
454 	la->la_hold = m;
455 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
456 		rt->rt_flags &= ~RTF_REJECT;
457 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
458 			rt->rt_expire = time_second;
459 			if (la->la_asked++ < arp_maxtries) {
460 				arprequest(ifp,
461 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
462 					   &SIN(dst)->sin_addr,
463 					   IF_LLADDR(ifp));
464 			} else {
465 				rt->rt_flags |= RTF_REJECT;
466 				rt->rt_expire += arpt_down;
467 				la->la_asked = 0;
468 				la->la_preempt = arp_maxtries;
469 			}
470 		}
471 	}
472 	return (0);
473 }
474 
475 /*
476  * Common length and type checks are done here,
477  * then the protocol-specific routine is called.
478  */
479 static void
480 arpintr(struct netmsg *msg)
481 {
482 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
483 	struct arphdr *ar;
484 	u_short ar_hrd;
485 
486 	if (m->m_len < sizeof(struct arphdr) &&
487 	    (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
488 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
489 		return;
490 	}
491 	ar = mtod(m, struct arphdr *);
492 
493 	ar_hrd = ntohs(ar->ar_hrd);
494 	if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
495 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
496 		    (unsigned char *)&ar->ar_hrd, "");
497 		m_freem(m);
498 		return;
499 	}
500 
501 	if (m->m_pkthdr.len < arphdr_len(ar)) {
502 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
503 			log(LOG_ERR, "arp: runt packet\n");
504 			return;
505 		}
506 		ar = mtod(m, struct arphdr *);
507 	}
508 
509 	switch (ntohs(ar->ar_pro)) {
510 #ifdef INET
511 	case ETHERTYPE_IP:
512 		in_arpinput(m);
513 		return;
514 #endif
515 	}
516 	m_freem(m);
517 	/* msg was embedded in the mbuf, do not reply! */
518 }
519 
520 #ifdef INET
521 /*
522  * ARP for Internet protocols on 10 Mb/s Ethernet.
523  * Algorithm is that given in RFC 826.
524  * In addition, a sanity check is performed on the sender
525  * protocol address, to catch impersonators.
526  * We no longer handle negotiations for use of trailer protocol:
527  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
528  * along with IP replies if we wanted trailers sent to us,
529  * and also sent them in response to IP replies.
530  * This allowed either end to announce the desire to receive
531  * trailer packets.
532  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
533  * but formerly didn't normally send requests.
534  */
535 static int log_arp_wrong_iface = 1;
536 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
537 	   &log_arp_wrong_iface, 0,
538 	   "log arp packets arriving on the wrong interface");
539 
540 static void
541 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
542 		 boolean_t dologging)
543 {
544 	struct arphdr *ah = mtod(m, struct arphdr *);
545 	struct ifnet *ifp = m->m_pkthdr.rcvif;
546 	struct llinfo_arp *la;
547 	struct sockaddr_dl *sdl;
548 	struct rtentry *rt;
549 
550 	la = arplookup(saddr, create, FALSE);
551 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
552 		struct in_addr isaddr = { saddr };
553 
554 		/* the following is not an error when doing bridging */
555 		if (rt->rt_ifp != ifp) {
556 			if (dologging && log_arp_wrong_iface) {
557 				log(LOG_ERR,
558 				    "arp: %s is on %s "
559 				    "but got reply from %*D on %s\n",
560 				    inet_ntoa(isaddr),
561 				    rt->rt_ifp->if_xname,
562 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
563 				    ifp->if_xname);
564 			}
565 			return;
566 		}
567 		if (sdl->sdl_alen &&
568 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
569 			if (rt->rt_expire != 0) {
570 				if (dologging) {
571 			    		log(LOG_INFO,
572 			    		"arp: %s moved from %*D to %*D on %s\n",
573 			    		inet_ntoa(isaddr),
574 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
575 			    		":", ifp->if_addrlen,
576 			    		(u_char *)ar_sha(ah), ":",
577 			    		ifp->if_xname);
578 				}
579 			} else {
580 				if (dologging) {
581 					log(LOG_ERR,
582 					"arp: %*D attempts to modify "
583 					"permanent entry for %s on %s\n",
584 					ifp->if_addrlen, (u_char *)ar_sha(ah),
585 					":", inet_ntoa(isaddr), ifp->if_xname);
586 				}
587 				return;
588 			}
589 		}
590 		/*
591 		 * sanity check for the address length.
592 		 * XXX this does not work for protocols with variable address
593 		 * length. -is
594 		 */
595 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
596 			log(LOG_WARNING,
597 			    "arp from %*D: new addr len %d, was %d",
598 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
599 			    ah->ar_hln, sdl->sdl_alen);
600 		}
601 		if (ifp->if_addrlen != ah->ar_hln) {
602 			if (dologging) {
603 				log(LOG_WARNING,
604 				"arp from %*D: addr len: new %d, i/f %d "
605 				"(ignored)",
606 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
607 				ah->ar_hln, ifp->if_addrlen);
608 			}
609 			return;
610 		}
611 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
612 		if (rt->rt_expire != 0)
613 			rt->rt_expire = time_second + arpt_keep;
614 		rt->rt_flags &= ~RTF_REJECT;
615 		la->la_asked = 0;
616 		la->la_preempt = arp_maxtries;
617 
618 		/*
619 		 * This particular cpu might have been holding an mbuf
620 		 * pending ARP resolution.  If so, transmit the mbuf now.
621 		 */
622 		if (la->la_hold != NULL) {
623 			m_adj(la->la_hold, sizeof(struct ether_header));
624 			ifp->if_output(ifp, la->la_hold, rt_key(rt), rt);
625 			la->la_hold = NULL;
626 		}
627 	}
628 }
629 
630 #ifdef SMP
631 
632 struct netmsg_arp_update {
633 	struct netmsg	netmsg;
634 	struct mbuf	*m;
635 	in_addr_t	saddr;
636 	boolean_t	create;
637 };
638 
639 static void arp_update_msghandler(struct netmsg *);
640 
641 #endif
642 
643 /*
644  * Called from arpintr() - this routine is run from a single cpu.
645  */
646 static void
647 in_arpinput(struct mbuf *m)
648 {
649 	struct arphdr *ah;
650 	struct ifnet *ifp = m->m_pkthdr.rcvif;
651 	struct ether_header *eh;
652 	struct rtentry *rt;
653 	struct ifaddr_container *ifac;
654 	struct in_ifaddr_container *iac;
655 	struct in_ifaddr *ia;
656 	struct sockaddr sa;
657 	struct in_addr isaddr, itaddr, myaddr;
658 #ifdef SMP
659 	struct netmsg_arp_update msg;
660 #endif
661 	u_int8_t *enaddr = NULL;
662 	int op;
663 	int req_len;
664 
665 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
666 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
667 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
668 		return;
669 	}
670 
671 	ah = mtod(m, struct arphdr *);
672 	op = ntohs(ah->ar_op);
673 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
674 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
675 	/*
676 	 * Check both target and sender IP addresses:
677 	 *
678 	 * If we receive the packet on the interface owning the address,
679 	 * then accept the address.
680 	 *
681 	 * For a bridge, we accept the address if the receive interface and
682 	 * the interface owning the address are on the same bridge.
683 	 * (This will change slightly when we have clusters of interfaces).
684 	 */
685 	LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
686 		ia = iac->ia;
687 
688 		/* Skip all ia's which don't match */
689 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
690 			continue;
691 
692 		if (ia->ia_ifp == ifp)
693 			goto match;
694 
695 		if (ifp->if_bridge && ia->ia_ifp &&
696 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
697 			goto match;
698 
699 #ifdef CARP
700 		/*
701 		 * If the interface does not match, but the recieving interface
702 		 * is part of carp, we call carp_iamatch to see if this is a
703 		 * request for the virtual host ip.
704 		 * XXX: This is really ugly!
705 		 */
706 		if (ifp->if_carp != NULL &&
707 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
708 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
709 			goto match;
710 #endif
711 	}
712 	LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
713 		ia = iac->ia;
714 
715 		/* Skip all ia's which don't match */
716 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
717 			continue;
718 
719 		if (ia->ia_ifp == ifp)
720 			goto match;
721 
722 		if (ifp->if_bridge && ia->ia_ifp &&
723 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
724 			goto match;
725 	}
726 	/*
727 	 * No match, use the first inet address on the receive interface
728 	 * as a dummy address for the rest of the function.
729 	 */
730 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
731 		struct ifaddr *ifa = ifac->ifa;
732 
733 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
734 			ia = ifatoia(ifa);
735 			goto match;
736 		}
737 	}
738 	/*
739 	 * If we got here, we didn't find any suitable interface,
740 	 * so drop the packet.
741 	 */
742 	m_freem(m);
743 	return;
744 
745 match:
746 	if (!enaddr)
747 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
748 	myaddr = ia->ia_addr.sin_addr;
749 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
750 		m_freem(m);	/* it's from me, ignore it. */
751 		return;
752 	}
753 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
754 		log(LOG_ERR,
755 		    "arp: link address is broadcast for IP address %s!\n",
756 		    inet_ntoa(isaddr));
757 		m_freem(m);
758 		return;
759 	}
760 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
761 		log(LOG_ERR,
762 		   "arp: %*D is using my IP address %s!\n",
763 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
764 		   inet_ntoa(isaddr));
765 		itaddr = myaddr;
766 		goto reply;
767 	}
768 #ifdef SMP
769 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
770 		    arp_update_msghandler);
771 	msg.m = m;
772 	msg.saddr = isaddr.s_addr;
773 	msg.create = (itaddr.s_addr == myaddr.s_addr);
774 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
775 #else
776 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
777 			 TRUE);
778 #endif
779 reply:
780 	if (op != ARPOP_REQUEST) {
781 		m_freem(m);
782 		return;
783 	}
784 	if (itaddr.s_addr == myaddr.s_addr) {
785 		/* I am the target */
786 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
787 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
788 	} else {
789 		struct llinfo_arp *la;
790 
791 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
792 		if (la == NULL) {
793 			struct sockaddr_in sin;
794 
795 			if (!arp_proxyall) {
796 				m_freem(m);
797 				return;
798 			}
799 
800 			bzero(&sin, sizeof sin);
801 			sin.sin_family = AF_INET;
802 			sin.sin_len = sizeof sin;
803 			sin.sin_addr = itaddr;
804 
805 			rt = rtpurelookup((struct sockaddr *)&sin);
806 			if (rt == NULL) {
807 				m_freem(m);
808 				return;
809 			}
810 			--rt->rt_refcnt;
811 			/*
812 			 * Don't send proxies for nodes on the same interface
813 			 * as this one came out of, or we'll get into a fight
814 			 * over who claims what Ether address.
815 			 */
816 			if (rt->rt_ifp == ifp) {
817 				m_freem(m);
818 				return;
819 			}
820 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
821 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
822 #ifdef DEBUG_PROXY
823 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
824 #endif
825 		} else {
826 			struct sockaddr_dl *sdl;
827 
828 			rt = la->la_rt;
829 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
830 			sdl = SDL(rt->rt_gateway);
831 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
832 		}
833 	}
834 
835 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
836 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
837 	ah->ar_op = htons(ARPOP_REPLY);
838 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
839 	switch (ifp->if_type) {
840 	case IFT_ETHER:
841 	/*
842 	 * May not be correct for types not explictly
843 	 * listed, but it is our best guess.
844 	 */
845 	default:
846 		eh = (struct ether_header *)sa.sa_data;
847 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
848 		eh->ether_type = htons(ETHERTYPE_ARP);
849 		break;
850 	}
851 	sa.sa_family = AF_UNSPEC;
852 	sa.sa_len = sizeof sa;
853 	ifp->if_output(ifp, m, &sa, NULL);
854 }
855 
856 #ifdef SMP
857 
858 static void
859 arp_update_msghandler(struct netmsg *netmsg)
860 {
861 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
862 	int nextcpu;
863 
864 	arp_update_oncpu(msg->m, msg->saddr, msg->create, mycpuid == 0);
865 
866 	nextcpu = mycpuid + 1;
867 	if (nextcpu < ncpus)
868 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
869 	else
870 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
871 }
872 
873 #endif	/* SMP */
874 
875 #endif	/* INET */
876 
877 /*
878  * Free an arp entry.  If the arp entry is actively referenced or represents
879  * a static entry we only clear it back to an unresolved state, otherwise
880  * we destroy the entry entirely.
881  *
882  * Note that static entries are created when route add ... -interface is used
883  * to create an interface route to a (direct) destination.
884  */
885 static void
886 arptfree(struct llinfo_arp *la)
887 {
888 	struct rtentry *rt = la->la_rt;
889 	struct sockaddr_dl *sdl;
890 
891 	if (rt == NULL)
892 		panic("arptfree");
893 	sdl = SDL(rt->rt_gateway);
894 	if (sdl != NULL &&
895 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
896 	     (rt->rt_flags & RTF_STATIC))) {
897 		sdl->sdl_alen = 0;
898 		la->la_preempt = la->la_asked = 0;
899 		rt->rt_flags &= ~RTF_REJECT;
900 		return;
901 	}
902 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
903 }
904 
905 /*
906  * Lookup or enter a new address in arptab.
907  */
908 static struct llinfo_arp *
909 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
910 {
911 	struct rtentry *rt;
912 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
913 	const char *why = NULL;
914 
915 	sin.sin_addr.s_addr = addr;
916 	sin.sin_other = proxy ? SIN_PROXY : 0;
917 	if (create)
918 		rt = rtlookup((struct sockaddr *)&sin);
919 	else
920 		rt = rtpurelookup((struct sockaddr *)&sin);
921 	if (rt == NULL)
922 		return (NULL);
923 	rt->rt_refcnt--;
924 
925 	if (rt->rt_flags & RTF_GATEWAY)
926 		why = "host is not on local network";
927 	else if (!(rt->rt_flags & RTF_LLINFO))
928 		why = "could not allocate llinfo";
929 	else if (rt->rt_gateway->sa_family != AF_LINK)
930 		why = "gateway route is not ours";
931 
932 	if (why) {
933 		if (create) {
934 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
935 			    inet_ntoa(sin.sin_addr), why);
936 		}
937 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
938 			/* No references to this route.  Purge it. */
939 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
940 				  rt_mask(rt), rt->rt_flags, NULL);
941 		}
942 		return (NULL);
943 	}
944 	return (rt->rt_llinfo);
945 }
946 
947 void
948 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
949 {
950 	ASSERT_SERIALIZED(ifp->if_serializer);
951 
952 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
953 		lwkt_serialize_exit(ifp->if_serializer);
954 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
955 			   IF_LLADDR(ifp));
956 		lwkt_serialize_enter(ifp->if_serializer);
957 	}
958 	ifa->ifa_rtrequest = arp_rtrequest;
959 	ifa->ifa_flags |= RTF_CLONING;
960 }
961 
962 void
963 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
964 {
965 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
966 
967 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
968 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
969 			   enaddr);
970 	ifa->ifa_rtrequest = arp_rtrequest;
971 	ifa->ifa_flags |= RTF_CLONING;
972 }
973 
974 static void
975 arp_init(void)
976 {
977 	int cpu;
978 
979 	for (cpu = 0; cpu < ncpus2; cpu++)
980 		LIST_INIT(&llinfo_arp_list[cpu]);
981 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr,
982 			NETISR_FLAG_NOTMPSAFE);
983 }
984 
985 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
986