xref: /dflybsd-src/sys/netinet/if_ether.c (revision f0a5c1029a4d9e5497fe457e38db8db7a021dc86)
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 1.46 2008/05/14 11:59:24 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 
105 #ifdef CARP
106 #include <netinet/ip_carp.h>
107 #endif
108 
109 #define SIN(s) ((struct sockaddr_in *)s)
110 #define SDL(s) ((struct sockaddr_dl *)s)
111 
112 SYSCTL_DECL(_net_link_ether);
113 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
114 
115 /* timer values */
116 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
117 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
118 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
119 
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
121 	   &arpt_prune, 0, "");
122 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
123 	   &arpt_keep, 0, "");
124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
125 	   &arpt_down, 0, "");
126 
127 #define	rt_expire	rt_rmx.rmx_expire
128 
129 struct llinfo_arp {
130 	LIST_ENTRY(llinfo_arp) la_le;
131 	struct	rtentry *la_rt;
132 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
133 	u_short	la_preempt;	/* countdown for pre-expiry arps */
134 	u_short	la_asked;	/* #times we QUERIED following expiration */
135 };
136 
137 static	LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
138 
139 static int	arp_maxtries = 5;
140 static int	useloopback = 1; /* use loopback interface for local traffic */
141 static int	arp_proxyall = 0;
142 
143 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
144 	   &arp_maxtries, 0, "");
145 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
146 	   &useloopback, 0, "");
147 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
148 	   &arp_proxyall, 0, "");
149 
150 void 		arprequest_acces(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,  u_char *enaddr);
151 static void	arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
152 static void	arprequest (struct ifnet *,
153 			struct in_addr *, struct in_addr *, u_char *);
154 static void	arpintr(struct netmsg *);
155 static void	arptfree (struct llinfo_arp *);
156 static void	arptimer (void *);
157 static struct llinfo_arp
158 		*arplookup (in_addr_t addr, boolean_t create, boolean_t proxy);
159 #ifdef INET
160 static void	in_arpinput (struct mbuf *);
161 #endif
162 
163 static struct callout	arptimer_ch[MAXCPU];
164 
165 /*
166  * Timeout routine.  Age arp_tab entries periodically.
167  */
168 /* ARGSUSED */
169 static void
170 arptimer(void *ignored_arg)
171 {
172 	struct llinfo_arp *la, *nla;
173 
174 	crit_enter();
175 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
176 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
177 			arptfree(la);
178 	}
179 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
180 	crit_exit();
181 }
182 
183 /*
184  * Parallel to llc_rtrequest.
185  */
186 static void
187 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
188 {
189 	struct sockaddr *gate = rt->rt_gateway;
190 	struct llinfo_arp *la = rt->rt_llinfo;
191 
192 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
193 	static boolean_t arpinit_done[MAXCPU];
194 
195 	if (!arpinit_done[mycpuid]) {
196 		arpinit_done[mycpuid] = TRUE;
197 		callout_init(&arptimer_ch[mycpuid]);
198 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
199 	}
200 	if (rt->rt_flags & RTF_GATEWAY)
201 		return;
202 
203 	switch (req) {
204 	case RTM_ADD:
205 		/*
206 		 * XXX: If this is a manually added route to interface
207 		 * such as older version of routed or gated might provide,
208 		 * restore cloning bit.
209 		 */
210 		if (!(rt->rt_flags & RTF_HOST) &&
211 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
212 			rt->rt_flags |= RTF_CLONING;
213 		if (rt->rt_flags & RTF_CLONING) {
214 			/*
215 			 * Case 1: This route should come from a route to iface.
216 			 */
217 			rt_setgate(rt, rt_key(rt),
218 				   (struct sockaddr *)&null_sdl);
219 			gate = rt->rt_gateway;
220 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
221 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
222 			rt->rt_expire = time_second;
223 			break;
224 		}
225 		/* Announce a new entry if requested. */
226 		if (rt->rt_flags & RTF_ANNOUNCE) {
227 			arprequest(rt->rt_ifp,
228 			    &SIN(rt_key(rt))->sin_addr,
229 			    &SIN(rt_key(rt))->sin_addr,
230 			    LLADDR(SDL(gate)));
231 		}
232 		/*FALLTHROUGH*/
233 	case RTM_RESOLVE:
234 		if (gate->sa_family != AF_LINK ||
235 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
236 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
237 			break;
238 		}
239 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
240 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
241 		if (la != NULL)
242 			break; /* This happens on a route change */
243 		/*
244 		 * Case 2:  This route may come from cloning, or a manual route
245 		 * add with a LL address.
246 		 */
247 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
248 		rt->rt_llinfo = la;
249 		if (la == NULL) {
250 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
251 			break;
252 		}
253 		bzero(la, sizeof *la);
254 		la->la_rt = rt;
255 		rt->rt_flags |= RTF_LLINFO;
256 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
257 
258 #ifdef INET
259 		/*
260 		 * This keeps the multicast addresses from showing up
261 		 * in `arp -a' listings as unresolved.  It's not actually
262 		 * functional.  Then the same for broadcast.
263 		 */
264 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
265 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
266 					       LLADDR(SDL(gate)));
267 			SDL(gate)->sdl_alen = 6;
268 			rt->rt_expire = 0;
269 		}
270 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
271 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
272 			       rt->rt_ifp->if_addrlen);
273 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
274 			rt->rt_expire = 0;
275 		}
276 #endif
277 
278 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
279 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
280 			/*
281 			 * This test used to be
282 			 *	if (loif.if_flags & IFF_UP)
283 			 * It allowed local traffic to be forced
284 			 * through the hardware by configuring the
285 			 * loopback down.  However, it causes problems
286 			 * during network configuration for boards
287 			 * that can't receive packets they send.  It
288 			 * is now necessary to clear "useloopback" and
289 			 * remove the route to force traffic out to
290 			 * the hardware.
291 			 */
292 			rt->rt_expire = 0;
293 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
294 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
295 			if (useloopback)
296 				rt->rt_ifp = loif;
297 		}
298 		break;
299 
300 	case RTM_DELETE:
301 		if (la == NULL)
302 			break;
303 		LIST_REMOVE(la, la_le);
304 		rt->rt_llinfo = NULL;
305 		rt->rt_flags &= ~RTF_LLINFO;
306 		if (la->la_hold != NULL)
307 			m_freem(la->la_hold);
308 		Free(la);
309 	}
310 }
311 
312 /*
313  * Broadcast an ARP request. Caller specifies:
314  *	- arp header source ip address
315  *	- arp header target ip address
316  *	- arp header source ethernet address
317  */
318 static void
319 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
320 	   u_char *enaddr)
321 {
322 	struct mbuf *m;
323 	struct ether_header *eh;
324 	struct arphdr *ah;
325 	struct sockaddr sa;
326 	u_short ar_hrd;
327 
328 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
329 		return;
330 	m->m_pkthdr.rcvif = (struct ifnet *)NULL;
331 
332 	switch (ifp->if_type) {
333 	case IFT_ETHER:
334 		/*
335 		 * This may not be correct for types not explicitly
336 		 * listed, but this is our best guess
337 		 */
338 	default:
339 		ar_hrd = htons(ARPHRD_ETHER);
340 
341 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
342 		m->m_pkthdr.len = m->m_len;
343 		MH_ALIGN(m, m->m_len);
344 
345 		eh = (struct ether_header *)sa.sa_data;
346 		/* if_output() will not swap */
347 		eh->ether_type = htons(ETHERTYPE_ARP);
348 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
349 
350 		ah = mtod(m, struct arphdr *);
351 		break;
352 	}
353 
354 	ah->ar_hrd = ar_hrd;
355 	ah->ar_pro = htons(ETHERTYPE_IP);
356 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
357 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
358 	ah->ar_op = htons(ARPOP_REQUEST);
359 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
360 	memset(ar_tha(ah), 0, ah->ar_hln);
361 	memcpy(ar_spa(ah), sip, ah->ar_pln);
362 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
363 
364 	sa.sa_family = AF_UNSPEC;
365 	sa.sa_len = sizeof sa;
366 	ifp->if_output(ifp, m, &sa, (struct rtentry *)NULL);
367 }
368 
369 /*
370  * Resolve an IP address into an ethernet address.  If success,
371  * desten is filled in.  If there is no entry in arptab,
372  * set one up and broadcast a request for the IP address.
373  * Hold onto this mbuf and resend it once the address
374  * is finally resolved.  A return value of 1 indicates
375  * that desten has been filled in and the packet should be sent
376  * normally; a 0 return indicates that the packet has been
377  * taken over here, either now or for later transmission.
378  */
379 int
380 arpresolve(
381 	struct ifnet *ifp,
382 	struct rtentry *rt0,
383 	struct mbuf *m,
384 	struct sockaddr *dst,
385 	u_char *desten)
386 {
387 	struct rtentry *rt;
388 	struct llinfo_arp *la = NULL;
389 	struct sockaddr_dl *sdl;
390 
391 	if (m->m_flags & M_BCAST) {	/* broadcast */
392 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
393 		return (1);
394 	}
395 	if (m->m_flags & M_MCAST) {/* multicast */
396 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
397 		return (1);
398 	}
399 	if (rt0 != NULL) {
400 		if (rt_llroute(dst, rt0, &rt) != 0) {
401 			m_freem(m);
402 			return 0;
403 		}
404 		la = rt->rt_llinfo;
405 	}
406 	if (la == NULL) {
407 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
408 		if (la != NULL)
409 			rt = la->la_rt;
410 	}
411 	if (la == NULL || rt == NULL) {
412 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
413 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
414 		    rt ? "rt" : "");
415 		m_freem(m);
416 		return (0);
417 	}
418 	sdl = SDL(rt->rt_gateway);
419 	/*
420 	 * Check the address family and length is valid, the address
421 	 * is resolved; otherwise, try to resolve.
422 	 */
423 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
424 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
425 		/*
426 		 * If entry has an expiry time and it is approaching,
427 		 * see if we need to send an ARP request within this
428 		 * arpt_down interval.
429 		 */
430 		if ((rt->rt_expire != 0) &&
431 		    (time_second + la->la_preempt > rt->rt_expire)) {
432 			arprequest(ifp,
433 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
434 				   &SIN(dst)->sin_addr,
435 				   IF_LLADDR(ifp));
436 			la->la_preempt--;
437 		}
438 
439 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
440 		return 1;
441 	}
442 	/*
443 	 * If ARP is disabled on this interface, stop.
444 	 * XXX
445 	 * Probably should not allocate empty llinfo struct if we are
446 	 * not going to be sending out an arp request.
447 	 */
448 	if (ifp->if_flags & IFF_NOARP) {
449 		m_freem(m);
450 		return (0);
451 	}
452 	/*
453 	 * There is an arptab entry, but no ethernet address
454 	 * response yet.  Replace the held mbuf with this
455 	 * latest one.
456 	 */
457 	if (la->la_hold != NULL)
458 		m_freem(la->la_hold);
459 	la->la_hold = m;
460 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
461 		rt->rt_flags &= ~RTF_REJECT;
462 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
463 			rt->rt_expire = time_second;
464 			if (la->la_asked++ < arp_maxtries) {
465 				arprequest(ifp,
466 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
467 					   &SIN(dst)->sin_addr,
468 					   IF_LLADDR(ifp));
469 			} else {
470 				rt->rt_flags |= RTF_REJECT;
471 				rt->rt_expire += arpt_down;
472 				la->la_asked = 0;
473 				la->la_preempt = arp_maxtries;
474 			}
475 
476 		}
477 	}
478 	return (0);
479 }
480 
481 /*
482  * Common length and type checks are done here,
483  * then the protocol-specific routine is called.
484  */
485 static void
486 arpintr(struct netmsg *msg)
487 {
488 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
489 	struct arphdr *ar;
490 	u_short ar_hrd;
491 
492 	if (m->m_len < sizeof(struct arphdr) &&
493 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
494 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
495 		goto out2;
496 	}
497 	ar = mtod(m, struct arphdr *);
498 
499 	ar_hrd = ntohs(ar->ar_hrd);
500 	if (ar_hrd != ARPHRD_ETHER &&
501 	    ar_hrd != ARPHRD_IEEE802) {
502 		log(LOG_ERR,
503 		    "arp: unknown hardware address format (0x%2D)\n",
504 		    (unsigned char *)&ar->ar_hrd, "");
505 		goto out1;
506 	}
507 
508 	if (m->m_pkthdr.len < arphdr_len(ar) &&
509 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
510 		log(LOG_ERR, "arp: runt packet\n");
511 		goto out1;
512 	}
513 
514 	switch (ntohs(ar->ar_pro)) {
515 #ifdef INET
516 		case ETHERTYPE_IP:
517 			in_arpinput(m);
518 			goto out2;
519 #endif
520 	}
521 out1:
522 	m_freem(m);
523 out2:
524 	;
525 	/* msg was embedded in the mbuf, do not reply! */
526 }
527 
528 #ifdef INET
529 /*
530  * ARP for Internet protocols on 10 Mb/s Ethernet.
531  * Algorithm is that given in RFC 826.
532  * In addition, a sanity check is performed on the sender
533  * protocol address, to catch impersonators.
534  * We no longer handle negotiations for use of trailer protocol:
535  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
536  * along with IP replies if we wanted trailers sent to us,
537  * and also sent them in response to IP replies.
538  * This allowed either end to announce the desire to receive
539  * trailer packets.
540  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
541  * but formerly didn't normally send requests.
542  */
543 static int log_arp_wrong_iface = 1;
544 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
545 	&log_arp_wrong_iface, 0,
546 	"log arp packets arriving on the wrong interface");
547 
548 static void
549 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
550 		 boolean_t dologging)
551 {
552 	struct arphdr *ah = mtod(m, struct arphdr *);
553 	struct ifnet *ifp = m->m_pkthdr.rcvif;
554 	struct llinfo_arp *la;
555 	struct sockaddr_dl *sdl;
556 	struct rtentry *rt;
557 	int cpu = mycpuid;
558 
559 	la = arplookup(saddr, create, FALSE);
560 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
561 		struct in_addr isaddr = { saddr };
562 
563 		/* the following is not an error when doing bridging */
564 		if (rt->rt_ifp != ifp) {
565 			if (dologging && log_arp_wrong_iface && cpu == 0) {
566 				log(LOG_ERR,
567 				    "arp: %s is on %s "
568 				    "but got reply from %*D on %s\n",
569 				    inet_ntoa(isaddr),
570 				    rt->rt_ifp->if_xname,
571 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
572 				    ifp->if_xname);
573 			}
574 			return;
575 		}
576 		if (sdl->sdl_alen &&
577 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
578 			if (rt->rt_expire != 0) {
579 				if (dologging && cpu == 0) {
580 			    		log(LOG_INFO,
581 			    		"arp: %s moved from %*D to %*D on %s\n",
582 			    		inet_ntoa(isaddr),
583 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
584 			    		":", ifp->if_addrlen,
585 			    		(u_char *)ar_sha(ah), ":",
586 			    		ifp->if_xname);
587 				}
588 			} else {
589 				if (dologging && cpu == 0) {
590 					log(LOG_ERR,
591 					"arp: %*D attempts to modify permanent entry for %s on %s\n",
592 					ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
593 					inet_ntoa(isaddr), ifp->if_xname);
594 				}
595 				return;
596 			}
597 		}
598 		/*
599 		 * sanity check for the address length.
600 		 * XXX this does not work for protocols with variable address
601 		 * length. -is
602 		 */
603 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln &&
604 		    cpu == 0)
605 		{
606 			log(LOG_WARNING,
607 			    "arp from %*D: new addr len %d, was %d",
608 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
609 			    ah->ar_hln, sdl->sdl_alen);
610 		}
611 		if (ifp->if_addrlen != ah->ar_hln) {
612 			if (dologging && cpu == 0) {
613 				log(LOG_WARNING,
614 				"arp from %*D: addr len: new %d, i/f %d (ignored)",
615 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
616 				ah->ar_hln, ifp->if_addrlen);
617 			}
618 			return;
619 		}
620 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
621 		if (rt->rt_expire != 0)
622 			rt->rt_expire = time_second + arpt_keep;
623 		rt->rt_flags &= ~RTF_REJECT;
624 		la->la_asked = 0;
625 		la->la_preempt = arp_maxtries;
626 
627 		/*
628 		 * This particular cpu might have been holding an mbuf
629 		 * pending ARP resolution.  If so, transmit the mbuf now.
630 		 */
631 		if (la->la_hold != NULL) {
632 			m_adj(la->la_hold, sizeof(struct ether_header));
633 			ifp->if_output(ifp, la->la_hold, rt_key(rt), rt);
634 			la->la_hold = NULL;
635 		}
636 	}
637 }
638 
639 #ifdef SMP
640 
641 struct netmsg_arp_update {
642 	struct netmsg	netmsg;
643 	struct mbuf	*m;
644 	in_addr_t	saddr;
645 	boolean_t	create;
646 };
647 
648 static void arp_update_msghandler(struct netmsg *netmsg);
649 
650 #endif
651 
652 /*
653  * Called from arpintr() - this routine is run from a single cpu.
654  */
655 static void
656 in_arpinput(struct mbuf *m)
657 {
658 	struct arphdr *ah;
659 	struct ifnet *ifp = m->m_pkthdr.rcvif;
660 	struct ether_header *eh;
661 	struct rtentry *rt;
662 	struct ifaddr_container *ifac;
663 	struct in_ifaddr *ia;
664 	struct sockaddr sa;
665 	struct in_addr isaddr, itaddr, myaddr;
666 #ifdef SMP
667 	struct netmsg_arp_update msg;
668 #endif
669 	u_int8_t *enaddr = NULL;
670 	int op;
671 	int req_len;
672 
673 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
674 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
675 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
676 		return;
677 	}
678 
679 	ah = mtod(m, struct arphdr *);
680 	op = ntohs(ah->ar_op);
681 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
682 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
683 	/*
684 	 * Check both target and sender IP addresses:
685 	 *
686 	 * If we receive the packet on the interface owning the address,
687 	 * then accept the address.
688 	 *
689 	 * For a bridge, we accept the address if the receive interface and
690 	 * the interface owning the address are on the same bridge.
691 	 * (This will change slightly when we have clusters of interfaces).
692 	 */
693 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
694 		/* Skip all ia's which don't match */
695 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
696 			continue;
697 
698 		if (ia->ia_ifp == ifp)
699 			goto match;
700 
701 		if (ifp->if_bridge && ia->ia_ifp &&
702 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
703 			goto match;
704 
705 #ifdef CARP
706 	/*
707 	 * If the interface does not match, but the recieving interface
708 	 * is part of carp, we call carp_iamatch to see if this is a
709 	 * request for the virtual host ip.
710 	 * XXX: This is really ugly!
711 	 */
712         	if (ifp->if_carp != NULL &&
713 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
714 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
715                         goto match;
716 #endif
717 	}
718 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) {
719 		/* Skip all ia's which don't match */
720 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
721 			continue;
722 
723 		if (ia->ia_ifp == ifp)
724 			goto match;
725 
726 		if (ifp->if_bridge && ia->ia_ifp &&
727 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
728 			goto match;
729 	}
730 	/*
731 	 * No match, use the first inet address on the receive interface
732 	 * as a dummy address for the rest of the function.
733 	 */
734 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
735 		struct ifaddr *ifa = ifac->ifa;
736 
737 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
738 			ia = ifatoia(ifa);
739 			goto match;
740 		}
741 	}
742 	/*
743 	 * If we got here, we didn't find any suitable interface,
744 	 * so drop the packet.
745 	 */
746 	m_freem(m);
747 	return;
748 
749 match:
750 	if (!enaddr)
751 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
752 	myaddr = ia->ia_addr.sin_addr;
753 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
754 		m_freem(m);	/* it's from me, ignore it. */
755 		return;
756 	}
757 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
758 		log(LOG_ERR,
759 		    "arp: link address is broadcast for IP address %s!\n",
760 		    inet_ntoa(isaddr));
761 		m_freem(m);
762 		return;
763 	}
764 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
765 		log(LOG_ERR,
766 		   "arp: %*D is using my IP address %s!\n",
767 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
768 		   inet_ntoa(isaddr));
769 		itaddr = myaddr;
770 		goto reply;
771 	}
772 #ifdef SMP
773 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
774 		    arp_update_msghandler);
775 	msg.m = m;
776 	msg.saddr = isaddr.s_addr;
777 	msg.create = (itaddr.s_addr == myaddr.s_addr);
778 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
779 #endif
780 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
781 			 TRUE);
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), enaddr, ah->ar_hln);
791 	} else {
792 		struct llinfo_arp *la;
793 
794 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
795 		if (la == NULL) {
796 			struct sockaddr_in sin;
797 
798 			if (!arp_proxyall) {
799 				m_freem(m);
800 				return;
801 			}
802 
803 			bzero(&sin, sizeof sin);
804 			sin.sin_family = AF_INET;
805 			sin.sin_len = sizeof sin;
806 			sin.sin_addr = itaddr;
807 
808 			rt = rtpurelookup((struct sockaddr *)&sin);
809 			if (rt == NULL) {
810 				m_freem(m);
811 				return;
812 			}
813 			--rt->rt_refcnt;
814 			/*
815 			 * Don't send proxies for nodes on the same interface
816 			 * as this one came out of, or we'll get into a fight
817 			 * over who claims what Ether address.
818 			 */
819 			if (rt->rt_ifp == ifp) {
820 				m_freem(m);
821 				return;
822 			}
823 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
824 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
825 #ifdef DEBUG_PROXY
826 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
827 #endif
828 		} else {
829 			struct sockaddr_dl *sdl;
830 
831 			rt = la->la_rt;
832 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
833 			sdl = SDL(rt->rt_gateway);
834 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
835 		}
836 	}
837 
838 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
839 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
840 	ah->ar_op = htons(ARPOP_REPLY);
841 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
842 	switch (ifp->if_type) {
843 	case IFT_ETHER:
844 	/*
845 	 * May not be correct for types not explictly
846 	 * listed, but it is our best guess.
847 	 */
848 	default:
849 		eh = (struct ether_header *)sa.sa_data;
850 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
851 		eh->ether_type = htons(ETHERTYPE_ARP);
852 		break;
853 	}
854 	sa.sa_family = AF_UNSPEC;
855 	sa.sa_len = sizeof sa;
856 	ifp->if_output(ifp, m, &sa, (struct rtentry *)0);
857 	return;
858 }
859 
860 #ifdef SMP
861 
862 static
863 void
864 arp_update_msghandler(struct netmsg *netmsg)
865 {
866 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
867 	int nextcpu;
868 
869 	arp_update_oncpu(msg->m, msg->saddr, msg->create, FALSE);
870 
871 	nextcpu = mycpuid + 1;
872 	if (nextcpu < ncpus) {
873 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
874 	} else {
875 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
876 	}
877 }
878 
879 #endif
880 
881 #endif
882 
883 /*
884  * Free an arp entry.  If the arp entry is actively referenced or represents
885  * a static entry we only clear it back to an unresolved state, otherwise
886  * we destroy the entry entirely.
887  *
888  * Note that static entries are created when route add ... -interface is used
889  * to create an interface route to a (direct) destination.
890  */
891 static void
892 arptfree(struct llinfo_arp *la)
893 {
894 	struct rtentry *rt = la->la_rt;
895 	struct sockaddr_dl *sdl;
896 
897 	if (rt == NULL)
898 		panic("arptfree");
899 	sdl = SDL(rt->rt_gateway);
900 	if (sdl != NULL &&
901 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
902 	     (rt->rt_flags & RTF_STATIC))) {
903 		sdl->sdl_alen = 0;
904 		la->la_preempt = la->la_asked = 0;
905 		rt->rt_flags &= ~RTF_REJECT;
906 		return;
907 	}
908 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
909 }
910 
911 /*
912  * Lookup or enter a new address in arptab.
913  */
914 static struct llinfo_arp *
915 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
916 {
917 	struct rtentry *rt;
918 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
919 	const char *why = NULL;
920 
921 	sin.sin_addr.s_addr = addr;
922 	sin.sin_other = proxy ? SIN_PROXY : 0;
923 	if (create)
924 		rt = rtlookup((struct sockaddr *)&sin);
925 	else
926 		rt = rtpurelookup((struct sockaddr *)&sin);
927 	if (rt == NULL)
928 		return (NULL);
929 	rt->rt_refcnt--;
930 
931 	if (rt->rt_flags & RTF_GATEWAY)
932 		why = "host is not on local network";
933 	else if (!(rt->rt_flags & RTF_LLINFO))
934 		why = "could not allocate llinfo";
935 	else if (rt->rt_gateway->sa_family != AF_LINK)
936 		why = "gateway route is not ours";
937 
938 	if (why) {
939 		if (create) {
940 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
941 			    inet_ntoa(sin.sin_addr), why);
942 		}
943 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
944 			/* No references to this route.  Purge it. */
945 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
946 				  rt_mask(rt), rt->rt_flags, NULL);
947 		}
948 		return (NULL);
949 	}
950 	return (rt->rt_llinfo);
951 }
952 
953 void
954 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
955 {
956 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
957 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
958 			   IF_LLADDR(ifp));
959 	ifa->ifa_rtrequest = arp_rtrequest;
960 	ifa->ifa_flags |= RTF_CLONING;
961 }
962 
963 void
964 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
965 {
966 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
967 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
968 			   enaddr);
969 	ifa->ifa_rtrequest = arp_rtrequest;
970 	ifa->ifa_flags |= RTF_CLONING;
971 }
972 
973 static void
974 arp_init(void)
975 {
976 	int cpu;
977 
978 	for (cpu = 0; cpu < ncpus2; cpu++)
979 		LIST_INIT(&llinfo_arp_list[cpu]);
980 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr);
981 }
982 
983 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
984