xref: /dflybsd-src/sys/netinet/if_ether.c (revision 212efd99d654b838bcfc3fc5965862b0a610f81e)
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 1.57 2008/10/07 22:30:31 thomas 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 #include <sys/lock.h>
89 
90 #include <net/if.h>
91 #include <net/if_dl.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #include <net/if_llc.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/if_ether.h>
100 
101 #include <sys/thread2.h>
102 #include <sys/msgport2.h>
103 #include <net/netmsg2.h>
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 static int	arp_mpsafe = 0;
151 TUNABLE_INT("net.link.ether.inet.arp_mpsafe", &arp_mpsafe);
152 
153 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
154 static void	arprequest(struct ifnet *, struct in_addr *, struct in_addr *,
155 			   const u_char *);
156 static void	arprequest_async(struct ifnet *, struct in_addr *,
157 				 struct in_addr *, const u_char *);
158 static void	arpintr(struct netmsg *);
159 static void	arptfree(struct llinfo_arp *);
160 static void	arptimer(void *);
161 static struct llinfo_arp *
162 		arplookup(in_addr_t, boolean_t, boolean_t);
163 #ifdef INET
164 static void	in_arpinput(struct mbuf *);
165 #endif
166 
167 static struct callout	arptimer_ch[MAXCPU];
168 
169 /*
170  * Timeout routine.  Age arp_tab entries periodically.
171  */
172 /* ARGSUSED */
173 static void
174 arptimer(void *ignored_arg)
175 {
176 	struct llinfo_arp *la, *nla;
177 
178 	crit_enter();
179 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
180 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
181 			arptfree(la);
182 	}
183 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
184 	crit_exit();
185 }
186 
187 /*
188  * Parallel to llc_rtrequest.
189  */
190 static void
191 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
192 {
193 	struct sockaddr *gate = rt->rt_gateway;
194 	struct llinfo_arp *la = rt->rt_llinfo;
195 
196 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
197 	static boolean_t arpinit_done[MAXCPU];
198 
199 	if (!arpinit_done[mycpuid]) {
200 		arpinit_done[mycpuid] = TRUE;
201 		callout_init(&arptimer_ch[mycpuid]);
202 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
203 	}
204 	if (rt->rt_flags & RTF_GATEWAY)
205 		return;
206 
207 	switch (req) {
208 	case RTM_ADD:
209 		/*
210 		 * XXX: If this is a manually added route to interface
211 		 * such as older version of routed or gated might provide,
212 		 * restore cloning bit.
213 		 */
214 		if (!(rt->rt_flags & RTF_HOST) &&
215 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
216 			rt->rt_flags |= RTF_CLONING;
217 		if (rt->rt_flags & RTF_CLONING) {
218 			/*
219 			 * Case 1: This route should come from a route to iface.
220 			 */
221 			rt_setgate(rt, rt_key(rt),
222 				   (struct sockaddr *)&null_sdl);
223 			gate = rt->rt_gateway;
224 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
225 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
226 			rt->rt_expire = time_second;
227 			break;
228 		}
229 		/* Announce a new entry if requested. */
230 		if (rt->rt_flags & RTF_ANNOUNCE) {
231 			arprequest_async(rt->rt_ifp,
232 			    &SIN(rt_key(rt))->sin_addr,
233 			    &SIN(rt_key(rt))->sin_addr,
234 			    LLADDR(SDL(gate)));
235 		}
236 		/*FALLTHROUGH*/
237 	case RTM_RESOLVE:
238 		if (gate->sa_family != AF_LINK ||
239 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
240 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
241 			break;
242 		}
243 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
244 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
245 		if (la != NULL)
246 			break; /* This happens on a route change */
247 		/*
248 		 * Case 2:  This route may come from cloning, or a manual route
249 		 * add with a LL address.
250 		 */
251 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
252 		rt->rt_llinfo = la;
253 		if (la == NULL) {
254 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
255 			break;
256 		}
257 		bzero(la, sizeof *la);
258 		la->la_rt = rt;
259 		rt->rt_flags |= RTF_LLINFO;
260 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
261 
262 #ifdef INET
263 		/*
264 		 * This keeps the multicast addresses from showing up
265 		 * in `arp -a' listings as unresolved.  It's not actually
266 		 * functional.  Then the same for broadcast.
267 		 */
268 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
269 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
270 					       LLADDR(SDL(gate)));
271 			SDL(gate)->sdl_alen = 6;
272 			rt->rt_expire = 0;
273 		}
274 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
275 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
276 			       rt->rt_ifp->if_addrlen);
277 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
278 			rt->rt_expire = 0;
279 		}
280 #endif
281 
282 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
283 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
284 			/*
285 			 * This test used to be
286 			 *	if (loif.if_flags & IFF_UP)
287 			 * It allowed local traffic to be forced
288 			 * through the hardware by configuring the
289 			 * loopback down.  However, it causes problems
290 			 * during network configuration for boards
291 			 * that can't receive packets they send.  It
292 			 * is now necessary to clear "useloopback" and
293 			 * remove the route to force traffic out to
294 			 * the hardware.
295 			 */
296 			rt->rt_expire = 0;
297 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
298 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
299 			if (useloopback)
300 				rt->rt_ifp = loif;
301 		}
302 		break;
303 
304 	case RTM_DELETE:
305 		if (la == NULL)
306 			break;
307 		LIST_REMOVE(la, la_le);
308 		rt->rt_llinfo = NULL;
309 		rt->rt_flags &= ~RTF_LLINFO;
310 		if (la->la_hold != NULL)
311 			m_freem(la->la_hold);
312 		Free(la);
313 		break;
314 	}
315 }
316 
317 static struct mbuf *
318 arpreq_alloc(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
319 	     const u_char *enaddr)
320 {
321 	struct mbuf *m;
322 	struct arphdr *ah;
323 	u_short ar_hrd;
324 
325 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
326 		return NULL;
327 	m->m_pkthdr.rcvif = NULL;
328 
329 	switch (ifp->if_type) {
330 	case IFT_ETHER:
331 		/*
332 		 * This may not be correct for types not explicitly
333 		 * listed, but this is our best guess
334 		 */
335 	default:
336 		ar_hrd = htons(ARPHRD_ETHER);
337 
338 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
339 		m->m_pkthdr.len = m->m_len;
340 		MH_ALIGN(m, m->m_len);
341 
342 		ah = mtod(m, struct arphdr *);
343 		break;
344 	}
345 
346 	ah->ar_hrd = ar_hrd;
347 	ah->ar_pro = htons(ETHERTYPE_IP);
348 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
349 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
350 	ah->ar_op = htons(ARPOP_REQUEST);
351 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
352 	memset(ar_tha(ah), 0, ah->ar_hln);
353 	memcpy(ar_spa(ah), sip, ah->ar_pln);
354 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
355 
356 	return m;
357 }
358 
359 static void
360 arpreq_send(struct ifnet *ifp, struct mbuf *m)
361 {
362 	struct sockaddr sa;
363 	struct ether_header *eh;
364 
365 	switch (ifp->if_type) {
366 	case IFT_ETHER:
367 		/*
368 		 * This may not be correct for types not explicitly
369 		 * listed, but this is our best guess
370 		 */
371 	default:
372 		eh = (struct ether_header *)sa.sa_data;
373 		/* if_output() will not swap */
374 		eh->ether_type = htons(ETHERTYPE_ARP);
375 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
376 		break;
377 	}
378 
379 	sa.sa_family = AF_UNSPEC;
380 	sa.sa_len = sizeof(sa);
381 	ifp->if_output(ifp, m, &sa, NULL);
382 }
383 
384 static void
385 arpreq_send_handler(struct netmsg *nmsg)
386 {
387 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
388 	struct ifnet *ifp = nmsg->nm_lmsg.u.ms_resultp;
389 
390 	arpreq_send(ifp, m);
391 	/* nmsg was embedded in the mbuf, do not reply! */
392 }
393 
394 /*
395  * Broadcast an ARP request. Caller specifies:
396  *	- arp header source ip address
397  *	- arp header target ip address
398  *	- arp header source ethernet address
399  *
400  * NOTE: Caller MUST NOT hold ifp's serializer
401  */
402 static void
403 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
404 	   const u_char *enaddr)
405 {
406 	struct mbuf *m;
407 
408 	m = arpreq_alloc(ifp, sip, tip, enaddr);
409 	if (m == NULL)
410 		return;
411 	arpreq_send(ifp, m);
412 }
413 
414 /*
415  * Same as arprequest(), except:
416  * - Caller is allowed to hold ifp's serializer
417  * - Network output is done in TDF_NETWORK kernel thread
418  */
419 static void
420 arprequest_async(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
421 		 const u_char *enaddr)
422 {
423 	struct mbuf *m;
424 	struct netmsg_packet *pmsg;
425 
426 	m = arpreq_alloc(ifp, sip, tip, enaddr);
427 	if (m == NULL)
428 		return;
429 
430 	pmsg = &m->m_hdr.mh_netmsg;
431 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
432 		    arpreq_send_handler);
433 	pmsg->nm_packet = m;
434 	pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = ifp;
435 
436 	lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->nm_netmsg.nm_lmsg);
437 }
438 
439 /*
440  * Resolve an IP address into an ethernet address.  If success,
441  * desten is filled in.  If there is no entry in arptab,
442  * set one up and broadcast a request for the IP address.
443  * Hold onto this mbuf and resend it once the address
444  * is finally resolved.  A return value of 1 indicates
445  * that desten has been filled in and the packet should be sent
446  * normally; a 0 return indicates that the packet has been
447  * taken over here, either now or for later transmission.
448  */
449 int
450 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
451 	   struct sockaddr *dst, u_char *desten)
452 {
453 	struct rtentry *rt;
454 	struct llinfo_arp *la = NULL;
455 	struct sockaddr_dl *sdl;
456 
457 	if (m->m_flags & M_BCAST) {	/* broadcast */
458 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
459 		return (1);
460 	}
461 	if (m->m_flags & M_MCAST) {/* multicast */
462 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
463 		return (1);
464 	}
465 	if (rt0 != NULL) {
466 		if (rt_llroute(dst, rt0, &rt) != 0) {
467 			m_freem(m);
468 			return 0;
469 		}
470 		la = rt->rt_llinfo;
471 	}
472 	if (la == NULL) {
473 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
474 		if (la != NULL)
475 			rt = la->la_rt;
476 	}
477 	if (la == NULL || rt == NULL) {
478 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
479 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
480 		    rt ? "rt" : "");
481 		m_freem(m);
482 		return (0);
483 	}
484 	sdl = SDL(rt->rt_gateway);
485 	/*
486 	 * Check the address family and length is valid, the address
487 	 * is resolved; otherwise, try to resolve.
488 	 */
489 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
490 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
491 		/*
492 		 * If entry has an expiry time and it is approaching,
493 		 * see if we need to send an ARP request within this
494 		 * arpt_down interval.
495 		 */
496 		if ((rt->rt_expire != 0) &&
497 		    (time_second + la->la_preempt > rt->rt_expire)) {
498 			arprequest(ifp,
499 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
500 				   &SIN(dst)->sin_addr,
501 				   IF_LLADDR(ifp));
502 			la->la_preempt--;
503 		}
504 
505 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
506 		return 1;
507 	}
508 	/*
509 	 * If ARP is disabled on this interface, stop.
510 	 * XXX
511 	 * Probably should not allocate empty llinfo struct if we are
512 	 * not going to be sending out an arp request.
513 	 */
514 	if (ifp->if_flags & IFF_NOARP) {
515 		m_freem(m);
516 		return (0);
517 	}
518 	/*
519 	 * There is an arptab entry, but no ethernet address
520 	 * response yet.  Replace the held mbuf with this
521 	 * latest one.
522 	 */
523 	if (la->la_hold != NULL)
524 		m_freem(la->la_hold);
525 	la->la_hold = m;
526 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
527 		rt->rt_flags &= ~RTF_REJECT;
528 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
529 			rt->rt_expire = time_second;
530 			if (la->la_asked++ < arp_maxtries) {
531 				arprequest(ifp,
532 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
533 					   &SIN(dst)->sin_addr,
534 					   IF_LLADDR(ifp));
535 			} else {
536 				rt->rt_flags |= RTF_REJECT;
537 				rt->rt_expire += arpt_down;
538 				la->la_asked = 0;
539 				la->la_preempt = arp_maxtries;
540 			}
541 		}
542 	}
543 	return (0);
544 }
545 
546 /*
547  * Common length and type checks are done here,
548  * then the protocol-specific routine is called.
549  */
550 static void
551 arpintr(struct netmsg *msg)
552 {
553 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
554 	struct arphdr *ar;
555 	u_short ar_hrd;
556 
557 	if (m->m_len < sizeof(struct arphdr) &&
558 	    (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
559 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
560 		return;
561 	}
562 	ar = mtod(m, struct arphdr *);
563 
564 	ar_hrd = ntohs(ar->ar_hrd);
565 	if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
566 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
567 		    (unsigned char *)&ar->ar_hrd, "");
568 		m_freem(m);
569 		return;
570 	}
571 
572 	if (m->m_pkthdr.len < arphdr_len(ar)) {
573 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
574 			log(LOG_ERR, "arp: runt packet\n");
575 			return;
576 		}
577 		ar = mtod(m, struct arphdr *);
578 	}
579 
580 	switch (ntohs(ar->ar_pro)) {
581 #ifdef INET
582 	case ETHERTYPE_IP:
583 		in_arpinput(m);
584 		return;
585 #endif
586 	}
587 	m_freem(m);
588 	/* msg was embedded in the mbuf, do not reply! */
589 }
590 
591 #ifdef INET
592 /*
593  * ARP for Internet protocols on 10 Mb/s Ethernet.
594  * Algorithm is that given in RFC 826.
595  * In addition, a sanity check is performed on the sender
596  * protocol address, to catch impersonators.
597  * We no longer handle negotiations for use of trailer protocol:
598  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
599  * along with IP replies if we wanted trailers sent to us,
600  * and also sent them in response to IP replies.
601  * This allowed either end to announce the desire to receive
602  * trailer packets.
603  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
604  * but formerly didn't normally send requests.
605  */
606 static int log_arp_wrong_iface = 1;
607 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
608 	   &log_arp_wrong_iface, 0,
609 	   "log arp packets arriving on the wrong interface");
610 
611 static void
612 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
613 		 boolean_t dologging)
614 {
615 	struct arphdr *ah = mtod(m, struct arphdr *);
616 	struct ifnet *ifp = m->m_pkthdr.rcvif;
617 	struct llinfo_arp *la;
618 	struct sockaddr_dl *sdl;
619 	struct rtentry *rt;
620 
621 	la = arplookup(saddr, create, FALSE);
622 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
623 		struct in_addr isaddr = { saddr };
624 
625 		/* the following is not an error when doing bridging */
626 		if (rt->rt_ifp != ifp) {
627 			if (dologging && log_arp_wrong_iface) {
628 				log(LOG_ERR,
629 				    "arp: %s is on %s "
630 				    "but got reply from %*D on %s\n",
631 				    inet_ntoa(isaddr),
632 				    rt->rt_ifp->if_xname,
633 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
634 				    ifp->if_xname);
635 			}
636 			return;
637 		}
638 		if (sdl->sdl_alen &&
639 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
640 			if (rt->rt_expire != 0) {
641 				if (dologging) {
642 			    		log(LOG_INFO,
643 			    		"arp: %s moved from %*D to %*D on %s\n",
644 			    		inet_ntoa(isaddr),
645 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
646 			    		":", ifp->if_addrlen,
647 			    		(u_char *)ar_sha(ah), ":",
648 			    		ifp->if_xname);
649 				}
650 			} else {
651 				if (dologging) {
652 					log(LOG_ERR,
653 					"arp: %*D attempts to modify "
654 					"permanent entry for %s on %s\n",
655 					ifp->if_addrlen, (u_char *)ar_sha(ah),
656 					":", inet_ntoa(isaddr), ifp->if_xname);
657 				}
658 				return;
659 			}
660 		}
661 		/*
662 		 * sanity check for the address length.
663 		 * XXX this does not work for protocols with variable address
664 		 * length. -is
665 		 */
666 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
667 			log(LOG_WARNING,
668 			    "arp from %*D: new addr len %d, was %d",
669 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
670 			    ah->ar_hln, sdl->sdl_alen);
671 		}
672 		if (ifp->if_addrlen != ah->ar_hln) {
673 			if (dologging) {
674 				log(LOG_WARNING,
675 				"arp from %*D: addr len: new %d, i/f %d "
676 				"(ignored)",
677 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
678 				ah->ar_hln, ifp->if_addrlen);
679 			}
680 			return;
681 		}
682 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
683 		if (rt->rt_expire != 0)
684 			rt->rt_expire = time_second + arpt_keep;
685 		rt->rt_flags &= ~RTF_REJECT;
686 		la->la_asked = 0;
687 		la->la_preempt = arp_maxtries;
688 
689 		/*
690 		 * This particular cpu might have been holding an mbuf
691 		 * pending ARP resolution.  If so, transmit the mbuf now.
692 		 */
693 		if (la->la_hold != NULL) {
694 			m_adj(la->la_hold, sizeof(struct ether_header));
695 			ifp->if_output(ifp, la->la_hold, rt_key(rt), rt);
696 			la->la_hold = NULL;
697 		}
698 	}
699 }
700 
701 #ifdef SMP
702 
703 struct netmsg_arp_update {
704 	struct netmsg	netmsg;
705 	struct mbuf	*m;
706 	in_addr_t	saddr;
707 	boolean_t	create;
708 };
709 
710 static void arp_update_msghandler(struct netmsg *);
711 
712 #endif
713 
714 /*
715  * Called from arpintr() - this routine is run from a single cpu.
716  */
717 static void
718 in_arpinput(struct mbuf *m)
719 {
720 	struct arphdr *ah;
721 	struct ifnet *ifp = m->m_pkthdr.rcvif;
722 	struct ether_header *eh;
723 	struct rtentry *rt;
724 	struct ifaddr_container *ifac;
725 	struct in_ifaddr_container *iac;
726 	struct in_ifaddr *ia;
727 	struct sockaddr sa;
728 	struct in_addr isaddr, itaddr, myaddr;
729 #ifdef SMP
730 	struct netmsg_arp_update msg;
731 #endif
732 	u_int8_t *enaddr = NULL;
733 	int op;
734 	int req_len;
735 
736 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
737 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
738 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
739 		return;
740 	}
741 
742 	ah = mtod(m, struct arphdr *);
743 	op = ntohs(ah->ar_op);
744 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
745 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
746 	/*
747 	 * Check both target and sender IP addresses:
748 	 *
749 	 * If we receive the packet on the interface owning the address,
750 	 * then accept the address.
751 	 *
752 	 * For a bridge, we accept the address if the receive interface and
753 	 * the interface owning the address are on the same bridge.
754 	 * (This will change slightly when we have clusters of interfaces).
755 	 */
756 	LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
757 		ia = iac->ia;
758 
759 		/* Skip all ia's which don't match */
760 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
761 			continue;
762 
763 		if (ia->ia_ifp == ifp)
764 			goto match;
765 
766 		if (ifp->if_bridge && ia->ia_ifp &&
767 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
768 			goto match;
769 
770 #ifdef CARP
771 		/*
772 		 * If the interface does not match, but the recieving interface
773 		 * is part of carp, we call carp_iamatch to see if this is a
774 		 * request for the virtual host ip.
775 		 * XXX: This is really ugly!
776 		 */
777 		if (ifp->if_carp != NULL) {
778 			get_mplock();
779 			if (ifp->if_carp != NULL &&
780 			    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr)) {
781 				rel_mplock();
782 				goto match;
783 			}
784 			rel_mplock();
785 		}
786 #endif
787 	}
788 	LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
789 		ia = iac->ia;
790 
791 		/* Skip all ia's which don't match */
792 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
793 			continue;
794 
795 		if (ia->ia_ifp == ifp)
796 			goto match;
797 
798 		if (ifp->if_bridge && ia->ia_ifp &&
799 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
800 			goto match;
801 	}
802 	/*
803 	 * No match, use the first inet address on the receive interface
804 	 * as a dummy address for the rest of the function.
805 	 */
806 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
807 		struct ifaddr *ifa = ifac->ifa;
808 
809 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
810 			ia = ifatoia(ifa);
811 			goto match;
812 		}
813 	}
814 	/*
815 	 * If we got here, we didn't find any suitable interface,
816 	 * so drop the packet.
817 	 */
818 	m_freem(m);
819 	return;
820 
821 match:
822 	if (!enaddr)
823 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
824 	myaddr = ia->ia_addr.sin_addr;
825 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
826 		m_freem(m);	/* it's from me, ignore it. */
827 		return;
828 	}
829 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
830 		log(LOG_ERR,
831 		    "arp: link address is broadcast for IP address %s!\n",
832 		    inet_ntoa(isaddr));
833 		m_freem(m);
834 		return;
835 	}
836 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
837 		log(LOG_ERR,
838 		   "arp: %*D is using my IP address %s!\n",
839 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
840 		   inet_ntoa(isaddr));
841 		itaddr = myaddr;
842 		goto reply;
843 	}
844 #ifdef SMP
845 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
846 		    arp_update_msghandler);
847 	msg.m = m;
848 	msg.saddr = isaddr.s_addr;
849 	msg.create = (itaddr.s_addr == myaddr.s_addr);
850 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
851 #else
852 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
853 			 TRUE);
854 #endif
855 reply:
856 	if (op != ARPOP_REQUEST) {
857 		m_freem(m);
858 		return;
859 	}
860 	if (itaddr.s_addr == myaddr.s_addr) {
861 		/* I am the target */
862 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
863 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
864 	} else {
865 		struct llinfo_arp *la;
866 
867 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
868 		if (la == NULL) {
869 			struct sockaddr_in sin;
870 
871 			if (!arp_proxyall) {
872 				m_freem(m);
873 				return;
874 			}
875 
876 			bzero(&sin, sizeof sin);
877 			sin.sin_family = AF_INET;
878 			sin.sin_len = sizeof sin;
879 			sin.sin_addr = itaddr;
880 
881 			rt = rtpurelookup((struct sockaddr *)&sin);
882 			if (rt == NULL) {
883 				m_freem(m);
884 				return;
885 			}
886 			--rt->rt_refcnt;
887 			/*
888 			 * Don't send proxies for nodes on the same interface
889 			 * as this one came out of, or we'll get into a fight
890 			 * over who claims what Ether address.
891 			 */
892 			if (rt->rt_ifp == ifp) {
893 				m_freem(m);
894 				return;
895 			}
896 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
897 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
898 #ifdef DEBUG_PROXY
899 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
900 #endif
901 		} else {
902 			struct sockaddr_dl *sdl;
903 
904 			rt = la->la_rt;
905 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
906 			sdl = SDL(rt->rt_gateway);
907 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
908 		}
909 	}
910 
911 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
912 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
913 	ah->ar_op = htons(ARPOP_REPLY);
914 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
915 	switch (ifp->if_type) {
916 	case IFT_ETHER:
917 	/*
918 	 * May not be correct for types not explictly
919 	 * listed, but it is our best guess.
920 	 */
921 	default:
922 		eh = (struct ether_header *)sa.sa_data;
923 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
924 		eh->ether_type = htons(ETHERTYPE_ARP);
925 		break;
926 	}
927 	sa.sa_family = AF_UNSPEC;
928 	sa.sa_len = sizeof sa;
929 	ifp->if_output(ifp, m, &sa, NULL);
930 }
931 
932 #ifdef SMP
933 
934 static void
935 arp_update_msghandler(struct netmsg *netmsg)
936 {
937 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
938 	int nextcpu;
939 
940 	arp_update_oncpu(msg->m, msg->saddr, msg->create, mycpuid == 0);
941 
942 	nextcpu = mycpuid + 1;
943 	if (nextcpu < ncpus)
944 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
945 	else
946 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
947 }
948 
949 #endif	/* SMP */
950 
951 #endif	/* INET */
952 
953 /*
954  * Free an arp entry.  If the arp entry is actively referenced or represents
955  * a static entry we only clear it back to an unresolved state, otherwise
956  * we destroy the entry entirely.
957  *
958  * Note that static entries are created when route add ... -interface is used
959  * to create an interface route to a (direct) destination.
960  */
961 static void
962 arptfree(struct llinfo_arp *la)
963 {
964 	struct rtentry *rt = la->la_rt;
965 	struct sockaddr_dl *sdl;
966 
967 	if (rt == NULL)
968 		panic("arptfree");
969 	sdl = SDL(rt->rt_gateway);
970 	if (sdl != NULL &&
971 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
972 	     (rt->rt_flags & RTF_STATIC))) {
973 		sdl->sdl_alen = 0;
974 		la->la_preempt = la->la_asked = 0;
975 		rt->rt_flags &= ~RTF_REJECT;
976 		return;
977 	}
978 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
979 }
980 
981 /*
982  * Lookup or enter a new address in arptab.
983  */
984 static struct llinfo_arp *
985 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
986 {
987 	struct rtentry *rt;
988 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
989 	const char *why = NULL;
990 
991 	sin.sin_addr.s_addr = addr;
992 	sin.sin_other = proxy ? SIN_PROXY : 0;
993 	if (create)
994 		rt = rtlookup((struct sockaddr *)&sin);
995 	else
996 		rt = rtpurelookup((struct sockaddr *)&sin);
997 	if (rt == NULL)
998 		return (NULL);
999 	rt->rt_refcnt--;
1000 
1001 	if (rt->rt_flags & RTF_GATEWAY)
1002 		why = "host is not on local network";
1003 	else if (!(rt->rt_flags & RTF_LLINFO))
1004 		why = "could not allocate llinfo";
1005 	else if (rt->rt_gateway->sa_family != AF_LINK)
1006 		why = "gateway route is not ours";
1007 
1008 	if (why) {
1009 		if (create) {
1010 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1011 			    inet_ntoa(sin.sin_addr), why);
1012 		}
1013 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1014 			/* No references to this route.  Purge it. */
1015 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1016 				  rt_mask(rt), rt->rt_flags, NULL);
1017 		}
1018 		return (NULL);
1019 	}
1020 	return (rt->rt_llinfo);
1021 }
1022 
1023 void
1024 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1025 {
1026 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1027 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1028 				 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1029 	}
1030 	ifa->ifa_rtrequest = arp_rtrequest;
1031 	ifa->ifa_flags |= RTF_CLONING;
1032 }
1033 
1034 void
1035 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
1036 {
1037 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1038 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1039 				 &IA_SIN(ifa)->sin_addr, enaddr);
1040 	}
1041 	ifa->ifa_rtrequest = arp_rtrequest;
1042 	ifa->ifa_flags |= RTF_CLONING;
1043 }
1044 
1045 static void
1046 arp_init(void)
1047 {
1048 	uint32_t flags;
1049 	int cpu;
1050 
1051 	for (cpu = 0; cpu < ncpus2; cpu++)
1052 		LIST_INIT(&llinfo_arp_list[cpu]);
1053 
1054 	if (arp_mpsafe) {
1055 		flags = NETISR_FLAG_MPSAFE;
1056 		kprintf("arp: MPSAFE\n");
1057 	} else {
1058 		flags = NETISR_FLAG_NOTMPSAFE;
1059 	}
1060 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr, flags);
1061 }
1062 
1063 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1064