xref: /dflybsd-src/sys/netinet/if_ether.c (revision c6ddf9d06769c30e7be20c1d090f3f4c6b2919fb)
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 1.56 2008/10/04 11:26:09 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 int	arp_mpsafe = 0;
150 TUNABLE_INT("net.link.ether.inet.arp_mpsafe", &arp_mpsafe);
151 
152 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
153 static void	arprequest(struct ifnet *, struct in_addr *, struct in_addr *,
154 			   const u_char *);
155 static void	arprequest_async(struct ifnet *, struct in_addr *,
156 				 struct in_addr *, const u_char *);
157 static void	arpintr(struct netmsg *);
158 static void	arptfree(struct llinfo_arp *);
159 static void	arptimer(void *);
160 static struct llinfo_arp *
161 		arplookup(in_addr_t, boolean_t, boolean_t);
162 #ifdef INET
163 static void	in_arpinput(struct mbuf *);
164 #endif
165 
166 static struct callout	arptimer_ch[MAXCPU];
167 
168 /*
169  * Timeout routine.  Age arp_tab entries periodically.
170  */
171 /* ARGSUSED */
172 static void
173 arptimer(void *ignored_arg)
174 {
175 	struct llinfo_arp *la, *nla;
176 
177 	crit_enter();
178 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
179 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
180 			arptfree(la);
181 	}
182 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
183 	crit_exit();
184 }
185 
186 /*
187  * Parallel to llc_rtrequest.
188  */
189 static void
190 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
191 {
192 	struct sockaddr *gate = rt->rt_gateway;
193 	struct llinfo_arp *la = rt->rt_llinfo;
194 
195 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
196 	static boolean_t arpinit_done[MAXCPU];
197 
198 	if (!arpinit_done[mycpuid]) {
199 		arpinit_done[mycpuid] = TRUE;
200 		callout_init(&arptimer_ch[mycpuid]);
201 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
202 	}
203 	if (rt->rt_flags & RTF_GATEWAY)
204 		return;
205 
206 	switch (req) {
207 	case RTM_ADD:
208 		/*
209 		 * XXX: If this is a manually added route to interface
210 		 * such as older version of routed or gated might provide,
211 		 * restore cloning bit.
212 		 */
213 		if (!(rt->rt_flags & RTF_HOST) &&
214 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
215 			rt->rt_flags |= RTF_CLONING;
216 		if (rt->rt_flags & RTF_CLONING) {
217 			/*
218 			 * Case 1: This route should come from a route to iface.
219 			 */
220 			rt_setgate(rt, rt_key(rt),
221 				   (struct sockaddr *)&null_sdl);
222 			gate = rt->rt_gateway;
223 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
224 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
225 			rt->rt_expire = time_second;
226 			break;
227 		}
228 		/* Announce a new entry if requested. */
229 		if (rt->rt_flags & RTF_ANNOUNCE) {
230 			arprequest_async(rt->rt_ifp,
231 			    &SIN(rt_key(rt))->sin_addr,
232 			    &SIN(rt_key(rt))->sin_addr,
233 			    LLADDR(SDL(gate)));
234 		}
235 		/*FALLTHROUGH*/
236 	case RTM_RESOLVE:
237 		if (gate->sa_family != AF_LINK ||
238 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
239 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
240 			break;
241 		}
242 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
243 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
244 		if (la != NULL)
245 			break; /* This happens on a route change */
246 		/*
247 		 * Case 2:  This route may come from cloning, or a manual route
248 		 * add with a LL address.
249 		 */
250 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
251 		rt->rt_llinfo = la;
252 		if (la == NULL) {
253 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
254 			break;
255 		}
256 		bzero(la, sizeof *la);
257 		la->la_rt = rt;
258 		rt->rt_flags |= RTF_LLINFO;
259 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
260 
261 #ifdef INET
262 		/*
263 		 * This keeps the multicast addresses from showing up
264 		 * in `arp -a' listings as unresolved.  It's not actually
265 		 * functional.  Then the same for broadcast.
266 		 */
267 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
268 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
269 					       LLADDR(SDL(gate)));
270 			SDL(gate)->sdl_alen = 6;
271 			rt->rt_expire = 0;
272 		}
273 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
274 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
275 			       rt->rt_ifp->if_addrlen);
276 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
277 			rt->rt_expire = 0;
278 		}
279 #endif
280 
281 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
282 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
283 			/*
284 			 * This test used to be
285 			 *	if (loif.if_flags & IFF_UP)
286 			 * It allowed local traffic to be forced
287 			 * through the hardware by configuring the
288 			 * loopback down.  However, it causes problems
289 			 * during network configuration for boards
290 			 * that can't receive packets they send.  It
291 			 * is now necessary to clear "useloopback" and
292 			 * remove the route to force traffic out to
293 			 * the hardware.
294 			 */
295 			rt->rt_expire = 0;
296 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
297 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
298 			if (useloopback)
299 				rt->rt_ifp = loif;
300 		}
301 		break;
302 
303 	case RTM_DELETE:
304 		if (la == NULL)
305 			break;
306 		LIST_REMOVE(la, la_le);
307 		rt->rt_llinfo = NULL;
308 		rt->rt_flags &= ~RTF_LLINFO;
309 		if (la->la_hold != NULL)
310 			m_freem(la->la_hold);
311 		Free(la);
312 		break;
313 	}
314 }
315 
316 static struct mbuf *
317 arpreq_alloc(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
318 	     const u_char *enaddr)
319 {
320 	struct mbuf *m;
321 	struct arphdr *ah;
322 	u_short ar_hrd;
323 
324 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
325 		return NULL;
326 	m->m_pkthdr.rcvif = NULL;
327 
328 	switch (ifp->if_type) {
329 	case IFT_ETHER:
330 		/*
331 		 * This may not be correct for types not explicitly
332 		 * listed, but this is our best guess
333 		 */
334 	default:
335 		ar_hrd = htons(ARPHRD_ETHER);
336 
337 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
338 		m->m_pkthdr.len = m->m_len;
339 		MH_ALIGN(m, m->m_len);
340 
341 		ah = mtod(m, struct arphdr *);
342 		break;
343 	}
344 
345 	ah->ar_hrd = ar_hrd;
346 	ah->ar_pro = htons(ETHERTYPE_IP);
347 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
348 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
349 	ah->ar_op = htons(ARPOP_REQUEST);
350 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
351 	memset(ar_tha(ah), 0, ah->ar_hln);
352 	memcpy(ar_spa(ah), sip, ah->ar_pln);
353 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
354 
355 	return m;
356 }
357 
358 static void
359 arpreq_send(struct ifnet *ifp, struct mbuf *m)
360 {
361 	struct sockaddr sa;
362 	struct ether_header *eh;
363 
364 	switch (ifp->if_type) {
365 	case IFT_ETHER:
366 		/*
367 		 * This may not be correct for types not explicitly
368 		 * listed, but this is our best guess
369 		 */
370 	default:
371 		eh = (struct ether_header *)sa.sa_data;
372 		/* if_output() will not swap */
373 		eh->ether_type = htons(ETHERTYPE_ARP);
374 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
375 		break;
376 	}
377 
378 	sa.sa_family = AF_UNSPEC;
379 	sa.sa_len = sizeof(sa);
380 	ifp->if_output(ifp, m, &sa, NULL);
381 }
382 
383 static void
384 arpreq_send_handler(struct netmsg *nmsg)
385 {
386 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
387 	struct ifnet *ifp = nmsg->nm_lmsg.u.ms_resultp;
388 
389 	arpreq_send(ifp, m);
390 	/* nmsg was embedded in the mbuf, do not reply! */
391 }
392 
393 /*
394  * Broadcast an ARP request. Caller specifies:
395  *	- arp header source ip address
396  *	- arp header target ip address
397  *	- arp header source ethernet address
398  *
399  * NOTE: Caller MUST NOT hold ifp's serializer
400  */
401 static void
402 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
403 	   const u_char *enaddr)
404 {
405 	struct mbuf *m;
406 
407 	m = arpreq_alloc(ifp, sip, tip, enaddr);
408 	if (m == NULL)
409 		return;
410 	arpreq_send(ifp, m);
411 }
412 
413 /*
414  * Same as arprequest(), except:
415  * - Caller is allowed to hold ifp's serializer
416  * - Network output is done in TDF_NETWORK kernel thread
417  */
418 static void
419 arprequest_async(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
420 		 const u_char *enaddr)
421 {
422 	struct mbuf *m;
423 	struct netmsg_packet *pmsg;
424 
425 	m = arpreq_alloc(ifp, sip, tip, enaddr);
426 	if (m == NULL)
427 		return;
428 
429 	pmsg = &m->m_hdr.mh_netmsg;
430 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
431 		    arpreq_send_handler);
432 	pmsg->nm_packet = m;
433 	pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = ifp;
434 
435 	lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->nm_netmsg.nm_lmsg);
436 }
437 
438 /*
439  * Resolve an IP address into an ethernet address.  If success,
440  * desten is filled in.  If there is no entry in arptab,
441  * set one up and broadcast a request for the IP address.
442  * Hold onto this mbuf and resend it once the address
443  * is finally resolved.  A return value of 1 indicates
444  * that desten has been filled in and the packet should be sent
445  * normally; a 0 return indicates that the packet has been
446  * taken over here, either now or for later transmission.
447  */
448 int
449 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
450 	   struct sockaddr *dst, u_char *desten)
451 {
452 	struct rtentry *rt;
453 	struct llinfo_arp *la = NULL;
454 	struct sockaddr_dl *sdl;
455 
456 	if (m->m_flags & M_BCAST) {	/* broadcast */
457 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
458 		return (1);
459 	}
460 	if (m->m_flags & M_MCAST) {/* multicast */
461 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
462 		return (1);
463 	}
464 	if (rt0 != NULL) {
465 		if (rt_llroute(dst, rt0, &rt) != 0) {
466 			m_freem(m);
467 			return 0;
468 		}
469 		la = rt->rt_llinfo;
470 	}
471 	if (la == NULL) {
472 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
473 		if (la != NULL)
474 			rt = la->la_rt;
475 	}
476 	if (la == NULL || rt == NULL) {
477 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
478 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
479 		    rt ? "rt" : "");
480 		m_freem(m);
481 		return (0);
482 	}
483 	sdl = SDL(rt->rt_gateway);
484 	/*
485 	 * Check the address family and length is valid, the address
486 	 * is resolved; otherwise, try to resolve.
487 	 */
488 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
489 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
490 		/*
491 		 * If entry has an expiry time and it is approaching,
492 		 * see if we need to send an ARP request within this
493 		 * arpt_down interval.
494 		 */
495 		if ((rt->rt_expire != 0) &&
496 		    (time_second + la->la_preempt > rt->rt_expire)) {
497 			arprequest(ifp,
498 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
499 				   &SIN(dst)->sin_addr,
500 				   IF_LLADDR(ifp));
501 			la->la_preempt--;
502 		}
503 
504 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
505 		return 1;
506 	}
507 	/*
508 	 * If ARP is disabled on this interface, stop.
509 	 * XXX
510 	 * Probably should not allocate empty llinfo struct if we are
511 	 * not going to be sending out an arp request.
512 	 */
513 	if (ifp->if_flags & IFF_NOARP) {
514 		m_freem(m);
515 		return (0);
516 	}
517 	/*
518 	 * There is an arptab entry, but no ethernet address
519 	 * response yet.  Replace the held mbuf with this
520 	 * latest one.
521 	 */
522 	if (la->la_hold != NULL)
523 		m_freem(la->la_hold);
524 	la->la_hold = m;
525 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
526 		rt->rt_flags &= ~RTF_REJECT;
527 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
528 			rt->rt_expire = time_second;
529 			if (la->la_asked++ < arp_maxtries) {
530 				arprequest(ifp,
531 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
532 					   &SIN(dst)->sin_addr,
533 					   IF_LLADDR(ifp));
534 			} else {
535 				rt->rt_flags |= RTF_REJECT;
536 				rt->rt_expire += arpt_down;
537 				la->la_asked = 0;
538 				la->la_preempt = arp_maxtries;
539 			}
540 		}
541 	}
542 	return (0);
543 }
544 
545 /*
546  * Common length and type checks are done here,
547  * then the protocol-specific routine is called.
548  */
549 static void
550 arpintr(struct netmsg *msg)
551 {
552 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
553 	struct arphdr *ar;
554 	u_short ar_hrd;
555 
556 	if (m->m_len < sizeof(struct arphdr) &&
557 	    (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
558 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
559 		return;
560 	}
561 	ar = mtod(m, struct arphdr *);
562 
563 	ar_hrd = ntohs(ar->ar_hrd);
564 	if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
565 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
566 		    (unsigned char *)&ar->ar_hrd, "");
567 		m_freem(m);
568 		return;
569 	}
570 
571 	if (m->m_pkthdr.len < arphdr_len(ar)) {
572 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
573 			log(LOG_ERR, "arp: runt packet\n");
574 			return;
575 		}
576 		ar = mtod(m, struct arphdr *);
577 	}
578 
579 	switch (ntohs(ar->ar_pro)) {
580 #ifdef INET
581 	case ETHERTYPE_IP:
582 		in_arpinput(m);
583 		return;
584 #endif
585 	}
586 	m_freem(m);
587 	/* msg was embedded in the mbuf, do not reply! */
588 }
589 
590 #ifdef INET
591 /*
592  * ARP for Internet protocols on 10 Mb/s Ethernet.
593  * Algorithm is that given in RFC 826.
594  * In addition, a sanity check is performed on the sender
595  * protocol address, to catch impersonators.
596  * We no longer handle negotiations for use of trailer protocol:
597  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
598  * along with IP replies if we wanted trailers sent to us,
599  * and also sent them in response to IP replies.
600  * This allowed either end to announce the desire to receive
601  * trailer packets.
602  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
603  * but formerly didn't normally send requests.
604  */
605 static int log_arp_wrong_iface = 1;
606 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
607 	   &log_arp_wrong_iface, 0,
608 	   "log arp packets arriving on the wrong interface");
609 
610 static void
611 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
612 		 boolean_t dologging)
613 {
614 	struct arphdr *ah = mtod(m, struct arphdr *);
615 	struct ifnet *ifp = m->m_pkthdr.rcvif;
616 	struct llinfo_arp *la;
617 	struct sockaddr_dl *sdl;
618 	struct rtentry *rt;
619 
620 	la = arplookup(saddr, create, FALSE);
621 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
622 		struct in_addr isaddr = { saddr };
623 
624 		/* the following is not an error when doing bridging */
625 		if (rt->rt_ifp != ifp) {
626 			if (dologging && log_arp_wrong_iface) {
627 				log(LOG_ERR,
628 				    "arp: %s is on %s "
629 				    "but got reply from %*D on %s\n",
630 				    inet_ntoa(isaddr),
631 				    rt->rt_ifp->if_xname,
632 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
633 				    ifp->if_xname);
634 			}
635 			return;
636 		}
637 		if (sdl->sdl_alen &&
638 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
639 			if (rt->rt_expire != 0) {
640 				if (dologging) {
641 			    		log(LOG_INFO,
642 			    		"arp: %s moved from %*D to %*D on %s\n",
643 			    		inet_ntoa(isaddr),
644 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
645 			    		":", ifp->if_addrlen,
646 			    		(u_char *)ar_sha(ah), ":",
647 			    		ifp->if_xname);
648 				}
649 			} else {
650 				if (dologging) {
651 					log(LOG_ERR,
652 					"arp: %*D attempts to modify "
653 					"permanent entry for %s on %s\n",
654 					ifp->if_addrlen, (u_char *)ar_sha(ah),
655 					":", inet_ntoa(isaddr), ifp->if_xname);
656 				}
657 				return;
658 			}
659 		}
660 		/*
661 		 * sanity check for the address length.
662 		 * XXX this does not work for protocols with variable address
663 		 * length. -is
664 		 */
665 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
666 			log(LOG_WARNING,
667 			    "arp from %*D: new addr len %d, was %d",
668 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
669 			    ah->ar_hln, sdl->sdl_alen);
670 		}
671 		if (ifp->if_addrlen != ah->ar_hln) {
672 			if (dologging) {
673 				log(LOG_WARNING,
674 				"arp from %*D: addr len: new %d, i/f %d "
675 				"(ignored)",
676 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
677 				ah->ar_hln, ifp->if_addrlen);
678 			}
679 			return;
680 		}
681 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
682 		if (rt->rt_expire != 0)
683 			rt->rt_expire = time_second + arpt_keep;
684 		rt->rt_flags &= ~RTF_REJECT;
685 		la->la_asked = 0;
686 		la->la_preempt = arp_maxtries;
687 
688 		/*
689 		 * This particular cpu might have been holding an mbuf
690 		 * pending ARP resolution.  If so, transmit the mbuf now.
691 		 */
692 		if (la->la_hold != NULL) {
693 			m_adj(la->la_hold, sizeof(struct ether_header));
694 			ifp->if_output(ifp, la->la_hold, rt_key(rt), rt);
695 			la->la_hold = NULL;
696 		}
697 	}
698 }
699 
700 #ifdef SMP
701 
702 struct netmsg_arp_update {
703 	struct netmsg	netmsg;
704 	struct mbuf	*m;
705 	in_addr_t	saddr;
706 	boolean_t	create;
707 };
708 
709 static void arp_update_msghandler(struct netmsg *);
710 
711 #endif
712 
713 /*
714  * Called from arpintr() - this routine is run from a single cpu.
715  */
716 static void
717 in_arpinput(struct mbuf *m)
718 {
719 	struct arphdr *ah;
720 	struct ifnet *ifp = m->m_pkthdr.rcvif;
721 	struct ether_header *eh;
722 	struct rtentry *rt;
723 	struct ifaddr_container *ifac;
724 	struct in_ifaddr_container *iac;
725 	struct in_ifaddr *ia;
726 	struct sockaddr sa;
727 	struct in_addr isaddr, itaddr, myaddr;
728 #ifdef SMP
729 	struct netmsg_arp_update msg;
730 #endif
731 	u_int8_t *enaddr = NULL;
732 	int op;
733 	int req_len;
734 
735 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
736 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
737 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
738 		return;
739 	}
740 
741 	ah = mtod(m, struct arphdr *);
742 	op = ntohs(ah->ar_op);
743 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
744 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
745 	/*
746 	 * Check both target and sender IP addresses:
747 	 *
748 	 * If we receive the packet on the interface owning the address,
749 	 * then accept the address.
750 	 *
751 	 * For a bridge, we accept the address if the receive interface and
752 	 * the interface owning the address are on the same bridge.
753 	 * (This will change slightly when we have clusters of interfaces).
754 	 */
755 	LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
756 		ia = iac->ia;
757 
758 		/* Skip all ia's which don't match */
759 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
760 			continue;
761 
762 		if (ia->ia_ifp == ifp)
763 			goto match;
764 
765 		if (ifp->if_bridge && ia->ia_ifp &&
766 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
767 			goto match;
768 
769 #ifdef CARP
770 		/*
771 		 * If the interface does not match, but the recieving interface
772 		 * is part of carp, we call carp_iamatch to see if this is a
773 		 * request for the virtual host ip.
774 		 * XXX: This is really ugly!
775 		 */
776 		if (ifp->if_carp != NULL) {
777 			get_mplock();
778 			if (ifp->if_carp != NULL &&
779 			    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr)) {
780 				rel_mplock();
781 				goto match;
782 			}
783 			rel_mplock();
784 		}
785 #endif
786 	}
787 	LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
788 		ia = iac->ia;
789 
790 		/* Skip all ia's which don't match */
791 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
792 			continue;
793 
794 		if (ia->ia_ifp == ifp)
795 			goto match;
796 
797 		if (ifp->if_bridge && ia->ia_ifp &&
798 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
799 			goto match;
800 	}
801 	/*
802 	 * No match, use the first inet address on the receive interface
803 	 * as a dummy address for the rest of the function.
804 	 */
805 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
806 		struct ifaddr *ifa = ifac->ifa;
807 
808 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
809 			ia = ifatoia(ifa);
810 			goto match;
811 		}
812 	}
813 	/*
814 	 * If we got here, we didn't find any suitable interface,
815 	 * so drop the packet.
816 	 */
817 	m_freem(m);
818 	return;
819 
820 match:
821 	if (!enaddr)
822 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
823 	myaddr = ia->ia_addr.sin_addr;
824 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
825 		m_freem(m);	/* it's from me, ignore it. */
826 		return;
827 	}
828 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
829 		log(LOG_ERR,
830 		    "arp: link address is broadcast for IP address %s!\n",
831 		    inet_ntoa(isaddr));
832 		m_freem(m);
833 		return;
834 	}
835 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
836 		log(LOG_ERR,
837 		   "arp: %*D is using my IP address %s!\n",
838 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
839 		   inet_ntoa(isaddr));
840 		itaddr = myaddr;
841 		goto reply;
842 	}
843 #ifdef SMP
844 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
845 		    arp_update_msghandler);
846 	msg.m = m;
847 	msg.saddr = isaddr.s_addr;
848 	msg.create = (itaddr.s_addr == myaddr.s_addr);
849 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
850 #else
851 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
852 			 TRUE);
853 #endif
854 reply:
855 	if (op != ARPOP_REQUEST) {
856 		m_freem(m);
857 		return;
858 	}
859 	if (itaddr.s_addr == myaddr.s_addr) {
860 		/* I am the target */
861 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
862 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
863 	} else {
864 		struct llinfo_arp *la;
865 
866 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
867 		if (la == NULL) {
868 			struct sockaddr_in sin;
869 
870 			if (!arp_proxyall) {
871 				m_freem(m);
872 				return;
873 			}
874 
875 			bzero(&sin, sizeof sin);
876 			sin.sin_family = AF_INET;
877 			sin.sin_len = sizeof sin;
878 			sin.sin_addr = itaddr;
879 
880 			rt = rtpurelookup((struct sockaddr *)&sin);
881 			if (rt == NULL) {
882 				m_freem(m);
883 				return;
884 			}
885 			--rt->rt_refcnt;
886 			/*
887 			 * Don't send proxies for nodes on the same interface
888 			 * as this one came out of, or we'll get into a fight
889 			 * over who claims what Ether address.
890 			 */
891 			if (rt->rt_ifp == ifp) {
892 				m_freem(m);
893 				return;
894 			}
895 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
896 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
897 #ifdef DEBUG_PROXY
898 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
899 #endif
900 		} else {
901 			struct sockaddr_dl *sdl;
902 
903 			rt = la->la_rt;
904 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
905 			sdl = SDL(rt->rt_gateway);
906 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
907 		}
908 	}
909 
910 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
911 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
912 	ah->ar_op = htons(ARPOP_REPLY);
913 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
914 	switch (ifp->if_type) {
915 	case IFT_ETHER:
916 	/*
917 	 * May not be correct for types not explictly
918 	 * listed, but it is our best guess.
919 	 */
920 	default:
921 		eh = (struct ether_header *)sa.sa_data;
922 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
923 		eh->ether_type = htons(ETHERTYPE_ARP);
924 		break;
925 	}
926 	sa.sa_family = AF_UNSPEC;
927 	sa.sa_len = sizeof sa;
928 	ifp->if_output(ifp, m, &sa, NULL);
929 }
930 
931 #ifdef SMP
932 
933 static void
934 arp_update_msghandler(struct netmsg *netmsg)
935 {
936 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
937 	int nextcpu;
938 
939 	arp_update_oncpu(msg->m, msg->saddr, msg->create, mycpuid == 0);
940 
941 	nextcpu = mycpuid + 1;
942 	if (nextcpu < ncpus)
943 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
944 	else
945 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
946 }
947 
948 #endif	/* SMP */
949 
950 #endif	/* INET */
951 
952 /*
953  * Free an arp entry.  If the arp entry is actively referenced or represents
954  * a static entry we only clear it back to an unresolved state, otherwise
955  * we destroy the entry entirely.
956  *
957  * Note that static entries are created when route add ... -interface is used
958  * to create an interface route to a (direct) destination.
959  */
960 static void
961 arptfree(struct llinfo_arp *la)
962 {
963 	struct rtentry *rt = la->la_rt;
964 	struct sockaddr_dl *sdl;
965 
966 	if (rt == NULL)
967 		panic("arptfree");
968 	sdl = SDL(rt->rt_gateway);
969 	if (sdl != NULL &&
970 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
971 	     (rt->rt_flags & RTF_STATIC))) {
972 		sdl->sdl_alen = 0;
973 		la->la_preempt = la->la_asked = 0;
974 		rt->rt_flags &= ~RTF_REJECT;
975 		return;
976 	}
977 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
978 }
979 
980 /*
981  * Lookup or enter a new address in arptab.
982  */
983 static struct llinfo_arp *
984 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
985 {
986 	struct rtentry *rt;
987 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
988 	const char *why = NULL;
989 
990 	sin.sin_addr.s_addr = addr;
991 	sin.sin_other = proxy ? SIN_PROXY : 0;
992 	if (create)
993 		rt = rtlookup((struct sockaddr *)&sin);
994 	else
995 		rt = rtpurelookup((struct sockaddr *)&sin);
996 	if (rt == NULL)
997 		return (NULL);
998 	rt->rt_refcnt--;
999 
1000 	if (rt->rt_flags & RTF_GATEWAY)
1001 		why = "host is not on local network";
1002 	else if (!(rt->rt_flags & RTF_LLINFO))
1003 		why = "could not allocate llinfo";
1004 	else if (rt->rt_gateway->sa_family != AF_LINK)
1005 		why = "gateway route is not ours";
1006 
1007 	if (why) {
1008 		if (create) {
1009 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1010 			    inet_ntoa(sin.sin_addr), why);
1011 		}
1012 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1013 			/* No references to this route.  Purge it. */
1014 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1015 				  rt_mask(rt), rt->rt_flags, NULL);
1016 		}
1017 		return (NULL);
1018 	}
1019 	return (rt->rt_llinfo);
1020 }
1021 
1022 void
1023 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1024 {
1025 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1026 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1027 				 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1028 	}
1029 	ifa->ifa_rtrequest = arp_rtrequest;
1030 	ifa->ifa_flags |= RTF_CLONING;
1031 }
1032 
1033 void
1034 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
1035 {
1036 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1037 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1038 				 &IA_SIN(ifa)->sin_addr, enaddr);
1039 	}
1040 	ifa->ifa_rtrequest = arp_rtrequest;
1041 	ifa->ifa_flags |= RTF_CLONING;
1042 }
1043 
1044 static void
1045 arp_init(void)
1046 {
1047 	uint32_t flags;
1048 	int cpu;
1049 
1050 	for (cpu = 0; cpu < ncpus2; cpu++)
1051 		LIST_INIT(&llinfo_arp_list[cpu]);
1052 
1053 	if (arp_mpsafe) {
1054 		flags = NETISR_FLAG_MPSAFE;
1055 		kprintf("arp: MPSAFE\n");
1056 	} else {
1057 		flags = NETISR_FLAG_NOTMPSAFE;
1058 	}
1059 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr, flags);
1060 }
1061 
1062 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1063