xref: /dflybsd-src/sys/netinet/if_ether.c (revision 10cf3cb7c694a65a355fa433dae9cbace95a62a2)
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 1.59 2008/11/22 11:03:35 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 #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 	struct	lwkt_port *la_msgport; /* last packet's msgport */
134 	u_short	la_preempt;	/* countdown for pre-expiry arps */
135 	u_short	la_asked;	/* #times we QUERIED following expiration */
136 };
137 
138 static	LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
139 
140 static int	arp_maxtries = 5;
141 static int	useloopback = 1; /* use loopback interface for local traffic */
142 static int	arp_proxyall = 0;
143 
144 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
145 	   &arp_maxtries, 0, "");
146 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
147 	   &useloopback, 0, "");
148 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
149 	   &arp_proxyall, 0, "");
150 
151 static int	arp_mpsafe = 0;
152 TUNABLE_INT("net.link.ether.inet.arp_mpsafe", &arp_mpsafe);
153 
154 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
155 static void	arprequest(struct ifnet *, const struct in_addr *,
156 			   const struct in_addr *, const u_char *);
157 static void	arprequest_async(struct ifnet *, const struct in_addr *,
158 				 const struct in_addr *, const u_char *);
159 static void	arpintr(struct netmsg *);
160 static void	arptfree(struct llinfo_arp *);
161 static void	arptimer(void *);
162 static struct llinfo_arp *
163 		arplookup(in_addr_t, boolean_t, boolean_t);
164 #ifdef INET
165 static void	in_arpinput(struct mbuf *);
166 #endif
167 
168 static struct callout	arptimer_ch[MAXCPU];
169 
170 /*
171  * Timeout routine.  Age arp_tab entries periodically.
172  */
173 /* ARGSUSED */
174 static void
175 arptimer(void *ignored_arg)
176 {
177 	struct llinfo_arp *la, *nla;
178 
179 	crit_enter();
180 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
181 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
182 			arptfree(la);
183 	}
184 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
185 	crit_exit();
186 }
187 
188 /*
189  * Parallel to llc_rtrequest.
190  */
191 static void
192 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
193 {
194 	struct sockaddr *gate = rt->rt_gateway;
195 	struct llinfo_arp *la = rt->rt_llinfo;
196 
197 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
198 	static boolean_t arpinit_done[MAXCPU];
199 
200 	if (!arpinit_done[mycpuid]) {
201 		arpinit_done[mycpuid] = TRUE;
202 		callout_init(&arptimer_ch[mycpuid]);
203 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
204 	}
205 	if (rt->rt_flags & RTF_GATEWAY)
206 		return;
207 
208 	switch (req) {
209 	case RTM_ADD:
210 		/*
211 		 * XXX: If this is a manually added route to interface
212 		 * such as older version of routed or gated might provide,
213 		 * restore cloning bit.
214 		 */
215 		if (!(rt->rt_flags & RTF_HOST) &&
216 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
217 			rt->rt_flags |= RTF_CLONING;
218 		if (rt->rt_flags & RTF_CLONING) {
219 			/*
220 			 * Case 1: This route should come from a route to iface.
221 			 */
222 			rt_setgate(rt, rt_key(rt),
223 				   (struct sockaddr *)&null_sdl);
224 			gate = rt->rt_gateway;
225 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
226 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
227 			rt->rt_expire = time_second;
228 			break;
229 		}
230 		/* Announce a new entry if requested. */
231 		if (rt->rt_flags & RTF_ANNOUNCE) {
232 			arprequest_async(rt->rt_ifp,
233 			    &SIN(rt_key(rt))->sin_addr,
234 			    &SIN(rt_key(rt))->sin_addr,
235 			    LLADDR(SDL(gate)));
236 		}
237 		/*FALLTHROUGH*/
238 	case RTM_RESOLVE:
239 		if (gate->sa_family != AF_LINK ||
240 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
241 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
242 			break;
243 		}
244 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
245 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
246 		if (la != NULL)
247 			break; /* This happens on a route change */
248 		/*
249 		 * Case 2:  This route may come from cloning, or a manual route
250 		 * add with a LL address.
251 		 */
252 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
253 		rt->rt_llinfo = la;
254 		if (la == NULL) {
255 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
256 			break;
257 		}
258 		bzero(la, sizeof *la);
259 		la->la_rt = rt;
260 		rt->rt_flags |= RTF_LLINFO;
261 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
262 
263 #ifdef INET
264 		/*
265 		 * This keeps the multicast addresses from showing up
266 		 * in `arp -a' listings as unresolved.  It's not actually
267 		 * functional.  Then the same for broadcast.
268 		 */
269 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
270 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
271 					       LLADDR(SDL(gate)));
272 			SDL(gate)->sdl_alen = 6;
273 			rt->rt_expire = 0;
274 		}
275 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
276 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
277 			       rt->rt_ifp->if_addrlen);
278 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
279 			rt->rt_expire = 0;
280 		}
281 #endif
282 
283 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
284 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
285 			/*
286 			 * This test used to be
287 			 *	if (loif.if_flags & IFF_UP)
288 			 * It allowed local traffic to be forced
289 			 * through the hardware by configuring the
290 			 * loopback down.  However, it causes problems
291 			 * during network configuration for boards
292 			 * that can't receive packets they send.  It
293 			 * is now necessary to clear "useloopback" and
294 			 * remove the route to force traffic out to
295 			 * the hardware.
296 			 */
297 			rt->rt_expire = 0;
298 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
299 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
300 			if (useloopback)
301 				rt->rt_ifp = loif;
302 		}
303 		break;
304 
305 	case RTM_DELETE:
306 		if (la == NULL)
307 			break;
308 		LIST_REMOVE(la, la_le);
309 		rt->rt_llinfo = NULL;
310 		rt->rt_flags &= ~RTF_LLINFO;
311 		if (la->la_hold != NULL)
312 			m_freem(la->la_hold);
313 		Free(la);
314 		break;
315 	}
316 }
317 
318 static struct mbuf *
319 arpreq_alloc(struct ifnet *ifp, const struct in_addr *sip,
320 	     const struct in_addr *tip, const u_char *enaddr)
321 {
322 	struct mbuf *m;
323 	struct arphdr *ah;
324 	u_short ar_hrd;
325 
326 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
327 		return NULL;
328 	m->m_pkthdr.rcvif = NULL;
329 
330 	switch (ifp->if_type) {
331 	case IFT_ETHER:
332 		/*
333 		 * This may not be correct for types not explicitly
334 		 * listed, but this is our best guess
335 		 */
336 	default:
337 		ar_hrd = htons(ARPHRD_ETHER);
338 
339 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
340 		m->m_pkthdr.len = m->m_len;
341 		MH_ALIGN(m, m->m_len);
342 
343 		ah = mtod(m, struct arphdr *);
344 		break;
345 	}
346 
347 	ah->ar_hrd = ar_hrd;
348 	ah->ar_pro = htons(ETHERTYPE_IP);
349 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
350 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
351 	ah->ar_op = htons(ARPOP_REQUEST);
352 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
353 	memset(ar_tha(ah), 0, ah->ar_hln);
354 	memcpy(ar_spa(ah), sip, ah->ar_pln);
355 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
356 
357 	return m;
358 }
359 
360 static void
361 arpreq_send(struct ifnet *ifp, struct mbuf *m)
362 {
363 	struct sockaddr sa;
364 	struct ether_header *eh;
365 
366 	switch (ifp->if_type) {
367 	case IFT_ETHER:
368 		/*
369 		 * This may not be correct for types not explicitly
370 		 * listed, but this is our best guess
371 		 */
372 	default:
373 		eh = (struct ether_header *)sa.sa_data;
374 		/* if_output() will not swap */
375 		eh->ether_type = htons(ETHERTYPE_ARP);
376 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
377 		break;
378 	}
379 
380 	sa.sa_family = AF_UNSPEC;
381 	sa.sa_len = sizeof(sa);
382 	ifp->if_output(ifp, m, &sa, NULL);
383 }
384 
385 static void
386 arpreq_send_handler(struct netmsg *nmsg)
387 {
388 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
389 	struct ifnet *ifp = nmsg->nm_lmsg.u.ms_resultp;
390 
391 	arpreq_send(ifp, m);
392 	/* nmsg was embedded in the mbuf, do not reply! */
393 }
394 
395 /*
396  * Broadcast an ARP request. Caller specifies:
397  *	- arp header source ip address
398  *	- arp header target ip address
399  *	- arp header source ethernet address
400  *
401  * NOTE: Caller MUST NOT hold ifp's serializer
402  */
403 static void
404 arprequest(struct ifnet *ifp, const struct in_addr *sip,
405 	   const struct in_addr *tip, const u_char *enaddr)
406 {
407 	struct mbuf *m;
408 
409 	m = arpreq_alloc(ifp, sip, tip, enaddr);
410 	if (m == NULL)
411 		return;
412 	arpreq_send(ifp, m);
413 }
414 
415 /*
416  * Same as arprequest(), except:
417  * - Caller is allowed to hold ifp's serializer
418  * - Network output is done in TDF_NETWORK kernel thread
419  */
420 static void
421 arprequest_async(struct ifnet *ifp, const struct in_addr *sip,
422 		 const struct in_addr *tip, const u_char *enaddr)
423 {
424 	struct mbuf *m;
425 	struct netmsg_packet *pmsg;
426 
427 	m = arpreq_alloc(ifp, sip, tip, enaddr);
428 	if (m == NULL)
429 		return;
430 
431 	pmsg = &m->m_hdr.mh_netmsg;
432 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
433 		    arpreq_send_handler);
434 	pmsg->nm_packet = m;
435 	pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = ifp;
436 
437 	lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->nm_netmsg.nm_lmsg);
438 }
439 
440 /*
441  * Resolve an IP address into an ethernet address.  If success,
442  * desten is filled in.  If there is no entry in arptab,
443  * set one up and broadcast a request for the IP address.
444  * Hold onto this mbuf and resend it once the address
445  * is finally resolved.  A return value of 1 indicates
446  * that desten has been filled in and the packet should be sent
447  * normally; a 0 return indicates that the packet has been
448  * taken over here, either now or for later transmission.
449  */
450 int
451 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
452 	   struct sockaddr *dst, u_char *desten)
453 {
454 	struct rtentry *rt;
455 	struct llinfo_arp *la = NULL;
456 	struct sockaddr_dl *sdl;
457 
458 	if (m->m_flags & M_BCAST) {	/* broadcast */
459 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
460 		return (1);
461 	}
462 	if (m->m_flags & M_MCAST) {/* multicast */
463 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
464 		return (1);
465 	}
466 	if (rt0 != NULL) {
467 		if (rt_llroute(dst, rt0, &rt) != 0) {
468 			m_freem(m);
469 			return 0;
470 		}
471 		la = rt->rt_llinfo;
472 	}
473 	if (la == NULL) {
474 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
475 		if (la != NULL)
476 			rt = la->la_rt;
477 	}
478 	if (la == NULL || rt == NULL) {
479 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
480 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
481 		    rt ? "rt" : "");
482 		m_freem(m);
483 		return (0);
484 	}
485 	sdl = SDL(rt->rt_gateway);
486 	/*
487 	 * Check the address family and length is valid, the address
488 	 * is resolved; otherwise, try to resolve.
489 	 */
490 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
491 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
492 		/*
493 		 * If entry has an expiry time and it is approaching,
494 		 * see if we need to send an ARP request within this
495 		 * arpt_down interval.
496 		 */
497 		if ((rt->rt_expire != 0) &&
498 		    (time_second + la->la_preempt > rt->rt_expire)) {
499 			arprequest(ifp,
500 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
501 				   &SIN(dst)->sin_addr,
502 				   IF_LLADDR(ifp));
503 			la->la_preempt--;
504 		}
505 
506 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
507 		return 1;
508 	}
509 	/*
510 	 * If ARP is disabled on this interface, stop.
511 	 * XXX
512 	 * Probably should not allocate empty llinfo struct if we are
513 	 * not going to be sending out an arp request.
514 	 */
515 	if (ifp->if_flags & IFF_NOARP) {
516 		m_freem(m);
517 		return (0);
518 	}
519 	/*
520 	 * There is an arptab entry, but no ethernet address
521 	 * response yet.  Replace the held mbuf with this
522 	 * latest one.
523 	 */
524 	if (la->la_hold != NULL)
525 		m_freem(la->la_hold);
526 	la->la_hold = m;
527 	la->la_msgport = curnetport;
528 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
529 		rt->rt_flags &= ~RTF_REJECT;
530 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
531 			rt->rt_expire = time_second;
532 			if (la->la_asked++ < arp_maxtries) {
533 				arprequest(ifp,
534 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
535 					   &SIN(dst)->sin_addr,
536 					   IF_LLADDR(ifp));
537 			} else {
538 				rt->rt_flags |= RTF_REJECT;
539 				rt->rt_expire += arpt_down;
540 				la->la_asked = 0;
541 				la->la_preempt = arp_maxtries;
542 			}
543 		}
544 	}
545 	return (0);
546 }
547 
548 /*
549  * Common length and type checks are done here,
550  * then the protocol-specific routine is called.
551  */
552 static void
553 arpintr(struct netmsg *msg)
554 {
555 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
556 	struct arphdr *ar;
557 	u_short ar_hrd;
558 
559 	if (m->m_len < sizeof(struct arphdr) &&
560 	    (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
561 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
562 		return;
563 	}
564 	ar = mtod(m, struct arphdr *);
565 
566 	ar_hrd = ntohs(ar->ar_hrd);
567 	if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
568 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
569 		    (unsigned char *)&ar->ar_hrd, "");
570 		m_freem(m);
571 		return;
572 	}
573 
574 	if (m->m_pkthdr.len < arphdr_len(ar)) {
575 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
576 			log(LOG_ERR, "arp: runt packet\n");
577 			return;
578 		}
579 		ar = mtod(m, struct arphdr *);
580 	}
581 
582 	switch (ntohs(ar->ar_pro)) {
583 #ifdef INET
584 	case ETHERTYPE_IP:
585 		in_arpinput(m);
586 		return;
587 #endif
588 	}
589 	m_freem(m);
590 	/* msg was embedded in the mbuf, do not reply! */
591 }
592 
593 #ifdef INET
594 /*
595  * ARP for Internet protocols on 10 Mb/s Ethernet.
596  * Algorithm is that given in RFC 826.
597  * In addition, a sanity check is performed on the sender
598  * protocol address, to catch impersonators.
599  * We no longer handle negotiations for use of trailer protocol:
600  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
601  * along with IP replies if we wanted trailers sent to us,
602  * and also sent them in response to IP replies.
603  * This allowed either end to announce the desire to receive
604  * trailer packets.
605  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
606  * but formerly didn't normally send requests.
607  */
608 static int log_arp_wrong_iface = 1;
609 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
610 	   &log_arp_wrong_iface, 0,
611 	   "log arp packets arriving on the wrong interface");
612 
613 static void
614 arp_hold_output(struct netmsg *nmsg)
615 {
616 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
617 	struct rtentry *rt;
618 	struct ifnet *ifp;
619 
620 	rt = nmsg->nm_lmsg.u.ms_resultp;
621 	ifp = m->m_pkthdr.rcvif;
622 	m->m_pkthdr.rcvif = NULL;
623 
624 	ifp->if_output(ifp, m, rt_key(rt), rt);
625 
626 	/* Drop the reference count bumped by the sender */
627 	RTFREE(rt);
628 
629 	/* nmsg was embedded in the mbuf, do not reply! */
630 }
631 
632 static void
633 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
634 		 boolean_t dologging)
635 {
636 	struct arphdr *ah = mtod(m, struct arphdr *);
637 	struct ifnet *ifp = m->m_pkthdr.rcvif;
638 	struct llinfo_arp *la;
639 	struct sockaddr_dl *sdl;
640 	struct rtentry *rt;
641 
642 	la = arplookup(saddr, create, FALSE);
643 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
644 		struct in_addr isaddr = { saddr };
645 
646 		/* the following is not an error when doing bridging */
647 		if (rt->rt_ifp != ifp) {
648 			if (dologging && log_arp_wrong_iface) {
649 				log(LOG_ERR,
650 				    "arp: %s is on %s "
651 				    "but got reply from %*D on %s\n",
652 				    inet_ntoa(isaddr),
653 				    rt->rt_ifp->if_xname,
654 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
655 				    ifp->if_xname);
656 			}
657 			return;
658 		}
659 		if (sdl->sdl_alen &&
660 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
661 			if (rt->rt_expire != 0) {
662 				if (dologging) {
663 			    		log(LOG_INFO,
664 			    		"arp: %s moved from %*D to %*D on %s\n",
665 			    		inet_ntoa(isaddr),
666 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
667 			    		":", ifp->if_addrlen,
668 			    		(u_char *)ar_sha(ah), ":",
669 			    		ifp->if_xname);
670 				}
671 			} else {
672 				if (dologging) {
673 					log(LOG_ERR,
674 					"arp: %*D attempts to modify "
675 					"permanent entry for %s on %s\n",
676 					ifp->if_addrlen, (u_char *)ar_sha(ah),
677 					":", inet_ntoa(isaddr), ifp->if_xname);
678 				}
679 				return;
680 			}
681 		}
682 		/*
683 		 * sanity check for the address length.
684 		 * XXX this does not work for protocols with variable address
685 		 * length. -is
686 		 */
687 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
688 			log(LOG_WARNING,
689 			    "arp from %*D: new addr len %d, was %d",
690 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
691 			    ah->ar_hln, sdl->sdl_alen);
692 		}
693 		if (ifp->if_addrlen != ah->ar_hln) {
694 			if (dologging) {
695 				log(LOG_WARNING,
696 				"arp from %*D: addr len: new %d, i/f %d "
697 				"(ignored)",
698 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
699 				ah->ar_hln, ifp->if_addrlen);
700 			}
701 			return;
702 		}
703 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
704 		if (rt->rt_expire != 0)
705 			rt->rt_expire = time_second + arpt_keep;
706 		rt->rt_flags &= ~RTF_REJECT;
707 		la->la_asked = 0;
708 		la->la_preempt = arp_maxtries;
709 
710 		/*
711 		 * This particular cpu might have been holding an mbuf
712 		 * pending ARP resolution.  If so, transmit the mbuf now.
713 		 */
714 		if (la->la_hold != NULL) {
715 			struct mbuf *m = la->la_hold;
716 			struct lwkt_port *port = la->la_msgport;
717 			struct netmsg_packet *pmsg;
718 
719 			la->la_hold = NULL;
720 			la->la_msgport = NULL;
721 
722 			m_adj(m, sizeof(struct ether_header));
723 
724 			/*
725 			 * Make sure that this rtentry will not be freed
726 			 * before the packet is processed on the target
727 			 * msgport.  The reference count will be dropped
728 			 * in the handler associated with this packet.
729 			 */
730 			rt->rt_refcnt++;
731 
732 			pmsg = &m->m_hdr.mh_netmsg;
733 			netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport,
734 				    MSGF_MPSAFE | MSGF_PRIORITY,
735 				    arp_hold_output);
736 			pmsg->nm_packet = m;
737 
738 			/* Record necessary information */
739 			m->m_pkthdr.rcvif = ifp;
740 			pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = rt;
741 
742 			lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
743 		}
744 	}
745 }
746 
747 #ifdef SMP
748 
749 struct netmsg_arp_update {
750 	struct netmsg	netmsg;
751 	struct mbuf	*m;
752 	in_addr_t	saddr;
753 	boolean_t	create;
754 };
755 
756 static void arp_update_msghandler(struct netmsg *);
757 
758 #endif
759 
760 /*
761  * Called from arpintr() - this routine is run from a single cpu.
762  */
763 static void
764 in_arpinput(struct mbuf *m)
765 {
766 	struct arphdr *ah;
767 	struct ifnet *ifp = m->m_pkthdr.rcvif;
768 	struct ether_header *eh;
769 	struct rtentry *rt;
770 	struct ifaddr_container *ifac;
771 	struct in_ifaddr_container *iac;
772 	struct in_ifaddr *ia = NULL;
773 	struct sockaddr sa;
774 	struct in_addr isaddr, itaddr, myaddr;
775 #ifdef SMP
776 	struct netmsg_arp_update msg;
777 #endif
778 	uint8_t *enaddr = NULL;
779 	int op;
780 	int req_len;
781 
782 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
783 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
784 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
785 		return;
786 	}
787 
788 	ah = mtod(m, struct arphdr *);
789 	op = ntohs(ah->ar_op);
790 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
791 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
792 
793 	myaddr.s_addr = INADDR_ANY;
794 #ifdef CARP
795 	if (ifp->if_carp != NULL) {
796 		get_mplock();
797 		if (ifp->if_carp != NULL &&
798 		    carp_iamatch(ifp->if_carp, &itaddr, &isaddr, &enaddr)) {
799 			rel_mplock();
800 			myaddr = itaddr;
801 			goto match;
802 		}
803 		rel_mplock();
804 	}
805 #endif
806 
807 	/*
808 	 * Check both target and sender IP addresses:
809 	 *
810 	 * If we receive the packet on the interface owning the address,
811 	 * then accept the address.
812 	 *
813 	 * For a bridge, we accept the address if the receive interface and
814 	 * the interface owning the address are on the same bridge.
815 	 * (This will change slightly when we have clusters of interfaces).
816 	 */
817 	LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
818 		ia = iac->ia;
819 
820 		/* Skip all ia's which don't match */
821 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
822 			continue;
823 #ifdef CARP
824 		if (ia->ia_ifp->if_type == IFT_CARP)
825 			continue;
826 #endif
827 		if (ia->ia_ifp == ifp)
828 			goto match;
829 
830 		if (ifp->if_bridge && ia->ia_ifp &&
831 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
832 			goto match;
833 	}
834 	LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
835 		ia = iac->ia;
836 
837 		/* Skip all ia's which don't match */
838 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
839 			continue;
840 #ifdef CARP
841 		if (ia->ia_ifp->if_type == IFT_CARP)
842 			continue;
843 #endif
844 		if (ia->ia_ifp == ifp)
845 			goto match;
846 
847 		if (ifp->if_bridge && ia->ia_ifp &&
848 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
849 			goto match;
850 	}
851 	/*
852 	 * No match, use the first inet address on the receive interface
853 	 * as a dummy address for the rest of the function.
854 	 */
855 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
856 		struct ifaddr *ifa = ifac->ifa;
857 
858 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
859 			ia = ifatoia(ifa);
860 			goto match;
861 		}
862 	}
863 	/*
864 	 * If we got here, we didn't find any suitable interface,
865 	 * so drop the packet.
866 	 */
867 	m_freem(m);
868 	return;
869 
870 match:
871 	if (!enaddr)
872 		enaddr = (uint8_t *)IF_LLADDR(ifp);
873 	if (myaddr.s_addr == INADDR_ANY)
874 		myaddr = ia->ia_addr.sin_addr;
875 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
876 		m_freem(m);	/* it's from me, ignore it. */
877 		return;
878 	}
879 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
880 		log(LOG_ERR,
881 		    "arp: link address is broadcast for IP address %s!\n",
882 		    inet_ntoa(isaddr));
883 		m_freem(m);
884 		return;
885 	}
886 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
887 		log(LOG_ERR,
888 		   "arp: %*D is using my IP address %s!\n",
889 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
890 		   inet_ntoa(isaddr));
891 		itaddr = myaddr;
892 		goto reply;
893 	}
894 #ifdef SMP
895 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
896 		    arp_update_msghandler);
897 	msg.m = m;
898 	msg.saddr = isaddr.s_addr;
899 	msg.create = (itaddr.s_addr == myaddr.s_addr);
900 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
901 #else
902 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
903 			 TRUE);
904 #endif
905 reply:
906 	if (op != ARPOP_REQUEST) {
907 		m_freem(m);
908 		return;
909 	}
910 	if (itaddr.s_addr == myaddr.s_addr) {
911 		/* I am the target */
912 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
913 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
914 	} else {
915 		struct llinfo_arp *la;
916 
917 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
918 		if (la == NULL) {
919 			struct sockaddr_in sin;
920 
921 			if (!arp_proxyall) {
922 				m_freem(m);
923 				return;
924 			}
925 
926 			bzero(&sin, sizeof sin);
927 			sin.sin_family = AF_INET;
928 			sin.sin_len = sizeof sin;
929 			sin.sin_addr = itaddr;
930 
931 			rt = rtpurelookup((struct sockaddr *)&sin);
932 			if (rt == NULL) {
933 				m_freem(m);
934 				return;
935 			}
936 			--rt->rt_refcnt;
937 			/*
938 			 * Don't send proxies for nodes on the same interface
939 			 * as this one came out of, or we'll get into a fight
940 			 * over who claims what Ether address.
941 			 */
942 			if (rt->rt_ifp == ifp) {
943 				m_freem(m);
944 				return;
945 			}
946 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
947 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
948 #ifdef DEBUG_PROXY
949 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
950 #endif
951 		} else {
952 			struct sockaddr_dl *sdl;
953 
954 			rt = la->la_rt;
955 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
956 			sdl = SDL(rt->rt_gateway);
957 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
958 		}
959 	}
960 
961 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
962 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
963 	ah->ar_op = htons(ARPOP_REPLY);
964 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
965 	switch (ifp->if_type) {
966 	case IFT_ETHER:
967 	/*
968 	 * May not be correct for types not explictly
969 	 * listed, but it is our best guess.
970 	 */
971 	default:
972 		eh = (struct ether_header *)sa.sa_data;
973 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
974 		eh->ether_type = htons(ETHERTYPE_ARP);
975 		break;
976 	}
977 	sa.sa_family = AF_UNSPEC;
978 	sa.sa_len = sizeof sa;
979 	ifp->if_output(ifp, m, &sa, NULL);
980 }
981 
982 #ifdef SMP
983 
984 static void
985 arp_update_msghandler(struct netmsg *netmsg)
986 {
987 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
988 	int nextcpu;
989 
990 	arp_update_oncpu(msg->m, msg->saddr, msg->create, mycpuid == 0);
991 
992 	nextcpu = mycpuid + 1;
993 	if (nextcpu < ncpus)
994 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
995 	else
996 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
997 }
998 
999 #endif	/* SMP */
1000 
1001 #endif	/* INET */
1002 
1003 /*
1004  * Free an arp entry.  If the arp entry is actively referenced or represents
1005  * a static entry we only clear it back to an unresolved state, otherwise
1006  * we destroy the entry entirely.
1007  *
1008  * Note that static entries are created when route add ... -interface is used
1009  * to create an interface route to a (direct) destination.
1010  */
1011 static void
1012 arptfree(struct llinfo_arp *la)
1013 {
1014 	struct rtentry *rt = la->la_rt;
1015 	struct sockaddr_dl *sdl;
1016 
1017 	if (rt == NULL)
1018 		panic("arptfree");
1019 	sdl = SDL(rt->rt_gateway);
1020 	if (sdl != NULL &&
1021 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
1022 	     (rt->rt_flags & RTF_STATIC))) {
1023 		sdl->sdl_alen = 0;
1024 		la->la_preempt = la->la_asked = 0;
1025 		rt->rt_flags &= ~RTF_REJECT;
1026 		return;
1027 	}
1028 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1029 }
1030 
1031 /*
1032  * Lookup or enter a new address in arptab.
1033  */
1034 static struct llinfo_arp *
1035 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
1036 {
1037 	struct rtentry *rt;
1038 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
1039 	const char *why = NULL;
1040 
1041 	sin.sin_addr.s_addr = addr;
1042 	sin.sin_other = proxy ? SIN_PROXY : 0;
1043 	if (create)
1044 		rt = rtlookup((struct sockaddr *)&sin);
1045 	else
1046 		rt = rtpurelookup((struct sockaddr *)&sin);
1047 	if (rt == NULL)
1048 		return (NULL);
1049 	rt->rt_refcnt--;
1050 
1051 	if (rt->rt_flags & RTF_GATEWAY)
1052 		why = "host is not on local network";
1053 	else if (!(rt->rt_flags & RTF_LLINFO))
1054 		why = "could not allocate llinfo";
1055 	else if (rt->rt_gateway->sa_family != AF_LINK)
1056 		why = "gateway route is not ours";
1057 
1058 	if (why) {
1059 		if (create) {
1060 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1061 			    inet_ntoa(sin.sin_addr), why);
1062 		}
1063 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1064 			/* No references to this route.  Purge it. */
1065 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1066 				  rt_mask(rt), rt->rt_flags, NULL);
1067 		}
1068 		return (NULL);
1069 	}
1070 	return (rt->rt_llinfo);
1071 }
1072 
1073 void
1074 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1075 {
1076 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1077 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1078 				 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1079 	}
1080 	ifa->ifa_rtrequest = arp_rtrequest;
1081 	ifa->ifa_flags |= RTF_CLONING;
1082 }
1083 
1084 void
1085 arp_iainit(struct ifnet *ifp, const struct in_addr *addr, const u_char *enaddr)
1086 {
1087 	if (addr->s_addr != INADDR_ANY)
1088 		arprequest_async(ifp, addr, addr, enaddr);
1089 }
1090 
1091 static void
1092 arp_init(void)
1093 {
1094 	uint32_t flags;
1095 	int cpu;
1096 
1097 	for (cpu = 0; cpu < ncpus2; cpu++)
1098 		LIST_INIT(&llinfo_arp_list[cpu]);
1099 
1100 	if (arp_mpsafe) {
1101 		flags = NETISR_FLAG_MPSAFE;
1102 		kprintf("arp: MPSAFE\n");
1103 	} else {
1104 		flags = NETISR_FLAG_NOTMPSAFE;
1105 	}
1106 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr, flags);
1107 }
1108 
1109 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1110