xref: /freebsd/sys/netinet/if_ether.c (revision b6de9e91bd2c47efaeec72a08642f8fd99cc7b20)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30  * $FreeBSD$
31  */
32 
33 /*
34  * Ethernet address resolution protocol.
35  * TODO:
36  *	add "inuse/lock" bit (or ref. count) along with valid bit
37  */
38 
39 #include "opt_inet.h"
40 #include "opt_mac.h"
41 #include "opt_carp.h"
42 
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/queue.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/mac.h>
49 #include <sys/mbuf.h>
50 #include <sys/malloc.h>
51 #include <sys/socket.h>
52 #include <sys/syslog.h>
53 
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/route.h>
58 #include <net/netisr.h>
59 #include <net/if_llc.h>
60 #include <net/ethernet.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_var.h>
64 #include <netinet/if_ether.h>
65 
66 #include <net/if_arc.h>
67 #include <net/iso88025.h>
68 
69 #ifdef DEV_CARP
70 #include <netinet/ip_carp.h>
71 #endif
72 
73 #define SIN(s) ((struct sockaddr_in *)s)
74 #define SDL(s) ((struct sockaddr_dl *)s)
75 
76 SYSCTL_DECL(_net_link_ether);
77 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
78 
79 /* timer values */
80 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
81 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
82 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
83 
84 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
85 	   &arpt_prune, 0, "");
86 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
87 	   &arpt_keep, 0, "");
88 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
89 	   &arpt_down, 0, "");
90 
91 #define	rt_expire rt_rmx.rmx_expire
92 
93 struct llinfo_arp {
94 	LIST_ENTRY(llinfo_arp) la_le;
95 	struct	rtentry *la_rt;
96 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
97 	u_short	la_preempt;	/* countdown for pre-expiry arps */
98 	u_short	la_asked;	/* #times we QUERIED following expiration */
99 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
100 };
101 
102 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
103 
104 static struct	ifqueue arpintrq;
105 static int	arp_allocated;
106 
107 static int	arp_maxtries = 5;
108 static int	useloopback = 1; /* use loopback interface for local traffic */
109 static int	arp_proxyall = 0;
110 static struct callout arp_callout;
111 
112 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
113 	   &arp_maxtries, 0, "");
114 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
115 	   &useloopback, 0, "");
116 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
117 	   &arp_proxyall, 0, "");
118 
119 static void	arp_init(void);
120 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
121 static void	arprequest(struct ifnet *,
122 			struct in_addr *, struct in_addr *, u_char *);
123 static void	arpintr(struct mbuf *);
124 static void	arptimer(void *);
125 static struct rtentry
126 		*arplookup(u_long, int, int);
127 #ifdef INET
128 static void	in_arpinput(struct mbuf *);
129 #endif
130 
131 /*
132  * Timeout routine.  Age arp_tab entries periodically.
133  */
134 /* ARGSUSED */
135 static void
136 arptimer(void * __unused unused)
137 {
138 	struct llinfo_arp *la, *ola;
139 
140 	RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]);
141 	LIST_FOREACH_SAFE(la, &llinfo_arp, la_le, ola) {
142 		struct rtentry *rt = la->la_rt;
143 
144 		RT_LOCK(rt);
145 		if (rt->rt_expire && rt->rt_expire <= time_uptime) {
146 			struct sockaddr_dl *sdl = SDL(rt->rt_gateway);
147 
148 			KASSERT(sdl->sdl_family == AF_LINK, ("sdl_family %d",
149 			    sdl->sdl_family));
150 			if (rt->rt_refcnt > 1) {
151 				sdl->sdl_alen = 0;
152 				la->la_preempt = la->la_asked = 0;
153 				rt->rt_flags &= ~RTF_REJECT;
154 				RT_UNLOCK(rt);
155 				continue;
156 			}
157 			RT_UNLOCK(rt);
158 			/*
159 			 * XXX: LIST_REMOVE() is deep inside rtrequest().
160 			 */
161 			rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0,
162 			    NULL);
163 			continue;
164 		}
165 		RT_UNLOCK(rt);
166 	}
167 	RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]);
168 
169 	callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL);
170 }
171 
172 /*
173  * Parallel to llc_rtrequest.
174  */
175 static void
176 arp_rtrequest(req, rt, info)
177 	int req;
178 	struct rtentry *rt;
179 	struct rt_addrinfo *info;
180 {
181 	struct sockaddr *gate;
182 	struct llinfo_arp *la;
183 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
184 	struct in_ifaddr *ia;
185 	struct ifaddr *ifa;
186 
187 	RT_LOCK_ASSERT(rt);
188 
189 	if (rt->rt_flags & RTF_GATEWAY)
190 		return;
191 	gate = rt->rt_gateway;
192 	la = (struct llinfo_arp *)rt->rt_llinfo;
193 	switch (req) {
194 
195 	case RTM_ADD:
196 		/*
197 		 * XXX: If this is a manually added route to interface
198 		 * such as older version of routed or gated might provide,
199 		 * restore cloning bit.
200 		 */
201 		if ((rt->rt_flags & RTF_HOST) == 0 &&
202 		    rt_mask(rt) != NULL &&
203 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
204 			rt->rt_flags |= RTF_CLONING;
205 		if (rt->rt_flags & RTF_CLONING) {
206 			/*
207 			 * Case 1: This route should come from a route to iface.
208 			 */
209 			rt_setgate(rt, rt_key(rt),
210 					(struct sockaddr *)&null_sdl);
211 			gate = rt->rt_gateway;
212 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
213 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
214 			rt->rt_expire = time_uptime;
215 			break;
216 		}
217 		/* Announce a new entry if requested. */
218 		if (rt->rt_flags & RTF_ANNOUNCE)
219 			arprequest(rt->rt_ifp,
220 			    &SIN(rt_key(rt))->sin_addr,
221 			    &SIN(rt_key(rt))->sin_addr,
222 			    (u_char *)LLADDR(SDL(gate)));
223 		/*FALLTHROUGH*/
224 	case RTM_RESOLVE:
225 		if (gate->sa_family != AF_LINK ||
226 		    gate->sa_len < sizeof(null_sdl)) {
227 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
228 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
229 			    (gate->sa_family != AF_LINK) ?
230 			    " (!AF_LINK)": "");
231 			break;
232 		}
233 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
234 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
235 		if (la != 0)
236 			break; /* This happens on a route change */
237 		/*
238 		 * Case 2:  This route may come from cloning, or a manual route
239 		 * add with a LL address.
240 		 */
241 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
242 		rt->rt_llinfo = (caddr_t)la;
243 		if (la == 0) {
244 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
245 			break;
246 		}
247 		arp_allocated++;
248 		/*
249 		 * We are storing a route entry outside of radix tree. So,
250 		 * it can be found and accessed by other means than radix
251 		 * lookup. The routing code assumes that any rtentry detached
252 		 * from radix can be destroyed safely. To prevent this, we
253 		 * add an additional reference.
254 		 */
255 		RT_ADDREF(rt);
256 		la->la_rt = rt;
257 		rt->rt_flags |= RTF_LLINFO;
258 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
259 		LIST_INSERT_HEAD(&llinfo_arp, 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 		    rt->rt_ifp->if_type != IFT_ARCNET) {
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 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
283 			if (ia->ia_ifp == rt->rt_ifp &&
284 			    SIN(rt_key(rt))->sin_addr.s_addr ==
285 			    (IA_SIN(ia))->sin_addr.s_addr)
286 				break;
287 		}
288 		if (ia) {
289 		    /*
290 		     * This test used to be
291 		     *	if (loif.if_flags & IFF_UP)
292 		     * It allowed local traffic to be forced
293 		     * through the hardware by configuring the loopback down.
294 		     * However, it causes problems during network configuration
295 		     * for boards that can't receive packets they send.
296 		     * It is now necessary to clear "useloopback" and remove
297 		     * the route to force traffic out to the hardware.
298 		     */
299 			rt->rt_expire = 0;
300 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
301 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
302 			if (useloopback)
303 				rt->rt_ifp = loif;
304 
305 		    /*
306 		     * make sure to set rt->rt_ifa to the interface
307 		     * address we are using, otherwise we will have trouble
308 		     * with source address selection.
309 		     */
310 			ifa = &ia->ia_ifa;
311 			if (ifa != rt->rt_ifa) {
312 				IFAFREE(rt->rt_ifa);
313 				IFAREF(ifa);
314 				rt->rt_ifa = ifa;
315 			}
316 		}
317 		break;
318 
319 	case RTM_DELETE:
320 		if (la == 0)
321 			break;
322 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
323 		LIST_REMOVE(la, la_le);
324 		RT_REMREF(rt);
325 		rt->rt_llinfo = 0;
326 		rt->rt_flags &= ~RTF_LLINFO;
327 		if (la->la_hold)
328 			m_freem(la->la_hold);
329 		Free((caddr_t)la);
330 	}
331 }
332 
333 /*
334  * Broadcast an ARP request. Caller specifies:
335  *	- arp header source ip address
336  *	- arp header target ip address
337  *	- arp header source ethernet address
338  */
339 static void
340 arprequest(ifp, sip, tip, enaddr)
341 	struct ifnet *ifp;
342 	struct in_addr *sip, *tip;
343 	u_char *enaddr;
344 {
345 	struct mbuf *m;
346 	struct arphdr *ah;
347 	struct sockaddr sa;
348 
349 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
350 		return;
351 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
352 		2*ifp->if_data.ifi_addrlen;
353 	m->m_pkthdr.len = m->m_len;
354 	MH_ALIGN(m, m->m_len);
355 	ah = mtod(m, struct arphdr *);
356 	bzero((caddr_t)ah, m->m_len);
357 #ifdef MAC
358 	mac_create_mbuf_linklayer(ifp, m);
359 #endif
360 	ah->ar_pro = htons(ETHERTYPE_IP);
361 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
362 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
363 	ah->ar_op = htons(ARPOP_REQUEST);
364 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
365 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
366 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
367 	sa.sa_family = AF_ARP;
368 	sa.sa_len = 2;
369 	m->m_flags |= M_BCAST;
370 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
371 
372 	return;
373 }
374 
375 /*
376  * Resolve an IP address into an ethernet address.
377  * On input:
378  *    ifp is the interface we use
379  *    dst is the next hop,
380  *    rt0 is the route to the final destination (possibly useless)
381  *    m is the mbuf
382  *    desten is where we want the address.
383  *
384  * On success, desten is filled in and the function returns 0;
385  * If the packet must be held pending resolution, we return EWOULDBLOCK
386  * On other errors, we return the corresponding error code.
387  */
388 int
389 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
390 	struct sockaddr *dst, u_char *desten)
391 {
392 	struct llinfo_arp *la = NULL;
393 	struct rtentry *rt = NULL;
394 	struct sockaddr_dl *sdl;
395 	int error;
396 
397 	if (m->m_flags & M_BCAST) {	/* broadcast */
398 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
399 		return (0);
400 	}
401 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
402 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
403 		return (0);
404 	}
405 
406 	if (rt0 != NULL) {
407 		error = rt_check(&rt, &rt0, dst);
408 		if (error) {
409 			m_freem(m);
410 			return error;
411 		}
412 		la = (struct llinfo_arp *)rt->rt_llinfo;
413 		if (la == NULL)
414 			RT_UNLOCK(rt);
415 	}
416 	if (la == NULL) {
417 		/*
418 		 * We enter this block in case if rt0 was NULL,
419 		 * or if rt found by rt_check() didn't have llinfo.
420 		 */
421 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
422 		if (rt == NULL) {
423 			log(LOG_DEBUG,
424 			    "arpresolve: can't allocate route for %s\n",
425 			    inet_ntoa(SIN(dst)->sin_addr));
426 			m_freem(m);
427 			return (EINVAL); /* XXX */
428 		}
429 		la = (struct llinfo_arp *)rt->rt_llinfo;
430 		if (la == NULL) {
431 			RT_UNLOCK(rt);
432 			log(LOG_DEBUG,
433 			    "arpresolve: can't allocate llinfo for %s\n",
434 			    inet_ntoa(SIN(dst)->sin_addr));
435 			m_freem(m);
436 			return (EINVAL); /* XXX */
437 		}
438 	}
439 	sdl = SDL(rt->rt_gateway);
440 	/*
441 	 * Check the address family and length is valid, the address
442 	 * is resolved; otherwise, try to resolve.
443 	 */
444 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
445 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
446 
447 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
448 
449 		/*
450 		 * If entry has an expiry time and it is approaching,
451 		 * see if we need to send an ARP request within this
452 		 * arpt_down interval.
453 		 */
454 		if ((rt->rt_expire != 0) &&
455 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
456 			struct in_addr sin =
457 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
458 
459 			la->la_preempt--;
460 			RT_UNLOCK(rt);
461 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
462 			    IF_LLADDR(ifp));
463 			return (0);
464 		}
465 
466 		RT_UNLOCK(rt);
467 		return (0);
468 	}
469 	/*
470 	 * If ARP is disabled or static on this interface, stop.
471 	 * XXX
472 	 * Probably should not allocate empty llinfo struct if we are
473 	 * not going to be sending out an arp request.
474 	 */
475 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
476 		RT_UNLOCK(rt);
477 		m_freem(m);
478 		return (EINVAL);
479 	}
480 	/*
481 	 * There is an arptab entry, but no ethernet address
482 	 * response yet.  Replace the held mbuf with this
483 	 * latest one.
484 	 */
485 	if (la->la_hold)
486 		m_freem(la->la_hold);
487 	la->la_hold = m;
488 	if (rt->rt_expire) {
489 		rt->rt_flags &= ~RTF_REJECT;
490 		if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
491 			rt->rt_expire = time_uptime;
492 			if (la->la_asked++ < arp_maxtries) {
493 				struct in_addr sin =
494 				    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
495 
496 				RT_UNLOCK(rt);
497 				arprequest(ifp, &sin, &SIN(dst)->sin_addr,
498 				    IF_LLADDR(ifp));
499 				return (EWOULDBLOCK);
500 			} else {
501 				rt->rt_flags |= RTF_REJECT;
502 				rt->rt_expire += arpt_down;
503 				la->la_asked = 0;
504 				la->la_preempt = arp_maxtries;
505 			}
506 
507 		}
508 	}
509 	RT_UNLOCK(rt);
510 	return (EWOULDBLOCK);
511 }
512 
513 /*
514  * Common length and type checks are done here,
515  * then the protocol-specific routine is called.
516  */
517 static void
518 arpintr(struct mbuf *m)
519 {
520 	struct arphdr *ar;
521 
522 	if (m->m_len < sizeof(struct arphdr) &&
523 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
524 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
525 		return;
526 	}
527 	ar = mtod(m, struct arphdr *);
528 
529 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
530 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
531 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
532 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
533 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
534 		    (unsigned char *)&ar->ar_hrd, "");
535 		m_freem(m);
536 		return;
537 	}
538 
539 	if (m->m_len < arphdr_len(ar)) {
540 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
541 			log(LOG_ERR, "arp: runt packet\n");
542 			m_freem(m);
543 			return;
544 		}
545 		ar = mtod(m, struct arphdr *);
546 	}
547 
548 	switch (ntohs(ar->ar_pro)) {
549 #ifdef INET
550 	case ETHERTYPE_IP:
551 		in_arpinput(m);
552 		return;
553 #endif
554 	}
555 	m_freem(m);
556 }
557 
558 #ifdef INET
559 /*
560  * ARP for Internet protocols on 10 Mb/s Ethernet.
561  * Algorithm is that given in RFC 826.
562  * In addition, a sanity check is performed on the sender
563  * protocol address, to catch impersonators.
564  * We no longer handle negotiations for use of trailer protocol:
565  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
566  * along with IP replies if we wanted trailers sent to us,
567  * and also sent them in response to IP replies.
568  * This allowed either end to announce the desire to receive
569  * trailer packets.
570  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
571  * but formerly didn't normally send requests.
572  */
573 static int log_arp_wrong_iface = 1;
574 static int log_arp_movements = 1;
575 
576 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
577 	&log_arp_wrong_iface, 0,
578 	"log arp packets arriving on the wrong interface");
579 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
580         &log_arp_movements, 0,
581         "log arp replies from MACs different than the one in the cache");
582 
583 
584 static void
585 in_arpinput(m)
586 	struct mbuf *m;
587 {
588 	struct arphdr *ah;
589 	struct ifnet *ifp = m->m_pkthdr.rcvif;
590 	struct iso88025_header *th = (struct iso88025_header *)0;
591 	struct iso88025_sockaddr_dl_data *trld;
592 	struct llinfo_arp *la;
593 	struct rtentry *rt;
594 	struct ifaddr *ifa;
595 	struct in_ifaddr *ia;
596 	struct sockaddr_dl *sdl;
597 	struct sockaddr sa;
598 	struct in_addr isaddr, itaddr, myaddr;
599 	struct mbuf *hold;
600 	u_int8_t *enaddr = NULL;
601 	int op, rif_len;
602 	int req_len;
603 	int bridged = 0;
604 #ifdef DEV_CARP
605 	int carp_match = 0;
606 #endif
607 
608 	if (ifp->if_bridge)
609 		bridged = 1;
610 
611 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
612 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
613 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
614 		return;
615 	}
616 
617 	ah = mtod(m, struct arphdr *);
618 	op = ntohs(ah->ar_op);
619 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
620 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
621 
622 	/*
623 	 * For a bridge, we want to check the address irrespective
624 	 * of the receive interface. (This will change slightly
625 	 * when we have clusters of interfaces).
626 	 * If the interface does not match, but the recieving interface
627 	 * is part of carp, we call carp_iamatch to see if this is a
628 	 * request for the virtual host ip.
629 	 * XXX: This is really ugly!
630 	 */
631 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
632 		if ((bridged || (ia->ia_ifp == ifp)) &&
633 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
634 			goto match;
635 #ifdef DEV_CARP
636 		if (ifp->if_carp != NULL &&
637 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
638 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
639 			carp_match = 1;
640 			goto match;
641 		}
642 #endif
643 	}
644 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
645 		if ((bridged || (ia->ia_ifp == ifp)) &&
646 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
647 			goto match;
648 	/*
649 	 * No match, use the first inet address on the receive interface
650 	 * as a dummy address for the rest of the function.
651 	 */
652 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
653 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
654 			ia = ifatoia(ifa);
655 			goto match;
656 		}
657 	/*
658 	 * If bridging, fall back to using any inet address.
659 	 */
660 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
661 		goto drop;
662 match:
663 	if (!enaddr)
664 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
665 	myaddr = ia->ia_addr.sin_addr;
666 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
667 		goto drop;	/* it's from me, ignore it. */
668 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
669 		log(LOG_ERR,
670 		    "arp: link address is broadcast for IP address %s!\n",
671 		    inet_ntoa(isaddr));
672 		goto drop;
673 	}
674 	/*
675 	 * Warn if another host is using the same IP address, but only if the
676 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
677 	 * case we suppress the warning to avoid false positive complaints of
678 	 * potential misconfiguration.
679 	 */
680 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
681 		log(LOG_ERR,
682 		   "arp: %*D is using my IP address %s!\n",
683 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
684 		   inet_ntoa(isaddr));
685 		itaddr = myaddr;
686 		goto reply;
687 	}
688 	if (ifp->if_flags & IFF_STATICARP)
689 		goto reply;
690 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
691 	if (rt != NULL) {
692 		la = (struct llinfo_arp *)rt->rt_llinfo;
693 		if (la == NULL) {
694 			RT_UNLOCK(rt);
695 			goto reply;
696 		}
697 	} else
698 		goto reply;
699 
700 	/* The following is not an error when doing bridging. */
701 	if (!bridged && rt->rt_ifp != ifp
702 #ifdef DEV_CARP
703 	    && (ifp->if_type != IFT_CARP || !carp_match)
704 #endif
705 							) {
706 		if (log_arp_wrong_iface)
707 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
708 			    inet_ntoa(isaddr),
709 			    rt->rt_ifp->if_xname,
710 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
711 			    ifp->if_xname);
712 		RT_UNLOCK(rt);
713 		goto reply;
714 	}
715 	sdl = SDL(rt->rt_gateway);
716 	if (sdl->sdl_alen &&
717 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
718 		if (rt->rt_expire) {
719 		    if (log_arp_movements)
720 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
721 			    inet_ntoa(isaddr),
722 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
723 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
724 			    ifp->if_xname);
725 		} else {
726 		    log(LOG_ERR,
727 			"arp: %*D attempts to modify permanent entry for %s on %s\n",
728 			ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
729 			inet_ntoa(isaddr), ifp->if_xname);
730 		    RT_UNLOCK(rt);
731 		    goto reply;
732 		}
733 	}
734 	/*
735 	 * sanity check for the address length.
736 	 * XXX this does not work for protocols with variable address
737 	 * length. -is
738 	 */
739 	if (sdl->sdl_alen &&
740 	    sdl->sdl_alen != ah->ar_hln) {
741 		log(LOG_WARNING,
742 		    "arp from %*D: new addr len %d, was %d",
743 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
744 		    ah->ar_hln, sdl->sdl_alen);
745 	}
746 	if (ifp->if_addrlen != ah->ar_hln) {
747 		log(LOG_WARNING,
748 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
749 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
750 		    ah->ar_hln, ifp->if_addrlen);
751 		RT_UNLOCK(rt);
752 		goto reply;
753 	}
754 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
755 	    sdl->sdl_alen = ah->ar_hln);
756 	/*
757 	 * If we receive an arp from a token-ring station over
758 	 * a token-ring nic then try to save the source
759 	 * routing info.
760 	 */
761 	if (ifp->if_type == IFT_ISO88025) {
762 		th = (struct iso88025_header *)m->m_pkthdr.header;
763 		trld = SDL_ISO88025(sdl);
764 		rif_len = TR_RCF_RIFLEN(th->rcf);
765 		if ((th->iso88025_shost[0] & TR_RII) &&
766 		    (rif_len > 2)) {
767 			trld->trld_rcf = th->rcf;
768 			trld->trld_rcf ^= htons(TR_RCF_DIR);
769 			memcpy(trld->trld_route, th->rd, rif_len - 2);
770 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
771 			/*
772 			 * Set up source routing information for
773 			 * reply packet (XXX)
774 			 */
775 			m->m_data -= rif_len;
776 			m->m_len  += rif_len;
777 			m->m_pkthdr.len += rif_len;
778 		} else {
779 			th->iso88025_shost[0] &= ~TR_RII;
780 			trld->trld_rcf = 0;
781 		}
782 		m->m_data -= 8;
783 		m->m_len  += 8;
784 		m->m_pkthdr.len += 8;
785 		th->rcf = trld->trld_rcf;
786 	}
787 	if (rt->rt_expire)
788 		rt->rt_expire = time_uptime + arpt_keep;
789 	rt->rt_flags &= ~RTF_REJECT;
790 	la->la_asked = 0;
791 	la->la_preempt = arp_maxtries;
792 	hold = la->la_hold;
793 	la->la_hold = NULL;
794 	RT_UNLOCK(rt);
795 	if (hold != NULL)
796 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
797 
798 reply:
799 	if (op != ARPOP_REQUEST)
800 		goto drop;
801 	if (itaddr.s_addr == myaddr.s_addr) {
802 		/* I am the target */
803 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
804 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
805 	} else {
806 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
807 		if (rt == NULL) {
808 			struct sockaddr_in sin;
809 
810 			if (!arp_proxyall)
811 				goto drop;
812 
813 			bzero(&sin, sizeof sin);
814 			sin.sin_family = AF_INET;
815 			sin.sin_len = sizeof sin;
816 			sin.sin_addr = itaddr;
817 
818 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
819 			if (!rt)
820 				goto drop;
821 			/*
822 			 * Don't send proxies for nodes on the same interface
823 			 * as this one came out of, or we'll get into a fight
824 			 * over who claims what Ether address.
825 			 */
826 			if (rt->rt_ifp == ifp) {
827 				rtfree(rt);
828 				goto drop;
829 			}
830 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
831 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
832 			rtfree(rt);
833 
834 			/*
835 			 * Also check that the node which sent the ARP packet
836 			 * is on the the interface we expect it to be on. This
837 			 * avoids ARP chaos if an interface is connected to the
838 			 * wrong network.
839 			 */
840 			sin.sin_addr = isaddr;
841 
842 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
843 			if (!rt)
844 				goto drop;
845 			if (rt->rt_ifp != ifp) {
846 				log(LOG_INFO, "arp_proxy: ignoring request"
847 				    " from %s via %s, expecting %s\n",
848 				    inet_ntoa(isaddr), ifp->if_xname,
849 				    rt->rt_ifp->if_xname);
850 				rtfree(rt);
851 				goto drop;
852 			}
853 			rtfree(rt);
854 
855 #ifdef DEBUG_PROXY
856 			printf("arp: proxying for %s\n",
857 			       inet_ntoa(itaddr));
858 #endif
859 		} else {
860 			/*
861 			 * Return proxied ARP replies only on the interface
862 			 * where this network resides. Otherwise we may
863 			 * conflict with the host we are proxying for.
864 			 */
865 			if (rt->rt_ifp != ifp) {
866 				RT_UNLOCK(rt);
867 				goto drop;
868 			}
869 			sdl = SDL(rt->rt_gateway);
870 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
871 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
872 			RT_UNLOCK(rt);
873 		}
874 	}
875 
876 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
877 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
878 	ah->ar_op = htons(ARPOP_REPLY);
879 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
880 	m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
881 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
882 	m->m_pkthdr.len = m->m_len;
883 	sa.sa_family = AF_ARP;
884 	sa.sa_len = 2;
885 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
886 	return;
887 
888 drop:
889 	m_freem(m);
890 }
891 #endif
892 
893 /*
894  * Lookup or enter a new address in arptab.
895  */
896 static struct rtentry *
897 arplookup(addr, create, proxy)
898 	u_long addr;
899 	int create, proxy;
900 {
901 	struct rtentry *rt;
902 	struct sockaddr_inarp sin;
903 	const char *why = 0;
904 
905 	bzero(&sin, sizeof(sin));
906 	sin.sin_len = sizeof(sin);
907 	sin.sin_family = AF_INET;
908 	sin.sin_addr.s_addr = addr;
909 	if (proxy)
910 		sin.sin_other = SIN_PROXY;
911 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
912 	if (rt == 0)
913 		return (0);
914 
915 	if (rt->rt_flags & RTF_GATEWAY)
916 		why = "host is not on local network";
917 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
918 		why = "could not allocate llinfo";
919 	else if (rt->rt_gateway->sa_family != AF_LINK)
920 		why = "gateway route is not ours";
921 
922 	if (why) {
923 #define	ISDYNCLONE(_rt) \
924 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
925 		if (create)
926 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
927 			    inet_ntoa(sin.sin_addr), why);
928 		/*
929 		 * If there are no references to this Layer 2 route,
930 		 * and it is a cloned route, and not static, and
931 		 * arplookup() is creating the route, then purge
932 		 * it from the routing table as it is probably bogus.
933 		 */
934 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
935 			rtexpunge(rt);
936 		RTFREE_LOCKED(rt);
937 		return (0);
938 #undef ISDYNCLONE
939 	} else {
940 		RT_REMREF(rt);
941 		return (rt);
942 	}
943 }
944 
945 void
946 arp_ifinit(ifp, ifa)
947 	struct ifnet *ifp;
948 	struct ifaddr *ifa;
949 {
950 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
951 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
952 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
953 	ifa->ifa_rtrequest = arp_rtrequest;
954 	ifa->ifa_flags |= RTF_CLONING;
955 }
956 
957 void
958 arp_ifinit2(ifp, ifa, enaddr)
959 	struct ifnet *ifp;
960 	struct ifaddr *ifa;
961 	u_char *enaddr;
962 {
963 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
964 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
965 				&IA_SIN(ifa)->sin_addr, enaddr);
966 	ifa->ifa_rtrequest = arp_rtrequest;
967 	ifa->ifa_flags |= RTF_CLONING;
968 }
969 
970 static void
971 arp_init(void)
972 {
973 
974 	arpintrq.ifq_maxlen = 50;
975 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
976 	LIST_INIT(&llinfo_arp);
977 	callout_init(&arp_callout, CALLOUT_MPSAFE);
978 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
979 	callout_reset(&arp_callout, hz, arptimer, NULL);
980 }
981 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
982