xref: /freebsd/sys/netinet/if_ether.c (revision f856af0466c076beef4ea9b15d088e1119a945b8)
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/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/socket.h>
51 #include <sys/syslog.h>
52 
53 #include <net/if.h>
54 #include <net/if_dl.h>
55 #include <net/if_types.h>
56 #include <net/route.h>
57 #include <net/netisr.h>
58 #include <net/if_llc.h>
59 #include <net/ethernet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_var.h>
63 #include <netinet/if_ether.h>
64 
65 #include <net/if_arc.h>
66 #include <net/iso88025.h>
67 
68 #ifdef DEV_CARP
69 #include <netinet/ip_carp.h>
70 #endif
71 
72 #include <security/mac/mac_framework.h>
73 
74 #define SIN(s) ((struct sockaddr_in *)s)
75 #define SDL(s) ((struct sockaddr_dl *)s)
76 
77 SYSCTL_DECL(_net_link_ether);
78 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
79 
80 /* timer values */
81 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
82 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
83 
84 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
85 	   &arpt_prune, 0, "ARP table prune interval in seconds");
86 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
87 	   &arpt_keep, 0, "ARP entry lifetime in seconds");
88 
89 #define	rt_expire rt_rmx.rmx_expire
90 
91 struct llinfo_arp {
92 	LIST_ENTRY(llinfo_arp) la_le;
93 	struct	rtentry *la_rt;
94 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
95 	u_short	la_preempt;	/* countdown for pre-expiry arps */
96 	u_short	la_asked;	/* # requests sent */
97 };
98 
99 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
100 
101 static struct	ifqueue arpintrq;
102 static int	arp_allocated;
103 
104 static int	arp_maxtries = 5;
105 static int	useloopback = 1; /* use loopback interface for local traffic */
106 static int	arp_proxyall = 0;
107 static struct callout arp_callout;
108 
109 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
110 	   &arp_maxtries, 0, "ARP resolution attempts before returning error");
111 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
112 	   &useloopback, 0, "Use the loopback interface for local traffic");
113 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
114 	   &arp_proxyall, 0, "Enable proxy ARP for all suitable requests");
115 
116 static void	arp_init(void);
117 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
118 static void	arprequest(struct ifnet *,
119 			struct in_addr *, struct in_addr *, u_char *);
120 static void	arpintr(struct mbuf *);
121 static void	arptimer(void *);
122 static struct rtentry
123 		*arplookup(u_long, int, int);
124 #ifdef INET
125 static void	in_arpinput(struct mbuf *);
126 #endif
127 
128 /*
129  * Timeout routine.  Age arp_tab entries periodically.
130  */
131 /* ARGSUSED */
132 static void
133 arptimer(void * __unused unused)
134 {
135 	struct llinfo_arp *la, *ola;
136 
137 	RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]);
138 	LIST_FOREACH_SAFE(la, &llinfo_arp, la_le, ola) {
139 		struct rtentry *rt = la->la_rt;
140 
141 		RT_LOCK(rt);
142 		if (rt->rt_expire && rt->rt_expire <= time_uptime) {
143 			struct sockaddr_dl *sdl = SDL(rt->rt_gateway);
144 
145 			KASSERT(sdl->sdl_family == AF_LINK, ("sdl_family %d",
146 			    sdl->sdl_family));
147 			if (rt->rt_refcnt > 1) {
148 				sdl->sdl_alen = 0;
149 				la->la_preempt = la->la_asked = 0;
150 				RT_UNLOCK(rt);
151 				continue;
152 			}
153 			RT_UNLOCK(rt);
154 			/*
155 			 * XXX: LIST_REMOVE() is deep inside rtrequest().
156 			 */
157 			rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0,
158 			    NULL);
159 			continue;
160 		}
161 		RT_UNLOCK(rt);
162 	}
163 	RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]);
164 
165 	callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL);
166 }
167 
168 /*
169  * Parallel to llc_rtrequest.
170  */
171 static void
172 arp_rtrequest(req, rt, info)
173 	int req;
174 	struct rtentry *rt;
175 	struct rt_addrinfo *info;
176 {
177 	struct sockaddr *gate;
178 	struct llinfo_arp *la;
179 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
180 	struct in_ifaddr *ia;
181 	struct ifaddr *ifa;
182 
183 	RT_LOCK_ASSERT(rt);
184 
185 	if (rt->rt_flags & RTF_GATEWAY)
186 		return;
187 	gate = rt->rt_gateway;
188 	la = (struct llinfo_arp *)rt->rt_llinfo;
189 	switch (req) {
190 
191 	case RTM_ADD:
192 		/*
193 		 * XXX: If this is a manually added route to interface
194 		 * such as older version of routed or gated might provide,
195 		 * restore cloning bit.
196 		 */
197 		if ((rt->rt_flags & RTF_HOST) == 0 &&
198 		    rt_mask(rt) != NULL &&
199 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
200 			rt->rt_flags |= RTF_CLONING;
201 		if (rt->rt_flags & RTF_CLONING) {
202 			/*
203 			 * Case 1: This route should come from a route to iface.
204 			 */
205 			rt_setgate(rt, rt_key(rt),
206 					(struct sockaddr *)&null_sdl);
207 			gate = rt->rt_gateway;
208 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
209 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
210 			rt->rt_expire = time_uptime;
211 			break;
212 		}
213 		/* Announce a new entry if requested. */
214 		if (rt->rt_flags & RTF_ANNOUNCE)
215 			arprequest(rt->rt_ifp,
216 			    &SIN(rt_key(rt))->sin_addr,
217 			    &SIN(rt_key(rt))->sin_addr,
218 			    (u_char *)LLADDR(SDL(gate)));
219 		/*FALLTHROUGH*/
220 	case RTM_RESOLVE:
221 		if (gate->sa_family != AF_LINK ||
222 		    gate->sa_len < sizeof(null_sdl)) {
223 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
224 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
225 			    (gate->sa_family != AF_LINK) ?
226 			    " (!AF_LINK)": "");
227 			break;
228 		}
229 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
230 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
231 		if (la != 0)
232 			break; /* This happens on a route change */
233 		/*
234 		 * Case 2:  This route may come from cloning, or a manual route
235 		 * add with a LL address.
236 		 */
237 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
238 		rt->rt_llinfo = (caddr_t)la;
239 		if (la == 0) {
240 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
241 			break;
242 		}
243 		arp_allocated++;
244 		/*
245 		 * We are storing a route entry outside of radix tree. So,
246 		 * it can be found and accessed by other means than radix
247 		 * lookup. The routing code assumes that any rtentry detached
248 		 * from radix can be destroyed safely. To prevent this, we
249 		 * add an additional reference.
250 		 */
251 		RT_ADDREF(rt);
252 		la->la_rt = rt;
253 		rt->rt_flags |= RTF_LLINFO;
254 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
255 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
256 
257 #ifdef INET
258 		/*
259 		 * This keeps the multicast addresses from showing up
260 		 * in `arp -a' listings as unresolved.  It's not actually
261 		 * functional.  Then the same for broadcast.
262 		 */
263 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
264 		    rt->rt_ifp->if_type != IFT_ARCNET) {
265 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
266 					       LLADDR(SDL(gate)));
267 			SDL(gate)->sdl_alen = 6;
268 			rt->rt_expire = 0;
269 		}
270 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
271 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
272 			       rt->rt_ifp->if_addrlen);
273 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
274 			rt->rt_expire = 0;
275 		}
276 #endif
277 
278 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
279 			if (ia->ia_ifp == rt->rt_ifp &&
280 			    SIN(rt_key(rt))->sin_addr.s_addr ==
281 			    (IA_SIN(ia))->sin_addr.s_addr)
282 				break;
283 		}
284 		if (ia) {
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 loopback down.
290 		     * However, it causes problems during network configuration
291 		     * for boards that can't receive packets they send.
292 		     * It is now necessary to clear "useloopback" and remove
293 		     * the route to force traffic out to 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 				rt->rt_rmx.rmx_mtu = loif->if_mtu;
301 			}
302 
303 		    /*
304 		     * make sure to set rt->rt_ifa to the interface
305 		     * address we are using, otherwise we will have trouble
306 		     * with source address selection.
307 		     */
308 			ifa = &ia->ia_ifa;
309 			if (ifa != rt->rt_ifa) {
310 				IFAFREE(rt->rt_ifa);
311 				IFAREF(ifa);
312 				rt->rt_ifa = ifa;
313 			}
314 		}
315 		break;
316 
317 	case RTM_DELETE:
318 		if (la == 0)
319 			break;
320 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
321 		LIST_REMOVE(la, la_le);
322 		RT_REMREF(rt);
323 		rt->rt_llinfo = 0;
324 		rt->rt_flags &= ~RTF_LLINFO;
325 		if (la->la_hold)
326 			m_freem(la->la_hold);
327 		Free((caddr_t)la);
328 	}
329 }
330 
331 /*
332  * Broadcast an ARP request. Caller specifies:
333  *	- arp header source ip address
334  *	- arp header target ip address
335  *	- arp header source ethernet address
336  */
337 static void
338 arprequest(ifp, sip, tip, enaddr)
339 	struct ifnet *ifp;
340 	struct in_addr *sip, *tip;
341 	u_char *enaddr;
342 {
343 	struct mbuf *m;
344 	struct arphdr *ah;
345 	struct sockaddr sa;
346 
347 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
348 		return;
349 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
350 		2*ifp->if_data.ifi_addrlen;
351 	m->m_pkthdr.len = m->m_len;
352 	MH_ALIGN(m, m->m_len);
353 	ah = mtod(m, struct arphdr *);
354 	bzero((caddr_t)ah, m->m_len);
355 #ifdef MAC
356 	mac_create_mbuf_linklayer(ifp, m);
357 #endif
358 	ah->ar_pro = htons(ETHERTYPE_IP);
359 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
360 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
361 	ah->ar_op = htons(ARPOP_REQUEST);
362 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
363 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
364 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
365 	sa.sa_family = AF_ARP;
366 	sa.sa_len = 2;
367 	m->m_flags |= M_BCAST;
368 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
369 
370 	return;
371 }
372 
373 /*
374  * Resolve an IP address into an ethernet address.
375  * On input:
376  *    ifp is the interface we use
377  *    dst is the next hop,
378  *    rt0 is the route to the final destination (possibly useless)
379  *    m is the mbuf
380  *    desten is where we want the address.
381  *
382  * On success, desten is filled in and the function returns 0;
383  * If the packet must be held pending resolution, we return EWOULDBLOCK
384  * On other errors, we return the corresponding error code.
385  */
386 int
387 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
388 	struct sockaddr *dst, u_char *desten)
389 {
390 	struct llinfo_arp *la = NULL;
391 	struct rtentry *rt = NULL;
392 	struct sockaddr_dl *sdl;
393 	int error;
394 
395 	if (m->m_flags & M_BCAST) {	/* broadcast */
396 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
397 		return (0);
398 	}
399 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
400 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
401 		return (0);
402 	}
403 
404 	if (rt0 != NULL) {
405 		error = rt_check(&rt, &rt0, dst);
406 		if (error) {
407 			m_freem(m);
408 			return error;
409 		}
410 		la = (struct llinfo_arp *)rt->rt_llinfo;
411 		if (la == NULL)
412 			RT_UNLOCK(rt);
413 	}
414 	if (la == NULL) {
415 		/*
416 		 * We enter this block in case if rt0 was NULL,
417 		 * or if rt found by rt_check() didn't have llinfo.
418 		 */
419 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
420 		if (rt == NULL) {
421 			log(LOG_DEBUG,
422 			    "arpresolve: can't allocate route for %s\n",
423 			    inet_ntoa(SIN(dst)->sin_addr));
424 			m_freem(m);
425 			return (EINVAL); /* XXX */
426 		}
427 		la = (struct llinfo_arp *)rt->rt_llinfo;
428 		if (la == NULL) {
429 			RT_UNLOCK(rt);
430 			log(LOG_DEBUG,
431 			    "arpresolve: can't allocate llinfo for %s\n",
432 			    inet_ntoa(SIN(dst)->sin_addr));
433 			m_freem(m);
434 			return (EINVAL); /* XXX */
435 		}
436 	}
437 	sdl = SDL(rt->rt_gateway);
438 	/*
439 	 * Check the address family and length is valid, the address
440 	 * is resolved; otherwise, try to resolve.
441 	 */
442 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
443 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
444 
445 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
446 
447 		/*
448 		 * If entry has an expiry time and it is approaching,
449 		 * send an ARP request.
450 		 */
451 		if ((rt->rt_expire != 0) &&
452 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
453 			struct in_addr sin =
454 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
455 
456 			la->la_preempt--;
457 			RT_UNLOCK(rt);
458 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
459 			    IF_LLADDR(ifp));
460 			return (0);
461 		}
462 
463 		RT_UNLOCK(rt);
464 		return (0);
465 	}
466 	/*
467 	 * If ARP is disabled or static on this interface, stop.
468 	 * XXX
469 	 * Probably should not allocate empty llinfo struct if we are
470 	 * not going to be sending out an arp request.
471 	 */
472 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
473 		RT_UNLOCK(rt);
474 		m_freem(m);
475 		return (EINVAL);
476 	}
477 	/*
478 	 * There is an arptab entry, but no ethernet address
479 	 * response yet.  Replace the held mbuf with this
480 	 * latest one.
481 	 */
482 	if (la->la_hold)
483 		m_freem(la->la_hold);
484 	la->la_hold = m;
485 
486 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
487 
488 	/*
489 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
490 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
491 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
492 	 * ARP request, but not faster than one request per second.
493 	 */
494 	if (la->la_asked < arp_maxtries)
495 		error = EWOULDBLOCK;	/* First request. */
496 	else
497 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
498 
499 	if (la->la_asked++ == 0 || rt->rt_expire != time_uptime) {
500 		struct in_addr sin =
501 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
502 
503 		rt->rt_expire = time_uptime;
504 		RT_UNLOCK(rt);
505 
506 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
507 		    IF_LLADDR(ifp));
508 	} else
509 		RT_UNLOCK(rt);
510 
511 	return (error);
512 }
513 
514 /*
515  * Common length and type checks are done here,
516  * then the protocol-specific routine is called.
517  */
518 static void
519 arpintr(struct mbuf *m)
520 {
521 	struct arphdr *ar;
522 
523 	if (m->m_len < sizeof(struct arphdr) &&
524 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
525 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
526 		return;
527 	}
528 	ar = mtod(m, struct arphdr *);
529 
530 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
531 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
532 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
533 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
534 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
535 		    (unsigned char *)&ar->ar_hrd, "");
536 		m_freem(m);
537 		return;
538 	}
539 
540 	if (m->m_len < arphdr_len(ar)) {
541 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
542 			log(LOG_ERR, "arp: runt packet\n");
543 			m_freem(m);
544 			return;
545 		}
546 		ar = mtod(m, struct arphdr *);
547 	}
548 
549 	switch (ntohs(ar->ar_pro)) {
550 #ifdef INET
551 	case ETHERTYPE_IP:
552 		in_arpinput(m);
553 		return;
554 #endif
555 	}
556 	m_freem(m);
557 }
558 
559 #ifdef INET
560 /*
561  * ARP for Internet protocols on 10 Mb/s Ethernet.
562  * Algorithm is that given in RFC 826.
563  * In addition, a sanity check is performed on the sender
564  * protocol address, to catch impersonators.
565  * We no longer handle negotiations for use of trailer protocol:
566  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
567  * along with IP replies if we wanted trailers sent to us,
568  * and also sent them in response to IP replies.
569  * This allowed either end to announce the desire to receive
570  * trailer packets.
571  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
572  * but formerly didn't normally send requests.
573  */
574 static int log_arp_wrong_iface = 1;
575 static int log_arp_movements = 1;
576 static int log_arp_permanent_modify = 1;
577 
578 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
579 	&log_arp_wrong_iface, 0,
580 	"log arp packets arriving on the wrong interface");
581 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
582         &log_arp_movements, 0,
583         "log arp replies from MACs different than the one in the cache");
584 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
585         &log_arp_permanent_modify, 0,
586         "log arp replies from MACs different than the one in the permanent arp entry");
587 
588 
589 static void
590 in_arpinput(m)
591 	struct mbuf *m;
592 {
593 	struct arphdr *ah;
594 	struct ifnet *ifp = m->m_pkthdr.rcvif;
595 	struct llinfo_arp *la;
596 	struct rtentry *rt;
597 	struct ifaddr *ifa;
598 	struct in_ifaddr *ia;
599 	struct sockaddr_dl *sdl;
600 	struct sockaddr sa;
601 	struct in_addr isaddr, itaddr, myaddr;
602 	struct mbuf *hold;
603 	u_int8_t *enaddr = NULL;
604 	int op, rif_len;
605 	int req_len;
606 	int bridged = 0;
607 #ifdef DEV_CARP
608 	int carp_match = 0;
609 #endif
610 
611 	if (ifp->if_bridge)
612 		bridged = 1;
613 
614 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
615 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
616 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
617 		return;
618 	}
619 
620 	ah = mtod(m, struct arphdr *);
621 	op = ntohs(ah->ar_op);
622 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
623 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
624 
625 	/*
626 	 * For a bridge, we want to check the address irrespective
627 	 * of the receive interface. (This will change slightly
628 	 * when we have clusters of interfaces).
629 	 * If the interface does not match, but the recieving interface
630 	 * is part of carp, we call carp_iamatch to see if this is a
631 	 * request for the virtual host ip.
632 	 * XXX: This is really ugly!
633 	 */
634 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
635 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
636 		    (ia->ia_ifp == ifp)) &&
637 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
638 			goto match;
639 #ifdef DEV_CARP
640 		if (ifp->if_carp != NULL &&
641 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
642 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
643 			carp_match = 1;
644 			goto match;
645 		}
646 #endif
647 	}
648 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
649 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
650 		    (ia->ia_ifp == ifp)) &&
651 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
652 			goto match;
653 	/*
654 	 * No match, use the first inet address on the receive interface
655 	 * as a dummy address for the rest of the function.
656 	 */
657 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
658 		if (ifa->ifa_addr->sa_family == AF_INET) {
659 			ia = ifatoia(ifa);
660 			goto match;
661 		}
662 	/*
663 	 * If bridging, fall back to using any inet address.
664 	 */
665 	if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL)
666 		goto drop;
667 match:
668 	if (!enaddr)
669 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
670 	myaddr = ia->ia_addr.sin_addr;
671 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
672 		goto drop;	/* it's from me, ignore it. */
673 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
674 		log(LOG_ERR,
675 		    "arp: link address is broadcast for IP address %s!\n",
676 		    inet_ntoa(isaddr));
677 		goto drop;
678 	}
679 	/*
680 	 * Warn if another host is using the same IP address, but only if the
681 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
682 	 * case we suppress the warning to avoid false positive complaints of
683 	 * potential misconfiguration.
684 	 */
685 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
686 		log(LOG_ERR,
687 		   "arp: %*D is using my IP address %s!\n",
688 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
689 		   inet_ntoa(isaddr));
690 		itaddr = myaddr;
691 		goto reply;
692 	}
693 	if (ifp->if_flags & IFF_STATICARP)
694 		goto reply;
695 	rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
696 	if (rt != NULL) {
697 		la = (struct llinfo_arp *)rt->rt_llinfo;
698 		if (la == NULL) {
699 			RT_UNLOCK(rt);
700 			goto reply;
701 		}
702 	} else
703 		goto reply;
704 
705 	/* The following is not an error when doing bridging. */
706 	if (!bridged && rt->rt_ifp != ifp
707 #ifdef DEV_CARP
708 	    && (ifp->if_type != IFT_CARP || !carp_match)
709 #endif
710 							) {
711 		if (log_arp_wrong_iface)
712 			log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
713 			    inet_ntoa(isaddr),
714 			    rt->rt_ifp->if_xname,
715 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
716 			    ifp->if_xname);
717 		RT_UNLOCK(rt);
718 		goto reply;
719 	}
720 	sdl = SDL(rt->rt_gateway);
721 	if (sdl->sdl_alen &&
722 	    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
723 		if (rt->rt_expire) {
724 		    if (log_arp_movements)
725 		        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
726 			    inet_ntoa(isaddr),
727 			    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
728 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
729 			    ifp->if_xname);
730 		} else {
731 			RT_UNLOCK(rt);
732 			if (log_arp_permanent_modify)
733 				log(LOG_ERR, "arp: %*D attempts to modify "
734 				    "permanent entry for %s on %s\n",
735 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
736 				    inet_ntoa(isaddr), ifp->if_xname);
737 			goto reply;
738 		}
739 	}
740 	/*
741 	 * sanity check for the address length.
742 	 * XXX this does not work for protocols with variable address
743 	 * length. -is
744 	 */
745 	if (sdl->sdl_alen &&
746 	    sdl->sdl_alen != ah->ar_hln) {
747 		log(LOG_WARNING,
748 		    "arp from %*D: new addr len %d, was %d",
749 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
750 		    ah->ar_hln, sdl->sdl_alen);
751 	}
752 	if (ifp->if_addrlen != ah->ar_hln) {
753 		log(LOG_WARNING,
754 		    "arp from %*D: addr len: new %d, i/f %d (ignored)",
755 		    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
756 		    ah->ar_hln, ifp->if_addrlen);
757 		RT_UNLOCK(rt);
758 		goto reply;
759 	}
760 	(void)memcpy(LLADDR(sdl), ar_sha(ah),
761 	    sdl->sdl_alen = ah->ar_hln);
762 	/*
763 	 * If we receive an arp from a token-ring station over
764 	 * a token-ring nic then try to save the source
765 	 * routing info.
766 	 */
767 	if (ifp->if_type == IFT_ISO88025) {
768 		struct iso88025_header *th = NULL;
769 		struct iso88025_sockaddr_dl_data *trld;
770 
771 		th = (struct iso88025_header *)m->m_pkthdr.header;
772 		trld = SDL_ISO88025(sdl);
773 		rif_len = TR_RCF_RIFLEN(th->rcf);
774 		if ((th->iso88025_shost[0] & TR_RII) &&
775 		    (rif_len > 2)) {
776 			trld->trld_rcf = th->rcf;
777 			trld->trld_rcf ^= htons(TR_RCF_DIR);
778 			memcpy(trld->trld_route, th->rd, rif_len - 2);
779 			trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
780 			/*
781 			 * Set up source routing information for
782 			 * reply packet (XXX)
783 			 */
784 			m->m_data -= rif_len;
785 			m->m_len  += rif_len;
786 			m->m_pkthdr.len += rif_len;
787 		} else {
788 			th->iso88025_shost[0] &= ~TR_RII;
789 			trld->trld_rcf = 0;
790 		}
791 		m->m_data -= 8;
792 		m->m_len  += 8;
793 		m->m_pkthdr.len += 8;
794 		th->rcf = trld->trld_rcf;
795 	}
796 	if (rt->rt_expire)
797 		rt->rt_expire = time_uptime + arpt_keep;
798 	la->la_asked = 0;
799 	la->la_preempt = arp_maxtries;
800 	hold = la->la_hold;
801 	la->la_hold = NULL;
802 	RT_UNLOCK(rt);
803 	if (hold != NULL)
804 		(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
805 
806 reply:
807 	if (op != ARPOP_REQUEST)
808 		goto drop;
809 	if (itaddr.s_addr == myaddr.s_addr) {
810 		/* I am the target */
811 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
812 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
813 	} else {
814 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY);
815 		if (rt == NULL) {
816 			struct sockaddr_in sin;
817 
818 			if (!arp_proxyall)
819 				goto drop;
820 
821 			bzero(&sin, sizeof sin);
822 			sin.sin_family = AF_INET;
823 			sin.sin_len = sizeof sin;
824 			sin.sin_addr = itaddr;
825 
826 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
827 			if (!rt)
828 				goto drop;
829 			/*
830 			 * Don't send proxies for nodes on the same interface
831 			 * as this one came out of, or we'll get into a fight
832 			 * over who claims what Ether address.
833 			 */
834 			if (rt->rt_ifp == ifp) {
835 				rtfree(rt);
836 				goto drop;
837 			}
838 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
839 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
840 			rtfree(rt);
841 
842 			/*
843 			 * Also check that the node which sent the ARP packet
844 			 * is on the the interface we expect it to be on. This
845 			 * avoids ARP chaos if an interface is connected to the
846 			 * wrong network.
847 			 */
848 			sin.sin_addr = isaddr;
849 
850 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
851 			if (!rt)
852 				goto drop;
853 			if (rt->rt_ifp != ifp) {
854 				log(LOG_INFO, "arp_proxy: ignoring request"
855 				    " from %s via %s, expecting %s\n",
856 				    inet_ntoa(isaddr), ifp->if_xname,
857 				    rt->rt_ifp->if_xname);
858 				rtfree(rt);
859 				goto drop;
860 			}
861 			rtfree(rt);
862 
863 #ifdef DEBUG_PROXY
864 			printf("arp: proxying for %s\n",
865 			       inet_ntoa(itaddr));
866 #endif
867 		} else {
868 			/*
869 			 * Return proxied ARP replies only on the interface
870 			 * or bridge cluster where this network resides.
871 			 * Otherwise we may conflict with the host we are
872 			 * proxying for.
873 			 */
874 			if (rt->rt_ifp != ifp &&
875 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
876 			    ifp->if_bridge == NULL)) {
877 				RT_UNLOCK(rt);
878 				goto drop;
879 			}
880 			sdl = SDL(rt->rt_gateway);
881 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
882 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
883 			RT_UNLOCK(rt);
884 		}
885 	}
886 
887 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
888 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
889 	ah->ar_op = htons(ARPOP_REPLY);
890 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
891 	m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
892 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
893 	m->m_pkthdr.len = m->m_len;
894 	sa.sa_family = AF_ARP;
895 	sa.sa_len = 2;
896 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
897 	return;
898 
899 drop:
900 	m_freem(m);
901 }
902 #endif
903 
904 /*
905  * Lookup or enter a new address in arptab.
906  */
907 static struct rtentry *
908 arplookup(addr, create, proxy)
909 	u_long addr;
910 	int create, proxy;
911 {
912 	struct rtentry *rt;
913 	struct sockaddr_inarp sin;
914 	const char *why = 0;
915 
916 	bzero(&sin, sizeof(sin));
917 	sin.sin_len = sizeof(sin);
918 	sin.sin_family = AF_INET;
919 	sin.sin_addr.s_addr = addr;
920 	if (proxy)
921 		sin.sin_other = SIN_PROXY;
922 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
923 	if (rt == 0)
924 		return (0);
925 
926 	if (rt->rt_flags & RTF_GATEWAY)
927 		why = "host is not on local network";
928 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
929 		why = "could not allocate llinfo";
930 	else if (rt->rt_gateway->sa_family != AF_LINK)
931 		why = "gateway route is not ours";
932 
933 	if (why) {
934 #define	ISDYNCLONE(_rt) \
935 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
936 		if (create)
937 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
938 			    inet_ntoa(sin.sin_addr), why);
939 		/*
940 		 * If there are no references to this Layer 2 route,
941 		 * and it is a cloned route, and not static, and
942 		 * arplookup() is creating the route, then purge
943 		 * it from the routing table as it is probably bogus.
944 		 */
945 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
946 			rtexpunge(rt);
947 		RTFREE_LOCKED(rt);
948 		return (0);
949 #undef ISDYNCLONE
950 	} else {
951 		RT_REMREF(rt);
952 		return (rt);
953 	}
954 }
955 
956 void
957 arp_ifinit(ifp, ifa)
958 	struct ifnet *ifp;
959 	struct ifaddr *ifa;
960 {
961 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
962 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
963 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
964 	ifa->ifa_rtrequest = arp_rtrequest;
965 	ifa->ifa_flags |= RTF_CLONING;
966 }
967 
968 void
969 arp_ifinit2(ifp, ifa, enaddr)
970 	struct ifnet *ifp;
971 	struct ifaddr *ifa;
972 	u_char *enaddr;
973 {
974 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
975 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
976 				&IA_SIN(ifa)->sin_addr, enaddr);
977 	ifa->ifa_rtrequest = arp_rtrequest;
978 	ifa->ifa_flags |= RTF_CLONING;
979 }
980 
981 static void
982 arp_init(void)
983 {
984 
985 	arpintrq.ifq_maxlen = 50;
986 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
987 	LIST_INIT(&llinfo_arp);
988 	callout_init(&arp_callout, CALLOUT_MPSAFE);
989 	netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE);
990 	callout_reset(&arp_callout, hz, arptimer, NULL);
991 }
992 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
993