xref: /freebsd/sys/netinet/if_ether.c (revision 54ebdd631db8c0bba2baab0155f603a8b5cf014a)
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  */
31 
32 /*
33  * Ethernet address resolution protocol.
34  * TODO:
35  *	add "inuse/lock" bit (or ref. count) along with valid bit
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 #include "opt_mac.h"
43 #include "opt_carp.h"
44 
45 #include <sys/param.h>
46 #include <sys/kernel.h>
47 #include <sys/queue.h>
48 #include <sys/sysctl.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/malloc.h>
52 #include <sys/proc.h>
53 #include <sys/socket.h>
54 #include <sys/syslog.h>
55 #include <sys/vimage.h>
56 
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/route.h>
61 #include <net/netisr.h>
62 #include <net/if_llc.h>
63 #include <net/ethernet.h>
64 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/vinet.h>
70 
71 #include <net/if_arc.h>
72 #include <net/iso88025.h>
73 
74 #ifdef DEV_CARP
75 #include <netinet/ip_carp.h>
76 #endif
77 
78 #include <security/mac/mac_framework.h>
79 
80 #define SIN(s) ((struct sockaddr_in *)s)
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 SYSCTL_DECL(_net_link_ether);
84 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
85 
86 /* timer values */
87 #ifdef VIMAGE_GLOBALS
88 static int	arpt_keep; /* once resolved, good for 20 more minutes */
89 static int	arp_maxtries;
90 static int	useloopback; /* use loopback interface for local traffic */
91 static int	arp_proxyall;
92 #endif
93 
94 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, max_age,
95     CTLFLAG_RW, arpt_keep, 0, "ARP entry lifetime in seconds");
96 
97 #define	rt_expire rt_rmx.rmx_expire
98 
99 struct llinfo_arp {
100 	struct	callout la_timer;
101 	struct	rtentry *la_rt;
102 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
103 	u_short	la_preempt;	/* countdown for pre-expiry arps */
104 	u_short	la_asked;	/* # requests sent */
105 };
106 
107 static struct	ifqueue arpintrq;
108 
109 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries,
110 	CTLFLAG_RW, arp_maxtries, 0,
111 	"ARP resolution attempts before returning error");
112 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, useloopback,
113 	CTLFLAG_RW, useloopback, 0,
114 	"Use the loopback interface for local traffic");
115 SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall,
116 	CTLFLAG_RW, arp_proxyall, 0,
117 	"Enable proxy ARP for all suitable requests");
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, int);
127 #ifdef INET
128 static void	in_arpinput(struct mbuf *);
129 #endif
130 
131 /*
132  * Timeout routine.
133  */
134 static void
135 arptimer(void *arg)
136 {
137 	struct rtentry *rt = (struct rtentry *)arg;
138 
139 	RT_LOCK_ASSERT(rt);
140 	/*
141 	 * The lock is needed to close a theoretical race
142 	 * between spontaneous expiry and intentional removal.
143 	 * We still got an extra reference on rtentry, so can
144 	 * safely pass pointers to its contents.
145 	 */
146 	RT_UNLOCK(rt);
147 
148 	in_rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL,
149 	    rt->rt_fibnum);
150 }
151 
152 /*
153  * Parallel to llc_rtrequest.
154  */
155 static void
156 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
157 {
158 	INIT_VNET_NET(curvnet);
159 	INIT_VNET_INET(curvnet);
160 	struct sockaddr *gate;
161 	struct llinfo_arp *la;
162 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
163 	struct in_ifaddr *ia;
164 	struct ifaddr *ifa;
165 
166 	RT_LOCK_ASSERT(rt);
167 
168 	if (rt->rt_flags & RTF_GATEWAY)
169 		return;
170 	gate = rt->rt_gateway;
171 	la = (struct llinfo_arp *)rt->rt_llinfo;
172 	switch (req) {
173 
174 	case RTM_ADD:
175 		/*
176 		 * XXX: If this is a manually added route to interface
177 		 * such as older version of routed or gated might provide,
178 		 * restore cloning bit.
179 		 */
180 		if ((rt->rt_flags & RTF_HOST) == 0 &&
181 		    rt_mask(rt) != NULL &&
182 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
183 			rt->rt_flags |= RTF_CLONING;
184 		if (rt->rt_flags & RTF_CLONING) {
185 			/*
186 			 * Case 1: This route should come from a route to iface.
187 			 */
188 			rt_setgate(rt, rt_key(rt),
189 					(struct sockaddr *)&null_sdl);
190 			gate = rt->rt_gateway;
191 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
192 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
193 			rt->rt_expire = time_uptime;
194 			break;
195 		}
196 		/* Announce a new entry if requested. */
197 		if (rt->rt_flags & RTF_ANNOUNCE)
198 			arprequest(rt->rt_ifp,
199 			    &SIN(rt_key(rt))->sin_addr,
200 			    &SIN(rt_key(rt))->sin_addr,
201 			    (u_char *)LLADDR(SDL(gate)));
202 		/*FALLTHROUGH*/
203 	case RTM_RESOLVE:
204 		if (gate->sa_family != AF_LINK ||
205 		    gate->sa_len < sizeof(null_sdl)) {
206 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
207 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
208 			    (gate->sa_family != AF_LINK) ?
209 			    " (!AF_LINK)": "");
210 			break;
211 		}
212 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
213 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
214 		if (la != 0)
215 			break; /* This happens on a route change */
216 		/*
217 		 * Case 2:  This route may come from cloning, or a manual route
218 		 * add with a LL address.
219 		 */
220 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
221 		rt->rt_llinfo = (caddr_t)la;
222 		if (la == 0) {
223 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
224 			break;
225 		}
226 		/*
227 		 * We are storing a route entry outside of radix tree. So,
228 		 * it can be found and accessed by other means than radix
229 		 * lookup. The routing code assumes that any rtentry detached
230 		 * from radix can be destroyed safely. To prevent this, we
231 		 * add an additional reference.
232 		 */
233 		RT_ADDREF(rt);
234 		la->la_rt = rt;
235 		rt->rt_flags |= RTF_LLINFO;
236 		callout_init_mtx(&la->la_timer, &rt->rt_mtx,
237 		    CALLOUT_RETURNUNLOCKED);
238 
239 #ifdef INET
240 		/*
241 		 * This keeps the multicast addresses from showing up
242 		 * in `arp -a' listings as unresolved.  It's not actually
243 		 * functional.  Then the same for broadcast.
244 		 */
245 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
246 		    rt->rt_ifp->if_type != IFT_ARCNET) {
247 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
248 					       LLADDR(SDL(gate)));
249 			SDL(gate)->sdl_alen = 6;
250 			rt->rt_expire = 0;
251 		}
252 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
253 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
254 			       rt->rt_ifp->if_addrlen);
255 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
256 			rt->rt_expire = 0;
257 		}
258 #endif
259 
260 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
261 			if (ia->ia_ifp == rt->rt_ifp &&
262 			    SIN(rt_key(rt))->sin_addr.s_addr ==
263 			    (IA_SIN(ia))->sin_addr.s_addr)
264 				break;
265 		}
266 		if (ia) {
267 		    /*
268 		     * This test used to be
269 		     *	if (loif.if_flags & IFF_UP)
270 		     * It allowed local traffic to be forced
271 		     * through the hardware by configuring the loopback down.
272 		     * However, it causes problems during network configuration
273 		     * for boards that can't receive packets they send.
274 		     * It is now necessary to clear "useloopback" and remove
275 		     * the route to force traffic out to the hardware.
276 		     */
277 			rt->rt_expire = 0;
278 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
279 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
280 			if (V_useloopback) {
281 				rt->rt_ifp = V_loif;
282 				rt->rt_rmx.rmx_mtu = V_loif->if_mtu;
283 			}
284 
285 		    /*
286 		     * make sure to set rt->rt_ifa to the interface
287 		     * address we are using, otherwise we will have trouble
288 		     * with source address selection.
289 		     */
290 			ifa = &ia->ia_ifa;
291 			if (ifa != rt->rt_ifa) {
292 				IFAFREE(rt->rt_ifa);
293 				IFAREF(ifa);
294 				rt->rt_ifa = ifa;
295 			}
296 		}
297 		break;
298 
299 	case RTM_DELETE:
300 		if (la == NULL)	/* XXX: at least CARP does this. */
301 			break;
302 		callout_stop(&la->la_timer);
303 		rt->rt_llinfo = NULL;
304 		rt->rt_flags &= ~RTF_LLINFO;
305 		RT_REMREF(rt);
306 		if (la->la_hold)
307 			m_freem(la->la_hold);
308 		Free((caddr_t)la);
309 	}
310 }
311 
312 /*
313  * Broadcast an ARP request. Caller specifies:
314  *	- arp header source ip address
315  *	- arp header target ip address
316  *	- arp header source ethernet address
317  */
318 static void
319 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
320     u_char *enaddr)
321 {
322 	struct mbuf *m;
323 	struct arphdr *ah;
324 	struct sockaddr sa;
325 
326 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
327 		return;
328 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
329 		2*ifp->if_data.ifi_addrlen;
330 	m->m_pkthdr.len = m->m_len;
331 	MH_ALIGN(m, m->m_len);
332 	ah = mtod(m, struct arphdr *);
333 	bzero((caddr_t)ah, m->m_len);
334 #ifdef MAC
335 	mac_netinet_arp_send(ifp, m);
336 #endif
337 	ah->ar_pro = htons(ETHERTYPE_IP);
338 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
339 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
340 	ah->ar_op = htons(ARPOP_REQUEST);
341 	bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln);
342 	bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln);
343 	bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln);
344 	sa.sa_family = AF_ARP;
345 	sa.sa_len = 2;
346 	m->m_flags |= M_BCAST;
347 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
348 
349 	return;
350 }
351 
352 /*
353  * Resolve an IP address into an ethernet address.
354  * On input:
355  *    ifp is the interface we use
356  *    rt0 is the route to the final destination (possibly useless)
357  *    m is the mbuf. May be NULL if we don't have a packet.
358  *    dst is the next hop,
359  *    desten is where we want the address.
360  *
361  * On success, desten is filled in and the function returns 0;
362  * If the packet must be held pending resolution, we return EWOULDBLOCK
363  * On other errors, we return the corresponding error code.
364  * Note that m_freem() handles NULL.
365  */
366 int
367 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
368     struct sockaddr *dst, u_char *desten)
369 {
370 	INIT_VNET_INET(ifp->if_vnet);
371 	struct llinfo_arp *la = NULL;
372 	struct rtentry *rt = NULL;
373 	struct sockaddr_dl *sdl;
374 	int error;
375 	int fibnum = -1;
376 
377 	if (m) {
378 		if (m->m_flags & M_BCAST) {
379 			/* broadcast */
380 			(void)memcpy(desten,
381 			    ifp->if_broadcastaddr, ifp->if_addrlen);
382 			return (0);
383 		}
384 		if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
385 			/* multicast */
386 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
387 			return (0);
388 		}
389 		fibnum = M_GETFIB(m);
390 	}
391 
392 	if (rt0 != NULL) {
393 		/* Look for a cached arp (ll) entry. */
394 		if (m == NULL)
395 			fibnum = rt0->rt_fibnum;
396 		error = rt_check(&rt, &rt0, dst);
397 		if (error) {
398 			m_freem(m);
399 			return error;
400 		}
401 		la = (struct llinfo_arp *)rt->rt_llinfo;
402 		if (la == NULL)
403 			RT_UNLOCK(rt);
404 	}
405 
406 	/*
407 	 * If we had no mbuf and no route, then hope the caller
408 	 * has a fib in mind because we are running out of ideas.
409 	 * I think this should not happen in current code.
410 	 * (kmacy would know).
411 	 */
412 	if (fibnum == -1)
413 		fibnum = curthread->td_proc->p_fibnum; /* last gasp */
414 
415 	if (la == NULL) {
416 		/*
417 		 * We enter this block if rt0 was NULL,
418 		 * or if rt found by rt_check() didn't have llinfo.
419 		 * we should get a cloned route, which since it should
420 		 * come from the local interface should have a ll entry.
421 		 * It may be incomplete but that's ok.
422 		 */
423 		rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0, fibnum);
424 		if (rt == NULL) {
425 			log(LOG_DEBUG,
426 			    "arpresolve: can't allocate route for %s\n",
427 			    inet_ntoa(SIN(dst)->sin_addr));
428 			m_freem(m);
429 			return (EINVAL); /* XXX */
430 		}
431 		la = (struct llinfo_arp *)rt->rt_llinfo;
432 		if (la == NULL) {
433 			RT_UNLOCK(rt);
434 			log(LOG_DEBUG,
435 			    "arpresolve: can't allocate llinfo for %s\n",
436 			    inet_ntoa(SIN(dst)->sin_addr));
437 			m_freem(m);
438 			return (EINVAL); /* XXX */
439 		}
440 	}
441 	sdl = SDL(rt->rt_gateway);
442 	/*
443 	 * Check the address family and length is valid, the address
444 	 * is resolved; otherwise, try to resolve.
445 	 */
446 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
447 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
448 
449 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
450 
451 		/*
452 		 * If entry has an expiry time and it is approaching,
453 		 * send an ARP request.
454 		 */
455 		if ((rt->rt_expire != 0) &&
456 		    (time_uptime + la->la_preempt > rt->rt_expire)) {
457 			struct in_addr sin =
458 			    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
459 
460 			la->la_preempt--;
461 			RT_UNLOCK(rt);
462 			arprequest(ifp, &sin, &SIN(dst)->sin_addr,
463 			    IF_LLADDR(ifp));
464 			return (0);
465 		}
466 
467 		RT_UNLOCK(rt);
468 		return (0);
469 	}
470 	/*
471 	 * If ARP is disabled or static on this interface, stop.
472 	 * XXX
473 	 * Probably should not allocate empty llinfo struct if we are
474 	 * not going to be sending out an arp request.
475 	 */
476 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
477 		RT_UNLOCK(rt);
478 		m_freem(m);
479 		return (EINVAL);
480 	}
481 	/*
482 	 * There is an arptab entry, but no ethernet address
483 	 * response yet.  Replace the held mbuf with this
484 	 * latest one.
485 	 */
486 	if (m) {
487 		if (la->la_hold)
488 			m_freem(la->la_hold);
489 		la->la_hold = m;
490 	}
491 	KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry"));
492 
493 	/*
494 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
495 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
496 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
497 	 * ARP request, but not faster than one request per second.
498 	 */
499 	if (la->la_asked < V_arp_maxtries)
500 		error = EWOULDBLOCK;	/* First request. */
501 	else
502 		error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
503 
504 	if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
505 		struct in_addr sin =
506 		    SIN(rt->rt_ifa->ifa_addr)->sin_addr;
507 
508 		rt->rt_expire = time_uptime;
509 		callout_reset(&la->la_timer, hz, arptimer, rt);
510 		la->la_asked++;
511 		RT_UNLOCK(rt);
512 
513 		arprequest(ifp, &sin, &SIN(dst)->sin_addr,
514 		    IF_LLADDR(ifp));
515 	} else
516 		RT_UNLOCK(rt);
517 
518 	return (error);
519 }
520 
521 /*
522  * Common length and type checks are done here,
523  * then the protocol-specific routine is called.
524  */
525 static void
526 arpintr(struct mbuf *m)
527 {
528 	struct arphdr *ar;
529 
530 	if (m->m_len < sizeof(struct arphdr) &&
531 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
532 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
533 		return;
534 	}
535 	ar = mtod(m, struct arphdr *);
536 
537 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
538 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
539 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
540 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) {
541 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
542 		    (unsigned char *)&ar->ar_hrd, "");
543 		m_freem(m);
544 		return;
545 	}
546 
547 	if (m->m_len < arphdr_len(ar)) {
548 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
549 			log(LOG_ERR, "arp: runt packet\n");
550 			m_freem(m);
551 			return;
552 		}
553 		ar = mtod(m, struct arphdr *);
554 	}
555 
556 	switch (ntohs(ar->ar_pro)) {
557 #ifdef INET
558 	case ETHERTYPE_IP:
559 		in_arpinput(m);
560 		return;
561 #endif
562 	}
563 	m_freem(m);
564 }
565 
566 #ifdef INET
567 /*
568  * ARP for Internet protocols on 10 Mb/s Ethernet.
569  * Algorithm is that given in RFC 826.
570  * In addition, a sanity check is performed on the sender
571  * protocol address, to catch impersonators.
572  * We no longer handle negotiations for use of trailer protocol:
573  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
574  * along with IP replies if we wanted trailers sent to us,
575  * and also sent them in response to IP replies.
576  * This allowed either end to announce the desire to receive
577  * trailer packets.
578  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
579  * but formerly didn't normally send requests.
580  */
581 static int log_arp_wrong_iface = 1;
582 static int log_arp_movements = 1;
583 static int log_arp_permanent_modify = 1;
584 
585 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
586 	&log_arp_wrong_iface, 0,
587 	"log arp packets arriving on the wrong interface");
588 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
589         &log_arp_movements, 0,
590         "log arp replies from MACs different than the one in the cache");
591 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
592         &log_arp_permanent_modify, 0,
593         "log arp replies from MACs different than the one in the permanent arp entry");
594 
595 
596 static void
597 in_arpinput(struct mbuf *m)
598 {
599 	struct arphdr *ah;
600 	struct ifnet *ifp = m->m_pkthdr.rcvif;
601 	struct llinfo_arp *la;
602 	struct rtentry *rt;
603 	struct ifaddr *ifa;
604 	struct in_ifaddr *ia;
605 	struct sockaddr_dl *sdl;
606 	struct sockaddr sa;
607 	struct in_addr isaddr, itaddr, myaddr;
608 	struct mbuf *hold;
609 	u_int8_t *enaddr = NULL;
610 	int op, rif_len;
611 	int req_len;
612 	int bridged = 0, is_bridge = 0;
613 	u_int fibnum;
614 	u_int goodfib = 0;
615 	int firstpass = 1;
616 #ifdef DEV_CARP
617 	int carp_match = 0;
618 #endif
619 	struct sockaddr_in sin;
620 	sin.sin_len = sizeof(struct sockaddr_in);
621 	sin.sin_family = AF_INET;
622 	sin.sin_addr.s_addr = 0;
623 	INIT_VNET_INET(ifp->if_vnet);
624 
625 	if (ifp->if_bridge)
626 		bridged = 1;
627 	if (ifp->if_type == IFT_BRIDGE)
628 		is_bridge = 1;
629 
630 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
631 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
632 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
633 		return;
634 	}
635 
636 	ah = mtod(m, struct arphdr *);
637 	op = ntohs(ah->ar_op);
638 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
639 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
640 
641 	/*
642 	 * For a bridge, we want to check the address irrespective
643 	 * of the receive interface. (This will change slightly
644 	 * when we have clusters of interfaces).
645 	 * If the interface does not match, but the recieving interface
646 	 * is part of carp, we call carp_iamatch to see if this is a
647 	 * request for the virtual host ip.
648 	 * XXX: This is really ugly!
649 	 */
650 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
651 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
652 		    (ia->ia_ifp == ifp)) &&
653 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
654 			goto match;
655 #ifdef DEV_CARP
656 		if (ifp->if_carp != NULL &&
657 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
658 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
659 			carp_match = 1;
660 			goto match;
661 		}
662 #endif
663 	}
664 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
665 		if (((bridged && ia->ia_ifp->if_bridge != NULL) ||
666 		    (ia->ia_ifp == ifp)) &&
667 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
668 			goto match;
669 
670 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
671   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
672   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
673   addr == ia->ia_addr.sin_addr.s_addr)
674 	/*
675 	 * Check the case when bridge shares its MAC address with
676 	 * some of its children, so packets are claimed by bridge
677 	 * itself (bridge_input() does it first), but they are really
678 	 * meant to be destined to the bridge member.
679 	 */
680 	if (is_bridge) {
681 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
682 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
683 				ifp = ia->ia_ifp;
684 				goto match;
685 			}
686 		}
687 	}
688 #undef BDG_MEMBER_MATCHES_ARP
689 
690 	/*
691 	 * No match, use the first inet address on the receive interface
692 	 * as a dummy address for the rest of the function.
693 	 */
694 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
695 		if (ifa->ifa_addr->sa_family == AF_INET) {
696 			ia = ifatoia(ifa);
697 			goto match;
698 		}
699 	/*
700 	 * If bridging, fall back to using any inet address.
701 	 */
702 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL)
703 		goto drop;
704 match:
705 	if (!enaddr)
706 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
707 	myaddr = ia->ia_addr.sin_addr;
708 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
709 		goto drop;	/* it's from me, ignore it. */
710 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
711 		log(LOG_ERR,
712 		    "arp: link address is broadcast for IP address %s!\n",
713 		    inet_ntoa(isaddr));
714 		goto drop;
715 	}
716 	/*
717 	 * Warn if another host is using the same IP address, but only if the
718 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
719 	 * case we suppress the warning to avoid false positive complaints of
720 	 * potential misconfiguration.
721 	 */
722 	if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
723 		log(LOG_ERR,
724 		   "arp: %*D is using my IP address %s on %s!\n",
725 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
726 		   inet_ntoa(isaddr), ifp->if_xname);
727 		itaddr = myaddr;
728 		goto reply;
729 	}
730 	if (ifp->if_flags & IFF_STATICARP)
731 		goto reply;
732 	/*
733 	 * We look for any FIB that has this address to find
734 	 * the interface etc.
735 	 * For sanity checks that are FIB independent we abort the loop.
736 	 */
737 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
738 		rt = arplookup(isaddr.s_addr,
739 		    itaddr.s_addr == myaddr.s_addr, 0, fibnum);
740 		if (rt == NULL)
741 			continue;
742 
743 		sdl = SDL(rt->rt_gateway);
744 		/* Only call this once */
745 		if (firstpass) {
746 			sin.sin_addr.s_addr = isaddr.s_addr;
747 			EVENTHANDLER_INVOKE(route_arp_update_event, rt,
748 			    ar_sha(ah), (struct sockaddr *)&sin);
749 		}
750 
751 		la = (struct llinfo_arp *)rt->rt_llinfo;
752 		if (la == NULL) {
753 			RT_UNLOCK(rt);
754 			continue;
755 		}
756 
757 		if (firstpass) {
758 			/* The following is not an error when doing bridging. */
759 			if (!bridged && rt->rt_ifp != ifp
760 #ifdef DEV_CARP
761 			    && (ifp->if_type != IFT_CARP || !carp_match)
762 #endif
763 			    ) {
764 				if (log_arp_wrong_iface)
765 					log(LOG_ERR, "arp: %s is on %s "
766 						"but got reply from %*D "
767 						"on %s\n",
768 					    inet_ntoa(isaddr),
769 					    rt->rt_ifp->if_xname,
770 					    ifp->if_addrlen,
771 					    (u_char *)ar_sha(ah), ":",
772 					    ifp->if_xname);
773 				RT_UNLOCK(rt);
774 				break;
775 			}
776 			if (sdl->sdl_alen &&
777 			    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
778 				if (rt->rt_expire) {
779 				    if (log_arp_movements)
780 					log(LOG_INFO,
781 					    "arp: %s moved from %*D to %*D "
782 					    "on %s\n",
783 					    inet_ntoa(isaddr),
784 					    ifp->if_addrlen,
785 					    (u_char *)LLADDR(sdl), ":",
786 					    ifp->if_addrlen,
787 					    (u_char *)ar_sha(ah), ":",
788 					    ifp->if_xname);
789 				} else {
790 					RT_UNLOCK(rt);
791 					if (log_arp_permanent_modify)
792 						log(LOG_ERR,
793 						    "arp: %*D attempts to "
794 						    "modify permanent entry "
795 						    "for %s on %s\n",
796 						    ifp->if_addrlen,
797 						    (u_char *)ar_sha(ah), ":",
798 						    inet_ntoa(isaddr),
799 						    ifp->if_xname);
800 					break;
801 				}
802 			}
803 			/*
804 			 * sanity check for the address length.
805 			 * XXX this does not work for protocols
806 			 * with variable address length. -is
807 			 */
808 			if (sdl->sdl_alen &&
809 			    sdl->sdl_alen != ah->ar_hln) {
810 				log(LOG_WARNING,
811 				    "arp from %*D: new addr len %d, was %d",
812 				    ifp->if_addrlen, (u_char *) ar_sha(ah),
813 				    ":", ah->ar_hln, sdl->sdl_alen);
814 			}
815 			if (ifp->if_addrlen != ah->ar_hln) {
816 				log(LOG_WARNING,
817 				    "arp from %*D: addr len: "
818 				    "new %d, i/f %d (ignored)",
819 				    ifp->if_addrlen, (u_char *) ar_sha(ah),
820 				    ":", ah->ar_hln, ifp->if_addrlen);
821 				RT_UNLOCK(rt);
822 				break;
823 			}
824 			firstpass = 0;
825 			goodfib = fibnum;
826 		}
827 
828 		/* Copy in the information received. */
829 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
830 		    sdl->sdl_alen = ah->ar_hln);
831 		/*
832 		 * If we receive an arp from a token-ring station over
833 		 * a token-ring nic then try to save the source routing info.
834 		 * XXXMRT Only minimal Token Ring support for MRT.
835 		 * Only do this on the first pass as if modifies the mbuf.
836 		 */
837 		if (ifp->if_type == IFT_ISO88025) {
838 			struct iso88025_header *th = NULL;
839 			struct iso88025_sockaddr_dl_data *trld;
840 
841 			/* force the fib loop to end after this pass */
842 			fibnum = rt_numfibs - 1;
843 
844 			th = (struct iso88025_header *)m->m_pkthdr.header;
845 			trld = SDL_ISO88025(sdl);
846 			rif_len = TR_RCF_RIFLEN(th->rcf);
847 			if ((th->iso88025_shost[0] & TR_RII) &&
848 			    (rif_len > 2)) {
849 				trld->trld_rcf = th->rcf;
850 				trld->trld_rcf ^= htons(TR_RCF_DIR);
851 				memcpy(trld->trld_route, th->rd, rif_len - 2);
852 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
853 				/*
854 				 * Set up source routing information for
855 				 * reply packet (XXX)
856 				 */
857 				m->m_data -= rif_len;
858 				m->m_len  += rif_len;
859 				m->m_pkthdr.len += rif_len;
860 			} else {
861 				th->iso88025_shost[0] &= ~TR_RII;
862 				trld->trld_rcf = 0;
863 			}
864 			m->m_data -= 8;
865 			m->m_len  += 8;
866 			m->m_pkthdr.len += 8;
867 			th->rcf = trld->trld_rcf;
868 		}
869 
870 		if (rt->rt_expire) {
871 			rt->rt_expire = time_uptime + V_arpt_keep;
872 			callout_reset(&la->la_timer, hz * V_arpt_keep,
873 			    arptimer, rt);
874 		}
875 		la->la_asked = 0;
876 		la->la_preempt = V_arp_maxtries;
877 		hold = la->la_hold;
878 		la->la_hold = NULL;
879 		RT_UNLOCK(rt);
880 		if (hold != NULL)
881 			(*ifp->if_output)(ifp, hold, rt_key(rt), rt);
882 	} /* end of FIB loop */
883 reply:
884 
885 	/*
886 	 * Decide if we have to respond to something.
887 	 */
888 	if (op != ARPOP_REQUEST)
889 		goto drop;
890 	if (itaddr.s_addr == myaddr.s_addr) {
891 		/* Shortcut.. the receiving interface is the target. */
892 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
893 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
894 	} else {
895 		/* It's not asking for our address. But it still may
896 		 * be something we should answer.
897 		 *
898 		 * XXX MRT
899 		 * We assume that link level info is independent of
900 		 * the table used and so we use whichever we can and don't
901 		 * have a better option.
902 		 */
903 		/* Have we been asked to proxy for the target. */
904 		rt = arplookup(itaddr.s_addr, 0, SIN_PROXY, goodfib);
905 		if (rt == NULL) {
906 			/* Nope, only intersted now if proxying everything. */
907 			struct sockaddr_in sin;
908 
909 			if (!V_arp_proxyall)
910 				goto drop;
911 
912 			bzero(&sin, sizeof sin);
913 			sin.sin_family = AF_INET;
914 			sin.sin_len = sizeof sin;
915 			sin.sin_addr = itaddr;
916 
917 			/* XXX MRT use table 0 for arp reply  */
918 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
919 			if (!rt)
920 				goto drop;
921 			/*
922 			 * Don't send proxies for nodes on the same interface
923 			 * as this one came out of, or we'll get into a fight
924 			 * over who claims what Ether address.
925 			 */
926 			if (rt->rt_ifp == ifp) {
927 				RTFREE_LOCKED(rt);
928 				goto drop;
929 			}
930 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
931 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
932 			RTFREE_LOCKED(rt);
933 
934 			/*
935 			 * Also check that the node which sent the ARP packet
936 			 * is on the the interface we expect it to be on. This
937 			 * avoids ARP chaos if an interface is connected to the
938 			 * wrong network.
939 			 */
940 			sin.sin_addr = isaddr;
941 
942 			/* XXX MRT use table 0 for arp checks */
943 			rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
944 			if (!rt)
945 				goto drop;
946 			if (rt->rt_ifp != ifp) {
947 				log(LOG_INFO, "arp_proxy: ignoring request"
948 				    " from %s via %s, expecting %s\n",
949 				    inet_ntoa(isaddr), ifp->if_xname,
950 				    rt->rt_ifp->if_xname);
951 				RTFREE_LOCKED(rt);
952 				goto drop;
953 			}
954 			RTFREE_LOCKED(rt);
955 
956 #ifdef DEBUG_PROXY
957 			printf("arp: proxying for %s\n",
958 			       inet_ntoa(itaddr));
959 #endif
960 		} else {
961 			/*
962 			 * Return proxied ARP replies only on the interface
963 			 * or bridge cluster where this network resides.
964 			 * Otherwise we may conflict with the host we are
965 			 * proxying for.
966 			 */
967 			if (rt->rt_ifp != ifp &&
968 			    (rt->rt_ifp->if_bridge != ifp->if_bridge ||
969 			    ifp->if_bridge == NULL)) {
970 				RT_UNLOCK(rt);
971 				goto drop;
972 			}
973 			sdl = SDL(rt->rt_gateway);
974 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
975 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
976 			RT_UNLOCK(rt);
977 		}
978 	}
979 
980 	if (itaddr.s_addr == myaddr.s_addr &&
981 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
982 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
983 #ifdef DEBUG_LINKLOCAL
984 		printf("arp: sending reply for link-local addr %s\n",
985 		    inet_ntoa(itaddr));
986 #endif
987 		m->m_flags |= M_BCAST;
988 		m->m_flags &= ~M_MCAST;
989 	} else {
990 		/* default behaviour; never reply by broadcast. */
991 		m->m_flags &= ~(M_BCAST|M_MCAST);
992 	}
993 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
994 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
995 	ah->ar_op = htons(ARPOP_REPLY);
996 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
997 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
998 	m->m_pkthdr.len = m->m_len;
999 	sa.sa_family = AF_ARP;
1000 	sa.sa_len = 2;
1001 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1002 	return;
1003 
1004 drop:
1005 	m_freem(m);
1006 }
1007 #endif
1008 
1009 /*
1010  * Lookup or enter a new address in arptab.
1011  */
1012 static struct rtentry *
1013 arplookup(u_long addr, int create, int proxy, int fibnum)
1014 {
1015 	struct rtentry *rt;
1016 	struct sockaddr_inarp sin;
1017 	const char *why = 0;
1018 
1019 	bzero(&sin, sizeof(sin));
1020 	sin.sin_len = sizeof(sin);
1021 	sin.sin_family = AF_INET;
1022 	sin.sin_addr.s_addr = addr;
1023 	if (proxy)
1024 		sin.sin_other = SIN_PROXY;
1025 	rt = in_rtalloc1((struct sockaddr *)&sin, create, 0UL, fibnum);
1026 	if (rt == 0)
1027 		return (0);
1028 
1029 	if (rt->rt_flags & RTF_GATEWAY)
1030 		why = "host is not on local network";
1031 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
1032 		why = "could not allocate llinfo";
1033 	else if (rt->rt_gateway->sa_family != AF_LINK)
1034 		why = "gateway route is not ours";
1035 
1036 	if (why) {
1037 #define	ISDYNCLONE(_rt) \
1038 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
1039 		if (create)
1040 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1041 			    inet_ntoa(sin.sin_addr), why);
1042 		/*
1043 		 * If there are no references to this Layer 2 route,
1044 		 * and it is a cloned route, and not static, and
1045 		 * arplookup() is creating the route, then purge
1046 		 * it from the routing table as it is probably bogus.
1047 		 */
1048 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
1049 			rtexpunge(rt);
1050 		RTFREE_LOCKED(rt);
1051 		return (0);
1052 #undef ISDYNCLONE
1053 	} else {
1054 		RT_REMREF(rt);
1055 		return (rt);
1056 	}
1057 }
1058 
1059 void
1060 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1061 {
1062 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
1063 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1064 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1065 	ifa->ifa_rtrequest = arp_rtrequest;
1066 	ifa->ifa_flags |= RTF_CLONING;
1067 }
1068 
1069 void
1070 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
1071 {
1072 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
1073 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1074 				&IA_SIN(ifa)->sin_addr, enaddr);
1075 	ifa->ifa_rtrequest = arp_rtrequest;
1076 	ifa->ifa_flags |= RTF_CLONING;
1077 }
1078 
1079 static void
1080 arp_init(void)
1081 {
1082 	INIT_VNET_INET(curvnet);
1083 
1084 	V_arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
1085 	V_arp_maxtries = 5;
1086 	V_useloopback = 1; /* use loopback interface for local traffic */
1087 	V_arp_proxyall = 0;
1088 
1089 	arpintrq.ifq_maxlen = 50;
1090 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
1091 	netisr_register(NETISR_ARP, arpintr, &arpintrq, 0);
1092 }
1093 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1094