xref: /freebsd/sys/netinet/if_ether.c (revision 7773002178c8dbc52b44e4d705f07706409af8e4)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
34  * $FreeBSD$
35  */
36 
37 /*
38  * Ethernet address resolution protocol.
39  * TODO:
40  *	add "inuse/lock" bit (or ref. count) along with valid bit
41  */
42 
43 #include "opt_inet.h"
44 #include "opt_bdg.h"
45 #include "opt_mac.h"
46 
47 #include <sys/param.h>
48 #include <sys/kernel.h>
49 #include <sys/queue.h>
50 #include <sys/sysctl.h>
51 #include <sys/systm.h>
52 #include <sys/mac.h>
53 #include <sys/mbuf.h>
54 #include <sys/malloc.h>
55 #include <sys/socket.h>
56 #include <sys/syslog.h>
57 
58 #include <net/if.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/route.h>
62 #include <net/netisr.h>
63 #include <net/if_llc.h>
64 #ifdef BRIDGE
65 #include <net/ethernet.h>
66 #include <net/bridge.h>
67 #endif
68 
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/if_ether.h>
72 
73 #include <net/if_arc.h>
74 #include <net/iso88025.h>
75 
76 #define SIN(s) ((struct sockaddr_in *)s)
77 #define SDL(s) ((struct sockaddr_dl *)s)
78 
79 SYSCTL_DECL(_net_link_ether);
80 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 
82 /* timer values */
83 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
84 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
85 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
86 
87 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
88 	   &arpt_prune, 0, "");
89 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
90 	   &arpt_keep, 0, "");
91 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
92 	   &arpt_down, 0, "");
93 
94 #define	rt_expire rt_rmx.rmx_expire
95 
96 struct llinfo_arp {
97 	LIST_ENTRY(llinfo_arp) la_le;
98 	struct	rtentry *la_rt;
99 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
100 	u_short	la_preempt;	/* countdown for pre-expiry arps */
101 	u_short	la_asked;	/* #times we QUERIED following expiration */
102 #define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */
103 };
104 
105 static	LIST_HEAD(, llinfo_arp) llinfo_arp;
106 
107 static struct	ifqueue arpintrq;
108 static int	arp_allocated;
109 static int	arpinit_done;
110 
111 static int	arp_maxtries = 5;
112 static int	useloopback = 1; /* use loopback interface for local traffic */
113 static int	arp_proxyall = 0;
114 static struct callout arp_callout;
115 
116 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
117 	   &arp_maxtries, 0, "");
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
119 	   &useloopback, 0, "");
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
121 	   &arp_proxyall, 0, "");
122 
123 static void	arp_init(void);
124 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
125 static void	arprequest(struct ifnet *,
126 			struct in_addr *, struct in_addr *, u_char *);
127 static void	arpintr(struct mbuf *);
128 static void	arptfree(struct llinfo_arp *);
129 static void	arptimer(void *);
130 static struct llinfo_arp
131 		*arplookup(u_long, int, int);
132 #ifdef INET
133 static void	in_arpinput(struct mbuf *);
134 #endif
135 
136 /*
137  * Timeout routine.  Age arp_tab entries periodically.
138  */
139 /* ARGSUSED */
140 static void
141 arptimer(ignored_arg)
142 	void *ignored_arg;
143 {
144 	struct llinfo_arp *la, *ola;
145 
146 	RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]);
147 	la = LIST_FIRST(&llinfo_arp);
148 	while (la != NULL) {
149 		struct rtentry *rt = la->la_rt;
150 		ola = la;
151 		la = LIST_NEXT(la, la_le);
152 		if (rt->rt_expire && rt->rt_expire <= time_second)
153 			arptfree(ola);		/* timer has expired, clear */
154 	}
155 	RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]);
156 
157 	callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL);
158 }
159 
160 /*
161  * Parallel to llc_rtrequest.
162  */
163 static void
164 arp_rtrequest(req, rt, info)
165 	int req;
166 	register struct rtentry *rt;
167 	struct rt_addrinfo *info;
168 {
169 	register struct sockaddr *gate;
170 	register struct llinfo_arp *la;
171 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
172 
173 	RT_LOCK_ASSERT(rt);
174 
175 	if (!arpinit_done) {
176 		arpinit_done = 1;
177 		callout_reset(&arp_callout, hz, arptimer, NULL);
178 	}
179 	if (rt->rt_flags & RTF_GATEWAY)
180 		return;
181 	gate = rt->rt_gateway;
182 	la = (struct llinfo_arp *)rt->rt_llinfo;
183 	switch (req) {
184 
185 	case RTM_ADD:
186 		/*
187 		 * XXX: If this is a manually added route to interface
188 		 * such as older version of routed or gated might provide,
189 		 * restore cloning bit.
190 		 */
191 		if ((rt->rt_flags & RTF_HOST) == 0 &&
192 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
193 			rt->rt_flags |= RTF_CLONING;
194 		if (rt->rt_flags & RTF_CLONING) {
195 			/*
196 			 * Case 1: This route should come from a route to iface.
197 			 */
198 			rt_setgate(rt, rt_key(rt),
199 					(struct sockaddr *)&null_sdl);
200 			gate = rt->rt_gateway;
201 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
202 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
203 			rt->rt_expire = time_second;
204 			break;
205 		}
206 		/* Announce a new entry if requested. */
207 		if (rt->rt_flags & RTF_ANNOUNCE)
208 			arprequest(rt->rt_ifp,
209 			    &SIN(rt_key(rt))->sin_addr,
210 			    &SIN(rt_key(rt))->sin_addr,
211 			    (u_char *)LLADDR(SDL(gate)));
212 		/*FALLTHROUGH*/
213 	case RTM_RESOLVE:
214 		if (gate->sa_family != AF_LINK ||
215 		    gate->sa_len < sizeof(null_sdl)) {
216 			log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__,
217 			    inet_ntoa(SIN(rt_key(rt))->sin_addr),
218 			    (gate->sa_family != AF_LINK) ?
219 			    " (!AF_LINK)": "");
220 			break;
221 		}
222 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
223 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
224 		if (la != 0)
225 			break; /* This happens on a route change */
226 		/*
227 		 * Case 2:  This route may come from cloning, or a manual route
228 		 * add with a LL address.
229 		 */
230 		R_Zalloc(la, struct llinfo_arp *, sizeof(*la));
231 		rt->rt_llinfo = (caddr_t)la;
232 		if (la == 0) {
233 			log(LOG_DEBUG, "%s: malloc failed\n", __func__);
234 			break;
235 		}
236 		arp_allocated++;
237 		la->la_rt = rt;
238 		rt->rt_flags |= RTF_LLINFO;
239 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
240 		LIST_INSERT_HEAD(&llinfo_arp, la, la_le);
241 
242 #ifdef INET
243 		/*
244 		 * This keeps the multicast addresses from showing up
245 		 * in `arp -a' listings as unresolved.  It's not actually
246 		 * functional.  Then the same for broadcast.
247 		 */
248 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) &&
249 		    rt->rt_ifp->if_type != IFT_ARCNET) {
250 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
251 					       LLADDR(SDL(gate)));
252 			SDL(gate)->sdl_alen = 6;
253 			rt->rt_expire = 0;
254 		}
255 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
256 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
257 			       rt->rt_ifp->if_addrlen);
258 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
259 			rt->rt_expire = 0;
260 		}
261 #endif
262 
263 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
264 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
265 		    /*
266 		     * This test used to be
267 		     *	if (loif.if_flags & IFF_UP)
268 		     * It allowed local traffic to be forced
269 		     * through the hardware by configuring the loopback down.
270 		     * However, it causes problems during network configuration
271 		     * for boards that can't receive packets they send.
272 		     * It is now necessary to clear "useloopback" and remove
273 		     * the route to force traffic out to the hardware.
274 		     */
275 			rt->rt_expire = 0;
276 			Bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
277 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
278 			if (useloopback)
279 				rt->rt_ifp = loif;
280 
281 		}
282 		break;
283 
284 	case RTM_DELETE:
285 		if (la == 0)
286 			break;
287 		RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]);
288 		LIST_REMOVE(la, la_le);
289 		rt->rt_llinfo = 0;
290 		rt->rt_flags &= ~RTF_LLINFO;
291 		if (la->la_hold)
292 			m_freem(la->la_hold);
293 		Free((caddr_t)la);
294 	}
295 }
296 
297 /*
298  * Broadcast an ARP request. Caller specifies:
299  *	- arp header source ip address
300  *	- arp header target ip address
301  *	- arp header source ethernet address
302  */
303 static void
304 arprequest(ifp, sip, tip, enaddr)
305 	register struct ifnet *ifp;
306 	register struct in_addr *sip, *tip;
307 	register u_char *enaddr;
308 {
309 	register struct mbuf *m;
310 	register struct ether_header *eh;
311 	register struct arc_header *arh;
312 	register struct arphdr *ah;
313 	struct sockaddr sa;
314 	static u_char	llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
315 				   LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
316 	u_short ar_hrd;
317 
318 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
319 		return;
320 	m->m_pkthdr.rcvif = (struct ifnet *)0;
321 #ifdef MAC
322 	mac_create_mbuf_linklayer(ifp, m);
323 #endif
324 	switch (ifp->if_type) {
325 	case IFT_ARCNET:
326 		ar_hrd = htons(ARPHRD_ARCNET);
327 
328 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
329 		m->m_pkthdr.len = m->m_len;
330 		MH_ALIGN(m, m->m_len);
331 
332 		arh = (struct arc_header *)sa.sa_data;
333 		arh->arc_dhost = *ifp->if_broadcastaddr;
334 		arh->arc_type = ARCTYPE_ARP;
335 
336 		ah = mtod(m, struct arphdr *);
337 		break;
338 
339 	case IFT_ISO88025:
340 		ar_hrd = htons(ARPHRD_IEEE802);
341 
342 		m->m_len = sizeof(llcx) +
343 		    arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
344 		m->m_pkthdr.len = m->m_len;
345 		MH_ALIGN(m, m->m_len);
346 
347 		(void)memcpy(mtod(m, caddr_t), llcx, sizeof(llcx));
348 		(void)memcpy(sa.sa_data, ifp->if_broadcastaddr, 6);
349 		(void)memcpy(sa.sa_data + 6, enaddr, 6);
350 		sa.sa_data[6] |= TR_RII;
351 		sa.sa_data[12] = TR_AC;
352 		sa.sa_data[13] = TR_LLC_FRAME;
353 
354 		ah = (struct arphdr *)(mtod(m, char *) + sizeof(llcx));
355 		break;
356 	case IFT_FDDI:
357 	case IFT_ETHER:
358 		/*
359 		 * This may not be correct for types not explicitly
360 		 * listed, but this is our best guess
361 		 */
362 	default:
363 		ar_hrd = htons(ARPHRD_ETHER);
364 
365 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
366 		m->m_pkthdr.len = m->m_len;
367 		MH_ALIGN(m, m->m_len);
368 
369 		eh = (struct ether_header *)sa.sa_data;
370 		/* if_output will not swap */
371 		eh->ether_type = htons(ETHERTYPE_ARP);
372 		(void)memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
373 		    sizeof(eh->ether_dhost));
374 
375 		ah = mtod(m, struct arphdr *);
376 		break;
377 	}
378 
379 	ah->ar_hrd = ar_hrd;
380 	ah->ar_pro = htons(ETHERTYPE_IP);
381 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
382 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
383 	ah->ar_op = htons(ARPOP_REQUEST);
384 	(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
385 	memset(ar_tha(ah), 0, ah->ar_hln);
386 	(void)memcpy(ar_spa(ah), sip, ah->ar_pln);
387 	(void)memcpy(ar_tpa(ah), tip, ah->ar_pln);
388 
389 	sa.sa_family = AF_UNSPEC;
390 	sa.sa_len = sizeof(sa);
391 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
392 }
393 
394 /*
395  * Resolve an IP address into an ethernet address.  If success,
396  * desten is filled in.  If there is no entry in arptab,
397  * set one up and broadcast a request for the IP address.
398  * Hold onto this mbuf and resend it once the address
399  * is finally resolved.  A return value of 1 indicates
400  * that desten has been filled in and the packet should be sent
401  * normally; a 0 return indicates that the packet has been
402  * taken over here, either now or for later transmission.
403  */
404 int
405 arpresolve(ifp, rt, m, dst, desten, rt0)
406 	register struct ifnet *ifp;
407 	register struct rtentry *rt;
408 	struct mbuf *m;
409 	register struct sockaddr *dst;
410 	register u_char *desten;
411 	struct rtentry *rt0;
412 {
413 	struct llinfo_arp *la = 0;
414 	struct sockaddr_dl *sdl;
415 
416 	if (m->m_flags & M_BCAST) {	/* broadcast */
417 		(void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
418 		return (1);
419 	}
420 	if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */
421 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
422 		return(1);
423 	}
424 	if (rt)
425 		la = (struct llinfo_arp *)rt->rt_llinfo;
426 	if (la == 0) {
427 		la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0);
428 		if (la)
429 			rt = la->la_rt;
430 	}
431 	if (la == 0 || rt == 0) {
432 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
433 			inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "",
434 				rt ? "rt" : "");
435 		m_freem(m);
436 		return (0);
437 	}
438 	sdl = SDL(rt->rt_gateway);
439 	/*
440 	 * Check the address family and length is valid, the address
441 	 * is resolved; otherwise, try to resolve.
442 	 */
443 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
444 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
445 		/*
446 		 * If entry has an expiry time and it is approaching,
447 		 * see if we need to send an ARP request within this
448 		 * arpt_down interval.
449 		 */
450 		if ((rt->rt_expire != 0) &&
451 		    (time_second + la->la_preempt > rt->rt_expire)) {
452 			arprequest(ifp,
453 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
454 				   &SIN(dst)->sin_addr,
455 				   IF_LLADDR(ifp));
456 			la->la_preempt--;
457 		}
458 
459 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
460 		return 1;
461 	}
462 	/*
463 	 * If ARP is disabled or static on this interface, stop.
464 	 * XXX
465 	 * Probably should not allocate empty llinfo struct if we are
466 	 * not going to be sending out an arp request.
467 	 */
468 	if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
469 		m_freem(m);
470 		return (0);
471 	}
472 	/*
473 	 * There is an arptab entry, but no ethernet address
474 	 * response yet.  Replace the held mbuf with this
475 	 * latest one.
476 	 */
477 	if (la->la_hold)
478 		m_freem(la->la_hold);
479 	la->la_hold = m;
480 	if (rt->rt_expire) {
481 		RT_LOCK(rt);
482 		rt->rt_flags &= ~RTF_REJECT;
483 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
484 			rt->rt_expire = time_second;
485 			if (la->la_asked++ < arp_maxtries) {
486 				arprequest(ifp,
487 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
488 					   &SIN(dst)->sin_addr,
489 					   IF_LLADDR(ifp));
490 			} else {
491 				rt->rt_flags |= RTF_REJECT;
492 				rt->rt_expire += arpt_down;
493 				la->la_asked = 0;
494 				la->la_preempt = arp_maxtries;
495 			}
496 
497 		}
498 		RT_UNLOCK(rt);
499 	}
500 	return (0);
501 }
502 
503 /*
504  * Common length and type checks are done here,
505  * then the protocol-specific routine is called.
506  */
507 static void
508 arpintr(struct mbuf *m)
509 {
510 	struct arphdr *ar;
511 
512 	if (!arpinit_done) {
513 		/* NB: this race should not matter */
514 		arpinit_done = 1;
515 		callout_reset(&arp_callout, hz, arptimer, NULL);
516 	}
517 	if (m->m_len < sizeof(struct arphdr) &&
518 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
519 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
520 		return;
521 	}
522 	ar = mtod(m, struct arphdr *);
523 
524 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
525 	    ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
526 	    ntohs(ar->ar_hrd) != ARPHRD_ARCNET) {
527 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
528 		    (unsigned char *)&ar->ar_hrd, "");
529 		m_freem(m);
530 		return;
531 	}
532 
533 	if (m->m_pkthdr.len < arphdr_len(ar) &&
534 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
535 		log(LOG_ERR, "arp: runt packet\n");
536 		m_freem(m);
537 		return;
538 	}
539 
540 	switch (ntohs(ar->ar_pro)) {
541 #ifdef INET
542 	case ETHERTYPE_IP:
543 		in_arpinput(m);
544 		return;
545 #endif
546 	}
547 	m_freem(m);
548 }
549 
550 #ifdef INET
551 /*
552  * ARP for Internet protocols on 10 Mb/s Ethernet.
553  * Algorithm is that given in RFC 826.
554  * In addition, a sanity check is performed on the sender
555  * protocol address, to catch impersonators.
556  * We no longer handle negotiations for use of trailer protocol:
557  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
558  * along with IP replies if we wanted trailers sent to us,
559  * and also sent them in response to IP replies.
560  * This allowed either end to announce the desire to receive
561  * trailer packets.
562  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
563  * but formerly didn't normally send requests.
564  */
565 static int log_arp_wrong_iface = 1;
566 static int log_arp_movements = 1;
567 
568 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
569 	&log_arp_wrong_iface, 0,
570 	"log arp packets arriving on the wrong interface");
571 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
572         &log_arp_movements, 0,
573         "log arp replies from MACs different than the one in the cache");
574 
575 
576 static void
577 in_arpinput(m)
578 	struct mbuf *m;
579 {
580 	register struct arphdr *ah;
581 	register struct ifnet *ifp = m->m_pkthdr.rcvif;
582 	struct ether_header *eh;
583 	struct arc_header *arh;
584 	struct iso88025_header *th = (struct iso88025_header *)0;
585 	struct iso88025_sockaddr_dl_data *trld;
586 	register struct llinfo_arp *la = 0;
587 	register struct rtentry *rt;
588 	struct ifaddr *ifa;
589 	struct in_ifaddr *ia;
590 	struct sockaddr_dl *sdl;
591 	struct sockaddr sa;
592 	struct in_addr isaddr, itaddr, myaddr;
593 	int op, rif_len;
594 	int req_len;
595 
596 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
597 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
598 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
599 		return;
600 	}
601 
602 	ah = mtod(m, struct arphdr *);
603 	op = ntohs(ah->ar_op);
604 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
605 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
606 #ifdef BRIDGE
607 #define BRIDGE_TEST (do_bridge)
608 #else
609 #define BRIDGE_TEST (0) /* cc will optimise the test away */
610 #endif
611 	/*
612 	 * For a bridge, we want to check the address irrespective
613 	 * of the receive interface. (This will change slightly
614 	 * when we have clusters of interfaces).
615 	 */
616 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash)
617 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
618 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
619 			goto match;
620 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
621 		if ((BRIDGE_TEST || (ia->ia_ifp == ifp)) &&
622 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
623 			goto match;
624 	/*
625 	 * No match, use the first inet address on the receive interface
626 	 * as a dummy address for the rest of the function.
627 	 */
628 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
629 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
630 			ia = ifatoia(ifa);
631 			goto match;
632 		}
633 	/*
634 	 * If bridging, fall back to using any inet address.
635 	 */
636 	if (!BRIDGE_TEST ||
637 	    (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) {
638 		m_freem(m);
639 		return;
640 	}
641 match:
642 	myaddr = ia->ia_addr.sin_addr;
643 	if (!bcmp(ar_sha(ah), IF_LLADDR(ifp), ifp->if_addrlen)) {
644 		m_freem(m);	/* it's from me, ignore it. */
645 		return;
646 	}
647 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
648 		log(LOG_ERR,
649 		    "arp: link address is broadcast for IP address %s!\n",
650 		    inet_ntoa(isaddr));
651 		m_freem(m);
652 		return;
653 	}
654 	if (isaddr.s_addr == myaddr.s_addr) {
655 		log(LOG_ERR,
656 		   "arp: %*D is using my IP address %s!\n",
657 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
658 		   inet_ntoa(isaddr));
659 		itaddr = myaddr;
660 		goto reply;
661 	}
662 	if (ifp->if_flags & IFF_STATICARP)
663 		goto reply;
664 	la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0);
665 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
666 		/* the following is not an error when doing bridging */
667 		if (!BRIDGE_TEST && rt->rt_ifp != ifp) {
668 			if (log_arp_wrong_iface)
669 				log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n",
670 				    inet_ntoa(isaddr),
671 				    rt->rt_ifp->if_xname,
672 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
673 				    ifp->if_xname);
674 			goto reply;
675 		}
676 		if (sdl->sdl_alen &&
677 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
678 			if (rt->rt_expire) {
679 			    if (log_arp_movements)
680 			        log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n",
681 				    inet_ntoa(isaddr),
682 				    ifp->if_addrlen, (u_char *)LLADDR(sdl), ":",
683 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
684 				    ifp->if_xname);
685 			} else {
686 			    log(LOG_ERR,
687 				"arp: %*D attempts to modify permanent entry for %s on %s\n",
688 				ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
689 				inet_ntoa(isaddr), ifp->if_xname);
690 			    goto reply;
691 			}
692 		}
693 		/*
694 		 * sanity check for the address length.
695 		 * XXX this does not work for protocols with variable address
696 		 * length. -is
697 		 */
698 		if (sdl->sdl_alen &&
699 		    sdl->sdl_alen != ah->ar_hln) {
700 			log(LOG_WARNING,
701 			    "arp from %*D: new addr len %d, was %d",
702 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
703 			    ah->ar_hln, sdl->sdl_alen);
704 		}
705 		if (ifp->if_addrlen != ah->ar_hln) {
706 			log(LOG_WARNING,
707 			    "arp from %*D: addr len: new %d, i/f %d (ignored)",
708 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
709 			    ah->ar_hln, ifp->if_addrlen);
710 			goto reply;
711 		}
712 		(void)memcpy(LLADDR(sdl), ar_sha(ah),
713 		    sdl->sdl_alen = ah->ar_hln);
714 		/*
715 		 * If we receive an arp from a token-ring station over
716 		 * a token-ring nic then try to save the source
717 		 * routing info.
718 		 */
719 		if (ifp->if_type == IFT_ISO88025) {
720 			th = (struct iso88025_header *)m->m_pkthdr.header;
721 			trld = SDL_ISO88025(sdl);
722 			rif_len = TR_RCF_RIFLEN(th->rcf);
723 			if ((th->iso88025_shost[0] & TR_RII) &&
724 			    (rif_len > 2)) {
725 				trld->trld_rcf = th->rcf;
726 				trld->trld_rcf ^= htons(TR_RCF_DIR);
727 				memcpy(trld->trld_route, th->rd, rif_len - 2);
728 				trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK);
729 				/*
730 				 * Set up source routing information for
731 				 * reply packet (XXX)
732 				 */
733 				m->m_data -= rif_len;
734 				m->m_len  += rif_len;
735 				m->m_pkthdr.len += rif_len;
736 			} else {
737 				th->iso88025_shost[0] &= ~TR_RII;
738 				trld->trld_rcf = 0;
739 			}
740 			m->m_data -= 8;
741 			m->m_len  += 8;
742 			m->m_pkthdr.len += 8;
743 			th->rcf = trld->trld_rcf;
744 		}
745 		RT_LOCK(rt);
746 		if (rt->rt_expire)
747 			rt->rt_expire = time_second + arpt_keep;
748 		rt->rt_flags &= ~RTF_REJECT;
749 		RT_UNLOCK(rt);
750 		la->la_asked = 0;
751 		la->la_preempt = arp_maxtries;
752 		if (la->la_hold) {
753 			(*ifp->if_output)(ifp, la->la_hold,
754 				rt_key(rt), rt);
755 			la->la_hold = 0;
756 		}
757 	}
758 reply:
759 	if (op != ARPOP_REQUEST) {
760 		m_freem(m);
761 		return;
762 	}
763 	if (itaddr.s_addr == myaddr.s_addr) {
764 		/* I am the target */
765 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
766 		(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
767 	} else {
768 		la = arplookup(itaddr.s_addr, 0, SIN_PROXY);
769 		if (la == NULL) {
770 			struct sockaddr_in sin;
771 
772 			if (!arp_proxyall) {
773 				m_freem(m);
774 				return;
775 			}
776 
777 			bzero(&sin, sizeof sin);
778 			sin.sin_family = AF_INET;
779 			sin.sin_len = sizeof sin;
780 			sin.sin_addr = itaddr;
781 
782 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
783 			if (!rt) {
784 				m_freem(m);
785 				return;
786 			}
787 			/*
788 			 * Don't send proxies for nodes on the same interface
789 			 * as this one came out of, or we'll get into a fight
790 			 * over who claims what Ether address.
791 			 */
792 			if (rt->rt_ifp == ifp) {
793 				rtfree(rt);
794 				m_freem(m);
795 				return;
796 			}
797 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
798 			(void)memcpy(ar_sha(ah), IF_LLADDR(ifp), ah->ar_hln);
799 			rtfree(rt);
800 
801 			/*
802 			 * Also check that the node which sent the ARP packet
803 			 * is on the the interface we expect it to be on. This
804 			 * avoids ARP chaos if an interface is connected to the
805 			 * wrong network.
806 			 */
807 			sin.sin_addr = isaddr;
808 
809 			rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
810 			if (!rt) {
811 				m_freem(m);
812 				return;
813 			}
814 			if (rt->rt_ifp != ifp) {
815 				log(LOG_INFO, "arp_proxy: ignoring request"
816 				    " from %s via %s, expecting %s\n",
817 				    inet_ntoa(isaddr), ifp->if_xname,
818 				    rt->rt_ifp->if_xname);
819 				rtfree(rt);
820 				m_freem(m);
821 				return;
822 			}
823 			rtfree(rt);
824 
825 #ifdef DEBUG_PROXY
826 			printf("arp: proxying for %s\n",
827 			       inet_ntoa(itaddr));
828 #endif
829 		} else {
830 			rt = la->la_rt;
831 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
832 			sdl = SDL(rt->rt_gateway);
833 			(void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
834 		}
835 	}
836 
837 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
838 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
839 	ah->ar_op = htons(ARPOP_REPLY);
840 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
841 	switch (ifp->if_type) {
842 	case IFT_ARCNET:
843 		arh = (struct arc_header *)sa.sa_data;
844 		arh->arc_dhost = *ar_tha(ah);
845 		arh->arc_type = ARCTYPE_ARP;
846 		break;
847 
848 	case IFT_ISO88025:
849 		/* Re-arrange the source/dest address */
850 		memcpy(th->iso88025_dhost, th->iso88025_shost,
851 		    sizeof(th->iso88025_dhost));
852 		memcpy(th->iso88025_shost, IF_LLADDR(ifp),
853 		    sizeof(th->iso88025_shost));
854 		/* Set the source routing bit if neccesary */
855 		if (th->iso88025_dhost[0] & TR_RII) {
856 			th->iso88025_dhost[0] &= ~TR_RII;
857 			if (TR_RCF_RIFLEN(th->rcf) > 2)
858 				th->iso88025_shost[0] |= TR_RII;
859 		}
860 		/* Copy the addresses, ac and fc into sa_data */
861 		memcpy(sa.sa_data, th->iso88025_dhost,
862 		    sizeof(th->iso88025_dhost) * 2);
863 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2)] = TR_AC;
864 		sa.sa_data[(sizeof(th->iso88025_dhost) * 2) + 1] = TR_LLC_FRAME;
865 		break;
866 	case IFT_ETHER:
867 	case IFT_FDDI:
868 	/*
869 	 * May not be correct for types not explictly
870 	 * listed, but it is our best guess.
871 	 */
872 	default:
873 		eh = (struct ether_header *)sa.sa_data;
874 		(void)memcpy(eh->ether_dhost, ar_tha(ah),
875 		    sizeof(eh->ether_dhost));
876 		eh->ether_type = htons(ETHERTYPE_ARP);
877 		break;
878 	}
879 	sa.sa_family = AF_UNSPEC;
880 	sa.sa_len = sizeof(sa);
881 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
882 	return;
883 }
884 #endif
885 
886 /*
887  * Free an arp entry.
888  */
889 static void
890 arptfree(la)
891 	register struct llinfo_arp *la;
892 {
893 	register struct rtentry *rt = la->la_rt;
894 	register struct sockaddr_dl *sdl;
895 
896 	if (rt == 0)
897 		panic("arptfree");
898 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
899 	    sdl->sdl_family == AF_LINK) {
900 		sdl->sdl_alen = 0;
901 		la->la_preempt = la->la_asked = 0;
902 		RT_LOCK(rt);		/* XXX needed or move higher? */
903 		rt->rt_flags &= ~RTF_REJECT;
904 		RT_UNLOCK(rt);
905 		return;
906 	}
907 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
908 			0, (struct rtentry **)0);
909 }
910 /*
911  * Lookup or enter a new address in arptab.
912  */
913 static struct llinfo_arp *
914 arplookup(addr, create, proxy)
915 	u_long addr;
916 	int create, proxy;
917 {
918 	register struct rtentry *rt;
919 	struct sockaddr_inarp sin;
920 	const char *why = 0;
921 
922 	bzero(&sin, sizeof(sin));
923 	sin.sin_len = sizeof(sin);
924 	sin.sin_family = AF_INET;
925 	sin.sin_addr.s_addr = addr;
926 	if (proxy)
927 		sin.sin_other = SIN_PROXY;
928 	rt = rtalloc1((struct sockaddr *)&sin, create, 0UL);
929 	if (rt == 0)
930 		return (0);
931 
932 	if (rt->rt_flags & RTF_GATEWAY)
933 		why = "host is not on local network";
934 	else if ((rt->rt_flags & RTF_LLINFO) == 0)
935 		why = "could not allocate llinfo";
936 	else if (rt->rt_gateway->sa_family != AF_LINK)
937 		why = "gateway route is not ours";
938 
939 	if (why) {
940 #define	ISDYNCLONE(_rt) \
941 	(((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED)
942 		if (create)
943 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
944 			    inet_ntoa(sin.sin_addr), why);
945 		/*
946 		 * If there are no references to this Layer 2 route,
947 		 * and it is a cloned route, and not static, and
948 		 * arplookup() is creating the route, then purge
949 		 * it from the routing table as it is probably bogus.
950 		 */
951 		if (rt->rt_refcnt == 1 && ISDYNCLONE(rt))
952 			rtexpunge(rt);
953 		RTFREE_LOCKED(rt);
954 		return (0);
955 #undef ISDYNCLONE
956 	} else {
957 		rt->rt_refcnt--;
958 		RT_UNLOCK(rt);
959 		return ((struct llinfo_arp *)rt->rt_llinfo);
960 	}
961 }
962 
963 void
964 arp_ifinit(ifp, ifa)
965 	struct ifnet *ifp;
966 	struct ifaddr *ifa;
967 {
968 	if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
969 		arprequest(ifp, &IA_SIN(ifa)->sin_addr,
970 				&IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
971 	ifa->ifa_rtrequest = arp_rtrequest;
972 	ifa->ifa_flags |= RTF_CLONING;
973 }
974 
975 static void
976 arp_init(void)
977 {
978 
979 	arpintrq.ifq_maxlen = 50;
980 	mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF);
981 	LIST_INIT(&llinfo_arp);
982 	callout_init(&arp_callout, CALLOUT_MPSAFE);
983 	netisr_register(NETISR_ARP, arpintr, &arpintrq);
984 }
985 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
986