xref: /netbsd-src/sys/netinet/if_arp.c (revision cac8e449158efc7261bebc8657cbb0125a2cfdde)
1 /*	$NetBSD: if_arp.c,v 1.140 2008/05/13 18:24:01 dyoung Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998, 2000, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Public Access Networks Corporation ("Panix").  It was developed under
9  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
62  */
63 
64 /*
65  * Ethernet address resolution protocol.
66  * TODO:
67  *	add "inuse/lock" bit (or ref. count) along with valid bit
68  */
69 
70 #include <sys/cdefs.h>
71 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.140 2008/05/13 18:24:01 dyoung Exp $");
72 
73 #include "opt_ddb.h"
74 #include "opt_inet.h"
75 
76 #ifdef INET
77 
78 #include "bridge.h"
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/callout.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/socket.h>
86 #include <sys/time.h>
87 #include <sys/timetc.h>
88 #include <sys/kernel.h>
89 #include <sys/errno.h>
90 #include <sys/ioctl.h>
91 #include <sys/syslog.h>
92 #include <sys/proc.h>
93 #include <sys/protosw.h>
94 #include <sys/domain.h>
95 #include <sys/sysctl.h>
96 #include <sys/socketvar.h>
97 #include <sys/percpu.h>
98 
99 #include <net/ethertypes.h>
100 #include <net/if.h>
101 #include <net/if_dl.h>
102 #include <net/if_token.h>
103 #include <net/if_types.h>
104 #include <net/if_ether.h>
105 #include <net/route.h>
106 #include <net/net_stats.h>
107 
108 #include <netinet/in.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/if_inarp.h>
113 
114 #include "arcnet.h"
115 #if NARCNET > 0
116 #include <net/if_arc.h>
117 #endif
118 #include "fddi.h"
119 #if NFDDI > 0
120 #include <net/if_fddi.h>
121 #endif
122 #include "token.h"
123 #include "carp.h"
124 #if NCARP > 0
125 #include <netinet/ip_carp.h>
126 #endif
127 
128 #define SIN(s) ((struct sockaddr_in *)s)
129 #define SRP(s) ((struct sockaddr_inarp *)s)
130 
131 /*
132  * ARP trailer negotiation.  Trailer protocol is not IP specific,
133  * but ARP request/response use IP addresses.
134  */
135 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
136 
137 /* timer values */
138 int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
139 int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
140 int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
141 int	arpt_refresh = (5*60);	/* time left before refreshing */
142 #define	rt_expire rt_rmx.rmx_expire
143 #define	rt_pksent rt_rmx.rmx_pksent
144 
145 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
146 	    const struct sockaddr *);
147 static	void arptfree(struct llinfo_arp *);
148 static	void arptimer(void *);
149 static	struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *,
150 				      int, int, struct rtentry *);
151 static	struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *,
152 					  int, int);
153 static	void in_arpinput(struct mbuf *);
154 
155 LIST_HEAD(, llinfo_arp) llinfo_arp;
156 struct	ifqueue arpintrq = {
157 	.ifq_head = NULL,
158 	.ifq_tail = NULL,
159 	.ifq_len = 0,
160 	.ifq_maxlen = 50,
161 	.ifq_drops = 0,
162 };
163 int	arp_inuse, arp_allocated, arp_intimer;
164 int	arp_maxtries = 5;
165 int	useloopback = 1;	/* use loopback interface for local traffic */
166 int	arpinit_done = 0;
167 
168 static percpu_t *arpstat_percpu;
169 
170 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
171 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
172 
173 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
174 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
175 
176 struct	callout arptimer_ch;
177 
178 /* revarp state */
179 struct	in_addr myip, srv_ip;
180 int	myip_initialized = 0;
181 int	revarp_in_progress = 0;
182 struct	ifnet *myip_ifp = NULL;
183 
184 #ifdef DDB
185 static void db_print_sa(const struct sockaddr *);
186 static void db_print_ifa(struct ifaddr *);
187 static void db_print_llinfo(void *);
188 static int db_show_rtentry(struct rtentry *, void *);
189 #endif
190 
191 /*
192  * this should be elsewhere.
193  */
194 
195 static char *
196 lla_snprintf(u_int8_t *, int);
197 
198 static char *
199 lla_snprintf(u_int8_t *adrp, int len)
200 {
201 #define NUMBUFS 3
202 	static char buf[NUMBUFS][16*3];
203 	static int bnum = 0;
204 
205 	int i;
206 	char *p;
207 
208 	p = buf[bnum];
209 
210 	*p++ = hexdigits[(*adrp)>>4];
211 	*p++ = hexdigits[(*adrp++)&0xf];
212 
213 	for (i=1; i<len && i<16; i++) {
214 		*p++ = ':';
215 		*p++ = hexdigits[(*adrp)>>4];
216 		*p++ = hexdigits[(*adrp++)&0xf];
217 	}
218 
219 	*p = 0;
220 	p = buf[bnum];
221 	bnum = (bnum + 1) % NUMBUFS;
222 	return p;
223 }
224 
225 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
226 
227 const struct protosw arpsw[] = {
228 	{ .pr_type = 0,
229 	  .pr_domain = &arpdomain,
230 	  .pr_protocol = 0,
231 	  .pr_flags = 0,
232 	  .pr_input = 0,
233 	  .pr_output = 0,
234 	  .pr_ctlinput = 0,
235 	  .pr_ctloutput = 0,
236 	  .pr_usrreq =  0,
237 	  .pr_init = arp_init,
238 	  .pr_fasttimo = 0,
239 	  .pr_slowtimo = 0,
240 	  .pr_drain = arp_drain,
241 	}
242 };
243 
244 
245 struct domain arpdomain = {
246 	.dom_family = PF_ARP,
247 	.dom_name = "arp",
248 	.dom_protosw = arpsw,
249 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
250 };
251 
252 /*
253  * ARP table locking.
254  *
255  * to prevent lossage vs. the arp_drain routine (which may be called at
256  * any time, including in a device driver context), we do two things:
257  *
258  * 1) manipulation of la->la_hold is done at splnet() (for all of
259  * about two instructions).
260  *
261  * 2) manipulation of the arp table's linked list is done under the
262  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
263  * while the arp table is locked, we punt and try again later.
264  */
265 
266 static int	arp_locked;
267 static inline int arp_lock_try(int);
268 static inline void arp_unlock(void);
269 
270 static inline int
271 arp_lock_try(int recurse)
272 {
273 	int s;
274 
275 	/*
276 	 * Use splvm() -- we're blocking things that would cause
277 	 * mbuf allocation.
278 	 */
279 	s = splvm();
280 	if (!recurse && arp_locked) {
281 		splx(s);
282 		return 0;
283 	}
284 	arp_locked++;
285 	splx(s);
286 	return 1;
287 }
288 
289 static inline void
290 arp_unlock(void)
291 {
292 	int s;
293 
294 	s = splvm();
295 	arp_locked--;
296 	splx(s);
297 }
298 
299 #ifdef DIAGNOSTIC
300 #define	ARP_LOCK(recurse)						\
301 do {									\
302 	if (arp_lock_try(recurse) == 0) {				\
303 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
304 		panic("arp_lock");					\
305 	}								\
306 } while (/*CONSTCOND*/ 0)
307 #define	ARP_LOCK_CHECK()						\
308 do {									\
309 	if (arp_locked == 0) {						\
310 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
311 		panic("arp lock check");				\
312 	}								\
313 } while (/*CONSTCOND*/ 0)
314 #else
315 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
316 #define	ARP_LOCK_CHECK()	/* nothing */
317 #endif
318 
319 #define	ARP_UNLOCK()		arp_unlock()
320 
321 void
322 arp_init(void)
323 {
324 
325 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
326 }
327 
328 /*
329  * ARP protocol drain routine.  Called when memory is in short supply.
330  * Called at splvm();  don't acquire softnet_lock as can be called from
331  * hardware interrupt handlers.
332  */
333 void
334 arp_drain(void)
335 {
336 	struct llinfo_arp *la, *nla;
337 	int count = 0;
338 	struct mbuf *mold;
339 
340 	KERNEL_LOCK(1, NULL);
341 
342 	if (arp_lock_try(0) == 0) {
343 		printf("arp_drain: locked; punting\n");
344 		KERNEL_UNLOCK_ONE(NULL);
345 		return;
346 	}
347 
348 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
349 		nla = LIST_NEXT(la, la_list);
350 
351 		mold = la->la_hold;
352 		la->la_hold = 0;
353 
354 		if (mold) {
355 			m_freem(mold);
356 			count++;
357 		}
358 	}
359 	ARP_UNLOCK();
360 	ARP_STATADD(ARP_STAT_DFRDROPPED, count);
361 	KERNEL_UNLOCK_ONE(NULL);
362 }
363 
364 
365 /*
366  * Timeout routine.  Age arp_tab entries periodically.
367  */
368 /* ARGSUSED */
369 static void
370 arptimer(void *arg)
371 {
372 	struct llinfo_arp *la, *nla;
373 
374 	mutex_enter(softnet_lock);
375 	KERNEL_LOCK(1, NULL);
376 
377 	if (arp_lock_try(0) == 0) {
378 		/* get it later.. */
379 		KERNEL_UNLOCK_ONE(NULL);
380 		mutex_exit(softnet_lock);
381 		return;
382 	}
383 
384 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
385 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
386 		struct rtentry *rt = la->la_rt;
387 
388 		nla = LIST_NEXT(la, la_list);
389 		if (rt->rt_expire == 0)
390 			continue;
391 		if ((rt->rt_expire - time_second) < arpt_refresh &&
392 		    rt->rt_pksent > (time_second - arpt_keep)) {
393 			/*
394 			 * If the entry has been used during since last
395 			 * refresh, try to renew it before deleting.
396 			 */
397 			arprequest(rt->rt_ifp,
398 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
399 			    &satocsin(rt_getkey(rt))->sin_addr,
400 			    CLLADDR(rt->rt_ifp->if_sadl));
401 		} else if (rt->rt_expire <= time_second)
402 			arptfree(la); /* timer has expired; clear */
403 	}
404 
405 	ARP_UNLOCK();
406 
407 	KERNEL_UNLOCK_ONE(NULL);
408 	mutex_exit(softnet_lock);
409 }
410 
411 /*
412  * We set the gateway for RTF_CLONING routes to a "prototype"
413  * link-layer sockaddr whose interface type (if_type) and interface
414  * index (if_index) fields are prepared.
415  */
416 static struct sockaddr *
417 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
418     const struct sockaddr *netmask)
419 {
420 	const struct ifnet *ifp = rt->rt_ifp;
421 	uint8_t namelen = strlen(ifp->if_xname);
422 	uint8_t addrlen = ifp->if_addrlen;
423 
424 	/*
425 	 * XXX: If this is a manually added route to interface
426 	 * such as older version of routed or gated might provide,
427 	 * restore cloning bit.
428 	 */
429 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
430 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
431 		rt->rt_flags |= RTF_CLONING;
432 	if (rt->rt_flags & RTF_CLONING) {
433 		union {
434 			struct sockaddr sa;
435 			struct sockaddr_storage ss;
436 			struct sockaddr_dl sdl;
437 		} u;
438 		/*
439 		 * Case 1: This route should come from a route to iface.
440 		 */
441 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
442 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
443 		rt_setgate(rt, &u.sa);
444 		gate = rt->rt_gateway;
445 	}
446 	return gate;
447 }
448 
449 /*
450  * Parallel to llc_rtrequest.
451  */
452 void
453 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
454 {
455 	struct sockaddr *gate = rt->rt_gateway;
456 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
457 	size_t allocsize;
458 	struct mbuf *mold;
459 	int s;
460 	struct in_ifaddr *ia;
461 	struct ifaddr *ifa;
462 	struct ifnet *ifp = rt->rt_ifp;
463 	uint8_t namelen = strlen(ifp->if_xname);
464 	uint8_t addrlen = ifp->if_addrlen;
465 
466 	if (!arpinit_done) {
467 		arpinit_done = 1;
468 		/*
469 		 * We generate expiration times from time_second
470 		 * so avoid accidentally creating permanent routes.
471 		 */
472 		if (time_second == 0) {
473 			struct timespec ts;
474 			ts.tv_sec = 1;
475 			ts.tv_nsec = 0;
476 			tc_setclock(&ts);
477 		}
478 		callout_init(&arptimer_ch, CALLOUT_MPSAFE);
479 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
480 	}
481 
482 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
483 		if (req != RTM_ADD)
484 			return;
485 
486 		/*
487 		 * linklayers with particular link MTU limitation.
488 		 */
489 		switch(ifp->if_type) {
490 #if NFDDI > 0
491 		case IFT_FDDI:
492 			if (ifp->if_mtu > FDDIIPMTU)
493 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
494 			break;
495 #endif
496 #if NARC > 0
497 		case IFT_ARCNET:
498 		    {
499 			int arcipifmtu;
500 
501 			if (ifp->if_flags & IFF_LINK0)
502 				arcipifmtu = arc_ipmtu;
503 			else
504 				arcipifmtu = ARCMTU;
505 			if (ifp->if_mtu > arcipifmtu)
506 				rt->rt_rmx.rmx_mtu = arcipifmtu;
507 			break;
508 		    }
509 #endif
510 		}
511 		return;
512 	}
513 
514 	ARP_LOCK(1);		/* we may already be locked here. */
515 
516 	switch (req) {
517 
518 	case RTM_SETGATE:
519 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
520 		break;
521 	case RTM_ADD:
522 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
523 		if (rt->rt_flags & RTF_CLONING) {
524 			/*
525 			 * Give this route an expiration time, even though
526 			 * it's a "permanent" route, so that routes cloned
527 			 * from it do not need their expiration time set.
528 			 */
529 			rt->rt_expire = time_second;
530 			/*
531 			 * linklayers with particular link MTU limitation.
532 			 */
533 			switch (ifp->if_type) {
534 #if NFDDI > 0
535 			case IFT_FDDI:
536 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
537 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
538 				     (rt->rt_rmx.rmx_mtu == 0 &&
539 				      ifp->if_mtu > FDDIIPMTU)))
540 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
541 				break;
542 #endif
543 #if NARC > 0
544 			case IFT_ARCNET:
545 			    {
546 				int arcipifmtu;
547 				if (ifp->if_flags & IFF_LINK0)
548 					arcipifmtu = arc_ipmtu;
549 				else
550 					arcipifmtu = ARCMTU;
551 
552 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
553 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
554 				     (rt->rt_rmx.rmx_mtu == 0 &&
555 				      ifp->if_mtu > arcipifmtu)))
556 					rt->rt_rmx.rmx_mtu = arcipifmtu;
557 				break;
558 			    }
559 #endif
560 			}
561 			break;
562 		}
563 		/* Announce a new entry if requested. */
564 		if (rt->rt_flags & RTF_ANNOUNCE)
565 			arprequest(ifp,
566 			    &satocsin(rt_getkey(rt))->sin_addr,
567 			    &satocsin(rt_getkey(rt))->sin_addr,
568 			    CLLADDR(satocsdl(gate)));
569 		/*FALLTHROUGH*/
570 	case RTM_RESOLVE:
571 		if (gate->sa_family != AF_LINK ||
572 		    gate->sa_len < sockaddr_dl_measure(namelen, addrlen)) {
573 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
574 			break;
575 		}
576 		satosdl(gate)->sdl_type = ifp->if_type;
577 		satosdl(gate)->sdl_index = ifp->if_index;
578 		if (la != NULL)
579 			break; /* This happens on a route change */
580 		/*
581 		 * Case 2:  This route may come from cloning, or a manual route
582 		 * add with a LL address.
583 		 */
584 		switch (satocsdl(gate)->sdl_type) {
585 #if NTOKEN > 0
586 		case IFT_ISO88025:
587 			allocsize = sizeof(*la) + sizeof(struct token_rif);
588 			break;
589 #endif /* NTOKEN > 0 */
590 		default:
591 			allocsize = sizeof(*la);
592 		}
593 		R_Malloc(la, struct llinfo_arp *, allocsize);
594 		rt->rt_llinfo = (void *)la;
595 		if (la == NULL) {
596 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
597 			break;
598 		}
599 		arp_inuse++, arp_allocated++;
600 		memset(la, 0, allocsize);
601 		la->la_rt = rt;
602 		rt->rt_flags |= RTF_LLINFO;
603 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
604 
605 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
606 		while (ia && ia->ia_ifp != ifp)
607 			NEXT_IA_WITH_SAME_ADDR(ia);
608 		if (ia) {
609 			/*
610 			 * This test used to be
611 			 *	if (lo0ifp->if_flags & IFF_UP)
612 			 * It allowed local traffic to be forced through
613 			 * the hardware by configuring the loopback down.
614 			 * However, it causes problems during network
615 			 * configuration for boards that can't receive
616 			 * packets they send.  It is now necessary to clear
617 			 * "useloopback" and remove the route to force
618 			 * traffic out to the hardware.
619 			 *
620 			 * In 4.4BSD, the above "if" statement checked
621 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
622 			 * to the current form so that we can provide a
623 			 * better support for multiple IPv4 addresses on a
624 			 * interface.
625 			 */
626 			rt->rt_expire = 0;
627 			(void)sockaddr_dl_setaddr(satosdl(gate), gate->sa_len,
628 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen);
629 			if (useloopback)
630 				ifp = rt->rt_ifp = lo0ifp;
631 			/*
632 			 * make sure to set rt->rt_ifa to the interface
633 			 * address we are using, otherwise we will have trouble
634 			 * with source address selection.
635 			 */
636 			ifa = &ia->ia_ifa;
637 			if (ifa != rt->rt_ifa)
638 				rt_replace_ifa(rt, ifa);
639 		}
640 		break;
641 
642 	case RTM_DELETE:
643 		if (la == NULL)
644 			break;
645 		arp_inuse--;
646 		LIST_REMOVE(la, la_list);
647 		rt->rt_llinfo = NULL;
648 		rt->rt_flags &= ~RTF_LLINFO;
649 
650 		s = splnet();
651 		mold = la->la_hold;
652 		la->la_hold = 0;
653 		splx(s);
654 
655 		if (mold)
656 			m_freem(mold);
657 
658 		Free((void *)la);
659 	}
660 	ARP_UNLOCK();
661 }
662 
663 /*
664  * Broadcast an ARP request. Caller specifies:
665  *	- arp header source ip address
666  *	- arp header target ip address
667  *	- arp header source ethernet address
668  */
669 void
670 arprequest(struct ifnet *ifp,
671     const struct in_addr *sip, const struct in_addr *tip,
672     const u_int8_t *enaddr)
673 {
674 	struct mbuf *m;
675 	struct arphdr *ah;
676 	struct sockaddr sa;
677 	uint64_t *arps;
678 
679 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
680 		return;
681 	MCLAIM(m, &arpdomain.dom_mowner);
682 	switch (ifp->if_type) {
683 	case IFT_IEEE1394:
684 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
685 		    ifp->if_addrlen;
686 		break;
687 	default:
688 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
689 		    2 * ifp->if_addrlen;
690 		break;
691 	}
692 	m->m_pkthdr.len = m->m_len;
693 	MH_ALIGN(m, m->m_len);
694 	ah = mtod(m, struct arphdr *);
695 	memset(ah, 0, m->m_len);
696 	switch (ifp->if_type) {
697 	case IFT_IEEE1394:	/* RFC2734 */
698 		/* fill it now for ar_tpa computation */
699 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
700 		break;
701 	default:
702 		/* ifp->if_output will fill ar_hrd */
703 		break;
704 	}
705 	ah->ar_pro = htons(ETHERTYPE_IP);
706 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
707 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
708 	ah->ar_op = htons(ARPOP_REQUEST);
709 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
710 	memcpy(ar_spa(ah), sip, ah->ar_pln);
711 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
712 	sa.sa_family = AF_ARP;
713 	sa.sa_len = 2;
714 	m->m_flags |= M_BCAST;
715 	arps = ARP_STAT_GETREF();
716 	arps[ARP_STAT_SNDTOTAL]++;
717 	arps[ARP_STAT_SENDREQUEST]++;
718 	ARP_STAT_PUTREF();
719 	(*ifp->if_output)(ifp, m, &sa, NULL);
720 }
721 
722 /*
723  * Resolve an IP address into an ethernet address.  If success,
724  * desten is filled in.  If there is no entry in arptab,
725  * set one up and broadcast a request for the IP address.
726  * Hold onto this mbuf and resend it once the address
727  * is finally resolved.  A return value of 1 indicates
728  * that desten has been filled in and the packet should be sent
729  * normally; a 0 return indicates that the packet has been
730  * taken over here, either now or for later transmission.
731  */
732 int
733 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
734     const struct sockaddr *dst, u_char *desten)
735 {
736 	struct llinfo_arp *la;
737 	const struct sockaddr_dl *sdl;
738 	struct mbuf *mold;
739 	int s;
740 
741 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
742 		rt = la->la_rt;
743 
744 	if (la == NULL || rt == NULL) {
745 		ARP_STATINC(ARP_STAT_ALLOCFAIL);
746 		log(LOG_DEBUG,
747 		    "arpresolve: can't allocate llinfo on %s for %s\n",
748 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
749 		m_freem(m);
750 		return 0;
751 	}
752 	sdl = satocsdl(rt->rt_gateway);
753 	/*
754 	 * Check the address family and length is valid, the address
755 	 * is resolved; otherwise, try to resolve.
756 	 */
757 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
758 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
759 		memcpy(desten, CLLADDR(sdl),
760 		    min(sdl->sdl_alen, ifp->if_addrlen));
761 		rt->rt_pksent = time_second; /* Time for last pkt sent */
762 		return 1;
763 	}
764 	/*
765 	 * There is an arptab entry, but no ethernet address
766 	 * response yet.  Replace the held mbuf with this
767 	 * latest one.
768 	 */
769 
770 	ARP_STATINC(ARP_STAT_DFRTOTAL);
771 	s = splnet();
772 	mold = la->la_hold;
773 	la->la_hold = m;
774 	splx(s);
775 
776 	if (mold) {
777 		ARP_STATINC(ARP_STAT_DFRDROPPED);
778 		m_freem(mold);
779 	}
780 
781 	/*
782 	 * Re-send the ARP request when appropriate.
783 	 */
784 #ifdef	DIAGNOSTIC
785 	if (rt->rt_expire == 0) {
786 		/* This should never happen. (Should it? -gwr) */
787 		printf("arpresolve: unresolved and rt_expire == 0\n");
788 		/* Set expiration time to now (expired). */
789 		rt->rt_expire = time_second;
790 	}
791 #endif
792 	if (rt->rt_expire) {
793 		rt->rt_flags &= ~RTF_REJECT;
794 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
795 			rt->rt_expire = time_second;
796 			if (la->la_asked++ < arp_maxtries)
797 				arprequest(ifp,
798 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
799 				    &satocsin(dst)->sin_addr,
800 #if NCARP > 0
801 				    (rt->rt_ifp->if_type == IFT_CARP) ?
802 				    CLLADDR(rt->rt_ifp->if_sadl):
803 #endif
804 				    CLLADDR(ifp->if_sadl));
805 			else {
806 				rt->rt_flags |= RTF_REJECT;
807 				rt->rt_expire += arpt_down;
808 				la->la_asked = 0;
809 			}
810 		}
811 	}
812 	return 0;
813 }
814 
815 /*
816  * Common length and type checks are done here,
817  * then the protocol-specific routine is called.
818  */
819 void
820 arpintr(void)
821 {
822 	struct mbuf *m;
823 	struct arphdr *ar;
824 	int s;
825 	int arplen;
826 
827 	mutex_enter(softnet_lock);
828 	KERNEL_LOCK(1, NULL);
829 	while (arpintrq.ifq_head) {
830 		s = splnet();
831 		IF_DEQUEUE(&arpintrq, m);
832 		splx(s);
833 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
834 			panic("arpintr");
835 
836 		MCLAIM(m, &arpdomain.dom_mowner);
837 		ARP_STATINC(ARP_STAT_RCVTOTAL);
838 
839 		/*
840 		 * First, make sure we have at least struct arphdr.
841 		 */
842 		if (m->m_len < sizeof(struct arphdr) ||
843 		    (ar = mtod(m, struct arphdr *)) == NULL)
844 			goto badlen;
845 
846 		switch (m->m_pkthdr.rcvif->if_type) {
847 		case IFT_IEEE1394:
848 			arplen = sizeof(struct arphdr) +
849 			    ar->ar_hln + 2 * ar->ar_pln;
850 			break;
851 		default:
852 			arplen = sizeof(struct arphdr) +
853 			    2 * ar->ar_hln + 2 * ar->ar_pln;
854 			break;
855 		}
856 
857 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
858 		    m->m_len >= arplen)
859 			switch (ntohs(ar->ar_pro)) {
860 			case ETHERTYPE_IP:
861 			case ETHERTYPE_IPTRAILERS:
862 				in_arpinput(m);
863 				continue;
864 			default:
865 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
866 			}
867 		else {
868 badlen:
869 			ARP_STATINC(ARP_STAT_RCVBADLEN);
870 		}
871 		m_freem(m);
872 	}
873 	KERNEL_UNLOCK_ONE(NULL);
874 	mutex_exit(softnet_lock);
875 }
876 
877 /*
878  * ARP for Internet protocols on 10 Mb/s Ethernet.
879  * Algorithm is that given in RFC 826.
880  * In addition, a sanity check is performed on the sender
881  * protocol address, to catch impersonators.
882  * We no longer handle negotiations for use of trailer protocol:
883  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
884  * along with IP replies if we wanted trailers sent to us,
885  * and also sent them in response to IP replies.
886  * This allowed either end to announce the desire to receive
887  * trailer packets.
888  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
889  * but formerly didn't normally send requests.
890  */
891 static void
892 in_arpinput(struct mbuf *m)
893 {
894 	struct arphdr *ah;
895 	struct ifnet *ifp = m->m_pkthdr.rcvif;
896 	struct llinfo_arp *la = NULL;
897 	struct rtentry  *rt;
898 	struct in_ifaddr *ia;
899 #if NBRIDGE > 0
900 	struct in_ifaddr *bridge_ia = NULL;
901 #endif
902 #if NCARP > 0
903 	u_int32_t count = 0, index = 0;
904 #endif
905 	struct sockaddr_dl *sdl;
906 	struct sockaddr sa;
907 	struct in_addr isaddr, itaddr, myaddr;
908 	int op;
909 	struct mbuf *mold;
910 	void *tha;
911 	int s;
912 	uint64_t *arps;
913 
914 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
915 		goto out;
916 	ah = mtod(m, struct arphdr *);
917 	op = ntohs(ah->ar_op);
918 
919 	/*
920 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
921 	 * ar_tpa().
922 	 */
923 	switch (ifp->if_type) {
924 	case IFT_IEEE1394:
925 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
926 			;
927 		else {
928 			/* XXX this is to make sure we compute ar_tha right */
929 			/* XXX check ar_hrd more strictly? */
930 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
931 		}
932 		break;
933 	default:
934 		/* XXX check ar_hrd? */
935 		break;
936 	}
937 
938 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
939 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
940 
941 	if (m->m_flags & (M_BCAST|M_MCAST))
942 		ARP_STATINC(ARP_STAT_RCVMCAST);
943 
944 	/*
945 	 * If the target IP address is zero, ignore the packet.
946 	 * This prevents the code below from tring to answer
947 	 * when we are using IP address zero (booting).
948 	 */
949 	if (in_nullhost(itaddr)) {
950 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
951 		goto out;
952 	}
953 
954 	/*
955 	 * If the source IP address is zero, this is most likely a
956 	 * confused host trying to use IP address zero. (Windoze?)
957 	 * XXX: Should we bother trying to reply to these?
958 	 */
959 	if (in_nullhost(isaddr)) {
960 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
961 		goto out;
962 	}
963 
964 	/*
965 	 * Search for a matching interface address
966 	 * or any address on the interface to use
967 	 * as a dummy address in the rest of this function
968 	 */
969 
970 	INADDR_TO_IA(itaddr, ia);
971 	while (ia != NULL) {
972 #if NCARP > 0
973 		if (ia->ia_ifp->if_type == IFT_CARP &&
974 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
975 		    (IFF_UP|IFF_RUNNING))) {
976 			index++;
977 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
978 			    carp_iamatch(ia, ar_sha(ah),
979 			    &count, index)) {
980 				break;
981 				}
982 		} else
983 #endif
984 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
985 				break;
986 #if NBRIDGE > 0
987 		/*
988 		 * If the interface we received the packet on
989 		 * is part of a bridge, check to see if we need
990 		 * to "bridge" the packet to ourselves at this
991 		 * layer.  Note we still prefer a perfect match,
992 		 * but allow this weaker match if necessary.
993 		 */
994 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
995 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
996 			bridge_ia = ia;
997 #endif /* NBRIDGE > 0 */
998 
999 		NEXT_IA_WITH_SAME_ADDR(ia);
1000 	}
1001 
1002 #if NBRIDGE > 0
1003 	if (ia == NULL && bridge_ia != NULL) {
1004 		ia = bridge_ia;
1005 		ifp = bridge_ia->ia_ifp;
1006 	}
1007 #endif
1008 
1009 	if (ia == NULL) {
1010 		INADDR_TO_IA(isaddr, ia);
1011 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
1012 			NEXT_IA_WITH_SAME_ADDR(ia);
1013 
1014 		if (ia == NULL) {
1015 			IFP_TO_IA(ifp, ia);
1016 			if (ia == NULL) {
1017 				ARP_STATINC(ARP_STAT_RCVNOINT);
1018 				goto out;
1019 			}
1020 		}
1021 	}
1022 
1023 	myaddr = ia->ia_addr.sin_addr;
1024 
1025 	/* XXX checks for bridge case? */
1026 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
1027 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1028 		goto out;	/* it's from me, ignore it. */
1029 	}
1030 
1031 	/* XXX checks for bridge case? */
1032 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1033 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1034 		log(LOG_ERR,
1035 		    "%s: arp: link address is broadcast for IP address %s!\n",
1036 		    ifp->if_xname, in_fmtaddr(isaddr));
1037 		goto out;
1038 	}
1039 
1040 	if (in_hosteq(isaddr, myaddr)) {
1041 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1042 		log(LOG_ERR,
1043 		   "duplicate IP address %s sent from link address %s\n",
1044 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
1045 		itaddr = myaddr;
1046 		goto reply;
1047 	}
1048 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1049 	if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
1050 		if (sdl->sdl_alen &&
1051 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
1052 			if (rt->rt_flags & RTF_STATIC) {
1053 				ARP_STATINC(ARP_STAT_RCVOVERPERM);
1054 				log(LOG_INFO,
1055 				    "%s tried to overwrite permanent arp info"
1056 				    " for %s\n",
1057 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1058 				    in_fmtaddr(isaddr));
1059 				goto out;
1060 			} else if (rt->rt_ifp != ifp) {
1061 				ARP_STATINC(ARP_STAT_RCVOVERINT);
1062 				log(LOG_INFO,
1063 				    "%s on %s tried to overwrite "
1064 				    "arp info for %s on %s\n",
1065 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1066 				    ifp->if_xname, in_fmtaddr(isaddr),
1067 				    rt->rt_ifp->if_xname);
1068 				    goto out;
1069 			} else {
1070 				ARP_STATINC(ARP_STAT_RCVOVER);
1071 				log(LOG_INFO,
1072 				    "arp info overwritten for %s by %s\n",
1073 				    in_fmtaddr(isaddr),
1074 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1075 			}
1076 		}
1077 		/*
1078 		 * sanity check for the address length.
1079 		 * XXX this does not work for protocols with variable address
1080 		 * length. -is
1081 		 */
1082 		if (sdl->sdl_alen &&
1083 		    sdl->sdl_alen != ah->ar_hln) {
1084 			ARP_STATINC(ARP_STAT_RCVLENCHG);
1085 			log(LOG_WARNING,
1086 			    "arp from %s: new addr len %d, was %d",
1087 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1088 		}
1089 		if (ifp->if_addrlen != ah->ar_hln) {
1090 			ARP_STATINC(ARP_STAT_RCVBADLEN);
1091 			log(LOG_WARNING,
1092 			    "arp from %s: addr len: new %d, i/f %d (ignored)",
1093 			    in_fmtaddr(isaddr), ah->ar_hln,
1094 			    ifp->if_addrlen);
1095 			goto reply;
1096 		}
1097 #if NTOKEN > 0
1098 		/*
1099 		 * XXX uses m_data and assumes the complete answer including
1100 		 * XXX token-ring headers is in the same buf
1101 		 */
1102 		if (ifp->if_type == IFT_ISO88025) {
1103 			struct token_header *trh;
1104 
1105 			trh = (struct token_header *)M_TRHSTART(m);
1106 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1107 				struct token_rif	*rif;
1108 				size_t	riflen;
1109 
1110 				rif = TOKEN_RIF(trh);
1111 				riflen = (ntohs(rif->tr_rcf) &
1112 				    TOKEN_RCF_LEN_MASK) >> 8;
1113 
1114 				if (riflen > 2 &&
1115 				    riflen < sizeof(struct token_rif) &&
1116 				    (riflen & 1) == 0) {
1117 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1118 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1119 					memcpy(TOKEN_RIF(la), rif, riflen);
1120 				}
1121 			}
1122 		}
1123 #endif /* NTOKEN > 0 */
1124 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
1125 		    ah->ar_hln);
1126 		if (rt->rt_expire)
1127 			rt->rt_expire = time_second + arpt_keep;
1128 		rt->rt_flags &= ~RTF_REJECT;
1129 		la->la_asked = 0;
1130 
1131 		s = splnet();
1132 		mold = la->la_hold;
1133 		la->la_hold = 0;
1134 		splx(s);
1135 
1136 		if (mold) {
1137 			ARP_STATINC(ARP_STAT_DFRSENT);
1138 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
1139 		}
1140 	}
1141 reply:
1142 	if (op != ARPOP_REQUEST) {
1143 		if (op == ARPOP_REPLY)
1144 			ARP_STATINC(ARP_STAT_RCVREPLY);
1145 	out:
1146 		m_freem(m);
1147 		return;
1148 	}
1149 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1150 	if (in_hosteq(itaddr, myaddr)) {
1151 		/* I am the target */
1152 		tha = ar_tha(ah);
1153 		if (tha)
1154 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1155 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1156 	} else {
1157 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1158 		if (la == NULL)
1159 			goto out;
1160 		rt = la->la_rt;
1161 		if (rt->rt_ifp->if_type == IFT_CARP &&
1162 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1163 			goto out;
1164 		tha = ar_tha(ah);
1165 		if (tha)
1166 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1167 		sdl = satosdl(rt->rt_gateway);
1168 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
1169 	}
1170 
1171 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1172 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1173 	ah->ar_op = htons(ARPOP_REPLY);
1174 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1175 	switch (ifp->if_type) {
1176 	case IFT_IEEE1394:
1177 		/*
1178 		 * ieee1394 arp reply is broadcast
1179 		 */
1180 		m->m_flags &= ~M_MCAST;
1181 		m->m_flags |= M_BCAST;
1182 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1183 		break;
1184 
1185 	default:
1186 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1187 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1188 		break;
1189 	}
1190 	m->m_pkthdr.len = m->m_len;
1191 	sa.sa_family = AF_ARP;
1192 	sa.sa_len = 2;
1193 	arps = ARP_STAT_GETREF();
1194 	arps[ARP_STAT_SNDTOTAL]++;
1195 	arps[ARP_STAT_SNDREPLY]++;
1196 	ARP_STAT_PUTREF();
1197 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1198 	return;
1199 }
1200 
1201 /*
1202  * Free an arp entry.
1203  */
1204 static void arptfree(struct llinfo_arp *la)
1205 {
1206 	struct rtentry *rt = la->la_rt;
1207 	struct sockaddr_dl *sdl;
1208 
1209 	ARP_LOCK_CHECK();
1210 
1211 	if (rt == NULL)
1212 		panic("arptfree");
1213 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
1214 	    sdl->sdl_family == AF_LINK) {
1215 		sdl->sdl_alen = 0;
1216 		la->la_asked = 0;
1217 		rt->rt_flags &= ~RTF_REJECT;
1218 		return;
1219 	}
1220 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1221 }
1222 
1223 static struct llinfo_arp *
1224 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
1225 {
1226 	return arplookup1(m, addr, create, proxy, NULL);
1227 }
1228 
1229 /*
1230  * Lookup or enter a new address in arptab.
1231  */
1232 static struct llinfo_arp *
1233 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
1234     struct rtentry *rt0)
1235 {
1236 	struct arphdr *ah;
1237 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1238 	struct rtentry *rt;
1239 	struct sockaddr_inarp sin;
1240 	const char *why = NULL;
1241 
1242 	ah = mtod(m, struct arphdr *);
1243 	if (rt0 == NULL) {
1244 		memset(&sin, 0, sizeof(sin));
1245 		sin.sin_len = sizeof(sin);
1246 		sin.sin_family = AF_INET;
1247 		sin.sin_addr = *addr;
1248 		sin.sin_other = proxy ? SIN_PROXY : 0;
1249 		rt = rtalloc1(sintosa(&sin), create);
1250 		if (rt == NULL)
1251 			return NULL;
1252 		rt->rt_refcnt--;
1253 	} else
1254 		rt = rt0;
1255 
1256 #define	IS_LLINFO(__rt)							  \
1257 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
1258 	 (__rt)->rt_gateway->sa_family == AF_LINK)
1259 
1260 
1261 	if (IS_LLINFO(rt))
1262 		return (struct llinfo_arp *)rt->rt_llinfo;
1263 
1264 	if (create) {
1265 		if (rt->rt_flags & RTF_GATEWAY)
1266 			why = "host is not on local network";
1267 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1268 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
1269 			why = "could not allocate llinfo";
1270 		} else
1271 			why = "gateway route is not ours";
1272 		log(LOG_DEBUG, "arplookup: unable to enter address"
1273 		    " for %s@%s on %s (%s)\n",
1274 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1275 		    (ifp) ? ifp->if_xname : "null", why);
1276 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1277 			rtrequest(RTM_DELETE, rt_getkey(rt),
1278 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1279 		}
1280 	}
1281 	return NULL;
1282 }
1283 
1284 int
1285 arpioctl(u_long cmd, void *data)
1286 {
1287 
1288 	return EOPNOTSUPP;
1289 }
1290 
1291 void
1292 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1293 {
1294 	struct in_addr *ip;
1295 
1296 	/*
1297 	 * Warn the user if another station has this IP address,
1298 	 * but only if the interface IP address is not zero.
1299 	 */
1300 	ip = &IA_SIN(ifa)->sin_addr;
1301 	if (!in_nullhost(*ip))
1302 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
1303 
1304 	ifa->ifa_rtrequest = arp_rtrequest;
1305 	ifa->ifa_flags |= RTF_CLONING;
1306 }
1307 
1308 /*
1309  * Called from 10 Mb/s Ethernet interrupt handlers
1310  * when ether packet type ETHERTYPE_REVARP
1311  * is received.  Common length and type checks are done here,
1312  * then the protocol-specific routine is called.
1313  */
1314 void
1315 revarpinput(struct mbuf *m)
1316 {
1317 	struct arphdr *ar;
1318 
1319 	if (m->m_len < sizeof(struct arphdr))
1320 		goto out;
1321 	ar = mtod(m, struct arphdr *);
1322 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1323 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1324 		goto out;
1325 #endif
1326 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1327 		goto out;
1328 	switch (ntohs(ar->ar_pro)) {
1329 	case ETHERTYPE_IP:
1330 	case ETHERTYPE_IPTRAILERS:
1331 		in_revarpinput(m);
1332 		return;
1333 
1334 	default:
1335 		break;
1336 	}
1337 out:
1338 	m_freem(m);
1339 }
1340 
1341 /*
1342  * RARP for Internet protocols on 10 Mb/s Ethernet.
1343  * Algorithm is that given in RFC 903.
1344  * We are only using for bootstrap purposes to get an ip address for one of
1345  * our interfaces.  Thus we support no user-interface.
1346  *
1347  * Since the contents of the RARP reply are specific to the interface that
1348  * sent the request, this code must ensure that they are properly associated.
1349  *
1350  * Note: also supports ARP via RARP packets, per the RFC.
1351  */
1352 void
1353 in_revarpinput(struct mbuf *m)
1354 {
1355 	struct ifnet *ifp;
1356 	struct arphdr *ah;
1357 	void *tha;
1358 	int op;
1359 
1360 	ah = mtod(m, struct arphdr *);
1361 	op = ntohs(ah->ar_op);
1362 
1363 	switch (m->m_pkthdr.rcvif->if_type) {
1364 	case IFT_IEEE1394:
1365 		/* ARP without target hardware address is not supported */
1366 		goto out;
1367 	default:
1368 		break;
1369 	}
1370 
1371 	switch (op) {
1372 	case ARPOP_REQUEST:
1373 	case ARPOP_REPLY:	/* per RFC */
1374 		in_arpinput(m);
1375 		return;
1376 	case ARPOP_REVREPLY:
1377 		break;
1378 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1379 	default:
1380 		goto out;
1381 	}
1382 	if (!revarp_in_progress)
1383 		goto out;
1384 	ifp = m->m_pkthdr.rcvif;
1385 	if (ifp != myip_ifp) /* !same interface */
1386 		goto out;
1387 	if (myip_initialized)
1388 		goto wake;
1389 	tha = ar_tha(ah);
1390 	KASSERT(tha);
1391 	if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1392 		goto out;
1393 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1394 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1395 	myip_initialized = 1;
1396 wake:	/* Do wakeup every time in case it was missed. */
1397 	wakeup((void *)&myip);
1398 
1399 out:
1400 	m_freem(m);
1401 }
1402 
1403 /*
1404  * Send a RARP request for the ip address of the specified interface.
1405  * The request should be RFC 903-compliant.
1406  */
1407 void
1408 revarprequest(struct ifnet *ifp)
1409 {
1410 	struct sockaddr sa;
1411 	struct mbuf *m;
1412 	struct arphdr *ah;
1413 	void *tha;
1414 
1415 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1416 		return;
1417 	MCLAIM(m, &arpdomain.dom_mowner);
1418 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1419 	    2*ifp->if_addrlen;
1420 	m->m_pkthdr.len = m->m_len;
1421 	MH_ALIGN(m, m->m_len);
1422 	ah = mtod(m, struct arphdr *);
1423 	memset(ah, 0, m->m_len);
1424 	ah->ar_pro = htons(ETHERTYPE_IP);
1425 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1426 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1427 	ah->ar_op = htons(ARPOP_REVREQUEST);
1428 
1429 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1430 	tha = ar_tha(ah);
1431 	KASSERT(tha);
1432 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1433 
1434 	sa.sa_family = AF_ARP;
1435 	sa.sa_len = 2;
1436 	m->m_flags |= M_BCAST;
1437 	(*ifp->if_output)(ifp, m, &sa, NULL);
1438 
1439 }
1440 
1441 /*
1442  * RARP for the ip address of the specified interface, but also
1443  * save the ip address of the server that sent the answer.
1444  * Timeout if no response is received.
1445  */
1446 int
1447 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1448     struct in_addr *clnt_in)
1449 {
1450 	int result, count = 20;
1451 
1452 	myip_initialized = 0;
1453 	myip_ifp = ifp;
1454 
1455 	revarp_in_progress = 1;
1456 	while (count--) {
1457 		revarprequest(ifp);
1458 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1459 		if (result != EWOULDBLOCK)
1460 			break;
1461 	}
1462 	revarp_in_progress = 0;
1463 
1464 	if (!myip_initialized)
1465 		return ENETUNREACH;
1466 
1467 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1468 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1469 	return 0;
1470 }
1471 
1472 
1473 
1474 #ifdef DDB
1475 
1476 #include <machine/db_machdep.h>
1477 #include <ddb/db_interface.h>
1478 #include <ddb/db_output.h>
1479 
1480 static void
1481 db_print_sa(const struct sockaddr *sa)
1482 {
1483 	int len;
1484 	const u_char *p;
1485 
1486 	if (sa == NULL) {
1487 		db_printf("[NULL]");
1488 		return;
1489 	}
1490 
1491 	p = (const u_char *)sa;
1492 	len = sa->sa_len;
1493 	db_printf("[");
1494 	while (len > 0) {
1495 		db_printf("%d", *p);
1496 		p++; len--;
1497 		if (len) db_printf(",");
1498 	}
1499 	db_printf("]\n");
1500 }
1501 
1502 static void
1503 db_print_ifa(struct ifaddr *ifa)
1504 {
1505 	if (ifa == NULL)
1506 		return;
1507 	db_printf("  ifa_addr=");
1508 	db_print_sa(ifa->ifa_addr);
1509 	db_printf("  ifa_dsta=");
1510 	db_print_sa(ifa->ifa_dstaddr);
1511 	db_printf("  ifa_mask=");
1512 	db_print_sa(ifa->ifa_netmask);
1513 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1514 			  ifa->ifa_flags,
1515 			  ifa->ifa_refcnt,
1516 			  ifa->ifa_metric);
1517 }
1518 
1519 static void
1520 db_print_llinfo(void *li)
1521 {
1522 	struct llinfo_arp *la;
1523 
1524 	if (li == NULL)
1525 		return;
1526 	la = (struct llinfo_arp *)li;
1527 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1528 			  la->la_rt, la->la_hold, la->la_asked);
1529 }
1530 
1531 /*
1532  * Function to pass to rt_walktree().
1533  * Return non-zero error to abort walk.
1534  */
1535 static int
1536 db_show_rtentry(struct rtentry *rt, void *w)
1537 {
1538 	db_printf("rtentry=%p", rt);
1539 
1540 	db_printf(" flags=0x%x refcnt=%d use=%ld expire=%ld\n",
1541 			  rt->rt_flags, rt->rt_refcnt,
1542 			  rt->rt_use, rt->rt_expire);
1543 
1544 	db_printf(" key="); db_print_sa(rt_getkey(rt));
1545 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1546 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1547 
1548 	db_printf(" ifp=%p ", rt->rt_ifp);
1549 	if (rt->rt_ifp)
1550 		db_printf("(%s)", rt->rt_ifp->if_xname);
1551 	else
1552 		db_printf("(NULL)");
1553 
1554 	db_printf(" ifa=%p\n", rt->rt_ifa);
1555 	db_print_ifa(rt->rt_ifa);
1556 
1557 	db_printf(" gwroute=%p llinfo=%p\n",
1558 			  rt->rt_gwroute, rt->rt_llinfo);
1559 	db_print_llinfo(rt->rt_llinfo);
1560 
1561 	return 0;
1562 }
1563 
1564 /*
1565  * Function to print all the route trees.
1566  * Use this from ddb:  "show arptab"
1567  */
1568 void
1569 db_show_arptab(db_expr_t addr, bool have_addr,
1570     db_expr_t count, const char *modif)
1571 {
1572 	rt_walktree(AF_INET, db_show_rtentry, NULL);
1573 }
1574 #endif
1575 
1576 static int
1577 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1578 {
1579 
1580 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
1581 }
1582 
1583 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup")
1584 {
1585 	const struct sysctlnode *node;
1586 
1587 	sysctl_createv(clog, 0, NULL, NULL,
1588 			CTLFLAG_PERMANENT,
1589 			CTLTYPE_NODE, "net", NULL,
1590 			NULL, 0, NULL, 0,
1591 			CTL_NET, CTL_EOL);
1592 	sysctl_createv(clog, 0, NULL, NULL,
1593 			CTLFLAG_PERMANENT,
1594 			CTLTYPE_NODE, "inet", NULL,
1595 			NULL, 0, NULL, 0,
1596 			CTL_NET, PF_INET, CTL_EOL);
1597 	sysctl_createv(clog, 0, NULL, &node,
1598 			CTLFLAG_PERMANENT,
1599 			CTLTYPE_NODE, "arp",
1600 			SYSCTL_DESCR("Address Resolution Protocol"),
1601 			NULL, 0, NULL, 0,
1602 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1603 
1604 	sysctl_createv(clog, 0, NULL, NULL,
1605 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1606 			CTLTYPE_INT, "prune",
1607 			SYSCTL_DESCR("ARP cache pruning interval"),
1608 			NULL, 0, &arpt_prune, 0,
1609 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1610 
1611 	sysctl_createv(clog, 0, NULL, NULL,
1612 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1613 			CTLTYPE_INT, "keep",
1614 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1615 			NULL, 0, &arpt_keep, 0,
1616 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1617 
1618 	sysctl_createv(clog, 0, NULL, NULL,
1619 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1620 			CTLTYPE_INT, "down",
1621 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1622 			NULL, 0, &arpt_down, 0,
1623 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1624 
1625 	sysctl_createv(clog, 0, NULL, NULL,
1626 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1627 			CTLTYPE_INT, "refresh",
1628 			SYSCTL_DESCR("ARP entry refresh interval"),
1629 			NULL, 0, &arpt_refresh, 0,
1630 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1631 
1632 	sysctl_createv(clog, 0, NULL, NULL,
1633 			CTLFLAG_PERMANENT,
1634 			CTLTYPE_STRUCT, "stats",
1635 			SYSCTL_DESCR("ARP statistics"),
1636 			sysctl_net_inet_arp_stats, 0, NULL, 0,
1637 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1638 }
1639 
1640 #endif /* INET */
1641