xref: /netbsd-src/sys/netinet/if_arp.c (revision 1b9578b8c2c1f848eeb16dabbfd7d1f0d9fdefbd)
1 /*	$NetBSD: if_arp.c,v 1.151 2011/05/03 16:00:29 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 1.151 2011/05/03 16:00:29 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 static	void arp_drainstub(void);
155 
156 LIST_HEAD(, llinfo_arp) llinfo_arp;
157 struct	ifqueue arpintrq = {
158 	.ifq_head = NULL,
159 	.ifq_tail = NULL,
160 	.ifq_len = 0,
161 	.ifq_maxlen = 50,
162 	.ifq_drops = 0,
163 };
164 int	arp_inuse, arp_allocated, arp_intimer;
165 int	arp_maxtries = 5;
166 int	useloopback = 1;	/* use loopback interface for local traffic */
167 int	arpinit_done = 0;
168 
169 static percpu_t *arpstat_percpu;
170 
171 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
172 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
173 
174 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
175 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
176 
177 struct	callout arptimer_ch;
178 
179 /* revarp state */
180 struct	in_addr myip, srv_ip;
181 int	myip_initialized = 0;
182 int	revarp_in_progress = 0;
183 struct	ifnet *myip_ifp = NULL;
184 
185 #ifdef DDB
186 static void db_print_sa(const struct sockaddr *);
187 static void db_print_ifa(struct ifaddr *);
188 static void db_print_llinfo(void *);
189 static int db_show_rtentry(struct rtentry *, void *);
190 #endif
191 
192 static int arp_drainwanted;
193 
194 /*
195  * this should be elsewhere.
196  */
197 
198 static char *
199 lla_snprintf(u_int8_t *, int);
200 
201 static char *
202 lla_snprintf(u_int8_t *adrp, int len)
203 {
204 #define NUMBUFS 3
205 	static char buf[NUMBUFS][16*3];
206 	static int bnum = 0;
207 
208 	int i;
209 	char *p;
210 
211 	p = buf[bnum];
212 
213 	*p++ = hexdigits[(*adrp)>>4];
214 	*p++ = hexdigits[(*adrp++)&0xf];
215 
216 	for (i=1; i<len && i<16; i++) {
217 		*p++ = ':';
218 		*p++ = hexdigits[(*adrp)>>4];
219 		*p++ = hexdigits[(*adrp++)&0xf];
220 	}
221 
222 	*p = 0;
223 	p = buf[bnum];
224 	bnum = (bnum + 1) % NUMBUFS;
225 	return p;
226 }
227 
228 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
229 
230 static void
231 arp_fasttimo(void)
232 {
233 	if (arp_drainwanted) {
234 		arp_drain();
235 		arp_drainwanted = 0;
236 	}
237 }
238 
239 const struct protosw arpsw[] = {
240 	{ .pr_type = 0,
241 	  .pr_domain = &arpdomain,
242 	  .pr_protocol = 0,
243 	  .pr_flags = 0,
244 	  .pr_input = 0,
245 	  .pr_output = 0,
246 	  .pr_ctlinput = 0,
247 	  .pr_ctloutput = 0,
248 	  .pr_usrreq =  0,
249 	  .pr_init = arp_init,
250 	  .pr_fasttimo = arp_fasttimo,
251 	  .pr_slowtimo = 0,
252 	  .pr_drain = arp_drainstub,
253 	}
254 };
255 
256 
257 struct domain arpdomain = {
258 	.dom_family = PF_ARP,
259 	.dom_name = "arp",
260 	.dom_protosw = arpsw,
261 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
262 };
263 
264 /*
265  * ARP table locking.
266  *
267  * to prevent lossage vs. the arp_drain routine (which may be called at
268  * any time, including in a device driver context), we do two things:
269  *
270  * 1) manipulation of la->la_hold is done at splnet() (for all of
271  * about two instructions).
272  *
273  * 2) manipulation of the arp table's linked list is done under the
274  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
275  * while the arp table is locked, we punt and try again later.
276  */
277 
278 static int	arp_locked;
279 static inline int arp_lock_try(int);
280 static inline void arp_unlock(void);
281 
282 static inline int
283 arp_lock_try(int recurse)
284 {
285 	int s;
286 
287 	/*
288 	 * Use splvm() -- we're blocking things that would cause
289 	 * mbuf allocation.
290 	 */
291 	s = splvm();
292 	if (!recurse && arp_locked) {
293 		splx(s);
294 		return 0;
295 	}
296 	arp_locked++;
297 	splx(s);
298 	return 1;
299 }
300 
301 static inline void
302 arp_unlock(void)
303 {
304 	int s;
305 
306 	s = splvm();
307 	arp_locked--;
308 	splx(s);
309 }
310 
311 #ifdef DIAGNOSTIC
312 #define	ARP_LOCK(recurse)						\
313 do {									\
314 	if (arp_lock_try(recurse) == 0) {				\
315 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
316 		panic("arp_lock");					\
317 	}								\
318 } while (/*CONSTCOND*/ 0)
319 #define	ARP_LOCK_CHECK()						\
320 do {									\
321 	if (arp_locked == 0) {						\
322 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
323 		panic("arp lock check");				\
324 	}								\
325 } while (/*CONSTCOND*/ 0)
326 #else
327 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
328 #define	ARP_LOCK_CHECK()	/* nothing */
329 #endif
330 
331 #define	ARP_UNLOCK()		arp_unlock()
332 
333 static void sysctl_net_inet_arp_setup(struct sysctllog **);
334 
335 void
336 arp_init(void)
337 {
338 
339 	sysctl_net_inet_arp_setup(NULL);
340 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
341 }
342 
343 static void
344 arp_drainstub(void)
345 {
346 	arp_drainwanted = 1;
347 }
348 
349 /*
350  * ARP protocol drain routine.  Called when memory is in short supply.
351  * Called at splvm();  don't acquire softnet_lock as can be called from
352  * hardware interrupt handlers.
353  */
354 void
355 arp_drain(void)
356 {
357 	struct llinfo_arp *la, *nla;
358 	int count = 0;
359 	struct mbuf *mold;
360 
361 	KERNEL_LOCK(1, NULL);
362 
363 	if (arp_lock_try(0) == 0) {
364 		KERNEL_UNLOCK_ONE(NULL);
365 		return;
366 	}
367 
368 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
369 		nla = LIST_NEXT(la, la_list);
370 
371 		mold = la->la_hold;
372 		la->la_hold = 0;
373 
374 		if (mold) {
375 			m_freem(mold);
376 			count++;
377 		}
378 	}
379 	ARP_UNLOCK();
380 	ARP_STATADD(ARP_STAT_DFRDROPPED, count);
381 	KERNEL_UNLOCK_ONE(NULL);
382 }
383 
384 
385 /*
386  * Timeout routine.  Age arp_tab entries periodically.
387  */
388 /* ARGSUSED */
389 static void
390 arptimer(void *arg)
391 {
392 	struct llinfo_arp *la, *nla;
393 
394 	mutex_enter(softnet_lock);
395 	KERNEL_LOCK(1, NULL);
396 
397 	if (arp_lock_try(0) == 0) {
398 		/* get it later.. */
399 		KERNEL_UNLOCK_ONE(NULL);
400 		mutex_exit(softnet_lock);
401 		return;
402 	}
403 
404 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
405 	for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) {
406 		struct rtentry *rt = la->la_rt;
407 
408 		nla = LIST_NEXT(la, la_list);
409 		if (rt->rt_expire == 0)
410 			continue;
411 		if ((rt->rt_expire - time_second) < arpt_refresh &&
412 		    rt->rt_pksent > (time_second - arpt_keep)) {
413 			/*
414 			 * If the entry has been used during since last
415 			 * refresh, try to renew it before deleting.
416 			 */
417 			arprequest(rt->rt_ifp,
418 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
419 			    &satocsin(rt_getkey(rt))->sin_addr,
420 			    CLLADDR(rt->rt_ifp->if_sadl));
421 		} else if (rt->rt_expire <= time_second)
422 			arptfree(la); /* timer has expired; clear */
423 	}
424 
425 	ARP_UNLOCK();
426 
427 	KERNEL_UNLOCK_ONE(NULL);
428 	mutex_exit(softnet_lock);
429 }
430 
431 /*
432  * We set the gateway for RTF_CLONING routes to a "prototype"
433  * link-layer sockaddr whose interface type (if_type) and interface
434  * index (if_index) fields are prepared.
435  */
436 static struct sockaddr *
437 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
438     const struct sockaddr *netmask)
439 {
440 	const struct ifnet *ifp = rt->rt_ifp;
441 	uint8_t namelen = strlen(ifp->if_xname);
442 	uint8_t addrlen = ifp->if_addrlen;
443 
444 	/*
445 	 * XXX: If this is a manually added route to interface
446 	 * such as older version of routed or gated might provide,
447 	 * restore cloning bit.
448 	 */
449 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
450 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
451 		rt->rt_flags |= RTF_CLONING;
452 	if (rt->rt_flags & RTF_CLONING) {
453 		union {
454 			struct sockaddr sa;
455 			struct sockaddr_storage ss;
456 			struct sockaddr_dl sdl;
457 		} u;
458 		/*
459 		 * Case 1: This route should come from a route to iface.
460 		 */
461 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
462 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
463 		rt_setgate(rt, &u.sa);
464 		gate = rt->rt_gateway;
465 	}
466 	return gate;
467 }
468 
469 /*
470  * Parallel to llc_rtrequest.
471  */
472 void
473 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
474 {
475 	struct sockaddr *gate = rt->rt_gateway;
476 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
477 	size_t allocsize;
478 	struct mbuf *mold;
479 	int s;
480 	struct in_ifaddr *ia;
481 	struct ifaddr *ifa;
482 	struct ifnet *ifp = rt->rt_ifp;
483 
484 	if (!arpinit_done) {
485 		arpinit_done = 1;
486 		/*
487 		 * We generate expiration times from time_second
488 		 * so avoid accidentally creating permanent routes.
489 		 */
490 		if (time_second == 0) {
491 			struct timespec ts;
492 			ts.tv_sec = 1;
493 			ts.tv_nsec = 0;
494 			tc_setclock(&ts);
495 		}
496 		callout_init(&arptimer_ch, CALLOUT_MPSAFE);
497 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
498 	}
499 
500 	if (req == RTM_LLINFO_UPD) {
501 		struct in_addr *in;
502 
503 		if ((ifa = info->rti_ifa) == NULL)
504 			return;
505 
506 		in = &ifatoia(ifa)->ia_addr.sin_addr;
507 
508 		arprequest(ifa->ifa_ifp, in, in,
509 		    CLLADDR(ifa->ifa_ifp->if_sadl));
510 		return;
511 	}
512 
513 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
514 		if (req != RTM_ADD)
515 			return;
516 
517 		/*
518 		 * linklayers with particular link MTU limitation.
519 		 */
520 		switch(ifp->if_type) {
521 #if NFDDI > 0
522 		case IFT_FDDI:
523 			if (ifp->if_mtu > FDDIIPMTU)
524 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
525 			break;
526 #endif
527 #if NARC > 0
528 		case IFT_ARCNET:
529 		    {
530 			int arcipifmtu;
531 
532 			if (ifp->if_flags & IFF_LINK0)
533 				arcipifmtu = arc_ipmtu;
534 			else
535 				arcipifmtu = ARCMTU;
536 			if (ifp->if_mtu > arcipifmtu)
537 				rt->rt_rmx.rmx_mtu = arcipifmtu;
538 			break;
539 		    }
540 #endif
541 		}
542 		return;
543 	}
544 
545 	ARP_LOCK(1);		/* we may already be locked here. */
546 
547 	switch (req) {
548 	case RTM_SETGATE:
549 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
550 		break;
551 	case RTM_ADD:
552 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
553 		if (rt->rt_flags & RTF_CLONING) {
554 			/*
555 			 * Give this route an expiration time, even though
556 			 * it's a "permanent" route, so that routes cloned
557 			 * from it do not need their expiration time set.
558 			 */
559 			rt->rt_expire = time_second;
560 			/*
561 			 * linklayers with particular link MTU limitation.
562 			 */
563 			switch (ifp->if_type) {
564 #if NFDDI > 0
565 			case IFT_FDDI:
566 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
567 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
568 				     (rt->rt_rmx.rmx_mtu == 0 &&
569 				      ifp->if_mtu > FDDIIPMTU)))
570 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
571 				break;
572 #endif
573 #if NARC > 0
574 			case IFT_ARCNET:
575 			    {
576 				int arcipifmtu;
577 				if (ifp->if_flags & IFF_LINK0)
578 					arcipifmtu = arc_ipmtu;
579 				else
580 					arcipifmtu = ARCMTU;
581 
582 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
583 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
584 				     (rt->rt_rmx.rmx_mtu == 0 &&
585 				      ifp->if_mtu > arcipifmtu)))
586 					rt->rt_rmx.rmx_mtu = arcipifmtu;
587 				break;
588 			    }
589 #endif
590 			}
591 			break;
592 		}
593 		/* Announce a new entry if requested. */
594 		if (rt->rt_flags & RTF_ANNOUNCE) {
595 			arprequest(ifp,
596 			    &satocsin(rt_getkey(rt))->sin_addr,
597 			    &satocsin(rt_getkey(rt))->sin_addr,
598 			    CLLADDR(satocsdl(gate)));
599 		}
600 		/*FALLTHROUGH*/
601 	case RTM_RESOLVE:
602 		if (gate->sa_family != AF_LINK ||
603 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
604 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
605 			break;
606 		}
607 		satosdl(gate)->sdl_type = ifp->if_type;
608 		satosdl(gate)->sdl_index = ifp->if_index;
609 		if (la != NULL)
610 			break; /* This happens on a route change */
611 		/*
612 		 * Case 2:  This route may come from cloning, or a manual route
613 		 * add with a LL address.
614 		 */
615 		switch (ifp->if_type) {
616 #if NTOKEN > 0
617 		case IFT_ISO88025:
618 			allocsize = sizeof(*la) + sizeof(struct token_rif);
619 			break;
620 #endif /* NTOKEN > 0 */
621 		default:
622 			allocsize = sizeof(*la);
623 		}
624 		R_Malloc(la, struct llinfo_arp *, allocsize);
625 		rt->rt_llinfo = (void *)la;
626 		if (la == NULL) {
627 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
628 			break;
629 		}
630 		arp_inuse++, arp_allocated++;
631 		memset(la, 0, allocsize);
632 		la->la_rt = rt;
633 		rt->rt_flags |= RTF_LLINFO;
634 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
635 
636 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
637 		while (ia && ia->ia_ifp != ifp)
638 			NEXT_IA_WITH_SAME_ADDR(ia);
639 		if (ia) {
640 			/*
641 			 * This test used to be
642 			 *	if (lo0ifp->if_flags & IFF_UP)
643 			 * It allowed local traffic to be forced through
644 			 * the hardware by configuring the loopback down.
645 			 * However, it causes problems during network
646 			 * configuration for boards that can't receive
647 			 * packets they send.  It is now necessary to clear
648 			 * "useloopback" and remove the route to force
649 			 * traffic out to the hardware.
650 			 *
651 			 * In 4.4BSD, the above "if" statement checked
652 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
653 			 * to the current form so that we can provide a
654 			 * better support for multiple IPv4 addresses on a
655 			 * interface.
656 			 */
657 			rt->rt_expire = 0;
658 			if (sockaddr_dl_init(satosdl(gate), gate->sa_len,
659 			    ifp->if_index, ifp->if_type, NULL, 0,
660 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen) == NULL) {
661 				panic("%s(%s): sockaddr_dl_init cannot fail",
662 				    __func__, ifp->if_xname);
663 			}
664 			if (useloopback)
665 				ifp = rt->rt_ifp = lo0ifp;
666 			/*
667 			 * make sure to set rt->rt_ifa to the interface
668 			 * address we are using, otherwise we will have trouble
669 			 * with source address selection.
670 			 */
671 			ifa = &ia->ia_ifa;
672 			if (ifa != rt->rt_ifa)
673 				rt_replace_ifa(rt, ifa);
674 		}
675 		break;
676 
677 	case RTM_DELETE:
678 		if (la == NULL)
679 			break;
680 		arp_inuse--;
681 		LIST_REMOVE(la, la_list);
682 		rt->rt_llinfo = NULL;
683 		rt->rt_flags &= ~RTF_LLINFO;
684 
685 		s = splnet();
686 		mold = la->la_hold;
687 		la->la_hold = 0;
688 		splx(s);
689 
690 		if (mold)
691 			m_freem(mold);
692 
693 		Free((void *)la);
694 	}
695 	ARP_UNLOCK();
696 }
697 
698 /*
699  * Broadcast an ARP request. Caller specifies:
700  *	- arp header source ip address
701  *	- arp header target ip address
702  *	- arp header source ethernet address
703  */
704 void
705 arprequest(struct ifnet *ifp,
706     const struct in_addr *sip, const struct in_addr *tip,
707     const u_int8_t *enaddr)
708 {
709 	struct mbuf *m;
710 	struct arphdr *ah;
711 	struct sockaddr sa;
712 	uint64_t *arps;
713 
714 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
715 		return;
716 	MCLAIM(m, &arpdomain.dom_mowner);
717 	switch (ifp->if_type) {
718 	case IFT_IEEE1394:
719 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
720 		    ifp->if_addrlen;
721 		break;
722 	default:
723 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
724 		    2 * ifp->if_addrlen;
725 		break;
726 	}
727 	m->m_pkthdr.len = m->m_len;
728 	MH_ALIGN(m, m->m_len);
729 	ah = mtod(m, struct arphdr *);
730 	memset(ah, 0, m->m_len);
731 	switch (ifp->if_type) {
732 	case IFT_IEEE1394:	/* RFC2734 */
733 		/* fill it now for ar_tpa computation */
734 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
735 		break;
736 	default:
737 		/* ifp->if_output will fill ar_hrd */
738 		break;
739 	}
740 	ah->ar_pro = htons(ETHERTYPE_IP);
741 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
742 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
743 	ah->ar_op = htons(ARPOP_REQUEST);
744 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
745 	memcpy(ar_spa(ah), sip, ah->ar_pln);
746 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
747 	sa.sa_family = AF_ARP;
748 	sa.sa_len = 2;
749 	m->m_flags |= M_BCAST;
750 	arps = ARP_STAT_GETREF();
751 	arps[ARP_STAT_SNDTOTAL]++;
752 	arps[ARP_STAT_SENDREQUEST]++;
753 	ARP_STAT_PUTREF();
754 	(*ifp->if_output)(ifp, m, &sa, NULL);
755 }
756 
757 /*
758  * Resolve an IP address into an ethernet address.  If success,
759  * desten is filled in.  If there is no entry in arptab,
760  * set one up and broadcast a request for the IP address.
761  * Hold onto this mbuf and resend it once the address
762  * is finally resolved.  A return value of 1 indicates
763  * that desten has been filled in and the packet should be sent
764  * normally; a 0 return indicates that the packet has been
765  * taken over here, either now or for later transmission.
766  */
767 int
768 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
769     const struct sockaddr *dst, u_char *desten)
770 {
771 	struct llinfo_arp *la;
772 	const struct sockaddr_dl *sdl;
773 	struct mbuf *mold;
774 	int s;
775 
776 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
777 		rt = la->la_rt;
778 
779 	if (la == NULL || rt == NULL) {
780 		ARP_STATINC(ARP_STAT_ALLOCFAIL);
781 		log(LOG_DEBUG,
782 		    "arpresolve: can't allocate llinfo on %s for %s\n",
783 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
784 		m_freem(m);
785 		return 0;
786 	}
787 	sdl = satocsdl(rt->rt_gateway);
788 	/*
789 	 * Check the address family and length is valid, the address
790 	 * is resolved; otherwise, try to resolve.
791 	 */
792 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
793 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
794 		memcpy(desten, CLLADDR(sdl),
795 		    min(sdl->sdl_alen, ifp->if_addrlen));
796 		rt->rt_pksent = time_second; /* Time for last pkt sent */
797 		return 1;
798 	}
799 	/*
800 	 * There is an arptab entry, but no ethernet address
801 	 * response yet.  Replace the held mbuf with this
802 	 * latest one.
803 	 */
804 
805 	ARP_STATINC(ARP_STAT_DFRTOTAL);
806 	s = splnet();
807 	mold = la->la_hold;
808 	la->la_hold = m;
809 	splx(s);
810 
811 	if (mold) {
812 		ARP_STATINC(ARP_STAT_DFRDROPPED);
813 		m_freem(mold);
814 	}
815 
816 	/*
817 	 * Re-send the ARP request when appropriate.
818 	 */
819 #ifdef	DIAGNOSTIC
820 	if (rt->rt_expire == 0) {
821 		/* This should never happen. (Should it? -gwr) */
822 		printf("arpresolve: unresolved and rt_expire == 0\n");
823 		/* Set expiration time to now (expired). */
824 		rt->rt_expire = time_second;
825 	}
826 #endif
827 	if (rt->rt_expire) {
828 		rt->rt_flags &= ~RTF_REJECT;
829 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
830 			rt->rt_expire = time_second;
831 			if (la->la_asked++ < arp_maxtries) {
832 				arprequest(ifp,
833 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
834 				    &satocsin(dst)->sin_addr,
835 #if NCARP > 0
836 				    (rt->rt_ifp->if_type == IFT_CARP) ?
837 				    CLLADDR(rt->rt_ifp->if_sadl):
838 #endif
839 				    CLLADDR(ifp->if_sadl));
840 			} else {
841 				rt->rt_flags |= RTF_REJECT;
842 				rt->rt_expire += arpt_down;
843 				la->la_asked = 0;
844 			}
845 		}
846 	}
847 	return 0;
848 }
849 
850 /*
851  * Common length and type checks are done here,
852  * then the protocol-specific routine is called.
853  */
854 void
855 arpintr(void)
856 {
857 	struct mbuf *m;
858 	struct arphdr *ar;
859 	int s;
860 	int arplen;
861 
862 	mutex_enter(softnet_lock);
863 	KERNEL_LOCK(1, NULL);
864 	while (arpintrq.ifq_head) {
865 		s = splnet();
866 		IF_DEQUEUE(&arpintrq, m);
867 		splx(s);
868 		if (m == 0 || (m->m_flags & M_PKTHDR) == 0)
869 			panic("arpintr");
870 
871 		MCLAIM(m, &arpdomain.dom_mowner);
872 		ARP_STATINC(ARP_STAT_RCVTOTAL);
873 
874 		/*
875 		 * First, make sure we have at least struct arphdr.
876 		 */
877 		if (m->m_len < sizeof(struct arphdr) ||
878 		    (ar = mtod(m, struct arphdr *)) == NULL)
879 			goto badlen;
880 
881 		switch (m->m_pkthdr.rcvif->if_type) {
882 		case IFT_IEEE1394:
883 			arplen = sizeof(struct arphdr) +
884 			    ar->ar_hln + 2 * ar->ar_pln;
885 			break;
886 		default:
887 			arplen = sizeof(struct arphdr) +
888 			    2 * ar->ar_hln + 2 * ar->ar_pln;
889 			break;
890 		}
891 
892 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
893 		    m->m_len >= arplen)
894 			switch (ntohs(ar->ar_pro)) {
895 			case ETHERTYPE_IP:
896 			case ETHERTYPE_IPTRAILERS:
897 				in_arpinput(m);
898 				continue;
899 			default:
900 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
901 			}
902 		else {
903 badlen:
904 			ARP_STATINC(ARP_STAT_RCVBADLEN);
905 		}
906 		m_freem(m);
907 	}
908 	KERNEL_UNLOCK_ONE(NULL);
909 	mutex_exit(softnet_lock);
910 }
911 
912 /*
913  * ARP for Internet protocols on 10 Mb/s Ethernet.
914  * Algorithm is that given in RFC 826.
915  * In addition, a sanity check is performed on the sender
916  * protocol address, to catch impersonators.
917  * We no longer handle negotiations for use of trailer protocol:
918  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
919  * along with IP replies if we wanted trailers sent to us,
920  * and also sent them in response to IP replies.
921  * This allowed either end to announce the desire to receive
922  * trailer packets.
923  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
924  * but formerly didn't normally send requests.
925  */
926 static void
927 in_arpinput(struct mbuf *m)
928 {
929 	struct arphdr *ah;
930 	struct ifnet *ifp = m->m_pkthdr.rcvif;
931 	struct llinfo_arp *la = NULL;
932 	struct rtentry  *rt;
933 	struct in_ifaddr *ia;
934 #if NBRIDGE > 0
935 	struct in_ifaddr *bridge_ia = NULL;
936 #endif
937 #if NCARP > 0
938 	u_int32_t count = 0, index = 0;
939 #endif
940 	struct sockaddr_dl *sdl;
941 	struct sockaddr sa;
942 	struct in_addr isaddr, itaddr, myaddr;
943 	int op;
944 	struct mbuf *mold;
945 	void *tha;
946 	int s;
947 	uint64_t *arps;
948 
949 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
950 		goto out;
951 	ah = mtod(m, struct arphdr *);
952 	op = ntohs(ah->ar_op);
953 
954 	/*
955 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
956 	 * ar_tpa().
957 	 */
958 	switch (ifp->if_type) {
959 	case IFT_IEEE1394:
960 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
961 			;
962 		else {
963 			/* XXX this is to make sure we compute ar_tha right */
964 			/* XXX check ar_hrd more strictly? */
965 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
966 		}
967 		break;
968 	default:
969 		/* XXX check ar_hrd? */
970 		break;
971 	}
972 
973 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
974 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
975 
976 	if (m->m_flags & (M_BCAST|M_MCAST))
977 		ARP_STATINC(ARP_STAT_RCVMCAST);
978 
979 	/*
980 	 * If the target IP address is zero, ignore the packet.
981 	 * This prevents the code below from tring to answer
982 	 * when we are using IP address zero (booting).
983 	 */
984 	if (in_nullhost(itaddr)) {
985 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
986 		goto out;
987 	}
988 
989 
990 	/*
991 	 * Search for a matching interface address
992 	 * or any address on the interface to use
993 	 * as a dummy address in the rest of this function
994 	 */
995 
996 	INADDR_TO_IA(itaddr, ia);
997 	while (ia != NULL) {
998 #if NCARP > 0
999 		if (ia->ia_ifp->if_type == IFT_CARP &&
1000 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1001 		    (IFF_UP|IFF_RUNNING))) {
1002 			index++;
1003 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
1004 			    carp_iamatch(ia, ar_sha(ah),
1005 			    &count, index)) {
1006 				break;
1007 				}
1008 		} else
1009 #endif
1010 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
1011 				break;
1012 #if NBRIDGE > 0
1013 		/*
1014 		 * If the interface we received the packet on
1015 		 * is part of a bridge, check to see if we need
1016 		 * to "bridge" the packet to ourselves at this
1017 		 * layer.  Note we still prefer a perfect match,
1018 		 * but allow this weaker match if necessary.
1019 		 */
1020 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
1021 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
1022 			bridge_ia = ia;
1023 #endif /* NBRIDGE > 0 */
1024 
1025 		NEXT_IA_WITH_SAME_ADDR(ia);
1026 	}
1027 
1028 #if NBRIDGE > 0
1029 	if (ia == NULL && bridge_ia != NULL) {
1030 		ia = bridge_ia;
1031 		ifp = bridge_ia->ia_ifp;
1032 	}
1033 #endif
1034 
1035 	if (ia == NULL) {
1036 		INADDR_TO_IA(isaddr, ia);
1037 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
1038 			NEXT_IA_WITH_SAME_ADDR(ia);
1039 
1040 		if (ia == NULL) {
1041 			IFP_TO_IA(ifp, ia);
1042 			if (ia == NULL) {
1043 				ARP_STATINC(ARP_STAT_RCVNOINT);
1044 				goto out;
1045 			}
1046 		}
1047 	}
1048 
1049 	myaddr = ia->ia_addr.sin_addr;
1050 
1051 	/* XXX checks for bridge case? */
1052 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
1053 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
1054 		goto out;	/* it's from me, ignore it. */
1055 	}
1056 
1057 	/* XXX checks for bridge case? */
1058 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1059 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
1060 		log(LOG_ERR,
1061 		    "%s: arp: link address is broadcast for IP address %s!\n",
1062 		    ifp->if_xname, in_fmtaddr(isaddr));
1063 		goto out;
1064 	}
1065 
1066 	/*
1067 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
1068 	 */
1069 	if (in_nullhost(isaddr)) {
1070 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
1071 		goto reply;
1072 	}
1073 
1074 	if (in_hosteq(isaddr, myaddr)) {
1075 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
1076 		log(LOG_ERR,
1077 		   "duplicate IP address %s sent from link address %s\n",
1078 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
1079 		itaddr = myaddr;
1080 		goto reply;
1081 	}
1082 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
1083 	if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
1084 		if (sdl->sdl_alen &&
1085 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
1086 			if (rt->rt_flags & RTF_STATIC) {
1087 				ARP_STATINC(ARP_STAT_RCVOVERPERM);
1088 				log(LOG_INFO,
1089 				    "%s tried to overwrite permanent arp info"
1090 				    " for %s\n",
1091 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1092 				    in_fmtaddr(isaddr));
1093 				goto out;
1094 			} else if (rt->rt_ifp != ifp) {
1095 				ARP_STATINC(ARP_STAT_RCVOVERINT);
1096 				log(LOG_INFO,
1097 				    "%s on %s tried to overwrite "
1098 				    "arp info for %s on %s\n",
1099 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
1100 				    ifp->if_xname, in_fmtaddr(isaddr),
1101 				    rt->rt_ifp->if_xname);
1102 				    goto out;
1103 			} else {
1104 				ARP_STATINC(ARP_STAT_RCVOVER);
1105 				log(LOG_INFO,
1106 				    "arp info overwritten for %s by %s\n",
1107 				    in_fmtaddr(isaddr),
1108 				    lla_snprintf(ar_sha(ah), ah->ar_hln));
1109 			}
1110 		}
1111 		/*
1112 		 * sanity check for the address length.
1113 		 * XXX this does not work for protocols with variable address
1114 		 * length. -is
1115 		 */
1116 		if (sdl->sdl_alen &&
1117 		    sdl->sdl_alen != ah->ar_hln) {
1118 			ARP_STATINC(ARP_STAT_RCVLENCHG);
1119 			log(LOG_WARNING,
1120 			    "arp from %s: new addr len %d, was %d\n",
1121 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
1122 		}
1123 		if (ifp->if_addrlen != ah->ar_hln) {
1124 			ARP_STATINC(ARP_STAT_RCVBADLEN);
1125 			log(LOG_WARNING,
1126 			    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
1127 			    in_fmtaddr(isaddr), ah->ar_hln,
1128 			    ifp->if_addrlen);
1129 			goto reply;
1130 		}
1131 #if NTOKEN > 0
1132 		/*
1133 		 * XXX uses m_data and assumes the complete answer including
1134 		 * XXX token-ring headers is in the same buf
1135 		 */
1136 		if (ifp->if_type == IFT_ISO88025) {
1137 			struct token_header *trh;
1138 
1139 			trh = (struct token_header *)M_TRHSTART(m);
1140 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
1141 				struct token_rif	*rif;
1142 				size_t	riflen;
1143 
1144 				rif = TOKEN_RIF(trh);
1145 				riflen = (ntohs(rif->tr_rcf) &
1146 				    TOKEN_RCF_LEN_MASK) >> 8;
1147 
1148 				if (riflen > 2 &&
1149 				    riflen < sizeof(struct token_rif) &&
1150 				    (riflen & 1) == 0) {
1151 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
1152 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
1153 					memcpy(TOKEN_RIF(la), rif, riflen);
1154 				}
1155 			}
1156 		}
1157 #endif /* NTOKEN > 0 */
1158 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
1159 		    ah->ar_hln);
1160 		if (rt->rt_expire)
1161 			rt->rt_expire = time_second + arpt_keep;
1162 		rt->rt_flags &= ~RTF_REJECT;
1163 		la->la_asked = 0;
1164 
1165 		s = splnet();
1166 		mold = la->la_hold;
1167 		la->la_hold = 0;
1168 		splx(s);
1169 
1170 		if (mold) {
1171 			ARP_STATINC(ARP_STAT_DFRSENT);
1172 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
1173 		}
1174 	}
1175 reply:
1176 	if (op != ARPOP_REQUEST) {
1177 		if (op == ARPOP_REPLY)
1178 			ARP_STATINC(ARP_STAT_RCVREPLY);
1179 	out:
1180 		m_freem(m);
1181 		return;
1182 	}
1183 	ARP_STATINC(ARP_STAT_RCVREQUEST);
1184 	if (in_hosteq(itaddr, myaddr)) {
1185 		/* I am the target */
1186 		tha = ar_tha(ah);
1187 		if (tha)
1188 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1189 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1190 	} else {
1191 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
1192 		if (la == NULL)
1193 			goto out;
1194 		rt = la->la_rt;
1195 		if (rt->rt_ifp->if_type == IFT_CARP &&
1196 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
1197 			goto out;
1198 		tha = ar_tha(ah);
1199 		if (tha)
1200 			memcpy(tha, ar_sha(ah), ah->ar_hln);
1201 		sdl = satosdl(rt->rt_gateway);
1202 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
1203 	}
1204 
1205 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1206 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1207 	ah->ar_op = htons(ARPOP_REPLY);
1208 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1209 	switch (ifp->if_type) {
1210 	case IFT_IEEE1394:
1211 		/*
1212 		 * ieee1394 arp reply is broadcast
1213 		 */
1214 		m->m_flags &= ~M_MCAST;
1215 		m->m_flags |= M_BCAST;
1216 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
1217 		break;
1218 
1219 	default:
1220 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
1221 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1222 		break;
1223 	}
1224 	m->m_pkthdr.len = m->m_len;
1225 	sa.sa_family = AF_ARP;
1226 	sa.sa_len = 2;
1227 	arps = ARP_STAT_GETREF();
1228 	arps[ARP_STAT_SNDTOTAL]++;
1229 	arps[ARP_STAT_SNDREPLY]++;
1230 	ARP_STAT_PUTREF();
1231 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
1232 	return;
1233 }
1234 
1235 /*
1236  * Free an arp entry.
1237  */
1238 static void arptfree(struct llinfo_arp *la)
1239 {
1240 	struct rtentry *rt = la->la_rt;
1241 	struct sockaddr_dl *sdl;
1242 
1243 	ARP_LOCK_CHECK();
1244 
1245 	if (rt == NULL)
1246 		panic("arptfree");
1247 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
1248 	    sdl->sdl_family == AF_LINK) {
1249 		sdl->sdl_alen = 0;
1250 		la->la_asked = 0;
1251 		rt->rt_flags &= ~RTF_REJECT;
1252 		return;
1253 	}
1254 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1255 }
1256 
1257 static struct llinfo_arp *
1258 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
1259 {
1260 	return arplookup1(m, addr, create, proxy, NULL);
1261 }
1262 
1263 /*
1264  * Lookup or enter a new address in arptab.
1265  */
1266 static struct llinfo_arp *
1267 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
1268     struct rtentry *rt0)
1269 {
1270 	struct arphdr *ah;
1271 	struct ifnet *ifp = m->m_pkthdr.rcvif;
1272 	struct rtentry *rt;
1273 	struct sockaddr_inarp sin;
1274 	const char *why = NULL;
1275 
1276 	ah = mtod(m, struct arphdr *);
1277 	if (rt0 == NULL) {
1278 		memset(&sin, 0, sizeof(sin));
1279 		sin.sin_len = sizeof(sin);
1280 		sin.sin_family = AF_INET;
1281 		sin.sin_addr = *addr;
1282 		sin.sin_other = proxy ? SIN_PROXY : 0;
1283 		rt = rtalloc1(sintosa(&sin), create);
1284 		if (rt == NULL)
1285 			return NULL;
1286 		rt->rt_refcnt--;
1287 	} else
1288 		rt = rt0;
1289 
1290 #define	IS_LLINFO(__rt)							  \
1291 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
1292 	 (__rt)->rt_gateway->sa_family == AF_LINK)
1293 
1294 
1295 	if (IS_LLINFO(rt))
1296 		return (struct llinfo_arp *)rt->rt_llinfo;
1297 
1298 	if (create) {
1299 		if (rt->rt_flags & RTF_GATEWAY)
1300 			why = "host is not on local network";
1301 		else if ((rt->rt_flags & RTF_LLINFO) == 0) {
1302 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
1303 			why = "could not allocate llinfo";
1304 		} else
1305 			why = "gateway route is not ours";
1306 		log(LOG_DEBUG, "arplookup: unable to enter address"
1307 		    " for %s@%s on %s (%s)\n",
1308 		    in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln),
1309 		    (ifp) ? ifp->if_xname : "null", why);
1310 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
1311 			rtrequest(RTM_DELETE, rt_getkey(rt),
1312 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1313 		}
1314 	}
1315 	return NULL;
1316 }
1317 
1318 int
1319 arpioctl(u_long cmd, void *data)
1320 {
1321 
1322 	return EOPNOTSUPP;
1323 }
1324 
1325 void
1326 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1327 {
1328 	struct in_addr *ip;
1329 
1330 	/*
1331 	 * Warn the user if another station has this IP address,
1332 	 * but only if the interface IP address is not zero.
1333 	 */
1334 	ip = &IA_SIN(ifa)->sin_addr;
1335 	if (!in_nullhost(*ip))
1336 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
1337 
1338 	ifa->ifa_rtrequest = arp_rtrequest;
1339 	ifa->ifa_flags |= RTF_CLONING;
1340 }
1341 
1342 /*
1343  * Called from 10 Mb/s Ethernet interrupt handlers
1344  * when ether packet type ETHERTYPE_REVARP
1345  * is received.  Common length and type checks are done here,
1346  * then the protocol-specific routine is called.
1347  */
1348 void
1349 revarpinput(struct mbuf *m)
1350 {
1351 	struct arphdr *ar;
1352 
1353 	if (m->m_len < sizeof(struct arphdr))
1354 		goto out;
1355 	ar = mtod(m, struct arphdr *);
1356 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
1357 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
1358 		goto out;
1359 #endif
1360 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
1361 		goto out;
1362 	switch (ntohs(ar->ar_pro)) {
1363 	case ETHERTYPE_IP:
1364 	case ETHERTYPE_IPTRAILERS:
1365 		in_revarpinput(m);
1366 		return;
1367 
1368 	default:
1369 		break;
1370 	}
1371 out:
1372 	m_freem(m);
1373 }
1374 
1375 /*
1376  * RARP for Internet protocols on 10 Mb/s Ethernet.
1377  * Algorithm is that given in RFC 903.
1378  * We are only using for bootstrap purposes to get an ip address for one of
1379  * our interfaces.  Thus we support no user-interface.
1380  *
1381  * Since the contents of the RARP reply are specific to the interface that
1382  * sent the request, this code must ensure that they are properly associated.
1383  *
1384  * Note: also supports ARP via RARP packets, per the RFC.
1385  */
1386 void
1387 in_revarpinput(struct mbuf *m)
1388 {
1389 	struct ifnet *ifp;
1390 	struct arphdr *ah;
1391 	void *tha;
1392 	int op;
1393 
1394 	ah = mtod(m, struct arphdr *);
1395 	op = ntohs(ah->ar_op);
1396 
1397 	switch (m->m_pkthdr.rcvif->if_type) {
1398 	case IFT_IEEE1394:
1399 		/* ARP without target hardware address is not supported */
1400 		goto out;
1401 	default:
1402 		break;
1403 	}
1404 
1405 	switch (op) {
1406 	case ARPOP_REQUEST:
1407 	case ARPOP_REPLY:	/* per RFC */
1408 		in_arpinput(m);
1409 		return;
1410 	case ARPOP_REVREPLY:
1411 		break;
1412 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
1413 	default:
1414 		goto out;
1415 	}
1416 	if (!revarp_in_progress)
1417 		goto out;
1418 	ifp = m->m_pkthdr.rcvif;
1419 	if (ifp != myip_ifp) /* !same interface */
1420 		goto out;
1421 	if (myip_initialized)
1422 		goto wake;
1423 	tha = ar_tha(ah);
1424 	if (tha == NULL)
1425 		goto out;
1426 	if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
1427 		goto out;
1428 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
1429 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
1430 	myip_initialized = 1;
1431 wake:	/* Do wakeup every time in case it was missed. */
1432 	wakeup((void *)&myip);
1433 
1434 out:
1435 	m_freem(m);
1436 }
1437 
1438 /*
1439  * Send a RARP request for the ip address of the specified interface.
1440  * The request should be RFC 903-compliant.
1441  */
1442 void
1443 revarprequest(struct ifnet *ifp)
1444 {
1445 	struct sockaddr sa;
1446 	struct mbuf *m;
1447 	struct arphdr *ah;
1448 	void *tha;
1449 
1450 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
1451 		return;
1452 	MCLAIM(m, &arpdomain.dom_mowner);
1453 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
1454 	    2*ifp->if_addrlen;
1455 	m->m_pkthdr.len = m->m_len;
1456 	MH_ALIGN(m, m->m_len);
1457 	ah = mtod(m, struct arphdr *);
1458 	memset(ah, 0, m->m_len);
1459 	ah->ar_pro = htons(ETHERTYPE_IP);
1460 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
1461 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
1462 	ah->ar_op = htons(ARPOP_REVREQUEST);
1463 
1464 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
1465 	tha = ar_tha(ah);
1466 	if (tha == NULL)
1467 		return;
1468 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
1469 
1470 	sa.sa_family = AF_ARP;
1471 	sa.sa_len = 2;
1472 	m->m_flags |= M_BCAST;
1473 	(*ifp->if_output)(ifp, m, &sa, NULL);
1474 
1475 }
1476 
1477 /*
1478  * RARP for the ip address of the specified interface, but also
1479  * save the ip address of the server that sent the answer.
1480  * Timeout if no response is received.
1481  */
1482 int
1483 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
1484     struct in_addr *clnt_in)
1485 {
1486 	int result, count = 20;
1487 
1488 	myip_initialized = 0;
1489 	myip_ifp = ifp;
1490 
1491 	revarp_in_progress = 1;
1492 	while (count--) {
1493 		revarprequest(ifp);
1494 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
1495 		if (result != EWOULDBLOCK)
1496 			break;
1497 	}
1498 	revarp_in_progress = 0;
1499 
1500 	if (!myip_initialized)
1501 		return ENETUNREACH;
1502 
1503 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
1504 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
1505 	return 0;
1506 }
1507 
1508 
1509 
1510 #ifdef DDB
1511 
1512 #include <machine/db_machdep.h>
1513 #include <ddb/db_interface.h>
1514 #include <ddb/db_output.h>
1515 
1516 static void
1517 db_print_sa(const struct sockaddr *sa)
1518 {
1519 	int len;
1520 	const u_char *p;
1521 
1522 	if (sa == NULL) {
1523 		db_printf("[NULL]");
1524 		return;
1525 	}
1526 
1527 	p = (const u_char *)sa;
1528 	len = sa->sa_len;
1529 	db_printf("[");
1530 	while (len > 0) {
1531 		db_printf("%d", *p);
1532 		p++; len--;
1533 		if (len) db_printf(",");
1534 	}
1535 	db_printf("]\n");
1536 }
1537 
1538 static void
1539 db_print_ifa(struct ifaddr *ifa)
1540 {
1541 	if (ifa == NULL)
1542 		return;
1543 	db_printf("  ifa_addr=");
1544 	db_print_sa(ifa->ifa_addr);
1545 	db_printf("  ifa_dsta=");
1546 	db_print_sa(ifa->ifa_dstaddr);
1547 	db_printf("  ifa_mask=");
1548 	db_print_sa(ifa->ifa_netmask);
1549 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
1550 			  ifa->ifa_flags,
1551 			  ifa->ifa_refcnt,
1552 			  ifa->ifa_metric);
1553 }
1554 
1555 static void
1556 db_print_llinfo(void *li)
1557 {
1558 	struct llinfo_arp *la;
1559 
1560 	if (li == NULL)
1561 		return;
1562 	la = (struct llinfo_arp *)li;
1563 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
1564 			  la->la_rt, la->la_hold, la->la_asked);
1565 }
1566 
1567 /*
1568  * Function to pass to rt_walktree().
1569  * Return non-zero error to abort walk.
1570  */
1571 static int
1572 db_show_rtentry(struct rtentry *rt, void *w)
1573 {
1574 	db_printf("rtentry=%p", rt);
1575 
1576 	db_printf(" flags=0x%x refcnt=%d use=%"PRId64" expire=%"PRId64"\n",
1577 			  rt->rt_flags, rt->rt_refcnt,
1578 			  rt->rt_use, (uint64_t)rt->rt_expire);
1579 
1580 	db_printf(" key="); db_print_sa(rt_getkey(rt));
1581 	db_printf(" mask="); db_print_sa(rt_mask(rt));
1582 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
1583 
1584 	db_printf(" ifp=%p ", rt->rt_ifp);
1585 	if (rt->rt_ifp)
1586 		db_printf("(%s)", rt->rt_ifp->if_xname);
1587 	else
1588 		db_printf("(NULL)");
1589 
1590 	db_printf(" ifa=%p\n", rt->rt_ifa);
1591 	db_print_ifa(rt->rt_ifa);
1592 
1593 	db_printf(" gwroute=%p llinfo=%p\n",
1594 			  rt->rt_gwroute, rt->rt_llinfo);
1595 	db_print_llinfo(rt->rt_llinfo);
1596 
1597 	return 0;
1598 }
1599 
1600 /*
1601  * Function to print all the route trees.
1602  * Use this from ddb:  "show arptab"
1603  */
1604 void
1605 db_show_arptab(db_expr_t addr, bool have_addr,
1606     db_expr_t count, const char *modif)
1607 {
1608 	rt_walktree(AF_INET, db_show_rtentry, NULL);
1609 }
1610 #endif
1611 
1612 static int
1613 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
1614 {
1615 
1616 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
1617 }
1618 
1619 static void
1620 sysctl_net_inet_arp_setup(struct sysctllog **clog)
1621 {
1622 	const struct sysctlnode *node;
1623 
1624 	sysctl_createv(clog, 0, NULL, NULL,
1625 			CTLFLAG_PERMANENT,
1626 			CTLTYPE_NODE, "net", NULL,
1627 			NULL, 0, NULL, 0,
1628 			CTL_NET, CTL_EOL);
1629 	sysctl_createv(clog, 0, NULL, NULL,
1630 			CTLFLAG_PERMANENT,
1631 			CTLTYPE_NODE, "inet", NULL,
1632 			NULL, 0, NULL, 0,
1633 			CTL_NET, PF_INET, CTL_EOL);
1634 	sysctl_createv(clog, 0, NULL, &node,
1635 			CTLFLAG_PERMANENT,
1636 			CTLTYPE_NODE, "arp",
1637 			SYSCTL_DESCR("Address Resolution Protocol"),
1638 			NULL, 0, NULL, 0,
1639 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
1640 
1641 	sysctl_createv(clog, 0, NULL, NULL,
1642 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1643 			CTLTYPE_INT, "prune",
1644 			SYSCTL_DESCR("ARP cache pruning interval"),
1645 			NULL, 0, &arpt_prune, 0,
1646 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1647 
1648 	sysctl_createv(clog, 0, NULL, NULL,
1649 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1650 			CTLTYPE_INT, "keep",
1651 			SYSCTL_DESCR("Valid ARP entry lifetime"),
1652 			NULL, 0, &arpt_keep, 0,
1653 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1654 
1655 	sysctl_createv(clog, 0, NULL, NULL,
1656 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1657 			CTLTYPE_INT, "down",
1658 			SYSCTL_DESCR("Failed ARP entry lifetime"),
1659 			NULL, 0, &arpt_down, 0,
1660 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1661 
1662 	sysctl_createv(clog, 0, NULL, NULL,
1663 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1664 			CTLTYPE_INT, "refresh",
1665 			SYSCTL_DESCR("ARP entry refresh interval"),
1666 			NULL, 0, &arpt_refresh, 0,
1667 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1668 
1669 	sysctl_createv(clog, 0, NULL, NULL,
1670 			CTLFLAG_PERMANENT,
1671 			CTLTYPE_STRUCT, "stats",
1672 			SYSCTL_DESCR("ARP statistics"),
1673 			sysctl_net_inet_arp_stats, 0, NULL, 0,
1674 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
1675 }
1676 
1677 #endif /* INET */
1678