xref: /openbsd/sys/netinet6/nd6.c (revision 9593dc34)
1 /*	$OpenBSD: nd6.c,v 1.283 2024/09/04 07:54:52 mglocker Exp $	*/
2 /*	$KAME: nd6.c,v 1.280 2002/06/08 19:52:07 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/timeout.h>
36 #include <sys/malloc.h>
37 #include <sys/mbuf.h>
38 #include <sys/socket.h>
39 #include <sys/sockio.h>
40 #include <sys/time.h>
41 #include <sys/kernel.h>
42 #include <sys/pool.h>
43 #include <sys/errno.h>
44 #include <sys/ioctl.h>
45 #include <sys/syslog.h>
46 #include <sys/queue.h>
47 #include <sys/stdint.h>
48 #include <sys/task.h>
49 
50 #include <net/if.h>
51 #include <net/if_dl.h>
52 #include <net/if_types.h>
53 #include <net/route.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/if_ether.h>
57 #include <netinet/ip_ipsp.h>
58 
59 #include <netinet6/in6_var.h>
60 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
62 #include <netinet6/nd6.h>
63 #include <netinet/icmp6.h>
64 
65 /*
66  * Locks used to protect struct members in this file:
67  *	a	atomic operations
68  *	I	immutable after creation
69  *	K	kernel lock
70  *	m	nd6 mutex, needed when net lock is shared
71  *	N	net lock
72  */
73 
74 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
75 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
76 
77 /* timer values */
78 int	nd6_timer_next	= -1;	/* at which uptime nd6_timer runs */
79 time_t	nd6_expire_next	= -1;	/* at which uptime nd6_expire runs */
80 int	nd6_delay	= 5;	/* delay first probe time 5 second */
81 int	nd6_umaxtries	= 3;	/* maximum unicast query */
82 int	nd6_mmaxtries	= 3;	/* maximum multicast query */
83 int	nd6_gctimer	= (60 * 60 * 24); /* 1 day: garbage collection timer */
84 
85 /* preventing too many loops in ND option parsing */
86 int nd6_maxndopt = 10;	/* max # of ND options allowed */
87 
88 int nd6_maxnudhint = 0;	/* max # of subsequent upper layer hints */
89 
90 #ifdef ND6_DEBUG
91 int nd6_debug = 1;
92 #else
93 int nd6_debug = 0;
94 #endif
95 
96 /* llinfo_nd6 live time, rt_llinfo and RTF_LLINFO are protected by nd6_mtx */
97 struct mutex nd6_mtx = MUTEX_INITIALIZER(IPL_SOFTNET);
98 
99 TAILQ_HEAD(llinfo_nd6_head, llinfo_nd6) nd6_list =
100     TAILQ_HEAD_INITIALIZER(nd6_list);	/* [mN] list of llinfo_nd6 structures */
101 struct	pool nd6_pool;		/* [I] pool for llinfo_nd6 structures */
102 int	nd6_inuse;		/* [m] limit neighbor discovery routes */
103 unsigned int	ln_hold_total;	/* [a] packets currently in the nd6 queue */
104 
105 void nd6_timer(void *);
106 void nd6_slowtimo(void *);
107 void nd6_expire(void *);
108 void nd6_expire_timer(void *);
109 void nd6_invalidate(struct rtentry *);
110 void nd6_free(struct rtentry *, int);
111 int nd6_llinfo_timer(struct rtentry *, int);
112 
113 struct timeout nd6_timer_to;
114 struct timeout nd6_slowtimo_ch;
115 struct timeout nd6_expire_timeout;
116 struct task nd6_expire_task;
117 
118 void
nd6_init(void)119 nd6_init(void)
120 {
121 	pool_init(&nd6_pool, sizeof(struct llinfo_nd6), 0,
122 	    IPL_SOFTNET, 0, "nd6", NULL);
123 
124 	task_set(&nd6_expire_task, nd6_expire, NULL);
125 
126 	/* start timer */
127 	timeout_set_proc(&nd6_timer_to, nd6_timer, NULL);
128 	timeout_set_proc(&nd6_slowtimo_ch, nd6_slowtimo, NULL);
129 	timeout_add_sec(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL);
130 	timeout_set(&nd6_expire_timeout, nd6_expire_timer, NULL);
131 }
132 
133 void
nd6_ifattach(struct ifnet * ifp)134 nd6_ifattach(struct ifnet *ifp)
135 {
136 	struct nd_ifinfo *nd;
137 
138 	nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
139 
140 	nd->reachable = ND_COMPUTE_RTIME(REACHABLE_TIME);
141 
142 	ifp->if_nd = nd;
143 }
144 
145 void
nd6_ifdetach(struct ifnet * ifp)146 nd6_ifdetach(struct ifnet *ifp)
147 {
148 	struct nd_ifinfo *nd = ifp->if_nd;
149 
150 	free(nd, M_IP6NDP, sizeof(*nd));
151 }
152 
153 /*
154  * Parse multiple ND options.
155  * This function is much easier to use, for ND routines that do not need
156  * multiple options of the same type.
157  */
158 int
nd6_options(void * opt,int icmp6len,struct nd_opts * ndopts)159 nd6_options(void *opt, int icmp6len, struct nd_opts *ndopts)
160 {
161 	struct nd_opt_hdr *nd_opt, *next_opt, *last_opt;
162 	int i = 0;
163 
164 	bzero(ndopts, sizeof(*ndopts));
165 
166 	if (icmp6len == 0)
167 		return 0;
168 
169 	next_opt = opt;
170 	last_opt = (struct nd_opt_hdr *)((u_char *)opt + icmp6len);
171 
172 	while (next_opt != NULL) {
173 		int olen;
174 
175 		nd_opt = next_opt;
176 
177 		/* make sure nd_opt_len is inside the buffer */
178 		if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)last_opt)
179 			goto invalid;
180 
181 		/* every option must have a length greater than zero */
182 		olen = nd_opt->nd_opt_len << 3;
183 		if (olen == 0)
184 			goto invalid;
185 
186 		next_opt = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
187 		if (next_opt > last_opt) {
188 			/* option overruns the end of buffer */
189 			goto invalid;
190 		} else if (next_opt == last_opt) {
191 			/* reached the end of options chain */
192 			next_opt = NULL;
193 		}
194 
195 		switch (nd_opt->nd_opt_type) {
196 		case ND_OPT_SOURCE_LINKADDR:
197 			if (ndopts->nd_opts_src_lladdr != NULL)
198 				nd6log((LOG_INFO, "duplicated ND6 option found "
199 				    "(type=%d)\n", nd_opt->nd_opt_type));
200 			else
201 				ndopts->nd_opts_src_lladdr = nd_opt;
202 			break;
203 		case ND_OPT_TARGET_LINKADDR:
204 			if (ndopts->nd_opts_tgt_lladdr != NULL)
205 				nd6log((LOG_INFO, "duplicated ND6 option found "
206 				    "(type=%d)\n", nd_opt->nd_opt_type));
207 			else
208 				ndopts->nd_opts_tgt_lladdr = nd_opt;
209 			break;
210 		case ND_OPT_MTU:
211 		case ND_OPT_REDIRECTED_HEADER:
212 		case ND_OPT_PREFIX_INFORMATION:
213 		case ND_OPT_DNSSL:
214 		case ND_OPT_RDNSS:
215 			/* Don't warn, not used by kernel */
216 			break;
217 		default:
218 			/*
219 			 * Unknown options must be silently ignored,
220 			 * to accommodate future extension to the protocol.
221 			 */
222 			nd6log((LOG_DEBUG,
223 			    "nd6_options: unsupported option %d - "
224 			    "option ignored\n", nd_opt->nd_opt_type));
225 			break;
226 		}
227 
228 		i++;
229 		if (i > nd6_maxndopt) {
230 			icmp6stat_inc(icp6s_nd_toomanyopt);
231 			nd6log((LOG_INFO, "too many loop in nd opt\n"));
232 			break;
233 		}
234 	}
235 
236 	return 0;
237 
238 invalid:
239 	bzero(ndopts, sizeof(*ndopts));
240 	icmp6stat_inc(icp6s_nd_badopt);
241 	return -1;
242 }
243 
244 /*
245  * ND6 timer routine to handle ND6 entries
246  */
247 void
nd6_llinfo_settimer(const struct llinfo_nd6 * ln,unsigned int secs)248 nd6_llinfo_settimer(const struct llinfo_nd6 *ln, unsigned int secs)
249 {
250 	time_t expire = getuptime() + secs;
251 
252 	NET_ASSERT_LOCKED();
253 	KASSERT(!ISSET(ln->ln_rt->rt_flags, RTF_LOCAL));
254 
255 	ln->ln_rt->rt_expire = expire;
256 	if (!timeout_pending(&nd6_timer_to) || expire < nd6_timer_next) {
257 		nd6_timer_next = expire;
258 		timeout_add_sec(&nd6_timer_to, secs);
259 	}
260 }
261 
262 void
nd6_timer(void * unused)263 nd6_timer(void *unused)
264 {
265 	struct llinfo_nd6 *ln, *nln;
266 	time_t uptime, expire;
267 	int i_am_router = (atomic_load_int(&ip6_forwarding) != 0);
268 	int secs;
269 
270 	NET_LOCK();
271 
272 	uptime = getuptime();
273 	expire = uptime + nd6_gctimer;
274 
275 	/* Net lock is exclusive, no nd6 mutex needed for nd6_list here. */
276 	TAILQ_FOREACH_SAFE(ln, &nd6_list, ln_list, nln) {
277 		struct rtentry *rt = ln->ln_rt;
278 
279 		if (rt->rt_expire && rt->rt_expire <= uptime)
280 			if (nd6_llinfo_timer(rt, i_am_router))
281 				continue;
282 
283 		if (rt->rt_expire && rt->rt_expire < expire)
284 			expire = rt->rt_expire;
285 	}
286 
287 	secs = expire - uptime;
288 	if (secs < 0)
289 		secs = 0;
290 	if (!TAILQ_EMPTY(&nd6_list)) {
291 		nd6_timer_next = uptime + secs;
292 		timeout_add_sec(&nd6_timer_to, secs);
293 	}
294 
295 	NET_UNLOCK();
296 }
297 
298 /*
299  * ND timer state handling.
300  *
301  * Returns 1 if `rt' should no longer be used, 0 otherwise.
302  */
303 int
nd6_llinfo_timer(struct rtentry * rt,int i_am_router)304 nd6_llinfo_timer(struct rtentry *rt, int i_am_router)
305 {
306 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
307 	struct sockaddr_in6 *dst = satosin6(rt_key(rt));
308 	struct ifnet *ifp;
309 
310 	NET_ASSERT_LOCKED_EXCLUSIVE();
311 
312 	if ((ifp = if_get(rt->rt_ifidx)) == NULL)
313 		return 1;
314 
315 	switch (ln->ln_state) {
316 	case ND6_LLINFO_INCOMPLETE:
317 		if (ln->ln_asked < nd6_mmaxtries) {
318 			ln->ln_asked++;
319 			nd6_llinfo_settimer(ln, RETRANS_TIMER / 1000);
320 			nd6_ns_output(ifp, NULL, &dst->sin6_addr,
321 			    &ln->ln_saddr6, 0);
322 		} else {
323 			struct mbuf_list ml;
324 			struct mbuf *m;
325 			unsigned int len;
326 
327 			mq_delist(&ln->ln_mq, &ml);
328 			len = ml_len(&ml);
329 			while ((m = ml_dequeue(&ml)) != NULL) {
330 				/*
331 				 * Fake rcvif to make the ICMP error
332 				 * more helpful in diagnosing for the
333 				 * receiver.
334 				 * XXX: should we consider older rcvif?
335 				 */
336 				m->m_pkthdr.ph_ifidx = rt->rt_ifidx;
337 
338 				icmp6_error(m, ICMP6_DST_UNREACH,
339 				    ICMP6_DST_UNREACH_ADDR, 0);
340 			}
341 
342 			/* XXXSMP we also discard if other CPU enqueues */
343 			if (mq_len(&ln->ln_mq) > 0) {
344 				/* mbuf is back in queue. Discard. */
345 				atomic_sub_int(&ln_hold_total,
346 				    len + mq_purge(&ln->ln_mq));
347 			} else
348 				atomic_sub_int(&ln_hold_total, len);
349 
350 			nd6_free(rt, i_am_router);
351 			ln = NULL;
352 		}
353 		break;
354 
355 	case ND6_LLINFO_REACHABLE:
356 		if (!ND6_LLINFO_PERMANENT(ln)) {
357 			ln->ln_state = ND6_LLINFO_STALE;
358 			nd6_llinfo_settimer(ln, nd6_gctimer);
359 		}
360 		break;
361 
362 	case ND6_LLINFO_STALE:
363 	case ND6_LLINFO_PURGE:
364 		/* Garbage Collection(RFC 2461 5.3) */
365 		if (!ND6_LLINFO_PERMANENT(ln)) {
366 			nd6_free(rt, i_am_router);
367 			ln = NULL;
368 		}
369 		break;
370 
371 	case ND6_LLINFO_DELAY:
372 		/* We need NUD */
373 		ln->ln_asked = 1;
374 		ln->ln_state = ND6_LLINFO_PROBE;
375 		nd6_llinfo_settimer(ln, RETRANS_TIMER / 1000);
376 		nd6_ns_output(ifp, &dst->sin6_addr, &dst->sin6_addr,
377 		    &ln->ln_saddr6, 0);
378 		break;
379 
380 	case ND6_LLINFO_PROBE:
381 		if (ln->ln_asked < nd6_umaxtries) {
382 			ln->ln_asked++;
383 			nd6_llinfo_settimer(ln, RETRANS_TIMER / 1000);
384 			nd6_ns_output(ifp, &dst->sin6_addr, &dst->sin6_addr,
385 			    &ln->ln_saddr6, 0);
386 		} else {
387 			nd6_free(rt, i_am_router);
388 			ln = NULL;
389 		}
390 		break;
391 	}
392 
393 	if_put(ifp);
394 
395 	return (ln == NULL);
396 }
397 
398 void
nd6_expire_timer_update(struct in6_ifaddr * ia6)399 nd6_expire_timer_update(struct in6_ifaddr *ia6)
400 {
401 	time_t expire_time = INT64_MAX;
402 
403 	if (ia6->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME)
404 		expire_time = ia6->ia6_lifetime.ia6t_expire;
405 
406 	if (!(ia6->ia6_flags & IN6_IFF_DEPRECATED) &&
407 	    ia6->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME &&
408 	    expire_time > ia6->ia6_lifetime.ia6t_preferred)
409 		expire_time = ia6->ia6_lifetime.ia6t_preferred;
410 
411 	if (expire_time == INT64_MAX)
412 		return;
413 
414 	/*
415 	 * IFA6_IS_INVALID() and IFA6_IS_DEPRECATED() check for uptime
416 	 * greater than ia6t_expire or ia6t_preferred, not greater or equal.
417 	 * Schedule timeout one second later so that either IFA6_IS_INVALID()
418 	 * or IFA6_IS_DEPRECATED() is true.
419 	 */
420 	expire_time++;
421 
422 	if (!timeout_pending(&nd6_expire_timeout) ||
423 	    nd6_expire_next > expire_time) {
424 		int secs;
425 
426 		secs = expire_time - getuptime();
427 		if (secs < 0)
428 			secs = 0;
429 
430 		timeout_add_sec(&nd6_expire_timeout, secs);
431 		nd6_expire_next = expire_time;
432 	}
433 }
434 
435 /*
436  * Expire interface addresses.
437  */
438 void
nd6_expire(void * unused)439 nd6_expire(void *unused)
440 {
441 	struct ifnet *ifp;
442 
443 	NET_LOCK();
444 
445 	TAILQ_FOREACH(ifp, &ifnetlist, if_list) {
446 		struct ifaddr *ifa, *nifa;
447 		struct in6_ifaddr *ia6;
448 
449 		TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrlist, ifa_list, nifa) {
450 			if (ifa->ifa_addr->sa_family != AF_INET6)
451 				continue;
452 			ia6 = ifatoia6(ifa);
453 			/* check address lifetime */
454 			if (IFA6_IS_INVALID(ia6)) {
455 				in6_purgeaddr(&ia6->ia_ifa);
456 			} else {
457 				if (IFA6_IS_DEPRECATED(ia6))
458 					ia6->ia6_flags |= IN6_IFF_DEPRECATED;
459 				nd6_expire_timer_update(ia6);
460 			}
461 		}
462 	}
463 
464 	NET_UNLOCK();
465 }
466 
467 void
nd6_expire_timer(void * unused)468 nd6_expire_timer(void *unused)
469 {
470 	task_add(net_tq(0), &nd6_expire_task);
471 }
472 
473 /*
474  * Nuke neighbor cache/prefix/default router management table, right before
475  * ifp goes away.
476  */
477 void
nd6_purge(struct ifnet * ifp)478 nd6_purge(struct ifnet *ifp)
479 {
480 	struct llinfo_nd6 *ln, *nln;
481 	int i_am_router = (atomic_load_int(&ip6_forwarding) != 0);
482 
483 	NET_ASSERT_LOCKED_EXCLUSIVE();
484 
485 	/*
486 	 * Nuke neighbor cache entries for the ifp.
487 	 */
488 	TAILQ_FOREACH_SAFE(ln, &nd6_list, ln_list, nln) {
489 		struct rtentry *rt;
490 		struct sockaddr_dl *sdl;
491 
492 		rt = ln->ln_rt;
493 		if (rt != NULL && rt->rt_gateway != NULL &&
494 		    rt->rt_gateway->sa_family == AF_LINK) {
495 			sdl = satosdl(rt->rt_gateway);
496 			if (sdl->sdl_index == ifp->if_index)
497 				nd6_free(rt, i_am_router);
498 		}
499 	}
500 }
501 
502 struct rtentry *
nd6_lookup(const struct in6_addr * addr6,int create,struct ifnet * ifp,u_int rtableid)503 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp,
504     u_int rtableid)
505 {
506 	struct rtentry *rt;
507 	struct sockaddr_in6 sin6;
508 	int flags;
509 
510 	bzero(&sin6, sizeof(sin6));
511 	sin6.sin6_len = sizeof(struct sockaddr_in6);
512 	sin6.sin6_family = AF_INET6;
513 	sin6.sin6_addr = *addr6;
514 	flags = (create) ? RT_RESOLVE : 0;
515 
516 	rt = rtalloc(sin6tosa(&sin6), flags, rtableid);
517 	if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) {
518 		/*
519 		 * This is the case for the default route.
520 		 * If we want to create a neighbor cache for the address, we
521 		 * should free the route for the destination and allocate an
522 		 * interface route.
523 		 */
524 		if (create) {
525 			rtfree(rt);
526 			rt = NULL;
527 		}
528 	}
529 	if (rt == NULL) {
530 		if (create && ifp) {
531 			struct rt_addrinfo info;
532 			struct llinfo_nd6 *ln;
533 			struct ifaddr *ifa;
534 			int error;
535 
536 			/*
537 			 * If no route is available and create is set,
538 			 * we allocate a host route for the destination
539 			 * and treat it like an interface route.
540 			 * This hack is necessary for a neighbor which can't
541 			 * be covered by our own prefix.
542 			 */
543 			ifa = ifaof_ifpforaddr(sin6tosa(&sin6), ifp);
544 			if (ifa == NULL)
545 				return (NULL);
546 
547 			/*
548 			 * Create a new route.  RTF_LLINFO is necessary
549 			 * to create a Neighbor Cache entry for the
550 			 * destination in nd6_rtrequest which will be
551 			 * called in rtrequest.
552 			 */
553 			bzero(&info, sizeof(info));
554 			info.rti_ifa = ifa;
555 			info.rti_flags = RTF_HOST | RTF_LLINFO;
556 			info.rti_info[RTAX_DST] = sin6tosa(&sin6);
557 			info.rti_info[RTAX_GATEWAY] = sdltosa(ifp->if_sadl);
558 			error = rtrequest(RTM_ADD, &info, RTP_CONNECTED, &rt,
559 			    rtableid);
560 			if (error)
561 				return (NULL);
562 			mtx_enter(&nd6_mtx);
563 			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
564 			if (ln != NULL)
565 				ln->ln_state = ND6_LLINFO_NOSTATE;
566 			mtx_leave(&nd6_mtx);
567 		} else
568 			return (NULL);
569 	}
570 	/*
571 	 * Validation for the entry.
572 	 * Note that the check for rt_llinfo is necessary because a cloned
573 	 * route from a parent route that has the L flag (e.g. the default
574 	 * route to a p2p interface) may have the flag, too, while the
575 	 * destination is not actually a neighbor.
576 	 */
577 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
578 	    rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
579 	    (ifp != NULL && rt->rt_ifidx != ifp->if_index)) {
580 		if (create) {
581 			char addr[INET6_ADDRSTRLEN];
582 			nd6log((LOG_DEBUG, "%s: failed to lookup %s (if=%s)\n",
583 			    __func__,
584 			    inet_ntop(AF_INET6, addr6, addr, sizeof(addr)),
585 			    ifp ? ifp->if_xname : "unspec"));
586 		}
587 		rtfree(rt);
588 		return (NULL);
589 	}
590 	return (rt);
591 }
592 
593 /*
594  * Detect if a given IPv6 address identifies a neighbor on a given link.
595  * XXX: should take care of the destination of a p2p link?
596  */
597 int
nd6_is_addr_neighbor(const struct sockaddr_in6 * addr,struct ifnet * ifp)598 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
599 {
600 	struct in6_ifaddr *ia6;
601 	struct ifaddr *ifa;
602 	struct rtentry *rt;
603 
604 	/*
605 	 * A link-local address is always a neighbor.
606 	 * XXX: we should use the sin6_scope_id field rather than the embedded
607 	 * interface index.
608 	 * XXX: a link does not necessarily specify a single interface.
609 	 */
610 	if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
611 	    ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
612 		return (1);
613 
614 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
615 		if (ifa->ifa_addr->sa_family != AF_INET6)
616 			continue;
617 
618 		ia6 = ifatoia6(ifa);
619 
620 		/* Prefix check down below. */
621 		if (ia6->ia6_flags & IN6_IFF_AUTOCONF)
622 			continue;
623 
624 		if (IN6_ARE_MASKED_ADDR_EQUAL(&addr->sin6_addr,
625 		    &ia6->ia_addr.sin6_addr,
626 		    &ia6->ia_prefixmask.sin6_addr))
627 			return (1);
628 	}
629 
630 	/*
631 	 * Even if the address matches none of our addresses, it might be
632 	 * in the neighbor cache.
633 	 */
634 	rt = nd6_lookup(&addr->sin6_addr, 0, ifp, ifp->if_rdomain);
635 	if (rt != NULL) {
636 		rtfree(rt);
637 		return (1);
638 	}
639 
640 	return (0);
641 }
642 
643 void
nd6_invalidate(struct rtentry * rt)644 nd6_invalidate(struct rtentry *rt)
645 {
646 	struct llinfo_nd6 *ln;
647 	struct sockaddr_dl *sdl = satosdl(rt->rt_gateway);
648 
649 	mtx_enter(&nd6_mtx);
650 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
651 	if (ln == NULL) {
652 		mtx_leave(&nd6_mtx);
653 		return;
654 	}
655 	atomic_sub_int(&ln_hold_total, mq_purge(&ln->ln_mq));
656 	sdl->sdl_alen = 0;
657 	ln->ln_state = ND6_LLINFO_INCOMPLETE;
658 	ln->ln_asked = 0;
659 	mtx_leave(&nd6_mtx);
660 }
661 
662 /*
663  * Free an nd6 llinfo entry.
664  */
665 void
nd6_free(struct rtentry * rt,int i_am_router)666 nd6_free(struct rtentry *rt, int i_am_router)
667 {
668 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
669 	struct in6_addr in6 = satosin6(rt_key(rt))->sin6_addr;
670 	struct ifnet *ifp;
671 
672 	NET_ASSERT_LOCKED_EXCLUSIVE();
673 
674 	ifp = if_get(rt->rt_ifidx);
675 
676 	if (!i_am_router) {
677 		if (ln->ln_router) {
678 			/*
679 			 * rt6_flush must be called whether or not the neighbor
680 			 * is in the Default Router List.
681 			 * See a corresponding comment in nd6_na_input().
682 			 */
683 			rt6_flush(&in6, ifp);
684 		}
685 	}
686 
687 	KASSERT(!ISSET(rt->rt_flags, RTF_LOCAL));
688 	nd6_invalidate(rt);
689 
690 	/*
691 	 * Detach the route from the routing tree and the list of neighbor
692 	 * caches, and disable the route entry not to be used in already
693 	 * cached routes.
694 	 */
695 	if (!ISSET(rt->rt_flags, RTF_STATIC|RTF_CACHED))
696 		rtdeletemsg(rt, ifp, ifp->if_rdomain);
697 
698 	if_put(ifp);
699 }
700 
701 /*
702  * Upper-layer reachability hint for Neighbor Unreachability Detection.
703  *
704  * XXX cost-effective methods?
705  */
706 void
nd6_nud_hint(struct rtentry * rt)707 nd6_nud_hint(struct rtentry *rt)
708 {
709 	struct llinfo_nd6 *ln;
710 	struct ifnet *ifp;
711 
712 	NET_ASSERT_LOCKED_EXCLUSIVE();
713 
714 	ifp = if_get(rt->rt_ifidx);
715 	if (ifp == NULL)
716 		return;
717 
718 	if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
719 	    (rt->rt_flags & RTF_LLINFO) == 0 ||
720 	    rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
721 	    rt->rt_gateway->sa_family != AF_LINK) {
722 		/* This is not a host route. */
723 		goto out;
724 	}
725 
726 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
727 	if (ln->ln_state < ND6_LLINFO_REACHABLE)
728 		goto out;
729 
730 	/*
731 	 * if we get upper-layer reachability confirmation many times,
732 	 * it is possible we have false information.
733 	 */
734 	ln->ln_byhint++;
735 	if (ln->ln_byhint > nd6_maxnudhint)
736 		goto out;
737 
738 	ln->ln_state = ND6_LLINFO_REACHABLE;
739 	if (!ND6_LLINFO_PERMANENT(ln))
740 		nd6_llinfo_settimer(ln, ifp->if_nd->reachable);
741 out:
742 	if_put(ifp);
743 }
744 
745 void
nd6_rtrequest(struct ifnet * ifp,int req,struct rtentry * rt)746 nd6_rtrequest(struct ifnet *ifp, int req, struct rtentry *rt)
747 {
748 	struct sockaddr *gate = rt->rt_gateway;
749 	struct llinfo_nd6 *ln;
750 	struct ifaddr *ifa;
751 	struct in6_ifaddr *ifa6;
752 
753 	if (ISSET(rt->rt_flags, RTF_GATEWAY|RTF_MULTICAST|RTF_MPLS))
754 		return;
755 
756 	if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
757 		/*
758 		 * This is probably an interface direct route for a link
759 		 * which does not need neighbor caches (e.g. fe80::%lo0/64).
760 		 * We do not need special treatment below for such a route.
761 		 * Moreover, the RTF_LLINFO flag which would be set below
762 		 * would annoy the ndp(8) command.
763 		 */
764 		return;
765 	}
766 
767 	if (req == RTM_RESOLVE && nd6_need_cache(ifp) == 0) {
768 		/*
769 		 * For routing daemons like ospf6d we allow neighbor discovery
770 		 * based on the cloning route only.  This allows us to send
771 		 * packets directly into a network without having an address
772 		 * with matching prefix on the interface.  If the cloning
773 		 * route is used for an 6to4 interface, we would mistakenly
774 		 * make a neighbor cache for the host route, and would see
775 		 * strange neighbor solicitation for the corresponding
776 		 * destination.  In order to avoid confusion, we check if the
777 		 * interface is suitable for neighbor discovery, and stop the
778 		 * process if not.  Additionally, we remove the LLINFO flag
779 		 * so that ndp(8) will not try to get the neighbor information
780 		 * of the destination.
781 		 */
782 		rt->rt_flags &= ~RTF_LLINFO;
783 		return;
784 	}
785 
786 	switch (req) {
787 	case RTM_ADD:
788 		if (rt->rt_flags & RTF_CLONING) {
789 			rt->rt_expire = 0;
790 			break;
791 		}
792 		if ((rt->rt_flags & RTF_LOCAL) && rt->rt_llinfo == NULL)
793 			rt->rt_expire = 0;
794 		/* FALLTHROUGH */
795 	case RTM_RESOLVE:
796 		if (gate->sa_family != AF_LINK ||
797 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
798 			log(LOG_DEBUG, "%s: bad gateway value: %s\n",
799 			    __func__, ifp->if_xname);
800 			break;
801 		}
802 		satosdl(gate)->sdl_type = ifp->if_type;
803 		satosdl(gate)->sdl_index = ifp->if_index;
804 		/*
805 		 * Case 2: This route may come from cloning, or a manual route
806 		 * add with a LL address.
807 		 */
808 		ln = pool_get(&nd6_pool, PR_NOWAIT | PR_ZERO);
809 		if (ln == NULL) {
810 			log(LOG_DEBUG, "%s: pool get failed\n", __func__);
811 			break;
812 		}
813 
814 		mtx_enter(&nd6_mtx);
815 		if (rt->rt_llinfo != NULL) {
816 			/* we lost the race, another thread has entered it */
817 			mtx_leave(&nd6_mtx);
818 			pool_put(&nd6_pool, ln);
819 			break;
820 		}
821 		nd6_inuse++;
822 		mq_init(&ln->ln_mq, LN_HOLD_QUEUE, IPL_SOFTNET);
823 		rt->rt_llinfo = (caddr_t)ln;
824 		ln->ln_rt = rt;
825 		rt->rt_flags |= RTF_LLINFO;
826 		TAILQ_INSERT_HEAD(&nd6_list, ln, ln_list);
827 		/* this is required for "ndp" command. - shin */
828 		if (req == RTM_ADD) {
829 			/*
830 			 * gate should have some valid AF_LINK entry,
831 			 * and ln expire should have some lifetime
832 			 * which is specified by ndp command.
833 			 */
834 			ln->ln_state = ND6_LLINFO_REACHABLE;
835 			ln->ln_byhint = 0;
836 		} else {
837 			/*
838 			 * When req == RTM_RESOLVE, rt is created and
839 			 * initialized in rtrequest(), so rt_expire is 0.
840 			 */
841 			ln->ln_state = ND6_LLINFO_NOSTATE;
842 			nd6_llinfo_settimer(ln, 0);
843 		}
844 
845 		/*
846 		 * If we have too many cache entries, initiate immediate
847 		 * purging for some "less recently used" entries.  Note that
848 		 * we cannot directly call nd6_free() here because it would
849 		 * cause re-entering rtable related routines triggering
850 		 * lock-order-reversal problems.
851 		 */
852 		if (ip6_neighborgcthresh >= 0 &&
853 		    nd6_inuse >= ip6_neighborgcthresh) {
854 			int i;
855 
856 			for (i = 0; i < 10; i++) {
857 				struct llinfo_nd6 *ln_end;
858 
859 				ln_end = TAILQ_LAST(&nd6_list, llinfo_nd6_head);
860 				if (ln_end == ln)
861 					break;
862 
863 				/* Move this entry to the head */
864 				TAILQ_REMOVE(&nd6_list, ln_end, ln_list);
865 				TAILQ_INSERT_HEAD(&nd6_list, ln_end, ln_list);
866 
867 				if (ND6_LLINFO_PERMANENT(ln_end))
868 					continue;
869 
870 				if (ln_end->ln_state > ND6_LLINFO_INCOMPLETE)
871 					ln_end->ln_state = ND6_LLINFO_STALE;
872 				else
873 					ln_end->ln_state = ND6_LLINFO_PURGE;
874 				nd6_llinfo_settimer(ln_end, 0);
875 			}
876 		}
877 
878 		/*
879 		 * check if rt_key(rt) is one of my address assigned
880 		 * to the interface.
881 		 */
882 		ifa6 = in6ifa_ifpwithaddr(ifp,
883 		    &satosin6(rt_key(rt))->sin6_addr);
884 		ifa = ifa6 ? &ifa6->ia_ifa : NULL;
885 		if (ifa != NULL ||
886 		    (rt->rt_flags & RTF_ANNOUNCE)) {
887 			ln->ln_state = ND6_LLINFO_REACHABLE;
888 			ln->ln_byhint = 0;
889 			rt->rt_expire = 0;
890 		}
891 		mtx_leave(&nd6_mtx);
892 
893 		/* join solicited node multicast for proxy ND */
894 		if (ifa == NULL &&
895 		    (rt->rt_flags & RTF_ANNOUNCE) &&
896 		    (ifp->if_flags & IFF_MULTICAST)) {
897 			struct in6_addr llsol;
898 			int error;
899 
900 			llsol = satosin6(rt_key(rt))->sin6_addr;
901 			llsol.s6_addr16[0] = htons(0xff02);
902 			llsol.s6_addr16[1] = htons(ifp->if_index);
903 			llsol.s6_addr32[1] = 0;
904 			llsol.s6_addr32[2] = htonl(1);
905 			llsol.s6_addr8[12] = 0xff;
906 
907 			KERNEL_LOCK();
908 			if (in6_addmulti(&llsol, ifp, &error)) {
909 				char addr[INET6_ADDRSTRLEN];
910 				nd6log((LOG_ERR, "%s: failed to join "
911 				    "%s (errno=%d)\n", ifp->if_xname,
912 				    inet_ntop(AF_INET6, &llsol,
913 					addr, sizeof(addr)),
914 				    error));
915 			}
916 			KERNEL_UNLOCK();
917 		}
918 		break;
919 
920 	case RTM_DELETE:
921 		mtx_enter(&nd6_mtx);
922 		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
923 		if (ln == NULL) {
924 			/* we lost the race, another thread has removed it */
925 			mtx_leave(&nd6_mtx);
926 			break;
927 		}
928 		nd6_inuse--;
929 		TAILQ_REMOVE(&nd6_list, ln, ln_list);
930 		rt->rt_expire = 0;
931 		rt->rt_llinfo = NULL;
932 		rt->rt_flags &= ~RTF_LLINFO;
933 		atomic_sub_int(&ln_hold_total, mq_purge(&ln->ln_mq));
934 		mtx_leave(&nd6_mtx);
935 
936 		pool_put(&nd6_pool, ln);
937 
938 		/* leave from solicited node multicast for proxy ND */
939 		if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
940 		    (ifp->if_flags & IFF_MULTICAST) != 0) {
941 			struct in6_addr llsol;
942 			struct in6_multi *in6m;
943 
944 			llsol = satosin6(rt_key(rt))->sin6_addr;
945 			llsol.s6_addr16[0] = htons(0xff02);
946 			llsol.s6_addr16[1] = htons(ifp->if_index);
947 			llsol.s6_addr32[1] = 0;
948 			llsol.s6_addr32[2] = htonl(1);
949 			llsol.s6_addr8[12] = 0xff;
950 
951 			KERNEL_LOCK();
952 			IN6_LOOKUP_MULTI(llsol, ifp, in6m);
953 			if (in6m)
954 				in6_delmulti(in6m);
955 			KERNEL_UNLOCK();
956 		}
957 		break;
958 
959 	case RTM_INVALIDATE:
960 		if (!ISSET(rt->rt_flags, RTF_LOCAL))
961 			nd6_invalidate(rt);
962 		break;
963 	}
964 }
965 
966 int
nd6_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp)967 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
968 {
969 	struct in6_ndireq *ndi = (struct in6_ndireq *)data;
970 	struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
971 	struct rtentry *rt;
972 
973 	switch (cmd) {
974 	case SIOCGIFINFO_IN6:
975 		NET_LOCK_SHARED();
976 		ndi->ndi = *ifp->if_nd;
977 		NET_UNLOCK_SHARED();
978 		return (0);
979 	case SIOCGNBRINFO_IN6:
980 	{
981 		struct llinfo_nd6 *ln;
982 		struct in6_addr nb_addr = nbi->addr; /* make local for safety */
983 		time_t expire;
984 
985 		NET_LOCK_SHARED();
986 		/*
987 		 * XXX: KAME specific hack for scoped addresses
988 		 *      XXXX: for other scopes than link-local?
989 		 */
990 		if (IN6_IS_ADDR_LINKLOCAL(&nb_addr) ||
991 		    IN6_IS_ADDR_MC_LINKLOCAL(&nb_addr)) {
992 			u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
993 
994 			if (*idp == 0)
995 				*idp = htons(ifp->if_index);
996 		}
997 
998 		rt = nd6_lookup(&nb_addr, 0, ifp, ifp->if_rdomain);
999 		mtx_enter(&nd6_mtx);
1000 		if (rt == NULL ||
1001 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) {
1002 			mtx_leave(&nd6_mtx);
1003 			rtfree(rt);
1004 			NET_UNLOCK_SHARED();
1005 			return (EINVAL);
1006 		}
1007 		expire = ln->ln_rt->rt_expire;
1008 		if (expire != 0) {
1009 			expire -= getuptime();
1010 			expire += gettime();
1011 		}
1012 
1013 		nbi->state = ln->ln_state;
1014 		nbi->asked = ln->ln_asked;
1015 		nbi->isrouter = ln->ln_router;
1016 		nbi->expire = expire;
1017 		mtx_leave(&nd6_mtx);
1018 
1019 		rtfree(rt);
1020 		NET_UNLOCK_SHARED();
1021 		return (0);
1022 	}
1023 	}
1024 	return (0);
1025 }
1026 
1027 /*
1028  * Create neighbor cache entry and cache link-layer address,
1029  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
1030  *
1031  * type - ICMP6 type
1032  * code - type dependent information
1033  */
1034 void
nd6_cache_lladdr(struct ifnet * ifp,const struct in6_addr * from,char * lladdr,int lladdrlen,int type,int code,int i_am_router)1035 nd6_cache_lladdr(struct ifnet *ifp, const struct in6_addr *from, char *lladdr,
1036     int lladdrlen, int type, int code, int i_am_router)
1037 {
1038 	struct rtentry *rt;
1039 	struct llinfo_nd6 *ln;
1040 	int is_newentry;
1041 	struct sockaddr_dl *sdl;
1042 	int do_update;
1043 	int olladdr;
1044 	int llchange;
1045 	int newstate = 0;
1046 
1047 	NET_ASSERT_LOCKED_EXCLUSIVE();
1048 
1049 	if (!ifp)
1050 		panic("%s: ifp == NULL", __func__);
1051 	if (!from)
1052 		panic("%s: from == NULL", __func__);
1053 
1054 	/* nothing must be updated for unspecified address */
1055 	if (IN6_IS_ADDR_UNSPECIFIED(from))
1056 		return;
1057 
1058 	/*
1059 	 * Validation about ifp->if_addrlen and lladdrlen must be done in
1060 	 * the caller.
1061 	 *
1062 	 * XXX If the link does not have link-layer address, what should
1063 	 * we do? (ifp->if_addrlen == 0)
1064 	 * Spec says nothing in sections for RA, RS and NA.  There's small
1065 	 * description on it in NS section (RFC 2461 7.2.3).
1066 	 */
1067 
1068 	rt = nd6_lookup(from, 0, ifp, ifp->if_rdomain);
1069 	if (rt == NULL) {
1070 		rt = nd6_lookup(from, 1, ifp, ifp->if_rdomain);
1071 		is_newentry = 1;
1072 	} else {
1073 		/* do not overwrite local or static entry */
1074 		if (ISSET(rt->rt_flags, RTF_STATIC|RTF_LOCAL)) {
1075 			rtfree(rt);
1076 			return;
1077 		}
1078 		is_newentry = 0;
1079 	}
1080 
1081 	if (!rt)
1082 		return;
1083 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1084 fail:
1085 		nd6_free(rt, i_am_router);
1086 		rtfree(rt);
1087 		return;
1088 	}
1089 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1090 	if (ln == NULL)
1091 		goto fail;
1092 	if (rt->rt_gateway == NULL)
1093 		goto fail;
1094 	if (rt->rt_gateway->sa_family != AF_LINK)
1095 		goto fail;
1096 	sdl = satosdl(rt->rt_gateway);
1097 
1098 	olladdr = (sdl->sdl_alen) ? 1 : 0;
1099 	if (olladdr && lladdr) {
1100 		if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1101 			llchange = 1;
1102 		else
1103 			llchange = 0;
1104 	} else
1105 		llchange = 0;
1106 
1107 	/*
1108 	 * newentry olladdr  lladdr  llchange	(*=record)
1109 	 *	0	n	n	--	(1)
1110 	 *	0	y	n	--	(2)
1111 	 *	0	n	y	--	(3) * STALE
1112 	 *	0	y	y	n	(4) *
1113 	 *	0	y	y	y	(5) * STALE
1114 	 *	1	--	n	--	(6)   NOSTATE(= PASSIVE)
1115 	 *	1	--	y	--	(7) * STALE
1116 	 */
1117 
1118 	if (llchange) {
1119 		char addr[INET6_ADDRSTRLEN];
1120 		log(LOG_INFO, "ndp info overwritten for %s by %s on %s\n",
1121 		    inet_ntop(AF_INET6, from, addr, sizeof(addr)),
1122 		    ether_sprintf(lladdr), ifp->if_xname);
1123 	}
1124 	if (lladdr) {		/* (3-5) and (7) */
1125 		/*
1126 		 * Record source link-layer address
1127 		 * XXX is it dependent to ifp->if_type?
1128 		 */
1129 		sdl->sdl_alen = ifp->if_addrlen;
1130 		bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1131 	}
1132 
1133 	if (!is_newentry) {
1134 		if ((!olladdr && lladdr) ||		/* (3) */
1135 		    (olladdr && lladdr && llchange)) {	/* (5) */
1136 			do_update = 1;
1137 			newstate = ND6_LLINFO_STALE;
1138 		} else					/* (1-2,4) */
1139 			do_update = 0;
1140 	} else {
1141 		do_update = 1;
1142 		if (!lladdr)				/* (6) */
1143 			newstate = ND6_LLINFO_NOSTATE;
1144 		else					/* (7) */
1145 			newstate = ND6_LLINFO_STALE;
1146 	}
1147 
1148 	if (do_update) {
1149 		/*
1150 		 * Update the state of the neighbor cache.
1151 		 */
1152 		ln->ln_state = newstate;
1153 
1154 		if (ln->ln_state == ND6_LLINFO_STALE) {
1155 			/*
1156 			 * Since nd6_resolve() in ifp->if_output() will cause
1157 			 * state transition to DELAY and reset the timer,
1158 			 * we must set the timer now, although it is actually
1159 			 * meaningless.
1160 			 */
1161 			nd6_llinfo_settimer(ln, nd6_gctimer);
1162 			if_output_mq(ifp, &ln->ln_mq, &ln_hold_total,
1163 			    rt_key(rt), rt);
1164 		} else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1165 			/* probe right away */
1166 			nd6_llinfo_settimer(ln, 0);
1167 		}
1168 	}
1169 
1170 	/*
1171 	 * ICMP6 type dependent behavior.
1172 	 *
1173 	 * NS: clear IsRouter if new entry
1174 	 * RS: clear IsRouter
1175 	 * RA: set IsRouter if there's lladdr
1176 	 * redir: clear IsRouter if new entry
1177 	 *
1178 	 * RA case, (1):
1179 	 * The spec says that we must set IsRouter in the following cases:
1180 	 * - If lladdr exist, set IsRouter.  This means (1-5).
1181 	 * - If it is old entry (!newentry), set IsRouter.  This means (7).
1182 	 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1183 	 * A question arises for (1) case.  (1) case has no lladdr in the
1184 	 * neighbor cache, this is similar to (6).
1185 	 * This case is rare but we figured that we MUST NOT set IsRouter.
1186 	 *
1187 	 * newentry olladdr  lladdr  llchange	    NS  RS  RA	redir
1188 	 *							D R
1189 	 *	0	n	n	--	(1)	c   ?     s
1190 	 *	0	y	n	--	(2)	c   s     s
1191 	 *	0	n	y	--	(3)	c   s     s
1192 	 *	0	y	y	n	(4)	c   s     s
1193 	 *	0	y	y	y	(5)	c   s     s
1194 	 *	1	--	n	--	(6) c	c	c s
1195 	 *	1	--	y	--	(7) c	c   s	c s
1196 	 *
1197 	 *					(c=clear s=set)
1198 	 */
1199 	switch (type & 0xff) {
1200 	case ND_NEIGHBOR_SOLICIT:
1201 		/*
1202 		 * New entry must have is_router flag cleared.
1203 		 */
1204 		if (is_newentry)	/* (6-7) */
1205 			ln->ln_router = 0;
1206 		break;
1207 	case ND_REDIRECT:
1208 		/*
1209 		 * If the icmp is a redirect to a better router, always set the
1210 		 * is_router flag.  Otherwise, if the entry is newly created,
1211 		 * clear the flag.  [RFC 2461, sec 8.3]
1212 		 */
1213 		if (code == ND_REDIRECT_ROUTER)
1214 			ln->ln_router = 1;
1215 		else if (is_newentry) /* (6-7) */
1216 			ln->ln_router = 0;
1217 		break;
1218 	case ND_ROUTER_SOLICIT:
1219 		/*
1220 		 * is_router flag must always be cleared.
1221 		 */
1222 		ln->ln_router = 0;
1223 		break;
1224 	case ND_ROUTER_ADVERT:
1225 		/*
1226 		 * Mark an entry with lladdr as a router.
1227 		 */
1228 		if ((!is_newentry && (olladdr || lladdr)) ||	/* (2-5) */
1229 		    (is_newentry && lladdr)) {			/* (7) */
1230 			ln->ln_router = 1;
1231 		}
1232 		break;
1233 	}
1234 
1235 	rtfree(rt);
1236 }
1237 
1238 void
nd6_slowtimo(void * ignored_arg)1239 nd6_slowtimo(void *ignored_arg)
1240 {
1241 	struct nd_ifinfo *nd6if;
1242 	struct ifnet *ifp;
1243 
1244 	NET_LOCK();
1245 
1246 	timeout_add_sec(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL);
1247 
1248 	TAILQ_FOREACH(ifp, &ifnetlist, if_list) {
1249 		nd6if = ifp->if_nd;
1250 		if ((nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1251 			/*
1252 			 * Since reachable time rarely changes by router
1253 			 * advertisements, we SHOULD insure that a new random
1254 			 * value gets recomputed at least once every few hours.
1255 			 * (RFC 2461, 6.3.4)
1256 			 */
1257 			nd6if->recalctm = ND6_RECALC_REACHTM_INTERVAL;
1258 			nd6if->reachable = ND_COMPUTE_RTIME(REACHABLE_TIME);
1259 		}
1260 	}
1261 	NET_UNLOCK();
1262 }
1263 
1264 int
nd6_resolve(struct ifnet * ifp,struct rtentry * rt0,struct mbuf * m,struct sockaddr * dst,u_char * desten)1265 nd6_resolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
1266     struct sockaddr *dst, u_char *desten)
1267 {
1268 	struct sockaddr_dl *sdl;
1269 	struct rtentry *rt;
1270 	struct llinfo_nd6 *ln;
1271 	struct in6_addr saddr6;
1272 	time_t uptime;
1273 	int solicit = 0;
1274 
1275 	if (m->m_flags & M_MCAST) {
1276 		ETHER_MAP_IPV6_MULTICAST(&satosin6(dst)->sin6_addr, desten);
1277 		return (0);
1278 	}
1279 
1280 	uptime = getuptime();
1281 	rt = rt_getll(rt0);
1282 
1283 	if (ISSET(rt->rt_flags, RTF_REJECT) &&
1284 	    (rt->rt_expire == 0 || rt->rt_expire > uptime)) {
1285 		m_freem(m);
1286 		return (rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
1287 	}
1288 
1289 	/*
1290 	 * Address resolution or Neighbor Unreachability Detection
1291 	 * for the next hop.
1292 	 * At this point, the destination of the packet must be a unicast
1293 	 * or an anycast address(i.e. not a multicast).
1294 	 */
1295 	if (!ISSET(rt->rt_flags, RTF_LLINFO)) {
1296 		char addr[INET6_ADDRSTRLEN];
1297 		log(LOG_DEBUG, "%s: %s: route contains no ND information\n",
1298 		    __func__, inet_ntop(AF_INET6,
1299 		    &satosin6(rt_key(rt))->sin6_addr, addr, sizeof(addr)));
1300 		goto bad;
1301 	}
1302 
1303 	if (rt->rt_gateway->sa_family != AF_LINK) {
1304 		printf("%s: something odd happens\n", __func__);
1305 		goto bad;
1306 	}
1307 
1308 	mtx_enter(&nd6_mtx);
1309 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1310 	if (ln == NULL) {
1311 		mtx_leave(&nd6_mtx);
1312 		goto bad;
1313 	}
1314 
1315 	/*
1316 	 * Move this entry to the head of the queue so that it is less likely
1317 	 * for this entry to be a target of forced garbage collection (see
1318 	 * nd6_rtrequest()).
1319 	 */
1320 	TAILQ_REMOVE(&nd6_list, ln, ln_list);
1321 	TAILQ_INSERT_HEAD(&nd6_list, ln, ln_list);
1322 
1323 	/*
1324 	 * The first time we send a packet to a neighbor whose entry is
1325 	 * STALE, we have to change the state to DELAY and set a timer to
1326 	 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure we do
1327 	 * neighbor unreachability detection on expiration.
1328 	 * (RFC 2461 7.3.3)
1329 	 */
1330 	if (ln->ln_state == ND6_LLINFO_STALE) {
1331 		ln->ln_asked = 0;
1332 		ln->ln_state = ND6_LLINFO_DELAY;
1333 		nd6_llinfo_settimer(ln, nd6_delay);
1334 	}
1335 
1336 	/*
1337 	 * If the neighbor cache entry has a state other than INCOMPLETE
1338 	 * (i.e. its link-layer address is already resolved), just
1339 	 * send the packet.
1340 	 */
1341 	if (ln->ln_state > ND6_LLINFO_INCOMPLETE) {
1342 		mtx_leave(&nd6_mtx);
1343 
1344 		sdl = satosdl(rt->rt_gateway);
1345 		if (sdl->sdl_alen != ETHER_ADDR_LEN) {
1346 			char addr[INET6_ADDRSTRLEN];
1347 			log(LOG_DEBUG, "%s: %s: incorrect nd6 information\n",
1348 			    __func__,
1349 			    inet_ntop(AF_INET6, &satosin6(dst)->sin6_addr,
1350 				addr, sizeof(addr)));
1351 			goto bad;
1352 		}
1353 
1354 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
1355 		return (0);
1356 	}
1357 
1358 	/*
1359 	 * There is a neighbor cache entry, but no ethernet address
1360 	 * response yet.  Insert mbuf in hold queue if below limit.
1361 	 * If above the limit free the queue without queuing the new packet.
1362 	 */
1363 	if (ln->ln_state == ND6_LLINFO_NOSTATE)
1364 		ln->ln_state = ND6_LLINFO_INCOMPLETE;
1365 	/* source address of prompting packet is needed by nd6_ns_output() */
1366 	if (m->m_len >= sizeof(struct ip6_hdr)) {
1367 		memcpy(&ln->ln_saddr6, &mtod(m, struct ip6_hdr *)->ip6_src,
1368 		    sizeof(ln->ln_saddr6));
1369 	}
1370 	if (atomic_inc_int_nv(&ln_hold_total) <= LN_HOLD_TOTAL) {
1371 		if (mq_push(&ln->ln_mq, m) != 0)
1372 			atomic_dec_int(&ln_hold_total);
1373 	} else {
1374 		atomic_sub_int(&ln_hold_total, mq_purge(&ln->ln_mq) + 1);
1375 		m_freem(m);
1376 	}
1377 
1378 	/*
1379 	 * If there has been no NS for the neighbor after entering the
1380 	 * INCOMPLETE state, send the first solicitation.
1381 	 */
1382 	if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
1383 		ln->ln_asked++;
1384 		nd6_llinfo_settimer(ln, RETRANS_TIMER / 1000);
1385 		saddr6 = ln->ln_saddr6;
1386 		solicit = 1;
1387 	}
1388 	mtx_leave(&nd6_mtx);
1389 
1390 	if (solicit)
1391 		nd6_ns_output(ifp, NULL, &satosin6(dst)->sin6_addr, &saddr6, 0);
1392 	return (EAGAIN);
1393 
1394 bad:
1395 	m_freem(m);
1396 	return (EINVAL);
1397 }
1398 
1399 int
nd6_need_cache(struct ifnet * ifp)1400 nd6_need_cache(struct ifnet *ifp)
1401 {
1402 	/*
1403 	 * RFC2893 says:
1404 	 * - unidirectional tunnels needs no ND
1405 	 */
1406 	switch (ifp->if_type) {
1407 	case IFT_ETHER:
1408 	case IFT_IEEE80211:
1409 	case IFT_CARP:
1410 		return (1);
1411 	default:
1412 		return (0);
1413 	}
1414 }
1415