1 /* $OpenBSD: nd6.c,v 1.284 2025/01/31 11:44:47 bluhm 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();
713
714 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
715 (rt->rt_flags & RTF_LLINFO) == 0 ||
716 rt->rt_gateway == NULL ||
717 rt->rt_gateway->sa_family != AF_LINK) {
718 /* This is not a host route. */
719 return;
720 }
721
722 ifp = if_get(rt->rt_ifidx);
723 if (ifp == NULL)
724 return;
725
726 mtx_enter(&nd6_mtx);
727
728 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
729 if (ln == NULL)
730 goto out;
731 if (ln->ln_state < ND6_LLINFO_REACHABLE)
732 goto out;
733
734 /*
735 * if we get upper-layer reachability confirmation many times,
736 * it is possible we have false information.
737 */
738 ln->ln_byhint++;
739 if (ln->ln_byhint > nd6_maxnudhint)
740 goto out;
741
742 ln->ln_state = ND6_LLINFO_REACHABLE;
743 if (!ND6_LLINFO_PERMANENT(ln))
744 nd6_llinfo_settimer(ln, ifp->if_nd->reachable);
745 out:
746 mtx_leave(&nd6_mtx);
747 if_put(ifp);
748 }
749
750 void
nd6_rtrequest(struct ifnet * ifp,int req,struct rtentry * rt)751 nd6_rtrequest(struct ifnet *ifp, int req, struct rtentry *rt)
752 {
753 struct sockaddr *gate = rt->rt_gateway;
754 struct llinfo_nd6 *ln;
755 struct ifaddr *ifa;
756 struct in6_ifaddr *ifa6;
757
758 if (ISSET(rt->rt_flags, RTF_GATEWAY|RTF_MULTICAST|RTF_MPLS))
759 return;
760
761 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
762 /*
763 * This is probably an interface direct route for a link
764 * which does not need neighbor caches (e.g. fe80::%lo0/64).
765 * We do not need special treatment below for such a route.
766 * Moreover, the RTF_LLINFO flag which would be set below
767 * would annoy the ndp(8) command.
768 */
769 return;
770 }
771
772 if (req == RTM_RESOLVE && nd6_need_cache(ifp) == 0) {
773 /*
774 * For routing daemons like ospf6d we allow neighbor discovery
775 * based on the cloning route only. This allows us to send
776 * packets directly into a network without having an address
777 * with matching prefix on the interface. If the cloning
778 * route is used for an 6to4 interface, we would mistakenly
779 * make a neighbor cache for the host route, and would see
780 * strange neighbor solicitation for the corresponding
781 * destination. In order to avoid confusion, we check if the
782 * interface is suitable for neighbor discovery, and stop the
783 * process if not. Additionally, we remove the LLINFO flag
784 * so that ndp(8) will not try to get the neighbor information
785 * of the destination.
786 */
787 rt->rt_flags &= ~RTF_LLINFO;
788 return;
789 }
790
791 switch (req) {
792 case RTM_ADD:
793 if (rt->rt_flags & RTF_CLONING) {
794 rt->rt_expire = 0;
795 break;
796 }
797 if ((rt->rt_flags & RTF_LOCAL) && rt->rt_llinfo == NULL)
798 rt->rt_expire = 0;
799 /* FALLTHROUGH */
800 case RTM_RESOLVE:
801 if (gate->sa_family != AF_LINK ||
802 gate->sa_len < sizeof(struct sockaddr_dl)) {
803 log(LOG_DEBUG, "%s: bad gateway value: %s\n",
804 __func__, ifp->if_xname);
805 break;
806 }
807 satosdl(gate)->sdl_type = ifp->if_type;
808 satosdl(gate)->sdl_index = ifp->if_index;
809 /*
810 * Case 2: This route may come from cloning, or a manual route
811 * add with a LL address.
812 */
813 ln = pool_get(&nd6_pool, PR_NOWAIT | PR_ZERO);
814 if (ln == NULL) {
815 log(LOG_DEBUG, "%s: pool get failed\n", __func__);
816 break;
817 }
818
819 mtx_enter(&nd6_mtx);
820 if (rt->rt_llinfo != NULL) {
821 /* we lost the race, another thread has entered it */
822 mtx_leave(&nd6_mtx);
823 pool_put(&nd6_pool, ln);
824 break;
825 }
826 nd6_inuse++;
827 mq_init(&ln->ln_mq, LN_HOLD_QUEUE, IPL_SOFTNET);
828 rt->rt_llinfo = (caddr_t)ln;
829 ln->ln_rt = rt;
830 rt->rt_flags |= RTF_LLINFO;
831 TAILQ_INSERT_HEAD(&nd6_list, ln, ln_list);
832 /* this is required for "ndp" command. - shin */
833 if (req == RTM_ADD) {
834 /*
835 * gate should have some valid AF_LINK entry,
836 * and ln expire should have some lifetime
837 * which is specified by ndp command.
838 */
839 ln->ln_state = ND6_LLINFO_REACHABLE;
840 ln->ln_byhint = 0;
841 } else {
842 /*
843 * When req == RTM_RESOLVE, rt is created and
844 * initialized in rtrequest(), so rt_expire is 0.
845 */
846 ln->ln_state = ND6_LLINFO_NOSTATE;
847 nd6_llinfo_settimer(ln, 0);
848 }
849
850 /*
851 * If we have too many cache entries, initiate immediate
852 * purging for some "less recently used" entries. Note that
853 * we cannot directly call nd6_free() here because it would
854 * cause re-entering rtable related routines triggering
855 * lock-order-reversal problems.
856 */
857 if (ip6_neighborgcthresh >= 0 &&
858 nd6_inuse >= ip6_neighborgcthresh) {
859 int i;
860
861 for (i = 0; i < 10; i++) {
862 struct llinfo_nd6 *ln_end;
863
864 ln_end = TAILQ_LAST(&nd6_list, llinfo_nd6_head);
865 if (ln_end == ln)
866 break;
867
868 /* Move this entry to the head */
869 TAILQ_REMOVE(&nd6_list, ln_end, ln_list);
870 TAILQ_INSERT_HEAD(&nd6_list, ln_end, ln_list);
871
872 if (ND6_LLINFO_PERMANENT(ln_end))
873 continue;
874
875 if (ln_end->ln_state > ND6_LLINFO_INCOMPLETE)
876 ln_end->ln_state = ND6_LLINFO_STALE;
877 else
878 ln_end->ln_state = ND6_LLINFO_PURGE;
879 nd6_llinfo_settimer(ln_end, 0);
880 }
881 }
882
883 /*
884 * check if rt_key(rt) is one of my address assigned
885 * to the interface.
886 */
887 ifa6 = in6ifa_ifpwithaddr(ifp,
888 &satosin6(rt_key(rt))->sin6_addr);
889 ifa = ifa6 ? &ifa6->ia_ifa : NULL;
890 if (ifa != NULL ||
891 (rt->rt_flags & RTF_ANNOUNCE)) {
892 ln->ln_state = ND6_LLINFO_REACHABLE;
893 ln->ln_byhint = 0;
894 rt->rt_expire = 0;
895 }
896 mtx_leave(&nd6_mtx);
897
898 /* join solicited node multicast for proxy ND */
899 if (ifa == NULL &&
900 (rt->rt_flags & RTF_ANNOUNCE) &&
901 (ifp->if_flags & IFF_MULTICAST)) {
902 struct in6_addr llsol;
903 int error;
904
905 llsol = satosin6(rt_key(rt))->sin6_addr;
906 llsol.s6_addr16[0] = htons(0xff02);
907 llsol.s6_addr16[1] = htons(ifp->if_index);
908 llsol.s6_addr32[1] = 0;
909 llsol.s6_addr32[2] = htonl(1);
910 llsol.s6_addr8[12] = 0xff;
911
912 KERNEL_LOCK();
913 if (in6_addmulti(&llsol, ifp, &error)) {
914 char addr[INET6_ADDRSTRLEN];
915 nd6log((LOG_ERR, "%s: failed to join "
916 "%s (errno=%d)\n", ifp->if_xname,
917 inet_ntop(AF_INET6, &llsol,
918 addr, sizeof(addr)),
919 error));
920 }
921 KERNEL_UNLOCK();
922 }
923 break;
924
925 case RTM_DELETE:
926 mtx_enter(&nd6_mtx);
927 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
928 if (ln == NULL) {
929 /* we lost the race, another thread has removed it */
930 mtx_leave(&nd6_mtx);
931 break;
932 }
933 nd6_inuse--;
934 TAILQ_REMOVE(&nd6_list, ln, ln_list);
935 rt->rt_expire = 0;
936 rt->rt_llinfo = NULL;
937 rt->rt_flags &= ~RTF_LLINFO;
938 atomic_sub_int(&ln_hold_total, mq_purge(&ln->ln_mq));
939 mtx_leave(&nd6_mtx);
940
941 pool_put(&nd6_pool, ln);
942
943 /* leave from solicited node multicast for proxy ND */
944 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
945 (ifp->if_flags & IFF_MULTICAST) != 0) {
946 struct in6_addr llsol;
947 struct in6_multi *in6m;
948
949 llsol = satosin6(rt_key(rt))->sin6_addr;
950 llsol.s6_addr16[0] = htons(0xff02);
951 llsol.s6_addr16[1] = htons(ifp->if_index);
952 llsol.s6_addr32[1] = 0;
953 llsol.s6_addr32[2] = htonl(1);
954 llsol.s6_addr8[12] = 0xff;
955
956 KERNEL_LOCK();
957 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
958 if (in6m)
959 in6_delmulti(in6m);
960 KERNEL_UNLOCK();
961 }
962 break;
963
964 case RTM_INVALIDATE:
965 if (!ISSET(rt->rt_flags, RTF_LOCAL))
966 nd6_invalidate(rt);
967 break;
968 }
969 }
970
971 int
nd6_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp)972 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
973 {
974 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
975 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
976 struct rtentry *rt;
977
978 switch (cmd) {
979 case SIOCGIFINFO_IN6:
980 NET_LOCK_SHARED();
981 ndi->ndi = *ifp->if_nd;
982 NET_UNLOCK_SHARED();
983 return (0);
984 case SIOCGNBRINFO_IN6:
985 {
986 struct llinfo_nd6 *ln;
987 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
988 time_t expire;
989
990 NET_LOCK_SHARED();
991 /*
992 * XXX: KAME specific hack for scoped addresses
993 * XXXX: for other scopes than link-local?
994 */
995 if (IN6_IS_ADDR_LINKLOCAL(&nb_addr) ||
996 IN6_IS_ADDR_MC_LINKLOCAL(&nb_addr)) {
997 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
998
999 if (*idp == 0)
1000 *idp = htons(ifp->if_index);
1001 }
1002
1003 rt = nd6_lookup(&nb_addr, 0, ifp, ifp->if_rdomain);
1004 mtx_enter(&nd6_mtx);
1005 if (rt == NULL ||
1006 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) {
1007 mtx_leave(&nd6_mtx);
1008 rtfree(rt);
1009 NET_UNLOCK_SHARED();
1010 return (EINVAL);
1011 }
1012 expire = ln->ln_rt->rt_expire;
1013 if (expire != 0) {
1014 expire -= getuptime();
1015 expire += gettime();
1016 }
1017
1018 nbi->state = ln->ln_state;
1019 nbi->asked = ln->ln_asked;
1020 nbi->isrouter = ln->ln_router;
1021 nbi->expire = expire;
1022 mtx_leave(&nd6_mtx);
1023
1024 rtfree(rt);
1025 NET_UNLOCK_SHARED();
1026 return (0);
1027 }
1028 }
1029 return (0);
1030 }
1031
1032 /*
1033 * Create neighbor cache entry and cache link-layer address,
1034 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1035 *
1036 * type - ICMP6 type
1037 * code - type dependent information
1038 */
1039 void
nd6_cache_lladdr(struct ifnet * ifp,const struct in6_addr * from,char * lladdr,int lladdrlen,int type,int code,int i_am_router)1040 nd6_cache_lladdr(struct ifnet *ifp, const struct in6_addr *from, char *lladdr,
1041 int lladdrlen, int type, int code, int i_am_router)
1042 {
1043 struct rtentry *rt;
1044 struct llinfo_nd6 *ln;
1045 int is_newentry;
1046 struct sockaddr_dl *sdl;
1047 int do_update;
1048 int olladdr;
1049 int llchange;
1050 int newstate = 0;
1051
1052 NET_ASSERT_LOCKED_EXCLUSIVE();
1053
1054 if (!ifp)
1055 panic("%s: ifp == NULL", __func__);
1056 if (!from)
1057 panic("%s: from == NULL", __func__);
1058
1059 /* nothing must be updated for unspecified address */
1060 if (IN6_IS_ADDR_UNSPECIFIED(from))
1061 return;
1062
1063 /*
1064 * Validation about ifp->if_addrlen and lladdrlen must be done in
1065 * the caller.
1066 *
1067 * XXX If the link does not have link-layer address, what should
1068 * we do? (ifp->if_addrlen == 0)
1069 * Spec says nothing in sections for RA, RS and NA. There's small
1070 * description on it in NS section (RFC 2461 7.2.3).
1071 */
1072
1073 rt = nd6_lookup(from, 0, ifp, ifp->if_rdomain);
1074 if (rt == NULL) {
1075 rt = nd6_lookup(from, 1, ifp, ifp->if_rdomain);
1076 is_newentry = 1;
1077 } else {
1078 /* do not overwrite local or static entry */
1079 if (ISSET(rt->rt_flags, RTF_STATIC|RTF_LOCAL)) {
1080 rtfree(rt);
1081 return;
1082 }
1083 is_newentry = 0;
1084 }
1085
1086 if (!rt)
1087 return;
1088 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1089 fail:
1090 nd6_free(rt, i_am_router);
1091 rtfree(rt);
1092 return;
1093 }
1094 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1095 if (ln == NULL)
1096 goto fail;
1097 if (rt->rt_gateway == NULL)
1098 goto fail;
1099 if (rt->rt_gateway->sa_family != AF_LINK)
1100 goto fail;
1101 sdl = satosdl(rt->rt_gateway);
1102
1103 olladdr = (sdl->sdl_alen) ? 1 : 0;
1104 if (olladdr && lladdr) {
1105 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1106 llchange = 1;
1107 else
1108 llchange = 0;
1109 } else
1110 llchange = 0;
1111
1112 /*
1113 * newentry olladdr lladdr llchange (*=record)
1114 * 0 n n -- (1)
1115 * 0 y n -- (2)
1116 * 0 n y -- (3) * STALE
1117 * 0 y y n (4) *
1118 * 0 y y y (5) * STALE
1119 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1120 * 1 -- y -- (7) * STALE
1121 */
1122
1123 if (llchange) {
1124 char addr[INET6_ADDRSTRLEN];
1125 log(LOG_INFO, "ndp info overwritten for %s by %s on %s\n",
1126 inet_ntop(AF_INET6, from, addr, sizeof(addr)),
1127 ether_sprintf(lladdr), ifp->if_xname);
1128 }
1129 if (lladdr) { /* (3-5) and (7) */
1130 /*
1131 * Record source link-layer address
1132 * XXX is it dependent to ifp->if_type?
1133 */
1134 sdl->sdl_alen = ifp->if_addrlen;
1135 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1136 }
1137
1138 if (!is_newentry) {
1139 if ((!olladdr && lladdr) || /* (3) */
1140 (olladdr && lladdr && llchange)) { /* (5) */
1141 do_update = 1;
1142 newstate = ND6_LLINFO_STALE;
1143 } else /* (1-2,4) */
1144 do_update = 0;
1145 } else {
1146 do_update = 1;
1147 if (!lladdr) /* (6) */
1148 newstate = ND6_LLINFO_NOSTATE;
1149 else /* (7) */
1150 newstate = ND6_LLINFO_STALE;
1151 }
1152
1153 if (do_update) {
1154 /*
1155 * Update the state of the neighbor cache.
1156 */
1157 ln->ln_state = newstate;
1158
1159 if (ln->ln_state == ND6_LLINFO_STALE) {
1160 /*
1161 * Since nd6_resolve() in ifp->if_output() will cause
1162 * state transition to DELAY and reset the timer,
1163 * we must set the timer now, although it is actually
1164 * meaningless.
1165 */
1166 nd6_llinfo_settimer(ln, nd6_gctimer);
1167 if_output_mq(ifp, &ln->ln_mq, &ln_hold_total,
1168 rt_key(rt), rt);
1169 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1170 /* probe right away */
1171 nd6_llinfo_settimer(ln, 0);
1172 }
1173 }
1174
1175 /*
1176 * ICMP6 type dependent behavior.
1177 *
1178 * NS: clear IsRouter if new entry
1179 * RS: clear IsRouter
1180 * RA: set IsRouter if there's lladdr
1181 * redir: clear IsRouter if new entry
1182 *
1183 * RA case, (1):
1184 * The spec says that we must set IsRouter in the following cases:
1185 * - If lladdr exist, set IsRouter. This means (1-5).
1186 * - If it is old entry (!newentry), set IsRouter. This means (7).
1187 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1188 * A question arises for (1) case. (1) case has no lladdr in the
1189 * neighbor cache, this is similar to (6).
1190 * This case is rare but we figured that we MUST NOT set IsRouter.
1191 *
1192 * newentry olladdr lladdr llchange NS RS RA redir
1193 * D R
1194 * 0 n n -- (1) c ? s
1195 * 0 y n -- (2) c s s
1196 * 0 n y -- (3) c s s
1197 * 0 y y n (4) c s s
1198 * 0 y y y (5) c s s
1199 * 1 -- n -- (6) c c c s
1200 * 1 -- y -- (7) c c s c s
1201 *
1202 * (c=clear s=set)
1203 */
1204 switch (type & 0xff) {
1205 case ND_NEIGHBOR_SOLICIT:
1206 /*
1207 * New entry must have is_router flag cleared.
1208 */
1209 if (is_newentry) /* (6-7) */
1210 ln->ln_router = 0;
1211 break;
1212 case ND_REDIRECT:
1213 /*
1214 * If the icmp is a redirect to a better router, always set the
1215 * is_router flag. Otherwise, if the entry is newly created,
1216 * clear the flag. [RFC 2461, sec 8.3]
1217 */
1218 if (code == ND_REDIRECT_ROUTER)
1219 ln->ln_router = 1;
1220 else if (is_newentry) /* (6-7) */
1221 ln->ln_router = 0;
1222 break;
1223 case ND_ROUTER_SOLICIT:
1224 /*
1225 * is_router flag must always be cleared.
1226 */
1227 ln->ln_router = 0;
1228 break;
1229 case ND_ROUTER_ADVERT:
1230 /*
1231 * Mark an entry with lladdr as a router.
1232 */
1233 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1234 (is_newentry && lladdr)) { /* (7) */
1235 ln->ln_router = 1;
1236 }
1237 break;
1238 }
1239
1240 rtfree(rt);
1241 }
1242
1243 void
nd6_slowtimo(void * ignored_arg)1244 nd6_slowtimo(void *ignored_arg)
1245 {
1246 struct nd_ifinfo *nd6if;
1247 struct ifnet *ifp;
1248
1249 NET_LOCK();
1250
1251 timeout_add_sec(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL);
1252
1253 TAILQ_FOREACH(ifp, &ifnetlist, if_list) {
1254 nd6if = ifp->if_nd;
1255 if ((nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1256 /*
1257 * Since reachable time rarely changes by router
1258 * advertisements, we SHOULD insure that a new random
1259 * value gets recomputed at least once every few hours.
1260 * (RFC 2461, 6.3.4)
1261 */
1262 nd6if->recalctm = ND6_RECALC_REACHTM_INTERVAL;
1263 nd6if->reachable = ND_COMPUTE_RTIME(REACHABLE_TIME);
1264 }
1265 }
1266 NET_UNLOCK();
1267 }
1268
1269 int
nd6_resolve(struct ifnet * ifp,struct rtentry * rt0,struct mbuf * m,struct sockaddr * dst,u_char * desten)1270 nd6_resolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
1271 struct sockaddr *dst, u_char *desten)
1272 {
1273 struct sockaddr_dl *sdl;
1274 struct rtentry *rt;
1275 struct llinfo_nd6 *ln;
1276 struct in6_addr saddr6;
1277 time_t uptime;
1278 int solicit = 0;
1279
1280 if (m->m_flags & M_MCAST) {
1281 ETHER_MAP_IPV6_MULTICAST(&satosin6(dst)->sin6_addr, desten);
1282 return (0);
1283 }
1284
1285 uptime = getuptime();
1286 rt = rt_getll(rt0);
1287
1288 if (ISSET(rt->rt_flags, RTF_REJECT) &&
1289 (rt->rt_expire == 0 || rt->rt_expire > uptime)) {
1290 m_freem(m);
1291 return (rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
1292 }
1293
1294 /*
1295 * Address resolution or Neighbor Unreachability Detection
1296 * for the next hop.
1297 * At this point, the destination of the packet must be a unicast
1298 * or an anycast address(i.e. not a multicast).
1299 */
1300 if (!ISSET(rt->rt_flags, RTF_LLINFO)) {
1301 char addr[INET6_ADDRSTRLEN];
1302 log(LOG_DEBUG, "%s: %s: route contains no ND information\n",
1303 __func__, inet_ntop(AF_INET6,
1304 &satosin6(rt_key(rt))->sin6_addr, addr, sizeof(addr)));
1305 goto bad;
1306 }
1307
1308 if (rt->rt_gateway->sa_family != AF_LINK) {
1309 printf("%s: something odd happens\n", __func__);
1310 goto bad;
1311 }
1312
1313 mtx_enter(&nd6_mtx);
1314 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1315 if (ln == NULL) {
1316 mtx_leave(&nd6_mtx);
1317 goto bad;
1318 }
1319
1320 /*
1321 * Move this entry to the head of the queue so that it is less likely
1322 * for this entry to be a target of forced garbage collection (see
1323 * nd6_rtrequest()).
1324 */
1325 TAILQ_REMOVE(&nd6_list, ln, ln_list);
1326 TAILQ_INSERT_HEAD(&nd6_list, ln, ln_list);
1327
1328 /*
1329 * The first time we send a packet to a neighbor whose entry is
1330 * STALE, we have to change the state to DELAY and set a timer to
1331 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure we do
1332 * neighbor unreachability detection on expiration.
1333 * (RFC 2461 7.3.3)
1334 */
1335 if (ln->ln_state == ND6_LLINFO_STALE) {
1336 ln->ln_asked = 0;
1337 ln->ln_state = ND6_LLINFO_DELAY;
1338 nd6_llinfo_settimer(ln, nd6_delay);
1339 }
1340
1341 /*
1342 * If the neighbor cache entry has a state other than INCOMPLETE
1343 * (i.e. its link-layer address is already resolved), just
1344 * send the packet.
1345 */
1346 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) {
1347 mtx_leave(&nd6_mtx);
1348
1349 sdl = satosdl(rt->rt_gateway);
1350 if (sdl->sdl_alen != ETHER_ADDR_LEN) {
1351 char addr[INET6_ADDRSTRLEN];
1352 log(LOG_DEBUG, "%s: %s: incorrect nd6 information\n",
1353 __func__,
1354 inet_ntop(AF_INET6, &satosin6(dst)->sin6_addr,
1355 addr, sizeof(addr)));
1356 goto bad;
1357 }
1358
1359 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
1360 return (0);
1361 }
1362
1363 /*
1364 * There is a neighbor cache entry, but no ethernet address
1365 * response yet. Insert mbuf in hold queue if below limit.
1366 * If above the limit free the queue without queuing the new packet.
1367 */
1368 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1369 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1370 /* source address of prompting packet is needed by nd6_ns_output() */
1371 if (m->m_len >= sizeof(struct ip6_hdr)) {
1372 memcpy(&ln->ln_saddr6, &mtod(m, struct ip6_hdr *)->ip6_src,
1373 sizeof(ln->ln_saddr6));
1374 }
1375 if (atomic_inc_int_nv(&ln_hold_total) <= LN_HOLD_TOTAL) {
1376 if (mq_push(&ln->ln_mq, m) != 0)
1377 atomic_dec_int(&ln_hold_total);
1378 } else {
1379 atomic_sub_int(&ln_hold_total, mq_purge(&ln->ln_mq) + 1);
1380 m_freem(m);
1381 }
1382
1383 /*
1384 * If there has been no NS for the neighbor after entering the
1385 * INCOMPLETE state, send the first solicitation.
1386 */
1387 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
1388 ln->ln_asked++;
1389 nd6_llinfo_settimer(ln, RETRANS_TIMER / 1000);
1390 saddr6 = ln->ln_saddr6;
1391 solicit = 1;
1392 }
1393 mtx_leave(&nd6_mtx);
1394
1395 if (solicit)
1396 nd6_ns_output(ifp, NULL, &satosin6(dst)->sin6_addr, &saddr6, 0);
1397 return (EAGAIN);
1398
1399 bad:
1400 m_freem(m);
1401 return (EINVAL);
1402 }
1403
1404 int
nd6_need_cache(struct ifnet * ifp)1405 nd6_need_cache(struct ifnet *ifp)
1406 {
1407 /*
1408 * RFC2893 says:
1409 * - unidirectional tunnels needs no ND
1410 */
1411 switch (ifp->if_type) {
1412 case IFT_ETHER:
1413 case IFT_IEEE80211:
1414 case IFT_CARP:
1415 return (1);
1416 default:
1417 return (0);
1418 }
1419 }
1420