1/* $FreeBSD: head/sys/netinet6/nd6.c 173018 2007-10-26 13:18:38Z rwatson $ */
2/* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 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 "opt_inet.h"
34#include "opt_inet6.h"
35#include "opt_mac.h"
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/callout.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/socket.h>
43#include <sys/sockio.h>
44#include <sys/time.h>
45#include <sys/kernel.h>
46#include <sys/protosw.h>
47#include <sys/errno.h>
48#include <sys/syslog.h>
49#include <sys/queue.h>
50#include <sys/sysctl.h>
51
52#include <net/if.h>
53#include <net/if_arc.h>
54#include <net/if_dl.h>
55#include <net/if_types.h>
56#include <net/iso88025.h>
57#include <net/fddi.h>
58#include <net/route.h>
59
60#include <netinet/in.h>
61#include <netinet/if_ether.h>
62#include <netinet6/in6_var.h>
63#include <netinet/ip6.h>
64#include <netinet6/ip6_var.h>
65#include <netinet6/scope6_var.h>
66#include <netinet6/nd6.h>
67#include <netinet/icmp6.h>
68
69#include <sys/limits.h>
70
71#include <security/mac/mac_framework.h>
72
73#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
74#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
75
76#define SIN6(s) ((struct sockaddr_in6 *)s)
77#define SDL(s) ((struct sockaddr_dl *)s)
78
79/* timer values */
80int nd6_prune = 1; /* walk list every 1 seconds */
81int nd6_delay = 5; /* delay first probe time 5 second */
82int nd6_umaxtries = 3; /* maximum unicast query */
83int nd6_mmaxtries = 3; /* maximum multicast query */
84int nd6_useloopback = 1; /* use loopback interface for local traffic */
85int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
86
87/* preventing too many loops in ND option parsing */
88int nd6_maxndopt = 10; /* max # of ND options allowed */
89
90int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
91int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */
92
93#ifdef ND6_DEBUG
94int nd6_debug = 1;
95#else
96int nd6_debug = 0;
97#endif
98
99/* for debugging? */
100static int nd6_inuse, nd6_allocated;
101
102struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
103struct nd_drhead nd_defrouter;
104struct nd_prhead nd_prefix = { 0 };
105
106int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
107static struct sockaddr_in6 all1_sa;
108
109static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
110 struct ifnet *));
111static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *));
112static void nd6_slowtimo __P((void *));
113static int regen_tmpaddr __P((struct in6_ifaddr *));
114static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int));
115static void nd6_llinfo_timer __P((void *));
116static void clear_llinfo_pqueue __P((struct llinfo_nd6 *));
117
118struct callout nd6_slowtimo_ch;
119struct callout nd6_timer_ch;
120extern struct callout in6_tmpaddrtimer_ch;
121
122void
123nd6_init(void)
124{
125 static int nd6_init_done = 0;
126 int i;
127
128 if (nd6_init_done) {
129 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
130 return;
131 }
132
133 all1_sa.sin6_family = AF_INET6;
134 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
135 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
136 all1_sa.sin6_addr.s6_addr[i] = 0xff;
137
138 /* initialization of the default router list */
139 TAILQ_INIT(&nd_defrouter);
140
141 nd6_init_done = 1;
142
143 /* start timer */
144 callout_init(&nd6_slowtimo_ch, 0);
145 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
146 nd6_slowtimo, NULL);
147}
148
149struct nd_ifinfo *
150nd6_ifattach(struct ifnet *ifp)
151{
152 struct nd_ifinfo *nd;
153
154 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
155 bzero(nd, sizeof(*nd));
156
157 nd->initialized = 1;
158
159 nd->chlim = IPV6_DEFHLIM;
160 nd->basereachable = REACHABLE_TIME;
161 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
162 nd->retrans = RETRANS_TIMER;
163 /*
164 * Note that the default value of ip6_accept_rtadv is 0, which means
165 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
166 * here.
167 */
168 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
169
170 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
171 nd6_setmtu0(ifp, nd);
172
173 return nd;
174}
175
176void
177nd6_ifdetach(struct nd_ifinfo *nd)
178{
179
180 free(nd, M_IP6NDP);
181}
182
183/*
184 * Reset ND level link MTU. This function is called when the physical MTU
185 * changes, which means we might have to adjust the ND level MTU.
186 */
187void
188nd6_setmtu(struct ifnet *ifp)
189{
190
191 nd6_setmtu0(ifp, ND_IFINFO(ifp));
192}
193
194/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
195void
196nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
197{
198 u_int32_t omaxmtu;
199
200 omaxmtu = ndi->maxmtu;
201
202 switch (ifp->if_type) {
203 case IFT_ARCNET:
204 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
205 break;
206 case IFT_FDDI:
207 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
208 break;
209 case IFT_ISO88025:
210 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
211 break;
212 default:
213 ndi->maxmtu = ifp->if_mtu;
214 break;
215 }
216
217 /*
218 * Decreasing the interface MTU under IPV6 minimum MTU may cause
219 * undesirable situation. We thus notify the operator of the change
220 * explicitly. The check for omaxmtu is necessary to restrict the
221 * log to the case of changing the MTU, not initializing it.
222 */
223 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
224 log(LOG_NOTICE, "nd6_setmtu0: "
225 "new link MTU on %s (%lu) is too small for IPv6\n",
226 if_name(ifp), (unsigned long)ndi->maxmtu);
227 }
228
229 if (ndi->maxmtu > in6_maxmtu)
230 in6_setmaxmtu(); /* check all interfaces just in case */
231
232#undef MIN
233}
234
235void
236nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
237{
238
239 bzero(ndopts, sizeof(*ndopts));
240 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
241 ndopts->nd_opts_last
242 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
243
244 if (icmp6len == 0) {
245 ndopts->nd_opts_done = 1;
246 ndopts->nd_opts_search = NULL;
247 }
248}
249
250/*
251 * Take one ND option.
252 */
253struct nd_opt_hdr *
254nd6_option(union nd_opts *ndopts)
255{
256 struct nd_opt_hdr *nd_opt;
257 int olen;
258
259 if (ndopts == NULL)
260 panic("ndopts == NULL in nd6_option");
261 if (ndopts->nd_opts_last == NULL)
262 panic("uninitialized ndopts in nd6_option");
263 if (ndopts->nd_opts_search == NULL)
264 return NULL;
265 if (ndopts->nd_opts_done)
266 return NULL;
267
268 nd_opt = ndopts->nd_opts_search;
269
270 /* make sure nd_opt_len is inside the buffer */
271 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
272 bzero(ndopts, sizeof(*ndopts));
273 return NULL;
274 }
275
276 olen = nd_opt->nd_opt_len << 3;
277 if (olen == 0) {
278 /*
279 * Message validation requires that all included
280 * options have a length that is greater than zero.
281 */
282 bzero(ndopts, sizeof(*ndopts));
283 return NULL;
284 }
285
286 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
287 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
288 /* option overruns the end of buffer, invalid */
289 bzero(ndopts, sizeof(*ndopts));
290 return NULL;
291 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
292 /* reached the end of options chain */
293 ndopts->nd_opts_done = 1;
294 ndopts->nd_opts_search = NULL;
295 }
296 return nd_opt;
297}
298
299/*
300 * Parse multiple ND options.
301 * This function is much easier to use, for ND routines that do not need
302 * multiple options of the same type.
303 */
304int
305nd6_options(union nd_opts *ndopts)
306{
307 struct nd_opt_hdr *nd_opt;
308 int i = 0;
309
310 if (ndopts == NULL)
311 panic("ndopts == NULL in nd6_options");
312 if (ndopts->nd_opts_last == NULL)
313 panic("uninitialized ndopts in nd6_options");
314 if (ndopts->nd_opts_search == NULL)
315 return 0;
316
317 while (1) {
318 nd_opt = nd6_option(ndopts);
319 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
320 /*
321 * Message validation requires that all included
322 * options have a length that is greater than zero.
323 */
324 icmp6stat.icp6s_nd_badopt++;
325 bzero(ndopts, sizeof(*ndopts));
326 return -1;
327 }
328
329 if (nd_opt == NULL)
330 goto skip1;
331
332 switch (nd_opt->nd_opt_type) {
333 case ND_OPT_SOURCE_LINKADDR:
334 case ND_OPT_TARGET_LINKADDR:
335 case ND_OPT_MTU:
336 case ND_OPT_REDIRECTED_HEADER:
337 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
338 nd6log((LOG_INFO,
339 "duplicated ND6 option found (type=%d)\n",
340 nd_opt->nd_opt_type));
341 /* XXX bark? */
342 } else {
343 ndopts->nd_opt_array[nd_opt->nd_opt_type]
344 = nd_opt;
345 }
346 break;
347 case ND_OPT_PREFIX_INFORMATION:
348 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
349 ndopts->nd_opt_array[nd_opt->nd_opt_type]
350 = nd_opt;
351 }
352 ndopts->nd_opts_pi_end =
353 (struct nd_opt_prefix_info *)nd_opt;
354 break;
355 default:
356 /*
357 * Unknown options must be silently ignored,
358 * to accomodate future extension to the protocol.
359 */
360 nd6log((LOG_DEBUG,
361 "nd6_options: unsupported option %d - "
362 "option ignored\n", nd_opt->nd_opt_type));
363 }
364
365skip1:
366 i++;
367 if (i > nd6_maxndopt) {
368 icmp6stat.icp6s_nd_toomanyopt++;
369 nd6log((LOG_INFO, "too many loop in nd opt\n"));
370 break;
371 }
372
373 if (ndopts->nd_opts_done)
374 break;
375 }
376
377 return 0;
378}
379
380/*
381 * ND6 timer routine to handle ND6 entries
382 */
383void
384nd6_llinfo_settimer(struct llinfo_nd6 *ln, long tick)
385{
386 if (tick < 0) {
387 ln->ln_expire = 0;
388 ln->ln_ntick = 0;
389 callout_stop(&ln->ln_timer_ch);
390 } else {
391 ln->ln_expire = time_second + tick / hz;
392 if (tick > INT_MAX) {
393 ln->ln_ntick = tick - INT_MAX;
394 callout_reset(&ln->ln_timer_ch, INT_MAX,
395 nd6_llinfo_timer, ln);
396 } else {
397 ln->ln_ntick = 0;
398 callout_reset(&ln->ln_timer_ch, tick,
399 nd6_llinfo_timer, ln);
400 }
401 }
402}
403
404static void
405nd6_llinfo_timer(void *arg)
406{
407 struct llinfo_nd6 *ln;
408 struct rtentry *rt;
409 struct in6_addr *dst;
410 struct ifnet *ifp;
411 struct nd_ifinfo *ndi = NULL;
412
413 ln = (struct llinfo_nd6 *)arg;
414
415 if (ln->ln_ntick > 0) {
416 if (ln->ln_ntick > INT_MAX) {
417 ln->ln_ntick -= INT_MAX;
418 nd6_llinfo_settimer(ln, INT_MAX);
419 } else {
420 ln->ln_ntick = 0;
421 nd6_llinfo_settimer(ln, ln->ln_ntick);
422 }
423 return;
424 }
425
426 if ((rt = ln->ln_rt) == NULL)
427 panic("ln->ln_rt == NULL");
428 if ((ifp = rt->rt_ifp) == NULL)
429 panic("ln->ln_rt->rt_ifp == NULL");
430 ndi = ND_IFINFO(ifp);
431
432 /* sanity check */
433 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
434 panic("rt_llinfo(%p) is not equal to ln(%p)",
435 rt->rt_llinfo, ln);
436 if (rt_key(rt) == NULL)
437 panic("rt key is NULL in nd6_timer(ln=%p)", ln);
438
439 dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
440
441 switch (ln->ln_state) {
442 case ND6_LLINFO_INCOMPLETE:
443 if (ln->ln_asked < nd6_mmaxtries) {
444 ln->ln_asked++;
445 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
446 nd6_ns_output(ifp, NULL, dst, ln, 0);
447 } else {
448 struct mbuf *m = ln->ln_hold;
449 if (m) {
450 struct mbuf *m0;
451
452 /*
453 * assuming every packet in ln_hold has the
454 * same IP header
455 */
456 m0 = m->m_nextpkt;
457 m->m_nextpkt = NULL;
458 icmp6_error2(m, ICMP6_DST_UNREACH,
459 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp);
460
461 ln->ln_hold = m0;
462 clear_llinfo_pqueue(ln);
463 }
464 if (rt && rt->rt_llinfo)
465 (void)nd6_free(rt, 0);
466 ln = NULL;
467 }
468 break;
469 case ND6_LLINFO_REACHABLE:
470 if (!ND6_LLINFO_PERMANENT(ln)) {
471 ln->ln_state = ND6_LLINFO_STALE;
472 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
473 }
474 break;
475
476 case ND6_LLINFO_STALE:
477 /* Garbage Collection(RFC 2461 5.3) */
478 if (!ND6_LLINFO_PERMANENT(ln)) {
479 if (rt && rt->rt_llinfo)
480 (void)nd6_free(rt, 1);
481 ln = NULL;
482 }
483 break;
484
485 case ND6_LLINFO_DELAY:
486 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
487 /* We need NUD */
488 ln->ln_asked = 1;
489 ln->ln_state = ND6_LLINFO_PROBE;
490 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
491 nd6_ns_output(ifp, dst, dst, ln, 0);
492 } else {
493 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
494 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
495 }
496 break;
497 case ND6_LLINFO_PROBE:
498 if (ln->ln_asked < nd6_umaxtries) {
499 ln->ln_asked++;
500 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
501 nd6_ns_output(ifp, dst, dst, ln, 0);
502 } else if (rt->rt_ifa != NULL &&
503 rt->rt_ifa->ifa_addr->sa_family == AF_INET6 &&
504 (((struct in6_ifaddr *)rt->rt_ifa)->ia_flags & IFA_ROUTE)) {
505 /*
506 * This is an unreachable neighbor whose address is
507 * specified as the destination of a p2p interface
508 * (see in6_ifinit()). We should not free the entry
509 * since this is sort of a "static" entry generated
510 * via interface address configuration.
511 */
512 ln->ln_asked = 0;
513 ln->ln_expire = 0; /* make it permanent */
514 ln->ln_state = ND6_LLINFO_STALE;
515 } else {
516 if (rt && rt->rt_llinfo)
517 (void)nd6_free(rt, 0);
518 ln = NULL;
519 }
520 break;
521 }
522}
523
524
525/*
526 * ND6 timer routine to expire default route list and prefix list
527 */
528void
529nd6_timer(void *ignored_arg)
530{
531 int s;
532 struct nd_defrouter *dr;
533 struct nd_prefix *pr;
534 struct in6_ifaddr *ia6, *nia6;
535 struct in6_addrlifetime *lt6;
536
537 callout_reset(&nd6_timer_ch, nd6_prune * hz,
538 nd6_timer, NULL);
539
540 /* expire default router list */
541 s = splnet();
542 dr = TAILQ_FIRST(&nd_defrouter);
543 while (dr) {
544 if (dr->expire && dr->expire < time_second) {
545 struct nd_defrouter *t;
546 t = TAILQ_NEXT(dr, dr_entry);
547 defrtrlist_del(dr);
548 dr = t;
549 } else {
550 dr = TAILQ_NEXT(dr, dr_entry);
551 }
552 }
553
554 /*
555 * expire interface addresses.
556 * in the past the loop was inside prefix expiry processing.
557 * However, from a stricter speci-confrmance standpoint, we should
558 * rather separate address lifetimes and prefix lifetimes.
559 */
560 addrloop:
561 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
562 nia6 = ia6->ia_next;
563 /* check address lifetime */
564 lt6 = &ia6->ia6_lifetime;
565 if (IFA6_IS_INVALID(ia6)) {
566 int regen = 0;
567
568 /*
569 * If the expiring address is temporary, try
570 * regenerating a new one. This would be useful when
571 * we suspended a laptop PC, then turned it on after a
572 * period that could invalidate all temporary
573 * addresses. Although we may have to restart the
574 * loop (see below), it must be after purging the
575 * address. Otherwise, we'd see an infinite loop of
576 * regeneration.
577 */
578 if (ip6_use_tempaddr &&
579 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
580 if (regen_tmpaddr(ia6) == 0)
581 regen = 1;
582 }
583
584 in6_purgeaddr(&ia6->ia_ifa);
585
586 if (regen)
587 goto addrloop; /* XXX: see below */
588 } else if (IFA6_IS_DEPRECATED(ia6)) {
589 int oldflags = ia6->ia6_flags;
590
591 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
592
593 /*
594 * If a temporary address has just become deprecated,
595 * regenerate a new one if possible.
596 */
597 if (ip6_use_tempaddr &&
598 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
599 (oldflags & IN6_IFF_DEPRECATED) == 0) {
600
601 if (regen_tmpaddr(ia6) == 0) {
602 /*
603 * A new temporary address is
604 * generated.
605 * XXX: this means the address chain
606 * has changed while we are still in
607 * the loop. Although the change
608 * would not cause disaster (because
609 * it's not a deletion, but an
610 * addition,) we'd rather restart the
611 * loop just for safety. Or does this
612 * significantly reduce performance??
613 */
614 goto addrloop;
615 }
616 }
617 } else {
618 /*
619 * A new RA might have made a deprecated address
620 * preferred.
621 */
622 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
623 }
624 }
625
626 /* expire prefix list */
627 pr = nd_prefix.lh_first;
628 while (pr) {
629 /*
630 * check prefix lifetime.
631 * since pltime is just for autoconf, pltime processing for
632 * prefix is not necessary.
633 */
634 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
635 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
636 struct nd_prefix *t;
637 t = pr->ndpr_next;
638
639 /*
640 * address expiration and prefix expiration are
641 * separate. NEVER perform in6_purgeaddr here.
642 */
643
644 prelist_remove(pr);
645 pr = t;
646 } else
647 pr = pr->ndpr_next;
648 }
649 splx(s);
650}
651
652/*
653 * ia6 - deprecated/invalidated temporary address
654 */
655static int
656regen_tmpaddr(struct in6_ifaddr *ia6)
657{
658 struct ifaddr *ifa;
659 struct ifnet *ifp;
660 struct in6_ifaddr *public_ifa6 = NULL;
661
662 ifp = ia6->ia_ifa.ifa_ifp;
663 for (ifa = ifp->if_addrlist.tqh_first; ifa;
664 ifa = ifa->ifa_list.tqe_next) {
665 struct in6_ifaddr *it6;
666
667 if (ifa->ifa_addr->sa_family != AF_INET6)
668 continue;
669
670 it6 = (struct in6_ifaddr *)ifa;
671
672 /* ignore no autoconf addresses. */
673 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
674 continue;
675
676 /* ignore autoconf addresses with different prefixes. */
677 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
678 continue;
679
680 /*
681 * Now we are looking at an autoconf address with the same
682 * prefix as ours. If the address is temporary and is still
683 * preferred, do not create another one. It would be rare, but
684 * could happen, for example, when we resume a laptop PC after
685 * a long period.
686 */
687 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
688 !IFA6_IS_DEPRECATED(it6)) {
689 public_ifa6 = NULL;
690 break;
691 }
692
693 /*
694 * This is a public autoconf address that has the same prefix
695 * as ours. If it is preferred, keep it. We can't break the
696 * loop here, because there may be a still-preferred temporary
697 * address with the prefix.
698 */
699 if (!IFA6_IS_DEPRECATED(it6))
700 public_ifa6 = it6;
701 }
702
703 if (public_ifa6 != NULL) {
704 int e;
705
706 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
707 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
708 " tmp addr,errno=%d\n", e);
709 return (-1);
710 }
711 return (0);
712 }
713
714 return (-1);
715}
716
717/*
718 * Nuke neighbor cache/prefix/default router management table, right before
719 * ifp goes away.
720 */
721void
722nd6_purge(struct ifnet *ifp)
723{
724 struct llinfo_nd6 *ln, *nln;
725 struct nd_defrouter *dr, *ndr;
726 struct nd_prefix *pr, *npr;
727
728 /*
729 * Nuke default router list entries toward ifp.
730 * We defer removal of default router list entries that is installed
731 * in the routing table, in order to keep additional side effects as
732 * small as possible.
733 */
734 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
735 ndr = TAILQ_NEXT(dr, dr_entry);
736 if (dr->installed)
737 continue;
738
739 if (dr->ifp == ifp)
740 defrtrlist_del(dr);
741 }
742
743 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
744 ndr = TAILQ_NEXT(dr, dr_entry);
745 if (!dr->installed)
746 continue;
747
748 if (dr->ifp == ifp)
749 defrtrlist_del(dr);
750 }
751
752 /* Nuke prefix list entries toward ifp */
753 for (pr = nd_prefix.lh_first; pr; pr = npr) {
754 npr = pr->ndpr_next;
755 if (pr->ndpr_ifp == ifp) {
756 /*
757 * Because if_detach() does *not* release prefixes
758 * while purging addresses the reference count will
759 * still be above zero. We therefore reset it to
760 * make sure that the prefix really gets purged.
761 */
762 pr->ndpr_refcnt = 0;
763
764 /*
765 * Previously, pr->ndpr_addr is removed as well,
766 * but I strongly believe we don't have to do it.
767 * nd6_purge() is only called from in6_ifdetach(),
768 * which removes all the associated interface addresses
769 * by itself.
770 * (jinmei@kame.net 20010129)
771 */
772 prelist_remove(pr);
773 }
774 }
775
776 /* cancel default outgoing interface setting */
777 if (nd6_defifindex == ifp->if_index)
778 nd6_setdefaultiface(0);
779
780 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
781 /* refresh default router list */
782 defrouter_select();
783 }
784
785 /*
786 * Nuke neighbor cache entries for the ifp.
787 * Note that rt->rt_ifp may not be the same as ifp,
788 * due to KAME goto ours hack. See RTM_RESOLVE case in
789 * nd6_rtrequest(), and ip6_input().
790 */
791 ln = llinfo_nd6.ln_next;
792 while (ln && ln != &llinfo_nd6) {
793 struct rtentry *rt;
794 struct sockaddr_dl *sdl;
795
796 nln = ln->ln_next;
797 rt = ln->ln_rt;
798 if (rt && rt->rt_gateway &&
799 rt->rt_gateway->sa_family == AF_LINK) {
800 sdl = (struct sockaddr_dl *)rt->rt_gateway;
801 if (sdl->sdl_index == ifp->if_index)
802 nln = nd6_free(rt, 0);
803 }
804 ln = nln;
805 }
806}
807
808struct rtentry *
809nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp)
810{
811 struct rtentry *rt;
812 struct sockaddr_in6 sin6;
813 char ip6buf[INET6_ADDRSTRLEN];
814
815 bzero(&sin6, sizeof(sin6));
816 sin6.sin6_len = sizeof(struct sockaddr_in6);
817 sin6.sin6_family = AF_INET6;
818 sin6.sin6_addr = *addr6;
819 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
820 if (rt) {
821 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) {
822 /*
823 * This is the case for the default route.
824 * If we want to create a neighbor cache for the
825 * address, we should free the route for the
826 * destination and allocate an interface route.
827 */
828 RTFREE_LOCKED(rt);
829 rt = NULL;
830 }
831 }
832 if (rt == NULL) {
833 if (create && ifp) {
834 int e;
835
836 /*
837 * If no route is available and create is set,
838 * we allocate a host route for the destination
839 * and treat it like an interface route.
840 * This hack is necessary for a neighbor which can't
841 * be covered by our own prefix.
842 */
843 struct ifaddr *ifa =
844 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
845 if (ifa == NULL)
846 return (NULL);
847
848 /*
849 * Create a new route. RTF_LLINFO is necessary
850 * to create a Neighbor Cache entry for the
851 * destination in nd6_rtrequest which will be
852 * called in rtrequest via ifa->ifa_rtrequest.
853 */
854 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
855 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
856 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
857 ~RTF_CLONING, &rt)) != 0) {
858 log(LOG_ERR,
859 "nd6_lookup: failed to add route for a "
860 "neighbor(%s), errno=%d\n",
861 ip6_sprintf(ip6buf, addr6), e);
862 }
863 if (rt == NULL)
864 return (NULL);
865 RT_LOCK(rt);
866 if (rt->rt_llinfo) {
867 struct llinfo_nd6 *ln =
868 (struct llinfo_nd6 *)rt->rt_llinfo;
869 ln->ln_state = ND6_LLINFO_NOSTATE;
870 }
871 } else
872 return (NULL);
873 }
874 RT_LOCK_ASSERT(rt);
875 RT_REMREF(rt);
876 /*
877 * Validation for the entry.
878 * Note that the check for rt_llinfo is necessary because a cloned
879 * route from a parent route that has the L flag (e.g. the default
880 * route to a p2p interface) may have the flag, too, while the
881 * destination is not actually a neighbor.
882 * XXX: we can't use rt->rt_ifp to check for the interface, since
883 * it might be the loopback interface if the entry is for our
884 * own address on a non-loopback interface. Instead, we should
885 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
886 * interface.
887 * Note also that ifa_ifp and ifp may differ when we connect two
888 * interfaces to a same link, install a link prefix to an interface,
889 * and try to install a neighbor cache on an interface that does not
890 * have a route to the prefix.
891 */
892 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
893 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
894 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
895 if (create) {
896 nd6log((LOG_DEBUG,
897 "nd6_lookup: failed to lookup %s (if = %s)\n",
898 ip6_sprintf(ip6buf, addr6),
899 ifp ? if_name(ifp) : "unspec"));
900 }
901 RT_UNLOCK(rt);
902 return (NULL);
903 }
904 RT_UNLOCK(rt); /* XXX not ready to return rt locked */
905 return (rt);
906}
907
908/*
909 * Test whether a given IPv6 address is a neighbor or not, ignoring
910 * the actual neighbor cache. The neighbor cache is ignored in order
911 * to not reenter the routing code from within itself.
912 */
913static int
914nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
915{
916 struct nd_prefix *pr;
917 struct ifaddr *dstaddr;
918
919 /*
920 * A link-local address is always a neighbor.
921 * XXX: a link does not necessarily specify a single interface.
922 */
923 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
924 struct sockaddr_in6 sin6_copy;
925 u_int32_t zone;
926
927 /*
928 * We need sin6_copy since sa6_recoverscope() may modify the
929 * content (XXX).
930 */
931 sin6_copy = *addr;
932 if (sa6_recoverscope(&sin6_copy))
933 return (0); /* XXX: should be impossible */
934 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
935 return (0);
936 if (sin6_copy.sin6_scope_id == zone)
937 return (1);
938 else
939 return (0);
940 }
941
942 /*
943 * If the address matches one of our addresses,
944 * it should be a neighbor.
945 * If the address matches one of our on-link prefixes, it should be a
946 * neighbor.
947 */
948 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
949 if (pr->ndpr_ifp != ifp)
950 continue;
951
952 if (!(pr->ndpr_stateflags & NDPRF_ONLINK))
953 continue;
954
955 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
956 &addr->sin6_addr, &pr->ndpr_mask))
957 return (1);
958 }
959
960 /*
961 * If the address is assigned on the node of the other side of
962 * a p2p interface, the address should be a neighbor.
963 */
964 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
965 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp))
966 return (1);
967
968 /*
969 * If the default router list is empty, all addresses are regarded
970 * as on-link, and thus, as a neighbor.
971 * XXX: we restrict the condition to hosts, because routers usually do
972 * not have the "default router list".
973 */
974 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL &&
975 nd6_defifindex == ifp->if_index) {
976 return (1);
977 }
978
979 return (0);
980}
981
982
983/*
984 * Detect if a given IPv6 address identifies a neighbor on a given link.
985 * XXX: should take care of the destination of a p2p link?
986 */
987int
988nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
989{
990
991 if (nd6_is_new_addr_neighbor(addr, ifp))
992 return (1);
993
994 /*
995 * Even if the address matches none of our addresses, it might be
996 * in the neighbor cache.
997 */
998 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
999 return (1);
1000
1001 return (0);
1002}
1003
1004/*
1005 * Free an nd6 llinfo entry.
1006 * Since the function would cause significant changes in the kernel, DO NOT
1007 * make it global, unless you have a strong reason for the change, and are sure
1008 * that the change is safe.
1009 */
1010static struct llinfo_nd6 *
1011nd6_free(struct rtentry *rt, int gc)
1012{
1013 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1014 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
1015 struct nd_defrouter *dr;
1016
1017 /*
1018 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1019 * even though it is not harmful, it was not really necessary.
1020 */
1021
1022 /* cancel timer */
1023 nd6_llinfo_settimer(ln, -1);
1024
1025 if (!ip6_forwarding) {
1026 int s;
1027 s = splnet();
1028 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1029 rt->rt_ifp);
1030
1031 if (dr != NULL && dr->expire &&
1032 ln->ln_state == ND6_LLINFO_STALE && gc) {
1033 /*
1034 * If the reason for the deletion is just garbage
1035 * collection, and the neighbor is an active default
1036 * router, do not delete it. Instead, reset the GC
1037 * timer using the router's lifetime.
1038 * Simply deleting the entry would affect default
1039 * router selection, which is not necessarily a good
1040 * thing, especially when we're using router preference
1041 * values.
1042 * XXX: the check for ln_state would be redundant,
1043 * but we intentionally keep it just in case.
1044 */
1045 if (dr->expire > time_second)
1046 nd6_llinfo_settimer(ln,
1047 (dr->expire - time_second) * hz);
1048 else
1049 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1050 splx(s);
1051 return (ln->ln_next);
1052 }
1053
1054 if (ln->ln_router || dr) {
1055 /*
1056 * rt6_flush must be called whether or not the neighbor
1057 * is in the Default Router List.
1058 * See a corresponding comment in nd6_na_input().
1059 */
1060 rt6_flush(&in6, rt->rt_ifp);
1061 }
1062
1063 if (dr) {
1064 /*
1065 * Unreachablity of a router might affect the default
1066 * router selection and on-link detection of advertised
1067 * prefixes.
1068 */
1069
1070 /*
1071 * Temporarily fake the state to choose a new default
1072 * router and to perform on-link determination of
1073 * prefixes correctly.
1074 * Below the state will be set correctly,
1075 * or the entry itself will be deleted.
1076 */
1077 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1078
1079 /*
1080 * Since defrouter_select() does not affect the
1081 * on-link determination and MIP6 needs the check
1082 * before the default router selection, we perform
1083 * the check now.
1084 */
1085 pfxlist_onlink_check();
1086
1087 /*
1088 * refresh default router list
1089 */
1090 defrouter_select();
1091 }
1092 splx(s);
1093 }
1094
1095 /*
1096 * Before deleting the entry, remember the next entry as the
1097 * return value. We need this because pfxlist_onlink_check() above
1098 * might have freed other entries (particularly the old next entry) as
1099 * a side effect (XXX).
1100 */
1101 next = ln->ln_next;
1102
1103 /*
1104 * Detach the route from the routing tree and the list of neighbor
1105 * caches, and disable the route entry not to be used in already
1106 * cached routes.
1107 */
1108 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1109 rt_mask(rt), 0, (struct rtentry **)0);
1110
1111 return (next);
1112}
1113
1114/*
1115 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1116 *
1117 * XXX cost-effective methods?
1118 */
1119void
1120nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1121{
1122 struct llinfo_nd6 *ln;
1123
1124 /*
1125 * If the caller specified "rt", use that. Otherwise, resolve the
1126 * routing table by supplied "dst6".
1127 */
1128 if (rt == NULL) {
1129 if (dst6 == NULL)
1130 return;
1131 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL)
1132 return;
1133 }
1134
1135 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1136 (rt->rt_flags & RTF_LLINFO) == 0 ||
1137 rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
1138 rt->rt_gateway->sa_family != AF_LINK) {
1139 /* This is not a host route. */
1140 return;
1141 }
1142
1143 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1144 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1145 return;
1146
1147 /*
1148 * if we get upper-layer reachability confirmation many times,
1149 * it is possible we have false information.
1150 */
1151 if (!force) {
1152 ln->ln_byhint++;
1153 if (ln->ln_byhint > nd6_maxnudhint)
1154 return;
1155 }
1156
1157 ln->ln_state = ND6_LLINFO_REACHABLE;
1158 if (!ND6_LLINFO_PERMANENT(ln)) {
1159 nd6_llinfo_settimer(ln,
1160 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1161 }
1162}
1163
1164/*
1165 * info - XXX unused
1166 */
1167void
1168nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1169{
1170 struct sockaddr *gate = rt->rt_gateway;
1171 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1172 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1173 struct ifnet *ifp = rt->rt_ifp;
1174 struct ifaddr *ifa;
1175
1176 RT_LOCK_ASSERT(rt);
1177
1178 if ((rt->rt_flags & RTF_GATEWAY) != 0)
1179 return;
1180
1181 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1182 /*
1183 * This is probably an interface direct route for a link
1184 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1185 * We do not need special treatment below for such a route.
1186 * Moreover, the RTF_LLINFO flag which would be set below
1187 * would annoy the ndp(8) command.
1188 */
1189 return;
1190 }
1191
1192 if (req == RTM_RESOLVE &&
1193 (nd6_need_cache(ifp) == 0 || /* stf case */
1194 !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt),
1195 ifp))) {
1196 /*
1197 * FreeBSD and BSD/OS often make a cloned host route based
1198 * on a less-specific route (e.g. the default route).
1199 * If the less specific route does not have a "gateway"
1200 * (this is the case when the route just goes to a p2p or an
1201 * stf interface), we'll mistakenly make a neighbor cache for
1202 * the host route, and will see strange neighbor solicitation
1203 * for the corresponding destination. In order to avoid the
1204 * confusion, we check if the destination of the route is
1205 * a neighbor in terms of neighbor discovery, and stop the
1206 * process if not. Additionally, we remove the LLINFO flag
1207 * so that ndp(8) will not try to get the neighbor information
1208 * of the destination.
1209 */
1210 rt->rt_flags &= ~RTF_LLINFO;
1211 return;
1212 }
1213
1214 switch (req) {
1215 case RTM_ADD:
1216 /*
1217 * There is no backward compatibility :)
1218 *
1219 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1220 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1221 * rt->rt_flags |= RTF_CLONING;
1222 */
1223 if ((rt->rt_flags & RTF_CLONING) ||
1224 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) {
1225 /*
1226 * Case 1: This route should come from a route to
1227 * interface (RTF_CLONING case) or the route should be
1228 * treated as on-link but is currently not
1229 * (RTF_LLINFO && ln == NULL case).
1230 */
1231 rt_setgate(rt, rt_key(rt),
1232 (struct sockaddr *)&null_sdl);
1233 gate = rt->rt_gateway;
1234 SDL(gate)->sdl_type = ifp->if_type;
1235 SDL(gate)->sdl_index = ifp->if_index;
1236 if (ln)
1237 nd6_llinfo_settimer(ln, 0);
1238 if ((rt->rt_flags & RTF_CLONING) != 0)
1239 break;
1240 }
1241 /*
1242 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1243 * We don't do that here since llinfo is not ready yet.
1244 *
1245 * There are also couple of other things to be discussed:
1246 * - unsolicited NA code needs improvement beforehand
1247 * - RFC2461 says we MAY send multicast unsolicited NA
1248 * (7.2.6 paragraph 4), however, it also says that we
1249 * SHOULD provide a mechanism to prevent multicast NA storm.
1250 * we don't have anything like it right now.
1251 * note that the mechanism needs a mutual agreement
1252 * between proxies, which means that we need to implement
1253 * a new protocol, or a new kludge.
1254 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1255 * we need to check ip6forwarding before sending it.
1256 * (or should we allow proxy ND configuration only for
1257 * routers? there's no mention about proxy ND from hosts)
1258 */
1259 /* FALLTHROUGH */
1260 case RTM_RESOLVE:
1261 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1262 /*
1263 * Address resolution isn't necessary for a point to
1264 * point link, so we can skip this test for a p2p link.
1265 */
1266 if (gate->sa_family != AF_LINK ||
1267 gate->sa_len < sizeof(null_sdl)) {
1268 log(LOG_DEBUG,
1269 "nd6_rtrequest: bad gateway value: %s\n",
1270 if_name(ifp));
1271 break;
1272 }
1273 SDL(gate)->sdl_type = ifp->if_type;
1274 SDL(gate)->sdl_index = ifp->if_index;
1275 }
1276 if (ln != NULL)
1277 break; /* This happens on a route change */
1278 /*
1279 * Case 2: This route may come from cloning, or a manual route
1280 * add with a LL address.
1281 */
1282 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1283 rt->rt_llinfo = (caddr_t)ln;
1284 if (ln == NULL) {
1285 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1286 break;
1287 }
1288 nd6_inuse++;
1289 nd6_allocated++;
1290 bzero(ln, sizeof(*ln));
1291 RT_ADDREF(rt);
1292 ln->ln_rt = rt;
1293 callout_init(&ln->ln_timer_ch, 0);
1294
1295 /* this is required for "ndp" command. - shin */
1296 if (req == RTM_ADD) {
1297 /*
1298 * gate should have some valid AF_LINK entry,
1299 * and ln->ln_expire should have some lifetime
1300 * which is specified by ndp command.
1301 */
1302 ln->ln_state = ND6_LLINFO_REACHABLE;
1303 ln->ln_byhint = 0;
1304 } else {
1305 /*
1306 * When req == RTM_RESOLVE, rt is created and
1307 * initialized in rtrequest(), so rt_expire is 0.
1308 */
1309 ln->ln_state = ND6_LLINFO_NOSTATE;
1310 nd6_llinfo_settimer(ln, 0);
1311 }
1312 rt->rt_flags |= RTF_LLINFO;
1313 ln->ln_next = llinfo_nd6.ln_next;
1314 llinfo_nd6.ln_next = ln;
1315 ln->ln_prev = &llinfo_nd6;
1316 ln->ln_next->ln_prev = ln;
1317
1318 /*
1319 * check if rt_key(rt) is one of my address assigned
1320 * to the interface.
1321 */
1322 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1323 &SIN6(rt_key(rt))->sin6_addr);
1324 if (ifa) {
1325 caddr_t macp = nd6_ifptomac(ifp);
1326 nd6_llinfo_settimer(ln, -1);
1327 ln->ln_state = ND6_LLINFO_REACHABLE;
1328 ln->ln_byhint = 0;
1329 if (macp) {
1330 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1331 SDL(gate)->sdl_alen = ifp->if_addrlen;
1332 }
1333 if (nd6_useloopback) {
1334 rt->rt_ifp = &loif[0]; /* XXX */
1335 /*
1336 * Make sure rt_ifa be equal to the ifaddr
1337 * corresponding to the address.
1338 * We need this because when we refer
1339 * rt_ifa->ia6_flags in ip6_input, we assume
1340 * that the rt_ifa points to the address instead
1341 * of the loopback address.
1342 */
1343 if (ifa != rt->rt_ifa) {
1344 IFAFREE(rt->rt_ifa);
1345 IFAREF(ifa);
1346 rt->rt_ifa = ifa;
1347 }
1348 }
1349 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1350 nd6_llinfo_settimer(ln, -1);
1351 ln->ln_state = ND6_LLINFO_REACHABLE;
1352 ln->ln_byhint = 0;
1353
1354 /* join solicited node multicast for proxy ND */
1355 if (ifp->if_flags & IFF_MULTICAST) {
1356 struct in6_addr llsol;
1357 int error;
1358
1359 llsol = SIN6(rt_key(rt))->sin6_addr;
1360 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1361 llsol.s6_addr32[1] = 0;
1362 llsol.s6_addr32[2] = htonl(1);
1363 llsol.s6_addr8[12] = 0xff;
1364 if (in6_setscope(&llsol, ifp, NULL))
1365 break;
1366 if (in6_addmulti(&llsol, ifp,
1367 &error, 0) == NULL) {
1368 char ip6buf[INET6_ADDRSTRLEN];
1369 nd6log((LOG_ERR, "%s: failed to join "
1370 "%s (errno=%d)\n", if_name(ifp),
1371 ip6_sprintf(ip6buf, &llsol),
1372 error));
1373 }
1374 }
1375 }
1376 break;
1377
1378 case RTM_DELETE:
1379 if (ln == NULL)
1380 break;
1381 /* leave from solicited node multicast for proxy ND */
1382 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1383 (ifp->if_flags & IFF_MULTICAST) != 0) {
1384 struct in6_addr llsol;
1385 struct in6_multi *in6m;
1386
1387 llsol = SIN6(rt_key(rt))->sin6_addr;
1388 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1389 llsol.s6_addr32[1] = 0;
1390 llsol.s6_addr32[2] = htonl(1);
1391 llsol.s6_addr8[12] = 0xff;
1392 if (in6_setscope(&llsol, ifp, NULL) == 0) {
1393 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1394 if (in6m)
1395 in6_delmulti(in6m);
1396 } else
1397 ; /* XXX: should not happen. bark here? */
1398 }
1399 nd6_inuse--;
1400 ln->ln_next->ln_prev = ln->ln_prev;
1401 ln->ln_prev->ln_next = ln->ln_next;
1402 ln->ln_prev = NULL;
1403 nd6_llinfo_settimer(ln, -1);
1404 RT_REMREF(rt);
1405 rt->rt_llinfo = 0;
1406 rt->rt_flags &= ~RTF_LLINFO;
1407 clear_llinfo_pqueue(ln);
1408 Free((caddr_t)ln);
1409 }
1410}
1411
1412int
1413nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1414{
1415 struct in6_drlist *drl = (struct in6_drlist *)data;
1416 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1417 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1418 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1419 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1420 struct nd_defrouter *dr;
1421 struct nd_prefix *pr;
1422 struct rtentry *rt;
1423 int i = 0, error = 0;
1424 int s;
1425
1426 switch (cmd) {
1427 case SIOCGDRLST_IN6:
1428 /*
1429 * obsolete API, use sysctl under net.inet6.icmp6
1430 */
1431 bzero(drl, sizeof(*drl));
1432 s = splnet();
1433 dr = TAILQ_FIRST(&nd_defrouter);
1434 while (dr && i < DRLSTSIZ) {
1435 drl->defrouter[i].rtaddr = dr->rtaddr;
1436 in6_clearscope(&drl->defrouter[i].rtaddr);
1437
1438 drl->defrouter[i].flags = dr->flags;
1439 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1440 drl->defrouter[i].expire = dr->expire;
1441 drl->defrouter[i].if_index = dr->ifp->if_index;
1442 i++;
1443 dr = TAILQ_NEXT(dr, dr_entry);
1444 }
1445 splx(s);
1446 break;
1447 case SIOCGPRLST_IN6:
1448 /*
1449 * obsolete API, use sysctl under net.inet6.icmp6
1450 *
1451 * XXX the structure in6_prlist was changed in backward-
1452 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1453 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1454 */
1455 /*
1456 * XXX meaning of fields, especialy "raflags", is very
1457 * differnet between RA prefix list and RR/static prefix list.
1458 * how about separating ioctls into two?
1459 */
1460 bzero(oprl, sizeof(*oprl));
1461 s = splnet();
1462 pr = nd_prefix.lh_first;
1463 while (pr && i < PRLSTSIZ) {
1464 struct nd_pfxrouter *pfr;
1465 int j;
1466
1467 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1468 oprl->prefix[i].raflags = pr->ndpr_raf;
1469 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1470 oprl->prefix[i].vltime = pr->ndpr_vltime;
1471 oprl->prefix[i].pltime = pr->ndpr_pltime;
1472 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1473 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1474 oprl->prefix[i].expire = 0;
1475 else {
1476 time_t maxexpire;
1477
1478 /* XXX: we assume time_t is signed. */
1479 maxexpire = (-1) &
1480 ~((time_t)1 <<
1481 ((sizeof(maxexpire) * 8) - 1));
1482 if (pr->ndpr_vltime <
1483 maxexpire - pr->ndpr_lastupdate) {
1484 oprl->prefix[i].expire =
1485 pr->ndpr_lastupdate +
1486 pr->ndpr_vltime;
1487 } else
1488 oprl->prefix[i].expire = maxexpire;
1489 }
1490
1491 pfr = pr->ndpr_advrtrs.lh_first;
1492 j = 0;
1493 while (pfr) {
1494 if (j < DRLSTSIZ) {
1495#define RTRADDR oprl->prefix[i].advrtr[j]
1496 RTRADDR = pfr->router->rtaddr;
1497 in6_clearscope(&RTRADDR);
1498#undef RTRADDR
1499 }
1500 j++;
1501 pfr = pfr->pfr_next;
1502 }
1503 oprl->prefix[i].advrtrs = j;
1504 oprl->prefix[i].origin = PR_ORIG_RA;
1505
1506 i++;
1507 pr = pr->ndpr_next;
1508 }
1509 splx(s);
1510
1511 break;
1512 case OSIOCGIFINFO_IN6:
1513#define ND ndi->ndi
1514 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1515 bzero(&ND, sizeof(ND));
1516 ND.linkmtu = IN6_LINKMTU(ifp);
1517 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1518 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1519 ND.reachable = ND_IFINFO(ifp)->reachable;
1520 ND.retrans = ND_IFINFO(ifp)->retrans;
1521 ND.flags = ND_IFINFO(ifp)->flags;
1522 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1523 ND.chlim = ND_IFINFO(ifp)->chlim;
1524 break;
1525 case SIOCGIFINFO_IN6:
1526 ND = *ND_IFINFO(ifp);
1527 break;
1528 case SIOCSIFINFO_IN6:
1529 /*
1530 * used to change host variables from userland.
1531 * intented for a use on router to reflect RA configurations.
1532 */
1533 /* 0 means 'unspecified' */
1534 if (ND.linkmtu != 0) {
1535 if (ND.linkmtu < IPV6_MMTU ||
1536 ND.linkmtu > IN6_LINKMTU(ifp)) {
1537 error = EINVAL;
1538 break;
1539 }
1540 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1541 }
1542
1543 if (ND.basereachable != 0) {
1544 int obasereachable = ND_IFINFO(ifp)->basereachable;
1545
1546 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1547 if (ND.basereachable != obasereachable)
1548 ND_IFINFO(ifp)->reachable =
1549 ND_COMPUTE_RTIME(ND.basereachable);
1550 }
1551 if (ND.retrans != 0)
1552 ND_IFINFO(ifp)->retrans = ND.retrans;
1553 if (ND.chlim != 0)
1554 ND_IFINFO(ifp)->chlim = ND.chlim;
1555 /* FALLTHROUGH */
1556 case SIOCSIFINFO_FLAGS:
1557 ND_IFINFO(ifp)->flags = ND.flags;
1558 break;
1559#undef ND
1560 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1561 /* sync kernel routing table with the default router list */
1562 defrouter_reset();
1563 defrouter_select();
1564 break;
1565 case SIOCSPFXFLUSH_IN6:
1566 {
1567 /* flush all the prefix advertised by routers */
1568 struct nd_prefix *pr, *next;
1569
1570 s = splnet();
1571 for (pr = nd_prefix.lh_first; pr; pr = next) {
1572 struct in6_ifaddr *ia, *ia_next;
1573
1574 next = pr->ndpr_next;
1575
1576 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1577 continue; /* XXX */
1578
1579 /* do we really have to remove addresses as well? */
1580 for (ia = in6_ifaddr; ia; ia = ia_next) {
1581 /* ia might be removed. keep the next ptr. */
1582 ia_next = ia->ia_next;
1583
1584 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1585 continue;
1586
1587 if (ia->ia6_ndpr == pr)
1588 in6_purgeaddr(&ia->ia_ifa);
1589 }
1590 prelist_remove(pr);
1591 }
1592 splx(s);
1593 break;
1594 }
1595 case SIOCSRTRFLUSH_IN6:
1596 {
1597 /* flush all the default routers */
1598 struct nd_defrouter *dr, *next;
1599
1600 s = splnet();
1601 defrouter_reset();
1602 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = next) {
1603 next = TAILQ_NEXT(dr, dr_entry);
1604 defrtrlist_del(dr);
1605 }
1606 defrouter_select();
1607 splx(s);
1608 break;
1609 }
1610 case SIOCGNBRINFO_IN6:
1611 {
1612 struct llinfo_nd6 *ln;
1613 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1614
1615 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1616 return (error);
1617
1618 s = splnet();
1619 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1620 error = EINVAL;
1621 splx(s);
1622 break;
1623 }
1624 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1625 nbi->state = ln->ln_state;
1626 nbi->asked = ln->ln_asked;
1627 nbi->isrouter = ln->ln_router;
1628 nbi->expire = ln->ln_expire;
1629 splx(s);
1630
1631 break;
1632 }
1633 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1634 ndif->ifindex = nd6_defifindex;
1635 break;
1636 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1637 return (nd6_setdefaultiface(ndif->ifindex));
1638 }
1639 return (error);
1640}
1641
1642/*
1643 * Create neighbor cache entry and cache link-layer address,
1644 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1645 *
1646 * type - ICMP6 type
1647 * code - type dependent information
1648 */
1649struct rtentry *
1650nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1651 int lladdrlen, int type, int code)
1652{
1653 struct rtentry *rt = NULL;
1654 struct llinfo_nd6 *ln = NULL;
1655 int is_newentry;
1656 struct sockaddr_dl *sdl = NULL;
1657 int do_update;
1658 int olladdr;
1659 int llchange;
1660 int newstate = 0;
1661
1662 if (ifp == NULL)
1663 panic("ifp == NULL in nd6_cache_lladdr");
1664 if (from == NULL)
1665 panic("from == NULL in nd6_cache_lladdr");
1666
1667 /* nothing must be updated for unspecified address */
1668 if (IN6_IS_ADDR_UNSPECIFIED(from))
1669 return NULL;
1670
1671 /*
1672 * Validation about ifp->if_addrlen and lladdrlen must be done in
1673 * the caller.
1674 *
1675 * XXX If the link does not have link-layer adderss, what should
1676 * we do? (ifp->if_addrlen == 0)
1677 * Spec says nothing in sections for RA, RS and NA. There's small
1678 * description on it in NS section (RFC 2461 7.2.3).
1679 */
1680
1681 rt = nd6_lookup(from, 0, ifp);
1682 if (rt == NULL) {
1683 rt = nd6_lookup(from, 1, ifp);
1684 is_newentry = 1;
1685 } else {
1686 /* do nothing if static ndp is set */
1687 if (rt->rt_flags & RTF_STATIC)
1688 return NULL;
1689 is_newentry = 0;
1690 }
1691
1692 if (rt == NULL)
1693 return NULL;
1694 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1695fail:
1696 (void)nd6_free(rt, 0);
1697 return NULL;
1698 }
1699 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1700 if (ln == NULL)
1701 goto fail;
1702 if (rt->rt_gateway == NULL)
1703 goto fail;
1704 if (rt->rt_gateway->sa_family != AF_LINK)
1705 goto fail;
1706 sdl = SDL(rt->rt_gateway);
1707
1708 olladdr = (sdl->sdl_alen) ? 1 : 0;
1709 if (olladdr && lladdr) {
1710 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1711 llchange = 1;
1712 else
1713 llchange = 0;
1714 } else
1715 llchange = 0;
1716
1717 /*
1718 * newentry olladdr lladdr llchange (*=record)
1719 * 0 n n -- (1)
1720 * 0 y n -- (2)
1721 * 0 n y -- (3) * STALE
1722 * 0 y y n (4) *
1723 * 0 y y y (5) * STALE
1724 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1725 * 1 -- y -- (7) * STALE
1726 */
1727
1728 if (lladdr) { /* (3-5) and (7) */
1729 /*
1730 * Record source link-layer address
1731 * XXX is it dependent to ifp->if_type?
1732 */
1733 sdl->sdl_alen = ifp->if_addrlen;
1734 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1735 }
1736
1737 if (!is_newentry) {
1738 if ((!olladdr && lladdr != NULL) || /* (3) */
1739 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1740 do_update = 1;
1741 newstate = ND6_LLINFO_STALE;
1742 } else /* (1-2,4) */
1743 do_update = 0;
1744 } else {
1745 do_update = 1;
1746 if (lladdr == NULL) /* (6) */
1747 newstate = ND6_LLINFO_NOSTATE;
1748 else /* (7) */
1749 newstate = ND6_LLINFO_STALE;
1750 }
1751
1752 if (do_update) {
1753 /*
1754 * Update the state of the neighbor cache.
1755 */
1756 ln->ln_state = newstate;
1757
1758 if (ln->ln_state == ND6_LLINFO_STALE) {
1759 /*
1760 * XXX: since nd6_output() below will cause
1761 * state tansition to DELAY and reset the timer,
1762 * we must set the timer now, although it is actually
1763 * meaningless.
1764 */
1765 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1766
1767 if (ln->ln_hold) {
1768 struct mbuf *m_hold, *m_hold_next;
1769
1770 /*
1771 * reset the ln_hold in advance, to explicitly
1772 * prevent a ln_hold lookup in nd6_output()
1773 * (wouldn't happen, though...)
1774 */
1775 for (m_hold = ln->ln_hold, ln->ln_hold = NULL;
1776 m_hold; m_hold = m_hold_next) {
1777 m_hold_next = m_hold->m_nextpkt;
1778 m_hold->m_nextpkt = NULL;
1779
1780 /*
1781 * we assume ifp is not a p2p here, so
1782 * just set the 2nd argument as the
1783 * 1st one.
1784 */
1785 nd6_output(ifp, ifp, m_hold,
1786 (struct sockaddr_in6 *)rt_key(rt),
1787 rt);
1788 }
1789 }
1790 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1791 /* probe right away */
1792 nd6_llinfo_settimer((void *)ln, 0);
1793 }
1794 }
1795
1796 /*
1797 * ICMP6 type dependent behavior.
1798 *
1799 * NS: clear IsRouter if new entry
1800 * RS: clear IsRouter
1801 * RA: set IsRouter if there's lladdr
1802 * redir: clear IsRouter if new entry
1803 *
1804 * RA case, (1):
1805 * The spec says that we must set IsRouter in the following cases:
1806 * - If lladdr exist, set IsRouter. This means (1-5).
1807 * - If it is old entry (!newentry), set IsRouter. This means (7).
1808 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1809 * A quetion arises for (1) case. (1) case has no lladdr in the
1810 * neighbor cache, this is similar to (6).
1811 * This case is rare but we figured that we MUST NOT set IsRouter.
1812 *
1813 * newentry olladdr lladdr llchange NS RS RA redir
1814 * D R
1815 * 0 n n -- (1) c ? s
1816 * 0 y n -- (2) c s s
1817 * 0 n y -- (3) c s s
1818 * 0 y y n (4) c s s
1819 * 0 y y y (5) c s s
1820 * 1 -- n -- (6) c c c s
1821 * 1 -- y -- (7) c c s c s
1822 *
1823 * (c=clear s=set)
1824 */
1825 switch (type & 0xff) {
1826 case ND_NEIGHBOR_SOLICIT:
1827 /*
1828 * New entry must have is_router flag cleared.
1829 */
1830 if (is_newentry) /* (6-7) */
1831 ln->ln_router = 0;
1832 break;
1833 case ND_REDIRECT:
1834 /*
1835 * If the icmp is a redirect to a better router, always set the
1836 * is_router flag. Otherwise, if the entry is newly created,
1837 * clear the flag. [RFC 2461, sec 8.3]
1838 */
1839 if (code == ND_REDIRECT_ROUTER)
1840 ln->ln_router = 1;
1841 else if (is_newentry) /* (6-7) */
1842 ln->ln_router = 0;
1843 break;
1844 case ND_ROUTER_SOLICIT:
1845 /*
1846 * is_router flag must always be cleared.
1847 */
1848 ln->ln_router = 0;
1849 break;
1850 case ND_ROUTER_ADVERT:
1851 /*
1852 * Mark an entry with lladdr as a router.
1853 */
1854 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1855 (is_newentry && lladdr)) { /* (7) */
1856 ln->ln_router = 1;
1857 }
1858 break;
1859 }
1860
1861 /*
1862 * When the link-layer address of a router changes, select the
1863 * best router again. In particular, when the neighbor entry is newly
1864 * created, it might affect the selection policy.
1865 * Question: can we restrict the first condition to the "is_newentry"
1866 * case?
1867 * XXX: when we hear an RA from a new router with the link-layer
1868 * address option, defrouter_select() is called twice, since
1869 * defrtrlist_update called the function as well. However, I believe
1870 * we can compromise the overhead, since it only happens the first
1871 * time.
1872 * XXX: although defrouter_select() should not have a bad effect
1873 * for those are not autoconfigured hosts, we explicitly avoid such
1874 * cases for safety.
1875 */
1876 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1877 defrouter_select();
1878
1879 return rt;
1880}
1881
1882static void
1883nd6_slowtimo(void *ignored_arg)
1884{
1885 struct nd_ifinfo *nd6if;
1886 struct ifnet *ifp;
1887
1888 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1889 nd6_slowtimo, NULL);
1890 IFNET_RLOCK();
1891 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1892 nd6if = ND_IFINFO(ifp);
1893 if (nd6if->basereachable && /* already initialized */
1894 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1895 /*
1896 * Since reachable time rarely changes by router
1897 * advertisements, we SHOULD insure that a new random
1898 * value gets recomputed at least once every few hours.
1899 * (RFC 2461, 6.3.4)
1900 */
1901 nd6if->recalctm = nd6_recalc_reachtm_interval;
1902 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1903 }
1904 }
1905 IFNET_RUNLOCK();
1906}
1907
1908#define senderr(e) { error = (e); goto bad;}
1909int
1910nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1911 struct sockaddr_in6 *dst, struct rtentry *rt0)
1912{
1913 struct mbuf *m = m0;
1914 struct rtentry *rt = rt0;
1915 struct sockaddr_in6 *gw6 = NULL;
1916 struct llinfo_nd6 *ln = NULL;
1917 int error = 0;
1918
1919 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1920 goto sendpkt;
1921
1922 if (nd6_need_cache(ifp) == 0)
1923 goto sendpkt;
1924
1925 /*
1926 * next hop determination. This routine is derived from ether_output.
1927 */
1928 /* NB: the locking here is tortuous... */
1929 if (rt != NULL)
1930 RT_LOCK(rt);
1931again:
1932 if (rt != NULL) {
1933 if ((rt->rt_flags & RTF_UP) == 0) {
1934 RT_UNLOCK(rt);
1935 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL);
1936 if (rt != NULL) {
1937 RT_REMREF(rt);
1938 if (rt->rt_ifp != ifp)
1939 /*
1940 * XXX maybe we should update ifp too,
1941 * but the original code didn't and I
1942 * don't know what is correct here.
1943 */
1944 goto again;
1945 } else
1946 senderr(EHOSTUNREACH);
1947 }
1948
1949 if (rt->rt_flags & RTF_GATEWAY) {
1950 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1951
1952 /*
1953 * We skip link-layer address resolution and NUD
1954 * if the gateway is not a neighbor from ND point
1955 * of view, regardless of the value of nd_ifinfo.flags.
1956 * The second condition is a bit tricky; we skip
1957 * if the gateway is our own address, which is
1958 * sometimes used to install a route to a p2p link.
1959 */
1960 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1961 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1962 RT_UNLOCK(rt);
1963 /*
1964 * We allow this kind of tricky route only
1965 * when the outgoing interface is p2p.
1966 * XXX: we may need a more generic rule here.
1967 */
1968 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1969 senderr(EHOSTUNREACH);
1970
1971 goto sendpkt;
1972 }
1973
1974 if (rt->rt_gwroute == NULL)
1975 goto lookup;
1976 rt = rt->rt_gwroute;
1977 RT_LOCK(rt); /* NB: gwroute */
1978 if ((rt->rt_flags & RTF_UP) == 0) {
1979 RTFREE_LOCKED(rt); /* unlock gwroute */
1980 rt = rt0;
1981 rt0->rt_gwroute = NULL;
1982 lookup:
1983 RT_UNLOCK(rt0);
1984 rt = rtalloc1(rt->rt_gateway, 1, 0UL);
1985 if (rt == rt0) {
1986 RT_REMREF(rt0);
1987 RT_UNLOCK(rt0);
1988 senderr(EHOSTUNREACH);
1989 }
1990 RT_LOCK(rt0);
1991 if (rt0->rt_gwroute != NULL)
1992 RTFREE(rt0->rt_gwroute);
1993 rt0->rt_gwroute = rt;
1994 if (rt == NULL) {
1995 RT_UNLOCK(rt0);
1996 senderr(EHOSTUNREACH);
1997 }
1998 }
1999 RT_UNLOCK(rt0);
2000 }
2001 RT_UNLOCK(rt);
2002 }
2003
2004 /*
2005 * Address resolution or Neighbor Unreachability Detection
2006 * for the next hop.
2007 * At this point, the destination of the packet must be a unicast
2008 * or an anycast address(i.e. not a multicast).
2009 */
2010
2011 /* Look up the neighbor cache for the nexthop */
2012 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
2013 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2014 else {
2015 /*
2016 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2017 * the condition below is not very efficient. But we believe
2018 * it is tolerable, because this should be a rare case.
2019 */
2020 if (nd6_is_addr_neighbor(dst, ifp) &&
2021 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
2022 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2023 }
2024 if (ln == NULL || rt == NULL) {
2025 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2026 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2027 char ip6buf[INET6_ADDRSTRLEN];
2028 log(LOG_DEBUG,
2029 "nd6_output: can't allocate llinfo for %s "
2030 "(ln=%p, rt=%p)\n",
2031 ip6_sprintf(ip6buf, &dst->sin6_addr), ln, rt);
2032 senderr(EIO); /* XXX: good error? */
2033 }
2034
2035 goto sendpkt; /* send anyway */
2036 }
2037
2038 /* We don't have to do link-layer address resolution on a p2p link. */
2039 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2040 ln->ln_state < ND6_LLINFO_REACHABLE) {
2041 ln->ln_state = ND6_LLINFO_STALE;
2042 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
2043 }
2044
2045 /*
2046 * The first time we send a packet to a neighbor whose entry is
2047 * STALE, we have to change the state to DELAY and a sets a timer to
2048 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2049 * neighbor unreachability detection on expiration.
2050 * (RFC 2461 7.3.3)
2051 */
2052 if (ln->ln_state == ND6_LLINFO_STALE) {
2053 ln->ln_asked = 0;
2054 ln->ln_state = ND6_LLINFO_DELAY;
2055 nd6_llinfo_settimer(ln, (long)nd6_delay * hz);
2056 }
2057
2058 /*
2059 * If the neighbor cache entry has a state other than INCOMPLETE
2060 * (i.e. its link-layer address is already resolved), just
2061 * send the packet.
2062 */
2063 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2064 goto sendpkt;
2065
2066 /*
2067 * There is a neighbor cache entry, but no ethernet address
2068 * response yet. Append this latest packet to the end of the
2069 * packet queue in the mbuf, unless the number of the packet
2070 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2071 * the oldest packet in the queue will be removed.
2072 */
2073 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2074 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2075 if (ln->ln_hold) {
2076 struct mbuf *m_hold;
2077 int i;
2078
2079 i = 0;
2080 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) {
2081 i++;
2082 if (m_hold->m_nextpkt == NULL) {
2083 m_hold->m_nextpkt = m;
2084 break;
2085 }
2086 }
2087 while (i >= nd6_maxqueuelen) {
2088 m_hold = ln->ln_hold;
2089 ln->ln_hold = ln->ln_hold->m_nextpkt;
2090 m_freem(m_hold);
2091 i--;
2092 }
2093 } else {
2094 ln->ln_hold = m;
2095 }
2096
2097 /*
2098 * If there has been no NS for the neighbor after entering the
2099 * INCOMPLETE state, send the first solicitation.
2100 */
2101 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
2102 ln->ln_asked++;
2103 nd6_llinfo_settimer(ln,
2104 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2105 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2106 }
2107 return (0);
2108
2109 sendpkt:
2110 /* discard the packet if IPv6 operation is disabled on the interface */
2111 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2112 error = ENETDOWN; /* better error? */
2113 goto bad;
2114 }
2115
2116#ifdef MAC
2117 mac_mbuf_create_linklayer(ifp, m);
2118#endif
2119 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2120 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2121 rt));
2122 }
2123 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2124
2125 bad:
2126 if (m)
2127 m_freem(m);
2128 return (error);
2129}
2130#undef senderr
2131
2132int
2133nd6_need_cache(struct ifnet *ifp)
2134{
2135 /*
2136 * XXX: we currently do not make neighbor cache on any interface
2137 * other than ARCnet, Ethernet, FDDI and GIF.
2138 *
2139 * RFC2893 says:
2140 * - unidirectional tunnels needs no ND
2141 */
2142 switch (ifp->if_type) {
2143 case IFT_ARCNET:
2144 case IFT_ETHER:
2145 case IFT_FDDI:
2146 case IFT_IEEE1394:
2147#ifdef IFT_L2VLAN
2148 case IFT_L2VLAN:
2149#endif
2150#ifdef IFT_IEEE80211
2151 case IFT_IEEE80211:
2152#endif
2153#ifdef IFT_CARP
2154 case IFT_CARP:
2155#endif
2156 case IFT_GIF: /* XXX need more cases? */
2157 case IFT_PPP:
2158 case IFT_TUNNEL:
2159 case IFT_BRIDGE:
2160 case IFT_PROPVIRTUAL:
2161 return (1);
2162 default:
2163 return (0);
2164 }
2165}
2166
2167int
2168nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
2169 struct sockaddr *dst, u_char *desten)
2170{
2171 struct sockaddr_dl *sdl;
2172 struct rtentry *rt;
2173 int error;
2174
2175 if (m->m_flags & M_MCAST) {
2176 int i;
2177
2178 switch (ifp->if_type) {
2179 case IFT_ETHER:
2180 case IFT_FDDI:
2181#ifdef IFT_L2VLAN
2182 case IFT_L2VLAN:
2183#endif
2184#ifdef IFT_IEEE80211
2185 case IFT_IEEE80211:
2186#endif
2187 case IFT_BRIDGE:
2188 case IFT_ISO88025:
2189 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2190 desten);
2191 return (0);
2192 case IFT_IEEE1394:
2193 /*
2194 * netbsd can use if_broadcastaddr, but we don't do so
2195 * to reduce # of ifdef.
2196 */
2197 for (i = 0; i < ifp->if_addrlen; i++)
2198 desten[i] = ~0;
2199 return (0);
2200 case IFT_ARCNET:
2201 *desten = 0;
2202 return (0);
2203 default:
2204 m_freem(m);
2205 return (EAFNOSUPPORT);
2206 }
2207 }
2208
2209 if (rt0 == NULL) {
2210 /* this could happen, if we could not allocate memory */
2211 m_freem(m);
2212 return (ENOMEM);
2213 }
2214
2215 error = rt_check(&rt, &rt0, dst);
2216 if (error) {
2217 m_freem(m);
2218 return (error);
2219 }
2220 RT_UNLOCK(rt);
2221
2222 if (rt->rt_gateway->sa_family != AF_LINK) {
2223 printf("nd6_storelladdr: something odd happens\n");
2224 m_freem(m);
2225 return (EINVAL);
2226 }
2227 sdl = SDL(rt->rt_gateway);
2228 if (sdl->sdl_alen == 0) {
2229 /* this should be impossible, but we bark here for debugging */
2230 printf("nd6_storelladdr: sdl_alen == 0\n");
2231 m_freem(m);
2232 return (EINVAL);
2233 }
2234
2235 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2236 return (0);
2237}
2238
2239static void
2240clear_llinfo_pqueue(struct llinfo_nd6 *ln)
2241{
2242 struct mbuf *m_hold, *m_hold_next;
2243
2244 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) {
2245 m_hold_next = m_hold->m_nextpkt;
2246 m_hold->m_nextpkt = NULL;
2247 m_freem(m_hold);
2248 }
2249
2250 ln->ln_hold = NULL;
2251 return;
2252}
2253
2254static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2255static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2256#ifdef SYSCTL_DECL
2257SYSCTL_DECL(_net_inet6_icmp6);
2258#endif
2259SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2260 CTLFLAG_RD, nd6_sysctl_drlist, "");
2261SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2262 CTLFLAG_RD, nd6_sysctl_prlist, "");
2263SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2264 CTLFLAG_RW, &nd6_maxqueuelen, 1, "");
2265
2266static int
2267nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2268{
2269 int error;
2270 char buf[1024] __aligned(4);
2271 struct in6_defrouter *d, *de;
2272 struct nd_defrouter *dr;
2273
2274 if (req->newptr)
2275 return EPERM;
2276 error = 0;
2277
2278 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
2279 dr = TAILQ_NEXT(dr, dr_entry)) {
2280 d = (struct in6_defrouter *)buf;
2281 de = (struct in6_defrouter *)(buf + sizeof(buf));
2282
2283 if (d + 1 <= de) {
2284 bzero(d, sizeof(*d));
2285 d->rtaddr.sin6_family = AF_INET6;
2286 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2287 d->rtaddr.sin6_addr = dr->rtaddr;
2288 error = sa6_recoverscope(&d->rtaddr);
2289 if (error != 0)
2290 return (error);
2291 d->flags = dr->flags;
2292 d->rtlifetime = dr->rtlifetime;
2293 d->expire = dr->expire;
2294 d->if_index = dr->ifp->if_index;
2295 } else
2296 panic("buffer too short");
2297
2298 error = SYSCTL_OUT(req, buf, sizeof(*d));
2299 if (error)
2300 break;
2301 }
2302
2303 return (error);
2304}
2305
2306static int
2307nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2308{
2309 int error;
2310 char buf[1024] __aligned(4);
2311 struct in6_prefix *p, *pe;
2312 struct nd_prefix *pr;
2313 char ip6buf[INET6_ADDRSTRLEN];
2314
2315 if (req->newptr)
2316 return EPERM;
2317 error = 0;
2318
2319 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2320 u_short advrtrs;
2321 size_t advance;
2322 struct sockaddr_in6 *sin6, *s6;
2323 struct nd_pfxrouter *pfr;
2324
2325 p = (struct in6_prefix *)buf;
2326 pe = (struct in6_prefix *)(buf + sizeof(buf));
2327
2328 if (p + 1 <= pe) {
2329 bzero(p, sizeof(*p));
2330 sin6 = (struct sockaddr_in6 *)(p + 1);
2331
2332 p->prefix = pr->ndpr_prefix;
2333 if (sa6_recoverscope(&p->prefix)) {
2334 log(LOG_ERR,
2335 "scope error in prefix list (%s)\n",
2336 ip6_sprintf(ip6buf, &p->prefix.sin6_addr));
2337 /* XXX: press on... */
2338 }
2339 p->raflags = pr->ndpr_raf;
2340 p->prefixlen = pr->ndpr_plen;
2341 p->vltime = pr->ndpr_vltime;
2342 p->pltime = pr->ndpr_pltime;
2343 p->if_index = pr->ndpr_ifp->if_index;
2344 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2345 p->expire = 0;
2346 else {
2347 time_t maxexpire;
2348
2349 /* XXX: we assume time_t is signed. */
2350 maxexpire = (-1) &
2351 ~((time_t)1 <<
2352 ((sizeof(maxexpire) * 8) - 1));
2353 if (pr->ndpr_vltime <
2354 maxexpire - pr->ndpr_lastupdate) {
2355 p->expire = pr->ndpr_lastupdate +
2356 pr->ndpr_vltime;
2357 } else
2358 p->expire = maxexpire;
2359 }
2360 p->refcnt = pr->ndpr_refcnt;
2361 p->flags = pr->ndpr_stateflags;
2362 p->origin = PR_ORIG_RA;
2363 advrtrs = 0;
2364 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2365 pfr = pfr->pfr_next) {
2366 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2367 advrtrs++;
2368 continue;
2369 }
2370 s6 = &sin6[advrtrs];
2371 bzero(s6, sizeof(*s6));
2372 s6->sin6_family = AF_INET6;
2373 s6->sin6_len = sizeof(*sin6);
2374 s6->sin6_addr = pfr->router->rtaddr;
2375 if (sa6_recoverscope(s6)) {
2376 log(LOG_ERR,
2377 "scope error in "
2378 "prefix list (%s)\n",
2379 ip6_sprintf(ip6buf,
2380 &pfr->router->rtaddr));
2381 }
2382 advrtrs++;
2383 }
2384 p->advrtrs = advrtrs;
2385 } else
2386 panic("buffer too short");
2387
2388 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2389 error = SYSCTL_OUT(req, buf, advance);
2390 if (error)
2391 break;
2392 }
2393
2394 return (error);
2395}