FreeBSD/Linux Kernel Cross Reference
sys/netinet6/nd6.c
1 /* $NetBSD: nd6.c,v 1.109 2006/11/24 19:47:00 christos Exp $ */
2 /* $KAME: nd6.c,v 1.279 2002/06/08 11:16:51 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/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.109 2006/11/24 19:47:00 christos Exp $");
35
36 #include "opt_ipsec.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/callout.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/protosw.h>
48 #include <sys/errno.h>
49 #include <sys/ioctl.h>
50 #include <sys/syslog.h>
51 #include <sys/queue.h>
52
53 #include <net/if.h>
54 #include <net/if_dl.h>
55 #include <net/if_types.h>
56 #include <net/route.h>
57 #include <net/if_ether.h>
58 #include <net/if_fddi.h>
59 #include <net/if_arc.h>
60
61 #include <netinet/in.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 #ifdef IPSEC
70 #include <netinet6/ipsec.h>
71 #endif
72
73 #include <net/net_osdep.h>
74
75 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
76 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
77
78 #define SIN6(s) ((struct sockaddr_in6 *)s)
79 #define SDL(s) ((struct sockaddr_dl *)s)
80
81 /* timer values */
82 int nd6_prune = 1; /* walk list every 1 seconds */
83 int nd6_delay = 5; /* delay first probe time 5 second */
84 int nd6_umaxtries = 3; /* maximum unicast query */
85 int nd6_mmaxtries = 3; /* maximum multicast query */
86 int nd6_useloopback = 1; /* use loopback interface for local traffic */
87 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
88
89 /* preventing too many loops in ND option parsing */
90 int nd6_maxndopt = 10; /* max # of ND options allowed */
91
92 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
93
94 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */
95
96 #ifdef ND6_DEBUG
97 int nd6_debug = 1;
98 #else
99 int nd6_debug = 0;
100 #endif
101
102 /* for debugging? */
103 static int nd6_inuse, nd6_allocated;
104
105 struct llinfo_nd6 llinfo_nd6 = {
106 .ln_prev = &llinfo_nd6,
107 .ln_next = &llinfo_nd6,
108 };
109 struct nd_drhead nd_defrouter;
110 struct nd_prhead nd_prefix = { 0 };
111
112 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
113 static struct sockaddr_in6 all1_sa;
114
115 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *));
116 static void nd6_slowtimo __P((void *));
117 static int regen_tmpaddr __P((struct in6_ifaddr *));
118 static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int));
119 static void nd6_llinfo_timer __P((void *));
120 static void clear_llinfo_pqueue __P((struct llinfo_nd6 *));
121
122 struct callout nd6_slowtimo_ch = CALLOUT_INITIALIZER;
123 struct callout nd6_timer_ch = CALLOUT_INITIALIZER;
124 extern struct callout in6_tmpaddrtimer_ch;
125
126 static int fill_drlist __P((void *, size_t *, size_t));
127 static int fill_prlist __P((void *, size_t *, size_t));
128
129 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery");
130
131 void
132 nd6_init()
133 {
134 static int nd6_init_done = 0;
135 int i;
136
137 if (nd6_init_done) {
138 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
139 return;
140 }
141
142 all1_sa.sin6_family = AF_INET6;
143 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
144 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
145 all1_sa.sin6_addr.s6_addr[i] = 0xff;
146
147 /* initialization of the default router list */
148 TAILQ_INIT(&nd_defrouter);
149
150 nd6_init_done = 1;
151
152 /* start timer */
153 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
154 nd6_slowtimo, NULL);
155 }
156
157 struct nd_ifinfo *
158 nd6_ifattach(ifp)
159 struct ifnet *ifp;
160 {
161 struct nd_ifinfo *nd;
162
163 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
164 bzero(nd, sizeof(*nd));
165
166 nd->initialized = 1;
167
168 nd->chlim = IPV6_DEFHLIM;
169 nd->basereachable = REACHABLE_TIME;
170 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
171 nd->retrans = RETRANS_TIMER;
172 /*
173 * Note that the default value of ip6_accept_rtadv is 0, which means
174 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
175 * here.
176 */
177 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
178
179 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
180 nd6_setmtu0(ifp, nd);
181
182 return nd;
183 }
184
185 void
186 nd6_ifdetach(nd)
187 struct nd_ifinfo *nd;
188 {
189
190 free(nd, M_IP6NDP);
191 }
192
193 void
194 nd6_setmtu(ifp)
195 struct ifnet *ifp;
196 {
197 nd6_setmtu0(ifp, ND_IFINFO(ifp));
198 }
199
200 void
201 nd6_setmtu0(ifp, ndi)
202 struct ifnet *ifp;
203 struct nd_ifinfo *ndi;
204 {
205 u_int32_t omaxmtu;
206
207 omaxmtu = ndi->maxmtu;
208
209 switch (ifp->if_type) {
210 case IFT_ARCNET:
211 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
212 break;
213 case IFT_FDDI:
214 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
215 break;
216 default:
217 ndi->maxmtu = ifp->if_mtu;
218 break;
219 }
220
221 /*
222 * Decreasing the interface MTU under IPV6 minimum MTU may cause
223 * undesirable situation. We thus notify the operator of the change
224 * explicitly. The check for omaxmtu is necessary to restrict the
225 * log to the case of changing the MTU, not initializing it.
226 */
227 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
228 log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too"
229 " small for IPv6 which needs %lu\n",
230 if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long)
231 IPV6_MMTU);
232 }
233
234 if (ndi->maxmtu > in6_maxmtu)
235 in6_setmaxmtu(); /* check all interfaces just in case */
236 }
237
238 void
239 nd6_option_init(opt, icmp6len, ndopts)
240 void *opt;
241 int icmp6len;
242 union nd_opts *ndopts;
243 {
244
245 bzero(ndopts, sizeof(*ndopts));
246 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
247 ndopts->nd_opts_last
248 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
249
250 if (icmp6len == 0) {
251 ndopts->nd_opts_done = 1;
252 ndopts->nd_opts_search = NULL;
253 }
254 }
255
256 /*
257 * Take one ND option.
258 */
259 struct nd_opt_hdr *
260 nd6_option(ndopts)
261 union nd_opts *ndopts;
262 {
263 struct nd_opt_hdr *nd_opt;
264 int olen;
265
266 if (ndopts == NULL)
267 panic("ndopts == NULL in nd6_option");
268 if (ndopts->nd_opts_last == NULL)
269 panic("uninitialized ndopts in nd6_option");
270 if (ndopts->nd_opts_search == NULL)
271 return NULL;
272 if (ndopts->nd_opts_done)
273 return NULL;
274
275 nd_opt = ndopts->nd_opts_search;
276
277 /* make sure nd_opt_len is inside the buffer */
278 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
279 bzero(ndopts, sizeof(*ndopts));
280 return NULL;
281 }
282
283 olen = nd_opt->nd_opt_len << 3;
284 if (olen == 0) {
285 /*
286 * Message validation requires that all included
287 * options have a length that is greater than zero.
288 */
289 bzero(ndopts, sizeof(*ndopts));
290 return NULL;
291 }
292
293 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
294 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
295 /* option overruns the end of buffer, invalid */
296 bzero(ndopts, sizeof(*ndopts));
297 return NULL;
298 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
299 /* reached the end of options chain */
300 ndopts->nd_opts_done = 1;
301 ndopts->nd_opts_search = NULL;
302 }
303 return nd_opt;
304 }
305
306 /*
307 * Parse multiple ND options.
308 * This function is much easier to use, for ND routines that do not need
309 * multiple options of the same type.
310 */
311 int
312 nd6_options(ndopts)
313 union nd_opts *ndopts;
314 {
315 struct nd_opt_hdr *nd_opt;
316 int i = 0;
317
318 if (ndopts == NULL)
319 panic("ndopts == NULL in nd6_options");
320 if (ndopts->nd_opts_last == NULL)
321 panic("uninitialized ndopts in nd6_options");
322 if (ndopts->nd_opts_search == NULL)
323 return 0;
324
325 while (1) {
326 nd_opt = nd6_option(ndopts);
327 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
328 /*
329 * Message validation requires that all included
330 * options have a length that is greater than zero.
331 */
332 icmp6stat.icp6s_nd_badopt++;
333 bzero(ndopts, sizeof(*ndopts));
334 return -1;
335 }
336
337 if (nd_opt == NULL)
338 goto skip1;
339
340 switch (nd_opt->nd_opt_type) {
341 case ND_OPT_SOURCE_LINKADDR:
342 case ND_OPT_TARGET_LINKADDR:
343 case ND_OPT_MTU:
344 case ND_OPT_REDIRECTED_HEADER:
345 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
346 nd6log((LOG_INFO,
347 "duplicated ND6 option found (type=%d)\n",
348 nd_opt->nd_opt_type));
349 /* XXX bark? */
350 } else {
351 ndopts->nd_opt_array[nd_opt->nd_opt_type]
352 = nd_opt;
353 }
354 break;
355 case ND_OPT_PREFIX_INFORMATION:
356 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
357 ndopts->nd_opt_array[nd_opt->nd_opt_type]
358 = nd_opt;
359 }
360 ndopts->nd_opts_pi_end =
361 (struct nd_opt_prefix_info *)nd_opt;
362 break;
363 default:
364 /*
365 * Unknown options must be silently ignored,
366 * to accommodate future extension to the protocol.
367 */
368 nd6log((LOG_DEBUG,
369 "nd6_options: unsupported option %d - "
370 "option ignored\n", nd_opt->nd_opt_type));
371 }
372
373 skip1:
374 i++;
375 if (i > nd6_maxndopt) {
376 icmp6stat.icp6s_nd_toomanyopt++;
377 nd6log((LOG_INFO, "too many loop in nd opt\n"));
378 break;
379 }
380
381 if (ndopts->nd_opts_done)
382 break;
383 }
384
385 return 0;
386 }
387
388 /*
389 * ND6 timer routine to handle ND6 entries
390 */
391 void
392 nd6_llinfo_settimer(ln, xtick)
393 struct llinfo_nd6 *ln;
394 long xtick;
395 {
396 int s;
397
398 s = splsoftnet();
399
400 if (xtick < 0) {
401 ln->ln_expire = 0;
402 ln->ln_ntick = 0;
403 callout_stop(&ln->ln_timer_ch);
404 } else {
405 ln->ln_expire = time_second + xtick / hz;
406 if (xtick > INT_MAX) {
407 ln->ln_ntick = xtick - INT_MAX;
408 callout_reset(&ln->ln_timer_ch, INT_MAX,
409 nd6_llinfo_timer, ln);
410 } else {
411 ln->ln_ntick = 0;
412 callout_reset(&ln->ln_timer_ch, xtick,
413 nd6_llinfo_timer, ln);
414 }
415 }
416
417 splx(s);
418 }
419
420 static void
421 nd6_llinfo_timer(arg)
422 void *arg;
423 {
424 int s;
425 struct llinfo_nd6 *ln;
426 struct rtentry *rt;
427 const struct sockaddr_in6 *dst;
428 struct ifnet *ifp;
429 struct nd_ifinfo *ndi = NULL;
430
431 s = splsoftnet();
432
433 ln = (struct llinfo_nd6 *)arg;
434
435 if (ln->ln_ntick > 0) {
436 nd6_llinfo_settimer(ln, ln->ln_ntick);
437 splx(s);
438 return;
439 }
440
441 if ((rt = ln->ln_rt) == NULL)
442 panic("ln->ln_rt == NULL");
443 if ((ifp = rt->rt_ifp) == NULL)
444 panic("ln->ln_rt->rt_ifp == NULL");
445 ndi = ND_IFINFO(ifp);
446 dst = (struct sockaddr_in6 *)rt_key(rt);
447
448 /* sanity check */
449 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
450 panic("rt_llinfo(%p) is not equal to ln(%p)",
451 rt->rt_llinfo, ln);
452 if (!dst)
453 panic("dst=0 in nd6_timer(ln=%p)", ln);
454
455 switch (ln->ln_state) {
456 case ND6_LLINFO_INCOMPLETE:
457 if (ln->ln_asked < nd6_mmaxtries) {
458 ln->ln_asked++;
459 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
460 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
461 } else {
462 struct mbuf *m = ln->ln_hold;
463 if (m) {
464 struct mbuf *m0;
465
466 /*
467 * assuming every packet in ln_hold has
468 * the same IP header
469 */
470 m0 = m->m_nextpkt;
471 m->m_nextpkt = NULL;
472 icmp6_error2(m, ICMP6_DST_UNREACH,
473 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp);
474
475 ln->ln_hold = m0;
476 clear_llinfo_pqueue(ln);
477 }
478 (void)nd6_free(rt, 0);
479 ln = NULL;
480 }
481 break;
482 case ND6_LLINFO_REACHABLE:
483 if (!ND6_LLINFO_PERMANENT(ln)) {
484 ln->ln_state = ND6_LLINFO_STALE;
485 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
486 }
487 break;
488
489 case ND6_LLINFO_STALE:
490 /* Garbage Collection(RFC 2461 5.3) */
491 if (!ND6_LLINFO_PERMANENT(ln)) {
492 (void)nd6_free(rt, 1);
493 ln = NULL;
494 }
495 break;
496
497 case ND6_LLINFO_DELAY:
498 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
499 /* We need NUD */
500 ln->ln_asked = 1;
501 ln->ln_state = ND6_LLINFO_PROBE;
502 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
503 nd6_ns_output(ifp, &dst->sin6_addr,
504 &dst->sin6_addr, ln, 0);
505 } else {
506 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
507 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
508 }
509 break;
510 case ND6_LLINFO_PROBE:
511 if (ln->ln_asked < nd6_umaxtries) {
512 ln->ln_asked++;
513 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
514 nd6_ns_output(ifp, &dst->sin6_addr,
515 &dst->sin6_addr, ln, 0);
516 } else {
517 (void)nd6_free(rt, 0);
518 ln = NULL;
519 }
520 break;
521 }
522
523 splx(s);
524 }
525
526 /*
527 * ND6 timer routine to expire default route list and prefix list
528 */
529 void
530 nd6_timer(void *ignored_arg)
531 {
532 int s;
533 struct nd_defrouter *next_dr, *dr;
534 struct nd_prefix *next_pr, *pr;
535 struct in6_ifaddr *ia6, *nia6;
536 struct in6_addrlifetime *lt6;
537
538 s = splsoftnet();
539 callout_reset(&nd6_timer_ch, nd6_prune * hz,
540 nd6_timer, NULL);
541
542 /* expire default router list */
543
544 for (dr = TAILQ_FIRST(&nd_defrouter); dr != NULL; dr = next_dr) {
545 next_dr = TAILQ_NEXT(dr, dr_entry);
546 if (dr->expire && dr->expire < time_second) {
547 defrtrlist_del(dr);
548 }
549 }
550
551 /*
552 * expire interface addresses.
553 * in the past the loop was inside prefix expiry processing.
554 * However, from a stricter speci-confrmance standpoint, we should
555 * rather separate address lifetimes and prefix lifetimes.
556 */
557 addrloop:
558 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
559 nia6 = ia6->ia_next;
560 /* check address lifetime */
561 lt6 = &ia6->ia6_lifetime;
562 if (IFA6_IS_INVALID(ia6)) {
563 int regen = 0;
564
565 /*
566 * If the expiring address is temporary, try
567 * regenerating a new one. This would be useful when
568 * we suspended a laptop PC, then turned it on after a
569 * period that could invalidate all temporary
570 * addresses. Although we may have to restart the
571 * loop (see below), it must be after purging the
572 * address. Otherwise, we'd see an infinite loop of
573 * regeneration.
574 */
575 if (ip6_use_tempaddr &&
576 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
577 if (regen_tmpaddr(ia6) == 0)
578 regen = 1;
579 }
580
581 in6_purgeaddr(&ia6->ia_ifa);
582
583 if (regen)
584 goto addrloop; /* XXX: see below */
585 } else if (IFA6_IS_DEPRECATED(ia6)) {
586 int oldflags = ia6->ia6_flags;
587
588 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
589
590 /*
591 * If a temporary address has just become deprecated,
592 * regenerate a new one if possible.
593 */
594 if (ip6_use_tempaddr &&
595 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
596 (oldflags & IN6_IFF_DEPRECATED) == 0) {
597
598 if (regen_tmpaddr(ia6) == 0) {
599 /*
600 * A new temporary address is
601 * generated.
602 * XXX: this means the address chain
603 * has changed while we are still in
604 * the loop. Although the change
605 * would not cause disaster (because
606 * it's not a deletion, but an
607 * addition,) we'd rather restart the
608 * loop just for safety. Or does this
609 * significantly reduce performance??
610 */
611 goto addrloop;
612 }
613 }
614 } else {
615 /*
616 * A new RA might have made a deprecated address
617 * preferred.
618 */
619 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
620 }
621 }
622
623 /* expire prefix list */
624 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = next_pr) {
625 next_pr = LIST_NEXT(pr, ndpr_entry);
626 /*
627 * check prefix lifetime.
628 * since pltime is just for autoconf, pltime processing for
629 * prefix is not necessary.
630 */
631 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
632 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
633
634 /*
635 * address expiration and prefix expiration are
636 * separate. NEVER perform in6_purgeaddr here.
637 */
638
639 prelist_remove(pr);
640 }
641 }
642 splx(s);
643 }
644
645 static int
646 regen_tmpaddr(ia6)
647 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
648 {
649 struct ifaddr *ifa;
650 struct ifnet *ifp;
651 struct in6_ifaddr *public_ifa6 = NULL;
652
653 ifp = ia6->ia_ifa.ifa_ifp;
654 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
655 struct in6_ifaddr *it6;
656
657 if (ifa->ifa_addr->sa_family != AF_INET6)
658 continue;
659
660 it6 = (struct in6_ifaddr *)ifa;
661
662 /* ignore no autoconf addresses. */
663 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
664 continue;
665
666 /* ignore autoconf addresses with different prefixes. */
667 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
668 continue;
669
670 /*
671 * Now we are looking at an autoconf address with the same
672 * prefix as ours. If the address is temporary and is still
673 * preferred, do not create another one. It would be rare, but
674 * could happen, for example, when we resume a laptop PC after
675 * a long period.
676 */
677 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
678 !IFA6_IS_DEPRECATED(it6)) {
679 public_ifa6 = NULL;
680 break;
681 }
682
683 /*
684 * This is a public autoconf address that has the same prefix
685 * as ours. If it is preferred, keep it. We can't break the
686 * loop here, because there may be a still-preferred temporary
687 * address with the prefix.
688 */
689 if (!IFA6_IS_DEPRECATED(it6))
690 public_ifa6 = it6;
691 }
692
693 if (public_ifa6 != NULL) {
694 int e;
695
696 /*
697 * Random factor is introduced in the preferred lifetime, so
698 * we do not need additional delay (3rd arg to in6_tmpifadd).
699 */
700 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
701 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
702 " tmp addr, errno=%d\n", e);
703 return (-1);
704 }
705 return (0);
706 }
707
708 return (-1);
709 }
710
711 /*
712 * Nuke neighbor cache/prefix/default router management table, right before
713 * ifp goes away.
714 */
715 void
716 nd6_purge(ifp)
717 struct ifnet *ifp;
718 {
719 struct llinfo_nd6 *ln, *nln;
720 struct nd_defrouter *dr, *ndr;
721 struct nd_prefix *pr, *npr;
722
723 /*
724 * Nuke default router list entries toward ifp.
725 * We defer removal of default router list entries that is installed
726 * in the routing table, in order to keep additional side effects as
727 * small as possible.
728 */
729 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
730 ndr = TAILQ_NEXT(dr, dr_entry);
731 if (dr->installed)
732 continue;
733
734 if (dr->ifp == ifp)
735 defrtrlist_del(dr);
736 }
737 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
738 ndr = TAILQ_NEXT(dr, dr_entry);
739 if (!dr->installed)
740 continue;
741
742 if (dr->ifp == ifp)
743 defrtrlist_del(dr);
744 }
745
746 /* Nuke prefix list entries toward ifp */
747 for (pr = LIST_FIRST(&nd_prefix); pr != NULL; pr = npr) {
748 npr = LIST_NEXT(pr, ndpr_entry);
749 if (pr->ndpr_ifp == ifp) {
750 /*
751 * Because if_detach() does *not* release prefixes
752 * while purging addresses the reference count will
753 * still be above zero. We therefore reset it to
754 * make sure that the prefix really gets purged.
755 */
756 pr->ndpr_refcnt = 0;
757 /*
758 * Previously, pr->ndpr_addr is removed as well,
759 * but I strongly believe we don't have to do it.
760 * nd6_purge() is only called from in6_ifdetach(),
761 * which removes all the associated interface addresses
762 * by itself.
763 * (jinmei@kame.net 20010129)
764 */
765 prelist_remove(pr);
766 }
767 }
768
769 /* cancel default outgoing interface setting */
770 if (nd6_defifindex == ifp->if_index)
771 nd6_setdefaultiface(0);
772
773 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
774 /* refresh default router list */
775 defrouter_select();
776 }
777
778 /*
779 * Nuke neighbor cache entries for the ifp.
780 * Note that rt->rt_ifp may not be the same as ifp,
781 * due to KAME goto ours hack. See RTM_RESOLVE case in
782 * nd6_rtrequest(), and ip6_input().
783 */
784 ln = llinfo_nd6.ln_next;
785 while (ln && ln != &llinfo_nd6) {
786 struct rtentry *rt;
787 struct sockaddr_dl *sdl;
788
789 nln = ln->ln_next;
790 rt = ln->ln_rt;
791 if (rt && rt->rt_gateway &&
792 rt->rt_gateway->sa_family == AF_LINK) {
793 sdl = (struct sockaddr_dl *)rt->rt_gateway;
794 if (sdl->sdl_index == ifp->if_index)
795 nln = nd6_free(rt, 0);
796 }
797 ln = nln;
798 }
799 }
800
801 struct rtentry *
802 nd6_lookup(addr6, create, ifp)
803 struct in6_addr *addr6;
804 int create;
805 struct ifnet *ifp;
806 {
807 struct rtentry *rt;
808 struct sockaddr_in6 sin6;
809
810 bzero(&sin6, sizeof(sin6));
811 sin6.sin6_len = sizeof(struct sockaddr_in6);
812 sin6.sin6_family = AF_INET6;
813 sin6.sin6_addr = *addr6;
814 rt = rtalloc1((struct sockaddr *)&sin6, create);
815 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
816 /*
817 * This is the case for the default route.
818 * If we want to create a neighbor cache for the address, we
819 * should free the route for the destination and allocate an
820 * interface route.
821 */
822 if (create) {
823 RTFREE(rt);
824 rt = NULL;
825 }
826 }
827 if (rt == NULL) {
828 if (create && ifp) {
829 int e;
830
831 /*
832 * If no route is available and create is set,
833 * we allocate a host route for the destination
834 * and treat it like an interface route.
835 * This hack is necessary for a neighbor which can't
836 * be covered by our own prefix.
837 */
838 struct ifaddr *ifa =
839 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
840 if (ifa == NULL)
841 return (NULL);
842
843 /*
844 * Create a new route. RTF_LLINFO is necessary
845 * to create a Neighbor Cache entry for the
846 * destination in nd6_rtrequest which will be
847 * called in rtrequest via ifa->ifa_rtrequest.
848 */
849 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
850 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
851 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
852 ~RTF_CLONING, &rt)) != 0) {
853 #if 0
854 log(LOG_ERR,
855 "nd6_lookup: failed to add route for a "
856 "neighbor(%s), errno=%d\n",
857 ip6_sprintf(addr6), e);
858 #endif
859 return (NULL);
860 }
861 if (rt == NULL)
862 return (NULL);
863 if (rt->rt_llinfo) {
864 struct llinfo_nd6 *ln =
865 (struct llinfo_nd6 *)rt->rt_llinfo;
866 ln->ln_state = ND6_LLINFO_NOSTATE;
867 }
868 } else
869 return (NULL);
870 }
871 rt->rt_refcnt--;
872 /*
873 * Validation for the entry.
874 * Note that the check for rt_llinfo is necessary because a cloned
875 * route from a parent route that has the L flag (e.g. the default
876 * route to a p2p interface) may have the flag, too, while the
877 * destination is not actually a neighbor.
878 * XXX: we can't use rt->rt_ifp to check for the interface, since
879 * it might be the loopback interface if the entry is for our
880 * own address on a non-loopback interface. Instead, we should
881 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
882 * interface.
883 * Note also that ifa_ifp and ifp may differ when we connect two
884 * interfaces to a same link, install a link prefix to an interface,
885 * and try to install a neighbor cache on an interface that does not
886 * have a route to the prefix.
887 */
888 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
889 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
890 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
891 if (create) {
892 nd6log((LOG_DEBUG,
893 "nd6_lookup: failed to lookup %s (if = %s)\n",
894 ip6_sprintf(addr6),
895 ifp ? if_name(ifp) : "unspec"));
896 }
897 return (NULL);
898 }
899 return (rt);
900 }
901
902 /*
903 * Detect if a given IPv6 address identifies a neighbor on a given link.
904 * XXX: should take care of the destination of a p2p link?
905 */
906 int
907 nd6_is_addr_neighbor(addr, ifp)
908 struct sockaddr_in6 *addr;
909 struct ifnet *ifp;
910 {
911 struct nd_prefix *pr;
912
913 /*
914 * A link-local address is always a neighbor.
915 * XXX: a link does not necessarily specify a single interface.
916 */
917 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
918 struct sockaddr_in6 sin6_copy;
919 u_int32_t zone;
920
921 /*
922 * We need sin6_copy since sa6_recoverscope() may modify the
923 * content (XXX).
924 */
925 sin6_copy = *addr;
926 if (sa6_recoverscope(&sin6_copy))
927 return (0); /* XXX: should be impossible */
928 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
929 return (0);
930 if (sin6_copy.sin6_scope_id == zone)
931 return (1);
932 else
933 return (0);
934 }
935
936 /*
937 * If the address matches one of our on-link prefixes, it should be a
938 * neighbor.
939 */
940 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
941 if (pr->ndpr_ifp != ifp)
942 continue;
943
944 if (!(pr->ndpr_stateflags & NDPRF_ONLINK))
945 continue;
946
947 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
948 &addr->sin6_addr, &pr->ndpr_mask))
949 return (1);
950 }
951
952 /*
953 * If the default router list is empty, all addresses are regarded
954 * as on-link, and thus, as a neighbor.
955 * XXX: we restrict the condition to hosts, because routers usually do
956 * not have the "default router list".
957 */
958 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL &&
959 nd6_defifindex == ifp->if_index) {
960 return (1);
961 }
962
963 /*
964 * Even if the address matches none of our addresses, it might be
965 * in the neighbor cache.
966 */
967 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
968 return (1);
969
970 return (0);
971 }
972
973 /*
974 * Free an nd6 llinfo entry.
975 * Since the function would cause significant changes in the kernel, DO NOT
976 * make it global, unless you have a strong reason for the change, and are sure
977 * that the change is safe.
978 */
979 static struct llinfo_nd6 *
980 nd6_free(rt, gc)
981 struct rtentry *rt;
982 int gc;
983 {
984 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
985 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
986 struct nd_defrouter *dr;
987
988 /*
989 * we used to have pfctlinput(PRC_HOSTDEAD) here.
990 * even though it is not harmful, it was not really necessary.
991 */
992
993 /* cancel timer */
994 nd6_llinfo_settimer(ln, -1);
995
996 if (!ip6_forwarding) {
997 int s;
998 s = splsoftnet();
999 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1000 rt->rt_ifp);
1001
1002 if (dr != NULL && dr->expire &&
1003 ln->ln_state == ND6_LLINFO_STALE && gc) {
1004 /*
1005 * If the reason for the deletion is just garbage
1006 * collection, and the neighbor is an active default
1007 * router, do not delete it. Instead, reset the GC
1008 * timer using the router's lifetime.
1009 * Simply deleting the entry would affect default
1010 * router selection, which is not necessarily a good
1011 * thing, especially when we're using router preference
1012 * values.
1013 * XXX: the check for ln_state would be redundant,
1014 * but we intentionally keep it just in case.
1015 */
1016 if (dr->expire > time_second)
1017 nd6_llinfo_settimer(ln,
1018 (dr->expire - time_second) * hz);
1019 else
1020 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1021 splx(s);
1022 return (ln->ln_next);
1023 }
1024
1025 if (ln->ln_router || dr) {
1026 /*
1027 * rt6_flush must be called whether or not the neighbor
1028 * is in the Default Router List.
1029 * See a corresponding comment in nd6_na_input().
1030 */
1031 rt6_flush(&in6, rt->rt_ifp);
1032 }
1033
1034 if (dr) {
1035 /*
1036 * Unreachablity of a router might affect the default
1037 * router selection and on-link detection of advertised
1038 * prefixes.
1039 */
1040
1041 /*
1042 * Temporarily fake the state to choose a new default
1043 * router and to perform on-link determination of
1044 * prefixes correctly.
1045 * Below the state will be set correctly,
1046 * or the entry itself will be deleted.
1047 */
1048 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1049
1050 /*
1051 * Since defrouter_select() does not affect the
1052 * on-link determination and MIP6 needs the check
1053 * before the default router selection, we perform
1054 * the check now.
1055 */
1056 pfxlist_onlink_check();
1057
1058 /*
1059 * refresh default router list
1060 */
1061 defrouter_select();
1062 }
1063 splx(s);
1064 }
1065
1066 /*
1067 * Before deleting the entry, remember the next entry as the
1068 * return value. We need this because pfxlist_onlink_check() above
1069 * might have freed other entries (particularly the old next entry) as
1070 * a side effect (XXX).
1071 */
1072 next = ln->ln_next;
1073
1074 /*
1075 * Detach the route from the routing tree and the list of neighbor
1076 * caches, and disable the route entry not to be used in already
1077 * cached routes.
1078 */
1079 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1080 rt_mask(rt), 0, (struct rtentry **)0);
1081
1082 return (next);
1083 }
1084
1085 /*
1086 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1087 *
1088 * XXX cost-effective methods?
1089 */
1090 void
1091 nd6_nud_hint(rt, dst6, force)
1092 struct rtentry *rt;
1093 struct in6_addr *dst6;
1094 int force;
1095 {
1096 struct llinfo_nd6 *ln;
1097
1098 /*
1099 * If the caller specified "rt", use that. Otherwise, resolve the
1100 * routing table by supplied "dst6".
1101 */
1102 if (rt == NULL) {
1103 if (dst6 == NULL)
1104 return;
1105 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL)
1106 return;
1107 }
1108
1109 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1110 (rt->rt_flags & RTF_LLINFO) == 0 ||
1111 !rt->rt_llinfo || !rt->rt_gateway ||
1112 rt->rt_gateway->sa_family != AF_LINK) {
1113 /* This is not a host route. */
1114 return;
1115 }
1116
1117 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1118 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1119 return;
1120
1121 /*
1122 * if we get upper-layer reachability confirmation many times,
1123 * it is possible we have false information.
1124 */
1125 if (!force) {
1126 ln->ln_byhint++;
1127 if (ln->ln_byhint > nd6_maxnudhint)
1128 return;
1129 }
1130
1131 ln->ln_state = ND6_LLINFO_REACHABLE;
1132 if (!ND6_LLINFO_PERMANENT(ln)) {
1133 nd6_llinfo_settimer(ln,
1134 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1135 }
1136 }
1137
1138 void
1139 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1140 {
1141 struct sockaddr *gate = rt->rt_gateway;
1142 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1143 static const struct sockaddr_dl null_sdl = {
1144 .sdl_len = sizeof(null_sdl),
1145 .sdl_family = AF_LINK,
1146 };
1147 struct ifnet *ifp = rt->rt_ifp;
1148 struct ifaddr *ifa;
1149
1150 if ((rt->rt_flags & RTF_GATEWAY) != 0)
1151 return;
1152
1153 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1154 /*
1155 * This is probably an interface direct route for a link
1156 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1157 * We do not need special treatment below for such a route.
1158 * Moreover, the RTF_LLINFO flag which would be set below
1159 * would annoy the ndp(8) command.
1160 */
1161 return;
1162 }
1163
1164 if (req == RTM_RESOLVE &&
1165 (nd6_need_cache(ifp) == 0 || /* stf case */
1166 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
1167 /*
1168 * FreeBSD and BSD/OS often make a cloned host route based
1169 * on a less-specific route (e.g. the default route).
1170 * If the less specific route does not have a "gateway"
1171 * (this is the case when the route just goes to a p2p or an
1172 * stf interface), we'll mistakenly make a neighbor cache for
1173 * the host route, and will see strange neighbor solicitation
1174 * for the corresponding destination. In order to avoid the
1175 * confusion, we check if the destination of the route is
1176 * a neighbor in terms of neighbor discovery, and stop the
1177 * process if not. Additionally, we remove the LLINFO flag
1178 * so that ndp(8) will not try to get the neighbor information
1179 * of the destination.
1180 */
1181 rt->rt_flags &= ~RTF_LLINFO;
1182 return;
1183 }
1184
1185 switch (req) {
1186 case RTM_ADD:
1187 /*
1188 * There is no backward compatibility :)
1189 *
1190 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1191 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1192 * rt->rt_flags |= RTF_CLONING;
1193 */
1194 if ((rt->rt_flags & RTF_CLONING) ||
1195 ((rt->rt_flags & RTF_LLINFO) && !ln)) {
1196 /*
1197 * Case 1: This route should come from a route to
1198 * interface (RTF_CLONING case) or the route should be
1199 * treated as on-link but is currently not
1200 * (RTF_LLINFO && !ln case).
1201 */
1202 rt_setgate(rt, rt_key(rt),
1203 (const struct sockaddr *)&null_sdl);
1204 gate = rt->rt_gateway;
1205 SDL(gate)->sdl_type = ifp->if_type;
1206 SDL(gate)->sdl_index = ifp->if_index;
1207 if (ln)
1208 nd6_llinfo_settimer(ln, 0);
1209 if ((rt->rt_flags & RTF_CLONING) != 0)
1210 break;
1211 }
1212 /*
1213 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1214 * We don't do that here since llinfo is not ready yet.
1215 *
1216 * There are also couple of other things to be discussed:
1217 * - unsolicited NA code needs improvement beforehand
1218 * - RFC2461 says we MAY send multicast unsolicited NA
1219 * (7.2.6 paragraph 4), however, it also says that we
1220 * SHOULD provide a mechanism to prevent multicast NA storm.
1221 * we don't have anything like it right now.
1222 * note that the mechanism needs a mutual agreement
1223 * between proxies, which means that we need to implement
1224 * a new protocol, or a new kludge.
1225 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1226 * we need to check ip6forwarding before sending it.
1227 * (or should we allow proxy ND configuration only for
1228 * routers? there's no mention about proxy ND from hosts)
1229 */
1230 #if 0
1231 /* XXX it does not work */
1232 if (rt->rt_flags & RTF_ANNOUNCE)
1233 nd6_na_output(ifp,
1234 &SIN6(rt_key(rt))->sin6_addr,
1235 &SIN6(rt_key(rt))->sin6_addr,
1236 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1237 1, NULL);
1238 #endif
1239 /* FALLTHROUGH */
1240 case RTM_RESOLVE:
1241 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1242 /*
1243 * Address resolution isn't necessary for a point to
1244 * point link, so we can skip this test for a p2p link.
1245 */
1246 if (gate->sa_family != AF_LINK ||
1247 gate->sa_len < sizeof(null_sdl)) {
1248 log(LOG_DEBUG,
1249 "nd6_rtrequest: bad gateway value: %s\n",
1250 if_name(ifp));
1251 break;
1252 }
1253 SDL(gate)->sdl_type = ifp->if_type;
1254 SDL(gate)->sdl_index = ifp->if_index;
1255 }
1256 if (ln != NULL)
1257 break; /* This happens on a route change */
1258 /*
1259 * Case 2: This route may come from cloning, or a manual route
1260 * add with a LL address.
1261 */
1262 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1263 rt->rt_llinfo = (caddr_t)ln;
1264 if (ln == NULL) {
1265 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1266 break;
1267 }
1268 nd6_inuse++;
1269 nd6_allocated++;
1270 bzero(ln, sizeof(*ln));
1271 ln->ln_rt = rt;
1272 callout_init(&ln->ln_timer_ch);
1273 /* this is required for "ndp" command. - shin */
1274 if (req == RTM_ADD) {
1275 /*
1276 * gate should have some valid AF_LINK entry,
1277 * and ln->ln_expire should have some lifetime
1278 * which is specified by ndp command.
1279 */
1280 ln->ln_state = ND6_LLINFO_REACHABLE;
1281 ln->ln_byhint = 0;
1282 } else {
1283 /*
1284 * When req == RTM_RESOLVE, rt is created and
1285 * initialized in rtrequest(), so rt_expire is 0.
1286 */
1287 ln->ln_state = ND6_LLINFO_NOSTATE;
1288 nd6_llinfo_settimer(ln, 0);
1289 }
1290 rt->rt_flags |= RTF_LLINFO;
1291 ln->ln_next = llinfo_nd6.ln_next;
1292 llinfo_nd6.ln_next = ln;
1293 ln->ln_prev = &llinfo_nd6;
1294 ln->ln_next->ln_prev = ln;
1295
1296 /*
1297 * check if rt_key(rt) is one of my address assigned
1298 * to the interface.
1299 */
1300 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1301 &SIN6(rt_key(rt))->sin6_addr);
1302 if (ifa) {
1303 caddr_t macp = nd6_ifptomac(ifp);
1304 nd6_llinfo_settimer(ln, -1);
1305 ln->ln_state = ND6_LLINFO_REACHABLE;
1306 ln->ln_byhint = 0;
1307 if (macp) {
1308 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1309 SDL(gate)->sdl_alen = ifp->if_addrlen;
1310 }
1311 if (nd6_useloopback) {
1312 rt->rt_ifp = lo0ifp; /* XXX */
1313 /*
1314 * Make sure rt_ifa be equal to the ifaddr
1315 * corresponding to the address.
1316 * We need this because when we refer
1317 * rt_ifa->ia6_flags in ip6_input, we assume
1318 * that the rt_ifa points to the address instead
1319 * of the loopback address.
1320 */
1321 if (ifa != rt->rt_ifa)
1322 rt_replace_ifa(rt, ifa);
1323 }
1324 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1325 nd6_llinfo_settimer(ln, -1);
1326 ln->ln_state = ND6_LLINFO_REACHABLE;
1327 ln->ln_byhint = 0;
1328
1329 /* join solicited node multicast for proxy ND */
1330 if (ifp->if_flags & IFF_MULTICAST) {
1331 struct in6_addr llsol;
1332 int error;
1333
1334 llsol = SIN6(rt_key(rt))->sin6_addr;
1335 llsol.s6_addr32[0] = htonl(0xff020000);
1336 llsol.s6_addr32[1] = 0;
1337 llsol.s6_addr32[2] = htonl(1);
1338 llsol.s6_addr8[12] = 0xff;
1339 if (in6_setscope(&llsol, ifp, NULL))
1340 break;
1341 if (!in6_addmulti(&llsol, ifp, &error, 0)) {
1342 nd6log((LOG_ERR, "%s: failed to join "
1343 "%s (errno=%d)\n", if_name(ifp),
1344 ip6_sprintf(&llsol), error));
1345 }
1346 }
1347 }
1348 break;
1349
1350 case RTM_DELETE:
1351 if (ln == NULL)
1352 break;
1353 /* leave from solicited node multicast for proxy ND */
1354 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1355 (ifp->if_flags & IFF_MULTICAST) != 0) {
1356 struct in6_addr llsol;
1357 struct in6_multi *in6m;
1358
1359 llsol = SIN6(rt_key(rt))->sin6_addr;
1360 llsol.s6_addr32[0] = htonl(0xff020000);
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) == 0) {
1365 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1366 if (in6m)
1367 in6_delmulti(in6m);
1368 }
1369 }
1370 nd6_inuse--;
1371 ln->ln_next->ln_prev = ln->ln_prev;
1372 ln->ln_prev->ln_next = ln->ln_next;
1373 ln->ln_prev = NULL;
1374 nd6_llinfo_settimer(ln, -1);
1375 rt->rt_llinfo = 0;
1376 rt->rt_flags &= ~RTF_LLINFO;
1377 clear_llinfo_pqueue(ln);
1378 Free((caddr_t)ln);
1379 }
1380 }
1381
1382 int
1383 nd6_ioctl(cmd, data, ifp)
1384 u_long cmd;
1385 caddr_t data;
1386 struct ifnet *ifp;
1387 {
1388 struct in6_drlist *drl = (struct in6_drlist *)data;
1389 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1390 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1391 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1392 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1393 struct nd_defrouter *dr;
1394 struct nd_prefix *pr;
1395 struct rtentry *rt;
1396 int i = 0, error = 0;
1397 int s;
1398
1399 switch (cmd) {
1400 case SIOCGDRLST_IN6:
1401 /*
1402 * obsolete API, use sysctl under net.inet6.icmp6
1403 */
1404 bzero(drl, sizeof(*drl));
1405 s = splsoftnet();
1406 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1407 if (i >= DRLSTSIZ)
1408 break;
1409 drl->defrouter[i].rtaddr = dr->rtaddr;
1410 in6_clearscope(&drl->defrouter[i].rtaddr);
1411
1412 drl->defrouter[i].flags = dr->flags;
1413 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1414 drl->defrouter[i].expire = dr->expire;
1415 drl->defrouter[i].if_index = dr->ifp->if_index;
1416 i++;
1417 }
1418 splx(s);
1419 break;
1420 case SIOCGPRLST_IN6:
1421 /*
1422 * obsolete API, use sysctl under net.inet6.icmp6
1423 *
1424 * XXX the structure in6_prlist was changed in backward-
1425 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1426 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1427 */
1428 /*
1429 * XXX meaning of fields, especialy "raflags", is very
1430 * differnet between RA prefix list and RR/static prefix list.
1431 * how about separating ioctls into two?
1432 */
1433 bzero(oprl, sizeof(*oprl));
1434 s = splsoftnet();
1435 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1436 struct nd_pfxrouter *pfr;
1437 int j;
1438
1439 if (i >= PRLSTSIZ)
1440 break;
1441 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1442 oprl->prefix[i].raflags = pr->ndpr_raf;
1443 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1444 oprl->prefix[i].vltime = pr->ndpr_vltime;
1445 oprl->prefix[i].pltime = pr->ndpr_pltime;
1446 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1447 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1448 oprl->prefix[i].expire = 0;
1449 else {
1450 time_t maxexpire;
1451
1452 /* XXX: we assume time_t is signed. */
1453 maxexpire = (-1) &
1454 ~((time_t)1 <<
1455 ((sizeof(maxexpire) * 8) - 1));
1456 if (pr->ndpr_vltime <
1457 maxexpire - pr->ndpr_lastupdate) {
1458 oprl->prefix[i].expire =
1459 pr->ndpr_lastupdate +
1460 pr->ndpr_vltime;
1461 } else
1462 oprl->prefix[i].expire = maxexpire;
1463 }
1464
1465 j = 0;
1466 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1467 if (j < DRLSTSIZ) {
1468 #define RTRADDR oprl->prefix[i].advrtr[j]
1469 RTRADDR = pfr->router->rtaddr;
1470 in6_clearscope(&RTRADDR);
1471 #undef RTRADDR
1472 }
1473 j++;
1474 }
1475 oprl->prefix[i].advrtrs = j;
1476 oprl->prefix[i].origin = PR_ORIG_RA;
1477
1478 i++;
1479 }
1480 splx(s);
1481
1482 break;
1483 case OSIOCGIFINFO_IN6:
1484 #define ND ndi->ndi
1485 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1486 memset(&ND, 0, sizeof(ND));
1487 ND.linkmtu = IN6_LINKMTU(ifp);
1488 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1489 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1490 ND.reachable = ND_IFINFO(ifp)->reachable;
1491 ND.retrans = ND_IFINFO(ifp)->retrans;
1492 ND.flags = ND_IFINFO(ifp)->flags;
1493 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1494 ND.chlim = ND_IFINFO(ifp)->chlim;
1495 break;
1496 case SIOCGIFINFO_IN6:
1497 ND = *ND_IFINFO(ifp);
1498 break;
1499 case SIOCSIFINFO_IN6:
1500 /*
1501 * used to change host variables from userland.
1502 * intented for a use on router to reflect RA configurations.
1503 */
1504 /* 0 means 'unspecified' */
1505 if (ND.linkmtu != 0) {
1506 if (ND.linkmtu < IPV6_MMTU ||
1507 ND.linkmtu > IN6_LINKMTU(ifp)) {
1508 error = EINVAL;
1509 break;
1510 }
1511 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1512 }
1513
1514 if (ND.basereachable != 0) {
1515 int obasereachable = ND_IFINFO(ifp)->basereachable;
1516
1517 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1518 if (ND.basereachable != obasereachable)
1519 ND_IFINFO(ifp)->reachable =
1520 ND_COMPUTE_RTIME(ND.basereachable);
1521 }
1522 if (ND.retrans != 0)
1523 ND_IFINFO(ifp)->retrans = ND.retrans;
1524 if (ND.chlim != 0)
1525 ND_IFINFO(ifp)->chlim = ND.chlim;
1526 /* FALLTHROUGH */
1527 case SIOCSIFINFO_FLAGS:
1528 ND_IFINFO(ifp)->flags = ND.flags;
1529 break;
1530 #undef ND
1531 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1532 /* sync kernel routing table with the default router list */
1533 defrouter_reset();
1534 defrouter_select();
1535 break;
1536 case SIOCSPFXFLUSH_IN6:
1537 {
1538 /* flush all the prefix advertised by routers */
1539 struct nd_prefix *pfx, *next;
1540
1541 s = splsoftnet();
1542 for (pfx = LIST_FIRST(&nd_prefix); pfx; pfx = next) {
1543 struct in6_ifaddr *ia, *ia_next;
1544
1545 next = LIST_NEXT(pfx, ndpr_entry);
1546
1547 if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr))
1548 continue; /* XXX */
1549
1550 /* do we really have to remove addresses as well? */
1551 for (ia = in6_ifaddr; ia; ia = ia_next) {
1552 /* ia might be removed. keep the next ptr. */
1553 ia_next = ia->ia_next;
1554
1555 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1556 continue;
1557
1558 if (ia->ia6_ndpr == pfx)
1559 in6_purgeaddr(&ia->ia_ifa);
1560 }
1561 prelist_remove(pfx);
1562 }
1563 splx(s);
1564 break;
1565 }
1566 case SIOCSRTRFLUSH_IN6:
1567 {
1568 /* flush all the default routers */
1569 struct nd_defrouter *drtr, *next;
1570
1571 s = splsoftnet();
1572 defrouter_reset();
1573 for (drtr = TAILQ_FIRST(&nd_defrouter); drtr; drtr = next) {
1574 next = TAILQ_NEXT(drtr, dr_entry);
1575 defrtrlist_del(drtr);
1576 }
1577 defrouter_select();
1578 splx(s);
1579 break;
1580 }
1581 case SIOCGNBRINFO_IN6:
1582 {
1583 struct llinfo_nd6 *ln;
1584 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1585
1586 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1587 return (error);
1588
1589 s = splsoftnet();
1590 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL ||
1591 (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) {
1592 error = EINVAL;
1593 splx(s);
1594 break;
1595 }
1596 nbi->state = ln->ln_state;
1597 nbi->asked = ln->ln_asked;
1598 nbi->isrouter = ln->ln_router;
1599 nbi->expire = ln->ln_expire;
1600 splx(s);
1601
1602 break;
1603 }
1604 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1605 ndif->ifindex = nd6_defifindex;
1606 break;
1607 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1608 return (nd6_setdefaultiface(ndif->ifindex));
1609 }
1610 return (error);
1611 }
1612
1613 /*
1614 * Create neighbor cache entry and cache link-layer address,
1615 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1616 */
1617 struct rtentry *
1618 nd6_cache_lladdr(
1619 struct ifnet *ifp,
1620 struct in6_addr *from,
1621 char *lladdr,
1622 int lladdrlen,
1623 int type, /* ICMP6 type */
1624 int code /* type dependent information */
1625 )
1626 {
1627 struct rtentry *rt = NULL;
1628 struct llinfo_nd6 *ln = NULL;
1629 int is_newentry;
1630 struct sockaddr_dl *sdl = NULL;
1631 int do_update;
1632 int olladdr;
1633 int llchange;
1634 int newstate = 0;
1635
1636 if (ifp == NULL)
1637 panic("ifp == NULL in nd6_cache_lladdr");
1638 if (from == NULL)
1639 panic("from == NULL in nd6_cache_lladdr");
1640
1641 /* nothing must be updated for unspecified address */
1642 if (IN6_IS_ADDR_UNSPECIFIED(from))
1643 return NULL;
1644
1645 /*
1646 * Validation about ifp->if_addrlen and lladdrlen must be done in
1647 * the caller.
1648 *
1649 * XXX If the link does not have link-layer adderss, what should
1650 * we do? (ifp->if_addrlen == 0)
1651 * Spec says nothing in sections for RA, RS and NA. There's small
1652 * description on it in NS section (RFC 2461 7.2.3).
1653 */
1654
1655 rt = nd6_lookup(from, 0, ifp);
1656 if (rt == NULL) {
1657 #if 0
1658 /* nothing must be done if there's no lladdr */
1659 if (!lladdr || !lladdrlen)
1660 return NULL;
1661 #endif
1662
1663 rt = nd6_lookup(from, 1, ifp);
1664 is_newentry = 1;
1665 } else {
1666 /* do nothing if static ndp is set */
1667 if (rt->rt_flags & RTF_STATIC)
1668 return NULL;
1669 is_newentry = 0;
1670 }
1671
1672 if (rt == NULL)
1673 return NULL;
1674 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1675 fail:
1676 (void)nd6_free(rt, 0);
1677 return NULL;
1678 }
1679 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1680 if (ln == NULL)
1681 goto fail;
1682 if (rt->rt_gateway == NULL)
1683 goto fail;
1684 if (rt->rt_gateway->sa_family != AF_LINK)
1685 goto fail;
1686 sdl = SDL(rt->rt_gateway);
1687
1688 olladdr = (sdl->sdl_alen) ? 1 : 0;
1689 if (olladdr && lladdr) {
1690 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1691 llchange = 1;
1692 else
1693 llchange = 0;
1694 } else
1695 llchange = 0;
1696
1697 /*
1698 * newentry olladdr lladdr llchange (*=record)
1699 * 0 n n -- (1)
1700 * 0 y n -- (2)
1701 * 0 n y -- (3) * STALE
1702 * 0 y y n (4) *
1703 * 0 y y y (5) * STALE
1704 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1705 * 1 -- y -- (7) * STALE
1706 */
1707
1708 if (lladdr) { /* (3-5) and (7) */
1709 /*
1710 * Record source link-layer address
1711 * XXX is it dependent to ifp->if_type?
1712 */
1713 sdl->sdl_alen = ifp->if_addrlen;
1714 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1715 }
1716
1717 if (!is_newentry) {
1718 if ((!olladdr && lladdr) || /* (3) */
1719 (olladdr && lladdr && llchange)) { /* (5) */
1720 do_update = 1;
1721 newstate = ND6_LLINFO_STALE;
1722 } else /* (1-2,4) */
1723 do_update = 0;
1724 } else {
1725 do_update = 1;
1726 if (lladdr == NULL) /* (6) */
1727 newstate = ND6_LLINFO_NOSTATE;
1728 else /* (7) */
1729 newstate = ND6_LLINFO_STALE;
1730 }
1731
1732 if (do_update) {
1733 /*
1734 * Update the state of the neighbor cache.
1735 */
1736 ln->ln_state = newstate;
1737
1738 if (ln->ln_state == ND6_LLINFO_STALE) {
1739 /*
1740 * XXX: since nd6_output() below will cause
1741 * state tansition to DELAY and reset the timer,
1742 * we must set the timer now, although it is actually
1743 * meaningless.
1744 */
1745 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1746
1747 if (ln->ln_hold) {
1748 struct mbuf *m_hold, *m_hold_next;
1749 for (m_hold = ln->ln_hold; m_hold;
1750 m_hold = m_hold_next) {
1751 struct mbuf *mpkt = NULL;
1752
1753 m_hold_next = m_hold->m_nextpkt;
1754 mpkt = m_copym(m_hold, 0, M_COPYALL, M_DONTWAIT);
1755 if (mpkt == NULL) {
1756 m_freem(m_hold);
1757 break;
1758 }
1759 mpkt->m_nextpkt = NULL;
1760
1761 /*
1762 * we assume ifp is not a p2p here, so
1763 * just set the 2nd argument as the
1764 * 1st one.
1765 */
1766 nd6_output(ifp, ifp, mpkt,
1767 (struct sockaddr_in6 *)rt_key(rt),
1768 rt);
1769 }
1770 ln->ln_hold = NULL;
1771 }
1772 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1773 /* probe right away */
1774 nd6_llinfo_settimer((void *)ln, 0);
1775 }
1776 }
1777
1778 /*
1779 * ICMP6 type dependent behavior.
1780 *
1781 * NS: clear IsRouter if new entry
1782 * RS: clear IsRouter
1783 * RA: set IsRouter if there's lladdr
1784 * redir: clear IsRouter if new entry
1785 *
1786 * RA case, (1):
1787 * The spec says that we must set IsRouter in the following cases:
1788 * - If lladdr exist, set IsRouter. This means (1-5).
1789 * - If it is old entry (!newentry), set IsRouter. This means (7).
1790 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1791 * A quetion arises for (1) case. (1) case has no lladdr in the
1792 * neighbor cache, this is similar to (6).
1793 * This case is rare but we figured that we MUST NOT set IsRouter.
1794 *
1795 * newentry olladdr lladdr llchange NS RS RA redir
1796 * D R
1797 * 0 n n -- (1) c ? s
1798 * 0 y n -- (2) c s s
1799 * 0 n y -- (3) c s s
1800 * 0 y y n (4) c s s
1801 * 0 y y y (5) c s s
1802 * 1 -- n -- (6) c c c s
1803 * 1 -- y -- (7) c c s c s
1804 *
1805 * (c=clear s=set)
1806 */
1807 switch (type & 0xff) {
1808 case ND_NEIGHBOR_SOLICIT:
1809 /*
1810 * New entry must have is_router flag cleared.
1811 */
1812 if (is_newentry) /* (6-7) */
1813 ln->ln_router = 0;
1814 break;
1815 case ND_REDIRECT:
1816 /*
1817 * If the icmp is a redirect to a better router, always set the
1818 * is_router flag. Otherwise, if the entry is newly created,
1819 * clear the flag. [RFC 2461, sec 8.3]
1820 */
1821 if (code == ND_REDIRECT_ROUTER)
1822 ln->ln_router = 1;
1823 else if (is_newentry) /* (6-7) */
1824 ln->ln_router = 0;
1825 break;
1826 case ND_ROUTER_SOLICIT:
1827 /*
1828 * is_router flag must always be cleared.
1829 */
1830 ln->ln_router = 0;
1831 break;
1832 case ND_ROUTER_ADVERT:
1833 /*
1834 * Mark an entry with lladdr as a router.
1835 */
1836 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1837 (is_newentry && lladdr)) { /* (7) */
1838 ln->ln_router = 1;
1839 }
1840 break;
1841 }
1842
1843 /*
1844 * When the link-layer address of a router changes, select the
1845 * best router again. In particular, when the neighbor entry is newly
1846 * created, it might affect the selection policy.
1847 * Question: can we restrict the first condition to the "is_newentry"
1848 * case?
1849 * XXX: when we hear an RA from a new router with the link-layer
1850 * address option, defrouter_select() is called twice, since
1851 * defrtrlist_update called the function as well. However, I believe
1852 * we can compromise the overhead, since it only happens the first
1853 * time.
1854 * XXX: although defrouter_select() should not have a bad effect
1855 * for those are not autoconfigured hosts, we explicitly avoid such
1856 * cases for safety.
1857 */
1858 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1859 defrouter_select();
1860
1861 return rt;
1862 }
1863
1864 static void
1865 nd6_slowtimo(void *ignored_arg)
1866 {
1867 int s = splsoftnet();
1868 struct nd_ifinfo *nd6if;
1869 struct ifnet *ifp;
1870
1871 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1872 nd6_slowtimo, NULL);
1873 TAILQ_FOREACH(ifp, &ifnet, if_list) {
1874 nd6if = ND_IFINFO(ifp);
1875 if (nd6if->basereachable && /* already initialized */
1876 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1877 /*
1878 * Since reachable time rarely changes by router
1879 * advertisements, we SHOULD insure that a new random
1880 * value gets recomputed at least once every few hours.
1881 * (RFC 2461, 6.3.4)
1882 */
1883 nd6if->recalctm = nd6_recalc_reachtm_interval;
1884 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1885 }
1886 }
1887 splx(s);
1888 }
1889
1890 #define senderr(e) { error = (e); goto bad;}
1891 int
1892 nd6_output(ifp, origifp, m0, dst, rt0)
1893 struct ifnet *ifp;
1894 struct ifnet *origifp;
1895 struct mbuf *m0;
1896 struct sockaddr_in6 *dst;
1897 struct rtentry *rt0;
1898 {
1899 struct mbuf *m = m0;
1900 struct rtentry *rt = rt0;
1901 struct sockaddr_in6 *gw6 = NULL;
1902 struct llinfo_nd6 *ln = NULL;
1903 int error = 0;
1904
1905 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1906 goto sendpkt;
1907
1908 if (nd6_need_cache(ifp) == 0)
1909 goto sendpkt;
1910
1911 /*
1912 * next hop determination. This routine is derived from ether_output.
1913 */
1914 if (rt) {
1915 if ((rt->rt_flags & RTF_UP) == 0) {
1916 if ((rt0 = rt = rtalloc1((struct sockaddr *)dst,
1917 1)) != NULL)
1918 {
1919 rt->rt_refcnt--;
1920 if (rt->rt_ifp != ifp)
1921 senderr(EHOSTUNREACH);
1922 } else
1923 senderr(EHOSTUNREACH);
1924 }
1925
1926 if (rt->rt_flags & RTF_GATEWAY) {
1927 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1928
1929 /*
1930 * We skip link-layer address resolution and NUD
1931 * if the gateway is not a neighbor from ND point
1932 * of view, regardless of the value of nd_ifinfo.flags.
1933 * The second condition is a bit tricky; we skip
1934 * if the gateway is our own address, which is
1935 * sometimes used to install a route to a p2p link.
1936 */
1937 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1938 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1939 /*
1940 * We allow this kind of tricky route only
1941 * when the outgoing interface is p2p.
1942 * XXX: we may need a more generic rule here.
1943 */
1944 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1945 senderr(EHOSTUNREACH);
1946
1947 goto sendpkt;
1948 }
1949
1950 if (rt->rt_gwroute == 0)
1951 goto lookup;
1952 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1953 rtfree(rt); rt = rt0;
1954 lookup:
1955 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1);
1956 if ((rt = rt->rt_gwroute) == 0)
1957 senderr(EHOSTUNREACH);
1958 /* the "G" test below also prevents rt == rt0 */
1959 if ((rt->rt_flags & RTF_GATEWAY) ||
1960 (rt->rt_ifp != ifp)) {
1961 rt->rt_refcnt--;
1962 rt0->rt_gwroute = 0;
1963 senderr(EHOSTUNREACH);
1964 }
1965 }
1966 }
1967 }
1968
1969 /*
1970 * Address resolution or Neighbor Unreachability Detection
1971 * for the next hop.
1972 * At this point, the destination of the packet must be a unicast
1973 * or an anycast address(i.e. not a multicast).
1974 */
1975
1976 /* Look up the neighbor cache for the nexthop */
1977 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
1978 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1979 else {
1980 /*
1981 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1982 * the condition below is not very efficient. But we believe
1983 * it is tolerable, because this should be a rare case.
1984 */
1985 if (nd6_is_addr_neighbor(dst, ifp) &&
1986 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
1987 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1988 }
1989 if (ln == NULL || rt == NULL) {
1990 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1991 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1992 log(LOG_DEBUG,
1993 "nd6_output: can't allocate llinfo for %s "
1994 "(ln=%p, rt=%p)\n",
1995 ip6_sprintf(&dst->sin6_addr), ln, rt);
1996 senderr(EIO); /* XXX: good error? */
1997 }
1998
1999 goto sendpkt; /* send anyway */
2000 }
2001
2002 /* We don't have to do link-layer address resolution on a p2p link. */
2003 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2004 ln->ln_state < ND6_LLINFO_REACHABLE) {
2005 ln->ln_state = ND6_LLINFO_STALE;
2006 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
2007 }
2008
2009 /*
2010 * The first time we send a packet to a neighbor whose entry is
2011 * STALE, we have to change the state to DELAY and a sets a timer to
2012 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2013 * neighbor unreachability detection on expiration.
2014 * (RFC 2461 7.3.3)
2015 */
2016 if (ln->ln_state == ND6_LLINFO_STALE) {
2017 ln->ln_asked = 0;
2018 ln->ln_state = ND6_LLINFO_DELAY;
2019 nd6_llinfo_settimer(ln, (long)nd6_delay * hz);
2020 }
2021
2022 /*
2023 * If the neighbor cache entry has a state other than INCOMPLETE
2024 * (i.e. its link-layer address is already resolved), just
2025 * send the packet.
2026 */
2027 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2028 goto sendpkt;
2029
2030 /*
2031 * There is a neighbor cache entry, but no ethernet address
2032 * response yet. Append this latest packet to the end of the
2033 * packet queue in the mbuf, unless the number of the packet
2034 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2035 * the oldest packet in the queue will be removed.
2036 */
2037 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2038 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2039 if (ln->ln_hold) {
2040 struct mbuf *m_hold;
2041 int i;
2042
2043 i = 0;
2044 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) {
2045 i++;
2046 if (m_hold->m_nextpkt == NULL) {
2047 m_hold->m_nextpkt = m;
2048 break;
2049 }
2050 }
2051 while (i >= nd6_maxqueuelen) {
2052 m_hold = ln->ln_hold;
2053 ln->ln_hold = ln->ln_hold->m_nextpkt;
2054 m_freem(m_hold);
2055 i--;
2056 }
2057 } else {
2058 ln->ln_hold = m;
2059 }
2060
2061 /*
2062 * If there has been no NS for the neighbor after entering the
2063 * INCOMPLETE state, send the first solicitation.
2064 */
2065 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
2066 ln->ln_asked++;
2067 nd6_llinfo_settimer(ln,
2068 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2069 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2070 }
2071 return (0);
2072
2073 sendpkt:
2074 /* discard the packet if IPv6 operation is disabled on the interface */
2075 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2076 error = ENETDOWN; /* better error? */
2077 goto bad;
2078 }
2079
2080 #ifdef IPSEC
2081 /* clean ipsec history once it goes out of the node */
2082 ipsec_delaux(m);
2083 #endif
2084 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2085 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2086 rt));
2087 }
2088 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2089
2090 bad:
2091 if (m)
2092 m_freem(m);
2093 return (error);
2094 }
2095 #undef senderr
2096
2097 int
2098 nd6_need_cache(ifp)
2099 struct ifnet *ifp;
2100 {
2101 /*
2102 * XXX: we currently do not make neighbor cache on any interface
2103 * other than ARCnet, Ethernet, FDDI and GIF.
2104 *
2105 * RFC2893 says:
2106 * - unidirectional tunnels needs no ND
2107 */
2108 switch (ifp->if_type) {
2109 case IFT_ARCNET:
2110 case IFT_ETHER:
2111 case IFT_FDDI:
2112 case IFT_IEEE1394:
2113 case IFT_CARP:
2114 case IFT_GIF: /* XXX need more cases? */
2115 case IFT_PPP:
2116 case IFT_TUNNEL:
2117 return (1);
2118 default:
2119 return (0);
2120 }
2121 }
2122
2123 int
2124 nd6_storelladdr(ifp, rt, m, dst, desten)
2125 struct ifnet *ifp;
2126 struct rtentry *rt;
2127 struct mbuf *m;
2128 struct sockaddr *dst;
2129 u_char *desten;
2130 {
2131 struct sockaddr_dl *sdl;
2132
2133 if (m->m_flags & M_MCAST) {
2134 switch (ifp->if_type) {
2135 case IFT_ETHER:
2136 case IFT_FDDI:
2137 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2138 desten);
2139 return (1);
2140 case IFT_IEEE1394:
2141 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen);
2142 return (1);
2143 case IFT_ARCNET:
2144 *desten = 0;
2145 return (1);
2146 default:
2147 m_freem(m);
2148 return (0);
2149 }
2150 }
2151
2152 if (rt == NULL) {
2153 /* this could happen, if we could not allocate memory */
2154 m_freem(m);
2155 return (0);
2156 }
2157 if (rt->rt_gateway->sa_family != AF_LINK) {
2158 printf("nd6_storelladdr: something odd happens\n");
2159 m_freem(m);
2160 return (0);
2161 }
2162 sdl = SDL(rt->rt_gateway);
2163 if (sdl->sdl_alen == 0) {
2164 /* this should be impossible, but we bark here for debugging */
2165 printf("nd6_storelladdr: sdl_alen == 0, dst=%s, if=%s\n",
2166 ip6_sprintf(&SIN6(dst)->sin6_addr), if_name(ifp));
2167 m_freem(m);
2168 return (0);
2169 }
2170
2171 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2172 return (1);
2173 }
2174
2175 static void
2176 clear_llinfo_pqueue(ln)
2177 struct llinfo_nd6 *ln;
2178 {
2179 struct mbuf *m_hold, *m_hold_next;
2180
2181 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) {
2182 m_hold_next = m_hold->m_nextpkt;
2183 m_hold->m_nextpkt = NULL;
2184 m_freem(m_hold);
2185 }
2186
2187 ln->ln_hold = NULL;
2188 return;
2189 }
2190
2191 int
2192 nd6_sysctl(
2193 int name,
2194 void *oldp, /* syscall arg, need copyout */
2195 size_t *oldlenp,
2196 void *newp, /* syscall arg, need copyin */
2197 size_t newlen
2198 )
2199 {
2200 void *p;
2201 size_t ol;
2202 int error;
2203
2204 error = 0;
2205
2206 if (newp)
2207 return EPERM;
2208 if (oldp && !oldlenp)
2209 return EINVAL;
2210 ol = oldlenp ? *oldlenp : 0;
2211
2212 if (oldp) {
2213 p = malloc(*oldlenp, M_TEMP, M_WAITOK);
2214 if (p == NULL)
2215 return ENOMEM;
2216 } else
2217 p = NULL;
2218 switch (name) {
2219 case ICMPV6CTL_ND6_DRLIST:
2220 error = fill_drlist(p, oldlenp, ol);
2221 if (!error && p != NULL && oldp != NULL)
2222 error = copyout(p, oldp, *oldlenp);
2223 break;
2224
2225 case ICMPV6CTL_ND6_PRLIST:
2226 error = fill_prlist(p, oldlenp, ol);
2227 if (!error && p != NULL && oldp != NULL)
2228 error = copyout(p, oldp, *oldlenp);
2229 break;
2230
2231 case ICMPV6CTL_ND6_MAXQLEN:
2232 break;
2233
2234 default:
2235 error = ENOPROTOOPT;
2236 break;
2237 }
2238 if (p)
2239 free(p, M_TEMP);
2240
2241 return (error);
2242 }
2243
2244 static int
2245 fill_drlist(oldp, oldlenp, ol)
2246 void *oldp;
2247 size_t *oldlenp, ol;
2248 {
2249 int error = 0, s;
2250 struct in6_defrouter *d = NULL, *de = NULL;
2251 struct nd_defrouter *dr;
2252 size_t l;
2253
2254 s = splsoftnet();
2255
2256 if (oldp) {
2257 d = (struct in6_defrouter *)oldp;
2258 de = (struct in6_defrouter *)((caddr_t)oldp + *oldlenp);
2259 }
2260 l = 0;
2261
2262 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
2263
2264 if (oldp && d + 1 <= de) {
2265 bzero(d, sizeof(*d));
2266 d->rtaddr.sin6_family = AF_INET6;
2267 d->rtaddr.sin6_len = sizeof(struct sockaddr_in6);
2268 d->rtaddr.sin6_addr = dr->rtaddr;
2269 if (sa6_recoverscope(&d->rtaddr)) {
2270 log(LOG_ERR,
2271 "scope error in router list (%s)\n",
2272 ip6_sprintf(&d->rtaddr.sin6_addr));
2273 /* XXX: press on... */
2274 }
2275 d->flags = dr->flags;
2276 d->rtlifetime = dr->rtlifetime;
2277 d->expire = dr->expire;
2278 d->if_index = dr->ifp->if_index;
2279 }
2280
2281 l += sizeof(*d);
2282 if (d)
2283 d++;
2284 }
2285
2286 if (oldp) {
2287 if (l > ol)
2288 error = ENOMEM;
2289 }
2290 if (oldlenp)
2291 *oldlenp = l; /* (caddr_t)d - (caddr_t)oldp */
2292
2293 splx(s);
2294
2295 return (error);
2296 }
2297
2298 static int
2299 fill_prlist(oldp, oldlenp, ol)
2300 void *oldp;
2301 size_t *oldlenp, ol;
2302 {
2303 int error = 0, s;
2304 struct nd_prefix *pr;
2305 struct in6_prefix *p = NULL;
2306 struct in6_prefix *pe = NULL;
2307 size_t l;
2308
2309 s = splsoftnet();
2310
2311 if (oldp) {
2312 p = (struct in6_prefix *)oldp;
2313 pe = (struct in6_prefix *)((caddr_t)oldp + *oldlenp);
2314 }
2315 l = 0;
2316
2317 LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
2318 u_short advrtrs;
2319 size_t advance;
2320 struct sockaddr_in6 *sin6;
2321 struct sockaddr_in6 *s6;
2322 struct nd_pfxrouter *pfr;
2323
2324 if (oldp && p + 1 <= pe)
2325 {
2326 bzero(p, sizeof(*p));
2327 sin6 = (struct sockaddr_in6 *)(p + 1);
2328
2329 p->prefix = pr->ndpr_prefix;
2330 if (sa6_recoverscope(&p->prefix)) {
2331 log(LOG_ERR,
2332 "scope error in prefix list (%s)\n",
2333 ip6_sprintf(&p->prefix.sin6_addr));
2334 /* XXX: press on... */
2335 }
2336 p->raflags = pr->ndpr_raf;
2337 p->prefixlen = pr->ndpr_plen;
2338 p->vltime = pr->ndpr_vltime;
2339 p->pltime = pr->ndpr_pltime;
2340 p->if_index = pr->ndpr_ifp->if_index;
2341 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2342 p->expire = 0;
2343 else {
2344 time_t maxexpire;
2345
2346 /* XXX: we assume time_t is signed. */
2347 maxexpire = (-1) &
2348 ~((time_t)1 <<
2349 ((sizeof(maxexpire) * 8) - 1));
2350 if (pr->ndpr_vltime <
2351 maxexpire - pr->ndpr_lastupdate) {
2352 p->expire = pr->ndpr_lastupdate +
2353 pr->ndpr_vltime;
2354 } else
2355 p->expire = maxexpire;
2356 }
2357 p->refcnt = pr->ndpr_refcnt;
2358 p->flags = pr->ndpr_stateflags;
2359 p->origin = PR_ORIG_RA;
2360 advrtrs = 0;
2361 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2362 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2363 advrtrs++;
2364 continue;
2365 }
2366 s6 = &sin6[advrtrs];
2367 s6->sin6_family = AF_INET6;
2368 s6->sin6_len = sizeof(struct sockaddr_in6);
2369 s6->sin6_addr = pfr->router->rtaddr;
2370 s6->sin6_scope_id = 0;
2371 if (sa6_recoverscope(s6)) {
2372 log(LOG_ERR,
2373 "scope error in "
2374 "prefix list (%s)\n",
2375 ip6_sprintf(&pfr->router->rtaddr));
2376 }
2377 advrtrs++;
2378 }
2379 p->advrtrs = advrtrs;
2380 }
2381 else {
2382 advrtrs = 0;
2383 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2384 advrtrs++;
2385 }
2386
2387 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2388 l += advance;
2389 if (p)
2390 p = (struct in6_prefix *)((caddr_t)p + advance);
2391 }
2392
2393 if (oldp) {
2394 *oldlenp = l; /* (caddr_t)d - (caddr_t)oldp */
2395 if (l > ol)
2396 error = ENOMEM;
2397 } else
2398 *oldlenp = l;
2399
2400 splx(s);
2401
2402 return (error);
2403 }
Cache object: 990689eb75407c0b6893d5195d2bdb21
|