FreeBSD/Linux Kernel Cross Reference
sys/netinet6/nd6.c
1 /* $FreeBSD: releng/6.3/sys/netinet6/nd6.c 173201 2007-10-30 18:07:55Z jhb $ */
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/mac.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/syslog.h>
50 #include <sys/queue.h>
51 #include <sys/sysctl.h>
52
53 #include <net/if.h>
54 #include <net/if_arc.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/iso88025.h>
58 #include <net/fddi.h>
59 #include <net/route.h>
60
61 #include <netinet/in.h>
62 #include <netinet/if_ether.h>
63 #include <netinet6/in6_var.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet6/scope6_var.h>
67 #include <netinet6/nd6.h>
68 #include <netinet/icmp6.h>
69
70 #include <sys/limits.h>
71
72 #include <net/net_osdep.h>
73
74 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
75 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
76
77 #define SIN6(s) ((struct sockaddr_in6 *)s)
78 #define SDL(s) ((struct sockaddr_dl *)s)
79
80 /* timer values */
81 int nd6_prune = 1; /* walk list every 1 seconds */
82 int nd6_delay = 5; /* delay first probe time 5 second */
83 int nd6_umaxtries = 3; /* maximum unicast query */
84 int nd6_mmaxtries = 3; /* maximum multicast query */
85 int nd6_useloopback = 1; /* use loopback interface for local traffic */
86 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
87
88 /* preventing too many loops in ND option parsing */
89 int nd6_maxndopt = 10; /* max # of ND options allowed */
90
91 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
92 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */
93
94 #ifdef ND6_DEBUG
95 int nd6_debug = 1;
96 #else
97 int nd6_debug = 0;
98 #endif
99
100 /* for debugging? */
101 static int nd6_inuse, nd6_allocated;
102
103 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
104 struct nd_drhead nd_defrouter;
105 struct nd_prhead nd_prefix = { 0 };
106
107 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
108 static struct sockaddr_in6 all1_sa;
109
110 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
111 struct ifnet *));
112 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *));
113 static void nd6_slowtimo __P((void *));
114 static int regen_tmpaddr __P((struct in6_ifaddr *));
115 static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int));
116 static void nd6_llinfo_timer __P((void *));
117 static void clear_llinfo_pqueue __P((struct llinfo_nd6 *));
118
119 struct callout nd6_slowtimo_ch;
120 struct callout nd6_timer_ch;
121 extern struct callout in6_tmpaddrtimer_ch;
122
123 void
124 nd6_init()
125 {
126 static int nd6_init_done = 0;
127 int i;
128
129 if (nd6_init_done) {
130 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
131 return;
132 }
133
134 all1_sa.sin6_family = AF_INET6;
135 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
136 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
137 all1_sa.sin6_addr.s6_addr[i] = 0xff;
138
139 /* initialization of the default router list */
140 TAILQ_INIT(&nd_defrouter);
141
142 nd6_init_done = 1;
143
144 /* start timer */
145 callout_init(&nd6_slowtimo_ch, 0);
146 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
147 nd6_slowtimo, NULL);
148 }
149
150 struct nd_ifinfo *
151 nd6_ifattach(ifp)
152 struct ifnet *ifp;
153 {
154 struct nd_ifinfo *nd;
155
156 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
157 bzero(nd, sizeof(*nd));
158
159 nd->initialized = 1;
160
161 nd->chlim = IPV6_DEFHLIM;
162 nd->basereachable = REACHABLE_TIME;
163 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
164 nd->retrans = RETRANS_TIMER;
165 /*
166 * Note that the default value of ip6_accept_rtadv is 0, which means
167 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
168 * here.
169 */
170 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
171
172 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
173 nd6_setmtu0(ifp, nd);
174
175 return nd;
176 }
177
178 void
179 nd6_ifdetach(nd)
180 struct nd_ifinfo *nd;
181 {
182
183 free(nd, M_IP6NDP);
184 }
185
186 /*
187 * Reset ND level link MTU. This function is called when the physical MTU
188 * changes, which means we might have to adjust the ND level MTU.
189 */
190 void
191 nd6_setmtu(ifp)
192 struct ifnet *ifp;
193 {
194
195 nd6_setmtu0(ifp, ND_IFINFO(ifp));
196 }
197
198 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
199 void
200 nd6_setmtu0(ifp, ndi)
201 struct ifnet *ifp;
202 struct nd_ifinfo *ndi;
203 {
204 u_int32_t omaxmtu;
205
206 omaxmtu = ndi->maxmtu;
207
208 switch (ifp->if_type) {
209 case IFT_ARCNET:
210 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
211 break;
212 case IFT_FDDI:
213 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
214 break;
215 case IFT_ISO88025:
216 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
217 break;
218 default:
219 ndi->maxmtu = ifp->if_mtu;
220 break;
221 }
222
223 /*
224 * Decreasing the interface MTU under IPV6 minimum MTU may cause
225 * undesirable situation. We thus notify the operator of the change
226 * explicitly. The check for omaxmtu is necessary to restrict the
227 * log to the case of changing the MTU, not initializing it.
228 */
229 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
230 log(LOG_NOTICE, "nd6_setmtu0: "
231 "new link MTU on %s (%lu) is too small for IPv6\n",
232 if_name(ifp), (unsigned long)ndi->maxmtu);
233 }
234
235 if (ndi->maxmtu > in6_maxmtu)
236 in6_setmaxmtu(); /* check all interfaces just in case */
237
238 #undef MIN
239 }
240
241 void
242 nd6_option_init(opt, icmp6len, ndopts)
243 void *opt;
244 int icmp6len;
245 union nd_opts *ndopts;
246 {
247
248 bzero(ndopts, sizeof(*ndopts));
249 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
250 ndopts->nd_opts_last
251 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
252
253 if (icmp6len == 0) {
254 ndopts->nd_opts_done = 1;
255 ndopts->nd_opts_search = NULL;
256 }
257 }
258
259 /*
260 * Take one ND option.
261 */
262 struct nd_opt_hdr *
263 nd6_option(ndopts)
264 union nd_opts *ndopts;
265 {
266 struct nd_opt_hdr *nd_opt;
267 int olen;
268
269 if (ndopts == NULL)
270 panic("ndopts == NULL in nd6_option");
271 if (ndopts->nd_opts_last == NULL)
272 panic("uninitialized ndopts in nd6_option");
273 if (ndopts->nd_opts_search == NULL)
274 return NULL;
275 if (ndopts->nd_opts_done)
276 return NULL;
277
278 nd_opt = ndopts->nd_opts_search;
279
280 /* make sure nd_opt_len is inside the buffer */
281 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
282 bzero(ndopts, sizeof(*ndopts));
283 return NULL;
284 }
285
286 olen = nd_opt->nd_opt_len << 3;
287 if (olen == 0) {
288 /*
289 * Message validation requires that all included
290 * options have a length that is greater than zero.
291 */
292 bzero(ndopts, sizeof(*ndopts));
293 return NULL;
294 }
295
296 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
297 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
298 /* option overruns the end of buffer, invalid */
299 bzero(ndopts, sizeof(*ndopts));
300 return NULL;
301 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
302 /* reached the end of options chain */
303 ndopts->nd_opts_done = 1;
304 ndopts->nd_opts_search = NULL;
305 }
306 return nd_opt;
307 }
308
309 /*
310 * Parse multiple ND options.
311 * This function is much easier to use, for ND routines that do not need
312 * multiple options of the same type.
313 */
314 int
315 nd6_options(ndopts)
316 union nd_opts *ndopts;
317 {
318 struct nd_opt_hdr *nd_opt;
319 int i = 0;
320
321 if (ndopts == NULL)
322 panic("ndopts == NULL in nd6_options");
323 if (ndopts->nd_opts_last == NULL)
324 panic("uninitialized ndopts in nd6_options");
325 if (ndopts->nd_opts_search == NULL)
326 return 0;
327
328 while (1) {
329 nd_opt = nd6_option(ndopts);
330 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
331 /*
332 * Message validation requires that all included
333 * options have a length that is greater than zero.
334 */
335 icmp6stat.icp6s_nd_badopt++;
336 bzero(ndopts, sizeof(*ndopts));
337 return -1;
338 }
339
340 if (nd_opt == NULL)
341 goto skip1;
342
343 switch (nd_opt->nd_opt_type) {
344 case ND_OPT_SOURCE_LINKADDR:
345 case ND_OPT_TARGET_LINKADDR:
346 case ND_OPT_MTU:
347 case ND_OPT_REDIRECTED_HEADER:
348 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
349 nd6log((LOG_INFO,
350 "duplicated ND6 option found (type=%d)\n",
351 nd_opt->nd_opt_type));
352 /* XXX bark? */
353 } else {
354 ndopts->nd_opt_array[nd_opt->nd_opt_type]
355 = nd_opt;
356 }
357 break;
358 case ND_OPT_PREFIX_INFORMATION:
359 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
360 ndopts->nd_opt_array[nd_opt->nd_opt_type]
361 = nd_opt;
362 }
363 ndopts->nd_opts_pi_end =
364 (struct nd_opt_prefix_info *)nd_opt;
365 break;
366 default:
367 /*
368 * Unknown options must be silently ignored,
369 * to accomodate future extension to the protocol.
370 */
371 nd6log((LOG_DEBUG,
372 "nd6_options: unsupported option %d - "
373 "option ignored\n", nd_opt->nd_opt_type));
374 }
375
376 skip1:
377 i++;
378 if (i > nd6_maxndopt) {
379 icmp6stat.icp6s_nd_toomanyopt++;
380 nd6log((LOG_INFO, "too many loop in nd opt\n"));
381 break;
382 }
383
384 if (ndopts->nd_opts_done)
385 break;
386 }
387
388 return 0;
389 }
390
391 /*
392 * ND6 timer routine to handle ND6 entries
393 */
394 void
395 nd6_llinfo_settimer(ln, tick)
396 struct llinfo_nd6 *ln;
397 long tick;
398 {
399 if (tick < 0) {
400 ln->ln_expire = 0;
401 ln->ln_ntick = 0;
402 callout_stop(&ln->ln_timer_ch);
403 } else {
404 ln->ln_expire = time_second + tick / hz;
405 if (tick > INT_MAX) {
406 ln->ln_ntick = tick - INT_MAX;
407 callout_reset(&ln->ln_timer_ch, INT_MAX,
408 nd6_llinfo_timer, ln);
409 } else {
410 ln->ln_ntick = 0;
411 callout_reset(&ln->ln_timer_ch, tick,
412 nd6_llinfo_timer, ln);
413 }
414 }
415 }
416
417 static void
418 nd6_llinfo_timer(arg)
419 void *arg;
420 {
421 struct llinfo_nd6 *ln;
422 struct rtentry *rt;
423 struct in6_addr *dst;
424 struct ifnet *ifp;
425 struct nd_ifinfo *ndi = NULL;
426
427 ln = (struct llinfo_nd6 *)arg;
428
429 if (ln->ln_ntick > 0) {
430 if (ln->ln_ntick > INT_MAX) {
431 ln->ln_ntick -= INT_MAX;
432 nd6_llinfo_settimer(ln, INT_MAX);
433 } else {
434 ln->ln_ntick = 0;
435 nd6_llinfo_settimer(ln, ln->ln_ntick);
436 }
437 return;
438 }
439
440 if ((rt = ln->ln_rt) == NULL)
441 panic("ln->ln_rt == NULL");
442 if ((ifp = rt->rt_ifp) == NULL)
443 panic("ln->ln_rt->rt_ifp == NULL");
444 ndi = ND_IFINFO(ifp);
445
446 /* sanity check */
447 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
448 panic("rt_llinfo(%p) is not equal to ln(%p)",
449 rt->rt_llinfo, ln);
450 if (rt_key(rt) == NULL)
451 panic("rt key is NULL in nd6_timer(ln=%p)", ln);
452
453 dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
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, 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 the
468 * 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 if (rt)
479 (void)nd6_free(rt, 0);
480 ln = NULL;
481 }
482 break;
483 case ND6_LLINFO_REACHABLE:
484 if (!ND6_LLINFO_PERMANENT(ln)) {
485 ln->ln_state = ND6_LLINFO_STALE;
486 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
487 }
488 break;
489
490 case ND6_LLINFO_STALE:
491 /* Garbage Collection(RFC 2461 5.3) */
492 if (!ND6_LLINFO_PERMANENT(ln)) {
493 (void)nd6_free(rt, 1);
494 ln = NULL;
495 }
496 break;
497
498 case ND6_LLINFO_DELAY:
499 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
500 /* We need NUD */
501 ln->ln_asked = 1;
502 ln->ln_state = ND6_LLINFO_PROBE;
503 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
504 nd6_ns_output(ifp, dst, dst, 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, dst, ln, 0);
515 } else if (rt->rt_ifa != NULL &&
516 rt->rt_ifa->ifa_addr->sa_family == AF_INET6 &&
517 (((struct in6_ifaddr *)rt->rt_ifa)->ia_flags & IFA_ROUTE)) {
518 /*
519 * This is an unreachable neighbor whose address is
520 * specified as the destination of a p2p interface
521 * (see in6_ifinit()). We should not free the entry
522 * since this is sort of a "static" entry generated
523 * via interface address configuration.
524 */
525 ln->ln_asked = 0;
526 ln->ln_expire = 0; /* make it permanent */
527 ln->ln_state = ND6_LLINFO_STALE;
528 } else {
529 (void)nd6_free(rt, 0);
530 ln = NULL;
531 }
532 break;
533 }
534 }
535
536
537 /*
538 * ND6 timer routine to expire default route list and prefix list
539 */
540 void
541 nd6_timer(ignored_arg)
542 void *ignored_arg;
543 {
544 int s;
545 struct nd_defrouter *dr;
546 struct nd_prefix *pr;
547 struct in6_ifaddr *ia6, *nia6;
548 struct in6_addrlifetime *lt6;
549
550 callout_reset(&nd6_timer_ch, nd6_prune * hz,
551 nd6_timer, NULL);
552
553 /* expire default router list */
554 s = splnet();
555 dr = TAILQ_FIRST(&nd_defrouter);
556 while (dr) {
557 if (dr->expire && dr->expire < time_second) {
558 struct nd_defrouter *t;
559 t = TAILQ_NEXT(dr, dr_entry);
560 defrtrlist_del(dr);
561 dr = t;
562 } else {
563 dr = TAILQ_NEXT(dr, dr_entry);
564 }
565 }
566
567 /*
568 * expire interface addresses.
569 * in the past the loop was inside prefix expiry processing.
570 * However, from a stricter speci-confrmance standpoint, we should
571 * rather separate address lifetimes and prefix lifetimes.
572 */
573 addrloop:
574 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
575 nia6 = ia6->ia_next;
576 /* check address lifetime */
577 lt6 = &ia6->ia6_lifetime;
578 if (IFA6_IS_INVALID(ia6)) {
579 int regen = 0;
580
581 /*
582 * If the expiring address is temporary, try
583 * regenerating a new one. This would be useful when
584 * we suspended a laptop PC, then turned it on after a
585 * period that could invalidate all temporary
586 * addresses. Although we may have to restart the
587 * loop (see below), it must be after purging the
588 * address. Otherwise, we'd see an infinite loop of
589 * regeneration.
590 */
591 if (ip6_use_tempaddr &&
592 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
593 if (regen_tmpaddr(ia6) == 0)
594 regen = 1;
595 }
596
597 in6_purgeaddr(&ia6->ia_ifa);
598
599 if (regen)
600 goto addrloop; /* XXX: see below */
601 } else if (IFA6_IS_DEPRECATED(ia6)) {
602 int oldflags = ia6->ia6_flags;
603
604 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
605
606 /*
607 * If a temporary address has just become deprecated,
608 * regenerate a new one if possible.
609 */
610 if (ip6_use_tempaddr &&
611 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
612 (oldflags & IN6_IFF_DEPRECATED) == 0) {
613
614 if (regen_tmpaddr(ia6) == 0) {
615 /*
616 * A new temporary address is
617 * generated.
618 * XXX: this means the address chain
619 * has changed while we are still in
620 * the loop. Although the change
621 * would not cause disaster (because
622 * it's not a deletion, but an
623 * addition,) we'd rather restart the
624 * loop just for safety. Or does this
625 * significantly reduce performance??
626 */
627 goto addrloop;
628 }
629 }
630 } else {
631 /*
632 * A new RA might have made a deprecated address
633 * preferred.
634 */
635 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
636 }
637 }
638
639 /* expire prefix list */
640 pr = nd_prefix.lh_first;
641 while (pr) {
642 /*
643 * check prefix lifetime.
644 * since pltime is just for autoconf, pltime processing for
645 * prefix is not necessary.
646 */
647 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
648 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
649 struct nd_prefix *t;
650 t = pr->ndpr_next;
651
652 /*
653 * address expiration and prefix expiration are
654 * separate. NEVER perform in6_purgeaddr here.
655 */
656
657 prelist_remove(pr);
658 pr = t;
659 } else
660 pr = pr->ndpr_next;
661 }
662 splx(s);
663 }
664
665 static int
666 regen_tmpaddr(ia6)
667 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
668 {
669 struct ifaddr *ifa;
670 struct ifnet *ifp;
671 struct in6_ifaddr *public_ifa6 = NULL;
672
673 ifp = ia6->ia_ifa.ifa_ifp;
674 for (ifa = ifp->if_addrlist.tqh_first; ifa;
675 ifa = ifa->ifa_list.tqe_next) {
676 struct in6_ifaddr *it6;
677
678 if (ifa->ifa_addr->sa_family != AF_INET6)
679 continue;
680
681 it6 = (struct in6_ifaddr *)ifa;
682
683 /* ignore no autoconf addresses. */
684 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
685 continue;
686
687 /* ignore autoconf addresses with different prefixes. */
688 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
689 continue;
690
691 /*
692 * Now we are looking at an autoconf address with the same
693 * prefix as ours. If the address is temporary and is still
694 * preferred, do not create another one. It would be rare, but
695 * could happen, for example, when we resume a laptop PC after
696 * a long period.
697 */
698 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
699 !IFA6_IS_DEPRECATED(it6)) {
700 public_ifa6 = NULL;
701 break;
702 }
703
704 /*
705 * This is a public autoconf address that has the same prefix
706 * as ours. If it is preferred, keep it. We can't break the
707 * loop here, because there may be a still-preferred temporary
708 * address with the prefix.
709 */
710 if (!IFA6_IS_DEPRECATED(it6))
711 public_ifa6 = it6;
712 }
713
714 if (public_ifa6 != NULL) {
715 int e;
716
717 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
718 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
719 " tmp addr,errno=%d\n", e);
720 return (-1);
721 }
722 return (0);
723 }
724
725 return (-1);
726 }
727
728 /*
729 * Nuke neighbor cache/prefix/default router management table, right before
730 * ifp goes away.
731 */
732 void
733 nd6_purge(ifp)
734 struct ifnet *ifp;
735 {
736 struct llinfo_nd6 *ln, *nln;
737 struct nd_defrouter *dr, *ndr;
738 struct nd_prefix *pr, *npr;
739
740 /*
741 * Nuke default router list entries toward ifp.
742 * We defer removal of default router list entries that is installed
743 * in the routing table, in order to keep additional side effects as
744 * small as possible.
745 */
746 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
747 ndr = TAILQ_NEXT(dr, dr_entry);
748 if (dr->installed)
749 continue;
750
751 if (dr->ifp == ifp)
752 defrtrlist_del(dr);
753 }
754
755 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
756 ndr = TAILQ_NEXT(dr, dr_entry);
757 if (!dr->installed)
758 continue;
759
760 if (dr->ifp == ifp)
761 defrtrlist_del(dr);
762 }
763
764 /* Nuke prefix list entries toward ifp */
765 for (pr = nd_prefix.lh_first; pr; pr = npr) {
766 npr = pr->ndpr_next;
767 if (pr->ndpr_ifp == ifp) {
768 /*
769 * Because if_detach() does *not* release prefixes
770 * while purging addresses the reference count will
771 * still be above zero. We therefore reset it to
772 * make sure that the prefix really gets purged.
773 */
774 pr->ndpr_refcnt = 0;
775
776 /*
777 * Previously, pr->ndpr_addr is removed as well,
778 * but I strongly believe we don't have to do it.
779 * nd6_purge() is only called from in6_ifdetach(),
780 * which removes all the associated interface addresses
781 * by itself.
782 * (jinmei@kame.net 20010129)
783 */
784 prelist_remove(pr);
785 }
786 }
787
788 /* cancel default outgoing interface setting */
789 if (nd6_defifindex == ifp->if_index)
790 nd6_setdefaultiface(0);
791
792 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
793 /* refresh default router list */
794 defrouter_select();
795 }
796
797 /*
798 * Nuke neighbor cache entries for the ifp.
799 * Note that rt->rt_ifp may not be the same as ifp,
800 * due to KAME goto ours hack. See RTM_RESOLVE case in
801 * nd6_rtrequest(), and ip6_input().
802 */
803 ln = llinfo_nd6.ln_next;
804 while (ln && ln != &llinfo_nd6) {
805 struct rtentry *rt;
806 struct sockaddr_dl *sdl;
807
808 nln = ln->ln_next;
809 rt = ln->ln_rt;
810 if (rt && rt->rt_gateway &&
811 rt->rt_gateway->sa_family == AF_LINK) {
812 sdl = (struct sockaddr_dl *)rt->rt_gateway;
813 if (sdl->sdl_index == ifp->if_index)
814 nln = nd6_free(rt, 0);
815 }
816 ln = nln;
817 }
818 }
819
820 struct rtentry *
821 nd6_lookup(addr6, create, ifp)
822 struct in6_addr *addr6;
823 int create;
824 struct ifnet *ifp;
825 {
826 struct rtentry *rt;
827 struct sockaddr_in6 sin6;
828
829 bzero(&sin6, sizeof(sin6));
830 sin6.sin6_len = sizeof(struct sockaddr_in6);
831 sin6.sin6_family = AF_INET6;
832 sin6.sin6_addr = *addr6;
833 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
834 if (rt) {
835 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) {
836 /*
837 * This is the case for the default route.
838 * If we want to create a neighbor cache for the
839 * address, we should free the route for the
840 * destination and allocate an interface route.
841 */
842 RTFREE_LOCKED(rt);
843 rt = NULL;
844 }
845 }
846 if (rt == NULL) {
847 if (create && ifp) {
848 int e;
849
850 /*
851 * If no route is available and create is set,
852 * we allocate a host route for the destination
853 * and treat it like an interface route.
854 * This hack is necessary for a neighbor which can't
855 * be covered by our own prefix.
856 */
857 struct ifaddr *ifa =
858 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
859 if (ifa == NULL)
860 return (NULL);
861
862 /*
863 * Create a new route. RTF_LLINFO is necessary
864 * to create a Neighbor Cache entry for the
865 * destination in nd6_rtrequest which will be
866 * called in rtrequest via ifa->ifa_rtrequest.
867 */
868 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
869 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
870 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
871 ~RTF_CLONING, &rt)) != 0) {
872 log(LOG_ERR,
873 "nd6_lookup: failed to add route for a "
874 "neighbor(%s), errno=%d\n",
875 ip6_sprintf(addr6), e);
876 }
877 if (rt == NULL)
878 return (NULL);
879 RT_LOCK(rt);
880 if (rt->rt_llinfo) {
881 struct llinfo_nd6 *ln =
882 (struct llinfo_nd6 *)rt->rt_llinfo;
883 ln->ln_state = ND6_LLINFO_NOSTATE;
884 }
885 } else
886 return (NULL);
887 }
888 RT_LOCK_ASSERT(rt);
889 RT_REMREF(rt);
890 /*
891 * Validation for the entry.
892 * Note that the check for rt_llinfo is necessary because a cloned
893 * route from a parent route that has the L flag (e.g. the default
894 * route to a p2p interface) may have the flag, too, while the
895 * destination is not actually a neighbor.
896 * XXX: we can't use rt->rt_ifp to check for the interface, since
897 * it might be the loopback interface if the entry is for our
898 * own address on a non-loopback interface. Instead, we should
899 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
900 * interface.
901 * Note also that ifa_ifp and ifp may differ when we connect two
902 * interfaces to a same link, install a link prefix to an interface,
903 * and try to install a neighbor cache on an interface that does not
904 * have a route to the prefix.
905 */
906 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
907 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
908 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
909 if (create) {
910 nd6log((LOG_DEBUG,
911 "nd6_lookup: failed to lookup %s (if = %s)\n",
912 ip6_sprintf(addr6),
913 ifp ? if_name(ifp) : "unspec"));
914 }
915 RT_UNLOCK(rt);
916 return (NULL);
917 }
918 RT_UNLOCK(rt); /* XXX not ready to return rt locked */
919 return (rt);
920 }
921
922 /*
923 * Test whether a given IPv6 address is a neighbor or not, ignoring
924 * the actual neighbor cache. The neighbor cache is ignored in order
925 * to not reenter the routing code from within itself.
926 */
927 static int
928 nd6_is_new_addr_neighbor(addr, ifp)
929 struct sockaddr_in6 *addr;
930 struct ifnet *ifp;
931 {
932 struct nd_prefix *pr;
933 struct ifaddr *dstaddr;
934
935 /*
936 * A link-local address is always a neighbor.
937 * XXX: a link does not necessarily specify a single interface.
938 */
939 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
940 struct sockaddr_in6 sin6_copy;
941 u_int32_t zone;
942
943 /*
944 * We need sin6_copy since sa6_recoverscope() may modify the
945 * content (XXX).
946 */
947 sin6_copy = *addr;
948 if (sa6_recoverscope(&sin6_copy))
949 return (0); /* XXX: should be impossible */
950 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
951 return (0);
952 if (sin6_copy.sin6_scope_id == zone)
953 return (1);
954 else
955 return (0);
956 }
957
958 /*
959 * If the address matches one of our addresses,
960 * it should be a neighbor.
961 * If the address matches one of our on-link prefixes, it should be a
962 * neighbor.
963 */
964 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
965 if (pr->ndpr_ifp != ifp)
966 continue;
967
968 if (!(pr->ndpr_stateflags & NDPRF_ONLINK))
969 continue;
970
971 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
972 &addr->sin6_addr, &pr->ndpr_mask))
973 return (1);
974 }
975
976 /*
977 * If the address is assigned on the node of the other side of
978 * a p2p interface, the address should be a neighbor.
979 */
980 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
981 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp))
982 return (1);
983
984 /*
985 * If the default router list is empty, all addresses are regarded
986 * as on-link, and thus, as a neighbor.
987 * XXX: we restrict the condition to hosts, because routers usually do
988 * not have the "default router list".
989 */
990 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL &&
991 nd6_defifindex == ifp->if_index) {
992 return (1);
993 }
994
995 return (0);
996 }
997
998
999 /*
1000 * Detect if a given IPv6 address identifies a neighbor on a given link.
1001 * XXX: should take care of the destination of a p2p link?
1002 */
1003 int
1004 nd6_is_addr_neighbor(addr, ifp)
1005 struct sockaddr_in6 *addr;
1006 struct ifnet *ifp;
1007 {
1008
1009 if (nd6_is_new_addr_neighbor(addr, ifp))
1010 return (1);
1011
1012 /*
1013 * Even if the address matches none of our addresses, it might be
1014 * in the neighbor cache.
1015 */
1016 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
1017 return (1);
1018
1019 return (0);
1020 }
1021
1022 /*
1023 * Free an nd6 llinfo entry.
1024 * Since the function would cause significant changes in the kernel, DO NOT
1025 * make it global, unless you have a strong reason for the change, and are sure
1026 * that the change is safe.
1027 */
1028 static struct llinfo_nd6 *
1029 nd6_free(rt, gc)
1030 struct rtentry *rt;
1031 int gc;
1032 {
1033 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1034 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
1035 struct nd_defrouter *dr;
1036
1037 /*
1038 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1039 * even though it is not harmful, it was not really necessary.
1040 */
1041
1042 /* cancel timer */
1043 nd6_llinfo_settimer(ln, -1);
1044
1045 if (!ip6_forwarding) {
1046 int s;
1047 s = splnet();
1048 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1049 rt->rt_ifp);
1050
1051 if (dr != NULL && dr->expire &&
1052 ln->ln_state == ND6_LLINFO_STALE && gc) {
1053 /*
1054 * If the reason for the deletion is just garbage
1055 * collection, and the neighbor is an active default
1056 * router, do not delete it. Instead, reset the GC
1057 * timer using the router's lifetime.
1058 * Simply deleting the entry would affect default
1059 * router selection, which is not necessarily a good
1060 * thing, especially when we're using router preference
1061 * values.
1062 * XXX: the check for ln_state would be redundant,
1063 * but we intentionally keep it just in case.
1064 */
1065 if (dr->expire > time_second)
1066 nd6_llinfo_settimer(ln,
1067 (dr->expire - time_second) * hz);
1068 else
1069 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1070 splx(s);
1071 return (ln->ln_next);
1072 }
1073
1074 if (ln->ln_router || dr) {
1075 /*
1076 * rt6_flush must be called whether or not the neighbor
1077 * is in the Default Router List.
1078 * See a corresponding comment in nd6_na_input().
1079 */
1080 rt6_flush(&in6, rt->rt_ifp);
1081 }
1082
1083 if (dr) {
1084 /*
1085 * Unreachablity of a router might affect the default
1086 * router selection and on-link detection of advertised
1087 * prefixes.
1088 */
1089
1090 /*
1091 * Temporarily fake the state to choose a new default
1092 * router and to perform on-link determination of
1093 * prefixes correctly.
1094 * Below the state will be set correctly,
1095 * or the entry itself will be deleted.
1096 */
1097 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1098
1099 /*
1100 * Since defrouter_select() does not affect the
1101 * on-link determination and MIP6 needs the check
1102 * before the default router selection, we perform
1103 * the check now.
1104 */
1105 pfxlist_onlink_check();
1106
1107 /*
1108 * refresh default router list
1109 */
1110 defrouter_select();
1111 }
1112 splx(s);
1113 }
1114
1115 /*
1116 * Before deleting the entry, remember the next entry as the
1117 * return value. We need this because pfxlist_onlink_check() above
1118 * might have freed other entries (particularly the old next entry) as
1119 * a side effect (XXX).
1120 */
1121 next = ln->ln_next;
1122
1123 /*
1124 * Detach the route from the routing tree and the list of neighbor
1125 * caches, and disable the route entry not to be used in already
1126 * cached routes.
1127 */
1128 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1129 rt_mask(rt), 0, (struct rtentry **)0);
1130
1131 return (next);
1132 }
1133
1134 /*
1135 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1136 *
1137 * XXX cost-effective methods?
1138 */
1139 void
1140 nd6_nud_hint(rt, dst6, force)
1141 struct rtentry *rt;
1142 struct in6_addr *dst6;
1143 int force;
1144 {
1145 struct llinfo_nd6 *ln;
1146
1147 /*
1148 * If the caller specified "rt", use that. Otherwise, resolve the
1149 * routing table by supplied "dst6".
1150 */
1151 if (rt == NULL) {
1152 if (dst6 == NULL)
1153 return;
1154 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL)
1155 return;
1156 }
1157
1158 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1159 (rt->rt_flags & RTF_LLINFO) == 0 ||
1160 rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
1161 rt->rt_gateway->sa_family != AF_LINK) {
1162 /* This is not a host route. */
1163 return;
1164 }
1165
1166 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1167 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1168 return;
1169
1170 /*
1171 * if we get upper-layer reachability confirmation many times,
1172 * it is possible we have false information.
1173 */
1174 if (!force) {
1175 ln->ln_byhint++;
1176 if (ln->ln_byhint > nd6_maxnudhint)
1177 return;
1178 }
1179
1180 ln->ln_state = ND6_LLINFO_REACHABLE;
1181 if (!ND6_LLINFO_PERMANENT(ln)) {
1182 nd6_llinfo_settimer(ln,
1183 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1184 }
1185 }
1186
1187 void
1188 nd6_rtrequest(req, rt, info)
1189 int req;
1190 struct rtentry *rt;
1191 struct rt_addrinfo *info; /* xxx unused */
1192 {
1193 struct sockaddr *gate = rt->rt_gateway;
1194 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1195 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1196 struct ifnet *ifp = rt->rt_ifp;
1197 struct ifaddr *ifa;
1198
1199 RT_LOCK_ASSERT(rt);
1200
1201 if ((rt->rt_flags & RTF_GATEWAY) != 0)
1202 return;
1203
1204 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1205 /*
1206 * This is probably an interface direct route for a link
1207 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1208 * We do not need special treatment below for such a route.
1209 * Moreover, the RTF_LLINFO flag which would be set below
1210 * would annoy the ndp(8) command.
1211 */
1212 return;
1213 }
1214
1215 if (req == RTM_RESOLVE &&
1216 (nd6_need_cache(ifp) == 0 || /* stf case */
1217 !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt),
1218 ifp))) {
1219 /*
1220 * FreeBSD and BSD/OS often make a cloned host route based
1221 * on a less-specific route (e.g. the default route).
1222 * If the less specific route does not have a "gateway"
1223 * (this is the case when the route just goes to a p2p or an
1224 * stf interface), we'll mistakenly make a neighbor cache for
1225 * the host route, and will see strange neighbor solicitation
1226 * for the corresponding destination. In order to avoid the
1227 * confusion, we check if the destination of the route is
1228 * a neighbor in terms of neighbor discovery, and stop the
1229 * process if not. Additionally, we remove the LLINFO flag
1230 * so that ndp(8) will not try to get the neighbor information
1231 * of the destination.
1232 */
1233 rt->rt_flags &= ~RTF_LLINFO;
1234 return;
1235 }
1236
1237 switch (req) {
1238 case RTM_ADD:
1239 /*
1240 * There is no backward compatibility :)
1241 *
1242 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1243 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1244 * rt->rt_flags |= RTF_CLONING;
1245 */
1246 if ((rt->rt_flags & RTF_CLONING) ||
1247 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) {
1248 /*
1249 * Case 1: This route should come from a route to
1250 * interface (RTF_CLONING case) or the route should be
1251 * treated as on-link but is currently not
1252 * (RTF_LLINFO && ln == NULL case).
1253 */
1254 rt_setgate(rt, rt_key(rt),
1255 (struct sockaddr *)&null_sdl);
1256 gate = rt->rt_gateway;
1257 SDL(gate)->sdl_type = ifp->if_type;
1258 SDL(gate)->sdl_index = ifp->if_index;
1259 if (ln)
1260 nd6_llinfo_settimer(ln, 0);
1261 if ((rt->rt_flags & RTF_CLONING) != 0)
1262 break;
1263 }
1264 /*
1265 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1266 * We don't do that here since llinfo is not ready yet.
1267 *
1268 * There are also couple of other things to be discussed:
1269 * - unsolicited NA code needs improvement beforehand
1270 * - RFC2461 says we MAY send multicast unsolicited NA
1271 * (7.2.6 paragraph 4), however, it also says that we
1272 * SHOULD provide a mechanism to prevent multicast NA storm.
1273 * we don't have anything like it right now.
1274 * note that the mechanism needs a mutual agreement
1275 * between proxies, which means that we need to implement
1276 * a new protocol, or a new kludge.
1277 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1278 * we need to check ip6forwarding before sending it.
1279 * (or should we allow proxy ND configuration only for
1280 * routers? there's no mention about proxy ND from hosts)
1281 */
1282 /* FALLTHROUGH */
1283 case RTM_RESOLVE:
1284 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1285 /*
1286 * Address resolution isn't necessary for a point to
1287 * point link, so we can skip this test for a p2p link.
1288 */
1289 if (gate->sa_family != AF_LINK ||
1290 gate->sa_len < sizeof(null_sdl)) {
1291 log(LOG_DEBUG,
1292 "nd6_rtrequest: bad gateway value: %s\n",
1293 if_name(ifp));
1294 break;
1295 }
1296 SDL(gate)->sdl_type = ifp->if_type;
1297 SDL(gate)->sdl_index = ifp->if_index;
1298 }
1299 if (ln != NULL)
1300 break; /* This happens on a route change */
1301 /*
1302 * Case 2: This route may come from cloning, or a manual route
1303 * add with a LL address.
1304 */
1305 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1306 rt->rt_llinfo = (caddr_t)ln;
1307 if (ln == NULL) {
1308 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1309 break;
1310 }
1311 nd6_inuse++;
1312 nd6_allocated++;
1313 bzero(ln, sizeof(*ln));
1314 RT_ADDREF(rt);
1315 ln->ln_rt = rt;
1316 callout_init(&ln->ln_timer_ch, 0);
1317
1318 /* this is required for "ndp" command. - shin */
1319 if (req == RTM_ADD) {
1320 /*
1321 * gate should have some valid AF_LINK entry,
1322 * and ln->ln_expire should have some lifetime
1323 * which is specified by ndp command.
1324 */
1325 ln->ln_state = ND6_LLINFO_REACHABLE;
1326 ln->ln_byhint = 0;
1327 } else {
1328 /*
1329 * When req == RTM_RESOLVE, rt is created and
1330 * initialized in rtrequest(), so rt_expire is 0.
1331 */
1332 ln->ln_state = ND6_LLINFO_NOSTATE;
1333 nd6_llinfo_settimer(ln, 0);
1334 }
1335 rt->rt_flags |= RTF_LLINFO;
1336 ln->ln_next = llinfo_nd6.ln_next;
1337 llinfo_nd6.ln_next = ln;
1338 ln->ln_prev = &llinfo_nd6;
1339 ln->ln_next->ln_prev = ln;
1340
1341 /*
1342 * check if rt_key(rt) is one of my address assigned
1343 * to the interface.
1344 */
1345 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1346 &SIN6(rt_key(rt))->sin6_addr);
1347 if (ifa) {
1348 caddr_t macp = nd6_ifptomac(ifp);
1349 nd6_llinfo_settimer(ln, -1);
1350 ln->ln_state = ND6_LLINFO_REACHABLE;
1351 ln->ln_byhint = 0;
1352 if (macp) {
1353 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1354 SDL(gate)->sdl_alen = ifp->if_addrlen;
1355 }
1356 if (nd6_useloopback) {
1357 rt->rt_ifp = &loif[0]; /* XXX */
1358 /*
1359 * Make sure rt_ifa be equal to the ifaddr
1360 * corresponding to the address.
1361 * We need this because when we refer
1362 * rt_ifa->ia6_flags in ip6_input, we assume
1363 * that the rt_ifa points to the address instead
1364 * of the loopback address.
1365 */
1366 if (ifa != rt->rt_ifa) {
1367 IFAFREE(rt->rt_ifa);
1368 IFAREF(ifa);
1369 rt->rt_ifa = ifa;
1370 }
1371 }
1372 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1373 nd6_llinfo_settimer(ln, -1);
1374 ln->ln_state = ND6_LLINFO_REACHABLE;
1375 ln->ln_byhint = 0;
1376
1377 /* join solicited node multicast for proxy ND */
1378 if (ifp->if_flags & IFF_MULTICAST) {
1379 struct in6_addr llsol;
1380 int error;
1381
1382 llsol = SIN6(rt_key(rt))->sin6_addr;
1383 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1384 llsol.s6_addr32[1] = 0;
1385 llsol.s6_addr32[2] = htonl(1);
1386 llsol.s6_addr8[12] = 0xff;
1387 if (in6_setscope(&llsol, ifp, NULL))
1388 break;
1389 if (in6_addmulti(&llsol, ifp,
1390 &error, 0) == NULL) {
1391 nd6log((LOG_ERR, "%s: failed to join "
1392 "%s (errno=%d)\n", if_name(ifp),
1393 ip6_sprintf(&llsol), error));
1394 }
1395 }
1396 }
1397 break;
1398
1399 case RTM_DELETE:
1400 if (ln == NULL)
1401 break;
1402 /* leave from solicited node multicast for proxy ND */
1403 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1404 (ifp->if_flags & IFF_MULTICAST) != 0) {
1405 struct in6_addr llsol;
1406 struct in6_multi *in6m;
1407
1408 llsol = SIN6(rt_key(rt))->sin6_addr;
1409 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1410 llsol.s6_addr32[1] = 0;
1411 llsol.s6_addr32[2] = htonl(1);
1412 llsol.s6_addr8[12] = 0xff;
1413 if (in6_setscope(&llsol, ifp, NULL) == 0) {
1414 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1415 if (in6m)
1416 in6_delmulti(in6m);
1417 } else
1418 ; /* XXX: should not happen. bark here? */
1419 }
1420 nd6_inuse--;
1421 ln->ln_next->ln_prev = ln->ln_prev;
1422 ln->ln_prev->ln_next = ln->ln_next;
1423 ln->ln_prev = NULL;
1424 nd6_llinfo_settimer(ln, -1);
1425 RT_REMREF(rt);
1426 rt->rt_llinfo = 0;
1427 rt->rt_flags &= ~RTF_LLINFO;
1428 clear_llinfo_pqueue(ln);
1429 Free((caddr_t)ln);
1430 }
1431 }
1432
1433 int
1434 nd6_ioctl(cmd, data, ifp)
1435 u_long cmd;
1436 caddr_t data;
1437 struct ifnet *ifp;
1438 {
1439 struct in6_drlist *drl = (struct in6_drlist *)data;
1440 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1441 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1442 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1443 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1444 struct nd_defrouter *dr;
1445 struct nd_prefix *pr;
1446 struct rtentry *rt;
1447 int i = 0, error = 0;
1448 int s;
1449
1450 switch (cmd) {
1451 case SIOCGDRLST_IN6:
1452 /*
1453 * obsolete API, use sysctl under net.inet6.icmp6
1454 */
1455 bzero(drl, sizeof(*drl));
1456 s = splnet();
1457 dr = TAILQ_FIRST(&nd_defrouter);
1458 while (dr && i < DRLSTSIZ) {
1459 drl->defrouter[i].rtaddr = dr->rtaddr;
1460 in6_clearscope(&drl->defrouter[i].rtaddr);
1461
1462 drl->defrouter[i].flags = dr->flags;
1463 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1464 drl->defrouter[i].expire = dr->expire;
1465 drl->defrouter[i].if_index = dr->ifp->if_index;
1466 i++;
1467 dr = TAILQ_NEXT(dr, dr_entry);
1468 }
1469 splx(s);
1470 break;
1471 case SIOCGPRLST_IN6:
1472 /*
1473 * obsolete API, use sysctl under net.inet6.icmp6
1474 *
1475 * XXX the structure in6_prlist was changed in backward-
1476 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1477 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1478 */
1479 /*
1480 * XXX meaning of fields, especialy "raflags", is very
1481 * differnet between RA prefix list and RR/static prefix list.
1482 * how about separating ioctls into two?
1483 */
1484 bzero(oprl, sizeof(*oprl));
1485 s = splnet();
1486 pr = nd_prefix.lh_first;
1487 while (pr && i < PRLSTSIZ) {
1488 struct nd_pfxrouter *pfr;
1489 int j;
1490
1491 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1492 oprl->prefix[i].raflags = pr->ndpr_raf;
1493 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1494 oprl->prefix[i].vltime = pr->ndpr_vltime;
1495 oprl->prefix[i].pltime = pr->ndpr_pltime;
1496 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1497 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1498 oprl->prefix[i].expire = 0;
1499 else {
1500 time_t maxexpire;
1501
1502 /* XXX: we assume time_t is signed. */
1503 maxexpire = (-1) &
1504 ~((time_t)1 <<
1505 ((sizeof(maxexpire) * 8) - 1));
1506 if (pr->ndpr_vltime <
1507 maxexpire - pr->ndpr_lastupdate) {
1508 oprl->prefix[i].expire =
1509 pr->ndpr_lastupdate +
1510 pr->ndpr_vltime;
1511 } else
1512 oprl->prefix[i].expire = maxexpire;
1513 }
1514
1515 pfr = pr->ndpr_advrtrs.lh_first;
1516 j = 0;
1517 while (pfr) {
1518 if (j < DRLSTSIZ) {
1519 #define RTRADDR oprl->prefix[i].advrtr[j]
1520 RTRADDR = pfr->router->rtaddr;
1521 in6_clearscope(&RTRADDR);
1522 #undef RTRADDR
1523 }
1524 j++;
1525 pfr = pfr->pfr_next;
1526 }
1527 oprl->prefix[i].advrtrs = j;
1528 oprl->prefix[i].origin = PR_ORIG_RA;
1529
1530 i++;
1531 pr = pr->ndpr_next;
1532 }
1533 splx(s);
1534
1535 break;
1536 case OSIOCGIFINFO_IN6:
1537 #define ND ndi->ndi
1538 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1539 bzero(&ND, sizeof(ND));
1540 ND.linkmtu = IN6_LINKMTU(ifp);
1541 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1542 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1543 ND.reachable = ND_IFINFO(ifp)->reachable;
1544 ND.retrans = ND_IFINFO(ifp)->retrans;
1545 ND.flags = ND_IFINFO(ifp)->flags;
1546 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1547 ND.chlim = ND_IFINFO(ifp)->chlim;
1548 break;
1549 case SIOCGIFINFO_IN6:
1550 ND = *ND_IFINFO(ifp);
1551 break;
1552 case SIOCSIFINFO_IN6:
1553 /*
1554 * used to change host variables from userland.
1555 * intented for a use on router to reflect RA configurations.
1556 */
1557 /* 0 means 'unspecified' */
1558 if (ND.linkmtu != 0) {
1559 if (ND.linkmtu < IPV6_MMTU ||
1560 ND.linkmtu > IN6_LINKMTU(ifp)) {
1561 error = EINVAL;
1562 break;
1563 }
1564 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1565 }
1566
1567 if (ND.basereachable != 0) {
1568 int obasereachable = ND_IFINFO(ifp)->basereachable;
1569
1570 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1571 if (ND.basereachable != obasereachable)
1572 ND_IFINFO(ifp)->reachable =
1573 ND_COMPUTE_RTIME(ND.basereachable);
1574 }
1575 if (ND.retrans != 0)
1576 ND_IFINFO(ifp)->retrans = ND.retrans;
1577 if (ND.chlim != 0)
1578 ND_IFINFO(ifp)->chlim = ND.chlim;
1579 /* FALLTHROUGH */
1580 case SIOCSIFINFO_FLAGS:
1581 ND_IFINFO(ifp)->flags = ND.flags;
1582 break;
1583 #undef ND
1584 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1585 /* sync kernel routing table with the default router list */
1586 defrouter_reset();
1587 defrouter_select();
1588 break;
1589 case SIOCSPFXFLUSH_IN6:
1590 {
1591 /* flush all the prefix advertised by routers */
1592 struct nd_prefix *pr, *next;
1593
1594 s = splnet();
1595 for (pr = nd_prefix.lh_first; pr; pr = next) {
1596 struct in6_ifaddr *ia, *ia_next;
1597
1598 next = pr->ndpr_next;
1599
1600 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1601 continue; /* XXX */
1602
1603 /* do we really have to remove addresses as well? */
1604 for (ia = in6_ifaddr; ia; ia = ia_next) {
1605 /* ia might be removed. keep the next ptr. */
1606 ia_next = ia->ia_next;
1607
1608 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1609 continue;
1610
1611 if (ia->ia6_ndpr == pr)
1612 in6_purgeaddr(&ia->ia_ifa);
1613 }
1614 prelist_remove(pr);
1615 }
1616 splx(s);
1617 break;
1618 }
1619 case SIOCSRTRFLUSH_IN6:
1620 {
1621 /* flush all the default routers */
1622 struct nd_defrouter *dr, *next;
1623
1624 s = splnet();
1625 defrouter_reset();
1626 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = next) {
1627 next = TAILQ_NEXT(dr, dr_entry);
1628 defrtrlist_del(dr);
1629 }
1630 defrouter_select();
1631 splx(s);
1632 break;
1633 }
1634 case SIOCGNBRINFO_IN6:
1635 {
1636 struct llinfo_nd6 *ln;
1637 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1638
1639 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1640 return (error);
1641
1642 s = splnet();
1643 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1644 error = EINVAL;
1645 splx(s);
1646 break;
1647 }
1648 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1649 nbi->state = ln->ln_state;
1650 nbi->asked = ln->ln_asked;
1651 nbi->isrouter = ln->ln_router;
1652 nbi->expire = ln->ln_expire;
1653 splx(s);
1654
1655 break;
1656 }
1657 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1658 ndif->ifindex = nd6_defifindex;
1659 break;
1660 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1661 return (nd6_setdefaultiface(ndif->ifindex));
1662 }
1663 return (error);
1664 }
1665
1666 /*
1667 * Create neighbor cache entry and cache link-layer address,
1668 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1669 */
1670 struct rtentry *
1671 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
1672 struct ifnet *ifp;
1673 struct in6_addr *from;
1674 char *lladdr;
1675 int lladdrlen;
1676 int type; /* ICMP6 type */
1677 int code; /* type dependent information */
1678 {
1679 struct rtentry *rt = NULL;
1680 struct llinfo_nd6 *ln = NULL;
1681 int is_newentry;
1682 struct sockaddr_dl *sdl = NULL;
1683 int do_update;
1684 int olladdr;
1685 int llchange;
1686 int newstate = 0;
1687
1688 if (ifp == NULL)
1689 panic("ifp == NULL in nd6_cache_lladdr");
1690 if (from == NULL)
1691 panic("from == NULL in nd6_cache_lladdr");
1692
1693 /* nothing must be updated for unspecified address */
1694 if (IN6_IS_ADDR_UNSPECIFIED(from))
1695 return NULL;
1696
1697 /*
1698 * Validation about ifp->if_addrlen and lladdrlen must be done in
1699 * the caller.
1700 *
1701 * XXX If the link does not have link-layer adderss, what should
1702 * we do? (ifp->if_addrlen == 0)
1703 * Spec says nothing in sections for RA, RS and NA. There's small
1704 * description on it in NS section (RFC 2461 7.2.3).
1705 */
1706
1707 rt = nd6_lookup(from, 0, ifp);
1708 if (rt == NULL) {
1709 rt = nd6_lookup(from, 1, ifp);
1710 is_newentry = 1;
1711 } else {
1712 /* do nothing if static ndp is set */
1713 if (rt->rt_flags & RTF_STATIC)
1714 return NULL;
1715 is_newentry = 0;
1716 }
1717
1718 if (rt == NULL)
1719 return NULL;
1720 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1721 fail:
1722 (void)nd6_free(rt, 0);
1723 return NULL;
1724 }
1725 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1726 if (ln == NULL)
1727 goto fail;
1728 if (rt->rt_gateway == NULL)
1729 goto fail;
1730 if (rt->rt_gateway->sa_family != AF_LINK)
1731 goto fail;
1732 sdl = SDL(rt->rt_gateway);
1733
1734 olladdr = (sdl->sdl_alen) ? 1 : 0;
1735 if (olladdr && lladdr) {
1736 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1737 llchange = 1;
1738 else
1739 llchange = 0;
1740 } else
1741 llchange = 0;
1742
1743 /*
1744 * newentry olladdr lladdr llchange (*=record)
1745 * 0 n n -- (1)
1746 * 0 y n -- (2)
1747 * 0 n y -- (3) * STALE
1748 * 0 y y n (4) *
1749 * 0 y y y (5) * STALE
1750 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1751 * 1 -- y -- (7) * STALE
1752 */
1753
1754 if (lladdr) { /* (3-5) and (7) */
1755 /*
1756 * Record source link-layer address
1757 * XXX is it dependent to ifp->if_type?
1758 */
1759 sdl->sdl_alen = ifp->if_addrlen;
1760 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1761 }
1762
1763 if (!is_newentry) {
1764 if ((!olladdr && lladdr != NULL) || /* (3) */
1765 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1766 do_update = 1;
1767 newstate = ND6_LLINFO_STALE;
1768 } else /* (1-2,4) */
1769 do_update = 0;
1770 } else {
1771 do_update = 1;
1772 if (lladdr == NULL) /* (6) */
1773 newstate = ND6_LLINFO_NOSTATE;
1774 else /* (7) */
1775 newstate = ND6_LLINFO_STALE;
1776 }
1777
1778 if (do_update) {
1779 /*
1780 * Update the state of the neighbor cache.
1781 */
1782 ln->ln_state = newstate;
1783
1784 if (ln->ln_state == ND6_LLINFO_STALE) {
1785 /*
1786 * XXX: since nd6_output() below will cause
1787 * state tansition to DELAY and reset the timer,
1788 * we must set the timer now, although it is actually
1789 * meaningless.
1790 */
1791 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1792
1793 if (ln->ln_hold) {
1794 struct mbuf *m_hold, *m_hold_next;
1795
1796 /*
1797 * reset the ln_hold in advance, to explicitly
1798 * prevent a ln_hold lookup in nd6_output()
1799 * (wouldn't happen, though...)
1800 */
1801 for (m_hold = ln->ln_hold, ln->ln_hold = NULL;
1802 m_hold; m_hold = m_hold_next) {
1803 m_hold_next = m_hold->m_nextpkt;
1804 m_hold->m_nextpkt = NULL;
1805
1806 /*
1807 * we assume ifp is not a p2p here, so
1808 * just set the 2nd argument as the
1809 * 1st one.
1810 */
1811 nd6_output(ifp, ifp, m_hold,
1812 (struct sockaddr_in6 *)rt_key(rt),
1813 rt);
1814 }
1815 }
1816 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1817 /* probe right away */
1818 nd6_llinfo_settimer((void *)ln, 0);
1819 }
1820 }
1821
1822 /*
1823 * ICMP6 type dependent behavior.
1824 *
1825 * NS: clear IsRouter if new entry
1826 * RS: clear IsRouter
1827 * RA: set IsRouter if there's lladdr
1828 * redir: clear IsRouter if new entry
1829 *
1830 * RA case, (1):
1831 * The spec says that we must set IsRouter in the following cases:
1832 * - If lladdr exist, set IsRouter. This means (1-5).
1833 * - If it is old entry (!newentry), set IsRouter. This means (7).
1834 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1835 * A quetion arises for (1) case. (1) case has no lladdr in the
1836 * neighbor cache, this is similar to (6).
1837 * This case is rare but we figured that we MUST NOT set IsRouter.
1838 *
1839 * newentry olladdr lladdr llchange NS RS RA redir
1840 * D R
1841 * 0 n n -- (1) c ? s
1842 * 0 y n -- (2) c s s
1843 * 0 n y -- (3) c s s
1844 * 0 y y n (4) c s s
1845 * 0 y y y (5) c s s
1846 * 1 -- n -- (6) c c c s
1847 * 1 -- y -- (7) c c s c s
1848 *
1849 * (c=clear s=set)
1850 */
1851 switch (type & 0xff) {
1852 case ND_NEIGHBOR_SOLICIT:
1853 /*
1854 * New entry must have is_router flag cleared.
1855 */
1856 if (is_newentry) /* (6-7) */
1857 ln->ln_router = 0;
1858 break;
1859 case ND_REDIRECT:
1860 /*
1861 * If the icmp is a redirect to a better router, always set the
1862 * is_router flag. Otherwise, if the entry is newly created,
1863 * clear the flag. [RFC 2461, sec 8.3]
1864 */
1865 if (code == ND_REDIRECT_ROUTER)
1866 ln->ln_router = 1;
1867 else if (is_newentry) /* (6-7) */
1868 ln->ln_router = 0;
1869 break;
1870 case ND_ROUTER_SOLICIT:
1871 /*
1872 * is_router flag must always be cleared.
1873 */
1874 ln->ln_router = 0;
1875 break;
1876 case ND_ROUTER_ADVERT:
1877 /*
1878 * Mark an entry with lladdr as a router.
1879 */
1880 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1881 (is_newentry && lladdr)) { /* (7) */
1882 ln->ln_router = 1;
1883 }
1884 break;
1885 }
1886
1887 /*
1888 * When the link-layer address of a router changes, select the
1889 * best router again. In particular, when the neighbor entry is newly
1890 * created, it might affect the selection policy.
1891 * Question: can we restrict the first condition to the "is_newentry"
1892 * case?
1893 * XXX: when we hear an RA from a new router with the link-layer
1894 * address option, defrouter_select() is called twice, since
1895 * defrtrlist_update called the function as well. However, I believe
1896 * we can compromise the overhead, since it only happens the first
1897 * time.
1898 * XXX: although defrouter_select() should not have a bad effect
1899 * for those are not autoconfigured hosts, we explicitly avoid such
1900 * cases for safety.
1901 */
1902 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1903 defrouter_select();
1904
1905 return rt;
1906 }
1907
1908 static void
1909 nd6_slowtimo(ignored_arg)
1910 void *ignored_arg;
1911 {
1912 struct nd_ifinfo *nd6if;
1913 struct ifnet *ifp;
1914
1915 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1916 nd6_slowtimo, NULL);
1917 IFNET_RLOCK();
1918 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1919 nd6if = ND_IFINFO(ifp);
1920 if (nd6if->basereachable && /* already initialized */
1921 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1922 /*
1923 * Since reachable time rarely changes by router
1924 * advertisements, we SHOULD insure that a new random
1925 * value gets recomputed at least once every few hours.
1926 * (RFC 2461, 6.3.4)
1927 */
1928 nd6if->recalctm = nd6_recalc_reachtm_interval;
1929 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1930 }
1931 }
1932 IFNET_RUNLOCK();
1933 }
1934
1935 #define senderr(e) { error = (e); goto bad;}
1936 int
1937 nd6_output(ifp, origifp, m0, dst, rt0)
1938 struct ifnet *ifp;
1939 struct ifnet *origifp;
1940 struct mbuf *m0;
1941 struct sockaddr_in6 *dst;
1942 struct rtentry *rt0;
1943 {
1944 struct mbuf *m = m0;
1945 struct rtentry *rt = rt0;
1946 struct sockaddr_in6 *gw6 = NULL;
1947 struct llinfo_nd6 *ln = NULL;
1948 int error = 0;
1949
1950 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1951 goto sendpkt;
1952
1953 if (nd6_need_cache(ifp) == 0)
1954 goto sendpkt;
1955
1956 /*
1957 * next hop determination. This routine is derived from ether_output.
1958 */
1959 /* NB: the locking here is tortuous... */
1960 if (rt != NULL)
1961 RT_LOCK(rt);
1962 again:
1963 if (rt != NULL) {
1964 if ((rt->rt_flags & RTF_UP) == 0) {
1965 RT_UNLOCK(rt);
1966 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL);
1967 if (rt != NULL) {
1968 RT_REMREF(rt);
1969 if (rt->rt_ifp != ifp)
1970 /*
1971 * XXX maybe we should update ifp too,
1972 * but the original code didn't and I
1973 * don't know what is correct here.
1974 */
1975 goto again;
1976 } else
1977 senderr(EHOSTUNREACH);
1978 }
1979
1980 if (rt->rt_flags & RTF_GATEWAY) {
1981 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1982
1983 /*
1984 * We skip link-layer address resolution and NUD
1985 * if the gateway is not a neighbor from ND point
1986 * of view, regardless of the value of nd_ifinfo.flags.
1987 * The second condition is a bit tricky; we skip
1988 * if the gateway is our own address, which is
1989 * sometimes used to install a route to a p2p link.
1990 */
1991 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1992 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1993 RT_UNLOCK(rt);
1994 /*
1995 * We allow this kind of tricky route only
1996 * when the outgoing interface is p2p.
1997 * XXX: we may need a more generic rule here.
1998 */
1999 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2000 senderr(EHOSTUNREACH);
2001
2002 goto sendpkt;
2003 }
2004
2005 if (rt->rt_gwroute == NULL)
2006 goto lookup;
2007 rt = rt->rt_gwroute;
2008 RT_LOCK(rt); /* NB: gwroute */
2009 if ((rt->rt_flags & RTF_UP) == 0) {
2010 rtfree(rt); /* unlock gwroute */
2011 rt = rt0;
2012 rt0->rt_gwroute = NULL;
2013 lookup:
2014 RT_UNLOCK(rt0);
2015 rt = rtalloc1(rt->rt_gateway, 1, 0UL);
2016 if (rt == rt0) {
2017 RT_REMREF(rt0);
2018 RT_UNLOCK(rt0);
2019 senderr(EHOSTUNREACH);
2020 }
2021 RT_LOCK(rt0);
2022 if (rt0->rt_gwroute != NULL)
2023 RTFREE(rt0->rt_gwroute);
2024 rt0->rt_gwroute = rt;
2025 if (rt == NULL) {
2026 RT_UNLOCK(rt0);
2027 senderr(EHOSTUNREACH);
2028 }
2029 }
2030 RT_UNLOCK(rt0);
2031 }
2032 RT_UNLOCK(rt);
2033 }
2034
2035 /*
2036 * Address resolution or Neighbor Unreachability Detection
2037 * for the next hop.
2038 * At this point, the destination of the packet must be a unicast
2039 * or an anycast address(i.e. not a multicast).
2040 */
2041
2042 /* Look up the neighbor cache for the nexthop */
2043 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
2044 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2045 else {
2046 /*
2047 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2048 * the condition below is not very efficient. But we believe
2049 * it is tolerable, because this should be a rare case.
2050 */
2051 if (nd6_is_addr_neighbor(dst, ifp) &&
2052 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
2053 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2054 }
2055 if (ln == NULL || rt == NULL) {
2056 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2057 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2058 log(LOG_DEBUG,
2059 "nd6_output: can't allocate llinfo for %s "
2060 "(ln=%p, rt=%p)\n",
2061 ip6_sprintf(&dst->sin6_addr), ln, rt);
2062 senderr(EIO); /* XXX: good error? */
2063 }
2064
2065 goto sendpkt; /* send anyway */
2066 }
2067
2068 /* We don't have to do link-layer address resolution on a p2p link. */
2069 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2070 ln->ln_state < ND6_LLINFO_REACHABLE) {
2071 ln->ln_state = ND6_LLINFO_STALE;
2072 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
2073 }
2074
2075 /*
2076 * The first time we send a packet to a neighbor whose entry is
2077 * STALE, we have to change the state to DELAY and a sets a timer to
2078 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2079 * neighbor unreachability detection on expiration.
2080 * (RFC 2461 7.3.3)
2081 */
2082 if (ln->ln_state == ND6_LLINFO_STALE) {
2083 ln->ln_asked = 0;
2084 ln->ln_state = ND6_LLINFO_DELAY;
2085 nd6_llinfo_settimer(ln, (long)nd6_delay * hz);
2086 }
2087
2088 /*
2089 * If the neighbor cache entry has a state other than INCOMPLETE
2090 * (i.e. its link-layer address is already resolved), just
2091 * send the packet.
2092 */
2093 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2094 goto sendpkt;
2095
2096 /*
2097 * There is a neighbor cache entry, but no ethernet address
2098 * response yet. Append this latest packet to the end of the
2099 * packet queue in the mbuf, unless the number of the packet
2100 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2101 * the oldest packet in the queue will be removed.
2102 */
2103 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2104 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2105 if (ln->ln_hold) {
2106 struct mbuf *m_hold;
2107 int i;
2108
2109 i = 0;
2110 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) {
2111 i++;
2112 if (m_hold->m_nextpkt == NULL) {
2113 m_hold->m_nextpkt = m;
2114 break;
2115 }
2116 }
2117 while (i >= nd6_maxqueuelen) {
2118 m_hold = ln->ln_hold;
2119 ln->ln_hold = ln->ln_hold->m_nextpkt;
2120 m_freem(m_hold);
2121 i--;
2122 }
2123 } else {
2124 ln->ln_hold = m;
2125 }
2126
2127 /*
2128 * If there has been no NS for the neighbor after entering the
2129 * INCOMPLETE state, send the first solicitation.
2130 */
2131 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
2132 ln->ln_asked++;
2133 nd6_llinfo_settimer(ln,
2134 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2135 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2136 }
2137 return (0);
2138
2139 sendpkt:
2140 /* discard the packet if IPv6 operation is disabled on the interface */
2141 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2142 error = ENETDOWN; /* better error? */
2143 goto bad;
2144 }
2145
2146 #ifdef IPSEC
2147 /* clean ipsec history once it goes out of the node */
2148 ipsec_delaux(m);
2149 #endif
2150
2151 #ifdef MAC
2152 mac_create_mbuf_linklayer(ifp, m);
2153 #endif
2154 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2155 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2156 rt));
2157 }
2158 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2159
2160 bad:
2161 if (m)
2162 m_freem(m);
2163 return (error);
2164 }
2165 #undef senderr
2166
2167 int
2168 nd6_need_cache(ifp)
2169 struct ifnet *ifp;
2170 {
2171 /*
2172 * XXX: we currently do not make neighbor cache on any interface
2173 * other than ARCnet, Ethernet, FDDI and GIF.
2174 *
2175 * RFC2893 says:
2176 * - unidirectional tunnels needs no ND
2177 */
2178 switch (ifp->if_type) {
2179 case IFT_ARCNET:
2180 case IFT_ETHER:
2181 case IFT_FDDI:
2182 case IFT_IEEE1394:
2183 #ifdef IFT_L2VLAN
2184 case IFT_L2VLAN:
2185 #endif
2186 #ifdef IFT_IEEE80211
2187 case IFT_IEEE80211:
2188 #endif
2189 #ifdef IFT_CARP
2190 case IFT_CARP:
2191 #endif
2192 case IFT_GIF: /* XXX need more cases? */
2193 case IFT_PPP:
2194 case IFT_TUNNEL:
2195 case IFT_BRIDGE:
2196 case IFT_PROPVIRTUAL:
2197 return (1);
2198 default:
2199 return (0);
2200 }
2201 }
2202
2203 int
2204 nd6_storelladdr(ifp, rt0, m, dst, desten)
2205 struct ifnet *ifp;
2206 struct rtentry *rt0;
2207 struct mbuf *m;
2208 struct sockaddr *dst;
2209 u_char *desten;
2210 {
2211 struct sockaddr_dl *sdl;
2212 struct rtentry *rt;
2213 int error;
2214
2215 if (m->m_flags & M_MCAST) {
2216 int i;
2217
2218 switch (ifp->if_type) {
2219 case IFT_ETHER:
2220 case IFT_FDDI:
2221 #ifdef IFT_L2VLAN
2222 case IFT_L2VLAN:
2223 #endif
2224 #ifdef IFT_IEEE80211
2225 case IFT_IEEE80211:
2226 #endif
2227 case IFT_BRIDGE:
2228 case IFT_ISO88025:
2229 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2230 desten);
2231 return (0);
2232 case IFT_IEEE1394:
2233 /*
2234 * netbsd can use if_broadcastaddr, but we don't do so
2235 * to reduce # of ifdef.
2236 */
2237 for (i = 0; i < ifp->if_addrlen; i++)
2238 desten[i] = ~0;
2239 return (0);
2240 case IFT_ARCNET:
2241 *desten = 0;
2242 return (0);
2243 default:
2244 m_freem(m);
2245 return (EAFNOSUPPORT);
2246 }
2247 }
2248
2249 if (rt0 == NULL) {
2250 /* this could happen, if we could not allocate memory */
2251 m_freem(m);
2252 return (ENOMEM);
2253 }
2254
2255 error = rt_check(&rt, &rt0, dst);
2256 if (error) {
2257 m_freem(m);
2258 return (error);
2259 }
2260 RT_UNLOCK(rt);
2261
2262 if (rt->rt_gateway->sa_family != AF_LINK) {
2263 printf("nd6_storelladdr: something odd happens\n");
2264 m_freem(m);
2265 return (EINVAL);
2266 }
2267 sdl = SDL(rt->rt_gateway);
2268 if (sdl->sdl_alen == 0) {
2269 /* this should be impossible, but we bark here for debugging */
2270 printf("nd6_storelladdr: sdl_alen == 0\n");
2271 m_freem(m);
2272 return (EINVAL);
2273 }
2274
2275 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2276 return (0);
2277 }
2278
2279 static void
2280 clear_llinfo_pqueue(ln)
2281 struct llinfo_nd6 *ln;
2282 {
2283 struct mbuf *m_hold, *m_hold_next;
2284
2285 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) {
2286 m_hold_next = m_hold->m_nextpkt;
2287 m_hold->m_nextpkt = NULL;
2288 m_freem(m_hold);
2289 }
2290
2291 ln->ln_hold = NULL;
2292 return;
2293 }
2294
2295 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2296 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2297 #ifdef SYSCTL_DECL
2298 SYSCTL_DECL(_net_inet6_icmp6);
2299 #endif
2300 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2301 CTLFLAG_RD, nd6_sysctl_drlist, "");
2302 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2303 CTLFLAG_RD, nd6_sysctl_prlist, "");
2304 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2305 CTLFLAG_RW, &nd6_maxqueuelen, 1, "");
2306
2307 static int
2308 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2309 {
2310 int error;
2311 char buf[1024];
2312 struct in6_defrouter *d, *de;
2313 struct nd_defrouter *dr;
2314
2315 if (req->newptr)
2316 return EPERM;
2317 error = 0;
2318
2319 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
2320 dr = TAILQ_NEXT(dr, dr_entry)) {
2321 d = (struct in6_defrouter *)buf;
2322 de = (struct in6_defrouter *)(buf + sizeof(buf));
2323
2324 if (d + 1 <= de) {
2325 bzero(d, sizeof(*d));
2326 d->rtaddr.sin6_family = AF_INET6;
2327 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2328 d->rtaddr.sin6_addr = dr->rtaddr;
2329 sa6_recoverscope(&d->rtaddr);
2330 d->flags = dr->flags;
2331 d->rtlifetime = dr->rtlifetime;
2332 d->expire = dr->expire;
2333 d->if_index = dr->ifp->if_index;
2334 } else
2335 panic("buffer too short");
2336
2337 error = SYSCTL_OUT(req, buf, sizeof(*d));
2338 if (error)
2339 break;
2340 }
2341
2342 return (error);
2343 }
2344
2345 static int
2346 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2347 {
2348 int error;
2349 char buf[1024];
2350 struct in6_prefix *p, *pe;
2351 struct nd_prefix *pr;
2352
2353 if (req->newptr)
2354 return EPERM;
2355 error = 0;
2356
2357 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2358 u_short advrtrs;
2359 size_t advance;
2360 struct sockaddr_in6 *sin6, *s6;
2361 struct nd_pfxrouter *pfr;
2362
2363 p = (struct in6_prefix *)buf;
2364 pe = (struct in6_prefix *)(buf + sizeof(buf));
2365
2366 if (p + 1 <= pe) {
2367 bzero(p, sizeof(*p));
2368 sin6 = (struct sockaddr_in6 *)(p + 1);
2369
2370 p->prefix = pr->ndpr_prefix;
2371 if (sa6_recoverscope(&p->prefix)) {
2372 log(LOG_ERR,
2373 "scope error in prefix list (%s)\n",
2374 ip6_sprintf(&p->prefix.sin6_addr));
2375 /* XXX: press on... */
2376 }
2377 p->raflags = pr->ndpr_raf;
2378 p->prefixlen = pr->ndpr_plen;
2379 p->vltime = pr->ndpr_vltime;
2380 p->pltime = pr->ndpr_pltime;
2381 p->if_index = pr->ndpr_ifp->if_index;
2382 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2383 p->expire = 0;
2384 else {
2385 time_t maxexpire;
2386
2387 /* XXX: we assume time_t is signed. */
2388 maxexpire = (-1) &
2389 ~((time_t)1 <<
2390 ((sizeof(maxexpire) * 8) - 1));
2391 if (pr->ndpr_vltime <
2392 maxexpire - pr->ndpr_lastupdate) {
2393 p->expire = pr->ndpr_lastupdate +
2394 pr->ndpr_vltime;
2395 } else
2396 p->expire = maxexpire;
2397 }
2398 p->refcnt = pr->ndpr_refcnt;
2399 p->flags = pr->ndpr_stateflags;
2400 p->origin = PR_ORIG_RA;
2401 advrtrs = 0;
2402 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2403 pfr = pfr->pfr_next) {
2404 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2405 advrtrs++;
2406 continue;
2407 }
2408 s6 = &sin6[advrtrs];
2409 bzero(s6, sizeof(*s6));
2410 s6->sin6_family = AF_INET6;
2411 s6->sin6_len = sizeof(*sin6);
2412 s6->sin6_addr = pfr->router->rtaddr;
2413 if (sa6_recoverscope(s6)) {
2414 log(LOG_ERR,
2415 "scope error in "
2416 "prefix list (%s)\n",
2417 ip6_sprintf(&pfr->router->rtaddr));
2418 }
2419 advrtrs++;
2420 }
2421 p->advrtrs = advrtrs;
2422 } else
2423 panic("buffer too short");
2424
2425 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2426 error = SYSCTL_OUT(req, buf, advance);
2427 if (error)
2428 break;
2429 }
2430
2431 return (error);
2432 }
Cache object: 3af74c501951cac2575b4842da087761
|