]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet6/nd6.c
projects / apple / xnu.git / blob
? search:


f0c838d588a09a52a7009331ab8531f83b23ad96
[apple/xnu.git] / bsd / netinet6 / nd6.c
1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.20 2002/08/02 20:49:14 rwatson Exp $ */
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 /*
34 * XXX
35 * KAME 970409 note:
36 * BSD/OS version heavily modifies this code, related to llinfo.
37 * Since we don't have BSD/OS version of net/route.c in our hand,
38 * I left the code mostly as it was in 970310. -- itojun
39 */
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/sockio.h>
47 #include <sys/time.h>
48 #include <sys/kernel.h>
49 #include <sys/sysctl.h>
50 #include <sys/errno.h>
51 #include <sys/syslog.h>
52 #include <sys/protosw.h>
53 #include <kern/queue.h>
54 #include <kern/lock.h>
55
56 #define DONT_WARN_OBSOLETE
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/if_atm.h>
61 #include <net/route.h>
62 #include <net/dlil.h>
63
64 #include <netinet/in.h>
65 #include <netinet/if_ether.h>
66 #include <netinet/if_fddi.h>
67 #include <netinet6/in6_var.h>
68 #include <netinet/ip6.h>
69 #include <netinet6/ip6_var.h>
70 #include <netinet6/nd6.h>
71 #include <netinet6/in6_prefix.h>
72 #include <netinet/icmp6.h>
73
74 #include "loop.h"
75
76 #include <net/net_osdep.h>
77
78 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
79 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
80
81 #define SIN6(s) ((struct sockaddr_in6 *)s)
82 #define SDL(s) ((struct sockaddr_dl *)s)
83
84 /* timer values */
85 int nd6_prune = 1; /* walk list every 1 seconds */
86 int nd6_delay = 5; /* delay first probe time 5 second */
87 int nd6_umaxtries = 3; /* maximum unicast query */
88 int nd6_mmaxtries = 3; /* maximum multicast query */
89 int nd6_useloopback = 1; /* use loopback interface for local traffic */
90 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
91
92 /* preventing too many loops in ND option parsing */
93 int nd6_maxndopt = 10; /* max # of ND options allowed */
94
95 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
96
97 #if ND6_DEBUG
98 int nd6_debug = 1;
99 #else
100 int nd6_debug = 0;
101 #endif
102
103 /* for debugging? */
104 static int nd6_inuse, nd6_allocated;
105
106 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6, NULL, NULL, 0, 0, 0, 0, 0 };
107 size_t nd_ifinfo_indexlim = 32; /* increased for 5589193 */
108 struct nd_ifinfo *nd_ifinfo = NULL;
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 int regen_tmpaddr(struct in6_ifaddr *);
116 extern lck_mtx_t *rt_mtx;
117 extern lck_mtx_t *ip6_mutex;
118 extern lck_mtx_t *nd6_mutex;
119
120 static void nd6_slowtimo(void *ignored_arg);
121
122 void
123 nd6_init()
124 {
125 static int nd6_init_done = 0;
126 int i;
127
128 if (nd6_init_done) {
129 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
130 return;
131 }
132
133 all1_sa.sin6_family = AF_INET6;
134 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
135 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
136 all1_sa.sin6_addr.s6_addr[i] = 0xff;
137
138 /* initialization of the default router list */
139 TAILQ_INIT(&nd_defrouter);
140
141 nd6_init_done = 1;
142
143 /* start timer */
144 timeout(nd6_slowtimo, (caddr_t)0, ND6_SLOWTIMER_INTERVAL * hz);
145 }
146
147 void
148 nd6_ifattach(
149 struct ifnet *ifp)
150 {
151
152 /*
153 * We have some arrays that should be indexed by if_index.
154 * since if_index will grow dynamically, they should grow too.
155 */
156 if (nd_ifinfo == NULL || if_index >= nd_ifinfo_indexlim) {
157 size_t n;
158 caddr_t q;
159
160 while (if_index >= nd_ifinfo_indexlim)
161 nd_ifinfo_indexlim <<= 1;
162
163 /* grow nd_ifinfo */
164 n = nd_ifinfo_indexlim * sizeof(struct nd_ifinfo);
165 q = (caddr_t)_MALLOC(n, M_IP6NDP, M_WAITOK);
166 bzero(q, n);
167 if (nd_ifinfo) {
168 bcopy((caddr_t)nd_ifinfo, q, n/2);
169 /* Radar 5589193:
170 * SU fix purposely leaks the old nd_ifinfo array
171 * if we grow the arraw to more than 32 interfaces
172 * Fix for future release is to use proper locking.
173
174 FREE((caddr_t)nd_ifinfo, M_IP6NDP);
175 */
176 }
177 nd_ifinfo = (struct nd_ifinfo *)q;
178 }
179
180 #define ND nd_ifinfo[ifp->if_index]
181
182 /*
183 * Don't initialize if called twice.
184 * XXX: to detect this, we should choose a member that is never set
185 * before initialization of the ND structure itself. We formaly used
186 * the linkmtu member, which was not suitable because it could be
187 * initialized via "ifconfig mtu".
188 */
189 if (ND.basereachable)
190 return;
191
192 ND.linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu;
193 ND.chlim = IPV6_DEFHLIM;
194 ND.basereachable = REACHABLE_TIME;
195 ND.reachable = ND_COMPUTE_RTIME(ND.basereachable);
196 ND.retrans = RETRANS_TIMER;
197 ND.receivedra = 0;
198 ND.flags = ND6_IFF_PERFORMNUD;
199 nd6_setmtu(ifp);
200 #undef ND
201 }
202
203 /*
204 * Reset ND level link MTU. This function is called when the physical MTU
205 * changes, which means we might have to adjust the ND level MTU.
206 */
207 void
208 nd6_setmtu(struct ifnet *ifp)
209 {
210 struct nd_ifinfo *ndi;
211 u_long oldmaxmtu;
212
213 /*
214 * Make sure IPv6 is enabled for the interface first,
215 * because this can be called directly from SIOCSIFMTU for IPv4
216 */
217
218 if (ifp->if_index >= nd_ifinfo_indexlim) {
219 return; /* we're out of bound for nd_ifinfo */
220 }
221
222 ndi = &nd_ifinfo[ifp->if_index];
223 oldmaxmtu = ndi->maxmtu;
224
225 /*
226 * The ND level maxmtu is somewhat redundant to the interface MTU
227 * and is an implementation artifact of KAME. Instead of hard-
228 * limiting the maxmtu based on the interface type here, we simply
229 * take the if_mtu value since SIOCSIFMTU would have taken care of
230 * the sanity checks related to the maximum MTU allowed for the
231 * interface (a value that is known only by the interface layer),
232 * by sending the request down via ifnet_ioctl(). The use of the
233 * ND level maxmtu and linkmtu (the latter obtained via RA) are done
234 * via IN6_LINKMTU() which does further checking against if_mtu.
235 */
236 ndi->maxmtu = ifp->if_mtu;
237
238 /*
239 * Decreasing the interface MTU under IPV6 minimum MTU may cause
240 * undesirable situation. We thus notify the operator of the change
241 * explicitly. The check for oldmaxmtu is necessary to restrict the
242 * log to the case of changing the MTU, not initializing it.
243 */
244 if (oldmaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
245 log(LOG_NOTICE, "nd6_setmtu: "
246 "new link MTU on %s%d (%lu) is too small for IPv6\n",
247 ifp->if_name, ifp->if_unit, (unsigned long)ndi->maxmtu);
248 }
249
250 /* also adjust in6_maxmtu if necessary. */
251 if (ndi->maxmtu > in6_maxmtu)
252 in6_setmaxmtu();
253 }
254
255 void
256 nd6_option_init(
257 void *opt,
258 int icmp6len,
259 union nd_opts *ndopts)
260 {
261 bzero(ndopts, sizeof(*ndopts));
262 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
263 ndopts->nd_opts_last
264 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
265
266 if (icmp6len == 0) {
267 ndopts->nd_opts_done = 1;
268 ndopts->nd_opts_search = NULL;
269 }
270 }
271
272 /*
273 * Take one ND option.
274 */
275 struct nd_opt_hdr *
276 nd6_option(
277 union nd_opts *ndopts)
278 {
279 struct nd_opt_hdr *nd_opt;
280 int olen;
281
282 if (!ndopts)
283 panic("ndopts == NULL in nd6_option\n");
284 if (!ndopts->nd_opts_last)
285 panic("uninitialized ndopts in nd6_option\n");
286 if (!ndopts->nd_opts_search)
287 return NULL;
288 if (ndopts->nd_opts_done)
289 return NULL;
290
291 nd_opt = ndopts->nd_opts_search;
292
293 /* make sure nd_opt_len is inside the buffer */
294 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
295 bzero(ndopts, sizeof(*ndopts));
296 return NULL;
297 }
298
299 olen = nd_opt->nd_opt_len << 3;
300 if (olen == 0) {
301 /*
302 * Message validation requires that all included
303 * options have a length that is greater than zero.
304 */
305 bzero(ndopts, sizeof(*ndopts));
306 return NULL;
307 }
308
309 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
310 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
311 /* option overruns the end of buffer, invalid */
312 bzero(ndopts, sizeof(*ndopts));
313 return NULL;
314 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
315 /* reached the end of options chain */
316 ndopts->nd_opts_done = 1;
317 ndopts->nd_opts_search = NULL;
318 }
319 return nd_opt;
320 }
321
322 /*
323 * Parse multiple ND options.
324 * This function is much easier to use, for ND routines that do not need
325 * multiple options of the same type.
326 */
327 int
328 nd6_options(
329 union nd_opts *ndopts)
330 {
331 struct nd_opt_hdr *nd_opt;
332 int i = 0;
333
334 if (!ndopts)
335 panic("ndopts == NULL in nd6_options\n");
336 if (!ndopts->nd_opts_last)
337 panic("uninitialized ndopts in nd6_options\n");
338 if (!ndopts->nd_opts_search)
339 return 0;
340
341 while (1) {
342 nd_opt = nd6_option(ndopts);
343 if (!nd_opt && !ndopts->nd_opts_last) {
344 /*
345 * Message validation requires that all included
346 * options have a length that is greater than zero.
347 */
348 icmp6stat.icp6s_nd_badopt++;
349 bzero(ndopts, sizeof(*ndopts));
350 return -1;
351 }
352
353 if (!nd_opt)
354 goto skip1;
355
356 switch (nd_opt->nd_opt_type) {
357 case ND_OPT_SOURCE_LINKADDR:
358 case ND_OPT_TARGET_LINKADDR:
359 case ND_OPT_MTU:
360 case ND_OPT_REDIRECTED_HEADER:
361 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
362 nd6log((LOG_INFO,
363 "duplicated ND6 option found (type=%d)\n",
364 nd_opt->nd_opt_type));
365 /* XXX bark? */
366 } else {
367 ndopts->nd_opt_array[nd_opt->nd_opt_type]
368 = nd_opt;
369 }
370 break;
371 case ND_OPT_PREFIX_INFORMATION:
372 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
373 ndopts->nd_opt_array[nd_opt->nd_opt_type]
374 = nd_opt;
375 }
376 ndopts->nd_opts_pi_end =
377 (struct nd_opt_prefix_info *)nd_opt;
378 break;
379 default:
380 /*
381 * Unknown options must be silently ignored,
382 * to accomodate future extension to the protocol.
383 */
384 nd6log((LOG_DEBUG,
385 "nd6_options: unsupported option %d - "
386 "option ignored\n", nd_opt->nd_opt_type));
387 }
388
389 skip1:
390 i++;
391 if (i > nd6_maxndopt) {
392 icmp6stat.icp6s_nd_toomanyopt++;
393 nd6log((LOG_INFO, "too many loop in nd opt\n"));
394 break;
395 }
396
397 if (ndopts->nd_opts_done)
398 break;
399 }
400
401 return 0;
402 }
403
404 /*
405 * ND6 timer routine to expire default route list and prefix list
406 */
407 void
408 nd6_timer(
409 __unused void *ignored_arg)
410 {
411 struct llinfo_nd6 *ln;
412 struct nd_defrouter *dr;
413 struct nd_prefix *pr;
414 struct ifnet *ifp = NULL;
415 struct in6_ifaddr *ia6, *nia6;
416 struct in6_addrlifetime *lt6;
417 struct timeval timenow;
418
419 getmicrotime(&timenow);
420
421
422
423 ln = llinfo_nd6.ln_next;
424 while (ln && ln != &llinfo_nd6) {
425 struct rtentry *rt;
426 struct sockaddr_in6 *dst;
427 struct llinfo_nd6 *next = ln->ln_next;
428 /* XXX: used for the DELAY case only: */
429 struct nd_ifinfo *ndi = NULL;
430
431 if ((rt = ln->ln_rt) == NULL) {
432 ln = next;
433 continue;
434 }
435 if ((ifp = rt->rt_ifp) == NULL) {
436 ln = next;
437 continue;
438 }
439 ndi = &nd_ifinfo[ifp->if_index];
440 dst = (struct sockaddr_in6 *)rt_key(rt);
441
442 if (ln->ln_expire > timenow.tv_sec) {
443 ln = next;
444 continue;
445 }
446
447 /* sanity check */
448 if (!rt) {
449 printf("rt=0 in nd6_timer(ln=%p)\n", ln);
450 ln = next;
451 continue;
452 }
453 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) {
454 printf("rt_llinfo(%p) is not equal to ln(%p)\n",
455 rt->rt_llinfo, ln);
456 ln = next;
457 continue;
458 }
459 if (!dst) {
460 printf("dst=0 in nd6_timer(ln=%p)\n", ln);
461 ln = next;
462 continue;
463 }
464
465 switch (ln->ln_state) {
466 case ND6_LLINFO_INCOMPLETE:
467 if (ln->ln_asked < nd6_mmaxtries) {
468 ln->ln_asked++;
469 ln->ln_expire = timenow.tv_sec +
470 nd_ifinfo[ifp->if_index].retrans / 1000;
471 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
472 ln, 0, 0);
473 } else {
474 struct mbuf *m = ln->ln_hold;
475 ln->ln_hold = NULL;
476 if (m) {
477 if (rt->rt_ifp) {
478 /*
479 * Fake rcvif to make ICMP error
480 * more helpful in diagnosing
481 * for the receiver.
482 * XXX: should we consider
483 * older rcvif?
484 */
485 m->m_pkthdr.rcvif = rt->rt_ifp;
486 }
487 icmp6_error(m, ICMP6_DST_UNREACH,
488 ICMP6_DST_UNREACH_ADDR, 0);
489 ln->ln_hold = NULL;
490 }
491 next = nd6_free(rt);
492 }
493 break;
494 case ND6_LLINFO_REACHABLE:
495 if (ln->ln_expire) {
496 ln->ln_state = ND6_LLINFO_STALE;
497 ln->ln_expire = timenow.tv_sec + nd6_gctimer;
498 }
499 break;
500
501 case ND6_LLINFO_STALE:
502 /* Garbage Collection(RFC 2461 5.3) */
503 if (ln->ln_expire)
504 next = nd6_free(rt);
505 break;
506
507 case ND6_LLINFO_DELAY:
508 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
509 /* We need NUD */
510 ln->ln_asked = 1;
511 ln->ln_state = ND6_LLINFO_PROBE;
512 ln->ln_expire = timenow.tv_sec +
513 ndi->retrans / 1000;
514 nd6_ns_output(ifp, &dst->sin6_addr,
515 &dst->sin6_addr,
516 ln, 0, 0);
517 } else {
518 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
519 ln->ln_expire = timenow.tv_sec + nd6_gctimer;
520 }
521 break;
522 case ND6_LLINFO_PROBE:
523 if (ln->ln_asked < nd6_umaxtries) {
524 ln->ln_asked++;
525 ln->ln_expire = timenow.tv_sec +
526 nd_ifinfo[ifp->if_index].retrans / 1000;
527 nd6_ns_output(ifp, &dst->sin6_addr,
528 &dst->sin6_addr, ln, 0, 0);
529 } else {
530 next = nd6_free(rt);
531 }
532 break;
533 }
534 ln = next;
535 }
536
537 /* expire default router list */
538 lck_mtx_lock(nd6_mutex);
539 dr = TAILQ_FIRST(&nd_defrouter);
540 while (dr) {
541 if (dr->expire && dr->expire < timenow.tv_sec) {
542 struct nd_defrouter *t;
543 t = TAILQ_NEXT(dr, dr_entry);
544 defrtrlist_del(dr, 1);
545 dr = t;
546 } else {
547 dr = TAILQ_NEXT(dr, dr_entry);
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_ifaddrs; 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 /* NOTE: We have to drop the lock here because
578 * regen_tmpaddr() eventually calls in6_update_ifa(),
579 * which must take the lock and would otherwise cause a
580 * hang. This is safe because the goto addrloop
581 * leads to a reevaluation of the in6_ifaddrs list
582 */
583 lck_mtx_unlock(nd6_mutex);
584 if (regen_tmpaddr(ia6) == 0)
585 regen = 1;
586 lck_mtx_lock(nd6_mutex);
587 }
588
589 in6_purgeaddr(&ia6->ia_ifa, 1);
590
591 if (regen)
592 goto addrloop; /* XXX: see below */
593 }
594 if (IFA6_IS_DEPRECATED(ia6)) {
595 int oldflags = ia6->ia6_flags;
596
597 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
598
599 /*
600 * If a temporary address has just become deprecated,
601 * regenerate a new one if possible.
602 */
603 if (ip6_use_tempaddr &&
604 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
605 (oldflags & IN6_IFF_DEPRECATED) == 0) {
606
607 /* see NOTE above */
608 lck_mtx_unlock(nd6_mutex);
609 if (regen_tmpaddr(ia6) == 0) {
610 /*
611 * A new temporary address is
612 * generated.
613 * XXX: this means the address chain
614 * has changed while we are still in
615 * the loop. Although the change
616 * would not cause disaster (because
617 * it's not a deletion, but an
618 * addition,) we'd rather restart the
619 * loop just for safety. Or does this
620 * significantly reduce performance??
621 */
622 lck_mtx_lock(nd6_mutex);
623 goto addrloop;
624 }
625 lck_mtx_lock(nd6_mutex);
626 }
627 } else {
628 /*
629 * A new RA might have made a deprecated address
630 * preferred.
631 */
632 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
633 }
634 }
635
636 /* expire prefix list */
637 pr = nd_prefix.lh_first;
638 while (pr) {
639 /*
640 * check prefix lifetime.
641 * since pltime is just for autoconf, pltime processing for
642 * prefix is not necessary.
643 */
644 if (pr->ndpr_expire && pr->ndpr_expire < timenow.tv_sec) {
645 struct nd_prefix *t;
646 t = pr->ndpr_next;
647
648 /*
649 * address expiration and prefix expiration are
650 * separate. NEVER perform in6_purgeaddr here.
651 */
652
653 prelist_remove(pr, 1);
654 pr = t;
655 } else
656 pr = pr->ndpr_next;
657 }
658 lck_mtx_unlock(nd6_mutex);
659 timeout(nd6_timer, (caddr_t)0, nd6_prune * hz);
660 }
661
662 static int
663 regen_tmpaddr(
664 struct in6_ifaddr *ia6) /* deprecated/invalidated temporary address */
665 {
666 struct ifaddr *ifa;
667 struct ifnet *ifp;
668 struct in6_ifaddr *public_ifa6 = NULL;
669 struct timeval timenow;
670
671 getmicrotime(&timenow);
672
673 ifp = ia6->ia_ifa.ifa_ifp;
674 ifnet_lock_exclusive(ifp);
675 for (ifa = ifp->if_addrlist.tqh_first; ifa;
676 ifa = ifa->ifa_list.tqe_next)
677 {
678 struct in6_ifaddr *it6;
679
680 if (ifa->ifa_addr->sa_family != AF_INET6)
681 continue;
682
683 it6 = (struct in6_ifaddr *)ifa;
684
685 /* ignore no autoconf addresses. */
686 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
687 continue;
688
689 /* ignore autoconf addresses with different prefixes. */
690 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
691 continue;
692
693 /*
694 * Now we are looking at an autoconf address with the same
695 * prefix as ours. If the address is temporary and is still
696 * preferred, do not create another one. It would be rare, but
697 * could happen, for example, when we resume a laptop PC after
698 * a long period.
699 */
700 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
701 !IFA6_IS_DEPRECATED(it6)) {
702 public_ifa6 = NULL;
703 break;
704 }
705
706 /*
707 * This is a public autoconf address that has the same prefix
708 * as ours. If it is preferred, keep it. We can't break the
709 * loop here, because there may be a still-preferred temporary
710 * address with the prefix.
711 */
712 if (!IFA6_IS_DEPRECATED(it6))
713 public_ifa6 = it6;
714 }
715 ifnet_lock_done(ifp);
716
717 if (public_ifa6 != NULL) {
718 int e;
719
720 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
721 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
722 " tmp addr,errno=%d\n", e);
723 return(-1);
724 }
725 return(0);
726 }
727
728 return(-1);
729 }
730
731 /*
732 * Nuke neighbor cache/prefix/default router management table, right before
733 * ifp goes away.
734 */
735 void
736 nd6_purge(
737 struct ifnet *ifp)
738 {
739 struct llinfo_nd6 *ln, *nln;
740 struct nd_defrouter *dr, *ndr, drany;
741 struct nd_prefix *pr, *npr;
742
743 /* Nuke default router list entries toward ifp */
744 lck_mtx_lock(nd6_mutex);
745 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
746 /*
747 * The first entry of the list may be stored in
748 * the routing table, so we'll delete it later.
749 */
750 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
751 ndr = TAILQ_NEXT(dr, dr_entry);
752 if (dr->ifp == ifp)
753 defrtrlist_del(dr, 1);
754 }
755 dr = TAILQ_FIRST(&nd_defrouter);
756 if (dr->ifp == ifp)
757 defrtrlist_del(dr, 1);
758 }
759
760 /* Nuke prefix list entries toward ifp */
761 for (pr = nd_prefix.lh_first; pr; pr = npr) {
762 npr = pr->ndpr_next;
763 if (pr->ndpr_ifp == ifp) {
764 /*
765 * Previously, pr->ndpr_addr is removed as well,
766 * but I strongly believe we don't have to do it.
767 * nd6_purge() is only called from in6_ifdetach(),
768 * which removes all the associated interface addresses
769 * by itself.
770 * (jinmei@kame.net 20010129)
771 */
772 prelist_remove(pr, 1);
773 }
774 }
775
776 /* cancel default outgoing interface setting */
777 if (nd6_defifindex == ifp->if_index)
778 nd6_setdefaultiface(0);
779
780 if (!ip6_forwarding && (ip6_accept_rtadv || (ifp->if_eflags & IFEF_ACCEPT_RTADVD))) {
781 /* refresh default router list */
782 bzero(&drany, sizeof(drany));
783 defrouter_delreq(&drany, 0);
784 defrouter_select();
785 }
786 lck_mtx_unlock(nd6_mutex);
787
788 /*
789 * Nuke neighbor cache entries for the ifp.
790 * Note that rt->rt_ifp may not be the same as ifp,
791 * due to KAME goto ours hack. See RTM_RESOLVE case in
792 * nd6_rtrequest(), and ip6_input().
793 */
794 ln = llinfo_nd6.ln_next;
795 while (ln && ln != &llinfo_nd6) {
796 struct rtentry *rt;
797 struct sockaddr_dl *sdl;
798
799 nln = ln->ln_next;
800 rt = ln->ln_rt;
801 if (rt && rt->rt_gateway &&
802 rt->rt_gateway->sa_family == AF_LINK) {
803 sdl = (struct sockaddr_dl *)rt->rt_gateway;
804 if (sdl->sdl_index == ifp->if_index)
805 nln = nd6_free(rt);
806 }
807 ln = nln;
808 }
809 }
810
811 struct rtentry *
812 nd6_lookup(
813 struct in6_addr *addr6,
814 int create,
815 struct ifnet *ifp,
816 int rt_locked)
817 {
818 struct rtentry *rt;
819 struct sockaddr_in6 sin6;
820
821 bzero(&sin6, sizeof(sin6));
822 sin6.sin6_len = sizeof(struct sockaddr_in6);
823 sin6.sin6_family = AF_INET6;
824 sin6.sin6_addr = *addr6;
825 #if SCOPEDROUTING
826 sin6.sin6_scope_id = in6_addr2scopeid(ifp, addr6);
827 #endif
828 if (!rt_locked)
829 lck_mtx_lock(rt_mtx);
830 rt = rtalloc1_locked((struct sockaddr *)&sin6, create, 0UL);
831 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
832 /*
833 * This is the case for the default route.
834 * If we want to create a neighbor cache for the address, we
835 * should free the route for the destination and allocate an
836 * interface route.
837 */
838 if (create) {
839 rtfree_locked(rt);
840 rt = 0;
841 }
842 }
843 if (!rt) {
844 if (create && ifp) {
845 int e;
846
847 /*
848 * If no route is available and create is set,
849 * we allocate a host route for the destination
850 * and treat it like an interface route.
851 * This hack is necessary for a neighbor which can't
852 * be covered by our own prefix.
853 */
854 struct ifaddr *ifa =
855 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
856 if (ifa == NULL) {
857 if (!rt_locked)
858 lck_mtx_unlock(rt_mtx);
859 return(NULL);
860 }
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_locked(RTM_ADD, (struct sockaddr *)&sin6,
869 ifa->ifa_addr,
870 (struct sockaddr *)&all1_sa,
871 (ifa->ifa_flags |
872 RTF_HOST | RTF_LLINFO) &
873 ~RTF_CLONING,
874 &rt)) != 0) {
875 if (e != EEXIST)
876 log(LOG_ERR,
877 "nd6_lookup: failed to add route for a "
878 "neighbor(%s), errno=%d\n",
879 ip6_sprintf(addr6), e);
880 }
881 ifafree(ifa);
882 if (rt == NULL) {
883 if (!rt_locked)
884 lck_mtx_unlock(rt_mtx);
885 return(NULL);
886 }
887 if (rt->rt_llinfo) {
888 struct llinfo_nd6 *ln =
889 (struct llinfo_nd6 *)rt->rt_llinfo;
890 ln->ln_state = ND6_LLINFO_NOSTATE;
891 }
892 } else {
893 if (!rt_locked)
894 lck_mtx_unlock(rt_mtx);
895 return(NULL);
896 }
897 }
898 rtunref(rt);
899 /*
900 * Validation for the entry.
901 * Note that the check for rt_llinfo is necessary because a cloned
902 * route from a parent route that has the L flag (e.g. the default
903 * route to a p2p interface) may have the flag, too, while the
904 * destination is not actually a neighbor.
905 * XXX: we can't use rt->rt_ifp to check for the interface, since
906 * it might be the loopback interface if the entry is for our
907 * own address on a non-loopback interface. Instead, we should
908 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
909 * interface.
910 */
911 if ((ifp && ifp->if_type !=IFT_PPP) && ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
912 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
913 (ifp && rt->rt_ifa->ifa_ifp != ifp))) {
914 if (!rt_locked)
915 lck_mtx_unlock(rt_mtx);
916 if (create) {
917 log(LOG_DEBUG, "nd6_lookup: failed to lookup %s (if = %s)\n",
918 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
919 /* xxx more logs... kazu */
920 }
921 return(NULL);
922 }
923 if (!rt_locked)
924 lck_mtx_unlock(rt_mtx);
925 return(rt);
926 }
927
928 /*
929 * Detect if a given IPv6 address identifies a neighbor on a given link.
930 * XXX: should take care of the destination of a p2p link?
931 */
932 int
933 nd6_is_addr_neighbor(
934 struct sockaddr_in6 *addr,
935 struct ifnet *ifp,
936 int rt_locked)
937 {
938 struct ifaddr *ifa;
939 int i;
940
941 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
942 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
943
944 /*
945 * A link-local address is always a neighbor.
946 * XXX: we should use the sin6_scope_id field rather than the embedded
947 * interface index.
948 */
949 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
950 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
951 return(1);
952
953 /*
954 * If the address matches one of our addresses,
955 * it should be a neighbor.
956 */
957 ifnet_lock_shared(ifp);
958 for (ifa = ifp->if_addrlist.tqh_first;
959 ifa;
960 ifa = ifa->ifa_list.tqe_next)
961 {
962 if (ifa->ifa_addr->sa_family != AF_INET6)
963 continue;
964
965 for (i = 0; i < 4; i++) {
966 if ((IFADDR6(ifa).s6_addr32[i] ^
967 addr->sin6_addr.s6_addr32[i]) &
968 IFMASK6(ifa).s6_addr32[i])
969 continue;
970 }
971 ifnet_lock_done(ifp);
972 return(1);
973 }
974 ifnet_lock_done(ifp);
975
976 /*
977 * Even if the address matches none of our addresses, it might be
978 * in the neighbor cache.
979 */
980 if (nd6_lookup(&addr->sin6_addr, 0, ifp, rt_locked) != NULL)
981 return(1);
982
983 return(0);
984 #undef IFADDR6
985 #undef IFMASK6
986 }
987
988 /*
989 * Free an nd6 llinfo entry.
990 */
991 struct llinfo_nd6 *
992 nd6_free(
993 struct rtentry *rt)
994 {
995 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
996 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
997 struct nd_defrouter *dr;
998
999 /*
1000 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1001 * even though it is not harmful, it was not really necessary.
1002 */
1003
1004 if (!ip6_forwarding && (ip6_accept_rtadv || (rt->rt_ifp->if_eflags & IFEF_ACCEPT_RTADVD))) {
1005 lck_mtx_lock(nd6_mutex);
1006 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1007 rt->rt_ifp);
1008
1009 if ((ln && ln->ln_router) || dr) {
1010 /*
1011 * rt6_flush must be called whether or not the neighbor
1012 * is in the Default Router List.
1013 * See a corresponding comment in nd6_na_input().
1014 */
1015 rt6_flush(&in6, rt->rt_ifp);
1016 }
1017
1018 if (dr) {
1019 /*
1020 * Unreachablity of a router might affect the default
1021 * router selection and on-link detection of advertised
1022 * prefixes.
1023 */
1024
1025 /*
1026 * Temporarily fake the state to choose a new default
1027 * router and to perform on-link determination of
1028 * prefixes correctly.
1029 * Below the state will be set correctly,
1030 * or the entry itself will be deleted.
1031 */
1032 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1033
1034 /*
1035 * Since defrouter_select() does not affect the
1036 * on-link determination and MIP6 needs the check
1037 * before the default router selection, we perform
1038 * the check now.
1039 */
1040 pfxlist_onlink_check(1);
1041
1042 if (dr == TAILQ_FIRST(&nd_defrouter)) {
1043 /*
1044 * It is used as the current default router,
1045 * so we have to move it to the end of the
1046 * list and choose a new one.
1047 * XXX: it is not very efficient if this is
1048 * the only router.
1049 */
1050 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1051 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
1052
1053 defrouter_select();
1054 }
1055 }
1056 lck_mtx_unlock(nd6_mutex);
1057 }
1058
1059 /*
1060 * Before deleting the entry, remember the next entry as the
1061 * return value. We need this because pfxlist_onlink_check() above
1062 * might have freed other entries (particularly the old next entry) as
1063 * a side effect (XXX).
1064 */
1065 if (ln)
1066 next = ln->ln_next;
1067 else
1068 next = 0;
1069
1070 /*
1071 * Detach the route from the routing tree and the list of neighbor
1072 * caches, and disable the route entry not to be used in already
1073 * cached routes.
1074 */
1075 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1076 rt_mask(rt), 0, (struct rtentry **)0);
1077
1078 return(next);
1079 }
1080
1081 /*
1082 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1083 *
1084 * XXX cost-effective metods?
1085 */
1086 void
1087 nd6_nud_hint(
1088 struct rtentry *rt,
1089 struct in6_addr *dst6,
1090 int force)
1091 {
1092 struct llinfo_nd6 *ln;
1093 struct timeval timenow;
1094
1095 getmicrotime(&timenow);
1096
1097 /*
1098 * If the caller specified "rt", use that. Otherwise, resolve the
1099 * routing table by supplied "dst6".
1100 */
1101 if (!rt) {
1102 if (!dst6)
1103 return;
1104 if (!(rt = nd6_lookup(dst6, 0, NULL, 0)))
1105 return;
1106 }
1107
1108 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1109 (rt->rt_flags & RTF_LLINFO) == 0 ||
1110 !rt->rt_llinfo || !rt->rt_gateway ||
1111 rt->rt_gateway->sa_family != AF_LINK) {
1112 /* This is not a host route. */
1113 return;
1114 }
1115
1116 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1117 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1118 return;
1119
1120 /*
1121 * if we get upper-layer reachability confirmation many times,
1122 * it is possible we have false information.
1123 */
1124 if (!force) {
1125 ln->ln_byhint++;
1126 if (ln->ln_byhint > nd6_maxnudhint)
1127 return;
1128 }
1129
1130 ln->ln_state = ND6_LLINFO_REACHABLE;
1131 if (ln->ln_expire)
1132 ln->ln_expire = timenow.tv_sec +
1133 nd_ifinfo[rt->rt_ifp->if_index].reachable;
1134 }
1135
1136 void
1137 nd6_rtrequest(
1138 int req,
1139 struct rtentry *rt,
1140 __unused struct sockaddr *sa)
1141 {
1142 struct sockaddr *gate = rt->rt_gateway;
1143 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1144 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK, 0, 0, 0, 0, 0,
1145 {0,0,0,0,0,0,0,0,0,0,0,0,} };
1146 struct ifnet *ifp = rt->rt_ifp;
1147 struct ifaddr *ifa;
1148 struct timeval timenow;
1149
1150
1151 if ((rt->rt_flags & RTF_GATEWAY))
1152 return;
1153
1154 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1155 /*
1156 * This is probably an interface direct route for a link
1157 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1158 * We do not need special treatment below for such a route.
1159 * Moreover, the RTF_LLINFO flag which would be set below
1160 * would annoy the ndp(8) command.
1161 */
1162 return;
1163 }
1164
1165 if (req == RTM_RESOLVE &&
1166 (nd6_need_cache(ifp) == 0 || /* stf case */
1167 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp, 1))) {
1168 /*
1169 * FreeBSD and BSD/OS often make a cloned host route based
1170 * on a less-specific route (e.g. the default route).
1171 * If the less specific route does not have a "gateway"
1172 * (this is the case when the route just goes to a p2p or an
1173 * stf interface), we'll mistakenly make a neighbor cache for
1174 * the host route, and will see strange neighbor solicitation
1175 * for the corresponding destination. In order to avoid the
1176 * confusion, we check if the destination of the route is
1177 * a neighbor in terms of neighbor discovery, and stop the
1178 * process if not. Additionally, we remove the LLINFO flag
1179 * so that ndp(8) will not try to get the neighbor information
1180 * of the destination.
1181 */
1182 rt->rt_flags &= ~RTF_LLINFO;
1183 return;
1184 }
1185
1186 getmicrotime(&timenow);
1187 switch (req) {
1188 case RTM_ADD:
1189 /*
1190 * There is no backward compatibility :)
1191 *
1192 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1193 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1194 * rt->rt_flags |= RTF_CLONING;
1195 */
1196 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1197 /*
1198 * Case 1: This route should come from
1199 * a route to interface. RTF_LLINFO flag is set
1200 * for a host route whose destination should be
1201 * treated as on-link.
1202 */
1203 rt_setgate(rt, rt_key(rt),
1204 (struct sockaddr *)&null_sdl);
1205 gate = rt->rt_gateway;
1206 SDL(gate)->sdl_type = ifp->if_type;
1207 SDL(gate)->sdl_index = ifp->if_index;
1208 if (ln)
1209 ln->ln_expire = timenow.tv_sec;
1210 #if 1
1211 if (ln && ln->ln_expire == 0) {
1212 /* kludge for desktops */
1213 #if 0
1214 printf("nd6_rtequest: time.tv_sec is zero; "
1215 "treat it as 1\n");
1216 #endif
1217 ln->ln_expire = 1;
1218 }
1219 #endif
1220 if ((rt->rt_flags & RTF_CLONING))
1221 break;
1222 }
1223 /*
1224 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1225 * We don't do that here since llinfo is not ready yet.
1226 *
1227 * There are also couple of other things to be discussed:
1228 * - unsolicited NA code needs improvement beforehand
1229 * - RFC2461 says we MAY send multicast unsolicited NA
1230 * (7.2.6 paragraph 4), however, it also says that we
1231 * SHOULD provide a mechanism to prevent multicast NA storm.
1232 * we don't have anything like it right now.
1233 * note that the mechanism needs a mutual agreement
1234 * between proxies, which means that we need to implement
1235 * a new protocol, or a new kludge.
1236 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1237 * we need to check ip6forwarding before sending it.
1238 * (or should we allow proxy ND configuration only for
1239 * routers? there's no mention about proxy ND from hosts)
1240 */
1241 #if 0
1242 /* XXX it does not work */
1243 if (rt->rt_flags & RTF_ANNOUNCE)
1244 nd6_na_output(ifp,
1245 &SIN6(rt_key(rt))->sin6_addr,
1246 &SIN6(rt_key(rt))->sin6_addr,
1247 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1248 1, NULL);
1249 #endif
1250 /* FALLTHROUGH */
1251 case RTM_RESOLVE:
1252 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1253 /*
1254 * Address resolution isn't necessary for a point to
1255 * point link, so we can skip this test for a p2p link.
1256 */
1257 if (gate->sa_family != AF_LINK ||
1258 gate->sa_len < sizeof(null_sdl)) {
1259 log(LOG_DEBUG,
1260 "nd6_rtrequest: bad gateway value: %s\n",
1261 if_name(ifp));
1262 break;
1263 }
1264 SDL(gate)->sdl_type = ifp->if_type;
1265 SDL(gate)->sdl_index = ifp->if_index;
1266 }
1267 if (ln != NULL)
1268 break; /* This happens on a route change */
1269 /*
1270 * Case 2: This route may come from cloning, or a manual route
1271 * add with a LL address.
1272 */
1273 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1274 rt->rt_llinfo = (caddr_t)ln;
1275 if (!ln) {
1276 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1277 break;
1278 }
1279 nd6_inuse++;
1280 nd6_allocated++;
1281 Bzero(ln, sizeof(*ln));
1282 ln->ln_rt = rt;
1283 /* this is required for "ndp" command. - shin */
1284 if (req == RTM_ADD) {
1285 /*
1286 * gate should have some valid AF_LINK entry,
1287 * and ln->ln_expire should have some lifetime
1288 * which is specified by ndp command.
1289 */
1290 ln->ln_state = ND6_LLINFO_REACHABLE;
1291 ln->ln_byhint = 0;
1292 } else {
1293 /*
1294 * When req == RTM_RESOLVE, rt is created and
1295 * initialized in rtrequest(), so rt_expire is 0.
1296 */
1297 ln->ln_state = ND6_LLINFO_NOSTATE;
1298 ln->ln_expire = timenow.tv_sec;
1299 }
1300 rt->rt_flags |= RTF_LLINFO;
1301 ln->ln_next = llinfo_nd6.ln_next;
1302 llinfo_nd6.ln_next = ln;
1303 ln->ln_prev = &llinfo_nd6;
1304 ln->ln_next->ln_prev = ln;
1305
1306 /*
1307 * check if rt_key(rt) is one of my address assigned
1308 * to the interface.
1309 */
1310 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1311 &SIN6(rt_key(rt))->sin6_addr);
1312 if (ifa) {
1313 caddr_t macp = nd6_ifptomac(ifp);
1314 ln->ln_expire = 0;
1315 ln->ln_state = ND6_LLINFO_REACHABLE;
1316 ln->ln_byhint = 0;
1317 if (macp) {
1318 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1319 SDL(gate)->sdl_alen = ifp->if_addrlen;
1320 }
1321 if (nd6_useloopback) {
1322 rt->rt_ifp = lo_ifp; /* XXX */
1323 /*
1324 * Make sure rt_ifa be equal to the ifaddr
1325 * corresponding to the address.
1326 * We need this because when we refer
1327 * rt_ifa->ia6_flags in ip6_input, we assume
1328 * that the rt_ifa points to the address instead
1329 * of the loopback address.
1330 */
1331 if (ifa != rt->rt_ifa) {
1332 rtsetifa(rt, ifa);
1333 }
1334 }
1335 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1336 ln->ln_expire = 0;
1337 ln->ln_state = ND6_LLINFO_REACHABLE;
1338 ln->ln_byhint = 0;
1339
1340 /* join solicited node multicast for proxy ND */
1341 if (ifp->if_flags & IFF_MULTICAST) {
1342 struct in6_addr llsol;
1343 int error;
1344
1345 llsol = SIN6(rt_key(rt))->sin6_addr;
1346 llsol.s6_addr16[0] = htons(0xff02);
1347 llsol.s6_addr16[1] = htons(ifp->if_index);
1348 llsol.s6_addr32[1] = 0;
1349 llsol.s6_addr32[2] = htonl(1);
1350 llsol.s6_addr8[12] = 0xff;
1351
1352 if (!in6_addmulti(&llsol, ifp, &error, 0)) {
1353 nd6log((LOG_ERR, "%s: failed to join "
1354 "%s (errno=%d)\n", if_name(ifp),
1355 ip6_sprintf(&llsol), error));
1356 }
1357 }
1358 }
1359 break;
1360
1361 case RTM_DELETE:
1362 if (!ln)
1363 break;
1364 /* leave from solicited node multicast for proxy ND */
1365 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1366 (ifp->if_flags & IFF_MULTICAST) != 0) {
1367 struct in6_addr llsol;
1368 struct in6_multi *in6m;
1369
1370 llsol = SIN6(rt_key(rt))->sin6_addr;
1371 llsol.s6_addr16[0] = htons(0xff02);
1372 llsol.s6_addr16[1] = htons(ifp->if_index);
1373 llsol.s6_addr32[1] = 0;
1374 llsol.s6_addr32[2] = htonl(1);
1375 llsol.s6_addr8[12] = 0xff;
1376
1377 ifnet_lock_shared(ifp);
1378 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1379 ifnet_lock_done(ifp);
1380 if (in6m)
1381 in6_delmulti(in6m, 0);
1382 }
1383 nd6_inuse--;
1384 ln->ln_next->ln_prev = ln->ln_prev;
1385 ln->ln_prev->ln_next = ln->ln_next;
1386 ln->ln_prev = NULL;
1387 rt->rt_llinfo = 0;
1388 rt->rt_flags &= ~RTF_LLINFO;
1389 if (ln->ln_hold)
1390 m_freem(ln->ln_hold);
1391 ln->ln_hold = NULL;
1392 R_Free((caddr_t)ln);
1393 }
1394 }
1395
1396 int
1397 nd6_ioctl(
1398 u_long cmd,
1399 caddr_t data,
1400 struct ifnet *ifp)
1401 {
1402 struct in6_drlist *drl = (struct in6_drlist *)data;
1403 struct in6_prlist *prl = (struct in6_prlist *)data;
1404 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1405 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1406 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1407 struct nd_defrouter *dr, any;
1408 struct nd_prefix *pr;
1409 struct rtentry *rt;
1410 int i = 0, error = 0;
1411
1412 switch (cmd) {
1413 case SIOCGDRLST_IN6:
1414 /*
1415 * obsolete API, use sysctl under net.inet6.icmp6
1416 */
1417 lck_mtx_lock(nd6_mutex);
1418 bzero(drl, sizeof(*drl));
1419 dr = TAILQ_FIRST(&nd_defrouter);
1420 while (dr && i < DRLSTSIZ) {
1421 drl->defrouter[i].rtaddr = dr->rtaddr;
1422 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1423 /* XXX: need to this hack for KAME stack */
1424 drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1425 } else
1426 log(LOG_ERR,
1427 "default router list contains a "
1428 "non-linklocal address(%s)\n",
1429 ip6_sprintf(&drl->defrouter[i].rtaddr));
1430
1431 drl->defrouter[i].flags = dr->flags;
1432 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1433 drl->defrouter[i].expire = dr->expire;
1434 drl->defrouter[i].if_index = dr->ifp->if_index;
1435 i++;
1436 dr = TAILQ_NEXT(dr, dr_entry);
1437 }
1438 lck_mtx_unlock(nd6_mutex);
1439 break;
1440 case SIOCGPRLST_IN6:
1441 /*
1442 * obsolete API, use sysctl under net.inet6.icmp6
1443 */
1444 /*
1445 * XXX meaning of fields, especialy "raflags", is very
1446 * differnet between RA prefix list and RR/static prefix list.
1447 * how about separating ioctls into two?
1448 */
1449 bzero(prl, sizeof(*prl));
1450 lck_mtx_lock(nd6_mutex);
1451 pr = nd_prefix.lh_first;
1452 while (pr && i < PRLSTSIZ) {
1453 struct nd_pfxrouter *pfr;
1454 int j;
1455
1456 (void)in6_embedscope(&prl->prefix[i].prefix,
1457 &pr->ndpr_prefix, NULL, NULL);
1458 prl->prefix[i].raflags = pr->ndpr_raf;
1459 prl->prefix[i].prefixlen = pr->ndpr_plen;
1460 prl->prefix[i].vltime = pr->ndpr_vltime;
1461 prl->prefix[i].pltime = pr->ndpr_pltime;
1462 prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1463 prl->prefix[i].expire = pr->ndpr_expire;
1464
1465 pfr = pr->ndpr_advrtrs.lh_first;
1466 j = 0;
1467 while (pfr) {
1468 if (j < DRLSTSIZ) {
1469 #define RTRADDR prl->prefix[i].advrtr[j]
1470 RTRADDR = pfr->router->rtaddr;
1471 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1472 /* XXX: hack for KAME */
1473 RTRADDR.s6_addr16[1] = 0;
1474 } else
1475 log(LOG_ERR,
1476 "a router(%s) advertises "
1477 "a prefix with "
1478 "non-link local address\n",
1479 ip6_sprintf(&RTRADDR));
1480 #undef RTRADDR
1481 }
1482 j++;
1483 pfr = pfr->pfr_next;
1484 }
1485 prl->prefix[i].advrtrs = j;
1486 prl->prefix[i].origin = PR_ORIG_RA;
1487
1488 i++;
1489 pr = pr->ndpr_next;
1490 }
1491 {
1492 struct rr_prefix *rpp;
1493
1494 for (rpp = LIST_FIRST(&rr_prefix); rpp;
1495 rpp = LIST_NEXT(rpp, rp_entry)) {
1496 if (i >= PRLSTSIZ)
1497 break;
1498 (void)in6_embedscope(&prl->prefix[i].prefix,
1499 &pr->ndpr_prefix, NULL, NULL);
1500 prl->prefix[i].raflags = rpp->rp_raf;
1501 prl->prefix[i].prefixlen = rpp->rp_plen;
1502 prl->prefix[i].vltime = rpp->rp_vltime;
1503 prl->prefix[i].pltime = rpp->rp_pltime;
1504 prl->prefix[i].if_index = rpp->rp_ifp->if_index;
1505 prl->prefix[i].expire = rpp->rp_expire;
1506 prl->prefix[i].advrtrs = 0;
1507 prl->prefix[i].origin = rpp->rp_origin;
1508 i++;
1509 }
1510 }
1511 lck_mtx_unlock(nd6_mutex);
1512 break;
1513 case OSIOCGIFINFO_IN6:
1514 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1515 error = EINVAL;
1516 break;
1517 }
1518 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
1519 ndi->ndi.maxmtu = nd_ifinfo[ifp->if_index].maxmtu;
1520 ndi->ndi.basereachable =
1521 nd_ifinfo[ifp->if_index].basereachable;
1522 ndi->ndi.reachable = nd_ifinfo[ifp->if_index].reachable;
1523 ndi->ndi.retrans = nd_ifinfo[ifp->if_index].retrans;
1524 ndi->ndi.flags = nd_ifinfo[ifp->if_index].flags;
1525 ndi->ndi.recalctm = nd_ifinfo[ifp->if_index].recalctm;
1526 ndi->ndi.chlim = nd_ifinfo[ifp->if_index].chlim;
1527 ndi->ndi.receivedra = nd_ifinfo[ifp->if_index].receivedra;
1528 break;
1529 case SIOCGIFINFO_IN6:
1530 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1531 error = EINVAL;
1532 break;
1533 }
1534 ndi->ndi = nd_ifinfo[ifp->if_index];
1535 break;
1536 case SIOCSIFINFO_FLAGS:
1537 /* XXX: almost all other fields of ndi->ndi is unused */
1538 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1539 error = EINVAL;
1540 break;
1541 }
1542 nd_ifinfo[ifp->if_index].flags = ndi->ndi.flags;
1543 break;
1544 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1545 /* flush default router list */
1546 /*
1547 * xxx sumikawa: should not delete route if default
1548 * route equals to the top of default router list
1549 */
1550 bzero(&any, sizeof(any));
1551 lck_mtx_lock(nd6_mutex);
1552 defrouter_delreq(&any, 1);
1553 defrouter_select();
1554 lck_mtx_unlock(nd6_mutex);
1555 /* xxx sumikawa: flush prefix list */
1556 break;
1557 case SIOCSPFXFLUSH_IN6:
1558 {
1559 /* flush all the prefix advertised by routers */
1560 struct nd_prefix *next;
1561 lck_mtx_lock(nd6_mutex);
1562
1563 for (pr = nd_prefix.lh_first; pr; pr = next) {
1564 struct in6_ifaddr *ia, *ia_next;
1565
1566 next = pr->ndpr_next;
1567
1568 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1569 continue; /* XXX */
1570
1571 /* do we really have to remove addresses as well? */
1572 for (ia = in6_ifaddrs; ia; ia = ia_next) {
1573 /* ia might be removed. keep the next ptr. */
1574 ia_next = ia->ia_next;
1575
1576 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1577 continue;
1578
1579 if (ia->ia6_ndpr == pr)
1580 in6_purgeaddr(&ia->ia_ifa, 1);
1581 }
1582 prelist_remove(pr, 1);
1583 }
1584 lck_mtx_unlock(nd6_mutex);
1585 break;
1586 }
1587 case SIOCSRTRFLUSH_IN6:
1588 {
1589 /* flush all the default routers */
1590 struct nd_defrouter *next;
1591
1592 lck_mtx_lock(nd6_mutex);
1593 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1594 /*
1595 * The first entry of the list may be stored in
1596 * the routing table, so we'll delete it later.
1597 */
1598 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1599 next = TAILQ_NEXT(dr, dr_entry);
1600 defrtrlist_del(dr, 1);
1601 }
1602 defrtrlist_del(TAILQ_FIRST(&nd_defrouter), 1);
1603 }
1604 lck_mtx_unlock(nd6_mutex);
1605 break;
1606 }
1607 case SIOCGNBRINFO_IN6:
1608 {
1609 struct llinfo_nd6 *ln;
1610 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1611
1612 /*
1613 * XXX: KAME specific hack for scoped addresses
1614 * XXXX: for other scopes than link-local?
1615 */
1616 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1617 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1618 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1619
1620 if (*idp == 0)
1621 *idp = htons(ifp->if_index);
1622 }
1623
1624 if ((rt = nd6_lookup(&nb_addr, 0, ifp, 0)) == NULL) {
1625 error = EINVAL;
1626 break;
1627 }
1628 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1629 nbi->state = ln->ln_state;
1630 nbi->asked = ln->ln_asked;
1631 nbi->isrouter = ln->ln_router;
1632 nbi->expire = ln->ln_expire;
1633
1634 break;
1635 }
1636 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1637 ndif->ifindex = nd6_defifindex;
1638 break;
1639 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1640 return(nd6_setdefaultiface(ndif->ifindex));
1641 break;
1642 }
1643 return(error);
1644 }
1645
1646 /*
1647 * Create neighbor cache entry and cache link-layer address,
1648 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1649 */
1650 struct rtentry *
1651 nd6_cache_lladdr(
1652 struct ifnet *ifp,
1653 struct in6_addr *from,
1654 char *lladdr,
1655 __unused int lladdrlen,
1656 int type, /* ICMP6 type */
1657 int code) /* type dependent information */
1658 {
1659 struct rtentry *rt = NULL;
1660 struct llinfo_nd6 *ln = NULL;
1661 int is_newentry;
1662 struct sockaddr_dl *sdl = NULL;
1663 int do_update;
1664 int olladdr;
1665 int llchange;
1666 int newstate = 0;
1667 struct timeval timenow;
1668
1669 if (!ifp)
1670 panic("ifp == NULL in nd6_cache_lladdr");
1671 if (!from)
1672 panic("from == NULL in nd6_cache_lladdr");
1673
1674 /* nothing must be updated for unspecified address */
1675 if (IN6_IS_ADDR_UNSPECIFIED(from))
1676 return NULL;
1677
1678 /*
1679 * Validation about ifp->if_addrlen and lladdrlen must be done in
1680 * the caller.
1681 *
1682 * XXX If the link does not have link-layer adderss, what should
1683 * we do? (ifp->if_addrlen == 0)
1684 * Spec says nothing in sections for RA, RS and NA. There's small
1685 * description on it in NS section (RFC 2461 7.2.3).
1686 */
1687 getmicrotime(&timenow);
1688
1689 lck_mtx_lock(rt_mtx);
1690 rt = nd6_lookup(from, 0, ifp, 1);
1691 if (!rt) {
1692 #if 0
1693 /* nothing must be done if there's no lladdr */
1694 if (!lladdr || !lladdrlen)
1695 return NULL;
1696 #endif
1697
1698 rt = nd6_lookup(from, 1, ifp, 1);
1699 is_newentry = 1;
1700 } else {
1701 /* do nothing if static ndp is set */
1702 if (rt->rt_flags & RTF_STATIC) {
1703 lck_mtx_unlock(rt_mtx);
1704 return NULL;
1705 }
1706 is_newentry = 0;
1707 }
1708
1709 lck_mtx_unlock(rt_mtx);
1710
1711 if (!rt)
1712 return NULL;
1713 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1714 fail:
1715 (void)nd6_free(rt);
1716 return NULL;
1717 }
1718 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1719 if (!ln)
1720 goto fail;
1721 if (!rt->rt_gateway)
1722 goto fail;
1723 if (rt->rt_gateway->sa_family != AF_LINK)
1724 goto fail;
1725 sdl = SDL(rt->rt_gateway);
1726
1727 olladdr = (sdl->sdl_alen) ? 1 : 0;
1728 if (olladdr && lladdr) {
1729 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1730 llchange = 1;
1731 else
1732 llchange = 0;
1733 } else
1734 llchange = 0;
1735
1736 /*
1737 * newentry olladdr lladdr llchange (*=record)
1738 * 0 n n -- (1)
1739 * 0 y n -- (2)
1740 * 0 n y -- (3) * STALE
1741 * 0 y y n (4) *
1742 * 0 y y y (5) * STALE
1743 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1744 * 1 -- y -- (7) * STALE
1745 */
1746
1747 if (lladdr) { /* (3-5) and (7) */
1748 /*
1749 * Record source link-layer address
1750 * XXX is it dependent to ifp->if_type?
1751 */
1752 sdl->sdl_alen = ifp->if_addrlen;
1753 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1754 }
1755
1756 if (!is_newentry) {
1757 if ((!olladdr && lladdr) /* (3) */
1758 || (olladdr && lladdr && llchange)) { /* (5) */
1759 do_update = 1;
1760 newstate = ND6_LLINFO_STALE;
1761 } else /* (1-2,4) */
1762 do_update = 0;
1763 } else {
1764 do_update = 1;
1765 if (!lladdr) /* (6) */
1766 newstate = ND6_LLINFO_NOSTATE;
1767 else /* (7) */
1768 newstate = ND6_LLINFO_STALE;
1769 }
1770
1771 if (do_update) {
1772 /*
1773 * Update the state of the neighbor cache.
1774 */
1775 ln->ln_state = newstate;
1776
1777 if (ln->ln_state == ND6_LLINFO_STALE) {
1778 /*
1779 * XXX: since nd6_output() below will cause
1780 * state tansition to DELAY and reset the timer,
1781 * we must set the timer now, although it is actually
1782 * meaningless.
1783 */
1784 ln->ln_expire = timenow.tv_sec + nd6_gctimer;
1785
1786 if (ln->ln_hold) {
1787 /*
1788 * we assume ifp is not a p2p here, so just
1789 * set the 2nd argument as the 1st one.
1790 */
1791 nd6_output(ifp, ifp, ln->ln_hold,
1792 (struct sockaddr_in6 *)rt_key(rt),
1793 rt, 0);
1794 ln->ln_hold = NULL;
1795 }
1796 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1797 /* probe right away */
1798 ln->ln_expire = timenow.tv_sec;
1799 }
1800 }
1801
1802 /*
1803 * ICMP6 type dependent behavior.
1804 *
1805 * NS: clear IsRouter if new entry
1806 * RS: clear IsRouter
1807 * RA: set IsRouter if there's lladdr
1808 * redir: clear IsRouter if new entry
1809 *
1810 * RA case, (1):
1811 * The spec says that we must set IsRouter in the following cases:
1812 * - If lladdr exist, set IsRouter. This means (1-5).
1813 * - If it is old entry (!newentry), set IsRouter. This means (7).
1814 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1815 * A quetion arises for (1) case. (1) case has no lladdr in the
1816 * neighbor cache, this is similar to (6).
1817 * This case is rare but we figured that we MUST NOT set IsRouter.
1818 *
1819 * newentry olladdr lladdr llchange NS RS RA redir
1820 * D R
1821 * 0 n n -- (1) c ? s
1822 * 0 y n -- (2) c s s
1823 * 0 n y -- (3) c s s
1824 * 0 y y n (4) c s s
1825 * 0 y y y (5) c s s
1826 * 1 -- n -- (6) c c c s
1827 * 1 -- y -- (7) c c s c s
1828 *
1829 * (c=clear s=set)
1830 */
1831 switch (type & 0xff) {
1832 case ND_NEIGHBOR_SOLICIT:
1833 /*
1834 * New entry must have is_router flag cleared.
1835 */
1836 if (is_newentry) /* (6-7) */
1837 ln->ln_router = 0;
1838 break;
1839 case ND_REDIRECT:
1840 /*
1841 * If the icmp is a redirect to a better router, always set the
1842 * is_router flag. Otherwise, if the entry is newly created,
1843 * clear the flag. [RFC 2461, sec 8.3]
1844 */
1845 if (code == ND_REDIRECT_ROUTER)
1846 ln->ln_router = 1;
1847 else if (is_newentry) /* (6-7) */
1848 ln->ln_router = 0;
1849 break;
1850 case ND_ROUTER_SOLICIT:
1851 /*
1852 * is_router flag must always be cleared.
1853 */
1854 ln->ln_router = 0;
1855 break;
1856 case ND_ROUTER_ADVERT:
1857 /*
1858 * Mark an entry with lladdr as a router.
1859 */
1860 if ((!is_newentry && (olladdr || lladdr)) /* (2-5) */
1861 || (is_newentry && lladdr)) { /* (7) */
1862 ln->ln_router = 1;
1863 }
1864 break;
1865 }
1866
1867 /*
1868 * When the link-layer address of a router changes, select the
1869 * best router again. In particular, when the neighbor entry is newly
1870 * created, it might affect the selection policy.
1871 * Question: can we restrict the first condition to the "is_newentry"
1872 * case?
1873 * XXX: when we hear an RA from a new router with the link-layer
1874 * address option, defrouter_select() is called twice, since
1875 * defrtrlist_update called the function as well. However, I believe
1876 * we can compromise the overhead, since it only happens the first
1877 * time.
1878 * XXX: although defrouter_select() should not have a bad effect
1879 * for those are not autoconfigured hosts, we explicitly avoid such
1880 * cases for safety.
1881 */
1882 if (do_update && ln->ln_router && !ip6_forwarding && (ip6_accept_rtadv || (ifp->if_eflags & IFEF_ACCEPT_RTADVD))) {
1883 lck_mtx_lock(nd6_mutex);
1884 defrouter_select();
1885 lck_mtx_unlock(nd6_mutex);
1886 }
1887
1888 return rt;
1889 }
1890
1891 static void
1892 nd6_slowtimo(
1893 __unused void *ignored_arg)
1894 {
1895 int i;
1896 struct nd_ifinfo *nd6if;
1897
1898 lck_mtx_lock(nd6_mutex);
1899 for (i = 1; i < if_index + 1; i++) {
1900 if (!nd_ifinfo || i >= nd_ifinfo_indexlim)
1901 continue;
1902 nd6if = &nd_ifinfo[i];
1903 if (nd6if->basereachable && /* already initialized */
1904 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1905 /*
1906 * Since reachable time rarely changes by router
1907 * advertisements, we SHOULD insure that a new random
1908 * value gets recomputed at least once every few hours.
1909 * (RFC 2461, 6.3.4)
1910 */
1911 nd6if->recalctm = nd6_recalc_reachtm_interval;
1912 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1913 }
1914 }
1915 lck_mtx_unlock(nd6_mutex);
1916 timeout(nd6_slowtimo, (caddr_t)0, ND6_SLOWTIMER_INTERVAL * hz);
1917 }
1918
1919
1920 #define senderr(e) { error = (e); goto bad;}
1921 int
1922 nd6_output(
1923 struct ifnet *ifp,
1924 struct ifnet *origifp,
1925 struct mbuf *m0,
1926 struct sockaddr_in6 *dst,
1927 struct rtentry *rt0,
1928 int locked)
1929 {
1930 struct mbuf *m = m0;
1931 struct rtentry *rt = rt0;
1932 struct sockaddr_in6 *gw6 = NULL;
1933 struct llinfo_nd6 *ln = NULL;
1934 int error = 0;
1935 struct timeval timenow;
1936
1937 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1938 goto sendpkt;
1939
1940 if (nd6_need_cache(ifp) == 0)
1941 goto sendpkt;
1942
1943 /*
1944 * next hop determination. This routine is derived from ether_outpout.
1945 */
1946 lck_mtx_lock(rt_mtx);
1947 if (rt) {
1948 if ((rt->rt_flags & RTF_UP) == 0) {
1949 if ((rt0 = rt = rtalloc1_locked((struct sockaddr *)dst, 1, 0UL)) !=
1950 NULL)
1951 {
1952 rtunref(rt);
1953 if (rt->rt_ifp != ifp) {
1954 /* XXX: loop care? */
1955 lck_mtx_unlock(rt_mtx);
1956 return nd6_output(ifp, origifp, m0,
1957 dst, rt, locked);
1958 }
1959 } else {
1960 lck_mtx_unlock(rt_mtx);
1961 senderr(EHOSTUNREACH);
1962 }
1963 }
1964
1965 if (rt->rt_flags & RTF_GATEWAY) {
1966 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1967
1968 /*
1969 * We skip link-layer address resolution and NUD
1970 * if the gateway is not a neighbor from ND point
1971 * of view, regardless of the value of nd_ifinfo.flags.
1972 * The second condition is a bit tricky; we skip
1973 * if the gateway is our own address, which is
1974 * sometimes used to install a route to a p2p link.
1975 */
1976 if (!nd6_is_addr_neighbor(gw6, ifp, 1) ||
1977 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1978 /*
1979 * We allow this kind of tricky route only
1980 * when the outgoing interface is p2p.
1981 * XXX: we may need a more generic rule here.
1982 */
1983 lck_mtx_unlock(rt_mtx);
1984 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1985 senderr(EHOSTUNREACH);
1986
1987 goto sendpkt;
1988 }
1989
1990 if (rt->rt_gwroute == 0)
1991 goto lookup;
1992 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1993 rtfree_locked(rt); rt = rt0;
1994 lookup: rt->rt_gwroute = rtalloc1_locked(rt->rt_gateway, 1, 0UL);
1995 if ((rt = rt->rt_gwroute) == 0) {
1996 lck_mtx_unlock(rt_mtx);
1997 senderr(EHOSTUNREACH);
1998 }
1999 }
2000 }
2001 }
2002
2003 /*
2004 * Address resolution or Neighbor Unreachability Detection
2005 * for the next hop.
2006 * At this point, the destination of the packet must be a unicast
2007 * or an anycast address(i.e. not a multicast).
2008 */
2009
2010 /* Look up the neighbor cache for the nexthop */
2011 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
2012 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2013 else {
2014 /*
2015 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2016 * the condition below is not very efficient. But we believe
2017 * it is tolerable, because this should be a rare case.
2018 */
2019 if (nd6_is_addr_neighbor(dst, ifp, 1) &&
2020 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp, 1)) != NULL)
2021 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2022 }
2023 lck_mtx_unlock(rt_mtx);
2024 if (!ln || !rt) {
2025 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2026 !(nd_ifinfo[ifp->if_index].flags & ND6_IFF_PERFORMNUD)) {
2027 log(LOG_DEBUG,
2028 "nd6_output: can't allocate llinfo for %s "
2029 "(ln=%p, rt=%p)\n",
2030 ip6_sprintf(&dst->sin6_addr), ln, rt);
2031 senderr(EIO); /* XXX: good error? */
2032 }
2033
2034 goto sendpkt; /* send anyway */
2035 }
2036
2037 getmicrotime(&timenow);
2038
2039 /* We don't have to do link-layer address resolution on a p2p link. */
2040 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2041 ln->ln_state < ND6_LLINFO_REACHABLE) {
2042 ln->ln_state = ND6_LLINFO_STALE;
2043 ln->ln_expire = timenow.tv_sec + nd6_gctimer;
2044 }
2045
2046 /*
2047 * The first time we send a packet to a neighbor whose entry is
2048 * STALE, we have to change the state to DELAY and a sets a timer to
2049 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2050 * neighbor unreachability detection on expiration.
2051 * (RFC 2461 7.3.3)
2052 */
2053 if (ln->ln_state == ND6_LLINFO_STALE) {
2054 ln->ln_asked = 0;
2055 ln->ln_state = ND6_LLINFO_DELAY;
2056 ln->ln_expire = timenow.tv_sec + nd6_delay;
2057 }
2058
2059 /*
2060 * If the neighbor cache entry has a state other than INCOMPLETE
2061 * (i.e. its link-layer address is already resolved), just
2062 * send the packet.
2063 */
2064 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2065 goto sendpkt;
2066
2067 /*
2068 * There is a neighbor cache entry, but no ethernet address
2069 * response yet. Replace the held mbuf (if any) with this
2070 * latest one.
2071 *
2072 * This code conforms to the rate-limiting rule described in Section
2073 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
2074 * an NS below.
2075 */
2076 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2077 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2078 if (ln->ln_hold)
2079 m_freem(ln->ln_hold);
2080 ln->ln_hold = m;
2081 if (ln->ln_expire) {
2082 if (ln->ln_asked < nd6_mmaxtries &&
2083 ln->ln_expire < timenow.tv_sec) {
2084 ln->ln_asked++;
2085 ln->ln_expire = timenow.tv_sec +
2086 nd_ifinfo[ifp->if_index].retrans / 1000;
2087 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0, locked);
2088 }
2089 }
2090 return(0);
2091
2092 sendpkt:
2093 #ifdef __APPLE__
2094
2095 /* Make sure the HW checksum flags are cleaned before sending the packet */
2096
2097 m->m_pkthdr.csum_data = 0;
2098 m->m_pkthdr.csum_flags = 0;
2099
2100 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2101 m->m_pkthdr.rcvif = origifp; /* forwarding rules require the original scope_id */
2102 if (locked)
2103 lck_mtx_unlock(ip6_mutex);
2104 error = dlil_output(origifp, PF_INET6, m, (caddr_t)rt, (struct sockaddr *)dst, 0);
2105 if (locked)
2106 lck_mtx_lock(ip6_mutex);
2107 return error;
2108 } else {
2109 /* Do not allow loopback address to wind up on a wire */
2110 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2111
2112 if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
2113 IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst))) {
2114 ip6stat.ip6s_badscope++;
2115 /*
2116 * Do not simply drop the packet just like a firewall -- we want the
2117 * the application to feel the pain.
2118 * Return ENETUNREACH like ip6_output does in some similar cases.
2119 * This can startle the otherwise clueless process that specifies
2120 * loopback as the source address.
2121 */
2122 error = ENETUNREACH;
2123 goto bad;
2124 }
2125 }
2126
2127 m->m_pkthdr.rcvif = 0;
2128 if (locked)
2129 lck_mtx_unlock(ip6_mutex);
2130 error = dlil_output(ifp, PF_INET6, m, (caddr_t)rt, (struct sockaddr *)dst, 0);
2131 if (locked)
2132 lck_mtx_lock(ip6_mutex);
2133 return(error);
2134 #else
2135 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2136 return((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2137 rt));
2138 }
2139 return((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2140 #endif
2141
2142 bad:
2143 if (m)
2144 m_freem(m);
2145 return (error);
2146 }
2147 #undef senderr
2148
2149 int
2150 nd6_need_cache(
2151 struct ifnet *ifp)
2152 {
2153 /*
2154 * XXX: we currently do not make neighbor cache on any interface
2155 * other than ARCnet, Ethernet, FDDI and GIF.
2156 *
2157 * RFC2893 says:
2158 * - unidirectional tunnels needs no ND
2159 */
2160 switch (ifp->if_type) {
2161 case IFT_ARCNET:
2162 case IFT_ETHER:
2163 case IFT_FDDI:
2164 case IFT_IEEE1394:
2165 case IFT_L2VLAN:
2166 case IFT_IEEE8023ADLAG:
2167 #if IFT_IEEE80211
2168 case IFT_IEEE80211:
2169 #endif
2170 case IFT_GIF: /* XXX need more cases? */
2171 return(1);
2172 default:
2173 return(0);
2174 }
2175 }
2176
2177 int
2178 nd6_storelladdr(
2179 struct ifnet *ifp,
2180 struct rtentry *rt,
2181 struct mbuf *m,
2182 struct sockaddr *dst,
2183 u_char *desten)
2184 {
2185 int i;
2186 struct sockaddr_dl *sdl;
2187
2188 if (m->m_flags & M_MCAST) {
2189 switch (ifp->if_type) {
2190 case IFT_ETHER:
2191 case IFT_FDDI:
2192 case IFT_L2VLAN:
2193 case IFT_IEEE8023ADLAG:
2194 #if IFT_IEEE80211
2195 case IFT_IEEE80211:
2196 #endif
2197 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2198 desten);
2199 return(1);
2200 case IFT_IEEE1394:
2201 for (i = 0; i < ifp->if_addrlen; i++)
2202 desten[i] = ~0;
2203 return(1);
2204 case IFT_ARCNET:
2205 *desten = 0;
2206 return(1);
2207 default:
2208 return(0); /* caller will free mbuf */
2209 }
2210 }
2211
2212 if (rt == NULL) {
2213 /* this could happen, if we could not allocate memory */
2214 return(0); /* caller will free mbuf */
2215 }
2216 if (rt->rt_gateway->sa_family != AF_LINK) {
2217 printf("nd6_storelladdr: something odd happens\n");
2218 return(0); /* caller will free mbuf */
2219 }
2220 sdl = SDL(rt->rt_gateway);
2221 if (sdl->sdl_alen == 0) {
2222 /* this should be impossible, but we bark here for debugging */
2223 printf("nd6_storelladdr: sdl_alen == 0\n");
2224 return(0); /* caller will free mbuf */
2225 }
2226
2227 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2228 return(1);
2229 }
2230
2231 extern errno_t arp_route_to_gateway_route(const struct sockaddr *net_dest,
2232 route_t hint, route_t *out_route);
2233
2234 errno_t
2235 nd6_lookup_ipv6(
2236 ifnet_t ifp,
2237 const struct sockaddr_in6 *ip6_dest,
2238 struct sockaddr_dl *ll_dest,
2239 size_t ll_dest_len,
2240 route_t hint,
2241 mbuf_t packet)
2242 {
2243 route_t route = hint;
2244 errno_t result = 0;
2245 struct sockaddr_dl *sdl = NULL;
2246 size_t copy_len;
2247
2248 if (ip6_dest->sin6_family != AF_INET6)
2249 return EAFNOSUPPORT;
2250
2251 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2252 return ENETDOWN;
2253
2254 if (hint) {
2255 result = arp_route_to_gateway_route((const struct sockaddr*)ip6_dest, hint, &route);
2256 if (result != 0)
2257 return result;
2258 }
2259
2260 if ((packet->m_flags & M_MCAST) != 0) {
2261 return dlil_resolve_multi(ifp, (const struct sockaddr*)ip6_dest,
2262 (struct sockaddr *)ll_dest, ll_dest_len);
2263 }
2264
2265 if (route == NULL) {
2266 /* this could happen, if we could not allocate memory */
2267 return ENOBUFS;
2268 }
2269
2270 lck_mtx_lock(rt_mtx);
2271
2272 if (route->rt_gateway->sa_family != AF_LINK) {
2273 printf("nd6_lookup_ipv6: gateway address not AF_LINK\n");
2274 result = EADDRNOTAVAIL;
2275 goto done;
2276 }
2277
2278 sdl = SDL(route->rt_gateway);
2279 if (sdl->sdl_alen == 0) {
2280 /* this should be impossible, but we bark here for debugging */
2281 printf("nd6_storelladdr: sdl_alen == 0\n");
2282 result = EHOSTUNREACH;
2283 }
2284
2285 copy_len = sdl->sdl_len <= ll_dest_len ? sdl->sdl_len : ll_dest_len;
2286 bcopy(sdl, ll_dest, copy_len);
2287
2288 done:
2289 lck_mtx_unlock(rt_mtx);
2290 return result;
2291 }
2292
2293 SYSCTL_DECL(_net_inet6_icmp6);
2294
2295 static int
2296 nd6_sysctl_drlist SYSCTL_HANDLER_ARGS
2297 {
2298 #pragma unused(oidp, arg1, arg2)
2299 int error;
2300 char buf[1024];
2301 struct in6_defrouter *d, *de;
2302 struct nd_defrouter *dr;
2303
2304 if (req->newptr)
2305 return EPERM;
2306 error = 0;
2307
2308 lck_mtx_lock(nd6_mutex);
2309 for (dr = TAILQ_FIRST(&nd_defrouter);
2310 dr;
2311 dr = TAILQ_NEXT(dr, dr_entry)) {
2312 d = (struct in6_defrouter *)buf;
2313 de = (struct in6_defrouter *)(buf + sizeof(buf));
2314
2315 if (d + 1 <= de) {
2316 bzero(d, sizeof(*d));
2317 d->rtaddr.sin6_family = AF_INET6;
2318 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2319 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2320 dr->ifp) != 0)
2321 log(LOG_ERR,
2322 "scope error in "
2323 "default router list (%s)\n",
2324 ip6_sprintf(&dr->rtaddr));
2325 d->flags = dr->flags;
2326 d->rtlifetime = dr->rtlifetime;
2327 d->expire = dr->expire;
2328 d->if_index = dr->ifp->if_index;
2329 } else
2330 panic("buffer too short");
2331
2332 error = SYSCTL_OUT(req, buf, sizeof(*d));
2333 if (error)
2334 break;
2335 }
2336 lck_mtx_unlock(nd6_mutex);
2337 return error;
2338 }
2339
2340 static int
2341 nd6_sysctl_prlist SYSCTL_HANDLER_ARGS
2342 {
2343 #pragma unused(oidp, arg1, arg2)
2344 int error;
2345 char buf[1024];
2346 struct in6_prefix *p, *pe;
2347 struct nd_prefix *pr;
2348
2349 if (req->newptr)
2350 return EPERM;
2351 error = 0;
2352
2353 lck_mtx_lock(nd6_mutex);
2354
2355 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2356 u_short advrtrs = 0;
2357 size_t advance;
2358 struct sockaddr_in6 *sin6, *s6;
2359 struct nd_pfxrouter *pfr;
2360
2361 p = (struct in6_prefix *)buf;
2362 pe = (struct in6_prefix *)(buf + sizeof(buf));
2363
2364 if (p + 1 <= pe) {
2365 bzero(p, sizeof(*p));
2366 sin6 = (struct sockaddr_in6 *)(p + 1);
2367
2368 p->prefix = pr->ndpr_prefix;
2369 if (in6_recoverscope(&p->prefix,
2370 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2371 log(LOG_ERR,
2372 "scope error in prefix list (%s)\n",
2373 ip6_sprintf(&p->prefix.sin6_addr));
2374 p->raflags = pr->ndpr_raf;
2375 p->prefixlen = pr->ndpr_plen;
2376 p->vltime = pr->ndpr_vltime;
2377 p->pltime = pr->ndpr_pltime;
2378 p->if_index = pr->ndpr_ifp->if_index;
2379 p->expire = pr->ndpr_expire;
2380 p->refcnt = pr->ndpr_refcnt;
2381 p->flags = pr->ndpr_stateflags;
2382 p->origin = PR_ORIG_RA;
2383 advrtrs = 0;
2384 for (pfr = pr->ndpr_advrtrs.lh_first;
2385 pfr;
2386 pfr = pfr->pfr_next) {
2387 if ((void *)&sin6[advrtrs + 1] >
2388 (void *)pe) {
2389 advrtrs++;
2390 continue;
2391 }
2392 s6 = &sin6[advrtrs];
2393 bzero(s6, sizeof(*s6));
2394 s6->sin6_family = AF_INET6;
2395 s6->sin6_len = sizeof(*sin6);
2396 if (in6_recoverscope(s6,
2397 &pfr->router->rtaddr,
2398 pfr->router->ifp) != 0)
2399 log(LOG_ERR,
2400 "scope error in "
2401 "prefix list (%s)\n",
2402 ip6_sprintf(&pfr->router->rtaddr));
2403 advrtrs++;
2404 }
2405 p->advrtrs = advrtrs;
2406 } else
2407 panic("buffer too short");
2408
2409 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2410 error = SYSCTL_OUT(req, buf, advance);
2411 if (error)
2412 break;
2413 }
2414 lck_mtx_unlock(nd6_mutex);
2415 return error;
2416 }
2417 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2418 CTLFLAG_RD, 0, 0, nd6_sysctl_drlist, "S,in6_defrouter","");
2419 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2420 CTLFLAG_RD, 0, 0, nd6_sysctl_prlist, "S,in6_defrouter","");
2421
mirror of XNU