]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet6/nd6_rtr.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-2782.1.97.tar.gz
[apple/xnu.git] / bsd / netinet6 / nd6_rtr.c
1 /*
2 * Copyright (c) 2003-2014 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /*
30 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. Neither the name of the project nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
44 *
45 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55 * SUCH DAMAGE.
56 */
57
58
59 #include <sys/param.h>
60 #include <sys/systm.h>
61 #include <sys/malloc.h>
62 #include <sys/mbuf.h>
63 #include <sys/socket.h>
64 #include <sys/sockio.h>
65 #include <sys/time.h>
66 #include <sys/kernel.h>
67 #include <sys/errno.h>
68 #include <sys/syslog.h>
69 #include <sys/queue.h>
70 #include <sys/mcache.h>
71 #include <sys/protosw.h>
72
73 #include <dev/random/randomdev.h>
74
75 #include <kern/locks.h>
76 #include <kern/zalloc.h>
77 #include <machine/machine_routines.h>
78
79 #include <net/if.h>
80 #include <net/if_types.h>
81 #include <net/if_dl.h>
82 #include <net/route.h>
83 #include <net/radix.h>
84
85 #include <netinet/in.h>
86 #include <netinet6/in6_var.h>
87 #include <netinet6/in6_ifattach.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/ip6_var.h>
90 #include <netinet6/nd6.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/scope6_var.h>
93
94 #include <net/net_osdep.h>
95
96 static void defrouter_addreq(struct nd_defrouter *, boolean_t);
97 static void defrouter_delreq(struct nd_defrouter *);
98 static struct nd_defrouter *defrtrlist_update_common(struct nd_defrouter *,
99 boolean_t);
100 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
101
102 static struct in6_ifaddr *in6_pfx_newpersistaddr(struct nd_prefix *, int,
103 int *);
104 static void defrtrlist_sync(struct ifnet *);
105
106 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
107 struct nd_defrouter *);
108 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
109 static void pfxrtr_del(struct nd_pfxrouter *, struct nd_prefix *);
110 static struct nd_pfxrouter *find_pfxlist_reachable_router(struct nd_prefix *);
111 static void nd6_rtmsg(int, struct rtentry *);
112
113 static int nd6_prefix_onlink_common(struct nd_prefix *, boolean_t,
114 unsigned int);
115 static struct nd_prefix *nd6_prefix_equal_lookup(struct nd_prefix *, boolean_t);
116 static void nd6_prefix_sync(struct ifnet *);
117
118 static void in6_init_address_ltimes(struct nd_prefix *,
119 struct in6_addrlifetime *);
120
121 static int rt6_deleteroute(struct radix_node *, void *);
122
123 static struct nd_defrouter *nddr_alloc(int);
124 static void nddr_free(struct nd_defrouter *);
125 static void nddr_trace(struct nd_defrouter *, int);
126
127 static struct nd_prefix *ndpr_alloc(int);
128 static void ndpr_free(struct nd_prefix *);
129 static void ndpr_trace(struct nd_prefix *, int);
130
131 extern int nd6_recalc_reachtm_interval;
132
133 static struct ifnet *nd6_defifp;
134 int nd6_defifindex;
135 static unsigned int nd6_defrouter_genid;
136
137 int ip6_use_tempaddr = 1; /* use temp addr by default for testing now */
138
139 int nd6_accept_6to4 = 1;
140
141 int ip6_desync_factor;
142 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
143 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
144 /*
145 * shorter lifetimes for debugging purposes.
146 * u_int32_t ip6_temp_preferred_lifetime = 800;
147 * static u_int32_t ip6_temp_valid_lifetime = 1800;
148 */
149 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
150
151 extern lck_mtx_t *nd6_mutex;
152
153 /* Serialization variables for single thread access to nd_prefix */
154 static boolean_t nd_prefix_busy;
155 static void *nd_prefix_waitchan = &nd_prefix_busy;
156 static int nd_prefix_waiters = 0;
157
158 /* Serialization variables for single thread access to nd_defrouter */
159 static boolean_t nd_defrouter_busy;
160 static void *nd_defrouter_waitchan = &nd_defrouter_busy;
161 static int nd_defrouter_waiters = 0;
162
163 /* RTPREF_MEDIUM has to be 0! */
164 #define RTPREF_HIGH 1
165 #define RTPREF_MEDIUM 0
166 #define RTPREF_LOW (-1)
167 #define RTPREF_RESERVED (-2)
168 #define RTPREF_INVALID (-3) /* internal */
169
170 #define NDPR_TRACE_HIST_SIZE 32 /* size of trace history */
171
172 /* For gdb */
173 __private_extern__ unsigned int ndpr_trace_hist_size = NDPR_TRACE_HIST_SIZE;
174
175 struct nd_prefix_dbg {
176 struct nd_prefix ndpr_pr; /* nd_prefix */
177 u_int16_t ndpr_refhold_cnt; /* # of ref */
178 u_int16_t ndpr_refrele_cnt; /* # of rele */
179 /*
180 * Circular lists of ndpr_addref and ndpr_remref callers.
181 */
182 ctrace_t ndpr_refhold[NDPR_TRACE_HIST_SIZE];
183 ctrace_t ndpr_refrele[NDPR_TRACE_HIST_SIZE];
184 };
185
186 static unsigned int ndpr_debug; /* debug flags */
187 static unsigned int ndpr_size; /* size of zone element */
188 static struct zone *ndpr_zone; /* zone for nd_prefix */
189
190 #define NDPR_ZONE_MAX 64 /* maximum elements in zone */
191 #define NDPR_ZONE_NAME "nd6_prefix" /* zone name */
192
193 #define NDDR_TRACE_HIST_SIZE 32 /* size of trace history */
194
195 /* For gdb */
196 __private_extern__ unsigned int nddr_trace_hist_size = NDDR_TRACE_HIST_SIZE;
197
198 struct nd_defrouter_dbg {
199 struct nd_defrouter nddr_dr; /* nd_defrouter */
200 uint16_t nddr_refhold_cnt; /* # of ref */
201 uint16_t nddr_refrele_cnt; /* # of rele */
202 /*
203 * Circular lists of ndpr_addref and ndpr_remref callers.
204 */
205 ctrace_t nddr_refhold[NDDR_TRACE_HIST_SIZE];
206 ctrace_t nddr_refrele[NDDR_TRACE_HIST_SIZE];
207 };
208
209 static unsigned int nddr_debug; /* debug flags */
210 static unsigned int nddr_size; /* size of zone element */
211 static struct zone *nddr_zone; /* zone for nd_defrouter */
212
213 #define NDDR_ZONE_MAX 64 /* maximum elements in zone */
214 #define NDDR_ZONE_NAME "nd6_defrouter" /* zone name */
215
216 static unsigned int ndprtr_size; /* size of zone element */
217 static struct zone *ndprtr_zone; /* zone for nd_pfxrouter */
218
219 #define NDPRTR_ZONE_MAX 64 /* maximum elements in zone */
220 #define NDPRTR_ZONE_NAME "nd6_pfxrouter" /* zone name */
221
222 void
223 nd6_rtr_init(void)
224 {
225 PE_parse_boot_argn("ifa_debug", &ndpr_debug, sizeof (ndpr_debug));
226 PE_parse_boot_argn("ifa_debug", &nddr_debug, sizeof (nddr_debug));
227
228 ndpr_size = (ndpr_debug == 0) ? sizeof (struct nd_prefix) :
229 sizeof (struct nd_prefix_dbg);
230 ndpr_zone = zinit(ndpr_size, NDPR_ZONE_MAX * ndpr_size, 0,
231 NDPR_ZONE_NAME);
232 if (ndpr_zone == NULL) {
233 panic("%s: failed allocating %s", __func__, NDPR_ZONE_NAME);
234 /* NOTREACHED */
235 }
236 zone_change(ndpr_zone, Z_EXPAND, TRUE);
237 zone_change(ndpr_zone, Z_CALLERACCT, FALSE);
238
239 nddr_size = (nddr_debug == 0) ? sizeof (struct nd_defrouter) :
240 sizeof (struct nd_defrouter_dbg);
241 nddr_zone = zinit(nddr_size, NDDR_ZONE_MAX * nddr_size, 0,
242 NDDR_ZONE_NAME);
243 if (nddr_zone == NULL) {
244 panic("%s: failed allocating %s", __func__, NDDR_ZONE_NAME);
245 /* NOTREACHED */
246 }
247 zone_change(nddr_zone, Z_EXPAND, TRUE);
248 zone_change(nddr_zone, Z_CALLERACCT, FALSE);
249
250 ndprtr_size = sizeof (struct nd_pfxrouter);
251 ndprtr_zone = zinit(ndprtr_size, NDPRTR_ZONE_MAX * ndprtr_size, 0,
252 NDPRTR_ZONE_NAME);
253 if (ndprtr_zone == NULL) {
254 panic("%s: failed allocating %s", __func__, NDPRTR_ZONE_NAME);
255 /* NOTREACHED */
256 }
257 zone_change(ndprtr_zone, Z_EXPAND, TRUE);
258 zone_change(ndprtr_zone, Z_CALLERACCT, FALSE);
259 }
260
261 /*
262 * Receive Router Solicitation Message - just for routers.
263 * Router solicitation/advertisement is mostly managed by userland program
264 * (rtadvd) so here we have no function like nd6_ra_output().
265 *
266 * Based on RFC 2461
267 */
268 void
269 nd6_rs_input(
270 struct mbuf *m,
271 int off,
272 int icmp6len)
273 {
274 struct ifnet *ifp = m->m_pkthdr.rcvif;
275 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
276 struct nd_router_solicit *nd_rs;
277 struct in6_addr saddr6 = ip6->ip6_src;
278 char *lladdr = NULL;
279 int lladdrlen = 0;
280 union nd_opts ndopts;
281
282 /* Expect 32-bit aligned data pointer on strict-align platforms */
283 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
284
285 /* If I'm not a router, ignore it. */
286 if (!ip6_forwarding || !(ifp->if_eflags & IFEF_IPV6_ROUTER))
287 goto freeit;
288
289 /* Sanity checks */
290 if (ip6->ip6_hlim != 255) {
291 nd6log((LOG_ERR,
292 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
293 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
294 ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
295 goto bad;
296 }
297
298 /*
299 * Don't update the neighbor cache, if src = :: or a non-neighbor.
300 * The former case indicates that the src has no IP address assigned
301 * yet. See nd6_ns_input() for the latter case.
302 */
303 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
304 goto freeit;
305 } else {
306 struct sockaddr_in6 src_sa6;
307
308 bzero(&src_sa6, sizeof (src_sa6));
309 src_sa6.sin6_family = AF_INET6;
310 src_sa6.sin6_len = sizeof (src_sa6);
311 src_sa6.sin6_addr = ip6->ip6_src;
312 if (!nd6_is_addr_neighbor(&src_sa6, ifp, 0)) {
313 nd6log((LOG_INFO, "nd6_rs_input: "
314 "RS packet from non-neighbor\n"));
315 goto freeit;
316 }
317 }
318
319 IP6_EXTHDR_CHECK(m, off, icmp6len, return);
320 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
321 icmp6len -= sizeof (*nd_rs);
322 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
323 if (nd6_options(&ndopts) < 0) {
324 nd6log((LOG_INFO,
325 "nd6_rs_input: invalid ND option, ignored\n"));
326 /* nd6_options have incremented stats */
327 goto freeit;
328 }
329
330 if (ndopts.nd_opts_src_lladdr) {
331 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
332 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
333 }
334
335 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
336 nd6log((LOG_INFO,
337 "nd6_rs_input: lladdrlen mismatch for %s "
338 "(if %d, RS packet %d)\n",
339 ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
340 goto bad;
341 }
342
343 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
344
345 freeit:
346 m_freem(m);
347 return;
348
349 bad:
350 icmp6stat.icp6s_badrs++;
351 m_freem(m);
352 }
353
354 /*
355 * Receive Router Advertisement Message.
356 *
357 * Based on RFC 2461
358 * TODO: on-link bit on prefix information
359 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
360 */
361 void
362 nd6_ra_input(
363 struct mbuf *m,
364 int off,
365 int icmp6len)
366 {
367 struct ifnet *ifp = m->m_pkthdr.rcvif;
368 struct nd_ifinfo *ndi = NULL;
369 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
370 struct nd_router_advert *nd_ra;
371 struct in6_addr saddr6 = ip6->ip6_src;
372 int mcast = 0;
373 union nd_opts ndopts;
374 struct nd_defrouter *dr = NULL;
375 u_int32_t mtu = 0;
376 char *lladdr = NULL;
377 u_int32_t lladdrlen = 0;
378 struct nd_prefix_list *nd_prefix_list_head = NULL;
379 u_int32_t nd_prefix_list_length = 0;
380 struct in6_ifaddr *ia6 = NULL;
381 struct nd_prefix_list *prfl;
382 struct nd_defrouter dr0;
383 u_int32_t advreachable;
384
385
386 /* Expect 32-bit aligned data pointer on strict-align platforms */
387 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
388
389 /*
390 * Discard RA unless IFEF_ACCEPT_RTADV is set (as host), or when
391 * IFEF_IPV6_ROUTER is set (as router) but the RA is not locally
392 * generated. For convenience, we allow locally generated (rtadvd)
393 * RAs to be processed on the advertising interface, as a router.
394 *
395 * Note that we don't test against ip6_forwarding as we could be
396 * both a host and a router on different interfaces, hence the
397 * check against the per-interface flags.
398 */
399 if (!(ifp->if_eflags & (IFEF_ACCEPT_RTADV | IFEF_IPV6_ROUTER)) ||
400 ((ifp->if_eflags & IFEF_IPV6_ROUTER) &&
401 (ia6 = ifa_foraddr6(&saddr6)) == NULL))
402 goto freeit;
403
404 if (ia6 != NULL) {
405 IFA_REMREF(&ia6->ia_ifa);
406 ia6 = NULL;
407 }
408
409 if (ip6->ip6_hlim != 255) {
410 nd6log((LOG_ERR,
411 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
412 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
413 ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
414 goto bad;
415 }
416
417 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
418 nd6log((LOG_ERR,
419 "nd6_ra_input: src %s is not link-local\n",
420 ip6_sprintf(&saddr6)));
421 goto bad;
422 }
423
424 IP6_EXTHDR_CHECK(m, off, icmp6len, return);
425 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
426
427 icmp6len -= sizeof (*nd_ra);
428 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
429 if (nd6_options(&ndopts) < 0) {
430 nd6log((LOG_INFO,
431 "nd6_ra_input: invalid ND option, ignored\n"));
432 /* nd6_options have incremented stats */
433 goto freeit;
434 }
435
436 advreachable = nd_ra->nd_ra_reachable;
437
438 /* remember if this is a multicasted advertisement */
439 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
440 mcast = 1;
441
442 lck_rw_lock_shared(nd_if_rwlock);
443 if (ifp->if_index >= nd_ifinfo_indexlim) {
444 lck_rw_done(nd_if_rwlock);
445 goto freeit;
446 }
447 ndi = &nd_ifinfo[ifp->if_index];
448 VERIFY(ndi->initialized);
449 lck_mtx_lock(&ndi->lock);
450 bzero(&dr0, sizeof (dr0));
451 dr0.rtaddr = saddr6;
452 dr0.flags = nd_ra->nd_ra_flags_reserved;
453 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
454 dr0.expire = net_uptime() + dr0.rtlifetime;
455 dr0.ifp = ifp;
456 /* unspecified or not? (RFC 2461 6.3.4) */
457 if (advreachable) {
458 advreachable = ntohl(advreachable);
459 if (advreachable <= MAX_REACHABLE_TIME &&
460 ndi->basereachable != advreachable) {
461 ndi->basereachable = advreachable;
462 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
463 ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
464 }
465 }
466 if (nd_ra->nd_ra_retransmit)
467 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
468 if (nd_ra->nd_ra_curhoplimit)
469 ndi->chlim = nd_ra->nd_ra_curhoplimit;
470 lck_mtx_unlock(&ndi->lock);
471 lck_rw_done(nd_if_rwlock);
472 ndi = NULL;
473 lck_mtx_lock(nd6_mutex);
474 dr = defrtrlist_update(&dr0);
475 lck_mtx_unlock(nd6_mutex);
476
477 /*
478 * prefix
479 */
480 if (ndopts.nd_opts_pi) {
481 struct nd_opt_hdr *pt;
482 struct nd_opt_prefix_info *pi = NULL;
483 struct nd_prefix pr;
484
485 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
486 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
487 pt = (struct nd_opt_hdr *)((caddr_t)pt +
488 (pt->nd_opt_len << 3))) {
489 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
490 continue;
491 pi = (struct nd_opt_prefix_info *)pt;
492
493 if (pi->nd_opt_pi_len != 4) {
494 nd6log((LOG_INFO,
495 "nd6_ra_input: invalid option "
496 "len %d for prefix information option, "
497 "ignored\n", pi->nd_opt_pi_len));
498 continue;
499 }
500
501 if (128 < pi->nd_opt_pi_prefix_len) {
502 nd6log((LOG_INFO,
503 "nd6_ra_input: invalid prefix "
504 "len %d for prefix information option, "
505 "ignored\n", pi->nd_opt_pi_prefix_len));
506 continue;
507 }
508
509 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix) ||
510 IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
511 nd6log((LOG_INFO,
512 "%s: invalid prefix %s, ignored\n",
513 __func__,
514 ip6_sprintf(&pi->nd_opt_pi_prefix)));
515 continue;
516 }
517
518 bzero(&pr, sizeof (pr));
519 lck_mtx_init(&pr.ndpr_lock, ifa_mtx_grp, ifa_mtx_attr);
520 NDPR_LOCK(&pr);
521 pr.ndpr_prefix.sin6_family = AF_INET6;
522 pr.ndpr_prefix.sin6_len = sizeof (pr.ndpr_prefix);
523 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
524 pr.ndpr_ifp = m->m_pkthdr.rcvif;
525
526 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
527 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
528 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
529 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
530 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
531 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
532 pr.ndpr_pltime =
533 ntohl(pi->nd_opt_pi_preferred_time);
534
535 /*
536 * Exceptions to stateless autoconfiguration processing:
537 * + nd6_accept_6to4 == 0 && address has 6to4 prefix
538 * + ip6_only_allow_rfc4193_prefix != 0 &&
539 * address not RFC 4193
540 */
541 if (ip6_only_allow_rfc4193_prefix &&
542 !IN6_IS_ADDR_UNIQUE_LOCAL(&pi->nd_opt_pi_prefix)) {
543 nd6log((LOG_INFO,
544 "nd6_ra_input: no SLAAC on prefix %s "
545 "[not RFC 4193]\n",
546 ip6_sprintf(&pi->nd_opt_pi_prefix)));
547 pr.ndpr_raf_auto = 0;
548 } else if (!nd6_accept_6to4 &&
549 IN6_IS_ADDR_6TO4(&pi->nd_opt_pi_prefix)) {
550 nd6log((LOG_INFO,
551 "%s: no SLAAC on prefix %s "
552 "[6to4]\n", __func__,
553 ip6_sprintf(&pi->nd_opt_pi_prefix)));
554 pr.ndpr_raf_auto = 0;
555 }
556
557 if (in6_init_prefix_ltimes(&pr)) {
558 NDPR_UNLOCK(&pr);
559 lck_mtx_destroy(&pr.ndpr_lock, ifa_mtx_grp);
560 continue; /* prefix lifetime init failed */
561 } else {
562 NDPR_UNLOCK(&pr);
563 }
564 (void) prelist_update(&pr, dr, m, mcast);
565 lck_mtx_destroy(&pr.ndpr_lock, ifa_mtx_grp);
566
567 /*
568 * We have to copy the values out after the
569 * prelist_update call since some of these values won't
570 * be properly set until after the router advertisement
571 * updating can vet the values.
572 */
573 prfl = NULL;
574 MALLOC(prfl, struct nd_prefix_list *, sizeof (*prfl),
575 M_TEMP, M_WAITOK | M_ZERO);
576
577 if (prfl == NULL) {
578 log(LOG_DEBUG, "%s: unable to MALLOC RA prefix "
579 "structure\n", __func__);
580 continue;
581 }
582
583 /* this is only for nd6_post_msg(), otherwise unused */
584 bcopy(&pr.ndpr_prefix, &prfl->pr.ndpr_prefix,
585 sizeof (prfl->pr.ndpr_prefix));
586 prfl->pr.ndpr_raf = pr.ndpr_raf;
587 prfl->pr.ndpr_plen = pr.ndpr_plen;
588 prfl->pr.ndpr_vltime = pr.ndpr_vltime;
589 prfl->pr.ndpr_pltime = pr.ndpr_pltime;
590 prfl->pr.ndpr_expire = pr.ndpr_expire;
591 prfl->pr.ndpr_base_calendartime =
592 pr.ndpr_base_calendartime;
593 prfl->pr.ndpr_base_uptime = pr.ndpr_base_uptime;
594 prfl->pr.ndpr_stateflags = pr.ndpr_stateflags;
595 prfl->pr.ndpr_addrcnt = pr.ndpr_addrcnt;
596 prfl->pr.ndpr_ifp = pr.ndpr_ifp;
597
598 prfl->next = nd_prefix_list_head;
599 nd_prefix_list_head = prfl;
600 nd_prefix_list_length++;
601 }
602 }
603
604 /*
605 * MTU
606 */
607 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
608 mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
609
610 /* lower bound */
611 if (mtu < IPV6_MMTU) {
612 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
613 "mtu=%d sent from %s, ignoring\n",
614 mtu, ip6_sprintf(&ip6->ip6_src)));
615 goto skip;
616 }
617
618 lck_rw_lock_shared(nd_if_rwlock);
619 if (ifp->if_index >= nd_ifinfo_indexlim) {
620 lck_rw_done(nd_if_rwlock);
621 goto freeit;
622 }
623 ndi = &nd_ifinfo[ifp->if_index];
624 VERIFY(ndi->initialized);
625 lck_mtx_lock(&ndi->lock);
626 /* upper bound */
627 if (ndi->maxmtu) {
628 if (mtu <= ndi->maxmtu) {
629 int change = (ndi->linkmtu != mtu);
630
631 ndi->linkmtu = mtu;
632 lck_mtx_unlock(&ndi->lock);
633 lck_rw_done(nd_if_rwlock);
634 if (change) /* in6_maxmtu may change */
635 in6_setmaxmtu();
636 } else {
637 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
638 "mtu=%d sent from %s; "
639 "exceeds maxmtu %d, ignoring\n",
640 mtu, ip6_sprintf(&ip6->ip6_src),
641 ndi->maxmtu));
642 lck_mtx_unlock(&ndi->lock);
643 lck_rw_done(nd_if_rwlock);
644 }
645 } else {
646 lck_mtx_unlock(&ndi->lock);
647 lck_rw_done(nd_if_rwlock);
648 nd6log((LOG_INFO, "nd6_ra_input: mtu option "
649 "mtu=%d sent from %s; maxmtu unknown, "
650 "ignoring\n",
651 mtu, ip6_sprintf(&ip6->ip6_src)));
652 }
653 ndi = NULL;
654 }
655
656 skip:
657
658 /*
659 * Source link layer address
660 */
661 if (ndopts.nd_opts_src_lladdr) {
662 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
663 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
664 }
665
666 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
667 nd6log((LOG_INFO,
668 "nd6_ra_input: lladdrlen mismatch for %s "
669 "(if %d, RA packet %d)\n",
670 ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
671 goto bad;
672 }
673
674 nd6_cache_lladdr(ifp, &saddr6, lladdr, (int)lladdrlen,
675 ND_ROUTER_ADVERT, 0);
676
677 /* Post message */
678 nd6_post_msg(KEV_ND6_RA, nd_prefix_list_head, nd_prefix_list_length,
679 mtu, lladdr, lladdrlen);
680
681 /*
682 * Installing a link-layer address might change the state of the
683 * router's neighbor cache, which might also affect our on-link
684 * detection of adveritsed prefixes.
685 */
686 lck_mtx_lock(nd6_mutex);
687 pfxlist_onlink_check();
688 lck_mtx_unlock(nd6_mutex);
689
690 freeit:
691 m_freem(m);
692 if (dr)
693 NDDR_REMREF(dr);
694
695 prfl = NULL;
696 while ((prfl = nd_prefix_list_head) != NULL) {
697 nd_prefix_list_head = prfl->next;
698 FREE(prfl, M_TEMP);
699 }
700
701 return;
702
703 bad:
704 icmp6stat.icp6s_badra++;
705 goto freeit;
706 }
707
708 /*
709 * default router list proccessing sub routines
710 */
711
712 /* tell the change to user processes watching the routing socket. */
713 static void
714 nd6_rtmsg(cmd, rt)
715 int cmd;
716 struct rtentry *rt;
717 {
718 struct rt_addrinfo info;
719 struct ifnet *ifp = rt->rt_ifp;
720
721 RT_LOCK_ASSERT_HELD(rt);
722
723 bzero((caddr_t)&info, sizeof (info));
724 /* It's not necessary to lock ifp for if_lladdr */
725 info.rti_info[RTAX_DST] = rt_key(rt);
726 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
727 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
728 /*
729 * ifa_addr pointers for both should always be valid
730 * in this context; no need to hold locks.
731 */
732 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
733 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
734
735 rt_missmsg(cmd, &info, rt->rt_flags, 0);
736 }
737
738 static void
739 defrouter_addreq(struct nd_defrouter *new, boolean_t scoped)
740 {
741 struct sockaddr_in6 def, mask, gate;
742 struct rtentry *newrt = NULL;
743 unsigned int ifscope;
744 int err;
745
746 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_NOTOWNED);
747 NDDR_LOCK_ASSERT_NOTHELD(new);
748 /*
749 * We're free to lock and unlock NDDR because our callers
750 * are holding an extra reference for us.
751 */
752
753 NDDR_LOCK(new);
754 if (new->stateflags & NDDRF_INSTALLED)
755 goto out;
756
757 if (new->ifp->if_eflags & IFEF_IPV6_ROUTER) {
758 nd6log2((LOG_INFO, "%s: ignoring router %s, scoped=%d, "
759 "static=%d on advertising interface\n", if_name(new->ifp),
760 ip6_sprintf(&new->rtaddr), scoped,
761 (new->stateflags & NDDRF_STATIC) ? 1 : 0));
762 goto out;
763 }
764
765 nd6log2((LOG_INFO, "%s: adding default router %s, scoped=%d, "
766 "static=%d\n", if_name(new->ifp), ip6_sprintf(&new->rtaddr),
767 scoped, (new->stateflags & NDDRF_STATIC) ? 1 : 0));
768
769 Bzero(&def, sizeof (def));
770 Bzero(&mask, sizeof (mask));
771 Bzero(&gate, sizeof (gate));
772
773 def.sin6_len = mask.sin6_len = gate.sin6_len
774 = sizeof (struct sockaddr_in6);
775 def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
776 gate.sin6_addr = new->rtaddr;
777
778 ifscope = scoped ? new->ifp->if_index : IFSCOPE_NONE;
779 NDDR_UNLOCK(new);
780
781 err = rtrequest_scoped(RTM_ADD, (struct sockaddr *)&def,
782 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
783 RTF_GATEWAY, &newrt, ifscope);
784
785 if (newrt) {
786 RT_LOCK(newrt);
787 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
788 RT_REMREF_LOCKED(newrt);
789 RT_UNLOCK(newrt);
790 NDDR_LOCK(new);
791 new->stateflags |= NDDRF_INSTALLED;
792 if (ifscope != IFSCOPE_NONE)
793 new->stateflags |= NDDRF_IFSCOPE;
794 new->genid = nd6_defrouter_genid;
795 } else {
796 nd6log((LOG_ERR, "%s: failed to add default router "
797 "%s on %s scoped %d (errno = %d)\n", __func__,
798 ip6_sprintf(&gate.sin6_addr), if_name(new->ifp),
799 (ifscope != IFSCOPE_NONE), err));
800 NDDR_LOCK(new);
801 }
802 new->err = err;
803
804 out:
805 NDDR_UNLOCK(new);
806 }
807
808 struct nd_defrouter *
809 defrouter_lookup(
810 struct in6_addr *addr,
811 struct ifnet *ifp)
812 {
813 struct nd_defrouter *dr;
814
815 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
816
817 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
818 dr = TAILQ_NEXT(dr, dr_entry)) {
819 NDDR_LOCK(dr);
820 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
821 NDDR_ADDREF_LOCKED(dr);
822 NDDR_UNLOCK(dr);
823 return (dr);
824 }
825 NDDR_UNLOCK(dr);
826 }
827
828 return (NULL); /* search failed */
829 }
830
831 /*
832 * Remove the default route for a given router.
833 * This is just a subroutine function for defrouter_select(), and should
834 * not be called from anywhere else.
835 */
836 static void
837 defrouter_delreq(struct nd_defrouter *dr)
838 {
839 struct sockaddr_in6 def, mask, gate;
840 struct rtentry *oldrt = NULL;
841 unsigned int ifscope;
842 int err;
843
844 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_NOTOWNED);
845 NDDR_LOCK_ASSERT_NOTHELD(dr);
846 /*
847 * We're free to lock and unlock NDDR because our callers
848 * are holding an extra reference for us.
849 */
850 NDDR_LOCK(dr);
851 /* ifp would be NULL for the "drany" case */
852 if (dr->ifp != NULL && !(dr->stateflags & NDDRF_INSTALLED))
853 goto out;
854
855 nd6log2((LOG_INFO, "%s: removing default router %s, scoped=%d, "
856 "static=%d\n", dr->ifp != NULL ? if_name(dr->ifp) : "ANY",
857 ip6_sprintf(&dr->rtaddr), (dr->stateflags & NDDRF_IFSCOPE) ? 1 : 0,
858 (dr->stateflags & NDDRF_STATIC) ? 1 : 0));
859
860 Bzero(&def, sizeof (def));
861 Bzero(&mask, sizeof (mask));
862 Bzero(&gate, sizeof (gate));
863
864 def.sin6_len = mask.sin6_len = gate.sin6_len
865 = sizeof (struct sockaddr_in6);
866 def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
867 gate.sin6_addr = dr->rtaddr;
868
869 if (dr->ifp != NULL) {
870 ifscope = (dr->stateflags & NDDRF_IFSCOPE) ?
871 dr->ifp->if_index : IFSCOPE_NONE;
872 } else {
873 ifscope = IFSCOPE_NONE;
874 }
875 NDDR_UNLOCK(dr);
876 err = rtrequest_scoped(RTM_DELETE,
877 (struct sockaddr *)&def, (struct sockaddr *)&gate,
878 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, ifscope);
879
880 if (oldrt) {
881 RT_LOCK(oldrt);
882 nd6_rtmsg(RTM_DELETE, oldrt);
883 RT_UNLOCK(oldrt);
884 rtfree(oldrt);
885 } else if (err != ESRCH) {
886 nd6log((LOG_ERR, "%s: failed to delete default router "
887 "%s on %s scoped %d (errno = %d)\n", __func__,
888 ip6_sprintf(&gate.sin6_addr), dr->ifp != NULL ?
889 if_name(dr->ifp) : "ANY", (ifscope != IFSCOPE_NONE), err));
890 }
891 NDDR_LOCK(dr);
892 /* ESRCH means it's no longer in the routing table; ignore it */
893 if (oldrt != NULL || err == ESRCH) {
894 dr->stateflags &= ~NDDRF_INSTALLED;
895 if (ifscope != IFSCOPE_NONE)
896 dr->stateflags &= ~NDDRF_IFSCOPE;
897 }
898 dr->err = 0;
899 out:
900 NDDR_UNLOCK(dr);
901 }
902
903
904 /*
905 * remove all default routes from default router list
906 */
907 void
908 defrouter_reset(void)
909 {
910 struct nd_defrouter *dr, drany;
911
912 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
913
914 dr = TAILQ_FIRST(&nd_defrouter);
915 while (dr) {
916 NDDR_LOCK(dr);
917 if (dr->stateflags & NDDRF_INSTALLED) {
918 NDDR_ADDREF_LOCKED(dr);
919 NDDR_UNLOCK(dr);
920 lck_mtx_unlock(nd6_mutex);
921 defrouter_delreq(dr);
922 lck_mtx_lock(nd6_mutex);
923 NDDR_REMREF(dr);
924 dr = TAILQ_FIRST(&nd_defrouter);
925 } else {
926 NDDR_UNLOCK(dr);
927 dr = TAILQ_NEXT(dr, dr_entry);
928 }
929 }
930
931 /* Nuke primary (non-scoped) default router */
932 if (ip6_doscopedroute) {
933 bzero(&drany, sizeof (drany));
934 lck_mtx_init(&drany.nddr_lock, ifa_mtx_grp, ifa_mtx_attr);
935 lck_mtx_unlock(nd6_mutex);
936 defrouter_delreq(&drany);
937 lck_mtx_destroy(&drany.nddr_lock, ifa_mtx_grp);
938 lck_mtx_lock(nd6_mutex);
939 }
940
941 }
942
943 int
944 defrtrlist_ioctl(u_long cmd, caddr_t data)
945 {
946 struct nd_defrouter dr0;
947 unsigned int ifindex;
948 struct ifnet *dr_ifp;
949 int error = 0, add = 0;
950
951 switch (cmd) {
952 case SIOCDRADD_IN6_32: /* struct in6_defrouter_32 */
953 case SIOCDRADD_IN6_64: /* struct in6_defrouter_64 */
954 ++add;
955 /* FALLTHRU */
956 case SIOCDRDEL_IN6_32: /* struct in6_defrouter_32 */
957 case SIOCDRDEL_IN6_64: /* struct in6_defrouter_64 */
958 bzero(&dr0, sizeof (dr0));
959 if (cmd == SIOCDRADD_IN6_64 || cmd == SIOCDRDEL_IN6_64) {
960 struct in6_defrouter_64 *r_64 =
961 (struct in6_defrouter_64 *)(void *)data;
962 u_int16_t i;
963
964 bcopy(&r_64->rtaddr.sin6_addr, &dr0.rtaddr,
965 sizeof (dr0.rtaddr));
966 dr0.flags = r_64->flags;
967 bcopy(&r_64->if_index, &i, sizeof (i));
968 ifindex = i;
969 } else {
970 struct in6_defrouter_32 *r_32 =
971 (struct in6_defrouter_32 *)(void *)data;
972 u_int16_t i;
973
974 bcopy(&r_32->rtaddr.sin6_addr, &dr0.rtaddr,
975 sizeof (dr0.rtaddr));
976 dr0.flags = r_32->flags;
977 bcopy(&r_32->if_index, &i, sizeof (i));
978 ifindex = i;
979 }
980 ifnet_head_lock_shared();
981 /* Don't need to check is ifindex is < 0 since it's unsigned */
982 if (if_index < ifindex ||
983 (dr_ifp = ifindex2ifnet[ifindex]) == NULL) {
984 ifnet_head_done();
985 error = EINVAL;
986 break;
987 }
988 dr0.ifp = dr_ifp;
989 ifnet_head_done();
990
991 if (IN6_IS_SCOPE_EMBED(&dr0.rtaddr)) {
992 uint16_t *scope = &dr0.rtaddr.s6_addr16[1];
993
994 if (*scope == 0) {
995 *scope = htons(dr_ifp->if_index);
996 } else if (*scope != htons(dr_ifp->if_index)) {
997 error = EINVAL;
998 break;
999 }
1000 }
1001
1002 if (add)
1003 error = defrtrlist_add_static(&dr0);
1004 if (!add || error != 0) {
1005 int err = defrtrlist_del_static(&dr0);
1006 if (!add)
1007 error = err;
1008 }
1009 break;
1010
1011 default:
1012 error = EOPNOTSUPP; /* check for safety */
1013 break;
1014 }
1015
1016 return (error);
1017 }
1018
1019 void
1020 defrtrlist_del(struct nd_defrouter *dr)
1021 {
1022 struct nd_defrouter *deldr = NULL;
1023 struct nd_prefix *pr;
1024 struct ifnet *ifp = dr->ifp;
1025 boolean_t resetmtu;
1026
1027 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1028
1029 if (!ip6_doscopedroute && dr == TAILQ_FIRST(&nd_defrouter))
1030 deldr = dr; /* The router is primary. */
1031
1032 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1033 ++nd6_defrouter_genid;
1034 /*
1035 * Flush all the routing table entries that use the router
1036 * as a next hop.
1037 */
1038 if (ip6_doscopedroute || !ip6_forwarding) {
1039 /* above is a good condition? */
1040 NDDR_ADDREF(dr);
1041 lck_mtx_unlock(nd6_mutex);
1042 rt6_flush(&dr->rtaddr, ifp);
1043 lck_mtx_lock(nd6_mutex);
1044 NDDR_REMREF(dr);
1045 }
1046 nd6log2((LOG_INFO, "%s: freeing defrouter %s\n", if_name(dr->ifp),
1047 ip6_sprintf(&dr->rtaddr)));
1048 /*
1049 * Delete it from the routing table.
1050 */
1051 NDDR_ADDREF(dr);
1052 lck_mtx_unlock(nd6_mutex);
1053 defrouter_delreq(dr);
1054 lck_mtx_lock(nd6_mutex);
1055 NDDR_REMREF(dr);
1056
1057 /*
1058 * Also delete all the pointers to the router in each prefix lists.
1059 */
1060 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1061 struct nd_pfxrouter *pfxrtr;
1062
1063 NDPR_LOCK(pr);
1064 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
1065 pfxrtr_del(pfxrtr, pr);
1066 NDPR_UNLOCK(pr);
1067 }
1068
1069 pfxlist_onlink_check();
1070
1071 /*
1072 * If the router is the primary one, choose a new one. If Scoped
1073 * Routing is enabled, always try to pick another eligible router
1074 * on this interface.
1075 */
1076 if (deldr || ip6_doscopedroute)
1077 defrouter_select(ifp);
1078
1079 resetmtu = FALSE;
1080 lck_rw_lock_shared(nd_if_rwlock);
1081 if (ifp->if_index < nd_ifinfo_indexlim) {
1082 struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
1083 VERIFY(ndi->initialized);
1084 lck_mtx_lock(&ndi->lock);
1085 VERIFY(ndi->ndefrouters >= 0);
1086 if (ndi->ndefrouters > 0 && --ndi->ndefrouters == 0) {
1087 nd6_ifreset(ifp);
1088 resetmtu = TRUE;
1089 }
1090 lck_mtx_unlock(&ndi->lock);
1091 }
1092 lck_rw_done(nd_if_rwlock);
1093
1094 if (resetmtu)
1095 nd6_setmtu(ifp);
1096
1097 NDDR_REMREF(dr); /* remove list reference */
1098 }
1099
1100 int
1101 defrtrlist_add_static(struct nd_defrouter *new)
1102 {
1103 struct nd_defrouter *dr;
1104 int err = 0;
1105
1106 new->rtlifetime = -1;
1107 new->stateflags |= NDDRF_STATIC;
1108
1109 /* we only want the preference level */
1110 new->flags &= ND_RA_FLAG_RTPREF_MASK;
1111
1112 lck_mtx_lock(nd6_mutex);
1113 dr = defrouter_lookup(&new->rtaddr, new->ifp);
1114 if (dr != NULL && !(dr->stateflags & NDDRF_STATIC)) {
1115 err = EINVAL;
1116 } else {
1117 if (dr != NULL)
1118 NDDR_REMREF(dr);
1119 dr = defrtrlist_update(new);
1120 if (dr != NULL)
1121 err = dr->err;
1122 else
1123 err = ENOMEM;
1124 }
1125 if (dr != NULL)
1126 NDDR_REMREF(dr);
1127 lck_mtx_unlock(nd6_mutex);
1128
1129 return (err);
1130 }
1131
1132 int
1133 defrtrlist_del_static(struct nd_defrouter *new)
1134 {
1135 struct nd_defrouter *dr;
1136
1137 lck_mtx_lock(nd6_mutex);
1138 dr = defrouter_lookup(&new->rtaddr, new->ifp);
1139 if (dr == NULL || !(dr->stateflags & NDDRF_STATIC)) {
1140 if (dr != NULL)
1141 NDDR_REMREF(dr);
1142 dr = NULL;
1143 } else {
1144 defrtrlist_del(dr);
1145 NDDR_REMREF(dr);
1146 }
1147 lck_mtx_unlock(nd6_mutex);
1148
1149 return (dr != NULL ? 0 : EINVAL);
1150 }
1151
1152 /*
1153 * for default router selection
1154 * regards router-preference field as a 2-bit signed integer
1155 */
1156 static int
1157 rtpref(struct nd_defrouter *dr)
1158 {
1159 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
1160 case ND_RA_FLAG_RTPREF_HIGH:
1161 return (RTPREF_HIGH);
1162 case ND_RA_FLAG_RTPREF_MEDIUM:
1163 case ND_RA_FLAG_RTPREF_RSV:
1164 return (RTPREF_MEDIUM);
1165 case ND_RA_FLAG_RTPREF_LOW:
1166 return (RTPREF_LOW);
1167 default:
1168 /*
1169 * This case should never happen. If it did, it would mean a
1170 * serious bug of kernel internal. We thus always bark here.
1171 * Or, can we even panic?
1172 */
1173 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
1174 return (RTPREF_INVALID);
1175 }
1176 /* NOTREACHED */
1177 }
1178
1179 /*
1180 * Default Router Selection according to Section 6.3.6 of RFC 2461 and RFC 4191:
1181 *
1182 * 1) Routers that are reachable or probably reachable should be preferred.
1183 * If we have more than one (probably) reachable router, prefer ones
1184 * with the highest router preference.
1185 * 2) When no routers on the list are known to be reachable or
1186 * probably reachable, routers SHOULD be selected in a round-robin
1187 * fashion, regardless of router preference values.
1188 * 3) If the Default Router List is empty, assume that all
1189 * destinations are on-link.
1190 *
1191 * When Scoped Routing is enabled, the selection logic is amended as follows:
1192 *
1193 * a) When a default interface is specified, the primary/non-scoped default
1194 * router will be set to the reachable router on that link (if any) with
1195 * the highest router preference.
1196 * b) When there are more than one routers on the same link, the one with
1197 * the highest router preference will be installed, either as scoped or
1198 * non-scoped route entry. If they all share the same preference value,
1199 * the one installed will be the static or the first encountered reachable
1200 * router, i.e. static one wins over dynamic.
1201 * c) When no routers on the list are known to be reachable, or probably
1202 * reachable, no round-robin selection will take place when the default
1203 * interface is set.
1204 *
1205 * We assume nd_defrouter is sorted by router preference value.
1206 * Since the code below covers both with and without router preference cases,
1207 * we do not need to classify the cases by ifdef.
1208 */
1209 void
1210 defrouter_select(struct ifnet *ifp)
1211 {
1212 #pragma unused(ifp)
1213 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
1214 struct nd_defrouter *installed_dr0 = NULL;
1215 struct rtentry *rt = NULL;
1216 struct llinfo_nd6 *ln = NULL;
1217 int update = 0;
1218 boolean_t found_installedrt = FALSE;
1219
1220 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1221
1222 /*
1223 * We no longer install (default) interface route; only prefix routes
1224 * are installed as interface routes. Therefore, there is no harm in
1225 * going through this routine even if a default interface is specified,
1226 * which happens when Scoped Routing is enabled. But for efficiency,
1227 * we fall back to the original KAME logic when Scoped Routing is
1228 * not in effect.
1229 */
1230 if (ip6_forwarding && !ip6_doscopedroute) {
1231 nd6log((LOG_WARNING,
1232 "defrouter_select: called unexpectedly (forwarding=%d)\n",
1233 ip6_forwarding));
1234 return;
1235 }
1236
1237 /*
1238 * Let's handle easy case (3) first:
1239 * If default router list is empty, there's nothing to be done.
1240 */
1241 if (!TAILQ_FIRST(&nd_defrouter))
1242 return;
1243
1244 /*
1245 * Due to the number of times we drop nd6_mutex, we need to
1246 * serialize this function.
1247 */
1248 while (nd_defrouter_busy) {
1249 nd_defrouter_waiters++;
1250 msleep(nd_defrouter_waitchan, nd6_mutex, (PZERO-1),
1251 __func__, NULL);
1252 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1253 }
1254 nd_defrouter_busy = TRUE;
1255
1256 /*
1257 * Search for a (probably) reachable router from the list.
1258 * We just pick up the first reachable one (if any), assuming that
1259 * the ordering rule of the list described in defrtrlist_update().
1260 *
1261 * For all intents and purposes of Scoped Routing:
1262 * selected_dr = candidate for primary router
1263 * installed_dr = currently installed primary router
1264 */
1265 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1266 dr = TAILQ_NEXT(dr, dr_entry)) {
1267 boolean_t reachable, advrouter;
1268 struct in6_addr rtaddr;
1269 struct ifnet *drifp;
1270 struct nd_defrouter *drrele;
1271
1272 drrele = NULL;
1273 reachable = FALSE;
1274 NDDR_LOCK(dr);
1275 rtaddr = *(&dr->rtaddr);
1276 drifp = dr->ifp;
1277 advrouter = (drifp != NULL &&
1278 (drifp->if_eflags & IFEF_IPV6_ROUTER));
1279 NDDR_ADDREF_LOCKED(dr); /* for this for loop */
1280 NDDR_UNLOCK(dr);
1281
1282 lck_mtx_unlock(nd6_mutex);
1283 /* Callee returns a locked route upon success */
1284 if ((rt = nd6_lookup(&rtaddr, 0, drifp, 0)) != NULL) {
1285 RT_LOCK_ASSERT_HELD(rt);
1286 if ((ln = rt->rt_llinfo) != NULL &&
1287 ND6_IS_LLINFO_PROBREACH(ln)) {
1288 reachable = TRUE;
1289 if (selected_dr == NULL &&
1290 (!ip6_doscopedroute ||
1291 (drifp == nd6_defifp && !advrouter))) {
1292 selected_dr = dr;
1293 NDDR_ADDREF(selected_dr);
1294 }
1295 }
1296 RT_REMREF_LOCKED(rt);
1297 RT_UNLOCK(rt);
1298 rt = NULL;
1299 }
1300 lck_mtx_lock(nd6_mutex);
1301
1302 /* Handle case (b) */
1303 NDDR_LOCK(dr);
1304 if (ip6_doscopedroute && drifp == nd6_defifp && !advrouter &&
1305 (selected_dr == NULL || rtpref(dr) > rtpref(selected_dr) ||
1306 (rtpref(dr) == rtpref(selected_dr) &&
1307 (dr->stateflags & NDDRF_STATIC) &&
1308 !(selected_dr->stateflags & NDDRF_STATIC)))) {
1309 if (selected_dr) {
1310 /* Release it later on */
1311 VERIFY(drrele == NULL);
1312 drrele = selected_dr;
1313 }
1314 selected_dr = dr;
1315 NDDR_ADDREF_LOCKED(selected_dr);
1316 }
1317
1318 if (!(dr->stateflags & NDDRF_INSTALLED)) {
1319 /*
1320 * If the router hasn't been installed and it is
1321 * reachable, try to install it later on below.
1322 * If it's static, try to install it anyway.
1323 */
1324 if (!advrouter && (reachable ||
1325 (dr->stateflags & NDDRF_STATIC))) {
1326 dr->genid = -1;
1327 ++update;
1328 nd6log2((LOG_INFO, "%s: possible router %s, "
1329 "scoped=%d, static=%d\n", if_name(drifp),
1330 ip6_sprintf(&rtaddr),
1331 (dr->stateflags & NDDRF_IFSCOPE) ? 1 : 0,
1332 (dr->stateflags & NDDRF_STATIC) ? 1 : 0));
1333 }
1334 NDDR_UNLOCK(dr);
1335 NDDR_REMREF(dr); /* for this for loop */
1336 if (drrele != NULL)
1337 NDDR_REMREF(drrele);
1338 continue;
1339 }
1340
1341 /* Record the currently installed primary/non-scoped router */
1342 if (!ip6_doscopedroute || !(dr->stateflags & NDDRF_IFSCOPE)) {
1343 if (installed_dr == NULL) {
1344 installed_dr = dr;
1345 NDDR_ADDREF_LOCKED(installed_dr);
1346 } else {
1347 /* this should not happen; warn for diagnosis */
1348 log(LOG_ERR, "defrouter_select: more than one "
1349 "%s default router is installed\n",
1350 ip6_doscopedroute ? "non-scoped" : "");
1351 }
1352 }
1353 NDDR_UNLOCK(dr);
1354 NDDR_REMREF(dr); /* for this for loop */
1355 if (drrele != NULL)
1356 NDDR_REMREF(drrele);
1357 }
1358
1359 /* If none was selected, use the currently installed one */
1360 if (ip6_doscopedroute && selected_dr == NULL && installed_dr != NULL) {
1361 selected_dr = installed_dr;
1362 NDDR_ADDREF(selected_dr);
1363 }
1364
1365 /*
1366 * Install the unreachable one(s) if necesssary.
1367 */
1368 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1369 dr = TAILQ_NEXT(dr, dr_entry)) {
1370 struct nd_defrouter *_dr;
1371
1372 if (!ip6_doscopedroute)
1373 break;
1374
1375 NDDR_LOCK(dr);
1376
1377 /* If already (or will be) installed, skip */
1378 if ((dr->stateflags & NDDRF_INSTALLED) || dr->genid == -1) {
1379 NDDR_UNLOCK(dr);
1380 continue;
1381 }
1382
1383 /* See if there is already a default router for the link */
1384 for (_dr = TAILQ_FIRST(&nd_defrouter); _dr;
1385 _dr = TAILQ_NEXT(_dr, dr_entry)) {
1386 if (_dr != dr)
1387 NDDR_LOCK(_dr);
1388 if (_dr == dr || _dr->ifp != dr->ifp) {
1389 if (_dr != dr)
1390 NDDR_UNLOCK(_dr);
1391 continue;
1392 }
1393
1394 if ((_dr->stateflags & NDDRF_INSTALLED) ||
1395 _dr->genid == -1) {
1396 if (_dr != dr)
1397 NDDR_UNLOCK(_dr);
1398 break;
1399 }
1400 if (_dr != dr)
1401 NDDR_UNLOCK(_dr);
1402 }
1403
1404 /* If none so far, schedule it to be installed below */
1405 if (_dr == NULL && dr->ifp != NULL &&
1406 !(dr->ifp->if_eflags & IFEF_IPV6_ROUTER)) {
1407 dr->genid = -1;
1408 ++update;
1409 nd6log2((LOG_INFO, "%s: possible router %s, "
1410 "static=%d (unreachable)\n", if_name(dr->ifp),
1411 ip6_sprintf(&dr->rtaddr),
1412 (dr->stateflags & NDDRF_STATIC) ? 1 : 0));
1413 }
1414 NDDR_UNLOCK(dr);
1415 }
1416
1417 dr = selected_dr;
1418 if (dr != NULL) {
1419 nd6log2((LOG_INFO, "%s: considering primary default router %s, "
1420 "static=%d [round 1]\n", if_name(dr->ifp),
1421 ip6_sprintf(&dr->rtaddr),
1422 (dr->stateflags & NDDRF_STATIC) ? 1 : 0));
1423 }
1424
1425 /*
1426 * If none of the default routers was found to be reachable,
1427 * round-robin the list regardless of preference, except when
1428 * Scoped Routing is enabled per case (c).
1429 *
1430 * Otherwise, if we have an installed router, check if the selected
1431 * (reachable) router should really be preferred to the installed one.
1432 * We only prefer the new router when the old one is not reachable
1433 * or when the new one has a really higher preference value.
1434 */
1435 if (!ip6_doscopedroute && selected_dr == NULL) {
1436 if (installed_dr == NULL ||
1437 !TAILQ_NEXT(installed_dr, dr_entry)) {
1438 selected_dr = TAILQ_FIRST(&nd_defrouter);
1439 if (selected_dr)
1440 NDDR_ADDREF(selected_dr);
1441 } else {
1442 selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
1443 if (selected_dr)
1444 NDDR_ADDREF(selected_dr);
1445 }
1446 } else if (selected_dr != NULL && installed_dr != NULL) {
1447 lck_mtx_unlock(nd6_mutex);
1448 rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp, 0);
1449 if (rt) {
1450 RT_LOCK_ASSERT_HELD(rt);
1451 if ((ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1452 ND6_IS_LLINFO_PROBREACH(ln) &&
1453 (!ip6_doscopedroute ||
1454 installed_dr->ifp == nd6_defifp) &&
1455 rtpref(selected_dr) <= rtpref(installed_dr)) {
1456 NDDR_REMREF(selected_dr);
1457 selected_dr = installed_dr;
1458 NDDR_ADDREF(selected_dr);
1459 }
1460 RT_REMREF_LOCKED(rt);
1461 RT_UNLOCK(rt);
1462 rt = NULL;
1463 found_installedrt = TRUE;
1464 }
1465 lck_mtx_lock(nd6_mutex);
1466 }
1467
1468 if (ip6_doscopedroute) {
1469 /*
1470 * If the installed primary router is not on the current
1471 * IPv6 default interface, demote it to a scoped entry.
1472 */
1473 if (installed_dr != NULL && installed_dr->ifp != nd6_defifp &&
1474 !(installed_dr->stateflags & NDDRF_IFSCOPE)) {
1475 if (selected_dr != NULL &&
1476 selected_dr->ifp != nd6_defifp) {
1477 NDDR_REMREF(selected_dr);
1478 selected_dr = NULL;
1479 }
1480 ++update;
1481 }
1482
1483 /*
1484 * If the selected router is currently scoped, make sure
1485 * we update (it needs to be promoted to primary.)
1486 */
1487 if (selected_dr != NULL &&
1488 (selected_dr->stateflags & NDDRF_IFSCOPE))
1489 ++update;
1490
1491 /*
1492 * If the installed router is no longer reachable, remove
1493 * it and install the selected router instead.
1494 */
1495 if (installed_dr != NULL
1496 && selected_dr != NULL
1497 && installed_dr != selected_dr
1498 && found_installedrt == FALSE
1499 && installed_dr->ifp == selected_dr->ifp) {
1500 /* skip it below */
1501 installed_dr0 = installed_dr;
1502 /* NB: we previousled referenced installed_dr */
1503 installed_dr = NULL;
1504 selected_dr->genid = -1;
1505 ++update;
1506 }
1507 }
1508
1509 /*
1510 * If Scoped Routing is enabled and there's nothing to update,
1511 * just return. Otherwise, if Scoped Routing is disabled and if
1512 * the selected router is different than the installed one,
1513 * remove the installed router and install the selected one.
1514 */
1515 dr = selected_dr;
1516 VERIFY(dr != NULL || ip6_doscopedroute);
1517 if (!ip6_doscopedroute || !update) {
1518 if (dr == NULL)
1519 goto out;
1520
1521 if (dr != installed_dr) {
1522 nd6log2((LOG_INFO, "%s: no update, selected router %s, "
1523 "installed router %s\n", if_name(dr->ifp),
1524 ip6_sprintf(&dr->rtaddr), installed_dr != NULL ?
1525 ip6_sprintf(&installed_dr->rtaddr) : "NONE"));
1526 } else {
1527 nd6log2((LOG_INFO, "%s: no update, router is %s\n",
1528 if_name(dr->ifp), ip6_sprintf(&dr->rtaddr)));
1529 }
1530 if (!ip6_doscopedroute && installed_dr != dr) {
1531 /*
1532 * No need to ADDREF dr because at this point
1533 * dr points to selected_dr, which already holds
1534 * a reference.
1535 */
1536 lck_mtx_unlock(nd6_mutex);
1537 if (installed_dr) {
1538 defrouter_delreq(installed_dr);
1539 }
1540 defrouter_addreq(dr, FALSE);
1541 lck_mtx_lock(nd6_mutex);
1542 }
1543 goto out;
1544 }
1545
1546 /*
1547 * Scoped Routing is enabled and we need to update. The selected
1548 * router needs to be installed as primary/non-scoped entry. If
1549 * there is any existing entry that is non-scoped, remove it from
1550 * the routing table and reinstall it as scoped entry.
1551 */
1552 if (dr != NULL) {
1553 nd6log2((LOG_INFO, "%s: considering primary default router %s, "
1554 "static=%d [round 2]\n", if_name(dr->ifp),
1555 ip6_sprintf(&dr->rtaddr),
1556 (dr->stateflags & NDDRF_STATIC) ? 1 : 0));
1557 }
1558
1559 /*
1560 * On the following while loops we use two flags:
1561 * dr->genid
1562 * NDDRF_PROCESSED
1563 *
1564 * genid is used to skip entries that are not to be added/removed on the
1565 * second while loop.
1566 * NDDRF_PROCESSED is used to skip entries that were already
1567 * processed.
1568 * This is necessary because we drop the nd6_mutex and start the while
1569 * loop again.
1570 */
1571 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1572 NDDR_LOCK(dr);
1573 VERIFY((dr->stateflags & NDDRF_PROCESSED) == 0);
1574 NDDR_UNLOCK(dr);
1575 }
1576 /* Remove conflicting entries */
1577 dr = TAILQ_FIRST(&nd_defrouter);
1578 while (dr) {
1579 NDDR_LOCK(dr);
1580 if (!(dr->stateflags & NDDRF_INSTALLED) ||
1581 dr->stateflags & NDDRF_PROCESSED) {
1582 NDDR_UNLOCK(dr);
1583 dr = TAILQ_NEXT(dr, dr_entry);
1584 continue;
1585 }
1586 dr->stateflags |= NDDRF_PROCESSED;
1587
1588 /* A NULL selected_dr will remove primary default route */
1589 if ((dr == selected_dr && (dr->stateflags & NDDRF_IFSCOPE)) ||
1590 (dr != selected_dr && !(dr->stateflags & NDDRF_IFSCOPE))) {
1591 NDDR_ADDREF_LOCKED(dr);
1592 NDDR_UNLOCK(dr);
1593 lck_mtx_unlock(nd6_mutex);
1594 defrouter_delreq(dr);
1595 lck_mtx_lock(nd6_mutex);
1596 NDDR_LOCK(dr);
1597 if (dr && dr != installed_dr0)
1598 dr->genid = -1;
1599 NDDR_UNLOCK(dr);
1600 NDDR_REMREF(dr);
1601 /*
1602 * Since we lost nd6_mutex, we have to start over.
1603 */
1604 dr = TAILQ_FIRST(&nd_defrouter);
1605 continue;
1606 }
1607 NDDR_UNLOCK(dr);
1608 dr = TAILQ_NEXT(dr, dr_entry);
1609 }
1610
1611 /* -1 is a special number, make sure we don't use it for genid */
1612 if (++nd6_defrouter_genid == -1)
1613 nd6_defrouter_genid = 1;
1614
1615 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1616 NDDR_LOCK(dr);
1617 dr->stateflags &= ~NDDRF_PROCESSED;
1618 NDDR_UNLOCK(dr);
1619 }
1620 /* Add the entries back */
1621 dr = TAILQ_FIRST(&nd_defrouter);
1622 while (dr) {
1623 struct nd_defrouter *_dr;
1624
1625 NDDR_LOCK(dr);
1626 if (dr->stateflags & NDDRF_PROCESSED ||
1627 dr->genid != -1) {
1628 NDDR_UNLOCK(dr);
1629 dr = TAILQ_NEXT(dr, dr_entry);
1630 continue;
1631 }
1632 dr->stateflags |= NDDRF_PROCESSED;
1633
1634 /* Handle case (b) */
1635 for (_dr = TAILQ_FIRST(&nd_defrouter); _dr;
1636 _dr = TAILQ_NEXT(_dr, dr_entry)) {
1637 if (_dr == dr)
1638 continue;
1639 /*
1640 * This is safe because we previously checked if
1641 * _dr == dr.
1642 */
1643 NDDR_LOCK(_dr);
1644 if (_dr->ifp == dr->ifp && rtpref(_dr) >= rtpref(dr) &&
1645 (_dr->stateflags & NDDRF_INSTALLED)) {
1646 NDDR_ADDREF_LOCKED(_dr);
1647 NDDR_UNLOCK(_dr);
1648 break;
1649 }
1650 NDDR_UNLOCK(_dr);
1651 }
1652
1653 /* If same preference and i/f, static entry takes precedence */
1654 if (_dr != NULL && rtpref(_dr) == rtpref(dr) &&
1655 !(_dr->stateflags & NDDRF_STATIC) &&
1656 (dr->stateflags & NDDRF_STATIC)) {
1657 lck_mtx_unlock(nd6_mutex);
1658 defrouter_delreq(_dr);
1659 lck_mtx_lock(nd6_mutex);
1660 NDDR_REMREF(_dr);
1661 _dr = NULL;
1662 }
1663
1664 if (_dr == NULL && !(dr->stateflags & NDDRF_INSTALLED)) {
1665 NDDR_ADDREF_LOCKED(dr);
1666 NDDR_UNLOCK(dr);
1667 lck_mtx_unlock(nd6_mutex);
1668 defrouter_addreq(dr, (selected_dr == NULL ||
1669 dr->ifp != selected_dr->ifp));
1670 dr->genid = nd6_defrouter_genid;
1671 lck_mtx_lock(nd6_mutex);
1672 NDDR_REMREF(dr);
1673 /*
1674 * Since we lost nd6_mutex, we have to start over.
1675 */
1676 dr = TAILQ_FIRST(&nd_defrouter);
1677 continue;
1678 }
1679 NDDR_UNLOCK(dr);
1680 dr = TAILQ_NEXT(dr, dr_entry);
1681 }
1682 out:
1683 TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1684 NDDR_LOCK(dr);
1685 dr->stateflags &= ~NDDRF_PROCESSED;
1686 NDDR_UNLOCK(dr);
1687 }
1688 if (selected_dr)
1689 NDDR_REMREF(selected_dr);
1690 if (installed_dr)
1691 NDDR_REMREF(installed_dr);
1692 if (installed_dr0)
1693 NDDR_REMREF(installed_dr0);
1694 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1695 VERIFY(nd_defrouter_busy);
1696 nd_defrouter_busy = FALSE;
1697 if (nd_defrouter_waiters > 0) {
1698 nd_defrouter_waiters = 0;
1699 wakeup(nd_defrouter_waitchan);
1700 }
1701 }
1702
1703 static struct nd_defrouter *
1704 defrtrlist_update_common(struct nd_defrouter *new, boolean_t scoped)
1705 {
1706 struct nd_defrouter *dr, *n;
1707 struct ifnet *ifp = new->ifp;
1708 struct nd_ifinfo *ndi = NULL;
1709 struct timeval caltime;
1710
1711 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1712
1713 if ((dr = defrouter_lookup(&new->rtaddr, ifp)) != NULL) {
1714 /* entry exists */
1715 if (new->rtlifetime == 0) {
1716 defrtrlist_del(dr);
1717 NDDR_REMREF(dr);
1718 dr = NULL;
1719 } else {
1720 int oldpref = rtpref(dr);
1721
1722 /* override */
1723 dr->flags = new->flags; /* xxx flag check */
1724 dr->rtlifetime = new->rtlifetime;
1725 dr->expire = new->expire;
1726
1727 /*
1728 * If the preference does not change, there's no need
1729 * to sort the entries. If Scoped Routing is enabled,
1730 * put the primary/non-scoped router at the top of the
1731 * list of routers in the same preference band, unless
1732 * it's already at that position.
1733 */
1734 if (ip6_doscopedroute) {
1735 struct nd_defrouter *p = NULL;
1736
1737 /* same preference and scoped; just return */
1738 if (rtpref(new) == oldpref && scoped)
1739 return (dr);
1740
1741 n = TAILQ_FIRST(&nd_defrouter);
1742 while (n != NULL) {
1743 /* preference changed; sort it */
1744 if (rtpref(new) != oldpref)
1745 break;
1746
1747 /* not at the top of band; sort it */
1748 if (n != dr && rtpref(n) == oldpref &&
1749 (!p || rtpref(p) > rtpref(n)))
1750 break;
1751
1752 p = n;
1753 n = TAILQ_NEXT(n, dr_entry);
1754 }
1755
1756 /* nothing has changed, just return */
1757 if (n == NULL && (scoped ||
1758 !(dr->stateflags & NDDRF_IFSCOPE)))
1759 return (dr);
1760 } else if (rtpref(new) == oldpref) {
1761 return (dr);
1762 }
1763
1764 /*
1765 * preferred router may be changed, so relocate
1766 * this router.
1767 * XXX: calling TAILQ_REMOVE directly is a bad manner.
1768 * However, since defrtrlist_del() has many side
1769 * effects, we intentionally do so here.
1770 * defrouter_select() below will handle routing
1771 * changes later.
1772 */
1773 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1774 new->stateflags = dr->stateflags;
1775 new->stateflags &= ~NDDRF_PROCESSED;
1776
1777 n = dr;
1778 goto insert;
1779 }
1780 return (dr);
1781 }
1782
1783 VERIFY(dr == NULL);
1784
1785 /* entry does not exist */
1786 if (new->rtlifetime == 0) {
1787 return (NULL);
1788 }
1789
1790 n = nddr_alloc(M_WAITOK);
1791 if (n == NULL) {
1792 return (NULL);
1793 }
1794
1795 lck_rw_lock_shared(nd_if_rwlock);
1796 if (ifp->if_index >= nd_ifinfo_indexlim)
1797 goto freeit;
1798 ndi = &nd_ifinfo[ifp->if_index];
1799 VERIFY(ndi->initialized);
1800 lck_mtx_lock(&ndi->lock);
1801 if (ip6_maxifdefrouters >= 0 &&
1802 ndi->ndefrouters >= ip6_maxifdefrouters) {
1803 lck_mtx_unlock(&ndi->lock);
1804 freeit:
1805 lck_rw_done(nd_if_rwlock);
1806 nddr_free(n);
1807 return (NULL);
1808 }
1809
1810 NDDR_ADDREF(n); /* for the nd_defrouter list */
1811 NDDR_ADDREF(n); /* for the caller */
1812
1813 ++nd6_defrouter_genid;
1814 ndi->ndefrouters++;
1815 VERIFY(ndi->ndefrouters != 0);
1816 lck_mtx_unlock(&ndi->lock);
1817 lck_rw_done(nd_if_rwlock);
1818
1819 nd6log2((LOG_INFO, "%s: allocating defrouter %s\n", if_name(ifp),
1820 ip6_sprintf(&new->rtaddr)));
1821
1822 getmicrotime(&caltime);
1823 NDDR_LOCK(n);
1824 memcpy(&n->rtaddr, &new->rtaddr, sizeof (n->rtaddr));
1825 n->flags = new->flags;
1826 n->stateflags = new->stateflags;
1827 n->stateflags &= ~NDDRF_PROCESSED;
1828 n->rtlifetime = new->rtlifetime;
1829 n->expire = new->expire;
1830 n->base_calendartime = caltime.tv_sec;
1831 n->base_uptime = net_uptime();
1832 n->ifp = new->ifp;
1833 n->genid = new->genid;
1834 n->err = new->err;
1835 NDDR_UNLOCK(n);
1836 insert:
1837 /* get nd6_service() to be scheduled as soon as it's convenient */
1838 ++nd6_sched_timeout_want;
1839
1840 /*
1841 * Insert the new router in the Default Router List;
1842 * The Default Router List should be in the descending order
1843 * of router-preferece. When Scoped Routing is disabled, routers
1844 * with the same preference are sorted in the arriving time order;
1845 * otherwise, the first entry in the list of routers having the same
1846 * preference is the primary default router, when the interface used
1847 * by the entry is the default interface.
1848 */
1849
1850 /* insert at the end of the group */
1851 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1852 dr = TAILQ_NEXT(dr, dr_entry)) {
1853 if (rtpref(n) > rtpref(dr) ||
1854 (ip6_doscopedroute && !scoped && rtpref(n) == rtpref(dr)))
1855 break;
1856 }
1857 if (dr)
1858 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
1859 else
1860 TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
1861
1862 defrouter_select(ifp);
1863
1864 return (n);
1865 }
1866
1867 static struct nd_defrouter *
1868 defrtrlist_update(struct nd_defrouter *new)
1869 {
1870 struct nd_defrouter *dr;
1871
1872 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1873 dr = defrtrlist_update_common(new,
1874 (nd6_defifp != NULL && new->ifp != nd6_defifp));
1875
1876 return (dr);
1877 }
1878
1879 static void
1880 defrtrlist_sync(struct ifnet *ifp)
1881 {
1882 struct nd_defrouter *dr, new;
1883
1884 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1885
1886 if (!ip6_doscopedroute) {
1887 defrouter_select(ifp);
1888 return;
1889 }
1890
1891 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1892 dr = TAILQ_NEXT(dr, dr_entry)) {
1893 NDDR_LOCK(dr);
1894 if (dr->ifp == ifp && (dr->stateflags & NDDRF_INSTALLED))
1895 break;
1896 NDDR_UNLOCK(dr);
1897 }
1898
1899 if (dr == NULL) {
1900 defrouter_select(ifp);
1901 } else {
1902 memcpy(&new.rtaddr, &dr->rtaddr, sizeof (new.rtaddr));
1903 new.flags = dr->flags;
1904 new.stateflags = dr->stateflags;
1905 new.stateflags &= ~NDDRF_PROCESSED;
1906 new.rtlifetime = dr->rtlifetime;
1907 new.expire = dr->expire;
1908 new.ifp = dr->ifp;
1909 new.genid = dr->genid;
1910 new.err = dr->err;
1911 NDDR_UNLOCK(dr);
1912 dr = defrtrlist_update_common(&new, FALSE);
1913 if (dr)
1914 NDDR_REMREF(dr);
1915 }
1916 }
1917
1918 static struct nd_pfxrouter *
1919 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
1920 {
1921 struct nd_pfxrouter *search;
1922
1923 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1924 NDPR_LOCK_ASSERT_HELD(pr);
1925
1926 for (search = pr->ndpr_advrtrs.lh_first; search;
1927 search = search->pfr_next) {
1928 if (search->router == dr)
1929 break;
1930 }
1931
1932 return (search);
1933 }
1934
1935 static void
1936 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
1937 {
1938 struct nd_pfxrouter *new;
1939
1940 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1941 NDPR_LOCK_ASSERT_NOTHELD(pr);
1942
1943 new = zalloc(ndprtr_zone);
1944 if (new == NULL)
1945 return;
1946 bzero(new, sizeof (*new));
1947 new->router = dr;
1948
1949 NDPR_LOCK(pr);
1950 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
1951 pr->ndpr_genid++;
1952 NDPR_UNLOCK(pr);
1953
1954 pfxlist_onlink_check();
1955 }
1956
1957 static void
1958 pfxrtr_del(struct nd_pfxrouter *pfr, struct nd_prefix *pr)
1959 {
1960 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
1961 NDPR_LOCK_ASSERT_HELD(pr);
1962 pr->ndpr_genid++;
1963 LIST_REMOVE(pfr, pfr_entry);
1964 zfree(ndprtr_zone, pfr);
1965 }
1966
1967 struct nd_prefix *
1968 nd6_prefix_lookup(struct nd_prefix *pr)
1969 {
1970 struct nd_prefix *search;
1971
1972 lck_mtx_lock(nd6_mutex);
1973 for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
1974 NDPR_LOCK(search);
1975 if (pr->ndpr_ifp == search->ndpr_ifp &&
1976 pr->ndpr_plen == search->ndpr_plen &&
1977 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1978 &search->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1979 NDPR_ADDREF_LOCKED(search);
1980 NDPR_UNLOCK(search);
1981 break;
1982 }
1983 NDPR_UNLOCK(search);
1984 }
1985 lck_mtx_unlock(nd6_mutex);
1986
1987 return (search);
1988 }
1989
1990 static void
1991 purge_detached(struct ifnet *ifp)
1992 {
1993 struct nd_prefix *pr, *pr_next;
1994 struct in6_ifaddr *ia;
1995 struct ifaddr *ifa, *ifa_next;
1996 boolean_t removed = FALSE;
1997
1998 lck_mtx_lock(nd6_mutex);
1999
2000 pr = nd_prefix.lh_first;
2001 repeat:
2002 while (pr) {
2003 NDPR_LOCK(pr);
2004 pr_next = pr->ndpr_next;
2005 if (pr->ndpr_ifp != ifp ||
2006 IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
2007 ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
2008 !LIST_EMPTY(&pr->ndpr_advrtrs))) {
2009 NDPR_UNLOCK(pr);
2010 pr = pr_next;
2011 continue;
2012 }
2013 NDPR_ADDREF_LOCKED(pr);
2014 NDPR_UNLOCK(pr);
2015 ifnet_lock_shared(ifp);
2016 for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa_next) {
2017 IFA_LOCK(ifa);
2018 ifa_next = ifa->ifa_list.tqe_next;
2019 if (ifa->ifa_addr->sa_family != AF_INET6) {
2020 IFA_UNLOCK(ifa);
2021 continue;
2022 }
2023 ia = (struct in6_ifaddr *)ifa;
2024 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
2025 IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
2026 IFA_ADDREF_LOCKED(ifa); /* for us */
2027 IFA_UNLOCK(ifa);
2028 /*
2029 * Purging the address requires writer access
2030 * to the address list, so drop the ifnet lock
2031 * now and repeat from beginning.
2032 */
2033 ifnet_lock_done(ifp);
2034 lck_mtx_unlock(nd6_mutex);
2035 in6_purgeaddr(ifa);
2036 IFA_REMREF(ifa); /* drop ours */
2037 lck_mtx_lock(nd6_mutex);
2038 NDPR_REMREF(pr);
2039 pr = nd_prefix.lh_first;
2040 goto repeat;
2041 }
2042 IFA_UNLOCK(ifa);
2043 }
2044 ifnet_lock_done(ifp);
2045 NDPR_LOCK(pr);
2046 if (pr->ndpr_addrcnt == 0 &&
2047 !(pr->ndpr_stateflags & NDPRF_DEFUNCT)) {
2048 prelist_remove(pr);
2049 NDPR_UNLOCK(pr);
2050 removed = TRUE;
2051 /*
2052 * Reset the search from the beginning because
2053 * nd6_mutex may have been dropped in
2054 * prelist_remove().
2055 */
2056 pr_next = nd_prefix.lh_first;
2057 } else {
2058 NDPR_UNLOCK(pr);
2059 }
2060 NDPR_REMREF(pr);
2061 pr = pr_next;
2062 }
2063 if (removed)
2064 pfxlist_onlink_check();
2065 lck_mtx_unlock(nd6_mutex);
2066 }
2067
2068 int
2069 nd6_prelist_add(struct nd_prefix *pr, struct nd_defrouter *dr,
2070 struct nd_prefix **newp, boolean_t force_scoped)
2071 {
2072 struct nd_prefix *new = NULL;
2073 struct ifnet *ifp = pr->ndpr_ifp;
2074 struct nd_ifinfo *ndi = NULL;
2075 int i, error;
2076
2077 if (ip6_maxifprefixes >= 0) {
2078 lck_rw_lock_shared(nd_if_rwlock);
2079 if (ifp->if_index >= nd_ifinfo_indexlim) {
2080 lck_rw_done(nd_if_rwlock);
2081 return (EINVAL);
2082 }
2083 ndi = &nd_ifinfo[ifp->if_index];
2084 VERIFY(ndi->initialized);
2085 lck_mtx_lock(&ndi->lock);
2086 if (ndi->nprefixes >= ip6_maxifprefixes / 2) {
2087 lck_mtx_unlock(&ndi->lock);
2088 lck_rw_done(nd_if_rwlock);
2089 purge_detached(ifp);
2090 lck_rw_lock_shared(nd_if_rwlock);
2091 /*
2092 * Refresh pointer since nd_ifinfo[] may have grown;
2093 * repeating the bounds check against nd_ifinfo_indexlim
2094 * isn't necessary since the array never shrinks.
2095 */
2096 ndi = &nd_ifinfo[ifp->if_index];
2097 lck_mtx_lock(&ndi->lock);
2098 }
2099 if (ndi->nprefixes >= ip6_maxifprefixes) {
2100 lck_mtx_unlock(&ndi->lock);
2101 lck_rw_done(nd_if_rwlock);
2102 return (ENOMEM);
2103 }
2104 lck_mtx_unlock(&ndi->lock);
2105 lck_rw_done(nd_if_rwlock);
2106 }
2107
2108 new = ndpr_alloc(M_WAITOK);
2109 if (new == NULL)
2110 return (ENOMEM);
2111
2112 NDPR_LOCK(new);
2113 NDPR_LOCK(pr);
2114 new->ndpr_ifp = pr->ndpr_ifp;
2115 new->ndpr_prefix = pr->ndpr_prefix;
2116 new->ndpr_plen = pr->ndpr_plen;
2117 new->ndpr_vltime = pr->ndpr_vltime;
2118 new->ndpr_pltime = pr->ndpr_pltime;
2119 new->ndpr_flags = pr->ndpr_flags;
2120 if (pr->ndpr_stateflags & NDPRF_STATIC)
2121 new->ndpr_stateflags |= NDPRF_STATIC;
2122 NDPR_UNLOCK(pr);
2123 if ((error = in6_init_prefix_ltimes(new)) != 0) {
2124 NDPR_UNLOCK(new);
2125 ndpr_free(new);
2126 return (error);
2127 }
2128 new->ndpr_lastupdate = net_uptime();
2129 if (newp != NULL) {
2130 *newp = new;
2131 NDPR_ADDREF_LOCKED(new); /* for caller */
2132 }
2133 /* initialization */
2134 LIST_INIT(&new->ndpr_advrtrs);
2135 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
2136 /* make prefix in the canonical form */
2137 for (i = 0; i < 4; i++)
2138 new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
2139 new->ndpr_mask.s6_addr32[i];
2140
2141 NDPR_UNLOCK(new);
2142
2143 /* get nd6_service() to be scheduled as soon as it's convenient */
2144 ++nd6_sched_timeout_want;
2145
2146 lck_mtx_lock(nd6_mutex);
2147 /* link ndpr_entry to nd_prefix list */
2148 LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
2149 new->ndpr_debug |= IFD_ATTACHED;
2150 NDPR_ADDREF(new); /* for nd_prefix list */
2151
2152 /* ND_OPT_PI_FLAG_ONLINK processing */
2153 if (new->ndpr_raf_onlink) {
2154 int e;
2155
2156 if ((e = nd6_prefix_onlink_common(new, force_scoped,
2157 new->ndpr_ifp->if_index)) != 0) {
2158 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
2159 "the prefix %s/%d on-link %s on %s (errno=%d)\n",
2160 ip6_sprintf(&new->ndpr_prefix.sin6_addr),
2161 new->ndpr_plen, force_scoped ? "scoped" :
2162 "non-scoped", if_name(ifp), e));
2163 /* proceed anyway. XXX: is it correct? */
2164 }
2165 }
2166
2167 if (dr) {
2168 pfxrtr_add(new, dr);
2169 }
2170
2171 lck_rw_lock_shared(nd_if_rwlock);
2172 /*
2173 * Refresh pointer since nd_ifinfo[] may have grown;
2174 * repeating the bounds check against nd_ifinfo_indexlim
2175 * isn't necessary since the array never shrinks.
2176 */
2177 ndi = &nd_ifinfo[ifp->if_index];
2178 VERIFY(ndi->initialized);
2179 lck_mtx_lock(&ndi->lock);
2180 ndi->nprefixes++;
2181 VERIFY(ndi->nprefixes != 0);
2182 lck_mtx_unlock(&ndi->lock);
2183 lck_rw_done(nd_if_rwlock);
2184
2185 lck_mtx_unlock(nd6_mutex);
2186
2187 return (0);
2188 }
2189
2190 /*
2191 * Caller must have held an extra reference on nd_prefix.
2192 */
2193 void
2194 prelist_remove(struct nd_prefix *pr)
2195 {
2196 struct nd_pfxrouter *pfr, *next;
2197 struct ifnet *ifp = pr->ndpr_ifp;
2198 int e;
2199
2200 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
2201 NDPR_LOCK_ASSERT_HELD(pr);
2202
2203 if (pr->ndpr_stateflags & NDPRF_DEFUNCT)
2204 return;
2205
2206 /*
2207 * If there are no more addresses, defunct the prefix. This is needed
2208 * because we don't want multiple threads calling prelist_remove() for
2209 * the same prefix and this might happen because we unlock nd6_mutex
2210 * down below.
2211 */
2212 if (pr->ndpr_addrcnt == 0)
2213 pr->ndpr_stateflags |= NDPRF_DEFUNCT;
2214
2215 /* make sure to invalidate the prefix until it is really freed. */
2216 pr->ndpr_vltime = 0;
2217 pr->ndpr_pltime = 0;
2218
2219 /*
2220 * Though these flags are now meaningless, we'd rather keep the value
2221 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
2222 * when executing "ndp -p".
2223 */
2224 if (pr->ndpr_stateflags & NDPRF_ONLINK) {
2225 NDPR_ADDREF_LOCKED(pr);
2226 NDPR_UNLOCK(pr);
2227 lck_mtx_unlock(nd6_mutex);
2228 if ((e = nd6_prefix_offlink(pr)) != 0) {
2229 nd6log((LOG_ERR, "prelist_remove: failed to make "
2230 "%s/%d offlink on %s, errno=%d\n",
2231 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
2232 pr->ndpr_plen, if_name(ifp), e));
2233 /* what should we do? */
2234 }
2235 lck_mtx_lock(nd6_mutex);
2236 NDPR_LOCK(pr);
2237 if (NDPR_REMREF_LOCKED(pr) == NULL)
2238 return;
2239 }
2240
2241 if (pr->ndpr_addrcnt > 0) {
2242 /*
2243 * The state might have changed if we called
2244 * nd6_prefix_offlink().
2245 */
2246 pr->ndpr_stateflags &= ~NDPRF_DEFUNCT;
2247 return; /* notice here? */
2248 }
2249
2250 /* unlink ndpr_entry from nd_prefix list */
2251 LIST_REMOVE(pr, ndpr_entry);
2252 pr->ndpr_debug &= ~IFD_ATTACHED;
2253
2254 /* free list of routers that adversed the prefix */
2255 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
2256 next = pfr->pfr_next;
2257 pfxrtr_del(pfr, pr);
2258 }
2259
2260 lck_rw_lock_shared(nd_if_rwlock);
2261 if (ifp->if_index < nd_ifinfo_indexlim) {
2262 struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
2263 VERIFY(ndi->initialized);
2264 lck_mtx_lock(&ndi->lock);
2265 VERIFY(ndi->nprefixes > 0);
2266 ndi->nprefixes--;
2267 lck_mtx_unlock(&ndi->lock);
2268 }
2269 lck_rw_done(nd_if_rwlock);
2270
2271 /* This must not be the last reference to the nd_prefix */
2272 if (NDPR_REMREF_LOCKED(pr) == NULL) {
2273 panic("%s: unexpected (missing) refcnt ndpr=%p", __func__, pr);
2274 /* NOTREACHED */
2275 }
2276
2277 /*
2278 * Don't call pfxlist_onlink_check() here because we are
2279 * holding the NDPR lock and this could cause a deadlock when
2280 * there are multiple threads executing pfxlist_onlink_check().
2281 */
2282 }
2283
2284 int
2285 prelist_update(
2286 struct nd_prefix *new,
2287 struct nd_defrouter *dr, /* may be NULL */
2288 struct mbuf *m,
2289 int mcast)
2290 {
2291 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
2292 struct ifaddr *ifa;
2293 struct ifnet *ifp = new->ndpr_ifp;
2294 struct nd_prefix *pr;
2295 int error = 0;
2296 int newprefix = 0;
2297 int auth;
2298 struct in6_addrlifetime lt6_tmp;
2299 uint64_t timenow = net_uptime();
2300
2301 /* no need to lock "new" here, as it is local to the caller */
2302 NDPR_LOCK_ASSERT_NOTHELD(new);
2303
2304 auth = 0;
2305 if (m) {
2306 /*
2307 * Authenticity for NA consists authentication for
2308 * both IP header and IP datagrams, doesn't it ?
2309 */
2310 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
2311 auth = (m->m_flags & M_AUTHIPHDR) && (m->m_flags & M_AUTHIPDGM);
2312 #endif
2313 }
2314
2315 if ((pr = nd6_prefix_lookup(new)) != NULL) {
2316 /*
2317 * nd6_prefix_lookup() ensures that pr and new have the same
2318 * prefix on a same interface.
2319 */
2320
2321 /*
2322 * Update prefix information. Note that the on-link (L) bit
2323 * and the autonomous (A) bit should NOT be changed from 1
2324 * to 0.
2325 */
2326 lck_mtx_lock(nd6_mutex);
2327 NDPR_LOCK(pr);
2328 if (new->ndpr_raf_onlink == 1)
2329 pr->ndpr_raf_onlink = 1;
2330 if (new->ndpr_raf_auto == 1)
2331 pr->ndpr_raf_auto = 1;
2332 if (new->ndpr_raf_onlink) {
2333 pr->ndpr_vltime = new->ndpr_vltime;
2334 pr->ndpr_pltime = new->ndpr_pltime;
2335 (void) in6_init_prefix_ltimes(pr); /* XXX error case? */
2336 pr->ndpr_lastupdate = net_uptime();
2337 }
2338
2339 NDPR_ADDREF_LOCKED(pr);
2340 if (new->ndpr_raf_onlink &&
2341 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
2342 int e;
2343
2344 NDPR_UNLOCK(pr);
2345 if ((e = nd6_prefix_onlink(pr)) != 0) {
2346 nd6log((LOG_ERR,
2347 "prelist_update: failed to make "
2348 "the prefix %s/%d on-link on %s "
2349 "(errno=%d)\n",
2350 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
2351 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
2352 /* proceed anyway. XXX: is it correct? */
2353 }
2354 NDPR_LOCK(pr);
2355 }
2356
2357 if (dr && pfxrtr_lookup(pr, dr) == NULL) {
2358 NDPR_UNLOCK(pr);
2359 pfxrtr_add(pr, dr);
2360 } else {
2361 NDPR_UNLOCK(pr);
2362 }
2363 NDPR_REMREF(pr);
2364 lck_mtx_unlock(nd6_mutex);
2365 } else {
2366 struct nd_prefix *newpr = NULL;
2367
2368 newprefix = 1;
2369
2370 if (new->ndpr_vltime == 0)
2371 goto end;
2372 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
2373 goto end;
2374
2375 bzero(&new->ndpr_addr, sizeof (struct in6_addr));
2376
2377 error = nd6_prelist_add(new, dr, &newpr, FALSE);
2378 if (error != 0 || newpr == NULL) {
2379 nd6log((LOG_NOTICE, "prelist_update: "
2380 "nd6_prelist_add failed for %s/%d on %s "
2381 "errno=%d, returnpr=0x%llx\n",
2382 ip6_sprintf(&new->ndpr_prefix.sin6_addr),
2383 new->ndpr_plen, if_name(new->ndpr_ifp),
2384 error, (uint64_t)VM_KERNEL_ADDRPERM(newpr)));
2385 goto end; /* we should just give up in this case. */
2386 }
2387
2388 /*
2389 * XXX: from the ND point of view, we can ignore a prefix
2390 * with the on-link bit being zero. However, we need a
2391 * prefix structure for references from autoconfigured
2392 * addresses. Thus, we explicitly make sure that the prefix
2393 * itself expires now.
2394 */
2395 NDPR_LOCK(newpr);
2396 if (newpr->ndpr_raf_onlink == 0) {
2397 newpr->ndpr_vltime = 0;
2398 newpr->ndpr_pltime = 0;
2399 in6_init_prefix_ltimes(newpr);
2400 }
2401
2402 pr = newpr;
2403 NDPR_UNLOCK(newpr);
2404 }
2405
2406 /*
2407 * Address autoconfiguration based on Section 5.5.3 of RFC 4862.
2408 * Note that pr must be non NULL at this point.
2409 */
2410
2411 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
2412 if (!new->ndpr_raf_auto)
2413 goto afteraddrconf;
2414
2415 /*
2416 * 5.5.3 (b). the link-local prefix should have been ignored in
2417 * nd6_ra_input.
2418 */
2419
2420 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
2421 if (new->ndpr_pltime > new->ndpr_vltime) {
2422 error = EINVAL; /* XXX: won't be used */
2423 goto end;
2424 }
2425
2426 /*
2427 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
2428 * an address configured by stateless autoconfiguration already in the
2429 * list of addresses associated with the interface, and the Valid
2430 * Lifetime is not 0, form an address. We first check if we have
2431 * a matching prefix.
2432 */
2433 ifnet_lock_shared(ifp);
2434 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2435 struct in6_ifaddr *ifa6;
2436 u_int32_t remaininglifetime;
2437
2438 IFA_LOCK(ifa);
2439 if (ifa->ifa_addr->sa_family != AF_INET6) {
2440 IFA_UNLOCK(ifa);
2441 continue;
2442 }
2443 ifa6 = (struct in6_ifaddr *)ifa;
2444
2445 /*
2446 * We only consider autoconfigured addresses as per RFC 4862.
2447 */
2448 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF)) {
2449 IFA_UNLOCK(ifa);
2450 continue;
2451 }
2452 /*
2453 * Spec is not clear here, but I believe we should concentrate
2454 * on unicast (i.e. not anycast) addresses.
2455 * XXX: other ia6_flags? detached or duplicated?
2456 */
2457 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0) {
2458 IFA_UNLOCK(ifa);
2459 continue;
2460 }
2461 /*
2462 * Ignore the address if it is not associated with a prefix
2463 * or is associated with a prefix that is different from this
2464 * one. (pr is never NULL here)
2465 */
2466 if (ifa6->ia6_ndpr != pr) {
2467 IFA_UNLOCK(ifa);
2468 continue;
2469 }
2470
2471 if (ia6_match == NULL) { /* remember the first one */
2472 ia6_match = ifa6;
2473 IFA_ADDREF_LOCKED(ifa); /* for ia6_match */
2474 }
2475
2476 /*
2477 * An already autoconfigured address matched. Now that we
2478 * are sure there is at least one matched address, we can
2479 * proceed to 5.5.3. (e): update the lifetimes according to the
2480 * "two hours" rule and the privacy extension.
2481 */
2482 #define TWOHOUR (120*60)
2483
2484 /* retrieve time as uptime (last arg is 0) */
2485 in6ifa_getlifetime(ifa6, &lt6_tmp, 0);
2486
2487 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
2488 remaininglifetime = ND6_INFINITE_LIFETIME;
2489 else if (timenow - ifa6->ia6_updatetime > lt6_tmp.ia6t_vltime) {
2490 /*
2491 * The case of "invalid" address. We should usually
2492 * not see this case.
2493 */
2494 remaininglifetime = 0;
2495 } else {
2496 remaininglifetime = lt6_tmp.ia6t_vltime -
2497 (timenow - ifa6->ia6_updatetime);
2498 }
2499 /* when not updating, keep the current stored lifetime. */
2500 lt6_tmp.ia6t_vltime = remaininglifetime;
2501
2502 if (TWOHOUR < new->ndpr_vltime ||
2503 remaininglifetime < new->ndpr_vltime) {
2504 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
2505 } else if (remaininglifetime <= TWOHOUR) {
2506 if (auth) {
2507 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
2508 }
2509 } else {
2510 /*
2511 * new->ndpr_vltime <= TWOHOUR &&
2512 * TWOHOUR < remaininglifetime
2513 */
2514 lt6_tmp.ia6t_vltime = TWOHOUR;
2515 }
2516
2517 /* The 2 hour rule is not imposed for preferred lifetime. */
2518 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
2519
2520 /* Special handling for lifetimes of temporary addresses. */
2521 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
2522 u_int32_t maxvltime, maxpltime;
2523
2524 /* Constrain lifetimes to system limits. */
2525 if (lt6_tmp.ia6t_vltime > ip6_temp_valid_lifetime)
2526 lt6_tmp.ia6t_vltime = ip6_temp_valid_lifetime;
2527 if (lt6_tmp.ia6t_pltime > ip6_temp_preferred_lifetime)
2528 lt6_tmp.ia6t_pltime =
2529 ip6_temp_preferred_lifetime -
2530 ip6_desync_factor;
2531
2532 /*
2533 * According to RFC 4941, section 3.3 (1), we only
2534 * update the lifetimes when they are in the maximum
2535 * intervals.
2536 */
2537 if (ip6_temp_valid_lifetime >
2538 (u_int32_t)((timenow - ifa6->ia6_createtime) +
2539 ip6_desync_factor)) {
2540 maxvltime = ip6_temp_valid_lifetime -
2541 (timenow - ifa6->ia6_createtime) -
2542 ip6_desync_factor;
2543 } else
2544 maxvltime = 0;
2545 if (ip6_temp_preferred_lifetime >
2546 (u_int32_t)((timenow - ifa6->ia6_createtime) +
2547 ip6_desync_factor)) {
2548 maxpltime = ip6_temp_preferred_lifetime -
2549 (timenow - ifa6->ia6_createtime) -
2550 ip6_desync_factor;
2551 } else
2552 maxpltime = 0;
2553
2554 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
2555 lt6_tmp.ia6t_vltime > maxvltime)
2556 lt6_tmp.ia6t_vltime = maxvltime;
2557
2558 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
2559 lt6_tmp.ia6t_pltime > maxpltime)
2560 lt6_tmp.ia6t_pltime = maxpltime;
2561 }
2562
2563 in6_init_address_ltimes(pr, &lt6_tmp);
2564
2565 in6ifa_setlifetime(ifa6, &lt6_tmp);
2566 ifa6->ia6_updatetime = timenow;
2567 IFA_UNLOCK(ifa);
2568 }
2569 ifnet_lock_done(ifp);
2570 if (ia6_match == NULL && new->ndpr_vltime) {
2571 /*
2572 * 5.5.3 (d) (continued)
2573 * No address matched and the valid lifetime is non-zero.
2574 * Create a new address.
2575 */
2576
2577 if ((ia6 = in6_pfx_newpersistaddr(new, mcast, &error))
2578 != NULL) {
2579 /*
2580 * note that we should use pr (not new) for reference.
2581 */
2582 IFA_LOCK(&ia6->ia_ifa);
2583 NDPR_LOCK(pr);
2584 ia6->ia6_ndpr = pr;
2585 NDPR_ADDREF_LOCKED(pr); /* for addr reference */
2586 pr->ndpr_addrcnt++;
2587 VERIFY(pr->ndpr_addrcnt != 0);
2588 NDPR_UNLOCK(pr);
2589 IFA_UNLOCK(&ia6->ia_ifa);
2590
2591 /*
2592 * RFC 4941 3.3 (2).
2593 * When a new public address is created as described
2594 * in RFC 4862, also create a new temporary address.
2595 *
2596 * RFC 4941 3.5.
2597 * When an interface connects to a new link, a new
2598 * randomized interface identifier should be generated
2599 * immediately together with a new set of temporary
2600 * addresses. Thus, we specifiy 1 as the 2nd arg of
2601 * in6_tmpifadd().
2602 */
2603 if (ip6_use_tempaddr) {
2604 int e;
2605 if ((e = in6_tmpifadd(ia6, 1)) != 0) {
2606 nd6log((LOG_NOTICE, "prelist_update: "
2607 "failed to create a temporary "
2608 "address, errno=%d\n",
2609 e));
2610 }
2611 }
2612 IFA_REMREF(&ia6->ia_ifa);
2613 ia6 = NULL;
2614
2615 /*
2616 * A newly added address might affect the status
2617 * of other addresses, so we check and update it.
2618 * XXX: what if address duplication happens?
2619 */
2620 lck_mtx_lock(nd6_mutex);
2621 pfxlist_onlink_check();
2622 lck_mtx_unlock(nd6_mutex);
2623 }
2624 }
2625
2626 afteraddrconf:
2627
2628 end:
2629 if (pr != NULL)
2630 NDPR_REMREF(pr);
2631 if (ia6_match != NULL)
2632 IFA_REMREF(&ia6_match->ia_ifa);
2633 return (error);
2634 }
2635
2636 /*
2637 * Neighbor Discover Default Router structure reference counting routines.
2638 */
2639 static struct nd_defrouter *
2640 nddr_alloc(int how)
2641 {
2642 struct nd_defrouter *dr;
2643
2644 dr = (how == M_WAITOK) ? zalloc(nddr_zone) : zalloc_noblock(nddr_zone);
2645 if (dr != NULL) {
2646 bzero(dr, nddr_size);
2647 lck_mtx_init(&dr->nddr_lock, ifa_mtx_grp, ifa_mtx_attr);
2648 dr->nddr_debug |= IFD_ALLOC;
2649 if (nddr_debug != 0) {
2650 dr->nddr_debug |= IFD_DEBUG;
2651 dr->nddr_trace = nddr_trace;
2652 }
2653 }
2654 return (dr);
2655 }
2656
2657 static void
2658 nddr_free(struct nd_defrouter *dr)
2659 {
2660 NDDR_LOCK(dr);
2661 if (dr->nddr_debug & IFD_ATTACHED) {
2662 panic("%s: attached nddr %p is being freed", __func__, dr);
2663 /* NOTREACHED */
2664 } else if (!(dr->nddr_debug & IFD_ALLOC)) {
2665 panic("%s: nddr %p cannot be freed", __func__, dr);
2666 /* NOTREACHED */
2667 }
2668 dr->nddr_debug &= ~IFD_ALLOC;
2669 NDDR_UNLOCK(dr);
2670
2671 lck_mtx_destroy(&dr->nddr_lock, ifa_mtx_grp);
2672 zfree(nddr_zone, dr);
2673 }
2674
2675 static void
2676 nddr_trace(struct nd_defrouter *dr, int refhold)
2677 {
2678 struct nd_defrouter_dbg *dr_dbg = (struct nd_defrouter_dbg *)dr;
2679 ctrace_t *tr;
2680 uint32_t idx;
2681 uint16_t *cnt;
2682
2683 if (!(dr->nddr_debug & IFD_DEBUG)) {
2684 panic("%s: nddr %p has no debug structure", __func__, dr);
2685 /* NOTREACHED */
2686 }
2687 if (refhold) {
2688 cnt = &dr_dbg->nddr_refhold_cnt;
2689 tr = dr_dbg->nddr_refhold;
2690 } else {
2691 cnt = &dr_dbg->nddr_refrele_cnt;
2692 tr = dr_dbg->nddr_refrele;
2693 }
2694
2695 idx = atomic_add_16_ov(cnt, 1) % NDDR_TRACE_HIST_SIZE;
2696 ctrace_record(&tr[idx]);
2697 }
2698
2699 void
2700 nddr_addref(struct nd_defrouter *nddr, int locked)
2701 {
2702
2703 if (!locked)
2704 NDDR_LOCK_SPIN(nddr);
2705 else
2706 NDDR_LOCK_ASSERT_HELD(nddr);
2707
2708 if (++nddr->nddr_refcount == 0) {
2709 panic("%s: nddr %p wraparound refcnt\n", __func__, nddr);
2710 /* NOTREACHED */
2711 } else if (nddr->nddr_trace != NULL) {
2712 (*nddr->nddr_trace)(nddr, TRUE);
2713 }
2714
2715 if (!locked)
2716 NDDR_UNLOCK(nddr);
2717 }
2718
2719 struct nd_defrouter *
2720 nddr_remref(struct nd_defrouter *nddr, int locked)
2721 {
2722
2723 if (!locked)
2724 NDDR_LOCK_SPIN(nddr);
2725 else
2726 NDDR_LOCK_ASSERT_HELD(nddr);
2727
2728 if (nddr->nddr_refcount == 0) {
2729 panic("%s: nddr %p negative refcnt\n", __func__, nddr);
2730 /* NOTREACHED */
2731 } else if (nddr->nddr_trace != NULL) {
2732 (*nddr->nddr_trace)(nddr, FALSE);
2733 }
2734
2735 if (--nddr->nddr_refcount == 0) {
2736 NDDR_UNLOCK(nddr);
2737 nddr_free(nddr);
2738 nddr = NULL;
2739 }
2740
2741 if (!locked && nddr != NULL)
2742 NDDR_UNLOCK(nddr);
2743
2744 return (nddr);
2745 }
2746
2747 uint64_t
2748 nddr_getexpire(struct nd_defrouter *dr)
2749 {
2750 struct timeval caltime;
2751 uint64_t expiry;
2752
2753 if (dr->expire != 0) {
2754 /* account for system time change */
2755 getmicrotime(&caltime);
2756
2757 dr->base_calendartime +=
2758 NET_CALCULATE_CLOCKSKEW(caltime,
2759 dr->base_calendartime, net_uptime(), dr->base_uptime);
2760
2761 expiry = dr->base_calendartime +
2762 dr->expire - dr->base_uptime;
2763 } else {
2764 expiry = 0;
2765 }
2766 return (expiry);
2767 }
2768
2769 /*
2770 * Neighbor Discover Prefix structure reference counting routines.
2771 */
2772 static struct nd_prefix *
2773 ndpr_alloc(int how)
2774 {
2775 struct nd_prefix *pr;
2776
2777 pr = (how == M_WAITOK) ? zalloc(ndpr_zone) : zalloc_noblock(ndpr_zone);
2778 if (pr != NULL) {
2779 bzero(pr, ndpr_size);
2780 lck_mtx_init(&pr->ndpr_lock, ifa_mtx_grp, ifa_mtx_attr);
2781 RB_INIT(&pr->ndpr_prproxy_sols);
2782 pr->ndpr_debug |= IFD_ALLOC;
2783 if (ndpr_debug != 0) {
2784 pr->ndpr_debug |= IFD_DEBUG;
2785 pr->ndpr_trace = ndpr_trace;
2786 }
2787 }
2788 return (pr);
2789 }
2790
2791 static void
2792 ndpr_free(struct nd_prefix *pr)
2793 {
2794 NDPR_LOCK(pr);
2795 if (pr->ndpr_debug & IFD_ATTACHED) {
2796 panic("%s: attached ndpr %p is being freed", __func__, pr);
2797 /* NOTREACHED */
2798 } else if (!(pr->ndpr_debug & IFD_ALLOC)) {
2799 panic("%s: ndpr %p cannot be freed", __func__, pr);
2800 /* NOTREACHED */
2801 } else if (pr->ndpr_rt != NULL) {
2802 panic("%s: ndpr %p route %p not freed", __func__, pr,
2803 pr->ndpr_rt);
2804 /* NOTREACHED */
2805 } else if (pr->ndpr_prproxy_sols_cnt != 0) {
2806 panic("%s: ndpr %p non-zero solicitors count (%d)",
2807 __func__, pr, pr->ndpr_prproxy_sols_cnt);
2808 /* NOTREACHED */
2809 } else if (!RB_EMPTY(&pr->ndpr_prproxy_sols)) {
2810 panic("%s: ndpr %p non-empty solicitors tree", __func__, pr);
2811 /* NOTREACHED */
2812 }
2813 pr->ndpr_debug &= ~IFD_ALLOC;
2814 NDPR_UNLOCK(pr);
2815
2816 lck_mtx_destroy(&pr->ndpr_lock, ifa_mtx_grp);
2817 zfree(ndpr_zone, pr);
2818 }
2819
2820 static void
2821 ndpr_trace(struct nd_prefix *pr, int refhold)
2822 {
2823 struct nd_prefix_dbg *pr_dbg = (struct nd_prefix_dbg *)pr;
2824 ctrace_t *tr;
2825 u_int32_t idx;
2826 u_int16_t *cnt;
2827
2828 if (!(pr->ndpr_debug & IFD_DEBUG)) {
2829 panic("%s: ndpr %p has no debug structure", __func__, pr);
2830 /* NOTREACHED */
2831 }
2832 if (refhold) {
2833 cnt = &pr_dbg->ndpr_refhold_cnt;
2834 tr = pr_dbg->ndpr_refhold;
2835 } else {
2836 cnt = &pr_dbg->ndpr_refrele_cnt;
2837 tr = pr_dbg->ndpr_refrele;
2838 }
2839
2840 idx = atomic_add_16_ov(cnt, 1) % NDPR_TRACE_HIST_SIZE;
2841 ctrace_record(&tr[idx]);
2842 }
2843
2844 void
2845 ndpr_addref(struct nd_prefix *ndpr, int locked)
2846 {
2847 if (!locked)
2848 NDPR_LOCK_SPIN(ndpr);
2849 else
2850 NDPR_LOCK_ASSERT_HELD(ndpr);
2851
2852 if (++ndpr->ndpr_refcount == 0) {
2853 panic("%s: ndpr %p wraparound refcnt\n", __func__, ndpr);
2854 /* NOTREACHED */
2855 } else if (ndpr->ndpr_trace != NULL) {
2856 (*ndpr->ndpr_trace)(ndpr, TRUE);
2857 }
2858
2859 if (!locked)
2860 NDPR_UNLOCK(ndpr);
2861 }
2862
2863 struct nd_prefix *
2864 ndpr_remref(struct nd_prefix *ndpr, int locked)
2865 {
2866 if (!locked)
2867 NDPR_LOCK_SPIN(ndpr);
2868 else
2869 NDPR_LOCK_ASSERT_HELD(ndpr);
2870
2871 if (ndpr->ndpr_refcount == 0) {
2872 panic("%s: ndpr %p negative refcnt\n", __func__, ndpr);
2873 /* NOTREACHED */
2874 } else if (ndpr->ndpr_trace != NULL) {
2875 (*ndpr->ndpr_trace)(ndpr, FALSE);
2876 }
2877
2878 if (--ndpr->ndpr_refcount == 0) {
2879 if (ndpr->ndpr_addrcnt != 0) {
2880 panic("%s: freeing ndpr %p with outstanding address "
2881 "reference (%d)", __func__, ndpr,
2882 ndpr->ndpr_addrcnt);
2883 /* NOTREACHED */
2884 }
2885 NDPR_UNLOCK(ndpr);
2886 ndpr_free(ndpr);
2887 ndpr = NULL;
2888 }
2889
2890 if (!locked && ndpr != NULL)
2891 NDPR_UNLOCK(ndpr);
2892
2893 return (ndpr);
2894 }
2895
2896 uint64_t
2897 ndpr_getexpire(struct nd_prefix *pr)
2898 {
2899 struct timeval caltime;
2900 uint64_t expiry;
2901
2902 if (pr->ndpr_expire != 0 && pr->ndpr_vltime != ND6_INFINITE_LIFETIME) {
2903 /* account for system time change */
2904 getmicrotime(&caltime);
2905
2906 pr->ndpr_base_calendartime +=
2907 NET_CALCULATE_CLOCKSKEW(caltime,
2908 pr->ndpr_base_calendartime, net_uptime(),
2909 pr->ndpr_base_uptime);
2910
2911 expiry = pr->ndpr_base_calendartime +
2912 pr->ndpr_expire - pr->ndpr_base_uptime;
2913 } else {
2914 expiry = 0;
2915 }
2916 return (expiry);
2917 }
2918
2919 /*
2920 * A supplement function used in the on-link detection below;
2921 * detect if a given prefix has a (probably) reachable advertising router.
2922 * XXX: lengthy function name...
2923 *
2924 * Callers *must* increase the reference count of nd_prefix.
2925 */
2926 static struct nd_pfxrouter *
2927 find_pfxlist_reachable_router(struct nd_prefix *pr)
2928 {
2929 struct nd_pfxrouter *pfxrtr;
2930 struct rtentry *rt;
2931 struct llinfo_nd6 *ln;
2932 struct ifnet *ifp;
2933 struct in6_addr rtaddr;
2934 unsigned int genid;
2935
2936 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
2937 NDPR_LOCK_ASSERT_HELD(pr);
2938
2939 genid = pr->ndpr_genid;
2940 pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs);
2941 while (pfxrtr) {
2942 ifp = pfxrtr->router->ifp;
2943 rtaddr = pfxrtr->router->rtaddr;
2944 NDPR_UNLOCK(pr);
2945 lck_mtx_unlock(nd6_mutex);
2946 /* Callee returns a locked route upon success */
2947 if ((rt = nd6_lookup(&rtaddr, 0, ifp, 0)) != NULL) {
2948 RT_LOCK_ASSERT_HELD(rt);
2949 if ((ln = rt->rt_llinfo) != NULL &&
2950 ND6_IS_LLINFO_PROBREACH(ln)) {
2951 RT_REMREF_LOCKED(rt);
2952 RT_UNLOCK(rt);
2953 lck_mtx_lock(nd6_mutex);
2954 NDPR_LOCK(pr);
2955 break; /* found */
2956 }
2957 RT_REMREF_LOCKED(rt);
2958 RT_UNLOCK(rt);
2959 }
2960 lck_mtx_lock(nd6_mutex);
2961 NDPR_LOCK(pr);
2962 if (pr->ndpr_genid != genid) {
2963 pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs);
2964 genid = pr->ndpr_genid;
2965 } else
2966 pfxrtr = LIST_NEXT(pfxrtr, pfr_entry);
2967 }
2968 NDPR_LOCK_ASSERT_HELD(pr);
2969
2970 return (pfxrtr);
2971
2972 }
2973
2974 /*
2975 * Check if each prefix in the prefix list has at least one available router
2976 * that advertised the prefix (a router is "available" if its neighbor cache
2977 * entry is reachable or probably reachable).
2978 * If the check fails, the prefix may be off-link, because, for example,
2979 * we have moved from the network but the lifetime of the prefix has not
2980 * expired yet. So we should not use the prefix if there is another prefix
2981 * that has an available router.
2982 * But, if there is no prefix that has an available router, we still regards
2983 * all the prefixes as on-link. This is because we can't tell if all the
2984 * routers are simply dead or if we really moved from the network and there
2985 * is no router around us.
2986 */
2987 void
2988 pfxlist_onlink_check(void)
2989 {
2990 struct nd_prefix *pr, *prclear;
2991 struct in6_ifaddr *ifa;
2992 struct nd_defrouter *dr;
2993 struct nd_pfxrouter *pfxrtr = NULL;
2994 int err, i, found = 0;
2995 struct ifaddr **ifap = NULL;
2996 struct nd_prefix *ndpr;
2997
2998 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
2999
3000 while (nd_prefix_busy) {
3001 nd_prefix_waiters++;
3002 msleep(nd_prefix_waitchan, nd6_mutex, (PZERO-1),
3003 __func__, NULL);
3004 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
3005 }
3006 nd_prefix_busy = TRUE;
3007
3008 /*
3009 * Check if there is a prefix that has a reachable advertising
3010 * router.
3011 */
3012 pr = nd_prefix.lh_first;
3013 while (pr) {
3014 NDPR_LOCK(pr);
3015 if (pr->ndpr_stateflags & NDPRF_PROCESSED_ONLINK) {
3016 NDPR_UNLOCK(pr);
3017 pr = pr->ndpr_next;
3018 continue;
3019 }
3020 NDPR_ADDREF_LOCKED(pr);
3021 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr) &&
3022 (pr->ndpr_debug & IFD_ATTACHED)) {
3023 if (NDPR_REMREF_LOCKED(pr) == NULL)
3024 pr = NULL;
3025 else
3026 NDPR_UNLOCK(pr);
3027 break;
3028 }
3029 pr->ndpr_stateflags |= NDPRF_PROCESSED_ONLINK;
3030 NDPR_UNLOCK(pr);
3031 NDPR_REMREF(pr);
3032 /*
3033 * Since find_pfxlist_reachable_router() drops the nd6_mutex, we
3034 * have to start over, but the NDPRF_PROCESSED_ONLINK flag will
3035 * stop us from checking the same prefix twice.
3036 */
3037 pr = nd_prefix.lh_first;
3038 }
3039 LIST_FOREACH(prclear, &nd_prefix, ndpr_entry) {
3040 NDPR_LOCK(prclear);
3041 prclear->ndpr_stateflags &= ~NDPRF_PROCESSED_ONLINK;
3042 NDPR_UNLOCK(prclear);
3043 }
3044 /*
3045 * If we have no such prefix, check whether we still have a router
3046 * that does not advertise any prefixes.
3047 */
3048 if (pr == NULL) {
3049 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
3050 dr = TAILQ_NEXT(dr, dr_entry)) {
3051 struct nd_prefix *pr0;
3052
3053 for (pr0 = nd_prefix.lh_first; pr0;
3054 pr0 = pr0->ndpr_next) {
3055 NDPR_LOCK(pr0);
3056 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL) {
3057 NDPR_UNLOCK(pr0);
3058 break;
3059 }
3060 NDPR_UNLOCK(pr0);
3061 }
3062 if (pfxrtr != NULL)
3063 break;
3064 }
3065 }
3066 if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && pfxrtr == NULL)) {
3067 /*
3068 * There is at least one prefix that has a reachable router,
3069 * or at least a router which probably does not advertise
3070 * any prefixes. The latter would be the case when we move
3071 * to a new link where we have a router that does not provide
3072 * prefixes and we configure an address by hand.
3073 * Detach prefixes which have no reachable advertising
3074 * router, and attach other prefixes.
3075 */
3076 pr = nd_prefix.lh_first;
3077 while (pr) {
3078 NDPR_LOCK(pr);
3079 /*
3080 * We aren't interested prefixes already processed,
3081 * nor in prefixes without the L bit
3082 * set nor in static prefixes
3083 */
3084 if (pr->ndpr_raf_onlink == 0 ||
3085 pr->ndpr_stateflags & NDPRF_PROCESSED_ONLINK ||
3086 pr->ndpr_stateflags & NDPRF_STATIC) {
3087 NDPR_UNLOCK(pr);
3088 pr = pr->ndpr_next;
3089 continue;
3090 }
3091 NDPR_ADDREF_LOCKED(pr);
3092 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
3093 find_pfxlist_reachable_router(pr) == NULL &&
3094 (pr->ndpr_debug & IFD_ATTACHED))
3095 pr->ndpr_stateflags |= NDPRF_DETACHED;
3096 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
3097 find_pfxlist_reachable_router(pr) != NULL &&
3098 (pr->ndpr_debug & IFD_ATTACHED))
3099 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
3100 pr->ndpr_stateflags |= NDPRF_PROCESSED_ONLINK;
3101 NDPR_UNLOCK(pr);
3102 NDPR_REMREF(pr);
3103 /*
3104 * Since find_pfxlist_reachable_router() drops the
3105 * nd6_mutex, we have to start over, but the
3106 * NDPRF_PROCESSED_ONLINK flag will stop us from
3107 * checking the same prefix twice.
3108 */
3109 pr = nd_prefix.lh_first;
3110 }
3111 } else {
3112 /* there is no prefix that has a reachable router */
3113 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
3114 NDPR_LOCK(pr);
3115 if (pr->ndpr_raf_onlink == 0 ||
3116 pr->ndpr_stateflags & NDPRF_STATIC) {
3117 NDPR_UNLOCK(pr);
3118 continue;
3119 }
3120 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
3121 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
3122 NDPR_UNLOCK(pr);
3123 }
3124 }
3125 LIST_FOREACH(prclear, &nd_prefix, ndpr_entry) {
3126 NDPR_LOCK(prclear);
3127 prclear->ndpr_stateflags &= ~NDPRF_PROCESSED_ONLINK;
3128 NDPR_UNLOCK(prclear);
3129 }
3130 /*
3131 * Remove each interface route associated with a (just) detached
3132 * prefix, and reinstall the interface route for a (just) attached
3133 * prefix. Note that all attempt of reinstallation does not
3134 * necessarily success, when a same prefix is shared among multiple
3135 * interfaces. Such cases will be handled in nd6_prefix_onlink,
3136 * so we don't have to care about them.
3137 */
3138 pr = nd_prefix.lh_first;
3139 while (pr) {
3140 int e;
3141
3142 NDPR_LOCK(pr);
3143 if (pr->ndpr_raf_onlink == 0 ||
3144 pr->ndpr_stateflags & NDPRF_STATIC ||
3145 pr->ndpr_stateflags & NDPRF_PROCESSED_ONLINK ||
3146 pr->ndpr_stateflags & NDPRF_DEFUNCT) {
3147 NDPR_UNLOCK(pr);
3148 pr = pr->ndpr_next;
3149 continue;
3150 }
3151 pr->ndpr_stateflags |= NDPRF_PROCESSED_ONLINK;
3152 NDPR_ADDREF_LOCKED(pr);
3153 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
3154 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
3155 NDPR_UNLOCK(pr);
3156 lck_mtx_unlock(nd6_mutex);
3157 if ((e = nd6_prefix_offlink(pr)) != 0) {
3158 nd6log((LOG_ERR,
3159 "pfxlist_onlink_check: failed to "
3160 "make %s/%d offlink, errno=%d\n",
3161 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3162 pr->ndpr_plen, e));
3163 }
3164 lck_mtx_lock(nd6_mutex);
3165 NDPR_REMREF(pr);
3166 pr = nd_prefix.lh_first;
3167 continue;
3168 }
3169 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
3170 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
3171 pr->ndpr_raf_onlink) {
3172 NDPR_UNLOCK(pr);
3173 if ((e = nd6_prefix_onlink(pr)) != 0) {
3174 nd6log((LOG_ERR,
3175 "pfxlist_onlink_check: failed to "
3176 "make %s/%d offlink, errno=%d\n",
3177 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3178 pr->ndpr_plen, e));
3179 }
3180 NDPR_REMREF(pr);
3181 pr = nd_prefix.lh_first;
3182 continue;
3183 } else {
3184 NDPR_UNLOCK(pr);
3185 }
3186 NDPR_REMREF(pr);
3187 pr = pr->ndpr_next;
3188 }
3189 LIST_FOREACH(prclear, &nd_prefix, ndpr_entry) {
3190 NDPR_LOCK(prclear);
3191 prclear->ndpr_stateflags &= ~NDPRF_PROCESSED_ONLINK;
3192 NDPR_UNLOCK(prclear);
3193 }
3194 VERIFY(nd_prefix_busy);
3195 nd_prefix_busy = FALSE;
3196 if (nd_prefix_waiters > 0) {
3197 nd_prefix_waiters = 0;
3198 wakeup(nd_prefix_waitchan);
3199 }
3200
3201 /*
3202 * Changes on the prefix status might affect address status as well.
3203 * Make sure that all addresses derived from an attached prefix are
3204 * attached, and that all addresses derived from a detached prefix are
3205 * detached. Note, however, that a manually configured address should
3206 * always be attached.
3207 * The precise detection logic is same as the one for prefixes.
3208 *
3209 * ifnet_get_address_list_family_internal() may fail due to memory
3210 * pressure, but we will eventually be called again when we receive
3211 * another NA, RA, or when the link status changes.
3212 */
3213 err = ifnet_get_address_list_family_internal(NULL, &ifap, AF_INET6, 0,
3214 M_NOWAIT, 0);
3215 if (err != 0 || ifap == NULL) {
3216 nd6log((LOG_ERR, "%s: ifnet_get_address_list_family_internal "
3217 "failed", __func__));
3218 return;
3219 }
3220 for (i = 0; ifap[i]; i++) {
3221 ifa = ifatoia6(ifap[i]);
3222 IFA_LOCK(&ifa->ia_ifa);
3223 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0 ||
3224 (ifap[i]->ifa_debug & IFD_ATTACHED) == 0) {
3225 IFA_UNLOCK(&ifa->ia_ifa);
3226 continue;
3227 }
3228 if ((ndpr = ifa->ia6_ndpr) == NULL) {
3229 /*
3230 * This can happen when we first configure the address
3231 * (i.e. the address exists, but the prefix does not).
3232 * XXX: complicated relationships...
3233 */
3234 IFA_UNLOCK(&ifa->ia_ifa);
3235 continue;
3236 }
3237 IFA_UNLOCK(&ifa->ia_ifa);
3238
3239 NDPR_LOCK(ndpr);
3240 NDPR_ADDREF_LOCKED(ndpr);
3241 if (find_pfxlist_reachable_router(ndpr)) {
3242 if (NDPR_REMREF_LOCKED(ndpr) == NULL) {
3243 found = 0;
3244 } else {
3245 NDPR_UNLOCK(ndpr);
3246 found = 1;
3247 }
3248 break;
3249 }
3250 NDPR_UNLOCK(ndpr);
3251 NDPR_REMREF(ndpr);
3252 }
3253 if (found) {
3254 for (i = 0; ifap[i]; i++) {
3255 ifa = ifatoia6(ifap[i]);
3256 IFA_LOCK(&ifa->ia_ifa);
3257 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0 ||
3258 (ifap[i]->ifa_debug & IFD_ATTACHED) == 0) {
3259 IFA_UNLOCK(&ifa->ia_ifa);
3260 continue;
3261 }
3262 if ((ndpr = ifa->ia6_ndpr) == NULL) {
3263 /* XXX: see above. */
3264 IFA_UNLOCK(&ifa->ia_ifa);
3265 continue;
3266 }
3267 IFA_UNLOCK(&ifa->ia_ifa);
3268 NDPR_LOCK(ndpr);
3269 NDPR_ADDREF_LOCKED(ndpr);
3270 if (find_pfxlist_reachable_router(ndpr)) {
3271 NDPR_UNLOCK(ndpr);
3272 IFA_LOCK(&ifa->ia_ifa);
3273 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
3274 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
3275 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
3276 IFA_UNLOCK(&ifa->ia_ifa);
3277 nd6_dad_start((struct ifaddr *)ifa, 0);
3278 } else {
3279 IFA_UNLOCK(&ifa->ia_ifa);
3280 }
3281 } else {
3282 NDPR_UNLOCK(ndpr);
3283 IFA_LOCK(&ifa->ia_ifa);
3284 ifa->ia6_flags |= IN6_IFF_DETACHED;
3285 IFA_UNLOCK(&ifa->ia_ifa);
3286 }
3287 NDPR_REMREF(ndpr);
3288 }
3289 } else {
3290 for (i = 0; ifap[i]; i++) {
3291 ifa = ifatoia6(ifap[i]);
3292 IFA_LOCK(&ifa->ia_ifa);
3293 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0) {
3294 IFA_UNLOCK(&ifa->ia_ifa);
3295 continue;
3296 }
3297 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
3298 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
3299 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
3300 IFA_UNLOCK(&ifa->ia_ifa);
3301 /* Do we need a delay in this case? */
3302 nd6_dad_start((struct ifaddr *)ifa, 0);
3303 } else {
3304 IFA_UNLOCK(&ifa->ia_ifa);
3305 }
3306 }
3307 }
3308 ifnet_free_address_list(ifap);
3309 }
3310
3311 static struct nd_prefix *
3312 nd6_prefix_equal_lookup(struct nd_prefix *pr, boolean_t primary_only)
3313 {
3314 struct nd_prefix *opr;
3315
3316 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
3317
3318 for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
3319 if (opr == pr)
3320 continue;
3321
3322 NDPR_LOCK(opr);
3323 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
3324 NDPR_UNLOCK(opr);
3325 continue;
3326 }
3327 if (opr->ndpr_plen == pr->ndpr_plen &&
3328 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
3329 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen) &&
3330 (!primary_only ||
3331 !(opr->ndpr_stateflags & NDPRF_IFSCOPE))) {
3332 NDPR_ADDREF_LOCKED(opr);
3333 NDPR_UNLOCK(opr);
3334 return (opr);
3335 }
3336 NDPR_UNLOCK(opr);
3337 }
3338 return (NULL);
3339 }
3340
3341 /*
3342 * Synchronize the interface routes of similar prefixes on different
3343 * interfaces; the one using the default interface would be (re)installed
3344 * as a primary/non-scoped entry, and the rest as scoped entri(es).
3345 */
3346 static void
3347 nd6_prefix_sync(struct ifnet *ifp)
3348 {
3349 struct nd_prefix *pr, *opr;
3350 int err = 0;
3351
3352 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
3353
3354 if (!ip6_doscopedroute || ifp == NULL)
3355 return;
3356
3357 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
3358 NDPR_LOCK(pr);
3359 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
3360 NDPR_UNLOCK(pr);
3361 continue;
3362 }
3363 if (pr->ndpr_ifp == ifp &&
3364 (pr->ndpr_stateflags & NDPRF_IFSCOPE) &&
3365 !IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) {
3366 NDPR_UNLOCK(pr);
3367 break;
3368 }
3369 NDPR_UNLOCK(pr);
3370 }
3371
3372 if (pr == NULL)
3373 return;
3374
3375 /* Remove conflicting entries */
3376 opr = nd6_prefix_equal_lookup(pr, TRUE);
3377 if (opr != NULL) {
3378 lck_mtx_unlock(nd6_mutex);
3379 err = nd6_prefix_offlink(opr);
3380 lck_mtx_lock(nd6_mutex);
3381 if (err != 0) {
3382 nd6log((LOG_ERR,
3383 "%s: failed to make %s/%d offlink on %s, "
3384 "errno=%d\n", __func__,
3385 ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
3386 opr->ndpr_plen, if_name(opr->ndpr_ifp), err));
3387 }
3388 } else {
3389 nd6log((LOG_ERR,
3390 "%s: scoped %s/%d on %s has no matching unscoped prefix\n",
3391 __func__, ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3392 pr->ndpr_plen, if_name(pr->ndpr_ifp)));
3393 }
3394
3395 lck_mtx_unlock(nd6_mutex);
3396 err = nd6_prefix_offlink(pr);
3397 lck_mtx_lock(nd6_mutex);
3398 if (err != 0) {
3399 nd6log((LOG_ERR,
3400 "%s: failed to make %s/%d offlink on %s, errno=%d\n",
3401 __func__, ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3402 pr->ndpr_plen, if_name(pr->ndpr_ifp), err));
3403 }
3404
3405 /* Add the entries back */
3406 if (opr != NULL) {
3407 err = nd6_prefix_onlink_scoped(opr, opr->ndpr_ifp->if_index);
3408 if (err != 0) {
3409 nd6log((LOG_ERR,
3410 "%s: failed to make %s/%d scoped onlink on %s, "
3411 "errno=%d\n", __func__,
3412 ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
3413 opr->ndpr_plen, if_name(opr->ndpr_ifp), err));
3414 }
3415 }
3416
3417 err = nd6_prefix_onlink_scoped(pr, IFSCOPE_NONE);
3418 if (err != 0) {
3419 nd6log((LOG_ERR,
3420 "%s: failed to make %s/%d onlink on %s, errno=%d\n",
3421 __func__, ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3422 pr->ndpr_plen, if_name(pr->ndpr_ifp), err));
3423 }
3424
3425 if (err != 0) {
3426 nd6log((LOG_ERR,
3427 "%s: error promoting %s/%d to %s from %s\n",
3428 __func__, ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3429 pr->ndpr_plen, if_name(pr->ndpr_ifp),
3430 (opr != NULL) ? if_name(opr->ndpr_ifp) : "NONE"));
3431 } else {
3432 nd6log2((LOG_INFO,
3433 "%s: %s/%d promoted, previously on %s\n",
3434 if_name(pr->ndpr_ifp),
3435 ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen,
3436 (opr != NULL) ? if_name(opr->ndpr_ifp) : "NONE"));
3437 }
3438
3439 if (opr != NULL)
3440 NDPR_REMREF(opr);
3441 }
3442
3443 static int
3444 nd6_prefix_onlink_common(struct nd_prefix *pr, boolean_t force_scoped,
3445 unsigned int ifscope)
3446 {
3447 struct ifaddr *ifa;
3448 struct ifnet *ifp = pr->ndpr_ifp;
3449 struct sockaddr_in6 mask6, prefix;
3450 struct nd_prefix *opr;
3451 u_int32_t rtflags;
3452 int error = 0, prproxy = 0;
3453 struct rtentry *rt = NULL;
3454
3455 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_OWNED);
3456
3457 /* sanity check */
3458 NDPR_LOCK(pr);
3459 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
3460 nd6log((LOG_ERR,
3461 "%s: %s/%d on %s scoped=%d is already on-link\n",
3462 __func__, ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3463 pr->ndpr_plen, if_name(pr->ndpr_ifp),
3464 (pr->ndpr_stateflags & NDPRF_IFSCOPE) ? 1 : 0);
3465 NDPR_UNLOCK(pr);
3466 return (EEXIST));
3467 }
3468 NDPR_UNLOCK(pr);
3469
3470 /*
3471 * Add the interface route associated with the prefix. Before
3472 * installing the route, check if there's the same prefix on another
3473 * interface, and the prefix has already installed the interface route.
3474 */
3475 opr = nd6_prefix_equal_lookup(pr, FALSE);
3476 if (opr != NULL)
3477 NDPR_REMREF(opr);
3478
3479 if (!ip6_doscopedroute) {
3480 /* if an interface route already exists, just return */
3481 if (opr != NULL)
3482 return (0);
3483 ifscope = IFSCOPE_NONE;
3484 } else if (!force_scoped) {
3485 /*
3486 * If a primary/non-scoped interface route already exists,
3487 * install the new one as a scoped entry. If the existing
3488 * interface route is scoped, install new as non-scoped.
3489 */
3490 ifscope = (opr != NULL) ? ifp->if_index : IFSCOPE_NONE;
3491 opr = nd6_prefix_equal_lookup(pr, TRUE);
3492 if (opr != NULL)
3493 NDPR_REMREF(opr);
3494 else if (ifscope != IFSCOPE_NONE)
3495 ifscope = IFSCOPE_NONE;
3496 }
3497
3498 /*
3499 * We prefer link-local addresses as the associated interface address.
3500 */
3501 /* search for a link-local addr */
3502 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
3503 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
3504 if (ifa == NULL) {
3505 struct in6_ifaddr *ia6;
3506 ifnet_lock_shared(ifp);
3507 IFP_TO_IA6(ifp, ia6);
3508 ifnet_lock_done(ifp);
3509 if (ia6 != NULL)
3510 ifa = &ia6->ia_ifa;
3511 /* should we care about ia6_flags? */
3512 }
3513 NDPR_LOCK(pr);
3514 if (ifa == NULL) {
3515 /*
3516 * This can still happen, when, for example, we receive an RA
3517 * containing a prefix with the L bit set and the A bit clear,
3518 * after removing all IPv6 addresses on the receiving
3519 * interface. This should, of course, be rare though.
3520 */
3521 nd6log((LOG_NOTICE,
3522 "nd6_prefix_onlink: failed to find any ifaddr"
3523 " to add route for a prefix(%s/%d) on %s\n",
3524 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3525 pr->ndpr_plen, if_name(ifp)));
3526 NDPR_UNLOCK(pr);
3527 return (0);
3528 }
3529
3530 /*
3531 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
3532 * ifa->ifa_rtrequest = nd6_rtrequest;
3533 */
3534 bzero(&mask6, sizeof (mask6));
3535 mask6.sin6_len = sizeof (mask6);
3536 mask6.sin6_addr = pr->ndpr_mask;
3537 prefix = pr->ndpr_prefix;
3538 if ((rt = pr->ndpr_rt) != NULL)
3539 pr->ndpr_rt = NULL;
3540 NDPR_ADDREF_LOCKED(pr); /* keep reference for this routine */
3541 NDPR_UNLOCK(pr);
3542
3543 IFA_LOCK_SPIN(ifa);
3544 rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
3545 IFA_UNLOCK(ifa);
3546 if (nd6_need_cache(ifp)) {
3547 /* explicitly set in case ifa_flags does not set the flag. */
3548 rtflags |= RTF_CLONING;
3549 } else {
3550 /*
3551 * explicitly clear the cloning bit in case ifa_flags sets it.
3552 */
3553 rtflags &= ~RTF_CLONING;
3554 }
3555
3556 lck_mtx_unlock(nd6_mutex);
3557
3558 if (rt != NULL) {
3559 rtfree(rt);
3560 rt = NULL;
3561 }
3562
3563 error = rtrequest_scoped(RTM_ADD, (struct sockaddr *)&prefix,
3564 ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt,
3565 ifscope);
3566
3567 /*
3568 * Serialize the setting of NDPRF_PRPROXY.
3569 */
3570 lck_mtx_lock(&proxy6_lock);
3571
3572 if (rt != NULL) {
3573 RT_LOCK(rt);
3574 nd6_rtmsg(RTM_ADD, rt);
3575 RT_UNLOCK(rt);
3576 NDPR_LOCK(pr);
3577 } else {
3578 NDPR_LOCK(pr);
3579 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
3580 " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx,"
3581 " scoped=%d, errno = %d\n",
3582 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3583 pr->ndpr_plen, if_name(ifp),
3584 ip6_sprintf(&((struct sockaddr_in6 *)
3585 (void *)ifa->ifa_addr)->sin6_addr),
3586 ip6_sprintf(&mask6.sin6_addr), rtflags,
3587 (ifscope != IFSCOPE_NONE), error));
3588 }
3589 NDPR_LOCK_ASSERT_HELD(pr);
3590
3591 pr->ndpr_stateflags &= ~(NDPRF_IFSCOPE | NDPRF_PRPROXY);
3592
3593 /*
3594 * TODO: If the prefix route exists, we should really find it and
3595 * refer the prefix to it; otherwise ndpr_rt is NULL.
3596 */
3597 if (!(pr->ndpr_stateflags & NDPRF_DEFUNCT) &&
3598 (rt != NULL || error == EEXIST)) {
3599 struct nd_ifinfo *ndi;
3600
3601 VERIFY(pr->ndpr_prproxy_sols_cnt == 0);
3602 VERIFY(RB_EMPTY(&pr->ndpr_prproxy_sols));
3603
3604 lck_rw_lock_shared(nd_if_rwlock);
3605 ndi = ND_IFINFO(ifp);
3606 VERIFY(ndi != NULL && ndi->initialized);
3607 lck_mtx_lock(&ndi->lock);
3608
3609 pr->ndpr_rt = rt; /* keep reference from rtrequest */
3610 pr->ndpr_stateflags |= NDPRF_ONLINK;
3611 if (ifscope != IFSCOPE_NONE) {
3612 pr->ndpr_stateflags |= NDPRF_IFSCOPE;
3613 } else if ((rtflags & RTF_CLONING) &&
3614 (ndi->flags & ND6_IFF_PROXY_PREFIXES) &&
3615 !IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) {
3616 /*
3617 * At present, in order for the prefix to be eligible
3618 * as a proxying/proxied prefix, we require that the
3619 * prefix route entry be marked as a cloning route with
3620 * RTF_PROXY; i.e. nd6_need_cache() needs to return
3621 * true for the interface type, hence the test for
3622 * RTF_CLONING above.
3623 */
3624 pr->ndpr_stateflags |= NDPRF_PRPROXY;
3625 }
3626
3627 lck_mtx_unlock(&ndi->lock);
3628 lck_rw_done(nd_if_rwlock);
3629 } else if (rt != NULL && pr->ndpr_stateflags & NDPRF_DEFUNCT)
3630 rtfree(rt);
3631
3632 prproxy = (pr->ndpr_stateflags & NDPRF_PRPROXY);
3633 VERIFY(!prproxy || !(pr->ndpr_stateflags & NDPRF_IFSCOPE));
3634 NDPR_UNLOCK(pr);
3635
3636 IFA_REMREF(ifa);
3637
3638 /*
3639 * If this is an upstream prefix, find the downstream ones (if any)
3640 * and re-configure their prefix routes accordingly. Otherwise,
3641 * this could be potentially be a downstream prefix, and so find the
3642 * upstream prefix, if any.
3643 */
3644 nd6_prproxy_prelist_update(pr, prproxy ? pr : NULL);
3645
3646 NDPR_REMREF(pr); /* release reference for this routine */
3647 lck_mtx_unlock(&proxy6_lock);
3648
3649 lck_mtx_lock(nd6_mutex);
3650
3651 return (error);
3652 }
3653
3654 int
3655 nd6_prefix_onlink(struct nd_prefix *pr)
3656 {
3657 return (nd6_prefix_onlink_common(pr, FALSE, IFSCOPE_NONE));
3658 }
3659
3660 int
3661 nd6_prefix_onlink_scoped(struct nd_prefix *pr, unsigned int ifscope)
3662 {
3663 return (nd6_prefix_onlink_common(pr, TRUE, ifscope));
3664 }
3665
3666 int
3667 nd6_prefix_offlink(struct nd_prefix *pr)
3668 {
3669 int plen, error = 0, prproxy;
3670 struct ifnet *ifp = pr->ndpr_ifp;
3671 struct nd_prefix *opr;
3672 struct sockaddr_in6 sa6, mask6, prefix;
3673 struct rtentry *rt = NULL, *ndpr_rt = NULL;
3674 unsigned int ifscope;
3675
3676 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_NOTOWNED);
3677
3678 /* sanity check */
3679 NDPR_LOCK(pr);
3680 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
3681 nd6log((LOG_ERR,
3682 "nd6_prefix_offlink: %s/%d on %s scoped=%d is already "
3683 "off-link\n", ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
3684 pr->ndpr_plen, if_name(pr->ndpr_ifp),
3685 (pr->ndpr_stateflags & NDPRF_IFSCOPE) ? 1 : 0));
3686 NDPR_UNLOCK(pr);
3687 return (EEXIST);
3688 }
3689
3690 bzero(&sa6, sizeof (sa6));
3691 sa6.sin6_family = AF_INET6;
3692 sa6.sin6_len = sizeof (sa6);
3693 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
3694 sizeof (struct in6_addr));
3695 bzero(&mask6, sizeof (mask6));
3696 mask6.sin6_family = AF_INET6;
3697 mask6.sin6_len = sizeof (sa6);
3698 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof (struct in6_addr));
3699 prefix = pr->ndpr_prefix;
3700 plen = pr->ndpr_plen;
3701 if ((ndpr_rt = pr->ndpr_rt) != NULL)
3702 pr->ndpr_rt = NULL;
3703 NDPR_ADDREF_LOCKED(pr); /* keep reference for this routine */
3704 NDPR_UNLOCK(pr);
3705
3706 ifscope = (pr->ndpr_stateflags & NDPRF_IFSCOPE) ?
3707 ifp->if_index : IFSCOPE_NONE;
3708
3709 error = rtrequest_scoped(RTM_DELETE, (struct sockaddr *)&sa6,
3710 NULL, (struct sockaddr *)&mask6, 0, &rt, ifscope);
3711
3712 if (rt != NULL) {
3713 /* report the route deletion to the routing socket. */
3714 RT_LOCK(rt);
3715 nd6_rtmsg(RTM_DELETE, rt);
3716 RT_UNLOCK(rt);
3717 rtfree(rt);
3718
3719 /*
3720 * The following check takes place only when Scoped Routing
3721 * is not enabled. There might be the same prefix on another
3722 * interface, the prefix which could not be on-link just
3723 * because we have the interface route (see comments in
3724 * nd6_prefix_onlink). If there's one, try to make the prefix
3725 * on-link on the interface.
3726 */
3727 lck_mtx_lock(nd6_mutex);
3728 opr = nd_prefix.lh_first;
3729 while (opr) {
3730 /* does not apply in the Scoped Routing case */
3731 if (ip6_doscopedroute)
3732 break;
3733
3734 if (opr == pr) {
3735 opr = opr->ndpr_next;
3736 continue;
3737 }
3738
3739 NDPR_LOCK(opr);
3740 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
3741 NDPR_UNLOCK(opr);
3742 opr = opr->ndpr_next;
3743 continue;
3744 }
3745 /*
3746 * KAME specific: detached prefixes should not be
3747 * on-link.
3748 */
3749 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0) {
3750 NDPR_UNLOCK(opr);
3751 opr = opr->ndpr_next;
3752 continue;
3753 }
3754 if (opr->ndpr_plen == plen &&
3755 in6_are_prefix_equal(&prefix.sin6_addr,
3756 &opr->ndpr_prefix.sin6_addr, plen)) {
3757 int e;
3758
3759 NDPR_ADDREF_LOCKED(opr);
3760 NDPR_UNLOCK(opr);
3761 if ((e = nd6_prefix_onlink(opr)) != 0) {
3762 nd6log((LOG_ERR,
3763 "nd6_prefix_offlink: failed to "
3764 "recover a prefix %s/%d from %s "
3765 "to %s (errno = %d)\n",
3766 ip6_sprintf(
3767 &opr->ndpr_prefix.sin6_addr),
3768 opr->ndpr_plen, if_name(ifp),
3769 if_name(opr->ndpr_ifp), e));
3770 }
3771 NDPR_REMREF(opr);
3772 opr = nd_prefix.lh_first;
3773 } else {
3774 NDPR_UNLOCK(opr);
3775 opr = opr->ndpr_next;
3776 }
3777 }
3778 lck_mtx_unlock(nd6_mutex);
3779 } else {
3780 nd6log((LOG_ERR,
3781 "nd6_prefix_offlink: failed to delete route: "
3782 "%s/%d on %s, scoped %d, (errno = %d)\n",
3783 ip6_sprintf(&sa6.sin6_addr), plen, if_name(ifp),
3784 (ifscope != IFSCOPE_NONE), error));
3785 }
3786
3787 if (ndpr_rt != NULL)
3788 rtfree(ndpr_rt);
3789
3790 lck_mtx_lock(&proxy6_lock);
3791
3792 NDPR_LOCK(pr);
3793 prproxy = (pr->ndpr_stateflags & NDPRF_PRPROXY);
3794 VERIFY(!prproxy || !(pr->ndpr_stateflags & NDPRF_IFSCOPE));
3795 pr->ndpr_stateflags &= ~(NDPRF_ONLINK | NDPRF_IFSCOPE | NDPRF_PRPROXY);
3796 if (pr->ndpr_prproxy_sols_cnt > 0) {
3797 VERIFY(prproxy);
3798 nd6_prproxy_sols_reap(pr);
3799 VERIFY(pr->ndpr_prproxy_sols_cnt == 0);
3800 VERIFY(RB_EMPTY(&pr->ndpr_prproxy_sols));
3801 }
3802 NDPR_UNLOCK(pr);
3803
3804 /*
3805 * If this was an upstream prefix, find the downstream ones and do
3806 * some cleanups. If this was a downstream prefix, the prefix route
3807 * has been removed from the routing table above, but there may be
3808 * other tasks to perform.
3809 */
3810 nd6_prproxy_prelist_update(pr, prproxy ? pr : NULL);
3811
3812 NDPR_REMREF(pr); /* release reference for this routine */
3813 lck_mtx_unlock(&proxy6_lock);
3814
3815 return (error);
3816 }
3817
3818 static struct in6_ifaddr *
3819 in6_pfx_newpersistaddr(struct nd_prefix *pr, int mcast, int *errorp)
3820 {
3821 struct in6_ifaddr *ia6;
3822 struct ifnet *ifp;
3823 struct nd_ifinfo *ndi;
3824 struct in6_addr mask;
3825 struct in6_aliasreq ifra;
3826 int error, ifaupdate, iidlen, notcga;
3827
3828 VERIFY(pr != NULL);
3829 VERIFY(errorp != NULL);
3830
3831 NDPR_LOCK(pr);
3832 ifp = pr->ndpr_ifp;
3833 ia6 = NULL;
3834 error = 0;
3835
3836 /*
3837 * Prefix Length check:
3838 * If the sum of the prefix length and interface identifier
3839 * length does not equal 128 bits, the Prefix Information
3840 * option MUST be ignored. The length of the interface
3841 * identifier is defined in a separate link-type specific
3842 * document.
3843 */
3844 iidlen = in6_if2idlen(ifp);
3845 if (iidlen < 0) {
3846 error = EADDRNOTAVAIL;
3847 /* this should not happen, so we always log it. */
3848 log(LOG_ERR, "%s: IID length undefined (%s)\n",
3849 __func__, if_name(ifp));
3850 goto unlock1;
3851 } else if (iidlen != 64) {
3852 error = EADDRNOTAVAIL;
3853 /*
3854 * stateless autoconfiguration not yet well-defined for IID
3855 * lengths other than 64 octets. Just give up for now.
3856 */
3857 nd6log((LOG_INFO, "%s: IID length not 64 octets (%s)\n",
3858 __func__, if_name(ifp)));
3859 goto unlock1;
3860 }
3861
3862 if (iidlen + pr->ndpr_plen != 128) {
3863 error = EADDRNOTAVAIL;
3864 nd6log((LOG_INFO,
3865 "%s: invalid prefix length %d for %s, ignored\n",
3866 __func__, pr->ndpr_plen, if_name(ifp)));
3867 goto unlock1;
3868 }
3869
3870 lck_rw_lock_shared(nd_if_rwlock);
3871 if (ifp->if_index >= nd_ifinfo_indexlim) {
3872 error = EADDRNOTAVAIL;
3873 nd6log((LOG_INFO,
3874 "%s: invalid prefix length %d for %s, ignored\n",
3875 __func__, pr->ndpr_plen, if_name(ifp)));
3876 goto unlock2;
3877 }
3878
3879 bzero(&ifra, sizeof (ifra));
3880 strlcpy(ifra.ifra_name, if_name(ifp), sizeof (ifra.ifra_name));
3881 ifra.ifra_addr.sin6_family = AF_INET6;
3882 ifra.ifra_addr.sin6_len = sizeof (struct sockaddr_in6);
3883
3884 /* prefix */
3885 bcopy(&pr->ndpr_prefix.sin6_addr, &ifra.ifra_addr.sin6_addr,
3886 sizeof (ifra.ifra_addr.sin6_addr));
3887 in6_len2mask(&mask, pr->ndpr_plen);
3888 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
3889 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
3890 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
3891 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
3892
3893 ndi = &nd_ifinfo[ifp->if_index];
3894 VERIFY(ndi->initialized);
3895 lck_mtx_lock(&ndi->lock);
3896
3897 notcga = nd6_send_opstate == ND6_SEND_OPMODE_DISABLED ||
3898 (ndi->flags & ND6_IFF_INSECURE) != 0;
3899
3900 lck_mtx_unlock(&ndi->lock);
3901 lck_rw_done(nd_if_rwlock);
3902 NDPR_UNLOCK(pr);
3903
3904 if (notcga) {
3905 ia6 = in6ifa_ifpforlinklocal(ifp, 0);
3906 if (ia6 == NULL) {
3907 error = EADDRNOTAVAIL;
3908 nd6log((LOG_INFO, "%s: no link-local address (%s)\n",
3909 __func__, if_name(ifp)));
3910 goto done;
3911 }
3912
3913 IFA_LOCK(&ia6->ia_ifa);
3914 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
3915 (ia6->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
3916 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
3917 (ia6->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
3918 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
3919 (ia6->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
3920 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
3921 (ia6->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
3922 IFA_UNLOCK(&ia6->ia_ifa);
3923 IFA_REMREF(&ia6->ia_ifa);
3924 ia6 = NULL;
3925 } else {
3926 in6_cga_node_lock();
3927 error = in6_cga_generate(NULL, 0, &ifra.ifra_addr.sin6_addr);
3928 in6_cga_node_unlock();
3929 if (error == 0)
3930 ifra.ifra_flags |= IN6_IFF_SECURED;
3931 else {
3932 nd6log((LOG_ERR, "%s: no CGA available (%s)\n",
3933 __func__, if_name(ifp)));
3934 goto done;
3935 }
3936 }
3937
3938 VERIFY(ia6 == NULL);
3939
3940 /* new prefix mask. */
3941 ifra.ifra_prefixmask.sin6_len = sizeof (struct sockaddr_in6);
3942 ifra.ifra_prefixmask.sin6_family = AF_INET6;
3943 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
3944 sizeof (ifra.ifra_prefixmask.sin6_addr));
3945
3946 /* lifetimes. */
3947 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
3948 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
3949
3950 /* address flags */
3951 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
3952
3953 /*
3954 * Make sure that we do not have this address already. This should
3955 * usually not happen, but we can still see this case, e.g., if we
3956 * have manually configured the exact address to be configured.
3957 */
3958 if ((ia6 = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr))
3959 != NULL) {
3960 error = EADDRNOTAVAIL;
3961 IFA_REMREF(&ia6->ia_ifa);
3962 ia6 = NULL;
3963
3964 /* this should be rare enough to make an explicit log */
3965 log(LOG_INFO, "%s: %s is already configured!\n",
3966 __func__, ip6_sprintf(&ifra.ifra_addr.sin6_addr));
3967 goto done;
3968 }
3969
3970 /*
3971 * Allocate ifaddr structure, link into chain, etc.
3972 * If we are going to create a new address upon receiving a multicasted
3973 * RA, we need to impose a random delay before starting DAD.
3974 * [RFC 4862, Section 5.4.2]
3975 */
3976 ifaupdate = IN6_IFAUPDATE_NOWAIT;
3977 if (mcast)
3978 ifaupdate |= IN6_IFAUPDATE_DADDELAY;
3979 error = in6_update_ifa(ifp, &ifra, ifaupdate, &ia6);
3980 if (error != 0) {
3981 nd6log((LOG_ERR,
3982 "%s: failed to make ifaddr %s on %s (errno=%d)\n",
3983 __func__, ip6_sprintf(&ifra.ifra_addr.sin6_addr),
3984 if_name(ifp), error));
3985 error = EADDRNOTAVAIL;
3986 goto done;
3987 }
3988
3989 VERIFY(ia6 != NULL);
3990 in6_post_msg(ifp, KEV_INET6_NEW_RTADV_ADDR, ia6, NULL);
3991 goto done;
3992
3993 unlock2:
3994 lck_rw_done(nd_if_rwlock);
3995
3996 unlock1:
3997 NDPR_UNLOCK(pr);
3998
3999 done:
4000 *errorp = error;
4001 return (ia6);
4002 }
4003
4004 #define IA6_NONCONST(i) ((struct in6_ifaddr *)(uintptr_t)(i))
4005
4006 int
4007 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen)
4008 {
4009 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
4010 struct in6_ifaddr *ia, *newia;
4011 struct in6_aliasreq ifra;
4012 int i, error, ifaupdate;
4013 int trylimit = 3; /* XXX: adhoc value */
4014 u_int32_t randid[2];
4015 time_t vltime0, pltime0;
4016 uint64_t timenow = net_uptime();
4017 struct in6_addr addr;
4018 struct nd_prefix *ndpr;
4019
4020 bzero(&ifra, sizeof (ifra));
4021 strlcpy(ifra.ifra_name, if_name(ifp), sizeof (ifra.ifra_name));
4022 IFA_LOCK(&IA6_NONCONST(ia0)->ia_ifa);
4023 ifra.ifra_addr = ia0->ia_addr;
4024 /* copy prefix mask */
4025 ifra.ifra_prefixmask = ia0->ia_prefixmask;
4026 /* clear the old IFID */
4027 for (i = 0; i < 4; i++) {
4028 ifra.ifra_addr.sin6_addr.s6_addr32[i]
4029 &= ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
4030 }
4031 addr = ia0->ia_addr.sin6_addr;
4032 IFA_UNLOCK(&IA6_NONCONST(ia0)->ia_ifa);
4033
4034 again:
4035 in6_iid_mktmp(ifp, (u_int8_t *)randid,
4036 (const u_int8_t *)&addr.s6_addr[8], forcegen);
4037
4038 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
4039 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
4040 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
4041 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
4042
4043 /*
4044 * in6_iid_mktmp() quite likely provided a unique interface ID.
4045 * However, we may still have a chance to see collision, because
4046 * there may be a time lag between generation of the ID and generation
4047 * of the address. So, we'll do one more sanity check.
4048 */
4049 if ((ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr)) != NULL) {
4050 IFA_REMREF(&ia->ia_ifa);
4051 if (trylimit-- == 0) {
4052 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find "
4053 "a unique random IFID\n"));
4054 return (EEXIST);
4055 }
4056 forcegen = 1;
4057 goto again;
4058 }
4059
4060 /*
4061 * The Valid Lifetime is the lower of the Valid Lifetime of the
4062 * public address or TEMP_VALID_LIFETIME.
4063 * The Preferred Lifetime is the lower of the Preferred Lifetime
4064 * of the public address or TEMP_PREFERRED_LIFETIME -
4065 * DESYNC_FACTOR.
4066 */
4067 IFA_LOCK(&IA6_NONCONST(ia0)->ia_ifa);
4068 if (ia0->ia6_lifetime.ia6ti_vltime != ND6_INFINITE_LIFETIME) {
4069 vltime0 = IFA6_IS_INVALID(ia0, timenow) ? 0 :
4070 (ia0->ia6_lifetime.ia6ti_vltime -
4071 (timenow - ia0->ia6_updatetime));
4072 if (vltime0 > ip6_temp_valid_lifetime)
4073 vltime0 = ip6_temp_valid_lifetime;
4074 } else {
4075 vltime0 = ip6_temp_valid_lifetime;
4076 }
4077 if (ia0->ia6_lifetime.ia6ti_pltime != ND6_INFINITE_LIFETIME) {
4078 pltime0 = IFA6_IS_DEPRECATED(ia0, timenow) ? 0 :
4079 (ia0->ia6_lifetime.ia6ti_pltime -
4080 (timenow - ia0->ia6_updatetime));
4081 if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor)
4082 pltime0 = ip6_temp_preferred_lifetime -
4083 ip6_desync_factor;
4084 } else {
4085 pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
4086 }
4087 ifra.ifra_lifetime.ia6t_vltime = vltime0;
4088 ifra.ifra_lifetime.ia6t_pltime = pltime0;
4089 IFA_UNLOCK(&IA6_NONCONST(ia0)->ia_ifa);
4090 /*
4091 * A temporary address is created only if this calculated Preferred
4092 * Lifetime is greater than REGEN_ADVANCE time units.
4093 */
4094 if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
4095 return (0);
4096
4097 /* XXX: scope zone ID? */
4098
4099 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
4100
4101 /* allocate ifaddr structure, link into chain, etc. */
4102 ifaupdate = IN6_IFAUPDATE_NOWAIT | IN6_IFAUPDATE_DADDELAY;
4103 error = in6_update_ifa(ifp, &ifra, ifaupdate, &newia);
4104 if (error != 0) {
4105 nd6log((LOG_ERR, "in6_tmpifadd: failed to add address.\n"));
4106 return (error);
4107 }
4108 VERIFY(newia != NULL);
4109
4110 IFA_LOCK(&IA6_NONCONST(ia0)->ia_ifa);
4111 ndpr = ia0->ia6_ndpr;
4112 if (ndpr == NULL) {
4113 /*
4114 * We lost the race with another thread that has purged
4115 * ia0 address; in this case, purge the tmp addr as well.
4116 */
4117 nd6log((LOG_ERR, "in6_tmpifadd: no public address\n"));
4118 VERIFY(!(ia0->ia6_flags & IN6_IFF_AUTOCONF));
4119 IFA_UNLOCK(&IA6_NONCONST(ia0)->ia_ifa);
4120 in6_purgeaddr(&newia->ia_ifa);
4121 IFA_REMREF(&newia->ia_ifa);
4122 return (EADDRNOTAVAIL);
4123 }
4124 NDPR_ADDREF(ndpr); /* for us */
4125 IFA_UNLOCK(&IA6_NONCONST(ia0)->ia_ifa);
4126 IFA_LOCK(&newia->ia_ifa);
4127 if (newia->ia6_ndpr != NULL) {
4128 NDPR_LOCK(newia->ia6_ndpr);
4129 VERIFY(newia->ia6_ndpr->ndpr_addrcnt != 0);
4130 newia->ia6_ndpr->ndpr_addrcnt--;
4131 NDPR_UNLOCK(newia->ia6_ndpr);
4132 NDPR_REMREF(newia->ia6_ndpr); /* release addr reference */
4133 }
4134 newia->ia6_ndpr = ndpr;
4135 NDPR_LOCK(newia->ia6_ndpr);
4136 newia->ia6_ndpr->ndpr_addrcnt++;
4137 VERIFY(newia->ia6_ndpr->ndpr_addrcnt != 0);
4138 NDPR_ADDREF_LOCKED(newia->ia6_ndpr); /* for addr reference */
4139 NDPR_UNLOCK(newia->ia6_ndpr);
4140 IFA_UNLOCK(&newia->ia_ifa);
4141 /*
4142 * A newly added address might affect the status of other addresses.
4143 * XXX: when the temporary address is generated with a new public
4144 * address, the onlink check is redundant. However, it would be safe
4145 * to do the check explicitly everywhere a new address is generated,
4146 * and, in fact, we surely need the check when we create a new
4147 * temporary address due to deprecation of an old temporary address.
4148 */
4149 lck_mtx_lock(nd6_mutex);
4150 pfxlist_onlink_check();
4151 lck_mtx_unlock(nd6_mutex);
4152 IFA_REMREF(&newia->ia_ifa);
4153
4154 /* remove our reference */
4155 NDPR_REMREF(ndpr);
4156
4157 return (0);
4158 }
4159 #undef IA6_NONCONST
4160
4161 int
4162 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
4163 {
4164 struct timeval caltime;
4165 u_int64_t timenow = net_uptime();
4166
4167 NDPR_LOCK_ASSERT_HELD(ndpr);
4168
4169 getmicrotime(&caltime);
4170 ndpr->ndpr_base_calendartime = caltime.tv_sec;
4171 ndpr->ndpr_base_uptime = timenow;
4172
4173 /* check if preferred lifetime > valid lifetime. RFC 4862 5.5.3 (c) */
4174 if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
4175 nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
4176 "(%d) is greater than valid lifetime(%d)\n",
4177 (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
4178 return (EINVAL);
4179 }
4180 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
4181 ndpr->ndpr_preferred = 0;
4182 else
4183 ndpr->ndpr_preferred = timenow + ndpr->ndpr_pltime;
4184 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
4185 ndpr->ndpr_expire = 0;
4186 else
4187 ndpr->ndpr_expire = timenow + ndpr->ndpr_vltime;
4188
4189 return (0);
4190 }
4191
4192 static void
4193 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
4194 {
4195 #pragma unused(new)
4196 uint64_t timenow = net_uptime();
4197
4198 /* Valid lifetime must not be updated unless explicitly specified. */
4199 /* init ia6t_expire */
4200 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME) {
4201 lt6->ia6t_expire = 0;
4202 } else {
4203 lt6->ia6t_expire = timenow;
4204 lt6->ia6t_expire += lt6->ia6t_vltime;
4205 }
4206
4207 /* init ia6t_preferred */
4208 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME) {
4209 lt6->ia6t_preferred = 0;
4210 } else {
4211 lt6->ia6t_preferred = timenow;
4212 lt6->ia6t_preferred += lt6->ia6t_pltime;
4213 }
4214 }
4215
4216 /*
4217 * Delete all the routing table entries that use the specified gateway.
4218 * XXX: this function causes search through all entries of routing table, so
4219 * it shouldn't be called when acting as a router.
4220 */
4221 void
4222 rt6_flush(
4223 struct in6_addr *gateway,
4224 struct ifnet *ifp)
4225 {
4226 struct radix_node_head *rnh = rt_tables[AF_INET6];
4227
4228 /* We'll care only link-local addresses */
4229 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
4230 return;
4231 }
4232 lck_mtx_lock(rnh_lock);
4233 /* XXX: hack for KAME's link-local address kludge */
4234 gateway->s6_addr16[1] = htons(ifp->if_index);
4235
4236 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
4237 lck_mtx_unlock(rnh_lock);
4238 }
4239
4240 static int
4241 rt6_deleteroute(
4242 struct radix_node *rn,
4243 void *arg)
4244 {
4245 struct rtentry *rt = (struct rtentry *)rn;
4246 struct in6_addr *gate = (struct in6_addr *)arg;
4247
4248 lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED);
4249
4250 RT_LOCK(rt);
4251 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6) {
4252 RT_UNLOCK(rt);
4253 return (0);
4254 }
4255
4256 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
4257 RT_UNLOCK(rt);
4258 return (0);
4259 }
4260 /*
4261 * Do not delete a static route.
4262 * XXX: this seems to be a bit ad-hoc. Should we consider the
4263 * 'cloned' bit instead?
4264 */
4265 if ((rt->rt_flags & RTF_STATIC) != 0) {
4266 RT_UNLOCK(rt);
4267 return (0);
4268 }
4269 /*
4270 * We delete only host route. This means, in particular, we don't
4271 * delete default route.
4272 */
4273 if ((rt->rt_flags & RTF_HOST) == 0) {
4274 RT_UNLOCK(rt);
4275 return (0);
4276 }
4277
4278 /*
4279 * Safe to drop rt_lock and use rt_key, rt_gateway, since holding
4280 * rnh_lock here prevents another thread from calling rt_setgate()
4281 * on this route.
4282 */
4283 RT_UNLOCK(rt);
4284 return (rtrequest_locked(RTM_DELETE, rt_key(rt), rt->rt_gateway,
4285 rt_mask(rt), rt->rt_flags, 0));
4286 }
4287
4288 int
4289 nd6_setdefaultiface(
4290 int ifindex)
4291 {
4292 int error = 0;
4293 ifnet_t def_ifp = NULL;
4294
4295 lck_mtx_assert(nd6_mutex, LCK_MTX_ASSERT_NOTOWNED);
4296
4297 ifnet_head_lock_shared();
4298 if (ifindex < 0 || if_index < ifindex) {
4299 ifnet_head_done();
4300 return (EINVAL);
4301 }
4302 def_ifp = ifindex2ifnet[ifindex];
4303 ifnet_head_done();
4304
4305 lck_mtx_lock(nd6_mutex);
4306 if (nd6_defifindex != ifindex) {
4307 struct ifnet *odef_ifp = nd6_defifp;
4308
4309 nd6_defifindex = ifindex;
4310 if (nd6_defifindex > 0)
4311 nd6_defifp = def_ifp;
4312 else
4313 nd6_defifp = NULL;
4314
4315 if (nd6_defifp != NULL)
4316 nd6log((LOG_INFO, "%s: is now the default "
4317 "interface (was %s)\n", if_name(nd6_defifp),
4318 odef_ifp != NULL ? if_name(odef_ifp) : "NONE"));
4319 else
4320 nd6log((LOG_INFO, "No default interface set\n"));
4321
4322 /*
4323 * If the Default Router List is empty, install a route
4324 * to the specified interface as default or remove the default
4325 * route when the default interface becomes canceled.
4326 * The check for the queue is actually redundant, but
4327 * we do this here to avoid re-install the default route
4328 * if the list is NOT empty.
4329 */
4330 if (ip6_doscopedroute || TAILQ_FIRST(&nd_defrouter) == NULL) {
4331 defrtrlist_sync(nd6_defifp);
4332 nd6_prefix_sync(nd6_defifp);
4333 }
4334
4335 /*
4336 * Our current implementation assumes one-to-one mapping between
4337 * interfaces and links, so it would be natural to use the
4338 * default interface as the default link.
4339 */
4340 scope6_setdefault(nd6_defifp);
4341 }
4342 lck_mtx_unlock(nd6_mutex);
4343
4344 return (error);
4345 }
mirror of XNU