nd6_rtr.c revision 1.128 1 /* $NetBSD: nd6_rtr.c,v 1.128 2016/12/19 07:51:34 ozaki-r Exp $ */
2 /* $KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $ */
3
4 /*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.128 2016/12/19 07:51:34 ozaki-r Exp $");
35
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/ioctl.h>
46 #include <sys/syslog.h>
47 #include <sys/cprng.h>
48
49 #include <net/if.h>
50 #include <net/if_types.h>
51 #include <net/if_dl.h>
52
53 #include <netinet/in.h>
54 #include <netinet6/in6_var.h>
55 #include <netinet6/in6_ifattach.h>
56 #include <netinet/ip6.h>
57 #include <netinet6/ip6_var.h>
58 #include <netinet6/nd6.h>
59 #include <netinet/icmp6.h>
60 #include <netinet6/icmp6_private.h>
61 #include <netinet6/scope6_var.h>
62
63 #include <net/net_osdep.h>
64
65 static int rtpref(struct nd_defrouter *);
66 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
67 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
68 struct mbuf *, int);
69 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int, struct psref *);
70 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
71 struct nd_defrouter *);
72 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
73 static void pfxrtr_del(struct nd_pfxrouter *);
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 (struct nd_prefix *);
76
77 static void defrouter_addreq(struct nd_defrouter *);
78 static void defrouter_delreq(struct nd_defrouter *);
79
80 static int in6_init_prefix_ltimes(struct nd_prefix *);
81 static void in6_init_address_ltimes(struct nd_prefix *,
82 struct in6_addrlifetime *);
83 static void purge_detached(struct ifnet *);
84
85 static int rt6_deleteroute_matcher(struct rtentry *, void *);
86
87 static int nd6_prelist_add(struct nd_prefixctl *, struct nd_defrouter *,
88 struct nd_prefix **);
89 static int nd6_prefix_onlink(struct nd_prefix *);
90 static int nd6_prefix_offlink(struct nd_prefix *);
91 static struct nd_prefix *nd6_prefix_lookup(struct nd_prefixctl *);
92
93 extern int nd6_recalc_reachtm_interval;
94
95 int ip6_use_tempaddr = 0;
96
97 int ip6_desync_factor;
98 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
99 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
100 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
101
102 int nd6_numroutes = 0;
103
104 /* RTPREF_MEDIUM has to be 0! */
105 #define RTPREF_HIGH 1
106 #define RTPREF_MEDIUM 0
107 #define RTPREF_LOW (-1)
108 #define RTPREF_RESERVED (-2)
109 #define RTPREF_INVALID (-3) /* internal */
110
111 static inline bool
112 nd6_is_llinfo_probreach(struct nd_defrouter *dr)
113 {
114 struct llentry *ln = NULL;
115
116 ln = nd6_lookup(&dr->rtaddr, dr->ifp, false);
117 if (ln == NULL)
118 return false;
119
120 if (!ND6_IS_LLINFO_PROBREACH(ln)) {
121 LLE_RUNLOCK(ln);
122 return false;
123 }
124
125 LLE_RUNLOCK(ln);
126 return true;
127 }
128
129 /*
130 * Receive Router Solicitation Message - just for routers.
131 * Router solicitation/advertisement is mostly managed by a userland program
132 * (rtadvd) so here we have no function like nd6_ra_output().
133 *
134 * Based on RFC 2461
135 */
136 void
137 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
138 {
139 struct ifnet *ifp;
140 struct nd_ifinfo *ndi;
141 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
142 struct nd_router_solicit *nd_rs;
143 struct in6_addr saddr6 = ip6->ip6_src;
144 char *lladdr = NULL;
145 int lladdrlen = 0;
146 union nd_opts ndopts;
147 struct psref psref;
148
149 ifp = m_get_rcvif_psref(m, &psref);
150 if (ifp == NULL)
151 goto freeit;
152
153 ndi = ND_IFINFO(ifp);
154
155 /* If I'm not a router, ignore it. */
156 if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
157 goto freeit;
158
159 /* Sanity checks */
160 if (ip6->ip6_hlim != 255) {
161 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
162 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
163 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
164 goto bad;
165 }
166
167 /*
168 * Don't update the neighbor cache, if src = ::.
169 * This indicates that the src has no IP address assigned yet.
170 */
171 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
172 goto freeit;
173
174 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
175 if (nd_rs == NULL) {
176 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
177 return;
178 }
179
180 icmp6len -= sizeof(*nd_rs);
181 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
182 if (nd6_options(&ndopts) < 0) {
183 nd6log(LOG_INFO, "invalid ND option, ignored\n");
184 /* nd6_options have incremented stats */
185 goto freeit;
186 }
187
188 if (ndopts.nd_opts_src_lladdr) {
189 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
190 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
191 }
192
193 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
194 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
195 "(if %d, RS packet %d)\n",
196 ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2);
197 goto bad;
198 }
199
200 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
201
202 freeit:
203 m_put_rcvif_psref(ifp, &psref);
204 m_freem(m);
205 return;
206
207 bad:
208 ICMP6_STATINC(ICMP6_STAT_BADRS);
209 m_put_rcvif_psref(ifp, &psref);
210 m_freem(m);
211 }
212
213 /*
214 * Receive Router Advertisement Message.
215 *
216 * Based on RFC 2461
217 * TODO: on-link bit on prefix information
218 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
219 */
220 void
221 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
222 {
223 struct ifnet *ifp;
224 struct nd_ifinfo *ndi;
225 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
226 struct nd_router_advert *nd_ra;
227 struct in6_addr saddr6 = ip6->ip6_src;
228 #if 0
229 struct in6_addr daddr6 = ip6->ip6_dst;
230 int flags; /* = nd_ra->nd_ra_flags_reserved; */
231 int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
232 int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
233 #endif
234 int mcast = 0;
235 union nd_opts ndopts;
236 struct nd_defrouter *dr;
237 struct psref psref;
238
239 ifp = m_get_rcvif_psref(m, &psref);
240 if (ifp == NULL)
241 goto freeit;
242
243 ndi = ND_IFINFO(ifp);
244 /*
245 * We only accept RAs when
246 * the system-wide variable allows the acceptance, and the
247 * per-interface variable allows RAs on the receiving interface.
248 */
249 if (!nd6_accepts_rtadv(ndi))
250 goto freeit;
251
252 if (ip6->ip6_hlim != 255) {
253 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
254 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
255 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
256 goto bad;
257 }
258
259 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
260 nd6log(LOG_ERR, "src %s is not link-local\n",
261 ip6_sprintf(&saddr6));
262 goto bad;
263 }
264
265 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
266 if (nd_ra == NULL) {
267 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
268 m_put_rcvif_psref(ifp, &psref);
269 return;
270 }
271
272 icmp6len -= sizeof(*nd_ra);
273 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
274 if (nd6_options(&ndopts) < 0) {
275 nd6log(LOG_INFO, "invalid ND option, ignored\n");
276 /* nd6_options have incremented stats */
277 goto freeit;
278 }
279
280 {
281 struct nd_defrouter drtr;
282 u_int32_t advreachable = nd_ra->nd_ra_reachable;
283
284 /* remember if this is a multicasted advertisement */
285 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
286 mcast = 1;
287
288 memset(&drtr, 0, sizeof(drtr));
289 drtr.rtaddr = saddr6;
290 drtr.flags = nd_ra->nd_ra_flags_reserved;
291 drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
292 drtr.expire = time_uptime + drtr.rtlifetime;
293 drtr.ifp = ifp;
294 /* unspecified or not? (RFC 2461 6.3.4) */
295 if (advreachable) {
296 NTOHL(advreachable);
297 if (advreachable <= MAX_REACHABLE_TIME &&
298 ndi->basereachable != advreachable) {
299 ndi->basereachable = advreachable;
300 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
301 ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
302 }
303 }
304 if (nd_ra->nd_ra_retransmit)
305 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
306 if (nd_ra->nd_ra_curhoplimit) {
307 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
308 ndi->chlim = nd_ra->nd_ra_curhoplimit;
309 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
310 log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
311 "%s on %s (current=%d, received=%d), ignored\n",
312 ip6_sprintf(&ip6->ip6_src),
313 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
314 }
315 ND6_WLOCK();
316 dr = defrtrlist_update(&drtr);
317 }
318
319 /*
320 * prefix
321 */
322 if (ndopts.nd_opts_pi) {
323 struct nd_opt_hdr *pt;
324 struct nd_opt_prefix_info *pi = NULL;
325 struct nd_prefixctl prc;
326
327 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
328 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
329 pt = (struct nd_opt_hdr *)((char *)pt +
330 (pt->nd_opt_len << 3))) {
331 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
332 continue;
333 pi = (struct nd_opt_prefix_info *)pt;
334
335 if (pi->nd_opt_pi_len != 4) {
336 nd6log(LOG_INFO, "invalid option "
337 "len %d for prefix information option, "
338 "ignored\n", pi->nd_opt_pi_len);
339 continue;
340 }
341
342 if (128 < pi->nd_opt_pi_prefix_len) {
343 nd6log(LOG_INFO, "invalid prefix "
344 "len %d for prefix information option, "
345 "ignored\n", pi->nd_opt_pi_prefix_len);
346 continue;
347 }
348
349 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
350 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
351 nd6log(LOG_INFO,
352 "invalid prefix %s, ignored\n",
353 ip6_sprintf(&pi->nd_opt_pi_prefix));
354 continue;
355 }
356
357 memset(&prc, 0, sizeof(prc));
358 sockaddr_in6_init(&prc.ndprc_prefix,
359 &pi->nd_opt_pi_prefix, 0, 0, 0);
360 prc.ndprc_ifp = ifp;
361
362 prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
363 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
364 prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
365 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
366 prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
367 prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
368 prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
369
370 (void)prelist_update(&prc, dr, m, mcast);
371 }
372 }
373 ND6_UNLOCK();
374
375 /*
376 * MTU
377 */
378 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
379 u_long mtu;
380 u_long maxmtu;
381
382 mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
383
384 /* lower bound */
385 if (mtu < IPV6_MMTU) {
386 nd6log(LOG_INFO, "bogus mtu option "
387 "mtu=%lu sent from %s, ignoring\n",
388 mtu, ip6_sprintf(&ip6->ip6_src));
389 goto skip;
390 }
391
392 /* upper bound */
393 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
394 ? ndi->maxmtu : ifp->if_mtu;
395 if (mtu <= maxmtu) {
396 int change = (ndi->linkmtu != mtu);
397
398 ndi->linkmtu = mtu;
399 if (change) /* in6_maxmtu may change */
400 in6_setmaxmtu();
401 } else {
402 nd6log(LOG_INFO,
403 "bogus mtu mtu=%lu sent from %s; "
404 "exceeds maxmtu %lu, ignoring\n",
405 mtu, ip6_sprintf(&ip6->ip6_src), maxmtu);
406 }
407 }
408
409 skip:
410
411 /*
412 * Source link layer address
413 */
414 {
415 char *lladdr = NULL;
416 int lladdrlen = 0;
417
418 if (ndopts.nd_opts_src_lladdr) {
419 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
420 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
421 }
422
423 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
424 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
425 "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
426 ifp->if_addrlen, lladdrlen - 2);
427 goto bad;
428 }
429
430 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
431
432 /*
433 * Installing a link-layer address might change the state of the
434 * router's neighbor cache, which might also affect our on-link
435 * detection of adveritsed prefixes.
436 */
437 ND6_WLOCK();
438 nd6_pfxlist_onlink_check();
439 ND6_UNLOCK();
440 }
441
442 freeit:
443 m_put_rcvif_psref(ifp, &psref);
444 m_freem(m);
445 return;
446
447 bad:
448 ICMP6_STATINC(ICMP6_STAT_BADRA);
449 m_put_rcvif_psref(ifp, &psref);
450 m_freem(m);
451 }
452
453 /*
454 * default router list processing sub routines
455 */
456 static void
457 defrouter_addreq(struct nd_defrouter *newdr)
458 {
459 union {
460 struct sockaddr_in6 sin6;
461 struct sockaddr sa;
462 } def, mask, gate;
463 int s;
464 int error;
465
466 memset(&def, 0, sizeof(def));
467 memset(&mask, 0, sizeof(mask));
468 memset(&gate, 0,sizeof(gate)); /* for safety */
469
470 def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
471 sizeof(struct sockaddr_in6);
472 def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
473 gate.sin6.sin6_addr = newdr->rtaddr;
474 #ifndef SCOPEDROUTING
475 gate.sin6.sin6_scope_id = 0; /* XXX */
476 #endif
477
478 s = splsoftnet();
479 error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
480 RTF_GATEWAY);
481 if (error == 0) {
482 nd6_numroutes++;
483 newdr->installed = 1;
484 }
485 splx(s);
486 return;
487 }
488
489 struct nd_defrouter *
490 nd6_defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
491 {
492 struct nd_defrouter *dr;
493
494 ND6_ASSERT_LOCK();
495
496 ND_DEFROUTER_LIST_FOREACH(dr) {
497 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
498 break;
499 }
500
501 return dr; /* search failed */
502 }
503
504 void
505 nd6_defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
506 {
507 struct nd_defrouter *deldr = NULL;
508 struct nd_prefix *pr;
509 struct nd_ifinfo *ndi;
510
511 ND6_ASSERT_WLOCK();
512
513 if (ext == NULL)
514 ext = dr->ifp->if_afdata[AF_INET6];
515
516 /* detach already in progress, can not do anything */
517 if (ext == NULL)
518 return;
519
520 ndi = ext->nd_ifinfo;
521
522 /*
523 * Flush all the routing table entries that use the router
524 * as a next hop.
525 */
526 /* XXX: better condition? */
527 if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
528 nd6_rt_flush(&dr->rtaddr, dr->ifp);
529
530 if (dr->installed) {
531 deldr = dr;
532 defrouter_delreq(dr);
533 }
534 ND_DEFROUTER_LIST_REMOVE(dr);
535
536 /*
537 * Also delete all the pointers to the router in each prefix lists.
538 */
539 ND_PREFIX_LIST_FOREACH(pr) {
540 struct nd_pfxrouter *pfxrtr;
541 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
542 pfxrtr_del(pfxrtr);
543 }
544 nd6_pfxlist_onlink_check();
545
546 /*
547 * If the router is the primary one, choose a new one.
548 * Note that nd6_defrouter_select() will remove the current gateway
549 * from the routing table.
550 */
551 if (deldr)
552 nd6_defrouter_select();
553
554 ext->ndefrouters--;
555 if (ext->ndefrouters < 0) {
556 log(LOG_WARNING, "nd6_defrtrlist_del: negative count on %s\n",
557 dr->ifp->if_xname);
558 }
559
560 free(dr, M_IP6NDP);
561 }
562
563 /*
564 * Remove the default route for a given router.
565 * This is just a subroutine function for nd6_defrouter_select(), and should
566 * not be called from anywhere else.
567 */
568 static void
569 defrouter_delreq(struct nd_defrouter *dr)
570 {
571 union {
572 struct sockaddr_in6 sin6;
573 struct sockaddr sa;
574 } def, mask, gw;
575 int error;
576
577 #ifdef DIAGNOSTIC
578 if (dr == NULL)
579 panic("dr == NULL in defrouter_delreq");
580 #endif
581
582 memset(&def, 0, sizeof(def));
583 memset(&mask, 0, sizeof(mask));
584 memset(&gw, 0, sizeof(gw)); /* for safety */
585
586 def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
587 sizeof(struct sockaddr_in6);
588 def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
589 gw.sin6.sin6_addr = dr->rtaddr;
590 #ifndef SCOPEDROUTING
591 gw.sin6.sin6_scope_id = 0; /* XXX */
592 #endif
593
594 error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
595 RTF_GATEWAY);
596 if (error == 0)
597 nd6_numroutes--;
598
599 dr->installed = 0;
600 }
601
602 /*
603 * remove all default routes from default router list
604 */
605 void
606 nd6_defrouter_reset(void)
607 {
608 struct nd_defrouter *dr;
609
610 ND6_ASSERT_WLOCK();
611
612 ND_DEFROUTER_LIST_FOREACH(dr)
613 defrouter_delreq(dr);
614
615 /*
616 * XXX should we also nuke any default routers in the kernel, by
617 * going through them by rtalloc1()?
618 */
619 }
620
621 /*
622 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
623 * draft-ietf-ipngwg-router-selection:
624 * 1) Routers that are reachable or probably reachable should be preferred.
625 * If we have more than one (probably) reachable router, prefer ones
626 * with the highest router preference.
627 * 2) When no routers on the list are known to be reachable or
628 * probably reachable, routers SHOULD be selected in a round-robin
629 * fashion, regardless of router preference values.
630 * 3) If the Default Router List is empty, assume that all
631 * destinations are on-link.
632 *
633 * We assume nd_defrouter is sorted by router preference value.
634 * Since the code below covers both with and without router preference cases,
635 * we do not need to classify the cases by ifdef.
636 *
637 * At this moment, we do not try to install more than one default router,
638 * even when the multipath routing is available, because we're not sure about
639 * the benefits for stub hosts comparing to the risk of making the code
640 * complicated and the possibility of introducing bugs.
641 */
642 void
643 nd6_defrouter_select(void)
644 {
645 struct nd_ifinfo *ndi;
646 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
647
648 ND6_ASSERT_WLOCK();
649
650 /*
651 * This function should be called only when acting as an autoconfigured
652 * host. Although the remaining part of this function is not effective
653 * if the node is not an autoconfigured host, we explicitly exclude
654 * such cases here for safety.
655 */
656 if (ip6_forwarding) {
657 nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, "
658 "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv);
659 return;
660 }
661
662 /*
663 * Let's handle easy case (3) first:
664 * If default router list is empty, there's nothing to be done.
665 */
666 if (ND_DEFROUTER_LIST_EMPTY())
667 return;
668
669 /*
670 * Search for a (probably) reachable router from the list.
671 * We just pick up the first reachable one (if any), assuming that
672 * the ordering rule of the list described in defrtrlist_update().
673 */
674 ND_DEFROUTER_LIST_FOREACH(dr) {
675 ndi = ND_IFINFO(dr->ifp);
676 if (nd6_accepts_rtadv(ndi))
677 continue;
678
679 if (selected_dr == NULL &&
680 nd6_is_llinfo_probreach(dr))
681 selected_dr = dr;
682
683 if (dr->installed && !installed_dr)
684 installed_dr = dr;
685 else if (dr->installed && installed_dr) {
686 /* this should not happen. warn for diagnosis. */
687 log(LOG_ERR, "nd6_defrouter_select: more than one router"
688 " is installed\n");
689 }
690 }
691 /*
692 * If none of the default routers was found to be reachable,
693 * round-robin the list regardless of preference.
694 * Otherwise, if we have an installed router, check if the selected
695 * (reachable) router should really be preferred to the installed one.
696 * We only prefer the new router when the old one is not reachable
697 * or when the new one has a really higher preference value.
698 */
699 if (selected_dr == NULL) {
700 if (installed_dr == NULL ||
701 ND_DEFROUTER_LIST_NEXT(installed_dr) == NULL)
702 selected_dr = ND_DEFROUTER_LIST_FIRST();
703 else
704 selected_dr = ND_DEFROUTER_LIST_NEXT(installed_dr);
705 } else if (installed_dr &&
706 nd6_is_llinfo_probreach(installed_dr) &&
707 rtpref(selected_dr) <= rtpref(installed_dr)) {
708 selected_dr = installed_dr;
709 }
710
711 /*
712 * If the selected router is different than the installed one,
713 * remove the installed router and install the selected one.
714 * Note that the selected router is never NULL here.
715 */
716 if (installed_dr != selected_dr) {
717 if (installed_dr)
718 defrouter_delreq(installed_dr);
719 defrouter_addreq(selected_dr);
720 }
721
722 return;
723 }
724
725 /*
726 * for default router selection
727 * regards router-preference field as a 2-bit signed integer
728 */
729 static int
730 rtpref(struct nd_defrouter *dr)
731 {
732 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
733 case ND_RA_FLAG_RTPREF_HIGH:
734 return (RTPREF_HIGH);
735 case ND_RA_FLAG_RTPREF_MEDIUM:
736 case ND_RA_FLAG_RTPREF_RSV:
737 return (RTPREF_MEDIUM);
738 case ND_RA_FLAG_RTPREF_LOW:
739 return (RTPREF_LOW);
740 default:
741 /*
742 * This case should never happen. If it did, it would mean a
743 * serious bug of kernel internal. We thus always bark here.
744 * Or, can we even panic?
745 */
746 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
747 return (RTPREF_INVALID);
748 }
749 /* NOTREACHED */
750 }
751
752 static struct nd_defrouter *
753 defrtrlist_update(struct nd_defrouter *newdr)
754 {
755 struct nd_defrouter *dr, *n, *ret = NULL;
756 struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
757
758 ND6_ASSERT_WLOCK();
759
760 if ((dr = nd6_defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
761 /* entry exists */
762 if (newdr->rtlifetime == 0) {
763 nd6_defrtrlist_del(dr, ext);
764 dr = NULL;
765 } else {
766 int oldpref = rtpref(dr);
767
768 /* override */
769 dr->flags = newdr->flags; /* xxx flag check */
770 dr->rtlifetime = newdr->rtlifetime;
771 dr->expire = newdr->expire;
772
773 /*
774 * If the preference does not change, there's no need
775 * to sort the entries.
776 */
777 if (rtpref(newdr) == oldpref) {
778 ret = dr;
779 goto out;
780 }
781
782 /*
783 * preferred router may be changed, so relocate
784 * this router.
785 * XXX: calling TAILQ_REMOVE directly is a bad manner.
786 * However, since nd6_defrtrlist_del() has many side
787 * effects, we intentionally do so here.
788 * nd6_defrouter_select() below will handle routing
789 * changes later.
790 */
791 ND_DEFROUTER_LIST_REMOVE(dr);
792 n = dr;
793 goto insert;
794 }
795 ret = dr;
796 goto out;
797 }
798
799 if (ip6_maxifdefrouters >= 0 && ext->ndefrouters >= ip6_maxifdefrouters)
800 goto out;
801
802 /* entry does not exist */
803 if (newdr->rtlifetime == 0)
804 goto out;
805
806 if (ip6_rtadv_maxroutes <= nd6_numroutes) {
807 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
808 goto out;
809 }
810
811 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
812 if (n == NULL)
813 goto out;
814 memset(n, 0, sizeof(*n));
815 *n = *newdr;
816
817 insert:
818 /*
819 * Insert the new router in the Default Router List;
820 * The Default Router List should be in the descending order
821 * of router-preferece. Routers with the same preference are
822 * sorted in the arriving time order.
823 */
824
825 /* insert at the end of the group */
826 ND_DEFROUTER_LIST_FOREACH(dr) {
827 if (rtpref(n) > rtpref(dr))
828 break;
829 }
830 if (dr)
831 ND_DEFROUTER_LIST_INSERT_BEFORE(dr, n);
832 else
833 ND_DEFROUTER_LIST_INSERT_TAIL(n);
834
835 nd6_defrouter_select();
836
837 ext->ndefrouters++;
838
839 ret = n;
840 out:
841 return ret;
842 }
843
844 static struct nd_pfxrouter *
845 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
846 {
847 struct nd_pfxrouter *search;
848
849 ND6_ASSERT_LOCK();
850
851 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
852 if (search->router == dr)
853 break;
854 }
855
856 return (search);
857 }
858
859 static void
860 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
861 {
862 struct nd_pfxrouter *newpfr;
863
864 ND6_ASSERT_WLOCK();
865
866 newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
867 if (newpfr == NULL)
868 return;
869 newpfr->router = dr;
870
871 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
872
873 nd6_pfxlist_onlink_check();
874 }
875
876 static void
877 pfxrtr_del(struct nd_pfxrouter *pfr)
878 {
879 LIST_REMOVE(pfr, pfr_entry);
880 free(pfr, M_IP6NDP);
881 }
882
883 static struct nd_prefix *
884 nd6_prefix_lookup(struct nd_prefixctl *key)
885 {
886 struct nd_prefix *search;
887
888 ND_PREFIX_LIST_FOREACH(search) {
889 if (key->ndprc_ifp == search->ndpr_ifp &&
890 key->ndprc_plen == search->ndpr_plen &&
891 in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
892 &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
893 break;
894 }
895 }
896
897 return (search);
898 }
899
900 static void
901 purge_detached(struct ifnet *ifp)
902 {
903 struct nd_prefix *pr, *pr_next;
904 struct in6_ifaddr *ia;
905 struct ifaddr *ifa, *ifa_next;
906
907 restart:
908 ND6_ASSERT_WLOCK();
909
910 ND_PREFIX_LIST_FOREACH_SAFE(pr, pr_next) {
911 int s;
912
913 /*
914 * This function is called when we need to make more room for
915 * new prefixes rather than keeping old, possibly stale ones.
916 * Detached prefixes would be a good candidate; if all routers
917 * that advertised the prefix expired, the prefix is also
918 * probably stale.
919 */
920 if (pr->ndpr_ifp != ifp ||
921 IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
922 ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
923 !LIST_EMPTY(&pr->ndpr_advrtrs)))
924 continue;
925
926 s = pserialize_read_enter();
927 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) {
928 ifa_next = IFADDR_READER_NEXT(ifa);
929 if (ifa->ifa_addr->sa_family != AF_INET6)
930 continue;
931 ia = (struct in6_ifaddr *)ifa;
932 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
933 IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
934 pserialize_read_exit(s);
935 ND6_UNLOCK();
936
937 /* in6_purgeaddr may destroy pr. */
938 in6_purgeaddr(ifa);
939
940 ND6_WLOCK();
941 goto restart;
942 }
943 }
944 pserialize_read_exit(s);
945
946 KASSERT(pr->ndpr_refcnt == 0);
947 nd6_prelist_remove(pr);
948 }
949 }
950
951 static int
952 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
953 struct nd_prefix **newp)
954 {
955 struct nd_prefix *newpr = NULL;
956 int i, s;
957 int error;
958 struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
959
960 ND6_ASSERT_WLOCK();
961
962 if (ip6_maxifprefixes >= 0) {
963 if (ext->nprefixes >= ip6_maxifprefixes / 2)
964 purge_detached(prc->ndprc_ifp);
965 if (ext->nprefixes >= ip6_maxifprefixes)
966 return ENOMEM;
967 }
968
969 error = 0;
970 newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
971 if (newpr == NULL)
972 return ENOMEM;
973 newpr->ndpr_ifp = prc->ndprc_ifp;
974 newpr->ndpr_prefix = prc->ndprc_prefix;
975 newpr->ndpr_plen = prc->ndprc_plen;
976 newpr->ndpr_vltime = prc->ndprc_vltime;
977 newpr->ndpr_pltime = prc->ndprc_pltime;
978 newpr->ndpr_flags = prc->ndprc_flags;
979 if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
980 free(newpr, M_IP6NDP);
981 return(error);
982 }
983 newpr->ndpr_lastupdate = time_uptime;
984 if (newp != NULL)
985 *newp = newpr;
986
987 /* initialization */
988 LIST_INIT(&newpr->ndpr_advrtrs);
989 in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
990 /* make prefix in the canonical form */
991 for (i = 0; i < 4; i++) {
992 newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
993 newpr->ndpr_mask.s6_addr32[i];
994 }
995
996 s = splsoftnet();
997 /* link ndpr_entry to nd_prefix list */
998 ND_PREFIX_LIST_INSERT_HEAD(newpr);
999 splx(s);
1000
1001 /* ND_OPT_PI_FLAG_ONLINK processing */
1002 if (newpr->ndpr_raf_onlink) {
1003 int e;
1004
1005 if ((e = nd6_prefix_onlink(newpr)) != 0) {
1006 nd6log(LOG_ERR, "failed to make "
1007 "the prefix %s/%d on-link on %s (errno=%d)\n",
1008 ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
1009 prc->ndprc_plen, if_name(prc->ndprc_ifp), e);
1010 /* proceed anyway. XXX: is it correct? */
1011 }
1012 }
1013
1014 if (dr)
1015 pfxrtr_add(newpr, dr);
1016
1017 ext->nprefixes++;
1018
1019 return 0;
1020 }
1021
1022 void
1023 nd6_prefix_unref(struct nd_prefix *pr)
1024 {
1025
1026 ND6_WLOCK();
1027 pr->ndpr_refcnt--;
1028 if (pr->ndpr_refcnt == 0)
1029 nd6_prelist_remove(pr);
1030 ND6_UNLOCK();
1031 }
1032
1033 void
1034 nd6_prelist_remove(struct nd_prefix *pr)
1035 {
1036 struct nd_pfxrouter *pfr, *next;
1037 int e, s;
1038 struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
1039
1040 ND6_ASSERT_WLOCK();
1041 KASSERT(pr->ndpr_refcnt == 0);
1042
1043 /* make sure to invalidate the prefix until it is really freed. */
1044 pr->ndpr_vltime = 0;
1045 pr->ndpr_pltime = 0;
1046 #if 0
1047 /*
1048 * Though these flags are now meaningless, we'd rather keep the value
1049 * not to confuse users when executing "ndp -p".
1050 */
1051 pr->ndpr_raf_onlink = 0;
1052 pr->ndpr_raf_auto = 0;
1053 #endif
1054 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
1055 (e = nd6_prefix_offlink(pr)) != 0) {
1056 nd6log(LOG_ERR,
1057 "failed to make %s/%d offlink on %s, errno=%d\n",
1058 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1059 pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1060 /* what should we do? */
1061 }
1062
1063 s = splsoftnet();
1064 /* unlink ndpr_entry from nd_prefix list */
1065 ND_PREFIX_LIST_REMOVE(pr);
1066
1067 /* free list of routers that adversed the prefix */
1068 for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
1069 next = LIST_NEXT(pfr, pfr_entry);
1070
1071 free(pfr, M_IP6NDP);
1072 }
1073
1074 if (ext) {
1075 ext->nprefixes--;
1076 if (ext->nprefixes < 0) {
1077 log(LOG_WARNING, "nd6_prelist_remove: negative count on "
1078 "%s\n", pr->ndpr_ifp->if_xname);
1079 }
1080 }
1081 splx(s);
1082
1083 free(pr, M_IP6NDP);
1084
1085 nd6_pfxlist_onlink_check();
1086 }
1087
1088 static int
1089 prelist_update(struct nd_prefixctl *newprc,
1090 struct nd_defrouter *dr, /* may be NULL */
1091 struct mbuf *m,
1092 int mcast)
1093 {
1094 struct in6_ifaddr *ia6_match = NULL;
1095 struct ifaddr *ifa;
1096 struct ifnet *ifp = newprc->ndprc_ifp;
1097 struct nd_prefix *pr;
1098 int error = 0;
1099 int auth;
1100 struct in6_addrlifetime lt6_tmp;
1101 int ss;
1102
1103 ND6_ASSERT_WLOCK();
1104
1105 auth = 0;
1106 if (m) {
1107 /*
1108 * Authenticity for NA consists authentication for
1109 * both IP header and IP datagrams, doesn't it ?
1110 */
1111 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1112 auth = (m->m_flags & M_AUTHIPHDR
1113 && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1114 #endif
1115 }
1116
1117 if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
1118 /*
1119 * nd6_prefix_lookup() ensures that pr and newprc have the same
1120 * prefix on a same interface.
1121 */
1122
1123 /*
1124 * Update prefix information. Note that the on-link (L) bit
1125 * and the autonomous (A) bit should NOT be changed from 1
1126 * to 0.
1127 */
1128 if (newprc->ndprc_raf_onlink == 1)
1129 pr->ndpr_raf_onlink = 1;
1130 if (newprc->ndprc_raf_auto == 1)
1131 pr->ndpr_raf_auto = 1;
1132 if (newprc->ndprc_raf_onlink) {
1133 pr->ndpr_vltime = newprc->ndprc_vltime;
1134 pr->ndpr_pltime = newprc->ndprc_pltime;
1135 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1136 pr->ndpr_lastupdate = time_uptime;
1137 }
1138
1139 if (newprc->ndprc_raf_onlink &&
1140 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1141 int e;
1142
1143 if ((e = nd6_prefix_onlink(pr)) != 0) {
1144 nd6log(LOG_ERR,
1145 "failed to make "
1146 "the prefix %s/%d on-link on %s "
1147 "(errno=%d)\n",
1148 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1149 pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1150 /* proceed anyway. XXX: is it correct? */
1151 }
1152 }
1153
1154 if (dr && pfxrtr_lookup(pr, dr) == NULL)
1155 pfxrtr_add(pr, dr);
1156 } else {
1157 struct nd_prefix *newpr = NULL;
1158
1159 if (newprc->ndprc_vltime == 0)
1160 goto end;
1161 if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
1162 goto end;
1163
1164 if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1165 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1166 goto end;
1167 }
1168
1169 error = nd6_prelist_add(newprc, dr, &newpr);
1170 if (error != 0 || newpr == NULL) {
1171 nd6log(LOG_NOTICE,
1172 "nd6_prelist_add failed for %s/%d on %s "
1173 "errno=%d, returnpr=%p\n",
1174 ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
1175 newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
1176 error, newpr);
1177 goto end; /* we should just give up in this case. */
1178 }
1179
1180 /*
1181 * XXX: from the ND point of view, we can ignore a prefix
1182 * with the on-link bit being zero. However, we need a
1183 * prefix structure for references from autoconfigured
1184 * addresses. Thus, we explicitly make sure that the prefix
1185 * itself expires now.
1186 */
1187 if (newpr->ndpr_raf_onlink == 0) {
1188 newpr->ndpr_vltime = 0;
1189 newpr->ndpr_pltime = 0;
1190 in6_init_prefix_ltimes(newpr);
1191 }
1192
1193 pr = newpr;
1194 }
1195
1196 /*
1197 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1198 * Note that pr must be non NULL at this point.
1199 */
1200
1201 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1202 if (!newprc->ndprc_raf_auto)
1203 goto end;
1204
1205 /*
1206 * 5.5.3 (b). the link-local prefix should have been ignored in
1207 * nd6_ra_input.
1208 */
1209
1210 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1211 if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
1212 error = EINVAL; /* XXX: won't be used */
1213 goto end;
1214 }
1215
1216 /*
1217 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1218 * an address configured by stateless autoconfiguration already in the
1219 * list of addresses associated with the interface, and the Valid
1220 * Lifetime is not 0, form an address. We first check if we have
1221 * a matching prefix.
1222 * Note: we apply a clarification in rfc2462bis-02 here. We only
1223 * consider autoconfigured addresses while RFC2462 simply said
1224 * "address".
1225 */
1226 ss = pserialize_read_enter();
1227 IFADDR_READER_FOREACH(ifa, ifp) {
1228 struct in6_ifaddr *ia6;
1229 u_int32_t remaininglifetime;
1230
1231 if (ifa->ifa_addr->sa_family != AF_INET6)
1232 continue;
1233
1234 ia6 = (struct in6_ifaddr *)ifa;
1235
1236 /*
1237 * We only consider autoconfigured addresses as per rfc2462bis.
1238 */
1239 if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF))
1240 continue;
1241
1242 /*
1243 * Spec is not clear here, but I believe we should concentrate
1244 * on unicast (i.e. not anycast) addresses.
1245 * XXX: other ia6_flags? detached or duplicated?
1246 */
1247 if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1248 continue;
1249
1250 /*
1251 * Ignore the address if it is not associated with a prefix
1252 * or is associated with a prefix that is different from this
1253 * one. (pr is never NULL here)
1254 */
1255 if (ia6->ia6_ndpr != pr)
1256 continue;
1257
1258 if (ia6_match == NULL) /* remember the first one */
1259 ia6_match = ia6;
1260
1261 /*
1262 * An already autoconfigured address matched. Now that we
1263 * are sure there is at least one matched address, we can
1264 * proceed to 5.5.3. (e): update the lifetimes according to the
1265 * "two hours" rule and the privacy extension.
1266 * We apply some clarifications in rfc2462bis:
1267 * - use remaininglifetime instead of storedlifetime as a
1268 * variable name
1269 * - remove the dead code in the "two-hour" rule
1270 */
1271 #define TWOHOUR (120*60)
1272 lt6_tmp = ia6->ia6_lifetime;
1273 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1274 remaininglifetime = ND6_INFINITE_LIFETIME;
1275 else if (time_uptime - ia6->ia6_updatetime >
1276 lt6_tmp.ia6t_vltime) {
1277 /*
1278 * The case of "invalid" address. We should usually
1279 * not see this case.
1280 */
1281 remaininglifetime = 0;
1282 } else
1283 remaininglifetime = lt6_tmp.ia6t_vltime -
1284 (time_uptime - ia6->ia6_updatetime);
1285
1286 /* when not updating, keep the current stored lifetime. */
1287 lt6_tmp.ia6t_vltime = remaininglifetime;
1288
1289 if (TWOHOUR < newprc->ndprc_vltime ||
1290 remaininglifetime < newprc->ndprc_vltime) {
1291 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1292 } else if (remaininglifetime <= TWOHOUR) {
1293 if (auth)
1294 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1295 } else {
1296 /*
1297 * newprc->ndprc_vltime <= TWOHOUR &&
1298 * TWOHOUR < remaininglifetime
1299 */
1300 lt6_tmp.ia6t_vltime = TWOHOUR;
1301 }
1302
1303 /* The 2 hour rule is not imposed for preferred lifetime. */
1304 lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
1305
1306 in6_init_address_ltimes(pr, <6_tmp);
1307
1308 /*
1309 * We need to treat lifetimes for temporary addresses
1310 * differently, according to
1311 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1312 * we only update the lifetimes when they are in the maximum
1313 * intervals.
1314 */
1315 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1316 u_int32_t maxvltime, maxpltime;
1317
1318 if (ip6_temp_valid_lifetime >
1319 (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1320 ip6_desync_factor)) {
1321 maxvltime = ip6_temp_valid_lifetime -
1322 (time_uptime - ia6->ia6_createtime) -
1323 ip6_desync_factor;
1324 } else
1325 maxvltime = 0;
1326 if (ip6_temp_preferred_lifetime >
1327 (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1328 ip6_desync_factor)) {
1329 maxpltime = ip6_temp_preferred_lifetime -
1330 (time_uptime - ia6->ia6_createtime) -
1331 ip6_desync_factor;
1332 } else
1333 maxpltime = 0;
1334
1335 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1336 lt6_tmp.ia6t_vltime > maxvltime) {
1337 lt6_tmp.ia6t_vltime = maxvltime;
1338 }
1339 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1340 lt6_tmp.ia6t_pltime > maxpltime) {
1341 lt6_tmp.ia6t_pltime = maxpltime;
1342 }
1343 }
1344
1345 ia6->ia6_lifetime = lt6_tmp;
1346 ia6->ia6_updatetime = time_uptime;
1347 }
1348 pserialize_read_exit(ss);
1349
1350 if (ia6_match == NULL && newprc->ndprc_vltime) {
1351 int ifidlen;
1352 struct in6_ifaddr *ia6;
1353 struct psref psref;
1354
1355 /*
1356 * 5.5.3 (d) (continued)
1357 * No address matched and the valid lifetime is non-zero.
1358 * Create a new address.
1359 */
1360
1361 /*
1362 * Prefix Length check:
1363 * If the sum of the prefix length and interface identifier
1364 * length does not equal 128 bits, the Prefix Information
1365 * option MUST be ignored. The length of the interface
1366 * identifier is defined in a separate link-type specific
1367 * document.
1368 */
1369 ifidlen = in6_if2idlen(ifp);
1370 if (ifidlen < 0) {
1371 /* this should not happen, so we always log it. */
1372 log(LOG_ERR, "%s: IFID undefined (%s)\n",
1373 __func__, if_name(ifp));
1374 goto end;
1375 }
1376 if (ifidlen + pr->ndpr_plen != 128) {
1377 nd6log(LOG_INFO,
1378 "invalid prefixlen %d for %s, ignored\n",
1379 pr->ndpr_plen, if_name(ifp));
1380 goto end;
1381 }
1382
1383 if ((ia6 = in6_ifadd(newprc, mcast, &psref)) != NULL) {
1384 /*
1385 * note that we should use pr (not newprc) for reference.
1386 */
1387 pr->ndpr_refcnt++;
1388 ia6->ia6_ndpr = pr;
1389
1390 /* toggle onlink state if the address was assigned
1391 * a prefix route. */
1392 if (ia6->ia_flags & IFA_ROUTE)
1393 pr->ndpr_stateflags |= NDPRF_ONLINK;
1394
1395 /*
1396 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1397 * When a new public address is created as described
1398 * in RFC2462, also create a new temporary address.
1399 *
1400 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1401 * When an interface connects to a new link, a new
1402 * randomized interface identifier should be generated
1403 * immediately together with a new set of temporary
1404 * addresses. Thus, we specifiy 1 as the 2nd arg of
1405 * in6_tmpifadd().
1406 */
1407 if (ip6_use_tempaddr) {
1408 int e;
1409 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1410 nd6log(LOG_NOTICE,
1411 "failed to create a temporary "
1412 "address, errno=%d\n", e);
1413 }
1414 }
1415 ia6_release(ia6, &psref);
1416
1417 /*
1418 * A newly added address might affect the status
1419 * of other addresses, so we check and update it.
1420 * XXX: what if address duplication happens?
1421 */
1422 nd6_pfxlist_onlink_check();
1423 } else {
1424 /* just set an error. do not bark here. */
1425 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1426 }
1427 }
1428
1429 end:
1430 return error;
1431 }
1432
1433 /*
1434 * A supplement function used in the on-link detection below;
1435 * detect if a given prefix has a (probably) reachable advertising router.
1436 * XXX: lengthy function name...
1437 */
1438 static struct nd_pfxrouter *
1439 find_pfxlist_reachable_router(struct nd_prefix *pr)
1440 {
1441 struct nd_pfxrouter *pfxrtr;
1442
1443 for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1444 pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1445 if (pfxrtr->router->ifp->if_flags & IFF_UP &&
1446 pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
1447 nd6_is_llinfo_probreach(pfxrtr->router))
1448 break; /* found */
1449 }
1450
1451 return (pfxrtr);
1452 }
1453
1454 /*
1455 * Check if each prefix in the prefix list has at least one available router
1456 * that advertised the prefix (a router is "available" if its neighbor cache
1457 * entry is reachable or probably reachable).
1458 * If the check fails, the prefix may be off-link, because, for example,
1459 * we have moved from the network but the lifetime of the prefix has not
1460 * expired yet. So we should not use the prefix if there is another prefix
1461 * that has an available router.
1462 * But, if there is no prefix that has an available router, we still regards
1463 * all the prefixes as on-link. This is because we can't tell if all the
1464 * routers are simply dead or if we really moved from the network and there
1465 * is no router around us.
1466 */
1467 void
1468 nd6_pfxlist_onlink_check(void)
1469 {
1470 struct nd_prefix *pr;
1471 struct in6_ifaddr *ia;
1472 struct nd_defrouter *dr;
1473 struct nd_pfxrouter *pfxrtr = NULL;
1474 int s;
1475
1476 ND6_ASSERT_WLOCK();
1477
1478 /*
1479 * Check if there is a prefix that has a reachable advertising
1480 * router.
1481 */
1482 ND_PREFIX_LIST_FOREACH(pr) {
1483 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1484 break;
1485 }
1486 /*
1487 * If we have no such prefix, check whether we still have a router
1488 * that does not advertise any prefixes.
1489 */
1490 if (pr == NULL) {
1491 ND_DEFROUTER_LIST_FOREACH(dr) {
1492 struct nd_prefix *pr0;
1493
1494 ND_PREFIX_LIST_FOREACH(pr0) {
1495 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1496 break;
1497 }
1498 if (pfxrtr)
1499 break;
1500 }
1501 }
1502 if (pr != NULL || (!ND_DEFROUTER_LIST_EMPTY() && !pfxrtr)) {
1503 /*
1504 * There is at least one prefix that has a reachable router,
1505 * or at least a router which probably does not advertise
1506 * any prefixes. The latter would be the case when we move
1507 * to a new link where we have a router that does not provide
1508 * prefixes and we configure an address by hand.
1509 * Detach prefixes which have no reachable advertising
1510 * router, and attach other prefixes.
1511 */
1512 ND_PREFIX_LIST_FOREACH(pr) {
1513 /* XXX: a link-local prefix should never be detached */
1514 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1515 continue;
1516
1517 /*
1518 * we aren't interested in prefixes without the L bit
1519 * set.
1520 */
1521 if (pr->ndpr_raf_onlink == 0)
1522 continue;
1523
1524 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1525 find_pfxlist_reachable_router(pr) == NULL)
1526 pr->ndpr_stateflags |= NDPRF_DETACHED;
1527 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1528 find_pfxlist_reachable_router(pr) != 0)
1529 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1530 }
1531 } else {
1532 /* there is no prefix that has a reachable router */
1533 ND_PREFIX_LIST_FOREACH(pr) {
1534 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1535 continue;
1536
1537 if (pr->ndpr_raf_onlink == 0)
1538 continue;
1539
1540 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1541 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1542 }
1543 }
1544
1545 /*
1546 * Remove each interface route associated with a (just) detached
1547 * prefix, and reinstall the interface route for a (just) attached
1548 * prefix. Note that all attempt of reinstallation does not
1549 * necessarily success, when a same prefix is shared among multiple
1550 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1551 * so we don't have to care about them.
1552 */
1553 ND_PREFIX_LIST_FOREACH(pr) {
1554 int e;
1555
1556 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1557 continue;
1558
1559 if (pr->ndpr_raf_onlink == 0)
1560 continue;
1561
1562 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1563 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1564 if ((e = nd6_prefix_offlink(pr)) != 0) {
1565 nd6log(LOG_ERR,
1566 "failed to make %s/%d offlink, errno=%d\n",
1567 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1568 pr->ndpr_plen, e);
1569 }
1570 }
1571 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1572 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1573 pr->ndpr_raf_onlink) {
1574 if ((e = nd6_prefix_onlink(pr)) != 0) {
1575 nd6log(LOG_ERR,
1576 "failed to make %s/%d onlink, errno=%d\n",
1577 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1578 pr->ndpr_plen, e);
1579 }
1580 }
1581 }
1582
1583 /*
1584 * Changes on the prefix status might affect address status as well.
1585 * Make sure that all addresses derived from an attached prefix are
1586 * attached, and that all addresses derived from a detached prefix are
1587 * detached. Note, however, that a manually configured address should
1588 * always be attached.
1589 * The precise detection logic is same as the one for prefixes.
1590 */
1591 s = pserialize_read_enter();
1592 IN6_ADDRLIST_READER_FOREACH(ia) {
1593 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
1594 continue;
1595
1596 if (ia->ia6_ndpr == NULL) {
1597 /*
1598 * This can happen when we first configure the address
1599 * (i.e. the address exists, but the prefix does not).
1600 * XXX: complicated relationships...
1601 */
1602 continue;
1603 }
1604
1605 if (find_pfxlist_reachable_router(ia->ia6_ndpr))
1606 break;
1607 }
1608 pserialize_read_exit(s);
1609
1610 if (ia) {
1611 int bound = curlwp_bind();
1612
1613 s = pserialize_read_enter();
1614 IN6_ADDRLIST_READER_FOREACH(ia) {
1615 struct ifaddr *ifa = (struct ifaddr *)ia;
1616 struct psref psref;
1617
1618 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1619 continue;
1620
1621 if (ia->ia6_ndpr == NULL) /* XXX: see above. */
1622 continue;
1623
1624 ia6_acquire(ia, &psref);
1625 pserialize_read_exit(s);
1626
1627 if (find_pfxlist_reachable_router(ia->ia6_ndpr)) {
1628 if (ia->ia6_flags & IN6_IFF_DETACHED) {
1629 ia->ia6_flags &= ~IN6_IFF_DETACHED;
1630 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1631 nd6_dad_start(ifa,
1632 0);
1633 /* We will notify the routing socket
1634 * of the DAD result, so no need to
1635 * here */
1636 }
1637 } else {
1638 if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) {
1639 ia->ia6_flags |= IN6_IFF_DETACHED;
1640 rt_newaddrmsg(RTM_NEWADDR,
1641 ifa, 0, NULL);
1642 }
1643 }
1644
1645 s = pserialize_read_enter();
1646 ia6_release(ia, &psref);
1647 }
1648 pserialize_read_exit(s);
1649 curlwp_bindx(bound);
1650 }
1651 else {
1652 int bound = curlwp_bind();
1653
1654 s = pserialize_read_enter();
1655 IN6_ADDRLIST_READER_FOREACH(ia) {
1656 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1657 continue;
1658
1659 if (ia->ia6_flags & IN6_IFF_DETACHED) {
1660 struct ifaddr *ifa = (struct ifaddr *)ia;
1661 struct psref psref;
1662
1663 ia->ia6_flags &= ~IN6_IFF_DETACHED;
1664 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1665
1666 ia6_acquire(ia, &psref);
1667 pserialize_read_exit(s);
1668
1669 /* Do we need a delay in this case? */
1670 nd6_dad_start(ifa, 0);
1671
1672 s = pserialize_read_enter();
1673 ia6_release(ia, &psref);
1674 }
1675 }
1676 pserialize_read_exit(s);
1677 curlwp_bindx(bound);
1678 }
1679 }
1680
1681 static int
1682 nd6_prefix_onlink(struct nd_prefix *pr)
1683 {
1684 struct ifaddr *ifa;
1685 struct ifnet *ifp = pr->ndpr_ifp;
1686 struct sockaddr_in6 mask6;
1687 struct nd_prefix *opr;
1688 u_long rtflags;
1689 int error = 0;
1690 struct psref psref;
1691 int bound;
1692
1693 ND6_ASSERT_WLOCK();
1694
1695 /* sanity check */
1696 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1697 nd6log(LOG_ERR, "%s/%d is already on-link\n",
1698 ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
1699 return (EEXIST);
1700 }
1701
1702 /*
1703 * Add the interface route associated with the prefix. Before
1704 * installing the route, check if there's the same prefix on another
1705 * interface, and the prefix has already installed the interface route.
1706 * Although such a configuration is expected to be rare, we explicitly
1707 * allow it.
1708 */
1709 ND_PREFIX_LIST_FOREACH(opr) {
1710 if (opr == pr)
1711 continue;
1712
1713 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1714 continue;
1715
1716 if (opr->ndpr_plen == pr->ndpr_plen &&
1717 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1718 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1719 return (0);
1720 }
1721
1722 /*
1723 * We prefer link-local addresses as the associated interface address.
1724 */
1725 /* search for a link-local addr */
1726 bound = curlwp_bind();
1727 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
1728 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST, &psref);
1729 if (ifa == NULL) {
1730 int s = pserialize_read_enter();
1731 IFADDR_READER_FOREACH(ifa, ifp) {
1732 if (ifa->ifa_addr->sa_family == AF_INET6)
1733 break;
1734 }
1735 if (ifa != NULL)
1736 ifa_acquire(ifa, &psref);
1737 pserialize_read_exit(s);
1738 /* should we care about ia6_flags? */
1739 }
1740 if (ifa == NULL) {
1741 /*
1742 * This can still happen, when, for example, we receive an RA
1743 * containing a prefix with the L bit set and the A bit clear,
1744 * after removing all IPv6 addresses on the receiving
1745 * interface. This should, of course, be rare though.
1746 */
1747 nd6log(LOG_NOTICE, "failed to find any ifaddr"
1748 " to add route for a prefix(%s/%d) on %s\n",
1749 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1750 pr->ndpr_plen, if_name(ifp));
1751 curlwp_bindx(bound);
1752 return (0);
1753 }
1754
1755 /*
1756 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1757 * ifa->ifa_rtrequest = nd6_rtrequest;
1758 */
1759 memset(&mask6, 0, sizeof(mask6));
1760 mask6.sin6_family = AF_INET6;
1761 mask6.sin6_len = sizeof(mask6);
1762 mask6.sin6_addr = pr->ndpr_mask;
1763 /* rtrequest() will probably set RTF_UP, but we're not sure. */
1764 rtflags = ifa->ifa_flags | RTF_UP;
1765 if (nd6_need_cache(ifp)) {
1766 /* explicitly set in case ifa_flags does not set the flag. */
1767 rtflags |= RTF_CONNECTED;
1768 } else {
1769 /*
1770 * explicitly clear the cloning bit in case ifa_flags sets it.
1771 */
1772 rtflags &= ~RTF_CONNECTED;
1773 }
1774 error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix),
1775 ifa->ifa_addr, sin6tosa(&mask6), rtflags);
1776 if (error == 0) {
1777 nd6_numroutes++;
1778 pr->ndpr_stateflags |= NDPRF_ONLINK;
1779 } else {
1780 nd6log(LOG_ERR, "failed to add route for a"
1781 " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1782 "errno = %d\n",
1783 ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1784 pr->ndpr_plen, if_name(ifp),
1785 ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1786 ip6_sprintf(&mask6.sin6_addr), rtflags, error);
1787 }
1788 ifa_release(ifa, &psref);
1789 curlwp_bindx(bound);
1790
1791 return (error);
1792 }
1793
1794 static int
1795 nd6_prefix_offlink(struct nd_prefix *pr)
1796 {
1797 int error = 0;
1798 struct ifnet *ifp = pr->ndpr_ifp;
1799 struct nd_prefix *opr;
1800 struct sockaddr_in6 sa6, mask6;
1801
1802 ND6_ASSERT_WLOCK();
1803
1804 /* sanity check */
1805 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1806 nd6log(LOG_ERR, "%s/%d is already off-link\n",
1807 ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
1808 return (EEXIST);
1809 }
1810
1811 sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1812 sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1813 error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL,
1814 sin6tosa(&mask6), 0);
1815 if (error == 0) {
1816 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1817 nd6_numroutes--;
1818
1819 /*
1820 * There might be the same prefix on another interface,
1821 * the prefix which could not be on-link just because we have
1822 * the interface route (see comments in nd6_prefix_onlink).
1823 * If there's one, try to make the prefix on-link on the
1824 * interface.
1825 */
1826 ND_PREFIX_LIST_FOREACH(opr) {
1827 if (opr == pr)
1828 continue;
1829
1830 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1831 continue;
1832
1833 /*
1834 * KAME specific: detached prefixes should not be
1835 * on-link.
1836 */
1837 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1838 continue;
1839
1840 if (opr->ndpr_plen == pr->ndpr_plen &&
1841 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1842 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1843 int e;
1844
1845 if ((e = nd6_prefix_onlink(opr)) != 0) {
1846 nd6log(LOG_ERR, "failed to "
1847 "recover a prefix %s/%d from %s "
1848 "to %s (errno = %d)\n",
1849 ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1850 opr->ndpr_plen, if_name(ifp),
1851 if_name(opr->ndpr_ifp), e);
1852 }
1853 }
1854 }
1855 } else {
1856 /* XXX: can we still set the NDPRF_ONLINK flag? */
1857 nd6log(LOG_ERR, "failed to delete route: "
1858 "%s/%d on %s (errno = %d)\n",
1859 ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1860 error);
1861 }
1862
1863 return error;
1864 }
1865
1866 static struct in6_ifaddr *
1867 in6_ifadd(struct nd_prefixctl *prc, int mcast, struct psref *psref)
1868 {
1869 struct ifnet *ifp = prc->ndprc_ifp;
1870 struct ifaddr *ifa;
1871 struct in6_aliasreq ifra;
1872 struct in6_ifaddr *ia, *ib;
1873 int error, plen0;
1874 struct in6_addr mask;
1875 int prefixlen = prc->ndprc_plen;
1876 int updateflags;
1877 int s;
1878
1879 ND6_ASSERT_WLOCK();
1880
1881 in6_prefixlen2mask(&mask, prefixlen);
1882
1883 /*
1884 * find a link-local address (will be interface ID).
1885 * Is it really mandatory? Theoretically, a global or a site-local
1886 * address can be configured without a link-local address, if we
1887 * have a unique interface identifier...
1888 *
1889 * it is not mandatory to have a link-local address, we can generate
1890 * interface identifier on the fly. we do this because:
1891 * (1) it should be the easiest way to find interface identifier.
1892 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1893 * for multiple addresses on a single interface, and possible shortcut
1894 * of DAD. we omitted DAD for this reason in the past.
1895 * (3) a user can prevent autoconfiguration of global address
1896 * by removing link-local address by hand (this is partly because we
1897 * don't have other way to control the use of IPv6 on an interface.
1898 * this has been our design choice - cf. NRL's "ifconfig auto").
1899 * (4) it is easier to manage when an interface has addresses
1900 * with the same interface identifier, than to have multiple addresses
1901 * with different interface identifiers.
1902 */
1903 s = pserialize_read_enter();
1904 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1905 if (ifa)
1906 ib = (struct in6_ifaddr *)ifa;
1907 else {
1908 pserialize_read_exit(s);
1909 return NULL;
1910 }
1911
1912 #if 0 /* don't care link local addr state, and always do DAD */
1913 /* if link-local address is not eligible, do not autoconfigure. */
1914 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1915 printf("in6_ifadd: link-local address not ready\n");
1916 return NULL;
1917 }
1918 #endif
1919
1920 /* prefixlen + ifidlen must be equal to 128 */
1921 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1922 if (prefixlen != plen0) {
1923 nd6log(LOG_INFO, "wrong prefixlen for %s "
1924 "(prefix=%d ifid=%d)\n",
1925 if_name(ifp), prefixlen, 128 - plen0);
1926 pserialize_read_exit(s);
1927 return NULL;
1928 }
1929
1930 /* make ifaddr */
1931
1932 memset(&ifra, 0, sizeof(ifra));
1933 /*
1934 * in6_update_ifa() does not use ifra_name, but we accurately set it
1935 * for safety.
1936 */
1937 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1938 sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
1939 /* prefix */
1940 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1941 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1942 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1943 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1944
1945 /* interface ID */
1946 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1947 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1948 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1949 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1950 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1951 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1952 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1953 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1954 pserialize_read_exit(s);
1955
1956 /* new prefix mask. */
1957 sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1958
1959 /* lifetimes */
1960 ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
1961 ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
1962
1963 /* XXX: scope zone ID? */
1964
1965 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1966
1967 /*
1968 * Make sure that we do not have this address already. This should
1969 * usually not happen, but we can still see this case, e.g., if we
1970 * have manually configured the exact address to be configured.
1971 */
1972 s = pserialize_read_enter();
1973 if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1974 /* this should be rare enough to make an explicit log */
1975 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1976 ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1977 pserialize_read_exit(s);
1978 return (NULL);
1979 }
1980 pserialize_read_exit(s);
1981
1982 /*
1983 * Allocate ifaddr structure, link into chain, etc.
1984 * If we are going to create a new address upon receiving a multicasted
1985 * RA, we need to impose a random delay before starting DAD.
1986 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1987 */
1988 updateflags = 0;
1989 if (mcast)
1990 updateflags |= IN6_IFAUPDATE_DADDELAY;
1991 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1992 nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n",
1993 ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1994 error);
1995 return (NULL); /* ifaddr must not have been allocated. */
1996 }
1997
1998 ia = in6ifa_ifpwithaddr_psref(ifp, &ifra.ifra_addr.sin6_addr, psref);
1999
2000 return (ia); /* this is always non-NULL */
2001 }
2002
2003 int
2004 in6_tmpifadd(
2005 const struct in6_ifaddr *ia0, /* corresponding public address */
2006 int forcegen,
2007 int dad_delay)
2008 {
2009 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2010 struct in6_ifaddr *newia, *ia;
2011 struct in6_aliasreq ifra;
2012 int i, error;
2013 int trylimit = 3; /* XXX: adhoc value */
2014 int updateflags;
2015 u_int32_t randid[2];
2016 u_int32_t vltime0, pltime0;
2017 int s;
2018
2019 ND6_ASSERT_WLOCK();
2020
2021 memset(&ifra, 0, sizeof(ifra));
2022 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
2023 ifra.ifra_addr = ia0->ia_addr;
2024 /* copy prefix mask */
2025 ifra.ifra_prefixmask = ia0->ia_prefixmask;
2026 /* clear the old IFID */
2027 for (i = 0; i < 4; i++) {
2028 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
2029 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
2030 }
2031
2032 again:
2033 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2034 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2035 nd6log(LOG_NOTICE, "failed to find a good random IFID\n");
2036 return (EINVAL);
2037 }
2038 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2039 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2040 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2041 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2042
2043 /*
2044 * in6_get_tmpifid() quite likely provided a unique interface ID.
2045 * However, we may still have a chance to see collision, because
2046 * there may be a time lag between generation of the ID and generation
2047 * of the address. So, we'll do one more sanity check.
2048 */
2049 s = pserialize_read_enter();
2050 IN6_ADDRLIST_READER_FOREACH(ia) {
2051 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2052 &ifra.ifra_addr.sin6_addr)) {
2053 pserialize_read_exit(s);
2054 if (trylimit-- == 0) {
2055 /*
2056 * Give up. Something strange should have
2057 * happened.
2058 */
2059 nd6log(LOG_NOTICE,
2060 "failed to find a unique random IFID\n");
2061 return (EEXIST);
2062 }
2063 forcegen = 1;
2064 goto again;
2065 }
2066 }
2067 pserialize_read_exit(s);
2068
2069 /*
2070 * The Valid Lifetime is the lower of the Valid Lifetime of the
2071 * public address or TEMP_VALID_LIFETIME.
2072 * The Preferred Lifetime is the lower of the Preferred Lifetime
2073 * of the public address or TEMP_PREFERRED_LIFETIME -
2074 * DESYNC_FACTOR.
2075 */
2076 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2077 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2078 (ia0->ia6_lifetime.ia6t_vltime -
2079 (time_uptime - ia0->ia6_updatetime));
2080 if (vltime0 > ip6_temp_valid_lifetime)
2081 vltime0 = ip6_temp_valid_lifetime;
2082 } else
2083 vltime0 = ip6_temp_valid_lifetime;
2084 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2085 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2086 (ia0->ia6_lifetime.ia6t_pltime -
2087 (time_uptime - ia0->ia6_updatetime));
2088 if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
2089 pltime0 = ip6_temp_preferred_lifetime -
2090 ip6_desync_factor;
2091 }
2092 } else
2093 pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
2094 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2095 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2096
2097 /*
2098 * A temporary address is created only if this calculated Preferred
2099 * Lifetime is greater than REGEN_ADVANCE time units.
2100 */
2101 if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
2102 return (0);
2103
2104 /* XXX: scope zone ID? */
2105
2106 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2107
2108 /* allocate ifaddr structure, link into chain, etc. */
2109 updateflags = 0;
2110 if (dad_delay)
2111 updateflags |= IN6_IFAUPDATE_DADDELAY;
2112 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2113 return (error);
2114
2115 s = pserialize_read_enter();
2116 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2117 if (newia == NULL) { /* XXX: can it happen? */
2118 pserialize_read_exit(s);
2119 nd6log(LOG_ERR,
2120 "ifa update succeeded, but we got no ifaddr\n");
2121 return (EINVAL); /* XXX */
2122 }
2123 newia->ia6_ndpr = ia0->ia6_ndpr;
2124 newia->ia6_ndpr->ndpr_refcnt++;
2125 pserialize_read_exit(s);
2126
2127 /*
2128 * A newly added address might affect the status of other addresses.
2129 * XXX: when the temporary address is generated with a new public
2130 * address, the onlink check is redundant. However, it would be safe
2131 * to do the check explicitly everywhere a new address is generated,
2132 * and, in fact, we surely need the check when we create a new
2133 * temporary address due to deprecation of an old temporary address.
2134 */
2135 nd6_pfxlist_onlink_check();
2136
2137 return (0);
2138 }
2139
2140 static int
2141 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2142 {
2143
2144 ND6_ASSERT_WLOCK();
2145
2146 /* check if preferred lifetime > valid lifetime. RFC2462 5.5.3 (c) */
2147 if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
2148 nd6log(LOG_INFO, "preferred lifetime"
2149 "(%d) is greater than valid lifetime(%d)\n",
2150 (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime);
2151 return (EINVAL);
2152 }
2153 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2154 ndpr->ndpr_preferred = 0;
2155 else
2156 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2157 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2158 ndpr->ndpr_expire = 0;
2159 else
2160 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2161
2162 return 0;
2163 }
2164
2165 static void
2166 in6_init_address_ltimes(struct nd_prefix *newpr,
2167 struct in6_addrlifetime *lt6)
2168 {
2169
2170 /* Valid lifetime must not be updated unless explicitly specified. */
2171 /* init ia6t_expire */
2172 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2173 lt6->ia6t_expire = 0;
2174 else {
2175 lt6->ia6t_expire = time_uptime;
2176 lt6->ia6t_expire += lt6->ia6t_vltime;
2177 }
2178
2179 /* init ia6t_preferred */
2180 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2181 lt6->ia6t_preferred = 0;
2182 else {
2183 lt6->ia6t_preferred = time_uptime;
2184 lt6->ia6t_preferred += lt6->ia6t_pltime;
2185 }
2186 }
2187
2188 /*
2189 * Delete all the routing table entries that use the specified gateway.
2190 * XXX: this function causes search through all entries of routing table, so
2191 * it shouldn't be called when acting as a router.
2192 */
2193 void
2194 nd6_rt_flush(struct in6_addr *gateway, struct ifnet *ifp)
2195 {
2196 int s = splsoftnet();
2197
2198 /* We'll care only link-local addresses */
2199 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2200 splx(s);
2201 return;
2202 }
2203
2204 rt_delete_matched_entries(AF_INET6, rt6_deleteroute_matcher, gateway);
2205 splx(s);
2206 }
2207
2208 static int
2209 rt6_deleteroute_matcher(struct rtentry *rt, void *arg)
2210 {
2211 struct in6_addr *gate = (struct in6_addr *)arg;
2212
2213 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2214 return (0);
2215
2216 if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
2217 return (0);
2218
2219 /*
2220 * Do not delete a static route.
2221 * XXX: this seems to be a bit ad-hoc. Should we consider the
2222 * 'cloned' bit instead?
2223 */
2224 if ((rt->rt_flags & RTF_STATIC) != 0)
2225 return (0);
2226
2227 /*
2228 * We delete only host route. This means, in particular, we don't
2229 * delete default route.
2230 */
2231 if ((rt->rt_flags & RTF_HOST) == 0)
2232 return (0);
2233
2234 return 1;
2235 }
2236