nd6_rtr.c revision 1.132 1 /* $NetBSD: nd6_rtr.c,v 1.132 2017/02/22 03:02:55 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 1.132 2017/02/22 03:02:55 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 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
149
150 ifp = m_get_rcvif_psref(m, &psref);
151 if (ifp == NULL)
152 goto freeit;
153
154 ndi = ND_IFINFO(ifp);
155
156 /* If I'm not a router, ignore it. */
157 if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
158 goto freeit;
159
160 /* Sanity checks */
161 if (ip6->ip6_hlim != 255) {
162 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
163 ip6->ip6_hlim, IN6_PRINT(ip6bufs, &ip6->ip6_src),
164 IN6_PRINT(ip6bufd, &ip6->ip6_dst), if_name(ifp));
165 goto bad;
166 }
167
168 /*
169 * Don't update the neighbor cache, if src = ::.
170 * This indicates that the src has no IP address assigned yet.
171 */
172 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
173 goto freeit;
174
175 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
176 if (nd_rs == NULL) {
177 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
178 return;
179 }
180
181 icmp6len -= sizeof(*nd_rs);
182 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
183 if (nd6_options(&ndopts) < 0) {
184 nd6log(LOG_INFO, "invalid ND option, ignored\n");
185 /* nd6_options have incremented stats */
186 goto freeit;
187 }
188
189 if (ndopts.nd_opts_src_lladdr) {
190 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
191 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
192 }
193
194 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
195 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
196 "(if %d, RS packet %d)\n",
197 IN6_PRINT(ip6bufs, &saddr6),
198 ifp->if_addrlen, lladdrlen - 2);
199 goto bad;
200 }
201
202 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
203
204 freeit:
205 m_put_rcvif_psref(ifp, &psref);
206 m_freem(m);
207 return;
208
209 bad:
210 ICMP6_STATINC(ICMP6_STAT_BADRS);
211 m_put_rcvif_psref(ifp, &psref);
212 m_freem(m);
213 }
214
215 /*
216 * Receive Router Advertisement Message.
217 *
218 * Based on RFC 2461
219 * TODO: on-link bit on prefix information
220 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
221 */
222 void
223 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
224 {
225 struct ifnet *ifp;
226 struct nd_ifinfo *ndi;
227 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
228 struct nd_router_advert *nd_ra;
229 struct in6_addr saddr6 = ip6->ip6_src;
230 #if 0
231 struct in6_addr daddr6 = ip6->ip6_dst;
232 int flags; /* = nd_ra->nd_ra_flags_reserved; */
233 int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
234 int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
235 #endif
236 int mcast = 0;
237 union nd_opts ndopts;
238 struct nd_defrouter *dr;
239 struct psref psref;
240 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
241
242 ifp = m_get_rcvif_psref(m, &psref);
243 if (ifp == NULL)
244 goto freeit;
245
246 ndi = ND_IFINFO(ifp);
247 /*
248 * We only accept RAs when
249 * the system-wide variable allows the acceptance, and the
250 * per-interface variable allows RAs on the receiving interface.
251 */
252 if (!nd6_accepts_rtadv(ndi))
253 goto freeit;
254
255 if (ip6->ip6_hlim != 255) {
256 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
257 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
258 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
259 goto bad;
260 }
261
262 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
263 nd6log(LOG_ERR, "src %s is not link-local\n",
264 IN6_PRINT(ip6buf, &saddr6));
265 goto bad;
266 }
267
268 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
269 if (nd_ra == NULL) {
270 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
271 m_put_rcvif_psref(ifp, &psref);
272 return;
273 }
274
275 icmp6len -= sizeof(*nd_ra);
276 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
277 if (nd6_options(&ndopts) < 0) {
278 nd6log(LOG_INFO, "invalid ND option, ignored\n");
279 /* nd6_options have incremented stats */
280 goto freeit;
281 }
282
283 {
284 struct nd_defrouter drtr;
285 u_int32_t advreachable = nd_ra->nd_ra_reachable;
286
287 /* remember if this is a multicasted advertisement */
288 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
289 mcast = 1;
290
291 memset(&drtr, 0, sizeof(drtr));
292 drtr.rtaddr = saddr6;
293 drtr.flags = nd_ra->nd_ra_flags_reserved;
294 drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
295 drtr.expire = time_uptime + drtr.rtlifetime;
296 drtr.ifp = ifp;
297 /* unspecified or not? (RFC 2461 6.3.4) */
298 if (advreachable) {
299 NTOHL(advreachable);
300 if (advreachable <= MAX_REACHABLE_TIME &&
301 ndi->basereachable != advreachable) {
302 ndi->basereachable = advreachable;
303 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
304 ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
305 }
306 }
307 if (nd_ra->nd_ra_retransmit)
308 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
309 if (nd_ra->nd_ra_curhoplimit) {
310 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
311 ndi->chlim = nd_ra->nd_ra_curhoplimit;
312 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
313 log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
314 "%s on %s (current=%d, received=%d), ignored\n",
315 IN6_PRINT(ip6buf, &ip6->ip6_src),
316 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
317 }
318 ND6_WLOCK();
319 dr = defrtrlist_update(&drtr);
320 }
321
322 /*
323 * prefix
324 */
325 if (ndopts.nd_opts_pi) {
326 struct nd_opt_hdr *pt;
327 struct nd_opt_prefix_info *pi = NULL;
328 struct nd_prefixctl prc;
329
330 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
331 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
332 pt = (struct nd_opt_hdr *)((char *)pt +
333 (pt->nd_opt_len << 3))) {
334 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
335 continue;
336 pi = (struct nd_opt_prefix_info *)pt;
337
338 if (pi->nd_opt_pi_len != 4) {
339 nd6log(LOG_INFO, "invalid option "
340 "len %d for prefix information option, "
341 "ignored\n", pi->nd_opt_pi_len);
342 continue;
343 }
344
345 if (128 < pi->nd_opt_pi_prefix_len) {
346 nd6log(LOG_INFO, "invalid prefix "
347 "len %d for prefix information option, "
348 "ignored\n", pi->nd_opt_pi_prefix_len);
349 continue;
350 }
351
352 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
353 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
354 nd6log(LOG_INFO,
355 "invalid prefix %s, ignored\n",
356 IN6_PRINT(ip6buf, &pi->nd_opt_pi_prefix));
357 continue;
358 }
359
360 memset(&prc, 0, sizeof(prc));
361 sockaddr_in6_init(&prc.ndprc_prefix,
362 &pi->nd_opt_pi_prefix, 0, 0, 0);
363 prc.ndprc_ifp = ifp;
364
365 prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
366 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
367 prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
368 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
369 prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
370 prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
371 prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
372
373 (void)prelist_update(&prc, dr, m, mcast);
374 }
375 }
376 ND6_UNLOCK();
377
378 /*
379 * MTU
380 */
381 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
382 u_long mtu;
383 u_long maxmtu;
384
385 mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
386
387 /* lower bound */
388 if (mtu < IPV6_MMTU) {
389 nd6log(LOG_INFO, "bogus mtu option "
390 "mtu=%lu sent from %s, ignoring\n",
391 mtu, IN6_PRINT(ip6buf, &ip6->ip6_src));
392 goto skip;
393 }
394
395 /* upper bound */
396 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
397 ? ndi->maxmtu : ifp->if_mtu;
398 if (mtu <= maxmtu) {
399 int change = (ndi->linkmtu != mtu);
400
401 ndi->linkmtu = mtu;
402 if (change) /* in6_maxmtu may change */
403 in6_setmaxmtu();
404 } else {
405 nd6log(LOG_INFO,
406 "bogus mtu mtu=%lu sent from %s; "
407 "exceeds maxmtu %lu, ignoring\n",
408 mtu, IN6_PRINT(ip6buf, &ip6->ip6_src), maxmtu);
409 }
410 }
411
412 skip:
413
414 /*
415 * Source link layer address
416 */
417 {
418 char *lladdr = NULL;
419 int lladdrlen = 0;
420
421 if (ndopts.nd_opts_src_lladdr) {
422 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
423 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
424 }
425
426 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
427 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
428 "(if %d, RA packet %d)\n", IN6_PRINT(ip6buf, &saddr6),
429 ifp->if_addrlen, lladdrlen - 2);
430 goto bad;
431 }
432
433 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
434
435 /*
436 * Installing a link-layer address might change the state of the
437 * router's neighbor cache, which might also affect our on-link
438 * detection of adveritsed prefixes.
439 */
440 ND6_WLOCK();
441 nd6_pfxlist_onlink_check();
442 ND6_UNLOCK();
443 }
444
445 freeit:
446 m_put_rcvif_psref(ifp, &psref);
447 m_freem(m);
448 return;
449
450 bad:
451 ICMP6_STATINC(ICMP6_STAT_BADRA);
452 m_put_rcvif_psref(ifp, &psref);
453 m_freem(m);
454 }
455
456 /*
457 * default router list processing sub routines
458 */
459 static void
460 defrouter_addreq(struct nd_defrouter *newdr)
461 {
462 union {
463 struct sockaddr_in6 sin6;
464 struct sockaddr sa;
465 } def, mask, gate;
466 int s;
467 int error;
468
469 memset(&def, 0, sizeof(def));
470 memset(&mask, 0, sizeof(mask));
471 memset(&gate, 0,sizeof(gate)); /* for safety */
472
473 def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
474 sizeof(struct sockaddr_in6);
475 def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
476 gate.sin6.sin6_addr = newdr->rtaddr;
477 #ifndef SCOPEDROUTING
478 gate.sin6.sin6_scope_id = 0; /* XXX */
479 #endif
480
481 s = splsoftnet();
482 error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
483 RTF_GATEWAY);
484 if (error == 0) {
485 nd6_numroutes++;
486 newdr->installed = 1;
487 }
488 splx(s);
489 return;
490 }
491
492 struct nd_defrouter *
493 nd6_defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
494 {
495 struct nd_defrouter *dr;
496
497 ND6_ASSERT_LOCK();
498
499 ND_DEFROUTER_LIST_FOREACH(dr) {
500 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
501 break;
502 }
503
504 return dr; /* search failed */
505 }
506
507 void
508 nd6_defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
509 {
510 struct nd_defrouter *deldr = NULL;
511 struct nd_prefix *pr;
512 struct nd_ifinfo *ndi;
513
514 ND6_ASSERT_WLOCK();
515
516 if (ext == NULL)
517 ext = dr->ifp->if_afdata[AF_INET6];
518
519 /* detach already in progress, can not do anything */
520 if (ext == NULL)
521 return;
522
523 ndi = ext->nd_ifinfo;
524
525 /*
526 * Flush all the routing table entries that use the router
527 * as a next hop.
528 */
529 /* XXX: better condition? */
530 if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
531 nd6_rt_flush(&dr->rtaddr, dr->ifp);
532
533 if (dr->installed) {
534 deldr = dr;
535 defrouter_delreq(dr);
536 }
537 ND_DEFROUTER_LIST_REMOVE(dr);
538
539 /*
540 * Also delete all the pointers to the router in each prefix lists.
541 */
542 ND_PREFIX_LIST_FOREACH(pr) {
543 struct nd_pfxrouter *pfxrtr;
544 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
545 pfxrtr_del(pfxrtr);
546 }
547 nd6_pfxlist_onlink_check();
548
549 /*
550 * If the router is the primary one, choose a new one.
551 * Note that nd6_defrouter_select() will remove the current gateway
552 * from the routing table.
553 */
554 if (deldr)
555 nd6_defrouter_select();
556
557 ext->ndefrouters--;
558 if (ext->ndefrouters < 0) {
559 log(LOG_WARNING, "nd6_defrtrlist_del: negative count on %s\n",
560 dr->ifp->if_xname);
561 }
562
563 free(dr, M_IP6NDP);
564 }
565
566 /*
567 * Remove the default route for a given router.
568 * This is just a subroutine function for nd6_defrouter_select(), and should
569 * not be called from anywhere else.
570 */
571 static void
572 defrouter_delreq(struct nd_defrouter *dr)
573 {
574 union {
575 struct sockaddr_in6 sin6;
576 struct sockaddr sa;
577 } def, mask, gw;
578 int error;
579
580 #ifdef DIAGNOSTIC
581 if (dr == NULL)
582 panic("dr == NULL in defrouter_delreq");
583 #endif
584
585 memset(&def, 0, sizeof(def));
586 memset(&mask, 0, sizeof(mask));
587 memset(&gw, 0, sizeof(gw)); /* for safety */
588
589 def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
590 sizeof(struct sockaddr_in6);
591 def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
592 gw.sin6.sin6_addr = dr->rtaddr;
593 #ifndef SCOPEDROUTING
594 gw.sin6.sin6_scope_id = 0; /* XXX */
595 #endif
596
597 error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
598 RTF_GATEWAY);
599 if (error == 0)
600 nd6_numroutes--;
601
602 dr->installed = 0;
603 }
604
605 /*
606 * remove all default routes from default router list
607 */
608 void
609 nd6_defrouter_reset(void)
610 {
611 struct nd_defrouter *dr;
612
613 ND6_ASSERT_WLOCK();
614
615 ND_DEFROUTER_LIST_FOREACH(dr)
616 defrouter_delreq(dr);
617
618 /*
619 * XXX should we also nuke any default routers in the kernel, by
620 * going through them by rtalloc1()?
621 */
622 }
623
624 /*
625 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
626 * draft-ietf-ipngwg-router-selection:
627 * 1) Routers that are reachable or probably reachable should be preferred.
628 * If we have more than one (probably) reachable router, prefer ones
629 * with the highest router preference.
630 * 2) When no routers on the list are known to be reachable or
631 * probably reachable, routers SHOULD be selected in a round-robin
632 * fashion, regardless of router preference values.
633 * 3) If the Default Router List is empty, assume that all
634 * destinations are on-link.
635 *
636 * We assume nd_defrouter is sorted by router preference value.
637 * Since the code below covers both with and without router preference cases,
638 * we do not need to classify the cases by ifdef.
639 *
640 * At this moment, we do not try to install more than one default router,
641 * even when the multipath routing is available, because we're not sure about
642 * the benefits for stub hosts comparing to the risk of making the code
643 * complicated and the possibility of introducing bugs.
644 */
645 void
646 nd6_defrouter_select(void)
647 {
648 struct nd_ifinfo *ndi;
649 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
650
651 ND6_ASSERT_WLOCK();
652
653 /*
654 * This function should be called only when acting as an autoconfigured
655 * host. Although the remaining part of this function is not effective
656 * if the node is not an autoconfigured host, we explicitly exclude
657 * such cases here for safety.
658 */
659 if (ip6_forwarding) {
660 nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, "
661 "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv);
662 return;
663 }
664
665 /*
666 * Let's handle easy case (3) first:
667 * If default router list is empty, there's nothing to be done.
668 */
669 if (ND_DEFROUTER_LIST_EMPTY())
670 return;
671
672 /*
673 * Search for a (probably) reachable router from the list.
674 * We just pick up the first reachable one (if any), assuming that
675 * the ordering rule of the list described in defrtrlist_update().
676 */
677 ND_DEFROUTER_LIST_FOREACH(dr) {
678 ndi = ND_IFINFO(dr->ifp);
679 if (nd6_accepts_rtadv(ndi))
680 continue;
681
682 if (selected_dr == NULL &&
683 nd6_is_llinfo_probreach(dr))
684 selected_dr = dr;
685
686 if (dr->installed && !installed_dr)
687 installed_dr = dr;
688 else if (dr->installed && installed_dr) {
689 /* this should not happen. warn for diagnosis. */
690 log(LOG_ERR, "nd6_defrouter_select: more than one router"
691 " is installed\n");
692 }
693 }
694 /*
695 * If none of the default routers was found to be reachable,
696 * round-robin the list regardless of preference.
697 * Otherwise, if we have an installed router, check if the selected
698 * (reachable) router should really be preferred to the installed one.
699 * We only prefer the new router when the old one is not reachable
700 * or when the new one has a really higher preference value.
701 */
702 if (selected_dr == NULL) {
703 if (installed_dr == NULL ||
704 ND_DEFROUTER_LIST_NEXT(installed_dr) == NULL)
705 selected_dr = ND_DEFROUTER_LIST_FIRST();
706 else
707 selected_dr = ND_DEFROUTER_LIST_NEXT(installed_dr);
708 } else if (installed_dr &&
709 nd6_is_llinfo_probreach(installed_dr) &&
710 rtpref(selected_dr) <= rtpref(installed_dr)) {
711 selected_dr = installed_dr;
712 }
713
714 /*
715 * If the selected router is different than the installed one,
716 * remove the installed router and install the selected one.
717 * Note that the selected router is never NULL here.
718 */
719 if (installed_dr != selected_dr) {
720 if (installed_dr)
721 defrouter_delreq(installed_dr);
722 defrouter_addreq(selected_dr);
723 }
724
725 return;
726 }
727
728 /*
729 * for default router selection
730 * regards router-preference field as a 2-bit signed integer
731 */
732 static int
733 rtpref(struct nd_defrouter *dr)
734 {
735 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
736 case ND_RA_FLAG_RTPREF_HIGH:
737 return (RTPREF_HIGH);
738 case ND_RA_FLAG_RTPREF_MEDIUM:
739 case ND_RA_FLAG_RTPREF_RSV:
740 return (RTPREF_MEDIUM);
741 case ND_RA_FLAG_RTPREF_LOW:
742 return (RTPREF_LOW);
743 default:
744 /*
745 * This case should never happen. If it did, it would mean a
746 * serious bug of kernel internal. We thus always bark here.
747 * Or, can we even panic?
748 */
749 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
750 return (RTPREF_INVALID);
751 }
752 /* NOTREACHED */
753 }
754
755 static struct nd_defrouter *
756 defrtrlist_update(struct nd_defrouter *newdr)
757 {
758 struct nd_defrouter *dr, *n, *ret = NULL;
759 struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
760
761 ND6_ASSERT_WLOCK();
762
763 if ((dr = nd6_defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
764 /* entry exists */
765 if (newdr->rtlifetime == 0) {
766 nd6_defrtrlist_del(dr, ext);
767 dr = NULL;
768 } else {
769 int oldpref = rtpref(dr);
770
771 /* override */
772 dr->flags = newdr->flags; /* xxx flag check */
773 dr->rtlifetime = newdr->rtlifetime;
774 dr->expire = newdr->expire;
775
776 /*
777 * If the preference does not change, there's no need
778 * to sort the entries.
779 */
780 if (rtpref(newdr) == oldpref) {
781 ret = dr;
782 goto out;
783 }
784
785 /*
786 * preferred router may be changed, so relocate
787 * this router.
788 * XXX: calling TAILQ_REMOVE directly is a bad manner.
789 * However, since nd6_defrtrlist_del() has many side
790 * effects, we intentionally do so here.
791 * nd6_defrouter_select() below will handle routing
792 * changes later.
793 */
794 ND_DEFROUTER_LIST_REMOVE(dr);
795 n = dr;
796 goto insert;
797 }
798 ret = dr;
799 goto out;
800 }
801
802 if (ip6_maxifdefrouters >= 0 && ext->ndefrouters >= ip6_maxifdefrouters)
803 goto out;
804
805 /* entry does not exist */
806 if (newdr->rtlifetime == 0)
807 goto out;
808
809 if (ip6_rtadv_maxroutes <= nd6_numroutes) {
810 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
811 goto out;
812 }
813
814 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
815 if (n == NULL)
816 goto out;
817 memset(n, 0, sizeof(*n));
818 *n = *newdr;
819
820 insert:
821 /*
822 * Insert the new router in the Default Router List;
823 * The Default Router List should be in the descending order
824 * of router-preferece. Routers with the same preference are
825 * sorted in the arriving time order.
826 */
827
828 /* insert at the end of the group */
829 ND_DEFROUTER_LIST_FOREACH(dr) {
830 if (rtpref(n) > rtpref(dr))
831 break;
832 }
833 if (dr)
834 ND_DEFROUTER_LIST_INSERT_BEFORE(dr, n);
835 else
836 ND_DEFROUTER_LIST_INSERT_TAIL(n);
837
838 nd6_defrouter_select();
839
840 ext->ndefrouters++;
841
842 ret = n;
843 out:
844 return ret;
845 }
846
847 static struct nd_pfxrouter *
848 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
849 {
850 struct nd_pfxrouter *search;
851
852 ND6_ASSERT_LOCK();
853
854 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
855 if (search->router == dr)
856 break;
857 }
858
859 return (search);
860 }
861
862 static void
863 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
864 {
865 struct nd_pfxrouter *newpfr;
866
867 ND6_ASSERT_WLOCK();
868
869 newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
870 if (newpfr == NULL)
871 return;
872 newpfr->router = dr;
873
874 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
875
876 nd6_pfxlist_onlink_check();
877 }
878
879 static void
880 pfxrtr_del(struct nd_pfxrouter *pfr)
881 {
882 LIST_REMOVE(pfr, pfr_entry);
883 free(pfr, M_IP6NDP);
884 }
885
886 static struct nd_prefix *
887 nd6_prefix_lookup(struct nd_prefixctl *key)
888 {
889 struct nd_prefix *search;
890
891 ND_PREFIX_LIST_FOREACH(search) {
892 if (key->ndprc_ifp == search->ndpr_ifp &&
893 key->ndprc_plen == search->ndpr_plen &&
894 in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
895 &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
896 break;
897 }
898 }
899
900 return (search);
901 }
902
903 static void
904 purge_detached(struct ifnet *ifp)
905 {
906 struct nd_prefix *pr, *pr_next;
907 struct in6_ifaddr *ia;
908 struct ifaddr *ifa, *ifa_next;
909
910 restart:
911 ND6_ASSERT_WLOCK();
912
913 ND_PREFIX_LIST_FOREACH_SAFE(pr, pr_next) {
914 int s;
915
916 /*
917 * This function is called when we need to make more room for
918 * new prefixes rather than keeping old, possibly stale ones.
919 * Detached prefixes would be a good candidate; if all routers
920 * that advertised the prefix expired, the prefix is also
921 * probably stale.
922 */
923 if (pr->ndpr_ifp != ifp ||
924 IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
925 ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
926 !LIST_EMPTY(&pr->ndpr_advrtrs)))
927 continue;
928
929 s = pserialize_read_enter();
930 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) {
931 ifa_next = IFADDR_READER_NEXT(ifa);
932 if (ifa->ifa_addr->sa_family != AF_INET6)
933 continue;
934 ia = (struct in6_ifaddr *)ifa;
935 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
936 IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
937 pserialize_read_exit(s);
938 ND6_UNLOCK();
939
940 /* in6_purgeaddr may destroy pr. */
941 in6_purgeaddr(ifa);
942
943 ND6_WLOCK();
944 goto restart;
945 }
946 }
947 pserialize_read_exit(s);
948
949 KASSERT(pr->ndpr_refcnt == 0);
950 nd6_prelist_remove(pr);
951 }
952 }
953
954 static int
955 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
956 struct nd_prefix **newp)
957 {
958 struct nd_prefix *newpr = NULL;
959 int i, s;
960 int error;
961 struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
962
963 ND6_ASSERT_WLOCK();
964
965 if (ip6_maxifprefixes >= 0) {
966 if (ext->nprefixes >= ip6_maxifprefixes / 2)
967 purge_detached(prc->ndprc_ifp);
968 if (ext->nprefixes >= ip6_maxifprefixes)
969 return ENOMEM;
970 }
971
972 error = 0;
973 newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
974 if (newpr == NULL)
975 return ENOMEM;
976 newpr->ndpr_ifp = prc->ndprc_ifp;
977 newpr->ndpr_prefix = prc->ndprc_prefix;
978 newpr->ndpr_plen = prc->ndprc_plen;
979 newpr->ndpr_vltime = prc->ndprc_vltime;
980 newpr->ndpr_pltime = prc->ndprc_pltime;
981 newpr->ndpr_flags = prc->ndprc_flags;
982 if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
983 free(newpr, M_IP6NDP);
984 return(error);
985 }
986 newpr->ndpr_lastupdate = time_uptime;
987 if (newp != NULL)
988 *newp = newpr;
989
990 /* initialization */
991 LIST_INIT(&newpr->ndpr_advrtrs);
992 in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
993 /* make prefix in the canonical form */
994 for (i = 0; i < 4; i++) {
995 newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
996 newpr->ndpr_mask.s6_addr32[i];
997 }
998
999 s = splsoftnet();
1000 /* link ndpr_entry to nd_prefix list */
1001 ND_PREFIX_LIST_INSERT_HEAD(newpr);
1002 splx(s);
1003
1004 /* ND_OPT_PI_FLAG_ONLINK processing */
1005 if (newpr->ndpr_raf_onlink) {
1006 int e;
1007
1008 if ((e = nd6_prefix_onlink(newpr)) != 0) {
1009 char ip6buf[INET6_ADDRSTRLEN];
1010 nd6log(LOG_ERR, "failed to make "
1011 "the prefix %s/%d on-link on %s (errno=%d)\n",
1012 IN6_PRINT(ip6buf, &prc->ndprc_prefix.sin6_addr),
1013 prc->ndprc_plen, if_name(prc->ndprc_ifp), e);
1014 /* proceed anyway. XXX: is it correct? */
1015 }
1016 }
1017
1018 if (dr)
1019 pfxrtr_add(newpr, dr);
1020
1021 ext->nprefixes++;
1022
1023 return 0;
1024 }
1025
1026 void
1027 nd6_prefix_unref(struct nd_prefix *pr)
1028 {
1029
1030 ND6_WLOCK();
1031 pr->ndpr_refcnt--;
1032 if (pr->ndpr_refcnt == 0)
1033 nd6_prelist_remove(pr);
1034 ND6_UNLOCK();
1035 }
1036
1037 void
1038 nd6_invalidate_prefix(struct nd_prefix *pr)
1039 {
1040 int e;
1041
1042 ND6_ASSERT_WLOCK();
1043
1044 /* make sure to invalidate the prefix until it is really freed. */
1045 pr->ndpr_vltime = 0;
1046 pr->ndpr_pltime = 0;
1047 #if 0
1048 /*
1049 * Though these flags are now meaningless, we'd rather keep the value
1050 * not to confuse users when executing "ndp -p".
1051 */
1052 pr->ndpr_raf_onlink = 0;
1053 pr->ndpr_raf_auto = 0;
1054 #endif
1055 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
1056 (e = nd6_prefix_offlink(pr)) != 0) {
1057 char ip6buf[INET6_ADDRSTRLEN];
1058 nd6log(LOG_ERR,
1059 "failed to make %s/%d offlink on %s, errno=%d\n",
1060 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr),
1061 pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1062 /* what should we do? */
1063 }
1064 }
1065
1066 void
1067 nd6_prelist_remove(struct nd_prefix *pr)
1068 {
1069 struct nd_pfxrouter *pfr, *next;
1070 int s;
1071 struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
1072
1073 ND6_ASSERT_WLOCK();
1074 KASSERT(pr->ndpr_refcnt == 0);
1075
1076 nd6_invalidate_prefix(pr);
1077
1078 s = splsoftnet();
1079 /* unlink ndpr_entry from nd_prefix list */
1080 ND_PREFIX_LIST_REMOVE(pr);
1081
1082 /* free list of routers that adversed the prefix */
1083 for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
1084 next = LIST_NEXT(pfr, pfr_entry);
1085
1086 free(pfr, M_IP6NDP);
1087 }
1088
1089 if (ext) {
1090 ext->nprefixes--;
1091 if (ext->nprefixes < 0) {
1092 log(LOG_WARNING, "nd6_prelist_remove: negative count on "
1093 "%s\n", pr->ndpr_ifp->if_xname);
1094 }
1095 }
1096 splx(s);
1097
1098 free(pr, M_IP6NDP);
1099
1100 nd6_pfxlist_onlink_check();
1101 }
1102
1103 static int
1104 prelist_update(struct nd_prefixctl *newprc,
1105 struct nd_defrouter *dr, /* may be NULL */
1106 struct mbuf *m,
1107 int mcast)
1108 {
1109 struct in6_ifaddr *ia6_match = NULL;
1110 struct ifaddr *ifa;
1111 struct ifnet *ifp = newprc->ndprc_ifp;
1112 struct nd_prefix *pr;
1113 int error = 0;
1114 int auth;
1115 struct in6_addrlifetime lt6_tmp;
1116 int ss;
1117 char ip6buf[INET6_ADDRSTRLEN];
1118
1119 ND6_ASSERT_WLOCK();
1120
1121 auth = 0;
1122 if (m) {
1123 /*
1124 * Authenticity for NA consists authentication for
1125 * both IP header and IP datagrams, doesn't it ?
1126 */
1127 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1128 auth = (m->m_flags & M_AUTHIPHDR
1129 && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1130 #endif
1131 }
1132
1133 if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
1134 /*
1135 * nd6_prefix_lookup() ensures that pr and newprc have the same
1136 * prefix on a same interface.
1137 */
1138
1139 /*
1140 * Update prefix information. Note that the on-link (L) bit
1141 * and the autonomous (A) bit should NOT be changed from 1
1142 * to 0.
1143 */
1144 if (newprc->ndprc_raf_onlink == 1)
1145 pr->ndpr_raf_onlink = 1;
1146 if (newprc->ndprc_raf_auto == 1)
1147 pr->ndpr_raf_auto = 1;
1148 if (newprc->ndprc_raf_onlink) {
1149 pr->ndpr_vltime = newprc->ndprc_vltime;
1150 pr->ndpr_pltime = newprc->ndprc_pltime;
1151 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1152 pr->ndpr_lastupdate = time_uptime;
1153 }
1154
1155 if (newprc->ndprc_raf_onlink &&
1156 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1157 int e;
1158
1159 if ((e = nd6_prefix_onlink(pr)) != 0) {
1160 nd6log(LOG_ERR,
1161 "failed to make "
1162 "the prefix %s/%d on-link on %s "
1163 "(errno=%d)\n",
1164 IN6_PRINT(ip6buf,
1165 &pr->ndpr_prefix.sin6_addr),
1166 pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1167 /* proceed anyway. XXX: is it correct? */
1168 }
1169 }
1170
1171 if (dr && pfxrtr_lookup(pr, dr) == NULL)
1172 pfxrtr_add(pr, dr);
1173 } else {
1174 struct nd_prefix *newpr = NULL;
1175
1176 if (newprc->ndprc_vltime == 0)
1177 goto end;
1178 if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
1179 goto end;
1180
1181 if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1182 ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1183 goto end;
1184 }
1185
1186 error = nd6_prelist_add(newprc, dr, &newpr);
1187 if (error != 0 || newpr == NULL) {
1188 nd6log(LOG_NOTICE,
1189 "nd6_prelist_add failed for %s/%d on %s "
1190 "errno=%d, returnpr=%p\n",
1191 IN6_PRINT(ip6buf, &newprc->ndprc_prefix.sin6_addr),
1192 newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
1193 error, newpr);
1194 goto end; /* we should just give up in this case. */
1195 }
1196
1197 /*
1198 * XXX: from the ND point of view, we can ignore a prefix
1199 * with the on-link bit being zero. However, we need a
1200 * prefix structure for references from autoconfigured
1201 * addresses. Thus, we explicitly make sure that the prefix
1202 * itself expires now.
1203 */
1204 if (newpr->ndpr_raf_onlink == 0) {
1205 newpr->ndpr_vltime = 0;
1206 newpr->ndpr_pltime = 0;
1207 in6_init_prefix_ltimes(newpr);
1208 }
1209
1210 pr = newpr;
1211 }
1212
1213 /*
1214 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1215 * Note that pr must be non NULL at this point.
1216 */
1217
1218 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1219 if (!newprc->ndprc_raf_auto)
1220 goto end;
1221
1222 /*
1223 * 5.5.3 (b). the link-local prefix should have been ignored in
1224 * nd6_ra_input.
1225 */
1226
1227 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1228 if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
1229 error = EINVAL; /* XXX: won't be used */
1230 goto end;
1231 }
1232
1233 /*
1234 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1235 * an address configured by stateless autoconfiguration already in the
1236 * list of addresses associated with the interface, and the Valid
1237 * Lifetime is not 0, form an address. We first check if we have
1238 * a matching prefix.
1239 * Note: we apply a clarification in rfc2462bis-02 here. We only
1240 * consider autoconfigured addresses while RFC2462 simply said
1241 * "address".
1242 */
1243 ss = pserialize_read_enter();
1244 IFADDR_READER_FOREACH(ifa, ifp) {
1245 struct in6_ifaddr *ia6;
1246 u_int32_t remaininglifetime;
1247
1248 if (ifa->ifa_addr->sa_family != AF_INET6)
1249 continue;
1250
1251 ia6 = (struct in6_ifaddr *)ifa;
1252
1253 /*
1254 * We only consider autoconfigured addresses as per rfc2462bis.
1255 */
1256 if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF))
1257 continue;
1258
1259 /*
1260 * Spec is not clear here, but I believe we should concentrate
1261 * on unicast (i.e. not anycast) addresses.
1262 * XXX: other ia6_flags? detached or duplicated?
1263 */
1264 if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1265 continue;
1266
1267 /*
1268 * Ignore the address if it is not associated with a prefix
1269 * or is associated with a prefix that is different from this
1270 * one. (pr is never NULL here)
1271 */
1272 if (ia6->ia6_ndpr != pr)
1273 continue;
1274
1275 if (ia6_match == NULL) /* remember the first one */
1276 ia6_match = ia6;
1277
1278 /*
1279 * An already autoconfigured address matched. Now that we
1280 * are sure there is at least one matched address, we can
1281 * proceed to 5.5.3. (e): update the lifetimes according to the
1282 * "two hours" rule and the privacy extension.
1283 * We apply some clarifications in rfc2462bis:
1284 * - use remaininglifetime instead of storedlifetime as a
1285 * variable name
1286 * - remove the dead code in the "two-hour" rule
1287 */
1288 #define TWOHOUR (120*60)
1289 lt6_tmp = ia6->ia6_lifetime;
1290 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1291 remaininglifetime = ND6_INFINITE_LIFETIME;
1292 else if (time_uptime - ia6->ia6_updatetime >
1293 lt6_tmp.ia6t_vltime) {
1294 /*
1295 * The case of "invalid" address. We should usually
1296 * not see this case.
1297 */
1298 remaininglifetime = 0;
1299 } else
1300 remaininglifetime = lt6_tmp.ia6t_vltime -
1301 (time_uptime - ia6->ia6_updatetime);
1302
1303 /* when not updating, keep the current stored lifetime. */
1304 lt6_tmp.ia6t_vltime = remaininglifetime;
1305
1306 if (TWOHOUR < newprc->ndprc_vltime ||
1307 remaininglifetime < newprc->ndprc_vltime) {
1308 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1309 } else if (remaininglifetime <= TWOHOUR) {
1310 if (auth)
1311 lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1312 } else {
1313 /*
1314 * newprc->ndprc_vltime <= TWOHOUR &&
1315 * TWOHOUR < remaininglifetime
1316 */
1317 lt6_tmp.ia6t_vltime = TWOHOUR;
1318 }
1319
1320 /* The 2 hour rule is not imposed for preferred lifetime. */
1321 lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
1322
1323 in6_init_address_ltimes(pr, <6_tmp);
1324
1325 /*
1326 * We need to treat lifetimes for temporary addresses
1327 * differently, according to
1328 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1329 * we only update the lifetimes when they are in the maximum
1330 * intervals.
1331 */
1332 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1333 u_int32_t maxvltime, maxpltime;
1334
1335 if (ip6_temp_valid_lifetime >
1336 (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1337 ip6_desync_factor)) {
1338 maxvltime = ip6_temp_valid_lifetime -
1339 (time_uptime - ia6->ia6_createtime) -
1340 ip6_desync_factor;
1341 } else
1342 maxvltime = 0;
1343 if (ip6_temp_preferred_lifetime >
1344 (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1345 ip6_desync_factor)) {
1346 maxpltime = ip6_temp_preferred_lifetime -
1347 (time_uptime - ia6->ia6_createtime) -
1348 ip6_desync_factor;
1349 } else
1350 maxpltime = 0;
1351
1352 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1353 lt6_tmp.ia6t_vltime > maxvltime) {
1354 lt6_tmp.ia6t_vltime = maxvltime;
1355 }
1356 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1357 lt6_tmp.ia6t_pltime > maxpltime) {
1358 lt6_tmp.ia6t_pltime = maxpltime;
1359 }
1360 }
1361
1362 ia6->ia6_lifetime = lt6_tmp;
1363 ia6->ia6_updatetime = time_uptime;
1364 }
1365 pserialize_read_exit(ss);
1366
1367 if (ia6_match == NULL && newprc->ndprc_vltime) {
1368 int ifidlen;
1369 struct in6_ifaddr *ia6;
1370 struct psref psref;
1371
1372 /*
1373 * 5.5.3 (d) (continued)
1374 * No address matched and the valid lifetime is non-zero.
1375 * Create a new address.
1376 */
1377
1378 /*
1379 * Prefix Length check:
1380 * If the sum of the prefix length and interface identifier
1381 * length does not equal 128 bits, the Prefix Information
1382 * option MUST be ignored. The length of the interface
1383 * identifier is defined in a separate link-type specific
1384 * document.
1385 */
1386 ifidlen = in6_if2idlen(ifp);
1387 if (ifidlen < 0) {
1388 /* this should not happen, so we always log it. */
1389 log(LOG_ERR, "%s: IFID undefined (%s)\n",
1390 __func__, if_name(ifp));
1391 goto end;
1392 }
1393 if (ifidlen + pr->ndpr_plen != 128) {
1394 nd6log(LOG_INFO,
1395 "invalid prefixlen %d for %s, ignored\n",
1396 pr->ndpr_plen, if_name(ifp));
1397 goto end;
1398 }
1399
1400 if ((ia6 = in6_ifadd(newprc, mcast, &psref)) != NULL) {
1401 /*
1402 * note that we should use pr (not newprc) for reference.
1403 */
1404 pr->ndpr_refcnt++;
1405 ia6->ia6_ndpr = pr;
1406
1407 /* toggle onlink state if the address was assigned
1408 * a prefix route. */
1409 if (ia6->ia_flags & IFA_ROUTE)
1410 pr->ndpr_stateflags |= NDPRF_ONLINK;
1411
1412 /*
1413 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1414 * When a new public address is created as described
1415 * in RFC2462, also create a new temporary address.
1416 *
1417 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1418 * When an interface connects to a new link, a new
1419 * randomized interface identifier should be generated
1420 * immediately together with a new set of temporary
1421 * addresses. Thus, we specifiy 1 as the 2nd arg of
1422 * in6_tmpifadd().
1423 */
1424 if (ip6_use_tempaddr) {
1425 int e;
1426 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1427 nd6log(LOG_NOTICE,
1428 "failed to create a temporary "
1429 "address, errno=%d\n", e);
1430 }
1431 }
1432 ia6_release(ia6, &psref);
1433
1434 /*
1435 * A newly added address might affect the status
1436 * of other addresses, so we check and update it.
1437 * XXX: what if address duplication happens?
1438 */
1439 nd6_pfxlist_onlink_check();
1440 } else {
1441 /* just set an error. do not bark here. */
1442 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1443 }
1444 }
1445
1446 end:
1447 return error;
1448 }
1449
1450 /*
1451 * A supplement function used in the on-link detection below;
1452 * detect if a given prefix has a (probably) reachable advertising router.
1453 * XXX: lengthy function name...
1454 */
1455 static struct nd_pfxrouter *
1456 find_pfxlist_reachable_router(struct nd_prefix *pr)
1457 {
1458 struct nd_pfxrouter *pfxrtr;
1459
1460 for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1461 pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1462 if (pfxrtr->router->ifp->if_flags & IFF_UP &&
1463 pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
1464 nd6_is_llinfo_probreach(pfxrtr->router))
1465 break; /* found */
1466 }
1467
1468 return (pfxrtr);
1469 }
1470
1471 /*
1472 * Check if each prefix in the prefix list has at least one available router
1473 * that advertised the prefix (a router is "available" if its neighbor cache
1474 * entry is reachable or probably reachable).
1475 * If the check fails, the prefix may be off-link, because, for example,
1476 * we have moved from the network but the lifetime of the prefix has not
1477 * expired yet. So we should not use the prefix if there is another prefix
1478 * that has an available router.
1479 * But, if there is no prefix that has an available router, we still regards
1480 * all the prefixes as on-link. This is because we can't tell if all the
1481 * routers are simply dead or if we really moved from the network and there
1482 * is no router around us.
1483 */
1484 void
1485 nd6_pfxlist_onlink_check(void)
1486 {
1487 struct nd_prefix *pr;
1488 struct in6_ifaddr *ia;
1489 struct nd_defrouter *dr;
1490 struct nd_pfxrouter *pfxrtr = NULL;
1491 int s;
1492 char ip6buf[INET6_ADDRSTRLEN];
1493
1494 ND6_ASSERT_WLOCK();
1495
1496 /*
1497 * Check if there is a prefix that has a reachable advertising
1498 * router.
1499 */
1500 ND_PREFIX_LIST_FOREACH(pr) {
1501 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1502 break;
1503 }
1504 /*
1505 * If we have no such prefix, check whether we still have a router
1506 * that does not advertise any prefixes.
1507 */
1508 if (pr == NULL) {
1509 ND_DEFROUTER_LIST_FOREACH(dr) {
1510 struct nd_prefix *pr0;
1511
1512 ND_PREFIX_LIST_FOREACH(pr0) {
1513 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1514 break;
1515 }
1516 if (pfxrtr)
1517 break;
1518 }
1519 }
1520 if (pr != NULL || (!ND_DEFROUTER_LIST_EMPTY() && !pfxrtr)) {
1521 /*
1522 * There is at least one prefix that has a reachable router,
1523 * or at least a router which probably does not advertise
1524 * any prefixes. The latter would be the case when we move
1525 * to a new link where we have a router that does not provide
1526 * prefixes and we configure an address by hand.
1527 * Detach prefixes which have no reachable advertising
1528 * router, and attach other prefixes.
1529 */
1530 ND_PREFIX_LIST_FOREACH(pr) {
1531 /* XXX: a link-local prefix should never be detached */
1532 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1533 continue;
1534
1535 /*
1536 * we aren't interested in prefixes without the L bit
1537 * set.
1538 */
1539 if (pr->ndpr_raf_onlink == 0)
1540 continue;
1541
1542 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1543 find_pfxlist_reachable_router(pr) == NULL)
1544 pr->ndpr_stateflags |= NDPRF_DETACHED;
1545 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1546 find_pfxlist_reachable_router(pr) != 0)
1547 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1548 }
1549 } else {
1550 /* there is no prefix that has a reachable router */
1551 ND_PREFIX_LIST_FOREACH(pr) {
1552 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1553 continue;
1554
1555 if (pr->ndpr_raf_onlink == 0)
1556 continue;
1557
1558 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1559 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1560 }
1561 }
1562
1563 /*
1564 * Remove each interface route associated with a (just) detached
1565 * prefix, and reinstall the interface route for a (just) attached
1566 * prefix. Note that all attempt of reinstallation does not
1567 * necessarily success, when a same prefix is shared among multiple
1568 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1569 * so we don't have to care about them.
1570 */
1571 ND_PREFIX_LIST_FOREACH(pr) {
1572 int e;
1573
1574 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1575 continue;
1576
1577 if (pr->ndpr_raf_onlink == 0)
1578 continue;
1579
1580 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1581 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1582 if ((e = nd6_prefix_offlink(pr)) != 0) {
1583 nd6log(LOG_ERR,
1584 "failed to make %s/%d offlink, errno=%d\n",
1585 IN6_PRINT(ip6buf,
1586 &pr->ndpr_prefix.sin6_addr),
1587 pr->ndpr_plen, e);
1588 }
1589 }
1590 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1591 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1592 pr->ndpr_raf_onlink) {
1593 if ((e = nd6_prefix_onlink(pr)) != 0) {
1594 nd6log(LOG_ERR,
1595 "failed to make %s/%d onlink, errno=%d\n",
1596 IN6_PRINT(ip6buf,
1597 &pr->ndpr_prefix.sin6_addr),
1598 pr->ndpr_plen, e);
1599 }
1600 }
1601 }
1602
1603 /*
1604 * Changes on the prefix status might affect address status as well.
1605 * Make sure that all addresses derived from an attached prefix are
1606 * attached, and that all addresses derived from a detached prefix are
1607 * detached. Note, however, that a manually configured address should
1608 * always be attached.
1609 * The precise detection logic is same as the one for prefixes.
1610 */
1611 s = pserialize_read_enter();
1612 IN6_ADDRLIST_READER_FOREACH(ia) {
1613 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
1614 continue;
1615
1616 if (ia->ia6_ndpr == NULL) {
1617 /*
1618 * This can happen when we first configure the address
1619 * (i.e. the address exists, but the prefix does not).
1620 * XXX: complicated relationships...
1621 */
1622 continue;
1623 }
1624
1625 if (find_pfxlist_reachable_router(ia->ia6_ndpr))
1626 break;
1627 }
1628 pserialize_read_exit(s);
1629
1630 if (ia) {
1631 int bound = curlwp_bind();
1632
1633 s = pserialize_read_enter();
1634 IN6_ADDRLIST_READER_FOREACH(ia) {
1635 struct ifaddr *ifa = (struct ifaddr *)ia;
1636 struct psref psref;
1637
1638 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1639 continue;
1640
1641 if (ia->ia6_ndpr == NULL) /* XXX: see above. */
1642 continue;
1643
1644 ia6_acquire(ia, &psref);
1645 pserialize_read_exit(s);
1646
1647 if (find_pfxlist_reachable_router(ia->ia6_ndpr)) {
1648 if (ia->ia6_flags & IN6_IFF_DETACHED) {
1649 ia->ia6_flags &= ~IN6_IFF_DETACHED;
1650 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1651 nd6_dad_start(ifa,
1652 0);
1653 /* We will notify the routing socket
1654 * of the DAD result, so no need to
1655 * here */
1656 }
1657 } else {
1658 if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) {
1659 ia->ia6_flags |= IN6_IFF_DETACHED;
1660 rt_newaddrmsg(RTM_NEWADDR,
1661 ifa, 0, NULL);
1662 }
1663 }
1664
1665 s = pserialize_read_enter();
1666 ia6_release(ia, &psref);
1667 }
1668 pserialize_read_exit(s);
1669 curlwp_bindx(bound);
1670 }
1671 else {
1672 int bound = curlwp_bind();
1673
1674 s = pserialize_read_enter();
1675 IN6_ADDRLIST_READER_FOREACH(ia) {
1676 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1677 continue;
1678
1679 if (ia->ia6_flags & IN6_IFF_DETACHED) {
1680 struct ifaddr *ifa = (struct ifaddr *)ia;
1681 struct psref psref;
1682
1683 ia->ia6_flags &= ~IN6_IFF_DETACHED;
1684 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1685
1686 ia6_acquire(ia, &psref);
1687 pserialize_read_exit(s);
1688
1689 /* Do we need a delay in this case? */
1690 nd6_dad_start(ifa, 0);
1691
1692 s = pserialize_read_enter();
1693 ia6_release(ia, &psref);
1694 }
1695 }
1696 pserialize_read_exit(s);
1697 curlwp_bindx(bound);
1698 }
1699 }
1700
1701 static int
1702 nd6_prefix_onlink(struct nd_prefix *pr)
1703 {
1704 struct ifaddr *ifa;
1705 struct ifnet *ifp = pr->ndpr_ifp;
1706 struct sockaddr_in6 mask6;
1707 struct nd_prefix *opr;
1708 u_long rtflags;
1709 int error = 0;
1710 struct psref psref;
1711 int bound;
1712 char ip6buf[INET6_ADDRSTRLEN];
1713 char ip6bufp[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1714
1715 ND6_ASSERT_WLOCK();
1716
1717 /* sanity check */
1718 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1719 nd6log(LOG_ERR, "%s/%d is already on-link\n",
1720 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr),
1721 pr->ndpr_plen);
1722 return (EEXIST);
1723 }
1724
1725 /*
1726 * Add the interface route associated with the prefix. Before
1727 * installing the route, check if there's the same prefix on another
1728 * interface, and the prefix has already installed the interface route.
1729 * Although such a configuration is expected to be rare, we explicitly
1730 * allow it.
1731 */
1732 ND_PREFIX_LIST_FOREACH(opr) {
1733 if (opr == pr)
1734 continue;
1735
1736 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1737 continue;
1738
1739 if (opr->ndpr_plen == pr->ndpr_plen &&
1740 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1741 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1742 return (0);
1743 }
1744
1745 /*
1746 * We prefer link-local addresses as the associated interface address.
1747 */
1748 /* search for a link-local addr */
1749 bound = curlwp_bind();
1750 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
1751 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST, &psref);
1752 if (ifa == NULL) {
1753 int s = pserialize_read_enter();
1754 IFADDR_READER_FOREACH(ifa, ifp) {
1755 if (ifa->ifa_addr->sa_family == AF_INET6)
1756 break;
1757 }
1758 if (ifa != NULL)
1759 ifa_acquire(ifa, &psref);
1760 pserialize_read_exit(s);
1761 /* should we care about ia6_flags? */
1762 }
1763 if (ifa == NULL) {
1764 /*
1765 * This can still happen, when, for example, we receive an RA
1766 * containing a prefix with the L bit set and the A bit clear,
1767 * after removing all IPv6 addresses on the receiving
1768 * interface. This should, of course, be rare though.
1769 */
1770 nd6log(LOG_NOTICE, "failed to find any ifaddr"
1771 " to add route for a prefix(%s/%d) on %s\n",
1772 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr),
1773 pr->ndpr_plen, if_name(ifp));
1774 curlwp_bindx(bound);
1775 return (0);
1776 }
1777
1778 /*
1779 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1780 * ifa->ifa_rtrequest = nd6_rtrequest;
1781 */
1782 memset(&mask6, 0, sizeof(mask6));
1783 mask6.sin6_family = AF_INET6;
1784 mask6.sin6_len = sizeof(mask6);
1785 mask6.sin6_addr = pr->ndpr_mask;
1786 /* rtrequest() will probably set RTF_UP, but we're not sure. */
1787 rtflags = ifa->ifa_flags | RTF_UP;
1788 if (nd6_need_cache(ifp)) {
1789 /* explicitly set in case ifa_flags does not set the flag. */
1790 rtflags |= RTF_CONNECTED;
1791 } else {
1792 /*
1793 * explicitly clear the cloning bit in case ifa_flags sets it.
1794 */
1795 rtflags &= ~RTF_CONNECTED;
1796 }
1797 error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix),
1798 ifa->ifa_addr, sin6tosa(&mask6), rtflags);
1799 if (error == 0) {
1800 nd6_numroutes++;
1801 pr->ndpr_stateflags |= NDPRF_ONLINK;
1802 } else {
1803 nd6log(LOG_ERR, "failed to add route for a"
1804 " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1805 "errno = %d\n",
1806 IN6_PRINT(ip6bufp, &pr->ndpr_prefix.sin6_addr),
1807 pr->ndpr_plen, if_name(ifp),
1808 IN6_PRINT(ip6buf,
1809 &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1810 IN6_PRINT(ip6bufm, &mask6.sin6_addr), rtflags, error);
1811 }
1812 ifa_release(ifa, &psref);
1813 curlwp_bindx(bound);
1814
1815 return (error);
1816 }
1817
1818 static int
1819 nd6_prefix_offlink(struct nd_prefix *pr)
1820 {
1821 int error = 0;
1822 struct ifnet *ifp = pr->ndpr_ifp;
1823 struct nd_prefix *opr;
1824 struct sockaddr_in6 sa6, mask6;
1825 char ip6buf[INET6_ADDRSTRLEN];
1826
1827 ND6_ASSERT_WLOCK();
1828
1829 /* sanity check */
1830 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1831 nd6log(LOG_ERR, "%s/%d is already off-link\n",
1832 IN6_PRINT(ip6buf, &pr->ndpr_prefix.sin6_addr),
1833 pr->ndpr_plen);
1834 return (EEXIST);
1835 }
1836
1837 sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1838 sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1839 error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL,
1840 sin6tosa(&mask6), 0);
1841 if (error == 0) {
1842 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1843 nd6_numroutes--;
1844
1845 /*
1846 * There might be the same prefix on another interface,
1847 * the prefix which could not be on-link just because we have
1848 * the interface route (see comments in nd6_prefix_onlink).
1849 * If there's one, try to make the prefix on-link on the
1850 * interface.
1851 */
1852 ND_PREFIX_LIST_FOREACH(opr) {
1853 if (opr == pr)
1854 continue;
1855
1856 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1857 continue;
1858
1859 /*
1860 * KAME specific: detached prefixes should not be
1861 * on-link.
1862 */
1863 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1864 continue;
1865
1866 if (opr->ndpr_plen == pr->ndpr_plen &&
1867 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1868 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1869 int e;
1870
1871 if ((e = nd6_prefix_onlink(opr)) != 0) {
1872 nd6log(LOG_ERR, "failed to "
1873 "recover a prefix %s/%d from %s "
1874 "to %s (errno = %d)\n",
1875 IN6_PRINT(ip6buf,
1876 &opr->ndpr_prefix.sin6_addr),
1877 opr->ndpr_plen, if_name(ifp),
1878 if_name(opr->ndpr_ifp), e);
1879 }
1880 }
1881 }
1882 } else {
1883 /* XXX: can we still set the NDPRF_ONLINK flag? */
1884 nd6log(LOG_ERR, "failed to delete route: "
1885 "%s/%d on %s (errno = %d)\n",
1886 IN6_PRINT(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1887 if_name(ifp),
1888 error);
1889 }
1890
1891 return error;
1892 }
1893
1894 static struct in6_ifaddr *
1895 in6_ifadd(struct nd_prefixctl *prc, int mcast, struct psref *psref)
1896 {
1897 struct ifnet *ifp = prc->ndprc_ifp;
1898 struct ifaddr *ifa;
1899 struct in6_aliasreq ifra;
1900 struct in6_ifaddr *ia, *ib;
1901 int error, plen0;
1902 struct in6_addr mask;
1903 int prefixlen = prc->ndprc_plen;
1904 int updateflags;
1905 int s;
1906 char ip6buf[INET6_ADDRSTRLEN];
1907
1908 ND6_ASSERT_WLOCK();
1909
1910 in6_prefixlen2mask(&mask, prefixlen);
1911
1912 /*
1913 * find a link-local address (will be interface ID).
1914 * Is it really mandatory? Theoretically, a global or a site-local
1915 * address can be configured without a link-local address, if we
1916 * have a unique interface identifier...
1917 *
1918 * it is not mandatory to have a link-local address, we can generate
1919 * interface identifier on the fly. we do this because:
1920 * (1) it should be the easiest way to find interface identifier.
1921 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1922 * for multiple addresses on a single interface, and possible shortcut
1923 * of DAD. we omitted DAD for this reason in the past.
1924 * (3) a user can prevent autoconfiguration of global address
1925 * by removing link-local address by hand (this is partly because we
1926 * don't have other way to control the use of IPv6 on an interface.
1927 * this has been our design choice - cf. NRL's "ifconfig auto").
1928 * (4) it is easier to manage when an interface has addresses
1929 * with the same interface identifier, than to have multiple addresses
1930 * with different interface identifiers.
1931 */
1932 s = pserialize_read_enter();
1933 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1934 if (ifa)
1935 ib = (struct in6_ifaddr *)ifa;
1936 else {
1937 pserialize_read_exit(s);
1938 return NULL;
1939 }
1940
1941 #if 0 /* don't care link local addr state, and always do DAD */
1942 /* if link-local address is not eligible, do not autoconfigure. */
1943 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1944 printf("in6_ifadd: link-local address not ready\n");
1945 return NULL;
1946 }
1947 #endif
1948
1949 /* prefixlen + ifidlen must be equal to 128 */
1950 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1951 if (prefixlen != plen0) {
1952 nd6log(LOG_INFO, "wrong prefixlen for %s "
1953 "(prefix=%d ifid=%d)\n",
1954 if_name(ifp), prefixlen, 128 - plen0);
1955 pserialize_read_exit(s);
1956 return NULL;
1957 }
1958
1959 /* make ifaddr */
1960
1961 memset(&ifra, 0, sizeof(ifra));
1962 /*
1963 * in6_update_ifa() does not use ifra_name, but we accurately set it
1964 * for safety.
1965 */
1966 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1967 sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
1968 /* prefix */
1969 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1970 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1971 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1972 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1973
1974 /* interface ID */
1975 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1976 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1977 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1978 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1979 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1980 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1981 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1982 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1983 pserialize_read_exit(s);
1984
1985 /* new prefix mask. */
1986 sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1987
1988 /* lifetimes */
1989 ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
1990 ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
1991
1992 /* XXX: scope zone ID? */
1993
1994 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1995
1996 /*
1997 * Make sure that we do not have this address already. This should
1998 * usually not happen, but we can still see this case, e.g., if we
1999 * have manually configured the exact address to be configured.
2000 */
2001 s = pserialize_read_enter();
2002 if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
2003 /* this should be rare enough to make an explicit log */
2004 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
2005 IN6_PRINT(ip6buf, &ifra.ifra_addr.sin6_addr));
2006 pserialize_read_exit(s);
2007 return (NULL);
2008 }
2009 pserialize_read_exit(s);
2010
2011 /*
2012 * Allocate ifaddr structure, link into chain, etc.
2013 * If we are going to create a new address upon receiving a multicasted
2014 * RA, we need to impose a random delay before starting DAD.
2015 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
2016 */
2017 updateflags = 0;
2018 if (mcast)
2019 updateflags |= IN6_IFAUPDATE_DADDELAY;
2020 if ((error = in6_update_ifa(ifp, &ifra, updateflags)) != 0) {
2021 nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n",
2022 IN6_PRINT(ip6buf, &ifra.ifra_addr.sin6_addr), if_name(ifp),
2023 error);
2024 return (NULL); /* ifaddr must not have been allocated. */
2025 }
2026
2027 ia = in6ifa_ifpwithaddr_psref(ifp, &ifra.ifra_addr.sin6_addr, psref);
2028
2029 return (ia); /* this is always non-NULL */
2030 }
2031
2032 int
2033 in6_tmpifadd(
2034 const struct in6_ifaddr *ia0, /* corresponding public address */
2035 int forcegen,
2036 int dad_delay)
2037 {
2038 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2039 struct in6_ifaddr *newia, *ia;
2040 struct in6_aliasreq ifra;
2041 int i, error;
2042 int trylimit = 3; /* XXX: adhoc value */
2043 int updateflags;
2044 u_int32_t randid[2];
2045 u_int32_t vltime0, pltime0;
2046 int s;
2047
2048 ND6_ASSERT_WLOCK();
2049
2050 memset(&ifra, 0, sizeof(ifra));
2051 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
2052 ifra.ifra_addr = ia0->ia_addr;
2053 /* copy prefix mask */
2054 ifra.ifra_prefixmask = ia0->ia_prefixmask;
2055 /* clear the old IFID */
2056 for (i = 0; i < 4; i++) {
2057 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
2058 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
2059 }
2060
2061 again:
2062 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2063 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2064 nd6log(LOG_NOTICE, "failed to find a good random IFID\n");
2065 return (EINVAL);
2066 }
2067 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2068 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2069 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2070 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2071
2072 /*
2073 * in6_get_tmpifid() quite likely provided a unique interface ID.
2074 * However, we may still have a chance to see collision, because
2075 * there may be a time lag between generation of the ID and generation
2076 * of the address. So, we'll do one more sanity check.
2077 */
2078 s = pserialize_read_enter();
2079 IN6_ADDRLIST_READER_FOREACH(ia) {
2080 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2081 &ifra.ifra_addr.sin6_addr)) {
2082 pserialize_read_exit(s);
2083 if (trylimit-- == 0) {
2084 /*
2085 * Give up. Something strange should have
2086 * happened.
2087 */
2088 nd6log(LOG_NOTICE,
2089 "failed to find a unique random IFID\n");
2090 return (EEXIST);
2091 }
2092 forcegen = 1;
2093 goto again;
2094 }
2095 }
2096 pserialize_read_exit(s);
2097
2098 /*
2099 * The Valid Lifetime is the lower of the Valid Lifetime of the
2100 * public address or TEMP_VALID_LIFETIME.
2101 * The Preferred Lifetime is the lower of the Preferred Lifetime
2102 * of the public address or TEMP_PREFERRED_LIFETIME -
2103 * DESYNC_FACTOR.
2104 */
2105 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2106 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2107 (ia0->ia6_lifetime.ia6t_vltime -
2108 (time_uptime - ia0->ia6_updatetime));
2109 if (vltime0 > ip6_temp_valid_lifetime)
2110 vltime0 = ip6_temp_valid_lifetime;
2111 } else
2112 vltime0 = ip6_temp_valid_lifetime;
2113 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2114 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2115 (ia0->ia6_lifetime.ia6t_pltime -
2116 (time_uptime - ia0->ia6_updatetime));
2117 if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
2118 pltime0 = ip6_temp_preferred_lifetime -
2119 ip6_desync_factor;
2120 }
2121 } else
2122 pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
2123 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2124 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2125
2126 /*
2127 * A temporary address is created only if this calculated Preferred
2128 * Lifetime is greater than REGEN_ADVANCE time units.
2129 */
2130 if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
2131 return (0);
2132
2133 /* XXX: scope zone ID? */
2134
2135 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2136
2137 /* allocate ifaddr structure, link into chain, etc. */
2138 updateflags = 0;
2139 if (dad_delay)
2140 updateflags |= IN6_IFAUPDATE_DADDELAY;
2141 if ((error = in6_update_ifa(ifp, &ifra, updateflags)) != 0)
2142 return (error);
2143
2144 s = pserialize_read_enter();
2145 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2146 if (newia == NULL) { /* XXX: can it happen? */
2147 pserialize_read_exit(s);
2148 nd6log(LOG_ERR,
2149 "ifa update succeeded, but we got no ifaddr\n");
2150 return (EINVAL); /* XXX */
2151 }
2152 newia->ia6_ndpr = ia0->ia6_ndpr;
2153 newia->ia6_ndpr->ndpr_refcnt++;
2154 pserialize_read_exit(s);
2155
2156 /*
2157 * A newly added address might affect the status of other addresses.
2158 * XXX: when the temporary address is generated with a new public
2159 * address, the onlink check is redundant. However, it would be safe
2160 * to do the check explicitly everywhere a new address is generated,
2161 * and, in fact, we surely need the check when we create a new
2162 * temporary address due to deprecation of an old temporary address.
2163 */
2164 nd6_pfxlist_onlink_check();
2165
2166 return (0);
2167 }
2168
2169 static int
2170 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2171 {
2172
2173 ND6_ASSERT_WLOCK();
2174
2175 /* check if preferred lifetime > valid lifetime. RFC2462 5.5.3 (c) */
2176 if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
2177 nd6log(LOG_INFO, "preferred lifetime"
2178 "(%d) is greater than valid lifetime(%d)\n",
2179 (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime);
2180 return (EINVAL);
2181 }
2182 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2183 ndpr->ndpr_preferred = 0;
2184 else
2185 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2186 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2187 ndpr->ndpr_expire = 0;
2188 else
2189 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2190
2191 return 0;
2192 }
2193
2194 static void
2195 in6_init_address_ltimes(struct nd_prefix *newpr,
2196 struct in6_addrlifetime *lt6)
2197 {
2198
2199 /* Valid lifetime must not be updated unless explicitly specified. */
2200 /* init ia6t_expire */
2201 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2202 lt6->ia6t_expire = 0;
2203 else {
2204 lt6->ia6t_expire = time_uptime;
2205 lt6->ia6t_expire += lt6->ia6t_vltime;
2206 }
2207
2208 /* init ia6t_preferred */
2209 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2210 lt6->ia6t_preferred = 0;
2211 else {
2212 lt6->ia6t_preferred = time_uptime;
2213 lt6->ia6t_preferred += lt6->ia6t_pltime;
2214 }
2215 }
2216
2217 /*
2218 * Delete all the routing table entries that use the specified gateway.
2219 * XXX: this function causes search through all entries of routing table, so
2220 * it shouldn't be called when acting as a router.
2221 */
2222 void
2223 nd6_rt_flush(struct in6_addr *gateway, struct ifnet *ifp)
2224 {
2225 int s = splsoftnet();
2226
2227 /* We'll care only link-local addresses */
2228 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2229 splx(s);
2230 return;
2231 }
2232
2233 rt_delete_matched_entries(AF_INET6, rt6_deleteroute_matcher, gateway);
2234 splx(s);
2235 }
2236
2237 static int
2238 rt6_deleteroute_matcher(struct rtentry *rt, void *arg)
2239 {
2240 struct in6_addr *gate = (struct in6_addr *)arg;
2241
2242 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2243 return (0);
2244
2245 if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
2246 return (0);
2247
2248 /*
2249 * Do not delete a static route.
2250 * XXX: this seems to be a bit ad-hoc. Should we consider the
2251 * 'cloned' bit instead?
2252 */
2253 if ((rt->rt_flags & RTF_STATIC) != 0)
2254 return (0);
2255
2256 /*
2257 * We delete only host route. This means, in particular, we don't
2258 * delete default route.
2259 */
2260 if ((rt->rt_flags & RTF_HOST) == 0)
2261 return (0);
2262
2263 return 1;
2264 }
2265