xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: nd6_rtr.c,v 1.119 2016/08/16 10:31:57 roy 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 1.119 2016/08/16 10:31:57 roy 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 static void defrouter_delreq(struct nd_defrouter *);
77 
78 static int in6_init_prefix_ltimes(struct nd_prefix *);
79 static void in6_init_address_ltimes(struct nd_prefix *,
80 	struct in6_addrlifetime *);
81 static void purge_detached(struct ifnet *);
82 
83 static int rt6_deleteroute(struct rtentry *, void *);
84 
85 extern int nd6_recalc_reachtm_interval;
86 
87 static struct ifnet *nd6_defifp;
88 int nd6_defifindex;
89 
90 int ip6_use_tempaddr = 0;
91 
92 int ip6_desync_factor;
93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
95 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
96 
97 int nd6_numroutes = 0;
98 
99 /* RTPREF_MEDIUM has to be 0! */
100 #define RTPREF_HIGH	1
101 #define RTPREF_MEDIUM	0
102 #define RTPREF_LOW	(-1)
103 #define RTPREF_RESERVED	(-2)
104 #define RTPREF_INVALID	(-3)	/* internal */
105 
106 static inline bool
107 nd6_is_llinfo_probreach(struct nd_defrouter *dr)
108 {
109 	struct llentry *ln = NULL;
110 
111 	ln = nd6_lookup(&dr->rtaddr, dr->ifp, false);
112 	if (ln == NULL)
113 		return false;
114 
115 	if (!ND6_IS_LLINFO_PROBREACH(ln)) {
116 		LLE_RUNLOCK(ln);
117 		return false;
118 	}
119 
120 	LLE_RUNLOCK(ln);
121 	return true;
122 }
123 
124 /*
125  * Receive Router Solicitation Message - just for routers.
126  * Router solicitation/advertisement is mostly managed by a userland program
127  * (rtadvd) so here we have no function like nd6_ra_output().
128  *
129  * Based on RFC 2461
130  */
131 void
132 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
133 {
134 	struct ifnet *ifp;
135 	struct nd_ifinfo *ndi;
136 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
137 	struct nd_router_solicit *nd_rs;
138 	struct in6_addr saddr6 = ip6->ip6_src;
139 	char *lladdr = NULL;
140 	int lladdrlen = 0;
141 	union nd_opts ndopts;
142 	struct psref psref;
143 
144 	ifp = m_get_rcvif_psref(m, &psref);
145 	if (ifp == NULL)
146 		goto freeit;
147 
148 	ndi = ND_IFINFO(ifp);
149 
150 	/* If I'm not a router, ignore it. */
151 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
152 		goto freeit;
153 
154 	/* Sanity checks */
155 	if (ip6->ip6_hlim != 255) {
156 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
157 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
158 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
159 		goto bad;
160 	}
161 
162 	/*
163 	 * Don't update the neighbor cache, if src = ::.
164 	 * This indicates that the src has no IP address assigned yet.
165 	 */
166 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
167 		goto freeit;
168 
169 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
170 	if (nd_rs == NULL) {
171 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
172 		return;
173 	}
174 
175 	icmp6len -= sizeof(*nd_rs);
176 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
177 	if (nd6_options(&ndopts) < 0) {
178 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
179 		/* nd6_options have incremented stats */
180 		goto freeit;
181 	}
182 
183 	if (ndopts.nd_opts_src_lladdr) {
184 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
185 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
186 	}
187 
188 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
189 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
190 		    "(if %d, RS packet %d)\n",
191 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2);
192 		goto bad;
193 	}
194 
195 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
196 
197  freeit:
198 	m_put_rcvif_psref(ifp, &psref);
199 	m_freem(m);
200 	return;
201 
202  bad:
203 	ICMP6_STATINC(ICMP6_STAT_BADRS);
204 	m_put_rcvif_psref(ifp, &psref);
205 	m_freem(m);
206 }
207 
208 /*
209  * Receive Router Advertisement Message.
210  *
211  * Based on RFC 2461
212  * TODO: on-link bit on prefix information
213  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
214  */
215 void
216 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
217 {
218 	struct ifnet *ifp;
219 	struct nd_ifinfo *ndi;
220 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
221 	struct nd_router_advert *nd_ra;
222 	struct in6_addr saddr6 = ip6->ip6_src;
223 #if 0
224 	struct in6_addr daddr6 = ip6->ip6_dst;
225 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
226 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
227 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
228 #endif
229 	int mcast = 0;
230 	union nd_opts ndopts;
231 	struct nd_defrouter *dr;
232 	struct psref psref;
233 
234 	ifp = m_get_rcvif_psref(m, &psref);
235 	if (ifp == NULL)
236 		goto freeit;
237 
238 	ndi = ND_IFINFO(ifp);
239 	/*
240 	 * We only accept RAs when
241 	 * the system-wide variable allows the acceptance, and the
242 	 * per-interface variable allows RAs on the receiving interface.
243 	 */
244 	if (!nd6_accepts_rtadv(ndi))
245 		goto freeit;
246 
247 	if (ip6->ip6_hlim != 255) {
248 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
249 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
250 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
251 		goto bad;
252 	}
253 
254 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
255 		nd6log(LOG_ERR, "src %s is not link-local\n",
256 		    ip6_sprintf(&saddr6));
257 		goto bad;
258 	}
259 
260 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
261 	if (nd_ra == NULL) {
262 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
263 		m_put_rcvif_psref(ifp, &psref);
264 		return;
265 	}
266 
267 	icmp6len -= sizeof(*nd_ra);
268 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
269 	if (nd6_options(&ndopts) < 0) {
270 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
271 		/* nd6_options have incremented stats */
272 		goto freeit;
273 	}
274 
275     {
276 	struct nd_defrouter drtr;
277 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
278 
279 	/* remember if this is a multicasted advertisement */
280 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
281 		mcast = 1;
282 
283 	memset(&drtr, 0, sizeof(drtr));
284 	drtr.rtaddr = saddr6;
285 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
286 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
287 	drtr.expire = time_uptime + drtr.rtlifetime;
288 	drtr.ifp = ifp;
289 	/* unspecified or not? (RFC 2461 6.3.4) */
290 	if (advreachable) {
291 		NTOHL(advreachable);
292 		if (advreachable <= MAX_REACHABLE_TIME &&
293 		    ndi->basereachable != advreachable) {
294 			ndi->basereachable = advreachable;
295 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
296 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
297 		}
298 	}
299 	if (nd_ra->nd_ra_retransmit)
300 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
301 	if (nd_ra->nd_ra_curhoplimit) {
302 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
303 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
304 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
305 			log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
306 			   "%s on %s (current=%d, received=%d), ignored\n",
307 			   ip6_sprintf(&ip6->ip6_src),
308 			   if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
309 	}
310 	dr = defrtrlist_update(&drtr);
311     }
312 
313 	/*
314 	 * prefix
315 	 */
316 	if (ndopts.nd_opts_pi) {
317 		struct nd_opt_hdr *pt;
318 		struct nd_opt_prefix_info *pi = NULL;
319 		struct nd_prefixctl prc;
320 
321 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
322 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
323 		     pt = (struct nd_opt_hdr *)((char *)pt +
324 						(pt->nd_opt_len << 3))) {
325 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
326 				continue;
327 			pi = (struct nd_opt_prefix_info *)pt;
328 
329 			if (pi->nd_opt_pi_len != 4) {
330 				nd6log(LOG_INFO, "invalid option "
331 				    "len %d for prefix information option, "
332 				    "ignored\n", pi->nd_opt_pi_len);
333 				continue;
334 			}
335 
336 			if (128 < pi->nd_opt_pi_prefix_len) {
337 				nd6log(LOG_INFO, "invalid prefix "
338 				    "len %d for prefix information option, "
339 				    "ignored\n", pi->nd_opt_pi_prefix_len);
340 				continue;
341 			}
342 
343 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
344 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
345 				nd6log(LOG_INFO,
346 				    "invalid prefix %s, ignored\n",
347 				    ip6_sprintf(&pi->nd_opt_pi_prefix));
348 				continue;
349 			}
350 
351 			memset(&prc, 0, sizeof(prc));
352 			sockaddr_in6_init(&prc.ndprc_prefix,
353 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
354 			prc.ndprc_ifp = ifp;
355 
356 			prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
357 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
358 			prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
359 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
360 			prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
361 			prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
362 			prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
363 
364 			(void)prelist_update(&prc, dr, m, mcast);
365 		}
366 	}
367 
368 	/*
369 	 * MTU
370 	 */
371 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
372 		u_long mtu;
373 		u_long maxmtu;
374 
375 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
376 
377 		/* lower bound */
378 		if (mtu < IPV6_MMTU) {
379 			nd6log(LOG_INFO, "bogus mtu option "
380 			    "mtu=%lu sent from %s, ignoring\n",
381 			    mtu, ip6_sprintf(&ip6->ip6_src));
382 			goto skip;
383 		}
384 
385 		/* upper bound */
386 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
387 		    ? ndi->maxmtu : ifp->if_mtu;
388 		if (mtu <= maxmtu) {
389 			int change = (ndi->linkmtu != mtu);
390 
391 			ndi->linkmtu = mtu;
392 			if (change) /* in6_maxmtu may change */
393 				in6_setmaxmtu();
394 		} else {
395 			nd6log(LOG_INFO,
396 			    "bogus mtu mtu=%lu sent from %s; "
397 			    "exceeds maxmtu %lu, ignoring\n",
398 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu);
399 		}
400 	}
401 
402  skip:
403 
404 	/*
405 	 * Source link layer address
406 	 */
407     {
408 	char *lladdr = NULL;
409 	int lladdrlen = 0;
410 
411 	if (ndopts.nd_opts_src_lladdr) {
412 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
413 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
414 	}
415 
416 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
417 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
418 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
419 		    ifp->if_addrlen, lladdrlen - 2);
420 		goto bad;
421 	}
422 
423 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
424 
425 	/*
426 	 * Installing a link-layer address might change the state of the
427 	 * router's neighbor cache, which might also affect our on-link
428 	 * detection of adveritsed prefixes.
429 	 */
430 	pfxlist_onlink_check();
431     }
432 
433  freeit:
434 	m_put_rcvif_psref(ifp, &psref);
435 	m_freem(m);
436 	return;
437 
438  bad:
439 	ICMP6_STATINC(ICMP6_STAT_BADRA);
440 	m_put_rcvif_psref(ifp, &psref);
441 	m_freem(m);
442 }
443 
444 /*
445  * default router list processing sub routines
446  */
447 void
448 defrouter_addreq(struct nd_defrouter *newdr)
449 {
450 	union {
451 		struct sockaddr_in6 sin6;
452 		struct sockaddr sa;
453 	} def, mask, gate;
454 	int s;
455 	int error;
456 
457 	memset(&def, 0, sizeof(def));
458 	memset(&mask, 0, sizeof(mask));
459 	memset(&gate, 0,sizeof(gate)); /* for safety */
460 
461 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
462 	    sizeof(struct sockaddr_in6);
463 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
464 	gate.sin6.sin6_addr = newdr->rtaddr;
465 #ifndef SCOPEDROUTING
466 	gate.sin6.sin6_scope_id = 0;	/* XXX */
467 #endif
468 
469 	s = splsoftnet();
470 	error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
471 	    RTF_GATEWAY);
472 	if (error == 0) {
473 		nd6_numroutes++;
474 		newdr->installed = 1;
475 	}
476 	splx(s);
477 	return;
478 }
479 
480 struct nd_defrouter *
481 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
482 {
483 	struct nd_defrouter *dr;
484 
485 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
486 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
487 			break;
488 	}
489 
490 	return dr;		/* search failed */
491 }
492 
493 void
494 defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
495 {
496 	struct nd_defrouter *deldr = NULL;
497 	struct nd_prefix *pr;
498 	struct nd_ifinfo *ndi;
499 
500 	if (ext == NULL)
501 		ext = dr->ifp->if_afdata[AF_INET6];
502 
503 	/* detach already in progress, can not do anything */
504 	if (ext == NULL)
505 		return;
506 
507 	ndi = ext->nd_ifinfo;
508 
509 	/*
510 	 * Flush all the routing table entries that use the router
511 	 * as a next hop.
512 	 */
513 	/* XXX: better condition? */
514 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
515 		rt6_flush(&dr->rtaddr, dr->ifp);
516 
517 	if (dr->installed) {
518 		deldr = dr;
519 		defrouter_delreq(dr);
520 	}
521 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
522 
523 	/*
524 	 * Also delete all the pointers to the router in each prefix lists.
525 	 */
526 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
527 		struct nd_pfxrouter *pfxrtr;
528 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
529 			pfxrtr_del(pfxrtr);
530 	}
531 	pfxlist_onlink_check();
532 
533 	/*
534 	 * If the router is the primary one, choose a new one.
535 	 * Note that defrouter_select() will remove the current gateway
536 	 * from the routing table.
537 	 */
538 	if (deldr)
539 		defrouter_select();
540 
541 	ext->ndefrouters--;
542 	if (ext->ndefrouters < 0) {
543 		log(LOG_WARNING, "defrtrlist_del: negative count on %s\n",
544 		    dr->ifp->if_xname);
545 	}
546 
547 	free(dr, M_IP6NDP);
548 }
549 
550 /*
551  * Remove the default route for a given router.
552  * This is just a subroutine function for defrouter_select(), and should
553  * not be called from anywhere else.
554  */
555 static void
556 defrouter_delreq(struct nd_defrouter *dr)
557 {
558 	union {
559 		struct sockaddr_in6 sin6;
560 		struct sockaddr sa;
561 	} def, mask, gw;
562 	int error;
563 
564 #ifdef DIAGNOSTIC
565 	if (dr == NULL)
566 		panic("dr == NULL in defrouter_delreq");
567 #endif
568 
569 	memset(&def, 0, sizeof(def));
570 	memset(&mask, 0, sizeof(mask));
571 	memset(&gw, 0, sizeof(gw));	/* for safety */
572 
573 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
574 	    sizeof(struct sockaddr_in6);
575 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
576 	gw.sin6.sin6_addr = dr->rtaddr;
577 #ifndef SCOPEDROUTING
578 	gw.sin6.sin6_scope_id = 0;	/* XXX */
579 #endif
580 
581 	error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
582 	    RTF_GATEWAY);
583 	if (error == 0)
584 		nd6_numroutes--;
585 
586 	dr->installed = 0;
587 }
588 
589 /*
590  * remove all default routes from default router list
591  */
592 void
593 defrouter_reset(void)
594 {
595 	struct nd_defrouter *dr;
596 
597 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
598 	     dr = TAILQ_NEXT(dr, dr_entry))
599 		defrouter_delreq(dr);
600 
601 	/*
602 	 * XXX should we also nuke any default routers in the kernel, by
603 	 * going through them by rtalloc1()?
604 	 */
605 }
606 
607 /*
608  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
609  * draft-ietf-ipngwg-router-selection:
610  * 1) Routers that are reachable or probably reachable should be preferred.
611  *    If we have more than one (probably) reachable router, prefer ones
612  *    with the highest router preference.
613  * 2) When no routers on the list are known to be reachable or
614  *    probably reachable, routers SHOULD be selected in a round-robin
615  *    fashion, regardless of router preference values.
616  * 3) If the Default Router List is empty, assume that all
617  *    destinations are on-link.
618  *
619  * We assume nd_defrouter is sorted by router preference value.
620  * Since the code below covers both with and without router preference cases,
621  * we do not need to classify the cases by ifdef.
622  *
623  * At this moment, we do not try to install more than one default router,
624  * even when the multipath routing is available, because we're not sure about
625  * the benefits for stub hosts comparing to the risk of making the code
626  * complicated and the possibility of introducing bugs.
627  */
628 void
629 defrouter_select(void)
630 {
631 	struct nd_ifinfo *ndi;
632 	int s = splsoftnet();
633 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
634 
635 	/*
636 	 * This function should be called only when acting as an autoconfigured
637 	 * host.  Although the remaining part of this function is not effective
638 	 * if the node is not an autoconfigured host, we explicitly exclude
639 	 * such cases here for safety.
640 	 */
641 	if (ip6_forwarding) {
642 		nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, "
643 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv);
644 		splx(s);
645 		return;
646 	}
647 
648 	/*
649 	 * Let's handle easy case (3) first:
650 	 * If default router list is empty, there's nothing to be done.
651 	 */
652 	if (!TAILQ_FIRST(&nd_defrouter)) {
653 		splx(s);
654 		return;
655 	}
656 
657 	/*
658 	 * Search for a (probably) reachable router from the list.
659 	 * We just pick up the first reachable one (if any), assuming that
660 	 * the ordering rule of the list described in defrtrlist_update().
661 	 */
662 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
663 	     dr = TAILQ_NEXT(dr, dr_entry)) {
664 		ndi = ND_IFINFO(dr->ifp);
665 		if (nd6_accepts_rtadv(ndi))
666 			continue;
667 
668 		if (selected_dr == NULL &&
669 		    nd6_is_llinfo_probreach(dr))
670 			selected_dr = dr;
671 
672 		if (dr->installed && !installed_dr)
673 			installed_dr = dr;
674 		else if (dr->installed && installed_dr) {
675 			/* this should not happen.  warn for diagnosis. */
676 			log(LOG_ERR, "defrouter_select: more than one router"
677 			    " is installed\n");
678 		}
679 	}
680 	/*
681 	 * If none of the default routers was found to be reachable,
682 	 * round-robin the list regardless of preference.
683 	 * Otherwise, if we have an installed router, check if the selected
684 	 * (reachable) router should really be preferred to the installed one.
685 	 * We only prefer the new router when the old one is not reachable
686 	 * or when the new one has a really higher preference value.
687 	 */
688 	if (selected_dr == NULL) {
689 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
690 			selected_dr = TAILQ_FIRST(&nd_defrouter);
691 		else
692 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
693 	} else if (installed_dr &&
694 	    nd6_is_llinfo_probreach(installed_dr) &&
695 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
696 		selected_dr = installed_dr;
697 	}
698 
699 	/*
700 	 * If the selected router is different than the installed one,
701 	 * remove the installed router and install the selected one.
702 	 * Note that the selected router is never NULL here.
703 	 */
704 	if (installed_dr != selected_dr) {
705 		if (installed_dr)
706 			defrouter_delreq(installed_dr);
707 		defrouter_addreq(selected_dr);
708 	}
709 
710 	splx(s);
711 	return;
712 }
713 
714 /*
715  * for default router selection
716  * regards router-preference field as a 2-bit signed integer
717  */
718 static int
719 rtpref(struct nd_defrouter *dr)
720 {
721 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
722 	case ND_RA_FLAG_RTPREF_HIGH:
723 		return (RTPREF_HIGH);
724 	case ND_RA_FLAG_RTPREF_MEDIUM:
725 	case ND_RA_FLAG_RTPREF_RSV:
726 		return (RTPREF_MEDIUM);
727 	case ND_RA_FLAG_RTPREF_LOW:
728 		return (RTPREF_LOW);
729 	default:
730 		/*
731 		 * This case should never happen.  If it did, it would mean a
732 		 * serious bug of kernel internal.  We thus always bark here.
733 		 * Or, can we even panic?
734 		 */
735 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
736 		return (RTPREF_INVALID);
737 	}
738 	/* NOTREACHED */
739 }
740 
741 static struct nd_defrouter *
742 defrtrlist_update(struct nd_defrouter *newdr)
743 {
744 	struct nd_defrouter *dr, *n;
745 	struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
746 	int s = splsoftnet();
747 
748 	if ((dr = defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
749 		/* entry exists */
750 		if (newdr->rtlifetime == 0) {
751 			defrtrlist_del(dr, ext);
752 			dr = NULL;
753 		} else {
754 			int oldpref = rtpref(dr);
755 
756 			/* override */
757 			dr->flags = newdr->flags; /* xxx flag check */
758 			dr->rtlifetime = newdr->rtlifetime;
759 			dr->expire = newdr->expire;
760 
761 			/*
762 			 * If the preference does not change, there's no need
763 			 * to sort the entries.
764 			 */
765 			if (rtpref(newdr) == oldpref) {
766 				splx(s);
767 				return (dr);
768 			}
769 
770 			/*
771 			 * preferred router may be changed, so relocate
772 			 * this router.
773 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
774 			 * However, since defrtrlist_del() has many side
775 			 * effects, we intentionally do so here.
776 			 * defrouter_select() below will handle routing
777 			 * changes later.
778 			 */
779 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
780 			n = dr;
781 			goto insert;
782 		}
783 		splx(s);
784 		return (dr);
785 	}
786 
787 	if (ip6_maxifdefrouters >= 0 &&
788 	    ext->ndefrouters >= ip6_maxifdefrouters) {
789 		splx(s);
790 		return (NULL);
791 	}
792 
793 	/* entry does not exist */
794 	if (newdr->rtlifetime == 0) {
795 		splx(s);
796 		return (NULL);
797 	}
798 
799 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
800 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
801 		splx(s);
802 		return (NULL);
803 	}
804 
805 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
806 	if (n == NULL) {
807 		splx(s);
808 		return (NULL);
809 	}
810 	memset(n, 0, sizeof(*n));
811 	*n = *newdr;
812 
813 insert:
814 	/*
815 	 * Insert the new router in the Default Router List;
816 	 * The Default Router List should be in the descending order
817 	 * of router-preferece.  Routers with the same preference are
818 	 * sorted in the arriving time order.
819 	 */
820 
821 	/* insert at the end of the group */
822 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
823 	     dr = TAILQ_NEXT(dr, dr_entry)) {
824 		if (rtpref(n) > rtpref(dr))
825 			break;
826 	}
827 	if (dr)
828 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
829 	else
830 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
831 
832 	defrouter_select();
833 
834 	ext->ndefrouters++;
835 
836 	splx(s);
837 
838 	return (n);
839 }
840 
841 static struct nd_pfxrouter *
842 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
843 {
844 	struct nd_pfxrouter *search;
845 
846 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
847 		if (search->router == dr)
848 			break;
849 	}
850 
851 	return (search);
852 }
853 
854 static void
855 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
856 {
857 	struct nd_pfxrouter *newpfr;
858 
859 	newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
860 	if (newpfr == NULL)
861 		return;
862 	newpfr->router = dr;
863 
864 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
865 
866 	pfxlist_onlink_check();
867 }
868 
869 static void
870 pfxrtr_del(struct nd_pfxrouter *pfr)
871 {
872 	LIST_REMOVE(pfr, pfr_entry);
873 	free(pfr, M_IP6NDP);
874 }
875 
876 struct nd_prefix *
877 nd6_prefix_lookup(struct nd_prefixctl *key)
878 {
879 	struct nd_prefix *search;
880 
881 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
882 		if (key->ndprc_ifp == search->ndpr_ifp &&
883 		    key->ndprc_plen == search->ndpr_plen &&
884 		    in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
885 		    &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
886 			break;
887 		}
888 	}
889 
890 	return (search);
891 }
892 
893 static void
894 purge_detached(struct ifnet *ifp)
895 {
896 	struct nd_prefix *pr, *pr_next;
897 	struct in6_ifaddr *ia;
898 	struct ifaddr *ifa, *ifa_next;
899 
900 	for (pr = nd_prefix.lh_first; pr; pr = pr_next) {
901 		int s;
902 		pr_next = pr->ndpr_next;
903 
904 		/*
905 		 * This function is called when we need to make more room for
906 		 * new prefixes rather than keeping old, possibly stale ones.
907 		 * Detached prefixes would be a good candidate; if all routers
908 		 * that advertised the prefix expired, the prefix is also
909 		 * probably stale.
910 		 */
911 		if (pr->ndpr_ifp != ifp ||
912 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
913 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
914 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
915 			continue;
916 
917 	restart:
918 		s = pserialize_read_enter();
919 		for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) {
920 			ifa_next = IFADDR_READER_NEXT(ifa);
921 			if (ifa->ifa_addr->sa_family != AF_INET6)
922 				continue;
923 			ia = (struct in6_ifaddr *)ifa;
924 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
925 			    IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
926 				pserialize_read_exit(s);
927 				in6_purgeaddr(ifa);
928 				goto restart;
929 			}
930 		}
931 		pserialize_read_exit(s);
932 
933 		if (pr->ndpr_refcnt == 0)
934 			prelist_remove(pr);
935 	}
936 }
937 int
938 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
939 	struct nd_prefix **newp)
940 {
941 	struct nd_prefix *newpr = NULL;
942 	int i, s;
943 	int error;
944 	struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
945 
946 	if (ip6_maxifprefixes >= 0) {
947 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
948 			purge_detached(prc->ndprc_ifp);
949 		if (ext->nprefixes >= ip6_maxifprefixes)
950 			return ENOMEM;
951 	}
952 
953 	error = 0;
954 	newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
955 	if (newpr == NULL)
956 		return ENOMEM;
957 	newpr->ndpr_ifp = prc->ndprc_ifp;
958 	newpr->ndpr_prefix = prc->ndprc_prefix;
959 	newpr->ndpr_plen = prc->ndprc_plen;
960 	newpr->ndpr_vltime = prc->ndprc_vltime;
961 	newpr->ndpr_pltime = prc->ndprc_pltime;
962 	newpr->ndpr_flags = prc->ndprc_flags;
963 	if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
964 		free(newpr, M_IP6NDP);
965 		return(error);
966 	}
967 	newpr->ndpr_lastupdate = time_uptime;
968 	if (newp != NULL)
969 		*newp = newpr;
970 
971 	/* initialization */
972 	LIST_INIT(&newpr->ndpr_advrtrs);
973 	in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
974 	/* make prefix in the canonical form */
975 	for (i = 0; i < 4; i++) {
976 		newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
977 		    newpr->ndpr_mask.s6_addr32[i];
978 	}
979 
980 	s = splsoftnet();
981 	/* link ndpr_entry to nd_prefix list */
982 	LIST_INSERT_HEAD(&nd_prefix, newpr, ndpr_entry);
983 	splx(s);
984 
985 	/* ND_OPT_PI_FLAG_ONLINK processing */
986 	if (newpr->ndpr_raf_onlink) {
987 		int e;
988 
989 		if ((e = nd6_prefix_onlink(newpr)) != 0) {
990 			nd6log(LOG_ERR, "failed to make "
991 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
992 			    ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
993 			    prc->ndprc_plen, if_name(prc->ndprc_ifp), e);
994 			/* proceed anyway. XXX: is it correct? */
995 		}
996 	}
997 
998 	if (dr)
999 		pfxrtr_add(newpr, dr);
1000 
1001 	ext->nprefixes++;
1002 
1003 	return 0;
1004 }
1005 
1006 void
1007 prelist_remove(struct nd_prefix *pr)
1008 {
1009 	struct nd_pfxrouter *pfr, *next;
1010 	int e, s;
1011 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
1012 
1013 	/* make sure to invalidate the prefix until it is really freed. */
1014 	pr->ndpr_vltime = 0;
1015 	pr->ndpr_pltime = 0;
1016 #if 0
1017 	/*
1018 	 * Though these flags are now meaningless, we'd rather keep the value
1019 	 * not to confuse users when executing "ndp -p".
1020 	 */
1021 	pr->ndpr_raf_onlink = 0;
1022 	pr->ndpr_raf_auto = 0;
1023 #endif
1024 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
1025 	    (e = nd6_prefix_offlink(pr)) != 0) {
1026 		nd6log(LOG_ERR,
1027 		    "failed to make %s/%d offlink on %s, errno=%d\n",
1028 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1029 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1030 		/* what should we do? */
1031 	}
1032 
1033 	if (pr->ndpr_refcnt > 0)
1034 		return;		/* notice here? */
1035 
1036 	s = splsoftnet();
1037 	/* unlink ndpr_entry from nd_prefix list */
1038 	LIST_REMOVE(pr, ndpr_entry);
1039 
1040 	/* free list of routers that adversed the prefix */
1041 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
1042 		next = LIST_NEXT(pfr, pfr_entry);
1043 
1044 		free(pfr, M_IP6NDP);
1045 	}
1046 
1047 	if (ext) {
1048 		ext->nprefixes--;
1049 		if (ext->nprefixes < 0) {
1050 			log(LOG_WARNING, "prelist_remove: negative count on "
1051 			    "%s\n", pr->ndpr_ifp->if_xname);
1052 		}
1053 	}
1054 	splx(s);
1055 
1056 	free(pr, M_IP6NDP);
1057 
1058 	pfxlist_onlink_check();
1059 }
1060 
1061 static int
1062 prelist_update(struct nd_prefixctl *newprc,
1063 	struct nd_defrouter *dr, /* may be NULL */
1064 	struct mbuf *m,
1065 	int mcast)
1066 {
1067 	struct in6_ifaddr *ia6_match = NULL;
1068 	struct ifaddr *ifa;
1069 	struct ifnet *ifp = newprc->ndprc_ifp;
1070 	struct nd_prefix *pr;
1071 	int s = splsoftnet();
1072 	int error = 0;
1073 	int auth;
1074 	struct in6_addrlifetime lt6_tmp;
1075 	int ss;
1076 
1077 	auth = 0;
1078 	if (m) {
1079 		/*
1080 		 * Authenticity for NA consists authentication for
1081 		 * both IP header and IP datagrams, doesn't it ?
1082 		 */
1083 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1084 		auth = (m->m_flags & M_AUTHIPHDR
1085 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1086 #endif
1087 	}
1088 
1089 	if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
1090 		/*
1091 		 * nd6_prefix_lookup() ensures that pr and newprc have the same
1092 		 * prefix on a same interface.
1093 		 */
1094 
1095 		/*
1096 		 * Update prefix information.  Note that the on-link (L) bit
1097 		 * and the autonomous (A) bit should NOT be changed from 1
1098 		 * to 0.
1099 		 */
1100 		if (newprc->ndprc_raf_onlink == 1)
1101 			pr->ndpr_raf_onlink = 1;
1102 		if (newprc->ndprc_raf_auto == 1)
1103 			pr->ndpr_raf_auto = 1;
1104 		if (newprc->ndprc_raf_onlink) {
1105 			pr->ndpr_vltime = newprc->ndprc_vltime;
1106 			pr->ndpr_pltime = newprc->ndprc_pltime;
1107 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1108 			pr->ndpr_lastupdate = time_uptime;
1109 		}
1110 
1111 		if (newprc->ndprc_raf_onlink &&
1112 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1113 			int e;
1114 
1115 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1116 				nd6log(LOG_ERR,
1117 				    "failed to make "
1118 				    "the prefix %s/%d on-link on %s "
1119 				    "(errno=%d)\n",
1120 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1121 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
1122 				/* proceed anyway. XXX: is it correct? */
1123 			}
1124 		}
1125 
1126 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1127 			pfxrtr_add(pr, dr);
1128 	} else {
1129 		struct nd_prefix *newpr = NULL;
1130 
1131 		if (newprc->ndprc_vltime == 0)
1132 			goto end;
1133 		if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
1134 			goto end;
1135 
1136 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1137 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1138 			goto end;
1139 		}
1140 
1141 		error = nd6_prelist_add(newprc, dr, &newpr);
1142 		if (error != 0 || newpr == NULL) {
1143 			nd6log(LOG_NOTICE,
1144 			    "nd6_prelist_add failed for %s/%d on %s "
1145 			    "errno=%d, returnpr=%p\n",
1146 			    ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
1147 			    newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
1148 			    error, newpr);
1149 			goto end; /* we should just give up in this case. */
1150 		}
1151 
1152 		/*
1153 		 * XXX: from the ND point of view, we can ignore a prefix
1154 		 * with the on-link bit being zero.  However, we need a
1155 		 * prefix structure for references from autoconfigured
1156 		 * addresses.  Thus, we explicitly make sure that the prefix
1157 		 * itself expires now.
1158 		 */
1159 		if (newpr->ndpr_raf_onlink == 0) {
1160 			newpr->ndpr_vltime = 0;
1161 			newpr->ndpr_pltime = 0;
1162 			in6_init_prefix_ltimes(newpr);
1163 		}
1164 
1165 		pr = newpr;
1166 	}
1167 
1168 	/*
1169 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1170 	 * Note that pr must be non NULL at this point.
1171 	 */
1172 
1173 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1174 	if (!newprc->ndprc_raf_auto)
1175 		goto end;
1176 
1177 	/*
1178 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1179 	 * nd6_ra_input.
1180 	 */
1181 
1182 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1183 	if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
1184 		error = EINVAL;	/* XXX: won't be used */
1185 		goto end;
1186 	}
1187 
1188 	/*
1189 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1190 	 * an address configured by stateless autoconfiguration already in the
1191 	 * list of addresses associated with the interface, and the Valid
1192 	 * Lifetime is not 0, form an address.  We first check if we have
1193 	 * a matching prefix.
1194 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1195 	 * consider autoconfigured addresses while RFC2462 simply said
1196 	 * "address".
1197 	 */
1198 	ss = pserialize_read_enter();
1199 	IFADDR_READER_FOREACH(ifa, ifp) {
1200 		struct in6_ifaddr *ia6;
1201 		u_int32_t remaininglifetime;
1202 
1203 		if (ifa->ifa_addr->sa_family != AF_INET6)
1204 			continue;
1205 
1206 		ia6 = (struct in6_ifaddr *)ifa;
1207 
1208 		/*
1209 		 * We only consider autoconfigured addresses as per rfc2462bis.
1210 		 */
1211 		if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF))
1212 			continue;
1213 
1214 		/*
1215 		 * Spec is not clear here, but I believe we should concentrate
1216 		 * on unicast (i.e. not anycast) addresses.
1217 		 * XXX: other ia6_flags? detached or duplicated?
1218 		 */
1219 		if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1220 			continue;
1221 
1222 		/*
1223 		 * Ignore the address if it is not associated with a prefix
1224 		 * or is associated with a prefix that is different from this
1225 		 * one.  (pr is never NULL here)
1226 		 */
1227 		if (ia6->ia6_ndpr != pr)
1228 			continue;
1229 
1230 		if (ia6_match == NULL) /* remember the first one */
1231 			ia6_match = ia6;
1232 
1233 		/*
1234 		 * An already autoconfigured address matched.  Now that we
1235 		 * are sure there is at least one matched address, we can
1236 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1237 		 * "two hours" rule and the privacy extension.
1238 		 * We apply some clarifications in rfc2462bis:
1239 		 * - use remaininglifetime instead of storedlifetime as a
1240 		 *   variable name
1241 		 * - remove the dead code in the "two-hour" rule
1242 		 */
1243 #define TWOHOUR		(120*60)
1244 		lt6_tmp = ia6->ia6_lifetime;
1245 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1246 			remaininglifetime = ND6_INFINITE_LIFETIME;
1247 		else if (time_uptime - ia6->ia6_updatetime >
1248 			 lt6_tmp.ia6t_vltime) {
1249 			/*
1250 			 * The case of "invalid" address.  We should usually
1251 			 * not see this case.
1252 			 */
1253 			remaininglifetime = 0;
1254 		} else
1255 			remaininglifetime = lt6_tmp.ia6t_vltime -
1256 			    (time_uptime - ia6->ia6_updatetime);
1257 
1258 		/* when not updating, keep the current stored lifetime. */
1259 		lt6_tmp.ia6t_vltime = remaininglifetime;
1260 
1261 		if (TWOHOUR < newprc->ndprc_vltime ||
1262 		    remaininglifetime < newprc->ndprc_vltime) {
1263 			lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1264 		} else if (remaininglifetime <= TWOHOUR) {
1265 			if (auth)
1266 				lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1267 		} else {
1268 			/*
1269 			 * newprc->ndprc_vltime <= TWOHOUR &&
1270 			 * TWOHOUR < remaininglifetime
1271 			 */
1272 			lt6_tmp.ia6t_vltime = TWOHOUR;
1273 		}
1274 
1275 		/* The 2 hour rule is not imposed for preferred lifetime. */
1276 		lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
1277 
1278 		in6_init_address_ltimes(pr, &lt6_tmp);
1279 
1280 		/*
1281 		 * We need to treat lifetimes for temporary addresses
1282 		 * differently, according to
1283 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1284 		 * we only update the lifetimes when they are in the maximum
1285 		 * intervals.
1286 		 */
1287 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1288 			u_int32_t maxvltime, maxpltime;
1289 
1290 			if (ip6_temp_valid_lifetime >
1291 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1292 			    ip6_desync_factor)) {
1293 				maxvltime = ip6_temp_valid_lifetime -
1294 				    (time_uptime - ia6->ia6_createtime) -
1295 				    ip6_desync_factor;
1296 			} else
1297 				maxvltime = 0;
1298 			if (ip6_temp_preferred_lifetime >
1299 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
1300 			    ip6_desync_factor)) {
1301 				maxpltime = ip6_temp_preferred_lifetime -
1302 				    (time_uptime - ia6->ia6_createtime) -
1303 				    ip6_desync_factor;
1304 			} else
1305 				maxpltime = 0;
1306 
1307 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1308 			    lt6_tmp.ia6t_vltime > maxvltime) {
1309 				lt6_tmp.ia6t_vltime = maxvltime;
1310 			}
1311 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1312 			    lt6_tmp.ia6t_pltime > maxpltime) {
1313 				lt6_tmp.ia6t_pltime = maxpltime;
1314 			}
1315 		}
1316 
1317 		ia6->ia6_lifetime = lt6_tmp;
1318 		ia6->ia6_updatetime = time_uptime;
1319 	}
1320 	pserialize_read_exit(ss);
1321 
1322 	if (ia6_match == NULL && newprc->ndprc_vltime) {
1323 		int ifidlen;
1324 		struct in6_ifaddr *ia6;
1325 		struct psref psref;
1326 
1327 		/*
1328 		 * 5.5.3 (d) (continued)
1329 		 * No address matched and the valid lifetime is non-zero.
1330 		 * Create a new address.
1331 		 */
1332 
1333 		/*
1334 		 * Prefix Length check:
1335 		 * If the sum of the prefix length and interface identifier
1336 		 * length does not equal 128 bits, the Prefix Information
1337 		 * option MUST be ignored.  The length of the interface
1338 		 * identifier is defined in a separate link-type specific
1339 		 * document.
1340 		 */
1341 		ifidlen = in6_if2idlen(ifp);
1342 		if (ifidlen < 0) {
1343 			/* this should not happen, so we always log it. */
1344 			log(LOG_ERR, "%s: IFID undefined (%s)\n",
1345 			    __func__, if_name(ifp));
1346 			goto end;
1347 		}
1348 		if (ifidlen + pr->ndpr_plen != 128) {
1349 			nd6log(LOG_INFO,
1350 			    "invalid prefixlen %d for %s, ignored\n",
1351 			    pr->ndpr_plen, if_name(ifp));
1352 			goto end;
1353 		}
1354 
1355 		if ((ia6 = in6_ifadd(newprc, mcast, &psref)) != NULL) {
1356 			/*
1357 			 * note that we should use pr (not newprc) for reference.
1358 			 */
1359 			pr->ndpr_refcnt++;
1360 			ia6->ia6_ndpr = pr;
1361 
1362 			/* toggle onlink state if the address was assigned
1363 			 * a prefix route. */
1364 			if (ia6->ia_flags & IFA_ROUTE)
1365 				pr->ndpr_stateflags |= NDPRF_ONLINK;
1366 
1367 			/*
1368 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1369 			 * When a new public address is created as described
1370 			 * in RFC2462, also create a new temporary address.
1371 			 *
1372 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1373 			 * When an interface connects to a new link, a new
1374 			 * randomized interface identifier should be generated
1375 			 * immediately together with a new set of temporary
1376 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1377 			 * in6_tmpifadd().
1378 			 */
1379 			if (ip6_use_tempaddr) {
1380 				int e;
1381 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1382 					nd6log(LOG_NOTICE,
1383 					    "failed to create a temporary "
1384 					    "address, errno=%d\n", e);
1385 				}
1386 			}
1387 			ia6_release(ia6, &psref);
1388 
1389 			/*
1390 			 * A newly added address might affect the status
1391 			 * of other addresses, so we check and update it.
1392 			 * XXX: what if address duplication happens?
1393 			 */
1394 			pfxlist_onlink_check();
1395 		} else {
1396 			/* just set an error. do not bark here. */
1397 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1398 		}
1399 	}
1400 
1401  end:
1402 	splx(s);
1403 	return error;
1404 }
1405 
1406 /*
1407  * A supplement function used in the on-link detection below;
1408  * detect if a given prefix has a (probably) reachable advertising router.
1409  * XXX: lengthy function name...
1410  */
1411 static struct nd_pfxrouter *
1412 find_pfxlist_reachable_router(struct nd_prefix *pr)
1413 {
1414 	struct nd_pfxrouter *pfxrtr;
1415 
1416 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1417 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1418 		if (pfxrtr->router->ifp->if_flags & IFF_UP &&
1419 		    pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
1420 		    nd6_is_llinfo_probreach(pfxrtr->router))
1421 			break;	/* found */
1422 	}
1423 
1424 	return (pfxrtr);
1425 }
1426 
1427 /*
1428  * Check if each prefix in the prefix list has at least one available router
1429  * that advertised the prefix (a router is "available" if its neighbor cache
1430  * entry is reachable or probably reachable).
1431  * If the check fails, the prefix may be off-link, because, for example,
1432  * we have moved from the network but the lifetime of the prefix has not
1433  * expired yet.  So we should not use the prefix if there is another prefix
1434  * that has an available router.
1435  * But, if there is no prefix that has an available router, we still regards
1436  * all the prefixes as on-link.  This is because we can't tell if all the
1437  * routers are simply dead or if we really moved from the network and there
1438  * is no router around us.
1439  */
1440 void
1441 pfxlist_onlink_check(void)
1442 {
1443 	struct nd_prefix *pr;
1444 	struct in6_ifaddr *ia;
1445 	struct nd_defrouter *dr;
1446 	struct nd_pfxrouter *pfxrtr = NULL;
1447 	int s;
1448 
1449 	/*
1450 	 * Check if there is a prefix that has a reachable advertising
1451 	 * router.
1452 	 */
1453 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1454 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1455 			break;
1456 	}
1457 	/*
1458 	 * If we have no such prefix, check whether we still have a router
1459 	 * that does not advertise any prefixes.
1460 	 */
1461 	if (pr == NULL) {
1462 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1463 			struct nd_prefix *pr0;
1464 
1465 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1466 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1467 					break;
1468 			}
1469 			if (pfxrtr)
1470 				break;
1471 		}
1472 	}
1473 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1474 		/*
1475 		 * There is at least one prefix that has a reachable router,
1476 		 * or at least a router which probably does not advertise
1477 		 * any prefixes.  The latter would be the case when we move
1478 		 * to a new link where we have a router that does not provide
1479 		 * prefixes and we configure an address by hand.
1480 		 * Detach prefixes which have no reachable advertising
1481 		 * router, and attach other prefixes.
1482 		 */
1483 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1484 			/* XXX: a link-local prefix should never be detached */
1485 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1486 				continue;
1487 
1488 			/*
1489 			 * we aren't interested in prefixes without the L bit
1490 			 * set.
1491 			 */
1492 			if (pr->ndpr_raf_onlink == 0)
1493 				continue;
1494 
1495 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1496 			    find_pfxlist_reachable_router(pr) == NULL)
1497 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1498 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1499 			    find_pfxlist_reachable_router(pr) != 0)
1500 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1501 		}
1502 	} else {
1503 		/* there is no prefix that has a reachable router */
1504 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1505 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1506 				continue;
1507 
1508 			if (pr->ndpr_raf_onlink == 0)
1509 				continue;
1510 
1511 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1512 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1513 		}
1514 	}
1515 
1516 	/*
1517 	 * Remove each interface route associated with a (just) detached
1518 	 * prefix, and reinstall the interface route for a (just) attached
1519 	 * prefix.  Note that all attempt of reinstallation does not
1520 	 * necessarily success, when a same prefix is shared among multiple
1521 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1522 	 * so we don't have to care about them.
1523 	 */
1524 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1525 		int e;
1526 
1527 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1528 			continue;
1529 
1530 		if (pr->ndpr_raf_onlink == 0)
1531 			continue;
1532 
1533 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1534 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1535 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1536 				nd6log(LOG_ERR,
1537 				    "failed to make %s/%d offlink, errno=%d\n",
1538 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1539 				    pr->ndpr_plen, e);
1540 			}
1541 		}
1542 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1543 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1544 		    pr->ndpr_raf_onlink) {
1545 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1546 				nd6log(LOG_ERR,
1547 				    "failed to make %s/%d onlink, errno=%d\n",
1548 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1549 				    pr->ndpr_plen, e);
1550 			}
1551 		}
1552 	}
1553 
1554 	/*
1555 	 * Changes on the prefix status might affect address status as well.
1556 	 * Make sure that all addresses derived from an attached prefix are
1557 	 * attached, and that all addresses derived from a detached prefix are
1558 	 * detached.  Note, however, that a manually configured address should
1559 	 * always be attached.
1560 	 * The precise detection logic is same as the one for prefixes.
1561 	 */
1562 	s = pserialize_read_enter();
1563 	IN6_ADDRLIST_READER_FOREACH(ia) {
1564 		if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
1565 			continue;
1566 
1567 		if (ia->ia6_ndpr == NULL) {
1568 			/*
1569 			 * This can happen when we first configure the address
1570 			 * (i.e. the address exists, but the prefix does not).
1571 			 * XXX: complicated relationships...
1572 			 */
1573 			continue;
1574 		}
1575 
1576 		if (find_pfxlist_reachable_router(ia->ia6_ndpr))
1577 			break;
1578 	}
1579 	pserialize_read_exit(s);
1580 
1581 	if (ia) {
1582 		int bound = curlwp_bind();
1583 
1584 		s = pserialize_read_enter();
1585 		IN6_ADDRLIST_READER_FOREACH(ia) {
1586 			struct ifaddr *ifa = (struct ifaddr *)ia;
1587 			struct psref psref;
1588 
1589 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1590 				continue;
1591 
1592 			if (ia->ia6_ndpr == NULL) /* XXX: see above. */
1593 				continue;
1594 
1595 			ia6_acquire(ia, &psref);
1596 			pserialize_read_exit(s);
1597 
1598 			if (find_pfxlist_reachable_router(ia->ia6_ndpr)) {
1599 				if (ia->ia6_flags & IN6_IFF_DETACHED) {
1600 					ia->ia6_flags &= ~IN6_IFF_DETACHED;
1601 					ia->ia6_flags |= IN6_IFF_TENTATIVE;
1602 					nd6_dad_start(ifa,
1603 					    0);
1604 					/* We will notify the routing socket
1605 					 * of the DAD result, so no need to
1606 					 * here */
1607 				}
1608 			} else {
1609 				if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) {
1610 					ia->ia6_flags |= IN6_IFF_DETACHED;
1611 					rt_newaddrmsg(RTM_NEWADDR,
1612 					    ifa, 0, NULL);
1613 				}
1614 			}
1615 
1616 			s = pserialize_read_enter();
1617 			ia6_release(ia, &psref);
1618 		}
1619 		pserialize_read_exit(s);
1620 		curlwp_bindx(bound);
1621 	}
1622 	else {
1623 		int bound = curlwp_bind();
1624 
1625 		s = pserialize_read_enter();
1626 		IN6_ADDRLIST_READER_FOREACH(ia) {
1627 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1628 				continue;
1629 
1630 			if (ia->ia6_flags & IN6_IFF_DETACHED) {
1631 				struct ifaddr *ifa = (struct ifaddr *)ia;
1632 				struct psref psref;
1633 
1634 				ia->ia6_flags &= ~IN6_IFF_DETACHED;
1635 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
1636 
1637 				ia6_acquire(ia, &psref);
1638 				pserialize_read_exit(s);
1639 
1640 				/* Do we need a delay in this case? */
1641 				nd6_dad_start(ifa, 0);
1642 
1643 				s = pserialize_read_enter();
1644 				ia6_release(ia, &psref);
1645 			}
1646 		}
1647 		pserialize_read_exit(s);
1648 		curlwp_bindx(bound);
1649 	}
1650 }
1651 
1652 int
1653 nd6_prefix_onlink(struct nd_prefix *pr)
1654 {
1655 	struct ifaddr *ifa;
1656 	struct ifnet *ifp = pr->ndpr_ifp;
1657 	struct sockaddr_in6 mask6;
1658 	struct nd_prefix *opr;
1659 	u_long rtflags;
1660 	int error = 0;
1661 	struct psref psref;
1662 	int bound;
1663 
1664 	/* sanity check */
1665 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1666 		nd6log(LOG_ERR, "%s/%d is already on-link\n",
1667 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
1668 		return (EEXIST);
1669 	}
1670 
1671 	/*
1672 	 * Add the interface route associated with the prefix.  Before
1673 	 * installing the route, check if there's the same prefix on another
1674 	 * interface, and the prefix has already installed the interface route.
1675 	 * Although such a configuration is expected to be rare, we explicitly
1676 	 * allow it.
1677 	 */
1678 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1679 		if (opr == pr)
1680 			continue;
1681 
1682 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1683 			continue;
1684 
1685 		if (opr->ndpr_plen == pr->ndpr_plen &&
1686 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1687 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1688 			return (0);
1689 	}
1690 
1691 	/*
1692 	 * We prefer link-local addresses as the associated interface address.
1693 	 */
1694 	/* search for a link-local addr */
1695 	bound = curlwp_bind();
1696 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
1697 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST, &psref);
1698 	if (ifa == NULL) {
1699 		int s = pserialize_read_enter();
1700 		IFADDR_READER_FOREACH(ifa, ifp) {
1701 			if (ifa->ifa_addr->sa_family == AF_INET6)
1702 				break;
1703 		}
1704 		if (ifa != NULL)
1705 			ifa_acquire(ifa, &psref);
1706 		pserialize_read_exit(s);
1707 		/* should we care about ia6_flags? */
1708 	}
1709 	if (ifa == NULL) {
1710 		/*
1711 		 * This can still happen, when, for example, we receive an RA
1712 		 * containing a prefix with the L bit set and the A bit clear,
1713 		 * after removing all IPv6 addresses on the receiving
1714 		 * interface.  This should, of course, be rare though.
1715 		 */
1716 		nd6log(LOG_NOTICE, "failed to find any ifaddr"
1717 		    " to add route for a prefix(%s/%d) on %s\n",
1718 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1719 		    pr->ndpr_plen, if_name(ifp));
1720 		curlwp_bindx(bound);
1721 		return (0);
1722 	}
1723 
1724 	/*
1725 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1726 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1727 	 */
1728 	memset(&mask6, 0, sizeof(mask6));
1729 	mask6.sin6_family = AF_INET6;
1730 	mask6.sin6_len = sizeof(mask6);
1731 	mask6.sin6_addr = pr->ndpr_mask;
1732 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1733 	rtflags = ifa->ifa_flags | RTF_UP;
1734 	if (nd6_need_cache(ifp)) {
1735 		/* explicitly set in case ifa_flags does not set the flag. */
1736 		rtflags |= RTF_CONNECTED;
1737 	} else {
1738 		/*
1739 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1740 		 */
1741 		rtflags &= ~RTF_CONNECTED;
1742 	}
1743 	error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix),
1744 	    ifa->ifa_addr, sin6tosa(&mask6), rtflags);
1745 	if (error == 0) {
1746 		nd6_numroutes++;
1747 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1748 	} else {
1749 		nd6log(LOG_ERR, "failed to add route for a"
1750 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1751 		    "errno = %d\n",
1752 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1753 		    pr->ndpr_plen, if_name(ifp),
1754 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1755 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error);
1756 	}
1757 	ifa_release(ifa, &psref);
1758 	curlwp_bindx(bound);
1759 
1760 	return (error);
1761 }
1762 
1763 int
1764 nd6_prefix_offlink(struct nd_prefix *pr)
1765 {
1766 	int error = 0;
1767 	struct ifnet *ifp = pr->ndpr_ifp;
1768 	struct nd_prefix *opr;
1769 	struct sockaddr_in6 sa6, mask6;
1770 
1771 	/* sanity check */
1772 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1773 		nd6log(LOG_ERR, "%s/%d is already off-link\n",
1774 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
1775 		return (EEXIST);
1776 	}
1777 
1778 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1779 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1780 	error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL,
1781 	    sin6tosa(&mask6), 0);
1782 	if (error == 0) {
1783 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1784 		nd6_numroutes--;
1785 
1786 		/*
1787 		 * There might be the same prefix on another interface,
1788 		 * the prefix which could not be on-link just because we have
1789 		 * the interface route (see comments in nd6_prefix_onlink).
1790 		 * If there's one, try to make the prefix on-link on the
1791 		 * interface.
1792 		 */
1793 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1794 			if (opr == pr)
1795 				continue;
1796 
1797 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1798 				continue;
1799 
1800 			/*
1801 			 * KAME specific: detached prefixes should not be
1802 			 * on-link.
1803 			 */
1804 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1805 				continue;
1806 
1807 			if (opr->ndpr_plen == pr->ndpr_plen &&
1808 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1809 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1810 				int e;
1811 
1812 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1813 					nd6log(LOG_ERR, "failed to "
1814 					    "recover a prefix %s/%d from %s "
1815 					    "to %s (errno = %d)\n",
1816 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1817 					    opr->ndpr_plen, if_name(ifp),
1818 					    if_name(opr->ndpr_ifp), e);
1819 				}
1820 			}
1821 		}
1822 	} else {
1823 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1824 		nd6log(LOG_ERR, "failed to delete route: "
1825 		    "%s/%d on %s (errno = %d)\n",
1826 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1827 		    error);
1828 	}
1829 
1830 	return error;
1831 }
1832 
1833 static struct in6_ifaddr *
1834 in6_ifadd(struct nd_prefixctl *prc, int mcast, struct psref *psref)
1835 {
1836 	struct ifnet *ifp = prc->ndprc_ifp;
1837 	struct ifaddr *ifa;
1838 	struct in6_aliasreq ifra;
1839 	struct in6_ifaddr *ia, *ib;
1840 	int error, plen0;
1841 	struct in6_addr mask;
1842 	int prefixlen = prc->ndprc_plen;
1843 	int updateflags;
1844 	int s;
1845 
1846 	in6_prefixlen2mask(&mask, prefixlen);
1847 
1848 	/*
1849 	 * find a link-local address (will be interface ID).
1850 	 * Is it really mandatory? Theoretically, a global or a site-local
1851 	 * address can be configured without a link-local address, if we
1852 	 * have a unique interface identifier...
1853 	 *
1854 	 * it is not mandatory to have a link-local address, we can generate
1855 	 * interface identifier on the fly.  we do this because:
1856 	 * (1) it should be the easiest way to find interface identifier.
1857 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1858 	 * for multiple addresses on a single interface, and possible shortcut
1859 	 * of DAD.  we omitted DAD for this reason in the past.
1860 	 * (3) a user can prevent autoconfiguration of global address
1861 	 * by removing link-local address by hand (this is partly because we
1862 	 * don't have other way to control the use of IPv6 on an interface.
1863 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1864 	 * (4) it is easier to manage when an interface has addresses
1865 	 * with the same interface identifier, than to have multiple addresses
1866 	 * with different interface identifiers.
1867 	 */
1868 	s = pserialize_read_enter();
1869 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1870 	if (ifa)
1871 		ib = (struct in6_ifaddr *)ifa;
1872 	else {
1873 		pserialize_read_exit(s);
1874 		return NULL;
1875 	}
1876 
1877 #if 0 /* don't care link local addr state, and always do DAD */
1878 	/* if link-local address is not eligible, do not autoconfigure. */
1879 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1880 		printf("in6_ifadd: link-local address not ready\n");
1881 		return NULL;
1882 	}
1883 #endif
1884 
1885 	/* prefixlen + ifidlen must be equal to 128 */
1886 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1887 	if (prefixlen != plen0) {
1888 		nd6log(LOG_INFO, "wrong prefixlen for %s "
1889 		    "(prefix=%d ifid=%d)\n",
1890 		    if_name(ifp), prefixlen, 128 - plen0);
1891 		pserialize_read_exit(s);
1892 		return NULL;
1893 	}
1894 
1895 	/* make ifaddr */
1896 
1897 	memset(&ifra, 0, sizeof(ifra));
1898 	/*
1899 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1900 	 * for safety.
1901 	 */
1902 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1903 	sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
1904 	/* prefix */
1905 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1906 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1907 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1908 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1909 
1910 	/* interface ID */
1911 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1912 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1913 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1914 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1915 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1916 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1917 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1918 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1919 	pserialize_read_exit(s);
1920 
1921 	/* new prefix mask. */
1922 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1923 
1924 	/* lifetimes */
1925 	ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
1926 	ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
1927 
1928 	/* XXX: scope zone ID? */
1929 
1930 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1931 
1932 	/*
1933 	 * Make sure that we do not have this address already.  This should
1934 	 * usually not happen, but we can still see this case, e.g., if we
1935 	 * have manually configured the exact address to be configured.
1936 	 */
1937 	s = pserialize_read_enter();
1938 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1939 		/* this should be rare enough to make an explicit log */
1940 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1941 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1942 		pserialize_read_exit(s);
1943 		return (NULL);
1944 	}
1945 	pserialize_read_exit(s);
1946 
1947 	/*
1948 	 * Allocate ifaddr structure, link into chain, etc.
1949 	 * If we are going to create a new address upon receiving a multicasted
1950 	 * RA, we need to impose a random delay before starting DAD.
1951 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1952 	 */
1953 	updateflags = 0;
1954 	if (mcast)
1955 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1956 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1957 		nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n",
1958 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1959 		    error);
1960 		return (NULL);	/* ifaddr must not have been allocated. */
1961 	}
1962 
1963 	ia = in6ifa_ifpwithaddr_psref(ifp, &ifra.ifra_addr.sin6_addr, psref);
1964 
1965 	return (ia);		/* this is always non-NULL */
1966 }
1967 
1968 int
1969 in6_tmpifadd(
1970 	const struct in6_ifaddr *ia0, /* corresponding public address */
1971 	int forcegen,
1972 	int dad_delay)
1973 {
1974 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1975 	struct in6_ifaddr *newia, *ia;
1976 	struct in6_aliasreq ifra;
1977 	int i, error;
1978 	int trylimit = 3;	/* XXX: adhoc value */
1979 	int updateflags;
1980 	u_int32_t randid[2];
1981 	u_int32_t vltime0, pltime0;
1982 	int s;
1983 
1984 	memset(&ifra, 0, sizeof(ifra));
1985 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1986 	ifra.ifra_addr = ia0->ia_addr;
1987 	/* copy prefix mask */
1988 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1989 	/* clear the old IFID */
1990 	for (i = 0; i < 4; i++) {
1991 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1992 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1993 	}
1994 
1995   again:
1996 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1997 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1998 		nd6log(LOG_NOTICE, "failed to find a good random IFID\n");
1999 		return (EINVAL);
2000 	}
2001 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2002 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2003 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2004 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2005 
2006 	/*
2007 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
2008 	 * However, we may still have a chance to see collision, because
2009 	 * there may be a time lag between generation of the ID and generation
2010 	 * of the address.  So, we'll do one more sanity check.
2011 	 */
2012 	s = pserialize_read_enter();
2013 	IN6_ADDRLIST_READER_FOREACH(ia) {
2014 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2015 		    &ifra.ifra_addr.sin6_addr)) {
2016 			pserialize_read_exit(s);
2017 			if (trylimit-- == 0) {
2018 				/*
2019 				 * Give up.  Something strange should have
2020 				 * happened.
2021 				 */
2022 				nd6log(LOG_NOTICE,
2023 				    "failed to find a unique random IFID\n");
2024 				return (EEXIST);
2025 			}
2026 			forcegen = 1;
2027 			goto again;
2028 		}
2029 	}
2030 	pserialize_read_exit(s);
2031 
2032 	/*
2033 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
2034          * public address or TEMP_VALID_LIFETIME.
2035 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
2036          * of the public address or TEMP_PREFERRED_LIFETIME -
2037          * DESYNC_FACTOR.
2038 	 */
2039 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2040 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2041 		    (ia0->ia6_lifetime.ia6t_vltime -
2042 		    (time_uptime - ia0->ia6_updatetime));
2043 		if (vltime0 > ip6_temp_valid_lifetime)
2044 			vltime0 = ip6_temp_valid_lifetime;
2045 	} else
2046 		vltime0 = ip6_temp_valid_lifetime;
2047 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2048 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2049 		    (ia0->ia6_lifetime.ia6t_pltime -
2050 		    (time_uptime - ia0->ia6_updatetime));
2051 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
2052 			pltime0 = ip6_temp_preferred_lifetime -
2053 			    ip6_desync_factor;
2054 		}
2055 	} else
2056 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
2057 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2058 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2059 
2060 	/*
2061 	 * A temporary address is created only if this calculated Preferred
2062 	 * Lifetime is greater than REGEN_ADVANCE time units.
2063 	 */
2064 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
2065 		return (0);
2066 
2067 	/* XXX: scope zone ID? */
2068 
2069 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2070 
2071 	/* allocate ifaddr structure, link into chain, etc. */
2072 	updateflags = 0;
2073 	if (dad_delay)
2074 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2075 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2076 		return (error);
2077 
2078 	s = pserialize_read_enter();
2079 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2080 	if (newia == NULL) {	/* XXX: can it happen? */
2081 		pserialize_read_exit(s);
2082 		nd6log(LOG_ERR,
2083 		    "ifa update succeeded, but we got no ifaddr\n");
2084 		return (EINVAL); /* XXX */
2085 	}
2086 	newia->ia6_ndpr = ia0->ia6_ndpr;
2087 	newia->ia6_ndpr->ndpr_refcnt++;
2088 	pserialize_read_exit(s);
2089 
2090 	/*
2091 	 * A newly added address might affect the status of other addresses.
2092 	 * XXX: when the temporary address is generated with a new public
2093 	 * address, the onlink check is redundant.  However, it would be safe
2094 	 * to do the check explicitly everywhere a new address is generated,
2095 	 * and, in fact, we surely need the check when we create a new
2096 	 * temporary address due to deprecation of an old temporary address.
2097 	 */
2098 	pfxlist_onlink_check();
2099 
2100 	return (0);
2101 }
2102 
2103 static int
2104 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2105 {
2106 
2107 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
2108 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
2109 		nd6log(LOG_INFO, "preferred lifetime"
2110 		    "(%d) is greater than valid lifetime(%d)\n",
2111 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime);
2112 		return (EINVAL);
2113 	}
2114 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2115 		ndpr->ndpr_preferred = 0;
2116 	else
2117 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2118 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2119 		ndpr->ndpr_expire = 0;
2120 	else
2121 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2122 
2123 	return 0;
2124 }
2125 
2126 static void
2127 in6_init_address_ltimes(struct nd_prefix *newpr,
2128     struct in6_addrlifetime *lt6)
2129 {
2130 
2131 	/* Valid lifetime must not be updated unless explicitly specified. */
2132 	/* init ia6t_expire */
2133 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2134 		lt6->ia6t_expire = 0;
2135 	else {
2136 		lt6->ia6t_expire = time_uptime;
2137 		lt6->ia6t_expire += lt6->ia6t_vltime;
2138 	}
2139 
2140 	/* init ia6t_preferred */
2141 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2142 		lt6->ia6t_preferred = 0;
2143 	else {
2144 		lt6->ia6t_preferred = time_uptime;
2145 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2146 	}
2147 }
2148 
2149 /*
2150  * Delete all the routing table entries that use the specified gateway.
2151  * XXX: this function causes search through all entries of routing table, so
2152  * it shouldn't be called when acting as a router.
2153  */
2154 void
2155 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2156 {
2157 	int s = splsoftnet();
2158 
2159 	/* We'll care only link-local addresses */
2160 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2161 		splx(s);
2162 		return;
2163 	}
2164 
2165 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2166 	splx(s);
2167 }
2168 
2169 static int
2170 rt6_deleteroute(struct rtentry *rt, void *arg)
2171 {
2172 	struct in6_addr *gate = (struct in6_addr *)arg;
2173 
2174 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2175 		return (0);
2176 
2177 	if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
2178 		return (0);
2179 
2180 	/*
2181 	 * Do not delete a static route.
2182 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2183 	 * 'cloned' bit instead?
2184 	 */
2185 	if ((rt->rt_flags & RTF_STATIC) != 0)
2186 		return (0);
2187 
2188 	/*
2189 	 * We delete only host route. This means, in particular, we don't
2190 	 * delete default route.
2191 	 */
2192 	if ((rt->rt_flags & RTF_HOST) == 0)
2193 		return (0);
2194 
2195 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2196 	    rt_mask(rt), rt->rt_flags, NULL));
2197 }
2198 
2199 int
2200 nd6_setdefaultiface(int ifindex)
2201 {
2202 	ifnet_t *ifp;
2203 	int error = 0;
2204 	int s;
2205 
2206 	s = pserialize_read_enter();
2207 	ifp = if_byindex(ifindex);
2208 	if (ifp == NULL) {
2209 		pserialize_read_exit(s);
2210 		return EINVAL;
2211 	}
2212 	if (nd6_defifindex != ifindex) {
2213 		nd6_defifindex = ifindex;
2214 		nd6_defifp = nd6_defifindex > 0 ? ifp : NULL;
2215 
2216 		/*
2217 		 * Our current implementation assumes one-to-one maping between
2218 		 * interfaces and links, so it would be natural to use the
2219 		 * default interface as the default link.
2220 		 */
2221 		scope6_setdefault(nd6_defifp);
2222 	}
2223 	pserialize_read_exit(s);
2224 
2225 	return (error);
2226 }
2227