nd6_rtr.c revision 228966
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/netinet6/nd6_rtr.c 228966 2011-12-29 18:25:18Z jhb $");
34
35#include "opt_inet.h"
36#include "opt_inet6.h"
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/socket.h>
43#include <sys/sockio.h>
44#include <sys/time.h>
45#include <sys/kernel.h>
46#include <sys/lock.h>
47#include <sys/errno.h>
48#include <sys/rwlock.h>
49#include <sys/syslog.h>
50#include <sys/queue.h>
51
52#include <net/if.h>
53#include <net/if_types.h>
54#include <net/if_dl.h>
55#include <net/route.h>
56#include <net/radix.h>
57#include <net/vnet.h>
58
59#include <netinet/in.h>
60#include <net/if_llatbl.h>
61#include <netinet6/in6_var.h>
62#include <netinet6/in6_ifattach.h>
63#include <netinet/ip6.h>
64#include <netinet6/ip6_var.h>
65#include <netinet6/nd6.h>
66#include <netinet/icmp6.h>
67#include <netinet6/scope6_var.h>
68
69static int rtpref(struct nd_defrouter *);
70static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
71static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
72    struct mbuf *, int));
73static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
74static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
75	struct nd_defrouter *));
76static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
77static void pfxrtr_del(struct nd_pfxrouter *);
78static struct nd_pfxrouter *find_pfxlist_reachable_router
79(struct nd_prefix *);
80static void defrouter_delreq(struct nd_defrouter *);
81static void nd6_rtmsg(int, struct rtentry *);
82
83static int in6_init_prefix_ltimes(struct nd_prefix *);
84static void in6_init_address_ltimes __P((struct nd_prefix *,
85	struct in6_addrlifetime *));
86
87static int rt6_deleteroute(struct radix_node *, void *);
88
89VNET_DECLARE(int, nd6_recalc_reachtm_interval);
90#define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
91
92static VNET_DEFINE(struct ifnet *, nd6_defifp);
93VNET_DEFINE(int, nd6_defifindex);
94#define	V_nd6_defifp			VNET(nd6_defifp)
95
96VNET_DEFINE(int, ip6_use_tempaddr) = 0;
97
98VNET_DEFINE(int, ip6_desync_factor);
99VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
100VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
101
102VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
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/*
112 * Receive Router Solicitation Message - just for routers.
113 * Router solicitation/advertisement is mostly managed by userland program
114 * (rtadvd) so here we have no function like nd6_ra_output().
115 *
116 * Based on RFC 2461
117 */
118void
119nd6_rs_input(struct mbuf *m, int off, int icmp6len)
120{
121	struct ifnet *ifp = m->m_pkthdr.rcvif;
122	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123	struct nd_router_solicit *nd_rs;
124	struct in6_addr saddr6 = ip6->ip6_src;
125	char *lladdr = NULL;
126	int lladdrlen = 0;
127	union nd_opts ndopts;
128	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
129
130	/*
131	 * Accept RS only when V_ip6_forwarding=1 and the interface has
132	 * no ND6_IFF_ACCEPT_RTADV.
133	 */
134	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
135		goto freeit;
136
137	/* Sanity checks */
138	if (ip6->ip6_hlim != 255) {
139		nd6log((LOG_ERR,
140		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
141		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
142		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
143		goto bad;
144	}
145
146	/*
147	 * Don't update the neighbor cache, if src = ::.
148	 * This indicates that the src has no IP address assigned yet.
149	 */
150	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
151		goto freeit;
152
153#ifndef PULLDOWN_TEST
154	IP6_EXTHDR_CHECK(m, off, icmp6len,);
155	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
156#else
157	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
158	if (nd_rs == NULL) {
159		ICMP6STAT_INC(icp6s_tooshort);
160		return;
161	}
162#endif
163
164	icmp6len -= sizeof(*nd_rs);
165	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
166	if (nd6_options(&ndopts) < 0) {
167		nd6log((LOG_INFO,
168		    "nd6_rs_input: invalid ND option, ignored\n"));
169		/* nd6_options have incremented stats */
170		goto freeit;
171	}
172
173	if (ndopts.nd_opts_src_lladdr) {
174		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
175		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
176	}
177
178	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
179		nd6log((LOG_INFO,
180		    "nd6_rs_input: lladdrlen mismatch for %s "
181		    "(if %d, RS packet %d)\n",
182		    ip6_sprintf(ip6bufs, &saddr6),
183		    ifp->if_addrlen, lladdrlen - 2));
184		goto bad;
185	}
186
187	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
188
189 freeit:
190	m_freem(m);
191	return;
192
193 bad:
194	ICMP6STAT_INC(icp6s_badrs);
195	m_freem(m);
196}
197
198/*
199 * Receive Router Advertisement Message.
200 *
201 * Based on RFC 2461
202 * TODO: on-link bit on prefix information
203 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
204 */
205void
206nd6_ra_input(struct mbuf *m, int off, int icmp6len)
207{
208	struct ifnet *ifp = m->m_pkthdr.rcvif;
209	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
210	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
211	struct nd_router_advert *nd_ra;
212	struct in6_addr saddr6 = ip6->ip6_src;
213	int mcast = 0;
214	union nd_opts ndopts;
215	struct nd_defrouter *dr;
216	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
217
218	/*
219	 * We only accept RAs only when the per-interface flag
220	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
221	 */
222	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
223		goto freeit;
224
225	if (ip6->ip6_hlim != 255) {
226		nd6log((LOG_ERR,
227		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
228		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
229		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
230		goto bad;
231	}
232
233	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
234		nd6log((LOG_ERR,
235		    "nd6_ra_input: src %s is not link-local\n",
236		    ip6_sprintf(ip6bufs, &saddr6)));
237		goto bad;
238	}
239
240#ifndef PULLDOWN_TEST
241	IP6_EXTHDR_CHECK(m, off, icmp6len,);
242	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
243#else
244	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
245	if (nd_ra == NULL) {
246		ICMP6STAT_INC(icp6s_tooshort);
247		return;
248	}
249#endif
250
251	icmp6len -= sizeof(*nd_ra);
252	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
253	if (nd6_options(&ndopts) < 0) {
254		nd6log((LOG_INFO,
255		    "nd6_ra_input: invalid ND option, ignored\n"));
256		/* nd6_options have incremented stats */
257		goto freeit;
258	}
259
260    {
261	struct nd_defrouter dr0;
262	u_int32_t advreachable = nd_ra->nd_ra_reachable;
263
264	/* remember if this is a multicasted advertisement */
265	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
266		mcast = 1;
267
268	bzero(&dr0, sizeof(dr0));
269	dr0.rtaddr = saddr6;
270	dr0.flags  = nd_ra->nd_ra_flags_reserved;
271	/*
272	 * Effectively-disable routes from RA messages when
273	 * ND6_IFF_NO_RADR enabled on the receiving interface or
274	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
275	 */
276	if (ndi->flags & ND6_IFF_NO_RADR)
277		dr0.rtlifetime = 0;
278	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
279		dr0.rtlifetime = 0;
280	else
281		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
282	dr0.expire = time_second + dr0.rtlifetime;
283	dr0.ifp = ifp;
284	/* unspecified or not? (RFC 2461 6.3.4) */
285	if (advreachable) {
286		advreachable = ntohl(advreachable);
287		if (advreachable <= MAX_REACHABLE_TIME &&
288		    ndi->basereachable != advreachable) {
289			ndi->basereachable = advreachable;
290			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
291			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
292		}
293	}
294	if (nd_ra->nd_ra_retransmit)
295		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
296	if (nd_ra->nd_ra_curhoplimit)
297		ndi->chlim = nd_ra->nd_ra_curhoplimit;
298	dr = defrtrlist_update(&dr0);
299    }
300
301	/*
302	 * prefix
303	 */
304	if (ndopts.nd_opts_pi) {
305		struct nd_opt_hdr *pt;
306		struct nd_opt_prefix_info *pi = NULL;
307		struct nd_prefixctl pr;
308
309		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
310		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
311		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
312						(pt->nd_opt_len << 3))) {
313			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
314				continue;
315			pi = (struct nd_opt_prefix_info *)pt;
316
317			if (pi->nd_opt_pi_len != 4) {
318				nd6log((LOG_INFO,
319				    "nd6_ra_input: invalid option "
320				    "len %d for prefix information option, "
321				    "ignored\n", pi->nd_opt_pi_len));
322				continue;
323			}
324
325			if (128 < pi->nd_opt_pi_prefix_len) {
326				nd6log((LOG_INFO,
327				    "nd6_ra_input: invalid prefix "
328				    "len %d for prefix information option, "
329				    "ignored\n", pi->nd_opt_pi_prefix_len));
330				continue;
331			}
332
333			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
334			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
335				nd6log((LOG_INFO,
336				    "nd6_ra_input: invalid prefix "
337				    "%s, ignored\n",
338				    ip6_sprintf(ip6bufs,
339					&pi->nd_opt_pi_prefix)));
340				continue;
341			}
342
343			bzero(&pr, sizeof(pr));
344			pr.ndpr_prefix.sin6_family = AF_INET6;
345			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
346			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
347			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
348
349			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
350			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
351			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
352			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
353			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
354			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
355			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
356			(void)prelist_update(&pr, dr, m, mcast);
357		}
358	}
359
360	/*
361	 * MTU
362	 */
363	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
364		u_long mtu;
365		u_long maxmtu;
366
367		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
368
369		/* lower bound */
370		if (mtu < IPV6_MMTU) {
371			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
372			    "mtu=%lu sent from %s, ignoring\n",
373			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
374			goto skip;
375		}
376
377		/* upper bound */
378		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
379		    ? ndi->maxmtu : ifp->if_mtu;
380		if (mtu <= maxmtu) {
381			int change = (ndi->linkmtu != mtu);
382
383			ndi->linkmtu = mtu;
384			if (change) /* in6_maxmtu may change */
385				in6_setmaxmtu();
386		} else {
387			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
388			    "mtu=%lu sent from %s; "
389			    "exceeds maxmtu %lu, ignoring\n",
390			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
391		}
392	}
393
394 skip:
395
396	/*
397	 * Source link layer address
398	 */
399    {
400	char *lladdr = NULL;
401	int lladdrlen = 0;
402
403	if (ndopts.nd_opts_src_lladdr) {
404		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
405		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
406	}
407
408	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
409		nd6log((LOG_INFO,
410		    "nd6_ra_input: lladdrlen mismatch for %s "
411		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
412		    ifp->if_addrlen, lladdrlen - 2));
413		goto bad;
414	}
415
416	nd6_cache_lladdr(ifp, &saddr6, lladdr,
417	    lladdrlen, ND_ROUTER_ADVERT, 0);
418
419	/*
420	 * Installing a link-layer address might change the state of the
421	 * router's neighbor cache, which might also affect our on-link
422	 * detection of adveritsed prefixes.
423	 */
424	pfxlist_onlink_check();
425    }
426
427 freeit:
428	m_freem(m);
429	return;
430
431 bad:
432	ICMP6STAT_INC(icp6s_badra);
433	m_freem(m);
434}
435
436/*
437 * default router list proccessing sub routines
438 */
439
440/* tell the change to user processes watching the routing socket. */
441static void
442nd6_rtmsg(int cmd, struct rtentry *rt)
443{
444	struct rt_addrinfo info;
445	struct ifnet *ifp;
446	struct ifaddr *ifa;
447
448	bzero((caddr_t)&info, sizeof(info));
449	info.rti_info[RTAX_DST] = rt_key(rt);
450	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
451	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
452	ifp = rt->rt_ifp;
453	if (ifp != NULL) {
454		IF_ADDR_LOCK(ifp);
455		ifa = TAILQ_FIRST(&ifp->if_addrhead);
456		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
457		ifa_ref(ifa);
458		IF_ADDR_UNLOCK(ifp);
459		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
460	} else
461		ifa = NULL;
462
463	rt_missmsg(cmd, &info, rt->rt_flags, 0);
464	if (ifa != NULL)
465		ifa_free(ifa);
466}
467
468void
469defrouter_addreq(struct nd_defrouter *new)
470{
471	struct sockaddr_in6 def, mask, gate;
472	struct rtentry *newrt = NULL;
473	int s;
474	int error;
475
476	bzero(&def, sizeof(def));
477	bzero(&mask, sizeof(mask));
478	bzero(&gate, sizeof(gate));
479
480	def.sin6_len = mask.sin6_len = gate.sin6_len =
481	    sizeof(struct sockaddr_in6);
482	def.sin6_family = gate.sin6_family = AF_INET6;
483	gate.sin6_addr = new->rtaddr;
484
485	s = splnet();
486	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
487	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
488	    RTF_GATEWAY, &newrt);
489	if (newrt) {
490		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
491		RTFREE(newrt);
492	}
493	if (error == 0)
494		new->installed = 1;
495	splx(s);
496	return;
497}
498
499struct nd_defrouter *
500defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
501{
502	struct nd_defrouter *dr;
503
504	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
505		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
506			return (dr);
507	}
508
509	return (NULL);		/* search failed */
510}
511
512/*
513 * Remove the default route for a given router.
514 * This is just a subroutine function for defrouter_select(), and should
515 * not be called from anywhere else.
516 */
517static void
518defrouter_delreq(struct nd_defrouter *dr)
519{
520	struct sockaddr_in6 def, mask, gate;
521	struct rtentry *oldrt = NULL;
522
523	bzero(&def, sizeof(def));
524	bzero(&mask, sizeof(mask));
525	bzero(&gate, sizeof(gate));
526
527	def.sin6_len = mask.sin6_len = gate.sin6_len =
528	    sizeof(struct sockaddr_in6);
529	def.sin6_family = gate.sin6_family = AF_INET6;
530	gate.sin6_addr = dr->rtaddr;
531
532	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
533	    (struct sockaddr *)&gate,
534	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
535	if (oldrt) {
536		nd6_rtmsg(RTM_DELETE, oldrt);
537		RTFREE(oldrt);
538	}
539
540	dr->installed = 0;
541}
542
543/*
544 * remove all default routes from default router list
545 */
546void
547defrouter_reset(void)
548{
549	struct nd_defrouter *dr;
550
551	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
552		defrouter_delreq(dr);
553
554	/*
555	 * XXX should we also nuke any default routers in the kernel, by
556	 * going through them by rtalloc1()?
557	 */
558}
559
560void
561defrtrlist_del(struct nd_defrouter *dr)
562{
563	struct nd_defrouter *deldr = NULL;
564	struct nd_prefix *pr;
565
566	/*
567	 * Flush all the routing table entries that use the router
568	 * as a next hop.
569	 */
570	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
571		rt6_flush(&dr->rtaddr, dr->ifp);
572
573	if (dr->installed) {
574		deldr = dr;
575		defrouter_delreq(dr);
576	}
577	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
578
579	/*
580	 * Also delete all the pointers to the router in each prefix lists.
581	 */
582	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
583		struct nd_pfxrouter *pfxrtr;
584		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
585			pfxrtr_del(pfxrtr);
586	}
587	pfxlist_onlink_check();
588
589	/*
590	 * If the router is the primary one, choose a new one.
591	 * Note that defrouter_select() will remove the current gateway
592	 * from the routing table.
593	 */
594	if (deldr)
595		defrouter_select();
596
597	free(dr, M_IP6NDP);
598}
599
600/*
601 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
602 * draft-ietf-ipngwg-router-selection:
603 * 1) Routers that are reachable or probably reachable should be preferred.
604 *    If we have more than one (probably) reachable router, prefer ones
605 *    with the highest router preference.
606 * 2) When no routers on the list are known to be reachable or
607 *    probably reachable, routers SHOULD be selected in a round-robin
608 *    fashion, regardless of router preference values.
609 * 3) If the Default Router List is empty, assume that all
610 *    destinations are on-link.
611 *
612 * We assume nd_defrouter is sorted by router preference value.
613 * Since the code below covers both with and without router preference cases,
614 * we do not need to classify the cases by ifdef.
615 *
616 * At this moment, we do not try to install more than one default router,
617 * even when the multipath routing is available, because we're not sure about
618 * the benefits for stub hosts comparing to the risk of making the code
619 * complicated and the possibility of introducing bugs.
620 */
621void
622defrouter_select(void)
623{
624	int s = splnet();
625	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
626	struct llentry *ln = NULL;
627
628	/*
629	 * Let's handle easy case (3) first:
630	 * If default router list is empty, there's nothing to be done.
631	 */
632	if (TAILQ_EMPTY(&V_nd_defrouter)) {
633		splx(s);
634		return;
635	}
636
637	/*
638	 * Search for a (probably) reachable router from the list.
639	 * We just pick up the first reachable one (if any), assuming that
640	 * the ordering rule of the list described in defrtrlist_update().
641	 */
642	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
643		IF_AFDATA_LOCK(dr->ifp);
644		if (selected_dr == NULL &&
645		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
646		    ND6_IS_LLINFO_PROBREACH(ln)) {
647			selected_dr = dr;
648		}
649		IF_AFDATA_UNLOCK(dr->ifp);
650		if (ln != NULL) {
651			LLE_RUNLOCK(ln);
652			ln = NULL;
653		}
654
655		if (dr->installed && installed_dr == NULL)
656			installed_dr = dr;
657		else if (dr->installed && installed_dr) {
658			/* this should not happen.  warn for diagnosis. */
659			log(LOG_ERR, "defrouter_select: more than one router"
660			    " is installed\n");
661		}
662	}
663	/*
664	 * If none of the default routers was found to be reachable,
665	 * round-robin the list regardless of preference.
666	 * Otherwise, if we have an installed router, check if the selected
667	 * (reachable) router should really be preferred to the installed one.
668	 * We only prefer the new router when the old one is not reachable
669	 * or when the new one has a really higher preference value.
670	 */
671	if (selected_dr == NULL) {
672		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
673			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
674		else
675			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
676	} else if (installed_dr) {
677		IF_AFDATA_LOCK(installed_dr->ifp);
678		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
679		    ND6_IS_LLINFO_PROBREACH(ln) &&
680		    rtpref(selected_dr) <= rtpref(installed_dr)) {
681			selected_dr = installed_dr;
682		}
683		IF_AFDATA_UNLOCK(installed_dr->ifp);
684		if (ln != NULL)
685			LLE_RUNLOCK(ln);
686	}
687
688	/*
689	 * If the selected router is different than the installed one,
690	 * remove the installed router and install the selected one.
691	 * Note that the selected router is never NULL here.
692	 */
693	if (installed_dr != selected_dr) {
694		if (installed_dr)
695			defrouter_delreq(installed_dr);
696		defrouter_addreq(selected_dr);
697	}
698
699	splx(s);
700	return;
701}
702
703/*
704 * for default router selection
705 * regards router-preference field as a 2-bit signed integer
706 */
707static int
708rtpref(struct nd_defrouter *dr)
709{
710	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
711	case ND_RA_FLAG_RTPREF_HIGH:
712		return (RTPREF_HIGH);
713	case ND_RA_FLAG_RTPREF_MEDIUM:
714	case ND_RA_FLAG_RTPREF_RSV:
715		return (RTPREF_MEDIUM);
716	case ND_RA_FLAG_RTPREF_LOW:
717		return (RTPREF_LOW);
718	default:
719		/*
720		 * This case should never happen.  If it did, it would mean a
721		 * serious bug of kernel internal.  We thus always bark here.
722		 * Or, can we even panic?
723		 */
724		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
725		return (RTPREF_INVALID);
726	}
727	/* NOTREACHED */
728}
729
730static struct nd_defrouter *
731defrtrlist_update(struct nd_defrouter *new)
732{
733	struct nd_defrouter *dr, *n;
734	int s = splnet();
735
736	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
737		/* entry exists */
738		if (new->rtlifetime == 0) {
739			defrtrlist_del(dr);
740			dr = NULL;
741		} else {
742			int oldpref = rtpref(dr);
743
744			/* override */
745			dr->flags = new->flags; /* xxx flag check */
746			dr->rtlifetime = new->rtlifetime;
747			dr->expire = new->expire;
748
749			/*
750			 * If the preference does not change, there's no need
751			 * to sort the entries. Also make sure the selected
752			 * router is still installed in the kernel.
753			 */
754			if (dr->installed && rtpref(new) == oldpref) {
755				splx(s);
756				return (dr);
757			}
758
759			/*
760			 * preferred router may be changed, so relocate
761			 * this router.
762			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
763			 * However, since defrtrlist_del() has many side
764			 * effects, we intentionally do so here.
765			 * defrouter_select() below will handle routing
766			 * changes later.
767			 */
768			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
769			n = dr;
770			goto insert;
771		}
772		splx(s);
773		return (dr);
774	}
775
776	/* entry does not exist */
777	if (new->rtlifetime == 0) {
778		splx(s);
779		return (NULL);
780	}
781
782	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
783	if (n == NULL) {
784		splx(s);
785		return (NULL);
786	}
787	bzero(n, sizeof(*n));
788	*n = *new;
789
790insert:
791	/*
792	 * Insert the new router in the Default Router List;
793	 * The Default Router List should be in the descending order
794	 * of router-preferece.  Routers with the same preference are
795	 * sorted in the arriving time order.
796	 */
797
798	/* insert at the end of the group */
799	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
800		if (rtpref(n) > rtpref(dr))
801			break;
802	}
803	if (dr)
804		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
805	else
806		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
807
808	defrouter_select();
809
810	splx(s);
811
812	return (n);
813}
814
815static struct nd_pfxrouter *
816pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
817{
818	struct nd_pfxrouter *search;
819
820	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
821		if (search->router == dr)
822			break;
823	}
824
825	return (search);
826}
827
828static void
829pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
830{
831	struct nd_pfxrouter *new;
832
833	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
834	if (new == NULL)
835		return;
836	bzero(new, sizeof(*new));
837	new->router = dr;
838
839	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
840
841	pfxlist_onlink_check();
842}
843
844static void
845pfxrtr_del(struct nd_pfxrouter *pfr)
846{
847	LIST_REMOVE(pfr, pfr_entry);
848	free(pfr, M_IP6NDP);
849}
850
851struct nd_prefix *
852nd6_prefix_lookup(struct nd_prefixctl *key)
853{
854	struct nd_prefix *search;
855
856	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
857		if (key->ndpr_ifp == search->ndpr_ifp &&
858		    key->ndpr_plen == search->ndpr_plen &&
859		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
860		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
861			break;
862		}
863	}
864
865	return (search);
866}
867
868int
869nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
870    struct nd_prefix **newp)
871{
872	struct nd_prefix *new = NULL;
873	int error = 0;
874	int i, s;
875	char ip6buf[INET6_ADDRSTRLEN];
876
877	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
878	if (new == NULL)
879		return(ENOMEM);
880	bzero(new, sizeof(*new));
881	new->ndpr_ifp = pr->ndpr_ifp;
882	new->ndpr_prefix = pr->ndpr_prefix;
883	new->ndpr_plen = pr->ndpr_plen;
884	new->ndpr_vltime = pr->ndpr_vltime;
885	new->ndpr_pltime = pr->ndpr_pltime;
886	new->ndpr_flags = pr->ndpr_flags;
887	if ((error = in6_init_prefix_ltimes(new)) != 0) {
888		free(new, M_IP6NDP);
889		return(error);
890	}
891	new->ndpr_lastupdate = time_second;
892	if (newp != NULL)
893		*newp = new;
894
895	/* initialization */
896	LIST_INIT(&new->ndpr_advrtrs);
897	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
898	/* make prefix in the canonical form */
899	for (i = 0; i < 4; i++)
900		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
901		    new->ndpr_mask.s6_addr32[i];
902
903	s = splnet();
904	/* link ndpr_entry to nd_prefix list */
905	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
906	splx(s);
907
908	/* ND_OPT_PI_FLAG_ONLINK processing */
909	if (new->ndpr_raf_onlink) {
910		int e;
911
912		if ((e = nd6_prefix_onlink(new)) != 0) {
913			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
914			    "the prefix %s/%d on-link on %s (errno=%d)\n",
915			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
916			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
917			/* proceed anyway. XXX: is it correct? */
918		}
919	}
920
921	if (dr)
922		pfxrtr_add(new, dr);
923
924	return 0;
925}
926
927void
928prelist_remove(struct nd_prefix *pr)
929{
930	struct nd_pfxrouter *pfr, *next;
931	int e, s;
932	char ip6buf[INET6_ADDRSTRLEN];
933
934	/* make sure to invalidate the prefix until it is really freed. */
935	pr->ndpr_vltime = 0;
936	pr->ndpr_pltime = 0;
937
938	/*
939	 * Though these flags are now meaningless, we'd rather keep the value
940	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
941	 * when executing "ndp -p".
942	 */
943
944	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
945	    (e = nd6_prefix_offlink(pr)) != 0) {
946		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
947		    "on %s, errno=%d\n",
948		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
949		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
950		/* what should we do? */
951	}
952
953	if (pr->ndpr_refcnt > 0)
954		return;		/* notice here? */
955
956	s = splnet();
957
958	/* unlink ndpr_entry from nd_prefix list */
959	LIST_REMOVE(pr, ndpr_entry);
960
961	/* free list of routers that adversed the prefix */
962	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
963		free(pfr, M_IP6NDP);
964	}
965	splx(s);
966
967	free(pr, M_IP6NDP);
968
969	pfxlist_onlink_check();
970}
971
972/*
973 * dr - may be NULL
974 */
975
976static int
977prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
978    struct mbuf *m, int mcast)
979{
980	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
981	struct ifaddr *ifa;
982	struct ifnet *ifp = new->ndpr_ifp;
983	struct nd_prefix *pr;
984	int s = splnet();
985	int error = 0;
986	int newprefix = 0;
987	int auth;
988	struct in6_addrlifetime lt6_tmp;
989	char ip6buf[INET6_ADDRSTRLEN];
990
991	auth = 0;
992	if (m) {
993		/*
994		 * Authenticity for NA consists authentication for
995		 * both IP header and IP datagrams, doesn't it ?
996		 */
997#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
998		auth = ((m->m_flags & M_AUTHIPHDR) &&
999		    (m->m_flags & M_AUTHIPDGM));
1000#endif
1001	}
1002
1003	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1004		/*
1005		 * nd6_prefix_lookup() ensures that pr and new have the same
1006		 * prefix on a same interface.
1007		 */
1008
1009		/*
1010		 * Update prefix information.  Note that the on-link (L) bit
1011		 * and the autonomous (A) bit should NOT be changed from 1
1012		 * to 0.
1013		 */
1014		if (new->ndpr_raf_onlink == 1)
1015			pr->ndpr_raf_onlink = 1;
1016		if (new->ndpr_raf_auto == 1)
1017			pr->ndpr_raf_auto = 1;
1018		if (new->ndpr_raf_onlink) {
1019			pr->ndpr_vltime = new->ndpr_vltime;
1020			pr->ndpr_pltime = new->ndpr_pltime;
1021			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1022			pr->ndpr_lastupdate = time_second;
1023		}
1024
1025		if (new->ndpr_raf_onlink &&
1026		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1027			int e;
1028
1029			if ((e = nd6_prefix_onlink(pr)) != 0) {
1030				nd6log((LOG_ERR,
1031				    "prelist_update: failed to make "
1032				    "the prefix %s/%d on-link on %s "
1033				    "(errno=%d)\n",
1034				    ip6_sprintf(ip6buf,
1035					    &pr->ndpr_prefix.sin6_addr),
1036				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1037				/* proceed anyway. XXX: is it correct? */
1038			}
1039		}
1040
1041		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1042			pfxrtr_add(pr, dr);
1043	} else {
1044		struct nd_prefix *newpr = NULL;
1045
1046		newprefix = 1;
1047
1048		if (new->ndpr_vltime == 0)
1049			goto end;
1050		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1051			goto end;
1052
1053		error = nd6_prelist_add(new, dr, &newpr);
1054		if (error != 0 || newpr == NULL) {
1055			nd6log((LOG_NOTICE, "prelist_update: "
1056			    "nd6_prelist_add failed for %s/%d on %s "
1057			    "errno=%d, returnpr=%p\n",
1058			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1059			    new->ndpr_plen, if_name(new->ndpr_ifp),
1060			    error, newpr));
1061			goto end; /* we should just give up in this case. */
1062		}
1063
1064		/*
1065		 * XXX: from the ND point of view, we can ignore a prefix
1066		 * with the on-link bit being zero.  However, we need a
1067		 * prefix structure for references from autoconfigured
1068		 * addresses.  Thus, we explicitly make sure that the prefix
1069		 * itself expires now.
1070		 */
1071		if (newpr->ndpr_raf_onlink == 0) {
1072			newpr->ndpr_vltime = 0;
1073			newpr->ndpr_pltime = 0;
1074			in6_init_prefix_ltimes(newpr);
1075		}
1076
1077		pr = newpr;
1078	}
1079
1080	/*
1081	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1082	 * Note that pr must be non NULL at this point.
1083	 */
1084
1085	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1086	if (!new->ndpr_raf_auto)
1087		goto end;
1088
1089	/*
1090	 * 5.5.3 (b). the link-local prefix should have been ignored in
1091	 * nd6_ra_input.
1092	 */
1093
1094	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1095	if (new->ndpr_pltime > new->ndpr_vltime) {
1096		error = EINVAL;	/* XXX: won't be used */
1097		goto end;
1098	}
1099
1100	/*
1101	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1102	 * an address configured by stateless autoconfiguration already in the
1103	 * list of addresses associated with the interface, and the Valid
1104	 * Lifetime is not 0, form an address.  We first check if we have
1105	 * a matching prefix.
1106	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1107	 * consider autoconfigured addresses while RFC2462 simply said
1108	 * "address".
1109	 */
1110	IF_ADDR_LOCK(ifp);
1111	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1112		struct in6_ifaddr *ifa6;
1113		u_int32_t remaininglifetime;
1114
1115		if (ifa->ifa_addr->sa_family != AF_INET6)
1116			continue;
1117
1118		ifa6 = (struct in6_ifaddr *)ifa;
1119
1120		/*
1121		 * We only consider autoconfigured addresses as per rfc2462bis.
1122		 */
1123		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1124			continue;
1125
1126		/*
1127		 * Spec is not clear here, but I believe we should concentrate
1128		 * on unicast (i.e. not anycast) addresses.
1129		 * XXX: other ia6_flags? detached or duplicated?
1130		 */
1131		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1132			continue;
1133
1134		/*
1135		 * Ignore the address if it is not associated with a prefix
1136		 * or is associated with a prefix that is different from this
1137		 * one.  (pr is never NULL here)
1138		 */
1139		if (ifa6->ia6_ndpr != pr)
1140			continue;
1141
1142		if (ia6_match == NULL) /* remember the first one */
1143			ia6_match = ifa6;
1144
1145		/*
1146		 * An already autoconfigured address matched.  Now that we
1147		 * are sure there is at least one matched address, we can
1148		 * proceed to 5.5.3. (e): update the lifetimes according to the
1149		 * "two hours" rule and the privacy extension.
1150		 * We apply some clarifications in rfc2462bis:
1151		 * - use remaininglifetime instead of storedlifetime as a
1152		 *   variable name
1153		 * - remove the dead code in the "two-hour" rule
1154		 */
1155#define TWOHOUR		(120*60)
1156		lt6_tmp = ifa6->ia6_lifetime;
1157
1158		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1159			remaininglifetime = ND6_INFINITE_LIFETIME;
1160		else if (time_second - ifa6->ia6_updatetime >
1161			 lt6_tmp.ia6t_vltime) {
1162			/*
1163			 * The case of "invalid" address.  We should usually
1164			 * not see this case.
1165			 */
1166			remaininglifetime = 0;
1167		} else
1168			remaininglifetime = lt6_tmp.ia6t_vltime -
1169			    (time_second - ifa6->ia6_updatetime);
1170
1171		/* when not updating, keep the current stored lifetime. */
1172		lt6_tmp.ia6t_vltime = remaininglifetime;
1173
1174		if (TWOHOUR < new->ndpr_vltime ||
1175		    remaininglifetime < new->ndpr_vltime) {
1176			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1177		} else if (remaininglifetime <= TWOHOUR) {
1178			if (auth) {
1179				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1180			}
1181		} else {
1182			/*
1183			 * new->ndpr_vltime <= TWOHOUR &&
1184			 * TWOHOUR < remaininglifetime
1185			 */
1186			lt6_tmp.ia6t_vltime = TWOHOUR;
1187		}
1188
1189		/* The 2 hour rule is not imposed for preferred lifetime. */
1190		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1191
1192		in6_init_address_ltimes(pr, &lt6_tmp);
1193
1194		/*
1195		 * We need to treat lifetimes for temporary addresses
1196		 * differently, according to
1197		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1198		 * we only update the lifetimes when they are in the maximum
1199		 * intervals.
1200		 */
1201		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1202			u_int32_t maxvltime, maxpltime;
1203
1204			if (V_ip6_temp_valid_lifetime >
1205			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1206			    V_ip6_desync_factor)) {
1207				maxvltime = V_ip6_temp_valid_lifetime -
1208				    (time_second - ifa6->ia6_createtime) -
1209				    V_ip6_desync_factor;
1210			} else
1211				maxvltime = 0;
1212			if (V_ip6_temp_preferred_lifetime >
1213			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1214			    V_ip6_desync_factor)) {
1215				maxpltime = V_ip6_temp_preferred_lifetime -
1216				    (time_second - ifa6->ia6_createtime) -
1217				    V_ip6_desync_factor;
1218			} else
1219				maxpltime = 0;
1220
1221			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1222			    lt6_tmp.ia6t_vltime > maxvltime) {
1223				lt6_tmp.ia6t_vltime = maxvltime;
1224			}
1225			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1226			    lt6_tmp.ia6t_pltime > maxpltime) {
1227				lt6_tmp.ia6t_pltime = maxpltime;
1228			}
1229		}
1230		ifa6->ia6_lifetime = lt6_tmp;
1231		ifa6->ia6_updatetime = time_second;
1232	}
1233	IF_ADDR_UNLOCK(ifp);
1234	if (ia6_match == NULL && new->ndpr_vltime) {
1235		int ifidlen;
1236
1237		/*
1238		 * 5.5.3 (d) (continued)
1239		 * No address matched and the valid lifetime is non-zero.
1240		 * Create a new address.
1241		 */
1242
1243		/*
1244		 * Prefix Length check:
1245		 * If the sum of the prefix length and interface identifier
1246		 * length does not equal 128 bits, the Prefix Information
1247		 * option MUST be ignored.  The length of the interface
1248		 * identifier is defined in a separate link-type specific
1249		 * document.
1250		 */
1251		ifidlen = in6_if2idlen(ifp);
1252		if (ifidlen < 0) {
1253			/* this should not happen, so we always log it. */
1254			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1255			    if_name(ifp));
1256			goto end;
1257		}
1258		if (ifidlen + pr->ndpr_plen != 128) {
1259			nd6log((LOG_INFO,
1260			    "prelist_update: invalid prefixlen "
1261			    "%d for %s, ignored\n",
1262			    pr->ndpr_plen, if_name(ifp)));
1263			goto end;
1264		}
1265
1266		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1267			/*
1268			 * note that we should use pr (not new) for reference.
1269			 */
1270			pr->ndpr_refcnt++;
1271			ia6->ia6_ndpr = pr;
1272
1273			/*
1274			 * RFC 3041 3.3 (2).
1275			 * When a new public address is created as described
1276			 * in RFC2462, also create a new temporary address.
1277			 *
1278			 * RFC 3041 3.5.
1279			 * When an interface connects to a new link, a new
1280			 * randomized interface identifier should be generated
1281			 * immediately together with a new set of temporary
1282			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1283			 * in6_tmpifadd().
1284			 */
1285			if (V_ip6_use_tempaddr) {
1286				int e;
1287				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1288					nd6log((LOG_NOTICE, "prelist_update: "
1289					    "failed to create a temporary "
1290					    "address, errno=%d\n",
1291					    e));
1292				}
1293			}
1294			ifa_free(&ia6->ia_ifa);
1295
1296			/*
1297			 * A newly added address might affect the status
1298			 * of other addresses, so we check and update it.
1299			 * XXX: what if address duplication happens?
1300			 */
1301			pfxlist_onlink_check();
1302		} else {
1303			/* just set an error. do not bark here. */
1304			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1305		}
1306	}
1307
1308 end:
1309	splx(s);
1310	return error;
1311}
1312
1313/*
1314 * A supplement function used in the on-link detection below;
1315 * detect if a given prefix has a (probably) reachable advertising router.
1316 * XXX: lengthy function name...
1317 */
1318static struct nd_pfxrouter *
1319find_pfxlist_reachable_router(struct nd_prefix *pr)
1320{
1321	struct nd_pfxrouter *pfxrtr;
1322	struct llentry *ln;
1323	int canreach;
1324
1325	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1326		IF_AFDATA_LOCK(pfxrtr->router->ifp);
1327		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1328		IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1329		if (ln == NULL)
1330			continue;
1331		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1332		LLE_RUNLOCK(ln);
1333		if (canreach)
1334			break;
1335	}
1336	return (pfxrtr);
1337}
1338
1339/*
1340 * Check if each prefix in the prefix list has at least one available router
1341 * that advertised the prefix (a router is "available" if its neighbor cache
1342 * entry is reachable or probably reachable).
1343 * If the check fails, the prefix may be off-link, because, for example,
1344 * we have moved from the network but the lifetime of the prefix has not
1345 * expired yet.  So we should not use the prefix if there is another prefix
1346 * that has an available router.
1347 * But, if there is no prefix that has an available router, we still regards
1348 * all the prefixes as on-link.  This is because we can't tell if all the
1349 * routers are simply dead or if we really moved from the network and there
1350 * is no router around us.
1351 */
1352void
1353pfxlist_onlink_check()
1354{
1355	struct nd_prefix *pr;
1356	struct in6_ifaddr *ifa;
1357	struct nd_defrouter *dr;
1358	struct nd_pfxrouter *pfxrtr = NULL;
1359
1360	/*
1361	 * Check if there is a prefix that has a reachable advertising
1362	 * router.
1363	 */
1364	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1365		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1366			break;
1367	}
1368
1369	/*
1370	 * If we have no such prefix, check whether we still have a router
1371	 * that does not advertise any prefixes.
1372	 */
1373	if (pr == NULL) {
1374		TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1375			struct nd_prefix *pr0;
1376
1377			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1378				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1379					break;
1380			}
1381			if (pfxrtr != NULL)
1382				break;
1383		}
1384	}
1385	if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1386		/*
1387		 * There is at least one prefix that has a reachable router,
1388		 * or at least a router which probably does not advertise
1389		 * any prefixes.  The latter would be the case when we move
1390		 * to a new link where we have a router that does not provide
1391		 * prefixes and we configure an address by hand.
1392		 * Detach prefixes which have no reachable advertising
1393		 * router, and attach other prefixes.
1394		 */
1395		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1396			/* XXX: a link-local prefix should never be detached */
1397			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1398				continue;
1399
1400			/*
1401			 * we aren't interested in prefixes without the L bit
1402			 * set.
1403			 */
1404			if (pr->ndpr_raf_onlink == 0)
1405				continue;
1406
1407			if (pr->ndpr_raf_auto == 0)
1408				continue;
1409
1410			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1411			    find_pfxlist_reachable_router(pr) == NULL)
1412				pr->ndpr_stateflags |= NDPRF_DETACHED;
1413			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1414			    find_pfxlist_reachable_router(pr) != 0)
1415				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1416		}
1417	} else {
1418		/* there is no prefix that has a reachable router */
1419		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1420			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1421				continue;
1422
1423			if (pr->ndpr_raf_onlink == 0)
1424				continue;
1425
1426			if (pr->ndpr_raf_auto == 0)
1427				continue;
1428
1429			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1430				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1431		}
1432	}
1433
1434	/*
1435	 * Remove each interface route associated with a (just) detached
1436	 * prefix, and reinstall the interface route for a (just) attached
1437	 * prefix.  Note that all attempt of reinstallation does not
1438	 * necessarily success, when a same prefix is shared among multiple
1439	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1440	 * so we don't have to care about them.
1441	 */
1442	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1443		int e;
1444		char ip6buf[INET6_ADDRSTRLEN];
1445
1446		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1447			continue;
1448
1449		if (pr->ndpr_raf_onlink == 0)
1450			continue;
1451
1452		if (pr->ndpr_raf_auto == 0)
1453			continue;
1454
1455		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1456		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1457			if ((e = nd6_prefix_offlink(pr)) != 0) {
1458				nd6log((LOG_ERR,
1459				    "pfxlist_onlink_check: failed to "
1460				    "make %s/%d offlink, errno=%d\n",
1461				    ip6_sprintf(ip6buf,
1462					    &pr->ndpr_prefix.sin6_addr),
1463					    pr->ndpr_plen, e));
1464			}
1465		}
1466		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1467		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1468		    pr->ndpr_raf_onlink) {
1469			if ((e = nd6_prefix_onlink(pr)) != 0) {
1470				nd6log((LOG_ERR,
1471				    "pfxlist_onlink_check: failed to "
1472				    "make %s/%d onlink, errno=%d\n",
1473				    ip6_sprintf(ip6buf,
1474					    &pr->ndpr_prefix.sin6_addr),
1475					    pr->ndpr_plen, e));
1476			}
1477		}
1478	}
1479
1480	/*
1481	 * Changes on the prefix status might affect address status as well.
1482	 * Make sure that all addresses derived from an attached prefix are
1483	 * attached, and that all addresses derived from a detached prefix are
1484	 * detached.  Note, however, that a manually configured address should
1485	 * always be attached.
1486	 * The precise detection logic is same as the one for prefixes.
1487	 *
1488	 * XXXRW: in6_ifaddrhead locking.
1489	 */
1490	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1491		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1492			continue;
1493
1494		if (ifa->ia6_ndpr == NULL) {
1495			/*
1496			 * This can happen when we first configure the address
1497			 * (i.e. the address exists, but the prefix does not).
1498			 * XXX: complicated relationships...
1499			 */
1500			continue;
1501		}
1502
1503		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1504			break;
1505	}
1506	if (ifa) {
1507		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1508			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1509				continue;
1510
1511			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1512				continue;
1513
1514			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1515				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1516					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1517					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1518					nd6_dad_start((struct ifaddr *)ifa, 0);
1519				}
1520			} else {
1521				ifa->ia6_flags |= IN6_IFF_DETACHED;
1522			}
1523		}
1524	}
1525	else {
1526		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1527			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1528				continue;
1529
1530			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1531				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1532				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1533				/* Do we need a delay in this case? */
1534				nd6_dad_start((struct ifaddr *)ifa, 0);
1535			}
1536		}
1537	}
1538}
1539
1540int
1541nd6_prefix_onlink(struct nd_prefix *pr)
1542{
1543	struct ifaddr *ifa;
1544	struct ifnet *ifp = pr->ndpr_ifp;
1545	struct sockaddr_in6 mask6;
1546	struct nd_prefix *opr;
1547	u_long rtflags;
1548	int error = 0;
1549	struct radix_node_head *rnh;
1550	struct rtentry *rt = NULL;
1551	char ip6buf[INET6_ADDRSTRLEN];
1552	struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1553
1554	/* sanity check */
1555	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1556		nd6log((LOG_ERR,
1557		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1558		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1559		    pr->ndpr_plen));
1560		return (EEXIST);
1561	}
1562
1563	/*
1564	 * Add the interface route associated with the prefix.  Before
1565	 * installing the route, check if there's the same prefix on another
1566	 * interface, and the prefix has already installed the interface route.
1567	 * Although such a configuration is expected to be rare, we explicitly
1568	 * allow it.
1569	 */
1570	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1571		if (opr == pr)
1572			continue;
1573
1574		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1575			continue;
1576
1577		if (opr->ndpr_plen == pr->ndpr_plen &&
1578		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1579		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1580			return (0);
1581	}
1582
1583	/*
1584	 * We prefer link-local addresses as the associated interface address.
1585	 */
1586	/* search for a link-local addr */
1587	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1588	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1589	if (ifa == NULL) {
1590		/* XXX: freebsd does not have ifa_ifwithaf */
1591		IF_ADDR_LOCK(ifp);
1592		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1593			if (ifa->ifa_addr->sa_family == AF_INET6)
1594				break;
1595		}
1596		if (ifa != NULL)
1597			ifa_ref(ifa);
1598		IF_ADDR_UNLOCK(ifp);
1599		/* should we care about ia6_flags? */
1600	}
1601	if (ifa == NULL) {
1602		/*
1603		 * This can still happen, when, for example, we receive an RA
1604		 * containing a prefix with the L bit set and the A bit clear,
1605		 * after removing all IPv6 addresses on the receiving
1606		 * interface.  This should, of course, be rare though.
1607		 */
1608		nd6log((LOG_NOTICE,
1609		    "nd6_prefix_onlink: failed to find any ifaddr"
1610		    " to add route for a prefix(%s/%d) on %s\n",
1611		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1612		    pr->ndpr_plen, if_name(ifp)));
1613		return (0);
1614	}
1615
1616	/*
1617	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1618	 * ifa->ifa_rtrequest = nd6_rtrequest;
1619	 */
1620	bzero(&mask6, sizeof(mask6));
1621	mask6.sin6_len = sizeof(mask6);
1622	mask6.sin6_addr = pr->ndpr_mask;
1623	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1624	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1625	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1626	if (error == 0) {
1627		if (rt != NULL) /* this should be non NULL, though */ {
1628			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1629			/* XXX what if rhn == NULL? */
1630			RADIX_NODE_HEAD_LOCK(rnh);
1631			RT_LOCK(rt);
1632			if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1633				((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1634					rt->rt_ifp->if_type;
1635				((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1636					rt->rt_ifp->if_index;
1637			}
1638			RADIX_NODE_HEAD_UNLOCK(rnh);
1639			nd6_rtmsg(RTM_ADD, rt);
1640			RT_UNLOCK(rt);
1641		}
1642		pr->ndpr_stateflags |= NDPRF_ONLINK;
1643	} else {
1644		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1645		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1646		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1647		    "errno = %d\n",
1648		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1649		    pr->ndpr_plen, if_name(ifp),
1650		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1651		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1652	}
1653
1654	if (rt != NULL) {
1655		RT_LOCK(rt);
1656		RT_REMREF(rt);
1657		RT_UNLOCK(rt);
1658	}
1659	if (ifa != NULL)
1660		ifa_free(ifa);
1661
1662	return (error);
1663}
1664
1665int
1666nd6_prefix_offlink(struct nd_prefix *pr)
1667{
1668	int error = 0;
1669	struct ifnet *ifp = pr->ndpr_ifp;
1670	struct nd_prefix *opr;
1671	struct sockaddr_in6 sa6, mask6;
1672	struct rtentry *rt = NULL;
1673	char ip6buf[INET6_ADDRSTRLEN];
1674
1675	/* sanity check */
1676	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1677		nd6log((LOG_ERR,
1678		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1679		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1680		    pr->ndpr_plen));
1681		return (EEXIST);
1682	}
1683
1684	bzero(&sa6, sizeof(sa6));
1685	sa6.sin6_family = AF_INET6;
1686	sa6.sin6_len = sizeof(sa6);
1687	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1688	    sizeof(struct in6_addr));
1689	bzero(&mask6, sizeof(mask6));
1690	mask6.sin6_family = AF_INET6;
1691	mask6.sin6_len = sizeof(sa6);
1692	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1693	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1694	    (struct sockaddr *)&mask6, 0, &rt);
1695	if (error == 0) {
1696		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1697
1698		/* report the route deletion to the routing socket. */
1699		if (rt != NULL)
1700			nd6_rtmsg(RTM_DELETE, rt);
1701
1702		/*
1703		 * There might be the same prefix on another interface,
1704		 * the prefix which could not be on-link just because we have
1705		 * the interface route (see comments in nd6_prefix_onlink).
1706		 * If there's one, try to make the prefix on-link on the
1707		 * interface.
1708		 */
1709		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1710			if (opr == pr)
1711				continue;
1712
1713			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1714				continue;
1715
1716			/*
1717			 * KAME specific: detached prefixes should not be
1718			 * on-link.
1719			 */
1720			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1721				continue;
1722
1723			if (opr->ndpr_plen == pr->ndpr_plen &&
1724			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1725			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1726				int e;
1727
1728				if ((e = nd6_prefix_onlink(opr)) != 0) {
1729					nd6log((LOG_ERR,
1730					    "nd6_prefix_offlink: failed to "
1731					    "recover a prefix %s/%d from %s "
1732					    "to %s (errno = %d)\n",
1733					    ip6_sprintf(ip6buf,
1734						&opr->ndpr_prefix.sin6_addr),
1735					    opr->ndpr_plen, if_name(ifp),
1736					    if_name(opr->ndpr_ifp), e));
1737				}
1738			}
1739		}
1740	} else {
1741		/* XXX: can we still set the NDPRF_ONLINK flag? */
1742		nd6log((LOG_ERR,
1743		    "nd6_prefix_offlink: failed to delete route: "
1744		    "%s/%d on %s (errno = %d)\n",
1745		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1746		    if_name(ifp), error));
1747	}
1748
1749	if (rt != NULL) {
1750		RTFREE(rt);
1751	}
1752
1753	return (error);
1754}
1755
1756static struct in6_ifaddr *
1757in6_ifadd(struct nd_prefixctl *pr, int mcast)
1758{
1759	struct ifnet *ifp = pr->ndpr_ifp;
1760	struct ifaddr *ifa;
1761	struct in6_aliasreq ifra;
1762	struct in6_ifaddr *ia, *ib;
1763	int error, plen0;
1764	struct in6_addr mask;
1765	int prefixlen = pr->ndpr_plen;
1766	int updateflags;
1767	char ip6buf[INET6_ADDRSTRLEN];
1768
1769	in6_prefixlen2mask(&mask, prefixlen);
1770
1771	/*
1772	 * find a link-local address (will be interface ID).
1773	 * Is it really mandatory? Theoretically, a global or a site-local
1774	 * address can be configured without a link-local address, if we
1775	 * have a unique interface identifier...
1776	 *
1777	 * it is not mandatory to have a link-local address, we can generate
1778	 * interface identifier on the fly.  we do this because:
1779	 * (1) it should be the easiest way to find interface identifier.
1780	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1781	 * for multiple addresses on a single interface, and possible shortcut
1782	 * of DAD.  we omitted DAD for this reason in the past.
1783	 * (3) a user can prevent autoconfiguration of global address
1784	 * by removing link-local address by hand (this is partly because we
1785	 * don't have other way to control the use of IPv6 on an interface.
1786	 * this has been our design choice - cf. NRL's "ifconfig auto").
1787	 * (4) it is easier to manage when an interface has addresses
1788	 * with the same interface identifier, than to have multiple addresses
1789	 * with different interface identifiers.
1790	 */
1791	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1792	if (ifa)
1793		ib = (struct in6_ifaddr *)ifa;
1794	else
1795		return NULL;
1796
1797	/* prefixlen + ifidlen must be equal to 128 */
1798	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1799	if (prefixlen != plen0) {
1800		ifa_free(ifa);
1801		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1802		    "(prefix=%d ifid=%d)\n",
1803		    if_name(ifp), prefixlen, 128 - plen0));
1804		return NULL;
1805	}
1806
1807	/* make ifaddr */
1808
1809	bzero(&ifra, sizeof(ifra));
1810	/*
1811	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1812	 * for safety.
1813	 */
1814	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1815	ifra.ifra_addr.sin6_family = AF_INET6;
1816	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1817	/* prefix */
1818	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1819	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1820	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1821	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1822	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1823
1824	/* interface ID */
1825	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1826	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1827	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1828	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1829	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1830	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1831	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1832	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1833	ifa_free(ifa);
1834
1835	/* new prefix mask. */
1836	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1837	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1838	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1839	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1840
1841	/* lifetimes. */
1842	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1843	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1844
1845	/* XXX: scope zone ID? */
1846
1847	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1848
1849	/*
1850	 * Make sure that we do not have this address already.  This should
1851	 * usually not happen, but we can still see this case, e.g., if we
1852	 * have manually configured the exact address to be configured.
1853	 */
1854	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1855	    &ifra.ifra_addr.sin6_addr);
1856	if (ifa != NULL) {
1857		ifa_free(ifa);
1858		/* this should be rare enough to make an explicit log */
1859		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1860		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1861		return (NULL);
1862	}
1863
1864	/*
1865	 * Allocate ifaddr structure, link into chain, etc.
1866	 * If we are going to create a new address upon receiving a multicasted
1867	 * RA, we need to impose a random delay before starting DAD.
1868	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1869	 */
1870	updateflags = 0;
1871	if (mcast)
1872		updateflags |= IN6_IFAUPDATE_DADDELAY;
1873	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1874		nd6log((LOG_ERR,
1875		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1876		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1877		    if_name(ifp), error));
1878		return (NULL);	/* ifaddr must not have been allocated. */
1879	}
1880
1881	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1882	/*
1883	 * XXXRW: Assumption of non-NULLness here might not be true with
1884	 * fine-grained locking -- should we validate it?  Or just return
1885	 * earlier ifa rather than looking it up again?
1886	 */
1887	return (ia);		/* this is always non-NULL  and referenced. */
1888}
1889
1890/*
1891 * ia0 - corresponding public address
1892 */
1893int
1894in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1895{
1896	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1897	struct in6_ifaddr *newia, *ia;
1898	struct in6_aliasreq ifra;
1899	int i, error;
1900	int trylimit = 3;	/* XXX: adhoc value */
1901	int updateflags;
1902	u_int32_t randid[2];
1903	time_t vltime0, pltime0;
1904
1905	bzero(&ifra, sizeof(ifra));
1906	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1907	ifra.ifra_addr = ia0->ia_addr;
1908	/* copy prefix mask */
1909	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1910	/* clear the old IFID */
1911	for (i = 0; i < 4; i++) {
1912		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1913		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1914	}
1915
1916  again:
1917	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1918	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1919		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1920		    "random IFID\n"));
1921		return (EINVAL);
1922	}
1923	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1924	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1925	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1926	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1927
1928	/*
1929	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1930	 * However, we may still have a chance to see collision, because
1931	 * there may be a time lag between generation of the ID and generation
1932	 * of the address.  So, we'll do one more sanity check.
1933	 */
1934	IN6_IFADDR_RLOCK();
1935	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1936		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1937		    &ifra.ifra_addr.sin6_addr)) {
1938			if (trylimit-- == 0) {
1939				IN6_IFADDR_RUNLOCK();
1940				/*
1941				 * Give up.  Something strange should have
1942				 * happened.
1943				 */
1944				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1945				    "find a unique random IFID\n"));
1946				return (EEXIST);
1947			}
1948			IN6_IFADDR_RUNLOCK();
1949			forcegen = 1;
1950			goto again;
1951		}
1952	}
1953	IN6_IFADDR_RUNLOCK();
1954
1955	/*
1956	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1957         * public address or TEMP_VALID_LIFETIME.
1958	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1959         * of the public address or TEMP_PREFERRED_LIFETIME -
1960         * DESYNC_FACTOR.
1961	 */
1962	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1963		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1964		    (ia0->ia6_lifetime.ia6t_vltime -
1965		    (time_second - ia0->ia6_updatetime));
1966		if (vltime0 > V_ip6_temp_valid_lifetime)
1967			vltime0 = V_ip6_temp_valid_lifetime;
1968	} else
1969		vltime0 = V_ip6_temp_valid_lifetime;
1970	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1971		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1972		    (ia0->ia6_lifetime.ia6t_pltime -
1973		    (time_second - ia0->ia6_updatetime));
1974		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1975			pltime0 = V_ip6_temp_preferred_lifetime -
1976			    V_ip6_desync_factor;
1977		}
1978	} else
1979		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1980	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1981	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1982
1983	/*
1984	 * A temporary address is created only if this calculated Preferred
1985	 * Lifetime is greater than REGEN_ADVANCE time units.
1986	 */
1987	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1988		return (0);
1989
1990	/* XXX: scope zone ID? */
1991
1992	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1993
1994	/* allocate ifaddr structure, link into chain, etc. */
1995	updateflags = 0;
1996	if (delay)
1997		updateflags |= IN6_IFAUPDATE_DADDELAY;
1998	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1999		return (error);
2000
2001	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2002	if (newia == NULL) {	/* XXX: can it happen? */
2003		nd6log((LOG_ERR,
2004		    "in6_tmpifadd: ifa update succeeded, but we got "
2005		    "no ifaddr\n"));
2006		return (EINVAL); /* XXX */
2007	}
2008	newia->ia6_ndpr = ia0->ia6_ndpr;
2009	newia->ia6_ndpr->ndpr_refcnt++;
2010	ifa_free(&newia->ia_ifa);
2011
2012	/*
2013	 * A newly added address might affect the status of other addresses.
2014	 * XXX: when the temporary address is generated with a new public
2015	 * address, the onlink check is redundant.  However, it would be safe
2016	 * to do the check explicitly everywhere a new address is generated,
2017	 * and, in fact, we surely need the check when we create a new
2018	 * temporary address due to deprecation of an old temporary address.
2019	 */
2020	pfxlist_onlink_check();
2021
2022	return (0);
2023}
2024
2025static int
2026in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2027{
2028	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2029		ndpr->ndpr_preferred = 0;
2030	else
2031		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2032	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2033		ndpr->ndpr_expire = 0;
2034	else
2035		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2036
2037	return 0;
2038}
2039
2040static void
2041in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2042{
2043	/* init ia6t_expire */
2044	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2045		lt6->ia6t_expire = 0;
2046	else {
2047		lt6->ia6t_expire = time_second;
2048		lt6->ia6t_expire += lt6->ia6t_vltime;
2049	}
2050
2051	/* init ia6t_preferred */
2052	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2053		lt6->ia6t_preferred = 0;
2054	else {
2055		lt6->ia6t_preferred = time_second;
2056		lt6->ia6t_preferred += lt6->ia6t_pltime;
2057	}
2058}
2059
2060/*
2061 * Delete all the routing table entries that use the specified gateway.
2062 * XXX: this function causes search through all entries of routing table, so
2063 * it shouldn't be called when acting as a router.
2064 */
2065void
2066rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2067{
2068	struct radix_node_head *rnh;
2069	int s = splnet();
2070
2071	/* We'll care only link-local addresses */
2072	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2073		splx(s);
2074		return;
2075	}
2076
2077	rnh = rt_tables_get_rnh(0, AF_INET6);
2078	if (rnh == NULL)
2079		return;
2080
2081	RADIX_NODE_HEAD_LOCK(rnh);
2082	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2083	RADIX_NODE_HEAD_UNLOCK(rnh);
2084	splx(s);
2085}
2086
2087static int
2088rt6_deleteroute(struct radix_node *rn, void *arg)
2089{
2090#define SIN6(s)	((struct sockaddr_in6 *)s)
2091	struct rtentry *rt = (struct rtentry *)rn;
2092	struct in6_addr *gate = (struct in6_addr *)arg;
2093
2094	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2095		return (0);
2096
2097	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2098		return (0);
2099	}
2100
2101	/*
2102	 * Do not delete a static route.
2103	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2104	 * 'cloned' bit instead?
2105	 */
2106	if ((rt->rt_flags & RTF_STATIC) != 0)
2107		return (0);
2108
2109	/*
2110	 * We delete only host route. This means, in particular, we don't
2111	 * delete default route.
2112	 */
2113	if ((rt->rt_flags & RTF_HOST) == 0)
2114		return (0);
2115
2116	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2117	    rt_mask(rt), rt->rt_flags, 0));
2118#undef SIN6
2119}
2120
2121int
2122nd6_setdefaultiface(int ifindex)
2123{
2124	int error = 0;
2125
2126	if (ifindex < 0 || V_if_index < ifindex)
2127		return (EINVAL);
2128	if (ifindex != 0 && !ifnet_byindex(ifindex))
2129		return (EINVAL);
2130
2131	if (V_nd6_defifindex != ifindex) {
2132		V_nd6_defifindex = ifindex;
2133		if (V_nd6_defifindex > 0)
2134			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2135		else
2136			V_nd6_defifp = NULL;
2137
2138		/*
2139		 * Our current implementation assumes one-to-one maping between
2140		 * interfaces and links, so it would be natural to use the
2141		 * default interface as the default link.
2142		 */
2143		scope6_setdefault(V_nd6_defifp);
2144	}
2145
2146	return (error);
2147}
2148