nd6_rtr.c revision 185348
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 185348 2008-11-26 22:32:07Z zec $");
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/errno.h>
47#include <sys/syslog.h>
48#include <sys/queue.h>
49#include <sys/vimage.h>
50
51#include <net/if.h>
52#include <net/if_types.h>
53#include <net/if_dl.h>
54#include <net/route.h>
55#include <net/radix.h>
56
57#include <netinet/in.h>
58#include <netinet6/in6_var.h>
59#include <netinet6/in6_ifattach.h>
60#include <netinet/ip6.h>
61#include <netinet6/ip6_var.h>
62#include <netinet6/nd6.h>
63#include <netinet/icmp6.h>
64#include <netinet6/scope6_var.h>
65
66#define SDL(s)	((struct sockaddr_dl *)s)
67
68static int rtpref(struct nd_defrouter *);
69static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
70static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
71    struct mbuf *, int));
72static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
73static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
74	struct nd_defrouter *));
75static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
76static void pfxrtr_del(struct nd_pfxrouter *);
77static struct nd_pfxrouter *find_pfxlist_reachable_router
78(struct nd_prefix *);
79static void defrouter_delreq(struct nd_defrouter *);
80static void nd6_rtmsg(int, struct rtentry *);
81
82static int in6_init_prefix_ltimes(struct nd_prefix *);
83static void in6_init_address_ltimes __P((struct nd_prefix *,
84	struct in6_addrlifetime *));
85
86static int rt6_deleteroute(struct radix_node *, void *);
87
88#ifdef VIMAGE_GLOBALS
89extern int nd6_recalc_reachtm_interval;
90
91static struct ifnet *nd6_defifp;
92int nd6_defifindex;
93
94int ip6_use_tempaddr;
95int ip6_desync_factor;
96u_int32_t ip6_temp_preferred_lifetime;
97u_int32_t ip6_temp_valid_lifetime;
98int ip6_temp_regen_advance;
99#endif
100
101/* RTPREF_MEDIUM has to be 0! */
102#define RTPREF_HIGH	1
103#define RTPREF_MEDIUM	0
104#define RTPREF_LOW	(-1)
105#define RTPREF_RESERVED	(-2)
106#define RTPREF_INVALID	(-3)	/* internal */
107
108/*
109 * Receive Router Solicitation Message - just for routers.
110 * Router solicitation/advertisement is mostly managed by userland program
111 * (rtadvd) so here we have no function like nd6_ra_output().
112 *
113 * Based on RFC 2461
114 */
115void
116nd6_rs_input(struct mbuf *m, int off, int icmp6len)
117{
118	INIT_VNET_INET6(curvnet);
119	struct ifnet *ifp = m->m_pkthdr.rcvif;
120	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
121	struct nd_router_solicit *nd_rs;
122	struct in6_addr saddr6 = ip6->ip6_src;
123	char *lladdr = NULL;
124	int lladdrlen = 0;
125	union nd_opts ndopts;
126	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
127
128	/* If I'm not a router, ignore it. */
129	if (V_ip6_accept_rtadv != 0 || V_ip6_forwarding != 1)
130		goto freeit;
131
132	/* Sanity checks */
133	if (ip6->ip6_hlim != 255) {
134		nd6log((LOG_ERR,
135		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
136		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
137		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
138		goto bad;
139	}
140
141	/*
142	 * Don't update the neighbor cache, if src = ::.
143	 * This indicates that the src has no IP address assigned yet.
144	 */
145	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
146		goto freeit;
147
148#ifndef PULLDOWN_TEST
149	IP6_EXTHDR_CHECK(m, off, icmp6len,);
150	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
151#else
152	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
153	if (nd_rs == NULL) {
154		V_icmp6stat.icp6s_tooshort++;
155		return;
156	}
157#endif
158
159	icmp6len -= sizeof(*nd_rs);
160	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
161	if (nd6_options(&ndopts) < 0) {
162		nd6log((LOG_INFO,
163		    "nd6_rs_input: invalid ND option, ignored\n"));
164		/* nd6_options have incremented stats */
165		goto freeit;
166	}
167
168	if (ndopts.nd_opts_src_lladdr) {
169		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
170		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
171	}
172
173	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
174		nd6log((LOG_INFO,
175		    "nd6_rs_input: lladdrlen mismatch for %s "
176		    "(if %d, RS packet %d)\n",
177		    ip6_sprintf(ip6bufs, &saddr6),
178		    ifp->if_addrlen, lladdrlen - 2));
179		goto bad;
180	}
181
182	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
183
184 freeit:
185	m_freem(m);
186	return;
187
188 bad:
189	V_icmp6stat.icp6s_badrs++;
190	m_freem(m);
191}
192
193/*
194 * Receive Router Advertisement Message.
195 *
196 * Based on RFC 2461
197 * TODO: on-link bit on prefix information
198 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
199 */
200void
201nd6_ra_input(struct mbuf *m, int off, int icmp6len)
202{
203	INIT_VNET_INET6(curvnet);
204	struct ifnet *ifp = m->m_pkthdr.rcvif;
205	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
206	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
207	struct nd_router_advert *nd_ra;
208	struct in6_addr saddr6 = ip6->ip6_src;
209	int mcast = 0;
210	union nd_opts ndopts;
211	struct nd_defrouter *dr;
212	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
213
214	/*
215	 * We only accept RAs only when
216	 * the system-wide variable allows the acceptance, and
217	 * per-interface variable allows RAs on the receiving interface.
218	 */
219	if (V_ip6_accept_rtadv == 0)
220		goto freeit;
221	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
222		goto freeit;
223
224	if (ip6->ip6_hlim != 255) {
225		nd6log((LOG_ERR,
226		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
227		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
228		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
229		goto bad;
230	}
231
232	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
233		nd6log((LOG_ERR,
234		    "nd6_ra_input: src %s is not link-local\n",
235		    ip6_sprintf(ip6bufs, &saddr6)));
236		goto bad;
237	}
238
239#ifndef PULLDOWN_TEST
240	IP6_EXTHDR_CHECK(m, off, icmp6len,);
241	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
242#else
243	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
244	if (nd_ra == NULL) {
245		V_icmp6stat.icp6s_tooshort++;
246		return;
247	}
248#endif
249
250	icmp6len -= sizeof(*nd_ra);
251	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
252	if (nd6_options(&ndopts) < 0) {
253		nd6log((LOG_INFO,
254		    "nd6_ra_input: invalid ND option, ignored\n"));
255		/* nd6_options have incremented stats */
256		goto freeit;
257	}
258
259    {
260	struct nd_defrouter dr0;
261	u_int32_t advreachable = nd_ra->nd_ra_reachable;
262
263	/* remember if this is a multicasted advertisement */
264	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
265		mcast = 1;
266
267	bzero(&dr0, sizeof(dr0));
268	dr0.rtaddr = saddr6;
269	dr0.flags  = nd_ra->nd_ra_flags_reserved;
270	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
271	dr0.expire = time_second + dr0.rtlifetime;
272	dr0.ifp = ifp;
273	/* unspecified or not? (RFC 2461 6.3.4) */
274	if (advreachable) {
275		advreachable = ntohl(advreachable);
276		if (advreachable <= MAX_REACHABLE_TIME &&
277		    ndi->basereachable != advreachable) {
278			ndi->basereachable = advreachable;
279			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
280			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
281		}
282	}
283	if (nd_ra->nd_ra_retransmit)
284		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
285	if (nd_ra->nd_ra_curhoplimit)
286		ndi->chlim = nd_ra->nd_ra_curhoplimit;
287	dr = defrtrlist_update(&dr0);
288    }
289
290	/*
291	 * prefix
292	 */
293	if (ndopts.nd_opts_pi) {
294		struct nd_opt_hdr *pt;
295		struct nd_opt_prefix_info *pi = NULL;
296		struct nd_prefixctl pr;
297
298		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
299		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
300		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
301						(pt->nd_opt_len << 3))) {
302			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
303				continue;
304			pi = (struct nd_opt_prefix_info *)pt;
305
306			if (pi->nd_opt_pi_len != 4) {
307				nd6log((LOG_INFO,
308				    "nd6_ra_input: invalid option "
309				    "len %d for prefix information option, "
310				    "ignored\n", pi->nd_opt_pi_len));
311				continue;
312			}
313
314			if (128 < pi->nd_opt_pi_prefix_len) {
315				nd6log((LOG_INFO,
316				    "nd6_ra_input: invalid prefix "
317				    "len %d for prefix information option, "
318				    "ignored\n", pi->nd_opt_pi_prefix_len));
319				continue;
320			}
321
322			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
323			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
324				nd6log((LOG_INFO,
325				    "nd6_ra_input: invalid prefix "
326				    "%s, ignored\n",
327				    ip6_sprintf(ip6bufs,
328					&pi->nd_opt_pi_prefix)));
329				continue;
330			}
331
332			bzero(&pr, sizeof(pr));
333			pr.ndpr_prefix.sin6_family = AF_INET6;
334			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
335			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
336			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
337
338			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
339			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
340			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
341			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
342			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
343			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
344			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
345			(void)prelist_update(&pr, dr, m, mcast);
346		}
347	}
348
349	/*
350	 * MTU
351	 */
352	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
353		u_long mtu;
354		u_long maxmtu;
355
356		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
357
358		/* lower bound */
359		if (mtu < IPV6_MMTU) {
360			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
361			    "mtu=%lu sent from %s, ignoring\n",
362			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
363			goto skip;
364		}
365
366		/* upper bound */
367		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
368		    ? ndi->maxmtu : ifp->if_mtu;
369		if (mtu <= maxmtu) {
370			int change = (ndi->linkmtu != mtu);
371
372			ndi->linkmtu = mtu;
373			if (change) /* in6_maxmtu may change */
374				in6_setmaxmtu();
375		} else {
376			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
377			    "mtu=%lu sent from %s; "
378			    "exceeds maxmtu %lu, ignoring\n",
379			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
380		}
381	}
382
383 skip:
384
385	/*
386	 * Source link layer address
387	 */
388    {
389	char *lladdr = NULL;
390	int lladdrlen = 0;
391
392	if (ndopts.nd_opts_src_lladdr) {
393		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
394		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
395	}
396
397	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
398		nd6log((LOG_INFO,
399		    "nd6_ra_input: lladdrlen mismatch for %s "
400		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
401		    ifp->if_addrlen, lladdrlen - 2));
402		goto bad;
403	}
404
405	nd6_cache_lladdr(ifp, &saddr6, lladdr,
406	    lladdrlen, ND_ROUTER_ADVERT, 0);
407
408	/*
409	 * Installing a link-layer address might change the state of the
410	 * router's neighbor cache, which might also affect our on-link
411	 * detection of adveritsed prefixes.
412	 */
413	pfxlist_onlink_check();
414    }
415
416 freeit:
417	m_freem(m);
418	return;
419
420 bad:
421	V_icmp6stat.icp6s_badra++;
422	m_freem(m);
423}
424
425/*
426 * default router list proccessing sub routines
427 */
428
429/* tell the change to user processes watching the routing socket. */
430static void
431nd6_rtmsg(int cmd, struct rtentry *rt)
432{
433	struct rt_addrinfo info;
434
435	bzero((caddr_t)&info, sizeof(info));
436	info.rti_info[RTAX_DST] = rt_key(rt);
437	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
438	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
439	if (rt->rt_ifp) {
440		info.rti_info[RTAX_IFP] =
441		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
442		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
443	}
444
445	rt_missmsg(cmd, &info, rt->rt_flags, 0);
446}
447
448void
449defrouter_addreq(struct nd_defrouter *new)
450{
451	struct sockaddr_in6 def, mask, gate;
452	struct rtentry *newrt = NULL;
453	int s;
454	int error;
455
456	bzero(&def, sizeof(def));
457	bzero(&mask, sizeof(mask));
458	bzero(&gate, sizeof(gate));
459
460	def.sin6_len = mask.sin6_len = gate.sin6_len =
461	    sizeof(struct sockaddr_in6);
462	def.sin6_family = gate.sin6_family = AF_INET6;
463	gate.sin6_addr = new->rtaddr;
464
465	s = splnet();
466	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
467	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
468	    RTF_GATEWAY, &newrt);
469	if (newrt) {
470		RT_LOCK(newrt);
471		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
472		RT_REMREF(newrt);
473		RT_UNLOCK(newrt);
474	}
475	if (error == 0)
476		new->installed = 1;
477	splx(s);
478	return;
479}
480
481struct nd_defrouter *
482defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
483{
484	INIT_VNET_INET6(ifp->if_vnet);
485	struct nd_defrouter *dr;
486
487	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
488	     dr = TAILQ_NEXT(dr, dr_entry)) {
489		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
490			return (dr);
491	}
492
493	return (NULL);		/* search failed */
494}
495
496/*
497 * Remove the default route for a given router.
498 * This is just a subroutine function for defrouter_select(), and should
499 * not be called from anywhere else.
500 */
501static void
502defrouter_delreq(struct nd_defrouter *dr)
503{
504	struct sockaddr_in6 def, mask, gate;
505	struct rtentry *oldrt = NULL;
506
507	bzero(&def, sizeof(def));
508	bzero(&mask, sizeof(mask));
509	bzero(&gate, sizeof(gate));
510
511	def.sin6_len = mask.sin6_len = gate.sin6_len =
512	    sizeof(struct sockaddr_in6);
513	def.sin6_family = gate.sin6_family = AF_INET6;
514	gate.sin6_addr = dr->rtaddr;
515
516	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
517	    (struct sockaddr *)&gate,
518	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
519	if (oldrt) {
520		nd6_rtmsg(RTM_DELETE, oldrt);
521		RTFREE(oldrt);
522	}
523
524	dr->installed = 0;
525}
526
527/*
528 * remove all default routes from default router list
529 */
530void
531defrouter_reset(void)
532{
533	INIT_VNET_INET6(curvnet);
534	struct nd_defrouter *dr;
535
536	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
537	     dr = TAILQ_NEXT(dr, dr_entry))
538		defrouter_delreq(dr);
539
540	/*
541	 * XXX should we also nuke any default routers in the kernel, by
542	 * going through them by rtalloc1()?
543	 */
544}
545
546void
547defrtrlist_del(struct nd_defrouter *dr)
548{
549	INIT_VNET_INET6(curvnet);
550	struct nd_defrouter *deldr = NULL;
551	struct nd_prefix *pr;
552
553	/*
554	 * Flush all the routing table entries that use the router
555	 * as a next hop.
556	 */
557	if (!V_ip6_forwarding && V_ip6_accept_rtadv) /* XXX: better condition? */
558		rt6_flush(&dr->rtaddr, dr->ifp);
559
560	if (dr->installed) {
561		deldr = dr;
562		defrouter_delreq(dr);
563	}
564	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
565
566	/*
567	 * Also delete all the pointers to the router in each prefix lists.
568	 */
569	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
570		struct nd_pfxrouter *pfxrtr;
571		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
572			pfxrtr_del(pfxrtr);
573	}
574	pfxlist_onlink_check();
575
576	/*
577	 * If the router is the primary one, choose a new one.
578	 * Note that defrouter_select() will remove the current gateway
579	 * from the routing table.
580	 */
581	if (deldr)
582		defrouter_select();
583
584	free(dr, M_IP6NDP);
585}
586
587/*
588 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
589 * draft-ietf-ipngwg-router-selection:
590 * 1) Routers that are reachable or probably reachable should be preferred.
591 *    If we have more than one (probably) reachable router, prefer ones
592 *    with the highest router preference.
593 * 2) When no routers on the list are known to be reachable or
594 *    probably reachable, routers SHOULD be selected in a round-robin
595 *    fashion, regardless of router preference values.
596 * 3) If the Default Router List is empty, assume that all
597 *    destinations are on-link.
598 *
599 * We assume nd_defrouter is sorted by router preference value.
600 * Since the code below covers both with and without router preference cases,
601 * we do not need to classify the cases by ifdef.
602 *
603 * At this moment, we do not try to install more than one default router,
604 * even when the multipath routing is available, because we're not sure about
605 * the benefits for stub hosts comparing to the risk of making the code
606 * complicated and the possibility of introducing bugs.
607 */
608void
609defrouter_select(void)
610{
611	INIT_VNET_INET6(curvnet);
612	int s = splnet();
613	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
614	struct rtentry *rt = NULL;
615	struct llinfo_nd6 *ln = NULL;
616
617	/*
618	 * This function should be called only when acting as an autoconfigured
619	 * host.  Although the remaining part of this function is not effective
620	 * if the node is not an autoconfigured host, we explicitly exclude
621	 * such cases here for safety.
622	 */
623	if (V_ip6_forwarding || !V_ip6_accept_rtadv) {
624		nd6log((LOG_WARNING,
625		    "defrouter_select: called unexpectedly (forwarding=%d, "
626		    "accept_rtadv=%d)\n", V_ip6_forwarding, V_ip6_accept_rtadv));
627		splx(s);
628		return;
629	}
630
631	/*
632	 * Let's handle easy case (3) first:
633	 * If default router list is empty, there's nothing to be done.
634	 */
635	if (!TAILQ_FIRST(&V_nd_defrouter)) {
636		splx(s);
637		return;
638	}
639
640	/*
641	 * Search for a (probably) reachable router from the list.
642	 * We just pick up the first reachable one (if any), assuming that
643	 * the ordering rule of the list described in defrtrlist_update().
644	 */
645	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
646	     dr = TAILQ_NEXT(dr, dr_entry)) {
647		if (selected_dr == NULL &&
648		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
649		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
650		    ND6_IS_LLINFO_PROBREACH(ln)) {
651			selected_dr = dr;
652		}
653
654		if (dr->installed && installed_dr == NULL)
655			installed_dr = dr;
656		else if (dr->installed && installed_dr) {
657			/* this should not happen.  warn for diagnosis. */
658			log(LOG_ERR, "defrouter_select: more than one router"
659			    " is installed\n");
660		}
661	}
662	/*
663	 * If none of the default routers was found to be reachable,
664	 * round-robin the list regardless of preference.
665	 * Otherwise, if we have an installed router, check if the selected
666	 * (reachable) router should really be preferred to the installed one.
667	 * We only prefer the new router when the old one is not reachable
668	 * or when the new one has a really higher preference value.
669	 */
670	if (selected_dr == NULL) {
671		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
672			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
673		else
674			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
675	} else if (installed_dr &&
676	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
677	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
678	    ND6_IS_LLINFO_PROBREACH(ln) &&
679	    rtpref(selected_dr) <= rtpref(installed_dr)) {
680		selected_dr = installed_dr;
681	}
682
683	/*
684	 * If the selected router is different than the installed one,
685	 * remove the installed router and install the selected one.
686	 * Note that the selected router is never NULL here.
687	 */
688	if (installed_dr != selected_dr) {
689		if (installed_dr)
690			defrouter_delreq(installed_dr);
691		defrouter_addreq(selected_dr);
692	}
693
694	splx(s);
695	return;
696}
697
698/*
699 * for default router selection
700 * regards router-preference field as a 2-bit signed integer
701 */
702static int
703rtpref(struct nd_defrouter *dr)
704{
705	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
706	case ND_RA_FLAG_RTPREF_HIGH:
707		return (RTPREF_HIGH);
708	case ND_RA_FLAG_RTPREF_MEDIUM:
709	case ND_RA_FLAG_RTPREF_RSV:
710		return (RTPREF_MEDIUM);
711	case ND_RA_FLAG_RTPREF_LOW:
712		return (RTPREF_LOW);
713	default:
714		/*
715		 * This case should never happen.  If it did, it would mean a
716		 * serious bug of kernel internal.  We thus always bark here.
717		 * Or, can we even panic?
718		 */
719		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
720		return (RTPREF_INVALID);
721	}
722	/* NOTREACHED */
723}
724
725static struct nd_defrouter *
726defrtrlist_update(struct nd_defrouter *new)
727{
728	INIT_VNET_INET6(curvnet);
729	struct nd_defrouter *dr, *n;
730	int s = splnet();
731
732	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
733		/* entry exists */
734		if (new->rtlifetime == 0) {
735			defrtrlist_del(dr);
736			dr = NULL;
737		} else {
738			int oldpref = rtpref(dr);
739
740			/* override */
741			dr->flags = new->flags; /* xxx flag check */
742			dr->rtlifetime = new->rtlifetime;
743			dr->expire = new->expire;
744
745			/*
746			 * If the preference does not change, there's no need
747			 * to sort the entries.
748			 */
749			if (rtpref(new) == oldpref) {
750				splx(s);
751				return (dr);
752			}
753
754			/*
755			 * preferred router may be changed, so relocate
756			 * this router.
757			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
758			 * However, since defrtrlist_del() has many side
759			 * effects, we intentionally do so here.
760			 * defrouter_select() below will handle routing
761			 * changes later.
762			 */
763			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
764			n = dr;
765			goto insert;
766		}
767		splx(s);
768		return (dr);
769	}
770
771	/* entry does not exist */
772	if (new->rtlifetime == 0) {
773		splx(s);
774		return (NULL);
775	}
776
777	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
778	if (n == NULL) {
779		splx(s);
780		return (NULL);
781	}
782	bzero(n, sizeof(*n));
783	*n = *new;
784
785insert:
786	/*
787	 * Insert the new router in the Default Router List;
788	 * The Default Router List should be in the descending order
789	 * of router-preferece.  Routers with the same preference are
790	 * sorted in the arriving time order.
791	 */
792
793	/* insert at the end of the group */
794	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
795	     dr = TAILQ_NEXT(dr, dr_entry)) {
796		if (rtpref(n) > rtpref(dr))
797			break;
798	}
799	if (dr)
800		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
801	else
802		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
803
804	defrouter_select();
805
806	splx(s);
807
808	return (n);
809}
810
811static struct nd_pfxrouter *
812pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
813{
814	struct nd_pfxrouter *search;
815
816	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
817		if (search->router == dr)
818			break;
819	}
820
821	return (search);
822}
823
824static void
825pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
826{
827	struct nd_pfxrouter *new;
828
829	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
830	if (new == NULL)
831		return;
832	bzero(new, sizeof(*new));
833	new->router = dr;
834
835	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
836
837	pfxlist_onlink_check();
838}
839
840static void
841pfxrtr_del(struct nd_pfxrouter *pfr)
842{
843	LIST_REMOVE(pfr, pfr_entry);
844	free(pfr, M_IP6NDP);
845}
846
847struct nd_prefix *
848nd6_prefix_lookup(struct nd_prefixctl *key)
849{
850	INIT_VNET_INET6(curvnet);
851	struct nd_prefix *search;
852
853	for (search = V_nd_prefix.lh_first;
854	    search; search = search->ndpr_next) {
855		if (key->ndpr_ifp == search->ndpr_ifp &&
856		    key->ndpr_plen == search->ndpr_plen &&
857		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
858		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
859			break;
860		}
861	}
862
863	return (search);
864}
865
866int
867nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
868    struct nd_prefix **newp)
869{
870	INIT_VNET_INET6(curvnet);
871	struct nd_prefix *new = NULL;
872	int error = 0;
873	int i, s;
874	char ip6buf[INET6_ADDRSTRLEN];
875
876	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
877	if (new == NULL)
878		return(ENOMEM);
879	bzero(new, sizeof(*new));
880	new->ndpr_ifp = pr->ndpr_ifp;
881	new->ndpr_prefix = pr->ndpr_prefix;
882	new->ndpr_plen = pr->ndpr_plen;
883	new->ndpr_vltime = pr->ndpr_vltime;
884	new->ndpr_pltime = pr->ndpr_pltime;
885	new->ndpr_flags = pr->ndpr_flags;
886	if ((error = in6_init_prefix_ltimes(new)) != 0) {
887		free(new, M_IP6NDP);
888		return(error);
889	}
890	new->ndpr_lastupdate = time_second;
891	if (newp != NULL)
892		*newp = new;
893
894	/* initialization */
895	LIST_INIT(&new->ndpr_advrtrs);
896	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
897	/* make prefix in the canonical form */
898	for (i = 0; i < 4; i++)
899		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
900		    new->ndpr_mask.s6_addr32[i];
901
902	s = splnet();
903	/* link ndpr_entry to nd_prefix list */
904	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
905	splx(s);
906
907	/* ND_OPT_PI_FLAG_ONLINK processing */
908	if (new->ndpr_raf_onlink) {
909		int e;
910
911		if ((e = nd6_prefix_onlink(new)) != 0) {
912			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
913			    "the prefix %s/%d on-link on %s (errno=%d)\n",
914			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
915			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
916			/* proceed anyway. XXX: is it correct? */
917		}
918	}
919
920	if (dr)
921		pfxrtr_add(new, dr);
922
923	return 0;
924}
925
926void
927prelist_remove(struct nd_prefix *pr)
928{
929	INIT_VNET_INET6(curvnet);
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	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
963		next = pfr->pfr_next;
964
965		free(pfr, M_IP6NDP);
966	}
967	splx(s);
968
969	free(pr, M_IP6NDP);
970
971	pfxlist_onlink_check();
972}
973
974/*
975 * dr - may be NULL
976 */
977
978static int
979prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
980    struct mbuf *m, int mcast)
981{
982	INIT_VNET_INET6(curvnet);
983	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
984	struct ifaddr *ifa;
985	struct ifnet *ifp = new->ndpr_ifp;
986	struct nd_prefix *pr;
987	int s = splnet();
988	int error = 0;
989	int newprefix = 0;
990	int auth;
991	struct in6_addrlifetime lt6_tmp;
992	char ip6buf[INET6_ADDRSTRLEN];
993
994	auth = 0;
995	if (m) {
996		/*
997		 * Authenticity for NA consists authentication for
998		 * both IP header and IP datagrams, doesn't it ?
999		 */
1000#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1001		auth = ((m->m_flags & M_AUTHIPHDR) &&
1002		    (m->m_flags & M_AUTHIPDGM));
1003#endif
1004	}
1005
1006	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1007		/*
1008		 * nd6_prefix_lookup() ensures that pr and new have the same
1009		 * prefix on a same interface.
1010		 */
1011
1012		/*
1013		 * Update prefix information.  Note that the on-link (L) bit
1014		 * and the autonomous (A) bit should NOT be changed from 1
1015		 * to 0.
1016		 */
1017		if (new->ndpr_raf_onlink == 1)
1018			pr->ndpr_raf_onlink = 1;
1019		if (new->ndpr_raf_auto == 1)
1020			pr->ndpr_raf_auto = 1;
1021		if (new->ndpr_raf_onlink) {
1022			pr->ndpr_vltime = new->ndpr_vltime;
1023			pr->ndpr_pltime = new->ndpr_pltime;
1024			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1025			pr->ndpr_lastupdate = time_second;
1026		}
1027
1028		if (new->ndpr_raf_onlink &&
1029		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1030			int e;
1031
1032			if ((e = nd6_prefix_onlink(pr)) != 0) {
1033				nd6log((LOG_ERR,
1034				    "prelist_update: failed to make "
1035				    "the prefix %s/%d on-link on %s "
1036				    "(errno=%d)\n",
1037				    ip6_sprintf(ip6buf,
1038					    &pr->ndpr_prefix.sin6_addr),
1039				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1040				/* proceed anyway. XXX: is it correct? */
1041			}
1042		}
1043
1044		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1045			pfxrtr_add(pr, dr);
1046	} else {
1047		struct nd_prefix *newpr = NULL;
1048
1049		newprefix = 1;
1050
1051		if (new->ndpr_vltime == 0)
1052			goto end;
1053		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1054			goto end;
1055
1056		error = nd6_prelist_add(new, dr, &newpr);
1057		if (error != 0 || newpr == NULL) {
1058			nd6log((LOG_NOTICE, "prelist_update: "
1059			    "nd6_prelist_add failed for %s/%d on %s "
1060			    "errno=%d, returnpr=%p\n",
1061			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1062			    new->ndpr_plen, if_name(new->ndpr_ifp),
1063			    error, newpr));
1064			goto end; /* we should just give up in this case. */
1065		}
1066
1067		/*
1068		 * XXX: from the ND point of view, we can ignore a prefix
1069		 * with the on-link bit being zero.  However, we need a
1070		 * prefix structure for references from autoconfigured
1071		 * addresses.  Thus, we explicitly make sure that the prefix
1072		 * itself expires now.
1073		 */
1074		if (newpr->ndpr_raf_onlink == 0) {
1075			newpr->ndpr_vltime = 0;
1076			newpr->ndpr_pltime = 0;
1077			in6_init_prefix_ltimes(newpr);
1078		}
1079
1080		pr = newpr;
1081	}
1082
1083	/*
1084	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1085	 * Note that pr must be non NULL at this point.
1086	 */
1087
1088	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1089	if (!new->ndpr_raf_auto)
1090		goto end;
1091
1092	/*
1093	 * 5.5.3 (b). the link-local prefix should have been ignored in
1094	 * nd6_ra_input.
1095	 */
1096
1097	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1098	if (new->ndpr_pltime > new->ndpr_vltime) {
1099		error = EINVAL;	/* XXX: won't be used */
1100		goto end;
1101	}
1102
1103	/*
1104	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1105	 * an address configured by stateless autoconfiguration already in the
1106	 * list of addresses associated with the interface, and the Valid
1107	 * Lifetime is not 0, form an address.  We first check if we have
1108	 * a matching prefix.
1109	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1110	 * consider autoconfigured addresses while RFC2462 simply said
1111	 * "address".
1112	 */
1113	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1114		struct in6_ifaddr *ifa6;
1115		u_int32_t remaininglifetime;
1116
1117		if (ifa->ifa_addr->sa_family != AF_INET6)
1118			continue;
1119
1120		ifa6 = (struct in6_ifaddr *)ifa;
1121
1122		/*
1123		 * We only consider autoconfigured addresses as per rfc2462bis.
1124		 */
1125		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1126			continue;
1127
1128		/*
1129		 * Spec is not clear here, but I believe we should concentrate
1130		 * on unicast (i.e. not anycast) addresses.
1131		 * XXX: other ia6_flags? detached or duplicated?
1132		 */
1133		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1134			continue;
1135
1136		/*
1137		 * Ignore the address if it is not associated with a prefix
1138		 * or is associated with a prefix that is different from this
1139		 * one.  (pr is never NULL here)
1140		 */
1141		if (ifa6->ia6_ndpr != pr)
1142			continue;
1143
1144		if (ia6_match == NULL) /* remember the first one */
1145			ia6_match = ifa6;
1146
1147		/*
1148		 * An already autoconfigured address matched.  Now that we
1149		 * are sure there is at least one matched address, we can
1150		 * proceed to 5.5.3. (e): update the lifetimes according to the
1151		 * "two hours" rule and the privacy extension.
1152		 * We apply some clarifications in rfc2462bis:
1153		 * - use remaininglifetime instead of storedlifetime as a
1154		 *   variable name
1155		 * - remove the dead code in the "two-hour" rule
1156		 */
1157#define TWOHOUR		(120*60)
1158		lt6_tmp = ifa6->ia6_lifetime;
1159
1160		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1161			remaininglifetime = ND6_INFINITE_LIFETIME;
1162		else if (time_second - ifa6->ia6_updatetime >
1163			 lt6_tmp.ia6t_vltime) {
1164			/*
1165			 * The case of "invalid" address.  We should usually
1166			 * not see this case.
1167			 */
1168			remaininglifetime = 0;
1169		} else
1170			remaininglifetime = lt6_tmp.ia6t_vltime -
1171			    (time_second - ifa6->ia6_updatetime);
1172
1173		/* when not updating, keep the current stored lifetime. */
1174		lt6_tmp.ia6t_vltime = remaininglifetime;
1175
1176		if (TWOHOUR < new->ndpr_vltime ||
1177		    remaininglifetime < new->ndpr_vltime) {
1178			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1179		} else if (remaininglifetime <= TWOHOUR) {
1180			if (auth) {
1181				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1182			}
1183		} else {
1184			/*
1185			 * new->ndpr_vltime <= TWOHOUR &&
1186			 * TWOHOUR < remaininglifetime
1187			 */
1188			lt6_tmp.ia6t_vltime = TWOHOUR;
1189		}
1190
1191		/* The 2 hour rule is not imposed for preferred lifetime. */
1192		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1193
1194		in6_init_address_ltimes(pr, &lt6_tmp);
1195
1196		/*
1197		 * We need to treat lifetimes for temporary addresses
1198		 * differently, according to
1199		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1200		 * we only update the lifetimes when they are in the maximum
1201		 * intervals.
1202		 */
1203		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1204			u_int32_t maxvltime, maxpltime;
1205
1206			if (V_ip6_temp_valid_lifetime >
1207			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1208			    V_ip6_desync_factor)) {
1209				maxvltime = V_ip6_temp_valid_lifetime -
1210				    (time_second - ifa6->ia6_createtime) -
1211				    V_ip6_desync_factor;
1212			} else
1213				maxvltime = 0;
1214			if (V_ip6_temp_preferred_lifetime >
1215			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1216			    V_ip6_desync_factor)) {
1217				maxpltime = V_ip6_temp_preferred_lifetime -
1218				    (time_second - ifa6->ia6_createtime) -
1219				    V_ip6_desync_factor;
1220			} else
1221				maxpltime = 0;
1222
1223			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1224			    lt6_tmp.ia6t_vltime > maxvltime) {
1225				lt6_tmp.ia6t_vltime = maxvltime;
1226			}
1227			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1228			    lt6_tmp.ia6t_pltime > maxpltime) {
1229				lt6_tmp.ia6t_pltime = maxpltime;
1230			}
1231		}
1232		ifa6->ia6_lifetime = lt6_tmp;
1233		ifa6->ia6_updatetime = time_second;
1234	}
1235	if (ia6_match == NULL && new->ndpr_vltime) {
1236		int ifidlen;
1237
1238		/*
1239		 * 5.5.3 (d) (continued)
1240		 * No address matched and the valid lifetime is non-zero.
1241		 * Create a new address.
1242		 */
1243
1244		/*
1245		 * Prefix Length check:
1246		 * If the sum of the prefix length and interface identifier
1247		 * length does not equal 128 bits, the Prefix Information
1248		 * option MUST be ignored.  The length of the interface
1249		 * identifier is defined in a separate link-type specific
1250		 * document.
1251		 */
1252		ifidlen = in6_if2idlen(ifp);
1253		if (ifidlen < 0) {
1254			/* this should not happen, so we always log it. */
1255			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1256			    if_name(ifp));
1257			goto end;
1258		}
1259		if (ifidlen + pr->ndpr_plen != 128) {
1260			nd6log((LOG_INFO,
1261			    "prelist_update: invalid prefixlen "
1262			    "%d for %s, ignored\n",
1263			    pr->ndpr_plen, if_name(ifp)));
1264			goto end;
1265		}
1266
1267		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1268			/*
1269			 * note that we should use pr (not new) for reference.
1270			 */
1271			pr->ndpr_refcnt++;
1272			ia6->ia6_ndpr = pr;
1273
1274			/*
1275			 * RFC 3041 3.3 (2).
1276			 * When a new public address is created as described
1277			 * in RFC2462, also create a new temporary address.
1278			 *
1279			 * RFC 3041 3.5.
1280			 * When an interface connects to a new link, a new
1281			 * randomized interface identifier should be generated
1282			 * immediately together with a new set of temporary
1283			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1284			 * in6_tmpifadd().
1285			 */
1286			if (V_ip6_use_tempaddr) {
1287				int e;
1288				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1289					nd6log((LOG_NOTICE, "prelist_update: "
1290					    "failed to create a temporary "
1291					    "address, errno=%d\n",
1292					    e));
1293				}
1294			}
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 rtentry *rt;
1323	struct llinfo_nd6 *ln;
1324
1325	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1326	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1327		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1328		    pfxrtr->router->ifp)) &&
1329		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1330		    ND6_IS_LLINFO_PROBREACH(ln))
1331			break;	/* found */
1332	}
1333
1334	return (pfxrtr);
1335}
1336
1337/*
1338 * Check if each prefix in the prefix list has at least one available router
1339 * that advertised the prefix (a router is "available" if its neighbor cache
1340 * entry is reachable or probably reachable).
1341 * If the check fails, the prefix may be off-link, because, for example,
1342 * we have moved from the network but the lifetime of the prefix has not
1343 * expired yet.  So we should not use the prefix if there is another prefix
1344 * that has an available router.
1345 * But, if there is no prefix that has an available router, we still regards
1346 * all the prefixes as on-link.  This is because we can't tell if all the
1347 * routers are simply dead or if we really moved from the network and there
1348 * is no router around us.
1349 */
1350void
1351pfxlist_onlink_check()
1352{
1353	INIT_VNET_INET6(curvnet);
1354	struct nd_prefix *pr;
1355	struct in6_ifaddr *ifa;
1356	struct nd_defrouter *dr;
1357	struct nd_pfxrouter *pfxrtr = NULL;
1358
1359	/*
1360	 * Check if there is a prefix that has a reachable advertising
1361	 * router.
1362	 */
1363	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1364		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1365			break;
1366	}
1367
1368	/*
1369	 * If we have no such prefix, check whether we still have a router
1370	 * that does not advertise any prefixes.
1371	 */
1372	if (pr == NULL) {
1373		for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1374		    dr = TAILQ_NEXT(dr, dr_entry)) {
1375			struct nd_prefix *pr0;
1376
1377			for (pr0 = V_nd_prefix.lh_first; pr0;
1378			    pr0 = pr0->ndpr_next) {
1379				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1380					break;
1381			}
1382			if (pfxrtr != NULL)
1383				break;
1384		}
1385	}
1386	if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1387		/*
1388		 * There is at least one prefix that has a reachable router,
1389		 * or at least a router which probably does not advertise
1390		 * any prefixes.  The latter would be the case when we move
1391		 * to a new link where we have a router that does not provide
1392		 * prefixes and we configure an address by hand.
1393		 * Detach prefixes which have no reachable advertising
1394		 * router, and attach other prefixes.
1395		 */
1396		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1397			/* XXX: a link-local prefix should never be detached */
1398			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1399				continue;
1400
1401			/*
1402			 * we aren't interested in prefixes without the L bit
1403			 * set.
1404			 */
1405			if (pr->ndpr_raf_onlink == 0)
1406				continue;
1407
1408			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1409			    find_pfxlist_reachable_router(pr) == NULL)
1410				pr->ndpr_stateflags |= NDPRF_DETACHED;
1411			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1412			    find_pfxlist_reachable_router(pr) != 0)
1413				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1414		}
1415	} else {
1416		/* there is no prefix that has a reachable router */
1417		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1418			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1419				continue;
1420
1421			if (pr->ndpr_raf_onlink == 0)
1422				continue;
1423
1424			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1425				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1426		}
1427	}
1428
1429	/*
1430	 * Remove each interface route associated with a (just) detached
1431	 * prefix, and reinstall the interface route for a (just) attached
1432	 * prefix.  Note that all attempt of reinstallation does not
1433	 * necessarily success, when a same prefix is shared among multiple
1434	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1435	 * so we don't have to care about them.
1436	 */
1437	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1438		int e;
1439		char ip6buf[INET6_ADDRSTRLEN];
1440
1441		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1442			continue;
1443
1444		if (pr->ndpr_raf_onlink == 0)
1445			continue;
1446
1447		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1448		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1449			if ((e = nd6_prefix_offlink(pr)) != 0) {
1450				nd6log((LOG_ERR,
1451				    "pfxlist_onlink_check: failed to "
1452				    "make %s/%d offlink, errno=%d\n",
1453				    ip6_sprintf(ip6buf,
1454					    &pr->ndpr_prefix.sin6_addr),
1455					    pr->ndpr_plen, e));
1456			}
1457		}
1458		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1459		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1460		    pr->ndpr_raf_onlink) {
1461			if ((e = nd6_prefix_onlink(pr)) != 0) {
1462				nd6log((LOG_ERR,
1463				    "pfxlist_onlink_check: failed to "
1464				    "make %s/%d onlink, errno=%d\n",
1465				    ip6_sprintf(ip6buf,
1466					    &pr->ndpr_prefix.sin6_addr),
1467					    pr->ndpr_plen, e));
1468			}
1469		}
1470	}
1471
1472	/*
1473	 * Changes on the prefix status might affect address status as well.
1474	 * Make sure that all addresses derived from an attached prefix are
1475	 * attached, and that all addresses derived from a detached prefix are
1476	 * detached.  Note, however, that a manually configured address should
1477	 * always be attached.
1478	 * The precise detection logic is same as the one for prefixes.
1479	 */
1480	for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1481		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1482			continue;
1483
1484		if (ifa->ia6_ndpr == NULL) {
1485			/*
1486			 * This can happen when we first configure the address
1487			 * (i.e. the address exists, but the prefix does not).
1488			 * XXX: complicated relationships...
1489			 */
1490			continue;
1491		}
1492
1493		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1494			break;
1495	}
1496	if (ifa) {
1497		for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1498			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1499				continue;
1500
1501			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1502				continue;
1503
1504			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1505				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1506					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1507					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1508					nd6_dad_start((struct ifaddr *)ifa, 0);
1509				}
1510			} else {
1511				ifa->ia6_flags |= IN6_IFF_DETACHED;
1512			}
1513		}
1514	}
1515	else {
1516		for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1517			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1518				continue;
1519
1520			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1521				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1522				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1523				/* Do we need a delay in this case? */
1524				nd6_dad_start((struct ifaddr *)ifa, 0);
1525			}
1526		}
1527	}
1528}
1529
1530int
1531nd6_prefix_onlink(struct nd_prefix *pr)
1532{
1533	INIT_VNET_INET6(curvnet);
1534	struct ifaddr *ifa;
1535	struct ifnet *ifp = pr->ndpr_ifp;
1536	struct sockaddr_in6 mask6;
1537	struct nd_prefix *opr;
1538	u_long rtflags;
1539	int error = 0;
1540	struct rtentry *rt = NULL;
1541	char ip6buf[INET6_ADDRSTRLEN];
1542
1543	/* sanity check */
1544	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1545		nd6log((LOG_ERR,
1546		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1547		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1548		    pr->ndpr_plen));
1549		return (EEXIST);
1550	}
1551
1552	/*
1553	 * Add the interface route associated with the prefix.  Before
1554	 * installing the route, check if there's the same prefix on another
1555	 * interface, and the prefix has already installed the interface route.
1556	 * Although such a configuration is expected to be rare, we explicitly
1557	 * allow it.
1558	 */
1559	for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1560		if (opr == pr)
1561			continue;
1562
1563		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1564			continue;
1565
1566		if (opr->ndpr_plen == pr->ndpr_plen &&
1567		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1568		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1569			return (0);
1570	}
1571
1572	/*
1573	 * We prefer link-local addresses as the associated interface address.
1574	 */
1575	/* search for a link-local addr */
1576	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1577	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1578	if (ifa == NULL) {
1579		/* XXX: freebsd does not have ifa_ifwithaf */
1580		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1581			if (ifa->ifa_addr->sa_family == AF_INET6)
1582				break;
1583		}
1584		/* should we care about ia6_flags? */
1585	}
1586	if (ifa == NULL) {
1587		/*
1588		 * This can still happen, when, for example, we receive an RA
1589		 * containing a prefix with the L bit set and the A bit clear,
1590		 * after removing all IPv6 addresses on the receiving
1591		 * interface.  This should, of course, be rare though.
1592		 */
1593		nd6log((LOG_NOTICE,
1594		    "nd6_prefix_onlink: failed to find any ifaddr"
1595		    " to add route for a prefix(%s/%d) on %s\n",
1596		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1597		    pr->ndpr_plen, if_name(ifp)));
1598		return (0);
1599	}
1600
1601	/*
1602	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1603	 * ifa->ifa_rtrequest = nd6_rtrequest;
1604	 */
1605	bzero(&mask6, sizeof(mask6));
1606	mask6.sin6_len = sizeof(mask6);
1607	mask6.sin6_addr = pr->ndpr_mask;
1608	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1609	if (nd6_need_cache(ifp)) {
1610		/* explicitly set in case ifa_flags does not set the flag. */
1611		rtflags |= RTF_CLONING;
1612	} else {
1613		/*
1614		 * explicitly clear the cloning bit in case ifa_flags sets it.
1615		 */
1616		rtflags &= ~RTF_CLONING;
1617	}
1618	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1619	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1620	if (error == 0) {
1621		if (rt != NULL) /* this should be non NULL, though */
1622			nd6_rtmsg(RTM_ADD, rt);
1623		pr->ndpr_stateflags |= NDPRF_ONLINK;
1624	} else {
1625		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1626		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1627		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1628		    "errno = %d\n",
1629		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1630		    pr->ndpr_plen, if_name(ifp),
1631		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1632		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1633	}
1634
1635	if (rt != NULL) {
1636		RT_LOCK(rt);
1637		RT_REMREF(rt);
1638		RT_UNLOCK(rt);
1639	}
1640
1641	return (error);
1642}
1643
1644int
1645nd6_prefix_offlink(struct nd_prefix *pr)
1646{
1647	INIT_VNET_INET6(curvnet);
1648	int error = 0;
1649	struct ifnet *ifp = pr->ndpr_ifp;
1650	struct nd_prefix *opr;
1651	struct sockaddr_in6 sa6, mask6;
1652	struct rtentry *rt = NULL;
1653	char ip6buf[INET6_ADDRSTRLEN];
1654
1655	/* sanity check */
1656	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1657		nd6log((LOG_ERR,
1658		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1659		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1660		    pr->ndpr_plen));
1661		return (EEXIST);
1662	}
1663
1664	bzero(&sa6, sizeof(sa6));
1665	sa6.sin6_family = AF_INET6;
1666	sa6.sin6_len = sizeof(sa6);
1667	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1668	    sizeof(struct in6_addr));
1669	bzero(&mask6, sizeof(mask6));
1670	mask6.sin6_family = AF_INET6;
1671	mask6.sin6_len = sizeof(sa6);
1672	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1673	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1674	    (struct sockaddr *)&mask6, 0, &rt);
1675	if (error == 0) {
1676		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1677
1678		/* report the route deletion to the routing socket. */
1679		if (rt != NULL)
1680			nd6_rtmsg(RTM_DELETE, rt);
1681
1682		/*
1683		 * There might be the same prefix on another interface,
1684		 * the prefix which could not be on-link just because we have
1685		 * the interface route (see comments in nd6_prefix_onlink).
1686		 * If there's one, try to make the prefix on-link on the
1687		 * interface.
1688		 */
1689		for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1690			if (opr == pr)
1691				continue;
1692
1693			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1694				continue;
1695
1696			/*
1697			 * KAME specific: detached prefixes should not be
1698			 * on-link.
1699			 */
1700			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1701				continue;
1702
1703			if (opr->ndpr_plen == pr->ndpr_plen &&
1704			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1705			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1706				int e;
1707
1708				if ((e = nd6_prefix_onlink(opr)) != 0) {
1709					nd6log((LOG_ERR,
1710					    "nd6_prefix_offlink: failed to "
1711					    "recover a prefix %s/%d from %s "
1712					    "to %s (errno = %d)\n",
1713					    ip6_sprintf(ip6buf,
1714						&opr->ndpr_prefix.sin6_addr),
1715					    opr->ndpr_plen, if_name(ifp),
1716					    if_name(opr->ndpr_ifp), e));
1717				}
1718			}
1719		}
1720	} else {
1721		/* XXX: can we still set the NDPRF_ONLINK flag? */
1722		nd6log((LOG_ERR,
1723		    "nd6_prefix_offlink: failed to delete route: "
1724		    "%s/%d on %s (errno = %d)\n",
1725		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1726		    if_name(ifp), error));
1727	}
1728
1729	if (rt != NULL) {
1730		RTFREE(rt);
1731	}
1732
1733	return (error);
1734}
1735
1736static struct in6_ifaddr *
1737in6_ifadd(struct nd_prefixctl *pr, int mcast)
1738{
1739	INIT_VNET_INET6(curvnet);
1740	struct ifnet *ifp = pr->ndpr_ifp;
1741	struct ifaddr *ifa;
1742	struct in6_aliasreq ifra;
1743	struct in6_ifaddr *ia, *ib;
1744	int error, plen0;
1745	struct in6_addr mask;
1746	int prefixlen = pr->ndpr_plen;
1747	int updateflags;
1748	char ip6buf[INET6_ADDRSTRLEN];
1749
1750	in6_prefixlen2mask(&mask, prefixlen);
1751
1752	/*
1753	 * find a link-local address (will be interface ID).
1754	 * Is it really mandatory? Theoretically, a global or a site-local
1755	 * address can be configured without a link-local address, if we
1756	 * have a unique interface identifier...
1757	 *
1758	 * it is not mandatory to have a link-local address, we can generate
1759	 * interface identifier on the fly.  we do this because:
1760	 * (1) it should be the easiest way to find interface identifier.
1761	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1762	 * for multiple addresses on a single interface, and possible shortcut
1763	 * of DAD.  we omitted DAD for this reason in the past.
1764	 * (3) a user can prevent autoconfiguration of global address
1765	 * by removing link-local address by hand (this is partly because we
1766	 * don't have other way to control the use of IPv6 on an interface.
1767	 * this has been our design choice - cf. NRL's "ifconfig auto").
1768	 * (4) it is easier to manage when an interface has addresses
1769	 * with the same interface identifier, than to have multiple addresses
1770	 * with different interface identifiers.
1771	 */
1772	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1773	if (ifa)
1774		ib = (struct in6_ifaddr *)ifa;
1775	else
1776		return NULL;
1777
1778	/* prefixlen + ifidlen must be equal to 128 */
1779	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1780	if (prefixlen != plen0) {
1781		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1782		    "(prefix=%d ifid=%d)\n",
1783		    if_name(ifp), prefixlen, 128 - plen0));
1784		return NULL;
1785	}
1786
1787	/* make ifaddr */
1788
1789	bzero(&ifra, sizeof(ifra));
1790	/*
1791	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1792	 * for safety.
1793	 */
1794	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1795	ifra.ifra_addr.sin6_family = AF_INET6;
1796	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1797	/* prefix */
1798	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1799	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1800	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1801	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1802	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1803
1804	/* interface ID */
1805	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1806	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1807	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1808	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1809	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1810	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1811	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1812	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1813
1814	/* new prefix mask. */
1815	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1816	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1817	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1818	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1819
1820	/* lifetimes. */
1821	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1822	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1823
1824	/* XXX: scope zone ID? */
1825
1826	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1827
1828	/*
1829	 * Make sure that we do not have this address already.  This should
1830	 * usually not happen, but we can still see this case, e.g., if we
1831	 * have manually configured the exact address to be configured.
1832	 */
1833	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1834		/* this should be rare enough to make an explicit log */
1835		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1836		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1837		return (NULL);
1838	}
1839
1840	/*
1841	 * Allocate ifaddr structure, link into chain, etc.
1842	 * If we are going to create a new address upon receiving a multicasted
1843	 * RA, we need to impose a random delay before starting DAD.
1844	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1845	 */
1846	updateflags = 0;
1847	if (mcast)
1848		updateflags |= IN6_IFAUPDATE_DADDELAY;
1849	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1850		nd6log((LOG_ERR,
1851		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1852		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1853		    if_name(ifp), error));
1854		return (NULL);	/* ifaddr must not have been allocated. */
1855	}
1856
1857	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1858
1859	return (ia);		/* this is always non-NULL */
1860}
1861
1862/*
1863 * ia0 - corresponding public address
1864 */
1865int
1866in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1867{
1868	INIT_VNET_INET6(curvnet);
1869	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1870	struct in6_ifaddr *newia, *ia;
1871	struct in6_aliasreq ifra;
1872	int i, error;
1873	int trylimit = 3;	/* XXX: adhoc value */
1874	int updateflags;
1875	u_int32_t randid[2];
1876	time_t vltime0, pltime0;
1877
1878	bzero(&ifra, sizeof(ifra));
1879	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1880	ifra.ifra_addr = ia0->ia_addr;
1881	/* copy prefix mask */
1882	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1883	/* clear the old IFID */
1884	for (i = 0; i < 4; i++) {
1885		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1886		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1887	}
1888
1889  again:
1890	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1891	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1892		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1893		    "random IFID\n"));
1894		return (EINVAL);
1895	}
1896	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1897	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1898	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1899	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1900
1901	/*
1902	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1903	 * However, we may still have a chance to see collision, because
1904	 * there may be a time lag between generation of the ID and generation
1905	 * of the address.  So, we'll do one more sanity check.
1906	 */
1907	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1908		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1909		    &ifra.ifra_addr.sin6_addr)) {
1910			if (trylimit-- == 0) {
1911				/*
1912				 * Give up.  Something strange should have
1913				 * happened.
1914				 */
1915				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1916				    "find a unique random IFID\n"));
1917				return (EEXIST);
1918			}
1919			forcegen = 1;
1920			goto again;
1921		}
1922	}
1923
1924	/*
1925	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1926         * public address or TEMP_VALID_LIFETIME.
1927	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1928         * of the public address or TEMP_PREFERRED_LIFETIME -
1929         * DESYNC_FACTOR.
1930	 */
1931	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1932		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1933		    (ia0->ia6_lifetime.ia6t_vltime -
1934		    (time_second - ia0->ia6_updatetime));
1935		if (vltime0 > V_ip6_temp_valid_lifetime)
1936			vltime0 = V_ip6_temp_valid_lifetime;
1937	} else
1938		vltime0 = V_ip6_temp_valid_lifetime;
1939	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1940		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1941		    (ia0->ia6_lifetime.ia6t_pltime -
1942		    (time_second - ia0->ia6_updatetime));
1943		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1944			pltime0 = V_ip6_temp_preferred_lifetime -
1945			    V_ip6_desync_factor;
1946		}
1947	} else
1948		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1949	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1950	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1951
1952	/*
1953	 * A temporary address is created only if this calculated Preferred
1954	 * Lifetime is greater than REGEN_ADVANCE time units.
1955	 */
1956	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1957		return (0);
1958
1959	/* XXX: scope zone ID? */
1960
1961	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1962
1963	/* allocate ifaddr structure, link into chain, etc. */
1964	updateflags = 0;
1965	if (delay)
1966		updateflags |= IN6_IFAUPDATE_DADDELAY;
1967	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1968		return (error);
1969
1970	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1971	if (newia == NULL) {	/* XXX: can it happen? */
1972		nd6log((LOG_ERR,
1973		    "in6_tmpifadd: ifa update succeeded, but we got "
1974		    "no ifaddr\n"));
1975		return (EINVAL); /* XXX */
1976	}
1977	newia->ia6_ndpr = ia0->ia6_ndpr;
1978	newia->ia6_ndpr->ndpr_refcnt++;
1979
1980	/*
1981	 * A newly added address might affect the status of other addresses.
1982	 * XXX: when the temporary address is generated with a new public
1983	 * address, the onlink check is redundant.  However, it would be safe
1984	 * to do the check explicitly everywhere a new address is generated,
1985	 * and, in fact, we surely need the check when we create a new
1986	 * temporary address due to deprecation of an old temporary address.
1987	 */
1988	pfxlist_onlink_check();
1989
1990	return (0);
1991}
1992
1993static int
1994in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1995{
1996	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1997		ndpr->ndpr_preferred = 0;
1998	else
1999		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2000	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2001		ndpr->ndpr_expire = 0;
2002	else
2003		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2004
2005	return 0;
2006}
2007
2008static void
2009in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2010{
2011	/* init ia6t_expire */
2012	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2013		lt6->ia6t_expire = 0;
2014	else {
2015		lt6->ia6t_expire = time_second;
2016		lt6->ia6t_expire += lt6->ia6t_vltime;
2017	}
2018
2019	/* init ia6t_preferred */
2020	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2021		lt6->ia6t_preferred = 0;
2022	else {
2023		lt6->ia6t_preferred = time_second;
2024		lt6->ia6t_preferred += lt6->ia6t_pltime;
2025	}
2026}
2027
2028/*
2029 * Delete all the routing table entries that use the specified gateway.
2030 * XXX: this function causes search through all entries of routing table, so
2031 * it shouldn't be called when acting as a router.
2032 */
2033void
2034rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2035{
2036	INIT_VNET_NET(curvnet);
2037	struct radix_node_head *rnh = V_rt_tables[0][AF_INET6];
2038	int s = splnet();
2039
2040	/* We'll care only link-local addresses */
2041	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2042		splx(s);
2043		return;
2044	}
2045
2046	RADIX_NODE_HEAD_LOCK(rnh);
2047	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2048	RADIX_NODE_HEAD_UNLOCK(rnh);
2049	splx(s);
2050}
2051
2052static int
2053rt6_deleteroute(struct radix_node *rn, void *arg)
2054{
2055#define SIN6(s)	((struct sockaddr_in6 *)s)
2056	struct rtentry *rt = (struct rtentry *)rn;
2057	struct in6_addr *gate = (struct in6_addr *)arg;
2058
2059	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2060		return (0);
2061
2062	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2063		return (0);
2064	}
2065
2066	/*
2067	 * Do not delete a static route.
2068	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2069	 * 'cloned' bit instead?
2070	 */
2071	if ((rt->rt_flags & RTF_STATIC) != 0)
2072		return (0);
2073
2074	/*
2075	 * We delete only host route. This means, in particular, we don't
2076	 * delete default route.
2077	 */
2078	if ((rt->rt_flags & RTF_HOST) == 0)
2079		return (0);
2080
2081	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2082	    rt_mask(rt), rt->rt_flags, 0));
2083#undef SIN6
2084}
2085
2086int
2087nd6_setdefaultiface(int ifindex)
2088{
2089	INIT_VNET_NET(curvnet);
2090	INIT_VNET_INET6(curvnet);
2091	int error = 0;
2092
2093	if (ifindex < 0 || V_if_index < ifindex)
2094		return (EINVAL);
2095	if (ifindex != 0 && !ifnet_byindex(ifindex))
2096		return (EINVAL);
2097
2098	if (V_nd6_defifindex != ifindex) {
2099		V_nd6_defifindex = ifindex;
2100		if (V_nd6_defifindex > 0)
2101			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2102		else
2103			V_nd6_defifp = NULL;
2104
2105		/*
2106		 * Our current implementation assumes one-to-one maping between
2107		 * interfaces and links, so it would be natural to use the
2108		 * default interface as the default link.
2109		 */
2110		scope6_setdefault(V_nd6_defifp);
2111	}
2112
2113	return (error);
2114}
2115