2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 * $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
45 #include <sys/kernel.h>
47 #include <sys/errno.h>
48 #include <sys/rwlock.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
53 #include <net/if_types.h>
54 #include <net/if_dl.h>
55 #include <net/route.h>
56 #include <net/radix.h>
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>
69 static int rtpref(struct nd_defrouter *);
70 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
71 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
73 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
74 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
75 struct nd_defrouter *);
76 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
77 static void pfxrtr_del(struct nd_pfxrouter *);
78 static struct nd_pfxrouter *find_pfxlist_reachable_router
80 static void defrouter_delreq(struct nd_defrouter *);
81 static void nd6_rtmsg(int, struct rtentry *);
83 static int in6_init_prefix_ltimes(struct nd_prefix *);
84 static void in6_init_address_ltimes(struct nd_prefix *,
85 struct in6_addrlifetime *);
87 static int nd6_prefix_onlink(struct nd_prefix *);
88 static int nd6_prefix_offlink(struct nd_prefix *);
90 static int rt6_deleteroute(struct radix_node *, void *);
92 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
93 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
95 static VNET_DEFINE(struct ifnet *, nd6_defifp);
96 VNET_DEFINE(int, nd6_defifindex);
97 #define V_nd6_defifp VNET(nd6_defifp)
99 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
101 VNET_DEFINE(int, ip6_desync_factor);
102 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
103 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
105 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
107 /* RTPREF_MEDIUM has to be 0! */
108 #define RTPREF_HIGH 1
109 #define RTPREF_MEDIUM 0
110 #define RTPREF_LOW (-1)
111 #define RTPREF_RESERVED (-2)
112 #define RTPREF_INVALID (-3) /* internal */
115 * Receive Router Solicitation Message - just for routers.
116 * Router solicitation/advertisement is mostly managed by userland program
117 * (rtadvd) so here we have no function like nd6_ra_output().
122 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
124 struct ifnet *ifp = m->m_pkthdr.rcvif;
125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
126 struct nd_router_solicit *nd_rs;
127 struct in6_addr saddr6 = ip6->ip6_src;
130 union nd_opts ndopts;
131 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
134 * Accept RS only when V_ip6_forwarding=1 and the interface has
135 * no ND6_IFF_ACCEPT_RTADV.
137 if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
141 if (ip6->ip6_hlim != 255) {
143 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
144 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
145 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
150 * Don't update the neighbor cache, if src = ::.
151 * This indicates that the src has no IP address assigned yet.
153 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
156 #ifndef PULLDOWN_TEST
157 IP6_EXTHDR_CHECK(m, off, icmp6len,);
158 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
160 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
162 ICMP6STAT_INC(icp6s_tooshort);
167 icmp6len -= sizeof(*nd_rs);
168 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
169 if (nd6_options(&ndopts) < 0) {
171 "nd6_rs_input: invalid ND option, ignored\n"));
172 /* nd6_options have incremented stats */
176 if (ndopts.nd_opts_src_lladdr) {
177 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
178 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
181 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
183 "nd6_rs_input: lladdrlen mismatch for %s "
184 "(if %d, RS packet %d)\n",
185 ip6_sprintf(ip6bufs, &saddr6),
186 ifp->if_addrlen, lladdrlen - 2));
190 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
197 ICMP6STAT_INC(icp6s_badrs);
202 * Receive Router Advertisement Message.
205 * TODO: on-link bit on prefix information
206 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
209 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
211 struct ifnet *ifp = m->m_pkthdr.rcvif;
212 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
213 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
214 struct nd_router_advert *nd_ra;
215 struct in6_addr saddr6 = ip6->ip6_src;
217 union nd_opts ndopts;
218 struct nd_defrouter *dr;
219 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
222 * We only accept RAs only when the per-interface flag
223 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
225 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
228 if (ip6->ip6_hlim != 255) {
230 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
231 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
232 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
236 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
238 "nd6_ra_input: src %s is not link-local\n",
239 ip6_sprintf(ip6bufs, &saddr6)));
243 #ifndef PULLDOWN_TEST
244 IP6_EXTHDR_CHECK(m, off, icmp6len,);
245 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
247 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
249 ICMP6STAT_INC(icp6s_tooshort);
254 icmp6len -= sizeof(*nd_ra);
255 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
256 if (nd6_options(&ndopts) < 0) {
258 "nd6_ra_input: invalid ND option, ignored\n"));
259 /* nd6_options have incremented stats */
264 struct nd_defrouter dr0;
265 u_int32_t advreachable = nd_ra->nd_ra_reachable;
267 /* remember if this is a multicasted advertisement */
268 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
271 bzero(&dr0, sizeof(dr0));
273 dr0.flags = nd_ra->nd_ra_flags_reserved;
275 * Effectively-disable routes from RA messages when
276 * ND6_IFF_NO_RADR enabled on the receiving interface or
277 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
279 if (ndi->flags & ND6_IFF_NO_RADR)
281 else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
284 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
285 dr0.expire = time_uptime + dr0.rtlifetime;
287 /* unspecified or not? (RFC 2461 6.3.4) */
289 advreachable = ntohl(advreachable);
290 if (advreachable <= MAX_REACHABLE_TIME &&
291 ndi->basereachable != advreachable) {
292 ndi->basereachable = advreachable;
293 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
294 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
297 if (nd_ra->nd_ra_retransmit)
298 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
299 if (nd_ra->nd_ra_curhoplimit)
300 ndi->chlim = nd_ra->nd_ra_curhoplimit;
301 dr = defrtrlist_update(&dr0);
307 if (ndopts.nd_opts_pi) {
308 struct nd_opt_hdr *pt;
309 struct nd_opt_prefix_info *pi = NULL;
310 struct nd_prefixctl pr;
312 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
313 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
314 pt = (struct nd_opt_hdr *)((caddr_t)pt +
315 (pt->nd_opt_len << 3))) {
316 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
318 pi = (struct nd_opt_prefix_info *)pt;
320 if (pi->nd_opt_pi_len != 4) {
322 "nd6_ra_input: invalid option "
323 "len %d for prefix information option, "
324 "ignored\n", pi->nd_opt_pi_len));
328 if (128 < pi->nd_opt_pi_prefix_len) {
330 "nd6_ra_input: invalid prefix "
331 "len %d for prefix information option, "
332 "ignored\n", pi->nd_opt_pi_prefix_len));
336 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
337 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
339 "nd6_ra_input: invalid prefix "
342 &pi->nd_opt_pi_prefix)));
346 bzero(&pr, sizeof(pr));
347 pr.ndpr_prefix.sin6_family = AF_INET6;
348 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
349 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
350 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
352 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
353 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
354 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
355 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
356 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
357 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
358 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
359 (void)prelist_update(&pr, dr, m, mcast);
366 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
370 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
373 if (mtu < IPV6_MMTU) {
374 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
375 "mtu=%lu sent from %s, ignoring\n",
376 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
381 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
382 ? ndi->maxmtu : ifp->if_mtu;
384 int change = (ndi->linkmtu != mtu);
387 if (change) /* in6_maxmtu may change */
390 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
391 "mtu=%lu sent from %s; "
392 "exceeds maxmtu %lu, ignoring\n",
393 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
400 * Source link layer address
406 if (ndopts.nd_opts_src_lladdr) {
407 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
408 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
411 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
413 "nd6_ra_input: lladdrlen mismatch for %s "
414 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
415 ifp->if_addrlen, lladdrlen - 2));
419 nd6_cache_lladdr(ifp, &saddr6, lladdr,
420 lladdrlen, ND_ROUTER_ADVERT, 0);
423 * Installing a link-layer address might change the state of the
424 * router's neighbor cache, which might also affect our on-link
425 * detection of adveritsed prefixes.
427 pfxlist_onlink_check();
435 ICMP6STAT_INC(icp6s_badra);
440 * default router list proccessing sub routines
443 /* tell the change to user processes watching the routing socket. */
445 nd6_rtmsg(int cmd, struct rtentry *rt)
447 struct rt_addrinfo info;
451 bzero((caddr_t)&info, sizeof(info));
452 info.rti_info[RTAX_DST] = rt_key(rt);
453 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
454 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
458 ifa = TAILQ_FIRST(&ifp->if_addrhead);
459 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
461 IF_ADDR_RUNLOCK(ifp);
462 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
466 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
472 defrouter_addreq(struct nd_defrouter *new)
474 struct sockaddr_in6 def, mask, gate;
475 struct rtentry *newrt = NULL;
478 bzero(&def, sizeof(def));
479 bzero(&mask, sizeof(mask));
480 bzero(&gate, sizeof(gate));
482 def.sin6_len = mask.sin6_len = gate.sin6_len =
483 sizeof(struct sockaddr_in6);
484 def.sin6_family = gate.sin6_family = AF_INET6;
485 gate.sin6_addr = new->rtaddr;
487 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
488 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
489 RTF_GATEWAY, &newrt, RT_DEFAULT_FIB);
491 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
499 struct nd_defrouter *
500 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
502 struct nd_defrouter *dr;
504 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
505 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
509 return (NULL); /* search failed */
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.
518 defrouter_delreq(struct nd_defrouter *dr)
520 struct sockaddr_in6 def, mask, gate;
521 struct rtentry *oldrt = NULL;
523 bzero(&def, sizeof(def));
524 bzero(&mask, sizeof(mask));
525 bzero(&gate, sizeof(gate));
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;
532 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
533 (struct sockaddr *)&gate,
534 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB);
536 nd6_rtmsg(RTM_DELETE, oldrt);
544 * remove all default routes from default router list
547 defrouter_reset(void)
549 struct nd_defrouter *dr;
551 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
552 defrouter_delreq(dr);
555 * XXX should we also nuke any default routers in the kernel, by
556 * going through them by rtalloc1()?
561 defrtrlist_del(struct nd_defrouter *dr)
563 struct nd_defrouter *deldr = NULL;
564 struct nd_prefix *pr;
567 * Flush all the routing table entries that use the router
570 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
571 rt6_flush(&dr->rtaddr, dr->ifp);
575 defrouter_delreq(dr);
577 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
580 * Also delete all the pointers to the router in each prefix lists.
582 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
583 struct nd_pfxrouter *pfxrtr;
584 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
587 pfxlist_onlink_check();
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.
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.
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.
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.
622 defrouter_select(void)
624 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
625 struct llentry *ln = NULL;
628 * Let's handle easy case (3) first:
629 * If default router list is empty, there's nothing to be done.
631 if (TAILQ_EMPTY(&V_nd_defrouter))
635 * Search for a (probably) reachable router from the list.
636 * We just pick up the first reachable one (if any), assuming that
637 * the ordering rule of the list described in defrtrlist_update().
639 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
640 IF_AFDATA_RLOCK(dr->ifp);
641 if (selected_dr == NULL &&
642 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
643 ND6_IS_LLINFO_PROBREACH(ln)) {
646 IF_AFDATA_RUNLOCK(dr->ifp);
652 if (dr->installed && installed_dr == NULL)
654 else if (dr->installed && installed_dr) {
655 /* this should not happen. warn for diagnosis. */
656 log(LOG_ERR, "defrouter_select: more than one router"
661 * If none of the default routers was found to be reachable,
662 * round-robin the list regardless of preference.
663 * Otherwise, if we have an installed router, check if the selected
664 * (reachable) router should really be preferred to the installed one.
665 * We only prefer the new router when the old one is not reachable
666 * or when the new one has a really higher preference value.
668 if (selected_dr == NULL) {
669 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
670 selected_dr = TAILQ_FIRST(&V_nd_defrouter);
672 selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
673 } else if (installed_dr) {
674 IF_AFDATA_RLOCK(installed_dr->ifp);
675 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
676 ND6_IS_LLINFO_PROBREACH(ln) &&
677 rtpref(selected_dr) <= rtpref(installed_dr)) {
678 selected_dr = installed_dr;
680 IF_AFDATA_RUNLOCK(installed_dr->ifp);
686 * If the selected router is different than the installed one,
687 * remove the installed router and install the selected one.
688 * Note that the selected router is never NULL here.
690 if (installed_dr != selected_dr) {
692 defrouter_delreq(installed_dr);
693 defrouter_addreq(selected_dr);
700 * for default router selection
701 * regards router-preference field as a 2-bit signed integer
704 rtpref(struct nd_defrouter *dr)
706 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
707 case ND_RA_FLAG_RTPREF_HIGH:
708 return (RTPREF_HIGH);
709 case ND_RA_FLAG_RTPREF_MEDIUM:
710 case ND_RA_FLAG_RTPREF_RSV:
711 return (RTPREF_MEDIUM);
712 case ND_RA_FLAG_RTPREF_LOW:
716 * This case should never happen. If it did, it would mean a
717 * serious bug of kernel internal. We thus always bark here.
718 * Or, can we even panic?
720 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
721 return (RTPREF_INVALID);
726 static struct nd_defrouter *
727 defrtrlist_update(struct nd_defrouter *new)
729 struct nd_defrouter *dr, *n;
731 if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
733 if (new->rtlifetime == 0) {
737 int oldpref = rtpref(dr);
740 dr->flags = new->flags; /* xxx flag check */
741 dr->rtlifetime = new->rtlifetime;
742 dr->expire = new->expire;
745 * If the preference does not change, there's no need
746 * to sort the entries. Also make sure the selected
747 * router is still installed in the kernel.
749 if (dr->installed && rtpref(new) == oldpref)
753 * preferred router may be changed, so relocate
755 * XXX: calling TAILQ_REMOVE directly is a bad manner.
756 * However, since defrtrlist_del() has many side
757 * effects, we intentionally do so here.
758 * defrouter_select() below will handle routing
761 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
768 /* entry does not exist */
769 if (new->rtlifetime == 0)
772 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
775 bzero(n, sizeof(*n));
780 * Insert the new router in the Default Router List;
781 * The Default Router List should be in the descending order
782 * of router-preferece. Routers with the same preference are
783 * sorted in the arriving time order.
786 /* insert at the end of the group */
787 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
788 if (rtpref(n) > rtpref(dr))
792 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
794 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
801 static struct nd_pfxrouter *
802 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
804 struct nd_pfxrouter *search;
806 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
807 if (search->router == dr)
815 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
817 struct nd_pfxrouter *new;
819 new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
822 bzero(new, sizeof(*new));
825 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
827 pfxlist_onlink_check();
831 pfxrtr_del(struct nd_pfxrouter *pfr)
833 LIST_REMOVE(pfr, pfr_entry);
838 nd6_prefix_lookup(struct nd_prefixctl *key)
840 struct nd_prefix *search;
842 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
843 if (key->ndpr_ifp == search->ndpr_ifp &&
844 key->ndpr_plen == search->ndpr_plen &&
845 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
846 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
855 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
856 struct nd_prefix **newp)
858 struct nd_prefix *new = NULL;
861 char ip6buf[INET6_ADDRSTRLEN];
863 new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
866 bzero(new, sizeof(*new));
867 new->ndpr_ifp = pr->ndpr_ifp;
868 new->ndpr_prefix = pr->ndpr_prefix;
869 new->ndpr_plen = pr->ndpr_plen;
870 new->ndpr_vltime = pr->ndpr_vltime;
871 new->ndpr_pltime = pr->ndpr_pltime;
872 new->ndpr_flags = pr->ndpr_flags;
873 if ((error = in6_init_prefix_ltimes(new)) != 0) {
877 new->ndpr_lastupdate = time_uptime;
882 LIST_INIT(&new->ndpr_advrtrs);
883 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
884 /* make prefix in the canonical form */
885 for (i = 0; i < 4; i++)
886 new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
887 new->ndpr_mask.s6_addr32[i];
889 /* link ndpr_entry to nd_prefix list */
890 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
892 /* ND_OPT_PI_FLAG_ONLINK processing */
893 if (new->ndpr_raf_onlink) {
896 if ((e = nd6_prefix_onlink(new)) != 0) {
897 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
898 "the prefix %s/%d on-link on %s (errno=%d)\n",
899 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
900 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
901 /* proceed anyway. XXX: is it correct? */
912 prelist_remove(struct nd_prefix *pr)
914 struct nd_pfxrouter *pfr, *next;
916 char ip6buf[INET6_ADDRSTRLEN];
918 /* make sure to invalidate the prefix until it is really freed. */
923 * Though these flags are now meaningless, we'd rather keep the value
924 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
925 * when executing "ndp -p".
928 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
929 (e = nd6_prefix_offlink(pr)) != 0) {
930 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
932 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
933 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
934 /* what should we do? */
937 if (pr->ndpr_refcnt > 0)
938 return; /* notice here? */
940 /* unlink ndpr_entry from nd_prefix list */
941 LIST_REMOVE(pr, ndpr_entry);
943 /* free list of routers that adversed the prefix */
944 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
949 pfxlist_onlink_check();
957 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
958 struct mbuf *m, int mcast)
960 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
962 struct ifnet *ifp = new->ndpr_ifp;
963 struct nd_prefix *pr;
967 struct in6_addrlifetime lt6_tmp;
968 char ip6buf[INET6_ADDRSTRLEN];
973 * Authenticity for NA consists authentication for
974 * both IP header and IP datagrams, doesn't it ?
976 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
977 auth = ((m->m_flags & M_AUTHIPHDR) &&
978 (m->m_flags & M_AUTHIPDGM));
982 if ((pr = nd6_prefix_lookup(new)) != NULL) {
984 * nd6_prefix_lookup() ensures that pr and new have the same
985 * prefix on a same interface.
989 * Update prefix information. Note that the on-link (L) bit
990 * and the autonomous (A) bit should NOT be changed from 1
993 if (new->ndpr_raf_onlink == 1)
994 pr->ndpr_raf_onlink = 1;
995 if (new->ndpr_raf_auto == 1)
996 pr->ndpr_raf_auto = 1;
997 if (new->ndpr_raf_onlink) {
998 pr->ndpr_vltime = new->ndpr_vltime;
999 pr->ndpr_pltime = new->ndpr_pltime;
1000 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1001 pr->ndpr_lastupdate = time_uptime;
1004 if (new->ndpr_raf_onlink &&
1005 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1008 if ((e = nd6_prefix_onlink(pr)) != 0) {
1010 "prelist_update: failed to make "
1011 "the prefix %s/%d on-link on %s "
1014 &pr->ndpr_prefix.sin6_addr),
1015 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1016 /* proceed anyway. XXX: is it correct? */
1020 if (dr && pfxrtr_lookup(pr, dr) == NULL)
1023 struct nd_prefix *newpr = NULL;
1027 if (new->ndpr_vltime == 0)
1029 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1032 error = nd6_prelist_add(new, dr, &newpr);
1033 if (error != 0 || newpr == NULL) {
1034 nd6log((LOG_NOTICE, "prelist_update: "
1035 "nd6_prelist_add failed for %s/%d on %s "
1036 "errno=%d, returnpr=%p\n",
1037 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1038 new->ndpr_plen, if_name(new->ndpr_ifp),
1040 goto end; /* we should just give up in this case. */
1044 * XXX: from the ND point of view, we can ignore a prefix
1045 * with the on-link bit being zero. However, we need a
1046 * prefix structure for references from autoconfigured
1047 * addresses. Thus, we explicitly make sure that the prefix
1048 * itself expires now.
1050 if (newpr->ndpr_raf_onlink == 0) {
1051 newpr->ndpr_vltime = 0;
1052 newpr->ndpr_pltime = 0;
1053 in6_init_prefix_ltimes(newpr);
1060 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1061 * Note that pr must be non NULL at this point.
1064 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1065 if (!new->ndpr_raf_auto)
1069 * 5.5.3 (b). the link-local prefix should have been ignored in
1073 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1074 if (new->ndpr_pltime > new->ndpr_vltime) {
1075 error = EINVAL; /* XXX: won't be used */
1080 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1081 * an address configured by stateless autoconfiguration already in the
1082 * list of addresses associated with the interface, and the Valid
1083 * Lifetime is not 0, form an address. We first check if we have
1084 * a matching prefix.
1085 * Note: we apply a clarification in rfc2462bis-02 here. We only
1086 * consider autoconfigured addresses while RFC2462 simply said
1090 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1091 struct in6_ifaddr *ifa6;
1092 u_int32_t remaininglifetime;
1094 if (ifa->ifa_addr->sa_family != AF_INET6)
1097 ifa6 = (struct in6_ifaddr *)ifa;
1100 * We only consider autoconfigured addresses as per rfc2462bis.
1102 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1106 * Spec is not clear here, but I believe we should concentrate
1107 * on unicast (i.e. not anycast) addresses.
1108 * XXX: other ia6_flags? detached or duplicated?
1110 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1114 * Ignore the address if it is not associated with a prefix
1115 * or is associated with a prefix that is different from this
1116 * one. (pr is never NULL here)
1118 if (ifa6->ia6_ndpr != pr)
1121 if (ia6_match == NULL) /* remember the first one */
1125 * An already autoconfigured address matched. Now that we
1126 * are sure there is at least one matched address, we can
1127 * proceed to 5.5.3. (e): update the lifetimes according to the
1128 * "two hours" rule and the privacy extension.
1129 * We apply some clarifications in rfc2462bis:
1130 * - use remaininglifetime instead of storedlifetime as a
1132 * - remove the dead code in the "two-hour" rule
1134 #define TWOHOUR (120*60)
1135 lt6_tmp = ifa6->ia6_lifetime;
1137 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1138 remaininglifetime = ND6_INFINITE_LIFETIME;
1139 else if (time_uptime - ifa6->ia6_updatetime >
1140 lt6_tmp.ia6t_vltime) {
1142 * The case of "invalid" address. We should usually
1143 * not see this case.
1145 remaininglifetime = 0;
1147 remaininglifetime = lt6_tmp.ia6t_vltime -
1148 (time_uptime - ifa6->ia6_updatetime);
1150 /* when not updating, keep the current stored lifetime. */
1151 lt6_tmp.ia6t_vltime = remaininglifetime;
1153 if (TWOHOUR < new->ndpr_vltime ||
1154 remaininglifetime < new->ndpr_vltime) {
1155 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1156 } else if (remaininglifetime <= TWOHOUR) {
1158 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1162 * new->ndpr_vltime <= TWOHOUR &&
1163 * TWOHOUR < remaininglifetime
1165 lt6_tmp.ia6t_vltime = TWOHOUR;
1168 /* The 2 hour rule is not imposed for preferred lifetime. */
1169 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1171 in6_init_address_ltimes(pr, <6_tmp);
1174 * We need to treat lifetimes for temporary addresses
1175 * differently, according to
1176 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1177 * we only update the lifetimes when they are in the maximum
1180 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1181 u_int32_t maxvltime, maxpltime;
1183 if (V_ip6_temp_valid_lifetime >
1184 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1185 V_ip6_desync_factor)) {
1186 maxvltime = V_ip6_temp_valid_lifetime -
1187 (time_uptime - ifa6->ia6_createtime) -
1188 V_ip6_desync_factor;
1191 if (V_ip6_temp_preferred_lifetime >
1192 (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1193 V_ip6_desync_factor)) {
1194 maxpltime = V_ip6_temp_preferred_lifetime -
1195 (time_uptime - ifa6->ia6_createtime) -
1196 V_ip6_desync_factor;
1200 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1201 lt6_tmp.ia6t_vltime > maxvltime) {
1202 lt6_tmp.ia6t_vltime = maxvltime;
1204 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1205 lt6_tmp.ia6t_pltime > maxpltime) {
1206 lt6_tmp.ia6t_pltime = maxpltime;
1209 ifa6->ia6_lifetime = lt6_tmp;
1210 ifa6->ia6_updatetime = time_uptime;
1212 IF_ADDR_RUNLOCK(ifp);
1213 if (ia6_match == NULL && new->ndpr_vltime) {
1217 * 5.5.3 (d) (continued)
1218 * No address matched and the valid lifetime is non-zero.
1219 * Create a new address.
1223 * Prefix Length check:
1224 * If the sum of the prefix length and interface identifier
1225 * length does not equal 128 bits, the Prefix Information
1226 * option MUST be ignored. The length of the interface
1227 * identifier is defined in a separate link-type specific
1230 ifidlen = in6_if2idlen(ifp);
1232 /* this should not happen, so we always log it. */
1233 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1237 if (ifidlen + pr->ndpr_plen != 128) {
1239 "prelist_update: invalid prefixlen "
1240 "%d for %s, ignored\n",
1241 pr->ndpr_plen, if_name(ifp)));
1245 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1247 * note that we should use pr (not new) for reference.
1254 * When a new public address is created as described
1255 * in RFC2462, also create a new temporary address.
1258 * When an interface connects to a new link, a new
1259 * randomized interface identifier should be generated
1260 * immediately together with a new set of temporary
1261 * addresses. Thus, we specifiy 1 as the 2nd arg of
1264 if (V_ip6_use_tempaddr) {
1266 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1267 nd6log((LOG_NOTICE, "prelist_update: "
1268 "failed to create a temporary "
1269 "address, errno=%d\n",
1273 ifa_free(&ia6->ia_ifa);
1276 * A newly added address might affect the status
1277 * of other addresses, so we check and update it.
1278 * XXX: what if address duplication happens?
1280 pfxlist_onlink_check();
1282 /* just set an error. do not bark here. */
1283 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1292 * A supplement function used in the on-link detection below;
1293 * detect if a given prefix has a (probably) reachable advertising router.
1294 * XXX: lengthy function name...
1296 static struct nd_pfxrouter *
1297 find_pfxlist_reachable_router(struct nd_prefix *pr)
1299 struct nd_pfxrouter *pfxrtr;
1303 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1304 IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1305 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1306 IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1309 canreach = ND6_IS_LLINFO_PROBREACH(ln);
1318 * Check if each prefix in the prefix list has at least one available router
1319 * that advertised the prefix (a router is "available" if its neighbor cache
1320 * entry is reachable or probably reachable).
1321 * If the check fails, the prefix may be off-link, because, for example,
1322 * we have moved from the network but the lifetime of the prefix has not
1323 * expired yet. So we should not use the prefix if there is another prefix
1324 * that has an available router.
1325 * But, if there is no prefix that has an available router, we still regards
1326 * all the prefixes as on-link. This is because we can't tell if all the
1327 * routers are simply dead or if we really moved from the network and there
1328 * is no router around us.
1331 pfxlist_onlink_check()
1333 struct nd_prefix *pr;
1334 struct in6_ifaddr *ifa;
1335 struct nd_defrouter *dr;
1336 struct nd_pfxrouter *pfxrtr = NULL;
1339 * Check if there is a prefix that has a reachable advertising
1342 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1343 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1348 * If we have no such prefix, check whether we still have a router
1349 * that does not advertise any prefixes.
1352 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1353 struct nd_prefix *pr0;
1355 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1356 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1363 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1365 * There is at least one prefix that has a reachable router,
1366 * or at least a router which probably does not advertise
1367 * any prefixes. The latter would be the case when we move
1368 * to a new link where we have a router that does not provide
1369 * prefixes and we configure an address by hand.
1370 * Detach prefixes which have no reachable advertising
1371 * router, and attach other prefixes.
1373 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1374 /* XXX: a link-local prefix should never be detached */
1375 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1379 * we aren't interested in prefixes without the L bit
1382 if (pr->ndpr_raf_onlink == 0)
1385 if (pr->ndpr_raf_auto == 0)
1388 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1389 find_pfxlist_reachable_router(pr) == NULL)
1390 pr->ndpr_stateflags |= NDPRF_DETACHED;
1391 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1392 find_pfxlist_reachable_router(pr) != 0)
1393 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1396 /* there is no prefix that has a reachable router */
1397 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1398 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1401 if (pr->ndpr_raf_onlink == 0)
1404 if (pr->ndpr_raf_auto == 0)
1407 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1408 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1413 * Remove each interface route associated with a (just) detached
1414 * prefix, and reinstall the interface route for a (just) attached
1415 * prefix. Note that all attempt of reinstallation does not
1416 * necessarily success, when a same prefix is shared among multiple
1417 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1418 * so we don't have to care about them.
1420 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1422 char ip6buf[INET6_ADDRSTRLEN];
1424 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1427 if (pr->ndpr_raf_onlink == 0)
1430 if (pr->ndpr_raf_auto == 0)
1433 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1434 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1435 if ((e = nd6_prefix_offlink(pr)) != 0) {
1437 "pfxlist_onlink_check: failed to "
1438 "make %s/%d offlink, errno=%d\n",
1440 &pr->ndpr_prefix.sin6_addr),
1444 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1445 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1446 pr->ndpr_raf_onlink) {
1447 if ((e = nd6_prefix_onlink(pr)) != 0) {
1449 "pfxlist_onlink_check: failed to "
1450 "make %s/%d onlink, errno=%d\n",
1452 &pr->ndpr_prefix.sin6_addr),
1459 * Changes on the prefix status might affect address status as well.
1460 * Make sure that all addresses derived from an attached prefix are
1461 * attached, and that all addresses derived from a detached prefix are
1462 * detached. Note, however, that a manually configured address should
1463 * always be attached.
1464 * The precise detection logic is same as the one for prefixes.
1466 * XXXRW: in6_ifaddrhead locking.
1468 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1469 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1472 if (ifa->ia6_ndpr == NULL) {
1474 * This can happen when we first configure the address
1475 * (i.e. the address exists, but the prefix does not).
1476 * XXX: complicated relationships...
1481 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1485 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1486 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1489 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1492 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1493 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1494 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1495 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1496 nd6_dad_start((struct ifaddr *)ifa, 0);
1499 ifa->ia6_flags |= IN6_IFF_DETACHED;
1504 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1505 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1508 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1509 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1510 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1511 /* Do we need a delay in this case? */
1512 nd6_dad_start((struct ifaddr *)ifa, 0);
1519 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1521 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1522 struct radix_node_head *rnh;
1524 struct sockaddr_in6 mask6;
1526 int error, a_failure, fibnum;
1529 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1530 * ifa->ifa_rtrequest = nd6_rtrequest;
1532 bzero(&mask6, sizeof(mask6));
1533 mask6.sin6_len = sizeof(mask6);
1534 mask6.sin6_addr = pr->ndpr_mask;
1535 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1538 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1541 error = in6_rtrequest(RTM_ADD,
1542 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1543 (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1545 KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1546 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1549 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1550 /* XXX what if rhn == NULL? */
1551 RADIX_NODE_HEAD_LOCK(rnh);
1553 if (rt_setgate(rt, rt_key(rt),
1554 (struct sockaddr *)&null_sdl) == 0) {
1555 struct sockaddr_dl *dl;
1557 dl = (struct sockaddr_dl *)rt->rt_gateway;
1558 dl->sdl_type = rt->rt_ifp->if_type;
1559 dl->sdl_index = rt->rt_ifp->if_index;
1561 RADIX_NODE_HEAD_UNLOCK(rnh);
1562 nd6_rtmsg(RTM_ADD, rt);
1564 pr->ndpr_stateflags |= NDPRF_ONLINK;
1566 char ip6buf[INET6_ADDRSTRLEN];
1567 char ip6bufg[INET6_ADDRSTRLEN];
1568 char ip6bufm[INET6_ADDRSTRLEN];
1569 struct sockaddr_in6 *sin6;
1571 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1572 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1573 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1574 "flags=%lx errno = %d\n",
1575 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1576 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1577 ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1578 ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1581 /* Save last error to return, see rtinit(). */
1592 /* Return the last error we got. */
1597 nd6_prefix_onlink(struct nd_prefix *pr)
1600 struct ifnet *ifp = pr->ndpr_ifp;
1601 struct nd_prefix *opr;
1603 char ip6buf[INET6_ADDRSTRLEN];
1606 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1608 "nd6_prefix_onlink: %s/%d is already on-link\n",
1609 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1615 * Add the interface route associated with the prefix. Before
1616 * installing the route, check if there's the same prefix on another
1617 * interface, and the prefix has already installed the interface route.
1618 * Although such a configuration is expected to be rare, we explicitly
1621 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1625 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1628 if (opr->ndpr_plen == pr->ndpr_plen &&
1629 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1630 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1635 * We prefer link-local addresses as the associated interface address.
1637 /* search for a link-local addr */
1638 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1639 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1641 /* XXX: freebsd does not have ifa_ifwithaf */
1643 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1644 if (ifa->ifa_addr->sa_family == AF_INET6)
1649 IF_ADDR_RUNLOCK(ifp);
1650 /* should we care about ia6_flags? */
1654 * This can still happen, when, for example, we receive an RA
1655 * containing a prefix with the L bit set and the A bit clear,
1656 * after removing all IPv6 addresses on the receiving
1657 * interface. This should, of course, be rare though.
1660 "nd6_prefix_onlink: failed to find any ifaddr"
1661 " to add route for a prefix(%s/%d) on %s\n",
1662 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1663 pr->ndpr_plen, if_name(ifp)));
1667 error = nd6_prefix_onlink_rtrequest(pr, ifa);
1676 nd6_prefix_offlink(struct nd_prefix *pr)
1679 struct ifnet *ifp = pr->ndpr_ifp;
1680 struct nd_prefix *opr;
1681 struct sockaddr_in6 sa6, mask6;
1683 char ip6buf[INET6_ADDRSTRLEN];
1684 int fibnum, a_failure;
1687 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1689 "nd6_prefix_offlink: %s/%d is already off-link\n",
1690 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1695 bzero(&sa6, sizeof(sa6));
1696 sa6.sin6_family = AF_INET6;
1697 sa6.sin6_len = sizeof(sa6);
1698 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1699 sizeof(struct in6_addr));
1700 bzero(&mask6, sizeof(mask6));
1701 mask6.sin6_family = AF_INET6;
1702 mask6.sin6_len = sizeof(sa6);
1703 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1706 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1708 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1709 (struct sockaddr *)&mask6, 0, &rt, fibnum);
1711 /* report the route deletion to the routing socket. */
1713 nd6_rtmsg(RTM_DELETE, rt);
1715 /* Save last error to return, see rtinit(). */
1725 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1728 * There might be the same prefix on another interface,
1729 * the prefix which could not be on-link just because we have
1730 * the interface route (see comments in nd6_prefix_onlink).
1731 * If there's one, try to make the prefix on-link on the
1734 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1738 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1742 * KAME specific: detached prefixes should not be
1745 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1748 if (opr->ndpr_plen == pr->ndpr_plen &&
1749 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1750 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1753 if ((e = nd6_prefix_onlink(opr)) != 0) {
1755 "nd6_prefix_offlink: failed to "
1756 "recover a prefix %s/%d from %s "
1757 "to %s (errno = %d)\n",
1759 &opr->ndpr_prefix.sin6_addr),
1760 opr->ndpr_plen, if_name(ifp),
1761 if_name(opr->ndpr_ifp), e));
1767 /* XXX: can we still set the NDPRF_ONLINK flag? */
1769 "nd6_prefix_offlink: failed to delete route: "
1770 "%s/%d on %s (errno = %d)\n",
1771 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1772 if_name(ifp), error));
1776 lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6,
1777 (struct sockaddr *)&mask6, LLE_STATIC);
1782 static struct in6_ifaddr *
1783 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1785 struct ifnet *ifp = pr->ndpr_ifp;
1787 struct in6_aliasreq ifra;
1788 struct in6_ifaddr *ia, *ib;
1790 struct in6_addr mask;
1791 int prefixlen = pr->ndpr_plen;
1793 char ip6buf[INET6_ADDRSTRLEN];
1795 in6_prefixlen2mask(&mask, prefixlen);
1798 * find a link-local address (will be interface ID).
1799 * Is it really mandatory? Theoretically, a global or a site-local
1800 * address can be configured without a link-local address, if we
1801 * have a unique interface identifier...
1803 * it is not mandatory to have a link-local address, we can generate
1804 * interface identifier on the fly. we do this because:
1805 * (1) it should be the easiest way to find interface identifier.
1806 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1807 * for multiple addresses on a single interface, and possible shortcut
1808 * of DAD. we omitted DAD for this reason in the past.
1809 * (3) a user can prevent autoconfiguration of global address
1810 * by removing link-local address by hand (this is partly because we
1811 * don't have other way to control the use of IPv6 on an interface.
1812 * this has been our design choice - cf. NRL's "ifconfig auto").
1813 * (4) it is easier to manage when an interface has addresses
1814 * with the same interface identifier, than to have multiple addresses
1815 * with different interface identifiers.
1817 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1819 ib = (struct in6_ifaddr *)ifa;
1823 /* prefixlen + ifidlen must be equal to 128 */
1824 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1825 if (prefixlen != plen0) {
1827 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1828 "(prefix=%d ifid=%d)\n",
1829 if_name(ifp), prefixlen, 128 - plen0));
1835 bzero(&ifra, sizeof(ifra));
1837 * in6_update_ifa() does not use ifra_name, but we accurately set it
1840 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1841 ifra.ifra_addr.sin6_family = AF_INET6;
1842 ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1844 ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1845 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1846 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1847 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1848 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1851 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1852 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1853 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1854 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1855 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1856 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1857 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1858 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1861 /* new prefix mask. */
1862 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1863 ifra.ifra_prefixmask.sin6_family = AF_INET6;
1864 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1865 sizeof(ifra.ifra_prefixmask.sin6_addr));
1868 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1869 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1871 /* XXX: scope zone ID? */
1873 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1876 * Make sure that we do not have this address already. This should
1877 * usually not happen, but we can still see this case, e.g., if we
1878 * have manually configured the exact address to be configured.
1880 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1881 &ifra.ifra_addr.sin6_addr);
1884 /* this should be rare enough to make an explicit log */
1885 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1886 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1891 * Allocate ifaddr structure, link into chain, etc.
1892 * If we are going to create a new address upon receiving a multicasted
1893 * RA, we need to impose a random delay before starting DAD.
1894 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1898 updateflags |= IN6_IFAUPDATE_DADDELAY;
1899 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1901 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1902 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1903 if_name(ifp), error));
1904 return (NULL); /* ifaddr must not have been allocated. */
1907 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1909 * XXXRW: Assumption of non-NULLness here might not be true with
1910 * fine-grained locking -- should we validate it? Or just return
1911 * earlier ifa rather than looking it up again?
1913 return (ia); /* this is always non-NULL and referenced. */
1917 * ia0 - corresponding public address
1920 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1922 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1923 struct in6_ifaddr *newia, *ia;
1924 struct in6_aliasreq ifra;
1926 int trylimit = 3; /* XXX: adhoc value */
1928 u_int32_t randid[2];
1929 time_t vltime0, pltime0;
1931 bzero(&ifra, sizeof(ifra));
1932 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1933 ifra.ifra_addr = ia0->ia_addr;
1934 /* copy prefix mask */
1935 ifra.ifra_prefixmask = ia0->ia_prefixmask;
1936 /* clear the old IFID */
1937 for (i = 0; i < 4; i++) {
1938 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1939 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1943 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1944 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1945 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1949 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1950 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1951 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1952 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1955 * in6_get_tmpifid() quite likely provided a unique interface ID.
1956 * However, we may still have a chance to see collision, because
1957 * there may be a time lag between generation of the ID and generation
1958 * of the address. So, we'll do one more sanity check.
1961 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1962 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1963 &ifra.ifra_addr.sin6_addr)) {
1964 if (trylimit-- == 0) {
1965 IN6_IFADDR_RUNLOCK();
1967 * Give up. Something strange should have
1970 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1971 "find a unique random IFID\n"));
1974 IN6_IFADDR_RUNLOCK();
1979 IN6_IFADDR_RUNLOCK();
1982 * The Valid Lifetime is the lower of the Valid Lifetime of the
1983 * public address or TEMP_VALID_LIFETIME.
1984 * The Preferred Lifetime is the lower of the Preferred Lifetime
1985 * of the public address or TEMP_PREFERRED_LIFETIME -
1988 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1989 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1990 (ia0->ia6_lifetime.ia6t_vltime -
1991 (time_uptime - ia0->ia6_updatetime));
1992 if (vltime0 > V_ip6_temp_valid_lifetime)
1993 vltime0 = V_ip6_temp_valid_lifetime;
1995 vltime0 = V_ip6_temp_valid_lifetime;
1996 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1997 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1998 (ia0->ia6_lifetime.ia6t_pltime -
1999 (time_uptime - ia0->ia6_updatetime));
2000 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2001 pltime0 = V_ip6_temp_preferred_lifetime -
2002 V_ip6_desync_factor;
2005 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2006 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2007 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2010 * A temporary address is created only if this calculated Preferred
2011 * Lifetime is greater than REGEN_ADVANCE time units.
2013 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2016 /* XXX: scope zone ID? */
2018 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2020 /* allocate ifaddr structure, link into chain, etc. */
2023 updateflags |= IN6_IFAUPDATE_DADDELAY;
2024 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2027 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2028 if (newia == NULL) { /* XXX: can it happen? */
2030 "in6_tmpifadd: ifa update succeeded, but we got "
2032 return (EINVAL); /* XXX */
2034 newia->ia6_ndpr = ia0->ia6_ndpr;
2035 newia->ia6_ndpr->ndpr_refcnt++;
2036 ifa_free(&newia->ia_ifa);
2039 * A newly added address might affect the status of other addresses.
2040 * XXX: when the temporary address is generated with a new public
2041 * address, the onlink check is redundant. However, it would be safe
2042 * to do the check explicitly everywhere a new address is generated,
2043 * and, in fact, we surely need the check when we create a new
2044 * temporary address due to deprecation of an old temporary address.
2046 pfxlist_onlink_check();
2052 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2054 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2055 ndpr->ndpr_preferred = 0;
2057 ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2058 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2059 ndpr->ndpr_expire = 0;
2061 ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2067 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2069 /* init ia6t_expire */
2070 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2071 lt6->ia6t_expire = 0;
2073 lt6->ia6t_expire = time_uptime;
2074 lt6->ia6t_expire += lt6->ia6t_vltime;
2077 /* init ia6t_preferred */
2078 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2079 lt6->ia6t_preferred = 0;
2081 lt6->ia6t_preferred = time_uptime;
2082 lt6->ia6t_preferred += lt6->ia6t_pltime;
2087 * Delete all the routing table entries that use the specified gateway.
2088 * XXX: this function causes search through all entries of routing table, so
2089 * it shouldn't be called when acting as a router.
2092 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2094 struct radix_node_head *rnh;
2097 /* We'll care only link-local addresses */
2098 if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2101 /* XXX Do we really need to walk any but the default FIB? */
2102 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
2103 rnh = rt_tables_get_rnh(fibnum, AF_INET6);
2107 RADIX_NODE_HEAD_LOCK(rnh);
2108 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2109 RADIX_NODE_HEAD_UNLOCK(rnh);
2114 rt6_deleteroute(struct radix_node *rn, void *arg)
2116 #define SIN6(s) ((struct sockaddr_in6 *)s)
2117 struct rtentry *rt = (struct rtentry *)rn;
2118 struct in6_addr *gate = (struct in6_addr *)arg;
2120 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2123 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2128 * Do not delete a static route.
2129 * XXX: this seems to be a bit ad-hoc. Should we consider the
2130 * 'cloned' bit instead?
2132 if ((rt->rt_flags & RTF_STATIC) != 0)
2136 * We delete only host route. This means, in particular, we don't
2137 * delete default route.
2139 if ((rt->rt_flags & RTF_HOST) == 0)
2142 return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2143 rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum));
2148 nd6_setdefaultiface(int ifindex)
2152 if (ifindex < 0 || V_if_index < ifindex)
2154 if (ifindex != 0 && !ifnet_byindex(ifindex))
2157 if (V_nd6_defifindex != ifindex) {
2158 V_nd6_defifindex = ifindex;
2159 if (V_nd6_defifindex > 0)
2160 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2162 V_nd6_defifp = NULL;
2165 * Our current implementation assumes one-to-one maping between
2166 * interfaces and links, so it would be natural to use the
2167 * default interface as the default link.
2169 scope6_setdefault(V_nd6_defifp);