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 __P((struct nd_prefixctl *, struct nd_defrouter *,
73 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
74 static struct nd_pfxrouter *pfxrtr_lookup __P((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 __P((struct nd_prefix *,
85 struct in6_addrlifetime *));
87 static int rt6_deleteroute(struct radix_node *, void *);
89 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
90 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
92 static VNET_DEFINE(struct ifnet *, nd6_defifp);
93 VNET_DEFINE(int, nd6_defifindex);
94 #define V_nd6_defifp VNET(nd6_defifp)
96 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
98 VNET_DEFINE(int, ip6_desync_factor);
99 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
100 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
102 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
104 /* RTPREF_MEDIUM has to be 0! */
105 #define RTPREF_HIGH 1
106 #define RTPREF_MEDIUM 0
107 #define RTPREF_LOW (-1)
108 #define RTPREF_RESERVED (-2)
109 #define RTPREF_INVALID (-3) /* internal */
112 * Receive Router Solicitation Message - just for routers.
113 * Router solicitation/advertisement is mostly managed by userland program
114 * (rtadvd) so here we have no function like nd6_ra_output().
119 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
121 struct ifnet *ifp = m->m_pkthdr.rcvif;
122 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 struct nd_router_solicit *nd_rs;
124 struct in6_addr saddr6 = ip6->ip6_src;
127 union nd_opts ndopts;
128 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
130 /* If I'm not a router, ignore it. */
131 if (V_ip6_accept_rtadv != 0 || V_ip6_forwarding != 1)
135 if (ip6->ip6_hlim != 255) {
137 "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
138 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
139 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
144 * Don't update the neighbor cache, if src = ::.
145 * This indicates that the src has no IP address assigned yet.
147 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
150 #ifndef PULLDOWN_TEST
151 IP6_EXTHDR_CHECK(m, off, icmp6len,);
152 nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
154 IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
156 ICMP6STAT_INC(icp6s_tooshort);
161 icmp6len -= sizeof(*nd_rs);
162 nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
163 if (nd6_options(&ndopts) < 0) {
165 "nd6_rs_input: invalid ND option, ignored\n"));
166 /* nd6_options have incremented stats */
170 if (ndopts.nd_opts_src_lladdr) {
171 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
172 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
175 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
177 "nd6_rs_input: lladdrlen mismatch for %s "
178 "(if %d, RS packet %d)\n",
179 ip6_sprintf(ip6bufs, &saddr6),
180 ifp->if_addrlen, lladdrlen - 2));
184 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
191 ICMP6STAT_INC(icp6s_badrs);
196 * Receive Router Advertisement Message.
199 * TODO: on-link bit on prefix information
200 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
203 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
205 struct ifnet *ifp = m->m_pkthdr.rcvif;
206 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
207 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
208 struct nd_router_advert *nd_ra;
209 struct in6_addr saddr6 = ip6->ip6_src;
211 union nd_opts ndopts;
212 struct nd_defrouter *dr;
213 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
216 * We only accept RAs only when
217 * the system-wide variable allows the acceptance, and
218 * per-interface variable allows RAs on the receiving interface.
220 if (V_ip6_accept_rtadv == 0)
222 if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
225 if (ip6->ip6_hlim != 255) {
227 "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
228 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
229 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
233 if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
235 "nd6_ra_input: src %s is not link-local\n",
236 ip6_sprintf(ip6bufs, &saddr6)));
240 #ifndef PULLDOWN_TEST
241 IP6_EXTHDR_CHECK(m, off, icmp6len,);
242 nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
244 IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
246 ICMP6STAT_INC(icp6s_tooshort);
251 icmp6len -= sizeof(*nd_ra);
252 nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
253 if (nd6_options(&ndopts) < 0) {
255 "nd6_ra_input: invalid ND option, ignored\n"));
256 /* nd6_options have incremented stats */
261 struct nd_defrouter dr0;
262 u_int32_t advreachable = nd_ra->nd_ra_reachable;
264 /* remember if this is a multicasted advertisement */
265 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
268 bzero(&dr0, sizeof(dr0));
270 dr0.flags = nd_ra->nd_ra_flags_reserved;
271 dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
272 dr0.expire = time_second + dr0.rtlifetime;
274 /* unspecified or not? (RFC 2461 6.3.4) */
276 advreachable = ntohl(advreachable);
277 if (advreachable <= MAX_REACHABLE_TIME &&
278 ndi->basereachable != advreachable) {
279 ndi->basereachable = advreachable;
280 ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
281 ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
284 if (nd_ra->nd_ra_retransmit)
285 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
286 if (nd_ra->nd_ra_curhoplimit)
287 ndi->chlim = nd_ra->nd_ra_curhoplimit;
288 dr = defrtrlist_update(&dr0);
294 if (ndopts.nd_opts_pi) {
295 struct nd_opt_hdr *pt;
296 struct nd_opt_prefix_info *pi = NULL;
297 struct nd_prefixctl pr;
299 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
300 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
301 pt = (struct nd_opt_hdr *)((caddr_t)pt +
302 (pt->nd_opt_len << 3))) {
303 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
305 pi = (struct nd_opt_prefix_info *)pt;
307 if (pi->nd_opt_pi_len != 4) {
309 "nd6_ra_input: invalid option "
310 "len %d for prefix information option, "
311 "ignored\n", pi->nd_opt_pi_len));
315 if (128 < pi->nd_opt_pi_prefix_len) {
317 "nd6_ra_input: invalid prefix "
318 "len %d for prefix information option, "
319 "ignored\n", pi->nd_opt_pi_prefix_len));
323 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
324 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
326 "nd6_ra_input: invalid prefix "
329 &pi->nd_opt_pi_prefix)));
333 bzero(&pr, sizeof(pr));
334 pr.ndpr_prefix.sin6_family = AF_INET6;
335 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
336 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
337 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
339 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
340 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
341 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
342 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
343 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
344 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
345 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
346 (void)prelist_update(&pr, dr, m, mcast);
353 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
357 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
360 if (mtu < IPV6_MMTU) {
361 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
362 "mtu=%lu sent from %s, ignoring\n",
363 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
368 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
369 ? ndi->maxmtu : ifp->if_mtu;
371 int change = (ndi->linkmtu != mtu);
374 if (change) /* in6_maxmtu may change */
377 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
378 "mtu=%lu sent from %s; "
379 "exceeds maxmtu %lu, ignoring\n",
380 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
387 * Source link layer address
393 if (ndopts.nd_opts_src_lladdr) {
394 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
395 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
398 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
400 "nd6_ra_input: lladdrlen mismatch for %s "
401 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
402 ifp->if_addrlen, lladdrlen - 2));
406 nd6_cache_lladdr(ifp, &saddr6, lladdr,
407 lladdrlen, ND_ROUTER_ADVERT, 0);
410 * Installing a link-layer address might change the state of the
411 * router's neighbor cache, which might also affect our on-link
412 * detection of adveritsed prefixes.
414 pfxlist_onlink_check();
422 ICMP6STAT_INC(icp6s_badra);
427 * default router list proccessing sub routines
430 /* tell the change to user processes watching the routing socket. */
432 nd6_rtmsg(int cmd, struct rtentry *rt)
434 struct rt_addrinfo info;
438 bzero((caddr_t)&info, sizeof(info));
439 info.rti_info[RTAX_DST] = rt_key(rt);
440 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
441 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
445 ifa = TAILQ_FIRST(&ifp->if_addrhead);
446 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
449 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
453 rt_missmsg(cmd, &info, rt->rt_flags, 0);
459 defrouter_addreq(struct nd_defrouter *new)
461 struct sockaddr_in6 def, mask, gate;
462 struct rtentry *newrt = NULL;
466 bzero(&def, sizeof(def));
467 bzero(&mask, sizeof(mask));
468 bzero(&gate, sizeof(gate));
470 def.sin6_len = mask.sin6_len = gate.sin6_len =
471 sizeof(struct sockaddr_in6);
472 def.sin6_family = gate.sin6_family = AF_INET6;
473 gate.sin6_addr = new->rtaddr;
476 error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
477 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
478 RTF_GATEWAY, &newrt);
480 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
489 struct nd_defrouter *
490 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
492 struct nd_defrouter *dr;
494 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
495 dr = TAILQ_NEXT(dr, dr_entry)) {
496 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
500 return (NULL); /* search failed */
504 * Remove the default route for a given router.
505 * This is just a subroutine function for defrouter_select(), and should
506 * not be called from anywhere else.
509 defrouter_delreq(struct nd_defrouter *dr)
511 struct sockaddr_in6 def, mask, gate;
512 struct rtentry *oldrt = NULL;
514 bzero(&def, sizeof(def));
515 bzero(&mask, sizeof(mask));
516 bzero(&gate, sizeof(gate));
518 def.sin6_len = mask.sin6_len = gate.sin6_len =
519 sizeof(struct sockaddr_in6);
520 def.sin6_family = gate.sin6_family = AF_INET6;
521 gate.sin6_addr = dr->rtaddr;
523 rtrequest(RTM_DELETE, (struct sockaddr *)&def,
524 (struct sockaddr *)&gate,
525 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
527 nd6_rtmsg(RTM_DELETE, oldrt);
535 * remove all default routes from default router list
538 defrouter_reset(void)
540 struct nd_defrouter *dr;
542 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
543 dr = TAILQ_NEXT(dr, dr_entry))
544 defrouter_delreq(dr);
547 * XXX should we also nuke any default routers in the kernel, by
548 * going through them by rtalloc1()?
553 defrtrlist_del(struct nd_defrouter *dr)
555 struct nd_defrouter *deldr = NULL;
556 struct nd_prefix *pr;
559 * Flush all the routing table entries that use the router
562 if (!V_ip6_forwarding && V_ip6_accept_rtadv) /* XXX: better condition? */
563 rt6_flush(&dr->rtaddr, dr->ifp);
567 defrouter_delreq(dr);
569 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
572 * Also delete all the pointers to the router in each prefix lists.
574 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
575 struct nd_pfxrouter *pfxrtr;
576 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
579 pfxlist_onlink_check();
582 * If the router is the primary one, choose a new one.
583 * Note that defrouter_select() will remove the current gateway
584 * from the routing table.
593 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
594 * draft-ietf-ipngwg-router-selection:
595 * 1) Routers that are reachable or probably reachable should be preferred.
596 * If we have more than one (probably) reachable router, prefer ones
597 * with the highest router preference.
598 * 2) When no routers on the list are known to be reachable or
599 * probably reachable, routers SHOULD be selected in a round-robin
600 * fashion, regardless of router preference values.
601 * 3) If the Default Router List is empty, assume that all
602 * destinations are on-link.
604 * We assume nd_defrouter is sorted by router preference value.
605 * Since the code below covers both with and without router preference cases,
606 * we do not need to classify the cases by ifdef.
608 * At this moment, we do not try to install more than one default router,
609 * even when the multipath routing is available, because we're not sure about
610 * the benefits for stub hosts comparing to the risk of making the code
611 * complicated and the possibility of introducing bugs.
614 defrouter_select(void)
617 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
618 struct llentry *ln = NULL;
621 * This function should be called only when acting as an autoconfigured
622 * host. Although the remaining part of this function is not effective
623 * if the node is not an autoconfigured host, we explicitly exclude
624 * such cases here for safety.
626 if (V_ip6_forwarding || !V_ip6_accept_rtadv) {
628 "defrouter_select: called unexpectedly (forwarding=%d, "
629 "accept_rtadv=%d)\n", V_ip6_forwarding, V_ip6_accept_rtadv));
635 * Let's handle easy case (3) first:
636 * If default router list is empty, there's nothing to be done.
638 if (!TAILQ_FIRST(&V_nd_defrouter)) {
644 * Search for a (probably) reachable router from the list.
645 * We just pick up the first reachable one (if any), assuming that
646 * the ordering rule of the list described in defrtrlist_update().
648 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
649 dr = TAILQ_NEXT(dr, dr_entry)) {
650 IF_AFDATA_LOCK(dr->ifp);
651 if (selected_dr == NULL &&
652 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
653 ND6_IS_LLINFO_PROBREACH(ln)) {
656 IF_AFDATA_UNLOCK(dr->ifp);
662 if (dr->installed && installed_dr == NULL)
664 else if (dr->installed && installed_dr) {
665 /* this should not happen. warn for diagnosis. */
666 log(LOG_ERR, "defrouter_select: more than one router"
671 * If none of the default routers was found to be reachable,
672 * round-robin the list regardless of preference.
673 * Otherwise, if we have an installed router, check if the selected
674 * (reachable) router should really be preferred to the installed one.
675 * We only prefer the new router when the old one is not reachable
676 * or when the new one has a really higher preference value.
678 if (selected_dr == NULL) {
679 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
680 selected_dr = TAILQ_FIRST(&V_nd_defrouter);
682 selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
683 } else if (installed_dr) {
684 IF_AFDATA_LOCK(installed_dr->ifp);
685 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
686 ND6_IS_LLINFO_PROBREACH(ln) &&
687 rtpref(selected_dr) <= rtpref(installed_dr)) {
688 selected_dr = installed_dr;
690 IF_AFDATA_UNLOCK(installed_dr->ifp);
696 * If the selected router is different than the installed one,
697 * remove the installed router and install the selected one.
698 * Note that the selected router is never NULL here.
700 if (installed_dr != selected_dr) {
702 defrouter_delreq(installed_dr);
703 defrouter_addreq(selected_dr);
711 * for default router selection
712 * regards router-preference field as a 2-bit signed integer
715 rtpref(struct nd_defrouter *dr)
717 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
718 case ND_RA_FLAG_RTPREF_HIGH:
719 return (RTPREF_HIGH);
720 case ND_RA_FLAG_RTPREF_MEDIUM:
721 case ND_RA_FLAG_RTPREF_RSV:
722 return (RTPREF_MEDIUM);
723 case ND_RA_FLAG_RTPREF_LOW:
727 * This case should never happen. If it did, it would mean a
728 * serious bug of kernel internal. We thus always bark here.
729 * Or, can we even panic?
731 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
732 return (RTPREF_INVALID);
737 static struct nd_defrouter *
738 defrtrlist_update(struct nd_defrouter *new)
740 struct nd_defrouter *dr, *n;
743 if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
745 if (new->rtlifetime == 0) {
749 int oldpref = rtpref(dr);
752 dr->flags = new->flags; /* xxx flag check */
753 dr->rtlifetime = new->rtlifetime;
754 dr->expire = new->expire;
757 * If the preference does not change, there's no need
758 * to sort the entries.
760 if (rtpref(new) == oldpref) {
766 * preferred router may be changed, so relocate
768 * XXX: calling TAILQ_REMOVE directly is a bad manner.
769 * However, since defrtrlist_del() has many side
770 * effects, we intentionally do so here.
771 * defrouter_select() below will handle routing
774 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
782 /* entry does not exist */
783 if (new->rtlifetime == 0) {
788 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
793 bzero(n, sizeof(*n));
798 * Insert the new router in the Default Router List;
799 * The Default Router List should be in the descending order
800 * of router-preferece. Routers with the same preference are
801 * sorted in the arriving time order.
804 /* insert at the end of the group */
805 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
806 dr = TAILQ_NEXT(dr, dr_entry)) {
807 if (rtpref(n) > rtpref(dr))
811 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
813 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
822 static struct nd_pfxrouter *
823 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
825 struct nd_pfxrouter *search;
827 for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
828 if (search->router == dr)
836 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
838 struct nd_pfxrouter *new;
840 new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
843 bzero(new, sizeof(*new));
846 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
848 pfxlist_onlink_check();
852 pfxrtr_del(struct nd_pfxrouter *pfr)
854 LIST_REMOVE(pfr, pfr_entry);
859 nd6_prefix_lookup(struct nd_prefixctl *key)
861 struct nd_prefix *search;
863 for (search = V_nd_prefix.lh_first;
864 search; search = search->ndpr_next) {
865 if (key->ndpr_ifp == search->ndpr_ifp &&
866 key->ndpr_plen == search->ndpr_plen &&
867 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
868 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
877 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
878 struct nd_prefix **newp)
880 struct nd_prefix *new = NULL;
883 char ip6buf[INET6_ADDRSTRLEN];
885 new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
888 bzero(new, sizeof(*new));
889 new->ndpr_ifp = pr->ndpr_ifp;
890 new->ndpr_prefix = pr->ndpr_prefix;
891 new->ndpr_plen = pr->ndpr_plen;
892 new->ndpr_vltime = pr->ndpr_vltime;
893 new->ndpr_pltime = pr->ndpr_pltime;
894 new->ndpr_flags = pr->ndpr_flags;
895 if ((error = in6_init_prefix_ltimes(new)) != 0) {
899 new->ndpr_lastupdate = time_second;
904 LIST_INIT(&new->ndpr_advrtrs);
905 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
906 /* make prefix in the canonical form */
907 for (i = 0; i < 4; i++)
908 new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
909 new->ndpr_mask.s6_addr32[i];
912 /* link ndpr_entry to nd_prefix list */
913 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
916 /* ND_OPT_PI_FLAG_ONLINK processing */
917 if (new->ndpr_raf_onlink) {
920 if ((e = nd6_prefix_onlink(new)) != 0) {
921 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
922 "the prefix %s/%d on-link on %s (errno=%d)\n",
923 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
924 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
925 /* proceed anyway. XXX: is it correct? */
936 prelist_remove(struct nd_prefix *pr)
938 struct nd_pfxrouter *pfr, *next;
940 char ip6buf[INET6_ADDRSTRLEN];
942 /* make sure to invalidate the prefix until it is really freed. */
947 * Though these flags are now meaningless, we'd rather keep the value
948 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
949 * when executing "ndp -p".
952 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
953 (e = nd6_prefix_offlink(pr)) != 0) {
954 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
956 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
957 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
958 /* what should we do? */
961 if (pr->ndpr_refcnt > 0)
962 return; /* notice here? */
966 /* unlink ndpr_entry from nd_prefix list */
967 LIST_REMOVE(pr, ndpr_entry);
969 /* free list of routers that adversed the prefix */
970 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
971 next = pfr->pfr_next;
979 pfxlist_onlink_check();
987 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
988 struct mbuf *m, int mcast)
990 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
992 struct ifnet *ifp = new->ndpr_ifp;
993 struct nd_prefix *pr;
998 struct in6_addrlifetime lt6_tmp;
999 char ip6buf[INET6_ADDRSTRLEN];
1004 * Authenticity for NA consists authentication for
1005 * both IP header and IP datagrams, doesn't it ?
1007 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1008 auth = ((m->m_flags & M_AUTHIPHDR) &&
1009 (m->m_flags & M_AUTHIPDGM));
1013 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1015 * nd6_prefix_lookup() ensures that pr and new have the same
1016 * prefix on a same interface.
1020 * Update prefix information. Note that the on-link (L) bit
1021 * and the autonomous (A) bit should NOT be changed from 1
1024 if (new->ndpr_raf_onlink == 1)
1025 pr->ndpr_raf_onlink = 1;
1026 if (new->ndpr_raf_auto == 1)
1027 pr->ndpr_raf_auto = 1;
1028 if (new->ndpr_raf_onlink) {
1029 pr->ndpr_vltime = new->ndpr_vltime;
1030 pr->ndpr_pltime = new->ndpr_pltime;
1031 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1032 pr->ndpr_lastupdate = time_second;
1035 if (new->ndpr_raf_onlink &&
1036 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1039 if ((e = nd6_prefix_onlink(pr)) != 0) {
1041 "prelist_update: failed to make "
1042 "the prefix %s/%d on-link on %s "
1045 &pr->ndpr_prefix.sin6_addr),
1046 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1047 /* proceed anyway. XXX: is it correct? */
1051 if (dr && pfxrtr_lookup(pr, dr) == NULL)
1054 struct nd_prefix *newpr = NULL;
1058 if (new->ndpr_vltime == 0)
1060 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1063 error = nd6_prelist_add(new, dr, &newpr);
1064 if (error != 0 || newpr == NULL) {
1065 nd6log((LOG_NOTICE, "prelist_update: "
1066 "nd6_prelist_add failed for %s/%d on %s "
1067 "errno=%d, returnpr=%p\n",
1068 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1069 new->ndpr_plen, if_name(new->ndpr_ifp),
1071 goto end; /* we should just give up in this case. */
1075 * XXX: from the ND point of view, we can ignore a prefix
1076 * with the on-link bit being zero. However, we need a
1077 * prefix structure for references from autoconfigured
1078 * addresses. Thus, we explicitly make sure that the prefix
1079 * itself expires now.
1081 if (newpr->ndpr_raf_onlink == 0) {
1082 newpr->ndpr_vltime = 0;
1083 newpr->ndpr_pltime = 0;
1084 in6_init_prefix_ltimes(newpr);
1091 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1092 * Note that pr must be non NULL at this point.
1095 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1096 if (!new->ndpr_raf_auto)
1100 * 5.5.3 (b). the link-local prefix should have been ignored in
1104 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1105 if (new->ndpr_pltime > new->ndpr_vltime) {
1106 error = EINVAL; /* XXX: won't be used */
1111 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1112 * an address configured by stateless autoconfiguration already in the
1113 * list of addresses associated with the interface, and the Valid
1114 * Lifetime is not 0, form an address. We first check if we have
1115 * a matching prefix.
1116 * Note: we apply a clarification in rfc2462bis-02 here. We only
1117 * consider autoconfigured addresses while RFC2462 simply said
1121 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1122 struct in6_ifaddr *ifa6;
1123 u_int32_t remaininglifetime;
1125 if (ifa->ifa_addr->sa_family != AF_INET6)
1128 ifa6 = (struct in6_ifaddr *)ifa;
1131 * We only consider autoconfigured addresses as per rfc2462bis.
1133 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1137 * Spec is not clear here, but I believe we should concentrate
1138 * on unicast (i.e. not anycast) addresses.
1139 * XXX: other ia6_flags? detached or duplicated?
1141 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1145 * Ignore the address if it is not associated with a prefix
1146 * or is associated with a prefix that is different from this
1147 * one. (pr is never NULL here)
1149 if (ifa6->ia6_ndpr != pr)
1152 if (ia6_match == NULL) /* remember the first one */
1156 * An already autoconfigured address matched. Now that we
1157 * are sure there is at least one matched address, we can
1158 * proceed to 5.5.3. (e): update the lifetimes according to the
1159 * "two hours" rule and the privacy extension.
1160 * We apply some clarifications in rfc2462bis:
1161 * - use remaininglifetime instead of storedlifetime as a
1163 * - remove the dead code in the "two-hour" rule
1165 #define TWOHOUR (120*60)
1166 lt6_tmp = ifa6->ia6_lifetime;
1168 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1169 remaininglifetime = ND6_INFINITE_LIFETIME;
1170 else if (time_second - ifa6->ia6_updatetime >
1171 lt6_tmp.ia6t_vltime) {
1173 * The case of "invalid" address. We should usually
1174 * not see this case.
1176 remaininglifetime = 0;
1178 remaininglifetime = lt6_tmp.ia6t_vltime -
1179 (time_second - ifa6->ia6_updatetime);
1181 /* when not updating, keep the current stored lifetime. */
1182 lt6_tmp.ia6t_vltime = remaininglifetime;
1184 if (TWOHOUR < new->ndpr_vltime ||
1185 remaininglifetime < new->ndpr_vltime) {
1186 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1187 } else if (remaininglifetime <= TWOHOUR) {
1189 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1193 * new->ndpr_vltime <= TWOHOUR &&
1194 * TWOHOUR < remaininglifetime
1196 lt6_tmp.ia6t_vltime = TWOHOUR;
1199 /* The 2 hour rule is not imposed for preferred lifetime. */
1200 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1202 in6_init_address_ltimes(pr, <6_tmp);
1205 * We need to treat lifetimes for temporary addresses
1206 * differently, according to
1207 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1208 * we only update the lifetimes when they are in the maximum
1211 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1212 u_int32_t maxvltime, maxpltime;
1214 if (V_ip6_temp_valid_lifetime >
1215 (u_int32_t)((time_second - ifa6->ia6_createtime) +
1216 V_ip6_desync_factor)) {
1217 maxvltime = V_ip6_temp_valid_lifetime -
1218 (time_second - ifa6->ia6_createtime) -
1219 V_ip6_desync_factor;
1222 if (V_ip6_temp_preferred_lifetime >
1223 (u_int32_t)((time_second - ifa6->ia6_createtime) +
1224 V_ip6_desync_factor)) {
1225 maxpltime = V_ip6_temp_preferred_lifetime -
1226 (time_second - ifa6->ia6_createtime) -
1227 V_ip6_desync_factor;
1231 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1232 lt6_tmp.ia6t_vltime > maxvltime) {
1233 lt6_tmp.ia6t_vltime = maxvltime;
1235 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1236 lt6_tmp.ia6t_pltime > maxpltime) {
1237 lt6_tmp.ia6t_pltime = maxpltime;
1240 ifa6->ia6_lifetime = lt6_tmp;
1241 ifa6->ia6_updatetime = time_second;
1243 IF_ADDR_UNLOCK(ifp);
1244 if (ia6_match == NULL && new->ndpr_vltime) {
1248 * 5.5.3 (d) (continued)
1249 * No address matched and the valid lifetime is non-zero.
1250 * Create a new address.
1254 * Prefix Length check:
1255 * If the sum of the prefix length and interface identifier
1256 * length does not equal 128 bits, the Prefix Information
1257 * option MUST be ignored. The length of the interface
1258 * identifier is defined in a separate link-type specific
1261 ifidlen = in6_if2idlen(ifp);
1263 /* this should not happen, so we always log it. */
1264 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1268 if (ifidlen + pr->ndpr_plen != 128) {
1270 "prelist_update: invalid prefixlen "
1271 "%d for %s, ignored\n",
1272 pr->ndpr_plen, if_name(ifp)));
1276 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1278 * note that we should use pr (not new) for reference.
1285 * When a new public address is created as described
1286 * in RFC2462, also create a new temporary address.
1289 * When an interface connects to a new link, a new
1290 * randomized interface identifier should be generated
1291 * immediately together with a new set of temporary
1292 * addresses. Thus, we specifiy 1 as the 2nd arg of
1295 if (V_ip6_use_tempaddr) {
1297 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1298 nd6log((LOG_NOTICE, "prelist_update: "
1299 "failed to create a temporary "
1300 "address, errno=%d\n",
1304 ifa_free(&ia6->ia_ifa);
1307 * A newly added address might affect the status
1308 * of other addresses, so we check and update it.
1309 * XXX: what if address duplication happens?
1311 pfxlist_onlink_check();
1313 /* just set an error. do not bark here. */
1314 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1324 * A supplement function used in the on-link detection below;
1325 * detect if a given prefix has a (probably) reachable advertising router.
1326 * XXX: lengthy function name...
1328 static struct nd_pfxrouter *
1329 find_pfxlist_reachable_router(struct nd_prefix *pr)
1331 struct nd_pfxrouter *pfxrtr;
1335 for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr != NULL;
1336 pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1337 IF_AFDATA_LOCK(pfxrtr->router->ifp);
1338 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1339 IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1342 canreach = ND6_IS_LLINFO_PROBREACH(ln);
1351 * Check if each prefix in the prefix list has at least one available router
1352 * that advertised the prefix (a router is "available" if its neighbor cache
1353 * entry is reachable or probably reachable).
1354 * If the check fails, the prefix may be off-link, because, for example,
1355 * we have moved from the network but the lifetime of the prefix has not
1356 * expired yet. So we should not use the prefix if there is another prefix
1357 * that has an available router.
1358 * But, if there is no prefix that has an available router, we still regards
1359 * all the prefixes as on-link. This is because we can't tell if all the
1360 * routers are simply dead or if we really moved from the network and there
1361 * is no router around us.
1364 pfxlist_onlink_check()
1366 struct nd_prefix *pr;
1367 struct in6_ifaddr *ifa;
1368 struct nd_defrouter *dr;
1369 struct nd_pfxrouter *pfxrtr = NULL;
1372 * Check if there is a prefix that has a reachable advertising
1375 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1376 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1381 * If we have no such prefix, check whether we still have a router
1382 * that does not advertise any prefixes.
1385 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1386 dr = TAILQ_NEXT(dr, dr_entry)) {
1387 struct nd_prefix *pr0;
1389 for (pr0 = V_nd_prefix.lh_first; pr0;
1390 pr0 = pr0->ndpr_next) {
1391 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1398 if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1400 * There is at least one prefix that has a reachable router,
1401 * or at least a router which probably does not advertise
1402 * any prefixes. The latter would be the case when we move
1403 * to a new link where we have a router that does not provide
1404 * prefixes and we configure an address by hand.
1405 * Detach prefixes which have no reachable advertising
1406 * router, and attach other prefixes.
1408 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1409 /* XXX: a link-local prefix should never be detached */
1410 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1414 * we aren't interested in prefixes without the L bit
1417 if (pr->ndpr_raf_onlink == 0)
1420 if (pr->ndpr_raf_auto == 0)
1423 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1424 find_pfxlist_reachable_router(pr) == NULL)
1425 pr->ndpr_stateflags |= NDPRF_DETACHED;
1426 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1427 find_pfxlist_reachable_router(pr) != 0)
1428 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1431 /* there is no prefix that has a reachable router */
1432 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1433 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1436 if (pr->ndpr_raf_onlink == 0)
1439 if (pr->ndpr_raf_auto == 0)
1442 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1443 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1448 * Remove each interface route associated with a (just) detached
1449 * prefix, and reinstall the interface route for a (just) attached
1450 * prefix. Note that all attempt of reinstallation does not
1451 * necessarily success, when a same prefix is shared among multiple
1452 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1453 * so we don't have to care about them.
1455 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1457 char ip6buf[INET6_ADDRSTRLEN];
1459 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1462 if (pr->ndpr_raf_onlink == 0)
1465 if (pr->ndpr_raf_auto == 0)
1468 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1469 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1470 if ((e = nd6_prefix_offlink(pr)) != 0) {
1472 "pfxlist_onlink_check: failed to "
1473 "make %s/%d offlink, errno=%d\n",
1475 &pr->ndpr_prefix.sin6_addr),
1479 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1480 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1481 pr->ndpr_raf_onlink) {
1482 if ((e = nd6_prefix_onlink(pr)) != 0) {
1484 "pfxlist_onlink_check: failed to "
1485 "make %s/%d onlink, errno=%d\n",
1487 &pr->ndpr_prefix.sin6_addr),
1494 * Changes on the prefix status might affect address status as well.
1495 * Make sure that all addresses derived from an attached prefix are
1496 * attached, and that all addresses derived from a detached prefix are
1497 * detached. Note, however, that a manually configured address should
1498 * always be attached.
1499 * The precise detection logic is same as the one for prefixes.
1501 * XXXRW: in6_ifaddrhead locking.
1503 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1504 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1507 if (ifa->ia6_ndpr == NULL) {
1509 * This can happen when we first configure the address
1510 * (i.e. the address exists, but the prefix does not).
1511 * XXX: complicated relationships...
1516 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1520 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1521 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1524 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1527 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1528 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1529 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1530 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1531 nd6_dad_start((struct ifaddr *)ifa, 0);
1534 ifa->ia6_flags |= IN6_IFF_DETACHED;
1539 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1540 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1543 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1544 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1545 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1546 /* Do we need a delay in this case? */
1547 nd6_dad_start((struct ifaddr *)ifa, 0);
1554 nd6_prefix_onlink(struct nd_prefix *pr)
1557 struct ifnet *ifp = pr->ndpr_ifp;
1558 struct sockaddr_in6 mask6;
1559 struct nd_prefix *opr;
1562 struct radix_node_head *rnh;
1563 struct rtentry *rt = NULL;
1564 char ip6buf[INET6_ADDRSTRLEN];
1565 struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1568 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1570 "nd6_prefix_onlink: %s/%d is already on-link\n",
1571 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1577 * Add the interface route associated with the prefix. Before
1578 * installing the route, check if there's the same prefix on another
1579 * interface, and the prefix has already installed the interface route.
1580 * Although such a configuration is expected to be rare, we explicitly
1583 for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1587 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1590 if (opr->ndpr_plen == pr->ndpr_plen &&
1591 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1592 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1597 * We prefer link-local addresses as the associated interface address.
1599 /* search for a link-local addr */
1600 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1601 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1603 /* XXX: freebsd does not have ifa_ifwithaf */
1605 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1606 if (ifa->ifa_addr->sa_family == AF_INET6)
1611 IF_ADDR_UNLOCK(ifp);
1612 /* should we care about ia6_flags? */
1616 * This can still happen, when, for example, we receive an RA
1617 * containing a prefix with the L bit set and the A bit clear,
1618 * after removing all IPv6 addresses on the receiving
1619 * interface. This should, of course, be rare though.
1622 "nd6_prefix_onlink: failed to find any ifaddr"
1623 " to add route for a prefix(%s/%d) on %s\n",
1624 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1625 pr->ndpr_plen, if_name(ifp)));
1630 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1631 * ifa->ifa_rtrequest = nd6_rtrequest;
1633 bzero(&mask6, sizeof(mask6));
1634 mask6.sin6_len = sizeof(mask6);
1635 mask6.sin6_addr = pr->ndpr_mask;
1636 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1637 error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1638 ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1640 if (rt != NULL) /* this should be non NULL, though */ {
1641 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1642 /* XXX what if rhn == NULL? */
1643 RADIX_NODE_HEAD_LOCK(rnh);
1645 if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1646 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1647 rt->rt_ifp->if_type;
1648 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1649 rt->rt_ifp->if_index;
1651 RADIX_NODE_HEAD_UNLOCK(rnh);
1652 nd6_rtmsg(RTM_ADD, rt);
1655 pr->ndpr_stateflags |= NDPRF_ONLINK;
1657 char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1658 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1659 " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1661 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1662 pr->ndpr_plen, if_name(ifp),
1663 ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1664 ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1679 nd6_prefix_offlink(struct nd_prefix *pr)
1682 struct ifnet *ifp = pr->ndpr_ifp;
1683 struct nd_prefix *opr;
1684 struct sockaddr_in6 sa6, mask6;
1685 struct rtentry *rt = NULL;
1686 char ip6buf[INET6_ADDRSTRLEN];
1689 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1691 "nd6_prefix_offlink: %s/%d is already off-link\n",
1692 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1697 bzero(&sa6, sizeof(sa6));
1698 sa6.sin6_family = AF_INET6;
1699 sa6.sin6_len = sizeof(sa6);
1700 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1701 sizeof(struct in6_addr));
1702 bzero(&mask6, sizeof(mask6));
1703 mask6.sin6_family = AF_INET6;
1704 mask6.sin6_len = sizeof(sa6);
1705 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1706 error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1707 (struct sockaddr *)&mask6, 0, &rt);
1709 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1711 /* report the route deletion to the routing socket. */
1713 nd6_rtmsg(RTM_DELETE, rt);
1716 * There might be the same prefix on another interface,
1717 * the prefix which could not be on-link just because we have
1718 * the interface route (see comments in nd6_prefix_onlink).
1719 * If there's one, try to make the prefix on-link on the
1722 for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1726 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1730 * KAME specific: detached prefixes should not be
1733 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1736 if (opr->ndpr_plen == pr->ndpr_plen &&
1737 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1738 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1741 if ((e = nd6_prefix_onlink(opr)) != 0) {
1743 "nd6_prefix_offlink: failed to "
1744 "recover a prefix %s/%d from %s "
1745 "to %s (errno = %d)\n",
1747 &opr->ndpr_prefix.sin6_addr),
1748 opr->ndpr_plen, if_name(ifp),
1749 if_name(opr->ndpr_ifp), e));
1754 /* XXX: can we still set the NDPRF_ONLINK flag? */
1756 "nd6_prefix_offlink: failed to delete route: "
1757 "%s/%d on %s (errno = %d)\n",
1758 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1759 if_name(ifp), error));
1769 static struct in6_ifaddr *
1770 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1772 struct ifnet *ifp = pr->ndpr_ifp;
1774 struct in6_aliasreq ifra;
1775 struct in6_ifaddr *ia, *ib;
1777 struct in6_addr mask;
1778 int prefixlen = pr->ndpr_plen;
1780 char ip6buf[INET6_ADDRSTRLEN];
1782 in6_prefixlen2mask(&mask, prefixlen);
1785 * find a link-local address (will be interface ID).
1786 * Is it really mandatory? Theoretically, a global or a site-local
1787 * address can be configured without a link-local address, if we
1788 * have a unique interface identifier...
1790 * it is not mandatory to have a link-local address, we can generate
1791 * interface identifier on the fly. we do this because:
1792 * (1) it should be the easiest way to find interface identifier.
1793 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1794 * for multiple addresses on a single interface, and possible shortcut
1795 * of DAD. we omitted DAD for this reason in the past.
1796 * (3) a user can prevent autoconfiguration of global address
1797 * by removing link-local address by hand (this is partly because we
1798 * don't have other way to control the use of IPv6 on an interface.
1799 * this has been our design choice - cf. NRL's "ifconfig auto").
1800 * (4) it is easier to manage when an interface has addresses
1801 * with the same interface identifier, than to have multiple addresses
1802 * with different interface identifiers.
1804 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1806 ib = (struct in6_ifaddr *)ifa;
1810 /* prefixlen + ifidlen must be equal to 128 */
1811 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1812 if (prefixlen != plen0) {
1814 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1815 "(prefix=%d ifid=%d)\n",
1816 if_name(ifp), prefixlen, 128 - plen0));
1822 bzero(&ifra, sizeof(ifra));
1824 * in6_update_ifa() does not use ifra_name, but we accurately set it
1827 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1828 ifra.ifra_addr.sin6_family = AF_INET6;
1829 ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1831 ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1832 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1833 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1834 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1835 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1838 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1839 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1840 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1841 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1842 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1843 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1844 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1845 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1848 /* new prefix mask. */
1849 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1850 ifra.ifra_prefixmask.sin6_family = AF_INET6;
1851 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1852 sizeof(ifra.ifra_prefixmask.sin6_addr));
1855 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1856 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1858 /* XXX: scope zone ID? */
1860 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1863 * Make sure that we do not have this address already. This should
1864 * usually not happen, but we can still see this case, e.g., if we
1865 * have manually configured the exact address to be configured.
1867 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1868 &ifra.ifra_addr.sin6_addr);
1871 /* this should be rare enough to make an explicit log */
1872 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1873 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1878 * Allocate ifaddr structure, link into chain, etc.
1879 * If we are going to create a new address upon receiving a multicasted
1880 * RA, we need to impose a random delay before starting DAD.
1881 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1885 updateflags |= IN6_IFAUPDATE_DADDELAY;
1886 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1888 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1889 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1890 if_name(ifp), error));
1891 return (NULL); /* ifaddr must not have been allocated. */
1894 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1896 * XXXRW: Assumption of non-NULLness here might not be true with
1897 * fine-grained locking -- should we validate it? Or just return
1898 * earlier ifa rather than looking it up again?
1900 return (ia); /* this is always non-NULL and referenced. */
1904 * ia0 - corresponding public address
1907 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1909 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1910 struct in6_ifaddr *newia, *ia;
1911 struct in6_aliasreq ifra;
1913 int trylimit = 3; /* XXX: adhoc value */
1915 u_int32_t randid[2];
1916 time_t vltime0, pltime0;
1918 bzero(&ifra, sizeof(ifra));
1919 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1920 ifra.ifra_addr = ia0->ia_addr;
1921 /* copy prefix mask */
1922 ifra.ifra_prefixmask = ia0->ia_prefixmask;
1923 /* clear the old IFID */
1924 for (i = 0; i < 4; i++) {
1925 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1926 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1930 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1931 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1932 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1936 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1937 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1938 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1939 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1942 * in6_get_tmpifid() quite likely provided a unique interface ID.
1943 * However, we may still have a chance to see collision, because
1944 * there may be a time lag between generation of the ID and generation
1945 * of the address. So, we'll do one more sanity check.
1948 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1949 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1950 &ifra.ifra_addr.sin6_addr)) {
1951 if (trylimit-- == 0) {
1952 IN6_IFADDR_RUNLOCK();
1954 * Give up. Something strange should have
1957 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1958 "find a unique random IFID\n"));
1961 IN6_IFADDR_RUNLOCK();
1966 IN6_IFADDR_RUNLOCK();
1969 * The Valid Lifetime is the lower of the Valid Lifetime of the
1970 * public address or TEMP_VALID_LIFETIME.
1971 * The Preferred Lifetime is the lower of the Preferred Lifetime
1972 * of the public address or TEMP_PREFERRED_LIFETIME -
1975 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1976 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1977 (ia0->ia6_lifetime.ia6t_vltime -
1978 (time_second - ia0->ia6_updatetime));
1979 if (vltime0 > V_ip6_temp_valid_lifetime)
1980 vltime0 = V_ip6_temp_valid_lifetime;
1982 vltime0 = V_ip6_temp_valid_lifetime;
1983 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1984 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1985 (ia0->ia6_lifetime.ia6t_pltime -
1986 (time_second - ia0->ia6_updatetime));
1987 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1988 pltime0 = V_ip6_temp_preferred_lifetime -
1989 V_ip6_desync_factor;
1992 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1993 ifra.ifra_lifetime.ia6t_vltime = vltime0;
1994 ifra.ifra_lifetime.ia6t_pltime = pltime0;
1997 * A temporary address is created only if this calculated Preferred
1998 * Lifetime is greater than REGEN_ADVANCE time units.
2000 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2003 /* XXX: scope zone ID? */
2005 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2007 /* allocate ifaddr structure, link into chain, etc. */
2010 updateflags |= IN6_IFAUPDATE_DADDELAY;
2011 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2014 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2015 if (newia == NULL) { /* XXX: can it happen? */
2017 "in6_tmpifadd: ifa update succeeded, but we got "
2019 return (EINVAL); /* XXX */
2021 newia->ia6_ndpr = ia0->ia6_ndpr;
2022 newia->ia6_ndpr->ndpr_refcnt++;
2023 ifa_free(&newia->ia_ifa);
2026 * A newly added address might affect the status of other addresses.
2027 * XXX: when the temporary address is generated with a new public
2028 * address, the onlink check is redundant. However, it would be safe
2029 * to do the check explicitly everywhere a new address is generated,
2030 * and, in fact, we surely need the check when we create a new
2031 * temporary address due to deprecation of an old temporary address.
2033 pfxlist_onlink_check();
2039 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2041 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2042 ndpr->ndpr_preferred = 0;
2044 ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2045 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2046 ndpr->ndpr_expire = 0;
2048 ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2054 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2056 /* init ia6t_expire */
2057 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2058 lt6->ia6t_expire = 0;
2060 lt6->ia6t_expire = time_second;
2061 lt6->ia6t_expire += lt6->ia6t_vltime;
2064 /* init ia6t_preferred */
2065 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2066 lt6->ia6t_preferred = 0;
2068 lt6->ia6t_preferred = time_second;
2069 lt6->ia6t_preferred += lt6->ia6t_pltime;
2074 * Delete all the routing table entries that use the specified gateway.
2075 * XXX: this function causes search through all entries of routing table, so
2076 * it shouldn't be called when acting as a router.
2079 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2081 struct radix_node_head *rnh;
2084 /* We'll care only link-local addresses */
2085 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2090 rnh = rt_tables_get_rnh(0, AF_INET6);
2094 RADIX_NODE_HEAD_LOCK(rnh);
2095 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2096 RADIX_NODE_HEAD_UNLOCK(rnh);
2101 rt6_deleteroute(struct radix_node *rn, void *arg)
2103 #define SIN6(s) ((struct sockaddr_in6 *)s)
2104 struct rtentry *rt = (struct rtentry *)rn;
2105 struct in6_addr *gate = (struct in6_addr *)arg;
2107 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2110 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2115 * Do not delete a static route.
2116 * XXX: this seems to be a bit ad-hoc. Should we consider the
2117 * 'cloned' bit instead?
2119 if ((rt->rt_flags & RTF_STATIC) != 0)
2123 * We delete only host route. This means, in particular, we don't
2124 * delete default route.
2126 if ((rt->rt_flags & RTF_HOST) == 0)
2129 return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2130 rt_mask(rt), rt->rt_flags, 0));
2135 nd6_setdefaultiface(int ifindex)
2139 if (ifindex < 0 || V_if_index < ifindex)
2141 if (ifindex != 0 && !ifnet_byindex(ifindex))
2144 if (V_nd6_defifindex != ifindex) {
2145 V_nd6_defifindex = ifindex;
2146 if (V_nd6_defifindex > 0)
2147 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2149 V_nd6_defifp = NULL;
2152 * Our current implementation assumes one-to-one maping between
2153 * interfaces and links, so it would be natural to use the
2154 * default interface as the default link.
2156 scope6_setdefault(V_nd6_defifp);