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_second + 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 if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
301 ndi->chlim = nd_ra->nd_ra_curhoplimit;
302 else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
303 log(LOG_ERR, "RA with a lower CurHopLimit sent from "
304 "%s on %s (current = %d, received = %d). "
305 "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
306 if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
309 dr = defrtrlist_update(&dr0);
315 if (ndopts.nd_opts_pi) {
316 struct nd_opt_hdr *pt;
317 struct nd_opt_prefix_info *pi = NULL;
318 struct nd_prefixctl pr;
320 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
321 pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
322 pt = (struct nd_opt_hdr *)((caddr_t)pt +
323 (pt->nd_opt_len << 3))) {
324 if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
326 pi = (struct nd_opt_prefix_info *)pt;
328 if (pi->nd_opt_pi_len != 4) {
330 "nd6_ra_input: invalid option "
331 "len %d for prefix information option, "
332 "ignored\n", pi->nd_opt_pi_len));
336 if (128 < pi->nd_opt_pi_prefix_len) {
338 "nd6_ra_input: invalid prefix "
339 "len %d for prefix information option, "
340 "ignored\n", pi->nd_opt_pi_prefix_len));
344 if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
345 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
347 "nd6_ra_input: invalid prefix "
350 &pi->nd_opt_pi_prefix)));
354 bzero(&pr, sizeof(pr));
355 pr.ndpr_prefix.sin6_family = AF_INET6;
356 pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
357 pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
358 pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
360 pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
361 ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
362 pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
363 ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
364 pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
365 pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
366 pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
367 (void)prelist_update(&pr, dr, m, mcast);
374 if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
378 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
381 if (mtu < IPV6_MMTU) {
382 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
383 "mtu=%lu sent from %s, ignoring\n",
384 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
389 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
390 ? ndi->maxmtu : ifp->if_mtu;
392 int change = (ndi->linkmtu != mtu);
395 if (change) /* in6_maxmtu may change */
398 nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
399 "mtu=%lu sent from %s; "
400 "exceeds maxmtu %lu, ignoring\n",
401 mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
408 * Source link layer address
414 if (ndopts.nd_opts_src_lladdr) {
415 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
416 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
419 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
421 "nd6_ra_input: lladdrlen mismatch for %s "
422 "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
423 ifp->if_addrlen, lladdrlen - 2));
427 nd6_cache_lladdr(ifp, &saddr6, lladdr,
428 lladdrlen, ND_ROUTER_ADVERT, 0);
431 * Installing a link-layer address might change the state of the
432 * router's neighbor cache, which might also affect our on-link
433 * detection of adveritsed prefixes.
435 pfxlist_onlink_check();
443 ICMP6STAT_INC(icp6s_badra);
448 * default router list proccessing sub routines
451 /* tell the change to user processes watching the routing socket. */
453 nd6_rtmsg(int cmd, struct rtentry *rt)
455 struct rt_addrinfo info;
459 bzero((caddr_t)&info, sizeof(info));
460 info.rti_info[RTAX_DST] = rt_key(rt);
461 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
462 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
466 ifa = TAILQ_FIRST(&ifp->if_addrhead);
467 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
469 IF_ADDR_RUNLOCK(ifp);
470 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
474 rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
480 defrouter_addreq(struct nd_defrouter *new)
482 struct sockaddr_in6 def, mask, gate;
483 struct rtentry *newrt = NULL;
487 bzero(&def, sizeof(def));
488 bzero(&mask, sizeof(mask));
489 bzero(&gate, sizeof(gate));
491 def.sin6_len = mask.sin6_len = gate.sin6_len =
492 sizeof(struct sockaddr_in6);
493 def.sin6_family = gate.sin6_family = AF_INET6;
494 gate.sin6_addr = new->rtaddr;
497 error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
498 (struct sockaddr *)&gate, (struct sockaddr *)&mask,
499 RTF_GATEWAY, &newrt, RT_DEFAULT_FIB);
501 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
510 struct nd_defrouter *
511 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
513 struct nd_defrouter *dr;
515 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
516 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
520 return (NULL); /* search failed */
524 * Remove the default route for a given router.
525 * This is just a subroutine function for defrouter_select(), and should
526 * not be called from anywhere else.
529 defrouter_delreq(struct nd_defrouter *dr)
531 struct sockaddr_in6 def, mask, gate;
532 struct rtentry *oldrt = NULL;
534 bzero(&def, sizeof(def));
535 bzero(&mask, sizeof(mask));
536 bzero(&gate, sizeof(gate));
538 def.sin6_len = mask.sin6_len = gate.sin6_len =
539 sizeof(struct sockaddr_in6);
540 def.sin6_family = gate.sin6_family = AF_INET6;
541 gate.sin6_addr = dr->rtaddr;
543 in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
544 (struct sockaddr *)&gate,
545 (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB);
547 nd6_rtmsg(RTM_DELETE, oldrt);
555 * remove all default routes from default router list
558 defrouter_reset(void)
560 struct nd_defrouter *dr;
562 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
563 defrouter_delreq(dr);
566 * XXX should we also nuke any default routers in the kernel, by
567 * going through them by rtalloc1()?
572 defrtrlist_del(struct nd_defrouter *dr)
574 struct nd_defrouter *deldr = NULL;
575 struct nd_prefix *pr;
578 * Flush all the routing table entries that use the router
581 if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
582 rt6_flush(&dr->rtaddr, dr->ifp);
586 defrouter_delreq(dr);
588 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
591 * Also delete all the pointers to the router in each prefix lists.
593 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
594 struct nd_pfxrouter *pfxrtr;
595 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
598 pfxlist_onlink_check();
601 * If the router is the primary one, choose a new one.
602 * Note that defrouter_select() will remove the current gateway
603 * from the routing table.
612 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
613 * draft-ietf-ipngwg-router-selection:
614 * 1) Routers that are reachable or probably reachable should be preferred.
615 * If we have more than one (probably) reachable router, prefer ones
616 * with the highest router preference.
617 * 2) When no routers on the list are known to be reachable or
618 * probably reachable, routers SHOULD be selected in a round-robin
619 * fashion, regardless of router preference values.
620 * 3) If the Default Router List is empty, assume that all
621 * destinations are on-link.
623 * We assume nd_defrouter is sorted by router preference value.
624 * Since the code below covers both with and without router preference cases,
625 * we do not need to classify the cases by ifdef.
627 * At this moment, we do not try to install more than one default router,
628 * even when the multipath routing is available, because we're not sure about
629 * the benefits for stub hosts comparing to the risk of making the code
630 * complicated and the possibility of introducing bugs.
633 defrouter_select(void)
636 struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
637 struct llentry *ln = NULL;
640 * Let's handle easy case (3) first:
641 * If default router list is empty, there's nothing to be done.
643 if (TAILQ_EMPTY(&V_nd_defrouter)) {
649 * Search for a (probably) reachable router from the list.
650 * We just pick up the first reachable one (if any), assuming that
651 * the ordering rule of the list described in defrtrlist_update().
653 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
654 IF_AFDATA_RLOCK(dr->ifp);
655 if (selected_dr == NULL &&
656 (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
657 ND6_IS_LLINFO_PROBREACH(ln)) {
660 IF_AFDATA_RUNLOCK(dr->ifp);
666 if (dr->installed && installed_dr == NULL)
668 else if (dr->installed && installed_dr) {
669 /* this should not happen. warn for diagnosis. */
670 log(LOG_ERR, "defrouter_select: more than one router"
675 * If none of the default routers was found to be reachable,
676 * round-robin the list regardless of preference.
677 * Otherwise, if we have an installed router, check if the selected
678 * (reachable) router should really be preferred to the installed one.
679 * We only prefer the new router when the old one is not reachable
680 * or when the new one has a really higher preference value.
682 if (selected_dr == NULL) {
683 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
684 selected_dr = TAILQ_FIRST(&V_nd_defrouter);
686 selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
687 } else if (installed_dr) {
688 IF_AFDATA_RLOCK(installed_dr->ifp);
689 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
690 ND6_IS_LLINFO_PROBREACH(ln) &&
691 rtpref(selected_dr) <= rtpref(installed_dr)) {
692 selected_dr = installed_dr;
694 IF_AFDATA_RUNLOCK(installed_dr->ifp);
700 * If the selected router is different than the installed one,
701 * remove the installed router and install the selected one.
702 * Note that the selected router is never NULL here.
704 if (installed_dr != selected_dr) {
706 defrouter_delreq(installed_dr);
707 defrouter_addreq(selected_dr);
715 * for default router selection
716 * regards router-preference field as a 2-bit signed integer
719 rtpref(struct nd_defrouter *dr)
721 switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
722 case ND_RA_FLAG_RTPREF_HIGH:
723 return (RTPREF_HIGH);
724 case ND_RA_FLAG_RTPREF_MEDIUM:
725 case ND_RA_FLAG_RTPREF_RSV:
726 return (RTPREF_MEDIUM);
727 case ND_RA_FLAG_RTPREF_LOW:
731 * This case should never happen. If it did, it would mean a
732 * serious bug of kernel internal. We thus always bark here.
733 * Or, can we even panic?
735 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
736 return (RTPREF_INVALID);
741 static struct nd_defrouter *
742 defrtrlist_update(struct nd_defrouter *new)
744 struct nd_defrouter *dr, *n;
747 if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
749 if (new->rtlifetime == 0) {
753 int oldpref = rtpref(dr);
756 dr->flags = new->flags; /* xxx flag check */
757 dr->rtlifetime = new->rtlifetime;
758 dr->expire = new->expire;
761 * If the preference does not change, there's no need
762 * to sort the entries. Also make sure the selected
763 * router is still installed in the kernel.
765 if (dr->installed && rtpref(new) == oldpref) {
771 * preferred router may be changed, so relocate
773 * XXX: calling TAILQ_REMOVE directly is a bad manner.
774 * However, since defrtrlist_del() has many side
775 * effects, we intentionally do so here.
776 * defrouter_select() below will handle routing
779 TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
787 /* entry does not exist */
788 if (new->rtlifetime == 0) {
793 n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
798 bzero(n, sizeof(*n));
803 * Insert the new router in the Default Router List;
804 * The Default Router List should be in the descending order
805 * of router-preferece. Routers with the same preference are
806 * sorted in the arriving time order.
809 /* insert at the end of the group */
810 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
811 if (rtpref(n) > rtpref(dr))
815 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
817 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
826 static struct nd_pfxrouter *
827 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
829 struct nd_pfxrouter *search;
831 LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
832 if (search->router == dr)
840 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
842 struct nd_pfxrouter *new;
844 new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
847 bzero(new, sizeof(*new));
850 LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
852 pfxlist_onlink_check();
856 pfxrtr_del(struct nd_pfxrouter *pfr)
858 LIST_REMOVE(pfr, pfr_entry);
863 nd6_prefix_lookup(struct nd_prefixctl *key)
865 struct nd_prefix *search;
867 LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
868 if (key->ndpr_ifp == search->ndpr_ifp &&
869 key->ndpr_plen == search->ndpr_plen &&
870 in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
871 &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
880 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
881 struct nd_prefix **newp)
883 struct nd_prefix *new = NULL;
886 char ip6buf[INET6_ADDRSTRLEN];
888 new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
891 bzero(new, sizeof(*new));
892 new->ndpr_ifp = pr->ndpr_ifp;
893 new->ndpr_prefix = pr->ndpr_prefix;
894 new->ndpr_plen = pr->ndpr_plen;
895 new->ndpr_vltime = pr->ndpr_vltime;
896 new->ndpr_pltime = pr->ndpr_pltime;
897 new->ndpr_flags = pr->ndpr_flags;
898 if ((error = in6_init_prefix_ltimes(new)) != 0) {
902 new->ndpr_lastupdate = time_second;
907 LIST_INIT(&new->ndpr_advrtrs);
908 in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
909 /* make prefix in the canonical form */
910 for (i = 0; i < 4; i++)
911 new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
912 new->ndpr_mask.s6_addr32[i];
915 /* link ndpr_entry to nd_prefix list */
916 LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
919 /* ND_OPT_PI_FLAG_ONLINK processing */
920 if (new->ndpr_raf_onlink) {
923 if ((e = nd6_prefix_onlink(new)) != 0) {
924 nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
925 "the prefix %s/%d on-link on %s (errno=%d)\n",
926 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
927 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
928 /* proceed anyway. XXX: is it correct? */
939 prelist_remove(struct nd_prefix *pr)
941 struct nd_pfxrouter *pfr, *next;
943 char ip6buf[INET6_ADDRSTRLEN];
945 /* make sure to invalidate the prefix until it is really freed. */
950 * Though these flags are now meaningless, we'd rather keep the value
951 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
952 * when executing "ndp -p".
955 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
956 (e = nd6_prefix_offlink(pr)) != 0) {
957 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
959 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
960 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
961 /* what should we do? */
964 if (pr->ndpr_refcnt > 0)
965 return; /* notice here? */
969 /* unlink ndpr_entry from nd_prefix list */
970 LIST_REMOVE(pr, ndpr_entry);
972 /* free list of routers that adversed the prefix */
973 LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
980 pfxlist_onlink_check();
988 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
989 struct mbuf *m, int mcast)
991 struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
993 struct ifnet *ifp = new->ndpr_ifp;
994 struct nd_prefix *pr;
999 struct in6_addrlifetime lt6_tmp;
1000 char ip6buf[INET6_ADDRSTRLEN];
1005 * Authenticity for NA consists authentication for
1006 * both IP header and IP datagrams, doesn't it ?
1008 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1009 auth = ((m->m_flags & M_AUTHIPHDR) &&
1010 (m->m_flags & M_AUTHIPDGM));
1014 if ((pr = nd6_prefix_lookup(new)) != NULL) {
1016 * nd6_prefix_lookup() ensures that pr and new have the same
1017 * prefix on a same interface.
1021 * Update prefix information. Note that the on-link (L) bit
1022 * and the autonomous (A) bit should NOT be changed from 1
1025 if (new->ndpr_raf_onlink == 1)
1026 pr->ndpr_raf_onlink = 1;
1027 if (new->ndpr_raf_auto == 1)
1028 pr->ndpr_raf_auto = 1;
1029 if (new->ndpr_raf_onlink) {
1030 pr->ndpr_vltime = new->ndpr_vltime;
1031 pr->ndpr_pltime = new->ndpr_pltime;
1032 (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1033 pr->ndpr_lastupdate = time_second;
1036 if (new->ndpr_raf_onlink &&
1037 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1040 if ((e = nd6_prefix_onlink(pr)) != 0) {
1042 "prelist_update: failed to make "
1043 "the prefix %s/%d on-link on %s "
1046 &pr->ndpr_prefix.sin6_addr),
1047 pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1048 /* proceed anyway. XXX: is it correct? */
1052 if (dr && pfxrtr_lookup(pr, dr) == NULL)
1055 struct nd_prefix *newpr = NULL;
1059 if (new->ndpr_vltime == 0)
1061 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1064 error = nd6_prelist_add(new, dr, &newpr);
1065 if (error != 0 || newpr == NULL) {
1066 nd6log((LOG_NOTICE, "prelist_update: "
1067 "nd6_prelist_add failed for %s/%d on %s "
1068 "errno=%d, returnpr=%p\n",
1069 ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1070 new->ndpr_plen, if_name(new->ndpr_ifp),
1072 goto end; /* we should just give up in this case. */
1076 * XXX: from the ND point of view, we can ignore a prefix
1077 * with the on-link bit being zero. However, we need a
1078 * prefix structure for references from autoconfigured
1079 * addresses. Thus, we explicitly make sure that the prefix
1080 * itself expires now.
1082 if (newpr->ndpr_raf_onlink == 0) {
1083 newpr->ndpr_vltime = 0;
1084 newpr->ndpr_pltime = 0;
1085 in6_init_prefix_ltimes(newpr);
1092 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1093 * Note that pr must be non NULL at this point.
1096 /* 5.5.3 (a). Ignore the prefix without the A bit set. */
1097 if (!new->ndpr_raf_auto)
1101 * 5.5.3 (b). the link-local prefix should have been ignored in
1105 /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1106 if (new->ndpr_pltime > new->ndpr_vltime) {
1107 error = EINVAL; /* XXX: won't be used */
1112 * 5.5.3 (d). If the prefix advertised is not equal to the prefix of
1113 * an address configured by stateless autoconfiguration already in the
1114 * list of addresses associated with the interface, and the Valid
1115 * Lifetime is not 0, form an address. We first check if we have
1116 * a matching prefix.
1117 * Note: we apply a clarification in rfc2462bis-02 here. We only
1118 * consider autoconfigured addresses while RFC2462 simply said
1122 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1123 struct in6_ifaddr *ifa6;
1124 u_int32_t remaininglifetime;
1126 if (ifa->ifa_addr->sa_family != AF_INET6)
1129 ifa6 = (struct in6_ifaddr *)ifa;
1132 * We only consider autoconfigured addresses as per rfc2462bis.
1134 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1138 * Spec is not clear here, but I believe we should concentrate
1139 * on unicast (i.e. not anycast) addresses.
1140 * XXX: other ia6_flags? detached or duplicated?
1142 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1146 * Ignore the address if it is not associated with a prefix
1147 * or is associated with a prefix that is different from this
1148 * one. (pr is never NULL here)
1150 if (ifa6->ia6_ndpr != pr)
1153 if (ia6_match == NULL) /* remember the first one */
1157 * An already autoconfigured address matched. Now that we
1158 * are sure there is at least one matched address, we can
1159 * proceed to 5.5.3. (e): update the lifetimes according to the
1160 * "two hours" rule and the privacy extension.
1161 * We apply some clarifications in rfc2462bis:
1162 * - use remaininglifetime instead of storedlifetime as a
1164 * - remove the dead code in the "two-hour" rule
1166 #define TWOHOUR (120*60)
1167 lt6_tmp = ifa6->ia6_lifetime;
1169 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1170 remaininglifetime = ND6_INFINITE_LIFETIME;
1171 else if (time_second - ifa6->ia6_updatetime >
1172 lt6_tmp.ia6t_vltime) {
1174 * The case of "invalid" address. We should usually
1175 * not see this case.
1177 remaininglifetime = 0;
1179 remaininglifetime = lt6_tmp.ia6t_vltime -
1180 (time_second - ifa6->ia6_updatetime);
1182 /* when not updating, keep the current stored lifetime. */
1183 lt6_tmp.ia6t_vltime = remaininglifetime;
1185 if (TWOHOUR < new->ndpr_vltime ||
1186 remaininglifetime < new->ndpr_vltime) {
1187 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1188 } else if (remaininglifetime <= TWOHOUR) {
1190 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1194 * new->ndpr_vltime <= TWOHOUR &&
1195 * TWOHOUR < remaininglifetime
1197 lt6_tmp.ia6t_vltime = TWOHOUR;
1200 /* The 2 hour rule is not imposed for preferred lifetime. */
1201 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1203 in6_init_address_ltimes(pr, <6_tmp);
1206 * We need to treat lifetimes for temporary addresses
1207 * differently, according to
1208 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1209 * we only update the lifetimes when they are in the maximum
1212 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1213 u_int32_t maxvltime, maxpltime;
1215 if (V_ip6_temp_valid_lifetime >
1216 (u_int32_t)((time_second - ifa6->ia6_createtime) +
1217 V_ip6_desync_factor)) {
1218 maxvltime = V_ip6_temp_valid_lifetime -
1219 (time_second - ifa6->ia6_createtime) -
1220 V_ip6_desync_factor;
1223 if (V_ip6_temp_preferred_lifetime >
1224 (u_int32_t)((time_second - ifa6->ia6_createtime) +
1225 V_ip6_desync_factor)) {
1226 maxpltime = V_ip6_temp_preferred_lifetime -
1227 (time_second - ifa6->ia6_createtime) -
1228 V_ip6_desync_factor;
1232 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1233 lt6_tmp.ia6t_vltime > maxvltime) {
1234 lt6_tmp.ia6t_vltime = maxvltime;
1236 if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1237 lt6_tmp.ia6t_pltime > maxpltime) {
1238 lt6_tmp.ia6t_pltime = maxpltime;
1241 ifa6->ia6_lifetime = lt6_tmp;
1242 ifa6->ia6_updatetime = time_second;
1244 IF_ADDR_RUNLOCK(ifp);
1245 if (ia6_match == NULL && new->ndpr_vltime) {
1249 * 5.5.3 (d) (continued)
1250 * No address matched and the valid lifetime is non-zero.
1251 * Create a new address.
1255 * Prefix Length check:
1256 * If the sum of the prefix length and interface identifier
1257 * length does not equal 128 bits, the Prefix Information
1258 * option MUST be ignored. The length of the interface
1259 * identifier is defined in a separate link-type specific
1262 ifidlen = in6_if2idlen(ifp);
1264 /* this should not happen, so we always log it. */
1265 log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1269 if (ifidlen + pr->ndpr_plen != 128) {
1271 "prelist_update: invalid prefixlen "
1272 "%d for %s, ignored\n",
1273 pr->ndpr_plen, if_name(ifp)));
1277 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1279 * note that we should use pr (not new) for reference.
1286 * When a new public address is created as described
1287 * in RFC2462, also create a new temporary address.
1290 * When an interface connects to a new link, a new
1291 * randomized interface identifier should be generated
1292 * immediately together with a new set of temporary
1293 * addresses. Thus, we specifiy 1 as the 2nd arg of
1296 if (V_ip6_use_tempaddr) {
1298 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1299 nd6log((LOG_NOTICE, "prelist_update: "
1300 "failed to create a temporary "
1301 "address, errno=%d\n",
1305 ifa_free(&ia6->ia_ifa);
1308 * A newly added address might affect the status
1309 * of other addresses, so we check and update it.
1310 * XXX: what if address duplication happens?
1312 pfxlist_onlink_check();
1314 /* just set an error. do not bark here. */
1315 error = EADDRNOTAVAIL; /* XXX: might be unused. */
1325 * A supplement function used in the on-link detection below;
1326 * detect if a given prefix has a (probably) reachable advertising router.
1327 * XXX: lengthy function name...
1329 static struct nd_pfxrouter *
1330 find_pfxlist_reachable_router(struct nd_prefix *pr)
1332 struct nd_pfxrouter *pfxrtr;
1336 LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1337 IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1338 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1339 IF_AFDATA_RUNLOCK(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 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
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 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1386 struct nd_prefix *pr0;
1388 LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1389 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1396 if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1398 * There is at least one prefix that has a reachable router,
1399 * or at least a router which probably does not advertise
1400 * any prefixes. The latter would be the case when we move
1401 * to a new link where we have a router that does not provide
1402 * prefixes and we configure an address by hand.
1403 * Detach prefixes which have no reachable advertising
1404 * router, and attach other prefixes.
1406 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1407 /* XXX: a link-local prefix should never be detached */
1408 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1412 * we aren't interested in prefixes without the L bit
1415 if (pr->ndpr_raf_onlink == 0)
1418 if (pr->ndpr_raf_auto == 0)
1421 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1422 find_pfxlist_reachable_router(pr) == NULL)
1423 pr->ndpr_stateflags |= NDPRF_DETACHED;
1424 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1425 find_pfxlist_reachable_router(pr) != 0)
1426 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1429 /* there is no prefix that has a reachable router */
1430 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1431 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1434 if (pr->ndpr_raf_onlink == 0)
1437 if (pr->ndpr_raf_auto == 0)
1440 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1441 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1446 * Remove each interface route associated with a (just) detached
1447 * prefix, and reinstall the interface route for a (just) attached
1448 * prefix. Note that all attempt of reinstallation does not
1449 * necessarily success, when a same prefix is shared among multiple
1450 * interfaces. Such cases will be handled in nd6_prefix_onlink,
1451 * so we don't have to care about them.
1453 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1455 char ip6buf[INET6_ADDRSTRLEN];
1457 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1460 if (pr->ndpr_raf_onlink == 0)
1463 if (pr->ndpr_raf_auto == 0)
1466 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1467 (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1468 if ((e = nd6_prefix_offlink(pr)) != 0) {
1470 "pfxlist_onlink_check: failed to "
1471 "make %s/%d offlink, errno=%d\n",
1473 &pr->ndpr_prefix.sin6_addr),
1477 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1478 (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1479 pr->ndpr_raf_onlink) {
1480 if ((e = nd6_prefix_onlink(pr)) != 0) {
1482 "pfxlist_onlink_check: failed to "
1483 "make %s/%d onlink, errno=%d\n",
1485 &pr->ndpr_prefix.sin6_addr),
1492 * Changes on the prefix status might affect address status as well.
1493 * Make sure that all addresses derived from an attached prefix are
1494 * attached, and that all addresses derived from a detached prefix are
1495 * detached. Note, however, that a manually configured address should
1496 * always be attached.
1497 * The precise detection logic is same as the one for prefixes.
1499 * XXXRW: in6_ifaddrhead locking.
1501 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1502 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1505 if (ifa->ia6_ndpr == NULL) {
1507 * This can happen when we first configure the address
1508 * (i.e. the address exists, but the prefix does not).
1509 * XXX: complicated relationships...
1514 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1518 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1519 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1522 if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1525 if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1526 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1527 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1528 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1529 nd6_dad_start((struct ifaddr *)ifa, 0);
1532 ifa->ia6_flags |= IN6_IFF_DETACHED;
1537 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1538 if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1541 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1542 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1543 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1544 /* Do we need a delay in this case? */
1545 nd6_dad_start((struct ifaddr *)ifa, 0);
1552 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1554 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1555 struct radix_node_head *rnh;
1557 struct sockaddr_in6 mask6;
1559 int error, a_failure, fibnum;
1562 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1563 * ifa->ifa_rtrequest = nd6_rtrequest;
1565 bzero(&mask6, sizeof(mask6));
1566 mask6.sin6_len = sizeof(mask6);
1567 mask6.sin6_addr = pr->ndpr_mask;
1568 rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1571 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1574 error = in6_rtrequest(RTM_ADD,
1575 (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1576 (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1578 KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1579 "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1582 rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1583 /* XXX what if rhn == NULL? */
1584 RADIX_NODE_HEAD_LOCK(rnh);
1586 if (rt_setgate(rt, rt_key(rt),
1587 (struct sockaddr *)&null_sdl) == 0) {
1588 struct sockaddr_dl *dl;
1590 dl = (struct sockaddr_dl *)rt->rt_gateway;
1591 dl->sdl_type = rt->rt_ifp->if_type;
1592 dl->sdl_index = rt->rt_ifp->if_index;
1594 RADIX_NODE_HEAD_UNLOCK(rnh);
1595 nd6_rtmsg(RTM_ADD, rt);
1597 pr->ndpr_stateflags |= NDPRF_ONLINK;
1599 char ip6buf[INET6_ADDRSTRLEN];
1600 char ip6bufg[INET6_ADDRSTRLEN];
1601 char ip6bufm[INET6_ADDRSTRLEN];
1602 struct sockaddr_in6 *sin6;
1604 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1605 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1606 "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1607 "flags=%lx errno = %d\n",
1608 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1609 pr->ndpr_plen, if_name(pr->ndpr_ifp),
1610 ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1611 ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1614 /* Save last error to return, see rtinit(). */
1625 /* Return the last error we got. */
1630 nd6_prefix_onlink(struct nd_prefix *pr)
1633 struct ifnet *ifp = pr->ndpr_ifp;
1634 struct nd_prefix *opr;
1636 char ip6buf[INET6_ADDRSTRLEN];
1639 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1641 "nd6_prefix_onlink: %s/%d is already on-link\n",
1642 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1648 * Add the interface route associated with the prefix. Before
1649 * installing the route, check if there's the same prefix on another
1650 * interface, and the prefix has already installed the interface route.
1651 * Although such a configuration is expected to be rare, we explicitly
1654 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1658 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1661 if (opr->ndpr_plen == pr->ndpr_plen &&
1662 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1663 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1668 * We prefer link-local addresses as the associated interface address.
1670 /* search for a link-local addr */
1671 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1672 IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1674 /* XXX: freebsd does not have ifa_ifwithaf */
1676 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1677 if (ifa->ifa_addr->sa_family == AF_INET6)
1682 IF_ADDR_RUNLOCK(ifp);
1683 /* should we care about ia6_flags? */
1687 * This can still happen, when, for example, we receive an RA
1688 * containing a prefix with the L bit set and the A bit clear,
1689 * after removing all IPv6 addresses on the receiving
1690 * interface. This should, of course, be rare though.
1693 "nd6_prefix_onlink: failed to find any ifaddr"
1694 " to add route for a prefix(%s/%d) on %s\n",
1695 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1696 pr->ndpr_plen, if_name(ifp)));
1700 error = nd6_prefix_onlink_rtrequest(pr, ifa);
1709 nd6_prefix_offlink(struct nd_prefix *pr)
1712 struct ifnet *ifp = pr->ndpr_ifp;
1713 struct nd_prefix *opr;
1714 struct sockaddr_in6 sa6, mask6;
1716 char ip6buf[INET6_ADDRSTRLEN];
1717 int fibnum, a_failure;
1720 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1722 "nd6_prefix_offlink: %s/%d is already off-link\n",
1723 ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1728 bzero(&sa6, sizeof(sa6));
1729 sa6.sin6_family = AF_INET6;
1730 sa6.sin6_len = sizeof(sa6);
1731 bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1732 sizeof(struct in6_addr));
1733 bzero(&mask6, sizeof(mask6));
1734 mask6.sin6_family = AF_INET6;
1735 mask6.sin6_len = sizeof(sa6);
1736 bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1739 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1741 error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1742 (struct sockaddr *)&mask6, 0, &rt, fibnum);
1744 /* report the route deletion to the routing socket. */
1746 nd6_rtmsg(RTM_DELETE, rt);
1748 /* Save last error to return, see rtinit(). */
1757 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1760 * There might be the same prefix on another interface,
1761 * the prefix which could not be on-link just because we have
1762 * the interface route (see comments in nd6_prefix_onlink).
1763 * If there's one, try to make the prefix on-link on the
1766 LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1770 if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1774 * KAME specific: detached prefixes should not be
1777 if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1780 if (opr->ndpr_plen == pr->ndpr_plen &&
1781 in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1782 &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1785 if ((e = nd6_prefix_onlink(opr)) != 0) {
1787 "nd6_prefix_offlink: failed to "
1788 "recover a prefix %s/%d from %s "
1789 "to %s (errno = %d)\n",
1791 &opr->ndpr_prefix.sin6_addr),
1792 opr->ndpr_plen, if_name(ifp),
1793 if_name(opr->ndpr_ifp), e));
1798 /* XXX: can we still set the NDPRF_ONLINK flag? */
1800 "nd6_prefix_offlink: failed to delete route: "
1801 "%s/%d on %s (errno = %d)\n",
1802 ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1803 if_name(ifp), error));
1809 static struct in6_ifaddr *
1810 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1812 struct ifnet *ifp = pr->ndpr_ifp;
1814 struct in6_aliasreq ifra;
1815 struct in6_ifaddr *ia, *ib;
1817 struct in6_addr mask;
1818 int prefixlen = pr->ndpr_plen;
1820 char ip6buf[INET6_ADDRSTRLEN];
1822 in6_prefixlen2mask(&mask, prefixlen);
1825 * find a link-local address (will be interface ID).
1826 * Is it really mandatory? Theoretically, a global or a site-local
1827 * address can be configured without a link-local address, if we
1828 * have a unique interface identifier...
1830 * it is not mandatory to have a link-local address, we can generate
1831 * interface identifier on the fly. we do this because:
1832 * (1) it should be the easiest way to find interface identifier.
1833 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1834 * for multiple addresses on a single interface, and possible shortcut
1835 * of DAD. we omitted DAD for this reason in the past.
1836 * (3) a user can prevent autoconfiguration of global address
1837 * by removing link-local address by hand (this is partly because we
1838 * don't have other way to control the use of IPv6 on an interface.
1839 * this has been our design choice - cf. NRL's "ifconfig auto").
1840 * (4) it is easier to manage when an interface has addresses
1841 * with the same interface identifier, than to have multiple addresses
1842 * with different interface identifiers.
1844 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1846 ib = (struct in6_ifaddr *)ifa;
1850 /* prefixlen + ifidlen must be equal to 128 */
1851 plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1852 if (prefixlen != plen0) {
1854 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1855 "(prefix=%d ifid=%d)\n",
1856 if_name(ifp), prefixlen, 128 - plen0));
1862 bzero(&ifra, sizeof(ifra));
1864 * in6_update_ifa() does not use ifra_name, but we accurately set it
1867 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1868 ifra.ifra_addr.sin6_family = AF_INET6;
1869 ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1871 ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1872 ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1873 ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1874 ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1875 ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1878 ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1879 (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1880 ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1881 (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1882 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1883 (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1884 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1885 (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1888 /* new prefix mask. */
1889 ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1890 ifra.ifra_prefixmask.sin6_family = AF_INET6;
1891 bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1892 sizeof(ifra.ifra_prefixmask.sin6_addr));
1895 ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1896 ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1898 /* XXX: scope zone ID? */
1900 ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1903 * Make sure that we do not have this address already. This should
1904 * usually not happen, but we can still see this case, e.g., if we
1905 * have manually configured the exact address to be configured.
1907 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1908 &ifra.ifra_addr.sin6_addr);
1911 /* this should be rare enough to make an explicit log */
1912 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1913 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1918 * Allocate ifaddr structure, link into chain, etc.
1919 * If we are going to create a new address upon receiving a multicasted
1920 * RA, we need to impose a random delay before starting DAD.
1921 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1925 updateflags |= IN6_IFAUPDATE_DADDELAY;
1926 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1928 "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1929 ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1930 if_name(ifp), error));
1931 return (NULL); /* ifaddr must not have been allocated. */
1934 ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1936 * XXXRW: Assumption of non-NULLness here might not be true with
1937 * fine-grained locking -- should we validate it? Or just return
1938 * earlier ifa rather than looking it up again?
1940 return (ia); /* this is always non-NULL and referenced. */
1944 * ia0 - corresponding public address
1947 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1949 struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1950 struct in6_ifaddr *newia, *ia;
1951 struct in6_aliasreq ifra;
1953 int trylimit = 3; /* XXX: adhoc value */
1955 u_int32_t randid[2];
1956 time_t vltime0, pltime0;
1958 bzero(&ifra, sizeof(ifra));
1959 strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1960 ifra.ifra_addr = ia0->ia_addr;
1961 /* copy prefix mask */
1962 ifra.ifra_prefixmask = ia0->ia_prefixmask;
1963 /* clear the old IFID */
1964 for (i = 0; i < 4; i++) {
1965 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1966 ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1970 if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1971 (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1972 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1976 ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1977 (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1978 ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1979 (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1982 * in6_get_tmpifid() quite likely provided a unique interface ID.
1983 * However, we may still have a chance to see collision, because
1984 * there may be a time lag between generation of the ID and generation
1985 * of the address. So, we'll do one more sanity check.
1988 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1989 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1990 &ifra.ifra_addr.sin6_addr)) {
1991 if (trylimit-- == 0) {
1992 IN6_IFADDR_RUNLOCK();
1994 * Give up. Something strange should have
1997 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1998 "find a unique random IFID\n"));
2001 IN6_IFADDR_RUNLOCK();
2006 IN6_IFADDR_RUNLOCK();
2009 * The Valid Lifetime is the lower of the Valid Lifetime of the
2010 * public address or TEMP_VALID_LIFETIME.
2011 * The Preferred Lifetime is the lower of the Preferred Lifetime
2012 * of the public address or TEMP_PREFERRED_LIFETIME -
2015 if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2016 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2017 (ia0->ia6_lifetime.ia6t_vltime -
2018 (time_second - ia0->ia6_updatetime));
2019 if (vltime0 > V_ip6_temp_valid_lifetime)
2020 vltime0 = V_ip6_temp_valid_lifetime;
2022 vltime0 = V_ip6_temp_valid_lifetime;
2023 if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2024 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2025 (ia0->ia6_lifetime.ia6t_pltime -
2026 (time_second - ia0->ia6_updatetime));
2027 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2028 pltime0 = V_ip6_temp_preferred_lifetime -
2029 V_ip6_desync_factor;
2032 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2033 ifra.ifra_lifetime.ia6t_vltime = vltime0;
2034 ifra.ifra_lifetime.ia6t_pltime = pltime0;
2037 * A temporary address is created only if this calculated Preferred
2038 * Lifetime is greater than REGEN_ADVANCE time units.
2040 if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2043 /* XXX: scope zone ID? */
2045 ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2047 /* allocate ifaddr structure, link into chain, etc. */
2050 updateflags |= IN6_IFAUPDATE_DADDELAY;
2051 if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2054 newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2055 if (newia == NULL) { /* XXX: can it happen? */
2057 "in6_tmpifadd: ifa update succeeded, but we got "
2059 return (EINVAL); /* XXX */
2061 newia->ia6_ndpr = ia0->ia6_ndpr;
2062 newia->ia6_ndpr->ndpr_refcnt++;
2063 ifa_free(&newia->ia_ifa);
2066 * A newly added address might affect the status of other addresses.
2067 * XXX: when the temporary address is generated with a new public
2068 * address, the onlink check is redundant. However, it would be safe
2069 * to do the check explicitly everywhere a new address is generated,
2070 * and, in fact, we surely need the check when we create a new
2071 * temporary address due to deprecation of an old temporary address.
2073 pfxlist_onlink_check();
2079 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2081 if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2082 ndpr->ndpr_preferred = 0;
2084 ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2085 if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2086 ndpr->ndpr_expire = 0;
2088 ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2094 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2096 /* init ia6t_expire */
2097 if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2098 lt6->ia6t_expire = 0;
2100 lt6->ia6t_expire = time_second;
2101 lt6->ia6t_expire += lt6->ia6t_vltime;
2104 /* init ia6t_preferred */
2105 if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2106 lt6->ia6t_preferred = 0;
2108 lt6->ia6t_preferred = time_second;
2109 lt6->ia6t_preferred += lt6->ia6t_pltime;
2114 * Delete all the routing table entries that use the specified gateway.
2115 * XXX: this function causes search through all entries of routing table, so
2116 * it shouldn't be called when acting as a router.
2119 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2121 struct radix_node_head *rnh;
2125 /* We'll care only link-local addresses */
2126 if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2131 /* XXX Do we really need to walk any but the default FIB? */
2132 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
2133 rnh = rt_tables_get_rnh(fibnum, AF_INET6);
2137 RADIX_NODE_HEAD_LOCK(rnh);
2138 rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2139 RADIX_NODE_HEAD_UNLOCK(rnh);
2145 rt6_deleteroute(struct radix_node *rn, void *arg)
2147 #define SIN6(s) ((struct sockaddr_in6 *)s)
2148 struct rtentry *rt = (struct rtentry *)rn;
2149 struct in6_addr *gate = (struct in6_addr *)arg;
2151 if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2154 if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2159 * Do not delete a static route.
2160 * XXX: this seems to be a bit ad-hoc. Should we consider the
2161 * 'cloned' bit instead?
2163 if ((rt->rt_flags & RTF_STATIC) != 0)
2167 * We delete only host route. This means, in particular, we don't
2168 * delete default route.
2170 if ((rt->rt_flags & RTF_HOST) == 0)
2173 return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2174 rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum));
2179 nd6_setdefaultiface(int ifindex)
2183 if (ifindex < 0 || V_if_index < ifindex)
2185 if (ifindex != 0 && !ifnet_byindex(ifindex))
2188 if (V_nd6_defifindex != ifindex) {
2189 V_nd6_defifindex = ifindex;
2190 if (V_nd6_defifindex > 0)
2191 V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2193 V_nd6_defifp = NULL;
2196 * Our current implementation assumes one-to-one maping between
2197 * interfaces and links, so it would be natural to use the
2198 * default interface as the default link.
2200 scope6_setdefault(V_nd6_defifp);