1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.2.2.15 2003/05/06 06:46:58 suz Exp $ */
2 /* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include "opt_inet6.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/callout.h>
39 #include <sys/malloc.h>
41 #include <sys/socket.h>
42 #include <sys/sockio.h>
44 #include <sys/kernel.h>
45 #include <sys/protosw.h>
46 #include <sys/errno.h>
47 #include <sys/syslog.h>
48 #include <sys/queue.h>
49 #include <sys/sysctl.h>
50 #include <sys/mutex.h>
52 #include <sys/thread2.h>
53 #include <sys/mutex2.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 #include <net/netisr2.h>
60 #include <net/netmsg2.h>
62 #include <netinet/in.h>
63 #include <netinet/if_ether.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet/ip6.h>
66 #include <netinet6/ip6_var.h>
67 #include <netinet6/nd6.h>
68 #include <netinet/icmp6.h>
70 #include <net/net_osdep.h>
72 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
73 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
75 #define SIN6(s) ((struct sockaddr_in6 *)s)
76 #define SDL(s) ((struct sockaddr_dl *)s)
79 * Determine if the route entry is a direct neighbor on the specified
80 * interface. The interface test is not done if ifp is passed as NULL.
81 * The route entry is a neighbor if all of the following are true:
83 * RTF_GATEWAY is FALSE
85 * rt_gateway family is AF_LINK
86 * rt_llinfo is non-NULL
87 * The interfaces matches (or ifp is passed as NULL)
89 * NOTE: rt_llinfo can be NULL with LLINFO set, so both must be
92 * NOTE: We can't use rt->rt_ifp to check for the interface, since
93 * it may be the loopback interface if the entry is for our
94 * own address on a non-loopback interface. Instead, we use
95 * rt->rt_ifa->ifa_ifp which should specify the REAL interface.
97 #define ND6_IFP_MATCHES(ifp, ifa_ifp) \
99 (ifa_ifp) == (ifp) || \
100 (((ifp)->if_flags & IFF_ISBRIDGE) && \
101 (ifa_ifp)->if_bridge == (ifp)->if_softc) \
104 #define ND6_RTENTRY_IS_NEIGHBOR(rt, ifp) \
105 (((rt)->rt_flags & RTF_GATEWAY) == 0 && \
106 ((rt)->rt_flags & RTF_LLINFO) != 0 && \
107 (rt)->rt_gateway->sa_family == AF_LINK && \
109 ND6_IFP_MATCHES((ifp), (rt)->rt_ifa->ifa_ifp) \
112 #define ND6_RTENTRY_IS_LLCLONING(rt) \
113 (((rt)->rt_flags & (RTF_PRCLONING | RTF_LLINFO)) == \
114 (RTF_PRCLONING | RTF_LLINFO) || \
115 ((rt)->rt_flags & RTF_CLONING))
118 int nd6_prune = 1; /* walk list every 1 seconds */
119 int nd6_delay = 5; /* delay first probe time 5 second */
120 int nd6_umaxtries = 3; /* maximum unicast query */
121 int nd6_mmaxtries = 3; /* maximum multicast query */
122 int nd6_useloopback = 1; /* use loopback interface for local traffic */
123 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
125 /* preventing too many loops in ND option parsing */
126 int nd6_maxndopt = 10; /* max # of ND options allowed */
128 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
137 static int nd6_inuse, nd6_allocated;
139 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
140 struct nd_drhead nd_defrouter;
141 struct nd_prhead nd_prefix = { 0 };
142 struct mtx nd6_mtx = MTX_INITIALIZER("nd6");
144 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
145 static struct sockaddr_in6 all1_sa;
147 static void nd6_setmtu0 (struct ifnet *, struct nd_ifinfo *);
148 static int regen_tmpaddr (struct in6_ifaddr *);
149 static void nd6_slowtimo(void *);
150 static void nd6_slowtimo_dispatch(netmsg_t);
151 static void nd6_timer(void *);
152 static void nd6_timer_dispatch(netmsg_t);
154 static struct callout nd6_slowtimo_ch;
155 static struct netmsg_base nd6_slowtimo_netmsg;
157 static struct callout nd6_timer_ch;
158 static struct netmsg_base nd6_timer_netmsg;
163 static int nd6_init_done = 0;
167 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
171 all1_sa.sin6_family = AF_INET6;
172 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
173 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
174 all1_sa.sin6_addr.s6_addr[i] = 0xff;
176 /* initialization of the default router list */
177 TAILQ_INIT(&nd_defrouter);
182 callout_init_mp(&nd6_slowtimo_ch);
183 netmsg_init(&nd6_slowtimo_netmsg, NULL, &netisr_adone_rport,
184 MSGF_PRIORITY, nd6_slowtimo_dispatch);
185 callout_reset_bycpu(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
186 nd6_slowtimo, NULL, 0);
190 nd6_ifattach(struct ifnet *ifp)
192 struct nd_ifinfo *nd;
194 nd = (struct nd_ifinfo *)kmalloc(sizeof(*nd), M_IP6NDP,
199 nd->chlim = IPV6_DEFHLIM;
200 nd->basereachable = REACHABLE_TIME;
201 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
202 nd->retrans = RETRANS_TIMER;
205 * Note that the default value of ip6_accept_rtadv is 0, which means
206 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
209 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
211 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
212 nd6_setmtu0(ifp, nd);
217 nd6_ifdetach(struct nd_ifinfo *nd)
223 * Reset ND level link MTU. This function is called when the physical MTU
224 * changes, which means we might have to adjust the ND level MTU.
227 nd6_setmtu(struct ifnet *ifp)
229 nd6_setmtu0(ifp, ND_IFINFO(ifp));
232 struct netmsg_nd6setmtu {
233 struct netmsg_base nmsg;
235 struct nd_ifinfo *ndi;
238 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
240 nd6_setmtu0_dispatch(netmsg_t msg)
242 struct netmsg_nd6setmtu *nmsg = (struct netmsg_nd6setmtu *)msg;
243 struct ifnet *ifp = nmsg->ifp;
244 struct nd_ifinfo *ndi = nmsg->ndi;
247 omaxmtu = ndi->maxmtu;
249 switch (ifp->if_type) {
251 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
253 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */
254 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
257 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */
258 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
262 ndi->maxmtu = ifp->if_mtu;
267 * Decreasing the interface MTU under IPV6 minimum MTU may cause
268 * undesirable situation. We thus notify the operator of the change
269 * explicitly. The check for omaxmtu is necessary to restrict the
270 * log to the case of changing the MTU, not initializing it.
272 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
273 log(LOG_NOTICE, "nd6_setmtu0: "
274 "new link MTU on %s (%lu) is too small for IPv6\n",
275 if_name(ifp), (unsigned long)ndi->maxmtu);
278 if (ndi->maxmtu > in6_maxmtu)
279 in6_setmaxmtu(); /* check all interfaces just in case */
281 lwkt_replymsg(&nmsg->nmsg.lmsg, 0);
285 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
287 struct netmsg_nd6setmtu nmsg;
289 netmsg_init(&nmsg.nmsg, NULL, &curthread->td_msgport, 0,
290 nd6_setmtu0_dispatch);
293 lwkt_domsg(netisr_cpuport(0), &nmsg.nmsg.lmsg, 0);
297 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
299 bzero(ndopts, sizeof(*ndopts));
300 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
302 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
305 ndopts->nd_opts_done = 1;
306 ndopts->nd_opts_search = NULL;
311 * Take one ND option.
314 nd6_option(union nd_opts *ndopts)
316 struct nd_opt_hdr *nd_opt;
320 panic("ndopts == NULL in nd6_option");
321 if (!ndopts->nd_opts_last)
322 panic("uninitialized ndopts in nd6_option");
323 if (!ndopts->nd_opts_search)
325 if (ndopts->nd_opts_done)
328 nd_opt = ndopts->nd_opts_search;
330 /* make sure nd_opt_len is inside the buffer */
331 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
332 bzero(ndopts, sizeof(*ndopts));
336 olen = nd_opt->nd_opt_len << 3;
339 * Message validation requires that all included
340 * options have a length that is greater than zero.
342 bzero(ndopts, sizeof(*ndopts));
346 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
347 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
348 /* option overruns the end of buffer, invalid */
349 bzero(ndopts, sizeof(*ndopts));
351 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
352 /* reached the end of options chain */
353 ndopts->nd_opts_done = 1;
354 ndopts->nd_opts_search = NULL;
360 * Parse multiple ND options.
361 * This function is much easier to use, for ND routines that do not need
362 * multiple options of the same type.
365 nd6_options(union nd_opts *ndopts)
367 struct nd_opt_hdr *nd_opt;
371 panic("ndopts == NULL in nd6_options");
372 if (!ndopts->nd_opts_last)
373 panic("uninitialized ndopts in nd6_options");
374 if (!ndopts->nd_opts_search)
378 nd_opt = nd6_option(ndopts);
379 if (!nd_opt && !ndopts->nd_opts_last) {
381 * Message validation requires that all included
382 * options have a length that is greater than zero.
384 icmp6stat.icp6s_nd_badopt++;
385 bzero(ndopts, sizeof(*ndopts));
392 switch (nd_opt->nd_opt_type) {
393 case ND_OPT_SOURCE_LINKADDR:
394 case ND_OPT_TARGET_LINKADDR:
396 case ND_OPT_REDIRECTED_HEADER:
397 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
399 "duplicated ND6 option found (type=%d)\n",
400 nd_opt->nd_opt_type));
403 ndopts->nd_opt_array[nd_opt->nd_opt_type]
407 case ND_OPT_PREFIX_INFORMATION:
408 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
409 ndopts->nd_opt_array[nd_opt->nd_opt_type]
412 ndopts->nd_opts_pi_end =
413 (struct nd_opt_prefix_info *)nd_opt;
417 * Unknown options must be silently ignored,
418 * to accomodate future extension to the protocol.
421 "nd6_options: unsupported option %d - "
422 "option ignored\n", nd_opt->nd_opt_type));
427 if (i > nd6_maxndopt) {
428 icmp6stat.icp6s_nd_toomanyopt++;
429 nd6log((LOG_INFO, "too many loop in nd opt\n"));
433 if (ndopts->nd_opts_done)
441 * ND6 timer routine to expire default route list and prefix list
444 nd6_timer_dispatch(netmsg_t nmsg)
446 struct llinfo_nd6 *ln;
447 struct nd_defrouter *dr;
448 struct nd_prefix *pr;
450 struct in6_ifaddr *ia6, *nia6;
455 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
460 ln = llinfo_nd6.ln_next;
461 while (ln && ln != &llinfo_nd6) {
463 struct sockaddr_in6 *dst;
464 struct llinfo_nd6 *next = ln->ln_next;
465 /* XXX: used for the DELAY case only: */
466 struct nd_ifinfo *ndi = NULL;
468 if ((rt = ln->ln_rt) == NULL) {
472 if ((ifp = rt->rt_ifp) == NULL) {
476 ndi = ND_IFINFO(ifp);
477 dst = (struct sockaddr_in6 *)rt_key(rt);
479 if (ln->ln_expire > time_uptime) {
486 panic("rt=0 in nd6_timer(ln=%p)", ln);
487 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
488 panic("rt_llinfo(%p) is not equal to ln(%p)",
491 panic("dst=0 in nd6_timer(ln=%p)", ln);
493 switch (ln->ln_state) {
494 case ND6_LLINFO_WAITDELETE:
497 case ND6_LLINFO_INCOMPLETE:
498 if (ln->ln_asked++ >= nd6_mmaxtries) {
499 struct mbuf *m = ln->ln_hold;
503 * Fake rcvif to make ICMP error
504 * more helpful in diagnosing
506 * XXX: should we consider
509 m->m_pkthdr.rcvif = rt->rt_ifp;
513 * mbuf has empty MAC header, remove
514 * for icmp. XXX layer violation.
516 m_adj(m, ETHER_HDR_LEN);
517 icmp6_error(m, ICMP6_DST_UNREACH,
518 ICMP6_DST_UNREACH_ADDR, 0);
521 ln->ln_state = ND6_LLINFO_WAITDELETE;
522 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0);
524 ln->ln_expire = time_uptime +
525 ND_IFINFO(ifp)->retrans / 1000;
526 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
529 case ND6_LLINFO_REACHABLE:
531 ln->ln_state = ND6_LLINFO_STALE;
532 ln->ln_expire = time_uptime + nd6_gctimer;
536 case ND6_LLINFO_STALE:
537 /* Garbage Collection(RFC 2461 5.3) */
542 case ND6_LLINFO_DELAY:
543 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD)) {
546 ln->ln_state = ND6_LLINFO_PROBE;
547 ln->ln_expire = time_uptime +
549 nd6_ns_output(ifp, &dst->sin6_addr,
553 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
554 ln->ln_expire = time_uptime + nd6_gctimer;
557 case ND6_LLINFO_PROBE:
558 if (ln->ln_asked < nd6_umaxtries) {
560 ln->ln_expire = time_uptime +
561 ND_IFINFO(ifp)->retrans / 1000;
562 nd6_ns_output(ifp, &dst->sin6_addr,
563 &dst->sin6_addr, ln, 0);
565 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0);
573 /* expire default router list */
574 dr = TAILQ_FIRST(&nd_defrouter);
576 if (dr->expire && dr->expire < time_uptime) {
577 struct nd_defrouter *t;
578 t = TAILQ_NEXT(dr, dr_entry);
582 dr = TAILQ_NEXT(dr, dr_entry);
587 * expire interface addresses.
588 * in the past the loop was inside prefix expiry processing.
589 * However, from a stricter speci-confrmance standpoint, we should
590 * rather separate address lifetimes and prefix lifetimes.
593 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
595 /* check address lifetime */
596 if (IFA6_IS_INVALID(ia6)) {
600 * If the expiring address is temporary, try
601 * regenerating a new one. This would be useful when
602 * we suspended a laptop PC, then turned it on after a
603 * period that could invalidate all temporary
604 * addresses. Although we may have to restart the
605 * loop (see below), it must be after purging the
606 * address. Otherwise, we'd see an infinite loop of
609 if (ip6_use_tempaddr &&
610 (ia6->ia6_flags & IN6_IFF_TEMPORARY)) {
611 if (regen_tmpaddr(ia6) == 0)
615 in6_purgeaddr(&ia6->ia_ifa);
618 goto addrloop; /* XXX: see below */
619 /* ia6 is no longer good, continue on to next */
622 if (IFA6_IS_DEPRECATED(ia6)) {
623 int oldflags = ia6->ia6_flags;
625 if ((oldflags & IN6_IFF_DEPRECATED) == 0) {
626 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
627 in6_newaddrmsg((struct ifaddr *)ia6);
631 * If a temporary address has just become deprecated,
632 * regenerate a new one if possible.
634 if (ip6_use_tempaddr &&
635 (ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
636 !(oldflags & IN6_IFF_DEPRECATED)) {
638 if (regen_tmpaddr(ia6) == 0) {
640 * A new temporary address is
642 * XXX: this means the address chain
643 * has changed while we are still in
644 * the loop. Although the change
645 * would not cause disaster (because
646 * it's not a deletion, but an
647 * addition,) we'd rather restart the
648 * loop just for safety. Or does this
649 * significantly reduce performance??
656 * A new RA might have made a deprecated address
659 if (ia6->ia6_flags & IN6_IFF_DEPRECATED) {
660 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
661 in6_newaddrmsg((struct ifaddr *)ia6);
666 /* expire prefix list */
667 pr = nd_prefix.lh_first;
670 * check prefix lifetime.
671 * since pltime is just for autoconf, pltime processing for
672 * prefix is not necessary.
674 if (pr->ndpr_expire && pr->ndpr_expire < time_uptime) {
679 * address expiration and prefix expiration are
680 * separate. NEVER perform in6_purgeaddr here.
689 mtx_unlock(&nd6_mtx);
691 callout_reset(&nd6_timer_ch, nd6_prune * hz, nd6_timer, NULL);
695 nd6_timer(void *arg __unused)
697 struct lwkt_msg *lmsg = &nd6_timer_netmsg.lmsg;
699 KASSERT(mycpuid == 0, ("not on cpu0"));
701 if (lmsg->ms_flags & MSGF_DONE)
702 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg);
709 callout_init_mp(&nd6_timer_ch);
710 netmsg_init(&nd6_timer_netmsg, NULL, &netisr_adone_rport,
711 MSGF_PRIORITY, nd6_timer_dispatch);
712 callout_reset_bycpu(&nd6_timer_ch, hz, nd6_timer, NULL, 0);
716 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary
719 struct ifaddr_container *ifac;
721 struct in6_ifaddr *public_ifa6 = NULL;
723 ifp = ia6->ia_ifa.ifa_ifp;
724 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
725 struct ifaddr *ifa = ifac->ifa;
726 struct in6_ifaddr *it6;
728 if (ifa->ifa_addr->sa_family != AF_INET6)
731 it6 = (struct in6_ifaddr *)ifa;
733 /* ignore no autoconf addresses. */
734 if (!(it6->ia6_flags & IN6_IFF_AUTOCONF))
737 /* ignore autoconf addresses with different prefixes. */
738 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
742 * Now we are looking at an autoconf address with the same
743 * prefix as ours. If the address is temporary and is still
744 * preferred, do not create another one. It would be rare, but
745 * could happen, for example, when we resume a laptop PC after
748 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) &&
749 !IFA6_IS_DEPRECATED(it6)) {
755 * This is a public autoconf address that has the same prefix
756 * as ours. If it is preferred, keep it. We can't break the
757 * loop here, because there may be a still-preferred temporary
758 * address with the prefix.
760 if (!IFA6_IS_DEPRECATED(it6))
764 if (public_ifa6 != NULL) {
767 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
768 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
769 " tmp addr,errno=%d\n", e);
779 * Nuke neighbor cache/prefix/default router management table, right before
783 nd6_purge(struct ifnet *ifp)
785 struct llinfo_nd6 *ln, *nln;
786 struct nd_defrouter *dr, *ndr, drany;
787 struct nd_prefix *pr, *npr;
789 /* Nuke default router list entries toward ifp */
790 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
792 * The first entry of the list may be stored in
793 * the routing table, so we'll delete it later.
795 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
796 ndr = TAILQ_NEXT(dr, dr_entry);
800 dr = TAILQ_FIRST(&nd_defrouter);
805 /* Nuke prefix list entries toward ifp */
806 for (pr = nd_prefix.lh_first; pr; pr = npr) {
808 if (pr->ndpr_ifp == ifp) {
810 * Previously, pr->ndpr_addr is removed as well,
811 * but I strongly believe we don't have to do it.
812 * nd6_purge() is only called from in6_ifdetach(),
813 * which removes all the associated interface addresses
815 * (jinmei@kame.net 20010129)
821 /* cancel default outgoing interface setting */
822 if (nd6_defifindex == ifp->if_index)
823 nd6_setdefaultiface(0);
825 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
826 /* refresh default router list */
827 bzero(&drany, sizeof(drany));
828 defrouter_delreq(&drany, 0);
833 * Nuke neighbor cache entries for the ifp.
834 * Note that rt->rt_ifp may not be the same as ifp,
835 * due to KAME goto ours hack. See RTM_RESOLVE case in
836 * nd6_rtrequest(), and ip6_input().
838 ln = llinfo_nd6.ln_next;
839 while (ln && ln != &llinfo_nd6) {
841 struct sockaddr_dl *sdl;
845 if (rt && rt->rt_gateway &&
846 rt->rt_gateway->sa_family == AF_LINK) {
847 sdl = (struct sockaddr_dl *)rt->rt_gateway;
848 if (sdl->sdl_index == ifp->if_index)
856 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp)
859 struct sockaddr_in6 sin6;
861 bzero(&sin6, sizeof(sin6));
862 sin6.sin6_len = sizeof(struct sockaddr_in6);
863 sin6.sin6_family = AF_INET6;
864 sin6.sin6_addr = *addr6;
867 rt = rtlookup((struct sockaddr *)&sin6);
869 rt = rtpurelookup((struct sockaddr *)&sin6);
870 if (rt && !(rt->rt_flags & RTF_LLINFO)) {
872 * This is the case for the default route.
873 * If we want to create a neighbor cache for the address, we
874 * should free the route for the destination and allocate an
887 * If no route is available and create is set,
888 * we allocate a host route for the destination
889 * and treat it like an interface route.
890 * This hack is necessary for a neighbor which can't
891 * be covered by our own prefix.
895 ifa = ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
900 * Create a new route. RTF_LLINFO is necessary
901 * to create a Neighbor Cache entry for the
902 * destination in nd6_rtrequest which will be
903 * called in rtrequest via ifa->ifa_rtrequest.
905 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
906 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
907 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
908 ~RTF_CLONING, &rt)) != 0) {
910 "nd6_lookup: failed to add route for a "
911 "neighbor(%s), errno=%d\n",
912 ip6_sprintf(addr6), e);
917 struct llinfo_nd6 *ln =
918 (struct llinfo_nd6 *)rt->rt_llinfo;
920 ln->ln_state = ND6_LLINFO_NOSTATE;
927 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) {
930 "nd6_lookup: failed to lookup %s (if = %s)\n",
931 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
932 /* xxx more logs... kazu */
939 static struct rtentry *
940 nd6_neighbor_lookup(struct in6_addr *addr6, struct ifnet *ifp)
943 struct sockaddr_in6 sin6;
945 bzero(&sin6, sizeof(sin6));
946 sin6.sin6_len = sizeof(struct sockaddr_in6);
947 sin6.sin6_family = AF_INET6;
948 sin6.sin6_addr = *addr6;
950 rt = rtpurelookup((struct sockaddr *)&sin6);
955 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) {
956 if (nd6_onlink_ns_rfc4861 &&
957 (ND6_RTENTRY_IS_LLCLONING(rt) || /* not cloned yet */
958 (rt->rt_parent != NULL && /* cloning */
959 ND6_RTENTRY_IS_LLCLONING(rt->rt_parent)))) {
961 * If cloning ever happened or is happening,
962 * rtentry for addr6 would or will become a
973 * Detect if a given IPv6 address identifies a neighbor on a given link.
974 * XXX: should take care of the destination of a p2p link?
977 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
979 struct ifaddr_container *ifac;
982 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
983 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
986 * A link-local address is always a neighbor.
987 * XXX: we should use the sin6_scope_id field rather than the embedded
990 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
991 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
995 * If the address matches one of our addresses,
996 * it should be a neighbor.
998 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
999 struct ifaddr *ifa = ifac->ifa;
1001 if (ifa->ifa_addr->sa_family != AF_INET6)
1004 for (i = 0; i < 4; i++) {
1005 if ((IFADDR6(ifa).s6_addr32[i] ^
1006 addr->sin6_addr.s6_addr32[i]) &
1007 IFMASK6(ifa).s6_addr32[i])
1014 * Even if the address matches none of our addresses, it might be
1015 * in the neighbor cache.
1017 if (nd6_neighbor_lookup(&addr->sin6_addr, ifp) != NULL)
1026 * Free an nd6 llinfo entry.
1029 nd6_free(struct rtentry *rt)
1031 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1032 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
1033 struct nd_defrouter *dr;
1036 * we used to have kpfctlinput(PRC_HOSTDEAD) here.
1037 * even though it is not harmful, it was not really necessary.
1040 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
1042 dr = defrouter_lookup(
1043 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1046 if (ln->ln_router || dr) {
1048 * rt6_flush must be called whether or not the neighbor
1049 * is in the Default Router List.
1050 * See a corresponding comment in nd6_na_input().
1052 rt6_flush(&in6, rt->rt_ifp);
1057 * Unreachablity of a router might affect the default
1058 * router selection and on-link detection of advertised
1063 * Temporarily fake the state to choose a new default
1064 * router and to perform on-link determination of
1065 * prefixes correctly.
1066 * Below the state will be set correctly,
1067 * or the entry itself will be deleted.
1069 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1072 * Since defrouter_select() does not affect the
1073 * on-link determination and MIP6 needs the check
1074 * before the default router selection, we perform
1077 pfxlist_onlink_check();
1079 if (dr == TAILQ_FIRST(&nd_defrouter)) {
1081 * It is used as the current default router,
1082 * so we have to move it to the end of the
1083 * list and choose a new one.
1084 * XXX: it is not very efficient if this is
1087 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1088 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
1093 mtx_unlock(&nd6_mtx);
1097 * Before deleting the entry, remember the next entry as the
1098 * return value. We need this because pfxlist_onlink_check() above
1099 * might have freed other entries (particularly the old next entry) as
1100 * a side effect (XXX).
1105 * Detach the route from the routing tree and the list of neighbor
1106 * caches, and disable the route entry not to be used in already
1109 * ND expiry happens under one big timer.
1110 * To avoid overflowing the route socket, don't report this.
1111 * Now that RTM_MISS is reported when an address is unresolvable
1112 * the benefit of reporting this deletion is questionable.
1114 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1120 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1122 * XXX cost-effective metods?
1125 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1127 struct llinfo_nd6 *ln;
1130 * If the caller specified "rt", use that. Otherwise, resolve the
1131 * routing table by supplied "dst6".
1136 if (!(rt = nd6_lookup(dst6, 0, NULL)))
1140 if ((rt->rt_flags & RTF_GATEWAY) ||
1141 !(rt->rt_flags & RTF_LLINFO) ||
1142 rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
1143 rt->rt_gateway->sa_family != AF_LINK) {
1144 /* This is not a host route. */
1148 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1149 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1153 * if we get upper-layer reachability confirmation many times,
1154 * it is possible we have false information.
1158 if (ln->ln_byhint > nd6_maxnudhint)
1162 ln->ln_state = ND6_LLINFO_REACHABLE;
1164 ln->ln_expire = time_uptime +
1165 ND_IFINFO(rt->rt_ifp)->reachable;
1169 nd6_rtrequest(int req, struct rtentry *rt)
1171 struct sockaddr *gate = rt->rt_gateway;
1172 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1173 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1174 struct ifnet *ifp = rt->rt_ifp;
1177 if ((rt->rt_flags & RTF_GATEWAY))
1180 if (nd6_need_cache(ifp) == 0 && !(rt->rt_flags & RTF_HOST)) {
1182 * This is probably an interface direct route for a link
1183 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1184 * We do not need special treatment below for such a route.
1185 * Moreover, the RTF_LLINFO flag which would be set below
1186 * would annoy the ndp(8) command.
1191 if (req == RTM_RESOLVE &&
1192 (nd6_need_cache(ifp) == 0 || /* stf case */
1193 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
1195 * FreeBSD and BSD/OS often make a cloned host route based
1196 * on a less-specific route (e.g. the default route).
1197 * If the less specific route does not have a "gateway"
1198 * (this is the case when the route just goes to a p2p or an
1199 * stf interface), we'll mistakenly make a neighbor cache for
1200 * the host route, and will see strange neighbor solicitation
1201 * for the corresponding destination. In order to avoid the
1202 * confusion, we check if the destination of the route is
1203 * a neighbor in terms of neighbor discovery, and stop the
1204 * process if not. Additionally, we remove the LLINFO flag
1205 * so that ndp(8) will not try to get the neighbor information
1206 * of the destination.
1208 rt->rt_flags &= ~RTF_LLINFO;
1215 * There is no backward compatibility :)
1217 * if (!(rt->rt_flags & RTF_HOST) &&
1218 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1219 * rt->rt_flags |= RTF_CLONING;
1221 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1223 * Case 1: This route should come from
1224 * a route to interface. RTF_LLINFO flag is set
1225 * for a host route whose destination should be
1226 * treated as on-link.
1228 rt_setgate(rt, rt_key(rt),
1229 (struct sockaddr *)&null_sdl);
1230 gate = rt->rt_gateway;
1231 SDL(gate)->sdl_type = ifp->if_type;
1232 SDL(gate)->sdl_index = ifp->if_index;
1234 ln->ln_expire = time_uptime;
1235 if (ln && ln->ln_expire == 0) {
1236 /* kludge for desktops */
1239 if ((rt->rt_flags & RTF_CLONING))
1243 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1244 * We don't do that here since llinfo is not ready yet.
1246 * There are also couple of other things to be discussed:
1247 * - unsolicited NA code needs improvement beforehand
1248 * - RFC2461 says we MAY send multicast unsolicited NA
1249 * (7.2.6 paragraph 4), however, it also says that we
1250 * SHOULD provide a mechanism to prevent multicast NA storm.
1251 * we don't have anything like it right now.
1252 * note that the mechanism needs a mutual agreement
1253 * between proxies, which means that we need to implement
1254 * a new protocol, or a new kludge.
1255 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1256 * we need to check ip6forwarding before sending it.
1257 * (or should we allow proxy ND configuration only for
1258 * routers? there's no mention about proxy ND from hosts)
1261 /* XXX it does not work */
1262 if ((rt->rt_flags & RTF_ANNOUNCE) && mycpuid == 0) {
1264 &SIN6(rt_key(rt))->sin6_addr,
1265 &SIN6(rt_key(rt))->sin6_addr,
1266 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1272 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1274 * Address resolution isn't necessary for a point to
1275 * point link, so we can skip this test for a p2p link.
1277 if (gate->sa_family != AF_LINK ||
1278 gate->sa_len < sizeof(null_sdl)) {
1280 "nd6_rtrequest: bad gateway value: %s\n",
1284 SDL(gate)->sdl_type = ifp->if_type;
1285 SDL(gate)->sdl_index = ifp->if_index;
1288 break; /* This happens on a route change */
1290 * Case 2: This route may come from cloning, or a manual route
1291 * add with a LL address.
1293 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1294 rt->rt_llinfo = (caddr_t)ln;
1296 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1301 bzero(ln, sizeof(*ln));
1303 /* this is required for "ndp" command. - shin */
1304 if (req == RTM_ADD) {
1306 * gate should have some valid AF_LINK entry,
1307 * and ln->ln_expire should have some lifetime
1308 * which is specified by ndp command.
1310 ln->ln_state = ND6_LLINFO_REACHABLE;
1314 * When req == RTM_RESOLVE, rt is created and
1315 * initialized in rtrequest(), so rt_expire is 0.
1317 ln->ln_state = ND6_LLINFO_NOSTATE;
1318 ln->ln_expire = time_uptime;
1320 rt->rt_flags |= RTF_LLINFO;
1321 ln->ln_next = llinfo_nd6.ln_next;
1322 llinfo_nd6.ln_next = ln;
1323 ln->ln_prev = &llinfo_nd6;
1324 ln->ln_next->ln_prev = ln;
1327 * check if rt_key(rt) is one of my address assigned
1330 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1331 &SIN6(rt_key(rt))->sin6_addr);
1333 caddr_t macp = nd6_ifptomac(ifp);
1335 ln->ln_state = ND6_LLINFO_REACHABLE;
1338 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1339 SDL(gate)->sdl_alen = ifp->if_addrlen;
1341 if (nd6_useloopback) {
1342 rt->rt_ifp = loif; /* XXX */
1344 * Make sure rt_ifa be equal to the ifaddr
1345 * corresponding to the address.
1346 * We need this because when we refer
1347 * rt_ifa->ia6_flags in ip6_input, we assume
1348 * that the rt_ifa points to the address instead
1349 * of the loopback address.
1351 if (ifa != rt->rt_ifa) {
1352 IFAFREE(rt->rt_ifa);
1357 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1359 ln->ln_state = ND6_LLINFO_REACHABLE;
1363 * Join solicited node multicast for proxy ND, and only
1364 * join it once on cpu0.
1366 if ((ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) {
1367 struct in6_addr llsol;
1370 llsol = SIN6(rt_key(rt))->sin6_addr;
1371 llsol.s6_addr16[0] = htons(0xff02);
1372 llsol.s6_addr16[1] = htons(ifp->if_index);
1373 llsol.s6_addr32[1] = 0;
1374 llsol.s6_addr32[2] = htonl(1);
1375 llsol.s6_addr8[12] = 0xff;
1377 if (!in6_addmulti(&llsol, ifp, &error)) {
1378 nd6log((LOG_ERR, "%s: failed to join "
1379 "%s (errno=%d)\n", if_name(ifp),
1380 ip6_sprintf(&llsol), error));
1390 * Leave from solicited node multicast for proxy ND, and only
1391 * leave it once on cpu0 (since we joined it once on cpu0).
1393 if ((rt->rt_flags & RTF_ANNOUNCE) &&
1394 (ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) {
1395 struct in6_addr llsol;
1396 struct in6_multi *in6m;
1398 llsol = SIN6(rt_key(rt))->sin6_addr;
1399 llsol.s6_addr16[0] = htons(0xff02);
1400 llsol.s6_addr16[1] = htons(ifp->if_index);
1401 llsol.s6_addr32[1] = 0;
1402 llsol.s6_addr32[2] = htonl(1);
1403 llsol.s6_addr8[12] = 0xff;
1405 in6m = IN6_LOOKUP_MULTI(&llsol, ifp);
1410 ln->ln_next->ln_prev = ln->ln_prev;
1411 ln->ln_prev->ln_next = ln->ln_next;
1414 rt->rt_flags &= ~RTF_LLINFO;
1416 m_freem(ln->ln_hold);
1422 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1424 struct in6_drlist *drl = (struct in6_drlist *)data;
1425 struct in6_prlist *prl = (struct in6_prlist *)data;
1426 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1427 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1428 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1429 struct nd_defrouter *dr, any;
1430 struct nd_prefix *pr;
1432 int i = 0, error = 0;
1435 case SIOCGDRLST_IN6:
1437 * obsolete API, use sysctl under net.inet6.icmp6
1439 bzero(drl, sizeof(*drl));
1441 dr = TAILQ_FIRST(&nd_defrouter);
1442 while (dr && i < DRLSTSIZ) {
1443 drl->defrouter[i].rtaddr = dr->rtaddr;
1444 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1445 /* XXX: need to this hack for KAME stack */
1446 drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1449 "default router list contains a "
1450 "non-linklocal address(%s)\n",
1451 ip6_sprintf(&drl->defrouter[i].rtaddr));
1453 drl->defrouter[i].flags = dr->flags;
1454 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1455 drl->defrouter[i].expire = dr->expire;
1456 drl->defrouter[i].if_index = dr->ifp->if_index;
1458 dr = TAILQ_NEXT(dr, dr_entry);
1460 mtx_unlock(&nd6_mtx);
1462 case SIOCGPRLST_IN6:
1464 * obsolete API, use sysctl under net.inet6.icmp6
1467 * XXX meaning of fields, especialy "raflags", is very
1468 * differnet between RA prefix list and RR/static prefix list.
1469 * how about separating ioctls into two?
1471 bzero(prl, sizeof(*prl));
1473 pr = nd_prefix.lh_first;
1474 while (pr && i < PRLSTSIZ) {
1475 struct nd_pfxrouter *pfr;
1478 in6_embedscope(&prl->prefix[i].prefix,
1479 &pr->ndpr_prefix, NULL, NULL);
1480 prl->prefix[i].raflags = pr->ndpr_raf;
1481 prl->prefix[i].prefixlen = pr->ndpr_plen;
1482 prl->prefix[i].vltime = pr->ndpr_vltime;
1483 prl->prefix[i].pltime = pr->ndpr_pltime;
1484 prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1485 prl->prefix[i].expire = pr->ndpr_expire;
1487 pfr = pr->ndpr_advrtrs.lh_first;
1491 #define RTRADDR prl->prefix[i].advrtr[j]
1492 RTRADDR = pfr->router->rtaddr;
1493 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1494 /* XXX: hack for KAME */
1495 RTRADDR.s6_addr16[1] = 0;
1498 "a router(%s) advertises "
1500 "non-link local address\n",
1501 ip6_sprintf(&RTRADDR));
1505 pfr = pfr->pfr_next;
1507 prl->prefix[i].advrtrs = j;
1508 prl->prefix[i].origin = PR_ORIG_RA;
1513 mtx_unlock(&nd6_mtx);
1516 case OSIOCGIFINFO_IN6:
1517 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1518 bzero(&ndi->ndi, sizeof(ndi->ndi));
1519 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
1520 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu;
1521 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable;
1522 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable;
1523 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans;
1524 ndi->ndi.flags = ND_IFINFO(ifp)->flags;
1525 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm;
1526 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim;
1528 case SIOCGIFINFO_IN6:
1529 ndi->ndi = *ND_IFINFO(ifp);
1530 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
1532 case SIOCSIFINFO_IN6:
1534 * used to change host variables from userland.
1535 * intented for a use on router to reflect RA configurations.
1537 /* 0 means 'unspecified' */
1538 if (ndi->ndi.linkmtu != 0) {
1539 if (ndi->ndi.linkmtu < IPV6_MMTU ||
1540 ndi->ndi.linkmtu > IN6_LINKMTU(ifp)) {
1544 ND_IFINFO(ifp)->linkmtu = ndi->ndi.linkmtu;
1547 if (ndi->ndi.basereachable != 0) {
1548 int obasereachable = ND_IFINFO(ifp)->basereachable;
1550 ND_IFINFO(ifp)->basereachable = ndi->ndi.basereachable;
1551 if (ndi->ndi.basereachable != obasereachable)
1552 ND_IFINFO(ifp)->reachable =
1553 ND_COMPUTE_RTIME(ndi->ndi.basereachable);
1555 if (ndi->ndi.retrans != 0)
1556 ND_IFINFO(ifp)->retrans = ndi->ndi.retrans;
1557 if (ndi->ndi.chlim != 0)
1558 ND_IFINFO(ifp)->chlim = ndi->ndi.chlim;
1560 case SIOCSIFINFO_FLAGS:
1561 ND_IFINFO(ifp)->flags = ndi->ndi.flags;
1563 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1564 /* flush default router list */
1566 * xxx sumikawa: should not delete route if default
1567 * route equals to the top of default router list
1569 bzero(&any, sizeof(any));
1570 defrouter_delreq(&any, 0);
1572 /* xxx sumikawa: flush prefix list */
1574 case SIOCSPFXFLUSH_IN6:
1576 /* flush all the prefix advertised by routers */
1577 struct nd_prefix *pr, *next;
1580 for (pr = nd_prefix.lh_first; pr; pr = next) {
1581 struct in6_ifaddr *ia, *ia_next;
1583 next = pr->ndpr_next;
1585 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1588 /* do we really have to remove addresses as well? */
1589 for (ia = in6_ifaddr; ia; ia = ia_next) {
1590 /* ia might be removed. keep the next ptr. */
1591 ia_next = ia->ia_next;
1593 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
1596 if (ia->ia6_ndpr == pr)
1597 in6_purgeaddr(&ia->ia_ifa);
1601 mtx_unlock(&nd6_mtx);
1604 case SIOCSRTRFLUSH_IN6:
1606 /* flush all the default routers */
1607 struct nd_defrouter *dr, *next;
1610 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1612 * The first entry of the list may be stored in
1613 * the routing table, so we'll delete it later.
1615 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1616 next = TAILQ_NEXT(dr, dr_entry);
1619 defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
1621 mtx_unlock(&nd6_mtx);
1624 case SIOCGNBRINFO_IN6:
1626 struct llinfo_nd6 *ln;
1627 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1630 * XXX: KAME specific hack for scoped addresses
1631 * XXXX: for other scopes than link-local?
1633 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1634 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1635 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1638 *idp = htons(ifp->if_index);
1642 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1644 mtx_unlock(&nd6_mtx);
1647 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1648 nbi->state = ln->ln_state;
1649 nbi->asked = ln->ln_asked;
1650 nbi->isrouter = ln->ln_router;
1651 nbi->expire = ln->ln_expire;
1652 mtx_unlock(&nd6_mtx);
1656 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1657 ndif->ifindex = nd6_defifindex;
1659 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1660 return (nd6_setdefaultiface(ndif->ifindex));
1666 * Create neighbor cache entry and cache link-layer address,
1667 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1670 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1672 int type, /* ICMP6 type */
1673 int code /* type dependent information */)
1675 struct rtentry *rt = NULL;
1676 struct llinfo_nd6 *ln = NULL;
1678 struct sockaddr_dl *sdl = NULL;
1685 panic("ifp == NULL in nd6_cache_lladdr");
1687 panic("from == NULL in nd6_cache_lladdr");
1689 /* nothing must be updated for unspecified address */
1690 if (IN6_IS_ADDR_UNSPECIFIED(from))
1694 * Validation about ifp->if_addrlen and lladdrlen must be done in
1697 * XXX If the link does not have link-layer adderss, what should
1698 * we do? (ifp->if_addrlen == 0)
1699 * Spec says nothing in sections for RA, RS and NA. There's small
1700 * description on it in NS section (RFC 2461 7.2.3).
1703 rt = nd6_lookup(from, 0, ifp);
1706 /* nothing must be done if there's no lladdr */
1707 if (!lladdr || !lladdrlen)
1711 rt = nd6_lookup(from, 1, ifp);
1714 /* do nothing if static ndp is set */
1715 if (rt->rt_flags & RTF_STATIC)
1722 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1727 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1730 if (!rt->rt_gateway)
1732 if (rt->rt_gateway->sa_family != AF_LINK)
1734 sdl = SDL(rt->rt_gateway);
1736 olladdr = (sdl->sdl_alen) ? 1 : 0;
1737 if (olladdr && lladdr) {
1738 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1746 * newentry olladdr lladdr llchange (*=record)
1749 * 0 n y -- (3) * STALE
1751 * 0 y y y (5) * STALE
1752 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1753 * 1 -- y -- (7) * STALE
1756 if (lladdr) { /* (3-5) and (7) */
1758 * Record source link-layer address
1759 * XXX is it dependent to ifp->if_type?
1761 sdl->sdl_alen = ifp->if_addrlen;
1762 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1766 if ((!olladdr && lladdr) || /* (3) */
1767 (olladdr && lladdr && llchange)) { /* (5) */
1769 newstate = ND6_LLINFO_STALE;
1770 } else { /* (1-2,4) */
1775 if (!lladdr) /* (6) */
1776 newstate = ND6_LLINFO_NOSTATE;
1778 newstate = ND6_LLINFO_STALE;
1783 * Update the state of the neighbor cache.
1785 ln->ln_state = newstate;
1787 if (ln->ln_state == ND6_LLINFO_STALE) {
1789 * XXX: since nd6_output() below will cause
1790 * state tansition to DELAY and reset the timer,
1791 * we must set the timer now, although it is actually
1794 ln->ln_expire = time_uptime + nd6_gctimer;
1798 * we assume ifp is not a p2p here, so just
1799 * set the 2nd argument as the 1st one.
1801 nd6_output(ifp, ifp, ln->ln_hold,
1802 (struct sockaddr_in6 *)rt_key(rt), rt);
1805 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1806 /* probe right away */
1807 ln->ln_expire = time_uptime;
1812 * ICMP6 type dependent behavior.
1814 * NS: clear IsRouter if new entry
1815 * RS: clear IsRouter
1816 * RA: set IsRouter if there's lladdr
1817 * redir: clear IsRouter if new entry
1820 * The spec says that we must set IsRouter in the following cases:
1821 * - If lladdr exist, set IsRouter. This means (1-5).
1822 * - If it is old entry (!newentry), set IsRouter. This means (7).
1823 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1824 * A quetion arises for (1) case. (1) case has no lladdr in the
1825 * neighbor cache, this is similar to (6).
1826 * This case is rare but we figured that we MUST NOT set IsRouter.
1828 * newentry olladdr lladdr llchange NS RS RA redir
1830 * 0 n n -- (1) c ? s
1831 * 0 y n -- (2) c s s
1832 * 0 n y -- (3) c s s
1835 * 1 -- n -- (6) c c c s
1836 * 1 -- y -- (7) c c s c s
1840 switch (type & 0xff) {
1841 case ND_NEIGHBOR_SOLICIT:
1843 * New entry must have is_router flag cleared.
1845 if (is_newentry) /* (6-7) */
1850 * If the icmp is a redirect to a better router, always set the
1851 * is_router flag. Otherwise, if the entry is newly created,
1852 * clear the flag. [RFC 2461, sec 8.3]
1854 if (code == ND_REDIRECT_ROUTER)
1856 else if (is_newentry) /* (6-7) */
1859 case ND_ROUTER_SOLICIT:
1861 * is_router flag must always be cleared.
1865 case ND_ROUTER_ADVERT:
1867 * Mark an entry with lladdr as a router.
1869 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1870 (is_newentry && lladdr)) { /* (7) */
1876 if (llchange || lladdr)
1877 rt_rtmsg(llchange ? RTM_CHANGE : RTM_ADD, rt, rt->rt_ifp, 0);
1880 * When the link-layer address of a router changes, select the
1881 * best router again. In particular, when the neighbor entry is newly
1882 * created, it might affect the selection policy.
1883 * Question: can we restrict the first condition to the "is_newentry"
1885 * XXX: when we hear an RA from a new router with the link-layer
1886 * address option, defrouter_select() is called twice, since
1887 * defrtrlist_update called the function as well. However, I believe
1888 * we can compromise the overhead, since it only happens the first
1890 * XXX: although defrouter_select() should not have a bad effect
1891 * for those are not autoconfigured hosts, we explicitly avoid such
1894 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1901 nd6_slowtimo(void *arg __unused)
1903 struct lwkt_msg *lmsg = &nd6_slowtimo_netmsg.lmsg;
1905 KASSERT(mycpuid == 0, ("not on cpu0"));
1907 if (lmsg->ms_flags & MSGF_DONE)
1908 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg);
1913 nd6_slowtimo_dispatch(netmsg_t nmsg)
1915 const struct ifnet_array *arr;
1916 struct nd_ifinfo *nd6if;
1922 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
1925 arr = ifnet_array_get();
1928 for (i = 0; i < arr->ifnet_count; ++i) {
1929 struct ifnet *ifp = arr->ifnet_arr[i];
1931 if (ifp->if_afdata[AF_INET6] == NULL)
1933 nd6if = ND_IFINFO(ifp);
1934 if (nd6if->basereachable && /* already initialized */
1935 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1937 * Since reachable time rarely changes by router
1938 * advertisements, we SHOULD insure that a new random
1939 * value gets recomputed at least once every few hours.
1942 nd6if->recalctm = nd6_recalc_reachtm_interval;
1943 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1946 mtx_unlock(&nd6_mtx);
1948 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1949 nd6_slowtimo, NULL);
1953 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1954 struct sockaddr_in6 *dst, struct rtentry *rt)
1958 if (ifp->if_flags & IFF_LOOPBACK)
1959 error = ifp->if_output(origifp, m, (struct sockaddr *)dst, rt);
1961 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, rt);
1966 nd6_resolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
1967 struct sockaddr *dst0, u_char *desten)
1969 struct sockaddr_in6 *dst = SIN6(dst0);
1970 struct rtentry *rt = NULL;
1971 struct llinfo_nd6 *ln = NULL;
1974 if (m->m_flags & M_MCAST) {
1975 switch (ifp->if_type) {
1980 #ifdef IFT_IEEE80211
1983 ETHER_MAP_IPV6_MULTICAST(&dst->sin6_addr,
1987 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen);
1990 error = EAFNOSUPPORT;
1996 error = rt_llroute(dst0, rt0, &rt);
2003 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2004 * the condition below is not very efficient. But we believe
2005 * it is tolerable, because this should be a rare case.
2007 if (ln == NULL && nd6_is_addr_neighbor(dst, ifp)) {
2008 rt = nd6_lookup(&dst->sin6_addr, 1, ifp);
2013 if (ln == NULL || rt == NULL) {
2014 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
2015 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2017 "nd6_output: can't allocate llinfo for %s "
2019 ip6_sprintf(&dst->sin6_addr), ln, rt);
2026 /* We don't have to do link-layer address resolution on a p2p link. */
2027 if ((ifp->if_flags & IFF_POINTOPOINT) &&
2028 ln->ln_state < ND6_LLINFO_REACHABLE) {
2029 ln->ln_state = ND6_LLINFO_STALE;
2030 ln->ln_expire = time_uptime + nd6_gctimer;
2034 * The first time we send a packet to a neighbor whose entry is
2035 * STALE, we have to change the state to DELAY and a sets a timer to
2036 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2037 * neighbor unreachability detection on expiration.
2040 if (ln->ln_state == ND6_LLINFO_STALE) {
2042 ln->ln_state = ND6_LLINFO_DELAY;
2043 ln->ln_expire = time_uptime + nd6_delay;
2047 * If the neighbor cache entry has a state other than INCOMPLETE
2048 * (i.e. its link-layer address is already resolved), return it.
2050 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) {
2051 struct sockaddr_dl *sdl = SDL(rt->rt_gateway);
2053 KKASSERT(sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0);
2054 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2059 * There is a neighbor cache entry, but no ethernet address
2060 * response yet. Replace the held mbuf (if any) with this
2064 m_freem(ln->ln_hold);
2068 * This code conforms to the rate-limiting rule described in Section
2069 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
2072 if (ln->ln_state == ND6_LLINFO_NOSTATE ||
2073 ln->ln_state == ND6_LLINFO_WAITDELETE) {
2075 * This neighbor cache entry was just created; change its
2076 * state to INCOMPLETE and start its life cycle.
2078 * We force an NS output below by setting ln_expire to 1
2079 * (nd6_rtrequest() could set it to the current time_uptime)
2080 * and zeroing out ln_asked (XXX this may not be necessary).
2082 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2086 if (ln->ln_expire && ln->ln_expire < time_uptime && ln->ln_asked == 0) {
2088 ln->ln_expire = time_uptime + ND_IFINFO(ifp)->retrans / 1000;
2089 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2092 if (ln->ln_asked >= nd6_mmaxtries)
2093 return (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
2094 EHOSTUNREACH : EHOSTDOWN;
2103 nd6_need_cache(struct ifnet *ifp)
2106 * XXX: we currently do not make neighbor cache on any interface
2107 * other than Ethernet and GIF.
2110 * - unidirectional tunnels needs no ND
2112 switch (ifp->if_type) {
2118 #ifdef IFT_IEEE80211
2124 case IFT_GIF: /* XXX need more cases? */
2131 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2132 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2134 SYSCTL_DECL(_net_inet6_icmp6);
2136 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2137 CTLFLAG_RD, nd6_sysctl_drlist, "List default routers");
2138 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2139 CTLFLAG_RD, nd6_sysctl_prlist, "List prefixes");
2142 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2146 struct in6_defrouter *d, *de;
2147 struct nd_defrouter *dr;
2153 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
2154 dr = TAILQ_NEXT(dr, dr_entry)) {
2155 d = (struct in6_defrouter *)buf;
2156 de = (struct in6_defrouter *)(buf + sizeof(buf));
2159 bzero(d, sizeof(*d));
2160 d->rtaddr.sin6_family = AF_INET6;
2161 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2162 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2166 "default router list (%s)\n",
2167 ip6_sprintf(&dr->rtaddr));
2168 d->flags = dr->flags;
2169 d->rtlifetime = dr->rtlifetime;
2170 d->expire = dr->expire;
2171 d->if_index = dr->ifp->if_index;
2173 panic("buffer too short");
2175 error = SYSCTL_OUT(req, buf, sizeof(*d));
2183 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2187 struct in6_prefix *p, *pe;
2188 struct nd_prefix *pr;
2194 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2197 struct sockaddr_in6 *sin6, *s6;
2198 struct nd_pfxrouter *pfr;
2200 p = (struct in6_prefix *)buf;
2201 pe = (struct in6_prefix *)(buf + sizeof(buf));
2204 bzero(p, sizeof(*p));
2205 sin6 = (struct sockaddr_in6 *)(p + 1);
2207 p->prefix = pr->ndpr_prefix;
2208 if (in6_recoverscope(&p->prefix,
2209 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2211 "scope error in prefix list (%s)\n",
2212 ip6_sprintf(&p->prefix.sin6_addr));
2213 p->raflags = pr->ndpr_raf;
2214 p->prefixlen = pr->ndpr_plen;
2215 p->vltime = pr->ndpr_vltime;
2216 p->pltime = pr->ndpr_pltime;
2217 p->if_index = pr->ndpr_ifp->if_index;
2218 p->expire = pr->ndpr_expire;
2219 p->refcnt = pr->ndpr_refcnt;
2220 p->flags = pr->ndpr_stateflags;
2221 p->origin = PR_ORIG_RA;
2223 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2224 pfr = pfr->pfr_next) {
2225 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2229 s6 = &sin6[advrtrs];
2230 bzero(s6, sizeof(*s6));
2231 s6->sin6_family = AF_INET6;
2232 s6->sin6_len = sizeof(*sin6);
2233 if (in6_recoverscope(s6, &pfr->router->rtaddr,
2234 pfr->router->ifp) != 0)
2237 "prefix list (%s)\n",
2238 ip6_sprintf(&pfr->router->rtaddr));
2241 p->advrtrs = advrtrs;
2243 panic("buffer too short");
2246 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2247 error = SYSCTL_OUT(req, buf, advance);