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"
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/callout.h>
41 #include <sys/malloc.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
46 #include <sys/kernel.h>
47 #include <sys/protosw.h>
48 #include <sys/errno.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
51 #include <sys/sysctl.h>
54 #include <net/if_arc.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/iso88025.h>
59 #include <net/route.h>
61 #include <netinet/in.h>
62 #include <netinet/if_ether.h>
63 #include <netinet6/in6_var.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet6/scope6_var.h>
67 #include <netinet6/nd6.h>
68 #include <netinet/icmp6.h>
70 #include <sys/limits.h>
72 #include <net/net_osdep.h>
74 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
75 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
77 #define SIN6(s) ((struct sockaddr_in6 *)s)
78 #define SDL(s) ((struct sockaddr_dl *)s)
81 int nd6_prune = 1; /* walk list every 1 seconds */
82 int nd6_delay = 5; /* delay first probe time 5 second */
83 int nd6_umaxtries = 3; /* maximum unicast query */
84 int nd6_mmaxtries = 3; /* maximum multicast query */
85 int nd6_useloopback = 1; /* use loopback interface for local traffic */
86 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
88 /* preventing too many loops in ND option parsing */
89 int nd6_maxndopt = 10; /* max # of ND options allowed */
91 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
92 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */
101 static int nd6_inuse, nd6_allocated;
103 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
104 struct nd_drhead nd_defrouter;
105 struct nd_prhead nd_prefix = { 0 };
107 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
108 static struct sockaddr_in6 all1_sa;
110 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
112 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *));
113 static void nd6_slowtimo __P((void *));
114 static int regen_tmpaddr __P((struct in6_ifaddr *));
115 static struct llinfo_nd6 *nd6_free __P((struct rtentry *, int));
116 static void nd6_llinfo_timer __P((void *));
117 static void clear_llinfo_pqueue __P((struct llinfo_nd6 *));
119 struct callout nd6_slowtimo_ch;
120 struct callout nd6_timer_ch;
121 extern struct callout in6_tmpaddrtimer_ch;
126 static int nd6_init_done = 0;
130 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
134 all1_sa.sin6_family = AF_INET6;
135 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
136 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
137 all1_sa.sin6_addr.s6_addr[i] = 0xff;
139 /* initialization of the default router list */
140 TAILQ_INIT(&nd_defrouter);
145 callout_init(&nd6_slowtimo_ch, 0);
146 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
154 struct nd_ifinfo *nd;
156 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
157 bzero(nd, sizeof(*nd));
161 nd->chlim = IPV6_DEFHLIM;
162 nd->basereachable = REACHABLE_TIME;
163 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
164 nd->retrans = RETRANS_TIMER;
166 * Note that the default value of ip6_accept_rtadv is 0, which means
167 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
170 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
172 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
173 nd6_setmtu0(ifp, nd);
180 struct nd_ifinfo *nd;
187 * Reset ND level link MTU. This function is called when the physical MTU
188 * changes, which means we might have to adjust the ND level MTU.
195 nd6_setmtu0(ifp, ND_IFINFO(ifp));
198 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
200 nd6_setmtu0(ifp, ndi)
202 struct nd_ifinfo *ndi;
206 omaxmtu = ndi->maxmtu;
208 switch (ifp->if_type) {
210 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
213 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
216 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
219 ndi->maxmtu = ifp->if_mtu;
224 * Decreasing the interface MTU under IPV6 minimum MTU may cause
225 * undesirable situation. We thus notify the operator of the change
226 * explicitly. The check for omaxmtu is necessary to restrict the
227 * log to the case of changing the MTU, not initializing it.
229 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
230 log(LOG_NOTICE, "nd6_setmtu0: "
231 "new link MTU on %s (%lu) is too small for IPv6\n",
232 if_name(ifp), (unsigned long)ndi->maxmtu);
235 if (ndi->maxmtu > in6_maxmtu)
236 in6_setmaxmtu(); /* check all interfaces just in case */
242 nd6_option_init(opt, icmp6len, ndopts)
245 union nd_opts *ndopts;
248 bzero(ndopts, sizeof(*ndopts));
249 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
251 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
254 ndopts->nd_opts_done = 1;
255 ndopts->nd_opts_search = NULL;
260 * Take one ND option.
264 union nd_opts *ndopts;
266 struct nd_opt_hdr *nd_opt;
270 panic("ndopts == NULL in nd6_option");
271 if (ndopts->nd_opts_last == NULL)
272 panic("uninitialized ndopts in nd6_option");
273 if (ndopts->nd_opts_search == NULL)
275 if (ndopts->nd_opts_done)
278 nd_opt = ndopts->nd_opts_search;
280 /* make sure nd_opt_len is inside the buffer */
281 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
282 bzero(ndopts, sizeof(*ndopts));
286 olen = nd_opt->nd_opt_len << 3;
289 * Message validation requires that all included
290 * options have a length that is greater than zero.
292 bzero(ndopts, sizeof(*ndopts));
296 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
297 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
298 /* option overruns the end of buffer, invalid */
299 bzero(ndopts, sizeof(*ndopts));
301 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
302 /* reached the end of options chain */
303 ndopts->nd_opts_done = 1;
304 ndopts->nd_opts_search = NULL;
310 * Parse multiple ND options.
311 * This function is much easier to use, for ND routines that do not need
312 * multiple options of the same type.
316 union nd_opts *ndopts;
318 struct nd_opt_hdr *nd_opt;
322 panic("ndopts == NULL in nd6_options");
323 if (ndopts->nd_opts_last == NULL)
324 panic("uninitialized ndopts in nd6_options");
325 if (ndopts->nd_opts_search == NULL)
329 nd_opt = nd6_option(ndopts);
330 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
332 * Message validation requires that all included
333 * options have a length that is greater than zero.
335 icmp6stat.icp6s_nd_badopt++;
336 bzero(ndopts, sizeof(*ndopts));
343 switch (nd_opt->nd_opt_type) {
344 case ND_OPT_SOURCE_LINKADDR:
345 case ND_OPT_TARGET_LINKADDR:
347 case ND_OPT_REDIRECTED_HEADER:
348 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
350 "duplicated ND6 option found (type=%d)\n",
351 nd_opt->nd_opt_type));
354 ndopts->nd_opt_array[nd_opt->nd_opt_type]
358 case ND_OPT_PREFIX_INFORMATION:
359 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
360 ndopts->nd_opt_array[nd_opt->nd_opt_type]
363 ndopts->nd_opts_pi_end =
364 (struct nd_opt_prefix_info *)nd_opt;
368 * Unknown options must be silently ignored,
369 * to accomodate future extension to the protocol.
372 "nd6_options: unsupported option %d - "
373 "option ignored\n", nd_opt->nd_opt_type));
378 if (i > nd6_maxndopt) {
379 icmp6stat.icp6s_nd_toomanyopt++;
380 nd6log((LOG_INFO, "too many loop in nd opt\n"));
384 if (ndopts->nd_opts_done)
392 * ND6 timer routine to handle ND6 entries
395 nd6_llinfo_settimer(ln, tick)
396 struct llinfo_nd6 *ln;
402 callout_stop(&ln->ln_timer_ch);
404 ln->ln_expire = time_second + tick / hz;
405 if (tick > INT_MAX) {
406 ln->ln_ntick = tick - INT_MAX;
407 callout_reset(&ln->ln_timer_ch, INT_MAX,
408 nd6_llinfo_timer, ln);
411 callout_reset(&ln->ln_timer_ch, tick,
412 nd6_llinfo_timer, ln);
418 nd6_llinfo_timer(arg)
421 struct llinfo_nd6 *ln;
423 struct in6_addr *dst;
425 struct nd_ifinfo *ndi = NULL;
427 ln = (struct llinfo_nd6 *)arg;
429 if (ln->ln_ntick > 0) {
430 if (ln->ln_ntick > INT_MAX) {
431 ln->ln_ntick -= INT_MAX;
432 nd6_llinfo_settimer(ln, INT_MAX);
435 nd6_llinfo_settimer(ln, ln->ln_ntick);
440 if ((rt = ln->ln_rt) == NULL)
441 panic("ln->ln_rt == NULL");
442 if ((ifp = rt->rt_ifp) == NULL)
443 panic("ln->ln_rt->rt_ifp == NULL");
444 ndi = ND_IFINFO(ifp);
447 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
448 panic("rt_llinfo(%p) is not equal to ln(%p)",
450 if (rt_key(rt) == NULL)
451 panic("rt key is NULL in nd6_timer(ln=%p)", ln);
453 dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
455 switch (ln->ln_state) {
456 case ND6_LLINFO_INCOMPLETE:
457 if (ln->ln_asked < nd6_mmaxtries) {
459 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
460 nd6_ns_output(ifp, NULL, dst, ln, 0);
462 struct mbuf *m = ln->ln_hold;
467 * assuming every packet in ln_hold has the
472 icmp6_error2(m, ICMP6_DST_UNREACH,
473 ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp);
476 clear_llinfo_pqueue(ln);
479 (void)nd6_free(rt, 0);
483 case ND6_LLINFO_REACHABLE:
484 if (!ND6_LLINFO_PERMANENT(ln)) {
485 ln->ln_state = ND6_LLINFO_STALE;
486 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
490 case ND6_LLINFO_STALE:
491 /* Garbage Collection(RFC 2461 5.3) */
492 if (!ND6_LLINFO_PERMANENT(ln)) {
493 (void)nd6_free(rt, 1);
498 case ND6_LLINFO_DELAY:
499 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
502 ln->ln_state = ND6_LLINFO_PROBE;
503 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
504 nd6_ns_output(ifp, dst, dst, ln, 0);
506 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
507 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
510 case ND6_LLINFO_PROBE:
511 if (ln->ln_asked < nd6_umaxtries) {
513 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
514 nd6_ns_output(ifp, dst, dst, ln, 0);
516 (void)nd6_free(rt, 0);
525 * ND6 timer routine to expire default route list and prefix list
528 nd6_timer(ignored_arg)
532 struct nd_defrouter *dr;
533 struct nd_prefix *pr;
534 struct in6_ifaddr *ia6, *nia6;
535 struct in6_addrlifetime *lt6;
537 callout_reset(&nd6_timer_ch, nd6_prune * hz,
540 /* expire default router list */
542 dr = TAILQ_FIRST(&nd_defrouter);
544 if (dr->expire && dr->expire < time_second) {
545 struct nd_defrouter *t;
546 t = TAILQ_NEXT(dr, dr_entry);
550 dr = TAILQ_NEXT(dr, dr_entry);
555 * expire interface addresses.
556 * in the past the loop was inside prefix expiry processing.
557 * However, from a stricter speci-confrmance standpoint, we should
558 * rather separate address lifetimes and prefix lifetimes.
561 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
563 /* check address lifetime */
564 lt6 = &ia6->ia6_lifetime;
565 if (IFA6_IS_INVALID(ia6)) {
569 * If the expiring address is temporary, try
570 * regenerating a new one. This would be useful when
571 * we suspended a laptop PC, then turned it on after a
572 * period that could invalidate all temporary
573 * addresses. Although we may have to restart the
574 * loop (see below), it must be after purging the
575 * address. Otherwise, we'd see an infinite loop of
578 if (ip6_use_tempaddr &&
579 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
580 if (regen_tmpaddr(ia6) == 0)
584 in6_purgeaddr(&ia6->ia_ifa);
587 goto addrloop; /* XXX: see below */
588 } else if (IFA6_IS_DEPRECATED(ia6)) {
589 int oldflags = ia6->ia6_flags;
591 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
594 * If a temporary address has just become deprecated,
595 * regenerate a new one if possible.
597 if (ip6_use_tempaddr &&
598 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
599 (oldflags & IN6_IFF_DEPRECATED) == 0) {
601 if (regen_tmpaddr(ia6) == 0) {
603 * A new temporary address is
605 * XXX: this means the address chain
606 * has changed while we are still in
607 * the loop. Although the change
608 * would not cause disaster (because
609 * it's not a deletion, but an
610 * addition,) we'd rather restart the
611 * loop just for safety. Or does this
612 * significantly reduce performance??
619 * A new RA might have made a deprecated address
622 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
626 /* expire prefix list */
627 pr = nd_prefix.lh_first;
630 * check prefix lifetime.
631 * since pltime is just for autoconf, pltime processing for
632 * prefix is not necessary.
634 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
635 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
640 * address expiration and prefix expiration are
641 * separate. NEVER perform in6_purgeaddr here.
654 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
658 struct in6_ifaddr *public_ifa6 = NULL;
660 ifp = ia6->ia_ifa.ifa_ifp;
661 for (ifa = ifp->if_addrlist.tqh_first; ifa;
662 ifa = ifa->ifa_list.tqe_next) {
663 struct in6_ifaddr *it6;
665 if (ifa->ifa_addr->sa_family != AF_INET6)
668 it6 = (struct in6_ifaddr *)ifa;
670 /* ignore no autoconf addresses. */
671 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
674 /* ignore autoconf addresses with different prefixes. */
675 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
679 * Now we are looking at an autoconf address with the same
680 * prefix as ours. If the address is temporary and is still
681 * preferred, do not create another one. It would be rare, but
682 * could happen, for example, when we resume a laptop PC after
685 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
686 !IFA6_IS_DEPRECATED(it6)) {
692 * This is a public autoconf address that has the same prefix
693 * as ours. If it is preferred, keep it. We can't break the
694 * loop here, because there may be a still-preferred temporary
695 * address with the prefix.
697 if (!IFA6_IS_DEPRECATED(it6))
701 if (public_ifa6 != NULL) {
704 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
705 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
706 " tmp addr,errno=%d\n", e);
716 * Nuke neighbor cache/prefix/default router management table, right before
723 struct llinfo_nd6 *ln, *nln;
724 struct nd_defrouter *dr, *ndr;
725 struct nd_prefix *pr, *npr;
728 * Nuke default router list entries toward ifp.
729 * We defer removal of default router list entries that is installed
730 * in the routing table, in order to keep additional side effects as
733 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
734 ndr = TAILQ_NEXT(dr, dr_entry);
742 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = ndr) {
743 ndr = TAILQ_NEXT(dr, dr_entry);
751 /* Nuke prefix list entries toward ifp */
752 for (pr = nd_prefix.lh_first; pr; pr = npr) {
754 if (pr->ndpr_ifp == ifp) {
756 * Because if_detach() does *not* release prefixes
757 * while purging addresses the reference count will
758 * still be above zero. We therefore reset it to
759 * make sure that the prefix really gets purged.
764 * Previously, pr->ndpr_addr is removed as well,
765 * but I strongly believe we don't have to do it.
766 * nd6_purge() is only called from in6_ifdetach(),
767 * which removes all the associated interface addresses
769 * (jinmei@kame.net 20010129)
775 /* cancel default outgoing interface setting */
776 if (nd6_defifindex == ifp->if_index)
777 nd6_setdefaultiface(0);
779 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
780 /* refresh default router list */
785 * Nuke neighbor cache entries for the ifp.
786 * Note that rt->rt_ifp may not be the same as ifp,
787 * due to KAME goto ours hack. See RTM_RESOLVE case in
788 * nd6_rtrequest(), and ip6_input().
790 ln = llinfo_nd6.ln_next;
791 while (ln && ln != &llinfo_nd6) {
793 struct sockaddr_dl *sdl;
797 if (rt && rt->rt_gateway &&
798 rt->rt_gateway->sa_family == AF_LINK) {
799 sdl = (struct sockaddr_dl *)rt->rt_gateway;
800 if (sdl->sdl_index == ifp->if_index)
801 nln = nd6_free(rt, 0);
808 nd6_lookup(addr6, create, ifp)
809 struct in6_addr *addr6;
814 struct sockaddr_in6 sin6;
816 bzero(&sin6, sizeof(sin6));
817 sin6.sin6_len = sizeof(struct sockaddr_in6);
818 sin6.sin6_family = AF_INET6;
819 sin6.sin6_addr = *addr6;
820 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
822 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) {
824 * This is the case for the default route.
825 * If we want to create a neighbor cache for the
826 * address, we should free the route for the
827 * destination and allocate an interface route.
838 * If no route is available and create is set,
839 * we allocate a host route for the destination
840 * and treat it like an interface route.
841 * This hack is necessary for a neighbor which can't
842 * be covered by our own prefix.
845 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
850 * Create a new route. RTF_LLINFO is necessary
851 * to create a Neighbor Cache entry for the
852 * destination in nd6_rtrequest which will be
853 * called in rtrequest via ifa->ifa_rtrequest.
855 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
856 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
857 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
858 ~RTF_CLONING, &rt)) != 0) {
860 "nd6_lookup: failed to add route for a "
861 "neighbor(%s), errno=%d\n",
862 ip6_sprintf(addr6), e);
868 struct llinfo_nd6 *ln =
869 (struct llinfo_nd6 *)rt->rt_llinfo;
870 ln->ln_state = ND6_LLINFO_NOSTATE;
878 * Validation for the entry.
879 * Note that the check for rt_llinfo is necessary because a cloned
880 * route from a parent route that has the L flag (e.g. the default
881 * route to a p2p interface) may have the flag, too, while the
882 * destination is not actually a neighbor.
883 * XXX: we can't use rt->rt_ifp to check for the interface, since
884 * it might be the loopback interface if the entry is for our
885 * own address on a non-loopback interface. Instead, we should
886 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
888 * Note also that ifa_ifp and ifp may differ when we connect two
889 * interfaces to a same link, install a link prefix to an interface,
890 * and try to install a neighbor cache on an interface that does not
891 * have a route to the prefix.
893 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
894 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
895 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
898 "nd6_lookup: failed to lookup %s (if = %s)\n",
900 ifp ? if_name(ifp) : "unspec"));
905 RT_UNLOCK(rt); /* XXX not ready to return rt locked */
910 * Test whether a given IPv6 address is a neighbor or not, ignoring
911 * the actual neighbor cache. The neighbor cache is ignored in order
912 * to not reenter the routing code from within itself.
915 nd6_is_new_addr_neighbor(addr, ifp)
916 struct sockaddr_in6 *addr;
919 struct nd_prefix *pr;
920 struct ifaddr *dstaddr;
923 * A link-local address is always a neighbor.
924 * XXX: a link does not necessarily specify a single interface.
926 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
927 struct sockaddr_in6 sin6_copy;
931 * We need sin6_copy since sa6_recoverscope() may modify the
935 if (sa6_recoverscope(&sin6_copy))
936 return (0); /* XXX: should be impossible */
937 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
939 if (sin6_copy.sin6_scope_id == zone)
946 * If the address matches one of our addresses,
947 * it should be a neighbor.
948 * If the address matches one of our on-link prefixes, it should be a
951 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
952 if (pr->ndpr_ifp != ifp)
955 if (!(pr->ndpr_stateflags & NDPRF_ONLINK))
958 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
959 &addr->sin6_addr, &pr->ndpr_mask))
964 * If the address is assigned on the node of the other side of
965 * a p2p interface, the address should be a neighbor.
967 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
968 if ((dstaddr != NULL) && (dstaddr->ifa_ifp == ifp))
972 * If the default router list is empty, all addresses are regarded
973 * as on-link, and thus, as a neighbor.
974 * XXX: we restrict the condition to hosts, because routers usually do
975 * not have the "default router list".
977 if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL &&
978 nd6_defifindex == ifp->if_index) {
987 * Detect if a given IPv6 address identifies a neighbor on a given link.
988 * XXX: should take care of the destination of a p2p link?
991 nd6_is_addr_neighbor(addr, ifp)
992 struct sockaddr_in6 *addr;
996 if (nd6_is_new_addr_neighbor(addr, ifp))
1000 * Even if the address matches none of our addresses, it might be
1001 * in the neighbor cache.
1003 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
1010 * Free an nd6 llinfo entry.
1011 * Since the function would cause significant changes in the kernel, DO NOT
1012 * make it global, unless you have a strong reason for the change, and are sure
1013 * that the change is safe.
1015 static struct llinfo_nd6 *
1020 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1021 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
1022 struct nd_defrouter *dr;
1025 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1026 * even though it is not harmful, it was not really necessary.
1030 nd6_llinfo_settimer(ln, -1);
1032 if (!ip6_forwarding) {
1035 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1038 if (dr != NULL && dr->expire &&
1039 ln->ln_state == ND6_LLINFO_STALE && gc) {
1041 * If the reason for the deletion is just garbage
1042 * collection, and the neighbor is an active default
1043 * router, do not delete it. Instead, reset the GC
1044 * timer using the router's lifetime.
1045 * Simply deleting the entry would affect default
1046 * router selection, which is not necessarily a good
1047 * thing, especially when we're using router preference
1049 * XXX: the check for ln_state would be redundant,
1050 * but we intentionally keep it just in case.
1052 if (dr->expire > time_second)
1053 nd6_llinfo_settimer(ln,
1054 (dr->expire - time_second) * hz);
1056 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1058 return (ln->ln_next);
1061 if (ln->ln_router || dr) {
1063 * rt6_flush must be called whether or not the neighbor
1064 * is in the Default Router List.
1065 * See a corresponding comment in nd6_na_input().
1067 rt6_flush(&in6, rt->rt_ifp);
1072 * Unreachablity of a router might affect the default
1073 * router selection and on-link detection of advertised
1078 * Temporarily fake the state to choose a new default
1079 * router and to perform on-link determination of
1080 * prefixes correctly.
1081 * Below the state will be set correctly,
1082 * or the entry itself will be deleted.
1084 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1087 * Since defrouter_select() does not affect the
1088 * on-link determination and MIP6 needs the check
1089 * before the default router selection, we perform
1092 pfxlist_onlink_check();
1095 * refresh default router list
1103 * Before deleting the entry, remember the next entry as the
1104 * return value. We need this because pfxlist_onlink_check() above
1105 * might have freed other entries (particularly the old next entry) as
1106 * a side effect (XXX).
1111 * Detach the route from the routing tree and the list of neighbor
1112 * caches, and disable the route entry not to be used in already
1115 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1116 rt_mask(rt), 0, (struct rtentry **)0);
1122 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1124 * XXX cost-effective methods?
1127 nd6_nud_hint(rt, dst6, force)
1129 struct in6_addr *dst6;
1132 struct llinfo_nd6 *ln;
1135 * If the caller specified "rt", use that. Otherwise, resolve the
1136 * routing table by supplied "dst6".
1141 if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL)
1145 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1146 (rt->rt_flags & RTF_LLINFO) == 0 ||
1147 rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
1148 rt->rt_gateway->sa_family != AF_LINK) {
1149 /* This is not a host route. */
1153 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1154 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1158 * if we get upper-layer reachability confirmation many times,
1159 * it is possible we have false information.
1163 if (ln->ln_byhint > nd6_maxnudhint)
1167 ln->ln_state = ND6_LLINFO_REACHABLE;
1168 if (!ND6_LLINFO_PERMANENT(ln)) {
1169 nd6_llinfo_settimer(ln,
1170 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1175 nd6_rtrequest(req, rt, info)
1178 struct rt_addrinfo *info; /* xxx unused */
1180 struct sockaddr *gate = rt->rt_gateway;
1181 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1182 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1183 struct ifnet *ifp = rt->rt_ifp;
1188 if ((rt->rt_flags & RTF_GATEWAY) != 0)
1191 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1193 * This is probably an interface direct route for a link
1194 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1195 * We do not need special treatment below for such a route.
1196 * Moreover, the RTF_LLINFO flag which would be set below
1197 * would annoy the ndp(8) command.
1202 if (req == RTM_RESOLVE &&
1203 (nd6_need_cache(ifp) == 0 || /* stf case */
1204 !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt),
1207 * FreeBSD and BSD/OS often make a cloned host route based
1208 * on a less-specific route (e.g. the default route).
1209 * If the less specific route does not have a "gateway"
1210 * (this is the case when the route just goes to a p2p or an
1211 * stf interface), we'll mistakenly make a neighbor cache for
1212 * the host route, and will see strange neighbor solicitation
1213 * for the corresponding destination. In order to avoid the
1214 * confusion, we check if the destination of the route is
1215 * a neighbor in terms of neighbor discovery, and stop the
1216 * process if not. Additionally, we remove the LLINFO flag
1217 * so that ndp(8) will not try to get the neighbor information
1218 * of the destination.
1220 rt->rt_flags &= ~RTF_LLINFO;
1227 * There is no backward compatibility :)
1229 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1230 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1231 * rt->rt_flags |= RTF_CLONING;
1233 if ((rt->rt_flags & RTF_CLONING) ||
1234 ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) {
1236 * Case 1: This route should come from a route to
1237 * interface (RTF_CLONING case) or the route should be
1238 * treated as on-link but is currently not
1239 * (RTF_LLINFO && ln == NULL case).
1241 rt_setgate(rt, rt_key(rt),
1242 (struct sockaddr *)&null_sdl);
1243 gate = rt->rt_gateway;
1244 SDL(gate)->sdl_type = ifp->if_type;
1245 SDL(gate)->sdl_index = ifp->if_index;
1247 nd6_llinfo_settimer(ln, 0);
1248 if ((rt->rt_flags & RTF_CLONING) != 0)
1252 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1253 * We don't do that here since llinfo is not ready yet.
1255 * There are also couple of other things to be discussed:
1256 * - unsolicited NA code needs improvement beforehand
1257 * - RFC2461 says we MAY send multicast unsolicited NA
1258 * (7.2.6 paragraph 4), however, it also says that we
1259 * SHOULD provide a mechanism to prevent multicast NA storm.
1260 * we don't have anything like it right now.
1261 * note that the mechanism needs a mutual agreement
1262 * between proxies, which means that we need to implement
1263 * a new protocol, or a new kludge.
1264 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1265 * we need to check ip6forwarding before sending it.
1266 * (or should we allow proxy ND configuration only for
1267 * routers? there's no mention about proxy ND from hosts)
1271 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1273 * Address resolution isn't necessary for a point to
1274 * point link, so we can skip this test for a p2p link.
1276 if (gate->sa_family != AF_LINK ||
1277 gate->sa_len < sizeof(null_sdl)) {
1279 "nd6_rtrequest: bad gateway value: %s\n",
1283 SDL(gate)->sdl_type = ifp->if_type;
1284 SDL(gate)->sdl_index = ifp->if_index;
1287 break; /* This happens on a route change */
1289 * Case 2: This route may come from cloning, or a manual route
1290 * add with a LL address.
1292 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1293 rt->rt_llinfo = (caddr_t)ln;
1295 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1300 bzero(ln, sizeof(*ln));
1302 callout_init(&ln->ln_timer_ch, 0);
1304 /* this is required for "ndp" command. - shin */
1305 if (req == RTM_ADD) {
1307 * gate should have some valid AF_LINK entry,
1308 * and ln->ln_expire should have some lifetime
1309 * which is specified by ndp command.
1311 ln->ln_state = ND6_LLINFO_REACHABLE;
1315 * When req == RTM_RESOLVE, rt is created and
1316 * initialized in rtrequest(), so rt_expire is 0.
1318 ln->ln_state = ND6_LLINFO_NOSTATE;
1319 nd6_llinfo_settimer(ln, 0);
1321 rt->rt_flags |= RTF_LLINFO;
1322 ln->ln_next = llinfo_nd6.ln_next;
1323 llinfo_nd6.ln_next = ln;
1324 ln->ln_prev = &llinfo_nd6;
1325 ln->ln_next->ln_prev = ln;
1328 * check if rt_key(rt) is one of my address assigned
1331 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1332 &SIN6(rt_key(rt))->sin6_addr);
1334 caddr_t macp = nd6_ifptomac(ifp);
1335 nd6_llinfo_settimer(ln, -1);
1336 ln->ln_state = ND6_LLINFO_REACHABLE;
1339 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1340 SDL(gate)->sdl_alen = ifp->if_addrlen;
1342 if (nd6_useloopback) {
1343 rt->rt_ifp = &loif[0]; /* XXX */
1345 * Make sure rt_ifa be equal to the ifaddr
1346 * corresponding to the address.
1347 * We need this because when we refer
1348 * rt_ifa->ia6_flags in ip6_input, we assume
1349 * that the rt_ifa points to the address instead
1350 * of the loopback address.
1352 if (ifa != rt->rt_ifa) {
1353 IFAFREE(rt->rt_ifa);
1358 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1359 nd6_llinfo_settimer(ln, -1);
1360 ln->ln_state = ND6_LLINFO_REACHABLE;
1363 /* join solicited node multicast for proxy ND */
1364 if (ifp->if_flags & IFF_MULTICAST) {
1365 struct in6_addr llsol;
1368 llsol = SIN6(rt_key(rt))->sin6_addr;
1369 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1370 llsol.s6_addr32[1] = 0;
1371 llsol.s6_addr32[2] = htonl(1);
1372 llsol.s6_addr8[12] = 0xff;
1373 if (in6_setscope(&llsol, ifp, NULL))
1375 if (in6_addmulti(&llsol, ifp,
1376 &error, 0) == NULL) {
1377 nd6log((LOG_ERR, "%s: failed to join "
1378 "%s (errno=%d)\n", if_name(ifp),
1379 ip6_sprintf(&llsol), error));
1388 /* leave from solicited node multicast for proxy ND */
1389 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1390 (ifp->if_flags & IFF_MULTICAST) != 0) {
1391 struct in6_addr llsol;
1392 struct in6_multi *in6m;
1394 llsol = SIN6(rt_key(rt))->sin6_addr;
1395 llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1396 llsol.s6_addr32[1] = 0;
1397 llsol.s6_addr32[2] = htonl(1);
1398 llsol.s6_addr8[12] = 0xff;
1399 if (in6_setscope(&llsol, ifp, NULL) == 0) {
1400 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1404 ; /* XXX: should not happen. bark here? */
1407 ln->ln_next->ln_prev = ln->ln_prev;
1408 ln->ln_prev->ln_next = ln->ln_next;
1410 nd6_llinfo_settimer(ln, -1);
1412 rt->rt_flags &= ~RTF_LLINFO;
1413 clear_llinfo_pqueue(ln);
1419 nd6_ioctl(cmd, data, ifp)
1424 struct in6_drlist *drl = (struct in6_drlist *)data;
1425 struct in6_oprlist *oprl = (struct in6_oprlist *)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;
1430 struct nd_prefix *pr;
1432 int i = 0, error = 0;
1436 case SIOCGDRLST_IN6:
1438 * obsolete API, use sysctl under net.inet6.icmp6
1440 bzero(drl, sizeof(*drl));
1442 dr = TAILQ_FIRST(&nd_defrouter);
1443 while (dr && i < DRLSTSIZ) {
1444 drl->defrouter[i].rtaddr = dr->rtaddr;
1445 in6_clearscope(&drl->defrouter[i].rtaddr);
1447 drl->defrouter[i].flags = dr->flags;
1448 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1449 drl->defrouter[i].expire = dr->expire;
1450 drl->defrouter[i].if_index = dr->ifp->if_index;
1452 dr = TAILQ_NEXT(dr, dr_entry);
1456 case SIOCGPRLST_IN6:
1458 * obsolete API, use sysctl under net.inet6.icmp6
1460 * XXX the structure in6_prlist was changed in backward-
1461 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1462 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1465 * XXX meaning of fields, especialy "raflags", is very
1466 * differnet between RA prefix list and RR/static prefix list.
1467 * how about separating ioctls into two?
1469 bzero(oprl, sizeof(*oprl));
1471 pr = nd_prefix.lh_first;
1472 while (pr && i < PRLSTSIZ) {
1473 struct nd_pfxrouter *pfr;
1476 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1477 oprl->prefix[i].raflags = pr->ndpr_raf;
1478 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1479 oprl->prefix[i].vltime = pr->ndpr_vltime;
1480 oprl->prefix[i].pltime = pr->ndpr_pltime;
1481 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1482 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1483 oprl->prefix[i].expire = 0;
1487 /* XXX: we assume time_t is signed. */
1490 ((sizeof(maxexpire) * 8) - 1));
1491 if (pr->ndpr_vltime <
1492 maxexpire - pr->ndpr_lastupdate) {
1493 oprl->prefix[i].expire =
1494 pr->ndpr_lastupdate +
1497 oprl->prefix[i].expire = maxexpire;
1500 pfr = pr->ndpr_advrtrs.lh_first;
1504 #define RTRADDR oprl->prefix[i].advrtr[j]
1505 RTRADDR = pfr->router->rtaddr;
1506 in6_clearscope(&RTRADDR);
1510 pfr = pfr->pfr_next;
1512 oprl->prefix[i].advrtrs = j;
1513 oprl->prefix[i].origin = PR_ORIG_RA;
1521 case OSIOCGIFINFO_IN6:
1523 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1524 bzero(&ND, sizeof(ND));
1525 ND.linkmtu = IN6_LINKMTU(ifp);
1526 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1527 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1528 ND.reachable = ND_IFINFO(ifp)->reachable;
1529 ND.retrans = ND_IFINFO(ifp)->retrans;
1530 ND.flags = ND_IFINFO(ifp)->flags;
1531 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1532 ND.chlim = ND_IFINFO(ifp)->chlim;
1534 case SIOCGIFINFO_IN6:
1535 ND = *ND_IFINFO(ifp);
1537 case SIOCSIFINFO_IN6:
1539 * used to change host variables from userland.
1540 * intented for a use on router to reflect RA configurations.
1542 /* 0 means 'unspecified' */
1543 if (ND.linkmtu != 0) {
1544 if (ND.linkmtu < IPV6_MMTU ||
1545 ND.linkmtu > IN6_LINKMTU(ifp)) {
1549 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1552 if (ND.basereachable != 0) {
1553 int obasereachable = ND_IFINFO(ifp)->basereachable;
1555 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1556 if (ND.basereachable != obasereachable)
1557 ND_IFINFO(ifp)->reachable =
1558 ND_COMPUTE_RTIME(ND.basereachable);
1560 if (ND.retrans != 0)
1561 ND_IFINFO(ifp)->retrans = ND.retrans;
1563 ND_IFINFO(ifp)->chlim = ND.chlim;
1565 case SIOCSIFINFO_FLAGS:
1566 ND_IFINFO(ifp)->flags = ND.flags;
1569 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1570 /* sync kernel routing table with the default router 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) == 0)
1596 if (ia->ia6_ndpr == pr)
1597 in6_purgeaddr(&ia->ia_ifa);
1604 case SIOCSRTRFLUSH_IN6:
1606 /* flush all the default routers */
1607 struct nd_defrouter *dr, *next;
1611 for (dr = TAILQ_FIRST(&nd_defrouter); dr; dr = next) {
1612 next = TAILQ_NEXT(dr, dr_entry);
1619 case SIOCGNBRINFO_IN6:
1621 struct llinfo_nd6 *ln;
1622 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1624 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1628 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1633 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1634 nbi->state = ln->ln_state;
1635 nbi->asked = ln->ln_asked;
1636 nbi->isrouter = ln->ln_router;
1637 nbi->expire = ln->ln_expire;
1642 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1643 ndif->ifindex = nd6_defifindex;
1645 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1646 return (nd6_setdefaultiface(ndif->ifindex));
1652 * Create neighbor cache entry and cache link-layer address,
1653 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1656 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
1658 struct in6_addr *from;
1661 int type; /* ICMP6 type */
1662 int code; /* type dependent information */
1664 struct rtentry *rt = NULL;
1665 struct llinfo_nd6 *ln = NULL;
1667 struct sockaddr_dl *sdl = NULL;
1674 panic("ifp == NULL in nd6_cache_lladdr");
1676 panic("from == NULL in nd6_cache_lladdr");
1678 /* nothing must be updated for unspecified address */
1679 if (IN6_IS_ADDR_UNSPECIFIED(from))
1683 * Validation about ifp->if_addrlen and lladdrlen must be done in
1686 * XXX If the link does not have link-layer adderss, what should
1687 * we do? (ifp->if_addrlen == 0)
1688 * Spec says nothing in sections for RA, RS and NA. There's small
1689 * description on it in NS section (RFC 2461 7.2.3).
1692 rt = nd6_lookup(from, 0, ifp);
1694 rt = nd6_lookup(from, 1, ifp);
1697 /* do nothing if static ndp is set */
1698 if (rt->rt_flags & RTF_STATIC)
1705 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1707 (void)nd6_free(rt, 0);
1710 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1713 if (rt->rt_gateway == NULL)
1715 if (rt->rt_gateway->sa_family != AF_LINK)
1717 sdl = SDL(rt->rt_gateway);
1719 olladdr = (sdl->sdl_alen) ? 1 : 0;
1720 if (olladdr && lladdr) {
1721 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1729 * newentry olladdr lladdr llchange (*=record)
1732 * 0 n y -- (3) * STALE
1734 * 0 y y y (5) * STALE
1735 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1736 * 1 -- y -- (7) * STALE
1739 if (lladdr) { /* (3-5) and (7) */
1741 * Record source link-layer address
1742 * XXX is it dependent to ifp->if_type?
1744 sdl->sdl_alen = ifp->if_addrlen;
1745 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1749 if ((!olladdr && lladdr != NULL) || /* (3) */
1750 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1752 newstate = ND6_LLINFO_STALE;
1753 } else /* (1-2,4) */
1757 if (lladdr == NULL) /* (6) */
1758 newstate = ND6_LLINFO_NOSTATE;
1760 newstate = ND6_LLINFO_STALE;
1765 * Update the state of the neighbor cache.
1767 ln->ln_state = newstate;
1769 if (ln->ln_state == ND6_LLINFO_STALE) {
1771 * XXX: since nd6_output() below will cause
1772 * state tansition to DELAY and reset the timer,
1773 * we must set the timer now, although it is actually
1776 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1779 struct mbuf *m_hold, *m_hold_next;
1780 for (m_hold = ln->ln_hold; m_hold;
1781 m_hold = m_hold_next) {
1782 struct mbuf *mpkt = NULL;
1784 m_hold_next = m_hold->m_nextpkt;
1785 mpkt = m_copym(m_hold, 0, M_COPYALL, M_DONTWAIT);
1790 mpkt->m_nextpkt = NULL;
1793 * we assume ifp is not a p2p here, so
1794 * just set the 2nd argument as the
1797 nd6_output(ifp, ifp, mpkt,
1798 (struct sockaddr_in6 *)rt_key(rt),
1803 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1804 /* probe right away */
1805 nd6_llinfo_settimer((void *)ln, 0);
1810 * ICMP6 type dependent behavior.
1812 * NS: clear IsRouter if new entry
1813 * RS: clear IsRouter
1814 * RA: set IsRouter if there's lladdr
1815 * redir: clear IsRouter if new entry
1818 * The spec says that we must set IsRouter in the following cases:
1819 * - If lladdr exist, set IsRouter. This means (1-5).
1820 * - If it is old entry (!newentry), set IsRouter. This means (7).
1821 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1822 * A quetion arises for (1) case. (1) case has no lladdr in the
1823 * neighbor cache, this is similar to (6).
1824 * This case is rare but we figured that we MUST NOT set IsRouter.
1826 * newentry olladdr lladdr llchange NS RS RA redir
1828 * 0 n n -- (1) c ? s
1829 * 0 y n -- (2) c s s
1830 * 0 n y -- (3) c s s
1833 * 1 -- n -- (6) c c c s
1834 * 1 -- y -- (7) c c s c s
1838 switch (type & 0xff) {
1839 case ND_NEIGHBOR_SOLICIT:
1841 * New entry must have is_router flag cleared.
1843 if (is_newentry) /* (6-7) */
1848 * If the icmp is a redirect to a better router, always set the
1849 * is_router flag. Otherwise, if the entry is newly created,
1850 * clear the flag. [RFC 2461, sec 8.3]
1852 if (code == ND_REDIRECT_ROUTER)
1854 else if (is_newentry) /* (6-7) */
1857 case ND_ROUTER_SOLICIT:
1859 * is_router flag must always be cleared.
1863 case ND_ROUTER_ADVERT:
1865 * Mark an entry with lladdr as a router.
1867 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1868 (is_newentry && lladdr)) { /* (7) */
1875 * When the link-layer address of a router changes, select the
1876 * best router again. In particular, when the neighbor entry is newly
1877 * created, it might affect the selection policy.
1878 * Question: can we restrict the first condition to the "is_newentry"
1880 * XXX: when we hear an RA from a new router with the link-layer
1881 * address option, defrouter_select() is called twice, since
1882 * defrtrlist_update called the function as well. However, I believe
1883 * we can compromise the overhead, since it only happens the first
1885 * XXX: although defrouter_select() should not have a bad effect
1886 * for those are not autoconfigured hosts, we explicitly avoid such
1889 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1896 nd6_slowtimo(ignored_arg)
1899 struct nd_ifinfo *nd6if;
1902 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1903 nd6_slowtimo, NULL);
1905 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1906 nd6if = ND_IFINFO(ifp);
1907 if (nd6if->basereachable && /* already initialized */
1908 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1910 * Since reachable time rarely changes by router
1911 * advertisements, we SHOULD insure that a new random
1912 * value gets recomputed at least once every few hours.
1915 nd6if->recalctm = nd6_recalc_reachtm_interval;
1916 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1922 #define senderr(e) { error = (e); goto bad;}
1924 nd6_output(ifp, origifp, m0, dst, rt0)
1926 struct ifnet *origifp;
1928 struct sockaddr_in6 *dst;
1929 struct rtentry *rt0;
1931 struct mbuf *m = m0;
1932 struct rtentry *rt = rt0;
1933 struct sockaddr_in6 *gw6 = NULL;
1934 struct llinfo_nd6 *ln = NULL;
1937 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1940 if (nd6_need_cache(ifp) == 0)
1944 * next hop determination. This routine is derived from ether_output.
1948 if ((rt->rt_flags & RTF_UP) == 0) {
1949 rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL);
1953 if (rt->rt_ifp != ifp)
1955 * XXX maybe we should update ifp too,
1956 * but the original code didn't and I
1957 * don't know what is correct here.
1961 senderr(EHOSTUNREACH);
1964 if (rt->rt_flags & RTF_GATEWAY) {
1965 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1968 * We skip link-layer address resolution and NUD
1969 * if the gateway is not a neighbor from ND point
1970 * of view, regardless of the value of nd_ifinfo.flags.
1971 * The second condition is a bit tricky; we skip
1972 * if the gateway is our own address, which is
1973 * sometimes used to install a route to a p2p link.
1975 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1976 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1978 * We allow this kind of tricky route only
1979 * when the outgoing interface is p2p.
1980 * XXX: we may need a more generic rule here.
1982 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1983 senderr(EHOSTUNREACH);
1988 if (rt->rt_gwroute == 0)
1990 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1992 rtfree(rt); rt = rt0;
1994 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
1995 if ((rt = rt->rt_gwroute) == 0)
1996 senderr(EHOSTUNREACH);
2003 * Address resolution or Neighbor Unreachability Detection
2005 * At this point, the destination of the packet must be a unicast
2006 * or an anycast address(i.e. not a multicast).
2009 /* Look up the neighbor cache for the nexthop */
2010 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
2011 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2014 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2015 * the condition below is not very efficient. But we believe
2016 * it is tolerable, because this should be a rare case.
2018 if (nd6_is_addr_neighbor(dst, ifp) &&
2019 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
2020 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2022 if (ln == NULL || rt == NULL) {
2023 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2024 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2026 "nd6_output: can't allocate llinfo for %s "
2028 ip6_sprintf(&dst->sin6_addr), ln, rt);
2029 senderr(EIO); /* XXX: good error? */
2032 goto sendpkt; /* send anyway */
2035 /* We don't have to do link-layer address resolution on a p2p link. */
2036 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2037 ln->ln_state < ND6_LLINFO_REACHABLE) {
2038 ln->ln_state = ND6_LLINFO_STALE;
2039 nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
2043 * The first time we send a packet to a neighbor whose entry is
2044 * STALE, we have to change the state to DELAY and a sets a timer to
2045 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2046 * neighbor unreachability detection on expiration.
2049 if (ln->ln_state == ND6_LLINFO_STALE) {
2051 ln->ln_state = ND6_LLINFO_DELAY;
2052 nd6_llinfo_settimer(ln, (long)nd6_delay * hz);
2056 * If the neighbor cache entry has a state other than INCOMPLETE
2057 * (i.e. its link-layer address is already resolved), just
2060 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2064 * There is a neighbor cache entry, but no ethernet address
2065 * response yet. Append this latest packet to the end of the
2066 * packet queue in the mbuf, unless the number of the packet
2067 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2068 * the oldest packet in the queue will be removed.
2070 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2071 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2073 struct mbuf *m_hold;
2077 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) {
2079 if (m_hold->m_nextpkt == NULL) {
2080 m_hold->m_nextpkt = m;
2084 while (i >= nd6_maxqueuelen) {
2085 m_hold = ln->ln_hold;
2086 ln->ln_hold = ln->ln_hold->m_nextpkt;
2095 * If there has been no NS for the neighbor after entering the
2096 * INCOMPLETE state, send the first solicitation.
2098 if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
2100 nd6_llinfo_settimer(ln,
2101 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2102 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2107 /* discard the packet if IPv6 operation is disabled on the interface */
2108 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2109 error = ENETDOWN; /* better error? */
2114 /* clean ipsec history once it goes out of the node */
2119 mac_create_mbuf_linklayer(ifp, m);
2121 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2122 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2125 return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2139 * XXX: we currently do not make neighbor cache on any interface
2140 * other than ARCnet, Ethernet, FDDI and GIF.
2143 * - unidirectional tunnels needs no ND
2145 switch (ifp->if_type) {
2153 #ifdef IFT_IEEE80211
2159 case IFT_GIF: /* XXX need more cases? */
2170 nd6_storelladdr(ifp, rt0, m, dst, desten)
2172 struct rtentry *rt0;
2174 struct sockaddr *dst;
2177 struct sockaddr_dl *sdl;
2181 if (m->m_flags & M_MCAST) {
2184 switch (ifp->if_type) {
2190 #ifdef IFT_IEEE80211
2195 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2200 * netbsd can use if_broadcastaddr, but we don't do so
2201 * to reduce # of ifdef.
2203 for (i = 0; i < ifp->if_addrlen; i++)
2211 return (EAFNOSUPPORT);
2216 /* this could happen, if we could not allocate memory */
2221 error = rt_check(&rt, &rt0, dst);
2228 if (rt->rt_gateway->sa_family != AF_LINK) {
2229 printf("nd6_storelladdr: something odd happens\n");
2233 sdl = SDL(rt->rt_gateway);
2234 if (sdl->sdl_alen == 0) {
2235 /* this should be impossible, but we bark here for debugging */
2236 printf("nd6_storelladdr: sdl_alen == 0\n");
2241 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2246 clear_llinfo_pqueue(ln)
2247 struct llinfo_nd6 *ln;
2249 struct mbuf *m_hold, *m_hold_next;
2251 for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) {
2252 m_hold_next = m_hold->m_nextpkt;
2253 m_hold->m_nextpkt = NULL;
2261 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2262 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2264 SYSCTL_DECL(_net_inet6_icmp6);
2266 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2267 CTLFLAG_RD, nd6_sysctl_drlist, "");
2268 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2269 CTLFLAG_RD, nd6_sysctl_prlist, "");
2270 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2271 CTLFLAG_RW, &nd6_maxqueuelen, 1, "");
2274 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2278 struct in6_defrouter *d, *de;
2279 struct nd_defrouter *dr;
2285 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
2286 dr = TAILQ_NEXT(dr, dr_entry)) {
2287 d = (struct in6_defrouter *)buf;
2288 de = (struct in6_defrouter *)(buf + sizeof(buf));
2291 bzero(d, sizeof(*d));
2292 d->rtaddr.sin6_family = AF_INET6;
2293 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2294 d->rtaddr.sin6_addr = dr->rtaddr;
2295 sa6_recoverscope(&d->rtaddr);
2296 d->flags = dr->flags;
2297 d->rtlifetime = dr->rtlifetime;
2298 d->expire = dr->expire;
2299 d->if_index = dr->ifp->if_index;
2301 panic("buffer too short");
2303 error = SYSCTL_OUT(req, buf, sizeof(*d));
2312 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2316 struct in6_prefix *p, *pe;
2317 struct nd_prefix *pr;
2323 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2326 struct sockaddr_in6 *sin6, *s6;
2327 struct nd_pfxrouter *pfr;
2329 p = (struct in6_prefix *)buf;
2330 pe = (struct in6_prefix *)(buf + sizeof(buf));
2333 bzero(p, sizeof(*p));
2334 sin6 = (struct sockaddr_in6 *)(p + 1);
2336 p->prefix = pr->ndpr_prefix;
2337 if (sa6_recoverscope(&p->prefix)) {
2339 "scope error in prefix list (%s)\n",
2340 ip6_sprintf(&p->prefix.sin6_addr));
2341 /* XXX: press on... */
2343 p->raflags = pr->ndpr_raf;
2344 p->prefixlen = pr->ndpr_plen;
2345 p->vltime = pr->ndpr_vltime;
2346 p->pltime = pr->ndpr_pltime;
2347 p->if_index = pr->ndpr_ifp->if_index;
2348 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2353 /* XXX: we assume time_t is signed. */
2356 ((sizeof(maxexpire) * 8) - 1));
2357 if (pr->ndpr_vltime <
2358 maxexpire - pr->ndpr_lastupdate) {
2359 p->expire = pr->ndpr_lastupdate +
2362 p->expire = maxexpire;
2364 p->refcnt = pr->ndpr_refcnt;
2365 p->flags = pr->ndpr_stateflags;
2366 p->origin = PR_ORIG_RA;
2368 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2369 pfr = pfr->pfr_next) {
2370 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2374 s6 = &sin6[advrtrs];
2375 bzero(s6, sizeof(*s6));
2376 s6->sin6_family = AF_INET6;
2377 s6->sin6_len = sizeof(*sin6);
2378 s6->sin6_addr = pfr->router->rtaddr;
2379 if (sa6_recoverscope(s6)) {
2382 "prefix list (%s)\n",
2383 ip6_sprintf(&pfr->router->rtaddr));
2387 p->advrtrs = advrtrs;
2389 panic("buffer too short");
2391 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2392 error = SYSCTL_OUT(req, buf, advance);