1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.2.2.15 2003/05/06 06:46:58 suz Exp $ */
2 /* $DragonFly: src/sys/netinet6/nd6.c,v 1.9 2004/08/13 14:27:29 drhodus Exp $ */
3 /* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ */
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * BSD/OS version heavily modifies this code, related to llinfo.
38 * Since we don't have BSD/OS version of net/route.c in our hand,
39 * I left the code mostly as it was in 970310. -- itojun
43 #include "opt_inet6.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/callout.h>
48 #include <sys/malloc.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
53 #include <sys/kernel.h>
54 #include <sys/protosw.h>
55 #include <sys/errno.h>
56 #include <sys/syslog.h>
57 #include <sys/queue.h>
58 #include <sys/sysctl.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/if_atm.h>
64 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/if_fddi.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet/ip6.h>
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/nd6.h>
73 #include <netinet6/in6_prefix.h>
74 #include <netinet/icmp6.h>
78 #include <net/net_osdep.h>
80 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
81 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
83 #define SIN6(s) ((struct sockaddr_in6 *)s)
84 #define SDL(s) ((struct sockaddr_dl *)s)
87 int nd6_prune = 1; /* walk list every 1 seconds */
88 int nd6_delay = 5; /* delay first probe time 5 second */
89 int nd6_umaxtries = 3; /* maximum unicast query */
90 int nd6_mmaxtries = 3; /* maximum multicast query */
91 int nd6_useloopback = 1; /* use loopback interface for local traffic */
92 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
94 /* preventing too many loops in ND option parsing */
95 int nd6_maxndopt = 10; /* max # of ND options allowed */
97 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
106 static int nd6_inuse, nd6_allocated;
108 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
109 static size_t nd_ifinfo_indexlim = 8;
110 struct nd_ifinfo *nd_ifinfo = NULL;
111 struct nd_drhead nd_defrouter;
112 struct nd_prhead nd_prefix = { 0 };
114 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
115 static struct sockaddr_in6 all1_sa;
117 static void nd6_slowtimo (void *);
118 static int regen_tmpaddr (struct in6_ifaddr *);
120 struct callout nd6_slowtimo_ch;
121 struct callout nd6_timer_ch;
122 extern struct callout in6_tmpaddrtimer_ch;
127 static int nd6_init_done = 0;
131 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
135 all1_sa.sin6_family = AF_INET6;
136 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
137 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
138 all1_sa.sin6_addr.s6_addr[i] = 0xff;
140 /* initialization of the default router list */
141 TAILQ_INIT(&nd_defrouter);
146 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
151 nd6_ifattach(struct ifnet *ifp)
155 * We have some arrays that should be indexed by if_index.
156 * since if_index will grow dynamically, they should grow too.
158 if (nd_ifinfo == NULL || if_index >= nd_ifinfo_indexlim) {
162 while (if_index >= nd_ifinfo_indexlim)
163 nd_ifinfo_indexlim <<= 1;
166 n = nd_ifinfo_indexlim * sizeof(struct nd_ifinfo);
167 q = (caddr_t)malloc(n, M_IP6NDP, M_WAITOK);
170 bcopy((caddr_t)nd_ifinfo, q, n/2);
171 free((caddr_t)nd_ifinfo, M_IP6NDP);
173 nd_ifinfo = (struct nd_ifinfo *)q;
176 #define ND nd_ifinfo[ifp->if_index]
179 * Don't initialize if called twice.
180 * XXX: to detect this, we should choose a member that is never set
181 * before initialization of the ND structure itself. We formaly used
182 * the linkmtu member, which was not suitable because it could be
183 * initialized via "ifconfig mtu".
185 if (ND.basereachable)
188 ND.linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu;
189 ND.chlim = IPV6_DEFHLIM;
190 ND.basereachable = REACHABLE_TIME;
191 ND.reachable = ND_COMPUTE_RTIME(ND.basereachable);
192 ND.retrans = RETRANS_TIMER;
194 ND.flags = ND6_IFF_PERFORMNUD;
200 * Reset ND level link MTU. This function is called when the physical MTU
201 * changes, which means we might have to adjust the ND level MTU.
204 nd6_setmtu(struct ifnet *ifp)
206 struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
207 u_long oldmaxmtu = ndi->maxmtu;
208 u_long oldlinkmtu = ndi->linkmtu;
210 switch (ifp->if_type) {
211 case IFT_ARCNET: /* XXX MTU handling needs more work */
212 ndi->maxmtu = MIN(60480, ifp->if_mtu);
215 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
218 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
221 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
223 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */
224 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
227 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */
228 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
232 ndi->maxmtu = ifp->if_mtu;
236 if (oldmaxmtu != ndi->maxmtu) {
238 * If the ND level MTU is not set yet, or if the maxmtu
239 * is reset to a smaller value than the ND level MTU,
240 * also reset the ND level MTU.
242 if (ndi->linkmtu == 0 ||
243 ndi->maxmtu < ndi->linkmtu) {
244 ndi->linkmtu = ndi->maxmtu;
245 /* also adjust in6_maxmtu if necessary. */
246 if (oldlinkmtu == 0) {
248 * XXX: the case analysis is grotty, but
249 * it is not efficient to call in6_setmaxmtu()
250 * here when we are during the initialization
253 if (in6_maxmtu < ndi->linkmtu)
254 in6_maxmtu = ndi->linkmtu;
263 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
265 bzero(ndopts, sizeof(*ndopts));
266 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
268 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
271 ndopts->nd_opts_done = 1;
272 ndopts->nd_opts_search = NULL;
277 * Take one ND option.
280 nd6_option(union nd_opts *ndopts)
282 struct nd_opt_hdr *nd_opt;
286 panic("ndopts == NULL in nd6_option");
287 if (!ndopts->nd_opts_last)
288 panic("uninitialized ndopts in nd6_option");
289 if (!ndopts->nd_opts_search)
291 if (ndopts->nd_opts_done)
294 nd_opt = ndopts->nd_opts_search;
296 /* make sure nd_opt_len is inside the buffer */
297 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
298 bzero(ndopts, sizeof(*ndopts));
302 olen = nd_opt->nd_opt_len << 3;
305 * Message validation requires that all included
306 * options have a length that is greater than zero.
308 bzero(ndopts, sizeof(*ndopts));
312 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
313 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
314 /* option overruns the end of buffer, invalid */
315 bzero(ndopts, sizeof(*ndopts));
317 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
318 /* reached the end of options chain */
319 ndopts->nd_opts_done = 1;
320 ndopts->nd_opts_search = NULL;
326 * Parse multiple ND options.
327 * This function is much easier to use, for ND routines that do not need
328 * multiple options of the same type.
331 nd6_options(union nd_opts *ndopts)
333 struct nd_opt_hdr *nd_opt;
337 panic("ndopts == NULL in nd6_options");
338 if (!ndopts->nd_opts_last)
339 panic("uninitialized ndopts in nd6_options");
340 if (!ndopts->nd_opts_search)
344 nd_opt = nd6_option(ndopts);
345 if (!nd_opt && !ndopts->nd_opts_last) {
347 * Message validation requires that all included
348 * options have a length that is greater than zero.
350 icmp6stat.icp6s_nd_badopt++;
351 bzero(ndopts, sizeof(*ndopts));
358 switch (nd_opt->nd_opt_type) {
359 case ND_OPT_SOURCE_LINKADDR:
360 case ND_OPT_TARGET_LINKADDR:
362 case ND_OPT_REDIRECTED_HEADER:
363 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
365 "duplicated ND6 option found (type=%d)\n",
366 nd_opt->nd_opt_type));
369 ndopts->nd_opt_array[nd_opt->nd_opt_type]
373 case ND_OPT_PREFIX_INFORMATION:
374 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
375 ndopts->nd_opt_array[nd_opt->nd_opt_type]
378 ndopts->nd_opts_pi_end =
379 (struct nd_opt_prefix_info *)nd_opt;
383 * Unknown options must be silently ignored,
384 * to accomodate future extension to the protocol.
387 "nd6_options: unsupported option %d - "
388 "option ignored\n", nd_opt->nd_opt_type));
393 if (i > nd6_maxndopt) {
394 icmp6stat.icp6s_nd_toomanyopt++;
395 nd6log((LOG_INFO, "too many loop in nd opt\n"));
399 if (ndopts->nd_opts_done)
407 * ND6 timer routine to expire default route list and prefix list
410 nd6_timer(void *ignored_arg)
413 struct llinfo_nd6 *ln;
414 struct nd_defrouter *dr;
415 struct nd_prefix *pr;
417 struct in6_ifaddr *ia6, *nia6;
418 struct in6_addrlifetime *lt6;
421 callout_reset(&nd6_timer_ch, nd6_prune * hz,
424 ln = llinfo_nd6.ln_next;
425 while (ln && ln != &llinfo_nd6) {
427 struct sockaddr_in6 *dst;
428 struct llinfo_nd6 *next = ln->ln_next;
429 /* XXX: used for the DELAY case only: */
430 struct nd_ifinfo *ndi = NULL;
432 if ((rt = ln->ln_rt) == NULL) {
436 if ((ifp = rt->rt_ifp) == NULL) {
440 ndi = &nd_ifinfo[ifp->if_index];
441 dst = (struct sockaddr_in6 *)rt_key(rt);
443 if (ln->ln_expire > time_second) {
450 panic("rt=0 in nd6_timer(ln=%p)", ln);
451 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
452 panic("rt_llinfo(%p) is not equal to ln(%p)",
455 panic("dst=0 in nd6_timer(ln=%p)", ln);
457 switch (ln->ln_state) {
458 case ND6_LLINFO_INCOMPLETE:
459 if (ln->ln_asked < nd6_mmaxtries) {
461 ln->ln_expire = time_second +
462 nd_ifinfo[ifp->if_index].retrans / 1000;
463 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
466 struct mbuf *m = ln->ln_hold;
470 * Fake rcvif to make ICMP error
471 * more helpful in diagnosing
473 * XXX: should we consider
476 m->m_pkthdr.rcvif = rt->rt_ifp;
478 icmp6_error(m, ICMP6_DST_UNREACH,
479 ICMP6_DST_UNREACH_ADDR, 0);
485 case ND6_LLINFO_REACHABLE:
487 ln->ln_state = ND6_LLINFO_STALE;
488 ln->ln_expire = time_second + nd6_gctimer;
492 case ND6_LLINFO_STALE:
493 /* Garbage Collection(RFC 2461 5.3) */
498 case ND6_LLINFO_DELAY:
499 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
502 ln->ln_state = ND6_LLINFO_PROBE;
503 ln->ln_expire = time_second +
505 nd6_ns_output(ifp, &dst->sin6_addr,
509 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
510 ln->ln_expire = time_second + nd6_gctimer;
513 case ND6_LLINFO_PROBE:
514 if (ln->ln_asked < nd6_umaxtries) {
516 ln->ln_expire = time_second +
517 nd_ifinfo[ifp->if_index].retrans / 1000;
518 nd6_ns_output(ifp, &dst->sin6_addr,
519 &dst->sin6_addr, ln, 0);
528 /* expire default router list */
529 dr = TAILQ_FIRST(&nd_defrouter);
531 if (dr->expire && dr->expire < time_second) {
532 struct nd_defrouter *t;
533 t = TAILQ_NEXT(dr, dr_entry);
537 dr = TAILQ_NEXT(dr, dr_entry);
542 * expire interface addresses.
543 * in the past the loop was inside prefix expiry processing.
544 * However, from a stricter speci-confrmance standpoint, we should
545 * rather separate address lifetimes and prefix lifetimes.
548 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
550 /* check address lifetime */
551 lt6 = &ia6->ia6_lifetime;
552 if (IFA6_IS_INVALID(ia6)) {
556 * If the expiring address is temporary, try
557 * regenerating a new one. This would be useful when
558 * we suspended a laptop PC, then turned it on after a
559 * period that could invalidate all temporary
560 * addresses. Although we may have to restart the
561 * loop (see below), it must be after purging the
562 * address. Otherwise, we'd see an infinite loop of
565 if (ip6_use_tempaddr &&
566 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
567 if (regen_tmpaddr(ia6) == 0)
571 in6_purgeaddr(&ia6->ia_ifa);
574 goto addrloop; /* XXX: see below */
576 if (IFA6_IS_DEPRECATED(ia6)) {
577 int oldflags = ia6->ia6_flags;
579 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
582 * If a temporary address has just become deprecated,
583 * regenerate a new one if possible.
585 if (ip6_use_tempaddr &&
586 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
587 (oldflags & IN6_IFF_DEPRECATED) == 0) {
589 if (regen_tmpaddr(ia6) == 0) {
591 * A new temporary address is
593 * XXX: this means the address chain
594 * has changed while we are still in
595 * the loop. Although the change
596 * would not cause disaster (because
597 * it's not a deletion, but an
598 * addition,) we'd rather restart the
599 * loop just for safety. Or does this
600 * significantly reduce performance??
607 * A new RA might have made a deprecated address
610 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
614 /* expire prefix list */
615 pr = nd_prefix.lh_first;
618 * check prefix lifetime.
619 * since pltime is just for autoconf, pltime processing for
620 * prefix is not necessary.
622 if (pr->ndpr_expire && pr->ndpr_expire < time_second) {
627 * address expiration and prefix expiration are
628 * separate. NEVER perform in6_purgeaddr here.
640 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary
645 struct in6_ifaddr *public_ifa6 = NULL;
647 ifp = ia6->ia_ifa.ifa_ifp;
648 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_list) {
649 struct in6_ifaddr *it6;
651 if (ifa->ifa_addr->sa_family != AF_INET6)
654 it6 = (struct in6_ifaddr *)ifa;
656 /* ignore no autoconf addresses. */
657 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
660 /* ignore autoconf addresses with different prefixes. */
661 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
665 * Now we are looking at an autoconf address with the same
666 * prefix as ours. If the address is temporary and is still
667 * preferred, do not create another one. It would be rare, but
668 * could happen, for example, when we resume a laptop PC after
671 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
672 !IFA6_IS_DEPRECATED(it6)) {
678 * This is a public autoconf address that has the same prefix
679 * as ours. If it is preferred, keep it. We can't break the
680 * loop here, because there may be a still-preferred temporary
681 * address with the prefix.
683 if (!IFA6_IS_DEPRECATED(it6))
687 if (public_ifa6 != NULL) {
690 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
691 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
692 " tmp addr,errno=%d\n", e);
702 * Nuke neighbor cache/prefix/default router management table, right before
706 nd6_purge(struct ifnet *ifp)
708 struct llinfo_nd6 *ln, *nln;
709 struct nd_defrouter *dr, *ndr, drany;
710 struct nd_prefix *pr, *npr;
712 /* Nuke default router list entries toward ifp */
713 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
715 * The first entry of the list may be stored in
716 * the routing table, so we'll delete it later.
718 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
719 ndr = TAILQ_NEXT(dr, dr_entry);
723 dr = TAILQ_FIRST(&nd_defrouter);
728 /* Nuke prefix list entries toward ifp */
729 for (pr = nd_prefix.lh_first; pr; pr = npr) {
731 if (pr->ndpr_ifp == ifp) {
733 * Previously, pr->ndpr_addr is removed as well,
734 * but I strongly believe we don't have to do it.
735 * nd6_purge() is only called from in6_ifdetach(),
736 * which removes all the associated interface addresses
738 * (jinmei@kame.net 20010129)
744 /* cancel default outgoing interface setting */
745 if (nd6_defifindex == ifp->if_index)
746 nd6_setdefaultiface(0);
748 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
749 /* refresh default router list */
750 bzero(&drany, sizeof(drany));
751 defrouter_delreq(&drany, 0);
756 * Nuke neighbor cache entries for the ifp.
757 * Note that rt->rt_ifp may not be the same as ifp,
758 * due to KAME goto ours hack. See RTM_RESOLVE case in
759 * nd6_rtrequest(), and ip6_input().
761 ln = llinfo_nd6.ln_next;
762 while (ln && ln != &llinfo_nd6) {
764 struct sockaddr_dl *sdl;
768 if (rt && rt->rt_gateway &&
769 rt->rt_gateway->sa_family == AF_LINK) {
770 sdl = (struct sockaddr_dl *)rt->rt_gateway;
771 if (sdl->sdl_index == ifp->if_index)
779 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp)
782 struct sockaddr_in6 sin6;
784 bzero(&sin6, sizeof(sin6));
785 sin6.sin6_len = sizeof(struct sockaddr_in6);
786 sin6.sin6_family = AF_INET6;
787 sin6.sin6_addr = *addr6;
789 sin6.sin6_scope_id = in6_addr2scopeid(ifp, addr6);
791 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
792 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
794 * This is the case for the default route.
795 * If we want to create a neighbor cache for the address, we
796 * should free the route for the destination and allocate an
809 * If no route is available and create is set,
810 * we allocate a host route for the destination
811 * and treat it like an interface route.
812 * This hack is necessary for a neighbor which can't
813 * be covered by our own prefix.
816 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
821 * Create a new route. RTF_LLINFO is necessary
822 * to create a Neighbor Cache entry for the
823 * destination in nd6_rtrequest which will be
824 * called in rtrequest via ifa->ifa_rtrequest.
826 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
828 (struct sockaddr *)&all1_sa,
830 RTF_HOST | RTF_LLINFO) &
834 "nd6_lookup: failed to add route for a "
835 "neighbor(%s), errno=%d\n",
836 ip6_sprintf(addr6), e);
840 struct llinfo_nd6 *ln =
841 (struct llinfo_nd6 *)rt->rt_llinfo;
842 ln->ln_state = ND6_LLINFO_NOSTATE;
849 * Validation for the entry.
850 * Note that the check for rt_llinfo is necessary because a cloned
851 * route from a parent route that has the L flag (e.g. the default
852 * route to a p2p interface) may have the flag, too, while the
853 * destination is not actually a neighbor.
854 * XXX: we can't use rt->rt_ifp to check for the interface, since
855 * it might be the loopback interface if the entry is for our
856 * own address on a non-loopback interface. Instead, we should
857 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
860 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
861 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
862 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
864 log(LOG_DEBUG, "nd6_lookup: failed to lookup %s (if = %s)\n",
865 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
866 /* xxx more logs... kazu */
874 * Detect if a given IPv6 address identifies a neighbor on a given link.
875 * XXX: should take care of the destination of a p2p link?
878 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
883 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
884 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
887 * A link-local address is always a neighbor.
888 * XXX: we should use the sin6_scope_id field rather than the embedded
891 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
892 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
896 * If the address matches one of our addresses,
897 * it should be a neighbor.
899 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
900 if (ifa->ifa_addr->sa_family != AF_INET6)
903 for (i = 0; i < 4; i++) {
904 if ((IFADDR6(ifa).s6_addr32[i] ^
905 addr->sin6_addr.s6_addr32[i]) &
906 IFMASK6(ifa).s6_addr32[i])
913 * Even if the address matches none of our addresses, it might be
914 * in the neighbor cache.
916 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
925 * Free an nd6 llinfo entry.
928 nd6_free(struct rtentry *rt)
930 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
931 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
932 struct nd_defrouter *dr;
935 * we used to have pfctlinput(PRC_HOSTDEAD) here.
936 * even though it is not harmful, it was not really necessary.
939 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
942 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
945 if (ln->ln_router || dr) {
947 * rt6_flush must be called whether or not the neighbor
948 * is in the Default Router List.
949 * See a corresponding comment in nd6_na_input().
951 rt6_flush(&in6, rt->rt_ifp);
956 * Unreachablity of a router might affect the default
957 * router selection and on-link detection of advertised
962 * Temporarily fake the state to choose a new default
963 * router and to perform on-link determination of
964 * prefixes correctly.
965 * Below the state will be set correctly,
966 * or the entry itself will be deleted.
968 ln->ln_state = ND6_LLINFO_INCOMPLETE;
971 * Since defrouter_select() does not affect the
972 * on-link determination and MIP6 needs the check
973 * before the default router selection, we perform
976 pfxlist_onlink_check();
978 if (dr == TAILQ_FIRST(&nd_defrouter)) {
980 * It is used as the current default router,
981 * so we have to move it to the end of the
982 * list and choose a new one.
983 * XXX: it is not very efficient if this is
986 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
987 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
996 * Before deleting the entry, remember the next entry as the
997 * return value. We need this because pfxlist_onlink_check() above
998 * might have freed other entries (particularly the old next entry) as
999 * a side effect (XXX).
1004 * Detach the route from the routing tree and the list of neighbor
1005 * caches, and disable the route entry not to be used in already
1008 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1009 rt_mask(rt), 0, (struct rtentry **)0);
1015 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1017 * XXX cost-effective metods?
1020 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1022 struct llinfo_nd6 *ln;
1025 * If the caller specified "rt", use that. Otherwise, resolve the
1026 * routing table by supplied "dst6".
1031 if (!(rt = nd6_lookup(dst6, 0, NULL)))
1035 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1036 (rt->rt_flags & RTF_LLINFO) == 0 ||
1037 !rt->rt_llinfo || !rt->rt_gateway ||
1038 rt->rt_gateway->sa_family != AF_LINK) {
1039 /* This is not a host route. */
1043 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1044 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1048 * if we get upper-layer reachability confirmation many times,
1049 * it is possible we have false information.
1053 if (ln->ln_byhint > nd6_maxnudhint)
1057 ln->ln_state = ND6_LLINFO_REACHABLE;
1059 ln->ln_expire = time_second +
1060 nd_ifinfo[rt->rt_ifp->if_index].reachable;
1064 nd6_rtrequest(int req, struct rtentry *rt,
1065 struct rt_addrinfo *info) /* xxx unused */
1067 struct sockaddr *gate = rt->rt_gateway;
1068 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1069 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1070 struct ifnet *ifp = rt->rt_ifp;
1073 if ((rt->rt_flags & RTF_GATEWAY))
1076 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1078 * This is probably an interface direct route for a link
1079 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1080 * We do not need special treatment below for such a route.
1081 * Moreover, the RTF_LLINFO flag which would be set below
1082 * would annoy the ndp(8) command.
1087 if (req == RTM_RESOLVE &&
1088 (nd6_need_cache(ifp) == 0 || /* stf case */
1089 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
1091 * FreeBSD and BSD/OS often make a cloned host route based
1092 * on a less-specific route (e.g. the default route).
1093 * If the less specific route does not have a "gateway"
1094 * (this is the case when the route just goes to a p2p or an
1095 * stf interface), we'll mistakenly make a neighbor cache for
1096 * the host route, and will see strange neighbor solicitation
1097 * for the corresponding destination. In order to avoid the
1098 * confusion, we check if the destination of the route is
1099 * a neighbor in terms of neighbor discovery, and stop the
1100 * process if not. Additionally, we remove the LLINFO flag
1101 * so that ndp(8) will not try to get the neighbor information
1102 * of the destination.
1104 rt->rt_flags &= ~RTF_LLINFO;
1111 * There is no backward compatibility :)
1113 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1114 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1115 * rt->rt_flags |= RTF_CLONING;
1117 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1119 * Case 1: This route should come from
1120 * a route to interface. RTF_LLINFO flag is set
1121 * for a host route whose destination should be
1122 * treated as on-link.
1124 rt_setgate(rt, rt_key(rt),
1125 (struct sockaddr *)&null_sdl);
1126 gate = rt->rt_gateway;
1127 SDL(gate)->sdl_type = ifp->if_type;
1128 SDL(gate)->sdl_index = ifp->if_index;
1130 ln->ln_expire = time_second;
1132 if (ln && ln->ln_expire == 0) {
1133 /* kludge for desktops */
1135 printf("nd6_rtequest: time.tv_sec is zero; "
1141 if ((rt->rt_flags & RTF_CLONING))
1145 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1146 * We don't do that here since llinfo is not ready yet.
1148 * There are also couple of other things to be discussed:
1149 * - unsolicited NA code needs improvement beforehand
1150 * - RFC2461 says we MAY send multicast unsolicited NA
1151 * (7.2.6 paragraph 4), however, it also says that we
1152 * SHOULD provide a mechanism to prevent multicast NA storm.
1153 * we don't have anything like it right now.
1154 * note that the mechanism needs a mutual agreement
1155 * between proxies, which means that we need to implement
1156 * a new protocol, or a new kludge.
1157 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1158 * we need to check ip6forwarding before sending it.
1159 * (or should we allow proxy ND configuration only for
1160 * routers? there's no mention about proxy ND from hosts)
1163 /* XXX it does not work */
1164 if (rt->rt_flags & RTF_ANNOUNCE)
1166 &SIN6(rt_key(rt))->sin6_addr,
1167 &SIN6(rt_key(rt))->sin6_addr,
1168 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1173 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1175 * Address resolution isn't necessary for a point to
1176 * point link, so we can skip this test for a p2p link.
1178 if (gate->sa_family != AF_LINK ||
1179 gate->sa_len < sizeof(null_sdl)) {
1181 "nd6_rtrequest: bad gateway value: %s\n",
1185 SDL(gate)->sdl_type = ifp->if_type;
1186 SDL(gate)->sdl_index = ifp->if_index;
1189 break; /* This happens on a route change */
1191 * Case 2: This route may come from cloning, or a manual route
1192 * add with a LL address.
1194 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1195 rt->rt_llinfo = (caddr_t)ln;
1197 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1202 Bzero(ln, sizeof(*ln));
1204 /* this is required for "ndp" command. - shin */
1205 if (req == RTM_ADD) {
1207 * gate should have some valid AF_LINK entry,
1208 * and ln->ln_expire should have some lifetime
1209 * which is specified by ndp command.
1211 ln->ln_state = ND6_LLINFO_REACHABLE;
1215 * When req == RTM_RESOLVE, rt is created and
1216 * initialized in rtrequest(), so rt_expire is 0.
1218 ln->ln_state = ND6_LLINFO_NOSTATE;
1219 ln->ln_expire = time_second;
1221 rt->rt_flags |= RTF_LLINFO;
1222 ln->ln_next = llinfo_nd6.ln_next;
1223 llinfo_nd6.ln_next = ln;
1224 ln->ln_prev = &llinfo_nd6;
1225 ln->ln_next->ln_prev = ln;
1228 * check if rt_key(rt) is one of my address assigned
1231 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1232 &SIN6(rt_key(rt))->sin6_addr);
1234 caddr_t macp = nd6_ifptomac(ifp);
1236 ln->ln_state = ND6_LLINFO_REACHABLE;
1239 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1240 SDL(gate)->sdl_alen = ifp->if_addrlen;
1242 if (nd6_useloopback) {
1243 rt->rt_ifp = &loif[0]; /* XXX */
1245 * Make sure rt_ifa be equal to the ifaddr
1246 * corresponding to the address.
1247 * We need this because when we refer
1248 * rt_ifa->ia6_flags in ip6_input, we assume
1249 * that the rt_ifa points to the address instead
1250 * of the loopback address.
1252 if (ifa != rt->rt_ifa) {
1253 IFAFREE(rt->rt_ifa);
1258 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1260 ln->ln_state = ND6_LLINFO_REACHABLE;
1263 /* join solicited node multicast for proxy ND */
1264 if (ifp->if_flags & IFF_MULTICAST) {
1265 struct in6_addr llsol;
1268 llsol = SIN6(rt_key(rt))->sin6_addr;
1269 llsol.s6_addr16[0] = htons(0xff02);
1270 llsol.s6_addr16[1] = htons(ifp->if_index);
1271 llsol.s6_addr32[1] = 0;
1272 llsol.s6_addr32[2] = htonl(1);
1273 llsol.s6_addr8[12] = 0xff;
1275 if (!in6_addmulti(&llsol, ifp, &error)) {
1276 nd6log((LOG_ERR, "%s: failed to join "
1277 "%s (errno=%d)\n", if_name(ifp),
1278 ip6_sprintf(&llsol), error));
1287 /* leave from solicited node multicast for proxy ND */
1288 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1289 (ifp->if_flags & IFF_MULTICAST) != 0) {
1290 struct in6_addr llsol;
1291 struct in6_multi *in6m;
1293 llsol = SIN6(rt_key(rt))->sin6_addr;
1294 llsol.s6_addr16[0] = htons(0xff02);
1295 llsol.s6_addr16[1] = htons(ifp->if_index);
1296 llsol.s6_addr32[1] = 0;
1297 llsol.s6_addr32[2] = htonl(1);
1298 llsol.s6_addr8[12] = 0xff;
1300 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1305 ln->ln_next->ln_prev = ln->ln_prev;
1306 ln->ln_prev->ln_next = ln->ln_next;
1309 rt->rt_flags &= ~RTF_LLINFO;
1311 m_freem(ln->ln_hold);
1317 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1319 struct in6_drlist *drl = (struct in6_drlist *)data;
1320 struct in6_prlist *prl = (struct in6_prlist *)data;
1321 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1322 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1323 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1324 struct nd_defrouter *dr, any;
1325 struct nd_prefix *pr;
1327 int i = 0, error = 0;
1331 case SIOCGDRLST_IN6:
1333 * obsolete API, use sysctl under net.inet6.icmp6
1335 bzero(drl, sizeof(*drl));
1337 dr = TAILQ_FIRST(&nd_defrouter);
1338 while (dr && i < DRLSTSIZ) {
1339 drl->defrouter[i].rtaddr = dr->rtaddr;
1340 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1341 /* XXX: need to this hack for KAME stack */
1342 drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1345 "default router list contains a "
1346 "non-linklocal address(%s)\n",
1347 ip6_sprintf(&drl->defrouter[i].rtaddr));
1349 drl->defrouter[i].flags = dr->flags;
1350 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1351 drl->defrouter[i].expire = dr->expire;
1352 drl->defrouter[i].if_index = dr->ifp->if_index;
1354 dr = TAILQ_NEXT(dr, dr_entry);
1358 case SIOCGPRLST_IN6:
1360 * obsolete API, use sysctl under net.inet6.icmp6
1363 * XXX meaning of fields, especialy "raflags", is very
1364 * differnet between RA prefix list and RR/static prefix list.
1365 * how about separating ioctls into two?
1367 bzero(prl, sizeof(*prl));
1369 pr = nd_prefix.lh_first;
1370 while (pr && i < PRLSTSIZ) {
1371 struct nd_pfxrouter *pfr;
1374 (void)in6_embedscope(&prl->prefix[i].prefix,
1375 &pr->ndpr_prefix, NULL, NULL);
1376 prl->prefix[i].raflags = pr->ndpr_raf;
1377 prl->prefix[i].prefixlen = pr->ndpr_plen;
1378 prl->prefix[i].vltime = pr->ndpr_vltime;
1379 prl->prefix[i].pltime = pr->ndpr_pltime;
1380 prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1381 prl->prefix[i].expire = pr->ndpr_expire;
1383 pfr = pr->ndpr_advrtrs.lh_first;
1387 #define RTRADDR prl->prefix[i].advrtr[j]
1388 RTRADDR = pfr->router->rtaddr;
1389 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1390 /* XXX: hack for KAME */
1391 RTRADDR.s6_addr16[1] = 0;
1394 "a router(%s) advertises "
1396 "non-link local address\n",
1397 ip6_sprintf(&RTRADDR));
1401 pfr = pfr->pfr_next;
1403 prl->prefix[i].advrtrs = j;
1404 prl->prefix[i].origin = PR_ORIG_RA;
1410 struct rr_prefix *rpp;
1412 for (rpp = LIST_FIRST(&rr_prefix); rpp;
1413 rpp = LIST_NEXT(rpp, rp_entry)) {
1416 (void)in6_embedscope(&prl->prefix[i].prefix,
1417 &pr->ndpr_prefix, NULL, NULL);
1418 prl->prefix[i].raflags = rpp->rp_raf;
1419 prl->prefix[i].prefixlen = rpp->rp_plen;
1420 prl->prefix[i].vltime = rpp->rp_vltime;
1421 prl->prefix[i].pltime = rpp->rp_pltime;
1422 prl->prefix[i].if_index = rpp->rp_ifp->if_index;
1423 prl->prefix[i].expire = rpp->rp_expire;
1424 prl->prefix[i].advrtrs = 0;
1425 prl->prefix[i].origin = rpp->rp_origin;
1432 case OSIOCGIFINFO_IN6:
1433 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1437 ndi->ndi.linkmtu = nd_ifinfo[ifp->if_index].linkmtu;
1438 ndi->ndi.maxmtu = nd_ifinfo[ifp->if_index].maxmtu;
1439 ndi->ndi.basereachable =
1440 nd_ifinfo[ifp->if_index].basereachable;
1441 ndi->ndi.reachable = nd_ifinfo[ifp->if_index].reachable;
1442 ndi->ndi.retrans = nd_ifinfo[ifp->if_index].retrans;
1443 ndi->ndi.flags = nd_ifinfo[ifp->if_index].flags;
1444 ndi->ndi.recalctm = nd_ifinfo[ifp->if_index].recalctm;
1445 ndi->ndi.chlim = nd_ifinfo[ifp->if_index].chlim;
1446 ndi->ndi.receivedra = nd_ifinfo[ifp->if_index].receivedra;
1448 case SIOCGIFINFO_IN6:
1449 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1453 ndi->ndi = nd_ifinfo[ifp->if_index];
1455 case SIOCSIFINFO_FLAGS:
1456 /* XXX: almost all other fields of ndi->ndi is unused */
1457 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1461 nd_ifinfo[ifp->if_index].flags = ndi->ndi.flags;
1463 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1464 /* flush default router list */
1466 * xxx sumikawa: should not delete route if default
1467 * route equals to the top of default router list
1469 bzero(&any, sizeof(any));
1470 defrouter_delreq(&any, 0);
1472 /* xxx sumikawa: flush prefix list */
1474 case SIOCSPFXFLUSH_IN6:
1476 /* flush all the prefix advertised by routers */
1477 struct nd_prefix *pr, *next;
1480 for (pr = nd_prefix.lh_first; pr; pr = next) {
1481 struct in6_ifaddr *ia, *ia_next;
1483 next = pr->ndpr_next;
1485 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1488 /* do we really have to remove addresses as well? */
1489 for (ia = in6_ifaddr; ia; ia = ia_next) {
1490 /* ia might be removed. keep the next ptr. */
1491 ia_next = ia->ia_next;
1493 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1496 if (ia->ia6_ndpr == pr)
1497 in6_purgeaddr(&ia->ia_ifa);
1504 case SIOCSRTRFLUSH_IN6:
1506 /* flush all the default routers */
1507 struct nd_defrouter *dr, *next;
1510 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1512 * The first entry of the list may be stored in
1513 * the routing table, so we'll delete it later.
1515 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1516 next = TAILQ_NEXT(dr, dr_entry);
1519 defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
1524 case SIOCGNBRINFO_IN6:
1526 struct llinfo_nd6 *ln;
1527 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1530 * XXX: KAME specific hack for scoped addresses
1531 * XXXX: for other scopes than link-local?
1533 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1534 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1535 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1538 *idp = htons(ifp->if_index);
1542 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1547 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1548 nbi->state = ln->ln_state;
1549 nbi->asked = ln->ln_asked;
1550 nbi->isrouter = ln->ln_router;
1551 nbi->expire = ln->ln_expire;
1556 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1557 ndif->ifindex = nd6_defifindex;
1559 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1560 return(nd6_setdefaultiface(ndif->ifindex));
1567 * Create neighbor cache entry and cache link-layer address,
1568 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1571 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1573 int type, /* ICMP6 type */
1574 int code /* type dependent information */)
1576 struct rtentry *rt = NULL;
1577 struct llinfo_nd6 *ln = NULL;
1579 struct sockaddr_dl *sdl = NULL;
1586 panic("ifp == NULL in nd6_cache_lladdr");
1588 panic("from == NULL in nd6_cache_lladdr");
1590 /* nothing must be updated for unspecified address */
1591 if (IN6_IS_ADDR_UNSPECIFIED(from))
1595 * Validation about ifp->if_addrlen and lladdrlen must be done in
1598 * XXX If the link does not have link-layer adderss, what should
1599 * we do? (ifp->if_addrlen == 0)
1600 * Spec says nothing in sections for RA, RS and NA. There's small
1601 * description on it in NS section (RFC 2461 7.2.3).
1604 rt = nd6_lookup(from, 0, ifp);
1607 /* nothing must be done if there's no lladdr */
1608 if (!lladdr || !lladdrlen)
1612 rt = nd6_lookup(from, 1, ifp);
1615 /* do nothing if static ndp is set */
1616 if (rt->rt_flags & RTF_STATIC)
1623 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1628 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1631 if (!rt->rt_gateway)
1633 if (rt->rt_gateway->sa_family != AF_LINK)
1635 sdl = SDL(rt->rt_gateway);
1637 olladdr = (sdl->sdl_alen) ? 1 : 0;
1638 if (olladdr && lladdr) {
1639 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1647 * newentry olladdr lladdr llchange (*=record)
1650 * 0 n y -- (3) * STALE
1652 * 0 y y y (5) * STALE
1653 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1654 * 1 -- y -- (7) * STALE
1657 if (lladdr) { /* (3-5) and (7) */
1659 * Record source link-layer address
1660 * XXX is it dependent to ifp->if_type?
1662 sdl->sdl_alen = ifp->if_addrlen;
1663 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1667 if ((!olladdr && lladdr) /* (3) */
1668 || (olladdr && lladdr && llchange)) { /* (5) */
1670 newstate = ND6_LLINFO_STALE;
1671 } else /* (1-2,4) */
1675 if (!lladdr) /* (6) */
1676 newstate = ND6_LLINFO_NOSTATE;
1678 newstate = ND6_LLINFO_STALE;
1683 * Update the state of the neighbor cache.
1685 ln->ln_state = newstate;
1687 if (ln->ln_state == ND6_LLINFO_STALE) {
1689 * XXX: since nd6_output() below will cause
1690 * state tansition to DELAY and reset the timer,
1691 * we must set the timer now, although it is actually
1694 ln->ln_expire = time_second + nd6_gctimer;
1698 * we assume ifp is not a p2p here, so just
1699 * set the 2nd argument as the 1st one.
1701 nd6_output(ifp, ifp, ln->ln_hold,
1702 (struct sockaddr_in6 *)rt_key(rt),
1706 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1707 /* probe right away */
1708 ln->ln_expire = time_second;
1713 * ICMP6 type dependent behavior.
1715 * NS: clear IsRouter if new entry
1716 * RS: clear IsRouter
1717 * RA: set IsRouter if there's lladdr
1718 * redir: clear IsRouter if new entry
1721 * The spec says that we must set IsRouter in the following cases:
1722 * - If lladdr exist, set IsRouter. This means (1-5).
1723 * - If it is old entry (!newentry), set IsRouter. This means (7).
1724 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1725 * A quetion arises for (1) case. (1) case has no lladdr in the
1726 * neighbor cache, this is similar to (6).
1727 * This case is rare but we figured that we MUST NOT set IsRouter.
1729 * newentry olladdr lladdr llchange NS RS RA redir
1731 * 0 n n -- (1) c ? s
1732 * 0 y n -- (2) c s s
1733 * 0 n y -- (3) c s s
1736 * 1 -- n -- (6) c c c s
1737 * 1 -- y -- (7) c c s c s
1741 switch (type & 0xff) {
1742 case ND_NEIGHBOR_SOLICIT:
1744 * New entry must have is_router flag cleared.
1746 if (is_newentry) /* (6-7) */
1751 * If the icmp is a redirect to a better router, always set the
1752 * is_router flag. Otherwise, if the entry is newly created,
1753 * clear the flag. [RFC 2461, sec 8.3]
1755 if (code == ND_REDIRECT_ROUTER)
1757 else if (is_newentry) /* (6-7) */
1760 case ND_ROUTER_SOLICIT:
1762 * is_router flag must always be cleared.
1766 case ND_ROUTER_ADVERT:
1768 * Mark an entry with lladdr as a router.
1770 if ((!is_newentry && (olladdr || lladdr)) /* (2-5) */
1771 || (is_newentry && lladdr)) { /* (7) */
1778 * When the link-layer address of a router changes, select the
1779 * best router again. In particular, when the neighbor entry is newly
1780 * created, it might affect the selection policy.
1781 * Question: can we restrict the first condition to the "is_newentry"
1783 * XXX: when we hear an RA from a new router with the link-layer
1784 * address option, defrouter_select() is called twice, since
1785 * defrtrlist_update called the function as well. However, I believe
1786 * we can compromise the overhead, since it only happens the first
1788 * XXX: although defrouter_select() should not have a bad effect
1789 * for those are not autoconfigured hosts, we explicitly avoid such
1792 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1799 nd6_slowtimo(void *ignored_arg)
1803 struct nd_ifinfo *nd6if;
1805 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1806 nd6_slowtimo, NULL);
1807 for (i = 1; i < if_index + 1; i++) {
1808 if (!nd_ifinfo || i >= nd_ifinfo_indexlim)
1810 nd6if = &nd_ifinfo[i];
1811 if (nd6if->basereachable && /* already initialized */
1812 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1814 * Since reachable time rarely changes by router
1815 * advertisements, we SHOULD insure that a new random
1816 * value gets recomputed at least once every few hours.
1819 nd6if->recalctm = nd6_recalc_reachtm_interval;
1820 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1826 #define senderr(e) { error = (e); goto bad;}
1828 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1829 struct sockaddr_in6 *dst, struct rtentry *rt0)
1831 struct mbuf *m = m0;
1832 struct rtentry *rt = rt0;
1833 struct sockaddr_in6 *gw6 = NULL;
1834 struct llinfo_nd6 *ln = NULL;
1837 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1840 if (nd6_need_cache(ifp) == 0)
1844 * next hop determination. This routine is derived from ether_outpout.
1847 if ((rt->rt_flags & RTF_UP) == 0) {
1848 if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL)) !=
1852 if (rt->rt_ifp != ifp) {
1853 /* XXX: loop care? */
1854 return nd6_output(ifp, origifp, m0,
1858 senderr(EHOSTUNREACH);
1861 if (rt->rt_flags & RTF_GATEWAY) {
1862 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1865 * We skip link-layer address resolution and NUD
1866 * if the gateway is not a neighbor from ND point
1867 * of view, regardless of the value of nd_ifinfo.flags.
1868 * The second condition is a bit tricky; we skip
1869 * if the gateway is our own address, which is
1870 * sometimes used to install a route to a p2p link.
1872 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1873 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1875 * We allow this kind of tricky route only
1876 * when the outgoing interface is p2p.
1877 * XXX: we may need a more generic rule here.
1879 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1880 senderr(EHOSTUNREACH);
1885 if (rt->rt_gwroute == 0)
1887 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1888 rtfree(rt); rt = rt0;
1889 lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
1890 if ((rt = rt->rt_gwroute) == 0)
1891 senderr(EHOSTUNREACH);
1897 * Address resolution or Neighbor Unreachability Detection
1899 * At this point, the destination of the packet must be a unicast
1900 * or an anycast address(i.e. not a multicast).
1903 /* Look up the neighbor cache for the nexthop */
1904 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
1905 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1908 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1909 * the condition below is not very efficient. But we believe
1910 * it is tolerable, because this should be a rare case.
1912 if (nd6_is_addr_neighbor(dst, ifp) &&
1913 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
1914 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1917 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1918 !(nd_ifinfo[ifp->if_index].flags & ND6_IFF_PERFORMNUD)) {
1920 "nd6_output: can't allocate llinfo for %s "
1922 ip6_sprintf(&dst->sin6_addr), ln, rt);
1923 senderr(EIO); /* XXX: good error? */
1926 goto sendpkt; /* send anyway */
1929 /* We don't have to do link-layer address resolution on a p2p link. */
1930 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1931 ln->ln_state < ND6_LLINFO_REACHABLE) {
1932 ln->ln_state = ND6_LLINFO_STALE;
1933 ln->ln_expire = time_second + nd6_gctimer;
1937 * The first time we send a packet to a neighbor whose entry is
1938 * STALE, we have to change the state to DELAY and a sets a timer to
1939 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1940 * neighbor unreachability detection on expiration.
1943 if (ln->ln_state == ND6_LLINFO_STALE) {
1945 ln->ln_state = ND6_LLINFO_DELAY;
1946 ln->ln_expire = time_second + nd6_delay;
1950 * If the neighbor cache entry has a state other than INCOMPLETE
1951 * (i.e. its link-layer address is already resolved), just
1954 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1958 * There is a neighbor cache entry, but no ethernet address
1959 * response yet. Replace the held mbuf (if any) with this
1962 * This code conforms to the rate-limiting rule described in Section
1963 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
1966 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1967 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1969 m_freem(ln->ln_hold);
1971 if (ln->ln_expire) {
1972 if (ln->ln_asked < nd6_mmaxtries &&
1973 ln->ln_expire < time_second) {
1975 ln->ln_expire = time_second +
1976 nd_ifinfo[ifp->if_index].retrans / 1000;
1977 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1984 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1985 return((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
1988 return((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
1998 nd6_need_cache(struct ifnet *ifp)
2001 * XXX: we currently do not make neighbor cache on any interface
2002 * other than ARCnet, Ethernet, FDDI and GIF.
2005 * - unidirectional tunnels needs no ND
2007 switch (ifp->if_type) {
2015 #ifdef IFT_IEEE80211
2018 case IFT_GIF: /* XXX need more cases? */
2026 nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
2027 struct sockaddr *dst, u_char *desten)
2029 struct sockaddr_dl *sdl;
2031 if (m->m_flags & M_MCAST) {
2032 switch (ifp->if_type) {
2038 #ifdef IFT_IEEE80211
2041 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2045 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen);
2057 /* this could happen, if we could not allocate memory */
2061 if (rt->rt_gateway->sa_family != AF_LINK) {
2062 printf("nd6_storelladdr: something odd happens\n");
2066 sdl = SDL(rt->rt_gateway);
2067 if (sdl->sdl_alen == 0) {
2068 /* this should be impossible, but we bark here for debugging */
2069 printf("nd6_storelladdr: sdl_alen == 0\n");
2074 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2078 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2079 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2081 SYSCTL_DECL(_net_inet6_icmp6);
2083 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2084 CTLFLAG_RD, nd6_sysctl_drlist, "");
2085 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2086 CTLFLAG_RD, nd6_sysctl_prlist, "");
2089 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2093 struct in6_defrouter *d, *de;
2094 struct nd_defrouter *dr;
2100 for (dr = TAILQ_FIRST(&nd_defrouter);
2102 dr = TAILQ_NEXT(dr, dr_entry)) {
2103 d = (struct in6_defrouter *)buf;
2104 de = (struct in6_defrouter *)(buf + sizeof(buf));
2107 bzero(d, sizeof(*d));
2108 d->rtaddr.sin6_family = AF_INET6;
2109 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2110 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2114 "default router list (%s)\n",
2115 ip6_sprintf(&dr->rtaddr));
2116 d->flags = dr->flags;
2117 d->rtlifetime = dr->rtlifetime;
2118 d->expire = dr->expire;
2119 d->if_index = dr->ifp->if_index;
2121 panic("buffer too short");
2123 error = SYSCTL_OUT(req, buf, sizeof(*d));
2131 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2135 struct in6_prefix *p, *pe;
2136 struct nd_prefix *pr;
2142 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2145 struct sockaddr_in6 *sin6, *s6;
2146 struct nd_pfxrouter *pfr;
2148 p = (struct in6_prefix *)buf;
2149 pe = (struct in6_prefix *)(buf + sizeof(buf));
2152 bzero(p, sizeof(*p));
2153 sin6 = (struct sockaddr_in6 *)(p + 1);
2155 p->prefix = pr->ndpr_prefix;
2156 if (in6_recoverscope(&p->prefix,
2157 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2159 "scope error in prefix list (%s)\n",
2160 ip6_sprintf(&p->prefix.sin6_addr));
2161 p->raflags = pr->ndpr_raf;
2162 p->prefixlen = pr->ndpr_plen;
2163 p->vltime = pr->ndpr_vltime;
2164 p->pltime = pr->ndpr_pltime;
2165 p->if_index = pr->ndpr_ifp->if_index;
2166 p->expire = pr->ndpr_expire;
2167 p->refcnt = pr->ndpr_refcnt;
2168 p->flags = pr->ndpr_stateflags;
2169 p->origin = PR_ORIG_RA;
2171 for (pfr = pr->ndpr_advrtrs.lh_first;
2173 pfr = pfr->pfr_next) {
2174 if ((void *)&sin6[advrtrs + 1] >
2179 s6 = &sin6[advrtrs];
2180 bzero(s6, sizeof(*s6));
2181 s6->sin6_family = AF_INET6;
2182 s6->sin6_len = sizeof(*sin6);
2183 if (in6_recoverscope(s6,
2184 &pfr->router->rtaddr,
2185 pfr->router->ifp) != 0)
2188 "prefix list (%s)\n",
2189 ip6_sprintf(&pfr->router->rtaddr));
2192 p->advrtrs = advrtrs;
2194 panic("buffer too short");
2196 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2197 error = SYSCTL_OUT(req, buf, advance);