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 1.5 2003/08/23 11:02:45 rob 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,
156 * We have some arrays that should be indexed by if_index.
157 * since if_index will grow dynamically, they should grow too.
159 if (nd_ifinfo == NULL || if_index >= nd_ifinfo_indexlim) {
163 while (if_index >= nd_ifinfo_indexlim)
164 nd_ifinfo_indexlim <<= 1;
167 n = nd_ifinfo_indexlim * sizeof(struct nd_ifinfo);
168 q = (caddr_t)malloc(n, M_IP6NDP, M_WAITOK);
171 bcopy((caddr_t)nd_ifinfo, q, n/2);
172 free((caddr_t)nd_ifinfo, M_IP6NDP);
174 nd_ifinfo = (struct nd_ifinfo *)q;
177 #define ND nd_ifinfo[ifp->if_index]
180 * Don't initialize if called twice.
181 * XXX: to detect this, we should choose a member that is never set
182 * before initialization of the ND structure itself. We formaly used
183 * the linkmtu member, which was not suitable because it could be
184 * initialized via "ifconfig mtu".
186 if (ND.basereachable)
189 ND.linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu;
190 ND.chlim = IPV6_DEFHLIM;
191 ND.basereachable = REACHABLE_TIME;
192 ND.reachable = ND_COMPUTE_RTIME(ND.basereachable);
193 ND.retrans = RETRANS_TIMER;
195 ND.flags = ND6_IFF_PERFORMNUD;
201 * Reset ND level link MTU. This function is called when the physical MTU
202 * changes, which means we might have to adjust the ND level MTU.
208 struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
209 u_long oldmaxmtu = ndi->maxmtu;
210 u_long oldlinkmtu = ndi->linkmtu;
212 switch (ifp->if_type) {
213 case IFT_ARCNET: /* XXX MTU handling needs more work */
214 ndi->maxmtu = MIN(60480, ifp->if_mtu);
217 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
220 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
223 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
225 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */
226 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
229 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */
230 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
234 ndi->maxmtu = ifp->if_mtu;
238 if (oldmaxmtu != ndi->maxmtu) {
240 * If the ND level MTU is not set yet, or if the maxmtu
241 * is reset to a smaller value than the ND level MTU,
242 * also reset the ND level MTU.
244 if (ndi->linkmtu == 0 ||
245 ndi->maxmtu < ndi->linkmtu) {
246 ndi->linkmtu = ndi->maxmtu;
247 /* also adjust in6_maxmtu if necessary. */
248 if (oldlinkmtu == 0) {
250 * XXX: the case analysis is grotty, but
251 * it is not efficient to call in6_setmaxmtu()
252 * here when we are during the initialization
255 if (in6_maxmtu < ndi->linkmtu)
256 in6_maxmtu = ndi->linkmtu;
265 nd6_option_init(opt, icmp6len, ndopts)
268 union nd_opts *ndopts;
270 bzero(ndopts, sizeof(*ndopts));
271 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
273 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
276 ndopts->nd_opts_done = 1;
277 ndopts->nd_opts_search = NULL;
282 * Take one ND option.
286 union nd_opts *ndopts;
288 struct nd_opt_hdr *nd_opt;
292 panic("ndopts == NULL in nd6_option");
293 if (!ndopts->nd_opts_last)
294 panic("uninitialized ndopts in nd6_option");
295 if (!ndopts->nd_opts_search)
297 if (ndopts->nd_opts_done)
300 nd_opt = ndopts->nd_opts_search;
302 /* make sure nd_opt_len is inside the buffer */
303 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
304 bzero(ndopts, sizeof(*ndopts));
308 olen = nd_opt->nd_opt_len << 3;
311 * Message validation requires that all included
312 * options have a length that is greater than zero.
314 bzero(ndopts, sizeof(*ndopts));
318 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
319 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
320 /* option overruns the end of buffer, invalid */
321 bzero(ndopts, sizeof(*ndopts));
323 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
324 /* reached the end of options chain */
325 ndopts->nd_opts_done = 1;
326 ndopts->nd_opts_search = NULL;
332 * Parse multiple ND options.
333 * This function is much easier to use, for ND routines that do not need
334 * multiple options of the same type.
338 union nd_opts *ndopts;
340 struct nd_opt_hdr *nd_opt;
344 panic("ndopts == NULL in nd6_options");
345 if (!ndopts->nd_opts_last)
346 panic("uninitialized ndopts in nd6_options");
347 if (!ndopts->nd_opts_search)
351 nd_opt = nd6_option(ndopts);
352 if (!nd_opt && !ndopts->nd_opts_last) {
354 * Message validation requires that all included
355 * options have a length that is greater than zero.
357 icmp6stat.icp6s_nd_badopt++;
358 bzero(ndopts, sizeof(*ndopts));
365 switch (nd_opt->nd_opt_type) {
366 case ND_OPT_SOURCE_LINKADDR:
367 case ND_OPT_TARGET_LINKADDR:
369 case ND_OPT_REDIRECTED_HEADER:
370 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
372 "duplicated ND6 option found (type=%d)\n",
373 nd_opt->nd_opt_type));
376 ndopts->nd_opt_array[nd_opt->nd_opt_type]
380 case ND_OPT_PREFIX_INFORMATION:
381 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
382 ndopts->nd_opt_array[nd_opt->nd_opt_type]
385 ndopts->nd_opts_pi_end =
386 (struct nd_opt_prefix_info *)nd_opt;
390 * Unknown options must be silently ignored,
391 * to accomodate future extension to the protocol.
394 "nd6_options: unsupported option %d - "
395 "option ignored\n", nd_opt->nd_opt_type));
400 if (i > nd6_maxndopt) {
401 icmp6stat.icp6s_nd_toomanyopt++;
402 nd6log((LOG_INFO, "too many loop in nd opt\n"));
406 if (ndopts->nd_opts_done)
414 * ND6 timer routine to expire default route list and prefix list
417 nd6_timer(ignored_arg)
421 struct llinfo_nd6 *ln;
422 struct nd_defrouter *dr;
423 struct nd_prefix *pr;
425 struct in6_ifaddr *ia6, *nia6;
426 struct in6_addrlifetime *lt6;
429 callout_reset(&nd6_timer_ch, nd6_prune * hz,
432 ln = llinfo_nd6.ln_next;
433 while (ln && ln != &llinfo_nd6) {
435 struct sockaddr_in6 *dst;
436 struct llinfo_nd6 *next = ln->ln_next;
437 /* XXX: used for the DELAY case only: */
438 struct nd_ifinfo *ndi = NULL;
440 if ((rt = ln->ln_rt) == NULL) {
444 if ((ifp = rt->rt_ifp) == NULL) {
448 ndi = &nd_ifinfo[ifp->if_index];
449 dst = (struct sockaddr_in6 *)rt_key(rt);
451 if (ln->ln_expire > time_second) {
458 panic("rt=0 in nd6_timer(ln=%p)", ln);
459 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
460 panic("rt_llinfo(%p) is not equal to ln(%p)",
463 panic("dst=0 in nd6_timer(ln=%p)", ln);
465 switch (ln->ln_state) {
466 case ND6_LLINFO_INCOMPLETE:
467 if (ln->ln_asked < nd6_mmaxtries) {
469 ln->ln_expire = time_second +
470 nd_ifinfo[ifp->if_index].retrans / 1000;
471 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
474 struct mbuf *m = ln->ln_hold;
478 * Fake rcvif to make ICMP error
479 * more helpful in diagnosing
481 * XXX: should we consider
484 m->m_pkthdr.rcvif = rt->rt_ifp;
486 icmp6_error(m, ICMP6_DST_UNREACH,
487 ICMP6_DST_UNREACH_ADDR, 0);
493 case ND6_LLINFO_REACHABLE:
495 ln->ln_state = ND6_LLINFO_STALE;
496 ln->ln_expire = time_second + nd6_gctimer;
500 case ND6_LLINFO_STALE:
501 /* Garbage Collection(RFC 2461 5.3) */
506 case ND6_LLINFO_DELAY:
507 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
510 ln->ln_state = ND6_LLINFO_PROBE;
511 ln->ln_expire = time_second +
513 nd6_ns_output(ifp, &dst->sin6_addr,
517 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
518 ln->ln_expire = time_second + nd6_gctimer;
521 case ND6_LLINFO_PROBE:
522 if (ln->ln_asked < nd6_umaxtries) {
524 ln->ln_expire = time_second +
525 nd_ifinfo[ifp->if_index].retrans / 1000;
526 nd6_ns_output(ifp, &dst->sin6_addr,
527 &dst->sin6_addr, ln, 0);
536 /* expire default router list */
537 dr = TAILQ_FIRST(&nd_defrouter);
539 if (dr->expire && dr->expire < time_second) {
540 struct nd_defrouter *t;
541 t = TAILQ_NEXT(dr, dr_entry);
545 dr = TAILQ_NEXT(dr, dr_entry);
550 * expire interface addresses.
551 * in the past the loop was inside prefix expiry processing.
552 * However, from a stricter speci-confrmance standpoint, we should
553 * rather separate address lifetimes and prefix lifetimes.
556 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
558 /* check address lifetime */
559 lt6 = &ia6->ia6_lifetime;
560 if (IFA6_IS_INVALID(ia6)) {
564 * If the expiring address is temporary, try
565 * regenerating a new one. This would be useful when
566 * we suspended a laptop PC, then turned it on after a
567 * period that could invalidate all temporary
568 * addresses. Although we may have to restart the
569 * loop (see below), it must be after purging the
570 * address. Otherwise, we'd see an infinite loop of
573 if (ip6_use_tempaddr &&
574 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
575 if (regen_tmpaddr(ia6) == 0)
579 in6_purgeaddr(&ia6->ia_ifa);
582 goto addrloop; /* XXX: see below */
584 if (IFA6_IS_DEPRECATED(ia6)) {
585 int oldflags = ia6->ia6_flags;
587 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
590 * If a temporary address has just become deprecated,
591 * regenerate a new one if possible.
593 if (ip6_use_tempaddr &&
594 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
595 (oldflags & IN6_IFF_DEPRECATED) == 0) {
597 if (regen_tmpaddr(ia6) == 0) {
599 * A new temporary address is
601 * XXX: this means the address chain
602 * has changed while we are still in
603 * the loop. Although the change
604 * would not cause disaster (because
605 * it's not a deletion, but an
606 * addition,) we'd rather restart the
607 * loop just for safety. Or does this
608 * significantly reduce performance??
615 * A new RA might have made a deprecated address
618 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
622 /* expire prefix list */
623 pr = nd_prefix.lh_first;
626 * check prefix lifetime.
627 * since pltime is just for autoconf, pltime processing for
628 * prefix is not necessary.
630 if (pr->ndpr_expire && pr->ndpr_expire < time_second) {
635 * address expiration and prefix expiration are
636 * separate. NEVER perform in6_purgeaddr here.
649 struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
653 struct in6_ifaddr *public_ifa6 = NULL;
655 ifp = ia6->ia_ifa.ifa_ifp;
656 for (ifa = ifp->if_addrlist.tqh_first; ifa;
657 ifa = ifa->ifa_list.tqe_next)
659 struct in6_ifaddr *it6;
661 if (ifa->ifa_addr->sa_family != AF_INET6)
664 it6 = (struct in6_ifaddr *)ifa;
666 /* ignore no autoconf addresses. */
667 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
670 /* ignore autoconf addresses with different prefixes. */
671 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
675 * Now we are looking at an autoconf address with the same
676 * prefix as ours. If the address is temporary and is still
677 * preferred, do not create another one. It would be rare, but
678 * could happen, for example, when we resume a laptop PC after
681 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
682 !IFA6_IS_DEPRECATED(it6)) {
688 * This is a public autoconf address that has the same prefix
689 * as ours. If it is preferred, keep it. We can't break the
690 * loop here, because there may be a still-preferred temporary
691 * address with the prefix.
693 if (!IFA6_IS_DEPRECATED(it6))
697 if (public_ifa6 != NULL) {
700 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
701 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
702 " tmp addr,errno=%d\n", e);
712 * Nuke neighbor cache/prefix/default router management table, right before
719 struct llinfo_nd6 *ln, *nln;
720 struct nd_defrouter *dr, *ndr, drany;
721 struct nd_prefix *pr, *npr;
723 /* Nuke default router list entries toward ifp */
724 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
726 * The first entry of the list may be stored in
727 * the routing table, so we'll delete it later.
729 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
730 ndr = TAILQ_NEXT(dr, dr_entry);
734 dr = TAILQ_FIRST(&nd_defrouter);
739 /* Nuke prefix list entries toward ifp */
740 for (pr = nd_prefix.lh_first; pr; pr = npr) {
742 if (pr->ndpr_ifp == ifp) {
744 * Previously, pr->ndpr_addr is removed as well,
745 * but I strongly believe we don't have to do it.
746 * nd6_purge() is only called from in6_ifdetach(),
747 * which removes all the associated interface addresses
749 * (jinmei@kame.net 20010129)
755 /* cancel default outgoing interface setting */
756 if (nd6_defifindex == ifp->if_index)
757 nd6_setdefaultiface(0);
759 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
760 /* refresh default router list */
761 bzero(&drany, sizeof(drany));
762 defrouter_delreq(&drany, 0);
767 * Nuke neighbor cache entries for the ifp.
768 * Note that rt->rt_ifp may not be the same as ifp,
769 * due to KAME goto ours hack. See RTM_RESOLVE case in
770 * nd6_rtrequest(), and ip6_input().
772 ln = llinfo_nd6.ln_next;
773 while (ln && ln != &llinfo_nd6) {
775 struct sockaddr_dl *sdl;
779 if (rt && rt->rt_gateway &&
780 rt->rt_gateway->sa_family == AF_LINK) {
781 sdl = (struct sockaddr_dl *)rt->rt_gateway;
782 if (sdl->sdl_index == ifp->if_index)
790 nd6_lookup(addr6, create, ifp)
791 struct in6_addr *addr6;
796 struct sockaddr_in6 sin6;
798 bzero(&sin6, sizeof(sin6));
799 sin6.sin6_len = sizeof(struct sockaddr_in6);
800 sin6.sin6_family = AF_INET6;
801 sin6.sin6_addr = *addr6;
803 sin6.sin6_scope_id = in6_addr2scopeid(ifp, addr6);
805 rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
806 if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
808 * This is the case for the default route.
809 * If we want to create a neighbor cache for the address, we
810 * should free the route for the destination and allocate an
823 * If no route is available and create is set,
824 * we allocate a host route for the destination
825 * and treat it like an interface route.
826 * This hack is necessary for a neighbor which can't
827 * be covered by our own prefix.
830 ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
835 * Create a new route. RTF_LLINFO is necessary
836 * to create a Neighbor Cache entry for the
837 * destination in nd6_rtrequest which will be
838 * called in rtrequest via ifa->ifa_rtrequest.
840 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
842 (struct sockaddr *)&all1_sa,
844 RTF_HOST | RTF_LLINFO) &
848 "nd6_lookup: failed to add route for a "
849 "neighbor(%s), errno=%d\n",
850 ip6_sprintf(addr6), e);
854 struct llinfo_nd6 *ln =
855 (struct llinfo_nd6 *)rt->rt_llinfo;
856 ln->ln_state = ND6_LLINFO_NOSTATE;
863 * Validation for the entry.
864 * Note that the check for rt_llinfo is necessary because a cloned
865 * route from a parent route that has the L flag (e.g. the default
866 * route to a p2p interface) may have the flag, too, while the
867 * destination is not actually a neighbor.
868 * XXX: we can't use rt->rt_ifp to check for the interface, since
869 * it might be the loopback interface if the entry is for our
870 * own address on a non-loopback interface. Instead, we should
871 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
874 if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
875 rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
876 (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
878 log(LOG_DEBUG, "nd6_lookup: failed to lookup %s (if = %s)\n",
879 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
880 /* xxx more logs... kazu */
888 * Detect if a given IPv6 address identifies a neighbor on a given link.
889 * XXX: should take care of the destination of a p2p link?
892 nd6_is_addr_neighbor(addr, ifp)
893 struct sockaddr_in6 *addr;
899 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
900 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
903 * A link-local address is always a neighbor.
904 * XXX: we should use the sin6_scope_id field rather than the embedded
907 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
908 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
912 * If the address matches one of our addresses,
913 * it should be a neighbor.
915 for (ifa = ifp->if_addrlist.tqh_first;
917 ifa = ifa->ifa_list.tqe_next)
919 if (ifa->ifa_addr->sa_family != AF_INET6)
922 for (i = 0; i < 4; i++) {
923 if ((IFADDR6(ifa).s6_addr32[i] ^
924 addr->sin6_addr.s6_addr32[i]) &
925 IFMASK6(ifa).s6_addr32[i])
932 * Even if the address matches none of our addresses, it might be
933 * in the neighbor cache.
935 if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
944 * Free an nd6 llinfo entry.
950 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
951 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
952 struct nd_defrouter *dr;
955 * we used to have pfctlinput(PRC_HOSTDEAD) here.
956 * even though it is not harmful, it was not really necessary.
959 if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
962 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
965 if (ln->ln_router || dr) {
967 * rt6_flush must be called whether or not the neighbor
968 * is in the Default Router List.
969 * See a corresponding comment in nd6_na_input().
971 rt6_flush(&in6, rt->rt_ifp);
976 * Unreachablity of a router might affect the default
977 * router selection and on-link detection of advertised
982 * Temporarily fake the state to choose a new default
983 * router and to perform on-link determination of
984 * prefixes correctly.
985 * Below the state will be set correctly,
986 * or the entry itself will be deleted.
988 ln->ln_state = ND6_LLINFO_INCOMPLETE;
991 * Since defrouter_select() does not affect the
992 * on-link determination and MIP6 needs the check
993 * before the default router selection, we perform
996 pfxlist_onlink_check();
998 if (dr == TAILQ_FIRST(&nd_defrouter)) {
1000 * It is used as the current default router,
1001 * so we have to move it to the end of the
1002 * list and choose a new one.
1003 * XXX: it is not very efficient if this is
1006 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1007 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
1016 * Before deleting the entry, remember the next entry as the
1017 * return value. We need this because pfxlist_onlink_check() above
1018 * might have freed other entries (particularly the old next entry) as
1019 * a side effect (XXX).
1024 * Detach the route from the routing tree and the list of neighbor
1025 * caches, and disable the route entry not to be used in already
1028 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1029 rt_mask(rt), 0, (struct rtentry **)0);
1035 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1037 * XXX cost-effective metods?
1040 nd6_nud_hint(rt, dst6, force)
1042 struct in6_addr *dst6;
1045 struct llinfo_nd6 *ln;
1048 * If the caller specified "rt", use that. Otherwise, resolve the
1049 * routing table by supplied "dst6".
1054 if (!(rt = nd6_lookup(dst6, 0, NULL)))
1058 if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1059 (rt->rt_flags & RTF_LLINFO) == 0 ||
1060 !rt->rt_llinfo || !rt->rt_gateway ||
1061 rt->rt_gateway->sa_family != AF_LINK) {
1062 /* This is not a host route. */
1066 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1067 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1071 * if we get upper-layer reachability confirmation many times,
1072 * it is possible we have false information.
1076 if (ln->ln_byhint > nd6_maxnudhint)
1080 ln->ln_state = ND6_LLINFO_REACHABLE;
1082 ln->ln_expire = time_second +
1083 nd_ifinfo[rt->rt_ifp->if_index].reachable;
1087 nd6_rtrequest(req, rt, info)
1090 struct rt_addrinfo *info; /* xxx unused */
1092 struct sockaddr *gate = rt->rt_gateway;
1093 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1094 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1095 struct ifnet *ifp = rt->rt_ifp;
1098 if ((rt->rt_flags & RTF_GATEWAY))
1101 if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1103 * This is probably an interface direct route for a link
1104 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1105 * We do not need special treatment below for such a route.
1106 * Moreover, the RTF_LLINFO flag which would be set below
1107 * would annoy the ndp(8) command.
1112 if (req == RTM_RESOLVE &&
1113 (nd6_need_cache(ifp) == 0 || /* stf case */
1114 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
1116 * FreeBSD and BSD/OS often make a cloned host route based
1117 * on a less-specific route (e.g. the default route).
1118 * If the less specific route does not have a "gateway"
1119 * (this is the case when the route just goes to a p2p or an
1120 * stf interface), we'll mistakenly make a neighbor cache for
1121 * the host route, and will see strange neighbor solicitation
1122 * for the corresponding destination. In order to avoid the
1123 * confusion, we check if the destination of the route is
1124 * a neighbor in terms of neighbor discovery, and stop the
1125 * process if not. Additionally, we remove the LLINFO flag
1126 * so that ndp(8) will not try to get the neighbor information
1127 * of the destination.
1129 rt->rt_flags &= ~RTF_LLINFO;
1136 * There is no backward compatibility :)
1138 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1139 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1140 * rt->rt_flags |= RTF_CLONING;
1142 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1144 * Case 1: This route should come from
1145 * a route to interface. RTF_LLINFO flag is set
1146 * for a host route whose destination should be
1147 * treated as on-link.
1149 rt_setgate(rt, rt_key(rt),
1150 (struct sockaddr *)&null_sdl);
1151 gate = rt->rt_gateway;
1152 SDL(gate)->sdl_type = ifp->if_type;
1153 SDL(gate)->sdl_index = ifp->if_index;
1155 ln->ln_expire = time_second;
1157 if (ln && ln->ln_expire == 0) {
1158 /* kludge for desktops */
1160 printf("nd6_rtequest: time.tv_sec is zero; "
1166 if ((rt->rt_flags & RTF_CLONING))
1170 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1171 * We don't do that here since llinfo is not ready yet.
1173 * There are also couple of other things to be discussed:
1174 * - unsolicited NA code needs improvement beforehand
1175 * - RFC2461 says we MAY send multicast unsolicited NA
1176 * (7.2.6 paragraph 4), however, it also says that we
1177 * SHOULD provide a mechanism to prevent multicast NA storm.
1178 * we don't have anything like it right now.
1179 * note that the mechanism needs a mutual agreement
1180 * between proxies, which means that we need to implement
1181 * a new protocol, or a new kludge.
1182 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1183 * we need to check ip6forwarding before sending it.
1184 * (or should we allow proxy ND configuration only for
1185 * routers? there's no mention about proxy ND from hosts)
1188 /* XXX it does not work */
1189 if (rt->rt_flags & RTF_ANNOUNCE)
1191 &SIN6(rt_key(rt))->sin6_addr,
1192 &SIN6(rt_key(rt))->sin6_addr,
1193 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1198 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1200 * Address resolution isn't necessary for a point to
1201 * point link, so we can skip this test for a p2p link.
1203 if (gate->sa_family != AF_LINK ||
1204 gate->sa_len < sizeof(null_sdl)) {
1206 "nd6_rtrequest: bad gateway value: %s\n",
1210 SDL(gate)->sdl_type = ifp->if_type;
1211 SDL(gate)->sdl_index = ifp->if_index;
1214 break; /* This happens on a route change */
1216 * Case 2: This route may come from cloning, or a manual route
1217 * add with a LL address.
1219 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1220 rt->rt_llinfo = (caddr_t)ln;
1222 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1227 Bzero(ln, sizeof(*ln));
1229 /* this is required for "ndp" command. - shin */
1230 if (req == RTM_ADD) {
1232 * gate should have some valid AF_LINK entry,
1233 * and ln->ln_expire should have some lifetime
1234 * which is specified by ndp command.
1236 ln->ln_state = ND6_LLINFO_REACHABLE;
1240 * When req == RTM_RESOLVE, rt is created and
1241 * initialized in rtrequest(), so rt_expire is 0.
1243 ln->ln_state = ND6_LLINFO_NOSTATE;
1244 ln->ln_expire = time_second;
1246 rt->rt_flags |= RTF_LLINFO;
1247 ln->ln_next = llinfo_nd6.ln_next;
1248 llinfo_nd6.ln_next = ln;
1249 ln->ln_prev = &llinfo_nd6;
1250 ln->ln_next->ln_prev = ln;
1253 * check if rt_key(rt) is one of my address assigned
1256 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1257 &SIN6(rt_key(rt))->sin6_addr);
1259 caddr_t macp = nd6_ifptomac(ifp);
1261 ln->ln_state = ND6_LLINFO_REACHABLE;
1264 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1265 SDL(gate)->sdl_alen = ifp->if_addrlen;
1267 if (nd6_useloopback) {
1268 rt->rt_ifp = &loif[0]; /* XXX */
1270 * Make sure rt_ifa be equal to the ifaddr
1271 * corresponding to the address.
1272 * We need this because when we refer
1273 * rt_ifa->ia6_flags in ip6_input, we assume
1274 * that the rt_ifa points to the address instead
1275 * of the loopback address.
1277 if (ifa != rt->rt_ifa) {
1278 IFAFREE(rt->rt_ifa);
1283 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1285 ln->ln_state = ND6_LLINFO_REACHABLE;
1288 /* join solicited node multicast for proxy ND */
1289 if (ifp->if_flags & IFF_MULTICAST) {
1290 struct in6_addr llsol;
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 if (!in6_addmulti(&llsol, ifp, &error)) {
1301 nd6log((LOG_ERR, "%s: failed to join "
1302 "%s (errno=%d)\n", if_name(ifp),
1303 ip6_sprintf(&llsol), error));
1312 /* leave from solicited node multicast for proxy ND */
1313 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1314 (ifp->if_flags & IFF_MULTICAST) != 0) {
1315 struct in6_addr llsol;
1316 struct in6_multi *in6m;
1318 llsol = SIN6(rt_key(rt))->sin6_addr;
1319 llsol.s6_addr16[0] = htons(0xff02);
1320 llsol.s6_addr16[1] = htons(ifp->if_index);
1321 llsol.s6_addr32[1] = 0;
1322 llsol.s6_addr32[2] = htonl(1);
1323 llsol.s6_addr8[12] = 0xff;
1325 IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1330 ln->ln_next->ln_prev = ln->ln_prev;
1331 ln->ln_prev->ln_next = ln->ln_next;
1334 rt->rt_flags &= ~RTF_LLINFO;
1336 m_freem(ln->ln_hold);
1342 nd6_ioctl(cmd, data, ifp)
1347 struct in6_drlist *drl = (struct in6_drlist *)data;
1348 struct in6_prlist *prl = (struct in6_prlist *)data;
1349 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1350 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1351 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1352 struct nd_defrouter *dr, any;
1353 struct nd_prefix *pr;
1355 int i = 0, error = 0;
1359 case SIOCGDRLST_IN6:
1361 * obsolete API, use sysctl under net.inet6.icmp6
1363 bzero(drl, sizeof(*drl));
1365 dr = TAILQ_FIRST(&nd_defrouter);
1366 while (dr && i < DRLSTSIZ) {
1367 drl->defrouter[i].rtaddr = dr->rtaddr;
1368 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1369 /* XXX: need to this hack for KAME stack */
1370 drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1373 "default router list contains a "
1374 "non-linklocal address(%s)\n",
1375 ip6_sprintf(&drl->defrouter[i].rtaddr));
1377 drl->defrouter[i].flags = dr->flags;
1378 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1379 drl->defrouter[i].expire = dr->expire;
1380 drl->defrouter[i].if_index = dr->ifp->if_index;
1382 dr = TAILQ_NEXT(dr, dr_entry);
1386 case SIOCGPRLST_IN6:
1388 * obsolete API, use sysctl under net.inet6.icmp6
1391 * XXX meaning of fields, especialy "raflags", is very
1392 * differnet between RA prefix list and RR/static prefix list.
1393 * how about separating ioctls into two?
1395 bzero(prl, sizeof(*prl));
1397 pr = nd_prefix.lh_first;
1398 while (pr && i < PRLSTSIZ) {
1399 struct nd_pfxrouter *pfr;
1402 (void)in6_embedscope(&prl->prefix[i].prefix,
1403 &pr->ndpr_prefix, NULL, NULL);
1404 prl->prefix[i].raflags = pr->ndpr_raf;
1405 prl->prefix[i].prefixlen = pr->ndpr_plen;
1406 prl->prefix[i].vltime = pr->ndpr_vltime;
1407 prl->prefix[i].pltime = pr->ndpr_pltime;
1408 prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1409 prl->prefix[i].expire = pr->ndpr_expire;
1411 pfr = pr->ndpr_advrtrs.lh_first;
1415 #define RTRADDR prl->prefix[i].advrtr[j]
1416 RTRADDR = pfr->router->rtaddr;
1417 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1418 /* XXX: hack for KAME */
1419 RTRADDR.s6_addr16[1] = 0;
1422 "a router(%s) advertises "
1424 "non-link local address\n",
1425 ip6_sprintf(&RTRADDR));
1429 pfr = pfr->pfr_next;
1431 prl->prefix[i].advrtrs = j;
1432 prl->prefix[i].origin = PR_ORIG_RA;
1438 struct rr_prefix *rpp;
1440 for (rpp = LIST_FIRST(&rr_prefix); rpp;
1441 rpp = LIST_NEXT(rpp, rp_entry)) {
1444 (void)in6_embedscope(&prl->prefix[i].prefix,
1445 &pr->ndpr_prefix, NULL, NULL);
1446 prl->prefix[i].raflags = rpp->rp_raf;
1447 prl->prefix[i].prefixlen = rpp->rp_plen;
1448 prl->prefix[i].vltime = rpp->rp_vltime;
1449 prl->prefix[i].pltime = rpp->rp_pltime;
1450 prl->prefix[i].if_index = rpp->rp_ifp->if_index;
1451 prl->prefix[i].expire = rpp->rp_expire;
1452 prl->prefix[i].advrtrs = 0;
1453 prl->prefix[i].origin = rpp->rp_origin;
1460 case OSIOCGIFINFO_IN6:
1461 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1465 ndi->ndi.linkmtu = nd_ifinfo[ifp->if_index].linkmtu;
1466 ndi->ndi.maxmtu = nd_ifinfo[ifp->if_index].maxmtu;
1467 ndi->ndi.basereachable =
1468 nd_ifinfo[ifp->if_index].basereachable;
1469 ndi->ndi.reachable = nd_ifinfo[ifp->if_index].reachable;
1470 ndi->ndi.retrans = nd_ifinfo[ifp->if_index].retrans;
1471 ndi->ndi.flags = nd_ifinfo[ifp->if_index].flags;
1472 ndi->ndi.recalctm = nd_ifinfo[ifp->if_index].recalctm;
1473 ndi->ndi.chlim = nd_ifinfo[ifp->if_index].chlim;
1474 ndi->ndi.receivedra = nd_ifinfo[ifp->if_index].receivedra;
1476 case SIOCGIFINFO_IN6:
1477 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1481 ndi->ndi = nd_ifinfo[ifp->if_index];
1483 case SIOCSIFINFO_FLAGS:
1484 /* XXX: almost all other fields of ndi->ndi is unused */
1485 if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1489 nd_ifinfo[ifp->if_index].flags = ndi->ndi.flags;
1491 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1492 /* flush default router list */
1494 * xxx sumikawa: should not delete route if default
1495 * route equals to the top of default router list
1497 bzero(&any, sizeof(any));
1498 defrouter_delreq(&any, 0);
1500 /* xxx sumikawa: flush prefix list */
1502 case SIOCSPFXFLUSH_IN6:
1504 /* flush all the prefix advertised by routers */
1505 struct nd_prefix *pr, *next;
1508 for (pr = nd_prefix.lh_first; pr; pr = next) {
1509 struct in6_ifaddr *ia, *ia_next;
1511 next = pr->ndpr_next;
1513 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1516 /* do we really have to remove addresses as well? */
1517 for (ia = in6_ifaddr; ia; ia = ia_next) {
1518 /* ia might be removed. keep the next ptr. */
1519 ia_next = ia->ia_next;
1521 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1524 if (ia->ia6_ndpr == pr)
1525 in6_purgeaddr(&ia->ia_ifa);
1532 case SIOCSRTRFLUSH_IN6:
1534 /* flush all the default routers */
1535 struct nd_defrouter *dr, *next;
1538 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1540 * The first entry of the list may be stored in
1541 * the routing table, so we'll delete it later.
1543 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1544 next = TAILQ_NEXT(dr, dr_entry);
1547 defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
1552 case SIOCGNBRINFO_IN6:
1554 struct llinfo_nd6 *ln;
1555 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1558 * XXX: KAME specific hack for scoped addresses
1559 * XXXX: for other scopes than link-local?
1561 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1562 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1563 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1566 *idp = htons(ifp->if_index);
1570 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1575 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1576 nbi->state = ln->ln_state;
1577 nbi->asked = ln->ln_asked;
1578 nbi->isrouter = ln->ln_router;
1579 nbi->expire = ln->ln_expire;
1584 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1585 ndif->ifindex = nd6_defifindex;
1587 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1588 return(nd6_setdefaultiface(ndif->ifindex));
1595 * Create neighbor cache entry and cache link-layer address,
1596 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1599 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
1601 struct in6_addr *from;
1604 int type; /* ICMP6 type */
1605 int code; /* type dependent information */
1607 struct rtentry *rt = NULL;
1608 struct llinfo_nd6 *ln = NULL;
1610 struct sockaddr_dl *sdl = NULL;
1617 panic("ifp == NULL in nd6_cache_lladdr");
1619 panic("from == NULL in nd6_cache_lladdr");
1621 /* nothing must be updated for unspecified address */
1622 if (IN6_IS_ADDR_UNSPECIFIED(from))
1626 * Validation about ifp->if_addrlen and lladdrlen must be done in
1629 * XXX If the link does not have link-layer adderss, what should
1630 * we do? (ifp->if_addrlen == 0)
1631 * Spec says nothing in sections for RA, RS and NA. There's small
1632 * description on it in NS section (RFC 2461 7.2.3).
1635 rt = nd6_lookup(from, 0, ifp);
1638 /* nothing must be done if there's no lladdr */
1639 if (!lladdr || !lladdrlen)
1643 rt = nd6_lookup(from, 1, ifp);
1646 /* do nothing if static ndp is set */
1647 if (rt->rt_flags & RTF_STATIC)
1654 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1659 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1662 if (!rt->rt_gateway)
1664 if (rt->rt_gateway->sa_family != AF_LINK)
1666 sdl = SDL(rt->rt_gateway);
1668 olladdr = (sdl->sdl_alen) ? 1 : 0;
1669 if (olladdr && lladdr) {
1670 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1678 * newentry olladdr lladdr llchange (*=record)
1681 * 0 n y -- (3) * STALE
1683 * 0 y y y (5) * STALE
1684 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1685 * 1 -- y -- (7) * STALE
1688 if (lladdr) { /* (3-5) and (7) */
1690 * Record source link-layer address
1691 * XXX is it dependent to ifp->if_type?
1693 sdl->sdl_alen = ifp->if_addrlen;
1694 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1698 if ((!olladdr && lladdr) /* (3) */
1699 || (olladdr && lladdr && llchange)) { /* (5) */
1701 newstate = ND6_LLINFO_STALE;
1702 } else /* (1-2,4) */
1706 if (!lladdr) /* (6) */
1707 newstate = ND6_LLINFO_NOSTATE;
1709 newstate = ND6_LLINFO_STALE;
1714 * Update the state of the neighbor cache.
1716 ln->ln_state = newstate;
1718 if (ln->ln_state == ND6_LLINFO_STALE) {
1720 * XXX: since nd6_output() below will cause
1721 * state tansition to DELAY and reset the timer,
1722 * we must set the timer now, although it is actually
1725 ln->ln_expire = time_second + nd6_gctimer;
1729 * we assume ifp is not a p2p here, so just
1730 * set the 2nd argument as the 1st one.
1732 nd6_output(ifp, ifp, ln->ln_hold,
1733 (struct sockaddr_in6 *)rt_key(rt),
1737 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1738 /* probe right away */
1739 ln->ln_expire = time_second;
1744 * ICMP6 type dependent behavior.
1746 * NS: clear IsRouter if new entry
1747 * RS: clear IsRouter
1748 * RA: set IsRouter if there's lladdr
1749 * redir: clear IsRouter if new entry
1752 * The spec says that we must set IsRouter in the following cases:
1753 * - If lladdr exist, set IsRouter. This means (1-5).
1754 * - If it is old entry (!newentry), set IsRouter. This means (7).
1755 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1756 * A quetion arises for (1) case. (1) case has no lladdr in the
1757 * neighbor cache, this is similar to (6).
1758 * This case is rare but we figured that we MUST NOT set IsRouter.
1760 * newentry olladdr lladdr llchange NS RS RA redir
1762 * 0 n n -- (1) c ? s
1763 * 0 y n -- (2) c s s
1764 * 0 n y -- (3) c s s
1767 * 1 -- n -- (6) c c c s
1768 * 1 -- y -- (7) c c s c s
1772 switch (type & 0xff) {
1773 case ND_NEIGHBOR_SOLICIT:
1775 * New entry must have is_router flag cleared.
1777 if (is_newentry) /* (6-7) */
1782 * If the icmp is a redirect to a better router, always set the
1783 * is_router flag. Otherwise, if the entry is newly created,
1784 * clear the flag. [RFC 2461, sec 8.3]
1786 if (code == ND_REDIRECT_ROUTER)
1788 else if (is_newentry) /* (6-7) */
1791 case ND_ROUTER_SOLICIT:
1793 * is_router flag must always be cleared.
1797 case ND_ROUTER_ADVERT:
1799 * Mark an entry with lladdr as a router.
1801 if ((!is_newentry && (olladdr || lladdr)) /* (2-5) */
1802 || (is_newentry && lladdr)) { /* (7) */
1809 * When the link-layer address of a router changes, select the
1810 * best router again. In particular, when the neighbor entry is newly
1811 * created, it might affect the selection policy.
1812 * Question: can we restrict the first condition to the "is_newentry"
1814 * XXX: when we hear an RA from a new router with the link-layer
1815 * address option, defrouter_select() is called twice, since
1816 * defrtrlist_update called the function as well. However, I believe
1817 * we can compromise the overhead, since it only happens the first
1819 * XXX: although defrouter_select() should not have a bad effect
1820 * for those are not autoconfigured hosts, we explicitly avoid such
1823 if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
1830 nd6_slowtimo(ignored_arg)
1835 struct nd_ifinfo *nd6if;
1837 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1838 nd6_slowtimo, NULL);
1839 for (i = 1; i < if_index + 1; i++) {
1840 if (!nd_ifinfo || i >= nd_ifinfo_indexlim)
1842 nd6if = &nd_ifinfo[i];
1843 if (nd6if->basereachable && /* already initialized */
1844 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1846 * Since reachable time rarely changes by router
1847 * advertisements, we SHOULD insure that a new random
1848 * value gets recomputed at least once every few hours.
1851 nd6if->recalctm = nd6_recalc_reachtm_interval;
1852 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1858 #define senderr(e) { error = (e); goto bad;}
1860 nd6_output(ifp, origifp, m0, dst, rt0)
1862 struct ifnet *origifp;
1864 struct sockaddr_in6 *dst;
1865 struct rtentry *rt0;
1867 struct mbuf *m = m0;
1868 struct rtentry *rt = rt0;
1869 struct sockaddr_in6 *gw6 = NULL;
1870 struct llinfo_nd6 *ln = NULL;
1873 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1876 if (nd6_need_cache(ifp) == 0)
1880 * next hop determination. This routine is derived from ether_outpout.
1883 if ((rt->rt_flags & RTF_UP) == 0) {
1884 if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL)) !=
1888 if (rt->rt_ifp != ifp) {
1889 /* XXX: loop care? */
1890 return nd6_output(ifp, origifp, m0,
1894 senderr(EHOSTUNREACH);
1897 if (rt->rt_flags & RTF_GATEWAY) {
1898 gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1901 * We skip link-layer address resolution and NUD
1902 * if the gateway is not a neighbor from ND point
1903 * of view, regardless of the value of nd_ifinfo.flags.
1904 * The second condition is a bit tricky; we skip
1905 * if the gateway is our own address, which is
1906 * sometimes used to install a route to a p2p link.
1908 if (!nd6_is_addr_neighbor(gw6, ifp) ||
1909 in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1911 * We allow this kind of tricky route only
1912 * when the outgoing interface is p2p.
1913 * XXX: we may need a more generic rule here.
1915 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1916 senderr(EHOSTUNREACH);
1921 if (rt->rt_gwroute == 0)
1923 if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1924 rtfree(rt); rt = rt0;
1925 lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
1926 if ((rt = rt->rt_gwroute) == 0)
1927 senderr(EHOSTUNREACH);
1933 * Address resolution or Neighbor Unreachability Detection
1935 * At this point, the destination of the packet must be a unicast
1936 * or an anycast address(i.e. not a multicast).
1939 /* Look up the neighbor cache for the nexthop */
1940 if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
1941 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1944 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1945 * the condition below is not very efficient. But we believe
1946 * it is tolerable, because this should be a rare case.
1948 if (nd6_is_addr_neighbor(dst, ifp) &&
1949 (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
1950 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1953 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1954 !(nd_ifinfo[ifp->if_index].flags & ND6_IFF_PERFORMNUD)) {
1956 "nd6_output: can't allocate llinfo for %s "
1958 ip6_sprintf(&dst->sin6_addr), ln, rt);
1959 senderr(EIO); /* XXX: good error? */
1962 goto sendpkt; /* send anyway */
1965 /* We don't have to do link-layer address resolution on a p2p link. */
1966 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1967 ln->ln_state < ND6_LLINFO_REACHABLE) {
1968 ln->ln_state = ND6_LLINFO_STALE;
1969 ln->ln_expire = time_second + nd6_gctimer;
1973 * The first time we send a packet to a neighbor whose entry is
1974 * STALE, we have to change the state to DELAY and a sets a timer to
1975 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1976 * neighbor unreachability detection on expiration.
1979 if (ln->ln_state == ND6_LLINFO_STALE) {
1981 ln->ln_state = ND6_LLINFO_DELAY;
1982 ln->ln_expire = time_second + nd6_delay;
1986 * If the neighbor cache entry has a state other than INCOMPLETE
1987 * (i.e. its link-layer address is already resolved), just
1990 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1994 * There is a neighbor cache entry, but no ethernet address
1995 * response yet. Replace the held mbuf (if any) with this
1998 * This code conforms to the rate-limiting rule described in Section
1999 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
2002 if (ln->ln_state == ND6_LLINFO_NOSTATE)
2003 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2005 m_freem(ln->ln_hold);
2007 if (ln->ln_expire) {
2008 if (ln->ln_asked < nd6_mmaxtries &&
2009 ln->ln_expire < time_second) {
2011 ln->ln_expire = time_second +
2012 nd_ifinfo[ifp->if_index].retrans / 1000;
2013 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2020 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2021 return((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2024 return((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
2038 * XXX: we currently do not make neighbor cache on any interface
2039 * other than ARCnet, Ethernet, FDDI and GIF.
2042 * - unidirectional tunnels needs no ND
2044 switch (ifp->if_type) {
2052 #ifdef IFT_IEEE80211
2055 case IFT_GIF: /* XXX need more cases? */
2063 nd6_storelladdr(ifp, rt, m, dst, desten)
2067 struct sockaddr *dst;
2071 struct sockaddr_dl *sdl;
2073 if (m->m_flags & M_MCAST) {
2074 switch (ifp->if_type) {
2080 #ifdef IFT_IEEE80211
2083 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2088 * netbsd can use if_broadcastaddr, but we don't do so
2089 * to reduce # of ifdef.
2091 for (i = 0; i < ifp->if_addrlen; i++)
2104 /* this could happen, if we could not allocate memory */
2108 if (rt->rt_gateway->sa_family != AF_LINK) {
2109 printf("nd6_storelladdr: something odd happens\n");
2113 sdl = SDL(rt->rt_gateway);
2114 if (sdl->sdl_alen == 0) {
2115 /* this should be impossible, but we bark here for debugging */
2116 printf("nd6_storelladdr: sdl_alen == 0\n");
2121 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2125 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2126 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2128 SYSCTL_DECL(_net_inet6_icmp6);
2130 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2131 CTLFLAG_RD, nd6_sysctl_drlist, "");
2132 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2133 CTLFLAG_RD, nd6_sysctl_prlist, "");
2136 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2140 struct in6_defrouter *d, *de;
2141 struct nd_defrouter *dr;
2147 for (dr = TAILQ_FIRST(&nd_defrouter);
2149 dr = TAILQ_NEXT(dr, dr_entry)) {
2150 d = (struct in6_defrouter *)buf;
2151 de = (struct in6_defrouter *)(buf + sizeof(buf));
2154 bzero(d, sizeof(*d));
2155 d->rtaddr.sin6_family = AF_INET6;
2156 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2157 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2161 "default router list (%s)\n",
2162 ip6_sprintf(&dr->rtaddr));
2163 d->flags = dr->flags;
2164 d->rtlifetime = dr->rtlifetime;
2165 d->expire = dr->expire;
2166 d->if_index = dr->ifp->if_index;
2168 panic("buffer too short");
2170 error = SYSCTL_OUT(req, buf, sizeof(*d));
2178 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2182 struct in6_prefix *p, *pe;
2183 struct nd_prefix *pr;
2189 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2192 struct sockaddr_in6 *sin6, *s6;
2193 struct nd_pfxrouter *pfr;
2195 p = (struct in6_prefix *)buf;
2196 pe = (struct in6_prefix *)(buf + sizeof(buf));
2199 bzero(p, sizeof(*p));
2200 sin6 = (struct sockaddr_in6 *)(p + 1);
2202 p->prefix = pr->ndpr_prefix;
2203 if (in6_recoverscope(&p->prefix,
2204 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2206 "scope error in prefix list (%s)\n",
2207 ip6_sprintf(&p->prefix.sin6_addr));
2208 p->raflags = pr->ndpr_raf;
2209 p->prefixlen = pr->ndpr_plen;
2210 p->vltime = pr->ndpr_vltime;
2211 p->pltime = pr->ndpr_pltime;
2212 p->if_index = pr->ndpr_ifp->if_index;
2213 p->expire = pr->ndpr_expire;
2214 p->refcnt = pr->ndpr_refcnt;
2215 p->flags = pr->ndpr_stateflags;
2216 p->origin = PR_ORIG_RA;
2218 for (pfr = pr->ndpr_advrtrs.lh_first;
2220 pfr = pfr->pfr_next) {
2221 if ((void *)&sin6[advrtrs + 1] >
2226 s6 = &sin6[advrtrs];
2227 bzero(s6, sizeof(*s6));
2228 s6->sin6_family = AF_INET6;
2229 s6->sin6_len = sizeof(*sin6);
2230 if (in6_recoverscope(s6,
2231 &pfr->router->rtaddr,
2232 pfr->router->ifp) != 0)
2235 "prefix list (%s)\n",
2236 ip6_sprintf(&pfr->router->rtaddr));
2239 p->advrtrs = advrtrs;
2241 panic("buffer too short");
2243 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2244 error = SYSCTL_OUT(req, buf, advance);