2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/callout.h>
41 #include <sys/malloc.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
46 #include <sys/kernel.h>
47 #include <sys/protosw.h>
48 #include <sys/errno.h>
49 #include <sys/syslog.h>
51 #include <sys/rwlock.h>
52 #include <sys/queue.h>
53 #include <sys/sysctl.h>
56 #include <net/if_arc.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/iso88025.h>
61 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <net/if_llatbl.h>
66 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le))
67 #include <netinet/if_ether.h>
68 #include <netinet6/in6_var.h>
69 #include <netinet/ip6.h>
70 #include <netinet6/ip6_var.h>
71 #include <netinet6/scope6_var.h>
72 #include <netinet6/nd6.h>
73 #include <netinet6/in6_ifattach.h>
74 #include <netinet/icmp6.h>
76 #include <sys/limits.h>
78 #include <security/mac/mac_framework.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)
86 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
87 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
88 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
89 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
90 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
92 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
95 /* preventing too many loops in ND option parsing */
96 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
98 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
100 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
102 #define V_nd6_maxndopt VNET(nd6_maxndopt)
103 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
106 VNET_DEFINE(int, nd6_debug) = 1;
108 VNET_DEFINE(int, nd6_debug) = 0;
113 static int nd6_inuse, nd6_allocated;
116 VNET_DEFINE(struct nd_drhead, nd_defrouter);
117 VNET_DEFINE(struct nd_prhead, nd_prefix);
119 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
120 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
122 static struct sockaddr_in6 all1_sa;
124 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
126 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
127 static void nd6_slowtimo(void *);
128 static int regen_tmpaddr(struct in6_ifaddr *);
129 static struct llentry *nd6_free(struct llentry *, int);
130 static void nd6_llinfo_timer(void *);
131 static void clear_llinfo_pqueue(struct llentry *);
133 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
134 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
136 VNET_DEFINE(struct callout, nd6_timer_ch);
143 LIST_INIT(&V_nd_prefix);
145 all1_sa.sin6_family = AF_INET6;
146 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
147 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
148 all1_sa.sin6_addr.s6_addr[i] = 0xff;
150 /* initialization of the default router list */
151 TAILQ_INIT(&V_nd_defrouter);
154 callout_init(&V_nd6_slowtimo_ch, 0);
155 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
156 nd6_slowtimo, curvnet);
164 callout_drain(&V_nd6_slowtimo_ch);
165 callout_drain(&V_nd6_timer_ch);
170 nd6_ifattach(struct ifnet *ifp)
172 struct nd_ifinfo *nd;
174 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
175 bzero(nd, sizeof(*nd));
179 nd->chlim = IPV6_DEFHLIM;
180 nd->basereachable = REACHABLE_TIME;
181 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
182 nd->retrans = RETRANS_TIMER;
184 nd->flags = ND6_IFF_PERFORMNUD;
186 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */
187 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK))
188 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
190 /* A loopback interface does not need to accept RTADV. */
191 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK))
192 nd->flags |= ND6_IFF_ACCEPT_RTADV;
194 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
195 nd6_setmtu0(ifp, nd);
201 nd6_ifdetach(struct nd_ifinfo *nd)
208 * Reset ND level link MTU. This function is called when the physical MTU
209 * changes, which means we might have to adjust the ND level MTU.
212 nd6_setmtu(struct ifnet *ifp)
215 nd6_setmtu0(ifp, ND_IFINFO(ifp));
218 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
220 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
224 omaxmtu = ndi->maxmtu;
226 switch (ifp->if_type) {
228 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
231 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
234 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
237 ndi->maxmtu = ifp->if_mtu;
242 * Decreasing the interface MTU under IPV6 minimum MTU may cause
243 * undesirable situation. We thus notify the operator of the change
244 * explicitly. The check for omaxmtu is necessary to restrict the
245 * log to the case of changing the MTU, not initializing it.
247 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
248 log(LOG_NOTICE, "nd6_setmtu0: "
249 "new link MTU on %s (%lu) is too small for IPv6\n",
250 if_name(ifp), (unsigned long)ndi->maxmtu);
253 if (ndi->maxmtu > V_in6_maxmtu)
254 in6_setmaxmtu(); /* check all interfaces just in case */
259 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
262 bzero(ndopts, sizeof(*ndopts));
263 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
265 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
268 ndopts->nd_opts_done = 1;
269 ndopts->nd_opts_search = NULL;
274 * Take one ND option.
277 nd6_option(union nd_opts *ndopts)
279 struct nd_opt_hdr *nd_opt;
283 panic("ndopts == NULL in nd6_option");
284 if (ndopts->nd_opts_last == NULL)
285 panic("uninitialized ndopts in nd6_option");
286 if (ndopts->nd_opts_search == NULL)
288 if (ndopts->nd_opts_done)
291 nd_opt = ndopts->nd_opts_search;
293 /* make sure nd_opt_len is inside the buffer */
294 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
295 bzero(ndopts, sizeof(*ndopts));
299 olen = nd_opt->nd_opt_len << 3;
302 * Message validation requires that all included
303 * options have a length that is greater than zero.
305 bzero(ndopts, sizeof(*ndopts));
309 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
310 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
311 /* option overruns the end of buffer, invalid */
312 bzero(ndopts, sizeof(*ndopts));
314 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
315 /* reached the end of options chain */
316 ndopts->nd_opts_done = 1;
317 ndopts->nd_opts_search = NULL;
323 * Parse multiple ND options.
324 * This function is much easier to use, for ND routines that do not need
325 * multiple options of the same type.
328 nd6_options(union nd_opts *ndopts)
330 struct nd_opt_hdr *nd_opt;
334 panic("ndopts == NULL in nd6_options");
335 if (ndopts->nd_opts_last == NULL)
336 panic("uninitialized ndopts in nd6_options");
337 if (ndopts->nd_opts_search == NULL)
341 nd_opt = nd6_option(ndopts);
342 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
344 * Message validation requires that all included
345 * options have a length that is greater than zero.
347 ICMP6STAT_INC(icp6s_nd_badopt);
348 bzero(ndopts, sizeof(*ndopts));
355 switch (nd_opt->nd_opt_type) {
356 case ND_OPT_SOURCE_LINKADDR:
357 case ND_OPT_TARGET_LINKADDR:
359 case ND_OPT_REDIRECTED_HEADER:
360 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
362 "duplicated ND6 option found (type=%d)\n",
363 nd_opt->nd_opt_type));
366 ndopts->nd_opt_array[nd_opt->nd_opt_type]
370 case ND_OPT_PREFIX_INFORMATION:
371 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
372 ndopts->nd_opt_array[nd_opt->nd_opt_type]
375 ndopts->nd_opts_pi_end =
376 (struct nd_opt_prefix_info *)nd_opt;
380 * Unknown options must be silently ignored,
381 * to accomodate future extension to the protocol.
384 "nd6_options: unsupported option %d - "
385 "option ignored\n", nd_opt->nd_opt_type));
390 if (i > V_nd6_maxndopt) {
391 ICMP6STAT_INC(icp6s_nd_toomanyopt);
392 nd6log((LOG_INFO, "too many loop in nd opt\n"));
396 if (ndopts->nd_opts_done)
404 * ND6 timer routine to handle ND6 entries
407 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
414 canceled = callout_stop(&ln->ln_timer_ch);
416 ln->la_expire = time_second + tick / hz;
418 if (tick > INT_MAX) {
419 ln->ln_ntick = tick - INT_MAX;
420 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
421 nd6_llinfo_timer, ln);
424 canceled = callout_reset(&ln->ln_timer_ch, tick,
425 nd6_llinfo_timer, ln);
433 nd6_llinfo_settimer(struct llentry *ln, long tick)
437 nd6_llinfo_settimer_locked(ln, tick);
442 nd6_llinfo_timer(void *arg)
445 struct in6_addr *dst;
447 struct nd_ifinfo *ndi = NULL;
449 ln = (struct llentry *)arg;
451 panic("%s: NULL entry!\n", __func__);
455 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL)
456 panic("ln ifp == NULL");
458 CURVNET_SET(ifp->if_vnet);
460 if (ln->ln_ntick > 0) {
461 if (ln->ln_ntick > INT_MAX) {
462 ln->ln_ntick -= INT_MAX;
463 nd6_llinfo_settimer(ln, INT_MAX);
466 nd6_llinfo_settimer(ln, ln->ln_ntick);
471 ndi = ND_IFINFO(ifp);
472 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
473 if (ln->la_flags & LLE_STATIC) {
477 if (ln->la_flags & LLE_DELETED) {
478 (void)nd6_free(ln, 0);
483 switch (ln->ln_state) {
484 case ND6_LLINFO_INCOMPLETE:
485 if (ln->la_asked < V_nd6_mmaxtries) {
487 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
488 nd6_ns_output(ifp, NULL, dst, ln, 0);
490 struct mbuf *m = ln->la_hold;
495 * assuming every packet in la_hold has the
500 icmp6_error2(m, ICMP6_DST_UNREACH,
501 ICMP6_DST_UNREACH_ADDR, 0, ifp);
504 clear_llinfo_pqueue(ln);
506 (void)nd6_free(ln, 0);
510 case ND6_LLINFO_REACHABLE:
511 if (!ND6_LLINFO_PERMANENT(ln)) {
512 ln->ln_state = ND6_LLINFO_STALE;
513 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
517 case ND6_LLINFO_STALE:
518 /* Garbage Collection(RFC 2461 5.3) */
519 if (!ND6_LLINFO_PERMANENT(ln)) {
520 (void)nd6_free(ln, 1);
525 case ND6_LLINFO_DELAY:
526 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
529 ln->ln_state = ND6_LLINFO_PROBE;
530 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
531 nd6_ns_output(ifp, dst, dst, ln, 0);
533 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
534 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
537 case ND6_LLINFO_PROBE:
538 if (ln->la_asked < V_nd6_umaxtries) {
540 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
541 nd6_ns_output(ifp, dst, dst, ln, 0);
543 (void)nd6_free(ln, 0);
556 * ND6 timer routine to expire default route list and prefix list
561 CURVNET_SET((struct vnet *) arg);
563 struct nd_defrouter *dr;
564 struct nd_prefix *pr;
565 struct in6_ifaddr *ia6, *nia6;
566 struct in6_addrlifetime *lt6;
568 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
571 /* expire default router list */
573 dr = TAILQ_FIRST(&V_nd_defrouter);
575 if (dr->expire && dr->expire < time_second) {
576 struct nd_defrouter *t;
577 t = TAILQ_NEXT(dr, dr_entry);
581 dr = TAILQ_NEXT(dr, dr_entry);
586 * expire interface addresses.
587 * in the past the loop was inside prefix expiry processing.
588 * However, from a stricter speci-confrmance standpoint, we should
589 * rather separate address lifetimes and prefix lifetimes.
591 * XXXRW: in6_ifaddrhead locking.
594 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
595 /* check address lifetime */
596 lt6 = &ia6->ia6_lifetime;
597 if (IFA6_IS_INVALID(ia6)) {
601 * If the expiring address is temporary, try
602 * regenerating a new one. This would be useful when
603 * we suspended a laptop PC, then turned it on after a
604 * period that could invalidate all temporary
605 * addresses. Although we may have to restart the
606 * loop (see below), it must be after purging the
607 * address. Otherwise, we'd see an infinite loop of
610 if (V_ip6_use_tempaddr &&
611 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
612 if (regen_tmpaddr(ia6) == 0)
616 in6_purgeaddr(&ia6->ia_ifa);
619 goto addrloop; /* XXX: see below */
620 } else if (IFA6_IS_DEPRECATED(ia6)) {
621 int oldflags = ia6->ia6_flags;
623 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
626 * If a temporary address has just become deprecated,
627 * regenerate a new one if possible.
629 if (V_ip6_use_tempaddr &&
630 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
631 (oldflags & IN6_IFF_DEPRECATED) == 0) {
633 if (regen_tmpaddr(ia6) == 0) {
635 * A new temporary address is
637 * XXX: this means the address chain
638 * has changed while we are still in
639 * the loop. Although the change
640 * would not cause disaster (because
641 * it's not a deletion, but an
642 * addition,) we'd rather restart the
643 * loop just for safety. Or does this
644 * significantly reduce performance??
651 * A new RA might have made a deprecated address
654 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
658 /* expire prefix list */
659 pr = V_nd_prefix.lh_first;
662 * check prefix lifetime.
663 * since pltime is just for autoconf, pltime processing for
664 * prefix is not necessary.
666 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
667 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
672 * address expiration and prefix expiration are
673 * separate. NEVER perform in6_purgeaddr here.
686 * ia6 - deprecated/invalidated temporary address
689 regen_tmpaddr(struct in6_ifaddr *ia6)
693 struct in6_ifaddr *public_ifa6 = NULL;
695 ifp = ia6->ia_ifa.ifa_ifp;
697 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
698 struct in6_ifaddr *it6;
700 if (ifa->ifa_addr->sa_family != AF_INET6)
703 it6 = (struct in6_ifaddr *)ifa;
705 /* ignore no autoconf addresses. */
706 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
709 /* ignore autoconf addresses with different prefixes. */
710 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
714 * Now we are looking at an autoconf address with the same
715 * prefix as ours. If the address is temporary and is still
716 * preferred, do not create another one. It would be rare, but
717 * could happen, for example, when we resume a laptop PC after
720 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
721 !IFA6_IS_DEPRECATED(it6)) {
727 * This is a public autoconf address that has the same prefix
728 * as ours. If it is preferred, keep it. We can't break the
729 * loop here, because there may be a still-preferred temporary
730 * address with the prefix.
732 if (!IFA6_IS_DEPRECATED(it6))
735 if (public_ifa6 != NULL)
736 ifa_ref(&public_ifa6->ia_ifa);
740 if (public_ifa6 != NULL) {
743 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
744 ifa_free(&public_ifa6->ia_ifa);
745 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
746 " tmp addr,errno=%d\n", e);
749 ifa_free(&public_ifa6->ia_ifa);
757 * Nuke neighbor cache/prefix/default router management table, right before
761 nd6_purge(struct ifnet *ifp)
763 struct nd_defrouter *dr, *ndr;
764 struct nd_prefix *pr, *npr;
767 * Nuke default router list entries toward ifp.
768 * We defer removal of default router list entries that is installed
769 * in the routing table, in order to keep additional side effects as
772 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
773 ndr = TAILQ_NEXT(dr, dr_entry);
781 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
782 ndr = TAILQ_NEXT(dr, dr_entry);
790 /* Nuke prefix list entries toward ifp */
791 for (pr = V_nd_prefix.lh_first; pr; pr = npr) {
793 if (pr->ndpr_ifp == ifp) {
795 * Because if_detach() does *not* release prefixes
796 * while purging addresses the reference count will
797 * still be above zero. We therefore reset it to
798 * make sure that the prefix really gets purged.
803 * Previously, pr->ndpr_addr is removed as well,
804 * but I strongly believe we don't have to do it.
805 * nd6_purge() is only called from in6_ifdetach(),
806 * which removes all the associated interface addresses
808 * (jinmei@kame.net 20010129)
814 /* cancel default outgoing interface setting */
815 if (V_nd6_defifindex == ifp->if_index)
816 nd6_setdefaultiface(0);
818 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
819 /* Refresh default router list. */
824 * We do not nuke the neighbor cache entries here any more
825 * because the neighbor cache is kept in if_afdata[AF_INET6].
826 * nd6_purge() is invoked by in6_ifdetach() which is called
827 * from if_detach() where everything gets purged. So let
828 * in6_domifdetach() do the actual L2 table purging work.
833 * the caller acquires and releases the lock on the lltbls
834 * Returns the llentry locked
837 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
839 struct sockaddr_in6 sin6;
843 bzero(&sin6, sizeof(sin6));
844 sin6.sin6_len = sizeof(struct sockaddr_in6);
845 sin6.sin6_family = AF_INET6;
846 sin6.sin6_addr = *addr6;
848 IF_AFDATA_LOCK_ASSERT(ifp);
850 if (flags & ND6_CREATE)
851 llflags |= LLE_CREATE;
852 if (flags & ND6_EXCLUSIVE)
853 llflags |= LLE_EXCLUSIVE;
855 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
856 if ((ln != NULL) && (flags & LLE_CREATE)) {
857 ln->ln_state = ND6_LLINFO_NOSTATE;
858 callout_init(&ln->ln_timer_ch, 0);
865 * Test whether a given IPv6 address is a neighbor or not, ignoring
866 * the actual neighbor cache. The neighbor cache is ignored in order
867 * to not reenter the routing code from within itself.
870 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
872 struct nd_prefix *pr;
873 struct ifaddr *dstaddr;
876 * A link-local address is always a neighbor.
877 * XXX: a link does not necessarily specify a single interface.
879 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
880 struct sockaddr_in6 sin6_copy;
884 * We need sin6_copy since sa6_recoverscope() may modify the
888 if (sa6_recoverscope(&sin6_copy))
889 return (0); /* XXX: should be impossible */
890 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
892 if (sin6_copy.sin6_scope_id == zone)
899 * If the address matches one of our addresses,
900 * it should be a neighbor.
901 * If the address matches one of our on-link prefixes, it should be a
904 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
905 if (pr->ndpr_ifp != ifp)
908 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
910 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0);
914 * This is the case where multiple interfaces
915 * have the same prefix, but only one is installed
916 * into the routing table and that prefix entry
917 * is not the one being examined here. In the case
918 * where RADIX_MPATH is enabled, multiple route
919 * entries (of the same rt_key value) will be
920 * installed because the interface addresses all
923 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
924 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
931 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
932 &addr->sin6_addr, &pr->ndpr_mask))
937 * If the address is assigned on the node of the other side of
938 * a p2p interface, the address should be a neighbor.
940 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
941 if (dstaddr != NULL) {
942 if (dstaddr->ifa_ifp == ifp) {
950 * If the default router list is empty, all addresses are regarded
951 * as on-link, and thus, as a neighbor.
952 * XXX: we restrict the condition to hosts, because routers usually do
953 * not have the "default router list".
955 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL &&
956 V_nd6_defifindex == ifp->if_index) {
965 * Detect if a given IPv6 address identifies a neighbor on a given link.
966 * XXX: should take care of the destination of a p2p link?
969 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
974 IF_AFDATA_UNLOCK_ASSERT(ifp);
975 if (nd6_is_new_addr_neighbor(addr, ifp))
979 * Even if the address matches none of our addresses, it might be
980 * in the neighbor cache.
983 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
987 IF_AFDATA_UNLOCK(ifp);
992 * Free an nd6 llinfo entry.
993 * Since the function would cause significant changes in the kernel, DO NOT
994 * make it global, unless you have a strong reason for the change, and are sure
995 * that the change is safe.
997 static struct llentry *
998 nd6_free(struct llentry *ln, int gc)
1000 struct llentry *next;
1001 struct nd_defrouter *dr;
1002 struct ifnet *ifp=NULL;
1005 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1006 * even though it is not harmful, it was not really necessary.
1010 nd6_llinfo_settimer(ln, -1);
1012 if (!V_ip6_forwarding) {
1015 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1017 if (dr != NULL && dr->expire &&
1018 ln->ln_state == ND6_LLINFO_STALE && gc) {
1020 * If the reason for the deletion is just garbage
1021 * collection, and the neighbor is an active default
1022 * router, do not delete it. Instead, reset the GC
1023 * timer using the router's lifetime.
1024 * Simply deleting the entry would affect default
1025 * router selection, which is not necessarily a good
1026 * thing, especially when we're using router preference
1028 * XXX: the check for ln_state would be redundant,
1029 * but we intentionally keep it just in case.
1031 if (dr->expire > time_second)
1032 nd6_llinfo_settimer(ln,
1033 (dr->expire - time_second) * hz);
1035 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
1040 return (LIST_NEXT(ln, lle_next));
1043 if (ln->ln_router || dr) {
1045 * rt6_flush must be called whether or not the neighbor
1046 * is in the Default Router List.
1047 * See a corresponding comment in nd6_na_input().
1049 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1054 * Unreachablity of a router might affect the default
1055 * router selection and on-link detection of advertised
1060 * Temporarily fake the state to choose a new default
1061 * router and to perform on-link determination of
1062 * prefixes correctly.
1063 * Below the state will be set correctly,
1064 * or the entry itself will be deleted.
1066 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1069 * Since defrouter_select() does not affect the
1070 * on-link determination and MIP6 needs the check
1071 * before the default router selection, we perform
1074 pfxlist_onlink_check();
1077 * refresh default router list
1085 * Before deleting the entry, remember the next entry as the
1086 * return value. We need this because pfxlist_onlink_check() above
1087 * might have freed other entries (particularly the old next entry) as
1088 * a side effect (XXX).
1090 next = LIST_NEXT(ln, lle_next);
1092 ifp = ln->lle_tbl->llt_ifp;
1093 IF_AFDATA_LOCK(ifp);
1097 IF_AFDATA_UNLOCK(ifp);
1103 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1105 * XXX cost-effective methods?
1108 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1113 if ((dst6 == NULL) || (rt == NULL))
1117 IF_AFDATA_LOCK(ifp);
1118 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1119 IF_AFDATA_UNLOCK(ifp);
1123 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1127 * if we get upper-layer reachability confirmation many times,
1128 * it is possible we have false information.
1132 if (ln->ln_byhint > V_nd6_maxnudhint) {
1137 ln->ln_state = ND6_LLINFO_REACHABLE;
1138 if (!ND6_LLINFO_PERMANENT(ln)) {
1139 nd6_llinfo_settimer_locked(ln,
1140 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1148 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1150 struct in6_drlist *drl = (struct in6_drlist *)data;
1151 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1152 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1153 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1154 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1155 struct nd_defrouter *dr;
1156 struct nd_prefix *pr;
1157 int i = 0, error = 0;
1161 case SIOCGDRLST_IN6:
1163 * obsolete API, use sysctl under net.inet6.icmp6
1165 bzero(drl, sizeof(*drl));
1167 dr = TAILQ_FIRST(&V_nd_defrouter);
1168 while (dr && i < DRLSTSIZ) {
1169 drl->defrouter[i].rtaddr = dr->rtaddr;
1170 in6_clearscope(&drl->defrouter[i].rtaddr);
1172 drl->defrouter[i].flags = dr->flags;
1173 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1174 drl->defrouter[i].expire = dr->expire;
1175 drl->defrouter[i].if_index = dr->ifp->if_index;
1177 dr = TAILQ_NEXT(dr, dr_entry);
1181 case SIOCGPRLST_IN6:
1183 * obsolete API, use sysctl under net.inet6.icmp6
1185 * XXX the structure in6_prlist was changed in backward-
1186 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1187 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1190 * XXX meaning of fields, especialy "raflags", is very
1191 * differnet between RA prefix list and RR/static prefix list.
1192 * how about separating ioctls into two?
1194 bzero(oprl, sizeof(*oprl));
1196 pr = V_nd_prefix.lh_first;
1197 while (pr && i < PRLSTSIZ) {
1198 struct nd_pfxrouter *pfr;
1201 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1202 oprl->prefix[i].raflags = pr->ndpr_raf;
1203 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1204 oprl->prefix[i].vltime = pr->ndpr_vltime;
1205 oprl->prefix[i].pltime = pr->ndpr_pltime;
1206 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1207 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1208 oprl->prefix[i].expire = 0;
1212 /* XXX: we assume time_t is signed. */
1215 ((sizeof(maxexpire) * 8) - 1));
1216 if (pr->ndpr_vltime <
1217 maxexpire - pr->ndpr_lastupdate) {
1218 oprl->prefix[i].expire =
1219 pr->ndpr_lastupdate +
1222 oprl->prefix[i].expire = maxexpire;
1225 pfr = pr->ndpr_advrtrs.lh_first;
1229 #define RTRADDR oprl->prefix[i].advrtr[j]
1230 RTRADDR = pfr->router->rtaddr;
1231 in6_clearscope(&RTRADDR);
1235 pfr = pfr->pfr_next;
1237 oprl->prefix[i].advrtrs = j;
1238 oprl->prefix[i].origin = PR_ORIG_RA;
1246 case OSIOCGIFINFO_IN6:
1248 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1249 bzero(&ND, sizeof(ND));
1250 ND.linkmtu = IN6_LINKMTU(ifp);
1251 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1252 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1253 ND.reachable = ND_IFINFO(ifp)->reachable;
1254 ND.retrans = ND_IFINFO(ifp)->retrans;
1255 ND.flags = ND_IFINFO(ifp)->flags;
1256 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1257 ND.chlim = ND_IFINFO(ifp)->chlim;
1259 case SIOCGIFINFO_IN6:
1260 ND = *ND_IFINFO(ifp);
1262 case SIOCSIFINFO_IN6:
1264 * used to change host variables from userland.
1265 * intented for a use on router to reflect RA configurations.
1267 /* 0 means 'unspecified' */
1268 if (ND.linkmtu != 0) {
1269 if (ND.linkmtu < IPV6_MMTU ||
1270 ND.linkmtu > IN6_LINKMTU(ifp)) {
1274 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1277 if (ND.basereachable != 0) {
1278 int obasereachable = ND_IFINFO(ifp)->basereachable;
1280 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1281 if (ND.basereachable != obasereachable)
1282 ND_IFINFO(ifp)->reachable =
1283 ND_COMPUTE_RTIME(ND.basereachable);
1285 if (ND.retrans != 0)
1286 ND_IFINFO(ifp)->retrans = ND.retrans;
1288 ND_IFINFO(ifp)->chlim = ND.chlim;
1290 case SIOCSIFINFO_FLAGS:
1293 struct in6_ifaddr *ia;
1295 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1296 !(ND.flags & ND6_IFF_IFDISABLED)) {
1297 /* ifdisabled 1->0 transision */
1300 * If the interface is marked as ND6_IFF_IFDISABLED and
1301 * has an link-local address with IN6_IFF_DUPLICATED,
1302 * do not clear ND6_IFF_IFDISABLED.
1303 * See RFC 4862, Section 5.4.5.
1305 int duplicated_linklocal = 0;
1308 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1309 if (ifa->ifa_addr->sa_family != AF_INET6)
1311 ia = (struct in6_ifaddr *)ifa;
1312 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1313 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1314 duplicated_linklocal = 1;
1318 IF_ADDR_UNLOCK(ifp);
1320 if (duplicated_linklocal) {
1321 ND.flags |= ND6_IFF_IFDISABLED;
1322 log(LOG_ERR, "Cannot enable an interface"
1323 " with a link-local address marked"
1326 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1329 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1330 (ND.flags & ND6_IFF_IFDISABLED)) {
1331 /* ifdisabled 0->1 transision */
1332 /* Mark all IPv6 address as tentative. */
1334 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1336 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1337 if (ifa->ifa_addr->sa_family != AF_INET6)
1339 ia = (struct in6_ifaddr *)ifa;
1340 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1342 IF_ADDR_UNLOCK(ifp);
1345 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) &&
1346 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) {
1347 /* auto_linklocal 0->1 transision */
1349 /* If no link-local address on ifp, configure */
1350 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1351 in6_ifattach(ifp, NULL);
1354 ND_IFINFO(ifp)->flags = ND.flags;
1357 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1358 /* sync kernel routing table with the default router list */
1362 case SIOCSPFXFLUSH_IN6:
1364 /* flush all the prefix advertised by routers */
1365 struct nd_prefix *pr, *next;
1368 for (pr = V_nd_prefix.lh_first; pr; pr = next) {
1369 struct in6_ifaddr *ia, *ia_next;
1371 next = pr->ndpr_next;
1373 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1376 /* do we really have to remove addresses as well? */
1377 /* XXXRW: in6_ifaddrhead locking. */
1378 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1380 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1383 if (ia->ia6_ndpr == pr)
1384 in6_purgeaddr(&ia->ia_ifa);
1391 case SIOCSRTRFLUSH_IN6:
1393 /* flush all the default routers */
1394 struct nd_defrouter *dr, *next;
1398 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) {
1399 next = TAILQ_NEXT(dr, dr_entry);
1406 case SIOCGNBRINFO_IN6:
1409 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1411 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1414 IF_AFDATA_LOCK(ifp);
1415 ln = nd6_lookup(&nb_addr, 0, ifp);
1416 IF_AFDATA_UNLOCK(ifp);
1422 nbi->state = ln->ln_state;
1423 nbi->asked = ln->la_asked;
1424 nbi->isrouter = ln->ln_router;
1425 nbi->expire = ln->la_expire;
1429 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1430 ndif->ifindex = V_nd6_defifindex;
1432 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1433 return (nd6_setdefaultiface(ndif->ifindex));
1439 * Create neighbor cache entry and cache link-layer address,
1440 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1443 * code - type dependent information
1446 * The caller of this function already acquired the ndp
1447 * cache table lock because the cache entry is returned.
1450 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1451 int lladdrlen, int type, int code)
1453 struct llentry *ln = NULL;
1460 uint16_t router = 0;
1461 struct sockaddr_in6 sin6;
1462 struct mbuf *chain = NULL;
1463 int static_route = 0;
1465 IF_AFDATA_UNLOCK_ASSERT(ifp);
1468 panic("ifp == NULL in nd6_cache_lladdr");
1470 panic("from == NULL in nd6_cache_lladdr");
1472 /* nothing must be updated for unspecified address */
1473 if (IN6_IS_ADDR_UNSPECIFIED(from))
1477 * Validation about ifp->if_addrlen and lladdrlen must be done in
1480 * XXX If the link does not have link-layer adderss, what should
1481 * we do? (ifp->if_addrlen == 0)
1482 * Spec says nothing in sections for RA, RS and NA. There's small
1483 * description on it in NS section (RFC 2461 7.2.3).
1485 flags |= lladdr ? ND6_EXCLUSIVE : 0;
1486 IF_AFDATA_LOCK(ifp);
1487 ln = nd6_lookup(from, flags, ifp);
1490 flags |= LLE_EXCLUSIVE;
1491 ln = nd6_lookup(from, flags |ND6_CREATE, ifp);
1492 IF_AFDATA_UNLOCK(ifp);
1495 IF_AFDATA_UNLOCK(ifp);
1496 /* do nothing if static ndp is set */
1497 if (ln->la_flags & LLE_STATIC) {
1506 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1507 if (olladdr && lladdr) {
1508 llchange = bcmp(lladdr, &ln->ll_addr,
1514 * newentry olladdr lladdr llchange (*=record)
1517 * 0 n y -- (3) * STALE
1519 * 0 y y y (5) * STALE
1520 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1521 * 1 -- y -- (7) * STALE
1524 if (lladdr) { /* (3-5) and (7) */
1526 * Record source link-layer address
1527 * XXX is it dependent to ifp->if_type?
1529 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1530 ln->la_flags |= LLE_VALID;
1534 if ((!olladdr && lladdr != NULL) || /* (3) */
1535 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1537 newstate = ND6_LLINFO_STALE;
1538 } else /* (1-2,4) */
1542 if (lladdr == NULL) /* (6) */
1543 newstate = ND6_LLINFO_NOSTATE;
1545 newstate = ND6_LLINFO_STALE;
1550 * Update the state of the neighbor cache.
1552 ln->ln_state = newstate;
1554 if (ln->ln_state == ND6_LLINFO_STALE) {
1556 * XXX: since nd6_output() below will cause
1557 * state tansition to DELAY and reset the timer,
1558 * we must set the timer now, although it is actually
1561 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1564 struct mbuf *m_hold, *m_hold_next;
1567 * reset the la_hold in advance, to explicitly
1568 * prevent a la_hold lookup in nd6_output()
1569 * (wouldn't happen, though...)
1571 for (m_hold = ln->la_hold, ln->la_hold = NULL;
1572 m_hold; m_hold = m_hold_next) {
1573 m_hold_next = m_hold->m_nextpkt;
1574 m_hold->m_nextpkt = NULL;
1577 * we assume ifp is not a p2p here, so
1578 * just set the 2nd argument as the
1581 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
1584 * If we have mbufs in the chain we need to do
1585 * deferred transmit. Copy the address from the
1586 * llentry before dropping the lock down below.
1589 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
1591 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1592 /* probe right away */
1593 nd6_llinfo_settimer_locked((void *)ln, 0);
1598 * ICMP6 type dependent behavior.
1600 * NS: clear IsRouter if new entry
1601 * RS: clear IsRouter
1602 * RA: set IsRouter if there's lladdr
1603 * redir: clear IsRouter if new entry
1606 * The spec says that we must set IsRouter in the following cases:
1607 * - If lladdr exist, set IsRouter. This means (1-5).
1608 * - If it is old entry (!newentry), set IsRouter. This means (7).
1609 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1610 * A quetion arises for (1) case. (1) case has no lladdr in the
1611 * neighbor cache, this is similar to (6).
1612 * This case is rare but we figured that we MUST NOT set IsRouter.
1614 * newentry olladdr lladdr llchange NS RS RA redir
1616 * 0 n n -- (1) c ? s
1617 * 0 y n -- (2) c s s
1618 * 0 n y -- (3) c s s
1621 * 1 -- n -- (6) c c c s
1622 * 1 -- y -- (7) c c s c s
1626 switch (type & 0xff) {
1627 case ND_NEIGHBOR_SOLICIT:
1629 * New entry must have is_router flag cleared.
1631 if (is_newentry) /* (6-7) */
1636 * If the icmp is a redirect to a better router, always set the
1637 * is_router flag. Otherwise, if the entry is newly created,
1638 * clear the flag. [RFC 2461, sec 8.3]
1640 if (code == ND_REDIRECT_ROUTER)
1642 else if (is_newentry) /* (6-7) */
1645 case ND_ROUTER_SOLICIT:
1647 * is_router flag must always be cleared.
1651 case ND_ROUTER_ADVERT:
1653 * Mark an entry with lladdr as a router.
1655 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1656 (is_newentry && lladdr)) { /* (7) */
1663 static_route = (ln->la_flags & LLE_STATIC);
1664 router = ln->ln_router;
1666 if (flags & ND6_EXCLUSIVE)
1674 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
1677 * When the link-layer address of a router changes, select the
1678 * best router again. In particular, when the neighbor entry is newly
1679 * created, it might affect the selection policy.
1680 * Question: can we restrict the first condition to the "is_newentry"
1682 * XXX: when we hear an RA from a new router with the link-layer
1683 * address option, defrouter_select() is called twice, since
1684 * defrtrlist_update called the function as well. However, I believe
1685 * we can compromise the overhead, since it only happens the first
1687 * XXX: although defrouter_select() should not have a bad effect
1688 * for those are not autoconfigured hosts, we explicitly avoid such
1691 if (do_update && router && !V_ip6_forwarding &&
1692 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1694 * guaranteed recursion
1702 if (flags & ND6_EXCLUSIVE)
1713 nd6_slowtimo(void *arg)
1715 CURVNET_SET((struct vnet *) arg);
1716 struct nd_ifinfo *nd6if;
1719 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1720 nd6_slowtimo, curvnet);
1721 IFNET_RLOCK_NOSLEEP();
1722 for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
1723 ifp = TAILQ_NEXT(ifp, if_list)) {
1724 nd6if = ND_IFINFO(ifp);
1725 if (nd6if->basereachable && /* already initialized */
1726 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1728 * Since reachable time rarely changes by router
1729 * advertisements, we SHOULD insure that a new random
1730 * value gets recomputed at least once every few hours.
1733 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1734 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1737 IFNET_RUNLOCK_NOSLEEP();
1742 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1743 struct sockaddr_in6 *dst, struct rtentry *rt0)
1746 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
1751 * Note that I'm not enforcing any global serialization
1752 * lle state or asked changes here as the logic is too
1753 * complicated to avoid having to always acquire an exclusive
1758 #define senderr(e) { error = (e); goto bad;}
1761 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1762 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
1763 struct mbuf **chain)
1765 struct mbuf *m = m0;
1766 struct llentry *ln = lle;
1773 LLE_WLOCK_ASSERT(lle);
1775 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
1778 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1781 if (nd6_need_cache(ifp) == 0)
1785 * next hop determination. This routine is derived from ether_output.
1789 * Address resolution or Neighbor Unreachability Detection
1791 * At this point, the destination of the packet must be a unicast
1792 * or an anycast address(i.e. not a multicast).
1795 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0;
1798 IF_AFDATA_LOCK(ifp);
1799 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
1800 IF_AFDATA_UNLOCK(ifp);
1801 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1803 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1804 * the condition below is not very efficient. But we believe
1805 * it is tolerable, because this should be a rare case.
1807 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
1808 IF_AFDATA_LOCK(ifp);
1809 ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
1810 IF_AFDATA_UNLOCK(ifp);
1814 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1815 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1816 char ip6buf[INET6_ADDRSTRLEN];
1818 "nd6_output: can't allocate llinfo for %s "
1820 ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
1821 senderr(EIO); /* XXX: good error? */
1823 goto sendpkt; /* send anyway */
1826 /* We don't have to do link-layer address resolution on a p2p link. */
1827 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1828 ln->ln_state < ND6_LLINFO_REACHABLE) {
1829 if ((flags & LLE_EXCLUSIVE) == 0) {
1830 flags |= LLE_EXCLUSIVE;
1833 ln->ln_state = ND6_LLINFO_STALE;
1834 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1838 * The first time we send a packet to a neighbor whose entry is
1839 * STALE, we have to change the state to DELAY and a sets a timer to
1840 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1841 * neighbor unreachability detection on expiration.
1844 if (ln->ln_state == ND6_LLINFO_STALE) {
1845 if ((flags & LLE_EXCLUSIVE) == 0) {
1846 flags |= LLE_EXCLUSIVE;
1851 ln->ln_state = ND6_LLINFO_DELAY;
1852 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1856 * If the neighbor cache entry has a state other than INCOMPLETE
1857 * (i.e. its link-layer address is already resolved), just
1860 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1864 * There is a neighbor cache entry, but no ethernet address
1865 * response yet. Append this latest packet to the end of the
1866 * packet queue in the mbuf, unless the number of the packet
1867 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1868 * the oldest packet in the queue will be removed.
1870 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1871 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1873 if ((flags & LLE_EXCLUSIVE) == 0) {
1874 flags |= LLE_EXCLUSIVE;
1879 struct mbuf *m_hold;
1883 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
1885 if (m_hold->m_nextpkt == NULL) {
1886 m_hold->m_nextpkt = m;
1890 while (i >= V_nd6_maxqueuelen) {
1891 m_hold = ln->la_hold;
1892 ln->la_hold = ln->la_hold->m_nextpkt;
1900 * We did the lookup (no lle arg) so we
1901 * need to do the unlock here
1904 if (flags & LLE_EXCLUSIVE)
1911 * If there has been no NS for the neighbor after entering the
1912 * INCOMPLETE state, send the first solicitation.
1914 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
1917 nd6_llinfo_settimer(ln,
1918 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
1919 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1924 /* discard the packet if IPv6 operation is disabled on the interface */
1925 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1926 error = ENETDOWN; /* better error? */
1930 * ln is valid and the caller did not pass in
1933 if ((ln != NULL) && (lle == NULL)) {
1934 if (flags & LLE_EXCLUSIVE)
1941 mac_netinet6_nd6_send(ifp, m);
1944 * We were passed in a pointer to an lle with the lock held
1945 * this means that we can't call if_output as we will
1946 * recurse on the lle lock - so what we do is we create
1947 * a list of mbufs to send and transmit them in the caller
1948 * after the lock is dropped
1954 struct mbuf *m = *chain;
1957 * append mbuf to end of deferred chain
1959 while (m->m_nextpkt != NULL)
1965 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1966 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
1969 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
1974 * ln is valid and the caller did not pass in
1977 if ((ln != NULL) && (lle == NULL)) {
1978 if (flags & LLE_EXCLUSIVE)
1991 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
1992 struct sockaddr_in6 *dst, struct route *ro)
1994 struct mbuf *m, *m_head;
1995 struct ifnet *outifp;
1999 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2006 m_head = m_head->m_nextpkt;
2007 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);
2012 * note that intermediate errors are blindly ignored - but this is
2013 * the same convention as used with nd6_output when called by
2021 nd6_need_cache(struct ifnet *ifp)
2024 * XXX: we currently do not make neighbor cache on any interface
2025 * other than ARCnet, Ethernet, FDDI and GIF.
2028 * - unidirectional tunnels needs no ND
2030 switch (ifp->if_type) {
2038 #ifdef IFT_IEEE80211
2044 case IFT_GIF: /* XXX need more cases? */
2048 case IFT_PROPVIRTUAL:
2056 * the callers of this function need to be re-worked to drop
2057 * the lle lock, drop here for now
2060 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2061 struct sockaddr *dst, u_char *desten, struct llentry **lle)
2066 IF_AFDATA_UNLOCK_ASSERT(ifp);
2067 if (m->m_flags & M_MCAST) {
2070 switch (ifp->if_type) {
2076 #ifdef IFT_IEEE80211
2081 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2086 * netbsd can use if_broadcastaddr, but we don't do so
2087 * to reduce # of ifdef.
2089 for (i = 0; i < ifp->if_addrlen; i++)
2097 return (EAFNOSUPPORT);
2103 * the entry should have been created in nd6_store_lladdr
2105 IF_AFDATA_LOCK(ifp);
2106 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2107 IF_AFDATA_UNLOCK(ifp);
2108 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2111 /* this could happen, if we could not allocate memory */
2116 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2120 * A *small* use after free race exists here
2126 clear_llinfo_pqueue(struct llentry *ln)
2128 struct mbuf *m_hold, *m_hold_next;
2130 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2131 m_hold_next = m_hold->m_nextpkt;
2132 m_hold->m_nextpkt = NULL;
2140 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2141 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2143 SYSCTL_DECL(_net_inet6_icmp6);
2145 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2146 CTLFLAG_RD, nd6_sysctl_drlist, "");
2147 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2148 CTLFLAG_RD, nd6_sysctl_prlist, "");
2149 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2150 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2153 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2156 char buf[1024] __aligned(4);
2157 struct in6_defrouter *d, *de;
2158 struct nd_defrouter *dr;
2164 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
2165 dr = TAILQ_NEXT(dr, dr_entry)) {
2166 d = (struct in6_defrouter *)buf;
2167 de = (struct in6_defrouter *)(buf + sizeof(buf));
2170 bzero(d, sizeof(*d));
2171 d->rtaddr.sin6_family = AF_INET6;
2172 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2173 d->rtaddr.sin6_addr = dr->rtaddr;
2174 error = sa6_recoverscope(&d->rtaddr);
2177 d->flags = dr->flags;
2178 d->rtlifetime = dr->rtlifetime;
2179 d->expire = dr->expire;
2180 d->if_index = dr->ifp->if_index;
2182 panic("buffer too short");
2184 error = SYSCTL_OUT(req, buf, sizeof(*d));
2193 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2196 char buf[1024] __aligned(4);
2197 struct in6_prefix *p, *pe;
2198 struct nd_prefix *pr;
2199 char ip6buf[INET6_ADDRSTRLEN];
2205 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2208 struct sockaddr_in6 *sin6, *s6;
2209 struct nd_pfxrouter *pfr;
2211 p = (struct in6_prefix *)buf;
2212 pe = (struct in6_prefix *)(buf + sizeof(buf));
2215 bzero(p, sizeof(*p));
2216 sin6 = (struct sockaddr_in6 *)(p + 1);
2218 p->prefix = pr->ndpr_prefix;
2219 if (sa6_recoverscope(&p->prefix)) {
2221 "scope error in prefix list (%s)\n",
2222 ip6_sprintf(ip6buf, &p->prefix.sin6_addr));
2223 /* XXX: press on... */
2225 p->raflags = pr->ndpr_raf;
2226 p->prefixlen = pr->ndpr_plen;
2227 p->vltime = pr->ndpr_vltime;
2228 p->pltime = pr->ndpr_pltime;
2229 p->if_index = pr->ndpr_ifp->if_index;
2230 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2235 /* XXX: we assume time_t is signed. */
2238 ((sizeof(maxexpire) * 8) - 1));
2239 if (pr->ndpr_vltime <
2240 maxexpire - pr->ndpr_lastupdate) {
2241 p->expire = pr->ndpr_lastupdate +
2244 p->expire = maxexpire;
2246 p->refcnt = pr->ndpr_refcnt;
2247 p->flags = pr->ndpr_stateflags;
2248 p->origin = PR_ORIG_RA;
2250 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2251 pfr = pfr->pfr_next) {
2252 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2256 s6 = &sin6[advrtrs];
2257 bzero(s6, sizeof(*s6));
2258 s6->sin6_family = AF_INET6;
2259 s6->sin6_len = sizeof(*sin6);
2260 s6->sin6_addr = pfr->router->rtaddr;
2261 if (sa6_recoverscope(s6)) {
2264 "prefix list (%s)\n",
2266 &pfr->router->rtaddr));
2270 p->advrtrs = advrtrs;
2272 panic("buffer too short");
2274 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2275 error = SYSCTL_OUT(req, buf, advance);