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 <netinet/icmp6.h>
75 #include <sys/limits.h>
77 #include <security/mac/mac_framework.h>
79 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
80 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
82 #define SIN6(s) ((struct sockaddr_in6 *)s)
85 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
86 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
87 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
88 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
89 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
91 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
94 /* preventing too many loops in ND option parsing */
95 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
97 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
99 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
101 #define V_nd6_maxndopt VNET(nd6_maxndopt)
102 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
105 VNET_DEFINE(int, nd6_debug) = 1;
107 VNET_DEFINE(int, nd6_debug) = 0;
112 static int nd6_inuse, nd6_allocated;
115 VNET_DEFINE(struct nd_drhead, nd_defrouter);
116 VNET_DEFINE(struct nd_prhead, nd_prefix);
118 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
119 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
121 static struct sockaddr_in6 all1_sa;
123 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
125 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
126 static void nd6_slowtimo(void *);
127 static int regen_tmpaddr(struct in6_ifaddr *);
128 static struct llentry *nd6_free(struct llentry *, int);
129 static void nd6_llinfo_timer(void *);
130 static void clear_llinfo_pqueue(struct llentry *);
132 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
133 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
135 VNET_DEFINE(struct callout, nd6_timer_ch);
142 LIST_INIT(&V_nd_prefix);
144 all1_sa.sin6_family = AF_INET6;
145 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
146 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
147 all1_sa.sin6_addr.s6_addr[i] = 0xff;
149 /* initialization of the default router list */
150 TAILQ_INIT(&V_nd_defrouter);
153 callout_init(&V_nd6_slowtimo_ch, 0);
154 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
155 nd6_slowtimo, curvnet);
163 callout_drain(&V_nd6_slowtimo_ch);
164 callout_drain(&V_nd6_timer_ch);
169 nd6_ifattach(struct ifnet *ifp)
171 struct nd_ifinfo *nd;
173 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
174 bzero(nd, sizeof(*nd));
178 nd->chlim = IPV6_DEFHLIM;
179 nd->basereachable = REACHABLE_TIME;
180 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
181 nd->retrans = RETRANS_TIMER;
183 * Note that the default value of ip6_accept_rtadv is 0, which means
184 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
187 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
189 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
190 nd6_setmtu0(ifp, nd);
196 nd6_ifdetach(struct nd_ifinfo *nd)
203 * Reset ND level link MTU. This function is called when the physical MTU
204 * changes, which means we might have to adjust the ND level MTU.
207 nd6_setmtu(struct ifnet *ifp)
210 nd6_setmtu0(ifp, ND_IFINFO(ifp));
213 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
215 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
219 omaxmtu = ndi->maxmtu;
221 switch (ifp->if_type) {
223 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
226 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
229 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
232 ndi->maxmtu = ifp->if_mtu;
237 * Decreasing the interface MTU under IPV6 minimum MTU may cause
238 * undesirable situation. We thus notify the operator of the change
239 * explicitly. The check for omaxmtu is necessary to restrict the
240 * log to the case of changing the MTU, not initializing it.
242 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
243 log(LOG_NOTICE, "nd6_setmtu0: "
244 "new link MTU on %s (%lu) is too small for IPv6\n",
245 if_name(ifp), (unsigned long)ndi->maxmtu);
248 if (ndi->maxmtu > V_in6_maxmtu)
249 in6_setmaxmtu(); /* check all interfaces just in case */
254 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
257 bzero(ndopts, sizeof(*ndopts));
258 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
260 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
263 ndopts->nd_opts_done = 1;
264 ndopts->nd_opts_search = NULL;
269 * Take one ND option.
272 nd6_option(union nd_opts *ndopts)
274 struct nd_opt_hdr *nd_opt;
278 panic("ndopts == NULL in nd6_option");
279 if (ndopts->nd_opts_last == NULL)
280 panic("uninitialized ndopts in nd6_option");
281 if (ndopts->nd_opts_search == NULL)
283 if (ndopts->nd_opts_done)
286 nd_opt = ndopts->nd_opts_search;
288 /* make sure nd_opt_len is inside the buffer */
289 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
290 bzero(ndopts, sizeof(*ndopts));
294 olen = nd_opt->nd_opt_len << 3;
297 * Message validation requires that all included
298 * options have a length that is greater than zero.
300 bzero(ndopts, sizeof(*ndopts));
304 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
305 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
306 /* option overruns the end of buffer, invalid */
307 bzero(ndopts, sizeof(*ndopts));
309 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
310 /* reached the end of options chain */
311 ndopts->nd_opts_done = 1;
312 ndopts->nd_opts_search = NULL;
318 * Parse multiple ND options.
319 * This function is much easier to use, for ND routines that do not need
320 * multiple options of the same type.
323 nd6_options(union nd_opts *ndopts)
325 struct nd_opt_hdr *nd_opt;
329 panic("ndopts == NULL in nd6_options");
330 if (ndopts->nd_opts_last == NULL)
331 panic("uninitialized ndopts in nd6_options");
332 if (ndopts->nd_opts_search == NULL)
336 nd_opt = nd6_option(ndopts);
337 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
339 * Message validation requires that all included
340 * options have a length that is greater than zero.
342 ICMP6STAT_INC(icp6s_nd_badopt);
343 bzero(ndopts, sizeof(*ndopts));
350 switch (nd_opt->nd_opt_type) {
351 case ND_OPT_SOURCE_LINKADDR:
352 case ND_OPT_TARGET_LINKADDR:
354 case ND_OPT_REDIRECTED_HEADER:
355 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
357 "duplicated ND6 option found (type=%d)\n",
358 nd_opt->nd_opt_type));
361 ndopts->nd_opt_array[nd_opt->nd_opt_type]
365 case ND_OPT_PREFIX_INFORMATION:
366 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
367 ndopts->nd_opt_array[nd_opt->nd_opt_type]
370 ndopts->nd_opts_pi_end =
371 (struct nd_opt_prefix_info *)nd_opt;
375 * Unknown options must be silently ignored,
376 * to accomodate future extension to the protocol.
379 "nd6_options: unsupported option %d - "
380 "option ignored\n", nd_opt->nd_opt_type));
385 if (i > V_nd6_maxndopt) {
386 ICMP6STAT_INC(icp6s_nd_toomanyopt);
387 nd6log((LOG_INFO, "too many loop in nd opt\n"));
391 if (ndopts->nd_opts_done)
399 * ND6 timer routine to handle ND6 entries
402 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
409 canceled = callout_stop(&ln->ln_timer_ch);
411 ln->la_expire = time_second + tick / hz;
413 if (tick > INT_MAX) {
414 ln->ln_ntick = tick - INT_MAX;
415 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
416 nd6_llinfo_timer, ln);
419 canceled = callout_reset(&ln->ln_timer_ch, tick,
420 nd6_llinfo_timer, ln);
428 nd6_llinfo_settimer(struct llentry *ln, long tick)
432 nd6_llinfo_settimer_locked(ln, tick);
437 nd6_llinfo_timer(void *arg)
440 struct in6_addr *dst;
442 struct nd_ifinfo *ndi = NULL;
444 ln = (struct llentry *)arg;
446 panic("%s: NULL entry!\n", __func__);
450 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL)
451 panic("ln ifp == NULL");
453 CURVNET_SET(ifp->if_vnet);
455 if (ln->ln_ntick > 0) {
456 if (ln->ln_ntick > INT_MAX) {
457 ln->ln_ntick -= INT_MAX;
458 nd6_llinfo_settimer(ln, INT_MAX);
461 nd6_llinfo_settimer(ln, ln->ln_ntick);
466 ndi = ND_IFINFO(ifp);
467 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
468 if (ln->la_flags & LLE_STATIC) {
472 if (ln->la_flags & LLE_DELETED) {
473 (void)nd6_free(ln, 0);
478 switch (ln->ln_state) {
479 case ND6_LLINFO_INCOMPLETE:
480 if (ln->la_asked < V_nd6_mmaxtries) {
482 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
483 nd6_ns_output(ifp, NULL, dst, ln, 0);
485 struct mbuf *m = ln->la_hold;
490 * assuming every packet in la_hold has the
495 icmp6_error2(m, ICMP6_DST_UNREACH,
496 ICMP6_DST_UNREACH_ADDR, 0, ifp);
499 clear_llinfo_pqueue(ln);
501 (void)nd6_free(ln, 0);
505 case ND6_LLINFO_REACHABLE:
506 if (!ND6_LLINFO_PERMANENT(ln)) {
507 ln->ln_state = ND6_LLINFO_STALE;
508 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
512 case ND6_LLINFO_STALE:
513 /* Garbage Collection(RFC 2461 5.3) */
514 if (!ND6_LLINFO_PERMANENT(ln)) {
515 (void)nd6_free(ln, 1);
520 case ND6_LLINFO_DELAY:
521 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
524 ln->ln_state = ND6_LLINFO_PROBE;
525 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
526 nd6_ns_output(ifp, dst, dst, ln, 0);
528 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
529 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
532 case ND6_LLINFO_PROBE:
533 if (ln->la_asked < V_nd6_umaxtries) {
535 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
536 nd6_ns_output(ifp, dst, dst, ln, 0);
538 (void)nd6_free(ln, 0);
551 * ND6 timer routine to expire default route list and prefix list
556 CURVNET_SET((struct vnet *) arg);
558 struct nd_defrouter *dr;
559 struct nd_prefix *pr;
560 struct in6_ifaddr *ia6, *nia6;
561 struct in6_addrlifetime *lt6;
563 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
566 /* expire default router list */
568 dr = TAILQ_FIRST(&V_nd_defrouter);
570 if (dr->expire && dr->expire < time_second) {
571 struct nd_defrouter *t;
572 t = TAILQ_NEXT(dr, dr_entry);
576 dr = TAILQ_NEXT(dr, dr_entry);
581 * expire interface addresses.
582 * in the past the loop was inside prefix expiry processing.
583 * However, from a stricter speci-confrmance standpoint, we should
584 * rather separate address lifetimes and prefix lifetimes.
586 * XXXRW: in6_ifaddrhead locking.
589 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
590 /* check address lifetime */
591 lt6 = &ia6->ia6_lifetime;
592 if (IFA6_IS_INVALID(ia6)) {
596 * If the expiring address is temporary, try
597 * regenerating a new one. This would be useful when
598 * we suspended a laptop PC, then turned it on after a
599 * period that could invalidate all temporary
600 * addresses. Although we may have to restart the
601 * loop (see below), it must be after purging the
602 * address. Otherwise, we'd see an infinite loop of
605 if (V_ip6_use_tempaddr &&
606 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
607 if (regen_tmpaddr(ia6) == 0)
611 in6_purgeaddr(&ia6->ia_ifa);
614 goto addrloop; /* XXX: see below */
615 } else if (IFA6_IS_DEPRECATED(ia6)) {
616 int oldflags = ia6->ia6_flags;
618 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
621 * If a temporary address has just become deprecated,
622 * regenerate a new one if possible.
624 if (V_ip6_use_tempaddr &&
625 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
626 (oldflags & IN6_IFF_DEPRECATED) == 0) {
628 if (regen_tmpaddr(ia6) == 0) {
630 * A new temporary address is
632 * XXX: this means the address chain
633 * has changed while we are still in
634 * the loop. Although the change
635 * would not cause disaster (because
636 * it's not a deletion, but an
637 * addition,) we'd rather restart the
638 * loop just for safety. Or does this
639 * significantly reduce performance??
646 * A new RA might have made a deprecated address
649 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
653 /* expire prefix list */
654 pr = V_nd_prefix.lh_first;
657 * check prefix lifetime.
658 * since pltime is just for autoconf, pltime processing for
659 * prefix is not necessary.
661 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
662 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
667 * address expiration and prefix expiration are
668 * separate. NEVER perform in6_purgeaddr here.
681 * ia6 - deprecated/invalidated temporary address
684 regen_tmpaddr(struct in6_ifaddr *ia6)
688 struct in6_ifaddr *public_ifa6 = NULL;
690 ifp = ia6->ia_ifa.ifa_ifp;
692 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
693 struct in6_ifaddr *it6;
695 if (ifa->ifa_addr->sa_family != AF_INET6)
698 it6 = (struct in6_ifaddr *)ifa;
700 /* ignore no autoconf addresses. */
701 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
704 /* ignore autoconf addresses with different prefixes. */
705 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
709 * Now we are looking at an autoconf address with the same
710 * prefix as ours. If the address is temporary and is still
711 * preferred, do not create another one. It would be rare, but
712 * could happen, for example, when we resume a laptop PC after
715 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
716 !IFA6_IS_DEPRECATED(it6)) {
722 * This is a public autoconf address that has the same prefix
723 * as ours. If it is preferred, keep it. We can't break the
724 * loop here, because there may be a still-preferred temporary
725 * address with the prefix.
727 if (!IFA6_IS_DEPRECATED(it6))
730 if (public_ifa6 != NULL)
731 ifa_ref(&public_ifa6->ia_ifa);
735 if (public_ifa6 != NULL) {
738 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
739 ifa_free(&public_ifa6->ia_ifa);
740 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
741 " tmp addr,errno=%d\n", e);
744 ifa_free(&public_ifa6->ia_ifa);
752 * Nuke neighbor cache/prefix/default router management table, right before
756 nd6_purge(struct ifnet *ifp)
758 struct nd_defrouter *dr, *ndr;
759 struct nd_prefix *pr, *npr;
762 * Nuke default router list entries toward ifp.
763 * We defer removal of default router list entries that is installed
764 * in the routing table, in order to keep additional side effects as
767 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
768 ndr = TAILQ_NEXT(dr, dr_entry);
776 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
777 ndr = TAILQ_NEXT(dr, dr_entry);
785 /* Nuke prefix list entries toward ifp */
786 for (pr = V_nd_prefix.lh_first; pr; pr = npr) {
788 if (pr->ndpr_ifp == ifp) {
790 * Because if_detach() does *not* release prefixes
791 * while purging addresses the reference count will
792 * still be above zero. We therefore reset it to
793 * make sure that the prefix really gets purged.
798 * Previously, pr->ndpr_addr is removed as well,
799 * but I strongly believe we don't have to do it.
800 * nd6_purge() is only called from in6_ifdetach(),
801 * which removes all the associated interface addresses
803 * (jinmei@kame.net 20010129)
809 /* cancel default outgoing interface setting */
810 if (V_nd6_defifindex == ifp->if_index)
811 nd6_setdefaultiface(0);
813 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */
814 /* refresh default router list
823 * We do not nuke the neighbor cache entries here any more
824 * because the neighbor cache is kept in if_afdata[AF_INET6].
825 * nd6_purge() is invoked by in6_ifdetach() which is called
826 * from if_detach() where everything gets purged. So let
827 * in6_domifdetach() do the actual L2 table purging work.
832 * the caller acquires and releases the lock on the lltbls
833 * Returns the llentry locked
836 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
838 struct sockaddr_in6 sin6;
842 bzero(&sin6, sizeof(sin6));
843 sin6.sin6_len = sizeof(struct sockaddr_in6);
844 sin6.sin6_family = AF_INET6;
845 sin6.sin6_addr = *addr6;
847 IF_AFDATA_LOCK_ASSERT(ifp);
849 if (flags & ND6_CREATE)
850 llflags |= LLE_CREATE;
851 if (flags & ND6_EXCLUSIVE)
852 llflags |= LLE_EXCLUSIVE;
854 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
855 if ((ln != NULL) && (flags & LLE_CREATE)) {
856 ln->ln_state = ND6_LLINFO_NOSTATE;
857 callout_init(&ln->ln_timer_ch, 0);
864 * Test whether a given IPv6 address is a neighbor or not, ignoring
865 * the actual neighbor cache. The neighbor cache is ignored in order
866 * to not reenter the routing code from within itself.
869 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
871 struct nd_prefix *pr;
872 struct ifaddr *dstaddr;
875 * A link-local address is always a neighbor.
876 * XXX: a link does not necessarily specify a single interface.
878 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
879 struct sockaddr_in6 sin6_copy;
883 * We need sin6_copy since sa6_recoverscope() may modify the
887 if (sa6_recoverscope(&sin6_copy))
888 return (0); /* XXX: should be impossible */
889 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
891 if (sin6_copy.sin6_scope_id == zone)
898 * If the address matches one of our addresses,
899 * it should be a neighbor.
900 * If the address matches one of our on-link prefixes, it should be a
903 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
904 if (pr->ndpr_ifp != ifp)
907 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
909 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0);
913 * This is the case where multiple interfaces
914 * have the same prefix, but only one is installed
915 * into the routing table and that prefix entry
916 * is not the one being examined here. In the case
917 * where RADIX_MPATH is enabled, multiple route
918 * entries (of the same rt_key value) will be
919 * installed because the interface addresses all
922 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
923 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
930 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
931 &addr->sin6_addr, &pr->ndpr_mask))
936 * If the address is assigned on the node of the other side of
937 * a p2p interface, the address should be a neighbor.
939 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
940 if (dstaddr != NULL) {
941 if (dstaddr->ifa_ifp == ifp) {
949 * If the default router list is empty, all addresses are regarded
950 * as on-link, and thus, as a neighbor.
951 * XXX: we restrict the condition to hosts, because routers usually do
952 * not have the "default router list".
954 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL &&
955 V_nd6_defifindex == ifp->if_index) {
964 * Detect if a given IPv6 address identifies a neighbor on a given link.
965 * XXX: should take care of the destination of a p2p link?
968 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
973 IF_AFDATA_UNLOCK_ASSERT(ifp);
974 if (nd6_is_new_addr_neighbor(addr, ifp))
978 * Even if the address matches none of our addresses, it might be
979 * in the neighbor cache.
982 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
986 IF_AFDATA_UNLOCK(ifp);
991 * Free an nd6 llinfo entry.
992 * Since the function would cause significant changes in the kernel, DO NOT
993 * make it global, unless you have a strong reason for the change, and are sure
994 * that the change is safe.
996 static struct llentry *
997 nd6_free(struct llentry *ln, int gc)
999 struct llentry *next;
1000 struct nd_defrouter *dr;
1001 struct ifnet *ifp=NULL;
1004 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1005 * even though it is not harmful, it was not really necessary.
1009 nd6_llinfo_settimer(ln, -1);
1011 if (!V_ip6_forwarding) {
1014 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1016 if (dr != NULL && dr->expire &&
1017 ln->ln_state == ND6_LLINFO_STALE && gc) {
1019 * If the reason for the deletion is just garbage
1020 * collection, and the neighbor is an active default
1021 * router, do not delete it. Instead, reset the GC
1022 * timer using the router's lifetime.
1023 * Simply deleting the entry would affect default
1024 * router selection, which is not necessarily a good
1025 * thing, especially when we're using router preference
1027 * XXX: the check for ln_state would be redundant,
1028 * but we intentionally keep it just in case.
1030 if (dr->expire > time_second)
1031 nd6_llinfo_settimer(ln,
1032 (dr->expire - time_second) * hz);
1034 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
1039 return (LIST_NEXT(ln, lle_next));
1042 if (ln->ln_router || dr) {
1044 * rt6_flush must be called whether or not the neighbor
1045 * is in the Default Router List.
1046 * See a corresponding comment in nd6_na_input().
1048 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1053 * Unreachablity of a router might affect the default
1054 * router selection and on-link detection of advertised
1059 * Temporarily fake the state to choose a new default
1060 * router and to perform on-link determination of
1061 * prefixes correctly.
1062 * Below the state will be set correctly,
1063 * or the entry itself will be deleted.
1065 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1068 * Since defrouter_select() does not affect the
1069 * on-link determination and MIP6 needs the check
1070 * before the default router selection, we perform
1073 pfxlist_onlink_check();
1076 * refresh default router list
1084 * Before deleting the entry, remember the next entry as the
1085 * return value. We need this because pfxlist_onlink_check() above
1086 * might have freed other entries (particularly the old next entry) as
1087 * a side effect (XXX).
1089 next = LIST_NEXT(ln, lle_next);
1091 ifp = ln->lle_tbl->llt_ifp;
1092 IF_AFDATA_LOCK(ifp);
1096 IF_AFDATA_UNLOCK(ifp);
1102 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1104 * XXX cost-effective methods?
1107 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1112 if ((dst6 == NULL) || (rt == NULL))
1116 IF_AFDATA_LOCK(ifp);
1117 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1118 IF_AFDATA_UNLOCK(ifp);
1122 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1126 * if we get upper-layer reachability confirmation many times,
1127 * it is possible we have false information.
1131 if (ln->ln_byhint > V_nd6_maxnudhint) {
1136 ln->ln_state = ND6_LLINFO_REACHABLE;
1137 if (!ND6_LLINFO_PERMANENT(ln)) {
1138 nd6_llinfo_settimer_locked(ln,
1139 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1147 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1149 struct in6_drlist *drl = (struct in6_drlist *)data;
1150 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1151 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1152 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1153 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1154 struct nd_defrouter *dr;
1155 struct nd_prefix *pr;
1156 int i = 0, error = 0;
1160 case SIOCGDRLST_IN6:
1162 * obsolete API, use sysctl under net.inet6.icmp6
1164 bzero(drl, sizeof(*drl));
1166 dr = TAILQ_FIRST(&V_nd_defrouter);
1167 while (dr && i < DRLSTSIZ) {
1168 drl->defrouter[i].rtaddr = dr->rtaddr;
1169 in6_clearscope(&drl->defrouter[i].rtaddr);
1171 drl->defrouter[i].flags = dr->flags;
1172 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1173 drl->defrouter[i].expire = dr->expire;
1174 drl->defrouter[i].if_index = dr->ifp->if_index;
1176 dr = TAILQ_NEXT(dr, dr_entry);
1180 case SIOCGPRLST_IN6:
1182 * obsolete API, use sysctl under net.inet6.icmp6
1184 * XXX the structure in6_prlist was changed in backward-
1185 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1186 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1189 * XXX meaning of fields, especialy "raflags", is very
1190 * differnet between RA prefix list and RR/static prefix list.
1191 * how about separating ioctls into two?
1193 bzero(oprl, sizeof(*oprl));
1195 pr = V_nd_prefix.lh_first;
1196 while (pr && i < PRLSTSIZ) {
1197 struct nd_pfxrouter *pfr;
1200 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1201 oprl->prefix[i].raflags = pr->ndpr_raf;
1202 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1203 oprl->prefix[i].vltime = pr->ndpr_vltime;
1204 oprl->prefix[i].pltime = pr->ndpr_pltime;
1205 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1206 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1207 oprl->prefix[i].expire = 0;
1211 /* XXX: we assume time_t is signed. */
1214 ((sizeof(maxexpire) * 8) - 1));
1215 if (pr->ndpr_vltime <
1216 maxexpire - pr->ndpr_lastupdate) {
1217 oprl->prefix[i].expire =
1218 pr->ndpr_lastupdate +
1221 oprl->prefix[i].expire = maxexpire;
1224 pfr = pr->ndpr_advrtrs.lh_first;
1228 #define RTRADDR oprl->prefix[i].advrtr[j]
1229 RTRADDR = pfr->router->rtaddr;
1230 in6_clearscope(&RTRADDR);
1234 pfr = pfr->pfr_next;
1236 oprl->prefix[i].advrtrs = j;
1237 oprl->prefix[i].origin = PR_ORIG_RA;
1245 case OSIOCGIFINFO_IN6:
1247 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1248 bzero(&ND, sizeof(ND));
1249 ND.linkmtu = IN6_LINKMTU(ifp);
1250 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1251 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1252 ND.reachable = ND_IFINFO(ifp)->reachable;
1253 ND.retrans = ND_IFINFO(ifp)->retrans;
1254 ND.flags = ND_IFINFO(ifp)->flags;
1255 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1256 ND.chlim = ND_IFINFO(ifp)->chlim;
1258 case SIOCGIFINFO_IN6:
1259 ND = *ND_IFINFO(ifp);
1261 case SIOCSIFINFO_IN6:
1263 * used to change host variables from userland.
1264 * intented for a use on router to reflect RA configurations.
1266 /* 0 means 'unspecified' */
1267 if (ND.linkmtu != 0) {
1268 if (ND.linkmtu < IPV6_MMTU ||
1269 ND.linkmtu > IN6_LINKMTU(ifp)) {
1273 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1276 if (ND.basereachable != 0) {
1277 int obasereachable = ND_IFINFO(ifp)->basereachable;
1279 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1280 if (ND.basereachable != obasereachable)
1281 ND_IFINFO(ifp)->reachable =
1282 ND_COMPUTE_RTIME(ND.basereachable);
1284 if (ND.retrans != 0)
1285 ND_IFINFO(ifp)->retrans = ND.retrans;
1287 ND_IFINFO(ifp)->chlim = ND.chlim;
1289 case SIOCSIFINFO_FLAGS:
1290 ND_IFINFO(ifp)->flags = ND.flags;
1293 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1294 /* sync kernel routing table with the default router list */
1298 case SIOCSPFXFLUSH_IN6:
1300 /* flush all the prefix advertised by routers */
1301 struct nd_prefix *pr, *next;
1304 for (pr = V_nd_prefix.lh_first; pr; pr = next) {
1305 struct in6_ifaddr *ia, *ia_next;
1307 next = pr->ndpr_next;
1309 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1312 /* do we really have to remove addresses as well? */
1313 /* XXXRW: in6_ifaddrhead locking. */
1314 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1316 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1319 if (ia->ia6_ndpr == pr)
1320 in6_purgeaddr(&ia->ia_ifa);
1327 case SIOCSRTRFLUSH_IN6:
1329 /* flush all the default routers */
1330 struct nd_defrouter *dr, *next;
1334 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) {
1335 next = TAILQ_NEXT(dr, dr_entry);
1342 case SIOCGNBRINFO_IN6:
1345 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1347 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1350 IF_AFDATA_LOCK(ifp);
1351 ln = nd6_lookup(&nb_addr, 0, ifp);
1352 IF_AFDATA_UNLOCK(ifp);
1358 nbi->state = ln->ln_state;
1359 nbi->asked = ln->la_asked;
1360 nbi->isrouter = ln->ln_router;
1361 nbi->expire = ln->la_expire;
1365 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1366 ndif->ifindex = V_nd6_defifindex;
1368 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1369 return (nd6_setdefaultiface(ndif->ifindex));
1375 * Create neighbor cache entry and cache link-layer address,
1376 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1379 * code - type dependent information
1382 * The caller of this function already acquired the ndp
1383 * cache table lock because the cache entry is returned.
1386 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1387 int lladdrlen, int type, int code)
1389 struct llentry *ln = NULL;
1396 uint16_t router = 0;
1397 struct sockaddr_in6 sin6;
1398 struct mbuf *chain = NULL;
1399 int static_route = 0;
1401 IF_AFDATA_UNLOCK_ASSERT(ifp);
1404 panic("ifp == NULL in nd6_cache_lladdr");
1406 panic("from == NULL in nd6_cache_lladdr");
1408 /* nothing must be updated for unspecified address */
1409 if (IN6_IS_ADDR_UNSPECIFIED(from))
1413 * Validation about ifp->if_addrlen and lladdrlen must be done in
1416 * XXX If the link does not have link-layer adderss, what should
1417 * we do? (ifp->if_addrlen == 0)
1418 * Spec says nothing in sections for RA, RS and NA. There's small
1419 * description on it in NS section (RFC 2461 7.2.3).
1421 flags |= lladdr ? ND6_EXCLUSIVE : 0;
1422 IF_AFDATA_LOCK(ifp);
1423 ln = nd6_lookup(from, flags, ifp);
1426 flags |= LLE_EXCLUSIVE;
1427 ln = nd6_lookup(from, flags |ND6_CREATE, ifp);
1428 IF_AFDATA_UNLOCK(ifp);
1431 IF_AFDATA_UNLOCK(ifp);
1432 /* do nothing if static ndp is set */
1433 if (ln->la_flags & LLE_STATIC) {
1442 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1443 if (olladdr && lladdr) {
1444 llchange = bcmp(lladdr, &ln->ll_addr,
1450 * newentry olladdr lladdr llchange (*=record)
1453 * 0 n y -- (3) * STALE
1455 * 0 y y y (5) * STALE
1456 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1457 * 1 -- y -- (7) * STALE
1460 if (lladdr) { /* (3-5) and (7) */
1462 * Record source link-layer address
1463 * XXX is it dependent to ifp->if_type?
1465 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1466 ln->la_flags |= LLE_VALID;
1470 if ((!olladdr && lladdr != NULL) || /* (3) */
1471 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1473 newstate = ND6_LLINFO_STALE;
1474 } else /* (1-2,4) */
1478 if (lladdr == NULL) /* (6) */
1479 newstate = ND6_LLINFO_NOSTATE;
1481 newstate = ND6_LLINFO_STALE;
1486 * Update the state of the neighbor cache.
1488 ln->ln_state = newstate;
1490 if (ln->ln_state == ND6_LLINFO_STALE) {
1492 * XXX: since nd6_output() below will cause
1493 * state tansition to DELAY and reset the timer,
1494 * we must set the timer now, although it is actually
1497 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1500 struct mbuf *m_hold, *m_hold_next;
1503 * reset the la_hold in advance, to explicitly
1504 * prevent a la_hold lookup in nd6_output()
1505 * (wouldn't happen, though...)
1507 for (m_hold = ln->la_hold, ln->la_hold = NULL;
1508 m_hold; m_hold = m_hold_next) {
1509 m_hold_next = m_hold->m_nextpkt;
1510 m_hold->m_nextpkt = NULL;
1513 * we assume ifp is not a p2p here, so
1514 * just set the 2nd argument as the
1517 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
1520 * If we have mbufs in the chain we need to do
1521 * deferred transmit. Copy the address from the
1522 * llentry before dropping the lock down below.
1525 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
1527 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1528 /* probe right away */
1529 nd6_llinfo_settimer_locked((void *)ln, 0);
1534 * ICMP6 type dependent behavior.
1536 * NS: clear IsRouter if new entry
1537 * RS: clear IsRouter
1538 * RA: set IsRouter if there's lladdr
1539 * redir: clear IsRouter if new entry
1542 * The spec says that we must set IsRouter in the following cases:
1543 * - If lladdr exist, set IsRouter. This means (1-5).
1544 * - If it is old entry (!newentry), set IsRouter. This means (7).
1545 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1546 * A quetion arises for (1) case. (1) case has no lladdr in the
1547 * neighbor cache, this is similar to (6).
1548 * This case is rare but we figured that we MUST NOT set IsRouter.
1550 * newentry olladdr lladdr llchange NS RS RA redir
1552 * 0 n n -- (1) c ? s
1553 * 0 y n -- (2) c s s
1554 * 0 n y -- (3) c s s
1557 * 1 -- n -- (6) c c c s
1558 * 1 -- y -- (7) c c s c s
1562 switch (type & 0xff) {
1563 case ND_NEIGHBOR_SOLICIT:
1565 * New entry must have is_router flag cleared.
1567 if (is_newentry) /* (6-7) */
1572 * If the icmp is a redirect to a better router, always set the
1573 * is_router flag. Otherwise, if the entry is newly created,
1574 * clear the flag. [RFC 2461, sec 8.3]
1576 if (code == ND_REDIRECT_ROUTER)
1578 else if (is_newentry) /* (6-7) */
1581 case ND_ROUTER_SOLICIT:
1583 * is_router flag must always be cleared.
1587 case ND_ROUTER_ADVERT:
1589 * Mark an entry with lladdr as a router.
1591 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1592 (is_newentry && lladdr)) { /* (7) */
1599 static_route = (ln->la_flags & LLE_STATIC);
1600 router = ln->ln_router;
1602 if (flags & ND6_EXCLUSIVE)
1610 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
1613 * When the link-layer address of a router changes, select the
1614 * best router again. In particular, when the neighbor entry is newly
1615 * created, it might affect the selection policy.
1616 * Question: can we restrict the first condition to the "is_newentry"
1618 * XXX: when we hear an RA from a new router with the link-layer
1619 * address option, defrouter_select() is called twice, since
1620 * defrtrlist_update called the function as well. However, I believe
1621 * we can compromise the overhead, since it only happens the first
1623 * XXX: although defrouter_select() should not have a bad effect
1624 * for those are not autoconfigured hosts, we explicitly avoid such
1627 if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) {
1629 * guaranteed recursion
1637 if (flags & ND6_EXCLUSIVE)
1648 nd6_slowtimo(void *arg)
1650 CURVNET_SET((struct vnet *) arg);
1651 struct nd_ifinfo *nd6if;
1654 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1655 nd6_slowtimo, curvnet);
1656 IFNET_RLOCK_NOSLEEP();
1657 for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
1658 ifp = TAILQ_NEXT(ifp, if_list)) {
1659 nd6if = ND_IFINFO(ifp);
1660 if (nd6if->basereachable && /* already initialized */
1661 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1663 * Since reachable time rarely changes by router
1664 * advertisements, we SHOULD insure that a new random
1665 * value gets recomputed at least once every few hours.
1668 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1669 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1672 IFNET_RUNLOCK_NOSLEEP();
1677 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1678 struct sockaddr_in6 *dst, struct rtentry *rt0)
1681 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
1686 * Note that I'm not enforcing any global serialization
1687 * lle state or asked changes here as the logic is too
1688 * complicated to avoid having to always acquire an exclusive
1693 #define senderr(e) { error = (e); goto bad;}
1696 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1697 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
1698 struct mbuf **chain)
1700 struct mbuf *m = m0;
1701 struct llentry *ln = lle;
1708 LLE_WLOCK_ASSERT(lle);
1710 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
1713 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1716 if (nd6_need_cache(ifp) == 0)
1720 * next hop determination. This routine is derived from ether_output.
1724 * Address resolution or Neighbor Unreachability Detection
1726 * At this point, the destination of the packet must be a unicast
1727 * or an anycast address(i.e. not a multicast).
1730 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0;
1733 IF_AFDATA_LOCK(ifp);
1734 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
1735 IF_AFDATA_UNLOCK(ifp);
1736 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1738 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1739 * the condition below is not very efficient. But we believe
1740 * it is tolerable, because this should be a rare case.
1742 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
1743 IF_AFDATA_LOCK(ifp);
1744 ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
1745 IF_AFDATA_UNLOCK(ifp);
1749 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1750 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1751 char ip6buf[INET6_ADDRSTRLEN];
1753 "nd6_output: can't allocate llinfo for %s "
1755 ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
1756 senderr(EIO); /* XXX: good error? */
1758 goto sendpkt; /* send anyway */
1761 /* We don't have to do link-layer address resolution on a p2p link. */
1762 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1763 ln->ln_state < ND6_LLINFO_REACHABLE) {
1764 if ((flags & LLE_EXCLUSIVE) == 0) {
1765 flags |= LLE_EXCLUSIVE;
1768 ln->ln_state = ND6_LLINFO_STALE;
1769 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1773 * The first time we send a packet to a neighbor whose entry is
1774 * STALE, we have to change the state to DELAY and a sets a timer to
1775 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1776 * neighbor unreachability detection on expiration.
1779 if (ln->ln_state == ND6_LLINFO_STALE) {
1780 if ((flags & LLE_EXCLUSIVE) == 0) {
1781 flags |= LLE_EXCLUSIVE;
1786 ln->ln_state = ND6_LLINFO_DELAY;
1787 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1791 * If the neighbor cache entry has a state other than INCOMPLETE
1792 * (i.e. its link-layer address is already resolved), just
1795 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1799 * There is a neighbor cache entry, but no ethernet address
1800 * response yet. Append this latest packet to the end of the
1801 * packet queue in the mbuf, unless the number of the packet
1802 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1803 * the oldest packet in the queue will be removed.
1805 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1806 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1808 if ((flags & LLE_EXCLUSIVE) == 0) {
1809 flags |= LLE_EXCLUSIVE;
1814 struct mbuf *m_hold;
1818 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
1820 if (m_hold->m_nextpkt == NULL) {
1821 m_hold->m_nextpkt = m;
1825 while (i >= V_nd6_maxqueuelen) {
1826 m_hold = ln->la_hold;
1827 ln->la_hold = ln->la_hold->m_nextpkt;
1835 * We did the lookup (no lle arg) so we
1836 * need to do the unlock here
1839 if (flags & LLE_EXCLUSIVE)
1846 * If there has been no NS for the neighbor after entering the
1847 * INCOMPLETE state, send the first solicitation.
1849 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
1852 nd6_llinfo_settimer(ln,
1853 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
1854 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1859 /* discard the packet if IPv6 operation is disabled on the interface */
1860 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1861 error = ENETDOWN; /* better error? */
1865 * ln is valid and the caller did not pass in
1868 if ((ln != NULL) && (lle == NULL)) {
1869 if (flags & LLE_EXCLUSIVE)
1876 mac_netinet6_nd6_send(ifp, m);
1879 * We were passed in a pointer to an lle with the lock held
1880 * this means that we can't call if_output as we will
1881 * recurse on the lle lock - so what we do is we create
1882 * a list of mbufs to send and transmit them in the caller
1883 * after the lock is dropped
1889 struct mbuf *m = *chain;
1892 * append mbuf to end of deferred chain
1894 while (m->m_nextpkt != NULL)
1900 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1901 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
1904 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
1909 * ln is valid and the caller did not pass in
1912 if ((ln != NULL) && (lle == NULL)) {
1913 if (flags & LLE_EXCLUSIVE)
1926 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
1927 struct sockaddr_in6 *dst, struct route *ro)
1929 struct mbuf *m, *m_head;
1930 struct ifnet *outifp;
1934 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1941 m_head = m_head->m_nextpkt;
1942 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);
1947 * note that intermediate errors are blindly ignored - but this is
1948 * the same convention as used with nd6_output when called by
1956 nd6_need_cache(struct ifnet *ifp)
1959 * XXX: we currently do not make neighbor cache on any interface
1960 * other than ARCnet, Ethernet, FDDI and GIF.
1963 * - unidirectional tunnels needs no ND
1965 switch (ifp->if_type) {
1973 #ifdef IFT_IEEE80211
1979 case IFT_GIF: /* XXX need more cases? */
1983 case IFT_PROPVIRTUAL:
1991 * the callers of this function need to be re-worked to drop
1992 * the lle lock, drop here for now
1995 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
1996 struct sockaddr *dst, u_char *desten, struct llentry **lle)
2001 IF_AFDATA_UNLOCK_ASSERT(ifp);
2002 if (m->m_flags & M_MCAST) {
2005 switch (ifp->if_type) {
2011 #ifdef IFT_IEEE80211
2016 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2021 * netbsd can use if_broadcastaddr, but we don't do so
2022 * to reduce # of ifdef.
2024 for (i = 0; i < ifp->if_addrlen; i++)
2032 return (EAFNOSUPPORT);
2038 * the entry should have been created in nd6_store_lladdr
2040 IF_AFDATA_LOCK(ifp);
2041 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2042 IF_AFDATA_UNLOCK(ifp);
2043 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2046 /* this could happen, if we could not allocate memory */
2051 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2055 * A *small* use after free race exists here
2061 clear_llinfo_pqueue(struct llentry *ln)
2063 struct mbuf *m_hold, *m_hold_next;
2065 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2066 m_hold_next = m_hold->m_nextpkt;
2067 m_hold->m_nextpkt = NULL;
2075 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2076 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2078 SYSCTL_DECL(_net_inet6_icmp6);
2080 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2081 CTLFLAG_RD, nd6_sysctl_drlist, "");
2082 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2083 CTLFLAG_RD, nd6_sysctl_prlist, "");
2084 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2085 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2088 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2091 char buf[1024] __aligned(4);
2092 struct in6_defrouter *d, *de;
2093 struct nd_defrouter *dr;
2099 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
2100 dr = TAILQ_NEXT(dr, dr_entry)) {
2101 d = (struct in6_defrouter *)buf;
2102 de = (struct in6_defrouter *)(buf + sizeof(buf));
2105 bzero(d, sizeof(*d));
2106 d->rtaddr.sin6_family = AF_INET6;
2107 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2108 d->rtaddr.sin6_addr = dr->rtaddr;
2109 error = sa6_recoverscope(&d->rtaddr);
2112 d->flags = dr->flags;
2113 d->rtlifetime = dr->rtlifetime;
2114 d->expire = dr->expire;
2115 d->if_index = dr->ifp->if_index;
2117 panic("buffer too short");
2119 error = SYSCTL_OUT(req, buf, sizeof(*d));
2128 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2131 char buf[1024] __aligned(4);
2132 struct in6_prefix *p, *pe;
2133 struct nd_prefix *pr;
2134 char ip6buf[INET6_ADDRSTRLEN];
2140 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2143 struct sockaddr_in6 *sin6, *s6;
2144 struct nd_pfxrouter *pfr;
2146 p = (struct in6_prefix *)buf;
2147 pe = (struct in6_prefix *)(buf + sizeof(buf));
2150 bzero(p, sizeof(*p));
2151 sin6 = (struct sockaddr_in6 *)(p + 1);
2153 p->prefix = pr->ndpr_prefix;
2154 if (sa6_recoverscope(&p->prefix)) {
2156 "scope error in prefix list (%s)\n",
2157 ip6_sprintf(ip6buf, &p->prefix.sin6_addr));
2158 /* XXX: press on... */
2160 p->raflags = pr->ndpr_raf;
2161 p->prefixlen = pr->ndpr_plen;
2162 p->vltime = pr->ndpr_vltime;
2163 p->pltime = pr->ndpr_pltime;
2164 p->if_index = pr->ndpr_ifp->if_index;
2165 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2170 /* XXX: we assume time_t is signed. */
2173 ((sizeof(maxexpire) * 8) - 1));
2174 if (pr->ndpr_vltime <
2175 maxexpire - pr->ndpr_lastupdate) {
2176 p->expire = pr->ndpr_lastupdate +
2179 p->expire = maxexpire;
2181 p->refcnt = pr->ndpr_refcnt;
2182 p->flags = pr->ndpr_stateflags;
2183 p->origin = PR_ORIG_RA;
2185 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2186 pfr = pfr->pfr_next) {
2187 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2191 s6 = &sin6[advrtrs];
2192 bzero(s6, sizeof(*s6));
2193 s6->sin6_family = AF_INET6;
2194 s6->sin6_len = sizeof(*sin6);
2195 s6->sin6_addr = pfr->router->rtaddr;
2196 if (sa6_recoverscope(s6)) {
2199 "prefix list (%s)\n",
2201 &pfr->router->rtaddr));
2205 p->advrtrs = advrtrs;
2207 panic("buffer too short");
2209 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2210 error = SYSCTL_OUT(req, buf, advance);