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)
85 VNET_DEFINE(int, nd6_prune);
86 VNET_DEFINE(int, nd6_delay);
87 VNET_DEFINE(int, nd6_umaxtries);
88 VNET_DEFINE(int, nd6_mmaxtries);
89 VNET_DEFINE(int, nd6_useloopback);
90 VNET_DEFINE(int, nd6_gctimer);
92 /* preventing too many loops in ND option parsing */
93 static VNET_DEFINE(int, nd6_maxndopt);
94 VNET_DEFINE(int, nd6_maxnudhint);
95 static VNET_DEFINE(int, nd6_maxqueuelen);
96 #define V_nd6_maxndopt VNET(nd6_maxndopt)
97 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
99 VNET_DEFINE(int, nd6_debug);
103 static int nd6_inuse, nd6_allocated;
106 VNET_DEFINE(struct nd_drhead, nd_defrouter);
107 VNET_DEFINE(struct nd_prhead, nd_prefix);
109 VNET_DEFINE(int, nd6_recalc_reachtm_interval);
110 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
112 static struct sockaddr_in6 all1_sa;
114 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
116 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
117 static void nd6_slowtimo(void *);
118 static int regen_tmpaddr(struct in6_ifaddr *);
119 static struct llentry *nd6_free(struct llentry *, int);
120 static void nd6_llinfo_timer(void *);
121 static void clear_llinfo_pqueue(struct llentry *);
123 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
124 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
126 VNET_DEFINE(struct callout, nd6_timer_ch);
128 VNET_DECLARE(int, dad_ignore_ns);
129 VNET_DECLARE(int, dad_maxtry);
130 #define V_dad_ignore_ns VNET(dad_ignore_ns)
131 #define V_dad_maxtry VNET(dad_maxtry)
138 V_nd6_prune = 1; /* walk list every 1 seconds */
139 V_nd6_delay = 5; /* delay first probe time 5 second */
140 V_nd6_umaxtries = 3; /* maximum unicast query */
141 V_nd6_mmaxtries = 3; /* maximum multicast query */
142 V_nd6_useloopback = 1; /* use loopback interface for local traffic */
143 V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
145 /* preventing too many loops in ND option parsing */
146 V_nd6_maxndopt = 10; /* max # of ND options allowed */
148 V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
149 V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */
157 V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
159 V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/
160 V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */
163 * XXX just to get this to compile KMM
166 V_llinfo_nd6.ln_next = &V_llinfo_nd6;
167 V_llinfo_nd6.ln_prev = &V_llinfo_nd6;
169 LIST_INIT(&V_nd_prefix);
171 V_ip6_use_tempaddr = 0;
172 V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
173 V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
174 V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
176 V_ip6_desync_factor = 0;
178 all1_sa.sin6_family = AF_INET6;
179 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
180 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
181 all1_sa.sin6_addr.s6_addr[i] = 0xff;
183 /* initialization of the default router list */
184 TAILQ_INIT(&V_nd_defrouter);
186 callout_init(&V_nd6_slowtimo_ch, 0);
187 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
188 nd6_slowtimo, curvnet);
197 callout_drain(&V_nd6_slowtimo_ch);
198 callout_drain(&V_nd6_timer_ch);
203 nd6_ifattach(struct ifnet *ifp)
205 struct nd_ifinfo *nd;
207 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
208 bzero(nd, sizeof(*nd));
212 nd->chlim = IPV6_DEFHLIM;
213 nd->basereachable = REACHABLE_TIME;
214 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
215 nd->retrans = RETRANS_TIMER;
217 nd->flags = ND6_IFF_PERFORMNUD;
219 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. */
220 if (V_ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK))
221 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
223 /* A loopback interface does not need to accept RTADV. */
224 if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK))
225 nd->flags |= ND6_IFF_ACCEPT_RTADV;
227 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
228 nd6_setmtu0(ifp, nd);
234 nd6_ifdetach(struct nd_ifinfo *nd)
241 * Reset ND level link MTU. This function is called when the physical MTU
242 * changes, which means we might have to adjust the ND level MTU.
245 nd6_setmtu(struct ifnet *ifp)
248 nd6_setmtu0(ifp, ND_IFINFO(ifp));
251 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
253 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
257 omaxmtu = ndi->maxmtu;
259 switch (ifp->if_type) {
261 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
264 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
267 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
270 ndi->maxmtu = ifp->if_mtu;
275 * Decreasing the interface MTU under IPV6 minimum MTU may cause
276 * undesirable situation. We thus notify the operator of the change
277 * explicitly. The check for omaxmtu is necessary to restrict the
278 * log to the case of changing the MTU, not initializing it.
280 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
281 log(LOG_NOTICE, "nd6_setmtu0: "
282 "new link MTU on %s (%lu) is too small for IPv6\n",
283 if_name(ifp), (unsigned long)ndi->maxmtu);
286 if (ndi->maxmtu > V_in6_maxmtu)
287 in6_setmaxmtu(); /* check all interfaces just in case */
292 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
295 bzero(ndopts, sizeof(*ndopts));
296 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
298 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
301 ndopts->nd_opts_done = 1;
302 ndopts->nd_opts_search = NULL;
307 * Take one ND option.
310 nd6_option(union nd_opts *ndopts)
312 struct nd_opt_hdr *nd_opt;
316 panic("ndopts == NULL in nd6_option");
317 if (ndopts->nd_opts_last == NULL)
318 panic("uninitialized ndopts in nd6_option");
319 if (ndopts->nd_opts_search == NULL)
321 if (ndopts->nd_opts_done)
324 nd_opt = ndopts->nd_opts_search;
326 /* make sure nd_opt_len is inside the buffer */
327 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
328 bzero(ndopts, sizeof(*ndopts));
332 olen = nd_opt->nd_opt_len << 3;
335 * Message validation requires that all included
336 * options have a length that is greater than zero.
338 bzero(ndopts, sizeof(*ndopts));
342 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
343 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
344 /* option overruns the end of buffer, invalid */
345 bzero(ndopts, sizeof(*ndopts));
347 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
348 /* reached the end of options chain */
349 ndopts->nd_opts_done = 1;
350 ndopts->nd_opts_search = NULL;
356 * Parse multiple ND options.
357 * This function is much easier to use, for ND routines that do not need
358 * multiple options of the same type.
361 nd6_options(union nd_opts *ndopts)
363 struct nd_opt_hdr *nd_opt;
367 panic("ndopts == NULL in nd6_options");
368 if (ndopts->nd_opts_last == NULL)
369 panic("uninitialized ndopts in nd6_options");
370 if (ndopts->nd_opts_search == NULL)
374 nd_opt = nd6_option(ndopts);
375 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
377 * Message validation requires that all included
378 * options have a length that is greater than zero.
380 ICMP6STAT_INC(icp6s_nd_badopt);
381 bzero(ndopts, sizeof(*ndopts));
388 switch (nd_opt->nd_opt_type) {
389 case ND_OPT_SOURCE_LINKADDR:
390 case ND_OPT_TARGET_LINKADDR:
392 case ND_OPT_REDIRECTED_HEADER:
393 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
395 "duplicated ND6 option found (type=%d)\n",
396 nd_opt->nd_opt_type));
399 ndopts->nd_opt_array[nd_opt->nd_opt_type]
403 case ND_OPT_PREFIX_INFORMATION:
404 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
405 ndopts->nd_opt_array[nd_opt->nd_opt_type]
408 ndopts->nd_opts_pi_end =
409 (struct nd_opt_prefix_info *)nd_opt;
413 * Unknown options must be silently ignored,
414 * to accomodate future extension to the protocol.
417 "nd6_options: unsupported option %d - "
418 "option ignored\n", nd_opt->nd_opt_type));
423 if (i > V_nd6_maxndopt) {
424 ICMP6STAT_INC(icp6s_nd_toomanyopt);
425 nd6log((LOG_INFO, "too many loop in nd opt\n"));
429 if (ndopts->nd_opts_done)
437 * ND6 timer routine to handle ND6 entries
440 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
447 canceled = callout_stop(&ln->ln_timer_ch);
449 ln->la_expire = time_second + tick / hz;
451 if (tick > INT_MAX) {
452 ln->ln_ntick = tick - INT_MAX;
453 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
454 nd6_llinfo_timer, ln);
457 canceled = callout_reset(&ln->ln_timer_ch, tick,
458 nd6_llinfo_timer, ln);
466 nd6_llinfo_settimer(struct llentry *ln, long tick)
470 nd6_llinfo_settimer_locked(ln, tick);
475 nd6_llinfo_timer(void *arg)
478 struct in6_addr *dst;
480 struct nd_ifinfo *ndi = NULL;
482 ln = (struct llentry *)arg;
484 panic("%s: NULL entry!\n", __func__);
488 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL)
489 panic("ln ifp == NULL");
491 CURVNET_SET(ifp->if_vnet);
493 if (ln->ln_ntick > 0) {
494 if (ln->ln_ntick > INT_MAX) {
495 ln->ln_ntick -= INT_MAX;
496 nd6_llinfo_settimer(ln, INT_MAX);
499 nd6_llinfo_settimer(ln, ln->ln_ntick);
504 ndi = ND_IFINFO(ifp);
505 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
506 if (ln->la_flags & LLE_STATIC) {
510 if (ln->la_flags & LLE_DELETED) {
511 (void)nd6_free(ln, 0);
516 switch (ln->ln_state) {
517 case ND6_LLINFO_INCOMPLETE:
518 if (ln->la_asked < V_nd6_mmaxtries) {
520 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
521 nd6_ns_output(ifp, NULL, dst, ln, 0);
523 struct mbuf *m = ln->la_hold;
528 * assuming every packet in la_hold has the
533 icmp6_error2(m, ICMP6_DST_UNREACH,
534 ICMP6_DST_UNREACH_ADDR, 0, ifp);
537 clear_llinfo_pqueue(ln);
539 (void)nd6_free(ln, 0);
543 case ND6_LLINFO_REACHABLE:
544 if (!ND6_LLINFO_PERMANENT(ln)) {
545 ln->ln_state = ND6_LLINFO_STALE;
546 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
550 case ND6_LLINFO_STALE:
551 /* Garbage Collection(RFC 2461 5.3) */
552 if (!ND6_LLINFO_PERMANENT(ln)) {
553 (void)nd6_free(ln, 1);
558 case ND6_LLINFO_DELAY:
559 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
562 ln->ln_state = ND6_LLINFO_PROBE;
563 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
564 nd6_ns_output(ifp, dst, dst, ln, 0);
566 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
567 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
570 case ND6_LLINFO_PROBE:
571 if (ln->la_asked < V_nd6_umaxtries) {
573 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
574 nd6_ns_output(ifp, dst, dst, ln, 0);
576 (void)nd6_free(ln, 0);
589 * ND6 timer routine to expire default route list and prefix list
594 CURVNET_SET((struct vnet *) arg);
596 struct nd_defrouter *dr;
597 struct nd_prefix *pr;
598 struct in6_ifaddr *ia6, *nia6;
599 struct in6_addrlifetime *lt6;
601 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
604 /* expire default router list */
606 dr = TAILQ_FIRST(&V_nd_defrouter);
608 if (dr->expire && dr->expire < time_second) {
609 struct nd_defrouter *t;
610 t = TAILQ_NEXT(dr, dr_entry);
614 dr = TAILQ_NEXT(dr, dr_entry);
619 * expire interface addresses.
620 * in the past the loop was inside prefix expiry processing.
621 * However, from a stricter speci-confrmance standpoint, we should
622 * rather separate address lifetimes and prefix lifetimes.
624 * XXXRW: in6_ifaddrhead locking.
627 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
628 /* check address lifetime */
629 lt6 = &ia6->ia6_lifetime;
630 if (IFA6_IS_INVALID(ia6)) {
634 * If the expiring address is temporary, try
635 * regenerating a new one. This would be useful when
636 * we suspended a laptop PC, then turned it on after a
637 * period that could invalidate all temporary
638 * addresses. Although we may have to restart the
639 * loop (see below), it must be after purging the
640 * address. Otherwise, we'd see an infinite loop of
643 if (V_ip6_use_tempaddr &&
644 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
645 if (regen_tmpaddr(ia6) == 0)
649 in6_purgeaddr(&ia6->ia_ifa);
652 goto addrloop; /* XXX: see below */
653 } else if (IFA6_IS_DEPRECATED(ia6)) {
654 int oldflags = ia6->ia6_flags;
656 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
659 * If a temporary address has just become deprecated,
660 * regenerate a new one if possible.
662 if (V_ip6_use_tempaddr &&
663 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
664 (oldflags & IN6_IFF_DEPRECATED) == 0) {
666 if (regen_tmpaddr(ia6) == 0) {
668 * A new temporary address is
670 * XXX: this means the address chain
671 * has changed while we are still in
672 * the loop. Although the change
673 * would not cause disaster (because
674 * it's not a deletion, but an
675 * addition,) we'd rather restart the
676 * loop just for safety. Or does this
677 * significantly reduce performance??
684 * A new RA might have made a deprecated address
687 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
691 /* expire prefix list */
692 pr = V_nd_prefix.lh_first;
695 * check prefix lifetime.
696 * since pltime is just for autoconf, pltime processing for
697 * prefix is not necessary.
699 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
700 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
705 * address expiration and prefix expiration are
706 * separate. NEVER perform in6_purgeaddr here.
719 * ia6 - deprecated/invalidated temporary address
722 regen_tmpaddr(struct in6_ifaddr *ia6)
726 struct in6_ifaddr *public_ifa6 = NULL;
728 ifp = ia6->ia_ifa.ifa_ifp;
730 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
731 struct in6_ifaddr *it6;
733 if (ifa->ifa_addr->sa_family != AF_INET6)
736 it6 = (struct in6_ifaddr *)ifa;
738 /* ignore no autoconf addresses. */
739 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
742 /* ignore autoconf addresses with different prefixes. */
743 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
747 * Now we are looking at an autoconf address with the same
748 * prefix as ours. If the address is temporary and is still
749 * preferred, do not create another one. It would be rare, but
750 * could happen, for example, when we resume a laptop PC after
753 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
754 !IFA6_IS_DEPRECATED(it6)) {
760 * This is a public autoconf address that has the same prefix
761 * as ours. If it is preferred, keep it. We can't break the
762 * loop here, because there may be a still-preferred temporary
763 * address with the prefix.
765 if (!IFA6_IS_DEPRECATED(it6))
768 if (public_ifa6 != NULL)
769 ifa_ref(&public_ifa6->ia_ifa);
773 if (public_ifa6 != NULL) {
776 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
777 ifa_free(&public_ifa6->ia_ifa);
778 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
779 " tmp addr,errno=%d\n", e);
782 ifa_free(&public_ifa6->ia_ifa);
790 * Nuke neighbor cache/prefix/default router management table, right before
794 nd6_purge(struct ifnet *ifp)
796 struct nd_defrouter *dr, *ndr;
797 struct nd_prefix *pr, *npr;
800 * Nuke default router list entries toward ifp.
801 * We defer removal of default router list entries that is installed
802 * in the routing table, in order to keep additional side effects as
805 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
806 ndr = TAILQ_NEXT(dr, dr_entry);
814 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) {
815 ndr = TAILQ_NEXT(dr, dr_entry);
823 /* Nuke prefix list entries toward ifp */
824 for (pr = V_nd_prefix.lh_first; pr; pr = npr) {
826 if (pr->ndpr_ifp == ifp) {
828 * Because if_detach() does *not* release prefixes
829 * while purging addresses the reference count will
830 * still be above zero. We therefore reset it to
831 * make sure that the prefix really gets purged.
836 * Previously, pr->ndpr_addr is removed as well,
837 * but I strongly believe we don't have to do it.
838 * nd6_purge() is only called from in6_ifdetach(),
839 * which removes all the associated interface addresses
841 * (jinmei@kame.net 20010129)
847 /* cancel default outgoing interface setting */
848 if (V_nd6_defifindex == ifp->if_index)
849 nd6_setdefaultiface(0);
851 if (!V_ip6_forwarding && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
852 /* Refresh default router list. */
857 * We do not nuke the neighbor cache entries here any more
858 * because the neighbor cache is kept in if_afdata[AF_INET6].
859 * nd6_purge() is invoked by in6_ifdetach() which is called
860 * from if_detach() where everything gets purged. So let
861 * in6_domifdetach() do the actual L2 table purging work.
866 * the caller acquires and releases the lock on the lltbls
867 * Returns the llentry locked
870 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
872 struct sockaddr_in6 sin6;
876 bzero(&sin6, sizeof(sin6));
877 sin6.sin6_len = sizeof(struct sockaddr_in6);
878 sin6.sin6_family = AF_INET6;
879 sin6.sin6_addr = *addr6;
881 IF_AFDATA_LOCK_ASSERT(ifp);
883 if (flags & ND6_CREATE)
884 llflags |= LLE_CREATE;
885 if (flags & ND6_EXCLUSIVE)
886 llflags |= LLE_EXCLUSIVE;
888 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
889 if ((ln != NULL) && (flags & LLE_CREATE)) {
890 ln->ln_state = ND6_LLINFO_NOSTATE;
891 callout_init(&ln->ln_timer_ch, 0);
898 * Test whether a given IPv6 address is a neighbor or not, ignoring
899 * the actual neighbor cache. The neighbor cache is ignored in order
900 * to not reenter the routing code from within itself.
903 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
905 struct nd_prefix *pr;
906 struct ifaddr *dstaddr;
909 * A link-local address is always a neighbor.
910 * XXX: a link does not necessarily specify a single interface.
912 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
913 struct sockaddr_in6 sin6_copy;
917 * We need sin6_copy since sa6_recoverscope() may modify the
921 if (sa6_recoverscope(&sin6_copy))
922 return (0); /* XXX: should be impossible */
923 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
925 if (sin6_copy.sin6_scope_id == zone)
932 * If the address matches one of our addresses,
933 * it should be a neighbor.
934 * If the address matches one of our on-link prefixes, it should be a
937 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
938 if (pr->ndpr_ifp != ifp)
941 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
943 rt = rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0);
947 * This is the case where multiple interfaces
948 * have the same prefix, but only one is installed
949 * into the routing table and that prefix entry
950 * is not the one being examined here. In the case
951 * where RADIX_MPATH is enabled, multiple route
952 * entries (of the same rt_key value) will be
953 * installed because the interface addresses all
956 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
957 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
964 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
965 &addr->sin6_addr, &pr->ndpr_mask))
970 * If the address is assigned on the node of the other side of
971 * a p2p interface, the address should be a neighbor.
973 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
974 if (dstaddr != NULL) {
975 if (dstaddr->ifa_ifp == ifp) {
983 * If the default router list is empty, all addresses are regarded
984 * as on-link, and thus, as a neighbor.
985 * XXX: we restrict the condition to hosts, because routers usually do
986 * not have the "default router list".
988 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL &&
989 V_nd6_defifindex == ifp->if_index) {
998 * Detect if a given IPv6 address identifies a neighbor on a given link.
999 * XXX: should take care of the destination of a p2p link?
1002 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
1004 struct llentry *lle;
1007 IF_AFDATA_UNLOCK_ASSERT(ifp);
1008 if (nd6_is_new_addr_neighbor(addr, ifp))
1012 * Even if the address matches none of our addresses, it might be
1013 * in the neighbor cache.
1015 IF_AFDATA_LOCK(ifp);
1016 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1020 IF_AFDATA_UNLOCK(ifp);
1025 * Free an nd6 llinfo entry.
1026 * Since the function would cause significant changes in the kernel, DO NOT
1027 * make it global, unless you have a strong reason for the change, and are sure
1028 * that the change is safe.
1030 static struct llentry *
1031 nd6_free(struct llentry *ln, int gc)
1033 struct llentry *next;
1034 struct nd_defrouter *dr;
1035 struct ifnet *ifp=NULL;
1038 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1039 * even though it is not harmful, it was not really necessary.
1043 nd6_llinfo_settimer(ln, -1);
1045 if (!V_ip6_forwarding) {
1048 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1050 if (dr != NULL && dr->expire &&
1051 ln->ln_state == ND6_LLINFO_STALE && gc) {
1053 * If the reason for the deletion is just garbage
1054 * collection, and the neighbor is an active default
1055 * router, do not delete it. Instead, reset the GC
1056 * timer using the router's lifetime.
1057 * Simply deleting the entry would affect default
1058 * router selection, which is not necessarily a good
1059 * thing, especially when we're using router preference
1061 * XXX: the check for ln_state would be redundant,
1062 * but we intentionally keep it just in case.
1064 if (dr->expire > time_second)
1065 nd6_llinfo_settimer(ln,
1066 (dr->expire - time_second) * hz);
1068 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
1073 return (LIST_NEXT(ln, lle_next));
1076 if (ln->ln_router || dr) {
1078 * rt6_flush must be called whether or not the neighbor
1079 * is in the Default Router List.
1080 * See a corresponding comment in nd6_na_input().
1082 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp);
1087 * Unreachablity of a router might affect the default
1088 * router selection and on-link detection of advertised
1093 * Temporarily fake the state to choose a new default
1094 * router and to perform on-link determination of
1095 * prefixes correctly.
1096 * Below the state will be set correctly,
1097 * or the entry itself will be deleted.
1099 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1102 * Since defrouter_select() does not affect the
1103 * on-link determination and MIP6 needs the check
1104 * before the default router selection, we perform
1107 pfxlist_onlink_check();
1110 * refresh default router list
1118 * Before deleting the entry, remember the next entry as the
1119 * return value. We need this because pfxlist_onlink_check() above
1120 * might have freed other entries (particularly the old next entry) as
1121 * a side effect (XXX).
1123 next = LIST_NEXT(ln, lle_next);
1125 ifp = ln->lle_tbl->llt_ifp;
1126 IF_AFDATA_LOCK(ifp);
1129 IF_AFDATA_UNLOCK(ifp);
1135 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1137 * XXX cost-effective methods?
1140 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1145 if ((dst6 == NULL) || (rt == NULL))
1149 IF_AFDATA_LOCK(ifp);
1150 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1151 IF_AFDATA_UNLOCK(ifp);
1155 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1159 * if we get upper-layer reachability confirmation many times,
1160 * it is possible we have false information.
1164 if (ln->ln_byhint > V_nd6_maxnudhint) {
1169 ln->ln_state = ND6_LLINFO_REACHABLE;
1170 if (!ND6_LLINFO_PERMANENT(ln)) {
1171 nd6_llinfo_settimer(ln,
1172 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1180 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1182 struct in6_drlist *drl = (struct in6_drlist *)data;
1183 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1184 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1185 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1186 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1187 struct nd_defrouter *dr;
1188 struct nd_prefix *pr;
1189 int i = 0, error = 0;
1193 case SIOCGDRLST_IN6:
1195 * obsolete API, use sysctl under net.inet6.icmp6
1197 bzero(drl, sizeof(*drl));
1199 dr = TAILQ_FIRST(&V_nd_defrouter);
1200 while (dr && i < DRLSTSIZ) {
1201 drl->defrouter[i].rtaddr = dr->rtaddr;
1202 in6_clearscope(&drl->defrouter[i].rtaddr);
1204 drl->defrouter[i].flags = dr->flags;
1205 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1206 drl->defrouter[i].expire = dr->expire;
1207 drl->defrouter[i].if_index = dr->ifp->if_index;
1209 dr = TAILQ_NEXT(dr, dr_entry);
1213 case SIOCGPRLST_IN6:
1215 * obsolete API, use sysctl under net.inet6.icmp6
1217 * XXX the structure in6_prlist was changed in backward-
1218 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1219 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1222 * XXX meaning of fields, especialy "raflags", is very
1223 * differnet between RA prefix list and RR/static prefix list.
1224 * how about separating ioctls into two?
1226 bzero(oprl, sizeof(*oprl));
1228 pr = V_nd_prefix.lh_first;
1229 while (pr && i < PRLSTSIZ) {
1230 struct nd_pfxrouter *pfr;
1233 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1234 oprl->prefix[i].raflags = pr->ndpr_raf;
1235 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1236 oprl->prefix[i].vltime = pr->ndpr_vltime;
1237 oprl->prefix[i].pltime = pr->ndpr_pltime;
1238 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1239 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1240 oprl->prefix[i].expire = 0;
1244 /* XXX: we assume time_t is signed. */
1247 ((sizeof(maxexpire) * 8) - 1));
1248 if (pr->ndpr_vltime <
1249 maxexpire - pr->ndpr_lastupdate) {
1250 oprl->prefix[i].expire =
1251 pr->ndpr_lastupdate +
1254 oprl->prefix[i].expire = maxexpire;
1257 pfr = pr->ndpr_advrtrs.lh_first;
1261 #define RTRADDR oprl->prefix[i].advrtr[j]
1262 RTRADDR = pfr->router->rtaddr;
1263 in6_clearscope(&RTRADDR);
1267 pfr = pfr->pfr_next;
1269 oprl->prefix[i].advrtrs = j;
1270 oprl->prefix[i].origin = PR_ORIG_RA;
1278 case OSIOCGIFINFO_IN6:
1280 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1281 bzero(&ND, sizeof(ND));
1282 ND.linkmtu = IN6_LINKMTU(ifp);
1283 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1284 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1285 ND.reachable = ND_IFINFO(ifp)->reachable;
1286 ND.retrans = ND_IFINFO(ifp)->retrans;
1287 ND.flags = ND_IFINFO(ifp)->flags;
1288 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1289 ND.chlim = ND_IFINFO(ifp)->chlim;
1291 case SIOCGIFINFO_IN6:
1292 ND = *ND_IFINFO(ifp);
1294 case SIOCSIFINFO_IN6:
1296 * used to change host variables from userland.
1297 * intented for a use on router to reflect RA configurations.
1299 /* 0 means 'unspecified' */
1300 if (ND.linkmtu != 0) {
1301 if (ND.linkmtu < IPV6_MMTU ||
1302 ND.linkmtu > IN6_LINKMTU(ifp)) {
1306 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1309 if (ND.basereachable != 0) {
1310 int obasereachable = ND_IFINFO(ifp)->basereachable;
1312 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1313 if (ND.basereachable != obasereachable)
1314 ND_IFINFO(ifp)->reachable =
1315 ND_COMPUTE_RTIME(ND.basereachable);
1317 if (ND.retrans != 0)
1318 ND_IFINFO(ifp)->retrans = ND.retrans;
1320 ND_IFINFO(ifp)->chlim = ND.chlim;
1322 case SIOCSIFINFO_FLAGS:
1325 struct in6_ifaddr *ia;
1327 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1328 !(ND.flags & ND6_IFF_IFDISABLED)) {
1329 /* ifdisabled 1->0 transision */
1332 * If the interface is marked as ND6_IFF_IFDISABLED and
1333 * has an link-local address with IN6_IFF_DUPLICATED,
1334 * do not clear ND6_IFF_IFDISABLED.
1335 * See RFC 4862, Section 5.4.5.
1337 int duplicated_linklocal = 0;
1340 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1341 if (ifa->ifa_addr->sa_family != AF_INET6)
1343 ia = (struct in6_ifaddr *)ifa;
1344 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1345 IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1346 duplicated_linklocal = 1;
1350 IF_ADDR_UNLOCK(ifp);
1352 if (duplicated_linklocal) {
1353 ND.flags |= ND6_IFF_IFDISABLED;
1354 log(LOG_ERR, "Cannot enable an interface"
1355 " with a link-local address marked"
1358 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1361 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1362 (ND.flags & ND6_IFF_IFDISABLED)) {
1363 /* ifdisabled 0->1 transision */
1364 /* Mark all IPv6 address as tentative. */
1366 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1368 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1369 if (ifa->ifa_addr->sa_family != AF_INET6)
1371 ia = (struct in6_ifaddr *)ifa;
1372 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1374 IF_ADDR_UNLOCK(ifp);
1377 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) &&
1378 (ND.flags & ND6_IFF_AUTO_LINKLOCAL)) {
1379 /* auto_linklocal 0->1 transision */
1381 /* If no link-local address on ifp, configure */
1382 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1383 in6_ifattach(ifp, NULL);
1386 ND_IFINFO(ifp)->flags = ND.flags;
1389 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1390 /* sync kernel routing table with the default router list */
1394 case SIOCSPFXFLUSH_IN6:
1396 /* flush all the prefix advertised by routers */
1397 struct nd_prefix *pr, *next;
1400 for (pr = V_nd_prefix.lh_first; pr; pr = next) {
1401 struct in6_ifaddr *ia, *ia_next;
1403 next = pr->ndpr_next;
1405 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1408 /* do we really have to remove addresses as well? */
1409 /* XXXRW: in6_ifaddrhead locking. */
1410 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1412 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1415 if (ia->ia6_ndpr == pr)
1416 in6_purgeaddr(&ia->ia_ifa);
1423 case SIOCSRTRFLUSH_IN6:
1425 /* flush all the default routers */
1426 struct nd_defrouter *dr, *next;
1430 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) {
1431 next = TAILQ_NEXT(dr, dr_entry);
1438 case SIOCGNBRINFO_IN6:
1441 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1443 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1446 IF_AFDATA_LOCK(ifp);
1447 ln = nd6_lookup(&nb_addr, 0, ifp);
1448 IF_AFDATA_UNLOCK(ifp);
1454 nbi->state = ln->ln_state;
1455 nbi->asked = ln->la_asked;
1456 nbi->isrouter = ln->ln_router;
1457 nbi->expire = ln->la_expire;
1461 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1462 ndif->ifindex = V_nd6_defifindex;
1464 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1465 return (nd6_setdefaultiface(ndif->ifindex));
1471 * Create neighbor cache entry and cache link-layer address,
1472 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1475 * code - type dependent information
1478 * The caller of this function already acquired the ndp
1479 * cache table lock because the cache entry is returned.
1482 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1483 int lladdrlen, int type, int code)
1485 struct llentry *ln = NULL;
1492 uint16_t router = 0;
1493 struct sockaddr_in6 sin6;
1494 struct mbuf *chain = NULL;
1495 int static_route = 0;
1497 IF_AFDATA_UNLOCK_ASSERT(ifp);
1500 panic("ifp == NULL in nd6_cache_lladdr");
1502 panic("from == NULL in nd6_cache_lladdr");
1504 /* nothing must be updated for unspecified address */
1505 if (IN6_IS_ADDR_UNSPECIFIED(from))
1509 * Validation about ifp->if_addrlen and lladdrlen must be done in
1512 * XXX If the link does not have link-layer adderss, what should
1513 * we do? (ifp->if_addrlen == 0)
1514 * Spec says nothing in sections for RA, RS and NA. There's small
1515 * description on it in NS section (RFC 2461 7.2.3).
1517 flags |= lladdr ? ND6_EXCLUSIVE : 0;
1518 IF_AFDATA_LOCK(ifp);
1519 ln = nd6_lookup(from, flags, ifp);
1522 flags |= LLE_EXCLUSIVE;
1523 ln = nd6_lookup(from, flags |ND6_CREATE, ifp);
1524 IF_AFDATA_UNLOCK(ifp);
1527 IF_AFDATA_UNLOCK(ifp);
1528 /* do nothing if static ndp is set */
1529 if (ln->la_flags & LLE_STATIC) {
1538 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1539 if (olladdr && lladdr) {
1540 llchange = bcmp(lladdr, &ln->ll_addr,
1546 * newentry olladdr lladdr llchange (*=record)
1549 * 0 n y -- (3) * STALE
1551 * 0 y y y (5) * STALE
1552 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1553 * 1 -- y -- (7) * STALE
1556 if (lladdr) { /* (3-5) and (7) */
1558 * Record source link-layer address
1559 * XXX is it dependent to ifp->if_type?
1561 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1562 ln->la_flags |= LLE_VALID;
1566 if ((!olladdr && lladdr != NULL) || /* (3) */
1567 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1569 newstate = ND6_LLINFO_STALE;
1570 } else /* (1-2,4) */
1574 if (lladdr == NULL) /* (6) */
1575 newstate = ND6_LLINFO_NOSTATE;
1577 newstate = ND6_LLINFO_STALE;
1582 * Update the state of the neighbor cache.
1584 ln->ln_state = newstate;
1586 if (ln->ln_state == ND6_LLINFO_STALE) {
1588 * XXX: since nd6_output() below will cause
1589 * state tansition to DELAY and reset the timer,
1590 * we must set the timer now, although it is actually
1593 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1596 struct mbuf *m_hold, *m_hold_next;
1599 * reset the la_hold in advance, to explicitly
1600 * prevent a la_hold lookup in nd6_output()
1601 * (wouldn't happen, though...)
1603 for (m_hold = ln->la_hold, ln->la_hold = NULL;
1604 m_hold; m_hold = m_hold_next) {
1605 m_hold_next = m_hold->m_nextpkt;
1606 m_hold->m_nextpkt = NULL;
1609 * we assume ifp is not a p2p here, so
1610 * just set the 2nd argument as the
1613 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
1616 * If we have mbufs in the chain we need to do
1617 * deferred transmit. Copy the address from the
1618 * llentry before dropping the lock down below.
1621 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
1623 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1624 /* probe right away */
1625 nd6_llinfo_settimer_locked((void *)ln, 0);
1630 * ICMP6 type dependent behavior.
1632 * NS: clear IsRouter if new entry
1633 * RS: clear IsRouter
1634 * RA: set IsRouter if there's lladdr
1635 * redir: clear IsRouter if new entry
1638 * The spec says that we must set IsRouter in the following cases:
1639 * - If lladdr exist, set IsRouter. This means (1-5).
1640 * - If it is old entry (!newentry), set IsRouter. This means (7).
1641 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1642 * A quetion arises for (1) case. (1) case has no lladdr in the
1643 * neighbor cache, this is similar to (6).
1644 * This case is rare but we figured that we MUST NOT set IsRouter.
1646 * newentry olladdr lladdr llchange NS RS RA redir
1648 * 0 n n -- (1) c ? s
1649 * 0 y n -- (2) c s s
1650 * 0 n y -- (3) c s s
1653 * 1 -- n -- (6) c c c s
1654 * 1 -- y -- (7) c c s c s
1658 switch (type & 0xff) {
1659 case ND_NEIGHBOR_SOLICIT:
1661 * New entry must have is_router flag cleared.
1663 if (is_newentry) /* (6-7) */
1668 * If the icmp is a redirect to a better router, always set the
1669 * is_router flag. Otherwise, if the entry is newly created,
1670 * clear the flag. [RFC 2461, sec 8.3]
1672 if (code == ND_REDIRECT_ROUTER)
1674 else if (is_newentry) /* (6-7) */
1677 case ND_ROUTER_SOLICIT:
1679 * is_router flag must always be cleared.
1683 case ND_ROUTER_ADVERT:
1685 * Mark an entry with lladdr as a router.
1687 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1688 (is_newentry && lladdr)) { /* (7) */
1695 static_route = (ln->la_flags & LLE_STATIC);
1696 router = ln->ln_router;
1698 if (flags & ND6_EXCLUSIVE)
1706 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
1709 * When the link-layer address of a router changes, select the
1710 * best router again. In particular, when the neighbor entry is newly
1711 * created, it might affect the selection policy.
1712 * Question: can we restrict the first condition to the "is_newentry"
1714 * XXX: when we hear an RA from a new router with the link-layer
1715 * address option, defrouter_select() is called twice, since
1716 * defrtrlist_update called the function as well. However, I believe
1717 * we can compromise the overhead, since it only happens the first
1719 * XXX: although defrouter_select() should not have a bad effect
1720 * for those are not autoconfigured hosts, we explicitly avoid such
1723 if (do_update && router && !V_ip6_forwarding &&
1724 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1726 * guaranteed recursion
1734 if (flags & ND6_EXCLUSIVE)
1745 nd6_slowtimo(void *arg)
1747 CURVNET_SET((struct vnet *) arg);
1748 struct nd_ifinfo *nd6if;
1751 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1752 nd6_slowtimo, curvnet);
1753 IFNET_RLOCK_NOSLEEP();
1754 for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
1755 ifp = TAILQ_NEXT(ifp, if_list)) {
1756 nd6if = ND_IFINFO(ifp);
1757 if (nd6if->basereachable && /* already initialized */
1758 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1760 * Since reachable time rarely changes by router
1761 * advertisements, we SHOULD insure that a new random
1762 * value gets recomputed at least once every few hours.
1765 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1766 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1769 IFNET_RUNLOCK_NOSLEEP();
1774 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1775 struct sockaddr_in6 *dst, struct rtentry *rt0)
1778 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
1783 * Note that I'm not enforcing any global serialization
1784 * lle state or asked changes here as the logic is too
1785 * complicated to avoid having to always acquire an exclusive
1790 #define senderr(e) { error = (e); goto bad;}
1793 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1794 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
1795 struct mbuf **chain)
1797 struct mbuf *m = m0;
1798 struct llentry *ln = lle;
1805 LLE_WLOCK_ASSERT(lle);
1807 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
1810 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1813 if (nd6_need_cache(ifp) == 0)
1817 * next hop determination. This routine is derived from ether_output.
1821 * Address resolution or Neighbor Unreachability Detection
1823 * At this point, the destination of the packet must be a unicast
1824 * or an anycast address(i.e. not a multicast).
1827 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0;
1830 IF_AFDATA_LOCK(ifp);
1831 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
1832 IF_AFDATA_UNLOCK(ifp);
1833 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1835 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1836 * the condition below is not very efficient. But we believe
1837 * it is tolerable, because this should be a rare case.
1839 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
1840 IF_AFDATA_LOCK(ifp);
1841 ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
1842 IF_AFDATA_UNLOCK(ifp);
1846 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1847 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1848 char ip6buf[INET6_ADDRSTRLEN];
1850 "nd6_output: can't allocate llinfo for %s "
1852 ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
1853 senderr(EIO); /* XXX: good error? */
1855 goto sendpkt; /* send anyway */
1858 /* We don't have to do link-layer address resolution on a p2p link. */
1859 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1860 ln->ln_state < ND6_LLINFO_REACHABLE) {
1861 if ((flags & LLE_EXCLUSIVE) == 0) {
1862 flags |= LLE_EXCLUSIVE;
1865 ln->ln_state = ND6_LLINFO_STALE;
1866 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1870 * The first time we send a packet to a neighbor whose entry is
1871 * STALE, we have to change the state to DELAY and a sets a timer to
1872 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1873 * neighbor unreachability detection on expiration.
1876 if (ln->ln_state == ND6_LLINFO_STALE) {
1877 if ((flags & LLE_EXCLUSIVE) == 0) {
1878 flags |= LLE_EXCLUSIVE;
1883 ln->ln_state = ND6_LLINFO_DELAY;
1884 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1888 * If the neighbor cache entry has a state other than INCOMPLETE
1889 * (i.e. its link-layer address is already resolved), just
1892 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1896 * There is a neighbor cache entry, but no ethernet address
1897 * response yet. Append this latest packet to the end of the
1898 * packet queue in the mbuf, unless the number of the packet
1899 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1900 * the oldest packet in the queue will be removed.
1902 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1903 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1905 if ((flags & LLE_EXCLUSIVE) == 0) {
1906 flags |= LLE_EXCLUSIVE;
1911 struct mbuf *m_hold;
1915 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
1917 if (m_hold->m_nextpkt == NULL) {
1918 m_hold->m_nextpkt = m;
1922 while (i >= V_nd6_maxqueuelen) {
1923 m_hold = ln->la_hold;
1924 ln->la_hold = ln->la_hold->m_nextpkt;
1932 * We did the lookup (no lle arg) so we
1933 * need to do the unlock here
1936 if (flags & LLE_EXCLUSIVE)
1943 * If there has been no NS for the neighbor after entering the
1944 * INCOMPLETE state, send the first solicitation.
1946 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
1949 nd6_llinfo_settimer(ln,
1950 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
1951 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1956 /* discard the packet if IPv6 operation is disabled on the interface */
1957 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1958 error = ENETDOWN; /* better error? */
1962 * ln is valid and the caller did not pass in
1965 if ((ln != NULL) && (lle == NULL)) {
1966 if (flags & LLE_EXCLUSIVE)
1973 mac_netinet6_nd6_send(ifp, m);
1976 * We were passed in a pointer to an lle with the lock held
1977 * this means that we can't call if_output as we will
1978 * recurse on the lle lock - so what we do is we create
1979 * a list of mbufs to send and transmit them in the caller
1980 * after the lock is dropped
1986 struct mbuf *m = *chain;
1989 * append mbuf to end of deferred chain
1991 while (m->m_nextpkt != NULL)
1997 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1998 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2001 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
2006 * ln is valid and the caller did not pass in
2009 if ((ln != NULL) && (lle == NULL)) {
2010 if (flags & LLE_EXCLUSIVE)
2023 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2024 struct sockaddr_in6 *dst, struct route *ro)
2026 struct mbuf *m, *m_head;
2027 struct ifnet *outifp;
2031 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2038 m_head = m_head->m_nextpkt;
2039 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);
2044 * note that intermediate errors are blindly ignored - but this is
2045 * the same convention as used with nd6_output when called by
2053 nd6_need_cache(struct ifnet *ifp)
2056 * XXX: we currently do not make neighbor cache on any interface
2057 * other than ARCnet, Ethernet, FDDI and GIF.
2060 * - unidirectional tunnels needs no ND
2062 switch (ifp->if_type) {
2070 #ifdef IFT_IEEE80211
2076 case IFT_GIF: /* XXX need more cases? */
2080 case IFT_PROPVIRTUAL:
2088 * the callers of this function need to be re-worked to drop
2089 * the lle lock, drop here for now
2092 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2093 struct sockaddr *dst, u_char *desten, struct llentry **lle)
2098 IF_AFDATA_UNLOCK_ASSERT(ifp);
2099 if (m->m_flags & M_MCAST) {
2102 switch (ifp->if_type) {
2108 #ifdef IFT_IEEE80211
2113 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2118 * netbsd can use if_broadcastaddr, but we don't do so
2119 * to reduce # of ifdef.
2121 for (i = 0; i < ifp->if_addrlen; i++)
2129 return (EAFNOSUPPORT);
2135 * the entry should have been created in nd6_store_lladdr
2137 IF_AFDATA_LOCK(ifp);
2138 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2139 IF_AFDATA_UNLOCK(ifp);
2140 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2143 /* this could happen, if we could not allocate memory */
2148 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2152 * A *small* use after free race exists here
2158 clear_llinfo_pqueue(struct llentry *ln)
2160 struct mbuf *m_hold, *m_hold_next;
2162 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2163 m_hold_next = m_hold->m_nextpkt;
2164 m_hold->m_nextpkt = NULL;
2172 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2173 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2175 SYSCTL_DECL(_net_inet6_icmp6);
2177 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2178 CTLFLAG_RD, nd6_sysctl_drlist, "");
2179 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2180 CTLFLAG_RD, nd6_sysctl_prlist, "");
2181 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2182 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2185 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2188 char buf[1024] __aligned(4);
2189 struct in6_defrouter *d, *de;
2190 struct nd_defrouter *dr;
2196 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
2197 dr = TAILQ_NEXT(dr, dr_entry)) {
2198 d = (struct in6_defrouter *)buf;
2199 de = (struct in6_defrouter *)(buf + sizeof(buf));
2202 bzero(d, sizeof(*d));
2203 d->rtaddr.sin6_family = AF_INET6;
2204 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2205 d->rtaddr.sin6_addr = dr->rtaddr;
2206 error = sa6_recoverscope(&d->rtaddr);
2209 d->flags = dr->flags;
2210 d->rtlifetime = dr->rtlifetime;
2211 d->expire = dr->expire;
2212 d->if_index = dr->ifp->if_index;
2214 panic("buffer too short");
2216 error = SYSCTL_OUT(req, buf, sizeof(*d));
2225 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2228 char buf[1024] __aligned(4);
2229 struct in6_prefix *p, *pe;
2230 struct nd_prefix *pr;
2231 char ip6buf[INET6_ADDRSTRLEN];
2237 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2240 struct sockaddr_in6 *sin6, *s6;
2241 struct nd_pfxrouter *pfr;
2243 p = (struct in6_prefix *)buf;
2244 pe = (struct in6_prefix *)(buf + sizeof(buf));
2247 bzero(p, sizeof(*p));
2248 sin6 = (struct sockaddr_in6 *)(p + 1);
2250 p->prefix = pr->ndpr_prefix;
2251 if (sa6_recoverscope(&p->prefix)) {
2253 "scope error in prefix list (%s)\n",
2254 ip6_sprintf(ip6buf, &p->prefix.sin6_addr));
2255 /* XXX: press on... */
2257 p->raflags = pr->ndpr_raf;
2258 p->prefixlen = pr->ndpr_plen;
2259 p->vltime = pr->ndpr_vltime;
2260 p->pltime = pr->ndpr_pltime;
2261 p->if_index = pr->ndpr_ifp->if_index;
2262 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2267 /* XXX: we assume time_t is signed. */
2270 ((sizeof(maxexpire) * 8) - 1));
2271 if (pr->ndpr_vltime <
2272 maxexpire - pr->ndpr_lastupdate) {
2273 p->expire = pr->ndpr_lastupdate +
2276 p->expire = maxexpire;
2278 p->refcnt = pr->ndpr_refcnt;
2279 p->flags = pr->ndpr_stateflags;
2280 p->origin = PR_ORIG_RA;
2282 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2283 pfr = pfr->pfr_next) {
2284 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2288 s6 = &sin6[advrtrs];
2289 bzero(s6, sizeof(*s6));
2290 s6->sin6_family = AF_INET6;
2291 s6->sin6_len = sizeof(*sin6);
2292 s6->sin6_addr = pfr->router->rtaddr;
2293 if (sa6_recoverscope(s6)) {
2296 "prefix list (%s)\n",
2298 &pfr->router->rtaddr));
2302 p->advrtrs = advrtrs;
2304 panic("buffer too short");
2306 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2307 error = SYSCTL_OUT(req, buf, advance);