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>
75 #include <netinet6/send.h>
77 #include <sys/limits.h>
79 #include <security/mac/mac_framework.h>
81 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
82 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
84 #define SIN6(s) ((struct sockaddr_in6 *)s)
87 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
88 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
89 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
90 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
91 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
93 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
96 /* preventing too many loops in ND option parsing */
97 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
99 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
101 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
103 #define V_nd6_maxndopt VNET(nd6_maxndopt)
104 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
107 VNET_DEFINE(int, nd6_debug) = 1;
109 VNET_DEFINE(int, nd6_debug) = 0;
114 static int nd6_inuse, nd6_allocated;
117 VNET_DEFINE(struct nd_drhead, nd_defrouter);
118 VNET_DEFINE(struct nd_prhead, nd_prefix);
120 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
121 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
123 static struct sockaddr_in6 all1_sa;
125 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
127 static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *,
129 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
130 static void nd6_slowtimo(void *);
131 static int regen_tmpaddr(struct in6_ifaddr *);
132 static struct llentry *nd6_free(struct llentry *, int);
133 static void nd6_llinfo_timer(void *);
134 static void clear_llinfo_pqueue(struct llentry *);
136 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
137 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
139 VNET_DEFINE(struct callout, nd6_timer_ch);
146 LIST_INIT(&V_nd_prefix);
148 all1_sa.sin6_family = AF_INET6;
149 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
150 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
151 all1_sa.sin6_addr.s6_addr[i] = 0xff;
153 /* initialization of the default router list */
154 TAILQ_INIT(&V_nd_defrouter);
157 callout_init(&V_nd6_slowtimo_ch, 0);
158 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
159 nd6_slowtimo, curvnet);
167 callout_drain(&V_nd6_slowtimo_ch);
168 callout_drain(&V_nd6_timer_ch);
173 nd6_ifattach(struct ifnet *ifp)
175 struct nd_ifinfo *nd;
177 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO);
180 nd->chlim = IPV6_DEFHLIM;
181 nd->basereachable = REACHABLE_TIME;
182 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
183 nd->retrans = RETRANS_TIMER;
185 nd->flags = ND6_IFF_PERFORMNUD;
187 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
188 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
189 * default regardless of the V_ip6_auto_linklocal configuration to
190 * give a reasonable default behavior.
192 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
193 (ifp->if_flags & IFF_LOOPBACK))
194 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
196 * A loopback interface does not need to accept RTADV.
197 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
198 * default regardless of the V_ip6_accept_rtadv configuration to
199 * prevent the interface from accepting RA messages arrived
200 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
202 if (V_ip6_accept_rtadv &&
203 !(ifp->if_flags & IFF_LOOPBACK) &&
204 (ifp->if_type != IFT_BRIDGE))
205 nd->flags |= ND6_IFF_ACCEPT_RTADV;
206 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
207 nd->flags |= ND6_IFF_NO_RADR;
209 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
210 nd6_setmtu0(ifp, nd);
216 nd6_ifdetach(struct nd_ifinfo *nd)
223 * Reset ND level link MTU. This function is called when the physical MTU
224 * changes, which means we might have to adjust the ND level MTU.
227 nd6_setmtu(struct ifnet *ifp)
230 nd6_setmtu0(ifp, ND_IFINFO(ifp));
233 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
235 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
239 omaxmtu = ndi->maxmtu;
241 switch (ifp->if_type) {
243 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
246 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
249 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
252 ndi->maxmtu = ifp->if_mtu;
257 * Decreasing the interface MTU under IPV6 minimum MTU may cause
258 * undesirable situation. We thus notify the operator of the change
259 * explicitly. The check for omaxmtu is necessary to restrict the
260 * log to the case of changing the MTU, not initializing it.
262 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
263 log(LOG_NOTICE, "nd6_setmtu0: "
264 "new link MTU on %s (%lu) is too small for IPv6\n",
265 if_name(ifp), (unsigned long)ndi->maxmtu);
268 if (ndi->maxmtu > V_in6_maxmtu)
269 in6_setmaxmtu(); /* check all interfaces just in case */
274 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
277 bzero(ndopts, sizeof(*ndopts));
278 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
280 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
283 ndopts->nd_opts_done = 1;
284 ndopts->nd_opts_search = NULL;
289 * Take one ND option.
292 nd6_option(union nd_opts *ndopts)
294 struct nd_opt_hdr *nd_opt;
297 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
298 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
300 if (ndopts->nd_opts_search == NULL)
302 if (ndopts->nd_opts_done)
305 nd_opt = ndopts->nd_opts_search;
307 /* make sure nd_opt_len is inside the buffer */
308 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
309 bzero(ndopts, sizeof(*ndopts));
313 olen = nd_opt->nd_opt_len << 3;
316 * Message validation requires that all included
317 * options have a length that is greater than zero.
319 bzero(ndopts, sizeof(*ndopts));
323 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
324 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
325 /* option overruns the end of buffer, invalid */
326 bzero(ndopts, sizeof(*ndopts));
328 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
329 /* reached the end of options chain */
330 ndopts->nd_opts_done = 1;
331 ndopts->nd_opts_search = NULL;
337 * Parse multiple ND options.
338 * This function is much easier to use, for ND routines that do not need
339 * multiple options of the same type.
342 nd6_options(union nd_opts *ndopts)
344 struct nd_opt_hdr *nd_opt;
347 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
348 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
350 if (ndopts->nd_opts_search == NULL)
354 nd_opt = nd6_option(ndopts);
355 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
357 * Message validation requires that all included
358 * options have a length that is greater than zero.
360 ICMP6STAT_INC(icp6s_nd_badopt);
361 bzero(ndopts, sizeof(*ndopts));
368 switch (nd_opt->nd_opt_type) {
369 case ND_OPT_SOURCE_LINKADDR:
370 case ND_OPT_TARGET_LINKADDR:
372 case ND_OPT_REDIRECTED_HEADER:
373 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
375 "duplicated ND6 option found (type=%d)\n",
376 nd_opt->nd_opt_type));
379 ndopts->nd_opt_array[nd_opt->nd_opt_type]
383 case ND_OPT_PREFIX_INFORMATION:
384 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
385 ndopts->nd_opt_array[nd_opt->nd_opt_type]
388 ndopts->nd_opts_pi_end =
389 (struct nd_opt_prefix_info *)nd_opt;
391 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
392 case ND_OPT_RDNSS: /* RFC 6106 */
393 case ND_OPT_DNSSL: /* RFC 6106 */
395 * Silently ignore options we know and do not care about
401 * Unknown options must be silently ignored,
402 * to accomodate future extension to the protocol.
405 "nd6_options: unsupported option %d - "
406 "option ignored\n", nd_opt->nd_opt_type));
411 if (i > V_nd6_maxndopt) {
412 ICMP6STAT_INC(icp6s_nd_toomanyopt);
413 nd6log((LOG_INFO, "too many loop in nd opt\n"));
417 if (ndopts->nd_opts_done)
425 * ND6 timer routine to handle ND6 entries
428 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
432 LLE_WLOCK_ASSERT(ln);
437 canceled = callout_stop(&ln->ln_timer_ch);
439 ln->la_expire = time_second + tick / hz;
441 if (tick > INT_MAX) {
442 ln->ln_ntick = tick - INT_MAX;
443 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
444 nd6_llinfo_timer, ln);
447 canceled = callout_reset(&ln->ln_timer_ch, tick,
448 nd6_llinfo_timer, ln);
456 nd6_llinfo_settimer(struct llentry *ln, long tick)
460 nd6_llinfo_settimer_locked(ln, tick);
465 nd6_llinfo_timer(void *arg)
468 struct in6_addr *dst;
470 struct nd_ifinfo *ndi = NULL;
472 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
473 ln = (struct llentry *)arg;
474 LLE_WLOCK_ASSERT(ln);
475 ifp = ln->lle_tbl->llt_ifp;
477 CURVNET_SET(ifp->if_vnet);
479 if (ln->ln_ntick > 0) {
480 if (ln->ln_ntick > INT_MAX) {
481 ln->ln_ntick -= INT_MAX;
482 nd6_llinfo_settimer_locked(ln, INT_MAX);
485 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
490 ndi = ND_IFINFO(ifp);
491 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
492 if (ln->la_flags & LLE_STATIC) {
496 if (ln->la_flags & LLE_DELETED) {
497 (void)nd6_free(ln, 0);
502 switch (ln->ln_state) {
503 case ND6_LLINFO_INCOMPLETE:
504 if (ln->la_asked < V_nd6_mmaxtries) {
506 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
508 nd6_ns_output(ifp, NULL, dst, ln, 0);
511 struct mbuf *m = ln->la_hold;
516 * assuming every packet in la_hold has the
517 * same IP header. Send error after unlock.
522 clear_llinfo_pqueue(ln);
524 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT);
525 (void)nd6_free(ln, 0);
528 icmp6_error2(m, ICMP6_DST_UNREACH,
529 ICMP6_DST_UNREACH_ADDR, 0, ifp);
532 case ND6_LLINFO_REACHABLE:
533 if (!ND6_LLINFO_PERMANENT(ln)) {
534 ln->ln_state = ND6_LLINFO_STALE;
535 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
539 case ND6_LLINFO_STALE:
540 /* Garbage Collection(RFC 2461 5.3) */
541 if (!ND6_LLINFO_PERMANENT(ln)) {
542 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
543 (void)nd6_free(ln, 1);
548 case ND6_LLINFO_DELAY:
549 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
552 ln->ln_state = ND6_LLINFO_PROBE;
553 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
555 nd6_ns_output(ifp, dst, dst, ln, 0);
558 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
559 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
562 case ND6_LLINFO_PROBE:
563 if (ln->la_asked < V_nd6_umaxtries) {
565 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
567 nd6_ns_output(ifp, dst, dst, ln, 0);
570 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
571 (void)nd6_free(ln, 0);
576 panic("%s: paths in a dark night can be confusing: %d",
577 __func__, ln->ln_state);
587 * ND6 timer routine to expire default route list and prefix list
592 CURVNET_SET((struct vnet *) arg);
594 struct nd_defrouter *dr, *ndr;
595 struct nd_prefix *pr, *npr;
596 struct in6_ifaddr *ia6, *nia6;
598 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
601 /* expire default router list */
603 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
604 if (dr->expire && dr->expire < time_second)
609 * expire interface addresses.
610 * in the past the loop was inside prefix expiry processing.
611 * However, from a stricter speci-confrmance standpoint, we should
612 * rather separate address lifetimes and prefix lifetimes.
614 * XXXRW: in6_ifaddrhead locking.
617 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
618 /* check address lifetime */
619 if (IFA6_IS_INVALID(ia6)) {
623 * If the expiring address is temporary, try
624 * regenerating a new one. This would be useful when
625 * we suspended a laptop PC, then turned it on after a
626 * period that could invalidate all temporary
627 * addresses. Although we may have to restart the
628 * loop (see below), it must be after purging the
629 * address. Otherwise, we'd see an infinite loop of
632 if (V_ip6_use_tempaddr &&
633 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
634 if (regen_tmpaddr(ia6) == 0)
638 in6_purgeaddr(&ia6->ia_ifa);
641 goto addrloop; /* XXX: see below */
642 } else if (IFA6_IS_DEPRECATED(ia6)) {
643 int oldflags = ia6->ia6_flags;
645 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
648 * If a temporary address has just become deprecated,
649 * regenerate a new one if possible.
651 if (V_ip6_use_tempaddr &&
652 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
653 (oldflags & IN6_IFF_DEPRECATED) == 0) {
655 if (regen_tmpaddr(ia6) == 0) {
657 * A new temporary address is
659 * XXX: this means the address chain
660 * has changed while we are still in
661 * the loop. Although the change
662 * would not cause disaster (because
663 * it's not a deletion, but an
664 * addition,) we'd rather restart the
665 * loop just for safety. Or does this
666 * significantly reduce performance??
673 * A new RA might have made a deprecated address
676 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
680 /* expire prefix list */
681 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
683 * check prefix lifetime.
684 * since pltime is just for autoconf, pltime processing for
685 * prefix is not necessary.
687 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
688 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
691 * address expiration and prefix expiration are
692 * separate. NEVER perform in6_purgeaddr here.
702 * ia6 - deprecated/invalidated temporary address
705 regen_tmpaddr(struct in6_ifaddr *ia6)
709 struct in6_ifaddr *public_ifa6 = NULL;
711 ifp = ia6->ia_ifa.ifa_ifp;
713 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
714 struct in6_ifaddr *it6;
716 if (ifa->ifa_addr->sa_family != AF_INET6)
719 it6 = (struct in6_ifaddr *)ifa;
721 /* ignore no autoconf addresses. */
722 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
725 /* ignore autoconf addresses with different prefixes. */
726 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
730 * Now we are looking at an autoconf address with the same
731 * prefix as ours. If the address is temporary and is still
732 * preferred, do not create another one. It would be rare, but
733 * could happen, for example, when we resume a laptop PC after
736 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
737 !IFA6_IS_DEPRECATED(it6)) {
743 * This is a public autoconf address that has the same prefix
744 * as ours. If it is preferred, keep it. We can't break the
745 * loop here, because there may be a still-preferred temporary
746 * address with the prefix.
748 if (!IFA6_IS_DEPRECATED(it6))
751 if (public_ifa6 != NULL)
752 ifa_ref(&public_ifa6->ia_ifa);
754 IF_ADDR_RUNLOCK(ifp);
756 if (public_ifa6 != NULL) {
759 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
760 ifa_free(&public_ifa6->ia_ifa);
761 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
762 " tmp addr,errno=%d\n", e);
765 ifa_free(&public_ifa6->ia_ifa);
773 * Nuke neighbor cache/prefix/default router management table, right before
777 nd6_purge(struct ifnet *ifp)
779 struct nd_defrouter *dr, *ndr;
780 struct nd_prefix *pr, *npr;
783 * Nuke default router list entries toward ifp.
784 * We defer removal of default router list entries that is installed
785 * in the routing table, in order to keep additional side effects as
788 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
796 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
804 /* Nuke prefix list entries toward ifp */
805 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
806 if (pr->ndpr_ifp == ifp) {
808 * Because if_detach() does *not* release prefixes
809 * while purging addresses the reference count will
810 * still be above zero. We therefore reset it to
811 * make sure that the prefix really gets purged.
816 * Previously, pr->ndpr_addr is removed as well,
817 * but I strongly believe we don't have to do it.
818 * nd6_purge() is only called from in6_ifdetach(),
819 * which removes all the associated interface addresses
821 * (jinmei@kame.net 20010129)
827 /* cancel default outgoing interface setting */
828 if (V_nd6_defifindex == ifp->if_index)
829 nd6_setdefaultiface(0);
831 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
832 /* Refresh default router list. */
837 * We do not nuke the neighbor cache entries here any more
838 * because the neighbor cache is kept in if_afdata[AF_INET6].
839 * nd6_purge() is invoked by in6_ifdetach() which is called
840 * from if_detach() where everything gets purged. So let
841 * in6_domifdetach() do the actual L2 table purging work.
846 * the caller acquires and releases the lock on the lltbls
847 * Returns the llentry locked
850 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
852 struct sockaddr_in6 sin6;
856 bzero(&sin6, sizeof(sin6));
857 sin6.sin6_len = sizeof(struct sockaddr_in6);
858 sin6.sin6_family = AF_INET6;
859 sin6.sin6_addr = *addr6;
861 IF_AFDATA_LOCK_ASSERT(ifp);
864 if (flags & ND6_CREATE)
865 llflags |= LLE_CREATE;
866 if (flags & ND6_EXCLUSIVE)
867 llflags |= LLE_EXCLUSIVE;
869 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
870 if ((ln != NULL) && (llflags & LLE_CREATE))
871 ln->ln_state = ND6_LLINFO_NOSTATE;
877 * Test whether a given IPv6 address is a neighbor or not, ignoring
878 * the actual neighbor cache. The neighbor cache is ignored in order
879 * to not reenter the routing code from within itself.
882 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
884 struct nd_prefix *pr;
885 struct ifaddr *dstaddr;
888 * A link-local address is always a neighbor.
889 * XXX: a link does not necessarily specify a single interface.
891 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
892 struct sockaddr_in6 sin6_copy;
896 * We need sin6_copy since sa6_recoverscope() may modify the
900 if (sa6_recoverscope(&sin6_copy))
901 return (0); /* XXX: should be impossible */
902 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
904 if (sin6_copy.sin6_scope_id == zone)
911 * If the address matches one of our addresses,
912 * it should be a neighbor.
913 * If the address matches one of our on-link prefixes, it should be a
916 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
917 if (pr->ndpr_ifp != ifp)
920 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
923 /* Always use the default FIB here. */
924 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
925 0, 0, RT_DEFAULT_FIB);
929 * This is the case where multiple interfaces
930 * have the same prefix, but only one is installed
931 * into the routing table and that prefix entry
932 * is not the one being examined here. In the case
933 * where RADIX_MPATH is enabled, multiple route
934 * entries (of the same rt_key value) will be
935 * installed because the interface addresses all
938 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
939 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
946 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
947 &addr->sin6_addr, &pr->ndpr_mask))
952 * If the address is assigned on the node of the other side of
953 * a p2p interface, the address should be a neighbor.
955 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
956 if (dstaddr != NULL) {
957 if (dstaddr->ifa_ifp == ifp) {
965 * If the default router list is empty, all addresses are regarded
966 * as on-link, and thus, as a neighbor.
968 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
969 TAILQ_EMPTY(&V_nd_defrouter) &&
970 V_nd6_defifindex == ifp->if_index) {
979 * Detect if a given IPv6 address identifies a neighbor on a given link.
980 * XXX: should take care of the destination of a p2p link?
983 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
988 IF_AFDATA_UNLOCK_ASSERT(ifp);
989 if (nd6_is_new_addr_neighbor(addr, ifp))
993 * Even if the address matches none of our addresses, it might be
994 * in the neighbor cache.
996 IF_AFDATA_RLOCK(ifp);
997 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1001 IF_AFDATA_RUNLOCK(ifp);
1006 * Free an nd6 llinfo entry.
1007 * Since the function would cause significant changes in the kernel, DO NOT
1008 * make it global, unless you have a strong reason for the change, and are sure
1009 * that the change is safe.
1011 static struct llentry *
1012 nd6_free(struct llentry *ln, int gc)
1014 struct llentry *next;
1015 struct nd_defrouter *dr;
1018 LLE_WLOCK_ASSERT(ln);
1021 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1022 * even though it is not harmful, it was not really necessary.
1026 nd6_llinfo_settimer_locked(ln, -1);
1028 ifp = ln->lle_tbl->llt_ifp;
1030 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1031 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1033 if (dr != NULL && dr->expire &&
1034 ln->ln_state == ND6_LLINFO_STALE && gc) {
1036 * If the reason for the deletion is just garbage
1037 * collection, and the neighbor is an active default
1038 * router, do not delete it. Instead, reset the GC
1039 * timer using the router's lifetime.
1040 * Simply deleting the entry would affect default
1041 * router selection, which is not necessarily a good
1042 * thing, especially when we're using router preference
1044 * XXX: the check for ln_state would be redundant,
1045 * but we intentionally keep it just in case.
1047 if (dr->expire > time_second)
1048 nd6_llinfo_settimer_locked(ln,
1049 (dr->expire - time_second) * hz);
1051 nd6_llinfo_settimer_locked(ln,
1052 (long)V_nd6_gctimer * hz);
1054 next = LIST_NEXT(ln, lle_next);
1062 * Unreachablity of a router might affect the default
1063 * router selection and on-link detection of advertised
1068 * Temporarily fake the state to choose a new default
1069 * router and to perform on-link determination of
1070 * prefixes correctly.
1071 * Below the state will be set correctly,
1072 * or the entry itself will be deleted.
1074 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1077 if (ln->ln_router || dr) {
1080 * We need to unlock to avoid a LOR with rt6_flush() with the
1081 * rnh and for the calls to pfxlist_onlink_check() and
1082 * defrouter_select() in the block further down for calls
1083 * into nd6_lookup(). We still hold a ref.
1088 * rt6_flush must be called whether or not the neighbor
1089 * is in the Default Router List.
1090 * See a corresponding comment in nd6_na_input().
1092 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1097 * Since defrouter_select() does not affect the
1098 * on-link determination and MIP6 needs the check
1099 * before the default router selection, we perform
1102 pfxlist_onlink_check();
1105 * Refresh default router list.
1110 if (ln->ln_router || dr)
1115 * Before deleting the entry, remember the next entry as the
1116 * return value. We need this because pfxlist_onlink_check() above
1117 * might have freed other entries (particularly the old next entry) as
1118 * a side effect (XXX).
1120 next = LIST_NEXT(ln, lle_next);
1123 * Save to unlock. We still hold an extra reference and will not
1124 * free(9) in llentry_free() if someone else holds one as well.
1127 IF_AFDATA_LOCK(ifp);
1130 /* Guard against race with other llentry_free(). */
1131 if (ln->la_flags & LLE_LINKED) {
1135 LLE_FREE_LOCKED(ln);
1137 IF_AFDATA_UNLOCK(ifp);
1143 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1145 * XXX cost-effective methods?
1148 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1153 if ((dst6 == NULL) || (rt == NULL))
1157 IF_AFDATA_RLOCK(ifp);
1158 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1159 IF_AFDATA_RUNLOCK(ifp);
1163 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1167 * if we get upper-layer reachability confirmation many times,
1168 * it is possible we have false information.
1172 if (ln->ln_byhint > V_nd6_maxnudhint) {
1177 ln->ln_state = ND6_LLINFO_REACHABLE;
1178 if (!ND6_LLINFO_PERMANENT(ln)) {
1179 nd6_llinfo_settimer_locked(ln,
1180 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1188 * Rejuvenate this function for routing operations related
1192 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1194 struct sockaddr_in6 *gateway;
1195 struct nd_defrouter *dr;
1199 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1210 * Only indirect routes are interesting.
1212 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1215 * check for default route
1217 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1218 &SIN6(rt_key(rt))->sin6_addr)) {
1220 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1230 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1232 struct in6_drlist *drl = (struct in6_drlist *)data;
1233 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1234 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1235 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1236 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1237 struct nd_defrouter *dr;
1238 struct nd_prefix *pr;
1239 int i = 0, error = 0;
1243 case SIOCGDRLST_IN6:
1245 * obsolete API, use sysctl under net.inet6.icmp6
1247 bzero(drl, sizeof(*drl));
1249 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1252 drl->defrouter[i].rtaddr = dr->rtaddr;
1253 in6_clearscope(&drl->defrouter[i].rtaddr);
1255 drl->defrouter[i].flags = dr->flags;
1256 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1257 drl->defrouter[i].expire = dr->expire;
1258 drl->defrouter[i].if_index = dr->ifp->if_index;
1263 case SIOCGPRLST_IN6:
1265 * obsolete API, use sysctl under net.inet6.icmp6
1267 * XXX the structure in6_prlist was changed in backward-
1268 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1269 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1272 * XXX meaning of fields, especialy "raflags", is very
1273 * differnet between RA prefix list and RR/static prefix list.
1274 * how about separating ioctls into two?
1276 bzero(oprl, sizeof(*oprl));
1278 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1279 struct nd_pfxrouter *pfr;
1284 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1285 oprl->prefix[i].raflags = pr->ndpr_raf;
1286 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1287 oprl->prefix[i].vltime = pr->ndpr_vltime;
1288 oprl->prefix[i].pltime = pr->ndpr_pltime;
1289 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1290 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1291 oprl->prefix[i].expire = 0;
1295 /* XXX: we assume time_t is signed. */
1298 ((sizeof(maxexpire) * 8) - 1));
1299 if (pr->ndpr_vltime <
1300 maxexpire - pr->ndpr_lastupdate) {
1301 oprl->prefix[i].expire =
1302 pr->ndpr_lastupdate +
1305 oprl->prefix[i].expire = maxexpire;
1309 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1311 #define RTRADDR oprl->prefix[i].advrtr[j]
1312 RTRADDR = pfr->router->rtaddr;
1313 in6_clearscope(&RTRADDR);
1318 oprl->prefix[i].advrtrs = j;
1319 oprl->prefix[i].origin = PR_ORIG_RA;
1326 case OSIOCGIFINFO_IN6:
1328 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1329 bzero(&ND, sizeof(ND));
1330 ND.linkmtu = IN6_LINKMTU(ifp);
1331 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1332 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1333 ND.reachable = ND_IFINFO(ifp)->reachable;
1334 ND.retrans = ND_IFINFO(ifp)->retrans;
1335 ND.flags = ND_IFINFO(ifp)->flags;
1336 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1337 ND.chlim = ND_IFINFO(ifp)->chlim;
1339 case SIOCGIFINFO_IN6:
1340 ND = *ND_IFINFO(ifp);
1342 case SIOCSIFINFO_IN6:
1344 * used to change host variables from userland.
1345 * intented for a use on router to reflect RA configurations.
1347 /* 0 means 'unspecified' */
1348 if (ND.linkmtu != 0) {
1349 if (ND.linkmtu < IPV6_MMTU ||
1350 ND.linkmtu > IN6_LINKMTU(ifp)) {
1354 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1357 if (ND.basereachable != 0) {
1358 int obasereachable = ND_IFINFO(ifp)->basereachable;
1360 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1361 if (ND.basereachable != obasereachable)
1362 ND_IFINFO(ifp)->reachable =
1363 ND_COMPUTE_RTIME(ND.basereachable);
1365 if (ND.retrans != 0)
1366 ND_IFINFO(ifp)->retrans = ND.retrans;
1368 ND_IFINFO(ifp)->chlim = ND.chlim;
1370 case SIOCSIFINFO_FLAGS:
1373 struct in6_ifaddr *ia;
1375 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1376 !(ND.flags & ND6_IFF_IFDISABLED)) {
1377 /* ifdisabled 1->0 transision */
1380 * If the interface is marked as ND6_IFF_IFDISABLED and
1381 * has an link-local address with IN6_IFF_DUPLICATED,
1382 * do not clear ND6_IFF_IFDISABLED.
1383 * See RFC 4862, Section 5.4.5.
1385 int duplicated_linklocal = 0;
1388 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1389 if (ifa->ifa_addr->sa_family != AF_INET6)
1391 ia = (struct in6_ifaddr *)ifa;
1392 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1393 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1394 duplicated_linklocal = 1;
1398 IF_ADDR_RUNLOCK(ifp);
1400 if (duplicated_linklocal) {
1401 ND.flags |= ND6_IFF_IFDISABLED;
1402 log(LOG_ERR, "Cannot enable an interface"
1403 " with a link-local address marked"
1406 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1407 if (ifp->if_flags & IFF_UP)
1410 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1411 (ND.flags & ND6_IFF_IFDISABLED)) {
1412 /* ifdisabled 0->1 transision */
1413 /* Mark all IPv6 address as tentative. */
1415 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1417 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1418 if (ifa->ifa_addr->sa_family != AF_INET6)
1420 ia = (struct in6_ifaddr *)ifa;
1421 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1423 IF_ADDR_RUNLOCK(ifp);
1426 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1427 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1428 /* auto_linklocal 0->1 transision */
1430 /* If no link-local address on ifp, configure */
1431 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1432 in6_ifattach(ifp, NULL);
1433 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1434 ifp->if_flags & IFF_UP) {
1436 * When the IF already has
1437 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1438 * address is assigned, and IFF_UP, try to
1441 int haslinklocal = 0;
1444 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1445 if (ifa->ifa_addr->sa_family != AF_INET6)
1447 ia = (struct in6_ifaddr *)ifa;
1448 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1453 IF_ADDR_RUNLOCK(ifp);
1455 in6_ifattach(ifp, NULL);
1459 ND_IFINFO(ifp)->flags = ND.flags;
1462 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1463 /* sync kernel routing table with the default router list */
1467 case SIOCSPFXFLUSH_IN6:
1469 /* flush all the prefix advertised by routers */
1470 struct nd_prefix *pr, *next;
1473 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1474 struct in6_ifaddr *ia, *ia_next;
1476 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1479 /* do we really have to remove addresses as well? */
1480 /* XXXRW: in6_ifaddrhead locking. */
1481 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1483 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1486 if (ia->ia6_ndpr == pr)
1487 in6_purgeaddr(&ia->ia_ifa);
1494 case SIOCSRTRFLUSH_IN6:
1496 /* flush all the default routers */
1497 struct nd_defrouter *dr, *next;
1501 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1508 case SIOCGNBRINFO_IN6:
1511 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1513 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1516 IF_AFDATA_RLOCK(ifp);
1517 ln = nd6_lookup(&nb_addr, 0, ifp);
1518 IF_AFDATA_RUNLOCK(ifp);
1524 nbi->state = ln->ln_state;
1525 nbi->asked = ln->la_asked;
1526 nbi->isrouter = ln->ln_router;
1527 nbi->expire = ln->la_expire;
1531 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1532 ndif->ifindex = V_nd6_defifindex;
1534 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1535 return (nd6_setdefaultiface(ndif->ifindex));
1541 * Create neighbor cache entry and cache link-layer address,
1542 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1545 * code - type dependent information
1548 * The caller of this function already acquired the ndp
1549 * cache table lock because the cache entry is returned.
1552 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1553 int lladdrlen, int type, int code)
1555 struct llentry *ln = NULL;
1562 uint16_t router = 0;
1563 struct sockaddr_in6 sin6;
1564 struct mbuf *chain = NULL;
1565 int static_route = 0;
1567 IF_AFDATA_UNLOCK_ASSERT(ifp);
1569 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1570 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1572 /* nothing must be updated for unspecified address */
1573 if (IN6_IS_ADDR_UNSPECIFIED(from))
1577 * Validation about ifp->if_addrlen and lladdrlen must be done in
1580 * XXX If the link does not have link-layer adderss, what should
1581 * we do? (ifp->if_addrlen == 0)
1582 * Spec says nothing in sections for RA, RS and NA. There's small
1583 * description on it in NS section (RFC 2461 7.2.3).
1585 flags = lladdr ? ND6_EXCLUSIVE : 0;
1586 IF_AFDATA_RLOCK(ifp);
1587 ln = nd6_lookup(from, flags, ifp);
1588 IF_AFDATA_RUNLOCK(ifp);
1590 flags |= ND6_EXCLUSIVE;
1591 IF_AFDATA_LOCK(ifp);
1592 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1593 IF_AFDATA_UNLOCK(ifp);
1596 /* do nothing if static ndp is set */
1597 if (ln->la_flags & LLE_STATIC) {
1606 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1607 if (olladdr && lladdr) {
1608 llchange = bcmp(lladdr, &ln->ll_addr,
1614 * newentry olladdr lladdr llchange (*=record)
1617 * 0 n y -- (3) * STALE
1619 * 0 y y y (5) * STALE
1620 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1621 * 1 -- y -- (7) * STALE
1624 if (lladdr) { /* (3-5) and (7) */
1626 * Record source link-layer address
1627 * XXX is it dependent to ifp->if_type?
1629 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1630 ln->la_flags |= LLE_VALID;
1631 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1635 if ((!olladdr && lladdr != NULL) || /* (3) */
1636 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1638 newstate = ND6_LLINFO_STALE;
1639 } else /* (1-2,4) */
1643 if (lladdr == NULL) /* (6) */
1644 newstate = ND6_LLINFO_NOSTATE;
1646 newstate = ND6_LLINFO_STALE;
1651 * Update the state of the neighbor cache.
1653 ln->ln_state = newstate;
1655 if (ln->ln_state == ND6_LLINFO_STALE) {
1657 * XXX: since nd6_output() below will cause
1658 * state tansition to DELAY and reset the timer,
1659 * we must set the timer now, although it is actually
1662 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1665 struct mbuf *m_hold, *m_hold_next;
1668 * reset the la_hold in advance, to explicitly
1669 * prevent a la_hold lookup in nd6_output()
1670 * (wouldn't happen, though...)
1672 for (m_hold = ln->la_hold, ln->la_hold = NULL;
1673 m_hold; m_hold = m_hold_next) {
1674 m_hold_next = m_hold->m_nextpkt;
1675 m_hold->m_nextpkt = NULL;
1678 * we assume ifp is not a p2p here, so
1679 * just set the 2nd argument as the
1682 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
1685 * If we have mbufs in the chain we need to do
1686 * deferred transmit. Copy the address from the
1687 * llentry before dropping the lock down below.
1690 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
1692 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1693 /* probe right away */
1694 nd6_llinfo_settimer_locked((void *)ln, 0);
1699 * ICMP6 type dependent behavior.
1701 * NS: clear IsRouter if new entry
1702 * RS: clear IsRouter
1703 * RA: set IsRouter if there's lladdr
1704 * redir: clear IsRouter if new entry
1707 * The spec says that we must set IsRouter in the following cases:
1708 * - If lladdr exist, set IsRouter. This means (1-5).
1709 * - If it is old entry (!newentry), set IsRouter. This means (7).
1710 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1711 * A quetion arises for (1) case. (1) case has no lladdr in the
1712 * neighbor cache, this is similar to (6).
1713 * This case is rare but we figured that we MUST NOT set IsRouter.
1715 * newentry olladdr lladdr llchange NS RS RA redir
1717 * 0 n n -- (1) c ? s
1718 * 0 y n -- (2) c s s
1719 * 0 n y -- (3) c s s
1722 * 1 -- n -- (6) c c c s
1723 * 1 -- y -- (7) c c s c s
1727 switch (type & 0xff) {
1728 case ND_NEIGHBOR_SOLICIT:
1730 * New entry must have is_router flag cleared.
1732 if (is_newentry) /* (6-7) */
1737 * If the icmp is a redirect to a better router, always set the
1738 * is_router flag. Otherwise, if the entry is newly created,
1739 * clear the flag. [RFC 2461, sec 8.3]
1741 if (code == ND_REDIRECT_ROUTER)
1743 else if (is_newentry) /* (6-7) */
1746 case ND_ROUTER_SOLICIT:
1748 * is_router flag must always be cleared.
1752 case ND_ROUTER_ADVERT:
1754 * Mark an entry with lladdr as a router.
1756 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1757 (is_newentry && lladdr)) { /* (7) */
1764 static_route = (ln->la_flags & LLE_STATIC);
1765 router = ln->ln_router;
1767 if (flags & ND6_EXCLUSIVE)
1775 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
1778 * When the link-layer address of a router changes, select the
1779 * best router again. In particular, when the neighbor entry is newly
1780 * created, it might affect the selection policy.
1781 * Question: can we restrict the first condition to the "is_newentry"
1783 * XXX: when we hear an RA from a new router with the link-layer
1784 * address option, defrouter_select() is called twice, since
1785 * defrtrlist_update called the function as well. However, I believe
1786 * we can compromise the overhead, since it only happens the first
1788 * XXX: although defrouter_select() should not have a bad effect
1789 * for those are not autoconfigured hosts, we explicitly avoid such
1792 if (do_update && router &&
1793 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1795 * guaranteed recursion
1803 if (flags & ND6_EXCLUSIVE)
1814 nd6_slowtimo(void *arg)
1816 CURVNET_SET((struct vnet *) arg);
1817 struct nd_ifinfo *nd6if;
1820 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1821 nd6_slowtimo, curvnet);
1822 IFNET_RLOCK_NOSLEEP();
1823 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1824 nd6if = ND_IFINFO(ifp);
1825 if (nd6if->basereachable && /* already initialized */
1826 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1828 * Since reachable time rarely changes by router
1829 * advertisements, we SHOULD insure that a new random
1830 * value gets recomputed at least once every few hours.
1833 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1834 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1837 IFNET_RUNLOCK_NOSLEEP();
1842 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1843 struct sockaddr_in6 *dst, struct rtentry *rt0)
1846 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
1851 * Note that I'm not enforcing any global serialization
1852 * lle state or asked changes here as the logic is too
1853 * complicated to avoid having to always acquire an exclusive
1858 #define senderr(e) { error = (e); goto bad;}
1861 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1862 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
1863 struct mbuf **chain)
1865 struct mbuf *m = m0;
1867 struct llentry *ln = lle;
1868 struct ip6_hdr *ip6;
1876 LLE_WLOCK_ASSERT(lle);
1878 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
1881 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1884 if (nd6_need_cache(ifp) == 0)
1888 * next hop determination. This routine is derived from ether_output.
1892 * Address resolution or Neighbor Unreachability Detection
1894 * At this point, the destination of the packet must be a unicast
1895 * or an anycast address(i.e. not a multicast).
1898 flags = (lle != NULL) ? LLE_EXCLUSIVE : 0;
1901 IF_AFDATA_RLOCK(ifp);
1902 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
1903 IF_AFDATA_RUNLOCK(ifp);
1904 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1906 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1907 * the condition below is not very efficient. But we believe
1908 * it is tolerable, because this should be a rare case.
1910 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
1911 IF_AFDATA_LOCK(ifp);
1912 ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
1913 IF_AFDATA_UNLOCK(ifp);
1917 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1918 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1919 char ip6buf[INET6_ADDRSTRLEN];
1921 "nd6_output: can't allocate llinfo for %s "
1923 ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
1924 senderr(EIO); /* XXX: good error? */
1926 goto sendpkt; /* send anyway */
1929 /* We don't have to do link-layer address resolution on a p2p link. */
1930 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1931 ln->ln_state < ND6_LLINFO_REACHABLE) {
1932 if ((flags & LLE_EXCLUSIVE) == 0) {
1933 flags |= LLE_EXCLUSIVE;
1937 ln->ln_state = ND6_LLINFO_STALE;
1938 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1942 * The first time we send a packet to a neighbor whose entry is
1943 * STALE, we have to change the state to DELAY and a sets a timer to
1944 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1945 * neighbor unreachability detection on expiration.
1948 if (ln->ln_state == ND6_LLINFO_STALE) {
1949 if ((flags & LLE_EXCLUSIVE) == 0) {
1950 flags |= LLE_EXCLUSIVE;
1955 ln->ln_state = ND6_LLINFO_DELAY;
1956 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1960 * If the neighbor cache entry has a state other than INCOMPLETE
1961 * (i.e. its link-layer address is already resolved), just
1964 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1968 * There is a neighbor cache entry, but no ethernet address
1969 * response yet. Append this latest packet to the end of the
1970 * packet queue in the mbuf, unless the number of the packet
1971 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1972 * the oldest packet in the queue will be removed.
1974 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1975 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1977 if ((flags & LLE_EXCLUSIVE) == 0) {
1978 flags |= LLE_EXCLUSIVE;
1983 LLE_WLOCK_ASSERT(ln);
1986 struct mbuf *m_hold;
1990 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
1992 if (m_hold->m_nextpkt == NULL) {
1993 m_hold->m_nextpkt = m;
1997 while (i >= V_nd6_maxqueuelen) {
1998 m_hold = ln->la_hold;
1999 ln->la_hold = ln->la_hold->m_nextpkt;
2008 * If there has been no NS for the neighbor after entering the
2009 * INCOMPLETE state, send the first solicitation.
2011 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
2014 nd6_llinfo_settimer_locked(ln,
2015 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2017 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2018 if (lle != NULL && ln == lle)
2021 } else if (lle == NULL || ln != lle) {
2023 * We did the lookup (no lle arg) so we
2024 * need to do the unlock here.
2032 /* discard the packet if IPv6 operation is disabled on the interface */
2033 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2034 error = ENETDOWN; /* better error? */
2038 * ln is valid and the caller did not pass in
2041 if ((ln != NULL) && (lle == NULL)) {
2042 if (flags & LLE_EXCLUSIVE)
2049 mac_netinet6_nd6_send(ifp, m);
2053 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
2054 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
2055 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
2056 * to be diverted to user space. When re-injected into the kernel,
2057 * send_output() will directly dispatch them to the outgoing interface.
2059 if (send_sendso_input_hook != NULL) {
2060 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
2062 ip6 = mtod(m, struct ip6_hdr *);
2063 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2064 /* Use the SEND socket */
2065 error = send_sendso_input_hook(m, ifp, SND_OUT,
2067 /* -1 == no app on SEND socket */
2068 if (error == 0 || error != -1)
2074 * We were passed in a pointer to an lle with the lock held
2075 * this means that we can't call if_output as we will
2076 * recurse on the lle lock - so what we do is we create
2077 * a list of mbufs to send and transmit them in the caller
2078 * after the lock is dropped
2087 * append mbuf to end of deferred chain
2090 while (mb->m_nextpkt != NULL)
2096 /* Reset layer specific mbuf flags to avoid confusing lower layers. */
2097 m->m_flags &= ~(M_PROTOFLAGS);
2098 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
2099 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
2102 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
2107 * ln is valid and the caller did not pass in
2110 if ((ln != NULL) && (lle == NULL)) {
2111 if (flags & LLE_EXCLUSIVE)
2124 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2125 struct sockaddr_in6 *dst, struct route *ro)
2127 struct mbuf *m, *m_head;
2128 struct ifnet *outifp;
2132 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2139 m_head = m_head->m_nextpkt;
2140 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);
2145 * note that intermediate errors are blindly ignored - but this is
2146 * the same convention as used with nd6_output when called by
2154 nd6_need_cache(struct ifnet *ifp)
2157 * XXX: we currently do not make neighbor cache on any interface
2158 * other than ARCnet, Ethernet, FDDI and GIF.
2161 * - unidirectional tunnels needs no ND
2163 switch (ifp->if_type) {
2171 #ifdef IFT_IEEE80211
2177 case IFT_INFINIBAND:
2179 case IFT_PROPVIRTUAL:
2187 * the callers of this function need to be re-worked to drop
2188 * the lle lock, drop here for now
2191 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2192 struct sockaddr *dst, u_char *desten, struct llentry **lle)
2197 IF_AFDATA_UNLOCK_ASSERT(ifp);
2198 if (m != NULL && m->m_flags & M_MCAST) {
2201 switch (ifp->if_type) {
2207 #ifdef IFT_IEEE80211
2212 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2217 * netbsd can use if_broadcastaddr, but we don't do so
2218 * to reduce # of ifdef.
2220 for (i = 0; i < ifp->if_addrlen; i++)
2228 return (EAFNOSUPPORT);
2234 * the entry should have been created in nd6_store_lladdr
2236 IF_AFDATA_RLOCK(ifp);
2237 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2238 IF_AFDATA_RUNLOCK(ifp);
2239 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2242 /* this could happen, if we could not allocate memory */
2247 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2251 * A *small* use after free race exists here
2257 clear_llinfo_pqueue(struct llentry *ln)
2259 struct mbuf *m_hold, *m_hold_next;
2261 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2262 m_hold_next = m_hold->m_nextpkt;
2270 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2271 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2273 SYSCTL_DECL(_net_inet6_icmp6);
2275 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2276 CTLFLAG_RD, nd6_sysctl_drlist, "");
2277 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2278 CTLFLAG_RD, nd6_sysctl_prlist, "");
2279 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2280 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2281 SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2282 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2285 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2287 struct in6_defrouter d;
2288 struct nd_defrouter *dr;
2294 bzero(&d, sizeof(d));
2295 d.rtaddr.sin6_family = AF_INET6;
2296 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2301 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2302 d.rtaddr.sin6_addr = dr->rtaddr;
2303 error = sa6_recoverscope(&d.rtaddr);
2306 d.flags = dr->flags;
2307 d.rtlifetime = dr->rtlifetime;
2308 d.expire = dr->expire;
2309 d.if_index = dr->ifp->if_index;
2310 error = SYSCTL_OUT(req, &d, sizeof(d));
2318 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2320 struct in6_prefix p;
2321 struct sockaddr_in6 s6;
2322 struct nd_prefix *pr;
2323 struct nd_pfxrouter *pfr;
2326 char ip6buf[INET6_ADDRSTRLEN];
2331 bzero(&p, sizeof(p));
2332 p.origin = PR_ORIG_RA;
2333 bzero(&s6, sizeof(s6));
2334 s6.sin6_family = AF_INET6;
2335 s6.sin6_len = sizeof(s6);
2340 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2341 p.prefix = pr->ndpr_prefix;
2342 if (sa6_recoverscope(&p.prefix)) {
2343 log(LOG_ERR, "scope error in prefix list (%s)\n",
2344 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2345 /* XXX: press on... */
2347 p.raflags = pr->ndpr_raf;
2348 p.prefixlen = pr->ndpr_plen;
2349 p.vltime = pr->ndpr_vltime;
2350 p.pltime = pr->ndpr_pltime;
2351 p.if_index = pr->ndpr_ifp->if_index;
2352 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2355 /* XXX: we assume time_t is signed. */
2357 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2358 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2359 p.expire = pr->ndpr_lastupdate +
2362 p.expire = maxexpire;
2364 p.refcnt = pr->ndpr_refcnt;
2365 p.flags = pr->ndpr_stateflags;
2367 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2369 error = SYSCTL_OUT(req, &p, sizeof(p));
2372 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2373 s6.sin6_addr = pfr->router->rtaddr;
2374 if (sa6_recoverscope(&s6))
2376 "scope error in prefix list (%s)\n",
2377 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2378 error = SYSCTL_OUT(req, &s6, sizeof(s6));