2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/callout.h>
41 #include <sys/malloc.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
46 #include <sys/kernel.h>
47 #include <sys/protosw.h>
48 #include <sys/errno.h>
49 #include <sys/syslog.h>
51 #include <sys/rwlock.h>
52 #include <sys/queue.h>
53 #include <sys/sysctl.h>
56 #include <net/if_arc.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/iso88025.h>
61 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <net/if_llatbl.h>
66 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le))
67 #include <netinet/if_ether.h>
68 #include <netinet6/in6_var.h>
69 #include <netinet/ip6.h>
70 #include <netinet6/ip6_var.h>
71 #include <netinet6/scope6_var.h>
72 #include <netinet6/nd6.h>
73 #include <netinet/icmp6.h>
75 #include <sys/limits.h>
77 #include <security/mac/mac_framework.h>
79 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
80 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
82 #define SIN6(s) ((struct sockaddr_in6 *)s)
85 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
86 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
87 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
88 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
89 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
91 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
94 /* preventing too many loops in ND option parsing */
95 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
97 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
99 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
101 #define V_nd6_maxndopt VNET(nd6_maxndopt)
102 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
105 VNET_DEFINE(int, nd6_debug) = 1;
107 VNET_DEFINE(int, nd6_debug) = 0;
112 static int nd6_inuse, nd6_allocated;
115 VNET_DEFINE(struct nd_drhead, nd_defrouter);
116 VNET_DEFINE(struct nd_prhead, nd_prefix);
118 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
119 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
121 static struct sockaddr_in6 all1_sa;
123 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
125 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
126 static void nd6_slowtimo(void *);
127 static int regen_tmpaddr(struct in6_ifaddr *);
128 static struct llentry *nd6_free(struct llentry *, int);
129 static void nd6_llinfo_timer(void *);
130 static void clear_llinfo_pqueue(struct llentry *);
132 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
133 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
135 VNET_DEFINE(struct callout, nd6_timer_ch);
142 LIST_INIT(&V_nd_prefix);
144 all1_sa.sin6_family = AF_INET6;
145 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
146 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
147 all1_sa.sin6_addr.s6_addr[i] = 0xff;
149 /* initialization of the default router list */
150 TAILQ_INIT(&V_nd_defrouter);
153 callout_init(&V_nd6_slowtimo_ch, 0);
154 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
155 nd6_slowtimo, curvnet);
163 callout_drain(&V_nd6_slowtimo_ch);
164 callout_drain(&V_nd6_timer_ch);
169 nd6_ifattach(struct ifnet *ifp)
171 struct nd_ifinfo *nd;
173 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
174 bzero(nd, sizeof(*nd));
178 nd->chlim = IPV6_DEFHLIM;
179 nd->basereachable = REACHABLE_TIME;
180 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
181 nd->retrans = RETRANS_TIMER;
183 * Note that the default value of ip6_accept_rtadv is 0, which means
184 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
187 nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
189 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
190 nd6_setmtu0(ifp, nd);
196 nd6_ifdetach(struct nd_ifinfo *nd)
203 * Reset ND level link MTU. This function is called when the physical MTU
204 * changes, which means we might have to adjust the ND level MTU.
207 nd6_setmtu(struct ifnet *ifp)
210 nd6_setmtu0(ifp, ND_IFINFO(ifp));
213 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
215 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
219 omaxmtu = ndi->maxmtu;
221 switch (ifp->if_type) {
223 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
226 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
229 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
232 ndi->maxmtu = ifp->if_mtu;
237 * Decreasing the interface MTU under IPV6 minimum MTU may cause
238 * undesirable situation. We thus notify the operator of the change
239 * explicitly. The check for omaxmtu is necessary to restrict the
240 * log to the case of changing the MTU, not initializing it.
242 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
243 log(LOG_NOTICE, "nd6_setmtu0: "
244 "new link MTU on %s (%lu) is too small for IPv6\n",
245 if_name(ifp), (unsigned long)ndi->maxmtu);
248 if (ndi->maxmtu > V_in6_maxmtu)
249 in6_setmaxmtu(); /* check all interfaces just in case */
254 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
257 bzero(ndopts, sizeof(*ndopts));
258 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
260 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
263 ndopts->nd_opts_done = 1;
264 ndopts->nd_opts_search = NULL;
269 * Take one ND option.
272 nd6_option(union nd_opts *ndopts)
274 struct nd_opt_hdr *nd_opt;
278 panic("ndopts == NULL in nd6_option");
279 if (ndopts->nd_opts_last == NULL)
280 panic("uninitialized ndopts in nd6_option");
281 if (ndopts->nd_opts_search == NULL)
283 if (ndopts->nd_opts_done)
286 nd_opt = ndopts->nd_opts_search;
288 /* make sure nd_opt_len is inside the buffer */
289 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
290 bzero(ndopts, sizeof(*ndopts));
294 olen = nd_opt->nd_opt_len << 3;
297 * Message validation requires that all included
298 * options have a length that is greater than zero.
300 bzero(ndopts, sizeof(*ndopts));
304 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
305 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
306 /* option overruns the end of buffer, invalid */
307 bzero(ndopts, sizeof(*ndopts));
309 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
310 /* reached the end of options chain */
311 ndopts->nd_opts_done = 1;
312 ndopts->nd_opts_search = NULL;
318 * Parse multiple ND options.
319 * This function is much easier to use, for ND routines that do not need
320 * multiple options of the same type.
323 nd6_options(union nd_opts *ndopts)
325 struct nd_opt_hdr *nd_opt;
329 panic("ndopts == NULL in nd6_options");
330 if (ndopts->nd_opts_last == NULL)
331 panic("uninitialized ndopts in nd6_options");
332 if (ndopts->nd_opts_search == NULL)
336 nd_opt = nd6_option(ndopts);
337 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
339 * Message validation requires that all included
340 * options have a length that is greater than zero.
342 ICMP6STAT_INC(icp6s_nd_badopt);
343 bzero(ndopts, sizeof(*ndopts));
350 switch (nd_opt->nd_opt_type) {
351 case ND_OPT_SOURCE_LINKADDR:
352 case ND_OPT_TARGET_LINKADDR:
354 case ND_OPT_REDIRECTED_HEADER:
355 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
357 "duplicated ND6 option found (type=%d)\n",
358 nd_opt->nd_opt_type));
361 ndopts->nd_opt_array[nd_opt->nd_opt_type]
365 case ND_OPT_PREFIX_INFORMATION:
366 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
367 ndopts->nd_opt_array[nd_opt->nd_opt_type]
370 ndopts->nd_opts_pi_end =
371 (struct nd_opt_prefix_info *)nd_opt;
373 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
374 case ND_OPT_RDNSS: /* RFC 6106 */
375 case ND_OPT_DNSSL: /* RFC 6106 */
377 * Silently ignore options we know and do not care about
383 * Unknown options must be silently ignored,
384 * to accomodate future extension to the protocol.
387 "nd6_options: unsupported option %d - "
388 "option ignored\n", nd_opt->nd_opt_type));
393 if (i > V_nd6_maxndopt) {
394 ICMP6STAT_INC(icp6s_nd_toomanyopt);
395 nd6log((LOG_INFO, "too many loop in nd opt\n"));
399 if (ndopts->nd_opts_done)
407 * ND6 timer routine to handle ND6 entries
410 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
414 LLE_WLOCK_ASSERT(ln);
419 canceled = callout_stop(&ln->ln_timer_ch);
421 ln->la_expire = time_second + tick / hz;
423 if (tick > INT_MAX) {
424 ln->ln_ntick = tick - INT_MAX;
425 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
426 nd6_llinfo_timer, ln);
429 canceled = callout_reset(&ln->ln_timer_ch, tick,
430 nd6_llinfo_timer, ln);
438 nd6_llinfo_settimer(struct llentry *ln, long tick)
442 nd6_llinfo_settimer_locked(ln, tick);
447 nd6_llinfo_timer(void *arg)
450 struct in6_addr *dst;
452 struct nd_ifinfo *ndi = NULL;
454 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
455 ln = (struct llentry *)arg;
456 LLE_WLOCK_ASSERT(ln);
457 ifp = ln->lle_tbl->llt_ifp;
459 CURVNET_SET(ifp->if_vnet);
461 if (ln->ln_ntick > 0) {
462 if (ln->ln_ntick > INT_MAX) {
463 ln->ln_ntick -= INT_MAX;
464 nd6_llinfo_settimer_locked(ln, INT_MAX);
467 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
472 ndi = ND_IFINFO(ifp);
473 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
474 if (ln->la_flags & LLE_STATIC) {
478 if (ln->la_flags & LLE_DELETED) {
479 (void)nd6_free(ln, 0);
484 switch (ln->ln_state) {
485 case ND6_LLINFO_INCOMPLETE:
486 if (ln->la_asked < V_nd6_mmaxtries) {
488 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
490 nd6_ns_output(ifp, NULL, dst, ln, 0);
493 struct mbuf *m = ln->la_hold;
498 * assuming every packet in la_hold has the
499 * same IP header. Send error after unlock.
504 clear_llinfo_pqueue(ln);
506 (void)nd6_free(ln, 0);
509 icmp6_error2(m, ICMP6_DST_UNREACH,
510 ICMP6_DST_UNREACH_ADDR, 0, ifp);
513 case ND6_LLINFO_REACHABLE:
514 if (!ND6_LLINFO_PERMANENT(ln)) {
515 ln->ln_state = ND6_LLINFO_STALE;
516 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
520 case ND6_LLINFO_STALE:
521 /* Garbage Collection(RFC 2461 5.3) */
522 if (!ND6_LLINFO_PERMANENT(ln)) {
523 (void)nd6_free(ln, 1);
528 case ND6_LLINFO_DELAY:
529 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
532 ln->ln_state = ND6_LLINFO_PROBE;
533 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
535 nd6_ns_output(ifp, dst, dst, ln, 0);
538 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
539 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
542 case ND6_LLINFO_PROBE:
543 if (ln->la_asked < V_nd6_umaxtries) {
545 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
547 nd6_ns_output(ifp, dst, dst, ln, 0);
550 (void)nd6_free(ln, 0);
555 panic("%s: paths in a dark night can be confusing: %d",
556 __func__, ln->ln_state);
566 * ND6 timer routine to expire default route list and prefix list
571 CURVNET_SET((struct vnet *) arg);
573 struct nd_defrouter *dr, *ndr;
574 struct nd_prefix *pr, *npr;
575 struct in6_ifaddr *ia6, *nia6;
577 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
580 /* expire default router list */
582 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
583 if (dr->expire && dr->expire < time_second)
588 * expire interface addresses.
589 * in the past the loop was inside prefix expiry processing.
590 * However, from a stricter speci-confrmance standpoint, we should
591 * rather separate address lifetimes and prefix lifetimes.
593 * XXXRW: in6_ifaddrhead locking.
596 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
597 /* check address lifetime */
598 if (IFA6_IS_INVALID(ia6)) {
602 * If the expiring address is temporary, try
603 * regenerating a new one. This would be useful when
604 * we suspended a laptop PC, then turned it on after a
605 * period that could invalidate all temporary
606 * addresses. Although we may have to restart the
607 * loop (see below), it must be after purging the
608 * address. Otherwise, we'd see an infinite loop of
611 if (V_ip6_use_tempaddr &&
612 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
613 if (regen_tmpaddr(ia6) == 0)
617 in6_purgeaddr(&ia6->ia_ifa);
620 goto addrloop; /* XXX: see below */
621 } else if (IFA6_IS_DEPRECATED(ia6)) {
622 int oldflags = ia6->ia6_flags;
624 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
627 * If a temporary address has just become deprecated,
628 * regenerate a new one if possible.
630 if (V_ip6_use_tempaddr &&
631 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
632 (oldflags & IN6_IFF_DEPRECATED) == 0) {
634 if (regen_tmpaddr(ia6) == 0) {
636 * A new temporary address is
638 * XXX: this means the address chain
639 * has changed while we are still in
640 * the loop. Although the change
641 * would not cause disaster (because
642 * it's not a deletion, but an
643 * addition,) we'd rather restart the
644 * loop just for safety. Or does this
645 * significantly reduce performance??
652 * A new RA might have made a deprecated address
655 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
659 /* expire prefix list */
660 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
662 * check prefix lifetime.
663 * since pltime is just for autoconf, pltime processing for
664 * prefix is not necessary.
666 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
667 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
670 * address expiration and prefix expiration are
671 * separate. NEVER perform in6_purgeaddr here.
681 * ia6 - deprecated/invalidated temporary address
684 regen_tmpaddr(struct in6_ifaddr *ia6)
688 struct in6_ifaddr *public_ifa6 = NULL;
690 ifp = ia6->ia_ifa.ifa_ifp;
692 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
693 struct in6_ifaddr *it6;
695 if (ifa->ifa_addr->sa_family != AF_INET6)
698 it6 = (struct in6_ifaddr *)ifa;
700 /* ignore no autoconf addresses. */
701 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
704 /* ignore autoconf addresses with different prefixes. */
705 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
709 * Now we are looking at an autoconf address with the same
710 * prefix as ours. If the address is temporary and is still
711 * preferred, do not create another one. It would be rare, but
712 * could happen, for example, when we resume a laptop PC after
715 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
716 !IFA6_IS_DEPRECATED(it6)) {
722 * This is a public autoconf address that has the same prefix
723 * as ours. If it is preferred, keep it. We can't break the
724 * loop here, because there may be a still-preferred temporary
725 * address with the prefix.
727 if (!IFA6_IS_DEPRECATED(it6))
730 if (public_ifa6 != NULL)
731 ifa_ref(&public_ifa6->ia_ifa);
733 IF_ADDR_RUNLOCK(ifp);
735 if (public_ifa6 != NULL) {
738 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
739 ifa_free(&public_ifa6->ia_ifa);
740 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
741 " tmp addr,errno=%d\n", e);
744 ifa_free(&public_ifa6->ia_ifa);
752 * Nuke neighbor cache/prefix/default router management table, right before
756 nd6_purge(struct ifnet *ifp)
758 struct nd_defrouter *dr, *ndr;
759 struct nd_prefix *pr, *npr;
762 * Nuke default router list entries toward ifp.
763 * We defer removal of default router list entries that is installed
764 * in the routing table, in order to keep additional side effects as
767 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
775 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
783 /* Nuke prefix list entries toward ifp */
784 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
785 if (pr->ndpr_ifp == ifp) {
787 * Because if_detach() does *not* release prefixes
788 * while purging addresses the reference count will
789 * still be above zero. We therefore reset it to
790 * make sure that the prefix really gets purged.
795 * Previously, pr->ndpr_addr is removed as well,
796 * but I strongly believe we don't have to do it.
797 * nd6_purge() is only called from in6_ifdetach(),
798 * which removes all the associated interface addresses
800 * (jinmei@kame.net 20010129)
806 /* cancel default outgoing interface setting */
807 if (V_nd6_defifindex == ifp->if_index)
808 nd6_setdefaultiface(0);
810 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */
811 /* refresh default router list
820 * We do not nuke the neighbor cache entries here any more
821 * because the neighbor cache is kept in if_afdata[AF_INET6].
822 * nd6_purge() is invoked by in6_ifdetach() which is called
823 * from if_detach() where everything gets purged. So let
824 * in6_domifdetach() do the actual L2 table purging work.
829 * the caller acquires and releases the lock on the lltbls
830 * Returns the llentry locked
833 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
835 struct sockaddr_in6 sin6;
839 bzero(&sin6, sizeof(sin6));
840 sin6.sin6_len = sizeof(struct sockaddr_in6);
841 sin6.sin6_family = AF_INET6;
842 sin6.sin6_addr = *addr6;
844 IF_AFDATA_LOCK_ASSERT(ifp);
847 if (flags & ND6_CREATE)
848 llflags |= LLE_CREATE;
849 if (flags & ND6_EXCLUSIVE)
850 llflags |= LLE_EXCLUSIVE;
852 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
853 if ((ln != NULL) && (llflags & LLE_CREATE))
854 ln->ln_state = ND6_LLINFO_NOSTATE;
860 * Test whether a given IPv6 address is a neighbor or not, ignoring
861 * the actual neighbor cache. The neighbor cache is ignored in order
862 * to not reenter the routing code from within itself.
865 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
867 struct nd_prefix *pr;
868 struct ifaddr *dstaddr;
871 * A link-local address is always a neighbor.
872 * XXX: a link does not necessarily specify a single interface.
874 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
875 struct sockaddr_in6 sin6_copy;
879 * We need sin6_copy since sa6_recoverscope() may modify the
883 if (sa6_recoverscope(&sin6_copy))
884 return (0); /* XXX: should be impossible */
885 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
887 if (sin6_copy.sin6_scope_id == zone)
894 * If the address matches one of our addresses,
895 * it should be a neighbor.
896 * If the address matches one of our on-link prefixes, it should be a
899 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
900 if (pr->ndpr_ifp != ifp)
903 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
906 /* Always use the default FIB here. */
907 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
908 0, 0, RT_DEFAULT_FIB);
912 * This is the case where multiple interfaces
913 * have the same prefix, but only one is installed
914 * into the routing table and that prefix entry
915 * is not the one being examined here. In the case
916 * where RADIX_MPATH is enabled, multiple route
917 * entries (of the same rt_key value) will be
918 * installed because the interface addresses all
921 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
922 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
929 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
930 &addr->sin6_addr, &pr->ndpr_mask))
935 * If the address is assigned on the node of the other side of
936 * a p2p interface, the address should be a neighbor.
938 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
939 if (dstaddr != NULL) {
940 if (dstaddr->ifa_ifp == ifp) {
948 * If the default router list is empty, all addresses are regarded
949 * as on-link, and thus, as a neighbor.
950 * XXX: we restrict the condition to hosts, because routers usually do
951 * not have the "default router list".
953 if (!V_ip6_forwarding && TAILQ_EMPTY(&V_nd_defrouter) &&
954 V_nd6_defifindex == ifp->if_index) {
963 * Detect if a given IPv6 address identifies a neighbor on a given link.
964 * XXX: should take care of the destination of a p2p link?
967 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
972 IF_AFDATA_UNLOCK_ASSERT(ifp);
973 if (nd6_is_new_addr_neighbor(addr, ifp))
977 * Even if the address matches none of our addresses, it might be
978 * in the neighbor cache.
980 IF_AFDATA_RLOCK(ifp);
981 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
985 IF_AFDATA_RUNLOCK(ifp);
990 * Free an nd6 llinfo entry.
991 * Since the function would cause significant changes in the kernel, DO NOT
992 * make it global, unless you have a strong reason for the change, and are sure
993 * that the change is safe.
995 static struct llentry *
996 nd6_free(struct llentry *ln, int gc)
998 struct llentry *next;
999 struct nd_defrouter *dr;
1002 LLE_WLOCK_ASSERT(ln);
1005 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1006 * even though it is not harmful, it was not really necessary.
1010 nd6_llinfo_settimer_locked(ln, -1);
1012 ifp = ln->lle_tbl->llt_ifp;
1014 if (!V_ip6_forwarding) {
1016 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1018 if (dr != NULL && dr->expire &&
1019 ln->ln_state == ND6_LLINFO_STALE && gc) {
1021 * If the reason for the deletion is just garbage
1022 * collection, and the neighbor is an active default
1023 * router, do not delete it. Instead, reset the GC
1024 * timer using the router's lifetime.
1025 * Simply deleting the entry would affect default
1026 * router selection, which is not necessarily a good
1027 * thing, especially when we're using router preference
1029 * XXX: the check for ln_state would be redundant,
1030 * but we intentionally keep it just in case.
1032 if (dr->expire > time_second)
1033 nd6_llinfo_settimer_locked(ln,
1034 (dr->expire - time_second) * hz);
1036 nd6_llinfo_settimer_locked(ln,
1037 (long)V_nd6_gctimer * hz);
1039 next = LIST_NEXT(ln, lle_next);
1047 * Unreachablity of a router might affect the default
1048 * router selection and on-link detection of advertised
1053 * Temporarily fake the state to choose a new default
1054 * router and to perform on-link determination of
1055 * prefixes correctly.
1056 * Below the state will be set correctly,
1057 * or the entry itself will be deleted.
1059 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1062 if (ln->ln_router || dr) {
1065 * We need to unlock to avoid a LOR with rt6_flush() with the
1066 * rnh and for the calls to pfxlist_onlink_check() and
1067 * defrouter_select() in the block further down for calls
1068 * into nd6_lookup(). We still hold a ref.
1073 * rt6_flush must be called whether or not the neighbor
1074 * is in the Default Router List.
1075 * See a corresponding comment in nd6_na_input().
1077 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1082 * Since defrouter_select() does not affect the
1083 * on-link determination and MIP6 needs the check
1084 * before the default router selection, we perform
1087 pfxlist_onlink_check();
1090 * Refresh default router list.
1095 if (ln->ln_router || dr)
1100 * Before deleting the entry, remember the next entry as the
1101 * return value. We need this because pfxlist_onlink_check() above
1102 * might have freed other entries (particularly the old next entry) as
1103 * a side effect (XXX).
1105 next = LIST_NEXT(ln, lle_next);
1108 * Save to unlock. We still hold an extra reference and will not
1109 * free(9) in llentry_free() if someone else holds one as well.
1112 IF_AFDATA_LOCK(ifp);
1116 IF_AFDATA_UNLOCK(ifp);
1122 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1124 * XXX cost-effective methods?
1127 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1132 if ((dst6 == NULL) || (rt == NULL))
1136 IF_AFDATA_LOCK(ifp);
1137 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1138 IF_AFDATA_UNLOCK(ifp);
1142 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1146 * if we get upper-layer reachability confirmation many times,
1147 * it is possible we have false information.
1151 if (ln->ln_byhint > V_nd6_maxnudhint) {
1156 ln->ln_state = ND6_LLINFO_REACHABLE;
1157 if (!ND6_LLINFO_PERMANENT(ln)) {
1158 nd6_llinfo_settimer_locked(ln,
1159 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1167 * Rejuvenate this function for routing operations related
1171 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1173 struct sockaddr_in6 *gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1174 struct nd_defrouter *dr;
1175 struct ifnet *ifp = rt->rt_ifp;
1187 * Only indirect routes are interesting.
1189 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1192 * check for default route
1194 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1195 &SIN6(rt_key(rt))->sin6_addr)) {
1197 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1207 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1209 struct in6_drlist *drl = (struct in6_drlist *)data;
1210 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1211 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1212 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1213 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1214 struct nd_defrouter *dr;
1215 struct nd_prefix *pr;
1216 int i = 0, error = 0;
1220 case SIOCGDRLST_IN6:
1222 * obsolete API, use sysctl under net.inet6.icmp6
1224 bzero(drl, sizeof(*drl));
1226 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1229 drl->defrouter[i].rtaddr = dr->rtaddr;
1230 in6_clearscope(&drl->defrouter[i].rtaddr);
1232 drl->defrouter[i].flags = dr->flags;
1233 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1234 drl->defrouter[i].expire = dr->expire;
1235 drl->defrouter[i].if_index = dr->ifp->if_index;
1240 case SIOCGPRLST_IN6:
1242 * obsolete API, use sysctl under net.inet6.icmp6
1244 * XXX the structure in6_prlist was changed in backward-
1245 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1246 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1249 * XXX meaning of fields, especialy "raflags", is very
1250 * differnet between RA prefix list and RR/static prefix list.
1251 * how about separating ioctls into two?
1253 bzero(oprl, sizeof(*oprl));
1255 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1256 struct nd_pfxrouter *pfr;
1261 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1262 oprl->prefix[i].raflags = pr->ndpr_raf;
1263 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1264 oprl->prefix[i].vltime = pr->ndpr_vltime;
1265 oprl->prefix[i].pltime = pr->ndpr_pltime;
1266 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1267 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1268 oprl->prefix[i].expire = 0;
1272 /* XXX: we assume time_t is signed. */
1275 ((sizeof(maxexpire) * 8) - 1));
1276 if (pr->ndpr_vltime <
1277 maxexpire - pr->ndpr_lastupdate) {
1278 oprl->prefix[i].expire =
1279 pr->ndpr_lastupdate +
1282 oprl->prefix[i].expire = maxexpire;
1286 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1288 #define RTRADDR oprl->prefix[i].advrtr[j]
1289 RTRADDR = pfr->router->rtaddr;
1290 in6_clearscope(&RTRADDR);
1295 oprl->prefix[i].advrtrs = j;
1296 oprl->prefix[i].origin = PR_ORIG_RA;
1303 case OSIOCGIFINFO_IN6:
1305 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1306 bzero(&ND, sizeof(ND));
1307 ND.linkmtu = IN6_LINKMTU(ifp);
1308 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1309 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1310 ND.reachable = ND_IFINFO(ifp)->reachable;
1311 ND.retrans = ND_IFINFO(ifp)->retrans;
1312 ND.flags = ND_IFINFO(ifp)->flags;
1313 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1314 ND.chlim = ND_IFINFO(ifp)->chlim;
1316 case SIOCGIFINFO_IN6:
1317 ND = *ND_IFINFO(ifp);
1319 case SIOCSIFINFO_IN6:
1321 * used to change host variables from userland.
1322 * intented for a use on router to reflect RA configurations.
1324 /* 0 means 'unspecified' */
1325 if (ND.linkmtu != 0) {
1326 if (ND.linkmtu < IPV6_MMTU ||
1327 ND.linkmtu > IN6_LINKMTU(ifp)) {
1331 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1334 if (ND.basereachable != 0) {
1335 int obasereachable = ND_IFINFO(ifp)->basereachable;
1337 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1338 if (ND.basereachable != obasereachable)
1339 ND_IFINFO(ifp)->reachable =
1340 ND_COMPUTE_RTIME(ND.basereachable);
1342 if (ND.retrans != 0)
1343 ND_IFINFO(ifp)->retrans = ND.retrans;
1345 ND_IFINFO(ifp)->chlim = ND.chlim;
1347 case SIOCSIFINFO_FLAGS:
1348 ND_IFINFO(ifp)->flags = ND.flags;
1351 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1352 /* sync kernel routing table with the default router list */
1356 case SIOCSPFXFLUSH_IN6:
1358 /* flush all the prefix advertised by routers */
1359 struct nd_prefix *pr, *next;
1362 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1363 struct in6_ifaddr *ia, *ia_next;
1365 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1368 /* do we really have to remove addresses as well? */
1369 /* XXXRW: in6_ifaddrhead locking. */
1370 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1372 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1375 if (ia->ia6_ndpr == pr)
1376 in6_purgeaddr(&ia->ia_ifa);
1383 case SIOCSRTRFLUSH_IN6:
1385 /* flush all the default routers */
1386 struct nd_defrouter *dr, *next;
1390 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1397 case SIOCGNBRINFO_IN6:
1400 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1402 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1405 IF_AFDATA_RLOCK(ifp);
1406 ln = nd6_lookup(&nb_addr, 0, ifp);
1407 IF_AFDATA_RUNLOCK(ifp);
1413 nbi->state = ln->ln_state;
1414 nbi->asked = ln->la_asked;
1415 nbi->isrouter = ln->ln_router;
1416 nbi->expire = ln->la_expire;
1420 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1421 ndif->ifindex = V_nd6_defifindex;
1423 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1424 return (nd6_setdefaultiface(ndif->ifindex));
1430 * Create neighbor cache entry and cache link-layer address,
1431 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1434 * code - type dependent information
1437 * The caller of this function already acquired the ndp
1438 * cache table lock because the cache entry is returned.
1441 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1442 int lladdrlen, int type, int code)
1444 struct llentry *ln = NULL;
1451 uint16_t router = 0;
1452 struct sockaddr_in6 sin6;
1453 struct mbuf *chain = NULL;
1454 int static_route = 0;
1456 IF_AFDATA_UNLOCK_ASSERT(ifp);
1459 panic("ifp == NULL in nd6_cache_lladdr");
1461 panic("from == NULL in nd6_cache_lladdr");
1463 /* nothing must be updated for unspecified address */
1464 if (IN6_IS_ADDR_UNSPECIFIED(from))
1468 * Validation about ifp->if_addrlen and lladdrlen must be done in
1471 * XXX If the link does not have link-layer adderss, what should
1472 * we do? (ifp->if_addrlen == 0)
1473 * Spec says nothing in sections for RA, RS and NA. There's small
1474 * description on it in NS section (RFC 2461 7.2.3).
1476 flags = lladdr ? ND6_EXCLUSIVE : 0;
1477 IF_AFDATA_LOCK(ifp);
1478 ln = nd6_lookup(from, flags, ifp);
1481 flags |= ND6_EXCLUSIVE;
1482 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1483 IF_AFDATA_UNLOCK(ifp);
1486 IF_AFDATA_UNLOCK(ifp);
1487 /* do nothing if static ndp is set */
1488 if (ln->la_flags & LLE_STATIC) {
1497 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1498 if (olladdr && lladdr) {
1499 llchange = bcmp(lladdr, &ln->ll_addr,
1505 * newentry olladdr lladdr llchange (*=record)
1508 * 0 n y -- (3) * STALE
1510 * 0 y y y (5) * STALE
1511 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1512 * 1 -- y -- (7) * STALE
1515 if (lladdr) { /* (3-5) and (7) */
1517 * Record source link-layer address
1518 * XXX is it dependent to ifp->if_type?
1520 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1521 ln->la_flags |= LLE_VALID;
1525 if ((!olladdr && lladdr != NULL) || /* (3) */
1526 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1528 newstate = ND6_LLINFO_STALE;
1529 } else /* (1-2,4) */
1533 if (lladdr == NULL) /* (6) */
1534 newstate = ND6_LLINFO_NOSTATE;
1536 newstate = ND6_LLINFO_STALE;
1541 * Update the state of the neighbor cache.
1543 ln->ln_state = newstate;
1545 if (ln->ln_state == ND6_LLINFO_STALE) {
1547 * XXX: since nd6_output() below will cause
1548 * state tansition to DELAY and reset the timer,
1549 * we must set the timer now, although it is actually
1552 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1555 struct mbuf *m_hold, *m_hold_next;
1558 * reset the la_hold in advance, to explicitly
1559 * prevent a la_hold lookup in nd6_output()
1560 * (wouldn't happen, though...)
1562 for (m_hold = ln->la_hold, ln->la_hold = NULL;
1563 m_hold; m_hold = m_hold_next) {
1564 m_hold_next = m_hold->m_nextpkt;
1565 m_hold->m_nextpkt = NULL;
1568 * we assume ifp is not a p2p here, so
1569 * just set the 2nd argument as the
1572 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
1575 * If we have mbufs in the chain we need to do
1576 * deferred transmit. Copy the address from the
1577 * llentry before dropping the lock down below.
1580 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
1582 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1583 /* probe right away */
1584 nd6_llinfo_settimer_locked((void *)ln, 0);
1589 * ICMP6 type dependent behavior.
1591 * NS: clear IsRouter if new entry
1592 * RS: clear IsRouter
1593 * RA: set IsRouter if there's lladdr
1594 * redir: clear IsRouter if new entry
1597 * The spec says that we must set IsRouter in the following cases:
1598 * - If lladdr exist, set IsRouter. This means (1-5).
1599 * - If it is old entry (!newentry), set IsRouter. This means (7).
1600 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1601 * A quetion arises for (1) case. (1) case has no lladdr in the
1602 * neighbor cache, this is similar to (6).
1603 * This case is rare but we figured that we MUST NOT set IsRouter.
1605 * newentry olladdr lladdr llchange NS RS RA redir
1607 * 0 n n -- (1) c ? s
1608 * 0 y n -- (2) c s s
1609 * 0 n y -- (3) c s s
1612 * 1 -- n -- (6) c c c s
1613 * 1 -- y -- (7) c c s c s
1617 switch (type & 0xff) {
1618 case ND_NEIGHBOR_SOLICIT:
1620 * New entry must have is_router flag cleared.
1622 if (is_newentry) /* (6-7) */
1627 * If the icmp is a redirect to a better router, always set the
1628 * is_router flag. Otherwise, if the entry is newly created,
1629 * clear the flag. [RFC 2461, sec 8.3]
1631 if (code == ND_REDIRECT_ROUTER)
1633 else if (is_newentry) /* (6-7) */
1636 case ND_ROUTER_SOLICIT:
1638 * is_router flag must always be cleared.
1642 case ND_ROUTER_ADVERT:
1644 * Mark an entry with lladdr as a router.
1646 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1647 (is_newentry && lladdr)) { /* (7) */
1654 static_route = (ln->la_flags & LLE_STATIC);
1655 router = ln->ln_router;
1657 if (flags & ND6_EXCLUSIVE)
1665 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
1668 * When the link-layer address of a router changes, select the
1669 * best router again. In particular, when the neighbor entry is newly
1670 * created, it might affect the selection policy.
1671 * Question: can we restrict the first condition to the "is_newentry"
1673 * XXX: when we hear an RA from a new router with the link-layer
1674 * address option, defrouter_select() is called twice, since
1675 * defrtrlist_update called the function as well. However, I believe
1676 * we can compromise the overhead, since it only happens the first
1678 * XXX: although defrouter_select() should not have a bad effect
1679 * for those are not autoconfigured hosts, we explicitly avoid such
1682 if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) {
1684 * guaranteed recursion
1692 if (flags & ND6_EXCLUSIVE)
1703 nd6_slowtimo(void *arg)
1705 CURVNET_SET((struct vnet *) arg);
1706 struct nd_ifinfo *nd6if;
1709 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1710 nd6_slowtimo, curvnet);
1711 IFNET_RLOCK_NOSLEEP();
1712 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1713 nd6if = ND_IFINFO(ifp);
1714 if (nd6if->basereachable && /* already initialized */
1715 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1717 * Since reachable time rarely changes by router
1718 * advertisements, we SHOULD insure that a new random
1719 * value gets recomputed at least once every few hours.
1722 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1723 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1726 IFNET_RUNLOCK_NOSLEEP();
1731 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1732 struct sockaddr_in6 *dst, struct rtentry *rt0)
1735 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
1740 * Note that I'm not enforcing any global serialization
1741 * lle state or asked changes here as the logic is too
1742 * complicated to avoid having to always acquire an exclusive
1747 #define senderr(e) { error = (e); goto bad;}
1750 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
1751 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
1752 struct mbuf **chain)
1754 struct mbuf *m = m0;
1755 struct llentry *ln = lle;
1762 LLE_WLOCK_ASSERT(lle);
1764 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
1767 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1770 if (nd6_need_cache(ifp) == 0)
1774 * next hop determination. This routine is derived from ether_output.
1778 * Address resolution or Neighbor Unreachability Detection
1780 * At this point, the destination of the packet must be a unicast
1781 * or an anycast address(i.e. not a multicast).
1784 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0;
1787 IF_AFDATA_LOCK(ifp);
1788 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
1789 IF_AFDATA_UNLOCK(ifp);
1790 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1792 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1793 * the condition below is not very efficient. But we believe
1794 * it is tolerable, because this should be a rare case.
1796 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
1797 IF_AFDATA_LOCK(ifp);
1798 ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
1799 IF_AFDATA_UNLOCK(ifp);
1803 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1804 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1805 char ip6buf[INET6_ADDRSTRLEN];
1807 "nd6_output: can't allocate llinfo for %s "
1809 ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
1810 senderr(EIO); /* XXX: good error? */
1812 goto sendpkt; /* send anyway */
1815 /* We don't have to do link-layer address resolution on a p2p link. */
1816 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1817 ln->ln_state < ND6_LLINFO_REACHABLE) {
1818 if ((flags & LLE_EXCLUSIVE) == 0) {
1819 flags |= LLE_EXCLUSIVE;
1822 ln->ln_state = ND6_LLINFO_STALE;
1823 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
1827 * The first time we send a packet to a neighbor whose entry is
1828 * STALE, we have to change the state to DELAY and a sets a timer to
1829 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1830 * neighbor unreachability detection on expiration.
1833 if (ln->ln_state == ND6_LLINFO_STALE) {
1834 if ((flags & LLE_EXCLUSIVE) == 0) {
1835 flags |= LLE_EXCLUSIVE;
1840 ln->ln_state = ND6_LLINFO_DELAY;
1841 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1845 * If the neighbor cache entry has a state other than INCOMPLETE
1846 * (i.e. its link-layer address is already resolved), just
1849 if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1853 * There is a neighbor cache entry, but no ethernet address
1854 * response yet. Append this latest packet to the end of the
1855 * packet queue in the mbuf, unless the number of the packet
1856 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1857 * the oldest packet in the queue will be removed.
1859 if (ln->ln_state == ND6_LLINFO_NOSTATE)
1860 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1862 if ((flags & LLE_EXCLUSIVE) == 0) {
1863 flags |= LLE_EXCLUSIVE;
1868 LLE_WLOCK_ASSERT(ln);
1871 struct mbuf *m_hold;
1875 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
1877 if (m_hold->m_nextpkt == NULL) {
1878 m_hold->m_nextpkt = m;
1882 while (i >= V_nd6_maxqueuelen) {
1883 m_hold = ln->la_hold;
1884 ln->la_hold = ln->la_hold->m_nextpkt;
1893 * If there has been no NS for the neighbor after entering the
1894 * INCOMPLETE state, send the first solicitation.
1896 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
1899 nd6_llinfo_settimer_locked(ln,
1900 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
1902 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1903 if (lle != NULL && ln == lle)
1906 } else if (lle == NULL || ln != lle) {
1908 * We did the lookup (no lle arg) so we
1909 * need to do the unlock here.
1917 /* discard the packet if IPv6 operation is disabled on the interface */
1918 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1919 error = ENETDOWN; /* better error? */
1923 * ln is valid and the caller did not pass in
1926 if ((ln != NULL) && (lle == NULL)) {
1927 if (flags & LLE_EXCLUSIVE)
1934 mac_netinet6_nd6_send(ifp, m);
1937 * We were passed in a pointer to an lle with the lock held
1938 * this means that we can't call if_output as we will
1939 * recurse on the lle lock - so what we do is we create
1940 * a list of mbufs to send and transmit them in the caller
1941 * after the lock is dropped
1950 * append mbuf to end of deferred chain
1953 while (mb->m_nextpkt != NULL)
1959 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1960 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
1963 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
1968 * ln is valid and the caller did not pass in
1971 if ((ln != NULL) && (lle == NULL)) {
1972 if (flags & LLE_EXCLUSIVE)
1985 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
1986 struct sockaddr_in6 *dst, struct route *ro)
1988 struct mbuf *m, *m_head;
1989 struct ifnet *outifp;
1993 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2000 m_head = m_head->m_nextpkt;
2001 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);
2006 * note that intermediate errors are blindly ignored - but this is
2007 * the same convention as used with nd6_output when called by
2015 nd6_need_cache(struct ifnet *ifp)
2018 * XXX: we currently do not make neighbor cache on any interface
2019 * other than ARCnet, Ethernet, FDDI and GIF.
2022 * - unidirectional tunnels needs no ND
2024 switch (ifp->if_type) {
2032 #ifdef IFT_IEEE80211
2038 case IFT_GIF: /* XXX need more cases? */
2042 case IFT_PROPVIRTUAL:
2050 * the callers of this function need to be re-worked to drop
2051 * the lle lock, drop here for now
2054 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2055 struct sockaddr *dst, u_char *desten, struct llentry **lle)
2060 IF_AFDATA_UNLOCK_ASSERT(ifp);
2061 if (m->m_flags & M_MCAST) {
2064 switch (ifp->if_type) {
2070 #ifdef IFT_IEEE80211
2075 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2080 * netbsd can use if_broadcastaddr, but we don't do so
2081 * to reduce # of ifdef.
2083 for (i = 0; i < ifp->if_addrlen; i++)
2091 return (EAFNOSUPPORT);
2097 * the entry should have been created in nd6_store_lladdr
2099 IF_AFDATA_RLOCK(ifp);
2100 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2101 IF_AFDATA_RUNLOCK(ifp);
2102 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2105 /* this could happen, if we could not allocate memory */
2110 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2114 * A *small* use after free race exists here
2120 clear_llinfo_pqueue(struct llentry *ln)
2122 struct mbuf *m_hold, *m_hold_next;
2124 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2125 m_hold_next = m_hold->m_nextpkt;
2126 m_hold->m_nextpkt = NULL;
2134 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2135 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2137 SYSCTL_DECL(_net_inet6_icmp6);
2139 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2140 CTLFLAG_RD, nd6_sysctl_drlist, "");
2141 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2142 CTLFLAG_RD, nd6_sysctl_prlist, "");
2143 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2144 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2147 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2149 struct in6_defrouter d;
2150 struct nd_defrouter *dr;
2156 bzero(&d, sizeof(d));
2157 d.rtaddr.sin6_family = AF_INET6;
2158 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2163 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2164 d.rtaddr.sin6_addr = dr->rtaddr;
2165 error = sa6_recoverscope(&d.rtaddr);
2168 d.flags = dr->flags;
2169 d.rtlifetime = dr->rtlifetime;
2170 d.expire = dr->expire;
2171 d.if_index = dr->ifp->if_index;
2172 error = SYSCTL_OUT(req, &d, sizeof(d));
2180 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2182 struct in6_prefix p;
2183 struct sockaddr_in6 s6;
2184 struct nd_prefix *pr;
2185 struct nd_pfxrouter *pfr;
2188 char ip6buf[INET6_ADDRSTRLEN];
2193 bzero(&p, sizeof(p));
2194 p.origin = PR_ORIG_RA;
2195 bzero(&s6, sizeof(s6));
2196 s6.sin6_family = AF_INET6;
2197 s6.sin6_len = sizeof(s6);
2202 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2203 p.prefix = pr->ndpr_prefix;
2204 if (sa6_recoverscope(&p.prefix)) {
2205 log(LOG_ERR, "scope error in prefix list (%s)\n",
2206 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2207 /* XXX: press on... */
2209 p.raflags = pr->ndpr_raf;
2210 p.prefixlen = pr->ndpr_plen;
2211 p.vltime = pr->ndpr_vltime;
2212 p.pltime = pr->ndpr_pltime;
2213 p.if_index = pr->ndpr_ifp->if_index;
2214 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2217 /* XXX: we assume time_t is signed. */
2219 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2220 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2221 p.expire = pr->ndpr_lastupdate +
2224 p.expire = maxexpire;
2226 p.refcnt = pr->ndpr_refcnt;
2227 p.flags = pr->ndpr_stateflags;
2229 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2231 error = SYSCTL_OUT(req, &p, sizeof(p));
2234 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2235 s6.sin6_addr = pfr->router->rtaddr;
2236 if (sa6_recoverscope(&s6))
2238 "scope error in prefix list (%s)\n",
2239 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2240 error = SYSCTL_OUT(req, &s6, sizeof(s6));