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>
54 #include <sys/sysctl.h>
57 #include <net/if_var.h>
58 #include <net/if_arc.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/iso88025.h>
63 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <net/if_llatbl.h>
69 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le))
70 #include <netinet/if_ether.h>
71 #include <netinet6/in6_var.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet6/in6_ifattach.h>
77 #include <netinet/icmp6.h>
78 #include <netinet6/send.h>
80 #include <sys/limits.h>
82 #include <security/mac/mac_framework.h>
84 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
85 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
87 #define SIN6(s) ((const struct sockaddr_in6 *)(s))
90 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */
91 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */
92 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */
93 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */
94 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for
96 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage
99 /* preventing too many loops in ND option parsing */
100 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
102 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper
104 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
106 #define V_nd6_maxndopt VNET(nd6_maxndopt)
107 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen)
110 VNET_DEFINE(int, nd6_debug) = 1;
112 VNET_DEFINE(int, nd6_debug) = 0;
117 static int nd6_inuse, nd6_allocated;
120 VNET_DEFINE(struct nd_drhead, nd_defrouter);
121 VNET_DEFINE(struct nd_prhead, nd_prefix);
123 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
124 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
126 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
128 static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *,
130 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
131 static void nd6_slowtimo(void *);
132 static int regen_tmpaddr(struct in6_ifaddr *);
133 static struct llentry *nd6_free(struct llentry *, int);
134 static void nd6_llinfo_timer(void *);
135 static void clear_llinfo_pqueue(struct llentry *);
136 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
137 static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *,
138 struct sockaddr_in6 *);
139 static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *,
140 struct sockaddr_in6 *);
142 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
143 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
145 VNET_DEFINE(struct callout, nd6_timer_ch);
151 LIST_INIT(&V_nd_prefix);
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);
169 callout_drain(&V_nd6_slowtimo_ch);
170 callout_drain(&V_nd6_timer_ch);
175 nd6_ifattach(struct ifnet *ifp)
177 struct nd_ifinfo *nd;
179 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO);
182 nd->chlim = IPV6_DEFHLIM;
183 nd->basereachable = REACHABLE_TIME;
184 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
185 nd->retrans = RETRANS_TIMER;
187 nd->flags = ND6_IFF_PERFORMNUD;
189 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
190 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
191 * default regardless of the V_ip6_auto_linklocal configuration to
192 * give a reasonable default behavior.
194 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
195 (ifp->if_flags & IFF_LOOPBACK))
196 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
198 * A loopback interface does not need to accept RTADV.
199 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
200 * default regardless of the V_ip6_accept_rtadv configuration to
201 * prevent the interface from accepting RA messages arrived
202 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
204 if (V_ip6_accept_rtadv &&
205 !(ifp->if_flags & IFF_LOOPBACK) &&
206 (ifp->if_type != IFT_BRIDGE))
207 nd->flags |= ND6_IFF_ACCEPT_RTADV;
208 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
209 nd->flags |= ND6_IFF_NO_RADR;
211 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
212 nd6_setmtu0(ifp, nd);
218 nd6_ifdetach(struct nd_ifinfo *nd)
225 * Reset ND level link MTU. This function is called when the physical MTU
226 * changes, which means we might have to adjust the ND level MTU.
229 nd6_setmtu(struct ifnet *ifp)
232 nd6_setmtu0(ifp, ND_IFINFO(ifp));
235 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
237 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
241 omaxmtu = ndi->maxmtu;
243 switch (ifp->if_type) {
245 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
248 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
251 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
254 ndi->maxmtu = ifp->if_mtu;
259 * Decreasing the interface MTU under IPV6 minimum MTU may cause
260 * undesirable situation. We thus notify the operator of the change
261 * explicitly. The check for omaxmtu is necessary to restrict the
262 * log to the case of changing the MTU, not initializing it.
264 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
265 log(LOG_NOTICE, "nd6_setmtu0: "
266 "new link MTU on %s (%lu) is too small for IPv6\n",
267 if_name(ifp), (unsigned long)ndi->maxmtu);
270 if (ndi->maxmtu > V_in6_maxmtu)
271 in6_setmaxmtu(); /* check all interfaces just in case */
276 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
279 bzero(ndopts, sizeof(*ndopts));
280 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
282 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
285 ndopts->nd_opts_done = 1;
286 ndopts->nd_opts_search = NULL;
291 * Take one ND option.
294 nd6_option(union nd_opts *ndopts)
296 struct nd_opt_hdr *nd_opt;
299 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
300 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
302 if (ndopts->nd_opts_search == NULL)
304 if (ndopts->nd_opts_done)
307 nd_opt = ndopts->nd_opts_search;
309 /* make sure nd_opt_len is inside the buffer */
310 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
311 bzero(ndopts, sizeof(*ndopts));
315 olen = nd_opt->nd_opt_len << 3;
318 * Message validation requires that all included
319 * options have a length that is greater than zero.
321 bzero(ndopts, sizeof(*ndopts));
325 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
326 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
327 /* option overruns the end of buffer, invalid */
328 bzero(ndopts, sizeof(*ndopts));
330 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
331 /* reached the end of options chain */
332 ndopts->nd_opts_done = 1;
333 ndopts->nd_opts_search = NULL;
339 * Parse multiple ND options.
340 * This function is much easier to use, for ND routines that do not need
341 * multiple options of the same type.
344 nd6_options(union nd_opts *ndopts)
346 struct nd_opt_hdr *nd_opt;
349 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
350 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
352 if (ndopts->nd_opts_search == NULL)
356 nd_opt = nd6_option(ndopts);
357 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
359 * Message validation requires that all included
360 * options have a length that is greater than zero.
362 ICMP6STAT_INC(icp6s_nd_badopt);
363 bzero(ndopts, sizeof(*ndopts));
370 switch (nd_opt->nd_opt_type) {
371 case ND_OPT_SOURCE_LINKADDR:
372 case ND_OPT_TARGET_LINKADDR:
374 case ND_OPT_REDIRECTED_HEADER:
376 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
378 "duplicated ND6 option found (type=%d)\n",
379 nd_opt->nd_opt_type));
382 ndopts->nd_opt_array[nd_opt->nd_opt_type]
386 case ND_OPT_PREFIX_INFORMATION:
387 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
388 ndopts->nd_opt_array[nd_opt->nd_opt_type]
391 ndopts->nd_opts_pi_end =
392 (struct nd_opt_prefix_info *)nd_opt;
394 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
395 case ND_OPT_RDNSS: /* RFC 6106 */
396 case ND_OPT_DNSSL: /* RFC 6106 */
398 * Silently ignore options we know and do not care about
404 * Unknown options must be silently ignored,
405 * to accomodate future extension to the protocol.
408 "nd6_options: unsupported option %d - "
409 "option ignored\n", nd_opt->nd_opt_type));
414 if (i > V_nd6_maxndopt) {
415 ICMP6STAT_INC(icp6s_nd_toomanyopt);
416 nd6log((LOG_INFO, "too many loop in nd opt\n"));
420 if (ndopts->nd_opts_done)
428 * ND6 timer routine to handle ND6 entries
431 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
435 LLE_WLOCK_ASSERT(ln);
440 canceled = callout_stop(&ln->ln_timer_ch);
442 ln->la_expire = time_uptime + tick / hz;
444 if (tick > INT_MAX) {
445 ln->ln_ntick = tick - INT_MAX;
446 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
447 nd6_llinfo_timer, ln);
450 canceled = callout_reset(&ln->ln_timer_ch, tick,
451 nd6_llinfo_timer, ln);
459 nd6_llinfo_settimer(struct llentry *ln, long tick)
463 nd6_llinfo_settimer_locked(ln, tick);
468 nd6_llinfo_timer(void *arg)
471 struct in6_addr *dst;
473 struct nd_ifinfo *ndi = NULL;
475 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
476 ln = (struct llentry *)arg;
478 if (callout_pending(&ln->la_timer)) {
480 * Here we are a bit odd here in the treatment of
481 * active/pending. If the pending bit is set, it got
482 * rescheduled before I ran. The active
483 * bit we ignore, since if it was stopped
484 * in ll_tablefree() and was currently running
485 * it would have return 0 so the code would
486 * not have deleted it since the callout could
487 * not be stopped so we want to go through
488 * with the delete here now. If the callout
489 * was restarted, the pending bit will be back on and
490 * we just want to bail since the callout_reset would
491 * return 1 and our reference would have been removed
492 * by nd6_llinfo_settimer_locked above since canceled
498 ifp = ln->lle_tbl->llt_ifp;
499 CURVNET_SET(ifp->if_vnet);
501 if (ln->ln_ntick > 0) {
502 if (ln->ln_ntick > INT_MAX) {
503 ln->ln_ntick -= INT_MAX;
504 nd6_llinfo_settimer_locked(ln, INT_MAX);
507 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
512 ndi = ND_IFINFO(ifp);
513 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
514 if (ln->la_flags & LLE_STATIC) {
518 if (ln->la_flags & LLE_DELETED) {
519 (void)nd6_free(ln, 0);
524 switch (ln->ln_state) {
525 case ND6_LLINFO_INCOMPLETE:
526 if (ln->la_asked < V_nd6_mmaxtries) {
528 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
530 nd6_ns_output(ifp, NULL, dst, ln, NULL);
533 struct mbuf *m = ln->la_hold;
538 * assuming every packet in la_hold has the
539 * same IP header. Send error after unlock.
544 clear_llinfo_pqueue(ln);
546 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT);
547 (void)nd6_free(ln, 0);
550 icmp6_error2(m, ICMP6_DST_UNREACH,
551 ICMP6_DST_UNREACH_ADDR, 0, ifp);
554 case ND6_LLINFO_REACHABLE:
555 if (!ND6_LLINFO_PERMANENT(ln)) {
556 ln->ln_state = ND6_LLINFO_STALE;
557 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
561 case ND6_LLINFO_STALE:
562 /* Garbage Collection(RFC 2461 5.3) */
563 if (!ND6_LLINFO_PERMANENT(ln)) {
564 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
565 (void)nd6_free(ln, 1);
570 case ND6_LLINFO_DELAY:
571 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
574 ln->ln_state = ND6_LLINFO_PROBE;
575 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
577 nd6_ns_output(ifp, dst, dst, ln, NULL);
580 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
581 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
584 case ND6_LLINFO_PROBE:
585 if (ln->la_asked < V_nd6_umaxtries) {
587 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
589 nd6_ns_output(ifp, dst, dst, ln, NULL);
592 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
593 (void)nd6_free(ln, 0);
598 panic("%s: paths in a dark night can be confusing: %d",
599 __func__, ln->ln_state);
609 * ND6 timer routine to expire default route list and prefix list
614 CURVNET_SET((struct vnet *) arg);
615 struct nd_defrouter *dr, *ndr;
616 struct nd_prefix *pr, *npr;
617 struct in6_ifaddr *ia6, *nia6;
619 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
622 /* expire default router list */
623 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
624 if (dr->expire && dr->expire < time_uptime)
629 * expire interface addresses.
630 * in the past the loop was inside prefix expiry processing.
631 * However, from a stricter speci-confrmance standpoint, we should
632 * rather separate address lifetimes and prefix lifetimes.
634 * XXXRW: in6_ifaddrhead locking.
637 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
638 /* check address lifetime */
639 if (IFA6_IS_INVALID(ia6)) {
643 * If the expiring address is temporary, try
644 * regenerating a new one. This would be useful when
645 * we suspended a laptop PC, then turned it on after a
646 * period that could invalidate all temporary
647 * addresses. Although we may have to restart the
648 * loop (see below), it must be after purging the
649 * address. Otherwise, we'd see an infinite loop of
652 if (V_ip6_use_tempaddr &&
653 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
654 if (regen_tmpaddr(ia6) == 0)
658 in6_purgeaddr(&ia6->ia_ifa);
661 goto addrloop; /* XXX: see below */
662 } else if (IFA6_IS_DEPRECATED(ia6)) {
663 int oldflags = ia6->ia6_flags;
665 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
668 * If a temporary address has just become deprecated,
669 * regenerate a new one if possible.
671 if (V_ip6_use_tempaddr &&
672 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
673 (oldflags & IN6_IFF_DEPRECATED) == 0) {
675 if (regen_tmpaddr(ia6) == 0) {
677 * A new temporary address is
679 * XXX: this means the address chain
680 * has changed while we are still in
681 * the loop. Although the change
682 * would not cause disaster (because
683 * it's not a deletion, but an
684 * addition,) we'd rather restart the
685 * loop just for safety. Or does this
686 * significantly reduce performance??
693 * A new RA might have made a deprecated address
696 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
700 /* expire prefix list */
701 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
703 * check prefix lifetime.
704 * since pltime is just for autoconf, pltime processing for
705 * prefix is not necessary.
707 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
708 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) {
711 * address expiration and prefix expiration are
712 * separate. NEVER perform in6_purgeaddr here.
721 * ia6 - deprecated/invalidated temporary address
724 regen_tmpaddr(struct in6_ifaddr *ia6)
728 struct in6_ifaddr *public_ifa6 = NULL;
730 ifp = ia6->ia_ifa.ifa_ifp;
732 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
733 struct in6_ifaddr *it6;
735 if (ifa->ifa_addr->sa_family != AF_INET6)
738 it6 = (struct in6_ifaddr *)ifa;
740 /* ignore no autoconf addresses. */
741 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
744 /* ignore autoconf addresses with different prefixes. */
745 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
749 * Now we are looking at an autoconf address with the same
750 * prefix as ours. If the address is temporary and is still
751 * preferred, do not create another one. It would be rare, but
752 * could happen, for example, when we resume a laptop PC after
755 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
756 !IFA6_IS_DEPRECATED(it6)) {
762 * This is a public autoconf address that has the same prefix
763 * as ours. If it is preferred, keep it. We can't break the
764 * loop here, because there may be a still-preferred temporary
765 * address with the prefix.
767 if (!IFA6_IS_DEPRECATED(it6))
770 if (public_ifa6 != NULL)
771 ifa_ref(&public_ifa6->ia_ifa);
773 IF_ADDR_RUNLOCK(ifp);
775 if (public_ifa6 != NULL) {
778 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
779 ifa_free(&public_ifa6->ia_ifa);
780 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
781 " tmp addr,errno=%d\n", e);
784 ifa_free(&public_ifa6->ia_ifa);
792 * Nuke neighbor cache/prefix/default router management table, right before
796 nd6_purge(struct ifnet *ifp)
798 struct nd_defrouter *dr, *ndr;
799 struct nd_prefix *pr, *npr;
802 * Nuke default router list entries toward ifp.
803 * We defer removal of default router list entries that is installed
804 * in the routing table, in order to keep additional side effects as
807 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
815 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
823 /* Nuke prefix list entries toward ifp */
824 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
825 if (pr->ndpr_ifp == ifp) {
827 * Because if_detach() does *not* release prefixes
828 * while purging addresses the reference count will
829 * still be above zero. We therefore reset it to
830 * make sure that the prefix really gets purged.
835 * Previously, pr->ndpr_addr is removed as well,
836 * but I strongly believe we don't have to do it.
837 * nd6_purge() is only called from in6_ifdetach(),
838 * which removes all the associated interface addresses
840 * (jinmei@kame.net 20010129)
846 /* cancel default outgoing interface setting */
847 if (V_nd6_defifindex == ifp->if_index)
848 nd6_setdefaultiface(0);
850 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
851 /* Refresh default router list. */
856 * We do not nuke the neighbor cache entries here any more
857 * because the neighbor cache is kept in if_afdata[AF_INET6].
858 * nd6_purge() is invoked by in6_ifdetach() which is called
859 * from if_detach() where everything gets purged. So let
860 * in6_domifdetach() do the actual L2 table purging work.
865 * the caller acquires and releases the lock on the lltbls
866 * Returns the llentry locked
869 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
871 struct sockaddr_in6 sin6;
875 bzero(&sin6, sizeof(sin6));
876 sin6.sin6_len = sizeof(struct sockaddr_in6);
877 sin6.sin6_family = AF_INET6;
878 sin6.sin6_addr = *addr6;
880 IF_AFDATA_LOCK_ASSERT(ifp);
883 if (flags & ND6_CREATE)
884 llflags |= LLE_CREATE;
885 if (flags & ND6_EXCLUSIVE)
886 llflags |= LLE_EXCLUSIVE;
888 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
889 if ((ln != NULL) && (llflags & LLE_CREATE))
890 ln->ln_state = ND6_LLINFO_NOSTATE;
896 * Test whether a given IPv6 address is a neighbor or not, ignoring
897 * the actual neighbor cache. The neighbor cache is ignored in order
898 * to not reenter the routing code from within itself.
901 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
903 struct nd_prefix *pr;
904 struct ifaddr *dstaddr;
907 * A link-local address is always a neighbor.
908 * XXX: a link does not necessarily specify a single interface.
910 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
911 struct sockaddr_in6 sin6_copy;
915 * We need sin6_copy since sa6_recoverscope() may modify the
919 if (sa6_recoverscope(&sin6_copy))
920 return (0); /* XXX: should be impossible */
921 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
923 if (sin6_copy.sin6_scope_id == zone)
930 * If the address matches one of our addresses,
931 * it should be a neighbor.
932 * If the address matches one of our on-link prefixes, it should be a
935 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
936 if (pr->ndpr_ifp != ifp)
939 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
942 /* Always use the default FIB here. */
943 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
944 0, 0, RT_DEFAULT_FIB);
948 * This is the case where multiple interfaces
949 * have the same prefix, but only one is installed
950 * into the routing table and that prefix entry
951 * is not the one being examined here. In the case
952 * where RADIX_MPATH is enabled, multiple route
953 * entries (of the same rt_key value) will be
954 * installed because the interface addresses all
957 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
958 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
965 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
966 &addr->sin6_addr, &pr->ndpr_mask))
971 * If the address is assigned on the node of the other side of
972 * a p2p interface, the address should be a neighbor.
974 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr, RT_ALL_FIBS);
975 if (dstaddr != NULL) {
976 if (dstaddr->ifa_ifp == ifp) {
984 * If the default router list is empty, all addresses are regarded
985 * as on-link, and thus, as a neighbor.
987 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
988 TAILQ_EMPTY(&V_nd_defrouter) &&
989 V_nd6_defifindex == ifp->if_index) {
998 * Detect if a given IPv6 address identifies a neighbor on a given link.
999 * XXX: should take care of the destination of a p2p link?
1002 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
1004 struct llentry *lle;
1007 IF_AFDATA_UNLOCK_ASSERT(ifp);
1008 if (nd6_is_new_addr_neighbor(addr, ifp))
1012 * Even if the address matches none of our addresses, it might be
1013 * in the neighbor cache.
1015 IF_AFDATA_RLOCK(ifp);
1016 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1020 IF_AFDATA_RUNLOCK(ifp);
1025 * Free an nd6 llinfo entry.
1026 * Since the function would cause significant changes in the kernel, DO NOT
1027 * make it global, unless you have a strong reason for the change, and are sure
1028 * that the change is safe.
1030 static struct llentry *
1031 nd6_free(struct llentry *ln, int gc)
1033 struct llentry *next;
1034 struct nd_defrouter *dr;
1037 LLE_WLOCK_ASSERT(ln);
1040 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1041 * even though it is not harmful, it was not really necessary.
1045 nd6_llinfo_settimer_locked(ln, -1);
1047 ifp = ln->lle_tbl->llt_ifp;
1049 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1050 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1052 if (dr != NULL && dr->expire &&
1053 ln->ln_state == ND6_LLINFO_STALE && gc) {
1055 * If the reason for the deletion is just garbage
1056 * collection, and the neighbor is an active default
1057 * router, do not delete it. Instead, reset the GC
1058 * timer using the router's lifetime.
1059 * Simply deleting the entry would affect default
1060 * router selection, which is not necessarily a good
1061 * thing, especially when we're using router preference
1063 * XXX: the check for ln_state would be redundant,
1064 * but we intentionally keep it just in case.
1066 if (dr->expire > time_uptime)
1067 nd6_llinfo_settimer_locked(ln,
1068 (dr->expire - time_uptime) * hz);
1070 nd6_llinfo_settimer_locked(ln,
1071 (long)V_nd6_gctimer * hz);
1073 next = LIST_NEXT(ln, lle_next);
1081 * Unreachablity of a router might affect the default
1082 * router selection and on-link detection of advertised
1087 * Temporarily fake the state to choose a new default
1088 * router and to perform on-link determination of
1089 * prefixes correctly.
1090 * Below the state will be set correctly,
1091 * or the entry itself will be deleted.
1093 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1096 if (ln->ln_router || dr) {
1099 * We need to unlock to avoid a LOR with rt6_flush() with the
1100 * rnh and for the calls to pfxlist_onlink_check() and
1101 * defrouter_select() in the block further down for calls
1102 * into nd6_lookup(). We still hold a ref.
1107 * rt6_flush must be called whether or not the neighbor
1108 * is in the Default Router List.
1109 * See a corresponding comment in nd6_na_input().
1111 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1116 * Since defrouter_select() does not affect the
1117 * on-link determination and MIP6 needs the check
1118 * before the default router selection, we perform
1121 pfxlist_onlink_check();
1124 * Refresh default router list.
1129 if (ln->ln_router || dr)
1134 * Before deleting the entry, remember the next entry as the
1135 * return value. We need this because pfxlist_onlink_check() above
1136 * might have freed other entries (particularly the old next entry) as
1137 * a side effect (XXX).
1139 next = LIST_NEXT(ln, lle_next);
1142 * Save to unlock. We still hold an extra reference and will not
1143 * free(9) in llentry_free() if someone else holds one as well.
1146 IF_AFDATA_LOCK(ifp);
1149 /* Guard against race with other llentry_free(). */
1150 if (ln->la_flags & LLE_LINKED) {
1154 LLE_FREE_LOCKED(ln);
1156 IF_AFDATA_UNLOCK(ifp);
1162 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1164 * XXX cost-effective methods?
1167 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1172 if ((dst6 == NULL) || (rt == NULL))
1176 IF_AFDATA_RLOCK(ifp);
1177 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1178 IF_AFDATA_RUNLOCK(ifp);
1182 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1186 * if we get upper-layer reachability confirmation many times,
1187 * it is possible we have false information.
1191 if (ln->ln_byhint > V_nd6_maxnudhint) {
1196 ln->ln_state = ND6_LLINFO_REACHABLE;
1197 if (!ND6_LLINFO_PERMANENT(ln)) {
1198 nd6_llinfo_settimer_locked(ln,
1199 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1207 * Rejuvenate this function for routing operations related
1211 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1213 struct sockaddr_in6 *gateway;
1214 struct nd_defrouter *dr;
1217 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1228 * Only indirect routes are interesting.
1230 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1233 * check for default route
1235 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1236 &SIN6(rt_key(rt))->sin6_addr)) {
1238 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1248 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1250 struct in6_drlist *drl = (struct in6_drlist *)data;
1251 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1252 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1253 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1254 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1255 struct nd_defrouter *dr;
1256 struct nd_prefix *pr;
1257 int i = 0, error = 0;
1259 if (ifp->if_afdata[AF_INET6] == NULL)
1260 return (EPFNOSUPPORT);
1262 case SIOCGDRLST_IN6:
1264 * obsolete API, use sysctl under net.inet6.icmp6
1266 bzero(drl, sizeof(*drl));
1267 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1270 drl->defrouter[i].rtaddr = dr->rtaddr;
1271 in6_clearscope(&drl->defrouter[i].rtaddr);
1273 drl->defrouter[i].flags = dr->flags;
1274 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1275 drl->defrouter[i].expire = dr->expire +
1276 (time_second - time_uptime);
1277 drl->defrouter[i].if_index = dr->ifp->if_index;
1281 case SIOCGPRLST_IN6:
1283 * obsolete API, use sysctl under net.inet6.icmp6
1285 * XXX the structure in6_prlist was changed in backward-
1286 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1287 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1290 * XXX meaning of fields, especialy "raflags", is very
1291 * differnet between RA prefix list and RR/static prefix list.
1292 * how about separating ioctls into two?
1294 bzero(oprl, sizeof(*oprl));
1295 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1296 struct nd_pfxrouter *pfr;
1301 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1302 oprl->prefix[i].raflags = pr->ndpr_raf;
1303 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1304 oprl->prefix[i].vltime = pr->ndpr_vltime;
1305 oprl->prefix[i].pltime = pr->ndpr_pltime;
1306 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1307 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1308 oprl->prefix[i].expire = 0;
1312 /* XXX: we assume time_t is signed. */
1315 ((sizeof(maxexpire) * 8) - 1));
1316 if (pr->ndpr_vltime <
1317 maxexpire - pr->ndpr_lastupdate) {
1318 oprl->prefix[i].expire =
1319 pr->ndpr_lastupdate +
1321 (time_second - time_uptime);
1323 oprl->prefix[i].expire = maxexpire;
1327 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1329 #define RTRADDR oprl->prefix[i].advrtr[j]
1330 RTRADDR = pfr->router->rtaddr;
1331 in6_clearscope(&RTRADDR);
1336 oprl->prefix[i].advrtrs = j;
1337 oprl->prefix[i].origin = PR_ORIG_RA;
1343 case OSIOCGIFINFO_IN6:
1345 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1346 bzero(&ND, sizeof(ND));
1347 ND.linkmtu = IN6_LINKMTU(ifp);
1348 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1349 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1350 ND.reachable = ND_IFINFO(ifp)->reachable;
1351 ND.retrans = ND_IFINFO(ifp)->retrans;
1352 ND.flags = ND_IFINFO(ifp)->flags;
1353 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1354 ND.chlim = ND_IFINFO(ifp)->chlim;
1356 case SIOCGIFINFO_IN6:
1357 ND = *ND_IFINFO(ifp);
1359 case SIOCSIFINFO_IN6:
1361 * used to change host variables from userland.
1362 * intented for a use on router to reflect RA configurations.
1364 /* 0 means 'unspecified' */
1365 if (ND.linkmtu != 0) {
1366 if (ND.linkmtu < IPV6_MMTU ||
1367 ND.linkmtu > IN6_LINKMTU(ifp)) {
1371 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1374 if (ND.basereachable != 0) {
1375 int obasereachable = ND_IFINFO(ifp)->basereachable;
1377 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1378 if (ND.basereachable != obasereachable)
1379 ND_IFINFO(ifp)->reachable =
1380 ND_COMPUTE_RTIME(ND.basereachable);
1382 if (ND.retrans != 0)
1383 ND_IFINFO(ifp)->retrans = ND.retrans;
1385 ND_IFINFO(ifp)->chlim = ND.chlim;
1387 case SIOCSIFINFO_FLAGS:
1390 struct in6_ifaddr *ia;
1392 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1393 !(ND.flags & ND6_IFF_IFDISABLED)) {
1394 /* ifdisabled 1->0 transision */
1397 * If the interface is marked as ND6_IFF_IFDISABLED and
1398 * has an link-local address with IN6_IFF_DUPLICATED,
1399 * do not clear ND6_IFF_IFDISABLED.
1400 * See RFC 4862, Section 5.4.5.
1402 int duplicated_linklocal = 0;
1405 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1406 if (ifa->ifa_addr->sa_family != AF_INET6)
1408 ia = (struct in6_ifaddr *)ifa;
1409 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1410 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1411 duplicated_linklocal = 1;
1415 IF_ADDR_RUNLOCK(ifp);
1417 if (duplicated_linklocal) {
1418 ND.flags |= ND6_IFF_IFDISABLED;
1419 log(LOG_ERR, "Cannot enable an interface"
1420 " with a link-local address marked"
1423 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1424 if (ifp->if_flags & IFF_UP)
1427 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1428 (ND.flags & ND6_IFF_IFDISABLED)) {
1429 /* ifdisabled 0->1 transision */
1430 /* Mark all IPv6 address as tentative. */
1432 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1434 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1435 if (ifa->ifa_addr->sa_family != AF_INET6)
1437 ia = (struct in6_ifaddr *)ifa;
1438 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1440 IF_ADDR_RUNLOCK(ifp);
1443 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1444 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1445 /* auto_linklocal 0->1 transision */
1447 /* If no link-local address on ifp, configure */
1448 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1449 in6_ifattach(ifp, NULL);
1450 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1451 ifp->if_flags & IFF_UP) {
1453 * When the IF already has
1454 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1455 * address is assigned, and IFF_UP, try to
1458 int haslinklocal = 0;
1461 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1462 if (ifa->ifa_addr->sa_family != AF_INET6)
1464 ia = (struct in6_ifaddr *)ifa;
1465 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1470 IF_ADDR_RUNLOCK(ifp);
1472 in6_ifattach(ifp, NULL);
1476 ND_IFINFO(ifp)->flags = ND.flags;
1479 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1480 /* sync kernel routing table with the default router list */
1484 case SIOCSPFXFLUSH_IN6:
1486 /* flush all the prefix advertised by routers */
1487 struct nd_prefix *pr, *next;
1489 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1490 struct in6_ifaddr *ia, *ia_next;
1492 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1495 /* do we really have to remove addresses as well? */
1496 /* XXXRW: in6_ifaddrhead locking. */
1497 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1499 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1502 if (ia->ia6_ndpr == pr)
1503 in6_purgeaddr(&ia->ia_ifa);
1509 case SIOCSRTRFLUSH_IN6:
1511 /* flush all the default routers */
1512 struct nd_defrouter *dr, *next;
1515 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1521 case SIOCGNBRINFO_IN6:
1524 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1526 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1529 IF_AFDATA_RLOCK(ifp);
1530 ln = nd6_lookup(&nb_addr, 0, ifp);
1531 IF_AFDATA_RUNLOCK(ifp);
1537 nbi->state = ln->ln_state;
1538 nbi->asked = ln->la_asked;
1539 nbi->isrouter = ln->ln_router;
1540 if (ln->la_expire == 0)
1543 nbi->expire = ln->la_expire +
1544 (time_second - time_uptime);
1548 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1549 ndif->ifindex = V_nd6_defifindex;
1551 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1552 return (nd6_setdefaultiface(ndif->ifindex));
1558 * Create neighbor cache entry and cache link-layer address,
1559 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1562 * code - type dependent information
1565 * The caller of this function already acquired the ndp
1566 * cache table lock because the cache entry is returned.
1569 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1570 int lladdrlen, int type, int code)
1572 struct llentry *ln = NULL;
1579 uint16_t router = 0;
1580 struct sockaddr_in6 sin6;
1581 struct mbuf *chain = NULL;
1582 int static_route = 0;
1584 IF_AFDATA_UNLOCK_ASSERT(ifp);
1586 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1587 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1589 /* nothing must be updated for unspecified address */
1590 if (IN6_IS_ADDR_UNSPECIFIED(from))
1594 * Validation about ifp->if_addrlen and lladdrlen must be done in
1597 * XXX If the link does not have link-layer adderss, what should
1598 * we do? (ifp->if_addrlen == 0)
1599 * Spec says nothing in sections for RA, RS and NA. There's small
1600 * description on it in NS section (RFC 2461 7.2.3).
1602 flags = lladdr ? ND6_EXCLUSIVE : 0;
1603 IF_AFDATA_RLOCK(ifp);
1604 ln = nd6_lookup(from, flags, ifp);
1605 IF_AFDATA_RUNLOCK(ifp);
1607 flags |= ND6_EXCLUSIVE;
1608 IF_AFDATA_LOCK(ifp);
1609 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1610 IF_AFDATA_UNLOCK(ifp);
1613 /* do nothing if static ndp is set */
1614 if (ln->la_flags & LLE_STATIC) {
1623 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1624 if (olladdr && lladdr) {
1625 llchange = bcmp(lladdr, &ln->ll_addr,
1631 * newentry olladdr lladdr llchange (*=record)
1634 * 0 n y -- (3) * STALE
1636 * 0 y y y (5) * STALE
1637 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1638 * 1 -- y -- (7) * STALE
1641 if (lladdr) { /* (3-5) and (7) */
1643 * Record source link-layer address
1644 * XXX is it dependent to ifp->if_type?
1646 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1647 ln->la_flags |= LLE_VALID;
1648 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1652 if ((!olladdr && lladdr != NULL) || /* (3) */
1653 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1655 newstate = ND6_LLINFO_STALE;
1656 } else /* (1-2,4) */
1660 if (lladdr == NULL) /* (6) */
1661 newstate = ND6_LLINFO_NOSTATE;
1663 newstate = ND6_LLINFO_STALE;
1668 * Update the state of the neighbor cache.
1670 ln->ln_state = newstate;
1672 if (ln->ln_state == ND6_LLINFO_STALE) {
1673 if (ln->la_hold != NULL)
1674 nd6_grab_holdchain(ln, &chain, &sin6);
1675 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1676 /* probe right away */
1677 nd6_llinfo_settimer_locked((void *)ln, 0);
1682 * ICMP6 type dependent behavior.
1684 * NS: clear IsRouter if new entry
1685 * RS: clear IsRouter
1686 * RA: set IsRouter if there's lladdr
1687 * redir: clear IsRouter if new entry
1690 * The spec says that we must set IsRouter in the following cases:
1691 * - If lladdr exist, set IsRouter. This means (1-5).
1692 * - If it is old entry (!newentry), set IsRouter. This means (7).
1693 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1694 * A quetion arises for (1) case. (1) case has no lladdr in the
1695 * neighbor cache, this is similar to (6).
1696 * This case is rare but we figured that we MUST NOT set IsRouter.
1698 * newentry olladdr lladdr llchange NS RS RA redir
1700 * 0 n n -- (1) c ? s
1701 * 0 y n -- (2) c s s
1702 * 0 n y -- (3) c s s
1705 * 1 -- n -- (6) c c c s
1706 * 1 -- y -- (7) c c s c s
1710 switch (type & 0xff) {
1711 case ND_NEIGHBOR_SOLICIT:
1713 * New entry must have is_router flag cleared.
1715 if (is_newentry) /* (6-7) */
1720 * If the icmp is a redirect to a better router, always set the
1721 * is_router flag. Otherwise, if the entry is newly created,
1722 * clear the flag. [RFC 2461, sec 8.3]
1724 if (code == ND_REDIRECT_ROUTER)
1726 else if (is_newentry) /* (6-7) */
1729 case ND_ROUTER_SOLICIT:
1731 * is_router flag must always be cleared.
1735 case ND_ROUTER_ADVERT:
1737 * Mark an entry with lladdr as a router.
1739 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1740 (is_newentry && lladdr)) { /* (7) */
1747 static_route = (ln->la_flags & LLE_STATIC);
1748 router = ln->ln_router;
1750 if (flags & ND6_EXCLUSIVE)
1758 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
1761 * When the link-layer address of a router changes, select the
1762 * best router again. In particular, when the neighbor entry is newly
1763 * created, it might affect the selection policy.
1764 * Question: can we restrict the first condition to the "is_newentry"
1766 * XXX: when we hear an RA from a new router with the link-layer
1767 * address option, defrouter_select() is called twice, since
1768 * defrtrlist_update called the function as well. However, I believe
1769 * we can compromise the overhead, since it only happens the first
1771 * XXX: although defrouter_select() should not have a bad effect
1772 * for those are not autoconfigured hosts, we explicitly avoid such
1775 if (do_update && router &&
1776 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1778 * guaranteed recursion
1786 if (flags & ND6_EXCLUSIVE)
1797 nd6_slowtimo(void *arg)
1799 CURVNET_SET((struct vnet *) arg);
1800 struct nd_ifinfo *nd6if;
1803 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1804 nd6_slowtimo, curvnet);
1805 IFNET_RLOCK_NOSLEEP();
1806 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1807 if (ifp->if_afdata[AF_INET6] == NULL)
1809 nd6if = ND_IFINFO(ifp);
1810 if (nd6if->basereachable && /* already initialized */
1811 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1813 * Since reachable time rarely changes by router
1814 * advertisements, we SHOULD insure that a new random
1815 * value gets recomputed at least once every few hours.
1818 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1819 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1822 IFNET_RUNLOCK_NOSLEEP();
1827 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
1828 struct sockaddr_in6 *sin6)
1831 LLE_WLOCK_ASSERT(ln);
1833 *chain = ln->la_hold;
1835 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6));
1837 if (ln->ln_state == ND6_LLINFO_STALE) {
1840 * The first time we send a packet to a
1841 * neighbor whose entry is STALE, we have
1842 * to change the state to DELAY and a sets
1843 * a timer to expire in DELAY_FIRST_PROBE_TIME
1844 * seconds to ensure do neighbor unreachability
1845 * detection on expiration.
1849 ln->ln_state = ND6_LLINFO_DELAY;
1850 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1855 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1856 struct sockaddr_in6 *dst)
1860 struct ip6_hdr *ip6;
1864 mac_netinet6_nd6_send(ifp, m);
1868 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
1869 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
1870 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
1871 * to be diverted to user space. When re-injected into the kernel,
1872 * send_output() will directly dispatch them to the outgoing interface.
1874 if (send_sendso_input_hook != NULL) {
1875 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
1877 ip6 = mtod(m, struct ip6_hdr *);
1878 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
1879 /* Use the SEND socket */
1880 error = send_sendso_input_hook(m, ifp, SND_OUT,
1882 /* -1 == no app on SEND socket */
1883 if (error == 0 || error != -1)
1888 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1889 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
1890 mtod(m, struct ip6_hdr *));
1892 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1895 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL);
1900 * IPv6 packet output - light version.
1901 * Checks if destination LLE exists and is in proper state
1902 * (e.g no modification required). If not true, fall back to
1906 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1907 struct sockaddr_in6 *dst, struct rtentry *rt0)
1909 struct llentry *ln = NULL;
1911 /* discard the packet if IPv6 operation is disabled on the interface */
1912 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1914 return (ENETDOWN); /* better error? */
1917 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1920 if (nd6_need_cache(ifp) == 0)
1923 IF_AFDATA_RLOCK(ifp);
1924 ln = nd6_lookup(&dst->sin6_addr, 0, ifp);
1925 IF_AFDATA_RUNLOCK(ifp);
1928 * Perform fast path for the following cases:
1929 * 1) lle state is REACHABLE
1930 * 2) lle state is DELAY (NS message sentNS message sent)
1932 * Every other case involves lle modification, so we handle
1935 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE &&
1936 ln->ln_state != ND6_LLINFO_DELAY)) {
1937 /* Fall back to slow processing path */
1940 return (nd6_output_lle(ifp, origifp, m, dst));
1947 return (nd6_output_ifp(ifp, origifp, m, dst));
1952 * Output IPv6 packet - heavy version.
1953 * Function assume that either
1954 * 1) destination LLE does not exist, is invalid or stale, so
1955 * ND6_EXCLUSIVE lock needs to be acquired
1956 * 2) destination lle is provided (with ND6_EXCLUSIVE lock),
1957 * in that case packets are queued in &chain.
1961 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1962 struct sockaddr_in6 *dst)
1964 struct llentry *lle = NULL;
1967 KASSERT(m != NULL, ("NULL mbuf, nothing to send"));
1968 /* discard the packet if IPv6 operation is disabled on the interface */
1969 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1971 return (ENETDOWN); /* better error? */
1974 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1977 if (nd6_need_cache(ifp) == 0)
1981 * Address resolution or Neighbor Unreachability Detection
1983 * At this point, the destination of the packet must be a unicast
1984 * or an anycast address(i.e. not a multicast).
1987 IF_AFDATA_RLOCK(ifp);
1988 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
1989 IF_AFDATA_RUNLOCK(ifp);
1990 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1992 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1993 * the condition below is not very efficient. But we believe
1994 * it is tolerable, because this should be a rare case.
1996 flags = ND6_CREATE | ND6_EXCLUSIVE;
1997 IF_AFDATA_LOCK(ifp);
1998 lle = nd6_lookup(&dst->sin6_addr, flags, ifp);
1999 IF_AFDATA_UNLOCK(ifp);
2003 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2004 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2005 char ip6buf[INET6_ADDRSTRLEN];
2007 "nd6_output: can't allocate llinfo for %s "
2009 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
2013 goto sendpkt; /* send anyway */
2016 LLE_WLOCK_ASSERT(lle);
2018 /* We don't have to do link-layer address resolution on a p2p link. */
2019 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2020 lle->ln_state < ND6_LLINFO_REACHABLE) {
2021 lle->ln_state = ND6_LLINFO_STALE;
2022 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz);
2026 * The first time we send a packet to a neighbor whose entry is
2027 * STALE, we have to change the state to DELAY and a sets a timer to
2028 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2029 * neighbor unreachability detection on expiration.
2032 if (lle->ln_state == ND6_LLINFO_STALE) {
2034 lle->ln_state = ND6_LLINFO_DELAY;
2035 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz);
2039 * If the neighbor cache entry has a state other than INCOMPLETE
2040 * (i.e. its link-layer address is already resolved), just
2043 if (lle->ln_state > ND6_LLINFO_INCOMPLETE)
2047 * There is a neighbor cache entry, but no ethernet address
2048 * response yet. Append this latest packet to the end of the
2049 * packet queue in the mbuf, unless the number of the packet
2050 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2051 * the oldest packet in the queue will be removed.
2053 if (lle->ln_state == ND6_LLINFO_NOSTATE)
2054 lle->ln_state = ND6_LLINFO_INCOMPLETE;
2056 if (lle->la_hold != NULL) {
2057 struct mbuf *m_hold;
2061 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2063 if (m_hold->m_nextpkt == NULL) {
2064 m_hold->m_nextpkt = m;
2068 while (i >= V_nd6_maxqueuelen) {
2069 m_hold = lle->la_hold;
2070 lle->la_hold = lle->la_hold->m_nextpkt;
2079 * If there has been no NS for the neighbor after entering the
2080 * INCOMPLETE state, send the first solicitation.
2082 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) {
2085 nd6_llinfo_settimer_locked(lle,
2086 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2088 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL);
2090 /* We did the lookup so we need to do the unlock here. */
2100 return (nd6_output_ifp(ifp, origifp, m, dst));
2105 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2106 struct sockaddr_in6 *dst)
2108 struct mbuf *m, *m_head;
2109 struct ifnet *outifp;
2113 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2120 m_head = m_head->m_nextpkt;
2121 error = nd6_output_ifp(ifp, origifp, m, dst);
2126 * note that intermediate errors are blindly ignored - but this is
2127 * the same convention as used with nd6_output when called by
2135 nd6_need_cache(struct ifnet *ifp)
2138 * XXX: we currently do not make neighbor cache on any interface
2139 * other than ARCnet, Ethernet, FDDI and GIF.
2142 * - unidirectional tunnels needs no ND
2144 switch (ifp->if_type) {
2152 #ifdef IFT_IEEE80211
2155 case IFT_INFINIBAND:
2157 case IFT_PROPVIRTUAL:
2165 * Add pernament ND6 link-layer record for given
2166 * interface address.
2168 * Very similar to IPv4 arp_ifinit(), but:
2169 * 1) IPv6 DAD is performed in different place
2170 * 2) It is called by IPv6 protocol stack in contrast to
2171 * arp_ifinit() which is typically called in SIOCSIFADDR
2172 * driver ioctl handler.
2176 nd6_add_ifa_lle(struct in6_ifaddr *ia)
2181 ifp = ia->ia_ifa.ifa_ifp;
2182 if (nd6_need_cache(ifp) == 0)
2184 IF_AFDATA_LOCK(ifp);
2185 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
2186 ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
2187 LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
2188 IF_AFDATA_UNLOCK(ifp);
2190 ln->la_expire = 0; /* for IPv6 this means permanent */
2191 ln->ln_state = ND6_LLINFO_REACHABLE;
2193 in6_newaddrmsg(ia, RTM_ADD);
2201 * Removes ALL lle records for interface address prefix.
2202 * XXXME: That's probably not we really want to do, we need
2203 * to remove address record only and keep other records
2204 * until we determine if given prefix is really going
2208 nd6_rem_ifa_lle(struct in6_ifaddr *ia)
2210 struct sockaddr_in6 mask, addr;
2213 in6_newaddrmsg(ia, RTM_DELETE);
2215 ifp = ia->ia_ifa.ifa_ifp;
2216 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
2217 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
2218 lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr,
2219 (struct sockaddr *)&mask, LLE_STATIC);
2223 * the callers of this function need to be re-worked to drop
2224 * the lle lock, drop here for now
2227 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2228 const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
2234 IF_AFDATA_UNLOCK_ASSERT(ifp);
2235 if (m != NULL && m->m_flags & M_MCAST) {
2236 switch (ifp->if_type) {
2242 #ifdef IFT_IEEE80211
2247 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2252 return (EAFNOSUPPORT);
2258 * the entry should have been created in nd6_store_lladdr
2260 IF_AFDATA_RLOCK(ifp);
2261 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2262 IF_AFDATA_RUNLOCK(ifp);
2263 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2266 /* this could happen, if we could not allocate memory */
2271 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2273 *pflags = ln->la_flags;
2276 * A *small* use after free race exists here
2282 clear_llinfo_pqueue(struct llentry *ln)
2284 struct mbuf *m_hold, *m_hold_next;
2286 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2287 m_hold_next = m_hold->m_nextpkt;
2295 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2296 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2298 SYSCTL_DECL(_net_inet6_icmp6);
2300 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2301 CTLFLAG_RD, nd6_sysctl_drlist, "");
2302 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2303 CTLFLAG_RD, nd6_sysctl_prlist, "");
2304 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2305 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2306 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2307 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2310 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2312 struct in6_defrouter d;
2313 struct nd_defrouter *dr;
2319 bzero(&d, sizeof(d));
2320 d.rtaddr.sin6_family = AF_INET6;
2321 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2326 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2327 d.rtaddr.sin6_addr = dr->rtaddr;
2328 error = sa6_recoverscope(&d.rtaddr);
2331 d.flags = dr->flags;
2332 d.rtlifetime = dr->rtlifetime;
2333 d.expire = dr->expire + (time_second - time_uptime);
2334 d.if_index = dr->ifp->if_index;
2335 error = SYSCTL_OUT(req, &d, sizeof(d));
2343 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2345 struct in6_prefix p;
2346 struct sockaddr_in6 s6;
2347 struct nd_prefix *pr;
2348 struct nd_pfxrouter *pfr;
2351 char ip6buf[INET6_ADDRSTRLEN];
2356 bzero(&p, sizeof(p));
2357 p.origin = PR_ORIG_RA;
2358 bzero(&s6, sizeof(s6));
2359 s6.sin6_family = AF_INET6;
2360 s6.sin6_len = sizeof(s6);
2365 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2366 p.prefix = pr->ndpr_prefix;
2367 if (sa6_recoverscope(&p.prefix)) {
2368 log(LOG_ERR, "scope error in prefix list (%s)\n",
2369 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2370 /* XXX: press on... */
2372 p.raflags = pr->ndpr_raf;
2373 p.prefixlen = pr->ndpr_plen;
2374 p.vltime = pr->ndpr_vltime;
2375 p.pltime = pr->ndpr_pltime;
2376 p.if_index = pr->ndpr_ifp->if_index;
2377 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2380 /* XXX: we assume time_t is signed. */
2382 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2383 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2384 p.expire = pr->ndpr_lastupdate +
2386 (time_second - time_uptime);
2388 p.expire = maxexpire;
2390 p.refcnt = pr->ndpr_refcnt;
2391 p.flags = pr->ndpr_stateflags;
2393 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2395 error = SYSCTL_OUT(req, &p, sizeof(p));
2398 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2399 s6.sin6_addr = pfr->router->rtaddr;
2400 if (sa6_recoverscope(&s6))
2402 "scope error in prefix list (%s)\n",
2403 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2404 error = SYSCTL_OUT(req, &s6, sizeof(s6));