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"
37 #include "opt_kdtrace.h"
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/callout.h>
42 #include <sys/malloc.h>
44 #include <sys/socket.h>
45 #include <sys/sockio.h>
47 #include <sys/kernel.h>
48 #include <sys/protosw.h>
49 #include <sys/errno.h>
50 #include <sys/syslog.h>
52 #include <sys/rwlock.h>
53 #include <sys/queue.h>
55 #include <sys/sysctl.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 int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *,
137 struct sockaddr_in6 *);
138 static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *,
139 struct sockaddr_in6 *);
141 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
142 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
144 VNET_DEFINE(struct callout, nd6_timer_ch);
150 LIST_INIT(&V_nd_prefix);
152 /* initialization of the default router list */
153 TAILQ_INIT(&V_nd_defrouter);
156 callout_init(&V_nd6_slowtimo_ch, 0);
157 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
158 nd6_slowtimo, curvnet);
168 callout_drain(&V_nd6_slowtimo_ch);
169 callout_drain(&V_nd6_timer_ch);
174 nd6_ifattach(struct ifnet *ifp)
176 struct nd_ifinfo *nd;
178 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO);
181 nd->chlim = IPV6_DEFHLIM;
182 nd->basereachable = REACHABLE_TIME;
183 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
184 nd->retrans = RETRANS_TIMER;
186 nd->flags = ND6_IFF_PERFORMNUD;
188 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
189 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
190 * default regardless of the V_ip6_auto_linklocal configuration to
191 * give a reasonable default behavior.
193 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
194 (ifp->if_flags & IFF_LOOPBACK))
195 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
197 * A loopback interface does not need to accept RTADV.
198 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
199 * default regardless of the V_ip6_accept_rtadv configuration to
200 * prevent the interface from accepting RA messages arrived
201 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
203 if (V_ip6_accept_rtadv &&
204 !(ifp->if_flags & IFF_LOOPBACK) &&
205 (ifp->if_type != IFT_BRIDGE))
206 nd->flags |= ND6_IFF_ACCEPT_RTADV;
207 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
208 nd->flags |= ND6_IFF_NO_RADR;
210 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
211 nd6_setmtu0(ifp, nd);
217 nd6_ifdetach(struct nd_ifinfo *nd)
224 * Reset ND level link MTU. This function is called when the physical MTU
225 * changes, which means we might have to adjust the ND level MTU.
228 nd6_setmtu(struct ifnet *ifp)
231 nd6_setmtu0(ifp, ND_IFINFO(ifp));
234 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
236 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
240 omaxmtu = ndi->maxmtu;
242 switch (ifp->if_type) {
244 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
247 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
250 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
253 ndi->maxmtu = ifp->if_mtu;
258 * Decreasing the interface MTU under IPV6 minimum MTU may cause
259 * undesirable situation. We thus notify the operator of the change
260 * explicitly. The check for omaxmtu is necessary to restrict the
261 * log to the case of changing the MTU, not initializing it.
263 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
264 log(LOG_NOTICE, "nd6_setmtu0: "
265 "new link MTU on %s (%lu) is too small for IPv6\n",
266 if_name(ifp), (unsigned long)ndi->maxmtu);
269 if (ndi->maxmtu > V_in6_maxmtu)
270 in6_setmaxmtu(); /* check all interfaces just in case */
275 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
278 bzero(ndopts, sizeof(*ndopts));
279 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
281 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
284 ndopts->nd_opts_done = 1;
285 ndopts->nd_opts_search = NULL;
290 * Take one ND option.
293 nd6_option(union nd_opts *ndopts)
295 struct nd_opt_hdr *nd_opt;
298 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
299 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
301 if (ndopts->nd_opts_search == NULL)
303 if (ndopts->nd_opts_done)
306 nd_opt = ndopts->nd_opts_search;
308 /* make sure nd_opt_len is inside the buffer */
309 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
310 bzero(ndopts, sizeof(*ndopts));
314 olen = nd_opt->nd_opt_len << 3;
317 * Message validation requires that all included
318 * options have a length that is greater than zero.
320 bzero(ndopts, sizeof(*ndopts));
324 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
325 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
326 /* option overruns the end of buffer, invalid */
327 bzero(ndopts, sizeof(*ndopts));
329 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
330 /* reached the end of options chain */
331 ndopts->nd_opts_done = 1;
332 ndopts->nd_opts_search = NULL;
338 * Parse multiple ND options.
339 * This function is much easier to use, for ND routines that do not need
340 * multiple options of the same type.
343 nd6_options(union nd_opts *ndopts)
345 struct nd_opt_hdr *nd_opt;
348 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
349 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
351 if (ndopts->nd_opts_search == NULL)
355 nd_opt = nd6_option(ndopts);
356 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
358 * Message validation requires that all included
359 * options have a length that is greater than zero.
361 ICMP6STAT_INC(icp6s_nd_badopt);
362 bzero(ndopts, sizeof(*ndopts));
369 switch (nd_opt->nd_opt_type) {
370 case ND_OPT_SOURCE_LINKADDR:
371 case ND_OPT_TARGET_LINKADDR:
373 case ND_OPT_REDIRECTED_HEADER:
375 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
377 "duplicated ND6 option found (type=%d)\n",
378 nd_opt->nd_opt_type));
381 ndopts->nd_opt_array[nd_opt->nd_opt_type]
385 case ND_OPT_PREFIX_INFORMATION:
386 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
387 ndopts->nd_opt_array[nd_opt->nd_opt_type]
390 ndopts->nd_opts_pi_end =
391 (struct nd_opt_prefix_info *)nd_opt;
393 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
394 case ND_OPT_RDNSS: /* RFC 6106 */
395 case ND_OPT_DNSSL: /* RFC 6106 */
397 * Silently ignore options we know and do not care about
403 * Unknown options must be silently ignored,
404 * to accomodate future extension to the protocol.
407 "nd6_options: unsupported option %d - "
408 "option ignored\n", nd_opt->nd_opt_type));
413 if (i > V_nd6_maxndopt) {
414 ICMP6STAT_INC(icp6s_nd_toomanyopt);
415 nd6log((LOG_INFO, "too many loop in nd opt\n"));
419 if (ndopts->nd_opts_done)
427 * ND6 timer routine to handle ND6 entries
430 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
434 LLE_WLOCK_ASSERT(ln);
439 canceled = callout_stop(&ln->ln_timer_ch);
441 ln->la_expire = time_uptime + tick / hz;
443 if (tick > INT_MAX) {
444 ln->ln_ntick = tick - INT_MAX;
445 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
446 nd6_llinfo_timer, ln);
449 canceled = callout_reset(&ln->ln_timer_ch, tick,
450 nd6_llinfo_timer, ln);
458 nd6_llinfo_settimer(struct llentry *ln, long tick)
462 nd6_llinfo_settimer_locked(ln, tick);
467 nd6_llinfo_timer(void *arg)
470 struct in6_addr *dst;
472 struct nd_ifinfo *ndi = NULL;
474 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
475 ln = (struct llentry *)arg;
477 if (callout_pending(&ln->la_timer)) {
479 * Here we are a bit odd here in the treatment of
480 * active/pending. If the pending bit is set, it got
481 * rescheduled before I ran. The active
482 * bit we ignore, since if it was stopped
483 * in ll_tablefree() and was currently running
484 * it would have return 0 so the code would
485 * not have deleted it since the callout could
486 * not be stopped so we want to go through
487 * with the delete here now. If the callout
488 * was restarted, the pending bit will be back on and
489 * we just want to bail since the callout_reset would
490 * return 1 and our reference would have been removed
491 * by nd6_llinfo_settimer_locked above since canceled
497 ifp = ln->lle_tbl->llt_ifp;
498 CURVNET_SET(ifp->if_vnet);
500 if (ln->ln_ntick > 0) {
501 if (ln->ln_ntick > INT_MAX) {
502 ln->ln_ntick -= INT_MAX;
503 nd6_llinfo_settimer_locked(ln, INT_MAX);
506 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
511 ndi = ND_IFINFO(ifp);
512 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
513 if (ln->la_flags & LLE_STATIC) {
517 if (ln->la_flags & LLE_DELETED) {
518 (void)nd6_free(ln, 0);
523 switch (ln->ln_state) {
524 case ND6_LLINFO_INCOMPLETE:
525 if (ln->la_asked < V_nd6_mmaxtries) {
527 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
529 nd6_ns_output(ifp, NULL, dst, ln, NULL);
532 struct mbuf *m = ln->la_hold;
537 * assuming every packet in la_hold has the
538 * same IP header. Send error after unlock.
543 clear_llinfo_pqueue(ln);
545 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT);
546 (void)nd6_free(ln, 0);
549 icmp6_error2(m, ICMP6_DST_UNREACH,
550 ICMP6_DST_UNREACH_ADDR, 0, ifp);
553 case ND6_LLINFO_REACHABLE:
554 if (!ND6_LLINFO_PERMANENT(ln)) {
555 ln->ln_state = ND6_LLINFO_STALE;
556 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
560 case ND6_LLINFO_STALE:
561 /* Garbage Collection(RFC 2461 5.3) */
562 if (!ND6_LLINFO_PERMANENT(ln)) {
563 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
564 (void)nd6_free(ln, 1);
569 case ND6_LLINFO_DELAY:
570 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
573 ln->ln_state = ND6_LLINFO_PROBE;
574 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
576 nd6_ns_output(ifp, dst, dst, ln, NULL);
579 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
580 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
583 case ND6_LLINFO_PROBE:
584 if (ln->la_asked < V_nd6_umaxtries) {
586 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
588 nd6_ns_output(ifp, dst, dst, ln, NULL);
591 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
592 (void)nd6_free(ln, 0);
597 panic("%s: paths in a dark night can be confusing: %d",
598 __func__, ln->ln_state);
608 * ND6 timer routine to expire default route list and prefix list
613 CURVNET_SET((struct vnet *) arg);
614 struct nd_defrouter *dr, *ndr;
615 struct nd_prefix *pr, *npr;
616 struct in6_ifaddr *ia6, *nia6;
618 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
621 /* expire default router list */
622 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
623 if (dr->expire && dr->expire < time_uptime)
628 * expire interface addresses.
629 * in the past the loop was inside prefix expiry processing.
630 * However, from a stricter speci-confrmance standpoint, we should
631 * rather separate address lifetimes and prefix lifetimes.
633 * XXXRW: in6_ifaddrhead locking.
636 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
637 /* check address lifetime */
638 if (IFA6_IS_INVALID(ia6)) {
642 * If the expiring address is temporary, try
643 * regenerating a new one. This would be useful when
644 * we suspended a laptop PC, then turned it on after a
645 * period that could invalidate all temporary
646 * addresses. Although we may have to restart the
647 * loop (see below), it must be after purging the
648 * address. Otherwise, we'd see an infinite loop of
651 if (V_ip6_use_tempaddr &&
652 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
653 if (regen_tmpaddr(ia6) == 0)
657 in6_purgeaddr(&ia6->ia_ifa);
660 goto addrloop; /* XXX: see below */
661 } else if (IFA6_IS_DEPRECATED(ia6)) {
662 int oldflags = ia6->ia6_flags;
664 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
667 * If a temporary address has just become deprecated,
668 * regenerate a new one if possible.
670 if (V_ip6_use_tempaddr &&
671 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
672 (oldflags & IN6_IFF_DEPRECATED) == 0) {
674 if (regen_tmpaddr(ia6) == 0) {
676 * A new temporary address is
678 * XXX: this means the address chain
679 * has changed while we are still in
680 * the loop. Although the change
681 * would not cause disaster (because
682 * it's not a deletion, but an
683 * addition,) we'd rather restart the
684 * loop just for safety. Or does this
685 * significantly reduce performance??
692 * A new RA might have made a deprecated address
695 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
699 /* expire prefix list */
700 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
702 * check prefix lifetime.
703 * since pltime is just for autoconf, pltime processing for
704 * prefix is not necessary.
706 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
707 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) {
710 * address expiration and prefix expiration are
711 * separate. NEVER perform in6_purgeaddr here.
720 * ia6 - deprecated/invalidated temporary address
723 regen_tmpaddr(struct in6_ifaddr *ia6)
727 struct in6_ifaddr *public_ifa6 = NULL;
729 ifp = ia6->ia_ifa.ifa_ifp;
731 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 struct in6_ifaddr *it6;
734 if (ifa->ifa_addr->sa_family != AF_INET6)
737 it6 = (struct in6_ifaddr *)ifa;
739 /* ignore no autoconf addresses. */
740 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
743 /* ignore autoconf addresses with different prefixes. */
744 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
748 * Now we are looking at an autoconf address with the same
749 * prefix as ours. If the address is temporary and is still
750 * preferred, do not create another one. It would be rare, but
751 * could happen, for example, when we resume a laptop PC after
754 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
755 !IFA6_IS_DEPRECATED(it6)) {
761 * This is a public autoconf address that has the same prefix
762 * as ours. If it is preferred, keep it. We can't break the
763 * loop here, because there may be a still-preferred temporary
764 * address with the prefix.
766 if (!IFA6_IS_DEPRECATED(it6))
769 if (public_ifa6 != NULL)
770 ifa_ref(&public_ifa6->ia_ifa);
771 IF_ADDR_RUNLOCK(ifp);
773 if (public_ifa6 != NULL) {
776 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
777 ifa_free(&public_ifa6->ia_ifa);
778 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
779 " tmp addr,errno=%d\n", e);
782 ifa_free(&public_ifa6->ia_ifa);
790 * Nuke neighbor cache/prefix/default router management table, right before
794 nd6_purge(struct ifnet *ifp)
796 struct nd_defrouter *dr, *ndr;
797 struct nd_prefix *pr, *npr;
800 * Nuke default router list entries toward ifp.
801 * We defer removal of default router list entries that is installed
802 * in the routing table, in order to keep additional side effects as
805 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
813 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
821 /* Nuke prefix list entries toward ifp */
822 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
823 if (pr->ndpr_ifp == ifp) {
825 * Because if_detach() does *not* release prefixes
826 * while purging addresses the reference count will
827 * still be above zero. We therefore reset it to
828 * make sure that the prefix really gets purged.
833 * Previously, pr->ndpr_addr is removed as well,
834 * but I strongly believe we don't have to do it.
835 * nd6_purge() is only called from in6_ifdetach(),
836 * which removes all the associated interface addresses
838 * (jinmei@kame.net 20010129)
844 /* cancel default outgoing interface setting */
845 if (V_nd6_defifindex == ifp->if_index)
846 nd6_setdefaultiface(0);
848 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
849 /* Refresh default router list. */
854 * We do not nuke the neighbor cache entries here any more
855 * because the neighbor cache is kept in if_afdata[AF_INET6].
856 * nd6_purge() is invoked by in6_ifdetach() which is called
857 * from if_detach() where everything gets purged. So let
858 * in6_domifdetach() do the actual L2 table purging work.
863 * the caller acquires and releases the lock on the lltbls
864 * Returns the llentry locked
867 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
869 struct sockaddr_in6 sin6;
873 bzero(&sin6, sizeof(sin6));
874 sin6.sin6_len = sizeof(struct sockaddr_in6);
875 sin6.sin6_family = AF_INET6;
876 sin6.sin6_addr = *addr6;
878 IF_AFDATA_LOCK_ASSERT(ifp);
881 if (flags & ND6_CREATE)
882 llflags |= LLE_CREATE;
883 if (flags & ND6_EXCLUSIVE)
884 llflags |= LLE_EXCLUSIVE;
886 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
887 if ((ln != NULL) && (llflags & LLE_CREATE))
888 ln->ln_state = ND6_LLINFO_NOSTATE;
894 * Test whether a given IPv6 address is a neighbor or not, ignoring
895 * the actual neighbor cache. The neighbor cache is ignored in order
896 * to not reenter the routing code from within itself.
899 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
901 struct nd_prefix *pr;
902 struct ifaddr *dstaddr;
905 * A link-local address is always a neighbor.
906 * XXX: a link does not necessarily specify a single interface.
908 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
909 struct sockaddr_in6 sin6_copy;
913 * We need sin6_copy since sa6_recoverscope() may modify the
917 if (sa6_recoverscope(&sin6_copy))
918 return (0); /* XXX: should be impossible */
919 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
921 if (sin6_copy.sin6_scope_id == zone)
928 * If the address matches one of our addresses,
929 * it should be a neighbor.
930 * If the address matches one of our on-link prefixes, it should be a
933 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
934 if (pr->ndpr_ifp != ifp)
937 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
940 /* Always use the default FIB here. */
941 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
942 0, 0, RT_DEFAULT_FIB);
946 * This is the case where multiple interfaces
947 * have the same prefix, but only one is installed
948 * into the routing table and that prefix entry
949 * is not the one being examined here. In the case
950 * where RADIX_MPATH is enabled, multiple route
951 * entries (of the same rt_key value) will be
952 * installed because the interface addresses all
955 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
956 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
963 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
964 &addr->sin6_addr, &pr->ndpr_mask))
969 * If the address is assigned on the node of the other side of
970 * a p2p interface, the address should be a neighbor.
972 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
973 if (dstaddr != NULL) {
974 if (dstaddr->ifa_ifp == ifp) {
982 * If the default router list is empty, all addresses are regarded
983 * as on-link, and thus, as a neighbor.
985 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
986 TAILQ_EMPTY(&V_nd_defrouter) &&
987 V_nd6_defifindex == ifp->if_index) {
996 * Detect if a given IPv6 address identifies a neighbor on a given link.
997 * XXX: should take care of the destination of a p2p link?
1000 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
1002 struct llentry *lle;
1005 IF_AFDATA_UNLOCK_ASSERT(ifp);
1006 if (nd6_is_new_addr_neighbor(addr, ifp))
1010 * Even if the address matches none of our addresses, it might be
1011 * in the neighbor cache.
1013 IF_AFDATA_RLOCK(ifp);
1014 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1018 IF_AFDATA_RUNLOCK(ifp);
1023 * Free an nd6 llinfo entry.
1024 * Since the function would cause significant changes in the kernel, DO NOT
1025 * make it global, unless you have a strong reason for the change, and are sure
1026 * that the change is safe.
1028 static struct llentry *
1029 nd6_free(struct llentry *ln, int gc)
1031 struct llentry *next;
1032 struct nd_defrouter *dr;
1035 LLE_WLOCK_ASSERT(ln);
1038 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1039 * even though it is not harmful, it was not really necessary.
1043 nd6_llinfo_settimer_locked(ln, -1);
1045 ifp = ln->lle_tbl->llt_ifp;
1047 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1048 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1050 if (dr != NULL && dr->expire &&
1051 ln->ln_state == ND6_LLINFO_STALE && gc) {
1053 * If the reason for the deletion is just garbage
1054 * collection, and the neighbor is an active default
1055 * router, do not delete it. Instead, reset the GC
1056 * timer using the router's lifetime.
1057 * Simply deleting the entry would affect default
1058 * router selection, which is not necessarily a good
1059 * thing, especially when we're using router preference
1061 * XXX: the check for ln_state would be redundant,
1062 * but we intentionally keep it just in case.
1064 if (dr->expire > time_uptime)
1065 nd6_llinfo_settimer_locked(ln,
1066 (dr->expire - time_uptime) * hz);
1068 nd6_llinfo_settimer_locked(ln,
1069 (long)V_nd6_gctimer * hz);
1071 next = LIST_NEXT(ln, lle_next);
1079 * Unreachablity of a router might affect the default
1080 * router selection and on-link detection of advertised
1085 * Temporarily fake the state to choose a new default
1086 * router and to perform on-link determination of
1087 * prefixes correctly.
1088 * Below the state will be set correctly,
1089 * or the entry itself will be deleted.
1091 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1094 if (ln->ln_router || dr) {
1097 * We need to unlock to avoid a LOR with rt6_flush() with the
1098 * rnh and for the calls to pfxlist_onlink_check() and
1099 * defrouter_select() in the block further down for calls
1100 * into nd6_lookup(). We still hold a ref.
1105 * rt6_flush must be called whether or not the neighbor
1106 * is in the Default Router List.
1107 * See a corresponding comment in nd6_na_input().
1109 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1114 * Since defrouter_select() does not affect the
1115 * on-link determination and MIP6 needs the check
1116 * before the default router selection, we perform
1119 pfxlist_onlink_check();
1122 * Refresh default router list.
1127 if (ln->ln_router || dr)
1132 * Before deleting the entry, remember the next entry as the
1133 * return value. We need this because pfxlist_onlink_check() above
1134 * might have freed other entries (particularly the old next entry) as
1135 * a side effect (XXX).
1137 next = LIST_NEXT(ln, lle_next);
1140 * Save to unlock. We still hold an extra reference and will not
1141 * free(9) in llentry_free() if someone else holds one as well.
1144 IF_AFDATA_LOCK(ifp);
1147 /* Guard against race with other llentry_free(). */
1148 if (ln->la_flags & LLE_LINKED) {
1152 LLE_FREE_LOCKED(ln);
1154 IF_AFDATA_UNLOCK(ifp);
1160 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1162 * XXX cost-effective methods?
1165 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1170 if ((dst6 == NULL) || (rt == NULL))
1174 IF_AFDATA_RLOCK(ifp);
1175 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1176 IF_AFDATA_RUNLOCK(ifp);
1180 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1184 * if we get upper-layer reachability confirmation many times,
1185 * it is possible we have false information.
1189 if (ln->ln_byhint > V_nd6_maxnudhint) {
1194 ln->ln_state = ND6_LLINFO_REACHABLE;
1195 if (!ND6_LLINFO_PERMANENT(ln)) {
1196 nd6_llinfo_settimer_locked(ln,
1197 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1205 * Rejuvenate this function for routing operations related
1209 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1211 struct sockaddr_in6 *gateway;
1212 struct nd_defrouter *dr;
1216 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1227 * Only indirect routes are interesting.
1229 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1232 * check for default route
1234 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1235 &SIN6(rt_key(rt))->sin6_addr)) {
1237 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1247 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1249 struct in6_drlist *drl = (struct in6_drlist *)data;
1250 struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1251 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1252 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1253 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1254 struct nd_defrouter *dr;
1255 struct nd_prefix *pr;
1256 int i = 0, error = 0;
1258 if (ifp->if_afdata[AF_INET6] == NULL)
1259 return (EPFNOSUPPORT);
1261 case SIOCGDRLST_IN6:
1263 * obsolete API, use sysctl under net.inet6.icmp6
1265 bzero(drl, sizeof(*drl));
1266 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1269 drl->defrouter[i].rtaddr = dr->rtaddr;
1270 in6_clearscope(&drl->defrouter[i].rtaddr);
1272 drl->defrouter[i].flags = dr->flags;
1273 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1274 drl->defrouter[i].expire = dr->expire +
1275 (time_second - time_uptime);
1276 drl->defrouter[i].if_index = dr->ifp->if_index;
1280 case SIOCGPRLST_IN6:
1282 * obsolete API, use sysctl under net.inet6.icmp6
1284 * XXX the structure in6_prlist was changed in backward-
1285 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6,
1286 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1289 * XXX meaning of fields, especialy "raflags", is very
1290 * differnet between RA prefix list and RR/static prefix list.
1291 * how about separating ioctls into two?
1293 bzero(oprl, sizeof(*oprl));
1294 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1295 struct nd_pfxrouter *pfr;
1300 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1301 oprl->prefix[i].raflags = pr->ndpr_raf;
1302 oprl->prefix[i].prefixlen = pr->ndpr_plen;
1303 oprl->prefix[i].vltime = pr->ndpr_vltime;
1304 oprl->prefix[i].pltime = pr->ndpr_pltime;
1305 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1306 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1307 oprl->prefix[i].expire = 0;
1311 /* XXX: we assume time_t is signed. */
1314 ((sizeof(maxexpire) * 8) - 1));
1315 if (pr->ndpr_vltime <
1316 maxexpire - pr->ndpr_lastupdate) {
1317 oprl->prefix[i].expire =
1318 pr->ndpr_lastupdate +
1320 (time_second - time_uptime);
1322 oprl->prefix[i].expire = maxexpire;
1326 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1328 #define RTRADDR oprl->prefix[i].advrtr[j]
1329 RTRADDR = pfr->router->rtaddr;
1330 in6_clearscope(&RTRADDR);
1335 oprl->prefix[i].advrtrs = j;
1336 oprl->prefix[i].origin = PR_ORIG_RA;
1342 case OSIOCGIFINFO_IN6:
1344 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1345 bzero(&ND, sizeof(ND));
1346 ND.linkmtu = IN6_LINKMTU(ifp);
1347 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1348 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1349 ND.reachable = ND_IFINFO(ifp)->reachable;
1350 ND.retrans = ND_IFINFO(ifp)->retrans;
1351 ND.flags = ND_IFINFO(ifp)->flags;
1352 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1353 ND.chlim = ND_IFINFO(ifp)->chlim;
1355 case SIOCGIFINFO_IN6:
1356 ND = *ND_IFINFO(ifp);
1358 case SIOCSIFINFO_IN6:
1360 * used to change host variables from userland.
1361 * intented for a use on router to reflect RA configurations.
1363 /* 0 means 'unspecified' */
1364 if (ND.linkmtu != 0) {
1365 if (ND.linkmtu < IPV6_MMTU ||
1366 ND.linkmtu > IN6_LINKMTU(ifp)) {
1370 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1373 if (ND.basereachable != 0) {
1374 int obasereachable = ND_IFINFO(ifp)->basereachable;
1376 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1377 if (ND.basereachable != obasereachable)
1378 ND_IFINFO(ifp)->reachable =
1379 ND_COMPUTE_RTIME(ND.basereachable);
1381 if (ND.retrans != 0)
1382 ND_IFINFO(ifp)->retrans = ND.retrans;
1384 ND_IFINFO(ifp)->chlim = ND.chlim;
1386 case SIOCSIFINFO_FLAGS:
1389 struct in6_ifaddr *ia;
1391 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1392 !(ND.flags & ND6_IFF_IFDISABLED)) {
1393 /* ifdisabled 1->0 transision */
1396 * If the interface is marked as ND6_IFF_IFDISABLED and
1397 * has an link-local address with IN6_IFF_DUPLICATED,
1398 * do not clear ND6_IFF_IFDISABLED.
1399 * See RFC 4862, Section 5.4.5.
1401 int duplicated_linklocal = 0;
1404 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1405 if (ifa->ifa_addr->sa_family != AF_INET6)
1407 ia = (struct in6_ifaddr *)ifa;
1408 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1409 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1410 duplicated_linklocal = 1;
1414 IF_ADDR_RUNLOCK(ifp);
1416 if (duplicated_linklocal) {
1417 ND.flags |= ND6_IFF_IFDISABLED;
1418 log(LOG_ERR, "Cannot enable an interface"
1419 " with a link-local address marked"
1422 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1423 if (ifp->if_flags & IFF_UP)
1426 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1427 (ND.flags & ND6_IFF_IFDISABLED)) {
1428 /* ifdisabled 0->1 transision */
1429 /* Mark all IPv6 address as tentative. */
1431 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1433 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1434 if (ifa->ifa_addr->sa_family != AF_INET6)
1436 ia = (struct in6_ifaddr *)ifa;
1437 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1439 IF_ADDR_RUNLOCK(ifp);
1442 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1443 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1444 /* auto_linklocal 0->1 transision */
1446 /* If no link-local address on ifp, configure */
1447 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1448 in6_ifattach(ifp, NULL);
1449 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1450 ifp->if_flags & IFF_UP) {
1452 * When the IF already has
1453 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1454 * address is assigned, and IFF_UP, try to
1457 int haslinklocal = 0;
1460 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1461 if (ifa->ifa_addr->sa_family != AF_INET6)
1463 ia = (struct in6_ifaddr *)ifa;
1464 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) {
1469 IF_ADDR_RUNLOCK(ifp);
1471 in6_ifattach(ifp, NULL);
1475 ND_IFINFO(ifp)->flags = ND.flags;
1478 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1479 /* sync kernel routing table with the default router list */
1483 case SIOCSPFXFLUSH_IN6:
1485 /* flush all the prefix advertised by routers */
1486 struct nd_prefix *pr, *next;
1488 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1489 struct in6_ifaddr *ia, *ia_next;
1491 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1494 /* do we really have to remove addresses as well? */
1495 /* XXXRW: in6_ifaddrhead locking. */
1496 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1498 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1501 if (ia->ia6_ndpr == pr)
1502 in6_purgeaddr(&ia->ia_ifa);
1508 case SIOCSRTRFLUSH_IN6:
1510 /* flush all the default routers */
1511 struct nd_defrouter *dr, *next;
1514 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1520 case SIOCGNBRINFO_IN6:
1523 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1525 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1528 IF_AFDATA_RLOCK(ifp);
1529 ln = nd6_lookup(&nb_addr, 0, ifp);
1530 IF_AFDATA_RUNLOCK(ifp);
1536 nbi->state = ln->ln_state;
1537 nbi->asked = ln->la_asked;
1538 nbi->isrouter = ln->ln_router;
1539 if (ln->la_expire == 0)
1542 nbi->expire = ln->la_expire +
1543 (time_second - time_uptime);
1547 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1548 ndif->ifindex = V_nd6_defifindex;
1550 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1551 return (nd6_setdefaultiface(ndif->ifindex));
1557 * Create neighbor cache entry and cache link-layer address,
1558 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1561 * code - type dependent information
1564 * The caller of this function already acquired the ndp
1565 * cache table lock because the cache entry is returned.
1568 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1569 int lladdrlen, int type, int code)
1571 struct llentry *ln = NULL;
1578 uint16_t router = 0;
1579 struct sockaddr_in6 sin6;
1580 struct mbuf *chain = NULL;
1581 int static_route = 0;
1583 IF_AFDATA_UNLOCK_ASSERT(ifp);
1585 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1586 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1588 /* nothing must be updated for unspecified address */
1589 if (IN6_IS_ADDR_UNSPECIFIED(from))
1593 * Validation about ifp->if_addrlen and lladdrlen must be done in
1596 * XXX If the link does not have link-layer adderss, what should
1597 * we do? (ifp->if_addrlen == 0)
1598 * Spec says nothing in sections for RA, RS and NA. There's small
1599 * description on it in NS section (RFC 2461 7.2.3).
1601 flags = lladdr ? ND6_EXCLUSIVE : 0;
1602 IF_AFDATA_RLOCK(ifp);
1603 ln = nd6_lookup(from, flags, ifp);
1604 IF_AFDATA_RUNLOCK(ifp);
1606 flags |= ND6_EXCLUSIVE;
1607 IF_AFDATA_LOCK(ifp);
1608 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1609 IF_AFDATA_UNLOCK(ifp);
1612 /* do nothing if static ndp is set */
1613 if (ln->la_flags & LLE_STATIC) {
1622 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1623 if (olladdr && lladdr) {
1624 llchange = bcmp(lladdr, &ln->ll_addr,
1630 * newentry olladdr lladdr llchange (*=record)
1633 * 0 n y -- (3) * STALE
1635 * 0 y y y (5) * STALE
1636 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1637 * 1 -- y -- (7) * STALE
1640 if (lladdr) { /* (3-5) and (7) */
1642 * Record source link-layer address
1643 * XXX is it dependent to ifp->if_type?
1645 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1646 ln->la_flags |= LLE_VALID;
1647 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1651 if ((!olladdr && lladdr != NULL) || /* (3) */
1652 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1654 newstate = ND6_LLINFO_STALE;
1655 } else /* (1-2,4) */
1659 if (lladdr == NULL) /* (6) */
1660 newstate = ND6_LLINFO_NOSTATE;
1662 newstate = ND6_LLINFO_STALE;
1667 * Update the state of the neighbor cache.
1669 ln->ln_state = newstate;
1671 if (ln->ln_state == ND6_LLINFO_STALE) {
1672 if (ln->la_hold != NULL)
1673 nd6_grab_holdchain(ln, &chain, &sin6);
1674 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1675 /* probe right away */
1676 nd6_llinfo_settimer_locked((void *)ln, 0);
1681 * ICMP6 type dependent behavior.
1683 * NS: clear IsRouter if new entry
1684 * RS: clear IsRouter
1685 * RA: set IsRouter if there's lladdr
1686 * redir: clear IsRouter if new entry
1689 * The spec says that we must set IsRouter in the following cases:
1690 * - If lladdr exist, set IsRouter. This means (1-5).
1691 * - If it is old entry (!newentry), set IsRouter. This means (7).
1692 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1693 * A quetion arises for (1) case. (1) case has no lladdr in the
1694 * neighbor cache, this is similar to (6).
1695 * This case is rare but we figured that we MUST NOT set IsRouter.
1697 * newentry olladdr lladdr llchange NS RS RA redir
1699 * 0 n n -- (1) c ? s
1700 * 0 y n -- (2) c s s
1701 * 0 n y -- (3) c s s
1704 * 1 -- n -- (6) c c c s
1705 * 1 -- y -- (7) c c s c s
1709 switch (type & 0xff) {
1710 case ND_NEIGHBOR_SOLICIT:
1712 * New entry must have is_router flag cleared.
1714 if (is_newentry) /* (6-7) */
1719 * If the icmp is a redirect to a better router, always set the
1720 * is_router flag. Otherwise, if the entry is newly created,
1721 * clear the flag. [RFC 2461, sec 8.3]
1723 if (code == ND_REDIRECT_ROUTER)
1725 else if (is_newentry) /* (6-7) */
1728 case ND_ROUTER_SOLICIT:
1730 * is_router flag must always be cleared.
1734 case ND_ROUTER_ADVERT:
1736 * Mark an entry with lladdr as a router.
1738 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1739 (is_newentry && lladdr)) { /* (7) */
1746 static_route = (ln->la_flags & LLE_STATIC);
1747 router = ln->ln_router;
1749 if (flags & ND6_EXCLUSIVE)
1757 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
1760 * When the link-layer address of a router changes, select the
1761 * best router again. In particular, when the neighbor entry is newly
1762 * created, it might affect the selection policy.
1763 * Question: can we restrict the first condition to the "is_newentry"
1765 * XXX: when we hear an RA from a new router with the link-layer
1766 * address option, defrouter_select() is called twice, since
1767 * defrtrlist_update called the function as well. However, I believe
1768 * we can compromise the overhead, since it only happens the first
1770 * XXX: although defrouter_select() should not have a bad effect
1771 * for those are not autoconfigured hosts, we explicitly avoid such
1774 if (do_update && router &&
1775 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1777 * guaranteed recursion
1785 if (flags & ND6_EXCLUSIVE)
1796 nd6_slowtimo(void *arg)
1798 CURVNET_SET((struct vnet *) arg);
1799 struct nd_ifinfo *nd6if;
1802 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1803 nd6_slowtimo, curvnet);
1804 IFNET_RLOCK_NOSLEEP();
1805 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1806 if (ifp->if_afdata[AF_INET6] == NULL)
1808 nd6if = ND_IFINFO(ifp);
1809 if (nd6if->basereachable && /* already initialized */
1810 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1812 * Since reachable time rarely changes by router
1813 * advertisements, we SHOULD insure that a new random
1814 * value gets recomputed at least once every few hours.
1817 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1818 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1821 IFNET_RUNLOCK_NOSLEEP();
1826 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
1827 struct sockaddr_in6 *sin6)
1830 LLE_WLOCK_ASSERT(ln);
1832 *chain = ln->la_hold;
1834 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6));
1836 if (ln->ln_state == ND6_LLINFO_STALE) {
1839 * The first time we send a packet to a
1840 * neighbor whose entry is STALE, we have
1841 * to change the state to DELAY and a sets
1842 * a timer to expire in DELAY_FIRST_PROBE_TIME
1843 * seconds to ensure do neighbor unreachability
1844 * detection on expiration.
1848 ln->ln_state = ND6_LLINFO_DELAY;
1849 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1854 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1855 struct sockaddr_in6 *dst)
1859 struct ip6_hdr *ip6;
1863 mac_netinet6_nd6_send(ifp, m);
1867 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
1868 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
1869 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
1870 * to be diverted to user space. When re-injected into the kernel,
1871 * send_output() will directly dispatch them to the outgoing interface.
1873 if (send_sendso_input_hook != NULL) {
1874 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
1876 ip6 = mtod(m, struct ip6_hdr *);
1877 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
1878 /* Use the SEND socket */
1879 error = send_sendso_input_hook(m, ifp, SND_OUT,
1881 /* -1 == no app on SEND socket */
1882 if (error == 0 || error != -1)
1887 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1888 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
1889 mtod(m, struct ip6_hdr *));
1891 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1894 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL);
1899 * IPv6 packet output - light version.
1900 * Checks if destination LLE exists and is in proper state
1901 * (e.g no modification required). If not true, fall back to
1905 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1906 struct sockaddr_in6 *dst, struct rtentry *rt0)
1908 struct llentry *ln = NULL;
1910 /* discard the packet if IPv6 operation is disabled on the interface */
1911 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1913 return (ENETDOWN); /* better error? */
1916 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1919 if (nd6_need_cache(ifp) == 0)
1922 IF_AFDATA_RLOCK(ifp);
1923 ln = nd6_lookup(&dst->sin6_addr, 0, ifp);
1924 IF_AFDATA_RUNLOCK(ifp);
1927 * Perform fast path for the following cases:
1928 * 1) lle state is REACHABLE
1929 * 2) lle state is DELAY (NS message sentNS message sent)
1931 * Every other case involves lle modification, so we handle
1934 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE &&
1935 ln->ln_state != ND6_LLINFO_DELAY)) {
1936 /* Fall back to slow processing path */
1939 return (nd6_output_lle(ifp, origifp, m, dst));
1946 return (nd6_output_ifp(ifp, origifp, m, dst));
1951 * Output IPv6 packet - heavy version.
1952 * Function assume that either
1953 * 1) destination LLE does not exist, is invalid or stale, so
1954 * ND6_EXCLUSIVE lock needs to be acquired
1955 * 2) destination lle is provided (with ND6_EXCLUSIVE lock),
1956 * in that case packets are queued in &chain.
1960 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1961 struct sockaddr_in6 *dst)
1963 struct llentry *lle = NULL;
1966 KASSERT(m != NULL, ("NULL mbuf, nothing to send"));
1967 /* discard the packet if IPv6 operation is disabled on the interface */
1968 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1970 return (ENETDOWN); /* better error? */
1973 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1976 if (nd6_need_cache(ifp) == 0)
1980 * Address resolution or Neighbor Unreachability Detection
1982 * At this point, the destination of the packet must be a unicast
1983 * or an anycast address(i.e. not a multicast).
1986 IF_AFDATA_RLOCK(ifp);
1987 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
1988 IF_AFDATA_RUNLOCK(ifp);
1989 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1991 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1992 * the condition below is not very efficient. But we believe
1993 * it is tolerable, because this should be a rare case.
1995 flags = ND6_CREATE | ND6_EXCLUSIVE;
1996 IF_AFDATA_LOCK(ifp);
1997 lle = nd6_lookup(&dst->sin6_addr, flags, ifp);
1998 IF_AFDATA_UNLOCK(ifp);
2002 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2003 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2004 char ip6buf[INET6_ADDRSTRLEN];
2006 "nd6_output: can't allocate llinfo for %s "
2008 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
2012 goto sendpkt; /* send anyway */
2015 LLE_WLOCK_ASSERT(lle);
2017 /* We don't have to do link-layer address resolution on a p2p link. */
2018 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2019 lle->ln_state < ND6_LLINFO_REACHABLE) {
2020 lle->ln_state = ND6_LLINFO_STALE;
2021 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz);
2025 * The first time we send a packet to a neighbor whose entry is
2026 * STALE, we have to change the state to DELAY and a sets a timer to
2027 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2028 * neighbor unreachability detection on expiration.
2031 if (lle->ln_state == ND6_LLINFO_STALE) {
2033 lle->ln_state = ND6_LLINFO_DELAY;
2034 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz);
2038 * If the neighbor cache entry has a state other than INCOMPLETE
2039 * (i.e. its link-layer address is already resolved), just
2042 if (lle->ln_state > ND6_LLINFO_INCOMPLETE)
2046 * There is a neighbor cache entry, but no ethernet address
2047 * response yet. Append this latest packet to the end of the
2048 * packet queue in the mbuf, unless the number of the packet
2049 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2050 * the oldest packet in the queue will be removed.
2052 if (lle->ln_state == ND6_LLINFO_NOSTATE)
2053 lle->ln_state = ND6_LLINFO_INCOMPLETE;
2055 if (lle->la_hold != NULL) {
2056 struct mbuf *m_hold;
2060 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2062 if (m_hold->m_nextpkt == NULL) {
2063 m_hold->m_nextpkt = m;
2067 while (i >= V_nd6_maxqueuelen) {
2068 m_hold = lle->la_hold;
2069 lle->la_hold = lle->la_hold->m_nextpkt;
2078 * If there has been no NS for the neighbor after entering the
2079 * INCOMPLETE state, send the first solicitation.
2081 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) {
2084 nd6_llinfo_settimer_locked(lle,
2085 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2087 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL);
2089 /* We did the lookup so we need to do the unlock here. */
2099 return (nd6_output_ifp(ifp, origifp, m, dst));
2104 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2105 struct sockaddr_in6 *dst)
2107 struct mbuf *m, *m_head;
2108 struct ifnet *outifp;
2112 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2119 m_head = m_head->m_nextpkt;
2120 error = nd6_output_ifp(ifp, origifp, m, dst);
2125 * note that intermediate errors are blindly ignored - but this is
2126 * the same convention as used with nd6_output when called by
2134 nd6_need_cache(struct ifnet *ifp)
2137 * XXX: we currently do not make neighbor cache on any interface
2138 * other than ARCnet, Ethernet, FDDI and GIF.
2141 * - unidirectional tunnels needs no ND
2143 switch (ifp->if_type) {
2151 #ifdef IFT_IEEE80211
2154 case IFT_INFINIBAND:
2156 case IFT_PROPVIRTUAL:
2164 * the callers of this function need to be re-worked to drop
2165 * the lle lock, drop here for now
2168 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2169 const struct sockaddr *dst, u_char *desten, struct llentry **lle)
2174 IF_AFDATA_UNLOCK_ASSERT(ifp);
2175 if (m != NULL && m->m_flags & M_MCAST) {
2178 switch (ifp->if_type) {
2184 #ifdef IFT_IEEE80211
2189 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2194 * netbsd can use if_broadcastaddr, but we don't do so
2195 * to reduce # of ifdef.
2197 for (i = 0; i < ifp->if_addrlen; i++)
2205 return (EAFNOSUPPORT);
2211 * the entry should have been created in nd6_store_lladdr
2213 IF_AFDATA_RLOCK(ifp);
2214 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2215 IF_AFDATA_RUNLOCK(ifp);
2216 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2219 /* this could happen, if we could not allocate memory */
2224 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2228 * A *small* use after free race exists here
2234 clear_llinfo_pqueue(struct llentry *ln)
2236 struct mbuf *m_hold, *m_hold_next;
2238 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2239 m_hold_next = m_hold->m_nextpkt;
2247 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2248 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2250 SYSCTL_DECL(_net_inet6_icmp6);
2252 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2253 CTLFLAG_RD, nd6_sysctl_drlist, "");
2254 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2255 CTLFLAG_RD, nd6_sysctl_prlist, "");
2256 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2257 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2258 SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2259 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2262 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2264 struct in6_defrouter d;
2265 struct nd_defrouter *dr;
2271 bzero(&d, sizeof(d));
2272 d.rtaddr.sin6_family = AF_INET6;
2273 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2278 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2279 d.rtaddr.sin6_addr = dr->rtaddr;
2280 error = sa6_recoverscope(&d.rtaddr);
2283 d.flags = dr->flags;
2284 d.rtlifetime = dr->rtlifetime;
2285 d.expire = dr->expire + (time_second - time_uptime);
2286 d.if_index = dr->ifp->if_index;
2287 error = SYSCTL_OUT(req, &d, sizeof(d));
2295 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2297 struct in6_prefix p;
2298 struct sockaddr_in6 s6;
2299 struct nd_prefix *pr;
2300 struct nd_pfxrouter *pfr;
2303 char ip6buf[INET6_ADDRSTRLEN];
2308 bzero(&p, sizeof(p));
2309 p.origin = PR_ORIG_RA;
2310 bzero(&s6, sizeof(s6));
2311 s6.sin6_family = AF_INET6;
2312 s6.sin6_len = sizeof(s6);
2317 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2318 p.prefix = pr->ndpr_prefix;
2319 if (sa6_recoverscope(&p.prefix)) {
2320 log(LOG_ERR, "scope error in prefix list (%s)\n",
2321 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2322 /* XXX: press on... */
2324 p.raflags = pr->ndpr_raf;
2325 p.prefixlen = pr->ndpr_plen;
2326 p.vltime = pr->ndpr_vltime;
2327 p.pltime = pr->ndpr_pltime;
2328 p.if_index = pr->ndpr_ifp->if_index;
2329 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2332 /* XXX: we assume time_t is signed. */
2334 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2335 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2336 p.expire = pr->ndpr_lastupdate +
2338 (time_second - time_uptime);
2340 p.expire = maxexpire;
2342 p.refcnt = pr->ndpr_refcnt;
2343 p.flags = pr->ndpr_stateflags;
2345 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2347 error = SYSCTL_OUT(req, &p, sizeof(p));
2350 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2351 s6.sin6_addr = pfr->router->rtaddr;
2352 if (sa6_recoverscope(&s6))
2354 "scope error in prefix list (%s)\n",
2355 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2356 error = SYSCTL_OUT(req, &s6, sizeof(s6));