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)
230 if (ifp->if_afdata[AF_INET6] == NULL)
233 nd6_setmtu0(ifp, ND_IFINFO(ifp));
236 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
238 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
242 omaxmtu = ndi->maxmtu;
244 switch (ifp->if_type) {
246 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
249 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
252 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
255 ndi->maxmtu = ifp->if_mtu;
260 * Decreasing the interface MTU under IPV6 minimum MTU may cause
261 * undesirable situation. We thus notify the operator of the change
262 * explicitly. The check for omaxmtu is necessary to restrict the
263 * log to the case of changing the MTU, not initializing it.
265 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
266 log(LOG_NOTICE, "nd6_setmtu0: "
267 "new link MTU on %s (%lu) is too small for IPv6\n",
268 if_name(ifp), (unsigned long)ndi->maxmtu);
271 if (ndi->maxmtu > V_in6_maxmtu)
272 in6_setmaxmtu(); /* check all interfaces just in case */
277 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
280 bzero(ndopts, sizeof(*ndopts));
281 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
283 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
286 ndopts->nd_opts_done = 1;
287 ndopts->nd_opts_search = NULL;
292 * Take one ND option.
295 nd6_option(union nd_opts *ndopts)
297 struct nd_opt_hdr *nd_opt;
300 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
301 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
303 if (ndopts->nd_opts_search == NULL)
305 if (ndopts->nd_opts_done)
308 nd_opt = ndopts->nd_opts_search;
310 /* make sure nd_opt_len is inside the buffer */
311 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
312 bzero(ndopts, sizeof(*ndopts));
316 olen = nd_opt->nd_opt_len << 3;
319 * Message validation requires that all included
320 * options have a length that is greater than zero.
322 bzero(ndopts, sizeof(*ndopts));
326 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
327 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
328 /* option overruns the end of buffer, invalid */
329 bzero(ndopts, sizeof(*ndopts));
331 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
332 /* reached the end of options chain */
333 ndopts->nd_opts_done = 1;
334 ndopts->nd_opts_search = NULL;
340 * Parse multiple ND options.
341 * This function is much easier to use, for ND routines that do not need
342 * multiple options of the same type.
345 nd6_options(union nd_opts *ndopts)
347 struct nd_opt_hdr *nd_opt;
350 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
351 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
353 if (ndopts->nd_opts_search == NULL)
357 nd_opt = nd6_option(ndopts);
358 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
360 * Message validation requires that all included
361 * options have a length that is greater than zero.
363 ICMP6STAT_INC(icp6s_nd_badopt);
364 bzero(ndopts, sizeof(*ndopts));
371 switch (nd_opt->nd_opt_type) {
372 case ND_OPT_SOURCE_LINKADDR:
373 case ND_OPT_TARGET_LINKADDR:
375 case ND_OPT_REDIRECTED_HEADER:
377 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
379 "duplicated ND6 option found (type=%d)\n",
380 nd_opt->nd_opt_type));
383 ndopts->nd_opt_array[nd_opt->nd_opt_type]
387 case ND_OPT_PREFIX_INFORMATION:
388 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
389 ndopts->nd_opt_array[nd_opt->nd_opt_type]
392 ndopts->nd_opts_pi_end =
393 (struct nd_opt_prefix_info *)nd_opt;
395 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
396 case ND_OPT_RDNSS: /* RFC 6106 */
397 case ND_OPT_DNSSL: /* RFC 6106 */
399 * Silently ignore options we know and do not care about
405 * Unknown options must be silently ignored,
406 * to accomodate future extension to the protocol.
409 "nd6_options: unsupported option %d - "
410 "option ignored\n", nd_opt->nd_opt_type));
415 if (i > V_nd6_maxndopt) {
416 ICMP6STAT_INC(icp6s_nd_toomanyopt);
417 nd6log((LOG_INFO, "too many loop in nd opt\n"));
421 if (ndopts->nd_opts_done)
429 * ND6 timer routine to handle ND6 entries
432 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
436 LLE_WLOCK_ASSERT(ln);
441 canceled = callout_stop(&ln->ln_timer_ch);
443 ln->la_expire = time_uptime + tick / hz;
445 if (tick > INT_MAX) {
446 ln->ln_ntick = tick - INT_MAX;
447 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
448 nd6_llinfo_timer, ln);
451 canceled = callout_reset(&ln->ln_timer_ch, tick,
452 nd6_llinfo_timer, ln);
460 nd6_llinfo_settimer(struct llentry *ln, long tick)
464 nd6_llinfo_settimer_locked(ln, tick);
469 nd6_llinfo_timer(void *arg)
472 struct in6_addr *dst;
474 struct nd_ifinfo *ndi = NULL;
476 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
477 ln = (struct llentry *)arg;
479 if (callout_pending(&ln->la_timer)) {
481 * Here we are a bit odd here in the treatment of
482 * active/pending. If the pending bit is set, it got
483 * rescheduled before I ran. The active
484 * bit we ignore, since if it was stopped
485 * in ll_tablefree() and was currently running
486 * it would have return 0 so the code would
487 * not have deleted it since the callout could
488 * not be stopped so we want to go through
489 * with the delete here now. If the callout
490 * was restarted, the pending bit will be back on and
491 * we just want to bail since the callout_reset would
492 * return 1 and our reference would have been removed
493 * by nd6_llinfo_settimer_locked above since canceled
499 ifp = ln->lle_tbl->llt_ifp;
500 CURVNET_SET(ifp->if_vnet);
502 if (ln->ln_ntick > 0) {
503 if (ln->ln_ntick > INT_MAX) {
504 ln->ln_ntick -= INT_MAX;
505 nd6_llinfo_settimer_locked(ln, INT_MAX);
508 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
513 ndi = ND_IFINFO(ifp);
514 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
515 if (ln->la_flags & LLE_STATIC) {
519 if (ln->la_flags & LLE_DELETED) {
520 (void)nd6_free(ln, 0);
525 switch (ln->ln_state) {
526 case ND6_LLINFO_INCOMPLETE:
527 if (ln->la_asked < V_nd6_mmaxtries) {
529 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
531 nd6_ns_output(ifp, NULL, dst, ln, NULL);
534 struct mbuf *m = ln->la_hold;
539 * assuming every packet in la_hold has the
540 * same IP header. Send error after unlock.
545 clear_llinfo_pqueue(ln);
547 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT);
548 (void)nd6_free(ln, 0);
551 icmp6_error2(m, ICMP6_DST_UNREACH,
552 ICMP6_DST_UNREACH_ADDR, 0, ifp);
555 case ND6_LLINFO_REACHABLE:
556 if (!ND6_LLINFO_PERMANENT(ln)) {
557 ln->ln_state = ND6_LLINFO_STALE;
558 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
562 case ND6_LLINFO_STALE:
563 /* Garbage Collection(RFC 2461 5.3) */
564 if (!ND6_LLINFO_PERMANENT(ln)) {
565 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
566 (void)nd6_free(ln, 1);
571 case ND6_LLINFO_DELAY:
572 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
575 ln->ln_state = ND6_LLINFO_PROBE;
576 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
578 nd6_ns_output(ifp, dst, dst, ln, NULL);
581 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
582 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
585 case ND6_LLINFO_PROBE:
586 if (ln->la_asked < V_nd6_umaxtries) {
588 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
590 nd6_ns_output(ifp, dst, dst, ln, NULL);
593 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
594 (void)nd6_free(ln, 0);
599 panic("%s: paths in a dark night can be confusing: %d",
600 __func__, ln->ln_state);
610 * ND6 timer routine to expire default route list and prefix list
615 CURVNET_SET((struct vnet *) arg);
616 struct nd_defrouter *dr, *ndr;
617 struct nd_prefix *pr, *npr;
618 struct in6_ifaddr *ia6, *nia6;
620 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
623 /* expire default router list */
624 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
625 if (dr->expire && dr->expire < time_uptime)
630 * expire interface addresses.
631 * in the past the loop was inside prefix expiry processing.
632 * However, from a stricter speci-confrmance standpoint, we should
633 * rather separate address lifetimes and prefix lifetimes.
635 * XXXRW: in6_ifaddrhead locking.
638 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
639 /* check address lifetime */
640 if (IFA6_IS_INVALID(ia6)) {
644 * If the expiring address is temporary, try
645 * regenerating a new one. This would be useful when
646 * we suspended a laptop PC, then turned it on after a
647 * period that could invalidate all temporary
648 * addresses. Although we may have to restart the
649 * loop (see below), it must be after purging the
650 * address. Otherwise, we'd see an infinite loop of
653 if (V_ip6_use_tempaddr &&
654 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
655 if (regen_tmpaddr(ia6) == 0)
659 in6_purgeaddr(&ia6->ia_ifa);
662 goto addrloop; /* XXX: see below */
663 } else if (IFA6_IS_DEPRECATED(ia6)) {
664 int oldflags = ia6->ia6_flags;
666 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
669 * If a temporary address has just become deprecated,
670 * regenerate a new one if possible.
672 if (V_ip6_use_tempaddr &&
673 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
674 (oldflags & IN6_IFF_DEPRECATED) == 0) {
676 if (regen_tmpaddr(ia6) == 0) {
678 * A new temporary address is
680 * XXX: this means the address chain
681 * has changed while we are still in
682 * the loop. Although the change
683 * would not cause disaster (because
684 * it's not a deletion, but an
685 * addition,) we'd rather restart the
686 * loop just for safety. Or does this
687 * significantly reduce performance??
692 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
694 * Schedule DAD for a tentative address. This happens
695 * if the interface was down or not running
696 * when the address was configured.
700 delay = arc4random() %
701 (MAX_RTR_SOLICITATION_DELAY * hz);
702 nd6_dad_start((struct ifaddr *)ia6, delay);
705 * Check status of the interface. If it is down,
706 * mark the address as tentative for future DAD.
708 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 ||
709 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING)
711 (ND_IFINFO(ia6->ia_ifp)->flags &
712 ND6_IFF_IFDISABLED) != 0) {
713 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
714 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
717 * A new RA might have made a deprecated address
720 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
724 /* expire prefix list */
725 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
727 * check prefix lifetime.
728 * since pltime is just for autoconf, pltime processing for
729 * prefix is not necessary.
731 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
732 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) {
735 * address expiration and prefix expiration are
736 * separate. NEVER perform in6_purgeaddr here.
745 * ia6 - deprecated/invalidated temporary address
748 regen_tmpaddr(struct in6_ifaddr *ia6)
752 struct in6_ifaddr *public_ifa6 = NULL;
754 ifp = ia6->ia_ifa.ifa_ifp;
756 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
757 struct in6_ifaddr *it6;
759 if (ifa->ifa_addr->sa_family != AF_INET6)
762 it6 = (struct in6_ifaddr *)ifa;
764 /* ignore no autoconf addresses. */
765 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
768 /* ignore autoconf addresses with different prefixes. */
769 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
773 * Now we are looking at an autoconf address with the same
774 * prefix as ours. If the address is temporary and is still
775 * preferred, do not create another one. It would be rare, but
776 * could happen, for example, when we resume a laptop PC after
779 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
780 !IFA6_IS_DEPRECATED(it6)) {
786 * This is a public autoconf address that has the same prefix
787 * as ours. If it is preferred, keep it. We can't break the
788 * loop here, because there may be a still-preferred temporary
789 * address with the prefix.
791 if (!IFA6_IS_DEPRECATED(it6))
794 if (public_ifa6 != NULL)
795 ifa_ref(&public_ifa6->ia_ifa);
796 IF_ADDR_RUNLOCK(ifp);
798 if (public_ifa6 != NULL) {
801 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
802 ifa_free(&public_ifa6->ia_ifa);
803 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
804 " tmp addr,errno=%d\n", e);
807 ifa_free(&public_ifa6->ia_ifa);
815 * Nuke neighbor cache/prefix/default router management table, right before
819 nd6_purge(struct ifnet *ifp)
821 struct nd_defrouter *dr, *ndr;
822 struct nd_prefix *pr, *npr;
825 * Nuke default router list entries toward ifp.
826 * We defer removal of default router list entries that is installed
827 * in the routing table, in order to keep additional side effects as
830 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
838 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
846 /* Nuke prefix list entries toward ifp */
847 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
848 if (pr->ndpr_ifp == ifp) {
850 * Because if_detach() does *not* release prefixes
851 * while purging addresses the reference count will
852 * still be above zero. We therefore reset it to
853 * make sure that the prefix really gets purged.
858 * Previously, pr->ndpr_addr is removed as well,
859 * but I strongly believe we don't have to do it.
860 * nd6_purge() is only called from in6_ifdetach(),
861 * which removes all the associated interface addresses
863 * (jinmei@kame.net 20010129)
869 /* cancel default outgoing interface setting */
870 if (V_nd6_defifindex == ifp->if_index)
871 nd6_setdefaultiface(0);
873 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
874 /* Refresh default router list. */
879 * We do not nuke the neighbor cache entries here any more
880 * because the neighbor cache is kept in if_afdata[AF_INET6].
881 * nd6_purge() is invoked by in6_ifdetach() which is called
882 * from if_detach() where everything gets purged. So let
883 * in6_domifdetach() do the actual L2 table purging work.
888 * the caller acquires and releases the lock on the lltbls
889 * Returns the llentry locked
892 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
894 struct sockaddr_in6 sin6;
898 bzero(&sin6, sizeof(sin6));
899 sin6.sin6_len = sizeof(struct sockaddr_in6);
900 sin6.sin6_family = AF_INET6;
901 sin6.sin6_addr = *addr6;
903 IF_AFDATA_LOCK_ASSERT(ifp);
906 if (flags & ND6_CREATE)
907 llflags |= LLE_CREATE;
908 if (flags & ND6_EXCLUSIVE)
909 llflags |= LLE_EXCLUSIVE;
911 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
912 if ((ln != NULL) && (llflags & LLE_CREATE))
913 ln->ln_state = ND6_LLINFO_NOSTATE;
919 * Test whether a given IPv6 address is a neighbor or not, ignoring
920 * the actual neighbor cache. The neighbor cache is ignored in order
921 * to not reenter the routing code from within itself.
924 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
926 struct nd_prefix *pr;
927 struct ifaddr *dstaddr;
930 * A link-local address is always a neighbor.
931 * XXX: a link does not necessarily specify a single interface.
933 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
934 struct sockaddr_in6 sin6_copy;
938 * We need sin6_copy since sa6_recoverscope() may modify the
942 if (sa6_recoverscope(&sin6_copy))
943 return (0); /* XXX: should be impossible */
944 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
946 if (sin6_copy.sin6_scope_id == zone)
953 * If the address matches one of our addresses,
954 * it should be a neighbor.
955 * If the address matches one of our on-link prefixes, it should be a
958 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
959 if (pr->ndpr_ifp != ifp)
962 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
965 /* Always use the default FIB here. */
966 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
967 0, 0, RT_DEFAULT_FIB);
971 * This is the case where multiple interfaces
972 * have the same prefix, but only one is installed
973 * into the routing table and that prefix entry
974 * is not the one being examined here. In the case
975 * where RADIX_MPATH is enabled, multiple route
976 * entries (of the same rt_key value) will be
977 * installed because the interface addresses all
980 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
981 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
988 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
989 &addr->sin6_addr, &pr->ndpr_mask))
994 * If the address is assigned on the node of the other side of
995 * a p2p interface, the address should be a neighbor.
997 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
998 if (dstaddr != NULL) {
999 if (dstaddr->ifa_ifp == ifp) {
1007 * If the default router list is empty, all addresses are regarded
1008 * as on-link, and thus, as a neighbor.
1010 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1011 TAILQ_EMPTY(&V_nd_defrouter) &&
1012 V_nd6_defifindex == ifp->if_index) {
1021 * Detect if a given IPv6 address identifies a neighbor on a given link.
1022 * XXX: should take care of the destination of a p2p link?
1025 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
1027 struct llentry *lle;
1030 IF_AFDATA_UNLOCK_ASSERT(ifp);
1031 if (nd6_is_new_addr_neighbor(addr, ifp))
1035 * Even if the address matches none of our addresses, it might be
1036 * in the neighbor cache.
1038 IF_AFDATA_RLOCK(ifp);
1039 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1043 IF_AFDATA_RUNLOCK(ifp);
1048 * Free an nd6 llinfo entry.
1049 * Since the function would cause significant changes in the kernel, DO NOT
1050 * make it global, unless you have a strong reason for the change, and are sure
1051 * that the change is safe.
1053 static struct llentry *
1054 nd6_free(struct llentry *ln, int gc)
1056 struct llentry *next;
1057 struct nd_defrouter *dr;
1060 LLE_WLOCK_ASSERT(ln);
1063 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1064 * even though it is not harmful, it was not really necessary.
1068 nd6_llinfo_settimer_locked(ln, -1);
1070 ifp = ln->lle_tbl->llt_ifp;
1072 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1073 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1075 if (dr != NULL && dr->expire &&
1076 ln->ln_state == ND6_LLINFO_STALE && gc) {
1078 * If the reason for the deletion is just garbage
1079 * collection, and the neighbor is an active default
1080 * router, do not delete it. Instead, reset the GC
1081 * timer using the router's lifetime.
1082 * Simply deleting the entry would affect default
1083 * router selection, which is not necessarily a good
1084 * thing, especially when we're using router preference
1086 * XXX: the check for ln_state would be redundant,
1087 * but we intentionally keep it just in case.
1089 if (dr->expire > time_uptime)
1090 nd6_llinfo_settimer_locked(ln,
1091 (dr->expire - time_uptime) * hz);
1093 nd6_llinfo_settimer_locked(ln,
1094 (long)V_nd6_gctimer * hz);
1096 next = LIST_NEXT(ln, lle_next);
1104 * Unreachablity of a router might affect the default
1105 * router selection and on-link detection of advertised
1110 * Temporarily fake the state to choose a new default
1111 * router and to perform on-link determination of
1112 * prefixes correctly.
1113 * Below the state will be set correctly,
1114 * or the entry itself will be deleted.
1116 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1119 if (ln->ln_router || dr) {
1122 * We need to unlock to avoid a LOR with rt6_flush() with the
1123 * rnh and for the calls to pfxlist_onlink_check() and
1124 * defrouter_select() in the block further down for calls
1125 * into nd6_lookup(). We still hold a ref.
1130 * rt6_flush must be called whether or not the neighbor
1131 * is in the Default Router List.
1132 * See a corresponding comment in nd6_na_input().
1134 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1139 * Since defrouter_select() does not affect the
1140 * on-link determination and MIP6 needs the check
1141 * before the default router selection, we perform
1144 pfxlist_onlink_check();
1147 * Refresh default router list.
1152 if (ln->ln_router || dr)
1157 * Before deleting the entry, remember the next entry as the
1158 * return value. We need this because pfxlist_onlink_check() above
1159 * might have freed other entries (particularly the old next entry) as
1160 * a side effect (XXX).
1162 next = LIST_NEXT(ln, lle_next);
1165 * Save to unlock. We still hold an extra reference and will not
1166 * free(9) in llentry_free() if someone else holds one as well.
1169 IF_AFDATA_LOCK(ifp);
1172 /* Guard against race with other llentry_free(). */
1173 if (ln->la_flags & LLE_LINKED) {
1177 LLE_FREE_LOCKED(ln);
1179 IF_AFDATA_UNLOCK(ifp);
1185 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1187 * XXX cost-effective methods?
1190 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1195 if ((dst6 == NULL) || (rt == NULL))
1199 IF_AFDATA_RLOCK(ifp);
1200 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1201 IF_AFDATA_RUNLOCK(ifp);
1205 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1209 * if we get upper-layer reachability confirmation many times,
1210 * it is possible we have false information.
1214 if (ln->ln_byhint > V_nd6_maxnudhint) {
1219 ln->ln_state = ND6_LLINFO_REACHABLE;
1220 if (!ND6_LLINFO_PERMANENT(ln)) {
1221 nd6_llinfo_settimer_locked(ln,
1222 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1230 * Rejuvenate this function for routing operations related
1234 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1236 struct sockaddr_in6 *gateway;
1237 struct nd_defrouter *dr;
1241 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1252 * Only indirect routes are interesting.
1254 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1257 * check for default route
1259 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1260 &SIN6(rt_key(rt))->sin6_addr)) {
1262 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1272 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1274 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1275 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1276 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1279 if (ifp->if_afdata[AF_INET6] == NULL)
1280 return (EPFNOSUPPORT);
1282 case OSIOCGIFINFO_IN6:
1284 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1285 bzero(&ND, sizeof(ND));
1286 ND.linkmtu = IN6_LINKMTU(ifp);
1287 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1288 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1289 ND.reachable = ND_IFINFO(ifp)->reachable;
1290 ND.retrans = ND_IFINFO(ifp)->retrans;
1291 ND.flags = ND_IFINFO(ifp)->flags;
1292 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1293 ND.chlim = ND_IFINFO(ifp)->chlim;
1295 case SIOCGIFINFO_IN6:
1296 ND = *ND_IFINFO(ifp);
1298 case SIOCSIFINFO_IN6:
1300 * used to change host variables from userland.
1301 * intented for a use on router to reflect RA configurations.
1303 /* 0 means 'unspecified' */
1304 if (ND.linkmtu != 0) {
1305 if (ND.linkmtu < IPV6_MMTU ||
1306 ND.linkmtu > IN6_LINKMTU(ifp)) {
1310 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1313 if (ND.basereachable != 0) {
1314 int obasereachable = ND_IFINFO(ifp)->basereachable;
1316 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1317 if (ND.basereachable != obasereachable)
1318 ND_IFINFO(ifp)->reachable =
1319 ND_COMPUTE_RTIME(ND.basereachable);
1321 if (ND.retrans != 0)
1322 ND_IFINFO(ifp)->retrans = ND.retrans;
1324 ND_IFINFO(ifp)->chlim = ND.chlim;
1326 case SIOCSIFINFO_FLAGS:
1329 struct in6_ifaddr *ia;
1331 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1332 !(ND.flags & ND6_IFF_IFDISABLED)) {
1333 /* ifdisabled 1->0 transision */
1336 * If the interface is marked as ND6_IFF_IFDISABLED and
1337 * has an link-local address with IN6_IFF_DUPLICATED,
1338 * do not clear ND6_IFF_IFDISABLED.
1339 * See RFC 4862, Section 5.4.5.
1342 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1343 if (ifa->ifa_addr->sa_family != AF_INET6)
1345 ia = (struct in6_ifaddr *)ifa;
1346 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1347 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1350 IF_ADDR_RUNLOCK(ifp);
1353 /* LLA is duplicated. */
1354 ND.flags |= ND6_IFF_IFDISABLED;
1355 log(LOG_ERR, "Cannot enable an interface"
1356 " with a link-local address marked"
1359 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1360 if (ifp->if_flags & IFF_UP)
1363 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1364 (ND.flags & ND6_IFF_IFDISABLED)) {
1365 /* ifdisabled 0->1 transision */
1366 /* Mark all IPv6 address as tentative. */
1368 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1369 if (V_ip6_dad_count > 0 &&
1370 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1372 TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1374 if (ifa->ifa_addr->sa_family !=
1377 ia = (struct in6_ifaddr *)ifa;
1378 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1380 IF_ADDR_RUNLOCK(ifp);
1384 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1385 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1386 /* auto_linklocal 0->1 transision */
1388 /* If no link-local address on ifp, configure */
1389 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1390 in6_ifattach(ifp, NULL);
1391 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1392 ifp->if_flags & IFF_UP) {
1394 * When the IF already has
1395 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1396 * address is assigned, and IFF_UP, try to
1400 TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1402 if (ifa->ifa_addr->sa_family !=
1405 ia = (struct in6_ifaddr *)ifa;
1406 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1409 IF_ADDR_RUNLOCK(ifp);
1411 /* No LLA is configured. */
1412 in6_ifattach(ifp, NULL);
1416 ND_IFINFO(ifp)->flags = ND.flags;
1419 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1420 /* sync kernel routing table with the default router list */
1424 case SIOCSPFXFLUSH_IN6:
1426 /* flush all the prefix advertised by routers */
1427 struct nd_prefix *pr, *next;
1429 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1430 struct in6_ifaddr *ia, *ia_next;
1432 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1435 /* do we really have to remove addresses as well? */
1436 /* XXXRW: in6_ifaddrhead locking. */
1437 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1439 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1442 if (ia->ia6_ndpr == pr)
1443 in6_purgeaddr(&ia->ia_ifa);
1449 case SIOCSRTRFLUSH_IN6:
1451 /* flush all the default routers */
1452 struct nd_defrouter *dr, *next;
1455 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1461 case SIOCGNBRINFO_IN6:
1464 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1466 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1469 IF_AFDATA_RLOCK(ifp);
1470 ln = nd6_lookup(&nb_addr, 0, ifp);
1471 IF_AFDATA_RUNLOCK(ifp);
1477 nbi->state = ln->ln_state;
1478 nbi->asked = ln->la_asked;
1479 nbi->isrouter = ln->ln_router;
1480 if (ln->la_expire == 0)
1483 nbi->expire = ln->la_expire +
1484 (time_second - time_uptime);
1488 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1489 ndif->ifindex = V_nd6_defifindex;
1491 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1492 return (nd6_setdefaultiface(ndif->ifindex));
1498 * Create neighbor cache entry and cache link-layer address,
1499 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1502 * code - type dependent information
1505 * The caller of this function already acquired the ndp
1506 * cache table lock because the cache entry is returned.
1509 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1510 int lladdrlen, int type, int code)
1512 struct llentry *ln = NULL;
1519 uint16_t router = 0;
1520 struct sockaddr_in6 sin6;
1521 struct mbuf *chain = NULL;
1522 int static_route = 0;
1524 IF_AFDATA_UNLOCK_ASSERT(ifp);
1526 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1527 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1529 /* nothing must be updated for unspecified address */
1530 if (IN6_IS_ADDR_UNSPECIFIED(from))
1534 * Validation about ifp->if_addrlen and lladdrlen must be done in
1537 * XXX If the link does not have link-layer adderss, what should
1538 * we do? (ifp->if_addrlen == 0)
1539 * Spec says nothing in sections for RA, RS and NA. There's small
1540 * description on it in NS section (RFC 2461 7.2.3).
1542 flags = lladdr ? ND6_EXCLUSIVE : 0;
1543 IF_AFDATA_RLOCK(ifp);
1544 ln = nd6_lookup(from, flags, ifp);
1545 IF_AFDATA_RUNLOCK(ifp);
1547 flags |= ND6_EXCLUSIVE;
1548 IF_AFDATA_LOCK(ifp);
1549 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1550 IF_AFDATA_UNLOCK(ifp);
1553 /* do nothing if static ndp is set */
1554 if (ln->la_flags & LLE_STATIC) {
1563 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1564 if (olladdr && lladdr) {
1565 llchange = bcmp(lladdr, &ln->ll_addr,
1571 * newentry olladdr lladdr llchange (*=record)
1574 * 0 n y -- (3) * STALE
1576 * 0 y y y (5) * STALE
1577 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1578 * 1 -- y -- (7) * STALE
1581 if (lladdr) { /* (3-5) and (7) */
1583 * Record source link-layer address
1584 * XXX is it dependent to ifp->if_type?
1586 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1587 ln->la_flags |= LLE_VALID;
1588 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1592 if ((!olladdr && lladdr != NULL) || /* (3) */
1593 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1595 newstate = ND6_LLINFO_STALE;
1596 } else /* (1-2,4) */
1600 if (lladdr == NULL) /* (6) */
1601 newstate = ND6_LLINFO_NOSTATE;
1603 newstate = ND6_LLINFO_STALE;
1608 * Update the state of the neighbor cache.
1610 ln->ln_state = newstate;
1612 if (ln->ln_state == ND6_LLINFO_STALE) {
1613 if (ln->la_hold != NULL)
1614 nd6_grab_holdchain(ln, &chain, &sin6);
1615 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1616 /* probe right away */
1617 nd6_llinfo_settimer_locked((void *)ln, 0);
1622 * ICMP6 type dependent behavior.
1624 * NS: clear IsRouter if new entry
1625 * RS: clear IsRouter
1626 * RA: set IsRouter if there's lladdr
1627 * redir: clear IsRouter if new entry
1630 * The spec says that we must set IsRouter in the following cases:
1631 * - If lladdr exist, set IsRouter. This means (1-5).
1632 * - If it is old entry (!newentry), set IsRouter. This means (7).
1633 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1634 * A quetion arises for (1) case. (1) case has no lladdr in the
1635 * neighbor cache, this is similar to (6).
1636 * This case is rare but we figured that we MUST NOT set IsRouter.
1638 * newentry olladdr lladdr llchange NS RS RA redir
1640 * 0 n n -- (1) c ? s
1641 * 0 y n -- (2) c s s
1642 * 0 n y -- (3) c s s
1645 * 1 -- n -- (6) c c c s
1646 * 1 -- y -- (7) c c s c s
1650 switch (type & 0xff) {
1651 case ND_NEIGHBOR_SOLICIT:
1653 * New entry must have is_router flag cleared.
1655 if (is_newentry) /* (6-7) */
1660 * If the icmp is a redirect to a better router, always set the
1661 * is_router flag. Otherwise, if the entry is newly created,
1662 * clear the flag. [RFC 2461, sec 8.3]
1664 if (code == ND_REDIRECT_ROUTER)
1666 else if (is_newentry) /* (6-7) */
1669 case ND_ROUTER_SOLICIT:
1671 * is_router flag must always be cleared.
1675 case ND_ROUTER_ADVERT:
1677 * Mark an entry with lladdr as a router.
1679 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1680 (is_newentry && lladdr)) { /* (7) */
1687 static_route = (ln->la_flags & LLE_STATIC);
1688 router = ln->ln_router;
1690 if (flags & ND6_EXCLUSIVE)
1698 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
1701 * When the link-layer address of a router changes, select the
1702 * best router again. In particular, when the neighbor entry is newly
1703 * created, it might affect the selection policy.
1704 * Question: can we restrict the first condition to the "is_newentry"
1706 * XXX: when we hear an RA from a new router with the link-layer
1707 * address option, defrouter_select() is called twice, since
1708 * defrtrlist_update called the function as well. However, I believe
1709 * we can compromise the overhead, since it only happens the first
1711 * XXX: although defrouter_select() should not have a bad effect
1712 * for those are not autoconfigured hosts, we explicitly avoid such
1715 if (do_update && router &&
1716 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1718 * guaranteed recursion
1726 if (flags & ND6_EXCLUSIVE)
1737 nd6_slowtimo(void *arg)
1739 CURVNET_SET((struct vnet *) arg);
1740 struct nd_ifinfo *nd6if;
1743 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1744 nd6_slowtimo, curvnet);
1745 IFNET_RLOCK_NOSLEEP();
1746 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1747 if (ifp->if_afdata[AF_INET6] == NULL)
1749 nd6if = ND_IFINFO(ifp);
1750 if (nd6if->basereachable && /* already initialized */
1751 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1753 * Since reachable time rarely changes by router
1754 * advertisements, we SHOULD insure that a new random
1755 * value gets recomputed at least once every few hours.
1758 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1759 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1762 IFNET_RUNLOCK_NOSLEEP();
1767 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
1768 struct sockaddr_in6 *sin6)
1771 LLE_WLOCK_ASSERT(ln);
1773 *chain = ln->la_hold;
1775 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6));
1777 if (ln->ln_state == ND6_LLINFO_STALE) {
1780 * The first time we send a packet to a
1781 * neighbor whose entry is STALE, we have
1782 * to change the state to DELAY and a sets
1783 * a timer to expire in DELAY_FIRST_PROBE_TIME
1784 * seconds to ensure do neighbor unreachability
1785 * detection on expiration.
1789 ln->ln_state = ND6_LLINFO_DELAY;
1790 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1795 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1796 struct sockaddr_in6 *dst)
1800 struct ip6_hdr *ip6;
1804 mac_netinet6_nd6_send(ifp, m);
1808 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
1809 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
1810 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
1811 * to be diverted to user space. When re-injected into the kernel,
1812 * send_output() will directly dispatch them to the outgoing interface.
1814 if (send_sendso_input_hook != NULL) {
1815 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
1817 ip6 = mtod(m, struct ip6_hdr *);
1818 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
1819 /* Use the SEND socket */
1820 error = send_sendso_input_hook(m, ifp, SND_OUT,
1822 /* -1 == no app on SEND socket */
1823 if (error == 0 || error != -1)
1828 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1829 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
1830 mtod(m, struct ip6_hdr *));
1832 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1835 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL);
1840 * IPv6 packet output - light version.
1841 * Checks if destination LLE exists and is in proper state
1842 * (e.g no modification required). If not true, fall back to
1846 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1847 struct sockaddr_in6 *dst, struct rtentry *rt0)
1849 struct llentry *ln = NULL;
1851 /* discard the packet if IPv6 operation is disabled on the interface */
1852 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1854 return (ENETDOWN); /* better error? */
1857 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1860 if (nd6_need_cache(ifp) == 0)
1863 IF_AFDATA_RLOCK(ifp);
1864 ln = nd6_lookup(&dst->sin6_addr, 0, ifp);
1865 IF_AFDATA_RUNLOCK(ifp);
1868 * Perform fast path for the following cases:
1869 * 1) lle state is REACHABLE
1870 * 2) lle state is DELAY (NS message sentNS message sent)
1872 * Every other case involves lle modification, so we handle
1875 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE &&
1876 ln->ln_state != ND6_LLINFO_DELAY)) {
1877 /* Fall back to slow processing path */
1880 return (nd6_output_lle(ifp, origifp, m, dst));
1887 return (nd6_output_ifp(ifp, origifp, m, dst));
1892 * Output IPv6 packet - heavy version.
1893 * Function assume that either
1894 * 1) destination LLE does not exist, is invalid or stale, so
1895 * ND6_EXCLUSIVE lock needs to be acquired
1896 * 2) destination lle is provided (with ND6_EXCLUSIVE lock),
1897 * in that case packets are queued in &chain.
1901 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1902 struct sockaddr_in6 *dst)
1904 struct llentry *lle = NULL;
1907 KASSERT(m != NULL, ("NULL mbuf, nothing to send"));
1908 /* discard the packet if IPv6 operation is disabled on the interface */
1909 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1911 return (ENETDOWN); /* better error? */
1914 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1917 if (nd6_need_cache(ifp) == 0)
1921 * Address resolution or Neighbor Unreachability Detection
1923 * At this point, the destination of the packet must be a unicast
1924 * or an anycast address(i.e. not a multicast).
1927 IF_AFDATA_RLOCK(ifp);
1928 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
1929 IF_AFDATA_RUNLOCK(ifp);
1930 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1932 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1933 * the condition below is not very efficient. But we believe
1934 * it is tolerable, because this should be a rare case.
1936 flags = ND6_CREATE | ND6_EXCLUSIVE;
1937 IF_AFDATA_LOCK(ifp);
1938 lle = nd6_lookup(&dst->sin6_addr, flags, ifp);
1939 IF_AFDATA_UNLOCK(ifp);
1943 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1944 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1945 char ip6buf[INET6_ADDRSTRLEN];
1947 "nd6_output: can't allocate llinfo for %s "
1949 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
1953 goto sendpkt; /* send anyway */
1956 LLE_WLOCK_ASSERT(lle);
1958 /* We don't have to do link-layer address resolution on a p2p link. */
1959 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1960 lle->ln_state < ND6_LLINFO_REACHABLE) {
1961 lle->ln_state = ND6_LLINFO_STALE;
1962 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz);
1966 * The first time we send a packet to a neighbor whose entry is
1967 * STALE, we have to change the state to DELAY and a sets a timer to
1968 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1969 * neighbor unreachability detection on expiration.
1972 if (lle->ln_state == ND6_LLINFO_STALE) {
1974 lle->ln_state = ND6_LLINFO_DELAY;
1975 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz);
1979 * If the neighbor cache entry has a state other than INCOMPLETE
1980 * (i.e. its link-layer address is already resolved), just
1983 if (lle->ln_state > ND6_LLINFO_INCOMPLETE)
1987 * There is a neighbor cache entry, but no ethernet address
1988 * response yet. Append this latest packet to the end of the
1989 * packet queue in the mbuf, unless the number of the packet
1990 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
1991 * the oldest packet in the queue will be removed.
1993 if (lle->ln_state == ND6_LLINFO_NOSTATE)
1994 lle->ln_state = ND6_LLINFO_INCOMPLETE;
1996 if (lle->la_hold != NULL) {
1997 struct mbuf *m_hold;
2001 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2003 if (m_hold->m_nextpkt == NULL) {
2004 m_hold->m_nextpkt = m;
2008 while (i >= V_nd6_maxqueuelen) {
2009 m_hold = lle->la_hold;
2010 lle->la_hold = lle->la_hold->m_nextpkt;
2019 * If there has been no NS for the neighbor after entering the
2020 * INCOMPLETE state, send the first solicitation.
2022 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) {
2025 nd6_llinfo_settimer_locked(lle,
2026 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2028 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL);
2030 /* We did the lookup so we need to do the unlock here. */
2040 return (nd6_output_ifp(ifp, origifp, m, dst));
2045 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2046 struct sockaddr_in6 *dst)
2048 struct mbuf *m, *m_head;
2049 struct ifnet *outifp;
2053 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2060 m_head = m_head->m_nextpkt;
2061 error = nd6_output_ifp(ifp, origifp, m, dst);
2066 * note that intermediate errors are blindly ignored - but this is
2067 * the same convention as used with nd6_output when called by
2075 nd6_need_cache(struct ifnet *ifp)
2078 * XXX: we currently do not make neighbor cache on any interface
2079 * other than ARCnet, Ethernet, FDDI and GIF.
2082 * - unidirectional tunnels needs no ND
2084 switch (ifp->if_type) {
2092 #ifdef IFT_IEEE80211
2095 case IFT_INFINIBAND:
2097 case IFT_PROPVIRTUAL:
2105 * the callers of this function need to be re-worked to drop
2106 * the lle lock, drop here for now
2109 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2110 const struct sockaddr *dst, u_char *desten, struct llentry **lle)
2115 IF_AFDATA_UNLOCK_ASSERT(ifp);
2116 if (m != NULL && m->m_flags & M_MCAST) {
2119 switch (ifp->if_type) {
2125 #ifdef IFT_IEEE80211
2130 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2135 * netbsd can use if_broadcastaddr, but we don't do so
2136 * to reduce # of ifdef.
2138 for (i = 0; i < ifp->if_addrlen; i++)
2146 return (EAFNOSUPPORT);
2152 * the entry should have been created in nd6_store_lladdr
2154 IF_AFDATA_RLOCK(ifp);
2155 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2156 IF_AFDATA_RUNLOCK(ifp);
2157 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2160 /* this could happen, if we could not allocate memory */
2165 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2169 * A *small* use after free race exists here
2175 clear_llinfo_pqueue(struct llentry *ln)
2177 struct mbuf *m_hold, *m_hold_next;
2179 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2180 m_hold_next = m_hold->m_nextpkt;
2188 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2189 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2191 SYSCTL_DECL(_net_inet6_icmp6);
2193 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2194 CTLFLAG_RD, nd6_sysctl_drlist, "");
2195 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2196 CTLFLAG_RD, nd6_sysctl_prlist, "");
2197 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2198 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2199 SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2200 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2203 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2205 struct in6_defrouter d;
2206 struct nd_defrouter *dr;
2212 bzero(&d, sizeof(d));
2213 d.rtaddr.sin6_family = AF_INET6;
2214 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2219 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2220 d.rtaddr.sin6_addr = dr->rtaddr;
2221 error = sa6_recoverscope(&d.rtaddr);
2224 d.flags = dr->flags;
2225 d.rtlifetime = dr->rtlifetime;
2226 d.expire = dr->expire + (time_second - time_uptime);
2227 d.if_index = dr->ifp->if_index;
2228 error = SYSCTL_OUT(req, &d, sizeof(d));
2236 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2238 struct in6_prefix p;
2239 struct sockaddr_in6 s6;
2240 struct nd_prefix *pr;
2241 struct nd_pfxrouter *pfr;
2244 char ip6buf[INET6_ADDRSTRLEN];
2249 bzero(&p, sizeof(p));
2250 p.origin = PR_ORIG_RA;
2251 bzero(&s6, sizeof(s6));
2252 s6.sin6_family = AF_INET6;
2253 s6.sin6_len = sizeof(s6);
2258 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2259 p.prefix = pr->ndpr_prefix;
2260 if (sa6_recoverscope(&p.prefix)) {
2261 log(LOG_ERR, "scope error in prefix list (%s)\n",
2262 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2263 /* XXX: press on... */
2265 p.raflags = pr->ndpr_raf;
2266 p.prefixlen = pr->ndpr_plen;
2267 p.vltime = pr->ndpr_vltime;
2268 p.pltime = pr->ndpr_pltime;
2269 p.if_index = pr->ndpr_ifp->if_index;
2270 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2273 /* XXX: we assume time_t is signed. */
2275 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2276 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2277 p.expire = pr->ndpr_lastupdate +
2279 (time_second - time_uptime);
2281 p.expire = maxexpire;
2283 p.refcnt = pr->ndpr_refcnt;
2284 p.flags = pr->ndpr_stateflags;
2286 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2288 error = SYSCTL_OUT(req, &p, sizeof(p));
2291 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2292 s6.sin6_addr = pfr->router->rtaddr;
2293 if (sa6_recoverscope(&s6))
2295 "scope error in prefix list (%s)\n",
2296 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2297 error = SYSCTL_OUT(req, &s6, sizeof(s6));