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;
115 VNET_DEFINE(struct nd_drhead, nd_defrouter);
116 VNET_DEFINE(struct nd_prhead, nd_prefix);
117 VNET_DEFINE(struct rwlock, nd6_lock);
119 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
120 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval)
122 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
124 static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *,
126 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
127 static void nd6_slowtimo(void *);
128 static int regen_tmpaddr(struct in6_ifaddr *);
129 static struct llentry *nd6_free(struct llentry **, int);
130 static void nd6_llinfo_timer(void *);
131 static void clear_llinfo_pqueue(struct llentry *);
132 static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *,
133 struct sockaddr_in6 *);
134 static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *,
135 struct sockaddr_in6 *);
137 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
138 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch)
140 VNET_DEFINE(struct callout, nd6_timer_ch);
146 rw_init(&V_nd6_lock, "nd6");
148 LIST_INIT(&V_nd_prefix);
150 /* initialization of the default router list */
151 TAILQ_INIT(&V_nd_defrouter);
154 callout_init(&V_nd6_slowtimo_ch, 0);
155 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
156 nd6_slowtimo, curvnet);
166 callout_drain(&V_nd6_slowtimo_ch);
167 callout_drain(&V_nd6_timer_ch);
168 rw_destroy(&V_nd6_lock);
173 nd6_ifattach(struct ifnet *ifp)
175 struct nd_ifinfo *nd;
177 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
180 nd->chlim = IPV6_DEFHLIM;
181 nd->basereachable = REACHABLE_TIME;
182 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
183 nd->retrans = RETRANS_TIMER;
185 nd->flags = ND6_IFF_PERFORMNUD;
187 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
188 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
189 * default regardless of the V_ip6_auto_linklocal configuration to
190 * give a reasonable default behavior.
192 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
193 (ifp->if_flags & IFF_LOOPBACK))
194 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
196 * A loopback interface does not need to accept RTADV.
197 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
198 * default regardless of the V_ip6_accept_rtadv configuration to
199 * prevent the interface from accepting RA messages arrived
200 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
202 if (V_ip6_accept_rtadv &&
203 !(ifp->if_flags & IFF_LOOPBACK) &&
204 (ifp->if_type != IFT_BRIDGE))
205 nd->flags |= ND6_IFF_ACCEPT_RTADV;
206 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
207 nd->flags |= ND6_IFF_NO_RADR;
209 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
210 nd6_setmtu0(ifp, nd);
216 nd6_ifdetach(struct nd_ifinfo *nd)
223 * Reset ND level link MTU. This function is called when the physical MTU
224 * changes, which means we might have to adjust the ND level MTU.
227 nd6_setmtu(struct ifnet *ifp)
229 if (ifp->if_afdata[AF_INET6] == NULL)
232 nd6_setmtu0(ifp, ND_IFINFO(ifp));
235 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
237 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
241 omaxmtu = ndi->maxmtu;
243 switch (ifp->if_type) {
245 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
248 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
251 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
254 ndi->maxmtu = ifp->if_mtu;
259 * Decreasing the interface MTU under IPV6 minimum MTU may cause
260 * undesirable situation. We thus notify the operator of the change
261 * explicitly. The check for omaxmtu is necessary to restrict the
262 * log to the case of changing the MTU, not initializing it.
264 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
265 log(LOG_NOTICE, "nd6_setmtu0: "
266 "new link MTU on %s (%lu) is too small for IPv6\n",
267 if_name(ifp), (unsigned long)ndi->maxmtu);
270 if (ndi->maxmtu > V_in6_maxmtu)
271 in6_setmaxmtu(); /* check all interfaces just in case */
276 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
279 bzero(ndopts, sizeof(*ndopts));
280 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
282 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
285 ndopts->nd_opts_done = 1;
286 ndopts->nd_opts_search = NULL;
291 * Take one ND option.
294 nd6_option(union nd_opts *ndopts)
296 struct nd_opt_hdr *nd_opt;
299 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
300 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
302 if (ndopts->nd_opts_search == NULL)
304 if (ndopts->nd_opts_done)
307 nd_opt = ndopts->nd_opts_search;
309 /* make sure nd_opt_len is inside the buffer */
310 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
311 bzero(ndopts, sizeof(*ndopts));
315 olen = nd_opt->nd_opt_len << 3;
318 * Message validation requires that all included
319 * options have a length that is greater than zero.
321 bzero(ndopts, sizeof(*ndopts));
325 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
326 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
327 /* option overruns the end of buffer, invalid */
328 bzero(ndopts, sizeof(*ndopts));
330 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
331 /* reached the end of options chain */
332 ndopts->nd_opts_done = 1;
333 ndopts->nd_opts_search = NULL;
339 * Parse multiple ND options.
340 * This function is much easier to use, for ND routines that do not need
341 * multiple options of the same type.
344 nd6_options(union nd_opts *ndopts)
346 struct nd_opt_hdr *nd_opt;
349 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
350 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
352 if (ndopts->nd_opts_search == NULL)
356 nd_opt = nd6_option(ndopts);
357 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
359 * Message validation requires that all included
360 * options have a length that is greater than zero.
362 ICMP6STAT_INC(icp6s_nd_badopt);
363 bzero(ndopts, sizeof(*ndopts));
370 switch (nd_opt->nd_opt_type) {
371 case ND_OPT_SOURCE_LINKADDR:
372 case ND_OPT_TARGET_LINKADDR:
374 case ND_OPT_REDIRECTED_HEADER:
376 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
378 "duplicated ND6 option found (type=%d)\n",
379 nd_opt->nd_opt_type));
382 ndopts->nd_opt_array[nd_opt->nd_opt_type]
386 case ND_OPT_PREFIX_INFORMATION:
387 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
388 ndopts->nd_opt_array[nd_opt->nd_opt_type]
391 ndopts->nd_opts_pi_end =
392 (struct nd_opt_prefix_info *)nd_opt;
394 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
395 case ND_OPT_RDNSS: /* RFC 6106 */
396 case ND_OPT_DNSSL: /* RFC 6106 */
398 * Silently ignore options we know and do not care about
404 * Unknown options must be silently ignored,
405 * to accomodate future extension to the protocol.
408 "nd6_options: unsupported option %d - "
409 "option ignored\n", nd_opt->nd_opt_type));
414 if (i > V_nd6_maxndopt) {
415 ICMP6STAT_INC(icp6s_nd_toomanyopt);
416 nd6log((LOG_INFO, "too many loop in nd opt\n"));
420 if (ndopts->nd_opts_done)
428 * ND6 timer routine to handle ND6 entries
431 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
435 LLE_WLOCK_ASSERT(ln);
440 canceled = callout_stop(&ln->ln_timer_ch);
442 ln->la_expire = time_uptime + tick / hz;
444 if (tick > INT_MAX) {
445 ln->ln_ntick = tick - INT_MAX;
446 canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
447 nd6_llinfo_timer, ln);
450 canceled = callout_reset(&ln->ln_timer_ch, tick,
451 nd6_llinfo_timer, ln);
459 nd6_llinfo_settimer(struct llentry *ln, long tick)
463 nd6_llinfo_settimer_locked(ln, tick);
468 nd6_llinfo_timer(void *arg)
471 struct in6_addr *dst;
473 struct nd_ifinfo *ndi;
475 KASSERT(arg != NULL, ("%s: arg NULL", __func__));
476 ln = (struct llentry *)arg;
477 ifp = ln->lle_tbl->llt_ifp;
478 CURVNET_SET(ifp->if_vnet);
482 if (callout_pending(&ln->la_timer)) {
484 * Here we are a bit odd here in the treatment of
485 * active/pending. If the pending bit is set, it got
486 * rescheduled before I ran. The active
487 * bit we ignore, since if it was stopped
488 * in ll_tablefree() and was currently running
489 * it would have return 0 so the code would
490 * not have deleted it since the callout could
491 * not be stopped so we want to go through
492 * with the delete here now. If the callout
493 * was restarted, the pending bit will be back on and
494 * we just want to bail since the callout_reset would
495 * return 1 and our reference would have been removed
496 * by nd6_llinfo_settimer_locked above since canceled
505 if (ln->ln_ntick > 0) {
506 if (ln->ln_ntick > INT_MAX) {
507 ln->ln_ntick -= INT_MAX;
508 nd6_llinfo_settimer_locked(ln, INT_MAX);
511 nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
516 ndi = ND_IFINFO(ifp);
517 dst = &L3_ADDR_SIN6(ln)->sin6_addr;
518 if (ln->la_flags & LLE_STATIC) {
522 if (ln->la_flags & LLE_DELETED) {
523 (void)nd6_free(&ln, 0);
527 switch (ln->ln_state) {
528 case ND6_LLINFO_INCOMPLETE:
529 if (ln->la_asked < V_nd6_mmaxtries) {
531 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
533 nd6_ns_output(ifp, NULL, dst, ln, NULL);
536 struct mbuf *m = ln->la_hold;
541 * assuming every packet in la_hold has the
542 * same IP header. Send error after unlock.
547 clear_llinfo_pqueue(ln);
549 (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 (void)nd6_free(&ln, 0);
597 panic("%s: paths in a dark night can be confusing: %d",
598 __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_drhead drq;
617 struct nd_defrouter *dr, *ndr;
618 struct nd_prefix *pr, *npr;
619 struct in6_ifaddr *ia6, *nia6;
621 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
626 /* expire default router list */
628 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr)
629 if (dr->expire && dr->expire < time_uptime)
630 defrouter_unlink(dr, &drq);
633 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
634 TAILQ_REMOVE(&drq, dr, dr_entry);
639 * expire interface addresses.
640 * in the past the loop was inside prefix expiry processing.
641 * However, from a stricter speci-confrmance standpoint, we should
642 * rather separate address lifetimes and prefix lifetimes.
644 * XXXRW: in6_ifaddrhead locking.
647 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
648 /* check address lifetime */
649 if (IFA6_IS_INVALID(ia6)) {
653 * If the expiring address is temporary, try
654 * regenerating a new one. This would be useful when
655 * we suspended a laptop PC, then turned it on after a
656 * period that could invalidate all temporary
657 * addresses. Although we may have to restart the
658 * loop (see below), it must be after purging the
659 * address. Otherwise, we'd see an infinite loop of
662 if (V_ip6_use_tempaddr &&
663 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
664 if (regen_tmpaddr(ia6) == 0)
668 in6_purgeaddr(&ia6->ia_ifa);
671 goto addrloop; /* XXX: see below */
672 } else if (IFA6_IS_DEPRECATED(ia6)) {
673 int oldflags = ia6->ia6_flags;
675 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
678 * If a temporary address has just become deprecated,
679 * regenerate a new one if possible.
681 if (V_ip6_use_tempaddr &&
682 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
683 (oldflags & IN6_IFF_DEPRECATED) == 0) {
685 if (regen_tmpaddr(ia6) == 0) {
687 * A new temporary address is
689 * XXX: this means the address chain
690 * has changed while we are still in
691 * the loop. Although the change
692 * would not cause disaster (because
693 * it's not a deletion, but an
694 * addition,) we'd rather restart the
695 * loop just for safety. Or does this
696 * significantly reduce performance??
701 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
703 * Schedule DAD for a tentative address. This happens
704 * if the interface was down or not running
705 * when the address was configured.
709 delay = arc4random() %
710 (MAX_RTR_SOLICITATION_DELAY * hz);
711 nd6_dad_start((struct ifaddr *)ia6, delay);
714 * Check status of the interface. If it is down,
715 * mark the address as tentative for future DAD.
717 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 ||
718 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING)
720 (ND_IFINFO(ia6->ia_ifp)->flags &
721 ND6_IFF_IFDISABLED) != 0) {
722 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
723 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
726 * A new RA might have made a deprecated address
729 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
733 /* expire prefix list */
734 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
736 * check prefix lifetime.
737 * since pltime is just for autoconf, pltime processing for
738 * prefix is not necessary.
740 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
741 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) {
744 * address expiration and prefix expiration are
745 * separate. NEVER perform in6_purgeaddr here.
754 * ia6 - deprecated/invalidated temporary address
757 regen_tmpaddr(struct in6_ifaddr *ia6)
761 struct in6_ifaddr *public_ifa6 = NULL;
763 ifp = ia6->ia_ifa.ifa_ifp;
765 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
766 struct in6_ifaddr *it6;
768 if (ifa->ifa_addr->sa_family != AF_INET6)
771 it6 = (struct in6_ifaddr *)ifa;
773 /* ignore no autoconf addresses. */
774 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
777 /* ignore autoconf addresses with different prefixes. */
778 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
782 * Now we are looking at an autoconf address with the same
783 * prefix as ours. If the address is temporary and is still
784 * preferred, do not create another one. It would be rare, but
785 * could happen, for example, when we resume a laptop PC after
788 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
789 !IFA6_IS_DEPRECATED(it6)) {
795 * This is a public autoconf address that has the same prefix
796 * as ours. If it is preferred, keep it. We can't break the
797 * loop here, because there may be a still-preferred temporary
798 * address with the prefix.
800 if (!IFA6_IS_DEPRECATED(it6))
803 if (public_ifa6 != NULL)
804 ifa_ref(&public_ifa6->ia_ifa);
805 IF_ADDR_RUNLOCK(ifp);
807 if (public_ifa6 != NULL) {
810 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
811 ifa_free(&public_ifa6->ia_ifa);
812 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
813 " tmp addr,errno=%d\n", e);
816 ifa_free(&public_ifa6->ia_ifa);
824 * Nuke neighbor cache/prefix/default router management table, right before
828 nd6_purge(struct ifnet *ifp)
830 struct nd_drhead drq;
831 struct nd_defrouter *dr, *ndr;
832 struct nd_prefix *pr, *npr;
837 * Nuke default router list entries toward ifp.
838 * We defer removal of default router list entries that is installed
839 * in the routing table, in order to keep additional side effects as
843 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
847 defrouter_unlink(dr, &drq);
850 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
854 defrouter_unlink(dr, &drq);
858 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
859 TAILQ_REMOVE(&drq, dr, dr_entry);
863 /* Nuke prefix list entries toward ifp */
864 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
865 if (pr->ndpr_ifp == ifp) {
867 * Because if_detach() does *not* release prefixes
868 * while purging addresses the reference count will
869 * still be above zero. We therefore reset it to
870 * make sure that the prefix really gets purged.
875 * Previously, pr->ndpr_addr is removed as well,
876 * but I strongly believe we don't have to do it.
877 * nd6_purge() is only called from in6_ifdetach(),
878 * which removes all the associated interface addresses
880 * (jinmei@kame.net 20010129)
886 /* cancel default outgoing interface setting */
887 if (V_nd6_defifindex == ifp->if_index)
888 nd6_setdefaultiface(0);
890 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
891 /* Refresh default router list. */
896 * We do not nuke the neighbor cache entries here any more
897 * because the neighbor cache is kept in if_afdata[AF_INET6].
898 * nd6_purge() is invoked by in6_ifdetach() which is called
899 * from if_detach() where everything gets purged. So let
900 * in6_domifdetach() do the actual L2 table purging work.
905 * the caller acquires and releases the lock on the lltbls
906 * Returns the llentry locked
909 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
911 struct sockaddr_in6 sin6;
915 bzero(&sin6, sizeof(sin6));
916 sin6.sin6_len = sizeof(struct sockaddr_in6);
917 sin6.sin6_family = AF_INET6;
918 sin6.sin6_addr = *addr6;
920 IF_AFDATA_LOCK_ASSERT(ifp);
923 if (flags & ND6_CREATE)
924 llflags |= LLE_CREATE;
925 if (flags & ND6_EXCLUSIVE)
926 llflags |= LLE_EXCLUSIVE;
928 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
929 if ((ln != NULL) && (llflags & LLE_CREATE))
930 ln->ln_state = ND6_LLINFO_NOSTATE;
936 * Test whether a given IPv6 address is a neighbor or not, ignoring
937 * the actual neighbor cache. The neighbor cache is ignored in order
938 * to not reenter the routing code from within itself.
941 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
943 struct nd_prefix *pr;
944 struct ifaddr *dstaddr;
947 * A link-local address is always a neighbor.
948 * XXX: a link does not necessarily specify a single interface.
950 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
951 struct sockaddr_in6 sin6_copy;
955 * We need sin6_copy since sa6_recoverscope() may modify the
959 if (sa6_recoverscope(&sin6_copy))
960 return (0); /* XXX: should be impossible */
961 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
963 if (sin6_copy.sin6_scope_id == zone)
970 * If the address matches one of our addresses,
971 * it should be a neighbor.
972 * If the address matches one of our on-link prefixes, it should be a
975 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
976 if (pr->ndpr_ifp != ifp)
979 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
982 /* Always use the default FIB here. */
983 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
984 0, 0, RT_DEFAULT_FIB);
988 * This is the case where multiple interfaces
989 * have the same prefix, but only one is installed
990 * into the routing table and that prefix entry
991 * is not the one being examined here. In the case
992 * where RADIX_MPATH is enabled, multiple route
993 * entries (of the same rt_key value) will be
994 * installed because the interface addresses all
997 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
998 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
1005 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1006 &addr->sin6_addr, &pr->ndpr_mask))
1011 * If the address is assigned on the node of the other side of
1012 * a p2p interface, the address should be a neighbor.
1014 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
1015 if (dstaddr != NULL) {
1016 if (dstaddr->ifa_ifp == ifp) {
1024 * If the default router list is empty, all addresses are regarded
1025 * as on-link, and thus, as a neighbor.
1027 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1028 TAILQ_EMPTY(&V_nd_defrouter) &&
1029 V_nd6_defifindex == ifp->if_index) {
1038 * Detect if a given IPv6 address identifies a neighbor on a given link.
1039 * XXX: should take care of the destination of a p2p link?
1042 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
1044 struct llentry *lle;
1047 IF_AFDATA_UNLOCK_ASSERT(ifp);
1048 if (nd6_is_new_addr_neighbor(addr, ifp))
1052 * Even if the address matches none of our addresses, it might be
1053 * in the neighbor cache.
1055 IF_AFDATA_RLOCK(ifp);
1056 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1060 IF_AFDATA_RUNLOCK(ifp);
1065 * Free an nd6 llinfo entry.
1066 * Since the function would cause significant changes in the kernel, DO NOT
1067 * make it global, unless you have a strong reason for the change, and are sure
1068 * that the change is safe.
1070 static struct llentry *
1071 nd6_free(struct llentry **lnp, int gc)
1074 struct llentry *ln, *next;
1075 struct nd_defrouter *dr;
1080 LLE_WLOCK_ASSERT(ln);
1083 ifp = ln->lle_tbl->llt_ifp;
1084 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0)
1085 dr = defrouter_lookup_locked(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1090 if ((ln->la_flags & LLE_DELETED) == 0)
1091 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
1094 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1095 * even though it is not harmful, it was not really necessary.
1099 nd6_llinfo_settimer_locked(ln, -1);
1101 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1102 if (dr != NULL && dr->expire &&
1103 ln->ln_state == ND6_LLINFO_STALE && gc) {
1105 * If the reason for the deletion is just garbage
1106 * collection, and the neighbor is an active default
1107 * router, do not delete it. Instead, reset the GC
1108 * timer using the router's lifetime.
1109 * Simply deleting the entry would affect default
1110 * router selection, which is not necessarily a good
1111 * thing, especially when we're using router preference
1113 * XXX: the check for ln_state would be redundant,
1114 * but we intentionally keep it just in case.
1116 if (dr->expire > time_uptime)
1117 nd6_llinfo_settimer_locked(ln,
1118 (dr->expire - time_uptime) * hz);
1120 nd6_llinfo_settimer_locked(ln,
1121 (long)V_nd6_gctimer * hz);
1123 next = LIST_NEXT(ln, lle_next);
1132 * Unreachablity of a router might affect the default
1133 * router selection and on-link detection of advertised
1138 * Temporarily fake the state to choose a new default
1139 * router and to perform on-link determination of
1140 * prefixes correctly.
1141 * Below the state will be set correctly,
1142 * or the entry itself will be deleted.
1144 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1147 if (ln->ln_router || dr) {
1150 * We need to unlock to avoid a LOR with rt6_flush() with the
1151 * rnh and for the calls to pfxlist_onlink_check() and
1152 * defrouter_select() in the block further down for calls
1153 * into nd6_lookup(). We still hold a ref.
1158 * rt6_flush must be called whether or not the neighbor
1159 * is in the Default Router List.
1160 * See a corresponding comment in nd6_na_input().
1162 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
1167 * Since defrouter_select() does not affect the
1168 * on-link determination and MIP6 needs the check
1169 * before the default router selection, we perform
1172 pfxlist_onlink_check();
1175 * Refresh default router list.
1180 if (ln->ln_router || dr)
1185 * Before deleting the entry, remember the next entry as the
1186 * return value. We need this because pfxlist_onlink_check() above
1187 * might have freed other entries (particularly the old next entry) as
1188 * a side effect (XXX).
1190 next = LIST_NEXT(ln, lle_next);
1193 * Save to unlock. We still hold an extra reference and will not
1194 * free(9) in llentry_free() if someone else holds one as well.
1197 IF_AFDATA_LOCK(ifp);
1200 /* Guard against race with other llentry_free(). */
1201 if (ln->la_flags & LLE_LINKED) {
1205 LLE_FREE_LOCKED(ln);
1207 IF_AFDATA_UNLOCK(ifp);
1215 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1217 * XXX cost-effective methods?
1220 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1225 if ((dst6 == NULL) || (rt == NULL))
1229 IF_AFDATA_RLOCK(ifp);
1230 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1231 IF_AFDATA_RUNLOCK(ifp);
1235 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1239 * if we get upper-layer reachability confirmation many times,
1240 * it is possible we have false information.
1244 if (ln->ln_byhint > V_nd6_maxnudhint) {
1249 ln->ln_state = ND6_LLINFO_REACHABLE;
1250 if (!ND6_LLINFO_PERMANENT(ln)) {
1251 nd6_llinfo_settimer_locked(ln,
1252 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1260 * Rejuvenate this function for routing operations related
1264 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1266 struct sockaddr_in6 *gateway;
1267 struct nd_defrouter *dr;
1271 gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1282 * Only indirect routes are interesting.
1284 if ((rt->rt_flags & RTF_GATEWAY) == 0)
1287 * check for default route
1289 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1290 &SIN6(rt_key(rt))->sin6_addr)) {
1291 dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1303 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1305 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1306 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1307 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1310 if (ifp->if_afdata[AF_INET6] == NULL)
1311 return (EPFNOSUPPORT);
1313 case OSIOCGIFINFO_IN6:
1315 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1316 bzero(&ND, sizeof(ND));
1317 ND.linkmtu = IN6_LINKMTU(ifp);
1318 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1319 ND.basereachable = ND_IFINFO(ifp)->basereachable;
1320 ND.reachable = ND_IFINFO(ifp)->reachable;
1321 ND.retrans = ND_IFINFO(ifp)->retrans;
1322 ND.flags = ND_IFINFO(ifp)->flags;
1323 ND.recalctm = ND_IFINFO(ifp)->recalctm;
1324 ND.chlim = ND_IFINFO(ifp)->chlim;
1326 case SIOCGIFINFO_IN6:
1327 ND = *ND_IFINFO(ifp);
1329 case SIOCSIFINFO_IN6:
1331 * used to change host variables from userland.
1332 * intented for a use on router to reflect RA configurations.
1334 /* 0 means 'unspecified' */
1335 if (ND.linkmtu != 0) {
1336 if (ND.linkmtu < IPV6_MMTU ||
1337 ND.linkmtu > IN6_LINKMTU(ifp)) {
1341 ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1344 if (ND.basereachable != 0) {
1345 int obasereachable = ND_IFINFO(ifp)->basereachable;
1347 ND_IFINFO(ifp)->basereachable = ND.basereachable;
1348 if (ND.basereachable != obasereachable)
1349 ND_IFINFO(ifp)->reachable =
1350 ND_COMPUTE_RTIME(ND.basereachable);
1352 if (ND.retrans != 0)
1353 ND_IFINFO(ifp)->retrans = ND.retrans;
1355 ND_IFINFO(ifp)->chlim = ND.chlim;
1357 case SIOCSIFINFO_FLAGS:
1360 struct in6_ifaddr *ia;
1362 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1363 !(ND.flags & ND6_IFF_IFDISABLED)) {
1364 /* ifdisabled 1->0 transision */
1367 * If the interface is marked as ND6_IFF_IFDISABLED and
1368 * has an link-local address with IN6_IFF_DUPLICATED,
1369 * do not clear ND6_IFF_IFDISABLED.
1370 * See RFC 4862, Section 5.4.5.
1373 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1374 if (ifa->ifa_addr->sa_family != AF_INET6)
1376 ia = (struct in6_ifaddr *)ifa;
1377 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1378 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1381 IF_ADDR_RUNLOCK(ifp);
1384 /* LLA is duplicated. */
1385 ND.flags |= ND6_IFF_IFDISABLED;
1386 log(LOG_ERR, "Cannot enable an interface"
1387 " with a link-local address marked"
1390 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1391 if (ifp->if_flags & IFF_UP)
1394 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1395 (ND.flags & ND6_IFF_IFDISABLED)) {
1396 /* ifdisabled 0->1 transision */
1397 /* Mark all IPv6 address as tentative. */
1399 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1400 if (V_ip6_dad_count > 0 &&
1401 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1403 TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1405 if (ifa->ifa_addr->sa_family !=
1408 ia = (struct in6_ifaddr *)ifa;
1409 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1411 IF_ADDR_RUNLOCK(ifp);
1415 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1416 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1417 /* auto_linklocal 0->1 transision */
1419 /* If no link-local address on ifp, configure */
1420 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1421 in6_ifattach(ifp, NULL);
1422 } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1423 ifp->if_flags & IFF_UP) {
1425 * When the IF already has
1426 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1427 * address is assigned, and IFF_UP, try to
1431 TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1433 if (ifa->ifa_addr->sa_family !=
1436 ia = (struct in6_ifaddr *)ifa;
1437 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1440 IF_ADDR_RUNLOCK(ifp);
1442 /* No LLA is configured. */
1443 in6_ifattach(ifp, NULL);
1447 ND_IFINFO(ifp)->flags = ND.flags;
1450 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1451 /* sync kernel routing table with the default router list */
1455 case SIOCSPFXFLUSH_IN6:
1457 /* flush all the prefix advertised by routers */
1458 struct nd_prefix *pr, *next;
1460 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1461 struct in6_ifaddr *ia, *ia_next;
1463 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1466 /* do we really have to remove addresses as well? */
1467 /* XXXRW: in6_ifaddrhead locking. */
1468 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1470 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1473 if (ia->ia6_ndpr == pr)
1474 in6_purgeaddr(&ia->ia_ifa);
1480 case SIOCSRTRFLUSH_IN6:
1482 /* flush all the default routers */
1483 struct nd_drhead drq;
1484 struct nd_defrouter *dr;
1491 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL)
1492 defrouter_unlink(dr, &drq);
1494 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
1495 TAILQ_REMOVE(&drq, dr, dr_entry);
1502 case SIOCGNBRINFO_IN6:
1505 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1507 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1510 IF_AFDATA_RLOCK(ifp);
1511 ln = nd6_lookup(&nb_addr, 0, ifp);
1512 IF_AFDATA_RUNLOCK(ifp);
1518 nbi->state = ln->ln_state;
1519 nbi->asked = ln->la_asked;
1520 nbi->isrouter = ln->ln_router;
1521 if (ln->la_expire == 0)
1524 nbi->expire = ln->la_expire +
1525 (time_second - time_uptime);
1529 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1530 ndif->ifindex = V_nd6_defifindex;
1532 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1533 return (nd6_setdefaultiface(ndif->ifindex));
1539 * Create neighbor cache entry and cache link-layer address,
1540 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1543 * code - type dependent information
1546 * The caller of this function already acquired the ndp
1547 * cache table lock because the cache entry is returned.
1550 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1551 int lladdrlen, int type, int code)
1553 struct llentry *ln = NULL;
1560 uint16_t router = 0;
1561 struct sockaddr_in6 sin6;
1562 struct mbuf *chain = NULL;
1563 int static_route = 0;
1565 IF_AFDATA_UNLOCK_ASSERT(ifp);
1567 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1568 KASSERT(from != NULL, ("%s: from == NULL", __func__));
1570 /* nothing must be updated for unspecified address */
1571 if (IN6_IS_ADDR_UNSPECIFIED(from))
1575 * Validation about ifp->if_addrlen and lladdrlen must be done in
1578 * XXX If the link does not have link-layer adderss, what should
1579 * we do? (ifp->if_addrlen == 0)
1580 * Spec says nothing in sections for RA, RS and NA. There's small
1581 * description on it in NS section (RFC 2461 7.2.3).
1583 flags = lladdr ? ND6_EXCLUSIVE : 0;
1584 IF_AFDATA_RLOCK(ifp);
1585 ln = nd6_lookup(from, flags, ifp);
1586 IF_AFDATA_RUNLOCK(ifp);
1588 flags |= ND6_EXCLUSIVE;
1589 IF_AFDATA_LOCK(ifp);
1590 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
1591 IF_AFDATA_UNLOCK(ifp);
1594 /* do nothing if static ndp is set */
1595 if (ln->la_flags & LLE_STATIC) {
1604 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1605 if (olladdr && lladdr) {
1606 llchange = bcmp(lladdr, &ln->ll_addr,
1612 * newentry olladdr lladdr llchange (*=record)
1615 * 0 n y -- (3) * STALE
1617 * 0 y y y (5) * STALE
1618 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1619 * 1 -- y -- (7) * STALE
1622 if (lladdr) { /* (3-5) and (7) */
1624 * Record source link-layer address
1625 * XXX is it dependent to ifp->if_type?
1627 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1628 ln->la_flags |= LLE_VALID;
1629 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1633 if ((!olladdr && lladdr != NULL) || /* (3) */
1634 (olladdr && lladdr != NULL && llchange)) { /* (5) */
1636 newstate = ND6_LLINFO_STALE;
1637 } else /* (1-2,4) */
1641 if (lladdr == NULL) /* (6) */
1642 newstate = ND6_LLINFO_NOSTATE;
1644 newstate = ND6_LLINFO_STALE;
1649 * Update the state of the neighbor cache.
1651 ln->ln_state = newstate;
1653 if (ln->ln_state == ND6_LLINFO_STALE) {
1654 if (ln->la_hold != NULL)
1655 nd6_grab_holdchain(ln, &chain, &sin6);
1656 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1657 /* probe right away */
1658 nd6_llinfo_settimer_locked((void *)ln, 0);
1663 * ICMP6 type dependent behavior.
1665 * NS: clear IsRouter if new entry
1666 * RS: clear IsRouter
1667 * RA: set IsRouter if there's lladdr
1668 * redir: clear IsRouter if new entry
1671 * The spec says that we must set IsRouter in the following cases:
1672 * - If lladdr exist, set IsRouter. This means (1-5).
1673 * - If it is old entry (!newentry), set IsRouter. This means (7).
1674 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1675 * A quetion arises for (1) case. (1) case has no lladdr in the
1676 * neighbor cache, this is similar to (6).
1677 * This case is rare but we figured that we MUST NOT set IsRouter.
1679 * newentry olladdr lladdr llchange NS RS RA redir
1681 * 0 n n -- (1) c ? s
1682 * 0 y n -- (2) c s s
1683 * 0 n y -- (3) c s s
1686 * 1 -- n -- (6) c c c s
1687 * 1 -- y -- (7) c c s c s
1691 switch (type & 0xff) {
1692 case ND_NEIGHBOR_SOLICIT:
1694 * New entry must have is_router flag cleared.
1696 if (is_newentry) /* (6-7) */
1701 * If the icmp is a redirect to a better router, always set the
1702 * is_router flag. Otherwise, if the entry is newly created,
1703 * clear the flag. [RFC 2461, sec 8.3]
1705 if (code == ND_REDIRECT_ROUTER)
1707 else if (is_newentry) /* (6-7) */
1710 case ND_ROUTER_SOLICIT:
1712 * is_router flag must always be cleared.
1716 case ND_ROUTER_ADVERT:
1718 * Mark an entry with lladdr as a router.
1720 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1721 (is_newentry && lladdr)) { /* (7) */
1728 static_route = (ln->la_flags & LLE_STATIC);
1729 router = ln->ln_router;
1731 if (flags & ND6_EXCLUSIVE)
1739 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
1742 * When the link-layer address of a router changes, select the
1743 * best router again. In particular, when the neighbor entry is newly
1744 * created, it might affect the selection policy.
1745 * Question: can we restrict the first condition to the "is_newentry"
1747 * XXX: when we hear an RA from a new router with the link-layer
1748 * address option, defrouter_select() is called twice, since
1749 * defrtrlist_update called the function as well. However, I believe
1750 * we can compromise the overhead, since it only happens the first
1752 * XXX: although defrouter_select() should not have a bad effect
1753 * for those are not autoconfigured hosts, we explicitly avoid such
1756 if (do_update && router &&
1757 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1759 * guaranteed recursion
1767 if (flags & ND6_EXCLUSIVE)
1778 nd6_slowtimo(void *arg)
1780 CURVNET_SET((struct vnet *) arg);
1781 struct nd_ifinfo *nd6if;
1784 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1785 nd6_slowtimo, curvnet);
1786 IFNET_RLOCK_NOSLEEP();
1787 TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1788 if (ifp->if_afdata[AF_INET6] == NULL)
1790 nd6if = ND_IFINFO(ifp);
1791 if (nd6if->basereachable && /* already initialized */
1792 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1794 * Since reachable time rarely changes by router
1795 * advertisements, we SHOULD insure that a new random
1796 * value gets recomputed at least once every few hours.
1799 nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1800 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1803 IFNET_RUNLOCK_NOSLEEP();
1808 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
1809 struct sockaddr_in6 *sin6)
1812 LLE_WLOCK_ASSERT(ln);
1814 *chain = ln->la_hold;
1816 memcpy(sin6, L3_ADDR_SIN6(ln), sizeof(*sin6));
1818 if (ln->ln_state == ND6_LLINFO_STALE) {
1821 * The first time we send a packet to a
1822 * neighbor whose entry is STALE, we have
1823 * to change the state to DELAY and a sets
1824 * a timer to expire in DELAY_FIRST_PROBE_TIME
1825 * seconds to ensure do neighbor unreachability
1826 * detection on expiration.
1830 ln->ln_state = ND6_LLINFO_DELAY;
1831 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
1836 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1837 struct sockaddr_in6 *dst)
1841 struct ip6_hdr *ip6;
1845 mac_netinet6_nd6_send(ifp, m);
1849 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
1850 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
1851 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
1852 * to be diverted to user space. When re-injected into the kernel,
1853 * send_output() will directly dispatch them to the outgoing interface.
1855 if (send_sendso_input_hook != NULL) {
1856 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
1858 ip6 = mtod(m, struct ip6_hdr *);
1859 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
1860 /* Use the SEND socket */
1861 error = send_sendso_input_hook(m, ifp, SND_OUT,
1863 /* -1 == no app on SEND socket */
1864 if (error == 0 || error != -1)
1869 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1870 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
1871 mtod(m, struct ip6_hdr *));
1873 if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1876 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL);
1881 * IPv6 packet output - light version.
1882 * Checks if destination LLE exists and is in proper state
1883 * (e.g no modification required). If not true, fall back to
1887 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1888 struct sockaddr_in6 *dst, struct rtentry *rt0)
1890 struct llentry *ln = NULL;
1892 /* discard the packet if IPv6 operation is disabled on the interface */
1893 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1895 return (ENETDOWN); /* better error? */
1898 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1901 if (nd6_need_cache(ifp) == 0)
1904 IF_AFDATA_RLOCK(ifp);
1905 ln = nd6_lookup(&dst->sin6_addr, 0, ifp);
1906 IF_AFDATA_RUNLOCK(ifp);
1909 * Perform fast path for the following cases:
1910 * 1) lle state is REACHABLE
1911 * 2) lle state is DELAY (NS message sentNS message sent)
1913 * Every other case involves lle modification, so we handle
1916 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE &&
1917 ln->ln_state != ND6_LLINFO_DELAY)) {
1918 /* Fall back to slow processing path */
1921 return (nd6_output_lle(ifp, origifp, m, dst));
1928 return (nd6_output_ifp(ifp, origifp, m, dst));
1933 * Output IPv6 packet - heavy version.
1934 * Function assume that either
1935 * 1) destination LLE does not exist, is invalid or stale, so
1936 * ND6_EXCLUSIVE lock needs to be acquired
1937 * 2) destination lle is provided (with ND6_EXCLUSIVE lock),
1938 * in that case packets are queued in &chain.
1942 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1943 struct sockaddr_in6 *dst)
1945 struct llentry *lle = NULL;
1948 KASSERT(m != NULL, ("NULL mbuf, nothing to send"));
1949 /* discard the packet if IPv6 operation is disabled on the interface */
1950 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1952 return (ENETDOWN); /* better error? */
1955 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1958 if (nd6_need_cache(ifp) == 0)
1962 * Address resolution or Neighbor Unreachability Detection
1964 * At this point, the destination of the packet must be a unicast
1965 * or an anycast address(i.e. not a multicast).
1968 IF_AFDATA_RLOCK(ifp);
1969 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
1970 IF_AFDATA_RUNLOCK(ifp);
1971 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) {
1973 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1974 * the condition below is not very efficient. But we believe
1975 * it is tolerable, because this should be a rare case.
1977 flags = ND6_CREATE | ND6_EXCLUSIVE;
1978 IF_AFDATA_LOCK(ifp);
1979 lle = nd6_lookup(&dst->sin6_addr, flags, ifp);
1980 IF_AFDATA_UNLOCK(ifp);
1984 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1985 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
1986 char ip6buf[INET6_ADDRSTRLEN];
1988 "nd6_output: can't allocate llinfo for %s "
1990 ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
1994 goto sendpkt; /* send anyway */
1997 LLE_WLOCK_ASSERT(lle);
1999 /* We don't have to do link-layer address resolution on a p2p link. */
2000 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2001 lle->ln_state < ND6_LLINFO_REACHABLE) {
2002 lle->ln_state = ND6_LLINFO_STALE;
2003 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz);
2007 * The first time we send a packet to a neighbor whose entry is
2008 * STALE, we have to change the state to DELAY and a sets a timer to
2009 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2010 * neighbor unreachability detection on expiration.
2013 if (lle->ln_state == ND6_LLINFO_STALE) {
2015 lle->ln_state = ND6_LLINFO_DELAY;
2016 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz);
2020 * If the neighbor cache entry has a state other than INCOMPLETE
2021 * (i.e. its link-layer address is already resolved), just
2024 if (lle->ln_state > ND6_LLINFO_INCOMPLETE)
2028 * There is a neighbor cache entry, but no ethernet address
2029 * response yet. Append this latest packet to the end of the
2030 * packet queue in the mbuf, unless the number of the packet
2031 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen,
2032 * the oldest packet in the queue will be removed.
2034 if (lle->ln_state == ND6_LLINFO_NOSTATE)
2035 lle->ln_state = ND6_LLINFO_INCOMPLETE;
2037 if (lle->la_hold != NULL) {
2038 struct mbuf *m_hold;
2042 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2044 if (m_hold->m_nextpkt == NULL) {
2045 m_hold->m_nextpkt = m;
2049 while (i >= V_nd6_maxqueuelen) {
2050 m_hold = lle->la_hold;
2051 lle->la_hold = lle->la_hold->m_nextpkt;
2060 * If there has been no NS for the neighbor after entering the
2061 * INCOMPLETE state, send the first solicitation.
2063 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) {
2066 nd6_llinfo_settimer_locked(lle,
2067 (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2069 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL);
2071 /* We did the lookup so we need to do the unlock here. */
2081 return (nd6_output_ifp(ifp, origifp, m, dst));
2086 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2087 struct sockaddr_in6 *dst)
2089 struct mbuf *m, *m_head;
2090 struct ifnet *outifp;
2094 if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2101 m_head = m_head->m_nextpkt;
2102 error = nd6_output_ifp(ifp, origifp, m, dst);
2107 * note that intermediate errors are blindly ignored - but this is
2108 * the same convention as used with nd6_output when called by
2116 nd6_need_cache(struct ifnet *ifp)
2119 * XXX: we currently do not make neighbor cache on any interface
2120 * other than ARCnet, Ethernet, FDDI and GIF.
2123 * - unidirectional tunnels needs no ND
2125 switch (ifp->if_type) {
2133 #ifdef IFT_IEEE80211
2136 case IFT_INFINIBAND:
2138 case IFT_PROPVIRTUAL:
2146 * the callers of this function need to be re-worked to drop
2147 * the lle lock, drop here for now
2150 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
2151 const struct sockaddr *dst, u_char *desten, struct llentry **lle)
2156 IF_AFDATA_UNLOCK_ASSERT(ifp);
2157 if (m != NULL && m->m_flags & M_MCAST) {
2160 switch (ifp->if_type) {
2166 #ifdef IFT_IEEE80211
2171 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2176 * netbsd can use if_broadcastaddr, but we don't do so
2177 * to reduce # of ifdef.
2179 for (i = 0; i < ifp->if_addrlen; i++)
2187 return (EAFNOSUPPORT);
2193 * the entry should have been created in nd6_store_lladdr
2195 IF_AFDATA_RLOCK(ifp);
2196 ln = lla_lookup(LLTABLE6(ifp), 0, dst);
2197 IF_AFDATA_RUNLOCK(ifp);
2198 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
2201 /* this could happen, if we could not allocate memory */
2206 bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2210 * A *small* use after free race exists here
2216 clear_llinfo_pqueue(struct llentry *ln)
2218 struct mbuf *m_hold, *m_hold_next;
2220 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2221 m_hold_next = m_hold->m_nextpkt;
2228 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2229 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2231 SYSCTL_DECL(_net_inet6_icmp6);
2233 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2234 CTLFLAG_RD, nd6_sysctl_drlist, "");
2235 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2236 CTLFLAG_RD, nd6_sysctl_prlist, "");
2237 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2238 CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2239 SYSCTL_VNET_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2240 CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2243 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2245 struct in6_defrouter d;
2246 struct nd_defrouter *dr;
2249 if (req->newptr != NULL)
2252 error = sysctl_wire_old_buffer(req, 0);
2256 bzero(&d, sizeof(d));
2257 d.rtaddr.sin6_family = AF_INET6;
2258 d.rtaddr.sin6_len = sizeof(d.rtaddr);
2261 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2262 d.rtaddr.sin6_addr = dr->rtaddr;
2263 error = sa6_recoverscope(&d.rtaddr);
2266 d.flags = dr->raflags;
2267 d.rtlifetime = dr->rtlifetime;
2268 d.expire = dr->expire + (time_second - time_uptime);
2269 d.if_index = dr->ifp->if_index;
2270 error = SYSCTL_OUT(req, &d, sizeof(d));
2279 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2281 struct in6_prefix p;
2282 struct sockaddr_in6 s6;
2283 struct nd_prefix *pr;
2284 struct nd_pfxrouter *pfr;
2287 char ip6buf[INET6_ADDRSTRLEN];
2292 bzero(&p, sizeof(p));
2293 p.origin = PR_ORIG_RA;
2294 bzero(&s6, sizeof(s6));
2295 s6.sin6_family = AF_INET6;
2296 s6.sin6_len = sizeof(s6);
2301 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2302 p.prefix = pr->ndpr_prefix;
2303 if (sa6_recoverscope(&p.prefix)) {
2304 log(LOG_ERR, "scope error in prefix list (%s)\n",
2305 ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2306 /* XXX: press on... */
2308 p.raflags = pr->ndpr_raf;
2309 p.prefixlen = pr->ndpr_plen;
2310 p.vltime = pr->ndpr_vltime;
2311 p.pltime = pr->ndpr_pltime;
2312 p.if_index = pr->ndpr_ifp->if_index;
2313 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2316 /* XXX: we assume time_t is signed. */
2318 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2319 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2320 p.expire = pr->ndpr_lastupdate +
2322 (time_second - time_uptime);
2324 p.expire = maxexpire;
2326 p.refcnt = pr->ndpr_refcnt;
2327 p.flags = pr->ndpr_stateflags;
2329 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2331 error = SYSCTL_OUT(req, &p, sizeof(p));
2334 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2335 s6.sin6_addr = pfr->router->rtaddr;
2336 if (sa6_recoverscope(&s6))
2338 "scope error in prefix list (%s)\n",
2339 ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2340 error = SYSCTL_OUT(req, &s6, sizeof(s6));