1 /* $FreeBSD: src/sys/netinet6/nd6.c,v 1.2.2.15 2003/05/06 06:46:58 suz Exp $ */
2 /* $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include "opt_inet6.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/callout.h>
39 #include <sys/malloc.h>
41 #include <sys/socket.h>
42 #include <sys/sockio.h>
44 #include <sys/kernel.h>
45 #include <sys/protosw.h>
46 #include <sys/errno.h>
47 #include <sys/syslog.h>
48 #include <sys/queue.h>
49 #include <sys/sysctl.h>
50 #include <sys/mutex.h>
52 #include <sys/thread2.h>
53 #include <sys/mutex2.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 #include <net/netisr2.h>
60 #include <net/netmsg2.h>
62 #include <netinet/in.h>
63 #include <netinet/if_ether.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet6/in6_ifattach.h>
66 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #include <netinet6/nd6.h>
69 #include <netinet/icmp6.h>
71 #include <net/net_osdep.h>
73 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
74 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
76 #define SIN6(s) ((struct sockaddr_in6 *)s)
77 #define SDL(s) ((struct sockaddr_dl *)s)
80 * Determine if the route entry is a direct neighbor on the specified
81 * interface. The interface test is not done if ifp is passed as NULL.
82 * The route entry is a neighbor if all of the following are true:
84 * RTF_GATEWAY is FALSE
86 * rt_gateway family is AF_LINK
87 * rt_llinfo is non-NULL
88 * The interfaces matches (or ifp is passed as NULL)
90 * NOTE: rt_llinfo can be NULL with LLINFO set, so both must be
93 * NOTE: We can't use rt->rt_ifp to check for the interface, since
94 * it may be the loopback interface if the entry is for our
95 * own address on a non-loopback interface. Instead, we use
96 * rt->rt_ifa->ifa_ifp which should specify the REAL interface.
98 #define ND6_IFP_MATCHES(ifp, ifa_ifp) \
100 (ifa_ifp) == (ifp) || \
101 (((ifp)->if_flags & IFF_ISBRIDGE) && \
102 (ifa_ifp)->if_bridge == (ifp)->if_softc) \
105 #define ND6_RTENTRY_IS_NEIGHBOR(rt, ifp) \
106 (((rt)->rt_flags & RTF_GATEWAY) == 0 && \
107 ((rt)->rt_flags & RTF_LLINFO) != 0 && \
108 (rt)->rt_gateway->sa_family == AF_LINK && \
110 ND6_IFP_MATCHES((ifp), (rt)->rt_ifa->ifa_ifp) \
113 #define ND6_RTENTRY_IS_LLCLONING(rt) \
114 (((rt)->rt_flags & (RTF_PRCLONING | RTF_LLINFO)) == \
115 (RTF_PRCLONING | RTF_LLINFO) || \
116 ((rt)->rt_flags & RTF_CLONING))
119 int nd6_prune = 1; /* walk list every 1 seconds */
120 int nd6_delay = 5; /* delay first probe time 5 second */
121 int nd6_umaxtries = 3; /* maximum unicast query */
122 int nd6_mmaxtries = 3; /* maximum multicast query */
123 int nd6_useloopback = 1; /* use loopback interface for local traffic */
124 int nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */
126 /* preventing too many loops in ND option parsing */
127 int nd6_maxndopt = 10; /* max # of ND options allowed */
129 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
138 static int nd6_inuse, nd6_allocated;
140 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
141 struct nd_drhead nd_defrouter;
142 struct nd_prhead nd_prefix = { 0 };
143 struct mtx nd6_mtx = MTX_INITIALIZER("nd6");
145 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
146 static struct sockaddr_in6 all1_sa;
148 static void nd6_setmtu0 (struct ifnet *, struct nd_ifinfo *);
149 static int regen_tmpaddr (struct in6_ifaddr *);
150 static void nd6_slowtimo(void *);
151 static void nd6_slowtimo_dispatch(netmsg_t);
152 static void nd6_timer(void *);
153 static void nd6_timer_dispatch(netmsg_t);
155 static struct callout nd6_slowtimo_ch;
156 static struct netmsg_base nd6_slowtimo_netmsg;
158 static struct callout nd6_timer_ch;
159 static struct netmsg_base nd6_timer_netmsg;
164 static int nd6_init_done = 0;
168 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
172 all1_sa.sin6_family = AF_INET6;
173 all1_sa.sin6_len = sizeof(struct sockaddr_in6);
174 for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
175 all1_sa.sin6_addr.s6_addr[i] = 0xff;
177 /* initialization of the default router list */
178 TAILQ_INIT(&nd_defrouter);
183 callout_init_mp(&nd6_slowtimo_ch);
184 netmsg_init(&nd6_slowtimo_netmsg, NULL, &netisr_adone_rport,
185 MSGF_PRIORITY, nd6_slowtimo_dispatch);
186 callout_reset_bycpu(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
187 nd6_slowtimo, NULL, 0);
191 nd6_ifattach(struct ifnet *ifp)
193 struct nd_ifinfo *nd;
195 nd = kmalloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
199 nd->chlim = IPV6_DEFHLIM;
200 nd->basereachable = REACHABLE_TIME;
201 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
202 nd->retrans = RETRANS_TIMER;
204 nd->flags = ND6_IFF_PERFORMNUD;
205 /* A loopback interface always has link-local address. */
206 if (ip6_auto_linklocal || (ifp->if_flags & IFF_LOOPBACK))
207 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
208 /* A loopback interface does not need to accept RAs. */
209 if (ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK))
210 nd->flags |= ND6_IFF_ACCEPT_RTADV;
212 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
213 nd6_setmtu0(ifp, nd);
219 nd6_ifdetach(struct nd_ifinfo *nd)
225 * Reset ND level link MTU. This function is called when the physical MTU
226 * changes, which means we might have to adjust the ND level MTU.
229 nd6_setmtu(struct ifnet *ifp)
231 nd6_setmtu0(ifp, ND_IFINFO(ifp));
234 struct netmsg_nd6setmtu {
235 struct netmsg_base nmsg;
237 struct nd_ifinfo *ndi;
240 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
242 nd6_setmtu0_dispatch(netmsg_t msg)
244 struct netmsg_nd6setmtu *nmsg = (struct netmsg_nd6setmtu *)msg;
245 struct ifnet *ifp = nmsg->ifp;
246 struct nd_ifinfo *ndi = nmsg->ndi;
249 omaxmtu = ndi->maxmtu;
251 switch (ifp->if_type) {
253 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
255 case IFT_IEEE1394: /* XXX should be IEEE1394MTU(1500) */
256 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
259 case IFT_IEEE80211: /* XXX should be IEEE80211MTU(1500) */
260 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
264 ndi->maxmtu = ifp->if_mtu;
269 * Decreasing the interface MTU under IPV6 minimum MTU may cause
270 * undesirable situation. We thus notify the operator of the change
271 * explicitly. The check for omaxmtu is necessary to restrict the
272 * log to the case of changing the MTU, not initializing it.
274 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
275 log(LOG_NOTICE, "nd6_setmtu0: "
276 "new link MTU on %s (%lu) is too small for IPv6\n",
277 if_name(ifp), (unsigned long)ndi->maxmtu);
280 if (ndi->maxmtu > in6_maxmtu)
281 in6_setmaxmtu(); /* check all interfaces just in case */
283 lwkt_replymsg(&nmsg->nmsg.lmsg, 0);
287 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
289 struct netmsg_nd6setmtu nmsg;
291 netmsg_init(&nmsg.nmsg, NULL, &curthread->td_msgport, 0,
292 nd6_setmtu0_dispatch);
295 lwkt_domsg(netisr_cpuport(0), &nmsg.nmsg.lmsg, 0);
299 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
301 bzero(ndopts, sizeof(*ndopts));
302 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
304 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
307 ndopts->nd_opts_done = 1;
308 ndopts->nd_opts_search = NULL;
313 * Take one ND option.
316 nd6_option(union nd_opts *ndopts)
318 struct nd_opt_hdr *nd_opt;
322 panic("ndopts == NULL in nd6_option");
323 if (!ndopts->nd_opts_last)
324 panic("uninitialized ndopts in nd6_option");
325 if (!ndopts->nd_opts_search)
327 if (ndopts->nd_opts_done)
330 nd_opt = ndopts->nd_opts_search;
332 /* make sure nd_opt_len is inside the buffer */
333 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
334 bzero(ndopts, sizeof(*ndopts));
338 olen = nd_opt->nd_opt_len << 3;
341 * Message validation requires that all included
342 * options have a length that is greater than zero.
344 bzero(ndopts, sizeof(*ndopts));
348 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
349 if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
350 /* option overruns the end of buffer, invalid */
351 bzero(ndopts, sizeof(*ndopts));
353 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
354 /* reached the end of options chain */
355 ndopts->nd_opts_done = 1;
356 ndopts->nd_opts_search = NULL;
362 * Parse multiple ND options.
363 * This function is much easier to use, for ND routines that do not need
364 * multiple options of the same type.
367 nd6_options(union nd_opts *ndopts)
369 struct nd_opt_hdr *nd_opt;
373 panic("ndopts == NULL in nd6_options");
374 if (!ndopts->nd_opts_last)
375 panic("uninitialized ndopts in nd6_options");
376 if (!ndopts->nd_opts_search)
380 nd_opt = nd6_option(ndopts);
381 if (!nd_opt && !ndopts->nd_opts_last) {
383 * Message validation requires that all included
384 * options have a length that is greater than zero.
386 icmp6stat.icp6s_nd_badopt++;
387 bzero(ndopts, sizeof(*ndopts));
394 switch (nd_opt->nd_opt_type) {
395 case ND_OPT_SOURCE_LINKADDR:
396 case ND_OPT_TARGET_LINKADDR:
398 case ND_OPT_REDIRECTED_HEADER:
399 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
401 "duplicated ND6 option found (type=%d)\n",
402 nd_opt->nd_opt_type));
405 ndopts->nd_opt_array[nd_opt->nd_opt_type]
409 case ND_OPT_PREFIX_INFORMATION:
410 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
411 ndopts->nd_opt_array[nd_opt->nd_opt_type]
414 ndopts->nd_opts_pi_end =
415 (struct nd_opt_prefix_info *)nd_opt;
419 * Unknown options must be silently ignored,
420 * to accomodate future extension to the protocol.
423 "nd6_options: unsupported option %d - "
424 "option ignored\n", nd_opt->nd_opt_type));
429 if (i > nd6_maxndopt) {
430 icmp6stat.icp6s_nd_toomanyopt++;
431 nd6log((LOG_INFO, "too many loop in nd opt\n"));
435 if (ndopts->nd_opts_done)
443 * ND6 timer routine to expire default route list and prefix list
446 nd6_timer_dispatch(netmsg_t nmsg)
448 struct llinfo_nd6 *ln;
449 struct nd_defrouter *dr;
450 struct nd_prefix *pr;
452 struct in6_ifaddr *ia6, *nia6;
457 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
462 ln = llinfo_nd6.ln_next;
463 while (ln && ln != &llinfo_nd6) {
465 struct sockaddr_in6 *dst;
466 struct llinfo_nd6 *next = ln->ln_next;
467 /* XXX: used for the DELAY case only: */
468 struct nd_ifinfo *ndi = NULL;
470 if ((rt = ln->ln_rt) == NULL) {
474 if ((ifp = rt->rt_ifp) == NULL) {
478 ndi = ND_IFINFO(ifp);
479 dst = (struct sockaddr_in6 *)rt_key(rt);
481 if (ln->ln_expire > time_uptime) {
488 panic("rt=0 in nd6_timer(ln=%p)", ln);
489 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
490 panic("rt_llinfo(%p) is not equal to ln(%p)",
493 panic("dst=0 in nd6_timer(ln=%p)", ln);
495 switch (ln->ln_state) {
496 case ND6_LLINFO_WAITDELETE:
499 case ND6_LLINFO_INCOMPLETE:
500 if (ln->ln_asked++ >= nd6_mmaxtries) {
501 struct mbuf *m = ln->ln_hold;
505 * Fake rcvif to make ICMP error
506 * more helpful in diagnosing
508 * XXX: should we consider
511 m->m_pkthdr.rcvif = rt->rt_ifp;
515 * mbuf has empty MAC header, remove
516 * for icmp. XXX layer violation.
518 m_adj(m, ETHER_HDR_LEN);
519 icmp6_error(m, ICMP6_DST_UNREACH,
520 ICMP6_DST_UNREACH_ADDR, 0);
523 ln->ln_state = ND6_LLINFO_WAITDELETE;
524 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0);
526 ln->ln_expire = time_uptime +
527 ND_IFINFO(ifp)->retrans / 1000;
528 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
531 case ND6_LLINFO_REACHABLE:
533 ln->ln_state = ND6_LLINFO_STALE;
534 ln->ln_expire = time_uptime + nd6_gctimer;
538 case ND6_LLINFO_STALE:
539 /* Garbage Collection(RFC 2461 5.3) */
544 case ND6_LLINFO_DELAY:
545 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD)) {
548 ln->ln_state = ND6_LLINFO_PROBE;
549 ln->ln_expire = time_uptime +
551 nd6_ns_output(ifp, &dst->sin6_addr,
555 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
556 ln->ln_expire = time_uptime + nd6_gctimer;
559 case ND6_LLINFO_PROBE:
560 if (ln->ln_asked < nd6_umaxtries) {
562 ln->ln_expire = time_uptime +
563 ND_IFINFO(ifp)->retrans / 1000;
564 nd6_ns_output(ifp, &dst->sin6_addr,
565 &dst->sin6_addr, ln, 0);
567 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0);
575 /* expire default router list */
576 dr = TAILQ_FIRST(&nd_defrouter);
578 if (dr->expire && dr->expire < time_uptime) {
579 struct nd_defrouter *t;
580 t = TAILQ_NEXT(dr, dr_entry);
584 dr = TAILQ_NEXT(dr, dr_entry);
589 * expire interface addresses.
590 * in the past the loop was inside prefix expiry processing.
591 * However, from a stricter speci-confrmance standpoint, we should
592 * rather separate address lifetimes and prefix lifetimes.
595 for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
597 /* check address lifetime */
598 if (IFA6_IS_INVALID(ia6)) {
602 * If the expiring address is temporary, try
603 * regenerating a new one. This would be useful when
604 * we suspended a laptop PC, then turned it on after a
605 * period that could invalidate all temporary
606 * addresses. Although we may have to restart the
607 * loop (see below), it must be after purging the
608 * address. Otherwise, we'd see an infinite loop of
611 if (ip6_use_tempaddr &&
612 (ia6->ia6_flags & IN6_IFF_TEMPORARY)) {
613 if (regen_tmpaddr(ia6) == 0)
617 in6_purgeaddr(&ia6->ia_ifa);
620 goto addrloop; /* XXX: see below */
621 /* ia6 is no longer good, continue on to next */
624 if (IFA6_IS_DEPRECATED(ia6)) {
625 int oldflags = ia6->ia6_flags;
627 if ((oldflags & IN6_IFF_DEPRECATED) == 0) {
628 ia6->ia6_flags |= IN6_IFF_DEPRECATED;
629 in6_newaddrmsg((struct ifaddr *)ia6);
633 * If a temporary address has just become deprecated,
634 * regenerate a new one if possible.
636 if (ip6_use_tempaddr &&
637 (ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
638 !(oldflags & IN6_IFF_DEPRECATED)) {
640 if (regen_tmpaddr(ia6) == 0) {
642 * A new temporary address is
644 * XXX: this means the address chain
645 * has changed while we are still in
646 * the loop. Although the change
647 * would not cause disaster (because
648 * it's not a deletion, but an
649 * addition,) we'd rather restart the
650 * loop just for safety. Or does this
651 * significantly reduce performance??
658 * A new RA might have made a deprecated address
661 if (ia6->ia6_flags & IN6_IFF_DEPRECATED) {
662 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
663 in6_newaddrmsg((struct ifaddr *)ia6);
668 /* expire prefix list */
669 pr = nd_prefix.lh_first;
672 * check prefix lifetime.
673 * since pltime is just for autoconf, pltime processing for
674 * prefix is not necessary.
676 if (pr->ndpr_expire && pr->ndpr_expire < time_uptime) {
681 * address expiration and prefix expiration are
682 * separate. NEVER perform in6_purgeaddr here.
691 mtx_unlock(&nd6_mtx);
693 callout_reset(&nd6_timer_ch, nd6_prune * hz, nd6_timer, NULL);
697 nd6_timer(void *arg __unused)
699 struct lwkt_msg *lmsg = &nd6_timer_netmsg.lmsg;
701 KASSERT(mycpuid == 0, ("not on cpu0"));
703 if (lmsg->ms_flags & MSGF_DONE)
704 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg);
711 callout_init_mp(&nd6_timer_ch);
712 netmsg_init(&nd6_timer_netmsg, NULL, &netisr_adone_rport,
713 MSGF_PRIORITY, nd6_timer_dispatch);
714 callout_reset_bycpu(&nd6_timer_ch, hz, nd6_timer, NULL, 0);
718 regen_tmpaddr(struct in6_ifaddr *ia6) /* deprecated/invalidated temporary
721 struct ifaddr_container *ifac;
723 struct in6_ifaddr *public_ifa6 = NULL;
725 ifp = ia6->ia_ifa.ifa_ifp;
726 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
727 struct ifaddr *ifa = ifac->ifa;
728 struct in6_ifaddr *it6;
730 if (ifa->ifa_addr->sa_family != AF_INET6)
733 it6 = (struct in6_ifaddr *)ifa;
735 /* ignore no autoconf addresses. */
736 if (!(it6->ia6_flags & IN6_IFF_AUTOCONF))
739 /* ignore autoconf addresses with different prefixes. */
740 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
744 * Now we are looking at an autoconf address with the same
745 * prefix as ours. If the address is temporary and is still
746 * preferred, do not create another one. It would be rare, but
747 * could happen, for example, when we resume a laptop PC after
750 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) &&
751 !IFA6_IS_DEPRECATED(it6)) {
757 * This is a public autoconf address that has the same prefix
758 * as ours. If it is preferred, keep it. We can't break the
759 * loop here, because there may be a still-preferred temporary
760 * address with the prefix.
762 if (!IFA6_IS_DEPRECATED(it6))
766 if (public_ifa6 != NULL) {
769 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
770 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
771 " tmp addr,errno=%d\n", e);
781 * Nuke neighbor cache/prefix/default router management table, right before
785 nd6_purge(struct ifnet *ifp)
787 struct llinfo_nd6 *ln, *nln;
788 struct nd_defrouter *dr, *ndr, drany;
789 struct nd_prefix *pr, *npr;
791 /* Nuke default router list entries toward ifp */
792 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
794 * The first entry of the list may be stored in
795 * the routing table, so we'll delete it later.
797 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
798 ndr = TAILQ_NEXT(dr, dr_entry);
802 dr = TAILQ_FIRST(&nd_defrouter);
807 /* Nuke prefix list entries toward ifp */
808 for (pr = nd_prefix.lh_first; pr; pr = npr) {
810 if (pr->ndpr_ifp == ifp) {
812 * Previously, pr->ndpr_addr is removed as well,
813 * but I strongly believe we don't have to do it.
814 * nd6_purge() is only called from in6_ifdetach(),
815 * which removes all the associated interface addresses
817 * (jinmei@kame.net 20010129)
823 /* cancel default outgoing interface setting */
824 if (nd6_defifindex == ifp->if_index)
825 nd6_setdefaultiface(0);
827 if (!ip6_forwarding &&
828 (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)) {
829 /* refresh default router list */
830 bzero(&drany, sizeof(drany));
831 defrouter_delreq(&drany, 0);
836 * Nuke neighbor cache entries for the ifp.
837 * Note that rt->rt_ifp may not be the same as ifp,
838 * due to KAME goto ours hack. See RTM_RESOLVE case in
839 * nd6_rtrequest(), and ip6_input().
841 ln = llinfo_nd6.ln_next;
842 while (ln && ln != &llinfo_nd6) {
844 struct sockaddr_dl *sdl;
848 if (rt && rt->rt_gateway &&
849 rt->rt_gateway->sa_family == AF_LINK) {
850 sdl = (struct sockaddr_dl *)rt->rt_gateway;
851 if (sdl->sdl_index == ifp->if_index)
859 nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp)
862 struct sockaddr_in6 sin6;
864 bzero(&sin6, sizeof(sin6));
865 sin6.sin6_len = sizeof(struct sockaddr_in6);
866 sin6.sin6_family = AF_INET6;
867 sin6.sin6_addr = *addr6;
870 rt = rtlookup((struct sockaddr *)&sin6);
872 rt = rtpurelookup((struct sockaddr *)&sin6);
873 if (rt && !(rt->rt_flags & RTF_LLINFO)) {
875 * This is the case for the default route.
876 * If we want to create a neighbor cache for the address, we
877 * should free the route for the destination and allocate an
890 * If no route is available and create is set,
891 * we allocate a host route for the destination
892 * and treat it like an interface route.
893 * This hack is necessary for a neighbor which can't
894 * be covered by our own prefix.
898 ifa = ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
903 * Create a new route. RTF_LLINFO is necessary
904 * to create a Neighbor Cache entry for the
905 * destination in nd6_rtrequest which will be
906 * called in rtrequest via ifa->ifa_rtrequest.
908 if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
909 ifa->ifa_addr, (struct sockaddr *)&all1_sa,
910 (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
911 ~RTF_CLONING, &rt)) != 0) {
913 "nd6_lookup: failed to add route for a "
914 "neighbor(%s), errno=%d\n",
915 ip6_sprintf(addr6), e);
920 struct llinfo_nd6 *ln =
921 (struct llinfo_nd6 *)rt->rt_llinfo;
923 ln->ln_state = ND6_LLINFO_NOSTATE;
930 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) {
933 "nd6_lookup: failed to lookup %s (if = %s)\n",
934 ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
935 /* xxx more logs... kazu */
942 static struct rtentry *
943 nd6_neighbor_lookup(struct in6_addr *addr6, struct ifnet *ifp)
946 struct sockaddr_in6 sin6;
948 bzero(&sin6, sizeof(sin6));
949 sin6.sin6_len = sizeof(struct sockaddr_in6);
950 sin6.sin6_family = AF_INET6;
951 sin6.sin6_addr = *addr6;
953 rt = rtpurelookup((struct sockaddr *)&sin6);
958 if (!ND6_RTENTRY_IS_NEIGHBOR(rt, ifp)) {
959 if (nd6_onlink_ns_rfc4861 &&
960 (ND6_RTENTRY_IS_LLCLONING(rt) || /* not cloned yet */
961 (rt->rt_parent != NULL && /* cloning */
962 ND6_RTENTRY_IS_LLCLONING(rt->rt_parent)))) {
964 * If cloning ever happened or is happening,
965 * rtentry for addr6 would or will become a
976 * Detect if a given IPv6 address identifies a neighbor on a given link.
977 * XXX: should take care of the destination of a p2p link?
980 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
982 struct ifaddr_container *ifac;
985 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
986 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
989 * A link-local address is always a neighbor.
990 * XXX: we should use the sin6_scope_id field rather than the embedded
993 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
994 ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
998 * If the address matches one of our addresses,
999 * it should be a neighbor.
1001 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1002 struct ifaddr *ifa = ifac->ifa;
1004 if (ifa->ifa_addr->sa_family != AF_INET6)
1007 for (i = 0; i < 4; i++) {
1008 if ((IFADDR6(ifa).s6_addr32[i] ^
1009 addr->sin6_addr.s6_addr32[i]) &
1010 IFMASK6(ifa).s6_addr32[i])
1017 * Even if the address matches none of our addresses, it might be
1018 * in the neighbor cache.
1020 if (nd6_neighbor_lookup(&addr->sin6_addr, ifp) != NULL)
1029 * Free an nd6 llinfo entry.
1032 nd6_free(struct rtentry *rt)
1034 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1035 struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
1036 struct nd_defrouter *dr;
1039 * we used to have kpfctlinput(PRC_HOSTDEAD) here.
1040 * even though it is not harmful, it was not really necessary.
1043 /* XXX: this condition too restrictive? */
1044 if (!ip6_forwarding &&
1045 (ND_IFINFO(rt->rt_ifp)->flags & ND6_IFF_ACCEPT_RTADV)) {
1047 dr = defrouter_lookup(
1048 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1051 if (ln->ln_router || dr) {
1053 * rt6_flush must be called whether or not the neighbor
1054 * is in the Default Router List.
1055 * See a corresponding comment in nd6_na_input().
1057 rt6_flush(&in6, rt->rt_ifp);
1062 * Unreachablity of a router might affect the default
1063 * router selection and on-link detection of advertised
1068 * Temporarily fake the state to choose a new default
1069 * router and to perform on-link determination of
1070 * prefixes correctly.
1071 * Below the state will be set correctly,
1072 * or the entry itself will be deleted.
1074 ln->ln_state = ND6_LLINFO_INCOMPLETE;
1077 * Since defrouter_select() does not affect the
1078 * on-link determination and MIP6 needs the check
1079 * before the default router selection, we perform
1082 pfxlist_onlink_check();
1084 if (dr == TAILQ_FIRST(&nd_defrouter)) {
1086 * It is used as the current default router,
1087 * so we have to move it to the end of the
1088 * list and choose a new one.
1089 * XXX: it is not very efficient if this is
1092 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1093 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
1098 mtx_unlock(&nd6_mtx);
1102 * Before deleting the entry, remember the next entry as the
1103 * return value. We need this because pfxlist_onlink_check() above
1104 * might have freed other entries (particularly the old next entry) as
1105 * a side effect (XXX).
1110 * Detach the route from the routing tree and the list of neighbor
1111 * caches, and disable the route entry not to be used in already
1114 * ND expiry happens under one big timer.
1115 * To avoid overflowing the route socket, don't report this.
1116 * Now that RTM_MISS is reported when an address is unresolvable
1117 * the benefit of reporting this deletion is questionable.
1119 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1125 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1127 * XXX cost-effective metods?
1130 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1132 struct llinfo_nd6 *ln;
1135 * If the caller specified "rt", use that. Otherwise, resolve the
1136 * routing table by supplied "dst6".
1141 if (!(rt = nd6_lookup(dst6, 0, NULL)))
1145 if ((rt->rt_flags & RTF_GATEWAY) ||
1146 !(rt->rt_flags & RTF_LLINFO) ||
1147 rt->rt_llinfo == NULL || rt->rt_gateway == NULL ||
1148 rt->rt_gateway->sa_family != AF_LINK) {
1149 /* This is not a host route. */
1153 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1154 if (ln->ln_state < ND6_LLINFO_REACHABLE)
1158 * if we get upper-layer reachability confirmation many times,
1159 * it is possible we have false information.
1163 if (ln->ln_byhint > nd6_maxnudhint)
1167 ln->ln_state = ND6_LLINFO_REACHABLE;
1169 ln->ln_expire = time_uptime +
1170 ND_IFINFO(rt->rt_ifp)->reachable;
1174 nd6_rtrequest(int req, struct rtentry *rt)
1176 struct sockaddr *gate = rt->rt_gateway;
1177 struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1178 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1179 struct ifnet *ifp = rt->rt_ifp;
1182 if ((rt->rt_flags & RTF_GATEWAY))
1185 if (nd6_need_cache(ifp) == 0 && !(rt->rt_flags & RTF_HOST)) {
1187 * This is probably an interface direct route for a link
1188 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1189 * We do not need special treatment below for such a route.
1190 * Moreover, the RTF_LLINFO flag which would be set below
1191 * would annoy the ndp(8) command.
1196 if (req == RTM_RESOLVE &&
1197 (nd6_need_cache(ifp) == 0 || /* stf case */
1198 !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
1200 * FreeBSD and BSD/OS often make a cloned host route based
1201 * on a less-specific route (e.g. the default route).
1202 * If the less specific route does not have a "gateway"
1203 * (this is the case when the route just goes to a p2p or an
1204 * stf interface), we'll mistakenly make a neighbor cache for
1205 * the host route, and will see strange neighbor solicitation
1206 * for the corresponding destination. In order to avoid the
1207 * confusion, we check if the destination of the route is
1208 * a neighbor in terms of neighbor discovery, and stop the
1209 * process if not. Additionally, we remove the LLINFO flag
1210 * so that ndp(8) will not try to get the neighbor information
1211 * of the destination.
1213 rt->rt_flags &= ~RTF_LLINFO;
1220 * There is no backward compatibility :)
1222 * if (!(rt->rt_flags & RTF_HOST) &&
1223 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1224 * rt->rt_flags |= RTF_CLONING;
1226 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1228 * Case 1: This route should come from
1229 * a route to interface. RTF_LLINFO flag is set
1230 * for a host route whose destination should be
1231 * treated as on-link.
1233 rt_setgate(rt, rt_key(rt),
1234 (struct sockaddr *)&null_sdl);
1235 gate = rt->rt_gateway;
1236 SDL(gate)->sdl_type = ifp->if_type;
1237 SDL(gate)->sdl_index = ifp->if_index;
1239 ln->ln_expire = time_uptime;
1240 if (ln && ln->ln_expire == 0) {
1241 /* kludge for desktops */
1244 if ((rt->rt_flags & RTF_CLONING))
1248 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1249 * We don't do that here since llinfo is not ready yet.
1251 * There are also couple of other things to be discussed:
1252 * - unsolicited NA code needs improvement beforehand
1253 * - RFC2461 says we MAY send multicast unsolicited NA
1254 * (7.2.6 paragraph 4), however, it also says that we
1255 * SHOULD provide a mechanism to prevent multicast NA storm.
1256 * we don't have anything like it right now.
1257 * note that the mechanism needs a mutual agreement
1258 * between proxies, which means that we need to implement
1259 * a new protocol, or a new kludge.
1260 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1261 * we need to check ip6forwarding before sending it.
1262 * (or should we allow proxy ND configuration only for
1263 * routers? there's no mention about proxy ND from hosts)
1266 /* XXX it does not work */
1267 if ((rt->rt_flags & RTF_ANNOUNCE) && mycpuid == 0) {
1269 &SIN6(rt_key(rt))->sin6_addr,
1270 &SIN6(rt_key(rt))->sin6_addr,
1271 ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1277 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1279 * Address resolution isn't necessary for a point to
1280 * point link, so we can skip this test for a p2p link.
1282 if (gate->sa_family != AF_LINK ||
1283 gate->sa_len < sizeof(null_sdl)) {
1285 "nd6_rtrequest: bad gateway value: %s\n",
1289 SDL(gate)->sdl_type = ifp->if_type;
1290 SDL(gate)->sdl_index = ifp->if_index;
1293 break; /* This happens on a route change */
1295 * Case 2: This route may come from cloning, or a manual route
1296 * add with a LL address.
1298 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1299 rt->rt_llinfo = (caddr_t)ln;
1301 log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1306 bzero(ln, sizeof(*ln));
1308 /* this is required for "ndp" command. - shin */
1309 if (req == RTM_ADD) {
1311 * gate should have some valid AF_LINK entry,
1312 * and ln->ln_expire should have some lifetime
1313 * which is specified by ndp command.
1315 ln->ln_state = ND6_LLINFO_REACHABLE;
1319 * When req == RTM_RESOLVE, rt is created and
1320 * initialized in rtrequest(), so rt_expire is 0.
1322 ln->ln_state = ND6_LLINFO_NOSTATE;
1323 ln->ln_expire = time_uptime;
1325 rt->rt_flags |= RTF_LLINFO;
1326 ln->ln_next = llinfo_nd6.ln_next;
1327 llinfo_nd6.ln_next = ln;
1328 ln->ln_prev = &llinfo_nd6;
1329 ln->ln_next->ln_prev = ln;
1332 * check if rt_key(rt) is one of my address assigned
1335 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1336 &SIN6(rt_key(rt))->sin6_addr);
1338 caddr_t macp = nd6_ifptomac(ifp);
1340 ln->ln_state = ND6_LLINFO_REACHABLE;
1343 bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1344 SDL(gate)->sdl_alen = ifp->if_addrlen;
1346 if (nd6_useloopback) {
1347 rt->rt_ifp = loif; /* XXX */
1349 * Make sure rt_ifa be equal to the ifaddr
1350 * corresponding to the address.
1351 * We need this because when we refer
1352 * rt_ifa->ia6_flags in ip6_input, we assume
1353 * that the rt_ifa points to the address instead
1354 * of the loopback address.
1356 if (ifa != rt->rt_ifa) {
1357 IFAFREE(rt->rt_ifa);
1362 } else if (rt->rt_flags & RTF_ANNOUNCE) {
1364 ln->ln_state = ND6_LLINFO_REACHABLE;
1368 * Join solicited node multicast for proxy ND, and only
1369 * join it once on cpu0.
1371 if ((ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) {
1372 struct in6_addr llsol;
1375 llsol = SIN6(rt_key(rt))->sin6_addr;
1376 llsol.s6_addr16[0] = htons(0xff02);
1377 llsol.s6_addr16[1] = htons(ifp->if_index);
1378 llsol.s6_addr32[1] = 0;
1379 llsol.s6_addr32[2] = htonl(1);
1380 llsol.s6_addr8[12] = 0xff;
1382 if (!in6_addmulti(&llsol, ifp, &error)) {
1383 nd6log((LOG_ERR, "%s: failed to join "
1384 "%s (errno=%d)\n", if_name(ifp),
1385 ip6_sprintf(&llsol), error));
1395 * Leave from solicited node multicast for proxy ND, and only
1396 * leave it once on cpu0 (since we joined it once on cpu0).
1398 if ((rt->rt_flags & RTF_ANNOUNCE) &&
1399 (ifp->if_flags & IFF_MULTICAST) && mycpuid == 0) {
1400 struct in6_addr llsol;
1401 struct in6_multi *in6m;
1403 llsol = SIN6(rt_key(rt))->sin6_addr;
1404 llsol.s6_addr16[0] = htons(0xff02);
1405 llsol.s6_addr16[1] = htons(ifp->if_index);
1406 llsol.s6_addr32[1] = 0;
1407 llsol.s6_addr32[2] = htonl(1);
1408 llsol.s6_addr8[12] = 0xff;
1410 in6m = IN6_LOOKUP_MULTI(&llsol, ifp);
1415 ln->ln_next->ln_prev = ln->ln_prev;
1416 ln->ln_prev->ln_next = ln->ln_next;
1419 rt->rt_flags &= ~RTF_LLINFO;
1421 m_freem(ln->ln_hold);
1427 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1429 struct in6_drlist *drl = (struct in6_drlist *)data;
1430 struct in6_prlist *prl = (struct in6_prlist *)data;
1431 struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1432 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1433 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1434 struct nd_defrouter *dr, any;
1435 struct nd_prefix *pr;
1437 int i = 0, error = 0;
1440 case SIOCGDRLST_IN6:
1442 * obsolete API, use sysctl under net.inet6.icmp6
1444 bzero(drl, sizeof(*drl));
1446 dr = TAILQ_FIRST(&nd_defrouter);
1447 while (dr && i < DRLSTSIZ) {
1448 drl->defrouter[i].rtaddr = dr->rtaddr;
1449 if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1450 /* XXX: need to this hack for KAME stack */
1451 drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1454 "default router list contains a "
1455 "non-linklocal address(%s)\n",
1456 ip6_sprintf(&drl->defrouter[i].rtaddr));
1458 drl->defrouter[i].flags = dr->flags;
1459 drl->defrouter[i].rtlifetime = dr->rtlifetime;
1460 drl->defrouter[i].expire = dr->expire;
1461 drl->defrouter[i].if_index = dr->ifp->if_index;
1463 dr = TAILQ_NEXT(dr, dr_entry);
1465 mtx_unlock(&nd6_mtx);
1467 case SIOCGPRLST_IN6:
1469 * obsolete API, use sysctl under net.inet6.icmp6
1472 * XXX meaning of fields, especialy "raflags", is very
1473 * differnet between RA prefix list and RR/static prefix list.
1474 * how about separating ioctls into two?
1476 bzero(prl, sizeof(*prl));
1478 pr = nd_prefix.lh_first;
1479 while (pr && i < PRLSTSIZ) {
1480 struct nd_pfxrouter *pfr;
1483 in6_embedscope(&prl->prefix[i].prefix,
1484 &pr->ndpr_prefix, NULL, NULL);
1485 prl->prefix[i].raflags = pr->ndpr_raf;
1486 prl->prefix[i].prefixlen = pr->ndpr_plen;
1487 prl->prefix[i].vltime = pr->ndpr_vltime;
1488 prl->prefix[i].pltime = pr->ndpr_pltime;
1489 prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1490 prl->prefix[i].expire = pr->ndpr_expire;
1492 pfr = pr->ndpr_advrtrs.lh_first;
1496 #define RTRADDR prl->prefix[i].advrtr[j]
1497 RTRADDR = pfr->router->rtaddr;
1498 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1499 /* XXX: hack for KAME */
1500 RTRADDR.s6_addr16[1] = 0;
1503 "a router(%s) advertises "
1505 "non-link local address\n",
1506 ip6_sprintf(&RTRADDR));
1510 pfr = pfr->pfr_next;
1512 prl->prefix[i].advrtrs = j;
1513 prl->prefix[i].origin = PR_ORIG_RA;
1518 mtx_unlock(&nd6_mtx);
1521 case OSIOCGIFINFO_IN6:
1522 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1523 bzero(&ndi->ndi, sizeof(ndi->ndi));
1524 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
1525 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu;
1526 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable;
1527 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable;
1528 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans;
1529 ndi->ndi.flags = ND_IFINFO(ifp)->flags;
1530 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm;
1531 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim;
1533 case SIOCGIFINFO_IN6:
1534 ndi->ndi = *ND_IFINFO(ifp);
1535 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
1537 case SIOCSIFINFO_IN6:
1539 * used to change host variables from userland.
1540 * intented for a use on router to reflect RA configurations.
1542 /* 0 means 'unspecified' */
1543 if (ndi->ndi.linkmtu != 0) {
1544 if (ndi->ndi.linkmtu < IPV6_MMTU ||
1545 ndi->ndi.linkmtu > IN6_LINKMTU(ifp)) {
1549 ND_IFINFO(ifp)->linkmtu = ndi->ndi.linkmtu;
1552 if (ndi->ndi.basereachable != 0) {
1553 int obasereachable = ND_IFINFO(ifp)->basereachable;
1555 ND_IFINFO(ifp)->basereachable = ndi->ndi.basereachable;
1556 if (ndi->ndi.basereachable != obasereachable)
1557 ND_IFINFO(ifp)->reachable =
1558 ND_COMPUTE_RTIME(ndi->ndi.basereachable);
1560 if (ndi->ndi.retrans != 0)
1561 ND_IFINFO(ifp)->retrans = ndi->ndi.retrans;
1562 if (ndi->ndi.chlim != 0)
1563 ND_IFINFO(ifp)->chlim = ndi->ndi.chlim;
1565 case SIOCSIFINFO_FLAGS:
1566 if ((ndi->ndi.flags & ND6_IFF_AUTO_LINKLOCAL) &&
1567 !(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1568 /* auto_linklocal 0->1 transision */
1569 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1570 in6_ifattach(ifp, NULL);
1572 ND_IFINFO(ifp)->flags = ndi->ndi.flags;
1574 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */
1575 /* flush default router list */
1577 * xxx sumikawa: should not delete route if default
1578 * route equals to the top of default router list
1580 bzero(&any, sizeof(any));
1581 defrouter_delreq(&any, 0);
1583 /* xxx sumikawa: flush prefix list */
1585 case SIOCSPFXFLUSH_IN6:
1587 /* flush all the prefix advertised by routers */
1588 struct nd_prefix *pr, *next;
1591 for (pr = nd_prefix.lh_first; pr; pr = next) {
1592 struct in6_ifaddr *ia, *ia_next;
1594 next = pr->ndpr_next;
1596 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1599 /* do we really have to remove addresses as well? */
1600 for (ia = in6_ifaddr; ia; ia = ia_next) {
1601 /* ia might be removed. keep the next ptr. */
1602 ia_next = ia->ia_next;
1604 if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
1607 if (ia->ia6_ndpr == pr)
1608 in6_purgeaddr(&ia->ia_ifa);
1612 mtx_unlock(&nd6_mtx);
1615 case SIOCSRTRFLUSH_IN6:
1617 /* flush all the default routers */
1618 struct nd_defrouter *dr, *next;
1621 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1623 * The first entry of the list may be stored in
1624 * the routing table, so we'll delete it later.
1626 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1627 next = TAILQ_NEXT(dr, dr_entry);
1630 defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
1632 mtx_unlock(&nd6_mtx);
1635 case SIOCGNBRINFO_IN6:
1637 struct llinfo_nd6 *ln;
1638 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1641 * XXX: KAME specific hack for scoped addresses
1642 * XXXX: for other scopes than link-local?
1644 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1645 IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1646 u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1649 *idp = htons(ifp->if_index);
1653 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1655 mtx_unlock(&nd6_mtx);
1658 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1659 nbi->state = ln->ln_state;
1660 nbi->asked = ln->ln_asked;
1661 nbi->isrouter = ln->ln_router;
1662 nbi->expire = ln->ln_expire;
1663 mtx_unlock(&nd6_mtx);
1667 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1668 ndif->ifindex = nd6_defifindex;
1670 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
1671 return (nd6_setdefaultiface(ndif->ifindex));
1677 * Create neighbor cache entry and cache link-layer address,
1678 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1681 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1683 int type, /* ICMP6 type */
1684 int code /* type dependent information */)
1686 struct rtentry *rt = NULL;
1687 struct llinfo_nd6 *ln = NULL;
1689 struct sockaddr_dl *sdl = NULL;
1696 panic("ifp == NULL in nd6_cache_lladdr");
1698 panic("from == NULL in nd6_cache_lladdr");
1700 /* nothing must be updated for unspecified address */
1701 if (IN6_IS_ADDR_UNSPECIFIED(from))
1705 * Validation about ifp->if_addrlen and lladdrlen must be done in
1708 * XXX If the link does not have link-layer adderss, what should
1709 * we do? (ifp->if_addrlen == 0)
1710 * Spec says nothing in sections for RA, RS and NA. There's small
1711 * description on it in NS section (RFC 2461 7.2.3).
1714 rt = nd6_lookup(from, 0, ifp);
1717 /* nothing must be done if there's no lladdr */
1718 if (!lladdr || !lladdrlen)
1722 rt = nd6_lookup(from, 1, ifp);
1725 /* do nothing if static ndp is set */
1726 if (rt->rt_flags & RTF_STATIC)
1733 if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1738 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1741 if (!rt->rt_gateway)
1743 if (rt->rt_gateway->sa_family != AF_LINK)
1745 sdl = SDL(rt->rt_gateway);
1747 olladdr = (sdl->sdl_alen) ? 1 : 0;
1748 if (olladdr && lladdr) {
1749 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1757 * newentry olladdr lladdr llchange (*=record)
1760 * 0 n y -- (3) * STALE
1762 * 0 y y y (5) * STALE
1763 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1764 * 1 -- y -- (7) * STALE
1767 if (lladdr) { /* (3-5) and (7) */
1769 * Record source link-layer address
1770 * XXX is it dependent to ifp->if_type?
1772 sdl->sdl_alen = ifp->if_addrlen;
1773 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1777 if ((!olladdr && lladdr) || /* (3) */
1778 (olladdr && lladdr && llchange)) { /* (5) */
1780 newstate = ND6_LLINFO_STALE;
1781 } else { /* (1-2,4) */
1786 if (!lladdr) /* (6) */
1787 newstate = ND6_LLINFO_NOSTATE;
1789 newstate = ND6_LLINFO_STALE;
1794 * Update the state of the neighbor cache.
1796 ln->ln_state = newstate;
1798 if (ln->ln_state == ND6_LLINFO_STALE) {
1800 * XXX: since nd6_output() below will cause
1801 * state tansition to DELAY and reset the timer,
1802 * we must set the timer now, although it is actually
1805 ln->ln_expire = time_uptime + nd6_gctimer;
1809 * we assume ifp is not a p2p here, so just
1810 * set the 2nd argument as the 1st one.
1812 nd6_output(ifp, ifp, ln->ln_hold,
1813 (struct sockaddr_in6 *)rt_key(rt), rt);
1816 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1817 /* probe right away */
1818 ln->ln_expire = time_uptime;
1823 * ICMP6 type dependent behavior.
1825 * NS: clear IsRouter if new entry
1826 * RS: clear IsRouter
1827 * RA: set IsRouter if there's lladdr
1828 * redir: clear IsRouter if new entry
1831 * The spec says that we must set IsRouter in the following cases:
1832 * - If lladdr exist, set IsRouter. This means (1-5).
1833 * - If it is old entry (!newentry), set IsRouter. This means (7).
1834 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1835 * A quetion arises for (1) case. (1) case has no lladdr in the
1836 * neighbor cache, this is similar to (6).
1837 * This case is rare but we figured that we MUST NOT set IsRouter.
1839 * newentry olladdr lladdr llchange NS RS RA redir
1841 * 0 n n -- (1) c ? s
1842 * 0 y n -- (2) c s s
1843 * 0 n y -- (3) c s s
1846 * 1 -- n -- (6) c c c s
1847 * 1 -- y -- (7) c c s c s
1851 switch (type & 0xff) {
1852 case ND_NEIGHBOR_SOLICIT:
1854 * New entry must have is_router flag cleared.
1856 if (is_newentry) /* (6-7) */
1861 * If the icmp is a redirect to a better router, always set the
1862 * is_router flag. Otherwise, if the entry is newly created,
1863 * clear the flag. [RFC 2461, sec 8.3]
1865 if (code == ND_REDIRECT_ROUTER)
1867 else if (is_newentry) /* (6-7) */
1870 case ND_ROUTER_SOLICIT:
1872 * is_router flag must always be cleared.
1876 case ND_ROUTER_ADVERT:
1878 * Mark an entry with lladdr as a router.
1880 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */
1881 (is_newentry && lladdr)) { /* (7) */
1887 if (llchange || lladdr)
1888 rt_rtmsg(llchange ? RTM_CHANGE : RTM_ADD, rt, rt->rt_ifp, 0);
1891 * When the link-layer address of a router changes, select the
1892 * best router again. In particular, when the neighbor entry is newly
1893 * created, it might affect the selection policy.
1894 * Question: can we restrict the first condition to the "is_newentry"
1896 * XXX: when we hear an RA from a new router with the link-layer
1897 * address option, defrouter_select() is called twice, since
1898 * defrtrlist_update called the function as well. However, I believe
1899 * we can compromise the overhead, since it only happens the first
1901 * XXX: although defrouter_select() should not have a bad effect
1902 * for those are not autoconfigured hosts, we explicitly avoid such
1905 if (do_update && ln->ln_router && !ip6_forwarding &&
1906 (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV))
1913 nd6_slowtimo(void *arg __unused)
1915 struct lwkt_msg *lmsg = &nd6_slowtimo_netmsg.lmsg;
1917 KASSERT(mycpuid == 0, ("not on cpu0"));
1919 if (lmsg->ms_flags & MSGF_DONE)
1920 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg);
1925 nd6_slowtimo_dispatch(netmsg_t nmsg)
1927 const struct ifnet_array *arr;
1928 struct nd_ifinfo *nd6if;
1934 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
1937 arr = ifnet_array_get();
1940 for (i = 0; i < arr->ifnet_count; ++i) {
1941 struct ifnet *ifp = arr->ifnet_arr[i];
1943 if (ifp->if_afdata[AF_INET6] == NULL)
1945 nd6if = ND_IFINFO(ifp);
1946 if (nd6if->basereachable && /* already initialized */
1947 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1949 * Since reachable time rarely changes by router
1950 * advertisements, we SHOULD insure that a new random
1951 * value gets recomputed at least once every few hours.
1954 nd6if->recalctm = nd6_recalc_reachtm_interval;
1955 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1958 mtx_unlock(&nd6_mtx);
1960 callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1961 nd6_slowtimo, NULL);
1965 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1966 struct sockaddr_in6 *dst, struct rtentry *rt)
1970 if (ifp->if_flags & IFF_LOOPBACK)
1971 error = ifp->if_output(origifp, m, (struct sockaddr *)dst, rt);
1973 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, rt);
1978 nd6_resolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
1979 struct sockaddr *dst0, u_char *desten)
1981 struct sockaddr_in6 *dst = SIN6(dst0);
1982 struct rtentry *rt = NULL;
1983 struct llinfo_nd6 *ln = NULL;
1986 if (m->m_flags & M_MCAST) {
1987 switch (ifp->if_type) {
1992 #ifdef IFT_IEEE80211
1995 ETHER_MAP_IPV6_MULTICAST(&dst->sin6_addr,
1999 bcopy(ifp->if_broadcastaddr, desten, ifp->if_addrlen);
2002 error = EAFNOSUPPORT;
2008 error = rt_llroute(dst0, rt0, &rt);
2015 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2016 * the condition below is not very efficient. But we believe
2017 * it is tolerable, because this should be a rare case.
2019 if (ln == NULL && nd6_is_addr_neighbor(dst, ifp)) {
2020 rt = nd6_lookup(&dst->sin6_addr, 1, ifp);
2025 if (ln == NULL || rt == NULL) {
2026 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
2027 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2029 "nd6_output: can't allocate llinfo for %s "
2031 ip6_sprintf(&dst->sin6_addr), ln, rt);
2038 /* We don't have to do link-layer address resolution on a p2p link. */
2039 if ((ifp->if_flags & IFF_POINTOPOINT) &&
2040 ln->ln_state < ND6_LLINFO_REACHABLE) {
2041 ln->ln_state = ND6_LLINFO_STALE;
2042 ln->ln_expire = time_uptime + nd6_gctimer;
2046 * The first time we send a packet to a neighbor whose entry is
2047 * STALE, we have to change the state to DELAY and a sets a timer to
2048 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2049 * neighbor unreachability detection on expiration.
2052 if (ln->ln_state == ND6_LLINFO_STALE) {
2054 ln->ln_state = ND6_LLINFO_DELAY;
2055 ln->ln_expire = time_uptime + nd6_delay;
2059 * If the neighbor cache entry has a state other than INCOMPLETE
2060 * (i.e. its link-layer address is already resolved), return it.
2062 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) {
2063 struct sockaddr_dl *sdl = SDL(rt->rt_gateway);
2065 KKASSERT(sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0);
2066 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2071 * There is a neighbor cache entry, but no ethernet address
2072 * response yet. Replace the held mbuf (if any) with this
2076 m_freem(ln->ln_hold);
2080 * This code conforms to the rate-limiting rule described in Section
2081 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
2084 if (ln->ln_state == ND6_LLINFO_NOSTATE ||
2085 ln->ln_state == ND6_LLINFO_WAITDELETE) {
2087 * This neighbor cache entry was just created; change its
2088 * state to INCOMPLETE and start its life cycle.
2090 * We force an NS output below by setting ln_expire to 1
2091 * (nd6_rtrequest() could set it to the current time_uptime)
2092 * and zeroing out ln_asked (XXX this may not be necessary).
2094 ln->ln_state = ND6_LLINFO_INCOMPLETE;
2098 if (ln->ln_expire && ln->ln_expire < time_uptime && ln->ln_asked == 0) {
2100 ln->ln_expire = time_uptime + ND_IFINFO(ifp)->retrans / 1000;
2101 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2104 if (ln->ln_asked >= nd6_mmaxtries)
2105 return (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
2106 EHOSTUNREACH : EHOSTDOWN;
2115 nd6_need_cache(struct ifnet *ifp)
2118 * XXX: we currently do not make neighbor cache on any interface
2119 * other than Ethernet and GIF.
2122 * - unidirectional tunnels needs no ND
2124 switch (ifp->if_type) {
2130 #ifdef IFT_IEEE80211
2136 case IFT_GIF: /* XXX need more cases? */
2143 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2144 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2146 SYSCTL_DECL(_net_inet6_icmp6);
2148 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2149 CTLFLAG_RD, nd6_sysctl_drlist, "List default routers");
2150 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2151 CTLFLAG_RD, nd6_sysctl_prlist, "List prefixes");
2154 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2158 struct in6_defrouter *d, *de;
2159 struct nd_defrouter *dr;
2165 for (dr = TAILQ_FIRST(&nd_defrouter); dr;
2166 dr = TAILQ_NEXT(dr, dr_entry)) {
2167 d = (struct in6_defrouter *)buf;
2168 de = (struct in6_defrouter *)(buf + sizeof(buf));
2171 bzero(d, sizeof(*d));
2172 d->rtaddr.sin6_family = AF_INET6;
2173 d->rtaddr.sin6_len = sizeof(d->rtaddr);
2174 if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2178 "default router list (%s)\n",
2179 ip6_sprintf(&dr->rtaddr));
2180 d->flags = dr->flags;
2181 d->rtlifetime = dr->rtlifetime;
2182 d->expire = dr->expire;
2183 d->if_index = dr->ifp->if_index;
2185 panic("buffer too short");
2187 error = SYSCTL_OUT(req, buf, sizeof(*d));
2195 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2199 struct in6_prefix *p, *pe;
2200 struct nd_prefix *pr;
2206 for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2209 struct sockaddr_in6 *sin6, *s6;
2210 struct nd_pfxrouter *pfr;
2212 p = (struct in6_prefix *)buf;
2213 pe = (struct in6_prefix *)(buf + sizeof(buf));
2216 bzero(p, sizeof(*p));
2217 sin6 = (struct sockaddr_in6 *)(p + 1);
2219 p->prefix = pr->ndpr_prefix;
2220 if (in6_recoverscope(&p->prefix,
2221 &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2223 "scope error in prefix list (%s)\n",
2224 ip6_sprintf(&p->prefix.sin6_addr));
2225 p->raflags = pr->ndpr_raf;
2226 p->prefixlen = pr->ndpr_plen;
2227 p->vltime = pr->ndpr_vltime;
2228 p->pltime = pr->ndpr_pltime;
2229 p->if_index = pr->ndpr_ifp->if_index;
2230 p->expire = pr->ndpr_expire;
2231 p->refcnt = pr->ndpr_refcnt;
2232 p->flags = pr->ndpr_stateflags;
2233 p->origin = PR_ORIG_RA;
2235 for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
2236 pfr = pfr->pfr_next) {
2237 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
2241 s6 = &sin6[advrtrs];
2242 bzero(s6, sizeof(*s6));
2243 s6->sin6_family = AF_INET6;
2244 s6->sin6_len = sizeof(*sin6);
2245 if (in6_recoverscope(s6, &pfr->router->rtaddr,
2246 pfr->router->ifp) != 0)
2249 "prefix list (%s)\n",
2250 ip6_sprintf(&pfr->router->rtaddr));
2253 p->advrtrs = advrtrs;
2255 panic("buffer too short");
2258 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2259 error = SYSCTL_OUT(req, buf, advance);