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 <netinet/ip6.h>
66 #include <netinet6/ip6_var.h>
67 #include <netinet6/nd6.h>
68 #include <netinet/icmp6.h>
70 #include <net/net_osdep.h>
72 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
73 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
75 #define SIN6(s) ((struct sockaddr_in6 *)s)
76 #define SDL(s) ((struct sockaddr_dl *)s)
79 * Note that the check for rt_llinfo is necessary because a cloned
80 * route from a parent route that has the L flag (e.g. the default
81 * route to a p2p interface) may have the flag, too, while the
82 * destination is not actually a neighbor.
83 * XXX: we can't use rt->rt_ifp to check for the interface, since
84 * it might be the loopback interface if the entry is for our
85 * own address on a non-loopback interface. Instead, we should
86 * use rt->rt_ifa->ifa_ifp, which would specify the REAL
89 #define ND6_RTENTRY_IS_NEIGHBOR(rt, ifp) \
90 !(((rt)->rt_flags & RTF_GATEWAY) || \
91 ((rt)->rt_flags & RTF_LLINFO) == 0 || \
92 (rt)->rt_gateway->sa_family != AF_LINK || \
93 (rt)->rt_llinfo == NULL || \
94 ((ifp) != NULL && (rt)->rt_ifa->ifa_ifp != (ifp)))
96 #define ND6_RTENTRY_IS_LLCLONING(rt) \
97 (((rt)->rt_flags & (RTF_PRCLONING | RTF_LLINFO)) == \
98 (RTF_PRCLONING | RTF_LLINFO) || \
99 ((rt)->rt_flags & RTF_CLONING))
102 int nd6_prune
= 1; /* walk list every 1 seconds */
103 int nd6_delay
= 5; /* delay first probe time 5 second */
104 int nd6_umaxtries
= 3; /* maximum unicast query */
105 int nd6_mmaxtries
= 3; /* maximum multicast query */
106 int nd6_useloopback
= 1; /* use loopback interface for local traffic */
107 int nd6_gctimer
= (60 * 60 * 24); /* 1 day: garbage collection timer */
109 /* preventing too many loops in ND option parsing */
110 int nd6_maxndopt
= 10; /* max # of ND options allowed */
112 int nd6_maxnudhint
= 0; /* max # of subsequent upper layer hints */
121 static int nd6_inuse
, nd6_allocated
;
123 struct llinfo_nd6 llinfo_nd6
= {&llinfo_nd6
, &llinfo_nd6
};
124 struct nd_drhead nd_defrouter
;
125 struct nd_prhead nd_prefix
= { 0 };
126 struct mtx nd6_mtx
= MTX_INITIALIZER("nd6");
128 int nd6_recalc_reachtm_interval
= ND6_RECALC_REACHTM_INTERVAL
;
129 static struct sockaddr_in6 all1_sa
;
131 static void nd6_setmtu0 (struct ifnet
*, struct nd_ifinfo
*);
132 static int regen_tmpaddr (struct in6_ifaddr
*);
133 static void nd6_slowtimo(void *);
134 static void nd6_slowtimo_dispatch(netmsg_t
);
135 static void nd6_timer(void *);
136 static void nd6_timer_dispatch(netmsg_t
);
138 static struct callout nd6_slowtimo_ch
;
139 static struct netmsg_base nd6_slowtimo_netmsg
;
141 static struct callout nd6_timer_ch
;
142 static struct netmsg_base nd6_timer_netmsg
;
147 static int nd6_init_done
= 0;
151 log(LOG_NOTICE
, "nd6_init called more than once(ignored)\n");
155 all1_sa
.sin6_family
= AF_INET6
;
156 all1_sa
.sin6_len
= sizeof(struct sockaddr_in6
);
157 for (i
= 0; i
< sizeof(all1_sa
.sin6_addr
); i
++)
158 all1_sa
.sin6_addr
.s6_addr
[i
] = 0xff;
160 /* initialization of the default router list */
161 TAILQ_INIT(&nd_defrouter
);
166 callout_init_mp(&nd6_slowtimo_ch
);
167 netmsg_init(&nd6_slowtimo_netmsg
, NULL
, &netisr_adone_rport
,
168 MSGF_PRIORITY
, nd6_slowtimo_dispatch
);
169 callout_reset_bycpu(&nd6_slowtimo_ch
, ND6_SLOWTIMER_INTERVAL
* hz
,
170 nd6_slowtimo
, NULL
, 0);
174 nd6_ifattach(struct ifnet
*ifp
)
176 struct nd_ifinfo
*nd
;
178 nd
= (struct nd_ifinfo
*)kmalloc(sizeof(*nd
), M_IP6NDP
,
183 nd
->chlim
= IPV6_DEFHLIM
;
184 nd
->basereachable
= REACHABLE_TIME
;
185 nd
->reachable
= ND_COMPUTE_RTIME(nd
->basereachable
);
186 nd
->retrans
= RETRANS_TIMER
;
189 * Note that the default value of ip6_accept_rtadv is 0, which means
190 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
193 nd
->flags
= (ND6_IFF_PERFORMNUD
| ND6_IFF_ACCEPT_RTADV
);
195 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
196 nd6_setmtu0(ifp
, nd
);
201 nd6_ifdetach(struct nd_ifinfo
*nd
)
207 * Reset ND level link MTU. This function is called when the physical MTU
208 * changes, which means we might have to adjust the ND level MTU.
211 nd6_setmtu(struct ifnet
*ifp
)
213 nd6_setmtu0(ifp
, ND_IFINFO(ifp
));
216 struct netmsg_nd6setmtu
{
217 struct netmsg_base nmsg
;
219 struct nd_ifinfo
*ndi
;
222 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
224 nd6_setmtu0_dispatch(netmsg_t msg
)
226 struct netmsg_nd6setmtu
*nmsg
= (struct netmsg_nd6setmtu
*)msg
;
227 struct ifnet
*ifp
= nmsg
->ifp
;
228 struct nd_ifinfo
*ndi
= nmsg
->ndi
;
231 omaxmtu
= ndi
->maxmtu
;
233 switch (ifp
->if_type
) {
235 ndi
->maxmtu
= MIN(ETHERMTU
, ifp
->if_mtu
);
237 case IFT_IEEE1394
: /* XXX should be IEEE1394MTU(1500) */
238 ndi
->maxmtu
= MIN(ETHERMTU
, ifp
->if_mtu
);
241 case IFT_IEEE80211
: /* XXX should be IEEE80211MTU(1500) */
242 ndi
->maxmtu
= MIN(ETHERMTU
, ifp
->if_mtu
);
246 ndi
->maxmtu
= ifp
->if_mtu
;
251 * Decreasing the interface MTU under IPV6 minimum MTU may cause
252 * undesirable situation. We thus notify the operator of the change
253 * explicitly. The check for omaxmtu is necessary to restrict the
254 * log to the case of changing the MTU, not initializing it.
256 if (omaxmtu
>= IPV6_MMTU
&& ndi
->maxmtu
< IPV6_MMTU
) {
257 log(LOG_NOTICE
, "nd6_setmtu0: "
258 "new link MTU on %s (%lu) is too small for IPv6\n",
259 if_name(ifp
), (unsigned long)ndi
->maxmtu
);
262 if (ndi
->maxmtu
> in6_maxmtu
)
263 in6_setmaxmtu(); /* check all interfaces just in case */
265 lwkt_replymsg(&nmsg
->nmsg
.lmsg
, 0);
269 nd6_setmtu0(struct ifnet
*ifp
, struct nd_ifinfo
*ndi
)
271 struct netmsg_nd6setmtu nmsg
;
273 netmsg_init(&nmsg
.nmsg
, NULL
, &curthread
->td_msgport
, 0,
274 nd6_setmtu0_dispatch
);
277 lwkt_domsg(netisr_cpuport(0), &nmsg
.nmsg
.lmsg
, 0);
281 nd6_option_init(void *opt
, int icmp6len
, union nd_opts
*ndopts
)
283 bzero(ndopts
, sizeof(*ndopts
));
284 ndopts
->nd_opts_search
= (struct nd_opt_hdr
*)opt
;
286 = (struct nd_opt_hdr
*)(((u_char
*)opt
) + icmp6len
);
289 ndopts
->nd_opts_done
= 1;
290 ndopts
->nd_opts_search
= NULL
;
295 * Take one ND option.
298 nd6_option(union nd_opts
*ndopts
)
300 struct nd_opt_hdr
*nd_opt
;
304 panic("ndopts == NULL in nd6_option");
305 if (!ndopts
->nd_opts_last
)
306 panic("uninitialized ndopts in nd6_option");
307 if (!ndopts
->nd_opts_search
)
309 if (ndopts
->nd_opts_done
)
312 nd_opt
= ndopts
->nd_opts_search
;
314 /* make sure nd_opt_len is inside the buffer */
315 if ((caddr_t
)&nd_opt
->nd_opt_len
>= (caddr_t
)ndopts
->nd_opts_last
) {
316 bzero(ndopts
, sizeof(*ndopts
));
320 olen
= nd_opt
->nd_opt_len
<< 3;
323 * Message validation requires that all included
324 * options have a length that is greater than zero.
326 bzero(ndopts
, sizeof(*ndopts
));
330 ndopts
->nd_opts_search
= (struct nd_opt_hdr
*)((caddr_t
)nd_opt
+ olen
);
331 if (ndopts
->nd_opts_search
> ndopts
->nd_opts_last
) {
332 /* option overruns the end of buffer, invalid */
333 bzero(ndopts
, sizeof(*ndopts
));
335 } else if (ndopts
->nd_opts_search
== ndopts
->nd_opts_last
) {
336 /* reached the end of options chain */
337 ndopts
->nd_opts_done
= 1;
338 ndopts
->nd_opts_search
= NULL
;
344 * Parse multiple ND options.
345 * This function is much easier to use, for ND routines that do not need
346 * multiple options of the same type.
349 nd6_options(union nd_opts
*ndopts
)
351 struct nd_opt_hdr
*nd_opt
;
355 panic("ndopts == NULL in nd6_options");
356 if (!ndopts
->nd_opts_last
)
357 panic("uninitialized ndopts in nd6_options");
358 if (!ndopts
->nd_opts_search
)
362 nd_opt
= nd6_option(ndopts
);
363 if (!nd_opt
&& !ndopts
->nd_opts_last
) {
365 * Message validation requires that all included
366 * options have a length that is greater than zero.
368 icmp6stat
.icp6s_nd_badopt
++;
369 bzero(ndopts
, sizeof(*ndopts
));
376 switch (nd_opt
->nd_opt_type
) {
377 case ND_OPT_SOURCE_LINKADDR
:
378 case ND_OPT_TARGET_LINKADDR
:
380 case ND_OPT_REDIRECTED_HEADER
:
381 if (ndopts
->nd_opt_array
[nd_opt
->nd_opt_type
]) {
383 "duplicated ND6 option found (type=%d)\n",
384 nd_opt
->nd_opt_type
));
387 ndopts
->nd_opt_array
[nd_opt
->nd_opt_type
]
391 case ND_OPT_PREFIX_INFORMATION
:
392 if (ndopts
->nd_opt_array
[nd_opt
->nd_opt_type
] == 0) {
393 ndopts
->nd_opt_array
[nd_opt
->nd_opt_type
]
396 ndopts
->nd_opts_pi_end
=
397 (struct nd_opt_prefix_info
*)nd_opt
;
401 * Unknown options must be silently ignored,
402 * to accomodate future extension to the protocol.
405 "nd6_options: unsupported option %d - "
406 "option ignored\n", nd_opt
->nd_opt_type
));
411 if (i
> nd6_maxndopt
) {
412 icmp6stat
.icp6s_nd_toomanyopt
++;
413 nd6log((LOG_INFO
, "too many loop in nd opt\n"));
417 if (ndopts
->nd_opts_done
)
425 * ND6 timer routine to expire default route list and prefix list
428 nd6_timer_dispatch(netmsg_t nmsg
)
430 struct llinfo_nd6
*ln
;
431 struct nd_defrouter
*dr
;
432 struct nd_prefix
*pr
;
434 struct in6_ifaddr
*ia6
, *nia6
;
439 lwkt_replymsg(&nmsg
->lmsg
, 0); /* reply ASAP */
444 ln
= llinfo_nd6
.ln_next
;
445 while (ln
&& ln
!= &llinfo_nd6
) {
447 struct sockaddr_in6
*dst
;
448 struct llinfo_nd6
*next
= ln
->ln_next
;
449 /* XXX: used for the DELAY case only: */
450 struct nd_ifinfo
*ndi
= NULL
;
452 if ((rt
= ln
->ln_rt
) == NULL
) {
456 if ((ifp
= rt
->rt_ifp
) == NULL
) {
460 ndi
= ND_IFINFO(ifp
);
461 dst
= (struct sockaddr_in6
*)rt_key(rt
);
463 if (ln
->ln_expire
> time_uptime
) {
470 panic("rt=0 in nd6_timer(ln=%p)", ln
);
471 if (rt
->rt_llinfo
&& (struct llinfo_nd6
*)rt
->rt_llinfo
!= ln
)
472 panic("rt_llinfo(%p) is not equal to ln(%p)",
475 panic("dst=0 in nd6_timer(ln=%p)", ln
);
477 switch (ln
->ln_state
) {
478 case ND6_LLINFO_INCOMPLETE
:
479 if (ln
->ln_asked
< nd6_mmaxtries
) {
481 ln
->ln_expire
= time_uptime
+
482 ND_IFINFO(ifp
)->retrans
/ 1000;
483 nd6_ns_output(ifp
, NULL
, &dst
->sin6_addr
,
486 struct mbuf
*m
= ln
->ln_hold
;
490 * Fake rcvif to make ICMP error
491 * more helpful in diagnosing
493 * XXX: should we consider
496 m
->m_pkthdr
.rcvif
= rt
->rt_ifp
;
498 icmp6_error(m
, ICMP6_DST_UNREACH
,
499 ICMP6_DST_UNREACH_ADDR
, 0);
505 case ND6_LLINFO_REACHABLE
:
507 ln
->ln_state
= ND6_LLINFO_STALE
;
508 ln
->ln_expire
= time_uptime
+ nd6_gctimer
;
512 case ND6_LLINFO_STALE
:
513 /* Garbage Collection(RFC 2461 5.3) */
518 case ND6_LLINFO_DELAY
:
519 if (ndi
&& (ndi
->flags
& ND6_IFF_PERFORMNUD
)) {
522 ln
->ln_state
= ND6_LLINFO_PROBE
;
523 ln
->ln_expire
= time_uptime
+
525 nd6_ns_output(ifp
, &dst
->sin6_addr
,
529 ln
->ln_state
= ND6_LLINFO_STALE
; /* XXX */
530 ln
->ln_expire
= time_uptime
+ nd6_gctimer
;
533 case ND6_LLINFO_PROBE
:
534 if (ln
->ln_asked
< nd6_umaxtries
) {
536 ln
->ln_expire
= time_uptime
+
537 ND_IFINFO(ifp
)->retrans
/ 1000;
538 nd6_ns_output(ifp
, &dst
->sin6_addr
,
539 &dst
->sin6_addr
, ln
, 0);
548 /* expire default router list */
549 dr
= TAILQ_FIRST(&nd_defrouter
);
551 if (dr
->expire
&& dr
->expire
< time_uptime
) {
552 struct nd_defrouter
*t
;
553 t
= TAILQ_NEXT(dr
, dr_entry
);
557 dr
= TAILQ_NEXT(dr
, dr_entry
);
562 * expire interface addresses.
563 * in the past the loop was inside prefix expiry processing.
564 * However, from a stricter speci-confrmance standpoint, we should
565 * rather separate address lifetimes and prefix lifetimes.
568 for (ia6
= in6_ifaddr
; ia6
; ia6
= nia6
) {
570 /* check address lifetime */
571 if (IFA6_IS_INVALID(ia6
)) {
575 * If the expiring address is temporary, try
576 * regenerating a new one. This would be useful when
577 * we suspended a laptop PC, then turned it on after a
578 * period that could invalidate all temporary
579 * addresses. Although we may have to restart the
580 * loop (see below), it must be after purging the
581 * address. Otherwise, we'd see an infinite loop of
584 if (ip6_use_tempaddr
&&
585 (ia6
->ia6_flags
& IN6_IFF_TEMPORARY
)) {
586 if (regen_tmpaddr(ia6
) == 0)
590 in6_purgeaddr(&ia6
->ia_ifa
);
593 goto addrloop
; /* XXX: see below */
595 if (IFA6_IS_DEPRECATED(ia6
)) {
596 int oldflags
= ia6
->ia6_flags
;
598 ia6
->ia6_flags
|= IN6_IFF_DEPRECATED
;
601 * If a temporary address has just become deprecated,
602 * regenerate a new one if possible.
604 if (ip6_use_tempaddr
&&
605 (ia6
->ia6_flags
& IN6_IFF_TEMPORARY
) &&
606 !(oldflags
& IN6_IFF_DEPRECATED
)) {
608 if (regen_tmpaddr(ia6
) == 0) {
610 * A new temporary address is
612 * XXX: this means the address chain
613 * has changed while we are still in
614 * the loop. Although the change
615 * would not cause disaster (because
616 * it's not a deletion, but an
617 * addition,) we'd rather restart the
618 * loop just for safety. Or does this
619 * significantly reduce performance??
626 * A new RA might have made a deprecated address
629 ia6
->ia6_flags
&= ~IN6_IFF_DEPRECATED
;
633 /* expire prefix list */
634 pr
= nd_prefix
.lh_first
;
637 * check prefix lifetime.
638 * since pltime is just for autoconf, pltime processing for
639 * prefix is not necessary.
641 if (pr
->ndpr_expire
&& pr
->ndpr_expire
< time_uptime
) {
646 * address expiration and prefix expiration are
647 * separate. NEVER perform in6_purgeaddr here.
656 mtx_unlock(&nd6_mtx
);
658 callout_reset(&nd6_timer_ch
, nd6_prune
* hz
, nd6_timer
, NULL
);
662 nd6_timer(void *arg __unused
)
664 struct lwkt_msg
*lmsg
= &nd6_timer_netmsg
.lmsg
;
666 KASSERT(mycpuid
== 0, ("not on cpu0"));
668 if (lmsg
->ms_flags
& MSGF_DONE
)
669 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg
);
676 callout_init_mp(&nd6_timer_ch
);
677 netmsg_init(&nd6_timer_netmsg
, NULL
, &netisr_adone_rport
,
678 MSGF_PRIORITY
, nd6_timer_dispatch
);
679 callout_reset_bycpu(&nd6_timer_ch
, hz
, nd6_timer
, NULL
, 0);
683 regen_tmpaddr(struct in6_ifaddr
*ia6
) /* deprecated/invalidated temporary
686 struct ifaddr_container
*ifac
;
688 struct in6_ifaddr
*public_ifa6
= NULL
;
690 ifp
= ia6
->ia_ifa
.ifa_ifp
;
691 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
692 struct ifaddr
*ifa
= ifac
->ifa
;
693 struct in6_ifaddr
*it6
;
695 if (ifa
->ifa_addr
->sa_family
!= AF_INET6
)
698 it6
= (struct in6_ifaddr
*)ifa
;
700 /* ignore no autoconf addresses. */
701 if (!(it6
->ia6_flags
& IN6_IFF_AUTOCONF
))
704 /* ignore autoconf addresses with different prefixes. */
705 if (it6
->ia6_ndpr
== NULL
|| it6
->ia6_ndpr
!= ia6
->ia6_ndpr
)
709 * Now we are looking at an autoconf address with the same
710 * prefix as ours. If the address is temporary and is still
711 * preferred, do not create another one. It would be rare, but
712 * could happen, for example, when we resume a laptop PC after
715 if ((it6
->ia6_flags
& IN6_IFF_TEMPORARY
) &&
716 !IFA6_IS_DEPRECATED(it6
)) {
722 * This is a public autoconf address that has the same prefix
723 * as ours. If it is preferred, keep it. We can't break the
724 * loop here, because there may be a still-preferred temporary
725 * address with the prefix.
727 if (!IFA6_IS_DEPRECATED(it6
))
731 if (public_ifa6
!= NULL
) {
734 if ((e
= in6_tmpifadd(public_ifa6
, 0)) != 0) {
735 log(LOG_NOTICE
, "regen_tmpaddr: failed to create a new"
736 " tmp addr,errno=%d\n", e
);
746 * Nuke neighbor cache/prefix/default router management table, right before
750 nd6_purge(struct ifnet
*ifp
)
752 struct llinfo_nd6
*ln
, *nln
;
753 struct nd_defrouter
*dr
, *ndr
, drany
;
754 struct nd_prefix
*pr
, *npr
;
756 /* Nuke default router list entries toward ifp */
757 if ((dr
= TAILQ_FIRST(&nd_defrouter
)) != NULL
) {
759 * The first entry of the list may be stored in
760 * the routing table, so we'll delete it later.
762 for (dr
= TAILQ_NEXT(dr
, dr_entry
); dr
; dr
= ndr
) {
763 ndr
= TAILQ_NEXT(dr
, dr_entry
);
767 dr
= TAILQ_FIRST(&nd_defrouter
);
772 /* Nuke prefix list entries toward ifp */
773 for (pr
= nd_prefix
.lh_first
; pr
; pr
= npr
) {
775 if (pr
->ndpr_ifp
== ifp
) {
777 * Previously, pr->ndpr_addr is removed as well,
778 * but I strongly believe we don't have to do it.
779 * nd6_purge() is only called from in6_ifdetach(),
780 * which removes all the associated interface addresses
782 * (jinmei@kame.net 20010129)
788 /* cancel default outgoing interface setting */
789 if (nd6_defifindex
== ifp
->if_index
)
790 nd6_setdefaultiface(0);
792 if (!ip6_forwarding
&& ip6_accept_rtadv
) { /* XXX: too restrictive? */
793 /* refresh default router list */
794 bzero(&drany
, sizeof(drany
));
795 defrouter_delreq(&drany
, 0);
800 * Nuke neighbor cache entries for the ifp.
801 * Note that rt->rt_ifp may not be the same as ifp,
802 * due to KAME goto ours hack. See RTM_RESOLVE case in
803 * nd6_rtrequest(), and ip6_input().
805 ln
= llinfo_nd6
.ln_next
;
806 while (ln
&& ln
!= &llinfo_nd6
) {
808 struct sockaddr_dl
*sdl
;
812 if (rt
&& rt
->rt_gateway
&&
813 rt
->rt_gateway
->sa_family
== AF_LINK
) {
814 sdl
= (struct sockaddr_dl
*)rt
->rt_gateway
;
815 if (sdl
->sdl_index
== ifp
->if_index
)
823 nd6_lookup(struct in6_addr
*addr6
, int create
, struct ifnet
*ifp
)
826 struct sockaddr_in6 sin6
;
828 bzero(&sin6
, sizeof(sin6
));
829 sin6
.sin6_len
= sizeof(struct sockaddr_in6
);
830 sin6
.sin6_family
= AF_INET6
;
831 sin6
.sin6_addr
= *addr6
;
834 rt
= rtlookup((struct sockaddr
*)&sin6
);
836 rt
= rtpurelookup((struct sockaddr
*)&sin6
);
837 if (rt
&& !(rt
->rt_flags
& RTF_LLINFO
)) {
839 * This is the case for the default route.
840 * If we want to create a neighbor cache for the address, we
841 * should free the route for the destination and allocate an
854 * If no route is available and create is set,
855 * we allocate a host route for the destination
856 * and treat it like an interface route.
857 * This hack is necessary for a neighbor which can't
858 * be covered by our own prefix.
862 ifa
= ifaof_ifpforaddr((struct sockaddr
*)&sin6
, ifp
);
867 * Create a new route. RTF_LLINFO is necessary
868 * to create a Neighbor Cache entry for the
869 * destination in nd6_rtrequest which will be
870 * called in rtrequest via ifa->ifa_rtrequest.
872 if ((e
= rtrequest(RTM_ADD
, (struct sockaddr
*)&sin6
,
873 ifa
->ifa_addr
, (struct sockaddr
*)&all1_sa
,
874 (ifa
->ifa_flags
| RTF_HOST
| RTF_LLINFO
) &
875 ~RTF_CLONING
, &rt
)) != 0) {
877 "nd6_lookup: failed to add route for a "
878 "neighbor(%s), errno=%d\n",
879 ip6_sprintf(addr6
), e
);
884 struct llinfo_nd6
*ln
=
885 (struct llinfo_nd6
*)rt
->rt_llinfo
;
887 ln
->ln_state
= ND6_LLINFO_NOSTATE
;
894 if (!ND6_RTENTRY_IS_NEIGHBOR(rt
, ifp
)) {
897 "nd6_lookup: failed to lookup %s (if = %s)\n",
898 ip6_sprintf(addr6
), ifp
? if_name(ifp
) : "unspec");
899 /* xxx more logs... kazu */
906 static struct rtentry
*
907 nd6_neighbor_lookup(struct in6_addr
*addr6
, struct ifnet
*ifp
)
910 struct sockaddr_in6 sin6
;
912 bzero(&sin6
, sizeof(sin6
));
913 sin6
.sin6_len
= sizeof(struct sockaddr_in6
);
914 sin6
.sin6_family
= AF_INET6
;
915 sin6
.sin6_addr
= *addr6
;
917 rt
= rtpurelookup((struct sockaddr
*)&sin6
);
922 if (!ND6_RTENTRY_IS_NEIGHBOR(rt
, ifp
)) {
923 if (nd6_onlink_ns_rfc4861
&&
924 (ND6_RTENTRY_IS_LLCLONING(rt
) || /* not cloned yet */
925 (rt
->rt_parent
!= NULL
&& /* cloning */
926 ND6_RTENTRY_IS_LLCLONING(rt
->rt_parent
)))) {
928 * If cloning ever happened or is happening,
929 * rtentry for addr6 would or will become a
940 * Detect if a given IPv6 address identifies a neighbor on a given link.
941 * XXX: should take care of the destination of a p2p link?
944 nd6_is_addr_neighbor(struct sockaddr_in6
*addr
, struct ifnet
*ifp
)
946 struct ifaddr_container
*ifac
;
949 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
950 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
953 * A link-local address is always a neighbor.
954 * XXX: we should use the sin6_scope_id field rather than the embedded
957 if (IN6_IS_ADDR_LINKLOCAL(&addr
->sin6_addr
) &&
958 ntohs(*(u_int16_t
*)&addr
->sin6_addr
.s6_addr
[2]) == ifp
->if_index
)
962 * If the address matches one of our addresses,
963 * it should be a neighbor.
965 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
966 struct ifaddr
*ifa
= ifac
->ifa
;
968 if (ifa
->ifa_addr
->sa_family
!= AF_INET6
)
971 for (i
= 0; i
< 4; i
++) {
972 if ((IFADDR6(ifa
).s6_addr32
[i
] ^
973 addr
->sin6_addr
.s6_addr32
[i
]) &
974 IFMASK6(ifa
).s6_addr32
[i
])
981 * Even if the address matches none of our addresses, it might be
982 * in the neighbor cache.
984 if (nd6_neighbor_lookup(&addr
->sin6_addr
, ifp
) != NULL
)
993 * Free an nd6 llinfo entry.
996 nd6_free(struct rtentry
*rt
)
998 struct llinfo_nd6
*ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
, *next
;
999 struct in6_addr in6
= ((struct sockaddr_in6
*)rt_key(rt
))->sin6_addr
;
1000 struct nd_defrouter
*dr
;
1003 * we used to have kpfctlinput(PRC_HOSTDEAD) here.
1004 * even though it is not harmful, it was not really necessary.
1007 if (!ip6_forwarding
&& ip6_accept_rtadv
) { /* XXX: too restrictive? */
1009 dr
= defrouter_lookup(
1010 &((struct sockaddr_in6
*)rt_key(rt
))->sin6_addr
,
1013 if (ln
->ln_router
|| dr
) {
1015 * rt6_flush must be called whether or not the neighbor
1016 * is in the Default Router List.
1017 * See a corresponding comment in nd6_na_input().
1019 rt6_flush(&in6
, rt
->rt_ifp
);
1024 * Unreachablity of a router might affect the default
1025 * router selection and on-link detection of advertised
1030 * Temporarily fake the state to choose a new default
1031 * router and to perform on-link determination of
1032 * prefixes correctly.
1033 * Below the state will be set correctly,
1034 * or the entry itself will be deleted.
1036 ln
->ln_state
= ND6_LLINFO_INCOMPLETE
;
1039 * Since defrouter_select() does not affect the
1040 * on-link determination and MIP6 needs the check
1041 * before the default router selection, we perform
1044 pfxlist_onlink_check();
1046 if (dr
== TAILQ_FIRST(&nd_defrouter
)) {
1048 * It is used as the current default router,
1049 * so we have to move it to the end of the
1050 * list and choose a new one.
1051 * XXX: it is not very efficient if this is
1054 TAILQ_REMOVE(&nd_defrouter
, dr
, dr_entry
);
1055 TAILQ_INSERT_TAIL(&nd_defrouter
, dr
, dr_entry
);
1060 mtx_unlock(&nd6_mtx
);
1064 * Before deleting the entry, remember the next entry as the
1065 * return value. We need this because pfxlist_onlink_check() above
1066 * might have freed other entries (particularly the old next entry) as
1067 * a side effect (XXX).
1072 * Detach the route from the routing tree and the list of neighbor
1073 * caches, and disable the route entry not to be used in already
1076 rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
), 0, NULL
);
1082 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1084 * XXX cost-effective metods?
1087 nd6_nud_hint(struct rtentry
*rt
, struct in6_addr
*dst6
, int force
)
1089 struct llinfo_nd6
*ln
;
1092 * If the caller specified "rt", use that. Otherwise, resolve the
1093 * routing table by supplied "dst6".
1098 if (!(rt
= nd6_lookup(dst6
, 0, NULL
)))
1102 if ((rt
->rt_flags
& RTF_GATEWAY
) ||
1103 !(rt
->rt_flags
& RTF_LLINFO
) ||
1104 rt
->rt_llinfo
== NULL
|| rt
->rt_gateway
== NULL
||
1105 rt
->rt_gateway
->sa_family
!= AF_LINK
) {
1106 /* This is not a host route. */
1110 ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1111 if (ln
->ln_state
< ND6_LLINFO_REACHABLE
)
1115 * if we get upper-layer reachability confirmation many times,
1116 * it is possible we have false information.
1120 if (ln
->ln_byhint
> nd6_maxnudhint
)
1124 ln
->ln_state
= ND6_LLINFO_REACHABLE
;
1126 ln
->ln_expire
= time_uptime
+
1127 ND_IFINFO(rt
->rt_ifp
)->reachable
;
1131 nd6_rtrequest(int req
, struct rtentry
*rt
)
1133 struct sockaddr
*gate
= rt
->rt_gateway
;
1134 struct llinfo_nd6
*ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1135 static struct sockaddr_dl null_sdl
= {sizeof(null_sdl
), AF_LINK
};
1136 struct ifnet
*ifp
= rt
->rt_ifp
;
1139 if ((rt
->rt_flags
& RTF_GATEWAY
))
1142 if (nd6_need_cache(ifp
) == 0 && !(rt
->rt_flags
& RTF_HOST
)) {
1144 * This is probably an interface direct route for a link
1145 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1146 * We do not need special treatment below for such a route.
1147 * Moreover, the RTF_LLINFO flag which would be set below
1148 * would annoy the ndp(8) command.
1153 if (req
== RTM_RESOLVE
&&
1154 (nd6_need_cache(ifp
) == 0 || /* stf case */
1155 !nd6_is_addr_neighbor((struct sockaddr_in6
*)rt_key(rt
), ifp
))) {
1157 * FreeBSD and BSD/OS often make a cloned host route based
1158 * on a less-specific route (e.g. the default route).
1159 * If the less specific route does not have a "gateway"
1160 * (this is the case when the route just goes to a p2p or an
1161 * stf interface), we'll mistakenly make a neighbor cache for
1162 * the host route, and will see strange neighbor solicitation
1163 * for the corresponding destination. In order to avoid the
1164 * confusion, we check if the destination of the route is
1165 * a neighbor in terms of neighbor discovery, and stop the
1166 * process if not. Additionally, we remove the LLINFO flag
1167 * so that ndp(8) will not try to get the neighbor information
1168 * of the destination.
1170 rt
->rt_flags
&= ~RTF_LLINFO
;
1177 * There is no backward compatibility :)
1179 * if (!(rt->rt_flags & RTF_HOST) &&
1180 * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1181 * rt->rt_flags |= RTF_CLONING;
1183 if (rt
->rt_flags
& (RTF_CLONING
| RTF_LLINFO
)) {
1185 * Case 1: This route should come from
1186 * a route to interface. RTF_LLINFO flag is set
1187 * for a host route whose destination should be
1188 * treated as on-link.
1190 rt_setgate(rt
, rt_key(rt
),
1191 (struct sockaddr
*)&null_sdl
,
1193 gate
= rt
->rt_gateway
;
1194 SDL(gate
)->sdl_type
= ifp
->if_type
;
1195 SDL(gate
)->sdl_index
= ifp
->if_index
;
1197 ln
->ln_expire
= time_uptime
;
1198 if (ln
&& ln
->ln_expire
== 0) {
1199 /* kludge for desktops */
1202 if ((rt
->rt_flags
& RTF_CLONING
))
1206 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1207 * We don't do that here since llinfo is not ready yet.
1209 * There are also couple of other things to be discussed:
1210 * - unsolicited NA code needs improvement beforehand
1211 * - RFC2461 says we MAY send multicast unsolicited NA
1212 * (7.2.6 paragraph 4), however, it also says that we
1213 * SHOULD provide a mechanism to prevent multicast NA storm.
1214 * we don't have anything like it right now.
1215 * note that the mechanism needs a mutual agreement
1216 * between proxies, which means that we need to implement
1217 * a new protocol, or a new kludge.
1218 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1219 * we need to check ip6forwarding before sending it.
1220 * (or should we allow proxy ND configuration only for
1221 * routers? there's no mention about proxy ND from hosts)
1224 /* XXX it does not work */
1225 if ((rt
->rt_flags
& RTF_ANNOUNCE
) && mycpuid
== 0) {
1227 &SIN6(rt_key(rt
))->sin6_addr
,
1228 &SIN6(rt_key(rt
))->sin6_addr
,
1229 ip6_forwarding
? ND_NA_FLAG_ROUTER
: 0,
1235 if ((ifp
->if_flags
& (IFF_POINTOPOINT
| IFF_LOOPBACK
)) == 0) {
1237 * Address resolution isn't necessary for a point to
1238 * point link, so we can skip this test for a p2p link.
1240 if (gate
->sa_family
!= AF_LINK
||
1241 gate
->sa_len
< sizeof(null_sdl
)) {
1243 "nd6_rtrequest: bad gateway value: %s\n",
1247 SDL(gate
)->sdl_type
= ifp
->if_type
;
1248 SDL(gate
)->sdl_index
= ifp
->if_index
;
1251 break; /* This happens on a route change */
1253 * Case 2: This route may come from cloning, or a manual route
1254 * add with a LL address.
1256 R_Malloc(ln
, struct llinfo_nd6
*, sizeof(*ln
));
1257 rt
->rt_llinfo
= (caddr_t
)ln
;
1259 log(LOG_DEBUG
, "nd6_rtrequest: malloc failed\n");
1264 bzero(ln
, sizeof(*ln
));
1266 /* this is required for "ndp" command. - shin */
1267 if (req
== RTM_ADD
) {
1269 * gate should have some valid AF_LINK entry,
1270 * and ln->ln_expire should have some lifetime
1271 * which is specified by ndp command.
1273 ln
->ln_state
= ND6_LLINFO_REACHABLE
;
1277 * When req == RTM_RESOLVE, rt is created and
1278 * initialized in rtrequest(), so rt_expire is 0.
1280 ln
->ln_state
= ND6_LLINFO_NOSTATE
;
1281 ln
->ln_expire
= time_uptime
;
1283 rt
->rt_flags
|= RTF_LLINFO
;
1284 ln
->ln_next
= llinfo_nd6
.ln_next
;
1285 llinfo_nd6
.ln_next
= ln
;
1286 ln
->ln_prev
= &llinfo_nd6
;
1287 ln
->ln_next
->ln_prev
= ln
;
1290 * check if rt_key(rt) is one of my address assigned
1293 ifa
= (struct ifaddr
*)in6ifa_ifpwithaddr(rt
->rt_ifp
,
1294 &SIN6(rt_key(rt
))->sin6_addr
);
1296 caddr_t macp
= nd6_ifptomac(ifp
);
1298 ln
->ln_state
= ND6_LLINFO_REACHABLE
;
1301 bcopy(macp
, LLADDR(SDL(gate
)), ifp
->if_addrlen
);
1302 SDL(gate
)->sdl_alen
= ifp
->if_addrlen
;
1304 if (nd6_useloopback
) {
1305 rt
->rt_ifp
= loif
; /* XXX */
1307 * Make sure rt_ifa be equal to the ifaddr
1308 * corresponding to the address.
1309 * We need this because when we refer
1310 * rt_ifa->ia6_flags in ip6_input, we assume
1311 * that the rt_ifa points to the address instead
1312 * of the loopback address.
1314 if (ifa
!= rt
->rt_ifa
) {
1315 IFAFREE(rt
->rt_ifa
);
1320 } else if (rt
->rt_flags
& RTF_ANNOUNCE
) {
1322 ln
->ln_state
= ND6_LLINFO_REACHABLE
;
1326 * Join solicited node multicast for proxy ND, and only
1327 * join it once on cpu0.
1329 if ((ifp
->if_flags
& IFF_MULTICAST
) && mycpuid
== 0) {
1330 struct in6_addr llsol
;
1333 llsol
= SIN6(rt_key(rt
))->sin6_addr
;
1334 llsol
.s6_addr16
[0] = htons(0xff02);
1335 llsol
.s6_addr16
[1] = htons(ifp
->if_index
);
1336 llsol
.s6_addr32
[1] = 0;
1337 llsol
.s6_addr32
[2] = htonl(1);
1338 llsol
.s6_addr8
[12] = 0xff;
1340 if (!in6_addmulti(&llsol
, ifp
, &error
)) {
1341 nd6log((LOG_ERR
, "%s: failed to join "
1342 "%s (errno=%d)\n", if_name(ifp
),
1343 ip6_sprintf(&llsol
), error
));
1353 * Leave from solicited node multicast for proxy ND, and only
1354 * leave it once on cpu0 (since we joined it once on cpu0).
1356 if ((rt
->rt_flags
& RTF_ANNOUNCE
) &&
1357 (ifp
->if_flags
& IFF_MULTICAST
) && mycpuid
== 0) {
1358 struct in6_addr llsol
;
1359 struct in6_multi
*in6m
;
1361 llsol
= SIN6(rt_key(rt
))->sin6_addr
;
1362 llsol
.s6_addr16
[0] = htons(0xff02);
1363 llsol
.s6_addr16
[1] = htons(ifp
->if_index
);
1364 llsol
.s6_addr32
[1] = 0;
1365 llsol
.s6_addr32
[2] = htonl(1);
1366 llsol
.s6_addr8
[12] = 0xff;
1368 in6m
= IN6_LOOKUP_MULTI(&llsol
, ifp
);
1373 ln
->ln_next
->ln_prev
= ln
->ln_prev
;
1374 ln
->ln_prev
->ln_next
= ln
->ln_next
;
1377 rt
->rt_flags
&= ~RTF_LLINFO
;
1379 m_freem(ln
->ln_hold
);
1385 nd6_ioctl(u_long cmd
, caddr_t data
, struct ifnet
*ifp
)
1387 struct in6_drlist
*drl
= (struct in6_drlist
*)data
;
1388 struct in6_prlist
*prl
= (struct in6_prlist
*)data
;
1389 struct in6_ndireq
*ndi
= (struct in6_ndireq
*)data
;
1390 struct in6_nbrinfo
*nbi
= (struct in6_nbrinfo
*)data
;
1391 struct in6_ndifreq
*ndif
= (struct in6_ndifreq
*)data
;
1392 struct nd_defrouter
*dr
, any
;
1393 struct nd_prefix
*pr
;
1395 int i
= 0, error
= 0;
1398 case SIOCGDRLST_IN6
:
1400 * obsolete API, use sysctl under net.inet6.icmp6
1402 bzero(drl
, sizeof(*drl
));
1404 dr
= TAILQ_FIRST(&nd_defrouter
);
1405 while (dr
&& i
< DRLSTSIZ
) {
1406 drl
->defrouter
[i
].rtaddr
= dr
->rtaddr
;
1407 if (IN6_IS_ADDR_LINKLOCAL(&drl
->defrouter
[i
].rtaddr
)) {
1408 /* XXX: need to this hack for KAME stack */
1409 drl
->defrouter
[i
].rtaddr
.s6_addr16
[1] = 0;
1412 "default router list contains a "
1413 "non-linklocal address(%s)\n",
1414 ip6_sprintf(&drl
->defrouter
[i
].rtaddr
));
1416 drl
->defrouter
[i
].flags
= dr
->flags
;
1417 drl
->defrouter
[i
].rtlifetime
= dr
->rtlifetime
;
1418 drl
->defrouter
[i
].expire
= dr
->expire
;
1419 drl
->defrouter
[i
].if_index
= dr
->ifp
->if_index
;
1421 dr
= TAILQ_NEXT(dr
, dr_entry
);
1423 mtx_unlock(&nd6_mtx
);
1425 case SIOCGPRLST_IN6
:
1427 * obsolete API, use sysctl under net.inet6.icmp6
1430 * XXX meaning of fields, especialy "raflags", is very
1431 * differnet between RA prefix list and RR/static prefix list.
1432 * how about separating ioctls into two?
1434 bzero(prl
, sizeof(*prl
));
1436 pr
= nd_prefix
.lh_first
;
1437 while (pr
&& i
< PRLSTSIZ
) {
1438 struct nd_pfxrouter
*pfr
;
1441 in6_embedscope(&prl
->prefix
[i
].prefix
,
1442 &pr
->ndpr_prefix
, NULL
, NULL
);
1443 prl
->prefix
[i
].raflags
= pr
->ndpr_raf
;
1444 prl
->prefix
[i
].prefixlen
= pr
->ndpr_plen
;
1445 prl
->prefix
[i
].vltime
= pr
->ndpr_vltime
;
1446 prl
->prefix
[i
].pltime
= pr
->ndpr_pltime
;
1447 prl
->prefix
[i
].if_index
= pr
->ndpr_ifp
->if_index
;
1448 prl
->prefix
[i
].expire
= pr
->ndpr_expire
;
1450 pfr
= pr
->ndpr_advrtrs
.lh_first
;
1454 #define RTRADDR prl->prefix[i].advrtr[j]
1455 RTRADDR
= pfr
->router
->rtaddr
;
1456 if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR
)) {
1457 /* XXX: hack for KAME */
1458 RTRADDR
.s6_addr16
[1] = 0;
1461 "a router(%s) advertises "
1463 "non-link local address\n",
1464 ip6_sprintf(&RTRADDR
));
1468 pfr
= pfr
->pfr_next
;
1470 prl
->prefix
[i
].advrtrs
= j
;
1471 prl
->prefix
[i
].origin
= PR_ORIG_RA
;
1476 mtx_unlock(&nd6_mtx
);
1479 case OSIOCGIFINFO_IN6
:
1480 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
1481 bzero(&ndi
->ndi
, sizeof(ndi
->ndi
));
1482 ndi
->ndi
.linkmtu
= IN6_LINKMTU(ifp
);
1483 ndi
->ndi
.maxmtu
= ND_IFINFO(ifp
)->maxmtu
;
1484 ndi
->ndi
.basereachable
= ND_IFINFO(ifp
)->basereachable
;
1485 ndi
->ndi
.reachable
= ND_IFINFO(ifp
)->reachable
;
1486 ndi
->ndi
.retrans
= ND_IFINFO(ifp
)->retrans
;
1487 ndi
->ndi
.flags
= ND_IFINFO(ifp
)->flags
;
1488 ndi
->ndi
.recalctm
= ND_IFINFO(ifp
)->recalctm
;
1489 ndi
->ndi
.chlim
= ND_IFINFO(ifp
)->chlim
;
1491 case SIOCGIFINFO_IN6
:
1492 ndi
->ndi
= *ND_IFINFO(ifp
);
1493 ndi
->ndi
.linkmtu
= IN6_LINKMTU(ifp
);
1495 case SIOCSIFINFO_FLAGS
:
1496 ND_IFINFO(ifp
)->flags
= ndi
->ndi
.flags
;
1498 case SIOCSNDFLUSH_IN6
: /* XXX: the ioctl name is confusing... */
1499 /* flush default router list */
1501 * xxx sumikawa: should not delete route if default
1502 * route equals to the top of default router list
1504 bzero(&any
, sizeof(any
));
1505 defrouter_delreq(&any
, 0);
1507 /* xxx sumikawa: flush prefix list */
1509 case SIOCSPFXFLUSH_IN6
:
1511 /* flush all the prefix advertised by routers */
1512 struct nd_prefix
*pr
, *next
;
1515 for (pr
= nd_prefix
.lh_first
; pr
; pr
= next
) {
1516 struct in6_ifaddr
*ia
, *ia_next
;
1518 next
= pr
->ndpr_next
;
1520 if (IN6_IS_ADDR_LINKLOCAL(&pr
->ndpr_prefix
.sin6_addr
))
1523 /* do we really have to remove addresses as well? */
1524 for (ia
= in6_ifaddr
; ia
; ia
= ia_next
) {
1525 /* ia might be removed. keep the next ptr. */
1526 ia_next
= ia
->ia_next
;
1528 if (!(ia
->ia6_flags
& IN6_IFF_AUTOCONF
))
1531 if (ia
->ia6_ndpr
== pr
)
1532 in6_purgeaddr(&ia
->ia_ifa
);
1536 mtx_unlock(&nd6_mtx
);
1539 case SIOCSRTRFLUSH_IN6
:
1541 /* flush all the default routers */
1542 struct nd_defrouter
*dr
, *next
;
1545 if ((dr
= TAILQ_FIRST(&nd_defrouter
)) != NULL
) {
1547 * The first entry of the list may be stored in
1548 * the routing table, so we'll delete it later.
1550 for (dr
= TAILQ_NEXT(dr
, dr_entry
); dr
; dr
= next
) {
1551 next
= TAILQ_NEXT(dr
, dr_entry
);
1554 defrtrlist_del(TAILQ_FIRST(&nd_defrouter
));
1556 mtx_unlock(&nd6_mtx
);
1559 case SIOCGNBRINFO_IN6
:
1561 struct llinfo_nd6
*ln
;
1562 struct in6_addr nb_addr
= nbi
->addr
; /* make local for safety */
1565 * XXX: KAME specific hack for scoped addresses
1566 * XXXX: for other scopes than link-local?
1568 if (IN6_IS_ADDR_LINKLOCAL(&nbi
->addr
) ||
1569 IN6_IS_ADDR_MC_LINKLOCAL(&nbi
->addr
)) {
1570 u_int16_t
*idp
= (u_int16_t
*)&nb_addr
.s6_addr
[2];
1573 *idp
= htons(ifp
->if_index
);
1577 if ((rt
= nd6_lookup(&nb_addr
, 0, ifp
)) == NULL
) {
1579 mtx_unlock(&nd6_mtx
);
1582 ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1583 nbi
->state
= ln
->ln_state
;
1584 nbi
->asked
= ln
->ln_asked
;
1585 nbi
->isrouter
= ln
->ln_router
;
1586 nbi
->expire
= ln
->ln_expire
;
1587 mtx_unlock(&nd6_mtx
);
1591 case SIOCGDEFIFACE_IN6
: /* XXX: should be implemented as a sysctl? */
1592 ndif
->ifindex
= nd6_defifindex
;
1594 case SIOCSDEFIFACE_IN6
: /* XXX: should be implemented as a sysctl? */
1595 return (nd6_setdefaultiface(ndif
->ifindex
));
1601 * Create neighbor cache entry and cache link-layer address,
1602 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1605 nd6_cache_lladdr(struct ifnet
*ifp
, struct in6_addr
*from
, char *lladdr
,
1607 int type
, /* ICMP6 type */
1608 int code
/* type dependent information */)
1610 struct rtentry
*rt
= NULL
;
1611 struct llinfo_nd6
*ln
= NULL
;
1613 struct sockaddr_dl
*sdl
= NULL
;
1620 panic("ifp == NULL in nd6_cache_lladdr");
1622 panic("from == NULL in nd6_cache_lladdr");
1624 /* nothing must be updated for unspecified address */
1625 if (IN6_IS_ADDR_UNSPECIFIED(from
))
1629 * Validation about ifp->if_addrlen and lladdrlen must be done in
1632 * XXX If the link does not have link-layer adderss, what should
1633 * we do? (ifp->if_addrlen == 0)
1634 * Spec says nothing in sections for RA, RS and NA. There's small
1635 * description on it in NS section (RFC 2461 7.2.3).
1638 rt
= nd6_lookup(from
, 0, ifp
);
1641 /* nothing must be done if there's no lladdr */
1642 if (!lladdr
|| !lladdrlen
)
1646 rt
= nd6_lookup(from
, 1, ifp
);
1649 /* do nothing if static ndp is set */
1650 if (rt
->rt_flags
& RTF_STATIC
)
1657 if ((rt
->rt_flags
& (RTF_GATEWAY
| RTF_LLINFO
)) != RTF_LLINFO
) {
1662 ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1665 if (!rt
->rt_gateway
)
1667 if (rt
->rt_gateway
->sa_family
!= AF_LINK
)
1669 sdl
= SDL(rt
->rt_gateway
);
1671 olladdr
= (sdl
->sdl_alen
) ? 1 : 0;
1672 if (olladdr
&& lladdr
) {
1673 if (bcmp(lladdr
, LLADDR(sdl
), ifp
->if_addrlen
))
1681 * newentry olladdr lladdr llchange (*=record)
1684 * 0 n y -- (3) * STALE
1686 * 0 y y y (5) * STALE
1687 * 1 -- n -- (6) NOSTATE(= PASSIVE)
1688 * 1 -- y -- (7) * STALE
1691 if (lladdr
) { /* (3-5) and (7) */
1693 * Record source link-layer address
1694 * XXX is it dependent to ifp->if_type?
1696 sdl
->sdl_alen
= ifp
->if_addrlen
;
1697 bcopy(lladdr
, LLADDR(sdl
), ifp
->if_addrlen
);
1701 if ((!olladdr
&& lladdr
) || /* (3) */
1702 (olladdr
&& lladdr
&& llchange
)) { /* (5) */
1704 newstate
= ND6_LLINFO_STALE
;
1705 } else { /* (1-2,4) */
1710 if (!lladdr
) /* (6) */
1711 newstate
= ND6_LLINFO_NOSTATE
;
1713 newstate
= ND6_LLINFO_STALE
;
1718 * Update the state of the neighbor cache.
1720 ln
->ln_state
= newstate
;
1722 if (ln
->ln_state
== ND6_LLINFO_STALE
) {
1724 * XXX: since nd6_output() below will cause
1725 * state tansition to DELAY and reset the timer,
1726 * we must set the timer now, although it is actually
1729 ln
->ln_expire
= time_uptime
+ nd6_gctimer
;
1733 * we assume ifp is not a p2p here, so just
1734 * set the 2nd argument as the 1st one.
1736 nd6_output(ifp
, ifp
, ln
->ln_hold
,
1737 (struct sockaddr_in6
*)rt_key(rt
), rt
);
1740 } else if (ln
->ln_state
== ND6_LLINFO_INCOMPLETE
) {
1741 /* probe right away */
1742 ln
->ln_expire
= time_uptime
;
1747 * ICMP6 type dependent behavior.
1749 * NS: clear IsRouter if new entry
1750 * RS: clear IsRouter
1751 * RA: set IsRouter if there's lladdr
1752 * redir: clear IsRouter if new entry
1755 * The spec says that we must set IsRouter in the following cases:
1756 * - If lladdr exist, set IsRouter. This means (1-5).
1757 * - If it is old entry (!newentry), set IsRouter. This means (7).
1758 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1759 * A quetion arises for (1) case. (1) case has no lladdr in the
1760 * neighbor cache, this is similar to (6).
1761 * This case is rare but we figured that we MUST NOT set IsRouter.
1763 * newentry olladdr lladdr llchange NS RS RA redir
1765 * 0 n n -- (1) c ? s
1766 * 0 y n -- (2) c s s
1767 * 0 n y -- (3) c s s
1770 * 1 -- n -- (6) c c c s
1771 * 1 -- y -- (7) c c s c s
1775 switch (type
& 0xff) {
1776 case ND_NEIGHBOR_SOLICIT
:
1778 * New entry must have is_router flag cleared.
1780 if (is_newentry
) /* (6-7) */
1785 * If the icmp is a redirect to a better router, always set the
1786 * is_router flag. Otherwise, if the entry is newly created,
1787 * clear the flag. [RFC 2461, sec 8.3]
1789 if (code
== ND_REDIRECT_ROUTER
)
1791 else if (is_newentry
) /* (6-7) */
1794 case ND_ROUTER_SOLICIT
:
1796 * is_router flag must always be cleared.
1800 case ND_ROUTER_ADVERT
:
1802 * Mark an entry with lladdr as a router.
1804 if ((!is_newentry
&& (olladdr
|| lladdr
)) || /* (2-5) */
1805 (is_newentry
&& lladdr
)) { /* (7) */
1812 * When the link-layer address of a router changes, select the
1813 * best router again. In particular, when the neighbor entry is newly
1814 * created, it might affect the selection policy.
1815 * Question: can we restrict the first condition to the "is_newentry"
1817 * XXX: when we hear an RA from a new router with the link-layer
1818 * address option, defrouter_select() is called twice, since
1819 * defrtrlist_update called the function as well. However, I believe
1820 * we can compromise the overhead, since it only happens the first
1822 * XXX: although defrouter_select() should not have a bad effect
1823 * for those are not autoconfigured hosts, we explicitly avoid such
1826 if (do_update
&& ln
->ln_router
&& !ip6_forwarding
&& ip6_accept_rtadv
)
1833 nd6_slowtimo(void *arg __unused
)
1835 struct lwkt_msg
*lmsg
= &nd6_slowtimo_netmsg
.lmsg
;
1837 KASSERT(mycpuid
== 0, ("not on cpu0"));
1839 if (lmsg
->ms_flags
& MSGF_DONE
)
1840 lwkt_sendmsg_oncpu(netisr_cpuport(0), lmsg
);
1845 nd6_slowtimo_dispatch(netmsg_t nmsg
)
1847 const struct ifnet_array
*arr
;
1848 struct nd_ifinfo
*nd6if
;
1854 lwkt_replymsg(&nmsg
->lmsg
, 0); /* reply ASAP */
1857 arr
= ifnet_array_get();
1860 for (i
= 0; i
< arr
->ifnet_count
; ++i
) {
1861 struct ifnet
*ifp
= arr
->ifnet_arr
[i
];
1863 if (ifp
->if_afdata
[AF_INET6
] == NULL
)
1865 nd6if
= ND_IFINFO(ifp
);
1866 if (nd6if
->basereachable
&& /* already initialized */
1867 (nd6if
->recalctm
-= ND6_SLOWTIMER_INTERVAL
) <= 0) {
1869 * Since reachable time rarely changes by router
1870 * advertisements, we SHOULD insure that a new random
1871 * value gets recomputed at least once every few hours.
1874 nd6if
->recalctm
= nd6_recalc_reachtm_interval
;
1875 nd6if
->reachable
= ND_COMPUTE_RTIME(nd6if
->basereachable
);
1878 mtx_unlock(&nd6_mtx
);
1880 callout_reset(&nd6_slowtimo_ch
, ND6_SLOWTIMER_INTERVAL
* hz
,
1881 nd6_slowtimo
, NULL
);
1884 #define gotoerr(e) { error = (e); goto bad;}
1887 nd6_output(struct ifnet
*ifp
, struct ifnet
*origifp
, struct mbuf
*m
,
1888 struct sockaddr_in6
*dst
, struct rtentry
*rt
)
1890 struct llinfo_nd6
*ln
= NULL
;
1893 if (IN6_IS_ADDR_MULTICAST(&dst
->sin6_addr
))
1896 if (nd6_need_cache(ifp
) == 0)
1900 * Next hop determination. This routine is derived from rt_llroute.
1903 if (!(rt
->rt_flags
& RTF_UP
)) {
1904 rt
= rtlookup((struct sockaddr
*)dst
);
1906 gotoerr(EHOSTUNREACH
);
1908 if (rt
->rt_ifp
!= ifp
) {
1909 /* XXX: loop care? */
1910 return nd6_output(ifp
, origifp
, m
, dst
, rt
);
1913 if (rt
->rt_flags
& RTF_GATEWAY
) {
1914 struct sockaddr_in6
*gw6
;
1917 * We skip link-layer address resolution and NUD
1918 * if the gateway is not a neighbor from ND point
1919 * of view, regardless of the value of nd_ifinfo.flags.
1920 * The second condition is a bit tricky; we skip
1921 * if the gateway is our own address, which is
1922 * sometimes used to install a route to a p2p link.
1924 gw6
= (struct sockaddr_in6
*)rt
->rt_gateway
;
1925 if (!nd6_is_addr_neighbor(gw6
, ifp
) ||
1926 in6ifa_ifpwithaddr(ifp
, &gw6
->sin6_addr
)) {
1928 * We allow this kind of tricky route only
1929 * when the outgoing interface is p2p.
1930 * XXX: we may need a more generic rule here.
1932 if (!(ifp
->if_flags
& IFF_POINTOPOINT
))
1933 gotoerr(EHOSTUNREACH
);
1938 if (rt
->rt_gwroute
== NULL
) {
1939 rt
->rt_gwroute
= rtlookup(rt
->rt_gateway
);
1940 if (rt
->rt_gwroute
== NULL
)
1941 gotoerr(EHOSTUNREACH
);
1942 } else if (!(rt
->rt_gwroute
->rt_flags
& RTF_UP
)) {
1943 rtfree(rt
->rt_gwroute
);
1944 rt
->rt_gwroute
= rtlookup(rt
->rt_gateway
);
1945 if (rt
->rt_gwroute
== NULL
)
1946 gotoerr(EHOSTUNREACH
);
1948 rt
= rt
->rt_gwroute
;
1953 * Address resolution or Neighbor Unreachability Detection
1955 * At this point, the destination of the packet must be a unicast
1956 * or an anycast address(i.e. not a multicast).
1959 /* Look up the neighbor cache for the nexthop */
1960 if (rt
&& (rt
->rt_flags
& RTF_LLINFO
))
1961 ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1964 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1965 * the condition below is not very efficient. But we believe
1966 * it is tolerable, because this should be a rare case.
1968 if (nd6_is_addr_neighbor(dst
, ifp
) &&
1969 (rt
= nd6_lookup(&dst
->sin6_addr
, 1, ifp
)) != NULL
)
1970 ln
= (struct llinfo_nd6
*)rt
->rt_llinfo
;
1973 if (!(ifp
->if_flags
& IFF_POINTOPOINT
) &&
1974 !(ND_IFINFO(ifp
)->flags
& ND6_IFF_PERFORMNUD
)) {
1976 "nd6_output: can't allocate llinfo for %s "
1978 ip6_sprintf(&dst
->sin6_addr
), ln
, rt
);
1979 gotoerr(EIO
); /* XXX: good error? */
1982 goto sendpkt
; /* send anyway */
1985 /* We don't have to do link-layer address resolution on a p2p link. */
1986 if ((ifp
->if_flags
& IFF_POINTOPOINT
) &&
1987 ln
->ln_state
< ND6_LLINFO_REACHABLE
) {
1988 ln
->ln_state
= ND6_LLINFO_STALE
;
1989 ln
->ln_expire
= time_uptime
+ nd6_gctimer
;
1993 * The first time we send a packet to a neighbor whose entry is
1994 * STALE, we have to change the state to DELAY and a sets a timer to
1995 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1996 * neighbor unreachability detection on expiration.
1999 if (ln
->ln_state
== ND6_LLINFO_STALE
) {
2001 ln
->ln_state
= ND6_LLINFO_DELAY
;
2002 ln
->ln_expire
= time_uptime
+ nd6_delay
;
2006 * If the neighbor cache entry has a state other than INCOMPLETE
2007 * (i.e. its link-layer address is already resolved), just
2010 if (ln
->ln_state
> ND6_LLINFO_INCOMPLETE
)
2014 * There is a neighbor cache entry, but no ethernet address
2015 * response yet. Replace the held mbuf (if any) with this
2018 * This code conforms to the rate-limiting rule described in Section
2019 * 7.2.2 of RFC 2461, because the timer is set correctly after sending
2022 if (ln
->ln_state
== ND6_LLINFO_NOSTATE
) {
2024 * This neighbor cache entry was just created; change its
2025 * state to INCOMPLETE and start its life cycle.
2027 * We force an NS output below by setting ln_expire to 1
2028 * (nd6_rtrequest() could set it to the current time_uptime)
2029 * and zeroing out ln_asked (XXX this may not be necessary).
2031 ln
->ln_state
= ND6_LLINFO_INCOMPLETE
;
2036 m_freem(ln
->ln_hold
);
2038 if (ln
->ln_expire
) {
2039 if (ln
->ln_asked
< nd6_mmaxtries
&&
2040 ln
->ln_expire
< time_uptime
) {
2042 ln
->ln_expire
= time_uptime
+
2043 ND_IFINFO(ifp
)->retrans
/ 1000;
2044 nd6_ns_output(ifp
, NULL
, &dst
->sin6_addr
, ln
, 0);
2050 if (ifp
->if_flags
& IFF_LOOPBACK
)
2051 error
= ifp
->if_output(origifp
, m
, (struct sockaddr
*)dst
, rt
);
2053 error
= ifp
->if_output(ifp
, m
, (struct sockaddr
*)dst
, rt
);
2063 nd6_need_cache(struct ifnet
*ifp
)
2066 * XXX: we currently do not make neighbor cache on any interface
2067 * other than Ethernet and GIF.
2070 * - unidirectional tunnels needs no ND
2072 switch (ifp
->if_type
) {
2078 #ifdef IFT_IEEE80211
2084 case IFT_GIF
: /* XXX need more cases? */
2092 nd6_storelladdr(struct ifnet
*ifp
, struct rtentry
*rt0
, struct mbuf
*m
,
2093 struct sockaddr
*dst
, u_char
*desten
)
2095 struct sockaddr_dl
*sdl
;
2099 if (m
->m_flags
& M_MCAST
) {
2100 switch (ifp
->if_type
) {
2105 #ifdef IFT_IEEE80211
2108 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst
)->sin6_addr
,
2112 bcopy(ifp
->if_broadcastaddr
, desten
, ifp
->if_addrlen
);
2120 /* this could happen, if we could not allocate memory */
2124 if (rt_llroute(dst
, rt0
, &rt
) != 0) {
2128 if (rt
->rt_gateway
->sa_family
!= AF_LINK
) {
2129 kprintf("nd6_storelladdr: something odd happens\n");
2133 sdl
= SDL(rt
->rt_gateway
);
2134 if (sdl
->sdl_alen
== 0) {
2135 /* this should be impossible, but we bark here for debugging */
2136 kprintf("nd6_storelladdr: sdl_alen == 0\n");
2141 bcopy(LLADDR(sdl
), desten
, sdl
->sdl_alen
);
2145 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS
);
2146 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS
);
2148 SYSCTL_DECL(_net_inet6_icmp6
);
2150 SYSCTL_NODE(_net_inet6_icmp6
, ICMPV6CTL_ND6_DRLIST
, nd6_drlist
,
2151 CTLFLAG_RD
, nd6_sysctl_drlist
, "List default routers");
2152 SYSCTL_NODE(_net_inet6_icmp6
, ICMPV6CTL_ND6_PRLIST
, nd6_prlist
,
2153 CTLFLAG_RD
, nd6_sysctl_prlist
, "List prefixes");
2156 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS
)
2160 struct in6_defrouter
*d
, *de
;
2161 struct nd_defrouter
*dr
;
2167 for (dr
= TAILQ_FIRST(&nd_defrouter
); dr
;
2168 dr
= TAILQ_NEXT(dr
, dr_entry
)) {
2169 d
= (struct in6_defrouter
*)buf
;
2170 de
= (struct in6_defrouter
*)(buf
+ sizeof(buf
));
2173 bzero(d
, sizeof(*d
));
2174 d
->rtaddr
.sin6_family
= AF_INET6
;
2175 d
->rtaddr
.sin6_len
= sizeof(d
->rtaddr
);
2176 if (in6_recoverscope(&d
->rtaddr
, &dr
->rtaddr
,
2180 "default router list (%s)\n",
2181 ip6_sprintf(&dr
->rtaddr
));
2182 d
->flags
= dr
->flags
;
2183 d
->rtlifetime
= dr
->rtlifetime
;
2184 d
->expire
= dr
->expire
;
2185 d
->if_index
= dr
->ifp
->if_index
;
2187 panic("buffer too short");
2189 error
= SYSCTL_OUT(req
, buf
, sizeof(*d
));
2197 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS
)
2201 struct in6_prefix
*p
, *pe
;
2202 struct nd_prefix
*pr
;
2208 for (pr
= nd_prefix
.lh_first
; pr
; pr
= pr
->ndpr_next
) {
2211 struct sockaddr_in6
*sin6
, *s6
;
2212 struct nd_pfxrouter
*pfr
;
2214 p
= (struct in6_prefix
*)buf
;
2215 pe
= (struct in6_prefix
*)(buf
+ sizeof(buf
));
2218 bzero(p
, sizeof(*p
));
2219 sin6
= (struct sockaddr_in6
*)(p
+ 1);
2221 p
->prefix
= pr
->ndpr_prefix
;
2222 if (in6_recoverscope(&p
->prefix
,
2223 &p
->prefix
.sin6_addr
, pr
->ndpr_ifp
) != 0)
2225 "scope error in prefix list (%s)\n",
2226 ip6_sprintf(&p
->prefix
.sin6_addr
));
2227 p
->raflags
= pr
->ndpr_raf
;
2228 p
->prefixlen
= pr
->ndpr_plen
;
2229 p
->vltime
= pr
->ndpr_vltime
;
2230 p
->pltime
= pr
->ndpr_pltime
;
2231 p
->if_index
= pr
->ndpr_ifp
->if_index
;
2232 p
->expire
= pr
->ndpr_expire
;
2233 p
->refcnt
= pr
->ndpr_refcnt
;
2234 p
->flags
= pr
->ndpr_stateflags
;
2235 p
->origin
= PR_ORIG_RA
;
2237 for (pfr
= pr
->ndpr_advrtrs
.lh_first
; pfr
;
2238 pfr
= pfr
->pfr_next
) {
2239 if ((void *)&sin6
[advrtrs
+ 1] > (void *)pe
) {
2243 s6
= &sin6
[advrtrs
];
2244 bzero(s6
, sizeof(*s6
));
2245 s6
->sin6_family
= AF_INET6
;
2246 s6
->sin6_len
= sizeof(*sin6
);
2247 if (in6_recoverscope(s6
, &pfr
->router
->rtaddr
,
2248 pfr
->router
->ifp
) != 0)
2251 "prefix list (%s)\n",
2252 ip6_sprintf(&pfr
->router
->rtaddr
));
2255 p
->advrtrs
= advrtrs
;
2257 panic("buffer too short");
2260 advance
= sizeof(*p
) + sizeof(*sin6
) * advrtrs
;
2261 error
= SYSCTL_OUT(req
, buf
, advance
);