2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
33 * Copyright (c) 1982, 1986, 1991, 1993
34 * The Regents of the University of California. All rights reserved.
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
67 #include "opt_inet6.h"
68 #include "opt_mpath.h"
70 #include <sys/param.h>
71 #include <sys/systm.h>
73 #include <sys/malloc.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sockio.h>
80 #include <sys/sysctl.h>
81 #include <sys/errno.h>
84 #include <sys/kernel.h>
85 #include <sys/rmlock.h>
89 #include <net/if_var.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/if_llatbl.h>
94 #include <net/radix_mpath.h>
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/ip.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/ip_var.h>
103 #include <netinet/udp.h>
104 #include <netinet/udp_var.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet6/in6_fib.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/nd6.h>
114 static struct mtx addrsel_lock;
115 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
116 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
117 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
118 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
120 static struct sx addrsel_sxlock;
121 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
122 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
123 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
124 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
125 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
127 #define ADDR_LABEL_NOTAPP (-1)
128 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
129 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
131 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
133 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
134 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
135 struct rtentry **, int, u_int);
136 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
137 struct ip6_moptions *, struct ifnet **,
138 struct ifnet *, u_int);
139 static int in6_selectsrc(uint32_t, struct sockaddr_in6 *,
140 struct ip6_pktopts *, struct inpcb *, struct ucred *,
141 struct ifnet **, struct in6_addr *);
143 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
145 static void init_policy_queue(void);
146 static int add_addrsel_policyent(struct in6_addrpolicy *);
147 static int delete_addrsel_policyent(struct in6_addrpolicy *);
148 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
150 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
151 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
154 * Return an IPv6 address, which is the most appropriate for a given
155 * destination and user specified options.
156 * If necessary, this function lookups the routing table and returns
157 * an entry to the caller for later use.
159 #define REPLACE(r) do {\
160 IP6STAT_INC(ip6s_sources_rule[(r)]); \
163 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
164 printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
168 #define NEXT(r) do {\
170 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
171 printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
173 goto next; /* XXX: we can't use 'continue' here */ \
175 #define BREAK(r) do { \
176 IP6STAT_INC(ip6s_sources_rule[(r)]); \
178 goto out; /* XXX: we can't use 'break' here */ \
182 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
183 struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred,
184 struct ifnet **ifpp, struct in6_addr *srcp)
186 struct rm_priotracker in6_ifa_tracker;
187 struct in6_addr dst, tmp;
188 struct ifnet *ifp = NULL, *oifp = NULL;
189 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
190 struct in6_pktinfo *pi = NULL;
191 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
192 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
196 struct ip6_moptions *mopts;
198 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
200 dst = dstsock->sin6_addr; /* make a copy for local operation */
203 * Save a possibly passed in ifp for in6_selectsrc. Only
204 * neighbor discovery code should use this feature, where
205 * we may know the interface but not the FIB number holding
206 * the connected subnet in case someone deleted it from the
207 * default FIB and we need to check the interface.
215 INP_LOCK_ASSERT(inp);
216 mopts = inp->in6p_moptions;
222 * If the source address is explicitly specified by the caller,
223 * check if the requested source address is indeed a unicast address
224 * assigned to the node, and can be used as the packet's source
225 * address. If everything is okay, use the address as source.
227 if (opts && (pi = opts->ip6po_pktinfo) &&
228 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
229 /* get the outgoing interface */
230 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
236 * determine the appropriate zone id of the source based on
237 * the zone of the destination and the outgoing interface.
238 * If the specified address is ambiguous wrt the scope zone,
239 * the interface must be specified; otherwise, ifa_ifwithaddr()
240 * will fail matching the address.
244 error = in6_setscope(&tmp, ifp, &odstzone);
248 if (cred != NULL && (error = prison_local_ip6(cred,
249 &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
253 * If IPV6_BINDANY socket option is set, we allow to specify
254 * non local addresses as source address in IPV6_PKTINFO
257 if ((inp->inp_flags & INP_BINDANY) == 0) {
258 ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */);
259 if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST |
260 IN6_IFF_NOTREADY))) {
262 ifa_free(&ia->ia_ifa);
263 return (EADDRNOTAVAIL);
265 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
266 ifa_free(&ia->ia_ifa);
268 bcopy(&tmp, srcp, sizeof(*srcp));
269 pi->ipi6_addr = tmp; /* XXX: this overrides pi */
276 * Otherwise, if the socket has already bound the source, just use it.
278 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
280 (error = prison_local_ip6(cred, &inp->in6p_laddr,
281 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
283 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
288 * Bypass source address selection and use the primary jail IP
291 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
295 * If the address is not specified, choose the best one based on
296 * the outgoing interface and the destination address.
298 /* get the outgoing interface */
299 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
300 (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0)
304 if (ifp == NULL) /* this should not happen */
305 panic("in6_selectsrc: NULL ifp");
307 error = in6_setscope(&dst, ifp, &odstzone);
312 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
313 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
314 int new_scope = -1, new_matchlen = -1;
315 struct in6_addrpolicy *new_policy = NULL;
316 u_int32_t srczone, osrczone, dstzone;
318 struct ifnet *ifp1 = ia->ia_ifp;
321 * We'll never take an address that breaks the scope zone
322 * of the destination. We also skip an address if its zone
323 * does not contain the outgoing interface.
324 * XXX: we should probably use sin6_scope_id here.
326 if (in6_setscope(&dst, ifp1, &dstzone) ||
327 odstzone != dstzone) {
330 src = ia->ia_addr.sin6_addr;
331 if (in6_setscope(&src, ifp, &osrczone) ||
332 in6_setscope(&src, ifp1, &srczone) ||
333 osrczone != srczone) {
337 /* avoid unusable addresses */
339 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
342 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
345 /* If jailed only take addresses of the jail into account. */
347 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
350 /* Rule 1: Prefer same address */
351 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
353 BREAK(1); /* there should be no better candidate */
359 /* Rule 2: Prefer appropriate scope */
361 dst_scope = in6_addrscope(&dst);
362 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
363 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
364 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
367 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
368 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
374 * Rule 3: Avoid deprecated addresses. Note that the case of
375 * !ip6_use_deprecated is already rejected above.
377 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
379 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
382 /* Rule 4: Prefer home addresses */
384 * XXX: This is a TODO. We should probably merge the MIP6
388 /* Rule 5: Prefer outgoing interface */
389 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
390 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
392 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
397 * Rule 6: Prefer matching label
398 * Note that best_policy should be non-NULL here.
400 if (dst_policy == NULL)
401 dst_policy = lookup_addrsel_policy(dstsock);
402 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
403 new_policy = lookup_addrsel_policy(&ia->ia_addr);
404 if (dst_policy->label == best_policy->label &&
405 dst_policy->label != new_policy->label)
407 if (dst_policy->label != best_policy->label &&
408 dst_policy->label == new_policy->label)
413 * Rule 7: Prefer public addresses.
414 * We allow users to reverse the logic by configuring
415 * a sysctl variable, so that privacy conscious users can
416 * always prefer temporary addresses.
419 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
420 prefer_tempaddr = V_ip6_prefer_tempaddr;
421 } else if (opts->ip6po_prefer_tempaddr ==
422 IP6PO_TEMPADDR_NOTPREFER) {
426 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
427 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
433 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
434 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
442 * Rule 8: prefer addresses on alive interfaces.
443 * This is a KAME specific rule.
445 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
446 !(ia->ia_ifp->if_flags & IFF_UP))
448 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
449 (ia->ia_ifp->if_flags & IFF_UP))
453 * Rule 9: prefer address with better virtual status.
455 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
457 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
461 * Rule 10: prefer address with `prefer_source' flag.
463 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
464 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
466 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
467 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
471 * Rule 14: Use longest matching prefix.
472 * Note: in the address selection draft, this rule is
473 * documented as "Rule 8". However, since it is also
474 * documented that this rule can be overridden, we assign
475 * a large number so that it is easy to assign smaller numbers
476 * to more preferred rules.
478 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
479 if (best_matchlen < new_matchlen)
481 if (new_matchlen < best_matchlen)
484 /* Rule 15 is reserved. */
487 * Last resort: just keep the current candidate.
488 * Or, do we need more rules?
494 best_scope = (new_scope >= 0 ? new_scope :
495 in6_addrscope(&ia_best->ia_addr.sin6_addr));
496 best_policy = (new_policy ? new_policy :
497 lookup_addrsel_policy(&ia_best->ia_addr));
498 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
499 in6_matchlen(&ia_best->ia_addr.sin6_addr,
509 if ((ia = ia_best) == NULL) {
510 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
511 IP6STAT_INC(ip6s_sources_none);
512 return (EADDRNOTAVAIL);
516 * At this point at least one of the addresses belonged to the jail
517 * but it could still be, that we want to further restrict it, e.g.
518 * theoratically IN6_IS_ADDR_LOOPBACK.
519 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
520 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
521 * let all others previously selected pass.
522 * Use tmp to not change ::1 on lo0 to the primary jail address.
524 tmp = ia->ia_addr.sin6_addr;
525 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
526 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
527 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
528 IP6STAT_INC(ip6s_sources_none);
529 return (EADDRNOTAVAIL);
535 bcopy(&tmp, srcp, sizeof(*srcp));
536 if (ia->ia_ifp == ifp)
537 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
539 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
540 if (dst_scope == best_scope)
541 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
543 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
544 if (IFA6_IS_DEPRECATED(ia))
545 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
546 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
551 * Select source address based on @inp, @dstsock and @opts.
552 * Stores selected address to @srcp. If @scope_ambiguous is set,
553 * embed scope from selected outgoing interface. If @hlim pointer
554 * is provided, stores calculated hop limit there.
555 * Returns 0 on success.
558 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
559 struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
560 struct in6_addr *srcp, int *hlim)
562 struct ifnet *retifp;
566 fibnum = inp->inp_inc.inc_fibnum;
569 error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
574 *hlim = in6_selecthlim(inp, retifp);
576 if (retifp == NULL || scope_ambiguous == 0)
580 * Application should provide a proper zone ID or the use of
581 * default zone IDs should be enabled. Unfortunately, some
582 * applications do not behave as it should, so we need a
583 * workaround. Even if an appropriate ID is not determined
584 * (when it's required), if we can determine the outgoing
585 * interface. determine the zone ID based on the interface.
587 error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
593 * Select source address based on @fibnum, @dst and @scopeid.
594 * Stores selected address to @srcp.
595 * Returns 0 on success.
597 * Used by non-socket based consumers (ND code mostly)
600 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
601 uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
604 struct ifnet *retifp;
605 struct sockaddr_in6 dst_sa;
609 bzero(&dst_sa, sizeof(dst_sa));
610 dst_sa.sin6_family = AF_INET6;
611 dst_sa.sin6_len = sizeof(dst_sa);
612 dst_sa.sin6_addr = *dst;
613 dst_sa.sin6_scope_id = scopeid;
614 sa6_embedscope(&dst_sa, 0);
616 error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
618 *hlim = in6_selecthlim(NULL, retifp);
624 * clone - meaningful only for bsdi and freebsd
627 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
628 struct ip6_moptions *mopts, struct route_in6 *ro,
629 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
632 struct ifnet *ifp = NULL;
633 struct rtentry *rt = NULL;
634 struct sockaddr_in6 *sin6_next;
635 struct in6_pktinfo *pi = NULL;
636 struct in6_addr *dst = &dstsock->sin6_addr;
639 char ip6buf[INET6_ADDRSTRLEN];
641 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
642 dstsock->sin6_addr.s6_addr32[1] == 0 &&
643 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
644 printf("in6_selectroute: strange destination %s\n",
645 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
647 printf("in6_selectroute: destination = %s%%%d\n",
648 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
649 dstsock->sin6_scope_id); /* for debug */
653 /* If the caller specify the outgoing interface explicitly, use it. */
654 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
655 /* XXX boundary check is assumed to be already done. */
656 ifp = ifnet_byindex(pi->ipi6_ifindex);
658 (norouteok || retrt == NULL ||
659 IN6_IS_ADDR_MULTICAST(dst))) {
661 * we do not have to check or get the route for
669 * If the destination address is a multicast address and the outgoing
670 * interface for the address is specified by the caller, use it.
672 if (IN6_IS_ADDR_MULTICAST(dst) &&
673 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
674 goto done; /* we do not need a route for multicast. */
677 * If destination address is LLA or link- or node-local multicast,
678 * use it's embedded scope zone id to determine outgoing interface.
680 if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
681 IN6_IS_ADDR_MC_NODELOCAL(dst)) {
682 zoneid = ntohs(in6_getscope(dst));
684 ifp = in6_getlinkifnet(zoneid);
691 * If the next hop address for the packet is specified by the caller,
692 * use it as the gateway.
694 if (opts && opts->ip6po_nexthop) {
695 struct route_in6 *ron;
697 sin6_next = satosin6(opts->ip6po_nexthop);
698 if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
700 * Next hop is LLA, thus it should be neighbor.
701 * Determine outgoing interface by zone index.
703 zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
705 ifp = in6_getlinkifnet(zoneid);
709 ron = &opts->ip6po_nextroute;
710 /* Use a cached route if it exists and is valid. */
711 if (ron->ro_rt != NULL && (
712 (ron->ro_rt->rt_flags & RTF_UP) == 0 ||
713 ron->ro_dst.sin6_family != AF_INET6 ||
714 !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
715 &sin6_next->sin6_addr)))
717 if (ron->ro_rt == NULL) {
718 ron->ro_dst = *sin6_next;
719 in6_rtalloc(ron, fibnum); /* multi path case? */
722 * The node identified by that address must be a
723 * neighbor of the sending host.
725 if (ron->ro_rt == NULL ||
726 (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0)
727 error = EHOSTUNREACH;
732 * Use a cached route if it exists and is valid, else try to allocate
733 * a new one. Note that we should check the address family of the
734 * cached destination, in case of sharing the cache with IPv4.
738 (!(ro->ro_rt->rt_flags & RTF_UP) ||
739 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
740 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
743 ro->ro_rt = (struct rtentry *)NULL;
745 if (ro->ro_rt == (struct rtentry *)NULL) {
746 struct sockaddr_in6 *sa6;
748 /* No route yet, so try to acquire one */
749 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
750 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
752 sa6->sin6_scope_id = 0;
755 rtalloc_mpath_fib((struct route *)ro,
756 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
758 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
759 &ro->ro_dst, 0, 0UL, fibnum);
761 RT_UNLOCK(ro->ro_rt);
766 * do not care about the result if we have the nexthop
767 * explicitly specified.
769 if (opts && opts->ip6po_nexthop)
773 ifp = ro->ro_rt->rt_ifp;
775 if (ifp == NULL) { /* can this really happen? */
780 if (ro->ro_rt == NULL)
781 error = EHOSTUNREACH;
785 * Check if the outgoing interface conflicts with
786 * the interface specified by ipi6_ifindex (if specified).
787 * Note that loopback interface is always okay.
788 * (this may happen when we are sending a packet to one of
789 * our own addresses.)
791 if (ifp && opts && opts->ip6po_pktinfo &&
792 opts->ip6po_pktinfo->ipi6_ifindex) {
793 if (!(ifp->if_flags & IFF_LOOPBACK) &&
795 opts->ip6po_pktinfo->ipi6_ifindex) {
796 error = EHOSTUNREACH;
803 if (ifp == NULL && rt == NULL) {
805 * This can happen if the caller did not pass a cached route
806 * nor any other hints. We treat this case an error.
808 error = EHOSTUNREACH;
810 if (error == EHOSTUNREACH)
811 IP6STAT_INC(ip6s_noroute);
813 if (retifp != NULL) {
817 * Adjust the "outgoing" interface. If we're going to loop
818 * the packet back to ourselves, the ifp would be the loopback
819 * interface. However, we'd rather know the interface associated
820 * to the destination address (which should probably be one of
821 * our own addresses.)
824 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
825 (rt->rt_gateway->sa_family == AF_LINK))
827 ifnet_byindex(((struct sockaddr_dl *)
828 rt->rt_gateway)->sdl_index);
833 *retrt = rt; /* rt may be NULL */
839 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
840 struct ip6_moptions *mopts, struct ifnet **retifp,
841 struct ifnet *oifp, u_int fibnum)
844 struct route_in6 sro;
845 struct rtentry *rt = NULL;
848 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
850 bzero(&sro, sizeof(sro));
853 error = selectroute(dstsock, opts, mopts, &sro, retifp, &rt, 1, fibnum);
856 rt_flags = rt->rt_flags;
857 if (rt && rt == sro.ro_rt)
861 /* Help ND. See oifp comment in in6_selectsrc(). */
862 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
870 * do not use a rejected or black hole route.
871 * XXX: this check should be done in the L2 output routine.
872 * However, if we skipped this check here, we'd see the following
874 * - install a rejected route for a scoped address prefix
876 * - send a packet to a destination that matches the scoped prefix,
877 * with ambiguity about the scope zone.
878 * - pick the outgoing interface from the route, and disambiguate the
879 * scope zone with the interface.
880 * - ip6_output() would try to get another route with the "new"
881 * destination, which may be valid.
882 * - we'd see no error on output.
883 * Although this may not be very harmful, it should still be confusing.
884 * We thus reject the case here.
887 if (rt_flags & (RTF_REJECT | RTF_BLACKHOLE)) {
888 error = (rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
896 * Public wrapper function to selectroute().
898 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
899 * in6_selectroute_fib() function is only there for backward compat on stable.
902 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
903 struct ip6_moptions *mopts, struct route_in6 *ro,
904 struct ifnet **retifp, struct rtentry **retrt)
907 return (selectroute(dstsock, opts, mopts, ro, retifp,
908 retrt, 0, RT_DEFAULT_FIB));
913 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
914 struct ip6_moptions *mopts, struct route_in6 *ro,
915 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
918 return (selectroute(dstsock, opts, mopts, ro, retifp,
924 * Default hop limit selection. The precedence is as follows:
925 * 1. Hoplimit value specified via ioctl.
926 * 2. (If the outgoing interface is detected) the current
927 * hop limit of the interface specified by router advertisement.
928 * 3. The system default hoplimit.
931 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
934 if (in6p && in6p->in6p_hops >= 0)
935 return (in6p->in6p_hops);
937 return (ND_IFINFO(ifp)->chlim);
938 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
939 struct nhop6_basic nh6;
941 uint32_t fibnum, scopeid;
944 fibnum = in6p->inp_inc.inc_fibnum;
945 in6_splitscope(&in6p->in6p_faddr, &dst, &scopeid);
946 if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){
947 hlim = ND_IFINFO(nh6.nh_ifp)->chlim;
951 return (V_ip6_defhlim);
955 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
956 * share this function by all *bsd*...
959 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
961 struct socket *so = inp->inp_socket;
963 int error, lookupflags = 0;
965 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
968 INP_WLOCK_ASSERT(inp);
969 INP_HASH_WLOCK_ASSERT(pcbinfo);
971 error = prison_local_ip6(cred, laddr,
972 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
976 /* XXX: this is redundant when called from in6_pcbbind */
977 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
978 lookupflags = INPLOOKUP_WILDCARD;
980 inp->inp_flags |= INP_ANONPORT;
982 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
986 inp->inp_lport = lport;
987 if (in_pcbinshash(inp) != 0) {
988 inp->in6p_laddr = in6addr_any;
997 addrsel_policy_init(void)
1000 init_policy_queue();
1002 /* initialize the "last resort" policy */
1003 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
1004 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
1006 if (!IS_DEFAULT_VNET(curvnet))
1009 ADDRSEL_LOCK_INIT();
1010 ADDRSEL_SXLOCK_INIT();
1013 static struct in6_addrpolicy *
1014 lookup_addrsel_policy(struct sockaddr_in6 *key)
1016 struct in6_addrpolicy *match = NULL;
1019 match = match_addrsel_policy(key);
1022 match = &V_defaultaddrpolicy;
1031 * Subroutines to manage the address selection policy table via sysctl.
1034 struct sysctl_req *w_req;
1037 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1038 SYSCTL_DECL(_net_inet6_ip6);
1039 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1040 CTLFLAG_RD, in6_src_sysctl, "");
1043 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1050 bzero(&w, sizeof(w));
1053 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1057 in6_src_ioctl(u_long cmd, caddr_t data)
1059 struct in6_addrpolicy ent0;
1061 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1062 return (EOPNOTSUPP); /* check for safety */
1064 ent0 = *(struct in6_addrpolicy *)data;
1066 if (ent0.label == ADDR_LABEL_NOTAPP)
1068 /* check if the prefix mask is consecutive. */
1069 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1071 /* clear trailing garbages (if any) of the prefix address. */
1072 IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1076 case SIOCAADDRCTL_POLICY:
1077 return (add_addrsel_policyent(&ent0));
1078 case SIOCDADDRCTL_POLICY:
1079 return (delete_addrsel_policyent(&ent0));
1082 return (0); /* XXX: compromise compilers */
1086 * The followings are implementation of the policy table using a
1087 * simple tail queue.
1088 * XXX such details should be hidden.
1089 * XXX implementation using binary tree should be more efficient.
1091 struct addrsel_policyent {
1092 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1093 struct in6_addrpolicy ape_policy;
1096 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1098 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1099 #define V_addrsel_policytab VNET(addrsel_policytab)
1102 init_policy_queue(void)
1105 TAILQ_INIT(&V_addrsel_policytab);
1109 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1111 struct addrsel_policyent *new, *pol;
1113 new = malloc(sizeof(*new), M_IFADDR,
1118 /* duplication check */
1119 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1120 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1121 &pol->ape_policy.addr.sin6_addr) &&
1122 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1123 &pol->ape_policy.addrmask.sin6_addr)) {
1126 free(new, M_IFADDR);
1127 return (EEXIST); /* or override it? */
1131 bzero(new, sizeof(*new));
1133 /* XXX: should validate entry */
1134 new->ape_policy = *newpolicy;
1136 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1144 delete_addrsel_policyent(struct in6_addrpolicy *key)
1146 struct addrsel_policyent *pol;
1151 /* search for the entry in the table */
1152 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1153 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1154 &pol->ape_policy.addr.sin6_addr) &&
1155 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1156 &pol->ape_policy.addrmask.sin6_addr)) {
1166 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1169 free(pol, M_IFADDR);
1175 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1177 struct addrsel_policyent *pol;
1181 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1182 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1192 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1195 struct walkarg *w = arg;
1197 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1202 static struct in6_addrpolicy *
1203 match_addrsel_policy(struct sockaddr_in6 *key)
1205 struct addrsel_policyent *pent;
1206 struct in6_addrpolicy *bestpol = NULL, *pol;
1207 int matchlen, bestmatchlen = -1;
1208 u_char *mp, *ep, *k, *p, m;
1210 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1213 pol = &pent->ape_policy;
1214 mp = (u_char *)&pol->addrmask.sin6_addr;
1215 ep = mp + 16; /* XXX: scope field? */
1216 k = (u_char *)&key->sin6_addr;
1217 p = (u_char *)&pol->addr.sin6_addr;
1218 for (; mp < ep && *mp; mp++, k++, p++) {
1221 goto next; /* not match */
1222 if (m == 0xff) /* short cut for a typical case */
1232 /* matched. check if this is better than the current best. */
1233 if (bestpol == NULL ||
1234 matchlen > bestmatchlen) {
1236 bestmatchlen = matchlen;