2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
35 * Copyright (c) 1982, 1986, 1991, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
69 #include "opt_inet6.h"
70 #include "opt_mpath.h"
72 #include <sys/param.h>
73 #include <sys/systm.h>
75 #include <sys/malloc.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sockio.h>
82 #include <sys/sysctl.h>
83 #include <sys/errno.h>
86 #include <sys/kernel.h>
87 #include <sys/rmlock.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/route.h>
94 #include <net/if_llatbl.h>
96 #include <net/radix_mpath.h>
99 #include <netinet/in.h>
100 #include <netinet/in_var.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/ip_var.h>
105 #include <netinet/udp.h>
106 #include <netinet/udp_var.h>
108 #include <netinet6/in6_var.h>
109 #include <netinet/ip6.h>
110 #include <netinet6/in6_fib.h>
111 #include <netinet6/in6_pcb.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet6/scope6_var.h>
114 #include <netinet6/nd6.h>
116 static struct mtx addrsel_lock;
117 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
118 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
119 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
120 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
122 static struct sx addrsel_sxlock;
123 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
124 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
125 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
126 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
127 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
129 #define ADDR_LABEL_NOTAPP (-1)
130 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
131 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
133 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
135 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
136 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
137 struct rtentry **, int, u_int);
138 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
139 struct ip6_moptions *, struct ifnet **,
140 struct ifnet *, u_int);
141 static int in6_selectsrc(uint32_t, struct sockaddr_in6 *,
142 struct ip6_pktopts *, struct inpcb *, struct ucred *,
143 struct ifnet **, struct in6_addr *);
145 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
147 static void init_policy_queue(void);
148 static int add_addrsel_policyent(struct in6_addrpolicy *);
149 static int delete_addrsel_policyent(struct in6_addrpolicy *);
150 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
152 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
153 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
156 * Return an IPv6 address, which is the most appropriate for a given
157 * destination and user specified options.
158 * If necessary, this function lookups the routing table and returns
159 * an entry to the caller for later use.
161 #define REPLACE(r) do {\
162 IP6STAT_INC(ip6s_sources_rule[(r)]); \
165 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
166 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)); \
170 #define NEXT(r) do {\
172 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
173 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)); \
175 goto next; /* XXX: we can't use 'continue' here */ \
177 #define BREAK(r) do { \
178 IP6STAT_INC(ip6s_sources_rule[(r)]); \
180 goto out; /* XXX: we can't use 'break' here */ \
184 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
185 struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred,
186 struct ifnet **ifpp, struct in6_addr *srcp)
188 struct rm_priotracker in6_ifa_tracker;
189 struct in6_addr dst, tmp;
190 struct ifnet *ifp = NULL, *oifp = NULL;
191 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
192 struct in6_pktinfo *pi = NULL;
193 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
194 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
198 struct ip6_moptions *mopts;
200 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
202 dst = dstsock->sin6_addr; /* make a copy for local operation */
205 * Save a possibly passed in ifp for in6_selectsrc. Only
206 * neighbor discovery code should use this feature, where
207 * we may know the interface but not the FIB number holding
208 * the connected subnet in case someone deleted it from the
209 * default FIB and we need to check the interface.
217 INP_LOCK_ASSERT(inp);
218 mopts = inp->in6p_moptions;
224 * If the source address is explicitly specified by the caller,
225 * check if the requested source address is indeed a unicast address
226 * assigned to the node, and can be used as the packet's source
227 * address. If everything is okay, use the address as source.
229 if (opts && (pi = opts->ip6po_pktinfo) &&
230 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
231 /* get the outgoing interface */
232 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
238 * determine the appropriate zone id of the source based on
239 * the zone of the destination and the outgoing interface.
240 * If the specified address is ambiguous wrt the scope zone,
241 * the interface must be specified; otherwise, ifa_ifwithaddr()
242 * will fail matching the address.
246 error = in6_setscope(&tmp, ifp, &odstzone);
250 if (cred != NULL && (error = prison_local_ip6(cred,
251 &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
255 * If IPV6_BINDANY socket option is set, we allow to specify
256 * non local addresses as source address in IPV6_PKTINFO
259 if ((inp->inp_flags & INP_BINDANY) == 0) {
260 ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */);
261 if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST |
262 IN6_IFF_NOTREADY))) {
264 ifa_free(&ia->ia_ifa);
265 return (EADDRNOTAVAIL);
267 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
268 ifa_free(&ia->ia_ifa);
270 bcopy(&tmp, srcp, sizeof(*srcp));
271 pi->ipi6_addr = tmp; /* XXX: this overrides pi */
278 * Otherwise, if the socket has already bound the source, just use it.
280 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
282 (error = prison_local_ip6(cred, &inp->in6p_laddr,
283 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
285 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
290 * Bypass source address selection and use the primary jail IP
293 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
297 * If the address is not specified, choose the best one based on
298 * the outgoing interface and the destination address.
300 /* get the outgoing interface */
301 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
302 (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0)
306 if (ifp == NULL) /* this should not happen */
307 panic("in6_selectsrc: NULL ifp");
309 error = in6_setscope(&dst, ifp, &odstzone);
314 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
315 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
316 int new_scope = -1, new_matchlen = -1;
317 struct in6_addrpolicy *new_policy = NULL;
318 u_int32_t srczone, osrczone, dstzone;
320 struct ifnet *ifp1 = ia->ia_ifp;
323 * We'll never take an address that breaks the scope zone
324 * of the destination. We also skip an address if its zone
325 * does not contain the outgoing interface.
326 * XXX: we should probably use sin6_scope_id here.
328 if (in6_setscope(&dst, ifp1, &dstzone) ||
329 odstzone != dstzone) {
332 src = ia->ia_addr.sin6_addr;
333 if (in6_setscope(&src, ifp, &osrczone) ||
334 in6_setscope(&src, ifp1, &srczone) ||
335 osrczone != srczone) {
339 /* avoid unusable addresses */
341 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
344 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
347 /* If jailed only take addresses of the jail into account. */
349 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
352 /* Rule 1: Prefer same address */
353 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
355 BREAK(1); /* there should be no better candidate */
361 /* Rule 2: Prefer appropriate scope */
363 dst_scope = in6_addrscope(&dst);
364 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
365 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
366 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
369 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
370 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
376 * Rule 3: Avoid deprecated addresses. Note that the case of
377 * !ip6_use_deprecated is already rejected above.
379 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
381 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
384 /* Rule 4: Prefer home addresses */
386 * XXX: This is a TODO. We should probably merge the MIP6
390 /* Rule 5: Prefer outgoing interface */
391 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
392 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
394 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
399 * Rule 6: Prefer matching label
400 * Note that best_policy should be non-NULL here.
402 if (dst_policy == NULL)
403 dst_policy = lookup_addrsel_policy(dstsock);
404 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
405 new_policy = lookup_addrsel_policy(&ia->ia_addr);
406 if (dst_policy->label == best_policy->label &&
407 dst_policy->label != new_policy->label)
409 if (dst_policy->label != best_policy->label &&
410 dst_policy->label == new_policy->label)
415 * Rule 7: Prefer public addresses.
416 * We allow users to reverse the logic by configuring
417 * a sysctl variable, so that privacy conscious users can
418 * always prefer temporary addresses.
421 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
422 prefer_tempaddr = V_ip6_prefer_tempaddr;
423 } else if (opts->ip6po_prefer_tempaddr ==
424 IP6PO_TEMPADDR_NOTPREFER) {
428 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
429 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
435 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
436 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
444 * Rule 8: prefer addresses on alive interfaces.
445 * This is a KAME specific rule.
447 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
448 !(ia->ia_ifp->if_flags & IFF_UP))
450 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
451 (ia->ia_ifp->if_flags & IFF_UP))
455 * Rule 9: prefer address with better virtual status.
457 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
459 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
463 * Rule 10: prefer address with `prefer_source' flag.
465 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
466 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
468 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
469 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
473 * Rule 14: Use longest matching prefix.
474 * Note: in the address selection draft, this rule is
475 * documented as "Rule 8". However, since it is also
476 * documented that this rule can be overridden, we assign
477 * a large number so that it is easy to assign smaller numbers
478 * to more preferred rules.
480 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
481 if (best_matchlen < new_matchlen)
483 if (new_matchlen < best_matchlen)
486 /* Rule 15 is reserved. */
489 * Last resort: just keep the current candidate.
490 * Or, do we need more rules?
496 best_scope = (new_scope >= 0 ? new_scope :
497 in6_addrscope(&ia_best->ia_addr.sin6_addr));
498 best_policy = (new_policy ? new_policy :
499 lookup_addrsel_policy(&ia_best->ia_addr));
500 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
501 in6_matchlen(&ia_best->ia_addr.sin6_addr,
511 if ((ia = ia_best) == NULL) {
512 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
513 IP6STAT_INC(ip6s_sources_none);
514 return (EADDRNOTAVAIL);
518 * At this point at least one of the addresses belonged to the jail
519 * but it could still be, that we want to further restrict it, e.g.
520 * theoratically IN6_IS_ADDR_LOOPBACK.
521 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
522 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
523 * let all others previously selected pass.
524 * Use tmp to not change ::1 on lo0 to the primary jail address.
526 tmp = ia->ia_addr.sin6_addr;
527 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
528 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
529 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
530 IP6STAT_INC(ip6s_sources_none);
531 return (EADDRNOTAVAIL);
537 bcopy(&tmp, srcp, sizeof(*srcp));
538 if (ia->ia_ifp == ifp)
539 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
541 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
542 if (dst_scope == best_scope)
543 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
545 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
546 if (IFA6_IS_DEPRECATED(ia))
547 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
548 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
553 * Select source address based on @inp, @dstsock and @opts.
554 * Stores selected address to @srcp. If @scope_ambiguous is set,
555 * embed scope from selected outgoing interface. If @hlim pointer
556 * is provided, stores calculated hop limit there.
557 * Returns 0 on success.
560 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
561 struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
562 struct in6_addr *srcp, int *hlim)
564 struct ifnet *retifp;
568 fibnum = inp->inp_inc.inc_fibnum;
571 error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
576 *hlim = in6_selecthlim(inp, retifp);
578 if (retifp == NULL || scope_ambiguous == 0)
582 * Application should provide a proper zone ID or the use of
583 * default zone IDs should be enabled. Unfortunately, some
584 * applications do not behave as it should, so we need a
585 * workaround. Even if an appropriate ID is not determined
586 * (when it's required), if we can determine the outgoing
587 * interface. determine the zone ID based on the interface.
589 error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
595 * Select source address based on @fibnum, @dst and @scopeid.
596 * Stores selected address to @srcp.
597 * Returns 0 on success.
599 * Used by non-socket based consumers (ND code mostly)
602 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
603 uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
606 struct ifnet *retifp;
607 struct sockaddr_in6 dst_sa;
611 bzero(&dst_sa, sizeof(dst_sa));
612 dst_sa.sin6_family = AF_INET6;
613 dst_sa.sin6_len = sizeof(dst_sa);
614 dst_sa.sin6_addr = *dst;
615 dst_sa.sin6_scope_id = scopeid;
616 sa6_embedscope(&dst_sa, 0);
618 error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
620 *hlim = in6_selecthlim(NULL, retifp);
626 * clone - meaningful only for bsdi and freebsd
629 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
630 struct ip6_moptions *mopts, struct route_in6 *ro,
631 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
634 struct ifnet *ifp = NULL;
635 struct rtentry *rt = NULL;
636 struct sockaddr_in6 *sin6_next;
637 struct in6_pktinfo *pi = NULL;
638 struct in6_addr *dst = &dstsock->sin6_addr;
641 char ip6buf[INET6_ADDRSTRLEN];
643 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
644 dstsock->sin6_addr.s6_addr32[1] == 0 &&
645 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
646 printf("in6_selectroute: strange destination %s\n",
647 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
649 printf("in6_selectroute: destination = %s%%%d\n",
650 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
651 dstsock->sin6_scope_id); /* for debug */
655 /* If the caller specify the outgoing interface explicitly, use it. */
656 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
657 /* XXX boundary check is assumed to be already done. */
658 ifp = ifnet_byindex(pi->ipi6_ifindex);
660 (norouteok || retrt == NULL ||
661 IN6_IS_ADDR_MULTICAST(dst))) {
663 * we do not have to check or get the route for
671 * If the destination address is a multicast address and the outgoing
672 * interface for the address is specified by the caller, use it.
674 if (IN6_IS_ADDR_MULTICAST(dst) &&
675 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
676 goto done; /* we do not need a route for multicast. */
679 * If destination address is LLA or link- or node-local multicast,
680 * use it's embedded scope zone id to determine outgoing interface.
682 if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
683 IN6_IS_ADDR_MC_NODELOCAL(dst)) {
684 zoneid = ntohs(in6_getscope(dst));
686 ifp = in6_getlinkifnet(zoneid);
693 * If the next hop address for the packet is specified by the caller,
694 * use it as the gateway.
696 if (opts && opts->ip6po_nexthop) {
697 struct route_in6 *ron;
699 sin6_next = satosin6(opts->ip6po_nexthop);
700 if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
702 * Next hop is LLA, thus it should be neighbor.
703 * Determine outgoing interface by zone index.
705 zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
707 ifp = in6_getlinkifnet(zoneid);
711 ron = &opts->ip6po_nextroute;
712 /* Use a cached route if it exists and is valid. */
713 if (ron->ro_rt != NULL && (
714 (ron->ro_rt->rt_flags & RTF_UP) == 0 ||
715 ron->ro_dst.sin6_family != AF_INET6 ||
716 !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
717 &sin6_next->sin6_addr)))
719 if (ron->ro_rt == NULL) {
720 ron->ro_dst = *sin6_next;
721 in6_rtalloc(ron, fibnum); /* multi path case? */
724 * The node identified by that address must be a
725 * neighbor of the sending host.
727 if (ron->ro_rt == NULL ||
728 (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0)
729 error = EHOSTUNREACH;
734 * Use a cached route if it exists and is valid, else try to allocate
735 * a new one. Note that we should check the address family of the
736 * cached destination, in case of sharing the cache with IPv4.
740 (!(ro->ro_rt->rt_flags & RTF_UP) ||
741 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
742 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
745 ro->ro_rt = (struct rtentry *)NULL;
747 if (ro->ro_rt == (struct rtentry *)NULL) {
748 struct sockaddr_in6 *sa6;
750 /* No route yet, so try to acquire one */
751 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
752 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
754 sa6->sin6_scope_id = 0;
757 rtalloc_mpath_fib((struct route *)ro,
758 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
760 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
761 &ro->ro_dst, 0, 0UL, fibnum);
763 RT_UNLOCK(ro->ro_rt);
768 * do not care about the result if we have the nexthop
769 * explicitly specified.
771 if (opts && opts->ip6po_nexthop)
775 ifp = ro->ro_rt->rt_ifp;
777 if (ifp == NULL) { /* can this really happen? */
782 if (ro->ro_rt == NULL)
783 error = EHOSTUNREACH;
787 * Check if the outgoing interface conflicts with
788 * the interface specified by ipi6_ifindex (if specified).
789 * Note that loopback interface is always okay.
790 * (this may happen when we are sending a packet to one of
791 * our own addresses.)
793 if (ifp && opts && opts->ip6po_pktinfo &&
794 opts->ip6po_pktinfo->ipi6_ifindex) {
795 if (!(ifp->if_flags & IFF_LOOPBACK) &&
797 opts->ip6po_pktinfo->ipi6_ifindex) {
798 error = EHOSTUNREACH;
805 if (ifp == NULL && rt == NULL) {
807 * This can happen if the caller did not pass a cached route
808 * nor any other hints. We treat this case an error.
810 error = EHOSTUNREACH;
812 if (error == EHOSTUNREACH)
813 IP6STAT_INC(ip6s_noroute);
815 if (retifp != NULL) {
819 * Adjust the "outgoing" interface. If we're going to loop
820 * the packet back to ourselves, the ifp would be the loopback
821 * interface. However, we'd rather know the interface associated
822 * to the destination address (which should probably be one of
823 * our own addresses.)
826 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
827 (rt->rt_gateway->sa_family == AF_LINK))
829 ifnet_byindex(((struct sockaddr_dl *)
830 rt->rt_gateway)->sdl_index);
835 *retrt = rt; /* rt may be NULL */
841 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
842 struct ip6_moptions *mopts, struct ifnet **retifp,
843 struct ifnet *oifp, u_int fibnum)
846 struct route_in6 sro;
847 struct rtentry *rt = NULL;
850 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
852 bzero(&sro, sizeof(sro));
855 error = selectroute(dstsock, opts, mopts, &sro, retifp, &rt, 1, fibnum);
858 rt_flags = rt->rt_flags;
859 if (rt && rt == sro.ro_rt)
863 /* Help ND. See oifp comment in in6_selectsrc(). */
864 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
872 * do not use a rejected or black hole route.
873 * XXX: this check should be done in the L2 output routine.
874 * However, if we skipped this check here, we'd see the following
876 * - install a rejected route for a scoped address prefix
878 * - send a packet to a destination that matches the scoped prefix,
879 * with ambiguity about the scope zone.
880 * - pick the outgoing interface from the route, and disambiguate the
881 * scope zone with the interface.
882 * - ip6_output() would try to get another route with the "new"
883 * destination, which may be valid.
884 * - we'd see no error on output.
885 * Although this may not be very harmful, it should still be confusing.
886 * We thus reject the case here.
889 if (rt_flags & (RTF_REJECT | RTF_BLACKHOLE)) {
890 error = (rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
898 * Public wrapper function to selectroute().
900 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
901 * in6_selectroute_fib() function is only there for backward compat on stable.
904 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
905 struct ip6_moptions *mopts, struct route_in6 *ro,
906 struct ifnet **retifp, struct rtentry **retrt)
909 return (selectroute(dstsock, opts, mopts, ro, retifp,
910 retrt, 0, RT_DEFAULT_FIB));
915 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
916 struct ip6_moptions *mopts, struct route_in6 *ro,
917 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
920 return (selectroute(dstsock, opts, mopts, ro, retifp,
926 * Default hop limit selection. The precedence is as follows:
927 * 1. Hoplimit value specified via ioctl.
928 * 2. (If the outgoing interface is detected) the current
929 * hop limit of the interface specified by router advertisement.
930 * 3. The system default hoplimit.
933 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
936 if (in6p && in6p->in6p_hops >= 0)
937 return (in6p->in6p_hops);
939 return (ND_IFINFO(ifp)->chlim);
940 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
941 struct nhop6_basic nh6;
943 uint32_t fibnum, scopeid;
946 fibnum = in6p->inp_inc.inc_fibnum;
947 in6_splitscope(&in6p->in6p_faddr, &dst, &scopeid);
948 if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){
949 hlim = ND_IFINFO(nh6.nh_ifp)->chlim;
953 return (V_ip6_defhlim);
957 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
958 * share this function by all *bsd*...
961 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
963 struct socket *so = inp->inp_socket;
965 int error, lookupflags = 0;
967 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
970 INP_WLOCK_ASSERT(inp);
971 INP_HASH_WLOCK_ASSERT(pcbinfo);
973 error = prison_local_ip6(cred, laddr,
974 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
978 /* XXX: this is redundant when called from in6_pcbbind */
979 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
980 lookupflags = INPLOOKUP_WILDCARD;
982 inp->inp_flags |= INP_ANONPORT;
984 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
988 inp->inp_lport = lport;
989 if (in_pcbinshash(inp) != 0) {
990 inp->in6p_laddr = in6addr_any;
999 addrsel_policy_init(void)
1002 init_policy_queue();
1004 /* initialize the "last resort" policy */
1005 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
1006 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
1008 if (!IS_DEFAULT_VNET(curvnet))
1011 ADDRSEL_LOCK_INIT();
1012 ADDRSEL_SXLOCK_INIT();
1015 static struct in6_addrpolicy *
1016 lookup_addrsel_policy(struct sockaddr_in6 *key)
1018 struct in6_addrpolicy *match = NULL;
1021 match = match_addrsel_policy(key);
1024 match = &V_defaultaddrpolicy;
1033 * Subroutines to manage the address selection policy table via sysctl.
1036 struct sysctl_req *w_req;
1039 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1040 SYSCTL_DECL(_net_inet6_ip6);
1041 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1042 CTLFLAG_RD, in6_src_sysctl, "");
1045 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1052 bzero(&w, sizeof(w));
1055 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1059 in6_src_ioctl(u_long cmd, caddr_t data)
1061 struct in6_addrpolicy ent0;
1063 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1064 return (EOPNOTSUPP); /* check for safety */
1066 ent0 = *(struct in6_addrpolicy *)data;
1068 if (ent0.label == ADDR_LABEL_NOTAPP)
1070 /* check if the prefix mask is consecutive. */
1071 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1073 /* clear trailing garbages (if any) of the prefix address. */
1074 IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1078 case SIOCAADDRCTL_POLICY:
1079 return (add_addrsel_policyent(&ent0));
1080 case SIOCDADDRCTL_POLICY:
1081 return (delete_addrsel_policyent(&ent0));
1084 return (0); /* XXX: compromise compilers */
1088 * The followings are implementation of the policy table using a
1089 * simple tail queue.
1090 * XXX such details should be hidden.
1091 * XXX implementation using binary tree should be more efficient.
1093 struct addrsel_policyent {
1094 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1095 struct in6_addrpolicy ape_policy;
1098 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1100 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1101 #define V_addrsel_policytab VNET(addrsel_policytab)
1104 init_policy_queue(void)
1107 TAILQ_INIT(&V_addrsel_policytab);
1111 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1113 struct addrsel_policyent *new, *pol;
1115 new = malloc(sizeof(*new), M_IFADDR,
1120 /* duplication check */
1121 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1122 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1123 &pol->ape_policy.addr.sin6_addr) &&
1124 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1125 &pol->ape_policy.addrmask.sin6_addr)) {
1128 free(new, M_IFADDR);
1129 return (EEXIST); /* or override it? */
1133 bzero(new, sizeof(*new));
1135 /* XXX: should validate entry */
1136 new->ape_policy = *newpolicy;
1138 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1146 delete_addrsel_policyent(struct in6_addrpolicy *key)
1148 struct addrsel_policyent *pol;
1153 /* search for the entry in the table */
1154 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1155 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1156 &pol->ape_policy.addr.sin6_addr) &&
1157 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1158 &pol->ape_policy.addrmask.sin6_addr)) {
1168 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1171 free(pol, M_IFADDR);
1177 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1179 struct addrsel_policyent *pol;
1183 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1184 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1194 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1197 struct walkarg *w = arg;
1199 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1204 static struct in6_addrpolicy *
1205 match_addrsel_policy(struct sockaddr_in6 *key)
1207 struct addrsel_policyent *pent;
1208 struct in6_addrpolicy *bestpol = NULL, *pol;
1209 int matchlen, bestmatchlen = -1;
1210 u_char *mp, *ep, *k, *p, m;
1212 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1215 pol = &pent->ape_policy;
1216 mp = (u_char *)&pol->addrmask.sin6_addr;
1217 ep = mp + 16; /* XXX: scope field? */
1218 k = (u_char *)&key->sin6_addr;
1219 p = (u_char *)&pol->addr.sin6_addr;
1220 for (; mp < ep && *mp; mp++, k++, p++) {
1223 goto next; /* not match */
1224 if (m == 0xff) /* short cut for a typical case */
1234 /* matched. check if this is better than the current best. */
1235 if (bestpol == NULL ||
1236 matchlen > bestmatchlen) {
1238 bestmatchlen = matchlen;