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 VNET_DEFINE_STATIC(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)]); \
164 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
165 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)); \
169 #define NEXT(r) do {\
171 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
172 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)); \
174 goto next; /* XXX: we can't use 'continue' here */ \
176 #define BREAK(r) do { \
177 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);
311 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
312 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
313 int new_scope = -1, new_matchlen = -1;
314 struct in6_addrpolicy *new_policy = NULL;
315 u_int32_t srczone, osrczone, dstzone;
317 struct ifnet *ifp1 = ia->ia_ifp;
320 * We'll never take an address that breaks the scope zone
321 * of the destination. We also skip an address if its zone
322 * does not contain the outgoing interface.
323 * XXX: we should probably use sin6_scope_id here.
325 if (in6_setscope(&dst, ifp1, &dstzone) ||
326 odstzone != dstzone) {
329 src = ia->ia_addr.sin6_addr;
330 if (in6_setscope(&src, ifp, &osrczone) ||
331 in6_setscope(&src, ifp1, &srczone) ||
332 osrczone != srczone) {
336 /* avoid unusable addresses */
338 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
341 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
344 /* If jailed only take addresses of the jail into account. */
346 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
349 /* Rule 1: Prefer same address */
350 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
352 BREAK(1); /* there should be no better candidate */
358 /* Rule 2: Prefer appropriate scope */
360 dst_scope = in6_addrscope(&dst);
361 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
362 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
363 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
366 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
367 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
373 * Rule 3: Avoid deprecated addresses. Note that the case of
374 * !ip6_use_deprecated is already rejected above.
376 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
378 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
381 /* Rule 4: Prefer home addresses */
383 * XXX: This is a TODO. We should probably merge the MIP6
387 /* Rule 5: Prefer outgoing interface */
388 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
389 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
391 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
396 * Rule 6: Prefer matching label
397 * Note that best_policy should be non-NULL here.
399 if (dst_policy == NULL)
400 dst_policy = lookup_addrsel_policy(dstsock);
401 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
402 new_policy = lookup_addrsel_policy(&ia->ia_addr);
403 if (dst_policy->label == best_policy->label &&
404 dst_policy->label != new_policy->label)
406 if (dst_policy->label != best_policy->label &&
407 dst_policy->label == new_policy->label)
412 * Rule 7: Prefer public addresses.
413 * We allow users to reverse the logic by configuring
414 * a sysctl variable, so that privacy conscious users can
415 * always prefer temporary addresses.
418 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
419 prefer_tempaddr = V_ip6_prefer_tempaddr;
420 } else if (opts->ip6po_prefer_tempaddr ==
421 IP6PO_TEMPADDR_NOTPREFER) {
425 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
426 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
432 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
433 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
441 * Rule 8: prefer addresses on alive interfaces.
442 * This is a KAME specific rule.
444 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
445 !(ia->ia_ifp->if_flags & IFF_UP))
447 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
448 (ia->ia_ifp->if_flags & IFF_UP))
452 * Rule 9: prefer address with better virtual status.
454 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
456 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
460 * Rule 10: prefer address with `prefer_source' flag.
462 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
463 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
465 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
466 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
470 * Rule 14: Use longest matching prefix.
471 * Note: in the address selection draft, this rule is
472 * documented as "Rule 8". However, since it is also
473 * documented that this rule can be overridden, we assign
474 * a large number so that it is easy to assign smaller numbers
475 * to more preferred rules.
477 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
478 if (best_matchlen < new_matchlen)
480 if (new_matchlen < best_matchlen)
483 /* Rule 15 is reserved. */
486 * Last resort: just keep the current candidate.
487 * Or, do we need more rules?
493 best_scope = (new_scope >= 0 ? new_scope :
494 in6_addrscope(&ia_best->ia_addr.sin6_addr));
495 best_policy = (new_policy ? new_policy :
496 lookup_addrsel_policy(&ia_best->ia_addr));
497 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
498 in6_matchlen(&ia_best->ia_addr.sin6_addr,
508 if ((ia = ia_best) == NULL) {
509 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
510 IP6STAT_INC(ip6s_sources_none);
511 return (EADDRNOTAVAIL);
515 * At this point at least one of the addresses belonged to the jail
516 * but it could still be, that we want to further restrict it, e.g.
517 * theoratically IN6_IS_ADDR_LOOPBACK.
518 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
519 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
520 * let all others previously selected pass.
521 * Use tmp to not change ::1 on lo0 to the primary jail address.
523 tmp = ia->ia_addr.sin6_addr;
524 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
525 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
526 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
527 IP6STAT_INC(ip6s_sources_none);
528 return (EADDRNOTAVAIL);
534 bcopy(&tmp, srcp, sizeof(*srcp));
535 if (ia->ia_ifp == ifp)
536 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
538 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
539 if (dst_scope == best_scope)
540 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
542 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
543 if (IFA6_IS_DEPRECATED(ia))
544 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
545 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
550 * Select source address based on @inp, @dstsock and @opts.
551 * Stores selected address to @srcp. If @scope_ambiguous is set,
552 * embed scope from selected outgoing interface. If @hlim pointer
553 * is provided, stores calculated hop limit there.
554 * Returns 0 on success.
557 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
558 struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
559 struct in6_addr *srcp, int *hlim)
561 struct ifnet *retifp;
565 fibnum = inp->inp_inc.inc_fibnum;
568 error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
573 *hlim = in6_selecthlim(inp, retifp);
575 if (retifp == NULL || scope_ambiguous == 0)
579 * Application should provide a proper zone ID or the use of
580 * default zone IDs should be enabled. Unfortunately, some
581 * applications do not behave as it should, so we need a
582 * workaround. Even if an appropriate ID is not determined
583 * (when it's required), if we can determine the outgoing
584 * interface. determine the zone ID based on the interface.
586 error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
592 * Select source address based on @fibnum, @dst and @scopeid.
593 * Stores selected address to @srcp.
594 * Returns 0 on success.
596 * Used by non-socket based consumers (ND code mostly)
599 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
600 uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
603 struct ifnet *retifp;
604 struct sockaddr_in6 dst_sa;
608 bzero(&dst_sa, sizeof(dst_sa));
609 dst_sa.sin6_family = AF_INET6;
610 dst_sa.sin6_len = sizeof(dst_sa);
611 dst_sa.sin6_addr = *dst;
612 dst_sa.sin6_scope_id = scopeid;
613 sa6_embedscope(&dst_sa, 0);
615 error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
617 *hlim = in6_selecthlim(NULL, retifp);
623 * clone - meaningful only for bsdi and freebsd
626 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
627 struct ip6_moptions *mopts, struct route_in6 *ro,
628 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
631 struct ifnet *ifp = NULL;
632 struct rtentry *rt = NULL;
633 struct sockaddr_in6 *sin6_next;
634 struct in6_pktinfo *pi = NULL;
635 struct in6_addr *dst = &dstsock->sin6_addr;
638 char ip6buf[INET6_ADDRSTRLEN];
640 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
641 dstsock->sin6_addr.s6_addr32[1] == 0 &&
642 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
643 printf("in6_selectroute: strange destination %s\n",
644 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
646 printf("in6_selectroute: destination = %s%%%d\n",
647 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
648 dstsock->sin6_scope_id); /* for debug */
652 /* If the caller specify the outgoing interface explicitly, use it. */
653 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
654 /* XXX boundary check is assumed to be already done. */
655 ifp = ifnet_byindex(pi->ipi6_ifindex);
657 (norouteok || retrt == NULL ||
658 IN6_IS_ADDR_MULTICAST(dst))) {
660 * we do not have to check or get the route for
668 * If the destination address is a multicast address and the outgoing
669 * interface for the address is specified by the caller, use it.
671 if (IN6_IS_ADDR_MULTICAST(dst) &&
672 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
673 goto done; /* we do not need a route for multicast. */
676 * If destination address is LLA or link- or node-local multicast,
677 * use it's embedded scope zone id to determine outgoing interface.
679 if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
680 IN6_IS_ADDR_MC_NODELOCAL(dst)) {
681 zoneid = ntohs(in6_getscope(dst));
683 ifp = in6_getlinkifnet(zoneid);
690 * If the next hop address for the packet is specified by the caller,
691 * use it as the gateway.
693 if (opts && opts->ip6po_nexthop) {
694 struct route_in6 *ron;
696 sin6_next = satosin6(opts->ip6po_nexthop);
697 if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
699 * Next hop is LLA, thus it should be neighbor.
700 * Determine outgoing interface by zone index.
702 zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
704 ifp = in6_getlinkifnet(zoneid);
708 ron = &opts->ip6po_nextroute;
709 /* Use a cached route if it exists and is valid. */
710 if (ron->ro_rt != NULL && (
711 (ron->ro_rt->rt_flags & RTF_UP) == 0 ||
712 ron->ro_dst.sin6_family != AF_INET6 ||
713 !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
714 &sin6_next->sin6_addr)))
716 if (ron->ro_rt == NULL) {
717 ron->ro_dst = *sin6_next;
718 in6_rtalloc(ron, fibnum); /* multi path case? */
721 * The node identified by that address must be a
722 * neighbor of the sending host.
724 if (ron->ro_rt == NULL ||
725 (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0)
726 error = EHOSTUNREACH;
731 * Use a cached route if it exists and is valid, else try to allocate
732 * a new one. Note that we should check the address family of the
733 * cached destination, in case of sharing the cache with IPv4.
737 (!(ro->ro_rt->rt_flags & RTF_UP) ||
738 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
739 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
742 ro->ro_rt = (struct rtentry *)NULL;
744 if (ro->ro_rt == (struct rtentry *)NULL) {
745 struct sockaddr_in6 *sa6;
747 /* No route yet, so try to acquire one */
748 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
749 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
751 sa6->sin6_scope_id = 0;
754 rtalloc_mpath_fib((struct route *)ro,
755 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
757 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
758 &ro->ro_dst, 0, 0UL, fibnum);
760 RT_UNLOCK(ro->ro_rt);
765 * do not care about the result if we have the nexthop
766 * explicitly specified.
768 if (opts && opts->ip6po_nexthop)
772 ifp = ro->ro_rt->rt_ifp;
774 if (ifp == NULL) { /* can this really happen? */
779 if (ro->ro_rt == NULL)
780 error = EHOSTUNREACH;
784 * Check if the outgoing interface conflicts with
785 * the interface specified by ipi6_ifindex (if specified).
786 * Note that loopback interface is always okay.
787 * (this may happen when we are sending a packet to one of
788 * our own addresses.)
790 if (ifp && opts && opts->ip6po_pktinfo &&
791 opts->ip6po_pktinfo->ipi6_ifindex) {
792 if (!(ifp->if_flags & IFF_LOOPBACK) &&
794 opts->ip6po_pktinfo->ipi6_ifindex) {
795 error = EHOSTUNREACH;
802 if (ifp == NULL && rt == NULL) {
804 * This can happen if the caller did not pass a cached route
805 * nor any other hints. We treat this case an error.
807 error = EHOSTUNREACH;
809 if (error == EHOSTUNREACH)
810 IP6STAT_INC(ip6s_noroute);
812 if (retifp != NULL) {
816 * Adjust the "outgoing" interface. If we're going to loop
817 * the packet back to ourselves, the ifp would be the loopback
818 * interface. However, we'd rather know the interface associated
819 * to the destination address (which should probably be one of
820 * our own addresses.)
823 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
824 (rt->rt_gateway->sa_family == AF_LINK))
826 ifnet_byindex(((struct sockaddr_dl *)
827 rt->rt_gateway)->sdl_index);
832 *retrt = rt; /* rt may be NULL */
838 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
839 struct ip6_moptions *mopts, struct ifnet **retifp,
840 struct ifnet *oifp, u_int fibnum)
843 struct route_in6 sro;
844 struct rtentry *rt = NULL;
847 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
849 bzero(&sro, sizeof(sro));
852 error = selectroute(dstsock, opts, mopts, &sro, retifp, &rt, 1, fibnum);
855 rt_flags = rt->rt_flags;
856 if (rt && rt == sro.ro_rt)
860 /* Help ND. See oifp comment in in6_selectsrc(). */
861 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
869 * do not use a rejected or black hole route.
870 * XXX: this check should be done in the L2 output routine.
871 * However, if we skipped this check here, we'd see the following
873 * - install a rejected route for a scoped address prefix
875 * - send a packet to a destination that matches the scoped prefix,
876 * with ambiguity about the scope zone.
877 * - pick the outgoing interface from the route, and disambiguate the
878 * scope zone with the interface.
879 * - ip6_output() would try to get another route with the "new"
880 * destination, which may be valid.
881 * - we'd see no error on output.
882 * Although this may not be very harmful, it should still be confusing.
883 * We thus reject the case here.
886 if (rt_flags & (RTF_REJECT | RTF_BLACKHOLE)) {
887 error = (rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
895 * Public wrapper function to selectroute().
897 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
898 * in6_selectroute_fib() function is only there for backward compat on stable.
901 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
902 struct ip6_moptions *mopts, struct route_in6 *ro,
903 struct ifnet **retifp, struct rtentry **retrt)
906 return (selectroute(dstsock, opts, mopts, ro, retifp,
907 retrt, 0, RT_DEFAULT_FIB));
912 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
913 struct ip6_moptions *mopts, struct route_in6 *ro,
914 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
917 return (selectroute(dstsock, opts, mopts, ro, retifp,
923 * Default hop limit selection. The precedence is as follows:
924 * 1. Hoplimit value specified via ioctl.
925 * 2. (If the outgoing interface is detected) the current
926 * hop limit of the interface specified by router advertisement.
927 * 3. The system default hoplimit.
930 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
933 if (in6p && in6p->in6p_hops >= 0)
934 return (in6p->in6p_hops);
936 return (ND_IFINFO(ifp)->chlim);
937 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
938 struct nhop6_basic nh6;
940 uint32_t fibnum, scopeid;
943 fibnum = in6p->inp_inc.inc_fibnum;
944 in6_splitscope(&in6p->in6p_faddr, &dst, &scopeid);
945 if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){
946 hlim = ND_IFINFO(nh6.nh_ifp)->chlim;
950 return (V_ip6_defhlim);
954 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
955 * share this function by all *bsd*...
958 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
960 struct socket *so = inp->inp_socket;
962 int error, lookupflags = 0;
964 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
967 INP_WLOCK_ASSERT(inp);
968 INP_HASH_WLOCK_ASSERT(pcbinfo);
970 error = prison_local_ip6(cred, laddr,
971 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
975 /* XXX: this is redundant when called from in6_pcbbind */
976 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
977 lookupflags = INPLOOKUP_WILDCARD;
979 inp->inp_flags |= INP_ANONPORT;
981 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
985 inp->inp_lport = lport;
986 if (in_pcbinshash(inp) != 0) {
987 inp->in6p_laddr = in6addr_any;
996 addrsel_policy_init(void)
1001 /* initialize the "last resort" policy */
1002 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
1003 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
1005 if (!IS_DEFAULT_VNET(curvnet))
1008 ADDRSEL_LOCK_INIT();
1009 ADDRSEL_SXLOCK_INIT();
1012 static struct in6_addrpolicy *
1013 lookup_addrsel_policy(struct sockaddr_in6 *key)
1015 struct in6_addrpolicy *match = NULL;
1018 match = match_addrsel_policy(key);
1021 match = &V_defaultaddrpolicy;
1030 * Subroutines to manage the address selection policy table via sysctl.
1033 struct sysctl_req *w_req;
1036 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1037 SYSCTL_DECL(_net_inet6_ip6);
1038 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1039 CTLFLAG_RD, in6_src_sysctl, "");
1042 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1049 bzero(&w, sizeof(w));
1052 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1056 in6_src_ioctl(u_long cmd, caddr_t data)
1058 struct in6_addrpolicy ent0;
1060 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1061 return (EOPNOTSUPP); /* check for safety */
1063 ent0 = *(struct in6_addrpolicy *)data;
1065 if (ent0.label == ADDR_LABEL_NOTAPP)
1067 /* check if the prefix mask is consecutive. */
1068 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1070 /* clear trailing garbages (if any) of the prefix address. */
1071 IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1075 case SIOCAADDRCTL_POLICY:
1076 return (add_addrsel_policyent(&ent0));
1077 case SIOCDADDRCTL_POLICY:
1078 return (delete_addrsel_policyent(&ent0));
1081 return (0); /* XXX: compromise compilers */
1085 * The followings are implementation of the policy table using a
1086 * simple tail queue.
1087 * XXX such details should be hidden.
1088 * XXX implementation using binary tree should be more efficient.
1090 struct addrsel_policyent {
1091 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1092 struct in6_addrpolicy ape_policy;
1095 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1097 VNET_DEFINE_STATIC(struct addrsel_policyhead, addrsel_policytab);
1098 #define V_addrsel_policytab VNET(addrsel_policytab)
1101 init_policy_queue(void)
1104 TAILQ_INIT(&V_addrsel_policytab);
1108 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1110 struct addrsel_policyent *new, *pol;
1112 new = malloc(sizeof(*new), M_IFADDR,
1117 /* duplication check */
1118 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1119 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1120 &pol->ape_policy.addr.sin6_addr) &&
1121 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1122 &pol->ape_policy.addrmask.sin6_addr)) {
1125 free(new, M_IFADDR);
1126 return (EEXIST); /* or override it? */
1130 bzero(new, sizeof(*new));
1132 /* XXX: should validate entry */
1133 new->ape_policy = *newpolicy;
1135 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1143 delete_addrsel_policyent(struct in6_addrpolicy *key)
1145 struct addrsel_policyent *pol;
1150 /* search for the entry in the table */
1151 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1152 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1153 &pol->ape_policy.addr.sin6_addr) &&
1154 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1155 &pol->ape_policy.addrmask.sin6_addr)) {
1165 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1168 free(pol, M_IFADDR);
1174 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1176 struct addrsel_policyent *pol;
1180 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1181 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1191 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1194 struct walkarg *w = arg;
1196 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1201 static struct in6_addrpolicy *
1202 match_addrsel_policy(struct sockaddr_in6 *key)
1204 struct addrsel_policyent *pent;
1205 struct in6_addrpolicy *bestpol = NULL, *pol;
1206 int matchlen, bestmatchlen = -1;
1207 u_char *mp, *ep, *k, *p, m;
1209 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1212 pol = &pent->ape_policy;
1213 mp = (u_char *)&pol->addrmask.sin6_addr;
1214 ep = mp + 16; /* XXX: scope field? */
1215 k = (u_char *)&key->sin6_addr;
1216 p = (u_char *)&pol->addr.sin6_addr;
1217 for (; mp < ep && *mp; mp++, k++, p++) {
1220 goto next; /* not match */
1221 if (m == 0xff) /* short cut for a typical case */
1231 /* matched. check if this is better than the current best. */
1232 if (bestpol == NULL ||
1233 matchlen > bestmatchlen) {
1235 bestmatchlen = matchlen;