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_pcb.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/nd6.h>
113 static struct mtx addrsel_lock;
114 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
115 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
116 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
117 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
119 static struct sx addrsel_sxlock;
120 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
121 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
122 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
123 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
124 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
126 #define ADDR_LABEL_NOTAPP (-1)
127 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
128 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
130 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
132 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
133 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
134 struct rtentry **, int, u_int);
135 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
136 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **,
137 struct ifnet *, u_int);
139 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
141 static void init_policy_queue(void);
142 static int add_addrsel_policyent(struct in6_addrpolicy *);
143 static int delete_addrsel_policyent(struct in6_addrpolicy *);
144 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
146 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
147 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
150 * Return an IPv6 address, which is the most appropriate for a given
151 * destination and user specified options.
152 * If necessary, this function lookups the routing table and returns
153 * an entry to the caller for later use.
155 #define REPLACE(r) do {\
156 IP6STAT_INC(ip6s_sources_rule[(r)]); \
159 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
160 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)); \
164 #define NEXT(r) do {\
166 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
167 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)); \
169 goto next; /* XXX: we can't use 'continue' here */ \
171 #define BREAK(r) do { \
172 IP6STAT_INC(ip6s_sources_rule[(r)]); \
174 goto out; /* XXX: we can't use 'break' here */ \
178 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
179 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
180 struct ifnet **ifpp, struct in6_addr *srcp)
182 struct rm_priotracker in6_ifa_tracker;
183 struct in6_addr dst, tmp;
184 struct ifnet *ifp = NULL, *oifp = NULL;
185 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
186 struct in6_pktinfo *pi = NULL;
187 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
188 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
192 struct ip6_moptions *mopts;
194 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
196 dst = dstsock->sin6_addr; /* make a copy for local operation */
199 * Save a possibly passed in ifp for in6_selectsrc. Only
200 * neighbor discovery code should use this feature, where
201 * we may know the interface but not the FIB number holding
202 * the connected subnet in case someone deleted it from the
203 * default FIB and we need to check the interface.
211 INP_LOCK_ASSERT(inp);
212 mopts = inp->in6p_moptions;
218 * If the source address is explicitly specified by the caller,
219 * check if the requested source address is indeed a unicast address
220 * assigned to the node, and can be used as the packet's source
221 * address. If everything is okay, use the address as source.
223 if (opts && (pi = opts->ip6po_pktinfo) &&
224 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
225 struct sockaddr_in6 srcsock;
226 struct in6_ifaddr *ia6;
228 /* get the outgoing interface */
229 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
230 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB))
235 * determine the appropriate zone id of the source based on
236 * the zone of the destination and the outgoing interface.
237 * If the specified address is ambiguous wrt the scope zone,
238 * the interface must be specified; otherwise, ifa_ifwithaddr()
239 * will fail matching the address.
241 bzero(&srcsock, sizeof(srcsock));
242 srcsock.sin6_family = AF_INET6;
243 srcsock.sin6_len = sizeof(srcsock);
244 srcsock.sin6_addr = pi->ipi6_addr;
246 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
250 if (cred != NULL && (error = prison_local_ip6(cred,
251 &srcsock.sin6_addr, (inp != NULL &&
252 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
255 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
256 (struct sockaddr *)&srcsock);
258 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
260 ifa_free(&ia6->ia_ifa);
261 return (EADDRNOTAVAIL);
263 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
266 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
267 ifa_free(&ia6->ia_ifa);
272 * Otherwise, if the socket has already bound the source, just use it.
274 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
276 (error = prison_local_ip6(cred, &inp->in6p_laddr,
277 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
279 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
284 * Bypass source address selection and use the primary jail IP
287 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
291 * If the address is not specified, choose the best one based on
292 * the outgoing interface and the destination address.
294 /* get the outgoing interface */
295 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
296 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0)
300 if (ifp == NULL) /* this should not happen */
301 panic("in6_selectsrc: NULL ifp");
303 error = in6_setscope(&dst, ifp, &odstzone);
308 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
309 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
310 int new_scope = -1, new_matchlen = -1;
311 struct in6_addrpolicy *new_policy = NULL;
312 u_int32_t srczone, osrczone, dstzone;
314 struct ifnet *ifp1 = ia->ia_ifp;
317 * We'll never take an address that breaks the scope zone
318 * of the destination. We also skip an address if its zone
319 * does not contain the outgoing interface.
320 * XXX: we should probably use sin6_scope_id here.
322 if (in6_setscope(&dst, ifp1, &dstzone) ||
323 odstzone != dstzone) {
326 src = ia->ia_addr.sin6_addr;
327 if (in6_setscope(&src, ifp, &osrczone) ||
328 in6_setscope(&src, ifp1, &srczone) ||
329 osrczone != srczone) {
333 /* avoid unusable addresses */
335 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
338 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
341 /* If jailed only take addresses of the jail into account. */
343 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
346 /* Rule 1: Prefer same address */
347 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
349 BREAK(1); /* there should be no better candidate */
355 /* Rule 2: Prefer appropriate scope */
357 dst_scope = in6_addrscope(&dst);
358 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
359 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
360 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
363 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
364 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
370 * Rule 3: Avoid deprecated addresses. Note that the case of
371 * !ip6_use_deprecated is already rejected above.
373 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
375 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
378 /* Rule 4: Prefer home addresses */
380 * XXX: This is a TODO. We should probably merge the MIP6
384 /* Rule 5: Prefer outgoing interface */
385 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
386 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
388 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
393 * Rule 6: Prefer matching label
394 * Note that best_policy should be non-NULL here.
396 if (dst_policy == NULL)
397 dst_policy = lookup_addrsel_policy(dstsock);
398 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
399 new_policy = lookup_addrsel_policy(&ia->ia_addr);
400 if (dst_policy->label == best_policy->label &&
401 dst_policy->label != new_policy->label)
403 if (dst_policy->label != best_policy->label &&
404 dst_policy->label == new_policy->label)
409 * Rule 7: Prefer public addresses.
410 * We allow users to reverse the logic by configuring
411 * a sysctl variable, so that privacy conscious users can
412 * always prefer temporary addresses.
415 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
416 prefer_tempaddr = V_ip6_prefer_tempaddr;
417 } else if (opts->ip6po_prefer_tempaddr ==
418 IP6PO_TEMPADDR_NOTPREFER) {
422 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
423 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
429 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
430 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
438 * Rule 8: prefer addresses on alive interfaces.
439 * This is a KAME specific rule.
441 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
442 !(ia->ia_ifp->if_flags & IFF_UP))
444 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
445 (ia->ia_ifp->if_flags & IFF_UP))
449 * Rule 9: prefer address with better virtual status.
451 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
453 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
457 * Rule 10: prefer address with `prefer_source' flag.
459 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
460 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
462 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
463 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
467 * Rule 14: Use longest matching prefix.
468 * Note: in the address selection draft, this rule is
469 * documented as "Rule 8". However, since it is also
470 * documented that this rule can be overridden, we assign
471 * a large number so that it is easy to assign smaller numbers
472 * to more preferred rules.
474 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
475 if (best_matchlen < new_matchlen)
477 if (new_matchlen < best_matchlen)
480 /* Rule 15 is reserved. */
483 * Last resort: just keep the current candidate.
484 * Or, do we need more rules?
490 best_scope = (new_scope >= 0 ? new_scope :
491 in6_addrscope(&ia_best->ia_addr.sin6_addr));
492 best_policy = (new_policy ? new_policy :
493 lookup_addrsel_policy(&ia_best->ia_addr));
494 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
495 in6_matchlen(&ia_best->ia_addr.sin6_addr,
505 if ((ia = ia_best) == NULL) {
506 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
507 IP6STAT_INC(ip6s_sources_none);
508 return (EADDRNOTAVAIL);
512 * At this point at least one of the addresses belonged to the jail
513 * but it could still be, that we want to further restrict it, e.g.
514 * theoratically IN6_IS_ADDR_LOOPBACK.
515 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
516 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
517 * let all others previously selected pass.
518 * Use tmp to not change ::1 on lo0 to the primary jail address.
520 tmp = ia->ia_addr.sin6_addr;
521 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
522 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
523 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
524 IP6STAT_INC(ip6s_sources_none);
525 return (EADDRNOTAVAIL);
531 bcopy(&tmp, srcp, sizeof(*srcp));
532 if (ia->ia_ifp == ifp)
533 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
535 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
536 if (dst_scope == best_scope)
537 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
539 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
540 if (IFA6_IS_DEPRECATED(ia))
541 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
542 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
547 * clone - meaningful only for bsdi and freebsd
550 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
551 struct ip6_moptions *mopts, struct route_in6 *ro,
552 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
555 struct ifnet *ifp = NULL;
556 struct rtentry *rt = NULL;
557 struct sockaddr_in6 *sin6_next;
558 struct in6_pktinfo *pi = NULL;
559 struct in6_addr *dst = &dstsock->sin6_addr;
562 char ip6buf[INET6_ADDRSTRLEN];
564 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
565 dstsock->sin6_addr.s6_addr32[1] == 0 &&
566 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
567 printf("in6_selectroute: strange destination %s\n",
568 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
570 printf("in6_selectroute: destination = %s%%%d\n",
571 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
572 dstsock->sin6_scope_id); /* for debug */
576 /* If the caller specify the outgoing interface explicitly, use it. */
577 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
578 /* XXX boundary check is assumed to be already done. */
579 ifp = ifnet_byindex(pi->ipi6_ifindex);
581 (norouteok || retrt == NULL ||
582 IN6_IS_ADDR_MULTICAST(dst))) {
584 * we do not have to check or get the route for
592 * If the destination address is a multicast address and the outgoing
593 * interface for the address is specified by the caller, use it.
595 if (IN6_IS_ADDR_MULTICAST(dst) &&
596 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
597 goto done; /* we do not need a route for multicast. */
600 * If destination address is LLA or link- or node-local multicast,
601 * use it's embedded scope zone id to determine outgoing interface.
603 if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
604 IN6_IS_ADDR_MC_NODELOCAL(dst)) {
605 zoneid = ntohs(in6_getscope(dst));
607 ifp = in6_getlinkifnet(zoneid);
614 * If the next hop address for the packet is specified by the caller,
615 * use it as the gateway.
617 if (opts && opts->ip6po_nexthop) {
618 struct route_in6 *ron;
620 sin6_next = satosin6(opts->ip6po_nexthop);
621 if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
623 * Next hop is LLA, thus it should be neighbor.
624 * Determine outgoing interface by zone index.
626 zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
628 ifp = in6_getlinkifnet(zoneid);
632 ron = &opts->ip6po_nextroute;
633 /* Use a cached route if it exists and is valid. */
634 if (ron->ro_rt != NULL && (
635 (ron->ro_rt->rt_flags & RTF_UP) == 0 ||
636 ron->ro_dst.sin6_family != AF_INET6 ||
637 !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
638 &sin6_next->sin6_addr)))
640 if (ron->ro_rt == NULL) {
641 ron->ro_dst = *sin6_next;
642 in6_rtalloc(ron, fibnum); /* multi path case? */
645 * The node identified by that address must be a
646 * neighbor of the sending host.
648 if (ron->ro_rt == NULL ||
649 (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0)
650 error = EHOSTUNREACH;
655 * Use a cached route if it exists and is valid, else try to allocate
656 * a new one. Note that we should check the address family of the
657 * cached destination, in case of sharing the cache with IPv4.
661 (!(ro->ro_rt->rt_flags & RTF_UP) ||
662 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
663 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
666 ro->ro_rt = (struct rtentry *)NULL;
668 if (ro->ro_rt == (struct rtentry *)NULL) {
669 struct sockaddr_in6 *sa6;
671 /* No route yet, so try to acquire one */
672 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
673 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
675 sa6->sin6_scope_id = 0;
678 rtalloc_mpath_fib((struct route *)ro,
679 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
681 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
682 &ro->ro_dst, 0, 0UL, fibnum);
684 RT_UNLOCK(ro->ro_rt);
689 * do not care about the result if we have the nexthop
690 * explicitly specified.
692 if (opts && opts->ip6po_nexthop)
696 ifp = ro->ro_rt->rt_ifp;
698 if (ifp == NULL) { /* can this really happen? */
703 if (ro->ro_rt == NULL)
704 error = EHOSTUNREACH;
708 * Check if the outgoing interface conflicts with
709 * the interface specified by ipi6_ifindex (if specified).
710 * Note that loopback interface is always okay.
711 * (this may happen when we are sending a packet to one of
712 * our own addresses.)
714 if (ifp && opts && opts->ip6po_pktinfo &&
715 opts->ip6po_pktinfo->ipi6_ifindex) {
716 if (!(ifp->if_flags & IFF_LOOPBACK) &&
718 opts->ip6po_pktinfo->ipi6_ifindex) {
719 error = EHOSTUNREACH;
726 if (ifp == NULL && rt == NULL) {
728 * This can happen if the caller did not pass a cached route
729 * nor any other hints. We treat this case an error.
731 error = EHOSTUNREACH;
733 if (error == EHOSTUNREACH)
734 IP6STAT_INC(ip6s_noroute);
736 if (retifp != NULL) {
740 * Adjust the "outgoing" interface. If we're going to loop
741 * the packet back to ourselves, the ifp would be the loopback
742 * interface. However, we'd rather know the interface associated
743 * to the destination address (which should probably be one of
744 * our own addresses.)
747 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
748 (rt->rt_gateway->sa_family == AF_LINK))
750 ifnet_byindex(((struct sockaddr_dl *)
751 rt->rt_gateway)->sdl_index);
756 *retrt = rt; /* rt may be NULL */
762 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
763 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp,
764 struct ifnet *oifp, u_int fibnum)
767 struct route_in6 sro;
768 struct rtentry *rt = NULL;
770 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
773 bzero(&sro, sizeof(sro));
777 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
778 &rt, 1, fibnum)) != 0) {
779 if (ro == &sro && rt && rt == sro.ro_rt)
781 /* Help ND. See oifp comment in in6_selectsrc(). */
782 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
790 * do not use a rejected or black hole route.
791 * XXX: this check should be done in the L2 output routine.
792 * However, if we skipped this check here, we'd see the following
794 * - install a rejected route for a scoped address prefix
796 * - send a packet to a destination that matches the scoped prefix,
797 * with ambiguity about the scope zone.
798 * - pick the outgoing interface from the route, and disambiguate the
799 * scope zone with the interface.
800 * - ip6_output() would try to get another route with the "new"
801 * destination, which may be valid.
802 * - we'd see no error on output.
803 * Although this may not be very harmful, it should still be confusing.
804 * We thus reject the case here.
806 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
807 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
809 if (ro == &sro && rt && rt == sro.ro_rt)
814 if (ro == &sro && rt && rt == sro.ro_rt)
820 * Public wrapper function to selectroute().
822 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
823 * in6_selectroute_fib() function is only there for backward compat on stable.
826 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
827 struct ip6_moptions *mopts, struct route_in6 *ro,
828 struct ifnet **retifp, struct rtentry **retrt)
831 return (selectroute(dstsock, opts, mopts, ro, retifp,
832 retrt, 0, RT_DEFAULT_FIB));
837 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
838 struct ip6_moptions *mopts, struct route_in6 *ro,
839 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
842 return (selectroute(dstsock, opts, mopts, ro, retifp,
848 * Default hop limit selection. The precedence is as follows:
849 * 1. Hoplimit value specified via ioctl.
850 * 2. (If the outgoing interface is detected) the current
851 * hop limit of the interface specified by router advertisement.
852 * 3. The system default hoplimit.
855 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
858 if (in6p && in6p->in6p_hops >= 0)
859 return (in6p->in6p_hops);
861 return (ND_IFINFO(ifp)->chlim);
862 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
863 struct route_in6 ro6;
866 bzero(&ro6, sizeof(ro6));
867 ro6.ro_dst.sin6_family = AF_INET6;
868 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
869 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
870 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum);
872 lifp = ro6.ro_rt->rt_ifp;
875 return (ND_IFINFO(lifp)->chlim);
878 return (V_ip6_defhlim);
882 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
883 * share this function by all *bsd*...
886 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
888 struct socket *so = inp->inp_socket;
890 int error, lookupflags = 0;
892 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
895 INP_WLOCK_ASSERT(inp);
896 INP_HASH_WLOCK_ASSERT(pcbinfo);
898 error = prison_local_ip6(cred, laddr,
899 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
903 /* XXX: this is redundant when called from in6_pcbbind */
904 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
905 lookupflags = INPLOOKUP_WILDCARD;
907 inp->inp_flags |= INP_ANONPORT;
909 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
913 inp->inp_lport = lport;
914 if (in_pcbinshash(inp) != 0) {
915 inp->in6p_laddr = in6addr_any;
924 addrsel_policy_init(void)
929 /* initialize the "last resort" policy */
930 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
931 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
933 if (!IS_DEFAULT_VNET(curvnet))
937 ADDRSEL_SXLOCK_INIT();
940 static struct in6_addrpolicy *
941 lookup_addrsel_policy(struct sockaddr_in6 *key)
943 struct in6_addrpolicy *match = NULL;
946 match = match_addrsel_policy(key);
949 match = &V_defaultaddrpolicy;
958 * Subroutines to manage the address selection policy table via sysctl.
961 struct sysctl_req *w_req;
964 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
965 SYSCTL_DECL(_net_inet6_ip6);
966 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
967 CTLFLAG_RD, in6_src_sysctl, "");
970 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
977 bzero(&w, sizeof(w));
980 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
984 in6_src_ioctl(u_long cmd, caddr_t data)
986 struct in6_addrpolicy ent0;
988 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
989 return (EOPNOTSUPP); /* check for safety */
991 ent0 = *(struct in6_addrpolicy *)data;
993 if (ent0.label == ADDR_LABEL_NOTAPP)
995 /* check if the prefix mask is consecutive. */
996 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
998 /* clear trailing garbages (if any) of the prefix address. */
999 IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1003 case SIOCAADDRCTL_POLICY:
1004 return (add_addrsel_policyent(&ent0));
1005 case SIOCDADDRCTL_POLICY:
1006 return (delete_addrsel_policyent(&ent0));
1009 return (0); /* XXX: compromise compilers */
1013 * The followings are implementation of the policy table using a
1014 * simple tail queue.
1015 * XXX such details should be hidden.
1016 * XXX implementation using binary tree should be more efficient.
1018 struct addrsel_policyent {
1019 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1020 struct in6_addrpolicy ape_policy;
1023 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1025 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1026 #define V_addrsel_policytab VNET(addrsel_policytab)
1029 init_policy_queue(void)
1032 TAILQ_INIT(&V_addrsel_policytab);
1036 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1038 struct addrsel_policyent *new, *pol;
1040 new = malloc(sizeof(*new), M_IFADDR,
1045 /* duplication check */
1046 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1047 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1048 &pol->ape_policy.addr.sin6_addr) &&
1049 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1050 &pol->ape_policy.addrmask.sin6_addr)) {
1053 free(new, M_IFADDR);
1054 return (EEXIST); /* or override it? */
1058 bzero(new, sizeof(*new));
1060 /* XXX: should validate entry */
1061 new->ape_policy = *newpolicy;
1063 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1071 delete_addrsel_policyent(struct in6_addrpolicy *key)
1073 struct addrsel_policyent *pol;
1078 /* search for the entry in the table */
1079 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1080 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1081 &pol->ape_policy.addr.sin6_addr) &&
1082 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1083 &pol->ape_policy.addrmask.sin6_addr)) {
1093 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1096 free(pol, M_IFADDR);
1102 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1104 struct addrsel_policyent *pol;
1108 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1109 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1119 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1122 struct walkarg *w = arg;
1124 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1129 static struct in6_addrpolicy *
1130 match_addrsel_policy(struct sockaddr_in6 *key)
1132 struct addrsel_policyent *pent;
1133 struct in6_addrpolicy *bestpol = NULL, *pol;
1134 int matchlen, bestmatchlen = -1;
1135 u_char *mp, *ep, *k, *p, m;
1137 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1140 pol = &pent->ape_policy;
1141 mp = (u_char *)&pol->addrmask.sin6_addr;
1142 ep = mp + 16; /* XXX: scope field? */
1143 k = (u_char *)&key->sin6_addr;
1144 p = (u_char *)&pol->addr.sin6_addr;
1145 for (; mp < ep && *mp; mp++, k++, p++) {
1148 goto next; /* not match */
1149 if (m == 0xff) /* short cut for a typical case */
1159 /* matched. check if this is better than the current best. */
1160 if (bestpol == NULL ||
1161 matchlen > bestmatchlen) {
1163 bestmatchlen = matchlen;