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>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/if_llatbl.h>
92 #include <net/radix_mpath.h>
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/ip.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/udp.h>
102 #include <netinet/udp_var.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_pcb.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/nd6.h>
111 static struct mtx addrsel_lock;
112 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
113 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
114 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
115 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
117 static struct sx addrsel_sxlock;
118 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
119 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
120 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
121 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
122 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
124 #define ADDR_LABEL_NOTAPP (-1)
125 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
126 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
128 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
130 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
131 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
132 struct rtentry **, int, u_int);
133 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
134 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **,
135 struct ifnet *, u_int);
137 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
139 static void init_policy_queue(void);
140 static int add_addrsel_policyent(struct in6_addrpolicy *);
141 static int delete_addrsel_policyent(struct in6_addrpolicy *);
142 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
144 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
145 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
148 * Return an IPv6 address, which is the most appropriate for a given
149 * destination and user specified options.
150 * If necessary, this function lookups the routing table and returns
151 * an entry to the caller for later use.
153 #define REPLACE(r) do {\
154 IP6STAT_INC(ip6s_sources_rule[(r)]); \
157 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
158 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)); \
162 #define NEXT(r) do {\
164 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
165 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)); \
167 goto next; /* XXX: we can't use 'continue' here */ \
169 #define BREAK(r) do { \
170 IP6STAT_INC(ip6s_sources_rule[(r)]); \
172 goto out; /* XXX: we can't use 'break' here */ \
176 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
177 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
178 struct ifnet **ifpp, struct in6_addr *srcp)
180 struct in6_addr dst, tmp;
181 struct ifnet *ifp = NULL, *oifp = NULL;
182 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
183 struct in6_pktinfo *pi = NULL;
184 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
185 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
189 struct ip6_moptions *mopts;
191 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
193 dst = dstsock->sin6_addr; /* make a copy for local operation */
196 * Save a possibly passed in ifp for in6_selectsrc. Only
197 * neighbor discovery code should use this feature, where
198 * we may know the interface but not the FIB number holding
199 * the connected subnet in case someone deleted it from the
200 * default FIB and we need to check the interface.
208 INP_LOCK_ASSERT(inp);
209 mopts = inp->in6p_moptions;
215 * If the source address is explicitly specified by the caller,
216 * check if the requested source address is indeed a unicast address
217 * assigned to the node, and can be used as the packet's source
218 * address. If everything is okay, use the address as source.
220 if (opts && (pi = opts->ip6po_pktinfo) &&
221 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
222 struct sockaddr_in6 srcsock;
223 struct in6_ifaddr *ia6;
225 /* get the outgoing interface */
226 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
227 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB))
232 * determine the appropriate zone id of the source based on
233 * the zone of the destination and the outgoing interface.
234 * If the specified address is ambiguous wrt the scope zone,
235 * the interface must be specified; otherwise, ifa_ifwithaddr()
236 * will fail matching the address.
238 bzero(&srcsock, sizeof(srcsock));
239 srcsock.sin6_family = AF_INET6;
240 srcsock.sin6_len = sizeof(srcsock);
241 srcsock.sin6_addr = pi->ipi6_addr;
243 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
247 if (cred != NULL && (error = prison_local_ip6(cred,
248 &srcsock.sin6_addr, (inp != NULL &&
249 (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 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
259 (struct sockaddr *)&srcsock);
260 if (ia6 == NULL || (ia6->ia6_flags & (IN6_IFF_ANYCAST |
261 IN6_IFF_NOTREADY))) {
263 ifa_free(&ia6->ia_ifa);
264 return (EADDRNOTAVAIL);
266 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
267 ifa_free(&ia6->ia_ifa);
269 bcopy(&srcsock.sin6_addr, srcp, sizeof(*srcp));
270 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
277 * Otherwise, if the socket has already bound the source, just use it.
279 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
281 (error = prison_local_ip6(cred, &inp->in6p_laddr,
282 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
284 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
289 * Bypass source address selection and use the primary jail IP
292 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
296 * If the address is not specified, choose the best one based on
297 * the outgoing interface and the destination address.
299 /* get the outgoing interface */
300 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
301 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0)
305 if (ifp == NULL) /* this should not happen */
306 panic("in6_selectsrc: NULL ifp");
308 error = in6_setscope(&dst, ifp, &odstzone);
314 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
315 int new_scope = -1, new_matchlen = -1;
316 struct in6_addrpolicy *new_policy = NULL;
317 u_int32_t srczone, osrczone, dstzone;
319 struct ifnet *ifp1 = ia->ia_ifp;
322 * We'll never take an address that breaks the scope zone
323 * of the destination. We also skip an address if its zone
324 * does not contain the outgoing interface.
325 * XXX: we should probably use sin6_scope_id here.
327 if (in6_setscope(&dst, ifp1, &dstzone) ||
328 odstzone != dstzone) {
331 src = ia->ia_addr.sin6_addr;
332 if (in6_setscope(&src, ifp, &osrczone) ||
333 in6_setscope(&src, ifp1, &srczone) ||
334 osrczone != srczone) {
338 /* avoid unusable addresses */
340 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
343 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
346 /* If jailed only take addresses of the jail into account. */
348 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
351 /* Rule 1: Prefer same address */
352 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
354 BREAK(1); /* there should be no better candidate */
360 /* Rule 2: Prefer appropriate scope */
362 dst_scope = in6_addrscope(&dst);
363 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
364 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
365 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
368 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
369 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
375 * Rule 3: Avoid deprecated addresses. Note that the case of
376 * !ip6_use_deprecated is already rejected above.
378 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
380 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
383 /* Rule 4: Prefer home addresses */
385 * XXX: This is a TODO. We should probably merge the MIP6
389 /* Rule 5: Prefer outgoing interface */
390 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
391 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
393 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
398 * Rule 6: Prefer matching label
399 * Note that best_policy should be non-NULL here.
401 if (dst_policy == NULL)
402 dst_policy = lookup_addrsel_policy(dstsock);
403 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
404 new_policy = lookup_addrsel_policy(&ia->ia_addr);
405 if (dst_policy->label == best_policy->label &&
406 dst_policy->label != new_policy->label)
408 if (dst_policy->label != best_policy->label &&
409 dst_policy->label == new_policy->label)
414 * Rule 7: Prefer public addresses.
415 * We allow users to reverse the logic by configuring
416 * a sysctl variable, so that privacy conscious users can
417 * always prefer temporary addresses.
420 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
421 prefer_tempaddr = V_ip6_prefer_tempaddr;
422 } else if (opts->ip6po_prefer_tempaddr ==
423 IP6PO_TEMPADDR_NOTPREFER) {
427 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
428 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
434 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
435 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
443 * Rule 8: prefer addresses on alive interfaces.
444 * This is a KAME specific rule.
446 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
447 !(ia->ia_ifp->if_flags & IFF_UP))
449 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
450 (ia->ia_ifp->if_flags & IFF_UP))
454 * Rule 9: prefer address with better virtual status.
456 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
458 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
462 * Rule 10: prefer address with `prefer_source' flag.
464 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
465 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
467 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
468 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
472 * Rule 14: Use longest matching prefix.
473 * Note: in the address selection draft, this rule is
474 * documented as "Rule 8". However, since it is also
475 * documented that this rule can be overridden, we assign
476 * a large number so that it is easy to assign smaller numbers
477 * to more preferred rules.
479 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
480 if (best_matchlen < new_matchlen)
482 if (new_matchlen < best_matchlen)
485 /* Rule 15 is reserved. */
488 * Last resort: just keep the current candidate.
489 * Or, do we need more rules?
495 best_scope = (new_scope >= 0 ? new_scope :
496 in6_addrscope(&ia_best->ia_addr.sin6_addr));
497 best_policy = (new_policy ? new_policy :
498 lookup_addrsel_policy(&ia_best->ia_addr));
499 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
500 in6_matchlen(&ia_best->ia_addr.sin6_addr,
510 if ((ia = ia_best) == NULL) {
511 IN6_IFADDR_RUNLOCK();
512 IP6STAT_INC(ip6s_sources_none);
513 return (EADDRNOTAVAIL);
517 * At this point at least one of the addresses belonged to the jail
518 * but it could still be, that we want to further restrict it, e.g.
519 * theoratically IN6_IS_ADDR_LOOPBACK.
520 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
521 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
522 * let all others previously selected pass.
523 * Use tmp to not change ::1 on lo0 to the primary jail address.
525 tmp = ia->ia_addr.sin6_addr;
526 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
527 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
528 IN6_IFADDR_RUNLOCK();
529 IP6STAT_INC(ip6s_sources_none);
530 return (EADDRNOTAVAIL);
536 bcopy(&tmp, srcp, sizeof(*srcp));
537 if (ia->ia_ifp == ifp)
538 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
540 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
541 if (dst_scope == best_scope)
542 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
544 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
545 if (IFA6_IS_DEPRECATED(ia))
546 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
547 IN6_IFADDR_RUNLOCK();
552 * clone - meaningful only for bsdi and freebsd
555 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
556 struct ip6_moptions *mopts, struct route_in6 *ro,
557 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
560 struct ifnet *ifp = NULL;
561 struct rtentry *rt = NULL;
562 struct sockaddr_in6 *sin6_next;
563 struct in6_pktinfo *pi = NULL;
564 struct in6_addr *dst = &dstsock->sin6_addr;
566 char ip6buf[INET6_ADDRSTRLEN];
568 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
569 dstsock->sin6_addr.s6_addr32[1] == 0 &&
570 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
571 printf("in6_selectroute: strange destination %s\n",
572 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
574 printf("in6_selectroute: destination = %s%%%d\n",
575 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
576 dstsock->sin6_scope_id); /* for debug */
580 /* If the caller specify the outgoing interface explicitly, use it. */
581 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
582 /* XXX boundary check is assumed to be already done. */
583 ifp = ifnet_byindex(pi->ipi6_ifindex);
585 (norouteok || retrt == NULL ||
586 IN6_IS_ADDR_MULTICAST(dst))) {
588 * we do not have to check or get the route for
597 * If the destination address is a multicast address and the outgoing
598 * interface for the address is specified by the caller, use it.
600 if (IN6_IS_ADDR_MULTICAST(dst) &&
601 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
602 goto done; /* we do not need a route for multicast. */
607 * If the next hop address for the packet is specified by the caller,
608 * use it as the gateway.
610 if (opts && opts->ip6po_nexthop) {
611 struct route_in6 *ron;
614 sin6_next = satosin6(opts->ip6po_nexthop);
616 /* at this moment, we only support AF_INET6 next hops */
617 if (sin6_next->sin6_family != AF_INET6) {
618 error = EAFNOSUPPORT; /* or should we proceed? */
623 * If the next hop is an IPv6 address, then the node identified
624 * by that address must be a neighbor of the sending host.
626 ron = &opts->ip6po_nextroute;
628 * XXX what do we do here?
633 if (ron->ro_rt == NULL) {
634 in6_rtalloc(ron, fibnum); /* multi path case? */
635 if (ron->ro_rt == NULL) {
641 error = EHOSTUNREACH;
648 IF_AFDATA_RLOCK(ifp);
649 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
650 IF_AFDATA_RUNLOCK(ifp);
654 error = EHOSTUNREACH;
659 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
660 (RTF_UP | RTF_LLINFO)) ||
661 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
662 &sin6_next->sin6_addr)) {
667 *satosin6(&ron->ro_dst) = *sin6_next;
669 if (ron->ro_rt == NULL) {
670 in6_rtalloc(ron, fibnum); /* multi path case? */
671 if (ron->ro_rt == NULL ||
672 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
677 error = EHOSTUNREACH;
684 * When cloning is required, try to allocate a route to the
685 * destination so that the caller can store path MTU
692 * Use a cached route if it exists and is valid, else try to allocate
693 * a new one. Note that we should check the address family of the
694 * cached destination, in case of sharing the cache with IPv4.
698 (!(ro->ro_rt->rt_flags & RTF_UP) ||
699 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
700 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
703 ro->ro_rt = (struct rtentry *)NULL;
705 if (ro->ro_rt == (struct rtentry *)NULL) {
706 struct sockaddr_in6 *sa6;
708 /* No route yet, so try to acquire one */
709 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
710 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
712 sa6->sin6_scope_id = 0;
715 rtalloc_mpath_fib((struct route *)ro,
716 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
718 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
719 &ro->ro_dst, 0, 0UL, fibnum);
721 RT_UNLOCK(ro->ro_rt);
726 * do not care about the result if we have the nexthop
727 * explicitly specified.
729 if (opts && opts->ip6po_nexthop)
733 ifp = ro->ro_rt->rt_ifp;
735 if (ifp == NULL) { /* can this really happen? */
740 if (ro->ro_rt == NULL)
741 error = EHOSTUNREACH;
745 * Check if the outgoing interface conflicts with
746 * the interface specified by ipi6_ifindex (if specified).
747 * Note that loopback interface is always okay.
748 * (this may happen when we are sending a packet to one of
749 * our own addresses.)
751 if (ifp && opts && opts->ip6po_pktinfo &&
752 opts->ip6po_pktinfo->ipi6_ifindex) {
753 if (!(ifp->if_flags & IFF_LOOPBACK) &&
755 opts->ip6po_pktinfo->ipi6_ifindex) {
756 error = EHOSTUNREACH;
763 if (ifp == NULL && rt == NULL) {
765 * This can happen if the caller did not pass a cached route
766 * nor any other hints. We treat this case an error.
768 error = EHOSTUNREACH;
770 if (error == EHOSTUNREACH)
771 IP6STAT_INC(ip6s_noroute);
773 if (retifp != NULL) {
777 * Adjust the "outgoing" interface. If we're going to loop
778 * the packet back to ourselves, the ifp would be the loopback
779 * interface. However, we'd rather know the interface associated
780 * to the destination address (which should probably be one of
781 * our own addresses.)
784 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
785 (rt->rt_gateway->sa_family == AF_LINK))
787 ifnet_byindex(((struct sockaddr_dl *)
788 rt->rt_gateway)->sdl_index);
793 *retrt = rt; /* rt may be NULL */
799 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
800 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp,
801 struct ifnet *oifp, u_int fibnum)
804 struct route_in6 sro;
805 struct rtentry *rt = NULL;
807 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
810 bzero(&sro, sizeof(sro));
814 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
815 &rt, 1, fibnum)) != 0) {
816 if (ro == &sro && rt && rt == sro.ro_rt)
818 /* Help ND. See oifp comment in in6_selectsrc(). */
819 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
827 * do not use a rejected or black hole route.
828 * XXX: this check should be done in the L2 output routine.
829 * However, if we skipped this check here, we'd see the following
831 * - install a rejected route for a scoped address prefix
833 * - send a packet to a destination that matches the scoped prefix,
834 * with ambiguity about the scope zone.
835 * - pick the outgoing interface from the route, and disambiguate the
836 * scope zone with the interface.
837 * - ip6_output() would try to get another route with the "new"
838 * destination, which may be valid.
839 * - we'd see no error on output.
840 * Although this may not be very harmful, it should still be confusing.
841 * We thus reject the case here.
843 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
844 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
846 if (ro == &sro && rt && rt == sro.ro_rt)
851 if (ro == &sro && rt && rt == sro.ro_rt)
857 * Public wrapper function to selectroute().
859 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
860 * in6_selectroute_fib() function is only there for backward compat on stable.
863 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
864 struct ip6_moptions *mopts, struct route_in6 *ro,
865 struct ifnet **retifp, struct rtentry **retrt)
868 return (selectroute(dstsock, opts, mopts, ro, retifp,
869 retrt, 0, RT_DEFAULT_FIB));
874 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
875 struct ip6_moptions *mopts, struct route_in6 *ro,
876 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
879 return (selectroute(dstsock, opts, mopts, ro, retifp,
885 * Default hop limit selection. The precedence is as follows:
886 * 1. Hoplimit value specified via ioctl.
887 * 2. (If the outgoing interface is detected) the current
888 * hop limit of the interface specified by router advertisement.
889 * 3. The system default hoplimit.
892 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
895 if (in6p && in6p->in6p_hops >= 0)
896 return (in6p->in6p_hops);
898 return (ND_IFINFO(ifp)->chlim);
899 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
900 struct route_in6 ro6;
903 bzero(&ro6, sizeof(ro6));
904 ro6.ro_dst.sin6_family = AF_INET6;
905 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
906 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
907 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum);
909 lifp = ro6.ro_rt->rt_ifp;
912 return (ND_IFINFO(lifp)->chlim);
915 return (V_ip6_defhlim);
919 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
920 * share this function by all *bsd*...
923 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
925 struct socket *so = inp->inp_socket;
927 int error, lookupflags = 0;
929 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
932 INP_WLOCK_ASSERT(inp);
933 INP_HASH_WLOCK_ASSERT(pcbinfo);
935 error = prison_local_ip6(cred, laddr,
936 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
940 /* XXX: this is redundant when called from in6_pcbbind */
941 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
942 lookupflags = INPLOOKUP_WILDCARD;
944 inp->inp_flags |= INP_ANONPORT;
946 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
950 inp->inp_lport = lport;
951 if (in_pcbinshash(inp) != 0) {
952 inp->in6p_laddr = in6addr_any;
961 addrsel_policy_init(void)
966 /* initialize the "last resort" policy */
967 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
968 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
970 if (!IS_DEFAULT_VNET(curvnet))
974 ADDRSEL_SXLOCK_INIT();
977 static struct in6_addrpolicy *
978 lookup_addrsel_policy(struct sockaddr_in6 *key)
980 struct in6_addrpolicy *match = NULL;
983 match = match_addrsel_policy(key);
986 match = &V_defaultaddrpolicy;
995 * Subroutines to manage the address selection policy table via sysctl.
998 struct sysctl_req *w_req;
1001 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1002 SYSCTL_DECL(_net_inet6_ip6);
1003 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1004 CTLFLAG_RD, in6_src_sysctl, "");
1007 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1014 bzero(&w, sizeof(w));
1017 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1021 in6_src_ioctl(u_long cmd, caddr_t data)
1024 struct in6_addrpolicy ent0;
1026 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1027 return (EOPNOTSUPP); /* check for safety */
1029 ent0 = *(struct in6_addrpolicy *)data;
1031 if (ent0.label == ADDR_LABEL_NOTAPP)
1033 /* check if the prefix mask is consecutive. */
1034 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1036 /* clear trailing garbages (if any) of the prefix address. */
1037 for (i = 0; i < 4; i++) {
1038 ent0.addr.sin6_addr.s6_addr32[i] &=
1039 ent0.addrmask.sin6_addr.s6_addr32[i];
1044 case SIOCAADDRCTL_POLICY:
1045 return (add_addrsel_policyent(&ent0));
1046 case SIOCDADDRCTL_POLICY:
1047 return (delete_addrsel_policyent(&ent0));
1050 return (0); /* XXX: compromise compilers */
1054 * The followings are implementation of the policy table using a
1055 * simple tail queue.
1056 * XXX such details should be hidden.
1057 * XXX implementation using binary tree should be more efficient.
1059 struct addrsel_policyent {
1060 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1061 struct in6_addrpolicy ape_policy;
1064 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1066 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1067 #define V_addrsel_policytab VNET(addrsel_policytab)
1070 init_policy_queue(void)
1073 TAILQ_INIT(&V_addrsel_policytab);
1077 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1079 struct addrsel_policyent *new, *pol;
1081 new = malloc(sizeof(*new), M_IFADDR,
1086 /* duplication check */
1087 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1088 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1089 &pol->ape_policy.addr.sin6_addr) &&
1090 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1091 &pol->ape_policy.addrmask.sin6_addr)) {
1094 free(new, M_IFADDR);
1095 return (EEXIST); /* or override it? */
1099 bzero(new, sizeof(*new));
1101 /* XXX: should validate entry */
1102 new->ape_policy = *newpolicy;
1104 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1112 delete_addrsel_policyent(struct in6_addrpolicy *key)
1114 struct addrsel_policyent *pol;
1119 /* search for the entry in the table */
1120 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1121 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1122 &pol->ape_policy.addr.sin6_addr) &&
1123 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1124 &pol->ape_policy.addrmask.sin6_addr)) {
1134 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1137 free(pol, M_IFADDR);
1143 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1145 struct addrsel_policyent *pol;
1149 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1150 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1160 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1163 struct walkarg *w = arg;
1165 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1170 static struct in6_addrpolicy *
1171 match_addrsel_policy(struct sockaddr_in6 *key)
1173 struct addrsel_policyent *pent;
1174 struct in6_addrpolicy *bestpol = NULL, *pol;
1175 int matchlen, bestmatchlen = -1;
1176 u_char *mp, *ep, *k, *p, m;
1178 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1181 pol = &pent->ape_policy;
1182 mp = (u_char *)&pol->addrmask.sin6_addr;
1183 ep = mp + 16; /* XXX: scope field? */
1184 k = (u_char *)&key->sin6_addr;
1185 p = (u_char *)&pol->addr.sin6_addr;
1186 for (; mp < ep && *mp; mp++, k++, p++) {
1189 goto next; /* not match */
1190 if (m == 0xff) /* short cut for a typical case */
1200 /* matched. check if this is better than the current best. */
1201 if (bestpol == NULL ||
1202 matchlen > bestmatchlen) {
1204 bestmatchlen = matchlen;