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/route.h>
89 #include <net/if_llatbl.h>
91 #include <net/radix_mpath.h>
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/ip.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/ip_var.h>
100 #include <netinet/udp.h>
101 #include <netinet/udp_var.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet/ip6.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/nd6.h>
110 static struct mtx addrsel_lock;
111 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
112 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
113 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
114 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
116 static struct sx addrsel_sxlock;
117 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
118 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
119 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
120 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
121 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
123 #define ADDR_LABEL_NOTAPP (-1)
125 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
126 VNET_DEFINE(int, ip6_prefer_tempaddr);
128 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
130 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *,
131 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
132 struct rtentry **, int));
133 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
134 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
136 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
138 static void init_policy_queue(void);
139 static int add_addrsel_policyent(struct in6_addrpolicy *);
140 static int delete_addrsel_policyent(struct in6_addrpolicy *);
141 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
143 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
144 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
147 * Return an IPv6 address, which is the most appropriate for a given
148 * destination and user specified options.
149 * If necessary, this function lookups the routing table and returns
150 * an entry to the caller for later use.
152 #define REPLACE(r) do {\
153 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
154 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
155 V_ip6stat.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 {\
163 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
164 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
165 V_ip6stat.ip6s_sources_rule[(r)]++; \
167 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
168 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)); \
170 goto next; /* XXX: we can't use 'continue' here */ \
172 #define BREAK(r) do { \
173 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
174 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
175 V_ip6stat.ip6s_sources_rule[(r)]++; \
176 goto out; /* XXX: we can't use 'break' here */ \
180 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
181 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
182 struct ifnet **ifpp, struct in6_addr *srcp)
185 struct ifnet *ifp = NULL;
186 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
187 struct in6_pktinfo *pi = NULL;
188 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
189 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
193 struct ip6_moptions *mopts;
195 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
197 dst = dstsock->sin6_addr; /* make a copy for local operation */
202 INP_LOCK_ASSERT(inp);
203 mopts = inp->in6p_moptions;
209 * If the source address is explicitly specified by the caller,
210 * check if the requested source address is indeed a unicast address
211 * assigned to the node, and can be used as the packet's source
212 * address. If everything is okay, use the address as source.
214 if (opts && (pi = opts->ip6po_pktinfo) &&
215 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
216 struct sockaddr_in6 srcsock;
217 struct in6_ifaddr *ia6;
219 /* get the outgoing interface */
220 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
224 * determine the appropriate zone id of the source based on
225 * the zone of the destination and the outgoing interface.
226 * If the specified address is ambiguous wrt the scope zone,
227 * the interface must be specified; otherwise, ifa_ifwithaddr()
228 * will fail matching the address.
230 bzero(&srcsock, sizeof(srcsock));
231 srcsock.sin6_family = AF_INET6;
232 srcsock.sin6_len = sizeof(srcsock);
233 srcsock.sin6_addr = pi->ipi6_addr;
235 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
239 if (cred != NULL && (error = prison_local_ip6(cred,
240 &srcsock.sin6_addr, (inp != NULL &&
241 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
244 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
245 (struct sockaddr *)&srcsock);
247 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
249 ifa_free(&ia6->ia_ifa);
250 return (EADDRNOTAVAIL);
252 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
255 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
256 ifa_free(&ia6->ia_ifa);
261 * Otherwise, if the socket has already bound the source, just use it.
263 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
265 (error = prison_local_ip6(cred, &inp->in6p_laddr,
266 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
268 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
273 * If the address is not specified, choose the best one based on
274 * the outgoing interface and the destination address.
276 /* get the outgoing interface */
277 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
281 if (ifp == NULL) /* this should not happen */
282 panic("in6_selectsrc: NULL ifp");
284 error = in6_setscope(&dst, ifp, &odstzone);
289 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
290 int new_scope = -1, new_matchlen = -1;
291 struct in6_addrpolicy *new_policy = NULL;
292 u_int32_t srczone, osrczone, dstzone;
294 struct ifnet *ifp1 = ia->ia_ifp;
297 * We'll never take an address that breaks the scope zone
298 * of the destination. We also skip an address if its zone
299 * does not contain the outgoing interface.
300 * XXX: we should probably use sin6_scope_id here.
302 if (in6_setscope(&dst, ifp1, &dstzone) ||
303 odstzone != dstzone) {
306 src = ia->ia_addr.sin6_addr;
307 if (in6_setscope(&src, ifp, &osrczone) ||
308 in6_setscope(&src, ifp1, &srczone) ||
309 osrczone != srczone) {
313 /* avoid unusable addresses */
315 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
318 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
322 prison_local_ip6(cred, &ia->ia_addr.sin6_addr,
324 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
327 /* Rule 1: Prefer same address */
328 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
330 BREAK(1); /* there should be no better candidate */
336 /* Rule 2: Prefer appropriate scope */
338 dst_scope = in6_addrscope(&dst);
339 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
340 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
341 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
344 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
345 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
351 * Rule 3: Avoid deprecated addresses. Note that the case of
352 * !ip6_use_deprecated is already rejected above.
354 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
356 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
359 /* Rule 4: Prefer home addresses */
361 * XXX: This is a TODO. We should probably merge the MIP6
365 /* Rule 5: Prefer outgoing interface */
366 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
368 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
372 * Rule 6: Prefer matching label
373 * Note that best_policy should be non-NULL here.
375 if (dst_policy == NULL)
376 dst_policy = lookup_addrsel_policy(dstsock);
377 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
378 new_policy = lookup_addrsel_policy(&ia->ia_addr);
379 if (dst_policy->label == best_policy->label &&
380 dst_policy->label != new_policy->label)
382 if (dst_policy->label != best_policy->label &&
383 dst_policy->label == new_policy->label)
388 * Rule 7: Prefer public addresses.
389 * We allow users to reverse the logic by configuring
390 * a sysctl variable, so that privacy conscious users can
391 * always prefer temporary addresses.
394 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
395 prefer_tempaddr = V_ip6_prefer_tempaddr;
396 } else if (opts->ip6po_prefer_tempaddr ==
397 IP6PO_TEMPADDR_NOTPREFER) {
401 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
402 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
408 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
409 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
417 * Rule 8: prefer addresses on alive interfaces.
418 * This is a KAME specific rule.
420 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
421 !(ia->ia_ifp->if_flags & IFF_UP))
423 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
424 (ia->ia_ifp->if_flags & IFF_UP))
428 * Rule 14: Use longest matching prefix.
429 * Note: in the address selection draft, this rule is
430 * documented as "Rule 8". However, since it is also
431 * documented that this rule can be overridden, we assign
432 * a large number so that it is easy to assign smaller numbers
433 * to more preferred rules.
435 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
436 if (best_matchlen < new_matchlen)
438 if (new_matchlen < best_matchlen)
441 /* Rule 15 is reserved. */
444 * Last resort: just keep the current candidate.
445 * Or, do we need more rules?
451 best_scope = (new_scope >= 0 ? new_scope :
452 in6_addrscope(&ia_best->ia_addr.sin6_addr));
453 best_policy = (new_policy ? new_policy :
454 lookup_addrsel_policy(&ia_best->ia_addr));
455 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
456 in6_matchlen(&ia_best->ia_addr.sin6_addr,
466 if ((ia = ia_best) == NULL) {
467 IN6_IFADDR_RUNLOCK();
468 return (EADDRNOTAVAIL);
474 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
475 IN6_IFADDR_RUNLOCK();
480 * clone - meaningful only for bsdi and freebsd
483 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
484 struct ip6_moptions *mopts, struct route_in6 *ro,
485 struct ifnet **retifp, struct rtentry **retrt, int norouteok)
488 struct ifnet *ifp = NULL;
489 struct rtentry *rt = NULL;
490 struct sockaddr_in6 *sin6_next;
491 struct in6_pktinfo *pi = NULL;
492 struct in6_addr *dst = &dstsock->sin6_addr;
494 char ip6buf[INET6_ADDRSTRLEN];
496 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
497 dstsock->sin6_addr.s6_addr32[1] == 0 &&
498 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
499 printf("in6_selectroute: strange destination %s\n",
500 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
502 printf("in6_selectroute: destination = %s%%%d\n",
503 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
504 dstsock->sin6_scope_id); /* for debug */
508 /* If the caller specify the outgoing interface explicitly, use it. */
509 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
510 /* XXX boundary check is assumed to be already done. */
511 ifp = ifnet_byindex(pi->ipi6_ifindex);
513 (norouteok || retrt == NULL ||
514 IN6_IS_ADDR_MULTICAST(dst))) {
516 * we do not have to check or get the route for
525 * If the destination address is a multicast address and the outgoing
526 * interface for the address is specified by the caller, use it.
528 if (IN6_IS_ADDR_MULTICAST(dst) &&
529 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
530 goto done; /* we do not need a route for multicast. */
535 * If the next hop address for the packet is specified by the caller,
536 * use it as the gateway.
538 if (opts && opts->ip6po_nexthop) {
539 struct route_in6 *ron;
542 sin6_next = satosin6(opts->ip6po_nexthop);
544 /* at this moment, we only support AF_INET6 next hops */
545 if (sin6_next->sin6_family != AF_INET6) {
546 error = EAFNOSUPPORT; /* or should we proceed? */
551 * If the next hop is an IPv6 address, then the node identified
552 * by that address must be a neighbor of the sending host.
554 ron = &opts->ip6po_nextroute;
556 * XXX what do we do here?
561 if (ron->ro_rt == NULL) {
562 rtalloc((struct route *)ron); /* multi path case? */
563 if (ron->ro_rt == NULL) {
568 error = EHOSTUNREACH;
576 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
577 IF_AFDATA_UNLOCK(ifp);
581 error = EHOSTUNREACH;
586 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
587 (RTF_UP | RTF_LLINFO)) ||
588 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
589 &sin6_next->sin6_addr)) {
594 *satosin6(&ron->ro_dst) = *sin6_next;
596 if (ron->ro_rt == NULL) {
597 rtalloc((struct route *)ron); /* multi path case? */
598 if (ron->ro_rt == NULL ||
599 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
604 error = EHOSTUNREACH;
611 * When cloning is required, try to allocate a route to the
612 * destination so that the caller can store path MTU
619 * Use a cached route if it exists and is valid, else try to allocate
620 * a new one. Note that we should check the address family of the
621 * cached destination, in case of sharing the cache with IPv4.
625 (!(ro->ro_rt->rt_flags & RTF_UP) ||
626 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
627 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
630 ro->ro_rt = (struct rtentry *)NULL;
632 if (ro->ro_rt == (struct rtentry *)NULL) {
633 struct sockaddr_in6 *sa6;
635 /* No route yet, so try to acquire one */
636 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
637 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
639 sa6->sin6_scope_id = 0;
642 rtalloc_mpath((struct route *)ro,
643 ntohl(sa6->sin6_addr.s6_addr32[3]));
645 ro->ro_rt = rtalloc1(&((struct route *)ro)
648 RT_UNLOCK(ro->ro_rt);
653 * do not care about the result if we have the nexthop
654 * explicitly specified.
656 if (opts && opts->ip6po_nexthop)
660 ifp = ro->ro_rt->rt_ifp;
662 if (ifp == NULL) { /* can this really happen? */
667 if (ro->ro_rt == NULL)
668 error = EHOSTUNREACH;
672 * Check if the outgoing interface conflicts with
673 * the interface specified by ipi6_ifindex (if specified).
674 * Note that loopback interface is always okay.
675 * (this may happen when we are sending a packet to one of
676 * our own addresses.)
678 if (ifp && opts && opts->ip6po_pktinfo &&
679 opts->ip6po_pktinfo->ipi6_ifindex) {
680 if (!(ifp->if_flags & IFF_LOOPBACK) &&
682 opts->ip6po_pktinfo->ipi6_ifindex) {
683 error = EHOSTUNREACH;
690 if (ifp == NULL && rt == NULL) {
692 * This can happen if the caller did not pass a cached route
693 * nor any other hints. We treat this case an error.
695 error = EHOSTUNREACH;
697 if (error == EHOSTUNREACH)
698 V_ip6stat.ip6s_noroute++;
703 *retrt = rt; /* rt may be NULL */
709 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
710 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp)
713 struct route_in6 sro;
714 struct rtentry *rt = NULL;
717 bzero(&sro, sizeof(sro));
721 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
723 if (ro == &sro && rt && rt == sro.ro_rt)
729 * do not use a rejected or black hole route.
730 * XXX: this check should be done in the L2 output routine.
731 * However, if we skipped this check here, we'd see the following
733 * - install a rejected route for a scoped address prefix
735 * - send a packet to a destination that matches the scoped prefix,
736 * with ambiguity about the scope zone.
737 * - pick the outgoing interface from the route, and disambiguate the
738 * scope zone with the interface.
739 * - ip6_output() would try to get another route with the "new"
740 * destination, which may be valid.
741 * - we'd see no error on output.
742 * Although this may not be very harmful, it should still be confusing.
743 * We thus reject the case here.
745 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
746 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
748 if (ro == &sro && rt && rt == sro.ro_rt)
754 * Adjust the "outgoing" interface. If we're going to loop the packet
755 * back to ourselves, the ifp would be the loopback interface.
756 * However, we'd rather know the interface associated to the
757 * destination address (which should probably be one of our own
760 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
761 *retifp = rt->rt_ifa->ifa_ifp;
763 if (ro == &sro && rt && rt == sro.ro_rt)
769 * clone - meaningful only for bsdi and freebsd
772 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
773 struct ip6_moptions *mopts, struct route_in6 *ro,
774 struct ifnet **retifp, struct rtentry **retrt)
777 return (selectroute(dstsock, opts, mopts, ro, retifp,
782 * Default hop limit selection. The precedence is as follows:
783 * 1. Hoplimit value specified via ioctl.
784 * 2. (If the outgoing interface is detected) the current
785 * hop limit of the interface specified by router advertisement.
786 * 3. The system default hoplimit.
789 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
792 if (in6p && in6p->in6p_hops >= 0)
793 return (in6p->in6p_hops);
795 return (ND_IFINFO(ifp)->chlim);
796 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
797 struct route_in6 ro6;
800 bzero(&ro6, sizeof(ro6));
801 ro6.ro_dst.sin6_family = AF_INET6;
802 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
803 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
804 rtalloc((struct route *)&ro6);
806 lifp = ro6.ro_rt->rt_ifp;
809 return (ND_IFINFO(lifp)->chlim);
811 return (V_ip6_defhlim);
813 return (V_ip6_defhlim);
817 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
818 * share this function by all *bsd*...
821 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
823 struct socket *so = inp->inp_socket;
824 u_int16_t lport = 0, first, last, *lastport;
825 int count, error, wild = 0, dorandom;
826 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
828 INP_INFO_WLOCK_ASSERT(pcbinfo);
829 INP_WLOCK_ASSERT(inp);
831 error = prison_local_ip6(cred, laddr,
832 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
836 /* XXX: this is redundant when called from in6_pcbbind */
837 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
838 wild = INPLOOKUP_WILDCARD;
840 inp->inp_flags |= INP_ANONPORT;
842 if (inp->inp_flags & INP_HIGHPORT) {
843 first = V_ipport_hifirstauto; /* sysctl */
844 last = V_ipport_hilastauto;
845 lastport = &pcbinfo->ipi_lasthi;
846 } else if (inp->inp_flags & INP_LOWPORT) {
847 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
850 first = V_ipport_lowfirstauto; /* 1023 */
851 last = V_ipport_lowlastauto; /* 600 */
852 lastport = &pcbinfo->ipi_lastlow;
854 first = V_ipport_firstauto; /* sysctl */
855 last = V_ipport_lastauto;
856 lastport = &pcbinfo->ipi_lastport;
860 * For UDP, use random port allocation as long as the user
861 * allows it. For TCP (and as of yet unknown) connections,
862 * use random port allocation only if the user allows it AND
863 * ipport_tick() allows it.
865 if (V_ipport_randomized &&
866 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
871 * It makes no sense to do random port allocation if
872 * we have the only port available.
876 /* Make sure to not include UDP packets in the count. */
877 if (pcbinfo != &V_udbinfo)
878 V_ipport_tcpallocs++;
881 * Instead of having two loops further down counting up or down
882 * make sure that first is always <= last and go with only one
883 * code path implementing all logic.
894 *lastport = first + (arc4random() % (last - first));
896 count = last - first;
899 if (count-- < 0) { /* completely used? */
900 /* Undo an address bind that may have occurred. */
901 inp->in6p_laddr = in6addr_any;
902 return (EADDRNOTAVAIL);
905 if (*lastport < first || *lastport > last)
907 lport = htons(*lastport);
908 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
911 inp->inp_lport = lport;
912 if (in_pcbinshash(inp) != 0) {
913 inp->in6p_laddr = in6addr_any;
922 addrsel_policy_init(void)
925 V_ip6_prefer_tempaddr = 0;
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 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)
987 struct in6_addrpolicy ent0;
989 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
990 return (EOPNOTSUPP); /* check for safety */
992 ent0 = *(struct in6_addrpolicy *)data;
994 if (ent0.label == ADDR_LABEL_NOTAPP)
996 /* check if the prefix mask is consecutive. */
997 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
999 /* clear trailing garbages (if any) of the prefix address. */
1000 for (i = 0; i < 4; i++) {
1001 ent0.addr.sin6_addr.s6_addr32[i] &=
1002 ent0.addrmask.sin6_addr.s6_addr32[i];
1007 case SIOCAADDRCTL_POLICY:
1008 return (add_addrsel_policyent(&ent0));
1009 case SIOCDADDRCTL_POLICY:
1010 return (delete_addrsel_policyent(&ent0));
1013 return (0); /* XXX: compromise compilers */
1017 * The followings are implementation of the policy table using a
1018 * simple tail queue.
1019 * XXX such details should be hidden.
1020 * XXX implementation using binary tree should be more efficient.
1022 struct addrsel_policyent {
1023 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1024 struct in6_addrpolicy ape_policy;
1027 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1029 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1030 #define V_addrsel_policytab VNET(addrsel_policytab)
1033 init_policy_queue(void)
1036 TAILQ_INIT(&V_addrsel_policytab);
1040 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1042 struct addrsel_policyent *new, *pol;
1044 new = malloc(sizeof(*new), M_IFADDR,
1049 /* duplication check */
1050 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1051 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1052 &pol->ape_policy.addr.sin6_addr) &&
1053 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1054 &pol->ape_policy.addrmask.sin6_addr)) {
1057 free(new, M_IFADDR);
1058 return (EEXIST); /* or override it? */
1062 bzero(new, sizeof(*new));
1064 /* XXX: should validate entry */
1065 new->ape_policy = *newpolicy;
1067 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1075 delete_addrsel_policyent(struct in6_addrpolicy *key)
1077 struct addrsel_policyent *pol;
1082 /* search for the entry in the table */
1083 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1084 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1085 &pol->ape_policy.addr.sin6_addr) &&
1086 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1087 &pol->ape_policy.addrmask.sin6_addr)) {
1097 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1105 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1108 struct addrsel_policyent *pol;
1112 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1113 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1123 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1126 struct walkarg *w = arg;
1128 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1133 static struct in6_addrpolicy *
1134 match_addrsel_policy(struct sockaddr_in6 *key)
1136 struct addrsel_policyent *pent;
1137 struct in6_addrpolicy *bestpol = NULL, *pol;
1138 int matchlen, bestmatchlen = -1;
1139 u_char *mp, *ep, *k, *p, m;
1141 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1144 pol = &pent->ape_policy;
1145 mp = (u_char *)&pol->addrmask.sin6_addr;
1146 ep = mp + 16; /* XXX: scope field? */
1147 k = (u_char *)&key->sin6_addr;
1148 p = (u_char *)&pol->addr.sin6_addr;
1149 for (; mp < ep && *mp; mp++, k++, p++) {
1152 goto next; /* not match */
1153 if (m == 0xff) /* short cut for a typical case */
1163 /* matched. check if this is better than the current best. */
1164 if (bestpol == NULL ||
1165 matchlen > bestmatchlen) {
1167 bestmatchlen = matchlen;