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)
126 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
127 VNET_DEFINE(int, ip6_prefer_tempaddr);
129 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
131 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *,
132 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
133 struct rtentry **, int));
134 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
135 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
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 __P((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 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
155 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
156 V_ip6stat.ip6s_sources_rule[(r)]++; \
158 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
159 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)); \
163 #define NEXT(r) do {\
164 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
165 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
166 V_ip6stat.ip6s_sources_rule[(r)]++; \
168 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
169 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)); \
171 goto next; /* XXX: we can't use 'continue' here */ \
173 #define BREAK(r) do { \
174 if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
175 sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
176 V_ip6stat.ip6s_sources_rule[(r)]++; \
177 goto out; /* XXX: we can't use 'break' here */ \
181 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
182 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
183 struct ifnet **ifpp, struct in6_addr *srcp)
186 struct ifnet *ifp = NULL;
187 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
188 struct in6_pktinfo *pi = NULL;
189 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
190 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
194 struct ip6_moptions *mopts;
196 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
198 dst = dstsock->sin6_addr; /* make a copy for local operation */
203 INP_LOCK_ASSERT(inp);
204 mopts = inp->in6p_moptions;
210 * If the source address is explicitly specified by the caller,
211 * check if the requested source address is indeed a unicast address
212 * assigned to the node, and can be used as the packet's source
213 * address. If everything is okay, use the address as source.
215 if (opts && (pi = opts->ip6po_pktinfo) &&
216 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
217 struct sockaddr_in6 srcsock;
218 struct in6_ifaddr *ia6;
220 /* get the outgoing interface */
221 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
225 * determine the appropriate zone id of the source based on
226 * the zone of the destination and the outgoing interface.
227 * If the specified address is ambiguous wrt the scope zone,
228 * the interface must be specified; otherwise, ifa_ifwithaddr()
229 * will fail matching the address.
231 bzero(&srcsock, sizeof(srcsock));
232 srcsock.sin6_family = AF_INET6;
233 srcsock.sin6_len = sizeof(srcsock);
234 srcsock.sin6_addr = pi->ipi6_addr;
236 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
240 if (cred != NULL && (error = prison_local_ip6(cred,
241 &srcsock.sin6_addr, (inp != NULL &&
242 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
245 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
246 (struct sockaddr *)&srcsock);
248 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
250 ifa_free(&ia6->ia_ifa);
251 return (EADDRNOTAVAIL);
253 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
256 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
257 ifa_free(&ia6->ia_ifa);
262 * Otherwise, if the socket has already bound the source, just use it.
264 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
266 (error = prison_local_ip6(cred, &inp->in6p_laddr,
267 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
269 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
274 * Bypass source address selection and use the primary jail IP
277 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
281 * If the address is not specified, choose the best one based on
282 * the outgoing interface and the destination address.
284 /* get the outgoing interface */
285 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
289 if (ifp == NULL) /* this should not happen */
290 panic("in6_selectsrc: NULL ifp");
292 error = in6_setscope(&dst, ifp, &odstzone);
297 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
298 int new_scope = -1, new_matchlen = -1;
299 struct in6_addrpolicy *new_policy = NULL;
300 u_int32_t srczone, osrczone, dstzone;
302 struct ifnet *ifp1 = ia->ia_ifp;
305 * We'll never take an address that breaks the scope zone
306 * of the destination. We also skip an address if its zone
307 * does not contain the outgoing interface.
308 * XXX: we should probably use sin6_scope_id here.
310 if (in6_setscope(&dst, ifp1, &dstzone) ||
311 odstzone != dstzone) {
314 src = ia->ia_addr.sin6_addr;
315 if (in6_setscope(&src, ifp, &osrczone) ||
316 in6_setscope(&src, ifp1, &srczone) ||
317 osrczone != srczone) {
321 /* avoid unusable addresses */
323 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
326 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
330 prison_local_ip6(cred, &ia->ia_addr.sin6_addr,
332 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
335 /* Rule 1: Prefer same address */
336 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
338 BREAK(1); /* there should be no better candidate */
344 /* Rule 2: Prefer appropriate scope */
346 dst_scope = in6_addrscope(&dst);
347 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
348 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
349 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
352 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
353 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
359 * Rule 3: Avoid deprecated addresses. Note that the case of
360 * !ip6_use_deprecated is already rejected above.
362 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
364 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
367 /* Rule 4: Prefer home addresses */
369 * XXX: This is a TODO. We should probably merge the MIP6
373 /* Rule 5: Prefer outgoing interface */
374 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
376 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
380 * Rule 6: Prefer matching label
381 * Note that best_policy should be non-NULL here.
383 if (dst_policy == NULL)
384 dst_policy = lookup_addrsel_policy(dstsock);
385 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
386 new_policy = lookup_addrsel_policy(&ia->ia_addr);
387 if (dst_policy->label == best_policy->label &&
388 dst_policy->label != new_policy->label)
390 if (dst_policy->label != best_policy->label &&
391 dst_policy->label == new_policy->label)
396 * Rule 7: Prefer public addresses.
397 * We allow users to reverse the logic by configuring
398 * a sysctl variable, so that privacy conscious users can
399 * always prefer temporary addresses.
402 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
403 prefer_tempaddr = V_ip6_prefer_tempaddr;
404 } else if (opts->ip6po_prefer_tempaddr ==
405 IP6PO_TEMPADDR_NOTPREFER) {
409 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
410 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
416 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
417 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
425 * Rule 8: prefer addresses on alive interfaces.
426 * This is a KAME specific rule.
428 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
429 !(ia->ia_ifp->if_flags & IFF_UP))
431 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
432 (ia->ia_ifp->if_flags & IFF_UP))
436 * Rule 14: Use longest matching prefix.
437 * Note: in the address selection draft, this rule is
438 * documented as "Rule 8". However, since it is also
439 * documented that this rule can be overridden, we assign
440 * a large number so that it is easy to assign smaller numbers
441 * to more preferred rules.
443 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
444 if (best_matchlen < new_matchlen)
446 if (new_matchlen < best_matchlen)
449 /* Rule 15 is reserved. */
452 * Last resort: just keep the current candidate.
453 * Or, do we need more rules?
459 best_scope = (new_scope >= 0 ? new_scope :
460 in6_addrscope(&ia_best->ia_addr.sin6_addr));
461 best_policy = (new_policy ? new_policy :
462 lookup_addrsel_policy(&ia_best->ia_addr));
463 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
464 in6_matchlen(&ia_best->ia_addr.sin6_addr,
474 if ((ia = ia_best) == NULL) {
475 IN6_IFADDR_RUNLOCK();
476 return (EADDRNOTAVAIL);
482 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
483 IN6_IFADDR_RUNLOCK();
488 * clone - meaningful only for bsdi and freebsd
491 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
492 struct ip6_moptions *mopts, struct route_in6 *ro,
493 struct ifnet **retifp, struct rtentry **retrt, int norouteok)
496 struct ifnet *ifp = NULL;
497 struct rtentry *rt = NULL;
498 struct sockaddr_in6 *sin6_next;
499 struct in6_pktinfo *pi = NULL;
500 struct in6_addr *dst = &dstsock->sin6_addr;
502 char ip6buf[INET6_ADDRSTRLEN];
504 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
505 dstsock->sin6_addr.s6_addr32[1] == 0 &&
506 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
507 printf("in6_selectroute: strange destination %s\n",
508 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
510 printf("in6_selectroute: destination = %s%%%d\n",
511 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
512 dstsock->sin6_scope_id); /* for debug */
516 /* If the caller specify the outgoing interface explicitly, use it. */
517 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
518 /* XXX boundary check is assumed to be already done. */
519 ifp = ifnet_byindex(pi->ipi6_ifindex);
521 (norouteok || retrt == NULL ||
522 IN6_IS_ADDR_MULTICAST(dst))) {
524 * we do not have to check or get the route for
533 * If the destination address is a multicast address and the outgoing
534 * interface for the address is specified by the caller, use it.
536 if (IN6_IS_ADDR_MULTICAST(dst) &&
537 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
538 goto done; /* we do not need a route for multicast. */
543 * If the next hop address for the packet is specified by the caller,
544 * use it as the gateway.
546 if (opts && opts->ip6po_nexthop) {
547 struct route_in6 *ron;
550 sin6_next = satosin6(opts->ip6po_nexthop);
552 /* at this moment, we only support AF_INET6 next hops */
553 if (sin6_next->sin6_family != AF_INET6) {
554 error = EAFNOSUPPORT; /* or should we proceed? */
559 * If the next hop is an IPv6 address, then the node identified
560 * by that address must be a neighbor of the sending host.
562 ron = &opts->ip6po_nextroute;
564 * XXX what do we do here?
569 if (ron->ro_rt == NULL) {
570 rtalloc((struct route *)ron); /* multi path case? */
571 if (ron->ro_rt == NULL) {
576 error = EHOSTUNREACH;
584 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
585 IF_AFDATA_UNLOCK(ifp);
589 error = EHOSTUNREACH;
594 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
595 (RTF_UP | RTF_LLINFO)) ||
596 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
597 &sin6_next->sin6_addr)) {
602 *satosin6(&ron->ro_dst) = *sin6_next;
604 if (ron->ro_rt == NULL) {
605 rtalloc((struct route *)ron); /* multi path case? */
606 if (ron->ro_rt == NULL ||
607 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
612 error = EHOSTUNREACH;
619 * When cloning is required, try to allocate a route to the
620 * destination so that the caller can store path MTU
627 * Use a cached route if it exists and is valid, else try to allocate
628 * a new one. Note that we should check the address family of the
629 * cached destination, in case of sharing the cache with IPv4.
633 (!(ro->ro_rt->rt_flags & RTF_UP) ||
634 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
635 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
638 ro->ro_rt = (struct rtentry *)NULL;
640 if (ro->ro_rt == (struct rtentry *)NULL) {
641 struct sockaddr_in6 *sa6;
643 /* No route yet, so try to acquire one */
644 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
645 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
647 sa6->sin6_scope_id = 0;
650 rtalloc_mpath((struct route *)ro,
651 ntohl(sa6->sin6_addr.s6_addr32[3]));
653 ro->ro_rt = rtalloc1(&((struct route *)ro)
656 RT_UNLOCK(ro->ro_rt);
661 * do not care about the result if we have the nexthop
662 * explicitly specified.
664 if (opts && opts->ip6po_nexthop)
668 ifp = ro->ro_rt->rt_ifp;
670 if (ifp == NULL) { /* can this really happen? */
675 if (ro->ro_rt == NULL)
676 error = EHOSTUNREACH;
680 * Check if the outgoing interface conflicts with
681 * the interface specified by ipi6_ifindex (if specified).
682 * Note that loopback interface is always okay.
683 * (this may happen when we are sending a packet to one of
684 * our own addresses.)
686 if (ifp && opts && opts->ip6po_pktinfo &&
687 opts->ip6po_pktinfo->ipi6_ifindex) {
688 if (!(ifp->if_flags & IFF_LOOPBACK) &&
690 opts->ip6po_pktinfo->ipi6_ifindex) {
691 error = EHOSTUNREACH;
698 if (ifp == NULL && rt == NULL) {
700 * This can happen if the caller did not pass a cached route
701 * nor any other hints. We treat this case an error.
703 error = EHOSTUNREACH;
705 if (error == EHOSTUNREACH)
706 V_ip6stat.ip6s_noroute++;
708 if (retifp != NULL) {
712 * Adjust the "outgoing" interface. If we're going to loop
713 * the packet back to ourselves, the ifp would be the loopback
714 * interface. However, we'd rather know the interface associated
715 * to the destination address (which should probably be one of
716 * our own addresses.)
719 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
720 (rt->rt_gateway->sa_family == AF_LINK))
722 ifnet_byindex(((struct sockaddr_dl *)
723 rt->rt_gateway)->sdl_index);
728 *retrt = rt; /* rt may be NULL */
734 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
735 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp)
738 struct route_in6 sro;
739 struct rtentry *rt = NULL;
742 bzero(&sro, sizeof(sro));
746 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
748 if (ro == &sro && rt && rt == sro.ro_rt)
754 * do not use a rejected or black hole route.
755 * XXX: this check should be done in the L2 output routine.
756 * However, if we skipped this check here, we'd see the following
758 * - install a rejected route for a scoped address prefix
760 * - send a packet to a destination that matches the scoped prefix,
761 * with ambiguity about the scope zone.
762 * - pick the outgoing interface from the route, and disambiguate the
763 * scope zone with the interface.
764 * - ip6_output() would try to get another route with the "new"
765 * destination, which may be valid.
766 * - we'd see no error on output.
767 * Although this may not be very harmful, it should still be confusing.
768 * We thus reject the case here.
770 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
771 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
773 if (ro == &sro && rt && rt == sro.ro_rt)
778 if (ro == &sro && rt && rt == sro.ro_rt)
784 * clone - meaningful only for bsdi and freebsd
787 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
788 struct ip6_moptions *mopts, struct route_in6 *ro,
789 struct ifnet **retifp, struct rtentry **retrt)
792 return (selectroute(dstsock, opts, mopts, ro, retifp,
797 * Default hop limit selection. The precedence is as follows:
798 * 1. Hoplimit value specified via ioctl.
799 * 2. (If the outgoing interface is detected) the current
800 * hop limit of the interface specified by router advertisement.
801 * 3. The system default hoplimit.
804 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
807 if (in6p && in6p->in6p_hops >= 0)
808 return (in6p->in6p_hops);
810 return (ND_IFINFO(ifp)->chlim);
811 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
812 struct route_in6 ro6;
815 bzero(&ro6, sizeof(ro6));
816 ro6.ro_dst.sin6_family = AF_INET6;
817 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
818 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
819 rtalloc((struct route *)&ro6);
821 lifp = ro6.ro_rt->rt_ifp;
824 return (ND_IFINFO(lifp)->chlim);
826 return (V_ip6_defhlim);
828 return (V_ip6_defhlim);
832 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
833 * share this function by all *bsd*...
836 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
838 struct socket *so = inp->inp_socket;
839 u_int16_t lport = 0, first, last, *lastport;
840 int count, error, wild = 0, dorandom;
841 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
843 INP_INFO_WLOCK_ASSERT(pcbinfo);
844 INP_WLOCK_ASSERT(inp);
846 error = prison_local_ip6(cred, laddr,
847 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
851 /* XXX: this is redundant when called from in6_pcbbind */
852 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
853 wild = INPLOOKUP_WILDCARD;
855 inp->inp_flags |= INP_ANONPORT;
857 if (inp->inp_flags & INP_HIGHPORT) {
858 first = V_ipport_hifirstauto; /* sysctl */
859 last = V_ipport_hilastauto;
860 lastport = &pcbinfo->ipi_lasthi;
861 } else if (inp->inp_flags & INP_LOWPORT) {
862 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
865 first = V_ipport_lowfirstauto; /* 1023 */
866 last = V_ipport_lowlastauto; /* 600 */
867 lastport = &pcbinfo->ipi_lastlow;
869 first = V_ipport_firstauto; /* sysctl */
870 last = V_ipport_lastauto;
871 lastport = &pcbinfo->ipi_lastport;
875 * For UDP, use random port allocation as long as the user
876 * allows it. For TCP (and as of yet unknown) connections,
877 * use random port allocation only if the user allows it AND
878 * ipport_tick() allows it.
880 if (V_ipport_randomized &&
881 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
886 * It makes no sense to do random port allocation if
887 * we have the only port available.
891 /* Make sure to not include UDP packets in the count. */
892 if (pcbinfo != &V_udbinfo)
893 V_ipport_tcpallocs++;
896 * Instead of having two loops further down counting up or down
897 * make sure that first is always <= last and go with only one
898 * code path implementing all logic.
909 *lastport = first + (arc4random() % (last - first));
911 count = last - first;
914 if (count-- < 0) { /* completely used? */
915 /* Undo an address bind that may have occurred. */
916 inp->in6p_laddr = in6addr_any;
917 return (EADDRNOTAVAIL);
920 if (*lastport < first || *lastport > last)
922 lport = htons(*lastport);
923 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
926 inp->inp_lport = lport;
927 if (in_pcbinshash(inp) != 0) {
928 inp->in6p_laddr = in6addr_any;
937 addrsel_policy_init(void)
940 V_ip6_prefer_tempaddr = 0;
944 /* initialize the "last resort" policy */
945 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
946 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
948 if (!IS_DEFAULT_VNET(curvnet))
952 ADDRSEL_SXLOCK_INIT();
955 static struct in6_addrpolicy *
956 lookup_addrsel_policy(struct sockaddr_in6 *key)
958 struct in6_addrpolicy *match = NULL;
961 match = match_addrsel_policy(key);
964 match = &V_defaultaddrpolicy;
973 * Subroutines to manage the address selection policy table via sysctl.
976 struct sysctl_req *w_req;
979 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
980 SYSCTL_DECL(_net_inet6_ip6);
981 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
982 CTLFLAG_RD, in6_src_sysctl, "");
985 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
992 bzero(&w, sizeof(w));
995 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
999 in6_src_ioctl(u_long cmd, caddr_t data)
1002 struct in6_addrpolicy ent0;
1004 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1005 return (EOPNOTSUPP); /* check for safety */
1007 ent0 = *(struct in6_addrpolicy *)data;
1009 if (ent0.label == ADDR_LABEL_NOTAPP)
1011 /* check if the prefix mask is consecutive. */
1012 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1014 /* clear trailing garbages (if any) of the prefix address. */
1015 for (i = 0; i < 4; i++) {
1016 ent0.addr.sin6_addr.s6_addr32[i] &=
1017 ent0.addrmask.sin6_addr.s6_addr32[i];
1022 case SIOCAADDRCTL_POLICY:
1023 return (add_addrsel_policyent(&ent0));
1024 case SIOCDADDRCTL_POLICY:
1025 return (delete_addrsel_policyent(&ent0));
1028 return (0); /* XXX: compromise compilers */
1032 * The followings are implementation of the policy table using a
1033 * simple tail queue.
1034 * XXX such details should be hidden.
1035 * XXX implementation using binary tree should be more efficient.
1037 struct addrsel_policyent {
1038 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1039 struct in6_addrpolicy ape_policy;
1042 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1044 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1045 #define V_addrsel_policytab VNET(addrsel_policytab)
1048 init_policy_queue(void)
1051 TAILQ_INIT(&V_addrsel_policytab);
1055 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1057 struct addrsel_policyent *new, *pol;
1059 new = malloc(sizeof(*new), M_IFADDR,
1064 /* duplication check */
1065 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1066 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1067 &pol->ape_policy.addr.sin6_addr) &&
1068 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1069 &pol->ape_policy.addrmask.sin6_addr)) {
1072 free(new, M_IFADDR);
1073 return (EEXIST); /* or override it? */
1077 bzero(new, sizeof(*new));
1079 /* XXX: should validate entry */
1080 new->ape_policy = *newpolicy;
1082 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1090 delete_addrsel_policyent(struct in6_addrpolicy *key)
1092 struct addrsel_policyent *pol;
1097 /* search for the entry in the table */
1098 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1099 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1100 &pol->ape_policy.addr.sin6_addr) &&
1101 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1102 &pol->ape_policy.addrmask.sin6_addr)) {
1112 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1120 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1123 struct addrsel_policyent *pol;
1127 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1128 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1138 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1141 struct walkarg *w = arg;
1143 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1148 static struct in6_addrpolicy *
1149 match_addrsel_policy(struct sockaddr_in6 *key)
1151 struct addrsel_policyent *pent;
1152 struct in6_addrpolicy *bestpol = NULL, *pol;
1153 int matchlen, bestmatchlen = -1;
1154 u_char *mp, *ep, *k, *p, m;
1156 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1159 pol = &pent->ape_policy;
1160 mp = (u_char *)&pol->addrmask.sin6_addr;
1161 ep = mp + 16; /* XXX: scope field? */
1162 k = (u_char *)&key->sin6_addr;
1163 p = (u_char *)&pol->addr.sin6_addr;
1164 for (; mp < ep && *mp; mp++, k++, p++) {
1167 goto next; /* not match */
1168 if (m == 0xff) /* short cut for a typical case */
1178 /* matched. check if this is better than the current best. */
1179 if (bestpol == NULL ||
1180 matchlen > bestmatchlen) {
1182 bestmatchlen = matchlen;