1 /* $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $ */
2 /* $DragonFly: src/sys/netinet6/ip6_output.c,v 1.18 2005/02/08 22:56:19 hsu Exp $ */
3 /* $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ */
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
69 #include "opt_ip6fw.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
74 #include <sys/param.h>
75 #include <sys/malloc.h>
77 #include <sys/errno.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
86 #include <net/route.h>
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet6/nd6.h>
99 #include <netinet6/ipsec.h>
101 #include <netinet6/ipsec6.h>
103 #include <netproto/key/key.h>
107 #include <netproto/ipsec/ipsec.h>
108 #include <netproto/ipsec/ipsec6.h>
109 #include <netproto/ipsec/key.h>
110 #endif /* FAST_IPSEC */
112 #include <net/ip6fw/ip6_fw.h>
114 #include <net/net_osdep.h>
116 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
119 struct mbuf *ip6e_ip6;
120 struct mbuf *ip6e_hbh;
121 struct mbuf *ip6e_dest1;
122 struct mbuf *ip6e_rthdr;
123 struct mbuf *ip6e_dest2;
126 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
127 struct socket *, struct sockopt *sopt);
128 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
129 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
130 static int ip6_copyexthdr (struct mbuf **, caddr_t, int);
131 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
133 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
134 static int ip6_splithdr (struct mbuf *, struct ip6_exthdrs *);
137 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
138 * header (with pri, len, nxt, hlim, src, dst).
139 * This function may modify ver and hlim only.
140 * The mbuf chain containing the packet will be freed.
141 * The mbuf opt, if present, will not be freed.
143 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
144 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
145 * which is rt_rmx.rmx_mtu.
148 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
149 int flags, struct ip6_moptions *im6o,
150 struct ifnet **ifpp, /* XXX: just for statistics */
153 struct ip6_hdr *ip6, *mhip6;
154 struct ifnet *ifp, *origifp;
156 int hlen, tlen, len, off;
157 struct route_in6 ip6route;
158 struct sockaddr_in6 *dst;
160 struct in6_ifaddr *ia = NULL;
162 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
163 struct ip6_exthdrs exthdrs;
164 struct in6_addr finaldst;
165 struct route_in6 *ro_pmtu = NULL;
169 int needipsectun = 0;
170 struct secpolicy *sp = NULL;
171 struct socket *so = inp ? inp->inp_socket : NULL;
173 ip6 = mtod(m, struct ip6_hdr *);
176 int needipsectun = 0;
177 struct secpolicy *sp = NULL;
179 ip6 = mtod(m, struct ip6_hdr *);
180 #endif /* FAST_IPSEC */
182 #define MAKE_EXTHDR(hp, mp) \
185 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
186 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
187 ((eh)->ip6e_len + 1) << 3); \
193 bzero(&exthdrs, sizeof(exthdrs));
196 /* Hop-by-Hop options header */
197 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
198 /* Destination options header(1st part) */
199 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
201 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
202 /* Destination options header(2nd part) */
203 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
207 /* get a security policy for this packet */
209 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
211 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
214 ipsec6stat.out_inval++;
221 switch (sp->policy) {
222 case IPSEC_POLICY_DISCARD:
224 * This packet is just discarded.
226 ipsec6stat.out_polvio++;
229 case IPSEC_POLICY_BYPASS:
230 case IPSEC_POLICY_NONE:
231 /* no need to do IPsec. */
235 case IPSEC_POLICY_IPSEC:
236 if (sp->req == NULL) {
237 /* acquire a policy */
238 error = key_spdacquire(sp);
244 case IPSEC_POLICY_ENTRUST:
246 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
250 /* get a security policy for this packet */
252 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
254 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
257 newipsecstat.ips_out_inval++;
264 switch (sp->policy) {
265 case IPSEC_POLICY_DISCARD:
267 * This packet is just discarded.
269 newipsecstat.ips_out_polvio++;
272 case IPSEC_POLICY_BYPASS:
273 case IPSEC_POLICY_NONE:
274 /* no need to do IPsec. */
278 case IPSEC_POLICY_IPSEC:
279 if (sp->req == NULL) {
280 /* acquire a policy */
281 error = key_spdacquire(sp);
287 case IPSEC_POLICY_ENTRUST:
289 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
291 #endif /* FAST_IPSEC */
294 * Calculate the total length of the extension header chain.
295 * Keep the length of the unfragmentable part for fragmentation.
298 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
299 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
300 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
301 unfragpartlen = optlen + sizeof(struct ip6_hdr);
302 /* NOTE: we don't add AH/ESP length here. do that later. */
303 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
306 * If we need IPsec, or there is at least one extension header,
307 * separate IP6 header from the payload.
309 if ((needipsec || optlen) && !hdrsplit) {
310 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
314 m = exthdrs.ip6e_ip6;
319 ip6 = mtod(m, struct ip6_hdr *);
321 /* adjust mbuf packet header length */
322 m->m_pkthdr.len += optlen;
323 plen = m->m_pkthdr.len - sizeof(*ip6);
325 /* If this is a jumbo payload, insert a jumbo payload option. */
326 if (plen > IPV6_MAXPACKET) {
328 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
332 m = exthdrs.ip6e_ip6;
336 ip6 = mtod(m, struct ip6_hdr *);
337 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
341 ip6->ip6_plen = htons(plen);
344 * Concatenate headers and fill in next header fields.
345 * Here we have, on "m"
347 * and we insert headers accordingly. Finally, we should be getting:
348 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
350 * during the header composing process, "m" points to IPv6 header.
351 * "mprev" points to an extension header prior to esp.
354 u_char *nexthdrp = &ip6->ip6_nxt;
355 struct mbuf *mprev = m;
358 * we treat dest2 specially. this makes IPsec processing
359 * much easier. the goal here is to make mprev point the
360 * mbuf prior to dest2.
362 * result: IPv6 dest2 payload
363 * m and mprev will point to IPv6 header.
365 if (exthdrs.ip6e_dest2) {
367 panic("assumption failed: hdr not split");
368 exthdrs.ip6e_dest2->m_next = m->m_next;
369 m->m_next = exthdrs.ip6e_dest2;
370 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
371 ip6->ip6_nxt = IPPROTO_DSTOPTS;
374 #define MAKE_CHAIN(m, mp, p, i)\
378 panic("assumption failed: hdr not split"); \
379 *mtod((m), u_char *) = *(p);\
381 p = mtod((m), u_char *);\
382 (m)->m_next = (mp)->m_next;\
388 * result: IPv6 hbh dest1 rthdr dest2 payload
389 * m will point to IPv6 header. mprev will point to the
390 * extension header prior to dest2 (rthdr in the above case).
392 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
393 nexthdrp, IPPROTO_HOPOPTS);
394 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
395 nexthdrp, IPPROTO_DSTOPTS);
396 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
397 nexthdrp, IPPROTO_ROUTING);
399 #if defined(IPSEC) || defined(FAST_IPSEC)
404 * pointers after IPsec headers are not valid any more.
405 * other pointers need a great care too.
406 * (IPsec routines should not mangle mbufs prior to AH/ESP)
408 exthdrs.ip6e_dest2 = NULL;
411 struct ip6_rthdr *rh = NULL;
413 struct ipsec_output_state state;
415 if (exthdrs.ip6e_rthdr) {
416 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
417 segleft_org = rh->ip6r_segleft;
418 rh->ip6r_segleft = 0;
421 bzero(&state, sizeof(state));
423 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
427 /* mbuf is already reclaimed in ipsec6_output_trans. */
437 printf("ip6_output (ipsec): error code %d\n", error);
440 /* don't show these error codes to the user */
446 if (exthdrs.ip6e_rthdr) {
447 /* ah6_output doesn't modify mbuf chain */
448 rh->ip6r_segleft = segleft_org;
456 * If there is a routing header, replace destination address field
457 * with the first hop of the routing header.
459 if (exthdrs.ip6e_rthdr) {
460 struct ip6_rthdr *rh =
461 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
462 struct ip6_rthdr *));
463 struct ip6_rthdr0 *rh0;
465 finaldst = ip6->ip6_dst;
466 switch (rh->ip6r_type) {
467 case IPV6_RTHDR_TYPE_0:
468 rh0 = (struct ip6_rthdr0 *)rh;
469 ip6->ip6_dst = rh0->ip6r0_addr[0];
470 bcopy((caddr_t)&rh0->ip6r0_addr[1],
471 (caddr_t)&rh0->ip6r0_addr[0],
472 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
474 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
476 default: /* is it possible? */
482 /* Source address validation */
483 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
484 (flags & IPV6_DADOUTPUT) == 0) {
486 ip6stat.ip6s_badscope++;
489 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
491 ip6stat.ip6s_badscope++;
495 ip6stat.ip6s_localout++;
502 bzero((caddr_t)ro, sizeof(*ro));
505 if (opt && opt->ip6po_rthdr)
506 ro = &opt->ip6po_route;
507 dst = (struct sockaddr_in6 *)&ro->ro_dst;
509 * If there is a cached route,
510 * check that it is to the same destination
511 * and is still up. If not, free it and try again.
513 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
514 dst->sin6_family != AF_INET6 ||
515 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
517 ro->ro_rt = (struct rtentry *)0;
519 if (ro->ro_rt == 0) {
520 bzero(dst, sizeof(*dst));
521 dst->sin6_family = AF_INET6;
522 dst->sin6_len = sizeof(struct sockaddr_in6);
523 dst->sin6_addr = ip6->ip6_dst;
525 #if defined(IPSEC) || defined(FAST_IPSEC)
526 if (needipsec && needipsectun) {
527 struct ipsec_output_state state;
530 * All the extension headers will become inaccessible
531 * (since they can be encrypted).
532 * Don't panic, we need no more updates to extension headers
533 * on inner IPv6 packet (since they are now encapsulated).
535 * IPv6 [ESP|AH] IPv6 [extension headers] payload
537 bzero(&exthdrs, sizeof(exthdrs));
538 exthdrs.ip6e_ip6 = m;
540 bzero(&state, sizeof(state));
542 state.ro = (struct route *)ro;
543 state.dst = (struct sockaddr *)dst;
545 error = ipsec6_output_tunnel(&state, sp, flags);
548 ro = (struct route_in6 *)state.ro;
549 dst = (struct sockaddr_in6 *)state.dst;
551 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
562 printf("ip6_output (ipsec): error code %d\n", error);
565 /* don't show these error codes to the user */
572 exthdrs.ip6e_ip6 = m;
576 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
579 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
580 #define sin6tosa(sin6) ((struct sockaddr *)(sin6))
582 * interface selection comes here
583 * if an interface is specified from an upper layer,
586 if (ro->ro_rt == 0) {
588 * non-bsdi always clone routes, if parent is
591 rtalloc((struct route *)ro);
593 if (ro->ro_rt == 0) {
594 ip6stat.ip6s_noroute++;
595 error = EHOSTUNREACH;
596 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
599 ia = ifatoia6(ro->ro_rt->rt_ifa);
600 ifp = ro->ro_rt->rt_ifp;
602 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
603 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
604 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
606 in6_ifstat_inc(ifp, ifs6_out_request);
609 * Check if the outgoing interface conflicts with
610 * the interface specified by ifi6_ifindex (if specified).
611 * Note that loopback interface is always okay.
612 * (this may happen when we are sending a packet to one of
613 * our own addresses.)
615 if (opt && opt->ip6po_pktinfo
616 && opt->ip6po_pktinfo->ipi6_ifindex) {
617 if (!(ifp->if_flags & IFF_LOOPBACK)
618 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
619 ip6stat.ip6s_noroute++;
620 in6_ifstat_inc(ifp, ifs6_out_discard);
621 error = EHOSTUNREACH;
626 if (opt && opt->ip6po_hlim != -1)
627 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
630 struct in6_multi *in6m;
632 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
635 * See if the caller provided any multicast options
639 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
640 if (im6o->im6o_multicast_ifp != NULL)
641 ifp = im6o->im6o_multicast_ifp;
643 ip6->ip6_hlim = ip6_defmcasthlim;
646 * See if the caller provided the outgoing interface
647 * as an ancillary data.
648 * Boundary check for ifindex is assumed to be already done.
650 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
651 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
654 * If the destination is a node-local scope multicast,
655 * the packet should be loop-backed only.
657 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
659 * If the outgoing interface is already specified,
660 * it should be a loopback interface.
662 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
663 ip6stat.ip6s_badscope++;
664 error = ENETUNREACH; /* XXX: better error? */
665 /* XXX correct ifp? */
666 in6_ifstat_inc(ifp, ifs6_out_discard);
673 if (opt && opt->ip6po_hlim != -1)
674 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
677 * If caller did not provide an interface lookup a
678 * default in the routing table. This is either a
679 * default for the speicfied group (i.e. a host
680 * route), or a multicast default (a route for the
684 if (ro->ro_rt == NULL) {
686 rtpurelookup((struct sockaddr *)&ro->ro_dst);
688 if (ro->ro_rt == NULL) {
689 ip6stat.ip6s_noroute++;
690 error = EHOSTUNREACH;
691 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
694 ia = ifatoia6(ro->ro_rt->rt_ifa);
695 ifp = ro->ro_rt->rt_ifp;
699 if ((flags & IPV6_FORWARDING) == 0)
700 in6_ifstat_inc(ifp, ifs6_out_request);
701 in6_ifstat_inc(ifp, ifs6_out_mcast);
704 * Confirm that the outgoing interface supports multicast.
706 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
707 ip6stat.ip6s_noroute++;
708 in6_ifstat_inc(ifp, ifs6_out_discard);
712 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
714 (im6o == NULL || im6o->im6o_multicast_loop)) {
716 * If we belong to the destination multicast group
717 * on the outgoing interface, and the caller did not
718 * forbid loopback, loop back a copy.
720 ip6_mloopback(ifp, m, dst);
723 * If we are acting as a multicast router, perform
724 * multicast forwarding as if the packet had just
725 * arrived on the interface to which we are about
726 * to send. The multicast forwarding function
727 * recursively calls this function, using the
728 * IPV6_FORWARDING flag to prevent infinite recursion.
730 * Multicasts that are looped back by ip6_mloopback(),
731 * above, will be forwarded by the ip6_input() routine,
734 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
735 if (ip6_mforward(ip6, ifp, m) != 0) {
742 * Multicasts with a hoplimit of zero may be looped back,
743 * above, but must not be transmitted on a network.
744 * Also, multicasts addressed to the loopback interface
745 * are not sent -- the above call to ip6_mloopback() will
746 * loop back a copy if this host actually belongs to the
747 * destination group on the loopback interface.
749 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
756 * Fill the outgoing inteface to tell the upper layer
757 * to increment per-interface statistics.
763 * Determine path MTU.
766 /* The first hop and the final destination may differ. */
767 struct sockaddr_in6 *sin6_fin =
768 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
770 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
771 !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
773 RTFREE(ro_pmtu->ro_rt);
774 ro_pmtu->ro_rt = (struct rtentry *)0;
776 if (ro_pmtu->ro_rt == 0) {
777 bzero(sin6_fin, sizeof(*sin6_fin));
778 sin6_fin->sin6_family = AF_INET6;
779 sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
780 sin6_fin->sin6_addr = finaldst;
782 rtalloc((struct route *)ro_pmtu);
785 if (ro_pmtu->ro_rt != NULL) {
786 u_int32_t ifmtu = ND_IFINFO(ifp)->linkmtu;
788 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
789 if (mtu > ifmtu || mtu == 0) {
791 * The MTU on the route is larger than the MTU on
792 * the interface! This shouldn't happen, unless the
793 * MTU of the interface has been changed after the
794 * interface was brought up. Change the MTU in the
795 * route to match the interface MTU (as long as the
796 * field isn't locked).
798 * if MTU on the route is 0, we need to fix the MTU.
799 * this case happens with path MTU discovery timeouts.
802 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
803 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
806 mtu = ND_IFINFO(ifp)->linkmtu;
810 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
812 if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
815 /* Fake scoped addresses */
816 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
818 * If source or destination address is a scoped address, and
819 * the packet is going to be sent to a loopback interface,
820 * we should keep the original interface.
824 * XXX: this is a very experimental and temporary solution.
825 * We eventually have sockaddr_in6 and use the sin6_scope_id
826 * field of the structure here.
827 * We rely on the consistency between two scope zone ids
828 * of source and destination, which should already be assured.
829 * Larger scopes than link will be supported in the future.
832 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
833 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
834 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
835 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
837 * XXX: origifp can be NULL even in those two cases above.
838 * For example, if we remove the (only) link-local address
839 * from the loopback interface, and try to send a link-local
840 * address without link-id information. Then the source
841 * address is ::1, and the destination address is the
842 * link-local address with its s6_addr16[1] being zero.
843 * What is worse, if the packet goes to the loopback interface
844 * by a default rejected route, the null pointer would be
845 * passed to looutput, and the kernel would hang.
846 * The following last resort would prevent such disaster.
854 * clear embedded scope identifiers if necessary.
855 * in6_clearscope will touch the addresses only when necessary.
857 in6_clearscope(&ip6->ip6_src);
858 in6_clearscope(&ip6->ip6_dst);
861 * Check with the firewall...
863 if (ip6_fw_enable && ip6_fw_chk_ptr) {
865 m->m_pkthdr.rcvif = NULL; /* XXX */
866 /* If ipfw says divert, we have to just drop packet */
867 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
878 * If the outgoing packet contains a hop-by-hop options header,
879 * it must be examined and processed even by the source node.
880 * (RFC 2460, section 4.)
882 if (exthdrs.ip6e_hbh) {
883 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
884 u_int32_t dummy1; /* XXX unused */
885 u_int32_t dummy2; /* XXX unused */
888 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
889 panic("ip6e_hbh is not continuous");
892 * XXX: if we have to send an ICMPv6 error to the sender,
893 * we need the M_LOOP flag since icmp6_error() expects
894 * the IPv6 and the hop-by-hop options header are
895 * continuous unless the flag is set.
897 m->m_flags |= M_LOOP;
898 m->m_pkthdr.rcvif = ifp;
899 if (ip6_process_hopopts(m,
900 (u_int8_t *)(hbh + 1),
901 ((hbh->ip6h_len + 1) << 3) -
902 sizeof(struct ip6_hbh),
903 &dummy1, &dummy2) < 0) {
904 /* m was already freed at this point */
905 error = EINVAL;/* better error? */
908 m->m_flags &= ~M_LOOP; /* XXX */
909 m->m_pkthdr.rcvif = NULL;
913 * Run through list of hooks for output packets.
915 if (pfil_has_hooks(&inet6_pfil_hook)) {
916 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
917 if (error != 0 || m == NULL)
919 ip6 = mtod(m, struct ip6_hdr *);
923 * Send the packet to the outgoing interface.
924 * If necessary, do IPv6 fragmentation before sending.
926 tlen = m->m_pkthdr.len;
930 * On any link that cannot convey a 1280-octet packet in one piece,
931 * link-specific fragmentation and reassembly must be provided at
932 * a layer below IPv6. [RFC 2460, sec.5]
933 * Thus if the interface has ability of link-level fragmentation,
934 * we can just send the packet even if the packet size is
935 * larger than the link's MTU.
936 * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
939 || ifp->if_flags & IFF_FRAGMENTABLE
943 /* Record statistics for this interface address. */
944 if (ia && !(flags & IPV6_FORWARDING)) {
945 ia->ia_ifa.if_opackets++;
946 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
949 /* clean ipsec history once it goes out of the node */
952 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
954 } else if (mtu < IPV6_MMTU) {
956 * note that path MTU is never less than IPV6_MMTU
960 in6_ifstat_inc(ifp, ifs6_out_fragfail);
962 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
964 in6_ifstat_inc(ifp, ifs6_out_fragfail);
967 struct mbuf **mnext, *m_frgpart;
968 struct ip6_frag *ip6f;
969 u_int32_t id = htonl(ip6_id++);
973 * Too large for the destination or interface;
974 * fragment if possible.
975 * Must be able to put at least 8 bytes per fragment.
977 hlen = unfragpartlen;
978 if (mtu > IPV6_MAXPACKET)
979 mtu = IPV6_MAXPACKET;
981 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
984 in6_ifstat_inc(ifp, ifs6_out_fragfail);
988 mnext = &m->m_nextpkt;
991 * Change the next header field of the last header in the
992 * unfragmentable part.
994 if (exthdrs.ip6e_rthdr) {
995 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
996 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
997 } else if (exthdrs.ip6e_dest1) {
998 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
999 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1000 } else if (exthdrs.ip6e_hbh) {
1001 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1002 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1004 nextproto = ip6->ip6_nxt;
1005 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1009 * Loop through length of segment after first fragment,
1010 * make new header and copy data of each part and link onto
1014 for (off = hlen; off < tlen; off += len) {
1015 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1018 ip6stat.ip6s_odropped++;
1021 m->m_pkthdr.rcvif = NULL;
1022 m->m_flags = m0->m_flags & M_COPYFLAGS;
1024 mnext = &m->m_nextpkt;
1025 m->m_data += max_linkhdr;
1026 mhip6 = mtod(m, struct ip6_hdr *);
1028 m->m_len = sizeof(*mhip6);
1029 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1031 ip6stat.ip6s_odropped++;
1034 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1035 if (off + len >= tlen)
1038 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1039 mhip6->ip6_plen = htons((u_short)(len + hlen +
1041 sizeof(struct ip6_hdr)));
1042 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1044 ip6stat.ip6s_odropped++;
1047 m_cat(m, m_frgpart);
1048 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1049 m->m_pkthdr.rcvif = (struct ifnet *)0;
1050 ip6f->ip6f_reserved = 0;
1051 ip6f->ip6f_ident = id;
1052 ip6f->ip6f_nxt = nextproto;
1053 ip6stat.ip6s_ofragments++;
1054 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1057 in6_ifstat_inc(ifp, ifs6_out_fragok);
1061 * Remove leading garbages.
1067 for (m0 = m; m; m = m0) {
1071 /* Record statistics for this interface address. */
1073 ia->ia_ifa.if_opackets++;
1074 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1077 /* clean ipsec history once it goes out of the node */
1080 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1086 ip6stat.ip6s_fragmented++;
1089 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1091 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1092 RTFREE(ro_pmtu->ro_rt);
1102 #endif /* FAST_IPSEC */
1107 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1108 m_freem(exthdrs.ip6e_dest1);
1109 m_freem(exthdrs.ip6e_rthdr);
1110 m_freem(exthdrs.ip6e_dest2);
1118 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1122 if (hlen > MCLBYTES)
1123 return(ENOBUFS); /* XXX */
1125 MGET(m, MB_DONTWAIT, MT_DATA);
1130 MCLGET(m, MB_DONTWAIT);
1131 if ((m->m_flags & M_EXT) == 0) {
1138 bcopy(hdr, mtod(m, caddr_t), hlen);
1145 * Insert jumbo payload option.
1148 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1154 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1157 * If there is no hop-by-hop options header, allocate new one.
1158 * If there is one but it doesn't have enough space to store the
1159 * jumbo payload option, allocate a cluster to store the whole options.
1160 * Otherwise, use it to store the options.
1162 if (exthdrs->ip6e_hbh == 0) {
1163 MGET(mopt, MB_DONTWAIT, MT_DATA);
1166 mopt->m_len = JUMBOOPTLEN;
1167 optbuf = mtod(mopt, u_char *);
1168 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1169 exthdrs->ip6e_hbh = mopt;
1171 struct ip6_hbh *hbh;
1173 mopt = exthdrs->ip6e_hbh;
1174 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1177 * - exthdrs->ip6e_hbh is not referenced from places
1178 * other than exthdrs.
1179 * - exthdrs->ip6e_hbh is not an mbuf chain.
1181 int oldoptlen = mopt->m_len;
1185 * XXX: give up if the whole (new) hbh header does
1186 * not fit even in an mbuf cluster.
1188 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1192 * As a consequence, we must always prepare a cluster
1195 MGET(n, MB_DONTWAIT, MT_DATA);
1197 MCLGET(n, MB_DONTWAIT);
1198 if ((n->m_flags & M_EXT) == 0) {
1205 n->m_len = oldoptlen + JUMBOOPTLEN;
1206 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1208 optbuf = mtod(n, caddr_t) + oldoptlen;
1210 mopt = exthdrs->ip6e_hbh = n;
1212 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1213 mopt->m_len += JUMBOOPTLEN;
1215 optbuf[0] = IP6OPT_PADN;
1219 * Adjust the header length according to the pad and
1220 * the jumbo payload option.
1222 hbh = mtod(mopt, struct ip6_hbh *);
1223 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1226 /* fill in the option. */
1227 optbuf[2] = IP6OPT_JUMBO;
1229 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1230 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1232 /* finally, adjust the packet header length */
1233 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1240 * Insert fragment header and copy unfragmentable header portions.
1243 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1244 struct ip6_frag **frghdrp)
1246 struct mbuf *n, *mlast;
1248 if (hlen > sizeof(struct ip6_hdr)) {
1249 n = m_copym(m0, sizeof(struct ip6_hdr),
1250 hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1257 /* Search for the last mbuf of unfragmentable part. */
1258 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1261 if ((mlast->m_flags & M_EXT) == 0 &&
1262 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1263 /* use the trailing space of the last mbuf for the fragment hdr */
1265 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1266 mlast->m_len += sizeof(struct ip6_frag);
1267 m->m_pkthdr.len += sizeof(struct ip6_frag);
1269 /* allocate a new mbuf for the fragment header */
1272 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1275 mfrg->m_len = sizeof(struct ip6_frag);
1276 *frghdrp = mtod(mfrg, struct ip6_frag *);
1277 mlast->m_next = mfrg;
1284 * IP6 socket option processing.
1287 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1290 struct inpcb *in6p = so->so_pcb;
1292 int level, op, optname;
1297 level = sopt->sopt_level;
1298 op = sopt->sopt_dir;
1299 optname = sopt->sopt_name;
1300 optlen = sopt->sopt_valsize;
1303 panic("ip6_ctloutput: arg soopt is NULL");
1309 privileged = (td == NULL || suser(td)) ? 0 : 1;
1311 if (level == IPPROTO_IPV6) {
1316 case IPV6_PKTOPTIONS:
1320 error = soopt_getm(sopt, &m); /* XXX */
1323 error = soopt_mcopyin(sopt, m); /* XXX */
1326 error = ip6_pcbopts(&in6p->in6p_outputopts,
1328 m_freem(m); /* XXX */
1333 * Use of some Hop-by-Hop options or some
1334 * Destination options, might require special
1335 * privilege. That is, normal applications
1336 * (without special privilege) might be forbidden
1337 * from setting certain options in outgoing packets,
1338 * and might never see certain options in received
1339 * packets. [RFC 2292 Section 6]
1340 * KAME specific note:
1341 * KAME prevents non-privileged users from sending or
1342 * receiving ANY hbh/dst options in order to avoid
1343 * overhead of parsing options in the kernel.
1345 case IPV6_UNICAST_HOPS:
1350 if (optlen != sizeof(int)) {
1354 error = sooptcopyin(sopt, &optval,
1355 sizeof optval, sizeof optval);
1360 case IPV6_UNICAST_HOPS:
1361 if (optval < -1 || optval >= 256)
1364 /* -1 = kernel default */
1365 in6p->in6p_hops = optval;
1367 if ((in6p->in6p_vflag &
1369 in6p->inp_ip_ttl = optval;
1372 #define OPTSET(bit) \
1375 in6p->in6p_flags |= (bit); \
1377 in6p->in6p_flags &= ~(bit); \
1379 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1382 in6p->in6p_cksum = optval;
1391 * make setsockopt(IPV6_V6ONLY)
1392 * available only prior to bind(2).
1393 * see ipng mailing list, Jun 22 2001.
1395 if (in6p->in6p_lport ||
1396 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1401 OPTSET(IN6P_IPV6_V6ONLY);
1403 in6p->in6p_vflag &= ~INP_IPV4;
1405 in6p->in6p_vflag |= INP_IPV4;
1416 if (optlen != sizeof(int)) {
1420 error = sooptcopyin(sopt, &optval,
1421 sizeof optval, sizeof optval);
1426 OPTSET(IN6P_PKTINFO);
1429 OPTSET(IN6P_HOPLIMIT);
1433 * Check super-user privilege.
1434 * See comments for IPV6_RECVHOPOPTS.
1438 OPTSET(IN6P_HOPOPTS);
1443 OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1452 case IPV6_MULTICAST_IF:
1453 case IPV6_MULTICAST_HOPS:
1454 case IPV6_MULTICAST_LOOP:
1455 case IPV6_JOIN_GROUP:
1456 case IPV6_LEAVE_GROUP:
1459 if (sopt->sopt_valsize > MLEN) {
1464 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1469 m->m_len = sopt->sopt_valsize;
1470 error = sooptcopyin(sopt, mtod(m, char *),
1471 m->m_len, m->m_len);
1472 error = ip6_setmoptions(sopt->sopt_name,
1473 &in6p->in6p_moptions,
1479 case IPV6_PORTRANGE:
1480 error = sooptcopyin(sopt, &optval,
1481 sizeof optval, sizeof optval);
1486 case IPV6_PORTRANGE_DEFAULT:
1487 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1488 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1491 case IPV6_PORTRANGE_HIGH:
1492 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1493 in6p->in6p_flags |= IN6P_HIGHPORT;
1496 case IPV6_PORTRANGE_LOW:
1497 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1498 in6p->in6p_flags |= IN6P_LOWPORT;
1507 #if defined(IPSEC) || defined(FAST_IPSEC)
1508 case IPV6_IPSEC_POLICY:
1514 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1516 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1519 req = mtod(m, caddr_t);
1522 error = ipsec6_set_policy(in6p, optname, req,
1527 #endif /* KAME IPSEC */
1535 struct mbuf **mp = &m;
1537 if (ip6_fw_ctl_ptr == NULL)
1540 if ((error = soopt_getm(sopt, &m)) != 0)
1543 if ((error = soopt_mcopyin(sopt, m)) != 0)
1545 error = (*ip6_fw_ctl_ptr)(optname, mp);
1551 error = ENOPROTOOPT;
1559 case IPV6_PKTOPTIONS:
1560 if (in6p->in6p_options) {
1562 m = m_copym(in6p->in6p_options,
1563 0, M_COPYALL, MB_WAIT);
1564 error = soopt_mcopyout(sopt, m);
1568 sopt->sopt_valsize = 0;
1571 case IPV6_UNICAST_HOPS:
1576 case IPV6_PORTRANGE:
1579 case IPV6_UNICAST_HOPS:
1580 optval = in6p->in6p_hops;
1584 optval = in6p->in6p_cksum;
1588 optval = OPTBIT(IN6P_FAITH);
1592 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1595 case IPV6_PORTRANGE:
1598 flags = in6p->in6p_flags;
1599 if (flags & IN6P_HIGHPORT)
1600 optval = IPV6_PORTRANGE_HIGH;
1601 else if (flags & IN6P_LOWPORT)
1602 optval = IPV6_PORTRANGE_LOW;
1608 error = sooptcopyout(sopt, &optval,
1617 if (optname == IPV6_HOPOPTS ||
1618 optname == IPV6_DSTOPTS ||
1623 optval = OPTBIT(IN6P_PKTINFO);
1626 optval = OPTBIT(IN6P_HOPLIMIT);
1631 optval = OPTBIT(IN6P_HOPOPTS);
1634 optval = OPTBIT(IN6P_RTHDR);
1639 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1642 error = sooptcopyout(sopt, &optval,
1646 case IPV6_MULTICAST_IF:
1647 case IPV6_MULTICAST_HOPS:
1648 case IPV6_MULTICAST_LOOP:
1649 case IPV6_JOIN_GROUP:
1650 case IPV6_LEAVE_GROUP:
1653 error = ip6_getmoptions(sopt->sopt_name,
1654 in6p->in6p_moptions, &m);
1656 error = sooptcopyout(sopt,
1657 mtod(m, char *), m->m_len);
1662 #if defined(IPSEC) || defined(FAST_IPSEC)
1663 case IPV6_IPSEC_POLICY:
1667 struct mbuf *m = NULL;
1668 struct mbuf **mp = &m;
1670 error = soopt_getm(sopt, &m); /* XXX */
1673 error = soopt_mcopyin(sopt, m); /* XXX */
1677 req = mtod(m, caddr_t);
1680 error = ipsec6_get_policy(in6p, req, len, mp);
1682 error = soopt_mcopyout(sopt, m); /*XXX*/
1683 if (error == 0 && m)
1687 #endif /* KAME IPSEC */
1692 struct mbuf **mp = &m;
1694 if (ip6_fw_ctl_ptr == NULL)
1698 error = (*ip6_fw_ctl_ptr)(optname, mp);
1700 error = soopt_mcopyout(sopt, m); /* XXX */
1701 if (error == 0 && m)
1707 error = ENOPROTOOPT;
1719 * Set up IP6 options in pcb for insertion in output packets or
1720 * specifying behavior of outgoing packets.
1723 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
1724 struct sockopt *sopt)
1726 struct ip6_pktopts *opt = *pktopt;
1728 struct thread *td = sopt->sopt_td;
1731 /* turn off any old options. */
1734 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1735 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1736 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1737 printf("ip6_pcbopts: all specified options are cleared.\n");
1739 ip6_clearpktopts(opt, 1, -1);
1741 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1744 if (!m || m->m_len == 0) {
1746 * Only turning off any previous options, regardless of
1747 * whether the opt is just created or given.
1749 free(opt, M_IP6OPT);
1753 /* set options specified by user. */
1756 if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1757 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1758 free(opt, M_IP6OPT);
1766 * initialize ip6_pktopts. beware that there are non-zero default values in
1770 init_ip6pktopts(struct ip6_pktopts *opt)
1773 bzero(opt, sizeof(*opt));
1774 opt->ip6po_hlim = -1; /* -1 means default hop limit */
1778 ip6_clearpktopts(struct ip6_pktopts *pktopt, int needfree, int optname)
1783 if (optname == -1) {
1784 if (needfree && pktopt->ip6po_pktinfo)
1785 free(pktopt->ip6po_pktinfo, M_IP6OPT);
1786 pktopt->ip6po_pktinfo = NULL;
1789 pktopt->ip6po_hlim = -1;
1790 if (optname == -1) {
1791 if (needfree && pktopt->ip6po_nexthop)
1792 free(pktopt->ip6po_nexthop, M_IP6OPT);
1793 pktopt->ip6po_nexthop = NULL;
1795 if (optname == -1) {
1796 if (needfree && pktopt->ip6po_hbh)
1797 free(pktopt->ip6po_hbh, M_IP6OPT);
1798 pktopt->ip6po_hbh = NULL;
1800 if (optname == -1) {
1801 if (needfree && pktopt->ip6po_dest1)
1802 free(pktopt->ip6po_dest1, M_IP6OPT);
1803 pktopt->ip6po_dest1 = NULL;
1805 if (optname == -1) {
1806 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1807 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1808 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1809 if (pktopt->ip6po_route.ro_rt) {
1810 RTFREE(pktopt->ip6po_route.ro_rt);
1811 pktopt->ip6po_route.ro_rt = NULL;
1814 if (optname == -1) {
1815 if (needfree && pktopt->ip6po_dest2)
1816 free(pktopt->ip6po_dest2, M_IP6OPT);
1817 pktopt->ip6po_dest2 = NULL;
1821 #define PKTOPT_EXTHDRCPY(type) \
1825 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1826 dst->type = malloc(hlen, M_IP6OPT, canwait);\
1827 if (dst->type == NULL && canwait == M_NOWAIT)\
1829 bcopy(src->type, dst->type, hlen);\
1833 struct ip6_pktopts *
1834 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
1836 struct ip6_pktopts *dst;
1839 printf("ip6_clearpktopts: invalid argument\n");
1843 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1844 if (dst == NULL && canwait == M_NOWAIT)
1846 bzero(dst, sizeof(*dst));
1848 dst->ip6po_hlim = src->ip6po_hlim;
1849 if (src->ip6po_pktinfo) {
1850 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1852 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1854 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1856 if (src->ip6po_nexthop) {
1857 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1859 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1861 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1862 src->ip6po_nexthop->sa_len);
1864 PKTOPT_EXTHDRCPY(ip6po_hbh);
1865 PKTOPT_EXTHDRCPY(ip6po_dest1);
1866 PKTOPT_EXTHDRCPY(ip6po_dest2);
1867 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1871 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1872 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1873 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1874 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1875 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1876 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1877 free(dst, M_IP6OPT);
1880 #undef PKTOPT_EXTHDRCPY
1883 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1888 ip6_clearpktopts(pktopt, 1, -1);
1890 free(pktopt, M_IP6OPT);
1894 * Set the IP6 multicast options in response to user setsockopt().
1897 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1900 u_int loop, ifindex;
1901 struct ipv6_mreq *mreq;
1903 struct ip6_moptions *im6o = *im6op;
1904 struct route_in6 ro;
1905 struct sockaddr_in6 *dst;
1906 struct in6_multi_mship *imm;
1907 struct thread *td = curthread; /* XXX */
1911 * No multicast option buffer attached to the pcb;
1912 * allocate one and initialize to default values.
1914 im6o = (struct ip6_moptions *)
1915 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1920 im6o->im6o_multicast_ifp = NULL;
1921 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1922 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1923 LIST_INIT(&im6o->im6o_memberships);
1928 case IPV6_MULTICAST_IF:
1930 * Select the interface for outgoing multicast packets.
1932 if (m == NULL || m->m_len != sizeof(u_int)) {
1936 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1937 if (ifindex < 0 || if_index < ifindex) {
1938 error = ENXIO; /* XXX EINVAL? */
1941 ifp = ifindex2ifnet[ifindex];
1942 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1943 error = EADDRNOTAVAIL;
1946 im6o->im6o_multicast_ifp = ifp;
1949 case IPV6_MULTICAST_HOPS:
1952 * Set the IP6 hoplimit for outgoing multicast packets.
1955 if (m == NULL || m->m_len != sizeof(int)) {
1959 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1960 if (optval < -1 || optval >= 256)
1962 else if (optval == -1)
1963 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1965 im6o->im6o_multicast_hlim = optval;
1969 case IPV6_MULTICAST_LOOP:
1971 * Set the loopback flag for outgoing multicast packets.
1972 * Must be zero or one.
1974 if (m == NULL || m->m_len != sizeof(u_int)) {
1978 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1983 im6o->im6o_multicast_loop = loop;
1986 case IPV6_JOIN_GROUP:
1988 * Add a multicast group membership.
1989 * Group must be a valid IP6 multicast address.
1991 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1995 mreq = mtod(m, struct ipv6_mreq *);
1996 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1998 * We use the unspecified address to specify to accept
1999 * all multicast addresses. Only super user is allowed
2007 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2013 * If the interface is specified, validate it.
2015 if (mreq->ipv6mr_interface < 0
2016 || if_index < mreq->ipv6mr_interface) {
2017 error = ENXIO; /* XXX EINVAL? */
2021 * If no interface was explicitly specified, choose an
2022 * appropriate one according to the given multicast address.
2024 if (mreq->ipv6mr_interface == 0) {
2026 * If the multicast address is in node-local scope,
2027 * the interface should be a loopback interface.
2028 * Otherwise, look up the routing table for the
2029 * address, and choose the outgoing interface.
2030 * XXX: is it a good approach?
2032 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2036 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2037 bzero(dst, sizeof(*dst));
2038 dst->sin6_len = sizeof(struct sockaddr_in6);
2039 dst->sin6_family = AF_INET6;
2040 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2041 rtalloc((struct route *)&ro);
2042 if (ro.ro_rt == NULL) {
2043 error = EADDRNOTAVAIL;
2046 ifp = ro.ro_rt->rt_ifp;
2050 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2053 * See if we found an interface, and confirm that it
2054 * supports multicast
2056 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2057 error = EADDRNOTAVAIL;
2061 * Put interface index into the multicast address,
2062 * if the address has link-local scope.
2064 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2065 mreq->ipv6mr_multiaddr.s6_addr16[1]
2066 = htons(mreq->ipv6mr_interface);
2069 * See if the membership already exists.
2071 for (imm = im6o->im6o_memberships.lh_first;
2072 imm != NULL; imm = imm->i6mm_chain.le_next)
2073 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2074 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2075 &mreq->ipv6mr_multiaddr))
2082 * Everything looks good; add a new record to the multicast
2083 * address list for the given interface.
2085 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2090 if ((imm->i6mm_maddr =
2091 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2092 free(imm, M_IPMADDR);
2095 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2098 case IPV6_LEAVE_GROUP:
2100 * Drop a multicast group membership.
2101 * Group must be a valid IP6 multicast address.
2103 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2107 mreq = mtod(m, struct ipv6_mreq *);
2108 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2113 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2118 * If an interface address was specified, get a pointer
2119 * to its ifnet structure.
2121 if (mreq->ipv6mr_interface < 0
2122 || if_index < mreq->ipv6mr_interface) {
2123 error = ENXIO; /* XXX EINVAL? */
2126 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2128 * Put interface index into the multicast address,
2129 * if the address has link-local scope.
2131 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2132 mreq->ipv6mr_multiaddr.s6_addr16[1]
2133 = htons(mreq->ipv6mr_interface);
2136 * Find the membership in the membership list.
2138 for (imm = im6o->im6o_memberships.lh_first;
2139 imm != NULL; imm = imm->i6mm_chain.le_next) {
2141 imm->i6mm_maddr->in6m_ifp == ifp) &&
2142 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2143 &mreq->ipv6mr_multiaddr))
2147 /* Unable to resolve interface */
2148 error = EADDRNOTAVAIL;
2152 * Give up the multicast address record to which the
2153 * membership points.
2155 LIST_REMOVE(imm, i6mm_chain);
2156 in6_delmulti(imm->i6mm_maddr);
2157 free(imm, M_IPMADDR);
2166 * If all options have default values, no need to keep the mbuf.
2168 if (im6o->im6o_multicast_ifp == NULL &&
2169 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2170 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2171 im6o->im6o_memberships.lh_first == NULL) {
2172 free(*im6op, M_IPMOPTS);
2180 * Return the IP6 multicast options in response to user getsockopt().
2183 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2185 u_int *hlim, *loop, *ifindex;
2187 *mp = m_get(MB_WAIT, MT_HEADER); /* XXX */
2191 case IPV6_MULTICAST_IF:
2192 ifindex = mtod(*mp, u_int *);
2193 (*mp)->m_len = sizeof(u_int);
2194 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2197 *ifindex = im6o->im6o_multicast_ifp->if_index;
2200 case IPV6_MULTICAST_HOPS:
2201 hlim = mtod(*mp, u_int *);
2202 (*mp)->m_len = sizeof(u_int);
2204 *hlim = ip6_defmcasthlim;
2206 *hlim = im6o->im6o_multicast_hlim;
2209 case IPV6_MULTICAST_LOOP:
2210 loop = mtod(*mp, u_int *);
2211 (*mp)->m_len = sizeof(u_int);
2213 *loop = ip6_defmcasthlim;
2215 *loop = im6o->im6o_multicast_loop;
2224 * Discard the IP6 multicast options.
2227 ip6_freemoptions(struct ip6_moptions *im6o)
2229 struct in6_multi_mship *imm;
2234 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2235 LIST_REMOVE(imm, i6mm_chain);
2236 if (imm->i6mm_maddr)
2237 in6_delmulti(imm->i6mm_maddr);
2238 free(imm, M_IPMADDR);
2240 free(im6o, M_IPMOPTS);
2244 * Set IPv6 outgoing packet options based on advanced API.
2247 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, int priv,
2250 struct cmsghdr *cm = 0;
2252 if (control == 0 || opt == 0)
2255 init_ip6pktopts(opt);
2258 * XXX: Currently, we assume all the optional information is stored
2261 if (control->m_next)
2264 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2265 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2266 cm = mtod(control, struct cmsghdr *);
2267 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2269 if (cm->cmsg_level != IPPROTO_IPV6)
2273 * XXX should check if RFC2292 API is mixed with 2292bis API
2275 switch (cm->cmsg_type) {
2277 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2280 /* XXX: Is it really WAITOK? */
2281 opt->ip6po_pktinfo =
2282 malloc(sizeof(struct in6_pktinfo),
2283 M_IP6OPT, M_WAITOK);
2284 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2285 sizeof(struct in6_pktinfo));
2287 opt->ip6po_pktinfo =
2288 (struct in6_pktinfo *)CMSG_DATA(cm);
2289 if (opt->ip6po_pktinfo->ipi6_ifindex &&
2290 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2291 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2292 htons(opt->ip6po_pktinfo->ipi6_ifindex);
2294 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2295 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2300 * Check if the requested source address is indeed a
2301 * unicast address assigned to the node, and can be
2302 * used as the packet's source address.
2304 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2305 struct in6_ifaddr *ia6;
2306 struct sockaddr_in6 sin6;
2308 bzero(&sin6, sizeof(sin6));
2309 sin6.sin6_len = sizeof(sin6);
2310 sin6.sin6_family = AF_INET6;
2312 opt->ip6po_pktinfo->ipi6_addr;
2313 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2315 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2316 IN6_IFF_NOTREADY)) != 0)
2317 return(EADDRNOTAVAIL);
2322 if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2325 opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2326 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2334 if (cm->cmsg_len < sizeof(u_char) ||
2335 /* check if cmsg_len is large enough for sa_len */
2336 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2340 opt->ip6po_nexthop =
2341 malloc(*CMSG_DATA(cm),
2342 M_IP6OPT, M_WAITOK);
2343 bcopy(CMSG_DATA(cm),
2347 opt->ip6po_nexthop =
2348 (struct sockaddr *)CMSG_DATA(cm);
2353 struct ip6_hbh *hbh;
2356 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2358 hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2359 hbhlen = (hbh->ip6h_len + 1) << 3;
2360 if (cm->cmsg_len != CMSG_LEN(hbhlen))
2365 malloc(hbhlen, M_IP6OPT, M_WAITOK);
2366 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2368 opt->ip6po_hbh = hbh;
2374 struct ip6_dest *dest, **newdest;
2377 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2379 dest = (struct ip6_dest *)CMSG_DATA(cm);
2380 destlen = (dest->ip6d_len + 1) << 3;
2381 if (cm->cmsg_len != CMSG_LEN(destlen))
2385 * The old advacned API is ambiguous on this
2386 * point. Our approach is to determine the
2387 * position based according to the existence
2388 * of a routing header. Note, however, that
2389 * this depends on the order of the extension
2390 * headers in the ancillary data; the 1st part
2391 * of the destination options header must
2392 * appear before the routing header in the
2393 * ancillary data, too.
2394 * RFC2292bis solved the ambiguity by
2395 * introducing separate cmsg types.
2397 if (opt->ip6po_rthdr == NULL)
2398 newdest = &opt->ip6po_dest1;
2400 newdest = &opt->ip6po_dest2;
2403 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2404 bcopy(dest, *newdest, destlen);
2413 struct ip6_rthdr *rth;
2416 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2418 rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2419 rthlen = (rth->ip6r_len + 1) << 3;
2420 if (cm->cmsg_len != CMSG_LEN(rthlen))
2423 switch (rth->ip6r_type) {
2424 case IPV6_RTHDR_TYPE_0:
2425 /* must contain one addr */
2426 if (rth->ip6r_len == 0)
2428 /* length must be even */
2429 if (rth->ip6r_len % 2)
2431 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2435 return(EINVAL); /* not supported */
2439 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2441 bcopy(rth, opt->ip6po_rthdr, rthlen);
2443 opt->ip6po_rthdr = rth;
2449 return(ENOPROTOOPT);
2457 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2458 * packet to the input queue of a specified interface. Note that this
2459 * calls the output routine of the loopback "driver", but with an interface
2460 * pointer that might NOT be &loif -- easier than replicating that code here.
2463 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2466 struct ip6_hdr *ip6;
2468 copym = m_copy(m, 0, M_COPYALL);
2473 * Make sure to deep-copy IPv6 header portion in case the data
2474 * is in an mbuf cluster, so that we can safely override the IPv6
2475 * header portion later.
2477 if ((copym->m_flags & M_EXT) != 0 ||
2478 copym->m_len < sizeof(struct ip6_hdr)) {
2479 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2485 if (copym->m_len < sizeof(*ip6)) {
2491 ip6 = mtod(copym, struct ip6_hdr *);
2493 * clear embedded scope identifiers if necessary.
2494 * in6_clearscope will touch the addresses only when necessary.
2496 in6_clearscope(&ip6->ip6_src);
2497 in6_clearscope(&ip6->ip6_dst);
2499 (void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2503 * Chop IPv6 header off from the payload.
2506 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2509 struct ip6_hdr *ip6;
2511 ip6 = mtod(m, struct ip6_hdr *);
2512 if (m->m_len > sizeof(*ip6)) {
2513 MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
2518 M_MOVE_PKTHDR(mh, m);
2519 MH_ALIGN(mh, sizeof(*ip6));
2520 m->m_len -= sizeof(*ip6);
2521 m->m_data += sizeof(*ip6);
2524 m->m_len = sizeof(*ip6);
2525 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2527 exthdrs->ip6e_ip6 = m;
2532 * Compute IPv6 extension header length.
2535 ip6_optlen(struct in6pcb *in6p)
2539 if (!in6p->in6p_outputopts)
2544 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2546 len += elen(in6p->in6p_outputopts->ip6po_hbh);
2547 if (in6p->in6p_outputopts->ip6po_rthdr)
2548 /* dest1 is valid with rthdr only */
2549 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2550 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2551 len += elen(in6p->in6p_outputopts->ip6po_dest2);