1 /* $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $ */
2 /* $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * Copyright (c) 1982, 1986, 1988, 1990, 1993
35 * The Regents of the University of California. All rights reserved.
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. All advertising materials mentioning features or use of this software
46 * must display the following acknowledgement:
47 * This product includes software developed by the University of
48 * California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
68 #include "opt_ip6fw.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
73 #include <sys/param.h>
74 #include <sys/malloc.h>
76 #include <sys/errno.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/systm.h>
81 #include <sys/kernel.h>
85 #include <sys/thread2.h>
86 #include <sys/msgport2.h>
89 #include <net/route.h>
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet6/in6_var.h>
95 #include <netinet/ip6.h>
96 #include <netinet/icmp6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/ip6protosw.h>
103 #include <netinet6/ipsec.h>
105 #include <netinet6/ipsec6.h>
107 #include <netproto/key/key.h>
111 #include <netproto/ipsec/ipsec.h>
112 #include <netproto/ipsec/ipsec6.h>
113 #include <netproto/ipsec/key.h>
116 #include <net/ip6fw/ip6_fw.h>
118 #include <net/net_osdep.h>
120 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
123 struct mbuf *ip6e_ip6;
124 struct mbuf *ip6e_hbh;
125 struct mbuf *ip6e_dest1;
126 struct mbuf *ip6e_rthdr;
127 struct mbuf *ip6e_dest2;
130 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
131 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
133 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
134 struct socket *, struct sockopt *);
135 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
136 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
137 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
138 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
139 struct ifnet *, struct in6_addr *, u_long *, int *);
140 static int copyexthdr (void *, struct mbuf **);
141 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
143 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
144 static struct mbuf *ip6_splithdr (struct mbuf *);
145 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
148 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
149 * header (with pri, len, nxt, hlim, src, dst).
150 * This function may modify ver and hlim only.
151 * The mbuf chain containing the packet will be freed.
152 * The mbuf opt, if present, will not be freed.
154 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
155 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
156 * which is rt_rmx.rmx_mtu.
159 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
160 int flags, struct ip6_moptions *im6o,
161 struct ifnet **ifpp, /* XXX: just for statistics */
164 struct ip6_hdr *ip6, *mhip6;
165 struct ifnet *ifp, *origifp;
169 int hlen, tlen, len, off;
170 struct route_in6 ip6route;
171 struct sockaddr_in6 *dst;
173 struct in6_ifaddr *ia = NULL;
175 int alwaysfrag, dontfrag;
176 u_int32_t optlen, plen = 0, unfragpartlen;
177 struct ip6_exthdrs exthdrs;
178 struct in6_addr finaldst;
179 struct route_in6 *ro_pmtu = NULL;
180 boolean_t hdrsplit = FALSE;
181 boolean_t needipsec = FALSE;
183 boolean_t needipsectun = FALSE;
184 struct secpolicy *sp = NULL;
185 struct socket *so = inp ? inp->inp_socket : NULL;
187 ip6 = mtod(m, struct ip6_hdr *);
190 boolean_t needipsectun = FALSE;
191 struct secpolicy *sp = NULL;
193 ip6 = mtod(m, struct ip6_hdr *);
196 bzero(&exthdrs, sizeof exthdrs);
199 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
201 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
203 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
205 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
210 /* get a security policy for this packet */
212 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
214 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
217 ipsec6stat.out_inval++;
224 switch (sp->policy) {
225 case IPSEC_POLICY_DISCARD:
227 * This packet is just discarded.
229 ipsec6stat.out_polvio++;
232 case IPSEC_POLICY_BYPASS:
233 case IPSEC_POLICY_NONE:
234 /* no need to do IPsec. */
238 case IPSEC_POLICY_IPSEC:
239 if (sp->req == NULL) {
240 error = key_spdacquire(sp); /* acquire a policy */
246 case IPSEC_POLICY_ENTRUST:
248 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
252 /* get a security policy for this packet */
254 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
256 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
259 newipsecstat.ips_out_inval++;
266 switch (sp->policy) {
267 case IPSEC_POLICY_DISCARD:
269 * This packet is just discarded.
271 newipsecstat.ips_out_polvio++;
274 case IPSEC_POLICY_BYPASS:
275 case IPSEC_POLICY_NONE:
276 /* no need to do IPsec. */
280 case IPSEC_POLICY_IPSEC:
281 if (sp->req == NULL) {
282 error = key_spdacquire(sp); /* acquire a policy */
288 case IPSEC_POLICY_ENTRUST:
290 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
292 #endif /* FAST_IPSEC */
295 * Calculate the total length of the extension header chain.
296 * Keep the length of the unfragmentable part for fragmentation.
298 optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
299 m_lengthm(exthdrs.ip6e_dest1, NULL) +
300 m_lengthm(exthdrs.ip6e_rthdr, NULL);
302 unfragpartlen = optlen + sizeof(struct ip6_hdr);
304 /* NOTE: we don't add AH/ESP length here. do that later. */
305 optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
308 * If we need IPsec, or there is at least one extension header,
309 * separate IP6 header from the payload.
311 if ((needipsec || optlen) && !hdrsplit) {
312 exthdrs.ip6e_ip6 = ip6_splithdr(m);
313 if (exthdrs.ip6e_ip6 == NULL) {
317 m = exthdrs.ip6e_ip6;
322 ip6 = mtod(m, struct ip6_hdr *);
324 /* adjust mbuf packet header length */
325 m->m_pkthdr.len += optlen;
326 plen = m->m_pkthdr.len - sizeof(*ip6);
328 /* If this is a jumbo payload, insert a jumbo payload option. */
329 if (plen > IPV6_MAXPACKET) {
331 exthdrs.ip6e_ip6 = ip6_splithdr(m);
332 if (exthdrs.ip6e_ip6 == NULL) {
336 m = exthdrs.ip6e_ip6;
340 ip6 = mtod(m, struct ip6_hdr *);
341 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
345 ip6->ip6_plen = htons(plen);
348 * Concatenate headers and fill in next header fields.
349 * Here we have, on "m"
351 * and we insert headers accordingly. Finally, we should be getting:
352 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
354 * during the header composing process, "m" points to IPv6 header.
355 * "mprev" points to an extension header prior to esp.
358 nexthdrp = &ip6->ip6_nxt;
362 * we treat dest2 specially. this makes IPsec processing
363 * much easier. the goal here is to make mprev point the
364 * mbuf prior to dest2.
366 * result: IPv6 dest2 payload
367 * m and mprev will point to IPv6 header.
369 if (exthdrs.ip6e_dest2) {
371 panic("assumption failed: hdr not split");
372 exthdrs.ip6e_dest2->m_next = m->m_next;
373 m->m_next = exthdrs.ip6e_dest2;
374 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
375 ip6->ip6_nxt = IPPROTO_DSTOPTS;
379 * Place m1 after mprev.
381 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
385 panic("assumption failed: hdr not split");\
386 *mtod(m1, u_char *) = *nexthdrp;\
388 nexthdrp = mtod(m1, u_char *);\
389 m1->m_next = mprev->m_next;\
396 * result: IPv6 hbh dest1 rthdr dest2 payload
397 * m will point to IPv6 header. mprev will point to the
398 * extension header prior to dest2 (rthdr in the above case).
400 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
401 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
402 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
404 #if defined(IPSEC) || defined(FAST_IPSEC)
406 struct ipsec_output_state state;
408 struct ip6_rthdr *rh = NULL;
411 * pointers after IPsec headers are not valid any more.
412 * other pointers need a great care too.
413 * (IPsec routines should not mangle mbufs prior to AH/ESP)
415 exthdrs.ip6e_dest2 = NULL;
417 if (exthdrs.ip6e_rthdr) {
418 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
419 segleft_org = rh->ip6r_segleft;
420 rh->ip6r_segleft = 0;
423 bzero(&state, sizeof state);
425 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
429 /* mbuf is already reclaimed in ipsec6_output_trans. */
439 kprintf("ip6_output (ipsec): error code %d\n",
443 /* don't show these error codes to the user */
449 if (exthdrs.ip6e_rthdr) {
450 /* ah6_output doesn't modify mbuf chain */
451 rh->ip6r_segleft = segleft_org;
457 * If there is a routing header, replace destination address field
458 * with the first hop of the routing header.
460 if (exthdrs.ip6e_rthdr) {
461 struct ip6_rthdr *rh;
463 finaldst = ip6->ip6_dst;
464 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
465 switch (rh->ip6r_type) {
466 default: /* is it possible? */
472 /* Source address validation */
473 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
474 !(flags & IPV6_DADOUTPUT)) {
476 ip6stat.ip6s_badscope++;
479 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
481 ip6stat.ip6s_badscope++;
485 ip6stat.ip6s_localout++;
492 bzero(ro, sizeof(*ro));
495 if (opt && opt->ip6po_rthdr)
496 ro = &opt->ip6po_route;
497 dst = (struct sockaddr_in6 *)&ro->ro_dst;
499 * If there is a cached route,
500 * check that it is to the same destination
501 * and is still up. If not, free it and try again.
503 if (ro->ro_rt != NULL &&
504 (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
505 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
509 if (ro->ro_rt == NULL) {
510 bzero(dst, sizeof(*dst));
511 dst->sin6_family = AF_INET6;
512 dst->sin6_len = sizeof(struct sockaddr_in6);
513 dst->sin6_addr = ip6->ip6_dst;
515 #if defined(IPSEC) || defined(FAST_IPSEC)
516 if (needipsec && needipsectun) {
517 struct ipsec_output_state state;
520 * All the extension headers will become inaccessible
521 * (since they can be encrypted).
522 * Don't panic, we need no more updates to extension headers
523 * on inner IPv6 packet (since they are now encapsulated).
525 * IPv6 [ESP|AH] IPv6 [extension headers] payload
527 bzero(&exthdrs, sizeof(exthdrs));
528 exthdrs.ip6e_ip6 = m;
530 bzero(&state, sizeof(state));
532 state.ro = (struct route *)ro;
533 state.dst = (struct sockaddr *)dst;
535 error = ipsec6_output_tunnel(&state, sp, flags);
538 ro = (struct route_in6 *)state.ro;
539 dst = (struct sockaddr_in6 *)state.dst;
541 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
552 kprintf("ip6_output (ipsec): error code %d\n", error);
555 /* don't show these error codes to the user */
562 exthdrs.ip6e_ip6 = m;
566 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
569 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
570 #define sin6tosa(sin6) ((struct sockaddr *)(sin6))
572 * interface selection comes here
573 * if an interface is specified from an upper layer,
576 if (ro->ro_rt == NULL) {
578 * non-bsdi always clone routes, if parent is
581 rtalloc((struct route *)ro);
583 if (ro->ro_rt == NULL) {
584 ip6stat.ip6s_noroute++;
585 error = EHOSTUNREACH;
586 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
589 ia = ifatoia6(ro->ro_rt->rt_ifa);
590 ifp = ro->ro_rt->rt_ifp;
592 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
593 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
594 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
596 in6_ifstat_inc(ifp, ifs6_out_request);
599 * Check if the outgoing interface conflicts with
600 * the interface specified by ifi6_ifindex (if specified).
601 * Note that loopback interface is always okay.
602 * (this may happen when we are sending a packet to one of
603 * our own addresses.)
605 if (opt && opt->ip6po_pktinfo
606 && opt->ip6po_pktinfo->ipi6_ifindex) {
607 if (!(ifp->if_flags & IFF_LOOPBACK)
608 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
609 ip6stat.ip6s_noroute++;
610 in6_ifstat_inc(ifp, ifs6_out_discard);
611 error = EHOSTUNREACH;
616 if (opt && opt->ip6po_hlim != -1)
617 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
620 struct in6_multi *in6m;
622 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
625 * See if the caller provided any multicast options
629 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
630 if (im6o->im6o_multicast_ifp != NULL)
631 ifp = im6o->im6o_multicast_ifp;
633 ip6->ip6_hlim = ip6_defmcasthlim;
636 * See if the caller provided the outgoing interface
637 * as an ancillary data.
638 * Boundary check for ifindex is assumed to be already done.
640 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
641 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
644 * If the destination is a node-local scope multicast,
645 * the packet should be loop-backed only.
647 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
649 * If the outgoing interface is already specified,
650 * it should be a loopback interface.
652 if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
653 ip6stat.ip6s_badscope++;
654 error = ENETUNREACH; /* XXX: better error? */
655 /* XXX correct ifp? */
656 in6_ifstat_inc(ifp, ifs6_out_discard);
663 if (opt && opt->ip6po_hlim != -1)
664 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
667 * If caller did not provide an interface lookup a
668 * default in the routing table. This is either a
669 * default for the speicfied group (i.e. a host
670 * route), or a multicast default (a route for the
674 if (ro->ro_rt == NULL) {
676 rtpurelookup((struct sockaddr *)&ro->ro_dst);
678 if (ro->ro_rt == NULL) {
679 ip6stat.ip6s_noroute++;
680 error = EHOSTUNREACH;
681 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
684 ia = ifatoia6(ro->ro_rt->rt_ifa);
685 ifp = ro->ro_rt->rt_ifp;
689 if (!(flags & IPV6_FORWARDING))
690 in6_ifstat_inc(ifp, ifs6_out_request);
691 in6_ifstat_inc(ifp, ifs6_out_mcast);
694 * Confirm that the outgoing interface supports multicast.
696 if (!(ifp->if_flags & IFF_MULTICAST)) {
697 ip6stat.ip6s_noroute++;
698 in6_ifstat_inc(ifp, ifs6_out_discard);
702 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
704 (im6o == NULL || im6o->im6o_multicast_loop)) {
706 * If we belong to the destination multicast group
707 * on the outgoing interface, and the caller did not
708 * forbid loopback, loop back a copy.
710 ip6_mloopback(ifp, m, dst);
713 * If we are acting as a multicast router, perform
714 * multicast forwarding as if the packet had just
715 * arrived on the interface to which we are about
716 * to send. The multicast forwarding function
717 * recursively calls this function, using the
718 * IPV6_FORWARDING flag to prevent infinite recursion.
720 * Multicasts that are looped back by ip6_mloopback(),
721 * above, will be forwarded by the ip6_input() routine,
724 if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
725 if (ip6_mforward(ip6, ifp, m) != 0) {
732 * Multicasts with a hoplimit of zero may be looped back,
733 * above, but must not be transmitted on a network.
734 * Also, multicasts addressed to the loopback interface
735 * are not sent -- the above call to ip6_mloopback() will
736 * loop back a copy if this host actually belongs to the
737 * destination group on the loopback interface.
739 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
746 * Fill the outgoing inteface to tell the upper layer
747 * to increment per-interface statistics.
752 /* Determine path MTU. */
753 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
758 * The caller of this function may specify to use the minimum MTU
760 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
761 * setting. The logic is a bit complicated; by default, unicast
762 * packets will follow path MTU while multicast packets will be sent at
763 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
764 * including unicast ones will be sent at the minimum MTU. Multicast
765 * packets will always be sent at the minimum MTU unless
766 * IP6PO_MINMTU_DISABLE is explicitly specified.
767 * See RFC 3542 for more details.
769 if (mtu > IPV6_MMTU) {
770 if ((flags & IPV6_MINMTU))
772 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
774 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
776 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
781 /* Fake scoped addresses */
782 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
784 * If source or destination address is a scoped address, and
785 * the packet is going to be sent to a loopback interface,
786 * we should keep the original interface.
790 * XXX: this is a very experimental and temporary solution.
791 * We eventually have sockaddr_in6 and use the sin6_scope_id
792 * field of the structure here.
793 * We rely on the consistency between two scope zone ids
794 * of source and destination, which should already be assured.
795 * Larger scopes than link will be supported in the future.
798 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
799 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
800 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
801 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
803 * XXX: origifp can be NULL even in those two cases above.
804 * For example, if we remove the (only) link-local address
805 * from the loopback interface, and try to send a link-local
806 * address without link-id information. Then the source
807 * address is ::1, and the destination address is the
808 * link-local address with its s6_addr16[1] being zero.
809 * What is worse, if the packet goes to the loopback interface
810 * by a default rejected route, the null pointer would be
811 * passed to looutput, and the kernel would hang.
812 * The following last resort would prevent such disaster.
820 * clear embedded scope identifiers if necessary.
821 * in6_clearscope will touch the addresses only when necessary.
823 in6_clearscope(&ip6->ip6_src);
824 in6_clearscope(&ip6->ip6_dst);
827 * Check with the firewall...
829 if (ip6_fw_enable && ip6_fw_chk_ptr) {
832 m->m_pkthdr.rcvif = NULL; /* XXX */
833 /* If ipfw says divert, we have to just drop packet */
834 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
845 * If the outgoing packet contains a hop-by-hop options header,
846 * it must be examined and processed even by the source node.
847 * (RFC 2460, section 4.)
849 if (exthdrs.ip6e_hbh) {
850 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
851 u_int32_t dummy1; /* XXX unused */
852 u_int32_t dummy2; /* XXX unused */
855 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
856 panic("ip6e_hbh is not continuous");
859 * XXX: if we have to send an ICMPv6 error to the sender,
860 * we need the M_LOOP flag since icmp6_error() expects
861 * the IPv6 and the hop-by-hop options header are
862 * continuous unless the flag is set.
864 m->m_flags |= M_LOOP;
865 m->m_pkthdr.rcvif = ifp;
866 if (ip6_process_hopopts(m,
867 (u_int8_t *)(hbh + 1),
868 ((hbh->ip6h_len + 1) << 3) -
869 sizeof(struct ip6_hbh),
870 &dummy1, &dummy2) < 0) {
871 /* m was already freed at this point */
872 error = EINVAL;/* better error? */
875 m->m_flags &= ~M_LOOP; /* XXX */
876 m->m_pkthdr.rcvif = NULL;
880 * Run through list of hooks for output packets.
882 if (pfil_has_hooks(&inet6_pfil_hook)) {
883 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
884 if (error != 0 || m == NULL)
886 ip6 = mtod(m, struct ip6_hdr *);
890 * Send the packet to the outgoing interface.
891 * If necessary, do IPv6 fragmentation before sending.
893 * the logic here is rather complex:
894 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
895 * 1-a: send as is if tlen <= path mtu
896 * 1-b: fragment if tlen > path mtu
898 * 2: if user asks us not to fragment (dontfrag == 1)
899 * 2-a: send as is if tlen <= interface mtu
900 * 2-b: error if tlen > interface mtu
902 * 3: if we always need to attach fragment header (alwaysfrag == 1)
905 * 4: if dontfrag == 1 && alwaysfrag == 1
906 * error, as we cannot handle this conflicting request
908 tlen = m->m_pkthdr.len;
910 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
914 if (dontfrag && alwaysfrag) { /* case 4 */
915 /* conflicting request - can't transmit */
919 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */
921 * Even if the DONTFRAG option is specified, we cannot send the
922 * packet when the data length is larger than the MTU of the
923 * outgoing interface.
924 * Notify the error by sending IPV6_PATHMTU ancillary data as
925 * well as returning an error code (the latter is not described
929 struct ip6ctlparam ip6cp;
931 mtu32 = (u_int32_t)mtu;
932 bzero(&ip6cp, sizeof(ip6cp));
933 ip6cp.ip6c_cmdarg = (void *)&mtu32;
934 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
942 * transmit packet without fragmentation
944 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
945 struct in6_ifaddr *ia6;
947 ip6 = mtod(m, struct ip6_hdr *);
948 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
950 /* Record statistics for this interface address. */
951 IFA_STAT_INC(&ia6->ia_ifa, opackets, 1);
952 IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len);
955 /* clean ipsec history once it goes out of the node */
958 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
963 * try to fragment the packet. case 1-b and 3
965 if (mtu < IPV6_MMTU) {
967 * note that path MTU is never less than IPV6_MMTU
971 in6_ifstat_inc(ifp, ifs6_out_fragfail);
973 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
975 in6_ifstat_inc(ifp, ifs6_out_fragfail);
978 struct mbuf **mnext, *m_frgpart;
979 struct ip6_frag *ip6f;
980 u_int32_t id = htonl(ip6_id++);
984 * Too large for the destination or interface;
985 * fragment if possible.
986 * Must be able to put at least 8 bytes per fragment.
988 hlen = unfragpartlen;
989 if (mtu > IPV6_MAXPACKET)
990 mtu = IPV6_MAXPACKET;
992 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
995 in6_ifstat_inc(ifp, ifs6_out_fragfail);
999 mnext = &m->m_nextpkt;
1002 * Change the next header field of the last header in the
1003 * unfragmentable part.
1005 if (exthdrs.ip6e_rthdr) {
1006 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1007 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1008 } else if (exthdrs.ip6e_dest1) {
1009 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1010 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1011 } else if (exthdrs.ip6e_hbh) {
1012 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1013 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1015 nextproto = ip6->ip6_nxt;
1016 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1020 * Loop through length of segment after first fragment,
1021 * make new header and copy data of each part and link onto
1025 for (off = hlen; off < tlen; off += len) {
1026 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1029 ip6stat.ip6s_odropped++;
1032 m->m_pkthdr.rcvif = NULL;
1033 m->m_flags = m0->m_flags & M_COPYFLAGS;
1035 mnext = &m->m_nextpkt;
1036 m->m_data += max_linkhdr;
1037 mhip6 = mtod(m, struct ip6_hdr *);
1039 m->m_len = sizeof(*mhip6);
1040 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1042 ip6stat.ip6s_odropped++;
1045 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1046 if (off + len >= tlen)
1049 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1050 mhip6->ip6_plen = htons((u_short)(len + hlen +
1052 sizeof(struct ip6_hdr)));
1053 if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1055 ip6stat.ip6s_odropped++;
1058 m_cat(m, m_frgpart);
1059 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1060 m->m_pkthdr.rcvif = NULL;
1061 ip6f->ip6f_reserved = 0;
1062 ip6f->ip6f_ident = id;
1063 ip6f->ip6f_nxt = nextproto;
1064 ip6stat.ip6s_ofragments++;
1065 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1068 in6_ifstat_inc(ifp, ifs6_out_fragok);
1072 * Remove leading garbages.
1076 m0->m_nextpkt = NULL;
1078 for (m0 = m; m; m = m0) {
1080 m->m_nextpkt = NULL;
1082 /* Record statistics for this interface address. */
1084 IFA_STAT_INC(&ia->ia_ifa, opackets, 1);
1085 IFA_STAT_INC(&ia->ia_ifa, obytes,
1089 /* clean ipsec history once it goes out of the node */
1092 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1098 ip6stat.ip6s_fragmented++;
1101 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1103 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1104 RTFREE(ro_pmtu->ro_rt);
1119 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1120 m_freem(exthdrs.ip6e_dest1);
1121 m_freem(exthdrs.ip6e_rthdr);
1122 m_freem(exthdrs.ip6e_dest2);
1130 copyexthdr(void *h, struct mbuf **mp)
1132 struct ip6_ext *hdr = h;
1139 hlen = (hdr->ip6e_len + 1) * 8;
1140 if (hlen > MCLBYTES)
1141 return ENOBUFS; /* XXX */
1143 m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1148 bcopy(hdr, mtod(m, caddr_t), hlen);
1155 * Insert jumbo payload option.
1158 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1164 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1167 * If there is no hop-by-hop options header, allocate new one.
1168 * If there is one but it doesn't have enough space to store the
1169 * jumbo payload option, allocate a cluster to store the whole options.
1170 * Otherwise, use it to store the options.
1172 if (exthdrs->ip6e_hbh == NULL) {
1173 MGET(mopt, MB_DONTWAIT, MT_DATA);
1176 mopt->m_len = JUMBOOPTLEN;
1177 optbuf = mtod(mopt, u_char *);
1178 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1179 exthdrs->ip6e_hbh = mopt;
1181 struct ip6_hbh *hbh;
1183 mopt = exthdrs->ip6e_hbh;
1184 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1187 * - exthdrs->ip6e_hbh is not referenced from places
1188 * other than exthdrs.
1189 * - exthdrs->ip6e_hbh is not an mbuf chain.
1191 int oldoptlen = mopt->m_len;
1195 * XXX: give up if the whole (new) hbh header does
1196 * not fit even in an mbuf cluster.
1198 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1202 * As a consequence, we must always prepare a cluster
1205 n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1208 n->m_len = oldoptlen + JUMBOOPTLEN;
1209 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1210 optbuf = mtod(n, caddr_t) + oldoptlen;
1212 mopt = exthdrs->ip6e_hbh = n;
1214 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1215 mopt->m_len += JUMBOOPTLEN;
1217 optbuf[0] = IP6OPT_PADN;
1221 * Adjust the header length according to the pad and
1222 * the jumbo payload option.
1224 hbh = mtod(mopt, struct ip6_hbh *);
1225 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1228 /* fill in the option. */
1229 optbuf[2] = IP6OPT_JUMBO;
1231 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1232 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1234 /* finally, adjust the packet header length */
1235 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1242 * Insert fragment header and copy unfragmentable header portions.
1245 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1246 struct ip6_frag **frghdrp)
1248 struct mbuf *n, *mlast;
1250 if (hlen > sizeof(struct ip6_hdr)) {
1251 n = m_copym(m0, sizeof(struct ip6_hdr),
1252 hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1259 /* Search for the last mbuf of unfragmentable part. */
1260 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1263 if (!(mlast->m_flags & M_EXT) &&
1264 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1265 /* use the trailing space of the last mbuf for the fragment hdr */
1267 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1268 mlast->m_len += sizeof(struct ip6_frag);
1269 m->m_pkthdr.len += sizeof(struct ip6_frag);
1271 /* allocate a new mbuf for the fragment header */
1274 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1277 mfrg->m_len = sizeof(struct ip6_frag);
1278 *frghdrp = mtod(mfrg, struct ip6_frag *);
1279 mlast->m_next = mfrg;
1286 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1287 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1294 if (ro_pmtu != ro) {
1295 /* The first hop and the final destination may differ. */
1296 struct sockaddr_in6 *sa6_dst =
1297 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1298 if (ro_pmtu->ro_rt &&
1299 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1300 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1301 RTFREE(ro_pmtu->ro_rt);
1302 ro_pmtu->ro_rt = NULL;
1304 if (ro_pmtu->ro_rt == NULL) {
1305 bzero(sa6_dst, sizeof(*sa6_dst));
1306 sa6_dst->sin6_family = AF_INET6;
1307 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1308 sa6_dst->sin6_addr = *dst;
1310 rtalloc((struct route *)ro_pmtu);
1313 if (ro_pmtu->ro_rt) {
1315 struct in_conninfo inc;
1317 bzero(&inc, sizeof(inc));
1318 inc.inc_flags = 1; /* IPv6 */
1319 inc.inc6_faddr = *dst;
1322 ifp = ro_pmtu->ro_rt->rt_ifp;
1323 ifmtu = IN6_LINKMTU(ifp);
1324 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1327 else if (mtu < IPV6_MMTU) {
1329 * RFC2460 section 5, last paragraph:
1330 * if we record ICMPv6 too big message with
1331 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1332 * or smaller, with framgent header attached.
1333 * (fragment header is needed regardless from the
1334 * packet size, for translators to identify packets)
1338 } else if (mtu > ifmtu) {
1340 * The MTU on the route is larger than the MTU on
1341 * the interface! This shouldn't happen, unless the
1342 * MTU of the interface has been changed after the
1343 * interface was brought up. Change the MTU in the
1344 * route to match the interface MTU (as long as the
1345 * field isn't locked).
1348 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1351 mtu = IN6_LINKMTU(ifp);
1353 error = EHOSTUNREACH; /* XXX */
1357 *alwaysfragp = alwaysfrag;
1362 * IP6 socket option processing.
1365 ip6_ctloutput_dispatch(netmsg_t msg)
1369 error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1370 msg->ctloutput.nm_sopt);
1371 lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1375 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1377 int optdatalen,uproto;
1379 struct inpcb *in6p = so->so_pcb;
1382 int level, op, optname;
1387 level = sopt->sopt_level;
1388 op = sopt->sopt_dir;
1389 optname = sopt->sopt_name;
1390 optlen = sopt->sopt_valsize;
1393 panic("ip6_ctloutput: arg soopt is NULL");
1399 uproto = (int)so->so_proto->pr_protocol;
1400 privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1402 if (level == IPPROTO_IPV6) {
1407 case IPV6_2292PKTOPTIONS:
1408 #ifdef IPV6_PKTOPTIONS
1409 case IPV6_PKTOPTIONS:
1414 error = soopt_getm(sopt, &m); /* XXX */
1417 soopt_to_mbuf(sopt, m); /* XXX */
1418 error = ip6_pcbopts(&in6p->in6p_outputopts,
1420 m_freem(m); /* XXX */
1425 * Use of some Hop-by-Hop options or some
1426 * Destination options, might require special
1427 * privilege. That is, normal applications
1428 * (without special privilege) might be forbidden
1429 * from setting certain options in outgoing packets,
1430 * and might never see certain options in received
1431 * packets. [RFC 2292 Section 6]
1432 * KAME specific note:
1433 * KAME prevents non-privileged users from sending or
1434 * receiving ANY hbh/dst options in order to avoid
1435 * overhead of parsing options in the kernel.
1437 case IPV6_RECVHOPOPTS:
1438 case IPV6_RECVDSTOPTS:
1439 case IPV6_RECVRTHDRDSTOPTS:
1442 case IPV6_RECVPKTINFO:
1443 case IPV6_RECVHOPLIMIT:
1444 case IPV6_RECVRTHDR:
1445 case IPV6_RECVPATHMTU:
1446 case IPV6_RECVTCLASS:
1447 case IPV6_AUTOFLOWLABEL:
1450 case IPV6_UNICAST_HOPS:
1454 if (optlen != sizeof(int)) {
1458 error = soopt_to_kbuf(sopt, &optval,
1459 sizeof optval, sizeof optval);
1464 case IPV6_UNICAST_HOPS:
1465 if (optval < -1 || optval >= 256)
1468 /* -1 = kernel default */
1469 in6p->in6p_hops = optval;
1471 if ((in6p->in6p_vflag &
1473 in6p->inp_ip_ttl = optval;
1476 #define OPTSET(bit) \
1479 in6p->in6p_flags |= (bit); \
1481 in6p->in6p_flags &= ~(bit); \
1483 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1485 * Although changed to RFC3542, It's better to also support RFC2292 API
1487 #define OPTSET2292(bit) \
1489 in6p->in6p_flags |= IN6P_RFC2292; \
1491 in6p->in6p_flags |= (bit); \
1493 in6p->in6p_flags &= ~(bit); \
1494 } while (/*CONSTCOND*/ 0)
1496 case IPV6_RECVPKTINFO:
1497 /* cannot mix with RFC2292 */
1498 if (OPTBIT(IN6P_RFC2292)) {
1502 OPTSET(IN6P_PKTINFO);
1507 struct ip6_pktopts **optp;
1509 /* cannot mix with RFC2292 */
1510 if (OPTBIT(IN6P_RFC2292)) {
1514 optp = &in6p->in6p_outputopts;
1515 error = ip6_pcbopt(IPV6_HOPLIMIT,
1516 (u_char *)&optval, sizeof(optval),
1521 case IPV6_RECVHOPLIMIT:
1522 /* cannot mix with RFC2292 */
1523 if (OPTBIT(IN6P_RFC2292)) {
1527 OPTSET(IN6P_HOPLIMIT);
1530 case IPV6_RECVHOPOPTS:
1531 /* cannot mix with RFC2292 */
1532 if (OPTBIT(IN6P_RFC2292)) {
1536 OPTSET(IN6P_HOPOPTS);
1539 case IPV6_RECVDSTOPTS:
1540 /* cannot mix with RFC2292 */
1541 if (OPTBIT(IN6P_RFC2292)) {
1545 OPTSET(IN6P_DSTOPTS);
1548 case IPV6_RECVRTHDRDSTOPTS:
1549 /* cannot mix with RFC2292 */
1550 if (OPTBIT(IN6P_RFC2292)) {
1554 OPTSET(IN6P_RTHDRDSTOPTS);
1557 case IPV6_RECVRTHDR:
1558 /* cannot mix with RFC2292 */
1559 if (OPTBIT(IN6P_RFC2292)) {
1566 case IPV6_RECVPATHMTU:
1568 * We ignore this option for TCP
1570 * (RFC3542 leaves this case
1573 if (uproto != IPPROTO_TCP)
1577 case IPV6_RECVTCLASS:
1578 /* cannot mix with RFC2292 XXX */
1579 if (OPTBIT(IN6P_RFC2292)) {
1583 OPTSET(IN6P_TCLASS);
1586 case IPV6_AUTOFLOWLABEL:
1587 OPTSET(IN6P_AUTOFLOWLABEL);
1596 * make setsockopt(IPV6_V6ONLY)
1597 * available only prior to bind(2).
1599 if (in6p->in6p_lport ||
1600 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1605 OPTSET(IN6P_IPV6_V6ONLY);
1607 in6p->in6p_vflag &= ~INP_IPV4;
1609 in6p->in6p_vflag |= INP_IPV4;
1616 case IPV6_USE_MIN_MTU:
1617 case IPV6_PREFER_TEMPADDR:
1618 if (optlen != sizeof(optval)) {
1622 error = soopt_to_kbuf(sopt, &optval,
1623 sizeof optval, sizeof optval);
1627 struct ip6_pktopts **optp;
1628 optp = &in6p->in6p_outputopts;
1629 error = ip6_pcbopt(optname,
1630 (u_char *)&optval, sizeof(optval),
1635 case IPV6_2292PKTINFO:
1636 case IPV6_2292HOPLIMIT:
1637 case IPV6_2292HOPOPTS:
1638 case IPV6_2292DSTOPTS:
1639 case IPV6_2292RTHDR:
1641 if (optlen != sizeof(int)) {
1645 error = soopt_to_kbuf(sopt, &optval,
1646 sizeof optval, sizeof optval);
1650 case IPV6_2292PKTINFO:
1651 OPTSET2292(IN6P_PKTINFO);
1653 case IPV6_2292HOPLIMIT:
1654 OPTSET2292(IN6P_HOPLIMIT);
1656 case IPV6_2292HOPOPTS:
1658 * Check super-user privilege.
1659 * See comments for IPV6_RECVHOPOPTS.
1663 OPTSET2292(IN6P_HOPOPTS);
1665 case IPV6_2292DSTOPTS:
1668 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1670 case IPV6_2292RTHDR:
1671 OPTSET2292(IN6P_RTHDR);
1680 case IPV6_RTHDRDSTOPTS:
1684 * New advanced API (RFC3542)
1687 u_char optbuf_storage[MCLBYTES];
1689 struct ip6_pktopts **optp;
1691 /* cannot mix with RFC2292 */
1692 if (OPTBIT(IN6P_RFC2292)) {
1698 * We only ensure valsize is not too large
1699 * here. Further validation will be done
1702 error = soopt_to_kbuf(sopt, optbuf_storage,
1703 sizeof(optbuf_storage), 0);
1706 optlen = sopt->sopt_valsize;
1707 optbuf = optbuf_storage;
1708 optp = &in6p->in6p_outputopts;
1709 error = ip6_pcbopt(optname, optbuf, optlen,
1715 case IPV6_MULTICAST_IF:
1716 case IPV6_MULTICAST_HOPS:
1717 case IPV6_MULTICAST_LOOP:
1718 case IPV6_JOIN_GROUP:
1719 case IPV6_LEAVE_GROUP:
1722 if (sopt->sopt_valsize > MLEN) {
1727 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1732 m->m_len = sopt->sopt_valsize;
1733 error = soopt_to_kbuf(sopt, mtod(m, char *),
1734 m->m_len, m->m_len);
1735 error = ip6_setmoptions(sopt->sopt_name,
1736 &in6p->in6p_moptions,
1742 case IPV6_PORTRANGE:
1743 error = soopt_to_kbuf(sopt, &optval,
1744 sizeof optval, sizeof optval);
1749 case IPV6_PORTRANGE_DEFAULT:
1750 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1751 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1754 case IPV6_PORTRANGE_HIGH:
1755 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1756 in6p->in6p_flags |= IN6P_HIGHPORT;
1759 case IPV6_PORTRANGE_LOW:
1760 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1761 in6p->in6p_flags |= IN6P_LOWPORT;
1770 #if defined(IPSEC) || defined(FAST_IPSEC)
1771 case IPV6_IPSEC_POLICY:
1777 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1779 soopt_to_mbuf(sopt, m); /* XXX */
1781 req = mtod(m, caddr_t);
1784 error = ipsec6_set_policy(in6p, optname, req,
1789 #endif /* KAME IPSEC */
1797 struct mbuf **mp = &m;
1799 if (ip6_fw_ctl_ptr == NULL)
1802 if ((error = soopt_getm(sopt, &m)) != 0)
1805 soopt_to_mbuf(sopt, m);
1806 error = (*ip6_fw_ctl_ptr)(optname, mp);
1812 error = ENOPROTOOPT;
1819 case IPV6_2292PKTOPTIONS:
1820 #ifdef IPV6_PKTOPTIONS
1821 case IPV6_PKTOPTIONS:
1824 * RFC3542 (effectively) deprecated the
1825 * semantics of the 2292-style pktoptions.
1826 * Since it was not reliable in nature (i.e.,
1827 * applications had to expect the lack of some
1828 * information after all), it would make sense
1829 * to simplify this part by always returning
1832 if (in6p->in6p_options) {
1834 m = m_copym(in6p->in6p_options,
1835 0, M_COPYALL, MB_WAIT);
1836 error = soopt_from_mbuf(sopt, m);
1840 sopt->sopt_valsize = 0;
1843 case IPV6_RECVHOPOPTS:
1844 case IPV6_RECVDSTOPTS:
1845 case IPV6_RECVRTHDRDSTOPTS:
1846 case IPV6_UNICAST_HOPS:
1847 case IPV6_RECVPKTINFO:
1848 case IPV6_RECVHOPLIMIT:
1849 case IPV6_RECVRTHDR:
1850 case IPV6_RECVPATHMTU:
1851 case IPV6_RECVTCLASS:
1852 case IPV6_AUTOFLOWLABEL:
1855 case IPV6_PORTRANGE:
1858 case IPV6_RECVHOPOPTS:
1859 optval = OPTBIT(IN6P_HOPOPTS);
1862 case IPV6_RECVDSTOPTS:
1863 optval = OPTBIT(IN6P_DSTOPTS);
1866 case IPV6_RECVRTHDRDSTOPTS:
1867 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1870 case IPV6_RECVPKTINFO:
1871 optval = OPTBIT(IN6P_PKTINFO);
1874 case IPV6_RECVHOPLIMIT:
1875 optval = OPTBIT(IN6P_HOPLIMIT);
1878 case IPV6_RECVRTHDR:
1879 optval = OPTBIT(IN6P_RTHDR);
1882 case IPV6_RECVPATHMTU:
1883 optval = OPTBIT(IN6P_MTU);
1886 case IPV6_RECVTCLASS:
1887 optval = OPTBIT(IN6P_TCLASS);
1890 case IPV6_AUTOFLOWLABEL:
1891 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1895 case IPV6_UNICAST_HOPS:
1896 optval = in6p->in6p_hops;
1900 optval = OPTBIT(IN6P_FAITH);
1904 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1907 case IPV6_PORTRANGE:
1910 flags = in6p->in6p_flags;
1911 if (flags & IN6P_HIGHPORT)
1912 optval = IPV6_PORTRANGE_HIGH;
1913 else if (flags & IN6P_LOWPORT)
1914 optval = IPV6_PORTRANGE_LOW;
1920 soopt_from_kbuf(sopt, &optval,
1927 struct ip6_mtuinfo mtuinfo;
1928 struct route_in6 sro;
1930 bzero(&sro, sizeof(sro));
1932 if (!(so->so_state & SS_ISCONNECTED))
1935 * XXX: we dot not consider the case of source
1936 * routing, or optional information to specify
1937 * the outgoing interface.
1939 error = ip6_getpmtu(&sro, NULL, NULL,
1940 &in6p->in6p_faddr, &pmtu, NULL);
1945 if (pmtu > IPV6_MAXPACKET)
1946 pmtu = IPV6_MAXPACKET;
1948 bzero(&mtuinfo, sizeof(mtuinfo));
1949 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1950 optdata = (void *)&mtuinfo;
1951 optdatalen = sizeof(mtuinfo);
1952 soopt_from_kbuf(sopt, optdata,
1957 case IPV6_2292PKTINFO:
1958 case IPV6_2292HOPLIMIT:
1959 case IPV6_2292HOPOPTS:
1960 case IPV6_2292RTHDR:
1961 case IPV6_2292DSTOPTS:
1962 if (optname == IPV6_2292HOPOPTS ||
1963 optname == IPV6_2292DSTOPTS ||
1967 case IPV6_2292PKTINFO:
1968 optval = OPTBIT(IN6P_PKTINFO);
1970 case IPV6_2292HOPLIMIT:
1971 optval = OPTBIT(IN6P_HOPLIMIT);
1973 case IPV6_2292HOPOPTS:
1976 optval = OPTBIT(IN6P_HOPOPTS);
1978 case IPV6_2292RTHDR:
1979 optval = OPTBIT(IN6P_RTHDR);
1981 case IPV6_2292DSTOPTS:
1984 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1987 soopt_from_kbuf(sopt, &optval,
1995 case IPV6_RTHDRDSTOPTS:
1999 case IPV6_USE_MIN_MTU:
2000 case IPV6_PREFER_TEMPADDR:
2001 error = ip6_getpcbopt(in6p->in6p_outputopts,
2005 case IPV6_MULTICAST_IF:
2006 case IPV6_MULTICAST_HOPS:
2007 case IPV6_MULTICAST_LOOP:
2008 case IPV6_JOIN_GROUP:
2009 case IPV6_LEAVE_GROUP:
2012 error = ip6_getmoptions(sopt->sopt_name,
2013 in6p->in6p_moptions, &m);
2015 soopt_from_kbuf(sopt,
2016 mtod(m, char *), m->m_len);
2021 #if defined(IPSEC) || defined(FAST_IPSEC)
2022 case IPV6_IPSEC_POLICY:
2026 struct mbuf *m = NULL;
2027 struct mbuf **mp = &m;
2029 error = soopt_getm(sopt, &m); /* XXX */
2032 soopt_to_mbuf(sopt, m); /* XXX */
2034 req = mtod(m, caddr_t);
2037 error = ipsec6_get_policy(in6p, req, len, mp);
2039 error = soopt_from_mbuf(sopt, m); /*XXX*/
2040 if (error == 0 && m != NULL)
2044 #endif /* KAME IPSEC */
2049 struct mbuf **mp = &m;
2051 if (ip6_fw_ctl_ptr == NULL)
2055 error = (*ip6_fw_ctl_ptr)(optname, mp);
2057 error = soopt_from_mbuf(sopt, m); /* XXX */
2058 if (error == 0 && m != NULL)
2064 error = ENOPROTOOPT;
2076 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2078 int error = 0, optval, optlen;
2079 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2080 struct in6pcb *in6p = sotoin6pcb(so);
2081 int level, op, optname;
2084 level = sopt->sopt_level;
2085 op = sopt->sopt_dir;
2086 optname = sopt->sopt_name;
2087 optlen = sopt->sopt_valsize;
2089 panic("ip6_raw_ctloutput: arg soopt is NULL");
2091 if (level != IPPROTO_IPV6) {
2098 * For ICMPv6 sockets, no modification allowed for checksum
2099 * offset, permit "no change" values to help existing apps.
2101 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2102 * for an ICMPv6 socket will fail."
2103 * The current behavior does not meet RFC3542.
2107 if (optlen != sizeof(int)) {
2111 error = soopt_to_kbuf(sopt, &optval,
2112 sizeof optval, sizeof optval);
2115 if ((optval % 2) != 0) {
2116 /* the API assumes even offset values */
2118 } else if (so->so_proto->pr_protocol ==
2120 if (optval != icmp6off)
2123 in6p->in6p_cksum = optval;
2127 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2130 optval = in6p->in6p_cksum;
2132 soopt_from_kbuf(sopt, &optval, sizeof(optval));
2142 error = ENOPROTOOPT;
2150 * Set up IP6 options in pcb for insertion in output packets or
2151 * specifying behavior of outgoing packets.
2154 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2155 struct socket *so, struct sockopt *sopt)
2158 struct ip6_pktopts *opt = *pktopt;
2161 /* turn off any old options. */
2164 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2165 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2166 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2167 kprintf("ip6_pcbopts: all specified options are cleared.\n");
2169 ip6_clearpktopts(opt, -1);
2171 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2174 if (!m || m->m_len == 0) {
2176 * Only turning off any previous options, regardless of
2177 * whether the opt is just created or given.
2179 kfree(opt, M_IP6OPT);
2183 /* set options specified by user. */
2184 if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2185 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2186 kfree(opt, M_IP6OPT);
2195 * Below three functions are introduced by merge to RFC3542
2199 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2201 void *optdata = NULL;
2203 struct ip6_ext *ip6e;
2205 struct in6_pktinfo null_pktinfo;
2206 int deftclass = 0, on;
2207 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2208 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2212 if (pktopt && pktopt->ip6po_pktinfo)
2213 optdata = (void *)pktopt->ip6po_pktinfo;
2215 /* XXX: we don't have to do this every time... */
2216 bzero(&null_pktinfo, sizeof(null_pktinfo));
2217 optdata = (void *)&null_pktinfo;
2219 optdatalen = sizeof(struct in6_pktinfo);
2222 if (pktopt && pktopt->ip6po_tclass >= 0)
2223 optdata = (void *)&pktopt->ip6po_tclass;
2225 optdata = (void *)&deftclass;
2226 optdatalen = sizeof(int);
2229 if (pktopt && pktopt->ip6po_hbh) {
2230 optdata = (void *)pktopt->ip6po_hbh;
2231 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2232 optdatalen = (ip6e->ip6e_len + 1) << 3;
2236 if (pktopt && pktopt->ip6po_rthdr) {
2237 optdata = (void *)pktopt->ip6po_rthdr;
2238 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2239 optdatalen = (ip6e->ip6e_len + 1) << 3;
2242 case IPV6_RTHDRDSTOPTS:
2243 if (pktopt && pktopt->ip6po_dest1) {
2244 optdata = (void *)pktopt->ip6po_dest1;
2245 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2246 optdatalen = (ip6e->ip6e_len + 1) << 3;
2250 if (pktopt && pktopt->ip6po_dest2) {
2251 optdata = (void *)pktopt->ip6po_dest2;
2252 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2253 optdatalen = (ip6e->ip6e_len + 1) << 3;
2257 if (pktopt && pktopt->ip6po_nexthop) {
2258 optdata = (void *)pktopt->ip6po_nexthop;
2259 optdatalen = pktopt->ip6po_nexthop->sa_len;
2262 case IPV6_USE_MIN_MTU:
2264 optdata = (void *)&pktopt->ip6po_minmtu;
2266 optdata = (void *)&defminmtu;
2267 optdatalen = sizeof(int);
2270 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2274 optdata = (void *)&on;
2275 optdatalen = sizeof(on);
2277 case IPV6_PREFER_TEMPADDR:
2279 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2281 optdata = (void *)&defpreftemp;
2282 optdatalen = sizeof(int);
2284 default: /* should not happen */
2286 panic("ip6_getpcbopt: unexpected option");
2288 return (ENOPROTOOPT);
2291 soopt_from_kbuf(sopt, optdata, optdatalen);
2297 * initialize ip6_pktopts. beware that there are non-zero default values in
2302 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2304 struct ip6_pktopts *opt;
2306 if (*pktopt == NULL) {
2307 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2308 init_ip6pktopts(*pktopt);
2312 return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2316 * initialize ip6_pktopts. beware that there are non-zero default values in
2320 init_ip6pktopts(struct ip6_pktopts *opt)
2323 bzero(opt, sizeof(*opt));
2324 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2325 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2326 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2327 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2331 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2336 if (optname == -1 || optname == IPV6_PKTINFO) {
2337 if (pktopt->ip6po_pktinfo)
2338 kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2339 pktopt->ip6po_pktinfo = NULL;
2341 if (optname == -1 || optname == IPV6_HOPLIMIT)
2342 pktopt->ip6po_hlim = -1;
2343 if (optname == -1 || optname == IPV6_TCLASS)
2344 pktopt->ip6po_tclass = -1;
2345 if (optname == -1 || optname == IPV6_NEXTHOP) {
2346 if (pktopt->ip6po_nextroute.ro_rt) {
2347 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2348 pktopt->ip6po_nextroute.ro_rt = NULL;
2350 if (pktopt->ip6po_nexthop)
2351 kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2352 pktopt->ip6po_nexthop = NULL;
2354 if (optname == -1 || optname == IPV6_HOPOPTS) {
2355 if (pktopt->ip6po_hbh)
2356 kfree(pktopt->ip6po_hbh, M_IP6OPT);
2357 pktopt->ip6po_hbh = NULL;
2359 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2360 if (pktopt->ip6po_dest1)
2361 kfree(pktopt->ip6po_dest1, M_IP6OPT);
2362 pktopt->ip6po_dest1 = NULL;
2364 if (optname == -1 || optname == IPV6_RTHDR) {
2365 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2366 kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2367 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2368 if (pktopt->ip6po_route.ro_rt) {
2369 RTFREE(pktopt->ip6po_route.ro_rt);
2370 pktopt->ip6po_route.ro_rt = NULL;
2373 if (optname == -1 || optname == IPV6_DSTOPTS) {
2374 if (pktopt->ip6po_dest2)
2375 kfree(pktopt->ip6po_dest2, M_IP6OPT);
2376 pktopt->ip6po_dest2 = NULL;
2380 #define PKTOPT_EXTHDRCPY(type) \
2384 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2385 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2386 if (dst->type == NULL)\
2388 bcopy(src->type, dst->type, hlen);\
2392 struct ip6_pktopts *
2393 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2395 struct ip6_pktopts *dst;
2398 kprintf("ip6_clearpktopts: invalid argument\n");
2402 dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2406 dst->ip6po_hlim = src->ip6po_hlim;
2407 if (src->ip6po_pktinfo) {
2408 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2410 if (dst->ip6po_pktinfo == NULL)
2412 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2414 if (src->ip6po_nexthop) {
2415 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2417 if (dst->ip6po_nexthop == NULL)
2419 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2420 src->ip6po_nexthop->sa_len);
2422 PKTOPT_EXTHDRCPY(ip6po_hbh);
2423 PKTOPT_EXTHDRCPY(ip6po_dest1);
2424 PKTOPT_EXTHDRCPY(ip6po_dest2);
2425 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2429 if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2430 if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2431 if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2432 if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2433 if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2434 if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2435 kfree(dst, M_IP6OPT);
2440 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2442 if (dst == NULL || src == NULL) {
2444 kprintf("ip6_clearpktopts: invalid argument\n");
2449 dst->ip6po_hlim = src->ip6po_hlim;
2450 dst->ip6po_tclass = src->ip6po_tclass;
2451 dst->ip6po_flags = src->ip6po_flags;
2452 if (src->ip6po_pktinfo) {
2453 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2455 if (dst->ip6po_pktinfo == NULL)
2457 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2459 if (src->ip6po_nexthop) {
2460 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2462 if (dst->ip6po_nexthop == NULL)
2464 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2465 src->ip6po_nexthop->sa_len);
2467 PKTOPT_EXTHDRCPY(ip6po_hbh);
2468 PKTOPT_EXTHDRCPY(ip6po_dest1);
2469 PKTOPT_EXTHDRCPY(ip6po_dest2);
2470 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2474 ip6_clearpktopts(dst, -1);
2477 #undef PKTOPT_EXTHDRCPY
2480 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2485 ip6_clearpktopts(pktopt, -1);
2487 kfree(pktopt, M_IP6OPT);
2491 * Set the IP6 multicast options in response to user setsockopt().
2494 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2497 u_int loop, ifindex;
2498 struct ipv6_mreq *mreq;
2500 struct ip6_moptions *im6o = *im6op;
2501 struct route_in6 ro;
2502 struct sockaddr_in6 *dst;
2503 struct in6_multi_mship *imm;
2504 struct thread *td = curthread; /* XXX */
2508 * No multicast option buffer attached to the pcb;
2509 * allocate one and initialize to default values.
2511 im6o = (struct ip6_moptions *)
2512 kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2515 im6o->im6o_multicast_ifp = NULL;
2516 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2517 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2518 LIST_INIT(&im6o->im6o_memberships);
2523 case IPV6_MULTICAST_IF:
2525 * Select the interface for outgoing multicast packets.
2527 if (m == NULL || m->m_len != sizeof(u_int)) {
2531 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2532 if (ifindex < 0 || if_index < ifindex) {
2533 error = ENXIO; /* XXX EINVAL? */
2536 ifp = ifindex2ifnet[ifindex];
2537 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2538 error = EADDRNOTAVAIL;
2541 im6o->im6o_multicast_ifp = ifp;
2544 case IPV6_MULTICAST_HOPS:
2547 * Set the IP6 hoplimit for outgoing multicast packets.
2550 if (m == NULL || m->m_len != sizeof(int)) {
2554 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2555 if (optval < -1 || optval >= 256)
2557 else if (optval == -1)
2558 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2560 im6o->im6o_multicast_hlim = optval;
2564 case IPV6_MULTICAST_LOOP:
2566 * Set the loopback flag for outgoing multicast packets.
2567 * Must be zero or one.
2569 if (m == NULL || m->m_len != sizeof(u_int)) {
2573 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2578 im6o->im6o_multicast_loop = loop;
2581 case IPV6_JOIN_GROUP:
2583 * Add a multicast group membership.
2584 * Group must be a valid IP6 multicast address.
2586 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2590 mreq = mtod(m, struct ipv6_mreq *);
2591 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2593 * We use the unspecified address to specify to accept
2594 * all multicast addresses. Only super user is allowed
2597 if (priv_check(td, PRIV_ROOT))
2602 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2608 * If the interface is specified, validate it.
2610 if (mreq->ipv6mr_interface < 0
2611 || if_index < mreq->ipv6mr_interface) {
2612 error = ENXIO; /* XXX EINVAL? */
2616 * If no interface was explicitly specified, choose an
2617 * appropriate one according to the given multicast address.
2619 if (mreq->ipv6mr_interface == 0) {
2621 * If the multicast address is in node-local scope,
2622 * the interface should be a loopback interface.
2623 * Otherwise, look up the routing table for the
2624 * address, and choose the outgoing interface.
2625 * XXX: is it a good approach?
2627 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2631 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2632 bzero(dst, sizeof(*dst));
2633 dst->sin6_len = sizeof(struct sockaddr_in6);
2634 dst->sin6_family = AF_INET6;
2635 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2636 rtalloc((struct route *)&ro);
2637 if (ro.ro_rt == NULL) {
2638 error = EADDRNOTAVAIL;
2641 ifp = ro.ro_rt->rt_ifp;
2645 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2648 * See if we found an interface, and confirm that it
2649 * supports multicast
2651 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2652 error = EADDRNOTAVAIL;
2656 * Put interface index into the multicast address,
2657 * if the address has link-local scope.
2659 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2660 mreq->ipv6mr_multiaddr.s6_addr16[1]
2661 = htons(mreq->ipv6mr_interface);
2664 * See if the membership already exists.
2666 for (imm = im6o->im6o_memberships.lh_first;
2667 imm != NULL; imm = imm->i6mm_chain.le_next)
2668 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2669 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2670 &mreq->ipv6mr_multiaddr))
2677 * Everything looks good; add a new record to the multicast
2678 * address list for the given interface.
2680 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2681 if ((imm->i6mm_maddr =
2682 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2683 kfree(imm, M_IPMADDR);
2686 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2689 case IPV6_LEAVE_GROUP:
2691 * Drop a multicast group membership.
2692 * Group must be a valid IP6 multicast address.
2694 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2698 mreq = mtod(m, struct ipv6_mreq *);
2699 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2700 if (priv_check(td, PRIV_ROOT)) {
2704 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2709 * If an interface address was specified, get a pointer
2710 * to its ifnet structure.
2712 if (mreq->ipv6mr_interface < 0
2713 || if_index < mreq->ipv6mr_interface) {
2714 error = ENXIO; /* XXX EINVAL? */
2717 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2719 * Put interface index into the multicast address,
2720 * if the address has link-local scope.
2722 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2723 mreq->ipv6mr_multiaddr.s6_addr16[1]
2724 = htons(mreq->ipv6mr_interface);
2727 * Find the membership in the membership list.
2729 for (imm = im6o->im6o_memberships.lh_first;
2730 imm != NULL; imm = imm->i6mm_chain.le_next) {
2732 imm->i6mm_maddr->in6m_ifp == ifp) &&
2733 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2734 &mreq->ipv6mr_multiaddr))
2738 /* Unable to resolve interface */
2739 error = EADDRNOTAVAIL;
2743 * Give up the multicast address record to which the
2744 * membership points.
2746 LIST_REMOVE(imm, i6mm_chain);
2747 in6_delmulti(imm->i6mm_maddr);
2748 kfree(imm, M_IPMADDR);
2757 * If all options have default values, no need to keep the mbuf.
2759 if (im6o->im6o_multicast_ifp == NULL &&
2760 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2761 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2762 im6o->im6o_memberships.lh_first == NULL) {
2763 kfree(*im6op, M_IPMOPTS);
2771 * Return the IP6 multicast options in response to user getsockopt().
2774 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2776 u_int *hlim, *loop, *ifindex;
2778 *mp = m_get(MB_WAIT, MT_HEADER); /* XXX */
2782 case IPV6_MULTICAST_IF:
2783 ifindex = mtod(*mp, u_int *);
2784 (*mp)->m_len = sizeof(u_int);
2785 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2788 *ifindex = im6o->im6o_multicast_ifp->if_index;
2791 case IPV6_MULTICAST_HOPS:
2792 hlim = mtod(*mp, u_int *);
2793 (*mp)->m_len = sizeof(u_int);
2795 *hlim = ip6_defmcasthlim;
2797 *hlim = im6o->im6o_multicast_hlim;
2800 case IPV6_MULTICAST_LOOP:
2801 loop = mtod(*mp, u_int *);
2802 (*mp)->m_len = sizeof(u_int);
2804 *loop = ip6_defmcasthlim;
2806 *loop = im6o->im6o_multicast_loop;
2810 return (EOPNOTSUPP);
2815 * Discard the IP6 multicast options.
2818 ip6_freemoptions(struct ip6_moptions *im6o)
2820 struct in6_multi_mship *imm;
2825 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2826 LIST_REMOVE(imm, i6mm_chain);
2827 if (imm->i6mm_maddr)
2828 in6_delmulti(imm->i6mm_maddr);
2829 kfree(imm, M_IPMADDR);
2831 kfree(im6o, M_IPMOPTS);
2835 * Set a particular packet option, as a sticky option or an ancillary data
2836 * item. "len" can be 0 only when it's a sticky option.
2837 * We have 4 cases of combination of "sticky" and "cmsg":
2838 * "sticky=0, cmsg=0": impossible
2839 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2840 * "sticky=1, cmsg=0": RFC3542 socket option
2841 * "sticky=1, cmsg=1": RFC2292 socket option
2844 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2845 int sticky, int cmsg, int uproto, int priv)
2847 int minmtupolicy, preftemp;
2850 if (!sticky && !cmsg) {
2851 kprintf("ip6_setpktoption: impossible case\n");
2856 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2857 * not be specified in the context of RFC3542. Conversely,
2858 * RFC3542 types should not be specified in the context of RFC2292.
2862 case IPV6_2292PKTINFO:
2863 case IPV6_2292HOPLIMIT:
2864 case IPV6_2292NEXTHOP:
2865 case IPV6_2292HOPOPTS:
2866 case IPV6_2292DSTOPTS:
2867 case IPV6_2292RTHDR:
2868 case IPV6_2292PKTOPTIONS:
2869 return (ENOPROTOOPT);
2872 if (sticky && cmsg) {
2879 case IPV6_RTHDRDSTOPTS:
2881 case IPV6_USE_MIN_MTU:
2884 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2885 return (ENOPROTOOPT);
2890 case IPV6_2292PKTINFO:
2893 struct in6_pktinfo *pktinfo;
2894 if (len != sizeof(struct in6_pktinfo))
2896 pktinfo = (struct in6_pktinfo *)buf;
2899 * An application can clear any sticky IPV6_PKTINFO option by
2900 * doing a "regular" setsockopt with ipi6_addr being
2901 * in6addr_any and ipi6_ifindex being zero.
2902 * [RFC 3542, Section 6]
2904 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2905 pktinfo->ipi6_ifindex == 0 &&
2906 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2907 ip6_clearpktopts(opt, optname);
2911 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2912 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2916 /* validate the interface index if specified. */
2917 if (pktinfo->ipi6_ifindex > if_index ||
2918 pktinfo->ipi6_ifindex < 0) {
2922 * Check if the requested source address is indeed a
2923 * unicast address assigned to the node, and can be
2924 * used as the packet's source address.
2926 if (opt->ip6po_pktinfo != NULL &&
2927 !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2928 struct in6_ifaddr *ia6;
2929 struct sockaddr_in6 sin6;
2931 bzero(&sin6, sizeof(sin6));
2932 sin6.sin6_len = sizeof(sin6);
2933 sin6.sin6_family = AF_INET6;
2935 opt->ip6po_pktinfo->ipi6_addr;
2936 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2938 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2939 IN6_IFF_NOTREADY)) != 0)
2940 return (EADDRNOTAVAIL);
2944 * We store the address anyway, and let in6_selectsrc()
2945 * validate the specified address. This is because ipi6_addr
2946 * may not have enough information about its scope zone, and
2947 * we may need additional information (such as outgoing
2948 * interface or the scope zone of a destination address) to
2949 * disambiguate the scope.
2950 * XXX: the delay of the validation may confuse the
2951 * application when it is used as a sticky option.
2953 if (opt->ip6po_pktinfo == NULL) {
2954 opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2955 M_IP6OPT, M_NOWAIT);
2956 if (opt->ip6po_pktinfo == NULL)
2959 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2963 case IPV6_2292HOPLIMIT:
2969 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2970 * to simplify the ordering among hoplimit options.
2972 if (optname == IPV6_HOPLIMIT && sticky)
2973 return (ENOPROTOOPT);
2975 if (len != sizeof(int))
2978 if (*hlimp < -1 || *hlimp > 255)
2981 opt->ip6po_hlim = *hlimp;
2989 if (len != sizeof(int))
2991 tclass = *(int *)buf;
2992 if (tclass < -1 || tclass > 255)
2995 opt->ip6po_tclass = tclass;
2999 case IPV6_2292NEXTHOP:
3004 if (len == 0) { /* just remove the option */
3005 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3009 /* check if cmsg_len is large enough for sa_len */
3010 if (len < sizeof(struct sockaddr) || len < *buf)
3013 switch (((struct sockaddr *)buf)->sa_family) {
3016 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3019 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3022 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3023 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3028 case AF_LINK: /* should eventually be supported */
3030 return (EAFNOSUPPORT);
3033 /* turn off the previous option, then set the new option. */
3034 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3035 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3036 if (opt->ip6po_nexthop == NULL)
3038 bcopy(buf, opt->ip6po_nexthop, *buf);
3041 case IPV6_2292HOPOPTS:
3044 struct ip6_hbh *hbh;
3048 * XXX: We don't allow a non-privileged user to set ANY HbH
3049 * options, since per-option restriction has too much
3055 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3056 break; /* just remove the option */
3059 /* message length validation */
3060 if (len < sizeof(struct ip6_hbh))
3062 hbh = (struct ip6_hbh *)buf;
3063 hbhlen = (hbh->ip6h_len + 1) << 3;
3067 /* turn off the previous option, then set the new option. */
3068 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3069 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3070 if (opt->ip6po_hbh == NULL)
3072 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3077 case IPV6_2292DSTOPTS:
3079 case IPV6_RTHDRDSTOPTS:
3081 struct ip6_dest *dest, **newdest = NULL;
3087 ip6_clearpktopts(opt, optname);
3088 break; /* just remove the option */
3091 /* message length validation */
3092 if (len < sizeof(struct ip6_dest))
3094 dest = (struct ip6_dest *)buf;
3095 destlen = (dest->ip6d_len + 1) << 3;
3100 * Determine the position that the destination options header
3101 * should be inserted; before or after the routing header.
3104 case IPV6_2292DSTOPTS:
3106 * The old advacned API is ambiguous on this point.
3107 * Our approach is to determine the position based
3108 * according to the existence of a routing header.
3109 * Note, however, that this depends on the order of the
3110 * extension headers in the ancillary data; the 1st
3111 * part of the destination options header must appear
3112 * before the routing header in the ancillary data,
3114 * RFC3542 solved the ambiguity by introducing
3115 * separate ancillary data or option types.
3117 if (opt->ip6po_rthdr == NULL)
3118 newdest = &opt->ip6po_dest1;
3120 newdest = &opt->ip6po_dest2;
3122 case IPV6_RTHDRDSTOPTS:
3123 newdest = &opt->ip6po_dest1;
3126 newdest = &opt->ip6po_dest2;
3130 /* turn off the previous option, then set the new option. */
3131 ip6_clearpktopts(opt, optname);
3132 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3133 if (*newdest == NULL)
3135 bcopy(dest, *newdest, destlen);
3140 case IPV6_2292RTHDR:
3143 struct ip6_rthdr *rth;
3147 ip6_clearpktopts(opt, IPV6_RTHDR);
3148 break; /* just remove the option */
3151 /* message length validation */
3152 if (len < sizeof(struct ip6_rthdr))
3154 rth = (struct ip6_rthdr *)buf;
3155 rthlen = (rth->ip6r_len + 1) << 3;
3159 switch (rth->ip6r_type) {
3161 return (EINVAL); /* not supported */
3164 /* turn off the previous option */
3165 ip6_clearpktopts(opt, IPV6_RTHDR);
3166 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3167 if (opt->ip6po_rthdr == NULL)
3169 bcopy(rth, opt->ip6po_rthdr, rthlen);
3174 case IPV6_USE_MIN_MTU:
3175 if (len != sizeof(int))
3177 minmtupolicy = *(int *)buf;
3178 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3179 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3180 minmtupolicy != IP6PO_MINMTU_ALL) {
3183 opt->ip6po_minmtu = minmtupolicy;
3187 if (len != sizeof(int))
3190 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3192 * we ignore this option for TCP sockets.
3193 * (RFC3542 leaves this case unspecified.)
3195 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3197 opt->ip6po_flags |= IP6PO_DONTFRAG;
3200 case IPV6_PREFER_TEMPADDR:
3201 if (len != sizeof(int))
3203 preftemp = *(int *)buf;
3204 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3205 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3206 preftemp != IP6PO_TEMPADDR_PREFER) {
3209 opt->ip6po_prefer_tempaddr = preftemp;
3213 return (ENOPROTOOPT);
3214 } /* end of switch */
3221 * Set IPv6 outgoing packet options based on advanced API.
3224 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3225 struct ip6_pktopts *stickyopt, int uproto, int priv)
3227 struct cmsghdr *cm = NULL;
3229 if (control == NULL || opt == NULL)
3232 init_ip6pktopts(opt);
3235 * XXX: Currently, we assume all the optional information is stored
3242 * If stickyopt is provided, make a local copy of the options
3243 * for this particular packet, then override them by ancillary
3245 * XXX: copypktopts() does not copy the cached route to a next
3246 * hop (if any). This is not very good in terms of efficiency,
3247 * but we can allow this since this option should be rarely
3250 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3255 * XXX: Currently, we assume all the optional information is stored
3258 if (control->m_next)
3261 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3262 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3265 if (control->m_len < CMSG_LEN(0))
3268 cm = mtod(control, struct cmsghdr *);
3269 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3271 if (cm->cmsg_level != IPPROTO_IPV6)
3274 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3275 cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3284 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3285 * packet to the input queue of a specified interface. Note that this
3286 * calls the output routine of the loopback "driver", but with an interface
3287 * pointer that might NOT be &loif -- easier than replicating that code here.
3290 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3293 struct ip6_hdr *ip6;
3295 copym = m_copy(m, 0, M_COPYALL);
3300 * Make sure to deep-copy IPv6 header portion in case the data
3301 * is in an mbuf cluster, so that we can safely override the IPv6
3302 * header portion later.
3304 if ((copym->m_flags & M_EXT) != 0 ||
3305 copym->m_len < sizeof(struct ip6_hdr)) {
3306 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3312 if (copym->m_len < sizeof(*ip6)) {
3318 ip6 = mtod(copym, struct ip6_hdr *);
3320 * clear embedded scope identifiers if necessary.
3321 * in6_clearscope will touch the addresses only when necessary.
3323 in6_clearscope(&ip6->ip6_src);
3324 in6_clearscope(&ip6->ip6_dst);
3326 if_simloop(ifp, copym, dst->sin6_family, 0);
3330 * Separate the IPv6 header from the payload into its own mbuf.
3332 * Returns the new mbuf chain or the original mbuf if no payload.
3333 * Returns NULL if can't allocate new mbuf for header.
3335 static struct mbuf *
3336 ip6_splithdr(struct mbuf *m)
3340 if (m->m_len <= sizeof(struct ip6_hdr)) /* no payload */
3343 MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
3346 mh->m_len = sizeof(struct ip6_hdr);
3347 M_MOVE_PKTHDR(mh, m);
3348 MH_ALIGN(mh, sizeof(struct ip6_hdr));
3349 bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3350 m->m_data += sizeof(struct ip6_hdr);
3351 m->m_len -= sizeof(struct ip6_hdr);
3357 * Compute IPv6 extension header length.
3360 ip6_optlen(struct in6pcb *in6p)
3364 if (!in6p->in6p_outputopts)
3369 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3371 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3372 if (in6p->in6p_outputopts->ip6po_rthdr)
3373 /* dest1 is valid with rthdr only */
3374 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3375 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3376 len += elen(in6p->in6p_outputopts->ip6po_dest2);