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 ia6->ia_ifa.if_opackets++;
952 ia6->ia_ifa.if_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 ia->ia_ifa.if_opackets++;
1085 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1088 /* clean ipsec history once it goes out of the node */
1091 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1097 ip6stat.ip6s_fragmented++;
1100 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1102 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1103 RTFREE(ro_pmtu->ro_rt);
1118 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1119 m_freem(exthdrs.ip6e_dest1);
1120 m_freem(exthdrs.ip6e_rthdr);
1121 m_freem(exthdrs.ip6e_dest2);
1129 copyexthdr(void *h, struct mbuf **mp)
1131 struct ip6_ext *hdr = h;
1138 hlen = (hdr->ip6e_len + 1) * 8;
1139 if (hlen > MCLBYTES)
1140 return ENOBUFS; /* XXX */
1142 m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1147 bcopy(hdr, mtod(m, caddr_t), hlen);
1154 * Insert jumbo payload option.
1157 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1163 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1166 * If there is no hop-by-hop options header, allocate new one.
1167 * If there is one but it doesn't have enough space to store the
1168 * jumbo payload option, allocate a cluster to store the whole options.
1169 * Otherwise, use it to store the options.
1171 if (exthdrs->ip6e_hbh == NULL) {
1172 MGET(mopt, MB_DONTWAIT, MT_DATA);
1175 mopt->m_len = JUMBOOPTLEN;
1176 optbuf = mtod(mopt, u_char *);
1177 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1178 exthdrs->ip6e_hbh = mopt;
1180 struct ip6_hbh *hbh;
1182 mopt = exthdrs->ip6e_hbh;
1183 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1186 * - exthdrs->ip6e_hbh is not referenced from places
1187 * other than exthdrs.
1188 * - exthdrs->ip6e_hbh is not an mbuf chain.
1190 int oldoptlen = mopt->m_len;
1194 * XXX: give up if the whole (new) hbh header does
1195 * not fit even in an mbuf cluster.
1197 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1201 * As a consequence, we must always prepare a cluster
1204 n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1207 n->m_len = oldoptlen + JUMBOOPTLEN;
1208 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1209 optbuf = mtod(n, caddr_t) + oldoptlen;
1211 mopt = exthdrs->ip6e_hbh = n;
1213 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1214 mopt->m_len += JUMBOOPTLEN;
1216 optbuf[0] = IP6OPT_PADN;
1220 * Adjust the header length according to the pad and
1221 * the jumbo payload option.
1223 hbh = mtod(mopt, struct ip6_hbh *);
1224 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1227 /* fill in the option. */
1228 optbuf[2] = IP6OPT_JUMBO;
1230 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1231 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1233 /* finally, adjust the packet header length */
1234 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1241 * Insert fragment header and copy unfragmentable header portions.
1244 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1245 struct ip6_frag **frghdrp)
1247 struct mbuf *n, *mlast;
1249 if (hlen > sizeof(struct ip6_hdr)) {
1250 n = m_copym(m0, sizeof(struct ip6_hdr),
1251 hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1258 /* Search for the last mbuf of unfragmentable part. */
1259 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1262 if (!(mlast->m_flags & M_EXT) &&
1263 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1264 /* use the trailing space of the last mbuf for the fragment hdr */
1266 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1267 mlast->m_len += sizeof(struct ip6_frag);
1268 m->m_pkthdr.len += sizeof(struct ip6_frag);
1270 /* allocate a new mbuf for the fragment header */
1273 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1276 mfrg->m_len = sizeof(struct ip6_frag);
1277 *frghdrp = mtod(mfrg, struct ip6_frag *);
1278 mlast->m_next = mfrg;
1285 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1286 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1293 if (ro_pmtu != ro) {
1294 /* The first hop and the final destination may differ. */
1295 struct sockaddr_in6 *sa6_dst =
1296 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1297 if (ro_pmtu->ro_rt &&
1298 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1299 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1300 RTFREE(ro_pmtu->ro_rt);
1301 ro_pmtu->ro_rt = NULL;
1303 if (ro_pmtu->ro_rt == NULL) {
1304 bzero(sa6_dst, sizeof(*sa6_dst));
1305 sa6_dst->sin6_family = AF_INET6;
1306 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1307 sa6_dst->sin6_addr = *dst;
1309 rtalloc((struct route *)ro_pmtu);
1312 if (ro_pmtu->ro_rt) {
1314 struct in_conninfo inc;
1316 bzero(&inc, sizeof(inc));
1317 inc.inc_flags = 1; /* IPv6 */
1318 inc.inc6_faddr = *dst;
1321 ifp = ro_pmtu->ro_rt->rt_ifp;
1322 ifmtu = IN6_LINKMTU(ifp);
1323 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1326 else if (mtu < IPV6_MMTU) {
1328 * RFC2460 section 5, last paragraph:
1329 * if we record ICMPv6 too big message with
1330 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1331 * or smaller, with framgent header attached.
1332 * (fragment header is needed regardless from the
1333 * packet size, for translators to identify packets)
1337 } else if (mtu > ifmtu) {
1339 * The MTU on the route is larger than the MTU on
1340 * the interface! This shouldn't happen, unless the
1341 * MTU of the interface has been changed after the
1342 * interface was brought up. Change the MTU in the
1343 * route to match the interface MTU (as long as the
1344 * field isn't locked).
1347 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1350 mtu = IN6_LINKMTU(ifp);
1352 error = EHOSTUNREACH; /* XXX */
1356 *alwaysfragp = alwaysfrag;
1361 * IP6 socket option processing.
1364 ip6_ctloutput_dispatch(netmsg_t msg)
1368 error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1369 msg->ctloutput.nm_sopt);
1370 lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1374 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1376 int optdatalen,uproto;
1378 struct inpcb *in6p = so->so_pcb;
1381 int level, op, optname;
1386 level = sopt->sopt_level;
1387 op = sopt->sopt_dir;
1388 optname = sopt->sopt_name;
1389 optlen = sopt->sopt_valsize;
1392 panic("ip6_ctloutput: arg soopt is NULL");
1398 uproto = (int)so->so_proto->pr_protocol;
1399 privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1401 if (level == IPPROTO_IPV6) {
1406 case IPV6_2292PKTOPTIONS:
1407 #ifdef IPV6_PKTOPTIONS
1408 case IPV6_PKTOPTIONS:
1413 error = soopt_getm(sopt, &m); /* XXX */
1416 soopt_to_mbuf(sopt, m); /* XXX */
1417 error = ip6_pcbopts(&in6p->in6p_outputopts,
1419 m_freem(m); /* XXX */
1424 * Use of some Hop-by-Hop options or some
1425 * Destination options, might require special
1426 * privilege. That is, normal applications
1427 * (without special privilege) might be forbidden
1428 * from setting certain options in outgoing packets,
1429 * and might never see certain options in received
1430 * packets. [RFC 2292 Section 6]
1431 * KAME specific note:
1432 * KAME prevents non-privileged users from sending or
1433 * receiving ANY hbh/dst options in order to avoid
1434 * overhead of parsing options in the kernel.
1436 case IPV6_RECVHOPOPTS:
1437 case IPV6_RECVDSTOPTS:
1438 case IPV6_RECVRTHDRDSTOPTS:
1441 case IPV6_RECVPKTINFO:
1442 case IPV6_RECVHOPLIMIT:
1443 case IPV6_RECVRTHDR:
1444 case IPV6_RECVPATHMTU:
1445 case IPV6_RECVTCLASS:
1446 case IPV6_AUTOFLOWLABEL:
1449 case IPV6_UNICAST_HOPS:
1453 if (optlen != sizeof(int)) {
1457 error = soopt_to_kbuf(sopt, &optval,
1458 sizeof optval, sizeof optval);
1463 case IPV6_UNICAST_HOPS:
1464 if (optval < -1 || optval >= 256)
1467 /* -1 = kernel default */
1468 in6p->in6p_hops = optval;
1470 if ((in6p->in6p_vflag &
1472 in6p->inp_ip_ttl = optval;
1475 #define OPTSET(bit) \
1478 in6p->in6p_flags |= (bit); \
1480 in6p->in6p_flags &= ~(bit); \
1482 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1484 * Although changed to RFC3542, It's better to also support RFC2292 API
1486 #define OPTSET2292(bit) \
1488 in6p->in6p_flags |= IN6P_RFC2292; \
1490 in6p->in6p_flags |= (bit); \
1492 in6p->in6p_flags &= ~(bit); \
1493 } while (/*CONSTCOND*/ 0)
1495 case IPV6_RECVPKTINFO:
1496 /* cannot mix with RFC2292 */
1497 if (OPTBIT(IN6P_RFC2292)) {
1501 OPTSET(IN6P_PKTINFO);
1506 struct ip6_pktopts **optp;
1508 /* cannot mix with RFC2292 */
1509 if (OPTBIT(IN6P_RFC2292)) {
1513 optp = &in6p->in6p_outputopts;
1514 error = ip6_pcbopt(IPV6_HOPLIMIT,
1515 (u_char *)&optval, sizeof(optval),
1520 case IPV6_RECVHOPLIMIT:
1521 /* cannot mix with RFC2292 */
1522 if (OPTBIT(IN6P_RFC2292)) {
1526 OPTSET(IN6P_HOPLIMIT);
1529 case IPV6_RECVHOPOPTS:
1530 /* cannot mix with RFC2292 */
1531 if (OPTBIT(IN6P_RFC2292)) {
1535 OPTSET(IN6P_HOPOPTS);
1538 case IPV6_RECVDSTOPTS:
1539 /* cannot mix with RFC2292 */
1540 if (OPTBIT(IN6P_RFC2292)) {
1544 OPTSET(IN6P_DSTOPTS);
1547 case IPV6_RECVRTHDRDSTOPTS:
1548 /* cannot mix with RFC2292 */
1549 if (OPTBIT(IN6P_RFC2292)) {
1553 OPTSET(IN6P_RTHDRDSTOPTS);
1556 case IPV6_RECVRTHDR:
1557 /* cannot mix with RFC2292 */
1558 if (OPTBIT(IN6P_RFC2292)) {
1565 case IPV6_RECVPATHMTU:
1567 * We ignore this option for TCP
1569 * (RFC3542 leaves this case
1572 if (uproto != IPPROTO_TCP)
1576 case IPV6_RECVTCLASS:
1577 /* cannot mix with RFC2292 XXX */
1578 if (OPTBIT(IN6P_RFC2292)) {
1582 OPTSET(IN6P_TCLASS);
1585 case IPV6_AUTOFLOWLABEL:
1586 OPTSET(IN6P_AUTOFLOWLABEL);
1595 * make setsockopt(IPV6_V6ONLY)
1596 * available only prior to bind(2).
1598 if (in6p->in6p_lport ||
1599 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1604 OPTSET(IN6P_IPV6_V6ONLY);
1606 in6p->in6p_vflag &= ~INP_IPV4;
1608 in6p->in6p_vflag |= INP_IPV4;
1615 case IPV6_USE_MIN_MTU:
1616 case IPV6_PREFER_TEMPADDR:
1617 if (optlen != sizeof(optval)) {
1621 error = soopt_to_kbuf(sopt, &optval,
1622 sizeof optval, sizeof optval);
1626 struct ip6_pktopts **optp;
1627 optp = &in6p->in6p_outputopts;
1628 error = ip6_pcbopt(optname,
1629 (u_char *)&optval, sizeof(optval),
1634 case IPV6_2292PKTINFO:
1635 case IPV6_2292HOPLIMIT:
1636 case IPV6_2292HOPOPTS:
1637 case IPV6_2292DSTOPTS:
1638 case IPV6_2292RTHDR:
1640 if (optlen != sizeof(int)) {
1644 error = soopt_to_kbuf(sopt, &optval,
1645 sizeof optval, sizeof optval);
1649 case IPV6_2292PKTINFO:
1650 OPTSET2292(IN6P_PKTINFO);
1652 case IPV6_2292HOPLIMIT:
1653 OPTSET2292(IN6P_HOPLIMIT);
1655 case IPV6_2292HOPOPTS:
1657 * Check super-user privilege.
1658 * See comments for IPV6_RECVHOPOPTS.
1662 OPTSET2292(IN6P_HOPOPTS);
1664 case IPV6_2292DSTOPTS:
1667 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1669 case IPV6_2292RTHDR:
1670 OPTSET2292(IN6P_RTHDR);
1679 case IPV6_RTHDRDSTOPTS:
1683 * New advanced API (RFC3542)
1686 u_char optbuf_storage[MCLBYTES];
1688 struct ip6_pktopts **optp;
1690 /* cannot mix with RFC2292 */
1691 if (OPTBIT(IN6P_RFC2292)) {
1697 * We only ensure valsize is not too large
1698 * here. Further validation will be done
1701 error = soopt_to_kbuf(sopt, optbuf_storage,
1702 sizeof(optbuf_storage), 0);
1705 optlen = sopt->sopt_valsize;
1706 optbuf = optbuf_storage;
1707 optp = &in6p->in6p_outputopts;
1708 error = ip6_pcbopt(optname, optbuf, optlen,
1714 case IPV6_MULTICAST_IF:
1715 case IPV6_MULTICAST_HOPS:
1716 case IPV6_MULTICAST_LOOP:
1717 case IPV6_JOIN_GROUP:
1718 case IPV6_LEAVE_GROUP:
1721 if (sopt->sopt_valsize > MLEN) {
1726 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1731 m->m_len = sopt->sopt_valsize;
1732 error = soopt_to_kbuf(sopt, mtod(m, char *),
1733 m->m_len, m->m_len);
1734 error = ip6_setmoptions(sopt->sopt_name,
1735 &in6p->in6p_moptions,
1741 case IPV6_PORTRANGE:
1742 error = soopt_to_kbuf(sopt, &optval,
1743 sizeof optval, sizeof optval);
1748 case IPV6_PORTRANGE_DEFAULT:
1749 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1750 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1753 case IPV6_PORTRANGE_HIGH:
1754 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1755 in6p->in6p_flags |= IN6P_HIGHPORT;
1758 case IPV6_PORTRANGE_LOW:
1759 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1760 in6p->in6p_flags |= IN6P_LOWPORT;
1769 #if defined(IPSEC) || defined(FAST_IPSEC)
1770 case IPV6_IPSEC_POLICY:
1776 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1778 soopt_to_mbuf(sopt, m); /* XXX */
1780 req = mtod(m, caddr_t);
1783 error = ipsec6_set_policy(in6p, optname, req,
1788 #endif /* KAME IPSEC */
1796 struct mbuf **mp = &m;
1798 if (ip6_fw_ctl_ptr == NULL)
1801 if ((error = soopt_getm(sopt, &m)) != 0)
1804 soopt_to_mbuf(sopt, m);
1805 error = (*ip6_fw_ctl_ptr)(optname, mp);
1811 error = ENOPROTOOPT;
1818 case IPV6_2292PKTOPTIONS:
1819 #ifdef IPV6_PKTOPTIONS
1820 case IPV6_PKTOPTIONS:
1823 * RFC3542 (effectively) deprecated the
1824 * semantics of the 2292-style pktoptions.
1825 * Since it was not reliable in nature (i.e.,
1826 * applications had to expect the lack of some
1827 * information after all), it would make sense
1828 * to simplify this part by always returning
1831 if (in6p->in6p_options) {
1833 m = m_copym(in6p->in6p_options,
1834 0, M_COPYALL, MB_WAIT);
1835 error = soopt_from_mbuf(sopt, m);
1839 sopt->sopt_valsize = 0;
1842 case IPV6_RECVHOPOPTS:
1843 case IPV6_RECVDSTOPTS:
1844 case IPV6_RECVRTHDRDSTOPTS:
1845 case IPV6_UNICAST_HOPS:
1846 case IPV6_RECVPKTINFO:
1847 case IPV6_RECVHOPLIMIT:
1848 case IPV6_RECVRTHDR:
1849 case IPV6_RECVPATHMTU:
1850 case IPV6_RECVTCLASS:
1851 case IPV6_AUTOFLOWLABEL:
1854 case IPV6_PORTRANGE:
1857 case IPV6_RECVHOPOPTS:
1858 optval = OPTBIT(IN6P_HOPOPTS);
1861 case IPV6_RECVDSTOPTS:
1862 optval = OPTBIT(IN6P_DSTOPTS);
1865 case IPV6_RECVRTHDRDSTOPTS:
1866 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1869 case IPV6_RECVPKTINFO:
1870 optval = OPTBIT(IN6P_PKTINFO);
1873 case IPV6_RECVHOPLIMIT:
1874 optval = OPTBIT(IN6P_HOPLIMIT);
1877 case IPV6_RECVRTHDR:
1878 optval = OPTBIT(IN6P_RTHDR);
1881 case IPV6_RECVPATHMTU:
1882 optval = OPTBIT(IN6P_MTU);
1885 case IPV6_RECVTCLASS:
1886 optval = OPTBIT(IN6P_TCLASS);
1889 case IPV6_AUTOFLOWLABEL:
1890 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1894 case IPV6_UNICAST_HOPS:
1895 optval = in6p->in6p_hops;
1899 optval = OPTBIT(IN6P_FAITH);
1903 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1906 case IPV6_PORTRANGE:
1909 flags = in6p->in6p_flags;
1910 if (flags & IN6P_HIGHPORT)
1911 optval = IPV6_PORTRANGE_HIGH;
1912 else if (flags & IN6P_LOWPORT)
1913 optval = IPV6_PORTRANGE_LOW;
1919 soopt_from_kbuf(sopt, &optval,
1926 struct ip6_mtuinfo mtuinfo;
1927 struct route_in6 sro;
1929 bzero(&sro, sizeof(sro));
1931 if (!(so->so_state & SS_ISCONNECTED))
1934 * XXX: we dot not consider the case of source
1935 * routing, or optional information to specify
1936 * the outgoing interface.
1938 error = ip6_getpmtu(&sro, NULL, NULL,
1939 &in6p->in6p_faddr, &pmtu, NULL);
1944 if (pmtu > IPV6_MAXPACKET)
1945 pmtu = IPV6_MAXPACKET;
1947 bzero(&mtuinfo, sizeof(mtuinfo));
1948 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1949 optdata = (void *)&mtuinfo;
1950 optdatalen = sizeof(mtuinfo);
1951 soopt_from_kbuf(sopt, optdata,
1956 case IPV6_2292PKTINFO:
1957 case IPV6_2292HOPLIMIT:
1958 case IPV6_2292HOPOPTS:
1959 case IPV6_2292RTHDR:
1960 case IPV6_2292DSTOPTS:
1961 if (optname == IPV6_2292HOPOPTS ||
1962 optname == IPV6_2292DSTOPTS ||
1966 case IPV6_2292PKTINFO:
1967 optval = OPTBIT(IN6P_PKTINFO);
1969 case IPV6_2292HOPLIMIT:
1970 optval = OPTBIT(IN6P_HOPLIMIT);
1972 case IPV6_2292HOPOPTS:
1975 optval = OPTBIT(IN6P_HOPOPTS);
1977 case IPV6_2292RTHDR:
1978 optval = OPTBIT(IN6P_RTHDR);
1980 case IPV6_2292DSTOPTS:
1983 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1986 soopt_from_kbuf(sopt, &optval,
1994 case IPV6_RTHDRDSTOPTS:
1998 case IPV6_USE_MIN_MTU:
1999 case IPV6_PREFER_TEMPADDR:
2000 error = ip6_getpcbopt(in6p->in6p_outputopts,
2004 case IPV6_MULTICAST_IF:
2005 case IPV6_MULTICAST_HOPS:
2006 case IPV6_MULTICAST_LOOP:
2007 case IPV6_JOIN_GROUP:
2008 case IPV6_LEAVE_GROUP:
2011 error = ip6_getmoptions(sopt->sopt_name,
2012 in6p->in6p_moptions, &m);
2014 soopt_from_kbuf(sopt,
2015 mtod(m, char *), m->m_len);
2020 #if defined(IPSEC) || defined(FAST_IPSEC)
2021 case IPV6_IPSEC_POLICY:
2025 struct mbuf *m = NULL;
2026 struct mbuf **mp = &m;
2028 error = soopt_getm(sopt, &m); /* XXX */
2031 soopt_to_mbuf(sopt, m); /* XXX */
2033 req = mtod(m, caddr_t);
2036 error = ipsec6_get_policy(in6p, req, len, mp);
2038 error = soopt_from_mbuf(sopt, m); /*XXX*/
2039 if (error == 0 && m != NULL)
2043 #endif /* KAME IPSEC */
2048 struct mbuf **mp = &m;
2050 if (ip6_fw_ctl_ptr == NULL)
2054 error = (*ip6_fw_ctl_ptr)(optname, mp);
2056 error = soopt_from_mbuf(sopt, m); /* XXX */
2057 if (error == 0 && m != NULL)
2063 error = ENOPROTOOPT;
2075 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2077 int error = 0, optval, optlen;
2078 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2079 struct in6pcb *in6p = sotoin6pcb(so);
2080 int level, op, optname;
2083 level = sopt->sopt_level;
2084 op = sopt->sopt_dir;
2085 optname = sopt->sopt_name;
2086 optlen = sopt->sopt_valsize;
2088 panic("ip6_raw_ctloutput: arg soopt is NULL");
2090 if (level != IPPROTO_IPV6) {
2097 * For ICMPv6 sockets, no modification allowed for checksum
2098 * offset, permit "no change" values to help existing apps.
2100 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2101 * for an ICMPv6 socket will fail."
2102 * The current behavior does not meet RFC3542.
2106 if (optlen != sizeof(int)) {
2110 error = soopt_to_kbuf(sopt, &optval,
2111 sizeof optval, sizeof optval);
2114 if ((optval % 2) != 0) {
2115 /* the API assumes even offset values */
2117 } else if (so->so_proto->pr_protocol ==
2119 if (optval != icmp6off)
2122 in6p->in6p_cksum = optval;
2126 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2129 optval = in6p->in6p_cksum;
2131 soopt_from_kbuf(sopt, &optval, sizeof(optval));
2141 error = ENOPROTOOPT;
2149 * Set up IP6 options in pcb for insertion in output packets or
2150 * specifying behavior of outgoing packets.
2153 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2154 struct socket *so, struct sockopt *sopt)
2157 struct ip6_pktopts *opt = *pktopt;
2160 /* turn off any old options. */
2163 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2164 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2165 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2166 kprintf("ip6_pcbopts: all specified options are cleared.\n");
2168 ip6_clearpktopts(opt, -1);
2170 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2173 if (!m || m->m_len == 0) {
2175 * Only turning off any previous options, regardless of
2176 * whether the opt is just created or given.
2178 kfree(opt, M_IP6OPT);
2182 /* set options specified by user. */
2183 if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2184 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2185 kfree(opt, M_IP6OPT);
2194 * Below three functions are introduced by merge to RFC3542
2198 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2200 void *optdata = NULL;
2202 struct ip6_ext *ip6e;
2204 struct in6_pktinfo null_pktinfo;
2205 int deftclass = 0, on;
2206 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2207 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2211 if (pktopt && pktopt->ip6po_pktinfo)
2212 optdata = (void *)pktopt->ip6po_pktinfo;
2214 /* XXX: we don't have to do this every time... */
2215 bzero(&null_pktinfo, sizeof(null_pktinfo));
2216 optdata = (void *)&null_pktinfo;
2218 optdatalen = sizeof(struct in6_pktinfo);
2221 if (pktopt && pktopt->ip6po_tclass >= 0)
2222 optdata = (void *)&pktopt->ip6po_tclass;
2224 optdata = (void *)&deftclass;
2225 optdatalen = sizeof(int);
2228 if (pktopt && pktopt->ip6po_hbh) {
2229 optdata = (void *)pktopt->ip6po_hbh;
2230 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2231 optdatalen = (ip6e->ip6e_len + 1) << 3;
2235 if (pktopt && pktopt->ip6po_rthdr) {
2236 optdata = (void *)pktopt->ip6po_rthdr;
2237 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2238 optdatalen = (ip6e->ip6e_len + 1) << 3;
2241 case IPV6_RTHDRDSTOPTS:
2242 if (pktopt && pktopt->ip6po_dest1) {
2243 optdata = (void *)pktopt->ip6po_dest1;
2244 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2245 optdatalen = (ip6e->ip6e_len + 1) << 3;
2249 if (pktopt && pktopt->ip6po_dest2) {
2250 optdata = (void *)pktopt->ip6po_dest2;
2251 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2252 optdatalen = (ip6e->ip6e_len + 1) << 3;
2256 if (pktopt && pktopt->ip6po_nexthop) {
2257 optdata = (void *)pktopt->ip6po_nexthop;
2258 optdatalen = pktopt->ip6po_nexthop->sa_len;
2261 case IPV6_USE_MIN_MTU:
2263 optdata = (void *)&pktopt->ip6po_minmtu;
2265 optdata = (void *)&defminmtu;
2266 optdatalen = sizeof(int);
2269 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2273 optdata = (void *)&on;
2274 optdatalen = sizeof(on);
2276 case IPV6_PREFER_TEMPADDR:
2278 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2280 optdata = (void *)&defpreftemp;
2281 optdatalen = sizeof(int);
2283 default: /* should not happen */
2285 panic("ip6_getpcbopt: unexpected option");
2287 return (ENOPROTOOPT);
2290 soopt_from_kbuf(sopt, optdata, optdatalen);
2296 * initialize ip6_pktopts. beware that there are non-zero default values in
2301 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2303 struct ip6_pktopts *opt;
2305 if (*pktopt == NULL) {
2306 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2307 init_ip6pktopts(*pktopt);
2311 return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2315 * initialize ip6_pktopts. beware that there are non-zero default values in
2319 init_ip6pktopts(struct ip6_pktopts *opt)
2322 bzero(opt, sizeof(*opt));
2323 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2324 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2325 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2326 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2330 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2335 if (optname == -1 || optname == IPV6_PKTINFO) {
2336 if (pktopt->ip6po_pktinfo)
2337 kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2338 pktopt->ip6po_pktinfo = NULL;
2340 if (optname == -1 || optname == IPV6_HOPLIMIT)
2341 pktopt->ip6po_hlim = -1;
2342 if (optname == -1 || optname == IPV6_TCLASS)
2343 pktopt->ip6po_tclass = -1;
2344 if (optname == -1 || optname == IPV6_NEXTHOP) {
2345 if (pktopt->ip6po_nextroute.ro_rt) {
2346 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2347 pktopt->ip6po_nextroute.ro_rt = NULL;
2349 if (pktopt->ip6po_nexthop)
2350 kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2351 pktopt->ip6po_nexthop = NULL;
2353 if (optname == -1 || optname == IPV6_HOPOPTS) {
2354 if (pktopt->ip6po_hbh)
2355 kfree(pktopt->ip6po_hbh, M_IP6OPT);
2356 pktopt->ip6po_hbh = NULL;
2358 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2359 if (pktopt->ip6po_dest1)
2360 kfree(pktopt->ip6po_dest1, M_IP6OPT);
2361 pktopt->ip6po_dest1 = NULL;
2363 if (optname == -1 || optname == IPV6_RTHDR) {
2364 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2365 kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2366 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2367 if (pktopt->ip6po_route.ro_rt) {
2368 RTFREE(pktopt->ip6po_route.ro_rt);
2369 pktopt->ip6po_route.ro_rt = NULL;
2372 if (optname == -1 || optname == IPV6_DSTOPTS) {
2373 if (pktopt->ip6po_dest2)
2374 kfree(pktopt->ip6po_dest2, M_IP6OPT);
2375 pktopt->ip6po_dest2 = NULL;
2379 #define PKTOPT_EXTHDRCPY(type) \
2383 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2384 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2385 if (dst->type == NULL)\
2387 bcopy(src->type, dst->type, hlen);\
2391 struct ip6_pktopts *
2392 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2394 struct ip6_pktopts *dst;
2397 kprintf("ip6_clearpktopts: invalid argument\n");
2401 dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2405 dst->ip6po_hlim = src->ip6po_hlim;
2406 if (src->ip6po_pktinfo) {
2407 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2409 if (dst->ip6po_pktinfo == NULL)
2411 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2413 if (src->ip6po_nexthop) {
2414 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2416 if (dst->ip6po_nexthop == NULL)
2418 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2419 src->ip6po_nexthop->sa_len);
2421 PKTOPT_EXTHDRCPY(ip6po_hbh);
2422 PKTOPT_EXTHDRCPY(ip6po_dest1);
2423 PKTOPT_EXTHDRCPY(ip6po_dest2);
2424 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2428 if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2429 if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2430 if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2431 if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2432 if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2433 if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2434 kfree(dst, M_IP6OPT);
2439 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2441 if (dst == NULL || src == NULL) {
2443 kprintf("ip6_clearpktopts: invalid argument\n");
2448 dst->ip6po_hlim = src->ip6po_hlim;
2449 dst->ip6po_tclass = src->ip6po_tclass;
2450 dst->ip6po_flags = src->ip6po_flags;
2451 if (src->ip6po_pktinfo) {
2452 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2454 if (dst->ip6po_pktinfo == NULL)
2456 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2458 if (src->ip6po_nexthop) {
2459 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2461 if (dst->ip6po_nexthop == NULL)
2463 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2464 src->ip6po_nexthop->sa_len);
2466 PKTOPT_EXTHDRCPY(ip6po_hbh);
2467 PKTOPT_EXTHDRCPY(ip6po_dest1);
2468 PKTOPT_EXTHDRCPY(ip6po_dest2);
2469 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2473 ip6_clearpktopts(dst, -1);
2476 #undef PKTOPT_EXTHDRCPY
2479 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2484 ip6_clearpktopts(pktopt, -1);
2486 kfree(pktopt, M_IP6OPT);
2490 * Set the IP6 multicast options in response to user setsockopt().
2493 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2496 u_int loop, ifindex;
2497 struct ipv6_mreq *mreq;
2499 struct ip6_moptions *im6o = *im6op;
2500 struct route_in6 ro;
2501 struct sockaddr_in6 *dst;
2502 struct in6_multi_mship *imm;
2503 struct thread *td = curthread; /* XXX */
2507 * No multicast option buffer attached to the pcb;
2508 * allocate one and initialize to default values.
2510 im6o = (struct ip6_moptions *)
2511 kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2514 im6o->im6o_multicast_ifp = NULL;
2515 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2516 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2517 LIST_INIT(&im6o->im6o_memberships);
2522 case IPV6_MULTICAST_IF:
2524 * Select the interface for outgoing multicast packets.
2526 if (m == NULL || m->m_len != sizeof(u_int)) {
2530 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2531 if (ifindex < 0 || if_index < ifindex) {
2532 error = ENXIO; /* XXX EINVAL? */
2535 ifp = ifindex2ifnet[ifindex];
2536 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2537 error = EADDRNOTAVAIL;
2540 im6o->im6o_multicast_ifp = ifp;
2543 case IPV6_MULTICAST_HOPS:
2546 * Set the IP6 hoplimit for outgoing multicast packets.
2549 if (m == NULL || m->m_len != sizeof(int)) {
2553 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2554 if (optval < -1 || optval >= 256)
2556 else if (optval == -1)
2557 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2559 im6o->im6o_multicast_hlim = optval;
2563 case IPV6_MULTICAST_LOOP:
2565 * Set the loopback flag for outgoing multicast packets.
2566 * Must be zero or one.
2568 if (m == NULL || m->m_len != sizeof(u_int)) {
2572 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2577 im6o->im6o_multicast_loop = loop;
2580 case IPV6_JOIN_GROUP:
2582 * Add a multicast group membership.
2583 * Group must be a valid IP6 multicast address.
2585 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2589 mreq = mtod(m, struct ipv6_mreq *);
2590 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2592 * We use the unspecified address to specify to accept
2593 * all multicast addresses. Only super user is allowed
2596 if (priv_check(td, PRIV_ROOT))
2601 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2607 * If the interface is specified, validate it.
2609 if (mreq->ipv6mr_interface < 0
2610 || if_index < mreq->ipv6mr_interface) {
2611 error = ENXIO; /* XXX EINVAL? */
2615 * If no interface was explicitly specified, choose an
2616 * appropriate one according to the given multicast address.
2618 if (mreq->ipv6mr_interface == 0) {
2620 * If the multicast address is in node-local scope,
2621 * the interface should be a loopback interface.
2622 * Otherwise, look up the routing table for the
2623 * address, and choose the outgoing interface.
2624 * XXX: is it a good approach?
2626 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2630 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2631 bzero(dst, sizeof(*dst));
2632 dst->sin6_len = sizeof(struct sockaddr_in6);
2633 dst->sin6_family = AF_INET6;
2634 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2635 rtalloc((struct route *)&ro);
2636 if (ro.ro_rt == NULL) {
2637 error = EADDRNOTAVAIL;
2640 ifp = ro.ro_rt->rt_ifp;
2644 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2647 * See if we found an interface, and confirm that it
2648 * supports multicast
2650 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2651 error = EADDRNOTAVAIL;
2655 * Put interface index into the multicast address,
2656 * if the address has link-local scope.
2658 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2659 mreq->ipv6mr_multiaddr.s6_addr16[1]
2660 = htons(mreq->ipv6mr_interface);
2663 * See if the membership already exists.
2665 for (imm = im6o->im6o_memberships.lh_first;
2666 imm != NULL; imm = imm->i6mm_chain.le_next)
2667 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2668 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2669 &mreq->ipv6mr_multiaddr))
2676 * Everything looks good; add a new record to the multicast
2677 * address list for the given interface.
2679 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2680 if ((imm->i6mm_maddr =
2681 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2682 kfree(imm, M_IPMADDR);
2685 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2688 case IPV6_LEAVE_GROUP:
2690 * Drop a multicast group membership.
2691 * Group must be a valid IP6 multicast address.
2693 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2697 mreq = mtod(m, struct ipv6_mreq *);
2698 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2699 if (priv_check(td, PRIV_ROOT)) {
2703 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2708 * If an interface address was specified, get a pointer
2709 * to its ifnet structure.
2711 if (mreq->ipv6mr_interface < 0
2712 || if_index < mreq->ipv6mr_interface) {
2713 error = ENXIO; /* XXX EINVAL? */
2716 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2718 * Put interface index into the multicast address,
2719 * if the address has link-local scope.
2721 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2722 mreq->ipv6mr_multiaddr.s6_addr16[1]
2723 = htons(mreq->ipv6mr_interface);
2726 * Find the membership in the membership list.
2728 for (imm = im6o->im6o_memberships.lh_first;
2729 imm != NULL; imm = imm->i6mm_chain.le_next) {
2731 imm->i6mm_maddr->in6m_ifp == ifp) &&
2732 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2733 &mreq->ipv6mr_multiaddr))
2737 /* Unable to resolve interface */
2738 error = EADDRNOTAVAIL;
2742 * Give up the multicast address record to which the
2743 * membership points.
2745 LIST_REMOVE(imm, i6mm_chain);
2746 in6_delmulti(imm->i6mm_maddr);
2747 kfree(imm, M_IPMADDR);
2756 * If all options have default values, no need to keep the mbuf.
2758 if (im6o->im6o_multicast_ifp == NULL &&
2759 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2760 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2761 im6o->im6o_memberships.lh_first == NULL) {
2762 kfree(*im6op, M_IPMOPTS);
2770 * Return the IP6 multicast options in response to user getsockopt().
2773 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2775 u_int *hlim, *loop, *ifindex;
2777 *mp = m_get(MB_WAIT, MT_HEADER); /* XXX */
2781 case IPV6_MULTICAST_IF:
2782 ifindex = mtod(*mp, u_int *);
2783 (*mp)->m_len = sizeof(u_int);
2784 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2787 *ifindex = im6o->im6o_multicast_ifp->if_index;
2790 case IPV6_MULTICAST_HOPS:
2791 hlim = mtod(*mp, u_int *);
2792 (*mp)->m_len = sizeof(u_int);
2794 *hlim = ip6_defmcasthlim;
2796 *hlim = im6o->im6o_multicast_hlim;
2799 case IPV6_MULTICAST_LOOP:
2800 loop = mtod(*mp, u_int *);
2801 (*mp)->m_len = sizeof(u_int);
2803 *loop = ip6_defmcasthlim;
2805 *loop = im6o->im6o_multicast_loop;
2809 return (EOPNOTSUPP);
2814 * Discard the IP6 multicast options.
2817 ip6_freemoptions(struct ip6_moptions *im6o)
2819 struct in6_multi_mship *imm;
2824 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2825 LIST_REMOVE(imm, i6mm_chain);
2826 if (imm->i6mm_maddr)
2827 in6_delmulti(imm->i6mm_maddr);
2828 kfree(imm, M_IPMADDR);
2830 kfree(im6o, M_IPMOPTS);
2834 * Set a particular packet option, as a sticky option or an ancillary data
2835 * item. "len" can be 0 only when it's a sticky option.
2836 * We have 4 cases of combination of "sticky" and "cmsg":
2837 * "sticky=0, cmsg=0": impossible
2838 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2839 * "sticky=1, cmsg=0": RFC3542 socket option
2840 * "sticky=1, cmsg=1": RFC2292 socket option
2843 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2844 int sticky, int cmsg, int uproto, int priv)
2846 int minmtupolicy, preftemp;
2849 if (!sticky && !cmsg) {
2850 kprintf("ip6_setpktoption: impossible case\n");
2855 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2856 * not be specified in the context of RFC3542. Conversely,
2857 * RFC3542 types should not be specified in the context of RFC2292.
2861 case IPV6_2292PKTINFO:
2862 case IPV6_2292HOPLIMIT:
2863 case IPV6_2292NEXTHOP:
2864 case IPV6_2292HOPOPTS:
2865 case IPV6_2292DSTOPTS:
2866 case IPV6_2292RTHDR:
2867 case IPV6_2292PKTOPTIONS:
2868 return (ENOPROTOOPT);
2871 if (sticky && cmsg) {
2878 case IPV6_RTHDRDSTOPTS:
2880 case IPV6_USE_MIN_MTU:
2883 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2884 return (ENOPROTOOPT);
2889 case IPV6_2292PKTINFO:
2892 struct in6_pktinfo *pktinfo;
2893 if (len != sizeof(struct in6_pktinfo))
2895 pktinfo = (struct in6_pktinfo *)buf;
2898 * An application can clear any sticky IPV6_PKTINFO option by
2899 * doing a "regular" setsockopt with ipi6_addr being
2900 * in6addr_any and ipi6_ifindex being zero.
2901 * [RFC 3542, Section 6]
2903 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2904 pktinfo->ipi6_ifindex == 0 &&
2905 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2906 ip6_clearpktopts(opt, optname);
2910 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2911 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2915 /* validate the interface index if specified. */
2916 if (pktinfo->ipi6_ifindex > if_index ||
2917 pktinfo->ipi6_ifindex < 0) {
2921 * Check if the requested source address is indeed a
2922 * unicast address assigned to the node, and can be
2923 * used as the packet's source address.
2925 if (opt->ip6po_pktinfo != NULL &&
2926 !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2927 struct in6_ifaddr *ia6;
2928 struct sockaddr_in6 sin6;
2930 bzero(&sin6, sizeof(sin6));
2931 sin6.sin6_len = sizeof(sin6);
2932 sin6.sin6_family = AF_INET6;
2934 opt->ip6po_pktinfo->ipi6_addr;
2935 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2937 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2938 IN6_IFF_NOTREADY)) != 0)
2939 return (EADDRNOTAVAIL);
2943 * We store the address anyway, and let in6_selectsrc()
2944 * validate the specified address. This is because ipi6_addr
2945 * may not have enough information about its scope zone, and
2946 * we may need additional information (such as outgoing
2947 * interface or the scope zone of a destination address) to
2948 * disambiguate the scope.
2949 * XXX: the delay of the validation may confuse the
2950 * application when it is used as a sticky option.
2952 if (opt->ip6po_pktinfo == NULL) {
2953 opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2954 M_IP6OPT, M_NOWAIT);
2955 if (opt->ip6po_pktinfo == NULL)
2958 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2962 case IPV6_2292HOPLIMIT:
2968 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2969 * to simplify the ordering among hoplimit options.
2971 if (optname == IPV6_HOPLIMIT && sticky)
2972 return (ENOPROTOOPT);
2974 if (len != sizeof(int))
2977 if (*hlimp < -1 || *hlimp > 255)
2980 opt->ip6po_hlim = *hlimp;
2988 if (len != sizeof(int))
2990 tclass = *(int *)buf;
2991 if (tclass < -1 || tclass > 255)
2994 opt->ip6po_tclass = tclass;
2998 case IPV6_2292NEXTHOP:
3003 if (len == 0) { /* just remove the option */
3004 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3008 /* check if cmsg_len is large enough for sa_len */
3009 if (len < sizeof(struct sockaddr) || len < *buf)
3012 switch (((struct sockaddr *)buf)->sa_family) {
3015 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3018 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3021 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3022 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3027 case AF_LINK: /* should eventually be supported */
3029 return (EAFNOSUPPORT);
3032 /* turn off the previous option, then set the new option. */
3033 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3034 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3035 if (opt->ip6po_nexthop == NULL)
3037 bcopy(buf, opt->ip6po_nexthop, *buf);
3040 case IPV6_2292HOPOPTS:
3043 struct ip6_hbh *hbh;
3047 * XXX: We don't allow a non-privileged user to set ANY HbH
3048 * options, since per-option restriction has too much
3054 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3055 break; /* just remove the option */
3058 /* message length validation */
3059 if (len < sizeof(struct ip6_hbh))
3061 hbh = (struct ip6_hbh *)buf;
3062 hbhlen = (hbh->ip6h_len + 1) << 3;
3066 /* turn off the previous option, then set the new option. */
3067 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3068 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3069 if (opt->ip6po_hbh == NULL)
3071 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3076 case IPV6_2292DSTOPTS:
3078 case IPV6_RTHDRDSTOPTS:
3080 struct ip6_dest *dest, **newdest = NULL;
3086 ip6_clearpktopts(opt, optname);
3087 break; /* just remove the option */
3090 /* message length validation */
3091 if (len < sizeof(struct ip6_dest))
3093 dest = (struct ip6_dest *)buf;
3094 destlen = (dest->ip6d_len + 1) << 3;
3099 * Determine the position that the destination options header
3100 * should be inserted; before or after the routing header.
3103 case IPV6_2292DSTOPTS:
3105 * The old advacned API is ambiguous on this point.
3106 * Our approach is to determine the position based
3107 * according to the existence of a routing header.
3108 * Note, however, that this depends on the order of the
3109 * extension headers in the ancillary data; the 1st
3110 * part of the destination options header must appear
3111 * before the routing header in the ancillary data,
3113 * RFC3542 solved the ambiguity by introducing
3114 * separate ancillary data or option types.
3116 if (opt->ip6po_rthdr == NULL)
3117 newdest = &opt->ip6po_dest1;
3119 newdest = &opt->ip6po_dest2;
3121 case IPV6_RTHDRDSTOPTS:
3122 newdest = &opt->ip6po_dest1;
3125 newdest = &opt->ip6po_dest2;
3129 /* turn off the previous option, then set the new option. */
3130 ip6_clearpktopts(opt, optname);
3131 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3132 if (*newdest == NULL)
3134 bcopy(dest, *newdest, destlen);
3139 case IPV6_2292RTHDR:
3142 struct ip6_rthdr *rth;
3146 ip6_clearpktopts(opt, IPV6_RTHDR);
3147 break; /* just remove the option */
3150 /* message length validation */
3151 if (len < sizeof(struct ip6_rthdr))
3153 rth = (struct ip6_rthdr *)buf;
3154 rthlen = (rth->ip6r_len + 1) << 3;
3158 switch (rth->ip6r_type) {
3160 return (EINVAL); /* not supported */
3163 /* turn off the previous option */
3164 ip6_clearpktopts(opt, IPV6_RTHDR);
3165 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3166 if (opt->ip6po_rthdr == NULL)
3168 bcopy(rth, opt->ip6po_rthdr, rthlen);
3173 case IPV6_USE_MIN_MTU:
3174 if (len != sizeof(int))
3176 minmtupolicy = *(int *)buf;
3177 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3178 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3179 minmtupolicy != IP6PO_MINMTU_ALL) {
3182 opt->ip6po_minmtu = minmtupolicy;
3186 if (len != sizeof(int))
3189 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3191 * we ignore this option for TCP sockets.
3192 * (RFC3542 leaves this case unspecified.)
3194 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3196 opt->ip6po_flags |= IP6PO_DONTFRAG;
3199 case IPV6_PREFER_TEMPADDR:
3200 if (len != sizeof(int))
3202 preftemp = *(int *)buf;
3203 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3204 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3205 preftemp != IP6PO_TEMPADDR_PREFER) {
3208 opt->ip6po_prefer_tempaddr = preftemp;
3212 return (ENOPROTOOPT);
3213 } /* end of switch */
3220 * Set IPv6 outgoing packet options based on advanced API.
3223 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3224 struct ip6_pktopts *stickyopt, int uproto, int priv)
3226 struct cmsghdr *cm = NULL;
3228 if (control == NULL || opt == NULL)
3231 init_ip6pktopts(opt);
3234 * XXX: Currently, we assume all the optional information is stored
3241 * If stickyopt is provided, make a local copy of the options
3242 * for this particular packet, then override them by ancillary
3244 * XXX: copypktopts() does not copy the cached route to a next
3245 * hop (if any). This is not very good in terms of efficiency,
3246 * but we can allow this since this option should be rarely
3249 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3254 * XXX: Currently, we assume all the optional information is stored
3257 if (control->m_next)
3260 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3261 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3264 if (control->m_len < CMSG_LEN(0))
3267 cm = mtod(control, struct cmsghdr *);
3268 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3270 if (cm->cmsg_level != IPPROTO_IPV6)
3273 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3274 cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3283 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3284 * packet to the input queue of a specified interface. Note that this
3285 * calls the output routine of the loopback "driver", but with an interface
3286 * pointer that might NOT be &loif -- easier than replicating that code here.
3289 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3292 struct ip6_hdr *ip6;
3294 copym = m_copy(m, 0, M_COPYALL);
3299 * Make sure to deep-copy IPv6 header portion in case the data
3300 * is in an mbuf cluster, so that we can safely override the IPv6
3301 * header portion later.
3303 if ((copym->m_flags & M_EXT) != 0 ||
3304 copym->m_len < sizeof(struct ip6_hdr)) {
3305 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3311 if (copym->m_len < sizeof(*ip6)) {
3317 ip6 = mtod(copym, struct ip6_hdr *);
3319 * clear embedded scope identifiers if necessary.
3320 * in6_clearscope will touch the addresses only when necessary.
3322 in6_clearscope(&ip6->ip6_src);
3323 in6_clearscope(&ip6->ip6_dst);
3325 if_simloop(ifp, copym, dst->sin6_family, 0);
3329 * Separate the IPv6 header from the payload into its own mbuf.
3331 * Returns the new mbuf chain or the original mbuf if no payload.
3332 * Returns NULL if can't allocate new mbuf for header.
3334 static struct mbuf *
3335 ip6_splithdr(struct mbuf *m)
3339 if (m->m_len <= sizeof(struct ip6_hdr)) /* no payload */
3342 MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
3345 mh->m_len = sizeof(struct ip6_hdr);
3346 M_MOVE_PKTHDR(mh, m);
3347 MH_ALIGN(mh, sizeof(struct ip6_hdr));
3348 bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3349 m->m_data += sizeof(struct ip6_hdr);
3350 m->m_len -= sizeof(struct ip6_hdr);
3356 * Compute IPv6 extension header length.
3359 ip6_optlen(struct in6pcb *in6p)
3363 if (!in6p->in6p_outputopts)
3368 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3370 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3371 if (in6p->in6p_outputopts->ip6po_rthdr)
3372 /* dest1 is valid with rthdr only */
3373 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3374 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3375 len += elen(in6p->in6p_outputopts->ip6po_dest2);