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. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
64 #include "opt_ip6fw.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
69 #include <sys/param.h>
70 #include <sys/malloc.h>
72 #include <sys/errno.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/systm.h>
77 #include <sys/kernel.h>
81 #include <sys/thread2.h>
82 #include <sys/msgport2.h>
85 #include <net/route.h>
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet6/nd6.h>
96 #include <netinet6/ip6protosw.h>
99 #include <netinet6/ipsec.h>
101 #include <netinet6/ipsec6.h>
103 #include <netproto/key/key.h>
107 #include <netproto/ipsec/ipsec.h>
108 #include <netproto/ipsec/ipsec6.h>
109 #include <netproto/ipsec/key.h>
112 #include <net/ip6fw/ip6_fw.h>
114 #include <net/net_osdep.h>
116 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
119 struct mbuf *ip6e_ip6;
120 struct mbuf *ip6e_hbh;
121 struct mbuf *ip6e_dest1;
122 struct mbuf *ip6e_rthdr;
123 struct mbuf *ip6e_dest2;
126 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
127 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
129 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
130 struct socket *, struct sockopt *);
131 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
132 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
133 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
134 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
135 struct ifnet *, struct in6_addr *, u_long *, int *);
136 static int copyexthdr (void *, struct mbuf **);
137 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
139 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
140 static struct mbuf *ip6_splithdr (struct mbuf *);
141 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
144 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
145 * header (with pri, len, nxt, hlim, src, dst).
146 * This function may modify ver and hlim only.
147 * The mbuf chain containing the packet will be freed.
148 * The mbuf opt, if present, will not be freed.
150 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
151 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
152 * which is rt_rmx.rmx_mtu.
155 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
156 int flags, struct ip6_moptions *im6o,
157 struct ifnet **ifpp, /* XXX: just for statistics */
160 struct ip6_hdr *ip6, *mhip6;
161 struct ifnet *ifp, *origifp;
165 int hlen, tlen, len, off;
166 struct route_in6 ip6route;
167 struct sockaddr_in6 *dst;
169 struct in6_ifaddr *ia = NULL;
171 int alwaysfrag, dontfrag;
172 u_int32_t optlen, plen = 0, unfragpartlen;
173 struct ip6_exthdrs exthdrs;
174 struct in6_addr finaldst;
175 struct route_in6 *ro_pmtu = NULL;
176 boolean_t hdrsplit = FALSE;
177 boolean_t needipsec = FALSE;
179 boolean_t needipsectun = FALSE;
180 struct secpolicy *sp = NULL;
181 struct socket *so = inp ? inp->inp_socket : NULL;
183 ip6 = mtod(m, struct ip6_hdr *);
186 boolean_t needipsectun = FALSE;
187 struct secpolicy *sp = NULL;
189 ip6 = mtod(m, struct ip6_hdr *);
192 bzero(&exthdrs, sizeof exthdrs);
195 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
197 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
199 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
201 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
206 /* get a security policy for this packet */
208 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
210 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
213 ipsec6stat.out_inval++;
220 switch (sp->policy) {
221 case IPSEC_POLICY_DISCARD:
223 * This packet is just discarded.
225 ipsec6stat.out_polvio++;
228 case IPSEC_POLICY_BYPASS:
229 case IPSEC_POLICY_NONE:
230 /* no need to do IPsec. */
234 case IPSEC_POLICY_IPSEC:
235 if (sp->req == NULL) {
236 error = key_spdacquire(sp); /* acquire a policy */
242 case IPSEC_POLICY_ENTRUST:
244 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
248 /* get a security policy for this packet */
250 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
252 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
255 newipsecstat.ips_out_inval++;
262 switch (sp->policy) {
263 case IPSEC_POLICY_DISCARD:
265 * This packet is just discarded.
267 newipsecstat.ips_out_polvio++;
270 case IPSEC_POLICY_BYPASS:
271 case IPSEC_POLICY_NONE:
272 /* no need to do IPsec. */
276 case IPSEC_POLICY_IPSEC:
277 if (sp->req == NULL) {
278 error = key_spdacquire(sp); /* acquire a policy */
284 case IPSEC_POLICY_ENTRUST:
286 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
288 #endif /* FAST_IPSEC */
291 * Calculate the total length of the extension header chain.
292 * Keep the length of the unfragmentable part for fragmentation.
294 optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
295 m_lengthm(exthdrs.ip6e_dest1, NULL) +
296 m_lengthm(exthdrs.ip6e_rthdr, NULL);
298 unfragpartlen = optlen + sizeof(struct ip6_hdr);
300 /* NOTE: we don't add AH/ESP length here. do that later. */
301 optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
304 * If we need IPsec, or there is at least one extension header,
305 * separate IP6 header from the payload.
307 if ((needipsec || optlen) && !hdrsplit) {
308 exthdrs.ip6e_ip6 = ip6_splithdr(m);
309 if (exthdrs.ip6e_ip6 == NULL) {
313 m = exthdrs.ip6e_ip6;
318 ip6 = mtod(m, struct ip6_hdr *);
320 /* adjust mbuf packet header length */
321 m->m_pkthdr.len += optlen;
322 plen = m->m_pkthdr.len - sizeof(*ip6);
324 /* If this is a jumbo payload, insert a jumbo payload option. */
325 if (plen > IPV6_MAXPACKET) {
327 exthdrs.ip6e_ip6 = ip6_splithdr(m);
328 if (exthdrs.ip6e_ip6 == NULL) {
332 m = exthdrs.ip6e_ip6;
336 ip6 = mtod(m, struct ip6_hdr *);
337 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
341 ip6->ip6_plen = htons(plen);
344 * Concatenate headers and fill in next header fields.
345 * Here we have, on "m"
347 * and we insert headers accordingly. Finally, we should be getting:
348 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
350 * during the header composing process, "m" points to IPv6 header.
351 * "mprev" points to an extension header prior to esp.
354 nexthdrp = &ip6->ip6_nxt;
358 * we treat dest2 specially. this makes IPsec processing
359 * much easier. the goal here is to make mprev point the
360 * mbuf prior to dest2.
362 * result: IPv6 dest2 payload
363 * m and mprev will point to IPv6 header.
365 if (exthdrs.ip6e_dest2) {
367 panic("assumption failed: hdr not split");
368 exthdrs.ip6e_dest2->m_next = m->m_next;
369 m->m_next = exthdrs.ip6e_dest2;
370 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
371 ip6->ip6_nxt = IPPROTO_DSTOPTS;
375 * Place m1 after mprev.
377 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
381 panic("assumption failed: hdr not split");\
382 *mtod(m1, u_char *) = *nexthdrp;\
384 nexthdrp = mtod(m1, u_char *);\
385 m1->m_next = mprev->m_next;\
392 * result: IPv6 hbh dest1 rthdr dest2 payload
393 * m will point to IPv6 header. mprev will point to the
394 * extension header prior to dest2 (rthdr in the above case).
396 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
397 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
398 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
400 #if defined(IPSEC) || defined(FAST_IPSEC)
402 struct ipsec_output_state state;
404 struct ip6_rthdr *rh = NULL;
407 * pointers after IPsec headers are not valid any more.
408 * other pointers need a great care too.
409 * (IPsec routines should not mangle mbufs prior to AH/ESP)
411 exthdrs.ip6e_dest2 = NULL;
413 if (exthdrs.ip6e_rthdr) {
414 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
415 segleft_org = rh->ip6r_segleft;
416 rh->ip6r_segleft = 0;
419 bzero(&state, sizeof state);
421 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
425 /* mbuf is already reclaimed in ipsec6_output_trans. */
435 kprintf("ip6_output (ipsec): error code %d\n",
439 /* don't show these error codes to the user */
445 if (exthdrs.ip6e_rthdr) {
446 /* ah6_output doesn't modify mbuf chain */
447 rh->ip6r_segleft = segleft_org;
453 * If there is a routing header, replace destination address field
454 * with the first hop of the routing header.
456 if (exthdrs.ip6e_rthdr) {
457 struct ip6_rthdr *rh;
459 finaldst = ip6->ip6_dst;
460 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
461 switch (rh->ip6r_type) {
462 default: /* is it possible? */
468 /* Source address validation */
469 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
470 !(flags & IPV6_DADOUTPUT)) {
472 ip6stat.ip6s_badscope++;
475 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
477 ip6stat.ip6s_badscope++;
481 ip6stat.ip6s_localout++;
488 bzero(ro, sizeof(*ro));
491 if (opt && opt->ip6po_rthdr)
492 ro = &opt->ip6po_route;
493 dst = (struct sockaddr_in6 *)&ro->ro_dst;
495 * If there is a cached route,
496 * check that it is to the same destination
497 * and is still up. If not, free it and try again.
499 if (ro->ro_rt != NULL &&
500 (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
501 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
505 if (ro->ro_rt == NULL) {
506 bzero(dst, sizeof(*dst));
507 dst->sin6_family = AF_INET6;
508 dst->sin6_len = sizeof(struct sockaddr_in6);
509 dst->sin6_addr = ip6->ip6_dst;
511 #if defined(IPSEC) || defined(FAST_IPSEC)
512 if (needipsec && needipsectun) {
513 struct ipsec_output_state state;
516 * All the extension headers will become inaccessible
517 * (since they can be encrypted).
518 * Don't panic, we need no more updates to extension headers
519 * on inner IPv6 packet (since they are now encapsulated).
521 * IPv6 [ESP|AH] IPv6 [extension headers] payload
523 bzero(&exthdrs, sizeof(exthdrs));
524 exthdrs.ip6e_ip6 = m;
526 bzero(&state, sizeof(state));
528 state.ro = (struct route *)ro;
529 state.dst = (struct sockaddr *)dst;
531 error = ipsec6_output_tunnel(&state, sp, flags);
534 ro = (struct route_in6 *)state.ro;
535 dst = (struct sockaddr_in6 *)state.dst;
537 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
548 kprintf("ip6_output (ipsec): error code %d\n", error);
551 /* don't show these error codes to the user */
558 exthdrs.ip6e_ip6 = m;
562 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
565 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
566 #define sin6tosa(sin6) ((struct sockaddr *)(sin6))
568 * interface selection comes here
569 * if an interface is specified from an upper layer,
572 if (ro->ro_rt == NULL) {
574 * non-bsdi always clone routes, if parent is
577 rtalloc((struct route *)ro);
579 if (ro->ro_rt == NULL) {
580 ip6stat.ip6s_noroute++;
581 error = EHOSTUNREACH;
582 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
585 ia = ifatoia6(ro->ro_rt->rt_ifa);
586 ifp = ro->ro_rt->rt_ifp;
588 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
589 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
590 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
592 in6_ifstat_inc(ifp, ifs6_out_request);
595 * Check if the outgoing interface conflicts with
596 * the interface specified by ifi6_ifindex (if specified).
597 * Note that loopback interface is always okay.
598 * (this may happen when we are sending a packet to one of
599 * our own addresses.)
601 if (opt && opt->ip6po_pktinfo
602 && opt->ip6po_pktinfo->ipi6_ifindex) {
603 if (!(ifp->if_flags & IFF_LOOPBACK)
604 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
605 ip6stat.ip6s_noroute++;
606 in6_ifstat_inc(ifp, ifs6_out_discard);
607 error = EHOSTUNREACH;
612 if (opt && opt->ip6po_hlim != -1)
613 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
616 struct in6_multi *in6m;
618 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
621 * See if the caller provided any multicast options
625 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
626 if (im6o->im6o_multicast_ifp != NULL)
627 ifp = im6o->im6o_multicast_ifp;
629 ip6->ip6_hlim = ip6_defmcasthlim;
632 * See if the caller provided the outgoing interface
633 * as an ancillary data.
634 * Boundary check for ifindex is assumed to be already done.
636 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
637 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
640 * If the destination is a node-local scope multicast,
641 * the packet should be loop-backed only.
643 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
645 * If the outgoing interface is already specified,
646 * it should be a loopback interface.
648 if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
649 ip6stat.ip6s_badscope++;
650 error = ENETUNREACH; /* XXX: better error? */
651 /* XXX correct ifp? */
652 in6_ifstat_inc(ifp, ifs6_out_discard);
659 if (opt && opt->ip6po_hlim != -1)
660 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
663 * If caller did not provide an interface lookup a
664 * default in the routing table. This is either a
665 * default for the speicfied group (i.e. a host
666 * route), or a multicast default (a route for the
670 if (ro->ro_rt == NULL) {
672 rtpurelookup((struct sockaddr *)&ro->ro_dst);
674 if (ro->ro_rt == NULL) {
675 ip6stat.ip6s_noroute++;
676 error = EHOSTUNREACH;
677 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
680 ia = ifatoia6(ro->ro_rt->rt_ifa);
681 ifp = ro->ro_rt->rt_ifp;
685 if (!(flags & IPV6_FORWARDING))
686 in6_ifstat_inc(ifp, ifs6_out_request);
687 in6_ifstat_inc(ifp, ifs6_out_mcast);
690 * Confirm that the outgoing interface supports multicast.
692 if (!(ifp->if_flags & IFF_MULTICAST)) {
693 ip6stat.ip6s_noroute++;
694 in6_ifstat_inc(ifp, ifs6_out_discard);
698 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp);
700 (im6o == NULL || im6o->im6o_multicast_loop)) {
702 * If we belong to the destination multicast group
703 * on the outgoing interface, and the caller did not
704 * forbid loopback, loop back a copy.
706 ip6_mloopback(ifp, m, dst);
709 * If we are acting as a multicast router, perform
710 * multicast forwarding as if the packet had just
711 * arrived on the interface to which we are about
712 * to send. The multicast forwarding function
713 * recursively calls this function, using the
714 * IPV6_FORWARDING flag to prevent infinite recursion.
716 * Multicasts that are looped back by ip6_mloopback(),
717 * above, will be forwarded by the ip6_input() routine,
720 if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
721 if (ip6_mforward(ip6, ifp, m) != 0) {
728 * Multicasts with a hoplimit of zero may be looped back,
729 * above, but must not be transmitted on a network.
730 * Also, multicasts addressed to the loopback interface
731 * are not sent -- the above call to ip6_mloopback() will
732 * loop back a copy if this host actually belongs to the
733 * destination group on the loopback interface.
735 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
742 * Fill the outgoing inteface to tell the upper layer
743 * to increment per-interface statistics.
748 /* Determine path MTU. */
749 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
754 * The caller of this function may specify to use the minimum MTU
756 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
757 * setting. The logic is a bit complicated; by default, unicast
758 * packets will follow path MTU while multicast packets will be sent at
759 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
760 * including unicast ones will be sent at the minimum MTU. Multicast
761 * packets will always be sent at the minimum MTU unless
762 * IP6PO_MINMTU_DISABLE is explicitly specified.
763 * See RFC 3542 for more details.
765 if (mtu > IPV6_MMTU) {
766 if ((flags & IPV6_MINMTU))
768 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
770 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
772 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
777 /* Fake scoped addresses */
778 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
780 * If source or destination address is a scoped address, and
781 * the packet is going to be sent to a loopback interface,
782 * we should keep the original interface.
786 * XXX: this is a very experimental and temporary solution.
787 * We eventually have sockaddr_in6 and use the sin6_scope_id
788 * field of the structure here.
789 * We rely on the consistency between two scope zone ids
790 * of source and destination, which should already be assured.
791 * Larger scopes than link will be supported in the future.
794 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
795 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
796 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
797 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
799 * XXX: origifp can be NULL even in those two cases above.
800 * For example, if we remove the (only) link-local address
801 * from the loopback interface, and try to send a link-local
802 * address without link-id information. Then the source
803 * address is ::1, and the destination address is the
804 * link-local address with its s6_addr16[1] being zero.
805 * What is worse, if the packet goes to the loopback interface
806 * by a default rejected route, the null pointer would be
807 * passed to looutput, and the kernel would hang.
808 * The following last resort would prevent such disaster.
816 * clear embedded scope identifiers if necessary.
817 * in6_clearscope will touch the addresses only when necessary.
819 in6_clearscope(&ip6->ip6_src);
820 in6_clearscope(&ip6->ip6_dst);
823 * Check with the firewall...
825 if (ip6_fw_enable && ip6_fw_chk_ptr) {
828 m->m_pkthdr.rcvif = NULL; /* XXX */
829 /* If ipfw says divert, we have to just drop packet */
830 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
841 * If the outgoing packet contains a hop-by-hop options header,
842 * it must be examined and processed even by the source node.
843 * (RFC 2460, section 4.)
845 if (exthdrs.ip6e_hbh) {
846 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
847 u_int32_t dummy1; /* XXX unused */
848 u_int32_t dummy2; /* XXX unused */
851 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
852 panic("ip6e_hbh is not continuous");
855 * XXX: if we have to send an ICMPv6 error to the sender,
856 * we need the M_LOOP flag since icmp6_error() expects
857 * the IPv6 and the hop-by-hop options header are
858 * continuous unless the flag is set.
860 m->m_flags |= M_LOOP;
861 m->m_pkthdr.rcvif = ifp;
862 if (ip6_process_hopopts(m,
863 (u_int8_t *)(hbh + 1),
864 ((hbh->ip6h_len + 1) << 3) -
865 sizeof(struct ip6_hbh),
866 &dummy1, &dummy2) < 0) {
867 /* m was already freed at this point */
868 error = EINVAL;/* better error? */
871 m->m_flags &= ~M_LOOP; /* XXX */
872 m->m_pkthdr.rcvif = NULL;
876 * Run through list of hooks for output packets.
878 if (pfil_has_hooks(&inet6_pfil_hook)) {
879 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
880 if (error != 0 || m == NULL)
882 ip6 = mtod(m, struct ip6_hdr *);
886 * Send the packet to the outgoing interface.
887 * If necessary, do IPv6 fragmentation before sending.
889 * the logic here is rather complex:
890 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
891 * 1-a: send as is if tlen <= path mtu
892 * 1-b: fragment if tlen > path mtu
894 * 2: if user asks us not to fragment (dontfrag == 1)
895 * 2-a: send as is if tlen <= interface mtu
896 * 2-b: error if tlen > interface mtu
898 * 3: if we always need to attach fragment header (alwaysfrag == 1)
901 * 4: if dontfrag == 1 && alwaysfrag == 1
902 * error, as we cannot handle this conflicting request
904 tlen = m->m_pkthdr.len;
906 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
910 if (dontfrag && alwaysfrag) { /* case 4 */
911 /* conflicting request - can't transmit */
915 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */
917 * Even if the DONTFRAG option is specified, we cannot send the
918 * packet when the data length is larger than the MTU of the
919 * outgoing interface.
920 * Notify the error by sending IPV6_PATHMTU ancillary data as
921 * well as returning an error code (the latter is not described
925 struct ip6ctlparam ip6cp;
927 mtu32 = (u_int32_t)mtu;
928 bzero(&ip6cp, sizeof(ip6cp));
929 ip6cp.ip6c_cmdarg = (void *)&mtu32;
930 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
938 * transmit packet without fragmentation
940 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
941 struct in6_ifaddr *ia6;
943 ip6 = mtod(m, struct ip6_hdr *);
944 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
946 /* Record statistics for this interface address. */
947 IFA_STAT_INC(&ia6->ia_ifa, opackets, 1);
948 IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len);
951 /* clean ipsec history once it goes out of the node */
954 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
959 * try to fragment the packet. case 1-b and 3
961 if (mtu < IPV6_MMTU) {
963 * note that path MTU is never less than IPV6_MMTU
967 in6_ifstat_inc(ifp, ifs6_out_fragfail);
969 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
971 in6_ifstat_inc(ifp, ifs6_out_fragfail);
974 struct mbuf **mnext, *m_frgpart;
975 struct ip6_frag *ip6f;
976 u_int32_t id = htonl(ip6_id++);
980 * Too large for the destination or interface;
981 * fragment if possible.
982 * Must be able to put at least 8 bytes per fragment.
984 hlen = unfragpartlen;
985 if (mtu > IPV6_MAXPACKET)
986 mtu = IPV6_MAXPACKET;
988 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
991 in6_ifstat_inc(ifp, ifs6_out_fragfail);
995 mnext = &m->m_nextpkt;
998 * Change the next header field of the last header in the
999 * unfragmentable part.
1001 if (exthdrs.ip6e_rthdr) {
1002 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1003 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1004 } else if (exthdrs.ip6e_dest1) {
1005 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1006 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1007 } else if (exthdrs.ip6e_hbh) {
1008 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1009 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1011 nextproto = ip6->ip6_nxt;
1012 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1016 * Loop through length of segment after first fragment,
1017 * make new header and copy data of each part and link onto
1021 for (off = hlen; off < tlen; off += len) {
1022 MGETHDR(m, M_NOWAIT, MT_HEADER);
1025 ip6stat.ip6s_odropped++;
1028 m->m_pkthdr.rcvif = NULL;
1029 m->m_flags = m0->m_flags & M_COPYFLAGS;
1031 mnext = &m->m_nextpkt;
1032 m->m_data += max_linkhdr;
1033 mhip6 = mtod(m, struct ip6_hdr *);
1035 m->m_len = sizeof(*mhip6);
1036 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1038 ip6stat.ip6s_odropped++;
1041 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1042 if (off + len >= tlen)
1045 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1046 mhip6->ip6_plen = htons((u_short)(len + hlen +
1048 sizeof(struct ip6_hdr)));
1049 if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1051 ip6stat.ip6s_odropped++;
1054 m_cat(m, m_frgpart);
1055 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1056 m->m_pkthdr.rcvif = NULL;
1057 ip6f->ip6f_reserved = 0;
1058 ip6f->ip6f_ident = id;
1059 ip6f->ip6f_nxt = nextproto;
1060 ip6stat.ip6s_ofragments++;
1061 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1064 in6_ifstat_inc(ifp, ifs6_out_fragok);
1068 * Remove leading garbages.
1072 m0->m_nextpkt = NULL;
1074 for (m0 = m; m; m = m0) {
1076 m->m_nextpkt = NULL;
1078 /* Record statistics for this interface address. */
1080 IFA_STAT_INC(&ia->ia_ifa, opackets, 1);
1081 IFA_STAT_INC(&ia->ia_ifa, obytes,
1085 /* clean ipsec history once it goes out of the node */
1088 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1094 ip6stat.ip6s_fragmented++;
1097 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1099 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1100 RTFREE(ro_pmtu->ro_rt);
1115 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1116 m_freem(exthdrs.ip6e_dest1);
1117 m_freem(exthdrs.ip6e_rthdr);
1118 m_freem(exthdrs.ip6e_dest2);
1126 copyexthdr(void *h, struct mbuf **mp)
1128 struct ip6_ext *hdr = h;
1135 hlen = (hdr->ip6e_len + 1) * 8;
1136 if (hlen > MCLBYTES)
1137 return ENOBUFS; /* XXX */
1139 m = m_getb(hlen, M_NOWAIT, MT_DATA, 0);
1144 bcopy(hdr, mtod(m, caddr_t), hlen);
1151 * Insert jumbo payload option.
1154 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1160 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1163 * If there is no hop-by-hop options header, allocate new one.
1164 * If there is one but it doesn't have enough space to store the
1165 * jumbo payload option, allocate a cluster to store the whole options.
1166 * Otherwise, use it to store the options.
1168 if (exthdrs->ip6e_hbh == NULL) {
1169 MGET(mopt, M_NOWAIT, MT_DATA);
1172 mopt->m_len = JUMBOOPTLEN;
1173 optbuf = mtod(mopt, u_char *);
1174 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1175 exthdrs->ip6e_hbh = mopt;
1177 struct ip6_hbh *hbh;
1179 mopt = exthdrs->ip6e_hbh;
1180 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1183 * - exthdrs->ip6e_hbh is not referenced from places
1184 * other than exthdrs.
1185 * - exthdrs->ip6e_hbh is not an mbuf chain.
1187 int oldoptlen = mopt->m_len;
1191 * XXX: give up if the whole (new) hbh header does
1192 * not fit even in an mbuf cluster.
1194 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1198 * As a consequence, we must always prepare a cluster
1201 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1204 n->m_len = oldoptlen + JUMBOOPTLEN;
1205 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1206 optbuf = mtod(n, caddr_t) + oldoptlen;
1208 mopt = exthdrs->ip6e_hbh = n;
1210 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1211 mopt->m_len += JUMBOOPTLEN;
1213 optbuf[0] = IP6OPT_PADN;
1217 * Adjust the header length according to the pad and
1218 * the jumbo payload option.
1220 hbh = mtod(mopt, struct ip6_hbh *);
1221 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1224 /* fill in the option. */
1225 optbuf[2] = IP6OPT_JUMBO;
1227 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1228 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1230 /* finally, adjust the packet header length */
1231 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1238 * Insert fragment header and copy unfragmentable header portions.
1241 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1242 struct ip6_frag **frghdrp)
1244 struct mbuf *n, *mlast;
1246 if (hlen > sizeof(struct ip6_hdr)) {
1247 n = m_copym(m0, sizeof(struct ip6_hdr),
1248 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1255 /* Search for the last mbuf of unfragmentable part. */
1256 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1259 if (!(mlast->m_flags & M_EXT) &&
1260 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1261 /* use the trailing space of the last mbuf for the fragment hdr */
1263 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1264 mlast->m_len += sizeof(struct ip6_frag);
1265 m->m_pkthdr.len += sizeof(struct ip6_frag);
1267 /* allocate a new mbuf for the fragment header */
1270 MGET(mfrg, M_NOWAIT, MT_DATA);
1273 mfrg->m_len = sizeof(struct ip6_frag);
1274 *frghdrp = mtod(mfrg, struct ip6_frag *);
1275 mlast->m_next = mfrg;
1282 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1283 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1290 if (ro_pmtu != ro) {
1291 /* The first hop and the final destination may differ. */
1292 struct sockaddr_in6 *sa6_dst =
1293 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1294 if (ro_pmtu->ro_rt &&
1295 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1296 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1297 RTFREE(ro_pmtu->ro_rt);
1298 ro_pmtu->ro_rt = NULL;
1300 if (ro_pmtu->ro_rt == NULL) {
1301 bzero(sa6_dst, sizeof(*sa6_dst));
1302 sa6_dst->sin6_family = AF_INET6;
1303 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1304 sa6_dst->sin6_addr = *dst;
1306 rtalloc((struct route *)ro_pmtu);
1309 if (ro_pmtu->ro_rt) {
1311 struct in_conninfo inc;
1313 bzero(&inc, sizeof(inc));
1314 inc.inc_flags = 1; /* IPv6 */
1315 inc.inc6_faddr = *dst;
1318 ifp = ro_pmtu->ro_rt->rt_ifp;
1319 ifmtu = IN6_LINKMTU(ifp);
1320 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1323 else if (mtu < IPV6_MMTU) {
1325 * RFC2460 section 5, last paragraph:
1326 * if we record ICMPv6 too big message with
1327 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1328 * or smaller, with framgent header attached.
1329 * (fragment header is needed regardless from the
1330 * packet size, for translators to identify packets)
1334 } else if (mtu > ifmtu) {
1336 * The MTU on the route is larger than the MTU on
1337 * the interface! This shouldn't happen, unless the
1338 * MTU of the interface has been changed after the
1339 * interface was brought up. Change the MTU in the
1340 * route to match the interface MTU (as long as the
1341 * field isn't locked).
1344 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1347 mtu = IN6_LINKMTU(ifp);
1349 error = EHOSTUNREACH; /* XXX */
1353 *alwaysfragp = alwaysfrag;
1358 * IP6 socket option processing.
1361 ip6_ctloutput_dispatch(netmsg_t msg)
1365 error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1366 msg->ctloutput.nm_sopt);
1367 lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1371 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1373 int optdatalen,uproto;
1375 struct inpcb *in6p = so->so_pcb;
1378 int level, op, optname;
1383 level = sopt->sopt_level;
1384 op = sopt->sopt_dir;
1385 optname = sopt->sopt_name;
1386 optlen = sopt->sopt_valsize;
1389 panic("ip6_ctloutput: arg soopt is NULL");
1395 uproto = (int)so->so_proto->pr_protocol;
1396 privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1398 if (level == IPPROTO_IPV6) {
1403 case IPV6_2292PKTOPTIONS:
1404 #ifdef IPV6_PKTOPTIONS
1405 case IPV6_PKTOPTIONS:
1410 error = soopt_getm(sopt, &m); /* XXX */
1413 soopt_to_mbuf(sopt, m); /* XXX */
1414 error = ip6_pcbopts(&in6p->in6p_outputopts,
1416 m_freem(m); /* XXX */
1421 * Use of some Hop-by-Hop options or some
1422 * Destination options, might require special
1423 * privilege. That is, normal applications
1424 * (without special privilege) might be forbidden
1425 * from setting certain options in outgoing packets,
1426 * and might never see certain options in received
1427 * packets. [RFC 2292 Section 6]
1428 * KAME specific note:
1429 * KAME prevents non-privileged users from sending or
1430 * receiving ANY hbh/dst options in order to avoid
1431 * overhead of parsing options in the kernel.
1433 case IPV6_RECVHOPOPTS:
1434 case IPV6_RECVDSTOPTS:
1435 case IPV6_RECVRTHDRDSTOPTS:
1438 case IPV6_RECVPKTINFO:
1439 case IPV6_RECVHOPLIMIT:
1440 case IPV6_RECVRTHDR:
1441 case IPV6_RECVPATHMTU:
1442 case IPV6_RECVTCLASS:
1443 case IPV6_AUTOFLOWLABEL:
1446 case IPV6_UNICAST_HOPS:
1450 if (optlen != sizeof(int)) {
1454 error = soopt_to_kbuf(sopt, &optval,
1455 sizeof optval, sizeof optval);
1460 case IPV6_UNICAST_HOPS:
1461 if (optval < -1 || optval >= 256)
1464 /* -1 = kernel default */
1465 in6p->in6p_hops = optval;
1468 #define OPTSET(bit) \
1471 in6p->in6p_flags |= (bit); \
1473 in6p->in6p_flags &= ~(bit); \
1475 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1477 * Although changed to RFC3542, It's better to also support RFC2292 API
1479 #define OPTSET2292(bit) \
1481 in6p->in6p_flags |= IN6P_RFC2292; \
1483 in6p->in6p_flags |= (bit); \
1485 in6p->in6p_flags &= ~(bit); \
1486 } while (/*CONSTCOND*/ 0)
1488 case IPV6_RECVPKTINFO:
1489 /* cannot mix with RFC2292 */
1490 if (OPTBIT(IN6P_RFC2292)) {
1494 OPTSET(IN6P_PKTINFO);
1499 struct ip6_pktopts **optp;
1501 /* cannot mix with RFC2292 */
1502 if (OPTBIT(IN6P_RFC2292)) {
1506 optp = &in6p->in6p_outputopts;
1507 error = ip6_pcbopt(IPV6_HOPLIMIT,
1508 (u_char *)&optval, sizeof(optval),
1513 case IPV6_RECVHOPLIMIT:
1514 /* cannot mix with RFC2292 */
1515 if (OPTBIT(IN6P_RFC2292)) {
1519 OPTSET(IN6P_HOPLIMIT);
1522 case IPV6_RECVHOPOPTS:
1523 /* cannot mix with RFC2292 */
1524 if (OPTBIT(IN6P_RFC2292)) {
1528 OPTSET(IN6P_HOPOPTS);
1531 case IPV6_RECVDSTOPTS:
1532 /* cannot mix with RFC2292 */
1533 if (OPTBIT(IN6P_RFC2292)) {
1537 OPTSET(IN6P_DSTOPTS);
1540 case IPV6_RECVRTHDRDSTOPTS:
1541 /* cannot mix with RFC2292 */
1542 if (OPTBIT(IN6P_RFC2292)) {
1546 OPTSET(IN6P_RTHDRDSTOPTS);
1549 case IPV6_RECVRTHDR:
1550 /* cannot mix with RFC2292 */
1551 if (OPTBIT(IN6P_RFC2292)) {
1558 case IPV6_RECVPATHMTU:
1560 * We ignore this option for TCP
1562 * (RFC3542 leaves this case
1565 if (uproto != IPPROTO_TCP)
1569 case IPV6_RECVTCLASS:
1570 /* cannot mix with RFC2292 XXX */
1571 if (OPTBIT(IN6P_RFC2292)) {
1575 OPTSET(IN6P_TCLASS);
1578 case IPV6_AUTOFLOWLABEL:
1579 OPTSET(IN6P_AUTOFLOWLABEL);
1588 * make setsockopt(IPV6_V6ONLY)
1589 * available only prior to bind(2).
1591 if (in6p->in6p_lport ||
1592 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1598 /* Don't allow v4-mapped */
1607 case IPV6_USE_MIN_MTU:
1608 case IPV6_PREFER_TEMPADDR:
1609 if (optlen != sizeof(optval)) {
1613 error = soopt_to_kbuf(sopt, &optval,
1614 sizeof optval, sizeof optval);
1618 struct ip6_pktopts **optp;
1619 optp = &in6p->in6p_outputopts;
1620 error = ip6_pcbopt(optname,
1621 (u_char *)&optval, sizeof(optval),
1626 case IPV6_2292PKTINFO:
1627 case IPV6_2292HOPLIMIT:
1628 case IPV6_2292HOPOPTS:
1629 case IPV6_2292DSTOPTS:
1630 case IPV6_2292RTHDR:
1632 if (optlen != sizeof(int)) {
1636 error = soopt_to_kbuf(sopt, &optval,
1637 sizeof optval, sizeof optval);
1641 case IPV6_2292PKTINFO:
1642 OPTSET2292(IN6P_PKTINFO);
1644 case IPV6_2292HOPLIMIT:
1645 OPTSET2292(IN6P_HOPLIMIT);
1647 case IPV6_2292HOPOPTS:
1649 * Check super-user privilege.
1650 * See comments for IPV6_RECVHOPOPTS.
1654 OPTSET2292(IN6P_HOPOPTS);
1656 case IPV6_2292DSTOPTS:
1659 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1661 case IPV6_2292RTHDR:
1662 OPTSET2292(IN6P_RTHDR);
1671 case IPV6_RTHDRDSTOPTS:
1675 * New advanced API (RFC3542)
1678 u_char optbuf_storage[MCLBYTES];
1680 struct ip6_pktopts **optp;
1682 /* cannot mix with RFC2292 */
1683 if (OPTBIT(IN6P_RFC2292)) {
1689 * We only ensure valsize is not too large
1690 * here. Further validation will be done
1693 error = soopt_to_kbuf(sopt, optbuf_storage,
1694 sizeof(optbuf_storage), 0);
1697 optlen = sopt->sopt_valsize;
1698 optbuf = optbuf_storage;
1699 optp = &in6p->in6p_outputopts;
1700 error = ip6_pcbopt(optname, optbuf, optlen,
1706 case IPV6_MULTICAST_IF:
1707 case IPV6_MULTICAST_HOPS:
1708 case IPV6_MULTICAST_LOOP:
1709 case IPV6_JOIN_GROUP:
1710 case IPV6_LEAVE_GROUP:
1713 if (sopt->sopt_valsize > MLEN) {
1718 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER);
1723 m->m_len = sopt->sopt_valsize;
1724 error = soopt_to_kbuf(sopt, mtod(m, char *),
1725 m->m_len, m->m_len);
1726 error = ip6_setmoptions(sopt->sopt_name,
1727 &in6p->in6p_moptions,
1733 case IPV6_PORTRANGE:
1734 error = soopt_to_kbuf(sopt, &optval,
1735 sizeof optval, sizeof optval);
1740 case IPV6_PORTRANGE_DEFAULT:
1741 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1742 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1745 case IPV6_PORTRANGE_HIGH:
1746 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1747 in6p->in6p_flags |= IN6P_HIGHPORT;
1750 case IPV6_PORTRANGE_LOW:
1751 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1752 in6p->in6p_flags |= IN6P_LOWPORT;
1761 #if defined(IPSEC) || defined(FAST_IPSEC)
1762 case IPV6_IPSEC_POLICY:
1768 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1770 soopt_to_mbuf(sopt, m); /* XXX */
1772 req = mtod(m, caddr_t);
1775 error = ipsec6_set_policy(in6p, optname, req,
1780 #endif /* KAME IPSEC */
1788 struct mbuf **mp = &m;
1790 if (ip6_fw_ctl_ptr == NULL)
1793 if ((error = soopt_getm(sopt, &m)) != 0)
1796 soopt_to_mbuf(sopt, m);
1797 error = (*ip6_fw_ctl_ptr)(optname, mp);
1803 error = ENOPROTOOPT;
1810 case IPV6_2292PKTOPTIONS:
1811 #ifdef IPV6_PKTOPTIONS
1812 case IPV6_PKTOPTIONS:
1815 * RFC3542 (effectively) deprecated the
1816 * semantics of the 2292-style pktoptions.
1817 * Since it was not reliable in nature (i.e.,
1818 * applications had to expect the lack of some
1819 * information after all), it would make sense
1820 * to simplify this part by always returning
1823 if (in6p->in6p_options) {
1825 m = m_copym(in6p->in6p_options,
1826 0, M_COPYALL, M_WAITOK);
1827 error = soopt_from_mbuf(sopt, m);
1831 sopt->sopt_valsize = 0;
1834 case IPV6_RECVHOPOPTS:
1835 case IPV6_RECVDSTOPTS:
1836 case IPV6_RECVRTHDRDSTOPTS:
1837 case IPV6_UNICAST_HOPS:
1838 case IPV6_RECVPKTINFO:
1839 case IPV6_RECVHOPLIMIT:
1840 case IPV6_RECVRTHDR:
1841 case IPV6_RECVPATHMTU:
1842 case IPV6_RECVTCLASS:
1843 case IPV6_AUTOFLOWLABEL:
1846 case IPV6_PORTRANGE:
1849 case IPV6_RECVHOPOPTS:
1850 optval = OPTBIT(IN6P_HOPOPTS);
1853 case IPV6_RECVDSTOPTS:
1854 optval = OPTBIT(IN6P_DSTOPTS);
1857 case IPV6_RECVRTHDRDSTOPTS:
1858 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1861 case IPV6_RECVPKTINFO:
1862 optval = OPTBIT(IN6P_PKTINFO);
1865 case IPV6_RECVHOPLIMIT:
1866 optval = OPTBIT(IN6P_HOPLIMIT);
1869 case IPV6_RECVRTHDR:
1870 optval = OPTBIT(IN6P_RTHDR);
1873 case IPV6_RECVPATHMTU:
1874 optval = OPTBIT(IN6P_MTU);
1877 case IPV6_RECVTCLASS:
1878 optval = OPTBIT(IN6P_TCLASS);
1881 case IPV6_AUTOFLOWLABEL:
1882 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1886 case IPV6_UNICAST_HOPS:
1887 optval = in6p->in6p_hops;
1891 optval = OPTBIT(IN6P_FAITH);
1898 case IPV6_PORTRANGE:
1901 flags = in6p->in6p_flags;
1902 if (flags & IN6P_HIGHPORT)
1903 optval = IPV6_PORTRANGE_HIGH;
1904 else if (flags & IN6P_LOWPORT)
1905 optval = IPV6_PORTRANGE_LOW;
1911 soopt_from_kbuf(sopt, &optval,
1918 struct ip6_mtuinfo mtuinfo;
1919 struct route_in6 sro;
1921 bzero(&sro, sizeof(sro));
1923 if (!(so->so_state & SS_ISCONNECTED))
1926 * XXX: we dot not consider the case of source
1927 * routing, or optional information to specify
1928 * the outgoing interface.
1930 error = ip6_getpmtu(&sro, NULL, NULL,
1931 &in6p->in6p_faddr, &pmtu, NULL);
1936 if (pmtu > IPV6_MAXPACKET)
1937 pmtu = IPV6_MAXPACKET;
1939 bzero(&mtuinfo, sizeof(mtuinfo));
1940 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1941 optdata = (void *)&mtuinfo;
1942 optdatalen = sizeof(mtuinfo);
1943 soopt_from_kbuf(sopt, optdata,
1948 case IPV6_2292PKTINFO:
1949 case IPV6_2292HOPLIMIT:
1950 case IPV6_2292HOPOPTS:
1951 case IPV6_2292RTHDR:
1952 case IPV6_2292DSTOPTS:
1953 if (optname == IPV6_2292HOPOPTS ||
1954 optname == IPV6_2292DSTOPTS ||
1958 case IPV6_2292PKTINFO:
1959 optval = OPTBIT(IN6P_PKTINFO);
1961 case IPV6_2292HOPLIMIT:
1962 optval = OPTBIT(IN6P_HOPLIMIT);
1964 case IPV6_2292HOPOPTS:
1967 optval = OPTBIT(IN6P_HOPOPTS);
1969 case IPV6_2292RTHDR:
1970 optval = OPTBIT(IN6P_RTHDR);
1972 case IPV6_2292DSTOPTS:
1975 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1978 soopt_from_kbuf(sopt, &optval,
1986 case IPV6_RTHDRDSTOPTS:
1990 case IPV6_USE_MIN_MTU:
1991 case IPV6_PREFER_TEMPADDR:
1992 error = ip6_getpcbopt(in6p->in6p_outputopts,
1996 case IPV6_MULTICAST_IF:
1997 case IPV6_MULTICAST_HOPS:
1998 case IPV6_MULTICAST_LOOP:
1999 case IPV6_JOIN_GROUP:
2000 case IPV6_LEAVE_GROUP:
2003 error = ip6_getmoptions(sopt->sopt_name,
2004 in6p->in6p_moptions, &m);
2006 soopt_from_kbuf(sopt,
2007 mtod(m, char *), m->m_len);
2012 #if defined(IPSEC) || defined(FAST_IPSEC)
2013 case IPV6_IPSEC_POLICY:
2017 struct mbuf *m = NULL;
2018 struct mbuf **mp = &m;
2020 error = soopt_getm(sopt, &m); /* XXX */
2023 soopt_to_mbuf(sopt, m); /* XXX */
2025 req = mtod(m, caddr_t);
2028 error = ipsec6_get_policy(in6p, req, len, mp);
2030 error = soopt_from_mbuf(sopt, m); /*XXX*/
2031 if (error == 0 && m != NULL)
2035 #endif /* KAME IPSEC */
2040 struct mbuf **mp = &m;
2042 if (ip6_fw_ctl_ptr == NULL)
2046 error = (*ip6_fw_ctl_ptr)(optname, mp);
2048 error = soopt_from_mbuf(sopt, m); /* XXX */
2049 if (error == 0 && m != NULL)
2055 error = ENOPROTOOPT;
2067 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2069 int error = 0, optval, optlen;
2070 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2071 struct in6pcb *in6p = sotoin6pcb(so);
2072 int level, op, optname;
2075 level = sopt->sopt_level;
2076 op = sopt->sopt_dir;
2077 optname = sopt->sopt_name;
2078 optlen = sopt->sopt_valsize;
2080 panic("ip6_raw_ctloutput: arg soopt is NULL");
2082 if (level != IPPROTO_IPV6) {
2089 * For ICMPv6 sockets, no modification allowed for checksum
2090 * offset, permit "no change" values to help existing apps.
2092 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2093 * for an ICMPv6 socket will fail."
2094 * The current behavior does not meet RFC3542.
2098 if (optlen != sizeof(int)) {
2102 error = soopt_to_kbuf(sopt, &optval,
2103 sizeof optval, sizeof optval);
2106 if ((optval % 2) != 0) {
2107 /* the API assumes even offset values */
2109 } else if (so->so_proto->pr_protocol ==
2111 if (optval != icmp6off)
2114 in6p->in6p_cksum = optval;
2118 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2121 optval = in6p->in6p_cksum;
2123 soopt_from_kbuf(sopt, &optval, sizeof(optval));
2133 error = ENOPROTOOPT;
2141 * Set up IP6 options in pcb for insertion in output packets or
2142 * specifying behavior of outgoing packets.
2145 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2146 struct socket *so, struct sockopt *sopt)
2149 struct ip6_pktopts *opt = *pktopt;
2152 /* turn off any old options. */
2155 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2156 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2157 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2158 kprintf("ip6_pcbopts: all specified options are cleared.\n");
2160 ip6_clearpktopts(opt, -1);
2162 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2165 if (!m || m->m_len == 0) {
2167 * Only turning off any previous options, regardless of
2168 * whether the opt is just created or given.
2170 kfree(opt, M_IP6OPT);
2174 /* set options specified by user. */
2175 if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2176 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2177 kfree(opt, M_IP6OPT);
2186 * Below three functions are introduced by merge to RFC3542
2190 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2192 void *optdata = NULL;
2194 struct ip6_ext *ip6e;
2196 struct in6_pktinfo null_pktinfo;
2197 int deftclass = 0, on;
2198 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2199 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2203 if (pktopt && pktopt->ip6po_pktinfo)
2204 optdata = (void *)pktopt->ip6po_pktinfo;
2206 /* XXX: we don't have to do this every time... */
2207 bzero(&null_pktinfo, sizeof(null_pktinfo));
2208 optdata = (void *)&null_pktinfo;
2210 optdatalen = sizeof(struct in6_pktinfo);
2213 if (pktopt && pktopt->ip6po_tclass >= 0)
2214 optdata = (void *)&pktopt->ip6po_tclass;
2216 optdata = (void *)&deftclass;
2217 optdatalen = sizeof(int);
2220 if (pktopt && pktopt->ip6po_hbh) {
2221 optdata = (void *)pktopt->ip6po_hbh;
2222 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2223 optdatalen = (ip6e->ip6e_len + 1) << 3;
2227 if (pktopt && pktopt->ip6po_rthdr) {
2228 optdata = (void *)pktopt->ip6po_rthdr;
2229 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2230 optdatalen = (ip6e->ip6e_len + 1) << 3;
2233 case IPV6_RTHDRDSTOPTS:
2234 if (pktopt && pktopt->ip6po_dest1) {
2235 optdata = (void *)pktopt->ip6po_dest1;
2236 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2237 optdatalen = (ip6e->ip6e_len + 1) << 3;
2241 if (pktopt && pktopt->ip6po_dest2) {
2242 optdata = (void *)pktopt->ip6po_dest2;
2243 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2244 optdatalen = (ip6e->ip6e_len + 1) << 3;
2248 if (pktopt && pktopt->ip6po_nexthop) {
2249 optdata = (void *)pktopt->ip6po_nexthop;
2250 optdatalen = pktopt->ip6po_nexthop->sa_len;
2253 case IPV6_USE_MIN_MTU:
2255 optdata = (void *)&pktopt->ip6po_minmtu;
2257 optdata = (void *)&defminmtu;
2258 optdatalen = sizeof(int);
2261 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2265 optdata = (void *)&on;
2266 optdatalen = sizeof(on);
2268 case IPV6_PREFER_TEMPADDR:
2270 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2272 optdata = (void *)&defpreftemp;
2273 optdatalen = sizeof(int);
2275 default: /* should not happen */
2277 panic("ip6_getpcbopt: unexpected option");
2279 return (ENOPROTOOPT);
2282 soopt_from_kbuf(sopt, optdata, optdatalen);
2288 * initialize ip6_pktopts. beware that there are non-zero default values in
2293 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2295 struct ip6_pktopts *opt;
2297 if (*pktopt == NULL) {
2298 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2299 init_ip6pktopts(*pktopt);
2303 return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2307 * initialize ip6_pktopts. beware that there are non-zero default values in
2311 init_ip6pktopts(struct ip6_pktopts *opt)
2314 bzero(opt, sizeof(*opt));
2315 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2316 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2317 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2318 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2322 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2327 if (optname == -1 || optname == IPV6_PKTINFO) {
2328 if (pktopt->ip6po_pktinfo)
2329 kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2330 pktopt->ip6po_pktinfo = NULL;
2332 if (optname == -1 || optname == IPV6_HOPLIMIT)
2333 pktopt->ip6po_hlim = -1;
2334 if (optname == -1 || optname == IPV6_TCLASS)
2335 pktopt->ip6po_tclass = -1;
2336 if (optname == -1 || optname == IPV6_NEXTHOP) {
2337 if (pktopt->ip6po_nextroute.ro_rt) {
2338 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2339 pktopt->ip6po_nextroute.ro_rt = NULL;
2341 if (pktopt->ip6po_nexthop)
2342 kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2343 pktopt->ip6po_nexthop = NULL;
2345 if (optname == -1 || optname == IPV6_HOPOPTS) {
2346 if (pktopt->ip6po_hbh)
2347 kfree(pktopt->ip6po_hbh, M_IP6OPT);
2348 pktopt->ip6po_hbh = NULL;
2350 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2351 if (pktopt->ip6po_dest1)
2352 kfree(pktopt->ip6po_dest1, M_IP6OPT);
2353 pktopt->ip6po_dest1 = NULL;
2355 if (optname == -1 || optname == IPV6_RTHDR) {
2356 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2357 kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2358 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2359 if (pktopt->ip6po_route.ro_rt) {
2360 RTFREE(pktopt->ip6po_route.ro_rt);
2361 pktopt->ip6po_route.ro_rt = NULL;
2364 if (optname == -1 || optname == IPV6_DSTOPTS) {
2365 if (pktopt->ip6po_dest2)
2366 kfree(pktopt->ip6po_dest2, M_IP6OPT);
2367 pktopt->ip6po_dest2 = NULL;
2371 #define PKTOPT_EXTHDRCPY(type) \
2375 (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2376 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2377 if (dst->type == NULL)\
2379 bcopy(src->type, dst->type, hlen);\
2383 struct ip6_pktopts *
2384 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2386 struct ip6_pktopts *dst;
2389 kprintf("ip6_clearpktopts: invalid argument\n");
2393 dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2397 dst->ip6po_hlim = src->ip6po_hlim;
2398 if (src->ip6po_pktinfo) {
2399 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2401 if (dst->ip6po_pktinfo == NULL)
2403 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2405 if (src->ip6po_nexthop) {
2406 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2408 if (dst->ip6po_nexthop == NULL)
2410 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2411 src->ip6po_nexthop->sa_len);
2413 PKTOPT_EXTHDRCPY(ip6po_hbh);
2414 PKTOPT_EXTHDRCPY(ip6po_dest1);
2415 PKTOPT_EXTHDRCPY(ip6po_dest2);
2416 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2420 if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2421 if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2422 if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2423 if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2424 if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2425 if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2426 kfree(dst, M_IP6OPT);
2431 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2433 if (dst == NULL || src == NULL) {
2435 kprintf("ip6_clearpktopts: invalid argument\n");
2440 dst->ip6po_hlim = src->ip6po_hlim;
2441 dst->ip6po_tclass = src->ip6po_tclass;
2442 dst->ip6po_flags = src->ip6po_flags;
2443 if (src->ip6po_pktinfo) {
2444 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2446 if (dst->ip6po_pktinfo == NULL)
2448 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2450 if (src->ip6po_nexthop) {
2451 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2453 if (dst->ip6po_nexthop == NULL)
2455 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2456 src->ip6po_nexthop->sa_len);
2458 PKTOPT_EXTHDRCPY(ip6po_hbh);
2459 PKTOPT_EXTHDRCPY(ip6po_dest1);
2460 PKTOPT_EXTHDRCPY(ip6po_dest2);
2461 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2465 ip6_clearpktopts(dst, -1);
2468 #undef PKTOPT_EXTHDRCPY
2471 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2476 ip6_clearpktopts(pktopt, -1);
2478 kfree(pktopt, M_IP6OPT);
2482 * Set the IP6 multicast options in response to user setsockopt().
2485 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2488 u_int loop, ifindex;
2489 struct ipv6_mreq *mreq;
2491 struct ip6_moptions *im6o = *im6op;
2492 struct route_in6 ro;
2493 struct sockaddr_in6 *dst;
2494 struct in6_multi_mship *imm;
2495 struct thread *td = curthread; /* XXX */
2499 * No multicast option buffer attached to the pcb;
2500 * allocate one and initialize to default values.
2502 im6o = (struct ip6_moptions *)
2503 kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2506 im6o->im6o_multicast_ifp = NULL;
2507 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2508 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2509 LIST_INIT(&im6o->im6o_memberships);
2514 case IPV6_MULTICAST_IF:
2516 * Select the interface for outgoing multicast packets.
2518 if (m == NULL || m->m_len != sizeof(u_int)) {
2522 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2523 if (ifindex < 0 || if_index < ifindex) {
2524 error = ENXIO; /* XXX EINVAL? */
2527 ifp = ifindex2ifnet[ifindex];
2528 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2529 error = EADDRNOTAVAIL;
2532 im6o->im6o_multicast_ifp = ifp;
2535 case IPV6_MULTICAST_HOPS:
2538 * Set the IP6 hoplimit for outgoing multicast packets.
2541 if (m == NULL || m->m_len != sizeof(int)) {
2545 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2546 if (optval < -1 || optval >= 256)
2548 else if (optval == -1)
2549 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2551 im6o->im6o_multicast_hlim = optval;
2555 case IPV6_MULTICAST_LOOP:
2557 * Set the loopback flag for outgoing multicast packets.
2558 * Must be zero or one.
2560 if (m == NULL || m->m_len != sizeof(u_int)) {
2564 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2569 im6o->im6o_multicast_loop = loop;
2572 case IPV6_JOIN_GROUP:
2574 * Add a multicast group membership.
2575 * Group must be a valid IP6 multicast address.
2577 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2581 mreq = mtod(m, struct ipv6_mreq *);
2582 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2584 * We use the unspecified address to specify to accept
2585 * all multicast addresses. Only super user is allowed
2588 if (priv_check(td, PRIV_ROOT))
2593 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2599 * If the interface is specified, validate it.
2601 if (mreq->ipv6mr_interface < 0
2602 || if_index < mreq->ipv6mr_interface) {
2603 error = ENXIO; /* XXX EINVAL? */
2607 * If no interface was explicitly specified, choose an
2608 * appropriate one according to the given multicast address.
2610 if (mreq->ipv6mr_interface == 0) {
2612 * If the multicast address is in node-local scope,
2613 * the interface should be a loopback interface.
2614 * Otherwise, look up the routing table for the
2615 * address, and choose the outgoing interface.
2616 * XXX: is it a good approach?
2618 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2622 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2623 bzero(dst, sizeof(*dst));
2624 dst->sin6_len = sizeof(struct sockaddr_in6);
2625 dst->sin6_family = AF_INET6;
2626 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2627 rtalloc((struct route *)&ro);
2628 if (ro.ro_rt == NULL) {
2629 error = EADDRNOTAVAIL;
2632 ifp = ro.ro_rt->rt_ifp;
2636 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2639 * See if we found an interface, and confirm that it
2640 * supports multicast
2642 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2643 error = EADDRNOTAVAIL;
2647 * Put interface index into the multicast address,
2648 * if the address has link-local scope.
2650 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2651 mreq->ipv6mr_multiaddr.s6_addr16[1]
2652 = htons(mreq->ipv6mr_interface);
2655 * See if the membership already exists.
2657 for (imm = im6o->im6o_memberships.lh_first;
2658 imm != NULL; imm = imm->i6mm_chain.le_next)
2659 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2660 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2661 &mreq->ipv6mr_multiaddr))
2668 * Everything looks good; add a new record to the multicast
2669 * address list for the given interface.
2671 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2672 if ((imm->i6mm_maddr =
2673 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2674 kfree(imm, M_IPMADDR);
2677 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2680 case IPV6_LEAVE_GROUP:
2682 * Drop a multicast group membership.
2683 * Group must be a valid IP6 multicast address.
2685 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2689 mreq = mtod(m, struct ipv6_mreq *);
2690 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2691 if (priv_check(td, PRIV_ROOT)) {
2695 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2700 * If an interface address was specified, get a pointer
2701 * to its ifnet structure.
2703 if (mreq->ipv6mr_interface < 0
2704 || if_index < mreq->ipv6mr_interface) {
2705 error = ENXIO; /* XXX EINVAL? */
2708 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2710 * Put interface index into the multicast address,
2711 * if the address has link-local scope.
2713 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2714 mreq->ipv6mr_multiaddr.s6_addr16[1]
2715 = htons(mreq->ipv6mr_interface);
2718 * Find the membership in the membership list.
2720 for (imm = im6o->im6o_memberships.lh_first;
2721 imm != NULL; imm = imm->i6mm_chain.le_next) {
2723 imm->i6mm_maddr->in6m_ifp == ifp) &&
2724 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2725 &mreq->ipv6mr_multiaddr))
2729 /* Unable to resolve interface */
2730 error = EADDRNOTAVAIL;
2734 * Give up the multicast address record to which the
2735 * membership points.
2737 LIST_REMOVE(imm, i6mm_chain);
2738 in6_delmulti(imm->i6mm_maddr);
2739 kfree(imm, M_IPMADDR);
2748 * If all options have default values, no need to keep the mbuf.
2750 if (im6o->im6o_multicast_ifp == NULL &&
2751 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2752 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2753 im6o->im6o_memberships.lh_first == NULL) {
2754 kfree(*im6op, M_IPMOPTS);
2762 * Return the IP6 multicast options in response to user getsockopt().
2765 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2767 u_int *hlim, *loop, *ifindex;
2769 *mp = m_get(M_WAITOK, MT_HEADER); /* XXX */
2773 case IPV6_MULTICAST_IF:
2774 ifindex = mtod(*mp, u_int *);
2775 (*mp)->m_len = sizeof(u_int);
2776 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2779 *ifindex = im6o->im6o_multicast_ifp->if_index;
2782 case IPV6_MULTICAST_HOPS:
2783 hlim = mtod(*mp, u_int *);
2784 (*mp)->m_len = sizeof(u_int);
2786 *hlim = ip6_defmcasthlim;
2788 *hlim = im6o->im6o_multicast_hlim;
2791 case IPV6_MULTICAST_LOOP:
2792 loop = mtod(*mp, u_int *);
2793 (*mp)->m_len = sizeof(u_int);
2795 *loop = ip6_defmcasthlim;
2797 *loop = im6o->im6o_multicast_loop;
2801 return (EOPNOTSUPP);
2806 * Discard the IP6 multicast options.
2809 ip6_freemoptions(struct ip6_moptions *im6o)
2811 struct in6_multi_mship *imm;
2816 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2817 LIST_REMOVE(imm, i6mm_chain);
2818 if (imm->i6mm_maddr)
2819 in6_delmulti(imm->i6mm_maddr);
2820 kfree(imm, M_IPMADDR);
2822 kfree(im6o, M_IPMOPTS);
2826 * Set a particular packet option, as a sticky option or an ancillary data
2827 * item. "len" can be 0 only when it's a sticky option.
2828 * We have 4 cases of combination of "sticky" and "cmsg":
2829 * "sticky=0, cmsg=0": impossible
2830 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2831 * "sticky=1, cmsg=0": RFC3542 socket option
2832 * "sticky=1, cmsg=1": RFC2292 socket option
2835 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2836 int sticky, int cmsg, int uproto, int priv)
2838 int minmtupolicy, preftemp;
2841 if (!sticky && !cmsg) {
2842 kprintf("ip6_setpktoption: impossible case\n");
2847 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2848 * not be specified in the context of RFC3542. Conversely,
2849 * RFC3542 types should not be specified in the context of RFC2292.
2853 case IPV6_2292PKTINFO:
2854 case IPV6_2292HOPLIMIT:
2855 case IPV6_2292NEXTHOP:
2856 case IPV6_2292HOPOPTS:
2857 case IPV6_2292DSTOPTS:
2858 case IPV6_2292RTHDR:
2859 case IPV6_2292PKTOPTIONS:
2860 return (ENOPROTOOPT);
2863 if (sticky && cmsg) {
2870 case IPV6_RTHDRDSTOPTS:
2872 case IPV6_USE_MIN_MTU:
2875 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2876 return (ENOPROTOOPT);
2881 case IPV6_2292PKTINFO:
2884 struct in6_pktinfo *pktinfo;
2885 if (len != sizeof(struct in6_pktinfo))
2887 pktinfo = (struct in6_pktinfo *)buf;
2890 * An application can clear any sticky IPV6_PKTINFO option by
2891 * doing a "regular" setsockopt with ipi6_addr being
2892 * in6addr_any and ipi6_ifindex being zero.
2893 * [RFC 3542, Section 6]
2895 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2896 pktinfo->ipi6_ifindex == 0 &&
2897 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2898 ip6_clearpktopts(opt, optname);
2902 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2903 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2907 /* validate the interface index if specified. */
2908 if (pktinfo->ipi6_ifindex > if_index ||
2909 pktinfo->ipi6_ifindex < 0) {
2913 * Check if the requested source address is indeed a
2914 * unicast address assigned to the node, and can be
2915 * used as the packet's source address.
2917 if (opt->ip6po_pktinfo != NULL &&
2918 !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2919 struct in6_ifaddr *ia6;
2920 struct sockaddr_in6 sin6;
2922 bzero(&sin6, sizeof(sin6));
2923 sin6.sin6_len = sizeof(sin6);
2924 sin6.sin6_family = AF_INET6;
2926 opt->ip6po_pktinfo->ipi6_addr;
2927 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2929 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2930 IN6_IFF_NOTREADY)) != 0)
2931 return (EADDRNOTAVAIL);
2935 * We store the address anyway, and let in6_selectsrc()
2936 * validate the specified address. This is because ipi6_addr
2937 * may not have enough information about its scope zone, and
2938 * we may need additional information (such as outgoing
2939 * interface or the scope zone of a destination address) to
2940 * disambiguate the scope.
2941 * XXX: the delay of the validation may confuse the
2942 * application when it is used as a sticky option.
2944 if (opt->ip6po_pktinfo == NULL) {
2945 opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2946 M_IP6OPT, M_NOWAIT);
2947 if (opt->ip6po_pktinfo == NULL)
2950 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2954 case IPV6_2292HOPLIMIT:
2960 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2961 * to simplify the ordering among hoplimit options.
2963 if (optname == IPV6_HOPLIMIT && sticky)
2964 return (ENOPROTOOPT);
2966 if (len != sizeof(int))
2969 if (*hlimp < -1 || *hlimp > 255)
2972 opt->ip6po_hlim = *hlimp;
2980 if (len != sizeof(int))
2982 tclass = *(int *)buf;
2983 if (tclass < -1 || tclass > 255)
2986 opt->ip6po_tclass = tclass;
2990 case IPV6_2292NEXTHOP:
2995 if (len == 0) { /* just remove the option */
2996 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3000 /* check if cmsg_len is large enough for sa_len */
3001 if (len < sizeof(struct sockaddr) || len < *buf)
3004 switch (((struct sockaddr *)buf)->sa_family) {
3007 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3010 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3013 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3014 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3019 case AF_LINK: /* should eventually be supported */
3021 return (EAFNOSUPPORT);
3024 /* turn off the previous option, then set the new option. */
3025 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3026 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3027 if (opt->ip6po_nexthop == NULL)
3029 bcopy(buf, opt->ip6po_nexthop, *buf);
3032 case IPV6_2292HOPOPTS:
3035 struct ip6_hbh *hbh;
3039 * XXX: We don't allow a non-privileged user to set ANY HbH
3040 * options, since per-option restriction has too much
3046 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3047 break; /* just remove the option */
3050 /* message length validation */
3051 if (len < sizeof(struct ip6_hbh))
3053 hbh = (struct ip6_hbh *)buf;
3054 hbhlen = (hbh->ip6h_len + 1) << 3;
3058 /* turn off the previous option, then set the new option. */
3059 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3060 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3061 if (opt->ip6po_hbh == NULL)
3063 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3068 case IPV6_2292DSTOPTS:
3070 case IPV6_RTHDRDSTOPTS:
3072 struct ip6_dest *dest, **newdest = NULL;
3078 ip6_clearpktopts(opt, optname);
3079 break; /* just remove the option */
3082 /* message length validation */
3083 if (len < sizeof(struct ip6_dest))
3085 dest = (struct ip6_dest *)buf;
3086 destlen = (dest->ip6d_len + 1) << 3;
3091 * Determine the position that the destination options header
3092 * should be inserted; before or after the routing header.
3095 case IPV6_2292DSTOPTS:
3097 * The old advacned API is ambiguous on this point.
3098 * Our approach is to determine the position based
3099 * according to the existence of a routing header.
3100 * Note, however, that this depends on the order of the
3101 * extension headers in the ancillary data; the 1st
3102 * part of the destination options header must appear
3103 * before the routing header in the ancillary data,
3105 * RFC3542 solved the ambiguity by introducing
3106 * separate ancillary data or option types.
3108 if (opt->ip6po_rthdr == NULL)
3109 newdest = &opt->ip6po_dest1;
3111 newdest = &opt->ip6po_dest2;
3113 case IPV6_RTHDRDSTOPTS:
3114 newdest = &opt->ip6po_dest1;
3117 newdest = &opt->ip6po_dest2;
3121 /* turn off the previous option, then set the new option. */
3122 ip6_clearpktopts(opt, optname);
3123 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3124 if (*newdest == NULL)
3126 bcopy(dest, *newdest, destlen);
3131 case IPV6_2292RTHDR:
3134 struct ip6_rthdr *rth;
3138 ip6_clearpktopts(opt, IPV6_RTHDR);
3139 break; /* just remove the option */
3142 /* message length validation */
3143 if (len < sizeof(struct ip6_rthdr))
3145 rth = (struct ip6_rthdr *)buf;
3146 rthlen = (rth->ip6r_len + 1) << 3;
3150 switch (rth->ip6r_type) {
3152 return (EINVAL); /* not supported */
3155 /* turn off the previous option */
3156 ip6_clearpktopts(opt, IPV6_RTHDR);
3157 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3158 if (opt->ip6po_rthdr == NULL)
3160 bcopy(rth, opt->ip6po_rthdr, rthlen);
3165 case IPV6_USE_MIN_MTU:
3166 if (len != sizeof(int))
3168 minmtupolicy = *(int *)buf;
3169 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3170 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3171 minmtupolicy != IP6PO_MINMTU_ALL) {
3174 opt->ip6po_minmtu = minmtupolicy;
3178 if (len != sizeof(int))
3181 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3183 * we ignore this option for TCP sockets.
3184 * (RFC3542 leaves this case unspecified.)
3186 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3188 opt->ip6po_flags |= IP6PO_DONTFRAG;
3191 case IPV6_PREFER_TEMPADDR:
3192 if (len != sizeof(int))
3194 preftemp = *(int *)buf;
3195 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3196 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3197 preftemp != IP6PO_TEMPADDR_PREFER) {
3200 opt->ip6po_prefer_tempaddr = preftemp;
3204 return (ENOPROTOOPT);
3205 } /* end of switch */
3212 * Set IPv6 outgoing packet options based on advanced API.
3215 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3216 struct ip6_pktopts *stickyopt, int uproto, int priv)
3218 struct cmsghdr *cm = NULL;
3220 if (control == NULL || opt == NULL)
3223 init_ip6pktopts(opt);
3226 * XXX: Currently, we assume all the optional information is stored
3233 * If stickyopt is provided, make a local copy of the options
3234 * for this particular packet, then override them by ancillary
3236 * XXX: copypktopts() does not copy the cached route to a next
3237 * hop (if any). This is not very good in terms of efficiency,
3238 * but we can allow this since this option should be rarely
3241 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3246 * XXX: Currently, we assume all the optional information is stored
3249 if (control->m_next)
3252 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3253 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3256 if (control->m_len < CMSG_LEN(0))
3259 cm = mtod(control, struct cmsghdr *);
3260 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3262 if (cm->cmsg_level != IPPROTO_IPV6)
3265 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3266 cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3275 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3276 * packet to the input queue of a specified interface. Note that this
3277 * calls the output routine of the loopback "driver", but with an interface
3278 * pointer that might NOT be &loif -- easier than replicating that code here.
3281 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3284 struct ip6_hdr *ip6;
3286 copym = m_copy(m, 0, M_COPYALL);
3291 * Make sure to deep-copy IPv6 header portion in case the data
3292 * is in an mbuf cluster, so that we can safely override the IPv6
3293 * header portion later.
3295 if ((copym->m_flags & M_EXT) != 0 ||
3296 copym->m_len < sizeof(struct ip6_hdr)) {
3297 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3303 if (copym->m_len < sizeof(*ip6)) {
3309 ip6 = mtod(copym, struct ip6_hdr *);
3311 * clear embedded scope identifiers if necessary.
3312 * in6_clearscope will touch the addresses only when necessary.
3314 in6_clearscope(&ip6->ip6_src);
3315 in6_clearscope(&ip6->ip6_dst);
3317 if_simloop(ifp, copym, dst->sin6_family, 0);
3321 * Separate the IPv6 header from the payload into its own mbuf.
3323 * Returns the new mbuf chain or the original mbuf if no payload.
3324 * Returns NULL if can't allocate new mbuf for header.
3326 static struct mbuf *
3327 ip6_splithdr(struct mbuf *m)
3331 if (m->m_len <= sizeof(struct ip6_hdr)) /* no payload */
3334 MGETHDR(mh, M_NOWAIT, MT_HEADER);
3337 mh->m_len = sizeof(struct ip6_hdr);
3338 M_MOVE_PKTHDR(mh, m);
3339 MH_ALIGN(mh, sizeof(struct ip6_hdr));
3340 bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3341 m->m_data += sizeof(struct ip6_hdr);
3342 m->m_len -= sizeof(struct ip6_hdr);
3348 * Compute IPv6 extension header length.
3351 ip6_optlen(struct in6pcb *in6p)
3355 if (!in6p->in6p_outputopts)
3360 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3362 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3363 if (in6p->in6p_outputopts->ip6po_rthdr)
3364 /* dest1 is valid with rthdr only */
3365 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3366 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3367 len += elen(in6p->in6p_outputopts->ip6po_dest2);