2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
33 * Copyright (c) 1982, 1986, 1988, 1990, 1993
34 * The Regents of the University of California. All rights reserved.
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
67 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
71 #include "opt_route.h"
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
77 #include <sys/errno.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
86 #include <machine/in_cksum.h>
89 #include <net/netisr.h>
90 #include <net/route.h>
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip_var.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/tcp_var.h>
103 #include <netinet6/nd6.h>
106 #include <netipsec/ipsec.h>
107 #include <netipsec/ipsec6.h>
108 #include <netipsec/key.h>
109 #include <netinet6/ip6_ipsec.h>
112 #include <netinet/sctp.h>
113 #include <netinet/sctp_crc32.h>
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/scope6_var.h>
120 #include <net/flowtable.h>
123 extern int in6_mcast_loop;
126 struct mbuf *ip6e_ip6;
127 struct mbuf *ip6e_hbh;
128 struct mbuf *ip6e_dest1;
129 struct mbuf *ip6e_rthdr;
130 struct mbuf *ip6e_dest2;
133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
134 struct ucred *, int);
135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
136 struct socket *, struct sockopt *);
137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
139 struct ucred *, int, int, int);
141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
147 struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
152 * Make an extension header from option data. hp is the source, and
153 * mp is the destination.
155 #define MAKE_EXTHDR(hp, mp) \
158 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
159 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
160 ((eh)->ip6e_len + 1) << 3); \
164 } while (/*CONSTCOND*/ 0)
167 * Form a chain of extension headers.
168 * m is the extension header mbuf
169 * mp is the previous mbuf in the chain
170 * p is the next header
171 * i is the type of option.
173 #define MAKE_CHAIN(m, mp, p, i)\
177 panic("assumption failed: hdr not split"); \
178 *mtod((m), u_char *) = *(p);\
180 p = mtod((m), u_char *);\
181 (m)->m_next = (mp)->m_next;\
185 } while (/*CONSTCOND*/ 0)
188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
192 csum = in_cksum_skip(m, offset + plen, offset);
193 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
195 offset += m->m_pkthdr.csum_data; /* checksum offset */
197 if (offset + sizeof(u_short) > m->m_len) {
198 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
199 "csum_flags=0x%04x\n", __func__, m->m_len, plen, offset,
200 m->m_pkthdr.csum_flags);
202 * XXX this should not happen, but if it does, the correct
203 * behavior may be to insert the checksum in the appropriate
204 * next mbuf in the chain.
208 *(u_short *)(m->m_data + offset) = csum;
212 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
213 * header (with pri, len, nxt, hlim, src, dst).
214 * This function may modify ver and hlim only.
215 * The mbuf chain containing the packet will be freed.
216 * The mbuf opt, if present, will not be freed.
217 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
218 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
219 * then result of route lookup is stored in ro->ro_rt.
221 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
222 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
223 * which is rt_rmx.rmx_mtu.
225 * ifpp - XXX: just for statistics
228 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
229 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
230 struct ifnet **ifpp, struct inpcb *inp)
232 struct ip6_hdr *ip6, *mhip6;
233 struct ifnet *ifp, *origifp;
235 struct mbuf *mprev = NULL;
236 int hlen, tlen, len, off;
237 struct route_in6 ip6route;
238 struct rtentry *rt = NULL;
239 struct sockaddr_in6 *dst, src_sa, dst_sa;
240 struct in6_addr odst;
242 struct in6_ifaddr *ia = NULL;
244 int alwaysfrag, dontfrag;
245 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
246 struct ip6_exthdrs exthdrs;
247 struct in6_addr finaldst, src0, dst0;
249 struct route_in6 *ro_pmtu = NULL;
254 struct ipsec_output_state state;
255 struct ip6_rthdr *rh = NULL;
256 int needipsectun = 0;
258 struct secpolicy *sp = NULL;
260 struct m_tag *fwd_tag = NULL;
262 ip6 = mtod(m, struct ip6_hdr *);
264 printf ("ip6 is NULL");
269 M_SETFIB(m, inp->inp_inc.inc_fibnum);
271 finaldst = ip6->ip6_dst;
272 bzero(&exthdrs, sizeof(exthdrs));
274 /* Hop-by-Hop options header */
275 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
276 /* Destination options header(1st part) */
277 if (opt->ip6po_rthdr) {
279 * Destination options header(1st part)
280 * This only makes sense with a routing header.
281 * See Section 9.2 of RFC 3542.
282 * Disabling this part just for MIP6 convenience is
283 * a bad idea. We need to think carefully about a
284 * way to make the advanced API coexist with MIP6
285 * options, which might automatically be inserted in
288 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
291 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
292 /* Destination options header(2nd part) */
293 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
298 * IPSec checking which handles several cases.
299 * FAST IPSEC: We re-injected the packet.
301 switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
303 case 1: /* Bad packet */
305 case -1: /* Do IPSec */
308 * Do delayed checksums now, as we may send before returning.
310 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
311 plen = m->m_pkthdr.len - sizeof(*ip6);
312 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
313 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
316 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
317 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
318 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
321 case 0: /* No IPSec */
328 * Calculate the total length of the extension header chain.
329 * Keep the length of the unfragmentable part for fragmentation.
332 if (exthdrs.ip6e_hbh)
333 optlen += exthdrs.ip6e_hbh->m_len;
334 if (exthdrs.ip6e_dest1)
335 optlen += exthdrs.ip6e_dest1->m_len;
336 if (exthdrs.ip6e_rthdr)
337 optlen += exthdrs.ip6e_rthdr->m_len;
338 unfragpartlen = optlen + sizeof(struct ip6_hdr);
340 /* NOTE: we don't add AH/ESP length here. do that later. */
341 if (exthdrs.ip6e_dest2)
342 optlen += exthdrs.ip6e_dest2->m_len;
345 * If we need IPsec, or there is at least one extension header,
346 * separate IP6 header from the payload.
348 if ((needipsec || optlen) && !hdrsplit) {
349 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
353 m = exthdrs.ip6e_ip6;
358 ip6 = mtod(m, struct ip6_hdr *);
360 /* adjust mbuf packet header length */
361 m->m_pkthdr.len += optlen;
362 plen = m->m_pkthdr.len - sizeof(*ip6);
364 /* If this is a jumbo payload, insert a jumbo payload option. */
365 if (plen > IPV6_MAXPACKET) {
367 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
371 m = exthdrs.ip6e_ip6;
375 ip6 = mtod(m, struct ip6_hdr *);
376 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
380 ip6->ip6_plen = htons(plen);
383 * Concatenate headers and fill in next header fields.
384 * Here we have, on "m"
386 * and we insert headers accordingly. Finally, we should be getting:
387 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
389 * during the header composing process, "m" points to IPv6 header.
390 * "mprev" points to an extension header prior to esp.
392 u_char *nexthdrp = &ip6->ip6_nxt;
396 * we treat dest2 specially. this makes IPsec processing
397 * much easier. the goal here is to make mprev point the
398 * mbuf prior to dest2.
400 * result: IPv6 dest2 payload
401 * m and mprev will point to IPv6 header.
403 if (exthdrs.ip6e_dest2) {
405 panic("assumption failed: hdr not split");
406 exthdrs.ip6e_dest2->m_next = m->m_next;
407 m->m_next = exthdrs.ip6e_dest2;
408 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
409 ip6->ip6_nxt = IPPROTO_DSTOPTS;
413 * result: IPv6 hbh dest1 rthdr dest2 payload
414 * m will point to IPv6 header. mprev will point to the
415 * extension header prior to dest2 (rthdr in the above case).
417 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
418 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
420 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
428 * pointers after IPsec headers are not valid any more.
429 * other pointers need a great care too.
430 * (IPsec routines should not mangle mbufs prior to AH/ESP)
432 exthdrs.ip6e_dest2 = NULL;
434 if (exthdrs.ip6e_rthdr) {
435 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
436 segleft_org = rh->ip6r_segleft;
437 rh->ip6r_segleft = 0;
440 bzero(&state, sizeof(state));
442 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
445 if (error == EJUSTRETURN) {
447 * We had a SP with a level of 'use' and no SA. We
448 * will just continue to process the packet without
453 /* mbuf is already reclaimed in ipsec6_output_trans. */
463 printf("[%s:%d] (ipsec): error code %d\n",
464 __func__, __LINE__, error);
467 /* don't show these error codes to the user */
472 } else if (!needipsectun) {
474 * In the FAST IPSec case we have already
475 * re-injected the packet and it has been freed
476 * by the ipsec_done() function. So, just clean
477 * up after ourselves.
482 if (exthdrs.ip6e_rthdr) {
483 /* ah6_output doesn't modify mbuf chain */
484 rh->ip6r_segleft = segleft_org;
490 * If there is a routing header, discard the packet.
492 if (exthdrs.ip6e_rthdr) {
497 /* Source address validation */
498 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
499 (flags & IPV6_UNSPECSRC) == 0) {
501 V_ip6stat.ip6s_badscope++;
504 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
506 V_ip6stat.ip6s_badscope++;
510 V_ip6stat.ip6s_localout++;
517 bzero((caddr_t)ro, sizeof(*ro));
520 if (opt && opt->ip6po_rthdr)
521 ro = &opt->ip6po_route;
522 dst = (struct sockaddr_in6 *)&ro->ro_dst;
524 if (ro->ro_rt == NULL) {
528 * The flow table returns route entries valid for up to 30
529 * seconds; we rely on the remainder of ip_output() taking no
530 * longer than that long for the stability of ro_rt. The
531 * flow ID assignment must have happened before this point.
533 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
535 flow_to_route_in6(fle, ro);
540 * if specified, try to fill in the traffic class field.
541 * do not override if a non-zero value is already set.
542 * we check the diffserv field and the ecn field separately.
544 if (opt && opt->ip6po_tclass >= 0) {
547 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
549 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
552 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
555 /* fill in or override the hop limit field, if necessary. */
556 if (opt && opt->ip6po_hlim != -1)
557 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
558 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
560 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
562 ip6->ip6_hlim = V_ip6_defmcasthlim;
567 * We may re-inject packets into the stack here.
569 if (needipsec && needipsectun) {
570 struct ipsec_output_state state;
573 * All the extension headers will become inaccessible
574 * (since they can be encrypted).
575 * Don't panic, we need no more updates to extension headers
576 * on inner IPv6 packet (since they are now encapsulated).
578 * IPv6 [ESP|AH] IPv6 [extension headers] payload
580 bzero(&exthdrs, sizeof(exthdrs));
581 exthdrs.ip6e_ip6 = m;
583 bzero(&state, sizeof(state));
585 state.ro = (struct route *)ro;
586 state.dst = (struct sockaddr *)dst;
588 error = ipsec6_output_tunnel(&state, sp, flags);
591 ro = (struct route_in6 *)state.ro;
592 dst = (struct sockaddr_in6 *)state.dst;
593 if (error == EJUSTRETURN) {
595 * We had a SP with a level of 'use' and no SA. We
596 * will just continue to process the packet without
601 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
612 printf("[%s:%d] (ipsec): error code %d\n",
613 __func__, __LINE__, error);
616 /* don't show these error codes to the user */
623 * In the FAST IPSec case we have already
624 * re-injected the packet and it has been freed
625 * by the ipsec_done() function. So, just clean
626 * up after ourselves.
632 exthdrs.ip6e_ip6 = m;
637 ip6 = mtod(m, struct ip6_hdr *);
639 if (ro->ro_rt && fwd_tag == NULL) {
641 ifp = ro->ro_rt->rt_ifp;
643 if (fwd_tag == NULL) {
644 bzero(&dst_sa, sizeof(dst_sa));
645 dst_sa.sin6_family = AF_INET6;
646 dst_sa.sin6_len = sizeof(dst_sa);
647 dst_sa.sin6_addr = ip6->ip6_dst;
649 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
650 &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
653 in6_ifstat_inc(ifp, ifs6_out_discard);
659 * If in6_selectroute() does not return a route entry,
660 * dst may not have been updated.
662 *dst = dst_sa; /* XXX */
666 * then rt (for unicast) and ifp must be non-NULL valid values.
668 if ((flags & IPV6_FORWARDING) == 0) {
669 /* XXX: the FORWARDING flag can be set for mrouting. */
670 in6_ifstat_inc(ifp, ifs6_out_request);
673 ia = (struct in6_ifaddr *)(rt->rt_ifa);
679 * The outgoing interface must be in the zone of source and
680 * destination addresses.
685 if (in6_setscope(&src0, origifp, &zone))
687 bzero(&src_sa, sizeof(src_sa));
688 src_sa.sin6_family = AF_INET6;
689 src_sa.sin6_len = sizeof(src_sa);
690 src_sa.sin6_addr = ip6->ip6_src;
691 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
695 if (in6_setscope(&dst0, origifp, &zone))
697 /* re-initialize to be sure */
698 bzero(&dst_sa, sizeof(dst_sa));
699 dst_sa.sin6_family = AF_INET6;
700 dst_sa.sin6_len = sizeof(dst_sa);
701 dst_sa.sin6_addr = ip6->ip6_dst;
702 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
706 /* We should use ia_ifp to support the case of
707 * sending packets to an address of our own.
709 if (ia != NULL && ia->ia_ifp)
712 /* scope check is done. */
716 V_ip6stat.ip6s_badscope++;
717 in6_ifstat_inc(origifp, ifs6_out_discard);
719 error = EHOSTUNREACH; /* XXX */
723 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
724 if (opt && opt->ip6po_nextroute.ro_rt) {
726 * The nexthop is explicitly specified by the
727 * application. We assume the next hop is an IPv6
730 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
732 else if ((rt->rt_flags & RTF_GATEWAY))
733 dst = (struct sockaddr_in6 *)rt->rt_gateway;
736 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
737 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
739 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
740 in6_ifstat_inc(ifp, ifs6_out_mcast);
742 * Confirm that the outgoing interface supports multicast.
744 if (!(ifp->if_flags & IFF_MULTICAST)) {
745 V_ip6stat.ip6s_noroute++;
746 in6_ifstat_inc(ifp, ifs6_out_discard);
750 if ((im6o == NULL && in6_mcast_loop) ||
751 (im6o && im6o->im6o_multicast_loop)) {
753 * Loop back multicast datagram if not expressly
754 * forbidden to do so, even if we have not joined
755 * the address; protocols will filter it later,
756 * thus deferring a hash lookup and lock acquisition
757 * at the expense of an m_copym().
759 ip6_mloopback(ifp, m, dst);
762 * If we are acting as a multicast router, perform
763 * multicast forwarding as if the packet had just
764 * arrived on the interface to which we are about
765 * to send. The multicast forwarding function
766 * recursively calls this function, using the
767 * IPV6_FORWARDING flag to prevent infinite recursion.
769 * Multicasts that are looped back by ip6_mloopback(),
770 * above, will be forwarded by the ip6_input() routine,
773 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
775 * XXX: ip6_mforward expects that rcvif is NULL
776 * when it is called from the originating path.
777 * However, it may not always be the case.
779 m->m_pkthdr.rcvif = NULL;
780 if (ip6_mforward(ip6, ifp, m) != 0) {
787 * Multicasts with a hoplimit of zero may be looped back,
788 * above, but must not be transmitted on a network.
789 * Also, multicasts addressed to the loopback interface
790 * are not sent -- the above call to ip6_mloopback() will
791 * loop back a copy if this host actually belongs to the
792 * destination group on the loopback interface.
794 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
795 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
802 * Fill the outgoing inteface to tell the upper layer
803 * to increment per-interface statistics.
808 /* Determine path MTU. */
809 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
810 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
814 * The caller of this function may specify to use the minimum MTU
816 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
817 * setting. The logic is a bit complicated; by default, unicast
818 * packets will follow path MTU while multicast packets will be sent at
819 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
820 * including unicast ones will be sent at the minimum MTU. Multicast
821 * packets will always be sent at the minimum MTU unless
822 * IP6PO_MINMTU_DISABLE is explicitly specified.
823 * See RFC 3542 for more details.
825 if (mtu > IPV6_MMTU) {
826 if ((flags & IPV6_MINMTU))
828 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
830 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
832 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
838 * clear embedded scope identifiers if necessary.
839 * in6_clearscope will touch the addresses only when necessary.
841 in6_clearscope(&ip6->ip6_src);
842 in6_clearscope(&ip6->ip6_dst);
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 dummy; /* XXX unused */
852 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
855 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
856 panic("ip6e_hbh is not contiguous");
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 * contiguous unless the flag is set.
864 m->m_flags |= M_LOOP;
865 m->m_pkthdr.rcvif = ifp;
866 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
867 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
868 &dummy, &plen) < 0) {
869 /* m was already freed at this point */
870 error = EINVAL;/* better error? */
873 m->m_flags &= ~M_LOOP; /* XXX */
874 m->m_pkthdr.rcvif = NULL;
877 /* Jump over all PFIL processing if hooks are not active. */
878 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
882 /* Run through list of hooks for output packets. */
883 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
884 if (error != 0 || m == NULL)
886 ip6 = mtod(m, struct ip6_hdr *);
888 /* See if destination IP address was changed by packet filter. */
889 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
890 m->m_flags |= M_SKIP_FIREWALL;
891 /* If destination is now ourself drop to ip6_input(). */
892 if (in6_localip(&ip6->ip6_dst)) {
893 m->m_flags |= M_FASTFWD_OURS;
894 if (m->m_pkthdr.rcvif == NULL)
895 m->m_pkthdr.rcvif = V_loif;
896 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
897 m->m_pkthdr.csum_flags |=
898 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
899 m->m_pkthdr.csum_data = 0xffff;
902 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
903 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
905 error = netisr_queue(NETISR_IPV6, m);
908 goto again; /* Redo the routing table lookup. */
911 /* See if local, if yes, send it to netisr. */
912 if (m->m_flags & M_FASTFWD_OURS) {
913 if (m->m_pkthdr.rcvif == NULL)
914 m->m_pkthdr.rcvif = V_loif;
915 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
916 m->m_pkthdr.csum_flags |=
917 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
918 m->m_pkthdr.csum_data = 0xffff;
921 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
922 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
924 error = netisr_queue(NETISR_IPV6, m);
927 /* Or forward to some other address? */
928 if ((m->m_flags & M_IP6_NEXTHOP) &&
929 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
930 dst = (struct sockaddr_in6 *)&ro->ro_dst;
931 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
932 m->m_flags |= M_SKIP_FIREWALL;
933 m->m_flags &= ~M_IP6_NEXTHOP;
934 m_tag_delete(m, fwd_tag);
940 * Send the packet to the outgoing interface.
941 * If necessary, do IPv6 fragmentation before sending.
943 * the logic here is rather complex:
944 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
945 * 1-a: send as is if tlen <= path mtu
946 * 1-b: fragment if tlen > path mtu
948 * 2: if user asks us not to fragment (dontfrag == 1)
949 * 2-a: send as is if tlen <= interface mtu
950 * 2-b: error if tlen > interface mtu
952 * 3: if we always need to attach fragment header (alwaysfrag == 1)
955 * 4: if dontfrag == 1 && alwaysfrag == 1
956 * error, as we cannot handle this conflicting request
958 sw_csum = m->m_pkthdr.csum_flags;
960 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
961 sw_csum &= ~ifp->if_hwassist;
965 * If we added extension headers, we will not do TSO and calculate the
966 * checksums ourselves for now.
967 * XXX-BZ Need a framework to know when the NIC can handle it, even
970 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
971 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
972 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
975 if (sw_csum & CSUM_SCTP_IPV6) {
976 sw_csum &= ~CSUM_SCTP_IPV6;
977 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
980 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
981 tlen = m->m_pkthdr.len;
983 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
987 if (dontfrag && alwaysfrag) { /* case 4 */
988 /* conflicting request - can't transmit */
992 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
994 * Even if the DONTFRAG option is specified, we cannot send the
995 * packet when the data length is larger than the MTU of the
996 * outgoing interface.
997 * Notify the error by sending IPV6_PATHMTU ancillary data as
998 * well as returning an error code (the latter is not described
1002 struct ip6ctlparam ip6cp;
1004 mtu32 = (u_int32_t)mtu;
1005 bzero(&ip6cp, sizeof(ip6cp));
1006 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1007 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1015 * transmit packet without fragmentation
1017 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1018 struct in6_ifaddr *ia6;
1020 ip6 = mtod(m, struct ip6_hdr *);
1021 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1023 /* Record statistics for this interface address. */
1024 ia6->ia_ifa.if_opackets++;
1025 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1026 ifa_free(&ia6->ia_ifa);
1028 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1033 * try to fragment the packet. case 1-b and 3
1035 if (mtu < IPV6_MMTU) {
1036 /* path MTU cannot be less than IPV6_MMTU */
1038 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1040 } else if (ip6->ip6_plen == 0) {
1041 /* jumbo payload cannot be fragmented */
1043 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1046 struct mbuf **mnext, *m_frgpart;
1047 struct ip6_frag *ip6f;
1048 u_int32_t id = htonl(ip6_randomid());
1051 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1054 * Too large for the destination or interface;
1055 * fragment if possible.
1056 * Must be able to put at least 8 bytes per fragment.
1058 hlen = unfragpartlen;
1059 if (mtu > IPV6_MAXPACKET)
1060 mtu = IPV6_MAXPACKET;
1062 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1065 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1070 * Verify that we have any chance at all of being able to queue
1071 * the packet or packet fragments
1073 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1074 < tlen /* - hlen */)) {
1076 V_ip6stat.ip6s_odropped++;
1082 * If the interface will not calculate checksums on
1083 * fragmented packets, then do it here.
1084 * XXX-BZ handle the hw offloading case. Need flags.
1086 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1087 in6_delayed_cksum(m, plen, hlen);
1088 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1091 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1092 sctp_delayed_cksum(m, hlen);
1093 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1096 mnext = &m->m_nextpkt;
1099 * Change the next header field of the last header in the
1100 * unfragmentable part.
1102 if (exthdrs.ip6e_rthdr) {
1103 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1104 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1105 } else if (exthdrs.ip6e_dest1) {
1106 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1107 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1108 } else if (exthdrs.ip6e_hbh) {
1109 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1110 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1112 nextproto = ip6->ip6_nxt;
1113 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1117 * Loop through length of segment after first fragment,
1118 * make new header and copy data of each part and link onto
1122 for (off = hlen; off < tlen; off += len) {
1123 m = m_gethdr(M_NOWAIT, MT_DATA);
1126 V_ip6stat.ip6s_odropped++;
1129 m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1131 mnext = &m->m_nextpkt;
1132 m->m_data += max_linkhdr;
1133 mhip6 = mtod(m, struct ip6_hdr *);
1135 m->m_len = sizeof(*mhip6);
1136 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1138 V_ip6stat.ip6s_odropped++;
1141 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1142 if (off + len >= tlen)
1145 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1146 mhip6->ip6_plen = htons((u_short)(len + hlen +
1147 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1148 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1150 V_ip6stat.ip6s_odropped++;
1153 m_cat(m, m_frgpart);
1154 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1155 m->m_pkthdr.rcvif = NULL;
1156 ip6f->ip6f_reserved = 0;
1157 ip6f->ip6f_ident = id;
1158 ip6f->ip6f_nxt = nextproto;
1159 V_ip6stat.ip6s_ofragments++;
1160 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1163 in6_ifstat_inc(ifp, ifs6_out_fragok);
1167 * Remove leading garbages.
1173 for (m0 = m; m; m = m0) {
1177 /* Record statistics for this interface address. */
1179 ia->ia_ifa.if_opackets++;
1180 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1182 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1188 V_ip6stat.ip6s_fragmented++;
1191 if (ro == &ip6route)
1193 if (ro_pmtu == &ip6route)
1203 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1204 m_freem(exthdrs.ip6e_dest1);
1205 m_freem(exthdrs.ip6e_rthdr);
1206 m_freem(exthdrs.ip6e_dest2);
1215 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1219 if (hlen > MCLBYTES)
1220 return (ENOBUFS); /* XXX */
1223 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1225 m = m_get(M_NOWAIT, MT_DATA);
1230 bcopy(hdr, mtod(m, caddr_t), hlen);
1237 * Insert jumbo payload option.
1240 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1246 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1249 * If there is no hop-by-hop options header, allocate new one.
1250 * If there is one but it doesn't have enough space to store the
1251 * jumbo payload option, allocate a cluster to store the whole options.
1252 * Otherwise, use it to store the options.
1254 if (exthdrs->ip6e_hbh == 0) {
1255 mopt = m_get(M_NOWAIT, MT_DATA);
1258 mopt->m_len = JUMBOOPTLEN;
1259 optbuf = mtod(mopt, u_char *);
1260 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1261 exthdrs->ip6e_hbh = mopt;
1263 struct ip6_hbh *hbh;
1265 mopt = exthdrs->ip6e_hbh;
1266 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1269 * - exthdrs->ip6e_hbh is not referenced from places
1270 * other than exthdrs.
1271 * - exthdrs->ip6e_hbh is not an mbuf chain.
1273 int oldoptlen = mopt->m_len;
1277 * XXX: give up if the whole (new) hbh header does
1278 * not fit even in an mbuf cluster.
1280 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1284 * As a consequence, we must always prepare a cluster
1287 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1290 n->m_len = oldoptlen + JUMBOOPTLEN;
1291 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1293 optbuf = mtod(n, caddr_t) + oldoptlen;
1295 mopt = exthdrs->ip6e_hbh = n;
1297 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1298 mopt->m_len += JUMBOOPTLEN;
1300 optbuf[0] = IP6OPT_PADN;
1304 * Adjust the header length according to the pad and
1305 * the jumbo payload option.
1307 hbh = mtod(mopt, struct ip6_hbh *);
1308 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1311 /* fill in the option. */
1312 optbuf[2] = IP6OPT_JUMBO;
1314 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1315 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1317 /* finally, adjust the packet header length */
1318 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1325 * Insert fragment header and copy unfragmentable header portions.
1328 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1329 struct ip6_frag **frghdrp)
1331 struct mbuf *n, *mlast;
1333 if (hlen > sizeof(struct ip6_hdr)) {
1334 n = m_copym(m0, sizeof(struct ip6_hdr),
1335 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1342 /* Search for the last mbuf of unfragmentable part. */
1343 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1346 if ((mlast->m_flags & M_EXT) == 0 &&
1347 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1348 /* use the trailing space of the last mbuf for the fragment hdr */
1349 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1351 mlast->m_len += sizeof(struct ip6_frag);
1352 m->m_pkthdr.len += sizeof(struct ip6_frag);
1354 /* allocate a new mbuf for the fragment header */
1357 mfrg = m_get(M_NOWAIT, MT_DATA);
1360 mfrg->m_len = sizeof(struct ip6_frag);
1361 *frghdrp = mtod(mfrg, struct ip6_frag *);
1362 mlast->m_next = mfrg;
1369 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1370 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1371 int *alwaysfragp, u_int fibnum)
1377 if (ro_pmtu != ro) {
1378 /* The first hop and the final destination may differ. */
1379 struct sockaddr_in6 *sa6_dst =
1380 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1381 if (ro_pmtu->ro_rt &&
1382 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1383 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1384 RTFREE(ro_pmtu->ro_rt);
1385 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1387 if (ro_pmtu->ro_rt == NULL) {
1388 bzero(sa6_dst, sizeof(*sa6_dst));
1389 sa6_dst->sin6_family = AF_INET6;
1390 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1391 sa6_dst->sin6_addr = *dst;
1393 in6_rtalloc(ro_pmtu, fibnum);
1396 if (ro_pmtu->ro_rt) {
1398 struct in_conninfo inc;
1400 bzero(&inc, sizeof(inc));
1401 inc.inc_flags |= INC_ISIPV6;
1402 inc.inc6_faddr = *dst;
1405 ifp = ro_pmtu->ro_rt->rt_ifp;
1406 ifmtu = IN6_LINKMTU(ifp);
1407 mtu = tcp_hc_getmtu(&inc);
1409 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1411 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1414 else if (mtu < IPV6_MMTU) {
1416 * RFC2460 section 5, last paragraph:
1417 * if we record ICMPv6 too big message with
1418 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1419 * or smaller, with framgent header attached.
1420 * (fragment header is needed regardless from the
1421 * packet size, for translators to identify packets)
1425 } else if (mtu > ifmtu) {
1427 * The MTU on the route is larger than the MTU on
1428 * the interface! This shouldn't happen, unless the
1429 * MTU of the interface has been changed after the
1430 * interface was brought up. Change the MTU in the
1431 * route to match the interface MTU (as long as the
1432 * field isn't locked).
1435 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1438 mtu = IN6_LINKMTU(ifp);
1440 error = EHOSTUNREACH; /* XXX */
1444 *alwaysfragp = alwaysfrag;
1449 * IP6 socket option processing.
1452 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1454 int optdatalen, uproto;
1456 struct inpcb *in6p = sotoinpcb(so);
1458 int level, op, optname;
1462 level = sopt->sopt_level;
1463 op = sopt->sopt_dir;
1464 optname = sopt->sopt_name;
1465 optlen = sopt->sopt_valsize;
1469 uproto = (int)so->so_proto->pr_protocol;
1471 if (level != IPPROTO_IPV6) {
1474 if (sopt->sopt_level == SOL_SOCKET &&
1475 sopt->sopt_dir == SOPT_SET) {
1476 switch (sopt->sopt_name) {
1479 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1480 if ((so->so_options &
1481 (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1482 in6p->inp_flags2 |= INP_REUSEPORT;
1484 in6p->inp_flags2 &= ~INP_REUSEPORT;
1491 if ((so->so_options & SO_REUSEPORT) != 0)
1492 in6p->inp_flags2 |= INP_REUSEPORT;
1494 in6p->inp_flags2 &= ~INP_REUSEPORT;
1500 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1508 } else { /* level == IPPROTO_IPV6 */
1513 case IPV6_2292PKTOPTIONS:
1514 #ifdef IPV6_PKTOPTIONS
1515 case IPV6_PKTOPTIONS:
1520 error = soopt_getm(sopt, &m); /* XXX */
1523 error = soopt_mcopyin(sopt, m); /* XXX */
1526 error = ip6_pcbopts(&in6p->in6p_outputopts,
1528 m_freem(m); /* XXX */
1533 * Use of some Hop-by-Hop options or some
1534 * Destination options, might require special
1535 * privilege. That is, normal applications
1536 * (without special privilege) might be forbidden
1537 * from setting certain options in outgoing packets,
1538 * and might never see certain options in received
1539 * packets. [RFC 2292 Section 6]
1540 * KAME specific note:
1541 * KAME prevents non-privileged users from sending or
1542 * receiving ANY hbh/dst options in order to avoid
1543 * overhead of parsing options in the kernel.
1545 case IPV6_RECVHOPOPTS:
1546 case IPV6_RECVDSTOPTS:
1547 case IPV6_RECVRTHDRDSTOPTS:
1549 error = priv_check(td,
1550 PRIV_NETINET_SETHDROPTS);
1555 case IPV6_UNICAST_HOPS:
1559 case IPV6_RECVPKTINFO:
1560 case IPV6_RECVHOPLIMIT:
1561 case IPV6_RECVRTHDR:
1562 case IPV6_RECVPATHMTU:
1563 case IPV6_RECVTCLASS:
1565 case IPV6_AUTOFLOWLABEL:
1567 if (optname == IPV6_BINDANY && td != NULL) {
1568 error = priv_check(td,
1569 PRIV_NETINET_BINDANY);
1574 if (optlen != sizeof(int)) {
1578 error = sooptcopyin(sopt, &optval,
1579 sizeof optval, sizeof optval);
1584 case IPV6_UNICAST_HOPS:
1585 if (optval < -1 || optval >= 256)
1588 /* -1 = kernel default */
1589 in6p->in6p_hops = optval;
1590 if ((in6p->inp_vflag &
1592 in6p->inp_ip_ttl = optval;
1595 #define OPTSET(bit) \
1599 in6p->inp_flags |= (bit); \
1601 in6p->inp_flags &= ~(bit); \
1602 INP_WUNLOCK(in6p); \
1603 } while (/*CONSTCOND*/ 0)
1604 #define OPTSET2292(bit) \
1607 in6p->inp_flags |= IN6P_RFC2292; \
1609 in6p->inp_flags |= (bit); \
1611 in6p->inp_flags &= ~(bit); \
1612 INP_WUNLOCK(in6p); \
1613 } while (/*CONSTCOND*/ 0)
1614 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1616 case IPV6_RECVPKTINFO:
1617 /* cannot mix with RFC2292 */
1618 if (OPTBIT(IN6P_RFC2292)) {
1622 OPTSET(IN6P_PKTINFO);
1627 struct ip6_pktopts **optp;
1629 /* cannot mix with RFC2292 */
1630 if (OPTBIT(IN6P_RFC2292)) {
1634 optp = &in6p->in6p_outputopts;
1635 error = ip6_pcbopt(IPV6_HOPLIMIT,
1636 (u_char *)&optval, sizeof(optval),
1637 optp, (td != NULL) ? td->td_ucred :
1642 case IPV6_RECVHOPLIMIT:
1643 /* cannot mix with RFC2292 */
1644 if (OPTBIT(IN6P_RFC2292)) {
1648 OPTSET(IN6P_HOPLIMIT);
1651 case IPV6_RECVHOPOPTS:
1652 /* cannot mix with RFC2292 */
1653 if (OPTBIT(IN6P_RFC2292)) {
1657 OPTSET(IN6P_HOPOPTS);
1660 case IPV6_RECVDSTOPTS:
1661 /* cannot mix with RFC2292 */
1662 if (OPTBIT(IN6P_RFC2292)) {
1666 OPTSET(IN6P_DSTOPTS);
1669 case IPV6_RECVRTHDRDSTOPTS:
1670 /* cannot mix with RFC2292 */
1671 if (OPTBIT(IN6P_RFC2292)) {
1675 OPTSET(IN6P_RTHDRDSTOPTS);
1678 case IPV6_RECVRTHDR:
1679 /* cannot mix with RFC2292 */
1680 if (OPTBIT(IN6P_RFC2292)) {
1691 case IPV6_RECVPATHMTU:
1693 * We ignore this option for TCP
1695 * (RFC3542 leaves this case
1698 if (uproto != IPPROTO_TCP)
1704 * make setsockopt(IPV6_V6ONLY)
1705 * available only prior to bind(2).
1706 * see ipng mailing list, Jun 22 2001.
1708 if (in6p->inp_lport ||
1709 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1713 OPTSET(IN6P_IPV6_V6ONLY);
1715 in6p->inp_vflag &= ~INP_IPV4;
1717 in6p->inp_vflag |= INP_IPV4;
1719 case IPV6_RECVTCLASS:
1720 /* cannot mix with RFC2292 XXX */
1721 if (OPTBIT(IN6P_RFC2292)) {
1725 OPTSET(IN6P_TCLASS);
1727 case IPV6_AUTOFLOWLABEL:
1728 OPTSET(IN6P_AUTOFLOWLABEL);
1732 OPTSET(INP_BINDANY);
1739 case IPV6_USE_MIN_MTU:
1740 case IPV6_PREFER_TEMPADDR:
1741 if (optlen != sizeof(optval)) {
1745 error = sooptcopyin(sopt, &optval,
1746 sizeof optval, sizeof optval);
1750 struct ip6_pktopts **optp;
1751 optp = &in6p->in6p_outputopts;
1752 error = ip6_pcbopt(optname,
1753 (u_char *)&optval, sizeof(optval),
1754 optp, (td != NULL) ? td->td_ucred :
1759 case IPV6_2292PKTINFO:
1760 case IPV6_2292HOPLIMIT:
1761 case IPV6_2292HOPOPTS:
1762 case IPV6_2292DSTOPTS:
1763 case IPV6_2292RTHDR:
1765 if (optlen != sizeof(int)) {
1769 error = sooptcopyin(sopt, &optval,
1770 sizeof optval, sizeof optval);
1774 case IPV6_2292PKTINFO:
1775 OPTSET2292(IN6P_PKTINFO);
1777 case IPV6_2292HOPLIMIT:
1778 OPTSET2292(IN6P_HOPLIMIT);
1780 case IPV6_2292HOPOPTS:
1782 * Check super-user privilege.
1783 * See comments for IPV6_RECVHOPOPTS.
1786 error = priv_check(td,
1787 PRIV_NETINET_SETHDROPTS);
1791 OPTSET2292(IN6P_HOPOPTS);
1793 case IPV6_2292DSTOPTS:
1795 error = priv_check(td,
1796 PRIV_NETINET_SETHDROPTS);
1800 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1802 case IPV6_2292RTHDR:
1803 OPTSET2292(IN6P_RTHDR);
1811 case IPV6_RTHDRDSTOPTS:
1814 /* new advanced API (RFC3542) */
1816 u_char optbuf_storage[MCLBYTES];
1818 struct ip6_pktopts **optp;
1820 /* cannot mix with RFC2292 */
1821 if (OPTBIT(IN6P_RFC2292)) {
1827 * We only ensure valsize is not too large
1828 * here. Further validation will be done
1831 error = sooptcopyin(sopt, optbuf_storage,
1832 sizeof(optbuf_storage), 0);
1835 optlen = sopt->sopt_valsize;
1836 optbuf = optbuf_storage;
1837 optp = &in6p->in6p_outputopts;
1838 error = ip6_pcbopt(optname, optbuf, optlen,
1839 optp, (td != NULL) ? td->td_ucred : NULL,
1845 case IPV6_MULTICAST_IF:
1846 case IPV6_MULTICAST_HOPS:
1847 case IPV6_MULTICAST_LOOP:
1848 case IPV6_JOIN_GROUP:
1849 case IPV6_LEAVE_GROUP:
1851 case MCAST_BLOCK_SOURCE:
1852 case MCAST_UNBLOCK_SOURCE:
1853 case MCAST_JOIN_GROUP:
1854 case MCAST_LEAVE_GROUP:
1855 case MCAST_JOIN_SOURCE_GROUP:
1856 case MCAST_LEAVE_SOURCE_GROUP:
1857 error = ip6_setmoptions(in6p, sopt);
1860 case IPV6_PORTRANGE:
1861 error = sooptcopyin(sopt, &optval,
1862 sizeof optval, sizeof optval);
1868 case IPV6_PORTRANGE_DEFAULT:
1869 in6p->inp_flags &= ~(INP_LOWPORT);
1870 in6p->inp_flags &= ~(INP_HIGHPORT);
1873 case IPV6_PORTRANGE_HIGH:
1874 in6p->inp_flags &= ~(INP_LOWPORT);
1875 in6p->inp_flags |= INP_HIGHPORT;
1878 case IPV6_PORTRANGE_LOW:
1879 in6p->inp_flags &= ~(INP_HIGHPORT);
1880 in6p->inp_flags |= INP_LOWPORT;
1891 case IPV6_IPSEC_POLICY:
1896 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1898 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1900 req = mtod(m, caddr_t);
1901 error = ipsec_set_policy(in6p, optname, req,
1902 m->m_len, (sopt->sopt_td != NULL) ?
1903 sopt->sopt_td->td_ucred : NULL);
1910 error = ENOPROTOOPT;
1918 case IPV6_2292PKTOPTIONS:
1919 #ifdef IPV6_PKTOPTIONS
1920 case IPV6_PKTOPTIONS:
1923 * RFC3542 (effectively) deprecated the
1924 * semantics of the 2292-style pktoptions.
1925 * Since it was not reliable in nature (i.e.,
1926 * applications had to expect the lack of some
1927 * information after all), it would make sense
1928 * to simplify this part by always returning
1931 sopt->sopt_valsize = 0;
1934 case IPV6_RECVHOPOPTS:
1935 case IPV6_RECVDSTOPTS:
1936 case IPV6_RECVRTHDRDSTOPTS:
1937 case IPV6_UNICAST_HOPS:
1938 case IPV6_RECVPKTINFO:
1939 case IPV6_RECVHOPLIMIT:
1940 case IPV6_RECVRTHDR:
1941 case IPV6_RECVPATHMTU:
1945 case IPV6_PORTRANGE:
1946 case IPV6_RECVTCLASS:
1947 case IPV6_AUTOFLOWLABEL:
1951 case IPV6_RECVHOPOPTS:
1952 optval = OPTBIT(IN6P_HOPOPTS);
1955 case IPV6_RECVDSTOPTS:
1956 optval = OPTBIT(IN6P_DSTOPTS);
1959 case IPV6_RECVRTHDRDSTOPTS:
1960 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1963 case IPV6_UNICAST_HOPS:
1964 optval = in6p->in6p_hops;
1967 case IPV6_RECVPKTINFO:
1968 optval = OPTBIT(IN6P_PKTINFO);
1971 case IPV6_RECVHOPLIMIT:
1972 optval = OPTBIT(IN6P_HOPLIMIT);
1975 case IPV6_RECVRTHDR:
1976 optval = OPTBIT(IN6P_RTHDR);
1979 case IPV6_RECVPATHMTU:
1980 optval = OPTBIT(IN6P_MTU);
1984 optval = OPTBIT(INP_FAITH);
1988 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1991 case IPV6_PORTRANGE:
1994 flags = in6p->inp_flags;
1995 if (flags & INP_HIGHPORT)
1996 optval = IPV6_PORTRANGE_HIGH;
1997 else if (flags & INP_LOWPORT)
1998 optval = IPV6_PORTRANGE_LOW;
2003 case IPV6_RECVTCLASS:
2004 optval = OPTBIT(IN6P_TCLASS);
2007 case IPV6_AUTOFLOWLABEL:
2008 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2012 optval = OPTBIT(INP_BINDANY);
2017 error = sooptcopyout(sopt, &optval,
2024 struct ip6_mtuinfo mtuinfo;
2025 struct route_in6 sro;
2027 bzero(&sro, sizeof(sro));
2029 if (!(so->so_state & SS_ISCONNECTED))
2032 * XXX: we dot not consider the case of source
2033 * routing, or optional information to specify
2034 * the outgoing interface.
2036 error = ip6_getpmtu(&sro, NULL, NULL,
2037 &in6p->in6p_faddr, &pmtu, NULL,
2043 if (pmtu > IPV6_MAXPACKET)
2044 pmtu = IPV6_MAXPACKET;
2046 bzero(&mtuinfo, sizeof(mtuinfo));
2047 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2048 optdata = (void *)&mtuinfo;
2049 optdatalen = sizeof(mtuinfo);
2050 error = sooptcopyout(sopt, optdata,
2055 case IPV6_2292PKTINFO:
2056 case IPV6_2292HOPLIMIT:
2057 case IPV6_2292HOPOPTS:
2058 case IPV6_2292RTHDR:
2059 case IPV6_2292DSTOPTS:
2061 case IPV6_2292PKTINFO:
2062 optval = OPTBIT(IN6P_PKTINFO);
2064 case IPV6_2292HOPLIMIT:
2065 optval = OPTBIT(IN6P_HOPLIMIT);
2067 case IPV6_2292HOPOPTS:
2068 optval = OPTBIT(IN6P_HOPOPTS);
2070 case IPV6_2292RTHDR:
2071 optval = OPTBIT(IN6P_RTHDR);
2073 case IPV6_2292DSTOPTS:
2074 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2077 error = sooptcopyout(sopt, &optval,
2084 case IPV6_RTHDRDSTOPTS:
2088 case IPV6_USE_MIN_MTU:
2089 case IPV6_PREFER_TEMPADDR:
2090 error = ip6_getpcbopt(in6p->in6p_outputopts,
2094 case IPV6_MULTICAST_IF:
2095 case IPV6_MULTICAST_HOPS:
2096 case IPV6_MULTICAST_LOOP:
2098 error = ip6_getmoptions(in6p, sopt);
2102 case IPV6_IPSEC_POLICY:
2106 struct mbuf *m = NULL;
2107 struct mbuf **mp = &m;
2108 size_t ovalsize = sopt->sopt_valsize;
2109 caddr_t oval = (caddr_t)sopt->sopt_val;
2111 error = soopt_getm(sopt, &m); /* XXX */
2114 error = soopt_mcopyin(sopt, m); /* XXX */
2117 sopt->sopt_valsize = ovalsize;
2118 sopt->sopt_val = oval;
2120 req = mtod(m, caddr_t);
2123 error = ipsec_get_policy(in6p, req, len, mp);
2125 error = soopt_mcopyout(sopt, m); /* XXX */
2126 if (error == 0 && m)
2133 error = ENOPROTOOPT;
2143 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2145 int error = 0, optval, optlen;
2146 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2147 struct inpcb *in6p = sotoinpcb(so);
2148 int level, op, optname;
2150 level = sopt->sopt_level;
2151 op = sopt->sopt_dir;
2152 optname = sopt->sopt_name;
2153 optlen = sopt->sopt_valsize;
2155 if (level != IPPROTO_IPV6) {
2162 * For ICMPv6 sockets, no modification allowed for checksum
2163 * offset, permit "no change" values to help existing apps.
2165 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2166 * for an ICMPv6 socket will fail."
2167 * The current behavior does not meet RFC3542.
2171 if (optlen != sizeof(int)) {
2175 error = sooptcopyin(sopt, &optval, sizeof(optval),
2179 if ((optval % 2) != 0) {
2180 /* the API assumes even offset values */
2182 } else if (so->so_proto->pr_protocol ==
2184 if (optval != icmp6off)
2187 in6p->in6p_cksum = optval;
2191 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2194 optval = in6p->in6p_cksum;
2196 error = sooptcopyout(sopt, &optval, sizeof(optval));
2206 error = ENOPROTOOPT;
2214 * Set up IP6 options in pcb for insertion in output packets or
2215 * specifying behavior of outgoing packets.
2218 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2219 struct socket *so, struct sockopt *sopt)
2221 struct ip6_pktopts *opt = *pktopt;
2223 struct thread *td = sopt->sopt_td;
2225 /* turn off any old options. */
2228 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2229 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2230 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2231 printf("ip6_pcbopts: all specified options are cleared.\n");
2233 ip6_clearpktopts(opt, -1);
2235 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2238 if (!m || m->m_len == 0) {
2240 * Only turning off any previous options, regardless of
2241 * whether the opt is just created or given.
2243 free(opt, M_IP6OPT);
2247 /* set options specified by user. */
2248 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2249 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2250 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2251 free(opt, M_IP6OPT);
2259 * initialize ip6_pktopts. beware that there are non-zero default values in
2263 ip6_initpktopts(struct ip6_pktopts *opt)
2266 bzero(opt, sizeof(*opt));
2267 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2268 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2269 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2270 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2274 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2275 struct ucred *cred, int uproto)
2277 struct ip6_pktopts *opt;
2279 if (*pktopt == NULL) {
2280 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2282 ip6_initpktopts(*pktopt);
2286 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2290 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2292 void *optdata = NULL;
2294 struct ip6_ext *ip6e;
2296 struct in6_pktinfo null_pktinfo;
2297 int deftclass = 0, on;
2298 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2299 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2303 if (pktopt && pktopt->ip6po_pktinfo)
2304 optdata = (void *)pktopt->ip6po_pktinfo;
2306 /* XXX: we don't have to do this every time... */
2307 bzero(&null_pktinfo, sizeof(null_pktinfo));
2308 optdata = (void *)&null_pktinfo;
2310 optdatalen = sizeof(struct in6_pktinfo);
2313 if (pktopt && pktopt->ip6po_tclass >= 0)
2314 optdata = (void *)&pktopt->ip6po_tclass;
2316 optdata = (void *)&deftclass;
2317 optdatalen = sizeof(int);
2320 if (pktopt && pktopt->ip6po_hbh) {
2321 optdata = (void *)pktopt->ip6po_hbh;
2322 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2323 optdatalen = (ip6e->ip6e_len + 1) << 3;
2327 if (pktopt && pktopt->ip6po_rthdr) {
2328 optdata = (void *)pktopt->ip6po_rthdr;
2329 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2330 optdatalen = (ip6e->ip6e_len + 1) << 3;
2333 case IPV6_RTHDRDSTOPTS:
2334 if (pktopt && pktopt->ip6po_dest1) {
2335 optdata = (void *)pktopt->ip6po_dest1;
2336 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2337 optdatalen = (ip6e->ip6e_len + 1) << 3;
2341 if (pktopt && pktopt->ip6po_dest2) {
2342 optdata = (void *)pktopt->ip6po_dest2;
2343 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2344 optdatalen = (ip6e->ip6e_len + 1) << 3;
2348 if (pktopt && pktopt->ip6po_nexthop) {
2349 optdata = (void *)pktopt->ip6po_nexthop;
2350 optdatalen = pktopt->ip6po_nexthop->sa_len;
2353 case IPV6_USE_MIN_MTU:
2355 optdata = (void *)&pktopt->ip6po_minmtu;
2357 optdata = (void *)&defminmtu;
2358 optdatalen = sizeof(int);
2361 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2365 optdata = (void *)&on;
2366 optdatalen = sizeof(on);
2368 case IPV6_PREFER_TEMPADDR:
2370 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2372 optdata = (void *)&defpreftemp;
2373 optdatalen = sizeof(int);
2375 default: /* should not happen */
2377 panic("ip6_getpcbopt: unexpected option\n");
2379 return (ENOPROTOOPT);
2382 error = sooptcopyout(sopt, optdata, optdatalen);
2388 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2393 if (optname == -1 || optname == IPV6_PKTINFO) {
2394 if (pktopt->ip6po_pktinfo)
2395 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2396 pktopt->ip6po_pktinfo = NULL;
2398 if (optname == -1 || optname == IPV6_HOPLIMIT)
2399 pktopt->ip6po_hlim = -1;
2400 if (optname == -1 || optname == IPV6_TCLASS)
2401 pktopt->ip6po_tclass = -1;
2402 if (optname == -1 || optname == IPV6_NEXTHOP) {
2403 if (pktopt->ip6po_nextroute.ro_rt) {
2404 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2405 pktopt->ip6po_nextroute.ro_rt = NULL;
2407 if (pktopt->ip6po_nexthop)
2408 free(pktopt->ip6po_nexthop, M_IP6OPT);
2409 pktopt->ip6po_nexthop = NULL;
2411 if (optname == -1 || optname == IPV6_HOPOPTS) {
2412 if (pktopt->ip6po_hbh)
2413 free(pktopt->ip6po_hbh, M_IP6OPT);
2414 pktopt->ip6po_hbh = NULL;
2416 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2417 if (pktopt->ip6po_dest1)
2418 free(pktopt->ip6po_dest1, M_IP6OPT);
2419 pktopt->ip6po_dest1 = NULL;
2421 if (optname == -1 || optname == IPV6_RTHDR) {
2422 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2423 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2424 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2425 if (pktopt->ip6po_route.ro_rt) {
2426 RTFREE(pktopt->ip6po_route.ro_rt);
2427 pktopt->ip6po_route.ro_rt = NULL;
2430 if (optname == -1 || optname == IPV6_DSTOPTS) {
2431 if (pktopt->ip6po_dest2)
2432 free(pktopt->ip6po_dest2, M_IP6OPT);
2433 pktopt->ip6po_dest2 = NULL;
2437 #define PKTOPT_EXTHDRCPY(type) \
2440 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2441 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2442 if (dst->type == NULL && canwait == M_NOWAIT)\
2444 bcopy(src->type, dst->type, hlen);\
2446 } while (/*CONSTCOND*/ 0)
2449 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2451 if (dst == NULL || src == NULL) {
2452 printf("ip6_clearpktopts: invalid argument\n");
2456 dst->ip6po_hlim = src->ip6po_hlim;
2457 dst->ip6po_tclass = src->ip6po_tclass;
2458 dst->ip6po_flags = src->ip6po_flags;
2459 dst->ip6po_minmtu = src->ip6po_minmtu;
2460 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2461 if (src->ip6po_pktinfo) {
2462 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2464 if (dst->ip6po_pktinfo == NULL)
2466 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2468 if (src->ip6po_nexthop) {
2469 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2471 if (dst->ip6po_nexthop == NULL)
2473 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2474 src->ip6po_nexthop->sa_len);
2476 PKTOPT_EXTHDRCPY(ip6po_hbh);
2477 PKTOPT_EXTHDRCPY(ip6po_dest1);
2478 PKTOPT_EXTHDRCPY(ip6po_dest2);
2479 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2483 ip6_clearpktopts(dst, -1);
2486 #undef PKTOPT_EXTHDRCPY
2488 struct ip6_pktopts *
2489 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2492 struct ip6_pktopts *dst;
2494 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2497 ip6_initpktopts(dst);
2499 if ((error = copypktopts(dst, src, canwait)) != 0) {
2500 free(dst, M_IP6OPT);
2508 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2513 ip6_clearpktopts(pktopt, -1);
2515 free(pktopt, M_IP6OPT);
2519 * Set IPv6 outgoing packet options based on advanced API.
2522 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2523 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2525 struct cmsghdr *cm = 0;
2527 if (control == NULL || opt == NULL)
2530 ip6_initpktopts(opt);
2535 * If stickyopt is provided, make a local copy of the options
2536 * for this particular packet, then override them by ancillary
2538 * XXX: copypktopts() does not copy the cached route to a next
2539 * hop (if any). This is not very good in terms of efficiency,
2540 * but we can allow this since this option should be rarely
2543 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2548 * XXX: Currently, we assume all the optional information is stored
2551 if (control->m_next)
2554 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2555 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2558 if (control->m_len < CMSG_LEN(0))
2561 cm = mtod(control, struct cmsghdr *);
2562 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2564 if (cm->cmsg_level != IPPROTO_IPV6)
2567 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2568 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2577 * Set a particular packet option, as a sticky option or an ancillary data
2578 * item. "len" can be 0 only when it's a sticky option.
2579 * We have 4 cases of combination of "sticky" and "cmsg":
2580 * "sticky=0, cmsg=0": impossible
2581 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2582 * "sticky=1, cmsg=0": RFC3542 socket option
2583 * "sticky=1, cmsg=1": RFC2292 socket option
2586 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2587 struct ucred *cred, int sticky, int cmsg, int uproto)
2589 int minmtupolicy, preftemp;
2592 if (!sticky && !cmsg) {
2594 printf("ip6_setpktopt: impossible case\n");
2600 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2601 * not be specified in the context of RFC3542. Conversely,
2602 * RFC3542 types should not be specified in the context of RFC2292.
2606 case IPV6_2292PKTINFO:
2607 case IPV6_2292HOPLIMIT:
2608 case IPV6_2292NEXTHOP:
2609 case IPV6_2292HOPOPTS:
2610 case IPV6_2292DSTOPTS:
2611 case IPV6_2292RTHDR:
2612 case IPV6_2292PKTOPTIONS:
2613 return (ENOPROTOOPT);
2616 if (sticky && cmsg) {
2623 case IPV6_RTHDRDSTOPTS:
2625 case IPV6_USE_MIN_MTU:
2628 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2629 return (ENOPROTOOPT);
2634 case IPV6_2292PKTINFO:
2637 struct ifnet *ifp = NULL;
2638 struct in6_pktinfo *pktinfo;
2640 if (len != sizeof(struct in6_pktinfo))
2643 pktinfo = (struct in6_pktinfo *)buf;
2646 * An application can clear any sticky IPV6_PKTINFO option by
2647 * doing a "regular" setsockopt with ipi6_addr being
2648 * in6addr_any and ipi6_ifindex being zero.
2649 * [RFC 3542, Section 6]
2651 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2652 pktinfo->ipi6_ifindex == 0 &&
2653 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2654 ip6_clearpktopts(opt, optname);
2658 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2659 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2663 /* validate the interface index if specified. */
2664 if (pktinfo->ipi6_ifindex > V_if_index ||
2665 pktinfo->ipi6_ifindex < 0) {
2668 if (pktinfo->ipi6_ifindex) {
2669 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2675 * We store the address anyway, and let in6_selectsrc()
2676 * validate the specified address. This is because ipi6_addr
2677 * may not have enough information about its scope zone, and
2678 * we may need additional information (such as outgoing
2679 * interface or the scope zone of a destination address) to
2680 * disambiguate the scope.
2681 * XXX: the delay of the validation may confuse the
2682 * application when it is used as a sticky option.
2684 if (opt->ip6po_pktinfo == NULL) {
2685 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2686 M_IP6OPT, M_NOWAIT);
2687 if (opt->ip6po_pktinfo == NULL)
2690 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2694 case IPV6_2292HOPLIMIT:
2700 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2701 * to simplify the ordering among hoplimit options.
2703 if (optname == IPV6_HOPLIMIT && sticky)
2704 return (ENOPROTOOPT);
2706 if (len != sizeof(int))
2709 if (*hlimp < -1 || *hlimp > 255)
2712 opt->ip6po_hlim = *hlimp;
2720 if (len != sizeof(int))
2722 tclass = *(int *)buf;
2723 if (tclass < -1 || tclass > 255)
2726 opt->ip6po_tclass = tclass;
2730 case IPV6_2292NEXTHOP:
2733 error = priv_check_cred(cred,
2734 PRIV_NETINET_SETHDROPTS, 0);
2739 if (len == 0) { /* just remove the option */
2740 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2744 /* check if cmsg_len is large enough for sa_len */
2745 if (len < sizeof(struct sockaddr) || len < *buf)
2748 switch (((struct sockaddr *)buf)->sa_family) {
2751 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2754 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2757 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2758 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2761 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2767 case AF_LINK: /* should eventually be supported */
2769 return (EAFNOSUPPORT);
2772 /* turn off the previous option, then set the new option. */
2773 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2774 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2775 if (opt->ip6po_nexthop == NULL)
2777 bcopy(buf, opt->ip6po_nexthop, *buf);
2780 case IPV6_2292HOPOPTS:
2783 struct ip6_hbh *hbh;
2787 * XXX: We don't allow a non-privileged user to set ANY HbH
2788 * options, since per-option restriction has too much
2792 error = priv_check_cred(cred,
2793 PRIV_NETINET_SETHDROPTS, 0);
2799 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2800 break; /* just remove the option */
2803 /* message length validation */
2804 if (len < sizeof(struct ip6_hbh))
2806 hbh = (struct ip6_hbh *)buf;
2807 hbhlen = (hbh->ip6h_len + 1) << 3;
2811 /* turn off the previous option, then set the new option. */
2812 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2813 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2814 if (opt->ip6po_hbh == NULL)
2816 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2821 case IPV6_2292DSTOPTS:
2823 case IPV6_RTHDRDSTOPTS:
2825 struct ip6_dest *dest, **newdest = NULL;
2828 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2829 error = priv_check_cred(cred,
2830 PRIV_NETINET_SETHDROPTS, 0);
2836 ip6_clearpktopts(opt, optname);
2837 break; /* just remove the option */
2840 /* message length validation */
2841 if (len < sizeof(struct ip6_dest))
2843 dest = (struct ip6_dest *)buf;
2844 destlen = (dest->ip6d_len + 1) << 3;
2849 * Determine the position that the destination options header
2850 * should be inserted; before or after the routing header.
2853 case IPV6_2292DSTOPTS:
2855 * The old advacned API is ambiguous on this point.
2856 * Our approach is to determine the position based
2857 * according to the existence of a routing header.
2858 * Note, however, that this depends on the order of the
2859 * extension headers in the ancillary data; the 1st
2860 * part of the destination options header must appear
2861 * before the routing header in the ancillary data,
2863 * RFC3542 solved the ambiguity by introducing
2864 * separate ancillary data or option types.
2866 if (opt->ip6po_rthdr == NULL)
2867 newdest = &opt->ip6po_dest1;
2869 newdest = &opt->ip6po_dest2;
2871 case IPV6_RTHDRDSTOPTS:
2872 newdest = &opt->ip6po_dest1;
2875 newdest = &opt->ip6po_dest2;
2879 /* turn off the previous option, then set the new option. */
2880 ip6_clearpktopts(opt, optname);
2881 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2882 if (*newdest == NULL)
2884 bcopy(dest, *newdest, destlen);
2889 case IPV6_2292RTHDR:
2892 struct ip6_rthdr *rth;
2896 ip6_clearpktopts(opt, IPV6_RTHDR);
2897 break; /* just remove the option */
2900 /* message length validation */
2901 if (len < sizeof(struct ip6_rthdr))
2903 rth = (struct ip6_rthdr *)buf;
2904 rthlen = (rth->ip6r_len + 1) << 3;
2908 switch (rth->ip6r_type) {
2909 case IPV6_RTHDR_TYPE_0:
2910 if (rth->ip6r_len == 0) /* must contain one addr */
2912 if (rth->ip6r_len % 2) /* length must be even */
2914 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2918 return (EINVAL); /* not supported */
2921 /* turn off the previous option */
2922 ip6_clearpktopts(opt, IPV6_RTHDR);
2923 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2924 if (opt->ip6po_rthdr == NULL)
2926 bcopy(rth, opt->ip6po_rthdr, rthlen);
2931 case IPV6_USE_MIN_MTU:
2932 if (len != sizeof(int))
2934 minmtupolicy = *(int *)buf;
2935 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2936 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2937 minmtupolicy != IP6PO_MINMTU_ALL) {
2940 opt->ip6po_minmtu = minmtupolicy;
2944 if (len != sizeof(int))
2947 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2949 * we ignore this option for TCP sockets.
2950 * (RFC3542 leaves this case unspecified.)
2952 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2954 opt->ip6po_flags |= IP6PO_DONTFRAG;
2957 case IPV6_PREFER_TEMPADDR:
2958 if (len != sizeof(int))
2960 preftemp = *(int *)buf;
2961 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2962 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2963 preftemp != IP6PO_TEMPADDR_PREFER) {
2966 opt->ip6po_prefer_tempaddr = preftemp;
2970 return (ENOPROTOOPT);
2971 } /* end of switch */
2977 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2978 * packet to the input queue of a specified interface. Note that this
2979 * calls the output routine of the loopback "driver", but with an interface
2980 * pointer that might NOT be &loif -- easier than replicating that code here.
2983 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2986 struct ip6_hdr *ip6;
2988 copym = m_copy(m, 0, M_COPYALL);
2993 * Make sure to deep-copy IPv6 header portion in case the data
2994 * is in an mbuf cluster, so that we can safely override the IPv6
2995 * header portion later.
2997 if ((copym->m_flags & M_EXT) != 0 ||
2998 copym->m_len < sizeof(struct ip6_hdr)) {
2999 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3005 if (copym->m_len < sizeof(*ip6)) {
3011 ip6 = mtod(copym, struct ip6_hdr *);
3013 * clear embedded scope identifiers if necessary.
3014 * in6_clearscope will touch the addresses only when necessary.
3016 in6_clearscope(&ip6->ip6_src);
3017 in6_clearscope(&ip6->ip6_dst);
3019 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3023 * Chop IPv6 header off from the payload.
3026 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3029 struct ip6_hdr *ip6;
3031 ip6 = mtod(m, struct ip6_hdr *);
3032 if (m->m_len > sizeof(*ip6)) {
3033 mh = m_gethdr(M_NOWAIT, MT_DATA);
3038 m_move_pkthdr(mh, m);
3039 MH_ALIGN(mh, sizeof(*ip6));
3040 m->m_len -= sizeof(*ip6);
3041 m->m_data += sizeof(*ip6);
3044 m->m_len = sizeof(*ip6);
3045 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3047 exthdrs->ip6e_ip6 = m;
3052 * Compute IPv6 extension header length.
3055 ip6_optlen(struct inpcb *in6p)
3059 if (!in6p->in6p_outputopts)
3064 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3066 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3067 if (in6p->in6p_outputopts->ip6po_rthdr)
3068 /* dest1 is valid with rthdr only */
3069 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3070 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3071 len += elen(in6p->in6p_outputopts->ip6po_dest2);