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 IP6STAT_INC(ip6s_badscope);
504 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
506 IP6STAT_INC(ip6s_badscope);
510 IP6STAT_INC(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 IP6STAT_INC(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 IP6STAT_INC(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 is not always the case, since
778 * some versions of MGETHDR() does not
779 * initialize the field.
781 m->m_pkthdr.rcvif = NULL;
782 if (ip6_mforward(ip6, ifp, m) != 0) {
789 * Multicasts with a hoplimit of zero may be looped back,
790 * above, but must not be transmitted on a network.
791 * Also, multicasts addressed to the loopback interface
792 * are not sent -- the above call to ip6_mloopback() will
793 * loop back a copy if this host actually belongs to the
794 * destination group on the loopback interface.
796 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
797 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
804 * Fill the outgoing inteface to tell the upper layer
805 * to increment per-interface statistics.
810 /* Determine path MTU. */
811 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
812 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
816 * The caller of this function may specify to use the minimum MTU
818 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
819 * setting. The logic is a bit complicated; by default, unicast
820 * packets will follow path MTU while multicast packets will be sent at
821 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
822 * including unicast ones will be sent at the minimum MTU. Multicast
823 * packets will always be sent at the minimum MTU unless
824 * IP6PO_MINMTU_DISABLE is explicitly specified.
825 * See RFC 3542 for more details.
827 if (mtu > IPV6_MMTU) {
828 if ((flags & IPV6_MINMTU))
830 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
832 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
834 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
840 * clear embedded scope identifiers if necessary.
841 * in6_clearscope will touch the addresses only when necessary.
843 in6_clearscope(&ip6->ip6_src);
844 in6_clearscope(&ip6->ip6_dst);
847 * If the outgoing packet contains a hop-by-hop options header,
848 * it must be examined and processed even by the source node.
849 * (RFC 2460, section 4.)
851 if (exthdrs.ip6e_hbh) {
852 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
853 u_int32_t dummy; /* XXX unused */
854 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
857 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
858 panic("ip6e_hbh is not contiguous");
861 * XXX: if we have to send an ICMPv6 error to the sender,
862 * we need the M_LOOP flag since icmp6_error() expects
863 * the IPv6 and the hop-by-hop options header are
864 * contiguous unless the flag is set.
866 m->m_flags |= M_LOOP;
867 m->m_pkthdr.rcvif = ifp;
868 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
869 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
870 &dummy, &plen) < 0) {
871 /* m was already freed at this point */
872 error = EINVAL;/* better error? */
875 m->m_flags &= ~M_LOOP; /* XXX */
876 m->m_pkthdr.rcvif = NULL;
879 /* Jump over all PFIL processing if hooks are not active. */
880 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
884 /* Run through list of hooks for output packets. */
885 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
886 if (error != 0 || m == NULL)
888 ip6 = mtod(m, struct ip6_hdr *);
890 /* See if destination IP address was changed by packet filter. */
891 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
892 m->m_flags |= M_SKIP_FIREWALL;
893 /* If destination is now ourself drop to ip6_input(). */
894 if (in6_localip(&ip6->ip6_dst)) {
895 m->m_flags |= M_FASTFWD_OURS;
896 if (m->m_pkthdr.rcvif == NULL)
897 m->m_pkthdr.rcvif = V_loif;
898 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
899 m->m_pkthdr.csum_flags |=
900 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
901 m->m_pkthdr.csum_data = 0xffff;
904 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
905 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
907 error = netisr_queue(NETISR_IPV6, m);
910 goto again; /* Redo the routing table lookup. */
913 /* See if local, if yes, send it to netisr. */
914 if (m->m_flags & M_FASTFWD_OURS) {
915 if (m->m_pkthdr.rcvif == NULL)
916 m->m_pkthdr.rcvif = V_loif;
917 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
918 m->m_pkthdr.csum_flags |=
919 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
920 m->m_pkthdr.csum_data = 0xffff;
923 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
924 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
926 error = netisr_queue(NETISR_IPV6, m);
929 /* Or forward to some other address? */
930 if ((m->m_flags & M_IP6_NEXTHOP) &&
931 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
932 dst = (struct sockaddr_in6 *)&ro->ro_dst;
933 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
934 m->m_flags |= M_SKIP_FIREWALL;
935 m->m_flags &= ~M_IP6_NEXTHOP;
936 m_tag_delete(m, fwd_tag);
942 * Send the packet to the outgoing interface.
943 * If necessary, do IPv6 fragmentation before sending.
945 * the logic here is rather complex:
946 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
947 * 1-a: send as is if tlen <= path mtu
948 * 1-b: fragment if tlen > path mtu
950 * 2: if user asks us not to fragment (dontfrag == 1)
951 * 2-a: send as is if tlen <= interface mtu
952 * 2-b: error if tlen > interface mtu
954 * 3: if we always need to attach fragment header (alwaysfrag == 1)
957 * 4: if dontfrag == 1 && alwaysfrag == 1
958 * error, as we cannot handle this conflicting request
960 sw_csum = m->m_pkthdr.csum_flags;
962 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
963 sw_csum &= ~ifp->if_hwassist;
967 * If we added extension headers, we will not do TSO and calculate the
968 * checksums ourselves for now.
969 * XXX-BZ Need a framework to know when the NIC can handle it, even
972 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
973 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
974 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
977 if (sw_csum & CSUM_SCTP_IPV6) {
978 sw_csum &= ~CSUM_SCTP_IPV6;
979 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
982 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
983 tlen = m->m_pkthdr.len;
985 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
989 if (dontfrag && alwaysfrag) { /* case 4 */
990 /* conflicting request - can't transmit */
994 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
996 * Even if the DONTFRAG option is specified, we cannot send the
997 * packet when the data length is larger than the MTU of the
998 * outgoing interface.
999 * Notify the error by sending IPV6_PATHMTU ancillary data as
1000 * well as returning an error code (the latter is not described
1004 struct ip6ctlparam ip6cp;
1006 mtu32 = (u_int32_t)mtu;
1007 bzero(&ip6cp, sizeof(ip6cp));
1008 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1009 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1017 * transmit packet without fragmentation
1019 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1020 struct in6_ifaddr *ia6;
1022 ip6 = mtod(m, struct ip6_hdr *);
1023 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1025 /* Record statistics for this interface address. */
1026 ia6->ia_ifa.if_opackets++;
1027 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1028 ifa_free(&ia6->ia_ifa);
1030 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1035 * try to fragment the packet. case 1-b and 3
1037 if (mtu < IPV6_MMTU) {
1038 /* path MTU cannot be less than IPV6_MMTU */
1040 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1042 } else if (ip6->ip6_plen == 0) {
1043 /* jumbo payload cannot be fragmented */
1045 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1048 struct mbuf **mnext, *m_frgpart;
1049 struct ip6_frag *ip6f;
1050 u_int32_t id = htonl(ip6_randomid());
1053 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1056 * Too large for the destination or interface;
1057 * fragment if possible.
1058 * Must be able to put at least 8 bytes per fragment.
1060 hlen = unfragpartlen;
1061 if (mtu > IPV6_MAXPACKET)
1062 mtu = IPV6_MAXPACKET;
1064 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1067 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1072 * Verify that we have any chance at all of being able to queue
1073 * the packet or packet fragments
1075 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1076 < tlen /* - hlen */)) {
1078 IP6STAT_INC(ip6s_odropped);
1084 * If the interface will not calculate checksums on
1085 * fragmented packets, then do it here.
1086 * XXX-BZ handle the hw offloading case. Need flags.
1088 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1089 in6_delayed_cksum(m, plen, hlen);
1090 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1093 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1094 sctp_delayed_cksum(m, hlen);
1095 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1098 mnext = &m->m_nextpkt;
1101 * Change the next header field of the last header in the
1102 * unfragmentable part.
1104 if (exthdrs.ip6e_rthdr) {
1105 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1106 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1107 } else if (exthdrs.ip6e_dest1) {
1108 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1109 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1110 } else if (exthdrs.ip6e_hbh) {
1111 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1112 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1114 nextproto = ip6->ip6_nxt;
1115 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1119 * Loop through length of segment after first fragment,
1120 * make new header and copy data of each part and link onto
1124 for (off = hlen; off < tlen; off += len) {
1125 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1128 IP6STAT_INC(ip6s_odropped);
1131 m->m_pkthdr.rcvif = NULL;
1132 m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1134 mnext = &m->m_nextpkt;
1135 m->m_data += max_linkhdr;
1136 mhip6 = mtod(m, struct ip6_hdr *);
1138 m->m_len = sizeof(*mhip6);
1139 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1141 IP6STAT_INC(ip6s_odropped);
1144 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1145 if (off + len >= tlen)
1148 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1149 mhip6->ip6_plen = htons((u_short)(len + hlen +
1150 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1151 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1153 IP6STAT_INC(ip6s_odropped);
1156 m_cat(m, m_frgpart);
1157 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1158 m->m_pkthdr.rcvif = NULL;
1159 ip6f->ip6f_reserved = 0;
1160 ip6f->ip6f_ident = id;
1161 ip6f->ip6f_nxt = nextproto;
1162 IP6STAT_INC(ip6s_ofragments);
1163 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1166 in6_ifstat_inc(ifp, ifs6_out_fragok);
1170 * Remove leading garbages.
1176 for (m0 = m; m; m = m0) {
1180 /* Record statistics for this interface address. */
1182 ia->ia_ifa.if_opackets++;
1183 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1185 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1191 IP6STAT_INC(ip6s_fragmented);
1194 if (ro == &ip6route)
1196 if (ro_pmtu == &ip6route)
1206 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1207 m_freem(exthdrs.ip6e_dest1);
1208 m_freem(exthdrs.ip6e_rthdr);
1209 m_freem(exthdrs.ip6e_dest2);
1218 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1222 if (hlen > MCLBYTES)
1223 return (ENOBUFS); /* XXX */
1225 MGET(m, M_DONTWAIT, MT_DATA);
1230 MCLGET(m, M_DONTWAIT);
1231 if ((m->m_flags & M_EXT) == 0) {
1238 bcopy(hdr, mtod(m, caddr_t), hlen);
1245 * Insert jumbo payload option.
1248 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1254 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1257 * If there is no hop-by-hop options header, allocate new one.
1258 * If there is one but it doesn't have enough space to store the
1259 * jumbo payload option, allocate a cluster to store the whole options.
1260 * Otherwise, use it to store the options.
1262 if (exthdrs->ip6e_hbh == 0) {
1263 MGET(mopt, M_DONTWAIT, MT_DATA);
1266 mopt->m_len = JUMBOOPTLEN;
1267 optbuf = mtod(mopt, u_char *);
1268 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1269 exthdrs->ip6e_hbh = mopt;
1271 struct ip6_hbh *hbh;
1273 mopt = exthdrs->ip6e_hbh;
1274 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1277 * - exthdrs->ip6e_hbh is not referenced from places
1278 * other than exthdrs.
1279 * - exthdrs->ip6e_hbh is not an mbuf chain.
1281 int oldoptlen = mopt->m_len;
1285 * XXX: give up if the whole (new) hbh header does
1286 * not fit even in an mbuf cluster.
1288 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1292 * As a consequence, we must always prepare a cluster
1295 MGET(n, M_DONTWAIT, MT_DATA);
1297 MCLGET(n, M_DONTWAIT);
1298 if ((n->m_flags & M_EXT) == 0) {
1305 n->m_len = oldoptlen + JUMBOOPTLEN;
1306 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1308 optbuf = mtod(n, caddr_t) + oldoptlen;
1310 mopt = exthdrs->ip6e_hbh = n;
1312 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1313 mopt->m_len += JUMBOOPTLEN;
1315 optbuf[0] = IP6OPT_PADN;
1319 * Adjust the header length according to the pad and
1320 * the jumbo payload option.
1322 hbh = mtod(mopt, struct ip6_hbh *);
1323 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1326 /* fill in the option. */
1327 optbuf[2] = IP6OPT_JUMBO;
1329 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1330 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1332 /* finally, adjust the packet header length */
1333 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1340 * Insert fragment header and copy unfragmentable header portions.
1343 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1344 struct ip6_frag **frghdrp)
1346 struct mbuf *n, *mlast;
1348 if (hlen > sizeof(struct ip6_hdr)) {
1349 n = m_copym(m0, sizeof(struct ip6_hdr),
1350 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1357 /* Search for the last mbuf of unfragmentable part. */
1358 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1361 if ((mlast->m_flags & M_EXT) == 0 &&
1362 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1363 /* use the trailing space of the last mbuf for the fragment hdr */
1364 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1366 mlast->m_len += sizeof(struct ip6_frag);
1367 m->m_pkthdr.len += sizeof(struct ip6_frag);
1369 /* allocate a new mbuf for the fragment header */
1372 MGET(mfrg, M_DONTWAIT, MT_DATA);
1375 mfrg->m_len = sizeof(struct ip6_frag);
1376 *frghdrp = mtod(mfrg, struct ip6_frag *);
1377 mlast->m_next = mfrg;
1384 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1385 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1386 int *alwaysfragp, u_int fibnum)
1392 if (ro_pmtu != ro) {
1393 /* The first hop and the final destination may differ. */
1394 struct sockaddr_in6 *sa6_dst =
1395 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1396 if (ro_pmtu->ro_rt &&
1397 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1398 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1399 RTFREE(ro_pmtu->ro_rt);
1400 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1402 if (ro_pmtu->ro_rt == NULL) {
1403 bzero(sa6_dst, sizeof(*sa6_dst));
1404 sa6_dst->sin6_family = AF_INET6;
1405 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1406 sa6_dst->sin6_addr = *dst;
1408 in6_rtalloc(ro_pmtu, fibnum);
1411 if (ro_pmtu->ro_rt) {
1413 struct in_conninfo inc;
1415 bzero(&inc, sizeof(inc));
1416 inc.inc_flags |= INC_ISIPV6;
1417 inc.inc6_faddr = *dst;
1420 ifp = ro_pmtu->ro_rt->rt_ifp;
1421 ifmtu = IN6_LINKMTU(ifp);
1422 mtu = tcp_hc_getmtu(&inc);
1424 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1426 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1429 else if (mtu < IPV6_MMTU) {
1431 * RFC2460 section 5, last paragraph:
1432 * if we record ICMPv6 too big message with
1433 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1434 * or smaller, with framgent header attached.
1435 * (fragment header is needed regardless from the
1436 * packet size, for translators to identify packets)
1440 } else if (mtu > ifmtu) {
1442 * The MTU on the route is larger than the MTU on
1443 * the interface! This shouldn't happen, unless the
1444 * MTU of the interface has been changed after the
1445 * interface was brought up. Change the MTU in the
1446 * route to match the interface MTU (as long as the
1447 * field isn't locked).
1450 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1453 mtu = IN6_LINKMTU(ifp);
1455 error = EHOSTUNREACH; /* XXX */
1459 *alwaysfragp = alwaysfrag;
1464 * IP6 socket option processing.
1467 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1469 int optdatalen, uproto;
1471 struct inpcb *in6p = sotoinpcb(so);
1473 int level, op, optname;
1477 level = sopt->sopt_level;
1478 op = sopt->sopt_dir;
1479 optname = sopt->sopt_name;
1480 optlen = sopt->sopt_valsize;
1484 uproto = (int)so->so_proto->pr_protocol;
1486 if (level != IPPROTO_IPV6) {
1489 if (sopt->sopt_level == SOL_SOCKET &&
1490 sopt->sopt_dir == SOPT_SET) {
1491 switch (sopt->sopt_name) {
1494 if ((so->so_options & SO_REUSEADDR) != 0)
1495 in6p->inp_flags2 |= INP_REUSEADDR;
1497 in6p->inp_flags2 &= ~INP_REUSEADDR;
1503 if ((so->so_options & SO_REUSEPORT) != 0)
1504 in6p->inp_flags2 |= INP_REUSEPORT;
1506 in6p->inp_flags2 &= ~INP_REUSEPORT;
1512 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1520 } else { /* level == IPPROTO_IPV6 */
1525 case IPV6_2292PKTOPTIONS:
1526 #ifdef IPV6_PKTOPTIONS
1527 case IPV6_PKTOPTIONS:
1532 error = soopt_getm(sopt, &m); /* XXX */
1535 error = soopt_mcopyin(sopt, m); /* XXX */
1538 error = ip6_pcbopts(&in6p->in6p_outputopts,
1540 m_freem(m); /* XXX */
1545 * Use of some Hop-by-Hop options or some
1546 * Destination options, might require special
1547 * privilege. That is, normal applications
1548 * (without special privilege) might be forbidden
1549 * from setting certain options in outgoing packets,
1550 * and might never see certain options in received
1551 * packets. [RFC 2292 Section 6]
1552 * KAME specific note:
1553 * KAME prevents non-privileged users from sending or
1554 * receiving ANY hbh/dst options in order to avoid
1555 * overhead of parsing options in the kernel.
1557 case IPV6_RECVHOPOPTS:
1558 case IPV6_RECVDSTOPTS:
1559 case IPV6_RECVRTHDRDSTOPTS:
1561 error = priv_check(td,
1562 PRIV_NETINET_SETHDROPTS);
1567 case IPV6_UNICAST_HOPS:
1571 case IPV6_RECVPKTINFO:
1572 case IPV6_RECVHOPLIMIT:
1573 case IPV6_RECVRTHDR:
1574 case IPV6_RECVPATHMTU:
1575 case IPV6_RECVTCLASS:
1577 case IPV6_AUTOFLOWLABEL:
1579 if (optname == IPV6_BINDANY && td != NULL) {
1580 error = priv_check(td,
1581 PRIV_NETINET_BINDANY);
1586 if (optlen != sizeof(int)) {
1590 error = sooptcopyin(sopt, &optval,
1591 sizeof optval, sizeof optval);
1596 case IPV6_UNICAST_HOPS:
1597 if (optval < -1 || optval >= 256)
1600 /* -1 = kernel default */
1601 in6p->in6p_hops = optval;
1602 if ((in6p->inp_vflag &
1604 in6p->inp_ip_ttl = optval;
1607 #define OPTSET(bit) \
1611 in6p->inp_flags |= (bit); \
1613 in6p->inp_flags &= ~(bit); \
1614 INP_WUNLOCK(in6p); \
1615 } while (/*CONSTCOND*/ 0)
1616 #define OPTSET2292(bit) \
1619 in6p->inp_flags |= IN6P_RFC2292; \
1621 in6p->inp_flags |= (bit); \
1623 in6p->inp_flags &= ~(bit); \
1624 INP_WUNLOCK(in6p); \
1625 } while (/*CONSTCOND*/ 0)
1626 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1628 case IPV6_RECVPKTINFO:
1629 /* cannot mix with RFC2292 */
1630 if (OPTBIT(IN6P_RFC2292)) {
1634 OPTSET(IN6P_PKTINFO);
1639 struct ip6_pktopts **optp;
1641 /* cannot mix with RFC2292 */
1642 if (OPTBIT(IN6P_RFC2292)) {
1646 optp = &in6p->in6p_outputopts;
1647 error = ip6_pcbopt(IPV6_HOPLIMIT,
1648 (u_char *)&optval, sizeof(optval),
1649 optp, (td != NULL) ? td->td_ucred :
1654 case IPV6_RECVHOPLIMIT:
1655 /* cannot mix with RFC2292 */
1656 if (OPTBIT(IN6P_RFC2292)) {
1660 OPTSET(IN6P_HOPLIMIT);
1663 case IPV6_RECVHOPOPTS:
1664 /* cannot mix with RFC2292 */
1665 if (OPTBIT(IN6P_RFC2292)) {
1669 OPTSET(IN6P_HOPOPTS);
1672 case IPV6_RECVDSTOPTS:
1673 /* cannot mix with RFC2292 */
1674 if (OPTBIT(IN6P_RFC2292)) {
1678 OPTSET(IN6P_DSTOPTS);
1681 case IPV6_RECVRTHDRDSTOPTS:
1682 /* cannot mix with RFC2292 */
1683 if (OPTBIT(IN6P_RFC2292)) {
1687 OPTSET(IN6P_RTHDRDSTOPTS);
1690 case IPV6_RECVRTHDR:
1691 /* cannot mix with RFC2292 */
1692 if (OPTBIT(IN6P_RFC2292)) {
1703 case IPV6_RECVPATHMTU:
1705 * We ignore this option for TCP
1707 * (RFC3542 leaves this case
1710 if (uproto != IPPROTO_TCP)
1716 * make setsockopt(IPV6_V6ONLY)
1717 * available only prior to bind(2).
1718 * see ipng mailing list, Jun 22 2001.
1720 if (in6p->inp_lport ||
1721 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1725 OPTSET(IN6P_IPV6_V6ONLY);
1727 in6p->inp_vflag &= ~INP_IPV4;
1729 in6p->inp_vflag |= INP_IPV4;
1731 case IPV6_RECVTCLASS:
1732 /* cannot mix with RFC2292 XXX */
1733 if (OPTBIT(IN6P_RFC2292)) {
1737 OPTSET(IN6P_TCLASS);
1739 case IPV6_AUTOFLOWLABEL:
1740 OPTSET(IN6P_AUTOFLOWLABEL);
1744 OPTSET(INP_BINDANY);
1751 case IPV6_USE_MIN_MTU:
1752 case IPV6_PREFER_TEMPADDR:
1753 if (optlen != sizeof(optval)) {
1757 error = sooptcopyin(sopt, &optval,
1758 sizeof optval, sizeof optval);
1762 struct ip6_pktopts **optp;
1763 optp = &in6p->in6p_outputopts;
1764 error = ip6_pcbopt(optname,
1765 (u_char *)&optval, sizeof(optval),
1766 optp, (td != NULL) ? td->td_ucred :
1771 case IPV6_2292PKTINFO:
1772 case IPV6_2292HOPLIMIT:
1773 case IPV6_2292HOPOPTS:
1774 case IPV6_2292DSTOPTS:
1775 case IPV6_2292RTHDR:
1777 if (optlen != sizeof(int)) {
1781 error = sooptcopyin(sopt, &optval,
1782 sizeof optval, sizeof optval);
1786 case IPV6_2292PKTINFO:
1787 OPTSET2292(IN6P_PKTINFO);
1789 case IPV6_2292HOPLIMIT:
1790 OPTSET2292(IN6P_HOPLIMIT);
1792 case IPV6_2292HOPOPTS:
1794 * Check super-user privilege.
1795 * See comments for IPV6_RECVHOPOPTS.
1798 error = priv_check(td,
1799 PRIV_NETINET_SETHDROPTS);
1803 OPTSET2292(IN6P_HOPOPTS);
1805 case IPV6_2292DSTOPTS:
1807 error = priv_check(td,
1808 PRIV_NETINET_SETHDROPTS);
1812 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1814 case IPV6_2292RTHDR:
1815 OPTSET2292(IN6P_RTHDR);
1823 case IPV6_RTHDRDSTOPTS:
1826 /* new advanced API (RFC3542) */
1828 u_char optbuf_storage[MCLBYTES];
1830 struct ip6_pktopts **optp;
1832 /* cannot mix with RFC2292 */
1833 if (OPTBIT(IN6P_RFC2292)) {
1839 * We only ensure valsize is not too large
1840 * here. Further validation will be done
1843 error = sooptcopyin(sopt, optbuf_storage,
1844 sizeof(optbuf_storage), 0);
1847 optlen = sopt->sopt_valsize;
1848 optbuf = optbuf_storage;
1849 optp = &in6p->in6p_outputopts;
1850 error = ip6_pcbopt(optname, optbuf, optlen,
1851 optp, (td != NULL) ? td->td_ucred : NULL,
1857 case IPV6_MULTICAST_IF:
1858 case IPV6_MULTICAST_HOPS:
1859 case IPV6_MULTICAST_LOOP:
1860 case IPV6_JOIN_GROUP:
1861 case IPV6_LEAVE_GROUP:
1863 case MCAST_BLOCK_SOURCE:
1864 case MCAST_UNBLOCK_SOURCE:
1865 case MCAST_JOIN_GROUP:
1866 case MCAST_LEAVE_GROUP:
1867 case MCAST_JOIN_SOURCE_GROUP:
1868 case MCAST_LEAVE_SOURCE_GROUP:
1869 error = ip6_setmoptions(in6p, sopt);
1872 case IPV6_PORTRANGE:
1873 error = sooptcopyin(sopt, &optval,
1874 sizeof optval, sizeof optval);
1880 case IPV6_PORTRANGE_DEFAULT:
1881 in6p->inp_flags &= ~(INP_LOWPORT);
1882 in6p->inp_flags &= ~(INP_HIGHPORT);
1885 case IPV6_PORTRANGE_HIGH:
1886 in6p->inp_flags &= ~(INP_LOWPORT);
1887 in6p->inp_flags |= INP_HIGHPORT;
1890 case IPV6_PORTRANGE_LOW:
1891 in6p->inp_flags &= ~(INP_HIGHPORT);
1892 in6p->inp_flags |= INP_LOWPORT;
1903 case IPV6_IPSEC_POLICY:
1908 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1910 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1912 req = mtod(m, caddr_t);
1913 error = ipsec_set_policy(in6p, optname, req,
1914 m->m_len, (sopt->sopt_td != NULL) ?
1915 sopt->sopt_td->td_ucred : NULL);
1922 error = ENOPROTOOPT;
1930 case IPV6_2292PKTOPTIONS:
1931 #ifdef IPV6_PKTOPTIONS
1932 case IPV6_PKTOPTIONS:
1935 * RFC3542 (effectively) deprecated the
1936 * semantics of the 2292-style pktoptions.
1937 * Since it was not reliable in nature (i.e.,
1938 * applications had to expect the lack of some
1939 * information after all), it would make sense
1940 * to simplify this part by always returning
1943 sopt->sopt_valsize = 0;
1946 case IPV6_RECVHOPOPTS:
1947 case IPV6_RECVDSTOPTS:
1948 case IPV6_RECVRTHDRDSTOPTS:
1949 case IPV6_UNICAST_HOPS:
1950 case IPV6_RECVPKTINFO:
1951 case IPV6_RECVHOPLIMIT:
1952 case IPV6_RECVRTHDR:
1953 case IPV6_RECVPATHMTU:
1957 case IPV6_PORTRANGE:
1958 case IPV6_RECVTCLASS:
1959 case IPV6_AUTOFLOWLABEL:
1963 case IPV6_RECVHOPOPTS:
1964 optval = OPTBIT(IN6P_HOPOPTS);
1967 case IPV6_RECVDSTOPTS:
1968 optval = OPTBIT(IN6P_DSTOPTS);
1971 case IPV6_RECVRTHDRDSTOPTS:
1972 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1975 case IPV6_UNICAST_HOPS:
1976 optval = in6p->in6p_hops;
1979 case IPV6_RECVPKTINFO:
1980 optval = OPTBIT(IN6P_PKTINFO);
1983 case IPV6_RECVHOPLIMIT:
1984 optval = OPTBIT(IN6P_HOPLIMIT);
1987 case IPV6_RECVRTHDR:
1988 optval = OPTBIT(IN6P_RTHDR);
1991 case IPV6_RECVPATHMTU:
1992 optval = OPTBIT(IN6P_MTU);
1996 optval = OPTBIT(INP_FAITH);
2000 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2003 case IPV6_PORTRANGE:
2006 flags = in6p->inp_flags;
2007 if (flags & INP_HIGHPORT)
2008 optval = IPV6_PORTRANGE_HIGH;
2009 else if (flags & INP_LOWPORT)
2010 optval = IPV6_PORTRANGE_LOW;
2015 case IPV6_RECVTCLASS:
2016 optval = OPTBIT(IN6P_TCLASS);
2019 case IPV6_AUTOFLOWLABEL:
2020 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2024 optval = OPTBIT(INP_BINDANY);
2029 error = sooptcopyout(sopt, &optval,
2036 struct ip6_mtuinfo mtuinfo;
2037 struct route_in6 sro;
2039 bzero(&sro, sizeof(sro));
2041 if (!(so->so_state & SS_ISCONNECTED))
2044 * XXX: we dot not consider the case of source
2045 * routing, or optional information to specify
2046 * the outgoing interface.
2048 error = ip6_getpmtu(&sro, NULL, NULL,
2049 &in6p->in6p_faddr, &pmtu, NULL,
2055 if (pmtu > IPV6_MAXPACKET)
2056 pmtu = IPV6_MAXPACKET;
2058 bzero(&mtuinfo, sizeof(mtuinfo));
2059 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2060 optdata = (void *)&mtuinfo;
2061 optdatalen = sizeof(mtuinfo);
2062 error = sooptcopyout(sopt, optdata,
2067 case IPV6_2292PKTINFO:
2068 case IPV6_2292HOPLIMIT:
2069 case IPV6_2292HOPOPTS:
2070 case IPV6_2292RTHDR:
2071 case IPV6_2292DSTOPTS:
2073 case IPV6_2292PKTINFO:
2074 optval = OPTBIT(IN6P_PKTINFO);
2076 case IPV6_2292HOPLIMIT:
2077 optval = OPTBIT(IN6P_HOPLIMIT);
2079 case IPV6_2292HOPOPTS:
2080 optval = OPTBIT(IN6P_HOPOPTS);
2082 case IPV6_2292RTHDR:
2083 optval = OPTBIT(IN6P_RTHDR);
2085 case IPV6_2292DSTOPTS:
2086 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2089 error = sooptcopyout(sopt, &optval,
2096 case IPV6_RTHDRDSTOPTS:
2100 case IPV6_USE_MIN_MTU:
2101 case IPV6_PREFER_TEMPADDR:
2102 error = ip6_getpcbopt(in6p->in6p_outputopts,
2106 case IPV6_MULTICAST_IF:
2107 case IPV6_MULTICAST_HOPS:
2108 case IPV6_MULTICAST_LOOP:
2110 error = ip6_getmoptions(in6p, sopt);
2114 case IPV6_IPSEC_POLICY:
2118 struct mbuf *m = NULL;
2119 struct mbuf **mp = &m;
2120 size_t ovalsize = sopt->sopt_valsize;
2121 caddr_t oval = (caddr_t)sopt->sopt_val;
2123 error = soopt_getm(sopt, &m); /* XXX */
2126 error = soopt_mcopyin(sopt, m); /* XXX */
2129 sopt->sopt_valsize = ovalsize;
2130 sopt->sopt_val = oval;
2132 req = mtod(m, caddr_t);
2135 error = ipsec_get_policy(in6p, req, len, mp);
2137 error = soopt_mcopyout(sopt, m); /* XXX */
2138 if (error == 0 && m)
2145 error = ENOPROTOOPT;
2155 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2157 int error = 0, optval, optlen;
2158 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2159 struct inpcb *in6p = sotoinpcb(so);
2160 int level, op, optname;
2162 level = sopt->sopt_level;
2163 op = sopt->sopt_dir;
2164 optname = sopt->sopt_name;
2165 optlen = sopt->sopt_valsize;
2167 if (level != IPPROTO_IPV6) {
2174 * For ICMPv6 sockets, no modification allowed for checksum
2175 * offset, permit "no change" values to help existing apps.
2177 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2178 * for an ICMPv6 socket will fail."
2179 * The current behavior does not meet RFC3542.
2183 if (optlen != sizeof(int)) {
2187 error = sooptcopyin(sopt, &optval, sizeof(optval),
2191 if ((optval % 2) != 0) {
2192 /* the API assumes even offset values */
2194 } else if (so->so_proto->pr_protocol ==
2196 if (optval != icmp6off)
2199 in6p->in6p_cksum = optval;
2203 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2206 optval = in6p->in6p_cksum;
2208 error = sooptcopyout(sopt, &optval, sizeof(optval));
2218 error = ENOPROTOOPT;
2226 * Set up IP6 options in pcb for insertion in output packets or
2227 * specifying behavior of outgoing packets.
2230 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2231 struct socket *so, struct sockopt *sopt)
2233 struct ip6_pktopts *opt = *pktopt;
2235 struct thread *td = sopt->sopt_td;
2237 /* turn off any old options. */
2240 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2241 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2242 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2243 printf("ip6_pcbopts: all specified options are cleared.\n");
2245 ip6_clearpktopts(opt, -1);
2247 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2250 if (!m || m->m_len == 0) {
2252 * Only turning off any previous options, regardless of
2253 * whether the opt is just created or given.
2255 free(opt, M_IP6OPT);
2259 /* set options specified by user. */
2260 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2261 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2262 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2263 free(opt, M_IP6OPT);
2271 * initialize ip6_pktopts. beware that there are non-zero default values in
2275 ip6_initpktopts(struct ip6_pktopts *opt)
2278 bzero(opt, sizeof(*opt));
2279 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2280 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2281 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2282 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2286 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2287 struct ucred *cred, int uproto)
2289 struct ip6_pktopts *opt;
2291 if (*pktopt == NULL) {
2292 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2294 ip6_initpktopts(*pktopt);
2298 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2302 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2304 void *optdata = NULL;
2306 struct ip6_ext *ip6e;
2308 struct in6_pktinfo null_pktinfo;
2309 int deftclass = 0, on;
2310 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2311 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2315 if (pktopt && pktopt->ip6po_pktinfo)
2316 optdata = (void *)pktopt->ip6po_pktinfo;
2318 /* XXX: we don't have to do this every time... */
2319 bzero(&null_pktinfo, sizeof(null_pktinfo));
2320 optdata = (void *)&null_pktinfo;
2322 optdatalen = sizeof(struct in6_pktinfo);
2325 if (pktopt && pktopt->ip6po_tclass >= 0)
2326 optdata = (void *)&pktopt->ip6po_tclass;
2328 optdata = (void *)&deftclass;
2329 optdatalen = sizeof(int);
2332 if (pktopt && pktopt->ip6po_hbh) {
2333 optdata = (void *)pktopt->ip6po_hbh;
2334 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2335 optdatalen = (ip6e->ip6e_len + 1) << 3;
2339 if (pktopt && pktopt->ip6po_rthdr) {
2340 optdata = (void *)pktopt->ip6po_rthdr;
2341 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2342 optdatalen = (ip6e->ip6e_len + 1) << 3;
2345 case IPV6_RTHDRDSTOPTS:
2346 if (pktopt && pktopt->ip6po_dest1) {
2347 optdata = (void *)pktopt->ip6po_dest1;
2348 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2349 optdatalen = (ip6e->ip6e_len + 1) << 3;
2353 if (pktopt && pktopt->ip6po_dest2) {
2354 optdata = (void *)pktopt->ip6po_dest2;
2355 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2356 optdatalen = (ip6e->ip6e_len + 1) << 3;
2360 if (pktopt && pktopt->ip6po_nexthop) {
2361 optdata = (void *)pktopt->ip6po_nexthop;
2362 optdatalen = pktopt->ip6po_nexthop->sa_len;
2365 case IPV6_USE_MIN_MTU:
2367 optdata = (void *)&pktopt->ip6po_minmtu;
2369 optdata = (void *)&defminmtu;
2370 optdatalen = sizeof(int);
2373 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2377 optdata = (void *)&on;
2378 optdatalen = sizeof(on);
2380 case IPV6_PREFER_TEMPADDR:
2382 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2384 optdata = (void *)&defpreftemp;
2385 optdatalen = sizeof(int);
2387 default: /* should not happen */
2389 panic("ip6_getpcbopt: unexpected option\n");
2391 return (ENOPROTOOPT);
2394 error = sooptcopyout(sopt, optdata, optdatalen);
2400 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2405 if (optname == -1 || optname == IPV6_PKTINFO) {
2406 if (pktopt->ip6po_pktinfo)
2407 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2408 pktopt->ip6po_pktinfo = NULL;
2410 if (optname == -1 || optname == IPV6_HOPLIMIT)
2411 pktopt->ip6po_hlim = -1;
2412 if (optname == -1 || optname == IPV6_TCLASS)
2413 pktopt->ip6po_tclass = -1;
2414 if (optname == -1 || optname == IPV6_NEXTHOP) {
2415 if (pktopt->ip6po_nextroute.ro_rt) {
2416 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2417 pktopt->ip6po_nextroute.ro_rt = NULL;
2419 if (pktopt->ip6po_nexthop)
2420 free(pktopt->ip6po_nexthop, M_IP6OPT);
2421 pktopt->ip6po_nexthop = NULL;
2423 if (optname == -1 || optname == IPV6_HOPOPTS) {
2424 if (pktopt->ip6po_hbh)
2425 free(pktopt->ip6po_hbh, M_IP6OPT);
2426 pktopt->ip6po_hbh = NULL;
2428 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2429 if (pktopt->ip6po_dest1)
2430 free(pktopt->ip6po_dest1, M_IP6OPT);
2431 pktopt->ip6po_dest1 = NULL;
2433 if (optname == -1 || optname == IPV6_RTHDR) {
2434 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2435 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2436 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2437 if (pktopt->ip6po_route.ro_rt) {
2438 RTFREE(pktopt->ip6po_route.ro_rt);
2439 pktopt->ip6po_route.ro_rt = NULL;
2442 if (optname == -1 || optname == IPV6_DSTOPTS) {
2443 if (pktopt->ip6po_dest2)
2444 free(pktopt->ip6po_dest2, M_IP6OPT);
2445 pktopt->ip6po_dest2 = NULL;
2449 #define PKTOPT_EXTHDRCPY(type) \
2452 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2453 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2454 if (dst->type == NULL && canwait == M_NOWAIT)\
2456 bcopy(src->type, dst->type, hlen);\
2458 } while (/*CONSTCOND*/ 0)
2461 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2463 if (dst == NULL || src == NULL) {
2464 printf("ip6_clearpktopts: invalid argument\n");
2468 dst->ip6po_hlim = src->ip6po_hlim;
2469 dst->ip6po_tclass = src->ip6po_tclass;
2470 dst->ip6po_flags = src->ip6po_flags;
2471 dst->ip6po_minmtu = src->ip6po_minmtu;
2472 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2473 if (src->ip6po_pktinfo) {
2474 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2476 if (dst->ip6po_pktinfo == NULL)
2478 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2480 if (src->ip6po_nexthop) {
2481 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2483 if (dst->ip6po_nexthop == NULL)
2485 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2486 src->ip6po_nexthop->sa_len);
2488 PKTOPT_EXTHDRCPY(ip6po_hbh);
2489 PKTOPT_EXTHDRCPY(ip6po_dest1);
2490 PKTOPT_EXTHDRCPY(ip6po_dest2);
2491 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2495 ip6_clearpktopts(dst, -1);
2498 #undef PKTOPT_EXTHDRCPY
2500 struct ip6_pktopts *
2501 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2504 struct ip6_pktopts *dst;
2506 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2509 ip6_initpktopts(dst);
2511 if ((error = copypktopts(dst, src, canwait)) != 0) {
2512 free(dst, M_IP6OPT);
2520 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2525 ip6_clearpktopts(pktopt, -1);
2527 free(pktopt, M_IP6OPT);
2531 * Set IPv6 outgoing packet options based on advanced API.
2534 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2535 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2537 struct cmsghdr *cm = 0;
2539 if (control == NULL || opt == NULL)
2542 ip6_initpktopts(opt);
2547 * If stickyopt is provided, make a local copy of the options
2548 * for this particular packet, then override them by ancillary
2550 * XXX: copypktopts() does not copy the cached route to a next
2551 * hop (if any). This is not very good in terms of efficiency,
2552 * but we can allow this since this option should be rarely
2555 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2560 * XXX: Currently, we assume all the optional information is stored
2563 if (control->m_next)
2566 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2567 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2570 if (control->m_len < CMSG_LEN(0))
2573 cm = mtod(control, struct cmsghdr *);
2574 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2576 if (cm->cmsg_level != IPPROTO_IPV6)
2579 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2580 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2589 * Set a particular packet option, as a sticky option or an ancillary data
2590 * item. "len" can be 0 only when it's a sticky option.
2591 * We have 4 cases of combination of "sticky" and "cmsg":
2592 * "sticky=0, cmsg=0": impossible
2593 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2594 * "sticky=1, cmsg=0": RFC3542 socket option
2595 * "sticky=1, cmsg=1": RFC2292 socket option
2598 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2599 struct ucred *cred, int sticky, int cmsg, int uproto)
2601 int minmtupolicy, preftemp;
2604 if (!sticky && !cmsg) {
2606 printf("ip6_setpktopt: impossible case\n");
2612 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2613 * not be specified in the context of RFC3542. Conversely,
2614 * RFC3542 types should not be specified in the context of RFC2292.
2618 case IPV6_2292PKTINFO:
2619 case IPV6_2292HOPLIMIT:
2620 case IPV6_2292NEXTHOP:
2621 case IPV6_2292HOPOPTS:
2622 case IPV6_2292DSTOPTS:
2623 case IPV6_2292RTHDR:
2624 case IPV6_2292PKTOPTIONS:
2625 return (ENOPROTOOPT);
2628 if (sticky && cmsg) {
2635 case IPV6_RTHDRDSTOPTS:
2637 case IPV6_USE_MIN_MTU:
2640 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2641 return (ENOPROTOOPT);
2646 case IPV6_2292PKTINFO:
2649 struct ifnet *ifp = NULL;
2650 struct in6_pktinfo *pktinfo;
2652 if (len != sizeof(struct in6_pktinfo))
2655 pktinfo = (struct in6_pktinfo *)buf;
2658 * An application can clear any sticky IPV6_PKTINFO option by
2659 * doing a "regular" setsockopt with ipi6_addr being
2660 * in6addr_any and ipi6_ifindex being zero.
2661 * [RFC 3542, Section 6]
2663 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2664 pktinfo->ipi6_ifindex == 0 &&
2665 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2666 ip6_clearpktopts(opt, optname);
2670 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2671 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2675 /* validate the interface index if specified. */
2676 if (pktinfo->ipi6_ifindex > V_if_index ||
2677 pktinfo->ipi6_ifindex < 0) {
2680 if (pktinfo->ipi6_ifindex) {
2681 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2687 * We store the address anyway, and let in6_selectsrc()
2688 * validate the specified address. This is because ipi6_addr
2689 * may not have enough information about its scope zone, and
2690 * we may need additional information (such as outgoing
2691 * interface or the scope zone of a destination address) to
2692 * disambiguate the scope.
2693 * XXX: the delay of the validation may confuse the
2694 * application when it is used as a sticky option.
2696 if (opt->ip6po_pktinfo == NULL) {
2697 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2698 M_IP6OPT, M_NOWAIT);
2699 if (opt->ip6po_pktinfo == NULL)
2702 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2706 case IPV6_2292HOPLIMIT:
2712 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2713 * to simplify the ordering among hoplimit options.
2715 if (optname == IPV6_HOPLIMIT && sticky)
2716 return (ENOPROTOOPT);
2718 if (len != sizeof(int))
2721 if (*hlimp < -1 || *hlimp > 255)
2724 opt->ip6po_hlim = *hlimp;
2732 if (len != sizeof(int))
2734 tclass = *(int *)buf;
2735 if (tclass < -1 || tclass > 255)
2738 opt->ip6po_tclass = tclass;
2742 case IPV6_2292NEXTHOP:
2745 error = priv_check_cred(cred,
2746 PRIV_NETINET_SETHDROPTS, 0);
2751 if (len == 0) { /* just remove the option */
2752 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2756 /* check if cmsg_len is large enough for sa_len */
2757 if (len < sizeof(struct sockaddr) || len < *buf)
2760 switch (((struct sockaddr *)buf)->sa_family) {
2763 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2766 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2769 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2770 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2773 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2779 case AF_LINK: /* should eventually be supported */
2781 return (EAFNOSUPPORT);
2784 /* turn off the previous option, then set the new option. */
2785 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2786 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2787 if (opt->ip6po_nexthop == NULL)
2789 bcopy(buf, opt->ip6po_nexthop, *buf);
2792 case IPV6_2292HOPOPTS:
2795 struct ip6_hbh *hbh;
2799 * XXX: We don't allow a non-privileged user to set ANY HbH
2800 * options, since per-option restriction has too much
2804 error = priv_check_cred(cred,
2805 PRIV_NETINET_SETHDROPTS, 0);
2811 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2812 break; /* just remove the option */
2815 /* message length validation */
2816 if (len < sizeof(struct ip6_hbh))
2818 hbh = (struct ip6_hbh *)buf;
2819 hbhlen = (hbh->ip6h_len + 1) << 3;
2823 /* turn off the previous option, then set the new option. */
2824 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2825 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2826 if (opt->ip6po_hbh == NULL)
2828 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2833 case IPV6_2292DSTOPTS:
2835 case IPV6_RTHDRDSTOPTS:
2837 struct ip6_dest *dest, **newdest = NULL;
2840 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2841 error = priv_check_cred(cred,
2842 PRIV_NETINET_SETHDROPTS, 0);
2848 ip6_clearpktopts(opt, optname);
2849 break; /* just remove the option */
2852 /* message length validation */
2853 if (len < sizeof(struct ip6_dest))
2855 dest = (struct ip6_dest *)buf;
2856 destlen = (dest->ip6d_len + 1) << 3;
2861 * Determine the position that the destination options header
2862 * should be inserted; before or after the routing header.
2865 case IPV6_2292DSTOPTS:
2867 * The old advacned API is ambiguous on this point.
2868 * Our approach is to determine the position based
2869 * according to the existence of a routing header.
2870 * Note, however, that this depends on the order of the
2871 * extension headers in the ancillary data; the 1st
2872 * part of the destination options header must appear
2873 * before the routing header in the ancillary data,
2875 * RFC3542 solved the ambiguity by introducing
2876 * separate ancillary data or option types.
2878 if (opt->ip6po_rthdr == NULL)
2879 newdest = &opt->ip6po_dest1;
2881 newdest = &opt->ip6po_dest2;
2883 case IPV6_RTHDRDSTOPTS:
2884 newdest = &opt->ip6po_dest1;
2887 newdest = &opt->ip6po_dest2;
2891 /* turn off the previous option, then set the new option. */
2892 ip6_clearpktopts(opt, optname);
2893 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2894 if (*newdest == NULL)
2896 bcopy(dest, *newdest, destlen);
2901 case IPV6_2292RTHDR:
2904 struct ip6_rthdr *rth;
2908 ip6_clearpktopts(opt, IPV6_RTHDR);
2909 break; /* just remove the option */
2912 /* message length validation */
2913 if (len < sizeof(struct ip6_rthdr))
2915 rth = (struct ip6_rthdr *)buf;
2916 rthlen = (rth->ip6r_len + 1) << 3;
2920 switch (rth->ip6r_type) {
2921 case IPV6_RTHDR_TYPE_0:
2922 if (rth->ip6r_len == 0) /* must contain one addr */
2924 if (rth->ip6r_len % 2) /* length must be even */
2926 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2930 return (EINVAL); /* not supported */
2933 /* turn off the previous option */
2934 ip6_clearpktopts(opt, IPV6_RTHDR);
2935 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2936 if (opt->ip6po_rthdr == NULL)
2938 bcopy(rth, opt->ip6po_rthdr, rthlen);
2943 case IPV6_USE_MIN_MTU:
2944 if (len != sizeof(int))
2946 minmtupolicy = *(int *)buf;
2947 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2948 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2949 minmtupolicy != IP6PO_MINMTU_ALL) {
2952 opt->ip6po_minmtu = minmtupolicy;
2956 if (len != sizeof(int))
2959 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2961 * we ignore this option for TCP sockets.
2962 * (RFC3542 leaves this case unspecified.)
2964 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2966 opt->ip6po_flags |= IP6PO_DONTFRAG;
2969 case IPV6_PREFER_TEMPADDR:
2970 if (len != sizeof(int))
2972 preftemp = *(int *)buf;
2973 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2974 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2975 preftemp != IP6PO_TEMPADDR_PREFER) {
2978 opt->ip6po_prefer_tempaddr = preftemp;
2982 return (ENOPROTOOPT);
2983 } /* end of switch */
2989 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2990 * packet to the input queue of a specified interface. Note that this
2991 * calls the output routine of the loopback "driver", but with an interface
2992 * pointer that might NOT be &loif -- easier than replicating that code here.
2995 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2998 struct ip6_hdr *ip6;
3000 copym = m_copy(m, 0, M_COPYALL);
3005 * Make sure to deep-copy IPv6 header portion in case the data
3006 * is in an mbuf cluster, so that we can safely override the IPv6
3007 * header portion later.
3009 if ((copym->m_flags & M_EXT) != 0 ||
3010 copym->m_len < sizeof(struct ip6_hdr)) {
3011 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3017 if (copym->m_len < sizeof(*ip6)) {
3023 ip6 = mtod(copym, struct ip6_hdr *);
3025 * clear embedded scope identifiers if necessary.
3026 * in6_clearscope will touch the addresses only when necessary.
3028 in6_clearscope(&ip6->ip6_src);
3029 in6_clearscope(&ip6->ip6_dst);
3031 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3035 * Chop IPv6 header off from the payload.
3038 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3041 struct ip6_hdr *ip6;
3043 ip6 = mtod(m, struct ip6_hdr *);
3044 if (m->m_len > sizeof(*ip6)) {
3045 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3050 M_MOVE_PKTHDR(mh, m);
3051 MH_ALIGN(mh, sizeof(*ip6));
3052 m->m_len -= sizeof(*ip6);
3053 m->m_data += sizeof(*ip6);
3056 m->m_len = sizeof(*ip6);
3057 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3059 exthdrs->ip6e_ip6 = m;
3064 * Compute IPv6 extension header length.
3067 ip6_optlen(struct inpcb *in6p)
3071 if (!in6p->in6p_outputopts)
3076 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3078 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3079 if (in6p->in6p_outputopts->ip6po_rthdr)
3080 /* dest1 is valid with rthdr only */
3081 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3082 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3083 len += elen(in6p->in6p_outputopts->ip6po_dest2);