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 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 V_ip6stat.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_NOWAIT, MT_HEADER);
1128 V_ip6stat.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 V_ip6stat.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 V_ip6stat.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 V_ip6stat.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 V_ip6stat.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_NOWAIT, MT_DATA);
1230 MCLGET(m, M_NOWAIT);
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_NOWAIT, 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_NOWAIT, MT_DATA);
1297 MCLGET(n, M_NOWAIT);
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_NOWAIT);
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_NOWAIT, 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 (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1495 if ((so->so_options &
1496 (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1497 in6p->inp_flags2 |= INP_REUSEPORT;
1499 in6p->inp_flags2 &= ~INP_REUSEPORT;
1506 if ((so->so_options & SO_REUSEPORT) != 0)
1507 in6p->inp_flags2 |= INP_REUSEPORT;
1509 in6p->inp_flags2 &= ~INP_REUSEPORT;
1515 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1523 } else { /* level == IPPROTO_IPV6 */
1528 case IPV6_2292PKTOPTIONS:
1529 #ifdef IPV6_PKTOPTIONS
1530 case IPV6_PKTOPTIONS:
1535 error = soopt_getm(sopt, &m); /* XXX */
1538 error = soopt_mcopyin(sopt, m); /* XXX */
1541 error = ip6_pcbopts(&in6p->in6p_outputopts,
1543 m_freem(m); /* XXX */
1548 * Use of some Hop-by-Hop options or some
1549 * Destination options, might require special
1550 * privilege. That is, normal applications
1551 * (without special privilege) might be forbidden
1552 * from setting certain options in outgoing packets,
1553 * and might never see certain options in received
1554 * packets. [RFC 2292 Section 6]
1555 * KAME specific note:
1556 * KAME prevents non-privileged users from sending or
1557 * receiving ANY hbh/dst options in order to avoid
1558 * overhead of parsing options in the kernel.
1560 case IPV6_RECVHOPOPTS:
1561 case IPV6_RECVDSTOPTS:
1562 case IPV6_RECVRTHDRDSTOPTS:
1564 error = priv_check(td,
1565 PRIV_NETINET_SETHDROPTS);
1570 case IPV6_UNICAST_HOPS:
1574 case IPV6_RECVPKTINFO:
1575 case IPV6_RECVHOPLIMIT:
1576 case IPV6_RECVRTHDR:
1577 case IPV6_RECVPATHMTU:
1578 case IPV6_RECVTCLASS:
1580 case IPV6_AUTOFLOWLABEL:
1582 if (optname == IPV6_BINDANY && td != NULL) {
1583 error = priv_check(td,
1584 PRIV_NETINET_BINDANY);
1589 if (optlen != sizeof(int)) {
1593 error = sooptcopyin(sopt, &optval,
1594 sizeof optval, sizeof optval);
1599 case IPV6_UNICAST_HOPS:
1600 if (optval < -1 || optval >= 256)
1603 /* -1 = kernel default */
1604 in6p->in6p_hops = optval;
1605 if ((in6p->inp_vflag &
1607 in6p->inp_ip_ttl = optval;
1610 #define OPTSET(bit) \
1614 in6p->inp_flags |= (bit); \
1616 in6p->inp_flags &= ~(bit); \
1617 INP_WUNLOCK(in6p); \
1618 } while (/*CONSTCOND*/ 0)
1619 #define OPTSET2292(bit) \
1622 in6p->inp_flags |= IN6P_RFC2292; \
1624 in6p->inp_flags |= (bit); \
1626 in6p->inp_flags &= ~(bit); \
1627 INP_WUNLOCK(in6p); \
1628 } while (/*CONSTCOND*/ 0)
1629 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1631 case IPV6_RECVPKTINFO:
1632 /* cannot mix with RFC2292 */
1633 if (OPTBIT(IN6P_RFC2292)) {
1637 OPTSET(IN6P_PKTINFO);
1642 struct ip6_pktopts **optp;
1644 /* cannot mix with RFC2292 */
1645 if (OPTBIT(IN6P_RFC2292)) {
1649 optp = &in6p->in6p_outputopts;
1650 error = ip6_pcbopt(IPV6_HOPLIMIT,
1651 (u_char *)&optval, sizeof(optval),
1652 optp, (td != NULL) ? td->td_ucred :
1657 case IPV6_RECVHOPLIMIT:
1658 /* cannot mix with RFC2292 */
1659 if (OPTBIT(IN6P_RFC2292)) {
1663 OPTSET(IN6P_HOPLIMIT);
1666 case IPV6_RECVHOPOPTS:
1667 /* cannot mix with RFC2292 */
1668 if (OPTBIT(IN6P_RFC2292)) {
1672 OPTSET(IN6P_HOPOPTS);
1675 case IPV6_RECVDSTOPTS:
1676 /* cannot mix with RFC2292 */
1677 if (OPTBIT(IN6P_RFC2292)) {
1681 OPTSET(IN6P_DSTOPTS);
1684 case IPV6_RECVRTHDRDSTOPTS:
1685 /* cannot mix with RFC2292 */
1686 if (OPTBIT(IN6P_RFC2292)) {
1690 OPTSET(IN6P_RTHDRDSTOPTS);
1693 case IPV6_RECVRTHDR:
1694 /* cannot mix with RFC2292 */
1695 if (OPTBIT(IN6P_RFC2292)) {
1706 case IPV6_RECVPATHMTU:
1708 * We ignore this option for TCP
1710 * (RFC3542 leaves this case
1713 if (uproto != IPPROTO_TCP)
1719 * make setsockopt(IPV6_V6ONLY)
1720 * available only prior to bind(2).
1721 * see ipng mailing list, Jun 22 2001.
1723 if (in6p->inp_lport ||
1724 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1728 OPTSET(IN6P_IPV6_V6ONLY);
1730 in6p->inp_vflag &= ~INP_IPV4;
1732 in6p->inp_vflag |= INP_IPV4;
1734 case IPV6_RECVTCLASS:
1735 /* cannot mix with RFC2292 XXX */
1736 if (OPTBIT(IN6P_RFC2292)) {
1740 OPTSET(IN6P_TCLASS);
1742 case IPV6_AUTOFLOWLABEL:
1743 OPTSET(IN6P_AUTOFLOWLABEL);
1747 OPTSET(INP_BINDANY);
1754 case IPV6_USE_MIN_MTU:
1755 case IPV6_PREFER_TEMPADDR:
1756 if (optlen != sizeof(optval)) {
1760 error = sooptcopyin(sopt, &optval,
1761 sizeof optval, sizeof optval);
1765 struct ip6_pktopts **optp;
1766 optp = &in6p->in6p_outputopts;
1767 error = ip6_pcbopt(optname,
1768 (u_char *)&optval, sizeof(optval),
1769 optp, (td != NULL) ? td->td_ucred :
1774 case IPV6_2292PKTINFO:
1775 case IPV6_2292HOPLIMIT:
1776 case IPV6_2292HOPOPTS:
1777 case IPV6_2292DSTOPTS:
1778 case IPV6_2292RTHDR:
1780 if (optlen != sizeof(int)) {
1784 error = sooptcopyin(sopt, &optval,
1785 sizeof optval, sizeof optval);
1789 case IPV6_2292PKTINFO:
1790 OPTSET2292(IN6P_PKTINFO);
1792 case IPV6_2292HOPLIMIT:
1793 OPTSET2292(IN6P_HOPLIMIT);
1795 case IPV6_2292HOPOPTS:
1797 * Check super-user privilege.
1798 * See comments for IPV6_RECVHOPOPTS.
1801 error = priv_check(td,
1802 PRIV_NETINET_SETHDROPTS);
1806 OPTSET2292(IN6P_HOPOPTS);
1808 case IPV6_2292DSTOPTS:
1810 error = priv_check(td,
1811 PRIV_NETINET_SETHDROPTS);
1815 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1817 case IPV6_2292RTHDR:
1818 OPTSET2292(IN6P_RTHDR);
1826 case IPV6_RTHDRDSTOPTS:
1829 /* new advanced API (RFC3542) */
1831 u_char optbuf_storage[MCLBYTES];
1833 struct ip6_pktopts **optp;
1835 /* cannot mix with RFC2292 */
1836 if (OPTBIT(IN6P_RFC2292)) {
1842 * We only ensure valsize is not too large
1843 * here. Further validation will be done
1846 error = sooptcopyin(sopt, optbuf_storage,
1847 sizeof(optbuf_storage), 0);
1850 optlen = sopt->sopt_valsize;
1851 optbuf = optbuf_storage;
1852 optp = &in6p->in6p_outputopts;
1853 error = ip6_pcbopt(optname, optbuf, optlen,
1854 optp, (td != NULL) ? td->td_ucred : NULL,
1860 case IPV6_MULTICAST_IF:
1861 case IPV6_MULTICAST_HOPS:
1862 case IPV6_MULTICAST_LOOP:
1863 case IPV6_JOIN_GROUP:
1864 case IPV6_LEAVE_GROUP:
1866 case MCAST_BLOCK_SOURCE:
1867 case MCAST_UNBLOCK_SOURCE:
1868 case MCAST_JOIN_GROUP:
1869 case MCAST_LEAVE_GROUP:
1870 case MCAST_JOIN_SOURCE_GROUP:
1871 case MCAST_LEAVE_SOURCE_GROUP:
1872 error = ip6_setmoptions(in6p, sopt);
1875 case IPV6_PORTRANGE:
1876 error = sooptcopyin(sopt, &optval,
1877 sizeof optval, sizeof optval);
1883 case IPV6_PORTRANGE_DEFAULT:
1884 in6p->inp_flags &= ~(INP_LOWPORT);
1885 in6p->inp_flags &= ~(INP_HIGHPORT);
1888 case IPV6_PORTRANGE_HIGH:
1889 in6p->inp_flags &= ~(INP_LOWPORT);
1890 in6p->inp_flags |= INP_HIGHPORT;
1893 case IPV6_PORTRANGE_LOW:
1894 in6p->inp_flags &= ~(INP_HIGHPORT);
1895 in6p->inp_flags |= INP_LOWPORT;
1906 case IPV6_IPSEC_POLICY:
1911 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1913 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1915 req = mtod(m, caddr_t);
1916 error = ipsec_set_policy(in6p, optname, req,
1917 m->m_len, (sopt->sopt_td != NULL) ?
1918 sopt->sopt_td->td_ucred : NULL);
1925 error = ENOPROTOOPT;
1933 case IPV6_2292PKTOPTIONS:
1934 #ifdef IPV6_PKTOPTIONS
1935 case IPV6_PKTOPTIONS:
1938 * RFC3542 (effectively) deprecated the
1939 * semantics of the 2292-style pktoptions.
1940 * Since it was not reliable in nature (i.e.,
1941 * applications had to expect the lack of some
1942 * information after all), it would make sense
1943 * to simplify this part by always returning
1946 sopt->sopt_valsize = 0;
1949 case IPV6_RECVHOPOPTS:
1950 case IPV6_RECVDSTOPTS:
1951 case IPV6_RECVRTHDRDSTOPTS:
1952 case IPV6_UNICAST_HOPS:
1953 case IPV6_RECVPKTINFO:
1954 case IPV6_RECVHOPLIMIT:
1955 case IPV6_RECVRTHDR:
1956 case IPV6_RECVPATHMTU:
1960 case IPV6_PORTRANGE:
1961 case IPV6_RECVTCLASS:
1962 case IPV6_AUTOFLOWLABEL:
1966 case IPV6_RECVHOPOPTS:
1967 optval = OPTBIT(IN6P_HOPOPTS);
1970 case IPV6_RECVDSTOPTS:
1971 optval = OPTBIT(IN6P_DSTOPTS);
1974 case IPV6_RECVRTHDRDSTOPTS:
1975 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1978 case IPV6_UNICAST_HOPS:
1979 optval = in6p->in6p_hops;
1982 case IPV6_RECVPKTINFO:
1983 optval = OPTBIT(IN6P_PKTINFO);
1986 case IPV6_RECVHOPLIMIT:
1987 optval = OPTBIT(IN6P_HOPLIMIT);
1990 case IPV6_RECVRTHDR:
1991 optval = OPTBIT(IN6P_RTHDR);
1994 case IPV6_RECVPATHMTU:
1995 optval = OPTBIT(IN6P_MTU);
1999 optval = OPTBIT(INP_FAITH);
2003 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2006 case IPV6_PORTRANGE:
2009 flags = in6p->inp_flags;
2010 if (flags & INP_HIGHPORT)
2011 optval = IPV6_PORTRANGE_HIGH;
2012 else if (flags & INP_LOWPORT)
2013 optval = IPV6_PORTRANGE_LOW;
2018 case IPV6_RECVTCLASS:
2019 optval = OPTBIT(IN6P_TCLASS);
2022 case IPV6_AUTOFLOWLABEL:
2023 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2027 optval = OPTBIT(INP_BINDANY);
2032 error = sooptcopyout(sopt, &optval,
2039 struct ip6_mtuinfo mtuinfo;
2040 struct route_in6 sro;
2042 bzero(&sro, sizeof(sro));
2044 if (!(so->so_state & SS_ISCONNECTED))
2047 * XXX: we dot not consider the case of source
2048 * routing, or optional information to specify
2049 * the outgoing interface.
2051 error = ip6_getpmtu(&sro, NULL, NULL,
2052 &in6p->in6p_faddr, &pmtu, NULL,
2058 if (pmtu > IPV6_MAXPACKET)
2059 pmtu = IPV6_MAXPACKET;
2061 bzero(&mtuinfo, sizeof(mtuinfo));
2062 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2063 optdata = (void *)&mtuinfo;
2064 optdatalen = sizeof(mtuinfo);
2065 error = sooptcopyout(sopt, optdata,
2070 case IPV6_2292PKTINFO:
2071 case IPV6_2292HOPLIMIT:
2072 case IPV6_2292HOPOPTS:
2073 case IPV6_2292RTHDR:
2074 case IPV6_2292DSTOPTS:
2076 case IPV6_2292PKTINFO:
2077 optval = OPTBIT(IN6P_PKTINFO);
2079 case IPV6_2292HOPLIMIT:
2080 optval = OPTBIT(IN6P_HOPLIMIT);
2082 case IPV6_2292HOPOPTS:
2083 optval = OPTBIT(IN6P_HOPOPTS);
2085 case IPV6_2292RTHDR:
2086 optval = OPTBIT(IN6P_RTHDR);
2088 case IPV6_2292DSTOPTS:
2089 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2092 error = sooptcopyout(sopt, &optval,
2099 case IPV6_RTHDRDSTOPTS:
2103 case IPV6_USE_MIN_MTU:
2104 case IPV6_PREFER_TEMPADDR:
2105 error = ip6_getpcbopt(in6p->in6p_outputopts,
2109 case IPV6_MULTICAST_IF:
2110 case IPV6_MULTICAST_HOPS:
2111 case IPV6_MULTICAST_LOOP:
2113 error = ip6_getmoptions(in6p, sopt);
2117 case IPV6_IPSEC_POLICY:
2121 struct mbuf *m = NULL;
2122 struct mbuf **mp = &m;
2123 size_t ovalsize = sopt->sopt_valsize;
2124 caddr_t oval = (caddr_t)sopt->sopt_val;
2126 error = soopt_getm(sopt, &m); /* XXX */
2129 error = soopt_mcopyin(sopt, m); /* XXX */
2132 sopt->sopt_valsize = ovalsize;
2133 sopt->sopt_val = oval;
2135 req = mtod(m, caddr_t);
2138 error = ipsec_get_policy(in6p, req, len, mp);
2140 error = soopt_mcopyout(sopt, m); /* XXX */
2141 if (error == 0 && m)
2148 error = ENOPROTOOPT;
2158 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2160 int error = 0, optval, optlen;
2161 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2162 struct inpcb *in6p = sotoinpcb(so);
2163 int level, op, optname;
2165 level = sopt->sopt_level;
2166 op = sopt->sopt_dir;
2167 optname = sopt->sopt_name;
2168 optlen = sopt->sopt_valsize;
2170 if (level != IPPROTO_IPV6) {
2177 * For ICMPv6 sockets, no modification allowed for checksum
2178 * offset, permit "no change" values to help existing apps.
2180 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2181 * for an ICMPv6 socket will fail."
2182 * The current behavior does not meet RFC3542.
2186 if (optlen != sizeof(int)) {
2190 error = sooptcopyin(sopt, &optval, sizeof(optval),
2194 if ((optval % 2) != 0) {
2195 /* the API assumes even offset values */
2197 } else if (so->so_proto->pr_protocol ==
2199 if (optval != icmp6off)
2202 in6p->in6p_cksum = optval;
2206 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2209 optval = in6p->in6p_cksum;
2211 error = sooptcopyout(sopt, &optval, sizeof(optval));
2221 error = ENOPROTOOPT;
2229 * Set up IP6 options in pcb for insertion in output packets or
2230 * specifying behavior of outgoing packets.
2233 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2234 struct socket *so, struct sockopt *sopt)
2236 struct ip6_pktopts *opt = *pktopt;
2238 struct thread *td = sopt->sopt_td;
2240 /* turn off any old options. */
2243 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2244 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2245 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2246 printf("ip6_pcbopts: all specified options are cleared.\n");
2248 ip6_clearpktopts(opt, -1);
2250 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2253 if (!m || m->m_len == 0) {
2255 * Only turning off any previous options, regardless of
2256 * whether the opt is just created or given.
2258 free(opt, M_IP6OPT);
2262 /* set options specified by user. */
2263 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2264 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2265 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2266 free(opt, M_IP6OPT);
2274 * initialize ip6_pktopts. beware that there are non-zero default values in
2278 ip6_initpktopts(struct ip6_pktopts *opt)
2281 bzero(opt, sizeof(*opt));
2282 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2283 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2284 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2285 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2289 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2290 struct ucred *cred, int uproto)
2292 struct ip6_pktopts *opt;
2294 if (*pktopt == NULL) {
2295 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2297 ip6_initpktopts(*pktopt);
2301 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2305 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2307 void *optdata = NULL;
2309 struct ip6_ext *ip6e;
2311 struct in6_pktinfo null_pktinfo;
2312 int deftclass = 0, on;
2313 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2314 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2318 if (pktopt && pktopt->ip6po_pktinfo)
2319 optdata = (void *)pktopt->ip6po_pktinfo;
2321 /* XXX: we don't have to do this every time... */
2322 bzero(&null_pktinfo, sizeof(null_pktinfo));
2323 optdata = (void *)&null_pktinfo;
2325 optdatalen = sizeof(struct in6_pktinfo);
2328 if (pktopt && pktopt->ip6po_tclass >= 0)
2329 optdata = (void *)&pktopt->ip6po_tclass;
2331 optdata = (void *)&deftclass;
2332 optdatalen = sizeof(int);
2335 if (pktopt && pktopt->ip6po_hbh) {
2336 optdata = (void *)pktopt->ip6po_hbh;
2337 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2338 optdatalen = (ip6e->ip6e_len + 1) << 3;
2342 if (pktopt && pktopt->ip6po_rthdr) {
2343 optdata = (void *)pktopt->ip6po_rthdr;
2344 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2345 optdatalen = (ip6e->ip6e_len + 1) << 3;
2348 case IPV6_RTHDRDSTOPTS:
2349 if (pktopt && pktopt->ip6po_dest1) {
2350 optdata = (void *)pktopt->ip6po_dest1;
2351 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2352 optdatalen = (ip6e->ip6e_len + 1) << 3;
2356 if (pktopt && pktopt->ip6po_dest2) {
2357 optdata = (void *)pktopt->ip6po_dest2;
2358 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2359 optdatalen = (ip6e->ip6e_len + 1) << 3;
2363 if (pktopt && pktopt->ip6po_nexthop) {
2364 optdata = (void *)pktopt->ip6po_nexthop;
2365 optdatalen = pktopt->ip6po_nexthop->sa_len;
2368 case IPV6_USE_MIN_MTU:
2370 optdata = (void *)&pktopt->ip6po_minmtu;
2372 optdata = (void *)&defminmtu;
2373 optdatalen = sizeof(int);
2376 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2380 optdata = (void *)&on;
2381 optdatalen = sizeof(on);
2383 case IPV6_PREFER_TEMPADDR:
2385 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2387 optdata = (void *)&defpreftemp;
2388 optdatalen = sizeof(int);
2390 default: /* should not happen */
2392 panic("ip6_getpcbopt: unexpected option\n");
2394 return (ENOPROTOOPT);
2397 error = sooptcopyout(sopt, optdata, optdatalen);
2403 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2408 if (optname == -1 || optname == IPV6_PKTINFO) {
2409 if (pktopt->ip6po_pktinfo)
2410 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2411 pktopt->ip6po_pktinfo = NULL;
2413 if (optname == -1 || optname == IPV6_HOPLIMIT)
2414 pktopt->ip6po_hlim = -1;
2415 if (optname == -1 || optname == IPV6_TCLASS)
2416 pktopt->ip6po_tclass = -1;
2417 if (optname == -1 || optname == IPV6_NEXTHOP) {
2418 if (pktopt->ip6po_nextroute.ro_rt) {
2419 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2420 pktopt->ip6po_nextroute.ro_rt = NULL;
2422 if (pktopt->ip6po_nexthop)
2423 free(pktopt->ip6po_nexthop, M_IP6OPT);
2424 pktopt->ip6po_nexthop = NULL;
2426 if (optname == -1 || optname == IPV6_HOPOPTS) {
2427 if (pktopt->ip6po_hbh)
2428 free(pktopt->ip6po_hbh, M_IP6OPT);
2429 pktopt->ip6po_hbh = NULL;
2431 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2432 if (pktopt->ip6po_dest1)
2433 free(pktopt->ip6po_dest1, M_IP6OPT);
2434 pktopt->ip6po_dest1 = NULL;
2436 if (optname == -1 || optname == IPV6_RTHDR) {
2437 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2438 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2439 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2440 if (pktopt->ip6po_route.ro_rt) {
2441 RTFREE(pktopt->ip6po_route.ro_rt);
2442 pktopt->ip6po_route.ro_rt = NULL;
2445 if (optname == -1 || optname == IPV6_DSTOPTS) {
2446 if (pktopt->ip6po_dest2)
2447 free(pktopt->ip6po_dest2, M_IP6OPT);
2448 pktopt->ip6po_dest2 = NULL;
2452 #define PKTOPT_EXTHDRCPY(type) \
2455 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2456 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2457 if (dst->type == NULL && canwait == M_NOWAIT)\
2459 bcopy(src->type, dst->type, hlen);\
2461 } while (/*CONSTCOND*/ 0)
2464 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2466 if (dst == NULL || src == NULL) {
2467 printf("ip6_clearpktopts: invalid argument\n");
2471 dst->ip6po_hlim = src->ip6po_hlim;
2472 dst->ip6po_tclass = src->ip6po_tclass;
2473 dst->ip6po_flags = src->ip6po_flags;
2474 dst->ip6po_minmtu = src->ip6po_minmtu;
2475 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2476 if (src->ip6po_pktinfo) {
2477 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2479 if (dst->ip6po_pktinfo == NULL)
2481 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2483 if (src->ip6po_nexthop) {
2484 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2486 if (dst->ip6po_nexthop == NULL)
2488 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2489 src->ip6po_nexthop->sa_len);
2491 PKTOPT_EXTHDRCPY(ip6po_hbh);
2492 PKTOPT_EXTHDRCPY(ip6po_dest1);
2493 PKTOPT_EXTHDRCPY(ip6po_dest2);
2494 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2498 ip6_clearpktopts(dst, -1);
2501 #undef PKTOPT_EXTHDRCPY
2503 struct ip6_pktopts *
2504 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2507 struct ip6_pktopts *dst;
2509 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2512 ip6_initpktopts(dst);
2514 if ((error = copypktopts(dst, src, canwait)) != 0) {
2515 free(dst, M_IP6OPT);
2523 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2528 ip6_clearpktopts(pktopt, -1);
2530 free(pktopt, M_IP6OPT);
2534 * Set IPv6 outgoing packet options based on advanced API.
2537 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2538 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2540 struct cmsghdr *cm = 0;
2542 if (control == NULL || opt == NULL)
2545 ip6_initpktopts(opt);
2550 * If stickyopt is provided, make a local copy of the options
2551 * for this particular packet, then override them by ancillary
2553 * XXX: copypktopts() does not copy the cached route to a next
2554 * hop (if any). This is not very good in terms of efficiency,
2555 * but we can allow this since this option should be rarely
2558 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2563 * XXX: Currently, we assume all the optional information is stored
2566 if (control->m_next)
2569 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2570 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2573 if (control->m_len < CMSG_LEN(0))
2576 cm = mtod(control, struct cmsghdr *);
2577 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2579 if (cm->cmsg_level != IPPROTO_IPV6)
2582 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2583 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2592 * Set a particular packet option, as a sticky option or an ancillary data
2593 * item. "len" can be 0 only when it's a sticky option.
2594 * We have 4 cases of combination of "sticky" and "cmsg":
2595 * "sticky=0, cmsg=0": impossible
2596 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2597 * "sticky=1, cmsg=0": RFC3542 socket option
2598 * "sticky=1, cmsg=1": RFC2292 socket option
2601 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2602 struct ucred *cred, int sticky, int cmsg, int uproto)
2604 int minmtupolicy, preftemp;
2607 if (!sticky && !cmsg) {
2609 printf("ip6_setpktopt: impossible case\n");
2615 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2616 * not be specified in the context of RFC3542. Conversely,
2617 * RFC3542 types should not be specified in the context of RFC2292.
2621 case IPV6_2292PKTINFO:
2622 case IPV6_2292HOPLIMIT:
2623 case IPV6_2292NEXTHOP:
2624 case IPV6_2292HOPOPTS:
2625 case IPV6_2292DSTOPTS:
2626 case IPV6_2292RTHDR:
2627 case IPV6_2292PKTOPTIONS:
2628 return (ENOPROTOOPT);
2631 if (sticky && cmsg) {
2638 case IPV6_RTHDRDSTOPTS:
2640 case IPV6_USE_MIN_MTU:
2643 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2644 return (ENOPROTOOPT);
2649 case IPV6_2292PKTINFO:
2652 struct ifnet *ifp = NULL;
2653 struct in6_pktinfo *pktinfo;
2655 if (len != sizeof(struct in6_pktinfo))
2658 pktinfo = (struct in6_pktinfo *)buf;
2661 * An application can clear any sticky IPV6_PKTINFO option by
2662 * doing a "regular" setsockopt with ipi6_addr being
2663 * in6addr_any and ipi6_ifindex being zero.
2664 * [RFC 3542, Section 6]
2666 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2667 pktinfo->ipi6_ifindex == 0 &&
2668 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2669 ip6_clearpktopts(opt, optname);
2673 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2674 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678 /* validate the interface index if specified. */
2679 if (pktinfo->ipi6_ifindex > V_if_index ||
2680 pktinfo->ipi6_ifindex < 0) {
2683 if (pktinfo->ipi6_ifindex) {
2684 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2690 * We store the address anyway, and let in6_selectsrc()
2691 * validate the specified address. This is because ipi6_addr
2692 * may not have enough information about its scope zone, and
2693 * we may need additional information (such as outgoing
2694 * interface or the scope zone of a destination address) to
2695 * disambiguate the scope.
2696 * XXX: the delay of the validation may confuse the
2697 * application when it is used as a sticky option.
2699 if (opt->ip6po_pktinfo == NULL) {
2700 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2701 M_IP6OPT, M_NOWAIT);
2702 if (opt->ip6po_pktinfo == NULL)
2705 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2709 case IPV6_2292HOPLIMIT:
2715 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2716 * to simplify the ordering among hoplimit options.
2718 if (optname == IPV6_HOPLIMIT && sticky)
2719 return (ENOPROTOOPT);
2721 if (len != sizeof(int))
2724 if (*hlimp < -1 || *hlimp > 255)
2727 opt->ip6po_hlim = *hlimp;
2735 if (len != sizeof(int))
2737 tclass = *(int *)buf;
2738 if (tclass < -1 || tclass > 255)
2741 opt->ip6po_tclass = tclass;
2745 case IPV6_2292NEXTHOP:
2748 error = priv_check_cred(cred,
2749 PRIV_NETINET_SETHDROPTS, 0);
2754 if (len == 0) { /* just remove the option */
2755 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2759 /* check if cmsg_len is large enough for sa_len */
2760 if (len < sizeof(struct sockaddr) || len < *buf)
2763 switch (((struct sockaddr *)buf)->sa_family) {
2766 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2769 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2772 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2773 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2776 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2782 case AF_LINK: /* should eventually be supported */
2784 return (EAFNOSUPPORT);
2787 /* turn off the previous option, then set the new option. */
2788 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2789 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2790 if (opt->ip6po_nexthop == NULL)
2792 bcopy(buf, opt->ip6po_nexthop, *buf);
2795 case IPV6_2292HOPOPTS:
2798 struct ip6_hbh *hbh;
2802 * XXX: We don't allow a non-privileged user to set ANY HbH
2803 * options, since per-option restriction has too much
2807 error = priv_check_cred(cred,
2808 PRIV_NETINET_SETHDROPTS, 0);
2814 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2815 break; /* just remove the option */
2818 /* message length validation */
2819 if (len < sizeof(struct ip6_hbh))
2821 hbh = (struct ip6_hbh *)buf;
2822 hbhlen = (hbh->ip6h_len + 1) << 3;
2826 /* turn off the previous option, then set the new option. */
2827 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2828 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2829 if (opt->ip6po_hbh == NULL)
2831 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2836 case IPV6_2292DSTOPTS:
2838 case IPV6_RTHDRDSTOPTS:
2840 struct ip6_dest *dest, **newdest = NULL;
2843 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2844 error = priv_check_cred(cred,
2845 PRIV_NETINET_SETHDROPTS, 0);
2851 ip6_clearpktopts(opt, optname);
2852 break; /* just remove the option */
2855 /* message length validation */
2856 if (len < sizeof(struct ip6_dest))
2858 dest = (struct ip6_dest *)buf;
2859 destlen = (dest->ip6d_len + 1) << 3;
2864 * Determine the position that the destination options header
2865 * should be inserted; before or after the routing header.
2868 case IPV6_2292DSTOPTS:
2870 * The old advacned API is ambiguous on this point.
2871 * Our approach is to determine the position based
2872 * according to the existence of a routing header.
2873 * Note, however, that this depends on the order of the
2874 * extension headers in the ancillary data; the 1st
2875 * part of the destination options header must appear
2876 * before the routing header in the ancillary data,
2878 * RFC3542 solved the ambiguity by introducing
2879 * separate ancillary data or option types.
2881 if (opt->ip6po_rthdr == NULL)
2882 newdest = &opt->ip6po_dest1;
2884 newdest = &opt->ip6po_dest2;
2886 case IPV6_RTHDRDSTOPTS:
2887 newdest = &opt->ip6po_dest1;
2890 newdest = &opt->ip6po_dest2;
2894 /* turn off the previous option, then set the new option. */
2895 ip6_clearpktopts(opt, optname);
2896 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2897 if (*newdest == NULL)
2899 bcopy(dest, *newdest, destlen);
2904 case IPV6_2292RTHDR:
2907 struct ip6_rthdr *rth;
2911 ip6_clearpktopts(opt, IPV6_RTHDR);
2912 break; /* just remove the option */
2915 /* message length validation */
2916 if (len < sizeof(struct ip6_rthdr))
2918 rth = (struct ip6_rthdr *)buf;
2919 rthlen = (rth->ip6r_len + 1) << 3;
2923 switch (rth->ip6r_type) {
2924 case IPV6_RTHDR_TYPE_0:
2925 if (rth->ip6r_len == 0) /* must contain one addr */
2927 if (rth->ip6r_len % 2) /* length must be even */
2929 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2933 return (EINVAL); /* not supported */
2936 /* turn off the previous option */
2937 ip6_clearpktopts(opt, IPV6_RTHDR);
2938 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2939 if (opt->ip6po_rthdr == NULL)
2941 bcopy(rth, opt->ip6po_rthdr, rthlen);
2946 case IPV6_USE_MIN_MTU:
2947 if (len != sizeof(int))
2949 minmtupolicy = *(int *)buf;
2950 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2951 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2952 minmtupolicy != IP6PO_MINMTU_ALL) {
2955 opt->ip6po_minmtu = minmtupolicy;
2959 if (len != sizeof(int))
2962 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2964 * we ignore this option for TCP sockets.
2965 * (RFC3542 leaves this case unspecified.)
2967 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2969 opt->ip6po_flags |= IP6PO_DONTFRAG;
2972 case IPV6_PREFER_TEMPADDR:
2973 if (len != sizeof(int))
2975 preftemp = *(int *)buf;
2976 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2977 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2978 preftemp != IP6PO_TEMPADDR_PREFER) {
2981 opt->ip6po_prefer_tempaddr = preftemp;
2985 return (ENOPROTOOPT);
2986 } /* end of switch */
2992 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2993 * packet to the input queue of a specified interface. Note that this
2994 * calls the output routine of the loopback "driver", but with an interface
2995 * pointer that might NOT be &loif -- easier than replicating that code here.
2998 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3001 struct ip6_hdr *ip6;
3003 copym = m_copy(m, 0, M_COPYALL);
3008 * Make sure to deep-copy IPv6 header portion in case the data
3009 * is in an mbuf cluster, so that we can safely override the IPv6
3010 * header portion later.
3012 if ((copym->m_flags & M_EXT) != 0 ||
3013 copym->m_len < sizeof(struct ip6_hdr)) {
3014 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3020 if (copym->m_len < sizeof(*ip6)) {
3026 ip6 = mtod(copym, struct ip6_hdr *);
3028 * clear embedded scope identifiers if necessary.
3029 * in6_clearscope will touch the addresses only when necessary.
3031 in6_clearscope(&ip6->ip6_src);
3032 in6_clearscope(&ip6->ip6_dst);
3034 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3038 * Chop IPv6 header off from the payload.
3041 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3044 struct ip6_hdr *ip6;
3046 ip6 = mtod(m, struct ip6_hdr *);
3047 if (m->m_len > sizeof(*ip6)) {
3048 MGETHDR(mh, M_NOWAIT, MT_HEADER);
3053 M_MOVE_PKTHDR(mh, m);
3054 MH_ALIGN(mh, sizeof(*ip6));
3055 m->m_len -= sizeof(*ip6);
3056 m->m_data += sizeof(*ip6);
3059 m->m_len = sizeof(*ip6);
3060 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3062 exthdrs->ip6e_ip6 = m;
3067 * Compute IPv6 extension header length.
3070 ip6_optlen(struct inpcb *in6p)
3074 if (!in6p->in6p_outputopts)
3079 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3081 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3082 if (in6p->in6p_outputopts->ip6po_rthdr)
3083 /* dest1 is valid with rthdr only */
3084 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3085 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3086 len += elen(in6p->in6p_outputopts->ip6po_dest2);