2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_kern_tls.h"
72 #include "opt_ratelimit.h"
73 #include "opt_route.h"
77 #include <sys/param.h>
78 #include <sys/kernel.h>
80 #include <sys/malloc.h>
82 #include <sys/errno.h>
85 #include <sys/protosw.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/syslog.h>
89 #include <sys/ucred.h>
91 #include <machine/in_cksum.h>
94 #include <net/if_var.h>
95 #include <net/if_llatbl.h>
96 #include <net/netisr.h>
97 #include <net/route.h>
98 #include <net/route/nhop.h>
100 #include <net/rss_config.h>
101 #include <net/vnet.h>
103 #include <netinet/in.h>
104 #include <netinet/in_var.h>
105 #include <netinet/ip_var.h>
106 #include <netinet6/in6_fib.h>
107 #include <netinet6/in6_var.h>
108 #include <netinet/ip6.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/tcp_var.h>
113 #include <netinet6/nd6.h>
114 #include <netinet6/in6_rss.h>
116 #include <netipsec/ipsec_support.h>
118 #include <netinet/sctp.h>
119 #include <netinet/sctp_crc32.h>
122 #include <netinet6/ip6protosw.h>
123 #include <netinet6/scope6_var.h>
125 extern int in6_mcast_loop;
128 struct mbuf *ip6e_ip6;
129 struct mbuf *ip6e_hbh;
130 struct mbuf *ip6e_dest1;
131 struct mbuf *ip6e_rthdr;
132 struct mbuf *ip6e_dest2;
135 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
137 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
138 struct ucred *, int);
139 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
140 struct socket *, struct sockopt *);
141 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
142 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
143 struct ucred *, int, int, int);
145 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
146 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
148 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
149 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
150 static int ip6_getpmtu(struct route_in6 *, int,
151 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
153 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
154 u_long *, int *, u_int);
155 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
156 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
160 * Make an extension header from option data. hp is the source,
161 * mp is the destination, and _ol is the optlen.
163 #define MAKE_EXTHDR(hp, mp, _ol) \
166 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
167 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
168 ((eh)->ip6e_len + 1) << 3); \
171 (_ol) += (*(mp))->m_len; \
173 } while (/*CONSTCOND*/ 0)
176 * Form a chain of extension headers.
177 * m is the extension header mbuf
178 * mp is the previous mbuf in the chain
179 * p is the next header
180 * i is the type of option.
182 #define MAKE_CHAIN(m, mp, p, i)\
186 panic("%s:%d: assumption failed: "\
187 "hdr not split: hdrsplit %d exthdrs %p",\
188 __func__, __LINE__, hdrsplit, &exthdrs);\
189 *mtod((m), u_char *) = *(p);\
191 p = mtod((m), u_char *);\
192 (m)->m_next = (mp)->m_next;\
196 } while (/*CONSTCOND*/ 0)
199 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
203 csum = in_cksum_skip(m, offset + plen, offset);
204 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
206 offset += m->m_pkthdr.csum_data; /* checksum offset */
208 if (offset + sizeof(csum) > m->m_len)
209 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
211 *(u_short *)mtodo(m, offset) = csum;
215 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
216 int plen, int optlen, bool frag)
219 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
220 "csum_flags %#x frag %d\n",
221 __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
223 if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
225 (csum_flags & CSUM_SCTP_IPV6) ||
227 (!frag && (ifp->if_capenable & IFCAP_NOMAP) == 0)) {
228 m = mb_unmapped_to_ext(m);
231 in6_ifstat_inc(ifp, ifs6_out_fragfail);
233 IP6STAT_INC(ip6s_odropped);
236 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
237 in6_delayed_cksum(m, plen - optlen,
238 sizeof(struct ip6_hdr) + optlen);
239 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
242 if (csum_flags & CSUM_SCTP_IPV6) {
243 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
244 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
253 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
254 int fraglen , uint32_t id)
256 struct mbuf *m, **mnext, *m_frgpart;
257 struct ip6_hdr *ip6, *mhip6;
258 struct ip6_frag *ip6f;
261 int tlen = m0->m_pkthdr.len;
263 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
266 ip6 = mtod(m, struct ip6_hdr *);
267 mnext = &m->m_nextpkt;
269 for (off = hlen; off < tlen; off += fraglen) {
270 m = m_gethdr(M_NOWAIT, MT_DATA);
272 IP6STAT_INC(ip6s_odropped);
277 * Make sure the complete packet header gets copied
278 * from the originating mbuf to the newly created
279 * mbuf. This also ensures that existing firewall
280 * classification(s), VLAN tags and so on get copied
281 * to the resulting fragmented packet(s):
283 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
285 IP6STAT_INC(ip6s_odropped);
290 mnext = &m->m_nextpkt;
291 m->m_data += max_linkhdr;
292 mhip6 = mtod(m, struct ip6_hdr *);
294 m->m_len = sizeof(*mhip6);
295 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
297 IP6STAT_INC(ip6s_odropped);
300 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
301 if (off + fraglen >= tlen)
302 fraglen = tlen - off;
304 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
305 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
306 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
307 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
308 IP6STAT_INC(ip6s_odropped);
312 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
313 ip6f->ip6f_reserved = 0;
314 ip6f->ip6f_ident = id;
315 ip6f->ip6f_nxt = nextproto;
316 IP6STAT_INC(ip6s_ofragments);
317 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
324 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
325 struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
329 struct ktls_session *tls = NULL;
331 struct m_snd_tag *mst;
334 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
339 * If this is an unencrypted TLS record, save a reference to
340 * the record. This local reference is used to call
341 * ktls_output_eagain after the mbuf has been freed (thus
342 * dropping the mbuf's reference) in if_output.
344 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
345 tls = ktls_hold(m->m_next->m_epg_tls);
349 * If a TLS session doesn't have a valid tag, it must
350 * have had an earlier ifp mismatch, so drop this
358 * Always stamp tags that include NIC ktls.
364 if (inp != NULL && mst == NULL) {
365 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
366 (inp->inp_snd_tag != NULL &&
367 inp->inp_snd_tag->ifp != ifp))
368 in_pcboutput_txrtlmt(inp, ifp, m);
370 if (inp->inp_snd_tag != NULL)
371 mst = inp->inp_snd_tag;
374 if (stamp_tag && mst != NULL) {
375 KASSERT(m->m_pkthdr.rcvif == NULL,
376 ("trying to add a send tag to a forwarded packet"));
377 if (mst->ifp != ifp) {
382 /* stamp send tag on mbuf */
383 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
384 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
387 error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
390 /* Check for route change invalidating send tags. */
394 error = ktls_output_eagain(inp, tls);
400 in_pcboutput_eagain(inp);
407 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
408 * nxt, hlim, src, dst).
409 * This function may modify ver and hlim only.
410 * The mbuf chain containing the packet will be freed.
411 * The mbuf opt, if present, will not be freed.
412 * If route_in6 ro is present and has ro_nh initialized, route lookup would be
413 * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
414 * then result of route lookup is stored in ro->ro_nh.
416 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
417 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu.
419 * ifpp - XXX: just for statistics
422 * XXX TODO: no flowid is assigned for outbound flows?
425 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
426 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
427 struct ifnet **ifpp, struct inpcb *inp)
430 struct ifnet *ifp, *origifp;
433 struct route_in6 *ro_pmtu;
434 struct nhop_object *nh;
435 struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
436 struct in6_addr odst;
440 struct in6_ifaddr *ia = NULL;
442 int alwaysfrag, dontfrag;
443 u_int32_t optlen, plen = 0, unfragpartlen;
444 struct ip6_exthdrs exthdrs;
445 struct in6_addr src0, dst0;
451 struct m_tag *fwd_tag = NULL;
457 INP_LOCK_ASSERT(inp);
458 M_SETFIB(m, inp->inp_inc.inc_fibnum);
459 if ((flags & IP_NODEFAULTFLOWID) == 0) {
460 /* Unconditionally set flowid. */
461 m->m_pkthdr.flowid = inp->inp_flowid;
462 M_HASHTYPE_SET(m, inp->inp_flowtype);
465 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
469 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
471 * IPSec checking which handles several cases.
472 * FAST IPSEC: We re-injected the packet.
473 * XXX: need scope argument.
475 if (IPSEC_ENABLED(ipv6)) {
476 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
477 if (error == EINPROGRESS)
484 /* Source address validation. */
485 ip6 = mtod(m, struct ip6_hdr *);
486 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
487 (flags & IPV6_UNSPECSRC) == 0) {
489 IP6STAT_INC(ip6s_badscope);
492 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
494 IP6STAT_INC(ip6s_badscope);
499 * If we are given packet options to add extension headers prepare them.
500 * Calculate the total length of the extension header chain.
501 * Keep the length of the unfragmentable part for fragmentation.
503 bzero(&exthdrs, sizeof(exthdrs));
505 unfragpartlen = sizeof(struct ip6_hdr);
507 /* Hop-by-Hop options header. */
508 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
510 /* Destination options header (1st part). */
511 if (opt->ip6po_rthdr) {
512 #ifndef RTHDR_SUPPORT_IMPLEMENTED
514 * If there is a routing header, discard the packet
515 * right away here. RH0/1 are obsolete and we do not
516 * currently support RH2/3/4.
517 * People trying to use RH253/254 may want to disable
519 * The moment we do support any routing header (again)
520 * this block should check the routing type more
528 * Destination options header (1st part).
529 * This only makes sense with a routing header.
530 * See Section 9.2 of RFC 3542.
531 * Disabling this part just for MIP6 convenience is
532 * a bad idea. We need to think carefully about a
533 * way to make the advanced API coexist with MIP6
534 * options, which might automatically be inserted in
537 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
540 /* Routing header. */
541 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
543 unfragpartlen += optlen;
546 * NOTE: we don't add AH/ESP length here (done in
547 * ip6_ipsec_output()).
550 /* Destination options header (2nd part). */
551 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
555 * If there is at least one extension header,
556 * separate IP6 header from the payload.
560 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
564 m = exthdrs.ip6e_ip6;
565 ip6 = mtod(m, struct ip6_hdr *);
569 /* Adjust mbuf packet header length. */
570 m->m_pkthdr.len += optlen;
571 plen = m->m_pkthdr.len - sizeof(*ip6);
573 /* If this is a jumbo payload, insert a jumbo payload option. */
574 if (plen > IPV6_MAXPACKET) {
576 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
580 m = exthdrs.ip6e_ip6;
581 ip6 = mtod(m, struct ip6_hdr *);
584 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
588 ip6->ip6_plen = htons(plen);
589 nexthdrp = &ip6->ip6_nxt;
593 * Concatenate headers and fill in next header fields.
594 * Here we have, on "m"
596 * and we insert headers accordingly.
597 * Finally, we should be getting:
598 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
600 * During the header composing process "m" points to IPv6
601 * header. "mprev" points to an extension header prior to esp.
606 * We treat dest2 specially. This makes IPsec processing
607 * much easier. The goal here is to make mprev point the
608 * mbuf prior to dest2.
610 * Result: IPv6 dest2 payload.
611 * m and mprev will point to IPv6 header.
613 if (exthdrs.ip6e_dest2) {
615 panic("%s:%d: assumption failed: "
616 "hdr not split: hdrsplit %d exthdrs %p",
617 __func__, __LINE__, hdrsplit, &exthdrs);
618 exthdrs.ip6e_dest2->m_next = m->m_next;
619 m->m_next = exthdrs.ip6e_dest2;
620 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
621 ip6->ip6_nxt = IPPROTO_DSTOPTS;
625 * Result: IPv6 hbh dest1 rthdr dest2 payload.
626 * m will point to IPv6 header. mprev will point to the
627 * extension header prior to dest2 (rthdr in the above case).
629 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
630 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
632 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
636 IP6STAT_INC(ip6s_localout);
640 if (opt && opt->ip6po_rthdr)
641 ro = &opt->ip6po_route;
643 dst = (struct sockaddr_in6 *)&ro->ro_dst;
646 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
650 * If specified, try to fill in the traffic class field.
651 * Do not override if a non-zero value is already set.
652 * We check the diffserv field and the ECN field separately.
654 if (opt && opt->ip6po_tclass >= 0) {
657 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
659 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
662 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
665 /* Fill in or override the hop limit field, if necessary. */
666 if (opt && opt->ip6po_hlim != -1)
667 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
668 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
670 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
672 ip6->ip6_hlim = V_ip6_defmcasthlim;
675 if (ro == NULL || ro->ro_nh == NULL) {
676 bzero(dst, sizeof(*dst));
677 dst->sin6_family = AF_INET6;
678 dst->sin6_len = sizeof(*dst);
679 dst->sin6_addr = ip6->ip6_dst;
682 * Validate route against routing table changes.
683 * Make sure that the address family is set in route.
689 if (ro->ro_nh != NULL && inp != NULL) {
690 ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
691 NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
694 if (ro->ro_nh != NULL && fwd_tag == NULL &&
695 (!NH_IS_VALID(ro->ro_nh) ||
696 ro->ro_dst.sin6_family != AF_INET6 ||
697 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
698 RO_INVALIDATE_CACHE(ro);
700 if (ro->ro_nh != NULL && fwd_tag == NULL &&
701 ro->ro_dst.sin6_family == AF_INET6 &&
702 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
707 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
709 if (fwd_tag == NULL) {
710 bzero(&dst_sa, sizeof(dst_sa));
711 dst_sa.sin6_family = AF_INET6;
712 dst_sa.sin6_len = sizeof(dst_sa);
713 dst_sa.sin6_addr = ip6->ip6_dst;
715 error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
716 &nh, fibnum, m->m_pkthdr.flowid);
718 IP6STAT_INC(ip6s_noroute);
720 in6_ifstat_inc(ifp, ifs6_out_discard);
728 * If in6_selectroute() does not return a nexthop
729 * dst may not have been updated.
731 *dst = dst_sa; /* XXX */
733 if (nh->nh_flags & NHF_HOST)
735 ia = (struct in6_ifaddr *)(nh->nh_ifa);
736 counter_u64_add(nh->nh_pksent, 1);
739 struct nhop6_extended nh6;
740 struct in6_addr kdst;
743 if (fwd_tag == NULL) {
744 bzero(&dst_sa, sizeof(dst_sa));
745 dst_sa.sin6_family = AF_INET6;
746 dst_sa.sin6_len = sizeof(dst_sa);
747 dst_sa.sin6_addr = ip6->ip6_dst;
750 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
752 (ifp = im6o->im6o_multicast_ifp) != NULL) {
753 /* We do not need a route lookup. */
754 *dst = dst_sa; /* XXX */
758 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
760 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
761 IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
763 ifp = in6_getlinkifnet(scopeid);
764 *dst = dst_sa; /* XXX */
769 error = fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0,
772 IP6STAT_INC(ip6s_noroute);
773 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
774 error = EHOSTUNREACH;;
780 dst->sin6_addr = nh6.nh_addr;
782 fib6_free_nh_ext(fibnum, &nh6);
787 /* Then nh (for unicast) and ifp must be non-NULL valid values. */
788 if ((flags & IPV6_FORWARDING) == 0) {
789 /* XXX: the FORWARDING flag can be set for mrouting. */
790 in6_ifstat_inc(ifp, ifs6_out_request);
793 /* Setup data structures for scope ID checks. */
795 bzero(&src_sa, sizeof(src_sa));
796 src_sa.sin6_family = AF_INET6;
797 src_sa.sin6_len = sizeof(src_sa);
798 src_sa.sin6_addr = ip6->ip6_src;
801 /* Re-initialize to be sure. */
802 bzero(&dst_sa, sizeof(dst_sa));
803 dst_sa.sin6_family = AF_INET6;
804 dst_sa.sin6_len = sizeof(dst_sa);
805 dst_sa.sin6_addr = ip6->ip6_dst;
807 /* Check for valid scope ID. */
808 if (in6_setscope(&src0, ifp, &zone) == 0 &&
809 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
810 in6_setscope(&dst0, ifp, &zone) == 0 &&
811 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
813 * The outgoing interface is in the zone of the source
814 * and destination addresses.
816 * Because the loopback interface cannot receive
817 * packets with a different scope ID than its own,
818 * there is a trick to pretend the outgoing packet
819 * was received by the real network interface, by
820 * setting "origifp" different from "ifp". This is
821 * only allowed when "ifp" is a loopback network
822 * interface. Refer to code in nd6_output_ifp() for
828 * We should use ia_ifp to support the case of sending
829 * packets to an address of our own.
831 if (ia != NULL && ia->ia_ifp)
834 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
835 sa6_recoverscope(&src_sa) != 0 ||
836 sa6_recoverscope(&dst_sa) != 0 ||
837 dst_sa.sin6_scope_id == 0 ||
838 (src_sa.sin6_scope_id != 0 &&
839 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
840 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
842 * If the destination network interface is not a
843 * loopback interface, or the destination network
844 * address has no scope ID, or the source address has
845 * a scope ID set which is different from the
846 * destination address one, or there is no network
847 * interface representing this scope ID, the address
848 * pair is considered invalid.
850 IP6STAT_INC(ip6s_badscope);
851 in6_ifstat_inc(ifp, ifs6_out_discard);
853 error = EHOSTUNREACH; /* XXX */
856 /* All scope ID checks are successful. */
858 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
859 if (opt && opt->ip6po_nextroute.ro_nh) {
861 * The nexthop is explicitly specified by the
862 * application. We assume the next hop is an IPv6
865 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
867 else if ((nh->nh_flags & NHF_GATEWAY))
871 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
872 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
874 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
875 in6_ifstat_inc(ifp, ifs6_out_mcast);
877 /* Confirm that the outgoing interface supports multicast. */
878 if (!(ifp->if_flags & IFF_MULTICAST)) {
879 IP6STAT_INC(ip6s_noroute);
880 in6_ifstat_inc(ifp, ifs6_out_discard);
884 if ((im6o == NULL && in6_mcast_loop) ||
885 (im6o && im6o->im6o_multicast_loop)) {
887 * Loop back multicast datagram if not expressly
888 * forbidden to do so, even if we have not joined
889 * the address; protocols will filter it later,
890 * thus deferring a hash lookup and lock acquisition
891 * at the expense of an m_copym().
893 ip6_mloopback(ifp, m);
896 * If we are acting as a multicast router, perform
897 * multicast forwarding as if the packet had just
898 * arrived on the interface to which we are about
899 * to send. The multicast forwarding function
900 * recursively calls this function, using the
901 * IPV6_FORWARDING flag to prevent infinite recursion.
903 * Multicasts that are looped back by ip6_mloopback(),
904 * above, will be forwarded by the ip6_input() routine,
907 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
909 * XXX: ip6_mforward expects that rcvif is NULL
910 * when it is called from the originating path.
911 * However, it may not always be the case.
913 m->m_pkthdr.rcvif = NULL;
914 if (ip6_mforward(ip6, ifp, m) != 0) {
921 * Multicasts with a hoplimit of zero may be looped back,
922 * above, but must not be transmitted on a network.
923 * Also, multicasts addressed to the loopback interface
924 * are not sent -- the above call to ip6_mloopback() will
925 * loop back a copy if this host actually belongs to the
926 * destination group on the loopback interface.
928 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
929 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
936 * Fill the outgoing inteface to tell the upper layer
937 * to increment per-interface statistics.
942 /* Determine path MTU. */
943 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
944 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
946 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
947 "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
948 ifp, alwaysfrag, fibnum));
951 * The caller of this function may specify to use the minimum MTU
953 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
954 * setting. The logic is a bit complicated; by default, unicast
955 * packets will follow path MTU while multicast packets will be sent at
956 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
957 * including unicast ones will be sent at the minimum MTU. Multicast
958 * packets will always be sent at the minimum MTU unless
959 * IP6PO_MINMTU_DISABLE is explicitly specified.
960 * See RFC 3542 for more details.
962 if (mtu > IPV6_MMTU) {
963 if ((flags & IPV6_MINMTU))
965 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
967 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
969 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
975 * Clear embedded scope identifiers if necessary.
976 * in6_clearscope() will touch the addresses only when necessary.
978 in6_clearscope(&ip6->ip6_src);
979 in6_clearscope(&ip6->ip6_dst);
982 * If the outgoing packet contains a hop-by-hop options header,
983 * it must be examined and processed even by the source node.
984 * (RFC 2460, section 4.)
986 if (exthdrs.ip6e_hbh) {
987 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
988 u_int32_t dummy; /* XXX unused */
989 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
992 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
993 panic("ip6e_hbh is not contiguous");
996 * XXX: if we have to send an ICMPv6 error to the sender,
997 * we need the M_LOOP flag since icmp6_error() expects
998 * the IPv6 and the hop-by-hop options header are
999 * contiguous unless the flag is set.
1001 m->m_flags |= M_LOOP;
1002 m->m_pkthdr.rcvif = ifp;
1003 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1004 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1005 &dummy, &plen) < 0) {
1006 /* m was already freed at this point. */
1007 error = EINVAL;/* better error? */
1010 m->m_flags &= ~M_LOOP; /* XXX */
1011 m->m_pkthdr.rcvif = NULL;
1014 /* Jump over all PFIL processing if hooks are not active. */
1015 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1018 odst = ip6->ip6_dst;
1019 /* Run through list of hooks for output packets. */
1020 switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1022 ip6 = mtod(m, struct ip6_hdr *);
1032 /* See if destination IP address was changed by packet filter. */
1033 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1034 m->m_flags |= M_SKIP_FIREWALL;
1035 /* If destination is now ourself drop to ip6_input(). */
1036 if (in6_localip(&ip6->ip6_dst)) {
1037 m->m_flags |= M_FASTFWD_OURS;
1038 if (m->m_pkthdr.rcvif == NULL)
1039 m->m_pkthdr.rcvif = V_loif;
1040 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1041 m->m_pkthdr.csum_flags |=
1042 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1043 m->m_pkthdr.csum_data = 0xffff;
1046 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1047 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1049 error = netisr_queue(NETISR_IPV6, m);
1053 RO_INVALIDATE_CACHE(ro);
1054 needfiblookup = 1; /* Redo the routing table lookup. */
1057 /* See if fib was changed by packet filter. */
1058 if (fibnum != M_GETFIB(m)) {
1059 m->m_flags |= M_SKIP_FIREWALL;
1060 fibnum = M_GETFIB(m);
1062 RO_INVALIDATE_CACHE(ro);
1068 /* See if local, if yes, send it to netisr. */
1069 if (m->m_flags & M_FASTFWD_OURS) {
1070 if (m->m_pkthdr.rcvif == NULL)
1071 m->m_pkthdr.rcvif = V_loif;
1072 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1073 m->m_pkthdr.csum_flags |=
1074 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1075 m->m_pkthdr.csum_data = 0xffff;
1078 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1079 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1081 error = netisr_queue(NETISR_IPV6, m);
1084 /* Or forward to some other address? */
1085 if ((m->m_flags & M_IP6_NEXTHOP) &&
1086 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1088 dst = (struct sockaddr_in6 *)&ro->ro_dst;
1091 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1092 m->m_flags |= M_SKIP_FIREWALL;
1093 m->m_flags &= ~M_IP6_NEXTHOP;
1094 m_tag_delete(m, fwd_tag);
1100 * Send the packet to the outgoing interface.
1101 * If necessary, do IPv6 fragmentation before sending.
1103 * The logic here is rather complex:
1104 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1105 * 1-a: send as is if tlen <= path mtu
1106 * 1-b: fragment if tlen > path mtu
1108 * 2: if user asks us not to fragment (dontfrag == 1)
1109 * 2-a: send as is if tlen <= interface mtu
1110 * 2-b: error if tlen > interface mtu
1112 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1115 * 4: if dontfrag == 1 && alwaysfrag == 1
1116 * error, as we cannot handle this conflicting request.
1118 sw_csum = m->m_pkthdr.csum_flags;
1120 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1121 sw_csum &= ~ifp->if_hwassist;
1125 * If we added extension headers, we will not do TSO and calculate the
1126 * checksums ourselves for now.
1127 * XXX-BZ Need a framework to know when the NIC can handle it, even
1130 error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1133 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1134 tlen = m->m_pkthdr.len;
1136 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1140 if (dontfrag && alwaysfrag) { /* Case 4. */
1141 /* Conflicting request - can't transmit. */
1145 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */
1147 * Even if the DONTFRAG option is specified, we cannot send the
1148 * packet when the data length is larger than the MTU of the
1149 * outgoing interface.
1150 * Notify the error by sending IPV6_PATHMTU ancillary data if
1151 * application wanted to know the MTU value. Also return an
1152 * error code (this is not described in the API spec).
1155 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1160 /* Transmit packet without fragmentation. */
1161 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1162 struct in6_ifaddr *ia6;
1164 ip6 = mtod(m, struct ip6_hdr *);
1165 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1167 /* Record statistics for this interface address. */
1168 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1169 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1171 ifa_free(&ia6->ia_ifa);
1173 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1174 (flags & IP_NO_SND_TAG_RL) ? false : true);
1178 /* Try to fragment the packet. Cases 1-b and 3. */
1179 if (mtu < IPV6_MMTU) {
1180 /* Path MTU cannot be less than IPV6_MMTU. */
1182 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1184 } else if (ip6->ip6_plen == 0) {
1185 /* Jumbo payload cannot be fragmented. */
1187 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1193 * Too large for the destination or interface;
1194 * fragment if possible.
1195 * Must be able to put at least 8 bytes per fragment.
1197 if (mtu > IPV6_MAXPACKET)
1198 mtu = IPV6_MAXPACKET;
1200 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1203 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1208 * If the interface will not calculate checksums on
1209 * fragmented packets, then do it here.
1210 * XXX-BZ handle the hw offloading case. Need flags.
1212 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1213 plen, optlen, true);
1218 * Change the next header field of the last header in the
1219 * unfragmentable part.
1221 if (exthdrs.ip6e_rthdr) {
1222 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1223 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1224 } else if (exthdrs.ip6e_dest1) {
1225 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1226 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1227 } else if (exthdrs.ip6e_hbh) {
1228 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1229 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1231 ip6 = mtod(m, struct ip6_hdr *);
1232 nextproto = ip6->ip6_nxt;
1233 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1237 * Loop through length of segment after first fragment,
1238 * make new header and copy data of each part and link onto
1242 id = htonl(ip6_randomid());
1243 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1247 in6_ifstat_inc(ifp, ifs6_out_fragok);
1250 /* Remove leading garbage. */
1259 /* Record statistics for this interface address. */
1261 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1262 counter_u64_add(ia->ia_ifa.ifa_obytes,
1265 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1272 IP6STAT_INC(ip6s_fragmented);
1278 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1279 m_freem(exthdrs.ip6e_dest1);
1280 m_freem(exthdrs.ip6e_rthdr);
1281 m_freem(exthdrs.ip6e_dest2);
1290 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1294 if (hlen > MCLBYTES)
1295 return (ENOBUFS); /* XXX */
1298 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1300 m = m_get(M_NOWAIT, MT_DATA);
1305 bcopy(hdr, mtod(m, caddr_t), hlen);
1312 * Insert jumbo payload option.
1315 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1321 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1324 * If there is no hop-by-hop options header, allocate new one.
1325 * If there is one but it doesn't have enough space to store the
1326 * jumbo payload option, allocate a cluster to store the whole options.
1327 * Otherwise, use it to store the options.
1329 if (exthdrs->ip6e_hbh == NULL) {
1330 mopt = m_get(M_NOWAIT, MT_DATA);
1333 mopt->m_len = JUMBOOPTLEN;
1334 optbuf = mtod(mopt, u_char *);
1335 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1336 exthdrs->ip6e_hbh = mopt;
1338 struct ip6_hbh *hbh;
1340 mopt = exthdrs->ip6e_hbh;
1341 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1344 * - exthdrs->ip6e_hbh is not referenced from places
1345 * other than exthdrs.
1346 * - exthdrs->ip6e_hbh is not an mbuf chain.
1348 int oldoptlen = mopt->m_len;
1352 * XXX: give up if the whole (new) hbh header does
1353 * not fit even in an mbuf cluster.
1355 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1359 * As a consequence, we must always prepare a cluster
1362 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1365 n->m_len = oldoptlen + JUMBOOPTLEN;
1366 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1368 optbuf = mtod(n, caddr_t) + oldoptlen;
1370 mopt = exthdrs->ip6e_hbh = n;
1372 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1373 mopt->m_len += JUMBOOPTLEN;
1375 optbuf[0] = IP6OPT_PADN;
1379 * Adjust the header length according to the pad and
1380 * the jumbo payload option.
1382 hbh = mtod(mopt, struct ip6_hbh *);
1383 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1386 /* fill in the option. */
1387 optbuf[2] = IP6OPT_JUMBO;
1389 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1390 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1392 /* finally, adjust the packet header length */
1393 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1400 * Insert fragment header and copy unfragmentable header portions.
1403 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1404 struct ip6_frag **frghdrp)
1406 struct mbuf *n, *mlast;
1408 if (hlen > sizeof(struct ip6_hdr)) {
1409 n = m_copym(m0, sizeof(struct ip6_hdr),
1410 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1417 /* Search for the last mbuf of unfragmentable part. */
1418 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1421 if (M_WRITABLE(mlast) &&
1422 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1423 /* use the trailing space of the last mbuf for the fragment hdr */
1424 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1426 mlast->m_len += sizeof(struct ip6_frag);
1427 m->m_pkthdr.len += sizeof(struct ip6_frag);
1429 /* allocate a new mbuf for the fragment header */
1432 mfrg = m_get(M_NOWAIT, MT_DATA);
1435 mfrg->m_len = sizeof(struct ip6_frag);
1436 *frghdrp = mtod(mfrg, struct ip6_frag *);
1437 mlast->m_next = mfrg;
1444 * Calculates IPv6 path mtu for destination @dst.
1445 * Resulting MTU is stored in @mtup.
1447 * Returns 0 on success.
1450 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1452 struct nhop6_extended nh6;
1453 struct in6_addr kdst;
1459 in6_splitscope(dst, &kdst, &scopeid);
1460 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1461 return (EHOSTUNREACH);
1466 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1467 fib6_free_nh_ext(fibnum, &nh6);
1473 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1474 * and cached data in @ro_pmtu.
1475 * MTU from (successful) route lookup is saved (along with dst)
1476 * inside @ro_pmtu to avoid subsequent route lookups after packet
1477 * filter processing.
1479 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1480 * Returns 0 on success.
1483 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1484 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1485 int *alwaysfragp, u_int fibnum, u_int proto)
1487 struct nhop6_basic nh6;
1488 struct in6_addr kdst;
1490 struct sockaddr_in6 *sa6_dst, sin6;
1494 if (ro_pmtu == NULL || do_lookup) {
1497 * Here ro_pmtu has final destination address, while
1498 * ro might represent immediate destination.
1499 * Use ro_pmtu destination since mtu might differ.
1501 if (ro_pmtu != NULL) {
1502 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1503 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1504 ro_pmtu->ro_mtu = 0;
1508 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1509 bzero(sa6_dst, sizeof(*sa6_dst));
1510 sa6_dst->sin6_family = AF_INET6;
1511 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1512 sa6_dst->sin6_addr = *dst;
1514 in6_splitscope(dst, &kdst, &scopeid);
1515 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1518 if (ro_pmtu != NULL)
1519 ro_pmtu->ro_mtu = mtu;
1522 mtu = ro_pmtu->ro_mtu;
1525 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1526 mtu = ro_pmtu->ro_nh->nh_mtu;
1528 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1532 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1533 * hostcache data for @dst.
1534 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1536 * Returns 0 on success.
1539 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1540 u_long *mtup, int *alwaysfragp, u_int proto)
1548 struct in_conninfo inc;
1550 bzero(&inc, sizeof(inc));
1551 inc.inc_flags |= INC_ISIPV6;
1552 inc.inc6_faddr = *dst;
1554 ifmtu = IN6_LINKMTU(ifp);
1556 /* TCP is known to react to pmtu changes so skip hc */
1557 if (proto != IPPROTO_TCP)
1558 mtu = tcp_hc_getmtu(&inc);
1561 mtu = min(mtu, rt_mtu);
1566 else if (mtu < IPV6_MMTU) {
1568 * RFC2460 section 5, last paragraph:
1569 * if we record ICMPv6 too big message with
1570 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1571 * or smaller, with framgent header attached.
1572 * (fragment header is needed regardless from the
1573 * packet size, for translators to identify packets)
1579 mtu = IN6_LINKMTU(ifp);
1581 error = EHOSTUNREACH; /* XXX */
1585 *alwaysfragp = alwaysfrag;
1590 * IP6 socket option processing.
1593 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1595 int optdatalen, uproto;
1597 struct inpcb *inp = sotoinpcb(so);
1599 int level, op, optname;
1603 uint32_t rss_bucket;
1608 * Don't use more than a quarter of mbuf clusters. N.B.:
1609 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1610 * on LP64 architectures, so cast to u_long to avoid undefined
1611 * behavior. ILP32 architectures cannot have nmbclusters
1612 * large enough to overflow for other reasons.
1614 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1616 level = sopt->sopt_level;
1617 op = sopt->sopt_dir;
1618 optname = sopt->sopt_name;
1619 optlen = sopt->sopt_valsize;
1623 uproto = (int)so->so_proto->pr_protocol;
1625 if (level != IPPROTO_IPV6) {
1628 if (sopt->sopt_level == SOL_SOCKET &&
1629 sopt->sopt_dir == SOPT_SET) {
1630 switch (sopt->sopt_name) {
1633 if ((so->so_options & SO_REUSEADDR) != 0)
1634 inp->inp_flags2 |= INP_REUSEADDR;
1636 inp->inp_flags2 &= ~INP_REUSEADDR;
1642 if ((so->so_options & SO_REUSEPORT) != 0)
1643 inp->inp_flags2 |= INP_REUSEPORT;
1645 inp->inp_flags2 &= ~INP_REUSEPORT;
1649 case SO_REUSEPORT_LB:
1651 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1652 inp->inp_flags2 |= INP_REUSEPORT_LB;
1654 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1660 inp->inp_inc.inc_fibnum = so->so_fibnum;
1664 case SO_MAX_PACING_RATE:
1667 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1678 } else { /* level == IPPROTO_IPV6 */
1683 case IPV6_2292PKTOPTIONS:
1684 #ifdef IPV6_PKTOPTIONS
1685 case IPV6_PKTOPTIONS:
1690 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1691 printf("ip6_ctloutput: mbuf limit hit\n");
1696 error = soopt_getm(sopt, &m); /* XXX */
1699 error = soopt_mcopyin(sopt, m); /* XXX */
1703 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1706 m_freem(m); /* XXX */
1711 * Use of some Hop-by-Hop options or some
1712 * Destination options, might require special
1713 * privilege. That is, normal applications
1714 * (without special privilege) might be forbidden
1715 * from setting certain options in outgoing packets,
1716 * and might never see certain options in received
1717 * packets. [RFC 2292 Section 6]
1718 * KAME specific note:
1719 * KAME prevents non-privileged users from sending or
1720 * receiving ANY hbh/dst options in order to avoid
1721 * overhead of parsing options in the kernel.
1723 case IPV6_RECVHOPOPTS:
1724 case IPV6_RECVDSTOPTS:
1725 case IPV6_RECVRTHDRDSTOPTS:
1727 error = priv_check(td,
1728 PRIV_NETINET_SETHDROPTS);
1733 case IPV6_UNICAST_HOPS:
1736 case IPV6_RECVPKTINFO:
1737 case IPV6_RECVHOPLIMIT:
1738 case IPV6_RECVRTHDR:
1739 case IPV6_RECVPATHMTU:
1740 case IPV6_RECVTCLASS:
1741 case IPV6_RECVFLOWID:
1743 case IPV6_RECVRSSBUCKETID:
1746 case IPV6_AUTOFLOWLABEL:
1747 case IPV6_ORIGDSTADDR:
1749 case IPV6_BINDMULTI:
1751 case IPV6_RSS_LISTEN_BUCKET:
1753 if (optname == IPV6_BINDANY && td != NULL) {
1754 error = priv_check(td,
1755 PRIV_NETINET_BINDANY);
1760 if (optlen != sizeof(int)) {
1764 error = sooptcopyin(sopt, &optval,
1765 sizeof optval, sizeof optval);
1770 case IPV6_UNICAST_HOPS:
1771 if (optval < -1 || optval >= 256)
1774 /* -1 = kernel default */
1775 inp->in6p_hops = optval;
1776 if ((inp->inp_vflag &
1778 inp->inp_ip_ttl = optval;
1781 #define OPTSET(bit) \
1785 inp->inp_flags |= (bit); \
1787 inp->inp_flags &= ~(bit); \
1789 } while (/*CONSTCOND*/ 0)
1790 #define OPTSET2292(bit) \
1793 inp->inp_flags |= IN6P_RFC2292; \
1795 inp->inp_flags |= (bit); \
1797 inp->inp_flags &= ~(bit); \
1799 } while (/*CONSTCOND*/ 0)
1800 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1802 #define OPTSET2_N(bit, val) do { \
1804 inp->inp_flags2 |= bit; \
1806 inp->inp_flags2 &= ~bit; \
1808 #define OPTSET2(bit, val) do { \
1810 OPTSET2_N(bit, val); \
1813 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1814 #define OPTSET2292_EXCLUSIVE(bit) \
1817 if (OPTBIT(IN6P_RFC2292)) { \
1821 inp->inp_flags |= (bit); \
1823 inp->inp_flags &= ~(bit); \
1826 } while (/*CONSTCOND*/ 0)
1828 case IPV6_RECVPKTINFO:
1829 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1834 struct ip6_pktopts **optp;
1836 /* cannot mix with RFC2292 */
1837 if (OPTBIT(IN6P_RFC2292)) {
1842 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1844 return (ECONNRESET);
1846 optp = &inp->in6p_outputopts;
1847 error = ip6_pcbopt(IPV6_HOPLIMIT,
1848 (u_char *)&optval, sizeof(optval),
1849 optp, (td != NULL) ? td->td_ucred :
1855 case IPV6_RECVHOPLIMIT:
1856 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1859 case IPV6_RECVHOPOPTS:
1860 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1863 case IPV6_RECVDSTOPTS:
1864 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1867 case IPV6_RECVRTHDRDSTOPTS:
1868 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1871 case IPV6_RECVRTHDR:
1872 OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1875 case IPV6_RECVPATHMTU:
1877 * We ignore this option for TCP
1879 * (RFC3542 leaves this case
1882 if (uproto != IPPROTO_TCP)
1886 case IPV6_RECVFLOWID:
1887 OPTSET2(INP_RECVFLOWID, optval);
1891 case IPV6_RECVRSSBUCKETID:
1892 OPTSET2(INP_RECVRSSBUCKETID, optval);
1898 if (inp->inp_lport ||
1899 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1901 * The socket is already bound.
1908 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1909 inp->inp_vflag &= ~INP_IPV4;
1911 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1912 inp->inp_vflag |= INP_IPV4;
1916 case IPV6_RECVTCLASS:
1917 /* cannot mix with RFC2292 XXX */
1918 OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1920 case IPV6_AUTOFLOWLABEL:
1921 OPTSET(IN6P_AUTOFLOWLABEL);
1924 case IPV6_ORIGDSTADDR:
1925 OPTSET2(INP_ORIGDSTADDR, optval);
1928 OPTSET(INP_BINDANY);
1931 case IPV6_BINDMULTI:
1932 OPTSET2(INP_BINDMULTI, optval);
1935 case IPV6_RSS_LISTEN_BUCKET:
1936 if ((optval >= 0) &&
1937 (optval < rss_getnumbuckets())) {
1939 inp->inp_rss_listen_bucket = optval;
1940 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1952 case IPV6_USE_MIN_MTU:
1953 case IPV6_PREFER_TEMPADDR:
1954 if (optlen != sizeof(optval)) {
1958 error = sooptcopyin(sopt, &optval,
1959 sizeof optval, sizeof optval);
1963 struct ip6_pktopts **optp;
1965 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1967 return (ECONNRESET);
1969 optp = &inp->in6p_outputopts;
1970 error = ip6_pcbopt(optname,
1971 (u_char *)&optval, sizeof(optval),
1972 optp, (td != NULL) ? td->td_ucred :
1978 case IPV6_2292PKTINFO:
1979 case IPV6_2292HOPLIMIT:
1980 case IPV6_2292HOPOPTS:
1981 case IPV6_2292DSTOPTS:
1982 case IPV6_2292RTHDR:
1984 if (optlen != sizeof(int)) {
1988 error = sooptcopyin(sopt, &optval,
1989 sizeof optval, sizeof optval);
1993 case IPV6_2292PKTINFO:
1994 OPTSET2292(IN6P_PKTINFO);
1996 case IPV6_2292HOPLIMIT:
1997 OPTSET2292(IN6P_HOPLIMIT);
1999 case IPV6_2292HOPOPTS:
2001 * Check super-user privilege.
2002 * See comments for IPV6_RECVHOPOPTS.
2005 error = priv_check(td,
2006 PRIV_NETINET_SETHDROPTS);
2010 OPTSET2292(IN6P_HOPOPTS);
2012 case IPV6_2292DSTOPTS:
2014 error = priv_check(td,
2015 PRIV_NETINET_SETHDROPTS);
2019 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2021 case IPV6_2292RTHDR:
2022 OPTSET2292(IN6P_RTHDR);
2030 case IPV6_RTHDRDSTOPTS:
2033 /* new advanced API (RFC3542) */
2035 u_char optbuf_storage[MCLBYTES];
2037 struct ip6_pktopts **optp;
2039 /* cannot mix with RFC2292 */
2040 if (OPTBIT(IN6P_RFC2292)) {
2046 * We only ensure valsize is not too large
2047 * here. Further validation will be done
2050 error = sooptcopyin(sopt, optbuf_storage,
2051 sizeof(optbuf_storage), 0);
2054 optlen = sopt->sopt_valsize;
2055 optbuf = optbuf_storage;
2057 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2059 return (ECONNRESET);
2061 optp = &inp->in6p_outputopts;
2062 error = ip6_pcbopt(optname, optbuf, optlen,
2063 optp, (td != NULL) ? td->td_ucred : NULL,
2070 case IPV6_MULTICAST_IF:
2071 case IPV6_MULTICAST_HOPS:
2072 case IPV6_MULTICAST_LOOP:
2073 case IPV6_JOIN_GROUP:
2074 case IPV6_LEAVE_GROUP:
2076 case MCAST_BLOCK_SOURCE:
2077 case MCAST_UNBLOCK_SOURCE:
2078 case MCAST_JOIN_GROUP:
2079 case MCAST_LEAVE_GROUP:
2080 case MCAST_JOIN_SOURCE_GROUP:
2081 case MCAST_LEAVE_SOURCE_GROUP:
2082 error = ip6_setmoptions(inp, sopt);
2085 case IPV6_PORTRANGE:
2086 error = sooptcopyin(sopt, &optval,
2087 sizeof optval, sizeof optval);
2093 case IPV6_PORTRANGE_DEFAULT:
2094 inp->inp_flags &= ~(INP_LOWPORT);
2095 inp->inp_flags &= ~(INP_HIGHPORT);
2098 case IPV6_PORTRANGE_HIGH:
2099 inp->inp_flags &= ~(INP_LOWPORT);
2100 inp->inp_flags |= INP_HIGHPORT;
2103 case IPV6_PORTRANGE_LOW:
2104 inp->inp_flags &= ~(INP_HIGHPORT);
2105 inp->inp_flags |= INP_LOWPORT;
2115 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2116 case IPV6_IPSEC_POLICY:
2117 if (IPSEC_ENABLED(ipv6)) {
2118 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2125 error = ENOPROTOOPT;
2133 case IPV6_2292PKTOPTIONS:
2134 #ifdef IPV6_PKTOPTIONS
2135 case IPV6_PKTOPTIONS:
2138 * RFC3542 (effectively) deprecated the
2139 * semantics of the 2292-style pktoptions.
2140 * Since it was not reliable in nature (i.e.,
2141 * applications had to expect the lack of some
2142 * information after all), it would make sense
2143 * to simplify this part by always returning
2146 sopt->sopt_valsize = 0;
2149 case IPV6_RECVHOPOPTS:
2150 case IPV6_RECVDSTOPTS:
2151 case IPV6_RECVRTHDRDSTOPTS:
2152 case IPV6_UNICAST_HOPS:
2153 case IPV6_RECVPKTINFO:
2154 case IPV6_RECVHOPLIMIT:
2155 case IPV6_RECVRTHDR:
2156 case IPV6_RECVPATHMTU:
2159 case IPV6_PORTRANGE:
2160 case IPV6_RECVTCLASS:
2161 case IPV6_AUTOFLOWLABEL:
2165 case IPV6_RECVFLOWID:
2167 case IPV6_RSSBUCKETID:
2168 case IPV6_RECVRSSBUCKETID:
2170 case IPV6_BINDMULTI:
2173 case IPV6_RECVHOPOPTS:
2174 optval = OPTBIT(IN6P_HOPOPTS);
2177 case IPV6_RECVDSTOPTS:
2178 optval = OPTBIT(IN6P_DSTOPTS);
2181 case IPV6_RECVRTHDRDSTOPTS:
2182 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2185 case IPV6_UNICAST_HOPS:
2186 optval = inp->in6p_hops;
2189 case IPV6_RECVPKTINFO:
2190 optval = OPTBIT(IN6P_PKTINFO);
2193 case IPV6_RECVHOPLIMIT:
2194 optval = OPTBIT(IN6P_HOPLIMIT);
2197 case IPV6_RECVRTHDR:
2198 optval = OPTBIT(IN6P_RTHDR);
2201 case IPV6_RECVPATHMTU:
2202 optval = OPTBIT(IN6P_MTU);
2206 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2209 case IPV6_PORTRANGE:
2212 flags = inp->inp_flags;
2213 if (flags & INP_HIGHPORT)
2214 optval = IPV6_PORTRANGE_HIGH;
2215 else if (flags & INP_LOWPORT)
2216 optval = IPV6_PORTRANGE_LOW;
2221 case IPV6_RECVTCLASS:
2222 optval = OPTBIT(IN6P_TCLASS);
2225 case IPV6_AUTOFLOWLABEL:
2226 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2229 case IPV6_ORIGDSTADDR:
2230 optval = OPTBIT2(INP_ORIGDSTADDR);
2234 optval = OPTBIT(INP_BINDANY);
2238 optval = inp->inp_flowid;
2242 optval = inp->inp_flowtype;
2245 case IPV6_RECVFLOWID:
2246 optval = OPTBIT2(INP_RECVFLOWID);
2249 case IPV6_RSSBUCKETID:
2251 rss_hash2bucket(inp->inp_flowid,
2255 optval = rss_bucket;
2260 case IPV6_RECVRSSBUCKETID:
2261 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2265 case IPV6_BINDMULTI:
2266 optval = OPTBIT2(INP_BINDMULTI);
2272 error = sooptcopyout(sopt, &optval,
2279 struct ip6_mtuinfo mtuinfo;
2280 struct in6_addr addr;
2282 if (!(so->so_state & SS_ISCONNECTED))
2285 * XXX: we dot not consider the case of source
2286 * routing, or optional information to specify
2287 * the outgoing interface.
2288 * Copy faddr out of inp to avoid holding lock
2289 * on inp during route lookup.
2292 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2294 error = ip6_getpmtu_ctl(so->so_fibnum,
2298 if (pmtu > IPV6_MAXPACKET)
2299 pmtu = IPV6_MAXPACKET;
2301 bzero(&mtuinfo, sizeof(mtuinfo));
2302 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2303 optdata = (void *)&mtuinfo;
2304 optdatalen = sizeof(mtuinfo);
2305 error = sooptcopyout(sopt, optdata,
2310 case IPV6_2292PKTINFO:
2311 case IPV6_2292HOPLIMIT:
2312 case IPV6_2292HOPOPTS:
2313 case IPV6_2292RTHDR:
2314 case IPV6_2292DSTOPTS:
2316 case IPV6_2292PKTINFO:
2317 optval = OPTBIT(IN6P_PKTINFO);
2319 case IPV6_2292HOPLIMIT:
2320 optval = OPTBIT(IN6P_HOPLIMIT);
2322 case IPV6_2292HOPOPTS:
2323 optval = OPTBIT(IN6P_HOPOPTS);
2325 case IPV6_2292RTHDR:
2326 optval = OPTBIT(IN6P_RTHDR);
2328 case IPV6_2292DSTOPTS:
2329 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2332 error = sooptcopyout(sopt, &optval,
2339 case IPV6_RTHDRDSTOPTS:
2343 case IPV6_USE_MIN_MTU:
2344 case IPV6_PREFER_TEMPADDR:
2345 error = ip6_getpcbopt(inp, optname, sopt);
2348 case IPV6_MULTICAST_IF:
2349 case IPV6_MULTICAST_HOPS:
2350 case IPV6_MULTICAST_LOOP:
2352 error = ip6_getmoptions(inp, sopt);
2355 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2356 case IPV6_IPSEC_POLICY:
2357 if (IPSEC_ENABLED(ipv6)) {
2358 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2364 error = ENOPROTOOPT;
2374 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2376 int error = 0, optval, optlen;
2377 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2378 struct inpcb *inp = sotoinpcb(so);
2379 int level, op, optname;
2381 level = sopt->sopt_level;
2382 op = sopt->sopt_dir;
2383 optname = sopt->sopt_name;
2384 optlen = sopt->sopt_valsize;
2386 if (level != IPPROTO_IPV6) {
2393 * For ICMPv6 sockets, no modification allowed for checksum
2394 * offset, permit "no change" values to help existing apps.
2396 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2397 * for an ICMPv6 socket will fail."
2398 * The current behavior does not meet RFC3542.
2402 if (optlen != sizeof(int)) {
2406 error = sooptcopyin(sopt, &optval, sizeof(optval),
2410 if (optval < -1 || (optval % 2) != 0) {
2412 * The API assumes non-negative even offset
2413 * values or -1 as a special value.
2416 } else if (so->so_proto->pr_protocol ==
2418 if (optval != icmp6off)
2421 inp->in6p_cksum = optval;
2425 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2428 optval = inp->in6p_cksum;
2430 error = sooptcopyout(sopt, &optval, sizeof(optval));
2440 error = ENOPROTOOPT;
2448 * Set up IP6 options in pcb for insertion in output packets or
2449 * specifying behavior of outgoing packets.
2452 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2453 struct socket *so, struct sockopt *sopt)
2455 struct ip6_pktopts *opt = *pktopt;
2457 struct thread *td = sopt->sopt_td;
2459 /* turn off any old options. */
2462 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2463 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2464 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2465 printf("ip6_pcbopts: all specified options are cleared.\n");
2467 ip6_clearpktopts(opt, -1);
2469 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2475 if (!m || m->m_len == 0) {
2477 * Only turning off any previous options, regardless of
2478 * whether the opt is just created or given.
2480 free(opt, M_IP6OPT);
2484 /* set options specified by user. */
2485 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2486 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2487 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2488 free(opt, M_IP6OPT);
2496 * initialize ip6_pktopts. beware that there are non-zero default values in
2500 ip6_initpktopts(struct ip6_pktopts *opt)
2503 bzero(opt, sizeof(*opt));
2504 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2505 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2506 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2507 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2511 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2512 struct ucred *cred, int uproto)
2514 struct ip6_pktopts *opt;
2516 if (*pktopt == NULL) {
2517 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2519 if (*pktopt == NULL)
2521 ip6_initpktopts(*pktopt);
2525 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2528 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2529 if (pktopt && pktopt->field) { \
2531 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2532 malloc_optdata = true; \
2534 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
2536 free(optdata, M_TEMP); \
2537 return (ECONNRESET); \
2539 pktopt = inp->in6p_outputopts; \
2540 if (pktopt && pktopt->field) { \
2541 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2542 bcopy(&pktopt->field, optdata, optdatalen); \
2544 free(optdata, M_TEMP); \
2546 malloc_optdata = false; \
2551 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2552 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2554 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2555 pktopt->field->sa_len)
2558 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2560 void *optdata = NULL;
2561 bool malloc_optdata = false;
2564 struct in6_pktinfo null_pktinfo;
2565 int deftclass = 0, on;
2566 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2567 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2568 struct ip6_pktopts *pktopt;
2571 pktopt = inp->in6p_outputopts;
2575 optdata = (void *)&null_pktinfo;
2576 if (pktopt && pktopt->ip6po_pktinfo) {
2577 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2578 sizeof(null_pktinfo));
2579 in6_clearscope(&null_pktinfo.ipi6_addr);
2581 /* XXX: we don't have to do this every time... */
2582 bzero(&null_pktinfo, sizeof(null_pktinfo));
2584 optdatalen = sizeof(struct in6_pktinfo);
2587 if (pktopt && pktopt->ip6po_tclass >= 0)
2588 deftclass = pktopt->ip6po_tclass;
2589 optdata = (void *)&deftclass;
2590 optdatalen = sizeof(int);
2593 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2596 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2598 case IPV6_RTHDRDSTOPTS:
2599 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2602 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2605 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2607 case IPV6_USE_MIN_MTU:
2609 defminmtu = pktopt->ip6po_minmtu;
2610 optdata = (void *)&defminmtu;
2611 optdatalen = sizeof(int);
2614 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2618 optdata = (void *)&on;
2619 optdatalen = sizeof(on);
2621 case IPV6_PREFER_TEMPADDR:
2623 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2624 optdata = (void *)&defpreftemp;
2625 optdatalen = sizeof(int);
2627 default: /* should not happen */
2629 panic("ip6_getpcbopt: unexpected option\n");
2632 return (ENOPROTOOPT);
2636 error = sooptcopyout(sopt, optdata, optdatalen);
2638 free(optdata, M_TEMP);
2644 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2649 if (optname == -1 || optname == IPV6_PKTINFO) {
2650 if (pktopt->ip6po_pktinfo)
2651 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2652 pktopt->ip6po_pktinfo = NULL;
2654 if (optname == -1 || optname == IPV6_HOPLIMIT)
2655 pktopt->ip6po_hlim = -1;
2656 if (optname == -1 || optname == IPV6_TCLASS)
2657 pktopt->ip6po_tclass = -1;
2658 if (optname == -1 || optname == IPV6_NEXTHOP) {
2659 if (pktopt->ip6po_nextroute.ro_nh) {
2660 NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2661 pktopt->ip6po_nextroute.ro_nh = NULL;
2663 if (pktopt->ip6po_nexthop)
2664 free(pktopt->ip6po_nexthop, M_IP6OPT);
2665 pktopt->ip6po_nexthop = NULL;
2667 if (optname == -1 || optname == IPV6_HOPOPTS) {
2668 if (pktopt->ip6po_hbh)
2669 free(pktopt->ip6po_hbh, M_IP6OPT);
2670 pktopt->ip6po_hbh = NULL;
2672 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2673 if (pktopt->ip6po_dest1)
2674 free(pktopt->ip6po_dest1, M_IP6OPT);
2675 pktopt->ip6po_dest1 = NULL;
2677 if (optname == -1 || optname == IPV6_RTHDR) {
2678 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2679 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2680 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2681 if (pktopt->ip6po_route.ro_nh) {
2682 NH_FREE(pktopt->ip6po_route.ro_nh);
2683 pktopt->ip6po_route.ro_nh = NULL;
2686 if (optname == -1 || optname == IPV6_DSTOPTS) {
2687 if (pktopt->ip6po_dest2)
2688 free(pktopt->ip6po_dest2, M_IP6OPT);
2689 pktopt->ip6po_dest2 = NULL;
2693 #define PKTOPT_EXTHDRCPY(type) \
2696 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2697 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2698 if (dst->type == NULL)\
2700 bcopy(src->type, dst->type, hlen);\
2702 } while (/*CONSTCOND*/ 0)
2705 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2707 if (dst == NULL || src == NULL) {
2708 printf("ip6_clearpktopts: invalid argument\n");
2712 dst->ip6po_hlim = src->ip6po_hlim;
2713 dst->ip6po_tclass = src->ip6po_tclass;
2714 dst->ip6po_flags = src->ip6po_flags;
2715 dst->ip6po_minmtu = src->ip6po_minmtu;
2716 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2717 if (src->ip6po_pktinfo) {
2718 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2720 if (dst->ip6po_pktinfo == NULL)
2722 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2724 if (src->ip6po_nexthop) {
2725 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2727 if (dst->ip6po_nexthop == NULL)
2729 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2730 src->ip6po_nexthop->sa_len);
2732 PKTOPT_EXTHDRCPY(ip6po_hbh);
2733 PKTOPT_EXTHDRCPY(ip6po_dest1);
2734 PKTOPT_EXTHDRCPY(ip6po_dest2);
2735 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2739 ip6_clearpktopts(dst, -1);
2742 #undef PKTOPT_EXTHDRCPY
2744 struct ip6_pktopts *
2745 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2748 struct ip6_pktopts *dst;
2750 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2753 ip6_initpktopts(dst);
2755 if ((error = copypktopts(dst, src, canwait)) != 0) {
2756 free(dst, M_IP6OPT);
2764 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2769 ip6_clearpktopts(pktopt, -1);
2771 free(pktopt, M_IP6OPT);
2775 * Set IPv6 outgoing packet options based on advanced API.
2778 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2779 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2781 struct cmsghdr *cm = NULL;
2783 if (control == NULL || opt == NULL)
2786 ip6_initpktopts(opt);
2791 * If stickyopt is provided, make a local copy of the options
2792 * for this particular packet, then override them by ancillary
2794 * XXX: copypktopts() does not copy the cached route to a next
2795 * hop (if any). This is not very good in terms of efficiency,
2796 * but we can allow this since this option should be rarely
2799 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2804 * XXX: Currently, we assume all the optional information is stored
2807 if (control->m_next)
2810 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2811 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2814 if (control->m_len < CMSG_LEN(0))
2817 cm = mtod(control, struct cmsghdr *);
2818 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2820 if (cm->cmsg_level != IPPROTO_IPV6)
2823 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2824 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2833 * Set a particular packet option, as a sticky option or an ancillary data
2834 * item. "len" can be 0 only when it's a sticky option.
2835 * We have 4 cases of combination of "sticky" and "cmsg":
2836 * "sticky=0, cmsg=0": impossible
2837 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2838 * "sticky=1, cmsg=0": RFC3542 socket option
2839 * "sticky=1, cmsg=1": RFC2292 socket option
2842 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2843 struct ucred *cred, int sticky, int cmsg, int uproto)
2845 int minmtupolicy, preftemp;
2848 if (!sticky && !cmsg) {
2850 printf("ip6_setpktopt: impossible case\n");
2856 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2857 * not be specified in the context of RFC3542. Conversely,
2858 * RFC3542 types should not be specified in the context of RFC2292.
2862 case IPV6_2292PKTINFO:
2863 case IPV6_2292HOPLIMIT:
2864 case IPV6_2292NEXTHOP:
2865 case IPV6_2292HOPOPTS:
2866 case IPV6_2292DSTOPTS:
2867 case IPV6_2292RTHDR:
2868 case IPV6_2292PKTOPTIONS:
2869 return (ENOPROTOOPT);
2872 if (sticky && cmsg) {
2879 case IPV6_RTHDRDSTOPTS:
2881 case IPV6_USE_MIN_MTU:
2884 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2885 return (ENOPROTOOPT);
2890 case IPV6_2292PKTINFO:
2893 struct ifnet *ifp = NULL;
2894 struct in6_pktinfo *pktinfo;
2896 if (len != sizeof(struct in6_pktinfo))
2899 pktinfo = (struct in6_pktinfo *)buf;
2902 * An application can clear any sticky IPV6_PKTINFO option by
2903 * doing a "regular" setsockopt with ipi6_addr being
2904 * in6addr_any and ipi6_ifindex being zero.
2905 * [RFC 3542, Section 6]
2907 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2908 pktinfo->ipi6_ifindex == 0 &&
2909 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2910 ip6_clearpktopts(opt, optname);
2914 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2915 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2918 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2920 /* validate the interface index if specified. */
2921 if (pktinfo->ipi6_ifindex > V_if_index)
2923 if (pktinfo->ipi6_ifindex) {
2924 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2928 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2929 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2933 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2934 struct in6_ifaddr *ia;
2936 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2937 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2939 return (EADDRNOTAVAIL);
2940 ifa_free(&ia->ia_ifa);
2943 * We store the address anyway, and let in6_selectsrc()
2944 * validate the specified address. This is because ipi6_addr
2945 * may not have enough information about its scope zone, and
2946 * we may need additional information (such as outgoing
2947 * interface or the scope zone of a destination address) to
2948 * disambiguate the scope.
2949 * XXX: the delay of the validation may confuse the
2950 * application when it is used as a sticky option.
2952 if (opt->ip6po_pktinfo == NULL) {
2953 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2954 M_IP6OPT, M_NOWAIT);
2955 if (opt->ip6po_pktinfo == NULL)
2958 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2962 case IPV6_2292HOPLIMIT:
2968 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2969 * to simplify the ordering among hoplimit options.
2971 if (optname == IPV6_HOPLIMIT && sticky)
2972 return (ENOPROTOOPT);
2974 if (len != sizeof(int))
2977 if (*hlimp < -1 || *hlimp > 255)
2980 opt->ip6po_hlim = *hlimp;
2988 if (len != sizeof(int))
2990 tclass = *(int *)buf;
2991 if (tclass < -1 || tclass > 255)
2994 opt->ip6po_tclass = tclass;
2998 case IPV6_2292NEXTHOP:
3001 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3006 if (len == 0) { /* just remove the option */
3007 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3011 /* check if cmsg_len is large enough for sa_len */
3012 if (len < sizeof(struct sockaddr) || len < *buf)
3015 switch (((struct sockaddr *)buf)->sa_family) {
3018 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3021 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3024 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3025 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3028 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3034 case AF_LINK: /* should eventually be supported */
3036 return (EAFNOSUPPORT);
3039 /* turn off the previous option, then set the new option. */
3040 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3041 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3042 if (opt->ip6po_nexthop == NULL)
3044 bcopy(buf, opt->ip6po_nexthop, *buf);
3047 case IPV6_2292HOPOPTS:
3050 struct ip6_hbh *hbh;
3054 * XXX: We don't allow a non-privileged user to set ANY HbH
3055 * options, since per-option restriction has too much
3059 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3065 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3066 break; /* just remove the option */
3069 /* message length validation */
3070 if (len < sizeof(struct ip6_hbh))
3072 hbh = (struct ip6_hbh *)buf;
3073 hbhlen = (hbh->ip6h_len + 1) << 3;
3077 /* turn off the previous option, then set the new option. */
3078 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3079 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3080 if (opt->ip6po_hbh == NULL)
3082 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3087 case IPV6_2292DSTOPTS:
3089 case IPV6_RTHDRDSTOPTS:
3091 struct ip6_dest *dest, **newdest = NULL;
3094 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3095 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3101 ip6_clearpktopts(opt, optname);
3102 break; /* just remove the option */
3105 /* message length validation */
3106 if (len < sizeof(struct ip6_dest))
3108 dest = (struct ip6_dest *)buf;
3109 destlen = (dest->ip6d_len + 1) << 3;
3114 * Determine the position that the destination options header
3115 * should be inserted; before or after the routing header.
3118 case IPV6_2292DSTOPTS:
3120 * The old advacned API is ambiguous on this point.
3121 * Our approach is to determine the position based
3122 * according to the existence of a routing header.
3123 * Note, however, that this depends on the order of the
3124 * extension headers in the ancillary data; the 1st
3125 * part of the destination options header must appear
3126 * before the routing header in the ancillary data,
3128 * RFC3542 solved the ambiguity by introducing
3129 * separate ancillary data or option types.
3131 if (opt->ip6po_rthdr == NULL)
3132 newdest = &opt->ip6po_dest1;
3134 newdest = &opt->ip6po_dest2;
3136 case IPV6_RTHDRDSTOPTS:
3137 newdest = &opt->ip6po_dest1;
3140 newdest = &opt->ip6po_dest2;
3144 /* turn off the previous option, then set the new option. */
3145 ip6_clearpktopts(opt, optname);
3146 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3147 if (*newdest == NULL)
3149 bcopy(dest, *newdest, destlen);
3154 case IPV6_2292RTHDR:
3157 struct ip6_rthdr *rth;
3161 ip6_clearpktopts(opt, IPV6_RTHDR);
3162 break; /* just remove the option */
3165 /* message length validation */
3166 if (len < sizeof(struct ip6_rthdr))
3168 rth = (struct ip6_rthdr *)buf;
3169 rthlen = (rth->ip6r_len + 1) << 3;
3173 switch (rth->ip6r_type) {
3174 case IPV6_RTHDR_TYPE_0:
3175 if (rth->ip6r_len == 0) /* must contain one addr */
3177 if (rth->ip6r_len % 2) /* length must be even */
3179 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3183 return (EINVAL); /* not supported */
3186 /* turn off the previous option */
3187 ip6_clearpktopts(opt, IPV6_RTHDR);
3188 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3189 if (opt->ip6po_rthdr == NULL)
3191 bcopy(rth, opt->ip6po_rthdr, rthlen);
3196 case IPV6_USE_MIN_MTU:
3197 if (len != sizeof(int))
3199 minmtupolicy = *(int *)buf;
3200 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3201 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3202 minmtupolicy != IP6PO_MINMTU_ALL) {
3205 opt->ip6po_minmtu = minmtupolicy;
3209 if (len != sizeof(int))
3212 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3214 * we ignore this option for TCP sockets.
3215 * (RFC3542 leaves this case unspecified.)
3217 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3219 opt->ip6po_flags |= IP6PO_DONTFRAG;
3222 case IPV6_PREFER_TEMPADDR:
3223 if (len != sizeof(int))
3225 preftemp = *(int *)buf;
3226 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3227 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3228 preftemp != IP6PO_TEMPADDR_PREFER) {
3231 opt->ip6po_prefer_tempaddr = preftemp;
3235 return (ENOPROTOOPT);
3236 } /* end of switch */
3242 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3243 * packet to the input queue of a specified interface. Note that this
3244 * calls the output routine of the loopback "driver", but with an interface
3245 * pointer that might NOT be &loif -- easier than replicating that code here.
3248 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3251 struct ip6_hdr *ip6;
3253 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3258 * Make sure to deep-copy IPv6 header portion in case the data
3259 * is in an mbuf cluster, so that we can safely override the IPv6
3260 * header portion later.
3262 if (!M_WRITABLE(copym) ||
3263 copym->m_len < sizeof(struct ip6_hdr)) {
3264 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3268 ip6 = mtod(copym, struct ip6_hdr *);
3270 * clear embedded scope identifiers if necessary.
3271 * in6_clearscope will touch the addresses only when necessary.
3273 in6_clearscope(&ip6->ip6_src);
3274 in6_clearscope(&ip6->ip6_dst);
3275 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3276 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3278 copym->m_pkthdr.csum_data = 0xffff;
3280 if_simloop(ifp, copym, AF_INET6, 0);
3284 * Chop IPv6 header off from the payload.
3287 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3290 struct ip6_hdr *ip6;
3292 ip6 = mtod(m, struct ip6_hdr *);
3293 if (m->m_len > sizeof(*ip6)) {
3294 mh = m_gethdr(M_NOWAIT, MT_DATA);
3299 m_move_pkthdr(mh, m);
3300 M_ALIGN(mh, sizeof(*ip6));
3301 m->m_len -= sizeof(*ip6);
3302 m->m_data += sizeof(*ip6);
3305 m->m_len = sizeof(*ip6);
3306 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3308 exthdrs->ip6e_ip6 = m;
3313 * Compute IPv6 extension header length.
3316 ip6_optlen(struct inpcb *inp)
3320 if (!inp->in6p_outputopts)
3325 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3327 len += elen(inp->in6p_outputopts->ip6po_hbh);
3328 if (inp->in6p_outputopts->ip6po_rthdr)
3329 /* dest1 is valid with rthdr only */
3330 len += elen(inp->in6p_outputopts->ip6po_dest1);
3331 len += elen(inp->in6p_outputopts->ip6po_rthdr);
3332 len += elen(inp->in6p_outputopts->ip6po_dest2);