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>
117 #if defined(SCTP) || defined(SCTP_SUPPORT)
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) ||
224 #if defined(SCTP) || defined(SCTP_SUPPORT)
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;
241 #if defined(SCTP) || defined(SCTP_SUPPORT)
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 nhop_object *nh;
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);
765 error = EHOSTUNREACH;
768 *dst = dst_sa; /* XXX */
773 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
775 IP6STAT_INC(ip6s_noroute);
776 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
777 error = EHOSTUNREACH;;
783 ia = ifatoia6(nh->nh_ifa);
784 if (nh->nh_flags & NHF_GATEWAY)
785 dst->sin6_addr = nh->gw6_sa.sin6_addr;
790 /* Then nh (for unicast) and ifp must be non-NULL valid values. */
791 if ((flags & IPV6_FORWARDING) == 0) {
792 /* XXX: the FORWARDING flag can be set for mrouting. */
793 in6_ifstat_inc(ifp, ifs6_out_request);
796 /* Setup data structures for scope ID checks. */
798 bzero(&src_sa, sizeof(src_sa));
799 src_sa.sin6_family = AF_INET6;
800 src_sa.sin6_len = sizeof(src_sa);
801 src_sa.sin6_addr = ip6->ip6_src;
804 /* Re-initialize to be sure. */
805 bzero(&dst_sa, sizeof(dst_sa));
806 dst_sa.sin6_family = AF_INET6;
807 dst_sa.sin6_len = sizeof(dst_sa);
808 dst_sa.sin6_addr = ip6->ip6_dst;
810 /* Check for valid scope ID. */
811 if (in6_setscope(&src0, ifp, &zone) == 0 &&
812 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
813 in6_setscope(&dst0, ifp, &zone) == 0 &&
814 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
816 * The outgoing interface is in the zone of the source
817 * and destination addresses.
819 * Because the loopback interface cannot receive
820 * packets with a different scope ID than its own,
821 * there is a trick to pretend the outgoing packet
822 * was received by the real network interface, by
823 * setting "origifp" different from "ifp". This is
824 * only allowed when "ifp" is a loopback network
825 * interface. Refer to code in nd6_output_ifp() for
831 * We should use ia_ifp to support the case of sending
832 * packets to an address of our own.
834 if (ia != NULL && ia->ia_ifp)
837 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
838 sa6_recoverscope(&src_sa) != 0 ||
839 sa6_recoverscope(&dst_sa) != 0 ||
840 dst_sa.sin6_scope_id == 0 ||
841 (src_sa.sin6_scope_id != 0 &&
842 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
843 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
845 * If the destination network interface is not a
846 * loopback interface, or the destination network
847 * address has no scope ID, or the source address has
848 * a scope ID set which is different from the
849 * destination address one, or there is no network
850 * interface representing this scope ID, the address
851 * pair is considered invalid.
853 IP6STAT_INC(ip6s_badscope);
854 in6_ifstat_inc(ifp, ifs6_out_discard);
856 error = EHOSTUNREACH; /* XXX */
859 /* All scope ID checks are successful. */
861 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
862 if (opt && opt->ip6po_nextroute.ro_nh) {
864 * The nexthop is explicitly specified by the
865 * application. We assume the next hop is an IPv6
868 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
870 else if ((nh->nh_flags & NHF_GATEWAY))
874 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
875 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
877 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
878 in6_ifstat_inc(ifp, ifs6_out_mcast);
880 /* Confirm that the outgoing interface supports multicast. */
881 if (!(ifp->if_flags & IFF_MULTICAST)) {
882 IP6STAT_INC(ip6s_noroute);
883 in6_ifstat_inc(ifp, ifs6_out_discard);
887 if ((im6o == NULL && in6_mcast_loop) ||
888 (im6o && im6o->im6o_multicast_loop)) {
890 * Loop back multicast datagram if not expressly
891 * forbidden to do so, even if we have not joined
892 * the address; protocols will filter it later,
893 * thus deferring a hash lookup and lock acquisition
894 * at the expense of an m_copym().
896 ip6_mloopback(ifp, m);
899 * If we are acting as a multicast router, perform
900 * multicast forwarding as if the packet had just
901 * arrived on the interface to which we are about
902 * to send. The multicast forwarding function
903 * recursively calls this function, using the
904 * IPV6_FORWARDING flag to prevent infinite recursion.
906 * Multicasts that are looped back by ip6_mloopback(),
907 * above, will be forwarded by the ip6_input() routine,
910 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
912 * XXX: ip6_mforward expects that rcvif is NULL
913 * when it is called from the originating path.
914 * However, it may not always be the case.
916 m->m_pkthdr.rcvif = NULL;
917 if (ip6_mforward(ip6, ifp, m) != 0) {
924 * Multicasts with a hoplimit of zero may be looped back,
925 * above, but must not be transmitted on a network.
926 * Also, multicasts addressed to the loopback interface
927 * are not sent -- the above call to ip6_mloopback() will
928 * loop back a copy if this host actually belongs to the
929 * destination group on the loopback interface.
931 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
932 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
939 * Fill the outgoing inteface to tell the upper layer
940 * to increment per-interface statistics.
945 /* Determine path MTU. */
946 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
947 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
949 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
950 "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
951 ifp, alwaysfrag, fibnum));
954 * The caller of this function may specify to use the minimum MTU
956 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
957 * setting. The logic is a bit complicated; by default, unicast
958 * packets will follow path MTU while multicast packets will be sent at
959 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
960 * including unicast ones will be sent at the minimum MTU. Multicast
961 * packets will always be sent at the minimum MTU unless
962 * IP6PO_MINMTU_DISABLE is explicitly specified.
963 * See RFC 3542 for more details.
965 if (mtu > IPV6_MMTU) {
966 if ((flags & IPV6_MINMTU))
968 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
970 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
972 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
978 * Clear embedded scope identifiers if necessary.
979 * in6_clearscope() will touch the addresses only when necessary.
981 in6_clearscope(&ip6->ip6_src);
982 in6_clearscope(&ip6->ip6_dst);
985 * If the outgoing packet contains a hop-by-hop options header,
986 * it must be examined and processed even by the source node.
987 * (RFC 2460, section 4.)
989 if (exthdrs.ip6e_hbh) {
990 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
991 u_int32_t dummy; /* XXX unused */
992 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
995 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
996 panic("ip6e_hbh is not contiguous");
999 * XXX: if we have to send an ICMPv6 error to the sender,
1000 * we need the M_LOOP flag since icmp6_error() expects
1001 * the IPv6 and the hop-by-hop options header are
1002 * contiguous unless the flag is set.
1004 m->m_flags |= M_LOOP;
1005 m->m_pkthdr.rcvif = ifp;
1006 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1007 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1008 &dummy, &plen) < 0) {
1009 /* m was already freed at this point. */
1010 error = EINVAL;/* better error? */
1013 m->m_flags &= ~M_LOOP; /* XXX */
1014 m->m_pkthdr.rcvif = NULL;
1017 /* Jump over all PFIL processing if hooks are not active. */
1018 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1021 odst = ip6->ip6_dst;
1022 /* Run through list of hooks for output packets. */
1023 switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1025 ip6 = mtod(m, struct ip6_hdr *);
1035 /* See if destination IP address was changed by packet filter. */
1036 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1037 m->m_flags |= M_SKIP_FIREWALL;
1038 /* If destination is now ourself drop to ip6_input(). */
1039 if (in6_localip(&ip6->ip6_dst)) {
1040 m->m_flags |= M_FASTFWD_OURS;
1041 if (m->m_pkthdr.rcvif == NULL)
1042 m->m_pkthdr.rcvif = V_loif;
1043 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1044 m->m_pkthdr.csum_flags |=
1045 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1046 m->m_pkthdr.csum_data = 0xffff;
1048 #if defined(SCTP) || defined(SCTP_SUPPORT)
1049 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1050 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1052 error = netisr_queue(NETISR_IPV6, m);
1056 RO_INVALIDATE_CACHE(ro);
1057 needfiblookup = 1; /* Redo the routing table lookup. */
1060 /* See if fib was changed by packet filter. */
1061 if (fibnum != M_GETFIB(m)) {
1062 m->m_flags |= M_SKIP_FIREWALL;
1063 fibnum = M_GETFIB(m);
1065 RO_INVALIDATE_CACHE(ro);
1071 /* See if local, if yes, send it to netisr. */
1072 if (m->m_flags & M_FASTFWD_OURS) {
1073 if (m->m_pkthdr.rcvif == NULL)
1074 m->m_pkthdr.rcvif = V_loif;
1075 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1076 m->m_pkthdr.csum_flags |=
1077 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1078 m->m_pkthdr.csum_data = 0xffff;
1080 #if defined(SCTP) || defined(SCTP_SUPPORT)
1081 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1082 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1084 error = netisr_queue(NETISR_IPV6, m);
1087 /* Or forward to some other address? */
1088 if ((m->m_flags & M_IP6_NEXTHOP) &&
1089 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1091 dst = (struct sockaddr_in6 *)&ro->ro_dst;
1094 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1095 m->m_flags |= M_SKIP_FIREWALL;
1096 m->m_flags &= ~M_IP6_NEXTHOP;
1097 m_tag_delete(m, fwd_tag);
1103 * Send the packet to the outgoing interface.
1104 * If necessary, do IPv6 fragmentation before sending.
1106 * The logic here is rather complex:
1107 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1108 * 1-a: send as is if tlen <= path mtu
1109 * 1-b: fragment if tlen > path mtu
1111 * 2: if user asks us not to fragment (dontfrag == 1)
1112 * 2-a: send as is if tlen <= interface mtu
1113 * 2-b: error if tlen > interface mtu
1115 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1118 * 4: if dontfrag == 1 && alwaysfrag == 1
1119 * error, as we cannot handle this conflicting request.
1121 sw_csum = m->m_pkthdr.csum_flags;
1123 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1124 sw_csum &= ~ifp->if_hwassist;
1128 * If we added extension headers, we will not do TSO and calculate the
1129 * checksums ourselves for now.
1130 * XXX-BZ Need a framework to know when the NIC can handle it, even
1133 error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1136 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1137 tlen = m->m_pkthdr.len;
1139 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1143 if (dontfrag && alwaysfrag) { /* Case 4. */
1144 /* Conflicting request - can't transmit. */
1148 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */
1150 * Even if the DONTFRAG option is specified, we cannot send the
1151 * packet when the data length is larger than the MTU of the
1152 * outgoing interface.
1153 * Notify the error by sending IPV6_PATHMTU ancillary data if
1154 * application wanted to know the MTU value. Also return an
1155 * error code (this is not described in the API spec).
1158 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1163 /* Transmit packet without fragmentation. */
1164 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1165 struct in6_ifaddr *ia6;
1167 ip6 = mtod(m, struct ip6_hdr *);
1168 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1170 /* Record statistics for this interface address. */
1171 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1172 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1174 ifa_free(&ia6->ia_ifa);
1176 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1177 (flags & IP_NO_SND_TAG_RL) ? false : true);
1181 /* Try to fragment the packet. Cases 1-b and 3. */
1182 if (mtu < IPV6_MMTU) {
1183 /* Path MTU cannot be less than IPV6_MMTU. */
1185 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1187 } else if (ip6->ip6_plen == 0) {
1188 /* Jumbo payload cannot be fragmented. */
1190 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1196 * Too large for the destination or interface;
1197 * fragment if possible.
1198 * Must be able to put at least 8 bytes per fragment.
1200 if (mtu > IPV6_MAXPACKET)
1201 mtu = IPV6_MAXPACKET;
1203 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1206 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1211 * If the interface will not calculate checksums on
1212 * fragmented packets, then do it here.
1213 * XXX-BZ handle the hw offloading case. Need flags.
1215 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1216 plen, optlen, true);
1221 * Change the next header field of the last header in the
1222 * unfragmentable part.
1224 if (exthdrs.ip6e_rthdr) {
1225 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1226 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1227 } else if (exthdrs.ip6e_dest1) {
1228 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1229 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1230 } else if (exthdrs.ip6e_hbh) {
1231 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1232 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1234 ip6 = mtod(m, struct ip6_hdr *);
1235 nextproto = ip6->ip6_nxt;
1236 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1240 * Loop through length of segment after first fragment,
1241 * make new header and copy data of each part and link onto
1245 id = htonl(ip6_randomid());
1246 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1250 in6_ifstat_inc(ifp, ifs6_out_fragok);
1253 /* Remove leading garbage. */
1262 /* Record statistics for this interface address. */
1264 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1265 counter_u64_add(ia->ia_ifa.ifa_obytes,
1268 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1275 IP6STAT_INC(ip6s_fragmented);
1281 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1282 m_freem(exthdrs.ip6e_dest1);
1283 m_freem(exthdrs.ip6e_rthdr);
1284 m_freem(exthdrs.ip6e_dest2);
1293 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1297 if (hlen > MCLBYTES)
1298 return (ENOBUFS); /* XXX */
1301 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1303 m = m_get(M_NOWAIT, MT_DATA);
1308 bcopy(hdr, mtod(m, caddr_t), hlen);
1315 * Insert jumbo payload option.
1318 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1324 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1327 * If there is no hop-by-hop options header, allocate new one.
1328 * If there is one but it doesn't have enough space to store the
1329 * jumbo payload option, allocate a cluster to store the whole options.
1330 * Otherwise, use it to store the options.
1332 if (exthdrs->ip6e_hbh == NULL) {
1333 mopt = m_get(M_NOWAIT, MT_DATA);
1336 mopt->m_len = JUMBOOPTLEN;
1337 optbuf = mtod(mopt, u_char *);
1338 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1339 exthdrs->ip6e_hbh = mopt;
1341 struct ip6_hbh *hbh;
1343 mopt = exthdrs->ip6e_hbh;
1344 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1347 * - exthdrs->ip6e_hbh is not referenced from places
1348 * other than exthdrs.
1349 * - exthdrs->ip6e_hbh is not an mbuf chain.
1351 int oldoptlen = mopt->m_len;
1355 * XXX: give up if the whole (new) hbh header does
1356 * not fit even in an mbuf cluster.
1358 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1362 * As a consequence, we must always prepare a cluster
1365 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1368 n->m_len = oldoptlen + JUMBOOPTLEN;
1369 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1371 optbuf = mtod(n, caddr_t) + oldoptlen;
1373 mopt = exthdrs->ip6e_hbh = n;
1375 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1376 mopt->m_len += JUMBOOPTLEN;
1378 optbuf[0] = IP6OPT_PADN;
1382 * Adjust the header length according to the pad and
1383 * the jumbo payload option.
1385 hbh = mtod(mopt, struct ip6_hbh *);
1386 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1389 /* fill in the option. */
1390 optbuf[2] = IP6OPT_JUMBO;
1392 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1393 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1395 /* finally, adjust the packet header length */
1396 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1403 * Insert fragment header and copy unfragmentable header portions.
1406 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1407 struct ip6_frag **frghdrp)
1409 struct mbuf *n, *mlast;
1411 if (hlen > sizeof(struct ip6_hdr)) {
1412 n = m_copym(m0, sizeof(struct ip6_hdr),
1413 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1420 /* Search for the last mbuf of unfragmentable part. */
1421 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1424 if (M_WRITABLE(mlast) &&
1425 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1426 /* use the trailing space of the last mbuf for the fragment hdr */
1427 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1429 mlast->m_len += sizeof(struct ip6_frag);
1430 m->m_pkthdr.len += sizeof(struct ip6_frag);
1432 /* allocate a new mbuf for the fragment header */
1435 mfrg = m_get(M_NOWAIT, MT_DATA);
1438 mfrg->m_len = sizeof(struct ip6_frag);
1439 *frghdrp = mtod(mfrg, struct ip6_frag *);
1440 mlast->m_next = mfrg;
1447 * Calculates IPv6 path mtu for destination @dst.
1448 * Resulting MTU is stored in @mtup.
1450 * Returns 0 on success.
1453 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1455 struct epoch_tracker et;
1456 struct nhop_object *nh;
1457 struct in6_addr kdst;
1461 in6_splitscope(dst, &kdst, &scopeid);
1463 NET_EPOCH_ENTER(et);
1464 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1466 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1468 error = EHOSTUNREACH;
1475 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1476 * and cached data in @ro_pmtu.
1477 * MTU from (successful) route lookup is saved (along with dst)
1478 * inside @ro_pmtu to avoid subsequent route lookups after packet
1479 * filter processing.
1481 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1482 * Returns 0 on success.
1485 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1486 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1487 int *alwaysfragp, u_int fibnum, u_int proto)
1489 struct nhop_object *nh;
1490 struct in6_addr kdst;
1492 struct sockaddr_in6 *sa6_dst, sin6;
1498 if (ro_pmtu == NULL || do_lookup) {
1501 * Here ro_pmtu has final destination address, while
1502 * ro might represent immediate destination.
1503 * Use ro_pmtu destination since mtu might differ.
1505 if (ro_pmtu != NULL) {
1506 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1507 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1508 ro_pmtu->ro_mtu = 0;
1512 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1513 bzero(sa6_dst, sizeof(*sa6_dst));
1514 sa6_dst->sin6_family = AF_INET6;
1515 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1516 sa6_dst->sin6_addr = *dst;
1518 in6_splitscope(dst, &kdst, &scopeid);
1519 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1522 if (ro_pmtu != NULL)
1523 ro_pmtu->ro_mtu = mtu;
1526 mtu = ro_pmtu->ro_mtu;
1529 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1530 mtu = ro_pmtu->ro_nh->nh_mtu;
1532 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1536 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1537 * hostcache data for @dst.
1538 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1540 * Returns 0 on success.
1543 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1544 u_long *mtup, int *alwaysfragp, u_int proto)
1552 struct in_conninfo inc;
1554 bzero(&inc, sizeof(inc));
1555 inc.inc_flags |= INC_ISIPV6;
1556 inc.inc6_faddr = *dst;
1558 ifmtu = IN6_LINKMTU(ifp);
1560 /* TCP is known to react to pmtu changes so skip hc */
1561 if (proto != IPPROTO_TCP)
1562 mtu = tcp_hc_getmtu(&inc);
1565 mtu = min(mtu, rt_mtu);
1570 else if (mtu < IPV6_MMTU) {
1572 * RFC2460 section 5, last paragraph:
1573 * if we record ICMPv6 too big message with
1574 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1575 * or smaller, with framgent header attached.
1576 * (fragment header is needed regardless from the
1577 * packet size, for translators to identify packets)
1583 mtu = IN6_LINKMTU(ifp);
1585 error = EHOSTUNREACH; /* XXX */
1589 *alwaysfragp = alwaysfrag;
1594 * IP6 socket option processing.
1597 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1599 int optdatalen, uproto;
1601 struct inpcb *inp = sotoinpcb(so);
1603 int level, op, optname;
1607 uint32_t rss_bucket;
1612 * Don't use more than a quarter of mbuf clusters. N.B.:
1613 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1614 * on LP64 architectures, so cast to u_long to avoid undefined
1615 * behavior. ILP32 architectures cannot have nmbclusters
1616 * large enough to overflow for other reasons.
1618 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1620 level = sopt->sopt_level;
1621 op = sopt->sopt_dir;
1622 optname = sopt->sopt_name;
1623 optlen = sopt->sopt_valsize;
1627 uproto = (int)so->so_proto->pr_protocol;
1629 if (level != IPPROTO_IPV6) {
1632 if (sopt->sopt_level == SOL_SOCKET &&
1633 sopt->sopt_dir == SOPT_SET) {
1634 switch (sopt->sopt_name) {
1637 if ((so->so_options & SO_REUSEADDR) != 0)
1638 inp->inp_flags2 |= INP_REUSEADDR;
1640 inp->inp_flags2 &= ~INP_REUSEADDR;
1646 if ((so->so_options & SO_REUSEPORT) != 0)
1647 inp->inp_flags2 |= INP_REUSEPORT;
1649 inp->inp_flags2 &= ~INP_REUSEPORT;
1653 case SO_REUSEPORT_LB:
1655 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1656 inp->inp_flags2 |= INP_REUSEPORT_LB;
1658 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1664 inp->inp_inc.inc_fibnum = so->so_fibnum;
1668 case SO_MAX_PACING_RATE:
1671 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1682 } else { /* level == IPPROTO_IPV6 */
1687 case IPV6_2292PKTOPTIONS:
1688 #ifdef IPV6_PKTOPTIONS
1689 case IPV6_PKTOPTIONS:
1694 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1695 printf("ip6_ctloutput: mbuf limit hit\n");
1700 error = soopt_getm(sopt, &m); /* XXX */
1703 error = soopt_mcopyin(sopt, m); /* XXX */
1707 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1710 m_freem(m); /* XXX */
1715 * Use of some Hop-by-Hop options or some
1716 * Destination options, might require special
1717 * privilege. That is, normal applications
1718 * (without special privilege) might be forbidden
1719 * from setting certain options in outgoing packets,
1720 * and might never see certain options in received
1721 * packets. [RFC 2292 Section 6]
1722 * KAME specific note:
1723 * KAME prevents non-privileged users from sending or
1724 * receiving ANY hbh/dst options in order to avoid
1725 * overhead of parsing options in the kernel.
1727 case IPV6_RECVHOPOPTS:
1728 case IPV6_RECVDSTOPTS:
1729 case IPV6_RECVRTHDRDSTOPTS:
1731 error = priv_check(td,
1732 PRIV_NETINET_SETHDROPTS);
1737 case IPV6_UNICAST_HOPS:
1740 case IPV6_RECVPKTINFO:
1741 case IPV6_RECVHOPLIMIT:
1742 case IPV6_RECVRTHDR:
1743 case IPV6_RECVPATHMTU:
1744 case IPV6_RECVTCLASS:
1745 case IPV6_RECVFLOWID:
1747 case IPV6_RECVRSSBUCKETID:
1750 case IPV6_AUTOFLOWLABEL:
1751 case IPV6_ORIGDSTADDR:
1753 case IPV6_BINDMULTI:
1755 case IPV6_RSS_LISTEN_BUCKET:
1757 if (optname == IPV6_BINDANY && td != NULL) {
1758 error = priv_check(td,
1759 PRIV_NETINET_BINDANY);
1764 if (optlen != sizeof(int)) {
1768 error = sooptcopyin(sopt, &optval,
1769 sizeof optval, sizeof optval);
1774 case IPV6_UNICAST_HOPS:
1775 if (optval < -1 || optval >= 256)
1778 /* -1 = kernel default */
1779 inp->in6p_hops = optval;
1780 if ((inp->inp_vflag &
1782 inp->inp_ip_ttl = optval;
1785 #define OPTSET(bit) \
1789 inp->inp_flags |= (bit); \
1791 inp->inp_flags &= ~(bit); \
1793 } while (/*CONSTCOND*/ 0)
1794 #define OPTSET2292(bit) \
1797 inp->inp_flags |= IN6P_RFC2292; \
1799 inp->inp_flags |= (bit); \
1801 inp->inp_flags &= ~(bit); \
1803 } while (/*CONSTCOND*/ 0)
1804 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1806 #define OPTSET2_N(bit, val) do { \
1808 inp->inp_flags2 |= bit; \
1810 inp->inp_flags2 &= ~bit; \
1812 #define OPTSET2(bit, val) do { \
1814 OPTSET2_N(bit, val); \
1817 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1818 #define OPTSET2292_EXCLUSIVE(bit) \
1821 if (OPTBIT(IN6P_RFC2292)) { \
1825 inp->inp_flags |= (bit); \
1827 inp->inp_flags &= ~(bit); \
1830 } while (/*CONSTCOND*/ 0)
1832 case IPV6_RECVPKTINFO:
1833 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1838 struct ip6_pktopts **optp;
1840 /* cannot mix with RFC2292 */
1841 if (OPTBIT(IN6P_RFC2292)) {
1846 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1848 return (ECONNRESET);
1850 optp = &inp->in6p_outputopts;
1851 error = ip6_pcbopt(IPV6_HOPLIMIT,
1852 (u_char *)&optval, sizeof(optval),
1853 optp, (td != NULL) ? td->td_ucred :
1859 case IPV6_RECVHOPLIMIT:
1860 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1863 case IPV6_RECVHOPOPTS:
1864 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1867 case IPV6_RECVDSTOPTS:
1868 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1871 case IPV6_RECVRTHDRDSTOPTS:
1872 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1875 case IPV6_RECVRTHDR:
1876 OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1879 case IPV6_RECVPATHMTU:
1881 * We ignore this option for TCP
1883 * (RFC3542 leaves this case
1886 if (uproto != IPPROTO_TCP)
1890 case IPV6_RECVFLOWID:
1891 OPTSET2(INP_RECVFLOWID, optval);
1895 case IPV6_RECVRSSBUCKETID:
1896 OPTSET2(INP_RECVRSSBUCKETID, optval);
1902 if (inp->inp_lport ||
1903 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1905 * The socket is already bound.
1912 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1913 inp->inp_vflag &= ~INP_IPV4;
1915 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1916 inp->inp_vflag |= INP_IPV4;
1920 case IPV6_RECVTCLASS:
1921 /* cannot mix with RFC2292 XXX */
1922 OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1924 case IPV6_AUTOFLOWLABEL:
1925 OPTSET(IN6P_AUTOFLOWLABEL);
1928 case IPV6_ORIGDSTADDR:
1929 OPTSET2(INP_ORIGDSTADDR, optval);
1932 OPTSET(INP_BINDANY);
1935 case IPV6_BINDMULTI:
1936 OPTSET2(INP_BINDMULTI, optval);
1939 case IPV6_RSS_LISTEN_BUCKET:
1940 if ((optval >= 0) &&
1941 (optval < rss_getnumbuckets())) {
1943 inp->inp_rss_listen_bucket = optval;
1944 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1956 case IPV6_USE_MIN_MTU:
1957 case IPV6_PREFER_TEMPADDR:
1958 if (optlen != sizeof(optval)) {
1962 error = sooptcopyin(sopt, &optval,
1963 sizeof optval, sizeof optval);
1967 struct ip6_pktopts **optp;
1969 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1971 return (ECONNRESET);
1973 optp = &inp->in6p_outputopts;
1974 error = ip6_pcbopt(optname,
1975 (u_char *)&optval, sizeof(optval),
1976 optp, (td != NULL) ? td->td_ucred :
1982 case IPV6_2292PKTINFO:
1983 case IPV6_2292HOPLIMIT:
1984 case IPV6_2292HOPOPTS:
1985 case IPV6_2292DSTOPTS:
1986 case IPV6_2292RTHDR:
1988 if (optlen != sizeof(int)) {
1992 error = sooptcopyin(sopt, &optval,
1993 sizeof optval, sizeof optval);
1997 case IPV6_2292PKTINFO:
1998 OPTSET2292(IN6P_PKTINFO);
2000 case IPV6_2292HOPLIMIT:
2001 OPTSET2292(IN6P_HOPLIMIT);
2003 case IPV6_2292HOPOPTS:
2005 * Check super-user privilege.
2006 * See comments for IPV6_RECVHOPOPTS.
2009 error = priv_check(td,
2010 PRIV_NETINET_SETHDROPTS);
2014 OPTSET2292(IN6P_HOPOPTS);
2016 case IPV6_2292DSTOPTS:
2018 error = priv_check(td,
2019 PRIV_NETINET_SETHDROPTS);
2023 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2025 case IPV6_2292RTHDR:
2026 OPTSET2292(IN6P_RTHDR);
2034 case IPV6_RTHDRDSTOPTS:
2037 /* new advanced API (RFC3542) */
2039 u_char optbuf_storage[MCLBYTES];
2041 struct ip6_pktopts **optp;
2043 /* cannot mix with RFC2292 */
2044 if (OPTBIT(IN6P_RFC2292)) {
2050 * We only ensure valsize is not too large
2051 * here. Further validation will be done
2054 error = sooptcopyin(sopt, optbuf_storage,
2055 sizeof(optbuf_storage), 0);
2058 optlen = sopt->sopt_valsize;
2059 optbuf = optbuf_storage;
2061 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2063 return (ECONNRESET);
2065 optp = &inp->in6p_outputopts;
2066 error = ip6_pcbopt(optname, optbuf, optlen,
2067 optp, (td != NULL) ? td->td_ucred : NULL,
2074 case IPV6_MULTICAST_IF:
2075 case IPV6_MULTICAST_HOPS:
2076 case IPV6_MULTICAST_LOOP:
2077 case IPV6_JOIN_GROUP:
2078 case IPV6_LEAVE_GROUP:
2080 case MCAST_BLOCK_SOURCE:
2081 case MCAST_UNBLOCK_SOURCE:
2082 case MCAST_JOIN_GROUP:
2083 case MCAST_LEAVE_GROUP:
2084 case MCAST_JOIN_SOURCE_GROUP:
2085 case MCAST_LEAVE_SOURCE_GROUP:
2086 error = ip6_setmoptions(inp, sopt);
2089 case IPV6_PORTRANGE:
2090 error = sooptcopyin(sopt, &optval,
2091 sizeof optval, sizeof optval);
2097 case IPV6_PORTRANGE_DEFAULT:
2098 inp->inp_flags &= ~(INP_LOWPORT);
2099 inp->inp_flags &= ~(INP_HIGHPORT);
2102 case IPV6_PORTRANGE_HIGH:
2103 inp->inp_flags &= ~(INP_LOWPORT);
2104 inp->inp_flags |= INP_HIGHPORT;
2107 case IPV6_PORTRANGE_LOW:
2108 inp->inp_flags &= ~(INP_HIGHPORT);
2109 inp->inp_flags |= INP_LOWPORT;
2119 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2120 case IPV6_IPSEC_POLICY:
2121 if (IPSEC_ENABLED(ipv6)) {
2122 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2129 error = ENOPROTOOPT;
2137 case IPV6_2292PKTOPTIONS:
2138 #ifdef IPV6_PKTOPTIONS
2139 case IPV6_PKTOPTIONS:
2142 * RFC3542 (effectively) deprecated the
2143 * semantics of the 2292-style pktoptions.
2144 * Since it was not reliable in nature (i.e.,
2145 * applications had to expect the lack of some
2146 * information after all), it would make sense
2147 * to simplify this part by always returning
2150 sopt->sopt_valsize = 0;
2153 case IPV6_RECVHOPOPTS:
2154 case IPV6_RECVDSTOPTS:
2155 case IPV6_RECVRTHDRDSTOPTS:
2156 case IPV6_UNICAST_HOPS:
2157 case IPV6_RECVPKTINFO:
2158 case IPV6_RECVHOPLIMIT:
2159 case IPV6_RECVRTHDR:
2160 case IPV6_RECVPATHMTU:
2163 case IPV6_PORTRANGE:
2164 case IPV6_RECVTCLASS:
2165 case IPV6_AUTOFLOWLABEL:
2169 case IPV6_RECVFLOWID:
2171 case IPV6_RSSBUCKETID:
2172 case IPV6_RECVRSSBUCKETID:
2174 case IPV6_BINDMULTI:
2177 case IPV6_RECVHOPOPTS:
2178 optval = OPTBIT(IN6P_HOPOPTS);
2181 case IPV6_RECVDSTOPTS:
2182 optval = OPTBIT(IN6P_DSTOPTS);
2185 case IPV6_RECVRTHDRDSTOPTS:
2186 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2189 case IPV6_UNICAST_HOPS:
2190 optval = inp->in6p_hops;
2193 case IPV6_RECVPKTINFO:
2194 optval = OPTBIT(IN6P_PKTINFO);
2197 case IPV6_RECVHOPLIMIT:
2198 optval = OPTBIT(IN6P_HOPLIMIT);
2201 case IPV6_RECVRTHDR:
2202 optval = OPTBIT(IN6P_RTHDR);
2205 case IPV6_RECVPATHMTU:
2206 optval = OPTBIT(IN6P_MTU);
2210 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2213 case IPV6_PORTRANGE:
2216 flags = inp->inp_flags;
2217 if (flags & INP_HIGHPORT)
2218 optval = IPV6_PORTRANGE_HIGH;
2219 else if (flags & INP_LOWPORT)
2220 optval = IPV6_PORTRANGE_LOW;
2225 case IPV6_RECVTCLASS:
2226 optval = OPTBIT(IN6P_TCLASS);
2229 case IPV6_AUTOFLOWLABEL:
2230 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2233 case IPV6_ORIGDSTADDR:
2234 optval = OPTBIT2(INP_ORIGDSTADDR);
2238 optval = OPTBIT(INP_BINDANY);
2242 optval = inp->inp_flowid;
2246 optval = inp->inp_flowtype;
2249 case IPV6_RECVFLOWID:
2250 optval = OPTBIT2(INP_RECVFLOWID);
2253 case IPV6_RSSBUCKETID:
2255 rss_hash2bucket(inp->inp_flowid,
2259 optval = rss_bucket;
2264 case IPV6_RECVRSSBUCKETID:
2265 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2269 case IPV6_BINDMULTI:
2270 optval = OPTBIT2(INP_BINDMULTI);
2276 error = sooptcopyout(sopt, &optval,
2283 struct ip6_mtuinfo mtuinfo;
2284 struct in6_addr addr;
2286 if (!(so->so_state & SS_ISCONNECTED))
2289 * XXX: we dot not consider the case of source
2290 * routing, or optional information to specify
2291 * the outgoing interface.
2292 * Copy faddr out of inp to avoid holding lock
2293 * on inp during route lookup.
2296 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2298 error = ip6_getpmtu_ctl(so->so_fibnum,
2302 if (pmtu > IPV6_MAXPACKET)
2303 pmtu = IPV6_MAXPACKET;
2305 bzero(&mtuinfo, sizeof(mtuinfo));
2306 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2307 optdata = (void *)&mtuinfo;
2308 optdatalen = sizeof(mtuinfo);
2309 error = sooptcopyout(sopt, optdata,
2314 case IPV6_2292PKTINFO:
2315 case IPV6_2292HOPLIMIT:
2316 case IPV6_2292HOPOPTS:
2317 case IPV6_2292RTHDR:
2318 case IPV6_2292DSTOPTS:
2320 case IPV6_2292PKTINFO:
2321 optval = OPTBIT(IN6P_PKTINFO);
2323 case IPV6_2292HOPLIMIT:
2324 optval = OPTBIT(IN6P_HOPLIMIT);
2326 case IPV6_2292HOPOPTS:
2327 optval = OPTBIT(IN6P_HOPOPTS);
2329 case IPV6_2292RTHDR:
2330 optval = OPTBIT(IN6P_RTHDR);
2332 case IPV6_2292DSTOPTS:
2333 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2336 error = sooptcopyout(sopt, &optval,
2343 case IPV6_RTHDRDSTOPTS:
2347 case IPV6_USE_MIN_MTU:
2348 case IPV6_PREFER_TEMPADDR:
2349 error = ip6_getpcbopt(inp, optname, sopt);
2352 case IPV6_MULTICAST_IF:
2353 case IPV6_MULTICAST_HOPS:
2354 case IPV6_MULTICAST_LOOP:
2356 error = ip6_getmoptions(inp, sopt);
2359 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2360 case IPV6_IPSEC_POLICY:
2361 if (IPSEC_ENABLED(ipv6)) {
2362 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2368 error = ENOPROTOOPT;
2378 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2380 int error = 0, optval, optlen;
2381 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2382 struct inpcb *inp = sotoinpcb(so);
2383 int level, op, optname;
2385 level = sopt->sopt_level;
2386 op = sopt->sopt_dir;
2387 optname = sopt->sopt_name;
2388 optlen = sopt->sopt_valsize;
2390 if (level != IPPROTO_IPV6) {
2397 * For ICMPv6 sockets, no modification allowed for checksum
2398 * offset, permit "no change" values to help existing apps.
2400 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2401 * for an ICMPv6 socket will fail."
2402 * The current behavior does not meet RFC3542.
2406 if (optlen != sizeof(int)) {
2410 error = sooptcopyin(sopt, &optval, sizeof(optval),
2414 if (optval < -1 || (optval % 2) != 0) {
2416 * The API assumes non-negative even offset
2417 * values or -1 as a special value.
2420 } else if (so->so_proto->pr_protocol ==
2422 if (optval != icmp6off)
2425 inp->in6p_cksum = optval;
2429 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2432 optval = inp->in6p_cksum;
2434 error = sooptcopyout(sopt, &optval, sizeof(optval));
2444 error = ENOPROTOOPT;
2452 * Set up IP6 options in pcb for insertion in output packets or
2453 * specifying behavior of outgoing packets.
2456 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2457 struct socket *so, struct sockopt *sopt)
2459 struct ip6_pktopts *opt = *pktopt;
2461 struct thread *td = sopt->sopt_td;
2463 /* turn off any old options. */
2466 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2467 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2468 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2469 printf("ip6_pcbopts: all specified options are cleared.\n");
2471 ip6_clearpktopts(opt, -1);
2473 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2479 if (!m || m->m_len == 0) {
2481 * Only turning off any previous options, regardless of
2482 * whether the opt is just created or given.
2484 free(opt, M_IP6OPT);
2488 /* set options specified by user. */
2489 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2490 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2491 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2492 free(opt, M_IP6OPT);
2500 * initialize ip6_pktopts. beware that there are non-zero default values in
2504 ip6_initpktopts(struct ip6_pktopts *opt)
2507 bzero(opt, sizeof(*opt));
2508 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2509 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2510 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2511 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2515 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2516 struct ucred *cred, int uproto)
2518 struct ip6_pktopts *opt;
2520 if (*pktopt == NULL) {
2521 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2523 if (*pktopt == NULL)
2525 ip6_initpktopts(*pktopt);
2529 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2532 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2533 if (pktopt && pktopt->field) { \
2535 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2536 malloc_optdata = true; \
2538 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
2540 free(optdata, M_TEMP); \
2541 return (ECONNRESET); \
2543 pktopt = inp->in6p_outputopts; \
2544 if (pktopt && pktopt->field) { \
2545 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2546 bcopy(&pktopt->field, optdata, optdatalen); \
2548 free(optdata, M_TEMP); \
2550 malloc_optdata = false; \
2555 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2556 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2558 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2559 pktopt->field->sa_len)
2562 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2564 void *optdata = NULL;
2565 bool malloc_optdata = false;
2568 struct in6_pktinfo null_pktinfo;
2569 int deftclass = 0, on;
2570 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2571 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2572 struct ip6_pktopts *pktopt;
2575 pktopt = inp->in6p_outputopts;
2579 optdata = (void *)&null_pktinfo;
2580 if (pktopt && pktopt->ip6po_pktinfo) {
2581 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2582 sizeof(null_pktinfo));
2583 in6_clearscope(&null_pktinfo.ipi6_addr);
2585 /* XXX: we don't have to do this every time... */
2586 bzero(&null_pktinfo, sizeof(null_pktinfo));
2588 optdatalen = sizeof(struct in6_pktinfo);
2591 if (pktopt && pktopt->ip6po_tclass >= 0)
2592 deftclass = pktopt->ip6po_tclass;
2593 optdata = (void *)&deftclass;
2594 optdatalen = sizeof(int);
2597 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2600 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2602 case IPV6_RTHDRDSTOPTS:
2603 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2606 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2609 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2611 case IPV6_USE_MIN_MTU:
2613 defminmtu = pktopt->ip6po_minmtu;
2614 optdata = (void *)&defminmtu;
2615 optdatalen = sizeof(int);
2618 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2622 optdata = (void *)&on;
2623 optdatalen = sizeof(on);
2625 case IPV6_PREFER_TEMPADDR:
2627 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2628 optdata = (void *)&defpreftemp;
2629 optdatalen = sizeof(int);
2631 default: /* should not happen */
2633 panic("ip6_getpcbopt: unexpected option\n");
2636 return (ENOPROTOOPT);
2640 error = sooptcopyout(sopt, optdata, optdatalen);
2642 free(optdata, M_TEMP);
2648 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2653 if (optname == -1 || optname == IPV6_PKTINFO) {
2654 if (pktopt->ip6po_pktinfo)
2655 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2656 pktopt->ip6po_pktinfo = NULL;
2658 if (optname == -1 || optname == IPV6_HOPLIMIT)
2659 pktopt->ip6po_hlim = -1;
2660 if (optname == -1 || optname == IPV6_TCLASS)
2661 pktopt->ip6po_tclass = -1;
2662 if (optname == -1 || optname == IPV6_NEXTHOP) {
2663 if (pktopt->ip6po_nextroute.ro_nh) {
2664 NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2665 pktopt->ip6po_nextroute.ro_nh = NULL;
2667 if (pktopt->ip6po_nexthop)
2668 free(pktopt->ip6po_nexthop, M_IP6OPT);
2669 pktopt->ip6po_nexthop = NULL;
2671 if (optname == -1 || optname == IPV6_HOPOPTS) {
2672 if (pktopt->ip6po_hbh)
2673 free(pktopt->ip6po_hbh, M_IP6OPT);
2674 pktopt->ip6po_hbh = NULL;
2676 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2677 if (pktopt->ip6po_dest1)
2678 free(pktopt->ip6po_dest1, M_IP6OPT);
2679 pktopt->ip6po_dest1 = NULL;
2681 if (optname == -1 || optname == IPV6_RTHDR) {
2682 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2683 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2684 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2685 if (pktopt->ip6po_route.ro_nh) {
2686 NH_FREE(pktopt->ip6po_route.ro_nh);
2687 pktopt->ip6po_route.ro_nh = NULL;
2690 if (optname == -1 || optname == IPV6_DSTOPTS) {
2691 if (pktopt->ip6po_dest2)
2692 free(pktopt->ip6po_dest2, M_IP6OPT);
2693 pktopt->ip6po_dest2 = NULL;
2697 #define PKTOPT_EXTHDRCPY(type) \
2700 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2701 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2702 if (dst->type == NULL)\
2704 bcopy(src->type, dst->type, hlen);\
2706 } while (/*CONSTCOND*/ 0)
2709 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2711 if (dst == NULL || src == NULL) {
2712 printf("ip6_clearpktopts: invalid argument\n");
2716 dst->ip6po_hlim = src->ip6po_hlim;
2717 dst->ip6po_tclass = src->ip6po_tclass;
2718 dst->ip6po_flags = src->ip6po_flags;
2719 dst->ip6po_minmtu = src->ip6po_minmtu;
2720 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2721 if (src->ip6po_pktinfo) {
2722 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2724 if (dst->ip6po_pktinfo == NULL)
2726 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2728 if (src->ip6po_nexthop) {
2729 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2731 if (dst->ip6po_nexthop == NULL)
2733 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2734 src->ip6po_nexthop->sa_len);
2736 PKTOPT_EXTHDRCPY(ip6po_hbh);
2737 PKTOPT_EXTHDRCPY(ip6po_dest1);
2738 PKTOPT_EXTHDRCPY(ip6po_dest2);
2739 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2743 ip6_clearpktopts(dst, -1);
2746 #undef PKTOPT_EXTHDRCPY
2748 struct ip6_pktopts *
2749 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2752 struct ip6_pktopts *dst;
2754 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2757 ip6_initpktopts(dst);
2759 if ((error = copypktopts(dst, src, canwait)) != 0) {
2760 free(dst, M_IP6OPT);
2768 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2773 ip6_clearpktopts(pktopt, -1);
2775 free(pktopt, M_IP6OPT);
2779 * Set IPv6 outgoing packet options based on advanced API.
2782 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2783 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2785 struct cmsghdr *cm = NULL;
2787 if (control == NULL || opt == NULL)
2790 ip6_initpktopts(opt);
2795 * If stickyopt is provided, make a local copy of the options
2796 * for this particular packet, then override them by ancillary
2798 * XXX: copypktopts() does not copy the cached route to a next
2799 * hop (if any). This is not very good in terms of efficiency,
2800 * but we can allow this since this option should be rarely
2803 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2808 * XXX: Currently, we assume all the optional information is stored
2811 if (control->m_next)
2814 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2815 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2818 if (control->m_len < CMSG_LEN(0))
2821 cm = mtod(control, struct cmsghdr *);
2822 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2824 if (cm->cmsg_level != IPPROTO_IPV6)
2827 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2828 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2837 * Set a particular packet option, as a sticky option or an ancillary data
2838 * item. "len" can be 0 only when it's a sticky option.
2839 * We have 4 cases of combination of "sticky" and "cmsg":
2840 * "sticky=0, cmsg=0": impossible
2841 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2842 * "sticky=1, cmsg=0": RFC3542 socket option
2843 * "sticky=1, cmsg=1": RFC2292 socket option
2846 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2847 struct ucred *cred, int sticky, int cmsg, int uproto)
2849 int minmtupolicy, preftemp;
2852 if (!sticky && !cmsg) {
2854 printf("ip6_setpktopt: impossible case\n");
2860 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2861 * not be specified in the context of RFC3542. Conversely,
2862 * RFC3542 types should not be specified in the context of RFC2292.
2866 case IPV6_2292PKTINFO:
2867 case IPV6_2292HOPLIMIT:
2868 case IPV6_2292NEXTHOP:
2869 case IPV6_2292HOPOPTS:
2870 case IPV6_2292DSTOPTS:
2871 case IPV6_2292RTHDR:
2872 case IPV6_2292PKTOPTIONS:
2873 return (ENOPROTOOPT);
2876 if (sticky && cmsg) {
2883 case IPV6_RTHDRDSTOPTS:
2885 case IPV6_USE_MIN_MTU:
2888 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2889 return (ENOPROTOOPT);
2894 case IPV6_2292PKTINFO:
2897 struct ifnet *ifp = NULL;
2898 struct in6_pktinfo *pktinfo;
2900 if (len != sizeof(struct in6_pktinfo))
2903 pktinfo = (struct in6_pktinfo *)buf;
2906 * An application can clear any sticky IPV6_PKTINFO option by
2907 * doing a "regular" setsockopt with ipi6_addr being
2908 * in6addr_any and ipi6_ifindex being zero.
2909 * [RFC 3542, Section 6]
2911 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2912 pktinfo->ipi6_ifindex == 0 &&
2913 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2914 ip6_clearpktopts(opt, optname);
2918 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2919 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2922 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2924 /* validate the interface index if specified. */
2925 if (pktinfo->ipi6_ifindex > V_if_index)
2927 if (pktinfo->ipi6_ifindex) {
2928 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2932 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2933 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2937 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2938 struct in6_ifaddr *ia;
2940 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2941 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2943 return (EADDRNOTAVAIL);
2944 ifa_free(&ia->ia_ifa);
2947 * We store the address anyway, and let in6_selectsrc()
2948 * validate the specified address. This is because ipi6_addr
2949 * may not have enough information about its scope zone, and
2950 * we may need additional information (such as outgoing
2951 * interface or the scope zone of a destination address) to
2952 * disambiguate the scope.
2953 * XXX: the delay of the validation may confuse the
2954 * application when it is used as a sticky option.
2956 if (opt->ip6po_pktinfo == NULL) {
2957 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2958 M_IP6OPT, M_NOWAIT);
2959 if (opt->ip6po_pktinfo == NULL)
2962 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2966 case IPV6_2292HOPLIMIT:
2972 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2973 * to simplify the ordering among hoplimit options.
2975 if (optname == IPV6_HOPLIMIT && sticky)
2976 return (ENOPROTOOPT);
2978 if (len != sizeof(int))
2981 if (*hlimp < -1 || *hlimp > 255)
2984 opt->ip6po_hlim = *hlimp;
2992 if (len != sizeof(int))
2994 tclass = *(int *)buf;
2995 if (tclass < -1 || tclass > 255)
2998 opt->ip6po_tclass = tclass;
3002 case IPV6_2292NEXTHOP:
3005 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3010 if (len == 0) { /* just remove the option */
3011 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3015 /* check if cmsg_len is large enough for sa_len */
3016 if (len < sizeof(struct sockaddr) || len < *buf)
3019 switch (((struct sockaddr *)buf)->sa_family) {
3022 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3025 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3028 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3029 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3032 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3038 case AF_LINK: /* should eventually be supported */
3040 return (EAFNOSUPPORT);
3043 /* turn off the previous option, then set the new option. */
3044 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3045 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3046 if (opt->ip6po_nexthop == NULL)
3048 bcopy(buf, opt->ip6po_nexthop, *buf);
3051 case IPV6_2292HOPOPTS:
3054 struct ip6_hbh *hbh;
3058 * XXX: We don't allow a non-privileged user to set ANY HbH
3059 * options, since per-option restriction has too much
3063 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3069 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3070 break; /* just remove the option */
3073 /* message length validation */
3074 if (len < sizeof(struct ip6_hbh))
3076 hbh = (struct ip6_hbh *)buf;
3077 hbhlen = (hbh->ip6h_len + 1) << 3;
3081 /* turn off the previous option, then set the new option. */
3082 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3083 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3084 if (opt->ip6po_hbh == NULL)
3086 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3091 case IPV6_2292DSTOPTS:
3093 case IPV6_RTHDRDSTOPTS:
3095 struct ip6_dest *dest, **newdest = NULL;
3098 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3099 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3105 ip6_clearpktopts(opt, optname);
3106 break; /* just remove the option */
3109 /* message length validation */
3110 if (len < sizeof(struct ip6_dest))
3112 dest = (struct ip6_dest *)buf;
3113 destlen = (dest->ip6d_len + 1) << 3;
3118 * Determine the position that the destination options header
3119 * should be inserted; before or after the routing header.
3122 case IPV6_2292DSTOPTS:
3124 * The old advacned API is ambiguous on this point.
3125 * Our approach is to determine the position based
3126 * according to the existence of a routing header.
3127 * Note, however, that this depends on the order of the
3128 * extension headers in the ancillary data; the 1st
3129 * part of the destination options header must appear
3130 * before the routing header in the ancillary data,
3132 * RFC3542 solved the ambiguity by introducing
3133 * separate ancillary data or option types.
3135 if (opt->ip6po_rthdr == NULL)
3136 newdest = &opt->ip6po_dest1;
3138 newdest = &opt->ip6po_dest2;
3140 case IPV6_RTHDRDSTOPTS:
3141 newdest = &opt->ip6po_dest1;
3144 newdest = &opt->ip6po_dest2;
3148 /* turn off the previous option, then set the new option. */
3149 ip6_clearpktopts(opt, optname);
3150 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3151 if (*newdest == NULL)
3153 bcopy(dest, *newdest, destlen);
3158 case IPV6_2292RTHDR:
3161 struct ip6_rthdr *rth;
3165 ip6_clearpktopts(opt, IPV6_RTHDR);
3166 break; /* just remove the option */
3169 /* message length validation */
3170 if (len < sizeof(struct ip6_rthdr))
3172 rth = (struct ip6_rthdr *)buf;
3173 rthlen = (rth->ip6r_len + 1) << 3;
3177 switch (rth->ip6r_type) {
3178 case IPV6_RTHDR_TYPE_0:
3179 if (rth->ip6r_len == 0) /* must contain one addr */
3181 if (rth->ip6r_len % 2) /* length must be even */
3183 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3187 return (EINVAL); /* not supported */
3190 /* turn off the previous option */
3191 ip6_clearpktopts(opt, IPV6_RTHDR);
3192 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3193 if (opt->ip6po_rthdr == NULL)
3195 bcopy(rth, opt->ip6po_rthdr, rthlen);
3200 case IPV6_USE_MIN_MTU:
3201 if (len != sizeof(int))
3203 minmtupolicy = *(int *)buf;
3204 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3205 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3206 minmtupolicy != IP6PO_MINMTU_ALL) {
3209 opt->ip6po_minmtu = minmtupolicy;
3213 if (len != sizeof(int))
3216 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3218 * we ignore this option for TCP sockets.
3219 * (RFC3542 leaves this case unspecified.)
3221 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3223 opt->ip6po_flags |= IP6PO_DONTFRAG;
3226 case IPV6_PREFER_TEMPADDR:
3227 if (len != sizeof(int))
3229 preftemp = *(int *)buf;
3230 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3231 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3232 preftemp != IP6PO_TEMPADDR_PREFER) {
3235 opt->ip6po_prefer_tempaddr = preftemp;
3239 return (ENOPROTOOPT);
3240 } /* end of switch */
3246 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3247 * packet to the input queue of a specified interface. Note that this
3248 * calls the output routine of the loopback "driver", but with an interface
3249 * pointer that might NOT be &loif -- easier than replicating that code here.
3252 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3255 struct ip6_hdr *ip6;
3257 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3262 * Make sure to deep-copy IPv6 header portion in case the data
3263 * is in an mbuf cluster, so that we can safely override the IPv6
3264 * header portion later.
3266 if (!M_WRITABLE(copym) ||
3267 copym->m_len < sizeof(struct ip6_hdr)) {
3268 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3272 ip6 = mtod(copym, struct ip6_hdr *);
3274 * clear embedded scope identifiers if necessary.
3275 * in6_clearscope will touch the addresses only when necessary.
3277 in6_clearscope(&ip6->ip6_src);
3278 in6_clearscope(&ip6->ip6_dst);
3279 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3280 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3282 copym->m_pkthdr.csum_data = 0xffff;
3284 if_simloop(ifp, copym, AF_INET6, 0);
3288 * Chop IPv6 header off from the payload.
3291 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3294 struct ip6_hdr *ip6;
3296 ip6 = mtod(m, struct ip6_hdr *);
3297 if (m->m_len > sizeof(*ip6)) {
3298 mh = m_gethdr(M_NOWAIT, MT_DATA);
3303 m_move_pkthdr(mh, m);
3304 M_ALIGN(mh, sizeof(*ip6));
3305 m->m_len -= sizeof(*ip6);
3306 m->m_data += sizeof(*ip6);
3309 m->m_len = sizeof(*ip6);
3310 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3312 exthdrs->ip6e_ip6 = m;
3317 * Compute IPv6 extension header length.
3320 ip6_optlen(struct inpcb *inp)
3324 if (!inp->in6p_outputopts)
3329 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3331 len += elen(inp->in6p_outputopts->ip6po_hbh);
3332 if (inp->in6p_outputopts->ip6po_rthdr)
3333 /* dest1 is valid with rthdr only */
3334 len += elen(inp->in6p_outputopts->ip6po_dest1);
3335 len += elen(inp->in6p_outputopts->ip6po_rthdr);
3336 len += elen(inp->in6p_outputopts->ip6po_dest2);