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);
159 * Make an extension header from option data. hp is the source,
160 * mp is the destination, and _ol is the optlen.
162 #define MAKE_EXTHDR(hp, mp, _ol) \
165 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
166 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
167 ((eh)->ip6e_len + 1) << 3); \
170 (_ol) += (*(mp))->m_len; \
172 } while (/*CONSTCOND*/ 0)
175 * Form a chain of extension headers.
176 * m is the extension header mbuf
177 * mp is the previous mbuf in the chain
178 * p is the next header
179 * i is the type of option.
181 #define MAKE_CHAIN(m, mp, p, i)\
185 panic("%s:%d: assumption failed: "\
186 "hdr not split: hdrsplit %d exthdrs %p",\
187 __func__, __LINE__, hdrsplit, &exthdrs);\
188 *mtod((m), u_char *) = *(p);\
190 p = mtod((m), u_char *);\
191 (m)->m_next = (mp)->m_next;\
195 } while (/*CONSTCOND*/ 0)
198 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
202 csum = in_cksum_skip(m, offset + plen, offset);
203 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
205 offset += m->m_pkthdr.csum_data; /* checksum offset */
207 if (offset + sizeof(csum) > m->m_len)
208 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
210 *(u_short *)mtodo(m, offset) = csum;
214 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
215 int plen, int optlen, bool frag)
218 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
219 "csum_flags %#x frag %d\n",
220 __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
222 if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
223 #if defined(SCTP) || defined(SCTP_SUPPORT)
224 (csum_flags & CSUM_SCTP_IPV6) ||
226 (!frag && (ifp->if_capenable & IFCAP_NOMAP) == 0)) {
227 m = mb_unmapped_to_ext(m);
230 in6_ifstat_inc(ifp, ifs6_out_fragfail);
232 IP6STAT_INC(ip6s_odropped);
235 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
236 in6_delayed_cksum(m, plen - optlen,
237 sizeof(struct ip6_hdr) + optlen);
238 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
240 #if defined(SCTP) || defined(SCTP_SUPPORT)
241 if (csum_flags & CSUM_SCTP_IPV6) {
242 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
243 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
252 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
253 int fraglen , uint32_t id)
255 struct mbuf *m, **mnext, *m_frgpart;
256 struct ip6_hdr *ip6, *mhip6;
257 struct ip6_frag *ip6f;
260 int tlen = m0->m_pkthdr.len;
262 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
265 ip6 = mtod(m, struct ip6_hdr *);
266 mnext = &m->m_nextpkt;
268 for (off = hlen; off < tlen; off += fraglen) {
269 m = m_gethdr(M_NOWAIT, MT_DATA);
271 IP6STAT_INC(ip6s_odropped);
276 * Make sure the complete packet header gets copied
277 * from the originating mbuf to the newly created
278 * mbuf. This also ensures that existing firewall
279 * classification(s), VLAN tags and so on get copied
280 * to the resulting fragmented packet(s):
282 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
284 IP6STAT_INC(ip6s_odropped);
289 mnext = &m->m_nextpkt;
290 m->m_data += max_linkhdr;
291 mhip6 = mtod(m, struct ip6_hdr *);
293 m->m_len = sizeof(*mhip6);
294 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
296 IP6STAT_INC(ip6s_odropped);
299 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
300 if (off + fraglen >= tlen)
301 fraglen = tlen - off;
303 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
304 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
305 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
306 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
307 IP6STAT_INC(ip6s_odropped);
311 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
312 ip6f->ip6f_reserved = 0;
313 ip6f->ip6f_ident = id;
314 ip6f->ip6f_nxt = nextproto;
315 IP6STAT_INC(ip6s_ofragments);
316 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
323 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
324 struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
328 struct ktls_session *tls = NULL;
330 struct m_snd_tag *mst;
333 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
338 * If this is an unencrypted TLS record, save a reference to
339 * the record. This local reference is used to call
340 * ktls_output_eagain after the mbuf has been freed (thus
341 * dropping the mbuf's reference) in if_output.
343 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
344 tls = ktls_hold(m->m_next->m_epg_tls);
348 * If a TLS session doesn't have a valid tag, it must
349 * have had an earlier ifp mismatch, so drop this
357 * Always stamp tags that include NIC ktls.
363 if (inp != NULL && mst == NULL) {
364 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
365 (inp->inp_snd_tag != NULL &&
366 inp->inp_snd_tag->ifp != ifp))
367 in_pcboutput_txrtlmt(inp, ifp, m);
369 if (inp->inp_snd_tag != NULL)
370 mst = inp->inp_snd_tag;
373 if (stamp_tag && mst != NULL) {
374 KASSERT(m->m_pkthdr.rcvif == NULL,
375 ("trying to add a send tag to a forwarded packet"));
376 if (mst->ifp != ifp) {
381 /* stamp send tag on mbuf */
382 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
383 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
386 error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
389 /* Check for route change invalidating send tags. */
393 error = ktls_output_eagain(inp, tls);
399 in_pcboutput_eagain(inp);
406 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
407 * nxt, hlim, src, dst).
408 * This function may modify ver and hlim only.
409 * The mbuf chain containing the packet will be freed.
410 * The mbuf opt, if present, will not be freed.
411 * If route_in6 ro is present and has ro_nh initialized, route lookup would be
412 * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
413 * then result of route lookup is stored in ro->ro_nh.
415 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
416 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu.
418 * ifpp - XXX: just for statistics
421 * XXX TODO: no flowid is assigned for outbound flows?
424 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
425 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
426 struct ifnet **ifpp, struct inpcb *inp)
429 struct ifnet *ifp, *origifp;
432 struct route_in6 *ro_pmtu;
433 struct nhop_object *nh;
434 struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
435 struct in6_addr odst;
439 struct in6_ifaddr *ia = NULL;
441 int alwaysfrag, dontfrag;
442 u_int32_t optlen, plen = 0, unfragpartlen;
443 struct ip6_exthdrs exthdrs;
444 struct in6_addr src0, dst0;
450 struct m_tag *fwd_tag = NULL;
456 INP_LOCK_ASSERT(inp);
457 M_SETFIB(m, inp->inp_inc.inc_fibnum);
458 if ((flags & IP_NODEFAULTFLOWID) == 0) {
459 /* Unconditionally set flowid. */
460 m->m_pkthdr.flowid = inp->inp_flowid;
461 M_HASHTYPE_SET(m, inp->inp_flowtype);
464 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
468 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
470 * IPSec checking which handles several cases.
471 * FAST IPSEC: We re-injected the packet.
472 * XXX: need scope argument.
474 if (IPSEC_ENABLED(ipv6)) {
475 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
476 if (error == EINPROGRESS)
483 /* Source address validation. */
484 ip6 = mtod(m, struct ip6_hdr *);
485 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
486 (flags & IPV6_UNSPECSRC) == 0) {
488 IP6STAT_INC(ip6s_badscope);
491 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
493 IP6STAT_INC(ip6s_badscope);
498 * If we are given packet options to add extension headers prepare them.
499 * Calculate the total length of the extension header chain.
500 * Keep the length of the unfragmentable part for fragmentation.
502 bzero(&exthdrs, sizeof(exthdrs));
504 unfragpartlen = sizeof(struct ip6_hdr);
506 /* Hop-by-Hop options header. */
507 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
509 /* Destination options header (1st part). */
510 if (opt->ip6po_rthdr) {
511 #ifndef RTHDR_SUPPORT_IMPLEMENTED
513 * If there is a routing header, discard the packet
514 * right away here. RH0/1 are obsolete and we do not
515 * currently support RH2/3/4.
516 * People trying to use RH253/254 may want to disable
518 * The moment we do support any routing header (again)
519 * this block should check the routing type more
527 * Destination options header (1st part).
528 * This only makes sense with a routing header.
529 * See Section 9.2 of RFC 3542.
530 * Disabling this part just for MIP6 convenience is
531 * a bad idea. We need to think carefully about a
532 * way to make the advanced API coexist with MIP6
533 * options, which might automatically be inserted in
536 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
539 /* Routing header. */
540 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
542 unfragpartlen += optlen;
545 * NOTE: we don't add AH/ESP length here (done in
546 * ip6_ipsec_output()).
549 /* Destination options header (2nd part). */
550 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
554 * If there is at least one extension header,
555 * separate IP6 header from the payload.
559 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
563 m = exthdrs.ip6e_ip6;
564 ip6 = mtod(m, struct ip6_hdr *);
568 /* Adjust mbuf packet header length. */
569 m->m_pkthdr.len += optlen;
570 plen = m->m_pkthdr.len - sizeof(*ip6);
572 /* If this is a jumbo payload, insert a jumbo payload option. */
573 if (plen > IPV6_MAXPACKET) {
575 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
579 m = exthdrs.ip6e_ip6;
580 ip6 = mtod(m, struct ip6_hdr *);
583 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
587 ip6->ip6_plen = htons(plen);
588 nexthdrp = &ip6->ip6_nxt;
592 * Concatenate headers and fill in next header fields.
593 * Here we have, on "m"
595 * and we insert headers accordingly.
596 * Finally, we should be getting:
597 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
599 * During the header composing process "m" points to IPv6
600 * header. "mprev" points to an extension header prior to esp.
605 * We treat dest2 specially. This makes IPsec processing
606 * much easier. The goal here is to make mprev point the
607 * mbuf prior to dest2.
609 * Result: IPv6 dest2 payload.
610 * m and mprev will point to IPv6 header.
612 if (exthdrs.ip6e_dest2) {
614 panic("%s:%d: assumption failed: "
615 "hdr not split: hdrsplit %d exthdrs %p",
616 __func__, __LINE__, hdrsplit, &exthdrs);
617 exthdrs.ip6e_dest2->m_next = m->m_next;
618 m->m_next = exthdrs.ip6e_dest2;
619 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
620 ip6->ip6_nxt = IPPROTO_DSTOPTS;
624 * Result: IPv6 hbh dest1 rthdr dest2 payload.
625 * m will point to IPv6 header. mprev will point to the
626 * extension header prior to dest2 (rthdr in the above case).
628 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
629 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
631 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
635 IP6STAT_INC(ip6s_localout);
639 if (opt && opt->ip6po_rthdr)
640 ro = &opt->ip6po_route;
642 dst = (struct sockaddr_in6 *)&ro->ro_dst;
645 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
649 * If specified, try to fill in the traffic class field.
650 * Do not override if a non-zero value is already set.
651 * We check the diffserv field and the ECN field separately.
653 if (opt && opt->ip6po_tclass >= 0) {
656 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
658 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
661 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
664 /* Fill in or override the hop limit field, if necessary. */
665 if (opt && opt->ip6po_hlim != -1)
666 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
667 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
669 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
671 ip6->ip6_hlim = V_ip6_defmcasthlim;
674 if (ro == NULL || ro->ro_nh == NULL) {
675 bzero(dst, sizeof(*dst));
676 dst->sin6_family = AF_INET6;
677 dst->sin6_len = sizeof(*dst);
678 dst->sin6_addr = ip6->ip6_dst;
681 * Validate route against routing table changes.
682 * Make sure that the address family is set in route.
688 if (ro->ro_nh != NULL && inp != NULL) {
689 ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
690 NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
693 if (ro->ro_nh != NULL && fwd_tag == NULL &&
694 (!NH_IS_VALID(ro->ro_nh) ||
695 ro->ro_dst.sin6_family != AF_INET6 ||
696 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
697 RO_INVALIDATE_CACHE(ro);
699 if (ro->ro_nh != NULL && fwd_tag == NULL &&
700 ro->ro_dst.sin6_family == AF_INET6 &&
701 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
706 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
708 if (fwd_tag == NULL) {
709 bzero(&dst_sa, sizeof(dst_sa));
710 dst_sa.sin6_family = AF_INET6;
711 dst_sa.sin6_len = sizeof(dst_sa);
712 dst_sa.sin6_addr = ip6->ip6_dst;
714 error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
715 &nh, fibnum, m->m_pkthdr.flowid);
717 IP6STAT_INC(ip6s_noroute);
719 in6_ifstat_inc(ifp, ifs6_out_discard);
727 * If in6_selectroute() does not return a nexthop
728 * dst may not have been updated.
730 *dst = dst_sa; /* XXX */
732 if (nh->nh_flags & NHF_HOST)
734 ia = (struct in6_ifaddr *)(nh->nh_ifa);
735 counter_u64_add(nh->nh_pksent, 1);
738 struct nhop_object *nh;
739 struct in6_addr kdst;
742 if (fwd_tag == NULL) {
743 bzero(&dst_sa, sizeof(dst_sa));
744 dst_sa.sin6_family = AF_INET6;
745 dst_sa.sin6_len = sizeof(dst_sa);
746 dst_sa.sin6_addr = ip6->ip6_dst;
749 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
751 (ifp = im6o->im6o_multicast_ifp) != NULL) {
752 /* We do not need a route lookup. */
753 *dst = dst_sa; /* XXX */
757 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
759 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
760 IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
762 ifp = in6_getlinkifnet(scopeid);
764 error = EHOSTUNREACH;
767 *dst = dst_sa; /* XXX */
772 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
774 IP6STAT_INC(ip6s_noroute);
775 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
776 error = EHOSTUNREACH;;
782 ia = ifatoia6(nh->nh_ifa);
783 if (nh->nh_flags & NHF_GATEWAY)
784 dst->sin6_addr = nh->gw6_sa.sin6_addr;
789 /* Then nh (for unicast) and ifp must be non-NULL valid values. */
790 if ((flags & IPV6_FORWARDING) == 0) {
791 /* XXX: the FORWARDING flag can be set for mrouting. */
792 in6_ifstat_inc(ifp, ifs6_out_request);
795 /* Setup data structures for scope ID checks. */
797 bzero(&src_sa, sizeof(src_sa));
798 src_sa.sin6_family = AF_INET6;
799 src_sa.sin6_len = sizeof(src_sa);
800 src_sa.sin6_addr = ip6->ip6_src;
803 /* Re-initialize to be sure. */
804 bzero(&dst_sa, sizeof(dst_sa));
805 dst_sa.sin6_family = AF_INET6;
806 dst_sa.sin6_len = sizeof(dst_sa);
807 dst_sa.sin6_addr = ip6->ip6_dst;
809 /* Check for valid scope ID. */
810 if (in6_setscope(&src0, ifp, &zone) == 0 &&
811 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
812 in6_setscope(&dst0, ifp, &zone) == 0 &&
813 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
815 * The outgoing interface is in the zone of the source
816 * and destination addresses.
818 * Because the loopback interface cannot receive
819 * packets with a different scope ID than its own,
820 * there is a trick to pretend the outgoing packet
821 * was received by the real network interface, by
822 * setting "origifp" different from "ifp". This is
823 * only allowed when "ifp" is a loopback network
824 * interface. Refer to code in nd6_output_ifp() for
830 * We should use ia_ifp to support the case of sending
831 * packets to an address of our own.
833 if (ia != NULL && ia->ia_ifp)
836 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
837 sa6_recoverscope(&src_sa) != 0 ||
838 sa6_recoverscope(&dst_sa) != 0 ||
839 dst_sa.sin6_scope_id == 0 ||
840 (src_sa.sin6_scope_id != 0 &&
841 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
842 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
844 * If the destination network interface is not a
845 * loopback interface, or the destination network
846 * address has no scope ID, or the source address has
847 * a scope ID set which is different from the
848 * destination address one, or there is no network
849 * interface representing this scope ID, the address
850 * pair is considered invalid.
852 IP6STAT_INC(ip6s_badscope);
853 in6_ifstat_inc(ifp, ifs6_out_discard);
855 error = EHOSTUNREACH; /* XXX */
858 /* All scope ID checks are successful. */
860 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
861 if (opt && opt->ip6po_nextroute.ro_nh) {
863 * The nexthop is explicitly specified by the
864 * application. We assume the next hop is an IPv6
867 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
869 else if ((nh->nh_flags & NHF_GATEWAY))
873 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
874 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
876 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
877 in6_ifstat_inc(ifp, ifs6_out_mcast);
879 /* Confirm that the outgoing interface supports multicast. */
880 if (!(ifp->if_flags & IFF_MULTICAST)) {
881 IP6STAT_INC(ip6s_noroute);
882 in6_ifstat_inc(ifp, ifs6_out_discard);
886 if ((im6o == NULL && in6_mcast_loop) ||
887 (im6o && im6o->im6o_multicast_loop)) {
889 * Loop back multicast datagram if not expressly
890 * forbidden to do so, even if we have not joined
891 * the address; protocols will filter it later,
892 * thus deferring a hash lookup and lock acquisition
893 * at the expense of an m_copym().
895 ip6_mloopback(ifp, m);
898 * If we are acting as a multicast router, perform
899 * multicast forwarding as if the packet had just
900 * arrived on the interface to which we are about
901 * to send. The multicast forwarding function
902 * recursively calls this function, using the
903 * IPV6_FORWARDING flag to prevent infinite recursion.
905 * Multicasts that are looped back by ip6_mloopback(),
906 * above, will be forwarded by the ip6_input() routine,
909 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
911 * XXX: ip6_mforward expects that rcvif is NULL
912 * when it is called from the originating path.
913 * However, it may not always be the case.
915 m->m_pkthdr.rcvif = NULL;
916 if (ip6_mforward(ip6, ifp, m) != 0) {
923 * Multicasts with a hoplimit of zero may be looped back,
924 * above, but must not be transmitted on a network.
925 * Also, multicasts addressed to the loopback interface
926 * are not sent -- the above call to ip6_mloopback() will
927 * loop back a copy if this host actually belongs to the
928 * destination group on the loopback interface.
930 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
931 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
938 * Fill the outgoing inteface to tell the upper layer
939 * to increment per-interface statistics.
944 /* Determine path MTU. */
945 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
946 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
948 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
949 "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
950 ifp, alwaysfrag, fibnum));
953 * The caller of this function may specify to use the minimum MTU
955 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
956 * setting. The logic is a bit complicated; by default, unicast
957 * packets will follow path MTU while multicast packets will be sent at
958 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
959 * including unicast ones will be sent at the minimum MTU. Multicast
960 * packets will always be sent at the minimum MTU unless
961 * IP6PO_MINMTU_DISABLE is explicitly specified.
962 * See RFC 3542 for more details.
964 if (mtu > IPV6_MMTU) {
965 if ((flags & IPV6_MINMTU))
967 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
969 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
971 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
977 * Clear embedded scope identifiers if necessary.
978 * in6_clearscope() will touch the addresses only when necessary.
980 in6_clearscope(&ip6->ip6_src);
981 in6_clearscope(&ip6->ip6_dst);
984 * If the outgoing packet contains a hop-by-hop options header,
985 * it must be examined and processed even by the source node.
986 * (RFC 2460, section 4.)
988 if (exthdrs.ip6e_hbh) {
989 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
990 u_int32_t dummy; /* XXX unused */
991 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
994 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
995 panic("ip6e_hbh is not contiguous");
998 * XXX: if we have to send an ICMPv6 error to the sender,
999 * we need the M_LOOP flag since icmp6_error() expects
1000 * the IPv6 and the hop-by-hop options header are
1001 * contiguous unless the flag is set.
1003 m->m_flags |= M_LOOP;
1004 m->m_pkthdr.rcvif = ifp;
1005 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1006 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1007 &dummy, &plen) < 0) {
1008 /* m was already freed at this point. */
1009 error = EINVAL;/* better error? */
1012 m->m_flags &= ~M_LOOP; /* XXX */
1013 m->m_pkthdr.rcvif = NULL;
1016 /* Jump over all PFIL processing if hooks are not active. */
1017 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1020 odst = ip6->ip6_dst;
1021 /* Run through list of hooks for output packets. */
1022 switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1024 ip6 = mtod(m, struct ip6_hdr *);
1034 /* See if destination IP address was changed by packet filter. */
1035 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1036 m->m_flags |= M_SKIP_FIREWALL;
1037 /* If destination is now ourself drop to ip6_input(). */
1038 if (in6_localip(&ip6->ip6_dst)) {
1039 m->m_flags |= M_FASTFWD_OURS;
1040 if (m->m_pkthdr.rcvif == NULL)
1041 m->m_pkthdr.rcvif = V_loif;
1042 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1043 m->m_pkthdr.csum_flags |=
1044 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1045 m->m_pkthdr.csum_data = 0xffff;
1047 #if defined(SCTP) || defined(SCTP_SUPPORT)
1048 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1049 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1051 error = netisr_queue(NETISR_IPV6, m);
1055 RO_INVALIDATE_CACHE(ro);
1056 needfiblookup = 1; /* Redo the routing table lookup. */
1059 /* See if fib was changed by packet filter. */
1060 if (fibnum != M_GETFIB(m)) {
1061 m->m_flags |= M_SKIP_FIREWALL;
1062 fibnum = M_GETFIB(m);
1064 RO_INVALIDATE_CACHE(ro);
1070 /* See if local, if yes, send it to netisr. */
1071 if (m->m_flags & M_FASTFWD_OURS) {
1072 if (m->m_pkthdr.rcvif == NULL)
1073 m->m_pkthdr.rcvif = V_loif;
1074 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1075 m->m_pkthdr.csum_flags |=
1076 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1077 m->m_pkthdr.csum_data = 0xffff;
1079 #if defined(SCTP) || defined(SCTP_SUPPORT)
1080 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1081 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1083 error = netisr_queue(NETISR_IPV6, m);
1086 /* Or forward to some other address? */
1087 if ((m->m_flags & M_IP6_NEXTHOP) &&
1088 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1090 dst = (struct sockaddr_in6 *)&ro->ro_dst;
1093 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1094 m->m_flags |= M_SKIP_FIREWALL;
1095 m->m_flags &= ~M_IP6_NEXTHOP;
1096 m_tag_delete(m, fwd_tag);
1102 * Send the packet to the outgoing interface.
1103 * If necessary, do IPv6 fragmentation before sending.
1105 * The logic here is rather complex:
1106 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1107 * 1-a: send as is if tlen <= path mtu
1108 * 1-b: fragment if tlen > path mtu
1110 * 2: if user asks us not to fragment (dontfrag == 1)
1111 * 2-a: send as is if tlen <= interface mtu
1112 * 2-b: error if tlen > interface mtu
1114 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1117 * 4: if dontfrag == 1 && alwaysfrag == 1
1118 * error, as we cannot handle this conflicting request.
1120 sw_csum = m->m_pkthdr.csum_flags;
1122 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1123 sw_csum &= ~ifp->if_hwassist;
1127 * If we added extension headers, we will not do TSO and calculate the
1128 * checksums ourselves for now.
1129 * XXX-BZ Need a framework to know when the NIC can handle it, even
1132 error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1135 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1136 tlen = m->m_pkthdr.len;
1138 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1142 if (dontfrag && alwaysfrag) { /* Case 4. */
1143 /* Conflicting request - can't transmit. */
1147 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */
1149 * Even if the DONTFRAG option is specified, we cannot send the
1150 * packet when the data length is larger than the MTU of the
1151 * outgoing interface.
1152 * Notify the error by sending IPV6_PATHMTU ancillary data if
1153 * application wanted to know the MTU value. Also return an
1154 * error code (this is not described in the API spec).
1157 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1162 /* Transmit packet without fragmentation. */
1163 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1164 struct in6_ifaddr *ia6;
1166 ip6 = mtod(m, struct ip6_hdr *);
1167 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1169 /* Record statistics for this interface address. */
1170 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1171 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1173 ifa_free(&ia6->ia_ifa);
1175 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1176 (flags & IP_NO_SND_TAG_RL) ? false : true);
1180 /* Try to fragment the packet. Cases 1-b and 3. */
1181 if (mtu < IPV6_MMTU) {
1182 /* Path MTU cannot be less than IPV6_MMTU. */
1184 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1186 } else if (ip6->ip6_plen == 0) {
1187 /* Jumbo payload cannot be fragmented. */
1189 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1195 * Too large for the destination or interface;
1196 * fragment if possible.
1197 * Must be able to put at least 8 bytes per fragment.
1199 if (mtu > IPV6_MAXPACKET)
1200 mtu = IPV6_MAXPACKET;
1202 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1205 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1210 * If the interface will not calculate checksums on
1211 * fragmented packets, then do it here.
1212 * XXX-BZ handle the hw offloading case. Need flags.
1214 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1215 plen, optlen, true);
1220 * Change the next header field of the last header in the
1221 * unfragmentable part.
1223 if (exthdrs.ip6e_rthdr) {
1224 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1225 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1226 } else if (exthdrs.ip6e_dest1) {
1227 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1228 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1229 } else if (exthdrs.ip6e_hbh) {
1230 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1231 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1233 ip6 = mtod(m, struct ip6_hdr *);
1234 nextproto = ip6->ip6_nxt;
1235 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1239 * Loop through length of segment after first fragment,
1240 * make new header and copy data of each part and link onto
1244 id = htonl(ip6_randomid());
1245 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1249 in6_ifstat_inc(ifp, ifs6_out_fragok);
1252 /* Remove leading garbage. */
1261 /* Record statistics for this interface address. */
1263 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1264 counter_u64_add(ia->ia_ifa.ifa_obytes,
1267 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1274 IP6STAT_INC(ip6s_fragmented);
1280 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1281 m_freem(exthdrs.ip6e_dest1);
1282 m_freem(exthdrs.ip6e_rthdr);
1283 m_freem(exthdrs.ip6e_dest2);
1292 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1296 if (hlen > MCLBYTES)
1297 return (ENOBUFS); /* XXX */
1300 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1302 m = m_get(M_NOWAIT, MT_DATA);
1307 bcopy(hdr, mtod(m, caddr_t), hlen);
1314 * Insert jumbo payload option.
1317 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1323 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1326 * If there is no hop-by-hop options header, allocate new one.
1327 * If there is one but it doesn't have enough space to store the
1328 * jumbo payload option, allocate a cluster to store the whole options.
1329 * Otherwise, use it to store the options.
1331 if (exthdrs->ip6e_hbh == NULL) {
1332 mopt = m_get(M_NOWAIT, MT_DATA);
1335 mopt->m_len = JUMBOOPTLEN;
1336 optbuf = mtod(mopt, u_char *);
1337 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1338 exthdrs->ip6e_hbh = mopt;
1340 struct ip6_hbh *hbh;
1342 mopt = exthdrs->ip6e_hbh;
1343 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1346 * - exthdrs->ip6e_hbh is not referenced from places
1347 * other than exthdrs.
1348 * - exthdrs->ip6e_hbh is not an mbuf chain.
1350 int oldoptlen = mopt->m_len;
1354 * XXX: give up if the whole (new) hbh header does
1355 * not fit even in an mbuf cluster.
1357 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1361 * As a consequence, we must always prepare a cluster
1364 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1367 n->m_len = oldoptlen + JUMBOOPTLEN;
1368 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1370 optbuf = mtod(n, caddr_t) + oldoptlen;
1372 mopt = exthdrs->ip6e_hbh = n;
1374 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1375 mopt->m_len += JUMBOOPTLEN;
1377 optbuf[0] = IP6OPT_PADN;
1381 * Adjust the header length according to the pad and
1382 * the jumbo payload option.
1384 hbh = mtod(mopt, struct ip6_hbh *);
1385 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1388 /* fill in the option. */
1389 optbuf[2] = IP6OPT_JUMBO;
1391 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1392 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1394 /* finally, adjust the packet header length */
1395 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1402 * Insert fragment header and copy unfragmentable header portions.
1405 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1406 struct ip6_frag **frghdrp)
1408 struct mbuf *n, *mlast;
1410 if (hlen > sizeof(struct ip6_hdr)) {
1411 n = m_copym(m0, sizeof(struct ip6_hdr),
1412 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1419 /* Search for the last mbuf of unfragmentable part. */
1420 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1423 if (M_WRITABLE(mlast) &&
1424 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1425 /* use the trailing space of the last mbuf for the fragment hdr */
1426 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1428 mlast->m_len += sizeof(struct ip6_frag);
1429 m->m_pkthdr.len += sizeof(struct ip6_frag);
1431 /* allocate a new mbuf for the fragment header */
1434 mfrg = m_get(M_NOWAIT, MT_DATA);
1437 mfrg->m_len = sizeof(struct ip6_frag);
1438 *frghdrp = mtod(mfrg, struct ip6_frag *);
1439 mlast->m_next = mfrg;
1446 * Calculates IPv6 path mtu for destination @dst.
1447 * Resulting MTU is stored in @mtup.
1449 * Returns 0 on success.
1452 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1454 struct epoch_tracker et;
1455 struct nhop_object *nh;
1456 struct in6_addr kdst;
1460 in6_splitscope(dst, &kdst, &scopeid);
1462 NET_EPOCH_ENTER(et);
1463 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1465 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1467 error = EHOSTUNREACH;
1474 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1475 * and cached data in @ro_pmtu.
1476 * MTU from (successful) route lookup is saved (along with dst)
1477 * inside @ro_pmtu to avoid subsequent route lookups after packet
1478 * filter processing.
1480 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1481 * Returns 0 on success.
1484 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1485 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1486 int *alwaysfragp, u_int fibnum, u_int proto)
1488 struct nhop_object *nh;
1489 struct in6_addr kdst;
1491 struct sockaddr_in6 *sa6_dst, sin6;
1497 if (ro_pmtu == NULL || do_lookup) {
1499 * Here ro_pmtu has final destination address, while
1500 * ro might represent immediate destination.
1501 * Use ro_pmtu destination since mtu might differ.
1503 if (ro_pmtu != NULL) {
1504 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1505 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1506 ro_pmtu->ro_mtu = 0;
1510 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1511 bzero(sa6_dst, sizeof(*sa6_dst));
1512 sa6_dst->sin6_family = AF_INET6;
1513 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1514 sa6_dst->sin6_addr = *dst;
1516 in6_splitscope(dst, &kdst, &scopeid);
1517 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1520 if (ro_pmtu != NULL)
1521 ro_pmtu->ro_mtu = mtu;
1524 mtu = ro_pmtu->ro_mtu;
1527 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1528 mtu = ro_pmtu->ro_nh->nh_mtu;
1530 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1534 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1535 * hostcache data for @dst.
1536 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1538 * Returns 0 on success.
1541 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1542 u_long *mtup, int *alwaysfragp, u_int proto)
1550 struct in_conninfo inc;
1552 bzero(&inc, sizeof(inc));
1553 inc.inc_flags |= INC_ISIPV6;
1554 inc.inc6_faddr = *dst;
1556 ifmtu = IN6_LINKMTU(ifp);
1558 /* TCP is known to react to pmtu changes so skip hc */
1559 if (proto != IPPROTO_TCP)
1560 mtu = tcp_hc_getmtu(&inc);
1563 mtu = min(mtu, rt_mtu);
1568 else if (mtu < IPV6_MMTU) {
1570 * RFC2460 section 5, last paragraph:
1571 * if we record ICMPv6 too big message with
1572 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1573 * or smaller, with framgent header attached.
1574 * (fragment header is needed regardless from the
1575 * packet size, for translators to identify packets)
1581 mtu = IN6_LINKMTU(ifp);
1583 error = EHOSTUNREACH; /* XXX */
1587 *alwaysfragp = alwaysfrag;
1592 * IP6 socket option processing.
1595 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1597 int optdatalen, uproto;
1599 struct inpcb *inp = sotoinpcb(so);
1601 int level, op, optname;
1605 uint32_t rss_bucket;
1610 * Don't use more than a quarter of mbuf clusters. N.B.:
1611 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1612 * on LP64 architectures, so cast to u_long to avoid undefined
1613 * behavior. ILP32 architectures cannot have nmbclusters
1614 * large enough to overflow for other reasons.
1616 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1618 level = sopt->sopt_level;
1619 op = sopt->sopt_dir;
1620 optname = sopt->sopt_name;
1621 optlen = sopt->sopt_valsize;
1625 uproto = (int)so->so_proto->pr_protocol;
1627 if (level != IPPROTO_IPV6) {
1630 if (sopt->sopt_level == SOL_SOCKET &&
1631 sopt->sopt_dir == SOPT_SET) {
1632 switch (sopt->sopt_name) {
1635 if ((so->so_options & SO_REUSEADDR) != 0)
1636 inp->inp_flags2 |= INP_REUSEADDR;
1638 inp->inp_flags2 &= ~INP_REUSEADDR;
1644 if ((so->so_options & SO_REUSEPORT) != 0)
1645 inp->inp_flags2 |= INP_REUSEPORT;
1647 inp->inp_flags2 &= ~INP_REUSEPORT;
1651 case SO_REUSEPORT_LB:
1653 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1654 inp->inp_flags2 |= INP_REUSEPORT_LB;
1656 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1662 inp->inp_inc.inc_fibnum = so->so_fibnum;
1666 case SO_MAX_PACING_RATE:
1669 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1680 } else { /* level == IPPROTO_IPV6 */
1684 case IPV6_2292PKTOPTIONS:
1685 #ifdef IPV6_PKTOPTIONS
1686 case IPV6_PKTOPTIONS:
1691 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1692 printf("ip6_ctloutput: mbuf limit hit\n");
1697 error = soopt_getm(sopt, &m); /* XXX */
1700 error = soopt_mcopyin(sopt, m); /* XXX */
1704 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1707 m_freem(m); /* XXX */
1712 * Use of some Hop-by-Hop options or some
1713 * Destination options, might require special
1714 * privilege. That is, normal applications
1715 * (without special privilege) might be forbidden
1716 * from setting certain options in outgoing packets,
1717 * and might never see certain options in received
1718 * packets. [RFC 2292 Section 6]
1719 * KAME specific note:
1720 * KAME prevents non-privileged users from sending or
1721 * receiving ANY hbh/dst options in order to avoid
1722 * overhead of parsing options in the kernel.
1724 case IPV6_RECVHOPOPTS:
1725 case IPV6_RECVDSTOPTS:
1726 case IPV6_RECVRTHDRDSTOPTS:
1728 error = priv_check(td,
1729 PRIV_NETINET_SETHDROPTS);
1734 case IPV6_UNICAST_HOPS:
1737 case IPV6_RECVPKTINFO:
1738 case IPV6_RECVHOPLIMIT:
1739 case IPV6_RECVRTHDR:
1740 case IPV6_RECVPATHMTU:
1741 case IPV6_RECVTCLASS:
1742 case IPV6_RECVFLOWID:
1744 case IPV6_RECVRSSBUCKETID:
1747 case IPV6_AUTOFLOWLABEL:
1748 case IPV6_ORIGDSTADDR:
1750 case IPV6_BINDMULTI:
1752 case IPV6_RSS_LISTEN_BUCKET:
1754 if (optname == IPV6_BINDANY && td != NULL) {
1755 error = priv_check(td,
1756 PRIV_NETINET_BINDANY);
1761 if (optlen != sizeof(int)) {
1765 error = sooptcopyin(sopt, &optval,
1766 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;
2132 case IPV6_2292PKTOPTIONS:
2133 #ifdef IPV6_PKTOPTIONS
2134 case IPV6_PKTOPTIONS:
2137 * RFC3542 (effectively) deprecated the
2138 * semantics of the 2292-style pktoptions.
2139 * Since it was not reliable in nature (i.e.,
2140 * applications had to expect the lack of some
2141 * information after all), it would make sense
2142 * to simplify this part by always returning
2145 sopt->sopt_valsize = 0;
2148 case IPV6_RECVHOPOPTS:
2149 case IPV6_RECVDSTOPTS:
2150 case IPV6_RECVRTHDRDSTOPTS:
2151 case IPV6_UNICAST_HOPS:
2152 case IPV6_RECVPKTINFO:
2153 case IPV6_RECVHOPLIMIT:
2154 case IPV6_RECVRTHDR:
2155 case IPV6_RECVPATHMTU:
2158 case IPV6_PORTRANGE:
2159 case IPV6_RECVTCLASS:
2160 case IPV6_AUTOFLOWLABEL:
2164 case IPV6_RECVFLOWID:
2166 case IPV6_RSSBUCKETID:
2167 case IPV6_RECVRSSBUCKETID:
2169 case IPV6_BINDMULTI:
2171 case IPV6_RECVHOPOPTS:
2172 optval = OPTBIT(IN6P_HOPOPTS);
2175 case IPV6_RECVDSTOPTS:
2176 optval = OPTBIT(IN6P_DSTOPTS);
2179 case IPV6_RECVRTHDRDSTOPTS:
2180 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2183 case IPV6_UNICAST_HOPS:
2184 optval = inp->in6p_hops;
2187 case IPV6_RECVPKTINFO:
2188 optval = OPTBIT(IN6P_PKTINFO);
2191 case IPV6_RECVHOPLIMIT:
2192 optval = OPTBIT(IN6P_HOPLIMIT);
2195 case IPV6_RECVRTHDR:
2196 optval = OPTBIT(IN6P_RTHDR);
2199 case IPV6_RECVPATHMTU:
2200 optval = OPTBIT(IN6P_MTU);
2204 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2207 case IPV6_PORTRANGE:
2210 flags = inp->inp_flags;
2211 if (flags & INP_HIGHPORT)
2212 optval = IPV6_PORTRANGE_HIGH;
2213 else if (flags & INP_LOWPORT)
2214 optval = IPV6_PORTRANGE_LOW;
2219 case IPV6_RECVTCLASS:
2220 optval = OPTBIT(IN6P_TCLASS);
2223 case IPV6_AUTOFLOWLABEL:
2224 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2227 case IPV6_ORIGDSTADDR:
2228 optval = OPTBIT2(INP_ORIGDSTADDR);
2232 optval = OPTBIT(INP_BINDANY);
2236 optval = inp->inp_flowid;
2240 optval = inp->inp_flowtype;
2243 case IPV6_RECVFLOWID:
2244 optval = OPTBIT2(INP_RECVFLOWID);
2247 case IPV6_RSSBUCKETID:
2249 rss_hash2bucket(inp->inp_flowid,
2253 optval = rss_bucket;
2258 case IPV6_RECVRSSBUCKETID:
2259 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2263 case IPV6_BINDMULTI:
2264 optval = OPTBIT2(INP_BINDMULTI);
2269 error = sooptcopyout(sopt, &optval,
2276 struct ip6_mtuinfo mtuinfo;
2277 struct in6_addr addr;
2279 if (!(so->so_state & SS_ISCONNECTED))
2282 * XXX: we dot not consider the case of source
2283 * routing, or optional information to specify
2284 * the outgoing interface.
2285 * Copy faddr out of inp to avoid holding lock
2286 * on inp during route lookup.
2289 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2291 error = ip6_getpmtu_ctl(so->so_fibnum,
2295 if (pmtu > IPV6_MAXPACKET)
2296 pmtu = IPV6_MAXPACKET;
2298 bzero(&mtuinfo, sizeof(mtuinfo));
2299 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2300 optdata = (void *)&mtuinfo;
2301 optdatalen = sizeof(mtuinfo);
2302 error = sooptcopyout(sopt, optdata,
2307 case IPV6_2292PKTINFO:
2308 case IPV6_2292HOPLIMIT:
2309 case IPV6_2292HOPOPTS:
2310 case IPV6_2292RTHDR:
2311 case IPV6_2292DSTOPTS:
2313 case IPV6_2292PKTINFO:
2314 optval = OPTBIT(IN6P_PKTINFO);
2316 case IPV6_2292HOPLIMIT:
2317 optval = OPTBIT(IN6P_HOPLIMIT);
2319 case IPV6_2292HOPOPTS:
2320 optval = OPTBIT(IN6P_HOPOPTS);
2322 case IPV6_2292RTHDR:
2323 optval = OPTBIT(IN6P_RTHDR);
2325 case IPV6_2292DSTOPTS:
2326 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2329 error = sooptcopyout(sopt, &optval,
2336 case IPV6_RTHDRDSTOPTS:
2340 case IPV6_USE_MIN_MTU:
2341 case IPV6_PREFER_TEMPADDR:
2342 error = ip6_getpcbopt(inp, optname, sopt);
2345 case IPV6_MULTICAST_IF:
2346 case IPV6_MULTICAST_HOPS:
2347 case IPV6_MULTICAST_LOOP:
2349 error = ip6_getmoptions(inp, sopt);
2352 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2353 case IPV6_IPSEC_POLICY:
2354 if (IPSEC_ENABLED(ipv6)) {
2355 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2361 error = ENOPROTOOPT;
2371 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2373 int error = 0, optval, optlen;
2374 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2375 struct inpcb *inp = sotoinpcb(so);
2376 int level, op, optname;
2378 level = sopt->sopt_level;
2379 op = sopt->sopt_dir;
2380 optname = sopt->sopt_name;
2381 optlen = sopt->sopt_valsize;
2383 if (level != IPPROTO_IPV6) {
2390 * For ICMPv6 sockets, no modification allowed for checksum
2391 * offset, permit "no change" values to help existing apps.
2393 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2394 * for an ICMPv6 socket will fail."
2395 * The current behavior does not meet RFC3542.
2399 if (optlen != sizeof(int)) {
2403 error = sooptcopyin(sopt, &optval, sizeof(optval),
2407 if (optval < -1 || (optval % 2) != 0) {
2409 * The API assumes non-negative even offset
2410 * values or -1 as a special value.
2413 } else if (so->so_proto->pr_protocol ==
2415 if (optval != icmp6off)
2418 inp->in6p_cksum = optval;
2422 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2425 optval = inp->in6p_cksum;
2427 error = sooptcopyout(sopt, &optval, sizeof(optval));
2437 error = ENOPROTOOPT;
2445 * Set up IP6 options in pcb for insertion in output packets or
2446 * specifying behavior of outgoing packets.
2449 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2450 struct socket *so, struct sockopt *sopt)
2452 struct ip6_pktopts *opt = *pktopt;
2454 struct thread *td = sopt->sopt_td;
2456 /* turn off any old options. */
2459 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2460 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2461 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2462 printf("ip6_pcbopts: all specified options are cleared.\n");
2464 ip6_clearpktopts(opt, -1);
2466 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2472 if (!m || m->m_len == 0) {
2474 * Only turning off any previous options, regardless of
2475 * whether the opt is just created or given.
2477 free(opt, M_IP6OPT);
2481 /* set options specified by user. */
2482 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2483 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2484 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2485 free(opt, M_IP6OPT);
2493 * initialize ip6_pktopts. beware that there are non-zero default values in
2497 ip6_initpktopts(struct ip6_pktopts *opt)
2500 bzero(opt, sizeof(*opt));
2501 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2502 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2503 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2504 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2508 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2509 struct ucred *cred, int uproto)
2511 struct ip6_pktopts *opt;
2513 if (*pktopt == NULL) {
2514 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2516 if (*pktopt == NULL)
2518 ip6_initpktopts(*pktopt);
2522 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2525 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2526 if (pktopt && pktopt->field) { \
2528 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2529 malloc_optdata = true; \
2531 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
2533 free(optdata, M_TEMP); \
2534 return (ECONNRESET); \
2536 pktopt = inp->in6p_outputopts; \
2537 if (pktopt && pktopt->field) { \
2538 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2539 bcopy(&pktopt->field, optdata, optdatalen); \
2541 free(optdata, M_TEMP); \
2543 malloc_optdata = false; \
2548 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2549 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2551 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2552 pktopt->field->sa_len)
2555 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2557 void *optdata = NULL;
2558 bool malloc_optdata = false;
2561 struct in6_pktinfo null_pktinfo;
2562 int deftclass = 0, on;
2563 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2564 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2565 struct ip6_pktopts *pktopt;
2568 pktopt = inp->in6p_outputopts;
2572 optdata = (void *)&null_pktinfo;
2573 if (pktopt && pktopt->ip6po_pktinfo) {
2574 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2575 sizeof(null_pktinfo));
2576 in6_clearscope(&null_pktinfo.ipi6_addr);
2578 /* XXX: we don't have to do this every time... */
2579 bzero(&null_pktinfo, sizeof(null_pktinfo));
2581 optdatalen = sizeof(struct in6_pktinfo);
2584 if (pktopt && pktopt->ip6po_tclass >= 0)
2585 deftclass = pktopt->ip6po_tclass;
2586 optdata = (void *)&deftclass;
2587 optdatalen = sizeof(int);
2590 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2593 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2595 case IPV6_RTHDRDSTOPTS:
2596 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2599 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2602 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2604 case IPV6_USE_MIN_MTU:
2606 defminmtu = pktopt->ip6po_minmtu;
2607 optdata = (void *)&defminmtu;
2608 optdatalen = sizeof(int);
2611 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2615 optdata = (void *)&on;
2616 optdatalen = sizeof(on);
2618 case IPV6_PREFER_TEMPADDR:
2620 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2621 optdata = (void *)&defpreftemp;
2622 optdatalen = sizeof(int);
2624 default: /* should not happen */
2626 panic("ip6_getpcbopt: unexpected option\n");
2629 return (ENOPROTOOPT);
2633 error = sooptcopyout(sopt, optdata, optdatalen);
2635 free(optdata, M_TEMP);
2641 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2646 if (optname == -1 || optname == IPV6_PKTINFO) {
2647 if (pktopt->ip6po_pktinfo)
2648 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2649 pktopt->ip6po_pktinfo = NULL;
2651 if (optname == -1 || optname == IPV6_HOPLIMIT)
2652 pktopt->ip6po_hlim = -1;
2653 if (optname == -1 || optname == IPV6_TCLASS)
2654 pktopt->ip6po_tclass = -1;
2655 if (optname == -1 || optname == IPV6_NEXTHOP) {
2656 if (pktopt->ip6po_nextroute.ro_nh) {
2657 NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2658 pktopt->ip6po_nextroute.ro_nh = NULL;
2660 if (pktopt->ip6po_nexthop)
2661 free(pktopt->ip6po_nexthop, M_IP6OPT);
2662 pktopt->ip6po_nexthop = NULL;
2664 if (optname == -1 || optname == IPV6_HOPOPTS) {
2665 if (pktopt->ip6po_hbh)
2666 free(pktopt->ip6po_hbh, M_IP6OPT);
2667 pktopt->ip6po_hbh = NULL;
2669 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2670 if (pktopt->ip6po_dest1)
2671 free(pktopt->ip6po_dest1, M_IP6OPT);
2672 pktopt->ip6po_dest1 = NULL;
2674 if (optname == -1 || optname == IPV6_RTHDR) {
2675 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2676 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2677 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2678 if (pktopt->ip6po_route.ro_nh) {
2679 NH_FREE(pktopt->ip6po_route.ro_nh);
2680 pktopt->ip6po_route.ro_nh = NULL;
2683 if (optname == -1 || optname == IPV6_DSTOPTS) {
2684 if (pktopt->ip6po_dest2)
2685 free(pktopt->ip6po_dest2, M_IP6OPT);
2686 pktopt->ip6po_dest2 = NULL;
2690 #define PKTOPT_EXTHDRCPY(type) \
2693 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2694 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2695 if (dst->type == NULL)\
2697 bcopy(src->type, dst->type, hlen);\
2699 } while (/*CONSTCOND*/ 0)
2702 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2704 if (dst == NULL || src == NULL) {
2705 printf("ip6_clearpktopts: invalid argument\n");
2709 dst->ip6po_hlim = src->ip6po_hlim;
2710 dst->ip6po_tclass = src->ip6po_tclass;
2711 dst->ip6po_flags = src->ip6po_flags;
2712 dst->ip6po_minmtu = src->ip6po_minmtu;
2713 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2714 if (src->ip6po_pktinfo) {
2715 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2717 if (dst->ip6po_pktinfo == NULL)
2719 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2721 if (src->ip6po_nexthop) {
2722 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2724 if (dst->ip6po_nexthop == NULL)
2726 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2727 src->ip6po_nexthop->sa_len);
2729 PKTOPT_EXTHDRCPY(ip6po_hbh);
2730 PKTOPT_EXTHDRCPY(ip6po_dest1);
2731 PKTOPT_EXTHDRCPY(ip6po_dest2);
2732 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2736 ip6_clearpktopts(dst, -1);
2739 #undef PKTOPT_EXTHDRCPY
2741 struct ip6_pktopts *
2742 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2745 struct ip6_pktopts *dst;
2747 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2750 ip6_initpktopts(dst);
2752 if ((error = copypktopts(dst, src, canwait)) != 0) {
2753 free(dst, M_IP6OPT);
2761 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2766 ip6_clearpktopts(pktopt, -1);
2768 free(pktopt, M_IP6OPT);
2772 * Set IPv6 outgoing packet options based on advanced API.
2775 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2776 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2778 struct cmsghdr *cm = NULL;
2780 if (control == NULL || opt == NULL)
2783 ip6_initpktopts(opt);
2788 * If stickyopt is provided, make a local copy of the options
2789 * for this particular packet, then override them by ancillary
2791 * XXX: copypktopts() does not copy the cached route to a next
2792 * hop (if any). This is not very good in terms of efficiency,
2793 * but we can allow this since this option should be rarely
2796 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2801 * XXX: Currently, we assume all the optional information is stored
2804 if (control->m_next)
2807 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2808 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2811 if (control->m_len < CMSG_LEN(0))
2814 cm = mtod(control, struct cmsghdr *);
2815 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2817 if (cm->cmsg_level != IPPROTO_IPV6)
2820 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2821 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2830 * Set a particular packet option, as a sticky option or an ancillary data
2831 * item. "len" can be 0 only when it's a sticky option.
2832 * We have 4 cases of combination of "sticky" and "cmsg":
2833 * "sticky=0, cmsg=0": impossible
2834 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2835 * "sticky=1, cmsg=0": RFC3542 socket option
2836 * "sticky=1, cmsg=1": RFC2292 socket option
2839 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2840 struct ucred *cred, int sticky, int cmsg, int uproto)
2842 int minmtupolicy, preftemp;
2845 if (!sticky && !cmsg) {
2847 printf("ip6_setpktopt: impossible case\n");
2853 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2854 * not be specified in the context of RFC3542. Conversely,
2855 * RFC3542 types should not be specified in the context of RFC2292.
2859 case IPV6_2292PKTINFO:
2860 case IPV6_2292HOPLIMIT:
2861 case IPV6_2292NEXTHOP:
2862 case IPV6_2292HOPOPTS:
2863 case IPV6_2292DSTOPTS:
2864 case IPV6_2292RTHDR:
2865 case IPV6_2292PKTOPTIONS:
2866 return (ENOPROTOOPT);
2869 if (sticky && cmsg) {
2876 case IPV6_RTHDRDSTOPTS:
2878 case IPV6_USE_MIN_MTU:
2881 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2882 return (ENOPROTOOPT);
2887 case IPV6_2292PKTINFO:
2890 struct ifnet *ifp = NULL;
2891 struct in6_pktinfo *pktinfo;
2893 if (len != sizeof(struct in6_pktinfo))
2896 pktinfo = (struct in6_pktinfo *)buf;
2899 * An application can clear any sticky IPV6_PKTINFO option by
2900 * doing a "regular" setsockopt with ipi6_addr being
2901 * in6addr_any and ipi6_ifindex being zero.
2902 * [RFC 3542, Section 6]
2904 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2905 pktinfo->ipi6_ifindex == 0 &&
2906 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2907 ip6_clearpktopts(opt, optname);
2911 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2912 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2915 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2917 /* validate the interface index if specified. */
2918 if (pktinfo->ipi6_ifindex > V_if_index)
2920 if (pktinfo->ipi6_ifindex) {
2921 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2925 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2926 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2930 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2931 struct in6_ifaddr *ia;
2933 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2934 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2936 return (EADDRNOTAVAIL);
2937 ifa_free(&ia->ia_ifa);
2940 * We store the address anyway, and let in6_selectsrc()
2941 * validate the specified address. This is because ipi6_addr
2942 * may not have enough information about its scope zone, and
2943 * we may need additional information (such as outgoing
2944 * interface or the scope zone of a destination address) to
2945 * disambiguate the scope.
2946 * XXX: the delay of the validation may confuse the
2947 * application when it is used as a sticky option.
2949 if (opt->ip6po_pktinfo == NULL) {
2950 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2951 M_IP6OPT, M_NOWAIT);
2952 if (opt->ip6po_pktinfo == NULL)
2955 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2959 case IPV6_2292HOPLIMIT:
2965 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2966 * to simplify the ordering among hoplimit options.
2968 if (optname == IPV6_HOPLIMIT && sticky)
2969 return (ENOPROTOOPT);
2971 if (len != sizeof(int))
2974 if (*hlimp < -1 || *hlimp > 255)
2977 opt->ip6po_hlim = *hlimp;
2985 if (len != sizeof(int))
2987 tclass = *(int *)buf;
2988 if (tclass < -1 || tclass > 255)
2991 opt->ip6po_tclass = tclass;
2995 case IPV6_2292NEXTHOP:
2998 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3003 if (len == 0) { /* just remove the option */
3004 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3008 /* check if cmsg_len is large enough for sa_len */
3009 if (len < sizeof(struct sockaddr) || len < *buf)
3012 switch (((struct sockaddr *)buf)->sa_family) {
3015 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3018 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3021 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3022 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3025 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3031 case AF_LINK: /* should eventually be supported */
3033 return (EAFNOSUPPORT);
3036 /* turn off the previous option, then set the new option. */
3037 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3038 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3039 if (opt->ip6po_nexthop == NULL)
3041 bcopy(buf, opt->ip6po_nexthop, *buf);
3044 case IPV6_2292HOPOPTS:
3047 struct ip6_hbh *hbh;
3051 * XXX: We don't allow a non-privileged user to set ANY HbH
3052 * options, since per-option restriction has too much
3056 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3062 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3063 break; /* just remove the option */
3066 /* message length validation */
3067 if (len < sizeof(struct ip6_hbh))
3069 hbh = (struct ip6_hbh *)buf;
3070 hbhlen = (hbh->ip6h_len + 1) << 3;
3074 /* turn off the previous option, then set the new option. */
3075 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3076 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3077 if (opt->ip6po_hbh == NULL)
3079 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3084 case IPV6_2292DSTOPTS:
3086 case IPV6_RTHDRDSTOPTS:
3088 struct ip6_dest *dest, **newdest = NULL;
3091 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3092 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3098 ip6_clearpktopts(opt, optname);
3099 break; /* just remove the option */
3102 /* message length validation */
3103 if (len < sizeof(struct ip6_dest))
3105 dest = (struct ip6_dest *)buf;
3106 destlen = (dest->ip6d_len + 1) << 3;
3111 * Determine the position that the destination options header
3112 * should be inserted; before or after the routing header.
3115 case IPV6_2292DSTOPTS:
3117 * The old advacned API is ambiguous on this point.
3118 * Our approach is to determine the position based
3119 * according to the existence of a routing header.
3120 * Note, however, that this depends on the order of the
3121 * extension headers in the ancillary data; the 1st
3122 * part of the destination options header must appear
3123 * before the routing header in the ancillary data,
3125 * RFC3542 solved the ambiguity by introducing
3126 * separate ancillary data or option types.
3128 if (opt->ip6po_rthdr == NULL)
3129 newdest = &opt->ip6po_dest1;
3131 newdest = &opt->ip6po_dest2;
3133 case IPV6_RTHDRDSTOPTS:
3134 newdest = &opt->ip6po_dest1;
3137 newdest = &opt->ip6po_dest2;
3141 /* turn off the previous option, then set the new option. */
3142 ip6_clearpktopts(opt, optname);
3143 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3144 if (*newdest == NULL)
3146 bcopy(dest, *newdest, destlen);
3151 case IPV6_2292RTHDR:
3154 struct ip6_rthdr *rth;
3158 ip6_clearpktopts(opt, IPV6_RTHDR);
3159 break; /* just remove the option */
3162 /* message length validation */
3163 if (len < sizeof(struct ip6_rthdr))
3165 rth = (struct ip6_rthdr *)buf;
3166 rthlen = (rth->ip6r_len + 1) << 3;
3170 switch (rth->ip6r_type) {
3171 case IPV6_RTHDR_TYPE_0:
3172 if (rth->ip6r_len == 0) /* must contain one addr */
3174 if (rth->ip6r_len % 2) /* length must be even */
3176 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3180 return (EINVAL); /* not supported */
3183 /* turn off the previous option */
3184 ip6_clearpktopts(opt, IPV6_RTHDR);
3185 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3186 if (opt->ip6po_rthdr == NULL)
3188 bcopy(rth, opt->ip6po_rthdr, rthlen);
3193 case IPV6_USE_MIN_MTU:
3194 if (len != sizeof(int))
3196 minmtupolicy = *(int *)buf;
3197 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3198 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3199 minmtupolicy != IP6PO_MINMTU_ALL) {
3202 opt->ip6po_minmtu = minmtupolicy;
3206 if (len != sizeof(int))
3209 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3211 * we ignore this option for TCP sockets.
3212 * (RFC3542 leaves this case unspecified.)
3214 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3216 opt->ip6po_flags |= IP6PO_DONTFRAG;
3219 case IPV6_PREFER_TEMPADDR:
3220 if (len != sizeof(int))
3222 preftemp = *(int *)buf;
3223 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3224 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3225 preftemp != IP6PO_TEMPADDR_PREFER) {
3228 opt->ip6po_prefer_tempaddr = preftemp;
3232 return (ENOPROTOOPT);
3233 } /* end of switch */
3239 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3240 * packet to the input queue of a specified interface. Note that this
3241 * calls the output routine of the loopback "driver", but with an interface
3242 * pointer that might NOT be &loif -- easier than replicating that code here.
3245 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3248 struct ip6_hdr *ip6;
3250 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3255 * Make sure to deep-copy IPv6 header portion in case the data
3256 * is in an mbuf cluster, so that we can safely override the IPv6
3257 * header portion later.
3259 if (!M_WRITABLE(copym) ||
3260 copym->m_len < sizeof(struct ip6_hdr)) {
3261 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3265 ip6 = mtod(copym, struct ip6_hdr *);
3267 * clear embedded scope identifiers if necessary.
3268 * in6_clearscope will touch the addresses only when necessary.
3270 in6_clearscope(&ip6->ip6_src);
3271 in6_clearscope(&ip6->ip6_dst);
3272 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3273 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3275 copym->m_pkthdr.csum_data = 0xffff;
3277 if_simloop(ifp, copym, AF_INET6, 0);
3281 * Chop IPv6 header off from the payload.
3284 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3287 struct ip6_hdr *ip6;
3289 ip6 = mtod(m, struct ip6_hdr *);
3290 if (m->m_len > sizeof(*ip6)) {
3291 mh = m_gethdr(M_NOWAIT, MT_DATA);
3296 m_move_pkthdr(mh, m);
3297 M_ALIGN(mh, sizeof(*ip6));
3298 m->m_len -= sizeof(*ip6);
3299 m->m_data += sizeof(*ip6);
3302 m->m_len = sizeof(*ip6);
3303 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3305 exthdrs->ip6e_ip6 = m;
3310 * Compute IPv6 extension header length.
3313 ip6_optlen(struct inpcb *inp)
3317 if (!inp->in6p_outputopts)
3322 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3324 len += elen(inp->in6p_outputopts->ip6po_hbh);
3325 if (inp->in6p_outputopts->ip6po_rthdr)
3326 /* dest1 is valid with rthdr only */
3327 len += elen(inp->in6p_outputopts->ip6po_dest1);
3328 len += elen(inp->in6p_outputopts->ip6po_rthdr);
3329 len += elen(inp->in6p_outputopts->ip6po_dest2);