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_ratelimit.h"
71 #include "opt_ipsec.h"
73 #include "opt_route.h"
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
80 #include <sys/errno.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/ucred.h>
89 #include <machine/in_cksum.h>
92 #include <net/if_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
97 #include <net/rss_config.h>
100 #include <netinet/in.h>
101 #include <netinet/in_var.h>
102 #include <netinet/ip_var.h>
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/tcp_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/in6_rss.h>
113 #include <netipsec/ipsec_support.h>
115 #include <netinet/sctp.h>
116 #include <netinet/sctp_crc32.h>
119 #include <netinet6/ip6protosw.h>
120 #include <netinet6/scope6_var.h>
122 extern int in6_mcast_loop;
125 struct mbuf *ip6e_ip6;
126 struct mbuf *ip6e_hbh;
127 struct mbuf *ip6e_dest1;
128 struct mbuf *ip6e_rthdr;
129 struct mbuf *ip6e_dest2;
132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 struct ucred *, int, int, int);
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static int ip6_getpmtu(struct route_in6 *, int,
148 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
151 u_long *, int *, u_int);
152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
157 * Make an extension header from option data. hp is the source, and
158 * mp is the destination.
160 #define MAKE_EXTHDR(hp, mp) \
163 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
164 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
165 ((eh)->ip6e_len + 1) << 3); \
169 } while (/*CONSTCOND*/ 0)
172 * Form a chain of extension headers.
173 * m is the extension header mbuf
174 * mp is the previous mbuf in the chain
175 * p is the next header
176 * i is the type of option.
178 #define MAKE_CHAIN(m, mp, p, i)\
182 panic("assumption failed: hdr not split"); \
183 *mtod((m), u_char *) = *(p);\
185 p = mtod((m), u_char *);\
186 (m)->m_next = (mp)->m_next;\
190 } while (/*CONSTCOND*/ 0)
193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
197 csum = in_cksum_skip(m, offset + plen, offset);
198 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
200 offset += m->m_pkthdr.csum_data; /* checksum offset */
202 if (offset + sizeof(csum) > m->m_len)
203 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
205 *(u_short *)mtodo(m, offset) = csum;
209 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
210 int fraglen , uint32_t id)
212 struct mbuf *m, **mnext, *m_frgpart;
213 struct ip6_hdr *ip6, *mhip6;
214 struct ip6_frag *ip6f;
217 int tlen = m0->m_pkthdr.len;
219 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
222 ip6 = mtod(m, struct ip6_hdr *);
223 mnext = &m->m_nextpkt;
225 for (off = hlen; off < tlen; off += fraglen) {
226 m = m_gethdr(M_NOWAIT, MT_DATA);
228 IP6STAT_INC(ip6s_odropped);
231 m->m_flags = m0->m_flags & M_COPYFLAGS;
233 mnext = &m->m_nextpkt;
234 m->m_data += max_linkhdr;
235 mhip6 = mtod(m, struct ip6_hdr *);
237 m->m_len = sizeof(*mhip6);
238 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
240 IP6STAT_INC(ip6s_odropped);
243 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
244 if (off + fraglen >= tlen)
245 fraglen = tlen - off;
247 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
248 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
249 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
250 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
251 IP6STAT_INC(ip6s_odropped);
255 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
256 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
257 m->m_pkthdr.rcvif = NULL;
258 ip6f->ip6f_reserved = 0;
259 ip6f->ip6f_ident = id;
260 ip6f->ip6f_nxt = nextproto;
261 IP6STAT_INC(ip6s_ofragments);
262 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
269 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
270 * header (with pri, len, nxt, hlim, src, dst).
271 * This function may modify ver and hlim only.
272 * The mbuf chain containing the packet will be freed.
273 * The mbuf opt, if present, will not be freed.
274 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
275 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
276 * then result of route lookup is stored in ro->ro_rt.
278 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
279 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
282 * ifpp - XXX: just for statistics
285 * XXX TODO: no flowid is assigned for outbound flows?
288 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
289 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
290 struct ifnet **ifpp, struct inpcb *inp)
293 struct ifnet *ifp, *origifp;
295 struct mbuf *mprev = NULL;
297 struct route_in6 ip6route;
298 struct rtentry *rt = NULL;
299 struct sockaddr_in6 *dst, src_sa, dst_sa;
300 struct in6_addr odst;
302 struct in6_ifaddr *ia = NULL;
304 int alwaysfrag, dontfrag;
305 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
306 struct ip6_exthdrs exthdrs;
307 struct in6_addr src0, dst0;
309 struct route_in6 *ro_pmtu = NULL;
314 struct m_tag *fwd_tag = NULL;
318 INP_LOCK_ASSERT(inp);
319 M_SETFIB(m, inp->inp_inc.inc_fibnum);
320 if ((flags & IP_NODEFAULTFLOWID) == 0) {
321 /* unconditionally set flowid */
322 m->m_pkthdr.flowid = inp->inp_flowid;
323 M_HASHTYPE_SET(m, inp->inp_flowtype);
327 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
329 * IPSec checking which handles several cases.
330 * FAST IPSEC: We re-injected the packet.
331 * XXX: need scope argument.
333 if (IPSEC_ENABLED(ipv6)) {
334 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
335 if (error == EINPROGRESS)
342 bzero(&exthdrs, sizeof(exthdrs));
344 /* Hop-by-Hop options header */
345 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
346 /* Destination options header(1st part) */
347 if (opt->ip6po_rthdr) {
349 * Destination options header(1st part)
350 * This only makes sense with a routing header.
351 * See Section 9.2 of RFC 3542.
352 * Disabling this part just for MIP6 convenience is
353 * a bad idea. We need to think carefully about a
354 * way to make the advanced API coexist with MIP6
355 * options, which might automatically be inserted in
358 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
361 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
362 /* Destination options header(2nd part) */
363 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
367 * Calculate the total length of the extension header chain.
368 * Keep the length of the unfragmentable part for fragmentation.
371 if (exthdrs.ip6e_hbh)
372 optlen += exthdrs.ip6e_hbh->m_len;
373 if (exthdrs.ip6e_dest1)
374 optlen += exthdrs.ip6e_dest1->m_len;
375 if (exthdrs.ip6e_rthdr)
376 optlen += exthdrs.ip6e_rthdr->m_len;
377 unfragpartlen = optlen + sizeof(struct ip6_hdr);
379 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
380 if (exthdrs.ip6e_dest2)
381 optlen += exthdrs.ip6e_dest2->m_len;
384 * If there is at least one extension header,
385 * separate IP6 header from the payload.
387 if (optlen && !hdrsplit) {
388 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
392 m = exthdrs.ip6e_ip6;
396 ip6 = mtod(m, struct ip6_hdr *);
398 /* adjust mbuf packet header length */
399 m->m_pkthdr.len += optlen;
400 plen = m->m_pkthdr.len - sizeof(*ip6);
402 /* If this is a jumbo payload, insert a jumbo payload option. */
403 if (plen > IPV6_MAXPACKET) {
405 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
409 m = exthdrs.ip6e_ip6;
413 ip6 = mtod(m, struct ip6_hdr *);
414 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
418 ip6->ip6_plen = htons(plen);
421 * Concatenate headers and fill in next header fields.
422 * Here we have, on "m"
424 * and we insert headers accordingly. Finally, we should be getting:
425 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
427 * during the header composing process, "m" points to IPv6 header.
428 * "mprev" points to an extension header prior to esp.
430 u_char *nexthdrp = &ip6->ip6_nxt;
434 * we treat dest2 specially. this makes IPsec processing
435 * much easier. the goal here is to make mprev point the
436 * mbuf prior to dest2.
438 * result: IPv6 dest2 payload
439 * m and mprev will point to IPv6 header.
441 if (exthdrs.ip6e_dest2) {
443 panic("assumption failed: hdr not split");
444 exthdrs.ip6e_dest2->m_next = m->m_next;
445 m->m_next = exthdrs.ip6e_dest2;
446 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
447 ip6->ip6_nxt = IPPROTO_DSTOPTS;
451 * result: IPv6 hbh dest1 rthdr dest2 payload
452 * m will point to IPv6 header. mprev will point to the
453 * extension header prior to dest2 (rthdr in the above case).
455 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
456 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
458 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
462 * If there is a routing header, discard the packet.
464 if (exthdrs.ip6e_rthdr) {
469 /* Source address validation */
470 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
471 (flags & IPV6_UNSPECSRC) == 0) {
473 IP6STAT_INC(ip6s_badscope);
476 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
478 IP6STAT_INC(ip6s_badscope);
482 IP6STAT_INC(ip6s_localout);
489 bzero((caddr_t)ro, sizeof(*ro));
492 if (opt && opt->ip6po_rthdr)
493 ro = &opt->ip6po_route;
494 dst = (struct sockaddr_in6 *)&ro->ro_dst;
495 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
498 * if specified, try to fill in the traffic class field.
499 * do not override if a non-zero value is already set.
500 * we check the diffserv field and the ecn field separately.
502 if (opt && opt->ip6po_tclass >= 0) {
505 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
507 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
510 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
513 /* fill in or override the hop limit field, if necessary. */
514 if (opt && opt->ip6po_hlim != -1)
515 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
516 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
518 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
520 ip6->ip6_hlim = V_ip6_defmcasthlim;
523 * Validate route against routing table additions;
524 * a better/more specific route might have been added.
525 * Make sure address family is set in route.
528 ro->ro_dst.sin6_family = AF_INET6;
529 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
531 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
532 ro->ro_dst.sin6_family == AF_INET6 &&
533 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
535 ifp = ro->ro_rt->rt_ifp;
538 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
540 if (fwd_tag == NULL) {
541 bzero(&dst_sa, sizeof(dst_sa));
542 dst_sa.sin6_family = AF_INET6;
543 dst_sa.sin6_len = sizeof(dst_sa);
544 dst_sa.sin6_addr = ip6->ip6_dst;
546 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
550 in6_ifstat_inc(ifp, ifs6_out_discard);
556 * If in6_selectroute() does not return a route entry,
557 * dst may not have been updated.
559 *dst = dst_sa; /* XXX */
563 * then rt (for unicast) and ifp must be non-NULL valid values.
565 if ((flags & IPV6_FORWARDING) == 0) {
566 /* XXX: the FORWARDING flag can be set for mrouting. */
567 in6_ifstat_inc(ifp, ifs6_out_request);
570 ia = (struct in6_ifaddr *)(rt->rt_ifa);
571 counter_u64_add(rt->rt_pksent, 1);
576 * The outgoing interface must be in the zone of source and
577 * destination addresses.
582 if (in6_setscope(&src0, origifp, &zone))
584 bzero(&src_sa, sizeof(src_sa));
585 src_sa.sin6_family = AF_INET6;
586 src_sa.sin6_len = sizeof(src_sa);
587 src_sa.sin6_addr = ip6->ip6_src;
588 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
592 if (in6_setscope(&dst0, origifp, &zone))
594 /* re-initialize to be sure */
595 bzero(&dst_sa, sizeof(dst_sa));
596 dst_sa.sin6_family = AF_INET6;
597 dst_sa.sin6_len = sizeof(dst_sa);
598 dst_sa.sin6_addr = ip6->ip6_dst;
599 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
603 /* We should use ia_ifp to support the case of
604 * sending packets to an address of our own.
606 if (ia != NULL && ia->ia_ifp)
609 /* scope check is done. */
613 IP6STAT_INC(ip6s_badscope);
614 in6_ifstat_inc(origifp, ifs6_out_discard);
616 error = EHOSTUNREACH; /* XXX */
620 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
621 if (opt && opt->ip6po_nextroute.ro_rt) {
623 * The nexthop is explicitly specified by the
624 * application. We assume the next hop is an IPv6
627 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
629 else if ((rt->rt_flags & RTF_GATEWAY))
630 dst = (struct sockaddr_in6 *)rt->rt_gateway;
633 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
634 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
636 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
637 in6_ifstat_inc(ifp, ifs6_out_mcast);
639 * Confirm that the outgoing interface supports multicast.
641 if (!(ifp->if_flags & IFF_MULTICAST)) {
642 IP6STAT_INC(ip6s_noroute);
643 in6_ifstat_inc(ifp, ifs6_out_discard);
647 if ((im6o == NULL && in6_mcast_loop) ||
648 (im6o && im6o->im6o_multicast_loop)) {
650 * Loop back multicast datagram if not expressly
651 * forbidden to do so, even if we have not joined
652 * the address; protocols will filter it later,
653 * thus deferring a hash lookup and lock acquisition
654 * at the expense of an m_copym().
656 ip6_mloopback(ifp, m);
659 * If we are acting as a multicast router, perform
660 * multicast forwarding as if the packet had just
661 * arrived on the interface to which we are about
662 * to send. The multicast forwarding function
663 * recursively calls this function, using the
664 * IPV6_FORWARDING flag to prevent infinite recursion.
666 * Multicasts that are looped back by ip6_mloopback(),
667 * above, will be forwarded by the ip6_input() routine,
670 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
672 * XXX: ip6_mforward expects that rcvif is NULL
673 * when it is called from the originating path.
674 * However, it may not always be the case.
676 m->m_pkthdr.rcvif = NULL;
677 if (ip6_mforward(ip6, ifp, m) != 0) {
684 * Multicasts with a hoplimit of zero may be looped back,
685 * above, but must not be transmitted on a network.
686 * Also, multicasts addressed to the loopback interface
687 * are not sent -- the above call to ip6_mloopback() will
688 * loop back a copy if this host actually belongs to the
689 * destination group on the loopback interface.
691 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
692 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
699 * Fill the outgoing inteface to tell the upper layer
700 * to increment per-interface statistics.
705 /* Determine path MTU. */
706 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
707 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
711 * The caller of this function may specify to use the minimum MTU
713 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
714 * setting. The logic is a bit complicated; by default, unicast
715 * packets will follow path MTU while multicast packets will be sent at
716 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
717 * including unicast ones will be sent at the minimum MTU. Multicast
718 * packets will always be sent at the minimum MTU unless
719 * IP6PO_MINMTU_DISABLE is explicitly specified.
720 * See RFC 3542 for more details.
722 if (mtu > IPV6_MMTU) {
723 if ((flags & IPV6_MINMTU))
725 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
727 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
729 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
735 * clear embedded scope identifiers if necessary.
736 * in6_clearscope will touch the addresses only when necessary.
738 in6_clearscope(&ip6->ip6_src);
739 in6_clearscope(&ip6->ip6_dst);
742 * If the outgoing packet contains a hop-by-hop options header,
743 * it must be examined and processed even by the source node.
744 * (RFC 2460, section 4.)
746 if (exthdrs.ip6e_hbh) {
747 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
748 u_int32_t dummy; /* XXX unused */
749 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
752 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
753 panic("ip6e_hbh is not contiguous");
756 * XXX: if we have to send an ICMPv6 error to the sender,
757 * we need the M_LOOP flag since icmp6_error() expects
758 * the IPv6 and the hop-by-hop options header are
759 * contiguous unless the flag is set.
761 m->m_flags |= M_LOOP;
762 m->m_pkthdr.rcvif = ifp;
763 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
764 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
765 &dummy, &plen) < 0) {
766 /* m was already freed at this point */
767 error = EINVAL;/* better error? */
770 m->m_flags &= ~M_LOOP; /* XXX */
771 m->m_pkthdr.rcvif = NULL;
774 /* Jump over all PFIL processing if hooks are not active. */
775 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
779 /* Run through list of hooks for output packets. */
780 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp);
781 if (error != 0 || m == NULL)
784 ip6 = mtod(m, struct ip6_hdr *);
787 /* See if destination IP address was changed by packet filter. */
788 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
789 m->m_flags |= M_SKIP_FIREWALL;
790 /* If destination is now ourself drop to ip6_input(). */
791 if (in6_localip(&ip6->ip6_dst)) {
792 m->m_flags |= M_FASTFWD_OURS;
793 if (m->m_pkthdr.rcvif == NULL)
794 m->m_pkthdr.rcvif = V_loif;
795 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
796 m->m_pkthdr.csum_flags |=
797 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
798 m->m_pkthdr.csum_data = 0xffff;
801 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
802 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
804 error = netisr_queue(NETISR_IPV6, m);
808 needfiblookup = 1; /* Redo the routing table lookup. */
810 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
814 /* See if fib was changed by packet filter. */
815 if (fibnum != M_GETFIB(m)) {
816 m->m_flags |= M_SKIP_FIREWALL;
817 fibnum = M_GETFIB(m);
821 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
827 /* See if local, if yes, send it to netisr. */
828 if (m->m_flags & M_FASTFWD_OURS) {
829 if (m->m_pkthdr.rcvif == NULL)
830 m->m_pkthdr.rcvif = V_loif;
831 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
832 m->m_pkthdr.csum_flags |=
833 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
834 m->m_pkthdr.csum_data = 0xffff;
837 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
838 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
840 error = netisr_queue(NETISR_IPV6, m);
843 /* Or forward to some other address? */
844 if ((m->m_flags & M_IP6_NEXTHOP) &&
845 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
846 dst = (struct sockaddr_in6 *)&ro->ro_dst;
847 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
848 m->m_flags |= M_SKIP_FIREWALL;
849 m->m_flags &= ~M_IP6_NEXTHOP;
850 m_tag_delete(m, fwd_tag);
856 * Send the packet to the outgoing interface.
857 * If necessary, do IPv6 fragmentation before sending.
859 * the logic here is rather complex:
860 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
861 * 1-a: send as is if tlen <= path mtu
862 * 1-b: fragment if tlen > path mtu
864 * 2: if user asks us not to fragment (dontfrag == 1)
865 * 2-a: send as is if tlen <= interface mtu
866 * 2-b: error if tlen > interface mtu
868 * 3: if we always need to attach fragment header (alwaysfrag == 1)
871 * 4: if dontfrag == 1 && alwaysfrag == 1
872 * error, as we cannot handle this conflicting request
874 sw_csum = m->m_pkthdr.csum_flags;
876 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
877 sw_csum &= ~ifp->if_hwassist;
881 * If we added extension headers, we will not do TSO and calculate the
882 * checksums ourselves for now.
883 * XXX-BZ Need a framework to know when the NIC can handle it, even
886 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
887 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
888 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
891 if (sw_csum & CSUM_SCTP_IPV6) {
892 sw_csum &= ~CSUM_SCTP_IPV6;
893 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
896 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
897 tlen = m->m_pkthdr.len;
899 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
903 if (dontfrag && alwaysfrag) { /* case 4 */
904 /* conflicting request - can't transmit */
908 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
910 * Even if the DONTFRAG option is specified, we cannot send the
911 * packet when the data length is larger than the MTU of the
912 * outgoing interface.
913 * Notify the error by sending IPV6_PATHMTU ancillary data if
914 * application wanted to know the MTU value. Also return an
915 * error code (this is not described in the API spec).
918 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
924 * transmit packet without fragmentation
926 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
927 struct in6_ifaddr *ia6;
929 ip6 = mtod(m, struct ip6_hdr *);
930 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
932 /* Record statistics for this interface address. */
933 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
934 counter_u64_add(ia6->ia_ifa.ifa_obytes,
936 ifa_free(&ia6->ia_ifa);
940 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
941 in_pcboutput_txrtlmt(inp, ifp, m);
942 /* stamp send tag on mbuf */
943 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
945 m->m_pkthdr.snd_tag = NULL;
948 error = nd6_output_ifp(ifp, origifp, m, dst,
951 /* check for route change */
953 in_pcboutput_eagain(inp);
959 * try to fragment the packet. case 1-b and 3
961 if (mtu < IPV6_MMTU) {
962 /* path MTU cannot be less than IPV6_MMTU */
964 in6_ifstat_inc(ifp, ifs6_out_fragfail);
966 } else if (ip6->ip6_plen == 0) {
967 /* jumbo payload cannot be fragmented */
969 in6_ifstat_inc(ifp, ifs6_out_fragfail);
975 * Too large for the destination or interface;
976 * fragment if possible.
977 * Must be able to put at least 8 bytes per fragment.
979 hlen = unfragpartlen;
980 if (mtu > IPV6_MAXPACKET)
981 mtu = IPV6_MAXPACKET;
983 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
986 in6_ifstat_inc(ifp, ifs6_out_fragfail);
991 * If the interface will not calculate checksums on
992 * fragmented packets, then do it here.
993 * XXX-BZ handle the hw offloading case. Need flags.
995 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
996 in6_delayed_cksum(m, plen, hlen);
997 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1000 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1001 sctp_delayed_cksum(m, hlen);
1002 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1006 * Change the next header field of the last header in the
1007 * unfragmentable part.
1009 if (exthdrs.ip6e_rthdr) {
1010 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1011 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1012 } else if (exthdrs.ip6e_dest1) {
1013 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1014 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1015 } else if (exthdrs.ip6e_hbh) {
1016 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1017 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1019 nextproto = ip6->ip6_nxt;
1020 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1024 * Loop through length of segment after first fragment,
1025 * make new header and copy data of each part and link onto
1029 id = htonl(ip6_randomid());
1030 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1033 in6_ifstat_inc(ifp, ifs6_out_fragok);
1037 * Remove leading garbages.
1047 /* Record statistics for this interface address. */
1049 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1050 counter_u64_add(ia->ia_ifa.ifa_obytes,
1055 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1056 in_pcboutput_txrtlmt(inp, ifp, m);
1057 /* stamp send tag on mbuf */
1058 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1060 m->m_pkthdr.snd_tag = NULL;
1063 error = nd6_output_ifp(ifp, origifp, m, dst,
1064 (struct route *)ro);
1066 /* check for route change */
1067 if (error == EAGAIN)
1068 in_pcboutput_eagain(inp);
1075 IP6STAT_INC(ip6s_fragmented);
1078 if (ro == &ip6route)
1083 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1084 m_freem(exthdrs.ip6e_dest1);
1085 m_freem(exthdrs.ip6e_rthdr);
1086 m_freem(exthdrs.ip6e_dest2);
1095 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1099 if (hlen > MCLBYTES)
1100 return (ENOBUFS); /* XXX */
1103 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1105 m = m_get(M_NOWAIT, MT_DATA);
1110 bcopy(hdr, mtod(m, caddr_t), hlen);
1117 * Insert jumbo payload option.
1120 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1126 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1129 * If there is no hop-by-hop options header, allocate new one.
1130 * If there is one but it doesn't have enough space to store the
1131 * jumbo payload option, allocate a cluster to store the whole options.
1132 * Otherwise, use it to store the options.
1134 if (exthdrs->ip6e_hbh == NULL) {
1135 mopt = m_get(M_NOWAIT, MT_DATA);
1138 mopt->m_len = JUMBOOPTLEN;
1139 optbuf = mtod(mopt, u_char *);
1140 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1141 exthdrs->ip6e_hbh = mopt;
1143 struct ip6_hbh *hbh;
1145 mopt = exthdrs->ip6e_hbh;
1146 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1149 * - exthdrs->ip6e_hbh is not referenced from places
1150 * other than exthdrs.
1151 * - exthdrs->ip6e_hbh is not an mbuf chain.
1153 int oldoptlen = mopt->m_len;
1157 * XXX: give up if the whole (new) hbh header does
1158 * not fit even in an mbuf cluster.
1160 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1164 * As a consequence, we must always prepare a cluster
1167 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1170 n->m_len = oldoptlen + JUMBOOPTLEN;
1171 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1173 optbuf = mtod(n, caddr_t) + oldoptlen;
1175 mopt = exthdrs->ip6e_hbh = n;
1177 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1178 mopt->m_len += JUMBOOPTLEN;
1180 optbuf[0] = IP6OPT_PADN;
1184 * Adjust the header length according to the pad and
1185 * the jumbo payload option.
1187 hbh = mtod(mopt, struct ip6_hbh *);
1188 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1191 /* fill in the option. */
1192 optbuf[2] = IP6OPT_JUMBO;
1194 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1195 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1197 /* finally, adjust the packet header length */
1198 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1205 * Insert fragment header and copy unfragmentable header portions.
1208 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1209 struct ip6_frag **frghdrp)
1211 struct mbuf *n, *mlast;
1213 if (hlen > sizeof(struct ip6_hdr)) {
1214 n = m_copym(m0, sizeof(struct ip6_hdr),
1215 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1222 /* Search for the last mbuf of unfragmentable part. */
1223 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1226 if (M_WRITABLE(mlast) &&
1227 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1228 /* use the trailing space of the last mbuf for the fragment hdr */
1229 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1231 mlast->m_len += sizeof(struct ip6_frag);
1232 m->m_pkthdr.len += sizeof(struct ip6_frag);
1234 /* allocate a new mbuf for the fragment header */
1237 mfrg = m_get(M_NOWAIT, MT_DATA);
1240 mfrg->m_len = sizeof(struct ip6_frag);
1241 *frghdrp = mtod(mfrg, struct ip6_frag *);
1242 mlast->m_next = mfrg;
1249 * Calculates IPv6 path mtu for destination @dst.
1250 * Resulting MTU is stored in @mtup.
1252 * Returns 0 on success.
1255 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1257 struct nhop6_extended nh6;
1258 struct in6_addr kdst;
1264 in6_splitscope(dst, &kdst, &scopeid);
1265 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1266 return (EHOSTUNREACH);
1271 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1272 fib6_free_nh_ext(fibnum, &nh6);
1278 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1279 * and cached data in @ro_pmtu.
1280 * MTU from (successful) route lookup is saved (along with dst)
1281 * inside @ro_pmtu to avoid subsequent route lookups after packet
1282 * filter processing.
1284 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1285 * Returns 0 on success.
1288 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1289 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1290 int *alwaysfragp, u_int fibnum, u_int proto)
1292 struct nhop6_basic nh6;
1293 struct in6_addr kdst;
1295 struct sockaddr_in6 *sa6_dst;
1302 * Here ro_pmtu has final destination address, while
1303 * ro might represent immediate destination.
1304 * Use ro_pmtu destination since mtu might differ.
1306 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1307 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1308 ro_pmtu->ro_mtu = 0;
1310 if (ro_pmtu->ro_mtu == 0) {
1311 bzero(sa6_dst, sizeof(*sa6_dst));
1312 sa6_dst->sin6_family = AF_INET6;
1313 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1314 sa6_dst->sin6_addr = *dst;
1316 in6_splitscope(dst, &kdst, &scopeid);
1317 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1319 ro_pmtu->ro_mtu = nh6.nh_mtu;
1322 mtu = ro_pmtu->ro_mtu;
1326 mtu = ro_pmtu->ro_rt->rt_mtu;
1328 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1332 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1333 * hostcache data for @dst.
1334 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1336 * Returns 0 on success.
1339 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1340 u_long *mtup, int *alwaysfragp, u_int proto)
1348 struct in_conninfo inc;
1350 bzero(&inc, sizeof(inc));
1351 inc.inc_flags |= INC_ISIPV6;
1352 inc.inc6_faddr = *dst;
1354 ifmtu = IN6_LINKMTU(ifp);
1356 /* TCP is known to react to pmtu changes so skip hc */
1357 if (proto != IPPROTO_TCP)
1358 mtu = tcp_hc_getmtu(&inc);
1361 mtu = min(mtu, rt_mtu);
1366 else if (mtu < IPV6_MMTU) {
1368 * RFC2460 section 5, last paragraph:
1369 * if we record ICMPv6 too big message with
1370 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1371 * or smaller, with framgent header attached.
1372 * (fragment header is needed regardless from the
1373 * packet size, for translators to identify packets)
1379 mtu = IN6_LINKMTU(ifp);
1381 error = EHOSTUNREACH; /* XXX */
1385 *alwaysfragp = alwaysfrag;
1390 * IP6 socket option processing.
1393 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1395 int optdatalen, uproto;
1397 struct inpcb *in6p = sotoinpcb(so);
1399 int level, op, optname;
1403 uint32_t rss_bucket;
1408 * Don't use more than a quarter of mbuf clusters. N.B.:
1409 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1410 * on LP64 architectures, so cast to u_long to avoid undefined
1411 * behavior. ILP32 architectures cannot have nmbclusters
1412 * large enough to overflow for other reasons.
1414 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1416 level = sopt->sopt_level;
1417 op = sopt->sopt_dir;
1418 optname = sopt->sopt_name;
1419 optlen = sopt->sopt_valsize;
1423 uproto = (int)so->so_proto->pr_protocol;
1425 if (level != IPPROTO_IPV6) {
1428 if (sopt->sopt_level == SOL_SOCKET &&
1429 sopt->sopt_dir == SOPT_SET) {
1430 switch (sopt->sopt_name) {
1433 if ((so->so_options & SO_REUSEADDR) != 0)
1434 in6p->inp_flags2 |= INP_REUSEADDR;
1436 in6p->inp_flags2 &= ~INP_REUSEADDR;
1442 if ((so->so_options & SO_REUSEPORT) != 0)
1443 in6p->inp_flags2 |= INP_REUSEPORT;
1445 in6p->inp_flags2 &= ~INP_REUSEPORT;
1449 case SO_REUSEPORT_LB:
1451 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1452 in6p->inp_flags2 |= INP_REUSEPORT_LB;
1454 in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1460 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1464 case SO_MAX_PACING_RATE:
1467 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1478 } else { /* level == IPPROTO_IPV6 */
1483 case IPV6_2292PKTOPTIONS:
1484 #ifdef IPV6_PKTOPTIONS
1485 case IPV6_PKTOPTIONS:
1490 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1491 printf("ip6_ctloutput: mbuf limit hit\n");
1496 error = soopt_getm(sopt, &m); /* XXX */
1499 error = soopt_mcopyin(sopt, m); /* XXX */
1502 error = ip6_pcbopts(&in6p->in6p_outputopts,
1504 m_freem(m); /* XXX */
1509 * Use of some Hop-by-Hop options or some
1510 * Destination options, might require special
1511 * privilege. That is, normal applications
1512 * (without special privilege) might be forbidden
1513 * from setting certain options in outgoing packets,
1514 * and might never see certain options in received
1515 * packets. [RFC 2292 Section 6]
1516 * KAME specific note:
1517 * KAME prevents non-privileged users from sending or
1518 * receiving ANY hbh/dst options in order to avoid
1519 * overhead of parsing options in the kernel.
1521 case IPV6_RECVHOPOPTS:
1522 case IPV6_RECVDSTOPTS:
1523 case IPV6_RECVRTHDRDSTOPTS:
1525 error = priv_check(td,
1526 PRIV_NETINET_SETHDROPTS);
1531 case IPV6_UNICAST_HOPS:
1534 case IPV6_RECVPKTINFO:
1535 case IPV6_RECVHOPLIMIT:
1536 case IPV6_RECVRTHDR:
1537 case IPV6_RECVPATHMTU:
1538 case IPV6_RECVTCLASS:
1539 case IPV6_RECVFLOWID:
1541 case IPV6_RECVRSSBUCKETID:
1544 case IPV6_AUTOFLOWLABEL:
1545 case IPV6_ORIGDSTADDR:
1547 case IPV6_BINDMULTI:
1549 case IPV6_RSS_LISTEN_BUCKET:
1551 if (optname == IPV6_BINDANY && td != NULL) {
1552 error = priv_check(td,
1553 PRIV_NETINET_BINDANY);
1558 if (optlen != sizeof(int)) {
1562 error = sooptcopyin(sopt, &optval,
1563 sizeof optval, sizeof optval);
1568 case IPV6_UNICAST_HOPS:
1569 if (optval < -1 || optval >= 256)
1572 /* -1 = kernel default */
1573 in6p->in6p_hops = optval;
1574 if ((in6p->inp_vflag &
1576 in6p->inp_ip_ttl = optval;
1579 #define OPTSET(bit) \
1583 in6p->inp_flags |= (bit); \
1585 in6p->inp_flags &= ~(bit); \
1586 INP_WUNLOCK(in6p); \
1587 } while (/*CONSTCOND*/ 0)
1588 #define OPTSET2292(bit) \
1591 in6p->inp_flags |= IN6P_RFC2292; \
1593 in6p->inp_flags |= (bit); \
1595 in6p->inp_flags &= ~(bit); \
1596 INP_WUNLOCK(in6p); \
1597 } while (/*CONSTCOND*/ 0)
1598 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1600 #define OPTSET2_N(bit, val) do { \
1602 in6p->inp_flags2 |= bit; \
1604 in6p->inp_flags2 &= ~bit; \
1606 #define OPTSET2(bit, val) do { \
1608 OPTSET2_N(bit, val); \
1609 INP_WUNLOCK(in6p); \
1611 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1612 #define OPTSET2292_EXCLUSIVE(bit) \
1615 if (OPTBIT(IN6P_RFC2292)) { \
1619 in6p->inp_flags |= (bit); \
1621 in6p->inp_flags &= ~(bit); \
1623 INP_WUNLOCK(in6p); \
1624 } while (/*CONSTCOND*/ 0)
1626 case IPV6_RECVPKTINFO:
1627 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1632 struct ip6_pktopts **optp;
1634 /* cannot mix with RFC2292 */
1635 if (OPTBIT(IN6P_RFC2292)) {
1640 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1642 return (ECONNRESET);
1644 optp = &in6p->in6p_outputopts;
1645 error = ip6_pcbopt(IPV6_HOPLIMIT,
1646 (u_char *)&optval, sizeof(optval),
1647 optp, (td != NULL) ? td->td_ucred :
1653 case IPV6_RECVHOPLIMIT:
1654 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1657 case IPV6_RECVHOPOPTS:
1658 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1661 case IPV6_RECVDSTOPTS:
1662 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1665 case IPV6_RECVRTHDRDSTOPTS:
1666 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1669 case IPV6_RECVRTHDR:
1670 OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1673 case IPV6_RECVPATHMTU:
1675 * We ignore this option for TCP
1677 * (RFC3542 leaves this case
1680 if (uproto != IPPROTO_TCP)
1684 case IPV6_RECVFLOWID:
1685 OPTSET2(INP_RECVFLOWID, optval);
1689 case IPV6_RECVRSSBUCKETID:
1690 OPTSET2(INP_RECVRSSBUCKETID, optval);
1696 * make setsockopt(IPV6_V6ONLY)
1697 * available only prior to bind(2).
1698 * see ipng mailing list, Jun 22 2001.
1700 if (in6p->inp_lport ||
1701 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1705 OPTSET(IN6P_IPV6_V6ONLY);
1707 in6p->inp_vflag &= ~INP_IPV4;
1709 in6p->inp_vflag |= INP_IPV4;
1711 case IPV6_RECVTCLASS:
1712 /* cannot mix with RFC2292 XXX */
1713 OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1715 case IPV6_AUTOFLOWLABEL:
1716 OPTSET(IN6P_AUTOFLOWLABEL);
1719 case IPV6_ORIGDSTADDR:
1720 OPTSET2(INP_ORIGDSTADDR, optval);
1723 OPTSET(INP_BINDANY);
1726 case IPV6_BINDMULTI:
1727 OPTSET2(INP_BINDMULTI, optval);
1730 case IPV6_RSS_LISTEN_BUCKET:
1731 if ((optval >= 0) &&
1732 (optval < rss_getnumbuckets())) {
1734 in6p->inp_rss_listen_bucket = optval;
1735 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1747 case IPV6_USE_MIN_MTU:
1748 case IPV6_PREFER_TEMPADDR:
1749 if (optlen != sizeof(optval)) {
1753 error = sooptcopyin(sopt, &optval,
1754 sizeof optval, sizeof optval);
1758 struct ip6_pktopts **optp;
1760 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1762 return (ECONNRESET);
1764 optp = &in6p->in6p_outputopts;
1765 error = ip6_pcbopt(optname,
1766 (u_char *)&optval, sizeof(optval),
1767 optp, (td != NULL) ? td->td_ucred :
1773 case IPV6_2292PKTINFO:
1774 case IPV6_2292HOPLIMIT:
1775 case IPV6_2292HOPOPTS:
1776 case IPV6_2292DSTOPTS:
1777 case IPV6_2292RTHDR:
1779 if (optlen != sizeof(int)) {
1783 error = sooptcopyin(sopt, &optval,
1784 sizeof optval, sizeof optval);
1788 case IPV6_2292PKTINFO:
1789 OPTSET2292(IN6P_PKTINFO);
1791 case IPV6_2292HOPLIMIT:
1792 OPTSET2292(IN6P_HOPLIMIT);
1794 case IPV6_2292HOPOPTS:
1796 * Check super-user privilege.
1797 * See comments for IPV6_RECVHOPOPTS.
1800 error = priv_check(td,
1801 PRIV_NETINET_SETHDROPTS);
1805 OPTSET2292(IN6P_HOPOPTS);
1807 case IPV6_2292DSTOPTS:
1809 error = priv_check(td,
1810 PRIV_NETINET_SETHDROPTS);
1814 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1816 case IPV6_2292RTHDR:
1817 OPTSET2292(IN6P_RTHDR);
1825 case IPV6_RTHDRDSTOPTS:
1828 /* new advanced API (RFC3542) */
1830 u_char optbuf_storage[MCLBYTES];
1832 struct ip6_pktopts **optp;
1834 /* cannot mix with RFC2292 */
1835 if (OPTBIT(IN6P_RFC2292)) {
1841 * We only ensure valsize is not too large
1842 * here. Further validation will be done
1845 error = sooptcopyin(sopt, optbuf_storage,
1846 sizeof(optbuf_storage), 0);
1849 optlen = sopt->sopt_valsize;
1850 optbuf = optbuf_storage;
1852 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1854 return (ECONNRESET);
1856 optp = &in6p->in6p_outputopts;
1857 error = ip6_pcbopt(optname, optbuf, optlen,
1858 optp, (td != NULL) ? td->td_ucred : NULL,
1865 case IPV6_MULTICAST_IF:
1866 case IPV6_MULTICAST_HOPS:
1867 case IPV6_MULTICAST_LOOP:
1868 case IPV6_JOIN_GROUP:
1869 case IPV6_LEAVE_GROUP:
1871 case MCAST_BLOCK_SOURCE:
1872 case MCAST_UNBLOCK_SOURCE:
1873 case MCAST_JOIN_GROUP:
1874 case MCAST_LEAVE_GROUP:
1875 case MCAST_JOIN_SOURCE_GROUP:
1876 case MCAST_LEAVE_SOURCE_GROUP:
1877 error = ip6_setmoptions(in6p, sopt);
1880 case IPV6_PORTRANGE:
1881 error = sooptcopyin(sopt, &optval,
1882 sizeof optval, sizeof optval);
1888 case IPV6_PORTRANGE_DEFAULT:
1889 in6p->inp_flags &= ~(INP_LOWPORT);
1890 in6p->inp_flags &= ~(INP_HIGHPORT);
1893 case IPV6_PORTRANGE_HIGH:
1894 in6p->inp_flags &= ~(INP_LOWPORT);
1895 in6p->inp_flags |= INP_HIGHPORT;
1898 case IPV6_PORTRANGE_LOW:
1899 in6p->inp_flags &= ~(INP_HIGHPORT);
1900 in6p->inp_flags |= INP_LOWPORT;
1910 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1911 case IPV6_IPSEC_POLICY:
1912 if (IPSEC_ENABLED(ipv6)) {
1913 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1920 error = ENOPROTOOPT;
1928 case IPV6_2292PKTOPTIONS:
1929 #ifdef IPV6_PKTOPTIONS
1930 case IPV6_PKTOPTIONS:
1933 * RFC3542 (effectively) deprecated the
1934 * semantics of the 2292-style pktoptions.
1935 * Since it was not reliable in nature (i.e.,
1936 * applications had to expect the lack of some
1937 * information after all), it would make sense
1938 * to simplify this part by always returning
1941 sopt->sopt_valsize = 0;
1944 case IPV6_RECVHOPOPTS:
1945 case IPV6_RECVDSTOPTS:
1946 case IPV6_RECVRTHDRDSTOPTS:
1947 case IPV6_UNICAST_HOPS:
1948 case IPV6_RECVPKTINFO:
1949 case IPV6_RECVHOPLIMIT:
1950 case IPV6_RECVRTHDR:
1951 case IPV6_RECVPATHMTU:
1954 case IPV6_PORTRANGE:
1955 case IPV6_RECVTCLASS:
1956 case IPV6_AUTOFLOWLABEL:
1960 case IPV6_RECVFLOWID:
1962 case IPV6_RSSBUCKETID:
1963 case IPV6_RECVRSSBUCKETID:
1965 case IPV6_BINDMULTI:
1968 case IPV6_RECVHOPOPTS:
1969 optval = OPTBIT(IN6P_HOPOPTS);
1972 case IPV6_RECVDSTOPTS:
1973 optval = OPTBIT(IN6P_DSTOPTS);
1976 case IPV6_RECVRTHDRDSTOPTS:
1977 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1980 case IPV6_UNICAST_HOPS:
1981 optval = in6p->in6p_hops;
1984 case IPV6_RECVPKTINFO:
1985 optval = OPTBIT(IN6P_PKTINFO);
1988 case IPV6_RECVHOPLIMIT:
1989 optval = OPTBIT(IN6P_HOPLIMIT);
1992 case IPV6_RECVRTHDR:
1993 optval = OPTBIT(IN6P_RTHDR);
1996 case IPV6_RECVPATHMTU:
1997 optval = OPTBIT(IN6P_MTU);
2001 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2004 case IPV6_PORTRANGE:
2007 flags = in6p->inp_flags;
2008 if (flags & INP_HIGHPORT)
2009 optval = IPV6_PORTRANGE_HIGH;
2010 else if (flags & INP_LOWPORT)
2011 optval = IPV6_PORTRANGE_LOW;
2016 case IPV6_RECVTCLASS:
2017 optval = OPTBIT(IN6P_TCLASS);
2020 case IPV6_AUTOFLOWLABEL:
2021 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2024 case IPV6_ORIGDSTADDR:
2025 optval = OPTBIT2(INP_ORIGDSTADDR);
2029 optval = OPTBIT(INP_BINDANY);
2033 optval = in6p->inp_flowid;
2037 optval = in6p->inp_flowtype;
2040 case IPV6_RECVFLOWID:
2041 optval = OPTBIT2(INP_RECVFLOWID);
2044 case IPV6_RSSBUCKETID:
2046 rss_hash2bucket(in6p->inp_flowid,
2050 optval = rss_bucket;
2055 case IPV6_RECVRSSBUCKETID:
2056 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2060 case IPV6_BINDMULTI:
2061 optval = OPTBIT2(INP_BINDMULTI);
2067 error = sooptcopyout(sopt, &optval,
2074 struct ip6_mtuinfo mtuinfo;
2075 struct in6_addr addr;
2077 if (!(so->so_state & SS_ISCONNECTED))
2080 * XXX: we dot not consider the case of source
2081 * routing, or optional information to specify
2082 * the outgoing interface.
2083 * Copy faddr out of in6p to avoid holding lock
2084 * on inp during route lookup.
2087 bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2089 error = ip6_getpmtu_ctl(so->so_fibnum,
2093 if (pmtu > IPV6_MAXPACKET)
2094 pmtu = IPV6_MAXPACKET;
2096 bzero(&mtuinfo, sizeof(mtuinfo));
2097 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2098 optdata = (void *)&mtuinfo;
2099 optdatalen = sizeof(mtuinfo);
2100 error = sooptcopyout(sopt, optdata,
2105 case IPV6_2292PKTINFO:
2106 case IPV6_2292HOPLIMIT:
2107 case IPV6_2292HOPOPTS:
2108 case IPV6_2292RTHDR:
2109 case IPV6_2292DSTOPTS:
2111 case IPV6_2292PKTINFO:
2112 optval = OPTBIT(IN6P_PKTINFO);
2114 case IPV6_2292HOPLIMIT:
2115 optval = OPTBIT(IN6P_HOPLIMIT);
2117 case IPV6_2292HOPOPTS:
2118 optval = OPTBIT(IN6P_HOPOPTS);
2120 case IPV6_2292RTHDR:
2121 optval = OPTBIT(IN6P_RTHDR);
2123 case IPV6_2292DSTOPTS:
2124 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2127 error = sooptcopyout(sopt, &optval,
2134 case IPV6_RTHDRDSTOPTS:
2138 case IPV6_USE_MIN_MTU:
2139 case IPV6_PREFER_TEMPADDR:
2140 error = ip6_getpcbopt(in6p, optname, sopt);
2143 case IPV6_MULTICAST_IF:
2144 case IPV6_MULTICAST_HOPS:
2145 case IPV6_MULTICAST_LOOP:
2147 error = ip6_getmoptions(in6p, sopt);
2150 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2151 case IPV6_IPSEC_POLICY:
2152 if (IPSEC_ENABLED(ipv6)) {
2153 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2159 error = ENOPROTOOPT;
2169 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2171 int error = 0, optval, optlen;
2172 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2173 struct inpcb *in6p = sotoinpcb(so);
2174 int level, op, optname;
2176 level = sopt->sopt_level;
2177 op = sopt->sopt_dir;
2178 optname = sopt->sopt_name;
2179 optlen = sopt->sopt_valsize;
2181 if (level != IPPROTO_IPV6) {
2188 * For ICMPv6 sockets, no modification allowed for checksum
2189 * offset, permit "no change" values to help existing apps.
2191 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2192 * for an ICMPv6 socket will fail."
2193 * The current behavior does not meet RFC3542.
2197 if (optlen != sizeof(int)) {
2201 error = sooptcopyin(sopt, &optval, sizeof(optval),
2205 if ((optval % 2) != 0) {
2206 /* the API assumes even offset values */
2208 } else if (so->so_proto->pr_protocol ==
2210 if (optval != icmp6off)
2213 in6p->in6p_cksum = optval;
2217 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2220 optval = in6p->in6p_cksum;
2222 error = sooptcopyout(sopt, &optval, sizeof(optval));
2232 error = ENOPROTOOPT;
2240 * Set up IP6 options in pcb for insertion in output packets or
2241 * specifying behavior of outgoing packets.
2244 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2245 struct socket *so, struct sockopt *sopt)
2247 struct ip6_pktopts *opt = *pktopt;
2249 struct thread *td = sopt->sopt_td;
2251 /* turn off any old options. */
2254 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2255 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2256 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2257 printf("ip6_pcbopts: all specified options are cleared.\n");
2259 ip6_clearpktopts(opt, -1);
2261 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2264 if (!m || m->m_len == 0) {
2266 * Only turning off any previous options, regardless of
2267 * whether the opt is just created or given.
2269 free(opt, M_IP6OPT);
2273 /* set options specified by user. */
2274 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2275 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2276 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2277 free(opt, M_IP6OPT);
2285 * initialize ip6_pktopts. beware that there are non-zero default values in
2289 ip6_initpktopts(struct ip6_pktopts *opt)
2292 bzero(opt, sizeof(*opt));
2293 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2294 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2295 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2296 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2300 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2301 struct ucred *cred, int uproto)
2303 struct ip6_pktopts *opt;
2305 if (*pktopt == NULL) {
2306 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2308 if (*pktopt == NULL)
2310 ip6_initpktopts(*pktopt);
2314 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2317 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2318 if (pktopt && pktopt->field) { \
2319 INP_RUNLOCK(in6p); \
2320 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2321 malloc_optdata = true; \
2323 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
2324 INP_RUNLOCK(in6p); \
2325 free(optdata, M_TEMP); \
2326 return (ECONNRESET); \
2328 pktopt = in6p->in6p_outputopts; \
2329 if (pktopt && pktopt->field) { \
2330 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2331 bcopy(&pktopt->field, optdata, optdatalen); \
2333 free(optdata, M_TEMP); \
2335 malloc_optdata = false; \
2340 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2341 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2343 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2344 pktopt->field->sa_len)
2347 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2349 void *optdata = NULL;
2350 bool malloc_optdata = false;
2353 struct in6_pktinfo null_pktinfo;
2354 int deftclass = 0, on;
2355 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2356 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2357 struct ip6_pktopts *pktopt;
2360 pktopt = in6p->in6p_outputopts;
2364 optdata = (void *)&null_pktinfo;
2365 if (pktopt && pktopt->ip6po_pktinfo) {
2366 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2367 sizeof(null_pktinfo));
2368 in6_clearscope(&null_pktinfo.ipi6_addr);
2370 /* XXX: we don't have to do this every time... */
2371 bzero(&null_pktinfo, sizeof(null_pktinfo));
2373 optdatalen = sizeof(struct in6_pktinfo);
2376 if (pktopt && pktopt->ip6po_tclass >= 0)
2377 deftclass = pktopt->ip6po_tclass;
2378 optdata = (void *)&deftclass;
2379 optdatalen = sizeof(int);
2382 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2385 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2387 case IPV6_RTHDRDSTOPTS:
2388 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2391 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2394 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2396 case IPV6_USE_MIN_MTU:
2398 defminmtu = pktopt->ip6po_minmtu;
2399 optdata = (void *)&defminmtu;
2400 optdatalen = sizeof(int);
2403 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2407 optdata = (void *)&on;
2408 optdatalen = sizeof(on);
2410 case IPV6_PREFER_TEMPADDR:
2412 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2413 optdata = (void *)&defpreftemp;
2414 optdatalen = sizeof(int);
2416 default: /* should not happen */
2418 panic("ip6_getpcbopt: unexpected option\n");
2421 return (ENOPROTOOPT);
2425 error = sooptcopyout(sopt, optdata, optdatalen);
2427 free(optdata, M_TEMP);
2433 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2438 if (optname == -1 || optname == IPV6_PKTINFO) {
2439 if (pktopt->ip6po_pktinfo)
2440 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2441 pktopt->ip6po_pktinfo = NULL;
2443 if (optname == -1 || optname == IPV6_HOPLIMIT)
2444 pktopt->ip6po_hlim = -1;
2445 if (optname == -1 || optname == IPV6_TCLASS)
2446 pktopt->ip6po_tclass = -1;
2447 if (optname == -1 || optname == IPV6_NEXTHOP) {
2448 if (pktopt->ip6po_nextroute.ro_rt) {
2449 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2450 pktopt->ip6po_nextroute.ro_rt = NULL;
2452 if (pktopt->ip6po_nexthop)
2453 free(pktopt->ip6po_nexthop, M_IP6OPT);
2454 pktopt->ip6po_nexthop = NULL;
2456 if (optname == -1 || optname == IPV6_HOPOPTS) {
2457 if (pktopt->ip6po_hbh)
2458 free(pktopt->ip6po_hbh, M_IP6OPT);
2459 pktopt->ip6po_hbh = NULL;
2461 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2462 if (pktopt->ip6po_dest1)
2463 free(pktopt->ip6po_dest1, M_IP6OPT);
2464 pktopt->ip6po_dest1 = NULL;
2466 if (optname == -1 || optname == IPV6_RTHDR) {
2467 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2468 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2469 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2470 if (pktopt->ip6po_route.ro_rt) {
2471 RTFREE(pktopt->ip6po_route.ro_rt);
2472 pktopt->ip6po_route.ro_rt = NULL;
2475 if (optname == -1 || optname == IPV6_DSTOPTS) {
2476 if (pktopt->ip6po_dest2)
2477 free(pktopt->ip6po_dest2, M_IP6OPT);
2478 pktopt->ip6po_dest2 = NULL;
2482 #define PKTOPT_EXTHDRCPY(type) \
2485 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2486 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2487 if (dst->type == NULL)\
2489 bcopy(src->type, dst->type, hlen);\
2491 } while (/*CONSTCOND*/ 0)
2494 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2496 if (dst == NULL || src == NULL) {
2497 printf("ip6_clearpktopts: invalid argument\n");
2501 dst->ip6po_hlim = src->ip6po_hlim;
2502 dst->ip6po_tclass = src->ip6po_tclass;
2503 dst->ip6po_flags = src->ip6po_flags;
2504 dst->ip6po_minmtu = src->ip6po_minmtu;
2505 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2506 if (src->ip6po_pktinfo) {
2507 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2509 if (dst->ip6po_pktinfo == NULL)
2511 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2513 if (src->ip6po_nexthop) {
2514 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2516 if (dst->ip6po_nexthop == NULL)
2518 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2519 src->ip6po_nexthop->sa_len);
2521 PKTOPT_EXTHDRCPY(ip6po_hbh);
2522 PKTOPT_EXTHDRCPY(ip6po_dest1);
2523 PKTOPT_EXTHDRCPY(ip6po_dest2);
2524 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2528 ip6_clearpktopts(dst, -1);
2531 #undef PKTOPT_EXTHDRCPY
2533 struct ip6_pktopts *
2534 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2537 struct ip6_pktopts *dst;
2539 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2542 ip6_initpktopts(dst);
2544 if ((error = copypktopts(dst, src, canwait)) != 0) {
2545 free(dst, M_IP6OPT);
2553 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2558 ip6_clearpktopts(pktopt, -1);
2560 free(pktopt, M_IP6OPT);
2564 * Set IPv6 outgoing packet options based on advanced API.
2567 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2568 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2570 struct cmsghdr *cm = NULL;
2572 if (control == NULL || opt == NULL)
2575 ip6_initpktopts(opt);
2580 * If stickyopt is provided, make a local copy of the options
2581 * for this particular packet, then override them by ancillary
2583 * XXX: copypktopts() does not copy the cached route to a next
2584 * hop (if any). This is not very good in terms of efficiency,
2585 * but we can allow this since this option should be rarely
2588 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2593 * XXX: Currently, we assume all the optional information is stored
2596 if (control->m_next)
2599 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2600 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2603 if (control->m_len < CMSG_LEN(0))
2606 cm = mtod(control, struct cmsghdr *);
2607 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2609 if (cm->cmsg_level != IPPROTO_IPV6)
2612 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2613 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2622 * Set a particular packet option, as a sticky option or an ancillary data
2623 * item. "len" can be 0 only when it's a sticky option.
2624 * We have 4 cases of combination of "sticky" and "cmsg":
2625 * "sticky=0, cmsg=0": impossible
2626 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2627 * "sticky=1, cmsg=0": RFC3542 socket option
2628 * "sticky=1, cmsg=1": RFC2292 socket option
2631 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2632 struct ucred *cred, int sticky, int cmsg, int uproto)
2634 int minmtupolicy, preftemp;
2637 if (!sticky && !cmsg) {
2639 printf("ip6_setpktopt: impossible case\n");
2645 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2646 * not be specified in the context of RFC3542. Conversely,
2647 * RFC3542 types should not be specified in the context of RFC2292.
2651 case IPV6_2292PKTINFO:
2652 case IPV6_2292HOPLIMIT:
2653 case IPV6_2292NEXTHOP:
2654 case IPV6_2292HOPOPTS:
2655 case IPV6_2292DSTOPTS:
2656 case IPV6_2292RTHDR:
2657 case IPV6_2292PKTOPTIONS:
2658 return (ENOPROTOOPT);
2661 if (sticky && cmsg) {
2668 case IPV6_RTHDRDSTOPTS:
2670 case IPV6_USE_MIN_MTU:
2673 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2674 return (ENOPROTOOPT);
2679 case IPV6_2292PKTINFO:
2682 struct ifnet *ifp = NULL;
2683 struct in6_pktinfo *pktinfo;
2685 if (len != sizeof(struct in6_pktinfo))
2688 pktinfo = (struct in6_pktinfo *)buf;
2691 * An application can clear any sticky IPV6_PKTINFO option by
2692 * doing a "regular" setsockopt with ipi6_addr being
2693 * in6addr_any and ipi6_ifindex being zero.
2694 * [RFC 3542, Section 6]
2696 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2697 pktinfo->ipi6_ifindex == 0 &&
2698 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2699 ip6_clearpktopts(opt, optname);
2703 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2704 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2707 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2709 /* validate the interface index if specified. */
2710 if (pktinfo->ipi6_ifindex > V_if_index)
2712 if (pktinfo->ipi6_ifindex) {
2713 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2717 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2718 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2722 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2723 struct in6_ifaddr *ia;
2725 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2726 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2728 return (EADDRNOTAVAIL);
2729 ifa_free(&ia->ia_ifa);
2732 * We store the address anyway, and let in6_selectsrc()
2733 * validate the specified address. This is because ipi6_addr
2734 * may not have enough information about its scope zone, and
2735 * we may need additional information (such as outgoing
2736 * interface or the scope zone of a destination address) to
2737 * disambiguate the scope.
2738 * XXX: the delay of the validation may confuse the
2739 * application when it is used as a sticky option.
2741 if (opt->ip6po_pktinfo == NULL) {
2742 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2743 M_IP6OPT, M_NOWAIT);
2744 if (opt->ip6po_pktinfo == NULL)
2747 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2751 case IPV6_2292HOPLIMIT:
2757 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2758 * to simplify the ordering among hoplimit options.
2760 if (optname == IPV6_HOPLIMIT && sticky)
2761 return (ENOPROTOOPT);
2763 if (len != sizeof(int))
2766 if (*hlimp < -1 || *hlimp > 255)
2769 opt->ip6po_hlim = *hlimp;
2777 if (len != sizeof(int))
2779 tclass = *(int *)buf;
2780 if (tclass < -1 || tclass > 255)
2783 opt->ip6po_tclass = tclass;
2787 case IPV6_2292NEXTHOP:
2790 error = priv_check_cred(cred,
2791 PRIV_NETINET_SETHDROPTS, 0);
2796 if (len == 0) { /* just remove the option */
2797 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2801 /* check if cmsg_len is large enough for sa_len */
2802 if (len < sizeof(struct sockaddr) || len < *buf)
2805 switch (((struct sockaddr *)buf)->sa_family) {
2808 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2811 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2814 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2815 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2818 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2824 case AF_LINK: /* should eventually be supported */
2826 return (EAFNOSUPPORT);
2829 /* turn off the previous option, then set the new option. */
2830 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2831 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2832 if (opt->ip6po_nexthop == NULL)
2834 bcopy(buf, opt->ip6po_nexthop, *buf);
2837 case IPV6_2292HOPOPTS:
2840 struct ip6_hbh *hbh;
2844 * XXX: We don't allow a non-privileged user to set ANY HbH
2845 * options, since per-option restriction has too much
2849 error = priv_check_cred(cred,
2850 PRIV_NETINET_SETHDROPTS, 0);
2856 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2857 break; /* just remove the option */
2860 /* message length validation */
2861 if (len < sizeof(struct ip6_hbh))
2863 hbh = (struct ip6_hbh *)buf;
2864 hbhlen = (hbh->ip6h_len + 1) << 3;
2868 /* turn off the previous option, then set the new option. */
2869 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2870 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2871 if (opt->ip6po_hbh == NULL)
2873 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2878 case IPV6_2292DSTOPTS:
2880 case IPV6_RTHDRDSTOPTS:
2882 struct ip6_dest *dest, **newdest = NULL;
2885 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2886 error = priv_check_cred(cred,
2887 PRIV_NETINET_SETHDROPTS, 0);
2893 ip6_clearpktopts(opt, optname);
2894 break; /* just remove the option */
2897 /* message length validation */
2898 if (len < sizeof(struct ip6_dest))
2900 dest = (struct ip6_dest *)buf;
2901 destlen = (dest->ip6d_len + 1) << 3;
2906 * Determine the position that the destination options header
2907 * should be inserted; before or after the routing header.
2910 case IPV6_2292DSTOPTS:
2912 * The old advacned API is ambiguous on this point.
2913 * Our approach is to determine the position based
2914 * according to the existence of a routing header.
2915 * Note, however, that this depends on the order of the
2916 * extension headers in the ancillary data; the 1st
2917 * part of the destination options header must appear
2918 * before the routing header in the ancillary data,
2920 * RFC3542 solved the ambiguity by introducing
2921 * separate ancillary data or option types.
2923 if (opt->ip6po_rthdr == NULL)
2924 newdest = &opt->ip6po_dest1;
2926 newdest = &opt->ip6po_dest2;
2928 case IPV6_RTHDRDSTOPTS:
2929 newdest = &opt->ip6po_dest1;
2932 newdest = &opt->ip6po_dest2;
2936 /* turn off the previous option, then set the new option. */
2937 ip6_clearpktopts(opt, optname);
2938 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2939 if (*newdest == NULL)
2941 bcopy(dest, *newdest, destlen);
2946 case IPV6_2292RTHDR:
2949 struct ip6_rthdr *rth;
2953 ip6_clearpktopts(opt, IPV6_RTHDR);
2954 break; /* just remove the option */
2957 /* message length validation */
2958 if (len < sizeof(struct ip6_rthdr))
2960 rth = (struct ip6_rthdr *)buf;
2961 rthlen = (rth->ip6r_len + 1) << 3;
2965 switch (rth->ip6r_type) {
2966 case IPV6_RTHDR_TYPE_0:
2967 if (rth->ip6r_len == 0) /* must contain one addr */
2969 if (rth->ip6r_len % 2) /* length must be even */
2971 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2975 return (EINVAL); /* not supported */
2978 /* turn off the previous option */
2979 ip6_clearpktopts(opt, IPV6_RTHDR);
2980 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2981 if (opt->ip6po_rthdr == NULL)
2983 bcopy(rth, opt->ip6po_rthdr, rthlen);
2988 case IPV6_USE_MIN_MTU:
2989 if (len != sizeof(int))
2991 minmtupolicy = *(int *)buf;
2992 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2993 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2994 minmtupolicy != IP6PO_MINMTU_ALL) {
2997 opt->ip6po_minmtu = minmtupolicy;
3001 if (len != sizeof(int))
3004 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3006 * we ignore this option for TCP sockets.
3007 * (RFC3542 leaves this case unspecified.)
3009 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3011 opt->ip6po_flags |= IP6PO_DONTFRAG;
3014 case IPV6_PREFER_TEMPADDR:
3015 if (len != sizeof(int))
3017 preftemp = *(int *)buf;
3018 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3019 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3020 preftemp != IP6PO_TEMPADDR_PREFER) {
3023 opt->ip6po_prefer_tempaddr = preftemp;
3027 return (ENOPROTOOPT);
3028 } /* end of switch */
3034 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3035 * packet to the input queue of a specified interface. Note that this
3036 * calls the output routine of the loopback "driver", but with an interface
3037 * pointer that might NOT be &loif -- easier than replicating that code here.
3040 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3043 struct ip6_hdr *ip6;
3045 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3050 * Make sure to deep-copy IPv6 header portion in case the data
3051 * is in an mbuf cluster, so that we can safely override the IPv6
3052 * header portion later.
3054 if (!M_WRITABLE(copym) ||
3055 copym->m_len < sizeof(struct ip6_hdr)) {
3056 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3060 ip6 = mtod(copym, struct ip6_hdr *);
3062 * clear embedded scope identifiers if necessary.
3063 * in6_clearscope will touch the addresses only when necessary.
3065 in6_clearscope(&ip6->ip6_src);
3066 in6_clearscope(&ip6->ip6_dst);
3067 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3068 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3070 copym->m_pkthdr.csum_data = 0xffff;
3072 if_simloop(ifp, copym, AF_INET6, 0);
3076 * Chop IPv6 header off from the payload.
3079 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3082 struct ip6_hdr *ip6;
3084 ip6 = mtod(m, struct ip6_hdr *);
3085 if (m->m_len > sizeof(*ip6)) {
3086 mh = m_gethdr(M_NOWAIT, MT_DATA);
3091 m_move_pkthdr(mh, m);
3092 M_ALIGN(mh, sizeof(*ip6));
3093 m->m_len -= sizeof(*ip6);
3094 m->m_data += sizeof(*ip6);
3097 m->m_len = sizeof(*ip6);
3098 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3100 exthdrs->ip6e_ip6 = m;
3105 * Compute IPv6 extension header length.
3108 ip6_optlen(struct inpcb *in6p)
3112 if (!in6p->in6p_outputopts)
3117 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3119 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3120 if (in6p->in6p_outputopts->ip6po_rthdr)
3121 /* dest1 is valid with rthdr only */
3122 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3123 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3124 len += elen(in6p->in6p_outputopts->ip6po_dest2);