2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
33 * Copyright (c) 1982, 1986, 1988, 1990, 1993
34 * The Regents of the University of California. All rights reserved.
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
67 #include "opt_inet6.h"
68 #include "opt_ratelimit.h"
69 #include "opt_ipsec.h"
71 #include "opt_route.h"
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
78 #include <sys/errno.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
87 #include <machine/in_cksum.h>
90 #include <net/if_var.h>
91 #include <net/if_llatbl.h>
92 #include <net/netisr.h>
93 #include <net/route.h>
95 #include <net/rss_config.h>
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip_var.h>
101 #include <netinet6/in6_fib.h>
102 #include <netinet6/in6_var.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp6.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/tcp_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/in6_rss.h>
111 #include <netipsec/ipsec_support.h>
113 #include <netinet/sctp.h>
114 #include <netinet/sctp_crc32.h>
117 #include <netinet6/ip6protosw.h>
118 #include <netinet6/scope6_var.h>
121 #include <net/flowtable.h>
124 extern int in6_mcast_loop;
127 struct mbuf *ip6e_ip6;
128 struct mbuf *ip6e_hbh;
129 struct mbuf *ip6e_dest1;
130 struct mbuf *ip6e_rthdr;
131 struct mbuf *ip6e_dest2;
134 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
137 struct ucred *, int);
138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
139 struct socket *, struct sockopt *);
140 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
142 struct ucred *, int, int, int);
144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
149 static int ip6_getpmtu(struct route_in6 *, int,
150 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
152 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
153 u_long *, int *, u_int);
154 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
155 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
159 * Make an extension header from option data. hp is the source, and
160 * mp is the destination.
162 #define MAKE_EXTHDR(hp, mp) \
165 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
166 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
167 ((eh)->ip6e_len + 1) << 3); \
171 } while (/*CONSTCOND*/ 0)
174 * Form a chain of extension headers.
175 * m is the extension header mbuf
176 * mp is the previous mbuf in the chain
177 * p is the next header
178 * i is the type of option.
180 #define MAKE_CHAIN(m, mp, p, i)\
184 panic("assumption failed: hdr not split"); \
185 *mtod((m), u_char *) = *(p);\
187 p = mtod((m), u_char *);\
188 (m)->m_next = (mp)->m_next;\
192 } while (/*CONSTCOND*/ 0)
195 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
199 csum = in_cksum_skip(m, offset + plen, offset);
200 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
202 offset += m->m_pkthdr.csum_data; /* checksum offset */
204 if (offset + sizeof(u_short) > m->m_len) {
205 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
206 "csum_flags=%b\n", __func__, m->m_len, plen, offset,
207 (int)m->m_pkthdr.csum_flags, CSUM_BITS);
209 * XXX this should not happen, but if it does, the correct
210 * behavior may be to insert the checksum in the appropriate
211 * next mbuf in the chain.
215 *(u_short *)(m->m_data + offset) = csum;
219 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
220 int mtu, uint32_t id)
222 struct mbuf *m, **mnext, *m_frgpart;
223 struct ip6_hdr *ip6, *mhip6;
224 struct ip6_frag *ip6f;
227 int tlen = m0->m_pkthdr.len;
230 ip6 = mtod(m, struct ip6_hdr *);
231 mnext = &m->m_nextpkt;
233 for (off = hlen; off < tlen; off += mtu) {
234 m = m_gethdr(M_NOWAIT, MT_DATA);
236 IP6STAT_INC(ip6s_odropped);
239 m->m_flags = m0->m_flags & M_COPYFLAGS;
241 mnext = &m->m_nextpkt;
242 m->m_data += max_linkhdr;
243 mhip6 = mtod(m, struct ip6_hdr *);
245 m->m_len = sizeof(*mhip6);
246 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
248 IP6STAT_INC(ip6s_odropped);
251 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
252 if (off + mtu >= tlen)
255 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
256 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
257 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
258 if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) {
259 IP6STAT_INC(ip6s_odropped);
263 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
264 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
265 m->m_pkthdr.rcvif = NULL;
266 ip6f->ip6f_reserved = 0;
267 ip6f->ip6f_ident = id;
268 ip6f->ip6f_nxt = nextproto;
269 IP6STAT_INC(ip6s_ofragments);
270 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
277 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
278 * header (with pri, len, nxt, hlim, src, dst).
279 * This function may modify ver and hlim only.
280 * The mbuf chain containing the packet will be freed.
281 * The mbuf opt, if present, will not be freed.
282 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
283 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
284 * then result of route lookup is stored in ro->ro_rt.
286 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
287 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
290 * ifpp - XXX: just for statistics
293 * XXX TODO: no flowid is assigned for outbound flows?
296 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
297 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
298 struct ifnet **ifpp, struct inpcb *inp)
301 struct ifnet *ifp, *origifp;
303 struct mbuf *mprev = NULL;
305 struct route_in6 ip6route;
306 struct rtentry *rt = NULL;
307 struct sockaddr_in6 *dst, src_sa, dst_sa;
308 struct in6_addr odst;
310 struct in6_ifaddr *ia = NULL;
312 int alwaysfrag, dontfrag;
313 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
314 struct ip6_exthdrs exthdrs;
315 struct in6_addr src0, dst0;
317 struct route_in6 *ro_pmtu = NULL;
322 struct m_tag *fwd_tag = NULL;
326 M_SETFIB(m, inp->inp_inc.inc_fibnum);
327 if ((flags & IP_NODEFAULTFLOWID) == 0) {
328 /* unconditionally set flowid */
329 m->m_pkthdr.flowid = inp->inp_flowid;
330 M_HASHTYPE_SET(m, inp->inp_flowtype);
334 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
336 * IPSec checking which handles several cases.
337 * FAST IPSEC: We re-injected the packet.
338 * XXX: need scope argument.
340 if (IPSEC_ENABLED(ipv6)) {
341 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
342 if (error == EINPROGRESS)
349 bzero(&exthdrs, sizeof(exthdrs));
351 /* Hop-by-Hop options header */
352 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
353 /* Destination options header(1st part) */
354 if (opt->ip6po_rthdr) {
356 * Destination options header(1st part)
357 * This only makes sense with a routing header.
358 * See Section 9.2 of RFC 3542.
359 * Disabling this part just for MIP6 convenience is
360 * a bad idea. We need to think carefully about a
361 * way to make the advanced API coexist with MIP6
362 * options, which might automatically be inserted in
365 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
368 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
369 /* Destination options header(2nd part) */
370 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
374 * Calculate the total length of the extension header chain.
375 * Keep the length of the unfragmentable part for fragmentation.
378 if (exthdrs.ip6e_hbh)
379 optlen += exthdrs.ip6e_hbh->m_len;
380 if (exthdrs.ip6e_dest1)
381 optlen += exthdrs.ip6e_dest1->m_len;
382 if (exthdrs.ip6e_rthdr)
383 optlen += exthdrs.ip6e_rthdr->m_len;
384 unfragpartlen = optlen + sizeof(struct ip6_hdr);
386 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
387 if (exthdrs.ip6e_dest2)
388 optlen += exthdrs.ip6e_dest2->m_len;
391 * If there is at least one extension header,
392 * separate IP6 header from the payload.
394 if (optlen && !hdrsplit) {
395 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
399 m = exthdrs.ip6e_ip6;
403 ip6 = mtod(m, struct ip6_hdr *);
405 /* adjust mbuf packet header length */
406 m->m_pkthdr.len += optlen;
407 plen = m->m_pkthdr.len - sizeof(*ip6);
409 /* If this is a jumbo payload, insert a jumbo payload option. */
410 if (plen > IPV6_MAXPACKET) {
412 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
416 m = exthdrs.ip6e_ip6;
420 ip6 = mtod(m, struct ip6_hdr *);
421 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
425 ip6->ip6_plen = htons(plen);
428 * Concatenate headers and fill in next header fields.
429 * Here we have, on "m"
431 * and we insert headers accordingly. Finally, we should be getting:
432 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
434 * during the header composing process, "m" points to IPv6 header.
435 * "mprev" points to an extension header prior to esp.
437 u_char *nexthdrp = &ip6->ip6_nxt;
441 * we treat dest2 specially. this makes IPsec processing
442 * much easier. the goal here is to make mprev point the
443 * mbuf prior to dest2.
445 * result: IPv6 dest2 payload
446 * m and mprev will point to IPv6 header.
448 if (exthdrs.ip6e_dest2) {
450 panic("assumption failed: hdr not split");
451 exthdrs.ip6e_dest2->m_next = m->m_next;
452 m->m_next = exthdrs.ip6e_dest2;
453 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
454 ip6->ip6_nxt = IPPROTO_DSTOPTS;
458 * result: IPv6 hbh dest1 rthdr dest2 payload
459 * m will point to IPv6 header. mprev will point to the
460 * extension header prior to dest2 (rthdr in the above case).
462 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
463 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
465 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
469 * If there is a routing header, discard the packet.
471 if (exthdrs.ip6e_rthdr) {
476 /* Source address validation */
477 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
478 (flags & IPV6_UNSPECSRC) == 0) {
480 IP6STAT_INC(ip6s_badscope);
483 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
485 IP6STAT_INC(ip6s_badscope);
489 IP6STAT_INC(ip6s_localout);
496 bzero((caddr_t)ro, sizeof(*ro));
498 ro->ro_flags |= RT_LLE_CACHE;
500 if (opt && opt->ip6po_rthdr)
501 ro = &opt->ip6po_route;
502 dst = (struct sockaddr_in6 *)&ro->ro_dst;
504 if (ro->ro_rt == NULL)
505 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
507 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
510 * if specified, try to fill in the traffic class field.
511 * do not override if a non-zero value is already set.
512 * we check the diffserv field and the ecn field separately.
514 if (opt && opt->ip6po_tclass >= 0) {
517 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
519 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
522 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
525 /* fill in or override the hop limit field, if necessary. */
526 if (opt && opt->ip6po_hlim != -1)
527 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
528 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
530 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
532 ip6->ip6_hlim = V_ip6_defmcasthlim;
535 * Validate route against routing table additions;
536 * a better/more specific route might have been added.
537 * Make sure address family is set in route.
540 ro->ro_dst.sin6_family = AF_INET6;
541 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
543 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
544 ro->ro_dst.sin6_family == AF_INET6 &&
545 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
547 ifp = ro->ro_rt->rt_ifp;
550 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
552 if (fwd_tag == NULL) {
553 bzero(&dst_sa, sizeof(dst_sa));
554 dst_sa.sin6_family = AF_INET6;
555 dst_sa.sin6_len = sizeof(dst_sa);
556 dst_sa.sin6_addr = ip6->ip6_dst;
558 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
562 in6_ifstat_inc(ifp, ifs6_out_discard);
568 * If in6_selectroute() does not return a route entry,
569 * dst may not have been updated.
571 *dst = dst_sa; /* XXX */
575 * then rt (for unicast) and ifp must be non-NULL valid values.
577 if ((flags & IPV6_FORWARDING) == 0) {
578 /* XXX: the FORWARDING flag can be set for mrouting. */
579 in6_ifstat_inc(ifp, ifs6_out_request);
582 ia = (struct in6_ifaddr *)(rt->rt_ifa);
583 counter_u64_add(rt->rt_pksent, 1);
588 * The outgoing interface must be in the zone of source and
589 * destination addresses.
594 if (in6_setscope(&src0, origifp, &zone))
596 bzero(&src_sa, sizeof(src_sa));
597 src_sa.sin6_family = AF_INET6;
598 src_sa.sin6_len = sizeof(src_sa);
599 src_sa.sin6_addr = ip6->ip6_src;
600 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
604 if (in6_setscope(&dst0, origifp, &zone))
606 /* re-initialize to be sure */
607 bzero(&dst_sa, sizeof(dst_sa));
608 dst_sa.sin6_family = AF_INET6;
609 dst_sa.sin6_len = sizeof(dst_sa);
610 dst_sa.sin6_addr = ip6->ip6_dst;
611 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
615 /* We should use ia_ifp to support the case of
616 * sending packets to an address of our own.
618 if (ia != NULL && ia->ia_ifp)
621 /* scope check is done. */
625 IP6STAT_INC(ip6s_badscope);
626 in6_ifstat_inc(origifp, ifs6_out_discard);
628 error = EHOSTUNREACH; /* XXX */
632 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
633 if (opt && opt->ip6po_nextroute.ro_rt) {
635 * The nexthop is explicitly specified by the
636 * application. We assume the next hop is an IPv6
639 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
641 else if ((rt->rt_flags & RTF_GATEWAY))
642 dst = (struct sockaddr_in6 *)rt->rt_gateway;
645 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
646 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
648 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
649 in6_ifstat_inc(ifp, ifs6_out_mcast);
651 * Confirm that the outgoing interface supports multicast.
653 if (!(ifp->if_flags & IFF_MULTICAST)) {
654 IP6STAT_INC(ip6s_noroute);
655 in6_ifstat_inc(ifp, ifs6_out_discard);
659 if ((im6o == NULL && in6_mcast_loop) ||
660 (im6o && im6o->im6o_multicast_loop)) {
662 * Loop back multicast datagram if not expressly
663 * forbidden to do so, even if we have not joined
664 * the address; protocols will filter it later,
665 * thus deferring a hash lookup and lock acquisition
666 * at the expense of an m_copym().
668 ip6_mloopback(ifp, m);
671 * If we are acting as a multicast router, perform
672 * multicast forwarding as if the packet had just
673 * arrived on the interface to which we are about
674 * to send. The multicast forwarding function
675 * recursively calls this function, using the
676 * IPV6_FORWARDING flag to prevent infinite recursion.
678 * Multicasts that are looped back by ip6_mloopback(),
679 * above, will be forwarded by the ip6_input() routine,
682 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
684 * XXX: ip6_mforward expects that rcvif is NULL
685 * when it is called from the originating path.
686 * However, it may not always be the case.
688 m->m_pkthdr.rcvif = NULL;
689 if (ip6_mforward(ip6, ifp, m) != 0) {
696 * Multicasts with a hoplimit of zero may be looped back,
697 * above, but must not be transmitted on a network.
698 * Also, multicasts addressed to the loopback interface
699 * are not sent -- the above call to ip6_mloopback() will
700 * loop back a copy if this host actually belongs to the
701 * destination group on the loopback interface.
703 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
704 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
711 * Fill the outgoing inteface to tell the upper layer
712 * to increment per-interface statistics.
717 /* Determine path MTU. */
718 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
719 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
723 * The caller of this function may specify to use the minimum MTU
725 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
726 * setting. The logic is a bit complicated; by default, unicast
727 * packets will follow path MTU while multicast packets will be sent at
728 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
729 * including unicast ones will be sent at the minimum MTU. Multicast
730 * packets will always be sent at the minimum MTU unless
731 * IP6PO_MINMTU_DISABLE is explicitly specified.
732 * See RFC 3542 for more details.
734 if (mtu > IPV6_MMTU) {
735 if ((flags & IPV6_MINMTU))
737 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
739 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
741 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
747 * clear embedded scope identifiers if necessary.
748 * in6_clearscope will touch the addresses only when necessary.
750 in6_clearscope(&ip6->ip6_src);
751 in6_clearscope(&ip6->ip6_dst);
754 * If the outgoing packet contains a hop-by-hop options header,
755 * it must be examined and processed even by the source node.
756 * (RFC 2460, section 4.)
758 if (exthdrs.ip6e_hbh) {
759 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
760 u_int32_t dummy; /* XXX unused */
761 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
764 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
765 panic("ip6e_hbh is not contiguous");
768 * XXX: if we have to send an ICMPv6 error to the sender,
769 * we need the M_LOOP flag since icmp6_error() expects
770 * the IPv6 and the hop-by-hop options header are
771 * contiguous unless the flag is set.
773 m->m_flags |= M_LOOP;
774 m->m_pkthdr.rcvif = ifp;
775 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
776 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
777 &dummy, &plen) < 0) {
778 /* m was already freed at this point */
779 error = EINVAL;/* better error? */
782 m->m_flags &= ~M_LOOP; /* XXX */
783 m->m_pkthdr.rcvif = NULL;
786 /* Jump over all PFIL processing if hooks are not active. */
787 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
791 /* Run through list of hooks for output packets. */
792 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
793 if (error != 0 || m == NULL)
796 ip6 = mtod(m, struct ip6_hdr *);
799 /* See if destination IP address was changed by packet filter. */
800 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
801 m->m_flags |= M_SKIP_FIREWALL;
802 /* If destination is now ourself drop to ip6_input(). */
803 if (in6_localip(&ip6->ip6_dst)) {
804 m->m_flags |= M_FASTFWD_OURS;
805 if (m->m_pkthdr.rcvif == NULL)
806 m->m_pkthdr.rcvif = V_loif;
807 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
808 m->m_pkthdr.csum_flags |=
809 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
810 m->m_pkthdr.csum_data = 0xffff;
813 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
814 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
816 error = netisr_queue(NETISR_IPV6, m);
820 needfiblookup = 1; /* Redo the routing table lookup. */
822 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
826 /* See if fib was changed by packet filter. */
827 if (fibnum != M_GETFIB(m)) {
828 m->m_flags |= M_SKIP_FIREWALL;
829 fibnum = M_GETFIB(m);
833 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
839 /* See if local, if yes, send it to netisr. */
840 if (m->m_flags & M_FASTFWD_OURS) {
841 if (m->m_pkthdr.rcvif == NULL)
842 m->m_pkthdr.rcvif = V_loif;
843 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
844 m->m_pkthdr.csum_flags |=
845 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
846 m->m_pkthdr.csum_data = 0xffff;
849 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
850 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
852 error = netisr_queue(NETISR_IPV6, m);
855 /* Or forward to some other address? */
856 if ((m->m_flags & M_IP6_NEXTHOP) &&
857 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
858 dst = (struct sockaddr_in6 *)&ro->ro_dst;
859 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
860 m->m_flags |= M_SKIP_FIREWALL;
861 m->m_flags &= ~M_IP6_NEXTHOP;
862 m_tag_delete(m, fwd_tag);
868 * Send the packet to the outgoing interface.
869 * If necessary, do IPv6 fragmentation before sending.
871 * the logic here is rather complex:
872 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
873 * 1-a: send as is if tlen <= path mtu
874 * 1-b: fragment if tlen > path mtu
876 * 2: if user asks us not to fragment (dontfrag == 1)
877 * 2-a: send as is if tlen <= interface mtu
878 * 2-b: error if tlen > interface mtu
880 * 3: if we always need to attach fragment header (alwaysfrag == 1)
883 * 4: if dontfrag == 1 && alwaysfrag == 1
884 * error, as we cannot handle this conflicting request
886 sw_csum = m->m_pkthdr.csum_flags;
888 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
889 sw_csum &= ~ifp->if_hwassist;
893 * If we added extension headers, we will not do TSO and calculate the
894 * checksums ourselves for now.
895 * XXX-BZ Need a framework to know when the NIC can handle it, even
898 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
899 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
900 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
903 if (sw_csum & CSUM_SCTP_IPV6) {
904 sw_csum &= ~CSUM_SCTP_IPV6;
905 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
908 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
909 tlen = m->m_pkthdr.len;
911 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
915 if (dontfrag && alwaysfrag) { /* case 4 */
916 /* conflicting request - can't transmit */
920 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
922 * Even if the DONTFRAG option is specified, we cannot send the
923 * packet when the data length is larger than the MTU of the
924 * outgoing interface.
925 * Notify the error by sending IPV6_PATHMTU ancillary data if
926 * application wanted to know the MTU value. Also return an
927 * error code (this is not described in the API spec).
930 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
936 * transmit packet without fragmentation
938 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
939 struct in6_ifaddr *ia6;
941 ip6 = mtod(m, struct ip6_hdr *);
942 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
944 /* Record statistics for this interface address. */
945 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
946 counter_u64_add(ia6->ia_ifa.ifa_obytes,
948 ifa_free(&ia6->ia_ifa);
952 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
953 in_pcboutput_txrtlmt(inp, ifp, m);
954 /* stamp send tag on mbuf */
955 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
957 m->m_pkthdr.snd_tag = NULL;
960 error = nd6_output_ifp(ifp, origifp, m, dst,
963 /* check for route change */
965 in_pcboutput_eagain(inp);
971 * try to fragment the packet. case 1-b and 3
973 if (mtu < IPV6_MMTU) {
974 /* path MTU cannot be less than IPV6_MMTU */
976 in6_ifstat_inc(ifp, ifs6_out_fragfail);
978 } else if (ip6->ip6_plen == 0) {
979 /* jumbo payload cannot be fragmented */
981 in6_ifstat_inc(ifp, ifs6_out_fragfail);
987 * Too large for the destination or interface;
988 * fragment if possible.
989 * Must be able to put at least 8 bytes per fragment.
991 hlen = unfragpartlen;
992 if (mtu > IPV6_MAXPACKET)
993 mtu = IPV6_MAXPACKET;
995 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
998 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1003 * If the interface will not calculate checksums on
1004 * fragmented packets, then do it here.
1005 * XXX-BZ handle the hw offloading case. Need flags.
1007 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1008 in6_delayed_cksum(m, plen, hlen);
1009 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1012 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1013 sctp_delayed_cksum(m, hlen);
1014 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1018 * Change the next header field of the last header in the
1019 * unfragmentable part.
1021 if (exthdrs.ip6e_rthdr) {
1022 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1023 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1024 } else if (exthdrs.ip6e_dest1) {
1025 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1026 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1027 } else if (exthdrs.ip6e_hbh) {
1028 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1029 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1031 nextproto = ip6->ip6_nxt;
1032 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1036 * Loop through length of segment after first fragment,
1037 * make new header and copy data of each part and link onto
1041 id = htonl(ip6_randomid());
1042 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1045 in6_ifstat_inc(ifp, ifs6_out_fragok);
1049 * Remove leading garbages.
1055 for (m0 = m; m; m = m0) {
1059 /* Record statistics for this interface address. */
1061 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1062 counter_u64_add(ia->ia_ifa.ifa_obytes,
1067 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1068 in_pcboutput_txrtlmt(inp, ifp, m);
1069 /* stamp send tag on mbuf */
1070 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1072 m->m_pkthdr.snd_tag = NULL;
1075 error = nd6_output_ifp(ifp, origifp, m, dst,
1076 (struct route *)ro);
1078 /* check for route change */
1079 if (error == EAGAIN)
1080 in_pcboutput_eagain(inp);
1087 IP6STAT_INC(ip6s_fragmented);
1090 if (ro == &ip6route)
1095 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1096 m_freem(exthdrs.ip6e_dest1);
1097 m_freem(exthdrs.ip6e_rthdr);
1098 m_freem(exthdrs.ip6e_dest2);
1107 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1111 if (hlen > MCLBYTES)
1112 return (ENOBUFS); /* XXX */
1115 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1117 m = m_get(M_NOWAIT, MT_DATA);
1122 bcopy(hdr, mtod(m, caddr_t), hlen);
1129 * Insert jumbo payload option.
1132 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1138 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1141 * If there is no hop-by-hop options header, allocate new one.
1142 * If there is one but it doesn't have enough space to store the
1143 * jumbo payload option, allocate a cluster to store the whole options.
1144 * Otherwise, use it to store the options.
1146 if (exthdrs->ip6e_hbh == NULL) {
1147 mopt = m_get(M_NOWAIT, MT_DATA);
1150 mopt->m_len = JUMBOOPTLEN;
1151 optbuf = mtod(mopt, u_char *);
1152 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1153 exthdrs->ip6e_hbh = mopt;
1155 struct ip6_hbh *hbh;
1157 mopt = exthdrs->ip6e_hbh;
1158 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1161 * - exthdrs->ip6e_hbh is not referenced from places
1162 * other than exthdrs.
1163 * - exthdrs->ip6e_hbh is not an mbuf chain.
1165 int oldoptlen = mopt->m_len;
1169 * XXX: give up if the whole (new) hbh header does
1170 * not fit even in an mbuf cluster.
1172 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1176 * As a consequence, we must always prepare a cluster
1179 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1182 n->m_len = oldoptlen + JUMBOOPTLEN;
1183 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1185 optbuf = mtod(n, caddr_t) + oldoptlen;
1187 mopt = exthdrs->ip6e_hbh = n;
1189 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1190 mopt->m_len += JUMBOOPTLEN;
1192 optbuf[0] = IP6OPT_PADN;
1196 * Adjust the header length according to the pad and
1197 * the jumbo payload option.
1199 hbh = mtod(mopt, struct ip6_hbh *);
1200 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1203 /* fill in the option. */
1204 optbuf[2] = IP6OPT_JUMBO;
1206 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1207 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1209 /* finally, adjust the packet header length */
1210 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1217 * Insert fragment header and copy unfragmentable header portions.
1220 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1221 struct ip6_frag **frghdrp)
1223 struct mbuf *n, *mlast;
1225 if (hlen > sizeof(struct ip6_hdr)) {
1226 n = m_copym(m0, sizeof(struct ip6_hdr),
1227 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1234 /* Search for the last mbuf of unfragmentable part. */
1235 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1238 if (M_WRITABLE(mlast) &&
1239 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1240 /* use the trailing space of the last mbuf for the fragment hdr */
1241 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1243 mlast->m_len += sizeof(struct ip6_frag);
1244 m->m_pkthdr.len += sizeof(struct ip6_frag);
1246 /* allocate a new mbuf for the fragment header */
1249 mfrg = m_get(M_NOWAIT, MT_DATA);
1252 mfrg->m_len = sizeof(struct ip6_frag);
1253 *frghdrp = mtod(mfrg, struct ip6_frag *);
1254 mlast->m_next = mfrg;
1261 * Calculates IPv6 path mtu for destination @dst.
1262 * Resulting MTU is stored in @mtup.
1264 * Returns 0 on success.
1267 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1269 struct nhop6_extended nh6;
1270 struct in6_addr kdst;
1276 in6_splitscope(dst, &kdst, &scopeid);
1277 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1278 return (EHOSTUNREACH);
1283 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1284 fib6_free_nh_ext(fibnum, &nh6);
1290 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1291 * and cached data in @ro_pmtu.
1292 * MTU from (successful) route lookup is saved (along with dst)
1293 * inside @ro_pmtu to avoid subsequent route lookups after packet
1294 * filter processing.
1296 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1297 * Returns 0 on success.
1300 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1301 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1302 int *alwaysfragp, u_int fibnum, u_int proto)
1304 struct nhop6_basic nh6;
1305 struct in6_addr kdst;
1307 struct sockaddr_in6 *sa6_dst;
1314 * Here ro_pmtu has final destination address, while
1315 * ro might represent immediate destination.
1316 * Use ro_pmtu destination since mtu might differ.
1318 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1319 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1320 ro_pmtu->ro_mtu = 0;
1322 if (ro_pmtu->ro_mtu == 0) {
1323 bzero(sa6_dst, sizeof(*sa6_dst));
1324 sa6_dst->sin6_family = AF_INET6;
1325 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1326 sa6_dst->sin6_addr = *dst;
1328 in6_splitscope(dst, &kdst, &scopeid);
1329 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1331 ro_pmtu->ro_mtu = nh6.nh_mtu;
1334 mtu = ro_pmtu->ro_mtu;
1338 mtu = ro_pmtu->ro_rt->rt_mtu;
1340 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1344 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1345 * hostcache data for @dst.
1346 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1348 * Returns 0 on success.
1351 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1352 u_long *mtup, int *alwaysfragp, u_int proto)
1360 struct in_conninfo inc;
1362 bzero(&inc, sizeof(inc));
1363 inc.inc_flags |= INC_ISIPV6;
1364 inc.inc6_faddr = *dst;
1366 ifmtu = IN6_LINKMTU(ifp);
1368 /* TCP is known to react to pmtu changes so skip hc */
1369 if (proto != IPPROTO_TCP)
1370 mtu = tcp_hc_getmtu(&inc);
1373 mtu = min(mtu, rt_mtu);
1378 else if (mtu < IPV6_MMTU) {
1380 * RFC2460 section 5, last paragraph:
1381 * if we record ICMPv6 too big message with
1382 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1383 * or smaller, with framgent header attached.
1384 * (fragment header is needed regardless from the
1385 * packet size, for translators to identify packets)
1391 mtu = IN6_LINKMTU(ifp);
1393 error = EHOSTUNREACH; /* XXX */
1397 *alwaysfragp = alwaysfrag;
1402 * IP6 socket option processing.
1405 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1407 int optdatalen, uproto;
1409 struct inpcb *in6p = sotoinpcb(so);
1411 int level, op, optname;
1415 uint32_t rss_bucket;
1420 * Don't use more than a quarter of mbuf clusters. N.B.:
1421 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1422 * on LP64 architectures, so cast to u_long to avoid undefined
1423 * behavior. ILP32 architectures cannot have nmbclusters
1424 * large enough to overflow for other reasons.
1426 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1428 level = sopt->sopt_level;
1429 op = sopt->sopt_dir;
1430 optname = sopt->sopt_name;
1431 optlen = sopt->sopt_valsize;
1435 uproto = (int)so->so_proto->pr_protocol;
1437 if (level != IPPROTO_IPV6) {
1440 if (sopt->sopt_level == SOL_SOCKET &&
1441 sopt->sopt_dir == SOPT_SET) {
1442 switch (sopt->sopt_name) {
1445 if ((so->so_options & SO_REUSEADDR) != 0)
1446 in6p->inp_flags2 |= INP_REUSEADDR;
1448 in6p->inp_flags2 &= ~INP_REUSEADDR;
1454 if ((so->so_options & SO_REUSEPORT) != 0)
1455 in6p->inp_flags2 |= INP_REUSEPORT;
1457 in6p->inp_flags2 &= ~INP_REUSEPORT;
1463 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1467 case SO_MAX_PACING_RATE:
1470 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1481 } else { /* level == IPPROTO_IPV6 */
1486 case IPV6_2292PKTOPTIONS:
1487 #ifdef IPV6_PKTOPTIONS
1488 case IPV6_PKTOPTIONS:
1493 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1494 printf("ip6_ctloutput: mbuf limit hit\n");
1499 error = soopt_getm(sopt, &m); /* XXX */
1502 error = soopt_mcopyin(sopt, m); /* XXX */
1505 error = ip6_pcbopts(&in6p->in6p_outputopts,
1507 m_freem(m); /* XXX */
1512 * Use of some Hop-by-Hop options or some
1513 * Destination options, might require special
1514 * privilege. That is, normal applications
1515 * (without special privilege) might be forbidden
1516 * from setting certain options in outgoing packets,
1517 * and might never see certain options in received
1518 * packets. [RFC 2292 Section 6]
1519 * KAME specific note:
1520 * KAME prevents non-privileged users from sending or
1521 * receiving ANY hbh/dst options in order to avoid
1522 * overhead of parsing options in the kernel.
1524 case IPV6_RECVHOPOPTS:
1525 case IPV6_RECVDSTOPTS:
1526 case IPV6_RECVRTHDRDSTOPTS:
1528 error = priv_check(td,
1529 PRIV_NETINET_SETHDROPTS);
1534 case IPV6_UNICAST_HOPS:
1537 case IPV6_RECVPKTINFO:
1538 case IPV6_RECVHOPLIMIT:
1539 case IPV6_RECVRTHDR:
1540 case IPV6_RECVPATHMTU:
1541 case IPV6_RECVTCLASS:
1542 case IPV6_RECVFLOWID:
1544 case IPV6_RECVRSSBUCKETID:
1547 case IPV6_AUTOFLOWLABEL:
1548 case IPV6_ORIGDSTADDR:
1550 case IPV6_BINDMULTI:
1552 case IPV6_RSS_LISTEN_BUCKET:
1554 if (optname == IPV6_BINDANY && td != NULL) {
1555 error = priv_check(td,
1556 PRIV_NETINET_BINDANY);
1561 if (optlen != sizeof(int)) {
1565 error = sooptcopyin(sopt, &optval,
1566 sizeof optval, sizeof optval);
1571 case IPV6_UNICAST_HOPS:
1572 if (optval < -1 || optval >= 256)
1575 /* -1 = kernel default */
1576 in6p->in6p_hops = optval;
1577 if ((in6p->inp_vflag &
1579 in6p->inp_ip_ttl = optval;
1582 #define OPTSET(bit) \
1586 in6p->inp_flags |= (bit); \
1588 in6p->inp_flags &= ~(bit); \
1589 INP_WUNLOCK(in6p); \
1590 } while (/*CONSTCOND*/ 0)
1591 #define OPTSET2292(bit) \
1594 in6p->inp_flags |= IN6P_RFC2292; \
1596 in6p->inp_flags |= (bit); \
1598 in6p->inp_flags &= ~(bit); \
1599 INP_WUNLOCK(in6p); \
1600 } while (/*CONSTCOND*/ 0)
1601 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1603 #define OPTSET2(bit, val) do { \
1606 in6p->inp_flags2 |= bit; \
1608 in6p->inp_flags2 &= ~bit; \
1609 INP_WUNLOCK(in6p); \
1611 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1613 case IPV6_RECVPKTINFO:
1614 /* cannot mix with RFC2292 */
1615 if (OPTBIT(IN6P_RFC2292)) {
1619 OPTSET(IN6P_PKTINFO);
1624 struct ip6_pktopts **optp;
1626 /* cannot mix with RFC2292 */
1627 if (OPTBIT(IN6P_RFC2292)) {
1631 optp = &in6p->in6p_outputopts;
1632 error = ip6_pcbopt(IPV6_HOPLIMIT,
1633 (u_char *)&optval, sizeof(optval),
1634 optp, (td != NULL) ? td->td_ucred :
1639 case IPV6_RECVHOPLIMIT:
1640 /* cannot mix with RFC2292 */
1641 if (OPTBIT(IN6P_RFC2292)) {
1645 OPTSET(IN6P_HOPLIMIT);
1648 case IPV6_RECVHOPOPTS:
1649 /* cannot mix with RFC2292 */
1650 if (OPTBIT(IN6P_RFC2292)) {
1654 OPTSET(IN6P_HOPOPTS);
1657 case IPV6_RECVDSTOPTS:
1658 /* cannot mix with RFC2292 */
1659 if (OPTBIT(IN6P_RFC2292)) {
1663 OPTSET(IN6P_DSTOPTS);
1666 case IPV6_RECVRTHDRDSTOPTS:
1667 /* cannot mix with RFC2292 */
1668 if (OPTBIT(IN6P_RFC2292)) {
1672 OPTSET(IN6P_RTHDRDSTOPTS);
1675 case IPV6_RECVRTHDR:
1676 /* cannot mix with RFC2292 */
1677 if (OPTBIT(IN6P_RFC2292)) {
1684 case IPV6_RECVPATHMTU:
1686 * We ignore this option for TCP
1688 * (RFC3542 leaves this case
1691 if (uproto != IPPROTO_TCP)
1695 case IPV6_RECVFLOWID:
1696 OPTSET2(INP_RECVFLOWID, optval);
1700 case IPV6_RECVRSSBUCKETID:
1701 OPTSET2(INP_RECVRSSBUCKETID, optval);
1707 * make setsockopt(IPV6_V6ONLY)
1708 * available only prior to bind(2).
1709 * see ipng mailing list, Jun 22 2001.
1711 if (in6p->inp_lport ||
1712 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1716 OPTSET(IN6P_IPV6_V6ONLY);
1718 in6p->inp_vflag &= ~INP_IPV4;
1720 in6p->inp_vflag |= INP_IPV4;
1722 case IPV6_RECVTCLASS:
1723 /* cannot mix with RFC2292 XXX */
1724 if (OPTBIT(IN6P_RFC2292)) {
1728 OPTSET(IN6P_TCLASS);
1730 case IPV6_AUTOFLOWLABEL:
1731 OPTSET(IN6P_AUTOFLOWLABEL);
1734 case IPV6_ORIGDSTADDR:
1735 OPTSET2(INP_ORIGDSTADDR, optval);
1738 OPTSET(INP_BINDANY);
1741 case IPV6_BINDMULTI:
1742 OPTSET2(INP_BINDMULTI, optval);
1745 case IPV6_RSS_LISTEN_BUCKET:
1746 if ((optval >= 0) &&
1747 (optval < rss_getnumbuckets())) {
1748 in6p->inp_rss_listen_bucket = optval;
1749 OPTSET2(INP_RSS_BUCKET_SET, 1);
1760 case IPV6_USE_MIN_MTU:
1761 case IPV6_PREFER_TEMPADDR:
1762 if (optlen != sizeof(optval)) {
1766 error = sooptcopyin(sopt, &optval,
1767 sizeof optval, sizeof optval);
1771 struct ip6_pktopts **optp;
1772 optp = &in6p->in6p_outputopts;
1773 error = ip6_pcbopt(optname,
1774 (u_char *)&optval, sizeof(optval),
1775 optp, (td != NULL) ? td->td_ucred :
1780 case IPV6_2292PKTINFO:
1781 case IPV6_2292HOPLIMIT:
1782 case IPV6_2292HOPOPTS:
1783 case IPV6_2292DSTOPTS:
1784 case IPV6_2292RTHDR:
1786 if (optlen != sizeof(int)) {
1790 error = sooptcopyin(sopt, &optval,
1791 sizeof optval, sizeof optval);
1795 case IPV6_2292PKTINFO:
1796 OPTSET2292(IN6P_PKTINFO);
1798 case IPV6_2292HOPLIMIT:
1799 OPTSET2292(IN6P_HOPLIMIT);
1801 case IPV6_2292HOPOPTS:
1803 * Check super-user privilege.
1804 * See comments for IPV6_RECVHOPOPTS.
1807 error = priv_check(td,
1808 PRIV_NETINET_SETHDROPTS);
1812 OPTSET2292(IN6P_HOPOPTS);
1814 case IPV6_2292DSTOPTS:
1816 error = priv_check(td,
1817 PRIV_NETINET_SETHDROPTS);
1821 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1823 case IPV6_2292RTHDR:
1824 OPTSET2292(IN6P_RTHDR);
1832 case IPV6_RTHDRDSTOPTS:
1835 /* new advanced API (RFC3542) */
1837 u_char optbuf_storage[MCLBYTES];
1839 struct ip6_pktopts **optp;
1841 /* cannot mix with RFC2292 */
1842 if (OPTBIT(IN6P_RFC2292)) {
1848 * We only ensure valsize is not too large
1849 * here. Further validation will be done
1852 error = sooptcopyin(sopt, optbuf_storage,
1853 sizeof(optbuf_storage), 0);
1856 optlen = sopt->sopt_valsize;
1857 optbuf = optbuf_storage;
1858 optp = &in6p->in6p_outputopts;
1859 error = ip6_pcbopt(optname, optbuf, optlen,
1860 optp, (td != NULL) ? td->td_ucred : NULL,
1866 case IPV6_MULTICAST_IF:
1867 case IPV6_MULTICAST_HOPS:
1868 case IPV6_MULTICAST_LOOP:
1869 case IPV6_JOIN_GROUP:
1870 case IPV6_LEAVE_GROUP:
1872 case MCAST_BLOCK_SOURCE:
1873 case MCAST_UNBLOCK_SOURCE:
1874 case MCAST_JOIN_GROUP:
1875 case MCAST_LEAVE_GROUP:
1876 case MCAST_JOIN_SOURCE_GROUP:
1877 case MCAST_LEAVE_SOURCE_GROUP:
1878 error = ip6_setmoptions(in6p, sopt);
1881 case IPV6_PORTRANGE:
1882 error = sooptcopyin(sopt, &optval,
1883 sizeof optval, sizeof optval);
1889 case IPV6_PORTRANGE_DEFAULT:
1890 in6p->inp_flags &= ~(INP_LOWPORT);
1891 in6p->inp_flags &= ~(INP_HIGHPORT);
1894 case IPV6_PORTRANGE_HIGH:
1895 in6p->inp_flags &= ~(INP_LOWPORT);
1896 in6p->inp_flags |= INP_HIGHPORT;
1899 case IPV6_PORTRANGE_LOW:
1900 in6p->inp_flags &= ~(INP_HIGHPORT);
1901 in6p->inp_flags |= INP_LOWPORT;
1911 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1912 case IPV6_IPSEC_POLICY:
1913 if (IPSEC_ENABLED(ipv6)) {
1914 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1921 error = ENOPROTOOPT;
1929 case IPV6_2292PKTOPTIONS:
1930 #ifdef IPV6_PKTOPTIONS
1931 case IPV6_PKTOPTIONS:
1934 * RFC3542 (effectively) deprecated the
1935 * semantics of the 2292-style pktoptions.
1936 * Since it was not reliable in nature (i.e.,
1937 * applications had to expect the lack of some
1938 * information after all), it would make sense
1939 * to simplify this part by always returning
1942 sopt->sopt_valsize = 0;
1945 case IPV6_RECVHOPOPTS:
1946 case IPV6_RECVDSTOPTS:
1947 case IPV6_RECVRTHDRDSTOPTS:
1948 case IPV6_UNICAST_HOPS:
1949 case IPV6_RECVPKTINFO:
1950 case IPV6_RECVHOPLIMIT:
1951 case IPV6_RECVRTHDR:
1952 case IPV6_RECVPATHMTU:
1955 case IPV6_PORTRANGE:
1956 case IPV6_RECVTCLASS:
1957 case IPV6_AUTOFLOWLABEL:
1961 case IPV6_RECVFLOWID:
1963 case IPV6_RSSBUCKETID:
1964 case IPV6_RECVRSSBUCKETID:
1966 case IPV6_BINDMULTI:
1969 case IPV6_RECVHOPOPTS:
1970 optval = OPTBIT(IN6P_HOPOPTS);
1973 case IPV6_RECVDSTOPTS:
1974 optval = OPTBIT(IN6P_DSTOPTS);
1977 case IPV6_RECVRTHDRDSTOPTS:
1978 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1981 case IPV6_UNICAST_HOPS:
1982 optval = in6p->in6p_hops;
1985 case IPV6_RECVPKTINFO:
1986 optval = OPTBIT(IN6P_PKTINFO);
1989 case IPV6_RECVHOPLIMIT:
1990 optval = OPTBIT(IN6P_HOPLIMIT);
1993 case IPV6_RECVRTHDR:
1994 optval = OPTBIT(IN6P_RTHDR);
1997 case IPV6_RECVPATHMTU:
1998 optval = OPTBIT(IN6P_MTU);
2002 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2005 case IPV6_PORTRANGE:
2008 flags = in6p->inp_flags;
2009 if (flags & INP_HIGHPORT)
2010 optval = IPV6_PORTRANGE_HIGH;
2011 else if (flags & INP_LOWPORT)
2012 optval = IPV6_PORTRANGE_LOW;
2017 case IPV6_RECVTCLASS:
2018 optval = OPTBIT(IN6P_TCLASS);
2021 case IPV6_AUTOFLOWLABEL:
2022 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2025 case IPV6_ORIGDSTADDR:
2026 optval = OPTBIT2(INP_ORIGDSTADDR);
2030 optval = OPTBIT(INP_BINDANY);
2034 optval = in6p->inp_flowid;
2038 optval = in6p->inp_flowtype;
2041 case IPV6_RECVFLOWID:
2042 optval = OPTBIT2(INP_RECVFLOWID);
2045 case IPV6_RSSBUCKETID:
2047 rss_hash2bucket(in6p->inp_flowid,
2051 optval = rss_bucket;
2056 case IPV6_RECVRSSBUCKETID:
2057 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2061 case IPV6_BINDMULTI:
2062 optval = OPTBIT2(INP_BINDMULTI);
2068 error = sooptcopyout(sopt, &optval,
2075 struct ip6_mtuinfo mtuinfo;
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.
2084 error = ip6_getpmtu_ctl(so->so_fibnum,
2085 &in6p->in6p_faddr, &pmtu);
2088 if (pmtu > IPV6_MAXPACKET)
2089 pmtu = IPV6_MAXPACKET;
2091 bzero(&mtuinfo, sizeof(mtuinfo));
2092 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2093 optdata = (void *)&mtuinfo;
2094 optdatalen = sizeof(mtuinfo);
2095 error = sooptcopyout(sopt, optdata,
2100 case IPV6_2292PKTINFO:
2101 case IPV6_2292HOPLIMIT:
2102 case IPV6_2292HOPOPTS:
2103 case IPV6_2292RTHDR:
2104 case IPV6_2292DSTOPTS:
2106 case IPV6_2292PKTINFO:
2107 optval = OPTBIT(IN6P_PKTINFO);
2109 case IPV6_2292HOPLIMIT:
2110 optval = OPTBIT(IN6P_HOPLIMIT);
2112 case IPV6_2292HOPOPTS:
2113 optval = OPTBIT(IN6P_HOPOPTS);
2115 case IPV6_2292RTHDR:
2116 optval = OPTBIT(IN6P_RTHDR);
2118 case IPV6_2292DSTOPTS:
2119 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2122 error = sooptcopyout(sopt, &optval,
2129 case IPV6_RTHDRDSTOPTS:
2133 case IPV6_USE_MIN_MTU:
2134 case IPV6_PREFER_TEMPADDR:
2135 error = ip6_getpcbopt(in6p->in6p_outputopts,
2139 case IPV6_MULTICAST_IF:
2140 case IPV6_MULTICAST_HOPS:
2141 case IPV6_MULTICAST_LOOP:
2143 error = ip6_getmoptions(in6p, sopt);
2146 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2147 case IPV6_IPSEC_POLICY:
2148 if (IPSEC_ENABLED(ipv6)) {
2149 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2155 error = ENOPROTOOPT;
2165 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2167 int error = 0, optval, optlen;
2168 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2169 struct inpcb *in6p = sotoinpcb(so);
2170 int level, op, optname;
2172 level = sopt->sopt_level;
2173 op = sopt->sopt_dir;
2174 optname = sopt->sopt_name;
2175 optlen = sopt->sopt_valsize;
2177 if (level != IPPROTO_IPV6) {
2184 * For ICMPv6 sockets, no modification allowed for checksum
2185 * offset, permit "no change" values to help existing apps.
2187 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2188 * for an ICMPv6 socket will fail."
2189 * The current behavior does not meet RFC3542.
2193 if (optlen != sizeof(int)) {
2197 error = sooptcopyin(sopt, &optval, sizeof(optval),
2201 if ((optval % 2) != 0) {
2202 /* the API assumes even offset values */
2204 } else if (so->so_proto->pr_protocol ==
2206 if (optval != icmp6off)
2209 in6p->in6p_cksum = optval;
2213 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2216 optval = in6p->in6p_cksum;
2218 error = sooptcopyout(sopt, &optval, sizeof(optval));
2228 error = ENOPROTOOPT;
2236 * Set up IP6 options in pcb for insertion in output packets or
2237 * specifying behavior of outgoing packets.
2240 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2241 struct socket *so, struct sockopt *sopt)
2243 struct ip6_pktopts *opt = *pktopt;
2245 struct thread *td = sopt->sopt_td;
2247 /* turn off any old options. */
2250 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2251 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2252 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2253 printf("ip6_pcbopts: all specified options are cleared.\n");
2255 ip6_clearpktopts(opt, -1);
2257 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2260 if (!m || m->m_len == 0) {
2262 * Only turning off any previous options, regardless of
2263 * whether the opt is just created or given.
2265 free(opt, M_IP6OPT);
2269 /* set options specified by user. */
2270 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2271 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2272 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2273 free(opt, M_IP6OPT);
2281 * initialize ip6_pktopts. beware that there are non-zero default values in
2285 ip6_initpktopts(struct ip6_pktopts *opt)
2288 bzero(opt, sizeof(*opt));
2289 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2290 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2291 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2292 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2296 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2297 struct ucred *cred, int uproto)
2299 struct ip6_pktopts *opt;
2301 if (*pktopt == NULL) {
2302 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2304 ip6_initpktopts(*pktopt);
2308 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2312 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2314 void *optdata = NULL;
2316 struct ip6_ext *ip6e;
2318 struct in6_pktinfo null_pktinfo;
2319 int deftclass = 0, on;
2320 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2321 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2325 optdata = (void *)&null_pktinfo;
2326 if (pktopt && pktopt->ip6po_pktinfo) {
2327 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2328 sizeof(null_pktinfo));
2329 in6_clearscope(&null_pktinfo.ipi6_addr);
2331 /* XXX: we don't have to do this every time... */
2332 bzero(&null_pktinfo, sizeof(null_pktinfo));
2334 optdatalen = sizeof(struct in6_pktinfo);
2337 if (pktopt && pktopt->ip6po_tclass >= 0)
2338 optdata = (void *)&pktopt->ip6po_tclass;
2340 optdata = (void *)&deftclass;
2341 optdatalen = sizeof(int);
2344 if (pktopt && pktopt->ip6po_hbh) {
2345 optdata = (void *)pktopt->ip6po_hbh;
2346 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2347 optdatalen = (ip6e->ip6e_len + 1) << 3;
2351 if (pktopt && pktopt->ip6po_rthdr) {
2352 optdata = (void *)pktopt->ip6po_rthdr;
2353 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2354 optdatalen = (ip6e->ip6e_len + 1) << 3;
2357 case IPV6_RTHDRDSTOPTS:
2358 if (pktopt && pktopt->ip6po_dest1) {
2359 optdata = (void *)pktopt->ip6po_dest1;
2360 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2361 optdatalen = (ip6e->ip6e_len + 1) << 3;
2365 if (pktopt && pktopt->ip6po_dest2) {
2366 optdata = (void *)pktopt->ip6po_dest2;
2367 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2368 optdatalen = (ip6e->ip6e_len + 1) << 3;
2372 if (pktopt && pktopt->ip6po_nexthop) {
2373 optdata = (void *)pktopt->ip6po_nexthop;
2374 optdatalen = pktopt->ip6po_nexthop->sa_len;
2377 case IPV6_USE_MIN_MTU:
2379 optdata = (void *)&pktopt->ip6po_minmtu;
2381 optdata = (void *)&defminmtu;
2382 optdatalen = sizeof(int);
2385 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2389 optdata = (void *)&on;
2390 optdatalen = sizeof(on);
2392 case IPV6_PREFER_TEMPADDR:
2394 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2396 optdata = (void *)&defpreftemp;
2397 optdatalen = sizeof(int);
2399 default: /* should not happen */
2401 panic("ip6_getpcbopt: unexpected option\n");
2403 return (ENOPROTOOPT);
2406 error = sooptcopyout(sopt, optdata, optdatalen);
2412 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2417 if (optname == -1 || optname == IPV6_PKTINFO) {
2418 if (pktopt->ip6po_pktinfo)
2419 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2420 pktopt->ip6po_pktinfo = NULL;
2422 if (optname == -1 || optname == IPV6_HOPLIMIT)
2423 pktopt->ip6po_hlim = -1;
2424 if (optname == -1 || optname == IPV6_TCLASS)
2425 pktopt->ip6po_tclass = -1;
2426 if (optname == -1 || optname == IPV6_NEXTHOP) {
2427 if (pktopt->ip6po_nextroute.ro_rt) {
2428 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2429 pktopt->ip6po_nextroute.ro_rt = NULL;
2431 if (pktopt->ip6po_nexthop)
2432 free(pktopt->ip6po_nexthop, M_IP6OPT);
2433 pktopt->ip6po_nexthop = NULL;
2435 if (optname == -1 || optname == IPV6_HOPOPTS) {
2436 if (pktopt->ip6po_hbh)
2437 free(pktopt->ip6po_hbh, M_IP6OPT);
2438 pktopt->ip6po_hbh = NULL;
2440 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2441 if (pktopt->ip6po_dest1)
2442 free(pktopt->ip6po_dest1, M_IP6OPT);
2443 pktopt->ip6po_dest1 = NULL;
2445 if (optname == -1 || optname == IPV6_RTHDR) {
2446 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2447 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2448 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2449 if (pktopt->ip6po_route.ro_rt) {
2450 RTFREE(pktopt->ip6po_route.ro_rt);
2451 pktopt->ip6po_route.ro_rt = NULL;
2454 if (optname == -1 || optname == IPV6_DSTOPTS) {
2455 if (pktopt->ip6po_dest2)
2456 free(pktopt->ip6po_dest2, M_IP6OPT);
2457 pktopt->ip6po_dest2 = NULL;
2461 #define PKTOPT_EXTHDRCPY(type) \
2464 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2465 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2466 if (dst->type == NULL && canwait == M_NOWAIT)\
2468 bcopy(src->type, dst->type, hlen);\
2470 } while (/*CONSTCOND*/ 0)
2473 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2475 if (dst == NULL || src == NULL) {
2476 printf("ip6_clearpktopts: invalid argument\n");
2480 dst->ip6po_hlim = src->ip6po_hlim;
2481 dst->ip6po_tclass = src->ip6po_tclass;
2482 dst->ip6po_flags = src->ip6po_flags;
2483 dst->ip6po_minmtu = src->ip6po_minmtu;
2484 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2485 if (src->ip6po_pktinfo) {
2486 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2488 if (dst->ip6po_pktinfo == NULL)
2490 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2492 if (src->ip6po_nexthop) {
2493 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2495 if (dst->ip6po_nexthop == NULL)
2497 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2498 src->ip6po_nexthop->sa_len);
2500 PKTOPT_EXTHDRCPY(ip6po_hbh);
2501 PKTOPT_EXTHDRCPY(ip6po_dest1);
2502 PKTOPT_EXTHDRCPY(ip6po_dest2);
2503 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2507 ip6_clearpktopts(dst, -1);
2510 #undef PKTOPT_EXTHDRCPY
2512 struct ip6_pktopts *
2513 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2516 struct ip6_pktopts *dst;
2518 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2521 ip6_initpktopts(dst);
2523 if ((error = copypktopts(dst, src, canwait)) != 0) {
2524 free(dst, M_IP6OPT);
2532 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2537 ip6_clearpktopts(pktopt, -1);
2539 free(pktopt, M_IP6OPT);
2543 * Set IPv6 outgoing packet options based on advanced API.
2546 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2547 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2549 struct cmsghdr *cm = NULL;
2551 if (control == NULL || opt == NULL)
2554 ip6_initpktopts(opt);
2559 * If stickyopt is provided, make a local copy of the options
2560 * for this particular packet, then override them by ancillary
2562 * XXX: copypktopts() does not copy the cached route to a next
2563 * hop (if any). This is not very good in terms of efficiency,
2564 * but we can allow this since this option should be rarely
2567 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2572 * XXX: Currently, we assume all the optional information is stored
2575 if (control->m_next)
2578 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2579 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2582 if (control->m_len < CMSG_LEN(0))
2585 cm = mtod(control, struct cmsghdr *);
2586 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2588 if (cm->cmsg_level != IPPROTO_IPV6)
2591 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2592 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2601 * Set a particular packet option, as a sticky option or an ancillary data
2602 * item. "len" can be 0 only when it's a sticky option.
2603 * We have 4 cases of combination of "sticky" and "cmsg":
2604 * "sticky=0, cmsg=0": impossible
2605 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2606 * "sticky=1, cmsg=0": RFC3542 socket option
2607 * "sticky=1, cmsg=1": RFC2292 socket option
2610 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2611 struct ucred *cred, int sticky, int cmsg, int uproto)
2613 int minmtupolicy, preftemp;
2616 if (!sticky && !cmsg) {
2618 printf("ip6_setpktopt: impossible case\n");
2624 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2625 * not be specified in the context of RFC3542. Conversely,
2626 * RFC3542 types should not be specified in the context of RFC2292.
2630 case IPV6_2292PKTINFO:
2631 case IPV6_2292HOPLIMIT:
2632 case IPV6_2292NEXTHOP:
2633 case IPV6_2292HOPOPTS:
2634 case IPV6_2292DSTOPTS:
2635 case IPV6_2292RTHDR:
2636 case IPV6_2292PKTOPTIONS:
2637 return (ENOPROTOOPT);
2640 if (sticky && cmsg) {
2647 case IPV6_RTHDRDSTOPTS:
2649 case IPV6_USE_MIN_MTU:
2652 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2653 return (ENOPROTOOPT);
2658 case IPV6_2292PKTINFO:
2661 struct ifnet *ifp = NULL;
2662 struct in6_pktinfo *pktinfo;
2664 if (len != sizeof(struct in6_pktinfo))
2667 pktinfo = (struct in6_pktinfo *)buf;
2670 * An application can clear any sticky IPV6_PKTINFO option by
2671 * doing a "regular" setsockopt with ipi6_addr being
2672 * in6addr_any and ipi6_ifindex being zero.
2673 * [RFC 3542, Section 6]
2675 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2676 pktinfo->ipi6_ifindex == 0 &&
2677 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678 ip6_clearpktopts(opt, optname);
2682 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2683 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2686 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2688 /* validate the interface index if specified. */
2689 if (pktinfo->ipi6_ifindex > V_if_index)
2691 if (pktinfo->ipi6_ifindex) {
2692 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2696 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2697 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2701 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2702 struct in6_ifaddr *ia;
2704 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2705 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2707 return (EADDRNOTAVAIL);
2708 ifa_free(&ia->ia_ifa);
2711 * We store the address anyway, and let in6_selectsrc()
2712 * validate the specified address. This is because ipi6_addr
2713 * may not have enough information about its scope zone, and
2714 * we may need additional information (such as outgoing
2715 * interface or the scope zone of a destination address) to
2716 * disambiguate the scope.
2717 * XXX: the delay of the validation may confuse the
2718 * application when it is used as a sticky option.
2720 if (opt->ip6po_pktinfo == NULL) {
2721 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2722 M_IP6OPT, M_NOWAIT);
2723 if (opt->ip6po_pktinfo == NULL)
2726 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2730 case IPV6_2292HOPLIMIT:
2736 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2737 * to simplify the ordering among hoplimit options.
2739 if (optname == IPV6_HOPLIMIT && sticky)
2740 return (ENOPROTOOPT);
2742 if (len != sizeof(int))
2745 if (*hlimp < -1 || *hlimp > 255)
2748 opt->ip6po_hlim = *hlimp;
2756 if (len != sizeof(int))
2758 tclass = *(int *)buf;
2759 if (tclass < -1 || tclass > 255)
2762 opt->ip6po_tclass = tclass;
2766 case IPV6_2292NEXTHOP:
2769 error = priv_check_cred(cred,
2770 PRIV_NETINET_SETHDROPTS, 0);
2775 if (len == 0) { /* just remove the option */
2776 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2780 /* check if cmsg_len is large enough for sa_len */
2781 if (len < sizeof(struct sockaddr) || len < *buf)
2784 switch (((struct sockaddr *)buf)->sa_family) {
2787 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2790 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2793 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2794 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2797 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2803 case AF_LINK: /* should eventually be supported */
2805 return (EAFNOSUPPORT);
2808 /* turn off the previous option, then set the new option. */
2809 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2810 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2811 if (opt->ip6po_nexthop == NULL)
2813 bcopy(buf, opt->ip6po_nexthop, *buf);
2816 case IPV6_2292HOPOPTS:
2819 struct ip6_hbh *hbh;
2823 * XXX: We don't allow a non-privileged user to set ANY HbH
2824 * options, since per-option restriction has too much
2828 error = priv_check_cred(cred,
2829 PRIV_NETINET_SETHDROPTS, 0);
2835 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2836 break; /* just remove the option */
2839 /* message length validation */
2840 if (len < sizeof(struct ip6_hbh))
2842 hbh = (struct ip6_hbh *)buf;
2843 hbhlen = (hbh->ip6h_len + 1) << 3;
2847 /* turn off the previous option, then set the new option. */
2848 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2849 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2850 if (opt->ip6po_hbh == NULL)
2852 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2857 case IPV6_2292DSTOPTS:
2859 case IPV6_RTHDRDSTOPTS:
2861 struct ip6_dest *dest, **newdest = NULL;
2864 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2865 error = priv_check_cred(cred,
2866 PRIV_NETINET_SETHDROPTS, 0);
2872 ip6_clearpktopts(opt, optname);
2873 break; /* just remove the option */
2876 /* message length validation */
2877 if (len < sizeof(struct ip6_dest))
2879 dest = (struct ip6_dest *)buf;
2880 destlen = (dest->ip6d_len + 1) << 3;
2885 * Determine the position that the destination options header
2886 * should be inserted; before or after the routing header.
2889 case IPV6_2292DSTOPTS:
2891 * The old advacned API is ambiguous on this point.
2892 * Our approach is to determine the position based
2893 * according to the existence of a routing header.
2894 * Note, however, that this depends on the order of the
2895 * extension headers in the ancillary data; the 1st
2896 * part of the destination options header must appear
2897 * before the routing header in the ancillary data,
2899 * RFC3542 solved the ambiguity by introducing
2900 * separate ancillary data or option types.
2902 if (opt->ip6po_rthdr == NULL)
2903 newdest = &opt->ip6po_dest1;
2905 newdest = &opt->ip6po_dest2;
2907 case IPV6_RTHDRDSTOPTS:
2908 newdest = &opt->ip6po_dest1;
2911 newdest = &opt->ip6po_dest2;
2915 /* turn off the previous option, then set the new option. */
2916 ip6_clearpktopts(opt, optname);
2917 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2918 if (*newdest == NULL)
2920 bcopy(dest, *newdest, destlen);
2925 case IPV6_2292RTHDR:
2928 struct ip6_rthdr *rth;
2932 ip6_clearpktopts(opt, IPV6_RTHDR);
2933 break; /* just remove the option */
2936 /* message length validation */
2937 if (len < sizeof(struct ip6_rthdr))
2939 rth = (struct ip6_rthdr *)buf;
2940 rthlen = (rth->ip6r_len + 1) << 3;
2944 switch (rth->ip6r_type) {
2945 case IPV6_RTHDR_TYPE_0:
2946 if (rth->ip6r_len == 0) /* must contain one addr */
2948 if (rth->ip6r_len % 2) /* length must be even */
2950 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2954 return (EINVAL); /* not supported */
2957 /* turn off the previous option */
2958 ip6_clearpktopts(opt, IPV6_RTHDR);
2959 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2960 if (opt->ip6po_rthdr == NULL)
2962 bcopy(rth, opt->ip6po_rthdr, rthlen);
2967 case IPV6_USE_MIN_MTU:
2968 if (len != sizeof(int))
2970 minmtupolicy = *(int *)buf;
2971 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2972 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2973 minmtupolicy != IP6PO_MINMTU_ALL) {
2976 opt->ip6po_minmtu = minmtupolicy;
2980 if (len != sizeof(int))
2983 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2985 * we ignore this option for TCP sockets.
2986 * (RFC3542 leaves this case unspecified.)
2988 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2990 opt->ip6po_flags |= IP6PO_DONTFRAG;
2993 case IPV6_PREFER_TEMPADDR:
2994 if (len != sizeof(int))
2996 preftemp = *(int *)buf;
2997 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2998 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2999 preftemp != IP6PO_TEMPADDR_PREFER) {
3002 opt->ip6po_prefer_tempaddr = preftemp;
3006 return (ENOPROTOOPT);
3007 } /* end of switch */
3013 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3014 * packet to the input queue of a specified interface. Note that this
3015 * calls the output routine of the loopback "driver", but with an interface
3016 * pointer that might NOT be &loif -- easier than replicating that code here.
3019 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3022 struct ip6_hdr *ip6;
3024 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3029 * Make sure to deep-copy IPv6 header portion in case the data
3030 * is in an mbuf cluster, so that we can safely override the IPv6
3031 * header portion later.
3033 if (!M_WRITABLE(copym) ||
3034 copym->m_len < sizeof(struct ip6_hdr)) {
3035 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3039 ip6 = mtod(copym, struct ip6_hdr *);
3041 * clear embedded scope identifiers if necessary.
3042 * in6_clearscope will touch the addresses only when necessary.
3044 in6_clearscope(&ip6->ip6_src);
3045 in6_clearscope(&ip6->ip6_dst);
3046 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3047 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3049 copym->m_pkthdr.csum_data = 0xffff;
3051 if_simloop(ifp, copym, AF_INET6, 0);
3055 * Chop IPv6 header off from the payload.
3058 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3061 struct ip6_hdr *ip6;
3063 ip6 = mtod(m, struct ip6_hdr *);
3064 if (m->m_len > sizeof(*ip6)) {
3065 mh = m_gethdr(M_NOWAIT, MT_DATA);
3070 m_move_pkthdr(mh, m);
3071 M_ALIGN(mh, sizeof(*ip6));
3072 m->m_len -= sizeof(*ip6);
3073 m->m_data += sizeof(*ip6);
3076 m->m_len = sizeof(*ip6);
3077 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3079 exthdrs->ip6e_ip6 = m;
3084 * Compute IPv6 extension header length.
3087 ip6_optlen(struct inpcb *in6p)
3091 if (!in6p->in6p_outputopts)
3096 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3098 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3099 if (in6p->in6p_outputopts->ip6po_rthdr)
3100 /* dest1 is valid with rthdr only */
3101 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3102 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3103 len += elen(in6p->in6p_outputopts->ip6po_dest2);