sys/netinet6/ip6_output.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  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.
  18  *
  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
  29  * SUCH DAMAGE.
  30  *
  31  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  32  */
  33
  34 /*-
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  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.
  49  *
  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
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  63  */
  64
  65 #include <sys/cdefs.h>
  66 __FBSDID("$FreeBSD$");
  67
  68 #include "opt_inet.h"
  69 #include "opt_inet6.h"
  70 #include "opt_ipsec.h"
  71 #include "opt_kern_tls.h"
  72 #include "opt_ratelimit.h"
  73 #include "opt_route.h"
  74 #include "opt_rss.h"
  75 #include "opt_sctp.h"
  76
  77 #include <sys/param.h>
  78 #include <sys/kernel.h>
  79 #include <sys/ktls.h>
  80 #include <sys/malloc.h>
  81 #include <sys/mbuf.h>
  82 #include <sys/errno.h>
  83 #include <sys/priv.h>
  84 #include <sys/proc.h>
  85 #include <sys/protosw.h>
  86 #include <sys/socket.h>
  87 #include <sys/socketvar.h>
  88 #include <sys/syslog.h>
  89 #include <sys/ucred.h>
  90
  91 #include <machine/in_cksum.h>
  92
  93 #include <net/if.h>
  94 #include <net/if_var.h>
  95 #include <net/if_private.h>
  96 #include <net/if_vlan_var.h>
  97 #include <net/if_llatbl.h>
  98 #include <net/ethernet.h>
  99 #include <net/netisr.h>
 100 #include <net/route.h>
 101 #include <net/route/nhop.h>
 102 #include <net/pfil.h>
 103 #include <net/rss_config.h>
 104 #include <net/vnet.h>
 105
 106 #include <netinet/in.h>
 107 #include <netinet/in_var.h>
 108 #include <netinet/ip_var.h>
 109 #include <netinet6/in6_fib.h>
 110 #include <netinet6/in6_var.h>
 111 #include <netinet/ip6.h>
 112 #include <netinet/icmp6.h>
 113 #include <netinet6/ip6_var.h>
 114 #include <netinet/in_pcb.h>
 115 #include <netinet/tcp_var.h>
 116 #include <netinet6/nd6.h>
 117 #include <netinet6/in6_rss.h>
 118
 119 #include <netipsec/ipsec_support.h>
 120 #if defined(SCTP) || defined(SCTP_SUPPORT)
 121 #include <netinet/sctp.h>
 122 #include <netinet/sctp_crc32.h>
 123 #endif
 124
 125 #include <netinet6/scope6_var.h>
 126
 127 extern int in6_mcast_loop;
 128
 129 struct ip6_exthdrs {
 130         struct mbuf *ip6e_ip6;
 131         struct mbuf *ip6e_hbh;
 132         struct mbuf *ip6e_dest1;
 133         struct mbuf *ip6e_rthdr;
 134         struct mbuf *ip6e_dest2;
 135 };
 136
 137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 138
 139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 140                            struct ucred *, int);
 141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 142         struct socket *, struct sockopt *);
 143 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
 144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 145         struct ucred *, int, int, int);
 146
 147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 149         struct ip6_frag **);
 150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 152 static int ip6_getpmtu(struct route_in6 *, int,
 153         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 154         u_int);
 155 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 156         u_long *, int *, u_int);
 157 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 158 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 159
 160 /*
 161  * Make an extension header from option data.  hp is the source,
 162  * mp is the destination, and _ol is the optlen.
 163  */
 164 #define MAKE_EXTHDR(hp, mp, _ol)                                        \
 165     do {                                                                \
 166         if (hp) {                                                       \
 167                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 168                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 169                     ((eh)->ip6e_len + 1) << 3);                         \
 170                 if (error)                                              \
 171                         goto freehdrs;                                  \
 172                 (_ol) += (*(mp))->m_len;                                \
 173         }                                                               \
 174     } while (/*CONSTCOND*/ 0)
 175
 176 /*
 177  * Form a chain of extension headers.
 178  * m is the extension header mbuf
 179  * mp is the previous mbuf in the chain
 180  * p is the next header
 181  * i is the type of option.
 182  */
 183 #define MAKE_CHAIN(m, mp, p, i)\
 184     do {\
 185         if (m) {\
 186                 if (!hdrsplit) \
 187                         panic("%s:%d: assumption failed: "\
 188                             "hdr not split: hdrsplit %d exthdrs %p",\
 189                             __func__, __LINE__, hdrsplit, &exthdrs);\
 190                 *mtod((m), u_char *) = *(p);\
 191                 *(p) = (i);\
 192                 p = mtod((m), u_char *);\
 193                 (m)->m_next = (mp)->m_next;\
 194                 (mp)->m_next = (m);\
 195                 (mp) = (m);\
 196         }\
 197     } while (/*CONSTCOND*/ 0)
 198
 199 void
 200 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 201 {
 202         u_short csum;
 203
 204         csum = in_cksum_skip(m, offset + plen, offset);
 205         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 206                 csum = 0xffff;
 207         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 208
 209         if (offset + sizeof(csum) > m->m_len)
 210                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 211         else
 212                 *(u_short *)mtodo(m, offset) = csum;
 213 }
 214
 215 static void
 216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
 217     int plen, int optlen)
 218 {
 219
 220         KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
 221             "csum_flags %#x",
 222             __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
 223
 224         if (csum_flags & CSUM_DELAY_DATA_IPV6) {
 225                 in6_delayed_cksum(m, plen - optlen,
 226                     sizeof(struct ip6_hdr) + optlen);
 227                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 228         }
 229 #if defined(SCTP) || defined(SCTP_SUPPORT)
 230         if (csum_flags & CSUM_SCTP_IPV6) {
 231                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
 232                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 233         }
 234 #endif
 235 }
 236
 237 int
 238 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 239     int fraglen , uint32_t id)
 240 {
 241         struct mbuf *m, **mnext, *m_frgpart;
 242         struct ip6_hdr *ip6, *mhip6;
 243         struct ip6_frag *ip6f;
 244         int off;
 245         int error;
 246         int tlen = m0->m_pkthdr.len;
 247
 248         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 249
 250         m = m0;
 251         ip6 = mtod(m, struct ip6_hdr *);
 252         mnext = &m->m_nextpkt;
 253
 254         for (off = hlen; off < tlen; off += fraglen) {
 255                 m = m_gethdr(M_NOWAIT, MT_DATA);
 256                 if (!m) {
 257                         IP6STAT_INC(ip6s_odropped);
 258                         return (ENOBUFS);
 259                 }
 260
 261                 /*
 262                  * Make sure the complete packet header gets copied
 263                  * from the originating mbuf to the newly created
 264                  * mbuf. This also ensures that existing firewall
 265                  * classification(s), VLAN tags and so on get copied
 266                  * to the resulting fragmented packet(s):
 267                  */
 268                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 269                         m_free(m);
 270                         IP6STAT_INC(ip6s_odropped);
 271                         return (ENOBUFS);
 272                 }
 273
 274                 *mnext = m;
 275                 mnext = &m->m_nextpkt;
 276                 m->m_data += max_linkhdr;
 277                 mhip6 = mtod(m, struct ip6_hdr *);
 278                 *mhip6 = *ip6;
 279                 m->m_len = sizeof(*mhip6);
 280                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 281                 if (error) {
 282                         IP6STAT_INC(ip6s_odropped);
 283                         return (error);
 284                 }
 285                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 286                 if (off + fraglen >= tlen)
 287                         fraglen = tlen - off;
 288                 else
 289                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 290                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 291                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 292                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 293                         IP6STAT_INC(ip6s_odropped);
 294                         return (ENOBUFS);
 295                 }
 296                 m_cat(m, m_frgpart);
 297                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 298                 ip6f->ip6f_reserved = 0;
 299                 ip6f->ip6f_ident = id;
 300                 ip6f->ip6f_nxt = nextproto;
 301                 IP6STAT_INC(ip6s_ofragments);
 302                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 303         }
 304
 305         return (0);
 306 }
 307
 308 static int
 309 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
 310     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
 311     bool stamp_tag)
 312 {
 313 #ifdef KERN_TLS
 314         struct ktls_session *tls = NULL;
 315 #endif
 316         struct m_snd_tag *mst;
 317         int error;
 318
 319         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 320         mst = NULL;
 321
 322 #ifdef KERN_TLS
 323         /*
 324          * If this is an unencrypted TLS record, save a reference to
 325          * the record.  This local reference is used to call
 326          * ktls_output_eagain after the mbuf has been freed (thus
 327          * dropping the mbuf's reference) in if_output.
 328          */
 329         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 330                 tls = ktls_hold(m->m_next->m_epg_tls);
 331                 mst = tls->snd_tag;
 332
 333                 /*
 334                  * If a TLS session doesn't have a valid tag, it must
 335                  * have had an earlier ifp mismatch, so drop this
 336                  * packet.
 337                  */
 338                 if (mst == NULL) {
 339                         m_freem(m);
 340                         error = EAGAIN;
 341                         goto done;
 342                 }
 343                 /*
 344                  * Always stamp tags that include NIC ktls.
 345                  */
 346                 stamp_tag = true;
 347         }
 348 #endif
 349 #ifdef RATELIMIT
 350         if (inp != NULL && mst == NULL) {
 351                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 352                     (inp->inp_snd_tag != NULL &&
 353                     inp->inp_snd_tag->ifp != ifp))
 354                         in_pcboutput_txrtlmt(inp, ifp, m);
 355
 356                 if (inp->inp_snd_tag != NULL)
 357                         mst = inp->inp_snd_tag;
 358         }
 359 #endif
 360         if (stamp_tag && mst != NULL) {
 361                 KASSERT(m->m_pkthdr.rcvif == NULL,
 362                     ("trying to add a send tag to a forwarded packet"));
 363                 if (mst->ifp != ifp) {
 364                         m_freem(m);
 365                         error = EAGAIN;
 366                         goto done;
 367                 }
 368
 369                 /* stamp send tag on mbuf */
 370                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 371                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 372         }
 373
 374         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 375
 376 done:
 377         /* Check for route change invalidating send tags. */
 378 #ifdef KERN_TLS
 379         if (tls != NULL) {
 380                 if (error == EAGAIN)
 381                         error = ktls_output_eagain(inp, tls);
 382                 ktls_free(tls);
 383         }
 384 #endif
 385 #ifdef RATELIMIT
 386         if (error == EAGAIN)
 387                 in_pcboutput_eagain(inp);
 388 #endif
 389         return (error);
 390 }
 391
 392 /*
 393  * IP6 output.
 394  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
 395  * nxt, hlim, src, dst).
 396  * This function may modify ver and hlim only.
 397  * The mbuf chain containing the packet will be freed.
 398  * The mbuf opt, if present, will not be freed.
 399  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
 400  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
 401  * then result of route lookup is stored in ro->ro_nh.
 402  *
 403  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
 404  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
 405  *
 406  * ifpp - XXX: just for statistics
 407  */
 408 int
 409 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 410     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 411     struct ifnet **ifpp, struct inpcb *inp)
 412 {
 413         struct ip6_hdr *ip6;
 414         struct ifnet *ifp, *origifp;
 415         struct mbuf *m = m0;
 416         struct mbuf *mprev;
 417         struct route_in6 *ro_pmtu;
 418         struct nhop_object *nh;
 419         struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
 420         struct in6_addr odst;
 421         u_char *nexthdrp;
 422         int tlen, len;
 423         int error = 0;
 424         int vlan_pcp = -1;
 425         struct in6_ifaddr *ia = NULL;
 426         u_long mtu;
 427         int alwaysfrag, dontfrag;
 428         u_int32_t optlen, plen = 0, unfragpartlen;
 429         struct ip6_exthdrs exthdrs;
 430         struct in6_addr src0, dst0;
 431         u_int32_t zone;
 432         bool hdrsplit;
 433         int sw_csum, tso;
 434         int needfiblookup;
 435         uint32_t fibnum;
 436         struct m_tag *fwd_tag = NULL;
 437         uint32_t id;
 438
 439         NET_EPOCH_ASSERT();
 440
 441         if (inp != NULL) {
 442                 INP_LOCK_ASSERT(inp);
 443                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 444                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 445                         /* Unconditionally set flowid. */
 446                         m->m_pkthdr.flowid = inp->inp_flowid;
 447                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 448                 }
 449                 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
 450                         vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
 451                             INP_2PCP_SHIFT;
 452 #ifdef NUMA
 453                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 454 #endif
 455         }
 456
 457 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 458         /*
 459          * IPSec checking which handles several cases.
 460          * FAST IPSEC: We re-injected the packet.
 461          * XXX: need scope argument.
 462          */
 463         if (IPSEC_ENABLED(ipv6)) {
 464                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 465                         if (error == EINPROGRESS)
 466                                 error = 0;
 467                         goto done;
 468                 }
 469         }
 470 #endif /* IPSEC */
 471
 472         /* Source address validation. */
 473         ip6 = mtod(m, struct ip6_hdr *);
 474         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 475             (flags & IPV6_UNSPECSRC) == 0) {
 476                 error = EOPNOTSUPP;
 477                 IP6STAT_INC(ip6s_badscope);
 478                 goto bad;
 479         }
 480         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 481                 error = EOPNOTSUPP;
 482                 IP6STAT_INC(ip6s_badscope);
 483                 goto bad;
 484         }
 485
 486         /*
 487          * If we are given packet options to add extension headers prepare them.
 488          * Calculate the total length of the extension header chain.
 489          * Keep the length of the unfragmentable part for fragmentation.
 490          */
 491         bzero(&exthdrs, sizeof(exthdrs));
 492         optlen = 0;
 493         unfragpartlen = sizeof(struct ip6_hdr);
 494         if (opt) {
 495                 /* Hop-by-Hop options header. */
 496                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
 497
 498                 /* Destination options header (1st part). */
 499                 if (opt->ip6po_rthdr) {
 500 #ifndef RTHDR_SUPPORT_IMPLEMENTED
 501                         /*
 502                          * If there is a routing header, discard the packet
 503                          * right away here. RH0/1 are obsolete and we do not
 504                          * currently support RH2/3/4.
 505                          * People trying to use RH253/254 may want to disable
 506                          * this check.
 507                          * The moment we do support any routing header (again)
 508                          * this block should check the routing type more
 509                          * selectively.
 510                          */
 511                         error = EINVAL;
 512                         goto bad;
 513 #endif
 514
 515                         /*
 516                          * Destination options header (1st part).
 517                          * This only makes sense with a routing header.
 518                          * See Section 9.2 of RFC 3542.
 519                          * Disabling this part just for MIP6 convenience is
 520                          * a bad idea.  We need to think carefully about a
 521                          * way to make the advanced API coexist with MIP6
 522                          * options, which might automatically be inserted in
 523                          * the kernel.
 524                          */
 525                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
 526                             optlen);
 527                 }
 528                 /* Routing header. */
 529                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
 530
 531                 unfragpartlen += optlen;
 532
 533                 /*
 534                  * NOTE: we don't add AH/ESP length here (done in
 535                  * ip6_ipsec_output()).
 536                  */
 537
 538                 /* Destination options header (2nd part). */
 539                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
 540         }
 541
 542         /*
 543          * If there is at least one extension header,
 544          * separate IP6 header from the payload.
 545          */
 546         hdrsplit = false;
 547         if (optlen) {
 548                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 549                         m = NULL;
 550                         goto freehdrs;
 551                 }
 552                 m = exthdrs.ip6e_ip6;
 553                 ip6 = mtod(m, struct ip6_hdr *);
 554                 hdrsplit = true;
 555         }
 556
 557         /* Adjust mbuf packet header length. */
 558         m->m_pkthdr.len += optlen;
 559         plen = m->m_pkthdr.len - sizeof(*ip6);
 560
 561         /* If this is a jumbo payload, insert a jumbo payload option. */
 562         if (plen > IPV6_MAXPACKET) {
 563                 if (!hdrsplit) {
 564                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 565                                 m = NULL;
 566                                 goto freehdrs;
 567                         }
 568                         m = exthdrs.ip6e_ip6;
 569                         ip6 = mtod(m, struct ip6_hdr *);
 570                         hdrsplit = true;
 571                 }
 572                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 573                         goto freehdrs;
 574                 ip6->ip6_plen = 0;
 575         } else
 576                 ip6->ip6_plen = htons(plen);
 577         nexthdrp = &ip6->ip6_nxt;
 578
 579         if (optlen) {
 580                 /*
 581                  * Concatenate headers and fill in next header fields.
 582                  * Here we have, on "m"
 583                  *      IPv6 payload
 584                  * and we insert headers accordingly.
 585                  * Finally, we should be getting:
 586                  *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
 587                  *
 588                  * During the header composing process "m" points to IPv6
 589                  * header.  "mprev" points to an extension header prior to esp.
 590                  */
 591                 mprev = m;
 592
 593                 /*
 594                  * We treat dest2 specially.  This makes IPsec processing
 595                  * much easier.  The goal here is to make mprev point the
 596                  * mbuf prior to dest2.
 597                  *
 598                  * Result: IPv6 dest2 payload.
 599                  * m and mprev will point to IPv6 header.
 600                  */
 601                 if (exthdrs.ip6e_dest2) {
 602                         if (!hdrsplit)
 603                                 panic("%s:%d: assumption failed: "
 604                                     "hdr not split: hdrsplit %d exthdrs %p",
 605                                     __func__, __LINE__, hdrsplit, &exthdrs);
 606                         exthdrs.ip6e_dest2->m_next = m->m_next;
 607                         m->m_next = exthdrs.ip6e_dest2;
 608                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 609                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 610                 }
 611
 612                 /*
 613                  * Result: IPv6 hbh dest1 rthdr dest2 payload.
 614                  * m will point to IPv6 header.  mprev will point to the
 615                  * extension header prior to dest2 (rthdr in the above case).
 616                  */
 617                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 618                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 619                            IPPROTO_DSTOPTS);
 620                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 621                            IPPROTO_ROUTING);
 622         }
 623
 624         IP6STAT_INC(ip6s_localout);
 625
 626         /* Route packet. */
 627         ro_pmtu = ro;
 628         if (opt && opt->ip6po_rthdr)
 629                 ro = &opt->ip6po_route;
 630         if (ro != NULL)
 631                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 632         else
 633                 dst = &sin6;
 634         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 635
 636 again:
 637         /*
 638          * If specified, try to fill in the traffic class field.
 639          * Do not override if a non-zero value is already set.
 640          * We check the diffserv field and the ECN field separately.
 641          */
 642         if (opt && opt->ip6po_tclass >= 0) {
 643                 int mask = 0;
 644
 645                 if (IPV6_DSCP(ip6) == 0)
 646                         mask |= 0xfc;
 647                 if (IPV6_ECN(ip6) == 0)
 648                         mask |= 0x03;
 649                 if (mask != 0)
 650                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 651         }
 652
 653         /* Fill in or override the hop limit field, if necessary. */
 654         if (opt && opt->ip6po_hlim != -1)
 655                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 656         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 657                 if (im6o != NULL)
 658                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 659                 else
 660                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 661         }
 662
 663         if (ro == NULL || ro->ro_nh == NULL) {
 664                 bzero(dst, sizeof(*dst));
 665                 dst->sin6_family = AF_INET6;
 666                 dst->sin6_len = sizeof(*dst);
 667                 dst->sin6_addr = ip6->ip6_dst;
 668         }
 669         /*
 670          * Validate route against routing table changes.
 671          * Make sure that the address family is set in route.
 672          */
 673         nh = NULL;
 674         ifp = NULL;
 675         mtu = 0;
 676         if (ro != NULL) {
 677                 if (ro->ro_nh != NULL && inp != NULL) {
 678                         ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
 679                         NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
 680                             fibnum);
 681                 }
 682                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 683                     (!NH_IS_VALID(ro->ro_nh) ||
 684                     ro->ro_dst.sin6_family != AF_INET6 ||
 685                     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
 686                         RO_INVALIDATE_CACHE(ro);
 687
 688                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 689                     ro->ro_dst.sin6_family == AF_INET6 &&
 690                     IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 691                         /* Nexthop is valid and contains valid ifp */
 692                         nh = ro->ro_nh;
 693                 } else {
 694                         if (ro->ro_lle)
 695                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 696                         ro->ro_lle = NULL;
 697                         if (fwd_tag == NULL) {
 698                                 bzero(&dst_sa, sizeof(dst_sa));
 699                                 dst_sa.sin6_family = AF_INET6;
 700                                 dst_sa.sin6_len = sizeof(dst_sa);
 701                                 dst_sa.sin6_addr = ip6->ip6_dst;
 702                         }
 703                         error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
 704                             &nh, fibnum, m->m_pkthdr.flowid);
 705                         if (error != 0) {
 706                                 IP6STAT_INC(ip6s_noroute);
 707                                 if (ifp != NULL)
 708                                         in6_ifstat_inc(ifp, ifs6_out_discard);
 709                                 goto bad;
 710                         }
 711                         /*
 712                          * At this point at least @ifp is not NULL
 713                          * Can be the case when dst is multicast, link-local or
 714                          * interface is explicitly specificed by the caller.
 715                          */
 716                 }
 717                 if (nh == NULL) {
 718                         /*
 719                          * If in6_selectroute() does not return a nexthop
 720                          * dst may not have been updated.
 721                          */
 722                         *dst = dst_sa;  /* XXX */
 723                         origifp = ifp;
 724                         mtu = ifp->if_mtu;
 725                 } else {
 726                         ifp = nh->nh_ifp;
 727                         origifp = nh->nh_aifp;
 728                         ia = (struct in6_ifaddr *)(nh->nh_ifa);
 729                         counter_u64_add(nh->nh_pksent, 1);
 730                 }
 731         } else {
 732                 struct nhop_object *nh;
 733                 struct in6_addr kdst;
 734                 uint32_t scopeid;
 735
 736                 if (fwd_tag == NULL) {
 737                         bzero(&dst_sa, sizeof(dst_sa));
 738                         dst_sa.sin6_family = AF_INET6;
 739                         dst_sa.sin6_len = sizeof(dst_sa);
 740                         dst_sa.sin6_addr = ip6->ip6_dst;
 741                 }
 742
 743                 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
 744                     im6o != NULL &&
 745                     (ifp = im6o->im6o_multicast_ifp) != NULL) {
 746                         /* We do not need a route lookup. */
 747                         *dst = dst_sa;  /* XXX */
 748                         origifp = ifp;
 749                         goto nonh6lookup;
 750                 }
 751
 752                 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
 753
 754                 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
 755                     IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
 756                         if (scopeid > 0) {
 757                                 ifp = in6_getlinkifnet(scopeid);
 758                                 if (ifp == NULL) {
 759                                         error = EHOSTUNREACH;
 760                                         goto bad;
 761                                 }
 762                                 *dst = dst_sa;  /* XXX */
 763                                 origifp = ifp;
 764                                 goto nonh6lookup;
 765                         }
 766                 }
 767
 768                 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
 769                     m->m_pkthdr.flowid);
 770                 if (nh == NULL) {
 771                         IP6STAT_INC(ip6s_noroute);
 772                         /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
 773                         error = EHOSTUNREACH;
 774                         goto bad;
 775                 }
 776
 777                 ifp = nh->nh_ifp;
 778                 origifp = nh->nh_aifp;
 779                 ia = ifatoia6(nh->nh_ifa);
 780                 if (nh->nh_flags & NHF_GATEWAY)
 781                         dst->sin6_addr = nh->gw6_sa.sin6_addr;
 782                 else if (fwd_tag != NULL)
 783                         dst->sin6_addr = dst_sa.sin6_addr;
 784 nonh6lookup:
 785                 ;
 786         }
 787         /*
 788          * At this point ifp MUST be pointing to the valid transmit ifp.
 789          * origifp MUST be valid and pointing to either the same ifp or,
 790          * in case of loopback output, to the interface which ip6_src
 791          * belongs to.
 792          * Examples:
 793          *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
 794          *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
 795          *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
 796          *
 797          * mtu can be 0 and will be refined later.
 798          */
 799         KASSERT((ifp != NULL), ("output interface must not be NULL"));
 800         KASSERT((origifp != NULL), ("output address interface must not be NULL"));
 801
 802         if ((flags & IPV6_FORWARDING) == 0) {
 803                 /* XXX: the FORWARDING flag can be set for mrouting. */
 804                 in6_ifstat_inc(ifp, ifs6_out_request);
 805         }
 806
 807         /* Setup data structures for scope ID checks. */
 808         src0 = ip6->ip6_src;
 809         bzero(&src_sa, sizeof(src_sa));
 810         src_sa.sin6_family = AF_INET6;
 811         src_sa.sin6_len = sizeof(src_sa);
 812         src_sa.sin6_addr = ip6->ip6_src;
 813
 814         dst0 = ip6->ip6_dst;
 815         /* Re-initialize to be sure. */
 816         bzero(&dst_sa, sizeof(dst_sa));
 817         dst_sa.sin6_family = AF_INET6;
 818         dst_sa.sin6_len = sizeof(dst_sa);
 819         dst_sa.sin6_addr = ip6->ip6_dst;
 820
 821         /* Check for valid scope ID. */
 822         if (in6_setscope(&src0, origifp, &zone) == 0 &&
 823             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 824             in6_setscope(&dst0, origifp, &zone) == 0 &&
 825             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 826                 /*
 827                  * The outgoing interface is in the zone of the source
 828                  * and destination addresses.
 829                  *
 830                  */
 831         } else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
 832             sa6_recoverscope(&src_sa) != 0 ||
 833             sa6_recoverscope(&dst_sa) != 0 ||
 834             dst_sa.sin6_scope_id == 0 ||
 835             (src_sa.sin6_scope_id != 0 &&
 836             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 837             ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
 838                 /*
 839                  * If the destination network interface is not a
 840                  * loopback interface, or the destination network
 841                  * address has no scope ID, or the source address has
 842                  * a scope ID set which is different from the
 843                  * destination address one, or there is no network
 844                  * interface representing this scope ID, the address
 845                  * pair is considered invalid.
 846                  */
 847                 IP6STAT_INC(ip6s_badscope);
 848                 in6_ifstat_inc(origifp, ifs6_out_discard);
 849                 if (error == 0)
 850                         error = EHOSTUNREACH; /* XXX */
 851                 goto bad;
 852         }
 853         /* All scope ID checks are successful. */
 854
 855         if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 856                 if (opt && opt->ip6po_nextroute.ro_nh) {
 857                         /*
 858                          * The nexthop is explicitly specified by the
 859                          * application.  We assume the next hop is an IPv6
 860                          * address.
 861                          */
 862                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 863                 }
 864                 else if ((nh->nh_flags & NHF_GATEWAY))
 865                         dst = &nh->gw6_sa;
 866         }
 867
 868         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 869                 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
 870         } else {
 871                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 872                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 873
 874                 /* Confirm that the outgoing interface supports multicast. */
 875                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 876                         IP6STAT_INC(ip6s_noroute);
 877                         in6_ifstat_inc(ifp, ifs6_out_discard);
 878                         error = ENETUNREACH;
 879                         goto bad;
 880                 }
 881                 if ((im6o == NULL && in6_mcast_loop) ||
 882                     (im6o && im6o->im6o_multicast_loop)) {
 883                         /*
 884                          * Loop back multicast datagram if not expressly
 885                          * forbidden to do so, even if we have not joined
 886                          * the address; protocols will filter it later,
 887                          * thus deferring a hash lookup and lock acquisition
 888                          * at the expense of an m_copym().
 889                          */
 890                         ip6_mloopback(ifp, m);
 891                 } else {
 892                         /*
 893                          * If we are acting as a multicast router, perform
 894                          * multicast forwarding as if the packet had just
 895                          * arrived on the interface to which we are about
 896                          * to send.  The multicast forwarding function
 897                          * recursively calls this function, using the
 898                          * IPV6_FORWARDING flag to prevent infinite recursion.
 899                          *
 900                          * Multicasts that are looped back by ip6_mloopback(),
 901                          * above, will be forwarded by the ip6_input() routine,
 902                          * if necessary.
 903                          */
 904                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 905                                 /*
 906                                  * XXX: ip6_mforward expects that rcvif is NULL
 907                                  * when it is called from the originating path.
 908                                  * However, it may not always be the case.
 909                                  */
 910                                 m->m_pkthdr.rcvif = NULL;
 911                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 912                                         m_freem(m);
 913                                         goto done;
 914                                 }
 915                         }
 916                 }
 917                 /*
 918                  * Multicasts with a hoplimit of zero may be looped back,
 919                  * above, but must not be transmitted on a network.
 920                  * Also, multicasts addressed to the loopback interface
 921                  * are not sent -- the above call to ip6_mloopback() will
 922                  * loop back a copy if this host actually belongs to the
 923                  * destination group on the loopback interface.
 924                  */
 925                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 926                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 927                         m_freem(m);
 928                         goto done;
 929                 }
 930         }
 931
 932         /*
 933          * Fill the outgoing inteface to tell the upper layer
 934          * to increment per-interface statistics.
 935          */
 936         if (ifpp)
 937                 *ifpp = ifp;
 938
 939         /* Determine path MTU. */
 940         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 941                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 942                 goto bad;
 943         KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
 944             "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
 945             ifp, alwaysfrag, fibnum));
 946
 947         /*
 948          * The caller of this function may specify to use the minimum MTU
 949          * in some cases.
 950          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 951          * setting.  The logic is a bit complicated; by default, unicast
 952          * packets will follow path MTU while multicast packets will be sent at
 953          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 954          * including unicast ones will be sent at the minimum MTU.  Multicast
 955          * packets will always be sent at the minimum MTU unless
 956          * IP6PO_MINMTU_DISABLE is explicitly specified.
 957          * See RFC 3542 for more details.
 958          */
 959         if (mtu > IPV6_MMTU) {
 960                 if ((flags & IPV6_MINMTU))
 961                         mtu = IPV6_MMTU;
 962                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 963                         mtu = IPV6_MMTU;
 964                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 965                          (opt == NULL ||
 966                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 967                         mtu = IPV6_MMTU;
 968                 }
 969         }
 970
 971         /*
 972          * Clear embedded scope identifiers if necessary.
 973          * in6_clearscope() will touch the addresses only when necessary.
 974          */
 975         in6_clearscope(&ip6->ip6_src);
 976         in6_clearscope(&ip6->ip6_dst);
 977
 978         /*
 979          * If the outgoing packet contains a hop-by-hop options header,
 980          * it must be examined and processed even by the source node.
 981          * (RFC 2460, section 4.)
 982          */
 983         if (exthdrs.ip6e_hbh) {
 984                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 985                 u_int32_t dummy; /* XXX unused */
 986                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 987
 988 #ifdef DIAGNOSTIC
 989                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 990                         panic("ip6e_hbh is not contiguous");
 991 #endif
 992                 /*
 993                  *  XXX: if we have to send an ICMPv6 error to the sender,
 994                  *       we need the M_LOOP flag since icmp6_error() expects
 995                  *       the IPv6 and the hop-by-hop options header are
 996                  *       contiguous unless the flag is set.
 997                  */
 998                 m->m_flags |= M_LOOP;
 999                 m->m_pkthdr.rcvif = ifp;
1000                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1001                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1002                     &dummy, &plen) < 0) {
1003                         /* m was already freed at this point. */
1004                         error = EINVAL;/* better error? */
1005                         goto done;
1006                 }
1007                 m->m_flags &= ~M_LOOP; /* XXX */
1008                 m->m_pkthdr.rcvif = NULL;
1009         }
1010
1011         /* Jump over all PFIL processing if hooks are not active. */
1012         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1013                 goto passout;
1014
1015         odst = ip6->ip6_dst;
1016         /* Run through list of hooks for output packets. */
1017         switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1018         case PFIL_PASS:
1019                 ip6 = mtod(m, struct ip6_hdr *);
1020                 break;
1021         case PFIL_DROPPED:
1022                 error = EACCES;
1023                 /* FALLTHROUGH */
1024         case PFIL_CONSUMED:
1025                 goto done;
1026         }
1027
1028         needfiblookup = 0;
1029         /* See if destination IP address was changed by packet filter. */
1030         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1031                 m->m_flags |= M_SKIP_FIREWALL;
1032                 /* If destination is now ourself drop to ip6_input(). */
1033                 if (in6_localip(&ip6->ip6_dst)) {
1034                         m->m_flags |= M_FASTFWD_OURS;
1035                         if (m->m_pkthdr.rcvif == NULL)
1036                                 m->m_pkthdr.rcvif = V_loif;
1037                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1038                                 m->m_pkthdr.csum_flags |=
1039                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1040                                 m->m_pkthdr.csum_data = 0xffff;
1041                         }
1042 #if defined(SCTP) || defined(SCTP_SUPPORT)
1043                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1044                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1045 #endif
1046                         error = netisr_queue(NETISR_IPV6, m);
1047                         goto done;
1048                 } else {
1049                         if (ro != NULL)
1050                                 RO_INVALIDATE_CACHE(ro);
1051                         needfiblookup = 1; /* Redo the routing table lookup. */
1052                 }
1053         }
1054         /* See if fib was changed by packet filter. */
1055         if (fibnum != M_GETFIB(m)) {
1056                 m->m_flags |= M_SKIP_FIREWALL;
1057                 fibnum = M_GETFIB(m);
1058                 if (ro != NULL)
1059                         RO_INVALIDATE_CACHE(ro);
1060                 needfiblookup = 1;
1061         }
1062         if (needfiblookup)
1063                 goto again;
1064
1065         /* See if local, if yes, send it to netisr. */
1066         if (m->m_flags & M_FASTFWD_OURS) {
1067                 if (m->m_pkthdr.rcvif == NULL)
1068                         m->m_pkthdr.rcvif = V_loif;
1069                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1070                         m->m_pkthdr.csum_flags |=
1071                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1072                         m->m_pkthdr.csum_data = 0xffff;
1073                 }
1074 #if defined(SCTP) || defined(SCTP_SUPPORT)
1075                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1076                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1077 #endif
1078                 error = netisr_queue(NETISR_IPV6, m);
1079                 goto done;
1080         }
1081         /* Or forward to some other address? */
1082         if ((m->m_flags & M_IP6_NEXTHOP) &&
1083             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1084                 if (ro != NULL)
1085                         dst = (struct sockaddr_in6 *)&ro->ro_dst;
1086                 else
1087                         dst = &sin6;
1088                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1089                 m->m_flags |= M_SKIP_FIREWALL;
1090                 m->m_flags &= ~M_IP6_NEXTHOP;
1091                 m_tag_delete(m, fwd_tag);
1092                 goto again;
1093         }
1094
1095 passout:
1096         if (vlan_pcp > -1)
1097                 EVL_APPLY_PRI(m, vlan_pcp);
1098
1099         /* Ensure the packet data is mapped if the interface requires it. */
1100         if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1101                 m = mb_unmapped_to_ext(m);
1102                 if (m == NULL) {
1103                         IP6STAT_INC(ip6s_odropped);
1104                         return (ENOBUFS);
1105                 }
1106         }
1107
1108         /*
1109          * Send the packet to the outgoing interface.
1110          * If necessary, do IPv6 fragmentation before sending.
1111          *
1112          * The logic here is rather complex:
1113          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1114          * 1-a: send as is if tlen <= path mtu
1115          * 1-b: fragment if tlen > path mtu
1116          *
1117          * 2: if user asks us not to fragment (dontfrag == 1)
1118          * 2-a: send as is if tlen <= interface mtu
1119          * 2-b: error if tlen > interface mtu
1120          *
1121          * 3: if we always need to attach fragment header (alwaysfrag == 1)
1122          *      always fragment
1123          *
1124          * 4: if dontfrag == 1 && alwaysfrag == 1
1125          *      error, as we cannot handle this conflicting request.
1126          */
1127         sw_csum = m->m_pkthdr.csum_flags;
1128         if (!hdrsplit) {
1129                 tso = ((sw_csum & ifp->if_hwassist &
1130                     (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1131                 sw_csum &= ~ifp->if_hwassist;
1132         } else
1133                 tso = 0;
1134         /*
1135          * If we added extension headers, we will not do TSO and calculate the
1136          * checksums ourselves for now.
1137          * XXX-BZ  Need a framework to know when the NIC can handle it, even
1138          * with ext. hdrs.
1139          */
1140         ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1141         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1142         tlen = m->m_pkthdr.len;
1143
1144         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1145                 dontfrag = 1;
1146         else
1147                 dontfrag = 0;
1148         if (dontfrag && alwaysfrag) {   /* Case 4. */
1149                 /* Conflicting request - can't transmit. */
1150                 error = EMSGSIZE;
1151                 goto bad;
1152         }
1153         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* Case 2-b. */
1154                 /*
1155                  * Even if the DONTFRAG option is specified, we cannot send the
1156                  * packet when the data length is larger than the MTU of the
1157                  * outgoing interface.
1158                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1159                  * application wanted to know the MTU value. Also return an
1160                  * error code (this is not described in the API spec).
1161                  */
1162                 if (inp != NULL)
1163                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1164                 error = EMSGSIZE;
1165                 goto bad;
1166         }
1167
1168         /* Transmit packet without fragmentation. */
1169         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1170                 struct in6_ifaddr *ia6;
1171
1172                 ip6 = mtod(m, struct ip6_hdr *);
1173                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1174                 if (ia6) {
1175                         /* Record statistics for this interface address. */
1176                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1177                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1178                             m->m_pkthdr.len);
1179                 }
1180                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1181                     (flags & IP_NO_SND_TAG_RL) ? false : true);
1182                 goto done;
1183         }
1184
1185         /* Try to fragment the packet.  Cases 1-b and 3. */
1186         if (mtu < IPV6_MMTU) {
1187                 /* Path MTU cannot be less than IPV6_MMTU. */
1188                 error = EMSGSIZE;
1189                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1190                 goto bad;
1191         } else if (ip6->ip6_plen == 0) {
1192                 /* Jumbo payload cannot be fragmented. */
1193                 error = EMSGSIZE;
1194                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1195                 goto bad;
1196         } else {
1197                 u_char nextproto;
1198
1199                 /*
1200                  * Too large for the destination or interface;
1201                  * fragment if possible.
1202                  * Must be able to put at least 8 bytes per fragment.
1203                  */
1204                 if (mtu > IPV6_MAXPACKET)
1205                         mtu = IPV6_MAXPACKET;
1206
1207                 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1208                 if (len < 8) {
1209                         error = EMSGSIZE;
1210                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1211                         goto bad;
1212                 }
1213
1214                 /*
1215                  * If the interface will not calculate checksums on
1216                  * fragmented packets, then do it here.
1217                  * XXX-BZ handle the hw offloading case.  Need flags.
1218                  */
1219                 ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1220                     optlen);
1221
1222                 /*
1223                  * Change the next header field of the last header in the
1224                  * unfragmentable part.
1225                  */
1226                 if (exthdrs.ip6e_rthdr) {
1227                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1228                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1229                 } else if (exthdrs.ip6e_dest1) {
1230                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1231                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1232                 } else if (exthdrs.ip6e_hbh) {
1233                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1234                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1235                 } else {
1236                         ip6 = mtod(m, struct ip6_hdr *);
1237                         nextproto = ip6->ip6_nxt;
1238                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1239                 }
1240
1241                 /*
1242                  * Loop through length of segment after first fragment,
1243                  * make new header and copy data of each part and link onto
1244                  * chain.
1245                  */
1246                 m0 = m;
1247                 id = htonl(ip6_randomid());
1248                 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1249                 if (error != 0)
1250                         goto sendorfree;
1251
1252                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1253         }
1254
1255         /* Remove leading garbage. */
1256 sendorfree:
1257         m = m0->m_nextpkt;
1258         m0->m_nextpkt = 0;
1259         m_freem(m0);
1260         for (; m; m = m0) {
1261                 m0 = m->m_nextpkt;
1262                 m->m_nextpkt = 0;
1263                 if (error == 0) {
1264                         /* Record statistics for this interface address. */
1265                         if (ia) {
1266                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1267                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1268                                     m->m_pkthdr.len);
1269                         }
1270                         if (vlan_pcp > -1)
1271                                 EVL_APPLY_PRI(m, vlan_pcp);
1272                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1273                             true);
1274                 } else
1275                         m_freem(m);
1276         }
1277
1278         if (error == 0)
1279                 IP6STAT_INC(ip6s_fragmented);
1280
1281 done:
1282         return (error);
1283
1284 freehdrs:
1285         m_freem(exthdrs.ip6e_hbh);      /* m_freem() checks if mbuf is NULL. */
1286         m_freem(exthdrs.ip6e_dest1);
1287         m_freem(exthdrs.ip6e_rthdr);
1288         m_freem(exthdrs.ip6e_dest2);
1289         /* FALLTHROUGH */
1290 bad:
1291         if (m)
1292                 m_freem(m);
1293         goto done;
1294 }
1295
1296 static int
1297 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1298 {
1299         struct mbuf *m;
1300
1301         if (hlen > MCLBYTES)
1302                 return (ENOBUFS); /* XXX */
1303
1304         if (hlen > MLEN)
1305                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1306         else
1307                 m = m_get(M_NOWAIT, MT_DATA);
1308         if (m == NULL)
1309                 return (ENOBUFS);
1310         m->m_len = hlen;
1311         if (hdr)
1312                 bcopy(hdr, mtod(m, caddr_t), hlen);
1313
1314         *mp = m;
1315         return (0);
1316 }
1317
1318 /*
1319  * Insert jumbo payload option.
1320  */
1321 static int
1322 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1323 {
1324         struct mbuf *mopt;
1325         u_char *optbuf;
1326         u_int32_t v;
1327
1328 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1329
1330         /*
1331          * If there is no hop-by-hop options header, allocate new one.
1332          * If there is one but it doesn't have enough space to store the
1333          * jumbo payload option, allocate a cluster to store the whole options.
1334          * Otherwise, use it to store the options.
1335          */
1336         if (exthdrs->ip6e_hbh == NULL) {
1337                 mopt = m_get(M_NOWAIT, MT_DATA);
1338                 if (mopt == NULL)
1339                         return (ENOBUFS);
1340                 mopt->m_len = JUMBOOPTLEN;
1341                 optbuf = mtod(mopt, u_char *);
1342                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1343                 exthdrs->ip6e_hbh = mopt;
1344         } else {
1345                 struct ip6_hbh *hbh;
1346
1347                 mopt = exthdrs->ip6e_hbh;
1348                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1349                         /*
1350                          * XXX assumption:
1351                          * - exthdrs->ip6e_hbh is not referenced from places
1352                          *   other than exthdrs.
1353                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1354                          */
1355                         int oldoptlen = mopt->m_len;
1356                         struct mbuf *n;
1357
1358                         /*
1359                          * XXX: give up if the whole (new) hbh header does
1360                          * not fit even in an mbuf cluster.
1361                          */
1362                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1363                                 return (ENOBUFS);
1364
1365                         /*
1366                          * As a consequence, we must always prepare a cluster
1367                          * at this point.
1368                          */
1369                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1370                         if (n == NULL)
1371                                 return (ENOBUFS);
1372                         n->m_len = oldoptlen + JUMBOOPTLEN;
1373                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1374                             oldoptlen);
1375                         optbuf = mtod(n, caddr_t) + oldoptlen;
1376                         m_freem(mopt);
1377                         mopt = exthdrs->ip6e_hbh = n;
1378                 } else {
1379                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1380                         mopt->m_len += JUMBOOPTLEN;
1381                 }
1382                 optbuf[0] = IP6OPT_PADN;
1383                 optbuf[1] = 1;
1384
1385                 /*
1386                  * Adjust the header length according to the pad and
1387                  * the jumbo payload option.
1388                  */
1389                 hbh = mtod(mopt, struct ip6_hbh *);
1390                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1391         }
1392
1393         /* fill in the option. */
1394         optbuf[2] = IP6OPT_JUMBO;
1395         optbuf[3] = 4;
1396         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1397         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1398
1399         /* finally, adjust the packet header length */
1400         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1401
1402         return (0);
1403 #undef JUMBOOPTLEN
1404 }
1405
1406 /*
1407  * Insert fragment header and copy unfragmentable header portions.
1408  */
1409 static int
1410 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1411     struct ip6_frag **frghdrp)
1412 {
1413         struct mbuf *n, *mlast;
1414
1415         if (hlen > sizeof(struct ip6_hdr)) {
1416                 n = m_copym(m0, sizeof(struct ip6_hdr),
1417                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1418                 if (n == NULL)
1419                         return (ENOBUFS);
1420                 m->m_next = n;
1421         } else
1422                 n = m;
1423
1424         /* Search for the last mbuf of unfragmentable part. */
1425         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1426                 ;
1427
1428         if (M_WRITABLE(mlast) &&
1429             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1430                 /* use the trailing space of the last mbuf for the fragment hdr */
1431                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1432                     mlast->m_len);
1433                 mlast->m_len += sizeof(struct ip6_frag);
1434                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1435         } else {
1436                 /* allocate a new mbuf for the fragment header */
1437                 struct mbuf *mfrg;
1438
1439                 mfrg = m_get(M_NOWAIT, MT_DATA);
1440                 if (mfrg == NULL)
1441                         return (ENOBUFS);
1442                 mfrg->m_len = sizeof(struct ip6_frag);
1443                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1444                 mlast->m_next = mfrg;
1445         }
1446
1447         return (0);
1448 }
1449
1450 /*
1451  * Calculates IPv6 path mtu for destination @dst.
1452  * Resulting MTU is stored in @mtup.
1453  *
1454  * Returns 0 on success.
1455  */
1456 static int
1457 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1458 {
1459         struct epoch_tracker et;
1460         struct nhop_object *nh;
1461         struct in6_addr kdst;
1462         uint32_t scopeid;
1463         int error;
1464
1465         in6_splitscope(dst, &kdst, &scopeid);
1466
1467         NET_EPOCH_ENTER(et);
1468         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1469         if (nh != NULL)
1470                 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1471         else
1472                 error = EHOSTUNREACH;
1473         NET_EPOCH_EXIT(et);
1474
1475         return (error);
1476 }
1477
1478 /*
1479  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1480  * and cached data in @ro_pmtu.
1481  * MTU from (successful) route lookup is saved (along with dst)
1482  * inside @ro_pmtu to avoid subsequent route lookups after packet
1483  * filter processing.
1484  *
1485  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1486  * Returns 0 on success.
1487  */
1488 static int
1489 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1490     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1491     int *alwaysfragp, u_int fibnum, u_int proto)
1492 {
1493         struct nhop_object *nh;
1494         struct in6_addr kdst;
1495         uint32_t scopeid;
1496         struct sockaddr_in6 *sa6_dst, sin6;
1497         u_long mtu;
1498
1499         NET_EPOCH_ASSERT();
1500
1501         mtu = 0;
1502         if (ro_pmtu == NULL || do_lookup) {
1503                 /*
1504                  * Here ro_pmtu has final destination address, while
1505                  * ro might represent immediate destination.
1506                  * Use ro_pmtu destination since mtu might differ.
1507                  */
1508                 if (ro_pmtu != NULL) {
1509                         sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1510                         if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1511                                 ro_pmtu->ro_mtu = 0;
1512                 } else
1513                         sa6_dst = &sin6;
1514
1515                 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1516                         bzero(sa6_dst, sizeof(*sa6_dst));
1517                         sa6_dst->sin6_family = AF_INET6;
1518                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1519                         sa6_dst->sin6_addr = *dst;
1520
1521                         in6_splitscope(dst, &kdst, &scopeid);
1522                         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1523                         if (nh != NULL) {
1524                                 mtu = nh->nh_mtu;
1525                                 if (ro_pmtu != NULL)
1526                                         ro_pmtu->ro_mtu = mtu;
1527                         }
1528                 } else
1529                         mtu = ro_pmtu->ro_mtu;
1530         }
1531
1532         if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1533                 mtu = ro_pmtu->ro_nh->nh_mtu;
1534
1535         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1536 }
1537
1538 /*
1539  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1540  * hostcache data for @dst.
1541  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1542  *
1543  * Returns 0 on success.
1544  */
1545 static int
1546 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1547     u_long *mtup, int *alwaysfragp, u_int proto)
1548 {
1549         u_long mtu = 0;
1550         int alwaysfrag = 0;
1551         int error = 0;
1552
1553         if (rt_mtu > 0) {
1554                 u_int32_t ifmtu;
1555                 struct in_conninfo inc;
1556
1557                 bzero(&inc, sizeof(inc));
1558                 inc.inc_flags |= INC_ISIPV6;
1559                 inc.inc6_faddr = *dst;
1560
1561                 ifmtu = IN6_LINKMTU(ifp);
1562
1563                 /* TCP is known to react to pmtu changes so skip hc */
1564                 if (proto != IPPROTO_TCP)
1565                         mtu = tcp_hc_getmtu(&inc);
1566
1567                 if (mtu)
1568                         mtu = min(mtu, rt_mtu);
1569                 else
1570                         mtu = rt_mtu;
1571                 if (mtu == 0)
1572                         mtu = ifmtu;
1573                 else if (mtu < IPV6_MMTU) {
1574                         /*
1575                          * RFC2460 section 5, last paragraph:
1576                          * if we record ICMPv6 too big message with
1577                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1578                          * or smaller, with framgent header attached.
1579                          * (fragment header is needed regardless from the
1580                          * packet size, for translators to identify packets)
1581                          */
1582                         alwaysfrag = 1;
1583                         mtu = IPV6_MMTU;
1584                 }
1585         } else if (ifp) {
1586                 mtu = IN6_LINKMTU(ifp);
1587         } else
1588                 error = EHOSTUNREACH; /* XXX */
1589
1590         *mtup = mtu;
1591         if (alwaysfragp)
1592                 *alwaysfragp = alwaysfrag;
1593         return (error);
1594 }
1595
1596 /*
1597  * IP6 socket option processing.
1598  */
1599 int
1600 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1601 {
1602         int optdatalen, uproto;
1603         void *optdata;
1604         struct inpcb *inp = sotoinpcb(so);
1605         int error, optval;
1606         int level, op, optname;
1607         int optlen;
1608         struct thread *td;
1609 #ifdef  RSS
1610         uint32_t rss_bucket;
1611         int retval;
1612 #endif
1613
1614 /*
1615  * Don't use more than a quarter of mbuf clusters.  N.B.:
1616  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1617  * on LP64 architectures, so cast to u_long to avoid undefined
1618  * behavior.  ILP32 architectures cannot have nmbclusters
1619  * large enough to overflow for other reasons.
1620  */
1621 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1622
1623         level = sopt->sopt_level;
1624         op = sopt->sopt_dir;
1625         optname = sopt->sopt_name;
1626         optlen = sopt->sopt_valsize;
1627         td = sopt->sopt_td;
1628         error = 0;
1629         optval = 0;
1630         uproto = (int)so->so_proto->pr_protocol;
1631
1632         if (level != IPPROTO_IPV6) {
1633                 error = EINVAL;
1634
1635                 if (sopt->sopt_level == SOL_SOCKET &&
1636                     sopt->sopt_dir == SOPT_SET) {
1637                         switch (sopt->sopt_name) {
1638                         case SO_REUSEADDR:
1639                                 INP_WLOCK(inp);
1640                                 if ((so->so_options & SO_REUSEADDR) != 0)
1641                                         inp->inp_flags2 |= INP_REUSEADDR;
1642                                 else
1643                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1644                                 INP_WUNLOCK(inp);
1645                                 error = 0;
1646                                 break;
1647                         case SO_REUSEPORT:
1648                                 INP_WLOCK(inp);
1649                                 if ((so->so_options & SO_REUSEPORT) != 0)
1650                                         inp->inp_flags2 |= INP_REUSEPORT;
1651                                 else
1652                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1653                                 INP_WUNLOCK(inp);
1654                                 error = 0;
1655                                 break;
1656                         case SO_REUSEPORT_LB:
1657                                 INP_WLOCK(inp);
1658                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1659                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1660                                 else
1661                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1662                                 INP_WUNLOCK(inp);
1663                                 error = 0;
1664                                 break;
1665                         case SO_SETFIB:
1666                                 INP_WLOCK(inp);
1667                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1668                                 INP_WUNLOCK(inp);
1669                                 error = 0;
1670                                 break;
1671                         case SO_MAX_PACING_RATE:
1672 #ifdef RATELIMIT
1673                                 INP_WLOCK(inp);
1674                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1675                                 INP_WUNLOCK(inp);
1676                                 error = 0;
1677 #else
1678                                 error = EOPNOTSUPP;
1679 #endif
1680                                 break;
1681                         default:
1682                                 break;
1683                         }
1684                 }
1685         } else {                /* level == IPPROTO_IPV6 */
1686                 switch (op) {
1687                 case SOPT_SET:
1688                         switch (optname) {
1689                         case IPV6_2292PKTOPTIONS:
1690 #ifdef IPV6_PKTOPTIONS
1691                         case IPV6_PKTOPTIONS:
1692 #endif
1693                         {
1694                                 struct mbuf *m;
1695
1696                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1697                                         printf("ip6_ctloutput: mbuf limit hit\n");
1698                                         error = ENOBUFS;
1699                                         break;
1700                                 }
1701
1702                                 error = soopt_getm(sopt, &m); /* XXX */
1703                                 if (error != 0)
1704                                         break;
1705                                 error = soopt_mcopyin(sopt, m); /* XXX */
1706                                 if (error != 0)
1707                                         break;
1708                                 INP_WLOCK(inp);
1709                                 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1710                                     so, sopt);
1711                                 INP_WUNLOCK(inp);
1712                                 m_freem(m); /* XXX */
1713                                 break;
1714                         }
1715
1716                         /*
1717                          * Use of some Hop-by-Hop options or some
1718                          * Destination options, might require special
1719                          * privilege.  That is, normal applications
1720                          * (without special privilege) might be forbidden
1721                          * from setting certain options in outgoing packets,
1722                          * and might never see certain options in received
1723                          * packets. [RFC 2292 Section 6]
1724                          * KAME specific note:
1725                          *  KAME prevents non-privileged users from sending or
1726                          *  receiving ANY hbh/dst options in order to avoid
1727                          *  overhead of parsing options in the kernel.
1728                          */
1729                         case IPV6_RECVHOPOPTS:
1730                         case IPV6_RECVDSTOPTS:
1731                         case IPV6_RECVRTHDRDSTOPTS:
1732                                 if (td != NULL) {
1733                                         error = priv_check(td,
1734                                             PRIV_NETINET_SETHDROPTS);
1735                                         if (error)
1736                                                 break;
1737                                 }
1738                                 /* FALLTHROUGH */
1739                         case IPV6_UNICAST_HOPS:
1740                         case IPV6_HOPLIMIT:
1741
1742                         case IPV6_RECVPKTINFO:
1743                         case IPV6_RECVHOPLIMIT:
1744                         case IPV6_RECVRTHDR:
1745                         case IPV6_RECVPATHMTU:
1746                         case IPV6_RECVTCLASS:
1747                         case IPV6_RECVFLOWID:
1748 #ifdef  RSS
1749                         case IPV6_RECVRSSBUCKETID:
1750 #endif
1751                         case IPV6_V6ONLY:
1752                         case IPV6_AUTOFLOWLABEL:
1753                         case IPV6_ORIGDSTADDR:
1754                         case IPV6_BINDANY:
1755                         case IPV6_VLAN_PCP:
1756                                 if (optname == IPV6_BINDANY && td != NULL) {
1757                                         error = priv_check(td,
1758                                             PRIV_NETINET_BINDANY);
1759                                         if (error)
1760                                                 break;
1761                                 }
1762
1763                                 if (optlen != sizeof(int)) {
1764                                         error = EINVAL;
1765                                         break;
1766                                 }
1767                                 error = sooptcopyin(sopt, &optval,
1768                                         sizeof optval, sizeof optval);
1769                                 if (error)
1770                                         break;
1771                                 switch (optname) {
1772                                 case IPV6_UNICAST_HOPS:
1773                                         if (optval < -1 || optval >= 256)
1774                                                 error = EINVAL;
1775                                         else {
1776                                                 /* -1 = kernel default */
1777                                                 inp->in6p_hops = optval;
1778                                                 if ((inp->inp_vflag &
1779                                                      INP_IPV4) != 0)
1780                                                         inp->inp_ip_ttl = optval;
1781                                         }
1782                                         break;
1783 #define OPTSET(bit) \
1784 do { \
1785         INP_WLOCK(inp); \
1786         if (optval) \
1787                 inp->inp_flags |= (bit); \
1788         else \
1789                 inp->inp_flags &= ~(bit); \
1790         INP_WUNLOCK(inp); \
1791 } while (/*CONSTCOND*/ 0)
1792 #define OPTSET2292(bit) \
1793 do { \
1794         INP_WLOCK(inp); \
1795         inp->inp_flags |= IN6P_RFC2292; \
1796         if (optval) \
1797                 inp->inp_flags |= (bit); \
1798         else \
1799                 inp->inp_flags &= ~(bit); \
1800         INP_WUNLOCK(inp); \
1801 } while (/*CONSTCOND*/ 0)
1802 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1803
1804 #define OPTSET2_N(bit, val) do {                                        \
1805         if (val)                                                        \
1806                 inp->inp_flags2 |= bit;                                 \
1807         else                                                            \
1808                 inp->inp_flags2 &= ~bit;                                \
1809 } while (0)
1810 #define OPTSET2(bit, val) do {                                          \
1811         INP_WLOCK(inp);                                                 \
1812         OPTSET2_N(bit, val);                                            \
1813         INP_WUNLOCK(inp);                                               \
1814 } while (0)
1815 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1816 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1817 do {                                                                    \
1818         INP_WLOCK(inp);                                                 \
1819         if (OPTBIT(IN6P_RFC2292)) {                                     \
1820                 error = EINVAL;                                         \
1821         } else {                                                        \
1822                 if (optval)                                             \
1823                         inp->inp_flags |= (bit);                        \
1824                 else                                                    \
1825                         inp->inp_flags &= ~(bit);                       \
1826         }                                                               \
1827         INP_WUNLOCK(inp);                                               \
1828 } while (/*CONSTCOND*/ 0)
1829
1830                                 case IPV6_RECVPKTINFO:
1831                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1832                                         break;
1833
1834                                 case IPV6_HOPLIMIT:
1835                                 {
1836                                         struct ip6_pktopts **optp;
1837
1838                                         /* cannot mix with RFC2292 */
1839                                         if (OPTBIT(IN6P_RFC2292)) {
1840                                                 error = EINVAL;
1841                                                 break;
1842                                         }
1843                                         INP_WLOCK(inp);
1844                                         if (inp->inp_flags & INP_DROPPED) {
1845                                                 INP_WUNLOCK(inp);
1846                                                 return (ECONNRESET);
1847                                         }
1848                                         optp = &inp->in6p_outputopts;
1849                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1850                                             (u_char *)&optval, sizeof(optval),
1851                                             optp, (td != NULL) ? td->td_ucred :
1852                                             NULL, uproto);
1853                                         INP_WUNLOCK(inp);
1854                                         break;
1855                                 }
1856
1857                                 case IPV6_RECVHOPLIMIT:
1858                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1859                                         break;
1860
1861                                 case IPV6_RECVHOPOPTS:
1862                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1863                                         break;
1864
1865                                 case IPV6_RECVDSTOPTS:
1866                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1867                                         break;
1868
1869                                 case IPV6_RECVRTHDRDSTOPTS:
1870                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1871                                         break;
1872
1873                                 case IPV6_RECVRTHDR:
1874                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1875                                         break;
1876
1877                                 case IPV6_RECVPATHMTU:
1878                                         /*
1879                                          * We ignore this option for TCP
1880                                          * sockets.
1881                                          * (RFC3542 leaves this case
1882                                          * unspecified.)
1883                                          */
1884                                         if (uproto != IPPROTO_TCP)
1885                                                 OPTSET(IN6P_MTU);
1886                                         break;
1887
1888                                 case IPV6_RECVFLOWID:
1889                                         OPTSET2(INP_RECVFLOWID, optval);
1890                                         break;
1891
1892 #ifdef  RSS
1893                                 case IPV6_RECVRSSBUCKETID:
1894                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1895                                         break;
1896 #endif
1897
1898                                 case IPV6_V6ONLY:
1899                                         INP_WLOCK(inp);
1900                                         if (inp->inp_lport ||
1901                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1902                                                 /*
1903                                                  * The socket is already bound.
1904                                                  */
1905                                                 INP_WUNLOCK(inp);
1906                                                 error = EINVAL;
1907                                                 break;
1908                                         }
1909                                         if (optval) {
1910                                                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1911                                                 inp->inp_vflag &= ~INP_IPV4;
1912                                         } else {
1913                                                 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1914                                                 inp->inp_vflag |= INP_IPV4;
1915                                         }
1916                                         INP_WUNLOCK(inp);
1917                                         break;
1918                                 case IPV6_RECVTCLASS:
1919                                         /* cannot mix with RFC2292 XXX */
1920                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1921                                         break;
1922                                 case IPV6_AUTOFLOWLABEL:
1923                                         OPTSET(IN6P_AUTOFLOWLABEL);
1924                                         break;
1925
1926                                 case IPV6_ORIGDSTADDR:
1927                                         OPTSET2(INP_ORIGDSTADDR, optval);
1928                                         break;
1929                                 case IPV6_BINDANY:
1930                                         OPTSET(INP_BINDANY);
1931                                         break;
1932                                 case IPV6_VLAN_PCP:
1933                                         if ((optval >= -1) && (optval <=
1934                                             (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1935                                                 if (optval == -1) {
1936                                                         INP_WLOCK(inp);
1937                                                         inp->inp_flags2 &=
1938                                                             ~(INP_2PCP_SET |
1939                                                             INP_2PCP_MASK);
1940                                                         INP_WUNLOCK(inp);
1941                                                 } else {
1942                                                         INP_WLOCK(inp);
1943                                                         inp->inp_flags2 |=
1944                                                             INP_2PCP_SET;
1945                                                         inp->inp_flags2 &=
1946                                                             ~INP_2PCP_MASK;
1947                                                         inp->inp_flags2 |=
1948                                                             optval <<
1949                                                             INP_2PCP_SHIFT;
1950                                                         INP_WUNLOCK(inp);
1951                                                 }
1952                                         } else
1953                                                 error = EINVAL;
1954                                         break;
1955                                 }
1956                                 break;
1957
1958                         case IPV6_TCLASS:
1959                         case IPV6_DONTFRAG:
1960                         case IPV6_USE_MIN_MTU:
1961                         case IPV6_PREFER_TEMPADDR:
1962                                 if (optlen != sizeof(optval)) {
1963                                         error = EINVAL;
1964                                         break;
1965                                 }
1966                                 error = sooptcopyin(sopt, &optval,
1967                                         sizeof optval, sizeof optval);
1968                                 if (error)
1969                                         break;
1970                                 {
1971                                         struct ip6_pktopts **optp;
1972                                         INP_WLOCK(inp);
1973                                         if (inp->inp_flags & INP_DROPPED) {
1974                                                 INP_WUNLOCK(inp);
1975                                                 return (ECONNRESET);
1976                                         }
1977                                         optp = &inp->in6p_outputopts;
1978                                         error = ip6_pcbopt(optname,
1979                                             (u_char *)&optval, sizeof(optval),
1980                                             optp, (td != NULL) ? td->td_ucred :
1981                                             NULL, uproto);
1982                                         INP_WUNLOCK(inp);
1983                                         break;
1984                                 }
1985
1986                         case IPV6_2292PKTINFO:
1987                         case IPV6_2292HOPLIMIT:
1988                         case IPV6_2292HOPOPTS:
1989                         case IPV6_2292DSTOPTS:
1990                         case IPV6_2292RTHDR:
1991                                 /* RFC 2292 */
1992                                 if (optlen != sizeof(int)) {
1993                                         error = EINVAL;
1994                                         break;
1995                                 }
1996                                 error = sooptcopyin(sopt, &optval,
1997                                         sizeof optval, sizeof optval);
1998                                 if (error)
1999                                         break;
2000                                 switch (optname) {
2001                                 case IPV6_2292PKTINFO:
2002                                         OPTSET2292(IN6P_PKTINFO);
2003                                         break;
2004                                 case IPV6_2292HOPLIMIT:
2005                                         OPTSET2292(IN6P_HOPLIMIT);
2006                                         break;
2007                                 case IPV6_2292HOPOPTS:
2008                                         /*
2009                                          * Check super-user privilege.
2010                                          * See comments for IPV6_RECVHOPOPTS.
2011                                          */
2012                                         if (td != NULL) {
2013                                                 error = priv_check(td,
2014                                                     PRIV_NETINET_SETHDROPTS);
2015                                                 if (error)
2016                                                         return (error);
2017                                         }
2018                                         OPTSET2292(IN6P_HOPOPTS);
2019                                         break;
2020                                 case IPV6_2292DSTOPTS:
2021                                         if (td != NULL) {
2022                                                 error = priv_check(td,
2023                                                     PRIV_NETINET_SETHDROPTS);
2024                                                 if (error)
2025                                                         return (error);
2026                                         }
2027                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2028                                         break;
2029                                 case IPV6_2292RTHDR:
2030                                         OPTSET2292(IN6P_RTHDR);
2031                                         break;
2032                                 }
2033                                 break;
2034                         case IPV6_PKTINFO:
2035                         case IPV6_HOPOPTS:
2036                         case IPV6_RTHDR:
2037                         case IPV6_DSTOPTS:
2038                         case IPV6_RTHDRDSTOPTS:
2039                         case IPV6_NEXTHOP:
2040                         {
2041                                 /* new advanced API (RFC3542) */
2042                                 u_char *optbuf;
2043                                 u_char optbuf_storage[MCLBYTES];
2044                                 int optlen;
2045                                 struct ip6_pktopts **optp;
2046
2047                                 /* cannot mix with RFC2292 */
2048                                 if (OPTBIT(IN6P_RFC2292)) {
2049                                         error = EINVAL;
2050                                         break;
2051                                 }
2052
2053                                 /*
2054                                  * We only ensure valsize is not too large
2055                                  * here.  Further validation will be done
2056                                  * later.
2057                                  */
2058                                 error = sooptcopyin(sopt, optbuf_storage,
2059                                     sizeof(optbuf_storage), 0);
2060                                 if (error)
2061                                         break;
2062                                 optlen = sopt->sopt_valsize;
2063                                 optbuf = optbuf_storage;
2064                                 INP_WLOCK(inp);
2065                                 if (inp->inp_flags & INP_DROPPED) {
2066                                         INP_WUNLOCK(inp);
2067                                         return (ECONNRESET);
2068                                 }
2069                                 optp = &inp->in6p_outputopts;
2070                                 error = ip6_pcbopt(optname, optbuf, optlen,
2071                                     optp, (td != NULL) ? td->td_ucred : NULL,
2072                                     uproto);
2073                                 INP_WUNLOCK(inp);
2074                                 break;
2075                         }
2076 #undef OPTSET
2077
2078                         case IPV6_MULTICAST_IF:
2079                         case IPV6_MULTICAST_HOPS:
2080                         case IPV6_MULTICAST_LOOP:
2081                         case IPV6_JOIN_GROUP:
2082                         case IPV6_LEAVE_GROUP:
2083                         case IPV6_MSFILTER:
2084                         case MCAST_BLOCK_SOURCE:
2085                         case MCAST_UNBLOCK_SOURCE:
2086                         case MCAST_JOIN_GROUP:
2087                         case MCAST_LEAVE_GROUP:
2088                         case MCAST_JOIN_SOURCE_GROUP:
2089                         case MCAST_LEAVE_SOURCE_GROUP:
2090                                 error = ip6_setmoptions(inp, sopt);
2091                                 break;
2092
2093                         case IPV6_PORTRANGE:
2094                                 error = sooptcopyin(sopt, &optval,
2095                                     sizeof optval, sizeof optval);
2096                                 if (error)
2097                                         break;
2098
2099                                 INP_WLOCK(inp);
2100                                 switch (optval) {
2101                                 case IPV6_PORTRANGE_DEFAULT:
2102                                         inp->inp_flags &= ~(INP_LOWPORT);
2103                                         inp->inp_flags &= ~(INP_HIGHPORT);
2104                                         break;
2105
2106                                 case IPV6_PORTRANGE_HIGH:
2107                                         inp->inp_flags &= ~(INP_LOWPORT);
2108                                         inp->inp_flags |= INP_HIGHPORT;
2109                                         break;
2110
2111                                 case IPV6_PORTRANGE_LOW:
2112                                         inp->inp_flags &= ~(INP_HIGHPORT);
2113                                         inp->inp_flags |= INP_LOWPORT;
2114                                         break;
2115
2116                                 default:
2117                                         error = EINVAL;
2118                                         break;
2119                                 }
2120                                 INP_WUNLOCK(inp);
2121                                 break;
2122
2123 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2124                         case IPV6_IPSEC_POLICY:
2125                                 if (IPSEC_ENABLED(ipv6)) {
2126                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2127                                         break;
2128                                 }
2129                                 /* FALLTHROUGH */
2130 #endif /* IPSEC */
2131
2132                         default:
2133                                 error = ENOPROTOOPT;
2134                                 break;
2135                         }
2136                         break;
2137
2138                 case SOPT_GET:
2139                         switch (optname) {
2140                         case IPV6_2292PKTOPTIONS:
2141 #ifdef IPV6_PKTOPTIONS
2142                         case IPV6_PKTOPTIONS:
2143 #endif
2144                                 /*
2145                                  * RFC3542 (effectively) deprecated the
2146                                  * semantics of the 2292-style pktoptions.
2147                                  * Since it was not reliable in nature (i.e.,
2148                                  * applications had to expect the lack of some
2149                                  * information after all), it would make sense
2150                                  * to simplify this part by always returning
2151                                  * empty data.
2152                                  */
2153                                 sopt->sopt_valsize = 0;
2154                                 break;
2155
2156                         case IPV6_RECVHOPOPTS:
2157                         case IPV6_RECVDSTOPTS:
2158                         case IPV6_RECVRTHDRDSTOPTS:
2159                         case IPV6_UNICAST_HOPS:
2160                         case IPV6_RECVPKTINFO:
2161                         case IPV6_RECVHOPLIMIT:
2162                         case IPV6_RECVRTHDR:
2163                         case IPV6_RECVPATHMTU:
2164
2165                         case IPV6_V6ONLY:
2166                         case IPV6_PORTRANGE:
2167                         case IPV6_RECVTCLASS:
2168                         case IPV6_AUTOFLOWLABEL:
2169                         case IPV6_BINDANY:
2170                         case IPV6_FLOWID:
2171                         case IPV6_FLOWTYPE:
2172                         case IPV6_RECVFLOWID:
2173 #ifdef  RSS
2174                         case IPV6_RSSBUCKETID:
2175                         case IPV6_RECVRSSBUCKETID:
2176 #endif
2177                         case IPV6_VLAN_PCP:
2178                                 switch (optname) {
2179                                 case IPV6_RECVHOPOPTS:
2180                                         optval = OPTBIT(IN6P_HOPOPTS);
2181                                         break;
2182
2183                                 case IPV6_RECVDSTOPTS:
2184                                         optval = OPTBIT(IN6P_DSTOPTS);
2185                                         break;
2186
2187                                 case IPV6_RECVRTHDRDSTOPTS:
2188                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2189                                         break;
2190
2191                                 case IPV6_UNICAST_HOPS:
2192                                         optval = inp->in6p_hops;
2193                                         break;
2194
2195                                 case IPV6_RECVPKTINFO:
2196                                         optval = OPTBIT(IN6P_PKTINFO);
2197                                         break;
2198
2199                                 case IPV6_RECVHOPLIMIT:
2200                                         optval = OPTBIT(IN6P_HOPLIMIT);
2201                                         break;
2202
2203                                 case IPV6_RECVRTHDR:
2204                                         optval = OPTBIT(IN6P_RTHDR);
2205                                         break;
2206
2207                                 case IPV6_RECVPATHMTU:
2208                                         optval = OPTBIT(IN6P_MTU);
2209                                         break;
2210
2211                                 case IPV6_V6ONLY:
2212                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2213                                         break;
2214
2215                                 case IPV6_PORTRANGE:
2216                                     {
2217                                         int flags;
2218                                         flags = inp->inp_flags;
2219                                         if (flags & INP_HIGHPORT)
2220                                                 optval = IPV6_PORTRANGE_HIGH;
2221                                         else if (flags & INP_LOWPORT)
2222                                                 optval = IPV6_PORTRANGE_LOW;
2223                                         else
2224                                                 optval = 0;
2225                                         break;
2226                                     }
2227                                 case IPV6_RECVTCLASS:
2228                                         optval = OPTBIT(IN6P_TCLASS);
2229                                         break;
2230
2231                                 case IPV6_AUTOFLOWLABEL:
2232                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2233                                         break;
2234
2235                                 case IPV6_ORIGDSTADDR:
2236                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2237                                         break;
2238
2239                                 case IPV6_BINDANY:
2240                                         optval = OPTBIT(INP_BINDANY);
2241                                         break;
2242
2243                                 case IPV6_FLOWID:
2244                                         optval = inp->inp_flowid;
2245                                         break;
2246
2247                                 case IPV6_FLOWTYPE:
2248                                         optval = inp->inp_flowtype;
2249                                         break;
2250
2251                                 case IPV6_RECVFLOWID:
2252                                         optval = OPTBIT2(INP_RECVFLOWID);
2253                                         break;
2254 #ifdef  RSS
2255                                 case IPV6_RSSBUCKETID:
2256                                         retval =
2257                                             rss_hash2bucket(inp->inp_flowid,
2258                                             inp->inp_flowtype,
2259                                             &rss_bucket);
2260                                         if (retval == 0)
2261                                                 optval = rss_bucket;
2262                                         else
2263                                                 error = EINVAL;
2264                                         break;
2265
2266                                 case IPV6_RECVRSSBUCKETID:
2267                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2268                                         break;
2269 #endif
2270
2271
2272                                 case IPV6_VLAN_PCP:
2273                                         if (OPTBIT2(INP_2PCP_SET)) {
2274                                                 optval = (inp->inp_flags2 &
2275                                                             INP_2PCP_MASK) >>
2276                                                             INP_2PCP_SHIFT;
2277                                         } else {
2278                                                 optval = -1;
2279                                         }
2280                                         break;
2281                                 }
2282
2283                                 if (error)
2284                                         break;
2285                                 error = sooptcopyout(sopt, &optval,
2286                                         sizeof optval);
2287                                 break;
2288
2289                         case IPV6_PATHMTU:
2290                         {
2291                                 u_long pmtu = 0;
2292                                 struct ip6_mtuinfo mtuinfo;
2293                                 struct in6_addr addr;
2294
2295                                 if (!(so->so_state & SS_ISCONNECTED))
2296                                         return (ENOTCONN);
2297                                 /*
2298                                  * XXX: we dot not consider the case of source
2299                                  * routing, or optional information to specify
2300                                  * the outgoing interface.
2301                                  * Copy faddr out of inp to avoid holding lock
2302                                  * on inp during route lookup.
2303                                  */
2304                                 INP_RLOCK(inp);
2305                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2306                                 INP_RUNLOCK(inp);
2307                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2308                                     &addr, &pmtu);
2309                                 if (error)
2310                                         break;
2311                                 if (pmtu > IPV6_MAXPACKET)
2312                                         pmtu = IPV6_MAXPACKET;
2313
2314                                 bzero(&mtuinfo, sizeof(mtuinfo));
2315                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2316                                 optdata = (void *)&mtuinfo;
2317                                 optdatalen = sizeof(mtuinfo);
2318                                 error = sooptcopyout(sopt, optdata,
2319                                     optdatalen);
2320                                 break;
2321                         }
2322
2323                         case IPV6_2292PKTINFO:
2324                         case IPV6_2292HOPLIMIT:
2325                         case IPV6_2292HOPOPTS:
2326                         case IPV6_2292RTHDR:
2327                         case IPV6_2292DSTOPTS:
2328                                 switch (optname) {
2329                                 case IPV6_2292PKTINFO:
2330                                         optval = OPTBIT(IN6P_PKTINFO);
2331                                         break;
2332                                 case IPV6_2292HOPLIMIT:
2333                                         optval = OPTBIT(IN6P_HOPLIMIT);
2334                                         break;
2335                                 case IPV6_2292HOPOPTS:
2336                                         optval = OPTBIT(IN6P_HOPOPTS);
2337                                         break;
2338                                 case IPV6_2292RTHDR:
2339                                         optval = OPTBIT(IN6P_RTHDR);
2340                                         break;
2341                                 case IPV6_2292DSTOPTS:
2342                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2343                                         break;
2344                                 }
2345                                 error = sooptcopyout(sopt, &optval,
2346                                     sizeof optval);
2347                                 break;
2348                         case IPV6_PKTINFO:
2349                         case IPV6_HOPOPTS:
2350                         case IPV6_RTHDR:
2351                         case IPV6_DSTOPTS:
2352                         case IPV6_RTHDRDSTOPTS:
2353                         case IPV6_NEXTHOP:
2354                         case IPV6_TCLASS:
2355                         case IPV6_DONTFRAG:
2356                         case IPV6_USE_MIN_MTU:
2357                         case IPV6_PREFER_TEMPADDR:
2358                                 error = ip6_getpcbopt(inp, optname, sopt);
2359                                 break;
2360
2361                         case IPV6_MULTICAST_IF:
2362                         case IPV6_MULTICAST_HOPS:
2363                         case IPV6_MULTICAST_LOOP:
2364                         case IPV6_MSFILTER:
2365                                 error = ip6_getmoptions(inp, sopt);
2366                                 break;
2367
2368 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2369                         case IPV6_IPSEC_POLICY:
2370                                 if (IPSEC_ENABLED(ipv6)) {
2371                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2372                                         break;
2373                                 }
2374                                 /* FALLTHROUGH */
2375 #endif /* IPSEC */
2376                         default:
2377                                 error = ENOPROTOOPT;
2378                                 break;
2379                         }
2380                         break;
2381                 }
2382         }
2383         return (error);
2384 }
2385
2386 int
2387 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2388 {
2389         int error = 0, optval, optlen;
2390         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2391         struct inpcb *inp = sotoinpcb(so);
2392         int level, op, optname;
2393
2394         level = sopt->sopt_level;
2395         op = sopt->sopt_dir;
2396         optname = sopt->sopt_name;
2397         optlen = sopt->sopt_valsize;
2398
2399         if (level != IPPROTO_IPV6) {
2400                 return (EINVAL);
2401         }
2402
2403         switch (optname) {
2404         case IPV6_CHECKSUM:
2405                 /*
2406                  * For ICMPv6 sockets, no modification allowed for checksum
2407                  * offset, permit "no change" values to help existing apps.
2408                  *
2409                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2410                  * for an ICMPv6 socket will fail."
2411                  * The current behavior does not meet RFC3542.
2412                  */
2413                 switch (op) {
2414                 case SOPT_SET:
2415                         if (optlen != sizeof(int)) {
2416                                 error = EINVAL;
2417                                 break;
2418                         }
2419                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2420                                             sizeof(optval));
2421                         if (error)
2422                                 break;
2423                         if (optval < -1 || (optval % 2) != 0) {
2424                                 /*
2425                                  * The API assumes non-negative even offset
2426                                  * values or -1 as a special value.
2427                                  */
2428                                 error = EINVAL;
2429                         } else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2430                                 if (optval != icmp6off)
2431                                         error = EINVAL;
2432                         } else
2433                                 inp->in6p_cksum = optval;
2434                         break;
2435
2436                 case SOPT_GET:
2437                         if (inp->inp_ip_p == IPPROTO_ICMPV6)
2438                                 optval = icmp6off;
2439                         else
2440                                 optval = inp->in6p_cksum;
2441
2442                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2443                         break;
2444
2445                 default:
2446                         error = EINVAL;
2447                         break;
2448                 }
2449                 break;
2450
2451         default:
2452                 error = ENOPROTOOPT;
2453                 break;
2454         }
2455
2456         return (error);
2457 }
2458
2459 /*
2460  * Set up IP6 options in pcb for insertion in output packets or
2461  * specifying behavior of outgoing packets.
2462  */
2463 static int
2464 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2465     struct socket *so, struct sockopt *sopt)
2466 {
2467         struct ip6_pktopts *opt = *pktopt;
2468         int error = 0;
2469         struct thread *td = sopt->sopt_td;
2470         struct epoch_tracker et;
2471
2472         /* turn off any old options. */
2473         if (opt) {
2474 #ifdef DIAGNOSTIC
2475                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2476                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2477                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2478                         printf("ip6_pcbopts: all specified options are cleared.\n");
2479 #endif
2480                 ip6_clearpktopts(opt, -1);
2481         } else {
2482                 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2483                 if (opt == NULL)
2484                         return (ENOMEM);
2485         }
2486         *pktopt = NULL;
2487
2488         if (!m || m->m_len == 0) {
2489                 /*
2490                  * Only turning off any previous options, regardless of
2491                  * whether the opt is just created or given.
2492                  */
2493                 free(opt, M_IP6OPT);
2494                 return (0);
2495         }
2496
2497         /*  set options specified by user. */
2498         NET_EPOCH_ENTER(et);
2499         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2500             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2501                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2502                 free(opt, M_IP6OPT);
2503                 NET_EPOCH_EXIT(et);
2504                 return (error);
2505         }
2506         NET_EPOCH_EXIT(et);
2507         *pktopt = opt;
2508         return (0);
2509 }
2510
2511 /*
2512  * initialize ip6_pktopts.  beware that there are non-zero default values in
2513  * the struct.
2514  */
2515 void
2516 ip6_initpktopts(struct ip6_pktopts *opt)
2517 {
2518
2519         bzero(opt, sizeof(*opt));
2520         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2521         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2522         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2523         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2524 }
2525
2526 static int
2527 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2528     struct ucred *cred, int uproto)
2529 {
2530         struct epoch_tracker et;
2531         struct ip6_pktopts *opt;
2532         int ret;
2533
2534         if (*pktopt == NULL) {
2535                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2536                     M_NOWAIT);
2537                 if (*pktopt == NULL)
2538                         return (ENOBUFS);
2539                 ip6_initpktopts(*pktopt);
2540         }
2541         opt = *pktopt;
2542
2543         NET_EPOCH_ENTER(et);
2544         ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2545         NET_EPOCH_EXIT(et);
2546
2547         return (ret);
2548 }
2549
2550 #define GET_PKTOPT_VAR(field, lenexpr) do {                             \
2551         if (pktopt && pktopt->field) {                                  \
2552                 INP_RUNLOCK(inp);                                       \
2553                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2554                 malloc_optdata = true;                                  \
2555                 INP_RLOCK(inp);                                         \
2556                 if (inp->inp_flags & INP_DROPPED) {                     \
2557                         INP_RUNLOCK(inp);                               \
2558                         free(optdata, M_TEMP);                          \
2559                         return (ECONNRESET);                            \
2560                 }                                                       \
2561                 pktopt = inp->in6p_outputopts;                          \
2562                 if (pktopt && pktopt->field) {                          \
2563                         optdatalen = min(lenexpr, sopt->sopt_valsize);  \
2564                         bcopy(pktopt->field, optdata, optdatalen);      \
2565                 } else {                                                \
2566                         free(optdata, M_TEMP);                          \
2567                         optdata = NULL;                                 \
2568                         malloc_optdata = false;                         \
2569                 }                                                       \
2570         }                                                               \
2571 } while(0)
2572
2573 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                 \
2574         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2575
2576 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                \
2577         pktopt->field->sa_len)
2578
2579 static int
2580 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2581 {
2582         void *optdata = NULL;
2583         bool malloc_optdata = false;
2584         int optdatalen = 0;
2585         int error = 0;
2586         struct in6_pktinfo null_pktinfo;
2587         int deftclass = 0, on;
2588         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2589         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2590         struct ip6_pktopts *pktopt;
2591
2592         INP_RLOCK(inp);
2593         pktopt = inp->in6p_outputopts;
2594
2595         switch (optname) {
2596         case IPV6_PKTINFO:
2597                 optdata = (void *)&null_pktinfo;
2598                 if (pktopt && pktopt->ip6po_pktinfo) {
2599                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2600                             sizeof(null_pktinfo));
2601                         in6_clearscope(&null_pktinfo.ipi6_addr);
2602                 } else {
2603                         /* XXX: we don't have to do this every time... */
2604                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2605                 }
2606                 optdatalen = sizeof(struct in6_pktinfo);
2607                 break;
2608         case IPV6_TCLASS:
2609                 if (pktopt && pktopt->ip6po_tclass >= 0)
2610                         deftclass = pktopt->ip6po_tclass;
2611                 optdata = (void *)&deftclass;
2612                 optdatalen = sizeof(int);
2613                 break;
2614         case IPV6_HOPOPTS:
2615                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2616                 break;
2617         case IPV6_RTHDR:
2618                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2619                 break;
2620         case IPV6_RTHDRDSTOPTS:
2621                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2622                 break;
2623         case IPV6_DSTOPTS:
2624                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2625                 break;
2626         case IPV6_NEXTHOP:
2627                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2628                 break;
2629         case IPV6_USE_MIN_MTU:
2630                 if (pktopt)
2631                         defminmtu = pktopt->ip6po_minmtu;
2632                 optdata = (void *)&defminmtu;
2633                 optdatalen = sizeof(int);
2634                 break;
2635         case IPV6_DONTFRAG:
2636                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2637                         on = 1;
2638                 else
2639                         on = 0;
2640                 optdata = (void *)&on;
2641                 optdatalen = sizeof(on);
2642                 break;
2643         case IPV6_PREFER_TEMPADDR:
2644                 if (pktopt)
2645                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2646                 optdata = (void *)&defpreftemp;
2647                 optdatalen = sizeof(int);
2648                 break;
2649         default:                /* should not happen */
2650 #ifdef DIAGNOSTIC
2651                 panic("ip6_getpcbopt: unexpected option\n");
2652 #endif
2653                 INP_RUNLOCK(inp);
2654                 return (ENOPROTOOPT);
2655         }
2656         INP_RUNLOCK(inp);
2657
2658         error = sooptcopyout(sopt, optdata, optdatalen);
2659         if (malloc_optdata)
2660                 free(optdata, M_TEMP);
2661
2662         return (error);
2663 }
2664
2665 void
2666 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2667 {
2668         if (pktopt == NULL)
2669                 return;
2670
2671         if (optname == -1 || optname == IPV6_PKTINFO) {
2672                 if (pktopt->ip6po_pktinfo)
2673                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2674                 pktopt->ip6po_pktinfo = NULL;
2675         }
2676         if (optname == -1 || optname == IPV6_HOPLIMIT)
2677                 pktopt->ip6po_hlim = -1;
2678         if (optname == -1 || optname == IPV6_TCLASS)
2679                 pktopt->ip6po_tclass = -1;
2680         if (optname == -1 || optname == IPV6_NEXTHOP) {
2681                 if (pktopt->ip6po_nextroute.ro_nh) {
2682                         NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2683                         pktopt->ip6po_nextroute.ro_nh = NULL;
2684                 }
2685                 if (pktopt->ip6po_nexthop)
2686                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2687                 pktopt->ip6po_nexthop = NULL;
2688         }
2689         if (optname == -1 || optname == IPV6_HOPOPTS) {
2690                 if (pktopt->ip6po_hbh)
2691                         free(pktopt->ip6po_hbh, M_IP6OPT);
2692                 pktopt->ip6po_hbh = NULL;
2693         }
2694         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2695                 if (pktopt->ip6po_dest1)
2696                         free(pktopt->ip6po_dest1, M_IP6OPT);
2697                 pktopt->ip6po_dest1 = NULL;
2698         }
2699         if (optname == -1 || optname == IPV6_RTHDR) {
2700                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2701                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2702                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2703                 if (pktopt->ip6po_route.ro_nh) {
2704                         NH_FREE(pktopt->ip6po_route.ro_nh);
2705                         pktopt->ip6po_route.ro_nh = NULL;
2706                 }
2707         }
2708         if (optname == -1 || optname == IPV6_DSTOPTS) {
2709                 if (pktopt->ip6po_dest2)
2710                         free(pktopt->ip6po_dest2, M_IP6OPT);
2711                 pktopt->ip6po_dest2 = NULL;
2712         }
2713 }
2714
2715 #define PKTOPT_EXTHDRCPY(type) \
2716 do {\
2717         if (src->type) {\
2718                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2719                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2720                 if (dst->type == NULL)\
2721                         goto bad;\
2722                 bcopy(src->type, dst->type, hlen);\
2723         }\
2724 } while (/*CONSTCOND*/ 0)
2725
2726 static int
2727 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2728 {
2729         if (dst == NULL || src == NULL)  {
2730                 printf("ip6_clearpktopts: invalid argument\n");
2731                 return (EINVAL);
2732         }
2733
2734         dst->ip6po_hlim = src->ip6po_hlim;
2735         dst->ip6po_tclass = src->ip6po_tclass;
2736         dst->ip6po_flags = src->ip6po_flags;
2737         dst->ip6po_minmtu = src->ip6po_minmtu;
2738         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2739         if (src->ip6po_pktinfo) {
2740                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2741                     M_IP6OPT, canwait);
2742                 if (dst->ip6po_pktinfo == NULL)
2743                         goto bad;
2744                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2745         }
2746         if (src->ip6po_nexthop) {
2747                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2748                     M_IP6OPT, canwait);
2749                 if (dst->ip6po_nexthop == NULL)
2750                         goto bad;
2751                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2752                     src->ip6po_nexthop->sa_len);
2753         }
2754         PKTOPT_EXTHDRCPY(ip6po_hbh);
2755         PKTOPT_EXTHDRCPY(ip6po_dest1);
2756         PKTOPT_EXTHDRCPY(ip6po_dest2);
2757         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2758         return (0);
2759
2760   bad:
2761         ip6_clearpktopts(dst, -1);
2762         return (ENOBUFS);
2763 }
2764 #undef PKTOPT_EXTHDRCPY
2765
2766 struct ip6_pktopts *
2767 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2768 {
2769         int error;
2770         struct ip6_pktopts *dst;
2771
2772         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2773         if (dst == NULL)
2774                 return (NULL);
2775         ip6_initpktopts(dst);
2776
2777         if ((error = copypktopts(dst, src, canwait)) != 0) {
2778                 free(dst, M_IP6OPT);
2779                 return (NULL);
2780         }
2781
2782         return (dst);
2783 }
2784
2785 void
2786 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2787 {
2788         if (pktopt == NULL)
2789                 return;
2790
2791         ip6_clearpktopts(pktopt, -1);
2792
2793         free(pktopt, M_IP6OPT);
2794 }
2795
2796 /*
2797  * Set IPv6 outgoing packet options based on advanced API.
2798  */
2799 int
2800 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2801     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2802 {
2803         struct cmsghdr *cm = NULL;
2804
2805         if (control == NULL || opt == NULL)
2806                 return (EINVAL);
2807
2808         /*
2809          * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2810          * are in the network epoch here.
2811          */
2812         NET_EPOCH_ASSERT();
2813
2814         ip6_initpktopts(opt);
2815         if (stickyopt) {
2816                 int error;
2817
2818                 /*
2819                  * If stickyopt is provided, make a local copy of the options
2820                  * for this particular packet, then override them by ancillary
2821                  * objects.
2822                  * XXX: copypktopts() does not copy the cached route to a next
2823                  * hop (if any).  This is not very good in terms of efficiency,
2824                  * but we can allow this since this option should be rarely
2825                  * used.
2826                  */
2827                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2828                         return (error);
2829         }
2830
2831         /*
2832          * XXX: Currently, we assume all the optional information is stored
2833          * in a single mbuf.
2834          */
2835         if (control->m_next)
2836                 return (EINVAL);
2837
2838         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2839             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2840                 int error;
2841
2842                 if (control->m_len < CMSG_LEN(0))
2843                         return (EINVAL);
2844
2845                 cm = mtod(control, struct cmsghdr *);
2846                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2847                         return (EINVAL);
2848                 if (cm->cmsg_level != IPPROTO_IPV6)
2849                         continue;
2850
2851                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2852                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2853                 if (error)
2854                         return (error);
2855         }
2856
2857         return (0);
2858 }
2859
2860 /*
2861  * Set a particular packet option, as a sticky option or an ancillary data
2862  * item.  "len" can be 0 only when it's a sticky option.
2863  * We have 4 cases of combination of "sticky" and "cmsg":
2864  * "sticky=0, cmsg=0": impossible
2865  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2866  * "sticky=1, cmsg=0": RFC3542 socket option
2867  * "sticky=1, cmsg=1": RFC2292 socket option
2868  */
2869 static int
2870 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2871     struct ucred *cred, int sticky, int cmsg, int uproto)
2872 {
2873         int minmtupolicy, preftemp;
2874         int error;
2875
2876         NET_EPOCH_ASSERT();
2877
2878         if (!sticky && !cmsg) {
2879 #ifdef DIAGNOSTIC
2880                 printf("ip6_setpktopt: impossible case\n");
2881 #endif
2882                 return (EINVAL);
2883         }
2884
2885         /*
2886          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2887          * not be specified in the context of RFC3542.  Conversely,
2888          * RFC3542 types should not be specified in the context of RFC2292.
2889          */
2890         if (!cmsg) {
2891                 switch (optname) {
2892                 case IPV6_2292PKTINFO:
2893                 case IPV6_2292HOPLIMIT:
2894                 case IPV6_2292NEXTHOP:
2895                 case IPV6_2292HOPOPTS:
2896                 case IPV6_2292DSTOPTS:
2897                 case IPV6_2292RTHDR:
2898                 case IPV6_2292PKTOPTIONS:
2899                         return (ENOPROTOOPT);
2900                 }
2901         }
2902         if (sticky && cmsg) {
2903                 switch (optname) {
2904                 case IPV6_PKTINFO:
2905                 case IPV6_HOPLIMIT:
2906                 case IPV6_NEXTHOP:
2907                 case IPV6_HOPOPTS:
2908                 case IPV6_DSTOPTS:
2909                 case IPV6_RTHDRDSTOPTS:
2910                 case IPV6_RTHDR:
2911                 case IPV6_USE_MIN_MTU:
2912                 case IPV6_DONTFRAG:
2913                 case IPV6_TCLASS:
2914                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2915                         return (ENOPROTOOPT);
2916                 }
2917         }
2918
2919         switch (optname) {
2920         case IPV6_2292PKTINFO:
2921         case IPV6_PKTINFO:
2922         {
2923                 struct ifnet *ifp = NULL;
2924                 struct in6_pktinfo *pktinfo;
2925
2926                 if (len != sizeof(struct in6_pktinfo))
2927                         return (EINVAL);
2928
2929                 pktinfo = (struct in6_pktinfo *)buf;
2930
2931                 /*
2932                  * An application can clear any sticky IPV6_PKTINFO option by
2933                  * doing a "regular" setsockopt with ipi6_addr being
2934                  * in6addr_any and ipi6_ifindex being zero.
2935                  * [RFC 3542, Section 6]
2936                  */
2937                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2938                     pktinfo->ipi6_ifindex == 0 &&
2939                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2940                         ip6_clearpktopts(opt, optname);
2941                         break;
2942                 }
2943
2944                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2945                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2946                         return (EINVAL);
2947                 }
2948                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2949                         return (EINVAL);
2950                 /* validate the interface index if specified. */
2951                 if (pktinfo->ipi6_ifindex) {
2952                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2953                         if (ifp == NULL)
2954                                 return (ENXIO);
2955                 }
2956                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2957                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2958                         return (ENETDOWN);
2959
2960                 if (ifp != NULL &&
2961                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2962                         struct in6_ifaddr *ia;
2963
2964                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2965                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2966                         if (ia == NULL)
2967                                 return (EADDRNOTAVAIL);
2968                         ifa_free(&ia->ia_ifa);
2969                 }
2970                 /*
2971                  * We store the address anyway, and let in6_selectsrc()
2972                  * validate the specified address.  This is because ipi6_addr
2973                  * may not have enough information about its scope zone, and
2974                  * we may need additional information (such as outgoing
2975                  * interface or the scope zone of a destination address) to
2976                  * disambiguate the scope.
2977                  * XXX: the delay of the validation may confuse the
2978                  * application when it is used as a sticky option.
2979                  */
2980                 if (opt->ip6po_pktinfo == NULL) {
2981                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2982                             M_IP6OPT, M_NOWAIT);
2983                         if (opt->ip6po_pktinfo == NULL)
2984                                 return (ENOBUFS);
2985                 }
2986                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2987                 break;
2988         }
2989
2990         case IPV6_2292HOPLIMIT:
2991         case IPV6_HOPLIMIT:
2992         {
2993                 int *hlimp;
2994
2995                 /*
2996                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2997                  * to simplify the ordering among hoplimit options.
2998                  */
2999                 if (optname == IPV6_HOPLIMIT && sticky)
3000                         return (ENOPROTOOPT);
3001
3002                 if (len != sizeof(int))
3003                         return (EINVAL);
3004                 hlimp = (int *)buf;
3005                 if (*hlimp < -1 || *hlimp > 255)
3006                         return (EINVAL);
3007
3008                 opt->ip6po_hlim = *hlimp;
3009                 break;
3010         }
3011
3012         case IPV6_TCLASS:
3013         {
3014                 int tclass;
3015
3016                 if (len != sizeof(int))
3017                         return (EINVAL);
3018                 tclass = *(int *)buf;
3019                 if (tclass < -1 || tclass > 255)
3020                         return (EINVAL);
3021
3022                 opt->ip6po_tclass = tclass;
3023                 break;
3024         }
3025
3026         case IPV6_2292NEXTHOP:
3027         case IPV6_NEXTHOP:
3028                 if (cred != NULL) {
3029                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3030                         if (error)
3031                                 return (error);
3032                 }
3033
3034                 if (len == 0) { /* just remove the option */
3035                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3036                         break;
3037                 }
3038
3039                 /* check if cmsg_len is large enough for sa_len */
3040                 if (len < sizeof(struct sockaddr) || len < *buf)
3041                         return (EINVAL);
3042
3043                 switch (((struct sockaddr *)buf)->sa_family) {
3044                 case AF_INET6:
3045                 {
3046                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3047                         int error;
3048
3049                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3050                                 return (EINVAL);
3051
3052                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3053                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3054                                 return (EINVAL);
3055                         }
3056                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3057                             != 0) {
3058                                 return (error);
3059                         }
3060                         break;
3061                 }
3062                 case AF_LINK:   /* should eventually be supported */
3063                 default:
3064                         return (EAFNOSUPPORT);
3065                 }
3066
3067                 /* turn off the previous option, then set the new option. */
3068                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3069                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3070                 if (opt->ip6po_nexthop == NULL)
3071                         return (ENOBUFS);
3072                 bcopy(buf, opt->ip6po_nexthop, *buf);
3073                 break;
3074
3075         case IPV6_2292HOPOPTS:
3076         case IPV6_HOPOPTS:
3077         {
3078                 struct ip6_hbh *hbh;
3079                 int hbhlen;
3080
3081                 /*
3082                  * XXX: We don't allow a non-privileged user to set ANY HbH
3083                  * options, since per-option restriction has too much
3084                  * overhead.
3085                  */
3086                 if (cred != NULL) {
3087                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3088                         if (error)
3089                                 return (error);
3090                 }
3091
3092                 if (len == 0) {
3093                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3094                         break;  /* just remove the option */
3095                 }
3096
3097                 /* message length validation */
3098                 if (len < sizeof(struct ip6_hbh))
3099                         return (EINVAL);
3100                 hbh = (struct ip6_hbh *)buf;
3101                 hbhlen = (hbh->ip6h_len + 1) << 3;
3102                 if (len != hbhlen)
3103                         return (EINVAL);
3104
3105                 /* turn off the previous option, then set the new option. */
3106                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3107                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3108                 if (opt->ip6po_hbh == NULL)
3109                         return (ENOBUFS);
3110                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3111
3112                 break;
3113         }
3114
3115         case IPV6_2292DSTOPTS:
3116         case IPV6_DSTOPTS:
3117         case IPV6_RTHDRDSTOPTS:
3118         {
3119                 struct ip6_dest *dest, **newdest = NULL;
3120                 int destlen;
3121
3122                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3123                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3124                         if (error)
3125                                 return (error);
3126                 }
3127
3128                 if (len == 0) {
3129                         ip6_clearpktopts(opt, optname);
3130                         break;  /* just remove the option */
3131                 }
3132
3133                 /* message length validation */
3134                 if (len < sizeof(struct ip6_dest))
3135                         return (EINVAL);
3136                 dest = (struct ip6_dest *)buf;
3137                 destlen = (dest->ip6d_len + 1) << 3;
3138                 if (len != destlen)
3139                         return (EINVAL);
3140
3141                 /*
3142                  * Determine the position that the destination options header
3143                  * should be inserted; before or after the routing header.
3144                  */
3145                 switch (optname) {
3146                 case IPV6_2292DSTOPTS:
3147                         /*
3148                          * The old advacned API is ambiguous on this point.
3149                          * Our approach is to determine the position based
3150                          * according to the existence of a routing header.
3151                          * Note, however, that this depends on the order of the
3152                          * extension headers in the ancillary data; the 1st
3153                          * part of the destination options header must appear
3154                          * before the routing header in the ancillary data,
3155                          * too.
3156                          * RFC3542 solved the ambiguity by introducing
3157                          * separate ancillary data or option types.
3158                          */
3159                         if (opt->ip6po_rthdr == NULL)
3160                                 newdest = &opt->ip6po_dest1;
3161                         else
3162                                 newdest = &opt->ip6po_dest2;
3163                         break;
3164                 case IPV6_RTHDRDSTOPTS:
3165                         newdest = &opt->ip6po_dest1;
3166                         break;
3167                 case IPV6_DSTOPTS:
3168                         newdest = &opt->ip6po_dest2;
3169                         break;
3170                 }
3171
3172                 /* turn off the previous option, then set the new option. */
3173                 ip6_clearpktopts(opt, optname);
3174                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3175                 if (*newdest == NULL)
3176                         return (ENOBUFS);
3177                 bcopy(dest, *newdest, destlen);
3178
3179                 break;
3180         }
3181
3182         case IPV6_2292RTHDR:
3183         case IPV6_RTHDR:
3184         {
3185                 struct ip6_rthdr *rth;
3186                 int rthlen;
3187
3188                 if (len == 0) {
3189                         ip6_clearpktopts(opt, IPV6_RTHDR);
3190                         break;  /* just remove the option */
3191                 }
3192
3193                 /* message length validation */
3194                 if (len < sizeof(struct ip6_rthdr))
3195                         return (EINVAL);
3196                 rth = (struct ip6_rthdr *)buf;
3197                 rthlen = (rth->ip6r_len + 1) << 3;
3198                 if (len != rthlen)
3199                         return (EINVAL);
3200
3201                 switch (rth->ip6r_type) {
3202                 case IPV6_RTHDR_TYPE_0:
3203                         if (rth->ip6r_len == 0) /* must contain one addr */
3204                                 return (EINVAL);
3205                         if (rth->ip6r_len % 2) /* length must be even */
3206                                 return (EINVAL);
3207                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3208                                 return (EINVAL);
3209                         break;
3210                 default:
3211                         return (EINVAL);        /* not supported */
3212                 }
3213
3214                 /* turn off the previous option */
3215                 ip6_clearpktopts(opt, IPV6_RTHDR);
3216                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3217                 if (opt->ip6po_rthdr == NULL)
3218                         return (ENOBUFS);
3219                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3220
3221                 break;
3222         }
3223
3224         case IPV6_USE_MIN_MTU:
3225                 if (len != sizeof(int))
3226                         return (EINVAL);
3227                 minmtupolicy = *(int *)buf;
3228                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3229                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3230                     minmtupolicy != IP6PO_MINMTU_ALL) {
3231                         return (EINVAL);
3232                 }
3233                 opt->ip6po_minmtu = minmtupolicy;
3234                 break;
3235
3236         case IPV6_DONTFRAG:
3237                 if (len != sizeof(int))
3238                         return (EINVAL);
3239
3240                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3241                         /*
3242                          * we ignore this option for TCP sockets.
3243                          * (RFC3542 leaves this case unspecified.)
3244                          */
3245                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3246                 } else
3247                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3248                 break;
3249
3250         case IPV6_PREFER_TEMPADDR:
3251                 if (len != sizeof(int))
3252                         return (EINVAL);
3253                 preftemp = *(int *)buf;
3254                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3255                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3256                     preftemp != IP6PO_TEMPADDR_PREFER) {
3257                         return (EINVAL);
3258                 }
3259                 opt->ip6po_prefer_tempaddr = preftemp;
3260                 break;
3261
3262         default:
3263                 return (ENOPROTOOPT);
3264         } /* end of switch */
3265
3266         return (0);
3267 }
3268
3269 /*
3270  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3271  * packet to the input queue of a specified interface.  Note that this
3272  * calls the output routine of the loopback "driver", but with an interface
3273  * pointer that might NOT be &loif -- easier than replicating that code here.
3274  */
3275 void
3276 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3277 {
3278         struct mbuf *copym;
3279         struct ip6_hdr *ip6;
3280
3281         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3282         if (copym == NULL)
3283                 return;
3284
3285         /*
3286          * Make sure to deep-copy IPv6 header portion in case the data
3287          * is in an mbuf cluster, so that we can safely override the IPv6
3288          * header portion later.
3289          */
3290         if (!M_WRITABLE(copym) ||
3291             copym->m_len < sizeof(struct ip6_hdr)) {
3292                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3293                 if (copym == NULL)
3294                         return;
3295         }
3296         ip6 = mtod(copym, struct ip6_hdr *);
3297         /*
3298          * clear embedded scope identifiers if necessary.
3299          * in6_clearscope will touch the addresses only when necessary.
3300          */
3301         in6_clearscope(&ip6->ip6_src);
3302         in6_clearscope(&ip6->ip6_dst);
3303         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3304                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3305                     CSUM_PSEUDO_HDR;
3306                 copym->m_pkthdr.csum_data = 0xffff;
3307         }
3308         if_simloop(ifp, copym, AF_INET6, 0);
3309 }
3310
3311 /*
3312  * Chop IPv6 header off from the payload.
3313  */
3314 static int
3315 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3316 {
3317         struct mbuf *mh;
3318         struct ip6_hdr *ip6;
3319
3320         ip6 = mtod(m, struct ip6_hdr *);
3321         if (m->m_len > sizeof(*ip6)) {
3322                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3323                 if (mh == NULL) {
3324                         m_freem(m);
3325                         return ENOBUFS;
3326                 }
3327                 m_move_pkthdr(mh, m);
3328                 M_ALIGN(mh, sizeof(*ip6));
3329                 m->m_len -= sizeof(*ip6);
3330                 m->m_data += sizeof(*ip6);
3331                 mh->m_next = m;
3332                 m = mh;
3333                 m->m_len = sizeof(*ip6);
3334                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3335         }
3336         exthdrs->ip6e_ip6 = m;
3337         return 0;
3338 }
3339
3340 /*
3341  * Compute IPv6 extension header length.
3342  */
3343 int
3344 ip6_optlen(struct inpcb *inp)
3345 {
3346         int len;
3347
3348         if (!inp->in6p_outputopts)
3349                 return 0;
3350
3351         len = 0;
3352 #define elen(x) \
3353     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3354
3355         len += elen(inp->in6p_outputopts->ip6po_hbh);
3356         if (inp->in6p_outputopts->ip6po_rthdr)
3357                 /* dest1 is valid with rthdr only */
3358                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3359         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3360         len += elen(inp->in6p_outputopts->ip6po_dest2);
3361         return len;
3362 #undef elen
3363 }