sys/netinet6/ip6_output.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipfw.h"
  69 #include "opt_ipsec.h"
  70 #include "opt_sctp.h"
  71 #include "opt_route.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/kernel.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/priv.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/socket.h>
  82 #include <sys/socketvar.h>
  83 #include <sys/syslog.h>
  84 #include <sys/ucred.h>
  85
  86 #include <machine/in_cksum.h>
  87
  88 #include <net/if.h>
  89 #include <net/netisr.h>
  90 #include <net/route.h>
  91 #include <net/pfil.h>
  92 #include <net/vnet.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet6/in6_var.h>
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/tcp_var.h>
 103 #include <netinet6/nd6.h>
 104
 105 #ifdef IPSEC
 106 #include <netipsec/ipsec.h>
 107 #include <netipsec/ipsec6.h>
 108 #include <netipsec/key.h>
 109 #include <netinet6/ip6_ipsec.h>
 110 #endif /* IPSEC */
 111 #ifdef SCTP
 112 #include <netinet/sctp.h>
 113 #include <netinet/sctp_crc32.h>
 114 #endif
 115
 116 #include <netinet6/ip6protosw.h>
 117 #include <netinet6/scope6_var.h>
 118
 119 #ifdef FLOWTABLE
 120 #include <net/flowtable.h>
 121 #endif
 122
 123 extern int in6_mcast_loop;
 124
 125 struct ip6_exthdrs {
 126         struct mbuf *ip6e_ip6;
 127         struct mbuf *ip6e_hbh;
 128         struct mbuf *ip6e_dest1;
 129         struct mbuf *ip6e_rthdr;
 130         struct mbuf *ip6e_dest2;
 131 };
 132
 133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 134                            struct ucred *, int);
 135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 136         struct socket *, struct sockopt *);
 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 139         struct ucred *, int, int, int);
 140
 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 143         struct ip6_frag **);
 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 147         struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
 148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 149
 150
 151 /*
 152  * Make an extension header from option data.  hp is the source, and
 153  * mp is the destination.
 154  */
 155 #define MAKE_EXTHDR(hp, mp)                                             \
 156     do {                                                                \
 157         if (hp) {                                                       \
 158                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 159                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 160                     ((eh)->ip6e_len + 1) << 3);                         \
 161                 if (error)                                              \
 162                         goto freehdrs;                                  \
 163         }                                                               \
 164     } while (/*CONSTCOND*/ 0)
 165
 166 /*
 167  * Form a chain of extension headers.
 168  * m is the extension header mbuf
 169  * mp is the previous mbuf in the chain
 170  * p is the next header
 171  * i is the type of option.
 172  */
 173 #define MAKE_CHAIN(m, mp, p, i)\
 174     do {\
 175         if (m) {\
 176                 if (!hdrsplit) \
 177                         panic("assumption failed: hdr not split"); \
 178                 *mtod((m), u_char *) = *(p);\
 179                 *(p) = (i);\
 180                 p = mtod((m), u_char *);\
 181                 (m)->m_next = (mp)->m_next;\
 182                 (mp)->m_next = (m);\
 183                 (mp) = (m);\
 184         }\
 185     } while (/*CONSTCOND*/ 0)
 186
 187 static void
 188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 189 {
 190         u_short csum;
 191
 192         csum = in_cksum_skip(m, offset + plen, offset);
 193         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 194                 csum = 0xffff;
 195         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 196
 197         if (offset + sizeof(u_short) > m->m_len) {
 198                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 199                     "csum_flags=0x%04x\n", __func__, m->m_len, plen, offset,
 200                     m->m_pkthdr.csum_flags);
 201                 /*
 202                  * XXX this should not happen, but if it does, the correct
 203                  * behavior may be to insert the checksum in the appropriate
 204                  * next mbuf in the chain.
 205                  */
 206                 return;
 207         }
 208         *(u_short *)(m->m_data + offset) = csum;
 209 }
 210
 211 /*
 212  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 213  * header (with pri, len, nxt, hlim, src, dst).
 214  * This function may modify ver and hlim only.
 215  * The mbuf chain containing the packet will be freed.
 216  * The mbuf opt, if present, will not be freed.
 217  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 218  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 219  * then result of route lookup is stored in ro->ro_rt.
 220  *
 221  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 222  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 223  * which is rt_rmx.rmx_mtu.
 224  *
 225  * ifpp - XXX: just for statistics
 226  */
 227 int
 228 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 229     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 230     struct ifnet **ifpp, struct inpcb *inp)
 231 {
 232         struct ip6_hdr *ip6, *mhip6;
 233         struct ifnet *ifp, *origifp;
 234         struct mbuf *m = m0;
 235         struct mbuf *mprev = NULL;
 236         int hlen, tlen, len, off;
 237         struct route_in6 ip6route;
 238         struct rtentry *rt = NULL;
 239         struct sockaddr_in6 *dst, src_sa, dst_sa;
 240         struct in6_addr odst;
 241         int error = 0;
 242         struct in6_ifaddr *ia = NULL;
 243         u_long mtu;
 244         int alwaysfrag, dontfrag;
 245         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 246         struct ip6_exthdrs exthdrs;
 247         struct in6_addr finaldst, src0, dst0;
 248         u_int32_t zone;
 249         struct route_in6 *ro_pmtu = NULL;
 250         int hdrsplit = 0;
 251         int needipsec = 0;
 252         int sw_csum, tso;
 253 #ifdef IPSEC
 254         struct ipsec_output_state state;
 255         struct ip6_rthdr *rh = NULL;
 256         int needipsectun = 0;
 257         int segleft_org = 0;
 258         struct secpolicy *sp = NULL;
 259 #endif /* IPSEC */
 260         struct m_tag *fwd_tag = NULL;
 261
 262         ip6 = mtod(m, struct ip6_hdr *);
 263         if (ip6 == NULL) {
 264                 printf ("ip6 is NULL");
 265                 goto bad;
 266         }
 267
 268         if (inp != NULL)
 269                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 270
 271         finaldst = ip6->ip6_dst;
 272         bzero(&exthdrs, sizeof(exthdrs));
 273         if (opt) {
 274                 /* Hop-by-Hop options header */
 275                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 276                 /* Destination options header(1st part) */
 277                 if (opt->ip6po_rthdr) {
 278                         /*
 279                          * Destination options header(1st part)
 280                          * This only makes sense with a routing header.
 281                          * See Section 9.2 of RFC 3542.
 282                          * Disabling this part just for MIP6 convenience is
 283                          * a bad idea.  We need to think carefully about a
 284                          * way to make the advanced API coexist with MIP6
 285                          * options, which might automatically be inserted in
 286                          * the kernel.
 287                          */
 288                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 289                 }
 290                 /* Routing header */
 291                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 292                 /* Destination options header(2nd part) */
 293                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 294         }
 295
 296 #ifdef IPSEC
 297         /*
 298          * IPSec checking which handles several cases.
 299          * FAST IPSEC: We re-injected the packet.
 300          */
 301         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 302         {
 303         case 1:                 /* Bad packet */
 304                 goto freehdrs;
 305         case -1:                /* Do IPSec */
 306                 needipsec = 1;
 307                 /*
 308                  * Do delayed checksums now, as we may send before returning.
 309                  */
 310                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 311                         plen = m->m_pkthdr.len - sizeof(*ip6);
 312                         in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 313                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 314                 }
 315 #ifdef SCTP
 316                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 317                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 318                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 319                 }
 320 #endif
 321         case 0:                 /* No IPSec */
 322         default:
 323                 break;
 324         }
 325 #endif /* IPSEC */
 326
 327         /*
 328          * Calculate the total length of the extension header chain.
 329          * Keep the length of the unfragmentable part for fragmentation.
 330          */
 331         optlen = 0;
 332         if (exthdrs.ip6e_hbh)
 333                 optlen += exthdrs.ip6e_hbh->m_len;
 334         if (exthdrs.ip6e_dest1)
 335                 optlen += exthdrs.ip6e_dest1->m_len;
 336         if (exthdrs.ip6e_rthdr)
 337                 optlen += exthdrs.ip6e_rthdr->m_len;
 338         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 339
 340         /* NOTE: we don't add AH/ESP length here. do that later. */
 341         if (exthdrs.ip6e_dest2)
 342                 optlen += exthdrs.ip6e_dest2->m_len;
 343
 344         /*
 345          * If we need IPsec, or there is at least one extension header,
 346          * separate IP6 header from the payload.
 347          */
 348         if ((needipsec || optlen) && !hdrsplit) {
 349                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 350                         m = NULL;
 351                         goto freehdrs;
 352                 }
 353                 m = exthdrs.ip6e_ip6;
 354                 hdrsplit++;
 355         }
 356
 357         /* adjust pointer */
 358         ip6 = mtod(m, struct ip6_hdr *);
 359
 360         /* adjust mbuf packet header length */
 361         m->m_pkthdr.len += optlen;
 362         plen = m->m_pkthdr.len - sizeof(*ip6);
 363
 364         /* If this is a jumbo payload, insert a jumbo payload option. */
 365         if (plen > IPV6_MAXPACKET) {
 366                 if (!hdrsplit) {
 367                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 368                                 m = NULL;
 369                                 goto freehdrs;
 370                         }
 371                         m = exthdrs.ip6e_ip6;
 372                         hdrsplit++;
 373                 }
 374                 /* adjust pointer */
 375                 ip6 = mtod(m, struct ip6_hdr *);
 376                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 377                         goto freehdrs;
 378                 ip6->ip6_plen = 0;
 379         } else
 380                 ip6->ip6_plen = htons(plen);
 381
 382         /*
 383          * Concatenate headers and fill in next header fields.
 384          * Here we have, on "m"
 385          *      IPv6 payload
 386          * and we insert headers accordingly.  Finally, we should be getting:
 387          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 388          *
 389          * during the header composing process, "m" points to IPv6 header.
 390          * "mprev" points to an extension header prior to esp.
 391          */
 392         u_char *nexthdrp = &ip6->ip6_nxt;
 393         mprev = m;
 394
 395         /*
 396          * we treat dest2 specially.  this makes IPsec processing
 397          * much easier.  the goal here is to make mprev point the
 398          * mbuf prior to dest2.
 399          *
 400          * result: IPv6 dest2 payload
 401          * m and mprev will point to IPv6 header.
 402          */
 403         if (exthdrs.ip6e_dest2) {
 404                 if (!hdrsplit)
 405                         panic("assumption failed: hdr not split");
 406                 exthdrs.ip6e_dest2->m_next = m->m_next;
 407                 m->m_next = exthdrs.ip6e_dest2;
 408                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 409                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 410         }
 411
 412         /*
 413          * result: IPv6 hbh dest1 rthdr dest2 payload
 414          * m will point to IPv6 header.  mprev will point to the
 415          * extension header prior to dest2 (rthdr in the above case).
 416          */
 417         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 418         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 419                    IPPROTO_DSTOPTS);
 420         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 421                    IPPROTO_ROUTING);
 422
 423 #ifdef IPSEC
 424         if (!needipsec)
 425                 goto skip_ipsec2;
 426
 427         /*
 428          * pointers after IPsec headers are not valid any more.
 429          * other pointers need a great care too.
 430          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 431          */
 432         exthdrs.ip6e_dest2 = NULL;
 433
 434         if (exthdrs.ip6e_rthdr) {
 435                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 436                 segleft_org = rh->ip6r_segleft;
 437                 rh->ip6r_segleft = 0;
 438         }
 439
 440         bzero(&state, sizeof(state));
 441         state.m = m;
 442         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 443                                     &needipsectun);
 444         m = state.m;
 445         if (error == EJUSTRETURN) {
 446                 /*
 447                  * We had a SP with a level of 'use' and no SA. We
 448                  * will just continue to process the packet without
 449                  * IPsec processing.
 450                  */
 451                 ;
 452         } else if (error) {
 453                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 454                 m = NULL;
 455                 switch (error) {
 456                 case EHOSTUNREACH:
 457                 case ENETUNREACH:
 458                 case EMSGSIZE:
 459                 case ENOBUFS:
 460                 case ENOMEM:
 461                         break;
 462                 default:
 463                         printf("[%s:%d] (ipsec): error code %d\n",
 464                             __func__, __LINE__, error);
 465                         /* FALLTHROUGH */
 466                 case ENOENT:
 467                         /* don't show these error codes to the user */
 468                         error = 0;
 469                         break;
 470                 }
 471                 goto bad;
 472         } else if (!needipsectun) {
 473                 /*
 474                  * In the FAST IPSec case we have already
 475                  * re-injected the packet and it has been freed
 476                  * by the ipsec_done() function.  So, just clean
 477                  * up after ourselves.
 478                  */
 479                 m = NULL;
 480                 goto done;
 481         }
 482         if (exthdrs.ip6e_rthdr) {
 483                 /* ah6_output doesn't modify mbuf chain */
 484                 rh->ip6r_segleft = segleft_org;
 485         }
 486 skip_ipsec2:;
 487 #endif /* IPSEC */
 488
 489         /*
 490          * If there is a routing header, discard the packet.
 491          */
 492         if (exthdrs.ip6e_rthdr) {
 493                  error = EINVAL;
 494                  goto bad;
 495         }
 496
 497         /* Source address validation */
 498         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 499             (flags & IPV6_UNSPECSRC) == 0) {
 500                 error = EOPNOTSUPP;
 501                 IP6STAT_INC(ip6s_badscope);
 502                 goto bad;
 503         }
 504         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 505                 error = EOPNOTSUPP;
 506                 IP6STAT_INC(ip6s_badscope);
 507                 goto bad;
 508         }
 509
 510         IP6STAT_INC(ip6s_localout);
 511
 512         /*
 513          * Route packet.
 514          */
 515         if (ro == 0) {
 516                 ro = &ip6route;
 517                 bzero((caddr_t)ro, sizeof(*ro));
 518         }
 519         ro_pmtu = ro;
 520         if (opt && opt->ip6po_rthdr)
 521                 ro = &opt->ip6po_route;
 522         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 523 #ifdef FLOWTABLE
 524         if (ro->ro_rt == NULL) {
 525                 struct flentry *fle;
 526
 527                 /*
 528                  * The flow table returns route entries valid for up to 30
 529                  * seconds; we rely on the remainder of ip_output() taking no
 530                  * longer than that long for the stability of ro_rt.  The
 531                  * flow ID assignment must have happened before this point.
 532                  */
 533                 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
 534                 if (fle != NULL)
 535                         flow_to_route_in6(fle, ro);
 536         }
 537 #endif
 538 again:
 539         /*
 540          * if specified, try to fill in the traffic class field.
 541          * do not override if a non-zero value is already set.
 542          * we check the diffserv field and the ecn field separately.
 543          */
 544         if (opt && opt->ip6po_tclass >= 0) {
 545                 int mask = 0;
 546
 547                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 548                         mask |= 0xfc;
 549                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 550                         mask |= 0x03;
 551                 if (mask != 0)
 552                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 553         }
 554
 555         /* fill in or override the hop limit field, if necessary. */
 556         if (opt && opt->ip6po_hlim != -1)
 557                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 558         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 559                 if (im6o != NULL)
 560                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 561                 else
 562                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 563         }
 564
 565 #ifdef IPSEC
 566         /*
 567          * We may re-inject packets into the stack here.
 568          */
 569         if (needipsec && needipsectun) {
 570                 struct ipsec_output_state state;
 571
 572                 /*
 573                  * All the extension headers will become inaccessible
 574                  * (since they can be encrypted).
 575                  * Don't panic, we need no more updates to extension headers
 576                  * on inner IPv6 packet (since they are now encapsulated).
 577                  *
 578                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 579                  */
 580                 bzero(&exthdrs, sizeof(exthdrs));
 581                 exthdrs.ip6e_ip6 = m;
 582
 583                 bzero(&state, sizeof(state));
 584                 state.m = m;
 585                 state.ro = (struct route *)ro;
 586                 state.dst = (struct sockaddr *)dst;
 587
 588                 error = ipsec6_output_tunnel(&state, sp, flags);
 589
 590                 m = state.m;
 591                 ro = (struct route_in6 *)state.ro;
 592                 dst = (struct sockaddr_in6 *)state.dst;
 593                 if (error == EJUSTRETURN) {
 594                         /*
 595                          * We had a SP with a level of 'use' and no SA. We
 596                          * will just continue to process the packet without
 597                          * IPsec processing.
 598                          */
 599                         ;
 600                 } else if (error) {
 601                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 602                         m0 = m = NULL;
 603                         m = NULL;
 604                         switch (error) {
 605                         case EHOSTUNREACH:
 606                         case ENETUNREACH:
 607                         case EMSGSIZE:
 608                         case ENOBUFS:
 609                         case ENOMEM:
 610                                 break;
 611                         default:
 612                                 printf("[%s:%d] (ipsec): error code %d\n",
 613                                     __func__, __LINE__, error);
 614                                 /* FALLTHROUGH */
 615                         case ENOENT:
 616                                 /* don't show these error codes to the user */
 617                                 error = 0;
 618                                 break;
 619                         }
 620                         goto bad;
 621                 } else {
 622                         /*
 623                          * In the FAST IPSec case we have already
 624                          * re-injected the packet and it has been freed
 625                          * by the ipsec_done() function.  So, just clean
 626                          * up after ourselves.
 627                          */
 628                         m = NULL;
 629                         goto done;
 630                 }
 631
 632                 exthdrs.ip6e_ip6 = m;
 633         }
 634 #endif /* IPSEC */
 635
 636         /* adjust pointer */
 637         ip6 = mtod(m, struct ip6_hdr *);
 638
 639         if (ro->ro_rt && fwd_tag == NULL) {
 640                 rt = ro->ro_rt;
 641                 ifp = ro->ro_rt->rt_ifp;
 642         } else {
 643                 if (fwd_tag == NULL) {
 644                         bzero(&dst_sa, sizeof(dst_sa));
 645                         dst_sa.sin6_family = AF_INET6;
 646                         dst_sa.sin6_len = sizeof(dst_sa);
 647                         dst_sa.sin6_addr = ip6->ip6_dst;
 648                 }
 649                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 650                     &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
 651                 if (error != 0) {
 652                         if (ifp != NULL)
 653                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 654                         goto bad;
 655                 }
 656         }
 657         if (rt == NULL) {
 658                 /*
 659                  * If in6_selectroute() does not return a route entry,
 660                  * dst may not have been updated.
 661                  */
 662                 *dst = dst_sa;  /* XXX */
 663         }
 664
 665         /*
 666          * then rt (for unicast) and ifp must be non-NULL valid values.
 667          */
 668         if ((flags & IPV6_FORWARDING) == 0) {
 669                 /* XXX: the FORWARDING flag can be set for mrouting. */
 670                 in6_ifstat_inc(ifp, ifs6_out_request);
 671         }
 672         if (rt != NULL) {
 673                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 674                 rt->rt_use++;
 675         }
 676
 677
 678         /*
 679          * The outgoing interface must be in the zone of source and
 680          * destination addresses.
 681          */
 682         origifp = ifp;
 683
 684         src0 = ip6->ip6_src;
 685         if (in6_setscope(&src0, origifp, &zone))
 686                 goto badscope;
 687         bzero(&src_sa, sizeof(src_sa));
 688         src_sa.sin6_family = AF_INET6;
 689         src_sa.sin6_len = sizeof(src_sa);
 690         src_sa.sin6_addr = ip6->ip6_src;
 691         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 692                 goto badscope;
 693
 694         dst0 = ip6->ip6_dst;
 695         if (in6_setscope(&dst0, origifp, &zone))
 696                 goto badscope;
 697         /* re-initialize to be sure */
 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         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 703                 goto badscope;
 704         }
 705
 706         /* We should use ia_ifp to support the case of
 707          * sending packets to an address of our own.
 708          */
 709         if (ia != NULL && ia->ia_ifp)
 710                 ifp = ia->ia_ifp;
 711
 712         /* scope check is done. */
 713         goto routefound;
 714
 715   badscope:
 716         IP6STAT_INC(ip6s_badscope);
 717         in6_ifstat_inc(origifp, ifs6_out_discard);
 718         if (error == 0)
 719                 error = EHOSTUNREACH; /* XXX */
 720         goto bad;
 721
 722   routefound:
 723         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 724                 if (opt && opt->ip6po_nextroute.ro_rt) {
 725                         /*
 726                          * The nexthop is explicitly specified by the
 727                          * application.  We assume the next hop is an IPv6
 728                          * address.
 729                          */
 730                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 731                 }
 732                 else if ((rt->rt_flags & RTF_GATEWAY))
 733                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 734         }
 735
 736         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 737                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 738         } else {
 739                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 740                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 741                 /*
 742                  * Confirm that the outgoing interface supports multicast.
 743                  */
 744                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 745                         IP6STAT_INC(ip6s_noroute);
 746                         in6_ifstat_inc(ifp, ifs6_out_discard);
 747                         error = ENETUNREACH;
 748                         goto bad;
 749                 }
 750                 if ((im6o == NULL && in6_mcast_loop) ||
 751                     (im6o && im6o->im6o_multicast_loop)) {
 752                         /*
 753                          * Loop back multicast datagram if not expressly
 754                          * forbidden to do so, even if we have not joined
 755                          * the address; protocols will filter it later,
 756                          * thus deferring a hash lookup and lock acquisition
 757                          * at the expense of an m_copym().
 758                          */
 759                         ip6_mloopback(ifp, m, dst);
 760                 } else {
 761                         /*
 762                          * If we are acting as a multicast router, perform
 763                          * multicast forwarding as if the packet had just
 764                          * arrived on the interface to which we are about
 765                          * to send.  The multicast forwarding function
 766                          * recursively calls this function, using the
 767                          * IPV6_FORWARDING flag to prevent infinite recursion.
 768                          *
 769                          * Multicasts that are looped back by ip6_mloopback(),
 770                          * above, will be forwarded by the ip6_input() routine,
 771                          * if necessary.
 772                          */
 773                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 774                                 /*
 775                                  * XXX: ip6_mforward expects that rcvif is NULL
 776                                  * when it is called from the originating path.
 777                                  * However, it may not always be the case.
 778                                  */
 779                                 m->m_pkthdr.rcvif = NULL;
 780                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 781                                         m_freem(m);
 782                                         goto done;
 783                                 }
 784                         }
 785                 }
 786                 /*
 787                  * Multicasts with a hoplimit of zero may be looped back,
 788                  * above, but must not be transmitted on a network.
 789                  * Also, multicasts addressed to the loopback interface
 790                  * are not sent -- the above call to ip6_mloopback() will
 791                  * loop back a copy if this host actually belongs to the
 792                  * destination group on the loopback interface.
 793                  */
 794                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 795                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 796                         m_freem(m);
 797                         goto done;
 798                 }
 799         }
 800
 801         /*
 802          * Fill the outgoing inteface to tell the upper layer
 803          * to increment per-interface statistics.
 804          */
 805         if (ifpp)
 806                 *ifpp = ifp;
 807
 808         /* Determine path MTU. */
 809         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 810             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 811                 goto bad;
 812
 813         /*
 814          * The caller of this function may specify to use the minimum MTU
 815          * in some cases.
 816          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 817          * setting.  The logic is a bit complicated; by default, unicast
 818          * packets will follow path MTU while multicast packets will be sent at
 819          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 820          * including unicast ones will be sent at the minimum MTU.  Multicast
 821          * packets will always be sent at the minimum MTU unless
 822          * IP6PO_MINMTU_DISABLE is explicitly specified.
 823          * See RFC 3542 for more details.
 824          */
 825         if (mtu > IPV6_MMTU) {
 826                 if ((flags & IPV6_MINMTU))
 827                         mtu = IPV6_MMTU;
 828                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 829                         mtu = IPV6_MMTU;
 830                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 831                          (opt == NULL ||
 832                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 833                         mtu = IPV6_MMTU;
 834                 }
 835         }
 836
 837         /*
 838          * clear embedded scope identifiers if necessary.
 839          * in6_clearscope will touch the addresses only when necessary.
 840          */
 841         in6_clearscope(&ip6->ip6_src);
 842         in6_clearscope(&ip6->ip6_dst);
 843
 844         /*
 845          * If the outgoing packet contains a hop-by-hop options header,
 846          * it must be examined and processed even by the source node.
 847          * (RFC 2460, section 4.)
 848          */
 849         if (exthdrs.ip6e_hbh) {
 850                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 851                 u_int32_t dummy; /* XXX unused */
 852                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 853
 854 #ifdef DIAGNOSTIC
 855                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 856                         panic("ip6e_hbh is not contiguous");
 857 #endif
 858                 /*
 859                  *  XXX: if we have to send an ICMPv6 error to the sender,
 860                  *       we need the M_LOOP flag since icmp6_error() expects
 861                  *       the IPv6 and the hop-by-hop options header are
 862                  *       contiguous unless the flag is set.
 863                  */
 864                 m->m_flags |= M_LOOP;
 865                 m->m_pkthdr.rcvif = ifp;
 866                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 867                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 868                     &dummy, &plen) < 0) {
 869                         /* m was already freed at this point */
 870                         error = EINVAL;/* better error? */
 871                         goto done;
 872                 }
 873                 m->m_flags &= ~M_LOOP; /* XXX */
 874                 m->m_pkthdr.rcvif = NULL;
 875         }
 876
 877         /* Jump over all PFIL processing if hooks are not active. */
 878         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 879                 goto passout;
 880
 881         odst = ip6->ip6_dst;
 882         /* Run through list of hooks for output packets. */
 883         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 884         if (error != 0 || m == NULL)
 885                 goto done;
 886         ip6 = mtod(m, struct ip6_hdr *);
 887
 888         /* See if destination IP address was changed by packet filter. */
 889         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 890                 m->m_flags |= M_SKIP_FIREWALL;
 891                 /* If destination is now ourself drop to ip6_input(). */
 892                 if (in6_localip(&ip6->ip6_dst)) {
 893                         m->m_flags |= M_FASTFWD_OURS;
 894                         if (m->m_pkthdr.rcvif == NULL)
 895                                 m->m_pkthdr.rcvif = V_loif;
 896                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 897                                 m->m_pkthdr.csum_flags |=
 898                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 899                                 m->m_pkthdr.csum_data = 0xffff;
 900                         }
 901 #ifdef SCTP
 902                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 903                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 904 #endif
 905                         error = netisr_queue(NETISR_IPV6, m);
 906                         goto done;
 907                 } else
 908                         goto again;     /* Redo the routing table lookup. */
 909         }
 910
 911         /* See if local, if yes, send it to netisr. */
 912         if (m->m_flags & M_FASTFWD_OURS) {
 913                 if (m->m_pkthdr.rcvif == NULL)
 914                         m->m_pkthdr.rcvif = V_loif;
 915                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 916                         m->m_pkthdr.csum_flags |=
 917                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 918                         m->m_pkthdr.csum_data = 0xffff;
 919                 }
 920 #ifdef SCTP
 921                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 922                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 923 #endif
 924                 error = netisr_queue(NETISR_IPV6, m);
 925                 goto done;
 926         }
 927         /* Or forward to some other address? */
 928         if ((m->m_flags & M_IP6_NEXTHOP) &&
 929             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 930                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 931                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 932                 m->m_flags |= M_SKIP_FIREWALL;
 933                 m->m_flags &= ~M_IP6_NEXTHOP;
 934                 m_tag_delete(m, fwd_tag);
 935                 goto again;
 936         }
 937
 938 passout:
 939         /*
 940          * Send the packet to the outgoing interface.
 941          * If necessary, do IPv6 fragmentation before sending.
 942          *
 943          * the logic here is rather complex:
 944          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 945          * 1-a: send as is if tlen <= path mtu
 946          * 1-b: fragment if tlen > path mtu
 947          *
 948          * 2: if user asks us not to fragment (dontfrag == 1)
 949          * 2-a: send as is if tlen <= interface mtu
 950          * 2-b: error if tlen > interface mtu
 951          *
 952          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 953          *      always fragment
 954          *
 955          * 4: if dontfrag == 1 && alwaysfrag == 1
 956          *      error, as we cannot handle this conflicting request
 957          */
 958         sw_csum = m->m_pkthdr.csum_flags;
 959         if (!hdrsplit) {
 960                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 961                 sw_csum &= ~ifp->if_hwassist;
 962         } else
 963                 tso = 0;
 964         /*
 965          * If we added extension headers, we will not do TSO and calculate the
 966          * checksums ourselves for now.
 967          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 968          * with ext. hdrs.
 969          */
 970         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 971                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 972                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 973         }
 974 #ifdef SCTP
 975         if (sw_csum & CSUM_SCTP_IPV6) {
 976                 sw_csum &= ~CSUM_SCTP_IPV6;
 977                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 978         }
 979 #endif
 980         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 981         tlen = m->m_pkthdr.len;
 982
 983         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 984                 dontfrag = 1;
 985         else
 986                 dontfrag = 0;
 987         if (dontfrag && alwaysfrag) {   /* case 4 */
 988                 /* conflicting request - can't transmit */
 989                 error = EMSGSIZE;
 990                 goto bad;
 991         }
 992         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 993                 /*
 994                  * Even if the DONTFRAG option is specified, we cannot send the
 995                  * packet when the data length is larger than the MTU of the
 996                  * outgoing interface.
 997                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 998                  * well as returning an error code (the latter is not described
 999                  * in the API spec.)
1000                  */
1001                 u_int32_t mtu32;
1002                 struct ip6ctlparam ip6cp;
1003
1004                 mtu32 = (u_int32_t)mtu;
1005                 bzero(&ip6cp, sizeof(ip6cp));
1006                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1007                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1008                     (void *)&ip6cp);
1009
1010                 error = EMSGSIZE;
1011                 goto bad;
1012         }
1013
1014         /*
1015          * transmit packet without fragmentation
1016          */
1017         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1018                 struct in6_ifaddr *ia6;
1019
1020                 ip6 = mtod(m, struct ip6_hdr *);
1021                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1022                 if (ia6) {
1023                         /* Record statistics for this interface address. */
1024                         ia6->ia_ifa.if_opackets++;
1025                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1026                         ifa_free(&ia6->ia_ifa);
1027                 }
1028                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1029                 goto done;
1030         }
1031
1032         /*
1033          * try to fragment the packet.  case 1-b and 3
1034          */
1035         if (mtu < IPV6_MMTU) {
1036                 /* path MTU cannot be less than IPV6_MMTU */
1037                 error = EMSGSIZE;
1038                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1039                 goto bad;
1040         } else if (ip6->ip6_plen == 0) {
1041                 /* jumbo payload cannot be fragmented */
1042                 error = EMSGSIZE;
1043                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1044                 goto bad;
1045         } else {
1046                 struct mbuf **mnext, *m_frgpart;
1047                 struct ip6_frag *ip6f;
1048                 u_int32_t id = htonl(ip6_randomid());
1049                 u_char nextproto;
1050
1051                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1052
1053                 /*
1054                  * Too large for the destination or interface;
1055                  * fragment if possible.
1056                  * Must be able to put at least 8 bytes per fragment.
1057                  */
1058                 hlen = unfragpartlen;
1059                 if (mtu > IPV6_MAXPACKET)
1060                         mtu = IPV6_MAXPACKET;
1061
1062                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1063                 if (len < 8) {
1064                         error = EMSGSIZE;
1065                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1066                         goto bad;
1067                 }
1068
1069                 /*
1070                  * Verify that we have any chance at all of being able to queue
1071                  *      the packet or packet fragments
1072                  */
1073                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1074                     < tlen  /* - hlen */)) {
1075                         error = ENOBUFS;
1076                         IP6STAT_INC(ip6s_odropped);
1077                         goto bad;
1078                 }
1079
1080
1081                 /*
1082                  * If the interface will not calculate checksums on
1083                  * fragmented packets, then do it here.
1084                  * XXX-BZ handle the hw offloading case.  Need flags.
1085                  */
1086                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1087                         in6_delayed_cksum(m, plen, hlen);
1088                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1089                 }
1090 #ifdef SCTP
1091                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1092                         sctp_delayed_cksum(m, hlen);
1093                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1094                 }
1095 #endif
1096                 mnext = &m->m_nextpkt;
1097
1098                 /*
1099                  * Change the next header field of the last header in the
1100                  * unfragmentable part.
1101                  */
1102                 if (exthdrs.ip6e_rthdr) {
1103                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1104                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1105                 } else if (exthdrs.ip6e_dest1) {
1106                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1107                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1108                 } else if (exthdrs.ip6e_hbh) {
1109                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1110                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1111                 } else {
1112                         nextproto = ip6->ip6_nxt;
1113                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1114                 }
1115
1116                 /*
1117                  * Loop through length of segment after first fragment,
1118                  * make new header and copy data of each part and link onto
1119                  * chain.
1120                  */
1121                 m0 = m;
1122                 for (off = hlen; off < tlen; off += len) {
1123                         m = m_gethdr(M_NOWAIT, MT_DATA);
1124                         if (!m) {
1125                                 error = ENOBUFS;
1126                                 IP6STAT_INC(ip6s_odropped);
1127                                 goto sendorfree;
1128                         }
1129                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1130                         *mnext = m;
1131                         mnext = &m->m_nextpkt;
1132                         m->m_data += max_linkhdr;
1133                         mhip6 = mtod(m, struct ip6_hdr *);
1134                         *mhip6 = *ip6;
1135                         m->m_len = sizeof(*mhip6);
1136                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1137                         if (error) {
1138                                 IP6STAT_INC(ip6s_odropped);
1139                                 goto sendorfree;
1140                         }
1141                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1142                         if (off + len >= tlen)
1143                                 len = tlen - off;
1144                         else
1145                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1146                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1147                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1148                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1149                                 error = ENOBUFS;
1150                                 IP6STAT_INC(ip6s_odropped);
1151                                 goto sendorfree;
1152                         }
1153                         m_cat(m, m_frgpart);
1154                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1155                         m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
1156                         m->m_pkthdr.rcvif = NULL;
1157                         ip6f->ip6f_reserved = 0;
1158                         ip6f->ip6f_ident = id;
1159                         ip6f->ip6f_nxt = nextproto;
1160                         IP6STAT_INC(ip6s_ofragments);
1161                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1162                 }
1163
1164                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1165         }
1166
1167         /*
1168          * Remove leading garbages.
1169          */
1170 sendorfree:
1171         m = m0->m_nextpkt;
1172         m0->m_nextpkt = 0;
1173         m_freem(m0);
1174         for (m0 = m; m; m = m0) {
1175                 m0 = m->m_nextpkt;
1176                 m->m_nextpkt = 0;
1177                 if (error == 0) {
1178                         /* Record statistics for this interface address. */
1179                         if (ia) {
1180                                 ia->ia_ifa.if_opackets++;
1181                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1182                         }
1183                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1184                 } else
1185                         m_freem(m);
1186         }
1187
1188         if (error == 0)
1189                 IP6STAT_INC(ip6s_fragmented);
1190
1191 done:
1192         if (ro == &ip6route)
1193                 RO_RTFREE(ro);
1194         if (ro_pmtu == &ip6route)
1195                 RO_RTFREE(ro_pmtu);
1196 #ifdef IPSEC
1197         if (sp != NULL)
1198                 KEY_FREESP(&sp);
1199 #endif
1200
1201         return (error);
1202
1203 freehdrs:
1204         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1205         m_freem(exthdrs.ip6e_dest1);
1206         m_freem(exthdrs.ip6e_rthdr);
1207         m_freem(exthdrs.ip6e_dest2);
1208         /* FALLTHROUGH */
1209 bad:
1210         if (m)
1211                 m_freem(m);
1212         goto done;
1213 }
1214
1215 static int
1216 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1217 {
1218         struct mbuf *m;
1219
1220         if (hlen > MCLBYTES)
1221                 return (ENOBUFS); /* XXX */
1222
1223         if (hlen > MLEN)
1224                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1225         else
1226                 m = m_get(M_NOWAIT, MT_DATA);
1227         if (m == NULL)
1228                 return (ENOBUFS);
1229         m->m_len = hlen;
1230         if (hdr)
1231                 bcopy(hdr, mtod(m, caddr_t), hlen);
1232
1233         *mp = m;
1234         return (0);
1235 }
1236
1237 /*
1238  * Insert jumbo payload option.
1239  */
1240 static int
1241 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1242 {
1243         struct mbuf *mopt;
1244         u_char *optbuf;
1245         u_int32_t v;
1246
1247 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1248
1249         /*
1250          * If there is no hop-by-hop options header, allocate new one.
1251          * If there is one but it doesn't have enough space to store the
1252          * jumbo payload option, allocate a cluster to store the whole options.
1253          * Otherwise, use it to store the options.
1254          */
1255         if (exthdrs->ip6e_hbh == 0) {
1256                 mopt = m_get(M_NOWAIT, MT_DATA);
1257                 if (mopt == NULL)
1258                         return (ENOBUFS);
1259                 mopt->m_len = JUMBOOPTLEN;
1260                 optbuf = mtod(mopt, u_char *);
1261                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1262                 exthdrs->ip6e_hbh = mopt;
1263         } else {
1264                 struct ip6_hbh *hbh;
1265
1266                 mopt = exthdrs->ip6e_hbh;
1267                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1268                         /*
1269                          * XXX assumption:
1270                          * - exthdrs->ip6e_hbh is not referenced from places
1271                          *   other than exthdrs.
1272                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1273                          */
1274                         int oldoptlen = mopt->m_len;
1275                         struct mbuf *n;
1276
1277                         /*
1278                          * XXX: give up if the whole (new) hbh header does
1279                          * not fit even in an mbuf cluster.
1280                          */
1281                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1282                                 return (ENOBUFS);
1283
1284                         /*
1285                          * As a consequence, we must always prepare a cluster
1286                          * at this point.
1287                          */
1288                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1289                         if (n == NULL)
1290                                 return (ENOBUFS);
1291                         n->m_len = oldoptlen + JUMBOOPTLEN;
1292                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1293                             oldoptlen);
1294                         optbuf = mtod(n, caddr_t) + oldoptlen;
1295                         m_freem(mopt);
1296                         mopt = exthdrs->ip6e_hbh = n;
1297                 } else {
1298                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1299                         mopt->m_len += JUMBOOPTLEN;
1300                 }
1301                 optbuf[0] = IP6OPT_PADN;
1302                 optbuf[1] = 1;
1303
1304                 /*
1305                  * Adjust the header length according to the pad and
1306                  * the jumbo payload option.
1307                  */
1308                 hbh = mtod(mopt, struct ip6_hbh *);
1309                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1310         }
1311
1312         /* fill in the option. */
1313         optbuf[2] = IP6OPT_JUMBO;
1314         optbuf[3] = 4;
1315         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1316         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1317
1318         /* finally, adjust the packet header length */
1319         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1320
1321         return (0);
1322 #undef JUMBOOPTLEN
1323 }
1324
1325 /*
1326  * Insert fragment header and copy unfragmentable header portions.
1327  */
1328 static int
1329 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1330     struct ip6_frag **frghdrp)
1331 {
1332         struct mbuf *n, *mlast;
1333
1334         if (hlen > sizeof(struct ip6_hdr)) {
1335                 n = m_copym(m0, sizeof(struct ip6_hdr),
1336                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1337                 if (n == 0)
1338                         return (ENOBUFS);
1339                 m->m_next = n;
1340         } else
1341                 n = m;
1342
1343         /* Search for the last mbuf of unfragmentable part. */
1344         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1345                 ;
1346
1347         if ((mlast->m_flags & M_EXT) == 0 &&
1348             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1349                 /* use the trailing space of the last mbuf for the fragment hdr */
1350                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1351                     mlast->m_len);
1352                 mlast->m_len += sizeof(struct ip6_frag);
1353                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1354         } else {
1355                 /* allocate a new mbuf for the fragment header */
1356                 struct mbuf *mfrg;
1357
1358                 mfrg = m_get(M_NOWAIT, MT_DATA);
1359                 if (mfrg == NULL)
1360                         return (ENOBUFS);
1361                 mfrg->m_len = sizeof(struct ip6_frag);
1362                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1363                 mlast->m_next = mfrg;
1364         }
1365
1366         return (0);
1367 }
1368
1369 static int
1370 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1371     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1372     int *alwaysfragp, u_int fibnum)
1373 {
1374         u_int32_t mtu = 0;
1375         int alwaysfrag = 0;
1376         int error = 0;
1377
1378         if (ro_pmtu != ro) {
1379                 /* The first hop and the final destination may differ. */
1380                 struct sockaddr_in6 *sa6_dst =
1381                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1382                 if (ro_pmtu->ro_rt &&
1383                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1384                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1385                         RTFREE(ro_pmtu->ro_rt);
1386                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1387                 }
1388                 if (ro_pmtu->ro_rt == NULL) {
1389                         bzero(sa6_dst, sizeof(*sa6_dst));
1390                         sa6_dst->sin6_family = AF_INET6;
1391                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1392                         sa6_dst->sin6_addr = *dst;
1393
1394                         in6_rtalloc(ro_pmtu, fibnum);
1395                 }
1396         }
1397         if (ro_pmtu->ro_rt) {
1398                 u_int32_t ifmtu;
1399                 struct in_conninfo inc;
1400
1401                 bzero(&inc, sizeof(inc));
1402                 inc.inc_flags |= INC_ISIPV6;
1403                 inc.inc6_faddr = *dst;
1404
1405                 if (ifp == NULL)
1406                         ifp = ro_pmtu->ro_rt->rt_ifp;
1407                 ifmtu = IN6_LINKMTU(ifp);
1408                 mtu = tcp_hc_getmtu(&inc);
1409                 if (mtu)
1410                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1411                 else
1412                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1413                 if (mtu == 0)
1414                         mtu = ifmtu;
1415                 else if (mtu < IPV6_MMTU) {
1416                         /*
1417                          * RFC2460 section 5, last paragraph:
1418                          * if we record ICMPv6 too big message with
1419                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1420                          * or smaller, with framgent header attached.
1421                          * (fragment header is needed regardless from the
1422                          * packet size, for translators to identify packets)
1423                          */
1424                         alwaysfrag = 1;
1425                         mtu = IPV6_MMTU;
1426                 } else if (mtu > ifmtu) {
1427                         /*
1428                          * The MTU on the route is larger than the MTU on
1429                          * the interface!  This shouldn't happen, unless the
1430                          * MTU of the interface has been changed after the
1431                          * interface was brought up.  Change the MTU in the
1432                          * route to match the interface MTU (as long as the
1433                          * field isn't locked).
1434                          */
1435                         mtu = ifmtu;
1436                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1437                 }
1438         } else if (ifp) {
1439                 mtu = IN6_LINKMTU(ifp);
1440         } else
1441                 error = EHOSTUNREACH; /* XXX */
1442
1443         *mtup = mtu;
1444         if (alwaysfragp)
1445                 *alwaysfragp = alwaysfrag;
1446         return (error);
1447 }
1448
1449 /*
1450  * IP6 socket option processing.
1451  */
1452 int
1453 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1454 {
1455         int optdatalen, uproto;
1456         void *optdata;
1457         struct inpcb *in6p = sotoinpcb(so);
1458         int error, optval;
1459         int level, op, optname;
1460         int optlen;
1461         struct thread *td;
1462
1463         level = sopt->sopt_level;
1464         op = sopt->sopt_dir;
1465         optname = sopt->sopt_name;
1466         optlen = sopt->sopt_valsize;
1467         td = sopt->sopt_td;
1468         error = 0;
1469         optval = 0;
1470         uproto = (int)so->so_proto->pr_protocol;
1471
1472         if (level != IPPROTO_IPV6) {
1473                 error = EINVAL;
1474
1475                 if (sopt->sopt_level == SOL_SOCKET &&
1476                     sopt->sopt_dir == SOPT_SET) {
1477                         switch (sopt->sopt_name) {
1478                         case SO_REUSEADDR:
1479                                 INP_WLOCK(in6p);
1480                                 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1481                                         if ((so->so_options &
1482                                             (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1483                                                 in6p->inp_flags2 |= INP_REUSEPORT;
1484                                         else
1485                                                 in6p->inp_flags2 &= ~INP_REUSEPORT;
1486                                 }
1487                                 INP_WUNLOCK(in6p);
1488                                 error = 0;
1489                                 break;
1490                         case SO_REUSEPORT:
1491                                 INP_WLOCK(in6p);
1492                                 if ((so->so_options & SO_REUSEPORT) != 0)
1493                                         in6p->inp_flags2 |= INP_REUSEPORT;
1494                                 else
1495                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1496                                 INP_WUNLOCK(in6p);
1497                                 error = 0;
1498                                 break;
1499                         case SO_SETFIB:
1500                                 INP_WLOCK(in6p);
1501                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1502                                 INP_WUNLOCK(in6p);
1503                                 error = 0;
1504                                 break;
1505                         default:
1506                                 break;
1507                         }
1508                 }
1509         } else {                /* level == IPPROTO_IPV6 */
1510                 switch (op) {
1511
1512                 case SOPT_SET:
1513                         switch (optname) {
1514                         case IPV6_2292PKTOPTIONS:
1515 #ifdef IPV6_PKTOPTIONS
1516                         case IPV6_PKTOPTIONS:
1517 #endif
1518                         {
1519                                 struct mbuf *m;
1520
1521                                 error = soopt_getm(sopt, &m); /* XXX */
1522                                 if (error != 0)
1523                                         break;
1524                                 error = soopt_mcopyin(sopt, m); /* XXX */
1525                                 if (error != 0)
1526                                         break;
1527                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1528                                                     m, so, sopt);
1529                                 m_freem(m); /* XXX */
1530                                 break;
1531                         }
1532
1533                         /*
1534                          * Use of some Hop-by-Hop options or some
1535                          * Destination options, might require special
1536                          * privilege.  That is, normal applications
1537                          * (without special privilege) might be forbidden
1538                          * from setting certain options in outgoing packets,
1539                          * and might never see certain options in received
1540                          * packets. [RFC 2292 Section 6]
1541                          * KAME specific note:
1542                          *  KAME prevents non-privileged users from sending or
1543                          *  receiving ANY hbh/dst options in order to avoid
1544                          *  overhead of parsing options in the kernel.
1545                          */
1546                         case IPV6_RECVHOPOPTS:
1547                         case IPV6_RECVDSTOPTS:
1548                         case IPV6_RECVRTHDRDSTOPTS:
1549                                 if (td != NULL) {
1550                                         error = priv_check(td,
1551                                             PRIV_NETINET_SETHDROPTS);
1552                                         if (error)
1553                                                 break;
1554                                 }
1555                                 /* FALLTHROUGH */
1556                         case IPV6_UNICAST_HOPS:
1557                         case IPV6_HOPLIMIT:
1558                         case IPV6_FAITH:
1559
1560                         case IPV6_RECVPKTINFO:
1561                         case IPV6_RECVHOPLIMIT:
1562                         case IPV6_RECVRTHDR:
1563                         case IPV6_RECVPATHMTU:
1564                         case IPV6_RECVTCLASS:
1565                         case IPV6_V6ONLY:
1566                         case IPV6_AUTOFLOWLABEL:
1567                         case IPV6_BINDANY:
1568                                 if (optname == IPV6_BINDANY && td != NULL) {
1569                                         error = priv_check(td,
1570                                             PRIV_NETINET_BINDANY);
1571                                         if (error)
1572                                                 break;
1573                                 }
1574
1575                                 if (optlen != sizeof(int)) {
1576                                         error = EINVAL;
1577                                         break;
1578                                 }
1579                                 error = sooptcopyin(sopt, &optval,
1580                                         sizeof optval, sizeof optval);
1581                                 if (error)
1582                                         break;
1583                                 switch (optname) {
1584
1585                                 case IPV6_UNICAST_HOPS:
1586                                         if (optval < -1 || optval >= 256)
1587                                                 error = EINVAL;
1588                                         else {
1589                                                 /* -1 = kernel default */
1590                                                 in6p->in6p_hops = optval;
1591                                                 if ((in6p->inp_vflag &
1592                                                      INP_IPV4) != 0)
1593                                                         in6p->inp_ip_ttl = optval;
1594                                         }
1595                                         break;
1596 #define OPTSET(bit) \
1597 do { \
1598         INP_WLOCK(in6p); \
1599         if (optval) \
1600                 in6p->inp_flags |= (bit); \
1601         else \
1602                 in6p->inp_flags &= ~(bit); \
1603         INP_WUNLOCK(in6p); \
1604 } while (/*CONSTCOND*/ 0)
1605 #define OPTSET2292(bit) \
1606 do { \
1607         INP_WLOCK(in6p); \
1608         in6p->inp_flags |= IN6P_RFC2292; \
1609         if (optval) \
1610                 in6p->inp_flags |= (bit); \
1611         else \
1612                 in6p->inp_flags &= ~(bit); \
1613         INP_WUNLOCK(in6p); \
1614 } while (/*CONSTCOND*/ 0)
1615 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1616
1617                                 case IPV6_RECVPKTINFO:
1618                                         /* cannot mix with RFC2292 */
1619                                         if (OPTBIT(IN6P_RFC2292)) {
1620                                                 error = EINVAL;
1621                                                 break;
1622                                         }
1623                                         OPTSET(IN6P_PKTINFO);
1624                                         break;
1625
1626                                 case IPV6_HOPLIMIT:
1627                                 {
1628                                         struct ip6_pktopts **optp;
1629
1630                                         /* cannot mix with RFC2292 */
1631                                         if (OPTBIT(IN6P_RFC2292)) {
1632                                                 error = EINVAL;
1633                                                 break;
1634                                         }
1635                                         optp = &in6p->in6p_outputopts;
1636                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1637                                             (u_char *)&optval, sizeof(optval),
1638                                             optp, (td != NULL) ? td->td_ucred :
1639                                             NULL, uproto);
1640                                         break;
1641                                 }
1642
1643                                 case IPV6_RECVHOPLIMIT:
1644                                         /* cannot mix with RFC2292 */
1645                                         if (OPTBIT(IN6P_RFC2292)) {
1646                                                 error = EINVAL;
1647                                                 break;
1648                                         }
1649                                         OPTSET(IN6P_HOPLIMIT);
1650                                         break;
1651
1652                                 case IPV6_RECVHOPOPTS:
1653                                         /* cannot mix with RFC2292 */
1654                                         if (OPTBIT(IN6P_RFC2292)) {
1655                                                 error = EINVAL;
1656                                                 break;
1657                                         }
1658                                         OPTSET(IN6P_HOPOPTS);
1659                                         break;
1660
1661                                 case IPV6_RECVDSTOPTS:
1662                                         /* cannot mix with RFC2292 */
1663                                         if (OPTBIT(IN6P_RFC2292)) {
1664                                                 error = EINVAL;
1665                                                 break;
1666                                         }
1667                                         OPTSET(IN6P_DSTOPTS);
1668                                         break;
1669
1670                                 case IPV6_RECVRTHDRDSTOPTS:
1671                                         /* cannot mix with RFC2292 */
1672                                         if (OPTBIT(IN6P_RFC2292)) {
1673                                                 error = EINVAL;
1674                                                 break;
1675                                         }
1676                                         OPTSET(IN6P_RTHDRDSTOPTS);
1677                                         break;
1678
1679                                 case IPV6_RECVRTHDR:
1680                                         /* cannot mix with RFC2292 */
1681                                         if (OPTBIT(IN6P_RFC2292)) {
1682                                                 error = EINVAL;
1683                                                 break;
1684                                         }
1685                                         OPTSET(IN6P_RTHDR);
1686                                         break;
1687
1688                                 case IPV6_FAITH:
1689                                         OPTSET(INP_FAITH);
1690                                         break;
1691
1692                                 case IPV6_RECVPATHMTU:
1693                                         /*
1694                                          * We ignore this option for TCP
1695                                          * sockets.
1696                                          * (RFC3542 leaves this case
1697                                          * unspecified.)
1698                                          */
1699                                         if (uproto != IPPROTO_TCP)
1700                                                 OPTSET(IN6P_MTU);
1701                                         break;
1702
1703                                 case IPV6_V6ONLY:
1704                                         /*
1705                                          * make setsockopt(IPV6_V6ONLY)
1706                                          * available only prior to bind(2).
1707                                          * see ipng mailing list, Jun 22 2001.
1708                                          */
1709                                         if (in6p->inp_lport ||
1710                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1711                                                 error = EINVAL;
1712                                                 break;
1713                                         }
1714                                         OPTSET(IN6P_IPV6_V6ONLY);
1715                                         if (optval)
1716                                                 in6p->inp_vflag &= ~INP_IPV4;
1717                                         else
1718                                                 in6p->inp_vflag |= INP_IPV4;
1719                                         break;
1720                                 case IPV6_RECVTCLASS:
1721                                         /* cannot mix with RFC2292 XXX */
1722                                         if (OPTBIT(IN6P_RFC2292)) {
1723                                                 error = EINVAL;
1724                                                 break;
1725                                         }
1726                                         OPTSET(IN6P_TCLASS);
1727                                         break;
1728                                 case IPV6_AUTOFLOWLABEL:
1729                                         OPTSET(IN6P_AUTOFLOWLABEL);
1730                                         break;
1731
1732                                 case IPV6_BINDANY:
1733                                         OPTSET(INP_BINDANY);
1734                                         break;
1735                                 }
1736                                 break;
1737
1738                         case IPV6_TCLASS:
1739                         case IPV6_DONTFRAG:
1740                         case IPV6_USE_MIN_MTU:
1741                         case IPV6_PREFER_TEMPADDR:
1742                                 if (optlen != sizeof(optval)) {
1743                                         error = EINVAL;
1744                                         break;
1745                                 }
1746                                 error = sooptcopyin(sopt, &optval,
1747                                         sizeof optval, sizeof optval);
1748                                 if (error)
1749                                         break;
1750                                 {
1751                                         struct ip6_pktopts **optp;
1752                                         optp = &in6p->in6p_outputopts;
1753                                         error = ip6_pcbopt(optname,
1754                                             (u_char *)&optval, sizeof(optval),
1755                                             optp, (td != NULL) ? td->td_ucred :
1756                                             NULL, uproto);
1757                                         break;
1758                                 }
1759
1760                         case IPV6_2292PKTINFO:
1761                         case IPV6_2292HOPLIMIT:
1762                         case IPV6_2292HOPOPTS:
1763                         case IPV6_2292DSTOPTS:
1764                         case IPV6_2292RTHDR:
1765                                 /* RFC 2292 */
1766                                 if (optlen != sizeof(int)) {
1767                                         error = EINVAL;
1768                                         break;
1769                                 }
1770                                 error = sooptcopyin(sopt, &optval,
1771                                         sizeof optval, sizeof optval);
1772                                 if (error)
1773                                         break;
1774                                 switch (optname) {
1775                                 case IPV6_2292PKTINFO:
1776                                         OPTSET2292(IN6P_PKTINFO);
1777                                         break;
1778                                 case IPV6_2292HOPLIMIT:
1779                                         OPTSET2292(IN6P_HOPLIMIT);
1780                                         break;
1781                                 case IPV6_2292HOPOPTS:
1782                                         /*
1783                                          * Check super-user privilege.
1784                                          * See comments for IPV6_RECVHOPOPTS.
1785                                          */
1786                                         if (td != NULL) {
1787                                                 error = priv_check(td,
1788                                                     PRIV_NETINET_SETHDROPTS);
1789                                                 if (error)
1790                                                         return (error);
1791                                         }
1792                                         OPTSET2292(IN6P_HOPOPTS);
1793                                         break;
1794                                 case IPV6_2292DSTOPTS:
1795                                         if (td != NULL) {
1796                                                 error = priv_check(td,
1797                                                     PRIV_NETINET_SETHDROPTS);
1798                                                 if (error)
1799                                                         return (error);
1800                                         }
1801                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1802                                         break;
1803                                 case IPV6_2292RTHDR:
1804                                         OPTSET2292(IN6P_RTHDR);
1805                                         break;
1806                                 }
1807                                 break;
1808                         case IPV6_PKTINFO:
1809                         case IPV6_HOPOPTS:
1810                         case IPV6_RTHDR:
1811                         case IPV6_DSTOPTS:
1812                         case IPV6_RTHDRDSTOPTS:
1813                         case IPV6_NEXTHOP:
1814                         {
1815                                 /* new advanced API (RFC3542) */
1816                                 u_char *optbuf;
1817                                 u_char optbuf_storage[MCLBYTES];
1818                                 int optlen;
1819                                 struct ip6_pktopts **optp;
1820
1821                                 /* cannot mix with RFC2292 */
1822                                 if (OPTBIT(IN6P_RFC2292)) {
1823                                         error = EINVAL;
1824                                         break;
1825                                 }
1826
1827                                 /*
1828                                  * We only ensure valsize is not too large
1829                                  * here.  Further validation will be done
1830                                  * later.
1831                                  */
1832                                 error = sooptcopyin(sopt, optbuf_storage,
1833                                     sizeof(optbuf_storage), 0);
1834                                 if (error)
1835                                         break;
1836                                 optlen = sopt->sopt_valsize;
1837                                 optbuf = optbuf_storage;
1838                                 optp = &in6p->in6p_outputopts;
1839                                 error = ip6_pcbopt(optname, optbuf, optlen,
1840                                     optp, (td != NULL) ? td->td_ucred : NULL,
1841                                     uproto);
1842                                 break;
1843                         }
1844 #undef OPTSET
1845
1846                         case IPV6_MULTICAST_IF:
1847                         case IPV6_MULTICAST_HOPS:
1848                         case IPV6_MULTICAST_LOOP:
1849                         case IPV6_JOIN_GROUP:
1850                         case IPV6_LEAVE_GROUP:
1851                         case IPV6_MSFILTER:
1852                         case MCAST_BLOCK_SOURCE:
1853                         case MCAST_UNBLOCK_SOURCE:
1854                         case MCAST_JOIN_GROUP:
1855                         case MCAST_LEAVE_GROUP:
1856                         case MCAST_JOIN_SOURCE_GROUP:
1857                         case MCAST_LEAVE_SOURCE_GROUP:
1858                                 error = ip6_setmoptions(in6p, sopt);
1859                                 break;
1860
1861                         case IPV6_PORTRANGE:
1862                                 error = sooptcopyin(sopt, &optval,
1863                                     sizeof optval, sizeof optval);
1864                                 if (error)
1865                                         break;
1866
1867                                 INP_WLOCK(in6p);
1868                                 switch (optval) {
1869                                 case IPV6_PORTRANGE_DEFAULT:
1870                                         in6p->inp_flags &= ~(INP_LOWPORT);
1871                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1872                                         break;
1873
1874                                 case IPV6_PORTRANGE_HIGH:
1875                                         in6p->inp_flags &= ~(INP_LOWPORT);
1876                                         in6p->inp_flags |= INP_HIGHPORT;
1877                                         break;
1878
1879                                 case IPV6_PORTRANGE_LOW:
1880                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1881                                         in6p->inp_flags |= INP_LOWPORT;
1882                                         break;
1883
1884                                 default:
1885                                         error = EINVAL;
1886                                         break;
1887                                 }
1888                                 INP_WUNLOCK(in6p);
1889                                 break;
1890
1891 #ifdef IPSEC
1892                         case IPV6_IPSEC_POLICY:
1893                         {
1894                                 caddr_t req;
1895                                 struct mbuf *m;
1896
1897                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1898                                         break;
1899                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1900                                         break;
1901                                 req = mtod(m, caddr_t);
1902                                 error = ipsec_set_policy(in6p, optname, req,
1903                                     m->m_len, (sopt->sopt_td != NULL) ?
1904                                     sopt->sopt_td->td_ucred : NULL);
1905                                 m_freem(m);
1906                                 break;
1907                         }
1908 #endif /* IPSEC */
1909
1910                         default:
1911                                 error = ENOPROTOOPT;
1912                                 break;
1913                         }
1914                         break;
1915
1916                 case SOPT_GET:
1917                         switch (optname) {
1918
1919                         case IPV6_2292PKTOPTIONS:
1920 #ifdef IPV6_PKTOPTIONS
1921                         case IPV6_PKTOPTIONS:
1922 #endif
1923                                 /*
1924                                  * RFC3542 (effectively) deprecated the
1925                                  * semantics of the 2292-style pktoptions.
1926                                  * Since it was not reliable in nature (i.e.,
1927                                  * applications had to expect the lack of some
1928                                  * information after all), it would make sense
1929                                  * to simplify this part by always returning
1930                                  * empty data.
1931                                  */
1932                                 sopt->sopt_valsize = 0;
1933                                 break;
1934
1935                         case IPV6_RECVHOPOPTS:
1936                         case IPV6_RECVDSTOPTS:
1937                         case IPV6_RECVRTHDRDSTOPTS:
1938                         case IPV6_UNICAST_HOPS:
1939                         case IPV6_RECVPKTINFO:
1940                         case IPV6_RECVHOPLIMIT:
1941                         case IPV6_RECVRTHDR:
1942                         case IPV6_RECVPATHMTU:
1943
1944                         case IPV6_FAITH:
1945                         case IPV6_V6ONLY:
1946                         case IPV6_PORTRANGE:
1947                         case IPV6_RECVTCLASS:
1948                         case IPV6_AUTOFLOWLABEL:
1949                         case IPV6_BINDANY:
1950                                 switch (optname) {
1951
1952                                 case IPV6_RECVHOPOPTS:
1953                                         optval = OPTBIT(IN6P_HOPOPTS);
1954                                         break;
1955
1956                                 case IPV6_RECVDSTOPTS:
1957                                         optval = OPTBIT(IN6P_DSTOPTS);
1958                                         break;
1959
1960                                 case IPV6_RECVRTHDRDSTOPTS:
1961                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1962                                         break;
1963
1964                                 case IPV6_UNICAST_HOPS:
1965                                         optval = in6p->in6p_hops;
1966                                         break;
1967
1968                                 case IPV6_RECVPKTINFO:
1969                                         optval = OPTBIT(IN6P_PKTINFO);
1970                                         break;
1971
1972                                 case IPV6_RECVHOPLIMIT:
1973                                         optval = OPTBIT(IN6P_HOPLIMIT);
1974                                         break;
1975
1976                                 case IPV6_RECVRTHDR:
1977                                         optval = OPTBIT(IN6P_RTHDR);
1978                                         break;
1979
1980                                 case IPV6_RECVPATHMTU:
1981                                         optval = OPTBIT(IN6P_MTU);
1982                                         break;
1983
1984                                 case IPV6_FAITH:
1985                                         optval = OPTBIT(INP_FAITH);
1986                                         break;
1987
1988                                 case IPV6_V6ONLY:
1989                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1990                                         break;
1991
1992                                 case IPV6_PORTRANGE:
1993                                     {
1994                                         int flags;
1995                                         flags = in6p->inp_flags;
1996                                         if (flags & INP_HIGHPORT)
1997                                                 optval = IPV6_PORTRANGE_HIGH;
1998                                         else if (flags & INP_LOWPORT)
1999                                                 optval = IPV6_PORTRANGE_LOW;
2000                                         else
2001                                                 optval = 0;
2002                                         break;
2003                                     }
2004                                 case IPV6_RECVTCLASS:
2005                                         optval = OPTBIT(IN6P_TCLASS);
2006                                         break;
2007
2008                                 case IPV6_AUTOFLOWLABEL:
2009                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2010                                         break;
2011
2012                                 case IPV6_BINDANY:
2013                                         optval = OPTBIT(INP_BINDANY);
2014                                         break;
2015                                 }
2016                                 if (error)
2017                                         break;
2018                                 error = sooptcopyout(sopt, &optval,
2019                                         sizeof optval);
2020                                 break;
2021
2022                         case IPV6_PATHMTU:
2023                         {
2024                                 u_long pmtu = 0;
2025                                 struct ip6_mtuinfo mtuinfo;
2026                                 struct route_in6 sro;
2027
2028                                 bzero(&sro, sizeof(sro));
2029
2030                                 if (!(so->so_state & SS_ISCONNECTED))
2031                                         return (ENOTCONN);
2032                                 /*
2033                                  * XXX: we dot not consider the case of source
2034                                  * routing, or optional information to specify
2035                                  * the outgoing interface.
2036                                  */
2037                                 error = ip6_getpmtu(&sro, NULL, NULL,
2038                                     &in6p->in6p_faddr, &pmtu, NULL,
2039                                     so->so_fibnum);
2040                                 if (sro.ro_rt)
2041                                         RTFREE(sro.ro_rt);
2042                                 if (error)
2043                                         break;
2044                                 if (pmtu > IPV6_MAXPACKET)
2045                                         pmtu = IPV6_MAXPACKET;
2046
2047                                 bzero(&mtuinfo, sizeof(mtuinfo));
2048                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2049                                 optdata = (void *)&mtuinfo;
2050                                 optdatalen = sizeof(mtuinfo);
2051                                 error = sooptcopyout(sopt, optdata,
2052                                     optdatalen);
2053                                 break;
2054                         }
2055
2056                         case IPV6_2292PKTINFO:
2057                         case IPV6_2292HOPLIMIT:
2058                         case IPV6_2292HOPOPTS:
2059                         case IPV6_2292RTHDR:
2060                         case IPV6_2292DSTOPTS:
2061                                 switch (optname) {
2062                                 case IPV6_2292PKTINFO:
2063                                         optval = OPTBIT(IN6P_PKTINFO);
2064                                         break;
2065                                 case IPV6_2292HOPLIMIT:
2066                                         optval = OPTBIT(IN6P_HOPLIMIT);
2067                                         break;
2068                                 case IPV6_2292HOPOPTS:
2069                                         optval = OPTBIT(IN6P_HOPOPTS);
2070                                         break;
2071                                 case IPV6_2292RTHDR:
2072                                         optval = OPTBIT(IN6P_RTHDR);
2073                                         break;
2074                                 case IPV6_2292DSTOPTS:
2075                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2076                                         break;
2077                                 }
2078                                 error = sooptcopyout(sopt, &optval,
2079                                     sizeof optval);
2080                                 break;
2081                         case IPV6_PKTINFO:
2082                         case IPV6_HOPOPTS:
2083                         case IPV6_RTHDR:
2084                         case IPV6_DSTOPTS:
2085                         case IPV6_RTHDRDSTOPTS:
2086                         case IPV6_NEXTHOP:
2087                         case IPV6_TCLASS:
2088                         case IPV6_DONTFRAG:
2089                         case IPV6_USE_MIN_MTU:
2090                         case IPV6_PREFER_TEMPADDR:
2091                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2092                                     optname, sopt);
2093                                 break;
2094
2095                         case IPV6_MULTICAST_IF:
2096                         case IPV6_MULTICAST_HOPS:
2097                         case IPV6_MULTICAST_LOOP:
2098                         case IPV6_MSFILTER:
2099                                 error = ip6_getmoptions(in6p, sopt);
2100                                 break;
2101
2102 #ifdef IPSEC
2103                         case IPV6_IPSEC_POLICY:
2104                           {
2105                                 caddr_t req = NULL;
2106                                 size_t len = 0;
2107                                 struct mbuf *m = NULL;
2108                                 struct mbuf **mp = &m;
2109                                 size_t ovalsize = sopt->sopt_valsize;
2110                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2111
2112                                 error = soopt_getm(sopt, &m); /* XXX */
2113                                 if (error != 0)
2114                                         break;
2115                                 error = soopt_mcopyin(sopt, m); /* XXX */
2116                                 if (error != 0)
2117                                         break;
2118                                 sopt->sopt_valsize = ovalsize;
2119                                 sopt->sopt_val = oval;
2120                                 if (m) {
2121                                         req = mtod(m, caddr_t);
2122                                         len = m->m_len;
2123                                 }
2124                                 error = ipsec_get_policy(in6p, req, len, mp);
2125                                 if (error == 0)
2126                                         error = soopt_mcopyout(sopt, m); /* XXX */
2127                                 if (error == 0 && m)
2128                                         m_freem(m);
2129                                 break;
2130                           }
2131 #endif /* IPSEC */
2132
2133                         default:
2134                                 error = ENOPROTOOPT;
2135                                 break;
2136                         }
2137                         break;
2138                 }
2139         }
2140         return (error);
2141 }
2142
2143 int
2144 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2145 {
2146         int error = 0, optval, optlen;
2147         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2148         struct inpcb *in6p = sotoinpcb(so);
2149         int level, op, optname;
2150
2151         level = sopt->sopt_level;
2152         op = sopt->sopt_dir;
2153         optname = sopt->sopt_name;
2154         optlen = sopt->sopt_valsize;
2155
2156         if (level != IPPROTO_IPV6) {
2157                 return (EINVAL);
2158         }
2159
2160         switch (optname) {
2161         case IPV6_CHECKSUM:
2162                 /*
2163                  * For ICMPv6 sockets, no modification allowed for checksum
2164                  * offset, permit "no change" values to help existing apps.
2165                  *
2166                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2167                  * for an ICMPv6 socket will fail."
2168                  * The current behavior does not meet RFC3542.
2169                  */
2170                 switch (op) {
2171                 case SOPT_SET:
2172                         if (optlen != sizeof(int)) {
2173                                 error = EINVAL;
2174                                 break;
2175                         }
2176                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2177                                             sizeof(optval));
2178                         if (error)
2179                                 break;
2180                         if ((optval % 2) != 0) {
2181                                 /* the API assumes even offset values */
2182                                 error = EINVAL;
2183                         } else if (so->so_proto->pr_protocol ==
2184                             IPPROTO_ICMPV6) {
2185                                 if (optval != icmp6off)
2186                                         error = EINVAL;
2187                         } else
2188                                 in6p->in6p_cksum = optval;
2189                         break;
2190
2191                 case SOPT_GET:
2192                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2193                                 optval = icmp6off;
2194                         else
2195                                 optval = in6p->in6p_cksum;
2196
2197                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2198                         break;
2199
2200                 default:
2201                         error = EINVAL;
2202                         break;
2203                 }
2204                 break;
2205
2206         default:
2207                 error = ENOPROTOOPT;
2208                 break;
2209         }
2210
2211         return (error);
2212 }
2213
2214 /*
2215  * Set up IP6 options in pcb for insertion in output packets or
2216  * specifying behavior of outgoing packets.
2217  */
2218 static int
2219 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2220     struct socket *so, struct sockopt *sopt)
2221 {
2222         struct ip6_pktopts *opt = *pktopt;
2223         int error = 0;
2224         struct thread *td = sopt->sopt_td;
2225
2226         /* turn off any old options. */
2227         if (opt) {
2228 #ifdef DIAGNOSTIC
2229                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2230                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2231                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2232                         printf("ip6_pcbopts: all specified options are cleared.\n");
2233 #endif
2234                 ip6_clearpktopts(opt, -1);
2235         } else
2236                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2237         *pktopt = NULL;
2238
2239         if (!m || m->m_len == 0) {
2240                 /*
2241                  * Only turning off any previous options, regardless of
2242                  * whether the opt is just created or given.
2243                  */
2244                 free(opt, M_IP6OPT);
2245                 return (0);
2246         }
2247
2248         /*  set options specified by user. */
2249         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2250             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2251                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2252                 free(opt, M_IP6OPT);
2253                 return (error);
2254         }
2255         *pktopt = opt;
2256         return (0);
2257 }
2258
2259 /*
2260  * initialize ip6_pktopts.  beware that there are non-zero default values in
2261  * the struct.
2262  */
2263 void
2264 ip6_initpktopts(struct ip6_pktopts *opt)
2265 {
2266
2267         bzero(opt, sizeof(*opt));
2268         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2269         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2270         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2271         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2272 }
2273
2274 static int
2275 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2276     struct ucred *cred, int uproto)
2277 {
2278         struct ip6_pktopts *opt;
2279
2280         if (*pktopt == NULL) {
2281                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2282                     M_WAITOK);
2283                 ip6_initpktopts(*pktopt);
2284         }
2285         opt = *pktopt;
2286
2287         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2288 }
2289
2290 static int
2291 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2292 {
2293         void *optdata = NULL;
2294         int optdatalen = 0;
2295         struct ip6_ext *ip6e;
2296         int error = 0;
2297         struct in6_pktinfo null_pktinfo;
2298         int deftclass = 0, on;
2299         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2300         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2301
2302         switch (optname) {
2303         case IPV6_PKTINFO:
2304                 if (pktopt && pktopt->ip6po_pktinfo)
2305                         optdata = (void *)pktopt->ip6po_pktinfo;
2306                 else {
2307                         /* XXX: we don't have to do this every time... */
2308                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2309                         optdata = (void *)&null_pktinfo;
2310                 }
2311                 optdatalen = sizeof(struct in6_pktinfo);
2312                 break;
2313         case IPV6_TCLASS:
2314                 if (pktopt && pktopt->ip6po_tclass >= 0)
2315                         optdata = (void *)&pktopt->ip6po_tclass;
2316                 else
2317                         optdata = (void *)&deftclass;
2318                 optdatalen = sizeof(int);
2319                 break;
2320         case IPV6_HOPOPTS:
2321                 if (pktopt && pktopt->ip6po_hbh) {
2322                         optdata = (void *)pktopt->ip6po_hbh;
2323                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2324                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2325                 }
2326                 break;
2327         case IPV6_RTHDR:
2328                 if (pktopt && pktopt->ip6po_rthdr) {
2329                         optdata = (void *)pktopt->ip6po_rthdr;
2330                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2331                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2332                 }
2333                 break;
2334         case IPV6_RTHDRDSTOPTS:
2335                 if (pktopt && pktopt->ip6po_dest1) {
2336                         optdata = (void *)pktopt->ip6po_dest1;
2337                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2338                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2339                 }
2340                 break;
2341         case IPV6_DSTOPTS:
2342                 if (pktopt && pktopt->ip6po_dest2) {
2343                         optdata = (void *)pktopt->ip6po_dest2;
2344                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2345                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2346                 }
2347                 break;
2348         case IPV6_NEXTHOP:
2349                 if (pktopt && pktopt->ip6po_nexthop) {
2350                         optdata = (void *)pktopt->ip6po_nexthop;
2351                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2352                 }
2353                 break;
2354         case IPV6_USE_MIN_MTU:
2355                 if (pktopt)
2356                         optdata = (void *)&pktopt->ip6po_minmtu;
2357                 else
2358                         optdata = (void *)&defminmtu;
2359                 optdatalen = sizeof(int);
2360                 break;
2361         case IPV6_DONTFRAG:
2362                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2363                         on = 1;
2364                 else
2365                         on = 0;
2366                 optdata = (void *)&on;
2367                 optdatalen = sizeof(on);
2368                 break;
2369         case IPV6_PREFER_TEMPADDR:
2370                 if (pktopt)
2371                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2372                 else
2373                         optdata = (void *)&defpreftemp;
2374                 optdatalen = sizeof(int);
2375                 break;
2376         default:                /* should not happen */
2377 #ifdef DIAGNOSTIC
2378                 panic("ip6_getpcbopt: unexpected option\n");
2379 #endif
2380                 return (ENOPROTOOPT);
2381         }
2382
2383         error = sooptcopyout(sopt, optdata, optdatalen);
2384
2385         return (error);
2386 }
2387
2388 void
2389 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2390 {
2391         if (pktopt == NULL)
2392                 return;
2393
2394         if (optname == -1 || optname == IPV6_PKTINFO) {
2395                 if (pktopt->ip6po_pktinfo)
2396                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2397                 pktopt->ip6po_pktinfo = NULL;
2398         }
2399         if (optname == -1 || optname == IPV6_HOPLIMIT)
2400                 pktopt->ip6po_hlim = -1;
2401         if (optname == -1 || optname == IPV6_TCLASS)
2402                 pktopt->ip6po_tclass = -1;
2403         if (optname == -1 || optname == IPV6_NEXTHOP) {
2404                 if (pktopt->ip6po_nextroute.ro_rt) {
2405                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2406                         pktopt->ip6po_nextroute.ro_rt = NULL;
2407                 }
2408                 if (pktopt->ip6po_nexthop)
2409                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2410                 pktopt->ip6po_nexthop = NULL;
2411         }
2412         if (optname == -1 || optname == IPV6_HOPOPTS) {
2413                 if (pktopt->ip6po_hbh)
2414                         free(pktopt->ip6po_hbh, M_IP6OPT);
2415                 pktopt->ip6po_hbh = NULL;
2416         }
2417         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2418                 if (pktopt->ip6po_dest1)
2419                         free(pktopt->ip6po_dest1, M_IP6OPT);
2420                 pktopt->ip6po_dest1 = NULL;
2421         }
2422         if (optname == -1 || optname == IPV6_RTHDR) {
2423                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2424                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2425                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2426                 if (pktopt->ip6po_route.ro_rt) {
2427                         RTFREE(pktopt->ip6po_route.ro_rt);
2428                         pktopt->ip6po_route.ro_rt = NULL;
2429                 }
2430         }
2431         if (optname == -1 || optname == IPV6_DSTOPTS) {
2432                 if (pktopt->ip6po_dest2)
2433                         free(pktopt->ip6po_dest2, M_IP6OPT);
2434                 pktopt->ip6po_dest2 = NULL;
2435         }
2436 }
2437
2438 #define PKTOPT_EXTHDRCPY(type) \
2439 do {\
2440         if (src->type) {\
2441                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2442                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2443                 if (dst->type == NULL && canwait == M_NOWAIT)\
2444                         goto bad;\
2445                 bcopy(src->type, dst->type, hlen);\
2446         }\
2447 } while (/*CONSTCOND*/ 0)
2448
2449 static int
2450 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2451 {
2452         if (dst == NULL || src == NULL)  {
2453                 printf("ip6_clearpktopts: invalid argument\n");
2454                 return (EINVAL);
2455         }
2456
2457         dst->ip6po_hlim = src->ip6po_hlim;
2458         dst->ip6po_tclass = src->ip6po_tclass;
2459         dst->ip6po_flags = src->ip6po_flags;
2460         dst->ip6po_minmtu = src->ip6po_minmtu;
2461         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2462         if (src->ip6po_pktinfo) {
2463                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2464                     M_IP6OPT, canwait);
2465                 if (dst->ip6po_pktinfo == NULL)
2466                         goto bad;
2467                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2468         }
2469         if (src->ip6po_nexthop) {
2470                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2471                     M_IP6OPT, canwait);
2472                 if (dst->ip6po_nexthop == NULL)
2473                         goto bad;
2474                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2475                     src->ip6po_nexthop->sa_len);
2476         }
2477         PKTOPT_EXTHDRCPY(ip6po_hbh);
2478         PKTOPT_EXTHDRCPY(ip6po_dest1);
2479         PKTOPT_EXTHDRCPY(ip6po_dest2);
2480         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2481         return (0);
2482
2483   bad:
2484         ip6_clearpktopts(dst, -1);
2485         return (ENOBUFS);
2486 }
2487 #undef PKTOPT_EXTHDRCPY
2488
2489 struct ip6_pktopts *
2490 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2491 {
2492         int error;
2493         struct ip6_pktopts *dst;
2494
2495         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2496         if (dst == NULL)
2497                 return (NULL);
2498         ip6_initpktopts(dst);
2499
2500         if ((error = copypktopts(dst, src, canwait)) != 0) {
2501                 free(dst, M_IP6OPT);
2502                 return (NULL);
2503         }
2504
2505         return (dst);
2506 }
2507
2508 void
2509 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2510 {
2511         if (pktopt == NULL)
2512                 return;
2513
2514         ip6_clearpktopts(pktopt, -1);
2515
2516         free(pktopt, M_IP6OPT);
2517 }
2518
2519 /*
2520  * Set IPv6 outgoing packet options based on advanced API.
2521  */
2522 int
2523 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2524     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2525 {
2526         struct cmsghdr *cm = 0;
2527
2528         if (control == NULL || opt == NULL)
2529                 return (EINVAL);
2530
2531         ip6_initpktopts(opt);
2532         if (stickyopt) {
2533                 int error;
2534
2535                 /*
2536                  * If stickyopt is provided, make a local copy of the options
2537                  * for this particular packet, then override them by ancillary
2538                  * objects.
2539                  * XXX: copypktopts() does not copy the cached route to a next
2540                  * hop (if any).  This is not very good in terms of efficiency,
2541                  * but we can allow this since this option should be rarely
2542                  * used.
2543                  */
2544                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2545                         return (error);
2546         }
2547
2548         /*
2549          * XXX: Currently, we assume all the optional information is stored
2550          * in a single mbuf.
2551          */
2552         if (control->m_next)
2553                 return (EINVAL);
2554
2555         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2556             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2557                 int error;
2558
2559                 if (control->m_len < CMSG_LEN(0))
2560                         return (EINVAL);
2561
2562                 cm = mtod(control, struct cmsghdr *);
2563                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2564                         return (EINVAL);
2565                 if (cm->cmsg_level != IPPROTO_IPV6)
2566                         continue;
2567
2568                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2569                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2570                 if (error)
2571                         return (error);
2572         }
2573
2574         return (0);
2575 }
2576
2577 /*
2578  * Set a particular packet option, as a sticky option or an ancillary data
2579  * item.  "len" can be 0 only when it's a sticky option.
2580  * We have 4 cases of combination of "sticky" and "cmsg":
2581  * "sticky=0, cmsg=0": impossible
2582  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2583  * "sticky=1, cmsg=0": RFC3542 socket option
2584  * "sticky=1, cmsg=1": RFC2292 socket option
2585  */
2586 static int
2587 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2588     struct ucred *cred, int sticky, int cmsg, int uproto)
2589 {
2590         int minmtupolicy, preftemp;
2591         int error;
2592
2593         if (!sticky && !cmsg) {
2594 #ifdef DIAGNOSTIC
2595                 printf("ip6_setpktopt: impossible case\n");
2596 #endif
2597                 return (EINVAL);
2598         }
2599
2600         /*
2601          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2602          * not be specified in the context of RFC3542.  Conversely,
2603          * RFC3542 types should not be specified in the context of RFC2292.
2604          */
2605         if (!cmsg) {
2606                 switch (optname) {
2607                 case IPV6_2292PKTINFO:
2608                 case IPV6_2292HOPLIMIT:
2609                 case IPV6_2292NEXTHOP:
2610                 case IPV6_2292HOPOPTS:
2611                 case IPV6_2292DSTOPTS:
2612                 case IPV6_2292RTHDR:
2613                 case IPV6_2292PKTOPTIONS:
2614                         return (ENOPROTOOPT);
2615                 }
2616         }
2617         if (sticky && cmsg) {
2618                 switch (optname) {
2619                 case IPV6_PKTINFO:
2620                 case IPV6_HOPLIMIT:
2621                 case IPV6_NEXTHOP:
2622                 case IPV6_HOPOPTS:
2623                 case IPV6_DSTOPTS:
2624                 case IPV6_RTHDRDSTOPTS:
2625                 case IPV6_RTHDR:
2626                 case IPV6_USE_MIN_MTU:
2627                 case IPV6_DONTFRAG:
2628                 case IPV6_TCLASS:
2629                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2630                         return (ENOPROTOOPT);
2631                 }
2632         }
2633
2634         switch (optname) {
2635         case IPV6_2292PKTINFO:
2636         case IPV6_PKTINFO:
2637         {
2638                 struct ifnet *ifp = NULL;
2639                 struct in6_pktinfo *pktinfo;
2640
2641                 if (len != sizeof(struct in6_pktinfo))
2642                         return (EINVAL);
2643
2644                 pktinfo = (struct in6_pktinfo *)buf;
2645
2646                 /*
2647                  * An application can clear any sticky IPV6_PKTINFO option by
2648                  * doing a "regular" setsockopt with ipi6_addr being
2649                  * in6addr_any and ipi6_ifindex being zero.
2650                  * [RFC 3542, Section 6]
2651                  */
2652                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2653                     pktinfo->ipi6_ifindex == 0 &&
2654                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2655                         ip6_clearpktopts(opt, optname);
2656                         break;
2657                 }
2658
2659                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2660                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2661                         return (EINVAL);
2662                 }
2663
2664                 /* validate the interface index if specified. */
2665                 if (pktinfo->ipi6_ifindex > V_if_index ||
2666                     pktinfo->ipi6_ifindex < 0) {
2667                          return (ENXIO);
2668                 }
2669                 if (pktinfo->ipi6_ifindex) {
2670                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2671                         if (ifp == NULL)
2672                                 return (ENXIO);
2673                 }
2674
2675                 /*
2676                  * We store the address anyway, and let in6_selectsrc()
2677                  * validate the specified address.  This is because ipi6_addr
2678                  * may not have enough information about its scope zone, and
2679                  * we may need additional information (such as outgoing
2680                  * interface or the scope zone of a destination address) to
2681                  * disambiguate the scope.
2682                  * XXX: the delay of the validation may confuse the
2683                  * application when it is used as a sticky option.
2684                  */
2685                 if (opt->ip6po_pktinfo == NULL) {
2686                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2687                             M_IP6OPT, M_NOWAIT);
2688                         if (opt->ip6po_pktinfo == NULL)
2689                                 return (ENOBUFS);
2690                 }
2691                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2692                 break;
2693         }
2694
2695         case IPV6_2292HOPLIMIT:
2696         case IPV6_HOPLIMIT:
2697         {
2698                 int *hlimp;
2699
2700                 /*
2701                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2702                  * to simplify the ordering among hoplimit options.
2703                  */
2704                 if (optname == IPV6_HOPLIMIT && sticky)
2705                         return (ENOPROTOOPT);
2706
2707                 if (len != sizeof(int))
2708                         return (EINVAL);
2709                 hlimp = (int *)buf;
2710                 if (*hlimp < -1 || *hlimp > 255)
2711                         return (EINVAL);
2712
2713                 opt->ip6po_hlim = *hlimp;
2714                 break;
2715         }
2716
2717         case IPV6_TCLASS:
2718         {
2719                 int tclass;
2720
2721                 if (len != sizeof(int))
2722                         return (EINVAL);
2723                 tclass = *(int *)buf;
2724                 if (tclass < -1 || tclass > 255)
2725                         return (EINVAL);
2726
2727                 opt->ip6po_tclass = tclass;
2728                 break;
2729         }
2730
2731         case IPV6_2292NEXTHOP:
2732         case IPV6_NEXTHOP:
2733                 if (cred != NULL) {
2734                         error = priv_check_cred(cred,
2735                             PRIV_NETINET_SETHDROPTS, 0);
2736                         if (error)
2737                                 return (error);
2738                 }
2739
2740                 if (len == 0) { /* just remove the option */
2741                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2742                         break;
2743                 }
2744
2745                 /* check if cmsg_len is large enough for sa_len */
2746                 if (len < sizeof(struct sockaddr) || len < *buf)
2747                         return (EINVAL);
2748
2749                 switch (((struct sockaddr *)buf)->sa_family) {
2750                 case AF_INET6:
2751                 {
2752                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2753                         int error;
2754
2755                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2756                                 return (EINVAL);
2757
2758                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2759                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2760                                 return (EINVAL);
2761                         }
2762                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2763                             != 0) {
2764                                 return (error);
2765                         }
2766                         break;
2767                 }
2768                 case AF_LINK:   /* should eventually be supported */
2769                 default:
2770                         return (EAFNOSUPPORT);
2771                 }
2772
2773                 /* turn off the previous option, then set the new option. */
2774                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2775                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2776                 if (opt->ip6po_nexthop == NULL)
2777                         return (ENOBUFS);
2778                 bcopy(buf, opt->ip6po_nexthop, *buf);
2779                 break;
2780
2781         case IPV6_2292HOPOPTS:
2782         case IPV6_HOPOPTS:
2783         {
2784                 struct ip6_hbh *hbh;
2785                 int hbhlen;
2786
2787                 /*
2788                  * XXX: We don't allow a non-privileged user to set ANY HbH
2789                  * options, since per-option restriction has too much
2790                  * overhead.
2791                  */
2792                 if (cred != NULL) {
2793                         error = priv_check_cred(cred,
2794                             PRIV_NETINET_SETHDROPTS, 0);
2795                         if (error)
2796                                 return (error);
2797                 }
2798
2799                 if (len == 0) {
2800                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2801                         break;  /* just remove the option */
2802                 }
2803
2804                 /* message length validation */
2805                 if (len < sizeof(struct ip6_hbh))
2806                         return (EINVAL);
2807                 hbh = (struct ip6_hbh *)buf;
2808                 hbhlen = (hbh->ip6h_len + 1) << 3;
2809                 if (len != hbhlen)
2810                         return (EINVAL);
2811
2812                 /* turn off the previous option, then set the new option. */
2813                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2814                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2815                 if (opt->ip6po_hbh == NULL)
2816                         return (ENOBUFS);
2817                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2818
2819                 break;
2820         }
2821
2822         case IPV6_2292DSTOPTS:
2823         case IPV6_DSTOPTS:
2824         case IPV6_RTHDRDSTOPTS:
2825         {
2826                 struct ip6_dest *dest, **newdest = NULL;
2827                 int destlen;
2828
2829                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2830                         error = priv_check_cred(cred,
2831                             PRIV_NETINET_SETHDROPTS, 0);
2832                         if (error)
2833                                 return (error);
2834                 }
2835
2836                 if (len == 0) {
2837                         ip6_clearpktopts(opt, optname);
2838                         break;  /* just remove the option */
2839                 }
2840
2841                 /* message length validation */
2842                 if (len < sizeof(struct ip6_dest))
2843                         return (EINVAL);
2844                 dest = (struct ip6_dest *)buf;
2845                 destlen = (dest->ip6d_len + 1) << 3;
2846                 if (len != destlen)
2847                         return (EINVAL);
2848
2849                 /*
2850                  * Determine the position that the destination options header
2851                  * should be inserted; before or after the routing header.
2852                  */
2853                 switch (optname) {
2854                 case IPV6_2292DSTOPTS:
2855                         /*
2856                          * The old advacned API is ambiguous on this point.
2857                          * Our approach is to determine the position based
2858                          * according to the existence of a routing header.
2859                          * Note, however, that this depends on the order of the
2860                          * extension headers in the ancillary data; the 1st
2861                          * part of the destination options header must appear
2862                          * before the routing header in the ancillary data,
2863                          * too.
2864                          * RFC3542 solved the ambiguity by introducing
2865                          * separate ancillary data or option types.
2866                          */
2867                         if (opt->ip6po_rthdr == NULL)
2868                                 newdest = &opt->ip6po_dest1;
2869                         else
2870                                 newdest = &opt->ip6po_dest2;
2871                         break;
2872                 case IPV6_RTHDRDSTOPTS:
2873                         newdest = &opt->ip6po_dest1;
2874                         break;
2875                 case IPV6_DSTOPTS:
2876                         newdest = &opt->ip6po_dest2;
2877                         break;
2878                 }
2879
2880                 /* turn off the previous option, then set the new option. */
2881                 ip6_clearpktopts(opt, optname);
2882                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2883                 if (*newdest == NULL)
2884                         return (ENOBUFS);
2885                 bcopy(dest, *newdest, destlen);
2886
2887                 break;
2888         }
2889
2890         case IPV6_2292RTHDR:
2891         case IPV6_RTHDR:
2892         {
2893                 struct ip6_rthdr *rth;
2894                 int rthlen;
2895
2896                 if (len == 0) {
2897                         ip6_clearpktopts(opt, IPV6_RTHDR);
2898                         break;  /* just remove the option */
2899                 }
2900
2901                 /* message length validation */
2902                 if (len < sizeof(struct ip6_rthdr))
2903                         return (EINVAL);
2904                 rth = (struct ip6_rthdr *)buf;
2905                 rthlen = (rth->ip6r_len + 1) << 3;
2906                 if (len != rthlen)
2907                         return (EINVAL);
2908
2909                 switch (rth->ip6r_type) {
2910                 case IPV6_RTHDR_TYPE_0:
2911                         if (rth->ip6r_len == 0) /* must contain one addr */
2912                                 return (EINVAL);
2913                         if (rth->ip6r_len % 2) /* length must be even */
2914                                 return (EINVAL);
2915                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2916                                 return (EINVAL);
2917                         break;
2918                 default:
2919                         return (EINVAL);        /* not supported */
2920                 }
2921
2922                 /* turn off the previous option */
2923                 ip6_clearpktopts(opt, IPV6_RTHDR);
2924                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2925                 if (opt->ip6po_rthdr == NULL)
2926                         return (ENOBUFS);
2927                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2928
2929                 break;
2930         }
2931
2932         case IPV6_USE_MIN_MTU:
2933                 if (len != sizeof(int))
2934                         return (EINVAL);
2935                 minmtupolicy = *(int *)buf;
2936                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2937                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2938                     minmtupolicy != IP6PO_MINMTU_ALL) {
2939                         return (EINVAL);
2940                 }
2941                 opt->ip6po_minmtu = minmtupolicy;
2942                 break;
2943
2944         case IPV6_DONTFRAG:
2945                 if (len != sizeof(int))
2946                         return (EINVAL);
2947
2948                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2949                         /*
2950                          * we ignore this option for TCP sockets.
2951                          * (RFC3542 leaves this case unspecified.)
2952                          */
2953                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2954                 } else
2955                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2956                 break;
2957
2958         case IPV6_PREFER_TEMPADDR:
2959                 if (len != sizeof(int))
2960                         return (EINVAL);
2961                 preftemp = *(int *)buf;
2962                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2963                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2964                     preftemp != IP6PO_TEMPADDR_PREFER) {
2965                         return (EINVAL);
2966                 }
2967                 opt->ip6po_prefer_tempaddr = preftemp;
2968                 break;
2969
2970         default:
2971                 return (ENOPROTOOPT);
2972         } /* end of switch */
2973
2974         return (0);
2975 }
2976
2977 /*
2978  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2979  * packet to the input queue of a specified interface.  Note that this
2980  * calls the output routine of the loopback "driver", but with an interface
2981  * pointer that might NOT be &loif -- easier than replicating that code here.
2982  */
2983 void
2984 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2985 {
2986         struct mbuf *copym;
2987         struct ip6_hdr *ip6;
2988
2989         copym = m_copy(m, 0, M_COPYALL);
2990         if (copym == NULL)
2991                 return;
2992
2993         /*
2994          * Make sure to deep-copy IPv6 header portion in case the data
2995          * is in an mbuf cluster, so that we can safely override the IPv6
2996          * header portion later.
2997          */
2998         if ((copym->m_flags & M_EXT) != 0 ||
2999             copym->m_len < sizeof(struct ip6_hdr)) {
3000                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3001                 if (copym == NULL)
3002                         return;
3003         }
3004
3005 #ifdef DIAGNOSTIC
3006         if (copym->m_len < sizeof(*ip6)) {
3007                 m_freem(copym);
3008                 return;
3009         }
3010 #endif
3011
3012         ip6 = mtod(copym, struct ip6_hdr *);
3013         /*
3014          * clear embedded scope identifiers if necessary.
3015          * in6_clearscope will touch the addresses only when necessary.
3016          */
3017         in6_clearscope(&ip6->ip6_src);
3018         in6_clearscope(&ip6->ip6_dst);
3019
3020         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3021 }
3022
3023 /*
3024  * Chop IPv6 header off from the payload.
3025  */
3026 static int
3027 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3028 {
3029         struct mbuf *mh;
3030         struct ip6_hdr *ip6;
3031
3032         ip6 = mtod(m, struct ip6_hdr *);
3033         if (m->m_len > sizeof(*ip6)) {
3034                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3035                 if (mh == NULL) {
3036                         m_freem(m);
3037                         return ENOBUFS;
3038                 }
3039                 m_move_pkthdr(mh, m);
3040                 MH_ALIGN(mh, sizeof(*ip6));
3041                 m->m_len -= sizeof(*ip6);
3042                 m->m_data += sizeof(*ip6);
3043                 mh->m_next = m;
3044                 m = mh;
3045                 m->m_len = sizeof(*ip6);
3046                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3047         }
3048         exthdrs->ip6e_ip6 = m;
3049         return 0;
3050 }
3051
3052 /*
3053  * Compute IPv6 extension header length.
3054  */
3055 int
3056 ip6_optlen(struct inpcb *in6p)
3057 {
3058         int len;
3059
3060         if (!in6p->in6p_outputopts)
3061                 return 0;
3062
3063         len = 0;
3064 #define elen(x) \
3065     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3066
3067         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3068         if (in6p->in6p_outputopts->ip6po_rthdr)
3069                 /* dest1 is valid with rthdr only */
3070                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3071         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3072         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3073         return len;
3074 #undef elen
3075 }