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;
 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                 V_ip6stat.ip6s_badscope++;
 502                 goto bad;
 503         }
 504         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 505                 error = EOPNOTSUPP;
 506                 V_ip6stat.ip6s_badscope++;
 507                 goto bad;
 508         }
 509
 510         V_ip6stat.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         bzero(&dst_sa, sizeof(dst_sa));
 640         dst_sa.sin6_family = AF_INET6;
 641         dst_sa.sin6_len = sizeof(dst_sa);
 642         dst_sa.sin6_addr = ip6->ip6_dst;
 643         if (ro->ro_rt) {
 644                 rt = ro->ro_rt;
 645                 ifp = ro->ro_rt->rt_ifp;
 646         } else if ((error = in6_selectroute_fib(&dst_sa, opt, im6o, ro,
 647             &ifp, &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) {
 648                 switch (error) {
 649                 case EHOSTUNREACH:
 650                         V_ip6stat.ip6s_noroute++;
 651                         break;
 652                 case EADDRNOTAVAIL:
 653                 default:
 654                         break; /* XXX statistics? */
 655                 }
 656                 if (ifp != NULL)
 657                         in6_ifstat_inc(ifp, ifs6_out_discard);
 658                 goto bad;
 659         }
 660         if (rt == NULL) {
 661                 /*
 662                  * If in6_selectroute() does not return a route entry,
 663                  * dst may not have been updated.
 664                  */
 665                 *dst = dst_sa;  /* XXX */
 666         }
 667
 668         /*
 669          * then rt (for unicast) and ifp must be non-NULL valid values.
 670          */
 671         if ((flags & IPV6_FORWARDING) == 0) {
 672                 /* XXX: the FORWARDING flag can be set for mrouting. */
 673                 in6_ifstat_inc(ifp, ifs6_out_request);
 674         }
 675         if (rt != NULL) {
 676                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 677                 rt->rt_use++;
 678         }
 679
 680
 681         /*
 682          * The outgoing interface must be in the zone of source and
 683          * destination addresses.
 684          */
 685         origifp = ifp;
 686
 687         src0 = ip6->ip6_src;
 688         if (in6_setscope(&src0, origifp, &zone))
 689                 goto badscope;
 690         bzero(&src_sa, sizeof(src_sa));
 691         src_sa.sin6_family = AF_INET6;
 692         src_sa.sin6_len = sizeof(src_sa);
 693         src_sa.sin6_addr = ip6->ip6_src;
 694         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 695                 goto badscope;
 696
 697         dst0 = ip6->ip6_dst;
 698         if (in6_setscope(&dst0, origifp, &zone))
 699                 goto badscope;
 700         /* re-initialize to be sure */
 701         bzero(&dst_sa, sizeof(dst_sa));
 702         dst_sa.sin6_family = AF_INET6;
 703         dst_sa.sin6_len = sizeof(dst_sa);
 704         dst_sa.sin6_addr = ip6->ip6_dst;
 705         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 706                 goto badscope;
 707         }
 708
 709         /* We should use ia_ifp to support the case of
 710          * sending packets to an address of our own.
 711          */
 712         if (ia != NULL && ia->ia_ifp)
 713                 ifp = ia->ia_ifp;
 714
 715         /* scope check is done. */
 716         goto routefound;
 717
 718   badscope:
 719         V_ip6stat.ip6s_badscope++;
 720         in6_ifstat_inc(origifp, ifs6_out_discard);
 721         if (error == 0)
 722                 error = EHOSTUNREACH; /* XXX */
 723         goto bad;
 724
 725   routefound:
 726         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 727                 if (opt && opt->ip6po_nextroute.ro_rt) {
 728                         /*
 729                          * The nexthop is explicitly specified by the
 730                          * application.  We assume the next hop is an IPv6
 731                          * address.
 732                          */
 733                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 734                 }
 735                 else if ((rt->rt_flags & RTF_GATEWAY))
 736                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 737         }
 738
 739         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 740                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 741         } else {
 742                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 743                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 744                 /*
 745                  * Confirm that the outgoing interface supports multicast.
 746                  */
 747                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 748                         V_ip6stat.ip6s_noroute++;
 749                         in6_ifstat_inc(ifp, ifs6_out_discard);
 750                         error = ENETUNREACH;
 751                         goto bad;
 752                 }
 753                 if ((im6o == NULL && in6_mcast_loop) ||
 754                     (im6o && im6o->im6o_multicast_loop)) {
 755                         /*
 756                          * Loop back multicast datagram if not expressly
 757                          * forbidden to do so, even if we have not joined
 758                          * the address; protocols will filter it later,
 759                          * thus deferring a hash lookup and lock acquisition
 760                          * at the expense of an m_copym().
 761                          */
 762                         ip6_mloopback(ifp, m, dst);
 763                 } else {
 764                         /*
 765                          * If we are acting as a multicast router, perform
 766                          * multicast forwarding as if the packet had just
 767                          * arrived on the interface to which we are about
 768                          * to send.  The multicast forwarding function
 769                          * recursively calls this function, using the
 770                          * IPV6_FORWARDING flag to prevent infinite recursion.
 771                          *
 772                          * Multicasts that are looped back by ip6_mloopback(),
 773                          * above, will be forwarded by the ip6_input() routine,
 774                          * if necessary.
 775                          */
 776                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 777                                 /*
 778                                  * XXX: ip6_mforward expects that rcvif is NULL
 779                                  * when it is called from the originating path.
 780                                  * However, it is not always the case, since
 781                                  * some versions of MGETHDR() does not
 782                                  * initialize the field.
 783                                  */
 784                                 m->m_pkthdr.rcvif = NULL;
 785                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 786                                         m_freem(m);
 787                                         goto done;
 788                                 }
 789                         }
 790                 }
 791                 /*
 792                  * Multicasts with a hoplimit of zero may be looped back,
 793                  * above, but must not be transmitted on a network.
 794                  * Also, multicasts addressed to the loopback interface
 795                  * are not sent -- the above call to ip6_mloopback() will
 796                  * loop back a copy if this host actually belongs to the
 797                  * destination group on the loopback interface.
 798                  */
 799                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 800                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 801                         m_freem(m);
 802                         goto done;
 803                 }
 804         }
 805
 806         /*
 807          * Fill the outgoing inteface to tell the upper layer
 808          * to increment per-interface statistics.
 809          */
 810         if (ifpp)
 811                 *ifpp = ifp;
 812
 813         /* Determine path MTU. */
 814         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 815             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 816                 goto bad;
 817
 818         /*
 819          * The caller of this function may specify to use the minimum MTU
 820          * in some cases.
 821          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 822          * setting.  The logic is a bit complicated; by default, unicast
 823          * packets will follow path MTU while multicast packets will be sent at
 824          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 825          * including unicast ones will be sent at the minimum MTU.  Multicast
 826          * packets will always be sent at the minimum MTU unless
 827          * IP6PO_MINMTU_DISABLE is explicitly specified.
 828          * See RFC 3542 for more details.
 829          */
 830         if (mtu > IPV6_MMTU) {
 831                 if ((flags & IPV6_MINMTU))
 832                         mtu = IPV6_MMTU;
 833                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 834                         mtu = IPV6_MMTU;
 835                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 836                          (opt == NULL ||
 837                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 838                         mtu = IPV6_MMTU;
 839                 }
 840         }
 841
 842         /*
 843          * clear embedded scope identifiers if necessary.
 844          * in6_clearscope will touch the addresses only when necessary.
 845          */
 846         in6_clearscope(&ip6->ip6_src);
 847         in6_clearscope(&ip6->ip6_dst);
 848
 849         /*
 850          * If the outgoing packet contains a hop-by-hop options header,
 851          * it must be examined and processed even by the source node.
 852          * (RFC 2460, section 4.)
 853          */
 854         if (exthdrs.ip6e_hbh) {
 855                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 856                 u_int32_t dummy; /* XXX unused */
 857                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 858
 859 #ifdef DIAGNOSTIC
 860                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 861                         panic("ip6e_hbh is not contiguous");
 862 #endif
 863                 /*
 864                  *  XXX: if we have to send an ICMPv6 error to the sender,
 865                  *       we need the M_LOOP flag since icmp6_error() expects
 866                  *       the IPv6 and the hop-by-hop options header are
 867                  *       contiguous unless the flag is set.
 868                  */
 869                 m->m_flags |= M_LOOP;
 870                 m->m_pkthdr.rcvif = ifp;
 871                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 872                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 873                     &dummy, &plen) < 0) {
 874                         /* m was already freed at this point */
 875                         error = EINVAL;/* better error? */
 876                         goto done;
 877                 }
 878                 m->m_flags &= ~M_LOOP; /* XXX */
 879                 m->m_pkthdr.rcvif = NULL;
 880         }
 881
 882         /* Jump over all PFIL processing if hooks are not active. */
 883         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 884                 goto passout;
 885
 886         odst = ip6->ip6_dst;
 887         /* Run through list of hooks for output packets. */
 888         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 889         if (error != 0 || m == NULL)
 890                 goto done;
 891         ip6 = mtod(m, struct ip6_hdr *);
 892
 893         /* See if destination IP address was changed by packet filter. */
 894         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 895                 m->m_flags |= M_SKIP_FIREWALL;
 896                 /* If destination is now ourself drop to ip6_input(). */
 897                 if (in6_localip(&ip6->ip6_dst)) {
 898                         m->m_flags |= M_FASTFWD_OURS;
 899                         if (m->m_pkthdr.rcvif == NULL)
 900                                 m->m_pkthdr.rcvif = V_loif;
 901                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 902                                 m->m_pkthdr.csum_flags |=
 903                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 904                                 m->m_pkthdr.csum_data = 0xffff;
 905                         }
 906 #ifdef SCTP
 907                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 908                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 909 #endif
 910                         error = netisr_queue(NETISR_IPV6, m);
 911                         goto done;
 912                 } else
 913                         goto again;     /* Redo the routing table lookup. */
 914         }
 915
 916         if (V_pfilforward == 0)
 917                 goto passout;
 918         /* See if local, if yes, send it to netisr. */
 919         if (m->m_flags & M_FASTFWD_OURS) {
 920                 if (m->m_pkthdr.rcvif == NULL)
 921                         m->m_pkthdr.rcvif = V_loif;
 922                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 923                         m->m_pkthdr.csum_flags |=
 924                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 925                         m->m_pkthdr.csum_data = 0xffff;
 926                 }
 927 #ifdef SCTP
 928                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 929                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 930 #endif
 931                 error = netisr_queue(NETISR_IPV6, m);
 932                 goto done;
 933         }
 934         /* Or forward to some other address? */
 935         fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 936         if (fwd_tag) {
 937                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 938                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
 939                 m->m_flags |= M_SKIP_FIREWALL;
 940                 m_tag_delete(m, fwd_tag);
 941                 goto again;
 942         }
 943
 944 passout:
 945         /*
 946          * Send the packet to the outgoing interface.
 947          * If necessary, do IPv6 fragmentation before sending.
 948          *
 949          * the logic here is rather complex:
 950          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 951          * 1-a: send as is if tlen <= path mtu
 952          * 1-b: fragment if tlen > path mtu
 953          *
 954          * 2: if user asks us not to fragment (dontfrag == 1)
 955          * 2-a: send as is if tlen <= interface mtu
 956          * 2-b: error if tlen > interface mtu
 957          *
 958          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 959          *      always fragment
 960          *
 961          * 4: if dontfrag == 1 && alwaysfrag == 1
 962          *      error, as we cannot handle this conflicting request
 963          */
 964         sw_csum = m->m_pkthdr.csum_flags;
 965         if (!hdrsplit) {
 966                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 967                 sw_csum &= ~ifp->if_hwassist;
 968         } else
 969                 tso = 0;
 970         /*
 971          * If we added extension headers, we will not do TSO and calculate the
 972          * checksums ourselves for now.
 973          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 974          * with ext. hdrs.
 975          */
 976         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 977                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 978                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 979         }
 980 #ifdef SCTP
 981         if (sw_csum & CSUM_SCTP_IPV6) {
 982                 sw_csum &= ~CSUM_SCTP_IPV6;
 983                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 984         }
 985 #endif
 986         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 987         tlen = m->m_pkthdr.len;
 988
 989         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 990                 dontfrag = 1;
 991         else
 992                 dontfrag = 0;
 993         if (dontfrag && alwaysfrag) {   /* case 4 */
 994                 /* conflicting request - can't transmit */
 995                 error = EMSGSIZE;
 996                 goto bad;
 997         }
 998         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 999                 /*
1000                  * Even if the DONTFRAG option is specified, we cannot send the
1001                  * packet when the data length is larger than the MTU of the
1002                  * outgoing interface.
1003                  * Notify the error by sending IPV6_PATHMTU ancillary data as
1004                  * well as returning an error code (the latter is not described
1005                  * in the API spec.)
1006                  */
1007                 u_int32_t mtu32;
1008                 struct ip6ctlparam ip6cp;
1009
1010                 mtu32 = (u_int32_t)mtu;
1011                 bzero(&ip6cp, sizeof(ip6cp));
1012                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1013                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1014                     (void *)&ip6cp);
1015
1016                 error = EMSGSIZE;
1017                 goto bad;
1018         }
1019
1020         /*
1021          * transmit packet without fragmentation
1022          */
1023         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1024                 struct in6_ifaddr *ia6;
1025
1026                 ip6 = mtod(m, struct ip6_hdr *);
1027                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1028                 if (ia6) {
1029                         /* Record statistics for this interface address. */
1030                         ia6->ia_ifa.if_opackets++;
1031                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1032                         ifa_free(&ia6->ia_ifa);
1033                 }
1034                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1035                 goto done;
1036         }
1037
1038         /*
1039          * try to fragment the packet.  case 1-b and 3
1040          */
1041         if (mtu < IPV6_MMTU) {
1042                 /* path MTU cannot be less than IPV6_MMTU */
1043                 error = EMSGSIZE;
1044                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1045                 goto bad;
1046         } else if (ip6->ip6_plen == 0) {
1047                 /* jumbo payload cannot be fragmented */
1048                 error = EMSGSIZE;
1049                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1050                 goto bad;
1051         } else {
1052                 struct mbuf **mnext, *m_frgpart;
1053                 struct ip6_frag *ip6f;
1054                 u_int32_t id = htonl(ip6_randomid());
1055                 u_char nextproto;
1056
1057                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1058
1059                 /*
1060                  * Too large for the destination or interface;
1061                  * fragment if possible.
1062                  * Must be able to put at least 8 bytes per fragment.
1063                  */
1064                 hlen = unfragpartlen;
1065                 if (mtu > IPV6_MAXPACKET)
1066                         mtu = IPV6_MAXPACKET;
1067
1068                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1069                 if (len < 8) {
1070                         error = EMSGSIZE;
1071                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1072                         goto bad;
1073                 }
1074
1075                 /*
1076                  * Verify that we have any chance at all of being able to queue
1077                  *      the packet or packet fragments
1078                  */
1079                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1080                     < tlen  /* - hlen */)) {
1081                         error = ENOBUFS;
1082                         V_ip6stat.ip6s_odropped++;
1083                         goto bad;
1084                 }
1085
1086
1087                 /*
1088                  * If the interface will not calculate checksums on
1089                  * fragmented packets, then do it here.
1090                  * XXX-BZ handle the hw offloading case.  Need flags.
1091                  */
1092                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1093                         in6_delayed_cksum(m, plen, hlen);
1094                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1095                 }
1096 #ifdef SCTP
1097                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1098                         sctp_delayed_cksum(m, hlen);
1099                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1100                 }
1101 #endif
1102                 mnext = &m->m_nextpkt;
1103
1104                 /*
1105                  * Change the next header field of the last header in the
1106                  * unfragmentable part.
1107                  */
1108                 if (exthdrs.ip6e_rthdr) {
1109                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1110                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1111                 } else if (exthdrs.ip6e_dest1) {
1112                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1113                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1114                 } else if (exthdrs.ip6e_hbh) {
1115                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1116                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1117                 } else {
1118                         nextproto = ip6->ip6_nxt;
1119                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1120                 }
1121
1122                 /*
1123                  * Loop through length of segment after first fragment,
1124                  * make new header and copy data of each part and link onto
1125                  * chain.
1126                  */
1127                 m0 = m;
1128                 for (off = hlen; off < tlen; off += len) {
1129                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1130                         if (!m) {
1131                                 error = ENOBUFS;
1132                                 V_ip6stat.ip6s_odropped++;
1133                                 goto sendorfree;
1134                         }
1135                         m->m_pkthdr.rcvif = NULL;
1136                         m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1137                         *mnext = m;
1138                         mnext = &m->m_nextpkt;
1139                         m->m_data += max_linkhdr;
1140                         mhip6 = mtod(m, struct ip6_hdr *);
1141                         *mhip6 = *ip6;
1142                         m->m_len = sizeof(*mhip6);
1143                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1144                         if (error) {
1145                                 V_ip6stat.ip6s_odropped++;
1146                                 goto sendorfree;
1147                         }
1148                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1149                         if (off + len >= tlen)
1150                                 len = tlen - off;
1151                         else
1152                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1153                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1154                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1155                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1156                                 error = ENOBUFS;
1157                                 V_ip6stat.ip6s_odropped++;
1158                                 goto sendorfree;
1159                         }
1160                         m_cat(m, m_frgpart);
1161                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1162                         m->m_pkthdr.rcvif = NULL;
1163                         ip6f->ip6f_reserved = 0;
1164                         ip6f->ip6f_ident = id;
1165                         ip6f->ip6f_nxt = nextproto;
1166                         V_ip6stat.ip6s_ofragments++;
1167                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1168                 }
1169
1170                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1171         }
1172
1173         /*
1174          * Remove leading garbages.
1175          */
1176 sendorfree:
1177         m = m0->m_nextpkt;
1178         m0->m_nextpkt = 0;
1179         m_freem(m0);
1180         for (m0 = m; m; m = m0) {
1181                 m0 = m->m_nextpkt;
1182                 m->m_nextpkt = 0;
1183                 if (error == 0) {
1184                         /* Record statistics for this interface address. */
1185                         if (ia) {
1186                                 ia->ia_ifa.if_opackets++;
1187                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1188                         }
1189                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1190                 } else
1191                         m_freem(m);
1192         }
1193
1194         if (error == 0)
1195                 V_ip6stat.ip6s_fragmented++;
1196
1197 done:
1198         if (ro == &ip6route)
1199                 RO_RTFREE(ro);
1200         if (ro_pmtu == &ip6route)
1201                 RO_RTFREE(ro_pmtu);
1202 #ifdef IPSEC
1203         if (sp != NULL)
1204                 KEY_FREESP(&sp);
1205 #endif
1206
1207         return (error);
1208
1209 freehdrs:
1210         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1211         m_freem(exthdrs.ip6e_dest1);
1212         m_freem(exthdrs.ip6e_rthdr);
1213         m_freem(exthdrs.ip6e_dest2);
1214         /* FALLTHROUGH */
1215 bad:
1216         if (m)
1217                 m_freem(m);
1218         goto done;
1219 }
1220
1221 static int
1222 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1223 {
1224         struct mbuf *m;
1225
1226         if (hlen > MCLBYTES)
1227                 return (ENOBUFS); /* XXX */
1228
1229         MGET(m, M_DONTWAIT, MT_DATA);
1230         if (!m)
1231                 return (ENOBUFS);
1232
1233         if (hlen > MLEN) {
1234                 MCLGET(m, M_DONTWAIT);
1235                 if ((m->m_flags & M_EXT) == 0) {
1236                         m_free(m);
1237                         return (ENOBUFS);
1238                 }
1239         }
1240         m->m_len = hlen;
1241         if (hdr)
1242                 bcopy(hdr, mtod(m, caddr_t), hlen);
1243
1244         *mp = m;
1245         return (0);
1246 }
1247
1248 /*
1249  * Insert jumbo payload option.
1250  */
1251 static int
1252 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1253 {
1254         struct mbuf *mopt;
1255         u_char *optbuf;
1256         u_int32_t v;
1257
1258 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1259
1260         /*
1261          * If there is no hop-by-hop options header, allocate new one.
1262          * If there is one but it doesn't have enough space to store the
1263          * jumbo payload option, allocate a cluster to store the whole options.
1264          * Otherwise, use it to store the options.
1265          */
1266         if (exthdrs->ip6e_hbh == 0) {
1267                 MGET(mopt, M_DONTWAIT, MT_DATA);
1268                 if (mopt == 0)
1269                         return (ENOBUFS);
1270                 mopt->m_len = JUMBOOPTLEN;
1271                 optbuf = mtod(mopt, u_char *);
1272                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1273                 exthdrs->ip6e_hbh = mopt;
1274         } else {
1275                 struct ip6_hbh *hbh;
1276
1277                 mopt = exthdrs->ip6e_hbh;
1278                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1279                         /*
1280                          * XXX assumption:
1281                          * - exthdrs->ip6e_hbh is not referenced from places
1282                          *   other than exthdrs.
1283                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1284                          */
1285                         int oldoptlen = mopt->m_len;
1286                         struct mbuf *n;
1287
1288                         /*
1289                          * XXX: give up if the whole (new) hbh header does
1290                          * not fit even in an mbuf cluster.
1291                          */
1292                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1293                                 return (ENOBUFS);
1294
1295                         /*
1296                          * As a consequence, we must always prepare a cluster
1297                          * at this point.
1298                          */
1299                         MGET(n, M_DONTWAIT, MT_DATA);
1300                         if (n) {
1301                                 MCLGET(n, M_DONTWAIT);
1302                                 if ((n->m_flags & M_EXT) == 0) {
1303                                         m_freem(n);
1304                                         n = NULL;
1305                                 }
1306                         }
1307                         if (!n)
1308                                 return (ENOBUFS);
1309                         n->m_len = oldoptlen + JUMBOOPTLEN;
1310                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1311                             oldoptlen);
1312                         optbuf = mtod(n, caddr_t) + oldoptlen;
1313                         m_freem(mopt);
1314                         mopt = exthdrs->ip6e_hbh = n;
1315                 } else {
1316                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1317                         mopt->m_len += JUMBOOPTLEN;
1318                 }
1319                 optbuf[0] = IP6OPT_PADN;
1320                 optbuf[1] = 1;
1321
1322                 /*
1323                  * Adjust the header length according to the pad and
1324                  * the jumbo payload option.
1325                  */
1326                 hbh = mtod(mopt, struct ip6_hbh *);
1327                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1328         }
1329
1330         /* fill in the option. */
1331         optbuf[2] = IP6OPT_JUMBO;
1332         optbuf[3] = 4;
1333         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1334         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1335
1336         /* finally, adjust the packet header length */
1337         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1338
1339         return (0);
1340 #undef JUMBOOPTLEN
1341 }
1342
1343 /*
1344  * Insert fragment header and copy unfragmentable header portions.
1345  */
1346 static int
1347 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1348     struct ip6_frag **frghdrp)
1349 {
1350         struct mbuf *n, *mlast;
1351
1352         if (hlen > sizeof(struct ip6_hdr)) {
1353                 n = m_copym(m0, sizeof(struct ip6_hdr),
1354                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1355                 if (n == 0)
1356                         return (ENOBUFS);
1357                 m->m_next = n;
1358         } else
1359                 n = m;
1360
1361         /* Search for the last mbuf of unfragmentable part. */
1362         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1363                 ;
1364
1365         if ((mlast->m_flags & M_EXT) == 0 &&
1366             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1367                 /* use the trailing space of the last mbuf for the fragment hdr */
1368                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1369                     mlast->m_len);
1370                 mlast->m_len += sizeof(struct ip6_frag);
1371                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1372         } else {
1373                 /* allocate a new mbuf for the fragment header */
1374                 struct mbuf *mfrg;
1375
1376                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1377                 if (mfrg == 0)
1378                         return (ENOBUFS);
1379                 mfrg->m_len = sizeof(struct ip6_frag);
1380                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1381                 mlast->m_next = mfrg;
1382         }
1383
1384         return (0);
1385 }
1386
1387 static int
1388 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1389     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1390     int *alwaysfragp, u_int fibnum)
1391 {
1392         u_int32_t mtu = 0;
1393         int alwaysfrag = 0;
1394         int error = 0;
1395
1396         if (ro_pmtu != ro) {
1397                 /* The first hop and the final destination may differ. */
1398                 struct sockaddr_in6 *sa6_dst =
1399                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1400                 if (ro_pmtu->ro_rt &&
1401                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1402                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1403                         RTFREE(ro_pmtu->ro_rt);
1404                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1405                 }
1406                 if (ro_pmtu->ro_rt == NULL) {
1407                         bzero(sa6_dst, sizeof(*sa6_dst));
1408                         sa6_dst->sin6_family = AF_INET6;
1409                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1410                         sa6_dst->sin6_addr = *dst;
1411
1412                         in6_rtalloc(ro_pmtu, fibnum);
1413                 }
1414         }
1415         if (ro_pmtu->ro_rt) {
1416                 u_int32_t ifmtu;
1417                 struct in_conninfo inc;
1418
1419                 bzero(&inc, sizeof(inc));
1420                 inc.inc_flags |= INC_ISIPV6;
1421                 inc.inc6_faddr = *dst;
1422
1423                 if (ifp == NULL)
1424                         ifp = ro_pmtu->ro_rt->rt_ifp;
1425                 ifmtu = IN6_LINKMTU(ifp);
1426                 mtu = tcp_hc_getmtu(&inc);
1427                 if (mtu)
1428                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1429                 else
1430                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1431                 if (mtu == 0)
1432                         mtu = ifmtu;
1433                 else if (mtu < IPV6_MMTU) {
1434                         /*
1435                          * RFC2460 section 5, last paragraph:
1436                          * if we record ICMPv6 too big message with
1437                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1438                          * or smaller, with framgent header attached.
1439                          * (fragment header is needed regardless from the
1440                          * packet size, for translators to identify packets)
1441                          */
1442                         alwaysfrag = 1;
1443                         mtu = IPV6_MMTU;
1444                 } else if (mtu > ifmtu) {
1445                         /*
1446                          * The MTU on the route is larger than the MTU on
1447                          * the interface!  This shouldn't happen, unless the
1448                          * MTU of the interface has been changed after the
1449                          * interface was brought up.  Change the MTU in the
1450                          * route to match the interface MTU (as long as the
1451                          * field isn't locked).
1452                          */
1453                         mtu = ifmtu;
1454                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1455                 }
1456         } else if (ifp) {
1457                 mtu = IN6_LINKMTU(ifp);
1458         } else
1459                 error = EHOSTUNREACH; /* XXX */
1460
1461         *mtup = mtu;
1462         if (alwaysfragp)
1463                 *alwaysfragp = alwaysfrag;
1464         return (error);
1465 }
1466
1467 /*
1468  * IP6 socket option processing.
1469  */
1470 int
1471 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1472 {
1473         int optdatalen, uproto;
1474         void *optdata;
1475         struct inpcb *in6p = sotoinpcb(so);
1476         int error, optval;
1477         int level, op, optname;
1478         int optlen;
1479         struct thread *td;
1480
1481         level = sopt->sopt_level;
1482         op = sopt->sopt_dir;
1483         optname = sopt->sopt_name;
1484         optlen = sopt->sopt_valsize;
1485         td = sopt->sopt_td;
1486         error = 0;
1487         optval = 0;
1488         uproto = (int)so->so_proto->pr_protocol;
1489
1490         if (level != IPPROTO_IPV6) {
1491                 error = EINVAL;
1492
1493                 if (sopt->sopt_level == SOL_SOCKET &&
1494                     sopt->sopt_dir == SOPT_SET) {
1495                         switch (sopt->sopt_name) {
1496                         case SO_REUSEADDR:
1497                                 INP_WLOCK(in6p);
1498                                 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1499                                         if ((so->so_options &
1500                                             (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1501                                                 in6p->inp_flags2 |= INP_REUSEPORT;
1502                                         else
1503                                                 in6p->inp_flags2 &= ~INP_REUSEPORT;
1504                                 }
1505                                 INP_WUNLOCK(in6p);
1506                                 error = 0;
1507                                 break;
1508                         case SO_REUSEPORT:
1509                                 INP_WLOCK(in6p);
1510                                 if ((so->so_options & SO_REUSEPORT) != 0)
1511                                         in6p->inp_flags2 |= INP_REUSEPORT;
1512                                 else
1513                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1514                                 INP_WUNLOCK(in6p);
1515                                 error = 0;
1516                                 break;
1517                         case SO_SETFIB:
1518                                 INP_WLOCK(in6p);
1519                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1520                                 INP_WUNLOCK(in6p);
1521                                 error = 0;
1522                                 break;
1523                         default:
1524                                 break;
1525                         }
1526                 }
1527         } else {                /* level == IPPROTO_IPV6 */
1528                 switch (op) {
1529
1530                 case SOPT_SET:
1531                         switch (optname) {
1532                         case IPV6_2292PKTOPTIONS:
1533 #ifdef IPV6_PKTOPTIONS
1534                         case IPV6_PKTOPTIONS:
1535 #endif
1536                         {
1537                                 struct mbuf *m;
1538
1539                                 error = soopt_getm(sopt, &m); /* XXX */
1540                                 if (error != 0)
1541                                         break;
1542                                 error = soopt_mcopyin(sopt, m); /* XXX */
1543                                 if (error != 0)
1544                                         break;
1545                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1546                                                     m, so, sopt);
1547                                 m_freem(m); /* XXX */
1548                                 break;
1549                         }
1550
1551                         /*
1552                          * Use of some Hop-by-Hop options or some
1553                          * Destination options, might require special
1554                          * privilege.  That is, normal applications
1555                          * (without special privilege) might be forbidden
1556                          * from setting certain options in outgoing packets,
1557                          * and might never see certain options in received
1558                          * packets. [RFC 2292 Section 6]
1559                          * KAME specific note:
1560                          *  KAME prevents non-privileged users from sending or
1561                          *  receiving ANY hbh/dst options in order to avoid
1562                          *  overhead of parsing options in the kernel.
1563                          */
1564                         case IPV6_RECVHOPOPTS:
1565                         case IPV6_RECVDSTOPTS:
1566                         case IPV6_RECVRTHDRDSTOPTS:
1567                                 if (td != NULL) {
1568                                         error = priv_check(td,
1569                                             PRIV_NETINET_SETHDROPTS);
1570                                         if (error)
1571                                                 break;
1572                                 }
1573                                 /* FALLTHROUGH */
1574                         case IPV6_UNICAST_HOPS:
1575                         case IPV6_HOPLIMIT:
1576                         case IPV6_FAITH:
1577
1578                         case IPV6_RECVPKTINFO:
1579                         case IPV6_RECVHOPLIMIT:
1580                         case IPV6_RECVRTHDR:
1581                         case IPV6_RECVPATHMTU:
1582                         case IPV6_RECVTCLASS:
1583                         case IPV6_V6ONLY:
1584                         case IPV6_AUTOFLOWLABEL:
1585                         case IPV6_BINDANY:
1586                                 if (optname == IPV6_BINDANY && td != NULL) {
1587                                         error = priv_check(td,
1588                                             PRIV_NETINET_BINDANY);
1589                                         if (error)
1590                                                 break;
1591                                 }
1592
1593                                 if (optlen != sizeof(int)) {
1594                                         error = EINVAL;
1595                                         break;
1596                                 }
1597                                 error = sooptcopyin(sopt, &optval,
1598                                         sizeof optval, sizeof optval);
1599                                 if (error)
1600                                         break;
1601                                 switch (optname) {
1602
1603                                 case IPV6_UNICAST_HOPS:
1604                                         if (optval < -1 || optval >= 256)
1605                                                 error = EINVAL;
1606                                         else {
1607                                                 /* -1 = kernel default */
1608                                                 in6p->in6p_hops = optval;
1609                                                 if ((in6p->inp_vflag &
1610                                                      INP_IPV4) != 0)
1611                                                         in6p->inp_ip_ttl = optval;
1612                                         }
1613                                         break;
1614 #define OPTSET(bit) \
1615 do { \
1616         INP_WLOCK(in6p); \
1617         if (optval) \
1618                 in6p->inp_flags |= (bit); \
1619         else \
1620                 in6p->inp_flags &= ~(bit); \
1621         INP_WUNLOCK(in6p); \
1622 } while (/*CONSTCOND*/ 0)
1623 #define OPTSET2292(bit) \
1624 do { \
1625         INP_WLOCK(in6p); \
1626         in6p->inp_flags |= IN6P_RFC2292; \
1627         if (optval) \
1628                 in6p->inp_flags |= (bit); \
1629         else \
1630                 in6p->inp_flags &= ~(bit); \
1631         INP_WUNLOCK(in6p); \
1632 } while (/*CONSTCOND*/ 0)
1633 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1634
1635                                 case IPV6_RECVPKTINFO:
1636                                         /* cannot mix with RFC2292 */
1637                                         if (OPTBIT(IN6P_RFC2292)) {
1638                                                 error = EINVAL;
1639                                                 break;
1640                                         }
1641                                         OPTSET(IN6P_PKTINFO);
1642                                         break;
1643
1644                                 case IPV6_HOPLIMIT:
1645                                 {
1646                                         struct ip6_pktopts **optp;
1647
1648                                         /* cannot mix with RFC2292 */
1649                                         if (OPTBIT(IN6P_RFC2292)) {
1650                                                 error = EINVAL;
1651                                                 break;
1652                                         }
1653                                         optp = &in6p->in6p_outputopts;
1654                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1655                                             (u_char *)&optval, sizeof(optval),
1656                                             optp, (td != NULL) ? td->td_ucred :
1657                                             NULL, uproto);
1658                                         break;
1659                                 }
1660
1661                                 case IPV6_RECVHOPLIMIT:
1662                                         /* cannot mix with RFC2292 */
1663                                         if (OPTBIT(IN6P_RFC2292)) {
1664                                                 error = EINVAL;
1665                                                 break;
1666                                         }
1667                                         OPTSET(IN6P_HOPLIMIT);
1668                                         break;
1669
1670                                 case IPV6_RECVHOPOPTS:
1671                                         /* cannot mix with RFC2292 */
1672                                         if (OPTBIT(IN6P_RFC2292)) {
1673                                                 error = EINVAL;
1674                                                 break;
1675                                         }
1676                                         OPTSET(IN6P_HOPOPTS);
1677                                         break;
1678
1679                                 case IPV6_RECVDSTOPTS:
1680                                         /* cannot mix with RFC2292 */
1681                                         if (OPTBIT(IN6P_RFC2292)) {
1682                                                 error = EINVAL;
1683                                                 break;
1684                                         }
1685                                         OPTSET(IN6P_DSTOPTS);
1686                                         break;
1687
1688                                 case IPV6_RECVRTHDRDSTOPTS:
1689                                         /* cannot mix with RFC2292 */
1690                                         if (OPTBIT(IN6P_RFC2292)) {
1691                                                 error = EINVAL;
1692                                                 break;
1693                                         }
1694                                         OPTSET(IN6P_RTHDRDSTOPTS);
1695                                         break;
1696
1697                                 case IPV6_RECVRTHDR:
1698                                         /* cannot mix with RFC2292 */
1699                                         if (OPTBIT(IN6P_RFC2292)) {
1700                                                 error = EINVAL;
1701                                                 break;
1702                                         }
1703                                         OPTSET(IN6P_RTHDR);
1704                                         break;
1705
1706                                 case IPV6_FAITH:
1707                                         OPTSET(INP_FAITH);
1708                                         break;
1709
1710                                 case IPV6_RECVPATHMTU:
1711                                         /*
1712                                          * We ignore this option for TCP
1713                                          * sockets.
1714                                          * (RFC3542 leaves this case
1715                                          * unspecified.)
1716                                          */
1717                                         if (uproto != IPPROTO_TCP)
1718                                                 OPTSET(IN6P_MTU);
1719                                         break;
1720
1721                                 case IPV6_V6ONLY:
1722                                         /*
1723                                          * make setsockopt(IPV6_V6ONLY)
1724                                          * available only prior to bind(2).
1725                                          * see ipng mailing list, Jun 22 2001.
1726                                          */
1727                                         if (in6p->inp_lport ||
1728                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1729                                                 error = EINVAL;
1730                                                 break;
1731                                         }
1732                                         OPTSET(IN6P_IPV6_V6ONLY);
1733                                         if (optval)
1734                                                 in6p->inp_vflag &= ~INP_IPV4;
1735                                         else
1736                                                 in6p->inp_vflag |= INP_IPV4;
1737                                         break;
1738                                 case IPV6_RECVTCLASS:
1739                                         /* cannot mix with RFC2292 XXX */
1740                                         if (OPTBIT(IN6P_RFC2292)) {
1741                                                 error = EINVAL;
1742                                                 break;
1743                                         }
1744                                         OPTSET(IN6P_TCLASS);
1745                                         break;
1746                                 case IPV6_AUTOFLOWLABEL:
1747                                         OPTSET(IN6P_AUTOFLOWLABEL);
1748                                         break;
1749
1750                                 case IPV6_BINDANY:
1751                                         OPTSET(INP_BINDANY);
1752                                         break;
1753                                 }
1754                                 break;
1755
1756                         case IPV6_TCLASS:
1757                         case IPV6_DONTFRAG:
1758                         case IPV6_USE_MIN_MTU:
1759                         case IPV6_PREFER_TEMPADDR:
1760                                 if (optlen != sizeof(optval)) {
1761                                         error = EINVAL;
1762                                         break;
1763                                 }
1764                                 error = sooptcopyin(sopt, &optval,
1765                                         sizeof optval, sizeof optval);
1766                                 if (error)
1767                                         break;
1768                                 {
1769                                         struct ip6_pktopts **optp;
1770                                         optp = &in6p->in6p_outputopts;
1771                                         error = ip6_pcbopt(optname,
1772                                             (u_char *)&optval, sizeof(optval),
1773                                             optp, (td != NULL) ? td->td_ucred :
1774                                             NULL, uproto);
1775                                         break;
1776                                 }
1777
1778                         case IPV6_2292PKTINFO:
1779                         case IPV6_2292HOPLIMIT:
1780                         case IPV6_2292HOPOPTS:
1781                         case IPV6_2292DSTOPTS:
1782                         case IPV6_2292RTHDR:
1783                                 /* RFC 2292 */
1784                                 if (optlen != sizeof(int)) {
1785                                         error = EINVAL;
1786                                         break;
1787                                 }
1788                                 error = sooptcopyin(sopt, &optval,
1789                                         sizeof optval, sizeof optval);
1790                                 if (error)
1791                                         break;
1792                                 switch (optname) {
1793                                 case IPV6_2292PKTINFO:
1794                                         OPTSET2292(IN6P_PKTINFO);
1795                                         break;
1796                                 case IPV6_2292HOPLIMIT:
1797                                         OPTSET2292(IN6P_HOPLIMIT);
1798                                         break;
1799                                 case IPV6_2292HOPOPTS:
1800                                         /*
1801                                          * Check super-user privilege.
1802                                          * See comments for IPV6_RECVHOPOPTS.
1803                                          */
1804                                         if (td != NULL) {
1805                                                 error = priv_check(td,
1806                                                     PRIV_NETINET_SETHDROPTS);
1807                                                 if (error)
1808                                                         return (error);
1809                                         }
1810                                         OPTSET2292(IN6P_HOPOPTS);
1811                                         break;
1812                                 case IPV6_2292DSTOPTS:
1813                                         if (td != NULL) {
1814                                                 error = priv_check(td,
1815                                                     PRIV_NETINET_SETHDROPTS);
1816                                                 if (error)
1817                                                         return (error);
1818                                         }
1819                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1820                                         break;
1821                                 case IPV6_2292RTHDR:
1822                                         OPTSET2292(IN6P_RTHDR);
1823                                         break;
1824                                 }
1825                                 break;
1826                         case IPV6_PKTINFO:
1827                         case IPV6_HOPOPTS:
1828                         case IPV6_RTHDR:
1829                         case IPV6_DSTOPTS:
1830                         case IPV6_RTHDRDSTOPTS:
1831                         case IPV6_NEXTHOP:
1832                         {
1833                                 /* new advanced API (RFC3542) */
1834                                 u_char *optbuf;
1835                                 u_char optbuf_storage[MCLBYTES];
1836                                 int optlen;
1837                                 struct ip6_pktopts **optp;
1838
1839                                 /* cannot mix with RFC2292 */
1840                                 if (OPTBIT(IN6P_RFC2292)) {
1841                                         error = EINVAL;
1842                                         break;
1843                                 }
1844
1845                                 /*
1846                                  * We only ensure valsize is not too large
1847                                  * here.  Further validation will be done
1848                                  * later.
1849                                  */
1850                                 error = sooptcopyin(sopt, optbuf_storage,
1851                                     sizeof(optbuf_storage), 0);
1852                                 if (error)
1853                                         break;
1854                                 optlen = sopt->sopt_valsize;
1855                                 optbuf = optbuf_storage;
1856                                 optp = &in6p->in6p_outputopts;
1857                                 error = ip6_pcbopt(optname, optbuf, optlen,
1858                                     optp, (td != NULL) ? td->td_ucred : NULL,
1859                                     uproto);
1860                                 break;
1861                         }
1862 #undef OPTSET
1863
1864                         case IPV6_MULTICAST_IF:
1865                         case IPV6_MULTICAST_HOPS:
1866                         case IPV6_MULTICAST_LOOP:
1867                         case IPV6_JOIN_GROUP:
1868                         case IPV6_LEAVE_GROUP:
1869                         case IPV6_MSFILTER:
1870                         case MCAST_BLOCK_SOURCE:
1871                         case MCAST_UNBLOCK_SOURCE:
1872                         case MCAST_JOIN_GROUP:
1873                         case MCAST_LEAVE_GROUP:
1874                         case MCAST_JOIN_SOURCE_GROUP:
1875                         case MCAST_LEAVE_SOURCE_GROUP:
1876                                 error = ip6_setmoptions(in6p, sopt);
1877                                 break;
1878
1879                         case IPV6_PORTRANGE:
1880                                 error = sooptcopyin(sopt, &optval,
1881                                     sizeof optval, sizeof optval);
1882                                 if (error)
1883                                         break;
1884
1885                                 INP_WLOCK(in6p);
1886                                 switch (optval) {
1887                                 case IPV6_PORTRANGE_DEFAULT:
1888                                         in6p->inp_flags &= ~(INP_LOWPORT);
1889                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1890                                         break;
1891
1892                                 case IPV6_PORTRANGE_HIGH:
1893                                         in6p->inp_flags &= ~(INP_LOWPORT);
1894                                         in6p->inp_flags |= INP_HIGHPORT;
1895                                         break;
1896
1897                                 case IPV6_PORTRANGE_LOW:
1898                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1899                                         in6p->inp_flags |= INP_LOWPORT;
1900                                         break;
1901
1902                                 default:
1903                                         error = EINVAL;
1904                                         break;
1905                                 }
1906                                 INP_WUNLOCK(in6p);
1907                                 break;
1908
1909 #ifdef IPSEC
1910                         case IPV6_IPSEC_POLICY:
1911                         {
1912                                 caddr_t req;
1913                                 struct mbuf *m;
1914
1915                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1916                                         break;
1917                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1918                                         break;
1919                                 req = mtod(m, caddr_t);
1920                                 error = ipsec_set_policy(in6p, optname, req,
1921                                     m->m_len, (sopt->sopt_td != NULL) ?
1922                                     sopt->sopt_td->td_ucred : NULL);
1923                                 m_freem(m);
1924                                 break;
1925                         }
1926 #endif /* IPSEC */
1927
1928                         default:
1929                                 error = ENOPROTOOPT;
1930                                 break;
1931                         }
1932                         break;
1933
1934                 case SOPT_GET:
1935                         switch (optname) {
1936
1937                         case IPV6_2292PKTOPTIONS:
1938 #ifdef IPV6_PKTOPTIONS
1939                         case IPV6_PKTOPTIONS:
1940 #endif
1941                                 /*
1942                                  * RFC3542 (effectively) deprecated the
1943                                  * semantics of the 2292-style pktoptions.
1944                                  * Since it was not reliable in nature (i.e.,
1945                                  * applications had to expect the lack of some
1946                                  * information after all), it would make sense
1947                                  * to simplify this part by always returning
1948                                  * empty data.
1949                                  */
1950                                 sopt->sopt_valsize = 0;
1951                                 break;
1952
1953                         case IPV6_RECVHOPOPTS:
1954                         case IPV6_RECVDSTOPTS:
1955                         case IPV6_RECVRTHDRDSTOPTS:
1956                         case IPV6_UNICAST_HOPS:
1957                         case IPV6_RECVPKTINFO:
1958                         case IPV6_RECVHOPLIMIT:
1959                         case IPV6_RECVRTHDR:
1960                         case IPV6_RECVPATHMTU:
1961
1962                         case IPV6_FAITH:
1963                         case IPV6_V6ONLY:
1964                         case IPV6_PORTRANGE:
1965                         case IPV6_RECVTCLASS:
1966                         case IPV6_AUTOFLOWLABEL:
1967                         case IPV6_BINDANY:
1968                                 switch (optname) {
1969
1970                                 case IPV6_RECVHOPOPTS:
1971                                         optval = OPTBIT(IN6P_HOPOPTS);
1972                                         break;
1973
1974                                 case IPV6_RECVDSTOPTS:
1975                                         optval = OPTBIT(IN6P_DSTOPTS);
1976                                         break;
1977
1978                                 case IPV6_RECVRTHDRDSTOPTS:
1979                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1980                                         break;
1981
1982                                 case IPV6_UNICAST_HOPS:
1983                                         optval = in6p->in6p_hops;
1984                                         break;
1985
1986                                 case IPV6_RECVPKTINFO:
1987                                         optval = OPTBIT(IN6P_PKTINFO);
1988                                         break;
1989
1990                                 case IPV6_RECVHOPLIMIT:
1991                                         optval = OPTBIT(IN6P_HOPLIMIT);
1992                                         break;
1993
1994                                 case IPV6_RECVRTHDR:
1995                                         optval = OPTBIT(IN6P_RTHDR);
1996                                         break;
1997
1998                                 case IPV6_RECVPATHMTU:
1999                                         optval = OPTBIT(IN6P_MTU);
2000                                         break;
2001
2002                                 case IPV6_FAITH:
2003                                         optval = OPTBIT(INP_FAITH);
2004                                         break;
2005
2006                                 case IPV6_V6ONLY:
2007                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2008                                         break;
2009
2010                                 case IPV6_PORTRANGE:
2011                                     {
2012                                         int flags;
2013                                         flags = in6p->inp_flags;
2014                                         if (flags & INP_HIGHPORT)
2015                                                 optval = IPV6_PORTRANGE_HIGH;
2016                                         else if (flags & INP_LOWPORT)
2017                                                 optval = IPV6_PORTRANGE_LOW;
2018                                         else
2019                                                 optval = 0;
2020                                         break;
2021                                     }
2022                                 case IPV6_RECVTCLASS:
2023                                         optval = OPTBIT(IN6P_TCLASS);
2024                                         break;
2025
2026                                 case IPV6_AUTOFLOWLABEL:
2027                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2028                                         break;
2029
2030                                 case IPV6_BINDANY:
2031                                         optval = OPTBIT(INP_BINDANY);
2032                                         break;
2033                                 }
2034                                 if (error)
2035                                         break;
2036                                 error = sooptcopyout(sopt, &optval,
2037                                         sizeof optval);
2038                                 break;
2039
2040                         case IPV6_PATHMTU:
2041                         {
2042                                 u_long pmtu = 0;
2043                                 struct ip6_mtuinfo mtuinfo;
2044                                 struct route_in6 sro;
2045
2046                                 bzero(&sro, sizeof(sro));
2047
2048                                 if (!(so->so_state & SS_ISCONNECTED))
2049                                         return (ENOTCONN);
2050                                 /*
2051                                  * XXX: we dot not consider the case of source
2052                                  * routing, or optional information to specify
2053                                  * the outgoing interface.
2054                                  */
2055                                 error = ip6_getpmtu(&sro, NULL, NULL,
2056                                     &in6p->in6p_faddr, &pmtu, NULL,
2057                                     so->so_fibnum);
2058                                 if (sro.ro_rt)
2059                                         RTFREE(sro.ro_rt);
2060                                 if (error)
2061                                         break;
2062                                 if (pmtu > IPV6_MAXPACKET)
2063                                         pmtu = IPV6_MAXPACKET;
2064
2065                                 bzero(&mtuinfo, sizeof(mtuinfo));
2066                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2067                                 optdata = (void *)&mtuinfo;
2068                                 optdatalen = sizeof(mtuinfo);
2069                                 error = sooptcopyout(sopt, optdata,
2070                                     optdatalen);
2071                                 break;
2072                         }
2073
2074                         case IPV6_2292PKTINFO:
2075                         case IPV6_2292HOPLIMIT:
2076                         case IPV6_2292HOPOPTS:
2077                         case IPV6_2292RTHDR:
2078                         case IPV6_2292DSTOPTS:
2079                                 switch (optname) {
2080                                 case IPV6_2292PKTINFO:
2081                                         optval = OPTBIT(IN6P_PKTINFO);
2082                                         break;
2083                                 case IPV6_2292HOPLIMIT:
2084                                         optval = OPTBIT(IN6P_HOPLIMIT);
2085                                         break;
2086                                 case IPV6_2292HOPOPTS:
2087                                         optval = OPTBIT(IN6P_HOPOPTS);
2088                                         break;
2089                                 case IPV6_2292RTHDR:
2090                                         optval = OPTBIT(IN6P_RTHDR);
2091                                         break;
2092                                 case IPV6_2292DSTOPTS:
2093                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2094                                         break;
2095                                 }
2096                                 error = sooptcopyout(sopt, &optval,
2097                                     sizeof optval);
2098                                 break;
2099                         case IPV6_PKTINFO:
2100                         case IPV6_HOPOPTS:
2101                         case IPV6_RTHDR:
2102                         case IPV6_DSTOPTS:
2103                         case IPV6_RTHDRDSTOPTS:
2104                         case IPV6_NEXTHOP:
2105                         case IPV6_TCLASS:
2106                         case IPV6_DONTFRAG:
2107                         case IPV6_USE_MIN_MTU:
2108                         case IPV6_PREFER_TEMPADDR:
2109                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2110                                     optname, sopt);
2111                                 break;
2112
2113                         case IPV6_MULTICAST_IF:
2114                         case IPV6_MULTICAST_HOPS:
2115                         case IPV6_MULTICAST_LOOP:
2116                         case IPV6_MSFILTER:
2117                                 error = ip6_getmoptions(in6p, sopt);
2118                                 break;
2119
2120 #ifdef IPSEC
2121                         case IPV6_IPSEC_POLICY:
2122                           {
2123                                 caddr_t req = NULL;
2124                                 size_t len = 0;
2125                                 struct mbuf *m = NULL;
2126                                 struct mbuf **mp = &m;
2127                                 size_t ovalsize = sopt->sopt_valsize;
2128                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2129
2130                                 error = soopt_getm(sopt, &m); /* XXX */
2131                                 if (error != 0)
2132                                         break;
2133                                 error = soopt_mcopyin(sopt, m); /* XXX */
2134                                 if (error != 0)
2135                                         break;
2136                                 sopt->sopt_valsize = ovalsize;
2137                                 sopt->sopt_val = oval;
2138                                 if (m) {
2139                                         req = mtod(m, caddr_t);
2140                                         len = m->m_len;
2141                                 }
2142                                 error = ipsec_get_policy(in6p, req, len, mp);
2143                                 if (error == 0)
2144                                         error = soopt_mcopyout(sopt, m); /* XXX */
2145                                 if (error == 0 && m)
2146                                         m_freem(m);
2147                                 break;
2148                           }
2149 #endif /* IPSEC */
2150
2151                         default:
2152                                 error = ENOPROTOOPT;
2153                                 break;
2154                         }
2155                         break;
2156                 }
2157         }
2158         return (error);
2159 }
2160
2161 int
2162 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2163 {
2164         int error = 0, optval, optlen;
2165         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2166         struct inpcb *in6p = sotoinpcb(so);
2167         int level, op, optname;
2168
2169         level = sopt->sopt_level;
2170         op = sopt->sopt_dir;
2171         optname = sopt->sopt_name;
2172         optlen = sopt->sopt_valsize;
2173
2174         if (level != IPPROTO_IPV6) {
2175                 return (EINVAL);
2176         }
2177
2178         switch (optname) {
2179         case IPV6_CHECKSUM:
2180                 /*
2181                  * For ICMPv6 sockets, no modification allowed for checksum
2182                  * offset, permit "no change" values to help existing apps.
2183                  *
2184                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2185                  * for an ICMPv6 socket will fail."
2186                  * The current behavior does not meet RFC3542.
2187                  */
2188                 switch (op) {
2189                 case SOPT_SET:
2190                         if (optlen != sizeof(int)) {
2191                                 error = EINVAL;
2192                                 break;
2193                         }
2194                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2195                                             sizeof(optval));
2196                         if (error)
2197                                 break;
2198                         if ((optval % 2) != 0) {
2199                                 /* the API assumes even offset values */
2200                                 error = EINVAL;
2201                         } else if (so->so_proto->pr_protocol ==
2202                             IPPROTO_ICMPV6) {
2203                                 if (optval != icmp6off)
2204                                         error = EINVAL;
2205                         } else
2206                                 in6p->in6p_cksum = optval;
2207                         break;
2208
2209                 case SOPT_GET:
2210                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2211                                 optval = icmp6off;
2212                         else
2213                                 optval = in6p->in6p_cksum;
2214
2215                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2216                         break;
2217
2218                 default:
2219                         error = EINVAL;
2220                         break;
2221                 }
2222                 break;
2223
2224         default:
2225                 error = ENOPROTOOPT;
2226                 break;
2227         }
2228
2229         return (error);
2230 }
2231
2232 /*
2233  * Set up IP6 options in pcb for insertion in output packets or
2234  * specifying behavior of outgoing packets.
2235  */
2236 static int
2237 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2238     struct socket *so, struct sockopt *sopt)
2239 {
2240         struct ip6_pktopts *opt = *pktopt;
2241         int error = 0;
2242         struct thread *td = sopt->sopt_td;
2243
2244         /* turn off any old options. */
2245         if (opt) {
2246 #ifdef DIAGNOSTIC
2247                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2248                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2249                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2250                         printf("ip6_pcbopts: all specified options are cleared.\n");
2251 #endif
2252                 ip6_clearpktopts(opt, -1);
2253         } else
2254                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2255         *pktopt = NULL;
2256
2257         if (!m || m->m_len == 0) {
2258                 /*
2259                  * Only turning off any previous options, regardless of
2260                  * whether the opt is just created or given.
2261                  */
2262                 free(opt, M_IP6OPT);
2263                 return (0);
2264         }
2265
2266         /*  set options specified by user. */
2267         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2268             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2269                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2270                 free(opt, M_IP6OPT);
2271                 return (error);
2272         }
2273         *pktopt = opt;
2274         return (0);
2275 }
2276
2277 /*
2278  * initialize ip6_pktopts.  beware that there are non-zero default values in
2279  * the struct.
2280  */
2281 void
2282 ip6_initpktopts(struct ip6_pktopts *opt)
2283 {
2284
2285         bzero(opt, sizeof(*opt));
2286         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2287         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2288         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2289         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2290 }
2291
2292 static int
2293 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2294     struct ucred *cred, int uproto)
2295 {
2296         struct ip6_pktopts *opt;
2297
2298         if (*pktopt == NULL) {
2299                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2300                     M_WAITOK);
2301                 ip6_initpktopts(*pktopt);
2302         }
2303         opt = *pktopt;
2304
2305         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2306 }
2307
2308 static int
2309 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2310 {
2311         void *optdata = NULL;
2312         int optdatalen = 0;
2313         struct ip6_ext *ip6e;
2314         int error = 0;
2315         struct in6_pktinfo null_pktinfo;
2316         int deftclass = 0, on;
2317         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2318         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2319
2320         switch (optname) {
2321         case IPV6_PKTINFO:
2322                 if (pktopt && pktopt->ip6po_pktinfo)
2323                         optdata = (void *)pktopt->ip6po_pktinfo;
2324                 else {
2325                         /* XXX: we don't have to do this every time... */
2326                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2327                         optdata = (void *)&null_pktinfo;
2328                 }
2329                 optdatalen = sizeof(struct in6_pktinfo);
2330                 break;
2331         case IPV6_TCLASS:
2332                 if (pktopt && pktopt->ip6po_tclass >= 0)
2333                         optdata = (void *)&pktopt->ip6po_tclass;
2334                 else
2335                         optdata = (void *)&deftclass;
2336                 optdatalen = sizeof(int);
2337                 break;
2338         case IPV6_HOPOPTS:
2339                 if (pktopt && pktopt->ip6po_hbh) {
2340                         optdata = (void *)pktopt->ip6po_hbh;
2341                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2342                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2343                 }
2344                 break;
2345         case IPV6_RTHDR:
2346                 if (pktopt && pktopt->ip6po_rthdr) {
2347                         optdata = (void *)pktopt->ip6po_rthdr;
2348                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2349                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2350                 }
2351                 break;
2352         case IPV6_RTHDRDSTOPTS:
2353                 if (pktopt && pktopt->ip6po_dest1) {
2354                         optdata = (void *)pktopt->ip6po_dest1;
2355                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2356                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2357                 }
2358                 break;
2359         case IPV6_DSTOPTS:
2360                 if (pktopt && pktopt->ip6po_dest2) {
2361                         optdata = (void *)pktopt->ip6po_dest2;
2362                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2363                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2364                 }
2365                 break;
2366         case IPV6_NEXTHOP:
2367                 if (pktopt && pktopt->ip6po_nexthop) {
2368                         optdata = (void *)pktopt->ip6po_nexthop;
2369                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2370                 }
2371                 break;
2372         case IPV6_USE_MIN_MTU:
2373                 if (pktopt)
2374                         optdata = (void *)&pktopt->ip6po_minmtu;
2375                 else
2376                         optdata = (void *)&defminmtu;
2377                 optdatalen = sizeof(int);
2378                 break;
2379         case IPV6_DONTFRAG:
2380                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2381                         on = 1;
2382                 else
2383                         on = 0;
2384                 optdata = (void *)&on;
2385                 optdatalen = sizeof(on);
2386                 break;
2387         case IPV6_PREFER_TEMPADDR:
2388                 if (pktopt)
2389                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2390                 else
2391                         optdata = (void *)&defpreftemp;
2392                 optdatalen = sizeof(int);
2393                 break;
2394         default:                /* should not happen */
2395 #ifdef DIAGNOSTIC
2396                 panic("ip6_getpcbopt: unexpected option\n");
2397 #endif
2398                 return (ENOPROTOOPT);
2399         }
2400
2401         error = sooptcopyout(sopt, optdata, optdatalen);
2402
2403         return (error);
2404 }
2405
2406 void
2407 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2408 {
2409         if (pktopt == NULL)
2410                 return;
2411
2412         if (optname == -1 || optname == IPV6_PKTINFO) {
2413                 if (pktopt->ip6po_pktinfo)
2414                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2415                 pktopt->ip6po_pktinfo = NULL;
2416         }
2417         if (optname == -1 || optname == IPV6_HOPLIMIT)
2418                 pktopt->ip6po_hlim = -1;
2419         if (optname == -1 || optname == IPV6_TCLASS)
2420                 pktopt->ip6po_tclass = -1;
2421         if (optname == -1 || optname == IPV6_NEXTHOP) {
2422                 if (pktopt->ip6po_nextroute.ro_rt) {
2423                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2424                         pktopt->ip6po_nextroute.ro_rt = NULL;
2425                 }
2426                 if (pktopt->ip6po_nexthop)
2427                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2428                 pktopt->ip6po_nexthop = NULL;
2429         }
2430         if (optname == -1 || optname == IPV6_HOPOPTS) {
2431                 if (pktopt->ip6po_hbh)
2432                         free(pktopt->ip6po_hbh, M_IP6OPT);
2433                 pktopt->ip6po_hbh = NULL;
2434         }
2435         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2436                 if (pktopt->ip6po_dest1)
2437                         free(pktopt->ip6po_dest1, M_IP6OPT);
2438                 pktopt->ip6po_dest1 = NULL;
2439         }
2440         if (optname == -1 || optname == IPV6_RTHDR) {
2441                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2442                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2443                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2444                 if (pktopt->ip6po_route.ro_rt) {
2445                         RTFREE(pktopt->ip6po_route.ro_rt);
2446                         pktopt->ip6po_route.ro_rt = NULL;
2447                 }
2448         }
2449         if (optname == -1 || optname == IPV6_DSTOPTS) {
2450                 if (pktopt->ip6po_dest2)
2451                         free(pktopt->ip6po_dest2, M_IP6OPT);
2452                 pktopt->ip6po_dest2 = NULL;
2453         }
2454 }
2455
2456 #define PKTOPT_EXTHDRCPY(type) \
2457 do {\
2458         if (src->type) {\
2459                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2460                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2461                 if (dst->type == NULL && canwait == M_NOWAIT)\
2462                         goto bad;\
2463                 bcopy(src->type, dst->type, hlen);\
2464         }\
2465 } while (/*CONSTCOND*/ 0)
2466
2467 static int
2468 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2469 {
2470         if (dst == NULL || src == NULL)  {
2471                 printf("ip6_clearpktopts: invalid argument\n");
2472                 return (EINVAL);
2473         }
2474
2475         dst->ip6po_hlim = src->ip6po_hlim;
2476         dst->ip6po_tclass = src->ip6po_tclass;
2477         dst->ip6po_flags = src->ip6po_flags;
2478         dst->ip6po_minmtu = src->ip6po_minmtu;
2479         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2480         if (src->ip6po_pktinfo) {
2481                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2482                     M_IP6OPT, canwait);
2483                 if (dst->ip6po_pktinfo == NULL)
2484                         goto bad;
2485                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2486         }
2487         if (src->ip6po_nexthop) {
2488                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2489                     M_IP6OPT, canwait);
2490                 if (dst->ip6po_nexthop == NULL)
2491                         goto bad;
2492                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2493                     src->ip6po_nexthop->sa_len);
2494         }
2495         PKTOPT_EXTHDRCPY(ip6po_hbh);
2496         PKTOPT_EXTHDRCPY(ip6po_dest1);
2497         PKTOPT_EXTHDRCPY(ip6po_dest2);
2498         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2499         return (0);
2500
2501   bad:
2502         ip6_clearpktopts(dst, -1);
2503         return (ENOBUFS);
2504 }
2505 #undef PKTOPT_EXTHDRCPY
2506
2507 struct ip6_pktopts *
2508 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2509 {
2510         int error;
2511         struct ip6_pktopts *dst;
2512
2513         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2514         if (dst == NULL)
2515                 return (NULL);
2516         ip6_initpktopts(dst);
2517
2518         if ((error = copypktopts(dst, src, canwait)) != 0) {
2519                 free(dst, M_IP6OPT);
2520                 return (NULL);
2521         }
2522
2523         return (dst);
2524 }
2525
2526 void
2527 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2528 {
2529         if (pktopt == NULL)
2530                 return;
2531
2532         ip6_clearpktopts(pktopt, -1);
2533
2534         free(pktopt, M_IP6OPT);
2535 }
2536
2537 /*
2538  * Set IPv6 outgoing packet options based on advanced API.
2539  */
2540 int
2541 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2542     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2543 {
2544         struct cmsghdr *cm = 0;
2545
2546         if (control == NULL || opt == NULL)
2547                 return (EINVAL);
2548
2549         ip6_initpktopts(opt);
2550         if (stickyopt) {
2551                 int error;
2552
2553                 /*
2554                  * If stickyopt is provided, make a local copy of the options
2555                  * for this particular packet, then override them by ancillary
2556                  * objects.
2557                  * XXX: copypktopts() does not copy the cached route to a next
2558                  * hop (if any).  This is not very good in terms of efficiency,
2559                  * but we can allow this since this option should be rarely
2560                  * used.
2561                  */
2562                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2563                         return (error);
2564         }
2565
2566         /*
2567          * XXX: Currently, we assume all the optional information is stored
2568          * in a single mbuf.
2569          */
2570         if (control->m_next)
2571                 return (EINVAL);
2572
2573         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2574             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2575                 int error;
2576
2577                 if (control->m_len < CMSG_LEN(0))
2578                         return (EINVAL);
2579
2580                 cm = mtod(control, struct cmsghdr *);
2581                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2582                         return (EINVAL);
2583                 if (cm->cmsg_level != IPPROTO_IPV6)
2584                         continue;
2585
2586                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2587                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2588                 if (error)
2589                         return (error);
2590         }
2591
2592         return (0);
2593 }
2594
2595 /*
2596  * Set a particular packet option, as a sticky option or an ancillary data
2597  * item.  "len" can be 0 only when it's a sticky option.
2598  * We have 4 cases of combination of "sticky" and "cmsg":
2599  * "sticky=0, cmsg=0": impossible
2600  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2601  * "sticky=1, cmsg=0": RFC3542 socket option
2602  * "sticky=1, cmsg=1": RFC2292 socket option
2603  */
2604 static int
2605 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2606     struct ucred *cred, int sticky, int cmsg, int uproto)
2607 {
2608         int minmtupolicy, preftemp;
2609         int error;
2610
2611         if (!sticky && !cmsg) {
2612 #ifdef DIAGNOSTIC
2613                 printf("ip6_setpktopt: impossible case\n");
2614 #endif
2615                 return (EINVAL);
2616         }
2617
2618         /*
2619          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2620          * not be specified in the context of RFC3542.  Conversely,
2621          * RFC3542 types should not be specified in the context of RFC2292.
2622          */
2623         if (!cmsg) {
2624                 switch (optname) {
2625                 case IPV6_2292PKTINFO:
2626                 case IPV6_2292HOPLIMIT:
2627                 case IPV6_2292NEXTHOP:
2628                 case IPV6_2292HOPOPTS:
2629                 case IPV6_2292DSTOPTS:
2630                 case IPV6_2292RTHDR:
2631                 case IPV6_2292PKTOPTIONS:
2632                         return (ENOPROTOOPT);
2633                 }
2634         }
2635         if (sticky && cmsg) {
2636                 switch (optname) {
2637                 case IPV6_PKTINFO:
2638                 case IPV6_HOPLIMIT:
2639                 case IPV6_NEXTHOP:
2640                 case IPV6_HOPOPTS:
2641                 case IPV6_DSTOPTS:
2642                 case IPV6_RTHDRDSTOPTS:
2643                 case IPV6_RTHDR:
2644                 case IPV6_USE_MIN_MTU:
2645                 case IPV6_DONTFRAG:
2646                 case IPV6_TCLASS:
2647                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2648                         return (ENOPROTOOPT);
2649                 }
2650         }
2651
2652         switch (optname) {
2653         case IPV6_2292PKTINFO:
2654         case IPV6_PKTINFO:
2655         {
2656                 struct ifnet *ifp = NULL;
2657                 struct in6_pktinfo *pktinfo;
2658
2659                 if (len != sizeof(struct in6_pktinfo))
2660                         return (EINVAL);
2661
2662                 pktinfo = (struct in6_pktinfo *)buf;
2663
2664                 /*
2665                  * An application can clear any sticky IPV6_PKTINFO option by
2666                  * doing a "regular" setsockopt with ipi6_addr being
2667                  * in6addr_any and ipi6_ifindex being zero.
2668                  * [RFC 3542, Section 6]
2669                  */
2670                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2671                     pktinfo->ipi6_ifindex == 0 &&
2672                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2673                         ip6_clearpktopts(opt, optname);
2674                         break;
2675                 }
2676
2677                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2678                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2679                         return (EINVAL);
2680                 }
2681
2682                 /* validate the interface index if specified. */
2683                 if (pktinfo->ipi6_ifindex > V_if_index ||
2684                     pktinfo->ipi6_ifindex < 0) {
2685                          return (ENXIO);
2686                 }
2687                 if (pktinfo->ipi6_ifindex) {
2688                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2689                         if (ifp == NULL)
2690                                 return (ENXIO);
2691                 }
2692
2693                 /*
2694                  * We store the address anyway, and let in6_selectsrc()
2695                  * validate the specified address.  This is because ipi6_addr
2696                  * may not have enough information about its scope zone, and
2697                  * we may need additional information (such as outgoing
2698                  * interface or the scope zone of a destination address) to
2699                  * disambiguate the scope.
2700                  * XXX: the delay of the validation may confuse the
2701                  * application when it is used as a sticky option.
2702                  */
2703                 if (opt->ip6po_pktinfo == NULL) {
2704                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2705                             M_IP6OPT, M_NOWAIT);
2706                         if (opt->ip6po_pktinfo == NULL)
2707                                 return (ENOBUFS);
2708                 }
2709                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2710                 break;
2711         }
2712
2713         case IPV6_2292HOPLIMIT:
2714         case IPV6_HOPLIMIT:
2715         {
2716                 int *hlimp;
2717
2718                 /*
2719                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2720                  * to simplify the ordering among hoplimit options.
2721                  */
2722                 if (optname == IPV6_HOPLIMIT && sticky)
2723                         return (ENOPROTOOPT);
2724
2725                 if (len != sizeof(int))
2726                         return (EINVAL);
2727                 hlimp = (int *)buf;
2728                 if (*hlimp < -1 || *hlimp > 255)
2729                         return (EINVAL);
2730
2731                 opt->ip6po_hlim = *hlimp;
2732                 break;
2733         }
2734
2735         case IPV6_TCLASS:
2736         {
2737                 int tclass;
2738
2739                 if (len != sizeof(int))
2740                         return (EINVAL);
2741                 tclass = *(int *)buf;
2742                 if (tclass < -1 || tclass > 255)
2743                         return (EINVAL);
2744
2745                 opt->ip6po_tclass = tclass;
2746                 break;
2747         }
2748
2749         case IPV6_2292NEXTHOP:
2750         case IPV6_NEXTHOP:
2751                 if (cred != NULL) {
2752                         error = priv_check_cred(cred,
2753                             PRIV_NETINET_SETHDROPTS, 0);
2754                         if (error)
2755                                 return (error);
2756                 }
2757
2758                 if (len == 0) { /* just remove the option */
2759                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2760                         break;
2761                 }
2762
2763                 /* check if cmsg_len is large enough for sa_len */
2764                 if (len < sizeof(struct sockaddr) || len < *buf)
2765                         return (EINVAL);
2766
2767                 switch (((struct sockaddr *)buf)->sa_family) {
2768                 case AF_INET6:
2769                 {
2770                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2771                         int error;
2772
2773                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2774                                 return (EINVAL);
2775
2776                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2777                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2778                                 return (EINVAL);
2779                         }
2780                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2781                             != 0) {
2782                                 return (error);
2783                         }
2784                         break;
2785                 }
2786                 case AF_LINK:   /* should eventually be supported */
2787                 default:
2788                         return (EAFNOSUPPORT);
2789                 }
2790
2791                 /* turn off the previous option, then set the new option. */
2792                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2793                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2794                 if (opt->ip6po_nexthop == NULL)
2795                         return (ENOBUFS);
2796                 bcopy(buf, opt->ip6po_nexthop, *buf);
2797                 break;
2798
2799         case IPV6_2292HOPOPTS:
2800         case IPV6_HOPOPTS:
2801         {
2802                 struct ip6_hbh *hbh;
2803                 int hbhlen;
2804
2805                 /*
2806                  * XXX: We don't allow a non-privileged user to set ANY HbH
2807                  * options, since per-option restriction has too much
2808                  * overhead.
2809                  */
2810                 if (cred != NULL) {
2811                         error = priv_check_cred(cred,
2812                             PRIV_NETINET_SETHDROPTS, 0);
2813                         if (error)
2814                                 return (error);
2815                 }
2816
2817                 if (len == 0) {
2818                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2819                         break;  /* just remove the option */
2820                 }
2821
2822                 /* message length validation */
2823                 if (len < sizeof(struct ip6_hbh))
2824                         return (EINVAL);
2825                 hbh = (struct ip6_hbh *)buf;
2826                 hbhlen = (hbh->ip6h_len + 1) << 3;
2827                 if (len != hbhlen)
2828                         return (EINVAL);
2829
2830                 /* turn off the previous option, then set the new option. */
2831                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2832                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2833                 if (opt->ip6po_hbh == NULL)
2834                         return (ENOBUFS);
2835                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2836
2837                 break;
2838         }
2839
2840         case IPV6_2292DSTOPTS:
2841         case IPV6_DSTOPTS:
2842         case IPV6_RTHDRDSTOPTS:
2843         {
2844                 struct ip6_dest *dest, **newdest = NULL;
2845                 int destlen;
2846
2847                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2848                         error = priv_check_cred(cred,
2849                             PRIV_NETINET_SETHDROPTS, 0);
2850                         if (error)
2851                                 return (error);
2852                 }
2853
2854                 if (len == 0) {
2855                         ip6_clearpktopts(opt, optname);
2856                         break;  /* just remove the option */
2857                 }
2858
2859                 /* message length validation */
2860                 if (len < sizeof(struct ip6_dest))
2861                         return (EINVAL);
2862                 dest = (struct ip6_dest *)buf;
2863                 destlen = (dest->ip6d_len + 1) << 3;
2864                 if (len != destlen)
2865                         return (EINVAL);
2866
2867                 /*
2868                  * Determine the position that the destination options header
2869                  * should be inserted; before or after the routing header.
2870                  */
2871                 switch (optname) {
2872                 case IPV6_2292DSTOPTS:
2873                         /*
2874                          * The old advacned API is ambiguous on this point.
2875                          * Our approach is to determine the position based
2876                          * according to the existence of a routing header.
2877                          * Note, however, that this depends on the order of the
2878                          * extension headers in the ancillary data; the 1st
2879                          * part of the destination options header must appear
2880                          * before the routing header in the ancillary data,
2881                          * too.
2882                          * RFC3542 solved the ambiguity by introducing
2883                          * separate ancillary data or option types.
2884                          */
2885                         if (opt->ip6po_rthdr == NULL)
2886                                 newdest = &opt->ip6po_dest1;
2887                         else
2888                                 newdest = &opt->ip6po_dest2;
2889                         break;
2890                 case IPV6_RTHDRDSTOPTS:
2891                         newdest = &opt->ip6po_dest1;
2892                         break;
2893                 case IPV6_DSTOPTS:
2894                         newdest = &opt->ip6po_dest2;
2895                         break;
2896                 }
2897
2898                 /* turn off the previous option, then set the new option. */
2899                 ip6_clearpktopts(opt, optname);
2900                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2901                 if (*newdest == NULL)
2902                         return (ENOBUFS);
2903                 bcopy(dest, *newdest, destlen);
2904
2905                 break;
2906         }
2907
2908         case IPV6_2292RTHDR:
2909         case IPV6_RTHDR:
2910         {
2911                 struct ip6_rthdr *rth;
2912                 int rthlen;
2913
2914                 if (len == 0) {
2915                         ip6_clearpktopts(opt, IPV6_RTHDR);
2916                         break;  /* just remove the option */
2917                 }
2918
2919                 /* message length validation */
2920                 if (len < sizeof(struct ip6_rthdr))
2921                         return (EINVAL);
2922                 rth = (struct ip6_rthdr *)buf;
2923                 rthlen = (rth->ip6r_len + 1) << 3;
2924                 if (len != rthlen)
2925                         return (EINVAL);
2926
2927                 switch (rth->ip6r_type) {
2928                 case IPV6_RTHDR_TYPE_0:
2929                         if (rth->ip6r_len == 0) /* must contain one addr */
2930                                 return (EINVAL);
2931                         if (rth->ip6r_len % 2) /* length must be even */
2932                                 return (EINVAL);
2933                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2934                                 return (EINVAL);
2935                         break;
2936                 default:
2937                         return (EINVAL);        /* not supported */
2938                 }
2939
2940                 /* turn off the previous option */
2941                 ip6_clearpktopts(opt, IPV6_RTHDR);
2942                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2943                 if (opt->ip6po_rthdr == NULL)
2944                         return (ENOBUFS);
2945                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2946
2947                 break;
2948         }
2949
2950         case IPV6_USE_MIN_MTU:
2951                 if (len != sizeof(int))
2952                         return (EINVAL);
2953                 minmtupolicy = *(int *)buf;
2954                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2955                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2956                     minmtupolicy != IP6PO_MINMTU_ALL) {
2957                         return (EINVAL);
2958                 }
2959                 opt->ip6po_minmtu = minmtupolicy;
2960                 break;
2961
2962         case IPV6_DONTFRAG:
2963                 if (len != sizeof(int))
2964                         return (EINVAL);
2965
2966                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2967                         /*
2968                          * we ignore this option for TCP sockets.
2969                          * (RFC3542 leaves this case unspecified.)
2970                          */
2971                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2972                 } else
2973                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2974                 break;
2975
2976         case IPV6_PREFER_TEMPADDR:
2977                 if (len != sizeof(int))
2978                         return (EINVAL);
2979                 preftemp = *(int *)buf;
2980                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2981                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2982                     preftemp != IP6PO_TEMPADDR_PREFER) {
2983                         return (EINVAL);
2984                 }
2985                 opt->ip6po_prefer_tempaddr = preftemp;
2986                 break;
2987
2988         default:
2989                 return (ENOPROTOOPT);
2990         } /* end of switch */
2991
2992         return (0);
2993 }
2994
2995 /*
2996  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2997  * packet to the input queue of a specified interface.  Note that this
2998  * calls the output routine of the loopback "driver", but with an interface
2999  * pointer that might NOT be &loif -- easier than replicating that code here.
3000  */
3001 void
3002 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3003 {
3004         struct mbuf *copym;
3005         struct ip6_hdr *ip6;
3006
3007         copym = m_copy(m, 0, M_COPYALL);
3008         if (copym == NULL)
3009                 return;
3010
3011         /*
3012          * Make sure to deep-copy IPv6 header portion in case the data
3013          * is in an mbuf cluster, so that we can safely override the IPv6
3014          * header portion later.
3015          */
3016         if ((copym->m_flags & M_EXT) != 0 ||
3017             copym->m_len < sizeof(struct ip6_hdr)) {
3018                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3019                 if (copym == NULL)
3020                         return;
3021         }
3022
3023 #ifdef DIAGNOSTIC
3024         if (copym->m_len < sizeof(*ip6)) {
3025                 m_freem(copym);
3026                 return;
3027         }
3028 #endif
3029
3030         ip6 = mtod(copym, struct ip6_hdr *);
3031         /*
3032          * clear embedded scope identifiers if necessary.
3033          * in6_clearscope will touch the addresses only when necessary.
3034          */
3035         in6_clearscope(&ip6->ip6_src);
3036         in6_clearscope(&ip6->ip6_dst);
3037
3038         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3039 }
3040
3041 /*
3042  * Chop IPv6 header off from the payload.
3043  */
3044 static int
3045 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3046 {
3047         struct mbuf *mh;
3048         struct ip6_hdr *ip6;
3049
3050         ip6 = mtod(m, struct ip6_hdr *);
3051         if (m->m_len > sizeof(*ip6)) {
3052                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3053                 if (mh == 0) {
3054                         m_freem(m);
3055                         return ENOBUFS;
3056                 }
3057                 M_MOVE_PKTHDR(mh, m);
3058                 MH_ALIGN(mh, sizeof(*ip6));
3059                 m->m_len -= sizeof(*ip6);
3060                 m->m_data += sizeof(*ip6);
3061                 mh->m_next = m;
3062                 m = mh;
3063                 m->m_len = sizeof(*ip6);
3064                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3065         }
3066         exthdrs->ip6e_ip6 = m;
3067         return 0;
3068 }
3069
3070 /*
3071  * Compute IPv6 extension header length.
3072  */
3073 int
3074 ip6_optlen(struct inpcb *in6p)
3075 {
3076         int len;
3077
3078         if (!in6p->in6p_outputopts)
3079                 return 0;
3080
3081         len = 0;
3082 #define elen(x) \
3083     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3084
3085         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3086         if (in6p->in6p_outputopts->ip6po_rthdr)
3087                 /* dest1 is valid with rthdr only */
3088                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3089         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3090         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3091         return len;
3092 #undef elen
3093 }