sys/netinet6/ip6_output.c

   1 /*      $FreeBSD$       */
   2 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   3
   4 /*-
   5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the project nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  */
  32
  33 /*-
  34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  35  *      The Regents of the University of California.  All rights reserved.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 4. Neither the name of the University nor the names of its contributors
  46  *    may be used to endorse or promote products derived from this software
  47  *    without specific prior written permission.
  48  *
  49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  59  * SUCH DAMAGE.
  60  *
  61  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  62  */
  63
  64 #include "opt_inet.h"
  65 #include "opt_inet6.h"
  66 #include "opt_ipsec.h"
  67
  68 #include <sys/param.h>
  69 #include <sys/malloc.h>
  70 #include <sys/mbuf.h>
  71 #include <sys/proc.h>
  72 #include <sys/errno.h>
  73 #include <sys/protosw.h>
  74 #include <sys/socket.h>
  75 #include <sys/socketvar.h>
  76 #include <sys/systm.h>
  77 #include <sys/kernel.h>
  78
  79 #include <net/if.h>
  80 #include <net/netisr.h>
  81 #include <net/route.h>
  82 #include <net/pfil.h>
  83
  84 #include <netinet/in.h>
  85 #include <netinet/in_var.h>
  86 #include <netinet6/in6_var.h>
  87 #include <netinet/ip6.h>
  88 #include <netinet/icmp6.h>
  89 #include <netinet6/ip6_var.h>
  90 #include <netinet/in_pcb.h>
  91 #include <netinet/tcp_var.h>
  92 #include <netinet6/nd6.h>
  93
  94 #ifdef IPSEC
  95 #include <netinet6/ipsec.h>
  96 #ifdef INET6
  97 #include <netinet6/ipsec6.h>
  98 #endif
  99 #include <netkey/key.h>
 100 #endif /* IPSEC */
 101
 102 #ifdef FAST_IPSEC
 103 #include <netipsec/ipsec.h>
 104 #include <netipsec/ipsec6.h>
 105 #include <netipsec/key.h>
 106 #endif /* FAST_IPSEC */
 107
 108 #include <netinet6/ip6protosw.h>
 109 #include <netinet6/scope6_var.h>
 110
 111 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
 112
 113 struct ip6_exthdrs {
 114         struct mbuf *ip6e_ip6;
 115         struct mbuf *ip6e_hbh;
 116         struct mbuf *ip6e_dest1;
 117         struct mbuf *ip6e_rthdr;
 118         struct mbuf *ip6e_dest2;
 119 };
 120
 121 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 122                            int, int));
 123 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 124         struct socket *, struct sockopt *));
 125 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
 126 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int,
 127         int, int, int));
 128
 129 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
 130 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
 131 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
 132 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 133         struct ip6_frag **));
 134 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
 135 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
 136 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 137         struct ifnet *, struct in6_addr *, u_long *, int *));
 138 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int));
 139
 140
 141 /*
 142  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 143  * header (with pri, len, nxt, hlim, src, dst).
 144  * This function may modify ver and hlim only.
 145  * The mbuf chain containing the packet will be freed.
 146  * The mbuf opt, if present, will not be freed.
 147  *
 148  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 149  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 150  * which is rt_rmx.rmx_mtu.
 151  */
 152 int
 153 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
 154         struct mbuf *m0;
 155         struct ip6_pktopts *opt;
 156         struct route_in6 *ro;
 157         int flags;
 158         struct ip6_moptions *im6o;
 159         struct ifnet **ifpp;            /* XXX: just for statistics */
 160         struct inpcb *inp;
 161 {
 162         struct ip6_hdr *ip6, *mhip6;
 163         struct ifnet *ifp, *origifp;
 164         struct mbuf *m = m0;
 165         int hlen, tlen, len, off;
 166         struct route_in6 ip6route;
 167         struct rtentry *rt = NULL;
 168         struct sockaddr_in6 *dst, src_sa, dst_sa;
 169         struct in6_addr odst;
 170         int error = 0;
 171         struct in6_ifaddr *ia = NULL;
 172         u_long mtu;
 173         int alwaysfrag, dontfrag;
 174         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 175         struct ip6_exthdrs exthdrs;
 176         struct in6_addr finaldst, src0, dst0;
 177         u_int32_t zone;
 178         struct route_in6 *ro_pmtu = NULL;
 179         int hdrsplit = 0;
 180         int needipsec = 0;
 181 #if defined(IPSEC) || defined(FAST_IPSEC)
 182         int needipsectun = 0;
 183         struct secpolicy *sp = NULL;
 184 #endif /*IPSEC || FAST_IPSEC*/
 185
 186         ip6 = mtod(m, struct ip6_hdr *);
 187         finaldst = ip6->ip6_dst;
 188
 189 #define MAKE_EXTHDR(hp, mp)                                             \
 190     do {                                                                \
 191         if (hp) {                                                       \
 192                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 193                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 194                     ((eh)->ip6e_len + 1) << 3);                         \
 195                 if (error)                                              \
 196                         goto freehdrs;                                  \
 197         }                                                               \
 198     } while (/*CONSTCOND*/ 0)
 199
 200         bzero(&exthdrs, sizeof(exthdrs));
 201
 202         if (opt) {
 203                 /* Hop-by-Hop options header */
 204                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 205                 /* Destination options header(1st part) */
 206                 if (opt->ip6po_rthdr) {
 207                         /*
 208                          * Destination options header(1st part)
 209                          * This only makes sence with a routing header.
 210                          * See Section 9.2 of RFC 3542.
 211                          * Disabling this part just for MIP6 convenience is
 212                          * a bad idea.  We need to think carefully about a
 213                          * way to make the advanced API coexist with MIP6
 214                          * options, which might automatically be inserted in
 215                          * the kernel.
 216                          */
 217                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 218                 }
 219                 /* Routing header */
 220                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 221                 /* Destination options header(2nd part) */
 222                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 223         }
 224
 225 #ifdef IPSEC
 226         /* get a security policy for this packet */
 227         if (inp == NULL)
 228                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 229         else
 230                 sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
 231
 232         if (sp == NULL) {
 233                 ipsec6stat.out_inval++;
 234                 goto freehdrs;
 235         }
 236
 237         error = 0;
 238
 239         /* check policy */
 240         switch (sp->policy) {
 241         case IPSEC_POLICY_DISCARD:
 242                 /*
 243                  * This packet is just discarded.
 244                  */
 245                 ipsec6stat.out_polvio++;
 246                 goto freehdrs;
 247
 248         case IPSEC_POLICY_BYPASS:
 249         case IPSEC_POLICY_NONE:
 250                 /* no need to do IPsec. */
 251                 needipsec = 0;
 252                 break;
 253
 254         case IPSEC_POLICY_IPSEC:
 255                 if (sp->req == NULL) {
 256                         /* acquire a policy */
 257                         error = key_spdacquire(sp);
 258                         goto freehdrs;
 259                 }
 260                 needipsec = 1;
 261                 break;
 262
 263         case IPSEC_POLICY_ENTRUST:
 264         default:
 265                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 266         }
 267 #endif /* IPSEC */
 268 #ifdef FAST_IPSEC
 269         /* get a security policy for this packet */
 270         if (inp == NULL)
 271                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 272         else
 273                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 274
 275         if (sp == NULL) {
 276                 newipsecstat.ips_out_inval++;
 277                 goto freehdrs;
 278         }
 279
 280         error = 0;
 281
 282         /* check policy */
 283         switch (sp->policy) {
 284         case IPSEC_POLICY_DISCARD:
 285                 /*
 286                  * This packet is just discarded.
 287                  */
 288                 newipsecstat.ips_out_polvio++;
 289                 goto freehdrs;
 290
 291         case IPSEC_POLICY_BYPASS:
 292         case IPSEC_POLICY_NONE:
 293                 /* no need to do IPsec. */
 294                 needipsec = 0;
 295                 break;
 296
 297         case IPSEC_POLICY_IPSEC:
 298                 if (sp->req == NULL) {
 299                         /* acquire a policy */
 300                         error = key_spdacquire(sp);
 301                         goto freehdrs;
 302                 }
 303                 needipsec = 1;
 304                 break;
 305
 306         case IPSEC_POLICY_ENTRUST:
 307         default:
 308                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 309         }
 310 #endif /* FAST_IPSEC */
 311
 312         /*
 313          * Calculate the total length of the extension header chain.
 314          * Keep the length of the unfragmentable part for fragmentation.
 315          */
 316         optlen = 0;
 317         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 318         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 319         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 320         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 321         /* NOTE: we don't add AH/ESP length here. do that later. */
 322         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 323
 324         /*
 325          * If we need IPsec, or there is at least one extension header,
 326          * separate IP6 header from the payload.
 327          */
 328         if ((needipsec || optlen) && !hdrsplit) {
 329                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 330                         m = NULL;
 331                         goto freehdrs;
 332                 }
 333                 m = exthdrs.ip6e_ip6;
 334                 hdrsplit++;
 335         }
 336
 337         /* adjust pointer */
 338         ip6 = mtod(m, struct ip6_hdr *);
 339
 340         /* adjust mbuf packet header length */
 341         m->m_pkthdr.len += optlen;
 342         plen = m->m_pkthdr.len - sizeof(*ip6);
 343
 344         /* If this is a jumbo payload, insert a jumbo payload option. */
 345         if (plen > IPV6_MAXPACKET) {
 346                 if (!hdrsplit) {
 347                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 348                                 m = NULL;
 349                                 goto freehdrs;
 350                         }
 351                         m = exthdrs.ip6e_ip6;
 352                         hdrsplit++;
 353                 }
 354                 /* adjust pointer */
 355                 ip6 = mtod(m, struct ip6_hdr *);
 356                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 357                         goto freehdrs;
 358                 ip6->ip6_plen = 0;
 359         } else
 360                 ip6->ip6_plen = htons(plen);
 361
 362         /*
 363          * Concatenate headers and fill in next header fields.
 364          * Here we have, on "m"
 365          *      IPv6 payload
 366          * and we insert headers accordingly.  Finally, we should be getting:
 367          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 368          *
 369          * during the header composing process, "m" points to IPv6 header.
 370          * "mprev" points to an extension header prior to esp.
 371          */
 372         {
 373                 u_char *nexthdrp = &ip6->ip6_nxt;
 374                 struct mbuf *mprev = m;
 375
 376                 /*
 377                  * we treat dest2 specially.  this makes IPsec processing
 378                  * much easier.  the goal here is to make mprev point the
 379                  * mbuf prior to dest2.
 380                  *
 381                  * result: IPv6 dest2 payload
 382                  * m and mprev will point to IPv6 header.
 383                  */
 384                 if (exthdrs.ip6e_dest2) {
 385                         if (!hdrsplit)
 386                                 panic("assumption failed: hdr not split");
 387                         exthdrs.ip6e_dest2->m_next = m->m_next;
 388                         m->m_next = exthdrs.ip6e_dest2;
 389                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 390                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 391                 }
 392
 393 #define MAKE_CHAIN(m, mp, p, i)\
 394     do {\
 395         if (m) {\
 396                 if (!hdrsplit) \
 397                         panic("assumption failed: hdr not split"); \
 398                 *mtod((m), u_char *) = *(p);\
 399                 *(p) = (i);\
 400                 p = mtod((m), u_char *);\
 401                 (m)->m_next = (mp)->m_next;\
 402                 (mp)->m_next = (m);\
 403                 (mp) = (m);\
 404         }\
 405     } while (/*CONSTCOND*/ 0)
 406                 /*
 407                  * result: IPv6 hbh dest1 rthdr dest2 payload
 408                  * m will point to IPv6 header.  mprev will point to the
 409                  * extension header prior to dest2 (rthdr in the above case).
 410                  */
 411                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 412                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 413                     IPPROTO_DSTOPTS);
 414                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 415                     IPPROTO_ROUTING);
 416
 417 #if defined(IPSEC) || defined(FAST_IPSEC)
 418                 if (!needipsec)
 419                         goto skip_ipsec2;
 420
 421                 /*
 422                  * pointers after IPsec headers are not valid any more.
 423                  * other pointers need a great care too.
 424                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 425                  */
 426                 exthdrs.ip6e_dest2 = NULL;
 427
 428             {
 429                 struct ip6_rthdr *rh = NULL;
 430                 int segleft_org = 0;
 431                 struct ipsec_output_state state;
 432
 433                 if (exthdrs.ip6e_rthdr) {
 434                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 435                         segleft_org = rh->ip6r_segleft;
 436                         rh->ip6r_segleft = 0;
 437                 }
 438
 439                 bzero(&state, sizeof(state));
 440                 state.m = m;
 441                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 442                     &needipsectun);
 443                 m = state.m;
 444                 if (error) {
 445                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 446                         m = NULL;
 447                         switch (error) {
 448                         case EHOSTUNREACH:
 449                         case ENETUNREACH:
 450                         case EMSGSIZE:
 451                         case ENOBUFS:
 452                         case ENOMEM:
 453                                 break;
 454                         default:
 455                                 printf("ip6_output (ipsec): error code %d\n", error);
 456                                 /* FALLTHROUGH */
 457                         case ENOENT:
 458                                 /* don't show these error codes to the user */
 459                                 error = 0;
 460                                 break;
 461                         }
 462                         goto bad;
 463                 }
 464                 if (exthdrs.ip6e_rthdr) {
 465                         /* ah6_output doesn't modify mbuf chain */
 466                         rh->ip6r_segleft = segleft_org;
 467                 }
 468             }
 469 skip_ipsec2:;
 470 #endif
 471         }
 472
 473         /*
 474          * If there is a routing header, replace the destination address field
 475          * with the first hop of the routing header.
 476          */
 477         if (exthdrs.ip6e_rthdr) {
 478                 struct ip6_rthdr *rh =
 479                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 480                                                   struct ip6_rthdr *));
 481                 struct ip6_rthdr0 *rh0;
 482                 struct in6_addr *addr;
 483                 struct sockaddr_in6 sa;
 484
 485                 switch (rh->ip6r_type) {
 486                 case IPV6_RTHDR_TYPE_0:
 487                          rh0 = (struct ip6_rthdr0 *)rh;
 488                          addr = (struct in6_addr *)(rh0 + 1);
 489
 490                          /*
 491                           * construct a sockaddr_in6 form of
 492                           * the first hop.
 493                           *
 494                           * XXX: we may not have enough
 495                           * information about its scope zone;
 496                           * there is no standard API to pass
 497                           * the information from the
 498                           * application.
 499                           */
 500                          bzero(&sa, sizeof(sa));
 501                          sa.sin6_family = AF_INET6;
 502                          sa.sin6_len = sizeof(sa);
 503                          sa.sin6_addr = addr[0];
 504                          if ((error = sa6_embedscope(&sa,
 505                              ip6_use_defzone)) != 0) {
 506                                  goto bad;
 507                          }
 508                          ip6->ip6_dst = sa.sin6_addr;
 509                          bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
 510                              * (rh0->ip6r0_segleft - 1));
 511                          addr[rh0->ip6r0_segleft - 1] = finaldst;
 512                          /* XXX */
 513                          in6_clearscope(addr + rh0->ip6r0_segleft - 1);
 514                          break;
 515                 default:        /* is it possible? */
 516                          error = EINVAL;
 517                          goto bad;
 518                 }
 519         }
 520
 521         /* Source address validation */
 522         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 523             (flags & IPV6_UNSPECSRC) == 0) {
 524                 error = EOPNOTSUPP;
 525                 ip6stat.ip6s_badscope++;
 526                 goto bad;
 527         }
 528         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 529                 error = EOPNOTSUPP;
 530                 ip6stat.ip6s_badscope++;
 531                 goto bad;
 532         }
 533
 534         ip6stat.ip6s_localout++;
 535
 536         /*
 537          * Route packet.
 538          */
 539         if (ro == 0) {
 540                 ro = &ip6route;
 541                 bzero((caddr_t)ro, sizeof(*ro));
 542         }
 543         ro_pmtu = ro;
 544         if (opt && opt->ip6po_rthdr)
 545                 ro = &opt->ip6po_route;
 546         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 547
 548 again:
 549         /*
 550          * if specified, try to fill in the traffic class field.
 551          * do not override if a non-zero value is already set.
 552          * we check the diffserv field and the ecn field separately.
 553          */
 554         if (opt && opt->ip6po_tclass >= 0) {
 555                 int mask = 0;
 556
 557                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 558                         mask |= 0xfc;
 559                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 560                         mask |= 0x03;
 561                 if (mask != 0)
 562                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 563         }
 564
 565         /* fill in or override the hop limit field, if necessary. */
 566         if (opt && opt->ip6po_hlim != -1)
 567                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 568         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 569                 if (im6o != NULL)
 570                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 571                 else
 572                         ip6->ip6_hlim = ip6_defmcasthlim;
 573         }
 574
 575 #if defined(IPSEC) || defined(FAST_IPSEC)
 576         if (needipsec && needipsectun) {
 577                 struct ipsec_output_state state;
 578
 579                 /*
 580                  * All the extension headers will become inaccessible
 581                  * (since they can be encrypted).
 582                  * Don't panic, we need no more updates to extension headers
 583                  * on inner IPv6 packet (since they are now encapsulated).
 584                  *
 585                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 586                  */
 587                 bzero(&exthdrs, sizeof(exthdrs));
 588                 exthdrs.ip6e_ip6 = m;
 589
 590                 bzero(&state, sizeof(state));
 591                 state.m = m;
 592                 state.ro = (struct route *)ro;
 593                 state.dst = (struct sockaddr *)dst;
 594
 595                 error = ipsec6_output_tunnel(&state, sp, flags);
 596
 597                 m = state.m;
 598                 ro = (struct route_in6 *)state.ro;
 599                 dst = (struct sockaddr_in6 *)state.dst;
 600                 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("ip6_output (ipsec): error code %d\n", error);
 613                                 /* FALLTHROUGH */
 614                         case ENOENT:
 615                                 /* don't show these error codes to the user */
 616                                 error = 0;
 617                                 break;
 618                         }
 619                         goto bad;
 620                 }
 621
 622                 exthdrs.ip6e_ip6 = m;
 623         }
 624 #endif /* IPSEC */
 625
 626         /* adjust pointer */
 627         ip6 = mtod(m, struct ip6_hdr *);
 628
 629         bzero(&dst_sa, sizeof(dst_sa));
 630         dst_sa.sin6_family = AF_INET6;
 631         dst_sa.sin6_len = sizeof(dst_sa);
 632         dst_sa.sin6_addr = ip6->ip6_dst;
 633         if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
 634             &ifp, &rt, 0)) != 0) {
 635                 switch (error) {
 636                 case EHOSTUNREACH:
 637                         ip6stat.ip6s_noroute++;
 638                         break;
 639                 case EADDRNOTAVAIL:
 640                 default:
 641                         break; /* XXX statistics? */
 642                 }
 643                 if (ifp != NULL)
 644                         in6_ifstat_inc(ifp, ifs6_out_discard);
 645                 goto bad;
 646         }
 647         if (rt == NULL) {
 648                 /*
 649                  * If in6_selectroute() does not return a route entry,
 650                  * dst may not have been updated.
 651                  */
 652                 *dst = dst_sa;  /* XXX */
 653         }
 654
 655         /*
 656          * then rt (for unicast) and ifp must be non-NULL valid values.
 657          */
 658         if ((flags & IPV6_FORWARDING) == 0) {
 659                 /* XXX: the FORWARDING flag can be set for mrouting. */
 660                 in6_ifstat_inc(ifp, ifs6_out_request);
 661         }
 662         if (rt != NULL) {
 663                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 664                 rt->rt_use++;
 665         }
 666
 667         /*
 668          * The outgoing interface must be in the zone of source and
 669          * destination addresses.  We should use ia_ifp to support the
 670          * case of sending packets to an address of our own.
 671          */
 672         if (ia != NULL && ia->ia_ifp)
 673                 origifp = ia->ia_ifp;
 674         else
 675                 origifp = ifp;
 676
 677         src0 = ip6->ip6_src;
 678         if (in6_setscope(&src0, origifp, &zone))
 679                 goto badscope;
 680         bzero(&src_sa, sizeof(src_sa));
 681         src_sa.sin6_family = AF_INET6;
 682         src_sa.sin6_len = sizeof(src_sa);
 683         src_sa.sin6_addr = ip6->ip6_src;
 684         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 685                 goto badscope;
 686
 687         dst0 = ip6->ip6_dst;
 688         if (in6_setscope(&dst0, origifp, &zone))
 689                 goto badscope;
 690         /* re-initialize to be sure */
 691         bzero(&dst_sa, sizeof(dst_sa));
 692         dst_sa.sin6_family = AF_INET6;
 693         dst_sa.sin6_len = sizeof(dst_sa);
 694         dst_sa.sin6_addr = ip6->ip6_dst;
 695         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 696                 goto badscope;
 697         }
 698
 699         /* scope check is done. */
 700         goto routefound;
 701
 702   badscope:
 703         ip6stat.ip6s_badscope++;
 704         in6_ifstat_inc(origifp, ifs6_out_discard);
 705         if (error == 0)
 706                 error = EHOSTUNREACH; /* XXX */
 707         goto bad;
 708
 709   routefound:
 710         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 711                 if (opt && opt->ip6po_nextroute.ro_rt) {
 712                         /*
 713                          * The nexthop is explicitly specified by the
 714                          * application.  We assume the next hop is an IPv6
 715                          * address.
 716                          */
 717                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 718                 }
 719                 else if ((rt->rt_flags & RTF_GATEWAY))
 720                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 721         }
 722
 723         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 724                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 725         } else {
 726                 struct  in6_multi *in6m;
 727
 728                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 729
 730                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 731
 732                 /*
 733                  * Confirm that the outgoing interface supports multicast.
 734                  */
 735                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 736                         ip6stat.ip6s_noroute++;
 737                         in6_ifstat_inc(ifp, ifs6_out_discard);
 738                         error = ENETUNREACH;
 739                         goto bad;
 740                 }
 741                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 742                 if (in6m != NULL &&
 743                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 744                         /*
 745                          * If we belong to the destination multicast group
 746                          * on the outgoing interface, and the caller did not
 747                          * forbid loopback, loop back a copy.
 748                          */
 749                         ip6_mloopback(ifp, m, dst);
 750                 } else {
 751                         /*
 752                          * If we are acting as a multicast router, perform
 753                          * multicast forwarding as if the packet had just
 754                          * arrived on the interface to which we are about
 755                          * to send.  The multicast forwarding function
 756                          * recursively calls this function, using the
 757                          * IPV6_FORWARDING flag to prevent infinite recursion.
 758                          *
 759                          * Multicasts that are looped back by ip6_mloopback(),
 760                          * above, will be forwarded by the ip6_input() routine,
 761                          * if necessary.
 762                          */
 763                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 764                                 /*
 765                                  * XXX: ip6_mforward expects that rcvif is NULL
 766                                  * when it is called from the originating path.
 767                                  * However, it is not always the case, since
 768                                  * some versions of MGETHDR() does not
 769                                  * initialize the field.
 770                                  */
 771                                 m->m_pkthdr.rcvif = NULL;
 772                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 773                                         m_freem(m);
 774                                         goto done;
 775                                 }
 776                         }
 777                 }
 778                 /*
 779                  * Multicasts with a hoplimit of zero may be looped back,
 780                  * above, but must not be transmitted on a network.
 781                  * Also, multicasts addressed to the loopback interface
 782                  * are not sent -- the above call to ip6_mloopback() will
 783                  * loop back a copy if this host actually belongs to the
 784                  * destination group on the loopback interface.
 785                  */
 786                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 787                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 788                         m_freem(m);
 789                         goto done;
 790                 }
 791         }
 792
 793         /*
 794          * Fill the outgoing inteface to tell the upper layer
 795          * to increment per-interface statistics.
 796          */
 797         if (ifpp)
 798                 *ifpp = ifp;
 799
 800         /* Determine path MTU. */
 801         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 802             &alwaysfrag)) != 0)
 803                 goto bad;
 804
 805         /*
 806          * The caller of this function may specify to use the minimum MTU
 807          * in some cases.
 808          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 809          * setting.  The logic is a bit complicated; by default, unicast
 810          * packets will follow path MTU while multicast packets will be sent at
 811          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 812          * including unicast ones will be sent at the minimum MTU.  Multicast
 813          * packets will always be sent at the minimum MTU unless
 814          * IP6PO_MINMTU_DISABLE is explicitly specified.
 815          * See RFC 3542 for more details.
 816          */
 817         if (mtu > IPV6_MMTU) {
 818                 if ((flags & IPV6_MINMTU))
 819                         mtu = IPV6_MMTU;
 820                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 821                         mtu = IPV6_MMTU;
 822                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 823                          (opt == NULL ||
 824                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 825                         mtu = IPV6_MMTU;
 826                 }
 827         }
 828
 829         /*
 830          * clear embedded scope identifiers if necessary.
 831          * in6_clearscope will touch the addresses only when necessary.
 832          */
 833         in6_clearscope(&ip6->ip6_src);
 834         in6_clearscope(&ip6->ip6_dst);
 835
 836         /*
 837          * If the outgoing packet contains a hop-by-hop options header,
 838          * it must be examined and processed even by the source node.
 839          * (RFC 2460, section 4.)
 840          */
 841         if (exthdrs.ip6e_hbh) {
 842                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 843                 u_int32_t dummy; /* XXX unused */
 844                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 845
 846 #ifdef DIAGNOSTIC
 847                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 848                         panic("ip6e_hbh is not continuous");
 849 #endif
 850                 /*
 851                  *  XXX: if we have to send an ICMPv6 error to the sender,
 852                  *       we need the M_LOOP flag since icmp6_error() expects
 853                  *       the IPv6 and the hop-by-hop options header are
 854                  *       continuous unless the flag is set.
 855                  */
 856                 m->m_flags |= M_LOOP;
 857                 m->m_pkthdr.rcvif = ifp;
 858                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 859                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 860                     &dummy, &plen) < 0) {
 861                         /* m was already freed at this point */
 862                         error = EINVAL;/* better error? */
 863                         goto done;
 864                 }
 865                 m->m_flags &= ~M_LOOP; /* XXX */
 866                 m->m_pkthdr.rcvif = NULL;
 867         }
 868
 869         /* Jump over all PFIL processing if hooks are not active. */
 870         if (!PFIL_HOOKED(&inet6_pfil_hook))
 871                 goto passout;
 872
 873         odst = ip6->ip6_dst;
 874         /* Run through list of hooks for output packets. */
 875         error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 876         if (error != 0 || m == NULL)
 877                 goto done;
 878         ip6 = mtod(m, struct ip6_hdr *);
 879
 880         /* See if destination IP address was changed by packet filter. */
 881         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 882                 m->m_flags |= M_SKIP_FIREWALL;
 883                 /* If destination is now ourself drop to ip6_input(). */
 884                 if (in6_localaddr(&ip6->ip6_dst)) {
 885                         if (m->m_pkthdr.rcvif == NULL)
 886                                 m->m_pkthdr.rcvif = loif;
 887                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 888                                 m->m_pkthdr.csum_flags |=
 889                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 890                                 m->m_pkthdr.csum_data = 0xffff;
 891                         }
 892                         m->m_pkthdr.csum_flags |=
 893                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 894                         error = netisr_queue(NETISR_IPV6, m);
 895                         goto done;
 896                 } else
 897                         goto again;     /* Redo the routing table lookup. */
 898         }
 899
 900         /* XXX: IPFIREWALL_FORWARD */
 901
 902 passout:
 903         /*
 904          * Send the packet to the outgoing interface.
 905          * If necessary, do IPv6 fragmentation before sending.
 906          *
 907          * the logic here is rather complex:
 908          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 909          * 1-a: send as is if tlen <= path mtu
 910          * 1-b: fragment if tlen > path mtu
 911          *
 912          * 2: if user asks us not to fragment (dontfrag == 1)
 913          * 2-a: send as is if tlen <= interface mtu
 914          * 2-b: error if tlen > interface mtu
 915          *
 916          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 917          *      always fragment
 918          *
 919          * 4: if dontfrag == 1 && alwaysfrag == 1
 920          *      error, as we cannot handle this conflicting request
 921          */
 922         tlen = m->m_pkthdr.len;
 923
 924         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 925                 dontfrag = 1;
 926         else
 927                 dontfrag = 0;
 928         if (dontfrag && alwaysfrag) {   /* case 4 */
 929                 /* conflicting request - can't transmit */
 930                 error = EMSGSIZE;
 931                 goto bad;
 932         }
 933         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 934                 /*
 935                  * Even if the DONTFRAG option is specified, we cannot send the
 936                  * packet when the data length is larger than the MTU of the
 937                  * outgoing interface.
 938                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 939                  * well as returning an error code (the latter is not described
 940                  * in the API spec.)
 941                  */
 942                 u_int32_t mtu32;
 943                 struct ip6ctlparam ip6cp;
 944
 945                 mtu32 = (u_int32_t)mtu;
 946                 bzero(&ip6cp, sizeof(ip6cp));
 947                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 948                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 949                     (void *)&ip6cp);
 950
 951                 error = EMSGSIZE;
 952                 goto bad;
 953         }
 954
 955         /*
 956          * transmit packet without fragmentation
 957          */
 958         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 959                 struct in6_ifaddr *ia6;
 960
 961                 ip6 = mtod(m, struct ip6_hdr *);
 962                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 963                 if (ia6) {
 964                         /* Record statistics for this interface address. */
 965                         ia6->ia_ifa.if_opackets++;
 966                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 967                 }
 968 #ifdef IPSEC
 969                 /* clean ipsec history once it goes out of the node */
 970                 ipsec_delaux(m);
 971 #endif
 972                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 973                 goto done;
 974         }
 975
 976         /*
 977          * try to fragment the packet.  case 1-b and 3
 978          */
 979         if (mtu < IPV6_MMTU) {
 980                 /* path MTU cannot be less than IPV6_MMTU */
 981                 error = EMSGSIZE;
 982                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 983                 goto bad;
 984         } else if (ip6->ip6_plen == 0) {
 985                 /* jumbo payload cannot be fragmented */
 986                 error = EMSGSIZE;
 987                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 988                 goto bad;
 989         } else {
 990                 struct mbuf **mnext, *m_frgpart;
 991                 struct ip6_frag *ip6f;
 992                 u_int32_t id = htonl(ip6_randomid());
 993                 u_char nextproto;
 994 #if 0
 995                 struct ip6ctlparam ip6cp;
 996                 u_int32_t mtu32;
 997 #endif
 998                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
 999
1000                 /*
1001                  * Too large for the destination or interface;
1002                  * fragment if possible.
1003                  * Must be able to put at least 8 bytes per fragment.
1004                  */
1005                 hlen = unfragpartlen;
1006                 if (mtu > IPV6_MAXPACKET)
1007                         mtu = IPV6_MAXPACKET;
1008
1009 #if 0
1010                 /*
1011                  * It is believed this code is a leftover from the
1012                  * development of the IPV6_RECVPATHMTU sockopt and
1013                  * associated work to implement RFC3542.
1014                  * It's not entirely clear what the intent of the API
1015                  * is at this point, so disable this code for now.
1016                  * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1017                  * will send notifications if the application requests.
1018                  */
1019
1020                 /* Notify a proper path MTU to applications. */
1021                 mtu32 = (u_int32_t)mtu;
1022                 bzero(&ip6cp, sizeof(ip6cp));
1023                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1024                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1025                     (void *)&ip6cp);
1026 #endif
1027
1028                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1029                 if (len < 8) {
1030                         error = EMSGSIZE;
1031                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1032                         goto bad;
1033                 }
1034
1035                 /*
1036                  * Verify that we have any chance at all of being able to queue
1037                  *      the packet or packet fragments
1038                  */
1039                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1040                     < tlen  /* - hlen */)) {
1041                         error = ENOBUFS;
1042                         ip6stat.ip6s_odropped++;
1043                         goto bad;
1044                 }
1045
1046                 mnext = &m->m_nextpkt;
1047
1048                 /*
1049                  * Change the next header field of the last header in the
1050                  * unfragmentable part.
1051                  */
1052                 if (exthdrs.ip6e_rthdr) {
1053                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1054                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1055                 } else if (exthdrs.ip6e_dest1) {
1056                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1057                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1058                 } else if (exthdrs.ip6e_hbh) {
1059                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1060                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1061                 } else {
1062                         nextproto = ip6->ip6_nxt;
1063                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1064                 }
1065
1066                 /*
1067                  * Loop through length of segment after first fragment,
1068                  * make new header and copy data of each part and link onto
1069                  * chain.
1070                  */
1071                 m0 = m;
1072                 for (off = hlen; off < tlen; off += len) {
1073                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1074                         if (!m) {
1075                                 error = ENOBUFS;
1076                                 ip6stat.ip6s_odropped++;
1077                                 goto sendorfree;
1078                         }
1079                         m->m_pkthdr.rcvif = NULL;
1080                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1081                         *mnext = m;
1082                         mnext = &m->m_nextpkt;
1083                         m->m_data += max_linkhdr;
1084                         mhip6 = mtod(m, struct ip6_hdr *);
1085                         *mhip6 = *ip6;
1086                         m->m_len = sizeof(*mhip6);
1087                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1088                         if (error) {
1089                                 ip6stat.ip6s_odropped++;
1090                                 goto sendorfree;
1091                         }
1092                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1093                         if (off + len >= tlen)
1094                                 len = tlen - off;
1095                         else
1096                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1097                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1098                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1099                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1100                                 error = ENOBUFS;
1101                                 ip6stat.ip6s_odropped++;
1102                                 goto sendorfree;
1103                         }
1104                         m_cat(m, m_frgpart);
1105                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1106                         m->m_pkthdr.rcvif = NULL;
1107                         ip6f->ip6f_reserved = 0;
1108                         ip6f->ip6f_ident = id;
1109                         ip6f->ip6f_nxt = nextproto;
1110                         ip6stat.ip6s_ofragments++;
1111                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1112                 }
1113
1114                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1115         }
1116
1117         /*
1118          * Remove leading garbages.
1119          */
1120 sendorfree:
1121         m = m0->m_nextpkt;
1122         m0->m_nextpkt = 0;
1123         m_freem(m0);
1124         for (m0 = m; m; m = m0) {
1125                 m0 = m->m_nextpkt;
1126                 m->m_nextpkt = 0;
1127                 if (error == 0) {
1128                         /* Record statistics for this interface address. */
1129                         if (ia) {
1130                                 ia->ia_ifa.if_opackets++;
1131                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1132                         }
1133 #ifdef IPSEC
1134                         /* clean ipsec history once it goes out of the node */
1135                         ipsec_delaux(m);
1136 #endif
1137                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1138                 } else
1139                         m_freem(m);
1140         }
1141
1142         if (error == 0)
1143                 ip6stat.ip6s_fragmented++;
1144
1145 done:
1146         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1147                 RTFREE(ro->ro_rt);
1148         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1149                 RTFREE(ro_pmtu->ro_rt);
1150         }
1151
1152 #ifdef IPSEC
1153         if (sp != NULL)
1154                 key_freesp(sp);
1155 #endif /* IPSEC */
1156 #ifdef FAST_IPSEC
1157         if (sp != NULL)
1158                 KEY_FREESP(&sp);
1159 #endif /* FAST_IPSEC */
1160
1161         return (error);
1162
1163 freehdrs:
1164         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1165         m_freem(exthdrs.ip6e_dest1);
1166         m_freem(exthdrs.ip6e_rthdr);
1167         m_freem(exthdrs.ip6e_dest2);
1168         /* FALLTHROUGH */
1169 bad:
1170         m_freem(m);
1171         goto done;
1172 }
1173
1174 static int
1175 ip6_copyexthdr(mp, hdr, hlen)
1176         struct mbuf **mp;
1177         caddr_t hdr;
1178         int hlen;
1179 {
1180         struct mbuf *m;
1181
1182         if (hlen > MCLBYTES)
1183                 return (ENOBUFS); /* XXX */
1184
1185         MGET(m, M_DONTWAIT, MT_DATA);
1186         if (!m)
1187                 return (ENOBUFS);
1188
1189         if (hlen > MLEN) {
1190                 MCLGET(m, M_DONTWAIT);
1191                 if ((m->m_flags & M_EXT) == 0) {
1192                         m_free(m);
1193                         return (ENOBUFS);
1194                 }
1195         }
1196         m->m_len = hlen;
1197         if (hdr)
1198                 bcopy(hdr, mtod(m, caddr_t), hlen);
1199
1200         *mp = m;
1201         return (0);
1202 }
1203
1204 /*
1205  * Insert jumbo payload option.
1206  */
1207 static int
1208 ip6_insert_jumboopt(exthdrs, plen)
1209         struct ip6_exthdrs *exthdrs;
1210         u_int32_t plen;
1211 {
1212         struct mbuf *mopt;
1213         u_char *optbuf;
1214         u_int32_t v;
1215
1216 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1217
1218         /*
1219          * If there is no hop-by-hop options header, allocate new one.
1220          * If there is one but it doesn't have enough space to store the
1221          * jumbo payload option, allocate a cluster to store the whole options.
1222          * Otherwise, use it to store the options.
1223          */
1224         if (exthdrs->ip6e_hbh == 0) {
1225                 MGET(mopt, M_DONTWAIT, MT_DATA);
1226                 if (mopt == 0)
1227                         return (ENOBUFS);
1228                 mopt->m_len = JUMBOOPTLEN;
1229                 optbuf = mtod(mopt, u_char *);
1230                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1231                 exthdrs->ip6e_hbh = mopt;
1232         } else {
1233                 struct ip6_hbh *hbh;
1234
1235                 mopt = exthdrs->ip6e_hbh;
1236                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1237                         /*
1238                          * XXX assumption:
1239                          * - exthdrs->ip6e_hbh is not referenced from places
1240                          *   other than exthdrs.
1241                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1242                          */
1243                         int oldoptlen = mopt->m_len;
1244                         struct mbuf *n;
1245
1246                         /*
1247                          * XXX: give up if the whole (new) hbh header does
1248                          * not fit even in an mbuf cluster.
1249                          */
1250                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1251                                 return (ENOBUFS);
1252
1253                         /*
1254                          * As a consequence, we must always prepare a cluster
1255                          * at this point.
1256                          */
1257                         MGET(n, M_DONTWAIT, MT_DATA);
1258                         if (n) {
1259                                 MCLGET(n, M_DONTWAIT);
1260                                 if ((n->m_flags & M_EXT) == 0) {
1261                                         m_freem(n);
1262                                         n = NULL;
1263                                 }
1264                         }
1265                         if (!n)
1266                                 return (ENOBUFS);
1267                         n->m_len = oldoptlen + JUMBOOPTLEN;
1268                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1269                             oldoptlen);
1270                         optbuf = mtod(n, caddr_t) + oldoptlen;
1271                         m_freem(mopt);
1272                         mopt = exthdrs->ip6e_hbh = n;
1273                 } else {
1274                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1275                         mopt->m_len += JUMBOOPTLEN;
1276                 }
1277                 optbuf[0] = IP6OPT_PADN;
1278                 optbuf[1] = 1;
1279
1280                 /*
1281                  * Adjust the header length according to the pad and
1282                  * the jumbo payload option.
1283                  */
1284                 hbh = mtod(mopt, struct ip6_hbh *);
1285                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1286         }
1287
1288         /* fill in the option. */
1289         optbuf[2] = IP6OPT_JUMBO;
1290         optbuf[3] = 4;
1291         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1292         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1293
1294         /* finally, adjust the packet header length */
1295         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1296
1297         return (0);
1298 #undef JUMBOOPTLEN
1299 }
1300
1301 /*
1302  * Insert fragment header and copy unfragmentable header portions.
1303  */
1304 static int
1305 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1306         struct mbuf *m0, *m;
1307         int hlen;
1308         struct ip6_frag **frghdrp;
1309 {
1310         struct mbuf *n, *mlast;
1311
1312         if (hlen > sizeof(struct ip6_hdr)) {
1313                 n = m_copym(m0, sizeof(struct ip6_hdr),
1314                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1315                 if (n == 0)
1316                         return (ENOBUFS);
1317                 m->m_next = n;
1318         } else
1319                 n = m;
1320
1321         /* Search for the last mbuf of unfragmentable part. */
1322         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1323                 ;
1324
1325         if ((mlast->m_flags & M_EXT) == 0 &&
1326             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1327                 /* use the trailing space of the last mbuf for the fragment hdr */
1328                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1329                     mlast->m_len);
1330                 mlast->m_len += sizeof(struct ip6_frag);
1331                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1332         } else {
1333                 /* allocate a new mbuf for the fragment header */
1334                 struct mbuf *mfrg;
1335
1336                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1337                 if (mfrg == 0)
1338                         return (ENOBUFS);
1339                 mfrg->m_len = sizeof(struct ip6_frag);
1340                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1341                 mlast->m_next = mfrg;
1342         }
1343
1344         return (0);
1345 }
1346
1347 static int
1348 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1349         struct route_in6 *ro_pmtu, *ro;
1350         struct ifnet *ifp;
1351         struct in6_addr *dst;
1352         u_long *mtup;
1353         int *alwaysfragp;
1354 {
1355         u_int32_t mtu = 0;
1356         int alwaysfrag = 0;
1357         int error = 0;
1358
1359         if (ro_pmtu != ro) {
1360                 /* The first hop and the final destination may differ. */
1361                 struct sockaddr_in6 *sa6_dst =
1362                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1363                 if (ro_pmtu->ro_rt &&
1364                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1365                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1366                         RTFREE(ro_pmtu->ro_rt);
1367                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1368                 }
1369                 if (ro_pmtu->ro_rt == NULL) {
1370                         bzero(sa6_dst, sizeof(*sa6_dst));
1371                         sa6_dst->sin6_family = AF_INET6;
1372                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1373                         sa6_dst->sin6_addr = *dst;
1374
1375                         rtalloc((struct route *)ro_pmtu);
1376                 }
1377         }
1378         if (ro_pmtu->ro_rt) {
1379                 u_int32_t ifmtu;
1380                 struct in_conninfo inc;
1381
1382                 bzero(&inc, sizeof(inc));
1383                 inc.inc_flags = 1; /* IPv6 */
1384                 inc.inc6_faddr = *dst;
1385
1386                 if (ifp == NULL)
1387                         ifp = ro_pmtu->ro_rt->rt_ifp;
1388                 ifmtu = IN6_LINKMTU(ifp);
1389                 mtu = tcp_hc_getmtu(&inc);
1390                 if (mtu)
1391                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1392                 else
1393                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1394                 if (mtu == 0)
1395                         mtu = ifmtu;
1396                 else if (mtu < IPV6_MMTU) {
1397                         /*
1398                          * RFC2460 section 5, last paragraph:
1399                          * if we record ICMPv6 too big message with
1400                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1401                          * or smaller, with framgent header attached.
1402                          * (fragment header is needed regardless from the
1403                          * packet size, for translators to identify packets)
1404                          */
1405                         alwaysfrag = 1;
1406                         mtu = IPV6_MMTU;
1407                 } else if (mtu > ifmtu) {
1408                         /*
1409                          * The MTU on the route is larger than the MTU on
1410                          * the interface!  This shouldn't happen, unless the
1411                          * MTU of the interface has been changed after the
1412                          * interface was brought up.  Change the MTU in the
1413                          * route to match the interface MTU (as long as the
1414                          * field isn't locked).
1415                          */
1416                         mtu = ifmtu;
1417                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1418                 }
1419         } else if (ifp) {
1420                 mtu = IN6_LINKMTU(ifp);
1421         } else
1422                 error = EHOSTUNREACH; /* XXX */
1423
1424         *mtup = mtu;
1425         if (alwaysfragp)
1426                 *alwaysfragp = alwaysfrag;
1427         return (error);
1428 }
1429
1430 /*
1431  * IP6 socket option processing.
1432  */
1433 int
1434 ip6_ctloutput(so, sopt)
1435         struct socket *so;
1436         struct sockopt *sopt;
1437 {
1438         int privileged, optdatalen, uproto;
1439         void *optdata;
1440         struct inpcb *in6p = sotoinpcb(so);
1441         int error, optval;
1442         int level, op, optname;
1443         int optlen;
1444         struct thread *td;
1445
1446         if (sopt) {
1447                 level = sopt->sopt_level;
1448                 op = sopt->sopt_dir;
1449                 optname = sopt->sopt_name;
1450                 optlen = sopt->sopt_valsize;
1451                 td = sopt->sopt_td;
1452         } else {
1453                 panic("ip6_ctloutput: arg soopt is NULL");
1454         }
1455         error = optval = 0;
1456
1457         privileged = (td == 0 || suser(td)) ? 0 : 1;
1458         uproto = (int)so->so_proto->pr_protocol;
1459
1460         if (level == IPPROTO_IPV6) {
1461                 switch (op) {
1462
1463                 case SOPT_SET:
1464                         switch (optname) {
1465                         case IPV6_2292PKTOPTIONS:
1466 #ifdef IPV6_PKTOPTIONS
1467                         case IPV6_PKTOPTIONS:
1468 #endif
1469                         {
1470                                 struct mbuf *m;
1471
1472                                 error = soopt_getm(sopt, &m); /* XXX */
1473                                 if (error != 0)
1474                                         break;
1475                                 error = soopt_mcopyin(sopt, m); /* XXX */
1476                                 if (error != 0)
1477                                         break;
1478                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1479                                                     m, so, sopt);
1480                                 m_freem(m); /* XXX */
1481                                 break;
1482                         }
1483
1484                         /*
1485                          * Use of some Hop-by-Hop options or some
1486                          * Destination options, might require special
1487                          * privilege.  That is, normal applications
1488                          * (without special privilege) might be forbidden
1489                          * from setting certain options in outgoing packets,
1490                          * and might never see certain options in received
1491                          * packets. [RFC 2292 Section 6]
1492                          * KAME specific note:
1493                          *  KAME prevents non-privileged users from sending or
1494                          *  receiving ANY hbh/dst options in order to avoid
1495                          *  overhead of parsing options in the kernel.
1496                          */
1497                         case IPV6_RECVHOPOPTS:
1498                         case IPV6_RECVDSTOPTS:
1499                         case IPV6_RECVRTHDRDSTOPTS:
1500                                 if (!privileged) {
1501                                         error = EPERM;
1502                                         break;
1503                                 }
1504                                 /* FALLTHROUGH */
1505                         case IPV6_UNICAST_HOPS:
1506                         case IPV6_HOPLIMIT:
1507                         case IPV6_FAITH:
1508
1509                         case IPV6_RECVPKTINFO:
1510                         case IPV6_RECVHOPLIMIT:
1511                         case IPV6_RECVRTHDR:
1512                         case IPV6_RECVPATHMTU:
1513                         case IPV6_RECVTCLASS:
1514                         case IPV6_V6ONLY:
1515                         case IPV6_AUTOFLOWLABEL:
1516                                 if (optlen != sizeof(int)) {
1517                                         error = EINVAL;
1518                                         break;
1519                                 }
1520                                 error = sooptcopyin(sopt, &optval,
1521                                         sizeof optval, sizeof optval);
1522                                 if (error)
1523                                         break;
1524                                 switch (optname) {
1525
1526                                 case IPV6_UNICAST_HOPS:
1527                                         if (optval < -1 || optval >= 256)
1528                                                 error = EINVAL;
1529                                         else {
1530                                                 /* -1 = kernel default */
1531                                                 in6p->in6p_hops = optval;
1532                                                 if ((in6p->in6p_vflag &
1533                                                      INP_IPV4) != 0)
1534                                                         in6p->inp_ip_ttl = optval;
1535                                         }
1536                                         break;
1537 #define OPTSET(bit) \
1538 do { \
1539         if (optval) \
1540                 in6p->in6p_flags |= (bit); \
1541         else \
1542                 in6p->in6p_flags &= ~(bit); \
1543 } while (/*CONSTCOND*/ 0)
1544 #define OPTSET2292(bit) \
1545 do { \
1546         in6p->in6p_flags |= IN6P_RFC2292; \
1547         if (optval) \
1548                 in6p->in6p_flags |= (bit); \
1549         else \
1550                 in6p->in6p_flags &= ~(bit); \
1551 } while (/*CONSTCOND*/ 0)
1552 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1553
1554                                 case IPV6_RECVPKTINFO:
1555                                         /* cannot mix with RFC2292 */
1556                                         if (OPTBIT(IN6P_RFC2292)) {
1557                                                 error = EINVAL;
1558                                                 break;
1559                                         }
1560                                         OPTSET(IN6P_PKTINFO);
1561                                         break;
1562
1563                                 case IPV6_HOPLIMIT:
1564                                 {
1565                                         struct ip6_pktopts **optp;
1566
1567                                         /* cannot mix with RFC2292 */
1568                                         if (OPTBIT(IN6P_RFC2292)) {
1569                                                 error = EINVAL;
1570                                                 break;
1571                                         }
1572                                         optp = &in6p->in6p_outputopts;
1573                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1574                                                            (u_char *)&optval,
1575                                                            sizeof(optval),
1576                                                            optp,
1577                                                            privileged, uproto);
1578                                         break;
1579                                 }
1580
1581                                 case IPV6_RECVHOPLIMIT:
1582                                         /* cannot mix with RFC2292 */
1583                                         if (OPTBIT(IN6P_RFC2292)) {
1584                                                 error = EINVAL;
1585                                                 break;
1586                                         }
1587                                         OPTSET(IN6P_HOPLIMIT);
1588                                         break;
1589
1590                                 case IPV6_RECVHOPOPTS:
1591                                         /* cannot mix with RFC2292 */
1592                                         if (OPTBIT(IN6P_RFC2292)) {
1593                                                 error = EINVAL;
1594                                                 break;
1595                                         }
1596                                         OPTSET(IN6P_HOPOPTS);
1597                                         break;
1598
1599                                 case IPV6_RECVDSTOPTS:
1600                                         /* cannot mix with RFC2292 */
1601                                         if (OPTBIT(IN6P_RFC2292)) {
1602                                                 error = EINVAL;
1603                                                 break;
1604                                         }
1605                                         OPTSET(IN6P_DSTOPTS);
1606                                         break;
1607
1608                                 case IPV6_RECVRTHDRDSTOPTS:
1609                                         /* cannot mix with RFC2292 */
1610                                         if (OPTBIT(IN6P_RFC2292)) {
1611                                                 error = EINVAL;
1612                                                 break;
1613                                         }
1614                                         OPTSET(IN6P_RTHDRDSTOPTS);
1615                                         break;
1616
1617                                 case IPV6_RECVRTHDR:
1618                                         /* cannot mix with RFC2292 */
1619                                         if (OPTBIT(IN6P_RFC2292)) {
1620                                                 error = EINVAL;
1621                                                 break;
1622                                         }
1623                                         OPTSET(IN6P_RTHDR);
1624                                         break;
1625
1626                                 case IPV6_FAITH:
1627                                         OPTSET(IN6P_FAITH);
1628                                         break;
1629
1630                                 case IPV6_RECVPATHMTU:
1631                                         /*
1632                                          * We ignore this option for TCP
1633                                          * sockets.
1634                                          * (RFC3542 leaves this case
1635                                          * unspecified.)
1636                                          */
1637                                         if (uproto != IPPROTO_TCP)
1638                                                 OPTSET(IN6P_MTU);
1639                                         break;
1640
1641                                 case IPV6_V6ONLY:
1642                                         /*
1643                                          * make setsockopt(IPV6_V6ONLY)
1644                                          * available only prior to bind(2).
1645                                          * see ipng mailing list, Jun 22 2001.
1646                                          */
1647                                         if (in6p->in6p_lport ||
1648                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1649                                                 error = EINVAL;
1650                                                 break;
1651                                         }
1652                                         OPTSET(IN6P_IPV6_V6ONLY);
1653                                         if (optval)
1654                                                 in6p->in6p_vflag &= ~INP_IPV4;
1655                                         else
1656                                                 in6p->in6p_vflag |= INP_IPV4;
1657                                         break;
1658                                 case IPV6_RECVTCLASS:
1659                                         /* cannot mix with RFC2292 XXX */
1660                                         if (OPTBIT(IN6P_RFC2292)) {
1661                                                 error = EINVAL;
1662                                                 break;
1663                                         }
1664                                         OPTSET(IN6P_TCLASS);
1665                                         break;
1666                                 case IPV6_AUTOFLOWLABEL:
1667                                         OPTSET(IN6P_AUTOFLOWLABEL);
1668                                         break;
1669
1670                                 }
1671                                 break;
1672
1673                         case IPV6_TCLASS:
1674                         case IPV6_DONTFRAG:
1675                         case IPV6_USE_MIN_MTU:
1676                         case IPV6_PREFER_TEMPADDR:
1677                                 if (optlen != sizeof(optval)) {
1678                                         error = EINVAL;
1679                                         break;
1680                                 }
1681                                 error = sooptcopyin(sopt, &optval,
1682                                         sizeof optval, sizeof optval);
1683                                 if (error)
1684                                         break;
1685                                 {
1686                                         struct ip6_pktopts **optp;
1687                                         optp = &in6p->in6p_outputopts;
1688                                         error = ip6_pcbopt(optname,
1689                                                            (u_char *)&optval,
1690                                                            sizeof(optval),
1691                                                            optp,
1692                                                            privileged, uproto);
1693                                         break;
1694                                 }
1695
1696                         case IPV6_2292PKTINFO:
1697                         case IPV6_2292HOPLIMIT:
1698                         case IPV6_2292HOPOPTS:
1699                         case IPV6_2292DSTOPTS:
1700                         case IPV6_2292RTHDR:
1701                                 /* RFC 2292 */
1702                                 if (optlen != sizeof(int)) {
1703                                         error = EINVAL;
1704                                         break;
1705                                 }
1706                                 error = sooptcopyin(sopt, &optval,
1707                                         sizeof optval, sizeof optval);
1708                                 if (error)
1709                                         break;
1710                                 switch (optname) {
1711                                 case IPV6_2292PKTINFO:
1712                                         OPTSET2292(IN6P_PKTINFO);
1713                                         break;
1714                                 case IPV6_2292HOPLIMIT:
1715                                         OPTSET2292(IN6P_HOPLIMIT);
1716                                         break;
1717                                 case IPV6_2292HOPOPTS:
1718                                         /*
1719                                          * Check super-user privilege.
1720                                          * See comments for IPV6_RECVHOPOPTS.
1721                                          */
1722                                         if (!privileged)
1723                                                 return (EPERM);
1724                                         OPTSET2292(IN6P_HOPOPTS);
1725                                         break;
1726                                 case IPV6_2292DSTOPTS:
1727                                         if (!privileged)
1728                                                 return (EPERM);
1729                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1730                                         break;
1731                                 case IPV6_2292RTHDR:
1732                                         OPTSET2292(IN6P_RTHDR);
1733                                         break;
1734                                 }
1735                                 break;
1736                         case IPV6_PKTINFO:
1737                         case IPV6_HOPOPTS:
1738                         case IPV6_RTHDR:
1739                         case IPV6_DSTOPTS:
1740                         case IPV6_RTHDRDSTOPTS:
1741                         case IPV6_NEXTHOP:
1742                         {
1743                                 /* new advanced API (RFC3542) */
1744                                 u_char *optbuf;
1745                                 u_char optbuf_storage[MCLBYTES];
1746                                 int optlen;
1747                                 struct ip6_pktopts **optp;
1748
1749                                 /* cannot mix with RFC2292 */
1750                                 if (OPTBIT(IN6P_RFC2292)) {
1751                                         error = EINVAL;
1752                                         break;
1753                                 }
1754
1755                                 /*
1756                                  * We only ensure valsize is not too large
1757                                  * here.  Further validation will be done
1758                                  * later.
1759                                  */
1760                                 error = sooptcopyin(sopt, optbuf_storage,
1761                                     sizeof(optbuf_storage), 0);
1762                                 if (error)
1763                                         break;
1764                                 optlen = sopt->sopt_valsize;
1765                                 optbuf = optbuf_storage;
1766                                 optp = &in6p->in6p_outputopts;
1767                                 error = ip6_pcbopt(optname,
1768                                                    optbuf, optlen,
1769                                                    optp, privileged, uproto);
1770                                 break;
1771                         }
1772 #undef OPTSET
1773
1774                         case IPV6_MULTICAST_IF:
1775                         case IPV6_MULTICAST_HOPS:
1776                         case IPV6_MULTICAST_LOOP:
1777                         case IPV6_JOIN_GROUP:
1778                         case IPV6_LEAVE_GROUP:
1779                             {
1780                                 if (sopt->sopt_valsize > MLEN) {
1781                                         error = EMSGSIZE;
1782                                         break;
1783                                 }
1784                                 /* XXX */
1785                             }
1786                             /* FALLTHROUGH */
1787                             {
1788                                 struct mbuf *m;
1789
1790                                 if (sopt->sopt_valsize > MCLBYTES) {
1791                                         error = EMSGSIZE;
1792                                         break;
1793                                 }
1794                                 /* XXX */
1795                                 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1796                                 if (m == 0) {
1797                                         error = ENOBUFS;
1798                                         break;
1799                                 }
1800                                 if (sopt->sopt_valsize > MLEN) {
1801                                         MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1802                                         if ((m->m_flags & M_EXT) == 0) {
1803                                                 m_free(m);
1804                                                 error = ENOBUFS;
1805                                                 break;
1806                                         }
1807                                 }
1808                                 m->m_len = sopt->sopt_valsize;
1809                                 error = sooptcopyin(sopt, mtod(m, char *),
1810                                                     m->m_len, m->m_len);
1811                                 if (error) {
1812                                         (void)m_free(m);
1813                                         break;
1814                                 }
1815                                 error = ip6_setmoptions(sopt->sopt_name,
1816                                                         &in6p->in6p_moptions,
1817                                                         m);
1818                                 (void)m_free(m);
1819                             }
1820                                 break;
1821
1822                         case IPV6_PORTRANGE:
1823                                 error = sooptcopyin(sopt, &optval,
1824                                     sizeof optval, sizeof optval);
1825                                 if (error)
1826                                         break;
1827
1828                                 switch (optval) {
1829                                 case IPV6_PORTRANGE_DEFAULT:
1830                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1831                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1832                                         break;
1833
1834                                 case IPV6_PORTRANGE_HIGH:
1835                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1836                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1837                                         break;
1838
1839                                 case IPV6_PORTRANGE_LOW:
1840                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1841                                         in6p->in6p_flags |= IN6P_LOWPORT;
1842                                         break;
1843
1844                                 default:
1845                                         error = EINVAL;
1846                                         break;
1847                                 }
1848                                 break;
1849
1850 #if defined(IPSEC) || defined(FAST_IPSEC)
1851                         case IPV6_IPSEC_POLICY:
1852                             {
1853                                 caddr_t req = NULL;
1854                                 size_t len = 0;
1855                                 struct mbuf *m;
1856
1857                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1858                                         break;
1859                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1860                                         break;
1861                                 if (m) {
1862                                         req = mtod(m, caddr_t);
1863                                         len = m->m_len;
1864                                 }
1865                                 error = ipsec6_set_policy(in6p, optname, req,
1866                                                           len, privileged);
1867                                 m_freem(m);
1868                             }
1869                                 break;
1870 #endif /* KAME IPSEC */
1871
1872                         default:
1873                                 error = ENOPROTOOPT;
1874                                 break;
1875                         }
1876                         break;
1877
1878                 case SOPT_GET:
1879                         switch (optname) {
1880
1881                         case IPV6_2292PKTOPTIONS:
1882 #ifdef IPV6_PKTOPTIONS
1883                         case IPV6_PKTOPTIONS:
1884 #endif
1885                                 /*
1886                                  * RFC3542 (effectively) deprecated the
1887                                  * semantics of the 2292-style pktoptions.
1888                                  * Since it was not reliable in nature (i.e.,
1889                                  * applications had to expect the lack of some
1890                                  * information after all), it would make sense
1891                                  * to simplify this part by always returning
1892                                  * empty data.
1893                                  */
1894                                 sopt->sopt_valsize = 0;
1895                                 break;
1896
1897                         case IPV6_RECVHOPOPTS:
1898                         case IPV6_RECVDSTOPTS:
1899                         case IPV6_RECVRTHDRDSTOPTS:
1900                         case IPV6_UNICAST_HOPS:
1901                         case IPV6_RECVPKTINFO:
1902                         case IPV6_RECVHOPLIMIT:
1903                         case IPV6_RECVRTHDR:
1904                         case IPV6_RECVPATHMTU:
1905
1906                         case IPV6_FAITH:
1907                         case IPV6_V6ONLY:
1908                         case IPV6_PORTRANGE:
1909                         case IPV6_RECVTCLASS:
1910                         case IPV6_AUTOFLOWLABEL:
1911                                 switch (optname) {
1912
1913                                 case IPV6_RECVHOPOPTS:
1914                                         optval = OPTBIT(IN6P_HOPOPTS);
1915                                         break;
1916
1917                                 case IPV6_RECVDSTOPTS:
1918                                         optval = OPTBIT(IN6P_DSTOPTS);
1919                                         break;
1920
1921                                 case IPV6_RECVRTHDRDSTOPTS:
1922                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1923                                         break;
1924
1925                                 case IPV6_UNICAST_HOPS:
1926                                         optval = in6p->in6p_hops;
1927                                         break;
1928
1929                                 case IPV6_RECVPKTINFO:
1930                                         optval = OPTBIT(IN6P_PKTINFO);
1931                                         break;
1932
1933                                 case IPV6_RECVHOPLIMIT:
1934                                         optval = OPTBIT(IN6P_HOPLIMIT);
1935                                         break;
1936
1937                                 case IPV6_RECVRTHDR:
1938                                         optval = OPTBIT(IN6P_RTHDR);
1939                                         break;
1940
1941                                 case IPV6_RECVPATHMTU:
1942                                         optval = OPTBIT(IN6P_MTU);
1943                                         break;
1944
1945                                 case IPV6_FAITH:
1946                                         optval = OPTBIT(IN6P_FAITH);
1947                                         break;
1948
1949                                 case IPV6_V6ONLY:
1950                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1951                                         break;
1952
1953                                 case IPV6_PORTRANGE:
1954                                     {
1955                                         int flags;
1956                                         flags = in6p->in6p_flags;
1957                                         if (flags & IN6P_HIGHPORT)
1958                                                 optval = IPV6_PORTRANGE_HIGH;
1959                                         else if (flags & IN6P_LOWPORT)
1960                                                 optval = IPV6_PORTRANGE_LOW;
1961                                         else
1962                                                 optval = 0;
1963                                         break;
1964                                     }
1965                                 case IPV6_RECVTCLASS:
1966                                         optval = OPTBIT(IN6P_TCLASS);
1967                                         break;
1968
1969                                 case IPV6_AUTOFLOWLABEL:
1970                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1971                                         break;
1972                                 }
1973                                 if (error)
1974                                         break;
1975                                 error = sooptcopyout(sopt, &optval,
1976                                         sizeof optval);
1977                                 break;
1978
1979                         case IPV6_PATHMTU:
1980                         {
1981                                 u_long pmtu = 0;
1982                                 struct ip6_mtuinfo mtuinfo;
1983                                 struct route_in6 sro;
1984
1985                                 bzero(&sro, sizeof(sro));
1986
1987                                 if (!(so->so_state & SS_ISCONNECTED))
1988                                         return (ENOTCONN);
1989                                 /*
1990                                  * XXX: we dot not consider the case of source
1991                                  * routing, or optional information to specify
1992                                  * the outgoing interface.
1993                                  */
1994                                 error = ip6_getpmtu(&sro, NULL, NULL,
1995                                     &in6p->in6p_faddr, &pmtu, NULL);
1996                                 if (sro.ro_rt)
1997                                         RTFREE(sro.ro_rt);
1998                                 if (error)
1999                                         break;
2000                                 if (pmtu > IPV6_MAXPACKET)
2001                                         pmtu = IPV6_MAXPACKET;
2002
2003                                 bzero(&mtuinfo, sizeof(mtuinfo));
2004                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2005                                 optdata = (void *)&mtuinfo;
2006                                 optdatalen = sizeof(mtuinfo);
2007                                 error = sooptcopyout(sopt, optdata,
2008                                     optdatalen);
2009                                 break;
2010                         }
2011
2012                         case IPV6_2292PKTINFO:
2013                         case IPV6_2292HOPLIMIT:
2014                         case IPV6_2292HOPOPTS:
2015                         case IPV6_2292RTHDR:
2016                         case IPV6_2292DSTOPTS:
2017                                 switch (optname) {
2018                                 case IPV6_2292PKTINFO:
2019                                         optval = OPTBIT(IN6P_PKTINFO);
2020                                         break;
2021                                 case IPV6_2292HOPLIMIT:
2022                                         optval = OPTBIT(IN6P_HOPLIMIT);
2023                                         break;
2024                                 case IPV6_2292HOPOPTS:
2025                                         optval = OPTBIT(IN6P_HOPOPTS);
2026                                         break;
2027                                 case IPV6_2292RTHDR:
2028                                         optval = OPTBIT(IN6P_RTHDR);
2029                                         break;
2030                                 case IPV6_2292DSTOPTS:
2031                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2032                                         break;
2033                                 }
2034                                 error = sooptcopyout(sopt, &optval,
2035                                     sizeof optval);
2036                                 break;
2037                         case IPV6_PKTINFO:
2038                         case IPV6_HOPOPTS:
2039                         case IPV6_RTHDR:
2040                         case IPV6_DSTOPTS:
2041                         case IPV6_RTHDRDSTOPTS:
2042                         case IPV6_NEXTHOP:
2043                         case IPV6_TCLASS:
2044                         case IPV6_DONTFRAG:
2045                         case IPV6_USE_MIN_MTU:
2046                         case IPV6_PREFER_TEMPADDR:
2047                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2048                                     optname, sopt);
2049                                 break;
2050
2051                         case IPV6_MULTICAST_IF:
2052                         case IPV6_MULTICAST_HOPS:
2053                         case IPV6_MULTICAST_LOOP:
2054                         case IPV6_JOIN_GROUP:
2055                         case IPV6_LEAVE_GROUP:
2056                             {
2057                                 struct mbuf *m;
2058                                 error = ip6_getmoptions(sopt->sopt_name,
2059                                     in6p->in6p_moptions, &m);
2060                                 if (error == 0)
2061                                         error = sooptcopyout(sopt,
2062                                             mtod(m, char *), m->m_len);
2063                                 m_freem(m);
2064                             }
2065                                 break;
2066
2067 #if defined(IPSEC) || defined(FAST_IPSEC)
2068                         case IPV6_IPSEC_POLICY:
2069                           {
2070                                 caddr_t req = NULL;
2071                                 size_t len = 0;
2072                                 struct mbuf *m = NULL;
2073                                 struct mbuf **mp = &m;
2074                                 size_t ovalsize = sopt->sopt_valsize;
2075                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2076
2077                                 error = soopt_getm(sopt, &m); /* XXX */
2078                                 if (error != 0)
2079                                         break;
2080                                 error = soopt_mcopyin(sopt, m); /* XXX */
2081                                 if (error != 0)
2082                                         break;
2083                                 sopt->sopt_valsize = ovalsize;
2084                                 sopt->sopt_val = oval;
2085                                 if (m) {
2086                                         req = mtod(m, caddr_t);
2087                                         len = m->m_len;
2088                                 }
2089                                 error = ipsec6_get_policy(in6p, req, len, mp);
2090                                 if (error == 0)
2091                                         error = soopt_mcopyout(sopt, m); /* XXX */
2092                                 if (error == 0 && m)
2093                                         m_freem(m);
2094                                 break;
2095                           }
2096 #endif /* KAME IPSEC */
2097
2098                         default:
2099                                 error = ENOPROTOOPT;
2100                                 break;
2101                         }
2102                         break;
2103                 }
2104         } else {                /* level != IPPROTO_IPV6 */
2105                 error = EINVAL;
2106         }
2107         return (error);
2108 }
2109
2110 int
2111 ip6_raw_ctloutput(so, sopt)
2112         struct socket *so;
2113         struct sockopt *sopt;
2114 {
2115         int error = 0, optval, optlen;
2116         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2117         struct in6pcb *in6p = sotoin6pcb(so);
2118         int level, op, optname;
2119
2120         if (sopt) {
2121                 level = sopt->sopt_level;
2122                 op = sopt->sopt_dir;
2123                 optname = sopt->sopt_name;
2124                 optlen = sopt->sopt_valsize;
2125         } else
2126                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2127
2128         if (level != IPPROTO_IPV6) {
2129                 return (EINVAL);
2130         }
2131
2132         switch (optname) {
2133         case IPV6_CHECKSUM:
2134                 /*
2135                  * For ICMPv6 sockets, no modification allowed for checksum
2136                  * offset, permit "no change" values to help existing apps.
2137                  *
2138                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2139                  * for an ICMPv6 socket will fail."
2140                  * The current behavior does not meet RFC3542.
2141                  */
2142                 switch (op) {
2143                 case SOPT_SET:
2144                         if (optlen != sizeof(int)) {
2145                                 error = EINVAL;
2146                                 break;
2147                         }
2148                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2149                                             sizeof(optval));
2150                         if (error)
2151                                 break;
2152                         if ((optval % 2) != 0) {
2153                                 /* the API assumes even offset values */
2154                                 error = EINVAL;
2155                         } else if (so->so_proto->pr_protocol ==
2156                             IPPROTO_ICMPV6) {
2157                                 if (optval != icmp6off)
2158                                         error = EINVAL;
2159                         } else
2160                                 in6p->in6p_cksum = optval;
2161                         break;
2162
2163                 case SOPT_GET:
2164                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2165                                 optval = icmp6off;
2166                         else
2167                                 optval = in6p->in6p_cksum;
2168
2169                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2170                         break;
2171
2172                 default:
2173                         error = EINVAL;
2174                         break;
2175                 }
2176                 break;
2177
2178         default:
2179                 error = ENOPROTOOPT;
2180                 break;
2181         }
2182
2183         return (error);
2184 }
2185
2186 /*
2187  * Set up IP6 options in pcb for insertion in output packets or
2188  * specifying behavior of outgoing packets.
2189  */
2190 static int
2191 ip6_pcbopts(pktopt, m, so, sopt)
2192         struct ip6_pktopts **pktopt;
2193         struct mbuf *m;
2194         struct socket *so;
2195         struct sockopt *sopt;
2196 {
2197         struct ip6_pktopts *opt = *pktopt;
2198         int error = 0;
2199         struct thread *td = sopt->sopt_td;
2200         int priv = 0;
2201
2202         /* turn off any old options. */
2203         if (opt) {
2204 #ifdef DIAGNOSTIC
2205                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2206                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2207                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2208                         printf("ip6_pcbopts: all specified options are cleared.\n");
2209 #endif
2210                 ip6_clearpktopts(opt, -1);
2211         } else
2212                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2213         *pktopt = NULL;
2214
2215         if (!m || m->m_len == 0) {
2216                 /*
2217                  * Only turning off any previous options, regardless of
2218                  * whether the opt is just created or given.
2219                  */
2220                 free(opt, M_IP6OPT);
2221                 return (0);
2222         }
2223
2224         /*  set options specified by user. */
2225         if (td && !suser(td))
2226                 priv = 1;
2227         if ((error = ip6_setpktopts(m, opt, NULL, priv,
2228             so->so_proto->pr_protocol)) != 0) {
2229                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2230                 free(opt, M_IP6OPT);
2231                 return (error);
2232         }
2233         *pktopt = opt;
2234         return (0);
2235 }
2236
2237 /*
2238  * initialize ip6_pktopts.  beware that there are non-zero default values in
2239  * the struct.
2240  */
2241 void
2242 ip6_initpktopts(opt)
2243         struct ip6_pktopts *opt;
2244 {
2245
2246         bzero(opt, sizeof(*opt));
2247         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2248         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2249         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2250         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2251 }
2252
2253 static int
2254 ip6_pcbopt(optname, buf, len, pktopt, priv, uproto)
2255         int optname, len, priv;
2256         u_char *buf;
2257         struct ip6_pktopts **pktopt;
2258         int uproto;
2259 {
2260         struct ip6_pktopts *opt;
2261
2262         if (*pktopt == NULL) {
2263                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2264                     M_WAITOK);
2265                 ip6_initpktopts(*pktopt);
2266         }
2267         opt = *pktopt;
2268
2269         return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2270 }
2271
2272 static int
2273 ip6_getpcbopt(pktopt, optname, sopt)
2274         struct ip6_pktopts *pktopt;
2275         struct sockopt *sopt;
2276         int optname;
2277 {
2278         void *optdata = NULL;
2279         int optdatalen = 0;
2280         struct ip6_ext *ip6e;
2281         int error = 0;
2282         struct in6_pktinfo null_pktinfo;
2283         int deftclass = 0, on;
2284         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2285         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2286
2287         switch (optname) {
2288         case IPV6_PKTINFO:
2289                 if (pktopt && pktopt->ip6po_pktinfo)
2290                         optdata = (void *)pktopt->ip6po_pktinfo;
2291                 else {
2292                         /* XXX: we don't have to do this every time... */
2293                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2294                         optdata = (void *)&null_pktinfo;
2295                 }
2296                 optdatalen = sizeof(struct in6_pktinfo);
2297                 break;
2298         case IPV6_TCLASS:
2299                 if (pktopt && pktopt->ip6po_tclass >= 0)
2300                         optdata = (void *)&pktopt->ip6po_tclass;
2301                 else
2302                         optdata = (void *)&deftclass;
2303                 optdatalen = sizeof(int);
2304                 break;
2305         case IPV6_HOPOPTS:
2306                 if (pktopt && pktopt->ip6po_hbh) {
2307                         optdata = (void *)pktopt->ip6po_hbh;
2308                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2309                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2310                 }
2311                 break;
2312         case IPV6_RTHDR:
2313                 if (pktopt && pktopt->ip6po_rthdr) {
2314                         optdata = (void *)pktopt->ip6po_rthdr;
2315                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2316                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2317                 }
2318                 break;
2319         case IPV6_RTHDRDSTOPTS:
2320                 if (pktopt && pktopt->ip6po_dest1) {
2321                         optdata = (void *)pktopt->ip6po_dest1;
2322                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2323                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2324                 }
2325                 break;
2326         case IPV6_DSTOPTS:
2327                 if (pktopt && pktopt->ip6po_dest2) {
2328                         optdata = (void *)pktopt->ip6po_dest2;
2329                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2330                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2331                 }
2332                 break;
2333         case IPV6_NEXTHOP:
2334                 if (pktopt && pktopt->ip6po_nexthop) {
2335                         optdata = (void *)pktopt->ip6po_nexthop;
2336                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2337                 }
2338                 break;
2339         case IPV6_USE_MIN_MTU:
2340                 if (pktopt)
2341                         optdata = (void *)&pktopt->ip6po_minmtu;
2342                 else
2343                         optdata = (void *)&defminmtu;
2344                 optdatalen = sizeof(int);
2345                 break;
2346         case IPV6_DONTFRAG:
2347                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2348                         on = 1;
2349                 else
2350                         on = 0;
2351                 optdata = (void *)&on;
2352                 optdatalen = sizeof(on);
2353                 break;
2354         case IPV6_PREFER_TEMPADDR:
2355                 if (pktopt)
2356                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2357                 else
2358                         optdata = (void *)&defpreftemp;
2359                 optdatalen = sizeof(int);
2360                 break;
2361         default:                /* should not happen */
2362 #ifdef DIAGNOSTIC
2363                 panic("ip6_getpcbopt: unexpected option\n");
2364 #endif
2365                 return (ENOPROTOOPT);
2366         }
2367
2368         error = sooptcopyout(sopt, optdata, optdatalen);
2369
2370         return (error);
2371 }
2372
2373 void
2374 ip6_clearpktopts(pktopt, optname)
2375         struct ip6_pktopts *pktopt;
2376         int optname;
2377 {
2378         if (pktopt == NULL)
2379                 return;
2380
2381         if (optname == -1 || optname == IPV6_PKTINFO) {
2382                 if (pktopt->ip6po_pktinfo)
2383                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2384                 pktopt->ip6po_pktinfo = NULL;
2385         }
2386         if (optname == -1 || optname == IPV6_HOPLIMIT)
2387                 pktopt->ip6po_hlim = -1;
2388         if (optname == -1 || optname == IPV6_TCLASS)
2389                 pktopt->ip6po_tclass = -1;
2390         if (optname == -1 || optname == IPV6_NEXTHOP) {
2391                 if (pktopt->ip6po_nextroute.ro_rt) {
2392                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2393                         pktopt->ip6po_nextroute.ro_rt = NULL;
2394                 }
2395                 if (pktopt->ip6po_nexthop)
2396                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2397                 pktopt->ip6po_nexthop = NULL;
2398         }
2399         if (optname == -1 || optname == IPV6_HOPOPTS) {
2400                 if (pktopt->ip6po_hbh)
2401                         free(pktopt->ip6po_hbh, M_IP6OPT);
2402                 pktopt->ip6po_hbh = NULL;
2403         }
2404         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2405                 if (pktopt->ip6po_dest1)
2406                         free(pktopt->ip6po_dest1, M_IP6OPT);
2407                 pktopt->ip6po_dest1 = NULL;
2408         }
2409         if (optname == -1 || optname == IPV6_RTHDR) {
2410                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2411                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2412                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2413                 if (pktopt->ip6po_route.ro_rt) {
2414                         RTFREE(pktopt->ip6po_route.ro_rt);
2415                         pktopt->ip6po_route.ro_rt = NULL;
2416                 }
2417         }
2418         if (optname == -1 || optname == IPV6_DSTOPTS) {
2419                 if (pktopt->ip6po_dest2)
2420                         free(pktopt->ip6po_dest2, M_IP6OPT);
2421                 pktopt->ip6po_dest2 = NULL;
2422         }
2423 }
2424
2425 #define PKTOPT_EXTHDRCPY(type) \
2426 do {\
2427         if (src->type) {\
2428                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2429                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2430                 if (dst->type == NULL && canwait == M_NOWAIT)\
2431                         goto bad;\
2432                 bcopy(src->type, dst->type, hlen);\
2433         }\
2434 } while (/*CONSTCOND*/ 0)
2435
2436 static int
2437 copypktopts(dst, src, canwait)
2438         struct ip6_pktopts *dst, *src;
2439         int canwait;
2440 {
2441         if (dst == NULL || src == NULL)  {
2442                 printf("ip6_clearpktopts: invalid argument\n");
2443                 return (EINVAL);
2444         }
2445
2446         dst->ip6po_hlim = src->ip6po_hlim;
2447         dst->ip6po_tclass = src->ip6po_tclass;
2448         dst->ip6po_flags = src->ip6po_flags;
2449         if (src->ip6po_pktinfo) {
2450                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2451                     M_IP6OPT, canwait);
2452                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2453                         goto bad;
2454                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2455         }
2456         if (src->ip6po_nexthop) {
2457                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2458                     M_IP6OPT, canwait);
2459                 if (dst->ip6po_nexthop == NULL)
2460                         goto bad;
2461                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2462                     src->ip6po_nexthop->sa_len);
2463         }
2464         PKTOPT_EXTHDRCPY(ip6po_hbh);
2465         PKTOPT_EXTHDRCPY(ip6po_dest1);
2466         PKTOPT_EXTHDRCPY(ip6po_dest2);
2467         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2468         return (0);
2469
2470   bad:
2471         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2472         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2473         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2474         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2475         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2476         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2477         return (ENOBUFS);
2478 }
2479 #undef PKTOPT_EXTHDRCPY
2480
2481 struct ip6_pktopts *
2482 ip6_copypktopts(src, canwait)
2483         struct ip6_pktopts *src;
2484         int canwait;
2485 {
2486         int error;
2487         struct ip6_pktopts *dst;
2488
2489         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2490         if (dst == NULL && canwait == M_NOWAIT)
2491                 return (NULL);
2492         ip6_initpktopts(dst);
2493
2494         if ((error = copypktopts(dst, src, canwait)) != 0) {
2495                 free(dst, M_IP6OPT);
2496                 return (NULL);
2497         }
2498
2499         return (dst);
2500 }
2501
2502 void
2503 ip6_freepcbopts(pktopt)
2504         struct ip6_pktopts *pktopt;
2505 {
2506         if (pktopt == NULL)
2507                 return;
2508
2509         ip6_clearpktopts(pktopt, -1);
2510
2511         free(pktopt, M_IP6OPT);
2512 }
2513
2514 /*
2515  * Set the IP6 multicast options in response to user setsockopt().
2516  */
2517 static int
2518 ip6_setmoptions(optname, im6op, m)
2519         int optname;
2520         struct ip6_moptions **im6op;
2521         struct mbuf *m;
2522 {
2523         int error = 0;
2524         u_int loop, ifindex;
2525         struct ipv6_mreq *mreq;
2526         struct ifnet *ifp;
2527         struct ip6_moptions *im6o = *im6op;
2528         struct route_in6 ro;
2529         struct in6_multi_mship *imm;
2530         struct thread *td = curthread;
2531
2532         if (im6o == NULL) {
2533                 /*
2534                  * No multicast option buffer attached to the pcb;
2535                  * allocate one and initialize to default values.
2536                  */
2537                 im6o = (struct ip6_moptions *)
2538                         malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2539
2540                 if (im6o == NULL)
2541                         return (ENOBUFS);
2542                 *im6op = im6o;
2543                 im6o->im6o_multicast_ifp = NULL;
2544                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2545                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2546                 LIST_INIT(&im6o->im6o_memberships);
2547         }
2548
2549         switch (optname) {
2550
2551         case IPV6_MULTICAST_IF:
2552                 /*
2553                  * Select the interface for outgoing multicast packets.
2554                  */
2555                 if (m == NULL || m->m_len != sizeof(u_int)) {
2556                         error = EINVAL;
2557                         break;
2558                 }
2559                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2560                 if (ifindex < 0 || if_index < ifindex) {
2561                         error = ENXIO;  /* XXX EINVAL? */
2562                         break;
2563                 }
2564                 ifp = ifnet_byindex(ifindex);
2565                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2566                         error = EADDRNOTAVAIL;
2567                         break;
2568                 }
2569                 im6o->im6o_multicast_ifp = ifp;
2570                 break;
2571
2572         case IPV6_MULTICAST_HOPS:
2573             {
2574                 /*
2575                  * Set the IP6 hoplimit for outgoing multicast packets.
2576                  */
2577                 int optval;
2578                 if (m == NULL || m->m_len != sizeof(int)) {
2579                         error = EINVAL;
2580                         break;
2581                 }
2582                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2583                 if (optval < -1 || optval >= 256)
2584                         error = EINVAL;
2585                 else if (optval == -1)
2586                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2587                 else
2588                         im6o->im6o_multicast_hlim = optval;
2589                 break;
2590             }
2591
2592         case IPV6_MULTICAST_LOOP:
2593                 /*
2594                  * Set the loopback flag for outgoing multicast packets.
2595                  * Must be zero or one.
2596                  */
2597                 if (m == NULL || m->m_len != sizeof(u_int)) {
2598                         error = EINVAL;
2599                         break;
2600                 }
2601                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2602                 if (loop > 1) {
2603                         error = EINVAL;
2604                         break;
2605                 }
2606                 im6o->im6o_multicast_loop = loop;
2607                 break;
2608
2609         case IPV6_JOIN_GROUP:
2610                 /*
2611                  * Add a multicast group membership.
2612                  * Group must be a valid IP6 multicast address.
2613                  */
2614                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2615                         error = EINVAL;
2616                         break;
2617                 }
2618                 mreq = mtod(m, struct ipv6_mreq *);
2619
2620                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2621                         /*
2622                          * We use the unspecified address to specify to accept
2623                          * all multicast addresses. Only super user is allowed
2624                          * to do this.
2625                          */
2626                         if (suser(td)) {
2627                                 error = EACCES;
2628                                 break;
2629                         }
2630                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2631                         error = EINVAL;
2632                         break;
2633                 }
2634
2635                 /*
2636                  * If no interface was explicitly specified, choose an
2637                  * appropriate one according to the given multicast address.
2638                  */
2639                 if (mreq->ipv6mr_interface == 0) {
2640                         struct sockaddr_in6 *dst;
2641
2642                         /*
2643                          * Look up the routing table for the
2644                          * address, and choose the outgoing interface.
2645                          *   XXX: is it a good approach?
2646                          */
2647                         ro.ro_rt = NULL;
2648                         dst = (struct sockaddr_in6 *)&ro.ro_dst;
2649                         bzero(dst, sizeof(*dst));
2650                         dst->sin6_family = AF_INET6;
2651                         dst->sin6_len = sizeof(*dst);
2652                         dst->sin6_addr = mreq->ipv6mr_multiaddr;
2653                         rtalloc((struct route *)&ro);
2654                         if (ro.ro_rt == NULL) {
2655                                 error = EADDRNOTAVAIL;
2656                                 break;
2657                         }
2658                         ifp = ro.ro_rt->rt_ifp;
2659                         RTFREE(ro.ro_rt);
2660                 } else {
2661                         /*
2662                          * If the interface is specified, validate it.
2663                          */
2664                         if (mreq->ipv6mr_interface < 0 ||
2665                             if_index < mreq->ipv6mr_interface) {
2666                                 error = ENXIO;  /* XXX EINVAL? */
2667                                 break;
2668                         }
2669                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2670                         if (!ifp) {
2671                                 error = ENXIO;  /* XXX EINVAL? */
2672                                 break;
2673                         }
2674                 }
2675
2676                 /*
2677                  * See if we found an interface, and confirm that it
2678                  * supports multicast
2679                  */
2680                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2681                         error = EADDRNOTAVAIL;
2682                         break;
2683                 }
2684
2685                 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2686                         error = EADDRNOTAVAIL; /* XXX: should not happen */
2687                         break;
2688                 }
2689
2690                 /*
2691                  * See if the membership already exists.
2692                  */
2693                 for (imm = im6o->im6o_memberships.lh_first;
2694                      imm != NULL; imm = imm->i6mm_chain.le_next)
2695                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2696                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2697                                                &mreq->ipv6mr_multiaddr))
2698                                 break;
2699                 if (imm != NULL) {
2700                         error = EADDRINUSE;
2701                         break;
2702                 }
2703                 /*
2704                  * Everything looks good; add a new record to the multicast
2705                  * address list for the given interface.
2706                  */
2707                 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2708                 if (imm == NULL)
2709                         break;
2710                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2711                 break;
2712
2713         case IPV6_LEAVE_GROUP:
2714                 /*
2715                  * Drop a multicast group membership.
2716                  * Group must be a valid IP6 multicast address.
2717                  */
2718                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2719                         error = EINVAL;
2720                         break;
2721                 }
2722                 mreq = mtod(m, struct ipv6_mreq *);
2723
2724                 /*
2725                  * If an interface address was specified, get a pointer
2726                  * to its ifnet structure.
2727                  */
2728                 if (mreq->ipv6mr_interface < 0 ||
2729                     if_index < mreq->ipv6mr_interface) {
2730                         error = ENXIO;  /* XXX EINVAL? */
2731                         break;
2732                 }
2733                 if (mreq->ipv6mr_interface == 0)
2734                         ifp = NULL;
2735                 else
2736                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2737
2738                 /* Fill in the scope zone ID */
2739                 if (ifp) {
2740                         if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2741                                 /* XXX: should not happen */
2742                                 error = EADDRNOTAVAIL;
2743                                 break;
2744                         }
2745                 } else if (mreq->ipv6mr_interface != 0) {
2746                         /*
2747                          * This case happens when the (positive) index is in
2748                          * the valid range, but the corresponding interface has
2749                          * been detached dynamically (XXX).
2750                          */
2751                         error = EADDRNOTAVAIL;
2752                         break;
2753                 } else {        /* ipv6mr_interface == 0 */
2754                         struct sockaddr_in6 sa6_mc;
2755
2756                         /*
2757                          * The API spec says as follows:
2758                          *  If the interface index is specified as 0, the
2759                          *  system may choose a multicast group membership to
2760                          *  drop by matching the multicast address only.
2761                          * On the other hand, we cannot disambiguate the scope
2762                          * zone unless an interface is provided.  Thus, we
2763                          * check if there's ambiguity with the default scope
2764                          * zone as the last resort.
2765                          */
2766                         bzero(&sa6_mc, sizeof(sa6_mc));
2767                         sa6_mc.sin6_family = AF_INET6;
2768                         sa6_mc.sin6_len = sizeof(sa6_mc);
2769                         sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2770                         error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2771                         if (error != 0)
2772                                 break;
2773                         mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2774                 }
2775
2776                 /*
2777                  * Find the membership in the membership list.
2778                  */
2779                 for (imm = im6o->im6o_memberships.lh_first;
2780                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2781                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2782                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2783                             &mreq->ipv6mr_multiaddr))
2784                                 break;
2785                 }
2786                 if (imm == NULL) {
2787                         /* Unable to resolve interface */
2788                         error = EADDRNOTAVAIL;
2789                         break;
2790                 }
2791                 /*
2792                  * Give up the multicast address record to which the
2793                  * membership points.
2794                  */
2795                 LIST_REMOVE(imm, i6mm_chain);
2796                 in6_delmulti(imm->i6mm_maddr);
2797                 free(imm, M_IP6MADDR);
2798                 break;
2799
2800         default:
2801                 error = EOPNOTSUPP;
2802                 break;
2803         }
2804
2805         /*
2806          * If all options have default values, no need to keep the mbuf.
2807          */
2808         if (im6o->im6o_multicast_ifp == NULL &&
2809             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2810             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2811             im6o->im6o_memberships.lh_first == NULL) {
2812                 free(*im6op, M_IP6MOPTS);
2813                 *im6op = NULL;
2814         }
2815
2816         return (error);
2817 }
2818
2819 /*
2820  * Return the IP6 multicast options in response to user getsockopt().
2821  */
2822 static int
2823 ip6_getmoptions(optname, im6o, mp)
2824         int optname;
2825         struct ip6_moptions *im6o;
2826         struct mbuf **mp;
2827 {
2828         u_int *hlim, *loop, *ifindex;
2829
2830         *mp = m_get(M_TRYWAIT, MT_HEADER);              /* XXX */
2831
2832         switch (optname) {
2833
2834         case IPV6_MULTICAST_IF:
2835                 ifindex = mtod(*mp, u_int *);
2836                 (*mp)->m_len = sizeof(u_int);
2837                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2838                         *ifindex = 0;
2839                 else
2840                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2841                 return (0);
2842
2843         case IPV6_MULTICAST_HOPS:
2844                 hlim = mtod(*mp, u_int *);
2845                 (*mp)->m_len = sizeof(u_int);
2846                 if (im6o == NULL)
2847                         *hlim = ip6_defmcasthlim;
2848                 else
2849                         *hlim = im6o->im6o_multicast_hlim;
2850                 return (0);
2851
2852         case IPV6_MULTICAST_LOOP:
2853                 loop = mtod(*mp, u_int *);
2854                 (*mp)->m_len = sizeof(u_int);
2855                 if (im6o == NULL)
2856                         *loop = ip6_defmcasthlim;
2857                 else
2858                         *loop = im6o->im6o_multicast_loop;
2859                 return (0);
2860
2861         default:
2862                 return (EOPNOTSUPP);
2863         }
2864 }
2865
2866 /*
2867  * Discard the IP6 multicast options.
2868  */
2869 void
2870 ip6_freemoptions(im6o)
2871         struct ip6_moptions *im6o;
2872 {
2873         struct in6_multi_mship *imm;
2874
2875         if (im6o == NULL)
2876                 return;
2877
2878         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2879                 LIST_REMOVE(imm, i6mm_chain);
2880                 if (imm->i6mm_maddr)
2881                         in6_delmulti(imm->i6mm_maddr);
2882                 free(imm, M_IP6MADDR);
2883         }
2884         free(im6o, M_IP6MOPTS);
2885 }
2886
2887 /*
2888  * Set IPv6 outgoing packet options based on advanced API.
2889  */
2890 int
2891 ip6_setpktopts(control, opt, stickyopt, priv, uproto)
2892         struct mbuf *control;
2893         struct ip6_pktopts *opt, *stickyopt;
2894         int priv, uproto;
2895 {
2896         struct cmsghdr *cm = 0;
2897
2898         if (control == NULL || opt == NULL)
2899                 return (EINVAL);
2900
2901         ip6_initpktopts(opt);
2902         if (stickyopt) {
2903                 int error;
2904
2905                 /*
2906                  * If stickyopt is provided, make a local copy of the options
2907                  * for this particular packet, then override them by ancillary
2908                  * objects.
2909                  * XXX: copypktopts() does not copy the cached route to a next
2910                  * hop (if any).  This is not very good in terms of efficiency,
2911                  * but we can allow this since this option should be rarely
2912                  * used.
2913                  */
2914                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2915                         return (error);
2916         }
2917
2918         /*
2919          * XXX: Currently, we assume all the optional information is stored
2920          * in a single mbuf.
2921          */
2922         if (control->m_next)
2923                 return (EINVAL);
2924
2925         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2926             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2927                 int error;
2928
2929                 if (control->m_len < CMSG_LEN(0))
2930                         return (EINVAL);
2931
2932                 cm = mtod(control, struct cmsghdr *);
2933                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2934                         return (EINVAL);
2935                 if (cm->cmsg_level != IPPROTO_IPV6)
2936                         continue;
2937
2938                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2939                     cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2940                 if (error)
2941                         return (error);
2942         }
2943
2944         return (0);
2945 }
2946
2947 /*
2948  * Set a particular packet option, as a sticky option or an ancillary data
2949  * item.  "len" can be 0 only when it's a sticky option.
2950  * We have 4 cases of combination of "sticky" and "cmsg":
2951  * "sticky=0, cmsg=0": impossible
2952  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2953  * "sticky=1, cmsg=0": RFC3542 socket option
2954  * "sticky=1, cmsg=1": RFC2292 socket option
2955  */
2956 static int
2957 ip6_setpktopt(optname, buf, len, opt, priv, sticky, cmsg, uproto)
2958         int optname, len, priv, sticky, cmsg, uproto;
2959         u_char *buf;
2960         struct ip6_pktopts *opt;
2961 {
2962         int minmtupolicy, preftemp;
2963
2964         if (!sticky && !cmsg) {
2965 #ifdef DIAGNOSTIC
2966                 printf("ip6_setpktopt: impossible case\n");
2967 #endif
2968                 return (EINVAL);
2969         }
2970
2971         /*
2972          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2973          * not be specified in the context of RFC3542.  Conversely,
2974          * RFC3542 types should not be specified in the context of RFC2292.
2975          */
2976         if (!cmsg) {
2977                 switch (optname) {
2978                 case IPV6_2292PKTINFO:
2979                 case IPV6_2292HOPLIMIT:
2980                 case IPV6_2292NEXTHOP:
2981                 case IPV6_2292HOPOPTS:
2982                 case IPV6_2292DSTOPTS:
2983                 case IPV6_2292RTHDR:
2984                 case IPV6_2292PKTOPTIONS:
2985                         return (ENOPROTOOPT);
2986                 }
2987         }
2988         if (sticky && cmsg) {
2989                 switch (optname) {
2990                 case IPV6_PKTINFO:
2991                 case IPV6_HOPLIMIT:
2992                 case IPV6_NEXTHOP:
2993                 case IPV6_HOPOPTS:
2994                 case IPV6_DSTOPTS:
2995                 case IPV6_RTHDRDSTOPTS:
2996                 case IPV6_RTHDR:
2997                 case IPV6_USE_MIN_MTU:
2998                 case IPV6_DONTFRAG:
2999                 case IPV6_TCLASS:
3000                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
3001                         return (ENOPROTOOPT);
3002                 }
3003         }
3004
3005         switch (optname) {
3006         case IPV6_2292PKTINFO:
3007         case IPV6_PKTINFO:
3008         {
3009                 struct ifnet *ifp = NULL;
3010                 struct in6_pktinfo *pktinfo;
3011
3012                 if (len != sizeof(struct in6_pktinfo))
3013                         return (EINVAL);
3014
3015                 pktinfo = (struct in6_pktinfo *)buf;
3016
3017                 /*
3018                  * An application can clear any sticky IPV6_PKTINFO option by
3019                  * doing a "regular" setsockopt with ipi6_addr being
3020                  * in6addr_any and ipi6_ifindex being zero.
3021                  * [RFC 3542, Section 6]
3022                  */
3023                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3024                     pktinfo->ipi6_ifindex == 0 &&
3025                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3026                         ip6_clearpktopts(opt, optname);
3027                         break;
3028                 }
3029
3030                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3031                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3032                         return (EINVAL);
3033                 }
3034
3035                 /* validate the interface index if specified. */
3036                 if (pktinfo->ipi6_ifindex > if_index ||
3037                     pktinfo->ipi6_ifindex < 0) {
3038                          return (ENXIO);
3039                 }
3040                 if (pktinfo->ipi6_ifindex) {
3041                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3042                         if (ifp == NULL)
3043                                 return (ENXIO);
3044                 }
3045
3046                 /*
3047                  * We store the address anyway, and let in6_selectsrc()
3048                  * validate the specified address.  This is because ipi6_addr
3049                  * may not have enough information about its scope zone, and
3050                  * we may need additional information (such as outgoing
3051                  * interface or the scope zone of a destination address) to
3052                  * disambiguate the scope.
3053                  * XXX: the delay of the validation may confuse the
3054                  * application when it is used as a sticky option.
3055                  */
3056                 if (opt->ip6po_pktinfo == NULL) {
3057                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3058                             M_IP6OPT, M_NOWAIT);
3059                         if (opt->ip6po_pktinfo == NULL)
3060                                 return (ENOBUFS);
3061                 }
3062                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3063                 break;
3064         }
3065
3066         case IPV6_2292HOPLIMIT:
3067         case IPV6_HOPLIMIT:
3068         {
3069                 int *hlimp;
3070
3071                 /*
3072                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3073                  * to simplify the ordering among hoplimit options.
3074                  */
3075                 if (optname == IPV6_HOPLIMIT && sticky)
3076                         return (ENOPROTOOPT);
3077
3078                 if (len != sizeof(int))
3079                         return (EINVAL);
3080                 hlimp = (int *)buf;
3081                 if (*hlimp < -1 || *hlimp > 255)
3082                         return (EINVAL);
3083
3084                 opt->ip6po_hlim = *hlimp;
3085                 break;
3086         }
3087
3088         case IPV6_TCLASS:
3089         {
3090                 int tclass;
3091
3092                 if (len != sizeof(int))
3093                         return (EINVAL);
3094                 tclass = *(int *)buf;
3095                 if (tclass < -1 || tclass > 255)
3096                         return (EINVAL);
3097
3098                 opt->ip6po_tclass = tclass;
3099                 break;
3100         }
3101
3102         case IPV6_2292NEXTHOP:
3103         case IPV6_NEXTHOP:
3104                 if (!priv)
3105                         return (EPERM);
3106
3107                 if (len == 0) { /* just remove the option */
3108                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3109                         break;
3110                 }
3111
3112                 /* check if cmsg_len is large enough for sa_len */
3113                 if (len < sizeof(struct sockaddr) || len < *buf)
3114                         return (EINVAL);
3115
3116                 switch (((struct sockaddr *)buf)->sa_family) {
3117                 case AF_INET6:
3118                 {
3119                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3120                         int error;
3121
3122                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3123                                 return (EINVAL);
3124
3125                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3126                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3127                                 return (EINVAL);
3128                         }
3129                         if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3130                             != 0) {
3131                                 return (error);
3132                         }
3133                         break;
3134                 }
3135                 case AF_LINK:   /* should eventually be supported */
3136                 default:
3137                         return (EAFNOSUPPORT);
3138                 }
3139
3140                 /* turn off the previous option, then set the new option. */
3141                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3142                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3143                 if (opt->ip6po_nexthop == NULL)
3144                         return (ENOBUFS);
3145                 bcopy(buf, opt->ip6po_nexthop, *buf);
3146                 break;
3147
3148         case IPV6_2292HOPOPTS:
3149         case IPV6_HOPOPTS:
3150         {
3151                 struct ip6_hbh *hbh;
3152                 int hbhlen;
3153
3154                 /*
3155                  * XXX: We don't allow a non-privileged user to set ANY HbH
3156                  * options, since per-option restriction has too much
3157                  * overhead.
3158                  */
3159                 if (!priv)
3160                         return (EPERM);
3161
3162                 if (len == 0) {
3163                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3164                         break;  /* just remove the option */
3165                 }
3166
3167                 /* message length validation */
3168                 if (len < sizeof(struct ip6_hbh))
3169                         return (EINVAL);
3170                 hbh = (struct ip6_hbh *)buf;
3171                 hbhlen = (hbh->ip6h_len + 1) << 3;
3172                 if (len != hbhlen)
3173                         return (EINVAL);
3174
3175                 /* turn off the previous option, then set the new option. */
3176                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3177                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3178                 if (opt->ip6po_hbh == NULL)
3179                         return (ENOBUFS);
3180                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3181
3182                 break;
3183         }
3184
3185         case IPV6_2292DSTOPTS:
3186         case IPV6_DSTOPTS:
3187         case IPV6_RTHDRDSTOPTS:
3188         {
3189                 struct ip6_dest *dest, **newdest = NULL;
3190                 int destlen;
3191
3192                 if (!priv)      /* XXX: see the comment for IPV6_HOPOPTS */
3193                         return (EPERM);
3194
3195                 if (len == 0) {
3196                         ip6_clearpktopts(opt, optname);
3197                         break;  /* just remove the option */
3198                 }
3199
3200                 /* message length validation */
3201                 if (len < sizeof(struct ip6_dest))
3202                         return (EINVAL);
3203                 dest = (struct ip6_dest *)buf;
3204                 destlen = (dest->ip6d_len + 1) << 3;
3205                 if (len != destlen)
3206                         return (EINVAL);
3207
3208                 /*
3209                  * Determine the position that the destination options header
3210                  * should be inserted; before or after the routing header.
3211                  */
3212                 switch (optname) {
3213                 case IPV6_2292DSTOPTS:
3214                         /*
3215                          * The old advacned API is ambiguous on this point.
3216                          * Our approach is to determine the position based
3217                          * according to the existence of a routing header.
3218                          * Note, however, that this depends on the order of the
3219                          * extension headers in the ancillary data; the 1st
3220                          * part of the destination options header must appear
3221                          * before the routing header in the ancillary data,
3222                          * too.
3223                          * RFC3542 solved the ambiguity by introducing
3224                          * separate ancillary data or option types.
3225                          */
3226                         if (opt->ip6po_rthdr == NULL)
3227                                 newdest = &opt->ip6po_dest1;
3228                         else
3229                                 newdest = &opt->ip6po_dest2;
3230                         break;
3231                 case IPV6_RTHDRDSTOPTS:
3232                         newdest = &opt->ip6po_dest1;
3233                         break;
3234                 case IPV6_DSTOPTS:
3235                         newdest = &opt->ip6po_dest2;
3236                         break;
3237                 }
3238
3239                 /* turn off the previous option, then set the new option. */
3240                 ip6_clearpktopts(opt, optname);
3241                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3242                 if (*newdest == NULL)
3243                         return (ENOBUFS);
3244                 bcopy(dest, *newdest, destlen);
3245
3246                 break;
3247         }
3248
3249         case IPV6_2292RTHDR:
3250         case IPV6_RTHDR:
3251         {
3252                 struct ip6_rthdr *rth;
3253                 int rthlen;
3254
3255                 if (len == 0) {
3256                         ip6_clearpktopts(opt, IPV6_RTHDR);
3257                         break;  /* just remove the option */
3258                 }
3259
3260                 /* message length validation */
3261                 if (len < sizeof(struct ip6_rthdr))
3262                         return (EINVAL);
3263                 rth = (struct ip6_rthdr *)buf;
3264                 rthlen = (rth->ip6r_len + 1) << 3;
3265                 if (len != rthlen)
3266                         return (EINVAL);
3267
3268                 switch (rth->ip6r_type) {
3269                 case IPV6_RTHDR_TYPE_0:
3270                         if (rth->ip6r_len == 0) /* must contain one addr */
3271                                 return (EINVAL);
3272                         if (rth->ip6r_len % 2) /* length must be even */
3273                                 return (EINVAL);
3274                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3275                                 return (EINVAL);
3276                         break;
3277                 default:
3278                         return (EINVAL);        /* not supported */
3279                 }
3280
3281                 /* turn off the previous option */
3282                 ip6_clearpktopts(opt, IPV6_RTHDR);
3283                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3284                 if (opt->ip6po_rthdr == NULL)
3285                         return (ENOBUFS);
3286                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3287
3288                 break;
3289         }
3290
3291         case IPV6_USE_MIN_MTU:
3292                 if (len != sizeof(int))
3293                         return (EINVAL);
3294                 minmtupolicy = *(int *)buf;
3295                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3296                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3297                     minmtupolicy != IP6PO_MINMTU_ALL) {
3298                         return (EINVAL);
3299                 }
3300                 opt->ip6po_minmtu = minmtupolicy;
3301                 break;
3302
3303         case IPV6_DONTFRAG:
3304                 if (len != sizeof(int))
3305                         return (EINVAL);
3306
3307                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3308                         /*
3309                          * we ignore this option for TCP sockets.
3310                          * (RFC3542 leaves this case unspecified.)
3311                          */
3312                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3313                 } else
3314                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3315                 break;
3316
3317         case IPV6_PREFER_TEMPADDR:
3318                 if (len != sizeof(int))
3319                         return (EINVAL);
3320                 preftemp = *(int *)buf;
3321                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3322                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3323                     preftemp != IP6PO_TEMPADDR_PREFER) {
3324                         return (EINVAL);
3325                 }
3326                 opt->ip6po_prefer_tempaddr = preftemp;
3327                 break;
3328
3329         default:
3330                 return (ENOPROTOOPT);
3331         } /* end of switch */
3332
3333         return (0);
3334 }
3335
3336 /*
3337  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3338  * packet to the input queue of a specified interface.  Note that this
3339  * calls the output routine of the loopback "driver", but with an interface
3340  * pointer that might NOT be &loif -- easier than replicating that code here.
3341  */
3342 void
3343 ip6_mloopback(ifp, m, dst)
3344         struct ifnet *ifp;
3345         struct mbuf *m;
3346         struct sockaddr_in6 *dst;
3347 {
3348         struct mbuf *copym;
3349         struct ip6_hdr *ip6;
3350
3351         copym = m_copy(m, 0, M_COPYALL);
3352         if (copym == NULL)
3353                 return;
3354
3355         /*
3356          * Make sure to deep-copy IPv6 header portion in case the data
3357          * is in an mbuf cluster, so that we can safely override the IPv6
3358          * header portion later.
3359          */
3360         if ((copym->m_flags & M_EXT) != 0 ||
3361             copym->m_len < sizeof(struct ip6_hdr)) {
3362                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3363                 if (copym == NULL)
3364                         return;
3365         }
3366
3367 #ifdef DIAGNOSTIC
3368         if (copym->m_len < sizeof(*ip6)) {
3369                 m_freem(copym);
3370                 return;
3371         }
3372 #endif
3373
3374         ip6 = mtod(copym, struct ip6_hdr *);
3375         /*
3376          * clear embedded scope identifiers if necessary.
3377          * in6_clearscope will touch the addresses only when necessary.
3378          */
3379         in6_clearscope(&ip6->ip6_src);
3380         in6_clearscope(&ip6->ip6_dst);
3381
3382         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3383 }
3384
3385 /*
3386  * Chop IPv6 header off from the payload.
3387  */
3388 static int
3389 ip6_splithdr(m, exthdrs)
3390         struct mbuf *m;
3391         struct ip6_exthdrs *exthdrs;
3392 {
3393         struct mbuf *mh;
3394         struct ip6_hdr *ip6;
3395
3396         ip6 = mtod(m, struct ip6_hdr *);
3397         if (m->m_len > sizeof(*ip6)) {
3398                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3399                 if (mh == 0) {
3400                         m_freem(m);
3401                         return ENOBUFS;
3402                 }
3403                 M_MOVE_PKTHDR(mh, m);
3404                 MH_ALIGN(mh, sizeof(*ip6));
3405                 m->m_len -= sizeof(*ip6);
3406                 m->m_data += sizeof(*ip6);
3407                 mh->m_next = m;
3408                 m = mh;
3409                 m->m_len = sizeof(*ip6);
3410                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3411         }
3412         exthdrs->ip6e_ip6 = m;
3413         return 0;
3414 }
3415
3416 /*
3417  * Compute IPv6 extension header length.
3418  */
3419 int
3420 ip6_optlen(in6p)
3421         struct in6pcb *in6p;
3422 {
3423         int len;
3424
3425         if (!in6p->in6p_outputopts)
3426                 return 0;
3427
3428         len = 0;
3429 #define elen(x) \
3430     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3431
3432         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3433         if (in6p->in6p_outputopts->ip6po_rthdr)
3434                 /* dest1 is valid with rthdr only */
3435                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3436         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3437         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3438         return len;
3439 #undef elen
3440 }