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_ip6fw.h"
  65 #include "opt_inet.h"
  66 #include "opt_inet6.h"
  67 #include "opt_ipsec.h"
  68
  69 #include <sys/param.h>
  70 #include <sys/malloc.h>
  71 #include <sys/mbuf.h>
  72 #include <sys/proc.h>
  73 #include <sys/errno.h>
  74 #include <sys/protosw.h>
  75 #include <sys/socket.h>
  76 #include <sys/socketvar.h>
  77 #include <sys/systm.h>
  78 #include <sys/kernel.h>
  79
  80 #include <net/if.h>
  81 #include <net/netisr.h>
  82 #include <net/route.h>
  83 #include <net/pfil.h>
  84
  85 #include <netinet/in.h>
  86 #include <netinet/in_var.h>
  87 #include <netinet6/in6_var.h>
  88 #include <netinet/ip6.h>
  89 #include <netinet/icmp6.h>
  90 #include <netinet6/ip6_var.h>
  91 #include <netinet/in_pcb.h>
  92 #include <netinet/tcp_var.h>
  93 #include <netinet6/nd6.h>
  94
  95 #ifdef IPSEC
  96 #include <netinet6/ipsec.h>
  97 #ifdef INET6
  98 #include <netinet6/ipsec6.h>
  99 #endif
 100 #include <netkey/key.h>
 101 #endif /* IPSEC */
 102
 103 #ifdef FAST_IPSEC
 104 #include <netipsec/ipsec.h>
 105 #include <netipsec/ipsec6.h>
 106 #include <netipsec/key.h>
 107 #endif /* FAST_IPSEC */
 108
 109 #include <netinet6/ip6_fw.h>
 110
 111 #include <net/net_osdep.h>
 112
 113 #include <netinet6/ip6protosw.h>
 114 #include <netinet6/scope6_var.h>
 115
 116 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
 117
 118 struct ip6_exthdrs {
 119         struct mbuf *ip6e_ip6;
 120         struct mbuf *ip6e_hbh;
 121         struct mbuf *ip6e_dest1;
 122         struct mbuf *ip6e_rthdr;
 123         struct mbuf *ip6e_dest2;
 124 };
 125
 126 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 127                            int, int));
 128 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 129         struct socket *, struct sockopt *));
 130 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
 131 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int,
 132         int, int, int));
 133
 134 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
 135 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
 136 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
 137 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 138         struct ip6_frag **));
 139 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
 140 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
 141 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 142         struct ifnet *, struct in6_addr *, u_long *, int *));
 143 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int));
 144
 145
 146 /*
 147  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 148  * header (with pri, len, nxt, hlim, src, dst).
 149  * This function may modify ver and hlim only.
 150  * The mbuf chain containing the packet will be freed.
 151  * The mbuf opt, if present, will not be freed.
 152  *
 153  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 154  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 155  * which is rt_rmx.rmx_mtu.
 156  */
 157 int
 158 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
 159         struct mbuf *m0;
 160         struct ip6_pktopts *opt;
 161         struct route_in6 *ro;
 162         int flags;
 163         struct ip6_moptions *im6o;
 164         struct ifnet **ifpp;            /* XXX: just for statistics */
 165         struct inpcb *inp;
 166 {
 167         struct ip6_hdr *ip6, *mhip6;
 168         struct ifnet *ifp, *origifp;
 169         struct mbuf *m = m0;
 170         int hlen, tlen, len, off;
 171         struct route_in6 ip6route;
 172         struct rtentry *rt = NULL;
 173         struct sockaddr_in6 *dst, src_sa, dst_sa;
 174         struct in6_addr odst;
 175         int error = 0;
 176         struct in6_ifaddr *ia = NULL;
 177         u_long mtu;
 178         int alwaysfrag, dontfrag;
 179         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 180         struct ip6_exthdrs exthdrs;
 181         struct in6_addr finaldst, src0, dst0;
 182         u_int32_t zone;
 183         struct route_in6 *ro_pmtu = NULL;
 184         int hdrsplit = 0;
 185         int needipsec = 0;
 186 #if defined(IPSEC) || defined(FAST_IPSEC)
 187         int needipsectun = 0;
 188         struct secpolicy *sp = NULL;
 189 #endif /*IPSEC || FAST_IPSEC*/
 190
 191         ip6 = mtod(m, struct ip6_hdr *);
 192         finaldst = ip6->ip6_dst;
 193
 194 #define MAKE_EXTHDR(hp, mp)                                             \
 195     do {                                                                \
 196         if (hp) {                                                       \
 197                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 198                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 199                     ((eh)->ip6e_len + 1) << 3);                         \
 200                 if (error)                                              \
 201                         goto freehdrs;                                  \
 202         }                                                               \
 203     } while (/*CONSTCOND*/ 0)
 204
 205         bzero(&exthdrs, sizeof(exthdrs));
 206
 207         if (opt) {
 208                 /* Hop-by-Hop options header */
 209                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 210                 /* Destination options header(1st part) */
 211                 if (opt->ip6po_rthdr) {
 212                         /*
 213                          * Destination options header(1st part)
 214                          * This only makes sence with a routing header.
 215                          * See Section 9.2 of RFC 3542.
 216                          * Disabling this part just for MIP6 convenience is
 217                          * a bad idea.  We need to think carefully about a
 218                          * way to make the advanced API coexist with MIP6
 219                          * options, which might automatically be inserted in
 220                          * the kernel.
 221                          */
 222                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 223                 }
 224                 /* Routing header */
 225                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 226                 /* Destination options header(2nd part) */
 227                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 228         }
 229
 230 #ifdef IPSEC
 231         /* get a security policy for this packet */
 232         if (inp == NULL)
 233                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 234         else
 235                 sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
 236
 237         if (sp == NULL) {
 238                 ipsec6stat.out_inval++;
 239                 goto freehdrs;
 240         }
 241
 242         error = 0;
 243
 244         /* check policy */
 245         switch (sp->policy) {
 246         case IPSEC_POLICY_DISCARD:
 247                 /*
 248                  * This packet is just discarded.
 249                  */
 250                 ipsec6stat.out_polvio++;
 251                 goto freehdrs;
 252
 253         case IPSEC_POLICY_BYPASS:
 254         case IPSEC_POLICY_NONE:
 255                 /* no need to do IPsec. */
 256                 needipsec = 0;
 257                 break;
 258
 259         case IPSEC_POLICY_IPSEC:
 260                 if (sp->req == NULL) {
 261                         /* acquire a policy */
 262                         error = key_spdacquire(sp);
 263                         goto freehdrs;
 264                 }
 265                 needipsec = 1;
 266                 break;
 267
 268         case IPSEC_POLICY_ENTRUST:
 269         default:
 270                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 271         }
 272 #endif /* IPSEC */
 273 #ifdef FAST_IPSEC
 274         /* get a security policy for this packet */
 275         if (inp == NULL)
 276                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 277         else
 278                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 279
 280         if (sp == NULL) {
 281                 newipsecstat.ips_out_inval++;
 282                 goto freehdrs;
 283         }
 284
 285         error = 0;
 286
 287         /* check policy */
 288         switch (sp->policy) {
 289         case IPSEC_POLICY_DISCARD:
 290                 /*
 291                  * This packet is just discarded.
 292                  */
 293                 newipsecstat.ips_out_polvio++;
 294                 goto freehdrs;
 295
 296         case IPSEC_POLICY_BYPASS:
 297         case IPSEC_POLICY_NONE:
 298                 /* no need to do IPsec. */
 299                 needipsec = 0;
 300                 break;
 301
 302         case IPSEC_POLICY_IPSEC:
 303                 if (sp->req == NULL) {
 304                         /* acquire a policy */
 305                         error = key_spdacquire(sp);
 306                         goto freehdrs;
 307                 }
 308                 needipsec = 1;
 309                 break;
 310
 311         case IPSEC_POLICY_ENTRUST:
 312         default:
 313                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 314         }
 315 #endif /* FAST_IPSEC */
 316
 317         /*
 318          * Calculate the total length of the extension header chain.
 319          * Keep the length of the unfragmentable part for fragmentation.
 320          */
 321         optlen = 0;
 322         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 323         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 324         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 325         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 326         /* NOTE: we don't add AH/ESP length here. do that later. */
 327         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 328
 329         /*
 330          * If we need IPsec, or there is at least one extension header,
 331          * separate IP6 header from the payload.
 332          */
 333         if ((needipsec || optlen) && !hdrsplit) {
 334                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 335                         m = NULL;
 336                         goto freehdrs;
 337                 }
 338                 m = exthdrs.ip6e_ip6;
 339                 hdrsplit++;
 340         }
 341
 342         /* adjust pointer */
 343         ip6 = mtod(m, struct ip6_hdr *);
 344
 345         /* adjust mbuf packet header length */
 346         m->m_pkthdr.len += optlen;
 347         plen = m->m_pkthdr.len - sizeof(*ip6);
 348
 349         /* If this is a jumbo payload, insert a jumbo payload option. */
 350         if (plen > IPV6_MAXPACKET) {
 351                 if (!hdrsplit) {
 352                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 353                                 m = NULL;
 354                                 goto freehdrs;
 355                         }
 356                         m = exthdrs.ip6e_ip6;
 357                         hdrsplit++;
 358                 }
 359                 /* adjust pointer */
 360                 ip6 = mtod(m, struct ip6_hdr *);
 361                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 362                         goto freehdrs;
 363                 ip6->ip6_plen = 0;
 364         } else
 365                 ip6->ip6_plen = htons(plen);
 366
 367         /*
 368          * Concatenate headers and fill in next header fields.
 369          * Here we have, on "m"
 370          *      IPv6 payload
 371          * and we insert headers accordingly.  Finally, we should be getting:
 372          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 373          *
 374          * during the header composing process, "m" points to IPv6 header.
 375          * "mprev" points to an extension header prior to esp.
 376          */
 377         {
 378                 u_char *nexthdrp = &ip6->ip6_nxt;
 379                 struct mbuf *mprev = m;
 380
 381                 /*
 382                  * we treat dest2 specially.  this makes IPsec processing
 383                  * much easier.  the goal here is to make mprev point the
 384                  * mbuf prior to dest2.
 385                  *
 386                  * result: IPv6 dest2 payload
 387                  * m and mprev will point to IPv6 header.
 388                  */
 389                 if (exthdrs.ip6e_dest2) {
 390                         if (!hdrsplit)
 391                                 panic("assumption failed: hdr not split");
 392                         exthdrs.ip6e_dest2->m_next = m->m_next;
 393                         m->m_next = exthdrs.ip6e_dest2;
 394                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 395                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 396                 }
 397
 398 #define MAKE_CHAIN(m, mp, p, i)\
 399     do {\
 400         if (m) {\
 401                 if (!hdrsplit) \
 402                         panic("assumption failed: hdr not split"); \
 403                 *mtod((m), u_char *) = *(p);\
 404                 *(p) = (i);\
 405                 p = mtod((m), u_char *);\
 406                 (m)->m_next = (mp)->m_next;\
 407                 (mp)->m_next = (m);\
 408                 (mp) = (m);\
 409         }\
 410     } while (/*CONSTCOND*/ 0)
 411                 /*
 412                  * result: IPv6 hbh dest1 rthdr dest2 payload
 413                  * m will point to IPv6 header.  mprev will point to the
 414                  * extension header prior to dest2 (rthdr in the above case).
 415                  */
 416                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 417                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 418                     IPPROTO_DSTOPTS);
 419                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 420                     IPPROTO_ROUTING);
 421
 422 #if defined(IPSEC) || defined(FAST_IPSEC)
 423                 if (!needipsec)
 424                         goto skip_ipsec2;
 425
 426                 /*
 427                  * pointers after IPsec headers are not valid any more.
 428                  * other pointers need a great care too.
 429                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 430                  */
 431                 exthdrs.ip6e_dest2 = NULL;
 432
 433             {
 434                 struct ip6_rthdr *rh = NULL;
 435                 int segleft_org = 0;
 436                 struct ipsec_output_state state;
 437
 438                 if (exthdrs.ip6e_rthdr) {
 439                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 440                         segleft_org = rh->ip6r_segleft;
 441                         rh->ip6r_segleft = 0;
 442                 }
 443
 444                 bzero(&state, sizeof(state));
 445                 state.m = m;
 446                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 447                     &needipsectun);
 448                 m = state.m;
 449                 if (error) {
 450                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 451                         m = NULL;
 452                         switch (error) {
 453                         case EHOSTUNREACH:
 454                         case ENETUNREACH:
 455                         case EMSGSIZE:
 456                         case ENOBUFS:
 457                         case ENOMEM:
 458                                 break;
 459                         default:
 460                                 printf("ip6_output (ipsec): error code %d\n", error);
 461                                 /* FALLTHROUGH */
 462                         case ENOENT:
 463                                 /* don't show these error codes to the user */
 464                                 error = 0;
 465                                 break;
 466                         }
 467                         goto bad;
 468                 }
 469                 if (exthdrs.ip6e_rthdr) {
 470                         /* ah6_output doesn't modify mbuf chain */
 471                         rh->ip6r_segleft = segleft_org;
 472                 }
 473             }
 474 skip_ipsec2:;
 475 #endif
 476         }
 477
 478         /*
 479          * If there is a routing header, replace the destination address field
 480          * with the first hop of the routing header.
 481          */
 482         if (exthdrs.ip6e_rthdr) {
 483                 struct ip6_rthdr *rh =
 484                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 485                                                   struct ip6_rthdr *));
 486                 struct ip6_rthdr0 *rh0;
 487                 struct in6_addr *addr;
 488                 struct sockaddr_in6 sa;
 489
 490                 switch (rh->ip6r_type) {
 491                 case IPV6_RTHDR_TYPE_0:
 492                          rh0 = (struct ip6_rthdr0 *)rh;
 493                          addr = (struct in6_addr *)(rh0 + 1);
 494
 495                          /*
 496                           * construct a sockaddr_in6 form of
 497                           * the first hop.
 498                           *
 499                           * XXX: we may not have enough
 500                           * information about its scope zone;
 501                           * there is no standard API to pass
 502                           * the information from the
 503                           * application.
 504                           */
 505                          bzero(&sa, sizeof(sa));
 506                          sa.sin6_family = AF_INET6;
 507                          sa.sin6_len = sizeof(sa);
 508                          sa.sin6_addr = addr[0];
 509                          if ((error = sa6_embedscope(&sa,
 510                              ip6_use_defzone)) != 0) {
 511                                  goto bad;
 512                          }
 513                          ip6->ip6_dst = sa.sin6_addr;
 514                          bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
 515                              * (rh0->ip6r0_segleft - 1));
 516                          addr[rh0->ip6r0_segleft - 1] = finaldst;
 517                          /* XXX */
 518                          in6_clearscope(addr + rh0->ip6r0_segleft - 1);
 519                          break;
 520                 default:        /* is it possible? */
 521                          error = EINVAL;
 522                          goto bad;
 523                 }
 524         }
 525
 526         /* Source address validation */
 527         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 528             (flags & IPV6_UNSPECSRC) == 0) {
 529                 error = EOPNOTSUPP;
 530                 ip6stat.ip6s_badscope++;
 531                 goto bad;
 532         }
 533         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 534                 error = EOPNOTSUPP;
 535                 ip6stat.ip6s_badscope++;
 536                 goto bad;
 537         }
 538
 539         ip6stat.ip6s_localout++;
 540
 541         /*
 542          * Route packet.
 543          */
 544         if (ro == 0) {
 545                 ro = &ip6route;
 546                 bzero((caddr_t)ro, sizeof(*ro));
 547         }
 548         ro_pmtu = ro;
 549         if (opt && opt->ip6po_rthdr)
 550                 ro = &opt->ip6po_route;
 551         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 552
 553 again:
 554         /*
 555          * if specified, try to fill in the traffic class field.
 556          * do not override if a non-zero value is already set.
 557          * we check the diffserv field and the ecn field separately.
 558          */
 559         if (opt && opt->ip6po_tclass >= 0) {
 560                 int mask = 0;
 561
 562                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 563                         mask |= 0xfc;
 564                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 565                         mask |= 0x03;
 566                 if (mask != 0)
 567                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 568         }
 569
 570         /* fill in or override the hop limit field, if necessary. */
 571         if (opt && opt->ip6po_hlim != -1)
 572                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 573         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 574                 if (im6o != NULL)
 575                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 576                 else
 577                         ip6->ip6_hlim = ip6_defmcasthlim;
 578         }
 579
 580 #if defined(IPSEC) || defined(FAST_IPSEC)
 581         if (needipsec && needipsectun) {
 582                 struct ipsec_output_state state;
 583
 584                 /*
 585                  * All the extension headers will become inaccessible
 586                  * (since they can be encrypted).
 587                  * Don't panic, we need no more updates to extension headers
 588                  * on inner IPv6 packet (since they are now encapsulated).
 589                  *
 590                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 591                  */
 592                 bzero(&exthdrs, sizeof(exthdrs));
 593                 exthdrs.ip6e_ip6 = m;
 594
 595                 bzero(&state, sizeof(state));
 596                 state.m = m;
 597                 state.ro = (struct route *)ro;
 598                 state.dst = (struct sockaddr *)dst;
 599
 600                 error = ipsec6_output_tunnel(&state, sp, flags);
 601
 602                 m = state.m;
 603                 ro = (struct route_in6 *)state.ro;
 604                 dst = (struct sockaddr_in6 *)state.dst;
 605                 if (error) {
 606                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 607                         m0 = m = NULL;
 608                         m = NULL;
 609                         switch (error) {
 610                         case EHOSTUNREACH:
 611                         case ENETUNREACH:
 612                         case EMSGSIZE:
 613                         case ENOBUFS:
 614                         case ENOMEM:
 615                                 break;
 616                         default:
 617                                 printf("ip6_output (ipsec): error code %d\n", error);
 618                                 /* FALLTHROUGH */
 619                         case ENOENT:
 620                                 /* don't show these error codes to the user */
 621                                 error = 0;
 622                                 break;
 623                         }
 624                         goto bad;
 625                 }
 626
 627                 exthdrs.ip6e_ip6 = m;
 628         }
 629 #endif /* IPSEC */
 630
 631         /* adjust pointer */
 632         ip6 = mtod(m, struct ip6_hdr *);
 633
 634         bzero(&dst_sa, sizeof(dst_sa));
 635         dst_sa.sin6_family = AF_INET6;
 636         dst_sa.sin6_len = sizeof(dst_sa);
 637         dst_sa.sin6_addr = ip6->ip6_dst;
 638         if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
 639             &ifp, &rt, 0)) != 0) {
 640                 switch (error) {
 641                 case EHOSTUNREACH:
 642                         ip6stat.ip6s_noroute++;
 643                         break;
 644                 case EADDRNOTAVAIL:
 645                 default:
 646                         break; /* XXX statistics? */
 647                 }
 648                 if (ifp != NULL)
 649                         in6_ifstat_inc(ifp, ifs6_out_discard);
 650                 goto bad;
 651         }
 652         if (rt == NULL) {
 653                 /*
 654                  * If in6_selectroute() does not return a route entry,
 655                  * dst may not have been updated.
 656                  */
 657                 *dst = dst_sa;  /* XXX */
 658         }
 659
 660         /*
 661          * then rt (for unicast) and ifp must be non-NULL valid values.
 662          */
 663         if ((flags & IPV6_FORWARDING) == 0) {
 664                 /* XXX: the FORWARDING flag can be set for mrouting. */
 665                 in6_ifstat_inc(ifp, ifs6_out_request);
 666         }
 667         if (rt != NULL) {
 668                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 669                 rt->rt_use++;
 670         }
 671
 672         /*
 673          * The outgoing interface must be in the zone of source and
 674          * destination addresses.  We should use ia_ifp to support the
 675          * case of sending packets to an address of our own.
 676          */
 677         if (ia != NULL && ia->ia_ifp)
 678                 origifp = ia->ia_ifp;
 679         else
 680                 origifp = ifp;
 681
 682         src0 = ip6->ip6_src;
 683         if (in6_setscope(&src0, origifp, &zone))
 684                 goto badscope;
 685         bzero(&src_sa, sizeof(src_sa));
 686         src_sa.sin6_family = AF_INET6;
 687         src_sa.sin6_len = sizeof(src_sa);
 688         src_sa.sin6_addr = ip6->ip6_src;
 689         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 690                 goto badscope;
 691
 692         dst0 = ip6->ip6_dst;
 693         if (in6_setscope(&dst0, origifp, &zone))
 694                 goto badscope;
 695         /* re-initialize to be sure */
 696         bzero(&dst_sa, sizeof(dst_sa));
 697         dst_sa.sin6_family = AF_INET6;
 698         dst_sa.sin6_len = sizeof(dst_sa);
 699         dst_sa.sin6_addr = ip6->ip6_dst;
 700         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 701                 goto badscope;
 702         }
 703
 704         /* scope check is done. */
 705         goto routefound;
 706
 707   badscope:
 708         ip6stat.ip6s_badscope++;
 709         in6_ifstat_inc(origifp, ifs6_out_discard);
 710         if (error == 0)
 711                 error = EHOSTUNREACH; /* XXX */
 712         goto bad;
 713
 714   routefound:
 715         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 716                 if (opt && opt->ip6po_nextroute.ro_rt) {
 717                         /*
 718                          * The nexthop is explicitly specified by the
 719                          * application.  We assume the next hop is an IPv6
 720                          * address.
 721                          */
 722                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 723                 }
 724                 else if ((rt->rt_flags & RTF_GATEWAY))
 725                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 726         }
 727
 728         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 729                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 730         } else {
 731                 struct  in6_multi *in6m;
 732
 733                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 734
 735                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 736
 737                 /*
 738                  * Confirm that the outgoing interface supports multicast.
 739                  */
 740                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 741                         ip6stat.ip6s_noroute++;
 742                         in6_ifstat_inc(ifp, ifs6_out_discard);
 743                         error = ENETUNREACH;
 744                         goto bad;
 745                 }
 746                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 747                 if (in6m != NULL &&
 748                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 749                         /*
 750                          * If we belong to the destination multicast group
 751                          * on the outgoing interface, and the caller did not
 752                          * forbid loopback, loop back a copy.
 753                          */
 754                         ip6_mloopback(ifp, m, dst);
 755                 } else {
 756                         /*
 757                          * If we are acting as a multicast router, perform
 758                          * multicast forwarding as if the packet had just
 759                          * arrived on the interface to which we are about
 760                          * to send.  The multicast forwarding function
 761                          * recursively calls this function, using the
 762                          * IPV6_FORWARDING flag to prevent infinite recursion.
 763                          *
 764                          * Multicasts that are looped back by ip6_mloopback(),
 765                          * above, will be forwarded by the ip6_input() routine,
 766                          * if necessary.
 767                          */
 768                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 769                                 /*
 770                                  * XXX: ip6_mforward expects that rcvif is NULL
 771                                  * when it is called from the originating path.
 772                                  * However, it is not always the case, since
 773                                  * some versions of MGETHDR() does not
 774                                  * initialize the field.
 775                                  */
 776                                 m->m_pkthdr.rcvif = NULL;
 777                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 778                                         m_freem(m);
 779                                         goto done;
 780                                 }
 781                         }
 782                 }
 783                 /*
 784                  * Multicasts with a hoplimit of zero may be looped back,
 785                  * above, but must not be transmitted on a network.
 786                  * Also, multicasts addressed to the loopback interface
 787                  * are not sent -- the above call to ip6_mloopback() will
 788                  * loop back a copy if this host actually belongs to the
 789                  * destination group on the loopback interface.
 790                  */
 791                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 792                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 793                         m_freem(m);
 794                         goto done;
 795                 }
 796         }
 797
 798         /*
 799          * Fill the outgoing inteface to tell the upper layer
 800          * to increment per-interface statistics.
 801          */
 802         if (ifpp)
 803                 *ifpp = ifp;
 804
 805         /* Determine path MTU. */
 806         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 807             &alwaysfrag)) != 0)
 808                 goto bad;
 809
 810         /*
 811          * The caller of this function may specify to use the minimum MTU
 812          * in some cases.
 813          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 814          * setting.  The logic is a bit complicated; by default, unicast
 815          * packets will follow path MTU while multicast packets will be sent at
 816          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 817          * including unicast ones will be sent at the minimum MTU.  Multicast
 818          * packets will always be sent at the minimum MTU unless
 819          * IP6PO_MINMTU_DISABLE is explicitly specified.
 820          * See RFC 3542 for more details.
 821          */
 822         if (mtu > IPV6_MMTU) {
 823                 if ((flags & IPV6_MINMTU))
 824                         mtu = IPV6_MMTU;
 825                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 826                         mtu = IPV6_MMTU;
 827                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 828                          (opt == NULL ||
 829                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 830                         mtu = IPV6_MMTU;
 831                 }
 832         }
 833
 834         /*
 835          * clear embedded scope identifiers if necessary.
 836          * in6_clearscope will touch the addresses only when necessary.
 837          */
 838         in6_clearscope(&ip6->ip6_src);
 839         in6_clearscope(&ip6->ip6_dst);
 840
 841         /*
 842          * Check with the firewall...
 843          */
 844         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 845                 u_short port = 0;
 846                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 847                 /* If ipfw says divert, we have to just drop packet */
 848                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 849                         m_freem(m);
 850                         goto done;
 851                 }
 852                 if (!m) {
 853                         error = EACCES;
 854                         goto done;
 855                 }
 856         }
 857
 858         /*
 859          * If the outgoing packet contains a hop-by-hop options header,
 860          * it must be examined and processed even by the source node.
 861          * (RFC 2460, section 4.)
 862          */
 863         if (exthdrs.ip6e_hbh) {
 864                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 865                 u_int32_t dummy; /* XXX unused */
 866                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 867
 868 #ifdef DIAGNOSTIC
 869                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 870                         panic("ip6e_hbh is not continuous");
 871 #endif
 872                 /*
 873                  *  XXX: if we have to send an ICMPv6 error to the sender,
 874                  *       we need the M_LOOP flag since icmp6_error() expects
 875                  *       the IPv6 and the hop-by-hop options header are
 876                  *       continuous unless the flag is set.
 877                  */
 878                 m->m_flags |= M_LOOP;
 879                 m->m_pkthdr.rcvif = ifp;
 880                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 881                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 882                     &dummy, &plen) < 0) {
 883                         /* m was already freed at this point */
 884                         error = EINVAL;/* better error? */
 885                         goto done;
 886                 }
 887                 m->m_flags &= ~M_LOOP; /* XXX */
 888                 m->m_pkthdr.rcvif = NULL;
 889         }
 890
 891         /* Jump over all PFIL processing if hooks are not active. */
 892         if (!PFIL_HOOKED(&inet6_pfil_hook))
 893                 goto passout;
 894
 895         odst = ip6->ip6_dst;
 896         /* Run through list of hooks for output packets. */
 897         error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 898         if (error != 0 || m == NULL)
 899                 goto done;
 900         ip6 = mtod(m, struct ip6_hdr *);
 901
 902         /* See if destination IP address was changed by packet filter. */
 903         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 904                 m->m_flags |= M_SKIP_FIREWALL;
 905                 /* If destination is now ourself drop to ip6_input(). */
 906                 if (in6_localaddr(&ip6->ip6_dst)) {
 907                         if (m->m_pkthdr.rcvif == NULL)
 908                                 m->m_pkthdr.rcvif = loif;
 909                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 910                                 m->m_pkthdr.csum_flags |=
 911                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 912                                 m->m_pkthdr.csum_data = 0xffff;
 913                         }
 914                         m->m_pkthdr.csum_flags |=
 915                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 916                         error = netisr_queue(NETISR_IPV6, m);
 917                         goto done;
 918                 } else
 919                         goto again;     /* Redo the routing table lookup. */
 920         }
 921
 922         /* XXX: IPFIREWALL_FORWARD */
 923
 924 passout:
 925         /*
 926          * Send the packet to the outgoing interface.
 927          * If necessary, do IPv6 fragmentation before sending.
 928          *
 929          * the logic here is rather complex:
 930          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 931          * 1-a: send as is if tlen <= path mtu
 932          * 1-b: fragment if tlen > path mtu
 933          *
 934          * 2: if user asks us not to fragment (dontfrag == 1)
 935          * 2-a: send as is if tlen <= interface mtu
 936          * 2-b: error if tlen > interface mtu
 937          *
 938          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 939          *      always fragment
 940          *
 941          * 4: if dontfrag == 1 && alwaysfrag == 1
 942          *      error, as we cannot handle this conflicting request
 943          */
 944         tlen = m->m_pkthdr.len;
 945
 946         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 947                 dontfrag = 1;
 948         else
 949                 dontfrag = 0;
 950         if (dontfrag && alwaysfrag) {   /* case 4 */
 951                 /* conflicting request - can't transmit */
 952                 error = EMSGSIZE;
 953                 goto bad;
 954         }
 955         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 956                 /*
 957                  * Even if the DONTFRAG option is specified, we cannot send the
 958                  * packet when the data length is larger than the MTU of the
 959                  * outgoing interface.
 960                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 961                  * well as returning an error code (the latter is not described
 962                  * in the API spec.)
 963                  */
 964                 u_int32_t mtu32;
 965                 struct ip6ctlparam ip6cp;
 966
 967                 mtu32 = (u_int32_t)mtu;
 968                 bzero(&ip6cp, sizeof(ip6cp));
 969                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 970                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 971                     (void *)&ip6cp);
 972
 973                 error = EMSGSIZE;
 974                 goto bad;
 975         }
 976
 977         /*
 978          * transmit packet without fragmentation
 979          */
 980         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 981                 struct in6_ifaddr *ia6;
 982
 983                 ip6 = mtod(m, struct ip6_hdr *);
 984                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 985                 if (ia6) {
 986                         /* Record statistics for this interface address. */
 987                         ia6->ia_ifa.if_opackets++;
 988                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 989                 }
 990 #ifdef IPSEC
 991                 /* clean ipsec history once it goes out of the node */
 992                 ipsec_delaux(m);
 993 #endif
 994                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 995                 goto done;
 996         }
 997
 998         /*
 999          * try to fragment the packet.  case 1-b and 3
1000          */
1001         if (mtu < IPV6_MMTU) {
1002                 /* path MTU cannot be less than IPV6_MMTU */
1003                 error = EMSGSIZE;
1004                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1005                 goto bad;
1006         } else if (ip6->ip6_plen == 0) {
1007                 /* jumbo payload cannot be fragmented */
1008                 error = EMSGSIZE;
1009                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1010                 goto bad;
1011         } else {
1012                 struct mbuf **mnext, *m_frgpart;
1013                 struct ip6_frag *ip6f;
1014                 u_int32_t id = htonl(ip6_randomid());
1015                 u_char nextproto;
1016 #if 0
1017                 struct ip6ctlparam ip6cp;
1018                 u_int32_t mtu32;
1019 #endif
1020                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1021
1022                 /*
1023                  * Too large for the destination or interface;
1024                  * fragment if possible.
1025                  * Must be able to put at least 8 bytes per fragment.
1026                  */
1027                 hlen = unfragpartlen;
1028                 if (mtu > IPV6_MAXPACKET)
1029                         mtu = IPV6_MAXPACKET;
1030
1031 #if 0
1032                 /*
1033                  * It is believed this code is a leftover from the
1034                  * development of the IPV6_RECVPATHMTU sockopt and
1035                  * associated work to implement RFC3542.
1036                  * It's not entirely clear what the intent of the API
1037                  * is at this point, so disable this code for now.
1038                  * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1039                  * will send notifications if the application requests.
1040                  */
1041
1042                 /* Notify a proper path MTU to applications. */
1043                 mtu32 = (u_int32_t)mtu;
1044                 bzero(&ip6cp, sizeof(ip6cp));
1045                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1046                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1047                     (void *)&ip6cp);
1048 #endif
1049
1050                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1051                 if (len < 8) {
1052                         error = EMSGSIZE;
1053                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1054                         goto bad;
1055                 }
1056
1057                 /*
1058                  * Verify that we have any chance at all of being able to queue
1059                  *      the packet or packet fragments
1060                  */
1061                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1062                     < tlen  /* - hlen */)) {
1063                         error = ENOBUFS;
1064                         ip6stat.ip6s_odropped++;
1065                         goto bad;
1066                 }
1067
1068                 mnext = &m->m_nextpkt;
1069
1070                 /*
1071                  * Change the next header field of the last header in the
1072                  * unfragmentable part.
1073                  */
1074                 if (exthdrs.ip6e_rthdr) {
1075                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1076                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1077                 } else if (exthdrs.ip6e_dest1) {
1078                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1079                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1080                 } else if (exthdrs.ip6e_hbh) {
1081                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1082                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1083                 } else {
1084                         nextproto = ip6->ip6_nxt;
1085                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1086                 }
1087
1088                 /*
1089                  * Loop through length of segment after first fragment,
1090                  * make new header and copy data of each part and link onto
1091                  * chain.
1092                  */
1093                 m0 = m;
1094                 for (off = hlen; off < tlen; off += len) {
1095                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1096                         if (!m) {
1097                                 error = ENOBUFS;
1098                                 ip6stat.ip6s_odropped++;
1099                                 goto sendorfree;
1100                         }
1101                         m->m_pkthdr.rcvif = NULL;
1102                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1103                         *mnext = m;
1104                         mnext = &m->m_nextpkt;
1105                         m->m_data += max_linkhdr;
1106                         mhip6 = mtod(m, struct ip6_hdr *);
1107                         *mhip6 = *ip6;
1108                         m->m_len = sizeof(*mhip6);
1109                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1110                         if (error) {
1111                                 ip6stat.ip6s_odropped++;
1112                                 goto sendorfree;
1113                         }
1114                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1115                         if (off + len >= tlen)
1116                                 len = tlen - off;
1117                         else
1118                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1119                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1120                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1121                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1122                                 error = ENOBUFS;
1123                                 ip6stat.ip6s_odropped++;
1124                                 goto sendorfree;
1125                         }
1126                         m_cat(m, m_frgpart);
1127                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1128                         m->m_pkthdr.rcvif = NULL;
1129                         ip6f->ip6f_reserved = 0;
1130                         ip6f->ip6f_ident = id;
1131                         ip6f->ip6f_nxt = nextproto;
1132                         ip6stat.ip6s_ofragments++;
1133                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1134                 }
1135
1136                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1137         }
1138
1139         /*
1140          * Remove leading garbages.
1141          */
1142 sendorfree:
1143         m = m0->m_nextpkt;
1144         m0->m_nextpkt = 0;
1145         m_freem(m0);
1146         for (m0 = m; m; m = m0) {
1147                 m0 = m->m_nextpkt;
1148                 m->m_nextpkt = 0;
1149                 if (error == 0) {
1150                         /* Record statistics for this interface address. */
1151                         if (ia) {
1152                                 ia->ia_ifa.if_opackets++;
1153                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1154                         }
1155 #ifdef IPSEC
1156                         /* clean ipsec history once it goes out of the node */
1157                         ipsec_delaux(m);
1158 #endif
1159                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1160                 } else
1161                         m_freem(m);
1162         }
1163
1164         if (error == 0)
1165                 ip6stat.ip6s_fragmented++;
1166
1167 done:
1168         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1169                 RTFREE(ro->ro_rt);
1170         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1171                 RTFREE(ro_pmtu->ro_rt);
1172         }
1173
1174 #ifdef IPSEC
1175         if (sp != NULL)
1176                 key_freesp(sp);
1177 #endif /* IPSEC */
1178 #ifdef FAST_IPSEC
1179         if (sp != NULL)
1180                 KEY_FREESP(&sp);
1181 #endif /* FAST_IPSEC */
1182
1183         return (error);
1184
1185 freehdrs:
1186         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1187         m_freem(exthdrs.ip6e_dest1);
1188         m_freem(exthdrs.ip6e_rthdr);
1189         m_freem(exthdrs.ip6e_dest2);
1190         /* FALLTHROUGH */
1191 bad:
1192         m_freem(m);
1193         goto done;
1194 }
1195
1196 static int
1197 ip6_copyexthdr(mp, hdr, hlen)
1198         struct mbuf **mp;
1199         caddr_t hdr;
1200         int hlen;
1201 {
1202         struct mbuf *m;
1203
1204         if (hlen > MCLBYTES)
1205                 return (ENOBUFS); /* XXX */
1206
1207         MGET(m, M_DONTWAIT, MT_DATA);
1208         if (!m)
1209                 return (ENOBUFS);
1210
1211         if (hlen > MLEN) {
1212                 MCLGET(m, M_DONTWAIT);
1213                 if ((m->m_flags & M_EXT) == 0) {
1214                         m_free(m);
1215                         return (ENOBUFS);
1216                 }
1217         }
1218         m->m_len = hlen;
1219         if (hdr)
1220                 bcopy(hdr, mtod(m, caddr_t), hlen);
1221
1222         *mp = m;
1223         return (0);
1224 }
1225
1226 /*
1227  * Insert jumbo payload option.
1228  */
1229 static int
1230 ip6_insert_jumboopt(exthdrs, plen)
1231         struct ip6_exthdrs *exthdrs;
1232         u_int32_t plen;
1233 {
1234         struct mbuf *mopt;
1235         u_char *optbuf;
1236         u_int32_t v;
1237
1238 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1239
1240         /*
1241          * If there is no hop-by-hop options header, allocate new one.
1242          * If there is one but it doesn't have enough space to store the
1243          * jumbo payload option, allocate a cluster to store the whole options.
1244          * Otherwise, use it to store the options.
1245          */
1246         if (exthdrs->ip6e_hbh == 0) {
1247                 MGET(mopt, M_DONTWAIT, MT_DATA);
1248                 if (mopt == 0)
1249                         return (ENOBUFS);
1250                 mopt->m_len = JUMBOOPTLEN;
1251                 optbuf = mtod(mopt, u_char *);
1252                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1253                 exthdrs->ip6e_hbh = mopt;
1254         } else {
1255                 struct ip6_hbh *hbh;
1256
1257                 mopt = exthdrs->ip6e_hbh;
1258                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1259                         /*
1260                          * XXX assumption:
1261                          * - exthdrs->ip6e_hbh is not referenced from places
1262                          *   other than exthdrs.
1263                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1264                          */
1265                         int oldoptlen = mopt->m_len;
1266                         struct mbuf *n;
1267
1268                         /*
1269                          * XXX: give up if the whole (new) hbh header does
1270                          * not fit even in an mbuf cluster.
1271                          */
1272                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1273                                 return (ENOBUFS);
1274
1275                         /*
1276                          * As a consequence, we must always prepare a cluster
1277                          * at this point.
1278                          */
1279                         MGET(n, M_DONTWAIT, MT_DATA);
1280                         if (n) {
1281                                 MCLGET(n, M_DONTWAIT);
1282                                 if ((n->m_flags & M_EXT) == 0) {
1283                                         m_freem(n);
1284                                         n = NULL;
1285                                 }
1286                         }
1287                         if (!n)
1288                                 return (ENOBUFS);
1289                         n->m_len = oldoptlen + JUMBOOPTLEN;
1290                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1291                             oldoptlen);
1292                         optbuf = mtod(n, caddr_t) + oldoptlen;
1293                         m_freem(mopt);
1294                         mopt = exthdrs->ip6e_hbh = n;
1295                 } else {
1296                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1297                         mopt->m_len += JUMBOOPTLEN;
1298                 }
1299                 optbuf[0] = IP6OPT_PADN;
1300                 optbuf[1] = 1;
1301
1302                 /*
1303                  * Adjust the header length according to the pad and
1304                  * the jumbo payload option.
1305                  */
1306                 hbh = mtod(mopt, struct ip6_hbh *);
1307                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1308         }
1309
1310         /* fill in the option. */
1311         optbuf[2] = IP6OPT_JUMBO;
1312         optbuf[3] = 4;
1313         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1314         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1315
1316         /* finally, adjust the packet header length */
1317         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1318
1319         return (0);
1320 #undef JUMBOOPTLEN
1321 }
1322
1323 /*
1324  * Insert fragment header and copy unfragmentable header portions.
1325  */
1326 static int
1327 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1328         struct mbuf *m0, *m;
1329         int hlen;
1330         struct ip6_frag **frghdrp;
1331 {
1332         struct mbuf *n, *mlast;
1333
1334         if (hlen > sizeof(struct ip6_hdr)) {
1335                 n = m_copym(m0, sizeof(struct ip6_hdr),
1336                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1337                 if (n == 0)
1338                         return (ENOBUFS);
1339                 m->m_next = n;
1340         } else
1341                 n = m;
1342
1343         /* Search for the last mbuf of unfragmentable part. */
1344         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1345                 ;
1346
1347         if ((mlast->m_flags & M_EXT) == 0 &&
1348             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1349                 /* use the trailing space of the last mbuf for the fragment hdr */
1350                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1351                     mlast->m_len);
1352                 mlast->m_len += sizeof(struct ip6_frag);
1353                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1354         } else {
1355                 /* allocate a new mbuf for the fragment header */
1356                 struct mbuf *mfrg;
1357
1358                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1359                 if (mfrg == 0)
1360                         return (ENOBUFS);
1361                 mfrg->m_len = sizeof(struct ip6_frag);
1362                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1363                 mlast->m_next = mfrg;
1364         }
1365
1366         return (0);
1367 }
1368
1369 static int
1370 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1371         struct route_in6 *ro_pmtu, *ro;
1372         struct ifnet *ifp;
1373         struct in6_addr *dst;
1374         u_long *mtup;
1375         int *alwaysfragp;
1376 {
1377         u_int32_t mtu = 0;
1378         int alwaysfrag = 0;
1379         int error = 0;
1380
1381         if (ro_pmtu != ro) {
1382                 /* The first hop and the final destination may differ. */
1383                 struct sockaddr_in6 *sa6_dst =
1384                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1385                 if (ro_pmtu->ro_rt &&
1386                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1387                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1388                         RTFREE(ro_pmtu->ro_rt);
1389                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1390                 }
1391                 if (ro_pmtu->ro_rt == NULL) {
1392                         bzero(sa6_dst, sizeof(*sa6_dst));
1393                         sa6_dst->sin6_family = AF_INET6;
1394                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1395                         sa6_dst->sin6_addr = *dst;
1396
1397                         rtalloc((struct route *)ro_pmtu);
1398                 }
1399         }
1400         if (ro_pmtu->ro_rt) {
1401                 u_int32_t ifmtu;
1402                 struct in_conninfo inc;
1403
1404                 bzero(&inc, sizeof(inc));
1405                 inc.inc_flags = 1; /* IPv6 */
1406                 inc.inc6_faddr = *dst;
1407
1408                 if (ifp == NULL)
1409                         ifp = ro_pmtu->ro_rt->rt_ifp;
1410                 ifmtu = IN6_LINKMTU(ifp);
1411                 mtu = tcp_hc_getmtu(&inc);
1412                 if (mtu)
1413                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1414                 else
1415                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1416                 if (mtu == 0)
1417                         mtu = ifmtu;
1418                 else if (mtu < IPV6_MMTU) {
1419                         /*
1420                          * RFC2460 section 5, last paragraph:
1421                          * if we record ICMPv6 too big message with
1422                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1423                          * or smaller, with framgent header attached.
1424                          * (fragment header is needed regardless from the
1425                          * packet size, for translators to identify packets)
1426                          */
1427                         alwaysfrag = 1;
1428                         mtu = IPV6_MMTU;
1429                 } else if (mtu > ifmtu) {
1430                         /*
1431                          * The MTU on the route is larger than the MTU on
1432                          * the interface!  This shouldn't happen, unless the
1433                          * MTU of the interface has been changed after the
1434                          * interface was brought up.  Change the MTU in the
1435                          * route to match the interface MTU (as long as the
1436                          * field isn't locked).
1437                          */
1438                         mtu = ifmtu;
1439                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1440                 }
1441         } else if (ifp) {
1442                 mtu = IN6_LINKMTU(ifp);
1443         } else
1444                 error = EHOSTUNREACH; /* XXX */
1445
1446         *mtup = mtu;
1447         if (alwaysfragp)
1448                 *alwaysfragp = alwaysfrag;
1449         return (error);
1450 }
1451
1452 /*
1453  * IP6 socket option processing.
1454  */
1455 int
1456 ip6_ctloutput(so, sopt)
1457         struct socket *so;
1458         struct sockopt *sopt;
1459 {
1460         int privileged, optdatalen, uproto;
1461         void *optdata;
1462         struct inpcb *in6p = sotoinpcb(so);
1463         int error, optval;
1464         int level, op, optname;
1465         int optlen;
1466         struct thread *td;
1467
1468         if (sopt) {
1469                 level = sopt->sopt_level;
1470                 op = sopt->sopt_dir;
1471                 optname = sopt->sopt_name;
1472                 optlen = sopt->sopt_valsize;
1473                 td = sopt->sopt_td;
1474         } else {
1475                 panic("ip6_ctloutput: arg soopt is NULL");
1476         }
1477         error = optval = 0;
1478
1479         privileged = (td == 0 || suser(td)) ? 0 : 1;
1480         uproto = (int)so->so_proto->pr_protocol;
1481
1482         if (level == IPPROTO_IPV6) {
1483                 switch (op) {
1484
1485                 case SOPT_SET:
1486                         switch (optname) {
1487                         case IPV6_2292PKTOPTIONS:
1488 #ifdef IPV6_PKTOPTIONS
1489                         case IPV6_PKTOPTIONS:
1490 #endif
1491                         {
1492                                 struct mbuf *m;
1493
1494                                 error = soopt_getm(sopt, &m); /* XXX */
1495                                 if (error != 0)
1496                                         break;
1497                                 error = soopt_mcopyin(sopt, m); /* XXX */
1498                                 if (error != 0)
1499                                         break;
1500                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1501                                                     m, so, sopt);
1502                                 m_freem(m); /* XXX */
1503                                 break;
1504                         }
1505
1506                         /*
1507                          * Use of some Hop-by-Hop options or some
1508                          * Destination options, might require special
1509                          * privilege.  That is, normal applications
1510                          * (without special privilege) might be forbidden
1511                          * from setting certain options in outgoing packets,
1512                          * and might never see certain options in received
1513                          * packets. [RFC 2292 Section 6]
1514                          * KAME specific note:
1515                          *  KAME prevents non-privileged users from sending or
1516                          *  receiving ANY hbh/dst options in order to avoid
1517                          *  overhead of parsing options in the kernel.
1518                          */
1519                         case IPV6_RECVHOPOPTS:
1520                         case IPV6_RECVDSTOPTS:
1521                         case IPV6_RECVRTHDRDSTOPTS:
1522                                 if (!privileged) {
1523                                         error = EPERM;
1524                                         break;
1525                                 }
1526                                 /* FALLTHROUGH */
1527                         case IPV6_UNICAST_HOPS:
1528                         case IPV6_HOPLIMIT:
1529                         case IPV6_FAITH:
1530
1531                         case IPV6_RECVPKTINFO:
1532                         case IPV6_RECVHOPLIMIT:
1533                         case IPV6_RECVRTHDR:
1534                         case IPV6_RECVPATHMTU:
1535                         case IPV6_RECVTCLASS:
1536                         case IPV6_V6ONLY:
1537                         case IPV6_AUTOFLOWLABEL:
1538                                 if (optlen != sizeof(int)) {
1539                                         error = EINVAL;
1540                                         break;
1541                                 }
1542                                 error = sooptcopyin(sopt, &optval,
1543                                         sizeof optval, sizeof optval);
1544                                 if (error)
1545                                         break;
1546                                 switch (optname) {
1547
1548                                 case IPV6_UNICAST_HOPS:
1549                                         if (optval < -1 || optval >= 256)
1550                                                 error = EINVAL;
1551                                         else {
1552                                                 /* -1 = kernel default */
1553                                                 in6p->in6p_hops = optval;
1554                                                 if ((in6p->in6p_vflag &
1555                                                      INP_IPV4) != 0)
1556                                                         in6p->inp_ip_ttl = optval;
1557                                         }
1558                                         break;
1559 #define OPTSET(bit) \
1560 do { \
1561         if (optval) \
1562                 in6p->in6p_flags |= (bit); \
1563         else \
1564                 in6p->in6p_flags &= ~(bit); \
1565 } while (/*CONSTCOND*/ 0)
1566 #define OPTSET2292(bit) \
1567 do { \
1568         in6p->in6p_flags |= IN6P_RFC2292; \
1569         if (optval) \
1570                 in6p->in6p_flags |= (bit); \
1571         else \
1572                 in6p->in6p_flags &= ~(bit); \
1573 } while (/*CONSTCOND*/ 0)
1574 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1575
1576                                 case IPV6_RECVPKTINFO:
1577                                         /* cannot mix with RFC2292 */
1578                                         if (OPTBIT(IN6P_RFC2292)) {
1579                                                 error = EINVAL;
1580                                                 break;
1581                                         }
1582                                         OPTSET(IN6P_PKTINFO);
1583                                         break;
1584
1585                                 case IPV6_HOPLIMIT:
1586                                 {
1587                                         struct ip6_pktopts **optp;
1588
1589                                         /* cannot mix with RFC2292 */
1590                                         if (OPTBIT(IN6P_RFC2292)) {
1591                                                 error = EINVAL;
1592                                                 break;
1593                                         }
1594                                         optp = &in6p->in6p_outputopts;
1595                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1596                                                            (u_char *)&optval,
1597                                                            sizeof(optval),
1598                                                            optp,
1599                                                            privileged, uproto);
1600                                         break;
1601                                 }
1602
1603                                 case IPV6_RECVHOPLIMIT:
1604                                         /* cannot mix with RFC2292 */
1605                                         if (OPTBIT(IN6P_RFC2292)) {
1606                                                 error = EINVAL;
1607                                                 break;
1608                                         }
1609                                         OPTSET(IN6P_HOPLIMIT);
1610                                         break;
1611
1612                                 case IPV6_RECVHOPOPTS:
1613                                         /* cannot mix with RFC2292 */
1614                                         if (OPTBIT(IN6P_RFC2292)) {
1615                                                 error = EINVAL;
1616                                                 break;
1617                                         }
1618                                         OPTSET(IN6P_HOPOPTS);
1619                                         break;
1620
1621                                 case IPV6_RECVDSTOPTS:
1622                                         /* cannot mix with RFC2292 */
1623                                         if (OPTBIT(IN6P_RFC2292)) {
1624                                                 error = EINVAL;
1625                                                 break;
1626                                         }
1627                                         OPTSET(IN6P_DSTOPTS);
1628                                         break;
1629
1630                                 case IPV6_RECVRTHDRDSTOPTS:
1631                                         /* cannot mix with RFC2292 */
1632                                         if (OPTBIT(IN6P_RFC2292)) {
1633                                                 error = EINVAL;
1634                                                 break;
1635                                         }
1636                                         OPTSET(IN6P_RTHDRDSTOPTS);
1637                                         break;
1638
1639                                 case IPV6_RECVRTHDR:
1640                                         /* cannot mix with RFC2292 */
1641                                         if (OPTBIT(IN6P_RFC2292)) {
1642                                                 error = EINVAL;
1643                                                 break;
1644                                         }
1645                                         OPTSET(IN6P_RTHDR);
1646                                         break;
1647
1648                                 case IPV6_FAITH:
1649                                         OPTSET(IN6P_FAITH);
1650                                         break;
1651
1652                                 case IPV6_RECVPATHMTU:
1653                                         /*
1654                                          * We ignore this option for TCP
1655                                          * sockets.
1656                                          * (RFC3542 leaves this case
1657                                          * unspecified.)
1658                                          */
1659                                         if (uproto != IPPROTO_TCP)
1660                                                 OPTSET(IN6P_MTU);
1661                                         break;
1662
1663                                 case IPV6_V6ONLY:
1664                                         /*
1665                                          * make setsockopt(IPV6_V6ONLY)
1666                                          * available only prior to bind(2).
1667                                          * see ipng mailing list, Jun 22 2001.
1668                                          */
1669                                         if (in6p->in6p_lport ||
1670                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1671                                                 error = EINVAL;
1672                                                 break;
1673                                         }
1674                                         OPTSET(IN6P_IPV6_V6ONLY);
1675                                         if (optval)
1676                                                 in6p->in6p_vflag &= ~INP_IPV4;
1677                                         else
1678                                                 in6p->in6p_vflag |= INP_IPV4;
1679                                         break;
1680                                 case IPV6_RECVTCLASS:
1681                                         /* cannot mix with RFC2292 XXX */
1682                                         if (OPTBIT(IN6P_RFC2292)) {
1683                                                 error = EINVAL;
1684                                                 break;
1685                                         }
1686                                         OPTSET(IN6P_TCLASS);
1687                                         break;
1688                                 case IPV6_AUTOFLOWLABEL:
1689                                         OPTSET(IN6P_AUTOFLOWLABEL);
1690                                         break;
1691
1692                                 }
1693                                 break;
1694
1695                         case IPV6_TCLASS:
1696                         case IPV6_DONTFRAG:
1697                         case IPV6_USE_MIN_MTU:
1698                         case IPV6_PREFER_TEMPADDR:
1699                                 if (optlen != sizeof(optval)) {
1700                                         error = EINVAL;
1701                                         break;
1702                                 }
1703                                 error = sooptcopyin(sopt, &optval,
1704                                         sizeof optval, sizeof optval);
1705                                 if (error)
1706                                         break;
1707                                 {
1708                                         struct ip6_pktopts **optp;
1709                                         optp = &in6p->in6p_outputopts;
1710                                         error = ip6_pcbopt(optname,
1711                                                            (u_char *)&optval,
1712                                                            sizeof(optval),
1713                                                            optp,
1714                                                            privileged, uproto);
1715                                         break;
1716                                 }
1717
1718                         case IPV6_2292PKTINFO:
1719                         case IPV6_2292HOPLIMIT:
1720                         case IPV6_2292HOPOPTS:
1721                         case IPV6_2292DSTOPTS:
1722                         case IPV6_2292RTHDR:
1723                                 /* RFC 2292 */
1724                                 if (optlen != sizeof(int)) {
1725                                         error = EINVAL;
1726                                         break;
1727                                 }
1728                                 error = sooptcopyin(sopt, &optval,
1729                                         sizeof optval, sizeof optval);
1730                                 if (error)
1731                                         break;
1732                                 switch (optname) {
1733                                 case IPV6_2292PKTINFO:
1734                                         OPTSET2292(IN6P_PKTINFO);
1735                                         break;
1736                                 case IPV6_2292HOPLIMIT:
1737                                         OPTSET2292(IN6P_HOPLIMIT);
1738                                         break;
1739                                 case IPV6_2292HOPOPTS:
1740                                         /*
1741                                          * Check super-user privilege.
1742                                          * See comments for IPV6_RECVHOPOPTS.
1743                                          */
1744                                         if (!privileged)
1745                                                 return (EPERM);
1746                                         OPTSET2292(IN6P_HOPOPTS);
1747                                         break;
1748                                 case IPV6_2292DSTOPTS:
1749                                         if (!privileged)
1750                                                 return (EPERM);
1751                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1752                                         break;
1753                                 case IPV6_2292RTHDR:
1754                                         OPTSET2292(IN6P_RTHDR);
1755                                         break;
1756                                 }
1757                                 break;
1758                         case IPV6_PKTINFO:
1759                         case IPV6_HOPOPTS:
1760                         case IPV6_RTHDR:
1761                         case IPV6_DSTOPTS:
1762                         case IPV6_RTHDRDSTOPTS:
1763                         case IPV6_NEXTHOP:
1764                         {
1765                                 /* new advanced API (RFC3542) */
1766                                 u_char *optbuf;
1767                                 u_char optbuf_storage[MCLBYTES];
1768                                 int optlen;
1769                                 struct ip6_pktopts **optp;
1770
1771                                 /* cannot mix with RFC2292 */
1772                                 if (OPTBIT(IN6P_RFC2292)) {
1773                                         error = EINVAL;
1774                                         break;
1775                                 }
1776
1777                                 /*
1778                                  * We only ensure valsize is not too large
1779                                  * here.  Further validation will be done
1780                                  * later.
1781                                  */
1782                                 error = sooptcopyin(sopt, optbuf_storage,
1783                                     sizeof(optbuf_storage), 0);
1784                                 if (error)
1785                                         break;
1786                                 optlen = sopt->sopt_valsize;
1787                                 optbuf = optbuf_storage;
1788                                 optp = &in6p->in6p_outputopts;
1789                                 error = ip6_pcbopt(optname,
1790                                                    optbuf, optlen,
1791                                                    optp, privileged, uproto);
1792                                 break;
1793                         }
1794 #undef OPTSET
1795
1796                         case IPV6_MULTICAST_IF:
1797                         case IPV6_MULTICAST_HOPS:
1798                         case IPV6_MULTICAST_LOOP:
1799                         case IPV6_JOIN_GROUP:
1800                         case IPV6_LEAVE_GROUP:
1801                             {
1802                                 if (sopt->sopt_valsize > MLEN) {
1803                                         error = EMSGSIZE;
1804                                         break;
1805                                 }
1806                                 /* XXX */
1807                             }
1808                             /* FALLTHROUGH */
1809                             {
1810                                 struct mbuf *m;
1811
1812                                 if (sopt->sopt_valsize > MCLBYTES) {
1813                                         error = EMSGSIZE;
1814                                         break;
1815                                 }
1816                                 /* XXX */
1817                                 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1818                                 if (m == 0) {
1819                                         error = ENOBUFS;
1820                                         break;
1821                                 }
1822                                 if (sopt->sopt_valsize > MLEN) {
1823                                         MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1824                                         if ((m->m_flags & M_EXT) == 0) {
1825                                                 m_free(m);
1826                                                 error = ENOBUFS;
1827                                                 break;
1828                                         }
1829                                 }
1830                                 m->m_len = sopt->sopt_valsize;
1831                                 error = sooptcopyin(sopt, mtod(m, char *),
1832                                                     m->m_len, m->m_len);
1833                                 if (error) {
1834                                         (void)m_free(m);
1835                                         break;
1836                                 }
1837                                 error = ip6_setmoptions(sopt->sopt_name,
1838                                                         &in6p->in6p_moptions,
1839                                                         m);
1840                                 (void)m_free(m);
1841                             }
1842                                 break;
1843
1844                         case IPV6_PORTRANGE:
1845                                 error = sooptcopyin(sopt, &optval,
1846                                     sizeof optval, sizeof optval);
1847                                 if (error)
1848                                         break;
1849
1850                                 switch (optval) {
1851                                 case IPV6_PORTRANGE_DEFAULT:
1852                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1853                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1854                                         break;
1855
1856                                 case IPV6_PORTRANGE_HIGH:
1857                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1858                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1859                                         break;
1860
1861                                 case IPV6_PORTRANGE_LOW:
1862                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1863                                         in6p->in6p_flags |= IN6P_LOWPORT;
1864                                         break;
1865
1866                                 default:
1867                                         error = EINVAL;
1868                                         break;
1869                                 }
1870                                 break;
1871
1872 #if defined(IPSEC) || defined(FAST_IPSEC)
1873                         case IPV6_IPSEC_POLICY:
1874                             {
1875                                 caddr_t req = NULL;
1876                                 size_t len = 0;
1877                                 struct mbuf *m;
1878
1879                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1880                                         break;
1881                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1882                                         break;
1883                                 if (m) {
1884                                         req = mtod(m, caddr_t);
1885                                         len = m->m_len;
1886                                 }
1887                                 error = ipsec6_set_policy(in6p, optname, req,
1888                                                           len, privileged);
1889                                 m_freem(m);
1890                             }
1891                                 break;
1892 #endif /* KAME IPSEC */
1893
1894                         case IPV6_FW_ADD:
1895                         case IPV6_FW_DEL:
1896                         case IPV6_FW_FLUSH:
1897                         case IPV6_FW_ZERO:
1898                             {
1899                                 struct mbuf *m;
1900                                 struct mbuf **mp = &m;
1901
1902                                 if (ip6_fw_ctl_ptr == NULL)
1903                                         return EINVAL;
1904                                 /* XXX */
1905                                 if ((error = soopt_getm(sopt, &m)) != 0)
1906                                         break;
1907                                 /* XXX */
1908                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
1909                                         break;
1910                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1911                                 m = *mp;
1912                             }
1913                                 break;
1914
1915                         default:
1916                                 error = ENOPROTOOPT;
1917                                 break;
1918                         }
1919                         break;
1920
1921                 case SOPT_GET:
1922                         switch (optname) {
1923
1924                         case IPV6_2292PKTOPTIONS:
1925 #ifdef IPV6_PKTOPTIONS
1926                         case IPV6_PKTOPTIONS:
1927 #endif
1928                                 /*
1929                                  * RFC3542 (effectively) deprecated the
1930                                  * semantics of the 2292-style pktoptions.
1931                                  * Since it was not reliable in nature (i.e.,
1932                                  * applications had to expect the lack of some
1933                                  * information after all), it would make sense
1934                                  * to simplify this part by always returning
1935                                  * empty data.
1936                                  */
1937                                 sopt->sopt_valsize = 0;
1938                                 break;
1939
1940                         case IPV6_RECVHOPOPTS:
1941                         case IPV6_RECVDSTOPTS:
1942                         case IPV6_RECVRTHDRDSTOPTS:
1943                         case IPV6_UNICAST_HOPS:
1944                         case IPV6_RECVPKTINFO:
1945                         case IPV6_RECVHOPLIMIT:
1946                         case IPV6_RECVRTHDR:
1947                         case IPV6_RECVPATHMTU:
1948
1949                         case IPV6_FAITH:
1950                         case IPV6_V6ONLY:
1951                         case IPV6_PORTRANGE:
1952                         case IPV6_RECVTCLASS:
1953                         case IPV6_AUTOFLOWLABEL:
1954                                 switch (optname) {
1955
1956                                 case IPV6_RECVHOPOPTS:
1957                                         optval = OPTBIT(IN6P_HOPOPTS);
1958                                         break;
1959
1960                                 case IPV6_RECVDSTOPTS:
1961                                         optval = OPTBIT(IN6P_DSTOPTS);
1962                                         break;
1963
1964                                 case IPV6_RECVRTHDRDSTOPTS:
1965                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1966                                         break;
1967
1968                                 case IPV6_UNICAST_HOPS:
1969                                         optval = in6p->in6p_hops;
1970                                         break;
1971
1972                                 case IPV6_RECVPKTINFO:
1973                                         optval = OPTBIT(IN6P_PKTINFO);
1974                                         break;
1975
1976                                 case IPV6_RECVHOPLIMIT:
1977                                         optval = OPTBIT(IN6P_HOPLIMIT);
1978                                         break;
1979
1980                                 case IPV6_RECVRTHDR:
1981                                         optval = OPTBIT(IN6P_RTHDR);
1982                                         break;
1983
1984                                 case IPV6_RECVPATHMTU:
1985                                         optval = OPTBIT(IN6P_MTU);
1986                                         break;
1987
1988                                 case IPV6_FAITH:
1989                                         optval = OPTBIT(IN6P_FAITH);
1990                                         break;
1991
1992                                 case IPV6_V6ONLY:
1993                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1994                                         break;
1995
1996                                 case IPV6_PORTRANGE:
1997                                     {
1998                                         int flags;
1999                                         flags = in6p->in6p_flags;
2000                                         if (flags & IN6P_HIGHPORT)
2001                                                 optval = IPV6_PORTRANGE_HIGH;
2002                                         else if (flags & IN6P_LOWPORT)
2003                                                 optval = IPV6_PORTRANGE_LOW;
2004                                         else
2005                                                 optval = 0;
2006                                         break;
2007                                     }
2008                                 case IPV6_RECVTCLASS:
2009                                         optval = OPTBIT(IN6P_TCLASS);
2010                                         break;
2011
2012                                 case IPV6_AUTOFLOWLABEL:
2013                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2014                                         break;
2015                                 }
2016                                 if (error)
2017                                         break;
2018                                 error = sooptcopyout(sopt, &optval,
2019                                         sizeof optval);
2020                                 break;
2021
2022                         case IPV6_PATHMTU:
2023                         {
2024                                 u_long pmtu = 0;
2025                                 struct ip6_mtuinfo mtuinfo;
2026                                 struct route_in6 sro;
2027
2028                                 bzero(&sro, sizeof(sro));
2029
2030                                 if (!(so->so_state & SS_ISCONNECTED))
2031                                         return (ENOTCONN);
2032                                 /*
2033                                  * XXX: we dot not consider the case of source
2034                                  * routing, or optional information to specify
2035                                  * the outgoing interface.
2036                                  */
2037                                 error = ip6_getpmtu(&sro, NULL, NULL,
2038                                     &in6p->in6p_faddr, &pmtu, NULL);
2039                                 if (sro.ro_rt)
2040                                         RTFREE(sro.ro_rt);
2041                                 if (error)
2042                                         break;
2043                                 if (pmtu > IPV6_MAXPACKET)
2044                                         pmtu = IPV6_MAXPACKET;
2045
2046                                 bzero(&mtuinfo, sizeof(mtuinfo));
2047                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2048                                 optdata = (void *)&mtuinfo;
2049                                 optdatalen = sizeof(mtuinfo);
2050                                 error = sooptcopyout(sopt, optdata,
2051                                     optdatalen);
2052                                 break;
2053                         }
2054
2055                         case IPV6_2292PKTINFO:
2056                         case IPV6_2292HOPLIMIT:
2057                         case IPV6_2292HOPOPTS:
2058                         case IPV6_2292RTHDR:
2059                         case IPV6_2292DSTOPTS:
2060                                 switch (optname) {
2061                                 case IPV6_2292PKTINFO:
2062                                         optval = OPTBIT(IN6P_PKTINFO);
2063                                         break;
2064                                 case IPV6_2292HOPLIMIT:
2065                                         optval = OPTBIT(IN6P_HOPLIMIT);
2066                                         break;
2067                                 case IPV6_2292HOPOPTS:
2068                                         optval = OPTBIT(IN6P_HOPOPTS);
2069                                         break;
2070                                 case IPV6_2292RTHDR:
2071                                         optval = OPTBIT(IN6P_RTHDR);
2072                                         break;
2073                                 case IPV6_2292DSTOPTS:
2074                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2075                                         break;
2076                                 }
2077                                 error = sooptcopyout(sopt, &optval,
2078                                     sizeof optval);
2079                                 break;
2080                         case IPV6_PKTINFO:
2081                         case IPV6_HOPOPTS:
2082                         case IPV6_RTHDR:
2083                         case IPV6_DSTOPTS:
2084                         case IPV6_RTHDRDSTOPTS:
2085                         case IPV6_NEXTHOP:
2086                         case IPV6_TCLASS:
2087                         case IPV6_DONTFRAG:
2088                         case IPV6_USE_MIN_MTU:
2089                         case IPV6_PREFER_TEMPADDR:
2090                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2091                                     optname, sopt);
2092                                 break;
2093
2094                         case IPV6_MULTICAST_IF:
2095                         case IPV6_MULTICAST_HOPS:
2096                         case IPV6_MULTICAST_LOOP:
2097                         case IPV6_JOIN_GROUP:
2098                         case IPV6_LEAVE_GROUP:
2099                             {
2100                                 struct mbuf *m;
2101                                 error = ip6_getmoptions(sopt->sopt_name,
2102                                     in6p->in6p_moptions, &m);
2103                                 if (error == 0)
2104                                         error = sooptcopyout(sopt,
2105                                             mtod(m, char *), m->m_len);
2106                                 m_freem(m);
2107                             }
2108                                 break;
2109
2110 #if defined(IPSEC) || defined(FAST_IPSEC)
2111                         case IPV6_IPSEC_POLICY:
2112                           {
2113                                 caddr_t req = NULL;
2114                                 size_t len = 0;
2115                                 struct mbuf *m = NULL;
2116                                 struct mbuf **mp = &m;
2117                                 size_t ovalsize = sopt->sopt_valsize;
2118                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2119
2120                                 error = soopt_getm(sopt, &m); /* XXX */
2121                                 if (error != 0)
2122                                         break;
2123                                 error = soopt_mcopyin(sopt, m); /* XXX */
2124                                 if (error != 0)
2125                                         break;
2126                                 sopt->sopt_valsize = ovalsize;
2127                                 sopt->sopt_val = oval;
2128                                 if (m) {
2129                                         req = mtod(m, caddr_t);
2130                                         len = m->m_len;
2131                                 }
2132                                 error = ipsec6_get_policy(in6p, req, len, mp);
2133                                 if (error == 0)
2134                                         error = soopt_mcopyout(sopt, m); /* XXX */
2135                                 if (error == 0 && m)
2136                                         m_freem(m);
2137                                 break;
2138                           }
2139 #endif /* KAME IPSEC */
2140
2141                         case IPV6_FW_GET:
2142                           {
2143                                 struct mbuf *m;
2144                                 struct mbuf **mp = &m;
2145
2146                                 if (ip6_fw_ctl_ptr == NULL)
2147                                 {
2148                                         return EINVAL;
2149                                 }
2150                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2151                                 if (error == 0)
2152                                         error = soopt_mcopyout(sopt, m); /* XXX */
2153                                 if (error == 0 && m)
2154                                         m_freem(m);
2155                           }
2156                                 break;
2157
2158                         default:
2159                                 error = ENOPROTOOPT;
2160                                 break;
2161                         }
2162                         break;
2163                 }
2164         } else {                /* level != IPPROTO_IPV6 */
2165                 error = EINVAL;
2166         }
2167         return (error);
2168 }
2169
2170 int
2171 ip6_raw_ctloutput(so, sopt)
2172         struct socket *so;
2173         struct sockopt *sopt;
2174 {
2175         int error = 0, optval, optlen;
2176         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2177         struct in6pcb *in6p = sotoin6pcb(so);
2178         int level, op, optname;
2179
2180         if (sopt) {
2181                 level = sopt->sopt_level;
2182                 op = sopt->sopt_dir;
2183                 optname = sopt->sopt_name;
2184                 optlen = sopt->sopt_valsize;
2185         } else
2186                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2187
2188         if (level != IPPROTO_IPV6) {
2189                 return (EINVAL);
2190         }
2191
2192         switch (optname) {
2193         case IPV6_CHECKSUM:
2194                 /*
2195                  * For ICMPv6 sockets, no modification allowed for checksum
2196                  * offset, permit "no change" values to help existing apps.
2197                  *
2198                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2199                  * for an ICMPv6 socket will fail."
2200                  * The current behavior does not meet RFC3542.
2201                  */
2202                 switch (op) {
2203                 case SOPT_SET:
2204                         if (optlen != sizeof(int)) {
2205                                 error = EINVAL;
2206                                 break;
2207                         }
2208                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2209                                             sizeof(optval));
2210                         if (error)
2211                                 break;
2212                         if ((optval % 2) != 0) {
2213                                 /* the API assumes even offset values */
2214                                 error = EINVAL;
2215                         } else if (so->so_proto->pr_protocol ==
2216                             IPPROTO_ICMPV6) {
2217                                 if (optval != icmp6off)
2218                                         error = EINVAL;
2219                         } else
2220                                 in6p->in6p_cksum = optval;
2221                         break;
2222
2223                 case SOPT_GET:
2224                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2225                                 optval = icmp6off;
2226                         else
2227                                 optval = in6p->in6p_cksum;
2228
2229                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2230                         break;
2231
2232                 default:
2233                         error = EINVAL;
2234                         break;
2235                 }
2236                 break;
2237
2238         default:
2239                 error = ENOPROTOOPT;
2240                 break;
2241         }
2242
2243         return (error);
2244 }
2245
2246 /*
2247  * Set up IP6 options in pcb for insertion in output packets or
2248  * specifying behavior of outgoing packets.
2249  */
2250 static int
2251 ip6_pcbopts(pktopt, m, so, sopt)
2252         struct ip6_pktopts **pktopt;
2253         struct mbuf *m;
2254         struct socket *so;
2255         struct sockopt *sopt;
2256 {
2257         struct ip6_pktopts *opt = *pktopt;
2258         int error = 0;
2259         struct thread *td = sopt->sopt_td;
2260         int priv = 0;
2261
2262         /* turn off any old options. */
2263         if (opt) {
2264 #ifdef DIAGNOSTIC
2265                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2266                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2267                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2268                         printf("ip6_pcbopts: all specified options are cleared.\n");
2269 #endif
2270                 ip6_clearpktopts(opt, -1);
2271         } else
2272                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2273         *pktopt = NULL;
2274
2275         if (!m || m->m_len == 0) {
2276                 /*
2277                  * Only turning off any previous options, regardless of
2278                  * whether the opt is just created or given.
2279                  */
2280                 free(opt, M_IP6OPT);
2281                 return (0);
2282         }
2283
2284         /*  set options specified by user. */
2285         if (td && !suser(td))
2286                 priv = 1;
2287         if ((error = ip6_setpktopts(m, opt, NULL, priv,
2288             so->so_proto->pr_protocol)) != 0) {
2289                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2290                 free(opt, M_IP6OPT);
2291                 return (error);
2292         }
2293         *pktopt = opt;
2294         return (0);
2295 }
2296
2297 /*
2298  * initialize ip6_pktopts.  beware that there are non-zero default values in
2299  * the struct.
2300  */
2301 void
2302 ip6_initpktopts(opt)
2303         struct ip6_pktopts *opt;
2304 {
2305
2306         bzero(opt, sizeof(*opt));
2307         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2308         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2309         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2310         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2311 }
2312
2313 static int
2314 ip6_pcbopt(optname, buf, len, pktopt, priv, uproto)
2315         int optname, len, priv;
2316         u_char *buf;
2317         struct ip6_pktopts **pktopt;
2318         int uproto;
2319 {
2320         struct ip6_pktopts *opt;
2321
2322         if (*pktopt == NULL) {
2323                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2324                     M_WAITOK);
2325                 ip6_initpktopts(*pktopt);
2326         }
2327         opt = *pktopt;
2328
2329         return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2330 }
2331
2332 static int
2333 ip6_getpcbopt(pktopt, optname, sopt)
2334         struct ip6_pktopts *pktopt;
2335         struct sockopt *sopt;
2336         int optname;
2337 {
2338         void *optdata = NULL;
2339         int optdatalen = 0;
2340         struct ip6_ext *ip6e;
2341         int error = 0;
2342         struct in6_pktinfo null_pktinfo;
2343         int deftclass = 0, on;
2344         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2345         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2346
2347         switch (optname) {
2348         case IPV6_PKTINFO:
2349                 if (pktopt && pktopt->ip6po_pktinfo)
2350                         optdata = (void *)pktopt->ip6po_pktinfo;
2351                 else {
2352                         /* XXX: we don't have to do this every time... */
2353                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2354                         optdata = (void *)&null_pktinfo;
2355                 }
2356                 optdatalen = sizeof(struct in6_pktinfo);
2357                 break;
2358         case IPV6_TCLASS:
2359                 if (pktopt && pktopt->ip6po_tclass >= 0)
2360                         optdata = (void *)&pktopt->ip6po_tclass;
2361                 else
2362                         optdata = (void *)&deftclass;
2363                 optdatalen = sizeof(int);
2364                 break;
2365         case IPV6_HOPOPTS:
2366                 if (pktopt && pktopt->ip6po_hbh) {
2367                         optdata = (void *)pktopt->ip6po_hbh;
2368                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2369                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2370                 }
2371                 break;
2372         case IPV6_RTHDR:
2373                 if (pktopt && pktopt->ip6po_rthdr) {
2374                         optdata = (void *)pktopt->ip6po_rthdr;
2375                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2376                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2377                 }
2378                 break;
2379         case IPV6_RTHDRDSTOPTS:
2380                 if (pktopt && pktopt->ip6po_dest1) {
2381                         optdata = (void *)pktopt->ip6po_dest1;
2382                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2383                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2384                 }
2385                 break;
2386         case IPV6_DSTOPTS:
2387                 if (pktopt && pktopt->ip6po_dest2) {
2388                         optdata = (void *)pktopt->ip6po_dest2;
2389                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2390                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2391                 }
2392                 break;
2393         case IPV6_NEXTHOP:
2394                 if (pktopt && pktopt->ip6po_nexthop) {
2395                         optdata = (void *)pktopt->ip6po_nexthop;
2396                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2397                 }
2398                 break;
2399         case IPV6_USE_MIN_MTU:
2400                 if (pktopt)
2401                         optdata = (void *)&pktopt->ip6po_minmtu;
2402                 else
2403                         optdata = (void *)&defminmtu;
2404                 optdatalen = sizeof(int);
2405                 break;
2406         case IPV6_DONTFRAG:
2407                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2408                         on = 1;
2409                 else
2410                         on = 0;
2411                 optdata = (void *)&on;
2412                 optdatalen = sizeof(on);
2413                 break;
2414         case IPV6_PREFER_TEMPADDR:
2415                 if (pktopt)
2416                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2417                 else
2418                         optdata = (void *)&defpreftemp;
2419                 optdatalen = sizeof(int);
2420                 break;
2421         default:                /* should not happen */
2422 #ifdef DIAGNOSTIC
2423                 panic("ip6_getpcbopt: unexpected option\n");
2424 #endif
2425                 return (ENOPROTOOPT);
2426         }
2427
2428         error = sooptcopyout(sopt, optdata, optdatalen);
2429
2430         return (error);
2431 }
2432
2433 void
2434 ip6_clearpktopts(pktopt, optname)
2435         struct ip6_pktopts *pktopt;
2436         int optname;
2437 {
2438         if (pktopt == NULL)
2439                 return;
2440
2441         if (optname == -1 || optname == IPV6_PKTINFO) {
2442                 if (pktopt->ip6po_pktinfo)
2443                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2444                 pktopt->ip6po_pktinfo = NULL;
2445         }
2446         if (optname == -1 || optname == IPV6_HOPLIMIT)
2447                 pktopt->ip6po_hlim = -1;
2448         if (optname == -1 || optname == IPV6_TCLASS)
2449                 pktopt->ip6po_tclass = -1;
2450         if (optname == -1 || optname == IPV6_NEXTHOP) {
2451                 if (pktopt->ip6po_nextroute.ro_rt) {
2452                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2453                         pktopt->ip6po_nextroute.ro_rt = NULL;
2454                 }
2455                 if (pktopt->ip6po_nexthop)
2456                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2457                 pktopt->ip6po_nexthop = NULL;
2458         }
2459         if (optname == -1 || optname == IPV6_HOPOPTS) {
2460                 if (pktopt->ip6po_hbh)
2461                         free(pktopt->ip6po_hbh, M_IP6OPT);
2462                 pktopt->ip6po_hbh = NULL;
2463         }
2464         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2465                 if (pktopt->ip6po_dest1)
2466                         free(pktopt->ip6po_dest1, M_IP6OPT);
2467                 pktopt->ip6po_dest1 = NULL;
2468         }
2469         if (optname == -1 || optname == IPV6_RTHDR) {
2470                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2471                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2472                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2473                 if (pktopt->ip6po_route.ro_rt) {
2474                         RTFREE(pktopt->ip6po_route.ro_rt);
2475                         pktopt->ip6po_route.ro_rt = NULL;
2476                 }
2477         }
2478         if (optname == -1 || optname == IPV6_DSTOPTS) {
2479                 if (pktopt->ip6po_dest2)
2480                         free(pktopt->ip6po_dest2, M_IP6OPT);
2481                 pktopt->ip6po_dest2 = NULL;
2482         }
2483 }
2484
2485 #define PKTOPT_EXTHDRCPY(type) \
2486 do {\
2487         if (src->type) {\
2488                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2489                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2490                 if (dst->type == NULL && canwait == M_NOWAIT)\
2491                         goto bad;\
2492                 bcopy(src->type, dst->type, hlen);\
2493         }\
2494 } while (/*CONSTCOND*/ 0)
2495
2496 static int
2497 copypktopts(dst, src, canwait)
2498         struct ip6_pktopts *dst, *src;
2499         int canwait;
2500 {
2501         if (dst == NULL || src == NULL)  {
2502                 printf("ip6_clearpktopts: invalid argument\n");
2503                 return (EINVAL);
2504         }
2505
2506         dst->ip6po_hlim = src->ip6po_hlim;
2507         dst->ip6po_tclass = src->ip6po_tclass;
2508         dst->ip6po_flags = src->ip6po_flags;
2509         if (src->ip6po_pktinfo) {
2510                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2511                     M_IP6OPT, canwait);
2512                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2513                         goto bad;
2514                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2515         }
2516         if (src->ip6po_nexthop) {
2517                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2518                     M_IP6OPT, canwait);
2519                 if (dst->ip6po_nexthop == NULL)
2520                         goto bad;
2521                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2522                     src->ip6po_nexthop->sa_len);
2523         }
2524         PKTOPT_EXTHDRCPY(ip6po_hbh);
2525         PKTOPT_EXTHDRCPY(ip6po_dest1);
2526         PKTOPT_EXTHDRCPY(ip6po_dest2);
2527         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2528         return (0);
2529
2530   bad:
2531         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2532         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2533         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2534         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2535         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2536         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2537         return (ENOBUFS);
2538 }
2539 #undef PKTOPT_EXTHDRCPY
2540
2541 struct ip6_pktopts *
2542 ip6_copypktopts(src, canwait)
2543         struct ip6_pktopts *src;
2544         int canwait;
2545 {
2546         int error;
2547         struct ip6_pktopts *dst;
2548
2549         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2550         if (dst == NULL && canwait == M_NOWAIT)
2551                 return (NULL);
2552         ip6_initpktopts(dst);
2553
2554         if ((error = copypktopts(dst, src, canwait)) != 0) {
2555                 free(dst, M_IP6OPT);
2556                 return (NULL);
2557         }
2558
2559         return (dst);
2560 }
2561
2562 void
2563 ip6_freepcbopts(pktopt)
2564         struct ip6_pktopts *pktopt;
2565 {
2566         if (pktopt == NULL)
2567                 return;
2568
2569         ip6_clearpktopts(pktopt, -1);
2570
2571         free(pktopt, M_IP6OPT);
2572 }
2573
2574 /*
2575  * Set the IP6 multicast options in response to user setsockopt().
2576  */
2577 static int
2578 ip6_setmoptions(optname, im6op, m)
2579         int optname;
2580         struct ip6_moptions **im6op;
2581         struct mbuf *m;
2582 {
2583         int error = 0;
2584         u_int loop, ifindex;
2585         struct ipv6_mreq *mreq;
2586         struct ifnet *ifp;
2587         struct ip6_moptions *im6o = *im6op;
2588         struct route_in6 ro;
2589         struct in6_multi_mship *imm;
2590         struct thread *td = curthread;
2591
2592         if (im6o == NULL) {
2593                 /*
2594                  * No multicast option buffer attached to the pcb;
2595                  * allocate one and initialize to default values.
2596                  */
2597                 im6o = (struct ip6_moptions *)
2598                         malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2599
2600                 if (im6o == NULL)
2601                         return (ENOBUFS);
2602                 *im6op = im6o;
2603                 im6o->im6o_multicast_ifp = NULL;
2604                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2605                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2606                 LIST_INIT(&im6o->im6o_memberships);
2607         }
2608
2609         switch (optname) {
2610
2611         case IPV6_MULTICAST_IF:
2612                 /*
2613                  * Select the interface for outgoing multicast packets.
2614                  */
2615                 if (m == NULL || m->m_len != sizeof(u_int)) {
2616                         error = EINVAL;
2617                         break;
2618                 }
2619                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2620                 if (ifindex < 0 || if_index < ifindex) {
2621                         error = ENXIO;  /* XXX EINVAL? */
2622                         break;
2623                 }
2624                 ifp = ifnet_byindex(ifindex);
2625                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2626                         error = EADDRNOTAVAIL;
2627                         break;
2628                 }
2629                 im6o->im6o_multicast_ifp = ifp;
2630                 break;
2631
2632         case IPV6_MULTICAST_HOPS:
2633             {
2634                 /*
2635                  * Set the IP6 hoplimit for outgoing multicast packets.
2636                  */
2637                 int optval;
2638                 if (m == NULL || m->m_len != sizeof(int)) {
2639                         error = EINVAL;
2640                         break;
2641                 }
2642                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2643                 if (optval < -1 || optval >= 256)
2644                         error = EINVAL;
2645                 else if (optval == -1)
2646                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2647                 else
2648                         im6o->im6o_multicast_hlim = optval;
2649                 break;
2650             }
2651
2652         case IPV6_MULTICAST_LOOP:
2653                 /*
2654                  * Set the loopback flag for outgoing multicast packets.
2655                  * Must be zero or one.
2656                  */
2657                 if (m == NULL || m->m_len != sizeof(u_int)) {
2658                         error = EINVAL;
2659                         break;
2660                 }
2661                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2662                 if (loop > 1) {
2663                         error = EINVAL;
2664                         break;
2665                 }
2666                 im6o->im6o_multicast_loop = loop;
2667                 break;
2668
2669         case IPV6_JOIN_GROUP:
2670                 /*
2671                  * Add a multicast group membership.
2672                  * Group must be a valid IP6 multicast address.
2673                  */
2674                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2675                         error = EINVAL;
2676                         break;
2677                 }
2678                 mreq = mtod(m, struct ipv6_mreq *);
2679
2680                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2681                         /*
2682                          * We use the unspecified address to specify to accept
2683                          * all multicast addresses. Only super user is allowed
2684                          * to do this.
2685                          */
2686                         if (suser(td)) {
2687                                 error = EACCES;
2688                                 break;
2689                         }
2690                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2691                         error = EINVAL;
2692                         break;
2693                 }
2694
2695                 /*
2696                  * If no interface was explicitly specified, choose an
2697                  * appropriate one according to the given multicast address.
2698                  */
2699                 if (mreq->ipv6mr_interface == 0) {
2700                         struct sockaddr_in6 *dst;
2701
2702                         /*
2703                          * Look up the routing table for the
2704                          * address, and choose the outgoing interface.
2705                          *   XXX: is it a good approach?
2706                          */
2707                         ro.ro_rt = NULL;
2708                         dst = (struct sockaddr_in6 *)&ro.ro_dst;
2709                         bzero(dst, sizeof(*dst));
2710                         dst->sin6_family = AF_INET6;
2711                         dst->sin6_len = sizeof(*dst);
2712                         dst->sin6_addr = mreq->ipv6mr_multiaddr;
2713                         rtalloc((struct route *)&ro);
2714                         if (ro.ro_rt == NULL) {
2715                                 error = EADDRNOTAVAIL;
2716                                 break;
2717                         }
2718                         ifp = ro.ro_rt->rt_ifp;
2719                         RTFREE(ro.ro_rt);
2720                 } else {
2721                         /*
2722                          * If the interface is specified, validate it.
2723                          */
2724                         if (mreq->ipv6mr_interface < 0 ||
2725                             if_index < mreq->ipv6mr_interface) {
2726                                 error = ENXIO;  /* XXX EINVAL? */
2727                                 break;
2728                         }
2729                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2730                         if (!ifp) {
2731                                 error = ENXIO;  /* XXX EINVAL? */
2732                                 break;
2733                         }
2734                 }
2735
2736                 /*
2737                  * See if we found an interface, and confirm that it
2738                  * supports multicast
2739                  */
2740                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2741                         error = EADDRNOTAVAIL;
2742                         break;
2743                 }
2744
2745                 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2746                         error = EADDRNOTAVAIL; /* XXX: should not happen */
2747                         break;
2748                 }
2749
2750                 /*
2751                  * See if the membership already exists.
2752                  */
2753                 for (imm = im6o->im6o_memberships.lh_first;
2754                      imm != NULL; imm = imm->i6mm_chain.le_next)
2755                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2756                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2757                                                &mreq->ipv6mr_multiaddr))
2758                                 break;
2759                 if (imm != NULL) {
2760                         error = EADDRINUSE;
2761                         break;
2762                 }
2763                 /*
2764                  * Everything looks good; add a new record to the multicast
2765                  * address list for the given interface.
2766                  */
2767                 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2768                 if (imm == NULL)
2769                         break;
2770                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2771                 break;
2772
2773         case IPV6_LEAVE_GROUP:
2774                 /*
2775                  * Drop a multicast group membership.
2776                  * Group must be a valid IP6 multicast address.
2777                  */
2778                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2779                         error = EINVAL;
2780                         break;
2781                 }
2782                 mreq = mtod(m, struct ipv6_mreq *);
2783
2784                 /*
2785                  * If an interface address was specified, get a pointer
2786                  * to its ifnet structure.
2787                  */
2788                 if (mreq->ipv6mr_interface < 0 ||
2789                     if_index < mreq->ipv6mr_interface) {
2790                         error = ENXIO;  /* XXX EINVAL? */
2791                         break;
2792                 }
2793                 if (mreq->ipv6mr_interface == 0)
2794                         ifp = NULL;
2795                 else
2796                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2797
2798                 /* Fill in the scope zone ID */
2799                 if (ifp) {
2800                         if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2801                                 /* XXX: should not happen */
2802                                 error = EADDRNOTAVAIL;
2803                                 break;
2804                         }
2805                 } else if (mreq->ipv6mr_interface != 0) {
2806                         /*
2807                          * This case happens when the (positive) index is in
2808                          * the valid range, but the corresponding interface has
2809                          * been detached dynamically (XXX).
2810                          */
2811                         error = EADDRNOTAVAIL;
2812                         break;
2813                 } else {        /* ipv6mr_interface == 0 */
2814                         struct sockaddr_in6 sa6_mc;
2815
2816                         /*
2817                          * The API spec says as follows:
2818                          *  If the interface index is specified as 0, the
2819                          *  system may choose a multicast group membership to
2820                          *  drop by matching the multicast address only.
2821                          * On the other hand, we cannot disambiguate the scope
2822                          * zone unless an interface is provided.  Thus, we
2823                          * check if there's ambiguity with the default scope
2824                          * zone as the last resort.
2825                          */
2826                         bzero(&sa6_mc, sizeof(sa6_mc));
2827                         sa6_mc.sin6_family = AF_INET6;
2828                         sa6_mc.sin6_len = sizeof(sa6_mc);
2829                         sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2830                         error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2831                         if (error != 0)
2832                                 break;
2833                         mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2834                 }
2835
2836                 /*
2837                  * Find the membership in the membership list.
2838                  */
2839                 for (imm = im6o->im6o_memberships.lh_first;
2840                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2841                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2842                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2843                             &mreq->ipv6mr_multiaddr))
2844                                 break;
2845                 }
2846                 if (imm == NULL) {
2847                         /* Unable to resolve interface */
2848                         error = EADDRNOTAVAIL;
2849                         break;
2850                 }
2851                 /*
2852                  * Give up the multicast address record to which the
2853                  * membership points.
2854                  */
2855                 LIST_REMOVE(imm, i6mm_chain);
2856                 in6_delmulti(imm->i6mm_maddr);
2857                 free(imm, M_IP6MADDR);
2858                 break;
2859
2860         default:
2861                 error = EOPNOTSUPP;
2862                 break;
2863         }
2864
2865         /*
2866          * If all options have default values, no need to keep the mbuf.
2867          */
2868         if (im6o->im6o_multicast_ifp == NULL &&
2869             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2870             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2871             im6o->im6o_memberships.lh_first == NULL) {
2872                 free(*im6op, M_IP6MOPTS);
2873                 *im6op = NULL;
2874         }
2875
2876         return (error);
2877 }
2878
2879 /*
2880  * Return the IP6 multicast options in response to user getsockopt().
2881  */
2882 static int
2883 ip6_getmoptions(optname, im6o, mp)
2884         int optname;
2885         struct ip6_moptions *im6o;
2886         struct mbuf **mp;
2887 {
2888         u_int *hlim, *loop, *ifindex;
2889
2890         *mp = m_get(M_TRYWAIT, MT_HEADER);              /* XXX */
2891
2892         switch (optname) {
2893
2894         case IPV6_MULTICAST_IF:
2895                 ifindex = mtod(*mp, u_int *);
2896                 (*mp)->m_len = sizeof(u_int);
2897                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2898                         *ifindex = 0;
2899                 else
2900                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2901                 return (0);
2902
2903         case IPV6_MULTICAST_HOPS:
2904                 hlim = mtod(*mp, u_int *);
2905                 (*mp)->m_len = sizeof(u_int);
2906                 if (im6o == NULL)
2907                         *hlim = ip6_defmcasthlim;
2908                 else
2909                         *hlim = im6o->im6o_multicast_hlim;
2910                 return (0);
2911
2912         case IPV6_MULTICAST_LOOP:
2913                 loop = mtod(*mp, u_int *);
2914                 (*mp)->m_len = sizeof(u_int);
2915                 if (im6o == NULL)
2916                         *loop = ip6_defmcasthlim;
2917                 else
2918                         *loop = im6o->im6o_multicast_loop;
2919                 return (0);
2920
2921         default:
2922                 return (EOPNOTSUPP);
2923         }
2924 }
2925
2926 /*
2927  * Discard the IP6 multicast options.
2928  */
2929 void
2930 ip6_freemoptions(im6o)
2931         struct ip6_moptions *im6o;
2932 {
2933         struct in6_multi_mship *imm;
2934
2935         if (im6o == NULL)
2936                 return;
2937
2938         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2939                 LIST_REMOVE(imm, i6mm_chain);
2940                 if (imm->i6mm_maddr)
2941                         in6_delmulti(imm->i6mm_maddr);
2942                 free(imm, M_IP6MADDR);
2943         }
2944         free(im6o, M_IP6MOPTS);
2945 }
2946
2947 /*
2948  * Set IPv6 outgoing packet options based on advanced API.
2949  */
2950 int
2951 ip6_setpktopts(control, opt, stickyopt, priv, uproto)
2952         struct mbuf *control;
2953         struct ip6_pktopts *opt, *stickyopt;
2954         int priv, uproto;
2955 {
2956         struct cmsghdr *cm = 0;
2957
2958         if (control == NULL || opt == NULL)
2959                 return (EINVAL);
2960
2961         ip6_initpktopts(opt);
2962         if (stickyopt) {
2963                 int error;
2964
2965                 /*
2966                  * If stickyopt is provided, make a local copy of the options
2967                  * for this particular packet, then override them by ancillary
2968                  * objects.
2969                  * XXX: copypktopts() does not copy the cached route to a next
2970                  * hop (if any).  This is not very good in terms of efficiency,
2971                  * but we can allow this since this option should be rarely
2972                  * used.
2973                  */
2974                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2975                         return (error);
2976         }
2977
2978         /*
2979          * XXX: Currently, we assume all the optional information is stored
2980          * in a single mbuf.
2981          */
2982         if (control->m_next)
2983                 return (EINVAL);
2984
2985         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2986             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2987                 int error;
2988
2989                 if (control->m_len < CMSG_LEN(0))
2990                         return (EINVAL);
2991
2992                 cm = mtod(control, struct cmsghdr *);
2993                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2994                         return (EINVAL);
2995                 if (cm->cmsg_level != IPPROTO_IPV6)
2996                         continue;
2997
2998                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2999                     cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
3000                 if (error)
3001                         return (error);
3002         }
3003
3004         return (0);
3005 }
3006
3007 /*
3008  * Set a particular packet option, as a sticky option or an ancillary data
3009  * item.  "len" can be 0 only when it's a sticky option.
3010  * We have 4 cases of combination of "sticky" and "cmsg":
3011  * "sticky=0, cmsg=0": impossible
3012  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
3013  * "sticky=1, cmsg=0": RFC3542 socket option
3014  * "sticky=1, cmsg=1": RFC2292 socket option
3015  */
3016 static int
3017 ip6_setpktopt(optname, buf, len, opt, priv, sticky, cmsg, uproto)
3018         int optname, len, priv, sticky, cmsg, uproto;
3019         u_char *buf;
3020         struct ip6_pktopts *opt;
3021 {
3022         int minmtupolicy, preftemp;
3023
3024         if (!sticky && !cmsg) {
3025 #ifdef DIAGNOSTIC
3026                 printf("ip6_setpktopt: impossible case\n");
3027 #endif
3028                 return (EINVAL);
3029         }
3030
3031         /*
3032          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3033          * not be specified in the context of RFC3542.  Conversely,
3034          * RFC3542 types should not be specified in the context of RFC2292.
3035          */
3036         if (!cmsg) {
3037                 switch (optname) {
3038                 case IPV6_2292PKTINFO:
3039                 case IPV6_2292HOPLIMIT:
3040                 case IPV6_2292NEXTHOP:
3041                 case IPV6_2292HOPOPTS:
3042                 case IPV6_2292DSTOPTS:
3043                 case IPV6_2292RTHDR:
3044                 case IPV6_2292PKTOPTIONS:
3045                         return (ENOPROTOOPT);
3046                 }
3047         }
3048         if (sticky && cmsg) {
3049                 switch (optname) {
3050                 case IPV6_PKTINFO:
3051                 case IPV6_HOPLIMIT:
3052                 case IPV6_NEXTHOP:
3053                 case IPV6_HOPOPTS:
3054                 case IPV6_DSTOPTS:
3055                 case IPV6_RTHDRDSTOPTS:
3056                 case IPV6_RTHDR:
3057                 case IPV6_USE_MIN_MTU:
3058                 case IPV6_DONTFRAG:
3059                 case IPV6_TCLASS:
3060                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
3061                         return (ENOPROTOOPT);
3062                 }
3063         }
3064
3065         switch (optname) {
3066         case IPV6_2292PKTINFO:
3067         case IPV6_PKTINFO:
3068         {
3069                 struct ifnet *ifp = NULL;
3070                 struct in6_pktinfo *pktinfo;
3071
3072                 if (len != sizeof(struct in6_pktinfo))
3073                         return (EINVAL);
3074
3075                 pktinfo = (struct in6_pktinfo *)buf;
3076
3077                 /*
3078                  * An application can clear any sticky IPV6_PKTINFO option by
3079                  * doing a "regular" setsockopt with ipi6_addr being
3080                  * in6addr_any and ipi6_ifindex being zero.
3081                  * [RFC 3542, Section 6]
3082                  */
3083                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3084                     pktinfo->ipi6_ifindex == 0 &&
3085                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3086                         ip6_clearpktopts(opt, optname);
3087                         break;
3088                 }
3089
3090                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3091                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3092                         return (EINVAL);
3093                 }
3094
3095                 /* validate the interface index if specified. */
3096                 if (pktinfo->ipi6_ifindex > if_index ||
3097                     pktinfo->ipi6_ifindex < 0) {
3098                          return (ENXIO);
3099                 }
3100                 if (pktinfo->ipi6_ifindex) {
3101                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3102                         if (ifp == NULL)
3103                                 return (ENXIO);
3104                 }
3105
3106                 /*
3107                  * We store the address anyway, and let in6_selectsrc()
3108                  * validate the specified address.  This is because ipi6_addr
3109                  * may not have enough information about its scope zone, and
3110                  * we may need additional information (such as outgoing
3111                  * interface or the scope zone of a destination address) to
3112                  * disambiguate the scope.
3113                  * XXX: the delay of the validation may confuse the
3114                  * application when it is used as a sticky option.
3115                  */
3116                 if (opt->ip6po_pktinfo == NULL) {
3117                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3118                             M_IP6OPT, M_NOWAIT);
3119                         if (opt->ip6po_pktinfo == NULL)
3120                                 return (ENOBUFS);
3121                 }
3122                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3123                 break;
3124         }
3125
3126         case IPV6_2292HOPLIMIT:
3127         case IPV6_HOPLIMIT:
3128         {
3129                 int *hlimp;
3130
3131                 /*
3132                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3133                  * to simplify the ordering among hoplimit options.
3134                  */
3135                 if (optname == IPV6_HOPLIMIT && sticky)
3136                         return (ENOPROTOOPT);
3137
3138                 if (len != sizeof(int))
3139                         return (EINVAL);
3140                 hlimp = (int *)buf;
3141                 if (*hlimp < -1 || *hlimp > 255)
3142                         return (EINVAL);
3143
3144                 opt->ip6po_hlim = *hlimp;
3145                 break;
3146         }
3147
3148         case IPV6_TCLASS:
3149         {
3150                 int tclass;
3151
3152                 if (len != sizeof(int))
3153                         return (EINVAL);
3154                 tclass = *(int *)buf;
3155                 if (tclass < -1 || tclass > 255)
3156                         return (EINVAL);
3157
3158                 opt->ip6po_tclass = tclass;
3159                 break;
3160         }
3161
3162         case IPV6_2292NEXTHOP:
3163         case IPV6_NEXTHOP:
3164                 if (!priv)
3165                         return (EPERM);
3166
3167                 if (len == 0) { /* just remove the option */
3168                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3169                         break;
3170                 }
3171
3172                 /* check if cmsg_len is large enough for sa_len */
3173                 if (len < sizeof(struct sockaddr) || len < *buf)
3174                         return (EINVAL);
3175
3176                 switch (((struct sockaddr *)buf)->sa_family) {
3177                 case AF_INET6:
3178                 {
3179                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3180                         int error;
3181
3182                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3183                                 return (EINVAL);
3184
3185                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3186                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3187                                 return (EINVAL);
3188                         }
3189                         if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3190                             != 0) {
3191                                 return (error);
3192                         }
3193                         break;
3194                 }
3195                 case AF_LINK:   /* should eventually be supported */
3196                 default:
3197                         return (EAFNOSUPPORT);
3198                 }
3199
3200                 /* turn off the previous option, then set the new option. */
3201                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3202                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3203                 if (opt->ip6po_nexthop == NULL)
3204                         return (ENOBUFS);
3205                 bcopy(buf, opt->ip6po_nexthop, *buf);
3206                 break;
3207
3208         case IPV6_2292HOPOPTS:
3209         case IPV6_HOPOPTS:
3210         {
3211                 struct ip6_hbh *hbh;
3212                 int hbhlen;
3213
3214                 /*
3215                  * XXX: We don't allow a non-privileged user to set ANY HbH
3216                  * options, since per-option restriction has too much
3217                  * overhead.
3218                  */
3219                 if (!priv)
3220                         return (EPERM);
3221
3222                 if (len == 0) {
3223                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3224                         break;  /* just remove the option */
3225                 }
3226
3227                 /* message length validation */
3228                 if (len < sizeof(struct ip6_hbh))
3229                         return (EINVAL);
3230                 hbh = (struct ip6_hbh *)buf;
3231                 hbhlen = (hbh->ip6h_len + 1) << 3;
3232                 if (len != hbhlen)
3233                         return (EINVAL);
3234
3235                 /* turn off the previous option, then set the new option. */
3236                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3237                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3238                 if (opt->ip6po_hbh == NULL)
3239                         return (ENOBUFS);
3240                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3241
3242                 break;
3243         }
3244
3245         case IPV6_2292DSTOPTS:
3246         case IPV6_DSTOPTS:
3247         case IPV6_RTHDRDSTOPTS:
3248         {
3249                 struct ip6_dest *dest, **newdest = NULL;
3250                 int destlen;
3251
3252                 if (!priv)      /* XXX: see the comment for IPV6_HOPOPTS */
3253                         return (EPERM);
3254
3255                 if (len == 0) {
3256                         ip6_clearpktopts(opt, optname);
3257                         break;  /* just remove the option */
3258                 }
3259
3260                 /* message length validation */
3261                 if (len < sizeof(struct ip6_dest))
3262                         return (EINVAL);
3263                 dest = (struct ip6_dest *)buf;
3264                 destlen = (dest->ip6d_len + 1) << 3;
3265                 if (len != destlen)
3266                         return (EINVAL);
3267
3268                 /*
3269                  * Determine the position that the destination options header
3270                  * should be inserted; before or after the routing header.
3271                  */
3272                 switch (optname) {
3273                 case IPV6_2292DSTOPTS:
3274                         /*
3275                          * The old advacned API is ambiguous on this point.
3276                          * Our approach is to determine the position based
3277                          * according to the existence of a routing header.
3278                          * Note, however, that this depends on the order of the
3279                          * extension headers in the ancillary data; the 1st
3280                          * part of the destination options header must appear
3281                          * before the routing header in the ancillary data,
3282                          * too.
3283                          * RFC3542 solved the ambiguity by introducing
3284                          * separate ancillary data or option types.
3285                          */
3286                         if (opt->ip6po_rthdr == NULL)
3287                                 newdest = &opt->ip6po_dest1;
3288                         else
3289                                 newdest = &opt->ip6po_dest2;
3290                         break;
3291                 case IPV6_RTHDRDSTOPTS:
3292                         newdest = &opt->ip6po_dest1;
3293                         break;
3294                 case IPV6_DSTOPTS:
3295                         newdest = &opt->ip6po_dest2;
3296                         break;
3297                 }
3298
3299                 /* turn off the previous option, then set the new option. */
3300                 ip6_clearpktopts(opt, optname);
3301                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3302                 if (*newdest == NULL)
3303                         return (ENOBUFS);
3304                 bcopy(dest, *newdest, destlen);
3305
3306                 break;
3307         }
3308
3309         case IPV6_2292RTHDR:
3310         case IPV6_RTHDR:
3311         {
3312                 struct ip6_rthdr *rth;
3313                 int rthlen;
3314
3315                 if (len == 0) {
3316                         ip6_clearpktopts(opt, IPV6_RTHDR);
3317                         break;  /* just remove the option */
3318                 }
3319
3320                 /* message length validation */
3321                 if (len < sizeof(struct ip6_rthdr))
3322                         return (EINVAL);
3323                 rth = (struct ip6_rthdr *)buf;
3324                 rthlen = (rth->ip6r_len + 1) << 3;
3325                 if (len != rthlen)
3326                         return (EINVAL);
3327
3328                 switch (rth->ip6r_type) {
3329                 case IPV6_RTHDR_TYPE_0:
3330                         if (rth->ip6r_len == 0) /* must contain one addr */
3331                                 return (EINVAL);
3332                         if (rth->ip6r_len % 2) /* length must be even */
3333                                 return (EINVAL);
3334                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3335                                 return (EINVAL);
3336                         break;
3337                 default:
3338                         return (EINVAL);        /* not supported */
3339                 }
3340
3341                 /* turn off the previous option */
3342                 ip6_clearpktopts(opt, IPV6_RTHDR);
3343                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3344                 if (opt->ip6po_rthdr == NULL)
3345                         return (ENOBUFS);
3346                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3347
3348                 break;
3349         }
3350
3351         case IPV6_USE_MIN_MTU:
3352                 if (len != sizeof(int))
3353                         return (EINVAL);
3354                 minmtupolicy = *(int *)buf;
3355                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3356                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3357                     minmtupolicy != IP6PO_MINMTU_ALL) {
3358                         return (EINVAL);
3359                 }
3360                 opt->ip6po_minmtu = minmtupolicy;
3361                 break;
3362
3363         case IPV6_DONTFRAG:
3364                 if (len != sizeof(int))
3365                         return (EINVAL);
3366
3367                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3368                         /*
3369                          * we ignore this option for TCP sockets.
3370                          * (RFC3542 leaves this case unspecified.)
3371                          */
3372                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3373                 } else
3374                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3375                 break;
3376
3377         case IPV6_PREFER_TEMPADDR:
3378                 if (len != sizeof(int))
3379                         return (EINVAL);
3380                 preftemp = *(int *)buf;
3381                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3382                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3383                     preftemp != IP6PO_TEMPADDR_PREFER) {
3384                         return (EINVAL);
3385                 }
3386                 opt->ip6po_prefer_tempaddr = preftemp;
3387                 break;
3388
3389         default:
3390                 return (ENOPROTOOPT);
3391         } /* end of switch */
3392
3393         return (0);
3394 }
3395
3396 /*
3397  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3398  * packet to the input queue of a specified interface.  Note that this
3399  * calls the output routine of the loopback "driver", but with an interface
3400  * pointer that might NOT be &loif -- easier than replicating that code here.
3401  */
3402 void
3403 ip6_mloopback(ifp, m, dst)
3404         struct ifnet *ifp;
3405         struct mbuf *m;
3406         struct sockaddr_in6 *dst;
3407 {
3408         struct mbuf *copym;
3409         struct ip6_hdr *ip6;
3410
3411         copym = m_copy(m, 0, M_COPYALL);
3412         if (copym == NULL)
3413                 return;
3414
3415         /*
3416          * Make sure to deep-copy IPv6 header portion in case the data
3417          * is in an mbuf cluster, so that we can safely override the IPv6
3418          * header portion later.
3419          */
3420         if ((copym->m_flags & M_EXT) != 0 ||
3421             copym->m_len < sizeof(struct ip6_hdr)) {
3422                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3423                 if (copym == NULL)
3424                         return;
3425         }
3426
3427 #ifdef DIAGNOSTIC
3428         if (copym->m_len < sizeof(*ip6)) {
3429                 m_freem(copym);
3430                 return;
3431         }
3432 #endif
3433
3434         ip6 = mtod(copym, struct ip6_hdr *);
3435         /*
3436          * clear embedded scope identifiers if necessary.
3437          * in6_clearscope will touch the addresses only when necessary.
3438          */
3439         in6_clearscope(&ip6->ip6_src);
3440         in6_clearscope(&ip6->ip6_dst);
3441
3442         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3443 }
3444
3445 /*
3446  * Chop IPv6 header off from the payload.
3447  */
3448 static int
3449 ip6_splithdr(m, exthdrs)
3450         struct mbuf *m;
3451         struct ip6_exthdrs *exthdrs;
3452 {
3453         struct mbuf *mh;
3454         struct ip6_hdr *ip6;
3455
3456         ip6 = mtod(m, struct ip6_hdr *);
3457         if (m->m_len > sizeof(*ip6)) {
3458                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3459                 if (mh == 0) {
3460                         m_freem(m);
3461                         return ENOBUFS;
3462                 }
3463                 M_MOVE_PKTHDR(mh, m);
3464                 MH_ALIGN(mh, sizeof(*ip6));
3465                 m->m_len -= sizeof(*ip6);
3466                 m->m_data += sizeof(*ip6);
3467                 mh->m_next = m;
3468                 m = mh;
3469                 m->m_len = sizeof(*ip6);
3470                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3471         }
3472         exthdrs->ip6e_ip6 = m;
3473         return 0;
3474 }
3475
3476 /*
3477  * Compute IPv6 extension header length.
3478  */
3479 int
3480 ip6_optlen(in6p)
3481         struct in6pcb *in6p;
3482 {
3483         int len;
3484
3485         if (!in6p->in6p_outputopts)
3486                 return 0;
3487
3488         len = 0;
3489 #define elen(x) \
3490     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3491
3492         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3493         if (in6p->in6p_outputopts->ip6po_rthdr)
3494                 /* dest1 is valid with rthdr only */
3495                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3496         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3497         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3498         return len;
3499 #undef elen
3500 }