sys/netinet6/ip6_output.c

   1 /*      $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $       */
   2 /*      $DragonFly: src/sys/netinet6/ip6_output.c,v 1.10 2004/05/20 18:30:36 cpressey Exp $     */
   3 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   4
   5 /*
   6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the name of the project nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. All advertising materials mentioning features or use of this software
  47  *    must display the following acknowledgement:
  48  *      This product includes software developed by the University of
  49  *      California, Berkeley and its contributors.
  50  * 4. Neither the name of the University nor the names of its contributors
  51  *    may be used to endorse or promote products derived from this software
  52  *    without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  64  * SUCH DAMAGE.
  65  *
  66  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  67  */
  68
  69 #include "opt_ip6fw.h"
  70 #include "opt_inet.h"
  71 #include "opt_inet6.h"
  72 #include "opt_ipsec.h"
  73 #include "opt_pfil_hooks.h"
  74
  75 #include <sys/param.h>
  76 #include <sys/malloc.h>
  77 #include <sys/mbuf.h>
  78 #include <sys/errno.h>
  79 #include <sys/protosw.h>
  80 #include <sys/socket.h>
  81 #include <sys/socketvar.h>
  82 #include <sys/systm.h>
  83 #include <sys/kernel.h>
  84 #include <sys/proc.h>
  85
  86 #include <net/if.h>
  87 #include <net/route.h>
  88 #ifdef PFIL_HOOKS
  89 #include <net/pfil.h>
  90 #endif
  91
  92 #include <netinet/in.h>
  93 #include <netinet/in_var.h>
  94 #include <netinet6/in6_var.h>
  95 #include <netinet/ip6.h>
  96 #include <netinet/icmp6.h>
  97 #include <netinet6/ip6_var.h>
  98 #include <netinet/in_pcb.h>
  99 #include <netinet6/nd6.h>
 100
 101 #ifdef IPSEC
 102 #include <netinet6/ipsec.h>
 103 #ifdef INET6
 104 #include <netinet6/ipsec6.h>
 105 #endif
 106 #include <netproto/key/key.h>
 107 #endif /* IPSEC */
 108
 109 #ifdef FAST_IPSEC
 110 #include "ipsec.h"
 111 #include "ipsec6.h"
 112 #include <netipsec/key.h>
 113 #endif /* FAST_IPSEC */
 114
 115 #include <net/ip6fw/ip6_fw.h>
 116
 117 #include <net/net_osdep.h>
 118
 119 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 120
 121 struct ip6_exthdrs {
 122         struct mbuf *ip6e_ip6;
 123         struct mbuf *ip6e_hbh;
 124         struct mbuf *ip6e_dest1;
 125         struct mbuf *ip6e_rthdr;
 126         struct mbuf *ip6e_dest2;
 127 };
 128
 129 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
 130                             struct socket *, struct sockopt *sopt);
 131 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 132 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 133 static int ip6_copyexthdr (struct mbuf **, caddr_t, int);
 134 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
 135                                   struct ip6_frag **);
 136 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 137 static int ip6_splithdr (struct mbuf *, struct ip6_exthdrs *);
 138
 139 /*
 140  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 141  * header (with pri, len, nxt, hlim, src, dst).
 142  * This function may modify ver and hlim only.
 143  * The mbuf chain containing the packet will be freed.
 144  * The mbuf opt, if present, will not be freed.
 145  *
 146  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 147  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 148  * which is rt_rmx.rmx_mtu.
 149  */
 150 int
 151 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 152            int flags, struct ip6_moptions *im6o,
 153            struct ifnet **ifpp,         /* XXX: just for statistics */
 154            struct inpcb *inp)
 155 {
 156         struct ip6_hdr *ip6, *mhip6;
 157         struct ifnet *ifp, *origifp;
 158         struct mbuf *m = m0;
 159         int hlen, tlen, len, off;
 160         struct route_in6 ip6route;
 161         struct sockaddr_in6 *dst;
 162         int error = 0;
 163         struct in6_ifaddr *ia = NULL;
 164         u_long mtu;
 165         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 166         struct ip6_exthdrs exthdrs;
 167         struct in6_addr finaldst;
 168         struct route_in6 *ro_pmtu = NULL;
 169         int hdrsplit = 0;
 170         int needipsec = 0;
 171 #ifdef IPSEC
 172         int needipsectun = 0;
 173         struct secpolicy *sp = NULL;
 174         struct socket *so = inp ? inp->inp_socket : NULL;
 175
 176         ip6 = mtod(m, struct ip6_hdr *);
 177 #endif /* IPSEC */
 178 #ifdef FAST_IPSEC
 179         int needipsectun = 0;
 180         struct secpolicy *sp = NULL;
 181
 182         ip6 = mtod(m, struct ip6_hdr *);
 183 #endif /* FAST_IPSEC */
 184
 185 #define MAKE_EXTHDR(hp, mp)                                             \
 186     do {                                                                \
 187         if (hp) {                                                       \
 188                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 189                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 190                                        ((eh)->ip6e_len + 1) << 3);      \
 191                 if (error)                                              \
 192                         goto freehdrs;                                  \
 193         }                                                               \
 194     } while (0)
 195
 196         bzero(&exthdrs, sizeof(exthdrs));
 197
 198         if (opt) {
 199                 /* Hop-by-Hop options header */
 200                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 201                 /* Destination options header(1st part) */
 202                 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 203                 /* Routing header */
 204                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 205                 /* Destination options header(2nd part) */
 206                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 207         }
 208
 209 #ifdef IPSEC
 210         /* get a security policy for this packet */
 211         if (so == NULL)
 212                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 213         else
 214                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 215
 216         if (sp == NULL) {
 217                 ipsec6stat.out_inval++;
 218                 goto freehdrs;
 219         }
 220
 221         error = 0;
 222
 223         /* check policy */
 224         switch (sp->policy) {
 225         case IPSEC_POLICY_DISCARD:
 226                 /*
 227                  * This packet is just discarded.
 228                  */
 229                 ipsec6stat.out_polvio++;
 230                 goto freehdrs;
 231
 232         case IPSEC_POLICY_BYPASS:
 233         case IPSEC_POLICY_NONE:
 234                 /* no need to do IPsec. */
 235                 needipsec = 0;
 236                 break;
 237
 238         case IPSEC_POLICY_IPSEC:
 239                 if (sp->req == NULL) {
 240                         /* acquire a policy */
 241                         error = key_spdacquire(sp);
 242                         goto freehdrs;
 243                 }
 244                 needipsec = 1;
 245                 break;
 246
 247         case IPSEC_POLICY_ENTRUST:
 248         default:
 249                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 250         }
 251 #endif /* IPSEC */
 252 #ifdef FAST_IPSEC
 253         /* get a security policy for this packet */
 254         if (inp == NULL)
 255                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 256         else
 257                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 258
 259         if (sp == NULL) {
 260                 newipsecstat.ips_out_inval++;
 261                 goto freehdrs;
 262         }
 263
 264         error = 0;
 265
 266         /* check policy */
 267         switch (sp->policy) {
 268         case IPSEC_POLICY_DISCARD:
 269                 /*
 270                  * This packet is just discarded.
 271                  */
 272                 newipsecstat.ips_out_polvio++;
 273                 goto freehdrs;
 274
 275         case IPSEC_POLICY_BYPASS:
 276         case IPSEC_POLICY_NONE:
 277                 /* no need to do IPsec. */
 278                 needipsec = 0;
 279                 break;
 280
 281         case IPSEC_POLICY_IPSEC:
 282                 if (sp->req == NULL) {
 283                         /* acquire a policy */
 284                         error = key_spdacquire(sp);
 285                         goto freehdrs;
 286                 }
 287                 needipsec = 1;
 288                 break;
 289
 290         case IPSEC_POLICY_ENTRUST:
 291         default:
 292                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 293         }
 294 #endif /* FAST_IPSEC */
 295
 296         /*
 297          * Calculate the total length of the extension header chain.
 298          * Keep the length of the unfragmentable part for fragmentation.
 299          */
 300         optlen = 0;
 301         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 302         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 303         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 304         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 305         /* NOTE: we don't add AH/ESP length here. do that later. */
 306         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 307
 308         /*
 309          * If we need IPsec, or there is at least one extension header,
 310          * separate IP6 header from the payload.
 311          */
 312         if ((needipsec || optlen) && !hdrsplit) {
 313                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 314                         m = NULL;
 315                         goto freehdrs;
 316                 }
 317                 m = exthdrs.ip6e_ip6;
 318                 hdrsplit++;
 319         }
 320
 321         /* adjust pointer */
 322         ip6 = mtod(m, struct ip6_hdr *);
 323
 324         /* adjust mbuf packet header length */
 325         m->m_pkthdr.len += optlen;
 326         plen = m->m_pkthdr.len - sizeof(*ip6);
 327
 328         /* If this is a jumbo payload, insert a jumbo payload option. */
 329         if (plen > IPV6_MAXPACKET) {
 330                 if (!hdrsplit) {
 331                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 332                                 m = NULL;
 333                                 goto freehdrs;
 334                         }
 335                         m = exthdrs.ip6e_ip6;
 336                         hdrsplit++;
 337                 }
 338                 /* adjust pointer */
 339                 ip6 = mtod(m, struct ip6_hdr *);
 340                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 341                         goto freehdrs;
 342                 ip6->ip6_plen = 0;
 343         } else
 344                 ip6->ip6_plen = htons(plen);
 345
 346         /*
 347          * Concatenate headers and fill in next header fields.
 348          * Here we have, on "m"
 349          *      IPv6 payload
 350          * and we insert headers accordingly.  Finally, we should be getting:
 351          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 352          *
 353          * during the header composing process, "m" points to IPv6 header.
 354          * "mprev" points to an extension header prior to esp.
 355          */
 356         {
 357                 u_char *nexthdrp = &ip6->ip6_nxt;
 358                 struct mbuf *mprev = m;
 359
 360                 /*
 361                  * we treat dest2 specially.  this makes IPsec processing
 362                  * much easier.  the goal here is to make mprev point the
 363                  * mbuf prior to dest2.
 364                  *
 365                  * result: IPv6 dest2 payload
 366                  * m and mprev will point to IPv6 header.
 367                  */
 368                 if (exthdrs.ip6e_dest2) {
 369                         if (!hdrsplit)
 370                                 panic("assumption failed: hdr not split");
 371                         exthdrs.ip6e_dest2->m_next = m->m_next;
 372                         m->m_next = exthdrs.ip6e_dest2;
 373                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 374                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 375                 }
 376
 377 #define MAKE_CHAIN(m, mp, p, i)\
 378     do {\
 379         if (m) {\
 380                 if (!hdrsplit) \
 381                         panic("assumption failed: hdr not split"); \
 382                 *mtod((m), u_char *) = *(p);\
 383                 *(p) = (i);\
 384                 p = mtod((m), u_char *);\
 385                 (m)->m_next = (mp)->m_next;\
 386                 (mp)->m_next = (m);\
 387                 (mp) = (m);\
 388         }\
 389     } while (0)
 390                 /*
 391                  * result: IPv6 hbh dest1 rthdr dest2 payload
 392                  * m will point to IPv6 header.  mprev will point to the
 393                  * extension header prior to dest2 (rthdr in the above case).
 394                  */
 395                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
 396                            nexthdrp, IPPROTO_HOPOPTS);
 397                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
 398                            nexthdrp, IPPROTO_DSTOPTS);
 399                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
 400                            nexthdrp, IPPROTO_ROUTING);
 401
 402 #if defined(IPSEC) || defined(FAST_IPSEC)
 403                 if (!needipsec)
 404                         goto skip_ipsec2;
 405
 406                 /*
 407                  * pointers after IPsec headers are not valid any more.
 408                  * other pointers need a great care too.
 409                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 410                  */
 411                 exthdrs.ip6e_dest2 = NULL;
 412
 413             {
 414                 struct ip6_rthdr *rh = NULL;
 415                 int segleft_org = 0;
 416                 struct ipsec_output_state state;
 417
 418                 if (exthdrs.ip6e_rthdr) {
 419                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 420                         segleft_org = rh->ip6r_segleft;
 421                         rh->ip6r_segleft = 0;
 422                 }
 423
 424                 bzero(&state, sizeof(state));
 425                 state.m = m;
 426                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 427                         &needipsectun);
 428                 m = state.m;
 429                 if (error) {
 430                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 431                         m = NULL;
 432                         switch (error) {
 433                         case EHOSTUNREACH:
 434                         case ENETUNREACH:
 435                         case EMSGSIZE:
 436                         case ENOBUFS:
 437                         case ENOMEM:
 438                                 break;
 439                         default:
 440                                 printf("ip6_output (ipsec): error code %d\n", error);
 441                                 /* fall through */
 442                         case ENOENT:
 443                                 /* don't show these error codes to the user */
 444                                 error = 0;
 445                                 break;
 446                         }
 447                         goto bad;
 448                 }
 449                 if (exthdrs.ip6e_rthdr) {
 450                         /* ah6_output doesn't modify mbuf chain */
 451                         rh->ip6r_segleft = segleft_org;
 452                 }
 453             }
 454 skip_ipsec2:;
 455 #endif
 456         }
 457
 458         /*
 459          * If there is a routing header, replace destination address field
 460          * with the first hop of the routing header.
 461          */
 462         if (exthdrs.ip6e_rthdr) {
 463                 struct ip6_rthdr *rh =
 464                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 465                                                   struct ip6_rthdr *));
 466                 struct ip6_rthdr0 *rh0;
 467
 468                 finaldst = ip6->ip6_dst;
 469                 switch (rh->ip6r_type) {
 470                 case IPV6_RTHDR_TYPE_0:
 471                          rh0 = (struct ip6_rthdr0 *)rh;
 472                          ip6->ip6_dst = rh0->ip6r0_addr[0];
 473                          bcopy((caddr_t)&rh0->ip6r0_addr[1],
 474                                (caddr_t)&rh0->ip6r0_addr[0],
 475                                sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
 476                                  );
 477                          rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
 478                          break;
 479                 default:        /* is it possible? */
 480                          error = EINVAL;
 481                          goto bad;
 482                 }
 483         }
 484
 485         /* Source address validation */
 486         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 487             (flags & IPV6_DADOUTPUT) == 0) {
 488                 error = EOPNOTSUPP;
 489                 ip6stat.ip6s_badscope++;
 490                 goto bad;
 491         }
 492         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 493                 error = EOPNOTSUPP;
 494                 ip6stat.ip6s_badscope++;
 495                 goto bad;
 496         }
 497
 498         ip6stat.ip6s_localout++;
 499
 500         /*
 501          * Route packet.
 502          */
 503         if (ro == 0) {
 504                 ro = &ip6route;
 505                 bzero((caddr_t)ro, sizeof(*ro));
 506         }
 507         ro_pmtu = ro;
 508         if (opt && opt->ip6po_rthdr)
 509                 ro = &opt->ip6po_route;
 510         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 511         /*
 512          * If there is a cached route,
 513          * check that it is to the same destination
 514          * and is still up. If not, free it and try again.
 515          */
 516         if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 517                          dst->sin6_family != AF_INET6 ||
 518                          !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 519                 RTFREE(ro->ro_rt);
 520                 ro->ro_rt = (struct rtentry *)0;
 521         }
 522         if (ro->ro_rt == 0) {
 523                 bzero(dst, sizeof(*dst));
 524                 dst->sin6_family = AF_INET6;
 525                 dst->sin6_len = sizeof(struct sockaddr_in6);
 526                 dst->sin6_addr = ip6->ip6_dst;
 527 #ifdef SCOPEDROUTING
 528                 /* XXX: sin6_scope_id should already be fixed at this point */
 529                 if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
 530                         dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
 531 #endif
 532         }
 533 #if defined(IPSEC) || defined(FAST_IPSEC)
 534         if (needipsec && needipsectun) {
 535                 struct ipsec_output_state state;
 536
 537                 /*
 538                  * All the extension headers will become inaccessible
 539                  * (since they can be encrypted).
 540                  * Don't panic, we need no more updates to extension headers
 541                  * on inner IPv6 packet (since they are now encapsulated).
 542                  *
 543                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 544                  */
 545                 bzero(&exthdrs, sizeof(exthdrs));
 546                 exthdrs.ip6e_ip6 = m;
 547
 548                 bzero(&state, sizeof(state));
 549                 state.m = m;
 550                 state.ro = (struct route *)ro;
 551                 state.dst = (struct sockaddr *)dst;
 552
 553                 error = ipsec6_output_tunnel(&state, sp, flags);
 554
 555                 m = state.m;
 556                 ro = (struct route_in6 *)state.ro;
 557                 dst = (struct sockaddr_in6 *)state.dst;
 558                 if (error) {
 559                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 560                         m0 = m = NULL;
 561                         m = NULL;
 562                         switch (error) {
 563                         case EHOSTUNREACH:
 564                         case ENETUNREACH:
 565                         case EMSGSIZE:
 566                         case ENOBUFS:
 567                         case ENOMEM:
 568                                 break;
 569                         default:
 570                                 printf("ip6_output (ipsec): error code %d\n", error);
 571                                 /* fall through */
 572                         case ENOENT:
 573                                 /* don't show these error codes to the user */
 574                                 error = 0;
 575                                 break;
 576                         }
 577                         goto bad;
 578                 }
 579
 580                 exthdrs.ip6e_ip6 = m;
 581         }
 582 #endif /* IPSEC */
 583
 584         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 585                 /* Unicast */
 586
 587 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 588 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 589                 /* xxx
 590                  * interface selection comes here
 591                  * if an interface is specified from an upper layer,
 592                  * ifp must point it.
 593                  */
 594                 if (ro->ro_rt == 0) {
 595                         /*
 596                          * non-bsdi always clone routes, if parent is
 597                          * PRF_CLONING.
 598                          */
 599                         rtalloc((struct route *)ro);
 600                 }
 601                 if (ro->ro_rt == 0) {
 602                         ip6stat.ip6s_noroute++;
 603                         error = EHOSTUNREACH;
 604                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 605                         goto bad;
 606                 }
 607                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 608                 ifp = ro->ro_rt->rt_ifp;
 609                 ro->ro_rt->rt_use++;
 610                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 611                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 612                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 613
 614                 in6_ifstat_inc(ifp, ifs6_out_request);
 615
 616                 /*
 617                  * Check if the outgoing interface conflicts with
 618                  * the interface specified by ifi6_ifindex (if specified).
 619                  * Note that loopback interface is always okay.
 620                  * (this may happen when we are sending a packet to one of
 621                  *  our own addresses.)
 622                  */
 623                 if (opt && opt->ip6po_pktinfo
 624                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 625                         if (!(ifp->if_flags & IFF_LOOPBACK)
 626                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 627                                 ip6stat.ip6s_noroute++;
 628                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 629                                 error = EHOSTUNREACH;
 630                                 goto bad;
 631                         }
 632                 }
 633
 634                 if (opt && opt->ip6po_hlim != -1)
 635                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 636         } else {
 637                 /* Multicast */
 638                 struct  in6_multi *in6m;
 639
 640                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 641
 642                 /*
 643                  * See if the caller provided any multicast options
 644                  */
 645                 ifp = NULL;
 646                 if (im6o != NULL) {
 647                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 648                         if (im6o->im6o_multicast_ifp != NULL)
 649                                 ifp = im6o->im6o_multicast_ifp;
 650                 } else
 651                         ip6->ip6_hlim = ip6_defmcasthlim;
 652
 653                 /*
 654                  * See if the caller provided the outgoing interface
 655                  * as an ancillary data.
 656                  * Boundary check for ifindex is assumed to be already done.
 657                  */
 658                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 659                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 660
 661                 /*
 662                  * If the destination is a node-local scope multicast,
 663                  * the packet should be loop-backed only.
 664                  */
 665                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 666                         /*
 667                          * If the outgoing interface is already specified,
 668                          * it should be a loopback interface.
 669                          */
 670                         if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
 671                                 ip6stat.ip6s_badscope++;
 672                                 error = ENETUNREACH; /* XXX: better error? */
 673                                 /* XXX correct ifp? */
 674                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 675                                 goto bad;
 676                         } else {
 677                                 ifp = &loif[0];
 678                         }
 679                 }
 680
 681                 if (opt && opt->ip6po_hlim != -1)
 682                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 683
 684                 /*
 685                  * If caller did not provide an interface lookup a
 686                  * default in the routing table.  This is either a
 687                  * default for the speicfied group (i.e. a host
 688                  * route), or a multicast default (a route for the
 689                  * ``net'' ff00::/8).
 690                  */
 691                 if (ifp == NULL) {
 692                         if (ro->ro_rt == 0) {
 693                                 ro->ro_rt = rtalloc1((struct sockaddr *)
 694                                                 &ro->ro_dst, 0, 0UL);
 695                         }
 696                         if (ro->ro_rt == 0) {
 697                                 ip6stat.ip6s_noroute++;
 698                                 error = EHOSTUNREACH;
 699                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 700                                 goto bad;
 701                         }
 702                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 703                         ifp = ro->ro_rt->rt_ifp;
 704                         ro->ro_rt->rt_use++;
 705                 }
 706
 707                 if ((flags & IPV6_FORWARDING) == 0)
 708                         in6_ifstat_inc(ifp, ifs6_out_request);
 709                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 710
 711                 /*
 712                  * Confirm that the outgoing interface supports multicast.
 713                  */
 714                 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 715                         ip6stat.ip6s_noroute++;
 716                         in6_ifstat_inc(ifp, ifs6_out_discard);
 717                         error = ENETUNREACH;
 718                         goto bad;
 719                 }
 720                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 721                 if (in6m != NULL &&
 722                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 723                         /*
 724                          * If we belong to the destination multicast group
 725                          * on the outgoing interface, and the caller did not
 726                          * forbid loopback, loop back a copy.
 727                          */
 728                         ip6_mloopback(ifp, m, dst);
 729                 } else {
 730                         /*
 731                          * If we are acting as a multicast router, perform
 732                          * multicast forwarding as if the packet had just
 733                          * arrived on the interface to which we are about
 734                          * to send.  The multicast forwarding function
 735                          * recursively calls this function, using the
 736                          * IPV6_FORWARDING flag to prevent infinite recursion.
 737                          *
 738                          * Multicasts that are looped back by ip6_mloopback(),
 739                          * above, will be forwarded by the ip6_input() routine,
 740                          * if necessary.
 741                          */
 742                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 743                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 744                                         m_freem(m);
 745                                         goto done;
 746                                 }
 747                         }
 748                 }
 749                 /*
 750                  * Multicasts with a hoplimit of zero may be looped back,
 751                  * above, but must not be transmitted on a network.
 752                  * Also, multicasts addressed to the loopback interface
 753                  * are not sent -- the above call to ip6_mloopback() will
 754                  * loop back a copy if this host actually belongs to the
 755                  * destination group on the loopback interface.
 756                  */
 757                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 758                         m_freem(m);
 759                         goto done;
 760                 }
 761         }
 762
 763         /*
 764          * Fill the outgoing inteface to tell the upper layer
 765          * to increment per-interface statistics.
 766          */
 767         if (ifpp)
 768                 *ifpp = ifp;
 769
 770         /*
 771          * Determine path MTU.
 772          */
 773         if (ro_pmtu != ro) {
 774                 /* The first hop and the final destination may differ. */
 775                 struct sockaddr_in6 *sin6_fin =
 776                         (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
 777                 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 778                                        !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
 779                                                            &finaldst))) {
 780                         RTFREE(ro_pmtu->ro_rt);
 781                         ro_pmtu->ro_rt = (struct rtentry *)0;
 782                 }
 783                 if (ro_pmtu->ro_rt == 0) {
 784                         bzero(sin6_fin, sizeof(*sin6_fin));
 785                         sin6_fin->sin6_family = AF_INET6;
 786                         sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
 787                         sin6_fin->sin6_addr = finaldst;
 788
 789                         rtalloc((struct route *)ro_pmtu);
 790                 }
 791         }
 792         if (ro_pmtu->ro_rt != NULL) {
 793                 u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
 794
 795                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
 796                 if (mtu > ifmtu || mtu == 0) {
 797                         /*
 798                          * The MTU on the route is larger than the MTU on
 799                          * the interface!  This shouldn't happen, unless the
 800                          * MTU of the interface has been changed after the
 801                          * interface was brought up.  Change the MTU in the
 802                          * route to match the interface MTU (as long as the
 803                          * field isn't locked).
 804                          *
 805                          * if MTU on the route is 0, we need to fix the MTU.
 806                          * this case happens with path MTU discovery timeouts.
 807                          */
 808                          mtu = ifmtu;
 809                          if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
 810                                  ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
 811                 }
 812         } else {
 813                 mtu = nd_ifinfo[ifp->if_index].linkmtu;
 814         }
 815
 816         /*
 817          * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
 818          */
 819         if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
 820                 mtu = IPV6_MMTU;
 821
 822         /* Fake scoped addresses */
 823         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 824                 /*
 825                  * If source or destination address is a scoped address, and
 826                  * the packet is going to be sent to a loopback interface,
 827                  * we should keep the original interface.
 828                  */
 829
 830                 /*
 831                  * XXX: this is a very experimental and temporary solution.
 832                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 833                  * field of the structure here.
 834                  * We rely on the consistency between two scope zone ids
 835                  * of source and destination, which should already be assured.
 836                  * Larger scopes than link will be supported in the future.
 837                  */
 838                 origifp = NULL;
 839                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 840                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 841                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 842                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 843                 /*
 844                  * XXX: origifp can be NULL even in those two cases above.
 845                  * For example, if we remove the (only) link-local address
 846                  * from the loopback interface, and try to send a link-local
 847                  * address without link-id information.  Then the source
 848                  * address is ::1, and the destination address is the
 849                  * link-local address with its s6_addr16[1] being zero.
 850                  * What is worse, if the packet goes to the loopback interface
 851                  * by a default rejected route, the null pointer would be
 852                  * passed to looutput, and the kernel would hang.
 853                  * The following last resort would prevent such disaster.
 854                  */
 855                 if (origifp == NULL)
 856                         origifp = ifp;
 857         }
 858         else
 859                 origifp = ifp;
 860 #ifndef SCOPEDROUTING
 861         /*
 862          * clear embedded scope identifiers if necessary.
 863          * in6_clearscope will touch the addresses only when necessary.
 864          */
 865         in6_clearscope(&ip6->ip6_src);
 866         in6_clearscope(&ip6->ip6_dst);
 867 #endif
 868
 869         /*
 870          * Check with the firewall...
 871          */
 872         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 873                 u_short port = 0;
 874                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 875                 /* If ipfw says divert, we have to just drop packet */
 876                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 877                         m_freem(m);
 878                         goto done;
 879                 }
 880                 if (!m) {
 881                         error = EACCES;
 882                         goto done;
 883                 }
 884         }
 885
 886         /*
 887          * If the outgoing packet contains a hop-by-hop options header,
 888          * it must be examined and processed even by the source node.
 889          * (RFC 2460, section 4.)
 890          */
 891         if (exthdrs.ip6e_hbh) {
 892                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 893                 u_int32_t dummy1; /* XXX unused */
 894                 u_int32_t dummy2; /* XXX unused */
 895
 896 #ifdef DIAGNOSTIC
 897                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 898                         panic("ip6e_hbh is not continuous");
 899 #endif
 900                 /*
 901                  *  XXX: if we have to send an ICMPv6 error to the sender,
 902                  *       we need the M_LOOP flag since icmp6_error() expects
 903                  *       the IPv6 and the hop-by-hop options header are
 904                  *       continuous unless the flag is set.
 905                  */
 906                 m->m_flags |= M_LOOP;
 907                 m->m_pkthdr.rcvif = ifp;
 908                 if (ip6_process_hopopts(m,
 909                                         (u_int8_t *)(hbh + 1),
 910                                         ((hbh->ip6h_len + 1) << 3) -
 911                                         sizeof(struct ip6_hbh),
 912                                         &dummy1, &dummy2) < 0) {
 913                         /* m was already freed at this point */
 914                         error = EINVAL;/* better error? */
 915                         goto done;
 916                 }
 917                 m->m_flags &= ~M_LOOP; /* XXX */
 918                 m->m_pkthdr.rcvif = NULL;
 919         }
 920
 921 #ifdef PFIL_HOOKS
 922         /*
 923          * Run through list of hooks for output packets.
 924          */
 925         error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 926         if (error != 0 || m == NULL)
 927                 goto done;
 928         ip6 = mtod(m, struct ip6_hdr *);
 929 #endif /* PFIL_HOOKS */
 930         /*
 931          * Send the packet to the outgoing interface.
 932          * If necessary, do IPv6 fragmentation before sending.
 933          */
 934         tlen = m->m_pkthdr.len;
 935         if (tlen <= mtu
 936 #ifdef notyet
 937             /*
 938              * On any link that cannot convey a 1280-octet packet in one piece,
 939              * link-specific fragmentation and reassembly must be provided at
 940              * a layer below IPv6. [RFC 2460, sec.5]
 941              * Thus if the interface has ability of link-level fragmentation,
 942              * we can just send the packet even if the packet size is
 943              * larger than the link's MTU.
 944              * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
 945              */
 946
 947             || ifp->if_flags & IFF_FRAGMENTABLE
 948 #endif
 949             )
 950         {
 951                 /* Record statistics for this interface address. */
 952                 if (ia && !(flags & IPV6_FORWARDING)) {
 953                         ia->ia_ifa.if_opackets++;
 954                         ia->ia_ifa.if_obytes += m->m_pkthdr.len;
 955                 }
 956 #ifdef IPSEC
 957                 /* clean ipsec history once it goes out of the node */
 958                 ipsec_delaux(m);
 959 #endif
 960                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 961                 goto done;
 962         } else if (mtu < IPV6_MMTU) {
 963                 /*
 964                  * note that path MTU is never less than IPV6_MMTU
 965                  * (see icmp6_input).
 966                  */
 967                 error = EMSGSIZE;
 968                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 969                 goto bad;
 970         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 971                 error = EMSGSIZE;
 972                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 973                 goto bad;
 974         } else {
 975                 struct mbuf **mnext, *m_frgpart;
 976                 struct ip6_frag *ip6f;
 977                 u_int32_t id = htonl(ip6_id++);
 978                 u_char nextproto;
 979
 980                 /*
 981                  * Too large for the destination or interface;
 982                  * fragment if possible.
 983                  * Must be able to put at least 8 bytes per fragment.
 984                  */
 985                 hlen = unfragpartlen;
 986                 if (mtu > IPV6_MAXPACKET)
 987                         mtu = IPV6_MAXPACKET;
 988
 989                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 990                 if (len < 8) {
 991                         error = EMSGSIZE;
 992                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 993                         goto bad;
 994                 }
 995
 996                 mnext = &m->m_nextpkt;
 997
 998                 /*
 999                  * Change the next header field of the last header in the
1000                  * unfragmentable part.
1001                  */
1002                 if (exthdrs.ip6e_rthdr) {
1003                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1004                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1005                 } else if (exthdrs.ip6e_dest1) {
1006                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1007                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1008                 } else if (exthdrs.ip6e_hbh) {
1009                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1010                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1011                 } else {
1012                         nextproto = ip6->ip6_nxt;
1013                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1014                 }
1015
1016                 /*
1017                  * Loop through length of segment after first fragment,
1018                  * make new header and copy data of each part and link onto
1019                  * chain.
1020                  */
1021                 m0 = m;
1022                 for (off = hlen; off < tlen; off += len) {
1023                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1024                         if (!m) {
1025                                 error = ENOBUFS;
1026                                 ip6stat.ip6s_odropped++;
1027                                 goto sendorfree;
1028                         }
1029                         m->m_pkthdr.rcvif = NULL;
1030                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1031                         *mnext = m;
1032                         mnext = &m->m_nextpkt;
1033                         m->m_data += max_linkhdr;
1034                         mhip6 = mtod(m, struct ip6_hdr *);
1035                         *mhip6 = *ip6;
1036                         m->m_len = sizeof(*mhip6);
1037                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1038                         if (error) {
1039                                 ip6stat.ip6s_odropped++;
1040                                 goto sendorfree;
1041                         }
1042                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1043                         if (off + len >= tlen)
1044                                 len = tlen - off;
1045                         else
1046                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1047                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1048                                                           sizeof(*ip6f) -
1049                                                           sizeof(struct ip6_hdr)));
1050                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1051                                 error = ENOBUFS;
1052                                 ip6stat.ip6s_odropped++;
1053                                 goto sendorfree;
1054                         }
1055                         m_cat(m, m_frgpart);
1056                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1057                         m->m_pkthdr.rcvif = (struct ifnet *)0;
1058                         ip6f->ip6f_reserved = 0;
1059                         ip6f->ip6f_ident = id;
1060                         ip6f->ip6f_nxt = nextproto;
1061                         ip6stat.ip6s_ofragments++;
1062                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1063                 }
1064
1065                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1066         }
1067
1068         /*
1069          * Remove leading garbages.
1070          */
1071 sendorfree:
1072         m = m0->m_nextpkt;
1073         m0->m_nextpkt = 0;
1074         m_freem(m0);
1075         for (m0 = m; m; m = m0) {
1076                 m0 = m->m_nextpkt;
1077                 m->m_nextpkt = 0;
1078                 if (error == 0) {
1079                         /* Record statistics for this interface address. */
1080                         if (ia) {
1081                                 ia->ia_ifa.if_opackets++;
1082                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1083                         }
1084 #ifdef IPSEC
1085                         /* clean ipsec history once it goes out of the node */
1086                         ipsec_delaux(m);
1087 #endif
1088                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1089                 } else
1090                         m_freem(m);
1091         }
1092
1093         if (error == 0)
1094                 ip6stat.ip6s_fragmented++;
1095
1096 done:
1097         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1098                 RTFREE(ro->ro_rt);
1099         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1100                 RTFREE(ro_pmtu->ro_rt);
1101         }
1102
1103 #ifdef IPSEC
1104         if (sp != NULL)
1105                 key_freesp(sp);
1106 #endif /* IPSEC */
1107 #ifdef FAST_IPSEC
1108         if (sp != NULL)
1109                 KEY_FREESP(&sp);
1110 #endif /* FAST_IPSEC */
1111
1112         return(error);
1113
1114 freehdrs:
1115         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1116         m_freem(exthdrs.ip6e_dest1);
1117         m_freem(exthdrs.ip6e_rthdr);
1118         m_freem(exthdrs.ip6e_dest2);
1119         /* fall through */
1120 bad:
1121         m_freem(m);
1122         goto done;
1123 }
1124
1125 static int
1126 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1127 {
1128         struct mbuf *m;
1129
1130         if (hlen > MCLBYTES)
1131                 return(ENOBUFS); /* XXX */
1132
1133         MGET(m, M_DONTWAIT, MT_DATA);
1134         if (!m)
1135                 return(ENOBUFS);
1136
1137         if (hlen > MLEN) {
1138                 MCLGET(m, M_DONTWAIT);
1139                 if ((m->m_flags & M_EXT) == 0) {
1140                         m_free(m);
1141                         return(ENOBUFS);
1142                 }
1143         }
1144         m->m_len = hlen;
1145         if (hdr)
1146                 bcopy(hdr, mtod(m, caddr_t), hlen);
1147
1148         *mp = m;
1149         return(0);
1150 }
1151
1152 /*
1153  * Insert jumbo payload option.
1154  */
1155 static int
1156 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1157 {
1158         struct mbuf *mopt;
1159         u_char *optbuf;
1160         u_int32_t v;
1161
1162 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1163
1164         /*
1165          * If there is no hop-by-hop options header, allocate new one.
1166          * If there is one but it doesn't have enough space to store the
1167          * jumbo payload option, allocate a cluster to store the whole options.
1168          * Otherwise, use it to store the options.
1169          */
1170         if (exthdrs->ip6e_hbh == 0) {
1171                 MGET(mopt, M_DONTWAIT, MT_DATA);
1172                 if (mopt == 0)
1173                         return(ENOBUFS);
1174                 mopt->m_len = JUMBOOPTLEN;
1175                 optbuf = mtod(mopt, u_char *);
1176                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1177                 exthdrs->ip6e_hbh = mopt;
1178         } else {
1179                 struct ip6_hbh *hbh;
1180
1181                 mopt = exthdrs->ip6e_hbh;
1182                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1183                         /*
1184                          * XXX assumption:
1185                          * - exthdrs->ip6e_hbh is not referenced from places
1186                          *   other than exthdrs.
1187                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1188                          */
1189                         int oldoptlen = mopt->m_len;
1190                         struct mbuf *n;
1191
1192                         /*
1193                          * XXX: give up if the whole (new) hbh header does
1194                          * not fit even in an mbuf cluster.
1195                          */
1196                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1197                                 return(ENOBUFS);
1198
1199                         /*
1200                          * As a consequence, we must always prepare a cluster
1201                          * at this point.
1202                          */
1203                         MGET(n, M_DONTWAIT, MT_DATA);
1204                         if (n) {
1205                                 MCLGET(n, M_DONTWAIT);
1206                                 if ((n->m_flags & M_EXT) == 0) {
1207                                         m_freem(n);
1208                                         n = NULL;
1209                                 }
1210                         }
1211                         if (!n)
1212                                 return(ENOBUFS);
1213                         n->m_len = oldoptlen + JUMBOOPTLEN;
1214                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1215                               oldoptlen);
1216                         optbuf = mtod(n, caddr_t) + oldoptlen;
1217                         m_freem(mopt);
1218                         mopt = exthdrs->ip6e_hbh = n;
1219                 } else {
1220                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1221                         mopt->m_len += JUMBOOPTLEN;
1222                 }
1223                 optbuf[0] = IP6OPT_PADN;
1224                 optbuf[1] = 1;
1225
1226                 /*
1227                  * Adjust the header length according to the pad and
1228                  * the jumbo payload option.
1229                  */
1230                 hbh = mtod(mopt, struct ip6_hbh *);
1231                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1232         }
1233
1234         /* fill in the option. */
1235         optbuf[2] = IP6OPT_JUMBO;
1236         optbuf[3] = 4;
1237         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1238         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1239
1240         /* finally, adjust the packet header length */
1241         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1242
1243         return(0);
1244 #undef JUMBOOPTLEN
1245 }
1246
1247 /*
1248  * Insert fragment header and copy unfragmentable header portions.
1249  */
1250 static int
1251 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1252                   struct ip6_frag **frghdrp)
1253 {
1254         struct mbuf *n, *mlast;
1255
1256         if (hlen > sizeof(struct ip6_hdr)) {
1257                 n = m_copym(m0, sizeof(struct ip6_hdr),
1258                             hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1259                 if (n == 0)
1260                         return(ENOBUFS);
1261                 m->m_next = n;
1262         } else
1263                 n = m;
1264
1265         /* Search for the last mbuf of unfragmentable part. */
1266         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1267                 ;
1268
1269         if ((mlast->m_flags & M_EXT) == 0 &&
1270             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1271                 /* use the trailing space of the last mbuf for the fragment hdr */
1272                 *frghdrp =
1273                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1274                 mlast->m_len += sizeof(struct ip6_frag);
1275                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1276         } else {
1277                 /* allocate a new mbuf for the fragment header */
1278                 struct mbuf *mfrg;
1279
1280                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1281                 if (mfrg == 0)
1282                         return(ENOBUFS);
1283                 mfrg->m_len = sizeof(struct ip6_frag);
1284                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1285                 mlast->m_next = mfrg;
1286         }
1287
1288         return(0);
1289 }
1290
1291 /*
1292  * IP6 socket option processing.
1293  */
1294 int
1295 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1296 {
1297         int privileged;
1298         struct inpcb *in6p = sotoinpcb(so);
1299         int error, optval;
1300         int level, op, optname;
1301         int optlen;
1302         struct thread *td;
1303
1304         if (sopt) {
1305                 level = sopt->sopt_level;
1306                 op = sopt->sopt_dir;
1307                 optname = sopt->sopt_name;
1308                 optlen = sopt->sopt_valsize;
1309                 td = sopt->sopt_td;
1310         } else {
1311                 panic("ip6_ctloutput: arg soopt is NULL");
1312                 /* NOT REACHED */
1313                 td = NULL;
1314         }
1315         error = optval = 0;
1316
1317         privileged = (td == NULL || suser(td)) ? 0 : 1;
1318
1319         if (level == IPPROTO_IPV6) {
1320                 switch (op) {
1321
1322                 case SOPT_SET:
1323                         switch (optname) {
1324                         case IPV6_PKTOPTIONS:
1325                         {
1326                                 struct mbuf *m;
1327
1328                                 error = soopt_getm(sopt, &m); /* XXX */
1329                                 if (error != NULL)
1330                                         break;
1331                                 error = soopt_mcopyin(sopt, m); /* XXX */
1332                                 if (error != NULL)
1333                                         break;
1334                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1335                                                     m, so, sopt);
1336                                 m_freem(m); /* XXX */
1337                                 break;
1338                         }
1339
1340                         /*
1341                          * Use of some Hop-by-Hop options or some
1342                          * Destination options, might require special
1343                          * privilege.  That is, normal applications
1344                          * (without special privilege) might be forbidden
1345                          * from setting certain options in outgoing packets,
1346                          * and might never see certain options in received
1347                          * packets. [RFC 2292 Section 6]
1348                          * KAME specific note:
1349                          *  KAME prevents non-privileged users from sending or
1350                          *  receiving ANY hbh/dst options in order to avoid
1351                          *  overhead of parsing options in the kernel.
1352                          */
1353                         case IPV6_UNICAST_HOPS:
1354                         case IPV6_CHECKSUM:
1355                         case IPV6_FAITH:
1356
1357                         case IPV6_V6ONLY:
1358                                 if (optlen != sizeof(int)) {
1359                                         error = EINVAL;
1360                                         break;
1361                                 }
1362                                 error = sooptcopyin(sopt, &optval,
1363                                         sizeof optval, sizeof optval);
1364                                 if (error)
1365                                         break;
1366                                 switch (optname) {
1367
1368                                 case IPV6_UNICAST_HOPS:
1369                                         if (optval < -1 || optval >= 256)
1370                                                 error = EINVAL;
1371                                         else {
1372                                                 /* -1 = kernel default */
1373                                                 in6p->in6p_hops = optval;
1374
1375                                                 if ((in6p->in6p_vflag &
1376                                                      INP_IPV4) != 0)
1377                                                         in6p->inp_ip_ttl = optval;
1378                                         }
1379                                         break;
1380 #define OPTSET(bit) \
1381 do { \
1382         if (optval) \
1383                 in6p->in6p_flags |= (bit); \
1384         else \
1385                 in6p->in6p_flags &= ~(bit); \
1386 } while (0)
1387 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1388
1389                                 case IPV6_CHECKSUM:
1390                                         in6p->in6p_cksum = optval;
1391                                         break;
1392
1393                                 case IPV6_FAITH:
1394                                         OPTSET(IN6P_FAITH);
1395                                         break;
1396
1397                                 case IPV6_V6ONLY:
1398                                         /*
1399                                          * make setsockopt(IPV6_V6ONLY)
1400                                          * available only prior to bind(2).
1401                                          * see ipng mailing list, Jun 22 2001.
1402                                          */
1403                                         if (in6p->in6p_lport ||
1404                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1405                                         {
1406                                                 error = EINVAL;
1407                                                 break;
1408                                         }
1409                                         OPTSET(IN6P_IPV6_V6ONLY);
1410                                         if (optval)
1411                                                 in6p->in6p_vflag &= ~INP_IPV4;
1412                                         else
1413                                                 in6p->in6p_vflag |= INP_IPV4;
1414                                         break;
1415                                 }
1416                                 break;
1417
1418                         case IPV6_PKTINFO:
1419                         case IPV6_HOPLIMIT:
1420                         case IPV6_HOPOPTS:
1421                         case IPV6_DSTOPTS:
1422                         case IPV6_RTHDR:
1423                                 /* RFC 2292 */
1424                                 if (optlen != sizeof(int)) {
1425                                         error = EINVAL;
1426                                         break;
1427                                 }
1428                                 error = sooptcopyin(sopt, &optval,
1429                                         sizeof optval, sizeof optval);
1430                                 if (error)
1431                                         break;
1432                                 switch (optname) {
1433                                 case IPV6_PKTINFO:
1434                                         OPTSET(IN6P_PKTINFO);
1435                                         break;
1436                                 case IPV6_HOPLIMIT:
1437                                         OPTSET(IN6P_HOPLIMIT);
1438                                         break;
1439                                 case IPV6_HOPOPTS:
1440                                         /*
1441                                          * Check super-user privilege.
1442                                          * See comments for IPV6_RECVHOPOPTS.
1443                                          */
1444                                         if (!privileged)
1445                                                 return(EPERM);
1446                                         OPTSET(IN6P_HOPOPTS);
1447                                         break;
1448                                 case IPV6_DSTOPTS:
1449                                         if (!privileged)
1450                                                 return(EPERM);
1451                                         OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1452                                         break;
1453                                 case IPV6_RTHDR:
1454                                         OPTSET(IN6P_RTHDR);
1455                                         break;
1456                                 }
1457                                 break;
1458 #undef OPTSET
1459
1460                         case IPV6_MULTICAST_IF:
1461                         case IPV6_MULTICAST_HOPS:
1462                         case IPV6_MULTICAST_LOOP:
1463                         case IPV6_JOIN_GROUP:
1464                         case IPV6_LEAVE_GROUP:
1465                             {
1466                                 struct mbuf *m;
1467                                 if (sopt->sopt_valsize > MLEN) {
1468                                         error = EMSGSIZE;
1469                                         break;
1470                                 }
1471                                 /* XXX */
1472                                 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_HEADER);
1473                                 if (m == 0) {
1474                                         error = ENOBUFS;
1475                                         break;
1476                                 }
1477                                 m->m_len = sopt->sopt_valsize;
1478                                 error = sooptcopyin(sopt, mtod(m, char *),
1479                                                     m->m_len, m->m_len);
1480                                 error = ip6_setmoptions(sopt->sopt_name,
1481                                                         &in6p->in6p_moptions,
1482                                                         m);
1483                                 (void)m_free(m);
1484                             }
1485                                 break;
1486
1487                         case IPV6_PORTRANGE:
1488                                 error = sooptcopyin(sopt, &optval,
1489                                     sizeof optval, sizeof optval);
1490                                 if (error)
1491                                         break;
1492
1493                                 switch (optval) {
1494                                 case IPV6_PORTRANGE_DEFAULT:
1495                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1496                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1497                                         break;
1498
1499                                 case IPV6_PORTRANGE_HIGH:
1500                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1501                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1502                                         break;
1503
1504                                 case IPV6_PORTRANGE_LOW:
1505                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1506                                         in6p->in6p_flags |= IN6P_LOWPORT;
1507                                         break;
1508
1509                                 default:
1510                                         error = EINVAL;
1511                                         break;
1512                                 }
1513                                 break;
1514
1515 #if defined(IPSEC) || defined(FAST_IPSEC)
1516                         case IPV6_IPSEC_POLICY:
1517                             {
1518                                 caddr_t req = NULL;
1519                                 size_t len = 0;
1520                                 struct mbuf *m;
1521
1522                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1523                                         break;
1524                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1525                                         break;
1526                                 if (m) {
1527                                         req = mtod(m, caddr_t);
1528                                         len = m->m_len;
1529                                 }
1530                                 error = ipsec6_set_policy(in6p, optname, req,
1531                                                           len, privileged);
1532                                 m_freem(m);
1533                             }
1534                                 break;
1535 #endif /* KAME IPSEC */
1536
1537                         case IPV6_FW_ADD:
1538                         case IPV6_FW_DEL:
1539                         case IPV6_FW_FLUSH:
1540                         case IPV6_FW_ZERO:
1541                             {
1542                                 struct mbuf *m;
1543                                 struct mbuf **mp = &m;
1544
1545                                 if (ip6_fw_ctl_ptr == NULL)
1546                                         return EINVAL;
1547                                 /* XXX */
1548                                 if ((error = soopt_getm(sopt, &m)) != 0)
1549                                         break;
1550                                 /* XXX */
1551                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
1552                                         break;
1553                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1554                                 m = *mp;
1555                             }
1556                                 break;
1557
1558                         default:
1559                                 error = ENOPROTOOPT;
1560                                 break;
1561                         }
1562                         break;
1563
1564                 case SOPT_GET:
1565                         switch (optname) {
1566
1567                         case IPV6_PKTOPTIONS:
1568                                 if (in6p->in6p_options) {
1569                                         struct mbuf *m;
1570                                         m = m_copym(in6p->in6p_options,
1571                                             0, M_COPYALL, M_WAIT);
1572                                         error = soopt_mcopyout(sopt, m);
1573                                         if (error == 0)
1574                                                 m_freem(m);
1575                                 } else
1576                                         sopt->sopt_valsize = 0;
1577                                 break;
1578
1579                         case IPV6_UNICAST_HOPS:
1580                         case IPV6_CHECKSUM:
1581
1582                         case IPV6_FAITH:
1583                         case IPV6_V6ONLY:
1584                         case IPV6_PORTRANGE:
1585                                 switch (optname) {
1586
1587                                 case IPV6_UNICAST_HOPS:
1588                                         optval = in6p->in6p_hops;
1589                                         break;
1590
1591                                 case IPV6_CHECKSUM:
1592                                         optval = in6p->in6p_cksum;
1593                                         break;
1594
1595                                 case IPV6_FAITH:
1596                                         optval = OPTBIT(IN6P_FAITH);
1597                                         break;
1598
1599                                 case IPV6_V6ONLY:
1600                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1601                                         break;
1602
1603                                 case IPV6_PORTRANGE:
1604                                     {
1605                                         int flags;
1606                                         flags = in6p->in6p_flags;
1607                                         if (flags & IN6P_HIGHPORT)
1608                                                 optval = IPV6_PORTRANGE_HIGH;
1609                                         else if (flags & IN6P_LOWPORT)
1610                                                 optval = IPV6_PORTRANGE_LOW;
1611                                         else
1612                                                 optval = 0;
1613                                         break;
1614                                     }
1615                                 }
1616                                 error = sooptcopyout(sopt, &optval,
1617                                         sizeof optval);
1618                                 break;
1619
1620                         case IPV6_PKTINFO:
1621                         case IPV6_HOPLIMIT:
1622                         case IPV6_HOPOPTS:
1623                         case IPV6_RTHDR:
1624                         case IPV6_DSTOPTS:
1625                                 if (optname == IPV6_HOPOPTS ||
1626                                     optname == IPV6_DSTOPTS ||
1627                                     !privileged)
1628                                         return(EPERM);
1629                                 switch (optname) {
1630                                 case IPV6_PKTINFO:
1631                                         optval = OPTBIT(IN6P_PKTINFO);
1632                                         break;
1633                                 case IPV6_HOPLIMIT:
1634                                         optval = OPTBIT(IN6P_HOPLIMIT);
1635                                         break;
1636                                 case IPV6_HOPOPTS:
1637                                         if (!privileged)
1638                                                 return(EPERM);
1639                                         optval = OPTBIT(IN6P_HOPOPTS);
1640                                         break;
1641                                 case IPV6_RTHDR:
1642                                         optval = OPTBIT(IN6P_RTHDR);
1643                                         break;
1644                                 case IPV6_DSTOPTS:
1645                                         if (!privileged)
1646                                                 return(EPERM);
1647                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1648                                         break;
1649                                 }
1650                                 error = sooptcopyout(sopt, &optval,
1651                                         sizeof optval);
1652                                 break;
1653
1654                         case IPV6_MULTICAST_IF:
1655                         case IPV6_MULTICAST_HOPS:
1656                         case IPV6_MULTICAST_LOOP:
1657                         case IPV6_JOIN_GROUP:
1658                         case IPV6_LEAVE_GROUP:
1659                             {
1660                                 struct mbuf *m;
1661                                 error = ip6_getmoptions(sopt->sopt_name,
1662                                                 in6p->in6p_moptions, &m);
1663                                 if (error == 0)
1664                                         error = sooptcopyout(sopt,
1665                                                 mtod(m, char *), m->m_len);
1666                                 m_freem(m);
1667                             }
1668                                 break;
1669
1670 #if defined(IPSEC) || defined(FAST_IPSEC)
1671                         case IPV6_IPSEC_POLICY:
1672                           {
1673                                 caddr_t req = NULL;
1674                                 size_t len = 0;
1675                                 struct mbuf *m = NULL;
1676                                 struct mbuf **mp = &m;
1677
1678                                 error = soopt_getm(sopt, &m); /* XXX */
1679                                 if (error != NULL)
1680                                         break;
1681                                 error = soopt_mcopyin(sopt, m); /* XXX */
1682                                 if (error != NULL)
1683                                         break;
1684                                 if (m) {
1685                                         req = mtod(m, caddr_t);
1686                                         len = m->m_len;
1687                                 }
1688                                 error = ipsec6_get_policy(in6p, req, len, mp);
1689                                 if (error == 0)
1690                                         error = soopt_mcopyout(sopt, m); /*XXX*/
1691                                 if (error == 0 && m)
1692                                         m_freem(m);
1693                                 break;
1694                           }
1695 #endif /* KAME IPSEC */
1696
1697                         case IPV6_FW_GET:
1698                           {
1699                                 struct mbuf *m;
1700                                 struct mbuf **mp = &m;
1701
1702                                 if (ip6_fw_ctl_ptr == NULL)
1703                                 {
1704                                         return EINVAL;
1705                                 }
1706                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1707                                 if (error == 0)
1708                                         error = soopt_mcopyout(sopt, m); /* XXX */
1709                                 if (error == 0 && m)
1710                                         m_freem(m);
1711                           }
1712                                 break;
1713
1714                         default:
1715                                 error = ENOPROTOOPT;
1716                                 break;
1717                         }
1718                         break;
1719                 }
1720         } else {
1721                 error = EINVAL;
1722         }
1723         return(error);
1724 }
1725
1726 /*
1727  * Set up IP6 options in pcb for insertion in output packets or
1728  * specifying behavior of outgoing packets.
1729  */
1730 static int
1731 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
1732             struct sockopt *sopt)
1733 {
1734         struct ip6_pktopts *opt = *pktopt;
1735         int error = 0;
1736         struct thread *td = sopt->sopt_td;
1737         int priv = 0;
1738
1739         /* turn off any old options. */
1740         if (opt) {
1741 #ifdef DIAGNOSTIC
1742                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1743                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1744                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1745                         printf("ip6_pcbopts: all specified options are cleared.\n");
1746 #endif
1747                 ip6_clearpktopts(opt, 1, -1);
1748         } else
1749                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1750         *pktopt = NULL;
1751
1752         if (!m || m->m_len == 0) {
1753                 /*
1754                  * Only turning off any previous options, regardless of
1755                  * whether the opt is just created or given.
1756                  */
1757                 free(opt, M_IP6OPT);
1758                 return(0);
1759         }
1760
1761         /*  set options specified by user. */
1762         if (suser(td) == 0)
1763                 priv = 1;
1764         if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1765                 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1766                 free(opt, M_IP6OPT);
1767                 return(error);
1768         }
1769         *pktopt = opt;
1770         return(0);
1771 }
1772
1773 /*
1774  * initialize ip6_pktopts.  beware that there are non-zero default values in
1775  * the struct.
1776  */
1777 void
1778 init_ip6pktopts(struct ip6_pktopts *opt)
1779 {
1780
1781         bzero(opt, sizeof(*opt));
1782         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
1783 }
1784
1785 void
1786 ip6_clearpktopts(struct ip6_pktopts *pktopt, int needfree, int optname)
1787 {
1788         if (pktopt == NULL)
1789                 return;
1790
1791         if (optname == -1) {
1792                 if (needfree && pktopt->ip6po_pktinfo)
1793                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
1794                 pktopt->ip6po_pktinfo = NULL;
1795         }
1796         if (optname == -1)
1797                 pktopt->ip6po_hlim = -1;
1798         if (optname == -1) {
1799                 if (needfree && pktopt->ip6po_nexthop)
1800                         free(pktopt->ip6po_nexthop, M_IP6OPT);
1801                 pktopt->ip6po_nexthop = NULL;
1802         }
1803         if (optname == -1) {
1804                 if (needfree && pktopt->ip6po_hbh)
1805                         free(pktopt->ip6po_hbh, M_IP6OPT);
1806                 pktopt->ip6po_hbh = NULL;
1807         }
1808         if (optname == -1) {
1809                 if (needfree && pktopt->ip6po_dest1)
1810                         free(pktopt->ip6po_dest1, M_IP6OPT);
1811                 pktopt->ip6po_dest1 = NULL;
1812         }
1813         if (optname == -1) {
1814                 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1815                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1816                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1817                 if (pktopt->ip6po_route.ro_rt) {
1818                         RTFREE(pktopt->ip6po_route.ro_rt);
1819                         pktopt->ip6po_route.ro_rt = NULL;
1820                 }
1821         }
1822         if (optname == -1) {
1823                 if (needfree && pktopt->ip6po_dest2)
1824                         free(pktopt->ip6po_dest2, M_IP6OPT);
1825                 pktopt->ip6po_dest2 = NULL;
1826         }
1827 }
1828
1829 #define PKTOPT_EXTHDRCPY(type) \
1830 do {\
1831         if (src->type) {\
1832                 int hlen =\
1833                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1834                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
1835                 if (dst->type == NULL && canwait == M_NOWAIT)\
1836                         goto bad;\
1837                 bcopy(src->type, dst->type, hlen);\
1838         }\
1839 } while (0)
1840
1841 struct ip6_pktopts *
1842 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
1843 {
1844         struct ip6_pktopts *dst;
1845
1846         if (src == NULL) {
1847                 printf("ip6_clearpktopts: invalid argument\n");
1848                 return(NULL);
1849         }
1850
1851         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1852         if (dst == NULL && canwait == M_NOWAIT)
1853                 return (NULL);
1854         bzero(dst, sizeof(*dst));
1855
1856         dst->ip6po_hlim = src->ip6po_hlim;
1857         if (src->ip6po_pktinfo) {
1858                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1859                                             M_IP6OPT, canwait);
1860                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1861                         goto bad;
1862                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1863         }
1864         if (src->ip6po_nexthop) {
1865                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1866                                             M_IP6OPT, canwait);
1867                 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1868                         goto bad;
1869                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1870                       src->ip6po_nexthop->sa_len);
1871         }
1872         PKTOPT_EXTHDRCPY(ip6po_hbh);
1873         PKTOPT_EXTHDRCPY(ip6po_dest1);
1874         PKTOPT_EXTHDRCPY(ip6po_dest2);
1875         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1876         return(dst);
1877
1878   bad:
1879         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1880         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1881         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1882         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1883         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1884         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1885         free(dst, M_IP6OPT);
1886         return(NULL);
1887 }
1888 #undef PKTOPT_EXTHDRCPY
1889
1890 void
1891 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1892 {
1893         if (pktopt == NULL)
1894                 return;
1895
1896         ip6_clearpktopts(pktopt, 1, -1);
1897
1898         free(pktopt, M_IP6OPT);
1899 }
1900
1901 /*
1902  * Set the IP6 multicast options in response to user setsockopt().
1903  */
1904 static int
1905 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1906 {
1907         int error = 0;
1908         u_int loop, ifindex;
1909         struct ipv6_mreq *mreq;
1910         struct ifnet *ifp;
1911         struct ip6_moptions *im6o = *im6op;
1912         struct route_in6 ro;
1913         struct sockaddr_in6 *dst;
1914         struct in6_multi_mship *imm;
1915         struct thread *td = curthread;  /* XXX */
1916
1917         if (im6o == NULL) {
1918                 /*
1919                  * No multicast option buffer attached to the pcb;
1920                  * allocate one and initialize to default values.
1921                  */
1922                 im6o = (struct ip6_moptions *)
1923                         malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1924
1925                 if (im6o == NULL)
1926                         return(ENOBUFS);
1927                 *im6op = im6o;
1928                 im6o->im6o_multicast_ifp = NULL;
1929                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1930                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1931                 LIST_INIT(&im6o->im6o_memberships);
1932         }
1933
1934         switch (optname) {
1935
1936         case IPV6_MULTICAST_IF:
1937                 /*
1938                  * Select the interface for outgoing multicast packets.
1939                  */
1940                 if (m == NULL || m->m_len != sizeof(u_int)) {
1941                         error = EINVAL;
1942                         break;
1943                 }
1944                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1945                 if (ifindex < 0 || if_index < ifindex) {
1946                         error = ENXIO;  /* XXX EINVAL? */
1947                         break;
1948                 }
1949                 ifp = ifindex2ifnet[ifindex];
1950                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1951                         error = EADDRNOTAVAIL;
1952                         break;
1953                 }
1954                 im6o->im6o_multicast_ifp = ifp;
1955                 break;
1956
1957         case IPV6_MULTICAST_HOPS:
1958             {
1959                 /*
1960                  * Set the IP6 hoplimit for outgoing multicast packets.
1961                  */
1962                 int optval;
1963                 if (m == NULL || m->m_len != sizeof(int)) {
1964                         error = EINVAL;
1965                         break;
1966                 }
1967                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1968                 if (optval < -1 || optval >= 256)
1969                         error = EINVAL;
1970                 else if (optval == -1)
1971                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1972                 else
1973                         im6o->im6o_multicast_hlim = optval;
1974                 break;
1975             }
1976
1977         case IPV6_MULTICAST_LOOP:
1978                 /*
1979                  * Set the loopback flag for outgoing multicast packets.
1980                  * Must be zero or one.
1981                  */
1982                 if (m == NULL || m->m_len != sizeof(u_int)) {
1983                         error = EINVAL;
1984                         break;
1985                 }
1986                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1987                 if (loop > 1) {
1988                         error = EINVAL;
1989                         break;
1990                 }
1991                 im6o->im6o_multicast_loop = loop;
1992                 break;
1993
1994         case IPV6_JOIN_GROUP:
1995                 /*
1996                  * Add a multicast group membership.
1997                  * Group must be a valid IP6 multicast address.
1998                  */
1999                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2000                         error = EINVAL;
2001                         break;
2002                 }
2003                 mreq = mtod(m, struct ipv6_mreq *);
2004                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2005                         /*
2006                          * We use the unspecified address to specify to accept
2007                          * all multicast addresses. Only super user is allowed
2008                          * to do this.
2009                          */
2010                         if (suser(td))
2011                         {
2012                                 error = EACCES;
2013                                 break;
2014                         }
2015                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2016                         error = EINVAL;
2017                         break;
2018                 }
2019
2020                 /*
2021                  * If the interface is specified, validate it.
2022                  */
2023                 if (mreq->ipv6mr_interface < 0
2024                  || if_index < mreq->ipv6mr_interface) {
2025                         error = ENXIO;  /* XXX EINVAL? */
2026                         break;
2027                 }
2028                 /*
2029                  * If no interface was explicitly specified, choose an
2030                  * appropriate one according to the given multicast address.
2031                  */
2032                 if (mreq->ipv6mr_interface == 0) {
2033                         /*
2034                          * If the multicast address is in node-local scope,
2035                          * the interface should be a loopback interface.
2036                          * Otherwise, look up the routing table for the
2037                          * address, and choose the outgoing interface.
2038                          *   XXX: is it a good approach?
2039                          */
2040                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2041                                 ifp = &loif[0];
2042                         } else {
2043                                 ro.ro_rt = NULL;
2044                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2045                                 bzero(dst, sizeof(*dst));
2046                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2047                                 dst->sin6_family = AF_INET6;
2048                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2049                                 rtalloc((struct route *)&ro);
2050                                 if (ro.ro_rt == NULL) {
2051                                         error = EADDRNOTAVAIL;
2052                                         break;
2053                                 }
2054                                 ifp = ro.ro_rt->rt_ifp;
2055                                 rtfree(ro.ro_rt);
2056                         }
2057                 } else
2058                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2059
2060                 /*
2061                  * See if we found an interface, and confirm that it
2062                  * supports multicast
2063                  */
2064                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2065                         error = EADDRNOTAVAIL;
2066                         break;
2067                 }
2068                 /*
2069                  * Put interface index into the multicast address,
2070                  * if the address has link-local scope.
2071                  */
2072                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2073                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2074                                 = htons(mreq->ipv6mr_interface);
2075                 }
2076                 /*
2077                  * See if the membership already exists.
2078                  */
2079                 for (imm = im6o->im6o_memberships.lh_first;
2080                      imm != NULL; imm = imm->i6mm_chain.le_next)
2081                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2082                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2083                                                &mreq->ipv6mr_multiaddr))
2084                                 break;
2085                 if (imm != NULL) {
2086                         error = EADDRINUSE;
2087                         break;
2088                 }
2089                 /*
2090                  * Everything looks good; add a new record to the multicast
2091                  * address list for the given interface.
2092                  */
2093                 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2094                 if (imm == NULL) {
2095                         error = ENOBUFS;
2096                         break;
2097                 }
2098                 if ((imm->i6mm_maddr =
2099                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2100                         free(imm, M_IPMADDR);
2101                         break;
2102                 }
2103                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2104                 break;
2105
2106         case IPV6_LEAVE_GROUP:
2107                 /*
2108                  * Drop a multicast group membership.
2109                  * Group must be a valid IP6 multicast address.
2110                  */
2111                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2112                         error = EINVAL;
2113                         break;
2114                 }
2115                 mreq = mtod(m, struct ipv6_mreq *);
2116                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2117                         if (suser(td)) {
2118                                 error = EACCES;
2119                                 break;
2120                         }
2121                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2122                         error = EINVAL;
2123                         break;
2124                 }
2125                 /*
2126                  * If an interface address was specified, get a pointer
2127                  * to its ifnet structure.
2128                  */
2129                 if (mreq->ipv6mr_interface < 0
2130                  || if_index < mreq->ipv6mr_interface) {
2131                         error = ENXIO;  /* XXX EINVAL? */
2132                         break;
2133                 }
2134                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2135                 /*
2136                  * Put interface index into the multicast address,
2137                  * if the address has link-local scope.
2138                  */
2139                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2140                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2141                                 = htons(mreq->ipv6mr_interface);
2142                 }
2143                 /*
2144                  * Find the membership in the membership list.
2145                  */
2146                 for (imm = im6o->im6o_memberships.lh_first;
2147                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2148                         if ((ifp == NULL ||
2149                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2150                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2151                                                &mreq->ipv6mr_multiaddr))
2152                                 break;
2153                 }
2154                 if (imm == NULL) {
2155                         /* Unable to resolve interface */
2156                         error = EADDRNOTAVAIL;
2157                         break;
2158                 }
2159                 /*
2160                  * Give up the multicast address record to which the
2161                  * membership points.
2162                  */
2163                 LIST_REMOVE(imm, i6mm_chain);
2164                 in6_delmulti(imm->i6mm_maddr);
2165                 free(imm, M_IPMADDR);
2166                 break;
2167
2168         default:
2169                 error = EOPNOTSUPP;
2170                 break;
2171         }
2172
2173         /*
2174          * If all options have default values, no need to keep the mbuf.
2175          */
2176         if (im6o->im6o_multicast_ifp == NULL &&
2177             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2178             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2179             im6o->im6o_memberships.lh_first == NULL) {
2180                 free(*im6op, M_IPMOPTS);
2181                 *im6op = NULL;
2182         }
2183
2184         return(error);
2185 }
2186
2187 /*
2188  * Return the IP6 multicast options in response to user getsockopt().
2189  */
2190 static int
2191 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2192 {
2193         u_int *hlim, *loop, *ifindex;
2194
2195         *mp = m_get(M_WAIT, MT_HEADER);         /* XXX */
2196
2197         switch (optname) {
2198
2199         case IPV6_MULTICAST_IF:
2200                 ifindex = mtod(*mp, u_int *);
2201                 (*mp)->m_len = sizeof(u_int);
2202                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2203                         *ifindex = 0;
2204                 else
2205                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2206                 return(0);
2207
2208         case IPV6_MULTICAST_HOPS:
2209                 hlim = mtod(*mp, u_int *);
2210                 (*mp)->m_len = sizeof(u_int);
2211                 if (im6o == NULL)
2212                         *hlim = ip6_defmcasthlim;
2213                 else
2214                         *hlim = im6o->im6o_multicast_hlim;
2215                 return(0);
2216
2217         case IPV6_MULTICAST_LOOP:
2218                 loop = mtod(*mp, u_int *);
2219                 (*mp)->m_len = sizeof(u_int);
2220                 if (im6o == NULL)
2221                         *loop = ip6_defmcasthlim;
2222                 else
2223                         *loop = im6o->im6o_multicast_loop;
2224                 return(0);
2225
2226         default:
2227                 return(EOPNOTSUPP);
2228         }
2229 }
2230
2231 /*
2232  * Discard the IP6 multicast options.
2233  */
2234 void
2235 ip6_freemoptions(struct ip6_moptions *im6o)
2236 {
2237         struct in6_multi_mship *imm;
2238
2239         if (im6o == NULL)
2240                 return;
2241
2242         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2243                 LIST_REMOVE(imm, i6mm_chain);
2244                 if (imm->i6mm_maddr)
2245                         in6_delmulti(imm->i6mm_maddr);
2246                 free(imm, M_IPMADDR);
2247         }
2248         free(im6o, M_IPMOPTS);
2249 }
2250
2251 /*
2252  * Set IPv6 outgoing packet options based on advanced API.
2253  */
2254 int
2255 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, int priv,
2256                   int needcopy)
2257 {
2258         struct cmsghdr *cm = 0;
2259
2260         if (control == 0 || opt == 0)
2261                 return(EINVAL);
2262
2263         init_ip6pktopts(opt);
2264
2265         /*
2266          * XXX: Currently, we assume all the optional information is stored
2267          * in a single mbuf.
2268          */
2269         if (control->m_next)
2270                 return(EINVAL);
2271
2272         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2273                      control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2274                 cm = mtod(control, struct cmsghdr *);
2275                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2276                         return(EINVAL);
2277                 if (cm->cmsg_level != IPPROTO_IPV6)
2278                         continue;
2279
2280                 /*
2281                  * XXX should check if RFC2292 API is mixed with 2292bis API
2282                  */
2283                 switch (cm->cmsg_type) {
2284                 case IPV6_PKTINFO:
2285                         if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2286                                 return(EINVAL);
2287                         if (needcopy) {
2288                                 /* XXX: Is it really WAITOK? */
2289                                 opt->ip6po_pktinfo =
2290                                         malloc(sizeof(struct in6_pktinfo),
2291                                                M_IP6OPT, M_WAITOK);
2292                                 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2293                                     sizeof(struct in6_pktinfo));
2294                         } else
2295                                 opt->ip6po_pktinfo =
2296                                         (struct in6_pktinfo *)CMSG_DATA(cm);
2297                         if (opt->ip6po_pktinfo->ipi6_ifindex &&
2298                             IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2299                                 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2300                                         htons(opt->ip6po_pktinfo->ipi6_ifindex);
2301
2302                         if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2303                          || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2304                                 return(ENXIO);
2305                         }
2306
2307                         /*
2308                          * Check if the requested source address is indeed a
2309                          * unicast address assigned to the node, and can be
2310                          * used as the packet's source address.
2311                          */
2312                         if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2313                                 struct in6_ifaddr *ia6;
2314                                 struct sockaddr_in6 sin6;
2315
2316                                 bzero(&sin6, sizeof(sin6));
2317                                 sin6.sin6_len = sizeof(sin6);
2318                                 sin6.sin6_family = AF_INET6;
2319                                 sin6.sin6_addr =
2320                                         opt->ip6po_pktinfo->ipi6_addr;
2321                                 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2322                                 if (ia6 == NULL ||
2323                                     (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2324                                                        IN6_IFF_NOTREADY)) != 0)
2325                                         return(EADDRNOTAVAIL);
2326                         }
2327                         break;
2328
2329                 case IPV6_HOPLIMIT:
2330                         if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2331                                 return(EINVAL);
2332
2333                         opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2334                         if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2335                                 return(EINVAL);
2336                         break;
2337
2338                 case IPV6_NEXTHOP:
2339                         if (!priv)
2340                                 return(EPERM);
2341
2342                         if (cm->cmsg_len < sizeof(u_char) ||
2343                             /* check if cmsg_len is large enough for sa_len */
2344                             cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2345                                 return(EINVAL);
2346
2347                         if (needcopy) {
2348                                 opt->ip6po_nexthop =
2349                                         malloc(*CMSG_DATA(cm),
2350                                                M_IP6OPT, M_WAITOK);
2351                                 bcopy(CMSG_DATA(cm),
2352                                       opt->ip6po_nexthop,
2353                                       *CMSG_DATA(cm));
2354                         } else
2355                                 opt->ip6po_nexthop =
2356                                         (struct sockaddr *)CMSG_DATA(cm);
2357                         break;
2358
2359                 case IPV6_HOPOPTS:
2360                 {
2361                         struct ip6_hbh *hbh;
2362                         int hbhlen;
2363
2364                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2365                                 return(EINVAL);
2366                         hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2367                         hbhlen = (hbh->ip6h_len + 1) << 3;
2368                         if (cm->cmsg_len != CMSG_LEN(hbhlen))
2369                                 return(EINVAL);
2370
2371                         if (needcopy) {
2372                                 opt->ip6po_hbh =
2373                                         malloc(hbhlen, M_IP6OPT, M_WAITOK);
2374                                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2375                         } else
2376                                 opt->ip6po_hbh = hbh;
2377                         break;
2378                 }
2379
2380                 case IPV6_DSTOPTS:
2381                 {
2382                         struct ip6_dest *dest, **newdest;
2383                         int destlen;
2384
2385                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2386                                 return(EINVAL);
2387                         dest = (struct ip6_dest *)CMSG_DATA(cm);
2388                         destlen = (dest->ip6d_len + 1) << 3;
2389                         if (cm->cmsg_len != CMSG_LEN(destlen))
2390                                 return(EINVAL);
2391
2392                         /*
2393                          * The old advacned API is ambiguous on this
2394                          * point. Our approach is to determine the
2395                          * position based according to the existence
2396                          * of a routing header. Note, however, that
2397                          * this depends on the order of the extension
2398                          * headers in the ancillary data; the 1st part
2399                          * of the destination options header must
2400                          * appear before the routing header in the
2401                          * ancillary data, too.
2402                          * RFC2292bis solved the ambiguity by
2403                          * introducing separate cmsg types.
2404                          */
2405                         if (opt->ip6po_rthdr == NULL)
2406                                 newdest = &opt->ip6po_dest1;
2407                         else
2408                                 newdest = &opt->ip6po_dest2;
2409
2410                         if (needcopy) {
2411                                 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2412                                 bcopy(dest, *newdest, destlen);
2413                         } else
2414                                 *newdest = dest;
2415
2416                         break;
2417                 }
2418
2419                 case IPV6_RTHDR:
2420                 {
2421                         struct ip6_rthdr *rth;
2422                         int rthlen;
2423
2424                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2425                                 return(EINVAL);
2426                         rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2427                         rthlen = (rth->ip6r_len + 1) << 3;
2428                         if (cm->cmsg_len != CMSG_LEN(rthlen))
2429                                 return(EINVAL);
2430
2431                         switch (rth->ip6r_type) {
2432                         case IPV6_RTHDR_TYPE_0:
2433                                 /* must contain one addr */
2434                                 if (rth->ip6r_len == 0)
2435                                         return(EINVAL);
2436                                 /* length must be even */
2437                                 if (rth->ip6r_len % 2)
2438                                         return(EINVAL);
2439                                 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2440                                         return(EINVAL);
2441                                 break;
2442                         default:
2443                                 return(EINVAL); /* not supported */
2444                         }
2445
2446                         if (needcopy) {
2447                                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2448                                                           M_WAITOK);
2449                                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2450                         } else
2451                                 opt->ip6po_rthdr = rth;
2452
2453                         break;
2454                 }
2455
2456                 default:
2457                         return(ENOPROTOOPT);
2458                 }
2459         }
2460
2461         return(0);
2462 }
2463
2464 /*
2465  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2466  * packet to the input queue of a specified interface.  Note that this
2467  * calls the output routine of the loopback "driver", but with an interface
2468  * pointer that might NOT be &loif -- easier than replicating that code here.
2469  */
2470 void
2471 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2472 {
2473         struct mbuf *copym;
2474         struct ip6_hdr *ip6;
2475
2476         copym = m_copy(m, 0, M_COPYALL);
2477         if (copym == NULL)
2478                 return;
2479
2480         /*
2481          * Make sure to deep-copy IPv6 header portion in case the data
2482          * is in an mbuf cluster, so that we can safely override the IPv6
2483          * header portion later.
2484          */
2485         if ((copym->m_flags & M_EXT) != 0 ||
2486             copym->m_len < sizeof(struct ip6_hdr)) {
2487                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2488                 if (copym == NULL)
2489                         return;
2490         }
2491
2492 #ifdef DIAGNOSTIC
2493         if (copym->m_len < sizeof(*ip6)) {
2494                 m_freem(copym);
2495                 return;
2496         }
2497 #endif
2498
2499         ip6 = mtod(copym, struct ip6_hdr *);
2500 #ifndef SCOPEDROUTING
2501         /*
2502          * clear embedded scope identifiers if necessary.
2503          * in6_clearscope will touch the addresses only when necessary.
2504          */
2505         in6_clearscope(&ip6->ip6_src);
2506         in6_clearscope(&ip6->ip6_dst);
2507 #endif
2508
2509         (void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2510 }
2511
2512 /*
2513  * Chop IPv6 header off from the payload.
2514  */
2515 static int
2516 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2517 {
2518         struct mbuf *mh;
2519         struct ip6_hdr *ip6;
2520
2521         ip6 = mtod(m, struct ip6_hdr *);
2522         if (m->m_len > sizeof(*ip6)) {
2523                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2524                 if (mh == 0) {
2525                         m_freem(m);
2526                         return ENOBUFS;
2527                 }
2528                 M_MOVE_PKTHDR(mh, m);
2529                 MH_ALIGN(mh, sizeof(*ip6));
2530                 m->m_len -= sizeof(*ip6);
2531                 m->m_data += sizeof(*ip6);
2532                 mh->m_next = m;
2533                 m = mh;
2534                 m->m_len = sizeof(*ip6);
2535                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2536         }
2537         exthdrs->ip6e_ip6 = m;
2538         return 0;
2539 }
2540
2541 /*
2542  * Compute IPv6 extension header length.
2543  */
2544 int
2545 ip6_optlen(struct in6pcb *in6p)
2546 {
2547         int len;
2548
2549         if (!in6p->in6p_outputopts)
2550                 return 0;
2551
2552         len = 0;
2553 #define elen(x) \
2554     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2555
2556         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2557         if (in6p->in6p_outputopts->ip6po_rthdr)
2558                 /* dest1 is valid with rthdr only */
2559                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2560         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2561         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2562         return len;
2563 #undef elen
2564 }