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 1.16 2005/01/06 17:59:32 hsu 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
  74 #include <sys/param.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/protosw.h>
  79 #include <sys/socket.h>
  80 #include <sys/socketvar.h>
  81 #include <sys/systm.h>
  82 #include <sys/kernel.h>
  83 #include <sys/proc.h>
  84
  85 #include <net/if.h>
  86 #include <net/route.h>
  87 #include <net/pfil.h>
  88
  89 #include <netinet/in.h>
  90 #include <netinet/in_var.h>
  91 #include <netinet6/in6_var.h>
  92 #include <netinet/ip6.h>
  93 #include <netinet/icmp6.h>
  94 #include <netinet6/ip6_var.h>
  95 #include <netinet/in_pcb.h>
  96 #include <netinet6/nd6.h>
  97
  98 #ifdef IPSEC
  99 #include <netinet6/ipsec.h>
 100 #ifdef INET6
 101 #include <netinet6/ipsec6.h>
 102 #endif
 103 #include <netproto/key/key.h>
 104 #endif /* IPSEC */
 105
 106 #ifdef FAST_IPSEC
 107 #include <netproto/ipsec/ipsec.h>
 108 #include <netproto/ipsec/ipsec6.h>
 109 #include <netproto/ipsec/key.h>
 110 #endif /* FAST_IPSEC */
 111
 112 #include <net/ip6fw/ip6_fw.h>
 113
 114 #include <net/net_osdep.h>
 115
 116 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 117
 118 struct ip6_exthdrs {
 119         struct mbuf *ip6e_ip6;
 120         struct mbuf *ip6e_hbh;
 121         struct mbuf *ip6e_dest1;
 122         struct mbuf *ip6e_rthdr;
 123         struct mbuf *ip6e_dest2;
 124 };
 125
 126 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
 127                             struct socket *, struct sockopt *sopt);
 128 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 129 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 130 static int ip6_copyexthdr (struct mbuf **, caddr_t, int);
 131 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
 132                                   struct ip6_frag **);
 133 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 134 static int ip6_splithdr (struct mbuf *, struct ip6_exthdrs *);
 135
 136 /*
 137  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 138  * header (with pri, len, nxt, hlim, src, dst).
 139  * This function may modify ver and hlim only.
 140  * The mbuf chain containing the packet will be freed.
 141  * The mbuf opt, if present, will not be freed.
 142  *
 143  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 144  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 145  * which is rt_rmx.rmx_mtu.
 146  */
 147 int
 148 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 149            int flags, struct ip6_moptions *im6o,
 150            struct ifnet **ifpp,         /* XXX: just for statistics */
 151            struct inpcb *inp)
 152 {
 153         struct ip6_hdr *ip6, *mhip6;
 154         struct ifnet *ifp, *origifp;
 155         struct mbuf *m = m0;
 156         int hlen, tlen, len, off;
 157         struct route_in6 ip6route;
 158         struct sockaddr_in6 *dst;
 159         int error = 0;
 160         struct in6_ifaddr *ia = NULL;
 161         u_long mtu;
 162         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 163         struct ip6_exthdrs exthdrs;
 164         struct in6_addr finaldst;
 165         struct route_in6 *ro_pmtu = NULL;
 166         int hdrsplit = 0;
 167         int needipsec = 0;
 168 #ifdef IPSEC
 169         int needipsectun = 0;
 170         struct secpolicy *sp = NULL;
 171         struct socket *so = inp ? inp->inp_socket : NULL;
 172
 173         ip6 = mtod(m, struct ip6_hdr *);
 174 #endif /* IPSEC */
 175 #ifdef FAST_IPSEC
 176         int needipsectun = 0;
 177         struct secpolicy *sp = NULL;
 178
 179         ip6 = mtod(m, struct ip6_hdr *);
 180 #endif /* FAST_IPSEC */
 181
 182 #define MAKE_EXTHDR(hp, mp)                                             \
 183     do {                                                                \
 184         if (hp) {                                                       \
 185                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 186                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 187                                        ((eh)->ip6e_len + 1) << 3);      \
 188                 if (error)                                              \
 189                         goto freehdrs;                                  \
 190         }                                                               \
 191     } while (0)
 192
 193         bzero(&exthdrs, sizeof(exthdrs));
 194
 195         if (opt) {
 196                 /* Hop-by-Hop options header */
 197                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 198                 /* Destination options header(1st part) */
 199                 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 200                 /* Routing header */
 201                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 202                 /* Destination options header(2nd part) */
 203                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 204         }
 205
 206 #ifdef IPSEC
 207         /* get a security policy for this packet */
 208         if (so == NULL)
 209                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 210         else
 211                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 212
 213         if (sp == NULL) {
 214                 ipsec6stat.out_inval++;
 215                 goto freehdrs;
 216         }
 217
 218         error = 0;
 219
 220         /* check policy */
 221         switch (sp->policy) {
 222         case IPSEC_POLICY_DISCARD:
 223                 /*
 224                  * This packet is just discarded.
 225                  */
 226                 ipsec6stat.out_polvio++;
 227                 goto freehdrs;
 228
 229         case IPSEC_POLICY_BYPASS:
 230         case IPSEC_POLICY_NONE:
 231                 /* no need to do IPsec. */
 232                 needipsec = 0;
 233                 break;
 234
 235         case IPSEC_POLICY_IPSEC:
 236                 if (sp->req == NULL) {
 237                         /* acquire a policy */
 238                         error = key_spdacquire(sp);
 239                         goto freehdrs;
 240                 }
 241                 needipsec = 1;
 242                 break;
 243
 244         case IPSEC_POLICY_ENTRUST:
 245         default:
 246                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 247         }
 248 #endif /* IPSEC */
 249 #ifdef FAST_IPSEC
 250         /* get a security policy for this packet */
 251         if (inp == NULL)
 252                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 253         else
 254                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 255
 256         if (sp == NULL) {
 257                 newipsecstat.ips_out_inval++;
 258                 goto freehdrs;
 259         }
 260
 261         error = 0;
 262
 263         /* check policy */
 264         switch (sp->policy) {
 265         case IPSEC_POLICY_DISCARD:
 266                 /*
 267                  * This packet is just discarded.
 268                  */
 269                 newipsecstat.ips_out_polvio++;
 270                 goto freehdrs;
 271
 272         case IPSEC_POLICY_BYPASS:
 273         case IPSEC_POLICY_NONE:
 274                 /* no need to do IPsec. */
 275                 needipsec = 0;
 276                 break;
 277
 278         case IPSEC_POLICY_IPSEC:
 279                 if (sp->req == NULL) {
 280                         /* acquire a policy */
 281                         error = key_spdacquire(sp);
 282                         goto freehdrs;
 283                 }
 284                 needipsec = 1;
 285                 break;
 286
 287         case IPSEC_POLICY_ENTRUST:
 288         default:
 289                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 290         }
 291 #endif /* FAST_IPSEC */
 292
 293         /*
 294          * Calculate the total length of the extension header chain.
 295          * Keep the length of the unfragmentable part for fragmentation.
 296          */
 297         optlen = 0;
 298         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 299         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 300         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 301         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 302         /* NOTE: we don't add AH/ESP length here. do that later. */
 303         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 304
 305         /*
 306          * If we need IPsec, or there is at least one extension header,
 307          * separate IP6 header from the payload.
 308          */
 309         if ((needipsec || optlen) && !hdrsplit) {
 310                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 311                         m = NULL;
 312                         goto freehdrs;
 313                 }
 314                 m = exthdrs.ip6e_ip6;
 315                 hdrsplit++;
 316         }
 317
 318         /* adjust pointer */
 319         ip6 = mtod(m, struct ip6_hdr *);
 320
 321         /* adjust mbuf packet header length */
 322         m->m_pkthdr.len += optlen;
 323         plen = m->m_pkthdr.len - sizeof(*ip6);
 324
 325         /* If this is a jumbo payload, insert a jumbo payload option. */
 326         if (plen > IPV6_MAXPACKET) {
 327                 if (!hdrsplit) {
 328                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 329                                 m = NULL;
 330                                 goto freehdrs;
 331                         }
 332                         m = exthdrs.ip6e_ip6;
 333                         hdrsplit++;
 334                 }
 335                 /* adjust pointer */
 336                 ip6 = mtod(m, struct ip6_hdr *);
 337                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 338                         goto freehdrs;
 339                 ip6->ip6_plen = 0;
 340         } else
 341                 ip6->ip6_plen = htons(plen);
 342
 343         /*
 344          * Concatenate headers and fill in next header fields.
 345          * Here we have, on "m"
 346          *      IPv6 payload
 347          * and we insert headers accordingly.  Finally, we should be getting:
 348          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 349          *
 350          * during the header composing process, "m" points to IPv6 header.
 351          * "mprev" points to an extension header prior to esp.
 352          */
 353         {
 354                 u_char *nexthdrp = &ip6->ip6_nxt;
 355                 struct mbuf *mprev = m;
 356
 357                 /*
 358                  * we treat dest2 specially.  this makes IPsec processing
 359                  * much easier.  the goal here is to make mprev point the
 360                  * mbuf prior to dest2.
 361                  *
 362                  * result: IPv6 dest2 payload
 363                  * m and mprev will point to IPv6 header.
 364                  */
 365                 if (exthdrs.ip6e_dest2) {
 366                         if (!hdrsplit)
 367                                 panic("assumption failed: hdr not split");
 368                         exthdrs.ip6e_dest2->m_next = m->m_next;
 369                         m->m_next = exthdrs.ip6e_dest2;
 370                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 371                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 372                 }
 373
 374 #define MAKE_CHAIN(m, mp, p, i)\
 375     do {\
 376         if (m) {\
 377                 if (!hdrsplit) \
 378                         panic("assumption failed: hdr not split"); \
 379                 *mtod((m), u_char *) = *(p);\
 380                 *(p) = (i);\
 381                 p = mtod((m), u_char *);\
 382                 (m)->m_next = (mp)->m_next;\
 383                 (mp)->m_next = (m);\
 384                 (mp) = (m);\
 385         }\
 386     } while (0)
 387                 /*
 388                  * result: IPv6 hbh dest1 rthdr dest2 payload
 389                  * m will point to IPv6 header.  mprev will point to the
 390                  * extension header prior to dest2 (rthdr in the above case).
 391                  */
 392                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
 393                            nexthdrp, IPPROTO_HOPOPTS);
 394                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
 395                            nexthdrp, IPPROTO_DSTOPTS);
 396                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
 397                            nexthdrp, IPPROTO_ROUTING);
 398
 399 #if defined(IPSEC) || defined(FAST_IPSEC)
 400                 if (!needipsec)
 401                         goto skip_ipsec2;
 402
 403                 /*
 404                  * pointers after IPsec headers are not valid any more.
 405                  * other pointers need a great care too.
 406                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 407                  */
 408                 exthdrs.ip6e_dest2 = NULL;
 409
 410             {
 411                 struct ip6_rthdr *rh = NULL;
 412                 int segleft_org = 0;
 413                 struct ipsec_output_state state;
 414
 415                 if (exthdrs.ip6e_rthdr) {
 416                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 417                         segleft_org = rh->ip6r_segleft;
 418                         rh->ip6r_segleft = 0;
 419                 }
 420
 421                 bzero(&state, sizeof(state));
 422                 state.m = m;
 423                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 424                         &needipsectun);
 425                 m = state.m;
 426                 if (error) {
 427                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 428                         m = NULL;
 429                         switch (error) {
 430                         case EHOSTUNREACH:
 431                         case ENETUNREACH:
 432                         case EMSGSIZE:
 433                         case ENOBUFS:
 434                         case ENOMEM:
 435                                 break;
 436                         default:
 437                                 printf("ip6_output (ipsec): error code %d\n", error);
 438                                 /* fall through */
 439                         case ENOENT:
 440                                 /* don't show these error codes to the user */
 441                                 error = 0;
 442                                 break;
 443                         }
 444                         goto bad;
 445                 }
 446                 if (exthdrs.ip6e_rthdr) {
 447                         /* ah6_output doesn't modify mbuf chain */
 448                         rh->ip6r_segleft = segleft_org;
 449                 }
 450             }
 451 skip_ipsec2:;
 452 #endif
 453         }
 454
 455         /*
 456          * If there is a routing header, replace destination address field
 457          * with the first hop of the routing header.
 458          */
 459         if (exthdrs.ip6e_rthdr) {
 460                 struct ip6_rthdr *rh =
 461                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 462                                                   struct ip6_rthdr *));
 463                 struct ip6_rthdr0 *rh0;
 464
 465                 finaldst = ip6->ip6_dst;
 466                 switch (rh->ip6r_type) {
 467                 case IPV6_RTHDR_TYPE_0:
 468                          rh0 = (struct ip6_rthdr0 *)rh;
 469                          ip6->ip6_dst = rh0->ip6r0_addr[0];
 470                          bcopy((caddr_t)&rh0->ip6r0_addr[1],
 471                                (caddr_t)&rh0->ip6r0_addr[0],
 472                                sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
 473                                  );
 474                          rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
 475                          break;
 476                 default:        /* is it possible? */
 477                          error = EINVAL;
 478                          goto bad;
 479                 }
 480         }
 481
 482         /* Source address validation */
 483         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 484             (flags & IPV6_DADOUTPUT) == 0) {
 485                 error = EOPNOTSUPP;
 486                 ip6stat.ip6s_badscope++;
 487                 goto bad;
 488         }
 489         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 490                 error = EOPNOTSUPP;
 491                 ip6stat.ip6s_badscope++;
 492                 goto bad;
 493         }
 494
 495         ip6stat.ip6s_localout++;
 496
 497         /*
 498          * Route packet.
 499          */
 500         if (ro == 0) {
 501                 ro = &ip6route;
 502                 bzero((caddr_t)ro, sizeof(*ro));
 503         }
 504         ro_pmtu = ro;
 505         if (opt && opt->ip6po_rthdr)
 506                 ro = &opt->ip6po_route;
 507         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 508         /*
 509          * If there is a cached route,
 510          * check that it is to the same destination
 511          * and is still up. If not, free it and try again.
 512          */
 513         if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 514                          dst->sin6_family != AF_INET6 ||
 515                          !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 516                 RTFREE(ro->ro_rt);
 517                 ro->ro_rt = (struct rtentry *)0;
 518         }
 519         if (ro->ro_rt == 0) {
 520                 bzero(dst, sizeof(*dst));
 521                 dst->sin6_family = AF_INET6;
 522                 dst->sin6_len = sizeof(struct sockaddr_in6);
 523                 dst->sin6_addr = ip6->ip6_dst;
 524 #ifdef SCOPEDROUTING
 525                 /* XXX: sin6_scope_id should already be fixed at this point */
 526                 if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
 527                         dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
 528 #endif
 529         }
 530 #if defined(IPSEC) || defined(FAST_IPSEC)
 531         if (needipsec && needipsectun) {
 532                 struct ipsec_output_state state;
 533
 534                 /*
 535                  * All the extension headers will become inaccessible
 536                  * (since they can be encrypted).
 537                  * Don't panic, we need no more updates to extension headers
 538                  * on inner IPv6 packet (since they are now encapsulated).
 539                  *
 540                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 541                  */
 542                 bzero(&exthdrs, sizeof(exthdrs));
 543                 exthdrs.ip6e_ip6 = m;
 544
 545                 bzero(&state, sizeof(state));
 546                 state.m = m;
 547                 state.ro = (struct route *)ro;
 548                 state.dst = (struct sockaddr *)dst;
 549
 550                 error = ipsec6_output_tunnel(&state, sp, flags);
 551
 552                 m = state.m;
 553                 ro = (struct route_in6 *)state.ro;
 554                 dst = (struct sockaddr_in6 *)state.dst;
 555                 if (error) {
 556                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 557                         m0 = m = NULL;
 558                         m = NULL;
 559                         switch (error) {
 560                         case EHOSTUNREACH:
 561                         case ENETUNREACH:
 562                         case EMSGSIZE:
 563                         case ENOBUFS:
 564                         case ENOMEM:
 565                                 break;
 566                         default:
 567                                 printf("ip6_output (ipsec): error code %d\n", error);
 568                                 /* fall through */
 569                         case ENOENT:
 570                                 /* don't show these error codes to the user */
 571                                 error = 0;
 572                                 break;
 573                         }
 574                         goto bad;
 575                 }
 576
 577                 exthdrs.ip6e_ip6 = m;
 578         }
 579 #endif /* IPSEC */
 580
 581         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 582                 /* Unicast */
 583
 584 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 585 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 586                 /* xxx
 587                  * interface selection comes here
 588                  * if an interface is specified from an upper layer,
 589                  * ifp must point it.
 590                  */
 591                 if (ro->ro_rt == 0) {
 592                         /*
 593                          * non-bsdi always clone routes, if parent is
 594                          * PRF_CLONING.
 595                          */
 596                         rtalloc((struct route *)ro);
 597                 }
 598                 if (ro->ro_rt == 0) {
 599                         ip6stat.ip6s_noroute++;
 600                         error = EHOSTUNREACH;
 601                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 602                         goto bad;
 603                 }
 604                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 605                 ifp = ro->ro_rt->rt_ifp;
 606                 ro->ro_rt->rt_use++;
 607                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 608                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 609                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 610
 611                 in6_ifstat_inc(ifp, ifs6_out_request);
 612
 613                 /*
 614                  * Check if the outgoing interface conflicts with
 615                  * the interface specified by ifi6_ifindex (if specified).
 616                  * Note that loopback interface is always okay.
 617                  * (this may happen when we are sending a packet to one of
 618                  *  our own addresses.)
 619                  */
 620                 if (opt && opt->ip6po_pktinfo
 621                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 622                         if (!(ifp->if_flags & IFF_LOOPBACK)
 623                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 624                                 ip6stat.ip6s_noroute++;
 625                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 626                                 error = EHOSTUNREACH;
 627                                 goto bad;
 628                         }
 629                 }
 630
 631                 if (opt && opt->ip6po_hlim != -1)
 632                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 633         } else {
 634                 /* Multicast */
 635                 struct  in6_multi *in6m;
 636
 637                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 638
 639                 /*
 640                  * See if the caller provided any multicast options
 641                  */
 642                 ifp = NULL;
 643                 if (im6o != NULL) {
 644                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 645                         if (im6o->im6o_multicast_ifp != NULL)
 646                                 ifp = im6o->im6o_multicast_ifp;
 647                 } else
 648                         ip6->ip6_hlim = ip6_defmcasthlim;
 649
 650                 /*
 651                  * See if the caller provided the outgoing interface
 652                  * as an ancillary data.
 653                  * Boundary check for ifindex is assumed to be already done.
 654                  */
 655                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 656                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 657
 658                 /*
 659                  * If the destination is a node-local scope multicast,
 660                  * the packet should be loop-backed only.
 661                  */
 662                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 663                         /*
 664                          * If the outgoing interface is already specified,
 665                          * it should be a loopback interface.
 666                          */
 667                         if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
 668                                 ip6stat.ip6s_badscope++;
 669                                 error = ENETUNREACH; /* XXX: better error? */
 670                                 /* XXX correct ifp? */
 671                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 672                                 goto bad;
 673                         } else {
 674                                 ifp = &loif[0];
 675                         }
 676                 }
 677
 678                 if (opt && opt->ip6po_hlim != -1)
 679                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 680
 681                 /*
 682                  * If caller did not provide an interface lookup a
 683                  * default in the routing table.  This is either a
 684                  * default for the speicfied group (i.e. a host
 685                  * route), or a multicast default (a route for the
 686                  * ``net'' ff00::/8).
 687                  */
 688                 if (ifp == NULL) {
 689                         if (ro->ro_rt == NULL) {
 690                                 ro->ro_rt =
 691                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 692                         }
 693                         if (ro->ro_rt == NULL) {
 694                                 ip6stat.ip6s_noroute++;
 695                                 error = EHOSTUNREACH;
 696                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 697                                 goto bad;
 698                         }
 699                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 700                         ifp = ro->ro_rt->rt_ifp;
 701                         ro->ro_rt->rt_use++;
 702                 }
 703
 704                 if ((flags & IPV6_FORWARDING) == 0)
 705                         in6_ifstat_inc(ifp, ifs6_out_request);
 706                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 707
 708                 /*
 709                  * Confirm that the outgoing interface supports multicast.
 710                  */
 711                 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 712                         ip6stat.ip6s_noroute++;
 713                         in6_ifstat_inc(ifp, ifs6_out_discard);
 714                         error = ENETUNREACH;
 715                         goto bad;
 716                 }
 717                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 718                 if (in6m != NULL &&
 719                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 720                         /*
 721                          * If we belong to the destination multicast group
 722                          * on the outgoing interface, and the caller did not
 723                          * forbid loopback, loop back a copy.
 724                          */
 725                         ip6_mloopback(ifp, m, dst);
 726                 } else {
 727                         /*
 728                          * If we are acting as a multicast router, perform
 729                          * multicast forwarding as if the packet had just
 730                          * arrived on the interface to which we are about
 731                          * to send.  The multicast forwarding function
 732                          * recursively calls this function, using the
 733                          * IPV6_FORWARDING flag to prevent infinite recursion.
 734                          *
 735                          * Multicasts that are looped back by ip6_mloopback(),
 736                          * above, will be forwarded by the ip6_input() routine,
 737                          * if necessary.
 738                          */
 739                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 740                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 741                                         m_freem(m);
 742                                         goto done;
 743                                 }
 744                         }
 745                 }
 746                 /*
 747                  * Multicasts with a hoplimit of zero may be looped back,
 748                  * above, but must not be transmitted on a network.
 749                  * Also, multicasts addressed to the loopback interface
 750                  * are not sent -- the above call to ip6_mloopback() will
 751                  * loop back a copy if this host actually belongs to the
 752                  * destination group on the loopback interface.
 753                  */
 754                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 755                         m_freem(m);
 756                         goto done;
 757                 }
 758         }
 759
 760         /*
 761          * Fill the outgoing inteface to tell the upper layer
 762          * to increment per-interface statistics.
 763          */
 764         if (ifpp)
 765                 *ifpp = ifp;
 766
 767         /*
 768          * Determine path MTU.
 769          */
 770         if (ro_pmtu != ro) {
 771                 /* The first hop and the final destination may differ. */
 772                 struct sockaddr_in6 *sin6_fin =
 773                         (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
 774
 775                 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 776                                        !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
 777                                                            &finaldst))) {
 778                         RTFREE(ro_pmtu->ro_rt);
 779                         ro_pmtu->ro_rt = (struct rtentry *)0;
 780                 }
 781                 if (ro_pmtu->ro_rt == 0) {
 782                         bzero(sin6_fin, sizeof(*sin6_fin));
 783                         sin6_fin->sin6_family = AF_INET6;
 784                         sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
 785                         sin6_fin->sin6_addr = finaldst;
 786
 787                         rtalloc((struct route *)ro_pmtu);
 788                 }
 789         }
 790         if (ro_pmtu->ro_rt != NULL) {
 791                 u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
 792
 793                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
 794                 if (mtu > ifmtu || mtu == 0) {
 795                         /*
 796                          * The MTU on the route is larger than the MTU on
 797                          * the interface!  This shouldn't happen, unless the
 798                          * MTU of the interface has been changed after the
 799                          * interface was brought up.  Change the MTU in the
 800                          * route to match the interface MTU (as long as the
 801                          * field isn't locked).
 802                          *
 803                          * if MTU on the route is 0, we need to fix the MTU.
 804                          * this case happens with path MTU discovery timeouts.
 805                          */
 806                          mtu = ifmtu;
 807                          if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
 808                                  ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
 809                 }
 810         } else {
 811                 mtu = nd_ifinfo[ifp->if_index].linkmtu;
 812         }
 813
 814         /*
 815          * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
 816          */
 817         if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
 818                 mtu = IPV6_MMTU;
 819
 820         /* Fake scoped addresses */
 821         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 822                 /*
 823                  * If source or destination address is a scoped address, and
 824                  * the packet is going to be sent to a loopback interface,
 825                  * we should keep the original interface.
 826                  */
 827
 828                 /*
 829                  * XXX: this is a very experimental and temporary solution.
 830                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 831                  * field of the structure here.
 832                  * We rely on the consistency between two scope zone ids
 833                  * of source and destination, which should already be assured.
 834                  * Larger scopes than link will be supported in the future.
 835                  */
 836                 origifp = NULL;
 837                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 838                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 839                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 840                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 841                 /*
 842                  * XXX: origifp can be NULL even in those two cases above.
 843                  * For example, if we remove the (only) link-local address
 844                  * from the loopback interface, and try to send a link-local
 845                  * address without link-id information.  Then the source
 846                  * address is ::1, and the destination address is the
 847                  * link-local address with its s6_addr16[1] being zero.
 848                  * What is worse, if the packet goes to the loopback interface
 849                  * by a default rejected route, the null pointer would be
 850                  * passed to looutput, and the kernel would hang.
 851                  * The following last resort would prevent such disaster.
 852                  */
 853                 if (origifp == NULL)
 854                         origifp = ifp;
 855         }
 856         else
 857                 origifp = ifp;
 858 #ifndef SCOPEDROUTING
 859         /*
 860          * clear embedded scope identifiers if necessary.
 861          * in6_clearscope will touch the addresses only when necessary.
 862          */
 863         in6_clearscope(&ip6->ip6_src);
 864         in6_clearscope(&ip6->ip6_dst);
 865 #endif
 866
 867         /*
 868          * Check with the firewall...
 869          */
 870         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 871                 u_short port = 0;
 872                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 873                 /* If ipfw says divert, we have to just drop packet */
 874                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 875                         m_freem(m);
 876                         goto done;
 877                 }
 878                 if (!m) {
 879                         error = EACCES;
 880                         goto done;
 881                 }
 882         }
 883
 884         /*
 885          * If the outgoing packet contains a hop-by-hop options header,
 886          * it must be examined and processed even by the source node.
 887          * (RFC 2460, section 4.)
 888          */
 889         if (exthdrs.ip6e_hbh) {
 890                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 891                 u_int32_t dummy1; /* XXX unused */
 892                 u_int32_t dummy2; /* XXX unused */
 893
 894 #ifdef DIAGNOSTIC
 895                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 896                         panic("ip6e_hbh is not continuous");
 897 #endif
 898                 /*
 899                  *  XXX: if we have to send an ICMPv6 error to the sender,
 900                  *       we need the M_LOOP flag since icmp6_error() expects
 901                  *       the IPv6 and the hop-by-hop options header are
 902                  *       continuous unless the flag is set.
 903                  */
 904                 m->m_flags |= M_LOOP;
 905                 m->m_pkthdr.rcvif = ifp;
 906                 if (ip6_process_hopopts(m,
 907                                         (u_int8_t *)(hbh + 1),
 908                                         ((hbh->ip6h_len + 1) << 3) -
 909                                         sizeof(struct ip6_hbh),
 910                                         &dummy1, &dummy2) < 0) {
 911                         /* m was already freed at this point */
 912                         error = EINVAL;/* better error? */
 913                         goto done;
 914                 }
 915                 m->m_flags &= ~M_LOOP; /* XXX */
 916                 m->m_pkthdr.rcvif = NULL;
 917         }
 918
 919         /*
 920          * Run through list of hooks for output packets.
 921          */
 922         if (pfil_has_hooks(&inet6_pfil_hook)) {
 923                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 924                 if (error != 0 || m == NULL)
 925                         goto done;
 926                 ip6 = mtod(m, struct ip6_hdr *);
 927         }
 928
 929         /*
 930          * Send the packet to the outgoing interface.
 931          * If necessary, do IPv6 fragmentation before sending.
 932          */
 933         tlen = m->m_pkthdr.len;
 934         if (tlen <= mtu
 935 #ifdef notyet
 936             /*
 937              * On any link that cannot convey a 1280-octet packet in one piece,
 938              * link-specific fragmentation and reassembly must be provided at
 939              * a layer below IPv6. [RFC 2460, sec.5]
 940              * Thus if the interface has ability of link-level fragmentation,
 941              * we can just send the packet even if the packet size is
 942              * larger than the link's MTU.
 943              * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
 944              */
 945
 946             || ifp->if_flags & IFF_FRAGMENTABLE
 947 #endif
 948             )
 949         {
 950                 /* Record statistics for this interface address. */
 951                 if (ia && !(flags & IPV6_FORWARDING)) {
 952                         ia->ia_ifa.if_opackets++;
 953                         ia->ia_ifa.if_obytes += m->m_pkthdr.len;
 954                 }
 955 #ifdef IPSEC
 956                 /* clean ipsec history once it goes out of the node */
 957                 ipsec_delaux(m);
 958 #endif
 959                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 960                 goto done;
 961         } else if (mtu < IPV6_MMTU) {
 962                 /*
 963                  * note that path MTU is never less than IPV6_MMTU
 964                  * (see icmp6_input).
 965                  */
 966                 error = EMSGSIZE;
 967                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 968                 goto bad;
 969         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 970                 error = EMSGSIZE;
 971                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 972                 goto bad;
 973         } else {
 974                 struct mbuf **mnext, *m_frgpart;
 975                 struct ip6_frag *ip6f;
 976                 u_int32_t id = htonl(ip6_id++);
 977                 u_char nextproto;
 978
 979                 /*
 980                  * Too large for the destination or interface;
 981                  * fragment if possible.
 982                  * Must be able to put at least 8 bytes per fragment.
 983                  */
 984                 hlen = unfragpartlen;
 985                 if (mtu > IPV6_MAXPACKET)
 986                         mtu = IPV6_MAXPACKET;
 987
 988                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 989                 if (len < 8) {
 990                         error = EMSGSIZE;
 991                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 992                         goto bad;
 993                 }
 994
 995                 mnext = &m->m_nextpkt;
 996
 997                 /*
 998                  * Change the next header field of the last header in the
 999                  * unfragmentable part.
1000                  */
1001                 if (exthdrs.ip6e_rthdr) {
1002                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1003                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1004                 } else if (exthdrs.ip6e_dest1) {
1005                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1006                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1007                 } else if (exthdrs.ip6e_hbh) {
1008                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1009                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1010                 } else {
1011                         nextproto = ip6->ip6_nxt;
1012                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1013                 }
1014
1015                 /*
1016                  * Loop through length of segment after first fragment,
1017                  * make new header and copy data of each part and link onto
1018                  * chain.
1019                  */
1020                 m0 = m;
1021                 for (off = hlen; off < tlen; off += len) {
1022                         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1023                         if (!m) {
1024                                 error = ENOBUFS;
1025                                 ip6stat.ip6s_odropped++;
1026                                 goto sendorfree;
1027                         }
1028                         m->m_pkthdr.rcvif = NULL;
1029                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1030                         *mnext = m;
1031                         mnext = &m->m_nextpkt;
1032                         m->m_data += max_linkhdr;
1033                         mhip6 = mtod(m, struct ip6_hdr *);
1034                         *mhip6 = *ip6;
1035                         m->m_len = sizeof(*mhip6);
1036                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1037                         if (error) {
1038                                 ip6stat.ip6s_odropped++;
1039                                 goto sendorfree;
1040                         }
1041                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1042                         if (off + len >= tlen)
1043                                 len = tlen - off;
1044                         else
1045                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1046                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1047                                                           sizeof(*ip6f) -
1048                                                           sizeof(struct ip6_hdr)));
1049                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1050                                 error = ENOBUFS;
1051                                 ip6stat.ip6s_odropped++;
1052                                 goto sendorfree;
1053                         }
1054                         m_cat(m, m_frgpart);
1055                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1056                         m->m_pkthdr.rcvif = (struct ifnet *)0;
1057                         ip6f->ip6f_reserved = 0;
1058                         ip6f->ip6f_ident = id;
1059                         ip6f->ip6f_nxt = nextproto;
1060                         ip6stat.ip6s_ofragments++;
1061                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1062                 }
1063
1064                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1065         }
1066
1067         /*
1068          * Remove leading garbages.
1069          */
1070 sendorfree:
1071         m = m0->m_nextpkt;
1072         m0->m_nextpkt = 0;
1073         m_freem(m0);
1074         for (m0 = m; m; m = m0) {
1075                 m0 = m->m_nextpkt;
1076                 m->m_nextpkt = 0;
1077                 if (error == 0) {
1078                         /* Record statistics for this interface address. */
1079                         if (ia) {
1080                                 ia->ia_ifa.if_opackets++;
1081                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1082                         }
1083 #ifdef IPSEC
1084                         /* clean ipsec history once it goes out of the node */
1085                         ipsec_delaux(m);
1086 #endif
1087                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1088                 } else
1089                         m_freem(m);
1090         }
1091
1092         if (error == 0)
1093                 ip6stat.ip6s_fragmented++;
1094
1095 done:
1096         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1097                 RTFREE(ro->ro_rt);
1098         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1099                 RTFREE(ro_pmtu->ro_rt);
1100         }
1101
1102 #ifdef IPSEC
1103         if (sp != NULL)
1104                 key_freesp(sp);
1105 #endif /* IPSEC */
1106 #ifdef FAST_IPSEC
1107         if (sp != NULL)
1108                 KEY_FREESP(&sp);
1109 #endif /* FAST_IPSEC */
1110
1111         return(error);
1112
1113 freehdrs:
1114         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1115         m_freem(exthdrs.ip6e_dest1);
1116         m_freem(exthdrs.ip6e_rthdr);
1117         m_freem(exthdrs.ip6e_dest2);
1118         /* fall through */
1119 bad:
1120         m_freem(m);
1121         goto done;
1122 }
1123
1124 static int
1125 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1126 {
1127         struct mbuf *m;
1128
1129         if (hlen > MCLBYTES)
1130                 return(ENOBUFS); /* XXX */
1131
1132         MGET(m, MB_DONTWAIT, MT_DATA);
1133         if (!m)
1134                 return(ENOBUFS);
1135
1136         if (hlen > MLEN) {
1137                 MCLGET(m, MB_DONTWAIT);
1138                 if ((m->m_flags & M_EXT) == 0) {
1139                         m_free(m);
1140                         return(ENOBUFS);
1141                 }
1142         }
1143         m->m_len = hlen;
1144         if (hdr)
1145                 bcopy(hdr, mtod(m, caddr_t), hlen);
1146
1147         *mp = m;
1148         return(0);
1149 }
1150
1151 /*
1152  * Insert jumbo payload option.
1153  */
1154 static int
1155 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1156 {
1157         struct mbuf *mopt;
1158         u_char *optbuf;
1159         u_int32_t v;
1160
1161 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1162
1163         /*
1164          * If there is no hop-by-hop options header, allocate new one.
1165          * If there is one but it doesn't have enough space to store the
1166          * jumbo payload option, allocate a cluster to store the whole options.
1167          * Otherwise, use it to store the options.
1168          */
1169         if (exthdrs->ip6e_hbh == 0) {
1170                 MGET(mopt, MB_DONTWAIT, MT_DATA);
1171                 if (mopt == 0)
1172                         return(ENOBUFS);
1173                 mopt->m_len = JUMBOOPTLEN;
1174                 optbuf = mtod(mopt, u_char *);
1175                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1176                 exthdrs->ip6e_hbh = mopt;
1177         } else {
1178                 struct ip6_hbh *hbh;
1179
1180                 mopt = exthdrs->ip6e_hbh;
1181                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1182                         /*
1183                          * XXX assumption:
1184                          * - exthdrs->ip6e_hbh is not referenced from places
1185                          *   other than exthdrs.
1186                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1187                          */
1188                         int oldoptlen = mopt->m_len;
1189                         struct mbuf *n;
1190
1191                         /*
1192                          * XXX: give up if the whole (new) hbh header does
1193                          * not fit even in an mbuf cluster.
1194                          */
1195                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1196                                 return(ENOBUFS);
1197
1198                         /*
1199                          * As a consequence, we must always prepare a cluster
1200                          * at this point.
1201                          */
1202                         MGET(n, MB_DONTWAIT, MT_DATA);
1203                         if (n) {
1204                                 MCLGET(n, MB_DONTWAIT);
1205                                 if ((n->m_flags & M_EXT) == 0) {
1206                                         m_freem(n);
1207                                         n = NULL;
1208                                 }
1209                         }
1210                         if (!n)
1211                                 return(ENOBUFS);
1212                         n->m_len = oldoptlen + JUMBOOPTLEN;
1213                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1214                               oldoptlen);
1215                         optbuf = mtod(n, caddr_t) + oldoptlen;
1216                         m_freem(mopt);
1217                         mopt = exthdrs->ip6e_hbh = n;
1218                 } else {
1219                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1220                         mopt->m_len += JUMBOOPTLEN;
1221                 }
1222                 optbuf[0] = IP6OPT_PADN;
1223                 optbuf[1] = 1;
1224
1225                 /*
1226                  * Adjust the header length according to the pad and
1227                  * the jumbo payload option.
1228                  */
1229                 hbh = mtod(mopt, struct ip6_hbh *);
1230                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1231         }
1232
1233         /* fill in the option. */
1234         optbuf[2] = IP6OPT_JUMBO;
1235         optbuf[3] = 4;
1236         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1237         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1238
1239         /* finally, adjust the packet header length */
1240         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1241
1242         return(0);
1243 #undef JUMBOOPTLEN
1244 }
1245
1246 /*
1247  * Insert fragment header and copy unfragmentable header portions.
1248  */
1249 static int
1250 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1251                   struct ip6_frag **frghdrp)
1252 {
1253         struct mbuf *n, *mlast;
1254
1255         if (hlen > sizeof(struct ip6_hdr)) {
1256                 n = m_copym(m0, sizeof(struct ip6_hdr),
1257                             hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1258                 if (n == 0)
1259                         return(ENOBUFS);
1260                 m->m_next = n;
1261         } else
1262                 n = m;
1263
1264         /* Search for the last mbuf of unfragmentable part. */
1265         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1266                 ;
1267
1268         if ((mlast->m_flags & M_EXT) == 0 &&
1269             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1270                 /* use the trailing space of the last mbuf for the fragment hdr */
1271                 *frghdrp =
1272                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1273                 mlast->m_len += sizeof(struct ip6_frag);
1274                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1275         } else {
1276                 /* allocate a new mbuf for the fragment header */
1277                 struct mbuf *mfrg;
1278
1279                 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1280                 if (mfrg == 0)
1281                         return(ENOBUFS);
1282                 mfrg->m_len = sizeof(struct ip6_frag);
1283                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1284                 mlast->m_next = mfrg;
1285         }
1286
1287         return(0);
1288 }
1289
1290 /*
1291  * IP6 socket option processing.
1292  */
1293 int
1294 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1295 {
1296         int privileged;
1297         struct inpcb *in6p = sotoinpcb(so);
1298         int error, optval;
1299         int level, op, optname;
1300         int optlen;
1301         struct thread *td;
1302
1303         if (sopt) {
1304                 level = sopt->sopt_level;
1305                 op = sopt->sopt_dir;
1306                 optname = sopt->sopt_name;
1307                 optlen = sopt->sopt_valsize;
1308                 td = sopt->sopt_td;
1309         } else {
1310                 panic("ip6_ctloutput: arg soopt is NULL");
1311                 /* NOT REACHED */
1312                 td = NULL;
1313         }
1314         error = optval = 0;
1315
1316         privileged = (td == NULL || suser(td)) ? 0 : 1;
1317
1318         if (level == IPPROTO_IPV6) {
1319                 switch (op) {
1320
1321                 case SOPT_SET:
1322                         switch (optname) {
1323                         case IPV6_PKTOPTIONS:
1324                         {
1325                                 struct mbuf *m;
1326
1327                                 error = soopt_getm(sopt, &m); /* XXX */
1328                                 if (error != NULL)
1329                                         break;
1330                                 error = soopt_mcopyin(sopt, m); /* XXX */
1331                                 if (error != NULL)
1332                                         break;
1333                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1334                                                     m, so, sopt);
1335                                 m_freem(m); /* XXX */
1336                                 break;
1337                         }
1338
1339                         /*
1340                          * Use of some Hop-by-Hop options or some
1341                          * Destination options, might require special
1342                          * privilege.  That is, normal applications
1343                          * (without special privilege) might be forbidden
1344                          * from setting certain options in outgoing packets,
1345                          * and might never see certain options in received
1346                          * packets. [RFC 2292 Section 6]
1347                          * KAME specific note:
1348                          *  KAME prevents non-privileged users from sending or
1349                          *  receiving ANY hbh/dst options in order to avoid
1350                          *  overhead of parsing options in the kernel.
1351                          */
1352                         case IPV6_UNICAST_HOPS:
1353                         case IPV6_CHECKSUM:
1354                         case IPV6_FAITH:
1355
1356                         case IPV6_V6ONLY:
1357                                 if (optlen != sizeof(int)) {
1358                                         error = EINVAL;
1359                                         break;
1360                                 }
1361                                 error = sooptcopyin(sopt, &optval,
1362                                         sizeof optval, sizeof optval);
1363                                 if (error)
1364                                         break;
1365                                 switch (optname) {
1366
1367                                 case IPV6_UNICAST_HOPS:
1368                                         if (optval < -1 || optval >= 256)
1369                                                 error = EINVAL;
1370                                         else {
1371                                                 /* -1 = kernel default */
1372                                                 in6p->in6p_hops = optval;
1373
1374                                                 if ((in6p->in6p_vflag &
1375                                                      INP_IPV4) != 0)
1376                                                         in6p->inp_ip_ttl = optval;
1377                                         }
1378                                         break;
1379 #define OPTSET(bit) \
1380 do { \
1381         if (optval) \
1382                 in6p->in6p_flags |= (bit); \
1383         else \
1384                 in6p->in6p_flags &= ~(bit); \
1385 } while (0)
1386 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1387
1388                                 case IPV6_CHECKSUM:
1389                                         in6p->in6p_cksum = optval;
1390                                         break;
1391
1392                                 case IPV6_FAITH:
1393                                         OPTSET(IN6P_FAITH);
1394                                         break;
1395
1396                                 case IPV6_V6ONLY:
1397                                         /*
1398                                          * make setsockopt(IPV6_V6ONLY)
1399                                          * available only prior to bind(2).
1400                                          * see ipng mailing list, Jun 22 2001.
1401                                          */
1402                                         if (in6p->in6p_lport ||
1403                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1404                                         {
1405                                                 error = EINVAL;
1406                                                 break;
1407                                         }
1408                                         OPTSET(IN6P_IPV6_V6ONLY);
1409                                         if (optval)
1410                                                 in6p->in6p_vflag &= ~INP_IPV4;
1411                                         else
1412                                                 in6p->in6p_vflag |= INP_IPV4;
1413                                         break;
1414                                 }
1415                                 break;
1416
1417                         case IPV6_PKTINFO:
1418                         case IPV6_HOPLIMIT:
1419                         case IPV6_HOPOPTS:
1420                         case IPV6_DSTOPTS:
1421                         case IPV6_RTHDR:
1422                                 /* RFC 2292 */
1423                                 if (optlen != sizeof(int)) {
1424                                         error = EINVAL;
1425                                         break;
1426                                 }
1427                                 error = sooptcopyin(sopt, &optval,
1428                                         sizeof optval, sizeof optval);
1429                                 if (error)
1430                                         break;
1431                                 switch (optname) {
1432                                 case IPV6_PKTINFO:
1433                                         OPTSET(IN6P_PKTINFO);
1434                                         break;
1435                                 case IPV6_HOPLIMIT:
1436                                         OPTSET(IN6P_HOPLIMIT);
1437                                         break;
1438                                 case IPV6_HOPOPTS:
1439                                         /*
1440                                          * Check super-user privilege.
1441                                          * See comments for IPV6_RECVHOPOPTS.
1442                                          */
1443                                         if (!privileged)
1444                                                 return(EPERM);
1445                                         OPTSET(IN6P_HOPOPTS);
1446                                         break;
1447                                 case IPV6_DSTOPTS:
1448                                         if (!privileged)
1449                                                 return(EPERM);
1450                                         OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1451                                         break;
1452                                 case IPV6_RTHDR:
1453                                         OPTSET(IN6P_RTHDR);
1454                                         break;
1455                                 }
1456                                 break;
1457 #undef OPTSET
1458
1459                         case IPV6_MULTICAST_IF:
1460                         case IPV6_MULTICAST_HOPS:
1461                         case IPV6_MULTICAST_LOOP:
1462                         case IPV6_JOIN_GROUP:
1463                         case IPV6_LEAVE_GROUP:
1464                             {
1465                                 struct mbuf *m;
1466                                 if (sopt->sopt_valsize > MLEN) {
1467                                         error = EMSGSIZE;
1468                                         break;
1469                                 }
1470                                 /* XXX */
1471                                 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1472                                 if (m == 0) {
1473                                         error = ENOBUFS;
1474                                         break;
1475                                 }
1476                                 m->m_len = sopt->sopt_valsize;
1477                                 error = sooptcopyin(sopt, mtod(m, char *),
1478                                                     m->m_len, m->m_len);
1479                                 error = ip6_setmoptions(sopt->sopt_name,
1480                                                         &in6p->in6p_moptions,
1481                                                         m);
1482                                 (void)m_free(m);
1483                             }
1484                                 break;
1485
1486                         case IPV6_PORTRANGE:
1487                                 error = sooptcopyin(sopt, &optval,
1488                                     sizeof optval, sizeof optval);
1489                                 if (error)
1490                                         break;
1491
1492                                 switch (optval) {
1493                                 case IPV6_PORTRANGE_DEFAULT:
1494                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1495                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1496                                         break;
1497
1498                                 case IPV6_PORTRANGE_HIGH:
1499                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1500                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1501                                         break;
1502
1503                                 case IPV6_PORTRANGE_LOW:
1504                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1505                                         in6p->in6p_flags |= IN6P_LOWPORT;
1506                                         break;
1507
1508                                 default:
1509                                         error = EINVAL;
1510                                         break;
1511                                 }
1512                                 break;
1513
1514 #if defined(IPSEC) || defined(FAST_IPSEC)
1515                         case IPV6_IPSEC_POLICY:
1516                             {
1517                                 caddr_t req = NULL;
1518                                 size_t len = 0;
1519                                 struct mbuf *m;
1520
1521                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1522                                         break;
1523                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1524                                         break;
1525                                 if (m) {
1526                                         req = mtod(m, caddr_t);
1527                                         len = m->m_len;
1528                                 }
1529                                 error = ipsec6_set_policy(in6p, optname, req,
1530                                                           len, privileged);
1531                                 m_freem(m);
1532                             }
1533                                 break;
1534 #endif /* KAME IPSEC */
1535
1536                         case IPV6_FW_ADD:
1537                         case IPV6_FW_DEL:
1538                         case IPV6_FW_FLUSH:
1539                         case IPV6_FW_ZERO:
1540                             {
1541                                 struct mbuf *m;
1542                                 struct mbuf **mp = &m;
1543
1544                                 if (ip6_fw_ctl_ptr == NULL)
1545                                         return EINVAL;
1546                                 /* XXX */
1547                                 if ((error = soopt_getm(sopt, &m)) != 0)
1548                                         break;
1549                                 /* XXX */
1550                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
1551                                         break;
1552                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1553                                 m = *mp;
1554                             }
1555                                 break;
1556
1557                         default:
1558                                 error = ENOPROTOOPT;
1559                                 break;
1560                         }
1561                         break;
1562
1563                 case SOPT_GET:
1564                         switch (optname) {
1565
1566                         case IPV6_PKTOPTIONS:
1567                                 if (in6p->in6p_options) {
1568                                         struct mbuf *m;
1569                                         m = m_copym(in6p->in6p_options,
1570                                             0, M_COPYALL, MB_WAIT);
1571                                         error = soopt_mcopyout(sopt, m);
1572                                         if (error == 0)
1573                                                 m_freem(m);
1574                                 } else
1575                                         sopt->sopt_valsize = 0;
1576                                 break;
1577
1578                         case IPV6_UNICAST_HOPS:
1579                         case IPV6_CHECKSUM:
1580
1581                         case IPV6_FAITH:
1582                         case IPV6_V6ONLY:
1583                         case IPV6_PORTRANGE:
1584                                 switch (optname) {
1585
1586                                 case IPV6_UNICAST_HOPS:
1587                                         optval = in6p->in6p_hops;
1588                                         break;
1589
1590                                 case IPV6_CHECKSUM:
1591                                         optval = in6p->in6p_cksum;
1592                                         break;
1593
1594                                 case IPV6_FAITH:
1595                                         optval = OPTBIT(IN6P_FAITH);
1596                                         break;
1597
1598                                 case IPV6_V6ONLY:
1599                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1600                                         break;
1601
1602                                 case IPV6_PORTRANGE:
1603                                     {
1604                                         int flags;
1605                                         flags = in6p->in6p_flags;
1606                                         if (flags & IN6P_HIGHPORT)
1607                                                 optval = IPV6_PORTRANGE_HIGH;
1608                                         else if (flags & IN6P_LOWPORT)
1609                                                 optval = IPV6_PORTRANGE_LOW;
1610                                         else
1611                                                 optval = 0;
1612                                         break;
1613                                     }
1614                                 }
1615                                 error = sooptcopyout(sopt, &optval,
1616                                         sizeof optval);
1617                                 break;
1618
1619                         case IPV6_PKTINFO:
1620                         case IPV6_HOPLIMIT:
1621                         case IPV6_HOPOPTS:
1622                         case IPV6_RTHDR:
1623                         case IPV6_DSTOPTS:
1624                                 if (optname == IPV6_HOPOPTS ||
1625                                     optname == IPV6_DSTOPTS ||
1626                                     !privileged)
1627                                         return(EPERM);
1628                                 switch (optname) {
1629                                 case IPV6_PKTINFO:
1630                                         optval = OPTBIT(IN6P_PKTINFO);
1631                                         break;
1632                                 case IPV6_HOPLIMIT:
1633                                         optval = OPTBIT(IN6P_HOPLIMIT);
1634                                         break;
1635                                 case IPV6_HOPOPTS:
1636                                         if (!privileged)
1637                                                 return(EPERM);
1638                                         optval = OPTBIT(IN6P_HOPOPTS);
1639                                         break;
1640                                 case IPV6_RTHDR:
1641                                         optval = OPTBIT(IN6P_RTHDR);
1642                                         break;
1643                                 case IPV6_DSTOPTS:
1644                                         if (!privileged)
1645                                                 return(EPERM);
1646                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1647                                         break;
1648                                 }
1649                                 error = sooptcopyout(sopt, &optval,
1650                                         sizeof optval);
1651                                 break;
1652
1653                         case IPV6_MULTICAST_IF:
1654                         case IPV6_MULTICAST_HOPS:
1655                         case IPV6_MULTICAST_LOOP:
1656                         case IPV6_JOIN_GROUP:
1657                         case IPV6_LEAVE_GROUP:
1658                             {
1659                                 struct mbuf *m;
1660                                 error = ip6_getmoptions(sopt->sopt_name,
1661                                                 in6p->in6p_moptions, &m);
1662                                 if (error == 0)
1663                                         error = sooptcopyout(sopt,
1664                                                 mtod(m, char *), m->m_len);
1665                                 m_freem(m);
1666                             }
1667                                 break;
1668
1669 #if defined(IPSEC) || defined(FAST_IPSEC)
1670                         case IPV6_IPSEC_POLICY:
1671                           {
1672                                 caddr_t req = NULL;
1673                                 size_t len = 0;
1674                                 struct mbuf *m = NULL;
1675                                 struct mbuf **mp = &m;
1676
1677                                 error = soopt_getm(sopt, &m); /* XXX */
1678                                 if (error != NULL)
1679                                         break;
1680                                 error = soopt_mcopyin(sopt, m); /* XXX */
1681                                 if (error != NULL)
1682                                         break;
1683                                 if (m) {
1684                                         req = mtod(m, caddr_t);
1685                                         len = m->m_len;
1686                                 }
1687                                 error = ipsec6_get_policy(in6p, req, len, mp);
1688                                 if (error == 0)
1689                                         error = soopt_mcopyout(sopt, m); /*XXX*/
1690                                 if (error == 0 && m)
1691                                         m_freem(m);
1692                                 break;
1693                           }
1694 #endif /* KAME IPSEC */
1695
1696                         case IPV6_FW_GET:
1697                           {
1698                                 struct mbuf *m;
1699                                 struct mbuf **mp = &m;
1700
1701                                 if (ip6_fw_ctl_ptr == NULL)
1702                                 {
1703                                         return EINVAL;
1704                                 }
1705                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1706                                 if (error == 0)
1707                                         error = soopt_mcopyout(sopt, m); /* XXX */
1708                                 if (error == 0 && m)
1709                                         m_freem(m);
1710                           }
1711                                 break;
1712
1713                         default:
1714                                 error = ENOPROTOOPT;
1715                                 break;
1716                         }
1717                         break;
1718                 }
1719         } else {
1720                 error = EINVAL;
1721         }
1722         return(error);
1723 }
1724
1725 /*
1726  * Set up IP6 options in pcb for insertion in output packets or
1727  * specifying behavior of outgoing packets.
1728  */
1729 static int
1730 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
1731             struct sockopt *sopt)
1732 {
1733         struct ip6_pktopts *opt = *pktopt;
1734         int error = 0;
1735         struct thread *td = sopt->sopt_td;
1736         int priv = 0;
1737
1738         /* turn off any old options. */
1739         if (opt) {
1740 #ifdef DIAGNOSTIC
1741                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1742                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1743                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1744                         printf("ip6_pcbopts: all specified options are cleared.\n");
1745 #endif
1746                 ip6_clearpktopts(opt, 1, -1);
1747         } else
1748                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1749         *pktopt = NULL;
1750
1751         if (!m || m->m_len == 0) {
1752                 /*
1753                  * Only turning off any previous options, regardless of
1754                  * whether the opt is just created or given.
1755                  */
1756                 free(opt, M_IP6OPT);
1757                 return(0);
1758         }
1759
1760         /*  set options specified by user. */
1761         if (suser(td) == 0)
1762                 priv = 1;
1763         if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1764                 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1765                 free(opt, M_IP6OPT);
1766                 return(error);
1767         }
1768         *pktopt = opt;
1769         return(0);
1770 }
1771
1772 /*
1773  * initialize ip6_pktopts.  beware that there are non-zero default values in
1774  * the struct.
1775  */
1776 void
1777 init_ip6pktopts(struct ip6_pktopts *opt)
1778 {
1779
1780         bzero(opt, sizeof(*opt));
1781         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
1782 }
1783
1784 void
1785 ip6_clearpktopts(struct ip6_pktopts *pktopt, int needfree, int optname)
1786 {
1787         if (pktopt == NULL)
1788                 return;
1789
1790         if (optname == -1) {
1791                 if (needfree && pktopt->ip6po_pktinfo)
1792                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
1793                 pktopt->ip6po_pktinfo = NULL;
1794         }
1795         if (optname == -1)
1796                 pktopt->ip6po_hlim = -1;
1797         if (optname == -1) {
1798                 if (needfree && pktopt->ip6po_nexthop)
1799                         free(pktopt->ip6po_nexthop, M_IP6OPT);
1800                 pktopt->ip6po_nexthop = NULL;
1801         }
1802         if (optname == -1) {
1803                 if (needfree && pktopt->ip6po_hbh)
1804                         free(pktopt->ip6po_hbh, M_IP6OPT);
1805                 pktopt->ip6po_hbh = NULL;
1806         }
1807         if (optname == -1) {
1808                 if (needfree && pktopt->ip6po_dest1)
1809                         free(pktopt->ip6po_dest1, M_IP6OPT);
1810                 pktopt->ip6po_dest1 = NULL;
1811         }
1812         if (optname == -1) {
1813                 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1814                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1815                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1816                 if (pktopt->ip6po_route.ro_rt) {
1817                         RTFREE(pktopt->ip6po_route.ro_rt);
1818                         pktopt->ip6po_route.ro_rt = NULL;
1819                 }
1820         }
1821         if (optname == -1) {
1822                 if (needfree && pktopt->ip6po_dest2)
1823                         free(pktopt->ip6po_dest2, M_IP6OPT);
1824                 pktopt->ip6po_dest2 = NULL;
1825         }
1826 }
1827
1828 #define PKTOPT_EXTHDRCPY(type) \
1829 do {\
1830         if (src->type) {\
1831                 int hlen =\
1832                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1833                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
1834                 if (dst->type == NULL && canwait == M_NOWAIT)\
1835                         goto bad;\
1836                 bcopy(src->type, dst->type, hlen);\
1837         }\
1838 } while (0)
1839
1840 struct ip6_pktopts *
1841 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
1842 {
1843         struct ip6_pktopts *dst;
1844
1845         if (src == NULL) {
1846                 printf("ip6_clearpktopts: invalid argument\n");
1847                 return(NULL);
1848         }
1849
1850         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1851         if (dst == NULL && canwait == M_NOWAIT)
1852                 return (NULL);
1853         bzero(dst, sizeof(*dst));
1854
1855         dst->ip6po_hlim = src->ip6po_hlim;
1856         if (src->ip6po_pktinfo) {
1857                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1858                                             M_IP6OPT, canwait);
1859                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1860                         goto bad;
1861                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1862         }
1863         if (src->ip6po_nexthop) {
1864                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1865                                             M_IP6OPT, canwait);
1866                 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1867                         goto bad;
1868                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1869                       src->ip6po_nexthop->sa_len);
1870         }
1871         PKTOPT_EXTHDRCPY(ip6po_hbh);
1872         PKTOPT_EXTHDRCPY(ip6po_dest1);
1873         PKTOPT_EXTHDRCPY(ip6po_dest2);
1874         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1875         return(dst);
1876
1877   bad:
1878         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1879         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1880         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1881         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1882         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1883         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1884         free(dst, M_IP6OPT);
1885         return(NULL);
1886 }
1887 #undef PKTOPT_EXTHDRCPY
1888
1889 void
1890 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1891 {
1892         if (pktopt == NULL)
1893                 return;
1894
1895         ip6_clearpktopts(pktopt, 1, -1);
1896
1897         free(pktopt, M_IP6OPT);
1898 }
1899
1900 /*
1901  * Set the IP6 multicast options in response to user setsockopt().
1902  */
1903 static int
1904 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1905 {
1906         int error = 0;
1907         u_int loop, ifindex;
1908         struct ipv6_mreq *mreq;
1909         struct ifnet *ifp;
1910         struct ip6_moptions *im6o = *im6op;
1911         struct route_in6 ro;
1912         struct sockaddr_in6 *dst;
1913         struct in6_multi_mship *imm;
1914         struct thread *td = curthread;  /* XXX */
1915
1916         if (im6o == NULL) {
1917                 /*
1918                  * No multicast option buffer attached to the pcb;
1919                  * allocate one and initialize to default values.
1920                  */
1921                 im6o = (struct ip6_moptions *)
1922                         malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1923
1924                 if (im6o == NULL)
1925                         return(ENOBUFS);
1926                 *im6op = im6o;
1927                 im6o->im6o_multicast_ifp = NULL;
1928                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1929                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1930                 LIST_INIT(&im6o->im6o_memberships);
1931         }
1932
1933         switch (optname) {
1934
1935         case IPV6_MULTICAST_IF:
1936                 /*
1937                  * Select the interface for outgoing multicast packets.
1938                  */
1939                 if (m == NULL || m->m_len != sizeof(u_int)) {
1940                         error = EINVAL;
1941                         break;
1942                 }
1943                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1944                 if (ifindex < 0 || if_index < ifindex) {
1945                         error = ENXIO;  /* XXX EINVAL? */
1946                         break;
1947                 }
1948                 ifp = ifindex2ifnet[ifindex];
1949                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1950                         error = EADDRNOTAVAIL;
1951                         break;
1952                 }
1953                 im6o->im6o_multicast_ifp = ifp;
1954                 break;
1955
1956         case IPV6_MULTICAST_HOPS:
1957             {
1958                 /*
1959                  * Set the IP6 hoplimit for outgoing multicast packets.
1960                  */
1961                 int optval;
1962                 if (m == NULL || m->m_len != sizeof(int)) {
1963                         error = EINVAL;
1964                         break;
1965                 }
1966                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1967                 if (optval < -1 || optval >= 256)
1968                         error = EINVAL;
1969                 else if (optval == -1)
1970                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1971                 else
1972                         im6o->im6o_multicast_hlim = optval;
1973                 break;
1974             }
1975
1976         case IPV6_MULTICAST_LOOP:
1977                 /*
1978                  * Set the loopback flag for outgoing multicast packets.
1979                  * Must be zero or one.
1980                  */
1981                 if (m == NULL || m->m_len != sizeof(u_int)) {
1982                         error = EINVAL;
1983                         break;
1984                 }
1985                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1986                 if (loop > 1) {
1987                         error = EINVAL;
1988                         break;
1989                 }
1990                 im6o->im6o_multicast_loop = loop;
1991                 break;
1992
1993         case IPV6_JOIN_GROUP:
1994                 /*
1995                  * Add a multicast group membership.
1996                  * Group must be a valid IP6 multicast address.
1997                  */
1998                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1999                         error = EINVAL;
2000                         break;
2001                 }
2002                 mreq = mtod(m, struct ipv6_mreq *);
2003                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2004                         /*
2005                          * We use the unspecified address to specify to accept
2006                          * all multicast addresses. Only super user is allowed
2007                          * to do this.
2008                          */
2009                         if (suser(td))
2010                         {
2011                                 error = EACCES;
2012                                 break;
2013                         }
2014                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2015                         error = EINVAL;
2016                         break;
2017                 }
2018
2019                 /*
2020                  * If the interface is specified, validate it.
2021                  */
2022                 if (mreq->ipv6mr_interface < 0
2023                  || if_index < mreq->ipv6mr_interface) {
2024                         error = ENXIO;  /* XXX EINVAL? */
2025                         break;
2026                 }
2027                 /*
2028                  * If no interface was explicitly specified, choose an
2029                  * appropriate one according to the given multicast address.
2030                  */
2031                 if (mreq->ipv6mr_interface == 0) {
2032                         /*
2033                          * If the multicast address is in node-local scope,
2034                          * the interface should be a loopback interface.
2035                          * Otherwise, look up the routing table for the
2036                          * address, and choose the outgoing interface.
2037                          *   XXX: is it a good approach?
2038                          */
2039                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2040                                 ifp = &loif[0];
2041                         } else {
2042                                 ro.ro_rt = NULL;
2043                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2044                                 bzero(dst, sizeof(*dst));
2045                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2046                                 dst->sin6_family = AF_INET6;
2047                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2048                                 rtalloc((struct route *)&ro);
2049                                 if (ro.ro_rt == NULL) {
2050                                         error = EADDRNOTAVAIL;
2051                                         break;
2052                                 }
2053                                 ifp = ro.ro_rt->rt_ifp;
2054                                 rtfree(ro.ro_rt);
2055                         }
2056                 } else
2057                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2058
2059                 /*
2060                  * See if we found an interface, and confirm that it
2061                  * supports multicast
2062                  */
2063                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2064                         error = EADDRNOTAVAIL;
2065                         break;
2066                 }
2067                 /*
2068                  * Put interface index into the multicast address,
2069                  * if the address has link-local scope.
2070                  */
2071                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2072                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2073                                 = htons(mreq->ipv6mr_interface);
2074                 }
2075                 /*
2076                  * See if the membership already exists.
2077                  */
2078                 for (imm = im6o->im6o_memberships.lh_first;
2079                      imm != NULL; imm = imm->i6mm_chain.le_next)
2080                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2081                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2082                                                &mreq->ipv6mr_multiaddr))
2083                                 break;
2084                 if (imm != NULL) {
2085                         error = EADDRINUSE;
2086                         break;
2087                 }
2088                 /*
2089                  * Everything looks good; add a new record to the multicast
2090                  * address list for the given interface.
2091                  */
2092                 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2093                 if (imm == NULL) {
2094                         error = ENOBUFS;
2095                         break;
2096                 }
2097                 if ((imm->i6mm_maddr =
2098                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2099                         free(imm, M_IPMADDR);
2100                         break;
2101                 }
2102                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2103                 break;
2104
2105         case IPV6_LEAVE_GROUP:
2106                 /*
2107                  * Drop a multicast group membership.
2108                  * Group must be a valid IP6 multicast address.
2109                  */
2110                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2111                         error = EINVAL;
2112                         break;
2113                 }
2114                 mreq = mtod(m, struct ipv6_mreq *);
2115                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2116                         if (suser(td)) {
2117                                 error = EACCES;
2118                                 break;
2119                         }
2120                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2121                         error = EINVAL;
2122                         break;
2123                 }
2124                 /*
2125                  * If an interface address was specified, get a pointer
2126                  * to its ifnet structure.
2127                  */
2128                 if (mreq->ipv6mr_interface < 0
2129                  || if_index < mreq->ipv6mr_interface) {
2130                         error = ENXIO;  /* XXX EINVAL? */
2131                         break;
2132                 }
2133                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2134                 /*
2135                  * Put interface index into the multicast address,
2136                  * if the address has link-local scope.
2137                  */
2138                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2139                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2140                                 = htons(mreq->ipv6mr_interface);
2141                 }
2142                 /*
2143                  * Find the membership in the membership list.
2144                  */
2145                 for (imm = im6o->im6o_memberships.lh_first;
2146                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2147                         if ((ifp == NULL ||
2148                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2149                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2150                                                &mreq->ipv6mr_multiaddr))
2151                                 break;
2152                 }
2153                 if (imm == NULL) {
2154                         /* Unable to resolve interface */
2155                         error = EADDRNOTAVAIL;
2156                         break;
2157                 }
2158                 /*
2159                  * Give up the multicast address record to which the
2160                  * membership points.
2161                  */
2162                 LIST_REMOVE(imm, i6mm_chain);
2163                 in6_delmulti(imm->i6mm_maddr);
2164                 free(imm, M_IPMADDR);
2165                 break;
2166
2167         default:
2168                 error = EOPNOTSUPP;
2169                 break;
2170         }
2171
2172         /*
2173          * If all options have default values, no need to keep the mbuf.
2174          */
2175         if (im6o->im6o_multicast_ifp == NULL &&
2176             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2177             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2178             im6o->im6o_memberships.lh_first == NULL) {
2179                 free(*im6op, M_IPMOPTS);
2180                 *im6op = NULL;
2181         }
2182
2183         return(error);
2184 }
2185
2186 /*
2187  * Return the IP6 multicast options in response to user getsockopt().
2188  */
2189 static int
2190 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2191 {
2192         u_int *hlim, *loop, *ifindex;
2193
2194         *mp = m_get(MB_WAIT, MT_HEADER);                /* XXX */
2195
2196         switch (optname) {
2197
2198         case IPV6_MULTICAST_IF:
2199                 ifindex = mtod(*mp, u_int *);
2200                 (*mp)->m_len = sizeof(u_int);
2201                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2202                         *ifindex = 0;
2203                 else
2204                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2205                 return(0);
2206
2207         case IPV6_MULTICAST_HOPS:
2208                 hlim = mtod(*mp, u_int *);
2209                 (*mp)->m_len = sizeof(u_int);
2210                 if (im6o == NULL)
2211                         *hlim = ip6_defmcasthlim;
2212                 else
2213                         *hlim = im6o->im6o_multicast_hlim;
2214                 return(0);
2215
2216         case IPV6_MULTICAST_LOOP:
2217                 loop = mtod(*mp, u_int *);
2218                 (*mp)->m_len = sizeof(u_int);
2219                 if (im6o == NULL)
2220                         *loop = ip6_defmcasthlim;
2221                 else
2222                         *loop = im6o->im6o_multicast_loop;
2223                 return(0);
2224
2225         default:
2226                 return(EOPNOTSUPP);
2227         }
2228 }
2229
2230 /*
2231  * Discard the IP6 multicast options.
2232  */
2233 void
2234 ip6_freemoptions(struct ip6_moptions *im6o)
2235 {
2236         struct in6_multi_mship *imm;
2237
2238         if (im6o == NULL)
2239                 return;
2240
2241         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2242                 LIST_REMOVE(imm, i6mm_chain);
2243                 if (imm->i6mm_maddr)
2244                         in6_delmulti(imm->i6mm_maddr);
2245                 free(imm, M_IPMADDR);
2246         }
2247         free(im6o, M_IPMOPTS);
2248 }
2249
2250 /*
2251  * Set IPv6 outgoing packet options based on advanced API.
2252  */
2253 int
2254 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, int priv,
2255                   int needcopy)
2256 {
2257         struct cmsghdr *cm = 0;
2258
2259         if (control == 0 || opt == 0)
2260                 return(EINVAL);
2261
2262         init_ip6pktopts(opt);
2263
2264         /*
2265          * XXX: Currently, we assume all the optional information is stored
2266          * in a single mbuf.
2267          */
2268         if (control->m_next)
2269                 return(EINVAL);
2270
2271         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2272                      control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2273                 cm = mtod(control, struct cmsghdr *);
2274                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2275                         return(EINVAL);
2276                 if (cm->cmsg_level != IPPROTO_IPV6)
2277                         continue;
2278
2279                 /*
2280                  * XXX should check if RFC2292 API is mixed with 2292bis API
2281                  */
2282                 switch (cm->cmsg_type) {
2283                 case IPV6_PKTINFO:
2284                         if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2285                                 return(EINVAL);
2286                         if (needcopy) {
2287                                 /* XXX: Is it really WAITOK? */
2288                                 opt->ip6po_pktinfo =
2289                                         malloc(sizeof(struct in6_pktinfo),
2290                                                M_IP6OPT, M_WAITOK);
2291                                 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2292                                     sizeof(struct in6_pktinfo));
2293                         } else
2294                                 opt->ip6po_pktinfo =
2295                                         (struct in6_pktinfo *)CMSG_DATA(cm);
2296                         if (opt->ip6po_pktinfo->ipi6_ifindex &&
2297                             IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2298                                 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2299                                         htons(opt->ip6po_pktinfo->ipi6_ifindex);
2300
2301                         if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2302                          || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2303                                 return(ENXIO);
2304                         }
2305
2306                         /*
2307                          * Check if the requested source address is indeed a
2308                          * unicast address assigned to the node, and can be
2309                          * used as the packet's source address.
2310                          */
2311                         if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2312                                 struct in6_ifaddr *ia6;
2313                                 struct sockaddr_in6 sin6;
2314
2315                                 bzero(&sin6, sizeof(sin6));
2316                                 sin6.sin6_len = sizeof(sin6);
2317                                 sin6.sin6_family = AF_INET6;
2318                                 sin6.sin6_addr =
2319                                         opt->ip6po_pktinfo->ipi6_addr;
2320                                 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2321                                 if (ia6 == NULL ||
2322                                     (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2323                                                        IN6_IFF_NOTREADY)) != 0)
2324                                         return(EADDRNOTAVAIL);
2325                         }
2326                         break;
2327
2328                 case IPV6_HOPLIMIT:
2329                         if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2330                                 return(EINVAL);
2331
2332                         opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2333                         if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2334                                 return(EINVAL);
2335                         break;
2336
2337                 case IPV6_NEXTHOP:
2338                         if (!priv)
2339                                 return(EPERM);
2340
2341                         if (cm->cmsg_len < sizeof(u_char) ||
2342                             /* check if cmsg_len is large enough for sa_len */
2343                             cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2344                                 return(EINVAL);
2345
2346                         if (needcopy) {
2347                                 opt->ip6po_nexthop =
2348                                         malloc(*CMSG_DATA(cm),
2349                                                M_IP6OPT, M_WAITOK);
2350                                 bcopy(CMSG_DATA(cm),
2351                                       opt->ip6po_nexthop,
2352                                       *CMSG_DATA(cm));
2353                         } else
2354                                 opt->ip6po_nexthop =
2355                                         (struct sockaddr *)CMSG_DATA(cm);
2356                         break;
2357
2358                 case IPV6_HOPOPTS:
2359                 {
2360                         struct ip6_hbh *hbh;
2361                         int hbhlen;
2362
2363                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2364                                 return(EINVAL);
2365                         hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2366                         hbhlen = (hbh->ip6h_len + 1) << 3;
2367                         if (cm->cmsg_len != CMSG_LEN(hbhlen))
2368                                 return(EINVAL);
2369
2370                         if (needcopy) {
2371                                 opt->ip6po_hbh =
2372                                         malloc(hbhlen, M_IP6OPT, M_WAITOK);
2373                                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2374                         } else
2375                                 opt->ip6po_hbh = hbh;
2376                         break;
2377                 }
2378
2379                 case IPV6_DSTOPTS:
2380                 {
2381                         struct ip6_dest *dest, **newdest;
2382                         int destlen;
2383
2384                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2385                                 return(EINVAL);
2386                         dest = (struct ip6_dest *)CMSG_DATA(cm);
2387                         destlen = (dest->ip6d_len + 1) << 3;
2388                         if (cm->cmsg_len != CMSG_LEN(destlen))
2389                                 return(EINVAL);
2390
2391                         /*
2392                          * The old advacned API is ambiguous on this
2393                          * point. Our approach is to determine the
2394                          * position based according to the existence
2395                          * of a routing header. Note, however, that
2396                          * this depends on the order of the extension
2397                          * headers in the ancillary data; the 1st part
2398                          * of the destination options header must
2399                          * appear before the routing header in the
2400                          * ancillary data, too.
2401                          * RFC2292bis solved the ambiguity by
2402                          * introducing separate cmsg types.
2403                          */
2404                         if (opt->ip6po_rthdr == NULL)
2405                                 newdest = &opt->ip6po_dest1;
2406                         else
2407                                 newdest = &opt->ip6po_dest2;
2408
2409                         if (needcopy) {
2410                                 *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2411                                 bcopy(dest, *newdest, destlen);
2412                         } else
2413                                 *newdest = dest;
2414
2415                         break;
2416                 }
2417
2418                 case IPV6_RTHDR:
2419                 {
2420                         struct ip6_rthdr *rth;
2421                         int rthlen;
2422
2423                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2424                                 return(EINVAL);
2425                         rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2426                         rthlen = (rth->ip6r_len + 1) << 3;
2427                         if (cm->cmsg_len != CMSG_LEN(rthlen))
2428                                 return(EINVAL);
2429
2430                         switch (rth->ip6r_type) {
2431                         case IPV6_RTHDR_TYPE_0:
2432                                 /* must contain one addr */
2433                                 if (rth->ip6r_len == 0)
2434                                         return(EINVAL);
2435                                 /* length must be even */
2436                                 if (rth->ip6r_len % 2)
2437                                         return(EINVAL);
2438                                 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2439                                         return(EINVAL);
2440                                 break;
2441                         default:
2442                                 return(EINVAL); /* not supported */
2443                         }
2444
2445                         if (needcopy) {
2446                                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2447                                                           M_WAITOK);
2448                                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2449                         } else
2450                                 opt->ip6po_rthdr = rth;
2451
2452                         break;
2453                 }
2454
2455                 default:
2456                         return(ENOPROTOOPT);
2457                 }
2458         }
2459
2460         return(0);
2461 }
2462
2463 /*
2464  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2465  * packet to the input queue of a specified interface.  Note that this
2466  * calls the output routine of the loopback "driver", but with an interface
2467  * pointer that might NOT be &loif -- easier than replicating that code here.
2468  */
2469 void
2470 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2471 {
2472         struct mbuf *copym;
2473         struct ip6_hdr *ip6;
2474
2475         copym = m_copy(m, 0, M_COPYALL);
2476         if (copym == NULL)
2477                 return;
2478
2479         /*
2480          * Make sure to deep-copy IPv6 header portion in case the data
2481          * is in an mbuf cluster, so that we can safely override the IPv6
2482          * header portion later.
2483          */
2484         if ((copym->m_flags & M_EXT) != 0 ||
2485             copym->m_len < sizeof(struct ip6_hdr)) {
2486                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2487                 if (copym == NULL)
2488                         return;
2489         }
2490
2491 #ifdef DIAGNOSTIC
2492         if (copym->m_len < sizeof(*ip6)) {
2493                 m_freem(copym);
2494                 return;
2495         }
2496 #endif
2497
2498         ip6 = mtod(copym, struct ip6_hdr *);
2499 #ifndef SCOPEDROUTING
2500         /*
2501          * clear embedded scope identifiers if necessary.
2502          * in6_clearscope will touch the addresses only when necessary.
2503          */
2504         in6_clearscope(&ip6->ip6_src);
2505         in6_clearscope(&ip6->ip6_dst);
2506 #endif
2507
2508         (void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2509 }
2510
2511 /*
2512  * Chop IPv6 header off from the payload.
2513  */
2514 static int
2515 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2516 {
2517         struct mbuf *mh;
2518         struct ip6_hdr *ip6;
2519
2520         ip6 = mtod(m, struct ip6_hdr *);
2521         if (m->m_len > sizeof(*ip6)) {
2522                 MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
2523                 if (mh == 0) {
2524                         m_freem(m);
2525                         return ENOBUFS;
2526                 }
2527                 M_MOVE_PKTHDR(mh, m);
2528                 MH_ALIGN(mh, sizeof(*ip6));
2529                 m->m_len -= sizeof(*ip6);
2530                 m->m_data += sizeof(*ip6);
2531                 mh->m_next = m;
2532                 m = mh;
2533                 m->m_len = sizeof(*ip6);
2534                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2535         }
2536         exthdrs->ip6e_ip6 = m;
2537         return 0;
2538 }
2539
2540 /*
2541  * Compute IPv6 extension header length.
2542  */
2543 int
2544 ip6_optlen(struct in6pcb *in6p)
2545 {
2546         int len;
2547
2548         if (!in6p->in6p_outputopts)
2549                 return 0;
2550
2551         len = 0;
2552 #define elen(x) \
2553     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2554
2555         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2556         if (in6p->in6p_outputopts->ip6po_rthdr)
2557                 /* dest1 is valid with rthdr only */
2558                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2559         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2560         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2561         return len;
2562 #undef elen
2563 }