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 1.35 2008/06/17 20:50:11 aggelos 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
 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 copyexthdr (void *, struct mbuf **);
 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 struct mbuf *ip6_splithdr (struct mbuf *);
 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         struct mbuf *mprev;
 157         u_char *nexthdrp;
 158         int hlen, tlen, len, off;
 159         struct route_in6 ip6route;
 160         struct sockaddr_in6 *dst;
 161         int error = 0;
 162         struct in6_ifaddr *ia = NULL;
 163         u_long mtu;
 164         u_int32_t optlen, plen = 0, unfragpartlen;
 165         struct ip6_exthdrs exthdrs;
 166         struct in6_addr finaldst;
 167         struct route_in6 *ro_pmtu = NULL;
 168         boolean_t hdrsplit = FALSE;
 169         boolean_t needipsec = FALSE;
 170 #ifdef IPSEC
 171         boolean_t needipsectun = FALSE;
 172         struct secpolicy *sp = NULL;
 173         struct socket *so = inp ? inp->inp_socket : NULL;
 174
 175         ip6 = mtod(m, struct ip6_hdr *);
 176 #endif
 177 #ifdef FAST_IPSEC
 178         boolean_t needipsectun = FALSE;
 179         struct secpolicy *sp = NULL;
 180
 181         ip6 = mtod(m, struct ip6_hdr *);
 182 #endif
 183
 184         bzero(&exthdrs, sizeof exthdrs);
 185
 186         if (opt) {
 187                 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
 188                         goto freehdrs;
 189                 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
 190                         goto freehdrs;
 191                 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
 192                         goto freehdrs;
 193                 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
 194                         goto freehdrs;
 195         }
 196
 197 #ifdef IPSEC
 198         /* get a security policy for this packet */
 199         if (so == NULL)
 200                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 201         else
 202                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 203
 204         if (sp == NULL) {
 205                 ipsec6stat.out_inval++;
 206                 goto freehdrs;
 207         }
 208
 209         error = 0;
 210
 211         /* check policy */
 212         switch (sp->policy) {
 213         case IPSEC_POLICY_DISCARD:
 214                 /*
 215                  * This packet is just discarded.
 216                  */
 217                 ipsec6stat.out_polvio++;
 218                 goto freehdrs;
 219
 220         case IPSEC_POLICY_BYPASS:
 221         case IPSEC_POLICY_NONE:
 222                 /* no need to do IPsec. */
 223                 needipsec = FALSE;
 224                 break;
 225
 226         case IPSEC_POLICY_IPSEC:
 227                 if (sp->req == NULL) {
 228                         error = key_spdacquire(sp);     /* acquire a policy */
 229                         goto freehdrs;
 230                 }
 231                 needipsec = TRUE;
 232                 break;
 233
 234         case IPSEC_POLICY_ENTRUST:
 235         default:
 236                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 237         }
 238 #endif /* IPSEC */
 239 #ifdef FAST_IPSEC
 240         /* get a security policy for this packet */
 241         if (inp == NULL)
 242                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 243         else
 244                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 245
 246         if (sp == NULL) {
 247                 newipsecstat.ips_out_inval++;
 248                 goto freehdrs;
 249         }
 250
 251         error = 0;
 252
 253         /* check policy */
 254         switch (sp->policy) {
 255         case IPSEC_POLICY_DISCARD:
 256                 /*
 257                  * This packet is just discarded.
 258                  */
 259                 newipsecstat.ips_out_polvio++;
 260                 goto freehdrs;
 261
 262         case IPSEC_POLICY_BYPASS:
 263         case IPSEC_POLICY_NONE:
 264                 /* no need to do IPsec. */
 265                 needipsec = FALSE;
 266                 break;
 267
 268         case IPSEC_POLICY_IPSEC:
 269                 if (sp->req == NULL) {
 270                         error = key_spdacquire(sp);     /* acquire a policy */
 271                         goto freehdrs;
 272                 }
 273                 needipsec = TRUE;
 274                 break;
 275
 276         case IPSEC_POLICY_ENTRUST:
 277         default:
 278                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 279         }
 280 #endif /* FAST_IPSEC */
 281
 282         /*
 283          * Calculate the total length of the extension header chain.
 284          * Keep the length of the unfragmentable part for fragmentation.
 285          */
 286         optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
 287             m_lengthm(exthdrs.ip6e_dest1, NULL) +
 288             m_lengthm(exthdrs.ip6e_rthdr, NULL);
 289
 290         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 291
 292         /* NOTE: we don't add AH/ESP length here. do that later. */
 293         optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
 294
 295         /*
 296          * If we need IPsec, or there is at least one extension header,
 297          * separate IP6 header from the payload.
 298          */
 299         if ((needipsec || optlen) && !hdrsplit) {
 300                 exthdrs.ip6e_ip6 = ip6_splithdr(m);
 301                 if (exthdrs.ip6e_ip6 == NULL) {
 302                         error = ENOBUFS;
 303                         goto freehdrs;
 304                 }
 305                 m = exthdrs.ip6e_ip6;
 306                 hdrsplit = TRUE;
 307         }
 308
 309         /* adjust pointer */
 310         ip6 = mtod(m, struct ip6_hdr *);
 311
 312         /* adjust mbuf packet header length */
 313         m->m_pkthdr.len += optlen;
 314         plen = m->m_pkthdr.len - sizeof(*ip6);
 315
 316         /* If this is a jumbo payload, insert a jumbo payload option. */
 317         if (plen > IPV6_MAXPACKET) {
 318                 if (!hdrsplit) {
 319                         exthdrs.ip6e_ip6 = ip6_splithdr(m);
 320                         if (exthdrs.ip6e_ip6 == NULL) {
 321                                 error = ENOBUFS;
 322                                 goto freehdrs;
 323                         }
 324                         m = exthdrs.ip6e_ip6;
 325                         hdrsplit = TRUE;
 326                 }
 327                 /* adjust pointer */
 328                 ip6 = mtod(m, struct ip6_hdr *);
 329                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 330                         goto freehdrs;
 331                 ip6->ip6_plen = 0;
 332         } else
 333                 ip6->ip6_plen = htons(plen);
 334
 335         /*
 336          * Concatenate headers and fill in next header fields.
 337          * Here we have, on "m"
 338          *      IPv6 payload
 339          * and we insert headers accordingly.  Finally, we should be getting:
 340          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 341          *
 342          * during the header composing process, "m" points to IPv6 header.
 343          * "mprev" points to an extension header prior to esp.
 344          */
 345
 346         nexthdrp = &ip6->ip6_nxt;
 347         mprev = m;
 348
 349         /*
 350          * we treat dest2 specially.  this makes IPsec processing
 351          * much easier.  the goal here is to make mprev point the
 352          * mbuf prior to dest2.
 353          *
 354          * result: IPv6 dest2 payload
 355          * m and mprev will point to IPv6 header.
 356          */
 357         if (exthdrs.ip6e_dest2) {
 358                 if (!hdrsplit)
 359                         panic("assumption failed: hdr not split");
 360                 exthdrs.ip6e_dest2->m_next = m->m_next;
 361                 m->m_next = exthdrs.ip6e_dest2;
 362                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 363                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 364         }
 365
 366 /*
 367  * Place m1 after mprev.
 368  */
 369 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
 370     do {\
 371         if (m1) {\
 372                 if (!hdrsplit)\
 373                         panic("assumption failed: hdr not split");\
 374                 *mtod(m1, u_char *) = *nexthdrp;\
 375                 *nexthdrp = (i);\
 376                 nexthdrp = mtod(m1, u_char *);\
 377                 m1->m_next = mprev->m_next;\
 378                 mprev->m_next = m1;\
 379                 mprev = m1;\
 380         }\
 381     } while (0)
 382
 383         /*
 384          * result: IPv6 hbh dest1 rthdr dest2 payload
 385          * m will point to IPv6 header.  mprev will point to the
 386          * extension header prior to dest2 (rthdr in the above case).
 387          */
 388         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 389         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
 390         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
 391
 392 #if defined(IPSEC) || defined(FAST_IPSEC)
 393         if (needipsec) {
 394                 struct ipsec_output_state state;
 395                 int segleft_org = 0;
 396                 struct ip6_rthdr *rh = NULL;
 397
 398                 /*
 399                  * pointers after IPsec headers are not valid any more.
 400                  * other pointers need a great care too.
 401                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 402                  */
 403                 exthdrs.ip6e_dest2 = NULL;
 404
 405                 if (exthdrs.ip6e_rthdr) {
 406                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 407                         segleft_org = rh->ip6r_segleft;
 408                         rh->ip6r_segleft = 0;
 409                 }
 410
 411                 bzero(&state, sizeof state);
 412                 state.m = m;
 413                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 414                                             &needipsectun);
 415                 m = state.m;
 416                 if (error) {
 417                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 418                         m = NULL;
 419                         switch (error) {
 420                         case EHOSTUNREACH:
 421                         case ENETUNREACH:
 422                         case EMSGSIZE:
 423                         case ENOBUFS:
 424                         case ENOMEM:
 425                                 break;
 426                         default:
 427                                 kprintf("ip6_output (ipsec): error code %d\n",
 428                                        error);
 429                                 /* fall through */
 430                         case ENOENT:
 431                                 /* don't show these error codes to the user */
 432                                 error = 0;
 433                                 break;
 434                         }
 435                         goto bad;
 436                 }
 437                 if (exthdrs.ip6e_rthdr) {
 438                         /* ah6_output doesn't modify mbuf chain */
 439                         rh->ip6r_segleft = segleft_org;
 440                 }
 441         }
 442 #endif
 443
 444         /*
 445          * If there is a routing header, replace destination address field
 446          * with the first hop of the routing header.
 447          */
 448         if (exthdrs.ip6e_rthdr) {
 449                 struct ip6_rthdr *rh;
 450                 struct ip6_rthdr0 *rh0;
 451
 452                 finaldst = ip6->ip6_dst;
 453                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 454                 switch (rh->ip6r_type) {
 455                 case IPV6_RTHDR_TYPE_0:
 456                          rh0 = (struct ip6_rthdr0 *)rh;
 457                          ip6->ip6_dst = rh0->ip6r0_addr[0];
 458                          bcopy(&rh0->ip6r0_addr[1], &rh0->ip6r0_addr[0],
 459                              sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1));
 460                          rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
 461                          break;
 462                 default:        /* is it possible? */
 463                          error = EINVAL;
 464                          goto bad;
 465                 }
 466         }
 467
 468         /* Source address validation */
 469         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 470             !(flags & IPV6_DADOUTPUT)) {
 471                 error = EOPNOTSUPP;
 472                 ip6stat.ip6s_badscope++;
 473                 goto bad;
 474         }
 475         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 476                 error = EOPNOTSUPP;
 477                 ip6stat.ip6s_badscope++;
 478                 goto bad;
 479         }
 480
 481         ip6stat.ip6s_localout++;
 482
 483         /*
 484          * Route packet.
 485          */
 486         if (ro == NULL) {
 487                 ro = &ip6route;
 488                 bzero(ro, sizeof(*ro));
 489         }
 490         ro_pmtu = ro;
 491         if (opt && opt->ip6po_rthdr)
 492                 ro = &opt->ip6po_route;
 493         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 494         /*
 495          * If there is a cached route,
 496          * check that it is to the same destination
 497          * and is still up. If not, free it and try again.
 498          */
 499         if (ro->ro_rt != NULL &&
 500             (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
 501              !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 502                 RTFREE(ro->ro_rt);
 503                 ro->ro_rt = NULL;
 504         }
 505         if (ro->ro_rt == NULL) {
 506                 bzero(dst, sizeof(*dst));
 507                 dst->sin6_family = AF_INET6;
 508                 dst->sin6_len = sizeof(struct sockaddr_in6);
 509                 dst->sin6_addr = ip6->ip6_dst;
 510         }
 511 #if defined(IPSEC) || defined(FAST_IPSEC)
 512         if (needipsec && needipsectun) {
 513                 struct ipsec_output_state state;
 514
 515                 /*
 516                  * All the extension headers will become inaccessible
 517                  * (since they can be encrypted).
 518                  * Don't panic, we need no more updates to extension headers
 519                  * on inner IPv6 packet (since they are now encapsulated).
 520                  *
 521                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 522                  */
 523                 bzero(&exthdrs, sizeof(exthdrs));
 524                 exthdrs.ip6e_ip6 = m;
 525
 526                 bzero(&state, sizeof(state));
 527                 state.m = m;
 528                 state.ro = (struct route *)ro;
 529                 state.dst = (struct sockaddr *)dst;
 530
 531                 error = ipsec6_output_tunnel(&state, sp, flags);
 532
 533                 m = state.m;
 534                 ro = (struct route_in6 *)state.ro;
 535                 dst = (struct sockaddr_in6 *)state.dst;
 536                 if (error) {
 537                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 538                         m0 = m = NULL;
 539                         m = NULL;
 540                         switch (error) {
 541                         case EHOSTUNREACH:
 542                         case ENETUNREACH:
 543                         case EMSGSIZE:
 544                         case ENOBUFS:
 545                         case ENOMEM:
 546                                 break;
 547                         default:
 548                                 kprintf("ip6_output (ipsec): error code %d\n", error);
 549                                 /* fall through */
 550                         case ENOENT:
 551                                 /* don't show these error codes to the user */
 552                                 error = 0;
 553                                 break;
 554                         }
 555                         goto bad;
 556                 }
 557
 558                 exthdrs.ip6e_ip6 = m;
 559         }
 560 #endif
 561
 562         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 563                 /* Unicast */
 564
 565 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 566 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 567                 /* xxx
 568                  * interface selection comes here
 569                  * if an interface is specified from an upper layer,
 570                  * ifp must point it.
 571                  */
 572                 if (ro->ro_rt == NULL) {
 573                         /*
 574                          * non-bsdi always clone routes, if parent is
 575                          * PRF_CLONING.
 576                          */
 577                         rtalloc((struct route *)ro);
 578                 }
 579                 if (ro->ro_rt == NULL) {
 580                         ip6stat.ip6s_noroute++;
 581                         error = EHOSTUNREACH;
 582                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 583                         goto bad;
 584                 }
 585                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 586                 ifp = ro->ro_rt->rt_ifp;
 587                 ro->ro_rt->rt_use++;
 588                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 589                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 590                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 591
 592                 in6_ifstat_inc(ifp, ifs6_out_request);
 593
 594                 /*
 595                  * Check if the outgoing interface conflicts with
 596                  * the interface specified by ifi6_ifindex (if specified).
 597                  * Note that loopback interface is always okay.
 598                  * (this may happen when we are sending a packet to one of
 599                  *  our own addresses.)
 600                  */
 601                 if (opt && opt->ip6po_pktinfo
 602                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 603                         if (!(ifp->if_flags & IFF_LOOPBACK)
 604                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 605                                 ip6stat.ip6s_noroute++;
 606                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 607                                 error = EHOSTUNREACH;
 608                                 goto bad;
 609                         }
 610                 }
 611
 612                 if (opt && opt->ip6po_hlim != -1)
 613                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 614         } else {
 615                 /* Multicast */
 616                 struct  in6_multi *in6m;
 617
 618                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 619
 620                 /*
 621                  * See if the caller provided any multicast options
 622                  */
 623                 ifp = NULL;
 624                 if (im6o != NULL) {
 625                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 626                         if (im6o->im6o_multicast_ifp != NULL)
 627                                 ifp = im6o->im6o_multicast_ifp;
 628                 } else
 629                         ip6->ip6_hlim = ip6_defmcasthlim;
 630
 631                 /*
 632                  * See if the caller provided the outgoing interface
 633                  * as an ancillary data.
 634                  * Boundary check for ifindex is assumed to be already done.
 635                  */
 636                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 637                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 638
 639                 /*
 640                  * If the destination is a node-local scope multicast,
 641                  * the packet should be loop-backed only.
 642                  */
 643                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 644                         /*
 645                          * If the outgoing interface is already specified,
 646                          * it should be a loopback interface.
 647                          */
 648                         if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
 649                                 ip6stat.ip6s_badscope++;
 650                                 error = ENETUNREACH; /* XXX: better error? */
 651                                 /* XXX correct ifp? */
 652                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 653                                 goto bad;
 654                         } else {
 655                                 ifp = &loif[0];
 656                         }
 657                 }
 658
 659                 if (opt && opt->ip6po_hlim != -1)
 660                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 661
 662                 /*
 663                  * If caller did not provide an interface lookup a
 664                  * default in the routing table.  This is either a
 665                  * default for the speicfied group (i.e. a host
 666                  * route), or a multicast default (a route for the
 667                  * ``net'' ff00::/8).
 668                  */
 669                 if (ifp == NULL) {
 670                         if (ro->ro_rt == NULL) {
 671                                 ro->ro_rt =
 672                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 673                         }
 674                         if (ro->ro_rt == NULL) {
 675                                 ip6stat.ip6s_noroute++;
 676                                 error = EHOSTUNREACH;
 677                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 678                                 goto bad;
 679                         }
 680                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 681                         ifp = ro->ro_rt->rt_ifp;
 682                         ro->ro_rt->rt_use++;
 683                 }
 684
 685                 if (!(flags & IPV6_FORWARDING))
 686                         in6_ifstat_inc(ifp, ifs6_out_request);
 687                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 688
 689                 /*
 690                  * Confirm that the outgoing interface supports multicast.
 691                  */
 692                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 693                         ip6stat.ip6s_noroute++;
 694                         in6_ifstat_inc(ifp, ifs6_out_discard);
 695                         error = ENETUNREACH;
 696                         goto bad;
 697                 }
 698                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 699                 if (in6m != NULL &&
 700                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 701                         /*
 702                          * If we belong to the destination multicast group
 703                          * on the outgoing interface, and the caller did not
 704                          * forbid loopback, loop back a copy.
 705                          */
 706                         ip6_mloopback(ifp, m, dst);
 707                 } else {
 708                         /*
 709                          * If we are acting as a multicast router, perform
 710                          * multicast forwarding as if the packet had just
 711                          * arrived on the interface to which we are about
 712                          * to send.  The multicast forwarding function
 713                          * recursively calls this function, using the
 714                          * IPV6_FORWARDING flag to prevent infinite recursion.
 715                          *
 716                          * Multicasts that are looped back by ip6_mloopback(),
 717                          * above, will be forwarded by the ip6_input() routine,
 718                          * if necessary.
 719                          */
 720                         if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
 721                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 722                                         m_freem(m);
 723                                         goto done;
 724                                 }
 725                         }
 726                 }
 727                 /*
 728                  * Multicasts with a hoplimit of zero may be looped back,
 729                  * above, but must not be transmitted on a network.
 730                  * Also, multicasts addressed to the loopback interface
 731                  * are not sent -- the above call to ip6_mloopback() will
 732                  * loop back a copy if this host actually belongs to the
 733                  * destination group on the loopback interface.
 734                  */
 735                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 736                         m_freem(m);
 737                         goto done;
 738                 }
 739         }
 740
 741         /*
 742          * Fill the outgoing inteface to tell the upper layer
 743          * to increment per-interface statistics.
 744          */
 745         if (ifpp)
 746                 *ifpp = ifp;
 747
 748         /*
 749          * Determine path MTU.
 750          */
 751         if (ro_pmtu != ro) {
 752                 /* The first hop and the final destination may differ. */
 753                 struct sockaddr_in6 *sin6_fin =
 754                         (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
 755
 756                 if (ro_pmtu->ro_rt != NULL &&
 757                     (!(ro->ro_rt->rt_flags & RTF_UP) ||
 758                      !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr, &finaldst))) {
 759                         RTFREE(ro_pmtu->ro_rt);
 760                         ro_pmtu->ro_rt = NULL;
 761                 }
 762                 if (ro_pmtu->ro_rt == NULL) {
 763                         bzero(sin6_fin, sizeof(*sin6_fin));
 764                         sin6_fin->sin6_family = AF_INET6;
 765                         sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
 766                         sin6_fin->sin6_addr = finaldst;
 767
 768                         rtalloc((struct route *)ro_pmtu);
 769                 }
 770         }
 771         if (ro_pmtu->ro_rt != NULL) {
 772                 u_int32_t ifmtu = ND_IFINFO(ifp)->linkmtu;
 773
 774                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
 775                 if (mtu > ifmtu || mtu == 0) {
 776                         /*
 777                          * The MTU on the route is larger than the MTU on
 778                          * the interface!  This shouldn't happen, unless the
 779                          * MTU of the interface has been changed after the
 780                          * interface was brought up.  Change the MTU in the
 781                          * route to match the interface MTU (as long as the
 782                          * field isn't locked).
 783                          *
 784                          * if MTU on the route is 0, we need to fix the MTU.
 785                          * this case happens with path MTU discovery timeouts.
 786                          */
 787                          mtu = ifmtu;
 788                          if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
 789                                  ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
 790                 }
 791         } else {
 792                 mtu = ND_IFINFO(ifp)->linkmtu;
 793         }
 794
 795         /*
 796          * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
 797          */
 798         if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
 799                 mtu = IPV6_MMTU;
 800
 801         /* Fake scoped addresses */
 802         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 803                 /*
 804                  * If source or destination address is a scoped address, and
 805                  * the packet is going to be sent to a loopback interface,
 806                  * we should keep the original interface.
 807                  */
 808
 809                 /*
 810                  * XXX: this is a very experimental and temporary solution.
 811                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 812                  * field of the structure here.
 813                  * We rely on the consistency between two scope zone ids
 814                  * of source and destination, which should already be assured.
 815                  * Larger scopes than link will be supported in the future.
 816                  */
 817                 origifp = NULL;
 818                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 819                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 820                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 821                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 822                 /*
 823                  * XXX: origifp can be NULL even in those two cases above.
 824                  * For example, if we remove the (only) link-local address
 825                  * from the loopback interface, and try to send a link-local
 826                  * address without link-id information.  Then the source
 827                  * address is ::1, and the destination address is the
 828                  * link-local address with its s6_addr16[1] being zero.
 829                  * What is worse, if the packet goes to the loopback interface
 830                  * by a default rejected route, the null pointer would be
 831                  * passed to looutput, and the kernel would hang.
 832                  * The following last resort would prevent such disaster.
 833                  */
 834                 if (origifp == NULL)
 835                         origifp = ifp;
 836         }
 837         else
 838                 origifp = ifp;
 839         /*
 840          * clear embedded scope identifiers if necessary.
 841          * in6_clearscope will touch the addresses only when necessary.
 842          */
 843         in6_clearscope(&ip6->ip6_src);
 844         in6_clearscope(&ip6->ip6_dst);
 845
 846         /*
 847          * Check with the firewall...
 848          */
 849         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 850                 u_short port = 0;
 851
 852                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 853                 /* If ipfw says divert, we have to just drop packet */
 854                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 855                         m_freem(m);
 856                         goto done;
 857                 }
 858                 if (!m) {
 859                         error = EACCES;
 860                         goto done;
 861                 }
 862         }
 863
 864         /*
 865          * If the outgoing packet contains a hop-by-hop options header,
 866          * it must be examined and processed even by the source node.
 867          * (RFC 2460, section 4.)
 868          */
 869         if (exthdrs.ip6e_hbh) {
 870                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 871                 u_int32_t dummy1; /* XXX unused */
 872                 u_int32_t dummy2; /* XXX unused */
 873
 874 #ifdef DIAGNOSTIC
 875                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 876                         panic("ip6e_hbh is not continuous");
 877 #endif
 878                 /*
 879                  *  XXX: if we have to send an ICMPv6 error to the sender,
 880                  *       we need the M_LOOP flag since icmp6_error() expects
 881                  *       the IPv6 and the hop-by-hop options header are
 882                  *       continuous unless the flag is set.
 883                  */
 884                 m->m_flags |= M_LOOP;
 885                 m->m_pkthdr.rcvif = ifp;
 886                 if (ip6_process_hopopts(m,
 887                                         (u_int8_t *)(hbh + 1),
 888                                         ((hbh->ip6h_len + 1) << 3) -
 889                                         sizeof(struct ip6_hbh),
 890                                         &dummy1, &dummy2) < 0) {
 891                         /* m was already freed at this point */
 892                         error = EINVAL;/* better error? */
 893                         goto done;
 894                 }
 895                 m->m_flags &= ~M_LOOP; /* XXX */
 896                 m->m_pkthdr.rcvif = NULL;
 897         }
 898
 899         /*
 900          * Run through list of hooks for output packets.
 901          */
 902         if (pfil_has_hooks(&inet6_pfil_hook)) {
 903                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 904                 if (error != 0 || m == NULL)
 905                         goto done;
 906                 ip6 = mtod(m, struct ip6_hdr *);
 907         }
 908
 909         /*
 910          * Send the packet to the outgoing interface.
 911          * If necessary, do IPv6 fragmentation before sending.
 912          */
 913         tlen = m->m_pkthdr.len;
 914         if (tlen <= mtu
 915 #ifdef notyet
 916             /*
 917              * On any link that cannot convey a 1280-octet packet in one piece,
 918              * link-specific fragmentation and reassembly must be provided at
 919              * a layer below IPv6. [RFC 2460, sec.5]
 920              * Thus if the interface has ability of link-level fragmentation,
 921              * we can just send the packet even if the packet size is
 922              * larger than the link's MTU.
 923              * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
 924              */
 925
 926             || ifp->if_flags & IFF_FRAGMENTABLE
 927 #endif
 928             )
 929         {
 930                 /* Record statistics for this interface address. */
 931                 if (ia && !(flags & IPV6_FORWARDING)) {
 932                         ia->ia_ifa.if_opackets++;
 933                         ia->ia_ifa.if_obytes += m->m_pkthdr.len;
 934                 }
 935 #ifdef IPSEC
 936                 /* clean ipsec history once it goes out of the node */
 937                 ipsec_delaux(m);
 938 #endif
 939                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 940                 goto done;
 941         } else if (mtu < IPV6_MMTU) {
 942                 /*
 943                  * note that path MTU is never less than IPV6_MMTU
 944                  * (see icmp6_input).
 945                  */
 946                 error = EMSGSIZE;
 947                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 948                 goto bad;
 949         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 950                 error = EMSGSIZE;
 951                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 952                 goto bad;
 953         } else {
 954                 struct mbuf **mnext, *m_frgpart;
 955                 struct ip6_frag *ip6f;
 956                 u_int32_t id = htonl(ip6_id++);
 957                 u_char nextproto;
 958
 959                 /*
 960                  * Too large for the destination or interface;
 961                  * fragment if possible.
 962                  * Must be able to put at least 8 bytes per fragment.
 963                  */
 964                 hlen = unfragpartlen;
 965                 if (mtu > IPV6_MAXPACKET)
 966                         mtu = IPV6_MAXPACKET;
 967
 968                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 969                 if (len < 8) {
 970                         error = EMSGSIZE;
 971                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 972                         goto bad;
 973                 }
 974
 975                 mnext = &m->m_nextpkt;
 976
 977                 /*
 978                  * Change the next header field of the last header in the
 979                  * unfragmentable part.
 980                  */
 981                 if (exthdrs.ip6e_rthdr) {
 982                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
 983                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
 984                 } else if (exthdrs.ip6e_dest1) {
 985                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
 986                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
 987                 } else if (exthdrs.ip6e_hbh) {
 988                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
 989                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
 990                 } else {
 991                         nextproto = ip6->ip6_nxt;
 992                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
 993                 }
 994
 995                 /*
 996                  * Loop through length of segment after first fragment,
 997                  * make new header and copy data of each part and link onto
 998                  * chain.
 999                  */
1000                 m0 = m;
1001                 for (off = hlen; off < tlen; off += len) {
1002                         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1003                         if (!m) {
1004                                 error = ENOBUFS;
1005                                 ip6stat.ip6s_odropped++;
1006                                 goto sendorfree;
1007                         }
1008                         m->m_pkthdr.rcvif = NULL;
1009                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1010                         *mnext = m;
1011                         mnext = &m->m_nextpkt;
1012                         m->m_data += max_linkhdr;
1013                         mhip6 = mtod(m, struct ip6_hdr *);
1014                         *mhip6 = *ip6;
1015                         m->m_len = sizeof(*mhip6);
1016                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1017                         if (error) {
1018                                 ip6stat.ip6s_odropped++;
1019                                 goto sendorfree;
1020                         }
1021                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1022                         if (off + len >= tlen)
1023                                 len = tlen - off;
1024                         else
1025                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1026                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1027                                                           sizeof(*ip6f) -
1028                                                           sizeof(struct ip6_hdr)));
1029                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1030                                 error = ENOBUFS;
1031                                 ip6stat.ip6s_odropped++;
1032                                 goto sendorfree;
1033                         }
1034                         m_cat(m, m_frgpart);
1035                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1036                         m->m_pkthdr.rcvif = (struct ifnet *)0;
1037                         ip6f->ip6f_reserved = 0;
1038                         ip6f->ip6f_ident = id;
1039                         ip6f->ip6f_nxt = nextproto;
1040                         ip6stat.ip6s_ofragments++;
1041                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1042                 }
1043
1044                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1045         }
1046
1047         /*
1048          * Remove leading garbages.
1049          */
1050 sendorfree:
1051         m = m0->m_nextpkt;
1052         m0->m_nextpkt = NULL;
1053         m_freem(m0);
1054         for (m0 = m; m; m = m0) {
1055                 m0 = m->m_nextpkt;
1056                 m->m_nextpkt = NULL;
1057                 if (error == 0) {
1058                         /* Record statistics for this interface address. */
1059                         if (ia) {
1060                                 ia->ia_ifa.if_opackets++;
1061                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1062                         }
1063 #ifdef IPSEC
1064                         /* clean ipsec history once it goes out of the node */
1065                         ipsec_delaux(m);
1066 #endif
1067                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1068                 } else
1069                         m_freem(m);
1070         }
1071
1072         if (error == 0)
1073                 ip6stat.ip6s_fragmented++;
1074
1075 done:
1076         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1077                 RTFREE(ro->ro_rt);
1078         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1079                 RTFREE(ro_pmtu->ro_rt);
1080         }
1081
1082 #ifdef IPSEC
1083         if (sp != NULL)
1084                 key_freesp(sp);
1085 #endif
1086 #ifdef FAST_IPSEC
1087         if (sp != NULL)
1088                 KEY_FREESP(&sp);
1089 #endif
1090
1091         return (error);
1092
1093 freehdrs:
1094         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1095         m_freem(exthdrs.ip6e_dest1);
1096         m_freem(exthdrs.ip6e_rthdr);
1097         m_freem(exthdrs.ip6e_dest2);
1098         /* fall through */
1099 bad:
1100         m_freem(m);
1101         goto done;
1102 }
1103
1104 static int
1105 copyexthdr(void *h, struct mbuf **mp)
1106 {
1107         struct ip6_ext *hdr = h;
1108         int hlen;
1109         struct mbuf *m;
1110
1111         if (hdr == NULL)
1112                 return 0;
1113
1114         hlen = (hdr->ip6e_len + 1) * 8;
1115         if (hlen > MCLBYTES)
1116                 return ENOBUFS; /* XXX */
1117
1118         m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1119         if (!m)
1120                 return ENOBUFS;
1121         m->m_len = hlen;
1122
1123         bcopy(hdr, mtod(m, caddr_t), hlen);
1124
1125         *mp = m;
1126         return 0;
1127 }
1128
1129 /*
1130  * Insert jumbo payload option.
1131  */
1132 static int
1133 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1134 {
1135         struct mbuf *mopt;
1136         u_char *optbuf;
1137         u_int32_t v;
1138
1139 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1140
1141         /*
1142          * If there is no hop-by-hop options header, allocate new one.
1143          * If there is one but it doesn't have enough space to store the
1144          * jumbo payload option, allocate a cluster to store the whole options.
1145          * Otherwise, use it to store the options.
1146          */
1147         if (exthdrs->ip6e_hbh == NULL) {
1148                 MGET(mopt, MB_DONTWAIT, MT_DATA);
1149                 if (mopt == NULL)
1150                         return (ENOBUFS);
1151                 mopt->m_len = JUMBOOPTLEN;
1152                 optbuf = mtod(mopt, u_char *);
1153                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1154                 exthdrs->ip6e_hbh = mopt;
1155         } else {
1156                 struct ip6_hbh *hbh;
1157
1158                 mopt = exthdrs->ip6e_hbh;
1159                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1160                         /*
1161                          * XXX assumption:
1162                          * - exthdrs->ip6e_hbh is not referenced from places
1163                          *   other than exthdrs.
1164                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1165                          */
1166                         int oldoptlen = mopt->m_len;
1167                         struct mbuf *n;
1168
1169                         /*
1170                          * XXX: give up if the whole (new) hbh header does
1171                          * not fit even in an mbuf cluster.
1172                          */
1173                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1174                                 return (ENOBUFS);
1175
1176                         /*
1177                          * As a consequence, we must always prepare a cluster
1178                          * at this point.
1179                          */
1180                         n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1181                         if (!n)
1182                                 return (ENOBUFS);
1183                         n->m_len = oldoptlen + JUMBOOPTLEN;
1184                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1185                         optbuf = mtod(n, caddr_t) + oldoptlen;
1186                         m_freem(mopt);
1187                         mopt = exthdrs->ip6e_hbh = n;
1188                 } else {
1189                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1190                         mopt->m_len += JUMBOOPTLEN;
1191                 }
1192                 optbuf[0] = IP6OPT_PADN;
1193                 optbuf[1] = 1;
1194
1195                 /*
1196                  * Adjust the header length according to the pad and
1197                  * the jumbo payload option.
1198                  */
1199                 hbh = mtod(mopt, struct ip6_hbh *);
1200                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1201         }
1202
1203         /* fill in the option. */
1204         optbuf[2] = IP6OPT_JUMBO;
1205         optbuf[3] = 4;
1206         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1207         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1208
1209         /* finally, adjust the packet header length */
1210         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1211
1212         return (0);
1213 #undef JUMBOOPTLEN
1214 }
1215
1216 /*
1217  * Insert fragment header and copy unfragmentable header portions.
1218  */
1219 static int
1220 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1221                   struct ip6_frag **frghdrp)
1222 {
1223         struct mbuf *n, *mlast;
1224
1225         if (hlen > sizeof(struct ip6_hdr)) {
1226                 n = m_copym(m0, sizeof(struct ip6_hdr),
1227                             hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1228                 if (n == NULL)
1229                         return (ENOBUFS);
1230                 m->m_next = n;
1231         } else
1232                 n = m;
1233
1234         /* Search for the last mbuf of unfragmentable part. */
1235         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1236                 ;
1237
1238         if (!(mlast->m_flags & M_EXT) &&
1239             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1240                 /* use the trailing space of the last mbuf for the fragment hdr */
1241                 *frghdrp =
1242                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1243                 mlast->m_len += sizeof(struct ip6_frag);
1244                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1245         } else {
1246                 /* allocate a new mbuf for the fragment header */
1247                 struct mbuf *mfrg;
1248
1249                 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1250                 if (mfrg == NULL)
1251                         return (ENOBUFS);
1252                 mfrg->m_len = sizeof(struct ip6_frag);
1253                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1254                 mlast->m_next = mfrg;
1255         }
1256
1257         return (0);
1258 }
1259
1260 /*
1261  * IP6 socket option processing.
1262  */
1263 int
1264 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1265 {
1266         int privileged;
1267         struct inpcb *in6p = so->so_pcb;
1268         int error, optval;
1269         int level, op, optname;
1270         int optlen;
1271         struct thread *td;
1272
1273         if (sopt) {
1274                 level = sopt->sopt_level;
1275                 op = sopt->sopt_dir;
1276                 optname = sopt->sopt_name;
1277                 optlen = sopt->sopt_valsize;
1278                 td = sopt->sopt_td;
1279         } else {
1280                 panic("ip6_ctloutput: arg soopt is NULL");
1281                 /* NOT REACHED */
1282                 td = NULL;
1283         }
1284         error = optval = 0;
1285
1286         privileged = (td == NULL || suser(td)) ? 0 : 1;
1287
1288         if (level == IPPROTO_IPV6) {
1289                 switch (op) {
1290
1291                 case SOPT_SET:
1292                         switch (optname) {
1293                         case IPV6_PKTOPTIONS:
1294                         {
1295                                 struct mbuf *m;
1296
1297                                 error = soopt_getm(sopt, &m); /* XXX */
1298                                 if (error != 0)
1299                                         break;
1300                                 soopt_to_mbuf(sopt, m); /* XXX */
1301                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1302                                                     m, so, sopt);
1303                                 m_freem(m); /* XXX */
1304                                 break;
1305                         }
1306
1307                         /*
1308                          * Use of some Hop-by-Hop options or some
1309                          * Destination options, might require special
1310                          * privilege.  That is, normal applications
1311                          * (without special privilege) might be forbidden
1312                          * from setting certain options in outgoing packets,
1313                          * and might never see certain options in received
1314                          * packets. [RFC 2292 Section 6]
1315                          * KAME specific note:
1316                          *  KAME prevents non-privileged users from sending or
1317                          *  receiving ANY hbh/dst options in order to avoid
1318                          *  overhead of parsing options in the kernel.
1319                          */
1320                         case IPV6_UNICAST_HOPS:
1321                         case IPV6_CHECKSUM:
1322                         case IPV6_FAITH:
1323
1324                         case IPV6_V6ONLY:
1325                                 if (optlen != sizeof(int)) {
1326                                         error = EINVAL;
1327                                         break;
1328                                 }
1329                                 error = soopt_to_kbuf(sopt, &optval,
1330                                         sizeof optval, sizeof optval);
1331                                 if (error)
1332                                         break;
1333                                 switch (optname) {
1334
1335                                 case IPV6_UNICAST_HOPS:
1336                                         if (optval < -1 || optval >= 256)
1337                                                 error = EINVAL;
1338                                         else {
1339                                                 /* -1 = kernel default */
1340                                                 in6p->in6p_hops = optval;
1341
1342                                                 if ((in6p->in6p_vflag &
1343                                                      INP_IPV4) != 0)
1344                                                         in6p->inp_ip_ttl = optval;
1345                                         }
1346                                         break;
1347 #define OPTSET(bit) \
1348 do { \
1349         if (optval) \
1350                 in6p->in6p_flags |= (bit); \
1351         else \
1352                 in6p->in6p_flags &= ~(bit); \
1353 } while (0)
1354 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1355
1356                                 case IPV6_CHECKSUM:
1357                                         in6p->in6p_cksum = optval;
1358                                         break;
1359
1360                                 case IPV6_FAITH:
1361                                         OPTSET(IN6P_FAITH);
1362                                         break;
1363
1364                                 case IPV6_V6ONLY:
1365                                         /*
1366                                          * make setsockopt(IPV6_V6ONLY)
1367                                          * available only prior to bind(2).
1368                                          * see ipng mailing list, Jun 22 2001.
1369                                          */
1370                                         if (in6p->in6p_lport ||
1371                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1372                                         {
1373                                                 error = EINVAL;
1374                                                 break;
1375                                         }
1376                                         OPTSET(IN6P_IPV6_V6ONLY);
1377                                         if (optval)
1378                                                 in6p->in6p_vflag &= ~INP_IPV4;
1379                                         else
1380                                                 in6p->in6p_vflag |= INP_IPV4;
1381                                         break;
1382                                 }
1383                                 break;
1384
1385                         case IPV6_PKTINFO:
1386                         case IPV6_HOPLIMIT:
1387                         case IPV6_HOPOPTS:
1388                         case IPV6_DSTOPTS:
1389                         case IPV6_RTHDR:
1390                                 /* RFC 2292 */
1391                                 if (optlen != sizeof(int)) {
1392                                         error = EINVAL;
1393                                         break;
1394                                 }
1395                                 error = soopt_to_kbuf(sopt, &optval,
1396                                         sizeof optval, sizeof optval);
1397                                 if (error)
1398                                         break;
1399                                 switch (optname) {
1400                                 case IPV6_PKTINFO:
1401                                         OPTSET(IN6P_PKTINFO);
1402                                         break;
1403                                 case IPV6_HOPLIMIT:
1404                                         OPTSET(IN6P_HOPLIMIT);
1405                                         break;
1406                                 case IPV6_HOPOPTS:
1407                                         /*
1408                                          * Check super-user privilege.
1409                                          * See comments for IPV6_RECVHOPOPTS.
1410                                          */
1411                                         if (!privileged)
1412                                                 return (EPERM);
1413                                         OPTSET(IN6P_HOPOPTS);
1414                                         break;
1415                                 case IPV6_DSTOPTS:
1416                                         if (!privileged)
1417                                                 return (EPERM);
1418                                         OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1419                                         break;
1420                                 case IPV6_RTHDR:
1421                                         OPTSET(IN6P_RTHDR);
1422                                         break;
1423                                 }
1424                                 break;
1425 #undef OPTSET
1426
1427                         case IPV6_MULTICAST_IF:
1428                         case IPV6_MULTICAST_HOPS:
1429                         case IPV6_MULTICAST_LOOP:
1430                         case IPV6_JOIN_GROUP:
1431                         case IPV6_LEAVE_GROUP:
1432                             {
1433                                 struct mbuf *m;
1434                                 if (sopt->sopt_valsize > MLEN) {
1435                                         error = EMSGSIZE;
1436                                         break;
1437                                 }
1438                                 /* XXX */
1439                                 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1440                                 if (m == NULL) {
1441                                         error = ENOBUFS;
1442                                         break;
1443                                 }
1444                                 m->m_len = sopt->sopt_valsize;
1445                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1446                                                     m->m_len, m->m_len);
1447                                 error = ip6_setmoptions(sopt->sopt_name,
1448                                                         &in6p->in6p_moptions,
1449                                                         m);
1450                                 m_free(m);
1451                             }
1452                                 break;
1453
1454                         case IPV6_PORTRANGE:
1455                                 error = soopt_to_kbuf(sopt, &optval,
1456                                     sizeof optval, sizeof optval);
1457                                 if (error)
1458                                         break;
1459
1460                                 switch (optval) {
1461                                 case IPV6_PORTRANGE_DEFAULT:
1462                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1463                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1464                                         break;
1465
1466                                 case IPV6_PORTRANGE_HIGH:
1467                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1468                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1469                                         break;
1470
1471                                 case IPV6_PORTRANGE_LOW:
1472                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1473                                         in6p->in6p_flags |= IN6P_LOWPORT;
1474                                         break;
1475
1476                                 default:
1477                                         error = EINVAL;
1478                                         break;
1479                                 }
1480                                 break;
1481
1482 #if defined(IPSEC) || defined(FAST_IPSEC)
1483                         case IPV6_IPSEC_POLICY:
1484                             {
1485                                 caddr_t req = NULL;
1486                                 size_t len = 0;
1487                                 struct mbuf *m;
1488
1489                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1490                                         break;
1491                                 soopt_to_mbuf(sopt, m));        /* XXX */
1492                                 if (m) {
1493                                         req = mtod(m, caddr_t);
1494                                         len = m->m_len;
1495                                 }
1496                                 error = ipsec6_set_policy(in6p, optname, req,
1497                                                           len, privileged);
1498                                 m_freem(m);
1499                             }
1500                                 break;
1501 #endif /* KAME IPSEC */
1502
1503                         case IPV6_FW_ADD:
1504                         case IPV6_FW_DEL:
1505                         case IPV6_FW_FLUSH:
1506                         case IPV6_FW_ZERO:
1507                             {
1508                                 struct mbuf *m;
1509                                 struct mbuf **mp = &m;
1510
1511                                 if (ip6_fw_ctl_ptr == NULL)
1512                                         return EINVAL;
1513                                 /* XXX */
1514                                 if ((error = soopt_getm(sopt, &m)) != 0)
1515                                         break;
1516                                 /* XXX */
1517                                 soopt_to_mbuf(sopt, m);
1518                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1519                                 m = *mp;
1520                             }
1521                                 break;
1522
1523                         default:
1524                                 error = ENOPROTOOPT;
1525                                 break;
1526                         }
1527                         break;
1528
1529                 case SOPT_GET:
1530                         switch (optname) {
1531
1532                         case IPV6_PKTOPTIONS:
1533                                 if (in6p->in6p_options) {
1534                                         struct mbuf *m;
1535                                         m = m_copym(in6p->in6p_options,
1536                                             0, M_COPYALL, MB_WAIT);
1537                                         error = soopt_from_mbuf(sopt, m);
1538                                         if (error == 0)
1539                                                 m_freem(m);
1540                                 } else
1541                                         sopt->sopt_valsize = 0;
1542                                 break;
1543
1544                         case IPV6_UNICAST_HOPS:
1545                         case IPV6_CHECKSUM:
1546
1547                         case IPV6_FAITH:
1548                         case IPV6_V6ONLY:
1549                         case IPV6_PORTRANGE:
1550                                 switch (optname) {
1551
1552                                 case IPV6_UNICAST_HOPS:
1553                                         optval = in6p->in6p_hops;
1554                                         break;
1555
1556                                 case IPV6_CHECKSUM:
1557                                         optval = in6p->in6p_cksum;
1558                                         break;
1559
1560                                 case IPV6_FAITH:
1561                                         optval = OPTBIT(IN6P_FAITH);
1562                                         break;
1563
1564                                 case IPV6_V6ONLY:
1565                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1566                                         break;
1567
1568                                 case IPV6_PORTRANGE:
1569                                     {
1570                                         int flags;
1571                                         flags = in6p->in6p_flags;
1572                                         if (flags & IN6P_HIGHPORT)
1573                                                 optval = IPV6_PORTRANGE_HIGH;
1574                                         else if (flags & IN6P_LOWPORT)
1575                                                 optval = IPV6_PORTRANGE_LOW;
1576                                         else
1577                                                 optval = 0;
1578                                         break;
1579                                     }
1580                                 }
1581                                 soopt_from_kbuf(sopt, &optval,
1582                                         sizeof optval);
1583                                 break;
1584
1585                         case IPV6_PKTINFO:
1586                         case IPV6_HOPLIMIT:
1587                         case IPV6_HOPOPTS:
1588                         case IPV6_RTHDR:
1589                         case IPV6_DSTOPTS:
1590                                 if (optname == IPV6_HOPOPTS ||
1591                                     optname == IPV6_DSTOPTS ||
1592                                     !privileged)
1593                                         return (EPERM);
1594                                 switch (optname) {
1595                                 case IPV6_PKTINFO:
1596                                         optval = OPTBIT(IN6P_PKTINFO);
1597                                         break;
1598                                 case IPV6_HOPLIMIT:
1599                                         optval = OPTBIT(IN6P_HOPLIMIT);
1600                                         break;
1601                                 case IPV6_HOPOPTS:
1602                                         if (!privileged)
1603                                                 return (EPERM);
1604                                         optval = OPTBIT(IN6P_HOPOPTS);
1605                                         break;
1606                                 case IPV6_RTHDR:
1607                                         optval = OPTBIT(IN6P_RTHDR);
1608                                         break;
1609                                 case IPV6_DSTOPTS:
1610                                         if (!privileged)
1611                                                 return (EPERM);
1612                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1613                                         break;
1614                                 }
1615                                 soopt_from_kbuf(sopt, &optval,
1616                                         sizeof optval);
1617                                 break;
1618
1619                         case IPV6_MULTICAST_IF:
1620                         case IPV6_MULTICAST_HOPS:
1621                         case IPV6_MULTICAST_LOOP:
1622                         case IPV6_JOIN_GROUP:
1623                         case IPV6_LEAVE_GROUP:
1624                             {
1625                                 struct mbuf *m;
1626                                 error = ip6_getmoptions(sopt->sopt_name,
1627                                                 in6p->in6p_moptions, &m);
1628                                 if (error == 0)
1629                                         soopt_from_kbuf(sopt,
1630                                                 mtod(m, char *), m->m_len);
1631                                 m_freem(m);
1632                             }
1633                                 break;
1634
1635 #if defined(IPSEC) || defined(FAST_IPSEC)
1636                         case IPV6_IPSEC_POLICY:
1637                           {
1638                                 caddr_t req = NULL;
1639                                 size_t len = 0;
1640                                 struct mbuf *m = NULL;
1641                                 struct mbuf **mp = &m;
1642
1643                                 error = soopt_getm(sopt, &m); /* XXX */
1644                                 if (error != 0)
1645                                         break;
1646                                 soopt_to_mbuf(sopt, m); /* XXX */
1647                                 if (m) {
1648                                         req = mtod(m, caddr_t);
1649                                         len = m->m_len;
1650                                 }
1651                                 error = ipsec6_get_policy(in6p, req, len, mp);
1652                                 if (error == 0)
1653                                         error = soopt_from_mbuf(sopt, m); /*XXX*/
1654                                 if (error == 0 && m != NULL)
1655                                         m_freem(m);
1656                                 break;
1657                           }
1658 #endif /* KAME IPSEC */
1659
1660                         case IPV6_FW_GET:
1661                           {
1662                                 struct mbuf *m;
1663                                 struct mbuf **mp = &m;
1664
1665                                 if (ip6_fw_ctl_ptr == NULL)
1666                                 {
1667                                         return EINVAL;
1668                                 }
1669                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1670                                 if (error == 0)
1671                                         error = soopt_from_mbuf(sopt, m); /* XXX */
1672                                 if (error == 0 && m != NULL)
1673                                         m_freem(m);
1674                           }
1675                                 break;
1676
1677                         default:
1678                                 error = ENOPROTOOPT;
1679                                 break;
1680                         }
1681                         break;
1682                 }
1683         } else {
1684                 error = EINVAL;
1685         }
1686         return (error);
1687 }
1688
1689 /*
1690  * Set up IP6 options in pcb for insertion in output packets or
1691  * specifying behavior of outgoing packets.
1692  */
1693 static int
1694 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
1695             struct sockopt *sopt)
1696 {
1697         struct ip6_pktopts *opt = *pktopt;
1698         int error = 0;
1699         struct thread *td = sopt->sopt_td;
1700         int priv = 0;
1701
1702         /* turn off any old options. */
1703         if (opt) {
1704 #ifdef DIAGNOSTIC
1705                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1706                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1707                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1708                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
1709 #endif
1710                 ip6_clearpktopts(opt, 1, -1);
1711         } else
1712                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1713         *pktopt = NULL;
1714
1715         if (m == NULL || m->m_len == 0) {
1716                 /*
1717                  * Only turning off any previous options, regardless of
1718                  * whether the opt is just created or given.
1719                  */
1720                 kfree(opt, M_IP6OPT);
1721                 return (0);
1722         }
1723
1724         /*  set options specified by user. */
1725         if (suser(td) == 0)
1726                 priv = 1;
1727         if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1728                 ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1729                 kfree(opt, M_IP6OPT);
1730                 return (error);
1731         }
1732         *pktopt = opt;
1733         return (0);
1734 }
1735
1736 /*
1737  * initialize ip6_pktopts.  beware that there are non-zero default values in
1738  * the struct.
1739  */
1740 void
1741 init_ip6pktopts(struct ip6_pktopts *opt)
1742 {
1743
1744         bzero(opt, sizeof(*opt));
1745         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
1746 }
1747
1748 void
1749 ip6_clearpktopts(struct ip6_pktopts *pktopt, int needfree, int optname)
1750 {
1751         if (pktopt == NULL)
1752                 return;
1753
1754         if (optname == -1) {
1755                 if (needfree && pktopt->ip6po_pktinfo)
1756                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
1757                 pktopt->ip6po_pktinfo = NULL;
1758         }
1759         if (optname == -1)
1760                 pktopt->ip6po_hlim = -1;
1761         if (optname == -1) {
1762                 if (needfree && pktopt->ip6po_nexthop)
1763                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
1764                 pktopt->ip6po_nexthop = NULL;
1765         }
1766         if (optname == -1) {
1767                 if (needfree && pktopt->ip6po_hbh)
1768                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
1769                 pktopt->ip6po_hbh = NULL;
1770         }
1771         if (optname == -1) {
1772                 if (needfree && pktopt->ip6po_dest1)
1773                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
1774                 pktopt->ip6po_dest1 = NULL;
1775         }
1776         if (optname == -1) {
1777                 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1778                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1779                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1780                 if (pktopt->ip6po_route.ro_rt) {
1781                         RTFREE(pktopt->ip6po_route.ro_rt);
1782                         pktopt->ip6po_route.ro_rt = NULL;
1783                 }
1784         }
1785         if (optname == -1) {
1786                 if (needfree && pktopt->ip6po_dest2)
1787                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
1788                 pktopt->ip6po_dest2 = NULL;
1789         }
1790 }
1791
1792 #define PKTOPT_EXTHDRCPY(type) \
1793 do {\
1794         if (src->type) {\
1795                 int hlen =\
1796                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1797                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
1798                 if (dst->type == NULL)\
1799                         goto bad;\
1800                 bcopy(src->type, dst->type, hlen);\
1801         }\
1802 } while (0)
1803
1804 struct ip6_pktopts *
1805 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
1806 {
1807         struct ip6_pktopts *dst;
1808
1809         if (src == NULL) {
1810                 kprintf("ip6_clearpktopts: invalid argument\n");
1811                 return (NULL);
1812         }
1813
1814         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
1815         if (dst == NULL)
1816                 return (NULL);
1817
1818         dst->ip6po_hlim = src->ip6po_hlim;
1819         if (src->ip6po_pktinfo) {
1820                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
1821                                             M_IP6OPT, canwait);
1822                 if (dst->ip6po_pktinfo == NULL)
1823                         goto bad;
1824                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1825         }
1826         if (src->ip6po_nexthop) {
1827                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
1828                                             M_IP6OPT, canwait);
1829                 if (dst->ip6po_nexthop == NULL)
1830                         goto bad;
1831                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1832                       src->ip6po_nexthop->sa_len);
1833         }
1834         PKTOPT_EXTHDRCPY(ip6po_hbh);
1835         PKTOPT_EXTHDRCPY(ip6po_dest1);
1836         PKTOPT_EXTHDRCPY(ip6po_dest2);
1837         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1838         return (dst);
1839
1840 bad:
1841         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
1842         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
1843         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
1844         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
1845         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
1846         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
1847         kfree(dst, M_IP6OPT);
1848         return (NULL);
1849 }
1850 #undef PKTOPT_EXTHDRCPY
1851
1852 void
1853 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1854 {
1855         if (pktopt == NULL)
1856                 return;
1857
1858         ip6_clearpktopts(pktopt, 1, -1);
1859
1860         kfree(pktopt, M_IP6OPT);
1861 }
1862
1863 /*
1864  * Set the IP6 multicast options in response to user setsockopt().
1865  */
1866 static int
1867 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1868 {
1869         int error = 0;
1870         u_int loop, ifindex;
1871         struct ipv6_mreq *mreq;
1872         struct ifnet *ifp;
1873         struct ip6_moptions *im6o = *im6op;
1874         struct route_in6 ro;
1875         struct sockaddr_in6 *dst;
1876         struct in6_multi_mship *imm;
1877         struct thread *td = curthread;  /* XXX */
1878
1879         if (im6o == NULL) {
1880                 /*
1881                  * No multicast option buffer attached to the pcb;
1882                  * allocate one and initialize to default values.
1883                  */
1884                 im6o = (struct ip6_moptions *)
1885                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1886
1887                 *im6op = im6o;
1888                 im6o->im6o_multicast_ifp = NULL;
1889                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1890                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1891                 LIST_INIT(&im6o->im6o_memberships);
1892         }
1893
1894         switch (optname) {
1895
1896         case IPV6_MULTICAST_IF:
1897                 /*
1898                  * Select the interface for outgoing multicast packets.
1899                  */
1900                 if (m == NULL || m->m_len != sizeof(u_int)) {
1901                         error = EINVAL;
1902                         break;
1903                 }
1904                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1905                 if (ifindex < 0 || if_index < ifindex) {
1906                         error = ENXIO;  /* XXX EINVAL? */
1907                         break;
1908                 }
1909                 ifp = ifindex2ifnet[ifindex];
1910                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1911                         error = EADDRNOTAVAIL;
1912                         break;
1913                 }
1914                 im6o->im6o_multicast_ifp = ifp;
1915                 break;
1916
1917         case IPV6_MULTICAST_HOPS:
1918             {
1919                 /*
1920                  * Set the IP6 hoplimit for outgoing multicast packets.
1921                  */
1922                 int optval;
1923                 if (m == NULL || m->m_len != sizeof(int)) {
1924                         error = EINVAL;
1925                         break;
1926                 }
1927                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1928                 if (optval < -1 || optval >= 256)
1929                         error = EINVAL;
1930                 else if (optval == -1)
1931                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1932                 else
1933                         im6o->im6o_multicast_hlim = optval;
1934                 break;
1935             }
1936
1937         case IPV6_MULTICAST_LOOP:
1938                 /*
1939                  * Set the loopback flag for outgoing multicast packets.
1940                  * Must be zero or one.
1941                  */
1942                 if (m == NULL || m->m_len != sizeof(u_int)) {
1943                         error = EINVAL;
1944                         break;
1945                 }
1946                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1947                 if (loop > 1) {
1948                         error = EINVAL;
1949                         break;
1950                 }
1951                 im6o->im6o_multicast_loop = loop;
1952                 break;
1953
1954         case IPV6_JOIN_GROUP:
1955                 /*
1956                  * Add a multicast group membership.
1957                  * Group must be a valid IP6 multicast address.
1958                  */
1959                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1960                         error = EINVAL;
1961                         break;
1962                 }
1963                 mreq = mtod(m, struct ipv6_mreq *);
1964                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1965                         /*
1966                          * We use the unspecified address to specify to accept
1967                          * all multicast addresses. Only super user is allowed
1968                          * to do this.
1969                          */
1970                         if (suser(td))
1971                         {
1972                                 error = EACCES;
1973                                 break;
1974                         }
1975                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1976                         error = EINVAL;
1977                         break;
1978                 }
1979
1980                 /*
1981                  * If the interface is specified, validate it.
1982                  */
1983                 if (mreq->ipv6mr_interface < 0
1984                  || if_index < mreq->ipv6mr_interface) {
1985                         error = ENXIO;  /* XXX EINVAL? */
1986                         break;
1987                 }
1988                 /*
1989                  * If no interface was explicitly specified, choose an
1990                  * appropriate one according to the given multicast address.
1991                  */
1992                 if (mreq->ipv6mr_interface == 0) {
1993                         /*
1994                          * If the multicast address is in node-local scope,
1995                          * the interface should be a loopback interface.
1996                          * Otherwise, look up the routing table for the
1997                          * address, and choose the outgoing interface.
1998                          *   XXX: is it a good approach?
1999                          */
2000                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2001                                 ifp = &loif[0];
2002                         } else {
2003                                 ro.ro_rt = NULL;
2004                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2005                                 bzero(dst, sizeof(*dst));
2006                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2007                                 dst->sin6_family = AF_INET6;
2008                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2009                                 rtalloc((struct route *)&ro);
2010                                 if (ro.ro_rt == NULL) {
2011                                         error = EADDRNOTAVAIL;
2012                                         break;
2013                                 }
2014                                 ifp = ro.ro_rt->rt_ifp;
2015                                 rtfree(ro.ro_rt);
2016                         }
2017                 } else
2018                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2019
2020                 /*
2021                  * See if we found an interface, and confirm that it
2022                  * supports multicast
2023                  */
2024                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2025                         error = EADDRNOTAVAIL;
2026                         break;
2027                 }
2028                 /*
2029                  * Put interface index into the multicast address,
2030                  * if the address has link-local scope.
2031                  */
2032                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2033                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2034                                 = htons(mreq->ipv6mr_interface);
2035                 }
2036                 /*
2037                  * See if the membership already exists.
2038                  */
2039                 for (imm = im6o->im6o_memberships.lh_first;
2040                      imm != NULL; imm = imm->i6mm_chain.le_next)
2041                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2042                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2043                                                &mreq->ipv6mr_multiaddr))
2044                                 break;
2045                 if (imm != NULL) {
2046                         error = EADDRINUSE;
2047                         break;
2048                 }
2049                 /*
2050                  * Everything looks good; add a new record to the multicast
2051                  * address list for the given interface.
2052                  */
2053                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2054                 if ((imm->i6mm_maddr =
2055                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2056                         kfree(imm, M_IPMADDR);
2057                         break;
2058                 }
2059                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2060                 break;
2061
2062         case IPV6_LEAVE_GROUP:
2063                 /*
2064                  * Drop a multicast group membership.
2065                  * Group must be a valid IP6 multicast address.
2066                  */
2067                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2068                         error = EINVAL;
2069                         break;
2070                 }
2071                 mreq = mtod(m, struct ipv6_mreq *);
2072                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2073                         if (suser(td)) {
2074                                 error = EACCES;
2075                                 break;
2076                         }
2077                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2078                         error = EINVAL;
2079                         break;
2080                 }
2081                 /*
2082                  * If an interface address was specified, get a pointer
2083                  * to its ifnet structure.
2084                  */
2085                 if (mreq->ipv6mr_interface < 0
2086                  || if_index < mreq->ipv6mr_interface) {
2087                         error = ENXIO;  /* XXX EINVAL? */
2088                         break;
2089                 }
2090                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2091                 /*
2092                  * Put interface index into the multicast address,
2093                  * if the address has link-local scope.
2094                  */
2095                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2096                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2097                                 = htons(mreq->ipv6mr_interface);
2098                 }
2099                 /*
2100                  * Find the membership in the membership list.
2101                  */
2102                 for (imm = im6o->im6o_memberships.lh_first;
2103                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2104                         if ((ifp == NULL ||
2105                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2106                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2107                                                &mreq->ipv6mr_multiaddr))
2108                                 break;
2109                 }
2110                 if (imm == NULL) {
2111                         /* Unable to resolve interface */
2112                         error = EADDRNOTAVAIL;
2113                         break;
2114                 }
2115                 /*
2116                  * Give up the multicast address record to which the
2117                  * membership points.
2118                  */
2119                 LIST_REMOVE(imm, i6mm_chain);
2120                 in6_delmulti(imm->i6mm_maddr);
2121                 kfree(imm, M_IPMADDR);
2122                 break;
2123
2124         default:
2125                 error = EOPNOTSUPP;
2126                 break;
2127         }
2128
2129         /*
2130          * If all options have default values, no need to keep the mbuf.
2131          */
2132         if (im6o->im6o_multicast_ifp == NULL &&
2133             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2134             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2135             im6o->im6o_memberships.lh_first == NULL) {
2136                 kfree(*im6op, M_IPMOPTS);
2137                 *im6op = NULL;
2138         }
2139
2140         return (error);
2141 }
2142
2143 /*
2144  * Return the IP6 multicast options in response to user getsockopt().
2145  */
2146 static int
2147 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2148 {
2149         u_int *hlim, *loop, *ifindex;
2150
2151         *mp = m_get(MB_WAIT, MT_HEADER);                /* XXX */
2152
2153         switch (optname) {
2154
2155         case IPV6_MULTICAST_IF:
2156                 ifindex = mtod(*mp, u_int *);
2157                 (*mp)->m_len = sizeof(u_int);
2158                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2159                         *ifindex = 0;
2160                 else
2161                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2162                 return (0);
2163
2164         case IPV6_MULTICAST_HOPS:
2165                 hlim = mtod(*mp, u_int *);
2166                 (*mp)->m_len = sizeof(u_int);
2167                 if (im6o == NULL)
2168                         *hlim = ip6_defmcasthlim;
2169                 else
2170                         *hlim = im6o->im6o_multicast_hlim;
2171                 return (0);
2172
2173         case IPV6_MULTICAST_LOOP:
2174                 loop = mtod(*mp, u_int *);
2175                 (*mp)->m_len = sizeof(u_int);
2176                 if (im6o == NULL)
2177                         *loop = ip6_defmcasthlim;
2178                 else
2179                         *loop = im6o->im6o_multicast_loop;
2180                 return (0);
2181
2182         default:
2183                 return (EOPNOTSUPP);
2184         }
2185 }
2186
2187 /*
2188  * Discard the IP6 multicast options.
2189  */
2190 void
2191 ip6_freemoptions(struct ip6_moptions *im6o)
2192 {
2193         struct in6_multi_mship *imm;
2194
2195         if (im6o == NULL)
2196                 return;
2197
2198         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2199                 LIST_REMOVE(imm, i6mm_chain);
2200                 if (imm->i6mm_maddr)
2201                         in6_delmulti(imm->i6mm_maddr);
2202                 kfree(imm, M_IPMADDR);
2203         }
2204         kfree(im6o, M_IPMOPTS);
2205 }
2206
2207 /*
2208  * Set IPv6 outgoing packet options based on advanced API.
2209  */
2210 int
2211 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, int priv,
2212                   int needcopy)
2213 {
2214         struct cmsghdr *cm = NULL;
2215
2216         if (control == NULL || opt == NULL)
2217                 return (EINVAL);
2218
2219         init_ip6pktopts(opt);
2220
2221         /*
2222          * XXX: Currently, we assume all the optional information is stored
2223          * in a single mbuf.
2224          */
2225         if (control->m_next)
2226                 return (EINVAL);
2227
2228         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2229                      control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2230                 cm = mtod(control, struct cmsghdr *);
2231                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2232                         return (EINVAL);
2233                 if (cm->cmsg_level != IPPROTO_IPV6)
2234                         continue;
2235
2236                 /*
2237                  * XXX should check if RFC2292 API is mixed with 2292bis API
2238                  */
2239                 switch (cm->cmsg_type) {
2240                 case IPV6_PKTINFO:
2241                         if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2242                                 return (EINVAL);
2243                         if (needcopy) {
2244                                 /* XXX: Is it really WAITOK? */
2245                                 opt->ip6po_pktinfo =
2246                                         kmalloc(sizeof(struct in6_pktinfo),
2247                                                M_IP6OPT, M_WAITOK);
2248                                 bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2249                                     sizeof(struct in6_pktinfo));
2250                         } else
2251                                 opt->ip6po_pktinfo =
2252                                         (struct in6_pktinfo *)CMSG_DATA(cm);
2253                         if (opt->ip6po_pktinfo->ipi6_ifindex &&
2254                             IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2255                                 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2256                                         htons(opt->ip6po_pktinfo->ipi6_ifindex);
2257
2258                         if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2259                          || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2260                                 return (ENXIO);
2261                         }
2262
2263                         /*
2264                          * Check if the requested source address is indeed a
2265                          * unicast address assigned to the node, and can be
2266                          * used as the packet's source address.
2267                          */
2268                         if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2269                                 struct in6_ifaddr *ia6;
2270                                 struct sockaddr_in6 sin6;
2271
2272                                 bzero(&sin6, sizeof(sin6));
2273                                 sin6.sin6_len = sizeof(sin6);
2274                                 sin6.sin6_family = AF_INET6;
2275                                 sin6.sin6_addr =
2276                                         opt->ip6po_pktinfo->ipi6_addr;
2277                                 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2278                                 if (ia6 == NULL ||
2279                                     (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2280                                                        IN6_IFF_NOTREADY)) != 0)
2281                                         return (EADDRNOTAVAIL);
2282                         }
2283                         break;
2284
2285                 case IPV6_HOPLIMIT:
2286                         if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2287                                 return (EINVAL);
2288
2289                         opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2290                         if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2291                                 return (EINVAL);
2292                         break;
2293
2294                 case IPV6_NEXTHOP:
2295                         if (!priv)
2296                                 return (EPERM);
2297
2298                         if (cm->cmsg_len < sizeof(u_char) ||
2299                             /* check if cmsg_len is large enough for sa_len */
2300                             cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2301                                 return (EINVAL);
2302
2303                         if (needcopy) {
2304                                 opt->ip6po_nexthop =
2305                                         kmalloc(*CMSG_DATA(cm),
2306                                                M_IP6OPT, M_WAITOK);
2307                                 bcopy(CMSG_DATA(cm),
2308                                       opt->ip6po_nexthop,
2309                                       *CMSG_DATA(cm));
2310                         } else
2311                                 opt->ip6po_nexthop =
2312                                         (struct sockaddr *)CMSG_DATA(cm);
2313                         break;
2314
2315                 case IPV6_HOPOPTS:
2316                 {
2317                         struct ip6_hbh *hbh;
2318                         int hbhlen;
2319
2320                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2321                                 return (EINVAL);
2322                         hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2323                         hbhlen = (hbh->ip6h_len + 1) << 3;
2324                         if (cm->cmsg_len != CMSG_LEN(hbhlen))
2325                                 return (EINVAL);
2326
2327                         if (needcopy) {
2328                                 opt->ip6po_hbh =
2329                                         kmalloc(hbhlen, M_IP6OPT, M_WAITOK);
2330                                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2331                         } else
2332                                 opt->ip6po_hbh = hbh;
2333                         break;
2334                 }
2335
2336                 case IPV6_DSTOPTS:
2337                 {
2338                         struct ip6_dest *dest, **newdest;
2339                         int destlen;
2340
2341                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2342                                 return (EINVAL);
2343                         dest = (struct ip6_dest *)CMSG_DATA(cm);
2344                         destlen = (dest->ip6d_len + 1) << 3;
2345                         if (cm->cmsg_len != CMSG_LEN(destlen))
2346                                 return (EINVAL);
2347
2348                         /*
2349                          * The old advacned API is ambiguous on this
2350                          * point. Our approach is to determine the
2351                          * position based according to the existence
2352                          * of a routing header. Note, however, that
2353                          * this depends on the order of the extension
2354                          * headers in the ancillary data; the 1st part
2355                          * of the destination options header must
2356                          * appear before the routing header in the
2357                          * ancillary data, too.
2358                          * RFC2292bis solved the ambiguity by
2359                          * introducing separate cmsg types.
2360                          */
2361                         if (opt->ip6po_rthdr == NULL)
2362                                 newdest = &opt->ip6po_dest1;
2363                         else
2364                                 newdest = &opt->ip6po_dest2;
2365
2366                         if (needcopy) {
2367                                 *newdest = kmalloc(destlen, M_IP6OPT, M_WAITOK);
2368                                 bcopy(dest, *newdest, destlen);
2369                         } else
2370                                 *newdest = dest;
2371
2372                         break;
2373                 }
2374
2375                 case IPV6_RTHDR:
2376                 {
2377                         struct ip6_rthdr *rth;
2378                         int rthlen;
2379
2380                         if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2381                                 return (EINVAL);
2382                         rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2383                         rthlen = (rth->ip6r_len + 1) << 3;
2384                         if (cm->cmsg_len != CMSG_LEN(rthlen))
2385                                 return (EINVAL);
2386
2387                         switch (rth->ip6r_type) {
2388                         case IPV6_RTHDR_TYPE_0:
2389                                 /* must contain one addr */
2390                                 if (rth->ip6r_len == 0)
2391                                         return (EINVAL);
2392                                 /* length must be even */
2393                                 if (rth->ip6r_len % 2)
2394                                         return (EINVAL);
2395                                 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2396                                         return (EINVAL);
2397                                 break;
2398                         default:
2399                                 return (EINVAL);        /* not supported */
2400                         }
2401
2402                         if (needcopy) {
2403                                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT,
2404                                                           M_WAITOK);
2405                                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2406                         } else
2407                                 opt->ip6po_rthdr = rth;
2408
2409                         break;
2410                 }
2411
2412                 default:
2413                         return (ENOPROTOOPT);
2414                 }
2415         }
2416
2417         return (0);
2418 }
2419
2420 /*
2421  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2422  * packet to the input queue of a specified interface.  Note that this
2423  * calls the output routine of the loopback "driver", but with an interface
2424  * pointer that might NOT be &loif -- easier than replicating that code here.
2425  */
2426 void
2427 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2428 {
2429         struct mbuf *copym;
2430         struct ip6_hdr *ip6;
2431
2432         copym = m_copy(m, 0, M_COPYALL);
2433         if (copym == NULL)
2434                 return;
2435
2436         /*
2437          * Make sure to deep-copy IPv6 header portion in case the data
2438          * is in an mbuf cluster, so that we can safely override the IPv6
2439          * header portion later.
2440          */
2441         if ((copym->m_flags & M_EXT) != 0 ||
2442             copym->m_len < sizeof(struct ip6_hdr)) {
2443                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2444                 if (copym == NULL)
2445                         return;
2446         }
2447
2448 #ifdef DIAGNOSTIC
2449         if (copym->m_len < sizeof(*ip6)) {
2450                 m_freem(copym);
2451                 return;
2452         }
2453 #endif
2454
2455         ip6 = mtod(copym, struct ip6_hdr *);
2456         /*
2457          * clear embedded scope identifiers if necessary.
2458          * in6_clearscope will touch the addresses only when necessary.
2459          */
2460         in6_clearscope(&ip6->ip6_src);
2461         in6_clearscope(&ip6->ip6_dst);
2462
2463         if_simloop(ifp, copym, dst->sin6_family, 0);
2464 }
2465
2466 /*
2467  * Separate the IPv6 header from the payload into its own mbuf.
2468  *
2469  * Returns the new mbuf chain or the original mbuf if no payload.
2470  * Returns NULL if can't allocate new mbuf for header.
2471  */
2472 static struct mbuf *
2473 ip6_splithdr(struct mbuf *m)
2474 {
2475         struct mbuf *mh;
2476
2477         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
2478                 return (m);
2479
2480         MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
2481         if (mh == NULL)
2482                 return (NULL);
2483         mh->m_len = sizeof(struct ip6_hdr);
2484         M_MOVE_PKTHDR(mh, m);
2485         MH_ALIGN(mh, sizeof(struct ip6_hdr));
2486         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
2487         m->m_data += sizeof(struct ip6_hdr);
2488         m->m_len -= sizeof(struct ip6_hdr);
2489         mh->m_next = m;
2490         return (mh);
2491 }
2492
2493 /*
2494  * Compute IPv6 extension header length.
2495  */
2496 int
2497 ip6_optlen(struct in6pcb *in6p)
2498 {
2499         int len;
2500
2501         if (!in6p->in6p_outputopts)
2502                 return 0;
2503
2504         len = 0;
2505 #define elen(x) \
2506     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2507
2508         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2509         if (in6p->in6p_outputopts->ip6po_rthdr)
2510                 /* dest1 is valid with rthdr only */
2511                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2512         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2513         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2514         return len;
2515 #undef elen
2516 }