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 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   3
   4 /*
   5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the project nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  */
  32
  33 /*
  34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  35  *      The Regents of the University of California.  All rights reserved.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. Neither the name of the University nor the names of its contributors
  46  *    may be used to endorse or promote products derived from this software
  47  *    without specific prior written permission.
  48  *
  49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  59  * SUCH DAMAGE.
  60  *
  61  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  62  */
  63
  64 #include "opt_ip6fw.h"
  65 #include "opt_inet.h"
  66 #include "opt_inet6.h"
  67 #include "opt_ipsec.h"
  68
  69 #include <sys/param.h>
  70 #include <sys/malloc.h>
  71 #include <sys/mbuf.h>
  72 #include <sys/errno.h>
  73 #include <sys/protosw.h>
  74 #include <sys/socket.h>
  75 #include <sys/socketvar.h>
  76 #include <sys/systm.h>
  77 #include <sys/kernel.h>
  78 #include <sys/proc.h>
  79 #include <sys/priv.h>
  80
  81 #include <sys/thread2.h>
  82 #include <sys/msgport2.h>
  83
  84 #include <net/if.h>
  85 #include <net/route.h>
  86 #include <net/pfil.h>
  87
  88 #include <netinet/in.h>
  89 #include <netinet/in_var.h>
  90 #include <netinet6/in6_var.h>
  91 #include <netinet/ip6.h>
  92 #include <netinet/icmp6.h>
  93 #include <netinet6/ip6_var.h>
  94 #include <netinet/in_pcb.h>
  95 #include <netinet6/nd6.h>
  96 #include <netinet6/ip6protosw.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_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
 127 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
 128          int, int, int, int);
 129 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
 130                             struct socket *, struct sockopt *);
 131 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 132 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 133 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 134 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 135         struct ifnet *, struct in6_addr *, u_long *, int *);
 136 static int copyexthdr (void *, struct mbuf **);
 137 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
 138                                   struct ip6_frag **);
 139 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 140 static struct mbuf *ip6_splithdr (struct mbuf *);
 141 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 142
 143 /*
 144  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 145  * header (with pri, len, nxt, hlim, src, dst).
 146  * This function may modify ver and hlim only.
 147  * The mbuf chain containing the packet will be freed.
 148  * The mbuf opt, if present, will not be freed.
 149  *
 150  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 151  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 152  * which is rt_rmx.rmx_mtu.
 153  */
 154 int
 155 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 156            int flags, struct ip6_moptions *im6o,
 157            struct ifnet **ifpp,         /* XXX: just for statistics */
 158            struct inpcb *inp)
 159 {
 160         struct ip6_hdr *ip6, *mhip6;
 161         struct ifnet *ifp, *origifp;
 162         struct mbuf *m = m0;
 163         struct mbuf *mprev;
 164         u_char *nexthdrp;
 165         int hlen, tlen, len, off;
 166         struct route_in6 ip6route;
 167         struct sockaddr_in6 *dst;
 168         int error = 0;
 169         struct in6_ifaddr *ia = NULL;
 170         u_long mtu;
 171         int alwaysfrag, dontfrag;
 172         u_int32_t optlen, plen = 0, unfragpartlen;
 173         struct ip6_exthdrs exthdrs;
 174         struct in6_addr finaldst;
 175         struct route_in6 *ro_pmtu = NULL;
 176         boolean_t hdrsplit = FALSE;
 177         boolean_t needipsec = FALSE;
 178 #ifdef IPSEC
 179         boolean_t needipsectun = FALSE;
 180         struct secpolicy *sp = NULL;
 181         struct socket *so = inp ? inp->inp_socket : NULL;
 182
 183         ip6 = mtod(m, struct ip6_hdr *);
 184 #endif
 185 #ifdef FAST_IPSEC
 186         boolean_t needipsectun = FALSE;
 187         struct secpolicy *sp = NULL;
 188
 189         ip6 = mtod(m, struct ip6_hdr *);
 190 #endif
 191
 192         bzero(&exthdrs, sizeof exthdrs);
 193
 194         if (opt) {
 195                 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
 196                         goto freehdrs;
 197                 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
 198                         goto freehdrs;
 199                 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
 200                         goto freehdrs;
 201                 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
 202                         goto freehdrs;
 203         }
 204
 205 #ifdef IPSEC
 206         /* get a security policy for this packet */
 207         if (so == NULL)
 208                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 209         else
 210                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 211
 212         if (sp == NULL) {
 213                 ipsec6stat.out_inval++;
 214                 goto freehdrs;
 215         }
 216
 217         error = 0;
 218
 219         /* check policy */
 220         switch (sp->policy) {
 221         case IPSEC_POLICY_DISCARD:
 222                 /*
 223                  * This packet is just discarded.
 224                  */
 225                 ipsec6stat.out_polvio++;
 226                 goto freehdrs;
 227
 228         case IPSEC_POLICY_BYPASS:
 229         case IPSEC_POLICY_NONE:
 230                 /* no need to do IPsec. */
 231                 needipsec = FALSE;
 232                 break;
 233
 234         case IPSEC_POLICY_IPSEC:
 235                 if (sp->req == NULL) {
 236                         error = key_spdacquire(sp);     /* acquire a policy */
 237                         goto freehdrs;
 238                 }
 239                 needipsec = TRUE;
 240                 break;
 241
 242         case IPSEC_POLICY_ENTRUST:
 243         default:
 244                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 245         }
 246 #endif /* IPSEC */
 247 #ifdef FAST_IPSEC
 248         /* get a security policy for this packet */
 249         if (inp == NULL)
 250                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 251         else
 252                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 253
 254         if (sp == NULL) {
 255                 newipsecstat.ips_out_inval++;
 256                 goto freehdrs;
 257         }
 258
 259         error = 0;
 260
 261         /* check policy */
 262         switch (sp->policy) {
 263         case IPSEC_POLICY_DISCARD:
 264                 /*
 265                  * This packet is just discarded.
 266                  */
 267                 newipsecstat.ips_out_polvio++;
 268                 goto freehdrs;
 269
 270         case IPSEC_POLICY_BYPASS:
 271         case IPSEC_POLICY_NONE:
 272                 /* no need to do IPsec. */
 273                 needipsec = FALSE;
 274                 break;
 275
 276         case IPSEC_POLICY_IPSEC:
 277                 if (sp->req == NULL) {
 278                         error = key_spdacquire(sp);     /* acquire a policy */
 279                         goto freehdrs;
 280                 }
 281                 needipsec = TRUE;
 282                 break;
 283
 284         case IPSEC_POLICY_ENTRUST:
 285         default:
 286                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 287         }
 288 #endif /* FAST_IPSEC */
 289
 290         /*
 291          * Calculate the total length of the extension header chain.
 292          * Keep the length of the unfragmentable part for fragmentation.
 293          */
 294         optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
 295             m_lengthm(exthdrs.ip6e_dest1, NULL) +
 296             m_lengthm(exthdrs.ip6e_rthdr, NULL);
 297
 298         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 299
 300         /* NOTE: we don't add AH/ESP length here. do that later. */
 301         optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
 302
 303         /*
 304          * If we need IPsec, or there is at least one extension header,
 305          * separate IP6 header from the payload.
 306          */
 307         if ((needipsec || optlen) && !hdrsplit) {
 308                 exthdrs.ip6e_ip6 = ip6_splithdr(m);
 309                 if (exthdrs.ip6e_ip6 == NULL) {
 310                         error = ENOBUFS;
 311                         goto freehdrs;
 312                 }
 313                 m = exthdrs.ip6e_ip6;
 314                 hdrsplit = TRUE;
 315         }
 316
 317         /* adjust pointer */
 318         ip6 = mtod(m, struct ip6_hdr *);
 319
 320         /* adjust mbuf packet header length */
 321         m->m_pkthdr.len += optlen;
 322         plen = m->m_pkthdr.len - sizeof(*ip6);
 323
 324         /* If this is a jumbo payload, insert a jumbo payload option. */
 325         if (plen > IPV6_MAXPACKET) {
 326                 if (!hdrsplit) {
 327                         exthdrs.ip6e_ip6 = ip6_splithdr(m);
 328                         if (exthdrs.ip6e_ip6 == NULL) {
 329                                 error = ENOBUFS;
 330                                 goto freehdrs;
 331                         }
 332                         m = exthdrs.ip6e_ip6;
 333                         hdrsplit = TRUE;
 334                 }
 335                 /* adjust pointer */
 336                 ip6 = mtod(m, struct ip6_hdr *);
 337                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 338                         goto freehdrs;
 339                 ip6->ip6_plen = 0;
 340         } else
 341                 ip6->ip6_plen = htons(plen);
 342
 343         /*
 344          * Concatenate headers and fill in next header fields.
 345          * Here we have, on "m"
 346          *      IPv6 payload
 347          * and we insert headers accordingly.  Finally, we should be getting:
 348          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 349          *
 350          * during the header composing process, "m" points to IPv6 header.
 351          * "mprev" points to an extension header prior to esp.
 352          */
 353
 354         nexthdrp = &ip6->ip6_nxt;
 355         mprev = m;
 356
 357         /*
 358          * we treat dest2 specially.  this makes IPsec processing
 359          * much easier.  the goal here is to make mprev point the
 360          * mbuf prior to dest2.
 361          *
 362          * result: IPv6 dest2 payload
 363          * m and mprev will point to IPv6 header.
 364          */
 365         if (exthdrs.ip6e_dest2) {
 366                 if (!hdrsplit)
 367                         panic("assumption failed: hdr not split");
 368                 exthdrs.ip6e_dest2->m_next = m->m_next;
 369                 m->m_next = exthdrs.ip6e_dest2;
 370                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 371                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 372         }
 373
 374 /*
 375  * Place m1 after mprev.
 376  */
 377 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
 378     do {\
 379         if (m1) {\
 380                 if (!hdrsplit)\
 381                         panic("assumption failed: hdr not split");\
 382                 *mtod(m1, u_char *) = *nexthdrp;\
 383                 *nexthdrp = (i);\
 384                 nexthdrp = mtod(m1, u_char *);\
 385                 m1->m_next = mprev->m_next;\
 386                 mprev->m_next = m1;\
 387                 mprev = m1;\
 388         }\
 389     } while (0)
 390
 391         /*
 392          * result: IPv6 hbh dest1 rthdr dest2 payload
 393          * m will point to IPv6 header.  mprev will point to the
 394          * extension header prior to dest2 (rthdr in the above case).
 395          */
 396         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 397         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
 398         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
 399
 400 #if defined(IPSEC) || defined(FAST_IPSEC)
 401         if (needipsec) {
 402                 struct ipsec_output_state state;
 403                 int segleft_org = 0;
 404                 struct ip6_rthdr *rh = NULL;
 405
 406                 /*
 407                  * pointers after IPsec headers are not valid any more.
 408                  * other pointers need a great care too.
 409                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 410                  */
 411                 exthdrs.ip6e_dest2 = NULL;
 412
 413                 if (exthdrs.ip6e_rthdr) {
 414                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 415                         segleft_org = rh->ip6r_segleft;
 416                         rh->ip6r_segleft = 0;
 417                 }
 418
 419                 bzero(&state, sizeof state);
 420                 state.m = m;
 421                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 422                                             &needipsectun);
 423                 m = state.m;
 424                 if (error) {
 425                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 426                         m = NULL;
 427                         switch (error) {
 428                         case EHOSTUNREACH:
 429                         case ENETUNREACH:
 430                         case EMSGSIZE:
 431                         case ENOBUFS:
 432                         case ENOMEM:
 433                                 break;
 434                         default:
 435                                 kprintf("ip6_output (ipsec): error code %d\n",
 436                                        error);
 437                                 /* fall through */
 438                         case ENOENT:
 439                                 /* don't show these error codes to the user */
 440                                 error = 0;
 441                                 break;
 442                         }
 443                         goto bad;
 444                 }
 445                 if (exthdrs.ip6e_rthdr) {
 446                         /* ah6_output doesn't modify mbuf chain */
 447                         rh->ip6r_segleft = segleft_org;
 448                 }
 449         }
 450 #endif
 451
 452         /*
 453          * If there is a routing header, replace destination address field
 454          * with the first hop of the routing header.
 455          */
 456         if (exthdrs.ip6e_rthdr) {
 457                 struct ip6_rthdr *rh;
 458
 459                 finaldst = ip6->ip6_dst;
 460                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 461                 switch (rh->ip6r_type) {
 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         /* Determine path MTU. */
 749         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 750             &alwaysfrag)) != 0)
 751                 goto bad;
 752
 753         /*
 754          * The caller of this function may specify to use the minimum MTU
 755          * in some cases.
 756          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 757          * setting.  The logic is a bit complicated; by default, unicast
 758          * packets will follow path MTU while multicast packets will be sent at
 759          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 760          * including unicast ones will be sent at the minimum MTU.  Multicast
 761          * packets will always be sent at the minimum MTU unless
 762          * IP6PO_MINMTU_DISABLE is explicitly specified.
 763          * See RFC 3542 for more details.
 764          */
 765         if (mtu > IPV6_MMTU) {
 766                 if ((flags & IPV6_MINMTU))
 767                         mtu = IPV6_MMTU;
 768                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 769                         mtu = IPV6_MMTU;
 770                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 771                          (opt == NULL ||
 772                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 773                         mtu = IPV6_MMTU;
 774                 }
 775         }
 776
 777         /* Fake scoped addresses */
 778         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 779                 /*
 780                  * If source or destination address is a scoped address, and
 781                  * the packet is going to be sent to a loopback interface,
 782                  * we should keep the original interface.
 783                  */
 784
 785                 /*
 786                  * XXX: this is a very experimental and temporary solution.
 787                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 788                  * field of the structure here.
 789                  * We rely on the consistency between two scope zone ids
 790                  * of source and destination, which should already be assured.
 791                  * Larger scopes than link will be supported in the future.
 792                  */
 793                 origifp = NULL;
 794                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 795                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 796                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 797                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 798                 /*
 799                  * XXX: origifp can be NULL even in those two cases above.
 800                  * For example, if we remove the (only) link-local address
 801                  * from the loopback interface, and try to send a link-local
 802                  * address without link-id information.  Then the source
 803                  * address is ::1, and the destination address is the
 804                  * link-local address with its s6_addr16[1] being zero.
 805                  * What is worse, if the packet goes to the loopback interface
 806                  * by a default rejected route, the null pointer would be
 807                  * passed to looutput, and the kernel would hang.
 808                  * The following last resort would prevent such disaster.
 809                  */
 810                 if (origifp == NULL)
 811                         origifp = ifp;
 812         }
 813         else
 814                 origifp = ifp;
 815         /*
 816          * clear embedded scope identifiers if necessary.
 817          * in6_clearscope will touch the addresses only when necessary.
 818          */
 819         in6_clearscope(&ip6->ip6_src);
 820         in6_clearscope(&ip6->ip6_dst);
 821
 822         /*
 823          * Check with the firewall...
 824          */
 825         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 826                 u_short port = 0;
 827
 828                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 829                 /* If ipfw says divert, we have to just drop packet */
 830                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 831                         m_freem(m);
 832                         goto done;
 833                 }
 834                 if (!m) {
 835                         error = EACCES;
 836                         goto done;
 837                 }
 838         }
 839
 840         /*
 841          * If the outgoing packet contains a hop-by-hop options header,
 842          * it must be examined and processed even by the source node.
 843          * (RFC 2460, section 4.)
 844          */
 845         if (exthdrs.ip6e_hbh) {
 846                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 847                 u_int32_t dummy1; /* XXX unused */
 848                 u_int32_t dummy2; /* XXX unused */
 849
 850 #ifdef DIAGNOSTIC
 851                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 852                         panic("ip6e_hbh is not continuous");
 853 #endif
 854                 /*
 855                  *  XXX: if we have to send an ICMPv6 error to the sender,
 856                  *       we need the M_LOOP flag since icmp6_error() expects
 857                  *       the IPv6 and the hop-by-hop options header are
 858                  *       continuous unless the flag is set.
 859                  */
 860                 m->m_flags |= M_LOOP;
 861                 m->m_pkthdr.rcvif = ifp;
 862                 if (ip6_process_hopopts(m,
 863                                         (u_int8_t *)(hbh + 1),
 864                                         ((hbh->ip6h_len + 1) << 3) -
 865                                         sizeof(struct ip6_hbh),
 866                                         &dummy1, &dummy2) < 0) {
 867                         /* m was already freed at this point */
 868                         error = EINVAL;/* better error? */
 869                         goto done;
 870                 }
 871                 m->m_flags &= ~M_LOOP; /* XXX */
 872                 m->m_pkthdr.rcvif = NULL;
 873         }
 874
 875         /*
 876          * Run through list of hooks for output packets.
 877          */
 878         if (pfil_has_hooks(&inet6_pfil_hook)) {
 879                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 880                 if (error != 0 || m == NULL)
 881                         goto done;
 882                 ip6 = mtod(m, struct ip6_hdr *);
 883         }
 884
 885         /*
 886          * Send the packet to the outgoing interface.
 887          * If necessary, do IPv6 fragmentation before sending.
 888          *
 889          * the logic here is rather complex:
 890          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 891          * 1-a: send as is if tlen <= path mtu
 892          * 1-b: fragment if tlen > path mtu
 893          *
 894          * 2: if user asks us not to fragment (dontfrag == 1)
 895          * 2-a: send as is if tlen <= interface mtu
 896          * 2-b: error if tlen > interface mtu
 897          *
 898          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 899          *      always fragment
 900          *
 901          * 4: if dontfrag == 1 && alwaysfrag == 1
 902          *      error, as we cannot handle this conflicting request
 903          */
 904         tlen = m->m_pkthdr.len;
 905
 906         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 907                 dontfrag = 1;
 908         else
 909                 dontfrag = 0;
 910         if (dontfrag && alwaysfrag) {   /* case 4 */
 911                 /* conflicting request - can't transmit */
 912                 error = EMSGSIZE;
 913                 goto bad;
 914         }
 915         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 916                 /*
 917                  * Even if the DONTFRAG option is specified, we cannot send the
 918                  * packet when the data length is larger than the MTU of the
 919                  * outgoing interface.
 920                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 921                  * well as returning an error code (the latter is not described
 922                  * in the API spec.)
 923                  */
 924                 u_int32_t mtu32;
 925                 struct ip6ctlparam ip6cp;
 926
 927                 mtu32 = (u_int32_t)mtu;
 928                 bzero(&ip6cp, sizeof(ip6cp));
 929                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 930                 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 931                     (void *)&ip6cp);
 932
 933                 error = EMSGSIZE;
 934                 goto bad;
 935         }
 936
 937         /*
 938          * transmit packet without fragmentation
 939          */
 940         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 941                 struct in6_ifaddr *ia6;
 942
 943                 ip6 = mtod(m, struct ip6_hdr *);
 944                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 945                 if (ia6) {
 946                         /* Record statistics for this interface address. */
 947                         IFA_STAT_INC(&ia6->ia_ifa, opackets, 1);
 948                         IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len);
 949                 }
 950 #ifdef IPSEC
 951                 /* clean ipsec history once it goes out of the node */
 952                 ipsec_delaux(m);
 953 #endif
 954                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 955                 goto done;
 956         }
 957
 958         /*
 959          * try to fragment the packet.  case 1-b and 3
 960          */
 961         if (mtu < IPV6_MMTU) {
 962                 /*
 963                  * note that path MTU is never less than IPV6_MMTU
 964                  * (see icmp6_input).
 965                  */
 966                 error = EMSGSIZE;
 967                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 968                 goto bad;
 969         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 970                 error = EMSGSIZE;
 971                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 972                 goto bad;
 973         } else {
 974                 struct mbuf **mnext, *m_frgpart;
 975                 struct ip6_frag *ip6f;
 976                 u_int32_t id = htonl(ip6_id++);
 977                 u_char nextproto;
 978
 979                 /*
 980                  * Too large for the destination or interface;
 981                  * fragment if possible.
 982                  * Must be able to put at least 8 bytes per fragment.
 983                  */
 984                 hlen = unfragpartlen;
 985                 if (mtu > IPV6_MAXPACKET)
 986                         mtu = IPV6_MAXPACKET;
 987
 988                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 989                 if (len < 8) {
 990                         error = EMSGSIZE;
 991                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 992                         goto bad;
 993                 }
 994
 995                 mnext = &m->m_nextpkt;
 996
 997                 /*
 998                  * Change the next header field of the last header in the
 999                  * unfragmentable part.
1000                  */
1001                 if (exthdrs.ip6e_rthdr) {
1002                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1003                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1004                 } else if (exthdrs.ip6e_dest1) {
1005                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1006                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1007                 } else if (exthdrs.ip6e_hbh) {
1008                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1009                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1010                 } else {
1011                         nextproto = ip6->ip6_nxt;
1012                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1013                 }
1014
1015                 /*
1016                  * Loop through length of segment after first fragment,
1017                  * make new header and copy data of each part and link onto
1018                  * chain.
1019                  */
1020                 m0 = m;
1021                 for (off = hlen; off < tlen; off += len) {
1022                         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1023                         if (!m) {
1024                                 error = ENOBUFS;
1025                                 ip6stat.ip6s_odropped++;
1026                                 goto sendorfree;
1027                         }
1028                         m->m_pkthdr.rcvif = NULL;
1029                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1030                         *mnext = m;
1031                         mnext = &m->m_nextpkt;
1032                         m->m_data += max_linkhdr;
1033                         mhip6 = mtod(m, struct ip6_hdr *);
1034                         *mhip6 = *ip6;
1035                         m->m_len = sizeof(*mhip6);
1036                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1037                         if (error) {
1038                                 ip6stat.ip6s_odropped++;
1039                                 goto sendorfree;
1040                         }
1041                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1042                         if (off + len >= tlen)
1043                                 len = tlen - off;
1044                         else
1045                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1046                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1047                                                           sizeof(*ip6f) -
1048                                                           sizeof(struct ip6_hdr)));
1049                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1050                                 error = ENOBUFS;
1051                                 ip6stat.ip6s_odropped++;
1052                                 goto sendorfree;
1053                         }
1054                         m_cat(m, m_frgpart);
1055                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1056                         m->m_pkthdr.rcvif = NULL;
1057                         ip6f->ip6f_reserved = 0;
1058                         ip6f->ip6f_ident = id;
1059                         ip6f->ip6f_nxt = nextproto;
1060                         ip6stat.ip6s_ofragments++;
1061                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1062                 }
1063
1064                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1065         }
1066
1067         /*
1068          * Remove leading garbages.
1069          */
1070 sendorfree:
1071         m = m0->m_nextpkt;
1072         m0->m_nextpkt = NULL;
1073         m_freem(m0);
1074         for (m0 = m; m; m = m0) {
1075                 m0 = m->m_nextpkt;
1076                 m->m_nextpkt = NULL;
1077                 if (error == 0) {
1078                         /* Record statistics for this interface address. */
1079                         if (ia) {
1080                                 IFA_STAT_INC(&ia->ia_ifa, opackets, 1);
1081                                 IFA_STAT_INC(&ia->ia_ifa, obytes,
1082                                     m->m_pkthdr.len);
1083                         }
1084 #ifdef IPSEC
1085                         /* clean ipsec history once it goes out of the node */
1086                         ipsec_delaux(m);
1087 #endif
1088                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1089                 } else
1090                         m_freem(m);
1091         }
1092
1093         if (error == 0)
1094                 ip6stat.ip6s_fragmented++;
1095
1096 done:
1097         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1098                 RTFREE(ro->ro_rt);
1099         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1100                 RTFREE(ro_pmtu->ro_rt);
1101         }
1102
1103 #ifdef IPSEC
1104         if (sp != NULL)
1105                 key_freesp(sp);
1106 #endif
1107 #ifdef FAST_IPSEC
1108         if (sp != NULL)
1109                 KEY_FREESP(&sp);
1110 #endif
1111
1112         return (error);
1113
1114 freehdrs:
1115         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1116         m_freem(exthdrs.ip6e_dest1);
1117         m_freem(exthdrs.ip6e_rthdr);
1118         m_freem(exthdrs.ip6e_dest2);
1119         /* fall through */
1120 bad:
1121         m_freem(m);
1122         goto done;
1123 }
1124
1125 static int
1126 copyexthdr(void *h, struct mbuf **mp)
1127 {
1128         struct ip6_ext *hdr = h;
1129         int hlen;
1130         struct mbuf *m;
1131
1132         if (hdr == NULL)
1133                 return 0;
1134
1135         hlen = (hdr->ip6e_len + 1) * 8;
1136         if (hlen > MCLBYTES)
1137                 return ENOBUFS; /* XXX */
1138
1139         m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1140         if (!m)
1141                 return ENOBUFS;
1142         m->m_len = hlen;
1143
1144         bcopy(hdr, mtod(m, caddr_t), hlen);
1145
1146         *mp = m;
1147         return 0;
1148 }
1149
1150 /*
1151  * Insert jumbo payload option.
1152  */
1153 static int
1154 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1155 {
1156         struct mbuf *mopt;
1157         u_char *optbuf;
1158         u_int32_t v;
1159
1160 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1161
1162         /*
1163          * If there is no hop-by-hop options header, allocate new one.
1164          * If there is one but it doesn't have enough space to store the
1165          * jumbo payload option, allocate a cluster to store the whole options.
1166          * Otherwise, use it to store the options.
1167          */
1168         if (exthdrs->ip6e_hbh == NULL) {
1169                 MGET(mopt, MB_DONTWAIT, MT_DATA);
1170                 if (mopt == NULL)
1171                         return (ENOBUFS);
1172                 mopt->m_len = JUMBOOPTLEN;
1173                 optbuf = mtod(mopt, u_char *);
1174                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1175                 exthdrs->ip6e_hbh = mopt;
1176         } else {
1177                 struct ip6_hbh *hbh;
1178
1179                 mopt = exthdrs->ip6e_hbh;
1180                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1181                         /*
1182                          * XXX assumption:
1183                          * - exthdrs->ip6e_hbh is not referenced from places
1184                          *   other than exthdrs.
1185                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1186                          */
1187                         int oldoptlen = mopt->m_len;
1188                         struct mbuf *n;
1189
1190                         /*
1191                          * XXX: give up if the whole (new) hbh header does
1192                          * not fit even in an mbuf cluster.
1193                          */
1194                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1195                                 return (ENOBUFS);
1196
1197                         /*
1198                          * As a consequence, we must always prepare a cluster
1199                          * at this point.
1200                          */
1201                         n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1202                         if (!n)
1203                                 return (ENOBUFS);
1204                         n->m_len = oldoptlen + JUMBOOPTLEN;
1205                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1206                         optbuf = mtod(n, caddr_t) + oldoptlen;
1207                         m_freem(mopt);
1208                         mopt = exthdrs->ip6e_hbh = n;
1209                 } else {
1210                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1211                         mopt->m_len += JUMBOOPTLEN;
1212                 }
1213                 optbuf[0] = IP6OPT_PADN;
1214                 optbuf[1] = 1;
1215
1216                 /*
1217                  * Adjust the header length according to the pad and
1218                  * the jumbo payload option.
1219                  */
1220                 hbh = mtod(mopt, struct ip6_hbh *);
1221                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1222         }
1223
1224         /* fill in the option. */
1225         optbuf[2] = IP6OPT_JUMBO;
1226         optbuf[3] = 4;
1227         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1228         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1229
1230         /* finally, adjust the packet header length */
1231         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1232
1233         return (0);
1234 #undef JUMBOOPTLEN
1235 }
1236
1237 /*
1238  * Insert fragment header and copy unfragmentable header portions.
1239  */
1240 static int
1241 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1242                   struct ip6_frag **frghdrp)
1243 {
1244         struct mbuf *n, *mlast;
1245
1246         if (hlen > sizeof(struct ip6_hdr)) {
1247                 n = m_copym(m0, sizeof(struct ip6_hdr),
1248                             hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1249                 if (n == NULL)
1250                         return (ENOBUFS);
1251                 m->m_next = n;
1252         } else
1253                 n = m;
1254
1255         /* Search for the last mbuf of unfragmentable part. */
1256         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1257                 ;
1258
1259         if (!(mlast->m_flags & M_EXT) &&
1260             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1261                 /* use the trailing space of the last mbuf for the fragment hdr */
1262                 *frghdrp =
1263                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1264                 mlast->m_len += sizeof(struct ip6_frag);
1265                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1266         } else {
1267                 /* allocate a new mbuf for the fragment header */
1268                 struct mbuf *mfrg;
1269
1270                 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1271                 if (mfrg == NULL)
1272                         return (ENOBUFS);
1273                 mfrg->m_len = sizeof(struct ip6_frag);
1274                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1275                 mlast->m_next = mfrg;
1276         }
1277
1278         return (0);
1279 }
1280
1281 static int
1282 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1283     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1284     int *alwaysfragp)
1285 {
1286         u_int32_t mtu = 0;
1287         int alwaysfrag = 0;
1288         int error = 0;
1289
1290         if (ro_pmtu != ro) {
1291                 /* The first hop and the final destination may differ. */
1292                 struct sockaddr_in6 *sa6_dst =
1293                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1294                 if (ro_pmtu->ro_rt &&
1295                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1296                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1297                         RTFREE(ro_pmtu->ro_rt);
1298                         ro_pmtu->ro_rt = NULL;
1299                 }
1300                 if (ro_pmtu->ro_rt == NULL) {
1301                         bzero(sa6_dst, sizeof(*sa6_dst));
1302                         sa6_dst->sin6_family = AF_INET6;
1303                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1304                         sa6_dst->sin6_addr = *dst;
1305
1306                         rtalloc((struct route *)ro_pmtu);
1307                 }
1308         }
1309         if (ro_pmtu->ro_rt) {
1310                 u_int32_t ifmtu;
1311                 struct in_conninfo inc;
1312
1313                 bzero(&inc, sizeof(inc));
1314                 inc.inc_flags = 1; /* IPv6 */
1315                 inc.inc6_faddr = *dst;
1316
1317                 if (ifp == NULL)
1318                         ifp = ro_pmtu->ro_rt->rt_ifp;
1319                 ifmtu = IN6_LINKMTU(ifp);
1320                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1321                 if (mtu == 0)
1322                         mtu = ifmtu;
1323                 else if (mtu < IPV6_MMTU) {
1324                         /*
1325                          * RFC2460 section 5, last paragraph:
1326                          * if we record ICMPv6 too big message with
1327                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1328                          * or smaller, with framgent header attached.
1329                          * (fragment header is needed regardless from the
1330                          * packet size, for translators to identify packets)
1331                          */
1332                         alwaysfrag = 1;
1333                         mtu = IPV6_MMTU;
1334                 } else if (mtu > ifmtu) {
1335                         /*
1336                          * The MTU on the route is larger than the MTU on
1337                          * the interface!  This shouldn't happen, unless the
1338                          * MTU of the interface has been changed after the
1339                          * interface was brought up.  Change the MTU in the
1340                          * route to match the interface MTU (as long as the
1341                          * field isn't locked).
1342                          */
1343                         mtu = ifmtu;
1344                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1345                 }
1346         } else if (ifp) {
1347                 mtu = IN6_LINKMTU(ifp);
1348         } else
1349                 error = EHOSTUNREACH; /* XXX */
1350
1351         *mtup = mtu;
1352         if (alwaysfragp)
1353                 *alwaysfragp = alwaysfrag;
1354         return (error);
1355 }
1356
1357 /*
1358  * IP6 socket option processing.
1359  */
1360 void
1361 ip6_ctloutput_dispatch(netmsg_t msg)
1362 {
1363         int error;
1364
1365         error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1366                               msg->ctloutput.nm_sopt);
1367         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1368 }
1369
1370 int
1371 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1372 {
1373         int optdatalen,uproto;
1374         int privileged;
1375         struct inpcb *in6p = so->so_pcb;
1376         void *optdata;
1377         int error, optval;
1378         int level, op, optname;
1379         int optlen;
1380         struct thread *td;
1381
1382         if (sopt) {
1383                 level = sopt->sopt_level;
1384                 op = sopt->sopt_dir;
1385                 optname = sopt->sopt_name;
1386                 optlen = sopt->sopt_valsize;
1387                 td = sopt->sopt_td;
1388         } else {
1389                 panic("ip6_ctloutput: arg soopt is NULL");
1390                 /* NOT REACHED */
1391                 td = NULL;
1392         }
1393         error = optval = 0;
1394
1395         uproto = (int)so->so_proto->pr_protocol;
1396         privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1397
1398         if (level == IPPROTO_IPV6) {
1399                 switch (op) {
1400
1401                 case SOPT_SET:
1402                         switch (optname) {
1403                         case IPV6_2292PKTOPTIONS:
1404 #ifdef IPV6_PKTOPTIONS
1405                         case IPV6_PKTOPTIONS:
1406 #endif
1407                         {
1408                                 struct mbuf *m;
1409
1410                                 error = soopt_getm(sopt, &m); /* XXX */
1411                                 if (error != 0)
1412                                         break;
1413                                 soopt_to_mbuf(sopt, m); /* XXX */
1414                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1415                                                     m, so, sopt);
1416                                 m_freem(m); /* XXX */
1417                                 break;
1418                         }
1419
1420                         /*
1421                          * Use of some Hop-by-Hop options or some
1422                          * Destination options, might require special
1423                          * privilege.  That is, normal applications
1424                          * (without special privilege) might be forbidden
1425                          * from setting certain options in outgoing packets,
1426                          * and might never see certain options in received
1427                          * packets. [RFC 2292 Section 6]
1428                          * KAME specific note:
1429                          *  KAME prevents non-privileged users from sending or
1430                          *  receiving ANY hbh/dst options in order to avoid
1431                          *  overhead of parsing options in the kernel.
1432                          */
1433                         case IPV6_RECVHOPOPTS:
1434                         case IPV6_RECVDSTOPTS:
1435                         case IPV6_RECVRTHDRDSTOPTS:
1436                                 if (!privileged)
1437                                         return (EPERM);
1438                         case IPV6_RECVPKTINFO:
1439                         case IPV6_RECVHOPLIMIT:
1440                         case IPV6_RECVRTHDR:
1441                         case IPV6_RECVPATHMTU:
1442                         case IPV6_RECVTCLASS:
1443                         case IPV6_AUTOFLOWLABEL:
1444                         case IPV6_HOPLIMIT:
1445                         /* FALLTHROUGH */
1446                         case IPV6_UNICAST_HOPS:
1447                         case IPV6_FAITH:
1448
1449                         case IPV6_V6ONLY:
1450                                 if (optlen != sizeof(int)) {
1451                                         error = EINVAL;
1452                                         break;
1453                                 }
1454                                 error = soopt_to_kbuf(sopt, &optval,
1455                                         sizeof optval, sizeof optval);
1456                                 if (error)
1457                                         break;
1458                                 switch (optname) {
1459
1460                                 case IPV6_UNICAST_HOPS:
1461                                         if (optval < -1 || optval >= 256)
1462                                                 error = EINVAL;
1463                                         else {
1464                                                 /* -1 = kernel default */
1465                                                 in6p->in6p_hops = optval;
1466
1467                                                 if ((in6p->in6p_vflag &
1468                                                      INP_IPV4) != 0)
1469                                                         in6p->inp_ip_ttl = optval;
1470                                         }
1471                                         break;
1472 #define OPTSET(bit) \
1473 do { \
1474         if (optval) \
1475                 in6p->in6p_flags |= (bit); \
1476         else \
1477                 in6p->in6p_flags &= ~(bit); \
1478 } while (0)
1479 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1480 /*
1481  * Although changed to RFC3542, It's better to also support RFC2292 API
1482  */
1483 #define OPTSET2292(bit) \
1484 do { \
1485         in6p->in6p_flags |= IN6P_RFC2292; \
1486         if (optval) \
1487                 in6p->in6p_flags |= (bit); \
1488         else \
1489                 in6p->in6p_flags &= ~(bit); \
1490 } while (/*CONSTCOND*/ 0)
1491
1492                                 case IPV6_RECVPKTINFO:
1493                                         /* cannot mix with RFC2292 */
1494                                         if (OPTBIT(IN6P_RFC2292)) {
1495                                                 error = EINVAL;
1496                                                 break;
1497                                         }
1498                                         OPTSET(IN6P_PKTINFO);
1499                                         break;
1500
1501                                 case IPV6_HOPLIMIT:
1502                                 {
1503                                         struct ip6_pktopts **optp;
1504
1505                                         /* cannot mix with RFC2292 */
1506                                         if (OPTBIT(IN6P_RFC2292)) {
1507                                                 error = EINVAL;
1508                                                 break;
1509                                         }
1510                                         optp = &in6p->in6p_outputopts;
1511                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1512                                             (u_char *)&optval, sizeof(optval),
1513                                             optp, uproto);
1514                                         break;
1515                                 }
1516
1517                                 case IPV6_RECVHOPLIMIT:
1518                                         /* cannot mix with RFC2292 */
1519                                         if (OPTBIT(IN6P_RFC2292)) {
1520                                                 error = EINVAL;
1521                                                 break;
1522                                         }
1523                                         OPTSET(IN6P_HOPLIMIT);
1524                                         break;
1525
1526                                 case IPV6_RECVHOPOPTS:
1527                                         /* cannot mix with RFC2292 */
1528                                         if (OPTBIT(IN6P_RFC2292)) {
1529                                                 error = EINVAL;
1530                                                 break;
1531                                         }
1532                                         OPTSET(IN6P_HOPOPTS);
1533                                         break;
1534
1535                                 case IPV6_RECVDSTOPTS:
1536                                         /* cannot mix with RFC2292 */
1537                                         if (OPTBIT(IN6P_RFC2292)) {
1538                                                 error = EINVAL;
1539                                                 break;
1540                                         }
1541                                         OPTSET(IN6P_DSTOPTS);
1542                                         break;
1543
1544                                 case IPV6_RECVRTHDRDSTOPTS:
1545                                         /* cannot mix with RFC2292 */
1546                                         if (OPTBIT(IN6P_RFC2292)) {
1547                                                 error = EINVAL;
1548                                                 break;
1549                                         }
1550                                         OPTSET(IN6P_RTHDRDSTOPTS);
1551                                         break;
1552
1553                                 case IPV6_RECVRTHDR:
1554                                         /* cannot mix with RFC2292 */
1555                                         if (OPTBIT(IN6P_RFC2292)) {
1556                                                 error = EINVAL;
1557                                                 break;
1558                                         }
1559                                         OPTSET(IN6P_RTHDR);
1560                                         break;
1561
1562                                 case IPV6_RECVPATHMTU:
1563                                         /*
1564                                          * We ignore this option for TCP
1565                                          * sockets.
1566                                          * (RFC3542 leaves this case
1567                                          * unspecified.)
1568                                          */
1569                                         if (uproto != IPPROTO_TCP)
1570                                                 OPTSET(IN6P_MTU);
1571                                         break;
1572
1573                                 case IPV6_RECVTCLASS:
1574                                         /* cannot mix with RFC2292 XXX */
1575                                         if (OPTBIT(IN6P_RFC2292)) {
1576                                                 error = EINVAL;
1577                                                 break;
1578                                         }
1579                                         OPTSET(IN6P_TCLASS);
1580                                         break;
1581
1582                                 case IPV6_AUTOFLOWLABEL:
1583                                         OPTSET(IN6P_AUTOFLOWLABEL);
1584                                         break;
1585
1586                                 case IPV6_FAITH:
1587                                         OPTSET(IN6P_FAITH);
1588                                         break;
1589
1590                                 case IPV6_V6ONLY:
1591                                         /*
1592                                          * make setsockopt(IPV6_V6ONLY)
1593                                          * available only prior to bind(2).
1594                                          */
1595                                         if (in6p->in6p_lport ||
1596                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1597                                         {
1598                                                 error = EINVAL;
1599                                                 break;
1600                                         }
1601                                         OPTSET(IN6P_IPV6_V6ONLY);
1602                                         if (optval)
1603                                                 in6p->in6p_vflag &= ~INP_IPV4;
1604                                         else
1605                                                 in6p->in6p_vflag |= INP_IPV4;
1606                                         break;
1607                                 }
1608                                 break;
1609
1610                         case IPV6_TCLASS:
1611                         case IPV6_DONTFRAG:
1612                         case IPV6_USE_MIN_MTU:
1613                         case IPV6_PREFER_TEMPADDR:
1614                                 if (optlen != sizeof(optval)) {
1615                                         error = EINVAL;
1616                                         break;
1617                                 }
1618                                 error = soopt_to_kbuf(sopt, &optval,
1619                                         sizeof optval, sizeof optval);
1620                                 if (error)
1621                                         break;
1622                                 {
1623                                         struct ip6_pktopts **optp;
1624                                         optp = &in6p->in6p_outputopts;
1625                                         error = ip6_pcbopt(optname,
1626                                             (u_char *)&optval, sizeof(optval),
1627                                             optp, uproto);
1628                                         break;
1629                                 }
1630
1631                         case IPV6_2292PKTINFO:
1632                         case IPV6_2292HOPLIMIT:
1633                         case IPV6_2292HOPOPTS:
1634                         case IPV6_2292DSTOPTS:
1635                         case IPV6_2292RTHDR:
1636                                 /* RFC 2292 */
1637                                 if (optlen != sizeof(int)) {
1638                                         error = EINVAL;
1639                                         break;
1640                                 }
1641                                 error = soopt_to_kbuf(sopt, &optval,
1642                                         sizeof optval, sizeof optval);
1643                                 if (error)
1644                                         break;
1645                                 switch (optname) {
1646                                 case IPV6_2292PKTINFO:
1647                                         OPTSET2292(IN6P_PKTINFO);
1648                                         break;
1649                                 case IPV6_2292HOPLIMIT:
1650                                         OPTSET2292(IN6P_HOPLIMIT);
1651                                         break;
1652                                 case IPV6_2292HOPOPTS:
1653                                         /*
1654                                          * Check super-user privilege.
1655                                          * See comments for IPV6_RECVHOPOPTS.
1656                                          */
1657                                         if (!privileged)
1658                                                 return (EPERM);
1659                                         OPTSET2292(IN6P_HOPOPTS);
1660                                         break;
1661                                 case IPV6_2292DSTOPTS:
1662                                         if (!privileged)
1663                                                 return (EPERM);
1664                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1665                                         break;
1666                                 case IPV6_2292RTHDR:
1667                                         OPTSET2292(IN6P_RTHDR);
1668                                         break;
1669                                 }
1670                                 break;
1671
1672                         case IPV6_PKTINFO:
1673                         case IPV6_HOPOPTS:
1674                         case IPV6_RTHDR:
1675                         case IPV6_DSTOPTS:
1676                         case IPV6_RTHDRDSTOPTS:
1677                         case IPV6_NEXTHOP:
1678                         {
1679                                 /*
1680                                  * New advanced API (RFC3542)
1681                                  */
1682                                 u_char *optbuf;
1683                                 u_char optbuf_storage[MCLBYTES];
1684                                 int optlen;
1685                                 struct ip6_pktopts **optp;
1686
1687                                 /* cannot mix with RFC2292 */
1688                                 if (OPTBIT(IN6P_RFC2292)) {
1689                                         error = EINVAL;
1690                                         break;
1691                                 }
1692
1693                                 /*
1694                                  * We only ensure valsize is not too large
1695                                  * here.  Further validation will be done
1696                                  * later.
1697                                  */
1698                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1699                                     sizeof(optbuf_storage), 0);
1700                                 if (error)
1701                                         break;
1702                                 optlen = sopt->sopt_valsize;
1703                                 optbuf = optbuf_storage;
1704                                 optp = &in6p->in6p_outputopts;
1705                                 error = ip6_pcbopt(optname, optbuf, optlen,
1706                                     optp, uproto);
1707                                 break;
1708                         }
1709 #undef OPTSET
1710
1711                         case IPV6_MULTICAST_IF:
1712                         case IPV6_MULTICAST_HOPS:
1713                         case IPV6_MULTICAST_LOOP:
1714                         case IPV6_JOIN_GROUP:
1715                         case IPV6_LEAVE_GROUP:
1716                             {
1717                                 struct mbuf *m;
1718                                 if (sopt->sopt_valsize > MLEN) {
1719                                         error = EMSGSIZE;
1720                                         break;
1721                                 }
1722                                 /* XXX */
1723                                 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1724                                 if (m == NULL) {
1725                                         error = ENOBUFS;
1726                                         break;
1727                                 }
1728                                 m->m_len = sopt->sopt_valsize;
1729                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1730                                                     m->m_len, m->m_len);
1731                                 error = ip6_setmoptions(sopt->sopt_name,
1732                                                         &in6p->in6p_moptions,
1733                                                         m);
1734                                 m_free(m);
1735                             }
1736                                 break;
1737
1738                         case IPV6_PORTRANGE:
1739                                 error = soopt_to_kbuf(sopt, &optval,
1740                                     sizeof optval, sizeof optval);
1741                                 if (error)
1742                                         break;
1743
1744                                 switch (optval) {
1745                                 case IPV6_PORTRANGE_DEFAULT:
1746                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1747                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1748                                         break;
1749
1750                                 case IPV6_PORTRANGE_HIGH:
1751                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1752                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1753                                         break;
1754
1755                                 case IPV6_PORTRANGE_LOW:
1756                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1757                                         in6p->in6p_flags |= IN6P_LOWPORT;
1758                                         break;
1759
1760                                 default:
1761                                         error = EINVAL;
1762                                         break;
1763                                 }
1764                                 break;
1765
1766 #if defined(IPSEC) || defined(FAST_IPSEC)
1767                         case IPV6_IPSEC_POLICY:
1768                             {
1769                                 caddr_t req = NULL;
1770                                 size_t len = 0;
1771                                 struct mbuf *m;
1772
1773                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1774                                         break;
1775                                 soopt_to_mbuf(sopt, m);         /* XXX */
1776                                 if (m) {
1777                                         req = mtod(m, caddr_t);
1778                                         len = m->m_len;
1779                                 }
1780                                 error = ipsec6_set_policy(in6p, optname, req,
1781                                                           len, privileged);
1782                                 m_freem(m);
1783                             }
1784                                 break;
1785 #endif /* KAME IPSEC */
1786
1787                         case IPV6_FW_ADD:
1788                         case IPV6_FW_DEL:
1789                         case IPV6_FW_FLUSH:
1790                         case IPV6_FW_ZERO:
1791                             {
1792                                 struct mbuf *m;
1793                                 struct mbuf **mp = &m;
1794
1795                                 if (ip6_fw_ctl_ptr == NULL)
1796                                         return EINVAL;
1797                                 /* XXX */
1798                                 if ((error = soopt_getm(sopt, &m)) != 0)
1799                                         break;
1800                                 /* XXX */
1801                                 soopt_to_mbuf(sopt, m);
1802                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1803                                 m = *mp;
1804                             }
1805                                 break;
1806
1807                         default:
1808                                 error = ENOPROTOOPT;
1809                                 break;
1810                         }
1811                         break;
1812
1813                 case SOPT_GET:
1814                         switch (optname) {
1815                         case IPV6_2292PKTOPTIONS:
1816 #ifdef IPV6_PKTOPTIONS
1817                         case IPV6_PKTOPTIONS:
1818 #endif
1819                                 /*
1820                                  * RFC3542 (effectively) deprecated the
1821                                  * semantics of the 2292-style pktoptions.
1822                                  * Since it was not reliable in nature (i.e.,
1823                                  * applications had to expect the lack of some
1824                                  * information after all), it would make sense
1825                                  * to simplify this part by always returning
1826                                  * empty data.
1827                                  */
1828                                 if (in6p->in6p_options) {
1829                                         struct mbuf *m;
1830                                         m = m_copym(in6p->in6p_options,
1831                                             0, M_COPYALL, MB_WAIT);
1832                                         error = soopt_from_mbuf(sopt, m);
1833                                         if (error == 0)
1834                                                 m_freem(m);
1835                                 } else
1836                                         sopt->sopt_valsize = 0;
1837                                 break;
1838
1839                         case IPV6_RECVHOPOPTS:
1840                         case IPV6_RECVDSTOPTS:
1841                         case IPV6_RECVRTHDRDSTOPTS:
1842                         case IPV6_UNICAST_HOPS:
1843                         case IPV6_RECVPKTINFO:
1844                         case IPV6_RECVHOPLIMIT:
1845                         case IPV6_RECVRTHDR:
1846                         case IPV6_RECVPATHMTU:
1847                         case IPV6_RECVTCLASS:
1848                         case IPV6_AUTOFLOWLABEL:
1849                         case IPV6_FAITH:
1850                         case IPV6_V6ONLY:
1851                         case IPV6_PORTRANGE:
1852                                 switch (optname) {
1853
1854                                 case IPV6_RECVHOPOPTS:
1855                                         optval = OPTBIT(IN6P_HOPOPTS);
1856                                         break;
1857
1858                                 case IPV6_RECVDSTOPTS:
1859                                         optval = OPTBIT(IN6P_DSTOPTS);
1860                                         break;
1861
1862                                 case IPV6_RECVRTHDRDSTOPTS:
1863                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1864                                         break;
1865
1866                                 case IPV6_RECVPKTINFO:
1867                                         optval = OPTBIT(IN6P_PKTINFO);
1868                                         break;
1869
1870                                 case IPV6_RECVHOPLIMIT:
1871                                         optval = OPTBIT(IN6P_HOPLIMIT);
1872                                         break;
1873
1874                                 case IPV6_RECVRTHDR:
1875                                         optval = OPTBIT(IN6P_RTHDR);
1876                                         break;
1877
1878                                 case IPV6_RECVPATHMTU:
1879                                         optval = OPTBIT(IN6P_MTU);
1880                                         break;
1881
1882                                 case IPV6_RECVTCLASS:
1883                                         optval = OPTBIT(IN6P_TCLASS);
1884                                         break;
1885
1886                                 case IPV6_AUTOFLOWLABEL:
1887                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1888                                         break;
1889
1890
1891                                 case IPV6_UNICAST_HOPS:
1892                                         optval = in6p->in6p_hops;
1893                                         break;
1894
1895                                 case IPV6_FAITH:
1896                                         optval = OPTBIT(IN6P_FAITH);
1897                                         break;
1898
1899                                 case IPV6_V6ONLY:
1900                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1901                                         break;
1902
1903                                 case IPV6_PORTRANGE:
1904                                     {
1905                                         int flags;
1906                                         flags = in6p->in6p_flags;
1907                                         if (flags & IN6P_HIGHPORT)
1908                                                 optval = IPV6_PORTRANGE_HIGH;
1909                                         else if (flags & IN6P_LOWPORT)
1910                                                 optval = IPV6_PORTRANGE_LOW;
1911                                         else
1912                                                 optval = 0;
1913                                         break;
1914                                     }
1915                                 }
1916                                 soopt_from_kbuf(sopt, &optval,
1917                                         sizeof optval);
1918                                 break;
1919
1920                         case IPV6_PATHMTU:
1921                         {
1922                                 u_long pmtu = 0;
1923                                 struct ip6_mtuinfo mtuinfo;
1924                                 struct route_in6 sro;
1925
1926                                 bzero(&sro, sizeof(sro));
1927
1928                                 if (!(so->so_state & SS_ISCONNECTED))
1929                                         return (ENOTCONN);
1930                                 /*
1931                                  * XXX: we dot not consider the case of source
1932                                  * routing, or optional information to specify
1933                                  * the outgoing interface.
1934                                  */
1935                                 error = ip6_getpmtu(&sro, NULL, NULL,
1936                                     &in6p->in6p_faddr, &pmtu, NULL);
1937                                 if (sro.ro_rt)
1938                                         RTFREE(sro.ro_rt);
1939                                 if (error)
1940                                         break;
1941                                 if (pmtu > IPV6_MAXPACKET)
1942                                         pmtu = IPV6_MAXPACKET;
1943
1944                                 bzero(&mtuinfo, sizeof(mtuinfo));
1945                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1946                                 optdata = (void *)&mtuinfo;
1947                                 optdatalen = sizeof(mtuinfo);
1948                                 soopt_from_kbuf(sopt, optdata,
1949                                     optdatalen);
1950                                 break;
1951                         }
1952
1953                         case IPV6_2292PKTINFO:
1954                         case IPV6_2292HOPLIMIT:
1955                         case IPV6_2292HOPOPTS:
1956                         case IPV6_2292RTHDR:
1957                         case IPV6_2292DSTOPTS:
1958                                 if (optname == IPV6_2292HOPOPTS ||
1959                                     optname == IPV6_2292DSTOPTS ||
1960                                     !privileged)
1961                                         return (EPERM);
1962                                 switch (optname) {
1963                                 case IPV6_2292PKTINFO:
1964                                         optval = OPTBIT(IN6P_PKTINFO);
1965                                         break;
1966                                 case IPV6_2292HOPLIMIT:
1967                                         optval = OPTBIT(IN6P_HOPLIMIT);
1968                                         break;
1969                                 case IPV6_2292HOPOPTS:
1970                                         if (!privileged)
1971                                                 return (EPERM);
1972                                         optval = OPTBIT(IN6P_HOPOPTS);
1973                                         break;
1974                                 case IPV6_2292RTHDR:
1975                                         optval = OPTBIT(IN6P_RTHDR);
1976                                         break;
1977                                 case IPV6_2292DSTOPTS:
1978                                         if (!privileged)
1979                                                 return (EPERM);
1980                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1981                                         break;
1982                                 }
1983                                 soopt_from_kbuf(sopt, &optval,
1984                                         sizeof optval);
1985                                 break;
1986
1987                         case IPV6_PKTINFO:
1988                         case IPV6_HOPOPTS:
1989                         case IPV6_RTHDR:
1990                         case IPV6_DSTOPTS:
1991                         case IPV6_RTHDRDSTOPTS:
1992                         case IPV6_NEXTHOP:
1993                         case IPV6_TCLASS:
1994                         case IPV6_DONTFRAG:
1995                         case IPV6_USE_MIN_MTU:
1996                         case IPV6_PREFER_TEMPADDR:
1997                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1998                                     optname, sopt);
1999                                 break;
2000
2001                         case IPV6_MULTICAST_IF:
2002                         case IPV6_MULTICAST_HOPS:
2003                         case IPV6_MULTICAST_LOOP:
2004                         case IPV6_JOIN_GROUP:
2005                         case IPV6_LEAVE_GROUP:
2006                             {
2007                                 struct mbuf *m;
2008                                 error = ip6_getmoptions(sopt->sopt_name,
2009                                                 in6p->in6p_moptions, &m);
2010                                 if (error == 0)
2011                                         soopt_from_kbuf(sopt,
2012                                                 mtod(m, char *), m->m_len);
2013                                 m_freem(m);
2014                             }
2015                                 break;
2016
2017 #if defined(IPSEC) || defined(FAST_IPSEC)
2018                         case IPV6_IPSEC_POLICY:
2019                           {
2020                                 caddr_t req = NULL;
2021                                 size_t len = 0;
2022                                 struct mbuf *m = NULL;
2023                                 struct mbuf **mp = &m;
2024
2025                                 error = soopt_getm(sopt, &m); /* XXX */
2026                                 if (error != 0)
2027                                         break;
2028                                 soopt_to_mbuf(sopt, m); /* XXX */
2029                                 if (m) {
2030                                         req = mtod(m, caddr_t);
2031                                         len = m->m_len;
2032                                 }
2033                                 error = ipsec6_get_policy(in6p, req, len, mp);
2034                                 if (error == 0)
2035                                         error = soopt_from_mbuf(sopt, m); /*XXX*/
2036                                 if (error == 0 && m != NULL)
2037                                         m_freem(m);
2038                                 break;
2039                           }
2040 #endif /* KAME IPSEC */
2041
2042                         case IPV6_FW_GET:
2043                           {
2044                                 struct mbuf *m;
2045                                 struct mbuf **mp = &m;
2046
2047                                 if (ip6_fw_ctl_ptr == NULL)
2048                                 {
2049                                         return EINVAL;
2050                                 }
2051                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2052                                 if (error == 0)
2053                                         error = soopt_from_mbuf(sopt, m); /* XXX */
2054                                 if (error == 0 && m != NULL)
2055                                         m_freem(m);
2056                           }
2057                                 break;
2058
2059                         default:
2060                                 error = ENOPROTOOPT;
2061                                 break;
2062                         }
2063                         break;
2064                 }
2065         } else {
2066                 error = EINVAL;
2067         }
2068         return (error);
2069 }
2070
2071 int
2072 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2073 {
2074         int error = 0, optval, optlen;
2075         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2076         struct in6pcb *in6p = sotoin6pcb(so);
2077         int level, op, optname;
2078
2079         if (sopt) {
2080                 level = sopt->sopt_level;
2081                 op = sopt->sopt_dir;
2082                 optname = sopt->sopt_name;
2083                 optlen = sopt->sopt_valsize;
2084         } else
2085                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2086
2087         if (level != IPPROTO_IPV6) {
2088                 return (EINVAL);
2089         }
2090
2091         switch (optname) {
2092         case IPV6_CHECKSUM:
2093                 /*
2094                  * For ICMPv6 sockets, no modification allowed for checksum
2095                  * offset, permit "no change" values to help existing apps.
2096                  *
2097                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2098                  * for an ICMPv6 socket will fail."
2099                  * The current behavior does not meet RFC3542.
2100                  */
2101                 switch (op) {
2102                 case SOPT_SET:
2103                         if (optlen != sizeof(int)) {
2104                                 error = EINVAL;
2105                                 break;
2106                         }
2107                         error = soopt_to_kbuf(sopt, &optval,
2108                                     sizeof optval, sizeof optval);
2109                         if (error)
2110                                 break;
2111                         if ((optval % 2) != 0) {
2112                                 /* the API assumes even offset values */
2113                                 error = EINVAL;
2114                         } else if (so->so_proto->pr_protocol ==
2115                             IPPROTO_ICMPV6) {
2116                                 if (optval != icmp6off)
2117                                         error = EINVAL;
2118                         } else
2119                                 in6p->in6p_cksum = optval;
2120                         break;
2121
2122                 case SOPT_GET:
2123                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2124                                 optval = icmp6off;
2125                         else
2126                                 optval = in6p->in6p_cksum;
2127
2128                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
2129                         break;
2130
2131                 default:
2132                         error = EINVAL;
2133                         break;
2134                 }
2135                 break;
2136
2137         default:
2138                 error = ENOPROTOOPT;
2139                 break;
2140         }
2141
2142         return (error);
2143 }
2144
2145 /*
2146  * Set up IP6 options in pcb for insertion in output packets or
2147  * specifying behavior of outgoing packets.
2148  */
2149 static int
2150 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2151     struct socket *so, struct sockopt *sopt)
2152 {
2153         int priv = 0;
2154         struct ip6_pktopts *opt = *pktopt;
2155         int error = 0;
2156
2157         /* turn off any old options. */
2158         if (opt) {
2159 #ifdef DIAGNOSTIC
2160                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2161                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2162                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2163                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
2164 #endif
2165                 ip6_clearpktopts(opt, -1);
2166         } else
2167                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2168         *pktopt = NULL;
2169
2170         if (!m || m->m_len == 0) {
2171                 /*
2172                  * Only turning off any previous options, regardless of
2173                  * whether the opt is just created or given.
2174                  */
2175                 kfree(opt, M_IP6OPT);
2176                 return (0);
2177         }
2178
2179         /*  set options specified by user. */
2180         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2181                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2182                 kfree(opt, M_IP6OPT);
2183                 return (error);
2184         }
2185         *pktopt = opt;
2186         return (0);
2187 }
2188
2189
2190 /*
2191  * Below three functions are introduced by merge to RFC3542
2192  */
2193
2194 static int
2195 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2196 {
2197         void *optdata = NULL;
2198         int optdatalen = 0;
2199         struct ip6_ext *ip6e;
2200         int error = 0;
2201         struct in6_pktinfo null_pktinfo;
2202         int deftclass = 0, on;
2203         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2204         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2205
2206         switch (optname) {
2207         case IPV6_PKTINFO:
2208                 if (pktopt && pktopt->ip6po_pktinfo)
2209                         optdata = (void *)pktopt->ip6po_pktinfo;
2210                 else {
2211                         /* XXX: we don't have to do this every time... */
2212                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2213                         optdata = (void *)&null_pktinfo;
2214                 }
2215                 optdatalen = sizeof(struct in6_pktinfo);
2216                 break;
2217         case IPV6_TCLASS:
2218                 if (pktopt && pktopt->ip6po_tclass >= 0)
2219                         optdata = (void *)&pktopt->ip6po_tclass;
2220                 else
2221                         optdata = (void *)&deftclass;
2222                 optdatalen = sizeof(int);
2223                 break;
2224         case IPV6_HOPOPTS:
2225                 if (pktopt && pktopt->ip6po_hbh) {
2226                         optdata = (void *)pktopt->ip6po_hbh;
2227                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2228                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2229                 }
2230                 break;
2231         case IPV6_RTHDR:
2232                 if (pktopt && pktopt->ip6po_rthdr) {
2233                         optdata = (void *)pktopt->ip6po_rthdr;
2234                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2235                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2236                 }
2237                 break;
2238         case IPV6_RTHDRDSTOPTS:
2239                 if (pktopt && pktopt->ip6po_dest1) {
2240                         optdata = (void *)pktopt->ip6po_dest1;
2241                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2242                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2243                 }
2244                 break;
2245         case IPV6_DSTOPTS:
2246                 if (pktopt && pktopt->ip6po_dest2) {
2247                         optdata = (void *)pktopt->ip6po_dest2;
2248                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2249                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2250                 }
2251                 break;
2252         case IPV6_NEXTHOP:
2253                 if (pktopt && pktopt->ip6po_nexthop) {
2254                         optdata = (void *)pktopt->ip6po_nexthop;
2255                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2256                 }
2257                 break;
2258         case IPV6_USE_MIN_MTU:
2259                 if (pktopt)
2260                         optdata = (void *)&pktopt->ip6po_minmtu;
2261                 else
2262                         optdata = (void *)&defminmtu;
2263                 optdatalen = sizeof(int);
2264                 break;
2265         case IPV6_DONTFRAG:
2266                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2267                         on = 1;
2268                 else
2269                         on = 0;
2270                 optdata = (void *)&on;
2271                 optdatalen = sizeof(on);
2272                 break;
2273         case IPV6_PREFER_TEMPADDR:
2274                 if (pktopt)
2275                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2276                 else
2277                         optdata = (void *)&defpreftemp;
2278                 optdatalen = sizeof(int);
2279                 break;
2280         default:                /* should not happen */
2281 #ifdef DIAGNOSTIC
2282                 panic("ip6_getpcbopt: unexpected option");
2283 #endif
2284                 return (ENOPROTOOPT);
2285         }
2286
2287         soopt_from_kbuf(sopt, optdata, optdatalen);
2288
2289         return (error);
2290 }
2291
2292 /*
2293  * initialize ip6_pktopts.  beware that there are non-zero default values in
2294  * the struct.
2295  */
2296
2297 static int
2298 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2299 {
2300         struct ip6_pktopts *opt;
2301         int priv =0;
2302         if (*pktopt == NULL) {
2303                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2304                 init_ip6pktopts(*pktopt);
2305         }
2306         opt = *pktopt;
2307
2308         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2309 }
2310
2311 /*
2312  * initialize ip6_pktopts.  beware that there are non-zero default values in
2313  * the struct.
2314  */
2315 void
2316 init_ip6pktopts(struct ip6_pktopts *opt)
2317 {
2318
2319         bzero(opt, sizeof(*opt));
2320         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2321         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2322         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2323         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2324 }
2325
2326 void
2327 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2328 {
2329         if (pktopt == NULL)
2330                 return;
2331
2332         if (optname == -1 || optname == IPV6_PKTINFO) {
2333                 if (pktopt->ip6po_pktinfo)
2334                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2335                 pktopt->ip6po_pktinfo = NULL;
2336         }
2337         if (optname == -1 || optname == IPV6_HOPLIMIT)
2338                 pktopt->ip6po_hlim = -1;
2339         if (optname == -1 || optname == IPV6_TCLASS)
2340                 pktopt->ip6po_tclass = -1;
2341         if (optname == -1 || optname == IPV6_NEXTHOP) {
2342                 if (pktopt->ip6po_nextroute.ro_rt) {
2343                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2344                         pktopt->ip6po_nextroute.ro_rt = NULL;
2345                 }
2346                 if (pktopt->ip6po_nexthop)
2347                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2348                 pktopt->ip6po_nexthop = NULL;
2349         }
2350         if (optname == -1 || optname == IPV6_HOPOPTS) {
2351                 if (pktopt->ip6po_hbh)
2352                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2353                 pktopt->ip6po_hbh = NULL;
2354         }
2355         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2356                 if (pktopt->ip6po_dest1)
2357                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2358                 pktopt->ip6po_dest1 = NULL;
2359         }
2360         if (optname == -1 || optname == IPV6_RTHDR) {
2361                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2362                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2363                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2364                 if (pktopt->ip6po_route.ro_rt) {
2365                         RTFREE(pktopt->ip6po_route.ro_rt);
2366                         pktopt->ip6po_route.ro_rt = NULL;
2367                 }
2368         }
2369         if (optname == -1 || optname == IPV6_DSTOPTS) {
2370                 if (pktopt->ip6po_dest2)
2371                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2372                 pktopt->ip6po_dest2 = NULL;
2373         }
2374 }
2375
2376 #define PKTOPT_EXTHDRCPY(type) \
2377 do {\
2378         if (src->type) {\
2379                 int hlen =\
2380                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2381                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2382                 if (dst->type == NULL)\
2383                         goto bad;\
2384                 bcopy(src->type, dst->type, hlen);\
2385         }\
2386 } while (0)
2387
2388 struct ip6_pktopts *
2389 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2390 {
2391         struct ip6_pktopts *dst;
2392
2393         if (src == NULL) {
2394                 kprintf("ip6_clearpktopts: invalid argument\n");
2395                 return (NULL);
2396         }
2397
2398         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2399         if (dst == NULL)
2400                 return (NULL);
2401
2402         dst->ip6po_hlim = src->ip6po_hlim;
2403         if (src->ip6po_pktinfo) {
2404                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2405                                             M_IP6OPT, canwait);
2406                 if (dst->ip6po_pktinfo == NULL)
2407                         goto bad;
2408                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2409         }
2410         if (src->ip6po_nexthop) {
2411                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2412                                             M_IP6OPT, canwait);
2413                 if (dst->ip6po_nexthop == NULL)
2414                         goto bad;
2415                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2416                       src->ip6po_nexthop->sa_len);
2417         }
2418         PKTOPT_EXTHDRCPY(ip6po_hbh);
2419         PKTOPT_EXTHDRCPY(ip6po_dest1);
2420         PKTOPT_EXTHDRCPY(ip6po_dest2);
2421         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2422         return (dst);
2423
2424 bad:
2425         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2426         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2427         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2428         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2429         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2430         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2431         kfree(dst, M_IP6OPT);
2432         return (NULL);
2433 }
2434
2435 static int
2436 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2437 {
2438         if (dst == NULL || src == NULL)  {
2439 #ifdef DIAGNOSTIC
2440                 kprintf("ip6_clearpktopts: invalid argument\n");
2441 #endif
2442                 return (EINVAL);
2443         }
2444
2445         dst->ip6po_hlim = src->ip6po_hlim;
2446         dst->ip6po_tclass = src->ip6po_tclass;
2447         dst->ip6po_flags = src->ip6po_flags;
2448         if (src->ip6po_pktinfo) {
2449                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2450                     M_IP6OPT, canwait);
2451                 if (dst->ip6po_pktinfo == NULL)
2452                         goto bad;
2453                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2454         }
2455         if (src->ip6po_nexthop) {
2456                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2457                     M_IP6OPT, canwait);
2458                 if (dst->ip6po_nexthop == NULL)
2459                         goto bad;
2460                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2461                     src->ip6po_nexthop->sa_len);
2462         }
2463         PKTOPT_EXTHDRCPY(ip6po_hbh);
2464         PKTOPT_EXTHDRCPY(ip6po_dest1);
2465         PKTOPT_EXTHDRCPY(ip6po_dest2);
2466         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2467         return (0);
2468
2469   bad:
2470         ip6_clearpktopts(dst, -1);
2471         return (ENOBUFS);
2472 }
2473 #undef PKTOPT_EXTHDRCPY
2474
2475 void
2476 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2477 {
2478         if (pktopt == NULL)
2479                 return;
2480
2481         ip6_clearpktopts(pktopt, -1);
2482
2483         kfree(pktopt, M_IP6OPT);
2484 }
2485
2486 /*
2487  * Set the IP6 multicast options in response to user setsockopt().
2488  */
2489 static int
2490 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2491 {
2492         int error = 0;
2493         u_int loop, ifindex;
2494         struct ipv6_mreq *mreq;
2495         struct ifnet *ifp;
2496         struct ip6_moptions *im6o = *im6op;
2497         struct route_in6 ro;
2498         struct sockaddr_in6 *dst;
2499         struct in6_multi_mship *imm;
2500         struct thread *td = curthread;  /* XXX */
2501
2502         if (im6o == NULL) {
2503                 /*
2504                  * No multicast option buffer attached to the pcb;
2505                  * allocate one and initialize to default values.
2506                  */
2507                 im6o = (struct ip6_moptions *)
2508                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2509
2510                 *im6op = im6o;
2511                 im6o->im6o_multicast_ifp = NULL;
2512                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2513                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2514                 LIST_INIT(&im6o->im6o_memberships);
2515         }
2516
2517         switch (optname) {
2518
2519         case IPV6_MULTICAST_IF:
2520                 /*
2521                  * Select the interface for outgoing multicast packets.
2522                  */
2523                 if (m == NULL || m->m_len != sizeof(u_int)) {
2524                         error = EINVAL;
2525                         break;
2526                 }
2527                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2528                 if (ifindex < 0 || if_index < ifindex) {
2529                         error = ENXIO;  /* XXX EINVAL? */
2530                         break;
2531                 }
2532                 ifp = ifindex2ifnet[ifindex];
2533                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2534                         error = EADDRNOTAVAIL;
2535                         break;
2536                 }
2537                 im6o->im6o_multicast_ifp = ifp;
2538                 break;
2539
2540         case IPV6_MULTICAST_HOPS:
2541             {
2542                 /*
2543                  * Set the IP6 hoplimit for outgoing multicast packets.
2544                  */
2545                 int optval;
2546                 if (m == NULL || m->m_len != sizeof(int)) {
2547                         error = EINVAL;
2548                         break;
2549                 }
2550                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2551                 if (optval < -1 || optval >= 256)
2552                         error = EINVAL;
2553                 else if (optval == -1)
2554                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2555                 else
2556                         im6o->im6o_multicast_hlim = optval;
2557                 break;
2558             }
2559
2560         case IPV6_MULTICAST_LOOP:
2561                 /*
2562                  * Set the loopback flag for outgoing multicast packets.
2563                  * Must be zero or one.
2564                  */
2565                 if (m == NULL || m->m_len != sizeof(u_int)) {
2566                         error = EINVAL;
2567                         break;
2568                 }
2569                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2570                 if (loop > 1) {
2571                         error = EINVAL;
2572                         break;
2573                 }
2574                 im6o->im6o_multicast_loop = loop;
2575                 break;
2576
2577         case IPV6_JOIN_GROUP:
2578                 /*
2579                  * Add a multicast group membership.
2580                  * Group must be a valid IP6 multicast address.
2581                  */
2582                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2583                         error = EINVAL;
2584                         break;
2585                 }
2586                 mreq = mtod(m, struct ipv6_mreq *);
2587                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2588                         /*
2589                          * We use the unspecified address to specify to accept
2590                          * all multicast addresses. Only super user is allowed
2591                          * to do this.
2592                          */
2593                         if (priv_check(td, PRIV_ROOT))
2594                         {
2595                                 error = EACCES;
2596                                 break;
2597                         }
2598                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2599                         error = EINVAL;
2600                         break;
2601                 }
2602
2603                 /*
2604                  * If the interface is specified, validate it.
2605                  */
2606                 if (mreq->ipv6mr_interface < 0
2607                  || if_index < mreq->ipv6mr_interface) {
2608                         error = ENXIO;  /* XXX EINVAL? */
2609                         break;
2610                 }
2611                 /*
2612                  * If no interface was explicitly specified, choose an
2613                  * appropriate one according to the given multicast address.
2614                  */
2615                 if (mreq->ipv6mr_interface == 0) {
2616                         /*
2617                          * If the multicast address is in node-local scope,
2618                          * the interface should be a loopback interface.
2619                          * Otherwise, look up the routing table for the
2620                          * address, and choose the outgoing interface.
2621                          *   XXX: is it a good approach?
2622                          */
2623                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2624                                 ifp = &loif[0];
2625                         } else {
2626                                 ro.ro_rt = NULL;
2627                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2628                                 bzero(dst, sizeof(*dst));
2629                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2630                                 dst->sin6_family = AF_INET6;
2631                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2632                                 rtalloc((struct route *)&ro);
2633                                 if (ro.ro_rt == NULL) {
2634                                         error = EADDRNOTAVAIL;
2635                                         break;
2636                                 }
2637                                 ifp = ro.ro_rt->rt_ifp;
2638                                 rtfree(ro.ro_rt);
2639                         }
2640                 } else
2641                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2642
2643                 /*
2644                  * See if we found an interface, and confirm that it
2645                  * supports multicast
2646                  */
2647                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2648                         error = EADDRNOTAVAIL;
2649                         break;
2650                 }
2651                 /*
2652                  * Put interface index into the multicast address,
2653                  * if the address has link-local scope.
2654                  */
2655                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2656                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2657                                 = htons(mreq->ipv6mr_interface);
2658                 }
2659                 /*
2660                  * See if the membership already exists.
2661                  */
2662                 for (imm = im6o->im6o_memberships.lh_first;
2663                      imm != NULL; imm = imm->i6mm_chain.le_next)
2664                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2665                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2666                                                &mreq->ipv6mr_multiaddr))
2667                                 break;
2668                 if (imm != NULL) {
2669                         error = EADDRINUSE;
2670                         break;
2671                 }
2672                 /*
2673                  * Everything looks good; add a new record to the multicast
2674                  * address list for the given interface.
2675                  */
2676                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2677                 if ((imm->i6mm_maddr =
2678                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2679                         kfree(imm, M_IPMADDR);
2680                         break;
2681                 }
2682                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2683                 break;
2684
2685         case IPV6_LEAVE_GROUP:
2686                 /*
2687                  * Drop a multicast group membership.
2688                  * Group must be a valid IP6 multicast address.
2689                  */
2690                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2691                         error = EINVAL;
2692                         break;
2693                 }
2694                 mreq = mtod(m, struct ipv6_mreq *);
2695                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2696                         if (priv_check(td, PRIV_ROOT)) {
2697                                 error = EACCES;
2698                                 break;
2699                         }
2700                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2701                         error = EINVAL;
2702                         break;
2703                 }
2704                 /*
2705                  * If an interface address was specified, get a pointer
2706                  * to its ifnet structure.
2707                  */
2708                 if (mreq->ipv6mr_interface < 0
2709                  || if_index < mreq->ipv6mr_interface) {
2710                         error = ENXIO;  /* XXX EINVAL? */
2711                         break;
2712                 }
2713                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2714                 /*
2715                  * Put interface index into the multicast address,
2716                  * if the address has link-local scope.
2717                  */
2718                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2719                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2720                                 = htons(mreq->ipv6mr_interface);
2721                 }
2722                 /*
2723                  * Find the membership in the membership list.
2724                  */
2725                 for (imm = im6o->im6o_memberships.lh_first;
2726                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2727                         if ((ifp == NULL ||
2728                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2729                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2730                                                &mreq->ipv6mr_multiaddr))
2731                                 break;
2732                 }
2733                 if (imm == NULL) {
2734                         /* Unable to resolve interface */
2735                         error = EADDRNOTAVAIL;
2736                         break;
2737                 }
2738                 /*
2739                  * Give up the multicast address record to which the
2740                  * membership points.
2741                  */
2742                 LIST_REMOVE(imm, i6mm_chain);
2743                 in6_delmulti(imm->i6mm_maddr);
2744                 kfree(imm, M_IPMADDR);
2745                 break;
2746
2747         default:
2748                 error = EOPNOTSUPP;
2749                 break;
2750         }
2751
2752         /*
2753          * If all options have default values, no need to keep the mbuf.
2754          */
2755         if (im6o->im6o_multicast_ifp == NULL &&
2756             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2757             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2758             im6o->im6o_memberships.lh_first == NULL) {
2759                 kfree(*im6op, M_IPMOPTS);
2760                 *im6op = NULL;
2761         }
2762
2763         return (error);
2764 }
2765
2766 /*
2767  * Return the IP6 multicast options in response to user getsockopt().
2768  */
2769 static int
2770 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2771 {
2772         u_int *hlim, *loop, *ifindex;
2773
2774         *mp = m_get(MB_WAIT, MT_HEADER);                /* XXX */
2775
2776         switch (optname) {
2777
2778         case IPV6_MULTICAST_IF:
2779                 ifindex = mtod(*mp, u_int *);
2780                 (*mp)->m_len = sizeof(u_int);
2781                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2782                         *ifindex = 0;
2783                 else
2784                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2785                 return (0);
2786
2787         case IPV6_MULTICAST_HOPS:
2788                 hlim = mtod(*mp, u_int *);
2789                 (*mp)->m_len = sizeof(u_int);
2790                 if (im6o == NULL)
2791                         *hlim = ip6_defmcasthlim;
2792                 else
2793                         *hlim = im6o->im6o_multicast_hlim;
2794                 return (0);
2795
2796         case IPV6_MULTICAST_LOOP:
2797                 loop = mtod(*mp, u_int *);
2798                 (*mp)->m_len = sizeof(u_int);
2799                 if (im6o == NULL)
2800                         *loop = ip6_defmcasthlim;
2801                 else
2802                         *loop = im6o->im6o_multicast_loop;
2803                 return (0);
2804
2805         default:
2806                 return (EOPNOTSUPP);
2807         }
2808 }
2809
2810 /*
2811  * Discard the IP6 multicast options.
2812  */
2813 void
2814 ip6_freemoptions(struct ip6_moptions *im6o)
2815 {
2816         struct in6_multi_mship *imm;
2817
2818         if (im6o == NULL)
2819                 return;
2820
2821         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2822                 LIST_REMOVE(imm, i6mm_chain);
2823                 if (imm->i6mm_maddr)
2824                         in6_delmulti(imm->i6mm_maddr);
2825                 kfree(imm, M_IPMADDR);
2826         }
2827         kfree(im6o, M_IPMOPTS);
2828 }
2829
2830 /*
2831  * Set a particular packet option, as a sticky option or an ancillary data
2832  * item.  "len" can be 0 only when it's a sticky option.
2833  * We have 4 cases of combination of "sticky" and "cmsg":
2834  * "sticky=0, cmsg=0": impossible
2835  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2836  * "sticky=1, cmsg=0": RFC3542 socket option
2837  * "sticky=1, cmsg=1": RFC2292 socket option
2838  */
2839 static int
2840 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2841      int sticky, int cmsg, int uproto, int priv)
2842 {
2843         int minmtupolicy, preftemp;
2844         //int error;
2845
2846         if (!sticky && !cmsg) {
2847                 kprintf("ip6_setpktoption: impossible case\n");
2848                 return (EINVAL);
2849         }
2850
2851         /*
2852          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2853          * not be specified in the context of RFC3542.  Conversely,
2854          * RFC3542 types should not be specified in the context of RFC2292.
2855          */
2856         if (!cmsg) {
2857                 switch (optname) {
2858                 case IPV6_2292PKTINFO:
2859                 case IPV6_2292HOPLIMIT:
2860                 case IPV6_2292NEXTHOP:
2861                 case IPV6_2292HOPOPTS:
2862                 case IPV6_2292DSTOPTS:
2863                 case IPV6_2292RTHDR:
2864                 case IPV6_2292PKTOPTIONS:
2865                         return (ENOPROTOOPT);
2866                 }
2867         }
2868         if (sticky && cmsg) {
2869                 switch (optname) {
2870                 case IPV6_PKTINFO:
2871                 case IPV6_HOPLIMIT:
2872                 case IPV6_NEXTHOP:
2873                 case IPV6_HOPOPTS:
2874                 case IPV6_DSTOPTS:
2875                 case IPV6_RTHDRDSTOPTS:
2876                 case IPV6_RTHDR:
2877                 case IPV6_USE_MIN_MTU:
2878                 case IPV6_DONTFRAG:
2879                 case IPV6_TCLASS:
2880                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2881                         return (ENOPROTOOPT);
2882                 }
2883         }
2884
2885         switch (optname) {
2886         case IPV6_2292PKTINFO:
2887         case IPV6_PKTINFO:
2888         {
2889                 struct in6_pktinfo *pktinfo;
2890                 if (len != sizeof(struct in6_pktinfo))
2891                         return (EINVAL);
2892                 pktinfo = (struct in6_pktinfo *)buf;
2893
2894                 /*
2895                  * An application can clear any sticky IPV6_PKTINFO option by
2896                  * doing a "regular" setsockopt with ipi6_addr being
2897                  * in6addr_any and ipi6_ifindex being zero.
2898                  * [RFC 3542, Section 6]
2899                  */
2900                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2901                     pktinfo->ipi6_ifindex == 0 &&
2902                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2903                         ip6_clearpktopts(opt, optname);
2904                         break;
2905                 }
2906
2907                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2908                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2909                         return (EINVAL);
2910                 }
2911
2912                 /* validate the interface index if specified. */
2913                 if (pktinfo->ipi6_ifindex > if_index ||
2914                     pktinfo->ipi6_ifindex < 0) {
2915                          return (ENXIO);
2916                 }
2917                 /*
2918                  * Check if the requested source address is indeed a
2919                  * unicast address assigned to the node, and can be
2920                  * used as the packet's source address.
2921                  */
2922                 if (opt->ip6po_pktinfo != NULL &&
2923                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2924                         struct in6_ifaddr *ia6;
2925                         struct sockaddr_in6 sin6;
2926
2927                         bzero(&sin6, sizeof(sin6));
2928                         sin6.sin6_len = sizeof(sin6);
2929                         sin6.sin6_family = AF_INET6;
2930                         sin6.sin6_addr =
2931                         opt->ip6po_pktinfo->ipi6_addr;
2932                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2933                         if (ia6 == NULL ||
2934                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2935                                         IN6_IFF_NOTREADY)) != 0)
2936                         return (EADDRNOTAVAIL);
2937                 }
2938
2939                 /*
2940                  * We store the address anyway, and let in6_selectsrc()
2941                  * validate the specified address.  This is because ipi6_addr
2942                  * may not have enough information about its scope zone, and
2943                  * we may need additional information (such as outgoing
2944                  * interface or the scope zone of a destination address) to
2945                  * disambiguate the scope.
2946                  * XXX: the delay of the validation may confuse the
2947                  * application when it is used as a sticky option.
2948                  */
2949                 if (opt->ip6po_pktinfo == NULL) {
2950                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2951                             M_IP6OPT, M_NOWAIT);
2952                         if (opt->ip6po_pktinfo == NULL)
2953                                 return (ENOBUFS);
2954                 }
2955                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2956                 break;
2957         }
2958
2959         case IPV6_2292HOPLIMIT:
2960         case IPV6_HOPLIMIT:
2961         {
2962                 int *hlimp;
2963
2964                 /*
2965                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2966                  * to simplify the ordering among hoplimit options.
2967                  */
2968                 if (optname == IPV6_HOPLIMIT && sticky)
2969                         return (ENOPROTOOPT);
2970
2971                 if (len != sizeof(int))
2972                         return (EINVAL);
2973                 hlimp = (int *)buf;
2974                 if (*hlimp < -1 || *hlimp > 255)
2975                         return (EINVAL);
2976
2977                 opt->ip6po_hlim = *hlimp;
2978                 break;
2979         }
2980
2981         case IPV6_TCLASS:
2982         {
2983                 int tclass;
2984
2985                 if (len != sizeof(int))
2986                         return (EINVAL);
2987                 tclass = *(int *)buf;
2988                 if (tclass < -1 || tclass > 255)
2989                         return (EINVAL);
2990
2991                 opt->ip6po_tclass = tclass;
2992                 break;
2993         }
2994
2995         case IPV6_2292NEXTHOP:
2996         case IPV6_NEXTHOP:
2997                 if (!priv)
2998                         return (EPERM);
2999
3000                 if (len == 0) { /* just remove the option */
3001                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3002                         break;
3003                 }
3004
3005                 /* check if cmsg_len is large enough for sa_len */
3006                 if (len < sizeof(struct sockaddr) || len < *buf)
3007                         return (EINVAL);
3008
3009                 switch (((struct sockaddr *)buf)->sa_family) {
3010                 case AF_INET6:
3011                 {
3012                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3013                         //int error;
3014
3015                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3016                                 return (EINVAL);
3017
3018                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3019                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3020                                 return (EINVAL);
3021                         }
3022                         break;
3023                 }
3024                 case AF_LINK:   /* should eventually be supported */
3025                 default:
3026                         return (EAFNOSUPPORT);
3027                 }
3028
3029                 /* turn off the previous option, then set the new option. */
3030                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3031                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3032                 if (opt->ip6po_nexthop == NULL)
3033                         return (ENOBUFS);
3034                 bcopy(buf, opt->ip6po_nexthop, *buf);
3035                 break;
3036
3037         case IPV6_2292HOPOPTS:
3038         case IPV6_HOPOPTS:
3039         {
3040                 struct ip6_hbh *hbh;
3041                 int hbhlen;
3042
3043                 /*
3044                  * XXX: We don't allow a non-privileged user to set ANY HbH
3045                  * options, since per-option restriction has too much
3046                  * overhead.
3047                  */
3048                 if (!priv)
3049                         return (EPERM);
3050                 if (len == 0) {
3051                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3052                         break;  /* just remove the option */
3053                 }
3054
3055                 /* message length validation */
3056                 if (len < sizeof(struct ip6_hbh))
3057                         return (EINVAL);
3058                 hbh = (struct ip6_hbh *)buf;
3059                 hbhlen = (hbh->ip6h_len + 1) << 3;
3060                 if (len != hbhlen)
3061                         return (EINVAL);
3062
3063                 /* turn off the previous option, then set the new option. */
3064                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3065                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3066                 if (opt->ip6po_hbh == NULL)
3067                         return (ENOBUFS);
3068                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3069
3070                 break;
3071         }
3072
3073         case IPV6_2292DSTOPTS:
3074         case IPV6_DSTOPTS:
3075         case IPV6_RTHDRDSTOPTS:
3076         {
3077                 struct ip6_dest *dest, **newdest = NULL;
3078                 int destlen;
3079                 if (!priv)
3080                         return (EPERM);
3081
3082                 if (len == 0) {
3083                         ip6_clearpktopts(opt, optname);
3084                         break;  /* just remove the option */
3085                 }
3086
3087                 /* message length validation */
3088                 if (len < sizeof(struct ip6_dest))
3089                         return (EINVAL);
3090                 dest = (struct ip6_dest *)buf;
3091                 destlen = (dest->ip6d_len + 1) << 3;
3092                 if (len != destlen)
3093                         return (EINVAL);
3094
3095                 /*
3096                  * Determine the position that the destination options header
3097                  * should be inserted; before or after the routing header.
3098                  */
3099                 switch (optname) {
3100                 case IPV6_2292DSTOPTS:
3101                         /*
3102                          * The old advacned API is ambiguous on this point.
3103                          * Our approach is to determine the position based
3104                          * according to the existence of a routing header.
3105                          * Note, however, that this depends on the order of the
3106                          * extension headers in the ancillary data; the 1st
3107                          * part of the destination options header must appear
3108                          * before the routing header in the ancillary data,
3109                          * too.
3110                          * RFC3542 solved the ambiguity by introducing
3111                          * separate ancillary data or option types.
3112                          */
3113                         if (opt->ip6po_rthdr == NULL)
3114                                 newdest = &opt->ip6po_dest1;
3115                         else
3116                                 newdest = &opt->ip6po_dest2;
3117                         break;
3118                 case IPV6_RTHDRDSTOPTS:
3119                         newdest = &opt->ip6po_dest1;
3120                         break;
3121                 case IPV6_DSTOPTS:
3122                         newdest = &opt->ip6po_dest2;
3123                         break;
3124                 }
3125
3126                 /* turn off the previous option, then set the new option. */
3127                 ip6_clearpktopts(opt, optname);
3128                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3129                 if (*newdest == NULL)
3130                         return (ENOBUFS);
3131                 bcopy(dest, *newdest, destlen);
3132
3133                 break;
3134         }
3135
3136         case IPV6_2292RTHDR:
3137         case IPV6_RTHDR:
3138         {
3139                 struct ip6_rthdr *rth;
3140                 int rthlen;
3141
3142                 if (len == 0) {
3143                         ip6_clearpktopts(opt, IPV6_RTHDR);
3144                         break;  /* just remove the option */
3145                 }
3146
3147                 /* message length validation */
3148                 if (len < sizeof(struct ip6_rthdr))
3149                         return (EINVAL);
3150                 rth = (struct ip6_rthdr *)buf;
3151                 rthlen = (rth->ip6r_len + 1) << 3;
3152                 if (len != rthlen)
3153                         return (EINVAL);
3154
3155                 switch (rth->ip6r_type) {
3156                 default:
3157                         return (EINVAL);        /* not supported */
3158                 }
3159
3160                 /* turn off the previous option */
3161                 ip6_clearpktopts(opt, IPV6_RTHDR);
3162                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3163                 if (opt->ip6po_rthdr == NULL)
3164                         return (ENOBUFS);
3165                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3166
3167                 break;
3168         }
3169
3170         case IPV6_USE_MIN_MTU:
3171                 if (len != sizeof(int))
3172                         return (EINVAL);
3173                 minmtupolicy = *(int *)buf;
3174                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3175                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3176                     minmtupolicy != IP6PO_MINMTU_ALL) {
3177                         return (EINVAL);
3178                 }
3179                 opt->ip6po_minmtu = minmtupolicy;
3180                 break;
3181
3182         case IPV6_DONTFRAG:
3183                 if (len != sizeof(int))
3184                         return (EINVAL);
3185
3186                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3187                         /*
3188                          * we ignore this option for TCP sockets.
3189                          * (RFC3542 leaves this case unspecified.)
3190                          */
3191                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3192                 } else
3193                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3194                 break;
3195
3196         case IPV6_PREFER_TEMPADDR:
3197                 if (len != sizeof(int))
3198                         return (EINVAL);
3199                 preftemp = *(int *)buf;
3200                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3201                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3202                     preftemp != IP6PO_TEMPADDR_PREFER) {
3203                         return (EINVAL);
3204                 }
3205                 opt->ip6po_prefer_tempaddr = preftemp;
3206                 break;
3207
3208         default:
3209                 return (ENOPROTOOPT);
3210         } /* end of switch */
3211
3212         return (0);
3213 }
3214
3215
3216 /*
3217  * Set IPv6 outgoing packet options based on advanced API.
3218  */
3219 int
3220 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3221     struct ip6_pktopts *stickyopt, int uproto, int priv)
3222 {
3223         struct cmsghdr *cm = NULL;
3224
3225         if (control == NULL || opt == NULL)
3226                 return (EINVAL);
3227
3228         init_ip6pktopts(opt);
3229
3230         /*
3231          * XXX: Currently, we assume all the optional information is stored
3232          * in a single mbuf.
3233          */
3234         if (stickyopt) {
3235                 int error;
3236
3237                 /*
3238                  * If stickyopt is provided, make a local copy of the options
3239                  * for this particular packet, then override them by ancillary
3240                  * objects.
3241                  * XXX: copypktopts() does not copy the cached route to a next
3242                  * hop (if any).  This is not very good in terms of efficiency,
3243                  * but we can allow this since this option should be rarely
3244                  * used.
3245                  */
3246                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3247                         return (error);
3248         }
3249
3250         /*
3251          * XXX: Currently, we assume all the optional information is stored
3252          * in a single mbuf.
3253          */
3254         if (control->m_next)
3255                 return (EINVAL);
3256
3257         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3258             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3259                 int error;
3260
3261                 if (control->m_len < CMSG_LEN(0))
3262                         return (EINVAL);
3263
3264                 cm = mtod(control, struct cmsghdr *);
3265                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3266                         return (EINVAL);
3267                 if (cm->cmsg_level != IPPROTO_IPV6)
3268                         continue;
3269
3270                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3271                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3272                 if (error)
3273                         return (error);
3274         }
3275
3276         return (0);
3277 }
3278
3279 /*
3280  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3281  * packet to the input queue of a specified interface.  Note that this
3282  * calls the output routine of the loopback "driver", but with an interface
3283  * pointer that might NOT be &loif -- easier than replicating that code here.
3284  */
3285 void
3286 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3287 {
3288         struct mbuf *copym;
3289         struct ip6_hdr *ip6;
3290
3291         copym = m_copy(m, 0, M_COPYALL);
3292         if (copym == NULL)
3293                 return;
3294
3295         /*
3296          * Make sure to deep-copy IPv6 header portion in case the data
3297          * is in an mbuf cluster, so that we can safely override the IPv6
3298          * header portion later.
3299          */
3300         if ((copym->m_flags & M_EXT) != 0 ||
3301             copym->m_len < sizeof(struct ip6_hdr)) {
3302                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3303                 if (copym == NULL)
3304                         return;
3305         }
3306
3307 #ifdef DIAGNOSTIC
3308         if (copym->m_len < sizeof(*ip6)) {
3309                 m_freem(copym);
3310                 return;
3311         }
3312 #endif
3313
3314         ip6 = mtod(copym, struct ip6_hdr *);
3315         /*
3316          * clear embedded scope identifiers if necessary.
3317          * in6_clearscope will touch the addresses only when necessary.
3318          */
3319         in6_clearscope(&ip6->ip6_src);
3320         in6_clearscope(&ip6->ip6_dst);
3321
3322         if_simloop(ifp, copym, dst->sin6_family, 0);
3323 }
3324
3325 /*
3326  * Separate the IPv6 header from the payload into its own mbuf.
3327  *
3328  * Returns the new mbuf chain or the original mbuf if no payload.
3329  * Returns NULL if can't allocate new mbuf for header.
3330  */
3331 static struct mbuf *
3332 ip6_splithdr(struct mbuf *m)
3333 {
3334         struct mbuf *mh;
3335
3336         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3337                 return (m);
3338
3339         MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
3340         if (mh == NULL)
3341                 return (NULL);
3342         mh->m_len = sizeof(struct ip6_hdr);
3343         M_MOVE_PKTHDR(mh, m);
3344         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3345         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3346         m->m_data += sizeof(struct ip6_hdr);
3347         m->m_len -= sizeof(struct ip6_hdr);
3348         mh->m_next = m;
3349         return (mh);
3350 }
3351
3352 /*
3353  * Compute IPv6 extension header length.
3354  */
3355 int
3356 ip6_optlen(struct in6pcb *in6p)
3357 {
3358         int len;
3359
3360         if (!in6p->in6p_outputopts)
3361                 return 0;
3362
3363         len = 0;
3364 #define elen(x) \
3365     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3366
3367         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3368         if (in6p->in6p_outputopts->ip6po_rthdr)
3369                 /* dest1 is valid with rthdr only */
3370                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3371         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3372         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3373         return len;
3374 #undef elen
3375 }